summaryrefslogtreecommitdiffstats
path: root/contrib/macfuse
diff options
context:
space:
mode:
authorAnoop C S <anoopcs@redhat.com>2017-07-27 14:15:22 +0530
committerJeff Darcy <jeff@pl.atyp.us>2017-08-09 12:58:07 +0000
commit071559cb4745669c87198128ab73d7af2017e9bd (patch)
treed825b74373bafcfb021778ed9b241903babc660d /contrib/macfuse
parent747a08d34e2a1e94d7fce68a3577370288bb1955 (diff)
contrib/fuse-lib: Fix compiler warning within switch case
Following warning is shown when compiled with gcc v7: mount.c: In function ‘gf_fuse_unmount_daemon’: mount.c:98:22: warning: statement will never be executed [-Wswitch-unreachable] char c = 0; ^ CCLD fuse.la We just need to move the relevant declaration statements to that switch case where it is valid and is being used. Change-Id: I8e50cc7cfcfc3bc88218cd69abbf516c08ee1648 Updates: #259 Signed-off-by: Anoop C S <anoopcs@redhat.com> Reviewed-on: https://review.gluster.org/17878 Smoke: Gluster Build System <jenkins@build.gluster.org> CentOS-regression: Gluster Build System <jenkins@build.gluster.org> Reviewed-by: Jeff Darcy <jeff@pl.atyp.us>
Diffstat (limited to 'contrib/macfuse')
0 files changed, 0 insertions, 0 deletions
> -rw-r--r--xlators/cluster/afr/src/afr-inode-read.c2932
-rw-r--r--xlators/cluster/afr/src/afr-inode-read.h32
-rw-r--r--xlators/cluster/afr/src/afr-inode-write.c3686
-rw-r--r--xlators/cluster/afr/src/afr-inode-write.h78
-rw-r--r--xlators/cluster/afr/src/afr-lk-common.c1615
-rw-r--r--xlators/cluster/afr/src/afr-mem-types.h56
-rw-r--r--xlators/cluster/afr/src/afr-messages.h188
-rw-r--r--xlators/cluster/afr/src/afr-open.c566
-rw-r--r--xlators/cluster/afr/src/afr-read-txn.c582
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-common.c4402
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-data.c1485
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-entry.c2076
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-metadata.c874
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-name.c1065
-rw-r--r--xlators/cluster/afr/src/afr-self-heal.h475
-rw-r--r--xlators/cluster/afr/src/afr-self-heald.c2508
-rw-r--r--xlators/cluster/afr/src/afr-self-heald.h86
-rw-r--r--xlators/cluster/afr/src/afr-transaction.c4348
-rw-r--r--xlators/cluster/afr/src/afr-transaction.h55
-rw-r--r--xlators/cluster/afr/src/afr.c2246
-rw-r--r--xlators/cluster/afr/src/afr.h1919
-rw-r--r--xlators/cluster/dht/src/Makefile.am26
-rw-r--r--xlators/cluster/dht/src/dht-common.c18464
-rw-r--r--xlators/cluster/dht/src/dht-common.h2200
-rw-r--r--xlators/cluster/dht/src/dht-diskusage.c857
-rw-r--r--xlators/cluster/dht/src/dht-hashfn.c157
-rw-r--r--xlators/cluster/dht/src/dht-helper.c3441
-rw-r--r--xlators/cluster/dht/src/dht-inode-read.c2550
-rw-r--r--xlators/cluster/dht/src/dht-inode-write.c2332
-rw-r--r--xlators/cluster/dht/src/dht-layout.c1265
-rw-r--r--xlators/cluster/dht/src/dht-linkfile.c520
-rw-r--r--xlators/cluster/dht/src/dht-lock.c2074
-rw-r--r--xlators/cluster/dht/src/dht-lock.h56
-rw-r--r--xlators/cluster/dht/src/dht-mem-types.h49
-rw-r--r--xlators/cluster/dht/src/dht-messages.h493
-rw-r--r--xlators/cluster/dht/src/dht-rebalance.c8637
-rw-r--r--xlators/cluster/dht/src/dht-rename.c3148
-rw-r--r--xlators/cluster/dht/src/dht-selfheal.c4669
-rw-r--r--xlators/cluster/dht/src/dht-shared.c1932
-rw-r--r--xlators/cluster/dht/src/dht.c158
-rw-r--r--xlators/cluster/dht/src/dht.sym8
-rw-r--r--xlators/cluster/dht/src/nufa.c1103
-rw-r--r--xlators/cluster/dht/src/nufa.sym8
-rw-r--r--xlators/cluster/dht/src/switch.c1482
-rw-r--r--xlators/cluster/dht/src/switch.sym8
-rw-r--r--xlators/cluster/dht/src/tier-common.c1249
-rw-r--r--xlators/cluster/dht/src/tier-common.h68
-rw-r--r--xlators/cluster/dht/src/tier.c3326
-rw-r--r--xlators/cluster/dht/src/tier.h112
-rw-r--r--xlators/cluster/dht/src/tier.sym9
-rw-r--r--xlators/cluster/dht/src/unittest/dht_layout_mock.c39
-rw-r--r--xlators/cluster/dht/src/unittest/dht_layout_unittest.c26
-rw-r--r--xlators/cluster/ec/src/ec-code-avx.c43
-rw-r--r--xlators/cluster/ec/src/ec-code-c.c902
-rw-r--r--xlators/cluster/ec/src/ec-code-c.h13
-rw-r--r--xlators/cluster/ec/src/ec-code-intel.c64
-rw-r--r--xlators/cluster/ec/src/ec-code-intel.h167
-rw-r--r--xlators/cluster/ec/src/ec-code-sse.c43
-rw-r--r--xlators/cluster/ec/src/ec-code-x64.c48
-rw-r--r--xlators/cluster/ec/src/ec-code.c394
-rw-r--r--xlators/cluster/ec/src/ec-code.h4
-rw-r--r--xlators/cluster/ec/src/ec-combine.c583
-rw-r--r--xlators/cluster/ec/src/ec-combine.h38
-rw-r--r--xlators/cluster/ec/src/ec-common.c1962
-rw-r--r--xlators/cluster/ec/src/ec-common.h289
-rw-r--r--xlators/cluster/ec/src/ec-data.c195
-rw-r--r--xlators/cluster/ec/src/ec-data.h29
-rw-r--r--xlators/cluster/ec/src/ec-dir-read.c369
-rw-r--r--xlators/cluster/ec/src/ec-dir-write.c679
-rw-r--r--xlators/cluster/ec/src/ec-fops.h430
-rw-r--r--xlators/cluster/ec/src/ec-galois.c3
-rw-r--r--xlators/cluster/ec/src/ec-galois.h18
-rw-r--r--xlators/cluster/ec/src/ec-generic.c878
-rw-r--r--xlators/cluster/ec/src/ec-gf8.c10679
-rw-r--r--xlators/cluster/ec/src/ec-heal.c5133
-rw-r--r--xlators/cluster/ec/src/ec-heald.c977
-rw-r--r--xlators/cluster/ec/src/ec-heald.h13
-rw-r--r--xlators/cluster/ec/src/ec-helpers.c346
-rw-r--r--xlators/cluster/ec/src/ec-helpers.h234
-rw-r--r--xlators/cluster/ec/src/ec-inode-read.c1136
-rw-r--r--xlators/cluster/ec/src/ec-inode-write.c1845
-rw-r--r--xlators/cluster/ec/src/ec-locks.c676
-rw-r--r--xlators/cluster/ec/src/ec-mem-types.h6
-rw-r--r--xlators/cluster/ec/src/ec-messages.h114
-rw-r--r--xlators/cluster/ec/src/ec-method.c48
-rw-r--r--xlators/cluster/ec/src/ec-method.h11
-rw-r--r--xlators/cluster/ec/src/ec-types.h798
-rw-r--r--xlators/cluster/ec/src/ec.c2121
-rw-r--r--xlators/cluster/ec/src/ec.h17
-rw-r--r--xlators/cluster/stripe/src/Makefile.am22
-rw-r--r--xlators/cluster/stripe/src/stripe-helpers.c677
-rw-r--r--xlators/cluster/stripe/src/stripe-mem-types.h31
-rw-r--r--xlators/cluster/stripe/src/stripe.c5782
-rw-r--r--xlators/cluster/stripe/src/stripe.h281
-rw-r--r--xlators/debug/delay-gen/src/delay-gen-mem-types.h7
-rw-r--r--xlators/debug/delay-gen/src/delay-gen-messages.h2
-rw-r--r--xlators/debug/delay-gen/src/delay-gen.c871
-rw-r--r--xlators/debug/delay-gen/src/delay-gen.h15
-rw-r--r--xlators/debug/error-gen/src/error-gen-mem-types.h6
-rw-r--r--xlators/debug/error-gen/src/error-gen.c2550
-rw-r--r--xlators/debug/error-gen/src/error-gen.h33
-rw-r--r--xlators/debug/io-stats/src/io-stats-mem-types.h17
-rw-r--r--xlators/debug/io-stats/src/io-stats.c6971
-rw-r--r--xlators/debug/sink/src/sink.c73
-rw-r--r--xlators/debug/trace/src/trace-mem-types.h7
-rw-r--r--xlators/debug/trace/src/trace.c5453
-rw-r--r--xlators/debug/trace/src/trace.h65
-rw-r--r--xlators/encryption/Makefile.am3
-rw-r--r--xlators/encryption/crypt/src/Makefile.am26
-rw-r--r--xlators/encryption/crypt/src/atom.c957
-rw-r--r--xlators/encryption/crypt/src/crypt-common.h141
-rw-r--r--xlators/encryption/crypt/src/crypt-mem-types.h45
-rw-r--r--xlators/encryption/crypt/src/crypt.c4525
-rw-r--r--xlators/encryption/crypt/src/crypt.h890
-rw-r--r--xlators/encryption/crypt/src/data.c764
-rw-r--r--xlators/encryption/crypt/src/keys.c310
-rw-r--r--xlators/encryption/crypt/src/metadata.c614
-rw-r--r--xlators/encryption/crypt/src/metadata.h74
-rw-r--r--xlators/encryption/rot-13/src/rot-13.c196
-rw-r--r--xlators/features/Makefile.am18
-rw-r--r--xlators/features/arbiter/src/arbiter-mem-types.h6
-rw-r--r--xlators/features/arbiter/src/arbiter.c533
-rw-r--r--xlators/features/arbiter/src/arbiter.h6
-rw-r--r--xlators/features/barrier/src/barrier-mem-types.h6
-rw-r--r--xlators/features/barrier/src/barrier.c1080
-rw-r--r--xlators/features/barrier/src/barrier.h129
-rw-r--r--xlators/features/bit-rot/src/bitd/bit-rot-bitd-messages.h133
-rw-r--r--xlators/features/bit-rot/src/bitd/bit-rot-scrub-status.c83
-rw-r--r--xlators/features/bit-rot/src/bitd/bit-rot-scrub-status.h28
-rw-r--r--xlators/features/bit-rot/src/bitd/bit-rot-scrub.c3178
-rw-r--r--xlators/features/bit-rot/src/bitd/bit-rot-scrub.h33
-rw-r--r--xlators/features/bit-rot/src/bitd/bit-rot-ssm.c129
-rw-r--r--xlators/features/bit-rot/src/bitd/bit-rot-ssm.h27
-rw-r--r--xlators/features/bit-rot/src/bitd/bit-rot.c3309
-rw-r--r--xlators/features/bit-rot/src/bitd/bit-rot.h337
-rw-r--r--xlators/features/bit-rot/src/stub/bit-rot-common.h181
-rw-r--r--xlators/features/bit-rot/src/stub/bit-rot-object-version.h12
-rw-r--r--xlators/features/bit-rot/src/stub/bit-rot-stub-helpers.c1181
-rw-r--r--xlators/features/bit-rot/src/stub/bit-rot-stub-mem-types.h35
-rw-r--r--xlators/features/bit-rot/src/stub/bit-rot-stub-messages.h127
-rw-r--r--xlators/features/bit-rot/src/stub/bit-rot-stub.c5175
-rw-r--r--xlators/features/bit-rot/src/stub/bit-rot-stub.h593
-rw-r--r--xlators/features/changelog/lib/examples/c/get-changes-multi.c80
-rw-r--r--xlators/features/changelog/lib/examples/c/get-changes.c102
-rw-r--r--xlators/features/changelog/lib/examples/c/get-history.c148
-rwxr-xr-x[-rw-r--r--]xlators/features/changelog/lib/examples/python/changes.py2
-rw-r--r--xlators/features/changelog/lib/examples/python/libgfchangelog.py4
-rw-r--r--xlators/features/changelog/lib/src/Makefile.am8
-rw-r--r--xlators/features/changelog/lib/src/changelog-lib-messages.h83
-rw-r--r--xlators/features/changelog/lib/src/gf-changelog-api.c313
-rw-r--r--xlators/features/changelog/lib/src/gf-changelog-helpers.c255
-rw-r--r--xlators/features/changelog/lib/src/gf-changelog-helpers.h244
-rw-r--r--xlators/features/changelog/lib/src/gf-changelog-journal-handler.c1634
-rw-r--r--xlators/features/changelog/lib/src/gf-changelog-journal.h94
-rw-r--r--xlators/features/changelog/lib/src/gf-changelog-reborp.c589
-rw-r--r--xlators/features/changelog/lib/src/gf-changelog-rpc.c91
-rw-r--r--xlators/features/changelog/lib/src/gf-changelog-rpc.h10
-rw-r--r--xlators/features/changelog/lib/src/gf-changelog.c889
-rw-r--r--xlators/features/changelog/lib/src/gf-history-changelog.c1531
-rw-r--r--xlators/features/changelog/src/changelog-barrier.c141
-rw-r--r--xlators/features/changelog/src/changelog-encoders.c298
-rw-r--r--xlators/features/changelog/src/changelog-encoders.h42
-rw-r--r--xlators/features/changelog/src/changelog-ev-handle.c582
-rw-r--r--xlators/features/changelog/src/changelog-ev-handle.h98
-rw-r--r--xlators/features/changelog/src/changelog-helpers.c2935
-rw-r--r--xlators/features/changelog/src/changelog-helpers.h894
-rw-r--r--xlators/features/changelog/src/changelog-mem-types.h32
-rw-r--r--xlators/features/changelog/src/changelog-messages.h205
-rw-r--r--xlators/features/changelog/src/changelog-misc.h161
-rw-r--r--xlators/features/changelog/src/changelog-rpc-common.c516
-rw-r--r--xlators/features/changelog/src/changelog-rpc-common.h59
-rw-r--r--xlators/features/changelog/src/changelog-rpc.c601
-rw-r--r--xlators/features/changelog/src/changelog-rpc.h10
-rw-r--r--xlators/features/changelog/src/changelog-rt.c63
-rw-r--r--xlators/features/changelog/src/changelog-rt.h14
-rw-r--r--xlators/features/changelog/src/changelog.c4678
-rw-r--r--xlators/features/changetimerecorder/src/Makefile.am26
-rw-r--r--xlators/features/changetimerecorder/src/changetimerecorder.c2523
-rw-r--r--xlators/features/changetimerecorder/src/changetimerecorder.h21
-rw-r--r--xlators/features/changetimerecorder/src/ctr-helper.c307
-rw-r--r--xlators/features/changetimerecorder/src/ctr-helper.h927
-rw-r--r--xlators/features/changetimerecorder/src/ctr-messages.h86
-rw-r--r--xlators/features/changetimerecorder/src/ctr-xlator-ctx.c409
-rw-r--r--xlators/features/changetimerecorder/src/ctr-xlator-ctx.h90
-rw-r--r--xlators/features/changetimerecorder/src/ctr_mem_types.h24
-rw-r--r--xlators/features/cloudsync/src/Makefile.am17
-rw-r--r--xlators/features/cloudsync/src/cloudsync-autogen-fops-tmpl.c8
-rw-r--r--xlators/features/cloudsync/src/cloudsync-autogen-fops-tmpl.h2
-rw-r--r--xlators/features/cloudsync/src/cloudsync-common.c58
-rw-r--r--xlators/features/cloudsync/src/cloudsync-common.h163
-rwxr-xr-x[-rw-r--r--]xlators/features/cloudsync/src/cloudsync-fops-c.py42
-rwxr-xr-x[-rw-r--r--]xlators/features/cloudsync/src/cloudsync-fops-h.py4
-rw-r--r--xlators/features/cloudsync/src/cloudsync-mem-types.h13
-rw-r--r--xlators/features/cloudsync/src/cloudsync-messages.h3
-rw-r--r--xlators/features/cloudsync/src/cloudsync-plugins/Makefile.am (renamed from xlators/features/changetimerecorder/Makefile.am)0
-rw-r--r--xlators/features/cloudsync/src/cloudsync-plugins/src/Makefile.am11
-rw-r--r--xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/Makefile.am (renamed from xlators/storage/bd/Makefile.am)0
-rw-r--r--xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/Makefile.am12
-rw-r--r--xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3-mem-types.h19
-rw-r--r--xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3.c584
-rw-r--r--xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3.h50
-rw-r--r--xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3.sym1
-rw-r--r--xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/Makefile.am (renamed from xlators/encryption/rot-13/Makefile.am)2
-rw-r--r--xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/Makefile.am12
-rw-r--r--xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/archivestore.h203
-rw-r--r--xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/cvlt-messages.h30
-rw-r--r--xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcloudsynccvlt.sym1
-rw-r--r--xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcvlt-mem-types.h19
-rw-r--r--xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcvlt.c842
-rw-r--r--xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcvlt.h84
-rw-r--r--xlators/features/cloudsync/src/cloudsync.c2827
-rw-r--r--xlators/features/cloudsync/src/cloudsync.h107
-rw-r--r--xlators/features/compress/src/cdc-helper.c762
-rw-r--r--xlators/features/compress/src/cdc-mem-types.h10
-rw-r--r--xlators/features/compress/src/cdc.c578
-rw-r--r--xlators/features/compress/src/cdc.h76
-rw-r--r--xlators/features/gfid-access/src/gfid-access-mem-types.h9
-rw-r--r--xlators/features/gfid-access/src/gfid-access.c2154
-rw-r--r--xlators/features/gfid-access/src/gfid-access.h132
-rw-r--r--xlators/features/glupy/Makefile.am3
-rw-r--r--xlators/features/glupy/doc/README.md44
-rw-r--r--xlators/features/glupy/doc/TESTING9
-rw-r--r--xlators/features/glupy/doc/test.vol10
-rw-r--r--xlators/features/glupy/examples/Makefile.am5
-rw-r--r--xlators/features/glupy/examples/debug-trace.py777
-rw-r--r--xlators/features/glupy/examples/helloworld.py21
-rw-r--r--xlators/features/glupy/examples/negative.py93
-rw-r--r--xlators/features/glupy/src/Makefile.am37
-rw-r--r--xlators/features/glupy/src/__init__.py.in4
-rw-r--r--xlators/features/glupy/src/glupy.c2515
-rw-r--r--xlators/features/glupy/src/glupy.h56
-rw-r--r--xlators/features/glupy/src/glupy.sym101
-rw-r--r--xlators/features/glupy/src/glupy/Makefile.am5
-rw-r--r--xlators/features/glupy/src/glupy/__init__.py852
-rw-r--r--xlators/features/glupy/src/setup.py.in24
-rw-r--r--xlators/features/index/src/index-mem-types.h16
-rw-r--r--xlators/features/index/src/index-messages.h20
-rw-r--r--xlators/features/index/src/index.c4119
-rw-r--r--xlators/features/index/src/index.h106
-rw-r--r--xlators/features/leases/src/leases-internal.c2093
-rw-r--r--xlators/features/leases/src/leases-mem-types.h21
-rw-r--r--xlators/features/leases/src/leases-messages.h21
-rw-r--r--xlators/features/leases/src/leases.c1557
-rw-r--r--xlators/features/leases/src/leases.h264
-rw-r--r--xlators/features/locks/src/clear.c717
-rw-r--r--xlators/features/locks/src/clear.h66
-rw-r--r--xlators/features/locks/src/common.c1953
-rw-r--r--xlators/features/locks/src/common.h253
-rw-r--r--xlators/features/locks/src/entrylk.c1647
-rw-r--r--xlators/features/locks/src/inodelk.c1714
-rw-r--r--xlators/features/locks/src/locks-mem-types.h24
-rw-r--r--xlators/features/locks/src/locks.h346
-rw-r--r--xlators/features/locks/src/pl-messages.h9
-rw-r--r--xlators/features/locks/src/posix.c7310
-rw-r--r--xlators/features/locks/src/reservelk.c567
-rw-r--r--xlators/features/locks/tests/unit-test.c100
-rw-r--r--xlators/features/marker/src/marker-common.c74
-rw-r--r--xlators/features/marker/src/marker-common.h6
-rw-r--r--xlators/features/marker/src/marker-mem-types.h23
-rw-r--r--xlators/features/marker/src/marker-quota-helper.c623
-rw-r--r--xlators/features/marker/src/marker-quota-helper.h78
-rw-r--r--xlators/features/marker/src/marker-quota.c3524
-rw-r--r--xlators/features/marker/src/marker-quota.h202
-rw-r--r--xlators/features/marker/src/marker.c4813
-rw-r--r--xlators/features/marker/src/marker.h227
-rw-r--r--xlators/features/metadisp/Makefile.am (renamed from xlators/encryption/crypt/Makefile.am)2
-rw-r--r--xlators/features/metadisp/src/Makefile.am38
-rw-r--r--xlators/features/metadisp/src/backend.c45
-rw-r--r--xlators/features/metadisp/src/fops-tmpl.c10
-rw-r--r--xlators/features/metadisp/src/gen-fops.py160
-rw-r--r--xlators/features/metadisp/src/metadisp-create.c101
-rw-r--r--xlators/features/metadisp/src/metadisp-fops.h51
-rw-r--r--xlators/features/metadisp/src/metadisp-fsync.c54
-rw-r--r--xlators/features/metadisp/src/metadisp-lookup.c90
-rw-r--r--xlators/features/metadisp/src/metadisp-open.c70
-rw-r--r--xlators/features/metadisp/src/metadisp-readdir.c65
-rw-r--r--xlators/features/metadisp/src/metadisp-setattr.c90
-rw-r--r--xlators/features/metadisp/src/metadisp-stat.c124
-rw-r--r--xlators/features/metadisp/src/metadisp-unlink.c160
-rw-r--r--xlators/features/metadisp/src/metadisp.c46
-rw-r--r--xlators/features/metadisp/src/metadisp.h45
-rw-r--r--xlators/features/namespace/src/namespace.c1803
-rw-r--r--xlators/features/namespace/src/namespace.h10
-rw-r--r--xlators/features/quiesce/src/quiesce-mem-types.h8
-rw-r--r--xlators/features/quiesce/src/quiesce-messages.h7
-rw-r--r--xlators/features/quiesce/src/quiesce.c3553
-rw-r--r--xlators/features/quiesce/src/quiesce.h74
-rw-r--r--xlators/features/quota/src/Makefile.am5
-rw-r--r--xlators/features/quota/src/quota-enforcer-client.c724
-rw-r--r--xlators/features/quota/src/quota-mem-types.h28
-rw-r--r--xlators/features/quota/src/quota-messages.h42
-rw-r--r--xlators/features/quota/src/quota.c8285
-rw-r--r--xlators/features/quota/src/quota.h414
-rw-r--r--xlators/features/quota/src/quotad-aggregator.c755
-rw-r--r--xlators/features/quota/src/quotad-aggregator.h29
-rw-r--r--xlators/features/quota/src/quotad-helpers.c116
-rw-r--r--xlators/features/quota/src/quotad-helpers.h4
-rw-r--r--xlators/features/quota/src/quotad.c321
-rw-r--r--xlators/features/quota/src/quotad.sym7
-rw-r--r--xlators/features/read-only/src/read-only-common.c557
-rw-r--r--xlators/features/read-only/src/read-only-common.h112
-rw-r--r--xlators/features/read-only/src/read-only-mem-types.h6
-rw-r--r--xlators/features/read-only/src/read-only.c181
-rw-r--r--xlators/features/read-only/src/read-only.h33
-rw-r--r--xlators/features/read-only/src/worm-helper.c610
-rw-r--r--xlators/features/read-only/src/worm-helper.h43
-rw-r--r--xlators/features/read-only/src/worm.c1157
-rw-r--r--xlators/features/sdfs/src/sdfs-messages.h16
-rw-r--r--xlators/features/sdfs/src/sdfs.c2124
-rw-r--r--xlators/features/sdfs/src/sdfs.h54
-rw-r--r--xlators/features/selinux/src/selinux-mem-types.h7
-rw-r--r--xlators/features/selinux/src/selinux-messages.h13
-rw-r--r--xlators/features/selinux/src/selinux.c422
-rw-r--r--xlators/features/selinux/src/selinux.h4
-rw-r--r--xlators/features/shard/src/shard-mem-types.h15
-rw-r--r--xlators/features/shard/src/shard-messages.h35
-rw-r--r--xlators/features/shard/src/shard.c10658
-rw-r--r--xlators/features/shard/src/shard.h529
-rw-r--r--xlators/features/snapview-client/src/Makefile.am2
-rw-r--r--xlators/features/snapview-client/src/snapview-client-mem-types.h12
-rw-r--r--xlators/features/snapview-client/src/snapview-client-messages.h71
-rw-r--r--xlators/features/snapview-client/src/snapview-client.c4127
-rw-r--r--xlators/features/snapview-client/src/snapview-client.h151
-rw-r--r--xlators/features/snapview-server/src/Makefile.am2
-rw-r--r--xlators/features/snapview-server/src/snapview-server-helpers.c896
-rw-r--r--xlators/features/snapview-server/src/snapview-server-mem-types.h15
-rw-r--r--xlators/features/snapview-server/src/snapview-server-messages.h54
-rw-r--r--xlators/features/snapview-server/src/snapview-server-mgmt.c856
-rw-r--r--xlators/features/snapview-server/src/snapview-server.c4177
-rw-r--r--xlators/features/snapview-server/src/snapview-server.h310
-rw-r--r--xlators/features/thin-arbiter/src/Makefile.am2
-rw-r--r--xlators/features/thin-arbiter/src/thin-arbiter-mem-types.h8
-rw-r--r--xlators/features/thin-arbiter/src/thin-arbiter-messages.h6
-rw-r--r--xlators/features/thin-arbiter/src/thin-arbiter.c456
-rw-r--r--xlators/features/thin-arbiter/src/thin-arbiter.h66
-rw-r--r--xlators/features/trash/src/trash-mem-types.h13
-rw-r--r--xlators/features/trash/src/trash.c4345
-rw-r--r--xlators/features/trash/src/trash.h111
-rw-r--r--xlators/features/upcall/src/upcall-cache-invalidation.h4
-rw-r--r--xlators/features/upcall/src/upcall-internal.c1038
-rw-r--r--xlators/features/upcall/src/upcall-mem-types.h13
-rw-r--r--xlators/features/upcall/src/upcall-messages.h9
-rw-r--r--xlators/features/upcall/src/upcall.c3198
-rw-r--r--xlators/features/upcall/src/upcall.h194
-rw-r--r--xlators/features/utime/src/Makefile.am4
-rw-r--r--xlators/features/utime/src/utime-autogen-fops-tmpl.c10
-rw-r--r--xlators/features/utime/src/utime-autogen-fops-tmpl.h2
-rwxr-xr-x[-rw-r--r--]xlators/features/utime/src/utime-gen-fops-c.py61
-rwxr-xr-x[-rw-r--r--]xlators/features/utime/src/utime-gen-fops-h.py17
-rw-r--r--xlators/features/utime/src/utime-helpers.c102
-rw-r--r--xlators/features/utime/src/utime-helpers.h11
-rw-r--r--xlators/features/utime/src/utime-mem-types.h (renamed from xlators/performance/decompounder/src/decompounder-mem-types.h)17
-rw-r--r--xlators/features/utime/src/utime-messages.h (renamed from xlators/performance/decompounder/src/decompounder-messages.h)16
-rw-r--r--xlators/features/utime/src/utime.c399
-rw-r--r--xlators/features/utime/src/utime.h10
-rw-r--r--xlators/lib/src/libxlator.c796
-rw-r--r--xlators/lib/src/libxlator.h112
-rw-r--r--xlators/meta/src/active-link.c25
-rw-r--r--xlators/meta/src/cmdline-file.c31
-rw-r--r--xlators/meta/src/frames-file.c163
-rw-r--r--xlators/meta/src/graph-dir.c129
-rw-r--r--xlators/meta/src/graphs-dir.c87
-rw-r--r--xlators/meta/src/history-file.c33
-rw-r--r--xlators/meta/src/logfile-link.c25
-rw-r--r--xlators/meta/src/logging-dir.c42
-rw-r--r--xlators/meta/src/loglevel-file.c40
-rw-r--r--xlators/meta/src/mallinfo-file.c25
-rw-r--r--xlators/meta/src/measure-file.c37
-rw-r--r--xlators/meta/src/meminfo-file.c33
-rw-r--r--xlators/meta/src/meta-defaults.c778
-rw-r--r--xlators/meta/src/meta-helpers.c430
-rw-r--r--xlators/meta/src/meta-hooks.h6
-rw-r--r--xlators/meta/src/meta-mem-types.h17
-rw-r--r--xlators/meta/src/meta.c294
-rw-r--r--xlators/meta/src/meta.h181
-rw-r--r--xlators/meta/src/name-file.c33
-rw-r--r--xlators/meta/src/option-file.c36
-rw-r--r--xlators/meta/src/options-dir.c62
-rw-r--r--xlators/meta/src/private-file.c33
-rw-r--r--xlators/meta/src/process_uuid-file.c27
-rw-r--r--xlators/meta/src/profile-file.c33
-rw-r--r--xlators/meta/src/root-dir.c105
-rw-r--r--xlators/meta/src/subvolume-link.c67
-rw-r--r--xlators/meta/src/subvolumes-dir.c65
-rw-r--r--xlators/meta/src/top-link.c31
-rw-r--r--xlators/meta/src/type-file.c33
-rw-r--r--xlators/meta/src/version-file.c28
-rw-r--r--xlators/meta/src/view-dir.c27
-rw-r--r--xlators/meta/src/volfile-file.c71
-rw-r--r--xlators/meta/src/xlator-dir.c130
-rw-r--r--xlators/mgmt/glusterd/src/Makefile.am34
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-bitd-svc.c287
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-bitd-svc.h16
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-bitrot.c1294
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-brick-ops.c5425
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-conn-helper.c4
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-conn-helper.h2
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-conn-mgmt.c223
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-conn-mgmt.h38
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-errno.h34
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-ganesha.c927
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-geo-rep.c11495
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-geo-rep.h33
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-gfproxyd-svc-helper.c279
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-gfproxyd-svc-helper.h34
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-gfproxyd-svc.c768
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-gfproxyd-svc.h22
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-handler.c11214
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-handshake.c4166
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-hooks.c982
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-hooks.h80
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-locks.c1390
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-locks.h37
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-log-ops.c458
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-mem-types.h102
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-messages.h1041
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-mgmt-handler.c1759
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-mgmt.c5060
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-mgmt.h86
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-mountbroker.c1237
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-mountbroker.h35
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-nfs-svc.c326
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-nfs-svc.h8
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-op-sm.c13950
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-op-sm.h327
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-peer-utils.c1568
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-peer-utils.h67
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-pmap.c980
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-pmap.h59
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-proc-mgmt.c181
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-proc-mgmt.h34
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-quota.c3701
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-quota.h4
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-quotad-svc.c318
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-quotad-svc.h10
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-rcu.h6
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-rebalance.c2186
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-replace-brick.c1234
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-reset-brick.c661
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-rpc-ops.c4152
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-scrub-svc.c286
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-scrub-svc.h22
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-server-quorum.c758
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-server-quorum.h26
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.c153
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-shd-svc-helper.h42
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-shd-svc.c931
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-shd-svc.h25
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-sm.c2602
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-sm.h253
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-snapd-svc-helper.c65
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-snapd-svc-helper.h16
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-snapd-svc.c755
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-snapd-svc.h20
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-snapshot-utils.c7165
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-snapshot-utils.h151
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-snapshot.c17699
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-statedump.c390
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-statedump.h4
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-store.c8431
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-store.h270
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-svc-helper.c1125
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-svc-helper.h53
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-svc-mgmt.c652
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-svc-mgmt.h101
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-syncop.c3368
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-syncop.h102
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-tier.c1381
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-tierd-svc-helper.c270
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-tierd-svc-helper.h37
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-tierd-svc.c488
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-tierd-svc.h42
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-utils.c24219
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-utils.h886
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-volgen.c10875
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-volgen.h349
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-volume-ops.c5385
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-volume-set.c6512
-rw-r--r--xlators/mgmt/glusterd/src/glusterd.c3793
-rw-r--r--xlators/mgmt/glusterd/src/glusterd.h1869
-rw-r--r--xlators/mount/fuse/src/fuse-bridge.c10502
-rw-r--r--xlators/mount/fuse/src/fuse-bridge.h814
-rw-r--r--xlators/mount/fuse/src/fuse-helpers.c1062
-rw-r--r--xlators/mount/fuse/src/fuse-mem-types.h27
-rw-r--r--xlators/mount/fuse/src/fuse-resolve.c1036
-rwxr-xr-xxlators/mount/fuse/utils/mount.glusterfs.in86
-rwxr-xr-xxlators/mount/fuse/utils/mount_glusterfs.in19
-rw-r--r--xlators/nfs/server/src/Makefile.am9
-rw-r--r--xlators/nfs/server/src/acl3.c1584
-rw-r--r--xlators/nfs/server/src/acl3.h18
-rw-r--r--xlators/nfs/server/src/auth-cache.c542
-rw-r--r--xlators/nfs/server/src/auth-cache.h25
-rw-r--r--xlators/nfs/server/src/exports.c1621
-rw-r--r--xlators/nfs/server/src/exports.h70
-rw-r--r--xlators/nfs/server/src/mount3-auth.c682
-rw-r--r--xlators/nfs/server/src/mount3-auth.h28
-rw-r--r--xlators/nfs/server/src/mount3.c6459
-rw-r--r--xlators/nfs/server/src/mount3.h210
-rw-r--r--xlators/nfs/server/src/mount3udp_svc.c344
-rw-r--r--xlators/nfs/server/src/netgroups.c1020
-rw-r--r--xlators/nfs/server/src/netgroups.h27
-rw-r--r--xlators/nfs/server/src/nfs-common.c648
-rw-r--r--xlators/nfs/server/src/nfs-common.h49
-rw-r--r--xlators/nfs/server/src/nfs-fops.c2361
-rw-r--r--xlators/nfs/server/src/nfs-fops.h304
-rw-r--r--xlators/nfs/server/src/nfs-generics.c335
-rw-r--r--xlators/nfs/server/src/nfs-generics.h146
-rw-r--r--xlators/nfs/server/src/nfs-inodes.c767
-rw-r--r--xlators/nfs/server/src/nfs-inodes.h59
-rw-r--r--xlators/nfs/server/src/nfs-mem-types.h71
-rw-r--r--xlators/nfs/server/src/nfs-messages.h283
-rw-r--r--xlators/nfs/server/src/nfs.c3468
-rw-r--r--xlators/nfs/server/src/nfs.h175
-rw-r--r--xlators/nfs/server/src/nfs3-fh.c216
-rw-r--r--xlators/nfs/server/src/nfs3-fh.h105
-rw-r--r--xlators/nfs/server/src/nfs3-helpers.c5065
-rw-r--r--xlators/nfs/server/src/nfs3-helpers.h280
-rw-r--r--xlators/nfs/server/src/nfs3.c9219
-rw-r--r--xlators/nfs/server/src/nfs3.h386
-rw-r--r--xlators/nfs/server/src/nfsserver.sym10
-rw-r--r--xlators/nfs/server/src/nlm4.c4355
-rw-r--r--xlators/nfs/server/src/nlm4.h126
-rw-r--r--xlators/nfs/server/src/nlmcbk_svc.c176
-rw-r--r--xlators/performance/Makefile.am2
-rw-r--r--xlators/performance/decompounder/Makefile.am1
-rw-r--r--xlators/performance/decompounder/src/Makefile.am19
-rw-r--r--xlators/performance/decompounder/src/decompounder.c948
-rw-r--r--xlators/performance/decompounder/src/decompounder.h75
-rw-r--r--xlators/performance/io-cache/src/io-cache-messages.h56
-rw-r--r--xlators/performance/io-cache/src/io-cache.c3094
-rw-r--r--xlators/performance/io-cache/src/io-cache.h400
-rw-r--r--xlators/performance/io-cache/src/ioc-inode.c298
-rw-r--r--xlators/performance/io-cache/src/ioc-mem-types.h24
-rw-r--r--xlators/performance/io-cache/src/page.c1456
-rw-r--r--xlators/performance/io-threads/src/io-threads-messages.h24
-rw-r--r--xlators/performance/io-threads/src/io-threads.c2057
-rw-r--r--xlators/performance/io-threads/src/io-threads.h111
-rw-r--r--xlators/performance/io-threads/src/iot-mem-types.h10
-rw-r--r--xlators/performance/md-cache/src/md-cache-mem-types.h14
-rw-r--r--xlators/performance/md-cache/src/md-cache-messages.h12
-rw-r--r--xlators/performance/md-cache/src/md-cache.c5458
-rw-r--r--xlators/performance/nl-cache/src/Makefile.am2
-rw-r--r--xlators/performance/nl-cache/src/nl-cache-helper.c1732
-rw-r--r--xlators/performance/nl-cache/src/nl-cache-mem-types.h20
-rw-r--r--xlators/performance/nl-cache/src/nl-cache-messages.h10
-rw-r--r--xlators/performance/nl-cache/src/nl-cache.c1272
-rw-r--r--xlators/performance/nl-cache/src/nl-cache.h170
-rw-r--r--xlators/performance/open-behind/src/open-behind-mem-types.h9
-rw-r--r--xlators/performance/open-behind/src/open-behind-messages.h13
-rw-r--r--xlators/performance/open-behind/src/open-behind.c1647
-rw-r--r--xlators/performance/quick-read/src/quick-read-mem-types.h16
-rw-r--r--xlators/performance/quick-read/src/quick-read-messages.h17
-rw-r--r--xlators/performance/quick-read/src/quick-read.c2249
-rw-r--r--xlators/performance/quick-read/src/quick-read.h88
-rw-r--r--xlators/performance/read-ahead/src/page.c861
-rw-r--r--xlators/performance/read-ahead/src/read-ahead-mem-types.h17
-rw-r--r--xlators/performance/read-ahead/src/read-ahead-messages.h15
-rw-r--r--xlators/performance/read-ahead/src/read-ahead.c1855
-rw-r--r--xlators/performance/read-ahead/src/read-ahead.h178
-rw-r--r--xlators/performance/readdir-ahead/src/readdir-ahead-mem-types.h12
-rw-r--r--xlators/performance/readdir-ahead/src/readdir-ahead-messages.h14
-rw-r--r--xlators/performance/readdir-ahead/src/readdir-ahead.c1713
-rw-r--r--xlators/performance/readdir-ahead/src/readdir-ahead.h99
-rw-r--r--xlators/performance/symlink-cache/Makefile.am3
-rw-r--r--xlators/performance/symlink-cache/src/Makefile.am16
-rw-r--r--xlators/performance/symlink-cache/src/symlink-cache-messages.h33
-rw-r--r--xlators/performance/symlink-cache/src/symlink-cache.c402
-rw-r--r--xlators/performance/write-behind/src/write-behind-mem-types.h16
-rw-r--r--xlators/performance/write-behind/src/write-behind-messages.h16
-rw-r--r--xlators/performance/write-behind/src/write-behind.c4749
-rw-r--r--xlators/playground/rot-13/Makefile.am (renamed from xlators/cluster/stripe/Makefile.am)0
-rw-r--r--xlators/playground/rot-13/src/Makefile.am (renamed from xlators/encryption/rot-13/src/Makefile.am)0
-rw-r--r--xlators/playground/rot-13/src/rot-13.c166
-rw-r--r--xlators/playground/rot-13/src/rot-13.h (renamed from xlators/encryption/rot-13/src/rot-13.h)4
-rw-r--r--xlators/playground/template/src/Makefile.am2
-rw-r--r--xlators/playground/template/src/template.c182
-rw-r--r--xlators/playground/template/src/template.h36
-rw-r--r--xlators/protocol/auth/addr/src/Makefile.am2
-rw-r--r--xlators/protocol/auth/addr/src/addr.c528
-rw-r--r--xlators/protocol/auth/login/src/Makefile.am2
-rw-r--r--xlators/protocol/auth/login/src/login.c351
-rw-r--r--xlators/protocol/client/src/client-callback.c456
-rw-r--r--xlators/protocol/client/src/client-common.c4568
-rw-r--r--xlators/protocol/client/src/client-common.h644
-rw-r--r--xlators/protocol/client/src/client-handshake.c2621
-rw-r--r--xlators/protocol/client/src/client-helpers.c3685
-rw-r--r--xlators/protocol/client/src/client-lk.c709
-rw-r--r--xlators/protocol/client/src/client-mem-types.h20
-rw-r--r--xlators/protocol/client/src/client-messages.h219
-rw-r--r--xlators/protocol/client/src/client-rpc-fops.c10479
-rw-r--r--xlators/protocol/client/src/client-rpc-fops_v2.c10368
-rw-r--r--xlators/protocol/client/src/client.c4368
-rw-r--r--xlators/protocol/client/src/client.h710
-rw-r--r--xlators/protocol/server/src/Makefile.am2
-rw-r--r--xlators/protocol/server/src/authenticate.c327
-rw-r--r--xlators/protocol/server/src/authenticate.h32
-rw-r--r--xlators/protocol/server/src/server-common.c1165
-rw-r--r--xlators/protocol/server/src/server-common.h213
-rw-r--r--xlators/protocol/server/src/server-handshake.c1573
-rw-r--r--xlators/protocol/server/src/server-helpers.c7196
-rw-r--r--xlators/protocol/server/src/server-helpers.h117
-rw-r--r--xlators/protocol/server/src/server-mem-types.h27
-rw-r--r--xlators/protocol/server/src/server-messages.h239
-rw-r--r--xlators/protocol/server/src/server-resolve.c981
-rw-r--r--xlators/protocol/server/src/server-rpc-fops.c8848
-rw-r--r--xlators/protocol/server/src/server-rpc-fops_v2.c8474
-rw-r--r--xlators/protocol/server/src/server.c3089
-rw-r--r--xlators/protocol/server/src/server.h350
-rw-r--r--xlators/storage/Makefile.am4
-rw-r--r--xlators/storage/bd/src/Makefile.am21
-rw-r--r--xlators/storage/bd/src/bd-aio.c523
-rw-r--r--xlators/storage/bd/src/bd-aio.h36
-rw-r--r--xlators/storage/bd/src/bd-helper.c1025
-rw-r--r--xlators/storage/bd/src/bd-mem-types.h27
-rw-r--r--xlators/storage/bd/src/bd.c2443
-rw-r--r--xlators/storage/bd/src/bd.h168
-rw-r--r--xlators/storage/posix/src/Makefile.am2
-rw-r--r--xlators/storage/posix/src/posix-aio.c970
-rw-r--r--xlators/storage/posix/src/posix-aio.h22
-rw-r--r--xlators/storage/posix/src/posix-common.c2413
-rw-r--r--xlators/storage/posix/src/posix-entry-ops.c4084
-rw-r--r--xlators/storage/posix/src/posix-gfid-path.c485
-rw-r--r--xlators/storage/posix/src/posix-gfid-path.h21
-rw-r--r--xlators/storage/posix/src/posix-handle.c1685
-rw-r--r--xlators/storage/posix/src/posix-handle.h348
-rw-r--r--xlators/storage/posix/src/posix-helpers.c5724
-rw-r--r--xlators/storage/posix/src/posix-inode-fd-ops.c9115
-rw-r--r--xlators/storage/posix/src/posix-inode-handle.h139
-rw-r--r--xlators/storage/posix/src/posix-mem-types.h22
-rw-r--r--xlators/storage/posix/src/posix-messages.h169
-rw-r--r--xlators/storage/posix/src/posix-metadata-disk.h20
-rw-r--r--xlators/storage/posix/src/posix-metadata.c1289
-rw-r--r--xlators/storage/posix/src/posix-metadata.h61
-rw-r--r--xlators/storage/posix/src/posix.c136
-rw-r--r--xlators/storage/posix/src/posix.h825
-rw-r--r--xlators/system/posix-acl/src/posix-acl-mem-types.h13
-rw-r--r--xlators/system/posix-acl/src/posix-acl-messages.h6
-rw-r--r--xlators/system/posix-acl/src/posix-acl-xattr.c229
-rw-r--r--xlators/system/posix-acl/src/posix-acl-xattr.h17
-rw-r--r--xlators/system/posix-acl/src/posix-acl.c3399
-rw-r--r--xlators/system/posix-acl/src/posix-acl.h33
-rw-r--r--xlators/xlator.sym10
649 files changed, 328356 insertions, 357509 deletions
diff --git a/xlators/Makefile.am b/xlators/Makefile.am
index 29549db724e..ef20cbb64fa 100644
--- a/xlators/Makefile.am
+++ b/xlators/Makefile.am
@@ -2,10 +2,10 @@ if BUILD_GNFS
GNFS_DIR = nfs
endif
-DIST_SUBDIRS = cluster storage protocol performance debug features encryption \
+DIST_SUBDIRS = cluster storage protocol performance debug features \
mount nfs mgmt system playground meta
-SUBDIRS = cluster storage protocol performance debug features encryption \
+SUBDIRS = cluster storage protocol performance debug features \
mount ${GNFS_DIR} mgmt system playground meta
EXTRA_DIST = xlator.sym
diff --git a/xlators/cluster/Makefile.am b/xlators/cluster/Makefile.am
index 903fbb39f12..8e067d5ab58 100644
--- a/xlators/cluster/Makefile.am
+++ b/xlators/cluster/Makefile.am
@@ -1,3 +1,3 @@
-SUBDIRS = stripe afr dht ec
+SUBDIRS = afr dht ec
CLEANFILES =
diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index ad037cd7060..032ab5c8001 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -15,22 +15,20 @@
#include <stdlib.h>
#include <signal.h>
-#include "glusterfs.h"
+#include <glusterfs/glusterfs.h>
#include "afr.h"
-#include "dict.h"
-#include "xlator.h"
-#include "hashfn.h"
-#include "logging.h"
-#include "list.h"
-#include "call-stub.h"
-#include "defaults.h"
-#include "common-utils.h"
-#include "compat-errno.h"
-#include "compat.h"
-#include "byte-order.h"
-#include "statedump.h"
-#include "events.h"
-#include "upcall-utils.h"
+#include <glusterfs/dict.h>
+#include <glusterfs/hashfn.h>
+#include <glusterfs/list.h>
+#include <glusterfs/call-stub.h>
+#include <glusterfs/defaults.h>
+#include <glusterfs/common-utils.h>
+#include <glusterfs/compat-errno.h>
+#include <glusterfs/compat.h>
+#include <glusterfs/byte-order.h>
+#include <glusterfs/statedump.h>
+#include <glusterfs/events.h>
+#include <glusterfs/upcall-utils.h>
#include "afr-inode-read.h"
#include "afr-inode-write.h"
@@ -42,157 +40,796 @@
#include "afr-messages.h"
int32_t
-afr_quorum_errno (afr_private_t *priv)
+afr_quorum_errno(afr_private_t *priv)
{
- return ENOTCONN;
+ return ENOTCONN;
+}
+
+gf_boolean_t
+afr_is_private_directory(afr_private_t *priv, uuid_t pargfid, const char *name,
+ pid_t pid)
+{
+ if (!__is_root_gfid(pargfid)) {
+ return _gf_false;
+ }
+
+ if (strcmp(name, GF_REPLICATE_TRASH_DIR) == 0) {
+ /*For backward compatibility /.landfill is private*/
+ return _gf_true;
+ }
+
+ if (pid == GF_CLIENT_PID_GSYNCD) {
+ /*geo-rep needs to create/sync private directory on slave because
+ * it appears in changelog*/
+ return _gf_false;
+ }
+
+ if (pid == GF_CLIENT_PID_GLFS_HEAL || pid == GF_CLIENT_PID_SELF_HEALD) {
+ if (strcmp(name, priv->anon_inode_name) == 0) {
+ /* anonymous-inode dir is private*/
+ return _gf_true;
+ }
+ } else {
+ if (strncmp(name, AFR_ANON_DIR_PREFIX, strlen(AFR_ANON_DIR_PREFIX)) ==
+ 0) {
+ /* anonymous-inode dir prefix is private for geo-rep to work*/
+ return _gf_true;
+ }
+ }
+
+ return _gf_false;
+}
+
+void
+afr_fill_success_replies(afr_local_t *local, afr_private_t *priv,
+ unsigned char *replies)
+{
+ int i = 0;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->replies[i].valid && local->replies[i].op_ret == 0) {
+ replies[i] = 1;
+ } else {
+ replies[i] = 0;
+ }
+ }
}
int
-afr_fav_child_reset_sink_xattrs (void *opaque);
+afr_fav_child_reset_sink_xattrs(void *opaque);
int
-afr_fav_child_reset_sink_xattrs_cbk (int ret, call_frame_t *frame,
- void *opaque);
+afr_fav_child_reset_sink_xattrs_cbk(int ret, call_frame_t *frame, void *opaque);
static void
-afr_discover_done (call_frame_t *frame, xlator_t *this);
+afr_discover_done(call_frame_t *frame, xlator_t *this);
-gf_boolean_t
-afr_is_consistent_io_possible (afr_local_t *local, afr_private_t *priv,
- int32_t *op_errno)
+int
+afr_dom_lock_acquire_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
{
- if (priv->consistent_io && local->call_count != priv->child_count) {
- gf_msg (THIS->name, GF_LOG_INFO, 0,
- AFR_MSG_SUBVOLS_DOWN, "All subvolumes are not up");
- if (op_errno)
- *op_errno = ENOTCONN;
- return _gf_false;
- }
- return _gf_true;
+ afr_local_t *local = frame->local;
+ afr_private_t *priv = this->private;
+ int i = (long)cookie;
+
+ local->cont.lk.dom_lock_op_ret[i] = op_ret;
+ local->cont.lk.dom_lock_op_errno[i] = op_errno;
+ if (op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, AFR_MSG_LK_HEAL_DOM,
+ "%s: Failed to acquire %s on %s",
+ uuid_utoa(local->fd->inode->gfid), AFR_LK_HEAL_DOM,
+ priv->children[i]->name);
+ } else {
+ local->cont.lk.dom_locked_nodes[i] = 1;
+ }
+
+ syncbarrier_wake(&local->barrier);
+
+ return 0;
}
-call_frame_t *
-afr_copy_frame (call_frame_t *base)
-{
- afr_local_t *local = NULL;
- call_frame_t *frame = NULL;
- int op_errno = 0;
+int
+afr_dom_lock_acquire(call_frame_t *frame)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ struct gf_flock flock = {
+ 0,
+ };
+ int i = 0;
+
+ priv = frame->this->private;
+ local = frame->local;
+ local->cont.lk.dom_locked_nodes = GF_CALLOC(
+ priv->child_count, sizeof(*local->cont.lk.locked_nodes),
+ gf_afr_mt_char);
+ if (!local->cont.lk.dom_locked_nodes) {
+ return -ENOMEM;
+ }
+ local->cont.lk.dom_lock_op_ret = GF_CALLOC(
+ priv->child_count, sizeof(*local->cont.lk.dom_lock_op_ret),
+ gf_afr_mt_int32_t);
+ if (!local->cont.lk.dom_lock_op_ret) {
+ return -ENOMEM; /* CALLOC'd members are freed in afr_local_cleanup. */
+ }
+ local->cont.lk.dom_lock_op_errno = GF_CALLOC(
+ priv->child_count, sizeof(*local->cont.lk.dom_lock_op_errno),
+ gf_afr_mt_int32_t);
+ if (!local->cont.lk.dom_lock_op_errno) {
+ return -ENOMEM; /* CALLOC'd members are freed in afr_local_cleanup. */
+ }
+ flock.l_type = F_WRLCK;
+
+ AFR_ONALL(frame, afr_dom_lock_acquire_cbk, finodelk, AFR_LK_HEAL_DOM,
+ local->fd, F_SETLK, &flock, NULL);
+
+ if (!afr_has_quorum(local->cont.lk.dom_locked_nodes, frame->this, NULL))
+ goto blocking_lock;
+
+ /*If any of the bricks returned EAGAIN, we still need blocking locks.*/
+ if (AFR_COUNT(local->cont.lk.dom_locked_nodes, priv->child_count) !=
+ priv->child_count) {
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->cont.lk.dom_lock_op_ret[i] == -1 &&
+ local->cont.lk.dom_lock_op_errno[i] == EAGAIN)
+ goto blocking_lock;
+ }
+ }
+
+ return 0;
- frame = copy_frame (base);
- if (!frame)
- return NULL;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local) {
- AFR_STACK_DESTROY (frame);
- return NULL;
- }
+blocking_lock:
+ afr_dom_lock_release(frame);
+ AFR_ONALL(frame, afr_dom_lock_acquire_cbk, finodelk, AFR_LK_HEAL_DOM,
+ local->fd, F_SETLKW, &flock, NULL);
+ if (!afr_has_quorum(local->cont.lk.dom_locked_nodes, frame->this, NULL)) {
+ afr_dom_lock_release(frame);
+ return -afr_quorum_errno(priv);
+ }
- return frame;
+ return 0;
}
-/* Check if an entry or inode could be undergoing a transaction. */
-gf_boolean_t
-afr_is_possibly_under_txn (afr_transaction_type type, afr_local_t *local,
- xlator_t *this)
+int
+afr_dom_lock_release_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
{
- int i = 0;
- int tmp = 0;
- afr_private_t *priv = NULL;
- GF_UNUSED char *key = NULL;
+ afr_local_t *local = frame->local;
+ afr_private_t *priv = this->private;
+ int i = (long)cookie;
- priv = this->private;
+ if (op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, AFR_MSG_LK_HEAL_DOM,
+ "%s: Failed to release %s on %s", local->loc.path,
+ AFR_LK_HEAL_DOM, priv->children[i]->name);
+ }
+ local->cont.lk.dom_locked_nodes[i] = 0;
- if (type == AFR_ENTRY_TRANSACTION)
- key = GLUSTERFS_PARENT_ENTRYLK;
- else if (type == AFR_DATA_TRANSACTION)
- /*FIXME: Use GLUSTERFS_INODELK_DOM_COUNT etc. once
- * pl_inodelk_xattr_fill supports separate keys for different
- * domains.*/
- key = GLUSTERFS_INODELK_COUNT;
+ syncbarrier_wake(&local->barrier);
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].xdata)
- continue;
- if (dict_get_int32 (local->replies[i].xdata, key, &tmp) == 0)
- if (tmp)
- return _gf_true;
- }
+ return 0;
+}
- return _gf_false;
+void
+afr_dom_lock_release(call_frame_t *frame)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ unsigned char *locked_on = NULL;
+ struct gf_flock flock = {
+ 0,
+ };
+
+ local = frame->local;
+ priv = frame->this->private;
+ locked_on = local->cont.lk.dom_locked_nodes;
+ if (AFR_COUNT(locked_on, priv->child_count) == 0)
+ return;
+ flock.l_type = F_UNLCK;
+
+ AFR_ONLIST(locked_on, frame, afr_dom_lock_release_cbk, finodelk,
+ AFR_LK_HEAL_DOM, local->fd, F_SETLK, &flock, NULL);
+
+ return;
}
static void
-afr_inode_ctx_destroy (afr_inode_ctx_t *ctx)
+afr_lk_heal_info_cleanup(afr_lk_heal_info_t *info)
{
- int i = 0;
-
- if (!ctx)
- return;
+ if (!info)
+ return;
+ if (info->xdata_req)
+ dict_unref(info->xdata_req);
+ if (info->fd)
+ fd_unref(info->fd);
+ GF_FREE(info->locked_nodes);
+ GF_FREE(info->child_up_event_gen);
+ GF_FREE(info->child_down_event_gen);
+ GF_FREE(info);
+}
- for (i = 0; i < AFR_NUM_CHANGE_LOGS; i++) {
- GF_FREE (ctx->pre_op_done[i]);
+static int
+afr_add_lock_to_saved_locks(call_frame_t *frame, xlator_t *this)
+{
+ afr_private_t *priv = this->private;
+ afr_local_t *local = frame->local;
+ afr_lk_heal_info_t *info = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int ret = -ENOMEM;
+
+ info = GF_CALLOC(sizeof(*info), 1, gf_afr_mt_lk_heal_info_t);
+ if (!info) {
+ goto cleanup;
+ }
+ INIT_LIST_HEAD(&info->pos);
+ info->fd = fd_ref(local->fd);
+ info->cmd = local->cont.lk.cmd;
+ info->pid = frame->root->pid;
+ info->flock = local->cont.lk.user_flock;
+ info->xdata_req = dict_copy_with_ref(local->xdata_req, NULL);
+ if (!info->xdata_req) {
+ goto cleanup;
+ }
+ info->lk_owner = frame->root->lk_owner;
+ info->locked_nodes = GF_MALLOC(
+ sizeof(*info->locked_nodes) * priv->child_count, gf_afr_mt_char);
+ if (!info->locked_nodes) {
+ goto cleanup;
+ }
+ memcpy(info->locked_nodes, local->cont.lk.locked_nodes,
+ sizeof(*info->locked_nodes) * priv->child_count);
+ info->child_up_event_gen = GF_CALLOC(sizeof(*info->child_up_event_gen),
+ priv->child_count, gf_afr_mt_int32_t);
+ if (!info->child_up_event_gen) {
+ goto cleanup;
+ }
+ info->child_down_event_gen = GF_CALLOC(sizeof(*info->child_down_event_gen),
+ priv->child_count,
+ gf_afr_mt_int32_t);
+ if (!info->child_down_event_gen) {
+ goto cleanup;
+ }
+
+ LOCK(&local->fd->lock);
+ {
+ fd_ctx = __afr_fd_ctx_get(local->fd, this);
+ if (fd_ctx)
+ fd_ctx->lk_heal_info = info;
+ }
+ UNLOCK(&local->fd->lock);
+ if (!fd_ctx) {
+ goto cleanup;
+ }
+
+ LOCK(&priv->lock);
+ {
+ list_add_tail(&info->pos, &priv->saved_locks);
+ }
+ UNLOCK(&priv->lock);
+
+ return 0;
+cleanup:
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_LK_HEAL_DOM,
+ "%s: Failed to add lock to healq",
+ uuid_utoa(local->fd->inode->gfid));
+ if (info) {
+ afr_lk_heal_info_cleanup(info);
+ if (fd_ctx) {
+ LOCK(&local->fd->lock);
+ {
+ fd_ctx->lk_heal_info = NULL;
+ }
+ UNLOCK(&local->fd->lock);
}
+ }
+ return ret;
+}
- GF_FREE (ctx);
+static int
+afr_remove_lock_from_saved_locks(afr_local_t *local, xlator_t *this)
+{
+ afr_private_t *priv = this->private;
+ struct gf_flock flock = local->cont.lk.user_flock;
+ afr_lk_heal_info_t *info = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int ret = -EINVAL;
+
+ fd_ctx = afr_fd_ctx_get(local->fd, this);
+ if (!fd_ctx || !fd_ctx->lk_heal_info) {
+ goto out;
+ }
+
+ info = fd_ctx->lk_heal_info;
+ if ((info->flock.l_start != flock.l_start) ||
+ (info->flock.l_whence != flock.l_whence) ||
+ (info->flock.l_len != flock.l_len)) {
+ /*TODO: Compare lkowners too.*/
+ goto out;
+ }
+
+ LOCK(&priv->lock);
+ {
+ list_del(&fd_ctx->lk_heal_info->pos);
+ }
+ UNLOCK(&priv->lock);
+
+ afr_lk_heal_info_cleanup(info);
+ fd_ctx->lk_heal_info = NULL;
+ ret = 0;
+out:
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_LK_HEAL_DOM,
+ "%s: Failed to remove lock from healq",
+ uuid_utoa(local->fd->inode->gfid));
+ return ret;
}
int
-__afr_inode_ctx_get (xlator_t *this, inode_t *inode, afr_inode_ctx_t **ctx)
+afr_lock_heal_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+ dict_t *xdata)
{
- uint64_t ctx_int = 0;
- int ret = -1;
- int i = -1;
- int num_locks = -1;
- afr_inode_ctx_t *ictx = NULL;
- afr_lock_t *lock = NULL;
- afr_private_t *priv = this->private;
+ afr_local_t *local = frame->local;
+ int i = (long)cookie;
- ret = __inode_ctx_get (inode, this, &ctx_int);
- if (ret == 0) {
- *ctx = (afr_inode_ctx_t *)ctx_int;
- return 0;
- }
+ local->replies[i].valid = 1;
+ local->replies[i].op_ret = op_ret;
+ local->replies[i].op_errno = op_errno;
+ if (op_ret != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, AFR_MSG_LK_HEAL_DOM,
+ "Failed to heal lock on child %d for %s", i,
+ uuid_utoa(local->fd->inode->gfid));
+ }
+ syncbarrier_wake(&local->barrier);
+ return 0;
+}
- ictx = GF_CALLOC (1, sizeof (afr_inode_ctx_t), gf_afr_mt_inode_ctx_t);
- if (!ictx)
- goto out;
+int
+afr_getlk_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
+{
+ afr_local_t *local = frame->local;
+ int i = (long)cookie;
- for (i = 0; i < AFR_NUM_CHANGE_LOGS; i++) {
- ictx->pre_op_done[i] = GF_CALLOC (sizeof *ictx->pre_op_done[i],
- priv->child_count,
- gf_afr_mt_int32_t);
- if (!ictx->pre_op_done[i]) {
- ret = -ENOMEM;
- goto out;
- }
+ local->replies[i].valid = 1;
+ local->replies[i].op_ret = op_ret;
+ local->replies[i].op_errno = op_errno;
+ if (op_ret != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, AFR_MSG_LK_HEAL_DOM,
+ "Failed getlk for %s", uuid_utoa(local->fd->inode->gfid));
+ } else {
+ local->cont.lk.getlk_rsp[i] = *lock;
+ }
+
+ syncbarrier_wake(&local->barrier);
+ return 0;
+}
+
+static gf_boolean_t
+afr_does_lk_owner_match(call_frame_t *frame, afr_private_t *priv,
+ afr_lk_heal_info_t *info)
+{
+ int i = 0;
+ afr_local_t *local = frame->local;
+ struct gf_flock flock = {
+ 0,
+ };
+ gf_boolean_t ret = _gf_true;
+ char *wind_on = alloca0(priv->child_count);
+ unsigned char *success_replies = alloca0(priv->child_count);
+ local->cont.lk.getlk_rsp = GF_CALLOC(sizeof(*local->cont.lk.getlk_rsp),
+ priv->child_count, gf_afr_mt_gf_lock);
+
+ flock = info->flock;
+ for (i = 0; i < priv->child_count; i++) {
+ if (info->locked_nodes[i])
+ wind_on[i] = 1;
+ }
+
+ AFR_ONLIST(wind_on, frame, afr_getlk_cbk, lk, info->fd, F_GETLK, &flock,
+ info->xdata_req);
+
+ afr_fill_success_replies(local, priv, success_replies);
+ if (AFR_COUNT(success_replies, priv->child_count) == 0) {
+ ret = _gf_false;
+ goto out;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid || local->replies[i].op_ret != 0)
+ continue;
+ if (local->cont.lk.getlk_rsp[i].l_type == F_UNLCK)
+ continue;
+ /*TODO: Do we really need to compare lkowner if F_UNLCK is true?*/
+ if (!is_same_lkowner(&local->cont.lk.getlk_rsp[i].l_owner,
+ &info->lk_owner)) {
+ ret = _gf_false;
+ break;
+ }
+ }
+out:
+ afr_local_replies_wipe(local, priv);
+ GF_FREE(local->cont.lk.getlk_rsp);
+ local->cont.lk.getlk_rsp = NULL;
+ return ret;
+}
+
+static void
+afr_mark_fd_bad(fd_t *fd, xlator_t *this)
+{
+ afr_fd_ctx_t *fd_ctx = NULL;
+
+ if (!fd)
+ return;
+ LOCK(&fd->lock);
+ {
+ fd_ctx = __afr_fd_ctx_get(fd, this);
+ if (fd_ctx) {
+ fd_ctx->is_fd_bad = _gf_true;
+ fd_ctx->lk_heal_info = NULL;
}
+ }
+ UNLOCK(&fd->lock);
+}
+
+static void
+afr_add_lock_to_lkhealq(afr_private_t *priv, afr_lk_heal_info_t *info)
+{
+ LOCK(&priv->lock);
+ {
+ list_del(&info->pos);
+ list_add_tail(&info->pos, &priv->lk_healq);
+ }
+ UNLOCK(&priv->lock);
+}
+
+static void
+afr_lock_heal_do(call_frame_t *frame, afr_private_t *priv,
+ afr_lk_heal_info_t *info)
+{
+ int i = 0;
+ int op_errno = 0;
+ int32_t *current_event_gen = NULL;
+ afr_local_t *local = frame->local;
+ xlator_t *this = frame->this;
+ char *wind_on = alloca0(priv->child_count);
+ gf_boolean_t retry = _gf_true;
+
+ frame->root->pid = info->pid;
+ lk_owner_copy(&frame->root->lk_owner, &info->lk_owner);
+
+ op_errno = -afr_dom_lock_acquire(frame);
+ if ((op_errno != 0)) {
+ goto release;
+ }
+
+ if (!afr_does_lk_owner_match(frame, priv, info)) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_LK_HEAL_DOM,
+ "Ignoring lock heal for %s since lk-onwers mismatch. "
+ "Lock possibly pre-empted by another client.",
+ uuid_utoa(info->fd->inode->gfid));
+ goto release;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (info->locked_nodes[i])
+ continue;
+ wind_on[i] = 1;
+ }
+
+ current_event_gen = alloca(priv->child_count);
+ memcpy(current_event_gen, info->child_up_event_gen,
+ priv->child_count * sizeof *current_event_gen);
+ AFR_ONLIST(wind_on, frame, afr_lock_heal_cbk, lk, info->fd, info->cmd,
+ &info->flock, info->xdata_req);
+
+ LOCK(&priv->lock);
+ {
+ for (i = 0; i < priv->child_count; i++) {
+ if (!wind_on[i])
+ continue;
+ if ((!local->replies[i].valid) || (local->replies[i].op_ret != 0)) {
+ continue;
+ }
- num_locks = sizeof(ictx->lock)/sizeof(afr_lock_t);
- for (i = 0; i < num_locks; i++) {
- lock = &ictx->lock[i];
- INIT_LIST_HEAD (&lock->post_op);
- INIT_LIST_HEAD (&lock->frozen);
- INIT_LIST_HEAD (&lock->waiting);
- INIT_LIST_HEAD (&lock->owners);
+ if ((current_event_gen[i] == info->child_up_event_gen[i]) &&
+ (current_event_gen[i] > info->child_down_event_gen[i])) {
+ info->locked_nodes[i] = 1;
+ retry = _gf_false;
+ list_del_init(&info->pos);
+ list_add_tail(&info->pos, &priv->saved_locks);
+ } else {
+ /*We received subsequent child up/down events while heal was in
+ * progress; don't mark child as healed. Attempt again on the
+ * new child up*/
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_LK_HEAL_DOM,
+ "Event gen mismatch: skipped healing lock on child %d "
+ "for %s.",
+ i, uuid_utoa(info->fd->inode->gfid));
+ }
}
+ }
+ UNLOCK(&priv->lock);
- ctx_int = (uint64_t)ictx;
- ret = __inode_ctx_set (inode, this, &ctx_int);
- if (ret) {
- goto out;
+release:
+ afr_dom_lock_release(frame);
+ if (retry)
+ afr_add_lock_to_lkhealq(priv, info);
+ return;
+}
+
+static int
+afr_lock_heal_done(int ret, call_frame_t *frame, void *opaque)
+{
+ STACK_DESTROY(frame->root);
+ return 0;
+}
+
+static int
+afr_lock_heal(void *opaque)
+{
+ call_frame_t *frame = (call_frame_t *)opaque;
+ call_frame_t *iter_frame = NULL;
+ xlator_t *this = frame->this;
+ afr_private_t *priv = this->private;
+ afr_lk_heal_info_t *info = NULL;
+ afr_lk_heal_info_t *tmp = NULL;
+ struct list_head healq = {
+ 0,
+ };
+ int ret = 0;
+
+ iter_frame = afr_copy_frame(frame);
+ if (!iter_frame) {
+ return ENOMEM;
+ }
+
+ INIT_LIST_HEAD(&healq);
+ LOCK(&priv->lock);
+ {
+ list_splice_init(&priv->lk_healq, &healq);
+ }
+ UNLOCK(&priv->lock);
+
+ list_for_each_entry_safe(info, tmp, &healq, pos)
+ {
+ GF_ASSERT((AFR_COUNT(info->locked_nodes, priv->child_count) <
+ priv->child_count));
+ ((afr_local_t *)(iter_frame->local))->fd = fd_ref(info->fd);
+ afr_lock_heal_do(iter_frame, priv, info);
+ AFR_STACK_RESET(iter_frame);
+ if (iter_frame->local == NULL) {
+ ret = ENOTCONN;
+ gf_msg(frame->this->name, GF_LOG_ERROR, ENOTCONN,
+ AFR_MSG_LK_HEAL_DOM,
+ "Aborting processing of lk_healq."
+ "Healing will be reattempted on next child up for locks "
+ "that are still in quorum.");
+ LOCK(&priv->lock);
+ {
+ list_add_tail(&healq, &priv->lk_healq);
+ }
+ UNLOCK(&priv->lock);
+ break;
}
+ }
- ictx->spb_choice = -1;
- ictx->read_subvol = 0;
- ictx->write_subvol = 0;
- ictx->lock_count = 0;
- ret = 0;
- *ctx = ictx;
+ AFR_STACK_DESTROY(iter_frame);
+ return ret;
+}
+
+static int
+__afr_lock_heal_synctask(xlator_t *this, afr_private_t *priv, int child)
+{
+ int ret = 0;
+ call_frame_t *frame = NULL;
+ afr_lk_heal_info_t *info = NULL;
+ afr_lk_heal_info_t *tmp = NULL;
+
+ if (priv->shd.iamshd)
+ return 0;
+
+ list_for_each_entry_safe(info, tmp, &priv->saved_locks, pos)
+ {
+ info->child_up_event_gen[child] = priv->event_generation;
+ list_del_init(&info->pos);
+ list_add_tail(&info->pos, &priv->lk_healq);
+ }
+
+ frame = create_frame(this, this->ctx->pool);
+ if (!frame)
+ return -1;
+
+ ret = synctask_new(this->ctx->env, afr_lock_heal, afr_lock_heal_done, frame,
+ frame);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, AFR_MSG_LK_HEAL_DOM,
+ "Failed to launch lock heal synctask");
+
+ return ret;
+}
+
+static int
+__afr_mark_pending_lk_heal(xlator_t *this, afr_private_t *priv, int child)
+{
+ afr_lk_heal_info_t *info = NULL;
+ afr_lk_heal_info_t *tmp = NULL;
+
+ if (priv->shd.iamshd)
+ return 0;
+ list_for_each_entry_safe(info, tmp, &priv->saved_locks, pos)
+ {
+ info->child_down_event_gen[child] = priv->event_generation;
+ if (info->locked_nodes[child] == 1)
+ info->locked_nodes[child] = 0;
+ if (!afr_has_quorum(info->locked_nodes, this, NULL)) {
+ /* Since the lock was lost on quorum no. of nodes, we should
+ * not attempt to heal it anymore. Some other client could have
+ * acquired the lock, modified data and released it and this
+ * client wouldn't know about it if we heal it.*/
+ afr_mark_fd_bad(info->fd, this);
+ list_del(&info->pos);
+ afr_lk_heal_info_cleanup(info);
+ /* We're not winding an unlock on the node where the lock is still
+ * present because when fencing logic switches over to the new
+ * client (since we marked the fd bad), it should preempt any
+ * existing lock. */
+ }
+ }
+ return 0;
+}
+
+gf_boolean_t
+afr_is_consistent_io_possible(afr_local_t *local, afr_private_t *priv,
+ int32_t *op_errno)
+{
+ if (priv->consistent_io && local->call_count != priv->child_count) {
+ gf_msg(THIS->name, GF_LOG_INFO, 0, AFR_MSG_SUBVOLS_DOWN,
+ "All subvolumes are not up");
+ if (op_errno)
+ *op_errno = ENOTCONN;
+ return _gf_false;
+ }
+ return _gf_true;
+}
+
+gf_boolean_t
+afr_is_lock_mode_mandatory(dict_t *xdata)
+{
+ int ret = 0;
+ uint32_t lk_mode = GF_LK_ADVISORY;
+
+ ret = dict_get_uint32(xdata, GF_LOCK_MODE, &lk_mode);
+ if (!ret && lk_mode == GF_LK_MANDATORY)
+ return _gf_true;
+
+ return _gf_false;
+}
+
+call_frame_t *
+afr_copy_frame(call_frame_t *base)
+{
+ afr_local_t *local = NULL;
+ call_frame_t *frame = NULL;
+ int op_errno = 0;
+
+ frame = copy_frame(base);
+ if (!frame)
+ return NULL;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local) {
+ AFR_STACK_DESTROY(frame);
+ return NULL;
+ }
+
+ return frame;
+}
+
+/* Check if an entry or inode could be undergoing a transaction. */
+gf_boolean_t
+afr_is_possibly_under_txn(afr_transaction_type type, afr_local_t *local,
+ xlator_t *this)
+{
+ int i = 0;
+ int tmp = 0;
+ afr_private_t *priv = NULL;
+ GF_UNUSED char *key = NULL;
+ int keylen = 0;
+
+ priv = this->private;
+
+ if (type == AFR_ENTRY_TRANSACTION) {
+ key = GLUSTERFS_PARENT_ENTRYLK;
+ keylen = SLEN(GLUSTERFS_PARENT_ENTRYLK);
+ } else if (type == AFR_DATA_TRANSACTION) {
+ /*FIXME: Use GLUSTERFS_INODELK_DOM_COUNT etc. once
+ * pl_inodelk_xattr_fill supports separate keys for different
+ * domains.*/
+ key = GLUSTERFS_INODELK_COUNT;
+ keylen = SLEN(GLUSTERFS_INODELK_COUNT);
+ }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].xdata)
+ continue;
+ if (dict_get_int32n(local->replies[i].xdata, key, keylen, &tmp) == 0)
+ if (tmp)
+ return _gf_true;
+ }
+
+ return _gf_false;
+}
+
+static void
+afr_inode_ctx_destroy(afr_inode_ctx_t *ctx)
+{
+ int i = 0;
+
+ if (!ctx)
+ return;
+
+ for (i = 0; i < AFR_NUM_CHANGE_LOGS; i++) {
+ GF_FREE(ctx->pre_op_done[i]);
+ }
+
+ GF_FREE(ctx);
+}
+
+int
+__afr_inode_ctx_get(xlator_t *this, inode_t *inode, afr_inode_ctx_t **ctx)
+{
+ uint64_t ctx_int = 0;
+ int ret = -1;
+ int i = -1;
+ int num_locks = -1;
+ afr_inode_ctx_t *ictx = NULL;
+ afr_lock_t *lock = NULL;
+ afr_private_t *priv = this->private;
+
+ ret = __inode_ctx_get(inode, this, &ctx_int);
+ if (ret == 0) {
+ *ctx = (afr_inode_ctx_t *)(uintptr_t)ctx_int;
+ return 0;
+ }
+
+ ictx = GF_CALLOC(1, sizeof(afr_inode_ctx_t), gf_afr_mt_inode_ctx_t);
+ if (!ictx)
+ goto out;
+
+ for (i = 0; i < AFR_NUM_CHANGE_LOGS; i++) {
+ ictx->pre_op_done[i] = GF_CALLOC(sizeof *ictx->pre_op_done[i],
+ priv->child_count, gf_afr_mt_int32_t);
+ if (!ictx->pre_op_done[i]) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ }
+
+ num_locks = sizeof(ictx->lock) / sizeof(afr_lock_t);
+ for (i = 0; i < num_locks; i++) {
+ lock = &ictx->lock[i];
+ INIT_LIST_HEAD(&lock->post_op);
+ INIT_LIST_HEAD(&lock->frozen);
+ INIT_LIST_HEAD(&lock->waiting);
+ INIT_LIST_HEAD(&lock->owners);
+ }
+
+ ctx_int = (uint64_t)(uintptr_t)ictx;
+ ret = __inode_ctx_set(inode, this, &ctx_int);
+ if (ret) {
+ goto out;
+ }
+
+ ictx->spb_choice = -1;
+ ictx->read_subvol = 0;
+ ictx->write_subvol = 0;
+ ictx->lock_count = 0;
+ ret = 0;
+ *ctx = ictx;
out:
- if (ret) {
- afr_inode_ctx_destroy (ictx);
- }
- return ret;
+ if (ret) {
+ afr_inode_ctx_destroy(ictx);
+ }
+ return ret;
}
/*
@@ -226,1694 +863,1681 @@ out:
*/
int
-__afr_set_in_flight_sb_status (xlator_t *this, afr_local_t *local,
- inode_t *inode)
-{
- int i = 0;
- int txn_type = 0;
- int count = 0;
- int index = -1;
- uint16_t datamap_old = 0;
- uint16_t metadatamap_old = 0;
- uint16_t datamap = 0;
- uint16_t metadatamap = 0;
- uint16_t tmp_map = 0;
- uint16_t mask = 0;
- uint32_t event = 0;
- uint64_t val = 0;
- afr_private_t *priv = NULL;
-
- priv = this->private;
- txn_type = local->transaction.type;
-
- if (txn_type == AFR_DATA_TRANSACTION)
- val = local->inode_ctx->write_subvol;
- else
- val = local->inode_ctx->read_subvol;
-
- metadatamap_old = metadatamap = (val & 0x000000000000ffff);
- datamap_old = datamap = (val & 0x00000000ffff0000) >> 16;
- event = (val & 0xffffffff00000000) >> 32;
-
- if (txn_type == AFR_DATA_TRANSACTION)
- tmp_map = datamap;
- else if (txn_type == AFR_METADATA_TRANSACTION)
- tmp_map = metadatamap;
-
- count = gf_bits_count (tmp_map);
-
- if (count == 1)
- index = gf_bits_index (tmp_map);
-
- for (i = 0; i < priv->child_count; i++) {
- mask = 0;
- if (!local->transaction.failed_subvols[i])
- continue;
-
- mask = 1 << i;
- if (txn_type == AFR_METADATA_TRANSACTION)
- metadatamap &= ~mask;
- else if (txn_type == AFR_DATA_TRANSACTION)
- datamap &= ~mask;
- }
-
- switch (txn_type) {
+__afr_set_in_flight_sb_status(xlator_t *this, afr_local_t *local,
+ inode_t *inode)
+{
+ int i = 0;
+ int txn_type = 0;
+ int count = 0;
+ int index = -1;
+ uint16_t datamap_old = 0;
+ uint16_t metadatamap_old = 0;
+ uint16_t datamap = 0;
+ uint16_t metadatamap = 0;
+ uint16_t tmp_map = 0;
+ uint16_t mask = 0;
+ uint32_t event = 0;
+ uint64_t val = 0;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ txn_type = local->transaction.type;
+
+ if (txn_type == AFR_DATA_TRANSACTION)
+ val = local->inode_ctx->write_subvol;
+ else
+ val = local->inode_ctx->read_subvol;
+
+ metadatamap_old = metadatamap = (val & 0x000000000000ffff);
+ datamap_old = datamap = (val & 0x00000000ffff0000) >> 16;
+ event = (val & 0xffffffff00000000) >> 32;
+
+ if (txn_type == AFR_DATA_TRANSACTION)
+ tmp_map = datamap;
+ else if (txn_type == AFR_METADATA_TRANSACTION)
+ tmp_map = metadatamap;
+
+ count = gf_bits_count(tmp_map);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.failed_subvols[i])
+ continue;
+
+ mask = 1 << i;
+ if (txn_type == AFR_METADATA_TRANSACTION)
+ metadatamap &= ~mask;
+ else if (txn_type == AFR_DATA_TRANSACTION)
+ datamap &= ~mask;
+ }
+
+ switch (txn_type) {
case AFR_METADATA_TRANSACTION:
- if ((metadatamap_old != 0) && (metadatamap == 0) &&
- (count == 1)) {
- local->transaction.in_flight_sb_errno =
- local->replies[index].op_errno;
- local->transaction.in_flight_sb = _gf_true;
- metadatamap |= (1 << index);
- }
- if (metadatamap_old != metadatamap) {
- event = 0;
- }
- break;
+ if ((metadatamap_old != 0) && (metadatamap == 0) && (count == 1)) {
+ index = gf_bits_index(tmp_map);
+ local->transaction.in_flight_sb_errno = local->replies[index]
+ .op_errno;
+ local->transaction.in_flight_sb = _gf_true;
+ metadatamap |= (1 << index);
+ }
+ if (metadatamap_old != metadatamap) {
+ __afr_inode_need_refresh_set(inode, this);
+ }
+ break;
case AFR_DATA_TRANSACTION:
- if ((datamap_old != 0) && (datamap == 0) && (count == 1)) {
- local->transaction.in_flight_sb_errno =
- local->replies[index].op_errno;
- local->transaction.in_flight_sb = _gf_true;
- datamap |= (1 << index);
- }
- if (datamap_old != datamap)
- event = 0;
- break;
+ if ((datamap_old != 0) && (datamap == 0) && (count == 1)) {
+ index = gf_bits_index(tmp_map);
+ local->transaction.in_flight_sb_errno = local->replies[index]
+ .op_errno;
+ local->transaction.in_flight_sb = _gf_true;
+ datamap |= (1 << index);
+ }
+ if (datamap_old != datamap)
+ __afr_inode_need_refresh_set(inode, this);
+ break;
default:
- break;
- }
+ break;
+ }
- val = ((uint64_t) metadatamap) |
- (((uint64_t) datamap) << 16) |
- (((uint64_t) event) << 32);
+ val = ((uint64_t)metadatamap) | (((uint64_t)datamap) << 16) |
+ (((uint64_t)event) << 32);
- if (txn_type == AFR_DATA_TRANSACTION)
- local->inode_ctx->write_subvol = val;
- local->inode_ctx->read_subvol = val;
+ if (txn_type == AFR_DATA_TRANSACTION)
+ local->inode_ctx->write_subvol = val;
+ local->inode_ctx->read_subvol = val;
- return 0;
+ return 0;
}
gf_boolean_t
-afr_is_symmetric_error (call_frame_t *frame, xlator_t *this)
+afr_is_symmetric_error(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int op_errno = 0;
- int i_errno = 0;
- gf_boolean_t matching_errors = _gf_true;
- int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int op_errno = 0;
+ int i_errno = 0;
+ gf_boolean_t matching_errors = _gf_true;
+ int i = 0;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
- if (local->replies[i].op_ret != -1) {
- /* Operation succeeded on at least one subvol,
- so it is not a failed-everywhere situation.
- */
- matching_errors = _gf_false;
- break;
- }
- i_errno = local->replies[i].op_errno;
-
- if (i_errno == ENOTCONN) {
- /* ENOTCONN is not a symmetric error. We do not
- know if the operation was performed on the
- backend or not.
- */
- matching_errors = _gf_false;
- break;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+ if (local->replies[i].op_ret != -1) {
+ /* Operation succeeded on at least one subvol,
+ so it is not a failed-everywhere situation.
+ */
+ matching_errors = _gf_false;
+ break;
+ }
+ i_errno = local->replies[i].op_errno;
- if (!op_errno) {
- op_errno = i_errno;
- } else if (op_errno != i_errno) {
- /* Mismatching op_errno's */
- matching_errors = _gf_false;
- break;
- }
+ if (i_errno == ENOTCONN) {
+ /* ENOTCONN is not a symmetric error. We do not
+ know if the operation was performed on the
+ backend or not.
+ */
+ matching_errors = _gf_false;
+ break;
}
- return matching_errors;
+ if (!op_errno) {
+ op_errno = i_errno;
+ } else if (op_errno != i_errno) {
+ /* Mismatching op_errno's */
+ matching_errors = _gf_false;
+ break;
+ }
+ }
+
+ return matching_errors;
}
int
-afr_set_in_flight_sb_status (xlator_t *this, call_frame_t *frame,
- inode_t *inode)
+afr_set_in_flight_sb_status(xlator_t *this, call_frame_t *frame, inode_t *inode)
{
- int ret = -1;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
+ int ret = -1;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- /* If this transaction saw no failures, then exit. */
- if (AFR_COUNT (local->transaction.failed_subvols,
- priv->child_count) == 0)
- return 0;
+ /* If this transaction saw no failures, then exit. */
+ if (AFR_COUNT(local->transaction.failed_subvols, priv->child_count) == 0)
+ return 0;
- if (afr_is_symmetric_error (frame, this))
- return 0;
+ if (afr_is_symmetric_error(frame, this))
+ return 0;
- LOCK (&inode->lock);
- {
- ret = __afr_set_in_flight_sb_status (this, local, inode);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __afr_set_in_flight_sb_status(this, local, inode);
+ }
+ UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
int
-__afr_inode_read_subvol_get_small (inode_t *inode, xlator_t *this,
- unsigned char *data, unsigned char *metadata,
- int *event_p)
-{
- afr_private_t *priv = NULL;
- int ret = -1;
- uint16_t datamap = 0;
- uint16_t metadatamap = 0;
- uint32_t event = 0;
- uint64_t val = 0;
- int i = 0;
- afr_inode_ctx_t *ctx = NULL;
-
- priv = this->private;
-
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret < 0)
- return ret;
+__afr_inode_read_subvol_get_small(inode_t *inode, xlator_t *this,
+ unsigned char *data, unsigned char *metadata,
+ int *event_p)
+{
+ afr_private_t *priv = NULL;
+ int ret = -1;
+ uint16_t datamap = 0;
+ uint16_t metadatamap = 0;
+ uint32_t event = 0;
+ uint64_t val = 0;
+ int i = 0;
+ afr_inode_ctx_t *ctx = NULL;
+
+ priv = this->private;
+
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
+ if (ret < 0)
+ return ret;
- val = ctx->read_subvol;
+ val = ctx->read_subvol;
- metadatamap = (val & 0x000000000000ffff);
- datamap = (val & 0x00000000ffff0000) >> 16;
- event = (val & 0xffffffff00000000) >> 32;
+ metadatamap = (val & 0x000000000000ffff);
+ datamap = (val & 0x00000000ffff0000) >> 16;
+ event = (val & 0xffffffff00000000) >> 32;
- for (i = 0; i < priv->child_count; i++) {
- if (metadata)
- metadata[i] = (metadatamap >> i) & 1;
- if (data)
- data[i] = (datamap >> i) & 1;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (metadata)
+ metadata[i] = (metadatamap >> i) & 1;
+ if (data)
+ data[i] = (datamap >> i) & 1;
+ }
- if (event_p)
- *event_p = event;
- return ret;
+ if (event_p)
+ *event_p = event;
+ return ret;
}
-
int
-__afr_inode_read_subvol_set_small (inode_t *inode, xlator_t *this,
- unsigned char *data, unsigned char *metadata,
- int event)
+__afr_inode_read_subvol_set_small(inode_t *inode, xlator_t *this,
+ unsigned char *data, unsigned char *metadata,
+ int event)
{
- afr_private_t *priv = NULL;
- uint16_t datamap = 0;
- uint16_t metadatamap = 0;
- uint64_t val = 0;
- int i = 0;
- int ret = -1;
- afr_inode_ctx_t *ctx = NULL;
+ afr_private_t *priv = NULL;
+ uint16_t datamap = 0;
+ uint16_t metadatamap = 0;
+ uint64_t val = 0;
+ int i = 0;
+ int ret = -1;
+ afr_inode_ctx_t *ctx = NULL;
- priv = this->private;
+ priv = this->private;
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret)
- goto out;
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
+ if (ret)
+ goto out;
- for (i = 0; i < priv->child_count; i++) {
- if (data[i])
- datamap |= (1 << i);
- if (metadata[i])
- metadatamap |= (1 << i);
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (data[i])
+ datamap |= (1 << i);
+ if (metadata[i])
+ metadatamap |= (1 << i);
+ }
- val = ((uint64_t) metadatamap) |
- (((uint64_t) datamap) << 16) |
- (((uint64_t) event) << 32);
+ val = ((uint64_t)metadatamap) | (((uint64_t)datamap) << 16) |
+ (((uint64_t)event) << 32);
- ctx->read_subvol = val;
+ ctx->read_subvol = val;
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-__afr_inode_event_gen_reset_small (inode_t *inode, xlator_t *this)
+__afr_inode_read_subvol_get(inode_t *inode, xlator_t *this, unsigned char *data,
+ unsigned char *metadata, int *event_p)
{
- int ret = -1;
- uint16_t datamap = 0;
- uint16_t metadatamap = 0;
- uint32_t event = 0;
- uint64_t val = 0;
- afr_inode_ctx_t *ctx = NULL;
-
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret)
- return ret;
-
- val = ctx->read_subvol;
-
- metadatamap = (val & 0x000000000000ffff) >> 0;
- datamap = (val & 0x00000000ffff0000) >> 16;
- event = 0;
+ afr_private_t *priv = NULL;
+ int ret = -1;
- val = ((uint64_t) metadatamap) |
- (((uint64_t) datamap) << 16) |
- (((uint64_t) event) << 32);
+ priv = this->private;
- ctx->read_subvol = val;
+ if (priv->child_count <= 16)
+ ret = __afr_inode_read_subvol_get_small(inode, this, data, metadata,
+ event_p);
+ else
+ /* TBD: allocate structure with array and read from it */
+ ret = -1;
- return ret;
+ return ret;
}
-
int
-__afr_inode_read_subvol_get (inode_t *inode, xlator_t *this,
- unsigned char *data, unsigned char *metadata,
- int *event_p)
+__afr_inode_split_brain_choice_get(inode_t *inode, xlator_t *this,
+ int *spb_choice)
{
- afr_private_t *priv = NULL;
- int ret = -1;
-
- priv = this->private;
+ afr_inode_ctx_t *ctx = NULL;
+ int ret = -1;
- if (priv->child_count <= 16)
- ret = __afr_inode_read_subvol_get_small (inode, this, data,
- metadata, event_p);
- else
- /* TBD: allocate structure with array and read from it */
- ret = -1;
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
+ if (ret < 0)
+ return ret;
- return ret;
+ *spb_choice = ctx->spb_choice;
+ return 0;
}
int
-__afr_inode_split_brain_choice_get (inode_t *inode, xlator_t *this,
- int *spb_choice)
+__afr_inode_read_subvol_set(inode_t *inode, xlator_t *this, unsigned char *data,
+ unsigned char *metadata, int event)
{
- afr_inode_ctx_t *ctx = NULL;
- int ret = -1;
+ afr_private_t *priv = NULL;
+ int ret = -1;
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret < 0)
- return ret;
+ priv = this->private;
- *spb_choice = ctx->spb_choice;
- return 0;
+ if (priv->child_count <= 16)
+ ret = __afr_inode_read_subvol_set_small(inode, this, data, metadata,
+ event);
+ else
+ ret = -1;
+
+ return ret;
}
int
-__afr_inode_read_subvol_set (inode_t *inode, xlator_t *this, unsigned char *data,
- unsigned char *metadata, int event)
+__afr_inode_split_brain_choice_set(inode_t *inode, xlator_t *this,
+ int spb_choice)
{
- afr_private_t *priv = NULL;
- int ret = -1;
+ afr_inode_ctx_t *ctx = NULL;
+ int ret = -1;
- priv = this->private;
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
+ if (ret)
+ goto out;
- if (priv->child_count <= 16)
- ret = __afr_inode_read_subvol_set_small (inode, this, data,
- metadata, event);
- else
- ret = -1;
+ ctx->spb_choice = spb_choice;
- return ret;
+ ret = 0;
+out:
+ return ret;
}
int
-__afr_inode_split_brain_choice_set (inode_t *inode, xlator_t *this,
- int spb_choice)
+afr_inode_read_subvol_get(inode_t *inode, xlator_t *this, unsigned char *data,
+ unsigned char *metadata, int *event_p)
{
- afr_inode_ctx_t *ctx = NULL;
- int ret = -1;
+ int ret = -1;
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret)
- goto out;
-
- ctx->spb_choice = spb_choice;
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- ret = 0;
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_read_subvol_get(inode, this, data, metadata, event_p);
+ }
+ UNLOCK(&inode->lock);
out:
- return ret;
+ return ret;
}
int
-__afr_inode_event_gen_reset (inode_t *inode, xlator_t *this)
+afr_inode_get_readable(call_frame_t *frame, inode_t *inode, xlator_t *this,
+ unsigned char *readable, int *event_p, int type)
{
- afr_private_t *priv = NULL;
- int ret = -1;
-
- priv = this->private;
-
- if (priv->child_count <= 16)
- ret = __afr_inode_event_gen_reset_small (inode, this);
- else
- ret = -1;
+ afr_private_t *priv = this->private;
+ afr_local_t *local = frame->local;
+ unsigned char *data = alloca0(priv->child_count);
+ unsigned char *metadata = alloca0(priv->child_count);
+ int data_count = 0;
+ int metadata_count = 0;
+ int event_generation = 0;
+ int ret = 0;
+
+ ret = afr_inode_read_subvol_get(inode, this, data, metadata,
+ &event_generation);
+ if (ret == -1)
+ return -EIO;
+
+ data_count = AFR_COUNT(data, priv->child_count);
+ metadata_count = AFR_COUNT(metadata, priv->child_count);
+
+ if (inode->ia_type == IA_IFDIR) {
+ /* For directories, allow even if it is in data split-brain. */
+ if (type == AFR_METADATA_TRANSACTION || local->op == GF_FOP_STAT ||
+ local->op == GF_FOP_FSTAT) {
+ if (!metadata_count)
+ return -EIO;
+ }
+ } else {
+ /* For files, abort in case of data/metadata split-brain. */
+ if (!data_count || !metadata_count) {
+ return -EIO;
+ }
+ }
- return ret;
+ if (type == AFR_METADATA_TRANSACTION && readable)
+ memcpy(readable, metadata, priv->child_count * sizeof *metadata);
+ if (type == AFR_DATA_TRANSACTION && readable) {
+ if (!data_count)
+ memcpy(readable, local->child_up,
+ priv->child_count * sizeof *readable);
+ else
+ memcpy(readable, data, priv->child_count * sizeof *data);
+ }
+ if (event_p)
+ *event_p = event_generation;
+ return 0;
}
-
-int
-afr_inode_read_subvol_get (inode_t *inode, xlator_t *this, unsigned char *data,
- unsigned char *metadata, int *event_p)
+static int
+afr_inode_split_brain_choice_get(inode_t *inode, xlator_t *this,
+ int *spb_choice)
{
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ int ret = -1;
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- LOCK(&inode->lock);
- {
- ret = __afr_inode_read_subvol_get (inode, this, data,
- metadata, event_p);
- }
- UNLOCK(&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_split_brain_choice_get(inode, this, spb_choice);
+ }
+ UNLOCK(&inode->lock);
out:
- return ret;
+ return ret;
}
+/*
+ * frame is used to get the favourite policy. Since
+ * afr_inode_split_brain_choice_get was called with afr_open, it is possible to
+ * have a frame with out local->replies. So in that case, frame is passed as
+ * null, hence this function will handle the frame NULL case.
+ */
int
-afr_inode_get_readable (call_frame_t *frame, inode_t *inode, xlator_t *this,
- unsigned char *readable, int *event_p, int type)
+afr_split_brain_read_subvol_get(inode_t *inode, xlator_t *this,
+ call_frame_t *frame, int *spb_subvol)
{
+ int ret = -1;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- afr_private_t *priv = this->private;
- afr_local_t *local = frame->local;
- unsigned char *data = alloca0 (priv->child_count);
- unsigned char *metadata = alloca0 (priv->child_count);
- int data_count = 0;
- int metadata_count = 0;
- int event_generation = 0;
- int ret = 0;
-
- ret = afr_inode_read_subvol_get (inode, this, data, metadata,
- &event_generation);
- if (ret == -1)
- return -EIO;
+ GF_VALIDATE_OR_GOTO("afr", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, spb_subvol, out);
- data_count = AFR_COUNT (data, priv->child_count);
- metadata_count = AFR_COUNT (metadata, priv->child_count);
+ priv = this->private;
- if (inode->ia_type == IA_IFDIR) {
- /* For directories, allow even if it is in data split-brain. */
- if (type == AFR_METADATA_TRANSACTION ||
- local->op == GF_FOP_STAT || local->op == GF_FOP_FSTAT) {
- if (!metadata_count)
- return -EIO;
- }
- } else {
- /* For files, abort in case of data/metadata split-brain. */
- if (!data_count || !metadata_count) {
- return -EIO;
- }
- }
-
- if (type == AFR_METADATA_TRANSACTION && readable)
- memcpy (readable, metadata, priv->child_count * sizeof *metadata);
- if (type == AFR_DATA_TRANSACTION && readable) {
- if (!data_count)
- memcpy (readable, local->child_up,
- priv->child_count * sizeof *readable);
- else
- memcpy (readable, data, priv->child_count * sizeof *data);
+ ret = afr_inode_split_brain_choice_get(inode, this, spb_subvol);
+ if (*spb_subvol < 0 && priv->fav_child_policy && frame && frame->local) {
+ local = frame->local;
+ *spb_subvol = afr_sh_get_fav_by_policy(this, local->replies, inode,
+ NULL);
+ if (*spb_subvol >= 0) {
+ ret = 0;
}
- if (event_p)
- *event_p = event_generation;
- return 0;
-}
+ }
-int
-afr_inode_split_brain_choice_get (inode_t *inode, xlator_t *this,
- int *spb_choice)
-{
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
-
- LOCK(&inode->lock);
- {
- ret = __afr_inode_split_brain_choice_get (inode, this,
- spb_choice);
- }
- UNLOCK(&inode->lock);
out:
- return ret;
+ return ret;
}
-
-
int
-afr_inode_read_subvol_set (inode_t *inode, xlator_t *this, unsigned char *data,
- unsigned char *metadata, int event)
+afr_inode_read_subvol_set(inode_t *inode, xlator_t *this, unsigned char *data,
+ unsigned char *metadata, int event)
{
- int ret = -1;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- LOCK(&inode->lock);
- {
- ret = __afr_inode_read_subvol_set (inode, this, data, metadata,
- event);
- }
- UNLOCK(&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_read_subvol_set(inode, this, data, metadata, event);
+ }
+ UNLOCK(&inode->lock);
out:
- return ret;
+ return ret;
}
-
int
-afr_inode_split_brain_choice_set (inode_t *inode, xlator_t *this,
- int spb_choice)
+afr_inode_split_brain_choice_set(inode_t *inode, xlator_t *this, int spb_choice)
{
- int ret = -1;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- LOCK(&inode->lock);
- {
- ret = __afr_inode_split_brain_choice_set (inode, this,
- spb_choice);
- }
- UNLOCK(&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_split_brain_choice_set(inode, this, spb_choice);
+ }
+ UNLOCK(&inode->lock);
out:
- return ret;
+ return ret;
}
-
/* The caller of this should perform afr_inode_refresh, if this function
* returns _gf_true
*/
gf_boolean_t
-afr_is_inode_refresh_reqd (inode_t *inode, xlator_t *this,
- int event_gen1, int event_gen2)
+afr_is_inode_refresh_reqd(inode_t *inode, xlator_t *this, int event_gen1,
+ int event_gen2)
{
- gf_boolean_t need_refresh = _gf_false;
- afr_inode_ctx_t *ctx = NULL;
- int ret = -1;
+ gf_boolean_t need_refresh = _gf_false;
+ afr_inode_ctx_t *ctx = NULL;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- LOCK(&inode->lock);
- {
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret)
- goto unlock;
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
+ if (ret)
+ goto unlock;
- need_refresh = ctx->need_refresh;
- /* Hoping that the caller will do inode_refresh followed by
- * this, hence setting the need_refresh to false */
- ctx->need_refresh = _gf_false;
- }
+ need_refresh = ctx->need_refresh;
+ /* Hoping that the caller will do inode_refresh followed by
+ * this, hence setting the need_refresh to false */
+ ctx->need_refresh = _gf_false;
+ }
unlock:
- UNLOCK(&inode->lock);
+ UNLOCK(&inode->lock);
- if (event_gen1 != event_gen2)
- need_refresh = _gf_true;
+ if (event_gen1 != event_gen2)
+ need_refresh = _gf_true;
out:
- return need_refresh;
+ return need_refresh;
}
-
-static int
-afr_inode_need_refresh_set (inode_t *inode, xlator_t *this)
+int
+__afr_inode_need_refresh_set(inode_t *inode, xlator_t *this)
{
- int ret = -1;
- afr_inode_ctx_t *ctx = NULL;
-
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ int ret = -1;
+ afr_inode_ctx_t *ctx = NULL;
- LOCK(&inode->lock);
- {
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret)
- goto unlock;
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
+ if (ret == 0) {
+ ctx->need_refresh = _gf_true;
+ }
- ctx->need_refresh = _gf_true;
- }
-unlock:
- UNLOCK(&inode->lock);
-out:
- return ret;
+ return ret;
}
int
-afr_inode_event_gen_reset (inode_t *inode, xlator_t *this)
+afr_inode_need_refresh_set(inode_t *inode, xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- LOCK(&inode->lock);
- {
- ret = __afr_inode_event_gen_reset (inode, this);
- }
- UNLOCK(&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_need_refresh_set(inode, this);
+ }
+ UNLOCK(&inode->lock);
out:
- return ret;
+ return ret;
}
int
-afr_spb_choice_timeout_cancel (xlator_t *this, inode_t *inode)
+afr_spb_choice_timeout_cancel(xlator_t *this, inode_t *inode)
{
- afr_inode_ctx_t *ctx = NULL;
- int ret = -1;
+ afr_inode_ctx_t *ctx = NULL;
+ int ret = -1;
- if (!inode)
- return ret;
+ if (!inode)
+ return ret;
- LOCK(&inode->lock);
- {
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret < 0 || !ctx) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
- "Failed to cancel split-brain choice timer.");
- goto out;
- }
- ctx->spb_choice = -1;
- if (ctx->timer) {
- gf_timer_call_cancel (this->ctx, ctx->timer);
- ctx->timer = NULL;
- }
- ret = 0;
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
+ if (ret < 0 || !ctx) {
+ UNLOCK(&inode->lock);
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
+ "Failed to cancel split-brain choice timer.");
+ goto out;
+ }
+ ctx->spb_choice = -1;
+ if (ctx->timer) {
+ gf_timer_call_cancel(this->ctx, ctx->timer);
+ ctx->timer = NULL;
}
+ ret = 0;
+ }
+ UNLOCK(&inode->lock);
out:
- UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
void
-afr_set_split_brain_choice_cbk (void *data)
+afr_set_split_brain_choice_cbk(void *data)
{
- inode_t *inode = data;
- xlator_t *this = THIS;
+ inode_t *inode = data;
+ xlator_t *this = THIS;
- afr_spb_choice_timeout_cancel (this, inode);
- inode_invalidate (inode);
- inode_unref (inode);
- return;
+ afr_spb_choice_timeout_cancel(this, inode);
+ inode_invalidate(inode);
+ inode_unref(inode);
+ return;
}
-
int
-afr_set_split_brain_choice (int ret, call_frame_t *frame, void *opaque)
-{
- int op_errno = ENOMEM;
- afr_private_t *priv = NULL;
- afr_inode_ctx_t *ctx = NULL;
- inode_t *inode = NULL;
- loc_t *loc = NULL;
- xlator_t *this = NULL;
- afr_spbc_timeout_t *data = opaque;
- struct timespec delta = {0, };
- gf_boolean_t timer_set = _gf_false;
- gf_boolean_t timer_cancelled = _gf_false;
- gf_boolean_t timer_reset = _gf_false;
- gf_boolean_t need_invalidate = _gf_true;
- int old_spb_choice = -1;
-
- frame = data->frame;
- loc = data->loc;
- this = frame->this;
- priv = this->private;
-
+afr_set_split_brain_choice(int ret, call_frame_t *frame, void *opaque)
+{
+ int op_errno = ENOMEM;
+ afr_private_t *priv = NULL;
+ afr_inode_ctx_t *ctx = NULL;
+ inode_t *inode = NULL;
+ loc_t *loc = NULL;
+ xlator_t *this = NULL;
+ afr_spbc_timeout_t *data = opaque;
+ struct timespec delta = {
+ 0,
+ };
+ gf_boolean_t timer_set = _gf_false;
+ gf_boolean_t timer_cancelled = _gf_false;
+ gf_boolean_t timer_reset = _gf_false;
+ int old_spb_choice = -1;
+
+ frame = data->frame;
+ loc = data->loc;
+ this = frame->this;
+ priv = this->private;
+
+ if (ret) {
+ op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ delta.tv_sec = priv->spb_choice_timeout;
+ delta.tv_nsec = 0;
+
+ if (!loc->inode) {
+ ret = -1;
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ if (!(data->d_spb || data->m_spb)) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
+ "Cannot set "
+ "replica.split-brain-choice on %s. File is"
+ " not in data/metadata split-brain.",
+ uuid_utoa(loc->gfid));
+ ret = -1;
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ /*
+ * we're ref'ing the inode before LOCK like it is done elsewhere in the
+ * code. If we ref after LOCK, coverity complains of possible deadlocks.
+ */
+ inode = inode_ref(loc->inode);
+
+ LOCK(&inode->lock);
+ {
+ ret = __afr_inode_ctx_get(this, inode, &ctx);
if (ret) {
- op_errno = -ret;
- ret = -1;
- goto out;
+ UNLOCK(&inode->lock);
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
+ "Failed to get inode_ctx for %s", loc->name);
+ goto post_unlock;
}
- delta.tv_sec = priv->spb_choice_timeout;
- delta.tv_nsec = 0;
+ old_spb_choice = ctx->spb_choice;
+ ctx->spb_choice = data->spb_child_index;
- if (!loc->inode) {
- ret = -1;
- op_errno = EINVAL;
- goto out;
- }
-
- if (!(data->d_spb || data->m_spb)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR, "Cannot set "
- "replica.split-brain-choice on %s. File is"
- " not in data/metadata split-brain.",
- uuid_utoa (loc->gfid));
- ret = -1;
- op_errno = EINVAL;
- goto out;
- }
-
- /*
- * we're ref'ing the inode before LOCK like it is done elsewhere in the
- * code. If we ref after LOCK, coverity complains of possible deadlocks.
+ /* Possible changes in spb-choice :
+ * valid to -1 : cancel timer and unref
+ * valid to valid : cancel timer and inject new one
+ * -1 to -1 : unref and do not do anything
+ * -1 to valid : inject timer
*/
- inode = inode_ref (loc->inode);
- LOCK(&inode->lock);
- {
- ret = __afr_inode_ctx_get (this, inode, &ctx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
- "Failed to get inode_ctx for %s", loc->name);
- goto unlock;
+ /* ctx->timer is NULL iff previous value of
+ * ctx->spb_choice is -1
+ */
+ if (ctx->timer) {
+ if (ctx->spb_choice == -1) {
+ if (!gf_timer_call_cancel(this->ctx, ctx->timer)) {
+ ctx->timer = NULL;
+ timer_cancelled = _gf_true;
}
-
- old_spb_choice = ctx->spb_choice;
- ctx->spb_choice = data->spb_child_index;
-
- /* Possible changes in spb-choice :
- * valid to -1 : cancel timer and unref
- * valid to valid : cancel timer and inject new one
- * -1 to -1 : unref and do not do anything
- * -1 to valid : inject timer
- */
-
- /* ctx->timer is NULL iff previous value of
- * ctx->spb_choice is -1
+ /* If timer cancel failed here it means that the
+ * previous cbk will be executed which will set
+ * spb_choice to -1. So we can consider the
+ * 'valid to -1' case to be a success
+ * (i.e. ret = 0) and goto unlock.
*/
- if (ctx->timer) {
- if (ctx->spb_choice == -1) {
- if (!gf_timer_call_cancel (this->ctx,
- ctx->timer)) {
- ctx->timer = NULL;
- timer_cancelled = _gf_true;
- }
- /* If timer cancel failed here it means that the
- * previous cbk will be executed which will set
- * spb_choice to -1. So we can consider the
- * 'valid to -1' case to be a sucess
- * (i.e. ret = 0) and goto unlock.
- */
- goto unlock;
- }
- goto reset_timer;
- } else {
- if (ctx->spb_choice == -1)
- goto unlock;
- goto set_timer;
- }
-
-reset_timer:
- ret = gf_timer_call_cancel (this->ctx, ctx->timer);
- if (ret != 0) {
- /* We need to bail out now instead of launching a new
- * timer. Otherwise the cbk of the previous timer event
- * will cancel the new ctx->timer.
- */
- ctx->spb_choice = old_spb_choice;
- ret = -1;
- op_errno = EAGAIN;
- goto unlock;
- }
- ctx->timer = NULL;
- timer_reset = _gf_true;
-
-set_timer:
- ctx->timer = gf_timer_call_after (this->ctx, delta,
- afr_set_split_brain_choice_cbk,
- inode);
- if (!ctx->timer) {
- ctx->spb_choice = old_spb_choice;
- ret = -1;
- op_errno = ENOMEM;
- }
- if (!timer_reset && ctx->timer)
- timer_set = _gf_true;
- if (timer_reset && !ctx->timer)
- timer_cancelled = _gf_true;
- need_invalidate = _gf_false;
+ goto unlock;
+ }
+ goto reset_timer;
+ } else {
+ if (ctx->spb_choice == -1)
+ goto unlock;
+ goto set_timer;
}
+
+ reset_timer:
+ ret = gf_timer_call_cancel(this->ctx, ctx->timer);
+ if (ret != 0) {
+ /* We need to bail out now instead of launching a new
+ * timer. Otherwise the cbk of the previous timer event
+ * will cancel the new ctx->timer.
+ */
+ ctx->spb_choice = old_spb_choice;
+ ret = -1;
+ op_errno = EAGAIN;
+ goto unlock;
+ }
+ ctx->timer = NULL;
+ timer_reset = _gf_true;
+
+ set_timer:
+ ctx->timer = gf_timer_call_after(this->ctx, delta,
+ afr_set_split_brain_choice_cbk, inode);
+ if (!ctx->timer) {
+ ctx->spb_choice = old_spb_choice;
+ ret = -1;
+ op_errno = ENOMEM;
+ }
+ if (!timer_reset && ctx->timer)
+ timer_set = _gf_true;
+ if (timer_reset && !ctx->timer)
+ timer_cancelled = _gf_true;
+ }
unlock:
- UNLOCK(&inode->lock);
- if (!timer_set)
- inode_unref (inode);
- if (timer_cancelled)
- inode_unref (inode);
- /*
- * We need to invalidate the inode to prevent the kernel from serving
- * reads from an older cached value despite a change in spb_choice to
- * a new value.
- */
- if (need_invalidate)
- inode_invalidate (inode);
+ UNLOCK(&inode->lock);
+post_unlock:
+ if (!timer_set)
+ inode_unref(inode);
+ if (timer_cancelled)
+ inode_unref(inode);
+ /*
+ * We need to invalidate the inode to prevent the kernel from serving
+ * reads from an older cached value despite a change in spb_choice to
+ * a new value.
+ */
+ inode_invalidate(inode);
out:
- GF_FREE (data);
- AFR_STACK_UNWIND (setxattr, frame, ret, op_errno, NULL);
- return 0;
+ GF_FREE(data);
+ AFR_STACK_UNWIND(setxattr, frame, ret, op_errno, NULL);
+ return 0;
}
int
-afr_accused_fill (xlator_t *this, dict_t *xdata, unsigned char *accused,
- afr_transaction_type type)
+afr_accused_fill(xlator_t *this, dict_t *xdata, unsigned char *accused,
+ afr_transaction_type type)
{
- afr_private_t *priv = NULL;
- int i = 0;
- int idx = afr_index_for_transaction_type (type);
- void *pending_raw = NULL;
- int pending[3];
- int ret = 0;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int idx = afr_index_for_transaction_type(type);
+ void *pending_raw = NULL;
+ int pending[3];
+ int ret = 0;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_get_ptr (xdata, priv->pending_key[i],
- &pending_raw);
- if (ret) /* no pending flags */
- continue;
- memcpy (pending, pending_raw, sizeof(pending));
+ for (i = 0; i < priv->child_count; i++) {
+ ret = dict_get_ptr(xdata, priv->pending_key[i], &pending_raw);
+ if (ret) /* no pending flags */
+ continue;
+ memcpy(pending, pending_raw, sizeof(pending));
- if (ntoh32 (pending[idx]))
- accused[i] = 1;
- }
+ if (ntoh32(pending[idx]))
+ accused[i] = 1;
+ }
- return 0;
+ return 0;
}
int
-afr_accuse_smallfiles (xlator_t *this, struct afr_reply *replies,
- unsigned char *data_accused)
+afr_accuse_smallfiles(xlator_t *this, struct afr_reply *replies,
+ unsigned char *data_accused)
{
- int i = 0;
- afr_private_t *priv = NULL;
- uint64_t maxsize = 0;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ uint64_t maxsize = 0;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (replies[i].valid && replies[i].xdata &&
- dict_get (replies[i].xdata, GLUSTERFS_BAD_INODE))
- continue;
- if (data_accused[i])
- continue;
- if (replies[i].poststat.ia_size > maxsize)
- maxsize = replies[i].poststat.ia_size;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (replies[i].valid && replies[i].xdata &&
+ dict_get_sizen(replies[i].xdata, GLUSTERFS_BAD_INODE))
+ continue;
+ if (data_accused[i])
+ continue;
+ if (replies[i].poststat.ia_size > maxsize)
+ maxsize = replies[i].poststat.ia_size;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (data_accused[i])
- continue;
- if (AFR_IS_ARBITER_BRICK(priv, i))
- continue;
- if (replies[i].poststat.ia_size < maxsize)
- data_accused[i] = 1;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (data_accused[i])
+ continue;
+ if (AFR_IS_ARBITER_BRICK(priv, i))
+ continue;
+ if (replies[i].poststat.ia_size < maxsize)
+ data_accused[i] = 1;
+ }
- return 0;
+ return 0;
}
int
-afr_readables_fill (call_frame_t *frame, xlator_t *this, inode_t *inode,
- unsigned char *data_accused,
- unsigned char *metadata_accused,
- unsigned char *data_readable,
- unsigned char *metadata_readable,
- struct afr_reply *replies)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- dict_t *xdata = NULL;
- int i = 0;
- int ret = 0;
- ia_type_t ia_type = IA_INVAL;
-
- local = frame->local;
- priv = this->private;
+afr_readables_fill(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ unsigned char *data_accused, unsigned char *metadata_accused,
+ unsigned char *data_readable,
+ unsigned char *metadata_readable, struct afr_reply *replies)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ dict_t *xdata = NULL;
+ int i = 0;
+ int ret = 0;
+ ia_type_t ia_type = IA_INVAL;
+
+ local = frame->local;
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ data_readable[i] = 1;
+ metadata_readable[i] = 1;
+ }
+ if (AFR_IS_ARBITER_BRICK(priv, ARBITER_BRICK_INDEX)) {
+ data_readable[ARBITER_BRICK_INDEX] = 0;
+ metadata_readable[ARBITER_BRICK_INDEX] = 0;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (replies) { /* Lookup */
+ if (!replies[i].valid || replies[i].op_ret == -1 ||
+ (replies[i].xdata &&
+ dict_get_sizen(replies[i].xdata, GLUSTERFS_BAD_INODE))) {
+ data_readable[i] = 0;
+ metadata_readable[i] = 0;
+ continue;
+ }
- for (i = 0; i < priv->child_count; i++) {
- data_readable[i] = 1;
- metadata_readable[i] = 1;
- }
- if (AFR_IS_ARBITER_BRICK (priv, ARBITER_BRICK_INDEX)) {
- data_readable[ARBITER_BRICK_INDEX] = 0;
- metadata_readable[ARBITER_BRICK_INDEX] = 0;
+ xdata = replies[i].xdata;
+ ia_type = replies[i].poststat.ia_type;
+ } else { /* pre-op xattrop */
+ xdata = local->transaction.changelog_xdata[i];
+ ia_type = inode->ia_type;
}
- for (i = 0; i < priv->child_count; i++) {
- if (replies) {/* Lookup */
- if (!replies[i].valid || replies[i].op_ret == -1 ||
- (replies[i].xdata && dict_get (replies[i].xdata,
- GLUSTERFS_BAD_INODE))) {
- data_readable[i] = 0;
- metadata_readable[i] = 0;
- continue;
- }
-
- xdata = replies[i].xdata;
- ia_type = replies[i].poststat.ia_type;
- } else {/* pre-op xattrop */
- xdata = local->transaction.changelog_xdata[i];
- ia_type = inode->ia_type;
- }
+ if (!xdata)
+ continue; /* mkdir_cbk sends NULL xdata_rsp. */
+ afr_accused_fill(this, xdata, data_accused,
+ (ia_type == IA_IFDIR) ? AFR_ENTRY_TRANSACTION
+ : AFR_DATA_TRANSACTION);
- afr_accused_fill (this, xdata, data_accused,
- (ia_type == IA_IFDIR) ?
- AFR_ENTRY_TRANSACTION : AFR_DATA_TRANSACTION);
+ afr_accused_fill(this, xdata, metadata_accused,
+ AFR_METADATA_TRANSACTION);
+ }
- afr_accused_fill (this, xdata,
- metadata_accused, AFR_METADATA_TRANSACTION);
- }
+ if (replies && ia_type != IA_INVAL && ia_type != IA_IFDIR &&
+ /* We want to accuse small files only when we know for
+ * sure that there is no IO happening. Otherwise, the
+ * ia_sizes obtained in post-refresh replies may
+ * mismatch due to a race between inode-refresh and
+ * ongoing writes, causing spurious heal launches*/
+ !afr_is_possibly_under_txn(AFR_DATA_TRANSACTION, local, this)) {
+ afr_accuse_smallfiles(this, replies, data_accused);
+ }
- if (replies && ia_type != IA_INVAL && ia_type != IA_IFDIR &&
- /* We want to accuse small files only when we know for
- * sure that there is no IO happening. Otherwise, the
- * ia_sizes obtained in post-refresh replies may
- * mismatch due to a race between inode-refresh and
- * ongoing writes, causing spurious heal launches*/
- !afr_is_possibly_under_txn (AFR_DATA_TRANSACTION, local, this)) {
- afr_accuse_smallfiles (this, replies, data_accused);
+ for (i = 0; i < priv->child_count; i++) {
+ if (data_accused[i]) {
+ data_readable[i] = 0;
+ ret = 1;
}
-
- for (i = 0; i < priv->child_count; i++) {
- if (data_accused[i]) {
- data_readable[i] = 0;
- ret = 1;
- }
- if (metadata_accused[i]) {
- metadata_readable[i] = 0;
- ret = 1;
- }
+ if (metadata_accused[i]) {
+ metadata_readable[i] = 0;
+ ret = 1;
}
- return ret;
+ }
+ return ret;
}
int
-afr_replies_interpret (call_frame_t *frame, xlator_t *this, inode_t *inode,
- gf_boolean_t *start_heal)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- struct afr_reply *replies = NULL;
- int event_generation = 0;
- int i = 0;
- unsigned char *data_accused = NULL;
- unsigned char *metadata_accused = NULL;
- unsigned char *data_readable = NULL;
- unsigned char *metadata_readable = NULL;
- int ret = 0;
-
- local = frame->local;
- priv = this->private;
- replies = local->replies;
- event_generation = local->event_generation;
-
- data_accused = alloca0 (priv->child_count);
- data_readable = alloca0 (priv->child_count);
- metadata_accused = alloca0 (priv->child_count);
- metadata_readable = alloca0 (priv->child_count);
-
- ret = afr_readables_fill (frame, this, inode, data_accused,
- metadata_accused, data_readable,
- metadata_readable, replies);
-
- for (i = 0; i < priv->child_count; i++) {
- if (start_heal && priv->child_up[i] &&
- (data_accused[i] || metadata_accused[i])) {
- *start_heal = _gf_true;
- break;
- }
- }
- afr_inode_read_subvol_set (inode, this, data_readable,
- metadata_readable, event_generation);
- return ret;
+afr_replies_interpret(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ gf_boolean_t *start_heal)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ struct afr_reply *replies = NULL;
+ int event_generation = 0;
+ int i = 0;
+ unsigned char *data_accused = NULL;
+ unsigned char *metadata_accused = NULL;
+ unsigned char *data_readable = NULL;
+ unsigned char *metadata_readable = NULL;
+ int ret = 0;
+
+ local = frame->local;
+ priv = this->private;
+ replies = local->replies;
+ event_generation = local->event_generation;
+
+ data_accused = alloca0(priv->child_count);
+ data_readable = alloca0(priv->child_count);
+ metadata_accused = alloca0(priv->child_count);
+ metadata_readable = alloca0(priv->child_count);
+
+ ret = afr_readables_fill(frame, this, inode, data_accused, metadata_accused,
+ data_readable, metadata_readable, replies);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (start_heal && priv->child_up[i] &&
+ (data_accused[i] || metadata_accused[i])) {
+ *start_heal = _gf_true;
+ break;
+ }
+ }
+ afr_inode_read_subvol_set(inode, this, data_readable, metadata_readable,
+ event_generation);
+ return ret;
}
int
-afr_refresh_selfheal_done (int ret, call_frame_t *heal, void *opaque)
+afr_refresh_selfheal_done(int ret, call_frame_t *heal, void *opaque)
{
- if (heal)
- AFR_STACK_DESTROY (heal);
- return 0;
+ if (heal)
+ AFR_STACK_DESTROY(heal);
+ return 0;
}
int
-afr_inode_refresh_err (call_frame_t *frame, xlator_t *this)
+afr_inode_refresh_err(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- int err = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int err = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (local->replies[i].valid && !local->replies[i].op_ret) {
- err = 0;
- goto ret;
- }
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->replies[i].valid && !local->replies[i].op_ret) {
+ err = 0;
+ goto ret;
+ }
+ }
- err = afr_final_errno (local, priv);
+ err = afr_final_errno(local, priv);
ret:
- return -err;
+ return err;
}
gf_boolean_t
-afr_selfheal_enabled (xlator_t *this)
+afr_selfheal_enabled(const xlator_t *this)
{
- afr_private_t *priv = NULL;
- gf_boolean_t data = _gf_false;
- int ret = 0;
+ const afr_private_t *priv = this->private;
- priv = this->private;
-
- ret = gf_string2boolean (priv->data_self_heal, &data);
- GF_ASSERT (!ret);
-
- return data || priv->metadata_self_heal || priv->entry_self_heal;
+ return priv->data_self_heal || priv->metadata_self_heal ||
+ priv->entry_self_heal;
}
-
int
-afr_txn_refresh_done (call_frame_t *frame, xlator_t *this, int err)
+afr_txn_refresh_done(call_frame_t *frame, xlator_t *this, int err)
{
+ call_frame_t *heal_frame = NULL;
+ afr_local_t *heal_local = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ inode_t *inode = NULL;
+ int event_generation = 0;
+ int read_subvol = -1;
+ int ret = 0;
- call_frame_t *heal_frame = NULL;
- afr_local_t *heal_local = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- inode_t *inode = NULL;
- int event_generation = 0;
- int read_subvol = -1;
- int op_errno = ENOMEM;
- int ret = 0;
+ local = frame->local;
+ inode = local->inode;
+ priv = this->private;
- local = frame->local;
- inode = local->inode;
- priv = this->private;
-
- if (err)
- goto refresh_done;
+ if (err)
+ goto refresh_done;
- if (local->op == GF_FOP_LOOKUP)
- goto refresh_done;
+ if (local->op == GF_FOP_LOOKUP)
+ goto refresh_done;
- ret = afr_inode_get_readable (frame, inode, this, local->readable,
- &event_generation,
- local->transaction.type);
+ ret = afr_inode_get_readable(frame, inode, this, local->readable,
+ &event_generation, local->transaction.type);
- if (ret == -EIO || (local->is_read_txn && !event_generation)) {
- /* No readable subvolume even after refresh ==> splitbrain.*/
- if (!priv->fav_child_policy) {
- err = -EIO;
- goto refresh_done;
- }
- read_subvol = afr_sh_get_fav_by_policy (this, local->replies,
- inode, NULL);
- if (read_subvol == -1) {
- err = -EIO;
- goto refresh_done;
- }
+ if (ret == -EIO) {
+ /* No readable subvolume even after refresh ==> splitbrain.*/
+ if (!priv->fav_child_policy) {
+ err = EIO;
+ goto refresh_done;
+ }
+ read_subvol = afr_sh_get_fav_by_policy(this, local->replies, inode,
+ NULL);
+ if (read_subvol == -1) {
+ err = EIO;
+ goto refresh_done;
+ }
- heal_frame = copy_frame (frame);
- if (!heal_frame) {
- err = -EIO;
- goto refresh_done;
- }
- heal_frame->root->pid = GF_CLIENT_PID_SELF_HEALD;
- heal_local = AFR_FRAME_INIT (heal_frame, op_errno);
- if (!heal_local) {
- err = -EIO;
- AFR_STACK_DESTROY (heal_frame);
- goto refresh_done;
- }
- heal_local->xdata_req = dict_new();
- if (!heal_local->xdata_req) {
- err = -EIO;
- AFR_STACK_DESTROY (heal_frame);
- goto refresh_done;
- }
- heal_local->heal_frame = frame;
- ret = synctask_new (this->ctx->env,
- afr_fav_child_reset_sink_xattrs,
- afr_fav_child_reset_sink_xattrs_cbk,
- heal_frame,
- heal_frame);
- return 0;
+ heal_frame = afr_frame_create(this, NULL);
+ if (!heal_frame) {
+ err = EIO;
+ goto refresh_done;
}
+ heal_local = heal_frame->local;
+ heal_local->xdata_req = dict_new();
+ if (!heal_local->xdata_req) {
+ err = EIO;
+ AFR_STACK_DESTROY(heal_frame);
+ goto refresh_done;
+ }
+ heal_local->heal_frame = frame;
+ ret = synctask_new(this->ctx->env, afr_fav_child_reset_sink_xattrs,
+ afr_fav_child_reset_sink_xattrs_cbk, heal_frame,
+ heal_frame);
+ return 0;
+ }
refresh_done:
- afr_local_replies_wipe (local, this->private);
- local->refreshfn (frame, this, err);
+ afr_local_replies_wipe(local, this->private);
+ local->refreshfn(frame, this, err);
- return 0;
+ return 0;
}
int
-afr_inode_refresh_done (call_frame_t *frame, xlator_t *this, int error)
-{
- call_frame_t *heal_frame = NULL;
- afr_local_t *local = NULL;
- gf_boolean_t start_heal = _gf_false;
- afr_local_t *heal_local = NULL;
- int op_errno = ENOMEM;
- int ret = 0;
- int err = 0;
-
- if (error != 0) {
- err = error;
- goto refresh_done;
- }
-
- local = frame->local;
-
- ret = afr_replies_interpret (frame, this, local->refreshinode,
- &start_heal);
-
- err = afr_inode_refresh_err (frame, this);
-
- if (ret && afr_selfheal_enabled (this) && start_heal) {
- heal_frame = copy_frame (frame);
- if (!heal_frame)
- goto refresh_done;
- heal_frame->root->pid = GF_CLIENT_PID_SELF_HEALD;
- heal_local = AFR_FRAME_INIT (heal_frame, op_errno);
- if (!heal_local) {
- AFR_STACK_DESTROY (heal_frame);
- goto refresh_done;
- }
- heal_local->refreshinode = inode_ref (local->refreshinode);
- heal_local->heal_frame = heal_frame;
- if (!afr_throttled_selfheal (heal_frame, this)) {
- AFR_STACK_DESTROY (heal_frame);
- goto refresh_done;
- }
+afr_inode_refresh_done(call_frame_t *frame, xlator_t *this, int error)
+{
+ call_frame_t *heal_frame = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ gf_boolean_t start_heal = _gf_false;
+ afr_local_t *heal_local = NULL;
+ unsigned char *success_replies = NULL;
+ int ret = 0;
+
+ if (error != 0) {
+ goto refresh_done;
+ }
+
+ local = frame->local;
+ priv = this->private;
+ success_replies = alloca0(priv->child_count);
+ afr_fill_success_replies(local, priv, success_replies);
+
+ if (priv->thin_arbiter_count && local->is_read_txn &&
+ AFR_COUNT(success_replies, priv->child_count) != priv->child_count) {
+ /* We need to query the good bricks and/or thin-arbiter.*/
+ if (success_replies[0]) {
+ local->read_txn_query_child = AFR_CHILD_ZERO;
+ } else if (success_replies[1]) {
+ local->read_txn_query_child = AFR_CHILD_ONE;
+ }
+ error = EINVAL;
+ goto refresh_done;
+ }
+
+ if (!afr_has_quorum(success_replies, this, frame)) {
+ error = afr_final_errno(frame->local, this->private);
+ if (!error)
+ error = afr_quorum_errno(priv);
+ goto refresh_done;
+ }
+
+ ret = afr_replies_interpret(frame, this, local->refreshinode, &start_heal);
+
+ if (ret && afr_selfheal_enabled(this) && start_heal) {
+ heal_frame = afr_frame_create(this, NULL);
+ if (!heal_frame)
+ goto refresh_done;
+ heal_local = heal_frame->local;
+ heal_local->refreshinode = inode_ref(local->refreshinode);
+ heal_local->heal_frame = heal_frame;
+ if (!afr_throttled_selfheal(heal_frame, this)) {
+ AFR_STACK_DESTROY(heal_frame);
+ goto refresh_done;
}
+ }
refresh_done:
- afr_txn_refresh_done (frame, this, err);
+ afr_txn_refresh_done(frame, this, error);
- return 0;
+ return 0;
}
void
-afr_inode_refresh_subvol_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *buf,
- dict_t *xdata, struct iatt *par)
-{
- afr_local_t *local = NULL;
- int call_child = (long) cookie;
- int8_t need_heal = 1;
- int call_count = 0;
- GF_UNUSED int ret = 0;
+afr_inode_refresh_subvol_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *buf,
+ dict_t *xdata, struct iatt *par)
+{
+ afr_local_t *local = NULL;
+ int call_child = (long)cookie;
+ int8_t need_heal = 1;
+ int call_count = 0;
+ int ret = 0;
+
+ local = frame->local;
+ local->replies[call_child].valid = 1;
+ local->replies[call_child].op_ret = op_ret;
+ local->replies[call_child].op_errno = op_errno;
+ if (op_ret != -1) {
+ local->replies[call_child].poststat = *buf;
+ if (par)
+ local->replies[call_child].postparent = *par;
+ if (xdata)
+ local->replies[call_child].xdata = dict_ref(xdata);
+ }
- local = frame->local;
- local->replies[call_child].valid = 1;
- local->replies[call_child].op_ret = op_ret;
- local->replies[call_child].op_errno = op_errno;
- if (op_ret != -1) {
- local->replies[call_child].poststat = *buf;
- if (par)
- local->replies[call_child].postparent = *par;
- if (xdata)
- local->replies[call_child].xdata = dict_ref (xdata);
- }
- if (xdata) {
- ret = dict_get_int8 (xdata, "link-count", &need_heal);
- local->replies[call_child].need_heal = need_heal;
- } else {
- local->replies[call_child].need_heal = need_heal;
+ if (xdata) {
+ ret = dict_get_int8(xdata, "link-count", &need_heal);
+ if (ret) {
+ gf_msg_debug(this->name, -ret, "Unable to get link count");
}
+ }
- call_count = afr_frame_return (frame);
- if (call_count == 0) {
- afr_set_need_heal (this, local);
- afr_inode_refresh_done (frame, this, 0);
+ local->replies[call_child].need_heal = need_heal;
+ call_count = afr_frame_return(frame);
+ if (call_count == 0) {
+ afr_set_need_heal(this, local);
+ ret = afr_inode_refresh_err(frame, this);
+ if (ret) {
+ gf_msg_debug(this->name, ret, "afr_inode_refresh_err failed");
}
-
+ afr_inode_refresh_done(frame, this, ret);
+ }
}
int
-afr_inode_refresh_subvol_with_lookup_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret,
- int op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata,
- struct iatt *par)
+afr_inode_refresh_subvol_with_lookup_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int op_ret,
+ int op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata,
+ struct iatt *par)
{
- afr_inode_refresh_subvol_cbk (frame, cookie, this, op_ret, op_errno,
- buf, xdata, par);
- return 0;
+ afr_inode_refresh_subvol_cbk(frame, cookie, this, op_ret, op_errno, buf,
+ xdata, par);
+ return 0;
}
-
int
-afr_inode_refresh_subvol_with_lookup (call_frame_t *frame, xlator_t *this,
- int i, inode_t *inode, uuid_t gfid,
- dict_t *xdata)
+afr_inode_refresh_subvol_with_lookup(call_frame_t *frame, xlator_t *this, int i,
+ inode_t *inode, uuid_t gfid, dict_t *xdata)
{
- loc_t loc = {0, };
- afr_private_t *priv = NULL;
+ loc_t loc = {
+ 0,
+ };
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- loc.inode = inode;
- if (gf_uuid_is_null (inode->gfid) && gfid) {
- /* To handle setattr/setxattr on yet to be linked inode from
- * dht */
- gf_uuid_copy (loc.gfid, gfid);
- } else {
- gf_uuid_copy (loc.gfid, inode->gfid);
- }
+ loc.inode = inode;
+ if (gf_uuid_is_null(inode->gfid) && gfid) {
+ /* To handle setattr/setxattr on yet to be linked inode from
+ * dht */
+ gf_uuid_copy(loc.gfid, gfid);
+ } else {
+ gf_uuid_copy(loc.gfid, inode->gfid);
+ }
- STACK_WIND_COOKIE (frame, afr_inode_refresh_subvol_with_lookup_cbk,
- (void *) (long) i, priv->children[i],
- priv->children[i]->fops->lookup, &loc, xdata);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_inode_refresh_subvol_with_lookup_cbk,
+ (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->lookup, &loc, xdata);
+ return 0;
}
int
-afr_inode_refresh_subvol_with_fstat_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *buf, dict_t *xdata)
+afr_inode_refresh_subvol_with_fstat_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
{
- afr_inode_refresh_subvol_cbk (frame, cookie, this, op_ret, op_errno,
- buf, xdata, NULL);
- return 0;
+ afr_inode_refresh_subvol_cbk(frame, cookie, this, op_ret, op_errno, buf,
+ xdata, NULL);
+ return 0;
}
int
-afr_inode_refresh_subvol_with_fstat (call_frame_t *frame, xlator_t *this, int i,
- dict_t *xdata)
+afr_inode_refresh_subvol_with_fstat(call_frame_t *frame, xlator_t *this, int i,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- STACK_WIND_COOKIE (frame, afr_inode_refresh_subvol_with_fstat_cbk,
- (void *) (long) i, priv->children[i],
- priv->children[i]->fops->fstat, local->fd, xdata);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_inode_refresh_subvol_with_fstat_cbk,
+ (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->fstat, local->fd, xdata);
+ return 0;
}
int
-afr_inode_refresh_do (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
- int i = 0;
- int ret = 0;
- dict_t *xdata = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
- unsigned char *wind_subvols = NULL;
-
- priv = this->private;
- local = frame->local;
- wind_subvols = alloca0 (priv->child_count);
-
- afr_local_replies_wipe (local, priv);
-
- if (local->fd) {
- fd_ctx = afr_fd_ctx_get (local->fd, this);
- if (!fd_ctx) {
- afr_inode_refresh_done (frame, this, EINVAL);
- return 0;
- }
- }
+afr_inode_refresh_do(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int i = 0;
+ int ret = 0;
+ dict_t *xdata = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ unsigned char *wind_subvols = NULL;
- xdata = dict_new ();
- if (!xdata) {
- afr_inode_refresh_done (frame, this, ENOMEM);
- return 0;
- }
+ priv = this->private;
+ local = frame->local;
+ wind_subvols = alloca0(priv->child_count);
- ret = afr_xattr_req_prepare (this, xdata);
- if (ret != 0) {
- dict_unref (xdata);
- afr_inode_refresh_done (frame, this, -ret);
- return 0;
- }
+ afr_local_replies_wipe(local, priv);
- ret = dict_set_str (xdata, "link-count", GF_XATTROP_INDEX_COUNT);
- if (ret) {
- gf_msg_debug (this->name, -ret,
- "Unable to set link-count in dict ");
+ if (local->fd) {
+ fd_ctx = afr_fd_ctx_get(local->fd, this);
+ if (!fd_ctx) {
+ afr_inode_refresh_done(frame, this, EINVAL);
+ return 0;
}
+ }
- ret = dict_set_str (xdata, GLUSTERFS_INODELK_DOM_COUNT, this->name);
- if (ret) {
- gf_msg_debug (this->name, -ret,
- "Unable to set inodelk-dom-count in dict ");
+ xdata = dict_new();
+ if (!xdata) {
+ afr_inode_refresh_done(frame, this, ENOMEM);
+ return 0;
+ }
- }
+ ret = afr_xattr_req_prepare(this, xdata);
+ if (ret != 0) {
+ dict_unref(xdata);
+ afr_inode_refresh_done(frame, this, -ret);
+ return 0;
+ }
- if (local->fd) {
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i] &&
- fd_ctx->opened_on[i] == AFR_FD_OPENED)
- wind_subvols[i] = 1;
- }
- } else {
- memcpy (wind_subvols, local->child_up,
- sizeof (*local->child_up) * priv->child_count);
- }
+ ret = dict_set_sizen_str_sizen(xdata, "link-count", GF_XATTROP_INDEX_COUNT);
+ if (ret) {
+ gf_msg_debug(this->name, -ret, "Unable to set link-count in dict ");
+ }
- local->call_count = AFR_COUNT (wind_subvols, priv->child_count);
+ ret = dict_set_str_sizen(xdata, GLUSTERFS_INODELK_DOM_COUNT, this->name);
+ if (ret) {
+ gf_msg_debug(this->name, -ret,
+ "Unable to set inodelk-dom-count in dict ");
+ }
- call_count = local->call_count;
- if (!call_count) {
- dict_unref (xdata);
- if (local->fd && AFR_COUNT(local->child_up, priv->child_count))
- afr_inode_refresh_done (frame, this, EBADFD);
- else
- afr_inode_refresh_done (frame, this, ENOTCONN);
- return 0;
+ if (local->fd) {
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i] && fd_ctx->opened_on[i] == AFR_FD_OPENED)
+ wind_subvols[i] = 1;
}
- for (i = 0; i < priv->child_count; i++) {
- if (!wind_subvols[i])
- continue;
+ } else {
+ memcpy(wind_subvols, local->child_up,
+ sizeof(*local->child_up) * priv->child_count);
+ }
- if (local->fd)
- afr_inode_refresh_subvol_with_fstat (frame, this, i,
- xdata);
- else
- afr_inode_refresh_subvol_with_lookup (frame, this, i,
- local->refreshinode,
- local->refreshgfid, xdata);
+ local->call_count = AFR_COUNT(wind_subvols, priv->child_count);
- if (!--call_count)
- break;
- }
+ call_count = local->call_count;
+ if (!call_count) {
+ dict_unref(xdata);
+ if (local->fd && AFR_COUNT(local->child_up, priv->child_count))
+ afr_inode_refresh_done(frame, this, EBADFD);
+ else
+ afr_inode_refresh_done(frame, this, ENOTCONN);
+ return 0;
+ }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!wind_subvols[i])
+ continue;
- dict_unref (xdata);
+ if (local->fd)
+ afr_inode_refresh_subvol_with_fstat(frame, this, i, xdata);
+ else
+ afr_inode_refresh_subvol_with_lookup(
+ frame, this, i, local->refreshinode, local->refreshgfid, xdata);
- return 0;
-}
+ if (!--call_count)
+ break;
+ }
+ dict_unref(xdata);
+
+ return 0;
+}
int
-afr_inode_refresh (call_frame_t *frame, xlator_t *this, inode_t *inode,
- uuid_t gfid, afr_inode_refresh_cbk_t refreshfn)
+afr_inode_refresh(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ uuid_t gfid, afr_inode_refresh_cbk_t refreshfn)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- local->refreshfn = refreshfn;
+ local->refreshfn = refreshfn;
- if (local->refreshinode) {
- inode_unref (local->refreshinode);
- local->refreshinode = NULL;
- }
+ if (local->refreshinode) {
+ inode_unref(local->refreshinode);
+ local->refreshinode = NULL;
+ }
- local->refreshinode = inode_ref (inode);
+ local->refreshinode = inode_ref(inode);
- if (gfid)
- gf_uuid_copy (local->refreshgfid, gfid);
- else
- gf_uuid_clear (local->refreshgfid);
+ if (gfid)
+ gf_uuid_copy(local->refreshgfid, gfid);
+ else
+ gf_uuid_clear(local->refreshgfid);
- afr_inode_refresh_do (frame, this);
+ afr_inode_refresh_do(frame, this);
- return 0;
+ return 0;
}
-
int
-afr_xattr_req_prepare (xlator_t *this, dict_t *xattr_req)
+afr_xattr_req_prepare(xlator_t *this, dict_t *xattr_req)
{
- int i = 0;
- afr_private_t *priv = NULL;
- int ret = 0;
-
- priv = this->private;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ int ret = 0;
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_set_uint64 (xattr_req, priv->pending_key[i],
- AFR_NUM_CHANGE_LOGS * sizeof(int));
- if (ret < 0)
- gf_msg (this->name, GF_LOG_WARNING,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Unable to set dict value for %s",
- priv->pending_key[i]);
- /* 3 = data+metadata+entry */
- }
- ret = dict_set_uint64 (xattr_req, AFR_DIRTY,
- AFR_NUM_CHANGE_LOGS * sizeof(int));
- if (ret) {
- gf_msg_debug (this->name, -ret, "failed to set dirty "
- "query flag");
- }
+ priv = this->private;
- ret = dict_set_int32 (xattr_req, "list-xattr", 1);
- if (ret) {
- gf_msg_debug (this->name, -ret,
- "Unable to set list-xattr in dict ");
- }
+ for (i = 0; i < priv->child_count; i++) {
+ ret = dict_set_uint64(xattr_req, priv->pending_key[i],
+ AFR_NUM_CHANGE_LOGS * sizeof(int));
+ if (ret < 0)
+ gf_msg(this->name, GF_LOG_WARNING, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Unable to set dict value for %s", priv->pending_key[i]);
+ /* 3 = data+metadata+entry */
+ }
+ ret = dict_set_uint64(xattr_req, AFR_DIRTY,
+ AFR_NUM_CHANGE_LOGS * sizeof(int));
+ if (ret) {
+ gf_msg_debug(this->name, -ret,
+ "failed to set dirty "
+ "query flag");
+ }
+
+ ret = dict_set_int32_sizen(xattr_req, "list-xattr", 1);
+ if (ret) {
+ gf_msg_debug(this->name, -ret, "Unable to set list-xattr in dict ");
+ }
+
+ return ret;
+}
- return ret;
+int
+afr_lookup_xattr_req_prepare(afr_local_t *local, xlator_t *this,
+ dict_t *xattr_req, loc_t *loc)
+{
+ int ret = -ENOMEM;
+
+ if (!local->xattr_req)
+ local->xattr_req = dict_new();
+
+ if (!local->xattr_req)
+ goto out;
+
+ if (xattr_req && (xattr_req != local->xattr_req))
+ dict_copy(xattr_req, local->xattr_req);
+
+ ret = afr_xattr_req_prepare(this, local->xattr_req);
+
+ ret = dict_set_uint64(local->xattr_req, GLUSTERFS_INODELK_COUNT, 0);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, AFR_MSG_DICT_SET_FAILED,
+ "%s: Unable to set dict value for %s", loc->path,
+ GLUSTERFS_INODELK_COUNT);
+ }
+ ret = dict_set_uint64(local->xattr_req, GLUSTERFS_ENTRYLK_COUNT, 0);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, AFR_MSG_DICT_SET_FAILED,
+ "%s: Unable to set dict value for %s", loc->path,
+ GLUSTERFS_ENTRYLK_COUNT);
+ }
+
+ ret = dict_set_uint32(local->xattr_req, GLUSTERFS_PARENT_ENTRYLK, 0);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, AFR_MSG_DICT_SET_FAILED,
+ "%s: Unable to set dict value for %s", loc->path,
+ GLUSTERFS_PARENT_ENTRYLK);
+ }
+
+ ret = dict_set_sizen_str_sizen(local->xattr_req, "link-count",
+ GF_XATTROP_INDEX_COUNT);
+ if (ret) {
+ gf_msg_debug(this->name, -ret, "Unable to set link-count in dict ");
+ }
+
+ ret = 0;
+out:
+ return ret;
}
int
-afr_lookup_xattr_req_prepare (afr_local_t *local, xlator_t *this,
- dict_t *xattr_req, loc_t *loc)
+afr_least_pending_reads_child(afr_private_t *priv, unsigned char *readable)
{
- int ret = -ENOMEM;
+ int i = 0;
+ int child = -1;
+ int64_t read_iter = -1;
+ int64_t pending_read = -1;
- if (!local->xattr_req)
- local->xattr_req = dict_new ();
+ for (i = 0; i < priv->child_count; i++) {
+ if (AFR_IS_ARBITER_BRICK(priv, i) || !readable[i])
+ continue;
+ read_iter = GF_ATOMIC_GET(priv->pending_reads[i]);
+ if (child == -1 || read_iter < pending_read) {
+ pending_read = read_iter;
+ child = i;
+ }
+ }
- if (!local->xattr_req)
- goto out;
+ return child;
+}
- if (xattr_req && (xattr_req != local->xattr_req))
- dict_copy (xattr_req, local->xattr_req);
+static int32_t
+afr_least_latency_child(afr_private_t *priv, unsigned char *readable)
+{
+ int32_t i = 0;
+ int child = -1;
- ret = afr_xattr_req_prepare (this, local->xattr_req);
+ for (i = 0; i < priv->child_count; i++) {
+ if (AFR_IS_ARBITER_BRICK(priv, i) || !readable[i] ||
+ priv->child_latency[i] < 0)
+ continue;
- ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_INODELK_COUNT, 0);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "%s: Unable to set dict value for %s",
- loc->path, GLUSTERFS_INODELK_COUNT);
- }
- ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_ENTRYLK_COUNT, 0);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "%s: Unable to set dict value for %s",
- loc->path, GLUSTERFS_ENTRYLK_COUNT);
+ if (child == -1 ||
+ priv->child_latency[i] < priv->child_latency[child]) {
+ child = i;
}
+ }
+ return child;
+}
- ret = dict_set_uint32 (local->xattr_req, GLUSTERFS_PARENT_ENTRYLK, 0);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "%s: Unable to set dict value for %s",
- loc->path, GLUSTERFS_PARENT_ENTRYLK);
- }
+static int32_t
+afr_least_latency_times_pending_reads_child(afr_private_t *priv,
+ unsigned char *readable)
+{
+ int32_t i = 0;
+ int child = -1;
+ int64_t pending_read = 0;
+ int64_t latency = -1;
+ int64_t least_latency = -1;
- ret = dict_set_str (local->xattr_req, "link-count",
- GF_XATTROP_INDEX_COUNT);
- if (ret) {
- gf_msg_debug (this->name, -ret,
- "Unable to set link-count in dict ");
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (AFR_IS_ARBITER_BRICK(priv, i) || !readable[i] ||
+ priv->child_latency[i] < 0)
+ continue;
- ret = 0;
-out:
- return ret;
-}
+ pending_read = GF_ATOMIC_GET(priv->pending_reads[i]);
+ latency = (pending_read + 1) * priv->child_latency[i];
-int
-afr_least_pending_reads_child (afr_private_t *priv)
-{
- int i = 0;
- int child = 0;
- int64_t read_iter = -1;
- int64_t pending_read = -1;
-
- pending_read = GF_ATOMIC_GET (priv->pending_reads[0]);
- for (i = 1; i < priv->child_count; i++) {
- if (AFR_IS_ARBITER_BRICK(priv, i))
- continue;
- read_iter = GF_ATOMIC_GET(priv->pending_reads[i]);
- if (read_iter < pending_read) {
- pending_read = read_iter;
- child = i;
- }
+ if (child == -1 || latency < least_latency) {
+ least_latency = latency;
+ child = i;
}
-
- return child;
+ }
+ return child;
}
int
-afr_hash_child (afr_read_subvol_args_t *args, afr_private_t *priv)
-{
- uuid_t gfid_copy = {0,};
- pid_t pid;
- int child = -1;
-
- switch (priv->hash_mode) {
- case 0:
- break;
- case 1:
- gf_uuid_copy (gfid_copy, args->gfid);
- child = SuperFastHash((char *)gfid_copy,
- sizeof(gfid_copy)) % priv->child_count;
- break;
- case 2:
- if (args->ia_type != IA_IFDIR) {
- /*
- * Why getpid? Because it's one of the cheapest calls
- * available - faster than gethostname etc. - and
- * returns a constant-length value that's sure to be
- * shorter than a UUID. It's still very unlikely to be
- * the same across clients, so it still provides good
- * mixing. We're not trying for perfection here. All we
- * need is a low probability that multiple clients
- * won't converge on the same subvolume.
- */
- pid = getpid();
- memcpy (gfid_copy, &pid, sizeof(pid));
- }
- child = SuperFastHash((char *)gfid_copy,
- sizeof(gfid_copy)) % priv->child_count;
- break;
- case 3:
- child = afr_least_pending_reads_child (priv);
- break;
- }
-
- return child;
+afr_hash_child(afr_read_subvol_args_t *args, afr_private_t *priv,
+ unsigned char *readable)
+{
+ uuid_t gfid_copy = {
+ 0,
+ };
+ pid_t pid;
+ int child = -1;
+
+ switch (priv->hash_mode) {
+ case AFR_READ_POLICY_FIRST_UP:
+ break;
+ case AFR_READ_POLICY_GFID_HASH:
+ gf_uuid_copy(gfid_copy, args->gfid);
+ child = SuperFastHash((char *)gfid_copy, sizeof(gfid_copy)) %
+ priv->child_count;
+ break;
+ case AFR_READ_POLICY_GFID_PID_HASH:
+ if (args->ia_type != IA_IFDIR) {
+ /*
+ * Why getpid? Because it's one of the cheapest calls
+ * available - faster than gethostname etc. - and
+ * returns a constant-length value that's sure to be
+ * shorter than a UUID. It's still very unlikely to be
+ * the same across clients, so it still provides good
+ * mixing. We're not trying for perfection here. All we
+ * need is a low probability that multiple clients
+ * won't converge on the same subvolume.
+ */
+ gf_uuid_copy(gfid_copy, args->gfid);
+ pid = getpid();
+ *(pid_t *)gfid_copy ^= pid;
+ }
+ child = SuperFastHash((char *)gfid_copy, sizeof(gfid_copy)) %
+ priv->child_count;
+ break;
+ case AFR_READ_POLICY_LESS_LOAD:
+ child = afr_least_pending_reads_child(priv, readable);
+ break;
+ case AFR_READ_POLICY_LEAST_LATENCY:
+ child = afr_least_latency_child(priv, readable);
+ break;
+ case AFR_READ_POLICY_LOAD_LATENCY_HYBRID:
+ child = afr_least_latency_times_pending_reads_child(priv, readable);
+ break;
+ }
+
+ return child;
}
int
-afr_read_subvol_select_by_policy (inode_t *inode, xlator_t *this,
- unsigned char *readable,
- afr_read_subvol_args_t *args)
+afr_read_subvol_select_by_policy(inode_t *inode, xlator_t *this,
+ unsigned char *readable,
+ afr_read_subvol_args_t *args)
{
- int i = 0;
- int read_subvol = -1;
- afr_private_t *priv = NULL;
- afr_read_subvol_args_t local_args = {0,};
+ int i = 0;
+ int read_subvol = -1;
+ afr_private_t *priv = NULL;
+ afr_read_subvol_args_t local_args = {
+ 0,
+ };
- priv = this->private;
+ priv = this->private;
- /* first preference - explicitly specified or local subvolume */
- if (priv->read_child >= 0 && readable[priv->read_child])
- return priv->read_child;
+ /* first preference - explicitly specified or local subvolume */
+ if (priv->read_child >= 0 && readable[priv->read_child])
+ return priv->read_child;
- if (inode_is_linked (inode)) {
- gf_uuid_copy (local_args.gfid, inode->gfid);
- local_args.ia_type = inode->ia_type;
- } else if (args) {
- local_args = *args;
- }
+ if (inode_is_linked(inode)) {
+ gf_uuid_copy(local_args.gfid, inode->gfid);
+ local_args.ia_type = inode->ia_type;
+ } else if (args) {
+ local_args = *args;
+ }
- /* second preference - use hashed mode */
- read_subvol = afr_hash_child (&local_args, priv);
- if (read_subvol >= 0 && readable[read_subvol])
- return read_subvol;
+ /* second preference - use hashed mode */
+ read_subvol = afr_hash_child(&local_args, priv, readable);
+ if (read_subvol >= 0 && readable[read_subvol])
+ return read_subvol;
- for (i = 0; i < priv->child_count; i++) {
- if (readable[i])
- return i;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (readable[i])
+ return i;
+ }
- /* no readable subvolumes, either split brain or all subvols down */
+ /* no readable subvolumes, either split brain or all subvols down */
- return -1;
+ return -1;
}
-
int
-afr_inode_read_subvol_type_get (inode_t *inode, xlator_t *this,
- unsigned char *readable, int *event_p,
- int type)
+afr_inode_read_subvol_type_get(inode_t *inode, xlator_t *this,
+ unsigned char *readable, int *event_p, int type)
{
- int ret = -1;
+ int ret = -1;
- if (type == AFR_METADATA_TRANSACTION)
- ret = afr_inode_read_subvol_get (inode, this, 0, readable,
- event_p);
- else
- ret = afr_inode_read_subvol_get (inode, this, readable, 0,
- event_p);
- return ret;
+ if (type == AFR_METADATA_TRANSACTION)
+ ret = afr_inode_read_subvol_get(inode, this, 0, readable, event_p);
+ else
+ ret = afr_inode_read_subvol_get(inode, this, readable, 0, event_p);
+ return ret;
}
void
-afr_readables_intersect_get (inode_t *inode, xlator_t *this, int *event,
- unsigned char *intersection)
+afr_readables_intersect_get(inode_t *inode, xlator_t *this, int *event,
+ unsigned char *intersection)
{
- afr_private_t *priv = NULL;
- unsigned char *data_readable = NULL;
- unsigned char *metadata_readable = NULL;
- unsigned char *intersect = NULL;
+ afr_private_t *priv = NULL;
+ unsigned char *data_readable = NULL;
+ unsigned char *metadata_readable = NULL;
+ unsigned char *intersect = NULL;
- priv = this->private;
- data_readable = alloca0 (priv->child_count);
- metadata_readable = alloca0 (priv->child_count);
- intersect = alloca0 (priv->child_count);
+ priv = this->private;
+ data_readable = alloca0(priv->child_count);
+ metadata_readable = alloca0(priv->child_count);
+ intersect = alloca0(priv->child_count);
- afr_inode_read_subvol_get (inode, this, data_readable,
- metadata_readable, event);
+ afr_inode_read_subvol_get(inode, this, data_readable, metadata_readable,
+ event);
- AFR_INTERSECT (intersect, data_readable, metadata_readable,
- priv->child_count);
- if (intersection)
- memcpy (intersection, intersect,
- sizeof (*intersection) * priv->child_count);
+ AFR_INTERSECT(intersect, data_readable, metadata_readable,
+ priv->child_count);
+ if (intersection)
+ memcpy(intersection, intersect,
+ sizeof(*intersection) * priv->child_count);
}
int
-afr_read_subvol_get (inode_t *inode, xlator_t *this, int *subvol_p,
- unsigned char *readables,
- int *event_p, afr_transaction_type type,
- afr_read_subvol_args_t *args)
+afr_read_subvol_get(inode_t *inode, xlator_t *this, int *subvol_p,
+ unsigned char *readables, int *event_p,
+ afr_transaction_type type, afr_read_subvol_args_t *args)
{
- afr_private_t *priv = NULL;
- unsigned char *readable = NULL;
- unsigned char *intersection = NULL;
- int subvol = -1;
- int event = 0;
+ afr_private_t *priv = NULL;
+ unsigned char *readable = NULL;
+ unsigned char *intersection = NULL;
+ int subvol = -1;
+ int event = 0;
- priv = this->private;
+ priv = this->private;
- readable = alloca0 (priv->child_count);
- intersection = alloca0 (priv->child_count);
+ readable = alloca0(priv->child_count);
+ intersection = alloca0(priv->child_count);
- afr_inode_read_subvol_type_get (inode, this, readable, &event, type);
+ afr_inode_read_subvol_type_get(inode, this, readable, &event, type);
- afr_readables_intersect_get (inode, this, &event, intersection);
+ afr_readables_intersect_get(inode, this, &event, intersection);
- if (AFR_COUNT (intersection, priv->child_count) > 0)
- subvol = afr_read_subvol_select_by_policy (inode, this,
- intersection, args);
- else
- subvol = afr_read_subvol_select_by_policy (inode, this,
- readable, args);
- if (subvol_p)
- *subvol_p = subvol;
- if (event_p)
- *event_p = event;
- if (readables)
- memcpy (readables, readable,
- sizeof (*readables) * priv->child_count);
- return subvol;
+ if (AFR_COUNT(intersection, priv->child_count) > 0)
+ subvol = afr_read_subvol_select_by_policy(inode, this, intersection,
+ args);
+ else
+ subvol = afr_read_subvol_select_by_policy(inode, this, readable, args);
+ if (subvol_p)
+ *subvol_p = subvol;
+ if (event_p)
+ *event_p = event;
+ if (readables)
+ memcpy(readables, readable, sizeof(*readables) * priv->child_count);
+ return subvol;
}
-
void
-afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this)
+afr_local_transaction_cleanup(afr_local_t *local, xlator_t *this)
{
- afr_private_t *priv = NULL;
- int i = 0;
-
- priv = this->private;
+ afr_private_t *priv = NULL;
+ int i = 0;
- afr_matrix_cleanup (local->pending, priv->child_count);
+ priv = this->private;
- GF_FREE (local->internal_lock.locked_nodes);
+ afr_matrix_cleanup(local->pending, priv->child_count);
- GF_FREE (local->internal_lock.lower_locked_nodes);
+ GF_FREE(local->internal_lock.lower_locked_nodes);
- afr_entry_lockee_cleanup (&local->internal_lock);
+ afr_lockees_cleanup(&local->internal_lock);
- GF_FREE (local->transaction.pre_op);
+ GF_FREE(local->transaction.pre_op);
- GF_FREE (local->transaction.pre_op_sources);
- if (local->transaction.changelog_xdata) {
- for (i = 0; i < priv->child_count; i++) {
- if (!local->transaction.changelog_xdata[i])
- continue;
- dict_unref (local->transaction.changelog_xdata[i]);
- }
- GF_FREE (local->transaction.changelog_xdata);
+ GF_FREE(local->transaction.pre_op_sources);
+ if (local->transaction.changelog_xdata) {
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.changelog_xdata[i])
+ continue;
+ dict_unref(local->transaction.changelog_xdata[i]);
}
+ GF_FREE(local->transaction.changelog_xdata);
+ }
- GF_FREE (local->transaction.failed_subvols);
-
- GF_FREE (local->transaction.basename);
- GF_FREE (local->transaction.new_basename);
+ GF_FREE(local->transaction.failed_subvols);
- loc_wipe (&local->transaction.parent_loc);
- loc_wipe (&local->transaction.new_parent_loc);
+ GF_FREE(local->transaction.basename);
+ GF_FREE(local->transaction.new_basename);
+ loc_wipe(&local->transaction.parent_loc);
+ loc_wipe(&local->transaction.new_parent_loc);
}
void
-afr_reply_wipe (struct afr_reply *reply)
+afr_reply_wipe(struct afr_reply *reply)
{
- if (reply->xdata) {
- dict_unref (reply->xdata);
- reply->xdata = NULL;
- }
+ if (reply->xdata) {
+ dict_unref(reply->xdata);
+ reply->xdata = NULL;
+ }
- if (reply->xattr) {
- dict_unref (reply->xattr);
- reply->xattr = NULL;
- }
+ if (reply->xattr) {
+ dict_unref(reply->xattr);
+ reply->xattr = NULL;
+ }
}
void
-afr_replies_wipe (struct afr_reply *replies, int count)
+afr_replies_wipe(struct afr_reply *replies, int count)
{
- int i = 0;
+ int i = 0;
- for (i = 0; i < count; i++) {
- afr_reply_wipe (&replies[i]);
- }
+ for (i = 0; i < count; i++) {
+ afr_reply_wipe(&replies[i]);
+ }
}
void
-afr_local_replies_wipe (afr_local_t *local, afr_private_t *priv)
+afr_local_replies_wipe(afr_local_t *local, afr_private_t *priv)
{
+ if (!local->replies)
+ return;
- if (!local->replies)
- return;
-
- afr_replies_wipe (local->replies, priv->child_count);
+ afr_replies_wipe(local->replies, priv->child_count);
- memset (local->replies, 0, sizeof(*local->replies) * priv->child_count);
+ memset(local->replies, 0, sizeof(*local->replies) * priv->child_count);
}
static gf_boolean_t
-afr_fop_lock_is_unlock (call_frame_t *frame)
+afr_fop_lock_is_unlock(call_frame_t *frame)
{
- afr_local_t *local = frame->local;
- switch (local->op) {
+ afr_local_t *local = frame->local;
+ switch (local->op) {
case GF_FOP_INODELK:
case GF_FOP_FINODELK:
- if ((F_UNLCK == local->cont.inodelk.in_flock.l_type) &&
- (local->cont.inodelk.in_cmd == F_SETLKW ||
- local->cont.inodelk.in_cmd == F_SETLK))
- return _gf_true;
- break;
+ if ((F_UNLCK == local->cont.inodelk.in_flock.l_type) &&
+ (local->cont.inodelk.in_cmd == F_SETLKW ||
+ local->cont.inodelk.in_cmd == F_SETLK))
+ return _gf_true;
+ break;
case GF_FOP_ENTRYLK:
case GF_FOP_FENTRYLK:
- if (ENTRYLK_UNLOCK == local->cont.entrylk.in_cmd)
- return _gf_true;
- break;
+ if (ENTRYLK_UNLOCK == local->cont.entrylk.in_cmd)
+ return _gf_true;
+ break;
default:
- return _gf_false;
- }
- return _gf_false;
+ return _gf_false;
+ }
+ return _gf_false;
}
static gf_boolean_t
-afr_lk_is_unlock (int32_t cmd, struct gf_flock *flock)
+afr_lk_is_unlock(int32_t cmd, struct gf_flock *flock)
{
- switch (cmd) {
+ switch (cmd) {
case F_RESLK_UNLCK:
- return _gf_true;
- break;
+ return _gf_true;
+ break;
#if F_SETLKW != F_SETLKW64
case F_SETLKW64:
@@ -1924,575 +2548,573 @@ afr_lk_is_unlock (int32_t cmd, struct gf_flock *flock)
case F_SETLK64:
#endif
case F_SETLK:
- if (F_UNLCK == flock->l_type)
- return _gf_true;
- break;
+ if (F_UNLCK == flock->l_type)
+ return _gf_true;
+ break;
default:
- return _gf_false;
- }
- return _gf_false;
+ return _gf_false;
+ }
+ return _gf_false;
}
void
-afr_handle_inconsistent_fop (call_frame_t *frame, int32_t *op_ret,
- int32_t *op_errno)
+afr_handle_inconsistent_fop(call_frame_t *frame, int32_t *op_ret,
+ int32_t *op_errno)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
- if (!frame || !frame->this || !frame->local || !frame->this->private)
- return;
+ if (!frame || !frame->this || !frame->local || !frame->this->private)
+ return;
- if (*op_ret < 0)
- return;
+ if (*op_ret < 0)
+ return;
- /* Failing inodelk/entrylk/lk here is not a good idea because we
- * need to cleanup the locks on the other bricks if we choose to fail
- * the fop here. The brick may go down just after unwind happens as well
- * so anyways the fop will fail when the next fop is sent so leaving
- * it like this for now.*/
- local = frame->local;
- switch (local->op) {
+ /* Failing inodelk/entrylk/lk here is not a good idea because we
+ * need to cleanup the locks on the other bricks if we choose to fail
+ * the fop here. The brick may go down just after unwind happens as well
+ * so anyways the fop will fail when the next fop is sent so leaving
+ * it like this for now.*/
+ local = frame->local;
+ switch (local->op) {
case GF_FOP_LOOKUP:
case GF_FOP_INODELK:
case GF_FOP_FINODELK:
case GF_FOP_ENTRYLK:
case GF_FOP_FENTRYLK:
case GF_FOP_LK:
- return;
+ return;
default:
- break;
- }
+ break;
+ }
- priv = frame->this->private;
- if (!priv->consistent_io)
- return;
+ priv = frame->this->private;
+ if (!priv->consistent_io)
+ return;
- if (local->event_generation &&
- (local->event_generation != priv->event_generation))
- goto inconsistent;
+ if (local->event_generation &&
+ (local->event_generation != priv->event_generation))
+ goto inconsistent;
- return;
+ return;
inconsistent:
- *op_ret = -1;
- *op_errno = ENOTCONN;
+ *op_ret = -1;
+ *op_errno = ENOTCONN;
}
void
-afr_local_cleanup (afr_local_t *local, xlator_t *this)
+afr_local_cleanup(afr_local_t *local, xlator_t *this)
{
- afr_private_t * priv = NULL;
-
- if (!local)
- return;
-
- syncbarrier_destroy (&local->barrier);
-
- afr_local_transaction_cleanup (local, this);
-
- priv = this->private;
-
- loc_wipe (&local->loc);
- loc_wipe (&local->newloc);
-
- if (local->fd)
- fd_unref (local->fd);
-
- if (local->xattr_req)
- dict_unref (local->xattr_req);
-
- if (local->xattr_rsp)
- dict_unref (local->xattr_rsp);
+ afr_private_t *priv = NULL;
- if (local->dict)
- dict_unref (local->dict);
+ if (!local)
+ return;
- afr_local_replies_wipe (local, priv);
- GF_FREE(local->replies);
+ syncbarrier_destroy(&local->barrier);
- GF_FREE (local->child_up);
+ afr_local_transaction_cleanup(local, this);
- GF_FREE (local->read_attempted);
+ priv = this->private;
- GF_FREE (local->readable);
- GF_FREE (local->readable2);
+ loc_wipe(&local->loc);
+ loc_wipe(&local->newloc);
- if (local->inode)
- inode_unref (local->inode);
+ if (local->fd)
+ fd_unref(local->fd);
- if (local->parent)
- inode_unref (local->parent);
+ if (local->xattr_req)
+ dict_unref(local->xattr_req);
- if (local->parent2)
- inode_unref (local->parent2);
+ if (local->xattr_rsp)
+ dict_unref(local->xattr_rsp);
- if (local->refreshinode)
- inode_unref (local->refreshinode);
+ if (local->dict)
+ dict_unref(local->dict);
- { /* getxattr */
- GF_FREE (local->cont.getxattr.name);
- }
+ afr_local_replies_wipe(local, priv);
+ GF_FREE(local->replies);
- { /* lk */
- GF_FREE (local->cont.lk.locked_nodes);
- }
+ GF_FREE(local->child_up);
- { /* create */
- if (local->cont.create.fd)
- fd_unref (local->cont.create.fd);
- if (local->cont.create.params)
- dict_unref (local->cont.create.params);
- }
+ GF_FREE(local->read_attempted);
+
+ GF_FREE(local->readable);
+ GF_FREE(local->readable2);
+
+ if (local->inode)
+ inode_unref(local->inode);
- { /* mknod */
- if (local->cont.mknod.params)
- dict_unref (local->cont.mknod.params);
- }
+ if (local->parent)
+ inode_unref(local->parent);
- { /* mkdir */
- if (local->cont.mkdir.params)
- dict_unref (local->cont.mkdir.params);
- }
+ if (local->parent2)
+ inode_unref(local->parent2);
- { /* symlink */
- if (local->cont.symlink.params)
- dict_unref (local->cont.symlink.params);
- }
+ if (local->refreshinode)
+ inode_unref(local->refreshinode);
- { /* writev */
- GF_FREE (local->cont.writev.vector);
- if (local->cont.writev.iobref)
- iobref_unref (local->cont.writev.iobref);
- }
+ { /* getxattr */
+ GF_FREE(local->cont.getxattr.name);
+ }
- { /* setxattr */
- if (local->cont.setxattr.dict)
- dict_unref (local->cont.setxattr.dict);
- }
+ { /* lk */
+ GF_FREE(local->cont.lk.locked_nodes);
+ GF_FREE(local->cont.lk.dom_locked_nodes);
+ GF_FREE(local->cont.lk.dom_lock_op_ret);
+ GF_FREE(local->cont.lk.dom_lock_op_errno);
+ }
- { /* fsetxattr */
- if (local->cont.fsetxattr.dict)
- dict_unref (local->cont.fsetxattr.dict);
- }
+ { /* create */
+ if (local->cont.create.fd)
+ fd_unref(local->cont.create.fd);
+ if (local->cont.create.params)
+ dict_unref(local->cont.create.params);
+ }
- { /* removexattr */
- GF_FREE (local->cont.removexattr.name);
- }
- { /* xattrop */
- if (local->cont.xattrop.xattr)
- dict_unref (local->cont.xattrop.xattr);
- }
- { /* symlink */
- GF_FREE (local->cont.symlink.linkpath);
- }
+ { /* mknod */
+ if (local->cont.mknod.params)
+ dict_unref(local->cont.mknod.params);
+ }
- { /* opendir */
- GF_FREE (local->cont.opendir.checksum);
- }
+ { /* mkdir */
+ if (local->cont.mkdir.params)
+ dict_unref(local->cont.mkdir.params);
+ }
- { /* open */
- if (local->cont.open.fd)
- fd_unref (local->cont.open.fd);
- }
+ { /* symlink */
+ if (local->cont.symlink.params)
+ dict_unref(local->cont.symlink.params);
+ }
- { /* readdirp */
- if (local->cont.readdir.dict)
- dict_unref (local->cont.readdir.dict);
- }
+ { /* writev */
+ GF_FREE(local->cont.writev.vector);
+ if (local->cont.writev.iobref)
+ iobref_unref(local->cont.writev.iobref);
+ }
- { /* inodelk */
- GF_FREE (local->cont.inodelk.volume);
- if (local->cont.inodelk.xdata)
- dict_unref (local->cont.inodelk.xdata);
- }
+ { /* setxattr */
+ if (local->cont.setxattr.dict)
+ dict_unref(local->cont.setxattr.dict);
+ }
- { /* entrylk */
- GF_FREE (local->cont.entrylk.volume);
- GF_FREE (local->cont.entrylk.basename);
- if (local->cont.entrylk.xdata)
- dict_unref (local->cont.entrylk.xdata);
- }
+ { /* fsetxattr */
+ if (local->cont.fsetxattr.dict)
+ dict_unref(local->cont.fsetxattr.dict);
+ }
- if (local->xdata_req)
- dict_unref (local->xdata_req);
+ { /* removexattr */
+ GF_FREE(local->cont.removexattr.name);
+ }
+ { /* xattrop */
+ if (local->cont.xattrop.xattr)
+ dict_unref(local->cont.xattrop.xattr);
+ }
+ { /* symlink */
+ GF_FREE(local->cont.symlink.linkpath);
+ }
- if (local->xdata_rsp)
- dict_unref (local->xdata_rsp);
+ { /* opendir */
+ GF_FREE(local->cont.opendir.checksum);
+ }
+
+ { /* open */
+ if (local->cont.open.fd)
+ fd_unref(local->cont.open.fd);
+ }
+
+ { /* readdirp */
+ if (local->cont.readdir.dict)
+ dict_unref(local->cont.readdir.dict);
+ }
+
+ { /* inodelk */
+ GF_FREE(local->cont.inodelk.volume);
+ if (local->cont.inodelk.xdata)
+ dict_unref(local->cont.inodelk.xdata);
+ }
+
+ { /* entrylk */
+ GF_FREE(local->cont.entrylk.volume);
+ GF_FREE(local->cont.entrylk.basename);
+ if (local->cont.entrylk.xdata)
+ dict_unref(local->cont.entrylk.xdata);
+ }
+
+ if (local->xdata_req)
+ dict_unref(local->xdata_req);
+
+ if (local->xdata_rsp)
+ dict_unref(local->xdata_rsp);
}
-
int
-afr_frame_return (call_frame_t *frame)
+afr_frame_return(call_frame_t *frame)
{
- afr_local_t *local = NULL;
- int call_count = 0;
+ afr_local_t *local = NULL;
+ int call_count = 0;
- local = frame->local;
+ local = frame->local;
- LOCK (&frame->lock);
- {
- call_count = --local->call_count;
- }
- UNLOCK (&frame->lock);
+ LOCK(&frame->lock);
+ {
+ call_count = --local->call_count;
+ }
+ UNLOCK(&frame->lock);
- return call_count;
+ return call_count;
}
-static char *afr_ignore_xattrs[] = {
- GF_SELINUX_XATTR_KEY,
- QUOTA_SIZE_KEY,
- NULL
-};
+static char *afr_ignore_xattrs[] = {GF_SELINUX_XATTR_KEY, QUOTA_SIZE_KEY, NULL};
gf_boolean_t
-afr_is_xattr_ignorable (char *key)
+afr_is_xattr_ignorable(char *key)
{
- int i = 0;
+ int i = 0;
- if (!strncmp (key, AFR_XATTR_PREFIX, strlen(AFR_XATTR_PREFIX)))
- return _gf_true;
- for (i = 0; afr_ignore_xattrs[i]; i++) {
- if (!strcmp (key, afr_ignore_xattrs[i]))
- return _gf_true;
- }
- return _gf_false;
+ if (!strncmp(key, AFR_XATTR_PREFIX, SLEN(AFR_XATTR_PREFIX)))
+ return _gf_true;
+ for (i = 0; afr_ignore_xattrs[i]; i++) {
+ if (!strcmp(key, afr_ignore_xattrs[i]))
+ return _gf_true;
+ }
+ return _gf_false;
}
static gf_boolean_t
-afr_xattr_match_needed (dict_t *this, char *key1, data_t *value1, void *data)
+afr_xattr_match_needed(dict_t *this, char *key1, data_t *value1, void *data)
{
- /* Ignore all non-disk (i.e. virtual) xattrs right away. */
- if (!gf_is_valid_xattr_namespace (key1))
- return _gf_false;
+ /* Ignore all non-disk (i.e. virtual) xattrs right away. */
+ if (!gf_is_valid_xattr_namespace(key1))
+ return _gf_false;
- /* Ignore on-disk xattrs that AFR doesn't need to heal. */
- if (!afr_is_xattr_ignorable (key1))
- return _gf_true;
+ /* Ignore on-disk xattrs that AFR doesn't need to heal. */
+ if (!afr_is_xattr_ignorable(key1))
+ return _gf_true;
- return _gf_false;
+ return _gf_false;
}
gf_boolean_t
-afr_xattrs_are_equal (dict_t *dict1, dict_t *dict2)
+afr_xattrs_are_equal(dict_t *dict1, dict_t *dict2)
{
- return are_dicts_equal (dict1, dict2, afr_xattr_match_needed, NULL);
+ return are_dicts_equal(dict1, dict2, afr_xattr_match_needed, NULL);
}
static int
-afr_get_parent_read_subvol (xlator_t *this, inode_t *parent,
- struct afr_reply *replies, unsigned char *readable)
+afr_get_parent_read_subvol(xlator_t *this, inode_t *parent,
+ struct afr_reply *replies, unsigned char *readable)
{
- int i = 0;
- int par_read_subvol = -1;
- int par_read_subvol_iter = -1;
- afr_private_t *priv = NULL;
-
- priv = this->private;
-
- if (parent)
- par_read_subvol = afr_data_subvol_get (parent, this, NULL, NULL,
- NULL, NULL);
+ int i = 0;
+ int par_read_subvol = -1;
+ int par_read_subvol_iter = -1;
+ afr_private_t *priv = NULL;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
+ priv = this->private;
- if (replies[i].op_ret < 0)
- continue;
+ if (parent)
+ par_read_subvol = afr_data_subvol_get(parent, this, NULL, NULL, NULL,
+ NULL);
- if (par_read_subvol_iter == -1) {
- par_read_subvol_iter = i;
- continue;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
- if ((par_read_subvol_iter != par_read_subvol) && readable[i])
- par_read_subvol_iter = i;
+ if (replies[i].op_ret < 0)
+ continue;
- if (i == par_read_subvol)
- par_read_subvol_iter = i;
+ if (par_read_subvol_iter == -1) {
+ par_read_subvol_iter = i;
+ continue;
}
- /* At the end of the for-loop, the only reason why @par_read_subvol_iter
- * could be -1 is when this LOOKUP has failed on all sub-volumes.
- * So it is okay to send an arbitrary subvolume (0 in this case)
- * as parent read subvol.
- */
- if (par_read_subvol_iter == -1)
- par_read_subvol_iter = 0;
- return par_read_subvol_iter;
+ if ((par_read_subvol_iter != par_read_subvol) && readable[i])
+ par_read_subvol_iter = i;
+ if (i == par_read_subvol)
+ par_read_subvol_iter = i;
+ }
+ /* At the end of the for-loop, the only reason why @par_read_subvol_iter
+ * could be -1 is when this LOOKUP has failed on all sub-volumes.
+ * So it is okay to send an arbitrary subvolume (0 in this case)
+ * as parent read subvol.
+ */
+ if (par_read_subvol_iter == -1)
+ par_read_subvol_iter = 0;
+
+ return par_read_subvol_iter;
}
int
-afr_read_subvol_decide (inode_t *inode, xlator_t *this,
- afr_read_subvol_args_t *args, unsigned char *readable)
+afr_read_subvol_decide(inode_t *inode, xlator_t *this,
+ afr_read_subvol_args_t *args, unsigned char *readable)
{
- int event = 0;
- afr_private_t *priv = NULL;
- unsigned char *intersection = NULL;
+ int event = 0;
+ afr_private_t *priv = NULL;
+ unsigned char *intersection = NULL;
- priv = this->private;
- intersection = alloca0 (priv->child_count);
+ priv = this->private;
+ intersection = alloca0(priv->child_count);
- afr_readables_intersect_get (inode, this, &event, intersection);
+ afr_readables_intersect_get(inode, this, &event, intersection);
- if (AFR_COUNT (intersection, priv->child_count) <= 0) {
- /* TODO: If we have one brick with valid data_readable and
- * another with metadata_readable, try to send an iatt with
- * valid bits from both.*/
- return -1;
- }
+ if (AFR_COUNT(intersection, priv->child_count) <= 0) {
+ /* TODO: If we have one brick with valid data_readable and
+ * another with metadata_readable, try to send an iatt with
+ * valid bits from both.*/
+ return -1;
+ }
- memcpy (readable, intersection, sizeof (*readable) * priv->child_count);
+ memcpy(readable, intersection, sizeof(*readable) * priv->child_count);
- return afr_read_subvol_select_by_policy (inode, this, intersection,
- args);
+ return afr_read_subvol_select_by_policy(inode, this, intersection, args);
}
static inline int
-afr_first_up_child (call_frame_t *frame, xlator_t *this)
+afr_first_up_child(call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++)
- if (local->replies[i].valid &&
- local->replies[i].op_ret == 0)
- return i;
- return -1;
+ for (i = 0; i < priv->child_count; i++)
+ if (local->replies[i].valid && local->replies[i].op_ret == 0)
+ return i;
+ return -1;
}
static void
-afr_attempt_readsubvol_set (call_frame_t *frame, xlator_t *this,
- unsigned char *success_replies,
- unsigned char *data_readable, int *read_subvol)
+afr_attempt_readsubvol_set(call_frame_t *frame, xlator_t *this,
+ unsigned char *success_replies,
+ unsigned char *data_readable, int *read_subvol)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int spb_choice = -1;
- int child_count = -1;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int spb_subvol = -1;
+ int child_count = -1;
- if (*read_subvol != -1)
- return;
+ if (*read_subvol != -1)
+ return;
- priv = this->private;
- local = frame->local;
- child_count = priv->child_count;
-
- afr_inode_split_brain_choice_get (local->inode, this,
- &spb_choice);
- if ((spb_choice >= 0) &&
- (AFR_COUNT(success_replies, child_count) == child_count)) {
- *read_subvol = spb_choice;
- } else if (!priv->quorum_count) {
- *read_subvol = afr_first_up_child (frame, this);
- } else if (priv->quorum_count &&
- afr_has_quorum (data_readable, this)) {
- /* read_subvol is guaranteed to be valid if we hit this path. */
- *read_subvol = afr_first_up_child (frame, this);
- } else {
- /* If quorum is enabled and we do not have a
- readable yet, it means all good copies are down.
- */
- local->op_ret = -1;
- local->op_errno = ENOTCONN;
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_READ_SUBVOL_ERROR, "no read "
- "subvols for %s", local->loc.path);
- }
- if (*read_subvol >= 0)
- dict_del (local->replies[*read_subvol].xdata, GF_CONTENT_KEY);
+ priv = this->private;
+ local = frame->local;
+ child_count = priv->child_count;
+
+ afr_split_brain_read_subvol_get(local->inode, this, frame, &spb_subvol);
+ if ((spb_subvol >= 0) &&
+ (AFR_COUNT(success_replies, child_count) == child_count)) {
+ *read_subvol = spb_subvol;
+ } else if (!priv->quorum_count ||
+ frame->root->pid == GF_CLIENT_PID_GLFS_HEAL) {
+ *read_subvol = afr_first_up_child(frame, this);
+ } else if (priv->quorum_count &&
+ afr_has_quorum(data_readable, this, NULL)) {
+ /* read_subvol is guaranteed to be valid if we hit this path. */
+ *read_subvol = afr_first_up_child(frame, this);
+ } else {
+ /* If quorum is enabled and we do not have a
+ readable yet, it means all good copies are down.
+ */
+ local->op_ret = -1;
+ local->op_errno = ENOTCONN;
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_READ_SUBVOL_ERROR,
+ "no read "
+ "subvols for %s",
+ local->loc.path);
+ }
+ if (*read_subvol >= 0)
+ dict_del_sizen(local->replies[*read_subvol].xdata, GF_CONTENT_KEY);
}
static void
-afr_lookup_done (call_frame_t *frame, xlator_t *this)
-{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = -1;
- int op_errno = 0;
- int read_subvol = 0;
- int par_read_subvol = 0;
- int ret = -1;
- unsigned char *readable = NULL;
- unsigned char *success_replies = NULL;
- int event = 0;
- struct afr_reply *replies = NULL;
- uuid_t read_gfid = {0, };
- gf_boolean_t locked_entry = _gf_false;
- gf_boolean_t can_interpret = _gf_true;
- inode_t *parent = NULL;
- ia_type_t ia_type = IA_INVAL;
- afr_read_subvol_args_t args = {0,};
- char *gfid_heal_msg = NULL;
-
- priv = this->private;
- local = frame->local;
- replies = local->replies;
- parent = local->loc.parent;
-
- locked_entry = afr_is_possibly_under_txn (AFR_ENTRY_TRANSACTION, local,
- this);
-
- readable = alloca0 (priv->child_count);
- success_replies = alloca0 (priv->child_count);
-
- afr_inode_read_subvol_get (parent, this, readable, NULL, &event);
- par_read_subvol = afr_get_parent_read_subvol (this, parent, replies,
- readable);
-
- /* First, check if we have a gfid-change from somewhere,
- If so, propagate that so that a fresh lookup can be
- issued
- */
- if (local->cont.lookup.needs_fresh_lookup) {
- local->op_ret = -1;
- local->op_errno = ESTALE;
- goto error;
- }
-
- op_errno = afr_final_errno (frame->local, this->private);
- local->op_errno = op_errno;
-
- read_subvol = -1;
- for (i = 0; i < priv->child_count; i++)
- if (replies[i].valid && replies[i].op_ret == 0)
- success_replies[i] = 1;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
-
- if (locked_entry && replies[i].op_ret == -1 &&
- replies[i].op_errno == ENOENT) {
- /* Second, check entry is still
- "underway" in creation */
- local->op_ret = -1;
- local->op_errno = ENOENT;
- goto error;
- }
-
- if (replies[i].op_ret == -1)
- continue;
-
- if (read_subvol == -1 || !readable[read_subvol]) {
- read_subvol = i;
- gf_uuid_copy (read_gfid, replies[i].poststat.ia_gfid);
- ia_type = replies[i].poststat.ia_type;
- local->op_ret = 0;
- }
- }
+afr_lookup_done(call_frame_t *frame, xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = -1;
+ int op_errno = 0;
+ int read_subvol = 0;
+ int par_read_subvol = 0;
+ int ret = -1;
+ unsigned char *readable = NULL;
+ unsigned char *success_replies = NULL;
+ int event = 0;
+ struct afr_reply *replies = NULL;
+ uuid_t read_gfid = {
+ 0,
+ };
+ gf_boolean_t locked_entry = _gf_false;
+ gf_boolean_t in_flight_create = _gf_false;
+ gf_boolean_t can_interpret = _gf_true;
+ inode_t *parent = NULL;
+ ia_type_t ia_type = IA_INVAL;
+ afr_read_subvol_args_t args = {
+ 0,
+ };
+ char *gfid_heal_msg = NULL;
+
+ priv = this->private;
+ local = frame->local;
+ replies = local->replies;
+ parent = local->loc.parent;
+
+ locked_entry = afr_is_possibly_under_txn(AFR_ENTRY_TRANSACTION, local,
+ this);
+
+ readable = alloca0(priv->child_count);
+ success_replies = alloca0(priv->child_count);
+
+ afr_inode_read_subvol_get(parent, this, readable, NULL, &event);
+ par_read_subvol = afr_get_parent_read_subvol(this, parent, replies,
+ readable);
+
+ /* First, check if we have a gfid-change from somewhere,
+ If so, propagate that so that a fresh lookup can be
+ issued
+ */
+ if (local->cont.lookup.needs_fresh_lookup) {
+ local->op_ret = -1;
+ local->op_errno = ESTALE;
+ goto error;
+ }
- if (read_subvol == -1)
- goto error;
- /* We now have a read_subvol, which is readable[] (if there
- were any). Next we look for GFID mismatches. We don't
- consider a GFID mismatch as an error if read_subvol is
- readable[] but the mismatching GFID subvol is not.
- */
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret == -1) {
- if (priv->child_up[i])
- can_interpret = _gf_false;
- continue;
- }
-
- if (!gf_uuid_compare (replies[i].poststat.ia_gfid, read_gfid))
- continue;
-
- can_interpret = _gf_false;
-
- if (locked_entry)
- continue;
-
- /* Now GFIDs mismatch. It's OK as long as this subvol
- is not readable[] but read_subvol is */
- if (readable[read_subvol] && !readable[i])
- continue;
-
- /* If we were called from glfsheal and there is still a gfid
- * mismatch, succeed the lookup and let glfsheal print the
- * response via gfid-heal-msg.*/
- if (!dict_get_str (local->xattr_req, "gfid-heal-msg",
- &gfid_heal_msg))
- goto cant_interpret;
-
- /* LOG ERROR */
- local->op_ret = -1;
- local->op_errno = EIO;
- goto error;
- }
-
- /* Forth, for the finalized GFID, pick the best subvolume
- to return stats from.
- */
- read_subvol = -1;
- memset (readable, 0, sizeof (*readable) * priv->child_count);
- if (can_interpret) {
- if (!afr_has_quorum (success_replies, this))
- goto cant_interpret;
- /* It is safe to call afr_replies_interpret() because we have
- a response from all the UP subvolumes and all of them resolved
- to the same GFID
- */
- gf_uuid_copy (args.gfid, read_gfid);
- args.ia_type = ia_type;
- ret = afr_replies_interpret (frame, this, local->inode, NULL);
- read_subvol = afr_read_subvol_decide (local->inode, this, &args,
- readable);
- if (read_subvol == -1)
- goto cant_interpret;
- if (ret) {
- afr_inode_event_gen_reset (local->inode, this);
- dict_del (local->replies[read_subvol].xdata,
- GF_CONTENT_KEY);
- }
- } else {
- cant_interpret:
- afr_attempt_readsubvol_set (frame, this, success_replies,
- readable, &read_subvol);
- if (read_subvol == -1) {
- goto error;
- }
- }
+ op_errno = afr_final_errno(frame->local, this->private);
+ local->op_errno = op_errno;
+
+ read_subvol = -1;
+ afr_fill_success_replies(local, priv, success_replies);
- afr_handle_quota_size (frame, this);
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
- afr_set_need_heal (this, local);
- if (AFR_IS_ARBITER_BRICK (priv, read_subvol) && local->op_ret == 0) {
- local->op_ret = -1;
- local->op_errno = ENOTCONN;
- gf_msg_debug(this->name, 0, "Arbiter cannot be a read subvol "
- "for %s", local->loc.path);
- goto error;
+ if (replies[i].op_ret == -1) {
+ if (locked_entry && replies[i].op_errno == ENOENT) {
+ in_flight_create = _gf_true;
+ }
+ continue;
}
- ret = dict_get_str (local->xattr_req, "gfid-heal-msg", &gfid_heal_msg);
- if (!ret) {
- ret = dict_set_str (local->replies[read_subvol].xdata,
- "gfid-heal-msg", gfid_heal_msg);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_DICT_SET_FAILED,
- "Error setting gfid-heal-msg dict");
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- }
+ if (read_subvol == -1 || !readable[read_subvol]) {
+ read_subvol = i;
+ gf_uuid_copy(read_gfid, replies[i].poststat.ia_gfid);
+ ia_type = replies[i].poststat.ia_type;
+ local->op_ret = 0;
}
+ }
- AFR_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
- local->inode, &local->replies[read_subvol].poststat,
- local->replies[read_subvol].xdata,
- &local->replies[par_read_subvol].postparent);
- return;
+ if (in_flight_create && !afr_has_quorum(success_replies, this, NULL)) {
+ local->op_ret = -1;
+ local->op_errno = ENOENT;
+ goto error;
+ }
+
+ if (read_subvol == -1)
+ goto error;
+ /* We now have a read_subvol, which is readable[] (if there
+ were any). Next we look for GFID mismatches. We don't
+ consider a GFID mismatch as an error if read_subvol is
+ readable[] but the mismatching GFID subvol is not.
+ */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret == -1) {
+ continue;
+ }
+
+ if (!gf_uuid_compare(replies[i].poststat.ia_gfid, read_gfid))
+ continue;
+
+ can_interpret = _gf_false;
+
+ if (locked_entry)
+ continue;
+
+ /* Now GFIDs mismatch. It's OK as long as this subvol
+ is not readable[] but read_subvol is */
+ if (readable[read_subvol] && !readable[i])
+ continue;
+
+ /* If we were called from glfsheal and there is still a gfid
+ * mismatch, succeed the lookup and let glfsheal print the
+ * response via gfid-heal-msg.*/
+ if (!dict_get_str_sizen(local->xattr_req, "gfid-heal-msg",
+ &gfid_heal_msg))
+ goto cant_interpret;
+
+ /* LOG ERROR */
+ local->op_ret = -1;
+ local->op_errno = EIO;
+ goto error;
+ }
+
+ /* Forth, for the finalized GFID, pick the best subvolume
+ to return stats from.
+ */
+ read_subvol = -1;
+ memset(readable, 0, sizeof(*readable) * priv->child_count);
+ if (can_interpret) {
+ if (!afr_has_quorum(success_replies, this, NULL))
+ goto cant_interpret;
+ /* It is safe to call afr_replies_interpret() because we have
+ a response from all the UP subvolumes and all of them resolved
+ to the same GFID
+ */
+ gf_uuid_copy(args.gfid, read_gfid);
+ args.ia_type = ia_type;
+ ret = afr_replies_interpret(frame, this, local->inode, NULL);
+ read_subvol = afr_read_subvol_decide(local->inode, this, &args,
+ readable);
+ if (read_subvol == -1)
+ goto cant_interpret;
+ if (ret) {
+ afr_inode_need_refresh_set(local->inode, this);
+ dict_del_sizen(local->replies[read_subvol].xdata, GF_CONTENT_KEY);
+ }
+ } else {
+ cant_interpret:
+ afr_attempt_readsubvol_set(frame, this, success_replies, readable,
+ &read_subvol);
+ if (read_subvol == -1) {
+ goto error;
+ }
+ }
+
+ afr_handle_quota_size(frame, this);
+
+ afr_set_need_heal(this, local);
+ if (AFR_IS_ARBITER_BRICK(priv, read_subvol) && local->op_ret == 0) {
+ local->op_ret = -1;
+ local->op_errno = ENOTCONN;
+ gf_msg_debug(this->name, 0,
+ "Arbiter cannot be a read subvol "
+ "for %s",
+ local->loc.path);
+ goto error;
+ }
+
+ ret = dict_get_str_sizen(local->xattr_req, "gfid-heal-msg", &gfid_heal_msg);
+ if (!ret) {
+ ret = dict_set_str_sizen(local->replies[read_subvol].xdata,
+ "gfid-heal-msg", gfid_heal_msg);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_DICT_SET_FAILED,
+ "Error setting gfid-heal-msg dict");
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ }
+ }
+
+ AFR_STACK_UNWIND(lookup, frame, local->op_ret, local->op_errno,
+ local->inode, &local->replies[read_subvol].poststat,
+ local->replies[read_subvol].xdata,
+ &local->replies[par_read_subvol].postparent);
+ return;
error:
- AFR_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno, NULL,
- NULL, NULL, NULL);
+ AFR_STACK_UNWIND(lookup, frame, local->op_ret, local->op_errno, NULL, NULL,
+ NULL, NULL);
}
/*
@@ -2500,751 +3122,845 @@ error:
* others in that they must be given higher priority while
* returning to the user.
*
- * The hierarchy is ENODATA > ENOENT > ESTALE > others
+ * The hierarchy is ENODATA > ENOENT > ESTALE > ENOSPC others
*/
int
-afr_higher_errno (int32_t old_errno, int32_t new_errno)
+afr_higher_errno(int32_t old_errno, int32_t new_errno)
{
- if (old_errno == ENODATA || new_errno == ENODATA)
- return ENODATA;
- if (old_errno == ENOENT || new_errno == ENOENT)
- return ENOENT;
- if (old_errno == ESTALE || new_errno == ESTALE)
- return ESTALE;
+ if (old_errno == ENODATA || new_errno == ENODATA)
+ return ENODATA;
+ if (old_errno == ENOENT || new_errno == ENOENT)
+ return ENOENT;
+ if (old_errno == ESTALE || new_errno == ESTALE)
+ return ESTALE;
+ if (old_errno == ENOSPC || new_errno == ENOSPC)
+ return ENOSPC;
- return new_errno;
+ return new_errno;
}
-
int
-afr_final_errno (afr_local_t *local, afr_private_t *priv)
+afr_final_errno(afr_local_t *local, afr_private_t *priv)
{
- int i = 0;
- int op_errno = 0;
- int tmp_errno = 0;
+ int i = 0;
+ int op_errno = 0;
+ int tmp_errno = 0;
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
- if (local->replies[i].op_ret >= 0)
- continue;
- tmp_errno = local->replies[i].op_errno;
- op_errno = afr_higher_errno (op_errno, tmp_errno);
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+ if (local->replies[i].op_ret >= 0)
+ continue;
+ tmp_errno = local->replies[i].op_errno;
+ op_errno = afr_higher_errno(op_errno, tmp_errno);
+ }
- return op_errno;
+ return op_errno;
}
static int32_t
-afr_local_discovery_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
-{
- int ret = 0;
- char *pathinfo = NULL;
- gf_boolean_t is_local = _gf_false;
- afr_private_t *priv = NULL;
- int32_t child_index = -1;
-
- if (op_ret != 0) {
- goto out;
- }
-
- priv = this->private;
- child_index = (int32_t)(long)cookie;
-
- ret = dict_get_str (dict, GF_XATTR_PATHINFO_KEY, &pathinfo);
- if (ret != 0) {
- goto out;
- }
-
- ret = glusterfs_is_local_pathinfo (pathinfo, &is_local);
- if (ret) {
- goto out;
- }
-
- /*
- * Note that one local subvolume will override another here. The only
- * way to avoid that would be to retain extra information about whether
- * the previous read_child is local, and it's just not worth it. Even
- * the slowest local subvolume is far preferable to a remote one.
- */
- if (is_local) {
- priv->local[child_index] = 1;
- /* Don't set arbiter as read child. */
- if (AFR_IS_ARBITER_BRICK(priv, child_index))
- goto out;
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_LOCAL_CHILD, "selecting local read_child %s",
- priv->children[child_index]->name);
-
- priv->read_child = child_index;
- }
+afr_local_discovery_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
+{
+ int ret = 0;
+ char *pathinfo = NULL;
+ gf_boolean_t is_local = _gf_false;
+ afr_private_t *priv = NULL;
+ int32_t child_index = -1;
+
+ if (op_ret != 0) {
+ goto out;
+ }
+
+ priv = this->private;
+ child_index = (int32_t)(long)cookie;
+
+ ret = dict_get_str_sizen(dict, GF_XATTR_PATHINFO_KEY, &pathinfo);
+ if (ret != 0) {
+ goto out;
+ }
+
+ ret = glusterfs_is_local_pathinfo(pathinfo, &is_local);
+ if (ret) {
+ goto out;
+ }
+
+ /*
+ * Note that one local subvolume will override another here. The only
+ * way to avoid that would be to retain extra information about whether
+ * the previous read_child is local, and it's just not worth it. Even
+ * the slowest local subvolume is far preferable to a remote one.
+ */
+ if (is_local) {
+ priv->local[child_index] = 1;
+ /* Don't set arbiter as read child. */
+ if (AFR_IS_ARBITER_BRICK(priv, child_index))
+ goto out;
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_LOCAL_CHILD,
+ "selecting local read_child %s",
+ priv->children[child_index]->name);
+
+ priv->read_child = child_index;
+ }
out:
- STACK_DESTROY(frame->root);
- return 0;
+ STACK_DESTROY(frame->root);
+ return 0;
}
static void
-afr_attempt_local_discovery (xlator_t *this, int32_t child_index)
+afr_attempt_local_discovery(xlator_t *this, int32_t child_index)
{
- call_frame_t *newframe = NULL;
- loc_t tmploc = {0,};
- afr_private_t *priv = this->private;
+ call_frame_t *newframe = NULL;
+ loc_t tmploc = {
+ 0,
+ };
+ afr_private_t *priv = this->private;
- newframe = create_frame(this,this->ctx->pool);
- if (!newframe) {
- return;
- }
+ newframe = create_frame(this, this->ctx->pool);
+ if (!newframe) {
+ return;
+ }
- tmploc.gfid[sizeof(tmploc.gfid)-1] = 1;
- STACK_WIND_COOKIE (newframe, afr_local_discovery_cbk,
- (void *)(long)child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->getxattr,
- &tmploc, GF_XATTR_PATHINFO_KEY, NULL);
+ tmploc.gfid[sizeof(tmploc.gfid) - 1] = 1;
+ STACK_WIND_COOKIE(newframe, afr_local_discovery_cbk,
+ (void *)(long)child_index, priv->children[child_index],
+ priv->children[child_index]->fops->getxattr, &tmploc,
+ GF_XATTR_PATHINFO_KEY, NULL);
}
int
-afr_lookup_sh_metadata_wrap (void *opaque)
-{
- call_frame_t *frame = opaque;
- afr_local_t *local = NULL;
- xlator_t *this = NULL;
- inode_t *inode = NULL;
- afr_private_t *priv = NULL;
- struct afr_reply *replies = NULL;
- int i= 0, first = -1;
- int ret = -1;
- dict_t *dict = NULL;
+afr_lookup_sh_metadata_wrap(void *opaque)
+{
+ call_frame_t *frame = opaque;
+ afr_local_t *local = NULL;
+ xlator_t *this = NULL;
+ inode_t *inode = NULL;
+ afr_private_t *priv = NULL;
+ struct afr_reply *replies = NULL;
+ int i = 0, first = -1;
+ int ret = -1;
+ dict_t *dict = NULL;
+
+ local = frame->local;
+ this = frame->this;
+ priv = this->private;
+ replies = local->replies;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret == -1)
+ continue;
+ first = i;
+ break;
+ }
+ if (first == -1)
+ goto out;
+
+ if (afr_selfheal_metadata_by_stbuf(this, &replies[first].poststat))
+ goto out;
+
+ afr_local_replies_wipe(local, this->private);
+
+ dict = dict_new();
+ if (!dict)
+ goto out;
+ if (local->xattr_req) {
+ dict_copy(local->xattr_req, dict);
+ }
+
+ ret = dict_set_sizen_str_sizen(dict, "link-count", GF_XATTROP_INDEX_COUNT);
+ if (ret) {
+ gf_msg_debug(this->name, -ret, "Unable to set link-count in dict ");
+ }
+
+ if (loc_is_nameless(&local->loc)) {
+ ret = afr_selfheal_unlocked_discover_on(frame, local->inode,
+ local->loc.gfid, local->replies,
+ local->child_up, dict);
+ } else {
+ inode = afr_selfheal_unlocked_lookup_on(frame, local->loc.parent,
+ local->loc.name, local->replies,
+ local->child_up, dict);
+ }
+ if (inode)
+ inode_unref(inode);
+out:
+ if (loc_is_nameless(&local->loc))
+ afr_discover_done(frame, this);
+ else
+ afr_lookup_done(frame, this);
- local = frame->local;
- this = frame->this;
- priv = this->private;
- replies = local->replies;
-
- for (i =0; i < priv->child_count; i++) {
- if(!replies[i].valid || replies[i].op_ret == -1)
- continue;
- first = i;
- break;
- }
- if (first == -1)
- goto out;
+ if (dict)
+ dict_unref(dict);
- if (afr_selfheal_metadata_by_stbuf (this, &replies[first].poststat))
- goto out;
+ return 0;
+}
- afr_local_replies_wipe (local, this->private);
+gf_boolean_t
+afr_is_pending_set(xlator_t *this, dict_t *xdata, int type)
+{
+ int idx = -1;
+ afr_private_t *priv = NULL;
+ void *pending_raw = NULL;
+ int *pending_int = NULL;
+ int i = 0;
- dict = dict_new ();
- if (!dict)
- goto out;
- ret = dict_set_str (dict, "link-count", GF_XATTROP_INDEX_COUNT);
- if (ret) {
- gf_msg_debug (this->name, -ret,
- "Unable to set link-count in dict ");
+ priv = this->private;
+ idx = afr_index_for_transaction_type(type);
+
+ if (dict_get_ptr(xdata, AFR_DIRTY, &pending_raw) == 0) {
+ if (pending_raw) {
+ pending_int = pending_raw;
+
+ if (ntoh32(pending_int[idx]))
+ return _gf_true;
}
+ }
- if (loc_is_nameless (&local->loc)) {
- ret = afr_selfheal_unlocked_discover_on (frame, local->inode,
- local->loc.gfid,
- local->replies,
- local->child_up);
- } else {
- inode = afr_selfheal_unlocked_lookup_on (frame,
- local->loc.parent,
- local->loc.name,
- local->replies,
- local->child_up, dict);
- }
- if (inode)
- inode_unref (inode);
-out:
- if (loc_is_nameless (&local->loc))
- afr_discover_done (frame, this);
- else
- afr_lookup_done (frame, this);
+ for (i = 0; i < priv->child_count; i++) {
+ if (dict_get_ptr(xdata, priv->pending_key[i], &pending_raw))
+ continue;
+ if (!pending_raw)
+ continue;
+ pending_int = pending_raw;
- if (dict)
- dict_unref (dict);
+ if (ntoh32(pending_int[idx]))
+ return _gf_true;
+ }
- return 0;
+ return _gf_false;
}
static gf_boolean_t
afr_can_start_metadata_self_heal(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- struct afr_reply *replies = NULL;
- int i = 0, first = -1;
- gf_boolean_t start = _gf_false;
- struct iatt stbuf = {0, };
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ struct afr_reply *replies = NULL;
+ int i = 0, first = -1;
+ gf_boolean_t start = _gf_false;
+ struct iatt stbuf = {
+ 0,
+ };
- local = frame->local;
- replies = local->replies;
- priv = this->private;
+ local = frame->local;
+ replies = local->replies;
+ priv = this->private;
- if (!priv->metadata_self_heal)
- return _gf_false;
+ if (!priv->metadata_self_heal)
+ return _gf_false;
- for (i = 0; i < priv->child_count; i++) {
- if(!replies[i].valid || replies[i].op_ret == -1)
- continue;
- if (first == -1) {
- first = i;
- stbuf = replies[i].poststat;
- continue;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret == -1)
+ continue;
+ if (first == -1) {
+ first = i;
+ stbuf = replies[i].poststat;
+ continue;
+ }
- if (gf_uuid_compare (stbuf.ia_gfid, replies[i].poststat.ia_gfid)) {
- start = _gf_false;
- break;
- }
- if (!IA_EQUAL (stbuf, replies[i].poststat, type)) {
- start = _gf_false;
- break;
- }
+ if (afr_is_pending_set(this, replies[i].xdata,
+ AFR_METADATA_TRANSACTION)) {
+ /* Let shd do the heal so that lookup is not blocked
+ * on getting metadata lock/doing the heal */
+ start = _gf_false;
+ break;
+ }
- /*Check if iattrs need heal*/
- if ((!IA_EQUAL (stbuf, replies[i].poststat, uid)) ||
- (!IA_EQUAL (stbuf, replies[i].poststat, gid)) ||
- (!IA_EQUAL (stbuf, replies[i].poststat, prot))) {
- start = _gf_true;
- continue;
- }
+ if (gf_uuid_compare(stbuf.ia_gfid, replies[i].poststat.ia_gfid)) {
+ start = _gf_false;
+ break;
+ }
+ if (!IA_EQUAL(stbuf, replies[i].poststat, type)) {
+ start = _gf_false;
+ break;
+ }
- /*Check if xattrs need heal*/
- if (!afr_xattrs_are_equal (replies[first].xdata,
- replies[i].xdata))
- start = _gf_true;
+ /*Check if iattrs need heal*/
+ if ((!IA_EQUAL(stbuf, replies[i].poststat, uid)) ||
+ (!IA_EQUAL(stbuf, replies[i].poststat, gid)) ||
+ (!IA_EQUAL(stbuf, replies[i].poststat, prot))) {
+ start = _gf_true;
+ continue;
}
- return start;
+ /*Check if xattrs need heal*/
+ if (!afr_xattrs_are_equal(replies[first].xdata, replies[i].xdata))
+ start = _gf_true;
+ }
+
+ return start;
}
int
-afr_lookup_metadata_heal_check (call_frame_t *frame, xlator_t *this)
+afr_lookup_metadata_heal_check(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *heal = NULL;
- afr_local_t *local = NULL;
- int ret = 0;
+ call_frame_t *heal = NULL;
+ afr_local_t *local = NULL;
+ int ret = 0;
- local = frame->local;
- if (!afr_can_start_metadata_self_heal (frame, this))
- goto out;
+ local = frame->local;
+ if (!afr_can_start_metadata_self_heal(frame, this))
+ goto out;
- heal = afr_frame_create (this, &ret);
- if (!heal) {
- ret = -ret;
- goto out;
- }
+ heal = afr_frame_create(this, &ret);
+ if (!heal) {
+ ret = -ret;
+ goto out;
+ }
- ret = synctask_new (this->ctx->env, afr_lookup_sh_metadata_wrap,
- afr_refresh_selfheal_done, heal, frame);
- if (ret)
- goto out;
- return ret;
+ ret = synctask_new(this->ctx->env, afr_lookup_sh_metadata_wrap,
+ afr_refresh_selfheal_done, heal, frame);
+ if (ret)
+ goto out;
+ return ret;
out:
- if (loc_is_nameless (&local->loc))
- afr_discover_done (frame, this);
- else
- afr_lookup_done (frame, this);
- if (heal)
- AFR_STACK_DESTROY (heal);
- return ret;
+ if (loc_is_nameless(&local->loc))
+ afr_discover_done(frame, this);
+ else
+ afr_lookup_done(frame, this);
+ if (heal)
+ AFR_STACK_DESTROY(heal);
+ return ret;
}
int
-afr_lookup_selfheal_wrap (void *opaque)
+afr_lookup_selfheal_wrap(void *opaque)
{
- int ret = 0;
- call_frame_t *frame = opaque;
- afr_local_t *local = NULL;
- xlator_t *this = NULL;
- inode_t *inode = NULL;
- uuid_t pargfid = {0,};
+ int ret = 0;
+ call_frame_t *frame = opaque;
+ afr_local_t *local = NULL;
+ xlator_t *this = NULL;
+ inode_t *inode = NULL;
+ uuid_t pargfid = {
+ 0,
+ };
- local = frame->local;
- this = frame->this;
- loc_pargfid (&local->loc, pargfid);
+ local = frame->local;
+ this = frame->this;
+ loc_pargfid(&local->loc, pargfid);
- ret = afr_selfheal_name (frame->this, pargfid, local->loc.name,
- &local->cont.lookup.gfid_req, local->xattr_req);
- if (ret == -EIO)
- goto unwind;
+ ret = afr_selfheal_name(frame->this, pargfid, local->loc.name,
+ &local->cont.lookup.gfid_req, local->xattr_req);
+ if (ret == -EIO)
+ goto unwind;
- afr_local_replies_wipe (local, this->private);
+ afr_local_replies_wipe(local, this->private);
- inode = afr_selfheal_unlocked_lookup_on (frame, local->loc.parent,
- local->loc.name, local->replies,
- local->child_up, NULL);
- if (inode)
- inode_unref (inode);
+ inode = afr_selfheal_unlocked_lookup_on(frame, local->loc.parent,
+ local->loc.name, local->replies,
+ local->child_up, local->xattr_req);
+ if (inode)
+ inode_unref(inode);
- afr_lookup_metadata_heal_check(frame, this);
- return 0;
+ afr_lookup_metadata_heal_check(frame, this);
+ return 0;
unwind:
- AFR_STACK_UNWIND (lookup, frame, -1, EIO, NULL, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(lookup, frame, -1, EIO, NULL, NULL, NULL, NULL);
+ return 0;
}
int
-afr_lookup_entry_heal (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- call_frame_t *heal = NULL;
- int i = 0, first = -1;
- gf_boolean_t need_heal = _gf_false;
- struct afr_reply *replies = NULL;
- int ret = 0;
-
- local = frame->local;
- replies = local->replies;
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
-
- if ((replies[i].op_ret == -1) &&
- (replies[i].op_errno == ENODATA))
- need_heal = _gf_true;
-
- if (first == -1) {
- first = i;
- continue;
- }
-
- if (replies[i].op_ret != replies[first].op_ret) {
- need_heal = _gf_true;
- break;
- }
-
- if (gf_uuid_compare (replies[i].poststat.ia_gfid,
- replies[first].poststat.ia_gfid)) {
- need_heal = _gf_true;
- break;
- }
- }
-
- if (need_heal) {
- heal = afr_frame_create (this, NULL);
- if (!heal)
- goto metadata_heal;
-
- ret = synctask_new (this->ctx->env, afr_lookup_selfheal_wrap,
- afr_refresh_selfheal_done, heal, frame);
- if (ret) {
- AFR_STACK_DESTROY (heal);
- goto metadata_heal;
- }
- return ret;
- }
-metadata_heal:
- ret = afr_lookup_metadata_heal_check (frame, this);
+afr_lookup_entry_heal(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ call_frame_t *heal = NULL;
+ int i = 0, first = -1;
+ gf_boolean_t name_state_mismatch = _gf_false;
+ struct afr_reply *replies = NULL;
+ int ret = 0;
+ unsigned char *par_readables = NULL;
+ unsigned char *success = NULL;
+ int32_t op_errno = 0;
+ uuid_t gfid = {0};
+
+ local = frame->local;
+ replies = local->replies;
+ priv = this->private;
+ par_readables = alloca0(priv->child_count);
+ success = alloca0(priv->child_count);
+
+ ret = afr_inode_read_subvol_get(local->loc.parent, this, par_readables,
+ NULL, NULL);
+ if (ret < 0 || AFR_COUNT(par_readables, priv->child_count) == 0) {
+ /* In this case set par_readables to all 1 so that name_heal
+ * need checks at the end of this function will flag missing
+ * entry when name state mismatches*/
+ memset(par_readables, 1, priv->child_count);
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
+
+ if (replies[i].op_ret == 0) {
+ if (gf_uuid_is_null(gfid)) {
+ gf_uuid_copy(gfid, replies[i].poststat.ia_gfid);
+ }
+ success[i] = 1;
+ } else {
+ if ((replies[i].op_errno != ENOTCONN) &&
+ (replies[i].op_errno != ENOENT) &&
+ (replies[i].op_errno != ESTALE)) {
+ op_errno = replies[i].op_errno;
+ }
+ }
- return ret;
-}
+ /*gfid is missing, needs heal*/
+ if ((replies[i].op_ret == -1) && (replies[i].op_errno == ENODATA)) {
+ goto name_heal;
+ }
+ if (first == -1) {
+ first = i;
+ continue;
+ }
-int
-afr_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode, struct iatt *buf,
- dict_t *xdata, struct iatt *postparent)
-{
- afr_local_t * local = NULL;
- int call_count = -1;
- int child_index = -1;
- GF_UNUSED int ret = 0;
- int8_t need_heal = 1;
-
- child_index = (long) cookie;
-
- local = frame->local;
-
- local->replies[child_index].valid = 1;
- local->replies[child_index].op_ret = op_ret;
- local->replies[child_index].op_errno = op_errno;
- /*
- * On revalidate lookup if the gfid-changed, afr should unwind the fop
- * with ESTALE so that a fresh lookup will be sent by the top xlator.
- * So remember it.
- */
- if (xdata && dict_get (xdata, "gfid-changed"))
- local->cont.lookup.needs_fresh_lookup = _gf_true;
+ if (replies[i].op_ret != replies[first].op_ret) {
+ name_state_mismatch = _gf_true;
+ }
- if (xdata) {
- ret = dict_get_int8 (xdata, "link-count", &need_heal);
- local->replies[child_index].need_heal = need_heal;
- } else {
- local->replies[child_index].need_heal = need_heal;
+ if (replies[i].op_ret == 0) {
+ /* Rename after this lookup may succeed if we don't do
+ * a name-heal and the destination may not have pending xattrs
+ * to indicate which name is good and which is bad so always do
+ * this heal*/
+ if (gf_uuid_compare(replies[i].poststat.ia_gfid, gfid)) {
+ goto name_heal;
+ }
}
- if (op_ret != -1) {
- local->replies[child_index].poststat = *buf;
- local->replies[child_index].postparent = *postparent;
- if (xdata)
- local->replies[child_index].xdata = dict_ref (xdata);
- }
+ }
- call_count = afr_frame_return (frame);
- if (call_count == 0) {
- afr_set_need_heal (this, local);
- afr_lookup_entry_heal (frame, this);
+ if (name_state_mismatch) {
+ if (!priv->quorum_count)
+ goto name_heal;
+ if (!afr_has_quorum(success, this, NULL))
+ goto name_heal;
+ if (op_errno)
+ goto name_heal;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
+ if (par_readables[i] && replies[i].op_ret < 0 &&
+ replies[i].op_errno != ENOTCONN) {
+ goto name_heal;
+ }
}
+ }
- return 0;
+ goto metadata_heal;
+
+name_heal:
+ heal = afr_frame_create(this, NULL);
+ if (!heal)
+ goto metadata_heal;
+
+ ret = synctask_new(this->ctx->env, afr_lookup_selfheal_wrap,
+ afr_refresh_selfheal_done, heal, frame);
+ if (ret) {
+ AFR_STACK_DESTROY(heal);
+ goto metadata_heal;
+ }
+ return ret;
+
+metadata_heal:
+ ret = afr_lookup_metadata_heal_check(frame, this);
+
+ return ret;
+}
+
+int
+afr_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, inode_t *inode, struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
+{
+ afr_local_t *local = NULL;
+ int call_count = -1;
+ int child_index = -1;
+ GF_UNUSED int ret = 0;
+ int8_t need_heal = 1;
+
+ child_index = (long)cookie;
+
+ local = frame->local;
+
+ local->replies[child_index].valid = 1;
+ local->replies[child_index].op_ret = op_ret;
+ local->replies[child_index].op_errno = op_errno;
+ /*
+ * On revalidate lookup if the gfid-changed, afr should unwind the fop
+ * with ESTALE so that a fresh lookup will be sent by the top xlator.
+ * So remember it.
+ */
+ if (xdata && dict_get_sizen(xdata, "gfid-changed"))
+ local->cont.lookup.needs_fresh_lookup = _gf_true;
+
+ if (xdata) {
+ ret = dict_get_int8(xdata, "link-count", &need_heal);
+ local->replies[child_index].need_heal = need_heal;
+ } else {
+ local->replies[child_index].need_heal = need_heal;
+ }
+ if (op_ret != -1) {
+ local->replies[child_index].poststat = *buf;
+ local->replies[child_index].postparent = *postparent;
+ if (xdata)
+ local->replies[child_index].xdata = dict_ref(xdata);
+ }
+
+ call_count = afr_frame_return(frame);
+ if (call_count == 0) {
+ afr_set_need_heal(this, local);
+ afr_lookup_entry_heal(frame, this);
+ }
+
+ return 0;
}
static void
-afr_discover_unwind (call_frame_t *frame, xlator_t *this)
+afr_discover_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = -1;
- int op_errno = 0;
- int read_subvol = -1;
- unsigned char *data_readable = NULL;
- unsigned char *success_replies = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int read_subvol = -1;
+ int ret = 0;
+ unsigned char *data_readable = NULL;
+ unsigned char *success_replies = NULL;
- priv = this->private;
- local = frame->local;
- data_readable = alloca0 (priv->child_count);
- success_replies = alloca0 (priv->child_count);
-
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
- if (local->replies[i].op_ret == 0) {
- success_replies[i] = 1;
- local->op_ret = 0;
- }
- }
+ priv = this->private;
+ local = frame->local;
+ data_readable = alloca0(priv->child_count);
+ success_replies = alloca0(priv->child_count);
- op_errno = afr_final_errno (frame->local, this->private);
+ afr_fill_success_replies(local, priv, success_replies);
+ if (AFR_COUNT(success_replies, priv->child_count) > 0)
+ local->op_ret = 0;
- if (local->op_ret < 0) {
- AFR_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL,
- NULL, NULL);
- return;
- }
+ if (local->op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = afr_final_errno(frame->local, this->private);
+ goto error;
+ }
- if (!afr_has_quorum (success_replies, this))
- goto unwind;
+ if (!afr_has_quorum(success_replies, this, frame))
+ goto unwind;
- afr_replies_interpret (frame, this, local->inode, NULL);
+ ret = afr_replies_interpret(frame, this, local->inode, NULL);
+ if (ret) {
+ afr_inode_need_refresh_set(local->inode, this);
+ }
- read_subvol = afr_read_subvol_decide (local->inode, this, NULL,
- data_readable);
+ read_subvol = afr_read_subvol_decide(local->inode, this, NULL,
+ data_readable);
unwind:
- afr_attempt_readsubvol_set (frame, this, success_replies, data_readable,
- &read_subvol);
- if (read_subvol == -1) {
- AFR_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
- NULL, NULL, NULL, NULL);
- return;
- }
+ afr_attempt_readsubvol_set(frame, this, success_replies, data_readable,
+ &read_subvol);
+ if (read_subvol == -1)
+ goto error;
- if (AFR_IS_ARBITER_BRICK (priv, read_subvol) && local->op_ret == 0) {
- local->op_ret = -1;
- local->op_errno = ENOTCONN;
- gf_msg_debug (this->name, 0, "Arbiter cannot be a read subvol "
- "for %s", local->loc.path);
- }
+ if (AFR_IS_ARBITER_BRICK(priv, read_subvol) && local->op_ret == 0) {
+ local->op_ret = -1;
+ local->op_errno = ENOTCONN;
+ gf_msg_debug(this->name, 0,
+ "Arbiter cannot be a read subvol "
+ "for %s",
+ local->loc.path);
+ }
+
+ AFR_STACK_UNWIND(lookup, frame, local->op_ret, local->op_errno,
+ local->inode, &local->replies[read_subvol].poststat,
+ local->replies[read_subvol].xdata,
+ &local->replies[read_subvol].postparent);
+ return;
- AFR_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
- local->inode, &local->replies[read_subvol].poststat,
- local->replies[read_subvol].xdata,
- &local->replies[read_subvol].postparent);
+error:
+ AFR_STACK_UNWIND(lookup, frame, local->op_ret, local->op_errno, NULL, NULL,
+ NULL, NULL);
}
static int
-afr_ta_id_file_check (void *opaque)
-{
- afr_private_t *priv = NULL;
- xlator_t *this = NULL;
- loc_t loc = {0, };
- struct iatt stbuf = {0,};
- dict_t *dict = NULL;
- uuid_t gfid = {0,};
- fd_t *fd = NULL;
- int ret = 0;
-
- this = opaque;
- priv = this->private;
-
- ret = afr_fill_ta_loc (this, &loc);
- if (ret)
- goto out;
-
- ret = syncop_lookup (priv->children[THIN_ARBITER_BRICK_INDEX], &loc,
- &stbuf, 0, 0, 0);
- if (ret == 0) {
- goto out;
- } else if (ret == -ENOENT) {
- fd = fd_create (loc.inode, getpid());
- if (!fd)
- goto out;
- dict = dict_new ();
- if (!dict)
- goto out;
- gf_uuid_generate (gfid);
- ret = dict_set_gfuuid (dict, "gfid-req", gfid, true);
- ret = syncop_create (priv->children[THIN_ARBITER_BRICK_INDEX],
- &loc, O_RDWR, 0664, fd, &stbuf, dict,
- NULL);
- }
+afr_ta_id_file_check(void *opaque)
+{
+ afr_private_t *priv = NULL;
+ xlator_t *this = NULL;
+ loc_t loc = {
+ 0,
+ };
+ struct iatt stbuf = {
+ 0,
+ };
+ dict_t *dict = NULL;
+ uuid_t gfid = {
+ 0,
+ };
+ fd_t *fd = NULL;
+ int ret = 0;
+
+ this = opaque;
+ priv = this->private;
+
+ ret = afr_fill_ta_loc(this, &loc, _gf_false);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to populate thin-arbiter loc for: %s.", loc.name);
+ goto out;
+ }
+
+ ret = syncop_lookup(priv->children[THIN_ARBITER_BRICK_INDEX], &loc, &stbuf,
+ 0, 0, 0);
+ if (ret == 0) {
+ goto out;
+ } else if (ret == -ENOENT) {
+ fd = fd_create(loc.inode, getpid());
+ if (!fd)
+ goto out;
+ dict = dict_new();
+ if (!dict)
+ goto out;
+ gf_uuid_generate(gfid);
+ ret = dict_set_gfuuid(dict, "gfid-req", gfid, true);
+ ret = syncop_create(priv->children[THIN_ARBITER_BRICK_INDEX], &loc,
+ O_RDWR, 0664, fd, &stbuf, dict, NULL);
+ }
out:
- if (ret == 0) {
- gf_uuid_copy (priv->ta_gfid, stbuf.ia_gfid);
- } else {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Failed to lookup/create thin-arbiter id file.");
- }
- if (dict)
- dict_unref (dict);
- if (fd)
- fd_unref (fd);
- loc_wipe (&loc);
+ if (ret == 0) {
+ gf_uuid_copy(priv->ta_gfid, stbuf.ia_gfid);
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to lookup/create thin-arbiter id file.");
+ }
+ if (dict)
+ dict_unref(dict);
+ if (fd)
+ fd_unref(fd);
+ loc_wipe(&loc);
- return 0;
+ return 0;
}
static int
-afr_ta_id_file_check_cbk (int ret, call_frame_t *ta_frame, void *opaque)
+afr_ta_id_file_check_cbk(int ret, call_frame_t *ta_frame, void *opaque)
{
- return 0;
+ return 0;
}
static void
-afr_discover_done (call_frame_t *frame, xlator_t *this)
+afr_discover_done(call_frame_t *frame, xlator_t *this)
{
- int ret = 0;
- afr_private_t *priv = NULL;
+ int ret = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
- if (!priv->thin_arbiter_count)
- goto unwind;
- if (!gf_uuid_is_null(priv->ta_gfid))
- goto unwind;
+ priv = this->private;
+ if (!priv->thin_arbiter_count)
+ goto unwind;
+ if (!gf_uuid_is_null(priv->ta_gfid))
+ goto unwind;
- ret = synctask_new (this->ctx->env, afr_ta_id_file_check,
- afr_ta_id_file_check_cbk, NULL, this);
- if (ret)
- goto unwind;
+ ret = synctask_new(this->ctx->env, afr_ta_id_file_check,
+ afr_ta_id_file_check_cbk, NULL, this);
+ if (ret)
+ goto unwind;
unwind:
- afr_discover_unwind (frame, this);
+ afr_discover_unwind(frame, this);
}
int
-afr_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode, struct iatt *buf,
- dict_t *xdata, struct iatt *postparent)
-{
- afr_local_t * local = NULL;
- int call_count = -1;
- int child_index = -1;
- GF_UNUSED int ret = 0;
- int8_t need_heal = 1;
-
- child_index = (long) cookie;
-
- local = frame->local;
-
- local->replies[child_index].valid = 1;
- local->replies[child_index].op_ret = op_ret;
- local->replies[child_index].op_errno = op_errno;
- if (op_ret != -1) {
- local->replies[child_index].poststat = *buf;
- local->replies[child_index].postparent = *postparent;
- if (xdata)
- local->replies[child_index].xdata = dict_ref (xdata);
- }
-
- if (local->do_discovery && (op_ret == 0))
- afr_attempt_local_discovery (this, child_index);
-
- if (xdata) {
- ret = dict_get_int8 (xdata, "link-count", &need_heal);
- local->replies[child_index].need_heal = need_heal;
- } else {
- local->replies[child_index].need_heal = need_heal;
- }
-
- call_count = afr_frame_return (frame);
- if (call_count == 0) {
- afr_set_need_heal (this, local);
- afr_lookup_metadata_heal_check (frame, this);
- }
+afr_discover_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, inode_t *inode, struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
+{
+ afr_local_t *local = NULL;
+ int call_count = -1;
+ int child_index = -1;
+ GF_UNUSED int ret = 0;
+ int8_t need_heal = 1;
+
+ child_index = (long)cookie;
+
+ local = frame->local;
+
+ local->replies[child_index].valid = 1;
+ local->replies[child_index].op_ret = op_ret;
+ local->replies[child_index].op_errno = op_errno;
+ if (op_ret != -1) {
+ local->replies[child_index].poststat = *buf;
+ local->replies[child_index].postparent = *postparent;
+ if (xdata)
+ local->replies[child_index].xdata = dict_ref(xdata);
+ }
+
+ if (local->do_discovery && (op_ret == 0))
+ afr_attempt_local_discovery(this, child_index);
+
+ if (xdata) {
+ ret = dict_get_int8(xdata, "link-count", &need_heal);
+ local->replies[child_index].need_heal = need_heal;
+ } else {
+ local->replies[child_index].need_heal = need_heal;
+ }
+
+ call_count = afr_frame_return(frame);
+ if (call_count == 0) {
+ afr_set_need_heal(this, local);
+ afr_lookup_metadata_heal_check(frame, this);
+ }
- return 0;
+ return 0;
}
-
int
-afr_discover_do (call_frame_t *frame, xlator_t *this, int err)
-{
- int ret = 0;
- int i = 0;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
-
- local = frame->local;
- priv = this->private;
-
- if (err) {
- local->op_errno = -err;
- goto out;
- }
-
- call_count = local->call_count = AFR_COUNT (local->child_up,
- priv->child_count);
-
- ret = afr_lookup_xattr_req_prepare (local, this, local->xattr_req,
- &local->loc);
- if (ret) {
- local->op_errno = -ret;
- goto out;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_discover_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->lookup,
- &local->loc, local->xattr_req);
- if (!--call_count)
- break;
- }
+afr_discover_do(call_frame_t *frame, xlator_t *this, int err)
+{
+ int ret = 0;
+ int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ if (err) {
+ local->op_errno = err;
+ goto out;
+ }
+
+ call_count = local->call_count = AFR_COUNT(local->child_up,
+ priv->child_count);
+
+ ret = afr_lookup_xattr_req_prepare(local, this, local->xattr_req,
+ &local->loc);
+ if (ret) {
+ local->op_errno = -ret;
+ goto out;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE(
+ frame, afr_discover_cbk, (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->lookup, &local->loc, local->xattr_req);
+ if (!--call_count)
+ break;
}
+ }
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (lookup, frame, -1, local->op_errno, 0, 0, 0, 0);
- return 0;
+ AFR_STACK_UNWIND(lookup, frame, -1, local->op_errno, 0, 0, 0, 0);
+ return 0;
}
-
int
-afr_discover (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr_req)
+afr_discover(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr_req)
{
- int op_errno = ENOMEM;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int event = 0;
+ int op_errno = ENOMEM;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int event = 0;
- priv = this->private;
+ priv = this->private;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- if (!local->call_count) {
- op_errno = ENOTCONN;
- goto out;
- }
+ if (!local->call_count) {
+ op_errno = ENOTCONN;
+ goto out;
+ }
- if (__is_root_gfid (loc->inode->gfid)) {
- if (!this->itable)
- this->itable = loc->inode->table;
- if (!priv->root_inode)
- priv->root_inode = inode_ref (loc->inode);
+ if (__is_root_gfid(loc->inode->gfid)) {
+ if (!priv->root_inode)
+ priv->root_inode = inode_ref(loc->inode);
- if (priv->choose_local && !priv->did_discovery) {
- /* Logic to detect which subvolumes of AFR are
- local, in order to prefer them for reads
- */
- local->do_discovery = _gf_true;
- priv->did_discovery = _gf_true;
- }
- }
+ if (priv->choose_local && !priv->did_discovery) {
+ /* Logic to detect which subvolumes of AFR are
+ local, in order to prefer them for reads
+ */
+ local->do_discovery = _gf_true;
+ priv->did_discovery = _gf_true;
+ }
+ }
- local->op = GF_FOP_LOOKUP;
+ local->op = GF_FOP_LOOKUP;
- loc_copy (&local->loc, loc);
+ loc_copy(&local->loc, loc);
- local->inode = inode_ref (loc->inode);
+ local->inode = inode_ref(loc->inode);
- if (xattr_req)
- /* If xattr_req was null, afr_lookup_xattr_req_prepare() will
- allocate one for us */
- local->xattr_req = dict_ref (xattr_req);
+ if (xattr_req) {
+ /* If xattr_req was null, afr_lookup_xattr_req_prepare() will
+ allocate one for us */
+ local->xattr_req = dict_copy_with_ref(xattr_req, NULL);
+ if (!local->xattr_req) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+ }
- if (gf_uuid_is_null (loc->inode->gfid)) {
- afr_discover_do (frame, this, 0);
- return 0;
- }
+ if (gf_uuid_is_null(loc->inode->gfid)) {
+ afr_discover_do(frame, this, 0);
+ return 0;
+ }
- afr_read_subvol_get (loc->inode, this, NULL, NULL, &event,
- AFR_DATA_TRANSACTION, NULL);
+ afr_read_subvol_get(loc->inode, this, NULL, NULL, &event,
+ AFR_DATA_TRANSACTION, NULL);
- if (afr_is_inode_refresh_reqd (loc->inode, this, event,
- local->event_generation))
- afr_inode_refresh (frame, this, loc->inode, NULL,
- afr_discover_do);
- else
- afr_discover_do (frame, this, 0);
+ afr_discover_do(frame, this, 0);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ return 0;
}
-
int
-afr_lookup_do (call_frame_t *frame, xlator_t *this, int err)
-{
- int ret = 0;
- int i = 0;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
-
- local = frame->local;
- priv = this->private;
-
- if (err < 0) {
- local->op_errno = -err;
- ret = -1;
- goto out;
- }
-
- call_count = local->call_count = AFR_COUNT (local->child_up,
- priv->child_count);
-
- ret = afr_lookup_xattr_req_prepare (local, this, local->xattr_req,
- &local->loc);
- if (ret) {
- local->op_errno = -ret;
- ret = -1;
- goto out;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_lookup_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->lookup,
- &local->loc, local->xattr_req);
- if (!--call_count)
- break;
- }
+afr_lookup_do(call_frame_t *frame, xlator_t *this, int err)
+{
+ int ret = 0;
+ int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ if (err < 0) {
+ local->op_errno = err;
+ goto out;
+ }
+
+ call_count = local->call_count = AFR_COUNT(local->child_up,
+ priv->child_count);
+
+ ret = afr_lookup_xattr_req_prepare(local, this, local->xattr_req,
+ &local->loc);
+ if (ret) {
+ local->op_errno = -ret;
+ goto out;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE(
+ frame, afr_lookup_cbk, (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->lookup, &local->loc, local->xattr_req);
+ if (!--call_count)
+ break;
}
- return 0;
+ }
+ return 0;
out:
- AFR_STACK_UNWIND (lookup, frame, -1, local->op_errno, 0, 0, 0, 0);
- return 0;
+ AFR_STACK_UNWIND(lookup, frame, -1, local->op_errno, 0, 0, 0, 0);
+ return 0;
}
/*
@@ -3284,1618 +4000,1722 @@ out:
*/
int
-afr_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr_req)
-{
- afr_local_t *local = NULL;
- int32_t op_errno = 0;
- int event = 0;
- int ret = 0;
-
- if (loc_is_nameless (loc)) {
- if (xattr_req)
- dict_del (xattr_req, "gfid-req");
- afr_discover (frame, this, loc, xattr_req);
- return 0;
- }
-
- if (__is_root_gfid (loc->parent->gfid)) {
- if (!strcmp (loc->name, GF_REPLICATE_TRASH_DIR)) {
- op_errno = EPERM;
- goto out;
- }
- }
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- if (!local->call_count) {
- op_errno = ENOTCONN;
- goto out;
- }
+afr_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr_req)
+{
+ afr_local_t *local = NULL;
+ int32_t op_errno = 0;
+ int event = 0;
+ int ret = 0;
- local->op = GF_FOP_LOOKUP;
+ if (loc_is_nameless(loc)) {
+ if (xattr_req)
+ dict_del_sizen(xattr_req, "gfid-req");
+ afr_discover(frame, this, loc, xattr_req);
+ return 0;
+ }
- loc_copy (&local->loc, loc);
+ if (afr_is_private_directory(this->private, loc->parent->gfid, loc->name,
+ frame->root->pid)) {
+ op_errno = EPERM;
+ goto out;
+ }
- local->inode = inode_ref (loc->inode);
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- if (xattr_req) {
- /* If xattr_req was null, afr_lookup_xattr_req_prepare() will
- allocate one for us */
- local->xattr_req = dict_copy_with_ref (xattr_req, NULL);
- if (!local->xattr_req) {
- op_errno = ENOMEM;
- goto out;
- }
- ret = dict_get_gfuuid (local->xattr_req, "gfid-req",
- &local->cont.lookup.gfid_req);
- if (ret == 0) {
- dict_del (local->xattr_req, "gfid-req");
- }
+ if (!local->call_count) {
+ op_errno = ENOTCONN;
+ goto out;
+ }
+
+ local->op = GF_FOP_LOOKUP;
+
+ loc_copy(&local->loc, loc);
+
+ local->inode = inode_ref(loc->inode);
+
+ if (xattr_req) {
+ /* If xattr_req was null, afr_lookup_xattr_req_prepare() will
+ allocate one for us */
+ local->xattr_req = dict_copy_with_ref(xattr_req, NULL);
+ if (!local->xattr_req) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+ ret = dict_get_gfuuid(local->xattr_req, "gfid-req",
+ &local->cont.lookup.gfid_req);
+ if (ret == 0) {
+ dict_del_sizen(local->xattr_req, "gfid-req");
}
+ }
- afr_read_subvol_get (loc->parent, this, NULL, NULL, &event,
- AFR_DATA_TRANSACTION, NULL);
+ afr_read_subvol_get(loc->parent, this, NULL, NULL, &event,
+ AFR_DATA_TRANSACTION, NULL);
- if (afr_is_inode_refresh_reqd (loc->inode, this, event,
- local->event_generation))
- afr_inode_refresh (frame, this, loc->parent, NULL,
- afr_lookup_do);
- else
- afr_lookup_do (frame, this, 0);
+ afr_lookup_do(frame, this, 0);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ AFR_STACK_UNWIND(lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
- return 0;
+ return 0;
}
void
-_afr_cleanup_fd_ctx (afr_fd_ctx_t *fd_ctx)
+_afr_cleanup_fd_ctx(xlator_t *this, afr_fd_ctx_t *fd_ctx)
{
- GF_FREE (fd_ctx->opened_on);
- GF_FREE (fd_ctx);
- return;
+ afr_private_t *priv = this->private;
+
+ if (fd_ctx->lk_heal_info) {
+ LOCK(&priv->lock);
+ {
+ list_del(&fd_ctx->lk_heal_info->pos);
+ }
+ afr_lk_heal_info_cleanup(fd_ctx->lk_heal_info);
+ fd_ctx->lk_heal_info = NULL;
+ }
+ GF_FREE(fd_ctx->opened_on);
+ GF_FREE(fd_ctx);
+ return;
}
int
-afr_cleanup_fd_ctx (xlator_t *this, fd_t *fd)
+afr_cleanup_fd_ctx(xlator_t *this, fd_t *fd)
{
- uint64_t ctx = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
- int ret = 0;
+ uint64_t ctx = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int ret = 0;
- ret = fd_ctx_get (fd, this, &ctx);
- if (ret < 0)
- goto out;
+ ret = fd_ctx_get(fd, this, &ctx);
+ if (ret < 0)
+ goto out;
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+ fd_ctx = (afr_fd_ctx_t *)(long)ctx;
- if (fd_ctx) {
- _afr_cleanup_fd_ctx (fd_ctx);
- }
+ if (fd_ctx) {
+ _afr_cleanup_fd_ctx(this, fd_ctx);
+ }
out:
- return 0;
+ return 0;
}
int
-afr_release (xlator_t *this, fd_t *fd)
+afr_release(xlator_t *this, fd_t *fd)
{
- afr_cleanup_fd_ctx (this, fd);
+ afr_cleanup_fd_ctx(this, fd);
- return 0;
+ return 0;
}
afr_fd_ctx_t *
-__afr_fd_ctx_get (fd_t *fd, xlator_t *this)
+__afr_fd_ctx_get(fd_t *fd, xlator_t *this)
{
- uint64_t ctx = 0;
- int ret = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
+ uint64_t ctx = 0;
+ int ret = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
- ret = __fd_ctx_get (fd, this, &ctx);
+ ret = __fd_ctx_get(fd, this, &ctx);
- if (ret < 0) {
- ret = __afr_fd_ctx_set (this, fd);
- if (ret < 0)
- goto out;
+ if (ret < 0) {
+ ret = __afr_fd_ctx_set(this, fd);
+ if (ret < 0)
+ goto out;
- ret = __fd_ctx_get (fd, this, &ctx);
- if (ret < 0)
- goto out;
- }
+ ret = __fd_ctx_get(fd, this, &ctx);
+ if (ret < 0)
+ goto out;
+ }
- fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+ fd_ctx = (afr_fd_ctx_t *)(long)ctx;
out:
- return fd_ctx;
+ return fd_ctx;
}
-
afr_fd_ctx_t *
-afr_fd_ctx_get (fd_t *fd, xlator_t *this)
+afr_fd_ctx_get(fd_t *fd, xlator_t *this)
{
- afr_fd_ctx_t *fd_ctx = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
- LOCK(&fd->lock);
- {
- fd_ctx = __afr_fd_ctx_get (fd, this);
- }
- UNLOCK(&fd->lock);
+ LOCK(&fd->lock);
+ {
+ fd_ctx = __afr_fd_ctx_get(fd, this);
+ }
+ UNLOCK(&fd->lock);
- return fd_ctx;
+ return fd_ctx;
}
-
int
-__afr_fd_ctx_set (xlator_t *this, fd_t *fd)
+__afr_fd_ctx_set(xlator_t *this, fd_t *fd)
{
- afr_private_t * priv = NULL;
- int ret = -1;
- uint64_t ctx = 0;
- afr_fd_ctx_t * fd_ctx = NULL;
- int i = 0;
+ afr_private_t *priv = NULL;
+ int ret = -1;
+ uint64_t ctx = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int i = 0;
- VALIDATE_OR_GOTO (this->private, out);
- VALIDATE_OR_GOTO (fd, out);
+ VALIDATE_OR_GOTO(this->private, out);
+ VALIDATE_OR_GOTO(fd, out);
- priv = this->private;
+ priv = this->private;
- ret = __fd_ctx_get (fd, this, &ctx);
+ ret = __fd_ctx_get(fd, this, &ctx);
- if (ret == 0)
- goto out;
+ if (ret == 0)
+ goto out;
- fd_ctx = GF_CALLOC (1, sizeof (afr_fd_ctx_t),
- gf_afr_mt_afr_fd_ctx_t);
- if (!fd_ctx) {
- ret = -ENOMEM;
- goto out;
- }
+ fd_ctx = GF_CALLOC(1, sizeof(afr_fd_ctx_t), gf_afr_mt_afr_fd_ctx_t);
+ if (!fd_ctx) {
+ ret = -ENOMEM;
+ goto out;
+ }
- fd_ctx->opened_on = GF_CALLOC (sizeof (*fd_ctx->opened_on),
- priv->child_count,
- gf_afr_mt_int32_t);
- if (!fd_ctx->opened_on) {
- ret = -ENOMEM;
- goto out;
- }
+ fd_ctx->opened_on = GF_CALLOC(sizeof(*fd_ctx->opened_on), priv->child_count,
+ gf_afr_mt_int32_t);
+ if (!fd_ctx->opened_on) {
+ ret = -ENOMEM;
+ goto out;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (fd_is_anonymous (fd))
- fd_ctx->opened_on[i] = AFR_FD_OPENED;
- else
- fd_ctx->opened_on[i] = AFR_FD_NOT_OPENED;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (fd_is_anonymous(fd))
+ fd_ctx->opened_on[i] = AFR_FD_OPENED;
+ else
+ fd_ctx->opened_on[i] = AFR_FD_NOT_OPENED;
+ }
- fd_ctx->readdir_subvol = -1;
+ fd_ctx->readdir_subvol = -1;
+ fd_ctx->lk_heal_info = NULL;
- ret = __fd_ctx_set (fd, this, (uint64_t)(long) fd_ctx);
- if (ret)
- gf_msg_debug (this->name, 0,
- "failed to set fd ctx (%p)", fd);
+ ret = __fd_ctx_set(fd, this, (uint64_t)(long)fd_ctx);
+ if (ret)
+ gf_msg_debug(this->name, 0, "failed to set fd ctx (%p)", fd);
out:
- if (ret && fd_ctx)
- _afr_cleanup_fd_ctx (fd_ctx);
- return ret;
+ if (ret && fd_ctx)
+ _afr_cleanup_fd_ctx(this, fd_ctx);
+ return ret;
}
-
/* {{{ flush */
int
-afr_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_flush_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int call_count = -1;
+ afr_local_t *local = NULL;
+ int call_count = -1;
- local = frame->local;
+ local = frame->local;
- LOCK (&frame->lock);
- {
- if (op_ret != -1) {
- local->op_ret = op_ret;
- if (!local->xdata_rsp && xdata)
- local->xdata_rsp = dict_ref (xdata);
- } else {
- local->op_errno = op_errno;
- }
+ LOCK(&frame->lock);
+ {
+ if (op_ret != -1) {
+ local->op_ret = op_ret;
+ if (!local->xdata_rsp && xdata)
+ local->xdata_rsp = dict_ref(xdata);
+ } else {
+ local->op_errno = op_errno;
}
- UNLOCK (&frame->lock);
-
- call_count = afr_frame_return (frame);
+ call_count = --local->call_count;
+ }
+ UNLOCK(&frame->lock);
- if (call_count == 0)
- AFR_STACK_UNWIND (flush, frame, local->op_ret,
- local->op_errno, local->xdata_rsp);
+ if (call_count == 0)
+ AFR_STACK_UNWIND(flush, frame, local->op_ret, local->op_errno,
+ local->xdata_rsp);
- return 0;
+ return 0;
}
static int
-afr_flush_wrapper (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
-{
- int i = 0;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = -1;
-
- priv = this->private;
- local = frame->local;
- call_count = local->call_count;
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_flush_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->flush,
- local->fd, xdata);
- if (!--call_count)
- break;
-
- }
+afr_flush_wrapper(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+{
+ int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+
+ priv = this->private;
+ local = frame->local;
+ call_count = local->call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE(frame, afr_flush_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->flush,
+ local->fd, xdata);
+ if (!--call_count)
+ break;
}
+ }
- return 0;
+ return 0;
}
-afr_local_t*
-afr_wakeup_same_fd_delayed_op (xlator_t *this, afr_lock_t *lock, fd_t *fd)
+afr_local_t *
+afr_wakeup_same_fd_delayed_op(xlator_t *this, afr_lock_t *lock, fd_t *fd)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- if (lock->delay_timer) {
- local = list_entry(lock->post_op.next, afr_local_t,
- transaction.owner_list);
- if (fd == local->fd) {
- if (gf_timer_call_cancel (this->ctx,
- lock->delay_timer)) {
- local = NULL;
- } else {
- lock->delay_timer = NULL;
- }
- } else {
- local = NULL;
- }
+ if (lock->delay_timer) {
+ local = list_entry(lock->post_op.next, afr_local_t,
+ transaction.owner_list);
+ if (fd == local->fd) {
+ if (gf_timer_call_cancel(this->ctx, lock->delay_timer)) {
+ local = NULL;
+ } else {
+ lock->delay_timer = NULL;
+ }
+ } else {
+ local = NULL;
}
+ }
- return local;
+ return local;
}
void
-afr_delayed_changelog_wake_resume (xlator_t *this, inode_t *inode,
- call_stub_t *stub)
-{
- afr_inode_ctx_t *ctx = NULL;
- afr_lock_t *lock = NULL;
- afr_local_t *metadata_local = NULL;
- afr_local_t *data_local = NULL;
- LOCK (&inode->lock);
- {
- (void)__afr_inode_ctx_get (this, inode, &ctx);
- lock = &ctx->lock[AFR_DATA_TRANSACTION];
- data_local = afr_wakeup_same_fd_delayed_op (this, lock,
- stub->args.fd);
- lock = &ctx->lock[AFR_METADATA_TRANSACTION];
- metadata_local = afr_wakeup_same_fd_delayed_op (this, lock,
- stub->args.fd);
- }
- UNLOCK (&inode->lock);
-
- if (data_local) {
- data_local->transaction.resume_stub = stub;
- } else if (metadata_local) {
- metadata_local->transaction.resume_stub = stub;
- } else {
- call_resume (stub);
- }
- if (data_local) {
- afr_delayed_changelog_wake_up_cbk (data_local);
- }
- if (metadata_local) {
- afr_delayed_changelog_wake_up_cbk (metadata_local);
- }
+afr_delayed_changelog_wake_resume(xlator_t *this, inode_t *inode,
+ call_stub_t *stub)
+{
+ afr_inode_ctx_t *ctx = NULL;
+ afr_lock_t *lock = NULL;
+ afr_local_t *metadata_local = NULL;
+ afr_local_t *data_local = NULL;
+ LOCK(&inode->lock);
+ {
+ (void)__afr_inode_ctx_get(this, inode, &ctx);
+ lock = &ctx->lock[AFR_DATA_TRANSACTION];
+ data_local = afr_wakeup_same_fd_delayed_op(this, lock, stub->args.fd);
+ lock = &ctx->lock[AFR_METADATA_TRANSACTION];
+ metadata_local = afr_wakeup_same_fd_delayed_op(this, lock,
+ stub->args.fd);
+ }
+ UNLOCK(&inode->lock);
+
+ if (data_local) {
+ data_local->transaction.resume_stub = stub;
+ } else if (metadata_local) {
+ metadata_local->transaction.resume_stub = stub;
+ } else {
+ call_resume(stub);
+ }
+ if (data_local) {
+ afr_delayed_changelog_wake_up_cbk(data_local);
+ }
+ if (metadata_local) {
+ afr_delayed_changelog_wake_up_cbk(metadata_local);
+ }
}
int
-afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+afr_flush(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_stub_t *stub = NULL;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ int op_errno = ENOMEM;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_FLUSH;
- if (!afr_is_consistent_io_possible (local, this->private, &op_errno))
- goto out;
+ local->op = GF_FOP_FLUSH;
+ if (!afr_is_consistent_io_possible(local, this->private, &op_errno))
+ goto out;
- local->fd = fd_ref(fd);
+ local->fd = fd_ref(fd);
- stub = fop_flush_stub (frame, afr_flush_wrapper, fd, xdata);
- if (!stub)
- goto out;
+ stub = fop_flush_stub(frame, afr_flush_wrapper, fd, xdata);
+ if (!stub)
+ goto out;
- afr_delayed_changelog_wake_resume (this, fd->inode, stub);
+ afr_delayed_changelog_wake_resume(this, fd->inode, stub);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (flush, frame, -1, op_errno, NULL);
- return 0;
+ AFR_STACK_UNWIND(flush, frame, -1, op_errno, NULL);
+ return 0;
}
int
-afr_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_fsyncdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int call_count = -1;
+ afr_local_t *local = NULL;
+ int call_count = -1;
- local = frame->local;
+ local = frame->local;
- LOCK (&frame->lock);
- {
- if (op_ret == 0) {
- local->op_ret = 0;
- if (!local->xdata_rsp && xdata)
- local->xdata_rsp = dict_ref (xdata);
- } else {
- local->op_errno = op_errno;
- }
+ LOCK(&frame->lock);
+ {
+ if (op_ret == 0) {
+ local->op_ret = 0;
+ if (!local->xdata_rsp && xdata)
+ local->xdata_rsp = dict_ref(xdata);
+ } else {
+ local->op_errno = op_errno;
}
- UNLOCK (&frame->lock);
+ call_count = --local->call_count;
+ }
+ UNLOCK(&frame->lock);
- call_count = afr_frame_return (frame);
+ if (call_count == 0)
+ AFR_STACK_UNWIND(fsyncdir, frame, local->op_ret, local->op_errno,
+ local->xdata_rsp);
- if (call_count == 0)
- AFR_STACK_UNWIND (fsyncdir, frame, local->op_ret,
- local->op_errno, local->xdata_rsp);
-
- return 0;
+ return 0;
}
-
int
-afr_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
- dict_t *xdata)
-{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = 0;
- int32_t call_count = 0;
- int32_t op_errno = ENOMEM;
-
- priv = this->private;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- local->op = GF_FOP_FSYNCDIR;
- if (!afr_is_consistent_io_possible (local, priv, &op_errno))
- goto out;
-
- call_count = local->call_count;
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND (frame, afr_fsyncdir_cbk,
- priv->children[i],
- priv->children[i]->fops->fsyncdir,
- fd, datasync, xdata);
- if (!--call_count)
- break;
- }
+afr_fsyncdir(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+ dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
+ int32_t call_count = 0;
+ int32_t op_errno = ENOMEM;
+
+ priv = this->private;
+
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
+
+ local->op = GF_FOP_FSYNCDIR;
+ if (!afr_is_consistent_io_possible(local, priv, &op_errno))
+ goto out;
+
+ call_count = local->call_count;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND(frame, afr_fsyncdir_cbk, priv->children[i],
+ priv->children[i]->fops->fsyncdir, fd, datasync, xdata);
+ if (!--call_count)
+ break;
}
+ }
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (fsyncdir, frame, -1, op_errno, NULL);
+ AFR_STACK_UNWIND(fsyncdir, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
/* }}} */
static int
-afr_serialized_lock_wind (call_frame_t *frame, xlator_t *this);
+afr_serialized_lock_wind(call_frame_t *frame, xlator_t *this);
static gf_boolean_t
-afr_is_conflicting_lock_present (int32_t op_ret, int32_t op_errno)
+afr_is_conflicting_lock_present(int32_t op_ret, int32_t op_errno)
{
- if (op_ret == -1 && op_errno == EAGAIN)
- return _gf_true;
- return _gf_false;
+ if (op_ret == -1 && op_errno == EAGAIN)
+ return _gf_true;
+ return _gf_false;
}
static void
-afr_fop_lock_unwind (call_frame_t *frame, glusterfs_fop_t op, int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
+afr_fop_lock_unwind(call_frame_t *frame, glusterfs_fop_t op, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
- switch (op) {
+ switch (op) {
case GF_FOP_INODELK:
- AFR_STACK_UNWIND (inodelk, frame, op_ret, op_errno, xdata);
- break;
+ AFR_STACK_UNWIND(inodelk, frame, op_ret, op_errno, xdata);
+ break;
case GF_FOP_FINODELK:
- AFR_STACK_UNWIND (finodelk, frame, op_ret, op_errno, xdata);
- break;
+ AFR_STACK_UNWIND(finodelk, frame, op_ret, op_errno, xdata);
+ break;
case GF_FOP_ENTRYLK:
- AFR_STACK_UNWIND (entrylk, frame, op_ret, op_errno, xdata);
- break;
+ AFR_STACK_UNWIND(entrylk, frame, op_ret, op_errno, xdata);
+ break;
case GF_FOP_FENTRYLK:
- AFR_STACK_UNWIND (fentrylk, frame, op_ret, op_errno, xdata);
- break;
+ AFR_STACK_UNWIND(fentrylk, frame, op_ret, op_errno, xdata);
+ break;
default:
- break;
- }
+ break;
+ }
}
static void
-afr_fop_lock_wind (call_frame_t *frame, xlator_t *this, int child_index,
- int32_t (*lock_cbk) (call_frame_t *, void *, xlator_t *,
- int32_t, int32_t, dict_t *))
+afr_fop_lock_wind(call_frame_t *frame, xlator_t *this, int child_index,
+ int32_t (*lock_cbk)(call_frame_t *, void *, xlator_t *,
+ int32_t, int32_t, dict_t *))
{
- afr_local_t *local = frame->local;
- afr_private_t *priv = this->private;
- int i = child_index;
+ afr_local_t *local = frame->local;
+ afr_private_t *priv = this->private;
+ int i = child_index;
- switch (local->op) {
+ switch (local->op) {
case GF_FOP_INODELK:
- STACK_WIND_COOKIE (frame, lock_cbk, (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->inodelk,
- (const char *)local->cont.inodelk.volume,
- &local->loc, local->cont.inodelk.cmd,
- &local->cont.inodelk.flock,
- local->cont.inodelk.xdata);
- break;
+ STACK_WIND_COOKIE(
+ frame, lock_cbk, (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->inodelk,
+ (const char *)local->cont.inodelk.volume, &local->loc,
+ local->cont.inodelk.cmd, &local->cont.inodelk.flock,
+ local->cont.inodelk.xdata);
+ break;
case GF_FOP_FINODELK:
- STACK_WIND_COOKIE (frame, lock_cbk, (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->finodelk,
- (const char *)local->cont.inodelk.volume,
- local->fd, local->cont.inodelk.cmd,
- &local->cont.inodelk.flock,
- local->cont.inodelk.xdata);
- break;
+ STACK_WIND_COOKIE(
+ frame, lock_cbk, (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->finodelk,
+ (const char *)local->cont.inodelk.volume, local->fd,
+ local->cont.inodelk.cmd, &local->cont.inodelk.flock,
+ local->cont.inodelk.xdata);
+ break;
case GF_FOP_ENTRYLK:
- STACK_WIND_COOKIE (frame, lock_cbk, (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->entrylk,
- local->cont.entrylk.volume, &local->loc,
- local->cont.entrylk.basename,
- local->cont.entrylk.cmd,
- local->cont.entrylk.type,
- local->cont.entrylk.xdata);
- break;
+ STACK_WIND_COOKIE(
+ frame, lock_cbk, (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->entrylk, local->cont.entrylk.volume,
+ &local->loc, local->cont.entrylk.basename,
+ local->cont.entrylk.cmd, local->cont.entrylk.type,
+ local->cont.entrylk.xdata);
+ break;
case GF_FOP_FENTRYLK:
- STACK_WIND_COOKIE (frame, lock_cbk, (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->fentrylk,
- local->cont.entrylk.volume, local->fd,
- local->cont.entrylk.basename,
- local->cont.entrylk.cmd,
- local->cont.entrylk.type,
- local->cont.entrylk.xdata);
- break;
+ STACK_WIND_COOKIE(
+ frame, lock_cbk, (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->fentrylk, local->cont.entrylk.volume,
+ local->fd, local->cont.entrylk.basename,
+ local->cont.entrylk.cmd, local->cont.entrylk.type,
+ local->cont.entrylk.xdata);
+ break;
default:
- break;
- }
+ break;
+ }
}
void
-afr_fop_lock_proceed (call_frame_t *frame)
+afr_fop_lock_proceed(call_frame_t *frame)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = frame->this->private;
-
- if (local->fop_lock_state != AFR_FOP_LOCK_PARALLEL) {
- afr_fop_lock_unwind (frame, local->op, local->op_ret,
- local->op_errno, local->xdata_rsp);
- return;
- }
- /* At least one child is up */
- /*
- * Non-blocking locks also need to be serialized. Otherwise there is
- * a chance that both the mounts which issued same non-blocking inodelk
- * may endup not acquiring the lock on any-brick.
- * Ex: Mount1 and Mount2
- * request for full length lock on file f1. Mount1 afr may acquire the
- * partial lock on brick-1 and may not acquire the lock on brick-2
- * because Mount2 already got the lock on brick-2, vice versa. Since
- * both the mounts only got partial locks, afr treats them as failure in
- * gaining the locks and unwinds with EAGAIN errno.
- */
- local->op_ret = -1;
- local->op_errno = EUCLEAN;
- local->fop_lock_state = AFR_FOP_LOCK_SERIAL;
- afr_local_replies_wipe (local, priv);
- if (local->xdata_rsp)
- dict_unref (local->xdata_rsp);
- local->xdata_rsp = NULL;
- switch (local->op) {
+ local = frame->local;
+ priv = frame->this->private;
+
+ if (local->fop_lock_state != AFR_FOP_LOCK_PARALLEL) {
+ afr_fop_lock_unwind(frame, local->op, local->op_ret, local->op_errno,
+ local->xdata_rsp);
+ return;
+ }
+ /* At least one child is up */
+ /*
+ * Non-blocking locks also need to be serialized. Otherwise there is
+ * a chance that both the mounts which issued same non-blocking inodelk
+ * may endup not acquiring the lock on any-brick.
+ * Ex: Mount1 and Mount2
+ * request for full length lock on file f1. Mount1 afr may acquire the
+ * partial lock on brick-1 and may not acquire the lock on brick-2
+ * because Mount2 already got the lock on brick-2, vice versa. Since
+ * both the mounts only got partial locks, afr treats them as failure in
+ * gaining the locks and unwinds with EAGAIN errno.
+ */
+ local->op_ret = -1;
+ local->op_errno = EUCLEAN;
+ local->fop_lock_state = AFR_FOP_LOCK_SERIAL;
+ afr_local_replies_wipe(local, priv);
+ if (local->xdata_rsp)
+ dict_unref(local->xdata_rsp);
+ local->xdata_rsp = NULL;
+ switch (local->op) {
case GF_FOP_INODELK:
case GF_FOP_FINODELK:
- local->cont.inodelk.cmd = local->cont.inodelk.in_cmd;
- local->cont.inodelk.flock = local->cont.inodelk.in_flock;
- if (local->cont.inodelk.xdata)
- dict_unref (local->cont.inodelk.xdata);
- local->cont.inodelk.xdata = NULL;
- if (local->xdata_req)
- local->cont.inodelk.xdata = dict_ref (local->xdata_req);
- break;
+ local->cont.inodelk.cmd = local->cont.inodelk.in_cmd;
+ local->cont.inodelk.flock = local->cont.inodelk.in_flock;
+ if (local->cont.inodelk.xdata)
+ dict_unref(local->cont.inodelk.xdata);
+ local->cont.inodelk.xdata = NULL;
+ if (local->xdata_req)
+ local->cont.inodelk.xdata = dict_ref(local->xdata_req);
+ break;
case GF_FOP_ENTRYLK:
case GF_FOP_FENTRYLK:
- local->cont.entrylk.cmd = local->cont.entrylk.in_cmd;
- if (local->cont.entrylk.xdata)
- dict_unref (local->cont.entrylk.xdata);
- local->cont.entrylk.xdata = NULL;
- if (local->xdata_req)
- local->cont.entrylk.xdata = dict_ref (local->xdata_req);
- break;
+ local->cont.entrylk.cmd = local->cont.entrylk.in_cmd;
+ if (local->cont.entrylk.xdata)
+ dict_unref(local->cont.entrylk.xdata);
+ local->cont.entrylk.xdata = NULL;
+ if (local->xdata_req)
+ local->cont.entrylk.xdata = dict_ref(local->xdata_req);
+ break;
default:
- break;
- }
- afr_serialized_lock_wind (frame, frame->this);
+ break;
+ }
+ afr_serialized_lock_wind(frame, frame->this);
}
static int32_t
-afr_unlock_partial_lock_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
+afr_unlock_partial_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = -1;
- int child_index = (long)cookie;
- uuid_t gfid = {0};
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ int child_index = (long)cookie;
+ uuid_t gfid = {0};
- local = frame->local;
- priv = this->private;
-
- if (op_ret < 0 && op_errno != ENOTCONN) {
- if (local->fd)
- gf_uuid_copy (gfid, local->fd->inode->gfid);
- else
- loc_gfid (&local->loc, gfid);
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- AFR_MSG_UNLOCK_FAIL,
- "%s: Failed to unlock %s on %s "
- "with lk_owner: %s", uuid_utoa (gfid),
- gf_fop_list[local->op],
- priv->children[child_index]->name,
- lkowner_utoa (&frame->root->lk_owner));
- }
-
- call_count = afr_frame_return (frame);
- if (call_count == 0)
- afr_fop_lock_proceed (frame);
+ local = frame->local;
+ priv = this->private;
- return 0;
+ if (op_ret < 0 && op_errno != ENOTCONN) {
+ if (local->fd)
+ gf_uuid_copy(gfid, local->fd->inode->gfid);
+ else
+ loc_gfid(&local->loc, gfid);
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, AFR_MSG_UNLOCK_FAIL,
+ "%s: Failed to unlock %s on %s "
+ "with lk_owner: %s",
+ uuid_utoa(gfid), gf_fop_list[local->op],
+ priv->children[child_index]->name,
+ lkowner_utoa(&frame->root->lk_owner));
+ }
+
+ call_count = afr_frame_return(frame);
+ if (call_count == 0)
+ afr_fop_lock_proceed(frame);
+
+ return 0;
}
static int32_t
-afr_unlock_locks_and_proceed (call_frame_t *frame, xlator_t *this,
+afr_unlock_locks_and_proceed(call_frame_t *frame, xlator_t *this,
int call_count)
{
- int i = 0;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
- if (call_count == 0) {
- afr_fop_lock_proceed (frame);
- goto out;
- }
+ if (call_count == 0) {
+ afr_fop_lock_proceed(frame);
+ goto out;
+ }
- local = frame->local;
- priv = this->private;
- local->call_count = call_count;
- switch (local->op) {
+ local = frame->local;
+ priv = this->private;
+ local->call_count = call_count;
+ switch (local->op) {
case GF_FOP_INODELK:
case GF_FOP_FINODELK:
- local->cont.inodelk.flock.l_type = F_UNLCK;
- local->cont.inodelk.cmd = F_SETLK;
- if (local->cont.inodelk.xdata)
- dict_unref (local->cont.inodelk.xdata);
- local->cont.inodelk.xdata = NULL;
- break;
+ local->cont.inodelk.flock.l_type = F_UNLCK;
+ local->cont.inodelk.cmd = F_SETLK;
+ if (local->cont.inodelk.xdata)
+ dict_unref(local->cont.inodelk.xdata);
+ local->cont.inodelk.xdata = NULL;
+ break;
case GF_FOP_ENTRYLK:
case GF_FOP_FENTRYLK:
- local->cont.entrylk.cmd = ENTRYLK_UNLOCK;
- if (local->cont.entrylk.xdata)
- dict_unref (local->cont.entrylk.xdata);
- local->cont.entrylk.xdata = NULL;
- break;
+ local->cont.entrylk.cmd = ENTRYLK_UNLOCK;
+ if (local->cont.entrylk.xdata)
+ dict_unref(local->cont.entrylk.xdata);
+ local->cont.entrylk.xdata = NULL;
+ break;
default:
- break;
- }
+ break;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
- if (local->replies[i].op_ret == -1)
- continue;
+ if (local->replies[i].op_ret == -1)
+ continue;
- afr_fop_lock_wind (frame, this, i, afr_unlock_partial_lock_cbk);
+ afr_fop_lock_wind(frame, this, i, afr_unlock_partial_lock_cbk);
- if (!--call_count)
- break;
- }
+ if (!--call_count)
+ break;
+ }
out:
- return 0;
+ return 0;
}
int32_t
-afr_fop_lock_done (call_frame_t *frame, xlator_t *this)
+afr_fop_lock_done(call_frame_t *frame, xlator_t *this)
{
- int i = 0;
- int lock_count = 0;
- unsigned char *success = NULL;
+ int i = 0;
+ int lock_count = 0;
+ unsigned char *success = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
- success = alloca0(priv->child_count);
+ local = frame->local;
+ priv = this->private;
+ success = alloca0(priv->child_count);
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
- if (local->replies[i].op_ret == 0) {
- lock_count++;
- success[i] = 1;
- }
-
- if (local->op_ret == -1 && local->op_errno == EAGAIN)
- continue;
-
- if ((local->replies[i].op_ret == -1) &&
- (local->replies[i].op_errno == EAGAIN)) {
- local->op_ret = -1;
- local->op_errno = EAGAIN;
- continue;
- }
+ if (local->replies[i].op_ret == 0) {
+ lock_count++;
+ success[i] = 1;
+ }
- if (local->replies[i].op_ret == 0)
- local->op_ret = 0;
+ if (local->op_ret == -1 && local->op_errno == EAGAIN)
+ continue;
- local->op_errno = local->replies[i].op_errno;
+ if ((local->replies[i].op_ret == -1) &&
+ (local->replies[i].op_errno == EAGAIN)) {
+ local->op_ret = -1;
+ local->op_errno = EAGAIN;
+ continue;
}
- if (afr_fop_lock_is_unlock (frame))
- goto unwind;
+ if (local->replies[i].op_ret == 0)
+ local->op_ret = 0;
- if (afr_is_conflicting_lock_present (local->op_ret, local->op_errno)) {
- afr_unlock_locks_and_proceed (frame, this, lock_count);
- } else if (priv->quorum_count && !afr_has_quorum (success, this)) {
- local->fop_lock_state = AFR_FOP_LOCK_QUORUM_FAILED;
- local->op_ret = -1;
- local->op_errno = afr_final_errno (local, priv);
- if (local->op_errno == 0)
- local->op_errno = afr_quorum_errno (priv);
- afr_unlock_locks_and_proceed (frame, this, lock_count);
- } else {
- goto unwind;
- }
+ local->op_errno = local->replies[i].op_errno;
+ }
- return 0;
+ if (afr_fop_lock_is_unlock(frame))
+ goto unwind;
+
+ if (afr_is_conflicting_lock_present(local->op_ret, local->op_errno)) {
+ afr_unlock_locks_and_proceed(frame, this, lock_count);
+ } else if (priv->quorum_count && !afr_has_quorum(success, this, NULL)) {
+ local->fop_lock_state = AFR_FOP_LOCK_QUORUM_FAILED;
+ local->op_ret = -1;
+ local->op_errno = afr_final_errno(local, priv);
+ if (local->op_errno == 0)
+ local->op_errno = afr_quorum_errno(priv);
+ afr_unlock_locks_and_proceed(frame, this, lock_count);
+ } else {
+ goto unwind;
+ }
+
+ return 0;
unwind:
- afr_fop_lock_unwind (frame, local->op, local->op_ret,
- local->op_errno, local->xdata_rsp);
- return 0;
+ afr_fop_lock_unwind(frame, local->op, local->op_ret, local->op_errno,
+ local->xdata_rsp);
+ return 0;
}
static int
-afr_common_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_common_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int child_index = (long)cookie;
+ afr_local_t *local = NULL;
+ int child_index = (long)cookie;
- local = frame->local;
+ local = frame->local;
- local->replies[child_index].valid = 1;
- local->replies[child_index].op_ret = op_ret;
- local->replies[child_index].op_errno = op_errno;
- if (op_ret == 0 && xdata) {
- local->replies[child_index].xdata = dict_ref (xdata);
- LOCK (&frame->lock);
- {
- if (!local->xdata_rsp)
- local->xdata_rsp = dict_ref (xdata);
- }
- UNLOCK (&frame->lock);
+ local->replies[child_index].valid = 1;
+ local->replies[child_index].op_ret = op_ret;
+ local->replies[child_index].op_errno = op_errno;
+ if (op_ret == 0 && xdata) {
+ local->replies[child_index].xdata = dict_ref(xdata);
+ LOCK(&frame->lock);
+ {
+ if (!local->xdata_rsp)
+ local->xdata_rsp = dict_ref(xdata);
}
- return 0;
+ UNLOCK(&frame->lock);
+ }
+ return 0;
}
static int32_t
-afr_serialized_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_serialized_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int child_index = (long)cookie;
- int next_child = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int child_index = (long)cookie;
+ int next_child = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- afr_common_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
+ afr_common_lock_cbk(frame, cookie, this, op_ret, op_errno, xdata);
- for (next_child = child_index + 1; next_child < priv->child_count;
- next_child++) {
- if (local->child_up[next_child])
- break;
- }
+ for (next_child = child_index + 1; next_child < priv->child_count;
+ next_child++) {
+ if (local->child_up[next_child])
+ break;
+ }
- if (afr_is_conflicting_lock_present (op_ret, op_errno) ||
- (next_child == priv->child_count)) {
- afr_fop_lock_done (frame, this);
- } else {
- afr_fop_lock_wind (frame, this, next_child,
- afr_serialized_lock_cbk);
- }
+ if (afr_is_conflicting_lock_present(op_ret, op_errno) ||
+ (next_child == priv->child_count)) {
+ afr_fop_lock_done(frame, this);
+ } else {
+ afr_fop_lock_wind(frame, this, next_child, afr_serialized_lock_cbk);
+ }
- return 0;
+ return 0;
}
static int
-afr_serialized_lock_wind (call_frame_t *frame, xlator_t *this)
+afr_serialized_lock_wind(call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- afr_fop_lock_wind (frame, this, i,
- afr_serialized_lock_cbk);
- break;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ afr_fop_lock_wind(frame, this, i, afr_serialized_lock_cbk);
+ break;
}
- return 0;
+ }
+ return 0;
}
static int32_t
-afr_parallel_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_parallel_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- int call_count = 0;
+ int call_count = 0;
- afr_common_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
+ afr_common_lock_cbk(frame, cookie, this, op_ret, op_errno, xdata);
- call_count = afr_frame_return (frame);
- if (call_count == 0)
- afr_fop_lock_done (frame, this);
+ call_count = afr_frame_return(frame);
+ if (call_count == 0)
+ afr_fop_lock_done(frame, this);
- return 0;
+ return 0;
}
static int
-afr_parallel_lock_wind (call_frame_t *frame, xlator_t *this)
+afr_parallel_lock_wind(call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int call_count = 0;
- int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int call_count = 0;
+ int i = 0;
- priv = this->private;
- local = frame->local;
- call_count = local->call_count;
+ priv = this->private;
+ local = frame->local;
+ call_count = local->call_count;
- for (i = 0; i < priv->child_count; i++) {
- if (!local->child_up[i])
- continue;
- afr_fop_lock_wind (frame, this, i, afr_parallel_lock_cbk);
- if (!--call_count)
- break;
- }
- return 0;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->child_up[i])
+ continue;
+ afr_fop_lock_wind(frame, this, i, afr_parallel_lock_cbk);
+ if (!--call_count)
+ break;
+ }
+ return 0;
}
static int
-afr_fop_handle_lock (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = frame->local;
- int op_errno = 0;
-
- if (!afr_fop_lock_is_unlock (frame)) {
- if (!afr_is_consistent_io_possible (local, this->private,
- &op_errno))
- goto out;
-
- switch (local->op) {
- case GF_FOP_INODELK:
- case GF_FOP_FINODELK:
- local->cont.inodelk.cmd = F_SETLK;
- break;
- case GF_FOP_ENTRYLK:
- case GF_FOP_FENTRYLK:
- local->cont.entrylk.cmd = ENTRYLK_LOCK_NB;
- break;
- default:
- break;
- }
+afr_fop_handle_lock(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = frame->local;
+ int op_errno = 0;
+
+ if (!afr_fop_lock_is_unlock(frame)) {
+ if (!afr_is_consistent_io_possible(local, this->private, &op_errno))
+ goto out;
+
+ switch (local->op) {
+ case GF_FOP_INODELK:
+ case GF_FOP_FINODELK:
+ local->cont.inodelk.cmd = F_SETLK;
+ break;
+ case GF_FOP_ENTRYLK:
+ case GF_FOP_FENTRYLK:
+ local->cont.entrylk.cmd = ENTRYLK_LOCK_NB;
+ break;
+ default:
+ break;
}
+ }
- if (local->xdata_req) {
- switch (local->op) {
- case GF_FOP_INODELK:
- case GF_FOP_FINODELK:
- local->cont.inodelk.xdata = dict_ref (local->xdata_req);
- break;
- case GF_FOP_ENTRYLK:
- case GF_FOP_FENTRYLK:
- local->cont.entrylk.xdata = dict_ref (local->xdata_req);
- break;
- default:
- break;
- }
+ if (local->xdata_req) {
+ switch (local->op) {
+ case GF_FOP_INODELK:
+ case GF_FOP_FINODELK:
+ local->cont.inodelk.xdata = dict_ref(local->xdata_req);
+ break;
+ case GF_FOP_ENTRYLK:
+ case GF_FOP_FENTRYLK:
+ local->cont.entrylk.xdata = dict_ref(local->xdata_req);
+ break;
+ default:
+ break;
}
+ }
- local->fop_lock_state = AFR_FOP_LOCK_PARALLEL;
- afr_parallel_lock_wind (frame, this);
+ local->fop_lock_state = AFR_FOP_LOCK_PARALLEL;
+ afr_parallel_lock_wind(frame, this);
out:
- return -op_errno;
+ return -op_errno;
}
static int32_t
-afr_handle_inodelk (call_frame_t *frame, glusterfs_fop_t fop,
- const char *volume, loc_t *loc, fd_t *fd, int32_t cmd,
- struct gf_flock *flock, dict_t *xdata)
-{
- afr_local_t *local = NULL;
- int32_t op_errno = ENOMEM;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- local->op = fop;
- if (loc)
- loc_copy (&local->loc, loc);
- if (fd)
- local->fd = fd_ref (fd);
-
- local->cont.inodelk.volume = gf_strdup (volume);
- if (!local->cont.inodelk.volume) {
- op_errno = ENOMEM;
- goto out;
- }
-
- local->cont.inodelk.in_cmd = cmd;
- local->cont.inodelk.cmd = cmd;
- local->cont.inodelk.in_flock = *flock;
- local->cont.inodelk.flock = *flock;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
-
- op_errno = -afr_fop_handle_lock (frame, frame->this);
- if (op_errno)
- goto out;
- return 0;
+afr_handle_inodelk(call_frame_t *frame, xlator_t *this, glusterfs_fop_t fop,
+ const char *volume, loc_t *loc, fd_t *fd, int32_t cmd,
+ struct gf_flock *flock, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ int32_t op_errno = ENOMEM;
+
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
+
+ local->op = fop;
+ if (loc)
+ loc_copy(&local->loc, loc);
+ if (fd && (flock->l_type != F_UNLCK)) {
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ local->fd = fd_ref(fd);
+ }
+
+ local->cont.inodelk.volume = gf_strdup(volume);
+ if (!local->cont.inodelk.volume) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ local->cont.inodelk.in_cmd = cmd;
+ local->cont.inodelk.cmd = cmd;
+ local->cont.inodelk.in_flock = *flock;
+ local->cont.inodelk.flock = *flock;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
+
+ op_errno = -afr_fop_handle_lock(frame, frame->this);
+ if (op_errno)
+ goto out;
+ return 0;
out:
- afr_fop_lock_unwind (frame, fop, -1, op_errno, NULL);
+ afr_fop_lock_unwind(frame, fop, -1, op_errno, NULL);
- return 0;
+ return 0;
}
int32_t
-afr_inodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, int32_t cmd,
- struct gf_flock *flock, dict_t *xdata)
+afr_inodelk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ int32_t cmd, struct gf_flock *flock, dict_t *xdata)
{
- afr_handle_inodelk (frame, GF_FOP_INODELK, volume, loc, NULL, cmd,
- flock, xdata);
- return 0;
+ afr_handle_inodelk(frame, this, GF_FOP_INODELK, volume, loc, NULL, cmd,
+ flock, xdata);
+ return 0;
}
int32_t
-afr_finodelk (call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
- int32_t cmd, struct gf_flock *flock, dict_t *xdata)
+afr_finodelk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ int32_t cmd, struct gf_flock *flock, dict_t *xdata)
{
- afr_handle_inodelk (frame, GF_FOP_FINODELK, volume, NULL, fd, cmd,
- flock, xdata);
- return 0;
+ afr_handle_inodelk(frame, this, GF_FOP_FINODELK, volume, NULL, fd, cmd,
+ flock, xdata);
+ return 0;
}
static int
-afr_handle_entrylk (call_frame_t *frame, glusterfs_fop_t fop,
- const char *volume, loc_t *loc, fd_t *fd,
- const char *basename, entrylk_cmd cmd,
- entrylk_type type, dict_t *xdata)
+afr_handle_entrylk(call_frame_t *frame, xlator_t *this, glusterfs_fop_t fop,
+ const char *volume, loc_t *loc, fd_t *fd,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int32_t op_errno = ENOMEM;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- local->op = fop;
- if (loc)
- loc_copy (&local->loc, loc);
- if (fd)
- local->fd = fd_ref (fd);
- local->cont.entrylk.cmd = cmd;
- local->cont.entrylk.in_cmd = cmd;
- local->cont.entrylk.type = type;
- local->cont.entrylk.volume = gf_strdup (volume);
- local->cont.entrylk.basename = gf_strdup (basename);
- if (!local->cont.entrylk.volume || !local->cont.entrylk.basename) {
- op_errno = ENOMEM;
- goto out;
- }
- if (xdata)
- local->xdata_req = dict_ref (xdata);
- op_errno = -afr_fop_handle_lock (frame, frame->this);
- if (op_errno)
- goto out;
-
- return 0;
+ afr_local_t *local = NULL;
+ int32_t op_errno = ENOMEM;
+
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
+
+ local->op = fop;
+ if (loc)
+ loc_copy(&local->loc, loc);
+ if (fd && (cmd != ENTRYLK_UNLOCK)) {
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ local->fd = fd_ref(fd);
+ }
+ local->cont.entrylk.cmd = cmd;
+ local->cont.entrylk.in_cmd = cmd;
+ local->cont.entrylk.type = type;
+ local->cont.entrylk.volume = gf_strdup(volume);
+ local->cont.entrylk.basename = gf_strdup(basename);
+ if (!local->cont.entrylk.volume || !local->cont.entrylk.basename) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
+ op_errno = -afr_fop_handle_lock(frame, frame->this);
+ if (op_errno)
+ goto out;
+
+ return 0;
out:
- afr_fop_lock_unwind (frame, fop, -1, op_errno, NULL);
- return 0;
+ afr_fop_lock_unwind(frame, fop, -1, op_errno, NULL);
+ return 0;
}
int
-afr_entrylk (call_frame_t *frame, xlator_t *this, const char *volume,
- loc_t *loc, const char *basename, entrylk_cmd cmd,
- entrylk_type type, dict_t *xdata)
+afr_entrylk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata)
{
- afr_handle_entrylk (frame, GF_FOP_ENTRYLK, volume, loc, NULL, basename,
- cmd, type, xdata);
- return 0;
+ afr_handle_entrylk(frame, this, GF_FOP_ENTRYLK, volume, loc, NULL, basename,
+ cmd, type, xdata);
+ return 0;
}
int
-afr_fentrylk (call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
- const char *basename, entrylk_cmd cmd, entrylk_type type,
- dict_t *xdata)
+afr_fentrylk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata)
{
- afr_handle_entrylk (frame, GF_FOP_FENTRYLK, volume, NULL, fd, basename,
- cmd, type, xdata);
- return 0;
+ afr_handle_entrylk(frame, this, GF_FOP_FENTRYLK, volume, NULL, fd, basename,
+ cmd, type, xdata);
+ return 0;
}
-
int
-afr_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
- int op_errno, struct statvfs *statvfs, dict_t *xdata)
+afr_statfs_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct statvfs *statvfs, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int call_count = 0;
- struct statvfs *buf = NULL;
+ afr_local_t *local = NULL;
+ int call_count = 0;
+ struct statvfs *buf = NULL;
- LOCK (&frame->lock);
- {
- local = frame->local;
-
- if (op_ret != 0) {
- local->op_errno = op_errno;
- goto unlock;
- }
-
- local->op_ret = op_ret;
-
- buf = &local->cont.statfs.buf;
- if (local->cont.statfs.buf_set) {
- if (statvfs->f_bavail < buf->f_bavail) {
- *buf = *statvfs;
- if (xdata) {
- if (local->xdata_rsp)
- dict_unref (local->xdata_rsp);
- local->xdata_rsp = dict_ref (xdata);
- }
- }
- } else {
- *buf = *statvfs;
- local->cont.statfs.buf_set = 1;
- if (xdata)
- local->xdata_rsp = dict_ref (xdata);
- }
+ local = frame->local;
+
+ LOCK(&frame->lock);
+ {
+ if (op_ret != 0) {
+ local->op_errno = op_errno;
+ goto unlock;
}
-unlock:
- UNLOCK (&frame->lock);
- call_count = afr_frame_return (frame);
+ local->op_ret = op_ret;
+
+ buf = &local->cont.statfs.buf;
+ if (local->cont.statfs.buf_set) {
+ if (statvfs->f_bavail < buf->f_bavail) {
+ *buf = *statvfs;
+ if (xdata) {
+ if (local->xdata_rsp)
+ dict_unref(local->xdata_rsp);
+ local->xdata_rsp = dict_ref(xdata);
+ }
+ }
+ } else {
+ *buf = *statvfs;
+ local->cont.statfs.buf_set = 1;
+ if (xdata)
+ local->xdata_rsp = dict_ref(xdata);
+ }
+ }
+unlock:
+ call_count = --local->call_count;
+ UNLOCK(&frame->lock);
- if (call_count == 0)
- AFR_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno,
- &local->cont.statfs.buf, local->xdata_rsp);
+ if (call_count == 0)
+ AFR_STACK_UNWIND(statfs, frame, local->op_ret, local->op_errno,
+ &local->cont.statfs.buf, local->xdata_rsp);
- return 0;
+ return 0;
}
-
int
-afr_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
-{
- afr_local_t * local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- int call_count = 0;
- int32_t op_errno = ENOMEM;
-
- priv = this->private;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- local->op = GF_FOP_STATFS;
- if (!afr_is_consistent_io_possible (local, priv, &op_errno))
- goto out;
-
- if (priv->arbiter_count == 1 && local->child_up[ARBITER_BRICK_INDEX])
- local->call_count--;
- call_count = local->call_count;
- if (!call_count) {
- op_errno = ENOTCONN;
- goto out;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- if (AFR_IS_ARBITER_BRICK(priv, i))
- continue;
- STACK_WIND (frame, afr_statfs_cbk,
- priv->children[i],
- priv->children[i]->fops->statfs,
- loc, xdata);
- if (!--call_count)
- break;
- }
+afr_statfs(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int call_count = 0;
+ int32_t op_errno = ENOMEM;
+
+ priv = this->private;
+
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
+
+ local->op = GF_FOP_STATFS;
+ if (!afr_is_consistent_io_possible(local, priv, &op_errno))
+ goto out;
+
+ if (priv->arbiter_count == 1 && local->child_up[ARBITER_BRICK_INDEX])
+ local->call_count--;
+ call_count = local->call_count;
+ if (!call_count) {
+ op_errno = ENOTCONN;
+ goto out;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ if (AFR_IS_ARBITER_BRICK(priv, i))
+ continue;
+ STACK_WIND(frame, afr_statfs_cbk, priv->children[i],
+ priv->children[i]->fops->statfs, loc, xdata);
+ if (!--call_count)
+ break;
}
+ }
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND(statfs, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
-
int32_t
-afr_lk_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
- dict_t *xdata)
+afr_lk_unlock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = this->private;
- int call_count = -1;
- int child_index = (long)cookie;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = this->private;
+ int call_count = -1;
+ int child_index = (long)cookie;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0 && op_errno != ENOTCONN && op_errno != EBADFD) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- AFR_MSG_UNLOCK_FAIL,
- "gfid=%s: unlock failed on subvolume %s "
- "with lock owner %s",
- uuid_utoa (local->fd->inode->gfid),
- priv->children[child_index]->name,
- lkowner_utoa (&frame->root->lk_owner));
- }
+ if (op_ret < 0 && op_errno != ENOTCONN && op_errno != EBADFD) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, AFR_MSG_UNLOCK_FAIL,
+ "gfid=%s: unlock failed on subvolume %s "
+ "with lock owner %s",
+ uuid_utoa(local->fd->inode->gfid),
+ priv->children[child_index]->name,
+ lkowner_utoa(&frame->root->lk_owner));
+ }
- call_count = afr_frame_return (frame);
- if (call_count == 0)
- AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
- NULL, local->xdata_rsp);
+ call_count = afr_frame_return(frame);
+ if (call_count == 0) {
+ AFR_STACK_UNWIND(lk, frame, local->op_ret, local->op_errno, NULL,
+ local->xdata_rsp);
+ }
- return 0;
+ return 0;
}
-
int32_t
-afr_lk_unlock (call_frame_t *frame, xlator_t *this)
+afr_lk_unlock(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- afr_private_t * priv = NULL;
- int i = 0;
- int call_count = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int call_count = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- call_count = afr_locked_nodes_count (local->cont.lk.locked_nodes,
- priv->child_count);
+ call_count = afr_locked_nodes_count(local->cont.lk.locked_nodes,
+ priv->child_count);
- if (call_count == 0) {
- AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
- NULL, local->xdata_rsp);
- return 0;
- }
+ if (call_count == 0) {
+ AFR_STACK_UNWIND(lk, frame, local->op_ret, local->op_errno, NULL,
+ local->xdata_rsp);
+ return 0;
+ }
- local->call_count = call_count;
+ local->call_count = call_count;
- local->cont.lk.user_flock.l_type = F_UNLCK;
+ local->cont.lk.user_flock.l_type = F_UNLCK;
- for (i = 0; i < priv->child_count; i++) {
- if (local->cont.lk.locked_nodes[i]) {
- STACK_WIND_COOKIE (frame, afr_lk_unlock_cbk,
- (void *) (long) i, priv->children[i],
- priv->children[i]->fops->lk,
- local->fd, F_SETLK,
- &local->cont.lk.user_flock, NULL);
-
- if (!--call_count)
- break;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->cont.lk.locked_nodes[i]) {
+ STACK_WIND_COOKIE(frame, afr_lk_unlock_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->lk,
+ local->fd, F_SETLK, &local->cont.lk.user_flock,
+ NULL);
+
+ if (!--call_count)
+ break;
}
+ }
- return 0;
+ return 0;
}
-
int32_t
-afr_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
- dict_t *xdata)
+afr_lk_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int child_index = -1;
-
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int child_index = -1;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- child_index = (long) cookie;
+ child_index = (long)cookie;
- afr_common_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
- if (op_ret < 0 && op_errno == EAGAIN) {
- local->op_ret = -1;
- local->op_errno = EAGAIN;
+ afr_common_lock_cbk(frame, cookie, this, op_ret, op_errno, xdata);
+ if (op_ret < 0 && op_errno == EAGAIN) {
+ local->op_ret = -1;
+ local->op_errno = EAGAIN;
- afr_lk_unlock (frame, this);
- return 0;
- }
+ afr_lk_unlock(frame, this);
+ return 0;
+ }
+
+ if (op_ret == 0) {
+ local->op_ret = 0;
+ local->op_errno = 0;
+ local->cont.lk.locked_nodes[child_index] = 1;
+ local->cont.lk.ret_flock = *lock;
+ }
+
+ child_index++;
+
+ if (child_index < priv->child_count) {
+ STACK_WIND_COOKIE(frame, afr_lk_cbk, (void *)(long)child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->lk, local->fd,
+ local->cont.lk.cmd, &local->cont.lk.user_flock,
+ local->xdata_req);
+ } else if (priv->quorum_count &&
+ !afr_has_quorum(local->cont.lk.locked_nodes, this, NULL)) {
+ local->op_ret = -1;
+ local->op_errno = afr_final_errno(local, priv);
- if (op_ret == 0) {
- local->op_ret = 0;
- local->op_errno = 0;
- local->cont.lk.locked_nodes[child_index] = 1;
- local->cont.lk.ret_flock = *lock;
- }
-
- child_index++;
-
- if (child_index < priv->child_count) {
- STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->lk,
- local->fd, local->cont.lk.cmd,
- &local->cont.lk.user_flock,
- local->xdata_req);
- } else if (priv->quorum_count &&
- !afr_has_quorum (local->cont.lk.locked_nodes, this)) {
- local->op_ret = -1;
- local->op_errno = afr_final_errno (local, priv);
-
- afr_lk_unlock (frame, this);
- } else {
- if (local->op_ret < 0)
- local->op_errno = afr_final_errno (local, priv);
+ afr_lk_unlock(frame, this);
+ } else {
+ if (local->op_ret < 0)
+ local->op_errno = afr_final_errno(local, priv);
- AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
- &local->cont.lk.ret_flock, local->xdata_rsp);
- }
+ AFR_STACK_UNWIND(lk, frame, local->op_ret, local->op_errno,
+ &local->cont.lk.ret_flock, local->xdata_rsp);
+ }
- return 0;
+ return 0;
}
int
-afr_lk (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
+afr_lk_transaction_cbk(int ret, call_frame_t *frame, void *opaque)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = 0;
- int32_t op_errno = ENOMEM;
+ return 0;
+}
- priv = this->private;
+int
+afr_lk_txn_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+ dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ int child_index = -1;
+
+ local = frame->local;
+ child_index = (long)cookie;
+ afr_common_lock_cbk(frame, cookie, this, op_ret, op_errno, xdata);
+ if (op_ret == 0) {
+ local->op_ret = 0;
+ local->op_errno = 0;
+ local->cont.lk.locked_nodes[child_index] = 1;
+ local->cont.lk.ret_flock = *lock;
+ }
+ syncbarrier_wake(&local->barrier);
+ return 0;
+}
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+int
+afr_lk_txn_unlock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+ dict_t *xdata)
+{
+ afr_local_t *local = frame->local;
+ afr_private_t *priv = this->private;
+ int child_index = (long)cookie;
- local->op = GF_FOP_LK;
- if (!afr_lk_is_unlock (cmd, flock) &&
- !afr_is_consistent_io_possible (local, priv, &op_errno))
- goto out;
+ if (op_ret < 0 && op_errno != ENOTCONN && op_errno != EBADFD) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, AFR_MSG_UNLOCK_FAIL,
+ "gfid=%s: unlock failed on subvolume %s "
+ "with lock owner %s",
+ uuid_utoa(local->fd->inode->gfid),
+ priv->children[child_index]->name,
+ lkowner_utoa(&frame->root->lk_owner));
+ }
+ return 0;
+}
+int
+afr_lk_transaction(void *opaque)
+{
+ call_frame_t *frame = NULL;
+ xlator_t *this = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ char *wind_on = NULL;
+ int op_errno = 0;
+ int i = 0;
+ int ret = 0;
+
+ frame = (call_frame_t *)opaque;
+ local = frame->local;
+ this = frame->this;
+ priv = this->private;
+ wind_on = alloca0(priv->child_count);
+
+ if (priv->arbiter_count || priv->child_count != 3) {
+ op_errno = ENOTSUP;
+ gf_msg(frame->this->name, GF_LOG_ERROR, op_errno, AFR_MSG_LK_HEAL_DOM,
+ "%s: Lock healing supported only for replica 3 volumes.",
+ uuid_utoa(local->fd->inode->gfid));
+ goto err;
+ }
+
+ op_errno = -afr_dom_lock_acquire(frame); // Released during
+ // AFR_STACK_UNWIND
+ if (op_errno != 0) {
+ goto err;
+ }
+ if (priv->quorum_count &&
+ !afr_has_quorum(local->cont.lk.dom_locked_nodes, this, NULL)) {
+ op_errno = afr_final_errno(local, priv);
+ goto err;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (priv->child_up[i] && local->cont.lk.dom_locked_nodes[i])
+ wind_on[i] = 1;
+ }
+ AFR_ONLIST(wind_on, frame, afr_lk_txn_wind_cbk, lk, local->fd,
+ local->cont.lk.cmd, &local->cont.lk.user_flock,
+ local->xdata_req);
+
+ if (priv->quorum_count &&
+ !afr_has_quorum(local->cont.lk.locked_nodes, this, NULL)) {
+ local->op_ret = -1;
+ local->op_errno = afr_final_errno(local, priv);
+ goto unlock;
+ } else {
+ if (local->cont.lk.user_flock.l_type == F_UNLCK)
+ ret = afr_remove_lock_from_saved_locks(local, this);
+ else
+ ret = afr_add_lock_to_saved_locks(frame, this);
+ if (ret) {
+ local->op_ret = -1;
+ local->op_errno = -ret;
+ goto unlock;
+ }
+ AFR_STACK_UNWIND(lk, frame, local->op_ret, local->op_errno,
+ &local->cont.lk.ret_flock, local->xdata_rsp);
+ }
- local->cont.lk.locked_nodes = GF_CALLOC (priv->child_count,
- sizeof (*local->cont.lk.locked_nodes),
- gf_afr_mt_char);
+ return 0;
- if (!local->cont.lk.locked_nodes) {
- op_errno = ENOMEM;
- goto out;
- }
+unlock:
+ local->cont.lk.user_flock.l_type = F_UNLCK;
+ AFR_ONLIST(local->cont.lk.locked_nodes, frame, afr_lk_txn_unlock_cbk, lk,
+ local->fd, F_SETLK, &local->cont.lk.user_flock, NULL);
+err:
+ AFR_STACK_UNWIND(lk, frame, -1, op_errno, NULL, NULL);
+ return -1;
+}
- local->fd = fd_ref (fd);
- local->cont.lk.cmd = cmd;
- local->cont.lk.user_flock = *flock;
- local->cont.lk.ret_flock = *flock;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+int
+afr_lk(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
+ struct gf_flock *flock, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int ret = 0;
+ int i = 0;
+ int32_t op_errno = ENOMEM;
+
+ priv = this->private;
+
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
+
+ local->op = GF_FOP_LK;
+ if (!afr_lk_is_unlock(cmd, flock)) {
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ if (!afr_is_consistent_io_possible(local, priv, &op_errno))
+ goto out;
+ }
+
+ local->cont.lk.locked_nodes = GF_CALLOC(
+ priv->child_count, sizeof(*local->cont.lk.locked_nodes),
+ gf_afr_mt_char);
+
+ if (!local->cont.lk.locked_nodes) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ local->fd = fd_ref(fd);
+ local->cont.lk.cmd = cmd;
+ local->cont.lk.user_flock = *flock;
+ local->cont.lk.ret_flock = *flock;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
+
+ if (afr_is_lock_mode_mandatory(xdata)) {
+ ret = synctask_new(this->ctx->env, afr_lk_transaction,
+ afr_lk_transaction_cbk, frame, frame);
+ if (ret) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+ return 0;
+ }
- STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) 0,
- priv->children[i],
- priv->children[i]->fops->lk,
- fd, cmd, flock, local->xdata_req);
+ STACK_WIND_COOKIE(frame, afr_lk_cbk, (void *)(long)0, priv->children[i],
+ priv->children[i]->fops->lk, fd, cmd, flock,
+ local->xdata_req);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (lk, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND(lk, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
int32_t
-afr_lease_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_lease *lease,
- dict_t *xdata)
+afr_lease_unlock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct gf_lease *lease,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int call_count = -1;
+ afr_local_t *local = NULL;
+ int call_count = -1;
- local = frame->local;
- call_count = afr_frame_return (frame);
+ local = frame->local;
+ call_count = afr_frame_return(frame);
- if (call_count == 0)
- AFR_STACK_UNWIND (lease, frame, local->op_ret, local->op_errno,
- lease, xdata);
+ if (call_count == 0)
+ AFR_STACK_UNWIND(lease, frame, local->op_ret, local->op_errno, lease,
+ xdata);
- return 0;
+ return 0;
}
int32_t
-afr_lease_unlock (call_frame_t *frame, xlator_t *this)
+afr_lease_unlock(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- int call_count = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int call_count = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- call_count = afr_locked_nodes_count (local->cont.lease.locked_nodes,
- priv->child_count);
+ call_count = afr_locked_nodes_count(local->cont.lease.locked_nodes,
+ priv->child_count);
- if (call_count == 0) {
- AFR_STACK_UNWIND (lease, frame, local->op_ret, local->op_errno,
- &local->cont.lease.ret_lease, NULL);
- return 0;
- }
+ if (call_count == 0) {
+ AFR_STACK_UNWIND(lease, frame, local->op_ret, local->op_errno,
+ &local->cont.lease.ret_lease, NULL);
+ return 0;
+ }
- local->call_count = call_count;
+ local->call_count = call_count;
- local->cont.lease.user_lease.cmd = GF_UNLK_LEASE;
+ local->cont.lease.user_lease.cmd = GF_UNLK_LEASE;
- for (i = 0; i < priv->child_count; i++) {
- if (local->cont.lease.locked_nodes[i]) {
- STACK_WIND (frame, afr_lease_unlock_cbk,
- priv->children[i],
- priv->children[i]->fops->lease,
- &local->loc, &local->cont.lease.user_lease, NULL);
-
- if (!--call_count)
- break;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->cont.lease.locked_nodes[i]) {
+ STACK_WIND(frame, afr_lease_unlock_cbk, priv->children[i],
+ priv->children[i]->fops->lease, &local->loc,
+ &local->cont.lease.user_lease, NULL);
+
+ if (!--call_count)
+ break;
}
+ }
- return 0;
+ return 0;
}
int32_t
-afr_lease_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_lease *lease,
- dict_t *xdata)
+afr_lease_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct gf_lease *lease, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int child_index = -1;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int child_index = -1;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- child_index = (long) cookie;
+ child_index = (long)cookie;
- afr_common_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
- if (op_ret < 0 && op_errno == EAGAIN) {
- local->op_ret = -1;
- local->op_errno = EAGAIN;
+ afr_common_lock_cbk(frame, cookie, this, op_ret, op_errno, xdata);
+ if (op_ret < 0 && op_errno == EAGAIN) {
+ local->op_ret = -1;
+ local->op_errno = EAGAIN;
- afr_lease_unlock (frame, this);
- return 0;
- }
+ afr_lease_unlock(frame, this);
+ return 0;
+ }
- if (op_ret == 0) {
- local->op_ret = 0;
- local->op_errno = 0;
- local->cont.lease.locked_nodes[child_index] = 1;
- local->cont.lease.ret_lease = *lease;
- }
-
- child_index++;
- if (child_index < priv->child_count) {
- STACK_WIND_COOKIE (frame, afr_lease_cbk, (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->lease,
- &local->loc, &local->cont.lease.user_lease, xdata);
- } else if (priv->quorum_count &&
- !afr_has_quorum (local->cont.lk.locked_nodes, this)) {
- local->op_ret = -1;
- local->op_errno = afr_final_errno (local, priv);
-
- afr_lease_unlock (frame, this);
- } else {
- if (local->op_ret < 0)
- local->op_errno = afr_final_errno (local, priv);
- AFR_STACK_UNWIND (lease, frame, local->op_ret, local->op_errno,
- &local->cont.lease.ret_lease, NULL);
- }
+ if (op_ret == 0) {
+ local->op_ret = 0;
+ local->op_errno = 0;
+ local->cont.lease.locked_nodes[child_index] = 1;
+ local->cont.lease.ret_lease = *lease;
+ }
+
+ child_index++;
+ if (child_index < priv->child_count) {
+ STACK_WIND_COOKIE(frame, afr_lease_cbk, (void *)(long)child_index,
+ priv->children[child_index],
+ priv->children[child_index]->fops->lease, &local->loc,
+ &local->cont.lease.user_lease, xdata);
+ } else if (priv->quorum_count &&
+ !afr_has_quorum(local->cont.lease.locked_nodes, this, NULL)) {
+ local->op_ret = -1;
+ local->op_errno = afr_final_errno(local, priv);
- return 0;
+ afr_lease_unlock(frame, this);
+ } else {
+ if (local->op_ret < 0)
+ local->op_errno = afr_final_errno(local, priv);
+ AFR_STACK_UNWIND(lease, frame, local->op_ret, local->op_errno,
+ &local->cont.lease.ret_lease, NULL);
+ }
+
+ return 0;
}
int
-afr_lease (call_frame_t *frame, xlator_t *this,
- loc_t *loc, struct gf_lease *lease, dict_t *xdata)
+afr_lease(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct gf_lease *lease, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int32_t op_errno = ENOMEM;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int32_t op_errno = ENOMEM;
- priv = this->private;
+ priv = this->private;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_LEASE;
- local->cont.lease.locked_nodes = GF_CALLOC (priv->child_count,
- sizeof (*local->cont.lease.locked_nodes),
- gf_afr_mt_char);
+ local->op = GF_FOP_LEASE;
+ local->cont.lease.locked_nodes = GF_CALLOC(
+ priv->child_count, sizeof(*local->cont.lease.locked_nodes),
+ gf_afr_mt_char);
- if (!local->cont.lease.locked_nodes) {
- op_errno = ENOMEM;
- goto out;
- }
+ if (!local->cont.lease.locked_nodes) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- loc_copy (&local->loc, loc);
- local->cont.lease.user_lease = *lease;
- local->cont.lease.ret_lease = *lease;
+ loc_copy(&local->loc, loc);
+ local->cont.lease.user_lease = *lease;
+ local->cont.lease.ret_lease = *lease;
- STACK_WIND_COOKIE (frame, afr_lease_cbk, (void *) (long) 0,
- priv->children[0],
- priv->children[0]->fops->lease,
- loc, lease, xdata);
+ STACK_WIND_COOKIE(frame, afr_lease_cbk, (void *)(long)0, priv->children[0],
+ priv->children[0]->fops->lease, loc, lease, xdata);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (lease, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND(lease, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
int
-afr_ipc_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- afr_local_t *local = NULL;
- int child_index = (long)cookie;
- int call_count = 0;
- gf_boolean_t failed = _gf_false;
- gf_boolean_t succeded = _gf_false;
- int i = 0;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- local->replies[child_index].valid = 1;
- local->replies[child_index].op_ret = op_ret;
- local->replies[child_index].op_errno = op_errno;
- if (xdata)
- local->replies[child_index].xdata = dict_ref (xdata);
-
- call_count = afr_frame_return (frame);
- if (call_count)
- goto out;
- /* If any of the subvolumes failed with other than ENOTCONN
- * return error else return success unless all the subvolumes
- * failed.
- * TODO: In case of failure, we need to unregister the xattrs
- * from the other subvolumes where it succeded (once upcall
- * fixes the Bz-1371622)*/
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
- if (local->replies[i].op_ret < 0 &&
- local->replies[i].op_errno != ENOTCONN) {
- local->op_ret = local->replies[i].op_ret;
- local->op_errno = local->replies[i].op_errno;
- if (local->xdata_rsp)
- dict_unref (local->xdata_rsp);
- local->xdata_rsp = NULL;
- if (local->replies[i].xdata) {
- local->xdata_rsp =
- dict_ref (local->replies[i].xdata);
- }
- failed = _gf_true;
- break;
- }
- if (local->replies[i].op_ret == 0) {
- succeded = _gf_true;
- local->op_ret = 0;
- local->op_errno = 0;
- if (!local->xdata_rsp && local->replies[i].xdata) {
- local->xdata_rsp =
- dict_ref (local->replies[i].xdata);
- }
- }
+afr_ipc_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ int child_index = (long)cookie;
+ int call_count = 0;
+ gf_boolean_t failed = _gf_false;
+ gf_boolean_t succeeded = _gf_false;
+ int i = 0;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ local->replies[child_index].valid = 1;
+ local->replies[child_index].op_ret = op_ret;
+ local->replies[child_index].op_errno = op_errno;
+ if (xdata)
+ local->replies[child_index].xdata = dict_ref(xdata);
+
+ call_count = afr_frame_return(frame);
+ if (call_count)
+ goto out;
+ /* If any of the subvolumes failed with other than ENOTCONN
+ * return error else return success unless all the subvolumes
+ * failed.
+ * TODO: In case of failure, we need to unregister the xattrs
+ * from the other subvolumes where it succeeded (once upcall
+ * fixes the Bz-1371622)*/
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+ if (local->replies[i].op_ret < 0 &&
+ local->replies[i].op_errno != ENOTCONN) {
+ local->op_ret = local->replies[i].op_ret;
+ local->op_errno = local->replies[i].op_errno;
+ if (local->xdata_rsp)
+ dict_unref(local->xdata_rsp);
+ local->xdata_rsp = NULL;
+ if (local->replies[i].xdata) {
+ local->xdata_rsp = dict_ref(local->replies[i].xdata);
+ }
+ failed = _gf_true;
+ break;
+ }
+ if (local->replies[i].op_ret == 0) {
+ succeeded = _gf_true;
+ local->op_ret = 0;
+ local->op_errno = 0;
+ if (!local->xdata_rsp && local->replies[i].xdata) {
+ local->xdata_rsp = dict_ref(local->replies[i].xdata);
+ }
}
+ }
- if (!succeded && !failed) {
- local->op_ret = -1;
- local->op_errno = ENOTCONN;
- }
+ if (!succeeded && !failed) {
+ local->op_ret = -1;
+ local->op_errno = ENOTCONN;
+ }
- AFR_STACK_UNWIND (ipc, frame, local->op_ret, local->op_errno,
- local->xdata_rsp);
+ AFR_STACK_UNWIND(ipc, frame, local->op_ret, local->op_errno,
+ local->xdata_rsp);
out:
- return 0;
+ return 0;
}
int
-afr_ipc (call_frame_t *frame, xlator_t *this, int32_t op, dict_t *xdata)
+afr_ipc(call_frame_t *frame, xlator_t *this, int32_t op, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int32_t op_errno = -1;
- afr_private_t *priv = NULL;
- int i = 0;
- int call_cnt = -1;
+ afr_local_t *local = NULL;
+ int32_t op_errno = -1;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int call_cnt = -1;
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
- if (op != GF_IPC_TARGET_UPCALL)
- goto wind_default;
+ if (op != GF_IPC_TARGET_UPCALL)
+ goto wind_default;
- VALIDATE_OR_GOTO (this->private, err);
- priv = this->private;
+ VALIDATE_OR_GOTO(this->private, err);
+ priv = this->private;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto err;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto err;
- call_cnt = local->call_count;
+ call_cnt = local->call_count;
- if (xdata) {
- for (i = 0; i < priv->child_count; i++) {
- if (dict_set_int8 (xdata, priv->pending_key[i], 0) < 0)
- goto err;
- }
+ if (xdata) {
+ for (i = 0; i < priv->child_count; i++) {
+ if (dict_set_int8(xdata, priv->pending_key[i], 0) < 0)
+ goto err;
}
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (!local->child_up[i])
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->child_up[i])
+ continue;
- STACK_WIND_COOKIE (frame, afr_ipc_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->ipc,
- op, xdata);
- if (!--call_cnt)
- break;
- }
- return 0;
+ STACK_WIND_COOKIE(frame, afr_ipc_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->ipc, op,
+ xdata);
+ if (!--call_cnt)
+ break;
+ }
+ return 0;
err:
- if (op_errno == -1)
- op_errno = errno;
- AFR_STACK_UNWIND (ipc, frame, -1, op_errno, NULL);
+ if (op_errno == -1)
+ op_errno = errno;
+ AFR_STACK_UNWIND(ipc, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
wind_default:
- STACK_WIND (frame, default_ipc_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->ipc, op, xdata);
- return 0;
+ STACK_WIND(frame, default_ipc_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ipc, op, xdata);
+ return 0;
}
int
-afr_forget (xlator_t *this, inode_t *inode)
+afr_forget(xlator_t *this, inode_t *inode)
{
- uint64_t ctx_int = 0;
- afr_inode_ctx_t *ctx = NULL;
-
- afr_spb_choice_timeout_cancel (this, inode);
- inode_ctx_del (inode, this, &ctx_int);
- if (!ctx_int)
- return 0;
+ uint64_t ctx_int = 0;
+ afr_inode_ctx_t *ctx = NULL;
- ctx = (afr_inode_ctx_t *)ctx_int;
- afr_inode_ctx_destroy (ctx);
+ afr_spb_choice_timeout_cancel(this, inode);
+ inode_ctx_del(inode, this, &ctx_int);
+ if (!ctx_int)
return 0;
+
+ ctx = (afr_inode_ctx_t *)(uintptr_t)ctx_int;
+ afr_inode_ctx_destroy(ctx);
+ return 0;
}
int
-afr_priv_dump (xlator_t *this)
-{
- afr_private_t *priv = NULL;
- char key_prefix[GF_DUMP_MAX_BUF_LEN];
- char key[GF_DUMP_MAX_BUF_LEN];
- int i = 0;
-
-
- GF_ASSERT (this);
- priv = this->private;
-
- GF_ASSERT (priv);
- snprintf(key_prefix, GF_DUMP_MAX_BUF_LEN, "%s.%s", this->type, this->name);
- gf_proc_dump_add_section(key_prefix);
- gf_proc_dump_write("child_count", "%u", priv->child_count);
- for (i = 0; i < priv->child_count; i++) {
- sprintf (key, "child_up[%d]", i);
- gf_proc_dump_write(key, "%d", priv->child_up[i]);
- sprintf (key, "pending_key[%d]", i);
- gf_proc_dump_write(key, "%s", priv->pending_key[i]);
- sprintf (key, "pending_reads[%d]", i);
- gf_proc_dump_write(key, "%"PRId64, GF_ATOMIC_GET(priv->pending_reads[i]));
- sprintf (key, "child_latency[%d]", i);
- gf_proc_dump_write(key, "%"PRId64, priv->child_latency[i]);
- }
- gf_proc_dump_write("data_self_heal", "%s", priv->data_self_heal);
- gf_proc_dump_write("metadata_self_heal", "%d", priv->metadata_self_heal);
- gf_proc_dump_write("entry_self_heal", "%d", priv->entry_self_heal);
- gf_proc_dump_write("read_child", "%d", priv->read_child);
- gf_proc_dump_write("wait_count", "%u", priv->wait_count);
- gf_proc_dump_write("heal-wait-queue-length", "%d",
- priv->heal_wait_qlen);
- gf_proc_dump_write("heal-waiters", "%d", priv->heal_waiters);
- gf_proc_dump_write("background-self-heal-count", "%d",
- priv->background_self_heal_count);
- gf_proc_dump_write("healers", "%d", priv->healers);
- gf_proc_dump_write("read-hash-mode", "%d", priv->hash_mode);
- if (priv->quorum_count == AFR_QUORUM_AUTO) {
- gf_proc_dump_write ("quorum-type", "auto");
- } else if (priv->quorum_count == 0) {
- gf_proc_dump_write ("quorum-type", "none");
- } else {
- gf_proc_dump_write("quorum-type", "fixed");
- gf_proc_dump_write("quorum-count", "%d", priv->quorum_count);
- }
-
- return 0;
+afr_priv_dump(xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ char key_prefix[GF_DUMP_MAX_BUF_LEN];
+ char key[GF_DUMP_MAX_BUF_LEN];
+ int i = 0;
+
+ GF_ASSERT(this);
+ priv = this->private;
+
+ GF_ASSERT(priv);
+ snprintf(key_prefix, GF_DUMP_MAX_BUF_LEN, "%s.%s", this->type, this->name);
+ gf_proc_dump_add_section("%s", key_prefix);
+ gf_proc_dump_write("child_count", "%u", priv->child_count);
+ for (i = 0; i < priv->child_count; i++) {
+ sprintf(key, "child_up[%d]", i);
+ gf_proc_dump_write(key, "%d", priv->child_up[i]);
+ sprintf(key, "pending_key[%d]", i);
+ gf_proc_dump_write(key, "%s", priv->pending_key[i]);
+ sprintf(key, "pending_reads[%d]", i);
+ gf_proc_dump_write(key, "%" PRId64,
+ GF_ATOMIC_GET(priv->pending_reads[i]));
+ sprintf(key, "child_latency[%d]", i);
+ gf_proc_dump_write(key, "%" PRId64, priv->child_latency[i]);
+ sprintf(key, "halo_child_up[%d]", i);
+ gf_proc_dump_write(key, "%d", priv->halo_child_up[i]);
+ }
+ gf_proc_dump_write("data_self_heal", "%d", priv->data_self_heal);
+ gf_proc_dump_write("metadata_self_heal", "%d", priv->metadata_self_heal);
+ gf_proc_dump_write("entry_self_heal", "%d", priv->entry_self_heal);
+ gf_proc_dump_write("read_child", "%d", priv->read_child);
+ gf_proc_dump_write("wait_count", "%u", priv->wait_count);
+ gf_proc_dump_write("heal-wait-queue-length", "%d", priv->heal_wait_qlen);
+ gf_proc_dump_write("heal-waiters", "%d", priv->heal_waiters);
+ gf_proc_dump_write("background-self-heal-count", "%d",
+ priv->background_self_heal_count);
+ gf_proc_dump_write("healers", "%d", priv->healers);
+ gf_proc_dump_write("read-hash-mode", "%d", priv->hash_mode);
+ gf_proc_dump_write("use-anonymous-inode", "%d", priv->use_anon_inode);
+ if (priv->quorum_count == AFR_QUORUM_AUTO) {
+ gf_proc_dump_write("quorum-type", "auto");
+ } else if (priv->quorum_count == 0) {
+ gf_proc_dump_write("quorum-type", "none");
+ } else {
+ gf_proc_dump_write("quorum-type", "fixed");
+ gf_proc_dump_write("quorum-count", "%d", priv->quorum_count);
+ }
+ gf_proc_dump_write("up", "%u", afr_has_quorum(priv->child_up, this, NULL));
+ if (priv->thin_arbiter_count) {
+ gf_proc_dump_write("ta_child_up", "%d", priv->ta_child_up);
+ gf_proc_dump_write("ta_bad_child_index", "%d",
+ priv->ta_bad_child_index);
+ gf_proc_dump_write("ta_notify_dom_lock_offset", "%" PRId64,
+ priv->ta_notify_dom_lock_offset);
+ }
+
+ return 0;
}
-
/**
* find_child_index - find the child's index in the array of subvolumes
* @this: AFR
@@ -4903,1834 +5723,2156 @@ afr_priv_dump (xlator_t *this)
*/
static int
-find_child_index (xlator_t *this, xlator_t *child)
+afr_find_child_index(xlator_t *this, xlator_t *child)
{
- afr_private_t *priv = NULL;
- int i = -1;
+ afr_private_t *priv = NULL;
+ int child_count = -1;
+ int i = -1;
- priv = this->private;
+ priv = this->private;
+ child_count = priv->child_count;
+ if (priv->thin_arbiter_count) {
+ child_count++;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if ((xlator_t *) child == priv->children[i])
- break;
- }
+ for (i = 0; i < child_count; i++) {
+ if ((xlator_t *)child == priv->children[i])
+ break;
+ }
- return i;
+ return i;
}
-static int
-__afr_get_up_children_count (afr_private_t *priv)
+int
+__afr_get_up_children_count(afr_private_t *priv)
{
- int up_children = 0;
- int i = 0;
+ int up_children = 0;
+ int i = 0;
- for (i = 0; i < priv->child_count; i++)
- if (priv->child_up[i] == 1)
- up_children++;
+ for (i = 0; i < priv->child_count; i++)
+ if (priv->child_up[i] == 1)
+ up_children++;
- return up_children;
+ return up_children;
}
-glusterfs_event_t
-__afr_transform_event_from_state (afr_private_t *priv)
+static int
+__get_heard_from_all_status(xlator_t *this)
{
- int i = 0;
- int up_children = 0;
+ afr_private_t *priv = this->private;
+ int i;
- if (AFR_COUNT (priv->last_event, priv->child_count) ==
- priv->child_count)
- /* have_heard_from_all. Let afr_notify() do the propagation. */
- return GF_EVENT_MAXVAL;
-
- up_children = __afr_get_up_children_count (priv);
- /* Treat the children with pending notification, as having sent a
- * GF_EVENT_CHILD_DOWN. i.e. set the event as GF_EVENT_SOME_DESCENDENT_DOWN,
- * as done in afr_notify() */
- for (i = 0; i < priv->child_count; i++) {
- if (priv->last_event[i])
- continue;
- priv->last_event[i] = GF_EVENT_SOME_DESCENDENT_DOWN;
- priv->child_up[i] = 0;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!priv->last_event[i]) {
+ return 0;
}
+ }
+ if (priv->thin_arbiter_count && !priv->ta_child_up) {
+ return 0;
+ }
+ return 1;
+}
- if (up_children)
- /* We received at least one child up */
- return GF_EVENT_CHILD_UP;
- else
- return GF_EVENT_CHILD_DOWN;
+glusterfs_event_t
+__afr_transform_event_from_state(xlator_t *this)
+{
+ int i = 0;
+ int up_children = 0;
+ afr_private_t *priv = this->private;
+ if (__get_heard_from_all_status(this))
+ /* have_heard_from_all. Let afr_notify() do the propagation. */
return GF_EVENT_MAXVAL;
-}
-static void
-afr_notify_cbk (void *data)
-{
- xlator_t *this = data;
- afr_private_t *priv = this->private;
- glusterfs_event_t event = GF_EVENT_MAXVAL;
- gf_boolean_t propagate = _gf_false;
+ up_children = __afr_get_up_children_count(priv);
+ /* Treat the children with pending notification, as having sent a
+ * GF_EVENT_CHILD_DOWN. i.e. set the event as GF_EVENT_SOME_DESCENDENT_DOWN,
+ * as done in afr_notify() */
+ for (i = 0; i < priv->child_count; i++) {
+ if (priv->last_event[i])
+ continue;
+ priv->last_event[i] = GF_EVENT_SOME_DESCENDENT_DOWN;
+ priv->child_up[i] = 0;
+ }
- LOCK (&priv->lock);
- {
- if (!priv->timer) {
- /*
- * Either child_up/child_down is already sent to parent.
- * This is a spurious wake up.
- */
- goto unlock;
- }
- priv->timer = NULL;
- event = __afr_transform_event_from_state (priv);
- if (event != GF_EVENT_MAXVAL)
- propagate = _gf_true;
- }
-unlock:
- UNLOCK (&priv->lock);
- if (propagate)
- default_notify (this, event, NULL);
+ if (up_children)
+ /* We received at least one child up */
+ return GF_EVENT_CHILD_UP;
+ else
+ return GF_EVENT_CHILD_DOWN;
+
+ return GF_EVENT_MAXVAL;
}
static void
-__afr_launch_notify_timer (xlator_t *this, afr_private_t *priv)
-{
-
- struct timespec delay = {0, };
-
- gf_msg_debug (this->name, 0, "Initiating child-down timer");
- delay.tv_sec = 10;
- delay.tv_nsec = 0;
- priv->timer = gf_timer_call_after (this->ctx, delay,
- afr_notify_cbk, this);
- if (priv->timer == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0, AFR_MSG_TIMER_CREATE_FAIL,
- "Cannot create timer for delayed initialization");
- }
+afr_notify_cbk(void *data)
+{
+ xlator_t *this = data;
+ afr_private_t *priv = this->private;
+ glusterfs_event_t event = GF_EVENT_MAXVAL;
+ gf_boolean_t propagate = _gf_false;
+
+ LOCK(&priv->lock);
+ {
+ if (!priv->timer) {
+ /*
+ * Either child_up/child_down is already sent to parent.
+ * This is a spurious wake up.
+ */
+ goto unlock;
+ }
+ priv->timer = NULL;
+ event = __afr_transform_event_from_state(this);
+ if (event != GF_EVENT_MAXVAL)
+ propagate = _gf_true;
+ }
+unlock:
+ UNLOCK(&priv->lock);
+ if (propagate)
+ default_notify(this, event, NULL);
}
-int
-__get_heard_from_all_status (xlator_t *this)
+static void
+__afr_launch_notify_timer(xlator_t *this, afr_private_t *priv)
{
- afr_private_t *priv = this->private;
- int heard_from_all = 1;
- int i = 0;
+ struct timespec delay = {
+ 0,
+ };
- for (i = 0; i < priv->child_count; i++) {
- if (!priv->last_event[i]) {
- heard_from_all = 0;
- break;
- }
- }
- return heard_from_all;
+ gf_msg_debug(this->name, 0, "Initiating child-down timer");
+ delay.tv_sec = 10;
+ delay.tv_nsec = 0;
+ priv->timer = gf_timer_call_after(this->ctx, delay, afr_notify_cbk, this);
+ if (priv->timer == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_TIMER_CREATE_FAIL,
+ "Cannot create timer for delayed initialization");
+ }
}
static int
-find_best_down_child (xlator_t *this)
+find_best_down_child(xlator_t *this)
{
- afr_private_t *priv = NULL;
- int i = -1;
- int32_t best_child = -1;
- int64_t best_latency = INT64_MAX;
+ afr_private_t *priv = NULL;
+ int i = -1;
+ int32_t best_child = -1;
+ int64_t best_latency = INT64_MAX;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (priv->child_up[i] &&
- priv->child_latency[i] >= 0 &&
- priv->child_latency[i] < best_latency) {
- best_child = i;
- best_latency = priv->child_latency[i];
- }
- }
- if (best_child >= 0) {
- gf_msg_debug (this->name, 0, "Found best down child (%d) "
- "@ %ld ms latency", best_child, best_latency);
+ for (i = 0; i < priv->child_count; i++) {
+ if (!priv->child_up[i] && priv->child_latency[i] >= 0 &&
+ priv->child_latency[i] < best_latency) {
+ best_child = i;
+ best_latency = priv->child_latency[i];
}
- return best_child;
+ }
+ if (best_child >= 0) {
+ gf_msg_debug(this->name, 0,
+ "Found best down child (%d) @ %" PRId64 " ms latency",
+ best_child, best_latency);
+ }
+ return best_child;
}
int
-find_worst_up_child (xlator_t *this)
+find_worst_up_child(xlator_t *this)
{
- afr_private_t *priv = NULL;
- int i = -1;
- int32_t worst_child = -1;
- int64_t worst_latency = INT64_MIN;
+ afr_private_t *priv = NULL;
+ int i = -1;
+ int32_t worst_child = -1;
+ int64_t worst_latency = INT64_MIN;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (priv->child_up[i] &&
- priv->child_latency[i] >= 0 &&
- priv->child_latency[i] > worst_latency) {
- worst_child = i;
- worst_latency = priv->child_latency[i];
- }
- }
- if (worst_child >= 0) {
- gf_msg_debug (this->name, 0, "Found worst up child (%d)"
- " @ %ld ms latency", worst_child, worst_latency);
+ for (i = 0; i < priv->child_count; i++) {
+ if (priv->child_up[i] && priv->child_latency[i] >= 0 &&
+ priv->child_latency[i] > worst_latency) {
+ worst_child = i;
+ worst_latency = priv->child_latency[i];
}
- return worst_child;
+ }
+ if (worst_child >= 0) {
+ gf_msg_debug(this->name, 0,
+ "Found worst up child (%d) @ %" PRId64 " ms latency",
+ worst_child, worst_latency);
+ }
+ return worst_child;
}
void
-__afr_handle_ping_event (xlator_t *this, xlator_t *child_xlator,
- const int idx, int64_t halo_max_latency_msec, int32_t *event,
- int64_t child_latency_msec)
+__afr_handle_ping_event(xlator_t *this, xlator_t *child_xlator, const int idx,
+ int64_t halo_max_latency_msec, int32_t *event,
+ int64_t child_latency_msec)
{
- afr_private_t *priv = NULL;
- int up_children = 0;
+ afr_private_t *priv = NULL;
+ int up_children = 0;
- priv = this->private;
+ priv = this->private;
- priv->child_latency[idx] = child_latency_msec;
- gf_msg_debug (child_xlator->name, 0, "Client ping @ %ld ms",
- child_latency_msec);
- if (priv->shd.iamshd)
- return;
-
- up_children = __afr_get_up_children_count (priv);
-
- if (child_latency_msec > halo_max_latency_msec &&
- priv->child_up[idx] == 1 &&
- up_children > priv->halo_min_replicas) {
- if ((up_children - 1) <
- priv->halo_min_replicas) {
- gf_log (child_xlator->name, GF_LOG_INFO,
- "Overriding halo threshold, "
- "min replicas: %d",
- priv->halo_min_replicas);
- } else {
- gf_log (child_xlator->name, GF_LOG_INFO,
- "Child latency (%ld ms) "
- "exceeds halo threshold (%ld), "
- "marking child down.",
- child_latency_msec,
- halo_max_latency_msec);
- *event = GF_EVENT_CHILD_DOWN;
- }
- } else if (child_latency_msec < halo_max_latency_msec &&
- priv->child_up[idx] == 0) {
- if (up_children < priv->halo_max_replicas) {
- gf_log (child_xlator->name, GF_LOG_INFO,
- "Child latency (%ld ms) "
- "below halo threshold (%ld), "
- "marking child up.",
- child_latency_msec,
- halo_max_latency_msec);
- *event = GF_EVENT_CHILD_UP;
- } else {
- gf_log (child_xlator->name, GF_LOG_INFO,
- "Not marking child %d up, "
- "max replicas (%d) reached.", idx,
- priv->halo_max_replicas);
- }
+ priv->child_latency[idx] = child_latency_msec;
+ gf_msg_debug(child_xlator->name, 0, "Client ping @ %" PRId64 " ms",
+ child_latency_msec);
+ if (priv->shd.iamshd)
+ return;
+
+ up_children = __afr_get_up_children_count(priv);
+
+ if (child_latency_msec > halo_max_latency_msec &&
+ priv->child_up[idx] == 1 && up_children > priv->halo_min_replicas) {
+ if ((up_children - 1) < priv->halo_min_replicas) {
+ gf_log(child_xlator->name, GF_LOG_INFO,
+ "Overriding halo threshold, "
+ "min replicas: %d",
+ priv->halo_min_replicas);
+ } else {
+ gf_log(child_xlator->name, GF_LOG_INFO,
+ "Child latency (%" PRId64
+ " ms) "
+ "exceeds halo threshold (%" PRId64
+ "), "
+ "marking child down.",
+ child_latency_msec, halo_max_latency_msec);
+ if (priv->halo_child_up[idx]) {
+ *event = GF_EVENT_CHILD_DOWN;
+ }
+ }
+ } else if (child_latency_msec < halo_max_latency_msec &&
+ priv->child_up[idx] == 0) {
+ if (up_children < priv->halo_max_replicas) {
+ gf_log(child_xlator->name, GF_LOG_INFO,
+ "Child latency (%" PRId64
+ " ms) "
+ "below halo threshold (%" PRId64
+ "), "
+ "marking child up.",
+ child_latency_msec, halo_max_latency_msec);
+ if (priv->halo_child_up[idx]) {
+ *event = GF_EVENT_CHILD_UP;
+ }
+ } else {
+ gf_log(child_xlator->name, GF_LOG_INFO,
+ "Not marking child %d up, "
+ "max replicas (%d) reached.",
+ idx, priv->halo_max_replicas);
}
+ }
}
static int64_t
-afr_get_halo_latency (xlator_t *this)
+afr_get_halo_latency(xlator_t *this)
{
- afr_private_t *priv = NULL;
- int64_t halo_max_latency_msec = 0;
+ afr_private_t *priv = NULL;
+ int64_t halo_max_latency_msec = 0;
- priv = this->private;
+ priv = this->private;
- if (priv->shd.iamshd) {
- halo_max_latency_msec = priv->shd.halo_max_latency_msec;
- } else if (priv->nfsd.iamnfsd) {
- halo_max_latency_msec =
- priv->nfsd.halo_max_latency_msec;
- } else {
- halo_max_latency_msec = priv->halo_max_latency_msec;
- }
- gf_msg_debug (this->name, 0, "Using halo latency %ld",
- halo_max_latency_msec);
- return halo_max_latency_msec;
+ if (priv->shd.iamshd) {
+ halo_max_latency_msec = priv->shd.halo_max_latency_msec;
+ } else if (priv->nfsd.iamnfsd) {
+ halo_max_latency_msec = priv->nfsd.halo_max_latency_msec;
+ } else {
+ halo_max_latency_msec = priv->halo_max_latency_msec;
+ }
+ gf_msg_debug(this->name, 0, "Using halo latency %" PRId64,
+ halo_max_latency_msec);
+ return halo_max_latency_msec;
}
void
-__afr_handle_child_up_event (xlator_t *this, xlator_t *child_xlator,
- const int idx, int64_t child_latency_msec,
- int32_t *event, int32_t *call_psh, int32_t *up_child)
-{
- afr_private_t *priv = NULL;
- int up_children = 0;
- int worst_up_child = -1;
- int64_t halo_max_latency_msec = afr_get_halo_latency (this);
-
- priv = this->private;
-
- /*
- * This only really counts if the child was never up
- * (value = -1) or had been down (value = 0). See
- * comment at GF_EVENT_CHILD_DOWN for a more detailed
- * explanation.
- */
- if (priv->child_up[idx] != 1) {
- priv->event_generation++;
- }
- priv->child_up[idx] = 1;
-
- *call_psh = 1;
- *up_child = idx;
- up_children = __afr_get_up_children_count (priv);
- /*
- * If this is an _actual_ CHILD_UP event, we
- * want to set the child_latency to MAX to indicate
- * the child needs ping data to be available before doing child-up
- */
- if (child_latency_msec < 0 && priv->halo_enabled) {
- /*set to INT64_MAX-1 so that it is found for best_down_child*/
- priv->child_latency[idx] = AFR_HALO_MAX_LATENCY;
- }
-
- /*
- * Handle the edge case where we exceed
- * halo_min_replicas and we've got a child which is
- * marked up as it was helping to satisfy the
- * halo_min_replicas even though it's latency exceeds
- * halo_max_latency_msec.
- */
- if (up_children > priv->halo_min_replicas) {
- worst_up_child = find_worst_up_child (this);
- if (worst_up_child >= 0 &&
- priv->child_latency[worst_up_child] >
- halo_max_latency_msec) {
- gf_msg_debug (this->name, 0, "Marking child %d down, "
- "doesn't meet halo threshold (%ld), and > "
- "halo_min_replicas (%d)",
- worst_up_child, halo_max_latency_msec,
- priv->halo_min_replicas);
- priv->child_up[worst_up_child] = 0;
- up_children--;
- }
- }
-
- if (up_children > priv->halo_max_replicas &&
- !priv->shd.iamshd) {
- worst_up_child = find_worst_up_child (this);
- if (worst_up_child < 0) {
- worst_up_child = idx;
- }
- priv->child_up[worst_up_child] = 0;
- up_children--;
- gf_msg_debug (this->name, 0, "Marking child %d down, "
- "up_children (%d) > halo_max_replicas (%d)",
- worst_up_child, up_children, priv->halo_max_replicas);
- }
-
- if (up_children == 1) {
- gf_msg (this->name, GF_LOG_INFO, 0, AFR_MSG_SUBVOL_UP,
- "Subvolume '%s' came back up; "
- "going online.",
- child_xlator->name);
- gf_event (EVENT_AFR_SUBVOL_UP, "subvol=%s", this->name);
- } else {
- *event = GF_EVENT_SOME_DESCENDENT_UP;
- }
+__afr_handle_child_up_event(xlator_t *this, xlator_t *child_xlator,
+ const int idx, int64_t child_latency_msec,
+ int32_t *event, int32_t *call_psh,
+ int32_t *up_child)
+{
+ afr_private_t *priv = NULL;
+ int up_children = 0;
+ int worst_up_child = -1;
+ int64_t halo_max_latency_msec = afr_get_halo_latency(this);
+
+ priv = this->private;
+
+ /*
+ * This only really counts if the child was never up
+ * (value = -1) or had been down (value = 0). See
+ * comment at GF_EVENT_CHILD_DOWN for a more detailed
+ * explanation.
+ */
+ if (priv->child_up[idx] != 1) {
+ priv->event_generation++;
+ }
+ priv->child_up[idx] = 1;
+
+ *call_psh = 1;
+ *up_child = idx;
+ up_children = __afr_get_up_children_count(priv);
+ /*
+ * If this is an _actual_ CHILD_UP event, we
+ * want to set the child_latency to MAX to indicate
+ * the child needs ping data to be available before doing child-up
+ */
+ if (!priv->halo_enabled)
+ goto out;
+
+ if (child_latency_msec < 0) {
+ /*set to INT64_MAX-1 so that it is found for best_down_child*/
+ priv->halo_child_up[idx] = 1;
+ if (priv->child_latency[idx] < 0) {
+ priv->child_latency[idx] = AFR_HALO_MAX_LATENCY;
+ }
+ }
+
+ /*
+ * Handle the edge case where we exceed
+ * halo_min_replicas and we've got a child which is
+ * marked up as it was helping to satisfy the
+ * halo_min_replicas even though it's latency exceeds
+ * halo_max_latency_msec.
+ */
+ if (up_children > priv->halo_min_replicas) {
+ worst_up_child = find_worst_up_child(this);
+ if (worst_up_child >= 0 &&
+ priv->child_latency[worst_up_child] > halo_max_latency_msec) {
+ gf_msg_debug(this->name, 0,
+ "Marking child %d down, "
+ "doesn't meet halo threshold (%" PRId64
+ "), and > "
+ "halo_min_replicas (%d)",
+ worst_up_child, halo_max_latency_msec,
+ priv->halo_min_replicas);
+ priv->child_up[worst_up_child] = 0;
+ up_children--;
+ }
+ }
+
+ if (up_children > priv->halo_max_replicas && !priv->shd.iamshd) {
+ worst_up_child = find_worst_up_child(this);
+ if (worst_up_child < 0) {
+ worst_up_child = idx;
+ }
+ priv->child_up[worst_up_child] = 0;
+ up_children--;
+ gf_msg_debug(this->name, 0,
+ "Marking child %d down, "
+ "up_children (%d) > halo_max_replicas (%d)",
+ worst_up_child, up_children, priv->halo_max_replicas);
+ }
+out:
+ if (up_children == 1) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_SUBVOL_UP,
+ "Subvolume '%s' came back up; "
+ "going online.",
+ child_xlator->name);
+ gf_event(EVENT_AFR_SUBVOL_UP, "client-pid=%d; subvol=%s",
+ this->ctx->cmd_args.client_pid, this->name);
+ } else {
+ *event = GF_EVENT_SOME_DESCENDENT_UP;
+ }
- priv->last_event[idx] = *event;
+ priv->last_event[idx] = *event;
}
void
-__afr_handle_child_down_event (xlator_t *this, xlator_t *child_xlator,
- int idx, int64_t child_latency_msec, int32_t *event,
- int32_t *call_psh, int32_t *up_child)
-{
- afr_private_t *priv = NULL;
- int i = 0;
- int up_children = 0;
- int down_children = 0;
- int best_down_child = -1;
-
- priv = this->private;
-
- /*
- * If a brick is down when we start, we'll get a
- * CHILD_DOWN to indicate its initial state. There
- * was never a CHILD_UP in this case, so if we
- * increment "down_count" the difference between than
- * and "up_count" will no longer be the number of
- * children that are currently up. This has serious
- * implications e.g. for quorum enforcement, so we
- * don't increment these values unless the event
- * represents an actual state transition between "up"
- * (value = 1) and anything else.
- */
- if (priv->child_up[idx] == 1) {
- priv->event_generation++;
- }
+__afr_handle_child_down_event(xlator_t *this, xlator_t *child_xlator, int idx,
+ int64_t child_latency_msec, int32_t *event,
+ int32_t *call_psh, int32_t *up_child)
+{
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int up_children = 0;
+ int down_children = 0;
+ int best_down_child = -1;
+
+ priv = this->private;
+
+ /*
+ * If a brick is down when we start, we'll get a
+ * CHILD_DOWN to indicate its initial state. There
+ * was never a CHILD_UP in this case, so if we
+ * increment "down_count" the difference between than
+ * and "up_count" will no longer be the number of
+ * children that are currently up. This has serious
+ * implications e.g. for quorum enforcement, so we
+ * don't increment these values unless the event
+ * represents an actual state transition between "up"
+ * (value = 1) and anything else.
+ */
+ if (priv->child_up[idx] == 1) {
+ priv->event_generation++;
+ }
+
+ /*
+ * If this is an _actual_ CHILD_DOWN event, we
+ * want to set the child_latency to < 0 to indicate
+ * the child is really disconnected.
+ */
+ if (child_latency_msec < 0) {
+ priv->child_latency[idx] = child_latency_msec;
+ priv->halo_child_up[idx] = 0;
+ }
+ priv->child_up[idx] = 0;
+
+ up_children = __afr_get_up_children_count(priv);
+ /*
+ * Handle the edge case where we need to find the
+ * next best child (to mark up) as marking this child
+ * down would cause us to fall below halo_min_replicas.
+ * We will also force the SHD to heal this child _now_
+ * as we want it to be up to date if we are going to
+ * begin using it synchronously.
+ */
+ if (priv->halo_enabled && up_children < priv->halo_min_replicas) {
+ best_down_child = find_best_down_child(this);
+ if (best_down_child >= 0) {
+ gf_msg_debug(this->name, 0,
+ "Swapping out child %d for "
+ "child %d to satisfy halo_min_replicas (%d).",
+ idx, best_down_child, priv->halo_min_replicas);
+ priv->child_up[best_down_child] = 1;
+ *call_psh = 1;
+ *up_child = best_down_child;
+ }
+ }
+ for (i = 0; i < priv->child_count; i++)
+ if (priv->child_up[i] == 0)
+ down_children++;
+ if (down_children == priv->child_count) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SUBVOLS_DOWN,
+ "All subvolumes are down. Going "
+ "offline until at least one of them "
+ "comes back up.");
+ gf_event(EVENT_AFR_SUBVOLS_DOWN, "client-pid=%d; subvol=%s",
+ this->ctx->cmd_args.client_pid, this->name);
+ } else {
+ *event = GF_EVENT_SOME_DESCENDENT_DOWN;
+ }
+ priv->last_event[idx] = *event;
+}
- /*
- * If this is an _actual_ CHILD_DOWN event, we
- * want to set the child_latency to < 0 to indicate
- * the child is really disconnected.
- */
- if (child_latency_msec < 0) {
- priv->child_latency[idx] = child_latency_msec;
- }
- priv->child_up[idx] = 0;
+void
+afr_ta_lock_release_synctask(xlator_t *this)
+{
+ call_frame_t *ta_frame = NULL;
+ int ret = 0;
- up_children = __afr_get_up_children_count (priv);
- /*
- * Handle the edge case where we need to find the
- * next best child (to mark up) as marking this child
- * down would cause us to fall below halo_min_replicas.
- * We will also force the SHD to heal this child _now_
- * as we want it to be up to date if we are going to
- * begin using it synchronously.
- */
- if (up_children < priv->halo_min_replicas) {
- best_down_child = find_best_down_child (this);
- if (best_down_child >= 0) {
- gf_msg_debug (this->name, 0,
- "Swapping out child %d for "
- "child %d to satisfy halo_min_replicas (%d).",
- idx, best_down_child, priv->halo_min_replicas);
- priv->child_up[best_down_child] = 1;
- *call_psh = 1;
- *up_child = best_down_child;
- }
- }
+ ta_frame = afr_ta_frame_create(this);
+ if (!ta_frame) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, AFR_MSG_THIN_ARB,
+ "Failed to create ta_frame");
+ return;
+ }
- for (i = 0; i < priv->child_count; i++)
- if (priv->child_up[i] == 0)
- down_children++;
- if (down_children == priv->child_count) {
- gf_msg (this->name, GF_LOG_ERROR, 0, AFR_MSG_SUBVOLS_DOWN,
- "All subvolumes are down. Going "
- "offline until atleast one of them "
- "comes back up.");
- gf_event (EVENT_AFR_SUBVOLS_DOWN, "subvol=%s", this->name);
- } else {
- *event = GF_EVENT_SOME_DESCENDENT_DOWN;
- }
- priv->last_event[idx] = *event;
+ ret = synctask_new(this->ctx->env, afr_release_notify_lock_for_ta,
+ afr_ta_lock_release_done, ta_frame, this);
+ if (ret) {
+ STACK_DESTROY(ta_frame->root);
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, AFR_MSG_THIN_ARB,
+ "Failed to release "
+ "AFR_TA_DOM_NOTIFY lock.");
+ }
}
-int32_t
-afr_notify (xlator_t *this, int32_t event,
- void *data, void *data2)
-{
- afr_private_t *priv = NULL;
- xlator_t *child_xlator = NULL;
- int i = -1;
- int propagate = 0;
- int had_heard_from_all = 0;
- int have_heard_from_all = 0;
- int idx = -1;
- int ret = -1;
- int call_psh = 0;
- int up_child = -1;
- dict_t *input = NULL;
- dict_t *output = NULL;
- gf_boolean_t had_quorum = _gf_false;
- gf_boolean_t has_quorum = _gf_false;
- struct gf_upcall *up_data = NULL;
- struct gf_upcall_cache_invalidation *up_ci = NULL;
- inode_table_t *itable = NULL;
- inode_t *inode = NULL;
- int64_t halo_max_latency_msec = 0;
- int64_t child_latency_msec = -1;
-
- child_xlator = (xlator_t *)data;
-
- priv = this->private;
-
- if (!priv)
- return 0;
-
- /*
- * We need to reset this in case children come up in "staggered"
- * fashion, so that we discover a late-arriving local subvolume. Note
- * that we could end up issuing N lookups to the first subvolume, and
- * O(N^2) overall, but N is small for AFR so it shouldn't be an issue.
- */
- priv->did_discovery = _gf_false;
+static void
+afr_handle_inodelk_contention(xlator_t *this, struct gf_upcall *upcall)
+{
+ struct gf_upcall_inodelk_contention *lc = NULL;
+ unsigned int inmem_count = 0;
+ unsigned int onwire_count = 0;
+ afr_private_t *priv = this->private;
+ lc = upcall->data;
- /* parent xlators dont need to know about every child_up, child_down
- * because of afr ha. If all subvolumes go down, child_down has
- * to be triggered. In that state when 1 subvolume comes up child_up
- * needs to be triggered. dht optimizes revalidate lookup by sending
- * it only to one of its subvolumes. When child up/down happens
- * for afr's subvolumes dht should be notified by child_modified. The
- * subsequent revalidate lookup happens on all the dht's subvolumes
- * which triggers afr self-heals if any.
- */
- idx = find_child_index (this, child_xlator);
- if (idx < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_CHILD_UP,
- "Received child_up from invalid subvolume");
- goto out;
- }
+ if (strcmp(lc->domain, AFR_TA_DOM_NOTIFY) != 0)
+ return;
- had_quorum = priv->quorum_count && afr_has_quorum (priv->child_up,
- this);
- if (priv->halo_enabled) {
- halo_max_latency_msec = afr_get_halo_latency (this);
-
- if (event == GF_EVENT_CHILD_PING) {
- /* Calculates the child latency and sets event
- */
- child_latency_msec = (int64_t)(uintptr_t)data2;
- LOCK (&priv->lock);
- {
- __afr_handle_ping_event (this, child_xlator,
- idx, halo_max_latency_msec, &event,
- child_latency_msec);
- }
- UNLOCK (&priv->lock);
- }
- }
+ if (priv->shd.iamshd) {
+ /* shd should ignore AFR_TA_DOM_NOTIFY release requests. */
+ return;
+ }
+ LOCK(&priv->lock);
+ {
+ if (priv->release_ta_notify_dom_lock == _gf_true) {
+ /* Ignore multiple release requests from shds.*/
+ UNLOCK(&priv->lock);
+ return;
+ }
+ priv->release_ta_notify_dom_lock = _gf_true;
+ inmem_count = priv->ta_in_mem_txn_count;
+ onwire_count = priv->ta_on_wire_txn_count;
+ }
+ UNLOCK(&priv->lock);
+ if (inmem_count || onwire_count)
+ /* lock release will happen in txn code path after
+ * in-memory or on-wire txns are over.*/
+ return;
- if (event == GF_EVENT_CHILD_PING) {
- /* This is the only xlator that handles PING, no reason to
- * propagate.
- */
- goto out;
+ afr_ta_lock_release_synctask(this);
+}
+
+static void
+afr_handle_upcall_event(xlator_t *this, struct gf_upcall *upcall)
+{
+ struct gf_upcall_cache_invalidation *up_ci = NULL;
+ afr_private_t *priv = this->private;
+ inode_t *inode = NULL;
+ inode_table_t *itable = NULL;
+ int i = 0;
+
+ switch (upcall->event_type) {
+ case GF_UPCALL_INODELK_CONTENTION:
+ afr_handle_inodelk_contention(this, upcall);
+ break;
+ case GF_UPCALL_CACHE_INVALIDATION:
+ up_ci = (struct gf_upcall_cache_invalidation *)upcall->data;
+
+ /* Since md-cache will be aggressively filtering
+ * lookups, the stale read issue will be more
+ * pronounced. Hence when a pending xattr is set notify
+ * all the md-cache clients to invalidate the existing
+ * stat cache and send the lookup next time */
+ if (!up_ci->dict)
+ break;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!dict_get(up_ci->dict, priv->pending_key[i]))
+ continue;
+ up_ci->flags |= UP_INVAL_ATTR;
+ itable = ((xlator_t *)this->graph->top)->itable;
+ /*Internal processes may not have itable for
+ *top xlator*/
+ if (itable)
+ inode = inode_find(itable, upcall->gfid);
+ if (inode)
+ afr_inode_need_refresh_set(inode, this);
+ break;
}
+ break;
+ default:
+ break;
+ }
+}
- if (event == GF_EVENT_TRANSLATOR_OP) {
- LOCK (&priv->lock);
- {
- had_heard_from_all = __get_heard_from_all_status (this);
- }
- UNLOCK (&priv->lock);
-
- if (!had_heard_from_all) {
- ret = -1;
- } else {
- input = data;
- output = data2;
- ret = afr_xl_op (this, input, output);
- }
- goto out;
+int32_t
+afr_notify(xlator_t *this, int32_t event, void *data, void *data2)
+{
+ afr_private_t *priv = NULL;
+ xlator_t *child_xlator = NULL;
+ int i = -1;
+ int propagate = 0;
+ int had_heard_from_all = 0;
+ int have_heard_from_all = 0;
+ int idx = -1;
+ int ret = -1;
+ int call_psh = 0;
+ int up_child = -1;
+ dict_t *input = NULL;
+ dict_t *output = NULL;
+ gf_boolean_t had_quorum = _gf_false;
+ gf_boolean_t has_quorum = _gf_false;
+ int64_t halo_max_latency_msec = 0;
+ int64_t child_latency_msec = -1;
+
+ child_xlator = (xlator_t *)data;
+
+ priv = this->private;
+
+ if (!priv)
+ return 0;
+
+ /*
+ * We need to reset this in case children come up in "staggered"
+ * fashion, so that we discover a late-arriving local subvolume. Note
+ * that we could end up issuing N lookups to the first subvolume, and
+ * O(N^2) overall, but N is small for AFR so it shouldn't be an issue.
+ */
+ priv->did_discovery = _gf_false;
+
+ /* parent xlators don't need to know about every child_up, child_down
+ * because of afr ha. If all subvolumes go down, child_down has
+ * to be triggered. In that state when 1 subvolume comes up child_up
+ * needs to be triggered. dht optimizes revalidate lookup by sending
+ * it only to one of its subvolumes. When child up/down happens
+ * for afr's subvolumes dht should be notified by child_modified. The
+ * subsequent revalidate lookup happens on all the dht's subvolumes
+ * which triggers afr self-heals if any.
+ */
+ idx = afr_find_child_index(this, child_xlator);
+ if (idx < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_CHILD_UP,
+ "Received child_up from invalid subvolume");
+ goto out;
+ }
+
+ had_quorum = priv->quorum_count &&
+ afr_has_quorum(priv->child_up, this, NULL);
+ if (event == GF_EVENT_CHILD_PING) {
+ child_latency_msec = (int64_t)(uintptr_t)data2;
+ if (priv->halo_enabled) {
+ halo_max_latency_msec = afr_get_halo_latency(this);
+
+ /* Calculates the child latency and sets event
+ */
+ LOCK(&priv->lock);
+ {
+ __afr_handle_ping_event(this, child_xlator, idx,
+ halo_max_latency_msec, &event,
+ child_latency_msec);
+ }
+ UNLOCK(&priv->lock);
+ } else {
+ LOCK(&priv->lock);
+ {
+ priv->child_latency[idx] = child_latency_msec;
+ }
+ UNLOCK(&priv->lock);
}
+ }
+
+ if (event == GF_EVENT_CHILD_PING) {
+ /* This is the only xlator that handles PING, no reason to
+ * propagate.
+ */
+ goto out;
+ }
- LOCK (&priv->lock);
+ if (event == GF_EVENT_TRANSLATOR_OP) {
+ LOCK(&priv->lock);
{
- had_heard_from_all = __get_heard_from_all_status (this);
- switch (event) {
- case GF_EVENT_PARENT_UP:
- __afr_launch_notify_timer (this, priv);
- propagate = 1;
- break;
- case GF_EVENT_CHILD_UP:
- __afr_handle_child_up_event (this, child_xlator,
- idx, child_latency_msec, &event, &call_psh,
- &up_child);
- break;
-
- case GF_EVENT_CHILD_DOWN:
- __afr_handle_child_down_event (this, child_xlator, idx,
- child_latency_msec, &event, &call_psh,
- &up_child);
- break;
-
- case GF_EVENT_CHILD_CONNECTING:
- priv->last_event[idx] = event;
-
- break;
-
- case GF_EVENT_SOME_DESCENDENT_DOWN:
- priv->last_event[idx] = event;
- break;
- case GF_EVENT_UPCALL:
- up_data = (struct gf_upcall *)data;
- if (up_data->event_type != GF_UPCALL_CACHE_INVALIDATION)
- break;
- up_ci = (struct gf_upcall_cache_invalidation *)up_data->data;
-
- /* Since md-cache will be aggressively filtering
- * lookups, the stale read issue will be more
- * pronounced. Hence when a pending xattr is set notify
- * all the md-cache clients to invalidate the existing
- * stat cache and send the lookup next time */
- if (!up_ci->dict)
- break;
- for (i = 0; i < priv->child_count; i++) {
- if (dict_get (up_ci->dict, priv->pending_key[i])) {
- up_ci->flags |= UP_INVAL_ATTR;
- itable = ((xlator_t *)this->graph->top)->itable;
- /*Internal processes may not have itable for top xlator*/
- if (itable)
- inode = inode_find (itable, up_data->gfid);
- if (inode)
- afr_inode_need_refresh_set (inode, this);
-
- break;
- }
- }
- break;
- default:
- propagate = 1;
- break;
- }
- have_heard_from_all = __get_heard_from_all_status (this);
- if (!had_heard_from_all && have_heard_from_all) {
- if (priv->timer) {
- gf_timer_call_cancel (this->ctx, priv->timer);
- priv->timer = NULL;
- }
- /* This is the first event which completes aggregation
- of events from all subvolumes. If at least one subvol
- had come up, propagate CHILD_UP, but only this time
- */
- event = GF_EVENT_CHILD_DOWN;
- for (i = 0; i < priv->child_count; i++) {
- if (priv->last_event[i] == GF_EVENT_CHILD_UP) {
- event = GF_EVENT_CHILD_UP;
- break;
- }
-
- if (priv->last_event[i] ==
- GF_EVENT_CHILD_CONNECTING) {
- event = GF_EVENT_CHILD_CONNECTING;
- /* continue to check other events for CHILD_UP */
- }
- }
- }
+ had_heard_from_all = __get_heard_from_all_status(this);
}
- UNLOCK (&priv->lock);
+ UNLOCK(&priv->lock);
- if (priv->quorum_count) {
- has_quorum = afr_has_quorum (priv->child_up, this);
- if (!had_quorum && has_quorum) {
- gf_msg (this->name, GF_LOG_INFO, 0, AFR_MSG_QUORUM_MET,
- "Client-quorum is met");
- gf_event (EVENT_AFR_QUORUM_MET,
- "subvol=%s", this->name);
+ if (!had_heard_from_all) {
+ ret = -1;
+ } else {
+ input = data;
+ output = data2;
+ ret = afr_xl_op(this, input, output);
+ }
+ goto out;
+ }
+
+ if (event == GF_EVENT_UPCALL) {
+ afr_handle_upcall_event(this, data);
+ }
+
+ LOCK(&priv->lock);
+ {
+ had_heard_from_all = __get_heard_from_all_status(this);
+ switch (event) {
+ case GF_EVENT_PARENT_UP:
+ __afr_launch_notify_timer(this, priv);
+ propagate = 1;
+ break;
+ case GF_EVENT_CHILD_UP:
+ if (priv->thin_arbiter_count &&
+ (idx == AFR_CHILD_THIN_ARBITER)) {
+ priv->ta_child_up = 1;
+ priv->ta_event_gen++;
+ break;
}
- if (had_quorum && !has_quorum) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_QUORUM_FAIL,
- "Client-quorum is not met");
- gf_event (EVENT_AFR_QUORUM_FAIL, "subvol=%s",
- this->name);
+ __afr_handle_child_up_event(this, child_xlator, idx,
+ child_latency_msec, &event,
+ &call_psh, &up_child);
+ __afr_lock_heal_synctask(this, priv, idx);
+ break;
+
+ case GF_EVENT_CHILD_DOWN:
+ if (priv->thin_arbiter_count &&
+ (idx == AFR_CHILD_THIN_ARBITER)) {
+ priv->ta_child_up = 0;
+ priv->ta_event_gen++;
+ afr_ta_locked_priv_invalidate(priv);
+ break;
}
- }
+ __afr_handle_child_down_event(this, child_xlator, idx,
+ child_latency_msec, &event,
+ &call_psh, &up_child);
+ __afr_mark_pending_lk_heal(this, priv, idx);
+ break;
- /* if all subvols have reported status, no need to hide anything
- or wait for anything else. Just propagate blindly */
- if (have_heard_from_all)
+ case GF_EVENT_CHILD_CONNECTING:
+ priv->last_event[idx] = event;
+
+ break;
+
+ case GF_EVENT_SOME_DESCENDENT_DOWN:
+ priv->last_event[idx] = event;
+ break;
+ default:
propagate = 1;
+ break;
+ }
+ have_heard_from_all = __get_heard_from_all_status(this);
+ if (!had_heard_from_all && have_heard_from_all) {
+ if (priv->timer) {
+ gf_timer_call_cancel(this->ctx, priv->timer);
+ priv->timer = NULL;
+ }
+ /* This is the first event which completes aggregation
+ of events from all subvolumes. If at least one subvol
+ had come up, propagate CHILD_UP, but only this time
+ */
+ event = GF_EVENT_CHILD_DOWN;
+ for (i = 0; i < priv->child_count; i++) {
+ if (priv->last_event[i] == GF_EVENT_CHILD_UP) {
+ event = GF_EVENT_CHILD_UP;
+ break;
+ }
- ret = 0;
- if (propagate)
- ret = default_notify (this, event, data);
-
- if ((!had_heard_from_all) || call_psh) {
- /* Launch self-heal on all local subvolumes if:
- * a) We have_heard_from_all for the first time
- * b) Already heard from everyone, but we now got a child-up
- * event.
- */
- if (have_heard_from_all && priv->shd.iamshd) {
- for (i = 0; i < priv->child_count; i++)
- if (priv->child_up[i])
- afr_selfheal_childup (this, i);
+ if (priv->last_event[i] == GF_EVENT_CHILD_CONNECTING) {
+ event = GF_EVENT_CHILD_CONNECTING;
+ /* continue to check other events for CHILD_UP */
}
+ }
}
+ }
+ UNLOCK(&priv->lock);
+
+ if (priv->quorum_count) {
+ has_quorum = afr_has_quorum(priv->child_up, this, NULL);
+ if (!had_quorum && has_quorum) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_QUORUM_MET,
+ "Client-quorum is met");
+ gf_event(EVENT_AFR_QUORUM_MET, "client-pid=%d; subvol=%s",
+ this->ctx->cmd_args.client_pid, this->name);
+ }
+ if (had_quorum && !has_quorum) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_QUORUM_FAIL,
+ "Client-quorum is not met");
+ gf_event(EVENT_AFR_QUORUM_FAIL, "client-pid=%d; subvol=%s",
+ this->ctx->cmd_args.client_pid, this->name);
+ }
+ }
+
+ /* if all subvols have reported status, no need to hide anything
+ or wait for anything else. Just propagate blindly */
+ if (have_heard_from_all)
+ propagate = 1;
+
+ ret = 0;
+ if (propagate)
+ ret = default_notify(this, event, data);
+
+ if ((!had_heard_from_all) || call_psh) {
+ /* Launch self-heal on all local subvolumes if:
+ * a) We have_heard_from_all for the first time
+ * b) Already heard from everyone, but we now got a child-up
+ * event.
+ */
+ if (have_heard_from_all) {
+ afr_selfheal_childup(this, priv);
+ }
+ }
out:
- return ret;
+ return ret;
}
int
-afr_local_init (afr_local_t *local, afr_private_t *priv, int32_t *op_errno)
+afr_local_init(afr_local_t *local, afr_private_t *priv, int32_t *op_errno)
{
- int __ret = -1;
- local->op_ret = -1;
- local->op_errno = EUCLEAN;
+ int __ret = -1;
+ local->op_ret = -1;
+ local->op_errno = EUCLEAN;
- __ret = syncbarrier_init (&local->barrier);
- if (__ret) {
- if (op_errno)
- *op_errno = __ret;
- goto out;
- }
-
- local->child_up = GF_CALLOC (priv->child_count,
- sizeof (*local->child_up),
- gf_afr_mt_char);
- if (!local->child_up) {
- if (op_errno)
- *op_errno = ENOMEM;
- goto out;
- }
-
- memcpy (local->child_up, priv->child_up,
- sizeof (*local->child_up) * priv->child_count);
- local->call_count = AFR_COUNT (local->child_up, priv->child_count);
- if (local->call_count == 0) {
- gf_msg (THIS->name, GF_LOG_INFO, 0,
- AFR_MSG_SUBVOLS_DOWN, "no subvolumes up");
- if (op_errno)
- *op_errno = ENOTCONN;
- goto out;
- }
-
- local->event_generation = priv->event_generation;
+ __ret = syncbarrier_init(&local->barrier);
+ if (__ret) {
+ if (op_errno)
+ *op_errno = __ret;
+ goto out;
+ }
- local->read_attempted = GF_CALLOC (priv->child_count, sizeof (char),
- gf_afr_mt_char);
- if (!local->read_attempted) {
- if (op_errno)
- *op_errno = ENOMEM;
- goto out;
- }
+ local->child_up = GF_MALLOC(priv->child_count * sizeof(*local->child_up),
+ gf_afr_mt_char);
+ if (!local->child_up) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ memcpy(local->child_up, priv->child_up,
+ sizeof(*local->child_up) * priv->child_count);
+ local->call_count = AFR_COUNT(local->child_up, priv->child_count);
+ if (local->call_count == 0) {
+ gf_msg(THIS->name, GF_LOG_INFO, 0, AFR_MSG_SUBVOLS_DOWN,
+ "no subvolumes up");
+ if (op_errno)
+ *op_errno = ENOTCONN;
+ goto out;
+ }
- local->readable = GF_CALLOC (priv->child_count, sizeof (char),
- gf_afr_mt_char);
- if (!local->readable) {
- if (op_errno)
- *op_errno = ENOMEM;
- goto out;
- }
+ local->event_generation = priv->event_generation;
- local->readable2 = GF_CALLOC (priv->child_count, sizeof (char),
+ local->read_attempted = GF_CALLOC(priv->child_count, sizeof(char),
gf_afr_mt_char);
- if (!local->readable2) {
- if (op_errno)
- *op_errno = ENOMEM;
- goto out;
- }
+ if (!local->read_attempted) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
- local->read_subvol = -1;
+ local->readable = GF_CALLOC(priv->child_count, sizeof(char),
+ gf_afr_mt_char);
+ if (!local->readable) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
- local->replies = GF_CALLOC(priv->child_count, sizeof(*local->replies),
- gf_afr_mt_reply_t);
- if (!local->replies) {
- if (op_errno)
- *op_errno = ENOMEM;
- goto out;
- }
+ local->readable2 = GF_CALLOC(priv->child_count, sizeof(char),
+ gf_afr_mt_char);
+ if (!local->readable2) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
- local->need_full_crawl = _gf_false;
+ local->read_subvol = -1;
- INIT_LIST_HEAD (&local->healer);
- return 0;
+ local->replies = GF_CALLOC(priv->child_count, sizeof(*local->replies),
+ gf_afr_mt_reply_t);
+ if (!local->replies) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ local->need_full_crawl = _gf_false;
+ if (priv->thin_arbiter_count) {
+ local->ta_child_up = priv->ta_child_up;
+ local->ta_failed_subvol = AFR_CHILD_UNKNOWN;
+ local->read_txn_query_child = AFR_CHILD_UNKNOWN;
+ local->ta_event_gen = priv->ta_event_gen;
+ local->fop_state = TA_SUCCESS;
+ }
+ local->is_new_entry = _gf_false;
+
+ INIT_LIST_HEAD(&local->healer);
+ return 0;
out:
- return -1;
+ return -1;
}
int
-afr_internal_lock_init (afr_internal_lock_t *lk, size_t child_count)
+afr_internal_lock_init(afr_internal_lock_t *lk, size_t child_count)
{
- int ret = -ENOMEM;
+ int ret = -ENOMEM;
- lk->locked_nodes = GF_CALLOC (sizeof (*lk->locked_nodes),
- child_count, gf_afr_mt_char);
- if (NULL == lk->locked_nodes)
- goto out;
-
- lk->lower_locked_nodes = GF_CALLOC (sizeof (*lk->lower_locked_nodes),
- child_count, gf_afr_mt_char);
- if (NULL == lk->lower_locked_nodes)
- goto out;
+ lk->lower_locked_nodes = GF_CALLOC(sizeof(*lk->lower_locked_nodes),
+ child_count, gf_afr_mt_char);
+ if (NULL == lk->lower_locked_nodes)
+ goto out;
- lk->lock_op_ret = -1;
- lk->lock_op_errno = EUCLEAN;
+ lk->lock_op_ret = -1;
+ lk->lock_op_errno = EUCLEAN;
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
void
-afr_matrix_cleanup (int32_t **matrix, unsigned int m)
+afr_matrix_cleanup(int32_t **matrix, unsigned int m)
{
- int i = 0;
+ int i = 0;
- if (!matrix)
- goto out;
- for (i = 0; i < m; i++) {
- GF_FREE (matrix[i]);
- }
+ if (!matrix)
+ goto out;
+ for (i = 0; i < m; i++) {
+ GF_FREE(matrix[i]);
+ }
- GF_FREE (matrix);
+ GF_FREE(matrix);
out:
- return;
+ return;
}
-int32_t**
-afr_matrix_create (unsigned int m, unsigned int n)
+int32_t **
+afr_matrix_create(unsigned int m, unsigned int n)
{
- int32_t **matrix = NULL;
- int i = 0;
+ int32_t **matrix = NULL;
+ int i = 0;
- matrix = GF_CALLOC (sizeof (*matrix), m, gf_afr_mt_int32_t);
- if (!matrix)
- goto out;
+ matrix = GF_CALLOC(sizeof(*matrix), m, gf_afr_mt_int32_t);
+ if (!matrix)
+ goto out;
- for (i = 0; i < m; i++) {
- matrix[i] = GF_CALLOC (sizeof (*matrix[i]), n,
- gf_afr_mt_int32_t);
- if (!matrix[i])
- goto out;
- }
- return matrix;
+ for (i = 0; i < m; i++) {
+ matrix[i] = GF_CALLOC(sizeof(*matrix[i]), n, gf_afr_mt_int32_t);
+ if (!matrix[i])
+ goto out;
+ }
+ return matrix;
out:
- afr_matrix_cleanup (matrix, m);
- return NULL;
+ afr_matrix_cleanup(matrix, m);
+ return NULL;
}
int
-afr_transaction_local_init (afr_local_t *local, xlator_t *this)
-{
- int ret = -ENOMEM;
- afr_private_t *priv = NULL;
-
- priv = this->private;
- ret = afr_internal_lock_init (&local->internal_lock, priv->child_count);
- if (ret < 0)
- goto out;
-
- ret = -ENOMEM;
- local->pre_op_compat = priv->pre_op_compat;
-
- local->transaction.pre_op = GF_CALLOC (sizeof (*local->transaction.pre_op),
- priv->child_count,
- gf_afr_mt_char);
- if (!local->transaction.pre_op)
- goto out;
-
- local->transaction.changelog_xdata =
- GF_CALLOC (sizeof (*local->transaction.changelog_xdata),
- priv->child_count, gf_afr_mt_dict_t);
- if (!local->transaction.changelog_xdata)
- goto out;
-
- if (priv->arbiter_count == 1) {
- local->transaction.pre_op_sources =
- GF_CALLOC (sizeof (*local->transaction.pre_op_sources),
- priv->child_count, gf_afr_mt_char);
- if (!local->transaction.pre_op_sources)
- goto out;
- }
-
- local->transaction.failed_subvols = GF_CALLOC (sizeof (*local->transaction.failed_subvols),
- priv->child_count,
- gf_afr_mt_char);
- if (!local->transaction.failed_subvols)
- goto out;
-
- local->pending = afr_matrix_create (priv->child_count,
- AFR_NUM_CHANGE_LOGS);
- if (!local->pending)
- goto out;
-
- ret = 0;
- INIT_LIST_HEAD (&local->transaction.wait_list);
- INIT_LIST_HEAD (&local->transaction.owner_list);
+afr_transaction_local_init(afr_local_t *local, xlator_t *this)
+{
+ int ret = -ENOMEM;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ INIT_LIST_HEAD(&local->transaction.wait_list);
+ INIT_LIST_HEAD(&local->transaction.owner_list);
+ INIT_LIST_HEAD(&local->ta_waitq);
+ INIT_LIST_HEAD(&local->ta_onwireq);
+ ret = afr_internal_lock_init(&local->internal_lock, priv->child_count);
+ if (ret < 0)
+ goto out;
+
+ ret = -ENOMEM;
+ local->pre_op_compat = priv->pre_op_compat;
+
+ local->transaction.pre_op = GF_CALLOC(sizeof(*local->transaction.pre_op),
+ priv->child_count, gf_afr_mt_char);
+ if (!local->transaction.pre_op)
+ goto out;
+
+ local->transaction.changelog_xdata = GF_CALLOC(
+ sizeof(*local->transaction.changelog_xdata), priv->child_count,
+ gf_afr_mt_dict_t);
+ if (!local->transaction.changelog_xdata)
+ goto out;
+
+ if (priv->arbiter_count == 1) {
+ local->transaction.pre_op_sources = GF_CALLOC(
+ sizeof(*local->transaction.pre_op_sources), priv->child_count,
+ gf_afr_mt_char);
+ if (!local->transaction.pre_op_sources)
+ goto out;
+ }
+
+ local->transaction.failed_subvols = GF_CALLOC(
+ sizeof(*local->transaction.failed_subvols), priv->child_count,
+ gf_afr_mt_char);
+ if (!local->transaction.failed_subvols)
+ goto out;
+
+ local->pending = afr_matrix_create(priv->child_count, AFR_NUM_CHANGE_LOGS);
+ if (!local->pending)
+ goto out;
+
+ ret = 0;
out:
- return ret;
+ return ret;
}
-
void
-afr_set_low_priority (call_frame_t *frame)
+afr_set_low_priority(call_frame_t *frame)
{
- frame->root->pid = LOW_PRIO_PROC_PID;
+ frame->root->pid = LOW_PRIO_PROC_PID;
}
-
void
-afr_priv_destroy (afr_private_t *priv)
-{
- int i = 0;
- int child_count = -1;
-
- if (!priv)
- goto out;
- GF_FREE (priv->last_event);
-
- child_count = priv->child_count;
- if (priv->thin_arbiter_count) {
- child_count++;
- }
- if (priv->pending_key) {
- for (i = 0; i < child_count; i++)
- GF_FREE (priv->pending_key[i]);
- }
-
- GF_FREE (priv->pending_reads);
- GF_FREE (priv->local);
- GF_FREE (priv->pending_key);
- GF_FREE (priv->children);
- GF_FREE (priv->child_up);
- GF_FREE (priv->child_latency);
- LOCK_DESTROY (&priv->lock);
-
- GF_FREE (priv);
+afr_priv_destroy(afr_private_t *priv)
+{
+ int i = 0;
+ int child_count = -1;
+
+ if (!priv)
+ goto out;
+
+ GF_FREE(priv->sh_domain);
+ GF_FREE(priv->last_event);
+
+ child_count = priv->child_count;
+ if (priv->thin_arbiter_count) {
+ child_count++;
+ }
+ if (priv->pending_key) {
+ for (i = 0; i < child_count; i++)
+ GF_FREE(priv->pending_key[i]);
+ }
+
+ GF_FREE(priv->pending_reads);
+ GF_FREE(priv->local);
+ GF_FREE(priv->pending_key);
+ GF_FREE(priv->children);
+ GF_FREE(priv->anon_inode);
+ GF_FREE(priv->child_up);
+ GF_FREE(priv->halo_child_up);
+ GF_FREE(priv->child_latency);
+ LOCK_DESTROY(&priv->lock);
+
+ GF_FREE(priv);
out:
- return;
+ return;
}
-int**
-afr_mark_pending_changelog (afr_private_t *priv, unsigned char *pending,
- dict_t *xattr, ia_type_t iat)
+int **
+afr_mark_pending_changelog(afr_private_t *priv, unsigned char *pending,
+ dict_t *xattr, ia_type_t iat)
{
- int i = 0;
- int **changelog = NULL;
- int idx = -1;
- int m_idx = 0;
- int d_idx = 0;
- int ret = 0;
+ int i = 0;
+ int **changelog = NULL;
+ int idx = -1;
+ int m_idx = 0;
+ int d_idx = 0;
+ int ret = 0;
- m_idx = afr_index_for_transaction_type (AFR_METADATA_TRANSACTION);
- d_idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
+ m_idx = afr_index_for_transaction_type(AFR_METADATA_TRANSACTION);
+ d_idx = afr_index_for_transaction_type(AFR_DATA_TRANSACTION);
- idx = afr_index_from_ia_type (iat);
+ idx = afr_index_from_ia_type(iat);
- changelog = afr_matrix_create (priv->child_count, AFR_NUM_CHANGE_LOGS);
- if (!changelog)
- goto out;
+ changelog = afr_matrix_create(priv->child_count, AFR_NUM_CHANGE_LOGS);
+ if (!changelog)
+ goto out;
- for (i = 0; i < priv->child_count; i++) {
- if (!pending[i])
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!pending[i])
+ continue;
- changelog[i][m_idx] = hton32(1);
- if (idx != -1)
- changelog[i][idx] = hton32(1);
- /* If the newentry marking is on a newly created directory,
- * then mark it with the full-heal indicator.
- */
- if ((IA_ISDIR (iat)) && (priv->esh_granular))
- changelog[i][d_idx] = hton32(1);
- }
- ret = afr_set_pending_dict (priv, xattr, changelog);
- if (ret < 0) {
- afr_matrix_cleanup (changelog, priv->child_count);
- return NULL;
- }
+ changelog[i][m_idx] = hton32(1);
+ if (idx != -1)
+ changelog[i][idx] = hton32(1);
+ /* If the newentry marking is on a newly created directory,
+ * then mark it with the full-heal indicator.
+ */
+ if ((IA_ISDIR(iat)) && (priv->esh_granular))
+ changelog[i][d_idx] = hton32(1);
+ }
+ ret = afr_set_pending_dict(priv, xattr, changelog);
+ if (ret < 0) {
+ afr_matrix_cleanup(changelog, priv->child_count);
+ return NULL;
+ }
out:
- return changelog;
+ return changelog;
}
-gf_boolean_t
-afr_decide_heal_info (afr_private_t *priv, unsigned char *sources, int source)
+static dict_t *
+afr_set_heal_info(char *status)
{
- int sources_count = 0;
+ dict_t *dict = NULL;
+ int ret = -1;
- if (source < 0)
- goto out;
-
- sources_count = AFR_COUNT (sources, priv->child_count);
- if (sources_count == priv->child_count)
- return _gf_false;
+ dict = dict_new();
+ if (!dict) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ ret = dict_set_dynstr_sizen(dict, "heal-info", status);
+ if (ret)
+ gf_msg("", GF_LOG_WARNING, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Failed to set heal-info key to "
+ "%s",
+ status);
out:
- return _gf_true;
-}
+ /* Any error other than EINVAL, dict_set_dynstr frees status */
+ if (ret == -ENOMEM || ret == -EINVAL) {
+ GF_FREE(status);
+ }
-int
-afr_selfheal_locked_metadata_inspect (call_frame_t *frame, xlator_t *this,
- inode_t *inode, gf_boolean_t *msh,
- gf_boolean_t *pending)
-{
- int ret = -1;
- unsigned char *locked_on = NULL;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *healed_sinks = NULL;
- unsigned char *undid_pending = NULL;
- struct afr_reply *locked_replies = NULL;
-
- afr_private_t *priv = this->private;
-
- locked_on = alloca0 (priv->child_count);
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
- undid_pending = alloca0 (priv->child_count);
-
- locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
-
- ret = afr_selfheal_inodelk (frame, this, inode, this->name,
- LLONG_MAX - 1, 0, locked_on);
- {
- if (ret == 0) {
- /* Not a single lock */
- ret = -afr_final_errno (frame->local, priv);
- if (ret == 0)
- ret = -ENOTCONN;/* all invalid responses */
- goto out;
- }
- ret = __afr_selfheal_metadata_prepare (frame, this, inode,
- locked_on, sources,
- sinks, healed_sinks,
- undid_pending,
- locked_replies,
- pending);
- *msh = afr_decide_heal_info (priv, sources, ret);
- }
- afr_selfheal_uninodelk (frame, this, inode, this->name,
- LLONG_MAX - 1, 0, locked_on);
-out:
- if (locked_replies)
- afr_replies_wipe (locked_replies, priv->child_count);
- return ret;
+ if (ret && dict) {
+ dict_unref(dict);
+ dict = NULL;
+ }
+ return dict;
}
-int
-afr_selfheal_locked_data_inspect (call_frame_t *frame, xlator_t *this,
- fd_t *fd, gf_boolean_t *dsh,
- gf_boolean_t *pflag)
-{
- int ret = -1;
- unsigned char *data_lock = NULL;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *healed_sinks = NULL;
- unsigned char *undid_pending = NULL;
- afr_private_t *priv = NULL;
- struct afr_reply *locked_replies = NULL;
- inode_t *inode = fd->inode;
-
- priv = this->private;
- data_lock = alloca0 (priv->child_count);
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
- undid_pending = alloca0 (priv->child_count);
-
- locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
-
- ret = afr_selfheal_inodelk (frame, this, inode, this->name,
- 0, 0, data_lock);
- {
- if (ret == 0) {
- ret = -afr_final_errno (frame->local, priv);
- if (ret == 0)
- ret = -ENOTCONN; /* all invalid responses */
- goto out;
- }
- ret = __afr_selfheal_data_prepare (frame, this, inode,
- data_lock, sources, sinks,
- healed_sinks, undid_pending,
- locked_replies, pflag);
- *dsh = afr_decide_heal_info (priv, sources, ret);
- }
- afr_selfheal_uninodelk (frame, this, inode, this->name, 0, 0,
- data_lock);
-out:
- if (locked_replies)
- afr_replies_wipe (locked_replies, priv->child_count);
- return ret;
-}
+static gf_boolean_t
+afr_is_dirty_count_non_unary_for_txn(xlator_t *this, struct afr_reply *replies,
+ afr_transaction_type type)
+{
+ afr_private_t *priv = this->private;
+ int *dirty = alloca0(priv->child_count * sizeof(int));
+ int i = 0;
-int
-afr_selfheal_locked_entry_inspect (call_frame_t *frame, xlator_t *this,
- inode_t *inode,
- gf_boolean_t *esh, gf_boolean_t *pflag)
-{
- int ret = -1;
- int source = -1;
- afr_private_t *priv = NULL;
- unsigned char *locked_on = NULL;
- unsigned char *data_lock = NULL;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *healed_sinks = NULL;
- struct afr_reply *locked_replies = NULL;
- gf_boolean_t granular_locks = _gf_false;
-
- priv = this->private;
- if (strcmp ("granular", priv->locking_scheme) == 0)
- granular_locks = _gf_true;
- locked_on = alloca0 (priv->child_count);
- data_lock = alloca0 (priv->child_count);
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
-
- locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
-
- if (!granular_locks) {
- ret = afr_selfheal_tryentrylk (frame, this, inode,
- priv->sh_domain, NULL, locked_on);
- }
- {
- if (!granular_locks && ret == 0) {
- ret = -afr_final_errno (frame->local, priv);
- if (ret == 0)
- ret = -ENOTCONN;/* all invalid responses */
- goto out;
- }
+ afr_selfheal_extract_xattr(this, replies, type, dirty, NULL);
+ for (i = 0; i < priv->child_count; i++) {
+ if (dirty[i] > 1)
+ return _gf_true;
+ }
- ret = afr_selfheal_entrylk (frame, this, inode, this->name,
- NULL, data_lock);
- {
- if (ret == 0) {
- ret = -afr_final_errno (frame->local, priv);
- if (ret == 0)
- ret = -ENOTCONN;
- /* all invalid responses */
- goto unlock;
- }
- ret = __afr_selfheal_entry_prepare (frame, this, inode,
- data_lock, sources,
- sinks, healed_sinks,
- locked_replies,
- &source, pflag);
- if ((ret == 0) && source < 0)
- ret = -EIO;
- *esh = afr_decide_heal_info (priv, sources, ret);
- }
- afr_selfheal_unentrylk (frame, this, inode, this->name, NULL,
- data_lock, NULL);
- }
-unlock:
- if (!granular_locks)
- afr_selfheal_unentrylk (frame, this, inode, priv->sh_domain,
- NULL, locked_on, NULL);
-out:
- if (locked_replies)
- afr_replies_wipe (locked_replies, priv->child_count);
- return ret;
+ return _gf_false;
}
-int
-afr_selfheal_locked_inspect (call_frame_t *frame, xlator_t *this, uuid_t gfid,
- inode_t **inode,
- gf_boolean_t *entry_selfheal,
- gf_boolean_t *data_selfheal,
- gf_boolean_t *metadata_selfheal,
- gf_boolean_t *pending)
-
-{
- int ret = -1;
- fd_t *fd = NULL;
- gf_boolean_t dsh = _gf_false;
- gf_boolean_t msh = _gf_false;
- gf_boolean_t esh = _gf_false;
-
- ret = afr_selfheal_unlocked_inspect (frame, this, gfid, inode,
- &dsh, &msh, &esh);
- if (ret)
- goto out;
-
- /* For every heal type hold locks and check if it indeed needs heal */
-
+static gf_boolean_t
+afr_is_dirty_count_non_unary(xlator_t *this, struct afr_reply *replies,
+ ia_type_t ia_type)
+{
+ gf_boolean_t data_chk = _gf_false;
+ gf_boolean_t mdata_chk = _gf_false;
+ gf_boolean_t entry_chk = _gf_false;
+
+ switch (ia_type) {
+ case IA_IFDIR:
+ mdata_chk = _gf_true;
+ entry_chk = _gf_true;
+ break;
+ case IA_IFREG:
+ mdata_chk = _gf_true;
+ data_chk = _gf_true;
+ break;
+ default:
+ /*IA_IFBLK, IA_IFCHR, IA_IFLNK, IA_IFIFO, IA_IFSOCK*/
+ mdata_chk = _gf_true;
+ break;
+ }
- /* Heal-info does an open() on the file being examined so that the
- * current eager-lock holding client, if present, at some point sees
- * open-fd count being > 1 and releases the eager-lock so that heal-info
- * doesn't remain blocked forever until IO completes.
- */
- if ((*inode)->ia_type == IA_IFREG) {
- ret = afr_selfheal_data_open (this, *inode, &fd);
- if (ret < 0) {
- gf_msg_debug (this->name, -ret, "%s: Failed to open",
- uuid_utoa ((*inode)->gfid));
- goto out;
- }
- }
+ if (data_chk && afr_is_dirty_count_non_unary_for_txn(
+ this, replies, AFR_DATA_TRANSACTION)) {
+ return _gf_true;
+ } else if (mdata_chk && afr_is_dirty_count_non_unary_for_txn(
+ this, replies, AFR_METADATA_TRANSACTION)) {
+ return _gf_true;
+ } else if (entry_chk && afr_is_dirty_count_non_unary_for_txn(
+ this, replies, AFR_ENTRY_TRANSACTION)) {
+ return _gf_true;
+ }
- if (msh) {
- ret = afr_selfheal_locked_metadata_inspect (frame, this,
- *inode, &msh,
- pending);
- if (ret == -EIO)
- goto out;
- }
+ return _gf_false;
+}
- if (dsh) {
- ret = afr_selfheal_locked_data_inspect (frame, this, fd,
- &dsh, pending);
- if (ret == -EIO || (ret == -EAGAIN))
- goto out;
+static int
+afr_update_heal_status(xlator_t *this, struct afr_reply *replies,
+ ia_type_t ia_type, gf_boolean_t *esh, gf_boolean_t *dsh,
+ gf_boolean_t *msh, unsigned char pending)
+{
+ int ret = -1;
+ GF_UNUSED int ret1 = 0;
+ int i = 0;
+ int io_domain_lk_count = 0;
+ int shd_domain_lk_count = 0;
+ afr_private_t *priv = NULL;
+ char *key1 = NULL;
+ char *key2 = NULL;
+
+ priv = this->private;
+ key1 = alloca0(strlen(GLUSTERFS_INODELK_DOM_PREFIX) + 2 +
+ strlen(this->name));
+ key2 = alloca0(strlen(GLUSTERFS_INODELK_DOM_PREFIX) + 2 +
+ strlen(priv->sh_domain));
+ sprintf(key1, "%s:%s", GLUSTERFS_INODELK_DOM_PREFIX, this->name);
+ sprintf(key2, "%s:%s", GLUSTERFS_INODELK_DOM_PREFIX, priv->sh_domain);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if ((replies[i].valid != 1) || (replies[i].op_ret != 0))
+ continue;
+ if (!io_domain_lk_count) {
+ ret1 = dict_get_int32(replies[i].xdata, key1, &io_domain_lk_count);
+ }
+ if (!shd_domain_lk_count) {
+ ret1 = dict_get_int32(replies[i].xdata, key2, &shd_domain_lk_count);
+ }
+ }
+
+ if (!pending) {
+ if ((afr_is_dirty_count_non_unary(this, replies, ia_type)) ||
+ (!io_domain_lk_count)) {
+ /* Needs heal. */
+ ret = 0;
+ } else {
+ /* No heal needed. */
+ *dsh = *esh = *msh = 0;
}
-
- if (esh) {
- ret = afr_selfheal_locked_entry_inspect (frame, this, *inode,
- &esh, pending);
+ } else {
+ if (shd_domain_lk_count) {
+ ret = -EAGAIN; /*For 'possibly-healing'. */
+ } else {
+ ret = 0; /*needs heal. Just set a non -ve value so that it is
+ assumed as the source index.*/
}
-
-out:
- *data_selfheal = dsh;
- *entry_selfheal = esh;
- *metadata_selfheal = msh;
- if (fd)
- fd_unref (fd);
- return ret;
+ }
+ return ret;
}
-dict_t*
-afr_set_heal_info (char *status)
-{
- dict_t *dict = NULL;
- int ret = -1;
-
- dict = dict_new ();
- if (!dict) {
- ret = -ENOMEM;
- goto out;
- }
-
- ret = dict_set_str (dict, "heal-info", status);
+/*return EIO, EAGAIN or pending*/
+int
+afr_lockless_inspect(call_frame_t *frame, xlator_t *this, uuid_t gfid,
+ inode_t **inode, gf_boolean_t *entry_selfheal,
+ gf_boolean_t *data_selfheal,
+ gf_boolean_t *metadata_selfheal, unsigned char *pending)
+{
+ int ret = -1;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ struct afr_reply *replies = NULL;
+ gf_boolean_t dsh = _gf_false;
+ gf_boolean_t msh = _gf_false;
+ gf_boolean_t esh = _gf_false;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ unsigned char *valid_on = NULL;
+ uint64_t *witness = NULL;
+
+ priv = this->private;
+ replies = alloca0(sizeof(*replies) * priv->child_count);
+ sources = alloca0(sizeof(*sources) * priv->child_count);
+ sinks = alloca0(sizeof(*sinks) * priv->child_count);
+ witness = alloca0(sizeof(*witness) * priv->child_count);
+ valid_on = alloca0(sizeof(*valid_on) * priv->child_count);
+
+ ret = afr_selfheal_unlocked_inspect(frame, this, gfid, inode, &dsh, &msh,
+ &esh, replies);
+ if (ret)
+ goto out;
+ for (i = 0; i < priv->child_count; i++) {
+ if (replies[i].valid && replies[i].op_ret == 0) {
+ valid_on[i] = 1;
+ }
+ }
+ if (msh) {
+ ret = afr_selfheal_find_direction(frame, this, replies,
+ AFR_METADATA_TRANSACTION, valid_on,
+ sources, sinks, witness, pending);
+ if (*pending & PFLAG_SBRAIN)
+ ret = -EIO;
+ if (ret)
+ goto out;
+ }
+ if (dsh) {
+ ret = afr_selfheal_find_direction(frame, this, replies,
+ AFR_DATA_TRANSACTION, valid_on,
+ sources, sinks, witness, pending);
+ if (*pending & PFLAG_SBRAIN)
+ ret = -EIO;
+ if (ret)
+ goto out;
+ }
+ if (esh) {
+ ret = afr_selfheal_find_direction(frame, this, replies,
+ AFR_ENTRY_TRANSACTION, valid_on,
+ sources, sinks, witness, pending);
+ if (*pending & PFLAG_SBRAIN)
+ ret = -EIO;
if (ret)
- gf_msg ("", GF_LOG_WARNING, -ret,
- AFR_MSG_DICT_SET_FAILED,
- "Failed to set heal-info key to "
- "%s", status);
+ goto out;
+ }
+
+ ret = afr_update_heal_status(this, replies, (*inode)->ia_type, &esh, &dsh,
+ &msh, *pending);
out:
- return dict;
+ *data_selfheal = dsh;
+ *entry_selfheal = esh;
+ *metadata_selfheal = msh;
+ if (replies)
+ afr_replies_wipe(replies, priv->child_count);
+ return ret;
}
int
-afr_get_heal_info (call_frame_t *frame, xlator_t *this, loc_t *loc)
-{
- gf_boolean_t data_selfheal = _gf_false;
- gf_boolean_t metadata_selfheal = _gf_false;
- gf_boolean_t entry_selfheal = _gf_false;
- gf_boolean_t pending = _gf_false;
- dict_t *dict = NULL;
- int ret = -1;
- int op_errno = 0;
- int size = 0;
- inode_t *inode = NULL;
- char *substr = NULL;
- char *status = NULL;
-
- ret = afr_selfheal_locked_inspect (frame, this, loc->gfid, &inode,
- &entry_selfheal,
- &data_selfheal, &metadata_selfheal,
- &pending);
-
- if (ret == -ENOMEM) {
- op_errno = -ret;
+afr_get_heal_info(call_frame_t *frame, xlator_t *this, loc_t *loc)
+{
+ gf_boolean_t data_selfheal = _gf_false;
+ gf_boolean_t metadata_selfheal = _gf_false;
+ gf_boolean_t entry_selfheal = _gf_false;
+ unsigned char pending = 0;
+ dict_t *dict = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+ inode_t *inode = NULL;
+ char *substr = NULL;
+ char *status = NULL;
+ call_frame_t *heal_frame = NULL;
+ afr_local_t *heal_local = NULL;
+
+ /*Use frame with lk-owner set*/
+ heal_frame = afr_frame_create(frame->this, &op_errno);
+ if (!heal_frame) {
+ ret = -1;
+ goto out;
+ }
+ heal_local = heal_frame->local;
+ heal_frame->local = frame->local;
+
+ ret = afr_lockless_inspect(heal_frame, this, loc->gfid, &inode,
+ &entry_selfheal, &data_selfheal,
+ &metadata_selfheal, &pending);
+
+ if (ret == -ENOMEM) {
+ ret = -1;
+ goto out;
+ }
+
+ if (pending & PFLAG_PENDING) {
+ gf_asprintf(&substr, "-pending");
+ if (!substr)
+ goto out;
+ }
+
+ if (ret == -EIO) {
+ ret = gf_asprintf(&status, "split-brain%s", substr ? substr : "");
+ if (ret < 0) {
+ goto out;
+ }
+ dict = afr_set_heal_info(status);
+ if (!dict) {
+ ret = -1;
+ goto out;
+ }
+ } else if (ret == -EAGAIN) {
+ ret = gf_asprintf(&status, "possibly-healing%s", substr ? substr : "");
+ if (ret < 0) {
+ goto out;
+ }
+ dict = afr_set_heal_info(status);
+ if (!dict) {
+ ret = -1;
+ goto out;
+ }
+ } else if (ret >= 0) {
+ /* value of ret = source index
+ * so ret >= 0 and at least one of the 3 booleans set to
+ * true means a source is identified; heal is required.
+ */
+ if (!data_selfheal && !entry_selfheal && !metadata_selfheal) {
+ status = gf_strdup("no-heal");
+ if (!status) {
ret = -1;
goto out;
+ }
+ dict = afr_set_heal_info(status);
+ if (!dict) {
+ ret = -1;
+ goto out;
+ }
+ } else {
+ ret = gf_asprintf(&status, "heal%s", substr ? substr : "");
+ if (ret < 0) {
+ goto out;
+ }
+ dict = afr_set_heal_info(status);
+ if (!dict) {
+ ret = -1;
+ goto out;
+ }
}
-
- if (pending) {
- size = strlen ("-pending") + 1;
- gf_asprintf (&substr, "-pending");
- if (!substr)
- goto out;
+ } else if (ret < 0) {
+ /* Apart from above checked -ve ret values, there are
+ * other possible ret values like ENOTCONN
+ * (returned when number of valid replies received are
+ * less than 2)
+ * in which case heal is required when one of the
+ * selfheal booleans is set.
+ */
+ if (data_selfheal || entry_selfheal || metadata_selfheal) {
+ ret = gf_asprintf(&status, "heal%s", substr ? substr : "");
+ if (ret < 0) {
+ goto out;
+ }
+ dict = afr_set_heal_info(status);
+ if (!dict) {
+ ret = -1;
+ goto out;
+ }
}
+ }
- if (ret == -EIO) {
- size += strlen ("split-brain") + 1;
- ret = gf_asprintf (&status, "split-brain%s",
- substr? substr : "");
- if (ret < 0)
- goto out;
- dict = afr_set_heal_info (status);
- } else if (ret == -EAGAIN) {
- size += strlen ("possibly-healing") + 1;
- ret = gf_asprintf (&status, "possibly-healing%s",
- substr? substr : "");
- if (ret < 0)
- goto out;
- dict = afr_set_heal_info (status);
- } else if (ret >= 0) {
- /* value of ret = source index
- * so ret >= 0 and at least one of the 3 booleans set to
- * true means a source is identified; heal is required.
- */
- if (!data_selfheal && !entry_selfheal &&
- !metadata_selfheal) {
- dict = afr_set_heal_info ("no-heal");
- } else {
- size += strlen ("heal") + 1;
- ret = gf_asprintf (&status, "heal%s",
- substr? substr : "");
- if (ret < 0)
- goto out;
- dict = afr_set_heal_info (status);
- }
- } else if (ret < 0) {
- /* Apart from above checked -ve ret values, there are
- * other possible ret values like ENOTCONN
- * (returned when number of valid replies received are
- * less than 2)
- * in which case heal is required when one of the
- * selfheal booleans is set.
- */
- if (data_selfheal || entry_selfheal ||
- metadata_selfheal) {
- size += strlen ("heal") + 1;
- ret = gf_asprintf (&status, "heal%s",
- substr? substr : "");
- if (ret < 0)
- goto out;
- dict = afr_set_heal_info (status);
- }
- }
- ret = 0;
+ ret = 0;
+ op_errno = 0;
out:
- AFR_STACK_UNWIND (getxattr, frame, ret, op_errno, dict, NULL);
- if (dict)
- dict_unref (dict);
- if (inode)
- inode_unref (inode);
- GF_FREE (substr);
- return ret;
+ if (heal_frame) {
+ heal_frame->local = heal_local;
+ AFR_STACK_DESTROY(heal_frame);
+ }
+ AFR_STACK_UNWIND(getxattr, frame, ret, op_errno, dict, NULL);
+ if (dict)
+ dict_unref(dict);
+ if (inode)
+ inode_unref(inode);
+ GF_FREE(substr);
+ return ret;
}
int
-_afr_is_split_brain (call_frame_t *frame, xlator_t *this,
- struct afr_reply *replies,
- afr_transaction_type type,
- gf_boolean_t *spb)
-{
- afr_private_t *priv = NULL;
- uint64_t *witness = NULL;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- int sources_count = 0;
- int ret = 0;
-
- priv = this->private;
-
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- witness = alloca0(priv->child_count * sizeof (*witness));
-
- ret = afr_selfheal_find_direction (frame, this, replies,
- type, priv->child_up, sources,
- sinks, witness, NULL);
- if (ret)
- return ret;
+_afr_is_split_brain(call_frame_t *frame, xlator_t *this,
+ struct afr_reply *replies, afr_transaction_type type,
+ gf_boolean_t *spb)
+{
+ afr_private_t *priv = NULL;
+ uint64_t *witness = NULL;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ int sources_count = 0;
+ int ret = 0;
+
+ priv = this->private;
+
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ witness = alloca0(priv->child_count * sizeof(*witness));
+
+ ret = afr_selfheal_find_direction(frame, this, replies, type,
+ priv->child_up, sources, sinks, witness,
+ NULL);
+ if (ret)
+ return ret;
- sources_count = AFR_COUNT (sources, priv->child_count);
- if (!sources_count)
- *spb = _gf_true;
+ sources_count = AFR_COUNT(sources, priv->child_count);
+ if (!sources_count)
+ *spb = _gf_true;
- return ret;
+ return ret;
}
int
-afr_is_split_brain (call_frame_t *frame, xlator_t *this, inode_t *inode,
- uuid_t gfid, gf_boolean_t *d_spb, gf_boolean_t *m_spb)
+afr_is_split_brain(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ uuid_t gfid, gf_boolean_t *d_spb, gf_boolean_t *m_spb)
{
- int ret = -1;
- afr_private_t *priv = NULL;
- struct afr_reply *replies = NULL;
+ int ret = -1;
+ afr_private_t *priv = NULL;
+ struct afr_reply *replies = NULL;
- priv = this->private;
+ priv = this->private;
- replies = alloca0 (sizeof (*replies) * priv->child_count);
+ replies = alloca0(sizeof(*replies) * priv->child_count);
- ret = afr_selfheal_unlocked_discover (frame, inode, gfid, replies);
- if (ret)
- goto out;
+ ret = afr_selfheal_unlocked_discover(frame, inode, gfid, replies);
+ if (ret)
+ goto out;
- if (!afr_can_decide_split_brain_source_sinks (replies,
- priv->child_count)) {
- ret = -EAGAIN;
- goto out;
- }
+ if (!afr_can_decide_split_brain_source_sinks(replies, priv->child_count)) {
+ ret = -EAGAIN;
+ goto out;
+ }
- ret = _afr_is_split_brain (frame, this, replies,
- AFR_DATA_TRANSACTION, d_spb);
- if (ret)
- goto out;
+ ret = _afr_is_split_brain(frame, this, replies, AFR_DATA_TRANSACTION,
+ d_spb);
+ if (ret)
+ goto out;
- ret = _afr_is_split_brain (frame, this, replies,
- AFR_METADATA_TRANSACTION, m_spb);
+ ret = _afr_is_split_brain(frame, this, replies, AFR_METADATA_TRANSACTION,
+ m_spb);
out:
- if (replies) {
- afr_replies_wipe (replies, priv->child_count);
- replies = NULL;
- }
- return ret;
+ if (replies) {
+ afr_replies_wipe(replies, priv->child_count);
+ replies = NULL;
+ }
+ return ret;
}
int
-afr_get_split_brain_status_cbk (int ret, call_frame_t *frame, void *opaque)
+afr_get_split_brain_status_cbk(int ret, call_frame_t *frame, void *opaque)
{
- GF_FREE (opaque);
- return 0;
+ GF_FREE(opaque);
+ return 0;
}
int
-afr_get_split_brain_status (void *opaque)
-{
- gf_boolean_t d_spb = _gf_false;
- gf_boolean_t m_spb = _gf_false;
- int ret = -1;
- int op_errno = 0;
- int i = 0;
- char *choices = NULL;
- char *status = NULL;
- dict_t *dict = NULL;
- inode_t *inode = NULL;
- afr_private_t *priv = NULL;
- xlator_t **children = NULL;
- call_frame_t *frame = NULL;
- xlator_t *this = NULL;
- loc_t *loc = NULL;
- afr_spb_status_t *data = NULL;
-
- data = opaque;
- frame = data->frame;
- this = frame->this;
- loc = data->loc;
- priv = this->private;
- children = priv->children;
-
- inode = afr_inode_find (this, loc->gfid);
- if (!inode)
- goto out;
+afr_get_split_brain_status(void *opaque)
+{
+ gf_boolean_t d_spb = _gf_false;
+ gf_boolean_t m_spb = _gf_false;
+ int ret = -1;
+ int op_errno = 0;
+ int i = 0;
+ char *choices = NULL;
+ char *status = NULL;
+ dict_t *dict = NULL;
+ inode_t *inode = NULL;
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ call_frame_t *frame = NULL;
+ xlator_t *this = NULL;
+ loc_t *loc = NULL;
+ afr_spb_status_t *data = NULL;
+
+ data = opaque;
+ frame = data->frame;
+ this = frame->this;
+ loc = data->loc;
+ priv = this->private;
+ children = priv->children;
+
+ inode = afr_inode_find(this, loc->gfid);
+ if (!inode)
+ goto out;
+
+ dict = dict_new();
+ if (!dict) {
+ op_errno = ENOMEM;
+ ret = -1;
+ goto out;
+ }
+
+ /* Calculation for string length :
+ * (child_count X length of child-name) + SLEN(" Choices :")
+ * child-name consists of :
+ * a) 251 = max characters for volname according to GD_VOLUME_NAME_MAX
+ * b) strlen("-client-00,") assuming 16 replicas
+ */
+ choices = alloca0(priv->child_count * (256 + SLEN("-client-00,")) +
+ SLEN(" Choices:"));
+
+ ret = afr_is_split_brain(frame, this, inode, loc->gfid, &d_spb, &m_spb);
+ if (ret) {
+ op_errno = -ret;
+ if (ret == -EAGAIN) {
+ ret = dict_set_sizen_str_sizen(dict, GF_AFR_SBRAIN_STATUS,
+ SBRAIN_HEAL_NO_GO_MSG);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret,
+ AFR_MSG_DICT_SET_FAILED,
+ "Failed to set GF_AFR_SBRAIN_STATUS in dict");
+ }
+ }
+ ret = -1;
+ goto out;
+ }
- dict = dict_new ();
- if (!dict) {
- op_errno = ENOMEM;
- ret = -1;
- goto out;
+ if (d_spb || m_spb) {
+ sprintf(choices, " Choices:");
+ for (i = 0; i < priv->child_count; i++) {
+ strcat(choices, children[i]->name);
+ strcat(choices, ",");
}
+ choices[strlen(choices) - 1] = '\0';
- /* Calculation for string length :
- * (child_count X length of child-name) + strlen (" Choices :")
- * child-name consists of :
- * a) 256 = max characters for volname according to GD_VOLUME_NAME_MAX
- * b) strlen ("-client-00,") assuming 16 replicas
- */
- choices = alloca0 (priv->child_count * (256 + strlen ("-client-00,")) +
- strlen (" Choices:"));
+ ret = gf_asprintf(&status,
+ "data-split-brain:%s "
+ "metadata-split-brain:%s%s",
+ (d_spb) ? "yes" : "no", (m_spb) ? "yes" : "no",
+ choices);
- ret = afr_is_split_brain (frame, this, inode, loc->gfid, &d_spb,
- &m_spb);
+ if (-1 == ret) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+ ret = dict_set_dynstr_sizen(dict, GF_AFR_SBRAIN_STATUS, status);
if (ret) {
- op_errno = -ret;
- if (ret == -EAGAIN) {
- ret = dict_set_str (dict, GF_AFR_SBRAIN_STATUS,
- SBRAIN_HEAL_NO_GO_MSG);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Failed to set GF_AFR_SBRAIN_STATUS in dict");
- }
- }
- ret = -1;
- goto out;
+ op_errno = -ret;
+ ret = -1;
+ goto out;
}
-
- if (d_spb || m_spb) {
- sprintf (choices, " Choices:");
- for (i = 0; i < priv->child_count; i++) {
- strcat (choices, children[i]->name);
- strcat (choices, ",");
- }
- choices[strlen (choices) - 1] = '\0';
-
- ret = gf_asprintf (&status, "data-split-brain:%s "
- "metadata-split-brain:%s%s",
- (d_spb) ? "yes" : "no",
- (m_spb) ? "yes" : "no", choices);
-
- if (-1 == ret) {
- op_errno = ENOMEM;
- goto out;
- }
- ret = dict_set_dynstr (dict, GF_AFR_SBRAIN_STATUS, status);
- if (ret) {
- op_errno = -ret;
- ret = -1;
- goto out;
- }
- } else {
- ret = dict_set_str (dict, GF_AFR_SBRAIN_STATUS,
- "The file is not under data or"
- " metadata split-brain");
- if (ret) {
- op_errno = -ret;
- ret = -1;
- goto out;
- }
+ } else {
+ ret = dict_set_sizen_str_sizen(dict, GF_AFR_SBRAIN_STATUS,
+ SFILE_NOT_UNDER_DATA);
+ if (ret) {
+ op_errno = -ret;
+ ret = -1;
+ goto out;
}
+ }
- ret = 0;
+ ret = 0;
out:
- AFR_STACK_UNWIND (getxattr, frame, ret, op_errno, dict, NULL);
- if (dict)
- dict_unref (dict);
- if (inode)
- inode_unref (inode);
- return ret;
+ AFR_STACK_UNWIND(getxattr, frame, ret, op_errno, dict, NULL);
+ if (dict)
+ dict_unref(dict);
+ if (inode)
+ inode_unref(inode);
+ return ret;
}
int32_t
afr_heal_splitbrain_file(call_frame_t *frame, xlator_t *this, loc_t *loc)
{
- int ret = 0;
- int op_errno = 0;
- dict_t *dict = NULL;
- afr_local_t *local = NULL;
-
- local = frame->local;
- dict = dict_new ();
- if (!dict) {
- op_errno = ENOMEM;
- ret = -1;
- goto out;
- }
-
- ret = afr_selfheal_do (frame, this, loc->gfid);
-
- if (ret == 1 || ret == 2) {
- ret = dict_set_str (dict, "sh-fail-msg",
- "File not in split-brain");
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Failed to set sh-fail-msg in dict");
- ret = 0;
- goto out;
- } else {
- if (local->xdata_rsp) {
- /* 'sh-fail-msg' has been set in the dict during self-heal.*/
- dict_copy (local->xdata_rsp, dict);
- ret = 0;
- } else if (ret < 0) {
- op_errno = -ret;
- ret = -1;
- }
+ int ret = 0;
+ int op_errno = 0;
+ dict_t *dict = NULL;
+ afr_local_t *local = NULL;
+ afr_local_t *heal_local = NULL;
+ call_frame_t *heal_frame = NULL;
+
+ local = frame->local;
+ dict = dict_new();
+ if (!dict) {
+ op_errno = ENOMEM;
+ ret = -1;
+ goto out;
+ }
+
+ heal_frame = afr_frame_create(this, &op_errno);
+ if (!heal_frame) {
+ ret = -1;
+ goto out;
+ }
+ heal_local = heal_frame->local;
+ heal_frame->local = frame->local;
+ /*Initiate heal with heal_frame with lk-owner set so that inodelk/entrylk
+ * work correctly*/
+ ret = afr_selfheal_do(heal_frame, this, loc->gfid);
+
+ if (ret == 1 || ret == 2) {
+ ret = dict_set_sizen_str_sizen(dict, "sh-fail-msg",
+ SFILE_NOT_IN_SPLIT_BRAIN);
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Failed to set sh-fail-msg in dict");
+ ret = 0;
+ goto out;
+ } else {
+ if (local->xdata_rsp) {
+ /* 'sh-fail-msg' has been set in the dict during self-heal.*/
+ dict_copy(local->xdata_rsp, dict);
+ ret = 0;
+ } else if (ret < 0) {
+ op_errno = -ret;
+ ret = -1;
}
+ }
out:
- if (local->op == GF_FOP_GETXATTR)
- AFR_STACK_UNWIND (getxattr, frame, ret, op_errno, dict, NULL);
- else if (local->op == GF_FOP_SETXATTR)
- AFR_STACK_UNWIND (setxattr, frame, ret, op_errno, NULL);
- if (dict)
- dict_unref(dict);
- return ret;
+ if (heal_frame) {
+ heal_frame->local = heal_local;
+ AFR_STACK_DESTROY(heal_frame);
+ }
+ if (local->op == GF_FOP_GETXATTR)
+ AFR_STACK_UNWIND(getxattr, frame, ret, op_errno, dict, NULL);
+ else if (local->op == GF_FOP_SETXATTR)
+ AFR_STACK_UNWIND(setxattr, frame, ret, op_errno, NULL);
+ if (dict)
+ dict_unref(dict);
+ return ret;
}
int
-afr_get_child_index_from_name (xlator_t *this, char *name)
+afr_get_child_index_from_name(xlator_t *this, char *name)
{
- afr_private_t *priv = this->private;
- int index = -1;
+ afr_private_t *priv = this->private;
+ int index = -1;
- for (index = 0; index < priv->child_count; index++) {
- if (!strcmp (priv->children[index]->name, name))
- goto out;
- }
- index = -1;
+ for (index = 0; index < priv->child_count; index++) {
+ if (!strcmp(priv->children[index]->name, name))
+ goto out;
+ }
+ index = -1;
out:
- return index;
+ return index;
}
void
-afr_priv_need_heal_set (afr_private_t *priv, gf_boolean_t need_heal)
+afr_priv_need_heal_set(afr_private_t *priv, gf_boolean_t need_heal)
{
- LOCK (&priv->lock);
- {
- priv->need_heal = need_heal;
- }
- UNLOCK (&priv->lock);
+ LOCK(&priv->lock);
+ {
+ priv->need_heal = need_heal;
+ }
+ UNLOCK(&priv->lock);
}
void
-afr_set_need_heal (xlator_t *this, afr_local_t *local)
+afr_set_need_heal(xlator_t *this, afr_local_t *local)
{
- int i = 0;
- afr_private_t *priv = this->private;
- gf_boolean_t need_heal = _gf_false;
+ int i = 0;
+ afr_private_t *priv = this->private;
+ gf_boolean_t need_heal = _gf_false;
- for (i = 0; i < priv->child_count; i++) {
- if (local->replies[i].valid && local->replies[i].need_heal) {
- need_heal = _gf_true;
- break;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->replies[i].valid && local->replies[i].need_heal) {
+ need_heal = _gf_true;
+ break;
}
- afr_priv_need_heal_set (priv, need_heal);
- return;
+ }
+ afr_priv_need_heal_set(priv, need_heal);
+ return;
}
gf_boolean_t
-afr_get_need_heal (xlator_t *this)
+afr_get_need_heal(xlator_t *this)
{
- afr_private_t *priv = this->private;
- gf_boolean_t need_heal = _gf_true;
+ afr_private_t *priv = this->private;
+ gf_boolean_t need_heal = _gf_true;
- LOCK (&priv->lock);
- {
- need_heal = priv->need_heal;
- }
- UNLOCK (&priv->lock);
- return need_heal;
+ LOCK(&priv->lock);
+ {
+ need_heal = priv->need_heal;
+ }
+ UNLOCK(&priv->lock);
+ return need_heal;
}
int
-afr_get_msg_id (char *op_type)
+afr_get_msg_id(char *op_type)
{
-
- if (!strcmp (op_type, GF_AFR_REPLACE_BRICK))
- return AFR_MSG_REPLACE_BRICK_STATUS;
- else if (!strcmp (op_type, GF_AFR_ADD_BRICK))
- return AFR_MSG_ADD_BRICK_STATUS;
- return -1;
+ if (!strcmp(op_type, GF_AFR_REPLACE_BRICK))
+ return AFR_MSG_REPLACE_BRICK_STATUS;
+ else if (!strcmp(op_type, GF_AFR_ADD_BRICK))
+ return AFR_MSG_ADD_BRICK_STATUS;
+ return -1;
}
int
-afr_fav_child_reset_sink_xattrs_cbk (int ret, call_frame_t *heal_frame,
- void *opaque)
+afr_fav_child_reset_sink_xattrs_cbk(int ret, call_frame_t *heal_frame,
+ void *opaque)
{
+ call_frame_t *txn_frame = NULL;
+ afr_local_t *local = NULL;
+ afr_local_t *heal_local = NULL;
+ xlator_t *this = NULL;
- call_frame_t *txn_frame = NULL;
- afr_local_t *local = NULL;
- afr_local_t *heal_local = NULL;
- xlator_t *this = NULL;
+ heal_local = heal_frame->local;
+ txn_frame = heal_local->heal_frame;
+ local = txn_frame->local;
+ this = txn_frame->this;
- heal_local = heal_frame->local;
- txn_frame = heal_local->heal_frame;
- local = txn_frame->local;
- this = txn_frame->this;
-
- /* Refresh the inode agan and proceed with the transaction.*/
- afr_inode_refresh (txn_frame, this, local->inode, NULL,
- local->refreshfn);
+ /* Refresh the inode agan and proceed with the transaction.*/
+ afr_inode_refresh(txn_frame, this, local->inode, NULL, local->refreshfn);
- AFR_STACK_DESTROY (heal_frame);
+ AFR_STACK_DESTROY(heal_frame);
- return 0;
+ return 0;
}
int
-afr_fav_child_reset_sink_xattrs (void *opaque)
-{
- call_frame_t *heal_frame = NULL;
- call_frame_t *txn_frame = NULL;
- xlator_t *this = NULL;
- gf_boolean_t d_spb = _gf_false;
- gf_boolean_t m_spb = _gf_false;
- afr_local_t *heal_local = NULL;
- afr_local_t *txn_local = NULL;
- afr_private_t *priv = NULL;
- inode_t *inode = NULL;
- unsigned char *locked_on = NULL;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *healed_sinks = NULL;
- unsigned char *undid_pending = NULL;
- struct afr_reply *locked_replies = NULL;
- int ret = 0;
-
- heal_frame = (call_frame_t *) opaque;
- heal_local = heal_frame->local;
- txn_frame = heal_local->heal_frame;
- txn_local = txn_frame->local;
- this = txn_frame->this;
- inode = txn_local->inode;
- priv = this->private;
- locked_on = alloca0 (priv->child_count);
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
- undid_pending = alloca0 (priv->child_count);
- locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
-
- ret = _afr_is_split_brain (txn_frame, this, txn_local->replies,
- AFR_DATA_TRANSACTION, &d_spb);
-
- ret = _afr_is_split_brain (txn_frame, this, txn_local->replies,
- AFR_METADATA_TRANSACTION, &m_spb);
-
- /* Take appropriate locks and reset sink xattrs. */
- if (d_spb) {
- ret = afr_selfheal_inodelk (heal_frame, this, inode, this->name,
- 0, 0, locked_on);
- {
- if (ret < AFR_SH_MIN_PARTICIPANTS)
- goto data_unlock;
- ret = __afr_selfheal_data_prepare (heal_frame, this,
- inode, locked_on,
- sources, sinks,
- healed_sinks,
- undid_pending,
- locked_replies,
- NULL);
- }
-data_unlock:
- afr_selfheal_uninodelk (heal_frame, this, inode, this->name,
- 0, 0, locked_on);
- }
-
- if (m_spb) {
- memset (locked_on, 0, sizeof (*locked_on) * priv->child_count);
- memset (undid_pending, 0,
- sizeof (*undid_pending) * priv->child_count);
- ret = afr_selfheal_inodelk (heal_frame, this, inode, this->name,
- LLONG_MAX-1, 0, locked_on);
- {
- if (ret < AFR_SH_MIN_PARTICIPANTS)
- goto mdata_unlock;
- ret = __afr_selfheal_metadata_prepare (heal_frame, this,
- inode, locked_on,
- sources, sinks,
- healed_sinks,
- undid_pending,
- locked_replies,
- NULL);
-
- }
-mdata_unlock:
- afr_selfheal_uninodelk (heal_frame, this, inode, this->name,
- LLONG_MAX-1, 0, locked_on);
+afr_fav_child_reset_sink_xattrs(void *opaque)
+{
+ call_frame_t *heal_frame = NULL;
+ call_frame_t *txn_frame = NULL;
+ xlator_t *this = NULL;
+ gf_boolean_t d_spb = _gf_false;
+ gf_boolean_t m_spb = _gf_false;
+ afr_local_t *heal_local = NULL;
+ afr_local_t *txn_local = NULL;
+ afr_private_t *priv = NULL;
+ inode_t *inode = NULL;
+ unsigned char *locked_on = NULL;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ unsigned char *healed_sinks = NULL;
+ unsigned char *undid_pending = NULL;
+ struct afr_reply *locked_replies = NULL;
+ int ret = 0;
+
+ heal_frame = (call_frame_t *)opaque;
+ heal_local = heal_frame->local;
+ txn_frame = heal_local->heal_frame;
+ txn_local = txn_frame->local;
+ this = txn_frame->this;
+ inode = txn_local->inode;
+ priv = this->private;
+ locked_on = alloca0(priv->child_count);
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ healed_sinks = alloca0(priv->child_count);
+ undid_pending = alloca0(priv->child_count);
+ locked_replies = alloca0(sizeof(*locked_replies) * priv->child_count);
+
+ ret = _afr_is_split_brain(txn_frame, this, txn_local->replies,
+ AFR_DATA_TRANSACTION, &d_spb);
+
+ ret = _afr_is_split_brain(txn_frame, this, txn_local->replies,
+ AFR_METADATA_TRANSACTION, &m_spb);
+
+ /* Take appropriate locks and reset sink xattrs. */
+ if (d_spb) {
+ ret = afr_selfheal_inodelk(heal_frame, this, inode, this->name, 0, 0,
+ locked_on);
+ {
+ if (ret < priv->child_count)
+ goto data_unlock;
+ ret = __afr_selfheal_data_prepare(
+ heal_frame, this, inode, locked_on, sources, sinks,
+ healed_sinks, undid_pending, locked_replies, NULL);
+ }
+ data_unlock:
+ afr_selfheal_uninodelk(heal_frame, this, inode, this->name, 0, 0,
+ locked_on);
+ }
+
+ if (m_spb) {
+ memset(locked_on, 0, sizeof(*locked_on) * priv->child_count);
+ memset(undid_pending, 0, sizeof(*undid_pending) * priv->child_count);
+ ret = afr_selfheal_inodelk(heal_frame, this, inode, this->name,
+ LLONG_MAX - 1, 0, locked_on);
+ {
+ if (ret < priv->child_count)
+ goto mdata_unlock;
+ ret = __afr_selfheal_metadata_prepare(
+ heal_frame, this, inode, locked_on, sources, sinks,
+ healed_sinks, undid_pending, locked_replies, NULL);
}
+ mdata_unlock:
+ afr_selfheal_uninodelk(heal_frame, this, inode, this->name,
+ LLONG_MAX - 1, 0, locked_on);
+ }
- return ret;
-
+ return ret;
}
/*
* Concatenates the xattrs in local->replies separated by a delimiter.
*/
int
-afr_serialize_xattrs_with_delimiter (call_frame_t *frame, xlator_t *this,
- char *buf, const char *default_str,
- int32_t *serz_len, char delimiter)
-{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- char *xattr = NULL;
- int i = 0;
- int len = 0;
- int ret = -1;
-
- priv = this->private;
- local = frame->local;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid || local->replies[i].op_ret) {
- buf = strncat (buf, default_str, strlen (default_str));
- len += strlen (default_str);
- buf[len++] = delimiter;
- buf[len] = '\0';
- } else {
- ret = dict_get_str (local->replies[i].xattr,
- local->cont.getxattr.name, &xattr);
- if (ret) {
- gf_msg ("TEST", GF_LOG_ERROR, -ret,
- AFR_MSG_DICT_GET_FAILED,
- "Failed to get the node_uuid of brick "
- "%d", i);
- goto out;
- }
- buf = strncat (buf, xattr, strlen (xattr));
- len += strlen (xattr);
- buf[len++] = delimiter;
- buf[len] = '\0';
- }
- }
- buf[--len] = '\0'; /*remove the last delimiter*/
- if (serz_len)
- *serz_len = ++len;
- ret = 0;
+afr_serialize_xattrs_with_delimiter(call_frame_t *frame, xlator_t *this,
+ char *buf, const char *default_str,
+ int32_t *serz_len, char delimiter)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ char *xattr = NULL;
+ int i = 0;
+ int len = 0;
+ int keylen = 0;
+ size_t str_len = 0;
+ int ret = -1;
+
+ priv = this->private;
+ local = frame->local;
+
+ keylen = strlen(local->cont.getxattr.name);
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid || local->replies[i].op_ret) {
+ str_len = strlen(default_str);
+ buf = strncat(buf, default_str, str_len);
+ len += str_len;
+ buf[len++] = delimiter;
+ buf[len] = '\0';
+ } else {
+ ret = dict_get_strn(local->replies[i].xattr,
+ local->cont.getxattr.name, keylen, &xattr);
+ if (ret) {
+ gf_msg("TEST", GF_LOG_ERROR, -ret, AFR_MSG_DICT_GET_FAILED,
+ "Failed to get the node_uuid of brick "
+ "%d",
+ i);
+ goto out;
+ }
+ str_len = strlen(xattr);
+ buf = strncat(buf, xattr, str_len);
+ len += str_len;
+ buf[len++] = delimiter;
+ buf[len] = '\0';
+ }
+ }
+ buf[--len] = '\0'; /*remove the last delimiter*/
+ if (serz_len)
+ *serz_len = ++len;
+ ret = 0;
out:
- return ret;
+ return ret;
}
uint64_t
-afr_write_subvol_get (call_frame_t *frame, xlator_t *this)
+afr_write_subvol_get(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- uint64_t write_subvol = 0;
+ afr_local_t *local = NULL;
+ uint64_t write_subvol = 0;
- local = frame->local;
- LOCK(&local->inode->lock);
- write_subvol = local->inode_ctx->write_subvol;
- UNLOCK (&local->inode->lock);
+ local = frame->local;
+ LOCK(&local->inode->lock);
+ write_subvol = local->inode_ctx->write_subvol;
+ UNLOCK(&local->inode->lock);
- return write_subvol;
+ return write_subvol;
}
int
-afr_write_subvol_set (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- unsigned char *data_accused = NULL;
- unsigned char *metadata_accused = NULL;
- unsigned char *data_readable = NULL;
- unsigned char *metadata_readable = NULL;
- uint16_t datamap = 0;
- uint16_t metadatamap = 0;
- uint64_t val = 0;
- int event = 0;
- int i = 0;
+afr_write_subvol_set(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ unsigned char *data_accused = NULL;
+ unsigned char *metadata_accused = NULL;
+ unsigned char *data_readable = NULL;
+ unsigned char *metadata_readable = NULL;
+ uint16_t datamap = 0;
+ uint16_t metadatamap = 0;
+ uint64_t val = 0;
+ int event = 0;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+ data_accused = alloca0(priv->child_count);
+ metadata_accused = alloca0(priv->child_count);
+ data_readable = alloca0(priv->child_count);
+ metadata_readable = alloca0(priv->child_count);
+ event = local->event_generation;
+
+ afr_readables_fill(frame, this, local->inode, data_accused,
+ metadata_accused, data_readable, metadata_readable,
+ NULL);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (data_readable[i])
+ datamap |= (1 << i);
+ if (metadata_readable[i])
+ metadatamap |= (1 << i);
+ }
+
+ val = ((uint64_t)metadatamap) | (((uint64_t)datamap) << 16) |
+ (((uint64_t)event) << 32);
+
+ LOCK(&local->inode->lock);
+ {
+ if (local->inode_ctx->write_subvol == 0 &&
+ local->transaction.type == AFR_DATA_TRANSACTION) {
+ local->inode_ctx->write_subvol = val;
+ }
+ }
+ UNLOCK(&local->inode->lock);
+
+ return 0;
+}
- local = frame->local;
- priv = this->private;
- data_accused = alloca0 (priv->child_count);
- metadata_accused = alloca0 (priv->child_count);
- data_readable = alloca0 (priv->child_count);
- metadata_readable = alloca0 (priv->child_count);
- event = local->event_generation;
+int
+afr_write_subvol_reset(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
- afr_readables_fill (frame, this, local->inode, data_accused,
- metadata_accused, data_readable, metadata_readable,
- NULL);
+ local = frame->local;
+ LOCK(&local->inode->lock);
+ {
+ GF_ASSERT(local->inode_ctx->lock_count > 0);
+ local->inode_ctx->lock_count--;
- for (i = 0; i < priv->child_count; i++) {
- if (data_readable[i])
- datamap |= (1 << i);
- if (metadata_readable[i])
- metadatamap |= (1 << i);
- }
+ if (!local->inode_ctx->lock_count)
+ local->inode_ctx->write_subvol = 0;
+ }
+ UNLOCK(&local->inode->lock);
- val = ((uint64_t) metadatamap) |
- (((uint64_t) datamap) << 16) |
- (((uint64_t) event) << 32);
+ return 0;
+}
- LOCK(&local->inode->lock);
- {
- if (local->inode_ctx->write_subvol == 0 &&
- local->transaction.type == AFR_DATA_TRANSACTION) {
- local->inode_ctx->write_subvol = val;
- }
- }
- UNLOCK (&local->inode->lock);
+int
+afr_set_inode_local(xlator_t *this, afr_local_t *local, inode_t *inode)
+{
+ int ret = 0;
- return 0;
+ local->inode = inode_ref(inode);
+ LOCK(&local->inode->lock);
+ {
+ ret = __afr_inode_ctx_get(this, local->inode, &local->inode_ctx);
+ }
+ UNLOCK(&local->inode->lock);
+ if (ret < 0) {
+ gf_msg_callingfn(
+ this->name, GF_LOG_ERROR, ENOMEM, AFR_MSG_INODE_CTX_GET_FAILED,
+ "Error getting inode ctx %s", uuid_utoa(local->inode->gfid));
+ }
+ return ret;
}
-int
-afr_write_subvol_reset (call_frame_t *frame, xlator_t *this)
+gf_boolean_t
+afr_ta_is_fop_called_from_synctask(xlator_t *this)
{
- afr_local_t *local = NULL;
+ struct synctask *task = NULL;
+ gf_lkowner_t tmp_owner = {
+ 0,
+ };
- local = frame->local;
- LOCK(&local->inode->lock);
- {
- GF_ASSERT (local->inode_ctx->lock_count > 0);
- local->inode_ctx->lock_count--;
+ task = synctask_get();
+ if (!task)
+ return _gf_false;
- if (!local->inode_ctx->lock_count)
- local->inode_ctx->write_subvol = 0;
- }
- UNLOCK(&local->inode->lock);
+ set_lk_owner_from_ptr(&tmp_owner, (void *)this);
- return 0;
+ if (!is_same_lkowner(&tmp_owner, &task->frame->root->lk_owner))
+ return _gf_false;
+
+ return _gf_true;
}
int
-afr_set_inode_local (xlator_t *this, afr_local_t *local, inode_t *inode)
+afr_ta_post_op_lock(xlator_t *this, loc_t *loc)
+{
+ int ret = 0;
+ uuid_t gfid = {
+ 0,
+ };
+ afr_private_t *priv = this->private;
+ gf_boolean_t locked = _gf_false;
+ struct gf_flock flock1 = {
+ 0,
+ };
+ struct gf_flock flock2 = {
+ 0,
+ };
+ int32_t cmd = 0;
+
+ /* Clients must take AFR_TA_DOM_NOTIFY lock only when the previous lock
+ * has been released in afr_notify due to upcall notification from shd.
+ */
+ GF_ASSERT(priv->ta_notify_dom_lock_offset == 0);
+
+ if (!priv->shd.iamshd)
+ GF_ASSERT(afr_ta_is_fop_called_from_synctask(this));
+ flock1.l_type = F_WRLCK;
+
+ while (!locked) {
+ if (priv->shd.iamshd) {
+ cmd = F_SETLKW;
+ flock1.l_start = 0;
+ flock1.l_len = 0;
+ } else {
+ cmd = F_SETLK;
+ gf_uuid_generate(gfid);
+ flock1.l_start = gfid_to_ino(gfid);
+ if (flock1.l_start < 0)
+ flock1.l_start = -flock1.l_start;
+ flock1.l_len = 1;
+ }
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_NOTIFY, loc, cmd, &flock1, NULL, NULL);
+ if (!ret) {
+ locked = _gf_true;
+ priv->ta_notify_dom_lock_offset = flock1.l_start;
+ } else if (ret == -EAGAIN) {
+ continue;
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to get "
+ "AFR_TA_DOM_NOTIFY lock on %s.",
+ loc->name);
+ goto out;
+ }
+ }
+
+ flock2.l_type = F_WRLCK;
+ flock2.l_start = 0;
+ flock2.l_len = 0;
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_MODIFY, loc, F_SETLKW, &flock2, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to get AFR_TA_DOM_MODIFY lock on %s.", loc->name);
+ flock1.l_type = F_UNLCK;
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_NOTIFY, loc, F_SETLK, &flock1, NULL,
+ NULL);
+ }
+out:
+ return ret;
+}
+
+int
+afr_ta_post_op_unlock(xlator_t *this, loc_t *loc)
+{
+ afr_private_t *priv = this->private;
+ struct gf_flock flock = {
+ 0,
+ };
+ int ret = 0;
+
+ if (!priv->shd.iamshd)
+ GF_ASSERT(afr_ta_is_fop_called_from_synctask(this));
+ flock.l_type = F_UNLCK;
+ flock.l_start = 0;
+ flock.l_len = 0;
+
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_MODIFY, loc, F_SETLK, &flock, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to unlock AFR_TA_DOM_MODIFY lock.");
+ goto out;
+ }
+
+ if (!priv->shd.iamshd)
+ /* Mounts (clients) will not release the AFR_TA_DOM_NOTIFY lock
+ * in post-op as they use it as a notification mechanism. When
+ * shd sends a lock request on TA during heal, the clients will
+ * receive a lock-contention upcall notification upon which they
+ * will release the AFR_TA_DOM_NOTIFY lock after completing the
+ * in flight I/O.*/
+ goto out;
+
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_NOTIFY, loc, F_SETLK, &flock, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to unlock AFR_TA_DOM_NOTIFY lock.");
+ }
+out:
+ return ret;
+}
+
+call_frame_t *
+afr_ta_frame_create(xlator_t *this)
{
- int ret = 0;
+ call_frame_t *frame = NULL;
+ void *lk_owner = NULL;
- local->inode = inode_ref (inode);
- LOCK(&local->inode->lock);
- {
- ret = __afr_inode_ctx_get (this, local->inode,
- &local->inode_ctx);
- }
- UNLOCK (&local->inode->lock);
- if (ret < 0) {
- gf_msg_callingfn (this->name, GF_LOG_ERROR, ENOMEM,
- AFR_MSG_INODE_CTX_GET_FAILED,
- "Error getting inode ctx %s",
- uuid_utoa (local->inode->gfid));
+ frame = create_frame(this, this->ctx->pool);
+ if (!frame)
+ return NULL;
+ lk_owner = (void *)this;
+ afr_set_lk_owner(frame, this, lk_owner);
+ return frame;
+}
+
+gf_boolean_t
+afr_ta_has_quorum(afr_private_t *priv, afr_local_t *local)
+{
+ int data_count = 0;
+
+ data_count = AFR_COUNT(local->child_up, priv->child_count);
+ if (data_count == 2) {
+ return _gf_true;
+ } else if (data_count == 1 && local->ta_child_up) {
+ return _gf_true;
+ }
+
+ return _gf_false;
+}
+
+static gf_boolean_t
+afr_is_add_replica_mount_lookup_on_root(call_frame_t *frame)
+{
+ afr_local_t *local = NULL;
+
+ if (frame->root->pid != GF_CLIENT_PID_ADD_REPLICA_MOUNT)
+ return _gf_false;
+
+ local = frame->local;
+
+ if (local->op != GF_FOP_LOOKUP)
+ /* TODO:If the replica count is being increased on a plain distribute
+ * volume that was never mounted, we need to allow setxattr on '/' with
+ * GF_CLIENT_PID_NO_ROOT_SQUASH to accomodate for DHT layout setting */
+ return _gf_false;
+
+ if (local->inode == NULL)
+ return _gf_false;
+
+ if (!__is_root_gfid(local->inode->gfid))
+ return _gf_false;
+
+ return _gf_true;
+}
+
+gf_boolean_t
+afr_lookup_has_quorum(call_frame_t *frame, const unsigned int up_children_count)
+{
+ if (frame && (up_children_count > 0) &&
+ afr_is_add_replica_mount_lookup_on_root(frame))
+ return _gf_true;
+
+ return _gf_false;
+}
+
+void
+afr_handle_replies_quorum(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = frame->local;
+ afr_private_t *priv = this->private;
+ unsigned char *success_replies = NULL;
+
+ success_replies = alloca0(priv->child_count);
+ afr_fill_success_replies(local, priv, success_replies);
+
+ if (priv->quorum_count && !afr_has_quorum(success_replies, this, NULL)) {
+ local->op_errno = afr_final_errno(local, priv);
+ if (!local->op_errno)
+ local->op_errno = afr_quorum_errno(priv);
+ local->op_ret = -1;
+ }
+}
+
+gf_boolean_t
+afr_ta_dict_contains_pending_xattr(dict_t *dict, afr_private_t *priv, int child)
+{
+ int *pending = NULL;
+ int ret = 0;
+ int i = 0;
+
+ ret = dict_get_ptr(dict, priv->pending_key[child], (void *)&pending);
+ if (ret == 0) {
+ for (i = 0; i < AFR_NUM_CHANGE_LOGS; i++) {
+ /* Not doing a ntoh32(pending) as we just want to check
+ * if it is non-zero or not. */
+ if (pending[i]) {
+ return _gf_true;
+ }
}
- return ret;
+ }
+
+ return _gf_false;
}
diff --git a/xlators/cluster/afr/src/afr-dir-read.c b/xlators/cluster/afr/src/afr-dir-read.c
index 5218d386c79..f8bf8340dab 100644
--- a/xlators/cluster/afr/src/afr-dir-read.c
+++ b/xlators/cluster/afr/src/afr-dir-read.c
@@ -8,348 +8,339 @@
cases as published by the Free Software Foundation.
*/
-
#include <libgen.h>
#include <unistd.h>
-#include <fnmatch.h>
#include <sys/time.h>
#include <stdlib.h>
#include <signal.h>
#include <string.h>
-#include "glusterfs.h"
-#include "dict.h"
-#include "xlator.h"
-#include "hashfn.h"
-#include "logging.h"
-#include "list.h"
-#include "call-stub.h"
-#include "defaults.h"
-#include "common-utils.h"
-#include "compat-errno.h"
-#include "compat.h"
-#include "checksum.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/dict.h>
+#include <glusterfs/list.h>
+#include <glusterfs/common-utils.h>
+#include <glusterfs/compat-errno.h>
+#include <glusterfs/compat.h>
#include "afr.h"
#include "afr-transaction.h"
-
int32_t
-afr_opendir_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- fd_t *fd, dict_t *xdata)
+afr_opendir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int call_count = -1;
- int32_t child_index = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
-
- local = frame->local;
- fd_ctx = local->fd_ctx;
- child_index = (long) cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
- } else {
- local->op_ret = op_ret;
- fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
- if (!local->xdata_rsp && xdata)
- local->xdata_rsp = dict_ref (xdata);
- }
+ afr_local_t *local = NULL;
+ int call_count = -1;
+ int32_t child_index = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+
+ local = frame->local;
+ fd_ctx = local->fd_ctx;
+ child_index = (long)cookie;
+
+ local->replies[child_index].valid = 1;
+ local->replies[child_index].op_ret = op_ret;
+ local->replies[child_index].op_errno = op_errno;
+
+ LOCK(&frame->lock);
+ {
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
+ } else {
+ local->op_ret = op_ret;
+ fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
+ if (!local->xdata_rsp && xdata)
+ local->xdata_rsp = dict_ref(xdata);
}
- UNLOCK (&frame->lock);
+ call_count = --local->call_count;
+ }
+ UNLOCK(&frame->lock);
- call_count = afr_frame_return (frame);
+ if (call_count == 0) {
+ afr_handle_replies_quorum(frame, this);
+ AFR_STACK_UNWIND(opendir, frame, local->op_ret, local->op_errno,
+ local->fd, NULL);
+ }
- if (call_count == 0)
- AFR_STACK_UNWIND (opendir, frame, local->op_ret,
- local->op_errno, local->fd, NULL);
- return 0;
+ return 0;
}
-
int
-afr_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
+afr_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- int i = 0;
- int call_count = -1;
- int32_t op_errno = ENOMEM;
- afr_fd_ctx_t *fd_ctx = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
+ int call_count = -1;
+ int32_t op_errno = ENOMEM;
+ afr_fd_ctx_t *fd_ctx = NULL;
- priv = this->private;
+ priv = this->private;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_OPENDIR;
- if (!afr_is_consistent_io_possible (local, priv, &op_errno))
- goto out;
+ local->op = GF_FOP_OPENDIR;
- fd_ctx = afr_fd_ctx_get (fd, this);
- if (!fd_ctx)
- goto out;
+ if (priv->quorum_count && !afr_has_quorum(local->child_up, this, NULL)) {
+ op_errno = afr_quorum_errno(priv);
+ goto out;
+ }
- loc_copy (&local->loc, loc);
+ if (!afr_is_consistent_io_possible(local, priv, &op_errno))
+ goto out;
- local->fd = fd_ref (fd);
- local->fd_ctx = fd_ctx;
+ fd_ctx = afr_fd_ctx_get(fd, this);
+ if (!fd_ctx)
+ goto out;
- call_count = local->call_count;
+ loc_copy(&local->loc, loc);
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_opendir_cbk,
- (void*) (long) i,
- priv->children[i],
- priv->children[i]->fops->opendir,
- loc, fd, NULL);
+ local->fd = fd_ref(fd);
+ local->fd_ctx = fd_ctx;
- if (!--call_count)
- break;
- }
+ call_count = local->call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE(frame, afr_opendir_cbk, (void *)(long)i,
+ priv->children[i],
+ priv->children[i]->fops->opendir, loc, fd, NULL);
+
+ if (!--call_count)
+ break;
}
+ }
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (opendir, frame, -1, op_errno, fd, NULL);
- return 0;
+ AFR_STACK_UNWIND(opendir, frame, -1, op_errno, fd, NULL);
+ return 0;
}
static int
-afr_validate_read_subvol (inode_t *inode, xlator_t *this, int par_read_subvol)
+afr_validate_read_subvol(inode_t *inode, xlator_t *this, int par_read_subvol)
{
- int gen = 0;
- int entry_read_subvol = 0;
- unsigned char *data_readable = NULL;
- unsigned char *metadata_readable = NULL;
- afr_private_t *priv = NULL;
-
- priv = this->private;
- data_readable = alloca0 (priv->child_count);
- metadata_readable = alloca0 (priv->child_count);
-
- afr_inode_read_subvol_get (inode, this, data_readable,
- metadata_readable, &gen);
-
- if (gen != priv->event_generation ||
- !data_readable[par_read_subvol] ||
- !metadata_readable[par_read_subvol])
- return -1;
-
- /* Once the control reaches the following statement, it means that the
- * parent's read subvol is perfectly readable. So calling
- * either afr_data_subvol_get() or afr_metadata_subvol_get() would
- * yield the same result. Hence, choosing afr_data_subvol_get() below.
- */
-
- if (!priv->consistent_metadata)
- return 0;
-
- /* For an inode fetched through readdirp which is yet to be linked,
- * inode ctx would not be initialised (yet). So this function returns
- * -1 above due to gen being 0, which is why it is OK to pass NULL for
- * read_subvol_args here.
- */
- entry_read_subvol = afr_data_subvol_get (inode, this, NULL, NULL,
- NULL, NULL);
- if (entry_read_subvol != par_read_subvol)
- return -1;
-
+ int gen = 0;
+ int entry_read_subvol = 0;
+ unsigned char *data_readable = NULL;
+ unsigned char *metadata_readable = NULL;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ data_readable = alloca0(priv->child_count);
+ metadata_readable = alloca0(priv->child_count);
+
+ afr_inode_read_subvol_get(inode, this, data_readable, metadata_readable,
+ &gen);
+
+ if (gen != priv->event_generation || !data_readable[par_read_subvol] ||
+ !metadata_readable[par_read_subvol])
+ return -1;
+
+ /* Once the control reaches the following statement, it means that the
+ * parent's read subvol is perfectly readable. So calling
+ * either afr_data_subvol_get() or afr_metadata_subvol_get() would
+ * yield the same result. Hence, choosing afr_data_subvol_get() below.
+ */
+
+ if (!priv->consistent_metadata)
return 0;
+ /* For an inode fetched through readdirp which is yet to be linked,
+ * inode ctx would not be initialised (yet). So this function returns
+ * -1 above due to gen being 0, which is why it is OK to pass NULL for
+ * read_subvol_args here.
+ */
+ entry_read_subvol = afr_data_subvol_get(inode, this, NULL, NULL, NULL,
+ NULL);
+ if (entry_read_subvol != par_read_subvol)
+ return -1;
+
+ return 0;
}
static void
-afr_readdir_transform_entries (gf_dirent_t *subvol_entries, int subvol,
- gf_dirent_t *entries, fd_t *fd)
+afr_readdir_transform_entries(call_frame_t *frame, gf_dirent_t *subvol_entries,
+ int subvol, gf_dirent_t *entries, fd_t *fd)
{
- int ret = -1;
- gf_dirent_t *entry = NULL;
- gf_dirent_t *tmp = NULL;
- xlator_t *this = NULL;
- afr_private_t *priv = NULL;
- gf_boolean_t need_heal = _gf_false;
- gf_boolean_t validate_subvol = _gf_false;
-
- this = THIS;
- priv = this->private;
-
- need_heal = afr_get_need_heal (this);
- validate_subvol = need_heal | priv->consistent_metadata;
-
- list_for_each_entry_safe (entry, tmp, &subvol_entries->list, list) {
- if (__is_root_gfid (fd->inode->gfid) &&
- !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR)) {
- continue;
- }
-
- list_del_init (&entry->list);
- list_add_tail (&entry->list, &entries->list);
-
- if (!validate_subvol)
- continue;
-
- if (entry->inode) {
- ret = afr_validate_read_subvol (entry->inode, this,
- subvol);
- if (ret == -1) {
- inode_unref (entry->inode);
- entry->inode = NULL;
- continue;
- }
- }
+ int ret = -1;
+ gf_dirent_t *entry = NULL;
+ gf_dirent_t *tmp = NULL;
+ xlator_t *this = NULL;
+ afr_private_t *priv = NULL;
+ gf_boolean_t need_heal = _gf_false;
+ gf_boolean_t validate_subvol = _gf_false;
+
+ this = THIS;
+ priv = this->private;
+
+ need_heal = afr_get_need_heal(this);
+ validate_subvol = need_heal | priv->consistent_metadata;
+
+ list_for_each_entry_safe(entry, tmp, &subvol_entries->list, list)
+ {
+ if (afr_is_private_directory(priv, fd->inode->gfid, entry->d_name,
+ frame->root->pid)) {
+ continue;
}
-}
+ list_del_init(&entry->list);
+ list_add_tail(&entry->list, &entries->list);
+
+ if (!validate_subvol)
+ continue;
+
+ if (entry->inode) {
+ ret = afr_validate_read_subvol(entry->inode, this, subvol);
+ if (ret == -1) {
+ inode_unref(entry->inode);
+ entry->inode = NULL;
+ continue;
+ }
+ }
+ }
+}
int32_t
-afr_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *subvol_entries,
- dict_t *xdata)
+afr_readdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *subvol_entries,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- gf_dirent_t entries;
+ afr_local_t *local = NULL;
+ gf_dirent_t entries;
- INIT_LIST_HEAD (&entries.list);
+ INIT_LIST_HEAD(&entries.list);
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0 && !local->cont.readdir.offset) {
- /* failover only if this was first readdir, detected
- by offset == 0 */
- local->op_ret = op_ret;
- local->op_errno = op_errno;
+ if (op_ret < 0 && !local->cont.readdir.offset) {
+ /* failover only if this was first readdir, detected
+ by offset == 0 */
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ afr_read_txn_continue(frame, this, (long)cookie);
+ return 0;
+ }
- if (op_ret >= 0)
- afr_readdir_transform_entries (subvol_entries, (long) cookie,
- &entries, local->fd);
+ if (op_ret >= 0)
+ afr_readdir_transform_entries(frame, subvol_entries, (long)cookie,
+ &entries, local->fd);
- AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, &entries, xdata);
+ AFR_STACK_UNWIND(readdir, frame, op_ret, op_errno, &entries, xdata);
- gf_dirent_free (&entries);
+ gf_dirent_free(&entries);
- return 0;
+ return 0;
}
-
int
-afr_readdir_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_readdir_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
-
- priv = this->private;
- local = frame->local;
- fd_ctx = afr_fd_ctx_get (local->fd, this);
-
- if (subvol == -1) {
- AFR_STACK_UNWIND (readdir, frame, local->op_ret,
- local->op_errno, 0, 0);
- return 0;
- }
-
- fd_ctx->readdir_subvol = subvol;
-
- if (local->op == GF_FOP_READDIR)
- STACK_WIND_COOKIE (frame, afr_readdir_cbk,
- (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->readdir,
- local->fd, local->cont.readdir.size,
- local->cont.readdir.offset,
- local->xdata_req);
- else
- STACK_WIND_COOKIE (frame, afr_readdir_cbk,
- (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->readdirp,
- local->fd, local->cont.readdir.size,
- local->cont.readdir.offset,
- local->xdata_req);
- return 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+
+ priv = this->private;
+ local = frame->local;
+ fd_ctx = afr_fd_ctx_get(local->fd, this);
+ if (!fd_ctx) {
+ local->op_errno = EINVAL;
+ local->op_ret = -1;
+ }
+
+ if (subvol == -1 || !fd_ctx) {
+ AFR_STACK_UNWIND(readdir, frame, local->op_ret, local->op_errno, 0, 0);
+ return 0;
+ }
+
+ fd_ctx->readdir_subvol = subvol;
+
+ if (local->op == GF_FOP_READDIR)
+ STACK_WIND_COOKIE(frame, afr_readdir_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->readdir, local->fd,
+ local->cont.readdir.size, local->cont.readdir.offset,
+ local->xdata_req);
+ else
+ STACK_WIND_COOKIE(frame, afr_readdir_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->readdirp, local->fd,
+ local->cont.readdir.size, local->cont.readdir.offset,
+ local->xdata_req);
+ return 0;
}
-
int
-afr_do_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, int whichop, dict_t *dict)
+afr_do_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, int whichop, dict_t *dict)
{
- afr_local_t *local = NULL;
- int32_t op_errno = 0;
- int subvol = -1;
- afr_fd_ctx_t *fd_ctx = NULL;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- fd_ctx = afr_fd_ctx_get (fd, this);
- if (!fd_ctx) {
- op_errno = EINVAL;
- goto out;
- }
-
- local->op = whichop;
- local->fd = fd_ref (fd);
- local->cont.readdir.size = size;
- local->cont.readdir.offset = offset;
- local->xdata_req = (dict)? dict_ref (dict) : NULL;
-
- subvol = fd_ctx->readdir_subvol;
-
- if (offset == 0 || subvol == -1) {
- /* First readdir has option of failing over and selecting
- an appropriate read subvolume */
- afr_read_txn (frame, this, fd->inode, afr_readdir_wind,
- AFR_DATA_TRANSACTION);
- } else {
- /* But continued readdirs MUST stick to the same subvolume
- without an option to failover */
- afr_readdir_wind (frame, this, subvol);
- }
-
- return 0;
+ afr_local_t *local = NULL;
+ int32_t op_errno = 0;
+ int subvol = -1;
+ afr_fd_ctx_t *fd_ctx = NULL;
+
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
+
+ fd_ctx = afr_fd_ctx_get(fd, this);
+ if (!fd_ctx) {
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ local->op = whichop;
+ local->fd = fd_ref(fd);
+ local->cont.readdir.size = size;
+ local->cont.readdir.offset = offset;
+ local->xdata_req = (dict) ? dict_ref(dict) : NULL;
+
+ subvol = fd_ctx->readdir_subvol;
+
+ if (offset == 0 || subvol == -1) {
+ /* First readdir has option of failing over and selecting
+ an appropriate read subvolume */
+ afr_read_txn(frame, this, fd->inode, afr_readdir_wind,
+ AFR_DATA_TRANSACTION);
+ } else {
+ /* But continued readdirs MUST stick to the same subvolume
+ without an option to failover */
+ afr_readdir_wind(frame, this, subvol);
+ }
+
+ return 0;
out:
- AFR_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(readdir, frame, -1, op_errno, NULL, NULL);
+ return 0;
}
-
int32_t
-afr_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, dict_t *xdata)
+afr_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *xdata)
{
- afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIR, xdata);
+ afr_do_readdir(frame, this, fd, size, offset, GF_FOP_READDIR, xdata);
- return 0;
+ return 0;
}
-
int32_t
-afr_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, dict_t *dict)
+afr_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *dict)
{
- afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIRP, dict);
+ afr_do_readdir(frame, this, fd, size, offset, GF_FOP_READDIRP, dict);
- return 0;
+ return 0;
}
-
int32_t
-afr_releasedir (xlator_t *this, fd_t *fd)
+afr_releasedir(xlator_t *this, fd_t *fd)
{
- afr_cleanup_fd_ctx (this, fd);
+ afr_cleanup_fd_ctx(this, fd);
- return 0;
+ return 0;
}
diff --git a/xlators/cluster/afr/src/afr-dir-read.h b/xlators/cluster/afr/src/afr-dir-read.h
index 09456d15949..773e925ec6c 100644
--- a/xlators/cluster/afr/src/afr-dir-read.h
+++ b/xlators/cluster/afr/src/afr-dir-read.h
@@ -11,26 +11,23 @@
#ifndef __DIR_READ_H__
#define __DIR_READ_H__
-
int32_t
-afr_opendir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, fd_t *fd, dict_t *xdata);
+afr_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata);
int32_t
-afr_releasedir (xlator_t *this, fd_t *fd);
+afr_releasedir(xlator_t *this, fd_t *fd);
int32_t
-afr_readdir (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset, dict_t *xdata);
-
+afr_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *xdata);
int32_t
-afr_readdirp (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset, dict_t *dict);
+afr_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *dict);
int32_t
-afr_checksum (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, dict_t *xdata);
-
+afr_checksum(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ dict_t *xdata);
#endif /* __DIR_READ_H__ */
diff --git a/xlators/cluster/afr/src/afr-dir-write.c b/xlators/cluster/afr/src/afr-dir-write.c
index 0e3ae18b46a..b7cceb79158 100644
--- a/xlators/cluster/afr/src/afr-dir-write.c
+++ b/xlators/cluster/afr/src/afr-dir-write.c
@@ -8,527 +8,493 @@
cases as published by the Free Software Foundation.
*/
-
#include <libgen.h>
#include <unistd.h>
-#include <fnmatch.h>
#include <sys/time.h>
#include <stdlib.h>
#include <signal.h>
-#include "glusterfs.h"
+#include <glusterfs/glusterfs.h>
#include "afr.h"
-#include "dict.h"
-#include "xlator.h"
-#include "hashfn.h"
-#include "logging.h"
-#include "list.h"
-#include "call-stub.h"
-#include "defaults.h"
-#include "common-utils.h"
-#include "compat-errno.h"
-#include "compat.h"
-#include "byte-order.h"
+#include <glusterfs/dict.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/list.h>
+#include <glusterfs/defaults.h>
+#include <glusterfs/common-utils.h>
+#include <glusterfs/compat-errno.h>
+#include <glusterfs/compat.h>
+#include <glusterfs/byte-order.h>
#include "afr.h"
#include "afr-transaction.h"
void
-afr_mark_entry_pending_changelog (call_frame_t *frame, xlator_t *this);
+afr_mark_entry_pending_changelog(call_frame_t *frame, xlator_t *this);
int
-afr_build_parent_loc (loc_t *parent, loc_t *child, int32_t *op_errno)
+afr_build_parent_loc(loc_t *parent, loc_t *child, int32_t *op_errno)
{
- int ret = -1;
- char *child_path = NULL;
-
- if (!child->parent) {
- if (op_errno)
- *op_errno = EINVAL;
- goto out;
- }
-
- child_path = gf_strdup (child->path);
- if (!child_path) {
- if (op_errno)
- *op_errno = ENOMEM;
- goto out;
- }
-
- parent->path = gf_strdup (dirname (child_path));
- if (!parent->path) {
- if (op_errno)
- *op_errno = ENOMEM;
- goto out;
- }
-
- parent->inode = inode_ref (child->parent);
- gf_uuid_copy (parent->gfid, child->pargfid);
-
- ret = 0;
+ int ret = -1;
+ char *child_path = NULL;
+
+ if (!child->parent) {
+ if (op_errno)
+ *op_errno = EINVAL;
+ goto out;
+ }
+
+ child_path = gf_strdup(child->path);
+ if (!child_path) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ parent->path = gf_strdup(dirname(child_path));
+ if (!parent->path) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ parent->inode = inode_ref(child->parent);
+ gf_uuid_copy(parent->gfid, child->pargfid);
+
+ ret = 0;
out:
- GF_FREE (child_path);
+ GF_FREE(child_path);
- return ret;
+ return ret;
}
-
static void
-__afr_dir_write_finalize (call_frame_t *frame, xlator_t *this)
+__afr_dir_write_finalize(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int inode_read_subvol = -1;
- int parent_read_subvol = -1;
- int parent2_read_subvol = -1;
- int i = 0;
- afr_read_subvol_args_t args = {0,};
-
- local = frame->local;
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
- if (local->replies[i].op_ret == -1)
- continue;
- gf_uuid_copy (args.gfid, local->replies[i].poststat.ia_gfid);
- args.ia_type = local->replies[i].poststat.ia_type;
- break;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int inode_read_subvol = -1;
+ int parent_read_subvol = -1;
+ int parent2_read_subvol = -1;
+ int i = 0;
+ afr_read_subvol_args_t args = {
+ 0,
+ };
+
+ local = frame->local;
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+ if (local->replies[i].op_ret == -1)
+ continue;
+ gf_uuid_copy(args.gfid, local->replies[i].poststat.ia_gfid);
+ args.ia_type = local->replies[i].poststat.ia_type;
+ break;
+ }
+
+ if (local->inode) {
+ if (local->op != GF_FOP_RENAME && local->op != GF_FOP_LINK)
+ afr_replies_interpret(frame, this, local->inode, NULL);
+
+ inode_read_subvol = afr_data_subvol_get(local->inode, this, NULL, NULL,
+ NULL, &args);
+ }
+
+ if (local->parent)
+ parent_read_subvol = afr_data_subvol_get(local->parent, this, NULL,
+ local->readable, NULL, NULL);
+
+ if (local->parent2)
+ parent2_read_subvol = afr_data_subvol_get(local->parent2, this, NULL,
+ local->readable2, NULL, NULL);
+
+ local->op_ret = -1;
+ local->op_errno = afr_final_errno(local, priv);
+ afr_pick_error_xdata(local, priv, local->parent, local->readable,
+ local->parent2, local->readable2);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+ if (local->replies[i].op_ret < 0) {
+ if (local->inode)
+ afr_inode_need_refresh_set(local->inode, this);
+ if (local->parent)
+ afr_inode_need_refresh_set(local->parent, this);
+ if (local->parent2)
+ afr_inode_need_refresh_set(local->parent2, this);
+ continue;
}
- if (local->inode) {
- afr_replies_interpret (frame, this, local->inode, NULL);
- inode_read_subvol = afr_data_subvol_get (local->inode, this,
- NULL, NULL, NULL, &args);
- }
-
- if (local->parent)
- parent_read_subvol = afr_data_subvol_get (local->parent, this,
- NULL, local->readable, NULL, NULL);
-
- if (local->parent2)
- parent2_read_subvol = afr_data_subvol_get (local->parent2, this,
- NULL, local->readable2, NULL, NULL);
-
- local->op_ret = -1;
- local->op_errno = afr_final_errno (local, priv);
- afr_pick_error_xdata (local, priv, local->parent, local->readable,
- local->parent2, local->readable2);
-
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
- if (local->replies[i].op_ret < 0) {
- if (local->inode)
- afr_inode_event_gen_reset (local->inode, this);
- if (local->parent)
- afr_inode_event_gen_reset (local->parent,
- this);
- if (local->parent2)
- afr_inode_event_gen_reset (local->parent2,
- this);
- continue;
- }
-
- if (local->op_ret == -1) {
- local->op_ret = local->replies[i].op_ret;
- local->op_errno = local->replies[i].op_errno;
-
- local->cont.dir_fop.buf =
- local->replies[i].poststat;
- local->cont.dir_fop.preparent =
- local->replies[i].preparent;
- local->cont.dir_fop.postparent =
- local->replies[i].postparent;
- local->cont.dir_fop.prenewparent =
- local->replies[i].preparent2;
- local->cont.dir_fop.postnewparent =
- local->replies[i].postparent2;
- if (local->xdata_rsp) {
- dict_unref (local->xdata_rsp);
- local->xdata_rsp = NULL;
- }
-
- if (local->replies[i].xdata)
- local->xdata_rsp =
- dict_ref (local->replies[i].xdata);
- continue;
- }
-
- if (i == inode_read_subvol) {
- local->cont.dir_fop.buf =
- local->replies[i].poststat;
- if (local->replies[i].xdata) {
- if (local->xdata_rsp)
- dict_unref (local->xdata_rsp);
- local->xdata_rsp =
- dict_ref (local->replies[i].xdata);
- }
- }
-
- if (i == parent_read_subvol) {
- local->cont.dir_fop.preparent =
- local->replies[i].preparent;
- local->cont.dir_fop.postparent =
- local->replies[i].postparent;
- }
-
- if (i == parent2_read_subvol) {
- local->cont.dir_fop.prenewparent =
- local->replies[i].preparent2;
- local->cont.dir_fop.postnewparent =
- local->replies[i].postparent2;
- }
- }
+ if (local->op_ret == -1) {
+ local->op_ret = local->replies[i].op_ret;
+ local->op_errno = local->replies[i].op_errno;
+
+ local->cont.dir_fop.buf = local->replies[i].poststat;
+ local->cont.dir_fop.preparent = local->replies[i].preparent;
+ local->cont.dir_fop.postparent = local->replies[i].postparent;
+ local->cont.dir_fop.prenewparent = local->replies[i].preparent2;
+ local->cont.dir_fop.postnewparent = local->replies[i].postparent2;
+ if (local->xdata_rsp) {
+ dict_unref(local->xdata_rsp);
+ local->xdata_rsp = NULL;
+ }
+
+ if (local->replies[i].xdata)
+ local->xdata_rsp = dict_ref(local->replies[i].xdata);
+ continue;
+ }
-}
+ if (i == inode_read_subvol) {
+ local->cont.dir_fop.buf = local->replies[i].poststat;
+ if (local->replies[i].xdata) {
+ if (local->xdata_rsp)
+ dict_unref(local->xdata_rsp);
+ local->xdata_rsp = dict_ref(local->replies[i].xdata);
+ }
+ }
+ if (i == parent_read_subvol) {
+ local->cont.dir_fop.preparent = local->replies[i].preparent;
+ local->cont.dir_fop.postparent = local->replies[i].postparent;
+ }
+
+ if (i == parent2_read_subvol) {
+ local->cont.dir_fop.prenewparent = local->replies[i].preparent2;
+ local->cont.dir_fop.postnewparent = local->replies[i].postparent2;
+ }
+ }
+}
static void
-__afr_dir_write_fill (call_frame_t *frame, xlator_t *this, int child_index,
- int op_ret, int op_errno, struct iatt *poststat,
- struct iatt *preparent, struct iatt *postparent,
- struct iatt *preparent2, struct iatt *postparent2,
- dict_t *xdata)
+__afr_dir_write_fill(call_frame_t *frame, xlator_t *this, int child_index,
+ int op_ret, int op_errno, struct iatt *poststat,
+ struct iatt *preparent, struct iatt *postparent,
+ struct iatt *preparent2, struct iatt *postparent2,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
-
- local = frame->local;
- fd_ctx = local->fd_ctx;
-
- local->replies[child_index].valid = 1;
- local->replies[child_index].op_ret = op_ret;
- local->replies[child_index].op_errno = op_errno;
- if (xdata)
- local->replies[child_index].xdata = dict_ref (xdata);
-
-
- if (op_ret >= 0) {
- if (poststat)
- local->replies[child_index].poststat = *poststat;
- if (preparent)
- local->replies[child_index].preparent = *preparent;
- if (postparent)
- local->replies[child_index].postparent = *postparent;
- if (preparent2)
- local->replies[child_index].preparent2 = *preparent2;
- if (postparent2)
- local->replies[child_index].postparent2 = *postparent2;
- if (fd_ctx)
- fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
- } else {
- if (op_errno != ENOTEMPTY)
- afr_transaction_fop_failed (frame, this, child_index);
- if (fd_ctx)
- fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
- }
-
- return;
+ afr_local_t *local = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+
+ local = frame->local;
+ fd_ctx = local->fd_ctx;
+
+ local->replies[child_index].valid = 1;
+ local->replies[child_index].op_ret = op_ret;
+ local->replies[child_index].op_errno = op_errno;
+ if (xdata)
+ local->replies[child_index].xdata = dict_ref(xdata);
+
+ if (op_ret >= 0) {
+ if (poststat)
+ local->replies[child_index].poststat = *poststat;
+ if (preparent)
+ local->replies[child_index].preparent = *preparent;
+ if (postparent)
+ local->replies[child_index].postparent = *postparent;
+ if (preparent2)
+ local->replies[child_index].preparent2 = *preparent2;
+ if (postparent2)
+ local->replies[child_index].postparent2 = *postparent2;
+ if (fd_ctx)
+ fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
+ } else {
+ if (op_errno != ENOTEMPTY)
+ afr_transaction_fop_failed(frame, this, child_index);
+ if (fd_ctx)
+ fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
+ }
+
+ return;
}
-
static int
-__afr_dir_write_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent,
- struct iatt *preparent2, struct iatt *postparent2,
- dict_t *xdata)
+__afr_dir_write_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *buf,
+ struct iatt *preparent, struct iatt *postparent,
+ struct iatt *preparent2, struct iatt *postparent2,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int child_index = (long) cookie;
- int call_count = -1;
- afr_private_t *priv = NULL;
-
- priv = this->private;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- __afr_dir_write_fill (frame, this, child_index, op_ret,
- op_errno, buf, preparent, postparent,
- preparent2, postparent2, xdata);
- }
- UNLOCK (&frame->lock);
- call_count = afr_frame_return (frame);
-
- if (call_count == 0) {
- __afr_dir_write_finalize (frame, this);
-
- if (afr_txn_nothing_failed (frame, this)) {
- /*if it did pre-op, it will do post-op changing ctime*/
- if (priv->consistent_metadata &&
- afr_needs_changelog_update (local))
- afr_zero_fill_stat (local);
- local->transaction.unwind (frame, this);
- }
-
- afr_mark_entry_pending_changelog (frame, this);
-
- afr_transaction_resume (frame, this);
+ afr_local_t *local = NULL;
+ int child_index = (long)cookie;
+ int call_count = -1;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ LOCK(&frame->lock);
+ {
+ __afr_dir_write_fill(frame, this, child_index, op_ret, op_errno, buf,
+ preparent, postparent, preparent2, postparent2,
+ xdata);
+ call_count = --local->call_count;
+ }
+ UNLOCK(&frame->lock);
+
+ if (call_count == 0) {
+ __afr_dir_write_finalize(frame, this);
+
+ if (afr_txn_nothing_failed(frame, this)) {
+ /*if it did pre-op, it will do post-op changing ctime*/
+ if (priv->consistent_metadata && afr_needs_changelog_update(local))
+ afr_zero_fill_stat(local);
+ local->transaction.unwind(frame, this);
}
- return 0;
-}
+ afr_mark_entry_pending_changelog(frame, this);
+
+ afr_transaction_resume(frame, this);
+ }
+ return 0;
+}
int
-afr_mark_new_entry_changelog_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- dict_t *xattr, dict_t *xdata)
+afr_mark_new_entry_changelog_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int op_ret, int op_errno,
+ dict_t *xattr, dict_t *xdata)
{
- int call_count = 0;
+ int call_count = 0;
- call_count = afr_frame_return (frame);
+ call_count = afr_frame_return(frame);
- if (call_count == 0)
- AFR_STACK_DESTROY (frame);
+ if (call_count == 0)
+ AFR_STACK_DESTROY(frame);
- return 0;
+ return 0;
}
-
void
-afr_mark_new_entry_changelog (call_frame_t *frame, xlator_t *this)
+afr_mark_new_entry_changelog(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *new_frame = NULL;
- afr_local_t *local = NULL;
- afr_local_t *new_local = NULL;
- afr_private_t *priv = NULL;
- dict_t *xattr = NULL;
- int32_t **changelog = NULL;
- int i = 0;
- int op_errno = ENOMEM;
- unsigned char *pending = NULL;
- int call_count = 0;
-
- local = frame->local;
- priv = this->private;
-
- new_frame = copy_frame (frame);
- if (!new_frame)
- goto out;
-
- new_local = AFR_FRAME_INIT (new_frame, op_errno);
- if (!new_local)
- goto out;
-
- xattr = dict_new ();
- if (!xattr)
- goto out;
-
- pending = alloca0 (priv->child_count);
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.pre_op[i] &&
- !local->transaction.failed_subvols[i]) {
- call_count ++;
- continue;
- }
- pending[i] = 1;
- }
-
- changelog = afr_mark_pending_changelog (priv, pending, xattr,
- local->cont.dir_fop.buf.ia_type);
- if (!changelog)
- goto out;
-
- new_local->pending = changelog;
- gf_uuid_copy (new_local->loc.gfid, local->cont.dir_fop.buf.ia_gfid);
- new_local->loc.inode = inode_ref (local->inode);
-
- new_local->call_count = call_count;
-
- for (i = 0; i < priv->child_count; i++) {
- if (pending[i])
- continue;
-
- STACK_WIND_COOKIE (new_frame, afr_mark_new_entry_changelog_cbk,
- (void *) (long) i, priv->children[i],
- priv->children[i]->fops->xattrop,
- &new_local->loc, GF_XATTROP_ADD_ARRAY,
- xattr, NULL);
- if (!--call_count)
- break;
+ call_frame_t *new_frame = NULL;
+ afr_local_t *local = NULL;
+ afr_local_t *new_local = NULL;
+ afr_private_t *priv = NULL;
+ dict_t *xattr = NULL;
+ int32_t **changelog = NULL;
+ int i = 0;
+ int op_errno = ENOMEM;
+ unsigned char *pending = NULL;
+ int call_count = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ new_frame = copy_frame(frame);
+ if (!new_frame)
+ goto out;
+
+ new_local = AFR_FRAME_INIT(new_frame, op_errno);
+ if (!new_local)
+ goto out;
+
+ xattr = dict_new();
+ if (!xattr)
+ goto out;
+
+ pending = alloca0(priv->child_count);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i] &&
+ !local->transaction.failed_subvols[i]) {
+ call_count++;
+ continue;
}
+ pending[i] = 1;
+ }
+
+ changelog = afr_mark_pending_changelog(priv, pending, xattr,
+ local->cont.dir_fop.buf.ia_type);
+ if (!changelog)
+ goto out;
+
+ new_local->pending = changelog;
+ gf_uuid_copy(new_local->loc.gfid, local->cont.dir_fop.buf.ia_gfid);
+ new_local->loc.inode = inode_ref(local->inode);
- new_frame = NULL;
+ new_local->call_count = call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (pending[i])
+ continue;
+
+ STACK_WIND_COOKIE(new_frame, afr_mark_new_entry_changelog_cbk,
+ (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->xattrop, &new_local->loc,
+ GF_XATTROP_ADD_ARRAY, xattr, NULL);
+ if (!--call_count)
+ break;
+ }
+
+ new_frame = NULL;
out:
- if (new_frame)
- AFR_STACK_DESTROY (new_frame);
- if (xattr)
- dict_unref (xattr);
- return;
+ if (new_frame)
+ AFR_STACK_DESTROY(new_frame);
+ if (xattr)
+ dict_unref(xattr);
+ return;
}
-
void
-afr_mark_entry_pending_changelog (call_frame_t *frame, xlator_t *this)
+afr_mark_entry_pending_changelog(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int pre_op_count = 0;
- int failed_count = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int pre_op_count = 0;
+ int failed_count = 0;
+ unsigned char *success_replies = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- if (local->op_ret < 0)
- return;
+ if (local->op_ret < 0)
+ return;
- if (local->op != GF_FOP_CREATE && local->op != GF_FOP_MKNOD &&
- local->op != GF_FOP_MKDIR)
- return;
+ if (local->op != GF_FOP_CREATE && local->op != GF_FOP_MKNOD &&
+ local->op != GF_FOP_MKDIR)
+ return;
- pre_op_count = AFR_COUNT (local->transaction.pre_op, priv->child_count);
- failed_count = AFR_COUNT (local->transaction.failed_subvols,
- priv->child_count);
+ pre_op_count = AFR_COUNT(local->transaction.pre_op, priv->child_count);
+ failed_count = AFR_COUNT(local->transaction.failed_subvols,
+ priv->child_count);
- if (pre_op_count == priv->child_count && !failed_count)
- return;
+ /* FOP succeeded on all bricks. */
+ if (pre_op_count == priv->child_count && !failed_count)
+ return;
- afr_mark_new_entry_changelog (frame, this);
+ /* FOP did not suceed on quorum no. of bricks. */
+ success_replies = alloca0(priv->child_count);
+ afr_fill_success_replies(local, priv, success_replies);
+ if (!afr_has_quorum(success_replies, this, NULL))
+ return;
+ if (priv->thin_arbiter_count) {
+ /*Mark new entry using ta file*/
+ local->is_new_entry = _gf_true;
return;
-}
+ }
+ afr_mark_new_entry_changelog(frame, this);
+
+ return;
+}
/* {{{ create */
int
-afr_create_unwind (call_frame_t *frame, xlator_t *this)
+afr_create_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- local = frame->local;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
+ local = frame->local;
- if (!main_frame)
- return 0;
+ main_frame = afr_transaction_detach_fop_frame(frame);
- AFR_STACK_UNWIND (create, main_frame, local->op_ret, local->op_errno,
- local->cont.create.fd, local->inode,
- &local->cont.dir_fop.buf,
- &local->cont.dir_fop.preparent,
- &local->cont.dir_fop.postparent, local->xdata_rsp);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(create, main_frame, local->op_ret, local->op_errno,
+ local->cont.create.fd, local->inode,
+ &local->cont.dir_fop.buf, &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent, local->xdata_rsp);
+ return 0;
+}
int
-afr_create_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- fd_t *fd, inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+afr_create_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
- preparent, postparent, NULL, NULL, xdata);
+ return __afr_dir_write_cbk(frame, cookie, this, op_ret, op_errno, buf,
+ preparent, postparent, NULL, NULL, xdata);
}
-
int
-afr_create_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_create_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- STACK_WIND_COOKIE (frame, afr_create_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->create,
- &local->loc, local->cont.create.flags,
- local->cont.create.mode, local->umask,
- local->cont.create.fd, local->xdata_req);
- return 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ STACK_WIND_COOKIE(frame, afr_create_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->create, &local->loc,
+ local->cont.create.flags, local->cont.create.mode,
+ local->umask, local->cont.create.fd, local->xdata_req);
+ return 0;
}
-
int
-afr_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+afr_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
-
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
-
- loc_copy (&local->loc, loc);
-
- local->fd_ctx = afr_fd_ctx_get (fd, this);
- if (!local->fd_ctx)
- goto out;
-
- local->inode = inode_ref (loc->inode);
- local->parent = inode_ref (loc->parent);
-
- local->op = GF_FOP_CREATE;
- local->cont.create.flags = flags;
- local->fd_ctx->flags = flags;
- local->cont.create.mode = mode;
- local->cont.create.fd = fd_ref (fd);
- local->umask = umask;
-
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
-
- if (!local->xdata_req)
- goto out;
-
- local->transaction.wind = afr_create_wind;
- local->transaction.unwind = afr_create_unwind;
-
- ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
- &op_errno);
- if (ret)
- goto out;
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
- int_lock = &local->internal_lock;
-
- int_lock->lockee_count = 0;
- ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
- &local->transaction.parent_loc,
- local->transaction.basename,
- priv->child_count);
- if (ret)
- goto out;
-
- int_lock->lockee_count++;
- ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- return 0;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
+
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, loc);
+
+ local->fd_ctx = afr_fd_ctx_get(fd, this);
+ if (!local->fd_ctx)
+ goto out;
+
+ local->inode = inode_ref(loc->inode);
+ local->parent = inode_ref(loc->parent);
+
+ local->op = GF_FOP_CREATE;
+ local->cont.create.flags = flags;
+ local->fd_ctx->flags = flags;
+ local->cont.create.mode = mode;
+ local->cont.create.fd = fd_ref(fd);
+ local->umask = umask;
+
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
+
+ if (!local->xdata_req)
+ goto out;
+
+ local->transaction.wind = afr_create_wind;
+ local->transaction.unwind = afr_create_unwind;
+
+ ret = afr_build_parent_loc(&local->transaction.parent_loc, loc, &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME(loc->path);
+ ret = afr_transaction(transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL, NULL,
- NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(create, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
/* }}} */
@@ -536,516 +502,436 @@ out:
/* {{{ mknod */
int
-afr_mknod_unwind (call_frame_t *frame, xlator_t *this)
+afr_mknod_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
-
- AFR_STACK_UNWIND (mknod, main_frame, local->op_ret, local->op_errno,
- local->inode, &local->cont.dir_fop.buf,
- &local->cont.dir_fop.preparent,
- &local->cont.dir_fop.postparent, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(mknod, main_frame, local->op_ret, local->op_errno,
+ local->inode, &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent, local->xdata_rsp);
+ return 0;
+}
int
-afr_mknod_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+afr_mknod_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
- preparent, postparent, NULL, NULL, xdata);
+ return __afr_dir_write_cbk(frame, cookie, this, op_ret, op_errno, buf,
+ preparent, postparent, NULL, NULL, xdata);
}
-
int
-afr_mknod_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_mknod_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- STACK_WIND_COOKIE (frame, afr_mknod_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->mknod,
- &local->loc, local->cont.mknod.mode,
- local->cont.mknod.dev, local->umask,
- local->xdata_req);
- return 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ STACK_WIND_COOKIE(frame, afr_mknod_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->mknod, &local->loc,
+ local->cont.mknod.mode, local->cont.mknod.dev,
+ local->umask, local->xdata_req);
+ return 0;
}
int
-afr_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- dev_t dev, mode_t umask, dict_t *xdata)
+afr_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t dev, mode_t umask, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
-
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
-
- loc_copy (&local->loc, loc);
- local->inode = inode_ref (loc->inode);
- local->parent = inode_ref (loc->parent);
-
- local->op = GF_FOP_MKNOD;
- local->cont.mknod.mode = mode;
- local->cont.mknod.dev = dev;
- local->umask = umask;
-
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
-
- if (!local->xdata_req)
- goto out;
-
- local->transaction.wind = afr_mknod_wind;
- local->transaction.unwind = afr_mknod_unwind;
-
- ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
- &op_errno);
- if (ret)
- goto out;
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
- int_lock = &local->internal_lock;
-
- int_lock->lockee_count = 0;
- ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
- &local->transaction.parent_loc,
- local->transaction.basename,
- priv->child_count);
- if (ret)
- goto out;
-
- int_lock->lockee_count++;
- ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- return 0;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
+
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, loc);
+ local->inode = inode_ref(loc->inode);
+ local->parent = inode_ref(loc->parent);
+
+ local->op = GF_FOP_MKNOD;
+ local->cont.mknod.mode = mode;
+ local->cont.mknod.dev = dev;
+ local->umask = umask;
+
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
+
+ if (!local->xdata_req)
+ goto out;
+
+ local->transaction.wind = afr_mknod_wind;
+ local->transaction.unwind = afr_mknod_unwind;
+
+ ret = afr_build_parent_loc(&local->transaction.parent_loc, loc, &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME(loc->path);
+ ret = afr_transaction(transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL,
- NULL);
- return 0;
+ AFR_STACK_UNWIND(mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+ return 0;
}
/* }}} */
/* {{{ mkdir */
-
int
-afr_mkdir_unwind (call_frame_t *frame, xlator_t *this)
+afr_mkdir_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- local = frame->local;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (mkdir, main_frame, local->op_ret, local->op_errno,
- local->inode, &local->cont.dir_fop.buf,
- &local->cont.dir_fop.preparent,
- &local->cont.dir_fop.postparent, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(mkdir, main_frame, local->op_ret, local->op_errno,
+ local->inode, &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent, local->xdata_rsp);
+ return 0;
+}
int
-afr_mkdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+afr_mkdir_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
- preparent, postparent, NULL, NULL, xdata);
+ return __afr_dir_write_cbk(frame, cookie, this, op_ret, op_errno, buf,
+ preparent, postparent, NULL, NULL, xdata);
}
-
int
-afr_mkdir_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_mkdir_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_mkdir_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->mkdir, &local->loc,
- local->cont.mkdir.mode, local->umask,
- local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_mkdir_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->mkdir, &local->loc,
+ local->cont.mkdir.mode, local->umask, local->xdata_req);
+ return 0;
}
-
int
-afr_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- mode_t umask, dict_t *xdata)
+afr_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
-
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
-
- loc_copy (&local->loc, loc);
- local->inode = inode_ref (loc->inode);
- local->parent = inode_ref (loc->parent);
-
- local->cont.mkdir.mode = mode;
- local->umask = umask;
-
- if (!xdata || !dict_get (xdata, "gfid-req")) {
- op_errno = EPERM;
- gf_msg_callingfn (this->name, GF_LOG_WARNING, op_errno,
- AFR_MSG_GFID_NULL, "mkdir: %s is received "
- "without gfid-req %p", loc->path, xdata);
- goto out;
- }
-
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- if (!local->xdata_req) {
- op_errno = ENOMEM;
- goto out;
- }
-
- local->op = GF_FOP_MKDIR;
- local->transaction.wind = afr_mkdir_wind;
- local->transaction.unwind = afr_mkdir_unwind;
-
- ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
- &op_errno);
- if (ret)
- goto out;
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
- int_lock = &local->internal_lock;
-
- int_lock->lockee_count = 0;
- ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
- &local->transaction.parent_loc,
- local->transaction.basename,
- priv->child_count);
- if (ret)
- goto out;
-
- int_lock->lockee_count++;
- ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- return 0;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
+
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, loc);
+ local->inode = inode_ref(loc->inode);
+ local->parent = inode_ref(loc->parent);
+
+ local->cont.mkdir.mode = mode;
+ local->umask = umask;
+
+ if (!xdata || !dict_get_sizen(xdata, "gfid-req")) {
+ op_errno = EPERM;
+ gf_msg_callingfn(this->name, GF_LOG_WARNING, op_errno,
+ AFR_MSG_GFID_NULL,
+ "mkdir: %s is received "
+ "without gfid-req %p",
+ loc->path, xdata);
+ goto out;
+ }
+
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ if (!local->xdata_req) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ local->op = GF_FOP_MKDIR;
+ local->transaction.wind = afr_mkdir_wind;
+ local->transaction.unwind = afr_mkdir_unwind;
+
+ ret = afr_build_parent_loc(&local->transaction.parent_loc, loc, &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME(loc->path);
+ ret = afr_transaction(transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL,
- NULL);
- return 0;
+ AFR_STACK_UNWIND(mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+ return 0;
}
/* }}} */
/* {{{ link */
-
int
-afr_link_unwind (call_frame_t *frame, xlator_t *this)
+afr_link_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- local = frame->local;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (link, main_frame, local->op_ret, local->op_errno,
- local->inode, &local->cont.dir_fop.buf,
- &local->cont.dir_fop.preparent,
- &local->cont.dir_fop.postparent, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(link, main_frame, local->op_ret, local->op_errno,
+ local->inode, &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent, local->xdata_rsp);
+ return 0;
+}
int
-afr_link_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+afr_link_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
- preparent, postparent, NULL, NULL, xdata);
+ return __afr_dir_write_cbk(frame, cookie, this, op_ret, op_errno, buf,
+ preparent, postparent, NULL, NULL, xdata);
}
-
int
-afr_link_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_link_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_link_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->link,
- &local->loc, &local->newloc, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_link_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->link, &local->loc,
+ &local->newloc, local->xdata_req);
+ return 0;
}
-
int
-afr_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
+afr_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
-
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
-
- loc_copy (&local->loc, oldloc);
- loc_copy (&local->newloc, newloc);
-
- local->inode = inode_ref (oldloc->inode);
- local->parent = inode_ref (newloc->parent);
-
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
-
- if (!local->xdata_req)
- goto out;
-
- local->op = GF_FOP_LINK;
-
- local->transaction.wind = afr_link_wind;
- local->transaction.unwind = afr_link_unwind;
-
- ret = afr_build_parent_loc (&local->transaction.parent_loc, newloc,
- &op_errno);
- if (ret)
- goto out;
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (newloc->path);
- int_lock = &local->internal_lock;
-
- int_lock->lockee_count = 0;
- ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
- &local->transaction.parent_loc,
- local->transaction.basename,
- priv->child_count);
- if (ret)
- goto out;
-
- int_lock->lockee_count++;
- ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
+
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, oldloc);
+ loc_copy(&local->newloc, newloc);
- return 0;
+ local->inode = inode_ref(oldloc->inode);
+ local->parent = inode_ref(newloc->parent);
+
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
+
+ if (!local->xdata_req)
+ goto out;
+
+ local->op = GF_FOP_LINK;
+
+ local->transaction.wind = afr_link_wind;
+ local->transaction.unwind = afr_link_unwind;
+
+ ret = afr_build_parent_loc(&local->transaction.parent_loc, newloc,
+ &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME(newloc->path);
+ ret = afr_transaction(transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL,
- NULL);
- return 0;
+ AFR_STACK_UNWIND(link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+ return 0;
}
/* }}} */
/* {{{ symlink */
-
int
-afr_symlink_unwind (call_frame_t *frame, xlator_t *this)
+afr_symlink_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- local = frame->local;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (symlink, main_frame, local->op_ret, local->op_errno,
- local->inode, &local->cont.dir_fop.buf,
- &local->cont.dir_fop.preparent,
- &local->cont.dir_fop.postparent, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(symlink, main_frame, local->op_ret, local->op_errno,
+ local->inode, &local->cont.dir_fop.buf,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent, local->xdata_rsp);
+ return 0;
+}
int
-afr_symlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+afr_symlink_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
- preparent, postparent, NULL, NULL, xdata);
+ return __afr_dir_write_cbk(frame, cookie, this, op_ret, op_errno, buf,
+ preparent, postparent, NULL, NULL, xdata);
}
-
int
-afr_symlink_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_symlink_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- STACK_WIND_COOKIE (frame, afr_symlink_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->symlink,
- local->cont.symlink.linkpath, &local->loc,
- local->umask, local->xdata_req);
- return 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ STACK_WIND_COOKIE(frame, afr_symlink_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->symlink,
+ local->cont.symlink.linkpath, &local->loc, local->umask,
+ local->xdata_req);
+ return 0;
}
-
int
-afr_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
- loc_t *loc, mode_t umask, dict_t *xdata)
+afr_symlink(call_frame_t *frame, xlator_t *this, const char *linkpath,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
-
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
-
- loc_copy (&local->loc, loc);
- local->inode = inode_ref (loc->inode);
- local->parent = inode_ref (loc->parent);
-
- local->cont.symlink.linkpath = gf_strdup (linkpath);
- local->umask = umask;
-
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
-
- if (!local->xdata_req)
- goto out;
-
- local->op = GF_FOP_SYMLINK;
- local->transaction.wind = afr_symlink_wind;
- local->transaction.unwind = afr_symlink_unwind;
-
- ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
- &op_errno);
- if (ret)
- goto out;
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
- int_lock = &local->internal_lock;
-
- int_lock->lockee_count = 0;
- ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
- &local->transaction.parent_loc,
- local->transaction.basename,
- priv->child_count);
- if (ret)
- goto out;
-
- int_lock->lockee_count++;
- ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- return 0;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
+
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, loc);
+ local->inode = inode_ref(loc->inode);
+ local->parent = inode_ref(loc->parent);
+
+ local->cont.symlink.linkpath = gf_strdup(linkpath);
+ local->umask = umask;
+
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
+
+ if (!local->xdata_req)
+ goto out;
+
+ local->op = GF_FOP_SYMLINK;
+ local->transaction.wind = afr_symlink_wind;
+ local->transaction.unwind = afr_symlink_unwind;
+
+ ret = afr_build_parent_loc(&local->transaction.parent_loc, loc, &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME(loc->path);
+ ret = afr_transaction(transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (symlink, frame, -1, op_errno, NULL, NULL, NULL,
- NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(symlink, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
/* }}} */
@@ -1053,159 +939,118 @@ out:
/* {{{ rename */
int
-afr_rename_unwind (call_frame_t *frame, xlator_t *this)
+afr_rename_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- local = frame->local;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (rename, main_frame, local->op_ret, local->op_errno,
- &local->cont.dir_fop.buf,
- &local->cont.dir_fop.preparent,
- &local->cont.dir_fop.postparent,
- &local->cont.dir_fop.prenewparent,
- &local->cont.dir_fop.postnewparent, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(rename, main_frame, local->op_ret, local->op_errno,
+ &local->cont.dir_fop.buf, &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent,
+ &local->cont.dir_fop.prenewparent,
+ &local->cont.dir_fop.postnewparent, local->xdata_rsp);
+ return 0;
+}
int
-afr_rename_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent,
- dict_t *xdata)
+afr_rename_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ struct iatt *preoldparent, struct iatt *postoldparent,
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
{
- return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
- preoldparent, postoldparent, prenewparent,
- postnewparent, xdata);
+ return __afr_dir_write_cbk(frame, cookie, this, op_ret, op_errno, buf,
+ preoldparent, postoldparent, prenewparent,
+ postnewparent, xdata);
}
-
int
-afr_rename_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_rename_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_rename_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->rename,
- &local->loc, &local->newloc, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_rename_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->rename, &local->loc,
+ &local->newloc, local->xdata_req);
+ return 0;
}
-
int
-afr_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
+afr_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
- int nlockee = 0;
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame) {
- op_errno = ENOMEM;
- goto out;
- }
-
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
-
- loc_copy (&local->loc, oldloc);
- loc_copy (&local->newloc, newloc);
-
- local->inode = inode_ref (oldloc->inode);
- local->parent = inode_ref (oldloc->parent);
- local->parent2 = inode_ref (newloc->parent);
-
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
-
- if (!local->xdata_req)
- goto out;
-
- local->op = GF_FOP_RENAME;
- local->transaction.wind = afr_rename_wind;
- local->transaction.unwind = afr_rename_unwind;
-
- ret = afr_build_parent_loc (&local->transaction.parent_loc, oldloc,
- &op_errno);
- if (ret)
- goto out;
- ret = afr_build_parent_loc (&local->transaction.new_parent_loc, newloc,
- &op_errno);
- if (ret)
- goto out;
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (oldloc->path);
- local->transaction.new_basename = AFR_BASENAME (newloc->path);
- int_lock = &local->internal_lock;
-
- int_lock->lockee_count = nlockee = 0;
- ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
- &local->transaction.new_parent_loc,
- local->transaction.new_basename,
- priv->child_count);
- if (ret)
- goto out;
-
- nlockee++;
- ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
- &local->transaction.parent_loc,
- local->transaction.basename,
- priv->child_count);
- if (ret)
- goto out;
-
- nlockee++;
- if (local->newloc.inode && IA_ISDIR (local->newloc.inode->ia_type)) {
- ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
- &local->newloc,
- NULL,
- priv->child_count);
- if (ret)
- goto out;
-
- nlockee++;
- }
- qsort (int_lock->lockee, nlockee, sizeof (*int_lock->lockee),
- afr_entry_lockee_cmp);
- int_lock->lockee_count = nlockee;
-
- ret = afr_transaction (transaction_frame, this, AFR_ENTRY_RENAME_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- return 0;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, oldloc);
+ loc_copy(&local->newloc, newloc);
+
+ local->inode = inode_ref(oldloc->inode);
+ local->parent = inode_ref(oldloc->parent);
+ local->parent2 = inode_ref(newloc->parent);
+
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
+
+ if (!local->xdata_req)
+ goto out;
+
+ local->op = GF_FOP_RENAME;
+ local->transaction.wind = afr_rename_wind;
+ local->transaction.unwind = afr_rename_unwind;
+
+ ret = afr_build_parent_loc(&local->transaction.parent_loc, oldloc,
+ &op_errno);
+ if (ret)
+ goto out;
+ ret = afr_build_parent_loc(&local->transaction.new_parent_loc, newloc,
+ &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME(oldloc->path);
+ local->transaction.new_basename = AFR_BASENAME(newloc->path);
+ ret = afr_transaction(transaction_frame, this,
+ AFR_ENTRY_RENAME_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
- NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(rename, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
/* }}} */
@@ -1213,259 +1058,205 @@ out:
/* {{{ unlink */
int
-afr_unlink_unwind (call_frame_t *frame, xlator_t *this)
+afr_unlink_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
-
- local = frame->local;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (unlink, main_frame, local->op_ret, local->op_errno,
- &local->cont.dir_fop.preparent,
- &local->cont.dir_fop.postparent, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(unlink, main_frame, local->op_ret, local->op_errno,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent, local->xdata_rsp);
+ return 0;
+}
int
-afr_unlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+afr_unlink_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, NULL,
- preparent, postparent, NULL, NULL, xdata);
+ return __afr_dir_write_cbk(frame, cookie, this, op_ret, op_errno, NULL,
+ preparent, postparent, NULL, NULL, xdata);
}
-
int
-afr_unlink_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_unlink_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_unlink_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->unlink,
- &local->loc, local->xflag, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_unlink_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->unlink, &local->loc,
+ local->xflag, local->xdata_req);
+ return 0;
}
-
int
-afr_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
- dict_t *xdata)
+afr_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
-
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
-
- loc_copy (&local->loc, loc);
- local->xflag = xflag;
-
- local->inode = inode_ref (loc->inode);
- local->parent = inode_ref (loc->parent);
-
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
-
- if (!local->xdata_req)
- goto out;
-
- local->op = GF_FOP_UNLINK;
- local->transaction.wind = afr_unlink_wind;
- local->transaction.unwind = afr_unlink_unwind;
-
- ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
- &op_errno);
- if (ret)
- goto out;
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
- int_lock = &local->internal_lock;
-
- int_lock->lockee_count = 0;
- ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
- &local->transaction.parent_loc,
- local->transaction.basename,
- priv->child_count);
- if (ret)
- goto out;
-
- int_lock->lockee_count++;
- ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- return 0;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
+
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, loc);
+ local->xflag = xflag;
+
+ local->inode = inode_ref(loc->inode);
+ local->parent = inode_ref(loc->parent);
+
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
+
+ if (!local->xdata_req)
+ goto out;
+
+ local->op = GF_FOP_UNLINK;
+ local->transaction.wind = afr_unlink_wind;
+ local->transaction.unwind = afr_unlink_unwind;
+
+ ret = afr_build_parent_loc(&local->transaction.parent_loc, loc, &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME(loc->path);
+ ret = afr_transaction(transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(unlink, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
/* }}} */
/* {{{ rmdir */
-
-
int
-afr_rmdir_unwind (call_frame_t *frame, xlator_t *this)
+afr_rmdir_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *main_frame = NULL;
- afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
-
- AFR_STACK_UNWIND (rmdir, main_frame, local->op_ret, local->op_errno,
- &local->cont.dir_fop.preparent,
- &local->cont.dir_fop.postparent, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(rmdir, main_frame, local->op_ret, local->op_errno,
+ &local->cont.dir_fop.preparent,
+ &local->cont.dir_fop.postparent, local->xdata_rsp);
+ return 0;
+}
int
-afr_rmdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+afr_rmdir_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, NULL,
- preparent, postparent, NULL, NULL, xdata);
+ return __afr_dir_write_cbk(frame, cookie, this, op_ret, op_errno, NULL,
+ preparent, postparent, NULL, NULL, xdata);
}
-
int
-afr_rmdir_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_rmdir_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_rmdir_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->rmdir,
- &local->loc, local->cont.rmdir.flags, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_rmdir_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->rmdir, &local->loc,
+ local->cont.rmdir.flags, local->xdata_req);
+ return 0;
}
-
int
-afr_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
- dict_t *xdata)
+afr_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
- int nlockee = 0;
-
- priv = this->private;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
-
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
-
-
- loc_copy (&local->loc, loc);
- local->inode = inode_ref (loc->inode);
- local->parent = inode_ref (loc->parent);
-
- local->cont.rmdir.flags = flags;
-
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
-
- if (!local->xdata_req)
- goto out;
-
- local->op = GF_FOP_RMDIR;
- local->transaction.wind = afr_rmdir_wind;
- local->transaction.unwind = afr_rmdir_unwind;
-
- ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
- &op_errno);
- if (ret)
- goto out;
-
- local->transaction.main_frame = frame;
- local->transaction.basename = AFR_BASENAME (loc->path);
- int_lock = &local->internal_lock;
-
- int_lock->lockee_count = nlockee = 0;
- ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
- &local->transaction.parent_loc,
- local->transaction.basename,
- priv->child_count);
- if (ret)
- goto out;
-
- nlockee++;
- ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
- &local->loc,
- NULL,
- priv->child_count);
- if (ret)
- goto out;
-
- nlockee++;
- qsort (int_lock->lockee, nlockee, sizeof (*int_lock->lockee),
- afr_entry_lockee_cmp);
- int_lock->lockee_count = nlockee;
-
- ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
-
- return 0;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
+
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
+
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, loc);
+ local->inode = inode_ref(loc->inode);
+ local->parent = inode_ref(loc->parent);
+
+ local->cont.rmdir.flags = flags;
+
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
+
+ if (!local->xdata_req)
+ goto out;
+
+ local->op = GF_FOP_RMDIR;
+ local->transaction.wind = afr_rmdir_wind;
+ local->transaction.unwind = afr_rmdir_unwind;
+
+ ret = afr_build_parent_loc(&local->transaction.parent_loc, loc, &op_errno);
+ if (ret)
+ goto out;
+
+ local->transaction.main_frame = frame;
+ local->transaction.basename = AFR_BASENAME(loc->path);
+ ret = afr_transaction(transaction_frame, this, AFR_ENTRY_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (rmdir, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(rmdir, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
/* }}} */
diff --git a/xlators/cluster/afr/src/afr-dir-write.h b/xlators/cluster/afr/src/afr-dir-write.h
index 02f0a3682d9..1d88c3b9b26 100644
--- a/xlators/cluster/afr/src/afr-dir-write.h
+++ b/xlators/cluster/afr/src/afr-dir-write.h
@@ -12,36 +12,35 @@
#define __DIR_WRITE_H__
int32_t
-afr_create (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, mode_t mode,
- mode_t umask, fd_t *fd, dict_t *xdata);
+afr_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata);
int32_t
-afr_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t dev, mode_t umask, dict_t *xdata);
+afr_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t dev, mode_t umask, dict_t *xdata);
int32_t
-afr_mkdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata);
+afr_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata);
int32_t
-afr_unlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int xflag, dict_t *xdata);
+afr_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata);
int32_t
-afr_rmdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int flags, dict_t *xdata);
+afr_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata);
int32_t
-afr_link (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata);
+afr_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata);
int32_t
-afr_rename (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata);
+afr_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata);
int
-afr_symlink (call_frame_t *frame, xlator_t *this,
- const char *linkpath, loc_t *oldloc, mode_t umask, dict_t *params);
+afr_symlink(call_frame_t *frame, xlator_t *this, const char *linkpath,
+ loc_t *oldloc, mode_t umask, dict_t *params);
#endif /* __DIR_WRITE_H__ */
diff --git a/xlators/cluster/afr/src/afr-inode-read.c b/xlators/cluster/afr/src/afr-inode-read.c
index 8d3c823043f..c5521704de2 100644
--- a/xlators/cluster/afr/src/afr-inode-read.c
+++ b/xlators/cluster/afr/src/afr-inode-read.c
@@ -8,7 +8,6 @@
cases as published by the Free Software Foundation.
*/
-
#include <libgen.h>
#include <unistd.h>
#include <fnmatch.h>
@@ -16,20 +15,17 @@
#include <stdlib.h>
#include <signal.h>
-#include "glusterfs.h"
+#include <glusterfs/glusterfs.h>
#include "afr.h"
-#include "dict.h"
-#include "xlator.h"
-#include "hashfn.h"
-#include "logging.h"
-#include "list.h"
-#include "call-stub.h"
-#include "byte-order.h"
-#include "defaults.h"
-#include "common-utils.h"
-#include "compat-errno.h"
-#include "compat.h"
-#include "quota-common-utils.h"
+#include <glusterfs/dict.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/list.h>
+#include <glusterfs/byte-order.h>
+#include <glusterfs/defaults.h>
+#include <glusterfs/common-utils.h>
+#include <glusterfs/compat-errno.h>
+#include <glusterfs/compat.h>
+#include <glusterfs/quota-common-utils.h>
#include "afr-transaction.h"
#include "afr-messages.h"
@@ -44,146 +40,146 @@
* */
int
-afr_handle_quota_size (call_frame_t *frame, xlator_t *this)
+afr_handle_quota_size(call_frame_t *frame, xlator_t *this)
{
- unsigned char *readable = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- struct afr_reply *replies = NULL;
- int i = 0;
- int ret = 0;
- quota_meta_t size = {0, };
- quota_meta_t max_size = {0, };
- int readable_cnt = 0;
- int read_subvol = -1;
-
- local = frame->local;
- priv = this->private;
- replies = local->replies;
-
- readable = alloca0 (priv->child_count);
-
- afr_inode_read_subvol_get (local->inode, this, readable, 0, 0);
-
- readable_cnt = AFR_COUNT (readable, priv->child_count);
-
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret == -1)
- continue;
- if (readable_cnt && !readable[i])
- continue;
- if (!replies[i].xdata)
- continue;
- ret = quota_dict_get_meta (replies[i].xdata, QUOTA_SIZE_KEY,
- &size);
- if (ret == -1)
- continue;
- if (read_subvol == -1)
- read_subvol = i;
- if (size.size > max_size.size ||
- (size.file_count + size.dir_count) >
- (max_size.file_count + max_size.dir_count))
- read_subvol = i;
-
- if (size.size > max_size.size)
- max_size.size = size.size;
- if (size.file_count > max_size.file_count)
- max_size.file_count = size.file_count;
- if (size.dir_count > max_size.dir_count)
- max_size.dir_count = size.dir_count;
- }
-
- if (max_size.size == 0 && max_size.file_count == 0 &&
- max_size.dir_count == 0)
- return read_subvol;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret == -1)
- continue;
- if (readable_cnt && !readable[i])
- continue;
- if (!replies[i].xdata)
- continue;
- quota_dict_set_meta (replies[i].xdata, QUOTA_SIZE_KEY,
- &max_size, IA_IFDIR);
- }
-
+ unsigned char *readable = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ struct afr_reply *replies = NULL;
+ int i = 0;
+ int ret = 0;
+ quota_meta_t size = {
+ 0,
+ };
+ quota_meta_t max_size = {
+ 0,
+ };
+ int readable_cnt = 0;
+ int read_subvol = -1;
+
+ local = frame->local;
+ priv = this->private;
+ replies = local->replies;
+
+ readable = alloca0(priv->child_count);
+
+ afr_inode_read_subvol_get(local->inode, this, readable, 0, 0);
+
+ readable_cnt = AFR_COUNT(readable, priv->child_count);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret == -1)
+ continue;
+ if (readable_cnt && !readable[i])
+ continue;
+ if (!replies[i].xdata)
+ continue;
+ ret = quota_dict_get_meta(replies[i].xdata, QUOTA_SIZE_KEY,
+ SLEN(QUOTA_SIZE_KEY), &size);
+ if (ret == -1)
+ continue;
+ if (read_subvol == -1)
+ read_subvol = i;
+ if (size.size > max_size.size ||
+ (size.file_count + size.dir_count) >
+ (max_size.file_count + max_size.dir_count))
+ read_subvol = i;
+
+ if (size.size > max_size.size)
+ max_size.size = size.size;
+ if (size.file_count > max_size.file_count)
+ max_size.file_count = size.file_count;
+ if (size.dir_count > max_size.dir_count)
+ max_size.dir_count = size.dir_count;
+ }
+
+ if (max_size.size == 0 && max_size.file_count == 0 &&
+ max_size.dir_count == 0)
return read_subvol;
-}
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret == -1)
+ continue;
+ if (readable_cnt && !readable[i])
+ continue;
+ if (!replies[i].xdata)
+ continue;
+ quota_dict_set_meta(replies[i].xdata, QUOTA_SIZE_KEY, &max_size,
+ IA_IFDIR);
+ }
+
+ return read_subvol;
+}
/* {{{ access */
int
-afr_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+afr_access_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ afr_read_txn_continue(frame, this, (long)cookie);
+ return 0;
+ }
- AFR_STACK_UNWIND (access, frame, op_ret, op_errno, xdata);
+ AFR_STACK_UNWIND(access, frame, op_ret, op_errno, xdata);
- return 0;
+ return 0;
}
-
int
-afr_access_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_access_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
-
- priv = this->private;
- local = frame->local;
-
- if (subvol == -1) {
- AFR_STACK_UNWIND (access, frame, local->op_ret,
- local->op_errno, 0);
- return 0;
- }
-
- STACK_WIND_COOKIE (frame, afr_access_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->access,
- &local->loc, local->cont.access.mask,
- local->xdata_req);
- return 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ if (subvol == -1) {
+ AFR_STACK_UNWIND(access, frame, local->op_ret, local->op_errno, 0);
+ return 0;
+ }
+
+ STACK_WIND_COOKIE(frame, afr_access_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->access, &local->loc,
+ local->cont.access.mask, local->xdata_req);
+ return 0;
}
int
-afr_access (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int mask, dict_t *xdata)
+afr_access(call_frame_t *frame, xlator_t *this, loc_t *loc, int mask,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int op_errno = 0;
+ afr_local_t *local = NULL;
+ int op_errno = 0;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_ACCESS;
- loc_copy (&local->loc, loc);
- local->cont.access.mask = mask;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ local->op = GF_FOP_ACCESS;
+ loc_copy(&local->loc, loc);
+ local->cont.access.mask = mask;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- afr_read_txn (frame, this, loc->inode, afr_access_wind,
- AFR_METADATA_TRANSACTION);
+ afr_read_txn(frame, this, loc->inode, afr_access_wind,
+ AFR_METADATA_TRANSACTION);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (access, frame, -1, op_errno, NULL);
+ AFR_STACK_UNWIND(access, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
/* }}} */
@@ -191,152 +187,140 @@ out:
/* {{{ stat */
int
-afr_stat_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct iatt *buf, dict_t *xdata)
+afr_stat_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *buf, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ afr_read_txn_continue(frame, this, (long)cookie);
+ return 0;
+ }
- AFR_STACK_UNWIND (stat, frame, op_ret, op_errno, buf, xdata);
+ AFR_STACK_UNWIND(stat, frame, op_ret, op_errno, buf, xdata);
- return 0;
+ return 0;
}
-
int
-afr_stat_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_stat_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
-
- priv = this->private;
- local = frame->local;
-
- if (subvol == -1) {
- AFR_STACK_UNWIND (stat, frame, local->op_ret, local->op_errno,
- 0, 0);
- return 0;
- }
-
- STACK_WIND_COOKIE (frame, afr_stat_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->stat,
- &local->loc, local->xdata_req);
- return 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ if (subvol == -1) {
+ AFR_STACK_UNWIND(stat, frame, local->op_ret, local->op_errno, 0, 0);
+ return 0;
+ }
+
+ STACK_WIND_COOKIE(
+ frame, afr_stat_cbk, (void *)(long)subvol, priv->children[subvol],
+ priv->children[subvol]->fops->stat, &local->loc, local->xdata_req);
+ return 0;
}
int
-afr_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+afr_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int op_errno = 0;
+ afr_local_t *local = NULL;
+ int op_errno = 0;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_STAT;
- loc_copy (&local->loc, loc);
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ local->op = GF_FOP_STAT;
+ loc_copy(&local->loc, loc);
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- afr_read_txn (frame, this, loc->inode, afr_stat_wind,
- AFR_DATA_TRANSACTION);
+ afr_read_txn(frame, this, loc->inode, afr_stat_wind, AFR_DATA_TRANSACTION);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND(stat, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
-
/* }}} */
/* {{{ fstat */
int
-afr_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- dict_t *xdata)
+afr_fstat_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *buf, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ afr_read_txn_continue(frame, this, (long)cookie);
+ return 0;
+ }
- AFR_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf, xdata);
+ AFR_STACK_UNWIND(fstat, frame, op_ret, op_errno, buf, xdata);
- return 0;
+ return 0;
}
-
int
-afr_fstat_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_fstat_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
-
- priv = this->private;
- local = frame->local;
-
- if (subvol == -1) {
- AFR_STACK_UNWIND (fstat, frame, local->op_ret, local->op_errno,
- 0, 0);
- return 0;
- }
-
- STACK_WIND_COOKIE (frame, afr_fstat_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->fstat,
- local->fd, local->xdata_req);
- return 0;
-}
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ if (subvol == -1) {
+ AFR_STACK_UNWIND(fstat, frame, local->op_ret, local->op_errno, 0, 0);
+ return 0;
+ }
+ STACK_WIND_COOKIE(
+ frame, afr_fstat_cbk, (void *)(long)subvol, priv->children[subvol],
+ priv->children[subvol]->fops->fstat, local->fd, local->xdata_req);
+ return 0;
+}
int32_t
-afr_fstat (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *xdata)
+afr_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int op_errno = 0;
+ afr_local_t *local = NULL;
+ int op_errno = 0;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_FSTAT;
- local->fd = fd_ref (fd);
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ local->op = GF_FOP_FSTAT;
+ local->fd = fd_ref(fd);
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- afr_read_txn (frame, this, fd->inode, afr_fstat_wind,
- AFR_DATA_TRANSACTION);
+ afr_read_txn(frame, this, fd->inode, afr_fstat_wind, AFR_DATA_TRANSACTION);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (fstat, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND(fstat, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
/* }}} */
@@ -344,1563 +328,1493 @@ out:
/* {{{ readlink */
int
-afr_readlink_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- const char *buf, struct iatt *sbuf, dict_t *xdata)
+afr_readlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, const char *buf,
+ struct iatt *sbuf, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = op_errno;
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ afr_read_txn_continue(frame, this, (long)cookie);
+ return 0;
+ }
- AFR_STACK_UNWIND (readlink, frame, op_ret, op_errno,
- buf, sbuf, xdata);
- return 0;
+ AFR_STACK_UNWIND(readlink, frame, op_ret, op_errno, buf, sbuf, xdata);
+ return 0;
}
int
-afr_readlink_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_readlink_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- if (subvol == -1) {
- AFR_STACK_UNWIND (readlink, frame, local->op_ret,
- local->op_errno, 0, 0, 0);
- return 0;
- }
-
- STACK_WIND_COOKIE (frame, afr_readlink_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->readlink,
- &local->loc, local->cont.readlink.size,
- local->xdata_req);
- return 0;
-}
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ local = frame->local;
+ priv = this->private;
+
+ if (subvol == -1) {
+ AFR_STACK_UNWIND(readlink, frame, local->op_ret, local->op_errno, 0, 0,
+ 0);
+ return 0;
+ }
+
+ STACK_WIND_COOKIE(frame, afr_readlink_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->readlink, &local->loc,
+ local->cont.readlink.size, local->xdata_req);
+ return 0;
+}
int
-afr_readlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc, size_t size, dict_t *xdata)
+afr_readlink(call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+ dict_t *xdata)
{
- afr_local_t * local = NULL;
- int32_t op_errno = 0;
+ afr_local_t *local = NULL;
+ int32_t op_errno = 0;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_READLINK;
- loc_copy (&local->loc, loc);
- local->cont.readlink.size = size;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ local->op = GF_FOP_READLINK;
+ loc_copy(&local->loc, loc);
+ local->cont.readlink.size = size;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- afr_read_txn (frame, this, loc->inode, afr_readlink_wind,
- AFR_DATA_TRANSACTION);
+ afr_read_txn(frame, this, loc->inode, afr_readlink_wind,
+ AFR_DATA_TRANSACTION);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND(readlink, frame, -1, op_errno, 0, 0, 0);
+ AFR_STACK_UNWIND(readlink, frame, -1, op_errno, 0, 0, 0);
- return 0;
+ return 0;
}
-
/* }}} */
/* {{{ getxattr */
struct _xattr_key {
- char *key;
- struct list_head list;
+ char *key;
+ struct list_head list;
};
-
int
-__gather_xattr_keys (dict_t *dict, char *key, data_t *value,
- void *data)
+__gather_xattr_keys(dict_t *dict, char *key, data_t *value, void *data)
{
- struct list_head * list = data;
- struct _xattr_key * xkey = NULL;
-
- if (!strncmp (key, AFR_XATTR_PREFIX,
- strlen (AFR_XATTR_PREFIX))) {
+ struct list_head *list = data;
+ struct _xattr_key *xkey = NULL;
- xkey = GF_CALLOC (1, sizeof (*xkey), gf_afr_mt_xattr_key);
- if (!xkey)
- return -1;
+ if (!strncmp(key, AFR_XATTR_PREFIX, SLEN(AFR_XATTR_PREFIX))) {
+ xkey = GF_MALLOC(sizeof(*xkey), gf_afr_mt_xattr_key);
+ if (!xkey)
+ return -1;
- xkey->key = key;
- INIT_LIST_HEAD (&xkey->list);
+ xkey->key = key;
+ INIT_LIST_HEAD(&xkey->list);
- list_add_tail (&xkey->list, list);
- }
- return 0;
+ list_add_tail(&xkey->list, list);
+ }
+ return 0;
}
-
void
-afr_filter_xattrs (dict_t *dict)
+afr_filter_xattrs(dict_t *dict)
{
- struct list_head keys = {0,};
- struct _xattr_key *key = NULL;
- struct _xattr_key *tmp = NULL;
+ struct list_head keys = {
+ 0,
+ };
+ struct _xattr_key *key = NULL;
+ struct _xattr_key *tmp = NULL;
- INIT_LIST_HEAD (&keys);
+ INIT_LIST_HEAD(&keys);
- dict_foreach (dict, __gather_xattr_keys,
- (void *) &keys);
+ dict_foreach(dict, __gather_xattr_keys, (void *)&keys);
- list_for_each_entry_safe (key, tmp, &keys, list) {
- dict_del (dict, key->key);
+ list_for_each_entry_safe(key, tmp, &keys, list)
+ {
+ dict_del(dict, key->key);
- list_del_init (&key->list);
+ list_del_init(&key->list);
- GF_FREE (key);
- }
+ GF_FREE(key);
+ }
}
-static
-gf_boolean_t
-afr_getxattr_ignorable_errnos (int32_t op_errno)
+static gf_boolean_t
+afr_getxattr_ignorable_errnos(int32_t op_errno)
{
- if (op_errno == ENODATA || op_errno == ENOTSUP || op_errno == ERANGE ||
- op_errno == ENAMETOOLONG)
- return _gf_true;
+ if (op_errno == ENODATA || op_errno == ENOTSUP || op_errno == ERANGE ||
+ op_errno == ENAMETOOLONG)
+ return _gf_true;
- return _gf_false;
+ return _gf_false;
}
int
-afr_getxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0 && !afr_getxattr_ignorable_errnos(op_errno)) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
+ if (op_ret < 0 && !afr_getxattr_ignorable_errnos(op_errno)) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ afr_read_txn_continue(frame, this, (long)cookie);
+ return 0;
+ }
- if (dict)
- afr_filter_xattrs (dict);
+ if (dict)
+ afr_filter_xattrs(dict);
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
+ AFR_STACK_UNWIND(getxattr, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ return 0;
}
-
int
-afr_getxattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_getxattr_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- if (subvol == -1) {
- AFR_STACK_UNWIND (getxattr, frame, local->op_ret,
- local->op_errno, NULL, NULL);
- return 0;
- }
-
- STACK_WIND_COOKIE (frame, afr_getxattr_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->getxattr,
- &local->loc, local->cont.getxattr.name,
- local->xdata_req);
- return 0;
-}
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ local = frame->local;
+ priv = this->private;
+
+ if (subvol == -1) {
+ AFR_STACK_UNWIND(getxattr, frame, local->op_ret, local->op_errno, NULL,
+ NULL);
+ return 0;
+ }
+
+ STACK_WIND_COOKIE(frame, afr_getxattr_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->getxattr, &local->loc,
+ local->cont.getxattr.name, local->xdata_req);
+ return 0;
+}
int32_t
-afr_getxattr_unwind (call_frame_t *frame, int op_ret, int op_errno,
- dict_t *dict, dict_t *xdata)
+afr_getxattr_unwind(call_frame_t *frame, int op_ret, int op_errno, dict_t *dict,
+ dict_t *xdata)
{
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ AFR_STACK_UNWIND(getxattr, frame, op_ret, op_errno, dict, xdata);
+ return 0;
}
int32_t
-afr_fgetxattr_clrlk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_fgetxattr_clrlk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- xlator_t **children = NULL;
- dict_t *xattr = NULL;
- char *tmp_report = NULL;
- char lk_summary[1024] = {0,};
- int serz_len = 0;
- int32_t callcnt = 0;
- long int cky = 0;
- int ret = 0;
-
- priv = this->private;
- children = priv->children;
-
- local = frame->local;
- cky = (long) cookie;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
- if (op_ret == -1)
- local->replies[cky].op_errno = op_errno;
-
- if (!local->dict)
- local->dict = dict_new ();
- if (local->dict) {
- ret = dict_get_str (dict, local->cont.getxattr.name,
- &tmp_report);
- if (ret)
- goto unlock;
- ret = dict_set_dynstr (local->dict,
- children[cky]->name,
- gf_strdup (tmp_report));
- if (ret)
- goto unlock;
- }
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ dict_t *xattr = NULL;
+ char *tmp_report = NULL;
+ char lk_summary[1024] = {
+ 0,
+ };
+ int serz_len = 0;
+ int32_t callcnt = 0;
+ long int cky = 0;
+ int ret = 0;
+ int keylen = 0;
+ int children_keylen = 0;
+
+ priv = this->private;
+ children = priv->children;
+
+ local = frame->local;
+ cky = (long)cookie;
+ keylen = strlen(local->cont.getxattr.name);
+ children_keylen = strlen(children[cky]->name);
+
+ LOCK(&frame->lock);
+ {
+ callcnt = --local->call_count;
+ if (op_ret == -1)
+ local->replies[cky].op_errno = op_errno;
+
+ if (!local->dict)
+ local->dict = dict_new();
+ if (local->dict) {
+ ret = dict_get_strn(dict, local->cont.getxattr.name, keylen,
+ &tmp_report);
+ if (ret)
+ goto unlock;
+ ret = dict_set_dynstrn(local->dict, children[cky]->name,
+ children_keylen, gf_strdup(tmp_report));
+ if (ret)
+ goto unlock;
}
+ }
unlock:
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- xattr = dict_new ();
- if (!xattr) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
- ret = dict_serialize_value_with_delim (local->dict,
- lk_summary,
- &serz_len, '\n');
- if (ret) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
- if (serz_len == -1)
- snprintf (lk_summary, sizeof (lk_summary),
- "No locks cleared.");
- ret = dict_set_dynstr (xattr, local->cont.getxattr.name,
- gf_strdup (lk_summary));
- if (ret) {
- op_ret = -1;
- op_errno = ENOMEM;
- gf_msg (this->name, GF_LOG_ERROR,
- ENOMEM, AFR_MSG_DICT_SET_FAILED,
- "Error setting dictionary");
- goto unwind;
- }
+ UNLOCK(&frame->lock);
+
+ if (!callcnt) {
+ xattr = dict_new();
+ if (!xattr) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ ret = dict_serialize_value_with_delim(local->dict, lk_summary,
+ &serz_len, '\n');
+ if (ret) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ if (serz_len == -1)
+ snprintf(lk_summary, sizeof(lk_summary), "No locks cleared.");
+ ret = dict_set_dynstrn(xattr, local->cont.getxattr.name, keylen,
+ gf_strdup(lk_summary));
+ if (ret) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, AFR_MSG_DICT_SET_FAILED,
+ "Error setting dictionary");
+ goto unwind;
+ }
- op_errno = afr_final_errno (local, priv);
+ op_errno = afr_final_errno(local, priv);
-unwind:
- AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, xattr,
- xdata);
- if (xattr)
- dict_unref (xattr);
- }
+ unwind:
+ AFR_STACK_UNWIND(fgetxattr, frame, op_ret, op_errno, xattr, xdata);
+ if (xattr)
+ dict_unref(xattr);
+ }
- return ret;
+ return ret;
}
int32_t
-afr_getxattr_clrlk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_getxattr_clrlk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- xlator_t **children = NULL;
- dict_t *xattr = NULL;
- char *tmp_report = NULL;
- char lk_summary[1024] = {0,};
- int serz_len = 0;
- int32_t callcnt = 0;
- long int cky = 0;
- int ret = 0;
-
- priv = this->private;
- children = priv->children;
-
- local = frame->local;
- cky = (long) cookie;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
- if (op_ret == -1)
- local->replies[cky].op_errno = op_errno;
-
- if (!local->dict)
- local->dict = dict_new ();
- if (local->dict) {
- ret = dict_get_str (dict, local->cont.getxattr.name,
- &tmp_report);
- if (ret)
- goto unlock;
- ret = dict_set_dynstr (local->dict,
- children[cky]->name,
- gf_strdup (tmp_report));
- if (ret)
- goto unlock;
- }
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ xlator_t **children = NULL;
+ dict_t *xattr = NULL;
+ char *tmp_report = NULL;
+ char lk_summary[1024] = {
+ 0,
+ };
+ int serz_len = 0;
+ int32_t callcnt = 0;
+ long int cky = 0;
+ int ret = 0;
+ int keylen = 0;
+ int children_keylen = 0;
+
+ priv = this->private;
+ children = priv->children;
+
+ local = frame->local;
+ cky = (long)cookie;
+
+ keylen = strlen(local->cont.getxattr.name);
+ children_keylen = strlen(children[cky]->name);
+
+ LOCK(&frame->lock);
+ {
+ callcnt = --local->call_count;
+ if (op_ret == -1)
+ local->replies[cky].op_errno = op_errno;
+
+ if (!local->dict)
+ local->dict = dict_new();
+ if (local->dict) {
+ ret = dict_get_strn(dict, local->cont.getxattr.name, keylen,
+ &tmp_report);
+ if (ret)
+ goto unlock;
+ ret = dict_set_dynstrn(local->dict, children[cky]->name,
+ children_keylen, gf_strdup(tmp_report));
+ if (ret)
+ goto unlock;
}
+ }
unlock:
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- xattr = dict_new ();
- if (!xattr) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
- ret = dict_serialize_value_with_delim (local->dict,
- lk_summary,
- &serz_len, '\n');
- if (ret) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
- if (serz_len == -1)
- snprintf (lk_summary, sizeof (lk_summary),
- "No locks cleared.");
- ret = dict_set_dynstr (xattr, local->cont.getxattr.name,
- gf_strdup (lk_summary));
- if (ret) {
- op_ret = -1;
- op_errno = ENOMEM;
- gf_msg (this->name, GF_LOG_ERROR,
- ENOMEM, AFR_MSG_DICT_SET_FAILED,
- "Error setting dictionary");
- goto unwind;
- }
+ UNLOCK(&frame->lock);
+
+ if (!callcnt) {
+ xattr = dict_new();
+ if (!xattr) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ ret = dict_serialize_value_with_delim(local->dict, lk_summary,
+ &serz_len, '\n');
+ if (ret) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ if (serz_len == -1)
+ snprintf(lk_summary, sizeof(lk_summary), "No locks cleared.");
+ ret = dict_set_dynstrn(xattr, local->cont.getxattr.name, keylen,
+ gf_strdup(lk_summary));
+ if (ret) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, AFR_MSG_DICT_SET_FAILED,
+ "Error setting dictionary");
+ goto unwind;
+ }
- op_errno = afr_final_errno (local, priv);
+ op_errno = afr_final_errno(local, priv);
-unwind:
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr, xdata);
+ unwind:
+ AFR_STACK_UNWIND(getxattr, frame, op_ret, op_errno, xattr, xdata);
- if (xattr)
- dict_unref (xattr);
- }
+ if (xattr)
+ dict_unref(xattr);
+ }
- return ret;
+ return ret;
}
/**
* node-uuid cbk uses next child querying mechanism
*/
int32_t
-afr_getxattr_node_uuid_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_getxattr_node_uuid_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- xlator_t **children = NULL;
- int unwind = 1;
- int curr_call_child = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ xlator_t **children = NULL;
+ int unwind = 1;
+ int curr_call_child = 0;
- priv = this->private;
- children = priv->children;
+ priv = this->private;
+ children = priv->children;
- local = frame->local;
+ local = frame->local;
- if (op_ret == -1) { /** query the _next_ child */
-
- /**
- * _current_ becomes _next_
- * If done with all childs and yet no success; give up !
- */
- curr_call_child = (int) ((long)cookie);
- if (++curr_call_child == priv->child_count)
- goto unwind;
-
- gf_msg_debug (this->name, op_errno,
- "op_ret (-1): Re-querying afr-child (%d/%d)",
- curr_call_child, priv->child_count);
-
- unwind = 0;
- STACK_WIND_COOKIE (frame, afr_getxattr_node_uuid_cbk,
- (void *) (long) curr_call_child,
- children[curr_call_child],
- children[curr_call_child]->fops->getxattr,
- &local->loc,
- local->cont.getxattr.name,
- local->xdata_req);
- }
+ if (op_ret == -1) { /** query the _next_ child */
- unwind:
- if (unwind)
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict,
- xdata);
+ /**
+ * _current_ becomes _next_
+ * If done with all children and yet no success; give up !
+ */
+ curr_call_child = (int)((long)cookie);
+ if (++curr_call_child == priv->child_count)
+ goto unwind;
+
+ gf_msg_debug(this->name, op_errno,
+ "op_ret (-1): Re-querying afr-child (%d/%d)",
+ curr_call_child, priv->child_count);
+
+ unwind = 0;
+ STACK_WIND_COOKIE(
+ frame, afr_getxattr_node_uuid_cbk, (void *)(long)curr_call_child,
+ children[curr_call_child],
+ children[curr_call_child]->fops->getxattr, &local->loc,
+ local->cont.getxattr.name, local->xdata_req);
+ }
- return 0;
+unwind:
+ if (unwind)
+ AFR_STACK_UNWIND(getxattr, frame, op_ret, op_errno, dict, xdata);
+
+ return 0;
}
/**
* list-node-uuids cbk returns the list of node_uuids for the subvolume.
*/
int32_t
-afr_getxattr_list_node_uuids_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *dict, dict_t *xdata)
+afr_getxattr_list_node_uuids_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int32_t callcnt = 0;
- int ret = 0;
- char *xattr_serz = NULL;
- long cky = 0;
- int32_t tlen = 0;
-
- local = frame->local;
- priv = this->private;
- cky = (long) cookie;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
- local->replies[cky].valid = 1;
- local->replies[cky].op_ret = op_ret;
- local->replies[cky].op_errno = op_errno;
-
- if (op_ret < 0)
- goto unlock;
-
- local->op_ret = 0;
-
- if (!local->xdata_rsp && xdata)
- local->xdata_rsp = dict_ref (xdata);
- local->replies[cky].xattr = dict_ref (dict);
- }
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int32_t callcnt = 0;
+ int ret = 0;
+ char *xattr_serz = NULL;
+ long cky = 0;
+ int32_t tlen = 0;
+
+ local = frame->local;
+ priv = this->private;
+ cky = (long)cookie;
+
+ LOCK(&frame->lock);
+ {
+ callcnt = --local->call_count;
+ local->replies[cky].valid = 1;
+ local->replies[cky].op_ret = op_ret;
+ local->replies[cky].op_errno = op_errno;
+
+ if (op_ret < 0)
+ goto unlock;
+
+ local->op_ret = 0;
+
+ if (!local->xdata_rsp && xdata)
+ local->xdata_rsp = dict_ref(xdata);
+ local->replies[cky].xattr = dict_ref(dict);
+ }
unlock:
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
+ UNLOCK(&frame->lock);
- if (local->op_ret != 0) {
- /* All bricks gave an error. */
- local->op_errno = afr_final_errno (local, priv);
- goto unwind;
- }
+ if (!callcnt) {
+ if (local->op_ret != 0) {
+ /* All bricks gave an error. */
+ local->op_errno = afr_final_errno(local, priv);
+ goto unwind;
+ }
- /*Since we store the UUID0_STR as node uuid for down bricks and
- *for non zero op_ret, assigning length to priv->child_count
- *number of uuids*/
- local->cont.getxattr.xattr_len = (strlen (UUID0_STR) + 2) *
- priv->child_count;
-
- if (!local->dict)
- local->dict = dict_new ();
- if (!local->dict) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unwind;
- }
+ /*Since we store the UUID0_STR as node uuid for down bricks and
+ *for non zero op_ret, assigning length to priv->child_count
+ *number of uuids*/
+ local->cont.getxattr.xattr_len = (SLEN(UUID0_STR) + 2) *
+ priv->child_count;
+
+ if (!local->dict)
+ local->dict = dict_new();
+ if (!local->dict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
- xattr_serz = GF_CALLOC (local->cont.getxattr.xattr_len,
- sizeof (char), gf_common_mt_char);
+ xattr_serz = GF_CALLOC(local->cont.getxattr.xattr_len, sizeof(char),
+ gf_common_mt_char);
- if (!xattr_serz) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unwind;
- }
-
- ret = afr_serialize_xattrs_with_delimiter (frame, this,
- xattr_serz,
- UUID0_STR, &tlen,
- ' ');
- if (ret) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unwind;
- }
- ret = dict_set_dynstr (local->dict,
- GF_XATTR_LIST_NODE_UUIDS_KEY,
- xattr_serz);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Cannot set node_uuid key in dict");
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- } else {
- local->op_ret = local->cont.getxattr.xattr_len - 1;
- local->op_errno = 0;
- }
+ if (!xattr_serz) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
-unwind:
- AFR_STACK_UNWIND (getxattr, frame, local->op_ret,
- local->op_errno, local->dict,
- local->xdata_rsp);
+ ret = afr_serialize_xattrs_with_delimiter(frame, this, xattr_serz,
+ UUID0_STR, &tlen, ' ');
+ if (ret) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ GF_FREE(xattr_serz);
+ goto unwind;
+ }
+ ret = dict_set_dynstr_sizen(local->dict, GF_XATTR_LIST_NODE_UUIDS_KEY,
+ xattr_serz);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Cannot set node_uuid key in dict");
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ if (ret == -EINVAL)
+ GF_FREE(xattr_serz);
+ } else {
+ local->op_ret = local->cont.getxattr.xattr_len - 1;
+ local->op_errno = 0;
}
- return ret;
-}
+ unwind:
+ AFR_STACK_UNWIND(getxattr, frame, local->op_ret, local->op_errno,
+ local->dict, local->xdata_rsp);
+ }
+ return ret;
+}
int32_t
-afr_getxattr_quota_size_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_getxattr_quota_size_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- int idx = (long) cookie;
- int call_count = 0;
- afr_local_t *local = frame->local;
- int read_subvol = -1;
-
- local->replies[idx].valid = 1;
- local->replies[idx].op_ret = op_ret;
- local->replies[idx].op_errno = op_errno;
- if (dict)
- local->replies[idx].xdata = dict_ref (dict);
- call_count = afr_frame_return (frame);
- if (call_count == 0) {
- local->inode = inode_ref (local->loc.inode);
- read_subvol = afr_handle_quota_size (frame, this);
- if (read_subvol != -1) {
- op_ret = local->replies[read_subvol].op_ret;
- op_errno = local->replies[read_subvol].op_errno;
- dict = local->replies[read_subvol].xdata;
- }
- AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict,
- xdata);
+ int idx = (long)cookie;
+ int call_count = 0;
+ afr_local_t *local = frame->local;
+ int read_subvol = -1;
+
+ local->replies[idx].valid = 1;
+ local->replies[idx].op_ret = op_ret;
+ local->replies[idx].op_errno = op_errno;
+ if (dict)
+ local->replies[idx].xdata = dict_ref(dict);
+ call_count = afr_frame_return(frame);
+ if (call_count == 0) {
+ local->inode = inode_ref(local->loc.inode);
+ read_subvol = afr_handle_quota_size(frame, this);
+ if (read_subvol != -1) {
+ op_ret = local->replies[read_subvol].op_ret;
+ op_errno = local->replies[read_subvol].op_errno;
+ dict = local->replies[read_subvol].xdata;
}
+ AFR_STACK_UNWIND(getxattr, frame, op_ret, op_errno, dict, xdata);
+ }
- return 0;
+ return 0;
}
int32_t
-afr_getxattr_lockinfo_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_getxattr_lockinfo_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- int call_cnt = 0, len = 0;
- char *lockinfo_buf = NULL;
- dict_t *lockinfo = NULL, *newdict = NULL;
- afr_local_t *local = NULL;
+ int call_cnt = 0, len = 0;
+ char *lockinfo_buf = NULL;
+ dict_t *lockinfo = NULL, *newdict = NULL;
+ afr_local_t *local = NULL;
- LOCK (&frame->lock);
- {
- local = frame->local;
-
- call_cnt = --local->call_count;
+ LOCK(&frame->lock);
+ {
+ local = frame->local;
- if ((op_ret < 0) || (!dict && !xdata)) {
- goto unlock;
- }
+ call_cnt = --local->call_count;
- if (xdata) {
- if (!local->xdata_rsp) {
- local->xdata_rsp = dict_new ();
- if (!local->xdata_rsp) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unlock;
- }
- }
- }
+ if ((op_ret < 0) || (!dict && !xdata)) {
+ goto unlock;
+ }
- if (!dict) {
- goto unlock;
+ if (xdata) {
+ if (!local->xdata_rsp) {
+ local->xdata_rsp = dict_new();
+ if (!local->xdata_rsp) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unlock;
}
+ }
+ }
- op_ret = dict_get_ptr_and_len (dict, GF_XATTR_LOCKINFO_KEY,
- (void **)&lockinfo_buf, &len);
+ if (!dict) {
+ goto unlock;
+ }
- if (!lockinfo_buf) {
- goto unlock;
- }
+ op_ret = dict_get_ptr_and_len(dict, GF_XATTR_LOCKINFO_KEY,
+ (void **)&lockinfo_buf, &len);
- if (!local->dict) {
- local->dict = dict_new ();
- if (!local->dict) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unlock;
- }
- }
- }
-unlock:
- UNLOCK (&frame->lock);
-
- if (lockinfo_buf != NULL) {
- lockinfo = dict_new ();
- if (lockinfo == NULL) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- } else {
- op_ret = dict_unserialize (lockinfo_buf, len,
- &lockinfo);
-
- if (lockinfo && local->dict) {
- dict_copy (lockinfo, local->dict);
- }
- }
+ if (!lockinfo_buf) {
+ goto unlock;
}
- if (xdata && local->xdata_rsp) {
- dict_copy (xdata, local->xdata_rsp);
+ if (!local->dict) {
+ local->dict = dict_new();
+ if (!local->dict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unlock;
+ }
}
+ }
+unlock:
+ UNLOCK(&frame->lock);
- if (!call_cnt) {
- newdict = dict_new ();
- if (!newdict) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unwind;
- }
-
- len = dict_serialized_length (local->dict);
- if (len <= 0) {
- goto unwind;
- }
-
- lockinfo_buf = GF_CALLOC (1, len, gf_common_mt_char);
- if (!lockinfo_buf) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unwind;
- }
+ if (lockinfo_buf != NULL) {
+ lockinfo = dict_new();
+ if (lockinfo == NULL) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ } else {
+ op_ret = dict_unserialize(lockinfo_buf, len, &lockinfo);
- op_ret = dict_serialize (local->dict, lockinfo_buf);
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = -op_ret;
- }
+ if (lockinfo && local->dict) {
+ dict_copy(lockinfo, local->dict);
+ }
+ }
+ }
+
+ if (xdata && local->xdata_rsp) {
+ dict_copy(xdata, local->xdata_rsp);
+ }
+
+ if (!call_cnt) {
+ newdict = dict_new();
+ if (!newdict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
- op_ret = dict_set_dynptr (newdict, GF_XATTR_LOCKINFO_KEY,
- (void *)lockinfo_buf, len);
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = -op_ret;
- goto unwind;
- }
+ op_ret = dict_allocate_and_serialize(
+ local->dict, (char **)&lockinfo_buf, (unsigned int *)&len);
+ if (op_ret != 0) {
+ local->op_ret = -1;
+ goto unwind;
+ }
- unwind:
- AFR_STACK_UNWIND (getxattr, frame, op_ret,
- op_errno, newdict,
- local->xdata_rsp);
+ op_ret = dict_set_dynptr(newdict, GF_XATTR_LOCKINFO_KEY,
+ (void *)lockinfo_buf, len);
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -op_ret;
+ goto unwind;
}
- dict_unref (lockinfo);
+ unwind:
+ AFR_STACK_UNWIND(getxattr, frame, op_ret, op_errno, newdict,
+ local->xdata_rsp);
+ }
- return 0;
+ dict_unref(lockinfo);
+
+ return 0;
}
int32_t
-afr_fgetxattr_lockinfo_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_fgetxattr_lockinfo_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- int call_cnt = 0, len = 0;
- char *lockinfo_buf = NULL;
- dict_t *lockinfo = NULL, *newdict = NULL;
- afr_local_t *local = NULL;
+ int call_cnt = 0, len = 0;
+ char *lockinfo_buf = NULL;
+ dict_t *lockinfo = NULL, *newdict = NULL;
+ afr_local_t *local = NULL;
- LOCK (&frame->lock);
- {
- local = frame->local;
-
- call_cnt = --local->call_count;
+ LOCK(&frame->lock);
+ {
+ local = frame->local;
- if ((op_ret < 0) || (!dict && !xdata)) {
- goto unlock;
- }
+ call_cnt = --local->call_count;
- if (xdata) {
- if (!local->xdata_rsp) {
- local->xdata_rsp = dict_new ();
- if (!local->xdata_rsp) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unlock;
- }
- }
- }
+ if ((op_ret < 0) || (!dict && !xdata)) {
+ goto unlock;
+ }
- if (!dict) {
- goto unlock;
+ if (xdata) {
+ if (!local->xdata_rsp) {
+ local->xdata_rsp = dict_new();
+ if (!local->xdata_rsp) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unlock;
}
+ }
+ }
- op_ret = dict_get_ptr_and_len (dict, GF_XATTR_LOCKINFO_KEY,
- (void **)&lockinfo_buf, &len);
+ if (!dict) {
+ goto unlock;
+ }
- if (!lockinfo_buf) {
- goto unlock;
- }
+ op_ret = dict_get_ptr_and_len(dict, GF_XATTR_LOCKINFO_KEY,
+ (void **)&lockinfo_buf, &len);
- if (!local->dict) {
- local->dict = dict_new ();
- if (!local->dict) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unlock;
- }
- }
- }
-unlock:
- UNLOCK (&frame->lock);
-
- if (lockinfo_buf != NULL) {
- lockinfo = dict_new ();
- if (lockinfo == NULL) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- } else {
- op_ret = dict_unserialize (lockinfo_buf, len,
- &lockinfo);
-
- if (lockinfo && local->dict) {
- dict_copy (lockinfo, local->dict);
- }
- }
+ if (!lockinfo_buf) {
+ goto unlock;
}
- if (xdata && local->xdata_rsp) {
- dict_copy (xdata, local->xdata_rsp);
+ if (!local->dict) {
+ local->dict = dict_new();
+ if (!local->dict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unlock;
+ }
}
+ }
+unlock:
+ UNLOCK(&frame->lock);
- if (!call_cnt) {
- newdict = dict_new ();
- if (!newdict) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unwind;
- }
-
- len = dict_serialized_length (local->dict);
- if (len <= 0) {
- goto unwind;
- }
-
- lockinfo_buf = GF_CALLOC (1, len, gf_common_mt_char);
- if (!lockinfo_buf) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unwind;
- }
+ if (lockinfo_buf != NULL) {
+ lockinfo = dict_new();
+ if (lockinfo == NULL) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ } else {
+ op_ret = dict_unserialize(lockinfo_buf, len, &lockinfo);
- op_ret = dict_serialize (local->dict, lockinfo_buf);
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = -op_ret;
- }
+ if (lockinfo && local->dict) {
+ dict_copy(lockinfo, local->dict);
+ }
+ }
+ }
+
+ if (xdata && local->xdata_rsp) {
+ dict_copy(xdata, local->xdata_rsp);
+ }
+
+ if (!call_cnt) {
+ newdict = dict_new();
+ if (!newdict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
- op_ret = dict_set_dynptr (newdict, GF_XATTR_LOCKINFO_KEY,
- (void *)lockinfo_buf, len);
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = -op_ret;
- goto unwind;
- }
+ op_ret = dict_allocate_and_serialize(
+ local->dict, (char **)&lockinfo_buf, (unsigned int *)&len);
+ if (op_ret != 0) {
+ local->op_ret = -1;
+ goto unwind;
+ }
- unwind:
- AFR_STACK_UNWIND (fgetxattr, frame, op_ret,
- op_errno, newdict,
- local->xdata_rsp);
+ op_ret = dict_set_dynptr(newdict, GF_XATTR_LOCKINFO_KEY,
+ (void *)lockinfo_buf, len);
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -op_ret;
+ goto unwind;
}
- dict_unref (lockinfo);
+ unwind:
+ AFR_STACK_UNWIND(fgetxattr, frame, op_ret, op_errno, newdict,
+ local->xdata_rsp);
+ }
- return 0;
+ dict_unref(lockinfo);
+
+ return 0;
}
int32_t
-afr_fgetxattr_pathinfo_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_fgetxattr_pathinfo_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int32_t callcnt = 0;
- int ret = 0;
- char *xattr = NULL;
- char *xattr_serz = NULL;
- char xattr_cky[1024] = {0,};
- dict_t *nxattr = NULL;
- long cky = 0;
- int32_t padding = 0;
- int32_t tlen = 0;
-
- if (!frame || !frame->local || !this) {
- gf_msg ("", GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_ARG, "possible NULL deref");
- goto out;
+ afr_local_t *local = NULL;
+ int32_t callcnt = 0;
+ int ret = 0;
+ char *xattr = NULL;
+ char *xattr_serz = NULL;
+ int keylen = 0;
+ char xattr_cky[1024] = {
+ 0,
+ };
+ int xattr_cky_len = 0;
+ dict_t *nxattr = NULL;
+ long cky = 0;
+ int32_t padding = 0;
+ int32_t tlen = 0;
+
+ if (!frame || !frame->local || !this) {
+ gf_msg("", GF_LOG_ERROR, 0, AFR_MSG_INVALID_ARG, "possible NULL deref");
+ goto out;
+ }
+
+ local = frame->local;
+ cky = (long)cookie;
+ keylen = strlen(local->cont.getxattr.name);
+ xattr_cky_len = snprintf(xattr_cky, sizeof(xattr_cky), "%s-%ld",
+ local->cont.getxattr.name, cky);
+ LOCK(&frame->lock);
+ {
+ callcnt = --local->call_count;
+
+ if (op_ret < 0) {
+ local->op_errno = op_errno;
+ } else {
+ local->op_ret = op_ret;
+ if (!local->xdata_rsp && xdata)
+ local->xdata_rsp = dict_ref(xdata);
}
- local = frame->local;
- cky = (long) cookie;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret < 0) {
- local->op_errno = op_errno;
- } else {
- local->op_ret = op_ret;
- if (!local->xdata_rsp && xdata)
- local->xdata_rsp = dict_ref (xdata);
- }
+ if (!dict || (op_ret < 0))
+ goto unlock;
- if (!dict || (op_ret < 0))
- goto unlock;
-
- if (!local->dict)
- local->dict = dict_new ();
-
- if (local->dict) {
- ret = dict_get_str (dict,
- local->cont.getxattr.name,
- &xattr);
- if (ret)
- goto unlock;
-
- xattr = gf_strdup (xattr);
-
- (void)snprintf (xattr_cky, 1024, "%s-%ld",
- local->cont.getxattr.name, cky);
- ret = dict_set_dynstr (local->dict,
- xattr_cky, xattr);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Cannot set xattr cookie key");
- goto unlock;
- }
-
- local->cont.getxattr.xattr_len
- += strlen (xattr) + 1;
- }
+ if (!local->dict) {
+ local->dict = dict_new();
+ if (!local->dict)
+ goto unlock;
+ }
+ ret = dict_get_strn(dict, local->cont.getxattr.name, keylen, &xattr);
+ if (ret)
+ goto unlock;
+
+ xattr = gf_strdup(xattr);
+
+ ret = dict_set_dynstrn(local->dict, xattr_cky, xattr_cky_len, xattr);
+ if (ret) {
+ UNLOCK(&frame->lock);
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Cannot set xattr cookie key");
+ goto post_unlock;
}
-unlock:
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (!local->cont.getxattr.xattr_len)
- goto unwind;
-
- nxattr = dict_new ();
- if (!nxattr)
- goto unwind;
-
- /* extra bytes for decorations (brackets and <>'s) */
- padding += strlen (this->name)
- + strlen (AFR_PATHINFO_HEADER) + 4;
- local->cont.getxattr.xattr_len += (padding + 2);
-
- xattr_serz = GF_CALLOC (local->cont.getxattr.xattr_len,
- sizeof (char), gf_common_mt_char);
-
- if (!xattr_serz)
- goto unwind;
-
- /* the xlator info */
- (void) sprintf (xattr_serz, "(<"AFR_PATHINFO_HEADER"%s> ",
- this->name);
-
- /* actual series of pathinfo */
- ret = dict_serialize_value_with_delim (local->dict,
- xattr_serz
- + strlen (xattr_serz),
- &tlen, ' ');
- if (ret) {
- goto unwind;
- }
- /* closing part */
- *(xattr_serz + padding + tlen) = ')';
- *(xattr_serz + padding + tlen + 1) = '\0';
+ local->cont.getxattr.xattr_len += strlen(xattr) + 1;
+ }
+unlock:
+ UNLOCK(&frame->lock);
+post_unlock:
+ if (!callcnt) {
+ if (!local->cont.getxattr.xattr_len)
+ goto unwind;
+
+ nxattr = dict_new();
+ if (!nxattr)
+ goto unwind;
+
+ /* extra bytes for decorations (brackets and <>'s) */
+ padding += strlen(this->name) + SLEN(AFR_PATHINFO_HEADER) + 4;
+ local->cont.getxattr.xattr_len += (padding + 2);
+
+ xattr_serz = GF_MALLOC(local->cont.getxattr.xattr_len,
+ gf_common_mt_char);
+
+ if (!xattr_serz)
+ goto unwind;
+
+ /* the xlator info */
+ int xattr_serz_len = sprintf(
+ xattr_serz, "(<" AFR_PATHINFO_HEADER "%s> ", this->name);
+
+ /* actual series of pathinfo */
+ ret = dict_serialize_value_with_delim(
+ local->dict, xattr_serz + xattr_serz_len, &tlen, ' ');
+ if (ret) {
+ GF_FREE(xattr_serz);
+ goto unwind;
+ }
- ret = dict_set_dynstr (nxattr, local->cont.getxattr.name,
- xattr_serz);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Cannot set pathinfo key in dict");
+ /* closing part */
+ *(xattr_serz + padding + tlen) = ')';
+ *(xattr_serz + padding + tlen + 1) = '\0';
+
+ ret = dict_set_dynstrn(nxattr, local->cont.getxattr.name, keylen,
+ xattr_serz);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Cannot set pathinfo key in dict");
+ if (ret == -EINVAL)
+ GF_FREE(xattr_serz);
+ }
- unwind:
- AFR_STACK_UNWIND (fgetxattr, frame, local->op_ret,
- local->op_errno, nxattr, local->xdata_rsp);
+ unwind:
+ AFR_STACK_UNWIND(fgetxattr, frame, local->op_ret, local->op_errno,
+ nxattr, local->xdata_rsp);
- if (nxattr)
- dict_unref (nxattr);
- }
+ if (nxattr)
+ dict_unref(nxattr);
+ }
out:
- return ret;
+ return ret;
}
int32_t
-afr_getxattr_pathinfo_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_getxattr_pathinfo_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int32_t callcnt = 0;
- int ret = 0;
- char *xattr = NULL;
- char *xattr_serz = NULL;
- char xattr_cky[1024] = {0,};
- dict_t *nxattr = NULL;
- long cky = 0;
- int32_t padding = 0;
- int32_t tlen = 0;
-
- if (!frame || !frame->local || !this) {
- gf_msg ("", GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_ARG, "possible NULL deref");
- goto out;
- }
-
- local = frame->local;
- cky = (long) cookie;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret < 0) {
- local->op_errno = op_errno;
- } else {
- local->op_ret = op_ret;
- if (!local->xdata_rsp && xdata)
- local->xdata_rsp = dict_ref (xdata);
- }
-
- if (!dict || (op_ret < 0))
- goto unlock;
-
- if (!local->dict)
- local->dict = dict_new ();
-
- if (local->dict) {
- ret = dict_get_str (dict,
- local->cont.getxattr.name,
- &xattr);
- if (ret)
- goto unlock;
-
- xattr = gf_strdup (xattr);
-
- (void)snprintf (xattr_cky, 1024, "%s-%ld",
- local->cont.getxattr.name, cky);
- ret = dict_set_dynstr (local->dict,
- xattr_cky, xattr);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret,
- AFR_MSG_DICT_SET_FAILED,
- "Cannot set xattr "
- "cookie key");
- goto unlock;
- }
-
- local->cont.getxattr.xattr_len += strlen (xattr) + 1;
- }
- }
-unlock:
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (!local->cont.getxattr.xattr_len)
- goto unwind;
-
- nxattr = dict_new ();
- if (!nxattr)
- goto unwind;
-
- /* extra bytes for decorations (brackets and <>'s) */
- padding += strlen (this->name) + strlen (AFR_PATHINFO_HEADER) + 4;
- local->cont.getxattr.xattr_len += (padding + 2);
-
- xattr_serz = GF_CALLOC (local->cont.getxattr.xattr_len,
- sizeof (char), gf_common_mt_char);
+ afr_local_t *local = NULL;
+ int32_t callcnt = 0;
+ int ret = 0;
+ char *xattr = NULL;
+ char *xattr_serz = NULL;
+ char xattr_cky[1024] = {
+ 0,
+ };
+ int keylen = 0;
+ int xattr_cky_len = 0;
+ dict_t *nxattr = NULL;
+ long cky = 0;
+ int32_t padding = 0;
+ int32_t tlen = 0;
+
+ if (!frame || !frame->local || !this) {
+ gf_msg("", GF_LOG_ERROR, 0, AFR_MSG_INVALID_ARG, "possible NULL deref");
+ goto out;
+ }
+
+ local = frame->local;
+ cky = (long)cookie;
+ keylen = strlen(local->cont.getxattr.name);
+ xattr_cky_len = snprintf(xattr_cky, sizeof(xattr_cky), "%s-%ld",
+ local->cont.getxattr.name, cky);
+ LOCK(&frame->lock);
+ {
+ callcnt = --local->call_count;
- if (!xattr_serz)
- goto unwind;
+ if (op_ret < 0) {
+ local->op_errno = op_errno;
+ } else {
+ local->op_ret = op_ret;
+ if (!local->xdata_rsp && xdata)
+ local->xdata_rsp = dict_ref(xdata);
+ }
- /* the xlator info */
- (void) sprintf (xattr_serz, "(<"AFR_PATHINFO_HEADER"%s> ",
- this->name);
+ if (!dict || (op_ret < 0))
+ goto unlock;
- /* actual series of pathinfo */
- ret = dict_serialize_value_with_delim (local->dict,
- xattr_serz + strlen (xattr_serz),
- &tlen, ' ');
- if (ret) {
- goto unwind;
- }
+ if (!local->dict) {
+ local->dict = dict_new();
+ if (!local->dict)
+ goto unlock;
+ }
+ ret = dict_get_strn(dict, local->cont.getxattr.name, keylen, &xattr);
+ if (ret)
+ goto unlock;
+
+ xattr = gf_strdup(xattr);
+
+ ret = dict_set_dynstrn(local->dict, xattr_cky, xattr_cky_len, xattr);
+ if (ret) {
+ UNLOCK(&frame->lock);
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Cannot set xattr cookie key");
+ goto post_unlock;
+ }
- /* closing part */
- *(xattr_serz + padding + tlen) = ')';
- *(xattr_serz + padding + tlen + 1) = '\0';
+ local->cont.getxattr.xattr_len += strlen(xattr) + 1;
+ }
+unlock:
+ UNLOCK(&frame->lock);
+post_unlock:
+ if (!callcnt) {
+ if (!local->cont.getxattr.xattr_len)
+ goto unwind;
+
+ nxattr = dict_new();
+ if (!nxattr)
+ goto unwind;
+
+ /* extra bytes for decorations (brackets and <>'s) */
+ padding += strlen(this->name) + SLEN(AFR_PATHINFO_HEADER) + 4;
+ local->cont.getxattr.xattr_len += (padding + 2);
+
+ xattr_serz = GF_MALLOC(local->cont.getxattr.xattr_len,
+ gf_common_mt_char);
+
+ if (!xattr_serz)
+ goto unwind;
+
+ /* the xlator info */
+ int xattr_serz_len = sprintf(
+ xattr_serz, "(<" AFR_PATHINFO_HEADER "%s> ", this->name);
+
+ /* actual series of pathinfo */
+ ret = dict_serialize_value_with_delim(
+ local->dict, xattr_serz + xattr_serz_len, &tlen, ' ');
+ if (ret) {
+ GF_FREE(xattr_serz);
+ goto unwind;
+ }
- ret = dict_set_dynstr (nxattr, local->cont.getxattr.name,
- xattr_serz);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Cannot set pathinfo key in dict");
+ /* closing part */
+ *(xattr_serz + padding + tlen) = ')';
+ *(xattr_serz + padding + tlen + 1) = '\0';
+
+ ret = dict_set_dynstrn(nxattr, local->cont.getxattr.name, keylen,
+ xattr_serz);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Cannot set pathinfo key in dict");
+ if (ret == -EINVAL)
+ GF_FREE(xattr_serz);
+ }
- unwind:
- AFR_STACK_UNWIND (getxattr, frame, local->op_ret,
- local->op_errno, nxattr, local->xdata_rsp);
+ unwind:
+ AFR_STACK_UNWIND(getxattr, frame, local->op_ret, local->op_errno,
+ nxattr, local->xdata_rsp);
- if (nxattr)
- dict_unref (nxattr);
- }
+ if (nxattr)
+ dict_unref(nxattr);
+ }
out:
- return ret;
+ return ret;
}
static int
-afr_aggregate_stime_xattr (dict_t *this, char *key, data_t *value, void *data)
+afr_aggregate_stime_xattr(dict_t *this, char *key, data_t *value, void *data)
{
- int ret = 0;
+ int ret = 0;
- if (fnmatch (GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0)
- ret = gf_get_max_stime (THIS, data, key, value);
+ if (fnmatch(GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0)
+ ret = gf_get_max_stime(THIS, data, key, value);
- return ret;
+ return ret;
}
int32_t
-afr_common_getxattr_stime_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_common_getxattr_stime_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int32_t callcnt = 0;
-
- if (!frame || !frame->local || !this) {
- gf_msg ("", GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_ARG, "possible NULL deref");
- goto out;
- }
+ afr_local_t *local = NULL;
+ int32_t callcnt = 0;
- local = frame->local;
+ if (!frame || !frame->local || !this) {
+ gf_msg("", GF_LOG_ERROR, 0, AFR_MSG_INVALID_ARG, "possible NULL deref");
+ goto out;
+ }
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
+ local = frame->local;
- if (!dict || (op_ret < 0)) {
- local->op_errno = op_errno;
- goto cleanup;
- }
+ LOCK(&frame->lock);
+ {
+ callcnt = --local->call_count;
- if (!local->dict)
- local->dict = dict_copy_with_ref (dict, NULL);
- else
- dict_foreach (dict, afr_aggregate_stime_xattr,
- local->dict);
- local->op_ret = 0;
+ if (!dict || (op_ret < 0)) {
+ local->op_errno = op_errno;
+ goto cleanup;
}
+ if (!local->dict)
+ local->dict = dict_copy_with_ref(dict, NULL);
+ else
+ dict_foreach(dict, afr_aggregate_stime_xattr, local->dict);
+ local->op_ret = 0;
+ }
+
cleanup:
- UNLOCK (&frame->lock);
+ UNLOCK(&frame->lock);
- if (!callcnt) {
- AFR_STACK_UNWIND (getxattr, frame, local->op_ret,
- local->op_errno, local->dict, xdata);
- }
+ if (!callcnt) {
+ AFR_STACK_UNWIND(getxattr, frame, local->op_ret, local->op_errno,
+ local->dict, xdata);
+ }
out:
- return 0;
+ return 0;
}
-
static gf_boolean_t
-afr_is_special_xattr (const char *name, fop_getxattr_cbk_t *cbk,
- gf_boolean_t is_fgetxattr)
+afr_is_special_xattr(const char *name, fop_getxattr_cbk_t *cbk,
+ gf_boolean_t is_fgetxattr)
{
- gf_boolean_t is_spl = _gf_true;
-
- GF_ASSERT (cbk);
- if (!cbk || !name) {
- is_spl = _gf_false;
- goto out;
+ gf_boolean_t is_spl = _gf_true;
+
+ GF_ASSERT(cbk);
+ if (!cbk || !name) {
+ is_spl = _gf_false;
+ goto out;
+ }
+
+ if (!strcmp(name, GF_XATTR_PATHINFO_KEY) ||
+ !strcmp(name, GF_XATTR_USER_PATHINFO_KEY)) {
+ if (is_fgetxattr) {
+ *cbk = afr_fgetxattr_pathinfo_cbk;
+ } else {
+ *cbk = afr_getxattr_pathinfo_cbk;
}
-
- if (!strcmp (name, GF_XATTR_PATHINFO_KEY) ||
- !strcmp (name, GF_XATTR_USER_PATHINFO_KEY)) {
- if (is_fgetxattr) {
- *cbk = afr_fgetxattr_pathinfo_cbk;
- } else {
- *cbk = afr_getxattr_pathinfo_cbk;
- }
- } else if (!strncmp (name, GF_XATTR_CLRLK_CMD,
- strlen (GF_XATTR_CLRLK_CMD))) {
- if (is_fgetxattr) {
- *cbk = afr_fgetxattr_clrlk_cbk;
- } else {
- *cbk = afr_getxattr_clrlk_cbk;
- }
- } else if (!strncmp (name, GF_XATTR_LOCKINFO_KEY,
- strlen (GF_XATTR_LOCKINFO_KEY))) {
- if (is_fgetxattr) {
- *cbk = afr_fgetxattr_lockinfo_cbk;
- } else {
- *cbk = afr_getxattr_lockinfo_cbk;
- }
- } else if (fnmatch (GF_XATTR_STIME_PATTERN, name, FNM_NOESCAPE) == 0) {
- *cbk = afr_common_getxattr_stime_cbk;
- } else if (strcmp (name, QUOTA_SIZE_KEY) == 0) {
- *cbk = afr_getxattr_quota_size_cbk;
- } else if (!strcmp (name, GF_XATTR_LIST_NODE_UUIDS_KEY)) {
- *cbk = afr_getxattr_list_node_uuids_cbk;
+ } else if (!strncmp(name, GF_XATTR_CLRLK_CMD, SLEN(GF_XATTR_CLRLK_CMD))) {
+ if (is_fgetxattr) {
+ *cbk = afr_fgetxattr_clrlk_cbk;
} else {
- is_spl = _gf_false;
+ *cbk = afr_getxattr_clrlk_cbk;
}
+ } else if (!strncmp(name, GF_XATTR_LOCKINFO_KEY,
+ SLEN(GF_XATTR_LOCKINFO_KEY))) {
+ if (is_fgetxattr) {
+ *cbk = afr_fgetxattr_lockinfo_cbk;
+ } else {
+ *cbk = afr_getxattr_lockinfo_cbk;
+ }
+ } else if (fnmatch(GF_XATTR_STIME_PATTERN, name, FNM_NOESCAPE) == 0) {
+ *cbk = afr_common_getxattr_stime_cbk;
+ } else if (strcmp(name, QUOTA_SIZE_KEY) == 0) {
+ *cbk = afr_getxattr_quota_size_cbk;
+ } else if (!strcmp(name, GF_XATTR_LIST_NODE_UUIDS_KEY)) {
+ *cbk = afr_getxattr_list_node_uuids_cbk;
+ } else {
+ is_spl = _gf_false;
+ }
out:
- return is_spl;
+ return is_spl;
}
static void
-afr_getxattr_all_subvols (xlator_t *this, call_frame_t *frame,
- const char *name, loc_t *loc,
- fop_getxattr_cbk_t cbk)
+afr_getxattr_all_subvols(xlator_t *this, call_frame_t *frame, const char *name,
+ loc_t *loc, fop_getxattr_cbk_t cbk)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = 0;
- int call_count = 0;
-
- priv = this->private;
-
- local = frame->local;
- //local->call_count set in afr_local_init
- call_count = local->call_count;
-
- if (!strcmp (name, GF_XATTR_LIST_NODE_UUIDS_KEY)) {
- GF_FREE (local->cont.getxattr.name);
- local->cont.getxattr.name = gf_strdup (GF_XATTR_NODE_UUID_KEY);
- }
-
- //If up-children count is 0, afr_local_init would have failed already
- //and the call would have unwound so not handling it here.
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, cbk,
- (void *) (long) i, priv->children[i],
- priv->children[i]->fops->getxattr,
- loc, local->cont.getxattr.name,
- NULL);
- if (!--call_count)
- break;
- }
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
+ int call_count = 0;
+
+ priv = this->private;
+
+ local = frame->local;
+ // local->call_count set in afr_local_init
+ call_count = local->call_count;
+
+ if (!strcmp(name, GF_XATTR_LIST_NODE_UUIDS_KEY)) {
+ GF_FREE(local->cont.getxattr.name);
+ local->cont.getxattr.name = gf_strdup(GF_XATTR_NODE_UUID_KEY);
+ }
+
+ // If up-children count is 0, afr_local_init would have failed already
+ // and the call would have unwound so not handling it here.
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE(frame, cbk, (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->getxattr, loc,
+ local->cont.getxattr.name, NULL);
+ if (!--call_count)
+ break;
}
- return;
+ }
+ return;
}
int
-afr_marker_populate_args (call_frame_t *frame, int type, int *gauge,
- xlator_t **subvols)
+afr_marker_populate_args(call_frame_t *frame, int type, int *gauge,
+ xlator_t **subvols)
{
- xlator_t *this = frame->this;
- afr_private_t *priv = this->private;
+ xlator_t *this = frame->this;
+ afr_private_t *priv = this->private;
- memcpy (subvols, priv->children, sizeof (*subvols) * priv->child_count);
+ memcpy(subvols, priv->children, sizeof(*subvols) * priv->child_count);
- if (type == MARKER_XTIME_TYPE) {
- /*Don't error out on ENOENT/ENOTCONN */
- gauge[MCNT_NOTFOUND] = 0;
- gauge[MCNT_ENOTCONN] = 0;
- }
- return priv->child_count;
+ if (type == MARKER_XTIME_TYPE) {
+ /*Don't error out on ENOENT/ENOTCONN */
+ gauge[MCNT_NOTFOUND] = 0;
+ gauge[MCNT_ENOTCONN] = 0;
+ }
+ return priv->child_count;
}
static int
-afr_handle_heal_xattrs (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *heal_op)
+afr_handle_heal_xattrs(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *heal_op)
{
- int ret = -1;
- afr_spb_status_t *data = NULL;
+ int ret = -1;
+ afr_spb_status_t *data = NULL;
- if (!strcmp (heal_op, GF_HEAL_INFO)) {
- afr_get_heal_info (frame, this, loc);
- ret = 0;
- goto out;
- }
+ if (!strcmp(heal_op, GF_HEAL_INFO)) {
+ afr_get_heal_info(frame, this, loc);
+ ret = 0;
+ goto out;
+ }
- if (!strcmp (heal_op, GF_AFR_HEAL_SBRAIN)) {
- afr_heal_splitbrain_file (frame, this, loc);
- ret = 0;
- goto out;
+ if (!strcmp(heal_op, GF_AFR_HEAL_SBRAIN)) {
+ afr_heal_splitbrain_file(frame, this, loc);
+ ret = 0;
+ goto out;
+ }
+
+ if (!strcmp(heal_op, GF_AFR_SBRAIN_STATUS)) {
+ data = GF_CALLOC(1, sizeof(*data), gf_afr_mt_spb_status_t);
+ if (!data) {
+ ret = 1;
+ goto out;
}
-
- if (!strcmp (heal_op, GF_AFR_SBRAIN_STATUS)) {
- data = GF_CALLOC (1, sizeof (*data), gf_afr_mt_spb_status_t);
- if (!data) {
- ret = 1;
- goto out;
- }
- data->frame = frame;
- data->loc = loc;
- ret = synctask_new (this->ctx->env,
- afr_get_split_brain_status,
- afr_get_split_brain_status_cbk,
- NULL, data);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN_STATUS,
- "Failed to create"
- " synctask. Unable to fetch split-brain status"
- " for %s.", loc->name);
- ret = 1;
- goto out;
- }
- goto out;
+ data->frame = frame;
+ data->loc = loc;
+ ret = synctask_new(this->ctx->env, afr_get_split_brain_status,
+ afr_get_split_brain_status_cbk, NULL, data);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN_STATUS,
+ "Failed to create"
+ " synctask. Unable to fetch split-brain status"
+ " for %s.",
+ loc->name);
+ ret = 1;
+ goto out;
}
+ goto out;
+ }
out:
- if (ret == 1) {
- AFR_STACK_UNWIND (getxattr, frame, -1, ENOMEM, NULL, NULL);
- if (data)
- GF_FREE (data);
- ret = 0;
- }
- return ret;
+ if (ret == 1) {
+ AFR_STACK_UNWIND(getxattr, frame, -1, ENOMEM, NULL, NULL);
+ if (data)
+ GF_FREE(data);
+ ret = 0;
+ }
+ return ret;
}
int32_t
-afr_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
+afr_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, const char *name,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- xlator_t **children = NULL;
- int i = 0;
- int32_t op_errno = 0;
- int ret = -1;
- fop_getxattr_cbk_t cbk = NULL;
-
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ xlator_t **children = NULL;
+ int i = 0;
+ int32_t op_errno = 0;
+ int ret = -1;
+ fop_getxattr_cbk_t cbk = NULL;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- priv = this->private;
+ priv = this->private;
- children = priv->children;
+ children = priv->children;
- loc_copy (&local->loc, loc);
+ loc_copy(&local->loc, loc);
- local->op = GF_FOP_GETXATTR;
+ local->op = GF_FOP_GETXATTR;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- if (!name)
- goto no_name;
+ if (!name)
+ goto no_name;
- local->cont.getxattr.name = gf_strdup (name);
+ local->cont.getxattr.name = gf_strdup(name);
- if (!local->cont.getxattr.name) {
- op_errno = ENOMEM;
- goto out;
- }
-
- if (!strncmp (name, AFR_XATTR_PREFIX,
- strlen (AFR_XATTR_PREFIX))) {
- op_errno = ENODATA;
- goto out;
- }
+ if (!local->cont.getxattr.name) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- if (cluster_handle_marker_getxattr (frame, loc, name, priv->vol_uuid,
- afr_getxattr_unwind,
- afr_marker_populate_args) == 0)
- return 0;
+ if (!strncmp(name, AFR_XATTR_PREFIX, SLEN(AFR_XATTR_PREFIX))) {
+ op_errno = ENODATA;
+ goto out;
+ }
- ret = afr_handle_heal_xattrs (frame, this, &local->loc, name);
- if (ret == 0)
- return 0;
+ if (cluster_handle_marker_getxattr(frame, loc, name, priv->vol_uuid,
+ afr_getxattr_unwind,
+ afr_marker_populate_args) == 0)
+ return 0;
- /*
- * Heal daemons don't have IO threads ... and as a result they
- * send this getxattr down and eventually crash :(
- */
- op_errno = -1;
- GF_CHECK_XATTR_KEY_AND_GOTO(name, IO_THREADS_QUEUE_SIZE_KEY, op_errno,
- out);
+ ret = afr_handle_heal_xattrs(frame, this, &local->loc, name);
+ if (ret == 0)
+ return 0;
- /*
- * Special xattrs which need responses from all subvols
- */
- if (afr_is_special_xattr (name, &cbk, 0)) {
- afr_getxattr_all_subvols (this, frame, name, loc, cbk);
- return 0;
- }
+ /*
+ * Heal daemons don't have IO threads ... and as a result they
+ * send this getxattr down and eventually crash :(
+ */
+ op_errno = -1;
+ GF_CHECK_XATTR_KEY_AND_GOTO(name, IO_THREADS_QUEUE_SIZE_KEY, op_errno, out);
+
+ /*
+ * Special xattrs which need responses from all subvols
+ */
+ if (afr_is_special_xattr(name, &cbk, 0)) {
+ afr_getxattr_all_subvols(this, frame, name, loc, cbk);
+ return 0;
+ }
- if (XATTR_IS_NODE_UUID (name)) {
- i = 0;
- STACK_WIND_COOKIE (frame, afr_getxattr_node_uuid_cbk,
- (void *) (long) i,
- children[i],
- children[i]->fops->getxattr,
- loc, name, xdata);
- return 0;
- }
+ if (XATTR_IS_NODE_UUID(name)) {
+ i = 0;
+ STACK_WIND_COOKIE(frame, afr_getxattr_node_uuid_cbk, (void *)(long)i,
+ children[i], children[i]->fops->getxattr, loc, name,
+ xdata);
+ return 0;
+ }
no_name:
- afr_read_txn (frame, this, local->loc.inode, afr_getxattr_wind,
- AFR_METADATA_TRANSACTION);
+ afr_read_txn(frame, this, local->loc.inode, afr_getxattr_wind,
+ AFR_METADATA_TRANSACTION);
- ret = 0;
+ ret = 0;
out:
- if (ret < 0)
- AFR_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
- return 0;
+ if (ret < 0)
+ AFR_STACK_UNWIND(getxattr, frame, -1, op_errno, NULL, NULL);
+ return 0;
}
/* {{{ fgetxattr */
-
int32_t
-afr_fgetxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+afr_fgetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = op_errno;
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ afr_read_txn_continue(frame, this, (long)cookie);
+ return 0;
+ }
- if (dict)
- afr_filter_xattrs (dict);
+ if (dict)
+ afr_filter_xattrs(dict);
- AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, dict, xdata);
+ AFR_STACK_UNWIND(fgetxattr, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ return 0;
}
int
-afr_fgetxattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_fgetxattr_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- if (subvol == -1) {
- AFR_STACK_UNWIND (fgetxattr, frame, local->op_ret,
- local->op_errno, NULL, NULL);
- return 0;
- }
-
- STACK_WIND_COOKIE (frame, afr_fgetxattr_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->fgetxattr,
- local->fd, local->cont.getxattr.name,
- local->xdata_req);
- return 0;
-}
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ local = frame->local;
+ priv = this->private;
-static void
-afr_fgetxattr_all_subvols (xlator_t *this, call_frame_t *frame,
- fop_fgetxattr_cbk_t cbk)
-{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = 0;
- int call_count = 0;
+ if (subvol == -1) {
+ AFR_STACK_UNWIND(fgetxattr, frame, local->op_ret, local->op_errno, NULL,
+ NULL);
+ return 0;
+ }
- priv = this->private;
+ STACK_WIND_COOKIE(frame, afr_fgetxattr_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->fgetxattr, local->fd,
+ local->cont.getxattr.name, local->xdata_req);
+ return 0;
+}
- local = frame->local;
- //local->call_count set in afr_local_init
- call_count = local->call_count;
-
- //If up-children count is 0, afr_local_init would have failed already
- //and the call would have unwound so not handling it here.
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->fgetxattr,
- local->fd, local->cont.getxattr.name,
- NULL);
- if (!--call_count)
- break;
- }
+static void
+afr_fgetxattr_all_subvols(xlator_t *this, call_frame_t *frame,
+ fop_fgetxattr_cbk_t cbk)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
+ int call_count = 0;
+
+ priv = this->private;
+
+ local = frame->local;
+ // local->call_count set in afr_local_init
+ call_count = local->call_count;
+
+ // If up-children count is 0, afr_local_init would have failed already
+ // and the call would have unwound so not handling it here.
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE(frame, cbk, (void *)(long)i, priv->children[i],
+ priv->children[i]->fops->fgetxattr, local->fd,
+ local->cont.getxattr.name, NULL);
+ if (!--call_count)
+ break;
}
+ }
- return;
+ return;
}
-
int
-afr_fgetxattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, const char *name, dict_t *xdata)
+afr_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int32_t op_errno = 0;
- fop_fgetxattr_cbk_t cbk = NULL;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- local->op = GF_FOP_FGETXATTR;
- local->fd = fd_ref (fd);
- if (name) {
- local->cont.getxattr.name = gf_strdup (name);
- if (!local->cont.getxattr.name) {
- op_errno = ENOMEM;
- goto out;
- }
- }
- if (xdata)
- local->xdata_req = dict_ref (xdata);
-
- /* pathinfo gets handled only in getxattr(), but we need to handle
- * lockinfo.
- * If we are doing fgetxattr with lockinfo as the key then we
- * collect information from all children.
- */
- if (afr_is_special_xattr (name, &cbk, 1)) {
- afr_fgetxattr_all_subvols (this, frame, cbk);
- return 0;
+ afr_local_t *local = NULL;
+ int32_t op_errno = 0;
+ fop_fgetxattr_cbk_t cbk = NULL;
+
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
+
+ local->op = GF_FOP_FGETXATTR;
+ local->fd = fd_ref(fd);
+ if (name) {
+ local->cont.getxattr.name = gf_strdup(name);
+ if (!local->cont.getxattr.name) {
+ op_errno = ENOMEM;
+ goto out;
}
+ }
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
+
+ /* pathinfo gets handled only in getxattr(), but we need to handle
+ * lockinfo.
+ * If we are doing fgetxattr with lockinfo as the key then we
+ * collect information from all children.
+ */
+ if (afr_is_special_xattr(name, &cbk, 1)) {
+ afr_fgetxattr_all_subvols(this, frame, cbk);
+ return 0;
+ }
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- afr_read_txn (frame, this, fd->inode, afr_fgetxattr_wind,
- AFR_METADATA_TRANSACTION);
+ afr_read_txn(frame, this, fd->inode, afr_fgetxattr_wind,
+ AFR_METADATA_TRANSACTION);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (fgetxattr, frame, -1, op_errno, NULL, NULL);
+ AFR_STACK_UNWIND(fgetxattr, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
-
/* }}} */
/* {{{ readv */
int
-afr_readv_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct iovec *vector, int32_t count, struct iatt *buf,
- struct iobref *iobref, dict_t *xdata)
+afr_readv_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iovec *vector, int32_t count,
+ struct iatt *buf, struct iobref *iobref, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = op_errno;
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ afr_read_txn_continue(frame, this, (long)cookie);
+ return 0;
+ }
- AFR_STACK_UNWIND (readv, frame, op_ret, op_errno,
- vector, count, buf, iobref, xdata);
- return 0;
+ AFR_STACK_UNWIND(readv, frame, op_ret, op_errno, vector, count, buf, iobref,
+ xdata);
+ return 0;
}
-
int
-afr_readv_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_readv_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- if (subvol == -1) {
- AFR_STACK_UNWIND (readv, frame, local->op_ret, local->op_errno,
- 0, 0, 0, 0, 0);
- return 0;
- }
-
- STACK_WIND_COOKIE (frame, afr_readv_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->readv,
- local->fd, local->cont.readv.size,
- local->cont.readv.offset, local->cont.readv.flags,
- local->xdata_req);
- return 0;
-}
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ local = frame->local;
+ priv = this->private;
+
+ if (subvol == -1) {
+ AFR_STACK_UNWIND(readv, frame, local->op_ret, local->op_errno, 0, 0, 0,
+ 0, 0);
+ return 0;
+ }
+
+ STACK_WIND_COOKIE(
+ frame, afr_readv_cbk, (void *)(long)subvol, priv->children[subvol],
+ priv->children[subvol]->fops->readv, local->fd, local->cont.readv.size,
+ local->cont.readv.offset, local->cont.readv.flags, local->xdata_req);
+ return 0;
+}
int
-afr_readv (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, uint32_t flags, dict_t *xdata)
+afr_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- afr_local_t * local = NULL;
- int32_t op_errno = 0;
+ afr_local_t *local = NULL;
+ int32_t op_errno = 0;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_READ;
- local->fd = fd_ref (fd);
- local->cont.readv.size = size;
- local->cont.readv.offset = offset;
- local->cont.readv.flags = flags;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ local->op = GF_FOP_READ;
+ local->fd = fd_ref(fd);
+ local->cont.readv.size = size;
+ local->cont.readv.offset = offset;
+ local->cont.readv.flags = flags;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- afr_read_txn (frame, this, fd->inode, afr_readv_wind,
- AFR_DATA_TRANSACTION);
+ afr_read_txn(frame, this, fd->inode, afr_readv_wind, AFR_DATA_TRANSACTION);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND(readv, frame, -1, op_errno, 0, 0, 0, 0, 0);
+ AFR_STACK_UNWIND(readv, frame, -1, op_errno, 0, 0, 0, 0, 0);
- return 0;
+ return 0;
}
/* }}} */
@@ -1908,77 +1822,73 @@ out:
/* {{{ seek */
int
-afr_seek_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, off_t offset, dict_t *xdata)
+afr_seek_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, off_t offset, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = op_errno;
-
- afr_read_txn_continue (frame, this, (long) cookie);
- return 0;
- }
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
- AFR_STACK_UNWIND (seek, frame, op_ret, op_errno, offset, xdata);
+ afr_read_txn_continue(frame, this, (long)cookie);
return 0;
-}
+ }
+ AFR_STACK_UNWIND(seek, frame, op_ret, op_errno, offset, xdata);
+ return 0;
+}
int
-afr_seek_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_seek_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- if (subvol == -1) {
- AFR_STACK_UNWIND (seek, frame, local->op_ret, local->op_errno,
- 0, NULL);
- return 0;
- }
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_seek_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->seek,
- local->fd, local->cont.seek.offset,
- local->cont.seek.what, local->xdata_req);
+ if (subvol == -1) {
+ AFR_STACK_UNWIND(seek, frame, local->op_ret, local->op_errno, 0, NULL);
return 0;
-}
+ }
+ STACK_WIND_COOKIE(
+ frame, afr_seek_cbk, (void *)(long)subvol, priv->children[subvol],
+ priv->children[subvol]->fops->seek, local->fd, local->cont.seek.offset,
+ local->cont.seek.what, local->xdata_req);
+ return 0;
+}
int
-afr_seek (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- gf_seek_what_t what, dict_t *xdata)
+afr_seek(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ gf_seek_what_t what, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int32_t op_errno = 0;
+ afr_local_t *local = NULL;
+ int32_t op_errno = 0;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->op = GF_FOP_SEEK;
- local->fd = fd_ref (fd);
- local->cont.seek.offset = offset;
- local->cont.seek.what = what;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ local->op = GF_FOP_SEEK;
+ local->fd = fd_ref(fd);
+ local->cont.seek.offset = offset;
+ local->cont.seek.what = what;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- afr_read_txn (frame, this, fd->inode, afr_seek_wind,
- AFR_DATA_TRANSACTION);
+ afr_read_txn(frame, this, fd->inode, afr_seek_wind, AFR_DATA_TRANSACTION);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (seek, frame, -1, op_errno, 0, NULL);
+ AFR_STACK_UNWIND(seek, frame, -1, op_errno, 0, NULL);
- return 0;
+ return 0;
}
/* }}} */
diff --git a/xlators/cluster/afr/src/afr-inode-read.h b/xlators/cluster/afr/src/afr-inode-read.h
index d128134ef2a..8c982bc7e6f 100644
--- a/xlators/cluster/afr/src/afr-inode-read.h
+++ b/xlators/cluster/afr/src/afr-inode-read.h
@@ -12,34 +12,34 @@
#define __INODE_READ_H__
int32_t
-afr_access (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t mask, dict_t *xdata);
+afr_access(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+ dict_t *xdata);
int32_t
-afr_stat (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xdata);
+afr_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata);
int32_t
-afr_fstat (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *xdata);
+afr_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata);
int32_t
-afr_readlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc, size_t size, dict_t *xdata);
+afr_readlink(call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+ dict_t *xdata);
int32_t
-afr_readv (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata);
+afr_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata);
int32_t
-afr_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata);
+afr_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, const char *name,
+ dict_t *xdata);
int32_t
-afr_fgetxattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, const char *name, dict_t *xdata);
-
+afr_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
+ dict_t *xdata);
int
-afr_handle_quota_size (call_frame_t *frame, xlator_t *this);
+afr_seek(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ gf_seek_what_t what, dict_t *xdata);
+int
+afr_handle_quota_size(call_frame_t *frame, xlator_t *this);
#endif /* __INODE_READ_H__ */
diff --git a/xlators/cluster/afr/src/afr-inode-write.c b/xlators/cluster/afr/src/afr-inode-write.c
index 9a9ef2e90a5..1d6e4f3570a 100644
--- a/xlators/cluster/afr/src/afr-inode-write.c
+++ b/xlators/cluster/afr/src/afr-inode-write.c
@@ -8,798 +8,779 @@
cases as published by the Free Software Foundation.
*/
-
-#include <libgen.h>
#include <unistd.h>
-#include <fnmatch.h>
#include <sys/time.h>
#include <stdlib.h>
#include <signal.h>
-#include "glusterfs.h"
+#include <glusterfs/glusterfs.h>
#include "afr.h"
-#include "dict.h"
-#include "xlator.h"
-#include "hashfn.h"
-#include "logging.h"
-#include "list.h"
-#include "call-stub.h"
-#include "defaults.h"
-#include "common-utils.h"
-#include "compat-errno.h"
-#include "compat.h"
+#include <glusterfs/dict.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/defaults.h>
+#include <glusterfs/common-utils.h>
+#include <glusterfs/compat-errno.h>
+#include <glusterfs/compat.h>
#include "protocol-common.h"
-#include "byte-order.h"
+#include <glusterfs/byte-order.h>
#include "afr-transaction.h"
#include "afr-self-heal.h"
#include "afr-messages.h"
static void
-__afr_inode_write_finalize (call_frame_t *frame, xlator_t *this)
-{
- int i = 0;
- int ret = 0;
- int read_subvol = 0;
- struct iatt *stbuf = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_read_subvol_args_t args = {0,};
-
- local = frame->local;
- priv = this->private;
- GF_VALIDATE_OR_GOTO (this->name, local->inode, out);
-
- /*This code needs to stay till DHT sends fops on linked
- * inodes*/
- if (!inode_is_linked (local->inode)) {
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
- if (local->replies[i].op_ret == -1)
- continue;
- if (!gf_uuid_is_null
- (local->replies[i].poststat.ia_gfid)) {
- gf_uuid_copy (args.gfid,
- local->replies[i].poststat.ia_gfid);
- args.ia_type =
- local->replies[i].poststat.ia_type;
- break;
- } else {
- ret = dict_get_bin (local->replies[i].xdata,
- DHT_IATT_IN_XDATA_KEY,
- (void **) &stbuf);
- if (ret)
- continue;
- gf_uuid_copy (args.gfid, stbuf->ia_gfid);
- args.ia_type = stbuf->ia_type;
- break;
- }
- }
+__afr_inode_write_finalize(call_frame_t *frame, xlator_t *this)
+{
+ int i = 0;
+ int ret = 0;
+ int read_subvol = 0;
+ struct iatt *stbuf = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_read_subvol_args_t args = {
+ 0,
+ };
+
+ local = frame->local;
+ priv = this->private;
+ GF_VALIDATE_OR_GOTO(this->name, local->inode, out);
+
+ /*This code needs to stay till DHT sends fops on linked
+ * inodes*/
+ if (!inode_is_linked(local->inode)) {
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+ if (local->replies[i].op_ret == -1)
+ continue;
+ if (!gf_uuid_is_null(local->replies[i].poststat.ia_gfid)) {
+ gf_uuid_copy(args.gfid, local->replies[i].poststat.ia_gfid);
+ args.ia_type = local->replies[i].poststat.ia_type;
+ break;
+ } else {
+ ret = dict_get_bin(local->replies[i].xdata,
+ DHT_IATT_IN_XDATA_KEY, (void **)&stbuf);
+ if (ret)
+ continue;
+ gf_uuid_copy(args.gfid, stbuf->ia_gfid);
+ args.ia_type = stbuf->ia_type;
+ break;
+ }
}
-
- if (local->transaction.type == AFR_METADATA_TRANSACTION) {
- read_subvol = afr_metadata_subvol_get (local->inode, this,
- NULL, local->readable,
- NULL, &args);
- } else {
- read_subvol = afr_data_subvol_get (local->inode, this,
- NULL, local->readable,
- NULL, &args);
+ }
+
+ if (local->transaction.type == AFR_METADATA_TRANSACTION) {
+ read_subvol = afr_metadata_subvol_get(local->inode, this, NULL,
+ local->readable, NULL, &args);
+ } else {
+ read_subvol = afr_data_subvol_get(local->inode, this, NULL,
+ local->readable, NULL, &args);
+ }
+
+ local->op_ret = -1;
+ local->op_errno = afr_final_errno(local, priv);
+ afr_pick_error_xdata(local, priv, local->inode, local->readable, NULL,
+ NULL);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+ if (local->replies[i].op_ret < 0)
+ continue;
+
+ /* Order of checks in the compound conditional
+ below is important.
+
+ - Highest precedence: largest op_ret
+ - Next precedence: if all op_rets are equal, read subvol
+ - Least precedence: any succeeded subvol
+ */
+ if ((local->op_ret < local->replies[i].op_ret) ||
+ ((local->op_ret == local->replies[i].op_ret) &&
+ (i == read_subvol))) {
+ local->op_ret = local->replies[i].op_ret;
+ local->op_errno = local->replies[i].op_errno;
+
+ local->cont.inode_wfop.prebuf = local->replies[i].prestat;
+ local->cont.inode_wfop.postbuf = local->replies[i].poststat;
+
+ if (local->replies[i].xdata) {
+ if (local->xdata_rsp)
+ dict_unref(local->xdata_rsp);
+ local->xdata_rsp = dict_ref(local->replies[i].xdata);
+ }
+ if (local->replies[i].xattr) {
+ if (local->xattr_rsp)
+ dict_unref(local->xattr_rsp);
+ local->xattr_rsp = dict_ref(local->replies[i].xattr);
+ }
}
+ }
- local->op_ret = -1;
- local->op_errno = afr_final_errno (local, priv);
- afr_pick_error_xdata (local, priv, local->inode, local->readable, NULL,
- NULL);
-
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
- if (local->replies[i].op_ret < 0)
- continue;
-
- /* Order of checks in the compound conditional
- below is important.
-
- - Highest precedence: largest op_ret
- - Next precendence: if all op_rets are equal, read subvol
- - Least precedence: any succeeded subvol
- */
- if ((local->op_ret < local->replies[i].op_ret) ||
- ((local->op_ret == local->replies[i].op_ret) &&
- (i == read_subvol))) {
-
- local->op_ret = local->replies[i].op_ret;
- local->op_errno = local->replies[i].op_errno;
-
- local->cont.inode_wfop.prebuf =
- local->replies[i].prestat;
- local->cont.inode_wfop.postbuf =
- local->replies[i].poststat;
-
- if (local->replies[i].xdata) {
- if (local->xdata_rsp)
- dict_unref (local->xdata_rsp);
- local->xdata_rsp =
- dict_ref (local->replies[i].xdata);
- }
- if (local->replies[i].xattr) {
- if (local->xattr_rsp)
- dict_unref (local->xattr_rsp);
- local->xattr_rsp =
- dict_ref (local->replies[i].xattr);
- }
- }
- }
-
- afr_set_in_flight_sb_status (this, frame, local->inode);
+ afr_set_in_flight_sb_status(this, frame, local->inode);
out:
- return;
+ return;
}
-
static void
-__afr_inode_write_fill (call_frame_t *frame, xlator_t *this, int child_index,
- int op_ret, int op_errno,
- struct iatt *prebuf, struct iatt *postbuf,
- dict_t *xattr, dict_t *xdata)
+__afr_inode_write_fill(call_frame_t *frame, xlator_t *this, int child_index,
+ int op_ret, int op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xattr, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- local->replies[child_index].valid = 1;
+ local->replies[child_index].valid = 1;
- if (AFR_IS_ARBITER_BRICK(priv, child_index) && op_ret == 1)
- op_ret = iov_length (local->cont.writev.vector,
- local->cont.writev.count);
+ if (AFR_IS_ARBITER_BRICK(priv, child_index) && op_ret == 1)
+ op_ret = iov_length(local->cont.writev.vector,
+ local->cont.writev.count);
- local->replies[child_index].op_ret = op_ret;
- local->replies[child_index].op_errno = op_errno;
- if (xdata)
- local->replies[child_index].xdata = dict_ref (xdata);
+ local->replies[child_index].op_ret = op_ret;
+ local->replies[child_index].op_errno = op_errno;
+ if (xdata)
+ local->replies[child_index].xdata = dict_ref(xdata);
- if (op_ret >= 0) {
- if (prebuf)
- local->replies[child_index].prestat = *prebuf;
- if (postbuf)
- local->replies[child_index].poststat = *postbuf;
- if (xattr)
- local->replies[child_index].xattr = dict_ref (xattr);
- } else {
- afr_transaction_fop_failed (frame, this, child_index);
- }
+ if (op_ret >= 0) {
+ if (prebuf)
+ local->replies[child_index].prestat = *prebuf;
+ if (postbuf)
+ local->replies[child_index].poststat = *postbuf;
+ if (xattr)
+ local->replies[child_index].xattr = dict_ref(xattr);
+ } else {
+ afr_transaction_fop_failed(frame, this, child_index);
+ }
- return;
+ return;
}
-
static int
-__afr_inode_write_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xattr, dict_t *xdata)
-{
- afr_local_t *local = NULL;
- int child_index = (long) cookie;
- int call_count = -1;
- afr_private_t *priv = NULL;
-
- priv = this->private;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- __afr_inode_write_fill (frame, this, child_index, op_ret,
- op_errno, prebuf, postbuf, xattr,
- xdata);
+__afr_inode_write_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xattr, dict_t *xdata)
+{
+ afr_local_t *local = NULL;
+ int child_index = (long)cookie;
+ int call_count = -1;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ LOCK(&frame->lock);
+ {
+ __afr_inode_write_fill(frame, this, child_index, op_ret, op_errno,
+ prebuf, postbuf, xattr, xdata);
+ call_count = --local->call_count;
+ }
+ UNLOCK(&frame->lock);
+
+ if (call_count == 0) {
+ __afr_inode_write_finalize(frame, this);
+
+ if (afr_txn_nothing_failed(frame, this)) {
+ /*if it did pre-op, it will do post-op changing ctime*/
+ if (priv->consistent_metadata && afr_needs_changelog_update(local))
+ afr_zero_fill_stat(local);
+ local->transaction.unwind(frame, this);
}
- UNLOCK (&frame->lock);
-
- call_count = afr_frame_return (frame);
-
- if (call_count == 0) {
- __afr_inode_write_finalize (frame, this);
- if (afr_txn_nothing_failed (frame, this)) {
- /*if it did pre-op, it will do post-op changing ctime*/
- if (priv->consistent_metadata &&
- afr_needs_changelog_update (local))
- afr_zero_fill_stat (local);
- local->transaction.unwind (frame, this);
- }
+ afr_transaction_resume(frame, this);
+ }
- afr_transaction_resume (frame, this);
- }
-
- return 0;
+ return 0;
}
/* {{{ writev */
void
-afr_writev_copy_outvars (call_frame_t *src_frame, call_frame_t *dst_frame)
+afr_writev_copy_outvars(call_frame_t *src_frame, call_frame_t *dst_frame)
{
- afr_local_t *src_local = NULL;
- afr_local_t *dst_local = NULL;
+ afr_local_t *src_local = NULL;
+ afr_local_t *dst_local = NULL;
- src_local = src_frame->local;
- dst_local = dst_frame->local;
+ src_local = src_frame->local;
+ dst_local = dst_frame->local;
- dst_local->op_ret = src_local->op_ret;
- dst_local->op_errno = src_local->op_errno;
- dst_local->cont.inode_wfop.prebuf = src_local->cont.inode_wfop.prebuf;
- dst_local->cont.inode_wfop.postbuf = src_local->cont.inode_wfop.postbuf;
- if (src_local->xdata_rsp)
- dst_local->xdata_rsp = dict_ref (src_local->xdata_rsp);
+ dst_local->op_ret = src_local->op_ret;
+ dst_local->op_errno = src_local->op_errno;
+ dst_local->cont.inode_wfop.prebuf = src_local->cont.inode_wfop.prebuf;
+ dst_local->cont.inode_wfop.postbuf = src_local->cont.inode_wfop.postbuf;
+ if (src_local->xdata_rsp)
+ dst_local->xdata_rsp = dict_ref(src_local->xdata_rsp);
}
void
-afr_writev_unwind (call_frame_t *frame, xlator_t *this)
+afr_writev_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- afr_private_t *priv = this->private;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = this->private;
- local = frame->local;
+ local = frame->local;
- if (priv->consistent_metadata)
- afr_zero_fill_stat (local);
+ if (priv->consistent_metadata)
+ afr_zero_fill_stat(local);
- AFR_STACK_UNWIND (writev, frame,
- local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf,
- local->xdata_rsp);
+ AFR_STACK_UNWIND(writev, frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
}
-
int
-afr_transaction_writev_unwind (call_frame_t *frame, xlator_t *this)
+afr_transaction_writev_unwind(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *fop_frame = NULL;
+ call_frame_t *fop_frame = NULL;
- fop_frame = afr_transaction_detach_fop_frame (frame);
+ fop_frame = afr_transaction_detach_fop_frame(frame);
- if (fop_frame) {
- afr_writev_copy_outvars (frame, fop_frame);
- afr_writev_unwind (fop_frame, this);
- }
- return 0;
+ if (fop_frame) {
+ afr_writev_copy_outvars(frame, fop_frame);
+ afr_writev_unwind(fop_frame, this);
+ }
+ return 0;
}
static void
-afr_writev_handle_short_writes (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
-
- local = frame->local;
- priv = this->private;
- /*
- * We already have the best case result of the writev calls staged
- * as the return value. Any writev that returns some value less
- * than the best case is now out of sync, so mark the fop as
- * failed. Note that fops that have returned with errors have
- * already been marked as failed.
- */
- for (i = 0; i < priv->child_count; i++) {
- if ((!local->replies[i].valid) ||
- (local->replies[i].op_ret == -1))
- continue;
-
- if (local->replies[i].op_ret < local->op_ret)
- afr_transaction_fop_failed (frame, this, i);
- }
+afr_writev_handle_short_writes(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+ /*
+ * We already have the best case result of the writev calls staged
+ * as the return value. Any writev that returns some value less
+ * than the best case is now out of sync, so mark the fop as
+ * failed. Note that fops that have returned with errors have
+ * already been marked as failed.
+ */
+ for (i = 0; i < priv->child_count; i++) {
+ if ((!local->replies[i].valid) || (local->replies[i].op_ret == -1))
+ continue;
+
+ if (local->replies[i].op_ret < local->op_ret)
+ afr_transaction_fop_failed(frame, this, i);
+ }
}
void
-afr_inode_write_fill (call_frame_t *frame, xlator_t *this, int child_index,
+afr_inode_write_fill(call_frame_t *frame, xlator_t *this, int child_index,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
struct iatt *postbuf, dict_t *xdata)
{
- int ret = 0;
- afr_local_t *local = frame->local;
- uint32_t open_fd_count = 0;
- uint32_t write_is_append = 0;
-
- LOCK (&frame->lock);
- {
- __afr_inode_write_fill (frame, this, child_index, op_ret,
- op_errno, prebuf, postbuf, NULL, xdata);
- if (op_ret == -1 || !xdata)
- goto unlock;
-
- write_is_append = 0;
- ret = dict_get_uint32 (xdata, GLUSTERFS_WRITE_IS_APPEND,
- &write_is_append);
- if (ret || !write_is_append)
- local->append_write = _gf_false;
-
- ret = dict_get_uint32 (xdata, GLUSTERFS_ACTIVE_FD_COUNT,
- &open_fd_count);
- if (ret < 0)
- goto unlock;
- if (open_fd_count > local->open_fd_count) {
- local->open_fd_count = open_fd_count;
- local->update_open_fd_count = _gf_true;
- }
+ int ret = 0;
+ afr_local_t *local = frame->local;
+ uint32_t open_fd_count = 0;
+ uint32_t write_is_append = 0;
+ int32_t num_inodelks = 0;
+
+ LOCK(&frame->lock);
+ {
+ __afr_inode_write_fill(frame, this, child_index, op_ret, op_errno,
+ prebuf, postbuf, NULL, xdata);
+ if (op_ret == -1 || !xdata)
+ goto unlock;
+
+ write_is_append = 0;
+ ret = dict_get_uint32(xdata, GLUSTERFS_WRITE_IS_APPEND,
+ &write_is_append);
+ if (ret || !write_is_append)
+ local->append_write = _gf_false;
+
+ ret = dict_get_uint32(xdata, GLUSTERFS_ACTIVE_FD_COUNT, &open_fd_count);
+ if (ret < 0)
+ goto unlock;
+ if (open_fd_count > local->open_fd_count) {
+ local->open_fd_count = open_fd_count;
+ local->update_open_fd_count = _gf_true;
}
+
+ ret = dict_get_int32_sizen(xdata, GLUSTERFS_INODELK_COUNT,
+ &num_inodelks);
+ if (ret < 0)
+ goto unlock;
+ if (num_inodelks > local->num_inodelks) {
+ local->num_inodelks = num_inodelks;
+ local->update_num_inodelks = _gf_true;
+ }
+ }
unlock:
- UNLOCK (&frame->lock);
+ UNLOCK(&frame->lock);
}
void
-afr_process_post_writev (call_frame_t *frame, xlator_t *this)
+afr_process_post_writev(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
+ afr_lock_t *lock = NULL;
- local = frame->local;
+ local = frame->local;
- if (!local->stable_write && !local->append_write)
- /* An appended write removes the necessity to
- fsync() the file. This is because self-heal
- has the logic to check for larger file when
- the xattrs are not reliably pointing at
- a stale file.
- */
- afr_fd_report_unstable_write (this, local);
+ if (!local->stable_write && !local->append_write)
+ /* An appended write removes the necessity to
+ fsync() the file. This is because self-heal
+ has the logic to check for larger file when
+ the xattrs are not reliably pointing at
+ a stale file.
+ */
+ afr_fd_report_unstable_write(this, local);
- __afr_inode_write_finalize (frame, this);
+ __afr_inode_write_finalize(frame, this);
- afr_writev_handle_short_writes (frame, this);
-
- if (local->update_open_fd_count)
- local->inode_ctx->open_fd_count = local->open_fd_count;
+ afr_writev_handle_short_writes(frame, this);
+ if (local->update_open_fd_count)
+ local->inode_ctx->open_fd_count = local->open_fd_count;
+ if (local->update_num_inodelks &&
+ local->transaction.type == AFR_DATA_TRANSACTION) {
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ lock->num_inodelks = local->num_inodelks;
+ }
}
int
-afr_writev_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+afr_writev_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- call_frame_t *fop_frame = NULL;
- int child_index = (long) cookie;
- int call_count = -1;
-
- afr_inode_write_fill (frame, this, child_index, op_ret, op_errno,
- prebuf, postbuf, xdata);
-
- call_count = afr_frame_return (frame);
-
- if (call_count == 0) {
- afr_process_post_writev (frame, this);
-
- if (!afr_txn_nothing_failed (frame, this)) {
- //Don't unwind until post-op is complete
- afr_transaction_resume (frame, this);
- } else {
- /*
- * Generally inode-write fops do transaction.unwind then
- * transaction.resume, but writev needs to make sure that
- * delayed post-op frame is placed in fdctx before unwind
- * happens. This prevents the race of flush doing the
- * changelog wakeup first in fuse thread and then this
- * writev placing its delayed post-op frame in fdctx.
- * This helps flush make sure all the delayed post-ops are
- * completed.
- */
-
- fop_frame = afr_transaction_detach_fop_frame (frame);
- afr_writev_copy_outvars (frame, fop_frame);
- afr_transaction_resume (frame, this);
- afr_writev_unwind (fop_frame, this);
- }
+ call_frame_t *fop_frame = NULL;
+ int child_index = (long)cookie;
+ int call_count = -1;
+
+ afr_inode_write_fill(frame, this, child_index, op_ret, op_errno, prebuf,
+ postbuf, xdata);
+
+ call_count = afr_frame_return(frame);
+
+ if (call_count == 0) {
+ afr_process_post_writev(frame, this);
+
+ if (!afr_txn_nothing_failed(frame, this)) {
+ // Don't unwind until post-op is complete
+ afr_transaction_resume(frame, this);
+ } else {
+ /*
+ * Generally inode-write fops do transaction.unwind then
+ * transaction.resume, but writev needs to make sure that
+ * delayed post-op frame is placed in fdctx before unwind
+ * happens. This prevents the race of flush doing the
+ * changelog wakeup first in fuse thread and then this
+ * writev placing its delayed post-op frame in fdctx.
+ * This helps flush make sure all the delayed post-ops are
+ * completed.
+ */
+
+ fop_frame = afr_transaction_detach_fop_frame(frame);
+ afr_writev_copy_outvars(frame, fop_frame);
+ afr_transaction_resume(frame, this);
+ afr_writev_unwind(fop_frame, this);
}
- return 0;
+ }
+ return 0;
}
static int
-afr_arbiter_writev_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_arbiter_writev_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = frame->local;
- afr_private_t *priv = this->private;
- static char byte = 0xFF;
- static struct iovec vector = {&byte, 1};
- int32_t count = 1;
+ afr_local_t *local = frame->local;
+ afr_private_t *priv = this->private;
+ static char byte = 0xFF;
+ static struct iovec vector = {&byte, 1};
+ int32_t count = 1;
- STACK_WIND_COOKIE (frame, afr_writev_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->writev,
- local->fd, &vector, count, local->cont.writev.offset,
- local->cont.writev.flags, local->cont.writev.iobref,
- local->xdata_req);
+ STACK_WIND_COOKIE(
+ frame, afr_writev_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol], priv->children[subvol]->fops->writev, local->fd,
+ &vector, count, local->cont.writev.offset, local->cont.writev.flags,
+ local->cont.writev.iobref, local->xdata_req);
- return 0;
+ return 0;
}
int
-afr_writev_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_writev_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- if (AFR_IS_ARBITER_BRICK(priv, subvol)) {
- afr_arbiter_writev_wind (frame, this, subvol);
- return 0;
- }
-
- STACK_WIND_COOKIE (frame, afr_writev_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->writev,
- local->fd, local->cont.writev.vector,
- local->cont.writev.count, local->cont.writev.offset,
- local->cont.writev.flags, local->cont.writev.iobref,
- local->xdata_req);
+ if (AFR_IS_ARBITER_BRICK(priv, subvol)) {
+ afr_arbiter_writev_wind(frame, this, subvol);
return 0;
-}
+ }
+ STACK_WIND_COOKIE(frame, afr_writev_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->writev, local->fd,
+ local->cont.writev.vector, local->cont.writev.count,
+ local->cont.writev.offset, local->cont.writev.flags,
+ local->cont.writev.iobref, local->xdata_req);
+ return 0;
+}
int
-afr_do_writev (call_frame_t *frame, xlator_t *this)
+afr_do_writev(call_frame_t *frame, xlator_t *this)
{
- call_frame_t *transaction_frame = NULL;
- afr_local_t *local = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
-
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ call_frame_t *transaction_frame = NULL;
+ afr_local_t *local = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- local = frame->local;
- transaction_frame->local = local;
- frame->local = NULL;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- if (!AFR_FRAME_INIT (frame, op_errno))
- goto out;
+ local = frame->local;
+ transaction_frame->local = local;
+ frame->local = NULL;
- local->op = GF_FOP_WRITE;
+ if (!AFR_FRAME_INIT(frame, op_errno))
+ goto out;
- local->transaction.wind = afr_writev_wind;
- local->transaction.unwind = afr_transaction_writev_unwind;
+ local->op = GF_FOP_WRITE;
- local->transaction.main_frame = frame;
+ local->transaction.wind = afr_writev_wind;
+ local->transaction.unwind = afr_transaction_writev_unwind;
- if (local->fd->flags & O_APPEND) {
- /*
- * Backend vfs ignores the 'offset' for append mode fd so
- * locking just the region provided for the writev does not
- * give consistency guarantee. The actual write may happen at a
- * completely different range than the one provided by the
- * offset, len in the fop. So lock the entire file.
- */
- local->transaction.start = 0;
- local->transaction.len = 0;
- } else {
- local->transaction.start = local->cont.writev.offset;
- local->transaction.len = iov_length (local->cont.writev.vector,
- local->cont.writev.count);
- }
-
- ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ local->transaction.main_frame = frame;
- return 0;
+ if (local->fd->flags & O_APPEND) {
+ /*
+ * Backend vfs ignores the 'offset' for append mode fd so
+ * locking just the region provided for the writev does not
+ * give consistency guarantee. The actual write may happen at a
+ * completely different range than the one provided by the
+ * offset, len in the fop. So lock the entire file.
+ */
+ local->transaction.start = 0;
+ local->transaction.len = 0;
+ } else {
+ local->transaction.start = local->cont.writev.offset;
+ local->transaction.len = iov_length(local->cont.writev.vector,
+ local->cont.writev.count);
+ }
+
+ ret = afr_transaction(transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
+
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(writev, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
-
int
-afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count, off_t offset,
- uint32_t flags, struct iobref *iobref, dict_t *xdata)
+afr_writev(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
+ int32_t count, off_t offset, uint32_t flags, struct iobref *iobref,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int op_errno = ENOMEM;
- int ret = -1;
+ afr_local_t *local = NULL;
+ int op_errno = ENOMEM;
+ int ret = -1;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.writev.vector = iov_dup (vector, count);
- if (!local->cont.writev.vector)
- goto out;
- local->cont.writev.count = count;
- local->cont.writev.offset = offset;
- local->cont.writev.flags = flags;
- local->cont.writev.iobref = iobref_ref (iobref);
+ local->cont.writev.vector = iov_dup(vector, count);
+ if (!local->cont.writev.vector)
+ goto out;
+ local->cont.writev.count = count;
+ local->cont.writev.offset = offset;
+ local->cont.writev.flags = flags;
+ local->cont.writev.iobref = iobref_ref(iobref);
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- if (dict_set_uint32 (local->xdata_req, GLUSTERFS_ACTIVE_FD_COUNT, 4)) {
- op_errno = ENOMEM;
- goto out;
- }
+ if (dict_set_uint32(local->xdata_req, GLUSTERFS_ACTIVE_FD_COUNT, 4)) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ if (dict_set_str_sizen(local->xdata_req, GLUSTERFS_INODELK_DOM_COUNT,
+ this->name)) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- if (dict_set_uint32 (local->xdata_req, GLUSTERFS_WRITE_IS_APPEND, 4)) {
- op_errno = ENOMEM;
- goto out;
- }
+ if (dict_set_uint32(local->xdata_req, GLUSTERFS_WRITE_IS_APPEND, 4)) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- /* Set append_write to be true speculatively. If on any
- server it turns not be true, we unset it in the
- callback.
- */
- local->append_write = _gf_true;
+ /* Set append_write to be true speculatively. If on any
+ server it turns not be true, we unset it in the
+ callback.
+ */
+ local->append_write = _gf_true;
- /* detect here, but set it in writev_wind_cbk *after* the unstable
- write is performed
- */
- local->stable_write = !!((fd->flags|flags)&(O_SYNC|O_DSYNC));
+ /* detect here, but set it in writev_wind_cbk *after* the unstable
+ write is performed
+ */
+ local->stable_write = !!((fd->flags | flags) & (O_SYNC | O_DSYNC));
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- afr_do_writev (frame, this);
+ afr_do_writev(frame, this);
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
+ AFR_STACK_UNWIND(writev, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
/* }}} */
/* {{{ truncate */
int
-afr_truncate_unwind (call_frame_t *frame, xlator_t *this)
+afr_truncate_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (truncate, main_frame, local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(truncate, main_frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ return 0;
+}
int
-afr_truncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+afr_truncate_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret == 0 && prebuf->ia_size != postbuf->ia_size)
- local->stable_write = _gf_false;
+ if (op_ret == 0 && prebuf->ia_size != postbuf->ia_size)
+ local->stable_write = _gf_false;
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- prebuf, postbuf, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, prebuf,
+ postbuf, NULL, xdata);
}
-
int
-afr_truncate_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_truncate_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_truncate_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->truncate,
- &local->loc, local->cont.truncate.offset,
- local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_truncate_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->truncate, &local->loc,
+ local->cont.truncate.offset, local->xdata_req);
+ return 0;
}
-
int
-afr_truncate (call_frame_t *frame, xlator_t *this,
- loc_t *loc, off_t offset, dict_t *xdata)
+afr_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- afr_local_t * local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.truncate.offset = offset;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ local->cont.truncate.offset = offset;
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->transaction.wind = afr_truncate_wind;
- local->transaction.unwind = afr_truncate_unwind;
+ local->transaction.wind = afr_truncate_wind;
+ local->transaction.unwind = afr_truncate_unwind;
- loc_copy (&local->loc, loc);
- ret = afr_set_inode_local (this, local, loc->inode);
- if (ret)
- goto out;
+ loc_copy(&local->loc, loc);
+ ret = afr_set_inode_local(this, local, loc->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_TRUNCATE;
+ local->op = GF_FOP_TRUNCATE;
- local->transaction.main_frame = frame;
- local->transaction.start = offset;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = offset;
+ local->transaction.len = 0;
- /* Set it true speculatively, will get reset in afr_truncate_wind_cbk
- if truncate was not a NOP */
- local->stable_write = _gf_true;
+ /* Set it true speculatively, will get reset in afr_truncate_wind_cbk
+ if truncate was not a NOP */
+ local->stable_write = _gf_true;
- ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(truncate, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
-
/* }}} */
/* {{{ ftruncate */
-
int
-afr_ftruncate_unwind (call_frame_t *frame, xlator_t *this)
+afr_ftruncate_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (ftruncate, main_frame, local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(ftruncate, main_frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ return 0;
+}
int
-afr_ftruncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+afr_ftruncate_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret == 0 && prebuf->ia_size != postbuf->ia_size)
- local->stable_write = _gf_false;
+ if (op_ret == 0 && prebuf->ia_size != postbuf->ia_size)
+ local->stable_write = _gf_false;
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- prebuf, postbuf, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, prebuf,
+ postbuf, NULL, xdata);
}
-
int
-afr_ftruncate_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_ftruncate_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_ftruncate_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->ftruncate,
- local->fd, local->cont.ftruncate.offset,
- local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_ftruncate_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->ftruncate, local->fd,
+ local->cont.ftruncate.offset, local->xdata_req);
+ return 0;
}
-
int
-afr_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- dict_t *xdata)
+afr_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.ftruncate.offset = offset;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ local->cont.ftruncate.offset = offset;
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_FTRUNCATE;
+ local->op = GF_FOP_FTRUNCATE;
- local->transaction.wind = afr_ftruncate_wind;
- local->transaction.unwind = afr_ftruncate_unwind;
+ local->transaction.wind = afr_ftruncate_wind;
+ local->transaction.unwind = afr_ftruncate_unwind;
- local->transaction.main_frame = frame;
+ local->transaction.main_frame = frame;
- local->transaction.start = local->cont.ftruncate.offset;
- local->transaction.len = 0;
+ local->transaction.start = local->cont.ftruncate.offset;
+ local->transaction.len = 0;
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- /* Set it true speculatively, will get reset in afr_ftruncate_wind_cbk
- if truncate was not a NOP */
- local->stable_write = _gf_true;
+ /* Set it true speculatively, will get reset in afr_ftruncate_wind_cbk
+ if truncate was not a NOP */
+ local->stable_write = _gf_true;
- ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- AFR_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
+ AFR_STACK_UNWIND(ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
/* }}} */
@@ -807,1813 +788,1778 @@ out:
/* {{{ setattr */
int
-afr_setattr_unwind (call_frame_t *frame, xlator_t *this)
+afr_setattr_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (setattr, main_frame, local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf,
- local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(setattr, main_frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ return 0;
+}
int
-afr_setattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- struct iatt *preop, struct iatt *postop, dict_t *xdata)
+afr_setattr_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *preop,
+ struct iatt *postop, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- preop, postop, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, preop,
+ postop, NULL, xdata);
}
-
int
-afr_setattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_setattr_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_setattr_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->setattr,
- &local->loc, &local->cont.setattr.in_buf,
- local->cont.setattr.valid, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_setattr_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->setattr, &local->loc,
+ &local->cont.setattr.in_buf, local->cont.setattr.valid,
+ local->xdata_req);
+ return 0;
}
-
int
-afr_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *buf,
- int32_t valid, dict_t *xdata)
+afr_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *buf,
+ int32_t valid, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.setattr.in_buf = *buf;
- local->cont.setattr.valid = valid;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ local->cont.setattr.in_buf = *buf;
+ local->cont.setattr.valid = valid;
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->transaction.wind = afr_setattr_wind;
- local->transaction.unwind = afr_setattr_unwind;
+ local->transaction.wind = afr_setattr_wind;
+ local->transaction.unwind = afr_setattr_unwind;
- loc_copy (&local->loc, loc);
- ret = afr_set_inode_local (this, local, loc->inode);
- if (ret)
- goto out;
+ loc_copy(&local->loc, loc);
+ ret = afr_set_inode_local(this, local, loc->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_SETATTR;
+ local->op = GF_FOP_SETATTR;
- local->transaction.main_frame = frame;
- local->transaction.start = LLONG_MAX - 1;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
- ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(setattr, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
/* {{{ fsetattr */
int
-afr_fsetattr_unwind (call_frame_t *frame, xlator_t *this)
+afr_fsetattr_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (fsetattr, main_frame, local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(fsetattr, main_frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ return 0;
+}
int
-afr_fsetattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop, dict_t *xdata)
+afr_fsetattr_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preop,
+ struct iatt *postop, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- preop, postop, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, preop,
+ postop, NULL, xdata);
}
-
int
-afr_fsetattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_fsetattr_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_fsetattr_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->fsetattr,
- local->fd, &local->cont.fsetattr.in_buf,
- local->cont.fsetattr.valid, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_fsetattr_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->fsetattr, local->fd,
+ &local->cont.fsetattr.in_buf, local->cont.fsetattr.valid,
+ local->xdata_req);
+ return 0;
}
-
int
-afr_fsetattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, struct iatt *buf, int32_t valid, dict_t *xdata)
+afr_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *buf,
+ int32_t valid, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.fsetattr.in_buf = *buf;
- local->cont.fsetattr.valid = valid;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ local->cont.fsetattr.in_buf = *buf;
+ local->cont.fsetattr.valid = valid;
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->transaction.wind = afr_fsetattr_wind;
- local->transaction.unwind = afr_fsetattr_unwind;
+ local->transaction.wind = afr_fsetattr_wind;
+ local->transaction.unwind = afr_fsetattr_unwind;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_FSETATTR;
+ local->op = GF_FOP_FSETATTR;
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- local->transaction.main_frame = frame;
- local->transaction.start = LLONG_MAX - 1;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
- ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
-
/* {{{ setxattr */
-
int
-afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
+afr_setxattr_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (setxattr, main_frame, local->op_ret, local->op_errno,
- local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(setxattr, main_frame, local->op_ret, local->op_errno,
+ local->xdata_rsp);
+ return 0;
+}
int
-afr_setxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_setxattr_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- NULL, NULL, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, NULL,
+ NULL, NULL, xdata);
}
-
int
-afr_setxattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_setxattr_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_setxattr_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->setxattr,
- &local->loc, local->cont.setxattr.dict,
- local->cont.setxattr.flags, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_setxattr_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->setxattr, &local->loc,
+ local->cont.setxattr.dict, local->cont.setxattr.flags,
+ local->xdata_req);
+ return 0;
}
int
-afr_emptyb_set_pending_changelog_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- dict_t *xattr, dict_t *xdata)
+afr_emptyb_set_pending_changelog_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int op_ret, int op_errno,
+ dict_t *xattr, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i, ret = 0;
- char *op_type = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i, ret = 0;
+ char *op_type = NULL;
- local = frame->local;
- priv = this->private;
- i = (long) cookie;
+ local = frame->local;
+ priv = this->private;
+ i = (long)cookie;
- local->replies[i].valid = 1;
- local->replies[i].op_ret = op_ret;
- local->replies[i].op_errno = op_errno;
+ local->replies[i].valid = 1;
+ local->replies[i].op_ret = op_ret;
+ local->replies[i].op_errno = op_errno;
- ret = dict_get_str (local->xdata_req, "replicate-brick-op", &op_type);
- if (ret)
- goto out;
+ ret = dict_get_str_sizen(local->xdata_req, "replicate-brick-op", &op_type);
+ if (ret)
+ goto out;
- gf_msg (this->name, op_ret ? GF_LOG_ERROR : GF_LOG_INFO,
- op_ret ? op_errno : 0,
- afr_get_msg_id (op_type),
- "Set of pending xattr %s on"
- " %s.", op_ret ? "failed" : "succeeded",
- priv->children[i]->name);
+ gf_smsg(this->name, op_ret ? GF_LOG_ERROR : GF_LOG_INFO,
+ op_ret ? op_errno : 0, AFR_MSG_SET_PEND_XATTR, "name=%s",
+ priv->children[i]->name, "op_ret=%s",
+ op_ret ? "failed" : "succeeded", NULL);
out:
- syncbarrier_wake (&local->barrier);
- return 0;
+ syncbarrier_wake(&local->barrier);
+ return 0;
}
int
-afr_emptyb_set_pending_changelog (call_frame_t *frame, xlator_t *this,
- unsigned char *locked_nodes)
+afr_emptyb_set_pending_changelog(call_frame_t *frame, xlator_t *this,
+ unsigned char *locked_nodes)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int ret = 0, i = 0;
-
- local = frame->local;
- priv = this->private;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int ret = 0, i = 0;
- AFR_ONLIST (locked_nodes, frame, afr_emptyb_set_pending_changelog_cbk,
- xattrop, &local->loc, GF_XATTROP_ADD_ARRAY,
- local->xattr_req, NULL);
-
- /* It is sufficient if xattrop was successful on one child */
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
-
- if (local->replies[i].op_ret == 0) {
- ret = 0;
- goto out;
- } else {
- ret = afr_higher_errno (ret,
- local->replies[i].op_errno);
- }
- }
-out:
- return -ret;
-}
+ local = frame->local;
+ priv = this->private;
-int
-_afr_handle_empty_brick_type (xlator_t *this, call_frame_t *frame,
- loc_t *loc, int empty_index,
- afr_transaction_type type,
- char *op_type)
-{
- int count = 0;
- int ret = -ENOMEM;
- int idx = -1;
- int d_idx = -1;
- unsigned char *locked_nodes = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ AFR_ONLIST(locked_nodes, frame, afr_emptyb_set_pending_changelog_cbk,
+ xattrop, &local->loc, GF_XATTROP_ADD_ARRAY, local->xattr_req,
+ NULL);
- priv = this->private;
- local = frame->local;
+ /* It is sufficient if xattrop was successful on one child */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
- locked_nodes = alloca0 (priv->child_count);
-
- idx = afr_index_for_transaction_type (type);
- d_idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
-
- local->pending = afr_matrix_create (priv->child_count,
- AFR_NUM_CHANGE_LOGS);
- if (!local->pending)
- goto out;
-
- local->pending[empty_index][idx] = hton32 (1);
-
- if ((priv->esh_granular) && (type == AFR_ENTRY_TRANSACTION))
- local->pending[empty_index][d_idx] = hton32 (1);
-
- local->xdata_req = dict_new ();
- if (!local->xdata_req)
- goto out;
-
- ret = dict_set_str (local->xdata_req, "replicate-brick-op", op_type);
- if (ret)
- goto out;
-
- local->xattr_req = dict_new ();
- if (!local->xattr_req)
- goto out;
-
- ret = afr_set_pending_dict (priv, local->xattr_req, local->pending);
- if (ret < 0)
- goto out;
-
- if (AFR_ENTRY_TRANSACTION == type) {
- count = afr_selfheal_entrylk (frame, this, loc->inode,
- this->name, NULL, locked_nodes);
+ if (local->replies[i].op_ret == 0) {
+ ret = 0;
+ goto out;
} else {
- count = afr_selfheal_inodelk (frame, this, loc->inode,
- this->name, LLONG_MAX - 1, 0,
- locked_nodes);
- }
-
- if (!count) {
- gf_msg (this->name, GF_LOG_ERROR, EAGAIN,
- AFR_MSG_REPLACE_BRICK_STATUS, "Couldn't acquire lock on"
- " any child.");
- ret = -EAGAIN;
- goto unlock;
+ ret = afr_higher_errno(ret, local->replies[i].op_errno);
}
+ }
+out:
+ return -ret;
+}
- ret = afr_emptyb_set_pending_changelog (frame, this, locked_nodes);
- if (ret)
- goto unlock;
- ret = 0;
+static int
+_afr_handle_empty_brick_type(xlator_t *this, call_frame_t *frame, loc_t *loc,
+ int empty_index, afr_transaction_type type,
+ char *op_type, const int op_type_len)
+{
+ int count = 0;
+ int ret = -ENOMEM;
+ int idx = -1;
+ int d_idx = -1;
+ unsigned char *locked_nodes = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ locked_nodes = alloca0(priv->child_count);
+
+ idx = afr_index_for_transaction_type(type);
+ d_idx = afr_index_for_transaction_type(AFR_DATA_TRANSACTION);
+
+ local->pending = afr_matrix_create(priv->child_count, AFR_NUM_CHANGE_LOGS);
+ if (!local->pending)
+ goto out;
+
+ local->pending[empty_index][idx] = hton32(1);
+
+ if ((priv->esh_granular) && (type == AFR_ENTRY_TRANSACTION))
+ local->pending[empty_index][d_idx] = hton32(1);
+
+ local->xdata_req = dict_new();
+ if (!local->xdata_req)
+ goto out;
+
+ ret = dict_set_nstrn(local->xdata_req, "replicate-brick-op",
+ SLEN("replicate-brick-op"), op_type, op_type_len);
+ if (ret)
+ goto out;
+
+ local->xattr_req = dict_new();
+ if (!local->xattr_req)
+ goto out;
+
+ ret = afr_set_pending_dict(priv, local->xattr_req, local->pending);
+ if (ret < 0)
+ goto out;
+
+ if (AFR_ENTRY_TRANSACTION == type) {
+ count = afr_selfheal_entrylk(frame, this, loc->inode, this->name, NULL,
+ locked_nodes);
+ } else {
+ count = afr_selfheal_inodelk(frame, this, loc->inode, this->name,
+ LLONG_MAX - 1, 0, locked_nodes);
+ }
+
+ if (!count) {
+ gf_smsg(this->name, GF_LOG_ERROR, EAGAIN, AFR_MSG_REPLACE_BRICK_STATUS,
+ NULL);
+ ret = -EAGAIN;
+ goto unlock;
+ }
+
+ ret = afr_emptyb_set_pending_changelog(frame, this, locked_nodes);
+ if (ret)
+ goto unlock;
+ ret = 0;
unlock:
- if (AFR_ENTRY_TRANSACTION == type) {
- afr_selfheal_unentrylk (frame, this, loc->inode, this->name,
- NULL, locked_nodes, NULL);
- } else {
- afr_selfheal_uninodelk (frame, this, loc->inode, this->name,
- LLONG_MAX - 1, 0, locked_nodes);
- }
+ if (AFR_ENTRY_TRANSACTION == type) {
+ afr_selfheal_unentrylk(frame, this, loc->inode, this->name, NULL,
+ locked_nodes, NULL);
+ } else {
+ afr_selfheal_uninodelk(frame, this, loc->inode, this->name,
+ LLONG_MAX - 1, 0, locked_nodes);
+ }
out:
- return ret;
+ return ret;
}
void
-afr_brick_args_cleanup (void *opaque)
-{
- afr_empty_brick_args_t *data = NULL;
-
- data = opaque;
- loc_wipe (&data->loc);
- GF_FREE (data);
-}
-
-int
-_afr_handle_empty_brick_cbk (int ret, call_frame_t *frame, void *opaque)
-{
- afr_brick_args_cleanup (opaque);
- return 0;
-}
-
-int
-_afr_handle_empty_brick (void *opaque)
-{
-
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int empty_index = -1;
- int ret = -1;
- int op_errno = ENOMEM;
- call_frame_t *frame = NULL;
- xlator_t *this = NULL;
- char *op_type = NULL;
- afr_empty_brick_args_t *data = NULL;
-
- data = opaque;
- frame = data->frame;
- empty_index = data->empty_index;
- op_type = data->op_type;
- this = frame->this;
- priv = this->private;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- loc_copy (&local->loc, &data->loc);
-
- gf_msg (this->name, GF_LOG_INFO, 0, 0, "New brick is : %s",
- priv->children[empty_index]->name);
-
- ret = _afr_handle_empty_brick_type (this, frame, &local->loc, empty_index,
- AFR_METADATA_TRANSACTION, op_type);
- if (ret) {
- op_errno = -ret;
- ret = -1;
- goto out;
- }
-
- dict_unref (local->xdata_req);
- dict_unref (local->xattr_req);
- afr_matrix_cleanup (local->pending, priv->child_count);
- local->pending = NULL;
- local->xattr_req = NULL;
- local->xdata_req = NULL;
-
- ret = _afr_handle_empty_brick_type (this, frame, &local->loc, empty_index,
- AFR_ENTRY_TRANSACTION, op_type);
- if (ret) {
- op_errno = -ret;
- ret = -1;
- goto out;
- }
- ret = 0;
+afr_brick_args_cleanup(void *opaque)
+{
+ afr_empty_brick_args_t *data = NULL;
+
+ data = opaque;
+ loc_wipe(&data->loc);
+ GF_FREE(data);
+}
+
+int
+_afr_handle_empty_brick_cbk(int ret, call_frame_t *frame, void *opaque)
+{
+ afr_brick_args_cleanup(opaque);
+ return 0;
+}
+
+int
+_afr_handle_empty_brick(void *opaque)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int empty_index = -1;
+ int ret = -1;
+ int op_errno = ENOMEM;
+ call_frame_t *frame = NULL;
+ xlator_t *this = NULL;
+ char *op_type = NULL;
+ int op_type_len = 0;
+ afr_empty_brick_args_t *data = NULL;
+ call_frame_t *op_frame = NULL;
+
+ data = opaque;
+ frame = data->frame;
+ empty_index = data->empty_index;
+ if (!data->op_type)
+ goto out;
+
+ op_frame = copy_frame(frame);
+ if (!op_frame) {
+ ret = -1;
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ op_type = data->op_type;
+ op_type_len = strlen(op_type);
+ this = op_frame->this;
+ priv = this->private;
+
+ afr_set_lk_owner(op_frame, this, op_frame->root);
+ local = AFR_FRAME_INIT(op_frame, op_errno);
+ if (!local)
+ goto out;
+
+ loc_copy(&local->loc, &data->loc);
+
+ gf_smsg(this->name, GF_LOG_INFO, 0, AFR_MSG_NEW_BRICK, "name=%s",
+ priv->children[empty_index]->name, NULL);
+
+ ret = _afr_handle_empty_brick_type(this, op_frame, &local->loc, empty_index,
+ AFR_METADATA_TRANSACTION, op_type,
+ op_type_len);
+ if (ret) {
+ op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ dict_unref(local->xdata_req);
+ dict_unref(local->xattr_req);
+ afr_matrix_cleanup(local->pending, priv->child_count);
+ local->pending = NULL;
+ local->xattr_req = NULL;
+ local->xdata_req = NULL;
+
+ ret = _afr_handle_empty_brick_type(this, op_frame, &local->loc, empty_index,
+ AFR_ENTRY_TRANSACTION, op_type,
+ op_type_len);
+ if (ret) {
+ op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+ ret = 0;
out:
- AFR_STACK_UNWIND (setxattr, frame, ret, op_errno, NULL);
- return 0;
+ if (op_frame) {
+ AFR_STACK_DESTROY(op_frame);
+ }
+ AFR_STACK_UNWIND(setxattr, frame, ret, op_errno, NULL);
+ return 0;
+}
+
+int
+afr_split_brain_resolve_do(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ char *data)
+{
+ afr_local_t *local = NULL;
+ int ret = -1;
+ int op_errno = EINVAL;
+
+ local = frame->local;
+ local->xdata_req = dict_new();
+
+ if (!local->xdata_req) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ ret = dict_set_int32_sizen(local->xdata_req, "heal-op",
+ GF_SHD_OP_SBRAIN_HEAL_FROM_BRICK);
+ if (ret) {
+ op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+ ret = dict_set_str_sizen(local->xdata_req, "child-name", data);
+ if (ret) {
+ op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+ /* set spb choice to -1 whether heal succeeds or not:
+ * If heal succeeds : spb-choice should be set to -1 as
+ * it is no longer valid; file is not
+ * in split-brain anymore.
+ * If heal doesn't succeed:
+ * spb-choice should be set to -1
+ * otherwise reads will be served
+ * from spb-choice which is misleading.
+ */
+ ret = afr_inode_split_brain_choice_set(loc->inode, this, -1);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_WARNING, 0, AFR_MSG_SPLIT_BRAIN_SET_FAILED,
+ NULL);
+ afr_heal_splitbrain_file(frame, this, loc);
+ ret = 0;
+out:
+ if (ret < 0)
+ AFR_STACK_UNWIND(setxattr, frame, -1, op_errno, NULL);
+ return 0;
}
-
int
-afr_split_brain_resolve_do (call_frame_t *frame, xlator_t *this, loc_t *loc,
- char *data)
+afr_get_split_brain_child_index(xlator_t *this, void *value, size_t len)
{
- afr_local_t *local = NULL;
- int ret = -1;
- int op_errno = EINVAL;
+ int spb_child_index = -1;
+ char *spb_child_str = NULL;
- local = frame->local;
- local->xdata_req = dict_new ();
+ spb_child_str = alloca0(len + 1);
+ memcpy(spb_child_str, value, len);
- if (!local->xdata_req) {
- op_errno = ENOMEM;
- goto out;
- }
+ if (!strcmp(spb_child_str, "none"))
+ return -2;
- ret = dict_set_int32 (local->xdata_req, "heal-op",
- GF_SHD_OP_SBRAIN_HEAL_FROM_BRICK);
- if (ret) {
- op_errno = -ret;
- ret = -1;
- goto out;
- }
- ret = dict_set_str (local->xdata_req, "child-name", data);
- if (ret) {
- op_errno = -ret;
- ret = -1;
- goto out;
- }
- /* set spb choice to -1 whether heal succeeds or not:
- * If heal succeeds : spb-choice should be set to -1 as
- * it is no longer valid; file is not
- * in split-brain anymore.
- * If heal doesn't succeed:
- * spb-choice should be set to -1
- * otherwise reads will be served
- * from spb-choice which is misleading.
- */
- ret = afr_inode_split_brain_choice_set (loc->inode, this, -1);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR, "Failed to set"
- "split-brain choice to -1");
- afr_heal_splitbrain_file (frame, this, loc);
- ret = 0;
-out:
- if (ret < 0)
- AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
- return 0;
+ spb_child_index = afr_get_child_index_from_name(this, spb_child_str);
+ if (spb_child_index < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_SUBVOL,
+ "subvol=%s", spb_child_str, NULL);
+ }
+ return spb_child_index;
}
int
-afr_get_split_brain_child_index (xlator_t *this, void *value, size_t len)
+afr_can_set_split_brain_choice(void *opaque)
{
- int spb_child_index = -1;
- char *spb_child_str = NULL;
+ afr_spbc_timeout_t *data = opaque;
+ call_frame_t *frame = NULL;
+ xlator_t *this = NULL;
+ loc_t *loc = NULL;
+ int ret = -1;
- spb_child_str = alloca0 (len + 1);
- memcpy (spb_child_str, value, len);
+ frame = data->frame;
+ loc = data->loc;
+ this = frame->this;
- if (!strcmp (spb_child_str, "none"))
- return -2;
+ ret = afr_is_split_brain(frame, this, loc->inode, loc->gfid, &data->d_spb,
+ &data->m_spb);
- spb_child_index = afr_get_child_index_from_name (this,
- spb_child_str);
- if (spb_child_index < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_SUBVOL, "Invalid subvol: %s",
- spb_child_str);
- }
- return spb_child_index;
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ AFR_MSG_SPLIT_BRAIN_DETERMINE_FAILED, "gfid=%s",
+ uuid_utoa(loc->gfid), NULL);
+ return ret;
}
int
-afr_can_set_split_brain_choice (void *opaque)
+afr_handle_split_brain_commands(xlator_t *this, call_frame_t *frame, loc_t *loc,
+ dict_t *dict)
{
- afr_spbc_timeout_t *data = opaque;
- call_frame_t *frame = NULL;
- xlator_t *this = NULL;
- loc_t *loc = NULL;
- int ret = -1;
+ void *choice_value = NULL;
+ void *resolve_value = NULL;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_spbc_timeout_t *data = NULL;
+ int len = 0;
+ int spb_child_index = -1;
+ int ret = -1;
+ int op_errno = EINVAL;
- frame = data->frame;
- loc = data->loc;
- this = frame->this;
+ priv = this->private;
- ret = afr_is_split_brain (frame, this, loc->inode, loc->gfid,
- &data->d_spb, &data->m_spb);
+ ret = dict_get_ptr_and_len(dict, GF_AFR_SBRAIN_CHOICE, &choice_value, &len);
+ ret = dict_get_ptr_and_len(dict, GF_AFR_SBRAIN_RESOLVE, &resolve_value,
+ &len);
+ if (!choice_value && !resolve_value) {
+ ret = -1;
+ goto out;
+ }
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
- "Failed to determine if %s"
- " is in split-brain. "
- "Aborting split-brain-choice set.",
- uuid_utoa (loc->gfid));
- return ret;
-}
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local) {
+ ret = 1;
+ goto out;
+ }
-int
-afr_handle_split_brain_commands (xlator_t *this, call_frame_t *frame,
- loc_t *loc, dict_t *dict)
-{
- void *value = NULL;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- afr_spbc_timeout_t *data = NULL;
- int len = 0;
- int spb_child_index = -1;
- int ret = -1;
- int op_errno = EINVAL;
-
- priv = this->private;
+ local->op = GF_FOP_SETXATTR;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local) {
+ if (choice_value) {
+ spb_child_index = afr_get_split_brain_child_index(this, choice_value,
+ len);
+ if (spb_child_index < 0) {
+ /* Case where value was "none" */
+ if (spb_child_index == -2)
+ spb_child_index = -1;
+ else {
ret = 1;
+ op_errno = EINVAL;
goto out;
+ }
}
- local->op = GF_FOP_SETXATTR;
-
- ret = dict_get_ptr_and_len (dict, GF_AFR_SBRAIN_CHOICE, &value,
- &len);
- if (value) {
- spb_child_index = afr_get_split_brain_child_index (this, value,
- len);
- if (spb_child_index < 0) {
- /* Case where value was "none" */
- if (spb_child_index == -2)
- spb_child_index = -1;
- else {
- ret = 1;
- op_errno = EINVAL;
- goto out;
- }
- }
-
- data = GF_CALLOC (1, sizeof (*data), gf_afr_mt_spbc_timeout_t);
- if (!data) {
- ret = 1;
- goto out;
- }
- data->spb_child_index = spb_child_index;
- data->frame = frame;
- loc_copy (&local->loc, loc);
- data->loc = &local->loc;
- ret = synctask_new (this->ctx->env,
- afr_can_set_split_brain_choice,
- afr_set_split_brain_choice, NULL, data);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
- "Failed to create"
- " synctask. Aborting split-brain choice set"
- " for %s", loc->name);
- ret = 1;
- op_errno = ENOMEM;
- goto out;
- }
- ret = 0;
- goto out;
+ data = GF_CALLOC(1, sizeof(*data), gf_afr_mt_spbc_timeout_t);
+ if (!data) {
+ ret = 1;
+ goto out;
}
+ data->spb_child_index = spb_child_index;
+ data->frame = frame;
+ loc_copy(&local->loc, loc);
+ data->loc = &local->loc;
+ ret = synctask_new(this->ctx->env, afr_can_set_split_brain_choice,
+ afr_set_split_brain_choice, NULL, data);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN_STATUS,
+ "name=%s", loc->name, NULL);
+ ret = 1;
+ op_errno = ENOMEM;
+ goto out;
+ }
+ ret = 0;
+ goto out;
+ }
- ret = dict_get_ptr_and_len (dict, GF_AFR_SBRAIN_RESOLVE, &value, &len);
- if (value) {
- spb_child_index = afr_get_split_brain_child_index (this, value,
- len);
- if (spb_child_index < 0) {
- ret = 1;
- goto out;
- }
-
- afr_split_brain_resolve_do (frame, this, loc,
- priv->children[spb_child_index]->name);
- ret = 0;
+ if (resolve_value) {
+ spb_child_index = afr_get_split_brain_child_index(this, resolve_value,
+ len);
+ if (spb_child_index < 0) {
+ ret = 1;
+ goto out;
}
+
+ afr_split_brain_resolve_do(frame, this, loc,
+ priv->children[spb_child_index]->name);
+ ret = 0;
+ }
out:
- /* key was correct but value was invalid when ret == 1 */
- if (ret == 1) {
- AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
- if (data)
- GF_FREE (data);
- ret = 0;
- }
- return ret;
+ /* key was correct but value was invalid when ret == 1 */
+ if (ret == 1) {
+ AFR_STACK_UNWIND(setxattr, frame, -1, op_errno, NULL);
+ if (data)
+ GF_FREE(data);
+ ret = 0;
+ }
+ return ret;
}
int
-afr_handle_spb_choice_timeout (xlator_t *this, call_frame_t *frame,
- dict_t *dict)
+afr_handle_spb_choice_timeout(xlator_t *this, call_frame_t *frame, dict_t *dict)
{
- int ret = -1;
- int op_errno = 0;
- uint64_t timeout = 0;
- afr_private_t *priv = NULL;
+ int ret = -1;
+ int op_errno = 0;
+ uint64_t timeout = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- ret = dict_get_uint64 (dict, GF_AFR_SPB_CHOICE_TIMEOUT, &timeout);
- if (!ret) {
- priv->spb_choice_timeout = timeout * 60;
- AFR_STACK_UNWIND (setxattr, frame, ret, op_errno, NULL);
- }
+ ret = dict_get_uint64(dict, GF_AFR_SPB_CHOICE_TIMEOUT, &timeout);
+ if (!ret) {
+ priv->spb_choice_timeout = timeout * 60;
+ AFR_STACK_UNWIND(setxattr, frame, ret, op_errno, NULL);
+ }
- return ret;
+ return ret;
}
int
-afr_handle_empty_brick (xlator_t *this, call_frame_t *frame, loc_t *loc,
- dict_t *dict)
+afr_handle_empty_brick(xlator_t *this, call_frame_t *frame, loc_t *loc,
+ dict_t *dict)
{
- int ret = -1;
- int ab_ret = -1;
- int empty_index = -1;
- int op_errno = EPERM;
- char *empty_brick = NULL;
- char *op_type = NULL;
- afr_empty_brick_args_t *data = NULL;
+ int ret = -1;
+ int ab_ret = -1;
+ int empty_index = -1;
+ int op_errno = EPERM;
+ char *empty_brick = NULL;
+ char *op_type = NULL;
+ afr_empty_brick_args_t *data = NULL;
- ret = dict_get_str (dict, GF_AFR_REPLACE_BRICK, &empty_brick);
- if (!ret)
- op_type = GF_AFR_REPLACE_BRICK;
+ ret = dict_get_str_sizen(dict, GF_AFR_REPLACE_BRICK, &empty_brick);
+ if (!ret)
+ op_type = GF_AFR_REPLACE_BRICK;
- ab_ret = dict_get_str (dict, GF_AFR_ADD_BRICK, &empty_brick);
- if (!ab_ret)
- op_type = GF_AFR_ADD_BRICK;
+ ab_ret = dict_get_str_sizen(dict, GF_AFR_ADD_BRICK, &empty_brick);
+ if (!ab_ret)
+ op_type = GF_AFR_ADD_BRICK;
- if (ret && ab_ret)
- goto out;
+ if (ret && ab_ret)
+ goto out;
- if (frame->root->pid != GF_CLIENT_PID_SELF_HEALD) {
- gf_msg (this->name, GF_LOG_ERROR, EPERM,
- afr_get_msg_id (op_type),
- "'%s' is an internal extended attribute.",
- op_type);
- ret = 1;
- goto out;
+ if (frame->root->pid != GF_CLIENT_PID_ADD_REPLICA_MOUNT) {
+ gf_smsg(this->name, GF_LOG_ERROR, EPERM, AFR_MSG_INTERNAL_ATTR,
+ "op_type=%s", op_type, NULL);
+ ret = 1;
+ goto out;
+ }
+ empty_index = afr_get_child_index_from_name(this, empty_brick);
+
+ if (empty_index < 0) {
+ /* Didn't belong to this replica pair
+ * Just do a no-op
+ */
+ AFR_STACK_UNWIND(setxattr, frame, 0, 0, NULL);
+ return 0;
+ } else {
+ data = GF_CALLOC(1, sizeof(*data), gf_afr_mt_empty_brick_t);
+ if (!data) {
+ ret = 1;
+ op_errno = ENOMEM;
+ goto out;
}
- empty_index = afr_get_child_index_from_name (this, empty_brick);
-
- if (empty_index < 0) {
- /* Didn't belong to this replica pair
- * Just do a no-op
- */
- AFR_STACK_UNWIND (setxattr, frame, 0, 0, NULL);
- return 0;
- } else {
- data = GF_CALLOC (1, sizeof (*data),
- gf_afr_mt_empty_brick_t);
- if (!data) {
- ret = 1;
- op_errno = ENOMEM;
- goto out;
- }
- data->frame = frame;
- loc_copy (&data->loc, loc);
- data->empty_index = empty_index;
- data->op_type = op_type;
- ret = synctask_new (this->ctx->env,
- _afr_handle_empty_brick,
- _afr_handle_empty_brick_cbk,
- NULL, data);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- afr_get_msg_id (op_type),
- "Failed to create synctask.");
- ret = 1;
- op_errno = ENOMEM;
- afr_brick_args_cleanup (data);
- goto out;
- }
+ data->frame = frame;
+ loc_copy(&data->loc, loc);
+ data->empty_index = empty_index;
+ data->op_type = op_type;
+ ret = synctask_new(this->ctx->env, _afr_handle_empty_brick,
+ _afr_handle_empty_brick_cbk, NULL, data);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN_STATUS,
+ NULL);
+ ret = 1;
+ op_errno = ENOMEM;
+ afr_brick_args_cleanup(data);
+ goto out;
}
- ret = 0;
+ }
+ ret = 0;
out:
- if (ret == 1) {
- AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
- ret = 0;
- }
- return ret;
+ if (ret == 1) {
+ AFR_STACK_UNWIND(setxattr, frame, -1, op_errno, NULL);
+ ret = 0;
+ }
+ return ret;
}
static int
-afr_handle_special_xattr (xlator_t *this, call_frame_t *frame, loc_t *loc,
- dict_t *dict)
+afr_handle_special_xattr(xlator_t *this, call_frame_t *frame, loc_t *loc,
+ dict_t *dict)
{
- int ret = -1;
+ int ret = -1;
- ret = afr_handle_split_brain_commands (this, frame, loc, dict);
- if (ret == 0)
- goto out;
+ ret = afr_handle_split_brain_commands(this, frame, loc, dict);
+ if (ret == 0)
+ goto out;
- ret = afr_handle_spb_choice_timeout (this, frame, dict);
- if (ret == 0)
- goto out;
+ ret = afr_handle_spb_choice_timeout(this, frame, dict);
+ if (ret == 0)
+ goto out;
- /* Applicable for replace-brick and add-brick commands */
- ret = afr_handle_empty_brick (this, frame, loc, dict);
+ /* Applicable for replace-brick and add-brick commands */
+ ret = afr_handle_empty_brick(this, frame, loc, dict);
out:
- return ret;
+ return ret;
}
int
-afr_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
- int32_t flags, dict_t *xdata)
+afr_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = EINVAL;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = EINVAL;
- GF_IF_INTERNAL_XATTR_GOTO ("trusted.afr.*", dict,
- op_errno, out);
+ GF_IF_INTERNAL_XATTR_GOTO("trusted.afr.*", dict, op_errno, out);
- GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.afr.*", dict,
- op_errno, out);
+ GF_IF_INTERNAL_XATTR_GOTO("trusted.glusterfs.afr.*", dict, op_errno, out);
- ret = afr_handle_special_xattr (this, frame, loc, dict);
- if (ret == 0)
- return 0;
+ ret = afr_handle_special_xattr(this, frame, loc, dict);
+ if (ret == 0)
+ return 0;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.setxattr.dict = dict_ref (dict);
- local->cont.setxattr.flags = flags;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ local->cont.setxattr.dict = dict_ref(dict);
+ local->cont.setxattr.flags = flags;
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->transaction.wind = afr_setxattr_wind;
- local->transaction.unwind = afr_setxattr_unwind;
+ local->transaction.wind = afr_setxattr_wind;
+ local->transaction.unwind = afr_setxattr_unwind;
- loc_copy (&local->loc, loc);
- ret = afr_set_inode_local (this, local, loc->inode);
- if (ret)
- goto out;
+ loc_copy(&local->loc, loc);
+ ret = afr_set_inode_local(this, local, loc->inode);
+ if (ret)
+ goto out;
- local->transaction.main_frame = frame;
- local->transaction.start = LLONG_MAX - 1;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
- local->op = GF_FOP_SETXATTR;
+ local->op = GF_FOP_SETXATTR;
- ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
+ AFR_STACK_UNWIND(setxattr, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
/* {{{ fsetxattr */
-
int
-afr_fsetxattr_unwind (call_frame_t *frame, xlator_t *this)
+afr_fsetxattr_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- local = frame->local;
+ local = frame->local;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
-
- AFR_STACK_UNWIND (fsetxattr, main_frame, local->op_ret, local->op_errno,
- local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(fsetxattr, main_frame, local->op_ret, local->op_errno,
+ local->xdata_rsp);
+ return 0;
+}
int
-afr_fsetxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_fsetxattr_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- NULL, NULL, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, NULL,
+ NULL, NULL, xdata);
}
-
int
-afr_fsetxattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_fsetxattr_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_fsetxattr_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->fsetxattr,
- local->fd, local->cont.fsetxattr.dict,
- local->cont.fsetxattr.flags, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_fsetxattr_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->fsetxattr, local->fd,
+ local->cont.fsetxattr.dict, local->cont.fsetxattr.flags,
+ local->xdata_req);
+ return 0;
}
-
int
-afr_fsetxattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *dict, int32_t flags, dict_t *xdata)
+afr_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- GF_IF_INTERNAL_XATTR_GOTO ("trusted.afr.*", dict,
- op_errno, out);
+ GF_IF_INTERNAL_XATTR_GOTO("trusted.afr.*", dict, op_errno, out);
- GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.afr.*", dict,
- op_errno, out);
+ GF_IF_INTERNAL_XATTR_GOTO("trusted.glusterfs.afr.*", dict, op_errno, out);
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.fsetxattr.dict = dict_ref (dict);
- local->cont.fsetxattr.flags = flags;
+ local->cont.fsetxattr.dict = dict_ref(dict);
+ local->cont.fsetxattr.flags = flags;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->transaction.wind = afr_fsetxattr_wind;
- local->transaction.unwind = afr_fsetxattr_unwind;
+ local->transaction.wind = afr_fsetxattr_wind;
+ local->transaction.unwind = afr_fsetxattr_unwind;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_FSETXATTR;
+ local->op = GF_FOP_FSETXATTR;
- local->transaction.main_frame = frame;
- local->transaction.start = LLONG_MAX - 1;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
- ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (fsetxattr, frame, -1, op_errno, NULL);
- return 0;
+ AFR_STACK_UNWIND(fsetxattr, frame, -1, op_errno, NULL);
+ return 0;
}
/* }}} */
-
/* {{{ removexattr */
-
int
-afr_removexattr_unwind (call_frame_t *frame, xlator_t *this)
+afr_removexattr_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- local = frame->local;
+ local = frame->local;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
-
- AFR_STACK_UNWIND (removexattr, main_frame, local->op_ret, local->op_errno,
- local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(removexattr, main_frame, local->op_ret, local->op_errno,
+ local->xdata_rsp);
+ return 0;
+}
int
-afr_removexattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_removexattr_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- NULL, NULL, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, NULL,
+ NULL, NULL, xdata);
}
-
int
-afr_removexattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_removexattr_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_removexattr_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->removexattr,
- &local->loc, local->cont.removexattr.name,
- local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_removexattr_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->removexattr, &local->loc,
+ local->cont.removexattr.name, local->xdata_req);
+ return 0;
}
-
int
-afr_removexattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
+afr_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- GF_IF_NATIVE_XATTR_GOTO ("trusted.afr.*",
- name, op_errno, out);
+ GF_IF_NATIVE_XATTR_GOTO("trusted.afr.*", name, op_errno, out);
- GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.afr.*",
- name, op_errno, out);
+ GF_IF_NATIVE_XATTR_GOTO("trusted.glusterfs.afr.*", name, op_errno, out);
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.removexattr.name = gf_strdup (name);
+ local->cont.removexattr.name = gf_strdup(name);
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->transaction.wind = afr_removexattr_wind;
- local->transaction.unwind = afr_removexattr_unwind;
+ local->transaction.wind = afr_removexattr_wind;
+ local->transaction.unwind = afr_removexattr_unwind;
- loc_copy (&local->loc, loc);
- ret = afr_set_inode_local (this, local, loc->inode);
- if (ret)
- goto out;
+ loc_copy(&local->loc, loc);
+ ret = afr_set_inode_local(this, local, loc->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_REMOVEXATTR;
+ local->op = GF_FOP_REMOVEXATTR;
- local->transaction.main_frame = frame;
- local->transaction.start = LLONG_MAX - 1;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
- ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
- return 0;
+ AFR_STACK_UNWIND(removexattr, frame, -1, op_errno, NULL);
+ return 0;
}
/* ffremovexattr */
int
-afr_fremovexattr_unwind (call_frame_t *frame, xlator_t *this)
+afr_fremovexattr_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- local = frame->local;
+ local = frame->local;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
-
- AFR_STACK_UNWIND (fremovexattr, main_frame, local->op_ret, local->op_errno,
- local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(fremovexattr, main_frame, local->op_ret, local->op_errno,
+ local->xdata_rsp);
+ return 0;
+}
int
-afr_fremovexattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+afr_fremovexattr_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- NULL, NULL, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, NULL,
+ NULL, NULL, xdata);
}
-
int
-afr_fremovexattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_fremovexattr_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_fremovexattr_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->fremovexattr,
- local->fd, local->cont.removexattr.name,
- local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_fremovexattr_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->fremovexattr, local->fd,
+ local->cont.removexattr.name, local->xdata_req);
+ return 0;
}
-
int
-afr_fremovexattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
+afr_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- GF_IF_NATIVE_XATTR_GOTO ("trusted.afr.*",
- name, op_errno, out);
+ GF_IF_NATIVE_XATTR_GOTO("trusted.afr.*", name, op_errno, out);
- GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.afr.*",
- name, op_errno, out);
+ GF_IF_NATIVE_XATTR_GOTO("trusted.glusterfs.afr.*", name, op_errno, out);
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.removexattr.name = gf_strdup (name);
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ local->cont.removexattr.name = gf_strdup(name);
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->transaction.wind = afr_fremovexattr_wind;
- local->transaction.unwind = afr_fremovexattr_unwind;
+ local->transaction.wind = afr_fremovexattr_wind;
+ local->transaction.unwind = afr_fremovexattr_unwind;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_FREMOVEXATTR;
+ local->op = GF_FOP_FREMOVEXATTR;
- local->transaction.main_frame = frame;
- local->transaction.start = LLONG_MAX - 1;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
- ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (fremovexattr, frame, -1, op_errno, NULL);
+ AFR_STACK_UNWIND(fremovexattr, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
-
int
-afr_fallocate_unwind (call_frame_t *frame, xlator_t *this)
+afr_fallocate_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (fallocate, main_frame, local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(fallocate, main_frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ return 0;
+}
int
-afr_fallocate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+afr_fallocate_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- prebuf, postbuf, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, prebuf,
+ postbuf, NULL, xdata);
}
-
int
-afr_fallocate_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_fallocate_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_fallocate_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->fallocate,
- local->fd, local->cont.fallocate.mode,
- local->cont.fallocate.offset,
- local->cont.fallocate.len, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_fallocate_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->fallocate, local->fd,
+ local->cont.fallocate.mode, local->cont.fallocate.offset,
+ local->cont.fallocate.len, local->xdata_req);
+ return 0;
}
-
int
-afr_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
- off_t offset, size_t len, dict_t *xdata)
+afr_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+ off_t offset, size_t len, dict_t *xdata)
{
- call_frame_t *transaction_frame = NULL;
- afr_local_t *local = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ call_frame_t *transaction_frame = NULL;
+ afr_local_t *local = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.fallocate.mode = mode;
- local->cont.fallocate.offset = offset;
- local->cont.fallocate.len = len;
+ local->cont.fallocate.mode = mode;
+ local->cont.fallocate.offset = offset;
+ local->cont.fallocate.len = len;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->op = GF_FOP_FALLOCATE;
+ local->op = GF_FOP_FALLOCATE;
- local->transaction.wind = afr_fallocate_wind;
- local->transaction.unwind = afr_fallocate_unwind;
+ local->transaction.wind = afr_fallocate_wind;
+ local->transaction.unwind = afr_fallocate_unwind;
- local->transaction.main_frame = frame;
+ local->transaction.main_frame = frame;
- local->transaction.start = local->cont.fallocate.offset;
- local->transaction.len = 0;
+ local->transaction.start = local->cont.fallocate.offset;
+ local->transaction.len = 0;
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(fallocate, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
-
/* }}} */
/* {{{ discard */
int
-afr_discard_unwind (call_frame_t *frame, xlator_t *this)
+afr_discard_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- local = frame->local;
+ local = frame->local;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
-
- AFR_STACK_UNWIND (discard, main_frame, local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(discard, main_frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ return 0;
+}
int
-afr_discard_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+afr_discard_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- prebuf, postbuf, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, prebuf,
+ postbuf, NULL, xdata);
}
-
int
-afr_discard_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_discard_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_discard_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->discard,
- local->fd, local->cont.discard.offset,
- local->cont.discard.len, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_discard_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->discard, local->fd,
+ local->cont.discard.offset, local->cont.discard.len,
+ local->xdata_req);
+ return 0;
}
-
int
-afr_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- size_t len, dict_t *xdata)
+afr_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.discard.offset = offset;
- local->cont.discard.len = len;
+ local->cont.discard.offset = offset;
+ local->cont.discard.len = len;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->op = GF_FOP_DISCARD;
+ local->op = GF_FOP_DISCARD;
- local->transaction.wind = afr_discard_wind;
- local->transaction.unwind = afr_discard_unwind;
+ local->transaction.wind = afr_discard_wind;
+ local->transaction.unwind = afr_discard_unwind;
- local->transaction.main_frame = frame;
+ local->transaction.main_frame = frame;
- local->transaction.start = local->cont.discard.offset;
- local->transaction.len = 0;
+ local->transaction.start = local->cont.discard.offset;
+ local->transaction.len = 0;
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (discard, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(discard, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
-
/* {{{ zerofill */
int
-afr_zerofill_unwind (call_frame_t *frame, xlator_t *this)
+afr_zerofill_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t * local = NULL;
- call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- local = frame->local;
+ local = frame->local;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
-
- AFR_STACK_UNWIND (discard, main_frame, local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
-}
+ AFR_STACK_UNWIND(discard, main_frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ return 0;
+}
int
-afr_zerofill_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+afr_zerofill_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
struct iatt *postbuf, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- prebuf, postbuf, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, prebuf,
+ postbuf, NULL, xdata);
}
-
int
-afr_zerofill_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_zerofill_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_zerofill_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->zerofill,
- local->fd, local->cont.zerofill.offset,
- local->cont.zerofill.len, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_zerofill_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->zerofill, local->fd,
+ local->cont.zerofill.offset, local->cont.zerofill.len,
+ local->xdata_req);
+ return 0;
}
int
-afr_zerofill (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+afr_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
size_t len, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.zerofill.offset = offset;
- local->cont.zerofill.len = len;
+ local->cont.zerofill.offset = offset;
+ local->cont.zerofill.len = len;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ if (xdata)
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ else
+ local->xdata_req = dict_new();
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->op = GF_FOP_ZEROFILL;
+ local->op = GF_FOP_ZEROFILL;
- local->transaction.wind = afr_zerofill_wind;
- local->transaction.unwind = afr_zerofill_unwind;
+ local->transaction.wind = afr_zerofill_wind;
+ local->transaction.unwind = afr_zerofill_unwind;
- local->transaction.main_frame = frame;
+ local->transaction.main_frame = frame;
- local->transaction.start = local->cont.discard.offset;
- local->transaction.len = len;
+ local->transaction.start = local->cont.zerofill.offset;
+ local->transaction.len = len;
- afr_fix_open (fd, this);
+ afr_fix_open(fd, this);
- ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (zerofill, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(zerofill, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
/* }}} */
int32_t
-afr_xattrop_wind_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xattr, dict_t *xdata)
+afr_xattrop_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- NULL, NULL, xattr, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, NULL,
+ NULL, xattr, xdata);
}
int
-afr_xattrop_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_xattrop_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_xattrop_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->xattrop,
- &local->loc, local->cont.xattrop.optype,
- local->cont.xattrop.xattr, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_xattrop_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->xattrop, &local->loc,
+ local->cont.xattrop.optype, local->cont.xattrop.xattr,
+ local->xdata_req);
+ return 0;
}
int
-afr_xattrop_unwind (call_frame_t *frame, xlator_t *this)
+afr_xattrop_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (xattrop, main_frame, local->op_ret, local->op_errno,
- local->xattr_rsp, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
+
+ AFR_STACK_UNWIND(xattrop, main_frame, local->op_ret, local->op_errno,
+ local->xattr_rsp, local->xdata_rsp);
+ return 0;
}
int32_t
-afr_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
+afr_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.xattrop.xattr = dict_ref (xattr);
- local->cont.xattrop.optype = optype;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ local->cont.xattrop.xattr = dict_ref(xattr);
+ local->cont.xattrop.optype = optype;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- local->transaction.wind = afr_xattrop_wind;
- local->transaction.unwind = afr_xattrop_unwind;
+ local->transaction.wind = afr_xattrop_wind;
+ local->transaction.unwind = afr_xattrop_unwind;
- loc_copy (&local->loc, loc);
- ret = afr_set_inode_local (this, local, loc->inode);
- if (ret)
- goto out;
+ loc_copy(&local->loc, loc);
+ ret = afr_set_inode_local(this, local, loc->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_XATTROP;
+ local->op = GF_FOP_XATTROP;
- local->transaction.main_frame = frame;
- local->transaction.start = LLONG_MAX - 1;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
- ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(xattrop, frame, -1, op_errno, NULL, NULL);
+ return 0;
}
int32_t
-afr_fxattrop_wind_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xattr, dict_t *xdata)
+afr_fxattrop_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- NULL, NULL, xattr, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, NULL,
+ NULL, xattr, xdata);
}
int
-afr_fxattrop_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_fxattrop_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_fxattrop_wind_cbk, (void *) (long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->fxattrop,
- local->fd, local->cont.xattrop.optype,
- local->cont.xattrop.xattr, local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_fxattrop_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->fxattrop, local->fd,
+ local->cont.xattrop.optype, local->cont.xattrop.xattr,
+ local->xdata_req);
+ return 0;
}
int
-afr_fxattrop_unwind (call_frame_t *frame, xlator_t *this)
+afr_fxattrop_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- call_frame_t *main_frame = NULL;
-
- local = frame->local;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ local = frame->local;
- AFR_STACK_UNWIND (fxattrop, main_frame, local->op_ret, local->op_errno,
- local->xattr_rsp, local->xdata_rsp);
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
return 0;
+
+ AFR_STACK_UNWIND(fxattrop, main_frame, local->op_ret, local->op_errno,
+ local->xattr_rsp, local->xdata_rsp);
+ return 0;
}
int32_t
-afr_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
- gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
+afr_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = ENOMEM;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- local->cont.xattrop.xattr = dict_ref (xattr);
- local->cont.xattrop.optype = optype;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
+ local->cont.xattrop.xattr = dict_ref(xattr);
+ local->cont.xattrop.optype = optype;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
- local->transaction.wind = afr_fxattrop_wind;
- local->transaction.unwind = afr_fxattrop_unwind;
+ local->transaction.wind = afr_fxattrop_wind;
+ local->transaction.unwind = afr_fxattrop_unwind;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_FXATTROP;
+ local->op = GF_FOP_FXATTROP;
- local->transaction.main_frame = frame;
- local->transaction.start = LLONG_MAX - 1;
- local->transaction.len = 0;
+ local->transaction.main_frame = frame;
+ local->transaction.start = LLONG_MAX - 1;
+ local->transaction.len = 0;
- ret = afr_transaction (transaction_frame, this,
- AFR_METADATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_METADATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL, NULL);
- return 0;
+ AFR_STACK_UNWIND(fxattrop, frame, -1, op_errno, NULL, NULL);
+ return 0;
}
-
int
-afr_fsync_unwind (call_frame_t *frame, xlator_t *this)
+afr_fsync_unwind(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- call_frame_t *main_frame = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
- local = frame->local;
+ local = frame->local;
- main_frame = afr_transaction_detach_fop_frame (frame);
- if (!main_frame)
- return 0;
+ main_frame = afr_transaction_detach_fop_frame(frame);
+ if (!main_frame)
+ return 0;
- AFR_STACK_UNWIND (fsync, main_frame, local->op_ret, local->op_errno,
- &local->cont.inode_wfop.prebuf,
- &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+ AFR_STACK_UNWIND(fsync, main_frame, local->op_ret, local->op_errno,
+ &local->cont.inode_wfop.prebuf,
+ &local->cont.inode_wfop.postbuf, local->xdata_rsp);
- return 0;
+ return 0;
}
-
int
-afr_fsync_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+afr_fsync_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
- prebuf, postbuf, NULL, xdata);
+ return __afr_inode_write_cbk(frame, cookie, this, op_ret, op_errno, prebuf,
+ postbuf, NULL, xdata);
}
-
int
-afr_fsync_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_fsync_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- STACK_WIND_COOKIE (frame, afr_fsync_wind_cbk, (void *)(long) subvol,
- priv->children[subvol],
- priv->children[subvol]->fops->fsync,
- local->fd, local->cont.fsync.datasync,
- local->xdata_req);
- return 0;
+ STACK_WIND_COOKIE(frame, afr_fsync_wind_cbk, (void *)(long)subvol,
+ priv->children[subvol],
+ priv->children[subvol]->fops->fsync, local->fd,
+ local->cont.fsync.datasync, local->xdata_req);
+ return 0;
}
int
-afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
- dict_t *xdata)
+afr_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int32_t op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int32_t op_errno = ENOMEM;
+ int8_t last_fsync = 0;
- transaction_frame = copy_frame (frame);
- if (!transaction_frame)
- goto out;
+ AFR_ERROR_OUT_IF_FDCTX_INVALID(fd, this, op_errno, out);
+ transaction_frame = copy_frame(frame);
+ if (!transaction_frame)
+ goto out;
- local = AFR_FRAME_INIT (transaction_frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(transaction_frame, op_errno);
+ if (!local)
+ goto out;
- if (xdata)
- local->xdata_req = dict_copy_with_ref (xdata, NULL);
- else
- local->xdata_req = dict_new ();
+ if (xdata) {
+ local->xdata_req = dict_copy_with_ref(xdata, NULL);
+ if (dict_get_int8(xdata, "last-fsync", &last_fsync) == 0) {
+ if (last_fsync) {
+ local->transaction.disable_delayed_post_op = _gf_true;
+ }
+ }
+ } else {
+ local->xdata_req = dict_new();
+ }
- if (!local->xdata_req)
- goto out;
+ if (!local->xdata_req)
+ goto out;
- local->fd = fd_ref (fd);
- ret = afr_set_inode_local (this, local, fd->inode);
- if (ret)
- goto out;
+ local->fd = fd_ref(fd);
+ ret = afr_set_inode_local(this, local, fd->inode);
+ if (ret)
+ goto out;
- local->op = GF_FOP_FSYNC;
- local->cont.fsync.datasync = datasync;
+ local->op = GF_FOP_FSYNC;
+ local->cont.fsync.datasync = datasync;
- if (afr_fd_has_witnessed_unstable_write (this, fd->inode)) {
- /* don't care. we only wanted to CLEAR the bit */
- }
+ if (afr_fd_has_witnessed_unstable_write(this, fd->inode)) {
+ /* don't care. we only wanted to CLEAR the bit */
+ }
- local->transaction.wind = afr_fsync_wind;
- local->transaction.unwind = afr_fsync_unwind;
+ local->transaction.wind = afr_fsync_wind;
+ local->transaction.unwind = afr_fsync_unwind;
- local->transaction.main_frame = frame;
+ local->transaction.main_frame = frame;
- ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
- if (ret < 0) {
- op_errno = -ret;
- goto out;
- }
+ ret = afr_transaction(transaction_frame, this, AFR_DATA_TRANSACTION);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto out;
+ }
- return 0;
+ return 0;
out:
- if (transaction_frame)
- AFR_STACK_DESTROY (transaction_frame);
+ if (transaction_frame)
+ AFR_STACK_DESTROY(transaction_frame);
- AFR_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL, NULL);
+ AFR_STACK_UNWIND(fsync, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
diff --git a/xlators/cluster/afr/src/afr-inode-write.h b/xlators/cluster/afr/src/afr-inode-write.h
index 1e8bb5c12b3..a787069b7a1 100644
--- a/xlators/cluster/afr/src/afr-inode-write.h
+++ b/xlators/cluster/afr/src/afr-inode-write.h
@@ -12,83 +12,83 @@
#define __INODE_WRITE_H__
int32_t
-afr_chmod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dict_t *xdata);
+afr_chmod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dict_t *xdata);
int32_t
-afr_chown (call_frame_t *frame, xlator_t *this,
- loc_t *loc, uid_t uid, gid_t gid, dict_t *xdata);
+afr_chown(call_frame_t *frame, xlator_t *this, loc_t *loc, uid_t uid, gid_t gid,
+ dict_t *xdata);
int
-afr_fchown (call_frame_t *frame, xlator_t *this,
- fd_t *fd, uid_t uid, gid_t gid, dict_t *xdata);
+afr_fchown(call_frame_t *frame, xlator_t *this, fd_t *fd, uid_t uid, gid_t gid,
+ dict_t *xdata);
int32_t
-afr_fchmod (call_frame_t *frame, xlator_t *this,
- fd_t *fd, mode_t mode, dict_t *xdata);
+afr_fchmod(call_frame_t *frame, xlator_t *this, fd_t *fd, mode_t mode,
+ dict_t *xdata);
int32_t
-afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count, off_t offset,
- uint32_t flags, struct iobref *iobref, dict_t *xdata);
+afr_writev(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
+ int32_t count, off_t offset, uint32_t flags, struct iobref *iobref,
+ dict_t *xdata);
int32_t
-afr_truncate (call_frame_t *frame, xlator_t *this,
- loc_t *loc, off_t offset, dict_t *xdata);
+afr_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata);
int32_t
-afr_ftruncate (call_frame_t *frame, xlator_t *this,
- fd_t *fd, off_t offset, dict_t *xdata);
+afr_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata);
int32_t
-afr_utimens (call_frame_t *frame, xlator_t *this,
- loc_t *loc, struct timespec tv[2], dict_t *xdata);
+afr_utimens(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct timespec tv[2], dict_t *xdata);
int
-afr_setattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, struct iatt *buf, int32_t valid, dict_t *xdata);
+afr_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *buf,
+ int32_t valid, dict_t *xdata);
int
-afr_fsetattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, struct iatt *buf, int32_t valid, dict_t *xdata);
+afr_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *buf,
+ int32_t valid, dict_t *xdata);
int32_t
-afr_setxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata);
+afr_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata);
int32_t
-afr_fsetxattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *dict, int32_t flags, dict_t *xdata);
+afr_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata);
int32_t
-afr_removexattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata);
+afr_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata);
int32_t
-afr_fremovexattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, const char *name, dict_t *xdata);
+afr_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata);
int
-afr_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- size_t len, dict_t *xdata);
+afr_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata);
int
-afr_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
- off_t offset, size_t len, dict_t *xdata);
+afr_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+ off_t offset, size_t len, dict_t *xdata);
int
afr_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
off_t len, dict_t *xdata);
int32_t
-afr_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata);
+afr_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata);
int32_t
-afr_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
- gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata);
+afr_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata);
int
-afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
- dict_t *xdata);
+afr_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+ dict_t *xdata);
#endif /* __INODE_WRITE_H__ */
diff --git a/xlators/cluster/afr/src/afr-lk-common.c b/xlators/cluster/afr/src/afr-lk-common.c
index be3de01924d..bc8eabe0f43 100644
--- a/xlators/cluster/afr/src/afr-lk-common.c
+++ b/xlators/cluster/afr/src/afr-lk-common.c
@@ -8,9 +8,9 @@
cases as published by the Free Software Foundation.
*/
-#include "dict.h"
-#include "byte-order.h"
-#include "common-utils.h"
+#include <glusterfs/dict.h>
+#include <glusterfs/byte-order.h>
+#include <glusterfs/common-utils.h>
#include "afr.h"
#include "afr-transaction.h"
@@ -18,1121 +18,774 @@
#include <signal.h>
+#define LOCKED_NO 0x0 /* no lock held */
+#define LOCKED_YES 0x1 /* for DATA, METADATA, ENTRY and higher_path */
+#define LOCKED_LOWER 0x2 /* for lower path */
-#define LOCKED_NO 0x0 /* no lock held */
-#define LOCKED_YES 0x1 /* for DATA, METADATA, ENTRY and higher_path */
-#define LOCKED_LOWER 0x2 /* for lower path */
+void
+afr_lockee_cleanup(afr_lockee_t *lockee)
+{
+ if (lockee->fd) {
+ fd_unref(lockee->fd);
+ lockee->fd = NULL;
+ } else {
+ loc_wipe(&lockee->loc);
+ }
+
+ GF_FREE(lockee->basename);
+ lockee->basename = NULL;
+ GF_FREE(lockee->locked_nodes);
+ lockee->locked_nodes = NULL;
+
+ return;
+}
-int
-afr_entry_lockee_cmp (const void *l1, const void *l2)
+void
+afr_lockees_cleanup(afr_internal_lock_t *int_lock)
{
- const afr_entry_lockee_t *r1 = l1;
- const afr_entry_lockee_t *r2 = l2;
- int ret = 0;
- uuid_t gfid1 = {0};
- uuid_t gfid2 = {0};
-
- loc_gfid ((loc_t*)&r1->loc, gfid1);
- loc_gfid ((loc_t*)&r2->loc, gfid2);
- ret = gf_uuid_compare (gfid1, gfid2);
- /*Entrylks with NULL basename are the 'smallest'*/
- if (ret == 0) {
- if (!r1->basename)
- return -1;
- if (!r2->basename)
- return 1;
- ret = strcmp (r1->basename, r2->basename);
- }
+ int i = 0;
+
+ for (i = 0; i < int_lock->lockee_count; i++) {
+ afr_lockee_cleanup(&int_lock->lockee[i]);
+ }
- if (ret <= 0)
- return -1;
- else
- return 1;
+ return;
+}
+int
+afr_entry_lockee_cmp(const void *l1, const void *l2)
+{
+ const afr_lockee_t *r1 = l1;
+ const afr_lockee_t *r2 = l2;
+ int ret = 0;
+ uuid_t gfid1 = {0};
+ uuid_t gfid2 = {0};
+
+ loc_gfid((loc_t *)&r1->loc, gfid1);
+ loc_gfid((loc_t *)&r2->loc, gfid2);
+ ret = gf_uuid_compare(gfid1, gfid2);
+ /*Entrylks with NULL basename are the 'smallest'*/
+ if (ret == 0) {
+ if (!r1->basename)
+ return -1;
+ if (!r2->basename)
+ return 1;
+ ret = strcmp(r1->basename, r2->basename);
+ }
+
+ if (ret <= 0)
+ return -1;
+ else
+ return 1;
}
-int afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index);
+int
+afr_lock_blocking(call_frame_t *frame, xlator_t *this, int child_index);
void
-afr_set_lk_owner (call_frame_t *frame, xlator_t *this, void *lk_owner)
+afr_set_lk_owner(call_frame_t *frame, xlator_t *this, void *lk_owner)
{
- gf_msg_trace (this->name, 0,
- "Setting lk-owner=%llu",
- (unsigned long long) (unsigned long)lk_owner);
+ gf_msg_trace(this->name, 0, "Setting lk-owner=%llu",
+ (unsigned long long)(unsigned long)lk_owner);
- set_lk_owner_from_ptr (&frame->root->lk_owner, lk_owner);
+ set_lk_owner_from_ptr(&frame->root->lk_owner, lk_owner);
}
int32_t
-internal_lock_count (call_frame_t *frame, xlator_t *this)
+internal_lock_count(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int32_t call_count = 0;
- int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int32_t call_count = 0;
+ int i = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i])
- ++call_count;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i])
+ ++call_count;
+ }
- return call_count;
+ return call_count;
}
int
-afr_is_inodelk_transaction(afr_transaction_type type)
+afr_add_entry_lockee(afr_local_t *local, loc_t *loc, char *basename,
+ int child_count)
{
- int ret = 0;
+ int ret = -ENOMEM;
+ afr_internal_lock_t *int_lock = &local->internal_lock;
+ afr_lockee_t *lockee = &int_lock->lockee[int_lock->lockee_count];
- switch (type) {
- case AFR_DATA_TRANSACTION:
- case AFR_METADATA_TRANSACTION:
- ret = 1;
- break;
+ GF_ASSERT(int_lock->lockee_count < AFR_LOCKEE_COUNT_MAX);
+ loc_copy(&lockee->loc, loc);
+ lockee->basename = (basename) ? gf_strdup(basename) : NULL;
+ if (basename && !lockee->basename)
+ goto out;
- case AFR_ENTRY_RENAME_TRANSACTION:
- case AFR_ENTRY_TRANSACTION:
- ret = 0;
- break;
+ lockee->locked_count = 0;
+ lockee->locked_nodes = GF_CALLOC(child_count, sizeof(*lockee->locked_nodes),
+ gf_afr_mt_afr_node_character);
- }
+ if (!lockee->locked_nodes)
+ goto out;
- return ret;
+ ret = 0;
+ int_lock->lockee_count++;
+out:
+ if (ret) {
+ afr_lockee_cleanup(lockee);
+ }
+ return ret;
}
int
-afr_init_entry_lockee (afr_entry_lockee_t *lockee, afr_local_t *local,
- loc_t *loc, char *basename, int child_count)
+afr_add_inode_lockee(afr_local_t *local, int child_count)
{
- int ret = -1;
+ int ret = -ENOMEM;
+ afr_internal_lock_t *int_lock = &local->internal_lock;
+ afr_lockee_t *lockee = &int_lock->lockee[int_lock->lockee_count];
- loc_copy (&lockee->loc, loc);
- lockee->basename = (basename)? gf_strdup (basename): NULL;
- if (basename && !lockee->basename)
- goto out;
+ if (local->fd) {
+ lockee->fd = fd_ref(local->fd);
+ } else {
+ loc_copy(&lockee->loc, &local->loc);
+ }
- lockee->locked_count = 0;
- lockee->locked_nodes = GF_CALLOC (child_count,
- sizeof (*lockee->locked_nodes),
- gf_afr_mt_afr_node_character);
+ lockee->locked_count = 0;
+ lockee->locked_nodes = GF_CALLOC(child_count, sizeof(*lockee->locked_nodes),
+ gf_afr_mt_afr_node_character);
- if (!lockee->locked_nodes)
- goto out;
+ if (!lockee->locked_nodes)
+ goto out;
- ret = 0;
+ ret = 0;
+ int_lock->lockee_count++;
out:
- return ret;
-
-}
-
-void
-afr_entry_lockee_cleanup (afr_internal_lock_t *int_lock)
-{
- int i = 0;
-
- for (i = 0; i < int_lock->lockee_count; i++) {
- loc_wipe (&int_lock->lockee[i].loc);
- if (int_lock->lockee[i].basename)
- GF_FREE (int_lock->lockee[i].basename);
- if (int_lock->lockee[i].locked_nodes)
- GF_FREE (int_lock->lockee[i].locked_nodes);
- }
-
- return;
-}
-
-static int
-initialize_entrylk_variables (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
-
- int i = 0;
-
- priv = this->private;
- local = frame->local;
- int_lock = &local->internal_lock;
-
- int_lock->entrylk_lock_count = 0;
- int_lock->lock_op_ret = -1;
- int_lock->lock_op_errno = 0;
-
- for (i = 0; i < AFR_LOCKEE_COUNT_MAX; i++) {
- if (!int_lock->lockee[i].locked_nodes)
- break;
- int_lock->lockee[i].locked_count = 0;
- memset (int_lock->lockee[i].locked_nodes, 0,
- sizeof (*int_lock->lockee[i].locked_nodes) *
- priv->child_count);
- }
-
- return 0;
+ if (ret) {
+ afr_lockee_cleanup(lockee);
+ }
+ return ret;
}
static int
-initialize_inodelk_variables (call_frame_t *frame, xlator_t *this)
+initialize_internal_lock_variables(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
-
- priv = this->private;
- local = frame->local;
- int_lock = &local->internal_lock;
-
- int_lock->lock_count = 0;
- int_lock->lk_attempted_count = 0;
- int_lock->lock_op_ret = -1;
- int_lock->lock_op_errno = 0;
-
- memset (int_lock->locked_nodes, 0,
- sizeof (*int_lock->locked_nodes) * priv->child_count);
-
- return 0;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_private_t *priv = NULL;
+
+ int i = 0;
+
+ priv = this->private;
+ local = frame->local;
+ int_lock = &local->internal_lock;
+
+ int_lock->lock_count = 0;
+ int_lock->lock_op_ret = -1;
+ int_lock->lock_op_errno = 0;
+ int_lock->lk_attempted_count = 0;
+
+ for (i = 0; i < AFR_LOCKEE_COUNT_MAX; i++) {
+ if (!int_lock->lockee[i].locked_nodes)
+ break;
+ int_lock->lockee[i].locked_count = 0;
+ memset(int_lock->lockee[i].locked_nodes, 0,
+ sizeof(*int_lock->lockee[i].locked_nodes) * priv->child_count);
+ }
+
+ return 0;
}
int
-afr_lockee_locked_nodes_count (afr_internal_lock_t *int_lock)
+afr_lockee_locked_nodes_count(afr_internal_lock_t *int_lock)
{
- int call_count = 0;
- int i = 0;
+ int call_count = 0;
+ int i = 0;
- for (i = 0; i < int_lock->lockee_count; i++)
- call_count += int_lock->lockee[i].locked_count;
+ for (i = 0; i < int_lock->lockee_count; i++)
+ call_count += int_lock->lockee[i].locked_count;
- return call_count;
+ return call_count;
}
int
-afr_locked_nodes_count (unsigned char *locked_nodes, int child_count)
+afr_locked_nodes_count(unsigned char *locked_nodes, int child_count)
{
- int i = 0;
- int call_count = 0;
+ int i = 0;
+ int call_count = 0;
- for (i = 0; i < child_count; i++) {
- if (locked_nodes[i] & LOCKED_YES)
- call_count++;
- }
+ for (i = 0; i < child_count; i++) {
+ if (locked_nodes[i] & LOCKED_YES)
+ call_count++;
+ }
- return call_count;
+ return call_count;
}
-/* FIXME: What if UNLOCK fails */
-static int32_t
-afr_unlock_common_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+static void
+afr_log_locks_failure(call_frame_t *frame, char *where, char *what,
+ int op_errno)
{
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- int call_count = 0;
- int ret = 0;
+ xlator_t *this = frame->this;
+ gf_lkowner_t *lk_owner = &frame->root->lk_owner;
+ afr_local_t *local = frame->local;
+ const char *fop = NULL;
+ char *gfid = NULL;
+ const char *name = NULL;
- local = frame->local;
- int_lock = &local->internal_lock;
-
- if (local->transaction.type == AFR_DATA_TRANSACTION && op_ret != 1)
- ret = afr_write_subvol_reset (frame, this);
-
- LOCK (&frame->lock);
- {
- call_count = --int_lock->lk_call_count;
- }
- UNLOCK (&frame->lock);
-
- if (call_count == 0) {
- gf_msg_trace (this->name, 0,
- "All internal locks unlocked");
- int_lock->lock_cbk (frame, this);
- }
-
- return ret;
-}
-
-void
-afr_update_uninodelk (afr_local_t *local, afr_internal_lock_t *int_lock,
- int32_t child_index)
-{
- int_lock->locked_nodes[child_index] &= LOCKED_NO;
+ fop = gf_fop_list[local->op];
+ switch (local->transaction.type) {
+ case AFR_ENTRY_RENAME_TRANSACTION:
+ case AFR_ENTRY_TRANSACTION:
+ switch (local->op) {
+ case GF_FOP_LINK:
+ gfid = uuid_utoa(local->newloc.pargfid);
+ name = local->newloc.name;
+ break;
+ default:
+ gfid = uuid_utoa(local->loc.pargfid);
+ name = local->loc.name;
+ break;
+ }
+ gf_msg(this->name, GF_LOG_WARNING, op_errno,
+ AFR_MSG_INTERNAL_LKS_FAILED,
+ "Unable to do entry %s with lk-owner:%s on %s "
+ "while attempting %s on {pgfid:%s, name:%s}.",
+ what, lkowner_utoa(lk_owner), where, fop, gfid, name);
+ break;
+ case AFR_DATA_TRANSACTION:
+ case AFR_METADATA_TRANSACTION:
+ gfid = uuid_utoa(local->inode->gfid);
+ gf_msg(this->name, GF_LOG_WARNING, op_errno,
+ AFR_MSG_INTERNAL_LKS_FAILED,
+ "Unable to do inode %s with lk-owner:%s on %s "
+ "while attempting %s on gfid:%s.",
+ what, lkowner_utoa(lk_owner), where, fop, gfid);
+ break;
+ }
}
static int32_t
-afr_unlock_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- int32_t child_index = (long)cookie;
- afr_private_t *priv = NULL;
-
- local = frame->local;
- int_lock = &local->internal_lock;
-
- priv = this->private;
-
- if (op_ret < 0 && op_errno != ENOTCONN && op_errno != EBADFD) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- AFR_MSG_UNLOCK_FAIL,
- "path=%s gfid=%s: unlock failed on subvolume %s "
- "with lock owner %s", local->loc.path,
- loc_gfid_utoa (&(local->loc)),
- priv->children[child_index]->name,
- lkowner_utoa (&frame->root->lk_owner));
- }
-
- afr_update_uninodelk (local, int_lock, child_index);
-
- afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno, xdata);
-
- return 0;
-
-}
-
-static int
-afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
+afr_unlock_common_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- struct gf_flock flock = {0,};
- int call_count = 0;
- int i = 0;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
-
- flock.l_start = int_lock->flock.l_start;
- flock.l_len = int_lock->flock.l_len;
- flock.l_type = F_UNLCK;
-
- call_count = afr_locked_nodes_count (int_lock->locked_nodes,
- priv->child_count);
-
- int_lock->lk_call_count = call_count;
-
- if (!call_count) {
- GF_ASSERT (!local->transaction.do_eager_unlock);
- gf_msg_trace (this->name, 0,
- "No internal locks unlocked");
-
- int_lock->lock_cbk (frame, this);
- goto out;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if ((int_lock->locked_nodes[i] & LOCKED_YES) != LOCKED_YES)
- continue;
-
- if (local->fd) {
- STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
- (void *) (long)i,
- priv->children[i],
- priv->children[i]->fops->finodelk,
- int_lock->domain, local->fd,
- F_SETLK, &flock, NULL);
- } else {
- STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
- (void *) (long)i,
- priv->children[i],
- priv->children[i]->fops->inodelk,
- int_lock->domain, &local->loc,
- F_SETLK, &flock, NULL);
- }
-
- if (!--call_count)
- break;
- }
-out:
- return 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ int lockee_num = 0;
+ int call_count = 0;
+ int child_index = 0;
+ int ret = 0;
+
+ local = frame->local;
+ int_lock = &local->internal_lock;
+ priv = this->private;
+ lockee_num = (int)((long)cookie) / priv->child_count;
+ child_index = (int)((long)cookie) % priv->child_count;
+
+ if (op_ret < 0 && op_errno != ENOTCONN && op_errno != EBADFD) {
+ afr_log_locks_failure(frame, priv->children[child_index]->name,
+ "unlock", op_errno);
+ }
+
+ int_lock->lockee[lockee_num].locked_nodes[child_index] &= LOCKED_NO;
+ if (local->transaction.type == AFR_DATA_TRANSACTION && op_ret != 1)
+ ret = afr_write_subvol_reset(frame, this);
+
+ LOCK(&frame->lock);
+ {
+ call_count = --int_lock->lk_call_count;
+ }
+ UNLOCK(&frame->lock);
+
+ if (call_count == 0) {
+ int_lock->lock_cbk(frame, this);
+ }
+
+ return ret;
}
-static int32_t
-afr_unlock_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+void
+afr_internal_lock_wind(call_frame_t *frame,
+ int32_t (*cbk)(call_frame_t *, void *, xlator_t *,
+ int32_t, int32_t, dict_t *),
+ void *cookie, int child, int lockee_num,
+ gf_boolean_t blocking, gf_boolean_t unlock)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_internal_lock_t *int_lock = NULL;
- int32_t child_index = 0;
- int lockee_no = 0;
-
- priv = this->private;
- lockee_no = (int)((long) cookie) / priv->child_count;
- child_index = (int) ((long) cookie) % priv->child_count;
-
- local = frame->local;
- int_lock = &local->internal_lock;
-
- if (op_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- AFR_MSG_ENTRY_UNLOCK_FAIL,
- "%s: unlock failed on %s", local->loc.path,
- priv->children[child_index]->name);
- }
-
- int_lock->lockee[lockee_no].locked_nodes[child_index] &= LOCKED_NO;
- afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno, NULL);
+ afr_local_t *local = frame->local;
+ xlator_t *this = frame->this;
+ afr_private_t *priv = this->private;
+ afr_internal_lock_t *int_lock = &local->internal_lock;
+ entrylk_cmd cmd = ENTRYLK_LOCK_NB;
+ int32_t cmd1 = F_SETLK;
+ struct gf_flock flock = {
+ 0,
+ };
+
+ switch (local->transaction.type) {
+ case AFR_ENTRY_TRANSACTION:
+ case AFR_ENTRY_RENAME_TRANSACTION:
+ if (unlock) {
+ cmd = ENTRYLK_UNLOCK;
+ } else if (blocking) { /*Doesn't make sense to have blocking
+ unlock*/
+ cmd = ENTRYLK_LOCK;
+ }
+
+ if (local->fd) {
+ STACK_WIND_COOKIE(frame, cbk, cookie, priv->children[child],
+ priv->children[child]->fops->fentrylk,
+ int_lock->domain,
+ int_lock->lockee[lockee_num].fd,
+ int_lock->lockee[lockee_num].basename, cmd,
+ ENTRYLK_WRLCK, NULL);
+ } else {
+ STACK_WIND_COOKIE(frame, cbk, cookie, priv->children[child],
+ priv->children[child]->fops->entrylk,
+ int_lock->domain,
+ &int_lock->lockee[lockee_num].loc,
+ int_lock->lockee[lockee_num].basename, cmd,
+ ENTRYLK_WRLCK, NULL);
+ }
+ break;
- return 0;
+ case AFR_DATA_TRANSACTION:
+ case AFR_METADATA_TRANSACTION:
+ flock = int_lock->lockee[lockee_num].flock;
+ if (unlock) {
+ flock.l_type = F_UNLCK;
+ } else if (blocking) { /*Doesn't make sense to have blocking
+ unlock*/
+ cmd1 = F_SETLKW;
+ }
+
+ if (local->fd) {
+ STACK_WIND_COOKIE(
+ frame, cbk, cookie, priv->children[child],
+ priv->children[child]->fops->finodelk, int_lock->domain,
+ int_lock->lockee[lockee_num].fd, cmd1, &flock, NULL);
+ } else {
+ STACK_WIND_COOKIE(
+ frame, cbk, cookie, priv->children[child],
+ priv->children[child]->fops->inodelk, int_lock->domain,
+ &int_lock->lockee[lockee_num].loc, cmd1, &flock, NULL);
+ }
+ break;
+ }
}
static int
-afr_unlock_entrylk (call_frame_t *frame, xlator_t *this)
+afr_unlock_now(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
- int index = 0;
- int lockee_no = 0;
- int copies = 0;
- int i = -1;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
- copies = priv->child_count;
-
- call_count = afr_lockee_locked_nodes_count (int_lock);
-
- int_lock->lk_call_count = call_count;
-
- if (!call_count){
- gf_msg_trace (this->name, 0,
- "No internal locks unlocked");
- int_lock->lock_cbk (frame, this);
- goto out;
- }
-
- for (i = 0; i < int_lock->lockee_count * priv->child_count; i++) {
- lockee_no = i / copies;
- index = i % copies;
- if (int_lock->lockee[lockee_no].locked_nodes[index] & LOCKED_YES) {
-
- STACK_WIND_COOKIE (frame, afr_unlock_entrylk_cbk,
- (void *) (long) i,
- priv->children[index],
- priv->children[index]->fops->entrylk,
- int_lock->domain,
- &int_lock->lockee[lockee_no].loc,
- int_lock->lockee[lockee_no].basename,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK, NULL);
-
- if (!--call_count)
- break;
- }
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int child_index = 0;
+ int lockee_num = 0;
+ int i = -1;
+
+ local = frame->local;
+ int_lock = &local->internal_lock;
+ priv = this->private;
+
+ call_count = afr_lockee_locked_nodes_count(int_lock);
+
+ int_lock->lk_call_count = call_count;
+
+ if (!call_count) {
+ gf_msg_trace(this->name, 0, "No internal locks unlocked");
+ int_lock->lock_cbk(frame, this);
+ goto out;
+ }
+
+ for (i = 0; i < int_lock->lockee_count * priv->child_count; i++) {
+ lockee_num = i / priv->child_count;
+ child_index = i % priv->child_count;
+ if (int_lock->lockee[lockee_num].locked_nodes[child_index] &
+ LOCKED_YES) {
+ afr_internal_lock_wind(frame, afr_unlock_common_cbk,
+ (void *)(long)i, child_index, lockee_num,
+ _gf_false, _gf_true);
+ if (!--call_count)
+ break;
}
+ }
out:
- return 0;
-
-}
-
-int32_t
-afr_unlock_now (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = frame->local;
-
- if (afr_is_inodelk_transaction(local->transaction.type))
- afr_unlock_inodelk (frame, this);
- else
- afr_unlock_entrylk (frame, this);
- return 0;
+ return 0;
}
static int32_t
-afr_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+afr_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int cky = (long) cookie;
- int child_index = 0;
- int lockee_no = 0;
-
- priv = this->private;
- local = frame->local;
- int_lock = &local->internal_lock;
-
- child_index = ((int)cky) % priv->child_count;
- lockee_no = ((int)cky) / priv->child_count;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- if (op_errno == ENOSYS) {
- /* return ENOTSUP */
- gf_msg (this->name, GF_LOG_ERROR, ENOSYS,
- AFR_MSG_LOCK_XLATOR_NOT_LOADED,
- "subvolume does not support locking. "
- "please load features/locks xlator on server");
- local->op_ret = op_ret;
- int_lock->lock_op_ret = op_ret;
- }
-
- local->op_errno = op_errno;
- int_lock->lock_op_errno = op_errno;
- }
-
- int_lock->lk_attempted_count++;
- }
- UNLOCK (&frame->lock);
-
- if ((op_ret == -1) &&
- (op_errno == ENOSYS)) {
- afr_unlock_now (frame, this);
- } else {
- if (op_ret == 0) {
- if (local->transaction.type == AFR_ENTRY_TRANSACTION ||
- local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
- int_lock->lockee[lockee_no].locked_nodes[child_index] |= LOCKED_YES;
- int_lock->lockee[lockee_no].locked_count++;
- int_lock->entrylk_lock_count++;
- } else {
- int_lock->locked_nodes[child_index] |= LOCKED_YES;
- int_lock->lock_count++;
-
- if (local->transaction.type ==
- AFR_DATA_TRANSACTION) {
- LOCK(&local->inode->lock);
- {
- local->inode_ctx->lock_count++;
- }
- UNLOCK (&local->inode->lock);
- }
- }
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int cky = (long)cookie;
+ int child_index = 0;
+ int lockee_num = 0;
+
+ priv = this->private;
+ local = frame->local;
+ int_lock = &local->internal_lock;
+
+ child_index = ((int)cky) % priv->child_count;
+ lockee_num = ((int)cky) / priv->child_count;
+
+ LOCK(&frame->lock);
+ {
+ if (op_ret == -1) {
+ if (op_errno == ENOSYS) {
+ /* return ENOTSUP */
+ gf_msg(this->name, GF_LOG_ERROR, ENOSYS,
+ AFR_MSG_LOCK_XLATOR_NOT_LOADED,
+ "subvolume does not support locking. "
+ "please load features/locks xlator on server");
+ local->op_ret = op_ret;
+ int_lock->lock_op_ret = op_ret;
+ }
+
+ local->op_errno = op_errno;
+ int_lock->lock_op_errno = op_errno;
+ }
+
+ int_lock->lk_attempted_count++;
+ }
+ UNLOCK(&frame->lock);
+
+ if ((op_ret == -1) && (op_errno == ENOSYS)) {
+ afr_unlock_now(frame, this);
+ } else {
+ if (op_ret == 0) {
+ int_lock->lockee[lockee_num]
+ .locked_nodes[child_index] |= LOCKED_YES;
+ int_lock->lockee[lockee_num].locked_count++;
+ int_lock->lock_count++;
+ if (local->transaction.type == AFR_DATA_TRANSACTION) {
+ LOCK(&local->inode->lock);
+ {
+ local->inode_ctx->lock_count++;
}
- afr_lock_blocking (frame, this, cky + 1);
+ UNLOCK(&local->inode->lock);
+ }
}
+ afr_lock_blocking(frame, this, cky + 1);
+ }
- return 0;
-}
-
-static int32_t
-afr_blocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- afr_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
- return 0;
-
-}
-
-static int32_t
-afr_blocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- afr_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
- return 0;
-}
-
-static gf_boolean_t
-afr_is_entrylk (afr_transaction_type trans_type)
-{
- if (afr_is_inodelk_transaction (trans_type))
- return _gf_false;
- return _gf_true;
+ return 0;
}
static gf_boolean_t
-_is_lock_wind_needed (afr_local_t *local, int child_index)
-{
- if (!local->child_up[child_index])
- return _gf_false;
-
- return _gf_true;
-}
-
-static void
-afr_log_entry_locks_failure(xlator_t *this, afr_local_t *local,
- afr_internal_lock_t *int_lock)
+_is_lock_wind_needed(afr_local_t *local, int child_index)
{
- const char *fop = NULL;
- char *pargfid = NULL;
- const char *name = NULL;
-
- fop = gf_fop_list[local->op];
+ if (!local->child_up[child_index])
+ return _gf_false;
- switch (local->op) {
- case GF_FOP_LINK:
- pargfid = uuid_utoa(local->newloc.pargfid);
- name = local->newloc.name;
- break;
- default:
- pargfid = uuid_utoa(local->loc.pargfid);
- name = local->loc.name;
- break;
- }
-
- gf_msg (this->name, GF_LOG_WARNING, 0, AFR_MSG_BLOCKING_LKS_FAILED,
- "Unable to obtain sufficient blocking entry locks on at least "
- "one child while attempting %s on {pgfid:%s, name:%s}.", fop,
- pargfid, name);
+ return _gf_true;
}
static gf_boolean_t
-is_blocking_locks_count_sufficient (call_frame_t *frame, xlator_t *this)
+is_blocking_locks_count_sufficient(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_internal_lock_t *int_lock = NULL;
- gf_boolean_t is_entrylk = _gf_false;
- int child = 0;
- int nlockee = 0;
- int lockee_count = 0;
- gf_boolean_t ret = _gf_true;
-
- local = frame->local;
- priv = this->private;
- int_lock = &local->internal_lock;
- lockee_count = int_lock->lockee_count;
- is_entrylk = afr_is_entrylk (local->transaction.type);
-
- if (!is_entrylk) {
- if (int_lock->lock_count == 0) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_BLOCKING_LKS_FAILED, "Unable to obtain "
- "blocking inode lock on even one child for "
- "gfid:%s.", uuid_utoa (local->inode->gfid));
- return _gf_false;
- } else {
- /*inodelk succeded on atleast one child. */
- return _gf_true;
- }
-
- } else {
- if (int_lock->entrylk_lock_count == 0) {
- afr_log_entry_locks_failure (this, local, int_lock);
- return _gf_false;
- }
- /* For FOPS that take multiple sets of locks (mkdir, rename),
- * there must be atleast one brick on which the locks from
- * all lock sets were successful. */
- for (child = 0; child < priv->child_count; child++) {
- ret = _gf_true;
- for (nlockee = 0; nlockee < lockee_count; nlockee++) {
- if (!(int_lock->lockee[nlockee].locked_nodes[child] & LOCKED_YES))
- ret = _gf_false;
- }
- if (ret)
- return ret;
- }
- if (!ret)
- afr_log_entry_locks_failure (this, local, int_lock);
- }
-
- return ret;
-
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ int child = 0;
+ int nlockee = 0;
+ int lockee_count = 0;
+ gf_boolean_t ret = _gf_true;
+
+ local = frame->local;
+ priv = this->private;
+ int_lock = &local->internal_lock;
+ lockee_count = int_lock->lockee_count;
+
+ if (int_lock->lock_count == 0) {
+ afr_log_locks_failure(frame, "any subvolume", "lock",
+ int_lock->lock_op_errno);
+ return _gf_false;
+ }
+ /* For FOPS that take multiple sets of locks (mkdir, rename),
+ * there must be at least one brick on which the locks from
+ * all lock sets were successful. */
+ for (child = 0; child < priv->child_count; child++) {
+ ret = _gf_true;
+ for (nlockee = 0; nlockee < lockee_count; nlockee++) {
+ if (!(int_lock->lockee[nlockee].locked_nodes[child] & LOCKED_YES))
+ ret = _gf_false;
+ }
+ if (ret)
+ return ret;
+ }
+ if (!ret)
+ afr_log_locks_failure(frame, "all", "lock", int_lock->lock_op_errno);
+
+ return ret;
}
int
-afr_lock_blocking (call_frame_t *frame, xlator_t *this, int cookie)
+afr_lock_blocking(call_frame_t *frame, xlator_t *this, int cookie)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- struct gf_flock flock = {0,};
- uint64_t ctx = 0;
- int ret = 0;
- int child_index = 0;
- int lockee_no = 0;
- gf_boolean_t is_entrylk = _gf_false;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
- child_index = cookie % priv->child_count;
- lockee_no = cookie / priv->child_count;
- is_entrylk = afr_is_entrylk (local->transaction.type);
-
-
- if (!is_entrylk) {
- flock.l_start = int_lock->flock.l_start;
- flock.l_len = int_lock->flock.l_len;
- flock.l_type = int_lock->flock.l_type;
- }
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ uint64_t ctx = 0;
+ int ret = 0;
+ int child_index = 0;
+ int lockee_num = 0;
- if (local->fd) {
- ret = fd_ctx_get (local->fd, this, &ctx);
+ local = frame->local;
+ int_lock = &local->internal_lock;
+ priv = this->private;
+ child_index = cookie % priv->child_count;
+ lockee_num = cookie / priv->child_count;
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_FD_CTX_GET_FAILED,
- "unable to get fd ctx for fd=%p",
- local->fd);
+ if (local->fd) {
+ ret = fd_ctx_get(local->fd, this, &ctx);
- local->op_ret = -1;
- int_lock->lock_op_ret = -1;
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_FD_CTX_GET_FAILED,
+ "unable to get fd ctx for fd=%p", local->fd);
- afr_unlock_now (frame, this);
+ local->op_ret = -1;
+ int_lock->lock_op_ret = -1;
- return 0;
- }
- }
-
- if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
- if (!is_blocking_locks_count_sufficient (frame, this)) {
-
- local->op_ret = -1;
- int_lock->lock_op_ret = -1;
+ afr_unlock_now(frame, this);
- afr_unlock_now(frame, this);
-
- return 0;
- }
+ return 0;
}
+ }
- if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
- /* we're done locking */
-
- gf_msg_debug (this->name, 0,
- "we're done locking");
+ if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
+ if (!is_blocking_locks_count_sufficient(frame, this)) {
+ local->op_ret = -1;
+ int_lock->lock_op_ret = -1;
- int_lock->lock_op_ret = 0;
- int_lock->lock_cbk (frame, this);
- return 0;
- }
+ afr_unlock_now(frame, this);
- if (!_is_lock_wind_needed (local, child_index)) {
- afr_lock_blocking (frame, this, cookie + 1);
- return 0;
+ return 0;
}
+ }
- switch (local->transaction.type) {
- case AFR_DATA_TRANSACTION:
- case AFR_METADATA_TRANSACTION:
+ if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
+ /* we're done locking */
- if (local->fd) {
- STACK_WIND_COOKIE (frame, afr_blocking_inodelk_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->finodelk,
- int_lock->domain, local->fd,
- F_SETLKW, &flock, NULL);
-
- } else {
- STACK_WIND_COOKIE (frame, afr_blocking_inodelk_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->inodelk,
- int_lock->domain, &local->loc,
- F_SETLKW, &flock, NULL);
- }
+ gf_msg_debug(this->name, 0, "we're done locking");
- break;
+ int_lock->lock_op_ret = 0;
+ int_lock->lock_cbk(frame, this);
+ return 0;
+ }
- case AFR_ENTRY_RENAME_TRANSACTION:
- case AFR_ENTRY_TRANSACTION:
- /*Accounting for child_index increments on 'down'
- *and 'fd-less' children */
-
- if (local->fd) {
- STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
- (void *) (long) cookie,
- priv->children[child_index],
- priv->children[child_index]->fops->fentrylk,
- int_lock->domain, local->fd,
- int_lock->lockee[lockee_no].basename,
- ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
- } else {
- STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
- (void *) (long) cookie,
- priv->children[child_index],
- priv->children[child_index]->fops->entrylk,
- int_lock->domain,
- &int_lock->lockee[lockee_no].loc,
- int_lock->lockee[lockee_no].basename,
- ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
- }
+ if (!_is_lock_wind_needed(local, child_index)) {
+ afr_lock_blocking(frame, this, cookie + 1);
+ return 0;
+ }
- break;
- }
+ afr_internal_lock_wind(frame, afr_lock_cbk, (void *)(long)cookie,
+ child_index, lockee_num, _gf_true, _gf_false);
- return 0;
+ return 0;
}
int32_t
-afr_blocking_lock (call_frame_t *frame, xlator_t *this)
+afr_blocking_lock(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int up_count = 0;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int up_count = 0;
- priv = this->private;
- local = frame->local;
- int_lock = &local->internal_lock;
+ priv = this->private;
+ local = frame->local;
+ int_lock = &local->internal_lock;
- switch (local->transaction.type) {
- case AFR_DATA_TRANSACTION:
- case AFR_METADATA_TRANSACTION:
- initialize_inodelk_variables (frame, this);
- break;
-
- case AFR_ENTRY_RENAME_TRANSACTION:
- case AFR_ENTRY_TRANSACTION:
- up_count = AFR_COUNT (local->child_up, priv->child_count);
- int_lock->lk_call_count = int_lock->lk_expected_count
- = (int_lock->lockee_count *
- up_count);
- initialize_entrylk_variables (frame, this);
- break;
- }
+ up_count = AFR_COUNT(local->child_up, priv->child_count);
+ int_lock->lk_call_count = int_lock->lk_expected_count =
+ (int_lock->lockee_count * up_count);
+ initialize_internal_lock_variables(frame, this);
- afr_lock_blocking (frame, this, 0);
+ afr_lock_blocking(frame, this, 0);
- return 0;
+ return 0;
}
static int32_t
-afr_nonblocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- int call_count = 0;
- int child_index = (long) cookie;
- int copies = 0;
- int index = 0;
- int lockee_no = 0;
- afr_private_t *priv = NULL;
-
- priv = this->private;
-
- copies = priv->child_count;
- index = child_index % copies;
- lockee_no = child_index / copies;
-
- local = frame->local;
- int_lock = &local->internal_lock;
-
- LOCK (&frame->lock);
- {
- if (op_ret < 0 ) {
- if (op_errno == ENOSYS) {
- /* return ENOTSUP */
- gf_msg (this->name, GF_LOG_ERROR,
- ENOSYS, AFR_MSG_LOCK_XLATOR_NOT_LOADED,
- "subvolume does not support "
- "locking. please load features/locks"
- " xlator on server");
- local->op_ret = op_ret;
- int_lock->lock_op_ret = op_ret;
-
- int_lock->lock_op_errno = op_errno;
- local->op_errno = op_errno;
- }
- } else if (op_ret == 0) {
- int_lock->lockee[lockee_no].locked_nodes[index] |= \
- LOCKED_YES;
- int_lock->lockee[lockee_no].locked_count++;
- int_lock->entrylk_lock_count++;
- }
-
- call_count = --int_lock->lk_call_count;
- }
- UNLOCK (&frame->lock);
-
- if (call_count == 0) {
- gf_msg_trace (this->name, 0,
- "Last locking reply received");
- /* all locks successful. Proceed to call FOP */
- if (int_lock->entrylk_lock_count ==
- int_lock->lk_expected_count) {
- gf_msg_trace (this->name, 0,
- "All servers locked. Calling the cbk");
- int_lock->lock_op_ret = 0;
- int_lock->lock_cbk (frame, this);
- }
- /* Not all locks were successful. Unlock and try locking
- again, this time with serially blocking locks */
- else {
- gf_msg_trace (this->name, 0,
- "%d servers locked. Trying again "
- "with blocking calls",
- int_lock->lock_count);
-
- afr_unlock_now(frame, this);
- }
- }
-
- return 0;
-}
-
-int
-afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this)
+afr_nb_internal_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
- int copies = 0;
- int index = 0;
- int lockee_no = 0;
- int32_t call_count = 0;
- int i = 0;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
-
- copies = priv->child_count;
- initialize_entrylk_variables (frame, this);
-
- if (local->fd) {
- fd_ctx = afr_fd_ctx_get (local->fd, this);
- if (!fd_ctx) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_FD_CTX_GET_FAILED,
- "unable to get fd ctx for fd=%p",
- local->fd);
-
- local->op_ret = -1;
- int_lock->lock_op_ret = -1;
- local->op_errno = EINVAL;
- int_lock->lock_op_errno = EINVAL;
-
- afr_unlock_now (frame, this);
- return -1;
- }
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ int call_count = 0;
+ int child_index = 0;
+ int lockee_num = 0;
+ afr_private_t *priv = NULL;
- call_count = int_lock->lockee_count * internal_lock_count (frame, this);
- int_lock->lk_call_count = call_count;
- int_lock->lk_expected_count = call_count;
+ priv = this->private;
- if (!call_count) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_INFO_COMMON,
- "fd not open on any subvolumes. aborting.");
- afr_unlock_now (frame, this);
- goto out;
- }
+ child_index = ((long)cookie) % priv->child_count;
+ lockee_num = ((long)cookie) / priv->child_count;
- /* Send non-blocking entrylk calls only on up children
- and where the fd has been opened */
- for (i = 0; i < int_lock->lockee_count*priv->child_count; i++) {
- index = i%copies;
- lockee_no = i/copies;
- if (local->child_up[index]) {
- STACK_WIND_COOKIE (frame, afr_nonblocking_entrylk_cbk,
- (void *) (long) i,
- priv->children[index],
- priv->children[index]->fops->fentrylk,
- this->name, local->fd,
- int_lock->lockee[lockee_no].basename,
- ENTRYLK_LOCK_NB, ENTRYLK_WRLCK,
- NULL);
- if (!--call_count)
- break;
- }
- }
- } else {
- call_count = int_lock->lockee_count * internal_lock_count (frame, this);
- int_lock->lk_call_count = call_count;
- int_lock->lk_expected_count = call_count;
-
- for (i = 0; i < int_lock->lockee_count*priv->child_count; i++) {
- index = i%copies;
- lockee_no = i/copies;
- if (local->child_up[index]) {
- STACK_WIND_COOKIE (frame, afr_nonblocking_entrylk_cbk,
- (void *) (long) i,
- priv->children[index],
- priv->children[index]->fops->entrylk,
- this->name, &int_lock->lockee[lockee_no].loc,
- int_lock->lockee[lockee_no].basename,
- ENTRYLK_LOCK_NB, ENTRYLK_WRLCK,
- NULL);
-
- if (!--call_count)
- break;
- }
- }
- }
-out:
- return 0;
-}
+ local = frame->local;
+ int_lock = &local->internal_lock;
-int32_t
-afr_nonblocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- int call_count = 0;
- int child_index = (long) cookie;
-
- local = frame->local;
- int_lock = &local->internal_lock;
-
- LOCK (&frame->lock);
+ if (op_ret == 0 && local->transaction.type == AFR_DATA_TRANSACTION) {
+ LOCK(&local->inode->lock);
{
- if (op_ret < 0) {
- if (op_errno == ENOSYS) {
- /* return ENOTSUP */
- gf_msg (this->name, GF_LOG_ERROR, ENOSYS,
- AFR_MSG_LOCK_XLATOR_NOT_LOADED,
- "subvolume does not support "
- "locking. please load features/locks"
- " xlator on server");
- local->op_ret = op_ret;
- int_lock->lock_op_ret = op_ret;
- int_lock->lock_op_errno = op_errno;
- local->op_errno = op_errno;
- }
- } else {
- int_lock->locked_nodes[child_index] |= LOCKED_YES;
- int_lock->lock_count++;
- }
-
- call_count = --int_lock->lk_call_count;
- }
- UNLOCK (&frame->lock);
-
- if (op_ret == 0 && local->transaction.type == AFR_DATA_TRANSACTION) {
- LOCK (&local->inode->lock);
- {
- local->inode_ctx->lock_count++;
- }
- UNLOCK (&local->inode->lock);
- }
- if (call_count == 0) {
- gf_msg_trace (this->name, 0,
- "Last inode locking reply received");
- /* all locks successful. Proceed to call FOP */
- if (int_lock->lock_count == int_lock->lk_expected_count) {
- gf_msg_trace (this->name, 0,
- "All servers locked. Calling the cbk");
- int_lock->lock_op_ret = 0;
- int_lock->lock_cbk (frame, this);
- }
- /* Not all locks were successful. Unlock and try locking
- again, this time with serially blocking locks */
- else {
- gf_msg_trace (this->name, 0,
- "%d servers locked. "
- "Trying again with blocking calls",
- int_lock->lock_count);
-
- afr_unlock_now(frame, this);
- }
+ local->inode_ctx->lock_count++;
}
+ UNLOCK(&local->inode->lock);
+ }
- return 0;
+ LOCK(&frame->lock);
+ {
+ if (op_ret < 0) {
+ if (op_errno == ENOSYS) {
+ /* return ENOTSUP */
+ gf_msg(this->name, GF_LOG_ERROR, ENOSYS,
+ AFR_MSG_LOCK_XLATOR_NOT_LOADED,
+ "subvolume does not support "
+ "locking. please load features/locks"
+ " xlator on server");
+ local->op_ret = op_ret;
+ int_lock->lock_op_ret = op_ret;
+
+ int_lock->lock_op_errno = op_errno;
+ local->op_errno = op_errno;
+ }
+ } else if (op_ret == 0) {
+ int_lock->lockee[lockee_num]
+ .locked_nodes[child_index] |= LOCKED_YES;
+ int_lock->lockee[lockee_num].locked_count++;
+ int_lock->lock_count++;
+ }
+
+ call_count = --int_lock->lk_call_count;
+ }
+ UNLOCK(&frame->lock);
+
+ if (call_count == 0) {
+ gf_msg_trace(this->name, 0, "Last locking reply received");
+ /* all locks successful. Proceed to call FOP */
+ if (int_lock->lock_count == int_lock->lk_expected_count) {
+ gf_msg_trace(this->name, 0, "All servers locked. Calling the cbk");
+ int_lock->lock_op_ret = 0;
+ int_lock->lock_cbk(frame, this);
+ }
+ /* Not all locks were successful. Unlock and try locking
+ again, this time with serially blocking locks */
+ else {
+ gf_msg_trace(this->name, 0,
+ "%d servers locked. Trying again "
+ "with blocking calls",
+ int_lock->lock_count);
+
+ afr_unlock_now(frame, this);
+ }
+ }
+
+ return 0;
}
int
-afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
+afr_lock_nonblocking(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
- int32_t call_count = 0;
- int i = 0;
- int ret = 0;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
-
- initialize_inodelk_variables (frame, this);
-
- if (local->fd) {
- fd_ctx = afr_fd_ctx_get (local->fd, this);
- if (!fd_ctx) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_FD_CTX_GET_FAILED,
- "unable to get fd ctx for fd=%p",
- local->fd);
-
- local->op_ret = -1;
- int_lock->lock_op_ret = -1;
- local->op_errno = EINVAL;
- int_lock->lock_op_errno = EINVAL;
-
- afr_unlock_now (frame, this);
- ret = -1;
- goto out;
- }
- }
-
- call_count = internal_lock_count (frame, this);
- int_lock->lk_call_count = call_count;
- int_lock->lk_expected_count = call_count;
-
- if (!call_count) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_SUBVOLS_DOWN,
- "All bricks are down, aborting.");
- afr_unlock_now (frame, this);
- goto out;
- }
-
- /* Send non-blocking inodelk calls only on up children
- and where the fd has been opened */
- for (i = 0; i < priv->child_count; i++) {
- if (!local->child_up[i])
- continue;
-
- if (local->fd) {
- STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->finodelk,
- int_lock->domain, local->fd,
- F_SETLK, &int_lock->flock, NULL);
- } else {
-
- STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->inodelk,
- int_lock->domain, &local->loc,
- F_SETLK, &int_lock->flock, NULL);
- }
- if (!--call_count)
- break;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int child = 0;
+ int lockee_num = 0;
+ int32_t call_count = 0;
+ int i = 0;
+ int ret = 0;
+
+ local = frame->local;
+ int_lock = &local->internal_lock;
+ priv = this->private;
+
+ initialize_internal_lock_variables(frame, this);
+
+ if (local->fd) {
+ fd_ctx = afr_fd_ctx_get(local->fd, this);
+ if (!fd_ctx) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_FD_CTX_GET_FAILED,
+ "unable to get fd ctx for fd=%p", local->fd);
+
+ local->op_ret = -1;
+ int_lock->lock_op_ret = -1;
+ local->op_errno = EINVAL;
+ int_lock->lock_op_errno = EINVAL;
+
+ afr_unlock_now(frame, this);
+ ret = -1;
+ goto out;
+ }
+ }
+
+ call_count = int_lock->lockee_count * internal_lock_count(frame, this);
+ int_lock->lk_call_count = call_count;
+ int_lock->lk_expected_count = call_count;
+
+ if (!call_count) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_INFO_COMMON,
+ "fd not open on any subvolumes. aborting.");
+ afr_unlock_now(frame, this);
+ goto out;
+ }
+
+ /* Send non-blocking lock calls only on up children
+ and where the fd has been opened */
+ for (i = 0; i < int_lock->lockee_count * priv->child_count; i++) {
+ child = i % priv->child_count;
+ lockee_num = i / priv->child_count;
+ if (local->child_up[child]) {
+ afr_internal_lock_wind(frame, afr_nb_internal_lock_cbk,
+ (void *)(long)i, child, lockee_num,
+ _gf_false, _gf_false);
+ if (!--call_count)
+ break;
}
+ }
out:
- return ret;
+ return ret;
}
int32_t
-afr_unlock (call_frame_t *frame, xlator_t *this)
+afr_unlock(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_lock_t *lock = NULL;
-
- local = frame->local;
-
- if (!local->transaction.eager_lock_on)
- goto out;
- lock = &local->inode_ctx->lock[local->transaction.type];
- LOCK (&local->inode->lock);
- {
- list_del_init (&local->transaction.owner_list);
- if (list_empty (&lock->owners) && list_empty (&lock->post_op)) {
- local->transaction.do_eager_unlock = _gf_true;
- /*TODO: Need to get metadata use on_disk and inherit/uninherit
- *GF_ASSERT (!local->inode_ctx->on_disk[local->transaction.type]);
- *GF_ASSERT (!local->inode_ctx->inherited[local->transaction.type]);
- */
- GF_ASSERT (lock->release);
- }
- }
- UNLOCK (&local->inode->lock);
- if (!local->transaction.do_eager_unlock) {
- local->internal_lock.lock_cbk (frame, this);
- return 0;
- }
+ afr_local_t *local = NULL;
+ afr_lock_t *lock = NULL;
+
+ local = frame->local;
+
+ if (!local->transaction.eager_lock_on)
+ goto out;
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ LOCK(&local->inode->lock);
+ {
+ list_del_init(&local->transaction.owner_list);
+ if (list_empty(&lock->owners) && list_empty(&lock->post_op)) {
+ local->transaction.do_eager_unlock = _gf_true;
+ /*TODO: Need to get metadata use on_disk and inherit/uninherit
+ *GF_ASSERT (!local->inode_ctx->on_disk[local->transaction.type]);
+ *GF_ASSERT (!local->inode_ctx->inherited[local->transaction.type]);
+ */
+ GF_ASSERT(lock->release);
+ }
+ }
+ UNLOCK(&local->inode->lock);
+ if (!local->transaction.do_eager_unlock) {
+ local->internal_lock.lock_cbk(frame, this);
+ return 0;
+ }
out:
- afr_unlock_now (frame, this);
- return 0;
+ afr_unlock_now(frame, this);
+ return 0;
}
diff --git a/xlators/cluster/afr/src/afr-mem-types.h b/xlators/cluster/afr/src/afr-mem-types.h
index 2e1117fc18c..816065fb57a 100644
--- a/xlators/cluster/afr/src/afr-mem-types.h
+++ b/xlators/cluster/afr/src/afr-mem-types.h
@@ -8,47 +8,31 @@
cases as published by the Free Software Foundation.
*/
-
#ifndef __AFR_MEM_TYPES_H__
#define __AFR_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_afr_mem_types_ {
- gf_afr_mt_iovec = gf_common_mt_end + 1,
- gf_afr_mt_afr_fd_ctx_t,
- gf_afr_mt_afr_private_t,
- gf_afr_mt_int32_t,
- gf_afr_mt_char,
- gf_afr_mt_xattr_key,
- gf_afr_mt_dict_t,
- gf_afr_mt_xlator_t,
- gf_afr_mt_iatt,
- gf_afr_mt_int,
- gf_afr_mt_afr_node_character,
- gf_afr_mt_sh_diff_loop_state,
- gf_afr_mt_uint8_t,
- gf_afr_mt_loc_t,
- gf_afr_mt_entry_name,
- gf_afr_mt_pump_priv,
- gf_afr_mt_locked_fd,
- gf_afr_mt_inode_ctx_t,
- gf_afr_fd_paused_call_t,
- gf_afr_mt_crawl_data_t,
- gf_afr_mt_brick_pos_t,
- gf_afr_mt_shd_bool_t,
- gf_afr_mt_shd_timer_t,
- gf_afr_mt_shd_event_t,
- gf_afr_mt_time_t,
- gf_afr_mt_pos_data_t,
- gf_afr_mt_reply_t,
- gf_afr_mt_subvol_healer_t,
- gf_afr_mt_spbc_timeout_t,
- gf_afr_mt_spb_status_t,
- gf_afr_mt_empty_brick_t,
- gf_afr_mt_child_latency_t,
- gf_afr_mt_atomic_t,
+ gf_afr_mt_afr_fd_ctx_t = gf_common_mt_end + 1,
+ gf_afr_mt_afr_private_t,
+ gf_afr_mt_int32_t,
+ gf_afr_mt_char,
+ gf_afr_mt_xattr_key,
+ gf_afr_mt_dict_t,
+ gf_afr_mt_xlator_t,
+ gf_afr_mt_afr_node_character,
+ gf_afr_mt_inode_ctx_t,
+ gf_afr_mt_shd_event_t,
+ gf_afr_mt_reply_t,
+ gf_afr_mt_subvol_healer_t,
+ gf_afr_mt_spbc_timeout_t,
+ gf_afr_mt_spb_status_t,
+ gf_afr_mt_empty_brick_t,
+ gf_afr_mt_child_latency_t,
+ gf_afr_mt_atomic_t,
+ gf_afr_mt_lk_heal_info_t,
+ gf_afr_mt_gf_lock,
gf_afr_mt_end
};
#endif
-
diff --git a/xlators/cluster/afr/src/afr-messages.h b/xlators/cluster/afr/src/afr-messages.h
index 8aa94730158..e73fd997765 100644
--- a/xlators/cluster/afr/src/afr-messages.h
+++ b/xlators/cluster/afr/src/afr-messages.h
@@ -11,7 +11,7 @@
#ifndef _AFR_MESSAGES_H_
#define _AFR_MESSAGES_H_
-#include "glfs-message-id.h"
+#include <glusterfs/glfs-message-id.h>
/* To add new message IDs, append new identifiers at the end of the list.
*
@@ -23,51 +23,145 @@
* glfs-message-id.h.
*/
-GLFS_MSGID(AFR,
- AFR_MSG_QUORUM_FAIL,
- AFR_MSG_QUORUM_MET,
- AFR_MSG_QUORUM_OVERRIDE,
- AFR_MSG_INVALID_CHILD_UP,
- AFR_MSG_SUBVOL_UP,
- AFR_MSG_SUBVOLS_DOWN,
- AFR_MSG_ENTRY_UNLOCK_FAIL,
- AFR_MSG_SPLIT_BRAIN,
- AFR_MSG_OPEN_FAIL,
- AFR_MSG_UNLOCK_FAIL,
- AFR_MSG_REPLACE_BRICK_STATUS,
- AFR_MSG_GFID_NULL,
- AFR_MSG_FD_CREATE_FAILED,
- AFR_MSG_DICT_SET_FAILED,
- AFR_MSG_EXPUNGING_FILE_OR_DIR,
- AFR_MSG_MIGRATION_IN_PROGRESS,
- AFR_MSG_CHILD_MISCONFIGURED,
- AFR_MSG_VOL_MISCONFIGURED,
- AFR_MSG_BLOCKING_LKS_FAILED,
- AFR_MSG_INVALID_FD,
- AFR_MSG_LOCK_INFO,
- AFR_MSG_LOCK_XLATOR_NOT_LOADED,
- AFR_MSG_FD_CTX_GET_FAILED,
- AFR_MSG_INVALID_SUBVOL,
- AFR_MSG_PUMP_XLATOR_ERROR,
- AFR_MSG_SELF_HEAL_INFO,
- AFR_MSG_READ_SUBVOL_ERROR,
- AFR_MSG_DICT_GET_FAILED,
- AFR_MSG_INFO_COMMON,
- AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR,
- AFR_MSG_LOCAL_CHILD,
- AFR_MSG_INVALID_DATA,
- AFR_MSG_INVALID_ARG,
- AFR_MSG_INDEX_DIR_GET_FAILED,
- AFR_MSG_FSYNC_FAILED,
- AFR_MSG_FAVORITE_CHILD,
- AFR_MSG_SELF_HEAL_FAILED,
- AFR_MSG_SPLIT_BRAIN_STATUS,
- AFR_MSG_ADD_BRICK_STATUS,
- AFR_MSG_NO_CHANGELOG,
- AFR_MSG_TIMER_CREATE_FAIL,
- AFR_MSG_SBRAIN_FAV_CHILD_POLICY,
- AFR_MSG_INODE_CTX_GET_FAILED,
- AFR_MSG_THIN_ARB
-);
+GLFS_MSGID(
+ AFR, AFR_MSG_QUORUM_FAIL, AFR_MSG_QUORUM_MET, AFR_MSG_QUORUM_OVERRIDE,
+ AFR_MSG_INVALID_CHILD_UP, AFR_MSG_SUBVOL_UP, AFR_MSG_SUBVOLS_DOWN,
+ AFR_MSG_ENTRY_UNLOCK_FAIL, AFR_MSG_SPLIT_BRAIN, AFR_MSG_OPEN_FAIL,
+ AFR_MSG_UNLOCK_FAIL, AFR_MSG_REPLACE_BRICK_STATUS, AFR_MSG_GFID_NULL,
+ AFR_MSG_FD_CREATE_FAILED, AFR_MSG_DICT_SET_FAILED,
+ AFR_MSG_EXPUNGING_FILE_OR_DIR, AFR_MSG_MIGRATION_IN_PROGRESS,
+ AFR_MSG_CHILD_MISCONFIGURED, AFR_MSG_VOL_MISCONFIGURED,
+ AFR_MSG_INTERNAL_LKS_FAILED, AFR_MSG_INVALID_FD, AFR_MSG_LOCK_INFO,
+ AFR_MSG_LOCK_XLATOR_NOT_LOADED, AFR_MSG_FD_CTX_GET_FAILED,
+ AFR_MSG_INVALID_SUBVOL, AFR_MSG_PUMP_XLATOR_ERROR, AFR_MSG_SELF_HEAL_INFO,
+ AFR_MSG_READ_SUBVOL_ERROR, AFR_MSG_DICT_GET_FAILED, AFR_MSG_INFO_COMMON,
+ AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR, AFR_MSG_LOCAL_CHILD, AFR_MSG_INVALID_DATA,
+ AFR_MSG_INVALID_ARG, AFR_MSG_INDEX_DIR_GET_FAILED, AFR_MSG_FSYNC_FAILED,
+ AFR_MSG_FAVORITE_CHILD, AFR_MSG_SELF_HEAL_FAILED,
+ AFR_MSG_SPLIT_BRAIN_STATUS, AFR_MSG_ADD_BRICK_STATUS, AFR_MSG_NO_CHANGELOG,
+ AFR_MSG_TIMER_CREATE_FAIL, AFR_MSG_SBRAIN_FAV_CHILD_POLICY,
+ AFR_MSG_INODE_CTX_GET_FAILED, AFR_MSG_THIN_ARB,
+ AFR_MSG_THIN_ARB_XATTROP_FAILED, AFR_MSG_THIN_ARB_LOC_POP_FAILED,
+ AFR_MSG_GET_PEND_VAL, AFR_MSG_THIN_ARB_SKIP_SHD, AFR_MSG_UNKNOWN_SET,
+ AFR_MSG_NO_XL_ID, AFR_MSG_SELF_HEAL_INFO_START,
+ AFR_MSG_SELF_HEAL_INFO_FINISH, AFR_MSG_INCRE_COUNT,
+ AFR_MSG_ADD_TO_OUTPUT_FAILED, AFR_MSG_SET_TIME_FAILED,
+ AFR_MSG_GFID_MISMATCH_DETECTED, AFR_MSG_GFID_HEAL_MSG,
+ AFR_MSG_THIN_ARB_LOOKUP_FAILED, AFR_MSG_DICT_CREATE_FAILED,
+ AFR_MSG_NO_MAJORITY_TO_RESOLVE, AFR_MSG_TYPE_MISMATCH,
+ AFR_MSG_SIZE_POLICY_NOT_APPLICABLE, AFR_MSG_NO_CHILD_SELECTED,
+ AFR_MSG_INVALID_CHILD, AFR_MSG_RESOLVE_CONFLICTING_DATA,
+ SERROR_GETTING_SRC_BRICK, SNO_DIFF_IN_MTIME, SNO_BIGGER_FILE,
+ SALL_BRICKS_UP_TO_RESOLVE, AFR_MSG_UNLOCK_FAILED, AFR_MSG_POST_OP_FAILED,
+ AFR_MSG_TA_FRAME_CREATE_FAILED, AFR_MSG_SET_KEY_XATTROP_FAILED,
+ AFR_MSG_BLOCKING_ENTRYLKS_FAILED, AFR_MSG_FOP_FAILED,
+ AFR_MSG_CLEAN_UP_FAILED, AFR_MSG_UNABLE_TO_FETCH, AFR_MSG_XATTR_SET_FAILED,
+ AFR_MSG_SPLIT_BRAIN_REPLICA, AFR_MSG_INODE_CTX_FAILED,
+ AFR_MSG_LOOKUP_FAILED, AFR_MSG_ALL_SUBVOLS_DOWN,
+ AFR_MSG_RELEASE_LOCK_FAILED, AFR_MSG_CLEAR_TIME_SPLIT_BRAIN,
+ AFR_MSG_READ_FAILED, AFR_MSG_LAUNCH_FAILED, AFR_MSG_READ_SUBVOL_NOT_UP,
+ AFR_MSG_LK_HEAL_DOM, AFR_MSG_NEW_BRICK, AFR_MSG_SPLIT_BRAIN_SET_FAILED,
+ AFR_MSG_SPLIT_BRAIN_DETERMINE_FAILED, AFR_MSG_HEALER_SPAWN_FAILED,
+ AFR_MSG_ADD_CRAWL_EVENT_FAILED, AFR_MSG_NULL_DEREF, AFR_MSG_SET_PEND_XATTR,
+ AFR_MSG_INTERNAL_ATTR);
+#define AFR_MSG_DICT_GET_FAILED_STR "Dict get failed"
+#define AFR_MSG_DICT_SET_FAILED_STR "Dict set failed"
+#define AFR_MSG_HEALER_SPAWN_FAILED_STR "Healer spawn failed"
+#define AFR_MSG_ADD_CRAWL_EVENT_FAILED_STR "Adding crawl event failed"
+#define AFR_MSG_INVALID_ARG_STR "Invalid argument"
+#define AFR_MSG_INDEX_DIR_GET_FAILED_STR "unable to get index-dir on "
+#define AFR_MSG_THIN_ARB_LOOKUP_FAILED_STR "Failed lookup on file"
+#define AFR_MSG_DICT_CREATE_FAILED_STR "Failed to create dict."
+#define AFR_MSG_THIN_ARB_XATTROP_FAILED_STR "Xattrop failed."
+#define AFR_MSG_THIN_ARB_LOC_POP_FAILED_STR \
+ "Failed to populate loc for thin-arbiter"
+#define AFR_MSG_GET_PEND_VAL_STR "Error getting value of pending"
+#define AFR_MSG_THIN_ARB_SKIP_SHD_STR "I am not the god shd. skipping."
+#define AFR_MSG_UNKNOWN_SET_STR "Unknown set"
+#define AFR_MSG_NO_XL_ID_STR "xl does not have id"
+#define AFR_MSG_SELF_HEAL_INFO_START_STR "starting full sweep on"
+#define AFR_MSG_SELF_HEAL_INFO_FINISH_STR "finished full sweep on"
+#define AFR_MSG_INCRE_COUNT_STR "Could not increment the counter."
+#define AFR_MSG_ADD_TO_OUTPUT_FAILED_STR "Could not add to output"
+#define AFR_MSG_SET_TIME_FAILED_STR "Could not set time"
+#define AFR_MSG_GFID_HEAL_MSG_STR "Error setting gfid-heal-msg dict"
+#define AFR_MSG_NO_MAJORITY_TO_RESOLVE_STR \
+ "No majority to resolve gfid split brain"
+#define AFR_MSG_GFID_MISMATCH_DETECTED_STR "Gfid mismatch dectected"
+#define AFR_MSG_SELF_HEAL_INFO_STR "performing selfheal"
+#define AFR_MSG_TYPE_MISMATCH_STR "TYPE mismatch"
+#define AFR_MSG_SIZE_POLICY_NOT_APPLICABLE_STR \
+ "Size policy is not applicable to directories."
+#define AFR_MSG_NO_CHILD_SELECTED_STR \
+ "No child selected by favorite-child policy"
+#define AFR_MSG_INVALID_CHILD_STR "Invalid child"
+#define AFR_MSG_RESOLVE_CONFLICTING_DATA_STR \
+ "selected as authentic to resolve conflicting data"
+#define SERROR_GETTING_SRC_BRICK_STR "Error getting the source brick"
+#define SNO_DIFF_IN_MTIME_STR "No difference in mtime"
+#define SNO_BIGGER_FILE_STR "No bigger file"
+#define SALL_BRICKS_UP_TO_RESOLVE_STR \
+ "All the bricks should be up to resolve the gfid split brain"
+#define AFR_MSG_UNLOCK_FAILED_STR "Failed to unlock"
+#define AFR_MSG_POST_OP_FAILED_STR "Post-op on thin-arbiter failed"
+#define AFR_MSG_TA_FRAME_CREATE_FAILED_STR "Failed to create ta_frame"
+#define AFR_MSG_SET_KEY_XATTROP_FAILED_STR "Could not set key during xattrop"
+#define AFR_MSG_BLOCKING_ENTRYLKS_FAILED_STR "Blocking entrylks failed"
+#define AFR_MSG_FSYNC_FAILED_STR "fsync failed"
+#define AFR_MSG_QUORUM_FAIL_STR "quorum is not met"
+#define AFR_MSG_FOP_FAILED_STR "Failing Fop"
+#define AFR_MSG_INVALID_SUBVOL_STR "not a subvolume"
+#define AFR_MSG_VOL_MISCONFIGURED_STR "Volume is dangling"
+#define AFR_MSG_CHILD_MISCONFIGURED_STR \
+ "replicate translator needs more than one subvolume defined"
+#define AFR_MSG_CLEAN_UP_FAILED_STR "Failed to clean up healer threads"
+#define AFR_MSG_QUORUM_OVERRIDE_STR "overriding quorum-count"
+#define AFR_MSG_UNABLE_TO_FETCH_STR \
+ "Unable to fetch afr-pending-xattr option from volfile. Falling back to " \
+ "using client translator names"
+#define AFR_MSG_NULL_DEREF_STR "possible NULL deref"
+#define AFR_MSG_XATTR_SET_FAILED_STR "Cannot set xattr cookie key"
+#define AFR_MSG_SPLIT_BRAIN_STATUS_STR "Failed to create synctask"
+#define AFR_MSG_SUBVOLS_DOWN_STR "All subvolumes are not up"
+#define AFR_MSG_SPLIT_BRAIN_CHOICE_ERROR_STR \
+ "Failed to cancel split-brain choice"
+#define AFR_MSG_SPLIT_BRAIN_REPLICA_STR \
+ "Cannot set replica. File is not in data/metadata split-brain"
+#define AFR_MSG_INODE_CTX_FAILED_STR "Failed to get inode_ctx"
+#define AFR_MSG_READ_SUBVOL_ERROR_STR "no read subvols"
+#define AFR_MSG_LOCAL_CHILD_STR "selecting local read-child"
+#define AFR_MSG_LOOKUP_FAILED_STR "Failed to lookup/create thin-arbiter id file"
+#define AFR_MSG_TIMER_CREATE_FAIL_STR \
+ "Cannot create timer for delayed initialization"
+#define AFR_MSG_SUBVOL_UP_STR "Subvolume came back up; going online"
+#define AFR_MSG_ALL_SUBVOLS_DOWN_STR \
+ "All subvolumes are down. Going offline until atleast one of them is up"
+#define AFR_MSG_RELEASE_LOCK_FAILED_STR "Failed to release lock"
+#define AFR_MSG_INVALID_CHILD_UP_STR "Received child_up from invalid subvolume"
+#define AFR_MSG_QUORUM_MET_STR "Client-quorum is met"
+#define AFR_MSG_EXPUNGING_FILE_OR_DIR_STR "expunging file or dir"
+#define AFR_MSG_SELF_HEAL_FAILED_STR "Invalid"
+#define AFR_MSG_SPLIT_BRAIN_STR "Skipping conservative mergeon the file"
+#define AFR_MSG_CLEAR_TIME_SPLIT_BRAIN_STR "clear time split brain"
+#define AFR_MSG_READ_FAILED_STR "Failing read since good brick is down"
+#define AFR_MSG_LAUNCH_FAILED_STR "Failed to launch synctask"
+#define AFR_MSG_READ_SUBVOL_NOT_UP_STR \
+ "read subvolume in this generation is not up"
+#define AFR_MSG_INTERNAL_LKS_FAILED_STR \
+ "Unable to work with lk-owner while attempting fop"
+#define AFR_MSG_LOCK_XLATOR_NOT_LOADED_STR \
+ "subvolume does not support locking. please load features/locks xlator " \
+ "on server."
+#define AFR_MSG_FD_CTX_GET_FAILED_STR "unable to get fd ctx"
+#define AFR_MSG_INFO_COMMON_STR "fd not open on any subvolumes, aborting."
+#define AFR_MSG_REPLACE_BRICK_STATUS_STR "Couldn't acquire lock on any child."
+#define AFR_MSG_NEW_BRICK_STR "New brick"
+#define AFR_MSG_SPLIT_BRAIN_SET_FAILED_STR \
+ "Failed to set split-brain choice to -1"
+#define AFR_MSG_SPLIT_BRAIN_DETERMINE_FAILED_STR \
+ "Failed to determine split-brain. Aborting split-brain-choice set"
+#define AFR_MSG_OPEN_FAIL_STR "Failed to open subvolume"
+#define AFR_MSG_SET_PEND_XATTR_STR "Set of pending xattr"
+#define AFR_MSG_INTERNAL_ATTR_STR "is an internal extended attribute"
#endif /* !_AFR_MESSAGES_H_ */
diff --git a/xlators/cluster/afr/src/afr-open.c b/xlators/cluster/afr/src/afr-open.c
index de125296bb3..64856042b65 100644
--- a/xlators/cluster/afr/src/afr-open.c
+++ b/xlators/cluster/afr/src/afr-open.c
@@ -8,364 +8,346 @@
cases as published by the Free Software Foundation.
*/
-#include <libgen.h>
#include <unistd.h>
-#include <fnmatch.h>
#include <sys/time.h>
#include <stdlib.h>
#include <signal.h>
-#include "glusterfs.h"
+#include <glusterfs/glusterfs.h>
#include "afr.h"
-#include "dict.h"
-#include "xlator.h"
-#include "hashfn.h"
-#include "logging.h"
-#include "list.h"
-#include "call-stub.h"
-#include "defaults.h"
-#include "common-utils.h"
-#include "compat-errno.h"
-#include "compat.h"
-#include "byte-order.h"
-#include "statedump.h"
-
-#include "afr-inode-read.h"
-#include "afr-inode-write.h"
-#include "afr-dir-read.h"
-#include "afr-dir-write.h"
-#include "afr-transaction.h"
+#include <glusterfs/dict.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/defaults.h>
+#include <glusterfs/common-utils.h>
+#include <glusterfs/compat-errno.h>
+#include <glusterfs/compat.h>
+#include <glusterfs/byte-order.h>
+#include <glusterfs/statedump.h>
+#include "afr-transaction.h"
gf_boolean_t
-afr_is_fd_fixable (fd_t *fd)
+afr_is_fd_fixable(fd_t *fd)
{
- if (!fd || !fd->inode)
- return _gf_false;
- else if (fd_is_anonymous (fd))
- return _gf_false;
- else if (gf_uuid_is_null (fd->inode->gfid))
- return _gf_false;
-
- return _gf_true;
+ if (!fd || !fd->inode)
+ return _gf_false;
+ else if (fd_is_anonymous(fd))
+ return _gf_false;
+ else if (gf_uuid_is_null(fd->inode->gfid))
+ return _gf_false;
+
+ return _gf_true;
}
-
int
-afr_open_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+afr_open_ftruncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- afr_local_t * local = frame->local;
+ afr_local_t *local = frame->local;
- AFR_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
- local->fd, xdata);
- return 0;
+ AFR_STACK_UNWIND(open, frame, local->op_ret, local->op_errno,
+ local->cont.open.fd, xdata);
+ return 0;
}
int
-afr_open_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- fd_t *fd, dict_t *xdata)
+afr_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int call_count = -1;
- int child_index = (long) cookie;
- afr_fd_ctx_t *fd_ctx = NULL;
-
- local = frame->local;
- fd_ctx = local->fd_ctx;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
- } else {
- local->op_ret = op_ret;
- fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
- if (!local->xdata_rsp && xdata)
- local->xdata_rsp = dict_ref (xdata);
- }
+ afr_local_t *local = NULL;
+ int call_count = -1;
+ int child_index = (long)cookie;
+ afr_fd_ctx_t *fd_ctx = NULL;
+
+ local = frame->local;
+ fd_ctx = local->fd_ctx;
+
+ local->replies[child_index].valid = 1;
+ local->replies[child_index].op_ret = op_ret;
+ local->replies[child_index].op_errno = op_errno;
+
+ LOCK(&frame->lock);
+ {
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
+ } else {
+ local->op_ret = op_ret;
+ fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
+ if (!local->xdata_rsp && xdata)
+ local->xdata_rsp = dict_ref(xdata);
}
- UNLOCK (&frame->lock);
-
- call_count = afr_frame_return (frame);
-
- if (call_count == 0) {
- if ((fd_ctx->flags & O_TRUNC) && (local->op_ret >= 0)) {
- STACK_WIND (frame, afr_open_ftruncate_cbk,
- this, this->fops->ftruncate,
- fd, 0, NULL);
- } else {
- AFR_STACK_UNWIND (open, frame, local->op_ret,
- local->op_errno, local->cont.open.fd,
- local->xdata_rsp);
- }
+ call_count = --local->call_count;
+ }
+ UNLOCK(&frame->lock);
+
+ if (call_count == 0) {
+ afr_handle_replies_quorum(frame, this);
+ if (local->op_ret == -1) {
+ AFR_STACK_UNWIND(open, frame, local->op_ret, local->op_errno, NULL,
+ NULL);
+ } else if (fd_ctx->flags & O_TRUNC) {
+ STACK_WIND(frame, afr_open_ftruncate_cbk, this,
+ this->fops->ftruncate, fd, 0, NULL);
+ } else {
+ AFR_STACK_UNWIND(open, frame, local->op_ret, local->op_errno,
+ local->cont.open.fd, local->xdata_rsp);
}
+ }
- return 0;
+ return 0;
}
-
int
-afr_open_continue (call_frame_t *frame, xlator_t *this, int err)
+afr_open_continue(call_frame_t *frame, xlator_t *this, int err)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = 0;
- int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int i = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- if (err) {
- AFR_STACK_UNWIND (open, frame, -1, -err, NULL, NULL);
- } else {
- local->call_count = AFR_COUNT (local->child_up,
- priv->child_count);
- call_count = local->call_count;
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- STACK_WIND_COOKIE (frame, afr_open_cbk,
- (void *)(long)i,
- priv->children[i],
- priv->children[i]->fops->open,
- &local->loc,
- (local->cont.open.flags & ~O_TRUNC),
- local->cont.open.fd,
- local->xdata_req);
- if (!--call_count)
- break;
- }
- }
+ if (err) {
+ AFR_STACK_UNWIND(open, frame, -1, err, NULL, NULL);
+ } else {
+ local->call_count = AFR_COUNT(local->child_up, priv->child_count);
+ call_count = local->call_count;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->child_up[i]) {
+ STACK_WIND_COOKIE(frame, afr_open_cbk, (void *)(long)i,
+ priv->children[i],
+ priv->children[i]->fops->open, &local->loc,
+ (local->cont.open.flags & ~O_TRUNC),
+ local->cont.open.fd, local->xdata_req);
+ if (!--call_count)
+ break;
+ }
}
- return 0;
+ }
+ return 0;
}
int
-afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- fd_t *fd, dict_t *xdata)
+afr_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int spb_choice = 0;
- int event_generation = 0;
- int ret = 0;
- int32_t op_errno = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
-
- //We can't let truncation to happen outside transaction.
-
- priv = this->private;
-
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
-
- local->op = GF_FOP_OPEN;
- fd_ctx = afr_fd_ctx_get (fd, this);
- if (!fd_ctx) {
- op_errno = ENOMEM;
- goto out;
- }
-
- if (!afr_is_consistent_io_possible (local, priv, &op_errno))
- goto out;
-
- local->inode = inode_ref (loc->inode);
- loc_copy (&local->loc, loc);
- local->fd_ctx = fd_ctx;
- fd_ctx->flags = flags;
- if (xdata)
- local->xdata_req = dict_ref (xdata);
-
- local->cont.open.flags = flags;
- local->cont.open.fd = fd_ref (fd);
-
- ret = afr_inode_get_readable (frame, local->inode, this,
- NULL, &event_generation,
- AFR_DATA_TRANSACTION);
- if ((ret < 0) &&
- (afr_inode_split_brain_choice_get (local->inode,
- this, &spb_choice) == 0) &&
- spb_choice < 0) {
- afr_inode_refresh (frame, this, local->inode,
- local->inode->gfid, afr_open_continue);
- } else {
- afr_open_continue (frame, this, 0);
- }
-
- return 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int spb_subvol = 0;
+ int event_generation = 0;
+ int ret = 0;
+ int32_t op_errno = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+
+ // We can't let truncation to happen outside transaction.
+
+ priv = this->private;
+
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
+
+ local->op = GF_FOP_OPEN;
+ fd_ctx = afr_fd_ctx_get(fd, this);
+ if (!fd_ctx) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ if (priv->quorum_count && !afr_has_quorum(local->child_up, this, NULL)) {
+ op_errno = afr_quorum_errno(priv);
+ goto out;
+ }
+
+ if (!afr_is_consistent_io_possible(local, priv, &op_errno))
+ goto out;
+
+ local->inode = inode_ref(loc->inode);
+ loc_copy(&local->loc, loc);
+ local->fd_ctx = fd_ctx;
+ fd_ctx->flags = flags;
+ if (xdata)
+ local->xdata_req = dict_ref(xdata);
+
+ local->cont.open.flags = flags;
+ local->cont.open.fd = fd_ref(fd);
+
+ ret = afr_inode_get_readable(frame, local->inode, this, NULL,
+ &event_generation, AFR_DATA_TRANSACTION);
+ if ((ret < 0) &&
+ (afr_split_brain_read_subvol_get(local->inode, this, NULL,
+ &spb_subvol) == 0) &&
+ spb_subvol < 0) {
+ afr_inode_refresh(frame, this, local->inode, local->inode->gfid,
+ afr_open_continue);
+ } else {
+ afr_open_continue(frame, this, 0);
+ }
+
+ return 0;
out:
- AFR_STACK_UNWIND (open, frame, -1, op_errno, fd, NULL);
+ AFR_STACK_UNWIND(open, frame, -1, op_errno, fd, NULL);
- return 0;
+ return 0;
}
int
-afr_openfd_fix_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd,
- dict_t *xdata)
+afr_openfd_fix_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
- int call_count = 0;
- int child_index = (long) cookie;
-
- priv = this->private;
- local = frame->local;
-
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int call_count = 0;
+ int child_index = (long)cookie;
+
+ priv = this->private;
+ local = frame->local;
+
+ if (op_ret >= 0) {
+ gf_msg_debug(this->name, 0,
+ "fd for %s opened "
+ "successfully on subvolume %s",
+ local->loc.path, priv->children[child_index]->name);
+ } else {
+ gf_smsg(this->name, fop_log_level(GF_FOP_OPEN, op_errno), op_errno,
+ AFR_MSG_OPEN_FAIL, "path=%s", local->loc.path, "subvolume=%s",
+ priv->children[child_index]->name, NULL);
+ }
+
+ fd_ctx = local->fd_ctx;
+
+ LOCK(&local->fd->lock);
+ {
if (op_ret >= 0) {
- gf_msg_debug (this->name, 0, "fd for %s opened "
- "successfully on subvolume %s", local->loc.path,
- priv->children[child_index]->name);
+ fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
} else {
- gf_msg (this->name, fop_log_level (GF_FOP_OPEN, op_errno),
- op_errno, AFR_MSG_OPEN_FAIL, "Failed to open %s on "
- "subvolume %s", local->loc.path,
- priv->children[child_index]->name);
+ fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
}
+ }
+ UNLOCK(&local->fd->lock);
- fd_ctx = local->fd_ctx;
+ call_count = afr_frame_return(frame);
+ if (call_count == 0)
+ AFR_STACK_DESTROY(frame);
- LOCK (&local->fd->lock);
- {
- if (op_ret >= 0) {
- fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
- } else {
- fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
- }
- }
- UNLOCK (&local->fd->lock);
+ return 0;
+}
- call_count = afr_frame_return (frame);
- if (call_count == 0)
- AFR_STACK_DESTROY (frame);
+static int
+afr_fd_ctx_need_open(fd_t *fd, xlator_t *this, unsigned char *need_open)
+{
+ afr_fd_ctx_t *fd_ctx = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int count = 0;
+ priv = this->private;
+
+ fd_ctx = afr_fd_ctx_get(fd, this);
+ if (!fd_ctx)
return 0;
-}
+ LOCK(&fd->lock);
+ {
+ for (i = 0; i < priv->child_count; i++) {
+ if (fd_ctx->opened_on[i] == AFR_FD_NOT_OPENED &&
+ priv->child_up[i]) {
+ fd_ctx->opened_on[i] = AFR_FD_OPENING;
+ need_open[i] = 1;
+ count++;
+ } else {
+ need_open[i] = 0;
+ }
+ }
+ }
+ UNLOCK(&fd->lock);
-static int
-afr_fd_ctx_need_open (fd_t *fd, xlator_t *this, unsigned char *need_open)
-{
- afr_fd_ctx_t *fd_ctx = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- int count = 0;
-
- priv = this->private;
-
- fd_ctx = afr_fd_ctx_get (fd, this);
- if (!fd_ctx)
- return 0;
-
- LOCK (&fd->lock);
- {
- for (i = 0; i < priv->child_count; i++) {
- if (fd_ctx->opened_on[i] == AFR_FD_NOT_OPENED &&
- priv->child_up[i]) {
- fd_ctx->opened_on[i] = AFR_FD_OPENING;
- need_open[i] = 1;
- count++;
- } else {
- need_open[i] = 0;
- }
- }
- }
- UNLOCK (&fd->lock);
-
- return count;
+ return count;
}
-
void
-afr_fix_open (fd_t *fd, xlator_t *this)
+afr_fix_open(fd_t *fd, xlator_t *this)
{
- afr_private_t *priv = NULL;
- int i = 0;
- call_frame_t *frame = NULL;
- afr_local_t *local = NULL;
- int ret = -1;
- int32_t op_errno = 0;
- afr_fd_ctx_t *fd_ctx = NULL;
- unsigned char *need_open = NULL;
- int call_count = 0;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
+ int ret = -1;
+ int32_t op_errno = 0;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ unsigned char *need_open = NULL;
+ int call_count = 0;
- priv = this->private;
+ priv = this->private;
- if (!afr_is_fd_fixable (fd))
- goto out;
+ if (!afr_is_fd_fixable(fd))
+ goto out;
- fd_ctx = afr_fd_ctx_get (fd, this);
- if (!fd_ctx)
- goto out;
+ fd_ctx = afr_fd_ctx_get(fd, this);
+ if (!fd_ctx)
+ goto out;
- need_open = alloca0 (priv->child_count);
+ need_open = alloca0(priv->child_count);
- call_count = afr_fd_ctx_need_open (fd, this, need_open);
- if (!call_count)
- goto out;
+ call_count = afr_fd_ctx_need_open(fd, this, need_open);
+ if (!call_count)
+ goto out;
- frame = create_frame (this, this->ctx->pool);
- if (!frame)
- goto out;
+ frame = create_frame(this, this->ctx->pool);
+ if (!frame)
+ goto out;
- local = AFR_FRAME_INIT (frame, op_errno);
- if (!local)
- goto out;
+ local = AFR_FRAME_INIT(frame, op_errno);
+ if (!local)
+ goto out;
- local->loc.inode = inode_ref (fd->inode);
- ret = loc_path (&local->loc, NULL);
- if (ret < 0)
- goto out;
+ local->loc.inode = inode_ref(fd->inode);
+ ret = loc_path(&local->loc, NULL);
+ if (ret < 0)
+ goto out;
- local->fd = fd_ref (fd);
- local->fd_ctx = fd_ctx;
+ local->fd = fd_ref(fd);
+ local->fd_ctx = fd_ctx;
- local->call_count = call_count;
+ local->call_count = call_count;
- gf_msg_debug (this->name, 0, "need open count: %d",
- call_count);
+ gf_msg_debug(this->name, 0, "need open count: %d", call_count);
- for (i = 0; i < priv->child_count; i++) {
- if (!need_open[i])
- continue;
-
- if (IA_IFDIR == fd->inode->ia_type) {
- gf_msg_debug (this->name, 0,
- "opening fd for dir %s on subvolume %s",
- local->loc.path, priv->children[i]->name);
-
- STACK_WIND_COOKIE (frame, afr_openfd_fix_open_cbk,
- (void*) (long) i,
- priv->children[i],
- priv->children[i]->fops->opendir,
- &local->loc, local->fd,
- NULL);
- } else {
- gf_msg_debug (this->name, 0,
- "opening fd for file %s on subvolume %s",
- local->loc.path, priv->children[i]->name);
-
- STACK_WIND_COOKIE (frame, afr_openfd_fix_open_cbk,
- (void *)(long) i,
- priv->children[i],
- priv->children[i]->fops->open,
- &local->loc,
- fd_ctx->flags & (~O_TRUNC),
- local->fd, NULL);
- }
-
- if (!--call_count)
- break;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!need_open[i])
+ continue;
+
+ if (IA_IFDIR == fd->inode->ia_type) {
+ gf_msg_debug(this->name, 0, "opening fd for dir %s on subvolume %s",
+ local->loc.path, priv->children[i]->name);
+
+ STACK_WIND_COOKIE(frame, afr_openfd_fix_open_cbk, (void *)(long)i,
+ priv->children[i],
+ priv->children[i]->fops->opendir, &local->loc,
+ local->fd, NULL);
+ } else {
+ gf_msg_debug(this->name, 0,
+ "opening fd for file %s on subvolume %s",
+ local->loc.path, priv->children[i]->name);
+
+ STACK_WIND_COOKIE(frame, afr_openfd_fix_open_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->open,
+ &local->loc, fd_ctx->flags & (~O_TRUNC),
+ local->fd, NULL);
}
- return;
+ if (!--call_count)
+ break;
+ }
+
+ return;
out:
- if (frame)
- AFR_STACK_DESTROY (frame);
+ if (frame)
+ AFR_STACK_DESTROY(frame);
}
diff --git a/xlators/cluster/afr/src/afr-read-txn.c b/xlators/cluster/afr/src/afr-read-txn.c
index a8a4090efd1..6fc2c75145c 100644
--- a/xlators/cluster/afr/src/afr-read-txn.c
+++ b/xlators/cluster/afr/src/afr-read-txn.c
@@ -13,140 +13,327 @@
#include "afr-messages.h"
void
-afr_pending_read_increment (afr_private_t *priv, int child_index)
+afr_pending_read_increment(afr_private_t *priv, int child_index)
{
- if (child_index < 0 || child_index > priv->child_count)
- return;
+ if (child_index < 0 || child_index > priv->child_count)
+ return;
- GF_ATOMIC_INC(priv->pending_reads[child_index]);
+ GF_ATOMIC_INC(priv->pending_reads[child_index]);
}
void
-afr_pending_read_decrement (afr_private_t *priv, int child_index)
+afr_pending_read_decrement(afr_private_t *priv, int child_index)
{
- if (child_index < 0 || child_index > priv->child_count)
- return;
+ if (child_index < 0 || child_index > priv->child_count)
+ return;
- GF_ATOMIC_DEC(priv->pending_reads[child_index]);
+ GF_ATOMIC_DEC(priv->pending_reads[child_index]);
}
void
-afr_read_txn_wind (call_frame_t *frame, xlator_t *this, int subvol)
+afr_read_txn_wind(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- afr_pending_read_decrement (priv, local->read_subvol);
- local->read_subvol = subvol;
- afr_pending_read_increment (priv, subvol);
- local->readfn (frame, this, subvol);
+ afr_pending_read_decrement(priv, local->read_subvol);
+ local->read_subvol = subvol;
+ afr_pending_read_increment(priv, subvol);
+ local->readfn(frame, this, subvol);
}
int
-afr_read_txn_next_subvol (call_frame_t *frame, xlator_t *this)
+afr_read_txn_next_subvol(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- int subvol = -1;
-
- local = frame->local;
- priv = this->private;
-
-
- for (i = 0; i < priv->child_count; i++) {
- if (!local->readable[i]) {
- /* don't even bother trying here.
- just mark as attempted and move on. */
- local->read_attempted[i] = 1;
- continue;
- }
-
- if (!local->read_attempted[i]) {
- subvol = i;
- break;
- }
- }
-
- /* If no more subvols were available for reading, we leave
- @subvol as -1, which is an indication we have run out of
- readable subvols. */
- if (subvol != -1)
- local->read_attempted[subvol] = 1;
- afr_read_txn_wind (frame, this, subvol);
-
- return 0;
-}
-
-int
-afr_read_txn_refresh_done (call_frame_t *frame, xlator_t *this, int err)
-{
- afr_local_t *local = NULL;
- int read_subvol = 0;
- inode_t *inode = NULL;
- int ret = -1;
- int spb_choice = -1;
-
- local = frame->local;
- inode = local->inode;
-
- if (err) {
- read_subvol = -1;
- goto readfn;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int subvol = -1;
+
+ local = frame->local;
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->readable[i]) {
+ /* don't even bother trying here.
+ just mark as attempted and move on. */
+ local->read_attempted[i] = 1;
+ continue;
}
- read_subvol = afr_read_subvol_select_by_policy (inode, this,
- local->readable, NULL);
- if (read_subvol == -1) {
- err = -EIO;
- goto readfn;
+ if (!local->read_attempted[i]) {
+ subvol = i;
+ break;
}
+ }
- if (local->read_attempted[read_subvol]) {
- afr_read_txn_next_subvol (frame, this);
- return 0;
- }
+ /* If no more subvols were available for reading, we leave
+ @subvol as -1, which is an indication we have run out of
+ readable subvols. */
+ if (subvol != -1)
+ local->read_attempted[subvol] = 1;
+ afr_read_txn_wind(frame, this, subvol);
- local->read_attempted[read_subvol] = 1;
-readfn:
- if (read_subvol == -1) {
- ret = afr_inode_split_brain_choice_get (inode, this,
- &spb_choice);
- if ((ret == 0) && spb_choice >= 0)
- read_subvol = spb_choice;
- }
+ return 0;
+}
- if (read_subvol == -1) {
- AFR_SET_ERROR_AND_CHECK_SPLIT_BRAIN (-1, -err);
- }
- afr_read_txn_wind (frame, this, read_subvol);
+static int
+afr_ta_read_txn_done(int ret, call_frame_t *ta_frame, void *opaque)
+{
+ STACK_DESTROY(ta_frame->root);
+ return 0;
+}
+
+static int
+afr_ta_read_txn(void *opaque)
+{
+ call_frame_t *frame = NULL;
+ xlator_t *this = NULL;
+ int read_subvol = -1;
+ int query_child = AFR_CHILD_UNKNOWN;
+ int possible_bad_child = AFR_CHILD_UNKNOWN;
+ int ret = 0;
+ int op_errno = ENOMEM;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ struct gf_flock flock = {
+ 0,
+ };
+ dict_t *xdata_req = NULL;
+ dict_t *xdata_rsp = NULL;
+ int **pending = NULL;
+ loc_t loc = {
+ 0,
+ };
+
+ frame = (call_frame_t *)opaque;
+ this = frame->this;
+ local = frame->local;
+ priv = this->private;
+ query_child = local->read_txn_query_child;
+
+ if (query_child == AFR_CHILD_ZERO) {
+ possible_bad_child = AFR_CHILD_ONE;
+ } else if (query_child == AFR_CHILD_ONE) {
+ possible_bad_child = AFR_CHILD_ZERO;
+ } else {
+ /*read_txn_query_child is AFR_CHILD_UNKNOWN*/
+ goto out;
+ }
+
+ /* Ask the query_child to see if it blames the possibly bad one. */
+ xdata_req = dict_new();
+ if (!xdata_req)
+ goto out;
+
+ pending = afr_matrix_create(priv->child_count, AFR_NUM_CHANGE_LOGS);
+ if (!pending)
+ goto out;
+
+ ret = afr_set_pending_dict(priv, xdata_req, pending);
+ if (ret < 0)
+ goto out;
+
+ if (local->fd) {
+ ret = syncop_fxattrop(priv->children[query_child], local->fd,
+ GF_XATTROP_ADD_ARRAY, xdata_req, NULL, &xdata_rsp,
+ NULL);
+ } else {
+ ret = syncop_xattrop(priv->children[query_child], &local->loc,
+ GF_XATTROP_ADD_ARRAY, xdata_req, NULL, &xdata_rsp,
+ NULL);
+ }
+ if (ret || !xdata_rsp) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed xattrop for gfid %s on %s",
+ uuid_utoa(local->inode->gfid),
+ priv->children[query_child]->name);
+ op_errno = -ret;
+ goto out;
+ }
+
+ if (afr_ta_dict_contains_pending_xattr(xdata_rsp, priv,
+ possible_bad_child)) {
+ read_subvol = query_child;
+ goto out;
+ }
+ dict_unref(xdata_rsp);
+ xdata_rsp = NULL;
+
+ /* It doesn't. So query thin-arbiter to see if it blames any data brick. */
+ ret = afr_fill_ta_loc(this, &loc, _gf_true);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to populate thin-arbiter loc for: %s.", loc.name);
+ goto out;
+ }
+ flock.l_type = F_WRLCK; /*start and length are already zero. */
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_MODIFY, &loc, F_SETLKW, &flock, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "gfid:%s: Failed to get AFR_TA_DOM_MODIFY lock on %s.",
+ uuid_utoa(local->inode->gfid),
+ priv->pending_key[THIN_ARBITER_BRICK_INDEX]);
+ op_errno = -ret;
+ goto out;
+ }
+
+ ret = syncop_xattrop(priv->children[THIN_ARBITER_BRICK_INDEX], &loc,
+ GF_XATTROP_ADD_ARRAY, xdata_req, NULL, &xdata_rsp,
+ NULL);
+ if (ret || !xdata_rsp) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "gfid:%s: Failed xattrop on %s.", uuid_utoa(local->inode->gfid),
+ priv->pending_key[THIN_ARBITER_BRICK_INDEX]);
+ op_errno = -ret;
+ goto unlock;
+ }
+
+ if (!afr_ta_dict_contains_pending_xattr(xdata_rsp, priv, query_child)) {
+ read_subvol = query_child;
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, EIO, AFR_MSG_THIN_ARB,
+ "Failing read for gfid %s since good brick %s is down",
+ uuid_utoa(local->inode->gfid),
+ priv->children[possible_bad_child]->name);
+ op_errno = EIO;
+ }
+
+unlock:
+ flock.l_type = F_UNLCK;
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_MODIFY, &loc, F_SETLK, &flock, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "gfid:%s: Failed to unlock AFR_TA_DOM_MODIFY lock on "
+ "%s.",
+ uuid_utoa(local->inode->gfid),
+ priv->pending_key[THIN_ARBITER_BRICK_INDEX]);
+ }
+out:
+ if (xdata_req)
+ dict_unref(xdata_req);
+ if (xdata_rsp)
+ dict_unref(xdata_rsp);
+ if (pending)
+ afr_matrix_cleanup(pending, priv->child_count);
+ loc_wipe(&loc);
+
+ if (read_subvol == -1) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ }
+ afr_read_txn_wind(frame, this, read_subvol);
+ return ret;
+}
- return 0;
+void
+afr_ta_read_txn_synctask(call_frame_t *frame, xlator_t *this)
+{
+ call_frame_t *ta_frame = NULL;
+ afr_local_t *local = NULL;
+ int ret = 0;
+
+ local = frame->local;
+ ta_frame = afr_ta_frame_create(this);
+ if (!ta_frame) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, AFR_MSG_THIN_ARB,
+ "Failed to create ta_frame");
+ goto out;
+ }
+ ret = synctask_new(this->ctx->env, afr_ta_read_txn, afr_ta_read_txn_done,
+ ta_frame, frame);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, AFR_MSG_THIN_ARB,
+ "Failed to launch "
+ "afr_ta_read_txn synctask for gfid %s.",
+ uuid_utoa(local->inode->gfid));
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ STACK_DESTROY(ta_frame->root);
+ goto out;
+ }
+ return;
+out:
+ afr_read_txn_wind(frame, this, -1);
}
+int
+afr_read_txn_refresh_done(call_frame_t *frame, xlator_t *this, int err)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int read_subvol = -1;
+ inode_t *inode = NULL;
+ int ret = -1;
+ int spb_subvol = -1;
+
+ local = frame->local;
+ inode = local->inode;
+ priv = this->private;
+
+ if (err) {
+ if (!priv->thin_arbiter_count)
+ goto readfn;
+ if (err != EINVAL)
+ goto readfn;
+ /* We need to query the good bricks and/or thin-arbiter.*/
+ afr_ta_read_txn_synctask(frame, this);
+ return 0;
+ }
+
+ read_subvol = afr_read_subvol_select_by_policy(inode, this, local->readable,
+ NULL);
+ if (read_subvol == -1) {
+ err = EIO;
+ goto readfn;
+ }
+
+ if (local->read_attempted[read_subvol]) {
+ afr_read_txn_next_subvol(frame, this);
+ return 0;
+ }
+
+ local->read_attempted[read_subvol] = 1;
+readfn:
+ if (read_subvol == -1) {
+ ret = afr_split_brain_read_subvol_get(inode, this, frame, &spb_subvol);
+ if ((ret == 0) && spb_subvol >= 0)
+ read_subvol = spb_subvol;
+ }
+
+ if (read_subvol == -1) {
+ AFR_SET_ERROR_AND_CHECK_SPLIT_BRAIN(-1, err);
+ }
+ afr_read_txn_wind(frame, this, read_subvol);
+
+ return 0;
+}
int
-afr_read_txn_continue (call_frame_t *frame, xlator_t *this, int subvol)
+afr_read_txn_continue(call_frame_t *frame, xlator_t *this, int subvol)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (!local->refreshed) {
- local->refreshed = _gf_true;
- afr_inode_refresh (frame, this, local->inode, NULL,
- afr_read_txn_refresh_done);
- } else {
- afr_read_txn_next_subvol (frame, this);
- }
+ if (!local->refreshed) {
+ local->refreshed = _gf_true;
+ afr_inode_refresh(frame, this, local->inode, NULL,
+ afr_read_txn_refresh_done);
+ } else {
+ afr_read_txn_next_subvol(frame, this);
+ }
- return 0;
+ return 0;
}
-
/* afr_read_txn_wipe:
clean internal variables in @local in order to make
@@ -155,27 +342,26 @@ afr_read_txn_continue (call_frame_t *frame, xlator_t *this, int subvol)
*/
void
-afr_read_txn_wipe (call_frame_t *frame, xlator_t *this)
+afr_read_txn_wipe(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- local->readfn = NULL;
+ local->readfn = NULL;
- if (local->inode)
- inode_unref (local->inode);
+ if (local->inode)
+ inode_unref(local->inode);
- for (i = 0; i < priv->child_count; i++) {
- local->read_attempted[i] = 0;
- local->readable[i] = 0;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ local->read_attempted[i] = 0;
+ local->readable[i] = 0;
+ }
}
-
/*
afr_read_txn:
@@ -204,87 +390,105 @@ afr_read_txn_wipe (call_frame_t *frame, xlator_t *this)
*/
int
-afr_read_txn (call_frame_t *frame, xlator_t *this, inode_t *inode,
- afr_read_txn_wind_t readfn, afr_transaction_type type)
+afr_read_txn(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ afr_read_txn_wind_t readfn, afr_transaction_type type)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- unsigned char *data = NULL;
- unsigned char *metadata = NULL;
- int read_subvol = -1;
- int event_generation = 0;
- int ret = -1;
-
- priv = this->private;
- local = frame->local;
- data = alloca0 (priv->child_count);
- metadata = alloca0 (priv->child_count);
-
- afr_read_txn_wipe (frame, this);
-
- local->readfn = readfn;
- local->inode = inode_ref (inode);
- local->is_read_txn = _gf_true;
-
- if (priv->quorum_count && !afr_has_quorum (local->child_up, this)) {
- local->op_ret = -1;
- local->op_errno = ENOTCONN;
- read_subvol = -1;
- goto read;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ unsigned char *data = NULL;
+ unsigned char *metadata = NULL;
+ int read_subvol = -1;
+ int event_generation = 0;
+ int ret = -1;
+
+ priv = this->private;
+ local = frame->local;
+ data = alloca0(priv->child_count);
+ metadata = alloca0(priv->child_count);
+
+ afr_read_txn_wipe(frame, this);
+
+ local->readfn = readfn;
+ local->inode = inode_ref(inode);
+ local->is_read_txn = _gf_true;
+ local->transaction.type = type;
+
+ if (priv->quorum_count && !afr_has_quorum(local->child_up, this, NULL)) {
+ local->op_ret = -1;
+ local->op_errno = afr_quorum_errno(priv);
+ goto read;
+ }
+
+ if (!afr_is_consistent_io_possible(local, priv, &local->op_errno)) {
+ local->op_ret = -1;
+ goto read;
+ }
+
+ if (priv->thin_arbiter_count && !afr_ta_has_quorum(priv, local)) {
+ local->op_ret = -1;
+ local->op_errno = -afr_quorum_errno(priv);
+ goto read;
+ }
+
+ if (priv->thin_arbiter_count &&
+ AFR_COUNT(local->child_up, priv->child_count) != priv->child_count) {
+ if (local->child_up[0]) {
+ local->read_txn_query_child = AFR_CHILD_ZERO;
+ } else if (local->child_up[1]) {
+ local->read_txn_query_child = AFR_CHILD_ONE;
}
-
- if (!afr_is_consistent_io_possible (local, priv, &local->op_errno)) {
- local->op_ret = -1;
- read_subvol = -1;
- goto read;
- }
-
- local->transaction.type = type;
- ret = afr_inode_read_subvol_get (inode, this, data, metadata,
- &event_generation);
- if (ret == -1)
- /* very first transaction on this inode */
- goto refresh;
- AFR_INTERSECT (local->readable, data, metadata, priv->child_count);
-
- gf_msg_debug (this->name, 0, "%s: generation now vs cached: %d, "
- "%d", uuid_utoa (inode->gfid), local->event_generation,
- event_generation);
- if (afr_is_inode_refresh_reqd (inode, this, local->event_generation,
- event_generation))
- /* servers have disconnected / reconnected, and possibly
- rebooted, very likely changing the state of freshness
- of copies */
- goto refresh;
-
- read_subvol = afr_read_subvol_select_by_policy (inode, this,
- local->readable, NULL);
-
- if (read_subvol < 0 || read_subvol > priv->child_count) {
- gf_msg_debug (this->name, 0, "Unreadable subvolume %d found "
- "with event generation %d for gfid %s.",
- read_subvol, event_generation, uuid_utoa(inode->gfid));
- goto refresh;
- }
-
- if (!local->child_up[read_subvol]) {
- /* should never happen, just in case */
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_READ_SUBVOL_ERROR, "subvolume %d is the "
- "read subvolume in this generation, but is not up",
- read_subvol);
- goto refresh;
- }
-
- local->read_attempted[read_subvol] = 1;
+ afr_ta_read_txn_synctask(frame, this);
+ return 0;
+ }
+
+ ret = afr_inode_read_subvol_get(inode, this, data, metadata,
+ &event_generation);
+ if (ret == -1)
+ /* very first transaction on this inode */
+ goto refresh;
+ AFR_INTERSECT(local->readable, data, metadata, priv->child_count);
+
+ gf_msg_debug(this->name, 0,
+ "%s: generation now vs cached: %d, "
+ "%d",
+ uuid_utoa(inode->gfid), local->event_generation,
+ event_generation);
+ if (afr_is_inode_refresh_reqd(inode, this, local->event_generation,
+ event_generation))
+ /* servers have disconnected / reconnected, and possibly
+ rebooted, very likely changing the state of freshness
+ of copies */
+ goto refresh;
+
+ read_subvol = afr_read_subvol_select_by_policy(inode, this, local->readable,
+ NULL);
+
+ if (read_subvol < 0 || read_subvol > priv->child_count) {
+ gf_msg_debug(this->name, 0,
+ "Unreadable subvolume %d found "
+ "with event generation %d for gfid %s.",
+ read_subvol, event_generation, uuid_utoa(inode->gfid));
+ goto refresh;
+ }
+
+ if (!local->child_up[read_subvol]) {
+ /* should never happen, just in case */
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_READ_SUBVOL_ERROR,
+ "subvolume %d is the "
+ "read subvolume in this generation, but is not up",
+ read_subvol);
+ goto refresh;
+ }
+
+ local->read_attempted[read_subvol] = 1;
read:
- afr_read_txn_wind (frame, this, read_subvol);
+ afr_read_txn_wind(frame, this, read_subvol);
- return 0;
+ return 0;
refresh:
- afr_inode_refresh (frame, this, inode, NULL, afr_read_txn_refresh_done);
+ afr_inode_refresh(frame, this, inode, NULL, afr_read_txn_refresh_done);
- return 0;
+ return 0;
}
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.c b/xlators/cluster/afr/src/afr-self-heal-common.c
index dc380c6d280..a580a1584cc 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.c
+++ b/xlators/cluster/afr/src/afr-self-heal-common.c
@@ -8,749 +8,808 @@
cases as published by the Free Software Foundation.
*/
-
#include "afr.h"
#include "afr-self-heal.h"
-#include "byte-order.h"
+#include <glusterfs/byte-order.h>
#include "protocol-common.h"
#include "afr-messages.h"
-#include "events.h"
+#include <glusterfs/events.h>
void
-afr_heal_synctask (xlator_t *this, afr_local_t *local);
+afr_heal_synctask(xlator_t *this, afr_local_t *local);
int
-afr_lookup_and_heal_gfid (xlator_t *this, inode_t *parent, const char *name,
- inode_t *inode, struct afr_reply *replies,
- int source, void *gfid)
-{
- afr_private_t *priv = NULL;
- call_frame_t *frame = NULL;
- afr_local_t *local = NULL;
- unsigned char *wind_on = NULL;
- ia_type_t ia_type = IA_INVAL;
- dict_t *xdata = NULL;
- loc_t loc = {0, };
- int ret = 0;
- int i = 0;
-
- priv = this->private;
- wind_on = alloca0 (priv->child_count);
+afr_lookup_and_heal_gfid(xlator_t *this, inode_t *parent, const char *name,
+ inode_t *inode, struct afr_reply *replies, int source,
+ unsigned char *sources, void *gfid, int *gfid_idx)
+{
+ afr_private_t *priv = NULL;
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
+ unsigned char *wind_on = NULL;
+ ia_type_t ia_type = IA_INVAL;
+ dict_t *xdata = NULL;
+ loc_t loc = {
+ 0,
+ };
+ int ret = 0;
+ int i = 0;
+
+ priv = this->private;
+ wind_on = alloca0(priv->child_count);
+ if (source >= 0 && replies[source].valid && replies[source].op_ret == 0)
ia_type = replies[source].poststat.ia_type;
- /* gfid heal on those subvolumes that do not have gfid associated
- * with the inode and update those replies.
- */
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret != 0)
- continue;
- if (!gf_uuid_is_null (replies[i].poststat.ia_gfid) ||
- replies[i].poststat.ia_type != ia_type)
- continue;
-
- wind_on[i] = 1;
+ if (ia_type != IA_INVAL)
+ goto heal;
+
+ /* If ia_type is still invalid, it means either
+ * (a)'source' was -1, i.e. parent dir pending xattrs are in split-brain
+ * (or) (b) The parent dir pending xattrs are all zeroes (i.e. all bricks
+ * are sources) and the 'source' we selected earlier might be the one where
+ * the file is not actually present.
+ *
+ * In both cases, let us pick a brick with a successful reply and use its
+ * ia_type.
+ * */
+ for (i = 0; i < priv->child_count; i++) {
+ if (source == -1) {
+ /* case (a) above. */
+ if (replies[i].valid && replies[i].op_ret == 0 &&
+ replies[i].poststat.ia_type != IA_INVAL) {
+ ia_type = replies[i].poststat.ia_type;
+ break;
+ }
+ } else {
+ /* case (b) above. */
+ if (i == source)
+ continue;
+ if (sources[i] && replies[i].valid && replies[i].op_ret == 0 &&
+ replies[i].poststat.ia_type != IA_INVAL) {
+ ia_type = replies[i].poststat.ia_type;
+ break;
+ }
}
+ }
- if (AFR_COUNT(wind_on, priv->child_count) == 0)
- return 0;
+heal:
+ /* gfid heal on those subvolumes that do not have gfid associated
+ * with the inode and update those replies.
+ */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret != 0)
+ continue;
- xdata = dict_new ();
- if (!xdata) {
- ret = -ENOMEM;
- goto out;
- }
+ if (gf_uuid_is_null(gfid) &&
+ !gf_uuid_is_null(replies[i].poststat.ia_gfid) &&
+ replies[i].poststat.ia_type == ia_type)
+ gfid = replies[i].poststat.ia_gfid;
- ret = dict_set_gfuuid (xdata, "gfid-req", gfid, true);
- if (ret) {
- ret = -ENOMEM;
- goto out;
- }
+ if (!gf_uuid_is_null(replies[i].poststat.ia_gfid) ||
+ replies[i].poststat.ia_type != ia_type)
+ continue;
- frame = afr_frame_create (this, &ret);
- if (!frame) {
- ret = -ret;
- goto out;
- }
+ wind_on[i] = 1;
+ }
- local = frame->local;
- loc.parent = inode_ref (parent);
- gf_uuid_copy (loc.pargfid, parent->gfid);
- loc.name = name;
- loc.inode = inode_ref (inode);
+ if (AFR_COUNT(wind_on, priv->child_count) == 0)
+ return 0;
- AFR_ONLIST (wind_on, frame, afr_selfheal_discover_cbk, lookup,
- &loc, xdata);
+ xdata = dict_new();
+ if (!xdata) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ ret = dict_set_gfuuid(xdata, "gfid-req", gfid, true);
+ if (ret) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ frame = afr_frame_create(this, &ret);
+ if (!frame) {
+ ret = -ret;
+ goto out;
+ }
+
+ local = frame->local;
+ loc.parent = inode_ref(parent);
+ gf_uuid_copy(loc.pargfid, parent->gfid);
+ loc.name = name;
+ loc.inode = inode_ref(inode);
+
+ AFR_ONLIST(wind_on, frame, afr_selfheal_discover_cbk, lookup, &loc, xdata);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!wind_on[i])
+ continue;
+ afr_reply_wipe(&replies[i]);
+ afr_reply_copy(&replies[i], &local->replies[i]);
+ }
+ if (gfid_idx && (*gfid_idx == -1)) {
+ /*Pick a brick where the gifd heal was successful.*/
for (i = 0; i < priv->child_count; i++) {
- if (!wind_on[i])
- continue;
- afr_reply_wipe (&replies[i]);
- afr_reply_copy (&replies[i], &local->replies[i]);
+ if (!wind_on[i])
+ continue;
+ if (replies[i].valid && replies[i].op_ret == 0 &&
+ !gf_uuid_is_null(replies[i].poststat.ia_gfid)) {
+ *gfid_idx = i;
+ break;
+ }
}
+ }
out:
- loc_wipe (&loc);
- if (frame)
- AFR_STACK_DESTROY (frame);
- if (xdata)
- dict_unref (xdata);
+ if (gfid_idx && (*gfid_idx == -1) && (ret == 0) && local) {
+ ret = -afr_final_errno(local, priv);
+ }
+ loc_wipe(&loc);
+ if (frame)
+ AFR_STACK_DESTROY(frame);
+ if (xdata)
+ dict_unref(xdata);
- return ret;
+ return ret;
}
int
-afr_gfid_sbrain_source_from_src_brick (xlator_t *this,
- struct afr_reply *replies,
- char *src_brick)
+afr_gfid_sbrain_source_from_src_brick(xlator_t *this, struct afr_reply *replies,
+ char *src_brick)
{
- int i = 0;
- afr_private_t *priv = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret == -1)
- continue;
- if (strcmp (priv->children[i]->name, src_brick) == 0)
- return i;
- }
- return -1;
+ priv = this->private;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret == -1)
+ continue;
+ if (strcmp(priv->children[i]->name, src_brick) == 0)
+ return i;
+ }
+ return -1;
}
int
-afr_selfheal_gfid_mismatch_by_majority (struct afr_reply *replies,
- int child_count)
+afr_selfheal_gfid_mismatch_by_majority(struct afr_reply *replies,
+ int child_count)
{
- int j = 0;
- int i = 0;
- int src = -1;
- int votes[child_count];
-
- for (i = 0; i < child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret == -1)
- continue;
-
- votes[i] = 1;
- for (j = i+1; j < child_count; j++) {
- if ((!gf_uuid_compare (replies[i].poststat.ia_gfid,
- replies[j].poststat.ia_gfid)))
- votes[i]++;
- if (votes[i] > child_count / 2) {
- src = i;
- goto out;
- }
- }
+ int j = 0;
+ int i = 0;
+ int votes;
+
+ for (i = 0; i < child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret == -1)
+ continue;
+
+ votes = 1;
+ for (j = i + 1; j < child_count; j++) {
+ if ((!gf_uuid_compare(replies[i].poststat.ia_gfid,
+ replies[j].poststat.ia_gfid)))
+ votes++;
+ if (votes > child_count / 2)
+ return i;
}
+ }
-out:
- return src;
+ return -1;
}
-int afr_gfid_sbrain_source_from_bigger_file (struct afr_reply *replies,
- int child_count)
+int
+afr_gfid_sbrain_source_from_bigger_file(struct afr_reply *replies,
+ int child_count)
{
- int i = 0;
- int src = -1;
- uint64_t size = 0;
+ int i = 0;
+ int src = -1;
+ uint64_t size = 0;
- for (i = 0; i < child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret == -1)
- continue;
- if (size < replies[i].poststat.ia_size) {
- src = i;
- size = replies[i].poststat.ia_size;
- } else if (replies[i].poststat.ia_size == size) {
- src = -1;
- }
+ for (i = 0; i < child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret == -1)
+ continue;
+ if (size < replies[i].poststat.ia_size) {
+ src = i;
+ size = replies[i].poststat.ia_size;
+ } else if (replies[i].poststat.ia_size == size) {
+ src = -1;
}
- return src;
-}
-
-int afr_gfid_sbrain_source_from_latest_mtime (struct afr_reply *replies,
- int child_count)
-{
- int i = 0;
- int src = -1;
- uint32_t mtime = 0;
- uint32_t mtime_nsec = 0;
-
- for (i = 0; i < child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret != 0)
- continue;
- if ((mtime < replies[i].poststat.ia_mtime) ||
- ((mtime == replies[i].poststat.ia_mtime) &&
- (mtime_nsec < replies[i].poststat.ia_mtime_nsec))) {
- src = i;
- mtime = replies[i].poststat.ia_mtime;
- mtime_nsec = replies[i].poststat.ia_mtime_nsec;
- } else if ((mtime == replies[i].poststat.ia_mtime) &&
- (mtime_nsec == replies[i].poststat.ia_mtime_nsec)) {
- src = -1;
- }
- }
- return src;
+ }
+ return src;
}
int
-afr_gfid_split_brain_source (xlator_t *this, struct afr_reply *replies,
- inode_t *inode, uuid_t pargfid, const char *bname,
- int src_idx, int child_idx,
- unsigned char *locked_on, int *src, dict_t *xdata)
-{
- afr_private_t *priv = NULL;
- char g1[64] = {0,};
- char g2[64] = {0,};
- int up_count = 0;
- int heal_op = -1;
- int ret = -1;
- char *src_brick = NULL;
-
- *src = -1;
- priv = this->private;
- up_count = AFR_COUNT (locked_on, priv->child_count);
- if (up_count != priv->child_count) {
- gf_msg (this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
- "All the bricks should be up to resolve the gfid split "
- "barin");
- if (xdata) {
- ret = dict_set_str (xdata, "gfid-heal-msg", "All the "
- "bricks should be up to resolve the"
- " gfid split barin");
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_DICT_SET_FAILED, "Error setting"
- " gfid-heal-msg dict");
- }
- goto out;
+afr_gfid_sbrain_source_from_latest_mtime(struct afr_reply *replies,
+ int child_count)
+{
+ int i = 0;
+ int src = -1;
+ uint32_t mtime = 0;
+ uint32_t mtime_nsec = 0;
+
+ for (i = 0; i < child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret != 0)
+ continue;
+ if ((mtime < replies[i].poststat.ia_mtime) ||
+ ((mtime == replies[i].poststat.ia_mtime) &&
+ (mtime_nsec < replies[i].poststat.ia_mtime_nsec))) {
+ src = i;
+ mtime = replies[i].poststat.ia_mtime;
+ mtime_nsec = replies[i].poststat.ia_mtime_nsec;
+ } else if ((mtime == replies[i].poststat.ia_mtime) &&
+ (mtime_nsec == replies[i].poststat.ia_mtime_nsec)) {
+ src = -1;
}
+ }
+ return src;
+}
+int
+afr_gfid_split_brain_source(xlator_t *this, struct afr_reply *replies,
+ inode_t *inode, uuid_t pargfid, const char *bname,
+ int src_idx, int child_idx,
+ unsigned char *locked_on, int *src, dict_t *xdata)
+{
+ afr_private_t *priv = NULL;
+ char g1[64] = {
+ 0,
+ };
+ char g2[64] = {
+ 0,
+ };
+ int up_count = 0;
+ int heal_op = -1;
+ int ret = -1;
+ char *src_brick = NULL;
+
+ *src = -1;
+ priv = this->private;
+ up_count = AFR_COUNT(locked_on, priv->child_count);
+ if (up_count != priv->child_count) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "All the bricks should be up to resolve the gfid split "
+ "barin");
if (xdata) {
- ret = dict_get_int32 (xdata, "heal-op", &heal_op);
- if (ret)
- goto fav_child;
- } else {
- goto fav_child;
+ ret = dict_set_sizen_str_sizen(xdata, "gfid-heal-msg",
+ SALL_BRICKS_UP_TO_RESOLVE);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_DICT_SET_FAILED,
+ "Error setting"
+ " gfid-heal-msg dict");
}
+ goto out;
+ }
+
+ if (xdata) {
+ ret = dict_get_int32_sizen(xdata, "heal-op", &heal_op);
+ if (ret)
+ goto fav_child;
+ } else {
+ goto fav_child;
+ }
- switch (heal_op) {
+ switch (heal_op) {
case GF_SHD_OP_SBRAIN_HEAL_FROM_BIGGER_FILE:
- *src = afr_gfid_sbrain_source_from_bigger_file (replies,
- priv->child_count);
- if (*src == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN, "No bigger file");
- if (xdata) {
- ret = dict_set_str (xdata, "gfid-heal-msg",
- "No bigger file");
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_DICT_SET_FAILED, "Error"
- " setting gfid-heal-msg dict");
- }
+ *src = afr_gfid_sbrain_source_from_bigger_file(replies,
+ priv->child_count);
+ if (*src == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ SNO_BIGGER_FILE);
+ if (xdata) {
+ ret = dict_set_sizen_str_sizen(xdata, "gfid-heal-msg",
+ SNO_BIGGER_FILE);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ AFR_MSG_DICT_SET_FAILED,
+ "Error"
+ " setting gfid-heal-msg dict");
}
- break;
+ }
+ break;
case GF_SHD_OP_SBRAIN_HEAL_FROM_LATEST_MTIME:
- *src = afr_gfid_sbrain_source_from_latest_mtime (replies,
- priv->child_count);
- if (*src == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN, "No difference in mtime");
- if (xdata) {
- ret = dict_set_str (xdata, "gfid-heal-msg",
- "No difference in mtime");
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_DICT_SET_FAILED, "Error"
- "setting gfid-heal-msg dict");
- }
+ *src = afr_gfid_sbrain_source_from_latest_mtime(replies,
+ priv->child_count);
+ if (*src == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ SNO_DIFF_IN_MTIME);
+ if (xdata) {
+ ret = dict_set_sizen_str_sizen(xdata, "gfid-heal-msg",
+ SNO_DIFF_IN_MTIME);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ AFR_MSG_DICT_SET_FAILED,
+ "Error"
+ "setting gfid-heal-msg dict");
}
- break;
+ }
+ break;
case GF_SHD_OP_SBRAIN_HEAL_FROM_BRICK:
- ret = dict_get_str (xdata, "child-name", &src_brick);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN, "Error getting the source "
- "brick");
- break;
- }
- *src = afr_gfid_sbrain_source_from_src_brick (this, replies,
- src_brick);
- if (*src == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN, "Error getting the source "
- "brick");
- if (xdata) {
- ret = dict_set_str (xdata, "gfid-heal-msg",
- "Error getting the source "
- "brick");
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_DICT_SET_FAILED, "Error"
- " setting gfid-heal-msg dict");
- }
- }
+ ret = dict_get_str_sizen(xdata, "child-name", &src_brick);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "Error getting the source "
+ "brick");
break;
+ }
+ *src = afr_gfid_sbrain_source_from_src_brick(this, replies,
+ src_brick);
+ if (*src == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ SERROR_GETTING_SRC_BRICK);
+ if (xdata) {
+ ret = dict_set_sizen_str_sizen(xdata, "gfid-heal-msg",
+ SERROR_GETTING_SRC_BRICK);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ AFR_MSG_DICT_SET_FAILED,
+ "Error"
+ " setting gfid-heal-msg dict");
+ }
+ }
+ break;
default:
- break;
- }
- goto out;
+ break;
+ }
+ goto out;
fav_child:
- switch (priv->fav_child_policy) {
+ switch (priv->fav_child_policy) {
case AFR_FAV_CHILD_BY_SIZE:
- *src = afr_sh_fav_by_size (this, replies, inode);
- break;
+ *src = afr_sh_fav_by_size(this, replies, inode);
+ break;
case AFR_FAV_CHILD_BY_MTIME:
- *src = afr_sh_fav_by_mtime (this, replies, inode);
- break;
+ *src = afr_sh_fav_by_mtime(this, replies, inode);
+ break;
case AFR_FAV_CHILD_BY_CTIME:
- *src = afr_sh_fav_by_ctime(this, replies, inode);
- break;
+ *src = afr_sh_fav_by_ctime(this, replies, inode);
+ break;
case AFR_FAV_CHILD_BY_MAJORITY:
- if (priv->child_count != 2)
- *src = afr_selfheal_gfid_mismatch_by_majority (replies,
- priv->child_count);
- else
- *src = -1;
-
- if (*src == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN, "No majority to resolve "
- "gfid split brain");
- }
- break;
+ if (priv->child_count != 2)
+ *src = afr_selfheal_gfid_mismatch_by_majority(
+ replies, priv->child_count);
+ else
+ *src = -1;
+
+ if (*src == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "No majority to resolve "
+ "gfid split brain");
+ }
+ break;
default:
- break;
- }
+ break;
+ }
out:
- if (*src == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
- "Gfid mismatch detected for <gfid:%s>/%s>, %s on %s and"
- " %s on %s.", uuid_utoa (pargfid), bname,
- uuid_utoa_r (replies[child_idx].poststat.ia_gfid, g1),
- priv->children[child_idx]->name,
- uuid_utoa_r (replies[src_idx].poststat.ia_gfid, g2),
- priv->children[src_idx]->name);
- gf_event (EVENT_AFR_SPLIT_BRAIN, "subvol=%s;type=gfid;file="
- "<gfid:%s>/%s>;count=2;child-%d=%s;gfid-%d=%s;"
- "child-%d=%s;gfid-%d=%s", this->name,
- uuid_utoa (pargfid), bname, child_idx,
- priv->children[child_idx]->name, child_idx,
- uuid_utoa_r (replies[child_idx].poststat.ia_gfid, g1),
- src_idx, priv->children[src_idx]->name, src_idx,
- uuid_utoa_r (replies[src_idx].poststat.ia_gfid, g2));
- return -1;
- }
- return 0;
+ if (*src == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "Gfid mismatch detected for <gfid:%s>/%s>, %s on %s and"
+ " %s on %s.",
+ uuid_utoa(pargfid), bname,
+ uuid_utoa_r(replies[child_idx].poststat.ia_gfid, g1),
+ priv->children[child_idx]->name,
+ uuid_utoa_r(replies[src_idx].poststat.ia_gfid, g2),
+ priv->children[src_idx]->name);
+ gf_event(EVENT_AFR_SPLIT_BRAIN,
+ "client-pid=%d;"
+ "subvol=%s;type=gfid;file="
+ "<gfid:%s>/%s>;count=2;child-%d=%s;gfid-%d=%s;"
+ "child-%d=%s;gfid-%d=%s",
+ this->ctx->cmd_args.client_pid, this->name, uuid_utoa(pargfid),
+ bname, child_idx, priv->children[child_idx]->name, child_idx,
+ uuid_utoa_r(replies[child_idx].poststat.ia_gfid, g1), src_idx,
+ priv->children[src_idx]->name, src_idx,
+ uuid_utoa_r(replies[src_idx].poststat.ia_gfid, g2));
+ return -1;
+ }
+ return 0;
}
-
int
-afr_selfheal_post_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
+afr_selfheal_post_op_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- syncbarrier_wake (&local->barrier);
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ syncbarrier_wake(&local->barrier);
- return 0;
+ return 0;
}
-
int
-afr_selfheal_post_op (call_frame_t *frame, xlator_t *this, inode_t *inode,
- int subvol, dict_t *xattr, dict_t *xdata)
+afr_selfheal_post_op(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ int subvol, dict_t *xattr, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- loc_t loc = {0, };
- int ret = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ loc_t loc = {
+ 0,
+ };
+ int ret = 0;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- local->op_ret = 0;
+ local->op_ret = 0;
- STACK_WIND (frame, afr_selfheal_post_op_cbk, priv->children[subvol],
- priv->children[subvol]->fops->xattrop, &loc,
- GF_XATTROP_ADD_ARRAY, xattr, xdata);
+ STACK_WIND(frame, afr_selfheal_post_op_cbk, priv->children[subvol],
+ priv->children[subvol]->fops->xattrop, &loc,
+ GF_XATTROP_ADD_ARRAY, xattr, xdata);
- syncbarrier_wait (&local->barrier, 1);
- if (local->op_ret < 0)
- ret = -local->op_errno;
+ syncbarrier_wait(&local->barrier, 1);
+ if (local->op_ret < 0)
+ ret = -local->op_errno;
- loc_wipe (&loc);
- local->op_ret = 0;
+ loc_wipe(&loc);
+ local->op_ret = 0;
- return ret;
+ return ret;
}
int
-afr_check_stale_error (struct afr_reply *replies, afr_private_t *priv)
+afr_check_stale_error(struct afr_reply *replies, afr_private_t *priv)
{
- int i = 0;
- int op_errno = 0;
- int tmp_errno = 0;
- int stale_count = 0;
+ int i = 0;
+ int op_errno = 0;
+ int tmp_errno = 0;
+ int stale_count = 0;
- for (i = 0; i < priv->child_count; i++) {
- tmp_errno = replies[i].op_errno;
- if (tmp_errno == ENOENT || tmp_errno == ESTALE) {
- op_errno = afr_higher_errno (op_errno, tmp_errno);
- stale_count++;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ tmp_errno = replies[i].op_errno;
+ if (tmp_errno == ENOENT || tmp_errno == ESTALE) {
+ op_errno = afr_higher_errno(op_errno, tmp_errno);
+ stale_count++;
}
- if (stale_count != priv->child_count)
- return -ENOTCONN;
- else
- return -op_errno;
+ }
+ if (stale_count != priv->child_count)
+ return -ENOTCONN;
+ else
+ return -op_errno;
}
int
-afr_sh_generic_fop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *pre, struct iatt *post,
- dict_t *xdata)
+afr_sh_generic_fop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *pre,
+ struct iatt *post, dict_t *xdata)
{
- int i = (long) cookie;
- afr_local_t *local = NULL;
+ int i = (long)cookie;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- local->replies[i].valid = 1;
- local->replies[i].op_ret = op_ret;
- local->replies[i].op_errno = op_errno;
- if (pre)
- local->replies[i].prestat = *pre;
- if (post)
- local->replies[i].poststat = *post;
- if (xdata)
- local->replies[i].xdata = dict_ref (xdata);
+ local->replies[i].valid = 1;
+ local->replies[i].op_ret = op_ret;
+ local->replies[i].op_errno = op_errno;
+ if (pre)
+ local->replies[i].prestat = *pre;
+ if (post)
+ local->replies[i].poststat = *post;
+ if (xdata)
+ local->replies[i].xdata = dict_ref(xdata);
- syncbarrier_wake (&local->barrier);
+ syncbarrier_wake(&local->barrier);
- return 0;
+ return 0;
}
int
-afr_selfheal_restore_time (call_frame_t *frame, xlator_t *this, inode_t *inode,
- int source, unsigned char *healed_sinks,
- struct afr_reply *replies)
+afr_selfheal_restore_time(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ int source, unsigned char *healed_sinks,
+ struct afr_reply *replies)
{
- loc_t loc = {0, };
+ loc_t loc = {
+ 0,
+ };
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- AFR_ONLIST (healed_sinks, frame, afr_sh_generic_fop_cbk, setattr, &loc,
- &replies[source].poststat,
- (GF_SET_ATTR_ATIME|GF_SET_ATTR_MTIME), NULL);
+ AFR_ONLIST(healed_sinks, frame, afr_sh_generic_fop_cbk, setattr, &loc,
+ &replies[source].poststat,
+ (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME | GF_SET_ATTR_CTIME),
+ NULL);
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return 0;
+ return 0;
}
dict_t *
-afr_selfheal_output_xattr (xlator_t *this, gf_boolean_t is_full_crawl,
- afr_transaction_type type, int *output_dirty,
- int **output_matrix, int subvol,
- int **full_heal_mtx_out)
-{
- int j = 0;
- int idx = 0;
- int d_idx = 0;
- int ret = 0;
- int *raw = 0;
- dict_t *xattr = NULL;
- afr_private_t *priv = NULL;
-
- priv = this->private;
- idx = afr_index_for_transaction_type (type);
- d_idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
-
- xattr = dict_new ();
- if (!xattr)
- return NULL;
-
- /* clear dirty */
- raw = GF_CALLOC (sizeof(int), AFR_NUM_CHANGE_LOGS, gf_afr_mt_int32_t);
- if (!raw)
- goto err;
-
- raw[idx] = hton32 (output_dirty[subvol]);
- ret = dict_set_bin (xattr, AFR_DIRTY, raw,
- sizeof(int) * AFR_NUM_CHANGE_LOGS);
- if (ret) {
- GF_FREE (raw);
- goto err;
- }
+afr_selfheal_output_xattr(xlator_t *this, gf_boolean_t is_full_crawl,
+ afr_transaction_type type, int *output_dirty,
+ int **output_matrix, int subvol,
+ int **full_heal_mtx_out)
+{
+ int j = 0;
+ int idx = 0;
+ int d_idx = 0;
+ int ret = 0;
+ int *raw = 0;
+ dict_t *xattr = NULL;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ idx = afr_index_for_transaction_type(type);
+ d_idx = afr_index_for_transaction_type(AFR_DATA_TRANSACTION);
+
+ xattr = dict_new();
+ if (!xattr)
+ return NULL;
- /* clear/set pending */
- for (j = 0; j < priv->child_count; j++) {
- raw = GF_CALLOC (sizeof(int), AFR_NUM_CHANGE_LOGS,
- gf_afr_mt_int32_t);
- if (!raw)
- goto err;
-
- raw[idx] = hton32 (output_matrix[subvol][j]);
- if (is_full_crawl)
- raw[d_idx] = hton32 (full_heal_mtx_out[subvol][j]);
-
- ret = dict_set_bin (xattr, priv->pending_key[j],
- raw, sizeof(int) * AFR_NUM_CHANGE_LOGS);
- if (ret) {
- GF_FREE (raw);
- goto err;
- }
- }
+ /* clear dirty */
+ raw = GF_CALLOC(sizeof(int), AFR_NUM_CHANGE_LOGS, gf_afr_mt_int32_t);
+ if (!raw)
+ goto err;
+
+ raw[idx] = hton32(output_dirty[subvol]);
+ ret = dict_set_bin(xattr, AFR_DIRTY, raw,
+ sizeof(int) * AFR_NUM_CHANGE_LOGS);
+ if (ret) {
+ GF_FREE(raw);
+ goto err;
+ }
+
+ /* clear/set pending */
+ for (j = 0; j < priv->child_count; j++) {
+ raw = GF_CALLOC(sizeof(int), AFR_NUM_CHANGE_LOGS, gf_afr_mt_int32_t);
+ if (!raw)
+ goto err;
+
+ raw[idx] = hton32(output_matrix[subvol][j]);
+ if (is_full_crawl)
+ raw[d_idx] = hton32(full_heal_mtx_out[subvol][j]);
+
+ ret = dict_set_bin(xattr, priv->pending_key[j], raw,
+ sizeof(int) * AFR_NUM_CHANGE_LOGS);
+ if (ret) {
+ GF_FREE(raw);
+ goto err;
+ }
+ }
- return xattr;
+ return xattr;
err:
- if (xattr)
- dict_unref (xattr);
- return NULL;
+ if (xattr)
+ dict_unref(xattr);
+ return NULL;
}
-
int
-afr_selfheal_undo_pending (call_frame_t *frame, xlator_t *this, inode_t *inode,
- unsigned char *sources, unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *undid_pending,
- afr_transaction_type type, struct afr_reply *replies,
- unsigned char *locked_on)
-{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = 0;
- int j = 0;
- unsigned char *pending = NULL;
- int *input_dirty = NULL;
- int **input_matrix = NULL;
- int **full_heal_mtx_in = NULL;
- int **full_heal_mtx_out = NULL;
- int *output_dirty = NULL;
- int **output_matrix = NULL;
- dict_t *xattr = NULL;
- dict_t *xdata = NULL;
-
- priv = this->private;
- local = frame->local;
-
- pending = alloca0 (priv->child_count);
-
- input_dirty = alloca0 (priv->child_count * sizeof (int));
- input_matrix = ALLOC_MATRIX (priv->child_count, int);
- full_heal_mtx_in = ALLOC_MATRIX (priv->child_count, int);
- full_heal_mtx_out = ALLOC_MATRIX (priv->child_count, int);
- output_dirty = alloca0 (priv->child_count * sizeof (int));
- output_matrix = ALLOC_MATRIX (priv->child_count, int);
-
- xdata = dict_new ();
- if (!xdata)
- return -1;
+afr_selfheal_undo_pending(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ unsigned char *sources, unsigned char *sinks,
+ unsigned char *healed_sinks,
+ unsigned char *undid_pending,
+ afr_transaction_type type, struct afr_reply *replies,
+ unsigned char *locked_on)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
+ int j = 0;
+ unsigned char *pending = NULL;
+ int *input_dirty = NULL;
+ int **input_matrix = NULL;
+ int **full_heal_mtx_in = NULL;
+ int **full_heal_mtx_out = NULL;
+ int *output_dirty = NULL;
+ int **output_matrix = NULL;
+ dict_t *xattr = NULL;
+ dict_t *xdata = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ pending = alloca0(priv->child_count);
+
+ input_dirty = alloca0(priv->child_count * sizeof(int));
+ input_matrix = ALLOC_MATRIX(priv->child_count, int);
+ full_heal_mtx_in = ALLOC_MATRIX(priv->child_count, int);
+ full_heal_mtx_out = ALLOC_MATRIX(priv->child_count, int);
+ output_dirty = alloca0(priv->child_count * sizeof(int));
+ output_matrix = ALLOC_MATRIX(priv->child_count, int);
+
+ xdata = dict_new();
+ if (!xdata)
+ return -1;
- afr_selfheal_extract_xattr (this, replies, type, input_dirty,
- input_matrix);
-
- if (local->need_full_crawl)
- afr_selfheal_extract_xattr (this, replies, AFR_DATA_TRANSACTION,
- NULL, full_heal_mtx_in);
-
- for (i = 0; i < priv->child_count; i++)
- if (sinks[i] && !healed_sinks[i])
- pending[i] = 1;
-
- for (i = 0; i < priv->child_count; i++) {
- for (j = 0; j < priv->child_count; j++) {
- if (pending[j]) {
- output_matrix[i][j] = 1;
- if (type == AFR_ENTRY_TRANSACTION)
- full_heal_mtx_out[i][j] = 1;
- } else if (locked_on[j]) {
- output_matrix[i][j] = -input_matrix[i][j];
- if (type == AFR_ENTRY_TRANSACTION)
- full_heal_mtx_out[i][j] = -full_heal_mtx_in[i][j];
- }
- }
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (!pending[i])
- output_dirty[i] = -input_dirty[i];
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (!locked_on[i])
- /* perform post-op only on subvols we had locked
- and inspected on.
- */
- continue;
- if (undid_pending[i])
- /* We already unset the pending xattrs in
- * _afr_fav_child_reset_sink_xattrs(). */
- continue;
-
- xattr = afr_selfheal_output_xattr (this, local->need_full_crawl,
- type, output_dirty,
- output_matrix, i,
- full_heal_mtx_out);
- if (!xattr) {
- continue;
- }
-
- if ((type == AFR_ENTRY_TRANSACTION) && (priv->esh_granular)) {
- if (xdata &&
- dict_set_int8 (xdata, GF_XATTROP_PURGE_INDEX, 1))
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_DICT_SET_FAILED, "Failed to set"
- " dict value for %s",
- GF_XATTROP_PURGE_INDEX);
- }
+ afr_selfheal_extract_xattr(this, replies, type, input_dirty, input_matrix);
+
+ if (local->need_full_crawl)
+ afr_selfheal_extract_xattr(this, replies, AFR_DATA_TRANSACTION, NULL,
+ full_heal_mtx_in);
+
+ for (i = 0; i < priv->child_count; i++)
+ if (sinks[i] && !healed_sinks[i])
+ pending[i] = 1;
+
+ for (i = 0; i < priv->child_count; i++) {
+ for (j = 0; j < priv->child_count; j++) {
+ if (pending[j]) {
+ output_matrix[i][j] = 1;
+ if (type == AFR_ENTRY_TRANSACTION)
+ full_heal_mtx_out[i][j] = 1;
+ } else if (locked_on[j]) {
+ output_matrix[i][j] = -input_matrix[i][j];
+ if (type == AFR_ENTRY_TRANSACTION)
+ full_heal_mtx_out[i][j] = -full_heal_mtx_in[i][j];
+ }
+ }
+ }
- afr_selfheal_post_op (frame, this, inode, i, xattr, xdata);
- dict_unref (xattr);
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!pending[i])
+ output_dirty[i] = -input_dirty[i];
+ }
- if (xdata)
- dict_unref (xdata);
+ for (i = 0; i < priv->child_count; i++) {
+ if (!locked_on[i])
+ /* perform post-op only on subvols we had locked
+ and inspected on.
+ */
+ continue;
+ if (undid_pending[i])
+ /* We already unset the pending xattrs in
+ * _afr_fav_child_reset_sink_xattrs(). */
+ continue;
- return 0;
+ xattr = afr_selfheal_output_xattr(this, local->need_full_crawl, type,
+ output_dirty, output_matrix, i,
+ full_heal_mtx_out);
+ if (!xattr) {
+ continue;
+ }
+
+ if ((type == AFR_ENTRY_TRANSACTION) && (priv->esh_granular)) {
+ if (xdata && dict_set_int8(xdata, GF_XATTROP_PURGE_INDEX, 1))
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_DICT_SET_FAILED,
+ "Failed to set"
+ " dict value for %s",
+ GF_XATTROP_PURGE_INDEX);
+ }
+
+ afr_selfheal_post_op(frame, this, inode, i, xattr, xdata);
+ dict_unref(xattr);
+ }
+
+ if (xdata)
+ dict_unref(xdata);
+
+ return 0;
}
void
-afr_reply_copy (struct afr_reply *dst, struct afr_reply *src)
-{
- dict_t *xdata = NULL;
-
- dst->valid = src->valid;
- dst->op_ret = src->op_ret;
- dst->op_errno = src->op_errno;
- dst->prestat = src->prestat;
- dst->poststat = src->poststat;
- dst->preparent = src->preparent;
- dst->postparent = src->postparent;
- dst->preparent2 = src->preparent2;
- dst->postparent2 = src->postparent2;
- if (src->xdata)
- xdata = dict_ref (src->xdata);
- else
- xdata = NULL;
- if (dst->xdata)
- dict_unref (dst->xdata);
- dst->xdata = xdata;
- if (xdata && dict_get_str_boolean (xdata, "fips-mode-rchecksum",
- _gf_false) == _gf_true) {
- memcpy (dst->checksum, src->checksum, SHA256_DIGEST_LENGTH);
- } else {
- memcpy (dst->checksum, src->checksum, MD5_DIGEST_LENGTH);
- }
- dst->fips_mode_rchecksum = src->fips_mode_rchecksum;
+afr_reply_copy(struct afr_reply *dst, struct afr_reply *src)
+{
+ dict_t *xdata = NULL;
+
+ dst->valid = src->valid;
+ dst->op_ret = src->op_ret;
+ dst->op_errno = src->op_errno;
+ dst->prestat = src->prestat;
+ dst->poststat = src->poststat;
+ dst->preparent = src->preparent;
+ dst->postparent = src->postparent;
+ dst->preparent2 = src->preparent2;
+ dst->postparent2 = src->postparent2;
+ if (src->xdata)
+ xdata = dict_ref(src->xdata);
+ else
+ xdata = NULL;
+ if (dst->xdata)
+ dict_unref(dst->xdata);
+ dst->xdata = xdata;
+ if (xdata && dict_get_str_boolean(xdata, "fips-mode-rchecksum",
+ _gf_false) == _gf_true) {
+ memcpy(dst->checksum, src->checksum, SHA256_DIGEST_LENGTH);
+ } else {
+ memcpy(dst->checksum, src->checksum, MD5_DIGEST_LENGTH);
+ }
+ dst->fips_mode_rchecksum = src->fips_mode_rchecksum;
}
void
-afr_replies_copy (struct afr_reply *dst, struct afr_reply *src, int count)
+afr_replies_copy(struct afr_reply *dst, struct afr_reply *src, int count)
{
- int i = 0;
+ int i = 0;
- if (dst == src)
- return;
+ if (dst == src)
+ return;
- for (i = 0; i < count; i++) {
- afr_reply_copy (&dst[i], &src[i]);
- }
+ for (i = 0; i < count; i++) {
+ afr_reply_copy(&dst[i], &src[i]);
+ }
}
int
-afr_selfheal_fill_dirty (xlator_t *this, int *dirty, int subvol,
- int idx, dict_t *xdata)
+afr_selfheal_fill_dirty(xlator_t *this, int *dirty, int subvol, int idx,
+ dict_t *xdata)
{
- void *pending_raw = NULL;
- int pending[3] = {0, };
+ void *pending_raw = NULL;
+ int pending[3] = {
+ 0,
+ };
- if (!dirty)
- return 0;
+ if (!dirty)
+ return 0;
- if (dict_get_ptr (xdata, AFR_DIRTY, &pending_raw))
- return -1;
+ if (dict_get_ptr(xdata, AFR_DIRTY, &pending_raw))
+ return -1;
- if (!pending_raw)
- return -1;
+ if (!pending_raw)
+ return -1;
- memcpy (pending, pending_raw, sizeof(pending));
+ memcpy(pending, pending_raw, sizeof(pending));
- dirty[subvol] = ntoh32 (pending[idx]);
+ dirty[subvol] = ntoh32(pending[idx]);
- return 0;
+ return 0;
}
-
int
-afr_selfheal_fill_matrix (xlator_t *this, int **matrix, int subvol,
- int idx, dict_t *xdata)
+afr_selfheal_fill_matrix(xlator_t *this, int **matrix, int subvol, int idx,
+ dict_t *xdata)
{
- int i = 0;
- void *pending_raw = NULL;
- int pending[3] = {0, };
- afr_private_t *priv = NULL;
+ int i = 0;
+ void *pending_raw = NULL;
+ int pending[3] = {
+ 0,
+ };
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (!matrix)
- return 0;
+ if (!matrix)
+ return 0;
- for (i = 0; i < priv->child_count; i++) {
- if (dict_get_ptr (xdata, priv->pending_key[i], &pending_raw))
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (dict_get_ptr(xdata, priv->pending_key[i], &pending_raw))
+ continue;
- if (!pending_raw)
- continue;
+ if (!pending_raw)
+ continue;
- memcpy (pending, pending_raw, sizeof(pending));
+ memcpy(pending, pending_raw, sizeof(pending));
- matrix[subvol][i] = ntoh32 (pending[idx]);
- }
+ matrix[subvol][i] = ntoh32(pending[idx]);
+ }
- return 0;
+ return 0;
}
-
int
-afr_selfheal_extract_xattr (xlator_t *this, struct afr_reply *replies,
- afr_transaction_type type, int *dirty, int **matrix)
+afr_selfheal_extract_xattr(xlator_t *this, struct afr_reply *replies,
+ afr_transaction_type type, int *dirty, int **matrix)
{
- afr_private_t *priv = NULL;
- int i = 0;
- dict_t *xdata = NULL;
- int idx = -1;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ dict_t *xdata = NULL;
+ int idx = -1;
- idx = afr_index_for_transaction_type (type);
+ idx = afr_index_for_transaction_type(type);
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret != 0)
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret != 0)
+ continue;
- if (!replies[i].xdata)
- continue;
+ if (!replies[i].xdata)
+ continue;
- xdata = replies[i].xdata;
+ xdata = replies[i].xdata;
- afr_selfheal_fill_dirty (this, dirty, i, idx, xdata);
- afr_selfheal_fill_matrix (this, matrix, i, idx, xdata);
- }
+ afr_selfheal_fill_dirty(this, dirty, i, idx, xdata);
+ afr_selfheal_fill_matrix(this, matrix, i, idx, xdata);
+ }
- return 0;
+ return 0;
}
/*
@@ -760,560 +819,566 @@ afr_selfheal_extract_xattr (xlator_t *this, struct afr_reply *replies,
* This can happen if data was directly modified in the backend or for snapshots
*/
void
-afr_mark_largest_file_as_source (xlator_t *this, unsigned char *sources,
- struct afr_reply *replies)
+afr_mark_largest_file_as_source(xlator_t *this, unsigned char *sources,
+ struct afr_reply *replies)
{
- int i = 0;
- afr_private_t *priv = NULL;
- uint64_t size = 0;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ uint64_t size = 0;
- /* Find source with biggest file size */
- priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
- if (!replies[i].valid || replies[i].op_ret != 0) {
- sources[i] = 0;
- continue;
- }
- if (size <= replies[i].poststat.ia_size) {
- size = replies[i].poststat.ia_size;
- }
+ /* Find source with biggest file size */
+ priv = this->private;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ if (!replies[i].valid || replies[i].op_ret != 0) {
+ sources[i] = 0;
+ continue;
}
-
- /* Mark sources with less size as not source */
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
- if (size > replies[i].poststat.ia_size)
- sources[i] = 0;
+ if (size <= replies[i].poststat.ia_size) {
+ size = replies[i].poststat.ia_size;
}
+ }
+
+ /* Mark sources with less size as not source */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ if (size > replies[i].poststat.ia_size)
+ sources[i] = 0;
+ }
}
void
-afr_mark_latest_mtime_file_as_source (xlator_t *this, unsigned char *sources,
- struct afr_reply *replies)
-{
- int i = 0;
- afr_private_t *priv = NULL;
- uint32_t mtime = 0;
- uint32_t mtime_nsec = 0;
-
- priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
- if (!replies[i].valid || replies[i].op_ret != 0) {
- sources[i] = 0;
- continue;
- }
- if ((mtime < replies[i].poststat.ia_mtime) ||
- ((mtime == replies[i].poststat.ia_mtime) &&
- (mtime_nsec < replies[i].poststat.ia_mtime_nsec))) {
- mtime = replies[i].poststat.ia_mtime;
- mtime_nsec = replies[i].poststat.ia_mtime_nsec;
- }
- }
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
- if ((mtime > replies[i].poststat.ia_mtime) ||
- ((mtime == replies[i].poststat.ia_mtime) &&
- (mtime_nsec > replies[i].poststat.ia_mtime_nsec))) {
- sources[i] = 0;
- }
- }
+afr_mark_latest_mtime_file_as_source(xlator_t *this, unsigned char *sources,
+ struct afr_reply *replies)
+{
+ int i = 0;
+ afr_private_t *priv = NULL;
+ uint32_t mtime = 0;
+ uint32_t mtime_nsec = 0;
+
+ priv = this->private;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ if (!replies[i].valid || replies[i].op_ret != 0) {
+ sources[i] = 0;
+ continue;
+ }
+ if ((mtime < replies[i].poststat.ia_mtime) ||
+ ((mtime == replies[i].poststat.ia_mtime) &&
+ (mtime_nsec < replies[i].poststat.ia_mtime_nsec))) {
+ mtime = replies[i].poststat.ia_mtime;
+ mtime_nsec = replies[i].poststat.ia_mtime_nsec;
+ }
+ }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ if ((mtime > replies[i].poststat.ia_mtime) ||
+ ((mtime == replies[i].poststat.ia_mtime) &&
+ (mtime_nsec > replies[i].poststat.ia_mtime_nsec))) {
+ sources[i] = 0;
+ }
+ }
}
void
-afr_mark_active_sinks (xlator_t *this, unsigned char *sources,
- unsigned char *locked_on, unsigned char *sinks)
+afr_mark_active_sinks(xlator_t *this, unsigned char *sources,
+ unsigned char *locked_on, unsigned char *sinks)
{
- int i = 0;
- afr_private_t *priv = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- memset (sinks, 0, sizeof (*sinks) * priv->child_count);
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i] && locked_on[i])
- sinks[i] = 1;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i] && locked_on[i])
+ sinks[i] = 1;
+ else
+ sinks[i] = 0;
+ }
}
gf_boolean_t
-afr_dict_contains_heal_op (call_frame_t *frame)
+afr_dict_contains_heal_op(call_frame_t *frame)
{
- afr_local_t *local = NULL;
- dict_t *xdata_req = NULL;
- int ret = 0;
- int heal_op = -1;
+ afr_local_t *local = NULL;
+ dict_t *xdata_req = NULL;
+ int ret = 0;
+ int heal_op = -1;
- local = frame->local;
- xdata_req = local->xdata_req;
- ret = dict_get_int32 (xdata_req, "heal-op", &heal_op);
- if (ret)
- return _gf_false;
- if (local->xdata_rsp == NULL) {
- local->xdata_rsp = dict_new();
- if (!local->xdata_rsp)
- return _gf_true;
- }
- ret = dict_set_str (local->xdata_rsp, "sh-fail-msg",
- "File not in split-brain");
+ local = frame->local;
+ xdata_req = local->xdata_req;
+ ret = dict_get_int32_sizen(xdata_req, "heal-op", &heal_op);
+ if (ret)
+ return _gf_false;
+ if (local->xdata_rsp == NULL) {
+ local->xdata_rsp = dict_new();
+ if (!local->xdata_rsp)
+ return _gf_true;
+ }
+ ret = dict_set_sizen_str_sizen(local->xdata_rsp, "sh-fail-msg",
+ SFILE_NOT_IN_SPLIT_BRAIN);
- return _gf_true;
+ return _gf_true;
}
gf_boolean_t
-afr_can_decide_split_brain_source_sinks (struct afr_reply *replies,
- int child_count)
+afr_can_decide_split_brain_source_sinks(struct afr_reply *replies,
+ int child_count)
{
- int i = 0;
+ int i = 0;
- for (i = 0; i < child_count; i++)
- if (replies[i].valid != 1 || replies[i].op_ret != 0)
- return _gf_false;
+ for (i = 0; i < child_count; i++)
+ if (replies[i].valid != 1 || replies[i].op_ret != 0)
+ return _gf_false;
- return _gf_true;
+ return _gf_true;
}
int
-afr_mark_split_brain_source_sinks_by_heal_op (call_frame_t *frame,
- xlator_t *this, unsigned char *sources,
- unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *locked_on,
- struct afr_reply *replies,
- afr_transaction_type type, int heal_op)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- dict_t *xdata_req = NULL;
- dict_t *xdata_rsp = NULL;
- int ret = 0;
- int i = 0;
- char *name = NULL;
- int source = -1;
-
- local = frame->local;
- priv = this->private;
- xdata_req = local->xdata_req;
-
- for (i = 0; i < priv->child_count; i++) {
- if (locked_on[i])
- if (sources[i] || !sinks[i] || !healed_sinks[i]) {
- ret = -1;
- goto out;
- }
- }
- if (local->xdata_rsp == NULL) {
- local->xdata_rsp = dict_new();
- if (!local->xdata_rsp) {
- ret = -1;
- goto out;
- }
- }
- xdata_rsp = local->xdata_rsp;
-
- if (!afr_can_decide_split_brain_source_sinks (replies,
- priv->child_count)) {
- ret = dict_set_str (xdata_rsp, "sh-fail-msg",
- SBRAIN_HEAL_NO_GO_MSG);
+afr_mark_split_brain_source_sinks_by_heal_op(
+ call_frame_t *frame, xlator_t *this, unsigned char *sources,
+ unsigned char *sinks, unsigned char *healed_sinks, unsigned char *locked_on,
+ struct afr_reply *replies, afr_transaction_type type, int heal_op)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ dict_t *xdata_req = NULL;
+ dict_t *xdata_rsp = NULL;
+ int ret = 0;
+ int i = 0;
+ char *name = NULL;
+ int source = -1;
+
+ local = frame->local;
+ priv = this->private;
+ xdata_req = local->xdata_req;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (locked_on[i])
+ if (sources[i] || !sinks[i] || !healed_sinks[i]) {
ret = -1;
goto out;
- }
+ }
+ }
+ if (local->xdata_rsp == NULL) {
+ local->xdata_rsp = dict_new();
+ if (!local->xdata_rsp) {
+ ret = -1;
+ goto out;
+ }
+ }
+ xdata_rsp = local->xdata_rsp;
+
+ if (!afr_can_decide_split_brain_source_sinks(replies, priv->child_count)) {
+ ret = dict_set_sizen_str_sizen(xdata_rsp, "sh-fail-msg",
+ SBRAIN_HEAL_NO_GO_MSG);
+ ret = -1;
+ goto out;
+ }
- for (i = 0 ; i < priv->child_count; i++)
- if (locked_on[i])
- sources[i] = 1;
- switch (heal_op) {
+ for (i = 0; i < priv->child_count; i++)
+ if (locked_on[i])
+ sources[i] = 1;
+ switch (heal_op) {
case GF_SHD_OP_SBRAIN_HEAL_FROM_BIGGER_FILE:
- if (type == AFR_METADATA_TRANSACTION) {
- ret = dict_set_str (xdata_rsp, "sh-fail-msg",
- "Use source-brick option to"
- " heal metadata split-brain");
- if (!ret)
- ret = -1;
- goto out;
- }
- afr_mark_largest_file_as_source (this, sources, replies);
- if (AFR_COUNT (sources, priv->child_count) != 1) {
- ret = dict_set_str (xdata_rsp, "sh-fail-msg",
- "No bigger file");
- if (!ret)
- ret = -1;
- goto out;
- }
- break;
+ if (type == AFR_METADATA_TRANSACTION) {
+ ret = dict_set_sizen_str_sizen(xdata_rsp, "sh-fail-msg",
+ SUSE_SOURCE_BRICK_TO_HEAL);
+ if (!ret)
+ ret = -1;
+ goto out;
+ }
+ afr_mark_largest_file_as_source(this, sources, replies);
+ if (AFR_COUNT(sources, priv->child_count) != 1) {
+ ret = dict_set_sizen_str_sizen(xdata_rsp, "sh-fail-msg",
+ SNO_BIGGER_FILE);
+ if (!ret)
+ ret = -1;
+ goto out;
+ }
+ break;
case GF_SHD_OP_SBRAIN_HEAL_FROM_LATEST_MTIME:
- if (type == AFR_METADATA_TRANSACTION) {
- ret = dict_set_str (xdata_rsp, "sh-fail-msg",
- "Use source-brick option to"
- " heal metadata split-brain");
- if (!ret)
- ret = -1;
- goto out;
- }
- afr_mark_latest_mtime_file_as_source (this, sources, replies);
- if (AFR_COUNT (sources, priv->child_count) != 1) {
- ret = dict_set_str (xdata_rsp, "sh-fail-msg",
- "No difference in mtime");
- if (!ret)
- ret = -1;
- goto out;
- }
- break;
+ if (type == AFR_METADATA_TRANSACTION) {
+ ret = dict_set_sizen_str_sizen(xdata_rsp, "sh-fail-msg",
+ SUSE_SOURCE_BRICK_TO_HEAL);
+ if (!ret)
+ ret = -1;
+ goto out;
+ }
+ afr_mark_latest_mtime_file_as_source(this, sources, replies);
+ if (AFR_COUNT(sources, priv->child_count) != 1) {
+ ret = dict_set_sizen_str_sizen(xdata_rsp, "sh-fail-msg",
+ SNO_DIFF_IN_MTIME);
+ if (!ret)
+ ret = -1;
+ goto out;
+ }
+ break;
case GF_SHD_OP_SBRAIN_HEAL_FROM_BRICK:
- ret = dict_get_str (xdata_req, "child-name", &name);
- if (ret)
- goto out;
- source = afr_get_child_index_from_name (this, name);
- if (source < 0) {
- ret = dict_set_str (xdata_rsp, "sh-fail-msg",
- "Invalid brick name");
- if (!ret)
- ret = -1;
- goto out;
- }
- if (locked_on[source] != 1) {
- ret = dict_set_str (xdata_rsp, "sh-fail-msg",
- "Brick is not up");
- if (!ret)
- ret = -1;
- goto out;
- }
- memset (sources, 0, sizeof (*sources) * priv->child_count);
- sources[source] = 1;
- break;
- default:
- ret = -1;
+ ret = dict_get_str_sizen(xdata_req, "child-name", &name);
+ if (ret)
goto out;
- }
- for (i = 0 ; i < priv->child_count; i++) {
- if (sources[i]) {
- source = i;
- break;
- }
- }
- sinks[source] = 0;
- healed_sinks[source] = 0;
- ret = source;
+ source = afr_get_child_index_from_name(this, name);
+ if (source < 0) {
+ ret = dict_set_sizen_str_sizen(xdata_rsp, "sh-fail-msg",
+ SINVALID_BRICK_NAME);
+ if (!ret)
+ ret = -1;
+ goto out;
+ }
+ if (locked_on[source] != 1) {
+ ret = dict_set_sizen_str_sizen(xdata_rsp, "sh-fail-msg",
+ SBRICK_IS_NOT_UP);
+ if (!ret)
+ ret = -1;
+ goto out;
+ }
+ memset(sources, 0, sizeof(*sources) * priv->child_count);
+ sources[source] = 1;
+ break;
+ default:
+ ret = -1;
+ goto out;
+ }
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i]) {
+ source = i;
+ break;
+ }
+ }
+ sinks[source] = 0;
+ healed_sinks[source] = 0;
+ ret = source;
out:
- if (ret < 0)
- memset (sources, 0, sizeof (*sources) * priv->child_count);
- return ret;
-
+ if (ret < 0)
+ memset(sources, 0, sizeof(*sources) * priv->child_count);
+ return ret;
}
int
-afr_sh_fav_by_majority (xlator_t *this, struct afr_reply *replies,
- inode_t *inode)
-{
- afr_private_t *priv;
- int vote_count = -1;
- int fav_child = -1;
- int i = 0;
- int k = 0;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (replies[i].valid == 1) {
- gf_msg_debug (this->name, 0, "Child:%s "
- "mtime_sec = %ld, size = %lu for gfid %s",
- priv->children[i]->name,
- replies[i].poststat.ia_mtime,
- replies[i].poststat.ia_size,
- uuid_utoa (inode->gfid));
- vote_count = 0;
- for (k = 0; k < priv->child_count; k++) {
- if ((replies[k].poststat.ia_mtime ==
- replies[i].poststat.ia_mtime) &&
- (replies[k].poststat.ia_size ==
- replies[i].poststat.ia_size)
- ) {
- vote_count++;
- }
- }
- if (vote_count > priv->child_count/2) {
- fav_child = i;
- break;
- }
+afr_sh_fav_by_majority(xlator_t *this, struct afr_reply *replies,
+ inode_t *inode)
+{
+ afr_private_t *priv;
+ int vote_count = -1;
+ int fav_child = -1;
+ int i = 0;
+ int k = 0;
+
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (replies[i].valid == 1) {
+ gf_msg_debug(this->name, 0,
+ "Child:%s mtime_sec = %" PRId64 ", size = %" PRIu64
+ " for gfid %s",
+ priv->children[i]->name, replies[i].poststat.ia_mtime,
+ replies[i].poststat.ia_size, uuid_utoa(inode->gfid));
+ vote_count = 0;
+ for (k = 0; k < priv->child_count; k++) {
+ if ((replies[k].poststat.ia_mtime ==
+ replies[i].poststat.ia_mtime) &&
+ (replies[k].poststat.ia_size ==
+ replies[i].poststat.ia_size)) {
+ vote_count++;
}
+ }
+ if (vote_count > priv->child_count / 2) {
+ fav_child = i;
+ break;
+ }
}
- return fav_child;
+ }
+ return fav_child;
}
/*
* afr_sh_fav_by_mtime: Choose favorite child by mtime.
*/
int
-afr_sh_fav_by_mtime (xlator_t *this, struct afr_reply *replies, inode_t *inode)
-{
- afr_private_t *priv;
- int fav_child = -1;
- int i = 0;
- uint32_t cmp_mtime = 0;
- uint32_t cmp_mtime_nsec = 0;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (replies[i].valid == 1) {
- gf_msg_debug (this->name, 0, "Child:%s "
- "mtime = %ld, mtime_nsec = %d for gfid %s",
- priv->children[i]->name,
- replies[i].poststat.ia_mtime,
- replies[i].poststat.ia_mtime_nsec,
- uuid_utoa (inode->gfid));
- if (replies[i].poststat.ia_mtime > cmp_mtime) {
- cmp_mtime = replies[i].poststat.ia_mtime;
- cmp_mtime_nsec =
- replies[i].poststat.ia_mtime_nsec;
- fav_child = i;
- } else if ((replies[i].poststat.ia_mtime == cmp_mtime)
- && (replies[i].poststat.ia_mtime_nsec >
- cmp_mtime_nsec)) {
- cmp_mtime = replies[i].poststat.ia_mtime;
- cmp_mtime_nsec =
- replies[i].poststat.ia_mtime_nsec;
- fav_child = i;
- }
- }
- }
- return fav_child;
+afr_sh_fav_by_mtime(xlator_t *this, struct afr_reply *replies, inode_t *inode)
+{
+ afr_private_t *priv;
+ int fav_child = -1;
+ int i = 0;
+ uint32_t cmp_mtime = 0;
+ uint32_t cmp_mtime_nsec = 0;
+
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (replies[i].valid == 1) {
+ gf_msg_debug(this->name, 0,
+ "Child:%s mtime = %" PRId64
+ ", mtime_nsec = %d for "
+ "gfid %s",
+ priv->children[i]->name, replies[i].poststat.ia_mtime,
+ replies[i].poststat.ia_mtime_nsec,
+ uuid_utoa(inode->gfid));
+ if (replies[i].poststat.ia_mtime > cmp_mtime) {
+ cmp_mtime = replies[i].poststat.ia_mtime;
+ cmp_mtime_nsec = replies[i].poststat.ia_mtime_nsec;
+ fav_child = i;
+ } else if ((replies[i].poststat.ia_mtime == cmp_mtime) &&
+ (replies[i].poststat.ia_mtime_nsec > cmp_mtime_nsec)) {
+ cmp_mtime = replies[i].poststat.ia_mtime;
+ cmp_mtime_nsec = replies[i].poststat.ia_mtime_nsec;
+ fav_child = i;
+ }
+ }
+ }
+ return fav_child;
}
/*
* afr_sh_fav_by_ctime: Choose favorite child by ctime.
*/
int
-afr_sh_fav_by_ctime (xlator_t *this, struct afr_reply *replies, inode_t *inode)
-{
- afr_private_t *priv;
- int fav_child = -1;
- int i = 0;
- uint32_t cmp_ctime = 0;
- uint32_t cmp_ctime_nsec = 0;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (replies[i].valid == 1) {
- gf_msg_debug (this->name, 0, "Child:%s "
- "ctime = %ld, ctime_nsec = %d for gfid %s",
- priv->children[i]->name,
- replies[i].poststat.ia_ctime,
- replies[i].poststat.ia_ctime_nsec,
- uuid_utoa (inode->gfid));
- if (replies[i].poststat.ia_ctime > cmp_ctime) {
- cmp_ctime = replies[i].poststat.ia_ctime;
- cmp_ctime_nsec =
- replies[i].poststat.ia_ctime_nsec;
- fav_child = i;
- } else if ((replies[i].poststat.ia_ctime == cmp_ctime)
- && (replies[i].poststat.ia_ctime_nsec >
- cmp_ctime_nsec)) {
- cmp_ctime = replies[i].poststat.ia_ctime;
- cmp_ctime_nsec =
- replies[i].poststat.ia_ctime_nsec;
- fav_child = i;
- }
- }
- }
- return fav_child;
+afr_sh_fav_by_ctime(xlator_t *this, struct afr_reply *replies, inode_t *inode)
+{
+ afr_private_t *priv;
+ int fav_child = -1;
+ int i = 0;
+ uint32_t cmp_ctime = 0;
+ uint32_t cmp_ctime_nsec = 0;
+
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (replies[i].valid == 1) {
+ gf_msg_debug(this->name, 0,
+ "Child:%s ctime = %" PRId64
+ ", ctime_nsec = %d for "
+ "gfid %s",
+ priv->children[i]->name, replies[i].poststat.ia_ctime,
+ replies[i].poststat.ia_ctime_nsec,
+ uuid_utoa(inode->gfid));
+ if (replies[i].poststat.ia_ctime > cmp_ctime) {
+ cmp_ctime = replies[i].poststat.ia_ctime;
+ cmp_ctime_nsec = replies[i].poststat.ia_ctime_nsec;
+ fav_child = i;
+ } else if ((replies[i].poststat.ia_ctime == cmp_ctime) &&
+ (replies[i].poststat.ia_ctime_nsec > cmp_ctime_nsec)) {
+ cmp_ctime = replies[i].poststat.ia_ctime;
+ cmp_ctime_nsec = replies[i].poststat.ia_ctime_nsec;
+ fav_child = i;
+ }
+ }
+ }
+ return fav_child;
}
/*
- * afr_sh_fav_by_size: Choose favorite child by size.
+ * afr_sh_fav_by_size: Choose favorite child by size
+ * when not all files are of zero size.
*/
int
-afr_sh_fav_by_size (xlator_t *this, struct afr_reply *replies, inode_t *inode)
-{
- afr_private_t *priv;
- int fav_child = -1;
- int i = 0;
- uint64_t cmp_sz = 0;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (replies[i].valid == 1) {
- gf_msg_debug (this->name, 0, "Child:%s "
- "file size = %lu for gfid %s",
- priv->children[i]->name,
- replies[i].poststat.ia_size,
- uuid_utoa (inode->gfid));
- if (replies[i].poststat.ia_size > cmp_sz) {
- cmp_sz = replies[i].poststat.ia_size;
- fav_child = i;
- }
- }
- }
- return fav_child;
+afr_sh_fav_by_size(xlator_t *this, struct afr_reply *replies, inode_t *inode)
+{
+ afr_private_t *priv;
+ int fav_child = -1;
+ int i = 0;
+ uint64_t cmp_sz = 0;
+
+ priv = this->private;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid) {
+ continue;
+ }
+ gf_msg_debug(this->name, 0,
+ "Child:%s file size = %" PRIu64 " for gfid %s",
+ priv->children[i]->name, replies[i].poststat.ia_size,
+ uuid_utoa(inode->gfid));
+ if (replies[i].poststat.ia_type == IA_IFDIR) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SBRAIN_FAV_CHILD_POLICY,
+ "Cannot perform selfheal on %s. "
+ "Size policy is not applicable to directories.",
+ uuid_utoa(inode->gfid));
+ break;
+ }
+ if (replies[i].poststat.ia_size > cmp_sz) {
+ cmp_sz = replies[i].poststat.ia_size;
+ fav_child = i;
+ } else if (replies[i].poststat.ia_size == cmp_sz) {
+ fav_child = -1;
+ }
+ }
+ if (fav_child == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "No bigger file");
+ }
+ return fav_child;
}
int
-afr_sh_get_fav_by_policy (xlator_t *this, struct afr_reply *replies,
- inode_t *inode, char **policy_str)
+afr_sh_get_fav_by_policy(xlator_t *this, struct afr_reply *replies,
+ inode_t *inode, char **policy_str)
{
- afr_private_t *priv = NULL;
- int fav_child = -1;
+ afr_private_t *priv = NULL;
+ int fav_child = -1;
- priv = this->private;
- if (!afr_can_decide_split_brain_source_sinks (replies,
- priv->child_count)) {
- return -1;
- }
+ priv = this->private;
+ if (!afr_can_decide_split_brain_source_sinks(replies, priv->child_count)) {
+ return -1;
+ }
- switch (priv->fav_child_policy) {
+ switch (priv->fav_child_policy) {
case AFR_FAV_CHILD_BY_SIZE:
- fav_child = afr_sh_fav_by_size (this, replies, inode);
- if (policy_str && fav_child >= 0) {
- *policy_str = "SIZE";
- }
- break;
+ fav_child = afr_sh_fav_by_size(this, replies, inode);
+ if (policy_str && fav_child >= 0) {
+ *policy_str = "SIZE";
+ }
+ break;
case AFR_FAV_CHILD_BY_CTIME:
- fav_child = afr_sh_fav_by_ctime (this, replies, inode);
- if (policy_str && fav_child >= 0) {
- *policy_str = "CTIME";
- }
- break;
+ fav_child = afr_sh_fav_by_ctime(this, replies, inode);
+ if (policy_str && fav_child >= 0) {
+ *policy_str = "CTIME";
+ }
+ break;
case AFR_FAV_CHILD_BY_MTIME:
- fav_child = afr_sh_fav_by_mtime (this, replies, inode);
- if (policy_str && fav_child >= 0) {
- *policy_str = "MTIME";
- }
- break;
+ fav_child = afr_sh_fav_by_mtime(this, replies, inode);
+ if (policy_str && fav_child >= 0) {
+ *policy_str = "MTIME";
+ }
+ break;
case AFR_FAV_CHILD_BY_MAJORITY:
- fav_child = afr_sh_fav_by_majority (this, replies, inode);
- if (policy_str && fav_child >= 0) {
- *policy_str = "MAJORITY";
- }
- break;
+ fav_child = afr_sh_fav_by_majority(this, replies, inode);
+ if (policy_str && fav_child >= 0) {
+ *policy_str = "MAJORITY";
+ }
+ break;
case AFR_FAV_CHILD_NONE:
default:
- break;
- }
+ break;
+ }
- return fav_child;
+ return fav_child;
}
int
-afr_mark_split_brain_source_sinks_by_policy (call_frame_t *frame,
- xlator_t *this,
- inode_t *inode,
- unsigned char *sources,
- unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *locked_on,
- struct afr_reply *replies,
- afr_transaction_type type)
-{
- afr_private_t *priv = NULL;
- int fav_child = -1;
- char mtime_str[256];
- char ctime_str[256];
- char *policy_str = NULL;
- struct tm *tm_ptr;
- time_t time;
-
- priv = this->private;
-
- fav_child = afr_sh_get_fav_by_policy (this, replies, inode,
- &policy_str);
- if (fav_child > priv->child_count - 1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SBRAIN_FAV_CHILD_POLICY, "Invalid child (%d) "
- "selected by policy %s.", fav_child, policy_str);
- } else if (fav_child >= 0) {
- time = replies[fav_child].poststat.ia_mtime;
- tm_ptr = localtime (&time);
- strftime (mtime_str, sizeof (mtime_str), "%Y-%m-%d %H:%M:%S",
- tm_ptr);
- time = replies[fav_child].poststat.ia_ctime;
- tm_ptr = localtime (&time);
- strftime (ctime_str, sizeof (ctime_str), "%Y-%m-%d %H:%M:%S",
- tm_ptr);
-
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_SBRAIN_FAV_CHILD_POLICY, "Source %s "
- "selected as authentic to resolve conflicting "
- "data in file (gfid:%s) by %s (%lu bytes @ %s mtime, "
- "%s ctime).",
- priv->children[fav_child]->name,
- uuid_utoa (inode->gfid),
- policy_str,
- replies[fav_child].poststat.ia_size,
- mtime_str,
- ctime_str);
-
- sources[fav_child] = 1;
- sinks[fav_child] = 0;
- healed_sinks[fav_child] = 0;
- }
- return fav_child;
+afr_mark_split_brain_source_sinks_by_policy(
+ call_frame_t *frame, xlator_t *this, inode_t *inode, unsigned char *sources,
+ unsigned char *sinks, unsigned char *healed_sinks, unsigned char *locked_on,
+ struct afr_reply *replies, afr_transaction_type type)
+{
+ afr_private_t *priv = NULL;
+ int fav_child = -1;
+ char mtime_str[256];
+ char ctime_str[256];
+ char *policy_str = NULL;
+ struct tm *tm_ptr;
+ time_t time;
+
+ priv = this->private;
+
+ fav_child = afr_sh_get_fav_by_policy(this, replies, inode, &policy_str);
+ if (fav_child == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SBRAIN_FAV_CHILD_POLICY,
+ "No child selected by favorite-child policy.");
+ } else if (fav_child > priv->child_count - 1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SBRAIN_FAV_CHILD_POLICY,
+ "Invalid child (%d) "
+ "selected by policy %s.",
+ fav_child, policy_str);
+ } else if (fav_child >= 0) {
+ time = replies[fav_child].poststat.ia_mtime;
+ tm_ptr = localtime(&time);
+ strftime(mtime_str, sizeof(mtime_str), "%Y-%m-%d %H:%M:%S", tm_ptr);
+ time = replies[fav_child].poststat.ia_ctime;
+ tm_ptr = localtime(&time);
+ strftime(ctime_str, sizeof(ctime_str), "%Y-%m-%d %H:%M:%S", tm_ptr);
+
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_SBRAIN_FAV_CHILD_POLICY,
+ "Source %s selected as authentic to resolve conflicting data "
+ "in file (gfid:%s) by %s (%" PRIu64
+ " bytes @ %s mtime, %s "
+ "ctime).",
+ priv->children[fav_child]->name, uuid_utoa(inode->gfid),
+ policy_str, replies[fav_child].poststat.ia_size, mtime_str,
+ ctime_str);
+
+ sources[fav_child] = 1;
+ sinks[fav_child] = 0;
+ healed_sinks[fav_child] = 0;
+ }
+ return fav_child;
}
-int
-afr_mark_source_sinks_if_file_empty (xlator_t *this, unsigned char *sources,
- unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *locked_on,
- struct afr_reply *replies,
- afr_transaction_type type)
-{
- int source = -1;
- int i = 0;
- afr_private_t *priv = this->private;
- struct iatt stbuf = {0, };
-
- if ((AFR_COUNT (locked_on, priv->child_count) < priv->child_count) ||
- (afr_success_count(replies, priv->child_count) < priv->child_count))
- return -1;
+gf_boolean_t
+afr_is_file_empty_on_all_children(afr_private_t *priv,
+ struct afr_reply *replies)
+{
+ int i = 0;
- if (type == AFR_DATA_TRANSACTION) {
- for (i = 0; i < priv->child_count; i++) {
- if (replies[i].poststat.ia_size != 0)
- return -1;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if ((!replies[i].valid) || (replies[i].op_ret != 0) ||
+ (replies[i].poststat.ia_size != 0))
+ return _gf_false;
+ }
- goto mark;
- }
+ return _gf_true;
+}
- /*For AFR_METADATA_TRANSACTION, metadata must be same on all bricks.*/
- stbuf = replies[0].poststat;
- for (i = 1; i < priv->child_count; i++) {
- if ((!IA_EQUAL (stbuf, replies[i].poststat, type)) ||
- (!IA_EQUAL (stbuf, replies[i].poststat, uid)) ||
- (!IA_EQUAL (stbuf, replies[i].poststat, gid)) ||
- (!IA_EQUAL (stbuf, replies[i].poststat, prot)))
- return -1;
- }
- for (i = 1; i < priv->child_count; i++) {
- if (!afr_xattrs_are_equal (replies[0].xdata,
- replies[i].xdata))
- return -1;
- }
+int
+afr_mark_source_sinks_if_file_empty(xlator_t *this, unsigned char *sources,
+ unsigned char *sinks,
+ unsigned char *healed_sinks,
+ unsigned char *locked_on,
+ struct afr_reply *replies,
+ afr_transaction_type type)
+{
+ int source = -1;
+ int i = 0;
+ afr_private_t *priv = this->private;
+ struct iatt stbuf = {
+ 0,
+ };
+
+ if ((AFR_COUNT(locked_on, priv->child_count) < priv->child_count) ||
+ (afr_success_count(replies, priv->child_count) < priv->child_count))
+ return -1;
+
+ if (type == AFR_DATA_TRANSACTION) {
+ if (!afr_is_file_empty_on_all_children(priv, replies))
+ return -1;
+ goto mark;
+ }
+
+ /*For AFR_METADATA_TRANSACTION, metadata must be same on all bricks.*/
+ stbuf = replies[0].poststat;
+ for (i = 1; i < priv->child_count; i++) {
+ if ((!IA_EQUAL(stbuf, replies[i].poststat, type)) ||
+ (!IA_EQUAL(stbuf, replies[i].poststat, uid)) ||
+ (!IA_EQUAL(stbuf, replies[i].poststat, gid)) ||
+ (!IA_EQUAL(stbuf, replies[i].poststat, prot)))
+ return -1;
+ }
+ for (i = 1; i < priv->child_count; i++) {
+ if (!afr_xattrs_are_equal(replies[0].xdata, replies[i].xdata))
+ return -1;
+ }
mark:
- /* data/metadata is same on all bricks. Pick one of them as source. Rest
- * are sinks.*/
- for (i = 0 ; i < priv->child_count; i++) {
- if (source == -1) {
- source = i;
- sources[i] = 1;
- sinks[i] = 0;
- healed_sinks[i] = 0;
- continue;
- }
- sources[i] = 0;
- sinks[i] = 1;
- healed_sinks[i] = 1;
+ /* data/metadata is same on all bricks. Pick one of them as source. Rest
+ * are sinks.*/
+ for (i = 0; i < priv->child_count; i++) {
+ if (source == -1) {
+ source = i;
+ sources[i] = 1;
+ sinks[i] = 0;
+ healed_sinks[i] = 0;
+ continue;
}
+ sources[i] = 0;
+ sinks[i] = 1;
+ healed_sinks[i] = 1;
+ }
- return source;
+ return source;
}
/* Return a source depending on the type of heal_op, and set sources[source],
@@ -1324,171 +1389,156 @@ mark:
* sinks[node] are 1. This should be the case if the file is in split-brain.
*/
int
-afr_mark_split_brain_source_sinks (call_frame_t *frame, xlator_t *this,
- inode_t *inode,
- unsigned char *sources,
- unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *locked_on,
- struct afr_reply *replies,
- afr_transaction_type type)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- dict_t *xdata_req = NULL;
- int heal_op = -1;
- int ret = -1;
- int source = -1;
-
- local = frame->local;
- priv = this->private;
- xdata_req = local->xdata_req;
-
- source = afr_mark_source_sinks_if_file_empty (this, sources, sinks,
- healed_sinks, locked_on,
- replies, type);
- if (source >= 0)
- return source;
-
- ret = dict_get_int32 (xdata_req, "heal-op", &heal_op);
- if (ret)
- goto autoheal;
-
- source = afr_mark_split_brain_source_sinks_by_heal_op (frame, this,
- sources, sinks,
- healed_sinks,
- locked_on, replies,
- type, heal_op);
+afr_mark_split_brain_source_sinks(
+ call_frame_t *frame, xlator_t *this, inode_t *inode, unsigned char *sources,
+ unsigned char *sinks, unsigned char *healed_sinks, unsigned char *locked_on,
+ struct afr_reply *replies, afr_transaction_type type)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ dict_t *xdata_req = NULL;
+ int heal_op = -1;
+ int ret = -1;
+ int source = -1;
+
+ local = frame->local;
+ priv = this->private;
+ xdata_req = local->xdata_req;
+
+ source = afr_mark_source_sinks_if_file_empty(
+ this, sources, sinks, healed_sinks, locked_on, replies, type);
+ if (source >= 0)
return source;
+ ret = dict_get_int32_sizen(xdata_req, "heal-op", &heal_op);
+ if (ret)
+ goto autoheal;
+
+ source = afr_mark_split_brain_source_sinks_by_heal_op(
+ frame, this, sources, sinks, healed_sinks, locked_on, replies, type,
+ heal_op);
+ return source;
+
autoheal:
- /* Automatically heal if fav_child_policy is set. */
- if (priv->fav_child_policy != AFR_FAV_CHILD_NONE) {
- source = afr_mark_split_brain_source_sinks_by_policy (frame,
- this,
- inode,
- sources,
- sinks,
- healed_sinks,
- locked_on,
- replies,
- type);
- if (source != -1) {
- ret = dict_set_int32 (xdata_req, "fav-child-policy", 1);
- if (ret)
- return -1;
- }
+ /* Automatically heal if fav_child_policy is set. */
+ if (priv->fav_child_policy != AFR_FAV_CHILD_NONE) {
+ source = afr_mark_split_brain_source_sinks_by_policy(
+ frame, this, inode, sources, sinks, healed_sinks, locked_on,
+ replies, type);
+ if (source != -1) {
+ ret = dict_set_int32_sizen(xdata_req, "fav-child-policy", 1);
+ if (ret)
+ return -1;
}
+ }
- return source;
+ return source;
}
int
-_afr_fav_child_reset_sink_xattrs (call_frame_t *frame, xlator_t *this,
- inode_t *inode, int source,
- unsigned char *healed_sinks,
- unsigned char *undid_pending,
- afr_transaction_type type,
- unsigned char *locked_on,
- struct afr_reply *replies)
+_afr_fav_child_reset_sink_xattrs(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, int source,
+ unsigned char *healed_sinks,
+ unsigned char *undid_pending,
+ afr_transaction_type type,
+ unsigned char *locked_on,
+ struct afr_reply *replies)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int *input_dirty = NULL;
- int **input_matrix = NULL;
- int *output_dirty = NULL;
- int **output_matrix = NULL;
- dict_t *xattr = NULL;
- dict_t *xdata = NULL;
- int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int *input_dirty = NULL;
+ int **input_matrix = NULL;
+ int *output_dirty = NULL;
+ int **output_matrix = NULL;
+ dict_t *xattr = NULL;
+ dict_t *xdata = NULL;
+ int i = 0;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- if (!dict_get (local->xdata_req, "fav-child-policy"))
- return 0;
+ if (!dict_get_sizen(local->xdata_req, "fav-child-policy"))
+ return 0;
- xdata = dict_new();
- if (!xdata)
- return -1;
+ xdata = dict_new();
+ if (!xdata)
+ return -1;
- input_dirty = alloca0 (priv->child_count * sizeof (int));
- input_matrix = ALLOC_MATRIX (priv->child_count, int);
- output_dirty = alloca0 (priv->child_count * sizeof (int));
- output_matrix = ALLOC_MATRIX (priv->child_count, int);
+ input_dirty = alloca0(priv->child_count * sizeof(int));
+ input_matrix = ALLOC_MATRIX(priv->child_count, int);
+ output_dirty = alloca0(priv->child_count * sizeof(int));
+ output_matrix = ALLOC_MATRIX(priv->child_count, int);
- afr_selfheal_extract_xattr (this, replies, type, input_dirty,
- input_matrix);
+ afr_selfheal_extract_xattr(this, replies, type, input_dirty, input_matrix);
- for (i = 0; i < priv->child_count; i++) {
- if (i == source || !healed_sinks[i])
- continue;
- output_dirty[i] = -input_dirty[i];
- output_matrix[i][source] = -input_matrix[i][source];
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == source || !healed_sinks[i])
+ continue;
+ output_dirty[i] = -input_dirty[i];
+ output_matrix[i][source] = -input_matrix[i][source];
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (!healed_sinks[i] || !locked_on[i])
- continue;
- xattr = afr_selfheal_output_xattr (this, _gf_false, type,
- output_dirty, output_matrix,
- i, NULL);
+ for (i = 0; i < priv->child_count; i++) {
+ if (!healed_sinks[i] || !locked_on[i])
+ continue;
+ xattr = afr_selfheal_output_xattr(this, _gf_false, type, output_dirty,
+ output_matrix, i, NULL);
- afr_selfheal_post_op (frame, this, inode, i, xattr, xdata);
+ afr_selfheal_post_op(frame, this, inode, i, xattr, xdata);
- undid_pending[i] = 1;
- dict_unref (xattr);
- }
+ undid_pending[i] = 1;
+ dict_unref(xattr);
+ }
- if (xdata)
- dict_unref (xdata);
+ if (xdata)
+ dict_unref(xdata);
- return 0;
+ return 0;
}
gf_boolean_t
-afr_does_witness_exist (xlator_t *this, uint64_t *witness)
+afr_does_witness_exist(xlator_t *this, uint64_t *witness)
{
- int i = 0;
- afr_private_t *priv = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (witness[i])
- return _gf_true;
- }
- return _gf_false;
+ for (i = 0; i < priv->child_count; i++) {
+ if (witness[i])
+ return _gf_true;
+ }
+ return _gf_false;
}
unsigned int
-afr_get_quorum_count (afr_private_t *priv)
+afr_get_quorum_count(afr_private_t *priv)
{
- if (priv->quorum_count == AFR_QUORUM_AUTO) {
- return priv->child_count/2 + 1;
- } else {
- return priv->quorum_count;
- }
+ if (priv->quorum_count == AFR_QUORUM_AUTO) {
+ return priv->child_count / 2 + 1;
+ } else {
+ return priv->quorum_count;
+ }
}
void
-afr_selfheal_post_op_failure_accounting (afr_private_t *priv, char *accused,
- unsigned char *sources,
- unsigned char *locked_on)
+afr_selfheal_post_op_failure_accounting(afr_private_t *priv, char *accused,
+ unsigned char *sources,
+ unsigned char *locked_on)
{
- int i = 0;
- unsigned int quorum_count = 0;
+ int i = 0;
+ unsigned int quorum_count = 0;
- if (AFR_COUNT (sources, priv->child_count) != 0)
- return;
+ if (AFR_COUNT(sources, priv->child_count) != 0)
+ return;
- quorum_count = afr_get_quorum_count (priv);
- for (i = 0; i < priv->child_count; i++) {
- if ((accused[i] < quorum_count) && locked_on[i]) {
- sources[i] = 1;
- }
+ quorum_count = afr_get_quorum_count(priv);
+ for (i = 0; i < priv->child_count; i++) {
+ if ((accused[i] < quorum_count) && locked_on[i]) {
+ sources[i] = 1;
}
- return;
+ }
+ return;
}
/*
@@ -1511,700 +1561,711 @@ afr_selfheal_post_op_failure_accounting (afr_private_t *priv, char *accused,
*/
int
-afr_selfheal_find_direction (call_frame_t *frame, xlator_t *this,
- struct afr_reply *replies,
- afr_transaction_type type,
- unsigned char *locked_on, unsigned char *sources,
- unsigned char *sinks, uint64_t *witness,
- gf_boolean_t *pflag)
-{
- afr_private_t *priv = NULL;
- int i = 0;
- int j = 0;
- int *dirty = NULL; /* Denotes if dirty xattr is set */
- int **matrix = NULL;/* Changelog matrix */
- char *accused = NULL;/* Accused others without any self-accusal */
- char *pending = NULL;/* Have pending operations on others */
- char *self_accused = NULL; /* Accused itself */
- int min_participants = -1;
-
- priv = this->private;
-
- dirty = alloca0 (priv->child_count * sizeof (int));
- accused = alloca0 (priv->child_count);
- pending = alloca0 (priv->child_count);
- self_accused = alloca0 (priv->child_count);
- matrix = ALLOC_MATRIX(priv->child_count, int);
- memset (witness, 0, sizeof (*witness) * priv->child_count);
-
- /* First construct the pending matrix for further analysis */
- afr_selfheal_extract_xattr (this, replies, type, dirty, matrix);
-
- if (pflag) {
- for (i = 0; i < priv->child_count; i++) {
- for (j = 0; j < priv->child_count; j++)
- if (matrix[i][j])
- *pflag = _gf_true;
- if (*pflag)
- break;
- }
+afr_selfheal_find_direction(call_frame_t *frame, xlator_t *this,
+ struct afr_reply *replies,
+ afr_transaction_type type, unsigned char *locked_on,
+ unsigned char *sources, unsigned char *sinks,
+ uint64_t *witness, unsigned char *pflag)
+{
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int j = 0;
+ int *dirty = NULL; /* Denotes if dirty xattr is set */
+ int **matrix = NULL; /* Changelog matrix */
+ char *accused = NULL; /* Accused others without any self-accusal */
+ char *pending = NULL; /* Have pending operations on others */
+ char *self_accused = NULL; /* Accused itself */
+
+ priv = this->private;
+
+ dirty = alloca0(priv->child_count * sizeof(int));
+ accused = alloca0(priv->child_count);
+ pending = alloca0(priv->child_count);
+ self_accused = alloca0(priv->child_count);
+ matrix = ALLOC_MATRIX(priv->child_count, int);
+ memset(witness, 0, sizeof(*witness) * priv->child_count);
+
+ /* First construct the pending matrix for further analysis */
+ afr_selfheal_extract_xattr(this, replies, type, dirty, matrix);
+
+ if (pflag) {
+ for (i = 0; i < priv->child_count; i++) {
+ for (j = 0; j < priv->child_count; j++)
+ if (matrix[i][j])
+ *pflag |= PFLAG_PENDING;
+ if (*pflag)
+ break;
}
+ }
- if (type == AFR_DATA_TRANSACTION) {
- min_participants = priv->child_count;
- } else {
- min_participants = AFR_SH_MIN_PARTICIPANTS;
- }
- if (afr_success_count (replies,
- priv->child_count) < min_participants) {
- /* Treat this just like locks not being acquired */
- return -ENOTCONN;
+ if (afr_success_count(replies, priv->child_count) < priv->child_count) {
+ /* Treat this just like locks not being acquired */
+ return -ENOTCONN;
+ }
+
+ /* short list all self-accused */
+ for (i = 0; i < priv->child_count; i++) {
+ if (matrix[i][i])
+ self_accused[i] = 1;
+ }
+
+ /* Next short list all accused to exclude them from being sources */
+ /* Self-accused can't accuse others as they are FOOLs */
+ for (i = 0; i < priv->child_count; i++) {
+ for (j = 0; j < priv->child_count; j++) {
+ if (matrix[i][j]) {
+ if (!self_accused[i])
+ accused[j] += 1;
+ if (i != j)
+ pending[i] += 1;
+ }
}
+ }
- /* short list all self-accused */
+ /* Short list all non-accused as sources */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!accused[i] && locked_on[i])
+ sources[i] = 1;
+ else
+ sources[i] = 0;
+ }
+
+ /* Everyone accused by non-self-accused sources are sinks */
+ memset(sinks, 0, priv->child_count);
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ if (self_accused[i])
+ continue;
+ for (j = 0; j < priv->child_count; j++) {
+ if (matrix[i][j])
+ sinks[j] = 1;
+ }
+ }
+
+ /* For breaking ties provide with number of fops they witnessed */
+
+ /*
+ * count the pending fops witnessed from itself to others when it is
+ * self-accused
+ */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!self_accused[i])
+ continue;
+ for (j = 0; j < priv->child_count; j++) {
+ if (i == j)
+ continue;
+ witness[i] += matrix[i][j];
+ }
+ }
+
+ if (type == AFR_DATA_TRANSACTION || type == AFR_METADATA_TRANSACTION)
+ afr_selfheal_post_op_failure_accounting(priv, accused, sources,
+ locked_on);
+
+ /* If no sources, all locked nodes are sinks - split brain */
+ if (AFR_COUNT(sources, priv->child_count) == 0) {
for (i = 0; i < priv->child_count; i++) {
- if (matrix[i][i])
- self_accused[i] = 1;
+ if (locked_on[i])
+ sinks[i] = 1;
}
+ if (pflag)
+ *pflag |= PFLAG_SBRAIN;
+ }
- /* Next short list all accused to exclude them from being sources */
- /* Self-accused can't accuse others as they are FOOLs */
- for (i = 0; i < priv->child_count; i++) {
- for (j = 0; j < priv->child_count; j++) {
- if (matrix[i][j]) {
- if (!self_accused[i])
- accused[j] += 1;
- if (i != j)
- pending[i] += 1;
- }
- }
- }
-
- /* Short list all non-accused as sources */
- memset (sources, 0, priv->child_count);
- for (i = 0; i < priv->child_count; i++) {
- if (!accused[i] && locked_on[i])
- sources[i] = 1;
- }
-
- /* Everyone accused by non-self-accused sources are sinks */
- memset (sinks, 0, priv->child_count);
+ /* One more class of witness similar to dirty in v2 is where no pending
+ * exists but we have self-accusing markers. This can happen in afr-v1
+ * if the brick crashes just after doing xattrop on self but
+ * before xattrop on the other xattrs on the brick in pre-op. */
+ if (AFR_COUNT(pending, priv->child_count) == 0) {
for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
- if (self_accused[i])
- continue;
- for (j = 0; j < priv->child_count; j++) {
- if (matrix[i][j])
- sinks[j] = 1;
- }
+ if (self_accused[i])
+ witness[i] += matrix[i][i];
}
-
- /* For breaking ties provide with number of fops they witnessed */
-
- /*
- * count the pending fops witnessed from itself to others when it is
- * self-accused
+ } else {
+ /* In afr-v1 if a file is self-accused and has pending
+ * operations on others then it is similar to 'dirty' in afr-v2.
+ * Consider such cases as witness.
*/
for (i = 0; i < priv->child_count; i++) {
- if (!self_accused[i])
- continue;
- for (j = 0; j < priv->child_count; j++) {
- if (i == j)
- continue;
- witness[i] += matrix[i][j];
- }
+ if (self_accused[i] && pending[i])
+ witness[i] += matrix[i][i];
}
+ }
- if (type == AFR_DATA_TRANSACTION)
- afr_selfheal_post_op_failure_accounting (priv, accused,
- sources, locked_on);
-
- /* If no sources, all locked nodes are sinks - split brain */
- if (AFR_COUNT (sources, priv->child_count) == 0) {
- for (i = 0; i < priv->child_count; i++) {
- if (locked_on[i])
- sinks[i] = 1;
- }
- }
+ /* count the number of dirty fops witnessed */
+ for (i = 0; i < priv->child_count; i++)
+ witness[i] += dirty[i];
- /* One more class of witness similar to dirty in v2 is where no pending
- * exists but we have self-accusing markers. This can happen in afr-v1
- * if the brick crashes just after doing xattrop on self but
- * before xattrop on the other xattrs on the brick in pre-op. */
- if (AFR_COUNT (pending, priv->child_count) == 0) {
- for (i = 0; i < priv->child_count; i++) {
- if (self_accused[i])
- witness[i] += matrix[i][i];
- }
- } else {
- /* In afr-v1 if a file is self-accused and has pending
- * operations on others then it is similar to 'dirty' in afr-v2.
- * Consider such cases as witness.
- */
- for (i = 0; i < priv->child_count; i++) {
- if (self_accused[i] && pending[i])
- witness[i] += matrix[i][i];
- }
- }
-
-
- /* count the number of dirty fops witnessed */
- for (i = 0; i < priv->child_count; i++)
- witness[i] += dirty[i];
-
- return 0;
+ return 0;
}
void
-afr_log_selfheal (uuid_t gfid, xlator_t *this, int ret, char *type,
- int source, unsigned char *sources,
- unsigned char *healed_sinks)
-{
- char *status = NULL;
- char *sinks_str = NULL;
- char *p = NULL;
- char *sources_str = NULL;
- char *q = NULL;
- afr_private_t *priv = NULL;
- gf_loglevel_t loglevel = GF_LOG_NONE;
- int i = 0;
-
- priv = this->private;
- sinks_str = alloca0 (priv->child_count * 8);
- p = sinks_str;
- sources_str = alloca0 (priv->child_count * 8);
- q = sources_str;
- for (i = 0; i < priv->child_count; i++) {
- if (healed_sinks[i])
- p += sprintf (p, "%d ", i);
- if (sources[i]) {
- if (source == i) {
- q += sprintf (q, "[%d] ", i);
- } else {
- q += sprintf (q, "%d ", i);
- }
- }
- }
-
- if (ret < 0) {
- status = "Failed";
- loglevel = GF_LOG_DEBUG;
- } else {
- status = "Completed";
- loglevel = GF_LOG_INFO;
- }
-
- gf_msg (this->name, loglevel, 0,
- AFR_MSG_SELF_HEAL_INFO, "%s %s selfheal on %s. "
- "sources=%s sinks=%s", status, type, uuid_utoa (gfid),
- sources_str, sinks_str);
+afr_log_selfheal(uuid_t gfid, xlator_t *this, int ret, char *type, int source,
+ unsigned char *sources, unsigned char *healed_sinks)
+{
+ char *status = NULL;
+ char *sinks_str = NULL;
+ char *p = NULL;
+ char *sources_str = NULL;
+ char *q = NULL;
+ afr_private_t *priv = NULL;
+ gf_loglevel_t loglevel = GF_LOG_NONE;
+ int i = 0;
+
+ priv = this->private;
+ sinks_str = alloca0(priv->child_count * 8);
+ p = sinks_str;
+ sources_str = alloca0(priv->child_count * 8);
+ q = sources_str;
+ for (i = 0; i < priv->child_count; i++) {
+ if (healed_sinks[i])
+ p += sprintf(p, "%d ", i);
+ if (sources[i]) {
+ if (source == i) {
+ q += sprintf(q, "[%d] ", i);
+ } else {
+ q += sprintf(q, "%d ", i);
+ }
+ }
+ }
+
+ if (ret < 0) {
+ status = "Failed";
+ loglevel = GF_LOG_DEBUG;
+ } else {
+ status = "Completed";
+ loglevel = GF_LOG_INFO;
+ }
+
+ gf_msg(this->name, loglevel, 0, AFR_MSG_SELF_HEAL_INFO,
+ "%s %s selfheal on %s. "
+ "sources=%s sinks=%s",
+ status, type, uuid_utoa(gfid), sources_str, sinks_str);
}
int
-afr_selfheal_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata, struct iatt *parbuf)
-{
- afr_local_t *local = NULL;
- int i = -1;
- GF_UNUSED int ret = -1;
- int8_t need_heal = 1;
-
- local = frame->local;
- i = (long) cookie;
-
- local->replies[i].valid = 1;
- local->replies[i].op_ret = op_ret;
- local->replies[i].op_errno = op_errno;
- if (buf)
- local->replies[i].poststat = *buf;
- if (parbuf)
- local->replies[i].postparent = *parbuf;
- if (xdata) {
- local->replies[i].xdata = dict_ref (xdata);
- ret = dict_get_int8 (xdata, "link-count", &need_heal);
- local->replies[i].need_heal = need_heal;
- } else {
- local->replies[i].need_heal = need_heal;
- }
-
- syncbarrier_wake (&local->barrier);
+afr_selfheal_discover_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata, struct iatt *parbuf)
+{
+ afr_local_t *local = NULL;
+ int i = -1;
+ GF_UNUSED int ret = -1;
+ int8_t need_heal = 1;
- return 0;
-}
+ local = frame->local;
+ i = (long)cookie;
+ local->replies[i].valid = 1;
+ local->replies[i].op_ret = op_ret;
+ local->replies[i].op_errno = op_errno;
+ if (buf)
+ local->replies[i].poststat = *buf;
+ if (parbuf)
+ local->replies[i].postparent = *parbuf;
+ if (xdata) {
+ local->replies[i].xdata = dict_ref(xdata);
+ ret = dict_get_int8(xdata, "link-count", &need_heal);
+ }
-inode_t *
-afr_selfheal_unlocked_lookup_on (call_frame_t *frame, inode_t *parent,
- const char *name, struct afr_reply *replies,
- unsigned char *lookup_on, dict_t *xattr)
-{
- loc_t loc = {0, };
- dict_t *xattr_req = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- inode_t *inode = NULL;
+ local->replies[i].need_heal = need_heal;
+ syncbarrier_wake(&local->barrier);
- local = frame->local;
- priv = frame->this->private;
+ return 0;
+}
- xattr_req = dict_new ();
- if (!xattr_req)
- return NULL;
+inode_t *
+afr_selfheal_unlocked_lookup_on(call_frame_t *frame, inode_t *parent,
+ const char *name, struct afr_reply *replies,
+ unsigned char *lookup_on, dict_t *xattr)
+{
+ loc_t loc = {
+ 0,
+ };
+ dict_t *xattr_req = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ inode_t *inode = NULL;
+
+ local = frame->local;
+ priv = frame->this->private;
+
+ xattr_req = dict_new();
+ if (!xattr_req)
+ return NULL;
- if (xattr)
- dict_copy (xattr, xattr_req);
+ if (xattr)
+ dict_copy(xattr, xattr_req);
- if (afr_xattr_req_prepare (frame->this, xattr_req) != 0) {
- dict_unref (xattr_req);
- return NULL;
- }
+ if (afr_xattr_req_prepare(frame->this, xattr_req) != 0) {
+ dict_unref(xattr_req);
+ return NULL;
+ }
- inode = inode_new (parent->table);
- if (!inode) {
- dict_unref (xattr_req);
- return NULL;
- }
+ inode = inode_new(parent->table);
+ if (!inode) {
+ dict_unref(xattr_req);
+ return NULL;
+ }
- loc.parent = inode_ref (parent);
- gf_uuid_copy (loc.pargfid, parent->gfid);
- loc.name = name;
- loc.inode = inode_ref (inode);
+ loc.parent = inode_ref(parent);
+ gf_uuid_copy(loc.pargfid, parent->gfid);
+ loc.name = name;
+ loc.inode = inode_ref(inode);
- AFR_ONLIST (lookup_on, frame, afr_selfheal_discover_cbk, lookup, &loc,
- xattr_req);
+ AFR_ONLIST(lookup_on, frame, afr_selfheal_discover_cbk, lookup, &loc,
+ xattr_req);
- afr_replies_copy (replies, local->replies, priv->child_count);
+ afr_replies_copy(replies, local->replies, priv->child_count);
- loc_wipe (&loc);
- dict_unref (xattr_req);
+ loc_wipe(&loc);
+ dict_unref(xattr_req);
- return inode;
+ return inode;
}
-int
-afr_selfheal_unlocked_discover_on (call_frame_t *frame, inode_t *inode,
- uuid_t gfid, struct afr_reply *replies,
- unsigned char *discover_on)
+static int
+afr_set_multi_dom_lock_count_request(xlator_t *this, dict_t *dict)
{
- loc_t loc = {0, };
- dict_t *xattr_req = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ int ret = 0;
+ afr_private_t *priv = NULL;
+ char *key1 = NULL;
+ char *key2 = NULL;
- local = frame->local;
- priv = frame->this->private;
+ priv = this->private;
+ key1 = alloca0(strlen(GLUSTERFS_INODELK_DOM_PREFIX) + 2 +
+ strlen(this->name));
+ key2 = alloca0(strlen(GLUSTERFS_INODELK_DOM_PREFIX) + 2 +
+ strlen(priv->sh_domain));
- xattr_req = dict_new ();
- if (!xattr_req)
- return -ENOMEM;
+ ret = dict_set_uint32(dict, GLUSTERFS_MULTIPLE_DOM_LK_CNT_REQUESTS, 1);
+ if (ret)
+ return ret;
- if (afr_xattr_req_prepare (frame->this, xattr_req) != 0) {
- dict_unref (xattr_req);
- return -ENOMEM;
- }
+ sprintf(key1, "%s:%s", GLUSTERFS_INODELK_DOM_PREFIX, this->name);
+ ret = dict_set_uint32(dict, key1, 1);
+ if (ret)
+ return ret;
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, gfid);
+ sprintf(key2, "%s:%s", GLUSTERFS_INODELK_DOM_PREFIX, priv->sh_domain);
+ ret = dict_set_uint32(dict, key2, 1);
+ if (ret)
+ return ret;
+
+ return 0;
+}
+
+int
+afr_selfheal_unlocked_discover_on(call_frame_t *frame, inode_t *inode,
+ uuid_t gfid, struct afr_reply *replies,
+ unsigned char *discover_on, dict_t *dict)
+{
+ loc_t loc = {
+ 0,
+ };
+ dict_t *xattr_req = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ local = frame->local;
+ priv = frame->this->private;
+
+ xattr_req = dict_new();
+ if (!xattr_req)
+ return -ENOMEM;
+ if (dict)
+ dict_copy(dict, xattr_req);
+
+ if (afr_xattr_req_prepare(frame->this, xattr_req) != 0) {
+ dict_unref(xattr_req);
+ return -ENOMEM;
+ }
+
+ if (afr_set_multi_dom_lock_count_request(frame->this, xattr_req)) {
+ dict_unref(xattr_req);
+ return -1;
+ }
- AFR_ONLIST (discover_on, frame, afr_selfheal_discover_cbk, lookup, &loc,
- xattr_req);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, gfid);
- afr_replies_copy (replies, local->replies, priv->child_count);
+ AFR_ONLIST(discover_on, frame, afr_selfheal_discover_cbk, lookup, &loc,
+ xattr_req);
- loc_wipe (&loc);
- dict_unref (xattr_req);
+ afr_replies_copy(replies, local->replies, priv->child_count);
- return 0;
+ loc_wipe(&loc);
+ dict_unref(xattr_req);
+
+ return 0;
}
int
-afr_selfheal_unlocked_discover (call_frame_t *frame, inode_t *inode,
- uuid_t gfid, struct afr_reply *replies)
+afr_selfheal_unlocked_discover(call_frame_t *frame, inode_t *inode, uuid_t gfid,
+ struct afr_reply *replies)
{
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ dict_t *dict = NULL;
+
+ local = frame->local;
- priv = frame->this->private;
+ if (local->xattr_req)
+ dict = local->xattr_req;
- return afr_selfheal_unlocked_discover_on (frame, inode, gfid, replies,
- priv->child_up);
+ return afr_selfheal_unlocked_discover_on(frame, inode, gfid, replies,
+ local->child_up, dict);
}
unsigned int
-afr_success_count (struct afr_reply *replies, unsigned int count)
+afr_success_count(struct afr_reply *replies, unsigned int count)
{
- int i = 0;
- unsigned int success = 0;
+ int i = 0;
+ unsigned int success = 0;
- for (i = 0; i < count; i++)
- if (replies[i].valid && replies[i].op_ret == 0)
- success++;
- return success;
+ for (i = 0; i < count; i++)
+ if (replies[i].valid && replies[i].op_ret == 0)
+ success++;
+ return success;
}
int
-afr_selfheal_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+afr_selfheal_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
{
- afr_local_t *local = NULL;
- int i = 0;
+ afr_local_t *local = NULL;
+ int i = 0;
- local = frame->local;
- i = (long) cookie;
+ local = frame->local;
+ i = (long)cookie;
- local->replies[i].valid = 1;
- local->replies[i].op_ret = op_ret;
- local->replies[i].op_errno = op_errno;
+ local->replies[i].valid = 1;
+ local->replies[i].op_ret = op_ret;
+ local->replies[i].op_errno = op_errno;
- syncbarrier_wake (&local->barrier);
+ syncbarrier_wake(&local->barrier);
- return 0;
+ return 0;
}
-
int
-afr_locked_fill (call_frame_t *frame, xlator_t *this,
- unsigned char *locked_on)
+afr_locked_fill(call_frame_t *frame, xlator_t *this, unsigned char *locked_on)
{
- int i = 0;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int count = 0;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int count = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (local->replies[i].valid && local->replies[i].op_ret == 0) {
- locked_on[i] = 1;
- count++;
- } else {
- locked_on[i] = 0;
- }
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->replies[i].valid && local->replies[i].op_ret == 0) {
+ locked_on[i] = 1;
+ count++;
+ } else {
+ locked_on[i] = 0;
+ }
+ }
- return count;
+ return count;
}
-
int
-afr_selfheal_tryinodelk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, off_t off, size_t size,
- unsigned char *locked_on)
+afr_selfheal_tryinodelk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, off_t off, size_t size,
+ unsigned char *locked_on)
{
- loc_t loc = {0,};
- struct gf_flock flock = {0, };
+ loc_t loc = {
+ 0,
+ };
+ struct gf_flock flock = {
+ 0,
+ };
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- flock.l_type = F_WRLCK;
- flock.l_start = off;
- flock.l_len = size;
+ flock.l_type = F_WRLCK;
+ flock.l_start = off;
+ flock.l_len = size;
- AFR_ONALL (frame, afr_selfheal_lock_cbk, inodelk, dom,
- &loc, F_SETLK, &flock, NULL);
+ AFR_ONALL(frame, afr_selfheal_lock_cbk, inodelk, dom, &loc, F_SETLK, &flock,
+ NULL);
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return afr_locked_fill (frame, this, locked_on);
+ return afr_locked_fill(frame, this, locked_on);
}
-
int
-afr_selfheal_inodelk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, off_t off, size_t size,
- unsigned char *locked_on)
+afr_selfheal_inodelk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, off_t off, size_t size,
+ unsigned char *locked_on)
{
- loc_t loc = {0,};
- struct gf_flock flock = {0, };
- afr_local_t *local = NULL;
- int i = 0;
- afr_private_t *priv = NULL;
+ loc_t loc = {
+ 0,
+ };
+ struct gf_flock flock = {
+ 0,
+ };
+ afr_local_t *local = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- flock.l_type = F_WRLCK;
- flock.l_start = off;
- flock.l_len = size;
+ flock.l_type = F_WRLCK;
+ flock.l_start = off;
+ flock.l_len = size;
- AFR_ONALL (frame, afr_selfheal_lock_cbk, inodelk, dom,
- &loc, F_SETLK, &flock, NULL);
+ AFR_ONALL(frame, afr_selfheal_lock_cbk, inodelk, dom, &loc, F_SETLK, &flock,
+ NULL);
- for (i = 0; i < priv->child_count; i++) {
- if (local->replies[i].op_ret == -1 &&
- local->replies[i].op_errno == EAGAIN) {
- afr_locked_fill (frame, this, locked_on);
- afr_selfheal_uninodelk (frame, this, inode, dom, off,
- size, locked_on);
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->replies[i].op_ret == -1 &&
+ local->replies[i].op_errno == EAGAIN) {
+ afr_locked_fill(frame, this, locked_on);
+ afr_selfheal_uninodelk(frame, this, inode, dom, off, size,
+ locked_on);
- AFR_SEQ (frame, afr_selfheal_lock_cbk, inodelk, dom,
- &loc, F_SETLKW, &flock, NULL);
- break;
- }
- }
+ AFR_SEQ(frame, afr_selfheal_lock_cbk, inodelk, dom, &loc, F_SETLKW,
+ &flock, NULL);
+ break;
+ }
+ }
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return afr_locked_fill (frame, this, locked_on);
+ return afr_locked_fill(frame, this, locked_on);
}
static void
-afr_get_lock_and_eagain_counts (afr_private_t *priv, struct afr_reply *replies,
- int *lock_count, int *eagain_count)
-{
- int i = 0;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
- if (replies[i].op_ret == 0) {
- (*lock_count)++;
- } else if (replies[i].op_ret == -1 &&
- replies[i].op_errno == EAGAIN) {
- (*eagain_count)++;
- }
- }
+afr_get_lock_and_eagain_counts(afr_private_t *priv, struct afr_reply *replies,
+ int *lock_count, int *eagain_count)
+{
+ int i = 0;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
+ if (replies[i].op_ret == 0) {
+ (*lock_count)++;
+ } else if (replies[i].op_ret == -1 && replies[i].op_errno == EAGAIN) {
+ (*eagain_count)++;
+ }
+ }
}
/*Do blocking locks if number of locks acquired is majority and there were some
* EAGAINs. Useful for odd-way replication*/
int
-afr_selfheal_tie_breaker_inodelk (call_frame_t *frame, xlator_t *this,
- inode_t *inode, char *dom, off_t off,
- size_t size, unsigned char *locked_on)
+afr_selfheal_tie_breaker_inodelk(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, char *dom, off_t off,
+ size_t size, unsigned char *locked_on)
{
- loc_t loc = {0,};
- struct gf_flock flock = {0, };
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int lock_count = 0;
- int eagain_count = 0;
+ loc_t loc = {
+ 0,
+ };
+ struct gf_flock flock = {
+ 0,
+ };
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int lock_count = 0;
+ int eagain_count = 0;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- flock.l_type = F_WRLCK;
- flock.l_start = off;
- flock.l_len = size;
+ flock.l_type = F_WRLCK;
+ flock.l_start = off;
+ flock.l_len = size;
- AFR_ONALL (frame, afr_selfheal_lock_cbk, inodelk, dom,
- &loc, F_SETLK, &flock, NULL);
+ AFR_ONALL(frame, afr_selfheal_lock_cbk, inodelk, dom, &loc, F_SETLK, &flock,
+ NULL);
- afr_get_lock_and_eagain_counts (priv, local->replies, &lock_count,
- &eagain_count);
+ afr_get_lock_and_eagain_counts(priv, local->replies, &lock_count,
+ &eagain_count);
- if (lock_count > priv->child_count/2 && eagain_count) {
- afr_locked_fill (frame, this, locked_on);
- afr_selfheal_uninodelk (frame, this, inode, dom, off,
- size, locked_on);
+ if (lock_count > priv->child_count / 2 && eagain_count) {
+ afr_locked_fill(frame, this, locked_on);
+ afr_selfheal_uninodelk(frame, this, inode, dom, off, size, locked_on);
- AFR_SEQ (frame, afr_selfheal_lock_cbk, inodelk, dom,
- &loc, F_SETLKW, &flock, NULL);
- }
+ AFR_SEQ(frame, afr_selfheal_lock_cbk, inodelk, dom, &loc, F_SETLKW,
+ &flock, NULL);
+ }
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return afr_locked_fill (frame, this, locked_on);
+ return afr_locked_fill(frame, this, locked_on);
}
int
-afr_selfheal_uninodelk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, off_t off, size_t size,
- const unsigned char *locked_on)
+afr_selfheal_uninodelk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, off_t off, size_t size,
+ const unsigned char *locked_on)
{
- loc_t loc = {0,};
- struct gf_flock flock = {0, };
-
+ loc_t loc = {
+ 0,
+ };
+ struct gf_flock flock = {
+ 0,
+ };
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- flock.l_type = F_UNLCK;
- flock.l_start = off;
- flock.l_len = size;
+ flock.l_type = F_UNLCK;
+ flock.l_start = off;
+ flock.l_len = size;
- AFR_ONLIST (locked_on, frame, afr_selfheal_lock_cbk, inodelk,
- dom, &loc, F_SETLK, &flock, NULL);
+ AFR_ONLIST(locked_on, frame, afr_selfheal_lock_cbk, inodelk, dom, &loc,
+ F_SETLK, &flock, NULL);
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return 0;
+ return 0;
}
-
int
-afr_selfheal_tryentrylk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, const char *name, unsigned char *locked_on)
+afr_selfheal_tryentrylk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, const char *name, unsigned char *locked_on)
{
- loc_t loc = {0,};
+ loc_t loc = {
+ 0,
+ };
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- AFR_ONALL (frame, afr_selfheal_lock_cbk, entrylk, dom,
- &loc, name, ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL);
+ AFR_ONALL(frame, afr_selfheal_lock_cbk, entrylk, dom, &loc, name,
+ ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL);
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return afr_locked_fill (frame, this, locked_on);
+ return afr_locked_fill(frame, this, locked_on);
}
-
int
-afr_selfheal_entrylk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, const char *name, unsigned char *locked_on)
+afr_selfheal_entrylk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, const char *name, unsigned char *locked_on)
{
- loc_t loc = {0,};
- afr_local_t *local = NULL;
- int i = 0;
- afr_private_t *priv = NULL;
+ loc_t loc = {
+ 0,
+ };
+ afr_local_t *local = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- AFR_ONALL (frame, afr_selfheal_lock_cbk, entrylk, dom, &loc,
- name, ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL);
+ AFR_ONALL(frame, afr_selfheal_lock_cbk, entrylk, dom, &loc, name,
+ ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL);
- for (i = 0; i < priv->child_count; i++) {
- if (local->replies[i].op_ret == -1 &&
- local->replies[i].op_errno == EAGAIN) {
- afr_locked_fill (frame, this, locked_on);
- afr_selfheal_unentrylk (frame, this, inode, dom, name,
- locked_on, NULL);
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->replies[i].op_ret == -1 &&
+ local->replies[i].op_errno == EAGAIN) {
+ afr_locked_fill(frame, this, locked_on);
+ afr_selfheal_unentrylk(frame, this, inode, dom, name, locked_on,
+ NULL);
- AFR_SEQ (frame, afr_selfheal_lock_cbk, entrylk, dom,
- &loc, name, ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
- break;
- }
- }
+ AFR_SEQ(frame, afr_selfheal_lock_cbk, entrylk, dom, &loc, name,
+ ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
+ break;
+ }
+ }
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return afr_locked_fill (frame, this, locked_on);
+ return afr_locked_fill(frame, this, locked_on);
}
int
-afr_selfheal_tie_breaker_entrylk (call_frame_t *frame, xlator_t *this,
- inode_t *inode, char *dom, const char *name,
- unsigned char *locked_on)
+afr_selfheal_tie_breaker_entrylk(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, char *dom, const char *name,
+ unsigned char *locked_on)
{
- loc_t loc = {0,};
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int lock_count = 0;
- int eagain_count = 0;
+ loc_t loc = {
+ 0,
+ };
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int lock_count = 0;
+ int eagain_count = 0;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- AFR_ONALL (frame, afr_selfheal_lock_cbk, entrylk, dom, &loc,
- name, ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL);
+ AFR_ONALL(frame, afr_selfheal_lock_cbk, entrylk, dom, &loc, name,
+ ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL);
- afr_get_lock_and_eagain_counts (priv, local->replies, &lock_count,
- &eagain_count);
+ afr_get_lock_and_eagain_counts(priv, local->replies, &lock_count,
+ &eagain_count);
- if (lock_count > priv->child_count/2 && eagain_count) {
- afr_locked_fill (frame, this, locked_on);
- afr_selfheal_unentrylk (frame, this, inode, dom, name,
- locked_on, NULL);
+ if (lock_count > priv->child_count / 2 && eagain_count) {
+ afr_locked_fill(frame, this, locked_on);
+ afr_selfheal_unentrylk(frame, this, inode, dom, name, locked_on, NULL);
- AFR_SEQ (frame, afr_selfheal_lock_cbk, entrylk, dom,
- &loc, name, ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
- }
+ AFR_SEQ(frame, afr_selfheal_lock_cbk, entrylk, dom, &loc, name,
+ ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
+ }
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return afr_locked_fill (frame, this, locked_on);
+ return afr_locked_fill(frame, this, locked_on);
}
-
int
-afr_selfheal_unentrylk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, const char *name, unsigned char *locked_on,
- dict_t *xdata)
-{
- loc_t loc = {0,};
-
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
-
- AFR_ONLIST (locked_on, frame, afr_selfheal_lock_cbk, entrylk,
- dom, &loc, name, ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
-
- loc_wipe (&loc);
-
- return 0;
-}
-
-
-gf_boolean_t
-afr_is_pending_set (xlator_t *this, dict_t *xdata, int type)
+afr_selfheal_unentrylk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, const char *name, unsigned char *locked_on,
+ dict_t *xdata)
{
- int idx = -1;
- afr_private_t *priv = NULL;
- void *pending_raw = NULL;
- int *pending_int = NULL;
- int i = 0;
-
- priv = this->private;
- idx = afr_index_for_transaction_type (type);
+ loc_t loc = {
+ 0,
+ };
- if (dict_get_ptr (xdata, AFR_DIRTY, &pending_raw) == 0) {
- if (pending_raw) {
- pending_int = pending_raw;
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- if (ntoh32 (pending_int[idx]))
- return _gf_true;
- }
- }
+ AFR_ONLIST(locked_on, frame, afr_selfheal_lock_cbk, entrylk, dom, &loc,
+ name, ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
- for (i = 0; i < priv->child_count; i++) {
- if (dict_get_ptr (xdata, priv->pending_key[i],
- &pending_raw))
- continue;
- if (!pending_raw)
- continue;
- pending_int = pending_raw;
+ loc_wipe(&loc);
- if (ntoh32 (pending_int[idx]))
- return _gf_true;
- }
-
- return _gf_false;
+ return 0;
}
-
gf_boolean_t
-afr_is_data_set (xlator_t *this, dict_t *xdata)
+afr_is_data_set(xlator_t *this, dict_t *xdata)
{
- return afr_is_pending_set (this, xdata, AFR_DATA_TRANSACTION);
+ return afr_is_pending_set(this, xdata, AFR_DATA_TRANSACTION);
}
gf_boolean_t
-afr_is_metadata_set (xlator_t *this, dict_t *xdata)
+afr_is_metadata_set(xlator_t *this, dict_t *xdata)
{
- return afr_is_pending_set (this, xdata, AFR_METADATA_TRANSACTION);
+ return afr_is_pending_set(this, xdata, AFR_METADATA_TRANSACTION);
}
gf_boolean_t
-afr_is_entry_set (xlator_t *this, dict_t *xdata)
+afr_is_entry_set(xlator_t *this, dict_t *xdata)
{
- return afr_is_pending_set (this, xdata, AFR_ENTRY_TRANSACTION);
+ return afr_is_pending_set(this, xdata, AFR_ENTRY_TRANSACTION);
}
/*
@@ -2217,317 +2278,310 @@ afr_is_entry_set (xlator_t *this, dict_t *xdata)
*/
int
-afr_selfheal_unlocked_inspect (call_frame_t *frame, xlator_t *this,
- uuid_t gfid, inode_t **link_inode,
- gf_boolean_t *data_selfheal,
- gf_boolean_t *metadata_selfheal,
- gf_boolean_t *entry_selfheal)
-{
- afr_private_t *priv = NULL;
- inode_t *inode = NULL;
- int i = 0;
- int valid_cnt = 0;
- struct iatt first = {0, };
- int first_idx = 0;
- struct afr_reply *replies = NULL;
- int ret = -1;
-
- priv = this->private;
-
- inode = afr_inode_find (this, gfid);
- if (!inode)
- goto out;
+afr_selfheal_unlocked_inspect(call_frame_t *frame, xlator_t *this, uuid_t gfid,
+ inode_t **link_inode, gf_boolean_t *data_selfheal,
+ gf_boolean_t *metadata_selfheal,
+ gf_boolean_t *entry_selfheal,
+ struct afr_reply *replies_dst)
+{
+ afr_private_t *priv = NULL;
+ inode_t *inode = NULL;
+ int i = 0;
+ int valid_cnt = 0;
+ struct iatt first = {
+ 0,
+ };
+ int first_idx = 0;
+ struct afr_reply *replies = NULL;
+ int ret = -1;
+
+ priv = this->private;
+
+ inode = afr_inode_find(this, gfid);
+ if (!inode)
+ goto out;
- replies = alloca0 (sizeof (*replies) * priv->child_count);
+ replies = alloca0(sizeof(*replies) * priv->child_count);
- ret = afr_selfheal_unlocked_discover (frame, inode, gfid, replies);
- if (ret)
- goto out;
+ ret = afr_selfheal_unlocked_discover(frame, inode, gfid, replies);
+ if (ret)
+ goto out;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
- if (replies[i].op_ret == -1)
- continue;
-
- /* The data segment of the changelog can be non-zero to indicate
- * the directory needs a full heal. So the check below ensures
- * it's not a directory before setting the data_selfheal boolean.
- */
- if (data_selfheal && !IA_ISDIR (replies[i].poststat.ia_type) &&
- afr_is_data_set (this, replies[i].xdata))
- *data_selfheal = _gf_true;
-
- if (metadata_selfheal &&
- afr_is_metadata_set (this, replies[i].xdata))
- *metadata_selfheal = _gf_true;
-
- if (entry_selfheal && afr_is_entry_set (this, replies[i].xdata))
- *entry_selfheal = _gf_true;
-
- valid_cnt++;
- if (valid_cnt == 1) {
- first = replies[i].poststat;
- first_idx = i;
- continue;
- }
-
- if (!IA_EQUAL (first, replies[i].poststat, type)) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN,
- "TYPE mismatch %d vs %d on %s for gfid:%s",
- (int) first.ia_type,
- (int) replies[i].poststat.ia_type,
- priv->children[i]->name,
- uuid_utoa (replies[i].poststat.ia_gfid));
- gf_event (EVENT_AFR_SPLIT_BRAIN, "subvol=%s;"
- "type=file;gfid=%s;"
- "ia_type-%d=%s;ia_type-%d=%s",
- this->name,
- uuid_utoa (replies[i].poststat.ia_gfid),
- first_idx,
- gf_inode_type_to_str (first.ia_type), i,
- gf_inode_type_to_str (replies[i].poststat.ia_type));
- ret = -EIO;
- goto out;
- }
-
- if (!IA_EQUAL (first, replies[i].poststat, uid)) {
- gf_msg_debug (this->name, 0,
- "UID mismatch "
- "%d vs %d on %s for gfid:%s",
- (int) first.ia_uid,
- (int) replies[i].poststat.ia_uid,
- priv->children[i]->name,
- uuid_utoa (replies[i].poststat.ia_gfid));
-
- if (metadata_selfheal)
- *metadata_selfheal = _gf_true;
- }
-
- if (!IA_EQUAL (first, replies[i].poststat, gid)) {
- gf_msg_debug (this->name, 0,
- "GID mismatch "
- "%d vs %d on %s for gfid:%s",
- (int) first.ia_uid,
- (int) replies[i].poststat.ia_uid,
- priv->children[i]->name,
- uuid_utoa (replies[i].poststat.ia_gfid));
-
- if (metadata_selfheal)
- *metadata_selfheal = _gf_true;
- }
-
- if (!IA_EQUAL (first, replies[i].poststat, prot)) {
- gf_msg_debug (this->name, 0,
- "MODE mismatch "
- "%d vs %d on %s for gfid:%s",
- (int) st_mode_from_ia (first.ia_prot, 0),
- (int) st_mode_from_ia
- (replies[i].poststat.ia_prot, 0),
- priv->children[i]->name,
- uuid_utoa (replies[i].poststat.ia_gfid));
-
- if (metadata_selfheal)
- *metadata_selfheal = _gf_true;
- }
-
- if (IA_ISREG(first.ia_type) &&
- !IA_EQUAL (first, replies[i].poststat, size)) {
- gf_msg_debug (this->name, 0,
- "SIZE mismatch "
- "%lld vs %lld on %s for gfid:%s",
- (long long) first.ia_size,
- (long long) replies[i].poststat.ia_size,
- priv->children[i]->name,
- uuid_utoa (replies[i].poststat.ia_gfid));
-
- if (data_selfheal)
- *data_selfheal = _gf_true;
- }
- }
-
- if (valid_cnt > 0 && link_inode) {
- *link_inode = inode_link (inode, NULL, NULL, &first);
- if (!*link_inode) {
- ret = -EINVAL;
- goto out;
- }
- } else if (valid_cnt < 2) {
- ret = afr_check_stale_error (replies, priv);
- goto out;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
+ if (replies[i].op_ret == -1)
+ continue;
- ret = 0;
+ /* The data segment of the changelog can be non-zero to indicate
+ * the directory needs a full heal. So the check below ensures
+ * it's not a directory before setting the data_selfheal boolean.
+ */
+ if (data_selfheal && !IA_ISDIR(replies[i].poststat.ia_type) &&
+ afr_is_data_set(this, replies[i].xdata))
+ *data_selfheal = _gf_true;
+
+ if (metadata_selfheal && afr_is_metadata_set(this, replies[i].xdata))
+ *metadata_selfheal = _gf_true;
+
+ if (entry_selfheal && afr_is_entry_set(this, replies[i].xdata))
+ *entry_selfheal = _gf_true;
+
+ valid_cnt++;
+ if (valid_cnt == 1) {
+ first = replies[i].poststat;
+ first_idx = i;
+ continue;
+ }
+
+ if (!IA_EQUAL(first, replies[i].poststat, type)) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "TYPE mismatch %d vs %d on %s for gfid:%s",
+ (int)first.ia_type, (int)replies[i].poststat.ia_type,
+ priv->children[i]->name,
+ uuid_utoa(replies[i].poststat.ia_gfid));
+ gf_event(EVENT_AFR_SPLIT_BRAIN,
+ "client-pid=%d;"
+ "subvol=%s;"
+ "type=file;gfid=%s;"
+ "ia_type-%d=%s;ia_type-%d=%s",
+ this->ctx->cmd_args.client_pid, this->name,
+ uuid_utoa(replies[i].poststat.ia_gfid), first_idx,
+ gf_inode_type_to_str(first.ia_type), i,
+ gf_inode_type_to_str(replies[i].poststat.ia_type));
+ ret = -EIO;
+ goto out;
+ }
+
+ if (!IA_EQUAL(first, replies[i].poststat, uid)) {
+ gf_msg_debug(this->name, 0,
+ "UID mismatch "
+ "%d vs %d on %s for gfid:%s",
+ (int)first.ia_uid, (int)replies[i].poststat.ia_uid,
+ priv->children[i]->name,
+ uuid_utoa(replies[i].poststat.ia_gfid));
+
+ if (metadata_selfheal)
+ *metadata_selfheal = _gf_true;
+ }
+
+ if (!IA_EQUAL(first, replies[i].poststat, gid)) {
+ gf_msg_debug(this->name, 0,
+ "GID mismatch "
+ "%d vs %d on %s for gfid:%s",
+ (int)first.ia_uid, (int)replies[i].poststat.ia_uid,
+ priv->children[i]->name,
+ uuid_utoa(replies[i].poststat.ia_gfid));
+
+ if (metadata_selfheal)
+ *metadata_selfheal = _gf_true;
+ }
+
+ if (!IA_EQUAL(first, replies[i].poststat, prot)) {
+ gf_msg_debug(this->name, 0,
+ "MODE mismatch "
+ "%d vs %d on %s for gfid:%s",
+ (int)st_mode_from_ia(first.ia_prot, 0),
+ (int)st_mode_from_ia(replies[i].poststat.ia_prot, 0),
+ priv->children[i]->name,
+ uuid_utoa(replies[i].poststat.ia_gfid));
+
+ if (metadata_selfheal)
+ *metadata_selfheal = _gf_true;
+ }
+
+ if (IA_ISREG(first.ia_type) &&
+ !IA_EQUAL(first, replies[i].poststat, size)) {
+ gf_msg_debug(this->name, 0,
+ "SIZE mismatch "
+ "%lld vs %lld on %s for gfid:%s",
+ (long long)first.ia_size,
+ (long long)replies[i].poststat.ia_size,
+ priv->children[i]->name,
+ uuid_utoa(replies[i].poststat.ia_gfid));
+
+ if (data_selfheal)
+ *data_selfheal = _gf_true;
+ }
+ }
+
+ if (valid_cnt > 0 && link_inode) {
+ *link_inode = inode_link(inode, NULL, NULL, &first);
+ if (!*link_inode) {
+ ret = -EINVAL;
+ goto out;
+ }
+ } else if (valid_cnt < 2) {
+ ret = afr_check_stale_error(replies, priv);
+ goto out;
+ }
+
+ ret = 0;
out:
- if (inode)
- inode_unref (inode);
- if (replies)
- afr_replies_wipe (replies, priv->child_count);
+ if (replies && replies_dst)
+ afr_replies_copy(replies_dst, replies, priv->child_count);
+ if (inode)
+ inode_unref(inode);
+ if (replies)
+ afr_replies_wipe(replies, priv->child_count);
- return ret;
+ return ret;
}
-
inode_t *
-afr_inode_find (xlator_t *this, uuid_t gfid)
+afr_inode_find(xlator_t *this, uuid_t gfid)
{
- inode_table_t *table = NULL;
- inode_t *inode = NULL;
+ inode_table_t *table = NULL;
+ inode_t *inode = NULL;
- table = this->itable;
- if (!table)
- return NULL;
+ table = this->itable;
+ if (!table)
+ return NULL;
- inode = inode_find (table, gfid);
- if (inode)
- return inode;
+ inode = inode_find(table, gfid);
+ if (inode)
+ return inode;
- inode = inode_new (table);
- if (!inode)
- return NULL;
+ inode = inode_new(table);
+ if (!inode)
+ return NULL;
- gf_uuid_copy (inode->gfid, gfid);
+ gf_uuid_copy(inode->gfid, gfid);
- return inode;
+ return inode;
}
-
call_frame_t *
-afr_frame_create (xlator_t *this, int32_t *op_errno)
+afr_frame_create(xlator_t *this, int32_t *op_errno)
{
- call_frame_t *frame = NULL;
- afr_local_t *local = NULL;
- pid_t pid = GF_CLIENT_PID_SELF_HEALD;
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
+ pid_t pid = GF_CLIENT_PID_SELF_HEALD;
- frame = create_frame (this, this->ctx->pool);
- if (!frame)
- return NULL;
+ frame = create_frame(this, this->ctx->pool);
+ if (!frame) {
+ if (op_errno)
+ *op_errno = ENOMEM;
+ return NULL;
+ }
- local = AFR_FRAME_INIT (frame, (*op_errno));
- if (!local) {
- STACK_DESTROY (frame->root);
- return NULL;
- }
+ local = AFR_FRAME_INIT(frame, (*op_errno));
+ if (!local) {
+ STACK_DESTROY(frame->root);
+ return NULL;
+ }
- syncopctx_setfspid (&pid);
+ syncopctx_setfspid(&pid);
- frame->root->pid = pid;
+ frame->root->pid = pid;
- afr_set_lk_owner (frame, this, frame->root);
+ afr_set_lk_owner(frame, this, frame->root);
- return frame;
+ return frame;
}
int
-afr_selfheal_newentry_mark (call_frame_t *frame, xlator_t *this, inode_t *inode,
- int source, struct afr_reply *replies,
- unsigned char *sources, unsigned char *newentry)
+afr_selfheal_newentry_mark(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ int source, struct afr_reply *replies,
+ unsigned char *sources, unsigned char *newentry)
{
- int ret = 0;
- int i = 0;
- afr_private_t *priv = NULL;
- dict_t *xattr = NULL;
- int **changelog = NULL;
+ int ret = 0;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ dict_t *xattr = NULL;
+ int **changelog = NULL;
- priv = this->private;
+ priv = this->private;
- gf_uuid_copy (inode->gfid, replies[source].poststat.ia_gfid);
+ gf_uuid_copy(inode->gfid, replies[source].poststat.ia_gfid);
- xattr = dict_new();
- if (!xattr)
- return -ENOMEM;
+ xattr = dict_new();
+ if (!xattr)
+ return -ENOMEM;
- changelog = afr_mark_pending_changelog (priv, newentry, xattr,
- replies[source].poststat.ia_type);
+ changelog = afr_mark_pending_changelog(priv, newentry, xattr,
+ replies[source].poststat.ia_type);
- if (!changelog) {
- ret = -ENOMEM;
- goto out;
- }
+ if (!changelog) {
+ ret = -ENOMEM;
+ goto out;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
- ret |= afr_selfheal_post_op (frame, this, inode, i, xattr,
- NULL);
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ ret |= afr_selfheal_post_op(frame, this, inode, i, xattr, NULL);
+ }
out:
- if (changelog)
- afr_matrix_cleanup (changelog, priv->child_count);
- if (xattr)
- dict_unref (xattr);
- return ret;
+ if (changelog)
+ afr_matrix_cleanup(changelog, priv->child_count);
+ if (xattr)
+ dict_unref(xattr);
+ return ret;
}
int
-afr_selfheal_do (call_frame_t *frame, xlator_t *this, uuid_t gfid)
-{
- int ret = -1;
- int entry_ret = 1;
- int metadata_ret = 1;
- int data_ret = 1;
- int or_ret = 0;
- inode_t *inode = NULL;
- fd_t *fd = NULL;
- gf_boolean_t data_selfheal = _gf_false;
- gf_boolean_t metadata_selfheal = _gf_false;
- gf_boolean_t entry_selfheal = _gf_false;
- afr_private_t *priv = NULL;
- gf_boolean_t dataheal_enabled = _gf_false;
-
- priv = this->private;
-
- ret = gf_string2boolean (priv->data_self_heal, &dataheal_enabled);
- if (ret)
- goto out;
-
- ret = afr_selfheal_unlocked_inspect (frame, this, gfid, &inode,
- &data_selfheal,
- &metadata_selfheal,
- &entry_selfheal);
- if (ret)
- goto out;
+afr_selfheal_do(call_frame_t *frame, xlator_t *this, uuid_t gfid)
+{
+ int ret = -1;
+ int entry_ret = 1;
+ int metadata_ret = 1;
+ int data_ret = 1;
+ int or_ret = 0;
+ inode_t *inode = NULL;
+ fd_t *fd = NULL;
+ gf_boolean_t data_selfheal = _gf_false;
+ gf_boolean_t metadata_selfheal = _gf_false;
+ gf_boolean_t entry_selfheal = _gf_false;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+
+ ret = afr_selfheal_unlocked_inspect(frame, this, gfid, &inode,
+ &data_selfheal, &metadata_selfheal,
+ &entry_selfheal, NULL);
+ if (ret)
+ goto out;
- if (!(data_selfheal || metadata_selfheal || entry_selfheal)) {
- ret = 2;
- goto out;
- }
+ if (!(data_selfheal || metadata_selfheal || entry_selfheal)) {
+ ret = 2;
+ goto out;
+ }
- if (inode->ia_type == IA_IFREG) {
- ret = afr_selfheal_data_open (this, inode, &fd);
- if (!fd) {
- ret = -EIO;
- goto out;
- }
+ if (inode->ia_type == IA_IFREG) {
+ ret = afr_selfheal_data_open(this, inode, &fd);
+ if (!fd) {
+ ret = -EIO;
+ goto out;
}
+ }
- if (data_selfheal && dataheal_enabled)
- data_ret = afr_selfheal_data (frame, this, fd);
+ if (data_selfheal && priv->data_self_heal)
+ data_ret = afr_selfheal_data(frame, this, fd);
- if (metadata_selfheal && priv->metadata_self_heal)
- metadata_ret = afr_selfheal_metadata (frame, this, inode);
+ if (metadata_selfheal && priv->metadata_self_heal)
+ metadata_ret = afr_selfheal_metadata(frame, this, inode);
- if (entry_selfheal && priv->entry_self_heal)
- entry_ret = afr_selfheal_entry (frame, this, inode);
+ if (entry_selfheal && priv->entry_self_heal)
+ entry_ret = afr_selfheal_entry(frame, this, inode);
- or_ret = (data_ret | metadata_ret | entry_ret);
+ or_ret = (data_ret | metadata_ret | entry_ret);
- if (data_ret == -EIO || metadata_ret == -EIO || entry_ret == -EIO)
- ret = -EIO;
- else if (data_ret == 1 && metadata_ret == 1 && entry_ret == 1)
- ret = 1;
- else if (or_ret < 0)
- ret = or_ret;
- else
- ret = 0;
+ if (data_ret == -EIO || metadata_ret == -EIO || entry_ret == -EIO)
+ ret = -EIO;
+ else if (data_ret == 1 && metadata_ret == 1 && entry_ret == 1)
+ ret = 1;
+ else if (or_ret < 0)
+ ret = or_ret;
+ else
+ ret = 0;
out:
- if (inode)
- inode_unref (inode);
- if (fd)
- fd_unref (fd);
- return ret;
+ if (inode)
+ inode_unref(inode);
+ if (fd)
+ fd_unref(fd);
+ return ret;
}
/*
* This is the entry point for healing a given GFID. The return values for this
@@ -2539,160 +2593,342 @@ out:
*/
int
-afr_selfheal (xlator_t *this, uuid_t gfid)
+afr_selfheal(xlator_t *this, uuid_t gfid)
{
- int ret = -1;
- call_frame_t *frame = NULL;
- afr_local_t *local = NULL;
+ int ret = -1;
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
- frame = afr_frame_create (this, NULL);
- if (!frame)
- return ret;
+ frame = afr_frame_create(this, NULL);
+ if (!frame)
+ return ret;
- local = frame->local;
- local->xdata_req = dict_new();
+ local = frame->local;
+ local->xdata_req = dict_new();
- ret = afr_selfheal_do (frame, this, gfid);
+ ret = afr_selfheal_do(frame, this, gfid);
- if (frame)
- AFR_STACK_DESTROY (frame);
+ if (frame)
+ AFR_STACK_DESTROY(frame);
- return ret;
+ return ret;
}
-afr_local_t*
-__afr_dequeue_heals (afr_private_t *priv)
+afr_local_t *
+__afr_dequeue_heals(afr_private_t *priv)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- if (list_empty (&priv->heal_waiting))
- goto none;
- if ((priv->background_self_heal_count > 0) &&
- (priv->healers >= priv->background_self_heal_count))
- goto none;
+ if (list_empty(&priv->heal_waiting))
+ goto none;
+ if ((priv->background_self_heal_count > 0) &&
+ (priv->healers >= priv->background_self_heal_count))
+ goto none;
- local = list_entry (priv->heal_waiting.next, afr_local_t, healer);
- priv->heal_waiters--;
- GF_ASSERT (priv->heal_waiters >= 0);
- list_del_init(&local->healer);
- list_add(&local->healer, &priv->healing);
- priv->healers++;
- return local;
+ local = list_entry(priv->heal_waiting.next, afr_local_t, healer);
+ priv->heal_waiters--;
+ GF_ASSERT(priv->heal_waiters >= 0);
+ list_del_init(&local->healer);
+ list_add(&local->healer, &priv->healing);
+ priv->healers++;
+ return local;
none:
- gf_msg_debug (THIS->name, 0, "Nothing dequeued. "
- "Num healers: %d, Num Waiters: %d",
- priv->healers, priv->heal_waiters);
- return NULL;
+ gf_msg_debug(THIS->name, 0,
+ "Nothing dequeued. "
+ "Num healers: %d, Num Waiters: %d",
+ priv->healers, priv->heal_waiters);
+ return NULL;
}
int
-afr_refresh_selfheal_wrap (void *opaque)
+afr_refresh_selfheal_wrap(void *opaque)
{
- call_frame_t *heal_frame = opaque;
- afr_local_t *local = heal_frame->local;
- int ret = 0;
+ call_frame_t *heal_frame = opaque;
+ afr_local_t *local = heal_frame->local;
+ int ret = 0;
- ret = afr_selfheal (heal_frame->this, local->refreshinode->gfid);
- return ret;
+ ret = afr_selfheal(heal_frame->this, local->refreshinode->gfid);
+ return ret;
}
int
-afr_refresh_heal_done (int ret, call_frame_t *frame, void *opaque)
-{
- call_frame_t *heal_frame = opaque;
- xlator_t *this = heal_frame->this;
- afr_private_t *priv = this->private;
- afr_local_t *local = heal_frame->local;
-
- LOCK (&priv->lock);
- {
- list_del_init(&local->healer);
- priv->healers--;
- GF_ASSERT (priv->healers >= 0);
- local = __afr_dequeue_heals (priv);
- }
- UNLOCK (&priv->lock);
+afr_refresh_heal_done(int ret, call_frame_t *frame, void *opaque)
+{
+ call_frame_t *heal_frame = opaque;
+ xlator_t *this = heal_frame->this;
+ afr_private_t *priv = this->private;
+ afr_local_t *local = heal_frame->local;
- AFR_STACK_DESTROY (heal_frame);
+ LOCK(&priv->lock);
+ {
+ list_del_init(&local->healer);
+ priv->healers--;
+ GF_ASSERT(priv->healers >= 0);
+ local = __afr_dequeue_heals(priv);
+ }
+ UNLOCK(&priv->lock);
- if (local)
- afr_heal_synctask (this, local);
- return 0;
+ AFR_STACK_DESTROY(heal_frame);
+ if (local)
+ afr_heal_synctask(this, local);
+ return 0;
}
void
-afr_heal_synctask (xlator_t *this, afr_local_t *local)
+afr_heal_synctask(xlator_t *this, afr_local_t *local)
{
- int ret = 0;
- call_frame_t *heal_frame = NULL;
+ int ret = 0;
+ call_frame_t *heal_frame = NULL;
- heal_frame = local->heal_frame;
- ret = synctask_new (this->ctx->env, afr_refresh_selfheal_wrap,
- afr_refresh_heal_done, heal_frame, heal_frame);
- if (ret < 0)
- /* Heal not launched. Will be queued when the next inode
- * refresh happens and shd hasn't healed it yet. */
- afr_refresh_heal_done (ret, heal_frame, heal_frame);
+ heal_frame = local->heal_frame;
+ ret = synctask_new(this->ctx->env, afr_refresh_selfheal_wrap,
+ afr_refresh_heal_done, heal_frame, heal_frame);
+ if (ret < 0)
+ /* Heal not launched. Will be queued when the next inode
+ * refresh happens and shd hasn't healed it yet. */
+ afr_refresh_heal_done(ret, heal_frame, heal_frame);
}
gf_boolean_t
-afr_throttled_selfheal (call_frame_t *frame, xlator_t *this)
-{
- gf_boolean_t can_heal = _gf_true;
- afr_private_t *priv = this->private;
- afr_local_t *local = frame->local;
-
- LOCK (&priv->lock);
- {
- if ((priv->background_self_heal_count > 0) &&
- (priv->heal_wait_qlen + priv->background_self_heal_count) >
- (priv->heal_waiters + priv->healers)) {
- list_add_tail(&local->healer, &priv->heal_waiting);
- priv->heal_waiters++;
- local = __afr_dequeue_heals (priv);
- } else {
- can_heal = _gf_false;
- }
- }
- UNLOCK (&priv->lock);
-
- if (can_heal) {
- if (local)
- afr_heal_synctask (this, local);
- else
- gf_msg_debug (this->name, 0, "Max number of heals are "
- "pending, background self-heal rejected.");
+afr_throttled_selfheal(call_frame_t *frame, xlator_t *this)
+{
+ gf_boolean_t can_heal = _gf_true;
+ afr_private_t *priv = this->private;
+ afr_local_t *local = frame->local;
+
+ LOCK(&priv->lock);
+ {
+ if ((priv->background_self_heal_count > 0) &&
+ (priv->heal_wait_qlen + priv->background_self_heal_count) >
+ (priv->heal_waiters + priv->healers)) {
+ list_add_tail(&local->healer, &priv->heal_waiting);
+ priv->heal_waiters++;
+ local = __afr_dequeue_heals(priv);
+ } else {
+ can_heal = _gf_false;
}
+ }
+ UNLOCK(&priv->lock);
+
+ if (can_heal) {
+ if (local)
+ afr_heal_synctask(this, local);
+ else
+ gf_msg_debug(this->name, 0,
+ "Max number of heals are "
+ "pending, background self-heal rejected.");
+ }
- return can_heal;
+ return can_heal;
}
int
-afr_choose_source_by_policy (afr_private_t *priv, unsigned char *sources,
- afr_transaction_type type)
+afr_choose_source_by_policy(afr_private_t *priv, unsigned char *sources,
+ afr_transaction_type type)
{
- int source = -1;
- int i = 0;
+ int source = -1;
+ int i = 0;
- /* Give preference to local child to save on bandwidth */
- for (i = 0; i < priv->child_count; i++) {
- if (priv->local[i] && sources[i]) {
- if ((type == AFR_DATA_TRANSACTION) &&
- AFR_IS_ARBITER_BRICK (priv, i))
- continue;
+ /* Give preference to local child to save on bandwidth */
+ for (i = 0; i < priv->child_count; i++) {
+ if (priv->local[i] && sources[i]) {
+ if ((type == AFR_DATA_TRANSACTION) && AFR_IS_ARBITER_BRICK(priv, i))
+ continue;
- source = i;
- goto out;
- }
+ source = i;
+ goto out;
}
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (sources[i]) {
- source = i;
- goto out;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i]) {
+ source = i;
+ goto out;
}
+ }
out:
- return source;
+ return source;
+}
+
+static int
+afr_anon_inode_mkdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ afr_local_t *local = frame->local;
+ int i = (long)cookie;
+
+ local->replies[i].valid = 1;
+ local->replies[i].op_ret = op_ret;
+ local->replies[i].op_errno = op_errno;
+ if (op_ret == 0) {
+ local->op_ret = 0;
+ local->replies[i].poststat = *buf;
+ local->replies[i].preparent = *preparent;
+ local->replies[i].postparent = *postparent;
+ }
+ if (xdata) {
+ local->replies[i].xdata = dict_ref(xdata);
+ }
+
+ syncbarrier_wake(&local->barrier);
+ return 0;
+}
+
+int
+afr_anon_inode_create(xlator_t *this, int child, inode_t **linked_inode)
+{
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = this->private;
+ unsigned char *mkdir_on = alloca0(priv->child_count);
+ unsigned char *lookup_on = alloca0(priv->child_count);
+ loc_t loc = {0};
+ int32_t op_errno = 0;
+ int32_t child_op_errno = 0;
+ struct iatt iatt = {0};
+ dict_t *xdata = NULL;
+ uuid_t anon_inode_gfid = {0};
+ int mkdir_count = 0;
+ int i = 0;
+
+ /*Try to mkdir everywhere and return success if the dir exists on 'child'
+ */
+
+ if (!priv->use_anon_inode) {
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ frame = afr_frame_create(this, &op_errno);
+ if (op_errno) {
+ goto out;
+ }
+ local = frame->local;
+ if (!local->child_up[child]) {
+ /*Other bricks may need mkdir so don't error out yet*/
+ child_op_errno = ENOTCONN;
+ }
+ gf_uuid_parse(priv->anon_gfid_str, anon_inode_gfid);
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->child_up[i])
+ continue;
+
+ if (priv->anon_inode[i]) {
+ mkdir_on[i] = 0;
+ } else {
+ mkdir_on[i] = 1;
+ mkdir_count++;
+ }
+ }
+
+ if (mkdir_count == 0) {
+ *linked_inode = inode_find(this->itable, anon_inode_gfid);
+ if (*linked_inode) {
+ op_errno = 0;
+ goto out;
+ }
+ }
+
+ loc.parent = inode_ref(this->itable->root);
+ loc.name = priv->anon_inode_name;
+ loc.inode = inode_new(this->itable);
+ if (!loc.inode) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ xdata = dict_new();
+ if (!xdata) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ op_errno = -dict_set_gfuuid(xdata, "gfid-req", anon_inode_gfid, _gf_true);
+ if (op_errno) {
+ goto out;
+ }
+
+ if (mkdir_count == 0) {
+ memcpy(lookup_on, local->child_up, priv->child_count);
+ goto lookup;
+ }
+
+ AFR_ONLIST(mkdir_on, frame, afr_anon_inode_mkdir_cbk, mkdir, &loc, 0755, 0,
+ xdata);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!mkdir_on[i]) {
+ continue;
+ }
+
+ if (local->replies[i].op_ret == 0) {
+ priv->anon_inode[i] = 1;
+ iatt = local->replies[i].poststat;
+ } else if (local->replies[i].op_ret < 0 &&
+ local->replies[i].op_errno == EEXIST) {
+ lookup_on[i] = 1;
+ } else if (i == child) {
+ child_op_errno = local->replies[i].op_errno;
+ }
+ }
+
+ if (AFR_COUNT(lookup_on, priv->child_count) == 0) {
+ goto link;
+ }
+
+lookup:
+ AFR_ONLIST(lookup_on, frame, afr_selfheal_discover_cbk, lookup, &loc,
+ xdata);
+ for (i = 0; i < priv->child_count; i++) {
+ if (!lookup_on[i]) {
+ continue;
+ }
+
+ if (local->replies[i].op_ret == 0) {
+ if (gf_uuid_compare(anon_inode_gfid,
+ local->replies[i].poststat.ia_gfid) == 0) {
+ priv->anon_inode[i] = 1;
+ iatt = local->replies[i].poststat;
+ } else {
+ if (i == child)
+ child_op_errno = EINVAL;
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_DATA,
+ "%s has gfid: %s", priv->anon_inode_name,
+ uuid_utoa(local->replies[i].poststat.ia_gfid));
+ }
+ } else if (i == child) {
+ child_op_errno = local->replies[i].op_errno;
+ }
+ }
+link:
+ if (!gf_uuid_is_null(iatt.ia_gfid)) {
+ *linked_inode = inode_link(loc.inode, loc.parent, loc.name, &iatt);
+ if (*linked_inode) {
+ op_errno = 0;
+ inode_lookup(*linked_inode);
+ } else {
+ op_errno = ENOMEM;
+ }
+ goto out;
+ }
+
+out:
+ if (xdata)
+ dict_unref(xdata);
+ loc_wipe(&loc);
+ /*child_op_errno takes precedence*/
+ if (child_op_errno == 0) {
+ child_op_errno = op_errno;
+ }
+
+ if (child_op_errno && *linked_inode) {
+ inode_unref(*linked_inode);
+ *linked_inode = NULL;
+ }
+ if (frame)
+ AFR_STACK_DESTROY(frame);
+ return -child_op_errno;
}
diff --git a/xlators/cluster/afr/src/afr-self-heal-data.c b/xlators/cluster/afr/src/afr-self-heal-data.c
index 72e0e85a0a5..37bcc2b3f9e 100644
--- a/xlators/cluster/afr/src/afr-self-heal-data.c
+++ b/xlators/cluster/afr/src/afr-self-heal-data.c
@@ -8,605 +8,592 @@
cases as published by the Free Software Foundation.
*/
-
#include "afr.h"
#include "afr-self-heal.h"
-#include "byte-order.h"
+#include <glusterfs/byte-order.h>
#include "protocol-common.h"
#include "afr-messages.h"
-#include "events.h"
-
-enum {
- AFR_SELFHEAL_DATA_FULL = 0,
- AFR_SELFHEAL_DATA_DIFF,
-};
-
+#include <glusterfs/events.h>
#define HAS_HOLES(i) ((i->ia_blocks * 512) < (i->ia_size))
static int
-__checksum_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, uint32_t weak, uint8_t *strong,
- dict_t *xdata)
+__checksum_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, uint32_t weak, uint8_t *strong, dict_t *xdata)
{
- afr_local_t *local = NULL;
- struct afr_reply *replies = NULL;
- int i = (long) cookie;
-
- local = frame->local;
- replies = local->replies;
-
- replies[i].valid = 1;
- replies[i].op_ret = op_ret;
- replies[i].op_errno = op_errno;
- if (xdata) {
- replies[i].buf_has_zeroes = dict_get_str_boolean (xdata,
- "buf-has-zeroes", _gf_false);
- replies[i].fips_mode_rchecksum = dict_get_str_boolean (xdata,
- "fips-mode-rchecksum", _gf_false);
- }
- if (strong) {
- if (replies[i].fips_mode_rchecksum) {
- memcpy (local->replies[i].checksum, strong,
- SHA256_DIGEST_LENGTH);
- } else {
- memcpy (local->replies[i].checksum, strong,
- MD5_DIGEST_LENGTH);
- }
+ afr_local_t *local = NULL;
+ struct afr_reply *replies = NULL;
+ int i = (long)cookie;
+
+ local = frame->local;
+ replies = local->replies;
+
+ replies[i].valid = 1;
+ replies[i].op_ret = op_ret;
+ replies[i].op_errno = op_errno;
+ if (xdata) {
+ replies[i].buf_has_zeroes = dict_get_str_boolean(
+ xdata, "buf-has-zeroes", _gf_false);
+ replies[i].fips_mode_rchecksum = dict_get_str_boolean(
+ xdata, "fips-mode-rchecksum", _gf_false);
+ }
+ if (strong) {
+ if (replies[i].fips_mode_rchecksum) {
+ memcpy(local->replies[i].checksum, strong, SHA256_DIGEST_LENGTH);
+ } else {
+ memcpy(local->replies[i].checksum, strong, MD5_DIGEST_LENGTH);
}
+ }
- syncbarrier_wake (&local->barrier);
- return 0;
+ syncbarrier_wake(&local->barrier);
+ return 0;
}
static gf_boolean_t
-__afr_can_skip_data_block_heal (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int source, unsigned char *healed_sinks,
- off_t offset, size_t size,
- struct iatt *poststat)
+__afr_can_skip_data_block_heal(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ int source, unsigned char *healed_sinks,
+ off_t offset, size_t size, struct iatt *poststat)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- unsigned char *wind_subvols = NULL;
- gf_boolean_t checksum_match = _gf_true;
- struct afr_reply *replies = NULL;
- dict_t *xdata = NULL;
- int i = 0;
-
- priv = this->private;
- local = frame->local;
- replies = local->replies;
-
- xdata = dict_new();
- if (!xdata)
- goto out;
- if (dict_set_int32 (xdata, "check-zero-filled", 1)) {
- dict_unref (xdata);
- goto out;
- }
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ unsigned char *wind_subvols = NULL;
+ gf_boolean_t checksum_match = _gf_true;
+ struct afr_reply *replies = NULL;
+ dict_t *xdata = NULL;
+ int i = 0;
+
+ priv = this->private;
+ local = frame->local;
+ replies = local->replies;
+
+ xdata = dict_new();
+ if (!xdata)
+ goto out;
+ if (dict_set_int32_sizen(xdata, "check-zero-filled", 1)) {
+ dict_unref(xdata);
+ goto out;
+ }
+
+ wind_subvols = alloca0(priv->child_count);
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == source || healed_sinks[i])
+ wind_subvols[i] = 1;
+ }
+
+ AFR_ONLIST(wind_subvols, frame, __checksum_cbk, rchecksum, fd, offset, size,
+ xdata);
+ if (xdata)
+ dict_unref(xdata);
+
+ if (!replies[source].valid || replies[source].op_ret != 0)
+ return _gf_false;
- wind_subvols = alloca0 (priv->child_count);
- for (i = 0; i < priv->child_count; i++) {
- if (i == source || healed_sinks[i])
- wind_subvols[i] = 1;
- }
-
- AFR_ONLIST (wind_subvols, frame, __checksum_cbk, rchecksum, fd,
- offset, size, xdata);
- if (xdata)
- dict_unref (xdata);
-
- if (!replies[source].valid || replies[source].op_ret != 0)
- return _gf_false;
-
- for (i = 0; i < priv->child_count; i++) {
- if (i == source)
- continue;
- if (replies[i].valid) {
- if (memcmp (replies[source].checksum,
- replies[i].checksum,
- replies[source].fips_mode_rchecksum ?
- SHA256_DIGEST_LENGTH : MD5_DIGEST_LENGTH)) {
- checksum_match = _gf_false;
- break;
- }
- }
- }
-
- if (checksum_match) {
- if (HAS_HOLES (poststat))
- return _gf_true;
-
- /* For non-sparse files, we might be better off writing the
- * zeroes to sinks to avoid mismatch of disk-usage in bricks. */
- if (local->replies[source].buf_has_zeroes)
- return _gf_false;
- else
- return _gf_true;
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == source)
+ continue;
+ if (replies[i].valid) {
+ if (memcmp(replies[source].checksum, replies[i].checksum,
+ replies[source].fips_mode_rchecksum
+ ? SHA256_DIGEST_LENGTH
+ : MD5_DIGEST_LENGTH)) {
+ checksum_match = _gf_false;
+ break;
+ }
}
+ }
+
+ if (checksum_match) {
+ if (HAS_HOLES(poststat))
+ return _gf_true;
+
+ /* For non-sparse files, we might be better off writing the
+ * zeroes to sinks to avoid mismatch of disk-usage in bricks. */
+ if (local->replies[source].buf_has_zeroes)
+ return _gf_false;
+ else
+ return _gf_true;
+ }
out:
- return _gf_false;
+ return _gf_false;
}
-
static gf_boolean_t
-__afr_is_sink_zero_filled (xlator_t *this, fd_t *fd, size_t size,
- off_t offset, int sink)
+__afr_is_sink_zero_filled(xlator_t *this, fd_t *fd, size_t size, off_t offset,
+ int sink)
{
- afr_private_t *priv = NULL;
- struct iobref *iobref = NULL;
- struct iovec *iovec = NULL;
- int count = 0;
- int ret = 0;
- gf_boolean_t zero_filled = _gf_false;
-
- priv = this->private;
- ret = syncop_readv (priv->children[sink], fd, size, offset, 0, &iovec,
- &count, &iobref, NULL, NULL);
- if (ret < 0)
- goto out;
- ret = iov_0filled (iovec, count);
- if (!ret)
- zero_filled = _gf_true;
+ afr_private_t *priv = NULL;
+ struct iobref *iobref = NULL;
+ struct iovec *iovec = NULL;
+ int count = 0;
+ int ret = 0;
+ gf_boolean_t zero_filled = _gf_false;
+
+ priv = this->private;
+ ret = syncop_readv(priv->children[sink], fd, size, offset, 0, &iovec,
+ &count, &iobref, NULL, NULL, NULL);
+ if (ret < 0)
+ goto out;
+ ret = iov_0filled(iovec, count);
+ if (!ret)
+ zero_filled = _gf_true;
out:
- if (iovec)
- GF_FREE (iovec);
- if (iobref)
- iobref_unref (iobref);
- return zero_filled;
+ if (iovec)
+ GF_FREE(iovec);
+ if (iobref)
+ iobref_unref(iobref);
+ return zero_filled;
}
static int
-__afr_selfheal_data_read_write (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int source, unsigned char *healed_sinks,
- off_t offset, size_t size,
- struct afr_reply *replies, int type)
+__afr_selfheal_data_read_write(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ int source, unsigned char *healed_sinks,
+ off_t offset, size_t size,
+ struct afr_reply *replies, int type)
{
- struct iovec *iovec = NULL;
- int count = 0;
- struct iobref *iobref = NULL;
- int ret = 0;
- int i = 0;
- afr_private_t *priv = NULL;
-
- priv = this->private;
-
- ret = syncop_readv (priv->children[source], fd, size, offset, 0,
- &iovec, &count, &iobref, NULL, NULL);
- if (ret <= 0)
- return ret;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!healed_sinks[i])
- continue;
-
- /*
- * TODO: Use fiemap() and discard() to heal holes
- * in the future.
- *
- * For now,
- *
- * - if the source had any holes at all,
- * AND
- * - if we are writing past the original file size
- * of the sink
- * AND
- * - is NOT the last block of the source file. if
- * the block contains EOF, it has to be written
- * in order to set the file size even if the
- * last block is 0-filled.
- * AND
- * - if the read buffer is filled with only 0's
- *
- * then, skip writing to this source. We don't depend
- * on the write to happen to update the size as we
- * have performed an ftruncate() upfront anyways.
- */
-#define is_last_block(o,b,s) ((s >= o) && (s <= (o + b)))
- if (HAS_HOLES ((&replies[source].poststat)) &&
- offset >= replies[i].poststat.ia_size &&
- !is_last_block (offset, size,
- replies[source].poststat.ia_size) &&
- (iov_0filled (iovec, count) == 0))
- continue;
-
- /* Avoid filling up sparse regions of the sink with 0-filled
- * writes.*/
- if (type == AFR_SELFHEAL_DATA_FULL &&
- HAS_HOLES ((&replies[source].poststat)) &&
- ((offset + size) <= replies[i].poststat.ia_size) &&
- (iov_0filled (iovec, count) == 0) &&
- __afr_is_sink_zero_filled (this, fd, size, offset, i)) {
- continue;
- }
-
- ret = syncop_writev (priv->children[i], fd, iovec, count,
- offset, iobref, 0, NULL, NULL);
- if (ret != iov_length (iovec, count)) {
- /* write() failed on this sink. unset the corresponding
- member in sinks[] (which is healed_sinks[] in the
- caller) so that this server does NOT get considered
- as successfully healed.
- */
- healed_sinks[i] = 0;
- }
- }
- if (iovec)
- GF_FREE (iovec);
- if (iobref)
- iobref_unref (iobref);
-
- return ret;
+ struct iovec *iovec = NULL;
+ int count = 0;
+ struct iobref *iobref = NULL;
+ int ret = 0;
+ int i = 0;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+
+ ret = syncop_readv(priv->children[source], fd, size, offset, 0, &iovec,
+ &count, &iobref, NULL, NULL, NULL);
+ if (ret <= 0)
+ return ret;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!healed_sinks[i])
+ continue;
+
+ /*
+ * TODO: Use fiemap() and discard() to heal holes
+ * in the future.
+ *
+ * For now,
+ *
+ * - if the source had any holes at all,
+ * AND
+ * - if we are writing past the original file size
+ * of the sink
+ * AND
+ * - is NOT the last block of the source file. if
+ * the block contains EOF, it has to be written
+ * in order to set the file size even if the
+ * last block is 0-filled.
+ * AND
+ * - if the read buffer is filled with only 0's
+ *
+ * then, skip writing to this source. We don't depend
+ * on the write to happen to update the size as we
+ * have performed an ftruncate() upfront anyways.
+ */
+#define is_last_block(o, b, s) ((s >= o) && (s <= (o + b)))
+ if (HAS_HOLES((&replies[source].poststat)) &&
+ offset >= replies[i].poststat.ia_size &&
+ !is_last_block(offset, size, replies[source].poststat.ia_size) &&
+ (iov_0filled(iovec, count) == 0))
+ continue;
+
+ /* Avoid filling up sparse regions of the sink with 0-filled
+ * writes.*/
+ if (type == AFR_SELFHEAL_DATA_FULL &&
+ HAS_HOLES((&replies[source].poststat)) &&
+ ((offset + size) <= replies[i].poststat.ia_size) &&
+ (iov_0filled(iovec, count) == 0) &&
+ __afr_is_sink_zero_filled(this, fd, size, offset, i)) {
+ continue;
+ }
+
+ ret = syncop_writev(priv->children[i], fd, iovec, count, offset, iobref,
+ 0, NULL, NULL, NULL, NULL);
+ if (ret != iov_length(iovec, count)) {
+ /* write() failed on this sink. unset the corresponding
+ member in sinks[] (which is healed_sinks[] in the
+ caller) so that this server does NOT get considered
+ as successfully healed.
+ */
+ healed_sinks[i] = 0;
+ }
+ }
+ if (iovec)
+ GF_FREE(iovec);
+ if (iobref)
+ iobref_unref(iobref);
+
+ return ret;
}
-static int
-afr_selfheal_data_block (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int source, unsigned char *healed_sinks, off_t offset,
- size_t size, int type, struct afr_reply *replies)
+static gf_boolean_t
+afr_source_sinks_locked(xlator_t *this, unsigned char *locked_on, int source,
+ unsigned char *healed_sinks)
{
- int ret = -1;
- int sink_count = 0;
- afr_private_t *priv = NULL;
- unsigned char *data_lock = NULL;
-
- priv = this->private;
- sink_count = AFR_COUNT (healed_sinks, priv->child_count);
- data_lock = alloca0 (priv->child_count);
-
- ret = afr_selfheal_inodelk (frame, this, fd->inode, this->name,
- offset, size, data_lock);
- {
- if (ret < sink_count) {
- ret = -ENOTCONN;
- goto unlock;
- }
-
- if (type == AFR_SELFHEAL_DATA_DIFF &&
- __afr_can_skip_data_block_heal (frame, this, fd, source,
- healed_sinks, offset, size,
- &replies[source].poststat)) {
- ret = 0;
- goto unlock;
- }
-
- ret = __afr_selfheal_data_read_write (frame, this, fd, source,
- healed_sinks, offset, size,
- replies, type);
- }
-unlock:
- afr_selfheal_uninodelk (frame, this, fd->inode, this->name,
- offset, size, data_lock);
- return ret;
-}
+ afr_private_t *priv = this->private;
+ int i = 0;
+ if (!locked_on[source])
+ return _gf_false;
+ for (i = 0; i < priv->child_count; i++) {
+ if (healed_sinks[i] && locked_on[i])
+ return _gf_true;
+ }
-static int
-afr_selfheal_data_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
- unsigned char *healed_sinks)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
-
- local = frame->local;
- priv = this->private;
-
- if (!priv->ensure_durability)
- return 0;
-
- AFR_ONLIST (healed_sinks, frame, afr_sh_generic_fop_cbk, fsync, fd, 0,
- NULL);
-
- for (i = 0; i < priv->child_count; i++)
- if (healed_sinks[i] && local->replies[i].op_ret != 0)
- /* fsync() failed. Do NOT consider this server
- as successfully healed. Mark it so.
- */
- healed_sinks[i] = 0;
- return 0;
+ return _gf_false;
}
static int
-afr_data_self_heal_type_get (afr_private_t *priv, unsigned char *healed_sinks,
- int source, struct afr_reply *replies)
+afr_selfheal_data_block(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ int source, unsigned char *healed_sinks, off_t offset,
+ size_t size, int type, struct afr_reply *replies)
{
- int type = AFR_SELFHEAL_DATA_FULL;
- int i = 0;
-
- if (priv->data_self_heal_algorithm == NULL) {
- type = AFR_SELFHEAL_DATA_FULL;
- for (i = 0; i < priv->child_count; i++) {
- if (!healed_sinks[i] && i != source)
- continue;
- if (replies[i].poststat.ia_size) {
- type = AFR_SELFHEAL_DATA_DIFF;
- break;
- }
- }
- } else if (strcmp (priv->data_self_heal_algorithm, "full") == 0) {
- type = AFR_SELFHEAL_DATA_FULL;
- } else if (strcmp (priv->data_self_heal_algorithm, "diff") == 0) {
- type = AFR_SELFHEAL_DATA_DIFF;
+ int ret = -1;
+ afr_private_t *priv = NULL;
+ unsigned char *data_lock = NULL;
+
+ priv = this->private;
+ data_lock = alloca0(priv->child_count);
+
+ ret = afr_selfheal_inodelk(frame, this, fd->inode, this->name, offset, size,
+ data_lock);
+ {
+ if (!afr_source_sinks_locked(this, data_lock, source, healed_sinks)) {
+ ret = -ENOTCONN;
+ goto unlock;
}
- return type;
-}
-static int
-afr_selfheal_data_do (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int source, unsigned char *healed_sinks,
- struct afr_reply *replies)
-{
- afr_private_t *priv = NULL;
- off_t off = 0;
- size_t block = 0;
- int type = AFR_SELFHEAL_DATA_FULL;
- int ret = -1;
- call_frame_t *iter_frame = NULL;
- unsigned char arbiter_sink_status = 0;
-
- priv = this->private;
- if (priv->arbiter_count) {
- arbiter_sink_status = healed_sinks[ARBITER_BRICK_INDEX];
- healed_sinks[ARBITER_BRICK_INDEX] = 0;
+ if (type == AFR_SELFHEAL_DATA_DIFF &&
+ __afr_can_skip_data_block_heal(frame, this, fd, source,
+ healed_sinks, offset, size,
+ &replies[source].poststat)) {
+ ret = 0;
+ goto unlock;
}
- block = 128 * 1024 * priv->data_self_heal_window_size;
-
- type = afr_data_self_heal_type_get (priv, healed_sinks, source,
- replies);
+ ret = __afr_selfheal_data_read_write(
+ frame, this, fd, source, healed_sinks, offset, size, replies, type);
+ }
+unlock:
+ afr_selfheal_uninodelk(frame, this, fd->inode, this->name, offset, size,
+ data_lock);
+ return ret;
+}
- iter_frame = afr_copy_frame (frame);
- if (!iter_frame) {
- ret = -ENOMEM;
- goto out;
- }
+static int
+afr_selfheal_data_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ unsigned char *healed_sinks)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
- for (off = 0; off < replies[source].poststat.ia_size; off += block) {
- if (AFR_COUNT (healed_sinks, priv->child_count) == 0) {
- ret = -ENOTCONN;
- goto out;
- }
+ local = frame->local;
+ priv = this->private;
- ret = afr_selfheal_data_block (iter_frame, this, fd, source,
- healed_sinks, off, block, type,
- replies);
- if (ret < 0)
- goto out;
+ if (!priv->ensure_durability)
+ return 0;
- AFR_STACK_RESET (iter_frame);
- if (iter_frame->local == NULL) {
- ret = -ENOTCONN;
- goto out;
- }
- }
+ AFR_ONLIST(healed_sinks, frame, afr_sh_generic_fop_cbk, fsync, fd, 0, NULL);
- ret = afr_selfheal_data_fsync (frame, this, fd, healed_sinks);
+ for (i = 0; i < priv->child_count; i++)
+ if (healed_sinks[i] && local->replies[i].op_ret != 0)
+ /* fsync() failed. Do NOT consider this server
+ as successfully healed. Mark it so.
+ */
+ healed_sinks[i] = 0;
+ return 0;
+}
-out:
- if (arbiter_sink_status)
- healed_sinks[ARBITER_BRICK_INDEX] = arbiter_sink_status;
+static int
+afr_data_self_heal_type_get(afr_private_t *priv, unsigned char *healed_sinks,
+ int source, struct afr_reply *replies)
+{
+ int type = AFR_SELFHEAL_DATA_FULL;
+ int i = 0;
- if (iter_frame)
- AFR_STACK_DESTROY (iter_frame);
- return ret;
+ if (priv->data_self_heal_algorithm == AFR_SELFHEAL_DATA_DYNAMIC) {
+ type = AFR_SELFHEAL_DATA_FULL;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!healed_sinks[i] && i != source)
+ continue;
+ if (replies[i].poststat.ia_size) {
+ type = AFR_SELFHEAL_DATA_DIFF;
+ break;
+ }
+ }
+ } else {
+ type = priv->data_self_heal_algorithm;
+ }
+ return type;
}
-
static int
-__afr_selfheal_truncate_sinks (call_frame_t *frame, xlator_t *this,
- fd_t *fd, unsigned char *healed_sinks,
- uint64_t size)
+afr_selfheal_data_do(call_frame_t *frame, xlator_t *this, fd_t *fd, int source,
+ unsigned char *healed_sinks, struct afr_reply *replies)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- unsigned char arbiter_sink_status = 0;
- int i = 0;
+ afr_private_t *priv = NULL;
+ off_t off = 0;
+ size_t block = 0;
+ int type = AFR_SELFHEAL_DATA_FULL;
+ int ret = -1;
+ call_frame_t *iter_frame = NULL;
+ unsigned char arbiter_sink_status = 0;
+
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_SELF_HEAL_INFO,
+ "performing data selfheal on %s", uuid_utoa(fd->inode->gfid));
+
+ priv = this->private;
+ if (priv->arbiter_count) {
+ arbiter_sink_status = healed_sinks[ARBITER_BRICK_INDEX];
+ healed_sinks[ARBITER_BRICK_INDEX] = 0;
+ }
+
+ block = 128 * 1024 * priv->data_self_heal_window_size;
+
+ type = afr_data_self_heal_type_get(priv, healed_sinks, source, replies);
+
+ iter_frame = afr_copy_frame(frame);
+ if (!iter_frame) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ for (off = 0; off < replies[source].poststat.ia_size; off += block) {
+ if (AFR_COUNT(healed_sinks, priv->child_count) == 0) {
+ ret = -ENOTCONN;
+ goto out;
+ }
- local = frame->local;
- priv = this->private;
+ ret = afr_selfheal_data_block(iter_frame, this, fd, source,
+ healed_sinks, off, block, type, replies);
+ if (ret < 0)
+ goto out;
- if (priv->arbiter_count) {
- arbiter_sink_status = healed_sinks[ARBITER_BRICK_INDEX];
- healed_sinks[ARBITER_BRICK_INDEX] = 0;
+ AFR_STACK_RESET(iter_frame);
+ if (iter_frame->local == NULL) {
+ ret = -ENOTCONN;
+ goto out;
}
+ }
- AFR_ONLIST (healed_sinks, frame, afr_sh_generic_fop_cbk, ftruncate, fd,
- size, NULL);
+ ret = afr_selfheal_data_fsync(frame, this, fd, healed_sinks);
- for (i = 0; i < priv->child_count; i++)
- if (healed_sinks[i] && local->replies[i].op_ret == -1)
- /* truncate() failed. Do NOT consider this server
- as successfully healed. Mark it so.
- */
- healed_sinks[i] = 0;
+out:
+ if (arbiter_sink_status)
+ healed_sinks[ARBITER_BRICK_INDEX] = arbiter_sink_status;
- if (arbiter_sink_status)
- healed_sinks[ARBITER_BRICK_INDEX] = arbiter_sink_status;
- return 0;
+ if (iter_frame)
+ AFR_STACK_DESTROY(iter_frame);
+ return ret;
+}
+
+static int
+__afr_selfheal_truncate_sinks(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ unsigned char *healed_sinks, uint64_t size)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ /* This will send truncate on the arbiter brick as well if it is marked as
+ * sink. If changelog is enabled on the volume it captures truncate as a
+ * data transactions on the arbiter brick. This will help geo-rep to
+ * properly sync the data from master to slave if arbiter is the ACTIVE
+ * brick during syncing and which had got some entries healed for data as
+ * part of self heal.
+ */
+ AFR_ONLIST(healed_sinks, frame, afr_sh_generic_fop_cbk, ftruncate, fd, size,
+ NULL);
+
+ for (i = 0; i < priv->child_count; i++)
+ if (healed_sinks[i] && local->replies[i].op_ret == -1)
+ /* truncate() failed. Do NOT consider this server
+ as successfully healed. Mark it so.
+ */
+ healed_sinks[i] = 0;
+
+ return 0;
}
gf_boolean_t
-afr_has_source_witnesses (xlator_t *this, unsigned char *sources,
- uint64_t *witness)
+afr_has_source_witnesses(xlator_t *this, unsigned char *sources,
+ uint64_t *witness)
{
- int i = 0;
- afr_private_t *priv = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (sources[i] && witness[i])
- return _gf_true;
- }
- return _gf_false;
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i] && witness[i])
+ return _gf_true;
+ }
+ return _gf_false;
}
static gf_boolean_t
-afr_does_size_mismatch (xlator_t *this, unsigned char *sources,
- struct afr_reply *replies)
+afr_does_size_mismatch(xlator_t *this, unsigned char *sources,
+ struct afr_reply *replies)
{
- int i = 0;
- afr_private_t *priv = NULL;
- struct iatt *min = NULL;
- struct iatt *max = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ struct iatt *min = NULL;
+ struct iatt *max = NULL;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
- if (replies[i].op_ret < 0)
- continue;
+ if (replies[i].op_ret < 0)
+ continue;
- if (!sources[i])
- continue;
+ if (!sources[i])
+ continue;
- if (AFR_IS_ARBITER_BRICK (priv, i) &&
- (replies[i].poststat.ia_size == 0))
- continue;
+ if (AFR_IS_ARBITER_BRICK(priv, i) && (replies[i].poststat.ia_size == 0))
+ continue;
- if (!min)
- min = &replies[i].poststat;
+ if (!min)
+ min = &replies[i].poststat;
- if (!max)
- max = &replies[i].poststat;
+ if (!max)
+ max = &replies[i].poststat;
- if (min->ia_size > replies[i].poststat.ia_size)
- min = &replies[i].poststat;
+ if (min->ia_size > replies[i].poststat.ia_size)
+ min = &replies[i].poststat;
- if (max->ia_size < replies[i].poststat.ia_size)
- max = &replies[i].poststat;
- }
+ if (max->ia_size < replies[i].poststat.ia_size)
+ max = &replies[i].poststat;
+ }
- if (min && max) {
- if (min->ia_size != max->ia_size)
- return _gf_true;
- }
+ if (min && max) {
+ if (min->ia_size != max->ia_size)
+ return _gf_true;
+ }
- return _gf_false;
+ return _gf_false;
}
static void
-afr_mark_biggest_witness_as_source (xlator_t *this, unsigned char *sources,
- uint64_t *witness)
+afr_mark_biggest_witness_as_source(xlator_t *this, unsigned char *sources,
+ uint64_t *witness)
{
- int i = 0;
- afr_private_t *priv = NULL;
- uint64_t biggest_witness = 0;
-
- priv = this->private;
- /* Find source with biggest witness count */
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
- if (biggest_witness < witness[i])
- biggest_witness = witness[i];
- }
-
- /* Mark files with less witness count as not source */
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
- if (witness[i] < biggest_witness)
- sources[i] = 0;
- }
-
- return;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ uint64_t biggest_witness = 0;
+
+ priv = this->private;
+ /* Find source with biggest witness count */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ if (biggest_witness < witness[i])
+ biggest_witness = witness[i];
+ }
+
+ /* Mark files with less witness count as not source */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ if (witness[i] < biggest_witness)
+ sources[i] = 0;
+ }
+
+ return;
}
/* This is a tie breaker function. Only one source be assigned here */
static void
-afr_mark_newest_file_as_source (xlator_t *this, unsigned char *sources,
- struct afr_reply *replies)
+afr_mark_newest_file_as_source(xlator_t *this, unsigned char *sources,
+ struct afr_reply *replies)
{
- int i = 0;
- afr_private_t *priv = NULL;
- int source = -1;
- uint32_t max_ctime = 0;
-
- priv = this->private;
- /* Find source with latest ctime */
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
-
- if (max_ctime <= replies[i].poststat.ia_ctime) {
- source = i;
- max_ctime = replies[i].poststat.ia_ctime;
- }
+ int i = 0;
+ afr_private_t *priv = NULL;
+ int source = -1;
+ uint32_t max_ctime = 0;
+
+ priv = this->private;
+ /* Find source with latest ctime */
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+
+ if (max_ctime <= replies[i].poststat.ia_ctime) {
+ source = i;
+ max_ctime = replies[i].poststat.ia_ctime;
}
+ }
- /* Only mark one of the files as source to break ties */
- memset (sources, 0, sizeof (*sources) * priv->child_count);
- sources[source] = 1;
+ /* Only mark one of the files as source to break ties */
+ memset(sources, 0, sizeof(*sources) * priv->child_count);
+ sources[source] = 1;
}
static int
-__afr_selfheal_data_finalize_source (call_frame_t *frame, xlator_t *this,
- inode_t *inode,
- unsigned char *sources,
- unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *locked_on,
- unsigned char *undid_pending,
- struct afr_reply *replies,
- uint64_t *witness)
+__afr_selfheal_data_finalize_source(
+ call_frame_t *frame, xlator_t *this, inode_t *inode, unsigned char *sources,
+ unsigned char *sinks, unsigned char *healed_sinks, unsigned char *locked_on,
+ unsigned char *undid_pending, struct afr_reply *replies, uint64_t *witness)
{
- afr_private_t *priv = NULL;
- int source = -1;
- int sources_count = 0;
- priv = this->private;
-
- sources_count = AFR_COUNT (sources, priv->child_count);
-
- if ((AFR_CMP (locked_on, healed_sinks, priv->child_count) == 0)
- || !sources_count) {
- /* split brain */
- source = afr_mark_split_brain_source_sinks (frame, this, inode,
- sources, sinks,
- healed_sinks,
- locked_on, replies,
- AFR_DATA_TRANSACTION);
- if (source < 0) {
- gf_event (EVENT_AFR_SPLIT_BRAIN, "subvol=%s;type=data;"
- "file=%s", this->name, uuid_utoa(inode->gfid));
- return -EIO;
- }
-
- _afr_fav_child_reset_sink_xattrs (frame, this, inode, source,
- healed_sinks, undid_pending,
- AFR_DATA_TRANSACTION,
- locked_on, replies);
- goto out;
- }
-
- /* No split brain at this point. If we were called from
- * afr_heal_splitbrain_file(), abort.*/
- if (afr_dict_contains_heal_op(frame))
- return -EIO;
-
- /* If there are no witnesses/size-mismatches on sources we are done*/
- if (!afr_does_size_mismatch (this, sources, replies) &&
- !afr_has_source_witnesses (this, sources, witness))
- goto out;
-
- afr_mark_largest_file_as_source (this, sources, replies);
- afr_mark_biggest_witness_as_source (this, sources, witness);
- afr_mark_newest_file_as_source (this, sources, replies);
- if (priv->arbiter_count)
- /* Choose non-arbiter brick as source for empty files. */
- afr_mark_source_sinks_if_file_empty (this, sources, sinks,
- healed_sinks, locked_on,
- replies,
- AFR_DATA_TRANSACTION);
+ afr_private_t *priv = NULL;
+ int source = -1;
+ int sources_count = 0;
+ priv = this->private;
+
+ sources_count = AFR_COUNT(sources, priv->child_count);
+
+ if ((AFR_CMP(locked_on, healed_sinks, priv->child_count) == 0) ||
+ !sources_count) {
+ /* split brain */
+ source = afr_mark_split_brain_source_sinks(
+ frame, this, inode, sources, sinks, healed_sinks, locked_on,
+ replies, AFR_DATA_TRANSACTION);
+ if (source < 0) {
+ gf_event(EVENT_AFR_SPLIT_BRAIN,
+ "client-pid=%d;"
+ "subvol=%s;type=data;"
+ "file=%s",
+ this->ctx->cmd_args.client_pid, this->name,
+ uuid_utoa(inode->gfid));
+ return -EIO;
+ }
+
+ _afr_fav_child_reset_sink_xattrs(
+ frame, this, inode, source, healed_sinks, undid_pending,
+ AFR_DATA_TRANSACTION, locked_on, replies);
+ goto out;
+ }
+
+ /* No split brain at this point. If we were called from
+ * afr_heal_splitbrain_file(), abort.*/
+ if (afr_dict_contains_heal_op(frame))
+ return -EIO;
+
+ /* If there are no witnesses/size-mismatches on sources we are done*/
+ if (!afr_does_size_mismatch(this, sources, replies) &&
+ !afr_has_source_witnesses(this, sources, witness))
+ goto out;
+
+ afr_mark_largest_file_as_source(this, sources, replies);
+ afr_mark_biggest_witness_as_source(this, sources, witness);
+ afr_mark_newest_file_as_source(this, sources, replies);
+ if (priv->arbiter_count)
+ /* Choose non-arbiter brick as source for empty files. */
+ afr_mark_source_sinks_if_file_empty(this, sources, sinks, healed_sinks,
+ locked_on, replies,
+ AFR_DATA_TRANSACTION);
out:
- afr_mark_active_sinks (this, sources, locked_on, healed_sinks);
- source = afr_choose_source_by_policy (priv, sources,
- AFR_DATA_TRANSACTION);
+ afr_mark_active_sinks(this, sources, locked_on, healed_sinks);
+ source = afr_choose_source_by_policy(priv, sources, AFR_DATA_TRANSACTION);
- return source;
+ return source;
}
/*
@@ -619,290 +606,286 @@ out:
* for self-healing, or -1 if no healing is necessary/split brain.
*/
int
-__afr_selfheal_data_prepare (call_frame_t *frame, xlator_t *this,
- inode_t *inode, unsigned char *locked_on,
- unsigned char *sources, unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *undid_pending,
- struct afr_reply *replies, gf_boolean_t *pflag)
+__afr_selfheal_data_prepare(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ unsigned char *locked_on, unsigned char *sources,
+ unsigned char *sinks, unsigned char *healed_sinks,
+ unsigned char *undid_pending,
+ struct afr_reply *replies, unsigned char *pflag)
{
- int ret = -1;
- int source = -1;
- afr_private_t *priv = NULL;
- uint64_t *witness = NULL;
-
- priv = this->private;
-
- ret = afr_selfheal_unlocked_discover (frame, inode, inode->gfid,
- replies);
-
- if (ret)
- return ret;
-
- witness = alloca0(priv->child_count * sizeof (*witness));
- ret = afr_selfheal_find_direction (frame, this, replies,
- AFR_DATA_TRANSACTION,
- locked_on, sources, sinks, witness,
- pflag);
- if (ret)
- return ret;
-
- /* Initialize the healed_sinks[] array optimistically to
- the intersection of to-be-healed (i.e sinks[]) and
- the list of servers which are up (i.e locked_on[]).
- As we encounter failures in the healing process, we
- will unmark the respective servers in the healed_sinks[]
- array.
- */
- AFR_INTERSECT (healed_sinks, sinks, locked_on, priv->child_count);
-
- source = __afr_selfheal_data_finalize_source (frame, this, inode,
- sources, sinks,
- healed_sinks,
- locked_on, undid_pending,
- replies, witness);
- if (source < 0)
- return -EIO;
-
- return source;
+ int ret = -1;
+ int source = -1;
+ afr_private_t *priv = NULL;
+ uint64_t *witness = NULL;
+
+ priv = this->private;
+
+ ret = afr_selfheal_unlocked_discover(frame, inode, inode->gfid, replies);
+
+ if (ret)
+ return ret;
+
+ witness = alloca0(priv->child_count * sizeof(*witness));
+ ret = afr_selfheal_find_direction(frame, this, replies,
+ AFR_DATA_TRANSACTION, locked_on, sources,
+ sinks, witness, pflag);
+ if (ret)
+ return ret;
+
+ /* Initialize the healed_sinks[] array optimistically to
+ the intersection of to-be-healed (i.e sinks[]) and
+ the list of servers which are up (i.e locked_on[]).
+ As we encounter failures in the healing process, we
+ will unmark the respective servers in the healed_sinks[]
+ array.
+ */
+ AFR_INTERSECT(healed_sinks, sinks, locked_on, priv->child_count);
+
+ source = __afr_selfheal_data_finalize_source(
+ frame, this, inode, sources, sinks, healed_sinks, locked_on,
+ undid_pending, replies, witness);
+ if (source < 0)
+ return -EIO;
+
+ return source;
}
-
static int
-__afr_selfheal_data (call_frame_t *frame, xlator_t *this, fd_t *fd,
- unsigned char *locked_on)
+__afr_selfheal_data(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ unsigned char *locked_on)
{
- afr_private_t *priv = NULL;
- int ret = -1;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *data_lock = NULL;
- unsigned char *healed_sinks = NULL;
- unsigned char *undid_pending = NULL;
- struct afr_reply *locked_replies = NULL;
- int source = -1;
- gf_boolean_t did_sh = _gf_true;
- gf_boolean_t is_arbiter_the_only_sink = _gf_false;
-
- priv = this->private;
-
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
- data_lock = alloca0 (priv->child_count);
- undid_pending = alloca0 (priv->child_count);
-
- locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
-
- ret = afr_selfheal_inodelk (frame, this, fd->inode, this->name, 0, 0,
- data_lock);
- {
- if (ret < priv->child_count) {
- gf_msg_debug (this->name, 0, "%s: Skipping "
- "self-heal as only %d number "
- "of subvolumes "
- "could be locked",
- uuid_utoa (fd->inode->gfid),
- ret);
- ret = -ENOTCONN;
- goto unlock;
- }
-
- ret = __afr_selfheal_data_prepare (frame, this, fd->inode,
- data_lock, sources, sinks,
- healed_sinks, undid_pending,
- locked_replies, NULL);
- if (ret < 0)
- goto unlock;
-
- if (AFR_COUNT(healed_sinks, priv->child_count) == 0) {
- did_sh = _gf_false;
- goto unlock;
- }
-
- source = ret;
-
- if (AFR_IS_ARBITER_BRICK(priv, source)) {
- did_sh = _gf_false;
- goto unlock;
- }
-
- if (priv->arbiter_count &&
- AFR_COUNT (healed_sinks, priv->child_count) == 1 &&
- healed_sinks[ARBITER_BRICK_INDEX]) {
- is_arbiter_the_only_sink = _gf_true;
- goto restore_time;
- }
-
- ret = __afr_selfheal_truncate_sinks (frame, this, fd, healed_sinks,
- locked_replies[source].poststat.ia_size);
- if (ret < 0)
- goto unlock;
-
- ret = 0;
-
- }
-unlock:
- afr_selfheal_uninodelk (frame, this, fd->inode, this->name, 0, 0,
- data_lock);
+ afr_private_t *priv = NULL;
+ int ret = -1;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ unsigned char *data_lock = NULL;
+ unsigned char *healed_sinks = NULL;
+ unsigned char *undid_pending = NULL;
+ struct afr_reply *locked_replies = NULL;
+ int source = -1;
+ gf_boolean_t did_sh = _gf_true;
+ gf_boolean_t is_arbiter_the_only_sink = _gf_false;
+ gf_boolean_t empty_file = _gf_false;
+
+ priv = this->private;
+
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ healed_sinks = alloca0(priv->child_count);
+ data_lock = alloca0(priv->child_count);
+ undid_pending = alloca0(priv->child_count);
+
+ locked_replies = alloca0(sizeof(*locked_replies) * priv->child_count);
+
+ ret = afr_selfheal_inodelk(frame, this, fd->inode, this->name, 0, 0,
+ data_lock);
+ {
+ if (ret < priv->child_count) {
+ gf_msg_debug(this->name, 0,
+ "%s: Skipping "
+ "self-heal as only %d number "
+ "of subvolumes "
+ "could be locked",
+ uuid_utoa(fd->inode->gfid), ret);
+ ret = -ENOTCONN;
+ goto unlock;
+ }
+
+ ret = __afr_selfheal_data_prepare(frame, this, fd->inode, data_lock,
+ sources, sinks, healed_sinks,
+ undid_pending, locked_replies, NULL);
+ if (ret < 0)
+ goto unlock;
+
+ if (AFR_COUNT(healed_sinks, priv->child_count) == 0) {
+ did_sh = _gf_false;
+ goto unlock;
+ }
+
+ source = ret;
+
+ if (AFR_IS_ARBITER_BRICK(priv, source)) {
+ empty_file = afr_is_file_empty_on_all_children(priv,
+ locked_replies);
+ if (empty_file)
+ goto restore_time;
+
+ did_sh = _gf_false;
+ goto unlock;
+ }
+
+ ret = __afr_selfheal_truncate_sinks(
+ frame, this, fd, healed_sinks,
+ locked_replies[source].poststat.ia_size);
if (ret < 0)
- goto out;
+ goto unlock;
- if (!did_sh)
- goto out;
+ if (priv->arbiter_count &&
+ AFR_COUNT(healed_sinks, priv->child_count) == 1 &&
+ healed_sinks[ARBITER_BRICK_INDEX]) {
+ is_arbiter_the_only_sink = _gf_true;
+ goto restore_time;
+ }
+ ret = 0;
+ }
+unlock:
+ afr_selfheal_uninodelk(frame, this, fd->inode, this->name, 0, 0, data_lock);
+ if (ret < 0)
+ goto out;
- ret = afr_selfheal_data_do (frame, this, fd, source, healed_sinks,
- locked_replies);
- if (ret)
- goto out;
+ if (!did_sh)
+ goto out;
+
+ ret = afr_selfheal_data_do(frame, this, fd, source, healed_sinks,
+ locked_replies);
+ if (ret)
+ goto out;
restore_time:
- afr_selfheal_restore_time (frame, this, fd->inode, source,
- healed_sinks, locked_replies);
-
- if (!is_arbiter_the_only_sink) {
- ret = afr_selfheal_inodelk (frame, this, fd->inode, this->name,
- 0, 0, data_lock);
- if (ret < priv->child_count) {
- ret = -ENOTCONN;
- did_sh = _gf_false;
- goto skip_undo_pending;
- }
+ afr_selfheal_restore_time(frame, this, fd->inode, source, healed_sinks,
+ locked_replies);
+
+ if (!is_arbiter_the_only_sink && !empty_file) {
+ ret = afr_selfheal_inodelk(frame, this, fd->inode, this->name, 0, 0,
+ data_lock);
+ if (ret < priv->child_count) {
+ ret = -ENOTCONN;
+ did_sh = _gf_false;
+ goto skip_undo_pending;
}
- ret = afr_selfheal_undo_pending (frame, this, fd->inode,
- sources, sinks, healed_sinks,
- undid_pending, AFR_DATA_TRANSACTION,
- locked_replies, data_lock);
+ }
+ ret = afr_selfheal_undo_pending(
+ frame, this, fd->inode, sources, sinks, healed_sinks, undid_pending,
+ AFR_DATA_TRANSACTION, locked_replies, data_lock);
skip_undo_pending:
- afr_selfheal_uninodelk (frame, this, fd->inode, this->name, 0, 0,
- data_lock);
+ afr_selfheal_uninodelk(frame, this, fd->inode, this->name, 0, 0, data_lock);
out:
- if (did_sh)
- afr_log_selfheal (fd->inode->gfid, this, ret, "data", source,
- sources, healed_sinks);
- else
- ret = 1;
+ if (did_sh)
+ afr_log_selfheal(fd->inode->gfid, this, ret, "data", source, sources,
+ healed_sinks);
+ else
+ ret = 1;
- if (locked_replies)
- afr_replies_wipe (locked_replies, priv->child_count);
+ if (locked_replies)
+ afr_replies_wipe(locked_replies, priv->child_count);
- return ret;
+ return ret;
}
int
-afr_selfheal_data_open_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- fd_t *fd, dict_t *xdata)
+afr_selfheal_data_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd,
+ dict_t *xdata)
{
- afr_local_t *local = NULL;
- int i = (long) cookie;
+ afr_local_t *local = NULL;
+ int i = (long)cookie;
- local = frame->local;
+ local = frame->local;
- local->replies[i].valid = 1;
- local->replies[i].op_ret = op_ret;
- local->replies[i].op_errno = op_errno;
+ local->replies[i].valid = 1;
+ local->replies[i].op_ret = op_ret;
+ local->replies[i].op_errno = op_errno;
- syncbarrier_wake (&local->barrier);
+ syncbarrier_wake(&local->barrier);
- return 0;
+ return 0;
}
int
-afr_selfheal_data_open (xlator_t *this, inode_t *inode, fd_t **fd)
+afr_selfheal_data_open(xlator_t *this, inode_t *inode, fd_t **fd)
{
- int ret = 0;
- fd_t *fd_tmp = NULL;
- loc_t loc = {0,};
- call_frame_t *frame = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
-
- priv = this->private;
-
- fd_tmp = fd_create (inode, 0);
- if (!fd_tmp)
- return -ENOMEM;
-
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
-
- frame = afr_frame_create (this, &ret);
- if (!frame) {
- ret = -ret;
- fd_unref (fd_tmp);
- goto out;
+ int ret = 0;
+ fd_t *fd_tmp = NULL;
+ loc_t loc = {
+ 0,
+ };
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+
+ priv = this->private;
+
+ fd_tmp = fd_create(inode, 0);
+ if (!fd_tmp)
+ return -ENOMEM;
+
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+
+ frame = afr_frame_create(this, &ret);
+ if (!frame) {
+ ret = -ret;
+ fd_unref(fd_tmp);
+ goto out;
+ }
+ local = frame->local;
+
+ AFR_ONLIST(local->child_up, frame, afr_selfheal_data_open_cbk, open, &loc,
+ O_RDWR | O_LARGEFILE, fd_tmp, NULL);
+
+ ret = -ENOTCONN;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+
+ if (local->replies[i].op_ret < 0) {
+ ret = -local->replies[i].op_errno;
+ continue;
}
- local = frame->local;
- AFR_ONLIST (local->child_up, frame, afr_selfheal_data_open_cbk, open,
- &loc, O_RDWR|O_LARGEFILE, fd_tmp, NULL);
+ ret = 0;
+ break;
+ }
- ret = -ENOTCONN;
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
-
- if (local->replies[i].op_ret < 0) {
- ret = -local->replies[i].op_errno;
- continue;
- }
+ if (ret < 0) {
+ fd_unref(fd_tmp);
+ goto out;
+ } else {
+ fd_bind(fd_tmp);
+ }
- ret = 0;
- break;
- }
-
- if (ret < 0) {
- fd_unref (fd_tmp);
- goto out;
- } else {
- fd_bind (fd_tmp);
- }
-
- *fd = fd_tmp;
+ *fd = fd_tmp;
out:
- loc_wipe (&loc);
- if (frame)
- AFR_STACK_DESTROY (frame);
- return ret;
+ loc_wipe(&loc);
+ if (frame)
+ AFR_STACK_DESTROY(frame);
+ return ret;
}
int
-afr_selfheal_data (call_frame_t *frame, xlator_t *this, fd_t *fd)
+afr_selfheal_data(call_frame_t *frame, xlator_t *this, fd_t *fd)
{
- afr_private_t *priv = NULL;
- unsigned char *locked_on = NULL;
- int ret = 0;
- inode_t *inode = fd->inode;
-
- priv = this->private;
-
- locked_on = alloca0 (priv->child_count);
-
- ret = afr_selfheal_tie_breaker_inodelk (frame, this, inode,
- priv->sh_domain, 0, 0,
- locked_on);
- {
- if (ret < priv->child_count) {
- gf_msg_debug (this->name, 0, "%s: Skipping "
- "self-heal as only %d number of "
- "subvolumes could be locked",
- uuid_utoa (fd->inode->gfid),
- ret);
- /* Either less than two subvols available, or another
- selfheal (from another server) is in progress. Skip
- for now in any case there isn't anything to do.
- */
- ret = -ENOTCONN;
- goto unlock;
- }
-
- ret = __afr_selfheal_data (frame, this, fd, locked_on);
- }
+ afr_private_t *priv = NULL;
+ unsigned char *locked_on = NULL;
+ int ret = 0;
+ inode_t *inode = fd->inode;
+
+ priv = this->private;
+
+ locked_on = alloca0(priv->child_count);
+
+ ret = afr_selfheal_tie_breaker_inodelk(frame, this, inode, priv->sh_domain,
+ 0, 0, locked_on);
+ {
+ if (ret < priv->child_count) {
+ gf_msg_debug(this->name, 0,
+ "%s: Skipping "
+ "self-heal as only %d number of "
+ "subvolumes could be locked",
+ uuid_utoa(fd->inode->gfid), ret);
+ /* Either less than two subvols available, or another
+ selfheal (from another server) is in progress. Skip
+ for now in any case there isn't anything to do.
+ */
+ ret = -ENOTCONN;
+ goto unlock;
+ }
+
+ ret = __afr_selfheal_data(frame, this, fd, locked_on);
+ }
unlock:
- afr_selfheal_uninodelk (frame, this, inode, priv->sh_domain, 0, 0,
- locked_on);
+ afr_selfheal_uninodelk(frame, this, inode, priv->sh_domain, 0, 0,
+ locked_on);
- return ret;
+ return ret;
}
diff --git a/xlators/cluster/afr/src/afr-self-heal-entry.c b/xlators/cluster/afr/src/afr-self-heal-entry.c
index b601040504e..64893f441e3 100644
--- a/xlators/cluster/afr/src/afr-self-heal-entry.c
+++ b/xlators/cluster/afr/src/afr-self-heal-entry.c
@@ -8,1115 +8,1269 @@
cases as published by the Free Software Foundation.
*/
-
#include "afr.h"
#include "afr-self-heal.h"
-#include "byte-order.h"
+#include <glusterfs/byte-order.h>
#include "afr-transaction.h"
#include "afr-messages.h"
-#include "syncop-utils.h"
-#include "events.h"
+#include <glusterfs/syncop-utils.h>
+#include <glusterfs/events.h>
-static int
-afr_selfheal_entry_delete (xlator_t *this, inode_t *dir, const char *name,
- inode_t *inode, int child, struct afr_reply *replies)
+int
+afr_selfheal_entry_anon_inode(xlator_t *this, inode_t *dir, const char *name,
+ inode_t *inode, int child,
+ struct afr_reply *replies,
+ gf_boolean_t *anon_inode)
{
- afr_private_t *priv = NULL;
- xlator_t *subvol = NULL;
- int ret = 0;
- loc_t loc = {0, };
- char g[64];
-
- priv = this->private;
-
- subvol = priv->children[child];
-
- loc.parent = inode_ref (dir);
- gf_uuid_copy (loc.pargfid, dir->gfid);
- loc.name = name;
- loc.inode = inode_ref (inode);
-
- if (replies[child].valid && replies[child].op_ret == 0) {
- switch (replies[child].poststat.ia_type) {
- case IA_IFDIR:
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_EXPUNGING_FILE_OR_DIR,
- "expunging dir %s/%s (%s) on %s",
- uuid_utoa (dir->gfid), name,
- uuid_utoa_r (replies[child].poststat.ia_gfid, g),
- subvol->name);
- ret = syncop_rmdir (subvol, &loc, 1, NULL, NULL);
- break;
- default:
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_EXPUNGING_FILE_OR_DIR,
- "expunging file %s/%s (%s) on %s",
- uuid_utoa (dir->gfid), name,
- uuid_utoa_r (replies[child].poststat.ia_gfid, g),
- subvol->name);
- ret = syncop_unlink (subvol, &loc, NULL, NULL);
- break;
- }
- }
-
- loc_wipe (&loc);
-
- return ret;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+ int ret = 0;
+ int i = 0;
+ char g[64] = {0};
+ unsigned char *lookup_success = NULL;
+ call_frame_t *frame = NULL;
+ loc_t loc2 = {
+ 0,
+ };
+ loc_t loc = {
+ 0,
+ };
+
+ priv = this->private;
+ subvol = priv->children[child];
+ lookup_success = alloca0(priv->child_count);
+ uuid_utoa_r(replies[child].poststat.ia_gfid, g);
+ loc.inode = inode_new(inode->table);
+ if (!loc.inode) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ if (replies[child].poststat.ia_type == IA_IFDIR) {
+ /* This directory may have sub-directory hierarchy which may need to
+ * be preserved for subsequent heals. So unconditionally move the
+ * directory to anonymous-inode directory*/
+ *anon_inode = _gf_true;
+ goto anon_inode;
+ }
+
+ frame = afr_frame_create(this, &ret);
+ if (!frame) {
+ ret = -ret;
+ goto out;
+ }
+ local = frame->local;
+ gf_uuid_copy(loc.gfid, replies[child].poststat.ia_gfid);
+ AFR_ONLIST(local->child_up, frame, afr_selfheal_discover_cbk, lookup, &loc,
+ NULL);
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->replies[i].op_ret == 0) {
+ lookup_success[i] = 1;
+ } else if (local->replies[i].op_errno != ENOENT &&
+ local->replies[i].op_errno != ESTALE) {
+ ret = -local->replies[i].op_errno;
+ }
+ }
+
+ if (priv->quorum_count) {
+ if (afr_has_quorum(lookup_success, this, NULL)) {
+ *anon_inode = _gf_true;
+ }
+ } else if (AFR_COUNT(lookup_success, priv->child_count) > 1) {
+ *anon_inode = _gf_true;
+ } else if (ret) {
+ goto out;
+ }
+
+anon_inode:
+ if (!*anon_inode) {
+ ret = 0;
+ goto out;
+ }
+
+ loc.parent = inode_ref(dir);
+ gf_uuid_copy(loc.pargfid, dir->gfid);
+ loc.name = name;
+
+ ret = afr_anon_inode_create(this, child, &loc2.parent);
+ if (ret < 0)
+ goto out;
+
+ loc2.name = g;
+ ret = syncop_rename(subvol, &loc, &loc2, NULL, NULL);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, AFR_MSG_EXPUNGING_FILE_OR_DIR,
+ "Rename to %s dir %s/%s (%s) on %s failed",
+ priv->anon_inode_name, uuid_utoa(dir->gfid), name, g,
+ subvol->name);
+ } else {
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_EXPUNGING_FILE_OR_DIR,
+ "Rename to %s dir %s/%s (%s) on %s successful",
+ priv->anon_inode_name, uuid_utoa(dir->gfid), name, g,
+ subvol->name);
+ }
+
+out:
+ loc_wipe(&loc);
+ loc_wipe(&loc2);
+ if (frame) {
+ AFR_STACK_DESTROY(frame);
+ }
+
+ return ret;
}
int
-afr_selfheal_recreate_entry (call_frame_t *frame, int dst, int source,
- unsigned char *sources, inode_t *dir,
- const char *name, inode_t *inode,
- struct afr_reply *replies)
+afr_selfheal_entry_delete(xlator_t *this, inode_t *dir, const char *name,
+ inode_t *inode, int child, struct afr_reply *replies)
{
- int ret = 0;
- loc_t loc = {0,};
- loc_t srcloc = {0,};
- xlator_t *this = frame->this;
- afr_private_t *priv = NULL;
- dict_t *xdata = NULL;
- struct iatt *iatt = NULL;
- char *linkname = NULL;
- mode_t mode = 0;
- struct iatt newent = {0,};
- unsigned char *newentry = NULL;
-
- priv = this->private;
- iatt = &replies[source].poststat;
- if (iatt->ia_type == IA_INVAL || gf_uuid_is_null (iatt->ia_gfid)) {
- gf_msg (this->name, GF_LOG_ERROR, 0, AFR_MSG_SELF_HEAL_FAILED,
- "Invalid ia_type (%d) or gfid(%s). source brick=%d, "
- "pargfid=%s, name=%s", iatt->ia_type,
- uuid_utoa(iatt->ia_gfid), source,
- uuid_utoa(dir->gfid), name);
- ret = -EINVAL;
- goto out;
- }
+ char g[64] = {0};
+ afr_private_t *priv = NULL;
+ xlator_t *subvol = NULL;
+ int ret = 0;
+ loc_t loc = {
+ 0,
+ };
+ gf_boolean_t anon_inode = _gf_false;
+
+ priv = this->private;
+ subvol = priv->children[child];
+
+ if ((!replies[child].valid) || (replies[child].op_ret < 0)) {
+ /*Nothing to do*/
+ ret = 0;
+ goto out;
+ }
+
+ if (priv->use_anon_inode) {
+ ret = afr_selfheal_entry_anon_inode(this, dir, name, inode, child,
+ replies, &anon_inode);
+ if (ret < 0 || anon_inode)
+ goto out;
+ }
+
+ loc.parent = inode_ref(dir);
+ loc.inode = inode_new(inode->table);
+ if (!loc.inode) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ loc.name = name;
+ switch (replies[child].poststat.ia_type) {
+ case IA_IFDIR:
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_EXPUNGING_FILE_OR_DIR,
+ "expunging dir %s/%s (%s) on %s", uuid_utoa(dir->gfid), name,
+ uuid_utoa_r(replies[child].poststat.ia_gfid, g),
+ subvol->name);
+ ret = syncop_rmdir(subvol, &loc, 1, NULL, NULL);
+ break;
+ default:
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_EXPUNGING_FILE_OR_DIR,
+ "expunging file %s/%s (%s) on %s", uuid_utoa(dir->gfid),
+ name, uuid_utoa_r(replies[child].poststat.ia_gfid, g),
+ subvol->name);
+ ret = syncop_unlink(subvol, &loc, NULL, NULL);
+ break;
+ }
- xdata = dict_new();
- if (!xdata)
- return -ENOMEM;
- newentry = alloca0 (priv->child_count);
- loc.parent = inode_ref (dir);
- gf_uuid_copy (loc.pargfid, dir->gfid);
- loc.name = name;
- loc.inode = inode_ref (inode);
-
- ret = afr_selfheal_entry_delete (this, dir, name, inode, dst, replies);
- if (ret)
- goto out;
-
- ret = dict_set_gfuuid (xdata, "gfid-req",
- replies[source].poststat.ia_gfid, true);
- if (ret)
- goto out;
-
- srcloc.inode = inode_ref (inode);
- gf_uuid_copy (srcloc.gfid, iatt->ia_gfid);
- if (iatt->ia_type != IA_IFDIR)
- ret = syncop_lookup (priv->children[dst], &srcloc, 0, 0, 0, 0);
- if (iatt->ia_type == IA_IFDIR || ret == -ENOENT || ret == -ESTALE) {
- newentry[dst] = 1;
- ret = afr_selfheal_newentry_mark (frame, this, inode, source,
- replies, sources, newentry);
- if (ret)
- goto out;
- }
+out:
+ loc_wipe(&loc);
+ return ret;
+}
- mode = st_mode_from_ia (iatt->ia_prot, iatt->ia_type);
-
- switch (iatt->ia_type) {
- case IA_IFDIR:
- ret = syncop_mkdir (priv->children[dst], &loc, mode, 0,
- xdata, NULL);
- break;
- case IA_IFLNK:
- if (!newentry[dst]) {
- ret = syncop_link (priv->children[dst], &srcloc, &loc,
- &newent, NULL, NULL);
- } else {
- ret = syncop_readlink (priv->children[source], &srcloc,
- &linkname, 4096, NULL, NULL);
- if (ret <= 0)
- goto out;
- ret = syncop_symlink (priv->children[dst], &loc,
- linkname, NULL, xdata, NULL);
- }
- break;
- default:
- ret = dict_set_int32 (xdata, GLUSTERFS_INTERNAL_FOP_KEY, 1);
- if (ret)
- goto out;
- ret = syncop_mknod (priv->children[dst], &loc, mode,
- makedev (ia_major(iatt->ia_rdev), ia_minor (iatt->ia_rdev)),
- &newent, xdata, NULL);
- break;
- }
+int
+afr_selfheal_recreate_entry(call_frame_t *frame, int dst, int source,
+ unsigned char *sources, inode_t *dir,
+ const char *name, inode_t *inode,
+ struct afr_reply *replies)
+{
+ int ret = 0;
+ loc_t loc = {
+ 0,
+ };
+ loc_t srcloc = {
+ 0,
+ };
+ loc_t anonloc = {
+ 0,
+ };
+ xlator_t *this = frame->this;
+ afr_private_t *priv = NULL;
+ dict_t *xdata = NULL;
+ struct iatt *iatt = NULL;
+ char *linkname = NULL;
+ mode_t mode = 0;
+ struct iatt newent = {
+ 0,
+ };
+ unsigned char *newentry = NULL;
+ char iatt_uuid_str[64] = {0};
+ char dir_uuid_str[64] = {0};
+
+ priv = this->private;
+ iatt = &replies[source].poststat;
+ uuid_utoa_r(iatt->ia_gfid, iatt_uuid_str);
+ if (iatt->ia_type == IA_INVAL || gf_uuid_is_null(iatt->ia_gfid)) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SELF_HEAL_FAILED,
+ "Invalid ia_type (%d) or gfid(%s). source brick=%d, "
+ "pargfid=%s, name=%s",
+ iatt->ia_type, iatt_uuid_str, source,
+ uuid_utoa_r(dir->gfid, dir_uuid_str), name);
+ ret = -EINVAL;
+ goto out;
+ }
+
+ xdata = dict_new();
+ if (!xdata)
+ return -ENOMEM;
+ newentry = alloca0(priv->child_count);
+ loc.parent = inode_ref(dir);
+ gf_uuid_copy(loc.pargfid, dir->gfid);
+ loc.name = name;
+ loc.inode = inode_ref(inode);
+
+ ret = afr_selfheal_entry_delete(this, dir, name, inode, dst, replies);
+ if (ret)
+ goto out;
+
+ ret = dict_set_gfuuid(xdata, "gfid-req", replies[source].poststat.ia_gfid,
+ true);
+ if (ret)
+ goto out;
+
+ srcloc.inode = inode_ref(inode);
+ gf_uuid_copy(srcloc.gfid, iatt->ia_gfid);
+ ret = syncop_lookup(priv->children[dst], &srcloc, 0, 0, 0, 0);
+ if (ret == -ENOENT || ret == -ESTALE) {
+ newentry[dst] = 1;
+ ret = afr_selfheal_newentry_mark(frame, this, inode, source, replies,
+ sources, newentry);
+ if (ret)
+ goto out;
+ } else if (ret == 0 && iatt->ia_type == IA_IFDIR && priv->use_anon_inode) {
+ // Try rename from hidden directory
+ ret = afr_anon_inode_create(this, dst, &anonloc.parent);
+ if (ret < 0)
+ goto out;
+ anonloc.inode = inode_ref(inode);
+ anonloc.name = iatt_uuid_str;
+ ret = syncop_rename(priv->children[dst], &anonloc, &loc, NULL, NULL);
+ if (ret == -ENOENT || ret == -ESTALE)
+ ret = -1; /*This sets 'mismatch' to true*/
+ goto out;
+ }
+
+ mode = st_mode_from_ia(iatt->ia_prot, iatt->ia_type);
+
+ switch (iatt->ia_type) {
+ case IA_IFDIR:
+ ret = syncop_mkdir(priv->children[dst], &loc, mode, 0, xdata, NULL);
+ break;
+ case IA_IFLNK:
+ if (!newentry[dst]) {
+ ret = syncop_link(priv->children[dst], &srcloc, &loc, &newent,
+ NULL, NULL);
+ } else {
+ ret = syncop_readlink(priv->children[source], &srcloc,
+ &linkname, 4096, NULL, NULL);
+ if (ret <= 0)
+ goto out;
+ ret = syncop_symlink(priv->children[dst], &loc, linkname, NULL,
+ xdata, NULL);
+ }
+ break;
+ default:
+ ret = dict_set_int32_sizen(xdata, GLUSTERFS_INTERNAL_FOP_KEY, 1);
+ if (ret)
+ goto out;
+ ret = syncop_mknod(
+ priv->children[dst], &loc, mode,
+ makedev(ia_major(iatt->ia_rdev), ia_minor(iatt->ia_rdev)),
+ &newent, xdata, NULL);
+ break;
+ }
out:
- if (xdata)
- dict_unref (xdata);
- GF_FREE (linkname);
- loc_wipe (&loc);
- loc_wipe (&srcloc);
- return ret;
+ if (xdata)
+ dict_unref(xdata);
+ GF_FREE(linkname);
+ loc_wipe(&loc);
+ loc_wipe(&srcloc);
+ loc_wipe(&anonloc);
+ return ret;
}
static int
-__afr_selfheal_heal_dirent (call_frame_t *frame, xlator_t *this, fd_t *fd,
- char *name, inode_t *inode, int source,
- unsigned char *sources, unsigned char *healed_sinks,
- unsigned char *locked_on, struct afr_reply *replies)
+__afr_selfheal_heal_dirent(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ char *name, inode_t *inode, int source,
+ unsigned char *sources, unsigned char *healed_sinks,
+ unsigned char *locked_on, struct afr_reply *replies)
{
- int ret = 0;
- afr_private_t *priv = NULL;
- int i = 0;
+ int ret = 0;
+ afr_private_t *priv = NULL;
+ int i = 0;
- priv = this->private;
+ priv = this->private;
- if (!replies[source].valid)
- return -EIO;
+ if (!replies[source].valid)
+ return -EIO;
- /* Skip healing this entry if the last lookup on it failed for reasons
- * other than ENOENT.
- */
- if ((replies[source].op_ret < 0) &&
- (replies[source].op_errno != ENOENT))
- return -replies[source].op_errno;
-
- if (replies[source].op_ret == 0) {
- ret = afr_lookup_and_heal_gfid (this, fd->inode, name,
- inode, replies, source,
- &replies[source].poststat.ia_gfid);
- if (ret)
- return ret;
+ /* Skip healing this entry if the last lookup on it failed for reasons
+ * other than ENOENT.
+ */
+ if ((replies[source].op_ret < 0) && (replies[source].op_errno != ENOENT))
+ return -replies[source].op_errno;
+
+ if (replies[source].op_ret == 0) {
+ ret = afr_lookup_and_heal_gfid(this, fd->inode, name, inode, replies,
+ source, sources,
+ &replies[source].poststat.ia_gfid, NULL);
+ if (ret)
+ return ret;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!healed_sinks[i])
+ continue;
+ if (replies[source].op_ret == -1 &&
+ replies[source].op_errno == ENOENT) {
+ ret = afr_selfheal_entry_delete(this, fd->inode, name, inode, i,
+ replies);
+ } else {
+ if (!gf_uuid_compare(replies[i].poststat.ia_gfid,
+ replies[source].poststat.ia_gfid))
+ continue;
+
+ ret = afr_selfheal_recreate_entry(frame, i, source, sources,
+ fd->inode, name, inode, replies);
}
+ if (ret < 0)
+ break;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (!healed_sinks[i])
- continue;
- if (replies[source].op_ret == -1 &&
- replies[source].op_errno == ENOENT) {
- ret = afr_selfheal_entry_delete (this, fd->inode, name,
- inode, i, replies);
- } else {
- if (!gf_uuid_compare (replies[i].poststat.ia_gfid,
- replies[source].poststat.ia_gfid))
- continue;
-
- ret = afr_selfheal_recreate_entry (frame, i, source,
- sources, fd->inode,
- name, inode,
- replies);
- }
- if (ret < 0)
- break;
- }
-
- return ret;
+ return ret;
}
static int
-afr_selfheal_detect_gfid_and_type_mismatch (xlator_t *this,
- struct afr_reply *replies,
- inode_t *inode,
- uuid_t pargfid,
- char *bname, int src_idx,
- unsigned char *locked_on,
- int *src)
+afr_selfheal_detect_gfid_and_type_mismatch(xlator_t *this,
+ struct afr_reply *replies,
+ inode_t *inode, uuid_t pargfid,
+ char *bname, int src_idx,
+ unsigned char *locked_on, int *src)
{
- int i = 0;
- int ret = -1;
- afr_private_t *priv = NULL;
- void *gfid = NULL;
- ia_type_t ia_type = IA_INVAL;
+ int i = 0;
+ int ret = -1;
+ afr_private_t *priv = NULL;
+ void *gfid = NULL;
+ ia_type_t ia_type = IA_INVAL;
- priv = this->private;
- gfid = &replies[src_idx].poststat.ia_gfid;
- ia_type = replies[src_idx].poststat.ia_type;
+ priv = this->private;
+ gfid = &replies[src_idx].poststat.ia_gfid;
+ ia_type = replies[src_idx].poststat.ia_type;
- for (i = 0; i < priv->child_count; i++) {
- if (i == src_idx)
- continue;
-
- if (!replies[i].valid)
- continue;
-
- if (replies[i].op_ret != 0)
- continue;
-
- if (gf_uuid_compare (gfid, replies[i].poststat.ia_gfid) &&
- (ia_type == replies[i].poststat.ia_type)) {
- ret = afr_gfid_split_brain_source (this, replies, inode,
- pargfid, bname,
- src_idx, i,
- locked_on, src,
- NULL);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN,
- "Skipping conservative merge on the "
- "file.");
- return ret;
- }
-
- if (ia_type != replies[i].poststat.ia_type) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_SPLIT_BRAIN, "Type mismatch detected "
- "for <gfid:%s>/%s>, %s on %s and %s on %s. "
- "Skipping conservative merge on the file.",
- uuid_utoa (pargfid), bname,
- gf_inode_type_to_str (replies[i].poststat.ia_type),
- priv->children[i]->name,
- gf_inode_type_to_str (replies[src_idx].poststat.ia_type),
- priv->children[src_idx]->name);
- gf_event (EVENT_AFR_SPLIT_BRAIN, "subvol=%s;type=file;"
- "file=<gfid:%s>/%s>;count=2;child-%d=%s;type-"
- "%d=%s;child-%d=%s;type-%d=%s",
- this->name, uuid_utoa (pargfid), bname, i,
- priv->children[i]->name, i,
- gf_inode_type_to_str(replies[i].poststat.ia_type),
- src_idx, priv->children[src_idx]->name, src_idx,
- gf_inode_type_to_str(replies[src_idx].poststat.ia_type));
- return -1;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == src_idx)
+ continue;
+
+ if (!replies[i].valid)
+ continue;
+
+ if (replies[i].op_ret != 0)
+ continue;
+
+ if (gf_uuid_is_null(replies[i].poststat.ia_gfid))
+ continue;
+
+ if (replies[i].poststat.ia_type == IA_INVAL)
+ continue;
+
+ if (ia_type == IA_INVAL || gf_uuid_is_null(gfid)) {
+ src_idx = i;
+ ia_type = replies[src_idx].poststat.ia_type;
+ gfid = &replies[src_idx].poststat.ia_gfid;
+ continue;
}
- return 0;
+ if (gf_uuid_compare(gfid, replies[i].poststat.ia_gfid) &&
+ (ia_type == replies[i].poststat.ia_type)) {
+ ret = afr_gfid_split_brain_source(this, replies, inode, pargfid,
+ bname, src_idx, i, locked_on, src,
+ NULL);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "Skipping conservative merge on the "
+ "file.");
+ return ret;
+ }
+
+ if (ia_type != replies[i].poststat.ia_type) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_SPLIT_BRAIN,
+ "Type mismatch detected "
+ "for <gfid:%s>/%s>, %s on %s and %s on %s. "
+ "Skipping conservative merge on the file.",
+ uuid_utoa(pargfid), bname,
+ gf_inode_type_to_str(replies[i].poststat.ia_type),
+ priv->children[i]->name,
+ gf_inode_type_to_str(replies[src_idx].poststat.ia_type),
+ priv->children[src_idx]->name);
+ gf_event(EVENT_AFR_SPLIT_BRAIN,
+ "client-pid=%d;"
+ "subvol=%s;type=file;"
+ "file=<gfid:%s>/%s>;count=2;child-%d=%s;type-"
+ "%d=%s;child-%d=%s;type-%d=%s",
+ this->ctx->cmd_args.client_pid, this->name,
+ uuid_utoa(pargfid), bname, i, priv->children[i]->name, i,
+ gf_inode_type_to_str(replies[i].poststat.ia_type), src_idx,
+ priv->children[src_idx]->name, src_idx,
+ gf_inode_type_to_str(replies[src_idx].poststat.ia_type));
+ return -1;
+ }
+ }
+
+ return 0;
}
static int
-__afr_selfheal_merge_dirent (call_frame_t *frame, xlator_t *this, fd_t *fd,
- char *name, inode_t *inode, unsigned char *sources,
- unsigned char *healed_sinks, unsigned char *locked_on,
- struct afr_reply *replies)
+__afr_selfheal_merge_dirent(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ char *name, inode_t *inode, unsigned char *sources,
+ unsigned char *healed_sinks,
+ unsigned char *locked_on, struct afr_reply *replies)
{
- int ret = 0;
- int i = 0;
- int source = -1;
- int src = -1;
- afr_private_t *priv = NULL;
+ int ret = 0;
+ int i = 0;
+ int source = -1;
+ int src = -1;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (replies[i].valid && replies[i].op_ret == 0) {
+ source = i;
+ break;
+ }
+ }
+
+ if (source == -1) {
+ /* entry got deleted in the mean time? */
+ return 0;
+ }
- priv = this->private;
+ /* Set all the sources as 1, otheriwse newentry_mark won't be set */
+ for (i = 0; i < priv->child_count; i++) {
+ if (replies[i].valid && replies[i].op_ret == 0) {
+ sources[i] = 1;
+ }
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (replies[i].valid && replies[i].op_ret == 0) {
- source = i;
- break;
- }
- }
-
- if (source == -1) {
- /* entry got deleted in the mean time? */
- return 0;
- }
-
- /* Set all the sources as 1, otheriwse newentry_mark won't be set */
- for (i = 0; i < priv->child_count; i++) {
- if (replies[i].valid && replies[i].op_ret == 0) {
- sources[i] = 1;
- }
- }
-
- ret = afr_lookup_and_heal_gfid (this, fd->inode, name, inode, replies,
- source,
- &replies[source].poststat.ia_gfid);
- if (ret)
- return ret;
+ ret = afr_lookup_and_heal_gfid(this, fd->inode, name, inode, replies,
+ source, sources,
+ &replies[source].poststat.ia_gfid, NULL);
+ if (ret)
+ return ret;
- /* In case of type mismatch / unable to resolve gfid mismatch on the
- * entry, return -1.*/
- ret = afr_selfheal_detect_gfid_and_type_mismatch (this, replies, inode,
- fd->inode->gfid,
- name, source,
- locked_on, &src);
+ /* In case of type mismatch / unable to resolve gfid mismatch on the
+ * entry, return -1.*/
+ ret = afr_selfheal_detect_gfid_and_type_mismatch(
+ this, replies, inode, fd->inode->gfid, name, source, locked_on, &src);
- if (ret < 0)
- return ret;
- if (src != -1) {
- source = src;
- for (i = 0; i < priv->child_count; i++) {
- if (i != src && replies[i].valid &&
- gf_uuid_compare (replies[src].poststat.ia_gfid,
- replies[i].poststat.ia_gfid)) {
- sources[i] = 0;
- }
- }
+ if (ret < 0)
+ return ret;
+ if (src != -1) {
+ source = src;
+ for (i = 0; i < priv->child_count; i++) {
+ if (i != src && replies[i].valid &&
+ gf_uuid_compare(replies[src].poststat.ia_gfid,
+ replies[i].poststat.ia_gfid)) {
+ sources[i] = 0;
+ }
}
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (i == source || !healed_sinks[i])
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == source || !healed_sinks[i])
+ continue;
- if (src != -1) {
- if (!gf_uuid_compare (replies[src].poststat.ia_gfid,
- replies[i].poststat.ia_gfid))
- continue;
- } else if (replies[i].op_errno != ENOENT) {
- continue;
- }
+ if (src != -1) {
+ if (!gf_uuid_compare(replies[src].poststat.ia_gfid,
+ replies[i].poststat.ia_gfid))
+ continue;
+ } else if (replies[i].op_errno != ENOENT) {
+ continue;
+ }
- ret |= afr_selfheal_recreate_entry (frame, i, source, sources,
- fd->inode, name, inode,
- replies);
- }
+ ret |= afr_selfheal_recreate_entry(frame, i, source, sources, fd->inode,
+ name, inode, replies);
+ }
- return ret;
+ return ret;
}
-
static int
-__afr_selfheal_entry_dirent (call_frame_t *frame, xlator_t *this, fd_t *fd,
- char *name, inode_t *inode, int source,
- unsigned char *sources, unsigned char *healed_sinks,
- unsigned char *locked_on,
- struct afr_reply *replies)
+__afr_selfheal_entry_dirent(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ char *name, inode_t *inode, int source,
+ unsigned char *sources, unsigned char *healed_sinks,
+ unsigned char *locked_on, struct afr_reply *replies)
{
- int ret = -1;
-
- if (source < 0)
- ret = __afr_selfheal_merge_dirent (frame, this, fd, name, inode,
- sources, healed_sinks,
- locked_on, replies);
- else
- ret = __afr_selfheal_heal_dirent (frame, this, fd, name, inode,
- source, sources, healed_sinks,
- locked_on, replies);
- return ret;
+ int ret = -1;
+
+ if (source < 0)
+ ret = __afr_selfheal_merge_dirent(frame, this, fd, name, inode, sources,
+ healed_sinks, locked_on, replies);
+ else
+ ret = __afr_selfheal_heal_dirent(frame, this, fd, name, inode, source,
+ sources, healed_sinks, locked_on,
+ replies);
+ return ret;
}
static gf_boolean_t
-is_full_heal_marker_present (xlator_t *this, dict_t *xdata, int idx)
+is_full_heal_marker_present(xlator_t *this, dict_t *xdata, int idx)
{
- int i = 0;
- int pending[3] = {0,};
- void *pending_raw = NULL;
- afr_private_t *priv = NULL;
+ int i = 0;
+ int pending[3] = {
+ 0,
+ };
+ void *pending_raw = NULL;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (!xdata)
- return _gf_false;
+ if (!xdata)
+ return _gf_false;
- /* Iterate over each of the priv->pending_keys[] elements and then
- * see if any of them have data segment non-zero. If they do, return
- * true. Else return false.
- */
- for (i = 0; i < priv->child_count; i++) {
- if (dict_get_ptr (xdata, priv->pending_key[i], &pending_raw))
- continue;
+ /* Iterate over each of the priv->pending_keys[] elements and then
+ * see if any of them have data segment non-zero. If they do, return
+ * true. Else return false.
+ */
+ for (i = 0; i < priv->child_count; i++) {
+ if (dict_get_ptr(xdata, priv->pending_key[i], &pending_raw))
+ continue;
- if (!pending_raw)
- continue;
+ if (!pending_raw)
+ continue;
- memcpy (pending, pending_raw, sizeof (pending));
- if (ntoh32 (pending[idx]))
- return _gf_true;
- }
+ memcpy(pending, pending_raw, sizeof(pending));
+ if (ntoh32(pending[idx]))
+ return _gf_true;
+ }
- return _gf_false;
+ return _gf_false;
}
static gf_boolean_t
-afr_need_full_heal (xlator_t *this, struct afr_reply *replies, int source,
- unsigned char *healed_sinks, afr_transaction_type type)
+afr_need_full_heal(xlator_t *this, struct afr_reply *replies, int source,
+ unsigned char *healed_sinks, afr_transaction_type type)
{
- int i = 0;
- int idx = 0;
- afr_private_t *priv = NULL;
+ int i = 0;
+ int idx = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (!priv->esh_granular)
- return _gf_true;
+ if (!priv->esh_granular)
+ return _gf_true;
- if (type != AFR_ENTRY_TRANSACTION)
- return _gf_true;
+ if (type != AFR_ENTRY_TRANSACTION)
+ return _gf_true;
- priv = this->private;
- idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
+ priv = this->private;
+ idx = afr_index_for_transaction_type(AFR_DATA_TRANSACTION);
- /* If there is a clear source, check whether the full-heal-indicator
- * is present in its xdata. Otherwise, we need to examine all the
- * participating bricks and then figure if *even* one of them has a
- * full-heal-indicator.
- */
+ /* If there is a clear source, check whether the full-heal-indicator
+ * is present in its xdata. Otherwise, we need to examine all the
+ * participating bricks and then figure if *even* one of them has a
+ * full-heal-indicator.
+ */
- if (source != -1) {
- if (is_full_heal_marker_present (this, replies[source].xdata,
- idx))
- return _gf_true;
- }
+ if (source != -1) {
+ if (is_full_heal_marker_present(this, replies[source].xdata, idx))
+ return _gf_true;
+ }
- /* else ..*/
+ /* else ..*/
- for (i = 0; i < priv->child_count; i++) {
- if (!healed_sinks[i])
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!healed_sinks[i])
+ continue;
- if (is_full_heal_marker_present (this, replies[i].xdata, idx))
- return _gf_true;
- }
+ if (is_full_heal_marker_present(this, replies[i].xdata, idx))
+ return _gf_true;
+ }
- return _gf_false;
+ return _gf_false;
}
static int
-__afr_selfheal_entry_finalize_source (xlator_t *this, unsigned char *sources,
- unsigned char *healed_sinks,
- unsigned char *locked_on,
- struct afr_reply *replies,
- uint64_t *witness)
+__afr_selfheal_entry_finalize_source(xlator_t *this, unsigned char *sources,
+ unsigned char *healed_sinks,
+ unsigned char *locked_on,
+ struct afr_reply *replies,
+ uint64_t *witness)
{
- afr_private_t *priv = NULL;
- int source = -1;
- int sources_count = 0;
+ afr_private_t *priv = NULL;
+ int source = -1;
+ int sources_count = 0;
+ int i = 0;
- priv = this->private;
+ priv = this->private;
- sources_count = AFR_COUNT (sources, priv->child_count);
+ sources_count = AFR_COUNT(sources, priv->child_count);
- if ((AFR_CMP (locked_on, healed_sinks, priv->child_count) == 0)
- || !sources_count || afr_does_witness_exist (this, witness)) {
+ if ((AFR_CMP(locked_on, healed_sinks, priv->child_count) == 0) ||
+ !sources_count || afr_does_witness_exist(this, witness)) {
+ memset(sources, 0, sizeof(*sources) * priv->child_count);
+ afr_mark_active_sinks(this, sources, locked_on, healed_sinks);
+ return -1;
+ }
- memset (sources, 0, sizeof (*sources) * priv->child_count);
- afr_mark_active_sinks (this, sources, locked_on, healed_sinks);
- return -1;
- }
+ source = afr_choose_source_by_policy(priv, sources, AFR_ENTRY_TRANSACTION);
+
+ /*If the selected source does not blame any other brick, then mark
+ * everything as sink to trigger conservative merge.
+ */
+ if (source != -1 && !AFR_COUNT(healed_sinks, priv->child_count)) {
+ for (i = 0; i < priv->child_count; i++) {
+ if (locked_on[i]) {
+ sources[i] = 0;
+ healed_sinks[i] = 1;
+ }
+ }
+ return -1;
+ }
- source = afr_choose_source_by_policy (priv, sources,
- AFR_ENTRY_TRANSACTION);
- return source;
+ return source;
}
int
-__afr_selfheal_entry_prepare (call_frame_t *frame, xlator_t *this,
- inode_t *inode, unsigned char *locked_on,
- unsigned char *sources, unsigned char *sinks,
- unsigned char *healed_sinks,
- struct afr_reply *replies, int *source_p,
- gf_boolean_t *pflag)
+__afr_selfheal_entry_prepare(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, unsigned char *locked_on,
+ unsigned char *sources, unsigned char *sinks,
+ unsigned char *healed_sinks,
+ struct afr_reply *replies, int *source_p,
+ unsigned char *pflag)
{
- int ret = -1;
- int source = -1;
- afr_private_t *priv = NULL;
- uint64_t *witness = NULL;
+ int ret = -1;
+ int source = -1;
+ afr_private_t *priv = NULL;
+ uint64_t *witness = NULL;
- priv = this->private;
+ priv = this->private;
- ret = afr_selfheal_unlocked_discover (frame, inode, inode->gfid,
- replies);
- if (ret)
- return ret;
-
- witness = alloca0 (sizeof (*witness) * priv->child_count);
- ret = afr_selfheal_find_direction (frame, this, replies,
- AFR_ENTRY_TRANSACTION,
- locked_on, sources, sinks, witness,
- pflag);
- if (ret)
- return ret;
-
- /* Initialize the healed_sinks[] array optimistically to
- the intersection of to-be-healed (i.e sinks[]) and
- the list of servers which are up (i.e locked_on[]).
-
- As we encounter failures in the healing process, we
- will unmark the respective servers in the healed_sinks[]
- array.
- */
- AFR_INTERSECT (healed_sinks, sinks, locked_on, priv->child_count);
-
- source = __afr_selfheal_entry_finalize_source (this, sources,
- healed_sinks,
- locked_on, replies,
- witness);
-
- if (source < 0) {
- /* If source is < 0 (typically split-brain), we perform a
- conservative merge of entries rather than erroring out */
- }
- *source_p = source;
-
- return ret;
-}
+ ret = afr_selfheal_unlocked_discover(frame, inode, inode->gfid, replies);
+ if (ret)
+ return ret;
-static int
-afr_selfheal_entry_dirent (call_frame_t *frame, xlator_t *this,
- fd_t *fd, char *name, inode_t *parent_idx_inode,
- xlator_t *subvol, gf_boolean_t full_crawl)
-{
- int ret = 0;
- int source = -1;
- unsigned char *locked_on = NULL;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *healed_sinks = NULL;
- inode_t *inode = NULL;
- struct afr_reply *replies = NULL;
- struct afr_reply *par_replies = NULL;
- afr_private_t *priv = NULL;
- dict_t *xattr = NULL;
-
- priv = this->private;
-
- xattr = dict_new ();
- if (!xattr)
- return -ENOMEM;
- ret = dict_set_int32 (xattr, GF_GFIDLESS_LOOKUP, 1);
- if (ret) {
- dict_unref (xattr);
- return -1;
- }
+ witness = alloca0(sizeof(*witness) * priv->child_count);
+ ret = afr_selfheal_find_direction(frame, this, replies,
+ AFR_ENTRY_TRANSACTION, locked_on, sources,
+ sinks, witness, pflag);
+ if (ret)
+ return ret;
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
- locked_on = alloca0 (priv->child_count);
-
- replies = alloca0 (priv->child_count * sizeof(*replies));
- par_replies = alloca0 (priv->child_count * sizeof(*par_replies));
-
- ret = afr_selfheal_entrylk (frame, this, fd->inode, this->name, NULL,
- locked_on);
- {
- if (ret < AFR_SH_MIN_PARTICIPANTS) {
- gf_msg_debug (this->name, 0, "%s: Skipping "
- "entry self-heal as only %d sub-volumes "
- " could be locked in %s domain",
- uuid_utoa (fd->inode->gfid),
- ret, this->name);
- ret = -ENOTCONN;
- goto unlock;
- }
-
- ret = __afr_selfheal_entry_prepare (frame, this, fd->inode,
- locked_on,
- sources, sinks,
- healed_sinks, par_replies,
- &source, NULL);
- if (ret < 0)
- goto unlock;
-
- inode = afr_selfheal_unlocked_lookup_on (frame, fd->inode, name,
- replies, locked_on,
- xattr);
- if (!inode) {
- ret = -ENOMEM;
- goto unlock;
- }
-
- ret = __afr_selfheal_entry_dirent (frame, this, fd, name, inode,
- source, sources, healed_sinks,
- locked_on, replies);
-
- if ((ret == 0) && (priv->esh_granular) && parent_idx_inode) {
- ret = afr_shd_index_purge (subvol, parent_idx_inode,
- name, inode->ia_type);
- /* Why is ret force-set to 0? We do not care about
- * index purge failing for full heal as it is quite
- * possible during replace-brick that not all files
- * and directories have their name indices present in
- * entry-changes/.
- */
- ret = 0;
- }
- }
+ /* Initialize the healed_sinks[] array optimistically to
+ the intersection of to-be-healed (i.e sinks[]) and
+ the list of servers which are up (i.e locked_on[]).
-unlock:
- afr_selfheal_unentrylk (frame, this, fd->inode, this->name, NULL,
- locked_on, NULL);
- if (inode)
- inode_unref (inode);
- if (replies)
- afr_replies_wipe (replies, priv->child_count);
- if (par_replies)
- afr_replies_wipe (par_replies, priv->child_count);
- if (xattr)
- dict_unref (xattr);
-
- return ret;
-}
+ As we encounter failures in the healing process, we
+ will unmark the respective servers in the healed_sinks[]
+ array.
+ */
+ AFR_INTERSECT(healed_sinks, sinks, locked_on, priv->child_count);
+ source = __afr_selfheal_entry_finalize_source(this, sources, healed_sinks,
+ locked_on, replies, witness);
-static inode_t *
-afr_shd_entry_changes_index_inode (xlator_t *this, xlator_t *subvol,
- uuid_t pargfid)
+ if (source < 0) {
+ /* If source is < 0 (typically split-brain), we perform a
+ conservative merge of entries rather than erroring out */
+ }
+ *source_p = source;
+
+ return ret;
+}
+
+static int
+afr_selfheal_entry_dirent(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ char *name, inode_t *parent_idx_inode,
+ xlator_t *subvol, gf_boolean_t full_crawl)
{
- int ret = -1;
- void *index_gfid = NULL;
- loc_t rootloc = {0,};
- loc_t loc = {0,};
- dict_t *xattr = NULL;
- inode_t *inode = NULL;
- struct iatt iatt = {0,};
-
- rootloc.inode = inode_ref (this->itable->root);
- gf_uuid_copy (rootloc.gfid, rootloc.inode->gfid);
-
- ret = syncop_getxattr (subvol, &rootloc, &xattr,
- GF_XATTROP_ENTRY_CHANGES_GFID, NULL, NULL);
- if (ret || !xattr) {
- errno = -ret;
- goto out;
+ int ret = 0;
+ int source = -1;
+ unsigned char *locked_on = NULL;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ unsigned char *healed_sinks = NULL;
+ inode_t *inode = NULL;
+ struct afr_reply *replies = NULL;
+ struct afr_reply *par_replies = NULL;
+ afr_private_t *priv = NULL;
+ dict_t *xattr = NULL;
+
+ priv = this->private;
+
+ if (afr_is_private_directory(priv, fd->inode->gfid, name,
+ GF_CLIENT_PID_SELF_HEALD)) {
+ return 0;
+ }
+
+ xattr = dict_new();
+ if (!xattr)
+ return -ENOMEM;
+ ret = dict_set_int32_sizen(xattr, GF_GFIDLESS_LOOKUP, 1);
+ if (ret) {
+ dict_unref(xattr);
+ return -1;
+ }
+
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ healed_sinks = alloca0(priv->child_count);
+ locked_on = alloca0(priv->child_count);
+
+ replies = alloca0(priv->child_count * sizeof(*replies));
+ par_replies = alloca0(priv->child_count * sizeof(*par_replies));
+
+ ret = afr_selfheal_entrylk(frame, this, fd->inode, this->name, NULL,
+ locked_on);
+ {
+ if (ret < priv->child_count) {
+ gf_msg_debug(this->name, 0,
+ "%s: Skipping "
+ "entry self-heal as only %d sub-volumes "
+ " could be locked in %s domain",
+ uuid_utoa(fd->inode->gfid), ret, this->name);
+ ret = -ENOTCONN;
+ goto unlock;
}
- ret = dict_get_ptr (xattr, GF_XATTROP_ENTRY_CHANGES_GFID, &index_gfid);
- if (ret) {
- errno = EINVAL;
- goto out;
- }
+ ret = __afr_selfheal_entry_prepare(frame, this, fd->inode, locked_on,
+ sources, sinks, healed_sinks,
+ par_replies, &source, NULL);
+ if (ret < 0)
+ goto unlock;
- loc.inode = inode_new (this->itable);
- if (!loc.inode) {
- errno = ENOMEM;
- goto out;
+ inode = afr_selfheal_unlocked_lookup_on(frame, fd->inode, name, replies,
+ locked_on, xattr);
+ if (!inode) {
+ ret = -ENOMEM;
+ goto unlock;
}
- gf_uuid_copy (loc.pargfid, index_gfid);
- loc.name = gf_strdup (uuid_utoa (pargfid));
-
- ret = syncop_lookup (subvol, &loc, &iatt, NULL, NULL, NULL);
- if (ret < 0) {
- errno = -ret;
- goto out;
+ ret = __afr_selfheal_entry_dirent(frame, this, fd, name, inode, source,
+ sources, healed_sinks, locked_on,
+ replies);
+
+ if ((ret == 0) && (priv->esh_granular) && parent_idx_inode) {
+ ret = afr_shd_entry_purge(subvol, parent_idx_inode, name,
+ inode->ia_type);
+ /* Why is ret force-set to 0? We do not care about
+ * index purge failing for full heal as it is quite
+ * possible during replace-brick that not all files
+ * and directories have their name indices present in
+ * entry-changes/.
+ */
+ ret = 0;
}
+ }
+
+unlock:
+ afr_selfheal_unentrylk(frame, this, fd->inode, this->name, NULL, locked_on,
+ NULL);
+ if (inode)
+ inode_unref(inode);
+ if (replies)
+ afr_replies_wipe(replies, priv->child_count);
+ if (par_replies)
+ afr_replies_wipe(par_replies, priv->child_count);
+ if (xattr)
+ dict_unref(xattr);
+
+ return ret;
+}
- inode = inode_link (loc.inode, NULL, NULL, &iatt);
+static inode_t *
+afr_shd_entry_changes_index_inode(xlator_t *this, xlator_t *subvol,
+ uuid_t pargfid)
+{
+ int ret = -1;
+ void *index_gfid = NULL;
+ loc_t rootloc = {
+ 0,
+ };
+ loc_t loc = {
+ 0,
+ };
+ dict_t *xattr = NULL;
+ inode_t *inode = NULL;
+ struct iatt iatt = {
+ 0,
+ };
+
+ rootloc.inode = inode_ref(this->itable->root);
+ gf_uuid_copy(rootloc.gfid, rootloc.inode->gfid);
+
+ ret = syncop_getxattr(subvol, &rootloc, &xattr,
+ GF_XATTROP_ENTRY_CHANGES_GFID, NULL, NULL);
+ if (ret || !xattr) {
+ errno = -ret;
+ goto out;
+ }
+
+ ret = dict_get_ptr(xattr, GF_XATTROP_ENTRY_CHANGES_GFID, &index_gfid);
+ if (ret) {
+ errno = EINVAL;
+ goto out;
+ }
+
+ loc.inode = inode_new(this->itable);
+ if (!loc.inode) {
+ errno = ENOMEM;
+ goto out;
+ }
+
+ gf_uuid_copy(loc.pargfid, index_gfid);
+ loc.name = gf_strdup(uuid_utoa(pargfid));
+
+ ret = syncop_lookup(subvol, &loc, &iatt, NULL, NULL, NULL);
+ if (ret < 0) {
+ errno = -ret;
+ goto out;
+ }
+
+ inode = inode_link(loc.inode, NULL, NULL, &iatt);
out:
- if (xattr)
- dict_unref (xattr);
- loc_wipe (&rootloc);
- GF_FREE ((char *)loc.name);
- loc_wipe (&loc);
+ if (xattr)
+ dict_unref(xattr);
+ loc_wipe(&rootloc);
+ GF_FREE((char *)loc.name);
+ loc_wipe(&loc);
- return inode;
+ return inode;
}
static int
-afr_selfheal_entry_do_subvol (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int child)
+afr_selfheal_entry_do_subvol(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ int child)
{
- int ret = 0;
- gf_dirent_t entries;
- gf_dirent_t *entry = NULL;
- off_t offset = 0;
- call_frame_t *iter_frame = NULL;
- xlator_t *subvol = NULL;
- afr_private_t *priv = NULL;
- gf_boolean_t mismatch = _gf_false;
- afr_local_t *local = NULL;
- loc_t loc = {0,};
-
- priv = this->private;
- subvol = priv->children[child];
-
- INIT_LIST_HEAD (&entries.list);
-
- local = frame->local;
-
- iter_frame = afr_copy_frame (frame);
- if (!iter_frame)
- return -ENOMEM;
-
- loc.inode = afr_shd_entry_changes_index_inode (this, subvol,
- fd->inode->gfid);
-
- while ((ret = syncop_readdir (subvol, fd, 131072, offset, &entries,
- NULL, NULL))) {
- if (ret > 0)
- ret = 0;
- list_for_each_entry (entry, &entries.list, list) {
- offset = entry->d_off;
-
- if (!strcmp (entry->d_name, ".") ||
- !strcmp (entry->d_name, ".."))
- continue;
-
- if (__is_root_gfid (fd->inode->gfid) &&
- !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR))
- continue;
-
- ret = afr_selfheal_entry_dirent (iter_frame, this, fd,
- entry->d_name,
- loc.inode, subvol,
- local->need_full_crawl);
- AFR_STACK_RESET (iter_frame);
- if (iter_frame->local == NULL) {
- ret = -ENOTCONN;
- break;
- }
-
- if (ret == -1) {
- /* gfid or type mismatch. */
- mismatch = _gf_true;
- ret = 0;
- }
- if (ret)
- break;
- }
-
- gf_dirent_free (&entries);
- if (ret)
- break;
- }
-
- loc_wipe (&loc);
-
- AFR_STACK_DESTROY (iter_frame);
- if (mismatch == _gf_true)
- /* undo pending will be skipped */
- ret = -1;
- return ret;
-}
+ int ret = 0;
+ gf_dirent_t entries;
+ gf_dirent_t *entry = NULL;
+ off_t offset = 0;
+ call_frame_t *iter_frame = NULL;
+ xlator_t *subvol = NULL;
+ afr_private_t *priv = NULL;
+ gf_boolean_t mismatch = _gf_false;
+ afr_local_t *local = NULL;
+ loc_t loc = {
+ 0,
+ };
+
+ priv = this->private;
+ subvol = priv->children[child];
+
+ INIT_LIST_HEAD(&entries.list);
+
+ local = frame->local;
+
+ iter_frame = afr_copy_frame(frame);
+ if (!iter_frame)
+ return -ENOMEM;
+
+ loc.inode = afr_shd_entry_changes_index_inode(this, subvol,
+ fd->inode->gfid);
+
+ while ((ret = syncop_readdir(subvol, fd, 131072, offset, &entries, NULL,
+ NULL))) {
+ if (ret > 0)
+ ret = 0;
+ list_for_each_entry(entry, &entries.list, list)
+ {
+ offset = entry->d_off;
-static int
-afr_selfheal_entry_granular_dirent (xlator_t *subvol, gf_dirent_t *entry,
- loc_t *parent, void *data)
-{
- int ret = 0;
- loc_t loc = {0,};
- struct iatt iatt = {0,};
- afr_granular_esh_args_t *args = data;
-
- /* Look up the actual inode associated with entry. If the lookup returns
- * ESTALE or ENOENT, then it means we have a stale index. Remove it.
- * This is analogous to the check in afr_shd_index_heal() except that
- * here it is achieved through LOOKUP and in afr_shd_index_heal() through
- * a GETXATTR.
- */
+ if (!strcmp(entry->d_name, ".") || !strcmp(entry->d_name, ".."))
+ continue;
+
+ ret = afr_selfheal_entry_dirent(iter_frame, this, fd, entry->d_name,
+ loc.inode, subvol,
+ local->need_full_crawl);
+ AFR_STACK_RESET(iter_frame);
+ if (iter_frame->local == NULL) {
+ ret = -ENOTCONN;
+ break;
+ }
- loc.inode = inode_new (args->xl->itable);
- loc.parent = inode_ref (args->heal_fd->inode);
- gf_uuid_copy (loc.pargfid, loc.parent->gfid);
- loc.name = entry->d_name;
-
- ret = syncop_lookup (args->xl, &loc, &iatt, NULL, NULL, NULL);
- if ((ret == -ENOENT) || (ret == -ESTALE)) {
- /* The name indices under the pgfid index dir are guaranteed
- * to be regular files. Hence the hardcoding.
- */
- afr_shd_index_purge (subvol, parent->inode, entry->d_name,
- IA_IFREG);
+ if (ret == -1) {
+ /* gfid or type mismatch. */
+ mismatch = _gf_true;
ret = 0;
- goto out;
+ }
+ if (ret)
+ break;
}
- /* TBD: afr_shd_zero_xattrop? */
- ret = afr_selfheal_entry_dirent (args->frame, args->xl, args->heal_fd,
- entry->d_name, parent->inode, subvol,
- _gf_false);
- AFR_STACK_RESET (args->frame);
- if (args->frame->local == NULL)
- ret = -ENOTCONN;
+ gf_dirent_free(&entries);
+ if (ret)
+ break;
+ }
- if (ret == -1)
- args->mismatch = _gf_true;
+ loc_wipe(&loc);
-out:
- loc_wipe (&loc);
- return 0;
+ AFR_STACK_DESTROY(iter_frame);
+ if (mismatch == _gf_true)
+ /* undo pending will be skipped */
+ ret = -1;
+ return ret;
}
static int
-afr_selfheal_entry_granular (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int subvol_idx, gf_boolean_t is_src)
+afr_selfheal_entry_granular_dirent(xlator_t *subvol, gf_dirent_t *entry,
+ loc_t *parent, void *data)
{
- int ret = 0;
- loc_t loc = {0,};
- xlator_t *subvol = NULL;
- afr_private_t *priv = NULL;
- afr_granular_esh_args_t args = {0,};
-
- priv = this->private;
- subvol = priv->children[subvol_idx];
-
- args.frame = afr_copy_frame (frame);
- args.xl = this;
- /* args.heal_fd represents the fd associated with the original directory
- * on which entry heal is being attempted.
+ int ret = 0;
+ loc_t loc = {
+ 0,
+ };
+ struct iatt iatt = {
+ 0,
+ };
+ afr_granular_esh_args_t *args = data;
+
+ /* Look up the actual inode associated with entry. If the lookup returns
+ * ESTALE or ENOENT, then it means we have a stale index. Remove it.
+ * This is analogous to the check in afr_shd_index_heal() except that
+ * here it is achieved through LOOKUP and in afr_shd_index_heal() through
+ * a GETXATTR.
+ */
+
+ loc.inode = inode_new(args->xl->itable);
+ loc.parent = inode_ref(args->heal_fd->inode);
+ gf_uuid_copy(loc.pargfid, loc.parent->gfid);
+ loc.name = entry->d_name;
+
+ ret = syncop_lookup(args->xl, &loc, &iatt, NULL, NULL, NULL);
+ if ((ret == -ENOENT) || (ret == -ESTALE)) {
+ /* The name indices under the pgfid index dir are guaranteed
+ * to be regular files. Hence the hardcoding.
*/
- args.heal_fd = fd;
+ afr_shd_entry_purge(subvol, parent->inode, entry->d_name, IA_IFREG);
+ ret = 0;
+ goto out;
+ }
+ /* TBD: afr_shd_zero_xattrop? */
+
+ ret = afr_selfheal_entry_dirent(args->frame, args->xl, args->heal_fd,
+ entry->d_name, parent->inode, subvol,
+ _gf_false);
+ AFR_STACK_RESET(args->frame);
+ if (args->frame->local == NULL)
+ ret = -ENOTCONN;
+
+ if (ret == -1)
+ args->mismatch = _gf_true;
- /* @subvol here represents the subvolume of AFR where
- * indices/entry-changes/<pargfid> will be processed
- */
- loc.inode = afr_shd_entry_changes_index_inode (this, subvol,
- fd->inode->gfid);
- if (!loc.inode) {
- /* If granular heal failed on the sink (as it might sometimes
- * because it is the src that would mostly contain the granular
- * changelogs and the sink's entry-changes would be empty),
- * do not treat heal as failure.
- */
- if (is_src)
- return -errno;
- else
- return 0;
- }
+out:
+ loc_wipe(&loc);
+ return 0;
+}
- ret = syncop_dir_scan (subvol, &loc, GF_CLIENT_PID_SELF_HEALD,
- &args, afr_selfheal_entry_granular_dirent);
+static int
+afr_selfheal_entry_granular(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ int subvol_idx, gf_boolean_t is_src)
+{
+ int ret = 0;
+ loc_t loc = {
+ 0,
+ };
+ xlator_t *subvol = NULL;
+ afr_private_t *priv = NULL;
+ afr_granular_esh_args_t args = {
+ 0,
+ };
+
+ priv = this->private;
+ subvol = priv->children[subvol_idx];
+
+ args.frame = afr_copy_frame(frame);
+ if (!args.frame)
+ goto out;
+ args.xl = this;
+ /* args.heal_fd represents the fd associated with the original directory
+ * on which entry heal is being attempted.
+ */
+ args.heal_fd = fd;
+
+ /* @subvol here represents the subvolume of AFR where
+ * indices/entry-changes/<pargfid> will be processed
+ */
+ loc.inode = afr_shd_entry_changes_index_inode(this, subvol,
+ fd->inode->gfid);
+ if (!loc.inode) {
+ /* If granular heal failed on the sink (as it might sometimes
+ * because it is the src that would mostly contain the granular
+ * changelogs and the sink's entry-changes would be empty),
+ * do not treat heal as failure.
+ */
+ if (is_src)
+ ret = -errno;
+ else
+ ret = 0;
+ goto out;
+ }
- loc_wipe (&loc);
+ ret = syncop_dir_scan(subvol, &loc, GF_CLIENT_PID_SELF_HEALD, &args,
+ afr_selfheal_entry_granular_dirent);
- if (args.mismatch == _gf_true)
- ret = -1;
+ loc_wipe(&loc);
- return ret;
+ if (args.mismatch == _gf_true)
+ ret = -1;
+out:
+ if (args.frame)
+ AFR_STACK_DESTROY(args.frame);
+ return ret;
}
static int
-afr_selfheal_entry_do (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int source, unsigned char *sources,
- unsigned char *healed_sinks)
+afr_selfheal_entry_do(call_frame_t *frame, xlator_t *this, fd_t *fd, int source,
+ unsigned char *sources, unsigned char *healed_sinks)
{
- int i = 0;
- int ret = 0;
- gf_boolean_t mismatch = _gf_false;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
-
- priv = this->private;
- local = frame->local;
-
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_SELF_HEAL_INFO, "performing entry selfheal on %s",
- uuid_utoa (fd->inode->gfid));
-
- for (i = 0; i < priv->child_count; i++) {
- /* Expunge */
- if (!healed_sinks[i])
- continue;
-
- if (!local->need_full_crawl)
- /* Why call afr_selfheal_entry_granular() on a "healed sink",
- * given that it is the source that contains the granular
- * indices?
- * If the index for this directory is non-existent or empty on
- * this subvol (=> clear sink), then it will return early
- * without failure status.
- * If the index is non-empty and it is yet a 'healed sink', then
- * it is due to a split-brain in which case we anyway need to
- * crawl the indices/entry-changes/pargfid directory.
- */
- ret = afr_selfheal_entry_granular (frame, this, fd, i,
- _gf_false);
- else
- ret = afr_selfheal_entry_do_subvol (frame, this, fd, i);
-
- if (ret == -1) {
- /* gfid or type mismatch. */
- mismatch = _gf_true;
- ret = 0;
- }
- if (ret)
- break;
- }
-
- if (!ret && source != -1) {
- /* Impunge */
- if (local->need_full_crawl)
- ret = afr_selfheal_entry_do_subvol (frame, this, fd,
- source);
- else
- ret = afr_selfheal_entry_granular (frame, this, fd,
- source, _gf_true);
+ int i = 0;
+ int ret = 0;
+ gf_boolean_t mismatch = _gf_false;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_SELF_HEAL_INFO,
+ "performing entry selfheal on %s", uuid_utoa(fd->inode->gfid));
+
+ for (i = 0; i < priv->child_count; i++) {
+ /* Expunge */
+ if (!healed_sinks[i])
+ continue;
+
+ if (!local->need_full_crawl)
+ /* Why call afr_selfheal_entry_granular() on a "healed sink",
+ * given that it is the source that contains the granular
+ * indices?
+ * If the index for this directory is non-existent or empty on
+ * this subvol (=> clear sink), then it will return early
+ * without failure status.
+ * If the index is non-empty and it is yet a 'healed sink', then
+ * it is due to a split-brain in which case we anyway need to
+ * crawl the indices/entry-changes/pargfid directory.
+ */
+ ret = afr_selfheal_entry_granular(frame, this, fd, i, _gf_false);
+ else
+ ret = afr_selfheal_entry_do_subvol(frame, this, fd, i);
+
+ if (ret == -1) {
+ /* gfid or type mismatch. */
+ mismatch = _gf_true;
+ ret = 0;
}
+ if (ret)
+ break;
+ }
- if (mismatch == _gf_true)
- /* undo pending will be skipped */
- ret = -1;
- return ret;
+ if (!ret && source != -1) {
+ /* Impunge */
+ if (local->need_full_crawl)
+ ret = afr_selfheal_entry_do_subvol(frame, this, fd, source);
+ else
+ ret = afr_selfheal_entry_granular(frame, this, fd, source,
+ _gf_true);
+ }
+
+ if (mismatch == _gf_true)
+ /* undo pending will be skipped */
+ ret = -1;
+ return ret;
}
static int
-__afr_selfheal_entry (call_frame_t *frame, xlator_t *this, fd_t *fd,
- unsigned char *locked_on)
+__afr_selfheal_entry(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ unsigned char *locked_on)
{
- int ret = -1;
- int source = -1;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *data_lock = NULL;
- unsigned char *postop_lock = NULL;
- unsigned char *healed_sinks = NULL;
- unsigned char *undid_pending = NULL;
- struct afr_reply *locked_replies = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- gf_boolean_t did_sh = _gf_true;
-
- priv = this->private;
- local = frame->local;
-
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
- undid_pending = alloca0 (priv->child_count);
- data_lock = alloca0 (priv->child_count);
- postop_lock = alloca0 (priv->child_count);
-
- locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
-
- ret = afr_selfheal_entrylk (frame, this, fd->inode, this->name, NULL,
- data_lock);
- {
- if (ret < AFR_SH_MIN_PARTICIPANTS) {
- gf_msg_debug (this->name, 0, "%s: Skipping "
- "entry self-heal as only %d sub-volumes could "
- "be locked in %s domain",
- uuid_utoa (fd->inode->gfid), ret,
- this->name);
- ret = -ENOTCONN;
- goto unlock;
- }
-
- ret = __afr_selfheal_entry_prepare (frame, this, fd->inode,
- data_lock, sources, sinks,
- healed_sinks,
- locked_replies, &source,
- NULL);
- if (AFR_COUNT(healed_sinks, priv->child_count) == 0) {
- did_sh = _gf_false;
- goto unlock;
- }
-
- local->need_full_crawl = afr_need_full_heal (this,
- locked_replies,
- source,
- healed_sinks,
- AFR_ENTRY_TRANSACTION);
- }
-unlock:
- afr_selfheal_unentrylk (frame, this, fd->inode, this->name, NULL,
- data_lock, NULL);
- if (ret < 0)
- goto out;
+ int ret = -1;
+ int source = -1;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ unsigned char *data_lock = NULL;
+ unsigned char *postop_lock = NULL;
+ unsigned char *healed_sinks = NULL;
+ unsigned char *undid_pending = NULL;
+ struct afr_reply *locked_replies = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ gf_boolean_t did_sh = _gf_true;
+
+ priv = this->private;
+ local = frame->local;
+
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ healed_sinks = alloca0(priv->child_count);
+ undid_pending = alloca0(priv->child_count);
+ data_lock = alloca0(priv->child_count);
+ postop_lock = alloca0(priv->child_count);
+
+ locked_replies = alloca0(sizeof(*locked_replies) * priv->child_count);
+
+ ret = afr_selfheal_entrylk(frame, this, fd->inode, this->name, NULL,
+ data_lock);
+ {
+ if (ret < priv->child_count) {
+ gf_msg_debug(this->name, 0,
+ "%s: Skipping "
+ "entry self-heal as only %d sub-volumes could "
+ "be locked in %s domain",
+ uuid_utoa(fd->inode->gfid), ret, this->name);
+ ret = -ENOTCONN;
+ goto unlock;
+ }
- if (!did_sh)
- goto out;
+ ret = __afr_selfheal_entry_prepare(frame, this, fd->inode, data_lock,
+ sources, sinks, healed_sinks,
+ locked_replies, &source, NULL);
+ if (AFR_COUNT(healed_sinks, priv->child_count) == 0) {
+ did_sh = _gf_false;
+ goto unlock;
+ }
- ret = afr_selfheal_entry_do (frame, this, fd, source, sources,
- healed_sinks);
- if (ret)
- goto out;
-
- /* Take entrylks in xlator domain before doing post-op (undo-pending) in
- * entry self-heal. This is to prevent a parallel name self-heal on
- * an entry under @fd->inode from reading pending xattrs while it is
- * being modified by SHD after entry sh below, given that
- * name self-heal takes locks ONLY in xlator domain and is free to read
- * pending changelog in the absence of the following locking.
- */
- ret = afr_selfheal_entrylk (frame, this, fd->inode, this->name, NULL,
- postop_lock);
- {
- if (AFR_CMP (data_lock, postop_lock, priv->child_count) != 0) {
- gf_msg_debug (this->name, 0, "%s: Skipping "
- "post-op after entry self-heal as %d "
- "sub-volumes, as opposed to %d, "
- "could be locked in %s domain",
- uuid_utoa (fd->inode->gfid),
- ret, AFR_COUNT (data_lock,
- priv->child_count), this->name);
- ret = -ENOTCONN;
- goto postop_unlock;
- }
-
- ret = afr_selfheal_undo_pending (frame, this, fd->inode,
- sources, sinks, healed_sinks,
- undid_pending,
- AFR_ENTRY_TRANSACTION,
- locked_replies, postop_lock);
+ local->need_full_crawl = afr_need_full_heal(
+ this, locked_replies, source, healed_sinks, AFR_ENTRY_TRANSACTION);
+ }
+unlock:
+ afr_selfheal_unentrylk(frame, this, fd->inode, this->name, NULL, data_lock,
+ NULL);
+ if (ret < 0)
+ goto out;
+
+ if (!did_sh)
+ goto out;
+
+ ret = afr_selfheal_entry_do(frame, this, fd, source, sources, healed_sinks);
+ if (ret)
+ goto out;
+
+ /* Take entrylks in xlator domain before doing post-op (undo-pending) in
+ * entry self-heal. This is to prevent a parallel name self-heal on
+ * an entry under @fd->inode from reading pending xattrs while it is
+ * being modified by SHD after entry sh below, given that
+ * name self-heal takes locks ONLY in xlator domain and is free to read
+ * pending changelog in the absence of the following locking.
+ */
+ ret = afr_selfheal_entrylk(frame, this, fd->inode, this->name, NULL,
+ postop_lock);
+ {
+ if (AFR_CMP(data_lock, postop_lock, priv->child_count) != 0) {
+ gf_msg_debug(this->name, 0,
+ "%s: Skipping "
+ "post-op after entry self-heal as %d "
+ "sub-volumes, as opposed to %d, "
+ "could be locked in %s domain",
+ uuid_utoa(fd->inode->gfid), ret,
+ AFR_COUNT(data_lock, priv->child_count), this->name);
+ ret = -ENOTCONN;
+ goto postop_unlock;
}
+
+ afr_selfheal_restore_time(frame, this, fd->inode, source, healed_sinks,
+ locked_replies);
+ ret = afr_selfheal_undo_pending(
+ frame, this, fd->inode, sources, sinks, healed_sinks, undid_pending,
+ AFR_ENTRY_TRANSACTION, locked_replies, postop_lock);
+ }
postop_unlock:
- afr_selfheal_unentrylk (frame, this, fd->inode, this->name, NULL,
- postop_lock, NULL);
+ afr_selfheal_unentrylk(frame, this, fd->inode, this->name, NULL,
+ postop_lock, NULL);
out:
- if (did_sh)
- afr_log_selfheal (fd->inode->gfid, this, ret, "entry", source,
- sources, healed_sinks);
- else
- ret = 1;
-
- if (locked_replies)
- afr_replies_wipe (locked_replies, priv->child_count);
- return ret;
+ if (did_sh)
+ afr_log_selfheal(fd->inode->gfid, this, ret, "entry", source, sources,
+ healed_sinks);
+ else
+ ret = 1;
+
+ if (locked_replies)
+ afr_replies_wipe(locked_replies, priv->child_count);
+ return ret;
}
-
static fd_t *
-afr_selfheal_data_opendir (xlator_t *this, inode_t *inode)
+afr_selfheal_data_opendir(xlator_t *this, inode_t *inode)
{
- loc_t loc = {0,};
- int ret = 0;
- fd_t *fd = NULL;
-
- fd = fd_create (inode, 0);
- if (!fd)
- return NULL;
-
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
-
- ret = syncop_opendir (this, &loc, fd, NULL, NULL);
- if (ret) {
- fd_unref (fd);
- fd = NULL;
- } else {
- fd_bind (fd);
- }
-
- loc_wipe (&loc);
- return fd;
+ loc_t loc = {
+ 0,
+ };
+ int ret = 0;
+ fd_t *fd = NULL;
+
+ fd = fd_create(inode, 0);
+ if (!fd)
+ return NULL;
+
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+
+ ret = syncop_opendir(this, &loc, fd, NULL, NULL);
+ if (ret) {
+ fd_unref(fd);
+ fd = NULL;
+ } else {
+ fd_bind(fd);
+ }
+
+ loc_wipe(&loc);
+ return fd;
}
-
int
-afr_selfheal_entry (call_frame_t *frame, xlator_t *this, inode_t *inode)
+afr_selfheal_entry(call_frame_t *frame, xlator_t *this, inode_t *inode)
{
- afr_private_t *priv = NULL;
- unsigned char *locked_on = NULL;
- fd_t *fd = NULL;
- int ret = 0;
-
- priv = this->private;
-
- fd = afr_selfheal_data_opendir (this, inode);
- if (!fd)
- return -EIO;
-
- locked_on = alloca0 (priv->child_count);
-
- ret = afr_selfheal_tie_breaker_entrylk (frame, this, inode,
- priv->sh_domain, NULL,
- locked_on);
- {
- if (ret < AFR_SH_MIN_PARTICIPANTS) {
- gf_msg_debug (this->name, 0, "%s: Skipping "
- "entry self-heal as only %d sub-volumes could "
- "be locked in %s domain",
- uuid_utoa (fd->inode->gfid), ret,
- priv->sh_domain);
- /* Either less than two subvols available, or another
- selfheal (from another server) is in progress. Skip
- for now in any case there isn't anything to do.
- */
- ret = -ENOTCONN;
- goto unlock;
- }
-
- ret = __afr_selfheal_entry (frame, this, fd, locked_on);
- }
+ afr_private_t *priv = NULL;
+ unsigned char *locked_on = NULL;
+ fd_t *fd = NULL;
+ int ret = 0;
+
+ priv = this->private;
+
+ fd = afr_selfheal_data_opendir(this, inode);
+ if (!fd)
+ return -EIO;
+
+ locked_on = alloca0(priv->child_count);
+
+ ret = afr_selfheal_tie_breaker_entrylk(frame, this, inode, priv->sh_domain,
+ NULL, locked_on);
+ {
+ if (ret < priv->child_count) {
+ gf_msg_debug(this->name, 0,
+ "%s: Skipping "
+ "entry self-heal as only %d sub-volumes could "
+ "be locked in %s domain",
+ uuid_utoa(fd->inode->gfid), ret, priv->sh_domain);
+ /* Either less than two subvols available, or another
+ selfheal (from another server) is in progress. Skip
+ for now in any case there isn't anything to do.
+ */
+ ret = -ENOTCONN;
+ goto unlock;
+ }
+
+ ret = __afr_selfheal_entry(frame, this, fd, locked_on);
+ }
unlock:
- afr_selfheal_unentrylk (frame, this, inode, priv->sh_domain, NULL,
- locked_on, NULL);
+ afr_selfheal_unentrylk(frame, this, inode, priv->sh_domain, NULL, locked_on,
+ NULL);
- if (fd)
- fd_unref (fd);
+ if (fd)
+ fd_unref(fd);
- return ret;
+ return ret;
}
diff --git a/xlators/cluster/afr/src/afr-self-heal-metadata.c b/xlators/cluster/afr/src/afr-self-heal-metadata.c
index 199f8961480..03f43bad16e 100644
--- a/xlators/cluster/afr/src/afr-self-heal-metadata.c
+++ b/xlators/cluster/afr/src/afr-self-heal-metadata.c
@@ -8,109 +8,108 @@
cases as published by the Free Software Foundation.
*/
-
#include "afr.h"
#include "afr-self-heal.h"
-#include "byte-order.h"
+#include <glusterfs/byte-order.h>
#include "protocol-common.h"
-#include "events.h"
+#include <glusterfs/events.h>
-#define AFR_HEAL_ATTR (GF_SET_ATTR_UID|GF_SET_ATTR_GID|GF_SET_ATTR_MODE)
+#define AFR_HEAL_ATTR (GF_SET_ATTR_UID | GF_SET_ATTR_GID | GF_SET_ATTR_MODE)
static gf_boolean_t
-_afr_ignorable_key_match (dict_t *d, char *k, data_t *val, void *mdata)
+_afr_ignorable_key_match(dict_t *d, char *k, data_t *val, void *mdata)
{
- return afr_is_xattr_ignorable (k);
+ return afr_is_xattr_ignorable(k);
}
void
-afr_delete_ignorable_xattrs (dict_t *xattr)
+afr_delete_ignorable_xattrs(dict_t *xattr)
{
- dict_foreach_match (xattr, _afr_ignorable_key_match, NULL,
- dict_remove_foreach_fn, NULL);
+ dict_foreach_match(xattr, _afr_ignorable_key_match, NULL,
+ dict_remove_foreach_fn, NULL);
}
int
-__afr_selfheal_metadata_do (call_frame_t *frame, xlator_t *this, inode_t *inode,
- int source, unsigned char *healed_sinks,
- struct afr_reply *locked_replies)
+__afr_selfheal_metadata_do(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ int source, unsigned char *healed_sinks,
+ struct afr_reply *locked_replies)
{
- int ret = -1;
- loc_t loc = {0,};
- dict_t *xattr = NULL;
- dict_t *old_xattr = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
-
- priv = this->private;
-
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
-
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_SELF_HEAL_INFO, "performing metadata selfheal on %s",
- uuid_utoa (inode->gfid));
-
- ret = syncop_getxattr (priv->children[source], &loc, &xattr, NULL,
- NULL, NULL);
- if (ret < 0) {
- ret = -EIO;
- goto out;
- }
-
- afr_delete_ignorable_xattrs (xattr);
-
- for (i = 0; i < priv->child_count; i++) {
- if (old_xattr) {
- dict_unref (old_xattr);
- old_xattr = NULL;
- }
-
- if (!healed_sinks[i])
- continue;
-
- ret = syncop_setattr (priv->children[i], &loc,
- &locked_replies[source].poststat,
- AFR_HEAL_ATTR, NULL, NULL, NULL, NULL);
- if (ret)
- healed_sinks[i] = 0;
-
- ret = syncop_getxattr (priv->children[i], &loc, &old_xattr, 0,
- NULL, NULL);
- if (old_xattr) {
- afr_delete_ignorable_xattrs (old_xattr);
- ret = syncop_removexattr (priv->children[i], &loc, "",
- old_xattr, NULL);
- if (ret)
- healed_sinks[i] = 0;
- }
-
- ret = syncop_setxattr (priv->children[i], &loc, xattr, 0, NULL,
- NULL);
- if (ret)
- healed_sinks[i] = 0;
- }
- ret = 0;
+ int ret = -1;
+ loc_t loc = {
+ 0,
+ };
+ dict_t *xattr = NULL;
+ dict_t *old_xattr = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+
+ priv = this->private;
+
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_SELF_HEAL_INFO,
+ "performing metadata selfheal on %s", uuid_utoa(inode->gfid));
+
+ ret = syncop_getxattr(priv->children[source], &loc, &xattr, NULL, NULL,
+ NULL);
+ if (ret < 0) {
+ ret = -EIO;
+ goto out;
+ }
+
+ afr_delete_ignorable_xattrs(xattr);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (old_xattr) {
+ dict_unref(old_xattr);
+ old_xattr = NULL;
+ }
+
+ if (!healed_sinks[i])
+ continue;
+
+ ret = syncop_setattr(priv->children[i], &loc,
+ &locked_replies[source].poststat, AFR_HEAL_ATTR,
+ NULL, NULL, NULL, NULL);
+ if (ret)
+ healed_sinks[i] = 0;
+
+ ret = syncop_getxattr(priv->children[i], &loc, &old_xattr, 0, NULL,
+ NULL);
+ if (old_xattr) {
+ afr_delete_ignorable_xattrs(old_xattr);
+ ret = syncop_removexattr(priv->children[i], &loc, "", old_xattr,
+ NULL);
+ if (ret)
+ healed_sinks[i] = 0;
+ }
+
+ ret = syncop_setxattr(priv->children[i], &loc, xattr, 0, NULL, NULL);
+ if (ret)
+ healed_sinks[i] = 0;
+ }
+ ret = 0;
out:
- loc_wipe (&loc);
- if (xattr)
- dict_unref (xattr);
- if (old_xattr)
- dict_unref (old_xattr);
+ loc_wipe(&loc);
+ if (xattr)
+ dict_unref(xattr);
+ if (old_xattr)
+ dict_unref(old_xattr);
- return ret;
+ return ret;
}
static uint64_t
mtime_ns(struct iatt *ia)
{
- uint64_t ret;
+ uint64_t ret;
- ret = (((uint64_t)(ia->ia_mtime)) * 1000000000)
- + (uint64_t)(ia->ia_mtime_nsec);
+ ret = (((uint64_t)(ia->ia_mtime)) * 1000000000) +
+ (uint64_t)(ia->ia_mtime_nsec);
- return ret;
+ return ret;
}
/*
@@ -123,382 +122,425 @@ mtime_ns(struct iatt *ia)
* the source with the most recent modification date.
*/
static int
-afr_dirtime_splitbrain_source (call_frame_t *frame, xlator_t *this,
- struct afr_reply *replies,
- unsigned char *locked_on)
+afr_dirtime_splitbrain_source(call_frame_t *frame, xlator_t *this,
+ struct afr_reply *replies,
+ unsigned char *locked_on)
{
- afr_private_t *priv = NULL;
- int source = -1;
- struct iatt source_ia;
- struct iatt child_ia;
- uint64_t mtime = 0;
- int i;
- int ret = -1;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!locked_on[i])
- continue;
-
- if (!replies[i].valid)
- continue;
-
- if (replies[i].op_ret != 0)
- continue;
-
- if (mtime_ns(&replies[i].poststat) <= mtime)
- continue;
+ afr_private_t *priv = NULL;
+ int source = -1;
+ struct iatt source_ia;
+ struct iatt child_ia;
+ uint64_t mtime = 0;
+ int i;
+ int ret = -1;
+
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!locked_on[i])
+ continue;
+
+ if (!replies[i].valid)
+ continue;
+
+ if (replies[i].op_ret != 0)
+ continue;
+
+ if (mtime_ns(&replies[i].poststat) <= mtime)
+ continue;
+
+ mtime = mtime_ns(&replies[i].poststat);
+ source = i;
+ }
+
+ if (source == -1)
+ goto out;
+
+ source_ia = replies[source].poststat;
+ if (source_ia.ia_type != IA_IFDIR)
+ goto out;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == source)
+ continue;
+
+ if (!replies[i].valid)
+ continue;
+
+ if (replies[i].op_ret != 0)
+ continue;
+
+ child_ia = replies[i].poststat;
+
+ if (!IA_EQUAL(source_ia, child_ia, gfid) ||
+ !IA_EQUAL(source_ia, child_ia, type) ||
+ !IA_EQUAL(source_ia, child_ia, prot) ||
+ !IA_EQUAL(source_ia, child_ia, uid) ||
+ !IA_EQUAL(source_ia, child_ia, gid) ||
+ !afr_xattrs_are_equal(replies[source].xdata, replies[i].xdata))
+ goto out;
+ }
+
+ /*
+ * Metadata split brain is just about [amc]time
+ * We return our source.
+ */
+ ret = source;
+out:
+ return ret;
+}
- mtime = mtime_ns(&replies[i].poststat);
- source = i;
+static int
+__afr_selfheal_metadata_mark_pending_xattrs(call_frame_t *frame, xlator_t *this,
+ inode_t *inode,
+ struct afr_reply *replies,
+ unsigned char *sources)
+{
+ int ret = 0;
+ int i = 0;
+ int m_idx = 0;
+ afr_private_t *priv = NULL;
+ int raw[AFR_NUM_CHANGE_LOGS] = {0};
+ dict_t *xattr = NULL;
+
+ priv = this->private;
+ m_idx = afr_index_for_transaction_type(AFR_METADATA_TRANSACTION);
+ raw[m_idx] = 1;
+
+ xattr = dict_new();
+ if (!xattr)
+ return -ENOMEM;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i])
+ continue;
+ ret = dict_set_static_bin(xattr, priv->pending_key[i], raw,
+ sizeof(int) * AFR_NUM_CHANGE_LOGS);
+ if (ret) {
+ ret = -1;
+ goto out;
}
-
- if (source == -1)
- goto out;
-
- source_ia = replies[source].poststat;
- if (source_ia.ia_type != IA_IFDIR)
- goto out;
-
- for (i = 0; i < priv->child_count; i++) {
- if (i == source)
- continue;
-
- if (!replies[i].valid)
- continue;
-
- if (replies[i].op_ret != 0)
- continue;
-
- child_ia = replies[i].poststat;
-
- if (!IA_EQUAL(source_ia, child_ia, gfid) ||
- !IA_EQUAL(source_ia, child_ia, type) ||
- !IA_EQUAL(source_ia, child_ia, prot) ||
- !IA_EQUAL(source_ia, child_ia, uid) ||
- !IA_EQUAL(source_ia, child_ia, gid) ||
- !afr_xattrs_are_equal (replies[source].xdata,
- replies[i].xdata))
- goto out;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
+ ret = afr_selfheal_post_op(frame, this, inode, i, xattr, NULL);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_INFO, -ret, AFR_MSG_SELF_HEAL_INFO,
+ "Failed to set pending metadata xattr on child %d for %s", i,
+ uuid_utoa(inode->gfid));
+ goto out;
}
+ }
+
+ afr_replies_wipe(replies, priv->child_count);
+ ret = afr_selfheal_unlocked_discover(frame, inode, inode->gfid, replies);
- /*
- * Metadata split brain is just about [amc]time
- * We return our source.
- */
- ret = source;
out:
- return ret;
+ if (xattr)
+ dict_unref(xattr);
+ return ret;
}
-
/*
* Look for mismatching uid/gid or mode or user xattrs even if
* AFR xattrs don't say so, and pick one arbitrarily as winner. */
static int
-__afr_selfheal_metadata_finalize_source (call_frame_t *frame, xlator_t *this,
- inode_t *inode,
- unsigned char *sources,
- unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *undid_pending,
- unsigned char *locked_on,
- struct afr_reply *replies)
+__afr_selfheal_metadata_finalize_source(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, unsigned char *sources,
+ unsigned char *sinks,
+ unsigned char *healed_sinks,
+ unsigned char *undid_pending,
+ unsigned char *locked_on,
+ struct afr_reply *replies)
{
- int i = 0;
- afr_private_t *priv = NULL;
- struct iatt srcstat = {0, };
- int source = -1;
- int sources_count = 0;
-
- priv = this->private;
-
- sources_count = AFR_COUNT (sources, priv->child_count);
-
- if ((AFR_CMP (locked_on, healed_sinks, priv->child_count) == 0)
- || !sources_count) {
-
- source = afr_mark_split_brain_source_sinks (frame, this, inode,
- sources, sinks,
- healed_sinks,
- locked_on, replies,
- AFR_METADATA_TRANSACTION);
- if (source >= 0) {
- _afr_fav_child_reset_sink_xattrs (frame, this, inode,
- source, healed_sinks,
- undid_pending,
- AFR_METADATA_TRANSACTION,
- locked_on, replies);
- goto out;
- }
-
- /* If this is a directory mtime/ctime only split brain
- use the most recent */
- source = afr_dirtime_splitbrain_source (frame, this,
- replies, locked_on);
- if (source != -1) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_SPLIT_BRAIN, "clear time "
- "split brain on %s",
- uuid_utoa (replies[source].poststat.ia_gfid));
- sources[source] = 1;
- healed_sinks[source] = 0;
- goto out;
- }
-
- if (!priv->metadata_splitbrain_forced_heal) {
- gf_event (EVENT_AFR_SPLIT_BRAIN, "subvol=%s;"
- "type=metadata;file=%s",
- this->name, uuid_utoa(inode->gfid));
- return -EIO;
- }
-
- /* Metadata split brain, select one subvol
- arbitrarily */
- for (i = 0; i < priv->child_count; i++) {
- if (locked_on[i] && healed_sinks[i]) {
- sources[i] = 1;
- healed_sinks[i] = 0;
- break;
- }
- }
- }
-
- /* No split brain at this point. If we were called from
- * afr_heal_splitbrain_file(), abort.*/
- if (afr_dict_contains_heal_op(frame))
- return -EIO;
-
- source = afr_choose_source_by_policy (priv, sources,
- AFR_METADATA_TRANSACTION);
- srcstat = replies[source].poststat;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i] || i == source)
- continue;
- if (!IA_EQUAL (srcstat, replies[i].poststat, type) ||
- !IA_EQUAL (srcstat, replies[i].poststat, uid) ||
- !IA_EQUAL (srcstat, replies[i].poststat, gid) ||
- !IA_EQUAL (srcstat, replies[i].poststat, prot)) {
- gf_msg_debug (this->name, 0, "%s: iatt mismatch "
- "for source(%d) vs (%d)",
- uuid_utoa
- (replies[source].poststat.ia_gfid),
- source, i);
- sources[i] = 0;
- healed_sinks[i] = 1;
- }
- }
-
- for (i =0; i < priv->child_count; i++) {
- if (!sources[i] || i == source)
- continue;
- if (!afr_xattrs_are_equal (replies[source].xdata,
- replies[i].xdata)) {
- gf_msg_debug (this->name, 0, "%s: xattr mismatch "
- "for source(%d) vs (%d)",
- uuid_utoa
- (replies[source].poststat.ia_gfid),
- source, i);
- sources[i] = 0;
- healed_sinks[i] = 1;
- }
+ int i = 0;
+ afr_private_t *priv = NULL;
+ struct iatt srcstat = {
+ 0,
+ };
+ int source = -1;
+ int sources_count = 0;
+ int ret = 0;
+
+ priv = this->private;
+
+ sources_count = AFR_COUNT(sources, priv->child_count);
+
+ if ((AFR_CMP(locked_on, healed_sinks, priv->child_count) == 0) ||
+ !sources_count) {
+ source = afr_mark_split_brain_source_sinks(
+ frame, this, inode, sources, sinks, healed_sinks, locked_on,
+ replies, AFR_METADATA_TRANSACTION);
+ if (source >= 0) {
+ _afr_fav_child_reset_sink_xattrs(
+ frame, this, inode, source, healed_sinks, undid_pending,
+ AFR_METADATA_TRANSACTION, locked_on, replies);
+ goto out;
+ }
+
+ /* If this is a directory mtime/ctime only split brain
+ use the most recent */
+ source = afr_dirtime_splitbrain_source(frame, this, replies, locked_on);
+ if (source != -1) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_SPLIT_BRAIN,
+ "clear time "
+ "split brain on %s",
+ uuid_utoa(replies[source].poststat.ia_gfid));
+ sources[source] = 1;
+ healed_sinks[source] = 0;
+ goto out;
+ }
+
+ if (!priv->metadata_splitbrain_forced_heal) {
+ gf_event(EVENT_AFR_SPLIT_BRAIN,
+ "client-pid=%d;"
+ "subvol=%s;"
+ "type=metadata;file=%s",
+ this->ctx->cmd_args.client_pid, this->name,
+ uuid_utoa(inode->gfid));
+ return -EIO;
}
+ /* Metadata split brain, select one subvol
+ arbitrarily */
+ for (i = 0; i < priv->child_count; i++) {
+ if (locked_on[i] && healed_sinks[i]) {
+ sources[i] = 1;
+ healed_sinks[i] = 0;
+ break;
+ }
+ }
+ }
+
+ /* No split brain at this point. If we were called from
+ * afr_heal_splitbrain_file(), abort.*/
+ if (afr_dict_contains_heal_op(frame))
+ return -EIO;
+
+ source = afr_choose_source_by_policy(priv, sources,
+ AFR_METADATA_TRANSACTION);
+ srcstat = replies[source].poststat;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i] || i == source)
+ continue;
+ if (!IA_EQUAL(srcstat, replies[i].poststat, type) ||
+ !IA_EQUAL(srcstat, replies[i].poststat, uid) ||
+ !IA_EQUAL(srcstat, replies[i].poststat, gid) ||
+ !IA_EQUAL(srcstat, replies[i].poststat, prot)) {
+ gf_msg_debug(this->name, 0,
+ "%s: iatt mismatch "
+ "for source(%d) vs (%d)",
+ uuid_utoa(replies[source].poststat.ia_gfid), source,
+ i);
+ sources[i] = 0;
+ healed_sinks[i] = 1;
+ }
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i] || i == source)
+ continue;
+ if (!afr_xattrs_are_equal(replies[source].xdata, replies[i].xdata)) {
+ gf_msg_debug(this->name, 0,
+ "%s: xattr mismatch "
+ "for source(%d) vs (%d)",
+ uuid_utoa(replies[source].poststat.ia_gfid), source,
+ i);
+ sources[i] = 0;
+ healed_sinks[i] = 1;
+ }
+ }
+ if ((sources_count == priv->child_count) && (source > -1) &&
+ (AFR_COUNT(healed_sinks, priv->child_count) != 0)) {
+ ret = __afr_selfheal_metadata_mark_pending_xattrs(frame, this, inode,
+ replies, sources);
+ if (ret < 0)
+ return ret;
+ }
out:
- afr_mark_active_sinks (this, sources, locked_on, healed_sinks);
- return source;
+ afr_mark_active_sinks(this, sources, locked_on, healed_sinks);
+ return source;
}
-
int
-__afr_selfheal_metadata_prepare (call_frame_t *frame, xlator_t *this, inode_t *inode,
- unsigned char *locked_on, unsigned char *sources,
- unsigned char *sinks, unsigned char *healed_sinks,
- unsigned char *undid_pending,
- struct afr_reply *replies, gf_boolean_t *pflag)
+__afr_selfheal_metadata_prepare(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, unsigned char *locked_on,
+ unsigned char *sources, unsigned char *sinks,
+ unsigned char *healed_sinks,
+ unsigned char *undid_pending,
+ struct afr_reply *replies, unsigned char *pflag)
{
- int ret = -1;
- int source = -1;
- afr_private_t *priv = NULL;
- int i = 0;
- uint64_t *witness = NULL;
+ int ret = -1;
+ int source = -1;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ uint64_t *witness = NULL;
- priv = this->private;
+ priv = this->private;
- ret = afr_selfheal_unlocked_discover (frame, inode, inode->gfid,
- replies);
- if (ret)
- return ret;
-
- witness = alloca0 (sizeof (*witness) * priv->child_count);
- ret = afr_selfheal_find_direction (frame, this, replies,
- AFR_METADATA_TRANSACTION,
- locked_on, sources, sinks, witness,
- pflag);
- if (ret)
- return ret;
-
- /* Initialize the healed_sinks[] array optimistically to
- the intersection of to-be-healed (i.e sinks[]) and
- the list of servers which are up (i.e locked_on[]).
-
- As we encounter failures in the healing process, we
- will unmark the respective servers in the healed_sinks[]
- array.
- */
- AFR_INTERSECT (healed_sinks, sinks, locked_on, priv->child_count);
-
- /* If any source has witness, pick first
- * witness source and make everybody else sinks */
- for (i = 0; i < priv->child_count; i++) {
- if (sources[i] && witness[i]) {
- source = i;
- break;
- }
- }
+ ret = afr_selfheal_unlocked_discover(frame, inode, inode->gfid, replies);
+ if (ret)
+ return ret;
- if (source != -1) {
- for (i = 0; i < priv->child_count; i++) {
- if (i != source && sources[i]) {
- sources[i] = 0;
- healed_sinks[i] = 1;
- }
- }
+ witness = alloca0(sizeof(*witness) * priv->child_count);
+ ret = afr_selfheal_find_direction(frame, this, replies,
+ AFR_METADATA_TRANSACTION, locked_on,
+ sources, sinks, witness, pflag);
+ if (ret)
+ return ret;
+
+ /* Initialize the healed_sinks[] array optimistically to
+ the intersection of to-be-healed (i.e sinks[]) and
+ the list of servers which are up (i.e locked_on[]).
+
+ As we encounter failures in the healing process, we
+ will unmark the respective servers in the healed_sinks[]
+ array.
+ */
+ AFR_INTERSECT(healed_sinks, sinks, locked_on, priv->child_count);
+
+ /* If any source has witness, pick first
+ * witness source and make everybody else sinks */
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i] && witness[i]) {
+ source = i;
+ break;
}
+ }
- source = __afr_selfheal_metadata_finalize_source (frame, this, inode,
- sources, sinks,
- healed_sinks,
- undid_pending,
- locked_on, replies);
+ if (source != -1) {
+ for (i = 0; i < priv->child_count; i++) {
+ if (i != source && sources[i]) {
+ sources[i] = 0;
+ healed_sinks[i] = 1;
+ }
+ }
+ }
- if (source < 0)
- return -EIO;
+ source = __afr_selfheal_metadata_finalize_source(
+ frame, this, inode, sources, sinks, healed_sinks, undid_pending,
+ locked_on, replies);
- return source;
-}
+ if (source < 0)
+ return -EIO;
-int
-afr_selfheal_metadata (call_frame_t *frame, xlator_t *this, inode_t *inode)
-{
- afr_private_t *priv = NULL;
- int ret = -1;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *data_lock = NULL;
- unsigned char *healed_sinks = NULL;
- unsigned char *undid_pending = NULL;
- struct afr_reply *locked_replies = NULL;
- gf_boolean_t did_sh = _gf_true;
- int source = -1;
-
- priv = this->private;
-
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
- undid_pending = alloca0 (priv->child_count);
- data_lock = alloca0 (priv->child_count);
-
- locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
-
- ret = afr_selfheal_inodelk (frame, this, inode, this->name,
- LLONG_MAX - 1, 0, data_lock);
- {
- if (ret < AFR_SH_MIN_PARTICIPANTS) {
- ret = -ENOTCONN;
- goto unlock;
- }
-
- ret = __afr_selfheal_metadata_prepare (frame, this, inode,
- data_lock, sources,
- sinks, healed_sinks,
- undid_pending,
- locked_replies, NULL);
- if (ret < 0)
- goto unlock;
-
- source = ret;
-
- if (AFR_COUNT (healed_sinks, priv->child_count) == 0) {
- did_sh = _gf_false;
- goto unlock;
- }
-
- ret = __afr_selfheal_metadata_do (frame, this, inode, source,
- healed_sinks, locked_replies);
- if (ret)
- goto unlock;
-
- /* Restore atime/mtime for files that don't need data heal as
- * restoring timestamps happens only as a part of data-heal.
- */
- if (!IA_ISREG (locked_replies[source].poststat.ia_type))
- afr_selfheal_restore_time (frame, this, inode, source,
- healed_sinks, locked_replies);
-
- ret = afr_selfheal_undo_pending (frame, this, inode, sources,
- sinks, healed_sinks,
- undid_pending,
- AFR_METADATA_TRANSACTION,
- locked_replies, data_lock);
- }
-unlock:
- afr_selfheal_uninodelk (frame, this, inode, this->name,
- LLONG_MAX -1, 0, data_lock);
-
- if (did_sh)
- afr_log_selfheal (inode->gfid, this, ret, "metadata", source,
- sources, healed_sinks);
- else
- ret = 1;
-
- if (locked_replies)
- afr_replies_wipe (locked_replies, priv->child_count);
- return ret;
+ return source;
}
int
-afr_selfheal_metadata_by_stbuf (xlator_t *this, struct iatt *stbuf)
+afr_selfheal_metadata(call_frame_t *frame, xlator_t *this, inode_t *inode)
{
- inode_t *inode = NULL;
- inode_t *link_inode = NULL;
- call_frame_t *frame = NULL;
- int ret = 0;
-
- if (gf_uuid_is_null (stbuf->ia_gfid)) {
- ret = -EINVAL;
- goto out;
+ afr_private_t *priv = NULL;
+ int ret = -1;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ unsigned char *data_lock = NULL;
+ unsigned char *healed_sinks = NULL;
+ unsigned char *undid_pending = NULL;
+ struct afr_reply *locked_replies = NULL;
+ gf_boolean_t did_sh = _gf_true;
+ int source = -1;
+
+ priv = this->private;
+
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ healed_sinks = alloca0(priv->child_count);
+ undid_pending = alloca0(priv->child_count);
+ data_lock = alloca0(priv->child_count);
+
+ locked_replies = alloca0(sizeof(*locked_replies) * priv->child_count);
+
+ ret = afr_selfheal_inodelk(frame, this, inode, this->name, LLONG_MAX - 1, 0,
+ data_lock);
+ {
+ if (ret < priv->child_count) {
+ ret = -ENOTCONN;
+ goto unlock;
}
- inode = inode_new (this->itable);
- if (!inode) {
- ret = -ENOMEM;
- goto out;
- }
+ ret = __afr_selfheal_metadata_prepare(
+ frame, this, inode, data_lock, sources, sinks, healed_sinks,
+ undid_pending, locked_replies, NULL);
+ if (ret < 0)
+ goto unlock;
- link_inode = inode_link (inode, NULL, NULL, stbuf);
- if (!link_inode) {
- ret = -ENOMEM;
- goto out;
- }
+ source = ret;
- frame = afr_frame_create (this, &ret);
- if (!frame) {
- ret = -ret;
- goto out;
+ if (AFR_COUNT(healed_sinks, priv->child_count) == 0) {
+ did_sh = _gf_false;
+ goto unlock;
}
- ret = afr_selfheal_metadata (frame, this, link_inode);
+ ret = __afr_selfheal_metadata_do(frame, this, inode, source,
+ healed_sinks, locked_replies);
+ if (ret)
+ goto unlock;
+
+ afr_selfheal_restore_time(frame, this, inode, source, healed_sinks,
+ locked_replies);
+
+ ret = afr_selfheal_undo_pending(
+ frame, this, inode, sources, sinks, healed_sinks, undid_pending,
+ AFR_METADATA_TRANSACTION, locked_replies, data_lock);
+ }
+unlock:
+ afr_selfheal_uninodelk(frame, this, inode, this->name, LLONG_MAX - 1, 0,
+ data_lock);
+
+ if (did_sh)
+ afr_log_selfheal(inode->gfid, this, ret, "metadata", source, sources,
+ healed_sinks);
+ else
+ ret = 1;
+
+ if (locked_replies)
+ afr_replies_wipe(locked_replies, priv->child_count);
+ return ret;
+}
+
+int
+afr_selfheal_metadata_by_stbuf(xlator_t *this, struct iatt *stbuf)
+{
+ inode_t *inode = NULL;
+ inode_t *link_inode = NULL;
+ call_frame_t *frame = NULL;
+ int ret = 0;
+
+ if (gf_uuid_is_null(stbuf->ia_gfid)) {
+ ret = -EINVAL;
+ goto out;
+ }
+
+ inode = inode_new(this->itable);
+ if (!inode) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ link_inode = inode_link(inode, NULL, NULL, stbuf);
+ if (!link_inode) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ frame = afr_frame_create(this, &ret);
+ if (!frame) {
+ ret = -ret;
+ goto out;
+ }
+
+ ret = afr_selfheal_metadata(frame, this, link_inode);
out:
- if (inode)
- inode_unref (inode);
- if (link_inode)
- inode_unref (link_inode);
- if (frame)
- AFR_STACK_DESTROY (frame);
- return ret;
+ if (inode)
+ inode_unref(inode);
+ if (link_inode)
+ inode_unref(link_inode);
+ if (frame)
+ AFR_STACK_DESTROY(frame);
+ return ret;
}
diff --git a/xlators/cluster/afr/src/afr-self-heal-name.c b/xlators/cluster/afr/src/afr-self-heal-name.c
index 556d14baf82..834aac86d48 100644
--- a/xlators/cluster/afr/src/afr-self-heal-name.c
+++ b/xlators/cluster/afr/src/afr-self-heal-name.c
@@ -8,690 +8,609 @@
cases as published by the Free Software Foundation.
*/
-
-#include "events.h"
+#include <glusterfs/events.h>
#include "afr.h"
#include "afr-self-heal.h"
#include "afr-messages.h"
int
-__afr_selfheal_assign_gfid (xlator_t *this, inode_t *parent, uuid_t pargfid,
- const char *bname, inode_t *inode,
- struct afr_reply *replies, void *gfid,
- unsigned char *locked_on, int source,
- gf_boolean_t is_gfid_absent)
+__afr_selfheal_assign_gfid(xlator_t *this, inode_t *parent, uuid_t pargfid,
+ const char *bname, inode_t *inode,
+ struct afr_reply *replies, void *gfid,
+ unsigned char *locked_on, int source,
+ unsigned char *sources, gf_boolean_t is_gfid_absent,
+ int *gfid_idx)
{
- int ret = 0;
- int up_count = 0;
- int locked_count = 0;
- afr_private_t *priv = NULL;
+ int ret = 0;
+ int up_count = 0;
+ int locked_count = 0;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- gf_uuid_copy (parent->gfid, pargfid);
+ gf_uuid_copy(parent->gfid, pargfid);
- if (is_gfid_absent) {
- /* Ensure all children of AFR are up before performing gfid heal, to
- * guard against the possibility of gfid split brain. */
+ if (is_gfid_absent) {
+ /* Ensure all children of AFR are up before performing gfid heal, to
+ * guard against the possibility of gfid split brain. */
- up_count = AFR_COUNT (priv->child_up, priv->child_count);
- if (up_count != priv->child_count) {
- ret = -EIO;
- goto out;
- }
+ up_count = AFR_COUNT(priv->child_up, priv->child_count);
+ if (up_count != priv->child_count) {
+ ret = -EIO;
+ goto out;
+ }
- locked_count = AFR_COUNT (locked_on, priv->child_count);
- if (locked_count != priv->child_count) {
- ret = -EIO;
- goto out;
- }
+ locked_count = AFR_COUNT(locked_on, priv->child_count);
+ if (locked_count != priv->child_count) {
+ ret = -EIO;
+ goto out;
}
+ }
- afr_lookup_and_heal_gfid (this, parent, bname, inode, replies, source,
- gfid);
+ ret = afr_lookup_and_heal_gfid(this, parent, bname, inode, replies, source,
+ sources, gfid, gfid_idx);
out:
- return ret;
+ return ret;
}
int
-__afr_selfheal_name_impunge (call_frame_t *frame, xlator_t *this,
- inode_t *parent, uuid_t pargfid,
- const char *bname, inode_t *inode,
- struct afr_reply *replies, int gfid_idx)
+__afr_selfheal_name_impunge(call_frame_t *frame, xlator_t *this,
+ inode_t *parent, uuid_t pargfid, const char *bname,
+ inode_t *inode, struct afr_reply *replies,
+ int gfid_idx)
{
- int i = 0;
- afr_private_t *priv = NULL;
- int ret = 0;
- unsigned char *sources = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ int ret = 0;
+ unsigned char *sources = NULL;
- priv = this->private;
+ priv = this->private;
- sources = alloca0 (priv->child_count);
+ sources = alloca0(priv->child_count);
- gf_uuid_copy (parent->gfid, pargfid);
+ gf_uuid_copy(parent->gfid, pargfid);
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret != 0)
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret != 0)
+ continue;
- if (gf_uuid_compare (replies[i].poststat.ia_gfid,
- replies[gfid_idx].poststat.ia_gfid) == 0) {
- sources[i] = 1;
- continue;
- }
+ if (gf_uuid_compare(replies[i].poststat.ia_gfid,
+ replies[gfid_idx].poststat.ia_gfid) == 0) {
+ sources[i] = 1;
+ continue;
}
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (sources[i])
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i])
+ continue;
- ret |= afr_selfheal_recreate_entry (frame, i, gfid_idx, sources,
- parent, bname, inode,
- replies);
- }
+ ret |= afr_selfheal_recreate_entry(frame, i, gfid_idx, sources, parent,
+ bname, inode, replies);
+ }
- return ret;
+ return ret;
}
-
int
-__afr_selfheal_name_expunge (xlator_t *this, inode_t *parent, uuid_t pargfid,
- const char *bname, inode_t *inode,
- struct afr_reply *replies)
-{
- loc_t loc = {0, };
- int i = 0;
- afr_private_t *priv = NULL;
- char g[64];
- int ret = 0;
-
- priv = this->private;
-
- loc.parent = inode_ref (parent);
- gf_uuid_copy (loc.pargfid, pargfid);
- loc.name = bname;
- loc.inode = inode_ref (inode);
-
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
-
- if (replies[i].op_ret)
- continue;
-
- switch (replies[i].poststat.ia_type) {
- case IA_IFDIR:
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_EXPUNGING_FILE_OR_DIR,
- "expunging dir %s/%s (%s) on %s",
- uuid_utoa (pargfid), bname,
- uuid_utoa_r (replies[i].poststat.ia_gfid, g),
- priv->children[i]->name);
-
- ret |= syncop_rmdir (priv->children[i], &loc, 1, NULL,
- NULL);
- break;
- default:
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_EXPUNGING_FILE_OR_DIR,
- "expunging file %s/%s (%s) on %s",
- uuid_utoa (pargfid), bname,
- uuid_utoa_r (replies[i].poststat.ia_gfid, g),
- priv->children[i]->name);
-
- ret |= syncop_unlink (priv->children[i], &loc, NULL,
- NULL);
- break;
- }
- }
-
- loc_wipe (&loc);
-
- return ret;
-
-}
-
-/* This function is to be called after ensuring that there is no gfid mismatch
- * for the inode across multiple sources
- */
-static int
-afr_selfheal_gfid_idx_get (xlator_t *this, struct afr_reply *replies,
- unsigned char *sources)
+__afr_selfheal_name_expunge(xlator_t *this, inode_t *parent, uuid_t pargfid,
+ const char *bname, inode_t *inode,
+ struct afr_reply *replies)
{
- int i = 0;
- int gfid_idx = -1;
- afr_private_t *priv = NULL;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ int ret = 0;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret != 0)
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
- if (!sources[i])
- continue;
+ if (replies[i].op_ret)
+ continue;
- if (gf_uuid_is_null (replies[i].poststat.ia_gfid))
- continue;
+ ret |= afr_selfheal_entry_delete(this, parent, bname, inode, i,
+ replies);
+ }
- gfid_idx = i;
- break;
- }
- return gfid_idx;
+ return ret;
}
static gf_boolean_t
-afr_selfheal_name_need_heal_check (xlator_t *this, struct afr_reply *replies)
+afr_selfheal_name_need_heal_check(xlator_t *this, struct afr_reply *replies)
{
- int i = 0;
- int first_idx = -1;
- gf_boolean_t need_heal = _gf_false;
- afr_private_t *priv = NULL;
-
- priv = this->private;
+ int i = 0;
+ int first_idx = -1;
+ gf_boolean_t need_heal = _gf_false;
+ afr_private_t *priv = NULL;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
+ priv = this->private;
- if ((replies[i].op_ret == -1) &&
- (replies[i].op_errno == ENODATA))
- need_heal = _gf_true;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
- if (first_idx == -1) {
- first_idx = i;
- continue;
- }
+ if ((replies[i].op_ret == -1) && (replies[i].op_errno == ENODATA))
+ need_heal = _gf_true;
- if (replies[i].op_ret != replies[first_idx].op_ret)
- need_heal = _gf_true;
+ if (first_idx == -1) {
+ first_idx = i;
+ continue;
+ }
- if (gf_uuid_compare (replies[i].poststat.ia_gfid,
- replies[first_idx].poststat.ia_gfid))
- need_heal = _gf_true;
+ if (replies[i].op_ret != replies[first_idx].op_ret)
+ need_heal = _gf_true;
- if ((replies[i].op_ret == 0) &&
- (gf_uuid_is_null(replies[i].poststat.ia_gfid)))
- need_heal = _gf_true;
+ if (gf_uuid_compare(replies[i].poststat.ia_gfid,
+ replies[first_idx].poststat.ia_gfid))
+ need_heal = _gf_true;
- }
+ if ((replies[i].op_ret == 0) &&
+ (gf_uuid_is_null(replies[i].poststat.ia_gfid)))
+ need_heal = _gf_true;
+ }
- return need_heal;
+ return need_heal;
}
static int
-afr_selfheal_name_type_mismatch_check (xlator_t *this, struct afr_reply *replies,
- int source, unsigned char *sources,
- uuid_t pargfid, const char *bname)
+afr_selfheal_name_type_mismatch_check(xlator_t *this, struct afr_reply *replies,
+ int source, unsigned char *sources,
+ uuid_t pargfid, const char *bname)
{
- int i = 0;
- int type_idx = -1;
- ia_type_t inode_type = IA_INVAL;
- ia_type_t inode_type1 = IA_INVAL;
- afr_private_t *priv = NULL;
+ int i = 0;
+ int type_idx = -1;
+ ia_type_t inode_type = IA_INVAL;
+ ia_type_t inode_type1 = IA_INVAL;
+ afr_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret != 0)
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret != 0)
+ continue;
- if (replies[i].poststat.ia_type == IA_INVAL)
- continue;
+ if (replies[i].poststat.ia_type == IA_INVAL)
+ continue;
- if (inode_type == IA_INVAL) {
- inode_type = replies[i].poststat.ia_type;
- type_idx = i;
- continue;
- }
- inode_type1 = replies[i].poststat.ia_type;
- if (sources[i] || source == -1) {
- if ((sources[type_idx] || source == -1) &&
- (inode_type != inode_type1)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_SPLIT_BRAIN,
- "Type mismatch for <gfid:%s>/%s: "
- "%s on %s and %s on %s",
- uuid_utoa(pargfid), bname,
- gf_inode_type_to_str (inode_type1),
- priv->children[i]->name,
- gf_inode_type_to_str (inode_type),
- priv->children[type_idx]->name);
- gf_event (EVENT_AFR_SPLIT_BRAIN,
- "subvol=%s;type=file;"
- "file=<gfid:%s>/%s;count=2;"
- "child-%d=%s;type-%d=%s;child-%d=%s;"
- "type-%d=%s", this->name,
- uuid_utoa (pargfid), bname, i,
- priv->children[i]->name, i,
- gf_inode_type_to_str (inode_type1),
- type_idx,
- priv->children[type_idx]->name,
- type_idx,
- gf_inode_type_to_str (inode_type));
- return -EIO;
- }
- inode_type = replies[i].poststat.ia_type;
- type_idx = i;
- }
+ if (inode_type == IA_INVAL) {
+ inode_type = replies[i].poststat.ia_type;
+ type_idx = i;
+ continue;
}
- return 0;
+ inode_type1 = replies[i].poststat.ia_type;
+ if (sources[i] || source == -1) {
+ if ((sources[type_idx] || source == -1) &&
+ (inode_type != inode_type1)) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_SPLIT_BRAIN,
+ "Type mismatch for <gfid:%s>/%s: "
+ "%s on %s and %s on %s",
+ uuid_utoa(pargfid), bname,
+ gf_inode_type_to_str(inode_type1),
+ priv->children[i]->name,
+ gf_inode_type_to_str(inode_type),
+ priv->children[type_idx]->name);
+ gf_event(EVENT_AFR_SPLIT_BRAIN,
+ "client-pid=%d;"
+ "subvol=%s;type=file;"
+ "file=<gfid:%s>/%s;count=2;"
+ "child-%d=%s;type-%d=%s;child-%d=%s;"
+ "type-%d=%s",
+ this->ctx->cmd_args.client_pid, this->name,
+ uuid_utoa(pargfid), bname, i, priv->children[i]->name,
+ i, gf_inode_type_to_str(inode_type1), type_idx,
+ priv->children[type_idx]->name, type_idx,
+ gf_inode_type_to_str(inode_type));
+ return -EIO;
+ }
+ inode_type = replies[i].poststat.ia_type;
+ type_idx = i;
+ }
+ }
+ return 0;
}
static int
-afr_selfheal_name_gfid_mismatch_check (xlator_t *this, struct afr_reply *replies,
- int source, unsigned char *sources,
- int *gfid_idx, uuid_t pargfid,
- const char *bname, inode_t *inode,
- unsigned char *locked_on, dict_t *xdata)
+afr_selfheal_name_gfid_mismatch_check(xlator_t *this, struct afr_reply *replies,
+ int source, unsigned char *sources,
+ int *gfid_idx, uuid_t pargfid,
+ const char *bname, inode_t *inode,
+ unsigned char *locked_on, dict_t *xdata)
{
- int i = 0;
- int gfid_idx_iter = -1;
- int ret = -1;
- void *gfid = NULL;
- void *gfid1 = NULL;
- afr_private_t *priv = NULL;
-
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid || replies[i].op_ret != 0)
- continue;
-
- if (gf_uuid_is_null (replies[i].poststat.ia_gfid))
- continue;
-
- if (!gfid) {
- gfid = &replies[i].poststat.ia_gfid;
- gfid_idx_iter = i;
- continue;
- }
-
- gfid1 = &replies[i].poststat.ia_gfid;
- if (sources[i] || source == -1) {
- if ((sources[gfid_idx_iter] || source == -1) &&
- gf_uuid_compare (gfid, gfid1)) {
- ret = afr_gfid_split_brain_source (this,
- replies,
- inode,
- pargfid,
- bname,
- gfid_idx_iter,
- i, locked_on,
- gfid_idx,
- xdata);
- if (!ret && *gfid_idx >= 0) {
- ret = dict_set_str (xdata,
- "gfid-heal-msg",
- "GFID split-brain "
- "resolved");
- if (ret)
- gf_msg (this->name,
- GF_LOG_ERROR, 0,
- AFR_MSG_DICT_SET_FAILED,
- "Error setting gfid-"
- "heal-msg dict");
- }
- return ret;
- }
- gfid = &replies[i].poststat.ia_gfid;
- gfid_idx_iter = i;
- }
- }
-
- *gfid_idx = gfid_idx_iter;
- return 0;
+ int i = 0;
+ int gfid_idx_iter = -1;
+ int ret = -1;
+ void *gfid = NULL;
+ void *gfid1 = NULL;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid || replies[i].op_ret != 0)
+ continue;
+
+ if (gf_uuid_is_null(replies[i].poststat.ia_gfid))
+ continue;
+
+ if (!gfid) {
+ gfid = &replies[i].poststat.ia_gfid;
+ gfid_idx_iter = i;
+ continue;
+ }
+
+ gfid1 = &replies[i].poststat.ia_gfid;
+ if (sources[i] || source == -1) {
+ if ((sources[gfid_idx_iter] || source == -1) &&
+ gf_uuid_compare(gfid, gfid1)) {
+ ret = afr_gfid_split_brain_source(this, replies, inode, pargfid,
+ bname, gfid_idx_iter, i,
+ locked_on, gfid_idx, xdata);
+ if (!ret && *gfid_idx >= 0) {
+ ret = dict_set_sizen_str_sizen(xdata, "gfid-heal-msg",
+ "GFID split-brain resolved");
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ AFR_MSG_DICT_SET_FAILED,
+ "Error setting gfid-"
+ "heal-msg dict");
+ }
+ return ret;
+ }
+ gfid = &replies[i].poststat.ia_gfid;
+ gfid_idx_iter = i;
+ }
+ }
+
+ *gfid_idx = gfid_idx_iter;
+ return 0;
}
static gf_boolean_t
-afr_selfheal_name_source_empty_check (xlator_t *this, struct afr_reply *replies,
- unsigned char *sources, int source)
+afr_selfheal_name_source_empty_check(xlator_t *this, struct afr_reply *replies,
+ unsigned char *sources, int source)
{
- int i = 0;
- afr_private_t *priv = NULL;
- gf_boolean_t source_is_empty = _gf_true;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ gf_boolean_t source_is_empty = _gf_true;
- priv = this->private;
+ priv = this->private;
- if (source == -1) {
- source_is_empty = _gf_false;
- goto out;
- }
+ if (source == -1) {
+ source_is_empty = _gf_false;
+ goto out;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (!sources[i])
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!sources[i])
+ continue;
- if (replies[i].op_ret == -1 && replies[i].op_errno == ENOENT)
- continue;
+ if (replies[i].op_ret == -1 && replies[i].op_errno == ENOENT)
+ continue;
- source_is_empty = _gf_false;
- break;
- }
+ source_is_empty = _gf_false;
+ break;
+ }
out:
- return source_is_empty;
+ return source_is_empty;
}
int
-__afr_selfheal_name_do (call_frame_t *frame, xlator_t *this, inode_t *parent,
- uuid_t pargfid, const char *bname, inode_t *inode,
- unsigned char *sources, unsigned char *sinks,
- unsigned char *healed_sinks, int source,
- unsigned char *locked_on, struct afr_reply *replies,
- void *gfid_req, dict_t *xdata)
+__afr_selfheal_name_do(call_frame_t *frame, xlator_t *this, inode_t *parent,
+ uuid_t pargfid, const char *bname, inode_t *inode,
+ unsigned char *sources, unsigned char *sinks,
+ unsigned char *healed_sinks, int source,
+ unsigned char *locked_on, struct afr_reply *replies,
+ void *gfid_req, dict_t *xdata)
{
- int gfid_idx = -1;
- int ret = -1;
- void *gfid = NULL;
- gf_boolean_t source_is_empty = _gf_true;
- gf_boolean_t need_heal = _gf_false;
- gf_boolean_t is_gfid_absent = _gf_false;
-
- need_heal = afr_selfheal_name_need_heal_check (this, replies);
- if (!need_heal)
- return 0;
-
- source_is_empty = afr_selfheal_name_source_empty_check (this, replies,
- sources,
- source);
- if (source_is_empty) {
- ret = __afr_selfheal_name_expunge (this, parent, pargfid,
- bname, inode, replies);
- if (ret == -EIO)
- ret = -1;
- return ret;
- }
-
- ret = afr_selfheal_name_type_mismatch_check (this, replies, source,
- sources, pargfid, bname);
- if (ret)
- return ret;
+ int gfid_idx = -1;
+ int ret = -1;
+ void *gfid = NULL;
+ gf_boolean_t source_is_empty = _gf_true;
+ gf_boolean_t need_heal = _gf_false;
+ gf_boolean_t is_gfid_absent = _gf_false;
+
+ need_heal = afr_selfheal_name_need_heal_check(this, replies);
+ if (!need_heal)
+ return 0;
- ret = afr_selfheal_name_gfid_mismatch_check (this, replies, source,
- sources, &gfid_idx,
- pargfid, bname, inode,
- locked_on, xdata);
- if (ret)
- return ret;
+ source_is_empty = afr_selfheal_name_source_empty_check(this, replies,
+ sources, source);
+ if (source_is_empty) {
+ ret = __afr_selfheal_name_expunge(this, parent, pargfid, bname, inode,
+ replies);
+ if (ret == -EIO)
+ ret = -1;
+ return ret;
+ }
- if (gfid_idx == -1) {
- if (!gfid_req || gf_uuid_is_null (gfid_req))
- return -1;
- gfid = gfid_req;
- } else {
- gfid = &replies[gfid_idx].poststat.ia_gfid;
- }
+ ret = afr_selfheal_name_type_mismatch_check(this, replies, source, sources,
+ pargfid, bname);
+ if (ret)
+ return ret;
- is_gfid_absent = (gfid_idx == -1) ? _gf_true : _gf_false;
- ret = __afr_selfheal_assign_gfid (this, parent, pargfid, bname, inode,
- replies, gfid, locked_on, source,
- is_gfid_absent);
- if (ret)
- return ret;
+ ret = afr_selfheal_name_gfid_mismatch_check(this, replies, source, sources,
+ &gfid_idx, pargfid, bname,
+ inode, locked_on, xdata);
+ if (ret)
+ return ret;
- if (gfid_idx == -1) {
- gfid_idx = afr_selfheal_gfid_idx_get (this, replies, sources);
- if (gfid_idx == -1)
- return -1;
- }
+ if (gfid_idx == -1) {
+ if (!gfid_req || gf_uuid_is_null(gfid_req))
+ return -1;
+ gfid = gfid_req;
+ } else {
+ gfid = &replies[gfid_idx].poststat.ia_gfid;
+ if (source == -1)
+ /* Either entry split-brain or dirty xattrs are present on parent.*/
+ source = gfid_idx;
+ }
+
+ is_gfid_absent = (gfid_idx == -1) ? _gf_true : _gf_false;
+ ret = __afr_selfheal_assign_gfid(this, parent, pargfid, bname, inode,
+ replies, gfid, locked_on, source, sources,
+ is_gfid_absent, &gfid_idx);
+ if (ret || (gfid_idx < 0))
+ return ret;
- ret = __afr_selfheal_name_impunge (frame, this, parent, pargfid,
- bname, inode, replies, gfid_idx);
- if (ret == -EIO)
- ret = -1;
+ ret = __afr_selfheal_name_impunge(frame, this, parent, pargfid, bname,
+ inode, replies, gfid_idx);
+ if (ret == -EIO)
+ ret = -1;
- return ret;
+ return ret;
}
-
int
-__afr_selfheal_name_finalize_source (xlator_t *this, unsigned char *sources,
- unsigned char *healed_sinks,
- unsigned char *locked_on,
- uint64_t *witness)
+__afr_selfheal_name_finalize_source(xlator_t *this, unsigned char *sources,
+ unsigned char *healed_sinks,
+ unsigned char *locked_on, uint64_t *witness)
{
- int i = 0;
- afr_private_t *priv = NULL;
- int source = -1;
- int sources_count = 0;
-
- priv = this->private;
-
- sources_count = AFR_COUNT (sources, priv->child_count);
-
- if ((AFR_CMP (locked_on, healed_sinks, priv->child_count) == 0)
- || !sources_count || afr_does_witness_exist (this, witness)) {
- memset (sources, 0, sizeof (*sources) * priv->child_count);
- afr_mark_active_sinks (this, sources, locked_on, healed_sinks);
- return -1;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (sources[i]) {
- source = i;
- break;
- }
- }
-
- return source;
+ int i = 0;
+ afr_private_t *priv = NULL;
+ int source = -1;
+ int sources_count = 0;
+
+ priv = this->private;
+
+ sources_count = AFR_COUNT(sources, priv->child_count);
+
+ if ((AFR_CMP(locked_on, healed_sinks, priv->child_count) == 0) ||
+ !sources_count || afr_does_witness_exist(this, witness)) {
+ memset(sources, 0, sizeof(*sources) * priv->child_count);
+ afr_mark_active_sinks(this, sources, locked_on, healed_sinks);
+ return -1;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (sources[i]) {
+ source = i;
+ break;
+ }
+ }
+
+ return source;
}
int
-__afr_selfheal_name_prepare (call_frame_t *frame, xlator_t *this, inode_t *parent,
- uuid_t pargfid, unsigned char *locked_on,
- unsigned char *sources, unsigned char *sinks,
- unsigned char *healed_sinks, int *source_p)
+__afr_selfheal_name_prepare(call_frame_t *frame, xlator_t *this,
+ inode_t *parent, uuid_t pargfid,
+ unsigned char *locked_on, unsigned char *sources,
+ unsigned char *sinks, unsigned char *healed_sinks,
+ int *source_p)
{
- int ret = -1;
- int source = -1;
- afr_private_t *priv = NULL;
- struct afr_reply *replies = NULL;
- uint64_t *witness = NULL;
-
- priv = this->private;
-
- replies = alloca0 (priv->child_count * sizeof(*replies));
-
- ret = afr_selfheal_unlocked_discover (frame, parent, pargfid, replies);
- if (ret)
- goto out;
-
- witness = alloca0 (sizeof (*witness) * priv->child_count);
- ret = afr_selfheal_find_direction (frame, this, replies,
- AFR_ENTRY_TRANSACTION,
- locked_on, sources, sinks, witness,
- NULL);
- if (ret)
- goto out;
-
- /* Initialize the healed_sinks[] array optimistically to
- the intersection of to-be-healed (i.e sinks[]) and
- the list of servers which are up (i.e locked_on[]).
-
- As we encounter failures in the healing process, we
- will unmark the respective servers in the healed_sinks[]
- array.
- */
- AFR_INTERSECT (healed_sinks, sinks, locked_on, priv->child_count);
-
- source = __afr_selfheal_name_finalize_source (this, sources,
- healed_sinks,
- locked_on, witness);
- if (source < 0) {
- /* If source is < 0 (typically split-brain), we perform a
- conservative merge of entries rather than erroring out */
- }
- *source_p = source;
+ int ret = -1;
+ int source = -1;
+ afr_private_t *priv = NULL;
+ struct afr_reply *replies = NULL;
+ uint64_t *witness = NULL;
+
+ priv = this->private;
+
+ replies = alloca0(priv->child_count * sizeof(*replies));
+
+ ret = afr_selfheal_unlocked_discover(frame, parent, pargfid, replies);
+ if (ret)
+ goto out;
+
+ witness = alloca0(sizeof(*witness) * priv->child_count);
+ ret = afr_selfheal_find_direction(frame, this, replies,
+ AFR_ENTRY_TRANSACTION, locked_on, sources,
+ sinks, witness, NULL);
+ if (ret)
+ goto out;
+
+ /* Initialize the healed_sinks[] array optimistically to
+ the intersection of to-be-healed (i.e sinks[]) and
+ the list of servers which are up (i.e locked_on[]).
+
+ As we encounter failures in the healing process, we
+ will unmark the respective servers in the healed_sinks[]
+ array.
+ */
+ AFR_INTERSECT(healed_sinks, sinks, locked_on, priv->child_count);
+
+ source = __afr_selfheal_name_finalize_source(this, sources, healed_sinks,
+ locked_on, witness);
+ if (source < 0) {
+ /* If source is < 0 (typically split-brain), we perform a
+ conservative merge of entries rather than erroring out */
+ }
+ *source_p = source;
out:
- if (replies)
- afr_replies_wipe (replies, priv->child_count);
+ if (replies)
+ afr_replies_wipe(replies, priv->child_count);
- return ret;
+ return ret;
}
-
int
-afr_selfheal_name_do (call_frame_t *frame, xlator_t *this, inode_t *parent,
- uuid_t pargfid, const char *bname, void *gfid_req,
- dict_t *xdata)
+afr_selfheal_name_do(call_frame_t *frame, xlator_t *this, inode_t *parent,
+ uuid_t pargfid, const char *bname, void *gfid_req,
+ dict_t *xdata)
{
- afr_private_t *priv = NULL;
- unsigned char *sources = NULL;
- unsigned char *sinks = NULL;
- unsigned char *healed_sinks = NULL;
- unsigned char *locked_on = NULL;
- int source = -1;
- struct afr_reply *replies = NULL;
- int ret = -1;
- inode_t *inode = NULL;
- dict_t *xattr = NULL;
-
- xattr = dict_new ();
- if (!xattr)
- return -ENOMEM;
-
- ret = dict_set_int32 (xattr, GF_GFIDLESS_LOOKUP, 1);
- if (ret) {
- dict_unref (xattr);
- return -1;
+ afr_private_t *priv = NULL;
+ unsigned char *sources = NULL;
+ unsigned char *sinks = NULL;
+ unsigned char *healed_sinks = NULL;
+ unsigned char *locked_on = NULL;
+ int source = -1;
+ struct afr_reply *replies = NULL;
+ int ret = -1;
+ inode_t *inode = NULL;
+ dict_t *xattr = NULL;
+
+ xattr = dict_new();
+ if (!xattr)
+ return -ENOMEM;
+
+ ret = dict_set_int32_sizen(xattr, GF_GFIDLESS_LOOKUP, 1);
+ if (ret) {
+ dict_unref(xattr);
+ return -1;
+ }
+
+ priv = this->private;
+
+ locked_on = alloca0(priv->child_count);
+ sources = alloca0(priv->child_count);
+ sinks = alloca0(priv->child_count);
+ healed_sinks = alloca0(priv->child_count);
+
+ replies = alloca0(priv->child_count * sizeof(*replies));
+
+ ret = afr_selfheal_entrylk(frame, this, parent, this->name, bname,
+ locked_on);
+ {
+ if (ret < priv->child_count) {
+ ret = -ENOTCONN;
+ goto unlock;
+ }
+
+ ret = __afr_selfheal_name_prepare(frame, this, parent, pargfid,
+ locked_on, sources, sinks,
+ healed_sinks, &source);
+ if (ret)
+ goto unlock;
+
+ inode = afr_selfheal_unlocked_lookup_on(frame, parent, bname, replies,
+ locked_on, xattr);
+ if (!inode) {
+ ret = -ENOMEM;
+ goto unlock;
}
- priv = this->private;
-
- locked_on = alloca0 (priv->child_count);
- sources = alloca0 (priv->child_count);
- sinks = alloca0 (priv->child_count);
- healed_sinks = alloca0 (priv->child_count);
-
- replies = alloca0 (priv->child_count * sizeof(*replies));
-
- ret = afr_selfheal_entrylk (frame, this, parent, this->name, bname,
- locked_on);
- {
- if (ret < AFR_SH_MIN_PARTICIPANTS) {
- ret = -ENOTCONN;
- goto unlock;
- }
-
- ret = __afr_selfheal_name_prepare (frame, this, parent, pargfid,
- locked_on, sources, sinks,
- healed_sinks, &source);
- if (ret)
- goto unlock;
-
- inode = afr_selfheal_unlocked_lookup_on (frame, parent, bname,
- replies, locked_on,
- xattr);
- if (!inode) {
- ret = -ENOMEM;
- goto unlock;
- }
-
- ret = __afr_selfheal_name_do (frame, this, parent, pargfid,
- bname, inode, sources, sinks,
- healed_sinks, source, locked_on,
- replies, gfid_req, xdata);
- }
+ ret = __afr_selfheal_name_do(frame, this, parent, pargfid, bname, inode,
+ sources, sinks, healed_sinks, source,
+ locked_on, replies, gfid_req, xdata);
+ }
unlock:
- afr_selfheal_unentrylk (frame, this, parent, this->name, bname,
- locked_on, NULL);
- if (inode)
- inode_unref (inode);
+ afr_selfheal_unentrylk(frame, this, parent, this->name, bname, locked_on,
+ NULL);
+ if (inode)
+ inode_unref(inode);
- if (replies)
- afr_replies_wipe (replies, priv->child_count);
- if (xattr)
- dict_unref (xattr);
+ if (replies)
+ afr_replies_wipe(replies, priv->child_count);
+ if (xattr)
+ dict_unref(xattr);
- return ret;
+ return ret;
}
-
int
-afr_selfheal_name_unlocked_inspect (call_frame_t *frame, xlator_t *this,
- inode_t *parent, uuid_t pargfid,
- const char *bname, gf_boolean_t *need_heal)
+afr_selfheal_name_unlocked_inspect(call_frame_t *frame, xlator_t *this,
+ inode_t *parent, uuid_t pargfid,
+ const char *bname, gf_boolean_t *need_heal)
{
- afr_private_t *priv = NULL;
- int i = 0;
- struct afr_reply *replies = NULL;
- inode_t *inode = NULL;
- int first_idx = -1;
-
- priv = this->private;
-
- replies = alloca0 (sizeof (*replies) * priv->child_count);
-
- inode = afr_selfheal_unlocked_lookup_on (frame, parent, bname,
- replies, priv->child_up, NULL);
- if (!inode)
- return -ENOMEM;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!replies[i].valid)
- continue;
-
- if ((replies[i].op_ret == -1) &&
- (replies[i].op_errno == ENODATA))
- *need_heal = _gf_true;
-
- if (first_idx == -1) {
- first_idx = i;
- continue;
- }
-
- if (replies[i].op_ret != replies[first_idx].op_ret)
- *need_heal = _gf_true;
-
- if (gf_uuid_compare (replies[i].poststat.ia_gfid,
- replies[first_idx].poststat.ia_gfid))
- *need_heal = _gf_true;
- }
-
- if (inode)
- inode_unref (inode);
- if (replies)
- afr_replies_wipe (replies, priv->child_count);
- return 0;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ struct afr_reply *replies = NULL;
+ inode_t *inode = NULL;
+ int first_idx = -1;
+ afr_local_t *local = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ replies = alloca0(sizeof(*replies) * priv->child_count);
+
+ inode = afr_selfheal_unlocked_lookup_on(frame, parent, bname, replies,
+ local->child_up, NULL);
+ if (!inode)
+ return -ENOMEM;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!replies[i].valid)
+ continue;
+
+ if ((replies[i].op_ret == -1) && (replies[i].op_errno == ENODATA)) {
+ *need_heal = _gf_true;
+ break;
+ }
+
+ if (first_idx == -1) {
+ first_idx = i;
+ continue;
+ }
+
+ if (replies[i].op_ret != replies[first_idx].op_ret) {
+ *need_heal = _gf_true;
+ break;
+ }
+
+ if (gf_uuid_compare(replies[i].poststat.ia_gfid,
+ replies[first_idx].poststat.ia_gfid)) {
+ *need_heal = _gf_true;
+ break;
+ }
+ }
+
+ if (inode)
+ inode_unref(inode);
+ if (replies)
+ afr_replies_wipe(replies, priv->child_count);
+ return 0;
}
int
-afr_selfheal_name (xlator_t *this, uuid_t pargfid, const char *bname,
- void *gfid_req, dict_t *xdata)
+afr_selfheal_name(xlator_t *this, uuid_t pargfid, const char *bname,
+ void *gfid_req, dict_t *xdata)
{
- inode_t *parent = NULL;
- call_frame_t *frame = NULL;
- int ret = -1;
- gf_boolean_t need_heal = _gf_false;
-
- parent = afr_inode_find (this, pargfid);
- if (!parent)
- goto out;
-
- frame = afr_frame_create (this, NULL);
- if (!frame)
- goto out;
-
- ret = afr_selfheal_name_unlocked_inspect (frame, this, parent, pargfid,
- bname, &need_heal);
- if (ret)
- goto out;
-
- if (need_heal) {
- ret = afr_selfheal_name_do (frame, this, parent, pargfid, bname,
- gfid_req, xdata);
- if (ret)
- goto out;
- }
+ inode_t *parent = NULL;
+ call_frame_t *frame = NULL;
+ int ret = -1;
+ gf_boolean_t need_heal = _gf_false;
+
+ parent = afr_inode_find(this, pargfid);
+ if (!parent)
+ goto out;
+
+ frame = afr_frame_create(this, NULL);
+ if (!frame)
+ goto out;
+
+ ret = afr_selfheal_name_unlocked_inspect(frame, this, parent, pargfid,
+ bname, &need_heal);
+ if (ret)
+ goto out;
+
+ if (need_heal) {
+ ret = afr_selfheal_name_do(frame, this, parent, pargfid, bname,
+ gfid_req, xdata);
+ if (ret)
+ goto out;
+ }
- ret = 0;
+ ret = 0;
out:
- if (parent)
- inode_unref (parent);
- if (frame)
- AFR_STACK_DESTROY (frame);
+ if (parent)
+ inode_unref(parent);
+ if (frame)
+ AFR_STACK_DESTROY(frame);
- return ret;
+ return ret;
}
diff --git a/xlators/cluster/afr/src/afr-self-heal.h b/xlators/cluster/afr/src/afr-self-heal.h
index cd67d2a3192..48e6dbcfb18 100644
--- a/xlators/cluster/afr/src/afr-self-heal.h
+++ b/xlators/cluster/afr/src/afr-self-heal.h
@@ -8,265 +8,285 @@
cases as published by the Free Software Foundation.
*/
-
#ifndef _AFR_SELFHEAL_H
#define _AFR_SELFHEAL_H
-#define AFR_SH_MIN_PARTICIPANTS 2
-
/* Perform fop on all UP subvolumes and wait for all callbacks to return */
-#define AFR_ONALL(frame, rfn, fop, args ...) do { \
- afr_local_t *__local = frame->local; \
- afr_private_t *__priv = frame->this->private; \
- int __i = 0, __count = 0; \
- unsigned char *__child_up = NULL; \
- \
- __child_up = alloca0 (__priv->child_count); \
- memcpy (__child_up, __priv->child_up, \
- sizeof (*__child_up) * __priv->child_count); \
- __count = AFR_COUNT (__child_up, __priv->child_count); \
- \
- __local->barrier.waitfor = __count; \
- afr_local_replies_wipe (__local, __priv); \
- \
- for (__i = 0; __i < __priv->child_count; __i++) { \
- if (!__child_up[__i]) \
- continue; \
- STACK_WIND_COOKIE (frame, rfn, (void *)(long) __i, \
- __priv->children[__i], \
- __priv->children[__i]->fops->fop, args); \
- } \
- syncbarrier_wait (&__local->barrier, __count); \
- } while (0)
-
+#define AFR_ONALL(frame, rfn, fop, args...) \
+ do { \
+ afr_local_t *__local = frame->local; \
+ afr_private_t *__priv = frame->this->private; \
+ int __i = 0, __count = 0; \
+ unsigned char *__child_up = alloca(__priv->child_count); \
+ \
+ memcpy(__child_up, __priv->child_up, \
+ sizeof(*__child_up) * __priv->child_count); \
+ __count = AFR_COUNT(__child_up, __priv->child_count); \
+ \
+ __local->barrier.waitfor = __count; \
+ afr_local_replies_wipe(__local, __priv); \
+ \
+ for (__i = 0; __i < __priv->child_count; __i++) { \
+ if (!__child_up[__i]) \
+ continue; \
+ STACK_WIND_COOKIE(frame, rfn, (void *)(long)__i, \
+ __priv->children[__i], \
+ __priv->children[__i]->fops->fop, args); \
+ } \
+ syncbarrier_wait(&__local->barrier, __count); \
+ } while (0)
/* Perform fop on all subvolumes represented by list[] array and wait
for all callbacks to return */
-#define AFR_ONLIST(list, frame, rfn, fop, args ...) do { \
- afr_local_t *__local = frame->local; \
- afr_private_t *__priv = frame->this->private; \
- int __i = 0; \
- int __count = AFR_COUNT (list, __priv->child_count); \
- \
- __local->barrier.waitfor = __count; \
- afr_local_replies_wipe (__local, __priv); \
- \
- for (__i = 0; __i < __priv->child_count; __i++) { \
- if (!list[__i]) continue; \
- STACK_WIND_COOKIE (frame, rfn, (void *)(long) __i, \
- __priv->children[__i], \
- __priv->children[__i]->fops->fop, args); \
- } \
- syncbarrier_wait (&__local->barrier, __count); \
- } while (0)
-
-
-#define AFR_SEQ(frame, rfn, fop, args ...) do { \
- afr_local_t *__local = frame->local; \
- afr_private_t *__priv = frame->this->private; \
- int __i = 0; \
- \
- afr_local_replies_wipe (__local, __priv); \
- \
- for (__i = 0; __i < __priv->child_count; __i++) { \
- if (!__priv->child_up[__i]) continue; \
- STACK_WIND_COOKIE (frame, rfn, (void *)(long) __i, \
- __priv->children[__i], \
- __priv->children[__i]->fops->fop, args); \
- syncbarrier_wait (&__local->barrier, 1); \
- } \
- } while (0)
-
-
-#define ALLOC_MATRIX(n, type) ({type **__ptr = NULL; \
- int __i; \
- __ptr = alloca0 (n * sizeof(type *)); \
- for (__i = 0; __i < n; __i++) __ptr[__i] = alloca0 (n * sizeof(type)); \
- __ptr;})
-
-
-#define IA_EQUAL(f,s,field) (memcmp (&(f.ia_##field), &(s.ia_##field), sizeof (s.ia_##field)) == 0)
-
-#define SBRAIN_HEAL_NO_GO_MSG "Failed to obtain replies from all bricks of "\
- "the replica (are they up?). Cannot resolve split-brain."
-int
-afr_selfheal (xlator_t *this, uuid_t gfid);
+#define AFR_ONLIST(list, frame, rfn, fop, args...) \
+ do { \
+ afr_local_t *__local = frame->local; \
+ afr_private_t *__priv = frame->this->private; \
+ int __i = 0; \
+ int __count = 0; \
+ unsigned char *__list = alloca(__priv->child_count); \
+ \
+ memcpy(__list, list, sizeof(*__list) * __priv->child_count); \
+ __count = AFR_COUNT(__list, __priv->child_count); \
+ __local->barrier.waitfor = __count; \
+ afr_local_replies_wipe(__local, __priv); \
+ \
+ for (__i = 0; __i < __priv->child_count; __i++) { \
+ if (!__list[__i]) \
+ continue; \
+ STACK_WIND_COOKIE(frame, rfn, (void *)(long)__i, \
+ __priv->children[__i], \
+ __priv->children[__i]->fops->fop, args); \
+ } \
+ syncbarrier_wait(&__local->barrier, __count); \
+ } while (0)
+
+#define AFR_SEQ(frame, rfn, fop, args...) \
+ do { \
+ afr_local_t *__local = frame->local; \
+ afr_private_t *__priv = frame->this->private; \
+ int __i = 0; \
+ \
+ afr_local_replies_wipe(__local, __priv); \
+ \
+ for (__i = 0; __i < __priv->child_count; __i++) { \
+ if (!__priv->child_up[__i]) \
+ continue; \
+ STACK_WIND_COOKIE(frame, rfn, (void *)(long)__i, \
+ __priv->children[__i], \
+ __priv->children[__i]->fops->fop, args); \
+ syncbarrier_wait(&__local->barrier, 1); \
+ } \
+ } while (0)
+
+#define ALLOC_MATRIX(n, type) \
+ ({ \
+ int __i; \
+ type **__ptr = alloca(n * sizeof(type *)); \
+ \
+ for (__i = 0; __i < n; __i++) \
+ __ptr[__i] = alloca0(n * sizeof(type)); \
+ __ptr; \
+ })
+
+#define IA_EQUAL(f, s, field) \
+ (memcmp(&(f.ia_##field), &(s.ia_##field), sizeof(s.ia_##field)) == 0)
+
+#define SBRAIN_HEAL_NO_GO_MSG \
+ "Failed to obtain replies from all bricks of " \
+ "the replica (are they up?). Cannot resolve split-brain."
+#define SFILE_NOT_IN_SPLIT_BRAIN "File not in split-brain"
+#define SNO_BIGGER_FILE "No bigger file"
+#define SNO_DIFF_IN_MTIME "No difference in mtime"
+#define SUSE_SOURCE_BRICK_TO_HEAL \
+ "Use source-brick option to heal metadata" \
+ " split-brain"
+#define SINVALID_BRICK_NAME "Invalid brick name"
+#define SBRICK_IS_NOT_UP "Brick is not up"
+#define SBRICK_NOT_CONNECTED "Brick is not connected"
+#define SLESS_THAN2_BRICKS_in_REP "< 2 bricks in replica are up"
+#define SBRICK_IS_REMOTE "Brick is remote"
+#define SSTARTED_SELF_HEAL "Started self-heal"
+#define SOP_NOT_SUPPORTED "Operation Not Supported"
+#define SFILE_NOT_UNDER_DATA \
+ "The file is not under data or metadata " \
+ "split-brain"
+#define SFILE_NOT_IN_SPLIT_BRAIN "File not in split-brain"
+#define SALL_BRICKS_UP_TO_RESOLVE \
+ "All the bricks should be up to resolve the" \
+ " gfid split brain"
+#define SERROR_GETTING_SRC_BRICK "Error getting the source brick"
+int
+afr_selfheal(xlator_t *this, uuid_t gfid);
gf_boolean_t
-afr_throttled_selfheal (call_frame_t *frame, xlator_t *this);
+afr_throttled_selfheal(call_frame_t *frame, xlator_t *this);
int
-afr_selfheal_name (xlator_t *this, uuid_t gfid, const char *name,
- void *gfid_req, dict_t *xdata);
+afr_selfheal_name(xlator_t *this, uuid_t gfid, const char *name, void *gfid_req,
+ dict_t *xdata);
int
-afr_selfheal_data (call_frame_t *frame, xlator_t *this, fd_t *fd);
+afr_selfheal_data(call_frame_t *frame, xlator_t *this, fd_t *fd);
int
-afr_selfheal_metadata (call_frame_t *frame, xlator_t *this, inode_t *inode);
+afr_selfheal_metadata(call_frame_t *frame, xlator_t *this, inode_t *inode);
int
-afr_selfheal_entry (call_frame_t *frame, xlator_t *this, inode_t *inode);
+afr_selfheal_entry(call_frame_t *frame, xlator_t *this, inode_t *inode);
int
-afr_lookup_and_heal_gfid (xlator_t *this, inode_t *parent, const char *name,
- inode_t *inode, struct afr_reply *replies, int source,
- void *gfid);
+afr_lookup_and_heal_gfid(xlator_t *this, inode_t *parent, const char *name,
+ inode_t *inode, struct afr_reply *replies, int source,
+ unsigned char *sources, void *gfid, int *gfid_idx);
int
-afr_selfheal_inodelk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, off_t off, size_t size,
- unsigned char *locked_on);
+afr_selfheal_inodelk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, off_t off, size_t size,
+ unsigned char *locked_on);
int
-afr_selfheal_tryinodelk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, off_t off, size_t size,
- unsigned char *locked_on);
+afr_selfheal_tryinodelk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, off_t off, size_t size,
+ unsigned char *locked_on);
int
-afr_selfheal_tie_breaker_inodelk (call_frame_t *frame, xlator_t *this,
- inode_t *inode, char *dom, off_t off,
- size_t size, unsigned char *locked_on);
+afr_selfheal_tie_breaker_inodelk(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, char *dom, off_t off,
+ size_t size, unsigned char *locked_on);
int
-afr_selfheal_uninodelk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, off_t off, size_t size,
- const unsigned char *locked_on);
+afr_selfheal_uninodelk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, off_t off, size_t size,
+ const unsigned char *locked_on);
int
-afr_selfheal_entrylk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, const char *name, unsigned char *locked_on);
+afr_selfheal_entrylk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, const char *name, unsigned char *locked_on);
int
-afr_selfheal_tryentrylk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, const char *name, unsigned char *locked_on);
+afr_selfheal_tryentrylk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, const char *name, unsigned char *locked_on);
int
-afr_selfheal_tie_breaker_entrylk (call_frame_t *frame, xlator_t *this,
- inode_t *inode, char *dom, const char *name,
- unsigned char *locked_on);
+afr_selfheal_tie_breaker_entrylk(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, char *dom, const char *name,
+ unsigned char *locked_on);
int
-afr_selfheal_unentrylk (call_frame_t *frame, xlator_t *this, inode_t *inode,
- char *dom, const char *name, unsigned char *locked_on,
- dict_t *xdata);
+afr_selfheal_unentrylk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ char *dom, const char *name, unsigned char *locked_on,
+ dict_t *xdata);
int
-afr_selfheal_unlocked_discover (call_frame_t *frame, inode_t *inode,
- uuid_t gfid, struct afr_reply *replies);
+afr_selfheal_unlocked_discover(call_frame_t *frame, inode_t *inode, uuid_t gfid,
+ struct afr_reply *replies);
int
-afr_selfheal_unlocked_discover_on (call_frame_t *frame, inode_t *inode,
+afr_selfheal_unlocked_discover_on(call_frame_t *frame, inode_t *inode,
uuid_t gfid, struct afr_reply *replies,
- unsigned char *discover_on);
+ unsigned char *discover_on, dict_t *dict);
inode_t *
-afr_selfheal_unlocked_lookup_on (call_frame_t *frame, inode_t *parent,
- const char *name, struct afr_reply *replies,
- unsigned char *lookup_on, dict_t *xattr);
+afr_selfheal_unlocked_lookup_on(call_frame_t *frame, inode_t *parent,
+ const char *name, struct afr_reply *replies,
+ unsigned char *lookup_on, dict_t *xattr);
int
-afr_selfheal_find_direction (call_frame_t *frame, xlator_t *this,
- struct afr_reply *replies,
- afr_transaction_type type,
- unsigned char *locked_on, unsigned char *sources,
- unsigned char *sinks, uint64_t *witness,
- gf_boolean_t *flag);
+afr_selfheal_find_direction(call_frame_t *frame, xlator_t *this,
+ struct afr_reply *replies,
+ afr_transaction_type type, unsigned char *locked_on,
+ unsigned char *sources, unsigned char *sinks,
+ uint64_t *witness, unsigned char *flag);
int
-afr_selfheal_fill_matrix (xlator_t *this, int **matrix, int subvol, int idx,
- dict_t *xdata);
+afr_selfheal_fill_matrix(xlator_t *this, int **matrix, int subvol, int idx,
+ dict_t *xdata);
int
-afr_selfheal_extract_xattr (xlator_t *this, struct afr_reply *replies,
- afr_transaction_type type, int *dirty, int **matrix);
+afr_selfheal_extract_xattr(xlator_t *this, struct afr_reply *replies,
+ afr_transaction_type type, int *dirty, int **matrix);
int
-afr_sh_generic_fop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *pre,
- struct iatt *post, dict_t *xdata);
+afr_sh_generic_fop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *pre,
+ struct iatt *post, dict_t *xdata);
int
-afr_selfheal_restore_time (call_frame_t *frame, xlator_t *this, inode_t *inode,
- int source, unsigned char *healed_sinks,
- struct afr_reply *replies);
+afr_selfheal_restore_time(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ int source, unsigned char *healed_sinks,
+ struct afr_reply *replies);
int
-afr_selfheal_undo_pending (call_frame_t *frame, xlator_t *this, inode_t *inode,
- unsigned char *sources, unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *undid_pending,
- afr_transaction_type type, struct afr_reply *replies,
- unsigned char *locked_on);
+afr_selfheal_undo_pending(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ unsigned char *sources, unsigned char *sinks,
+ unsigned char *healed_sinks,
+ unsigned char *undid_pending,
+ afr_transaction_type type, struct afr_reply *replies,
+ unsigned char *locked_on);
int
-afr_selfheal_recreate_entry (call_frame_t *frame, int dst, int source,
- unsigned char *sources,
- inode_t *dir, const char *name, inode_t *inode,
- struct afr_reply *replies);
+afr_selfheal_recreate_entry(call_frame_t *frame, int dst, int source,
+ unsigned char *sources, inode_t *dir,
+ const char *name, inode_t *inode,
+ struct afr_reply *replies);
int
-afr_selfheal_post_op (call_frame_t *frame, xlator_t *this, inode_t *inode,
- int subvol, dict_t *xattr, dict_t *xdata);
+afr_selfheal_post_op(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ int subvol, dict_t *xattr, dict_t *xdata);
call_frame_t *
-afr_frame_create (xlator_t *this, int32_t *op_errno);
+afr_frame_create(xlator_t *this, int32_t *op_errno);
inode_t *
-afr_inode_find (xlator_t *this, uuid_t gfid);
+afr_inode_find(xlator_t *this, uuid_t gfid);
int
-afr_selfheal_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata,
- struct iatt *parbuf);
+afr_selfheal_discover_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata, struct iatt *parbuf);
void
-afr_reply_copy (struct afr_reply *dst, struct afr_reply *src);
+afr_reply_copy(struct afr_reply *dst, struct afr_reply *src);
void
-afr_replies_copy (struct afr_reply *dst, struct afr_reply *src, int count);
+afr_replies_copy(struct afr_reply *dst, struct afr_reply *src, int count);
int
-afr_selfheal_newentry_mark (call_frame_t *frame, xlator_t *this, inode_t *inode,
- int source, struct afr_reply *replies,
- unsigned char *sources, unsigned char *newentry);
+afr_selfheal_newentry_mark(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ int source, struct afr_reply *replies,
+ unsigned char *sources, unsigned char *newentry);
unsigned int
-afr_success_count (struct afr_reply *replies, unsigned int count);
+afr_success_count(struct afr_reply *replies, unsigned int count);
void
-afr_log_selfheal (uuid_t gfid, xlator_t *this, int ret, char *type,
- int source, unsigned char *sources,
- unsigned char *healed_sinks);
+afr_log_selfheal(uuid_t gfid, xlator_t *this, int ret, char *type, int source,
+ unsigned char *sources, unsigned char *healed_sinks);
void
-afr_mark_largest_file_as_source (xlator_t *this, unsigned char *sources,
- struct afr_reply *replies);
+afr_mark_largest_file_as_source(xlator_t *this, unsigned char *sources,
+ struct afr_reply *replies);
void
-afr_mark_active_sinks (xlator_t *this, unsigned char *sources,
- unsigned char *locked_on, unsigned char *sinks);
+afr_mark_active_sinks(xlator_t *this, unsigned char *sources,
+ unsigned char *locked_on, unsigned char *sinks);
gf_boolean_t
-afr_dict_contains_heal_op (call_frame_t *frame);
+afr_dict_contains_heal_op(call_frame_t *frame);
gf_boolean_t
-afr_can_decide_split_brain_source_sinks (struct afr_reply *replies,
- int child_count);
+afr_can_decide_split_brain_source_sinks(struct afr_reply *replies,
+ int child_count);
int
-afr_mark_split_brain_source_sinks (call_frame_t *frame, xlator_t *this,
- inode_t *inode,
- unsigned char *sources,
- unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *locked_on,
- struct afr_reply *replies,
- afr_transaction_type type);
+afr_mark_split_brain_source_sinks(
+ call_frame_t *frame, xlator_t *this, inode_t *inode, unsigned char *sources,
+ unsigned char *sinks, unsigned char *healed_sinks, unsigned char *locked_on,
+ struct afr_reply *replies, afr_transaction_type type);
int
-afr_sh_get_fav_by_policy (xlator_t *this, struct afr_reply *replies,
- inode_t *inode, char **policy_str);
+afr_sh_get_fav_by_policy(xlator_t *this, struct afr_reply *replies,
+ inode_t *inode, char **policy_str);
int
-_afr_fav_child_reset_sink_xattrs (call_frame_t *frame, xlator_t *this,
+_afr_fav_child_reset_sink_xattrs(call_frame_t *frame, xlator_t *this,
inode_t *inode, int source,
unsigned char *healed_sinks,
unsigned char *undid_pending,
@@ -275,82 +295,83 @@ _afr_fav_child_reset_sink_xattrs (call_frame_t *frame, xlator_t *this,
struct afr_reply *replies);
int
-afr_get_child_index_from_name (xlator_t *this, char *name);
+afr_get_child_index_from_name(xlator_t *this, char *name);
gf_boolean_t
-afr_does_witness_exist (xlator_t *this, uint64_t *witness);
+afr_does_witness_exist(xlator_t *this, uint64_t *witness);
int
-__afr_selfheal_data_prepare (call_frame_t *frame, xlator_t *this,
- inode_t *inode, unsigned char *locked_on,
- unsigned char *sources,
- unsigned char *sinks, unsigned char *healed_sinks,
- unsigned char *undid_pending,
- struct afr_reply *replies, gf_boolean_t *flag);
+__afr_selfheal_data_prepare(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ unsigned char *locked_on, unsigned char *sources,
+ unsigned char *sinks, unsigned char *healed_sinks,
+ unsigned char *undid_pending,
+ struct afr_reply *replies, unsigned char *flag);
int
-__afr_selfheal_metadata_prepare (call_frame_t *frame, xlator_t *this,
- inode_t *inode, unsigned char *locked_on,
- unsigned char *sources,
- unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *undid_pending,
- struct afr_reply *replies,
- gf_boolean_t *flag);
+__afr_selfheal_metadata_prepare(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, unsigned char *locked_on,
+ unsigned char *sources, unsigned char *sinks,
+ unsigned char *healed_sinks,
+ unsigned char *undid_pending,
+ struct afr_reply *replies, unsigned char *flag);
int
-__afr_selfheal_entry_prepare (call_frame_t *frame, xlator_t *this,
- inode_t *inode, unsigned char *locked_on,
- unsigned char *sources,
- unsigned char *sinks,
- unsigned char *healed_sinks,
- struct afr_reply *replies, int *source_p,
- gf_boolean_t *flag);
+__afr_selfheal_entry_prepare(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, unsigned char *locked_on,
+ unsigned char *sources, unsigned char *sinks,
+ unsigned char *healed_sinks,
+ struct afr_reply *replies, int *source_p,
+ unsigned char *flag);
int
-afr_selfheal_unlocked_inspect (call_frame_t *frame, xlator_t *this,
- uuid_t gfid, inode_t **link_inode,
- gf_boolean_t *data_selfheal,
- gf_boolean_t *metadata_selfheal,
- gf_boolean_t *entry_selfheal);
+afr_selfheal_unlocked_inspect(call_frame_t *frame, xlator_t *this, uuid_t gfid,
+ inode_t **link_inode, gf_boolean_t *data_selfheal,
+ gf_boolean_t *metadata_selfheal,
+ gf_boolean_t *entry_selfheal,
+ struct afr_reply *replies);
int
-afr_selfheal_do (call_frame_t *frame, xlator_t *this, uuid_t gfid);
+afr_selfheal_do(call_frame_t *frame, xlator_t *this, uuid_t gfid);
int
-afr_selfheal_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata);
+afr_selfheal_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata);
int
-afr_locked_fill (call_frame_t *frame, xlator_t *this,
- unsigned char *locked_on);
+afr_locked_fill(call_frame_t *frame, xlator_t *this, unsigned char *locked_on);
int
-afr_choose_source_by_policy (afr_private_t *priv, unsigned char *sources,
- afr_transaction_type type);
+afr_choose_source_by_policy(afr_private_t *priv, unsigned char *sources,
+ afr_transaction_type type);
int
-afr_selfheal_metadata_by_stbuf (xlator_t *this, struct iatt *stbuf);
+afr_selfheal_metadata_by_stbuf(xlator_t *this, struct iatt *stbuf);
int
-afr_sh_fav_by_size (xlator_t *this, struct afr_reply *replies,
- inode_t *inode);
+afr_sh_fav_by_size(xlator_t *this, struct afr_reply *replies, inode_t *inode);
int
-afr_sh_fav_by_mtime (xlator_t *this, struct afr_reply *replies,
- inode_t *inode);
+afr_sh_fav_by_mtime(xlator_t *this, struct afr_reply *replies, inode_t *inode);
int
-afr_sh_fav_by_ctime (xlator_t *this, struct afr_reply *replies,
- inode_t *inode);
+afr_sh_fav_by_ctime(xlator_t *this, struct afr_reply *replies, inode_t *inode);
int
-afr_gfid_split_brain_source (xlator_t *this, struct afr_reply *replies,
- inode_t *inode, uuid_t pargfid, const char *bname,
- int src_idx, int child_idx,
- unsigned char *locked_on, int *src, dict_t *xdata);
+afr_gfid_split_brain_source(xlator_t *this, struct afr_reply *replies,
+ inode_t *inode, uuid_t pargfid, const char *bname,
+ int src_idx, int child_idx,
+ unsigned char *locked_on, int *src, dict_t *xdata);
int
-afr_mark_source_sinks_if_file_empty (xlator_t *this, unsigned char *sources,
- unsigned char *sinks,
- unsigned char *healed_sinks,
- unsigned char *locked_on,
- struct afr_reply *replies,
- afr_transaction_type type);
+afr_mark_source_sinks_if_file_empty(xlator_t *this, unsigned char *sources,
+ unsigned char *sinks,
+ unsigned char *healed_sinks,
+ unsigned char *locked_on,
+ struct afr_reply *replies,
+ afr_transaction_type type);
+
+gf_boolean_t
+afr_is_file_empty_on_all_children(afr_private_t *priv,
+ struct afr_reply *replies);
+int
+afr_selfheal_entry_delete(xlator_t *this, inode_t *dir, const char *name,
+ inode_t *inode, int child, struct afr_reply *replies);
+int
+afr_anon_inode_create(xlator_t *this, int child, inode_t **linked_inode);
#endif /* !_AFR_SELFHEAL_H */
diff --git a/xlators/cluster/afr/src/afr-self-heald.c b/xlators/cluster/afr/src/afr-self-heald.c
index 4667be63220..109fd4b7421 100644
--- a/xlators/cluster/afr/src/afr-self-heald.c
+++ b/xlators/cluster/afr/src/afr-self-heald.c
@@ -8,1387 +8,1709 @@
cases as published by the Free Software Foundation.
*/
-
#include "afr.h"
#include "afr-self-heal.h"
#include "afr-self-heald.h"
#include "protocol-common.h"
-#include "syncop-utils.h"
+#include <glusterfs/syncop-utils.h>
#include "afr-messages.h"
-#include "byte-order.h"
-
-#define SHD_INODE_LRU_LIMIT 2048
-#define AFR_EH_SPLIT_BRAIN_LIMIT 1024
-#define AFR_STATISTICS_HISTORY_SIZE 50
-
-
-#define ASSERT_LOCAL(this, healer) \
- if (!afr_shd_is_subvol_local(this, healer->subvol)) { \
- healer->local = _gf_false; \
- if (safe_break (healer)) { \
- break; \
- } else { \
- continue; \
- } \
- } else { \
- healer->local = _gf_true; \
- }
-
-
-#define NTH_INDEX_HEALER(this, n) &((((afr_private_t *)this->private))->shd.index_healers[n])
-#define NTH_FULL_HEALER(this, n) &((((afr_private_t *)this->private))->shd.full_healers[n])
+#include <glusterfs/byte-order.h>
+
+#define AFR_EH_SPLIT_BRAIN_LIMIT 1024
+#define AFR_STATISTICS_HISTORY_SIZE 50
+
+#define ASSERT_LOCAL(this, healer) \
+ if (!afr_shd_is_subvol_local(this, healer->subvol)) { \
+ healer->local = _gf_false; \
+ if (safe_break(healer)) { \
+ break; \
+ } else { \
+ continue; \
+ } \
+ } else { \
+ healer->local = _gf_true; \
+ }
+
+#define NTH_INDEX_HEALER(this, n) \
+ &((((afr_private_t *)this->private))->shd.index_healers[n])
+#define NTH_FULL_HEALER(this, n) \
+ &((((afr_private_t *)this->private))->shd.full_healers[n])
char *
-afr_subvol_name (xlator_t *this, int subvol)
+afr_subvol_name(xlator_t *this, int subvol)
{
- afr_private_t *priv = NULL;
+ afr_private_t *priv = NULL;
- priv = this->private;
- if (subvol < 0 || subvol > priv->child_count)
- return NULL;
+ priv = this->private;
+ if (subvol < 0 || subvol > priv->child_count)
+ return NULL;
- return priv->children[subvol]->name;
+ return priv->children[subvol]->name;
}
-
void
-afr_destroy_crawl_event_data (void *data)
+afr_destroy_crawl_event_data(void *data)
{
- return;
+ return;
}
-
void
-afr_destroy_shd_event_data (void *data)
+afr_destroy_shd_event_data(void *data)
{
- shd_event_t *shd_event = data;
-
- if (!shd_event)
- return;
- GF_FREE (shd_event->path);
+ shd_event_t *shd_event = data;
+ if (!shd_event)
return;
-}
+ GF_FREE(shd_event->path);
+ return;
+}
gf_boolean_t
-afr_shd_is_subvol_local (xlator_t *this, int subvol)
+afr_shd_is_subvol_local(xlator_t *this, int subvol)
{
- afr_private_t *priv = NULL;
- gf_boolean_t is_local = _gf_false;
- loc_t loc = {0, };
-
- loc.inode = this->itable->root;
- gf_uuid_copy (loc.gfid, loc.inode->gfid);
- priv = this->private;
- syncop_is_subvol_local(priv->children[subvol], &loc, &is_local);
- return is_local;
+ afr_private_t *priv = NULL;
+ gf_boolean_t is_local = _gf_false;
+ loc_t loc = {
+ 0,
+ };
+
+ loc.inode = this->itable->root;
+ gf_uuid_copy(loc.gfid, loc.inode->gfid);
+ priv = this->private;
+ syncop_is_subvol_local(priv->children[subvol], &loc, &is_local);
+ return is_local;
}
-
int
-__afr_shd_healer_wait (struct subvol_healer *healer)
+__afr_shd_healer_wait(struct subvol_healer *healer)
{
- afr_private_t *priv = NULL;
- struct timespec wait_till = {0, };
- int ret = 0;
+ afr_private_t *priv = NULL;
+ struct timespec wait_till = {
+ 0,
+ };
+ int ret = 0;
- priv = healer->this->private;
+ priv = healer->this->private;
disabled_loop:
- wait_till.tv_sec = time (NULL) + priv->shd.timeout;
+ wait_till.tv_sec = gf_time() + priv->shd.timeout;
- while (!healer->rerun) {
- ret = pthread_cond_timedwait (&healer->cond,
- &healer->mutex,
- &wait_till);
- if (ret == ETIMEDOUT)
- break;
- }
+ while (!healer->rerun) {
+ ret = pthread_cond_timedwait(&healer->cond, &healer->mutex, &wait_till);
+ if (ret == ETIMEDOUT)
+ break;
+ }
- ret = healer->rerun;
- healer->rerun = 0;
+ ret = healer->rerun;
+ healer->rerun = 0;
- if (!priv->shd.enabled)
- goto disabled_loop;
+ if (!priv->shd.enabled)
+ goto disabled_loop;
- return ret;
+ return ret;
}
-
int
-afr_shd_healer_wait (struct subvol_healer *healer)
+afr_shd_healer_wait(struct subvol_healer *healer)
{
- int ret = 0;
+ int ret = 0;
- pthread_mutex_lock (&healer->mutex);
- {
- ret = __afr_shd_healer_wait (healer);
- }
- pthread_mutex_unlock (&healer->mutex);
+ pthread_mutex_lock(&healer->mutex);
+ {
+ ret = __afr_shd_healer_wait(healer);
+ }
+ pthread_mutex_unlock(&healer->mutex);
- return ret;
+ return ret;
}
-
gf_boolean_t
-safe_break (struct subvol_healer *healer)
+safe_break(struct subvol_healer *healer)
{
- gf_boolean_t ret = _gf_false;
+ gf_boolean_t ret = _gf_false;
- pthread_mutex_lock (&healer->mutex);
- {
- if (healer->rerun)
- goto unlock;
+ pthread_mutex_lock(&healer->mutex);
+ {
+ if (healer->rerun)
+ goto unlock;
- healer->running = _gf_false;
- ret = _gf_true;
- }
+ healer->running = _gf_false;
+ ret = _gf_true;
+ }
unlock:
- pthread_mutex_unlock (&healer->mutex);
+ pthread_mutex_unlock(&healer->mutex);
- return ret;
+ return ret;
}
-
inode_t *
-afr_shd_inode_find (xlator_t *this, xlator_t *subvol, uuid_t gfid)
+afr_shd_inode_find(xlator_t *this, xlator_t *subvol, uuid_t gfid)
{
- int ret = 0;
- uint64_t val = IA_INVAL;
- dict_t *xdata = NULL;
- dict_t *rsp_dict = NULL;
- inode_t *inode = NULL;
-
- xdata = dict_new ();
- if (!xdata)
- goto out;
-
- ret = dict_set_int8 (xdata, GF_INDEX_IA_TYPE_GET_REQ, 1);
+ int ret = 0;
+ uint64_t val = IA_INVAL;
+ dict_t *xdata = NULL;
+ dict_t *rsp_dict = NULL;
+ inode_t *inode = NULL;
+
+ xdata = dict_new();
+ if (!xdata)
+ goto out;
+
+ ret = dict_set_int8(xdata, GF_INDEX_IA_TYPE_GET_REQ, 1);
+ if (ret)
+ goto out;
+
+ ret = syncop_inode_find(this, subvol, gfid, &inode, xdata, &rsp_dict);
+ if (ret < 0)
+ goto out;
+
+ if (rsp_dict) {
+ ret = dict_get_uint64(rsp_dict, GF_INDEX_IA_TYPE_GET_RSP, &val);
if (ret)
- goto out;
-
- ret = syncop_inode_find (this, subvol, gfid, &inode,
- xdata, &rsp_dict);
- if (ret < 0)
- goto out;
-
- if (rsp_dict) {
- ret = dict_get_uint64 (rsp_dict, GF_INDEX_IA_TYPE_GET_RSP,
- &val);
- if (ret)
- goto out;
- }
- ret = inode_ctx_set2 (inode, subvol, 0, &val);
+ goto out;
+ }
+ ret = inode_ctx_set2(inode, subvol, 0, &val);
out:
- if (ret && inode) {
- inode_unref (inode);
- inode = NULL;
- }
- if (xdata)
- dict_unref (xdata);
- if (rsp_dict)
- dict_unref (rsp_dict);
- return inode;
+ if (ret && inode) {
+ inode_unref(inode);
+ inode = NULL;
+ }
+ if (xdata)
+ dict_unref(xdata);
+ if (rsp_dict)
+ dict_unref(rsp_dict);
+ return inode;
}
-inode_t*
-afr_shd_index_inode (xlator_t *this, xlator_t *subvol, char *vgfid)
+inode_t *
+afr_shd_index_inode(xlator_t *this, xlator_t *subvol, char *vgfid)
{
- loc_t rootloc = {0, };
- inode_t *inode = NULL;
- int ret = 0;
- dict_t *xattr = NULL;
- void *index_gfid = NULL;
+ loc_t rootloc = {
+ 0,
+ };
+ inode_t *inode = NULL;
+ int ret = 0;
+ dict_t *xattr = NULL;
+ void *index_gfid = NULL;
- rootloc.inode = inode_ref (this->itable->root);
- gf_uuid_copy (rootloc.gfid, rootloc.inode->gfid);
+ rootloc.inode = inode_ref(this->itable->root);
+ gf_uuid_copy(rootloc.gfid, rootloc.inode->gfid);
- ret = syncop_getxattr (subvol, &rootloc, &xattr,
- vgfid, NULL, NULL);
- if (ret || !xattr) {
- errno = -ret;
- goto out;
- }
+ ret = syncop_getxattr(subvol, &rootloc, &xattr, vgfid, NULL, NULL);
+ if (ret || !xattr) {
+ errno = -ret;
+ goto out;
+ }
- ret = dict_get_ptr (xattr, vgfid, &index_gfid);
- if (ret)
- goto out;
+ ret = dict_get_ptr(xattr, vgfid, &index_gfid);
+ if (ret)
+ goto out;
- gf_msg_debug (this->name, 0, "%s dir gfid for %s: %s",
- vgfid, subvol->name, uuid_utoa (index_gfid));
+ gf_msg_debug(this->name, 0, "%s dir gfid for %s: %s", vgfid, subvol->name,
+ uuid_utoa(index_gfid));
- inode = afr_shd_inode_find (this, subvol, index_gfid);
+ inode = afr_shd_inode_find(this, subvol, index_gfid);
out:
- loc_wipe (&rootloc);
+ loc_wipe(&rootloc);
- if (xattr)
- dict_unref (xattr);
+ if (xattr)
+ dict_unref(xattr);
- return inode;
+ return inode;
}
int
-afr_shd_index_purge (xlator_t *subvol, inode_t *inode, char *name,
- ia_type_t type)
+afr_shd_entry_purge(xlator_t *subvol, inode_t *inode, char *name,
+ ia_type_t type)
{
- int ret = 0;
- loc_t loc = {0,};
+ int ret = 0;
+ loc_t loc = {
+ 0,
+ };
- loc.parent = inode_ref (inode);
- loc.name = name;
+ loc.parent = inode_ref(inode);
+ loc.name = name;
- if (IA_ISDIR (type))
- ret = syncop_rmdir (subvol, &loc, 1, NULL, NULL);
- else
- ret = syncop_unlink (subvol, &loc, NULL, NULL);
+ if (IA_ISDIR(type))
+ ret = syncop_rmdir(subvol, &loc, 1, NULL, NULL);
+ else
+ ret = syncop_unlink(subvol, &loc, NULL, NULL);
- loc_wipe (&loc);
- return ret;
+ loc_wipe(&loc);
+ return ret;
}
void
-afr_shd_zero_xattrop (xlator_t *this, uuid_t gfid)
+afr_shd_zero_xattrop(xlator_t *this, uuid_t gfid)
{
-
- call_frame_t *frame = NULL;
- inode_t *inode = NULL;
- afr_private_t *priv = NULL;
- dict_t *xattr = NULL;
- int ret = 0;
- int i = 0;
- int raw[AFR_NUM_CHANGE_LOGS] = {0};
-
- priv = this->private;
- frame = afr_frame_create (this, NULL);
- if (!frame)
- goto out;
- inode = afr_inode_find (this, gfid);
- if (!inode)
- goto out;
- xattr = dict_new();
- if (!xattr)
- goto out;
- ret = dict_set_static_bin (xattr, AFR_DIRTY, raw,
- sizeof(int) * AFR_NUM_CHANGE_LOGS);
+ call_frame_t *frame = NULL;
+ inode_t *inode = NULL;
+ afr_private_t *priv = NULL;
+ dict_t *xattr = NULL;
+ int ret = 0;
+ int i = 0;
+ int raw[AFR_NUM_CHANGE_LOGS] = {0};
+
+ priv = this->private;
+ frame = afr_frame_create(this, NULL);
+ if (!frame)
+ goto out;
+ inode = afr_inode_find(this, gfid);
+ if (!inode)
+ goto out;
+ xattr = dict_new();
+ if (!xattr)
+ goto out;
+ ret = dict_set_static_bin(xattr, AFR_DIRTY, raw,
+ sizeof(int) * AFR_NUM_CHANGE_LOGS);
+ if (ret)
+ goto out;
+ for (i = 0; i < priv->child_count; i++) {
+ ret = dict_set_static_bin(xattr, priv->pending_key[i], raw,
+ sizeof(int) * AFR_NUM_CHANGE_LOGS);
if (ret)
- goto out;
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_set_static_bin (xattr, priv->pending_key[i], raw,
- sizeof(int) * AFR_NUM_CHANGE_LOGS);
- if (ret)
- goto out;
- }
+ goto out;
+ }
- /*Send xattrop to all bricks. Doing a lookup to see if bricks are up or
- * has valid repies for this gfid seems a bit of an overkill.*/
- for (i = 0; i < priv->child_count; i++)
- afr_selfheal_post_op (frame, this, inode, i, xattr, NULL);
+ /*Send xattrop to all bricks. Doing a lookup to see if bricks are up or
+ * has valid repies for this gfid seems a bit of an overkill.*/
+ for (i = 0; i < priv->child_count; i++)
+ afr_selfheal_post_op(frame, this, inode, i, xattr, NULL);
out:
- if (frame)
- AFR_STACK_DESTROY (frame);
- if (inode)
- inode_unref (inode);
- if (xattr)
- dict_unref (xattr);
- return;
+ if (frame)
+ AFR_STACK_DESTROY(frame);
+ if (inode)
+ inode_unref(inode);
+ if (xattr)
+ dict_unref(xattr);
+ return;
}
int
-afr_shd_selfheal_name (struct subvol_healer *healer, int child, uuid_t parent,
- const char *bname)
+afr_shd_selfheal_name(struct subvol_healer *healer, int child, uuid_t parent,
+ const char *bname)
{
- int ret = -1;
+ int ret = -1;
- ret = afr_selfheal_name (THIS, parent, bname, NULL, NULL);
+ ret = afr_selfheal_name(THIS, parent, bname, NULL, NULL);
- return ret;
+ return ret;
}
int
-afr_shd_selfheal (struct subvol_healer *healer, int child, uuid_t gfid)
+afr_shd_selfheal(struct subvol_healer *healer, int child, uuid_t gfid)
{
- int ret = 0;
- eh_t *eh = NULL;
- afr_private_t *priv = NULL;
- afr_self_heald_t *shd = NULL;
- shd_event_t *shd_event = NULL;
- char *path = NULL;
- xlator_t *subvol = NULL;
- xlator_t *this = NULL;
- crawl_event_t *crawl_event = NULL;
-
- this = healer->this;
- priv = this->private;
- shd = &priv->shd;
- crawl_event = &healer->crawl_event;
-
- subvol = priv->children[child];
-
- //If this fails with ENOENT/ESTALE index is stale
- ret = syncop_gfid_to_path (this->itable, subvol, gfid, &path);
- if (ret < 0)
- return ret;
-
- ret = afr_selfheal (this, gfid);
-
- LOCK (&priv->lock);
- {
- if (ret == -EIO) {
- eh = shd->split_brain;
- crawl_event->split_brain_count++;
- } else if (ret < 0) {
- crawl_event->heal_failed_count++;
- } else if (ret == 0) {
- crawl_event->healed_count++;
- }
+ int ret = 0;
+ eh_t *eh = NULL;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ shd_event_t *shd_event = NULL;
+ char *path = NULL;
+ xlator_t *subvol = NULL;
+ xlator_t *this = NULL;
+ crawl_event_t *crawl_event = NULL;
+
+ this = healer->this;
+ priv = this->private;
+ shd = &priv->shd;
+ crawl_event = &healer->crawl_event;
+
+ subvol = priv->children[child];
+
+ // If this fails with ENOENT/ESTALE index is stale
+ ret = syncop_gfid_to_path(this->itable, subvol, gfid, &path);
+ if (ret < 0)
+ return ret;
+
+ ret = afr_selfheal(this, gfid);
+
+ LOCK(&priv->lock);
+ {
+ if (ret == -EIO) {
+ eh = shd->split_brain;
+ crawl_event->split_brain_count++;
+ } else if (ret < 0) {
+ crawl_event->heal_failed_count++;
+ } else if (ret == 0) {
+ crawl_event->healed_count++;
}
- UNLOCK (&priv->lock);
+ }
+ UNLOCK(&priv->lock);
- if (eh) {
- shd_event = GF_CALLOC (1, sizeof(*shd_event),
- gf_afr_mt_shd_event_t);
- if (!shd_event)
- goto out;
+ if (eh) {
+ shd_event = GF_CALLOC(1, sizeof(*shd_event), gf_afr_mt_shd_event_t);
+ if (!shd_event)
+ goto out;
- shd_event->child = child;
- shd_event->path = path;
+ shd_event->child = child;
+ shd_event->path = path;
- if (eh_save_history (eh, shd_event) < 0)
- goto out;
+ if (eh_save_history(eh, shd_event) < 0)
+ goto out;
- shd_event = NULL;
- path = NULL;
- }
+ shd_event = NULL;
+ path = NULL;
+ }
out:
- GF_FREE (shd_event);
- GF_FREE (path);
- return ret;
+ GF_FREE(shd_event);
+ GF_FREE(path);
+ return ret;
}
-
void
-afr_shd_sweep_prepare (struct subvol_healer *healer)
+afr_shd_sweep_prepare(struct subvol_healer *healer)
{
- crawl_event_t *event = NULL;
+ crawl_event_t *event = NULL;
- event = &healer->crawl_event;
+ event = &healer->crawl_event;
- event->healed_count = 0;
- event->split_brain_count = 0;
- event->heal_failed_count = 0;
+ event->healed_count = 0;
+ event->split_brain_count = 0;
+ event->heal_failed_count = 0;
- time (&event->start_time);
- event->end_time = 0;
+ event->start_time = gf_time();
+ event->end_time = 0;
+ _mask_cancellation();
}
-
void
-afr_shd_sweep_done (struct subvol_healer *healer)
+afr_shd_sweep_done(struct subvol_healer *healer)
{
- crawl_event_t *event = NULL;
- crawl_event_t *history = NULL;
- afr_self_heald_t *shd = NULL;
+ crawl_event_t *event = NULL;
+ crawl_event_t *history = NULL;
+ afr_self_heald_t *shd = NULL;
- event = &healer->crawl_event;
- shd = &(((afr_private_t *)healer->this->private)->shd);
+ event = &healer->crawl_event;
+ shd = &(((afr_private_t *)healer->this->private)->shd);
- time (&event->end_time);
- history = memdup (event, sizeof (*event));
- event->start_time = 0;
+ event->end_time = gf_time();
+ history = gf_memdup(event, sizeof(*event));
+ event->start_time = 0;
- if (!history)
- return;
+ if (!history)
+ return;
- if (eh_save_history (shd->statistics[healer->subvol], history) < 0)
- GF_FREE (history);
+ if (eh_save_history(shd->statistics[healer->subvol], history) < 0)
+ GF_FREE(history);
+ _unmask_cancellation();
}
int
-afr_shd_index_heal (xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
- void *data)
+afr_shd_index_heal(xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
+ void *data)
{
- struct subvol_healer *healer = data;
- afr_private_t *priv = NULL;
- uuid_t gfid = {0};
- int ret = 0;
- uint64_t val = IA_INVAL;
+ struct subvol_healer *healer = data;
+ afr_private_t *priv = NULL;
+ uuid_t gfid = {0};
+ int ret = 0;
+ uint64_t val = IA_INVAL;
- priv = healer->this->private;
- if (!priv->shd.enabled)
- return -EBUSY;
+ priv = healer->this->private;
+ if (!priv->shd.enabled)
+ return -EBUSY;
- gf_msg_debug (healer->this->name, 0, "got entry: %s from %s",
- entry->d_name, priv->children[healer->subvol]->name);
+ gf_msg_debug(healer->this->name, 0, "got entry: %s from %s", entry->d_name,
+ priv->children[healer->subvol]->name);
- ret = gf_uuid_parse (entry->d_name, gfid);
- if (ret)
- return 0;
+ ret = gf_uuid_parse(entry->d_name, gfid);
+ if (ret)
+ return 0;
- inode_ctx_get2 (parent->inode, subvol, NULL, &val);
+ inode_ctx_get2(parent->inode, subvol, NULL, &val);
- ret = afr_shd_selfheal (healer, healer->subvol, gfid);
+ ret = afr_shd_selfheal(healer, healer->subvol, gfid);
- if (ret == -ENOENT || ret == -ESTALE)
- afr_shd_index_purge (subvol, parent->inode, entry->d_name, val);
+ if (ret == -ENOENT || ret == -ESTALE)
+ afr_shd_entry_purge(subvol, parent->inode, entry->d_name, val);
- if (ret == 2)
- /* If bricks crashed in pre-op after creating indices/xattrop
- * link but before setting afr changelogs, we end up with stale
- * xattrop links but zero changelogs. Remove such entries by
- * sending a post-op with zero changelogs.
- */
- afr_shd_zero_xattrop (healer->this, gfid);
+ if (ret == 2)
+ /* If bricks crashed in pre-op after creating indices/xattrop
+ * link but before setting afr changelogs, we end up with stale
+ * xattrop links but zero changelogs. Remove such entries by
+ * sending a post-op with zero changelogs.
+ */
+ afr_shd_zero_xattrop(healer->this, gfid);
- return 0;
+ return 0;
}
int
-afr_shd_index_sweep (struct subvol_healer *healer, char *vgfid)
+afr_shd_index_sweep(struct subvol_healer *healer, char *vgfid)
{
- loc_t loc = {0};
- afr_private_t *priv = NULL;
- int ret = 0;
- xlator_t *subvol = NULL;
- dict_t *xdata = NULL;
- call_frame_t *frame = NULL;
-
- priv = healer->this->private;
- subvol = priv->children[healer->subvol];
-
- frame = afr_frame_create (healer->this, &ret);
- if (!frame) {
- ret = -ret;
- goto out;
- }
+ loc_t loc = {0};
+ afr_private_t *priv = NULL;
+ int ret = 0;
+ xlator_t *subvol = NULL;
+ dict_t *xdata = NULL;
+ call_frame_t *frame = NULL;
+
+ priv = healer->this->private;
+ subvol = priv->children[healer->subvol];
+
+ frame = afr_frame_create(healer->this, &ret);
+ if (!frame) {
+ ret = -ret;
+ goto out;
+ }
+
+ loc.inode = afr_shd_index_inode(healer->this, subvol, vgfid);
+ if (!loc.inode) {
+ gf_msg(healer->this->name, GF_LOG_WARNING, 0,
+ AFR_MSG_INDEX_DIR_GET_FAILED, "unable to get index-dir on %s",
+ subvol->name);
+ ret = -errno;
+ goto out;
+ }
+
+ xdata = dict_new();
+ if (!xdata || dict_set_int32_sizen(xdata, "get-gfid-type", 1)) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ ret = syncop_mt_dir_scan(frame, subvol, &loc, GF_CLIENT_PID_SELF_HEALD,
+ healer, afr_shd_index_heal, xdata,
+ priv->shd.max_threads, priv->shd.wait_qlength);
+
+ if (ret == 0)
+ ret = healer->crawl_event.healed_count;
- loc.inode = afr_shd_index_inode (healer->this, subvol, vgfid);
- if (!loc.inode) {
- gf_msg (healer->this->name, GF_LOG_WARNING,
- 0, AFR_MSG_INDEX_DIR_GET_FAILED,
- "unable to get index-dir on %s", subvol->name);
- ret = -errno;
- goto out;
- }
-
- xdata = dict_new ();
- if (!xdata || dict_set_int32 (xdata, "get-gfid-type", 1)) {
- ret = -ENOMEM;
- goto out;
- }
-
- ret = syncop_mt_dir_scan (frame, subvol, &loc, GF_CLIENT_PID_SELF_HEALD,
- healer, afr_shd_index_heal, xdata,
- priv->shd.max_threads, priv->shd.wait_qlength);
+out:
+ loc_wipe(&loc);
- if (ret == 0)
- ret = healer->crawl_event.healed_count;
+ if (xdata)
+ dict_unref(xdata);
+ if (frame)
+ AFR_STACK_DESTROY(frame);
+ return ret;
+}
+int
+afr_shd_index_sweep_all(struct subvol_healer *healer)
+{
+ int ret = 0;
+ int count = 0;
+
+ ret = afr_shd_index_sweep(healer, GF_XATTROP_INDEX_GFID);
+ if (ret < 0)
+ goto out;
+ count = ret;
+
+ ret = afr_shd_index_sweep(healer, GF_XATTROP_DIRTY_GFID);
+ if (ret < 0)
+ goto out;
+ count += ret;
+
+ ret = afr_shd_index_sweep(healer, GF_XATTROP_ENTRY_CHANGES_GFID);
+ if (ret < 0)
+ goto out;
+ count += ret;
out:
- loc_wipe (&loc);
-
- if (xdata)
- dict_unref (xdata);
- if (frame)
- AFR_STACK_DESTROY (frame);
- return ret;
+ if (ret < 0)
+ return ret;
+ else
+ return count;
}
int
-afr_shd_index_sweep_all (struct subvol_healer *healer)
+afr_shd_full_heal(xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
+ void *data)
{
- int ret = 0;
- int count = 0;
+ struct subvol_healer *healer = data;
+ xlator_t *this = healer->this;
+ afr_private_t *priv = NULL;
- ret = afr_shd_index_sweep (healer, GF_XATTROP_INDEX_GFID);
- if (ret < 0)
- goto out;
- count = ret;
+ priv = this->private;
- ret = afr_shd_index_sweep (healer, GF_XATTROP_DIRTY_GFID);
- if (ret < 0)
- goto out;
- count += ret;
+ if (this->cleanup_starting) {
+ return -ENOTCONN;
+ }
- ret = afr_shd_index_sweep (healer, GF_XATTROP_ENTRY_CHANGES_GFID);
- if (ret < 0)
- goto out;
- count += ret;
-out:
- if (ret < 0)
- return ret;
- else
- return count;
+ if (!priv->shd.enabled)
+ return -EBUSY;
+
+ afr_shd_selfheal_name(healer, healer->subvol, parent->inode->gfid,
+ entry->d_name);
+
+ afr_shd_selfheal(healer, healer->subvol, entry->d_stat.ia_gfid);
+
+ return 0;
}
int
-afr_shd_full_heal (xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
- void *data)
+afr_shd_full_sweep(struct subvol_healer *healer, inode_t *inode)
{
- struct subvol_healer *healer = data;
- xlator_t *this = healer->this;
- afr_private_t *priv = NULL;
+ afr_private_t *priv = NULL;
+ loc_t loc = {0};
- priv = this->private;
- if (!priv->shd.enabled)
- return -EBUSY;
+ priv = healer->this->private;
+ loc.inode = inode;
+ return syncop_ftw(priv->children[healer->subvol], &loc,
+ GF_CLIENT_PID_SELF_HEALD, healer, afr_shd_full_heal);
+}
- afr_shd_selfheal_name (healer, healer->subvol,
- parent->inode->gfid, entry->d_name);
+int
+afr_shd_fill_ta_loc(xlator_t *this, loc_t *loc)
+{
+ afr_private_t *priv = NULL;
+ struct iatt stbuf = {
+ 0,
+ };
+ int ret = -1;
+
+ priv = this->private;
+ loc->parent = inode_ref(this->itable->root);
+ gf_uuid_copy(loc->pargfid, loc->parent->gfid);
+ loc->name = priv->pending_key[THIN_ARBITER_BRICK_INDEX];
+ loc->inode = inode_new(loc->parent->table);
+ GF_CHECK_ALLOC(loc->inode, ret, out);
+
+ if (!gf_uuid_is_null(priv->ta_gfid))
+ goto assign_gfid;
+
+ ret = syncop_lookup(priv->children[THIN_ARBITER_BRICK_INDEX], loc, &stbuf,
+ 0, 0, 0);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed lookup on file %s.", loc->name);
+ goto out;
+ }
+
+ gf_uuid_copy(priv->ta_gfid, stbuf.ia_gfid);
+
+assign_gfid:
+ gf_uuid_copy(loc->gfid, priv->ta_gfid);
+ ret = 0;
- afr_shd_selfheal (healer, healer->subvol, entry->d_stat.ia_gfid);
+out:
+ if (ret)
+ loc_wipe(loc);
- return 0;
+ return ret;
}
int
-afr_shd_full_sweep (struct subvol_healer *healer, inode_t *inode)
+_afr_shd_ta_get_xattrs(xlator_t *this, loc_t *loc, dict_t **xdata)
{
- afr_private_t *priv = NULL;
- loc_t loc = {0};
-
- priv = healer->this->private;
- loc.inode = inode;
- return syncop_ftw (priv->children[healer->subvol], &loc,
- GF_CLIENT_PID_SELF_HEALD, healer,
- afr_shd_full_heal);
+ afr_private_t *priv = NULL;
+ dict_t *xattr = NULL;
+ int raw[AFR_NUM_CHANGE_LOGS] = {
+ 0,
+ };
+ int ret = -1;
+ int i = 0;
+
+ priv = this->private;
+
+ xattr = dict_new();
+ if (!xattr) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, AFR_MSG_DICT_GET_FAILED,
+ "Failed to create dict.");
+ goto out;
+ }
+ for (i = 0; i < priv->child_count; i++) {
+ ret = dict_set_static_bin(xattr, priv->pending_key[i], &raw,
+ AFR_NUM_CHANGE_LOGS * sizeof(int));
+ if (ret)
+ goto out;
+ }
+
+ ret = syncop_xattrop(priv->children[THIN_ARBITER_BRICK_INDEX], loc,
+ GF_XATTROP_ADD_ARRAY, xattr, NULL, xdata, NULL);
+ if (ret || !(*xdata)) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Xattrop failed on %s.", loc->name);
+ }
+
+out:
+ if (xattr)
+ dict_unref(xattr);
+
+ return ret;
}
void
-afr_shd_ta_set_xattrs (xlator_t *this, loc_t *loc, dict_t **xdata,
- int healer)
+afr_shd_ta_get_xattrs(xlator_t *this, loc_t *loc, struct subvol_healer *healer,
+ dict_t **xdata)
{
- afr_private_t *priv = NULL;
- dict_t *xattr = NULL;
- struct gf_flock flock = {0, };
- gf_boolean_t need_xattrop = _gf_false;
- void *pending_raw = NULL;
- int *raw = NULL;
- int pending[AFR_NUM_CHANGE_LOGS] = {0,};
- int i = 0;
- int j = 0;
- int val = 0;
- int ret = 0;
-
- priv = this->private;
-
- xattr = dict_new ();
- if (!xattr) {
- goto out;
+ int ret = 0;
+
+ loc_wipe(loc);
+ if (afr_shd_fill_ta_loc(this, loc)) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to populate thin-arbiter loc for: %s.", loc->name);
+ ret = -1;
+ goto out;
+ }
+
+ ret = afr_ta_post_op_lock(this, loc);
+ if (ret)
+ goto out;
+
+ ret = _afr_shd_ta_get_xattrs(this, loc, xdata);
+ if (ret) {
+ if (*xdata) {
+ dict_unref(*xdata);
+ *xdata = NULL;
}
+ }
- for (i = 0; i < priv->child_count; i++) {
- raw = GF_CALLOC (AFR_NUM_CHANGE_LOGS, sizeof(int),
- gf_afr_mt_int32_t);
- if (!raw) {
- goto out;
- }
-
- ret = dict_get_ptr (*xdata, priv->pending_key[i], &pending_raw);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- AFR_MSG_DICT_GET_FAILED, "Error getting value "
- "of pending key %s", priv->pending_key[i]);
- GF_FREE (raw);
- goto out;
- }
+ afr_ta_post_op_unlock(this, loc);
- memcpy (pending, pending_raw, sizeof (pending));
- for (j = 0; j < AFR_NUM_CHANGE_LOGS; j++) {
- val = ntoh32 (pending[j]);
- if (val) {
- if (i == healer) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_THIN_ARB, "I am "
- "not the good shd. Skipping. "
- "SHD = %d.", healer);
- GF_FREE (raw);
- goto out;
- }
- need_xattrop = _gf_true;
- raw[j] = hton32 (-val);
- }
- }
+out:
+ if (ret)
+ healer->rerun = 1;
+}
- ret = dict_set_bin (xattr, priv->pending_key[i], raw,
- AFR_NUM_CHANGE_LOGS * sizeof (int));
- if (ret) {
- GF_FREE (raw);
- goto out;
- }
+int
+afr_shd_ta_unset_xattrs(xlator_t *this, loc_t *loc, dict_t **xdata, int healer)
+{
+ afr_private_t *priv = NULL;
+ dict_t *xattr = NULL;
+ gf_boolean_t need_xattrop = _gf_false;
+ void *pending_raw = NULL;
+ int *raw = NULL;
+ int pending[AFR_NUM_CHANGE_LOGS] = {
+ 0,
+ };
+ int i = 0;
+ int j = 0;
+ int val = 0;
+ int ret = -1;
+
+ priv = this->private;
+
+ xattr = dict_new();
+ if (!xattr) {
+ goto out;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ raw = GF_CALLOC(AFR_NUM_CHANGE_LOGS, sizeof(int), gf_afr_mt_int32_t);
+ if (!raw) {
+ goto out;
+ }
- memset (pending, 0, sizeof (pending));
+ ret = dict_get_ptr(*xdata, priv->pending_key[i], &pending_raw);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_GET_FAILED,
+ "Error getting value "
+ "of pending key %s",
+ priv->pending_key[i]);
+ GF_FREE(raw);
+ goto out;
}
- if (!need_xattrop) {
- goto out;
+ memcpy(pending, pending_raw, sizeof(pending));
+ for (j = 0; j < AFR_NUM_CHANGE_LOGS; j++) {
+ val = ntoh32(pending[j]);
+ if (val) {
+ if (i == healer) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_THIN_ARB,
+ "I am "
+ "not the good shd. Skipping. "
+ "SHD = %d.",
+ healer);
+ ret = 0;
+ GF_FREE(raw);
+ goto out;
+ }
+ need_xattrop = _gf_true;
+ raw[j] = hton32(-val);
+ }
}
- flock.l_type = F_WRLCK;
- flock.l_start = 0;
- flock.l_len = 0;
+ ret = dict_set_bin(xattr, priv->pending_key[i], raw,
+ AFR_NUM_CHANGE_LOGS * sizeof(int));
+ if (ret) {
+ GF_FREE(raw);
+ goto out;
+ }
- ret = syncop_inodelk (priv->children[THIN_ARBITER_BRICK_INDEX],
- THIN_ARBITER_DOM1, loc, F_SETLKW, &flock,
- NULL, NULL);
- if (ret)
- goto out;
+ if (need_xattrop)
+ break;
+ }
- ret = syncop_xattrop (priv->children[THIN_ARBITER_BRICK_INDEX], loc,
- GF_XATTROP_ADD_ARRAY, xattr, NULL, NULL, NULL);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Xattrop failed.");
+ if (!need_xattrop) {
+ ret = 0;
+ goto out;
+ }
- flock.l_type = F_UNLCK;
- syncop_inodelk (priv->children[THIN_ARBITER_BRICK_INDEX],
- THIN_ARBITER_DOM1, loc, F_SETLKW, &flock, NULL, NULL);
+ ret = syncop_xattrop(priv->children[THIN_ARBITER_BRICK_INDEX], loc,
+ GF_XATTROP_ADD_ARRAY, xattr, NULL, NULL, NULL);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Xattrop failed.");
out:
- if (xattr)
- dict_unref (xattr);
- return;
+ if (xattr)
+ dict_unref(xattr);
+
+ return ret;
}
void
-afr_shd_ta_get_xattrs (xlator_t *this, loc_t *loc, dict_t **xdata)
+afr_shd_ta_check_and_unset_xattrs(xlator_t *this, loc_t *loc,
+ struct subvol_healer *healer,
+ dict_t *pre_crawl_xdata)
{
- afr_private_t *priv = NULL;
- dict_t *xattr = NULL;
- struct iatt stbuf = {0,};
- int *raw = NULL;
- int ret = 0;
- int i = 0;
-
- priv = this->private;
-
- loc->parent = inode_ref (this->itable->root);
- gf_uuid_copy (loc->pargfid, loc->parent->gfid);
- loc->name = priv->pending_key[THIN_ARBITER_BRICK_INDEX];
- loc->inode = inode_new (loc->parent->table);
- if (!loc->inode) {
- goto out;
- }
+ int ret_lock = 0;
+ int ret = 0;
+ dict_t *post_crawl_xdata = NULL;
- ret = syncop_lookup (priv->children[THIN_ARBITER_BRICK_INDEX], loc,
- &stbuf, 0, 0, 0);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Failed lookup on file %s.", loc->name);
- goto out;
- }
+ ret_lock = afr_ta_post_op_lock(this, loc);
+ if (ret_lock)
+ goto unref;
- gf_uuid_copy (priv->ta_gfid, stbuf.ia_gfid);
- gf_uuid_copy (loc->gfid, priv->ta_gfid);
+ ret = _afr_shd_ta_get_xattrs(this, loc, &post_crawl_xdata);
+ if (ret)
+ goto unref;
- xattr = dict_new ();
- if (!xattr) {
- gf_msg (this->name, GF_LOG_ERROR, 0, AFR_MSG_DICT_GET_FAILED,
- "Failed to create dict.");
- goto out;
+ if (!are_dicts_equal(pre_crawl_xdata, post_crawl_xdata, NULL, NULL)) {
+ ret = -1;
+ goto unref;
+ }
+
+ ret = afr_shd_ta_unset_xattrs(this, loc, &post_crawl_xdata, healer->subvol);
+
+unref:
+ if (post_crawl_xdata) {
+ dict_unref(post_crawl_xdata);
+ post_crawl_xdata = NULL;
+ }
+
+ if (ret || ret_lock)
+ healer->rerun = 1;
+
+ if (!ret_lock)
+ afr_ta_post_op_unlock(this, loc);
+}
+
+gf_boolean_t
+afr_bricks_available_for_heal(afr_private_t *priv)
+{
+ int up_children = 0;
+
+ up_children = __afr_get_up_children_count(priv);
+ if (up_children < 2) {
+ return _gf_false;
+ }
+ return _gf_true;
+}
+
+static gf_boolean_t
+afr_shd_ta_needs_heal(xlator_t *this, struct subvol_healer *healer)
+{
+ dict_t *xdata = NULL;
+ afr_private_t *priv = NULL;
+ loc_t loc = {
+ 0,
+ };
+ int ret = -1;
+ int i = 0;
+ gf_boolean_t need_heal = _gf_false;
+
+ priv = this->private;
+
+ ret = afr_shd_fill_ta_loc(this, &loc);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to populate thin-arbiter loc for: %s.", loc.name);
+ healer->rerun = 1;
+ goto out;
+ }
+
+ if (_afr_shd_ta_get_xattrs(this, &loc, &xdata)) {
+ healer->rerun = 1;
+ goto out;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (afr_ta_dict_contains_pending_xattr(xdata, priv, i)) {
+ need_heal = _gf_true;
+ break;
}
+ }
- for (i = 0; i < priv->child_count; i++) {
- raw = GF_CALLOC (AFR_NUM_CHANGE_LOGS, sizeof(int),
- gf_afr_mt_int32_t);
- if (!raw) {
- goto out;
- }
+out:
+ if (xdata)
+ dict_unref(xdata);
+ loc_wipe(&loc);
- ret = dict_set_bin (xattr, priv->pending_key[i], raw,
- AFR_NUM_CHANGE_LOGS * sizeof (int));
- if (ret) {
- GF_FREE (raw);
- goto out;
+ return need_heal;
+}
+
+static int
+afr_shd_anon_inode_cleaner(xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
+ void *data)
+{
+ struct subvol_healer *healer = data;
+ afr_private_t *priv = healer->this->private;
+ call_frame_t *frame = NULL;
+ afr_local_t *local = NULL;
+ int ret = 0;
+ loc_t loc = {0};
+ int count = 0;
+ int i = 0;
+ int op_errno = 0;
+ struct iatt *iatt = NULL;
+ gf_boolean_t multiple_links = _gf_false;
+ unsigned char *gfid_present = alloca0(priv->child_count);
+ unsigned char *entry_present = alloca0(priv->child_count);
+ char *type = "file";
+
+ frame = afr_frame_create(healer->this, &ret);
+ if (!frame) {
+ ret = -ret;
+ goto out;
+ }
+ local = frame->local;
+ if (AFR_COUNT(local->child_up, priv->child_count) != priv->child_count) {
+ gf_msg_debug(healer->this->name, 0,
+ "Not all bricks are up. Skipping "
+ "cleanup of %s on %s",
+ entry->d_name, subvol->name);
+ ret = 0;
+ goto out;
+ }
+
+ loc.inode = inode_new(parent->inode->table);
+ if (!loc.inode) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ ret = gf_uuid_parse(entry->d_name, loc.gfid);
+ if (ret) {
+ ret = 0;
+ goto out;
+ }
+ AFR_ONLIST(local->child_up, frame, afr_selfheal_discover_cbk, lookup, &loc,
+ NULL);
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->replies[i].op_ret == 0) {
+ count++;
+ gfid_present[i] = 1;
+ iatt = &local->replies[i].poststat;
+ if (iatt->ia_type == IA_IFDIR) {
+ type = "dir";
+ }
+
+ if (i == healer->subvol) {
+ if (local->replies[i].poststat.ia_nlink > 1) {
+ multiple_links = _gf_true;
}
+ }
+ } else if (local->replies[i].op_errno != ENOENT &&
+ local->replies[i].op_errno != ESTALE) {
+ /*We don't have complete view. Skip the entry*/
+ gf_msg_debug(healer->this->name, local->replies[i].op_errno,
+ "Skipping cleanup of %s on %s", entry->d_name,
+ subvol->name);
+ ret = 0;
+ goto out;
}
-
- ret = syncop_xattrop (priv->children[THIN_ARBITER_BRICK_INDEX],
- loc, GF_XATTROP_ADD_ARRAY, xattr, NULL, xdata,
- NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Xattrop failed.");
+ }
+
+ /*Inode is deleted from subvol*/
+ if (count == 1 || (iatt->ia_type != IA_IFDIR && multiple_links)) {
+ gf_msg(healer->this->name, GF_LOG_WARNING, 0,
+ AFR_MSG_EXPUNGING_FILE_OR_DIR, "expunging %s %s/%s on %s", type,
+ priv->anon_inode_name, entry->d_name, subvol->name);
+ ret = afr_shd_entry_purge(subvol, parent->inode, entry->d_name,
+ iatt->ia_type);
+ if (ret == -ENOENT || ret == -ESTALE)
+ ret = 0;
+ } else if (count > 1) {
+ loc_wipe(&loc);
+ loc.parent = inode_ref(parent->inode);
+ loc.name = entry->d_name;
+ loc.inode = inode_new(parent->inode->table);
+ if (!loc.inode) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ AFR_ONLIST(local->child_up, frame, afr_selfheal_discover_cbk, lookup,
+ &loc, NULL);
+ count = 0;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->replies[i].op_ret == 0) {
+ count++;
+ entry_present[i] = 1;
+ iatt = &local->replies[i].poststat;
+ } else if (local->replies[i].op_errno != ENOENT &&
+ local->replies[i].op_errno != ESTALE) {
+ /*We don't have complete view. Skip the entry*/
+ gf_msg_debug(healer->this->name, local->replies[i].op_errno,
+ "Skipping cleanup of %s on %s", entry->d_name,
+ subvol->name);
+ ret = 0;
+ goto out;
+ }
+ }
+ for (i = 0; i < priv->child_count; i++) {
+ if (gfid_present[i] && !entry_present[i]) {
+ /*Entry is not anonymous on at least one subvol*/
+ gf_msg_debug(healer->this->name, 0,
+ "Valid entry present on %s "
+ "Skipping cleanup of %s on %s",
+ priv->children[i]->name, entry->d_name,
+ subvol->name);
+ ret = 0;
goto out;
+ }
+ }
+
+ gf_msg(healer->this->name, GF_LOG_WARNING, 0,
+ AFR_MSG_EXPUNGING_FILE_OR_DIR,
+ "expunging %s %s/%s on all subvols", type, priv->anon_inode_name,
+ entry->d_name);
+ ret = 0;
+ for (i = 0; i < priv->child_count; i++) {
+ op_errno = -afr_shd_entry_purge(priv->children[i], loc.parent,
+ entry->d_name, iatt->ia_type);
+ if (op_errno != ENOENT && op_errno != ESTALE) {
+ ret |= -op_errno;
+ }
}
- if (!(*xdata))
- gf_msg (this->name, GF_LOG_ERROR, 0, AFR_MSG_DICT_GET_FAILED,
- "Xdata response is empty.");
+ }
out:
- if (xattr)
- dict_unref (xattr);
- return;
+ if (frame)
+ AFR_STACK_DESTROY(frame);
+ loc_wipe(&loc);
+ return ret;
+}
+
+static void
+afr_cleanup_anon_inode_dir(struct subvol_healer *healer)
+{
+ int ret = 0;
+ call_frame_t *frame = NULL;
+ afr_private_t *priv = healer->this->private;
+ loc_t loc = {0};
+
+ ret = afr_anon_inode_create(healer->this, healer->subvol, &loc.inode);
+ if (ret)
+ goto out;
+
+ frame = afr_frame_create(healer->this, &ret);
+ if (!frame) {
+ ret = -ret;
+ goto out;
+ }
+
+ ret = syncop_mt_dir_scan(frame, priv->children[healer->subvol], &loc,
+ GF_CLIENT_PID_SELF_HEALD, healer,
+ afr_shd_anon_inode_cleaner, NULL,
+ priv->shd.max_threads, priv->shd.wait_qlength);
+out:
+ if (frame)
+ AFR_STACK_DESTROY(frame);
+ loc_wipe(&loc);
+ return;
}
void *
-afr_shd_index_healer (void *data)
+afr_shd_index_healer(void *data)
{
- struct subvol_healer *healer = NULL;
- xlator_t *this = NULL;
- int ret = 0;
- afr_private_t *priv = NULL;
- dict_t *xdata = NULL;
- loc_t loc = {0, };
-
- healer = data;
- THIS = this = healer->this;
- priv = this->private;
-
- for (;;) {
- afr_shd_healer_wait (healer);
-
- ASSERT_LOCAL(this, healer);
- priv->local[healer->subvol] = healer->local;
-
- if (priv->thin_arbiter_count) {
- loc_wipe (&loc);
- afr_shd_ta_get_xattrs (this, &loc, &xdata);
- }
+ struct subvol_healer *healer = NULL;
+ xlator_t *this = NULL;
+ int ret = 0;
+ afr_private_t *priv = NULL;
+ dict_t *pre_crawl_xdata = NULL;
+ loc_t loc = {
+ 0,
+ };
+
+ healer = data;
+ THIS = this = healer->this;
+ priv = this->private;
+
+ for (;;) {
+ afr_shd_healer_wait(healer);
+
+ if (!afr_bricks_available_for_heal(priv))
+ continue;
+
+ ASSERT_LOCAL(this, healer);
+ priv->local[healer->subvol] = healer->local;
+
+ if (priv->thin_arbiter_count) {
+ if (afr_shd_ta_needs_heal(this, healer))
+ afr_shd_ta_get_xattrs(this, &loc, healer, &pre_crawl_xdata);
+ }
+ do {
+ gf_msg_debug(this->name, 0, "starting index sweep on subvol %s",
+ afr_subvol_name(this, healer->subvol));
+
+ afr_shd_sweep_prepare(healer);
+
+ ret = afr_shd_index_sweep_all(healer);
+
+ afr_shd_sweep_done(healer);
+ /*
+ As long as at least one gfid was
+ healed, keep retrying. We may have
+ just healed a directory and thereby
+ created entries for other gfids which
+ could not be healed thus far.
+ */
+
+ gf_msg_debug(this->name, 0, "finished index sweep on subvol %s",
+ afr_subvol_name(this, healer->subvol));
+ /*
+ Give a pause before retrying to avoid a busy loop
+ in case the only entry in index is because of
+ an ongoing I/O.
+ */
+ sleep(1);
+ } while (ret > 0);
+
+ if (ret == 0) {
+ afr_cleanup_anon_inode_dir(healer);
+ }
- do {
- gf_msg_debug (this->name, 0,
- "starting index sweep on subvol %s",
- afr_subvol_name (this, healer->subvol));
-
- afr_shd_sweep_prepare (healer);
-
- ret = afr_shd_index_sweep_all (healer);
-
- afr_shd_sweep_done (healer);
- /*
- As long as at least one gfid was
- healed, keep retrying. We may have
- just healed a directory and thereby
- created entries for other gfids which
- could not be healed thus far.
- */
-
- gf_msg_debug (this->name, 0,
- "finished index sweep on subvol %s",
- afr_subvol_name (this, healer->subvol));
- /*
- Give a pause before retrying to avoid a busy loop
- in case the only entry in index is because of
- an ongoing I/O.
- */
- sleep (1);
- } while (ret > 0);
-
- if (xdata && !healer->crawl_event.heal_failed_count) {
- afr_shd_ta_set_xattrs (this, &loc, &xdata,
- healer->subvol);
- dict_unref (xdata);
- xdata = NULL;
- }
- }
+ if (ret == 0 && pre_crawl_xdata &&
+ !healer->crawl_event.heal_failed_count) {
+ afr_shd_ta_check_and_unset_xattrs(this, &loc, healer,
+ pre_crawl_xdata);
+ }
- loc_wipe (&loc);
+ if (pre_crawl_xdata) {
+ dict_unref(pre_crawl_xdata);
+ pre_crawl_xdata = NULL;
+ }
+ }
- return NULL;
+ return NULL;
}
-
void *
-afr_shd_full_healer (void *data)
+afr_shd_full_healer(void *data)
{
- struct subvol_healer *healer = NULL;
- xlator_t *this = NULL;
- int run = 0;
+ struct subvol_healer *healer = NULL;
+ xlator_t *this = NULL;
+ int run = 0;
- healer = data;
- THIS = this = healer->this;
+ healer = data;
+ THIS = this = healer->this;
- for (;;) {
- pthread_mutex_lock (&healer->mutex);
- {
- run = __afr_shd_healer_wait (healer);
- if (!run)
- healer->running = _gf_false;
- }
- pthread_mutex_unlock (&healer->mutex);
+ for (;;) {
+ pthread_mutex_lock(&healer->mutex);
+ {
+ run = __afr_shd_healer_wait(healer);
+ if (!run)
+ healer->running = _gf_false;
+ }
+ pthread_mutex_unlock(&healer->mutex);
- if (!run)
- break;
+ if (!run)
+ break;
- ASSERT_LOCAL(this, healer);
+ ASSERT_LOCAL(this, healer);
- gf_msg (this->name, GF_LOG_INFO, 0, AFR_MSG_SELF_HEAL_INFO,
- "starting full sweep on subvol %s",
- afr_subvol_name (this, healer->subvol));
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_SELF_HEAL_INFO,
+ "starting full sweep on subvol %s",
+ afr_subvol_name(this, healer->subvol));
- afr_shd_sweep_prepare (healer);
+ afr_shd_sweep_prepare(healer);
- afr_shd_full_sweep (healer, this->itable->root);
+ afr_shd_full_sweep(healer, this->itable->root);
- afr_shd_sweep_done (healer);
+ afr_shd_sweep_done(healer);
- gf_msg (this->name, GF_LOG_INFO, 0, AFR_MSG_SELF_HEAL_INFO,
- "finished full sweep on subvol %s",
- afr_subvol_name (this, healer->subvol));
- }
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_SELF_HEAL_INFO,
+ "finished full sweep on subvol %s",
+ afr_subvol_name(this, healer->subvol));
+ }
- return NULL;
+ return NULL;
}
-
int
-afr_shd_healer_init (xlator_t *this, struct subvol_healer *healer)
+afr_shd_healer_init(xlator_t *this, struct subvol_healer *healer)
{
- int ret = 0;
+ int ret = 0;
- ret = pthread_mutex_init (&healer->mutex, NULL);
- if (ret)
- goto out;
+ ret = pthread_mutex_init(&healer->mutex, NULL);
+ if (ret)
+ goto out;
- ret = pthread_cond_init (&healer->cond, NULL);
- if (ret)
- goto out;
+ ret = pthread_cond_init(&healer->cond, NULL);
+ if (ret)
+ goto out;
- healer->this = this;
- healer->running = _gf_false;
- healer->rerun = _gf_false;
- healer->local = _gf_false;
+ healer->this = this;
+ healer->running = _gf_false;
+ healer->rerun = _gf_false;
+ healer->local = _gf_false;
out:
- return ret;
+ return ret;
}
-
int
-afr_shd_healer_spawn (xlator_t *this, struct subvol_healer *healer,
- void *(threadfn)(void *))
+afr_shd_healer_spawn(xlator_t *this, struct subvol_healer *healer,
+ void *(threadfn)(void *))
{
- int ret = 0;
-
- pthread_mutex_lock (&healer->mutex);
- {
- if (healer->running) {
- pthread_cond_signal (&healer->cond);
- } else {
- ret = gf_thread_create (&healer->thread, NULL,
- threadfn, healer, "shdheal");
- if (ret)
- goto unlock;
- healer->running = 1;
- }
-
- healer->rerun = 1;
- }
+ int ret = 0;
+
+ pthread_mutex_lock(&healer->mutex);
+ {
+ if (healer->running) {
+ pthread_cond_signal(&healer->cond);
+ } else {
+ ret = gf_thread_create(&healer->thread, NULL, threadfn, healer,
+ "shdheal");
+ if (ret)
+ goto unlock;
+ healer->running = 1;
+ }
+
+ healer->rerun = 1;
+ }
unlock:
- pthread_mutex_unlock (&healer->mutex);
+ pthread_mutex_unlock(&healer->mutex);
- return ret;
+ return ret;
}
-
int
-afr_shd_full_healer_spawn (xlator_t *this, int subvol)
+afr_shd_full_healer_spawn(xlator_t *this, int subvol)
{
- return afr_shd_healer_spawn (this, NTH_FULL_HEALER (this, subvol),
- afr_shd_full_healer);
+ return afr_shd_healer_spawn(this, NTH_FULL_HEALER(this, subvol),
+ afr_shd_full_healer);
}
-
int
-afr_shd_index_healer_spawn (xlator_t *this, int subvol)
+afr_shd_index_healer_spawn(xlator_t *this, int subvol)
{
- return afr_shd_healer_spawn (this, NTH_INDEX_HEALER (this, subvol),
- afr_shd_index_healer);
+ return afr_shd_healer_spawn(this, NTH_INDEX_HEALER(this, subvol),
+ afr_shd_index_healer);
}
-
int
-afr_shd_dict_add_crawl_event (xlator_t *this, dict_t *output,
- crawl_event_t *crawl_event)
+afr_shd_dict_add_crawl_event(xlator_t *this, dict_t *output,
+ crawl_event_t *crawl_event)
{
- int ret = 0;
- uint64_t count = 0;
- char key[256] = {0};
- int xl_id = 0;
- uint64_t healed_count = 0;
- uint64_t split_brain_count = 0;
- uint64_t heal_failed_count = 0;
- char *start_time_str = 0;
- char *end_time_str = NULL;
- char *crawl_type = NULL;
- int progress = -1;
- int child = -1;
-
- child = crawl_event->child;
- healed_count = crawl_event->healed_count;
- split_brain_count = crawl_event->split_brain_count;
- heal_failed_count = crawl_event->heal_failed_count;
- crawl_type = crawl_event->crawl_type;
-
- if (!crawl_event->start_time)
- goto out;
-
- start_time_str = gf_strdup (ctime (&crawl_event->start_time));
-
- if (crawl_event->end_time)
- end_time_str = gf_strdup (ctime (&crawl_event->end_time));
-
- ret = dict_get_int32 (output, this->name, &xl_id);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- AFR_MSG_DICT_GET_FAILED, "xl does not have id");
- goto out;
- }
-
- snprintf (key, sizeof (key), "statistics-%d-%d-count", xl_id, child);
- ret = dict_get_uint64 (output, key, &count);
-
-
- snprintf (key, sizeof (key), "statistics_healed_cnt-%d-%d-%"PRIu64,
- xl_id, child, count);
- ret = dict_set_uint64(output, key, healed_count);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not add statistics_healed_count to output");
- goto out;
- }
-
- snprintf (key, sizeof (key), "statistics_sb_cnt-%d-%d-%"PRIu64,
- xl_id, child, count);
- ret = dict_set_uint64 (output, key, split_brain_count);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not add statistics_split_brain_count to output");
- goto out;
- }
-
- snprintf (key, sizeof (key), "statistics_crawl_type-%d-%d-%"PRIu64,
- xl_id, child, count);
- ret = dict_set_str (output, key, crawl_type);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not add statistics_crawl_type to output");
- goto out;
- }
-
- snprintf (key, sizeof (key), "statistics_heal_failed_cnt-%d-%d-%"PRIu64,
- xl_id, child, count);
- ret = dict_set_uint64 (output, key, heal_failed_count);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not add statistics_healed_failed_count to output");
- goto out;
- }
-
- snprintf (key, sizeof (key), "statistics_strt_time-%d-%d-%"PRIu64,
- xl_id, child, count);
- ret = dict_set_dynstr (output, key, start_time_str);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not add statistics_crawl_start_time to output");
- goto out;
- } else {
- start_time_str = NULL;
- }
-
- if (!end_time_str)
- progress = 1;
- else
- progress = 0;
-
- snprintf (key, sizeof (key), "statistics_end_time-%d-%d-%"PRIu64,
- xl_id, child, count);
- if (!end_time_str)
- end_time_str = gf_strdup ("Could not determine the end time");
- ret = dict_set_dynstr (output, key, end_time_str);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not add statistics_crawl_end_time to output");
- goto out;
- } else {
- end_time_str = NULL;
- }
-
- snprintf (key, sizeof (key), "statistics_inprogress-%d-%d-%"PRIu64,
- xl_id, child, count);
-
- ret = dict_set_int32 (output, key, progress);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not add statistics_inprogress to output");
- goto out;
- }
-
- snprintf (key, sizeof (key), "statistics-%d-%d-count", xl_id, child);
- ret = dict_set_uint64 (output, key, count + 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not increment the counter.");
- goto out;
- }
+ int ret = 0;
+ uint64_t count = 0;
+ char key[128] = {0};
+ int keylen = 0;
+ char suffix[64] = {0};
+ int xl_id = 0;
+ uint64_t healed_count = 0;
+ uint64_t split_brain_count = 0;
+ uint64_t heal_failed_count = 0;
+ char *start_time_str = 0;
+ char *end_time_str = NULL;
+ char *crawl_type = NULL;
+ int progress = -1;
+ int child = -1;
+
+ child = crawl_event->child;
+ healed_count = crawl_event->healed_count;
+ split_brain_count = crawl_event->split_brain_count;
+ heal_failed_count = crawl_event->heal_failed_count;
+ crawl_type = crawl_event->crawl_type;
+
+ if (!crawl_event->start_time)
+ goto out;
+
+ start_time_str = gf_strdup(ctime(&crawl_event->start_time));
+
+ if (crawl_event->end_time)
+ end_time_str = gf_strdup(ctime(&crawl_event->end_time));
+
+ ret = dict_get_int32(output, this->name, &xl_id);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_GET_FAILED,
+ "xl does not have id");
+ goto out;
+ }
+
+ snprintf(key, sizeof(key), "statistics-%d-%d-count", xl_id, child);
+ ret = dict_get_uint64(output, key, &count);
+
+ snprintf(suffix, sizeof(suffix), "%d-%d-%" PRIu64, xl_id, child, count);
+ snprintf(key, sizeof(key), "statistics_healed_cnt-%s", suffix);
+ ret = dict_set_uint64(output, key, healed_count);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not add statistics_healed_count to output");
+ goto out;
+ }
+
+ snprintf(key, sizeof(key), "statistics_sb_cnt-%s", suffix);
+ ret = dict_set_uint64(output, key, split_brain_count);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not add statistics_split_brain_count to output");
+ goto out;
+ }
+
+ keylen = snprintf(key, sizeof(key), "statistics_crawl_type-%s", suffix);
+ ret = dict_set_strn(output, key, keylen, crawl_type);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not add statistics_crawl_type to output");
+ goto out;
+ }
+
+ snprintf(key, sizeof(key), "statistics_heal_failed_cnt-%s", suffix);
+ ret = dict_set_uint64(output, key, heal_failed_count);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not add statistics_healed_failed_count to output");
+ goto out;
+ }
+
+ keylen = snprintf(key, sizeof(key), "statistics_strt_time-%s", suffix);
+ ret = dict_set_dynstrn(output, key, keylen, start_time_str);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not add statistics_crawl_start_time to output");
+ goto out;
+ } else {
+ start_time_str = NULL;
+ }
+
+ if (!end_time_str)
+ progress = 1;
+ else
+ progress = 0;
+
+ keylen = snprintf(key, sizeof(key), "statistics_end_time-%s", suffix);
+ if (!end_time_str)
+ end_time_str = gf_strdup("Could not determine the end time");
+ ret = dict_set_dynstrn(output, key, keylen, end_time_str);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not add statistics_crawl_end_time to output");
+ goto out;
+ } else {
+ end_time_str = NULL;
+ }
+
+ keylen = snprintf(key, sizeof(key), "statistics_inprogress-%s", suffix);
+
+ ret = dict_set_int32n(output, key, keylen, progress);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not add statistics_inprogress to output");
+ goto out;
+ }
+
+ snprintf(key, sizeof(key), "statistics-%d-%d-count", xl_id, child);
+ ret = dict_set_uint64(output, key, count + 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not increment the counter.");
+ goto out;
+ }
out:
- GF_FREE (start_time_str);
- GF_FREE (end_time_str);
- return ret;
+ GF_FREE(start_time_str);
+ GF_FREE(end_time_str);
+ return ret;
}
-
int
-afr_shd_dict_add_path (xlator_t *this, dict_t *output, int child, char *path,
- struct timeval *tv)
+afr_shd_dict_add_path(xlator_t *this, dict_t *output, int child, char *path,
+ struct timeval *tv)
{
- int ret = -1;
- uint64_t count = 0;
- char key[256] = {0};
- int xl_id = 0;
-
- ret = dict_get_int32 (output, this->name, &xl_id);
+ int ret = -1;
+ uint64_t count = 0;
+ char key[64] = {0};
+ int keylen = 0;
+ char xl_id_child_str[32] = {0};
+ int xl_id = 0;
+
+ ret = dict_get_int32(output, this->name, &xl_id);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_GET_FAILED,
+ "xl does not have id");
+ goto out;
+ }
+
+ snprintf(xl_id_child_str, sizeof(xl_id_child_str), "%d-%d", xl_id, child);
+ snprintf(key, sizeof(key), "%s-count", xl_id_child_str);
+ ret = dict_get_uint64(output, key, &count);
+
+ keylen = snprintf(key, sizeof(key), "%s-%" PRIu64, xl_id_child_str, count);
+ ret = dict_set_dynstrn(output, key, keylen, path);
+
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "%s: Could not add to output", path);
+ goto out;
+ }
+
+ if (tv) {
+ snprintf(key, sizeof(key), "%s-%" PRIu64 "-time", xl_id_child_str,
+ count);
+ ret = dict_set_uint32(output, key, tv->tv_sec);
if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- AFR_MSG_DICT_GET_FAILED, "xl does not have id");
- goto out;
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "%s: Could not set time", path);
+ goto out;
}
+ }
- snprintf (key, sizeof (key), "%d-%d-count", xl_id, child);
- ret = dict_get_uint64 (output, key, &count);
-
- snprintf (key, sizeof (key), "%d-%d-%"PRIu64, xl_id, child, count);
- ret = dict_set_dynstr (output, key, path);
-
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- AFR_MSG_DICT_SET_FAILED, "%s: Could not add to output",
- path);
- goto out;
- }
+ snprintf(key, sizeof(key), "%s-count", xl_id_child_str);
- if (tv) {
- snprintf (key, sizeof (key), "%d-%d-%"PRIu64"-time", xl_id,
- child, count);
- ret = dict_set_uint32 (output, key, tv->tv_sec);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "%s: Could not set time",
- path);
- goto out;
- }
- }
-
- snprintf (key, sizeof (key), "%d-%d-count", xl_id, child);
-
- ret = dict_set_uint64 (output, key, count + 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- -ret, AFR_MSG_DICT_SET_FAILED,
- "Could not increment count");
- goto out;
- }
+ ret = dict_set_uint64(output, key, count + 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "Could not increment count");
+ goto out;
+ }
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-afr_add_shd_event (circular_buffer_t *cb, void *data)
+afr_add_shd_event(circular_buffer_t *cb, void *data)
{
- dict_t *output = NULL;
- xlator_t *this = THIS;
- afr_private_t *priv = NULL;
- afr_self_heald_t *shd = NULL;
- shd_event_t *shd_event = NULL;
- char *path = NULL;
-
- output = data;
- priv = this->private;
- shd = &priv->shd;
- shd_event = cb->data;
-
- if (!shd->index_healers[shd_event->child].local)
- return 0;
-
- path = gf_strdup (shd_event->path);
- if (!path)
- return -ENOMEM;
-
- afr_shd_dict_add_path (this, output, shd_event->child, path,
- &cb->tv);
- return 0;
+ dict_t *output = NULL;
+ xlator_t *this = THIS;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ shd_event_t *shd_event = NULL;
+ char *path = NULL;
+
+ output = data;
+ priv = this->private;
+ shd = &priv->shd;
+ shd_event = cb->data;
+
+ if (!shd->index_healers[shd_event->child].local)
+ return 0;
+
+ path = gf_strdup(shd_event->path);
+ if (!path)
+ return -ENOMEM;
+
+ afr_shd_dict_add_path(this, output, shd_event->child, path, &cb->tv);
+ return 0;
}
int
-afr_add_crawl_event (circular_buffer_t *cb, void *data)
+afr_add_crawl_event(circular_buffer_t *cb, void *data)
{
- dict_t *output = NULL;
- xlator_t *this = THIS;
- afr_private_t *priv = NULL;
- afr_self_heald_t *shd = NULL;
- crawl_event_t *crawl_event = NULL;
-
- output = data;
- priv = this->private;
- shd = &priv->shd;
- crawl_event = cb->data;
-
- if (!shd->index_healers[crawl_event->child].local)
- return 0;
+ dict_t *output = NULL;
+ xlator_t *this = THIS;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ crawl_event_t *crawl_event = NULL;
+
+ output = data;
+ priv = this->private;
+ shd = &priv->shd;
+ crawl_event = cb->data;
+
+ if (!shd->index_healers[crawl_event->child].local)
+ return 0;
- afr_shd_dict_add_crawl_event (this, output, crawl_event);
+ afr_shd_dict_add_crawl_event(this, output, crawl_event);
- return 0;
+ return 0;
}
-
int
-afr_selfheal_daemon_init (xlator_t *this)
+afr_selfheal_daemon_init(xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_self_heald_t *shd = NULL;
- int ret = -1;
- int i = 0;
-
- priv = this->private;
- shd = &priv->shd;
-
- this->itable = inode_table_new (SHD_INODE_LRU_LIMIT, this);
- if (!this->itable)
- goto out;
-
- shd->index_healers = GF_CALLOC (sizeof(*shd->index_healers),
- priv->child_count,
- gf_afr_mt_subvol_healer_t);
- if (!shd->index_healers)
- goto out;
-
- for (i = 0; i < priv->child_count; i++) {
- shd->index_healers[i].subvol = i;
- ret = afr_shd_healer_init (this, &shd->index_healers[i]);
- if (ret)
- goto out;
- }
-
- shd->full_healers = GF_CALLOC (sizeof(*shd->full_healers),
- priv->child_count,
- gf_afr_mt_subvol_healer_t);
- if (!shd->full_healers)
- goto out;
- for (i = 0; i < priv->child_count; i++) {
- shd->full_healers[i].subvol = i;
- ret = afr_shd_healer_init (this, &shd->full_healers[i]);
- if (ret)
- goto out;
- }
-
- shd->split_brain = eh_new (AFR_EH_SPLIT_BRAIN_LIMIT, _gf_false,
- afr_destroy_shd_event_data);
- if (!shd->split_brain)
- goto out;
-
- shd->statistics = GF_CALLOC (sizeof(eh_t *), priv->child_count,
- gf_common_mt_eh_t);
- if (!shd->statistics)
- goto out;
-
- for (i = 0; i < priv->child_count ; i++) {
- shd->statistics[i] = eh_new (AFR_STATISTICS_HISTORY_SIZE,
- _gf_false,
- afr_destroy_crawl_event_data);
- if (!shd->statistics[i])
- goto out;
- shd->full_healers[i].crawl_event.child = i;
- shd->full_healers[i].crawl_event.crawl_type = "FULL";
- shd->index_healers[i].crawl_event.child = i;
- shd->index_healers[i].crawl_event.crawl_type = "INDEX";
- }
-
- ret = 0;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ int ret = -1;
+ int i = 0;
+
+ priv = this->private;
+ shd = &priv->shd;
+
+ shd->index_healers = GF_CALLOC(sizeof(*shd->index_healers),
+ priv->child_count,
+ gf_afr_mt_subvol_healer_t);
+ if (!shd->index_healers)
+ goto out;
+
+ for (i = 0; i < priv->child_count; i++) {
+ shd->index_healers[i].subvol = i;
+ ret = afr_shd_healer_init(this, &shd->index_healers[i]);
+ if (ret)
+ goto out;
+ }
+
+ shd->full_healers = GF_CALLOC(sizeof(*shd->full_healers), priv->child_count,
+ gf_afr_mt_subvol_healer_t);
+ if (!shd->full_healers)
+ goto out;
+ for (i = 0; i < priv->child_count; i++) {
+ shd->full_healers[i].subvol = i;
+ ret = afr_shd_healer_init(this, &shd->full_healers[i]);
+ if (ret)
+ goto out;
+ }
+
+ shd->split_brain = eh_new(AFR_EH_SPLIT_BRAIN_LIMIT, _gf_false,
+ afr_destroy_shd_event_data);
+ if (!shd->split_brain)
+ goto out;
+
+ shd->statistics = GF_CALLOC(sizeof(eh_t *), priv->child_count,
+ gf_common_mt_eh_t);
+ if (!shd->statistics)
+ goto out;
+
+ for (i = 0; i < priv->child_count; i++) {
+ shd->statistics[i] = eh_new(AFR_STATISTICS_HISTORY_SIZE, _gf_false,
+ afr_destroy_crawl_event_data);
+ if (!shd->statistics[i])
+ goto out;
+ shd->full_healers[i].crawl_event.child = i;
+ shd->full_healers[i].crawl_event.crawl_type = "FULL";
+ shd->index_healers[i].crawl_event.child = i;
+ shd->index_healers[i].crawl_event.crawl_type = "INDEX";
+ }
+
+ ret = 0;
out:
- return ret;
+ return ret;
}
-
-int
-afr_selfheal_childup (xlator_t *this, int subvol)
+void
+afr_selfheal_childup(xlator_t *this, afr_private_t *priv)
{
- afr_shd_index_healer_spawn (this, subvol);
+ int subvol = 0;
- return 0;
-}
+ if (!priv->shd.iamshd)
+ return;
+ for (subvol = 0; subvol < priv->child_count; subvol++)
+ if (priv->child_up[subvol])
+ afr_shd_index_healer_spawn(this, subvol);
+ return;
+}
int
-afr_shd_get_index_count (xlator_t *this, int i, uint64_t *count)
+afr_shd_get_index_count(xlator_t *this, int i, uint64_t *count)
{
- afr_private_t *priv = NULL;
- xlator_t *subvol = NULL;
- loc_t rootloc = {0, };
- dict_t *xattr = NULL;
- int ret = -1;
+ afr_private_t *priv = NULL;
+ xlator_t *subvol = NULL;
+ loc_t rootloc = {
+ 0,
+ };
+ dict_t *xattr = NULL;
+ int ret = -1;
- priv = this->private;
- subvol = priv->children[i];
+ priv = this->private;
+ subvol = priv->children[i];
- rootloc.inode = inode_ref (this->itable->root);
- gf_uuid_copy (rootloc.gfid, rootloc.inode->gfid);
+ rootloc.inode = inode_ref(this->itable->root);
+ gf_uuid_copy(rootloc.gfid, rootloc.inode->gfid);
- ret = syncop_getxattr (subvol, &rootloc, &xattr,
- GF_XATTROP_INDEX_COUNT, NULL, NULL);
- if (ret < 0)
- goto out;
+ ret = syncop_getxattr(subvol, &rootloc, &xattr, GF_XATTROP_INDEX_COUNT,
+ NULL, NULL);
+ if (ret < 0)
+ goto out;
- ret = dict_get_uint64 (xattr, GF_XATTROP_INDEX_COUNT, count);
- if (ret)
- goto out;
+ ret = dict_get_uint64(xattr, GF_XATTROP_INDEX_COUNT, count);
+ if (ret)
+ goto out;
- ret = 0;
+ ret = 0;
out:
- if (xattr)
- dict_unref (xattr);
- loc_wipe (&rootloc);
+ if (xattr)
+ dict_unref(xattr);
+ loc_wipe(&rootloc);
- return ret;
+ return ret;
}
-
int
-afr_xl_op (xlator_t *this, dict_t *input, dict_t *output)
+afr_xl_op(xlator_t *this, dict_t *input, dict_t *output)
{
- gf_xl_afr_op_t op = GF_SHD_OP_INVALID;
- int ret = 0;
- int xl_id = 0;
- afr_private_t *priv = NULL;
- afr_self_heald_t *shd = NULL;
- struct subvol_healer *healer = NULL;
- int i = 0;
- char key[64];
- int op_ret = 0;
- uint64_t cnt = 0;
-
- priv = this->private;
- shd = &priv->shd;
-
- ret = dict_get_int32 (input, "xl-op", (int32_t*)&op);
- if (ret)
- goto out;
- ret = dict_get_int32 (input, this->name, &xl_id);
- if (ret)
- goto out;
- ret = dict_set_int32 (output, this->name, xl_id);
- if (ret)
- goto out;
- switch (op) {
+ gf_xl_afr_op_t op = GF_SHD_OP_INVALID;
+ int ret = 0;
+ int xl_id = 0;
+ afr_private_t *priv = NULL;
+ afr_self_heald_t *shd = NULL;
+ struct subvol_healer *healer = NULL;
+ int i = 0;
+ char key[64];
+ int keylen = 0;
+ int this_name_len = 0;
+ int op_ret = 0;
+ uint64_t cnt = 0;
+
+#define AFR_SET_DICT_AND_LOG(name, output, key, keylen, dict_str, \
+ dict_str_len) \
+ { \
+ int ret; \
+ \
+ ret = dict_set_nstrn(output, key, keylen, dict_str, dict_str_len); \
+ if (ret) { \
+ gf_smsg(name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED, \
+ "key=%s", key, "value=%s", dict_str, NULL); \
+ } \
+ }
+
+ priv = this->private;
+ shd = &priv->shd;
+
+ ret = dict_get_int32_sizen(input, "xl-op", (int32_t *)&op);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_GET_FAILED,
+ "key=xl-op", NULL);
+ goto out;
+ }
+ this_name_len = strlen(this->name);
+ ret = dict_get_int32n(input, this->name, this_name_len, &xl_id);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_GET_FAILED,
+ "key=%s", this->name, NULL);
+ goto out;
+ }
+ ret = dict_set_int32n(output, this->name, this_name_len, xl_id);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_DICT_SET_FAILED,
+ "key=%s", this->name, NULL);
+ goto out;
+ }
+ switch (op) {
case GF_SHD_OP_HEAL_INDEX:
- op_ret = 0;
-
- for (i = 0; i < priv->child_count; i++) {
- healer = &shd->index_healers[i];
- snprintf (key, sizeof (key), "%d-%d-status", xl_id, i);
-
- if (!priv->child_up[i]) {
- ret = dict_set_str (output, key,
- "Brick is not connected");
- op_ret = -1;
- } else if (AFR_COUNT (priv->child_up,
- priv->child_count) < 2) {
- ret = dict_set_str (output, key,
- "< 2 bricks in replica are up");
- op_ret = -1;
- } else if (!afr_shd_is_subvol_local (this, healer->subvol)) {
- ret = dict_set_str (output, key,
- "Brick is remote");
- } else {
- ret = dict_set_str (output, key,
- "Started self-heal");
- afr_shd_index_healer_spawn (this, i);
- }
- }
- break;
+ op_ret = 0;
+
+ for (i = 0; i < priv->child_count; i++) {
+ healer = &shd->index_healers[i];
+ keylen = snprintf(key, sizeof(key), "%d-%d-status", xl_id, i);
+
+ if (!priv->child_up[i]) {
+ AFR_SET_DICT_AND_LOG(this->name, output, key, keylen,
+ SBRICK_NOT_CONNECTED,
+ SLEN(SBRICK_NOT_CONNECTED));
+ op_ret = -1;
+ } else if (AFR_COUNT(priv->child_up, priv->child_count) < 2) {
+ AFR_SET_DICT_AND_LOG(this->name, output, key, keylen,
+ SLESS_THAN2_BRICKS_in_REP,
+ SLEN(SLESS_THAN2_BRICKS_in_REP));
+ op_ret = -1;
+ } else if (!afr_shd_is_subvol_local(this, healer->subvol)) {
+ AFR_SET_DICT_AND_LOG(this->name, output, key, keylen,
+ SBRICK_IS_REMOTE,
+ SLEN(SBRICK_IS_REMOTE));
+ } else {
+ AFR_SET_DICT_AND_LOG(this->name, output, key, keylen,
+ SSTARTED_SELF_HEAL,
+ SLEN(SSTARTED_SELF_HEAL));
+
+ ret = afr_shd_index_healer_spawn(this, i);
+
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret,
+ AFR_MSG_HEALER_SPAWN_FAILED, NULL);
+ }
+ }
+ }
+ break;
case GF_SHD_OP_HEAL_FULL:
- op_ret = -1;
-
- for (i = 0; i < priv->child_count; i++) {
- healer = &shd->full_healers[i];
- snprintf (key, sizeof (key), "%d-%d-status", xl_id, i);
-
- if (!priv->child_up[i]) {
- ret = dict_set_str (output, key,
- "Brick is not connected");
- } else if (AFR_COUNT (priv->child_up,
- priv->child_count) < 2) {
- ret = dict_set_str (output, key,
- "< 2 bricks in replica are up");
- } else if (!afr_shd_is_subvol_local (this, healer->subvol)) {
- ret = dict_set_str (output, key,
- "Brick is remote");
- } else {
- ret = dict_set_str (output, key,
- "Started self-heal");
- afr_shd_full_healer_spawn (this, i);
- op_ret = 0;
- }
- }
- break;
- case GF_SHD_OP_INDEX_SUMMARY:
- /* this case has been handled in glfs-heal.c */
- break;
- case GF_SHD_OP_HEALED_FILES:
- case GF_SHD_OP_HEAL_FAILED_FILES:
- for (i = 0; i < priv->child_count; i++) {
- snprintf (key, sizeof (key), "%d-%d-status", xl_id, i);
- ret = dict_set_str (output, key, "Operation Not "
- "Supported");
+ op_ret = -1;
+
+ for (i = 0; i < priv->child_count; i++) {
+ healer = &shd->full_healers[i];
+ keylen = snprintf(key, sizeof(key), "%d-%d-status", xl_id, i);
+
+ if (!priv->child_up[i]) {
+ AFR_SET_DICT_AND_LOG(this->name, output, key, keylen,
+ SBRICK_NOT_CONNECTED,
+ SLEN(SBRICK_NOT_CONNECTED));
+ } else if (AFR_COUNT(priv->child_up, priv->child_count) < 2) {
+ AFR_SET_DICT_AND_LOG(this->name, output, key, keylen,
+ SLESS_THAN2_BRICKS_in_REP,
+ SLEN(SLESS_THAN2_BRICKS_in_REP));
+ } else if (!afr_shd_is_subvol_local(this, healer->subvol)) {
+ AFR_SET_DICT_AND_LOG(this->name, output, key, keylen,
+ SBRICK_IS_REMOTE,
+ SLEN(SBRICK_IS_REMOTE));
+ } else {
+ AFR_SET_DICT_AND_LOG(this->name, output, key, keylen,
+ SSTARTED_SELF_HEAL,
+ SLEN(SSTARTED_SELF_HEAL));
+
+ ret = afr_shd_full_healer_spawn(this, i);
+
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret,
+ AFR_MSG_HEALER_SPAWN_FAILED, NULL);
+ }
+ op_ret = 0;
}
- break;
+ }
+ break;
+ case GF_SHD_OP_INDEX_SUMMARY:
+ /* this case has been handled in glfs-heal.c */
+ break;
case GF_SHD_OP_SPLIT_BRAIN_FILES:
- eh_dump (shd->split_brain, output, afr_add_shd_event);
- break;
+ eh_dump(shd->split_brain, output, afr_add_shd_event);
+ break;
case GF_SHD_OP_STATISTICS:
- for (i = 0; i < priv->child_count; i++) {
- eh_dump (shd->statistics[i], output,
- afr_add_crawl_event);
- afr_shd_dict_add_crawl_event (this, output,
- &shd->index_healers[i].crawl_event);
- afr_shd_dict_add_crawl_event (this, output,
- &shd->full_healers[i].crawl_event);
- }
- break;
+ for (i = 0; i < priv->child_count; i++) {
+ eh_dump(shd->statistics[i], output, afr_add_crawl_event);
+ ret = afr_shd_dict_add_crawl_event(
+ this, output, &shd->index_healers[i].crawl_event);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret,
+ AFR_MSG_ADD_CRAWL_EVENT_FAILED, NULL);
+ }
+
+ ret = afr_shd_dict_add_crawl_event(
+ this, output, &shd->full_healers[i].crawl_event);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret,
+ AFR_MSG_ADD_CRAWL_EVENT_FAILED, NULL);
+ }
+ }
+ break;
case GF_SHD_OP_STATISTICS_HEAL_COUNT:
case GF_SHD_OP_STATISTICS_HEAL_COUNT_PER_REPLICA:
- op_ret = -1;
-
- for (i = 0; i < priv->child_count; i++) {
- if (!priv->child_up[i]) {
- snprintf (key, sizeof (key), "%d-%d-status",
- xl_id, i);
- ret = dict_set_str (output, key,
- "Brick is not connected");
- } else {
- snprintf (key, sizeof (key), "%d-%d-hardlinks",
- xl_id, i);
- ret = afr_shd_get_index_count (this, i, &cnt);
- if (ret == 0) {
- ret = dict_set_uint64 (output, key, cnt);
- }
- op_ret = 0;
- }
- }
-
- break;
+ op_ret = -1;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!priv->child_up[i]) {
+ keylen = snprintf(key, sizeof(key), "%d-%d-status", xl_id,
+ i);
+ AFR_SET_DICT_AND_LOG(this->name, output, key, keylen,
+ SBRICK_NOT_CONNECTED,
+ SLEN(SBRICK_NOT_CONNECTED));
+ } else {
+ snprintf(key, sizeof(key), "%d-%d-hardlinks", xl_id, i);
+ ret = afr_shd_get_index_count(this, i, &cnt);
+ if (ret == 0) {
+ ret = dict_set_uint64(output, key, cnt);
+ }
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret,
+ AFR_MSG_DICT_SET_FAILED, NULL);
+ }
+ op_ret = 0;
+ }
+ }
+
+ break;
default:
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_ARG, "Unknown set op %d", op);
- break;
- }
+ gf_smsg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_ARG, "op=%d",
+ op, NULL);
+ break;
+ }
out:
- dict_del (output, this->name);
- return op_ret;
+ dict_deln(output, this->name, this_name_len);
+ return op_ret;
+
+#undef AFR_SET_DICT_AND_LOG
}
diff --git a/xlators/cluster/afr/src/afr-self-heald.h b/xlators/cluster/afr/src/afr-self-heald.h
index 4ac1d32f58a..18db728ea7b 100644
--- a/xlators/cluster/afr/src/afr-self-heald.h
+++ b/xlators/cluster/afr/src/afr-self-heald.h
@@ -8,74 +8,68 @@
cases as published by the Free Software Foundation.
*/
-
#ifndef _AFR_SELF_HEALD_H
#define _AFR_SELF_HEALD_H
#include <pthread.h>
-
typedef struct {
- int child;
- char *path;
+ char *path;
+ int child;
} shd_event_t;
typedef struct {
- int child;
- uint64_t healed_count;
- uint64_t split_brain_count;
- uint64_t heal_failed_count;
-
- /* If start_time is 0, it means crawler is not in progress
- and stats are not valid */
- time_t start_time;
- /* If start_time is NOT 0 and end_time is 0, it means
- cralwer is in progress */
- time_t end_time;
- char *crawl_type;
+ uint64_t healed_count;
+ uint64_t split_brain_count;
+ uint64_t heal_failed_count;
+
+ /* If start_time is 0, it means crawler is not in progress
+ and stats are not valid */
+ time_t start_time;
+ /* If start_time is NOT 0 and end_time is 0, it means
+ cralwer is in progress */
+ time_t end_time;
+ char *crawl_type;
+ int child;
} crawl_event_t;
struct subvol_healer {
- xlator_t *this;
- int subvol;
- gf_boolean_t local;
- gf_boolean_t running;
- gf_boolean_t rerun;
- crawl_event_t crawl_event;
- pthread_mutex_t mutex;
- pthread_cond_t cond;
- pthread_t thread;
+ xlator_t *this;
+ crawl_event_t crawl_event;
+ pthread_mutex_t mutex;
+ pthread_cond_t cond;
+ pthread_t thread;
+ int subvol;
+ gf_boolean_t local;
+ gf_boolean_t running;
+ gf_boolean_t rerun;
};
typedef struct {
- gf_boolean_t iamshd;
- gf_boolean_t enabled;
- int timeout;
- struct subvol_healer *index_healers;
- struct subvol_healer *full_healers;
-
- eh_t *split_brain;
- eh_t **statistics;
- uint32_t max_threads;
- uint32_t wait_qlength;
- uint32_t halo_max_latency_msec;
+ struct subvol_healer *index_healers;
+ struct subvol_healer *full_healers;
+
+ eh_t *split_brain;
+ eh_t **statistics;
+ int timeout;
+ uint32_t max_threads;
+ uint32_t wait_qlength;
+ uint32_t halo_max_latency_msec;
+ gf_boolean_t iamshd;
+ gf_boolean_t enabled;
} afr_self_heald_t;
-
-int
-afr_selfheal_childup (xlator_t *this, int subvol);
-
int
-afr_selfheal_daemon_init (xlator_t *this);
+afr_selfheal_daemon_init(xlator_t *this);
int
-afr_xl_op (xlator_t *this, dict_t *input, dict_t *output);
+afr_xl_op(xlator_t *this, dict_t *input, dict_t *output);
int
-afr_shd_gfid_to_path (xlator_t *this, xlator_t *subvol, uuid_t gfid,
- char **path_p);
+afr_shd_gfid_to_path(xlator_t *this, xlator_t *subvol, uuid_t gfid,
+ char **path_p);
int
-afr_shd_index_purge (xlator_t *subvol, inode_t *inode, char *name,
- ia_type_t type);
+afr_shd_entry_purge(xlator_t *subvol, inode_t *inode, char *name,
+ ia_type_t type);
#endif /* !_AFR_SELF_HEALD_H */
diff --git a/xlators/cluster/afr/src/afr-transaction.c b/xlators/cluster/afr/src/afr-transaction.c
index 9c0963781e3..a51f79b1f43 100644
--- a/xlators/cluster/afr/src/afr-transaction.c
+++ b/xlators/cluster/afr/src/afr-transaction.c
@@ -8,10 +8,10 @@
cases as published by the Free Software Foundation.
*/
-#include "dict.h"
-#include "byte-order.h"
-#include "common-utils.h"
-#include "timer.h"
+#include <glusterfs/dict.h>
+#include <glusterfs/byte-order.h>
+#include <glusterfs/common-utils.h>
+#include <glusterfs/timer.h>
#include "afr.h"
#include "afr-transaction.h"
@@ -21,2491 +21,2907 @@
#include <signal.h>
typedef enum {
- AFR_TRANSACTION_PRE_OP,
- AFR_TRANSACTION_POST_OP,
+ AFR_TRANSACTION_PRE_OP,
+ AFR_TRANSACTION_POST_OP,
} afr_xattrop_type_t;
static void
-afr_lock_resume_shared (struct list_head *list);
+afr_lock_resume_shared(struct list_head *list);
+
+static void
+afr_post_op_handle_success(call_frame_t *frame, xlator_t *this);
+
+static void
+afr_post_op_handle_failure(call_frame_t *frame, xlator_t *this, int op_errno);
void
-__afr_transaction_wake_shared (afr_local_t *local, struct list_head *shared);
+__afr_transaction_wake_shared(afr_local_t *local, struct list_head *shared);
void
-afr_changelog_post_op (call_frame_t *frame, xlator_t *this);
+afr_changelog_post_op_do(call_frame_t *frame, xlator_t *this);
int
-afr_changelog_post_op_safe (call_frame_t *frame, xlator_t *this);
+afr_changelog_post_op_safe(call_frame_t *frame, xlator_t *this);
gf_boolean_t
-afr_changelog_pre_op_uninherit (call_frame_t *frame, xlator_t *this);
+afr_changelog_pre_op_uninherit(call_frame_t *frame, xlator_t *this);
gf_boolean_t
-afr_changelog_pre_op_update (call_frame_t *frame, xlator_t *this);
+afr_changelog_pre_op_update(call_frame_t *frame, xlator_t *this);
int
-afr_changelog_call_count (afr_transaction_type type,
- unsigned char *pre_op_subvols,
- unsigned char *failed_subvols,
- unsigned int child_count);
+afr_changelog_call_count(afr_transaction_type type,
+ unsigned char *pre_op_subvols,
+ unsigned char *failed_subvols,
+ unsigned int child_count);
int
-afr_changelog_do (call_frame_t *frame, xlator_t *this, dict_t *xattr,
- afr_changelog_resume_t changelog_resume,
- afr_xattrop_type_t op);
+afr_changelog_do(call_frame_t *frame, xlator_t *this, dict_t *xattr,
+ afr_changelog_resume_t changelog_resume,
+ afr_xattrop_type_t op);
-void
-afr_zero_fill_stat (afr_local_t *local)
-{
- if (!local)
- return;
- if (local->transaction.type == AFR_DATA_TRANSACTION ||
- local->transaction.type == AFR_METADATA_TRANSACTION) {
- gf_zero_fill_stat (&local->cont.inode_wfop.prebuf);
- gf_zero_fill_stat (&local->cont.inode_wfop.postbuf);
- } else if (local->transaction.type == AFR_ENTRY_TRANSACTION ||
- local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
- gf_zero_fill_stat (&local->cont.dir_fop.buf);
- gf_zero_fill_stat (&local->cont.dir_fop.preparent);
- gf_zero_fill_stat (&local->cont.dir_fop.postparent);
- if (local->transaction.type == AFR_ENTRY_TRANSACTION)
- return;
- gf_zero_fill_stat (&local->cont.dir_fop.prenewparent);
- gf_zero_fill_stat (&local->cont.dir_fop.postnewparent);
- }
-}
+static void
+afr_ta_decide_post_op_state(call_frame_t *frame, xlator_t *this);
+
+static int
+afr_ta_post_op_do(void *opaque);
+
+static int
+afr_ta_post_op_synctask(xlator_t *this, afr_local_t *local);
+
+static int
+afr_changelog_post_op_done(call_frame_t *frame, xlator_t *this);
+
+static void
+afr_changelog_post_op_fail(call_frame_t *frame, xlator_t *this, int op_errno);
-/* In case of errors afr needs to choose which xdata from lower xlators it needs
- * to unwind with. The way it is done is by checking if there are
- * any good subvols which failed. Give preference to errnos other than
- * ENOTCONN even if the child is source */
void
-afr_pick_error_xdata (afr_local_t *local, afr_private_t *priv,
- inode_t *inode1, unsigned char *readable1,
- inode_t *inode2, unsigned char *readable2)
+afr_ta_locked_priv_invalidate(afr_private_t *priv)
{
- int s = -1;/*selection*/
- int i = 0;
- unsigned char *readable = NULL;
-
- if (local->xdata_rsp) {
- dict_unref (local->xdata_rsp);
- local->xdata_rsp = NULL;
- }
+ priv->ta_bad_child_index = AFR_CHILD_UNKNOWN;
+ priv->release_ta_notify_dom_lock = _gf_false;
+ priv->ta_notify_dom_lock_offset = 0;
+}
- readable = alloca0 (priv->child_count * sizeof (*readable));
- if (inode2 && readable2) {/*rename fop*/
- AFR_INTERSECT (readable, readable1, readable2,
- priv->child_count);
- } else {
- memcpy (readable, readable1,
- sizeof (*readable) * priv->child_count);
- }
+static void
+afr_ta_process_waitq(xlator_t *this)
+{
+ afr_local_t *entry = NULL;
+ afr_private_t *priv = this->private;
+ struct list_head waitq = {
+ 0,
+ };
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
+ INIT_LIST_HEAD(&waitq);
+ LOCK(&priv->lock);
+ list_splice_init(&priv->ta_waitq, &waitq);
+ UNLOCK(&priv->lock);
+ list_for_each_entry(entry, &waitq, ta_waitq)
+ {
+ afr_ta_decide_post_op_state(entry->transaction.frame, this);
+ }
+}
- if (local->replies[i].op_ret >= 0)
- continue;
+int
+afr_ta_lock_release_done(int ret, call_frame_t *ta_frame, void *opaque)
+{
+ afr_ta_process_waitq(ta_frame->this);
+ STACK_DESTROY(ta_frame->root);
+ return 0;
+}
- if (local->replies[i].op_errno == ENOTCONN)
- continue;
+int
+afr_release_notify_lock_for_ta(void *opaque)
+{
+ xlator_t *this = NULL;
+ afr_private_t *priv = NULL;
+ loc_t loc = {
+ 0,
+ };
+ struct gf_flock flock = {
+ 0,
+ };
+ int ret = -1;
+
+ this = (xlator_t *)opaque;
+ priv = this->private;
+ ret = afr_fill_ta_loc(this, &loc, _gf_true);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to populate loc for thin-arbiter.");
+ goto out;
+ }
+ flock.l_type = F_UNLCK;
+ flock.l_start = priv->ta_notify_dom_lock_offset;
+ flock.l_len = 1;
+ ret = syncop_inodelk(priv->children[THIN_ARBITER_BRICK_INDEX],
+ AFR_TA_DOM_NOTIFY, &loc, F_SETLK, &flock, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to unlock AFR_TA_DOM_NOTIFY lock.");
+ }
+
+ LOCK(&priv->lock);
+ {
+ afr_ta_locked_priv_invalidate(priv);
+ }
+ UNLOCK(&priv->lock);
+out:
+ loc_wipe(&loc);
+ return ret;
+}
- /*Order is important in the following condition*/
- if ((s < 0) || (!readable[s] && readable[i]))
- s = i;
- }
+void
+afr_zero_fill_stat(afr_local_t *local)
+{
+ if (!local)
+ return;
+ if (local->transaction.type == AFR_DATA_TRANSACTION ||
+ local->transaction.type == AFR_METADATA_TRANSACTION) {
+ gf_zero_fill_stat(&local->cont.inode_wfop.prebuf);
+ gf_zero_fill_stat(&local->cont.inode_wfop.postbuf);
+ } else if (local->transaction.type == AFR_ENTRY_TRANSACTION ||
+ local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
+ gf_zero_fill_stat(&local->cont.dir_fop.buf);
+ gf_zero_fill_stat(&local->cont.dir_fop.preparent);
+ gf_zero_fill_stat(&local->cont.dir_fop.postparent);
+ if (local->transaction.type == AFR_ENTRY_TRANSACTION)
+ return;
+ gf_zero_fill_stat(&local->cont.dir_fop.prenewparent);
+ gf_zero_fill_stat(&local->cont.dir_fop.postnewparent);
+ }
+}
- if (s != -1 && local->replies[s].xdata) {
- local->xdata_rsp = dict_ref (local->replies[s].xdata);
- } else if (s == -1) {
- for (i = 0; i < priv->child_count; i++) {
- if (!local->replies[i].valid)
- continue;
+/* In case of errors afr needs to choose which xdata from lower xlators it needs
+ * to unwind with. The way it is done is by checking if there are
+ * any good subvols which failed. Give preference to errnos other than
+ * ENOTCONN even if the child is source */
+void
+afr_pick_error_xdata(afr_local_t *local, afr_private_t *priv, inode_t *inode1,
+ unsigned char *readable1, inode_t *inode2,
+ unsigned char *readable2)
+{
+ int s = -1; /*selection*/
+ int i = 0;
+ unsigned char *readable = NULL;
+
+ if (local->xdata_rsp) {
+ dict_unref(local->xdata_rsp);
+ local->xdata_rsp = NULL;
+ }
+
+ readable = alloca0(priv->child_count * sizeof(*readable));
+ if (inode2 && readable2) { /*rename fop*/
+ AFR_INTERSECT(readable, readable1, readable2, priv->child_count);
+ } else {
+ memcpy(readable, readable1, sizeof(*readable) * priv->child_count);
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
+
+ if (local->replies[i].op_ret >= 0)
+ continue;
+
+ if (local->replies[i].op_errno == ENOTCONN)
+ continue;
+
+ /*Order is important in the following condition*/
+ if ((s < 0) || (!readable[s] && readable[i]))
+ s = i;
+ }
+
+ if (s != -1 && local->replies[s].xdata) {
+ local->xdata_rsp = dict_ref(local->replies[s].xdata);
+ } else if (s == -1) {
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->replies[i].valid)
+ continue;
- if (local->replies[i].op_ret >= 0)
- continue;
+ if (local->replies[i].op_ret >= 0)
+ continue;
- if (!local->replies[i].xdata)
- continue;
- local->xdata_rsp = dict_ref (local->replies[i].xdata);
- break;
- }
+ if (!local->replies[i].xdata)
+ continue;
+ local->xdata_rsp = dict_ref(local->replies[i].xdata);
+ break;
}
+ }
}
gf_boolean_t
-afr_needs_changelog_update (afr_local_t *local)
+afr_needs_changelog_update(afr_local_t *local)
{
- if (local->transaction.type == AFR_DATA_TRANSACTION)
- return _gf_true;
- if (!local->optimistic_change_log)
- return _gf_true;
- return _gf_false;
+ if (local->transaction.type == AFR_DATA_TRANSACTION)
+ return _gf_true;
+ if (!local->optimistic_change_log)
+ return _gf_true;
+ return _gf_false;
}
gf_boolean_t
-afr_changelog_has_quorum (afr_local_t *local, xlator_t *this)
+afr_changelog_has_quorum(afr_local_t *local, xlator_t *this)
{
- afr_private_t *priv = NULL;
- int i = 0;
- unsigned char *success_children = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ unsigned char *success_children = NULL;
- priv = this->private;
- success_children = alloca0 (priv->child_count);
+ priv = this->private;
+ success_children = alloca0(priv->child_count);
- for (i = 0; i < priv->child_count; i++) {
- if (!local->transaction.failed_subvols[i]) {
- success_children[i] = 1;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.failed_subvols[i]) {
+ success_children[i] = 1;
}
+ }
- if (afr_has_quorum (success_children, this)) {
- return _gf_true;
- }
+ if (afr_has_quorum(success_children, this, NULL)) {
+ return _gf_true;
+ }
- return _gf_false;
+ return _gf_false;
}
-
gf_boolean_t
-afr_is_write_subvol_valid (call_frame_t *frame, xlator_t *this)
+afr_is_write_subvol_valid(call_frame_t *frame, xlator_t *this)
{
- int i = 0;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- uint64_t write_subvol = 0;
- unsigned char *writable = NULL;
- uint16_t datamap = 0;
+ int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ uint64_t write_subvol = 0;
+ unsigned char *writable = NULL;
+ uint16_t datamap = 0;
- local = frame->local;
- priv = this->private;
- writable = alloca0 (priv->child_count);
+ local = frame->local;
+ priv = this->private;
+ writable = alloca0(priv->child_count);
- write_subvol = afr_write_subvol_get (frame, this);
- datamap = (write_subvol & 0x00000000ffff0000) >> 16;
- for (i = 0; i < priv->child_count; i++) {
- if (datamap & (1 << i))
- writable[i] = 1;
+ write_subvol = afr_write_subvol_get(frame, this);
+ datamap = (write_subvol & 0x00000000ffff0000) >> 16;
+ for (i = 0; i < priv->child_count; i++) {
+ if (datamap & (1 << i))
+ writable[i] = 1;
- if (writable[i] && !local->transaction.failed_subvols[i])
- return _gf_true;
- }
+ if (writable[i] && !local->transaction.failed_subvols[i])
+ return _gf_true;
+ }
- return _gf_false;
+ return _gf_false;
}
int
-afr_transaction_fop (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int call_count = -1;
- unsigned char *failed_subvols = NULL;
- int i = 0;
-
- local = frame->local;
- priv = this->private;
-
- failed_subvols = local->transaction.failed_subvols;
- call_count = priv->child_count - AFR_COUNT (failed_subvols,
- priv->child_count);
- /* Fail if pre-op did not succeed on quorum no. of bricks. */
- if (!afr_changelog_has_quorum (local, this) || !call_count) {
- local->op_ret = -1;
- /* local->op_errno is already captured in changelog cbk. */
- afr_transaction_resume (frame, this);
- return 0;
- }
+afr_transaction_fop(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = -1;
+ unsigned char *failed_subvols = NULL;
+ int i = 0;
+
+ local = frame->local;
+ priv = this->private;
+
+ failed_subvols = local->transaction.failed_subvols;
+ call_count = priv->child_count -
+ AFR_COUNT(failed_subvols, priv->child_count);
+ /* Fail if pre-op did not succeed on quorum no. of bricks. */
+ if (!afr_changelog_has_quorum(local, this) || !call_count) {
+ local->op_ret = -1;
+ /* local->op_errno is already captured in changelog cbk. */
+ afr_transaction_resume(frame, this);
+ return 0;
+ }
- /* Fail if at least one writeable brick isn't up.*/
- if (local->transaction.type == AFR_DATA_TRANSACTION &&
- !afr_is_write_subvol_valid (frame, this)) {
- local->op_ret = -1;
- local->op_errno = EIO;
- afr_transaction_resume (frame, this);
- return 0;
- }
+ /* Fail if at least one writeable brick isn't up.*/
+ if (local->transaction.type == AFR_DATA_TRANSACTION &&
+ !afr_is_write_subvol_valid(frame, this)) {
+ local->op_ret = -1;
+ local->op_errno = EIO;
+ afr_transaction_resume(frame, this);
+ return 0;
+ }
- local->call_count = call_count;
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.pre_op[i] && !failed_subvols[i]) {
- local->transaction.wind (frame, this, i);
+ local->call_count = call_count;
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i] && !failed_subvols[i]) {
+ local->transaction.wind(frame, this, i);
- if (!--call_count)
- break;
- }
+ if (!--call_count)
+ break;
}
+ }
- return 0;
+ return 0;
}
int
-afr_transaction_done (call_frame_t *frame, xlator_t *this)
+afr_transaction_done(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- gf_boolean_t unwind = _gf_false;
- afr_lock_t *lock = NULL;
- afr_local_t *lock_local = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ gf_boolean_t unwind = _gf_false;
+ afr_lock_t *lock = NULL;
+ afr_local_t *lock_local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- if (priv->consistent_metadata) {
- LOCK (&frame->lock);
- {
- unwind = (local->transaction.main_frame != NULL);
- }
- UNLOCK (&frame->lock);
- if (unwind)/*It definitely did post-op*/
- afr_zero_fill_stat (local);
+ if (priv->consistent_metadata) {
+ LOCK(&frame->lock);
+ {
+ unwind = (local->transaction.main_frame != NULL);
}
+ UNLOCK(&frame->lock);
+ if (unwind) /*It definitely did post-op*/
+ afr_zero_fill_stat(local);
+ }
- if (local->transaction.do_eager_unlock) {
- lock = &local->inode_ctx->lock[local->transaction.type];
- LOCK (&local->inode->lock);
- {
- lock->acquired = _gf_false;
- lock->release = _gf_false;
- list_splice_init (&lock->frozen,
- &lock->waiting);
- if (list_empty (&lock->waiting))
- goto unlock;
- lock_local = list_entry (lock->waiting.next,
- afr_local_t,
- transaction.wait_list);
- list_del_init (&lock_local->transaction.wait_list);
- list_add (&lock_local->transaction.owner_list,
- &lock->owners);
- }
-unlock:
- UNLOCK (&local->inode->lock);
- }
- if (lock_local) {
- afr_lock (lock_local->transaction.frame,
- lock_local->transaction.frame->this);
- }
- local->transaction.unwind (frame, this);
+ if (local->transaction.do_eager_unlock) {
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ LOCK(&local->inode->lock);
+ {
+ lock->acquired = _gf_false;
+ lock->release = _gf_false;
+ list_splice_init(&lock->frozen, &lock->waiting);
+ if (list_empty(&lock->waiting))
+ goto unlock;
+ lock_local = list_entry(lock->waiting.next, afr_local_t,
+ transaction.wait_list);
+ list_del_init(&lock_local->transaction.wait_list);
+ list_add(&lock_local->transaction.owner_list, &lock->owners);
+ }
+ unlock:
+ UNLOCK(&local->inode->lock);
+ }
+ if (lock_local) {
+ afr_lock(lock_local->transaction.frame,
+ lock_local->transaction.frame->this);
+ }
+ local->transaction.unwind(frame, this);
- AFR_STACK_DESTROY (frame);
+ GF_ASSERT(list_empty(&local->transaction.owner_list));
+ GF_ASSERT(list_empty(&local->transaction.wait_list));
+ AFR_STACK_DESTROY(frame);
- return 0;
+ return 0;
}
static void
-afr_lock_fail_shared (afr_local_t *local, struct list_head *list)
-{
- afr_local_t *each = NULL;
-
- while (!list_empty(list)) {
- each = list_entry (list->next, afr_local_t,
- transaction.wait_list);
- list_del_init(&each->transaction.wait_list);
- each->op_ret = -1;
- each->op_errno = local->op_errno;
- afr_transaction_done (each->transaction.frame,
- each->transaction.frame->this);
- }
+afr_lock_fail_shared(afr_local_t *local, struct list_head *list)
+{
+ afr_local_t *each = NULL;
+
+ while (!list_empty(list)) {
+ each = list_entry(list->next, afr_local_t, transaction.wait_list);
+ list_del_init(&each->transaction.wait_list);
+ each->op_ret = -1;
+ each->op_errno = local->op_errno;
+ afr_transaction_done(each->transaction.frame,
+ each->transaction.frame->this);
+ }
}
static void
-afr_handle_lock_acquire_failure (afr_local_t *local, gf_boolean_t locked)
+afr_handle_lock_acquire_failure(afr_local_t *local)
{
- struct list_head shared;
- afr_lock_t *lock = NULL;
+ struct list_head shared;
+ afr_lock_t *lock = NULL;
- if (!local->transaction.eager_lock_on)
- goto out;
+ if (!local->transaction.eager_lock_on)
+ goto out;
- lock = &local->inode_ctx->lock[local->transaction.type];
+ lock = &local->inode_ctx->lock[local->transaction.type];
- INIT_LIST_HEAD (&shared);
- LOCK (&local->inode->lock);
- {
- lock->release = _gf_true;
- list_splice_init (&lock->waiting, &shared);
- }
- UNLOCK (&local->inode->lock);
+ INIT_LIST_HEAD(&shared);
+ LOCK(&local->inode->lock);
+ {
+ lock->release = _gf_true;
+ list_splice_init(&lock->waiting, &shared);
+ }
+ UNLOCK(&local->inode->lock);
- afr_lock_fail_shared (local, &shared);
- local->transaction.do_eager_unlock = _gf_true;
+ afr_lock_fail_shared(local, &shared);
+ local->transaction.do_eager_unlock = _gf_true;
out:
- if (locked) {
- local->internal_lock.lock_cbk = afr_transaction_done;
- afr_unlock (local->transaction.frame,
- local->transaction.frame->this);
- } else {
- afr_transaction_done (local->transaction.frame,
- local->transaction.frame->this);
- }
+ local->internal_lock.lock_cbk = afr_transaction_done;
+ afr_unlock(local->transaction.frame, local->transaction.frame->this);
}
-call_frame_t*
-afr_transaction_detach_fop_frame (call_frame_t *frame)
+call_frame_t *
+afr_transaction_detach_fop_frame(call_frame_t *frame)
{
- afr_local_t * local = NULL;
- call_frame_t *fop_frame = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *fop_frame = NULL;
- local = frame->local;
+ local = frame->local;
- afr_handle_inconsistent_fop (frame, &local->op_ret, &local->op_errno);
- LOCK (&frame->lock);
- {
- fop_frame = local->transaction.main_frame;
- local->transaction.main_frame = NULL;
- }
- UNLOCK (&frame->lock);
+ afr_handle_inconsistent_fop(frame, &local->op_ret, &local->op_errno);
+ LOCK(&frame->lock);
+ {
+ fop_frame = local->transaction.main_frame;
+ local->transaction.main_frame = NULL;
+ }
+ UNLOCK(&frame->lock);
- return fop_frame;
+ return fop_frame;
}
-
static void
-afr_save_lk_owner (call_frame_t *frame)
+afr_save_lk_owner(call_frame_t *frame)
{
- afr_local_t * local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- local->saved_lk_owner = frame->root->lk_owner;
+ local->saved_lk_owner = frame->root->lk_owner;
}
-
static void
-afr_restore_lk_owner (call_frame_t *frame)
+afr_restore_lk_owner(call_frame_t *frame)
{
- afr_local_t * local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- frame->root->lk_owner = local->saved_lk_owner;
+ frame->root->lk_owner = local->saved_lk_owner;
}
void
-__mark_all_success (call_frame_t *frame, xlator_t *this)
+__mark_all_success(call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- local->transaction.failed_subvols[i] = 0;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ local->transaction.failed_subvols[i] = 0;
+ }
}
void
-afr_compute_pre_op_sources (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_transaction_type type = -1;
- dict_t *xdata = NULL;
- int **matrix = NULL;
- int idx = -1;
- int i = 0;
- int j = 0;
-
- priv = this->private;
- local = frame->local;
- type = local->transaction.type;
- idx = afr_index_for_transaction_type (type);
- matrix = ALLOC_MATRIX (priv->child_count, int);
-
- for (i = 0; i < priv->child_count; i++) {
- if (!local->transaction.changelog_xdata[i])
- continue;
- xdata = local->transaction.changelog_xdata[i];
- afr_selfheal_fill_matrix (this, matrix, i, idx, xdata);
- }
-
- memset (local->transaction.pre_op_sources, 1, priv->child_count);
-
- /*If lock or pre-op failed on a brick, it is not a source. */
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.failed_subvols[i])
- local->transaction.pre_op_sources[i] = 0;
- }
-
- /* If brick is blamed by others, it is not a source. */
- for (i = 0; i < priv->child_count; i++)
- for (j = 0; j < priv->child_count; j++)
- if (matrix[i][j] != 0)
- local->transaction.pre_op_sources[j] = 0;
+afr_compute_pre_op_sources(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_transaction_type type = -1;
+ dict_t *xdata = NULL;
+ int **matrix = NULL;
+ int idx = -1;
+ int i = 0;
+ int j = 0;
+
+ priv = this->private;
+ local = frame->local;
+ type = local->transaction.type;
+ idx = afr_index_for_transaction_type(type);
+ matrix = ALLOC_MATRIX(priv->child_count, int);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.changelog_xdata[i])
+ continue;
+ xdata = local->transaction.changelog_xdata[i];
+ afr_selfheal_fill_matrix(this, matrix, i, idx, xdata);
+ }
+
+ memset(local->transaction.pre_op_sources, 1, priv->child_count);
+
+ /*If lock or pre-op failed on a brick, it is not a source. */
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.failed_subvols[i])
+ local->transaction.pre_op_sources[i] = 0;
+ }
+
+ /* If brick is blamed by others, it is not a source. */
+ for (i = 0; i < priv->child_count; i++)
+ for (j = 0; j < priv->child_count; j++)
+ if (matrix[i][j] != 0)
+ local->transaction.pre_op_sources[j] = 0;
}
-gf_boolean_t
-afr_has_arbiter_fop_cbk_quorum (call_frame_t *frame)
+void
+afr_txn_arbitrate_fop(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- xlator_t *this = NULL;
- gf_boolean_t fop_failed = _gf_false;
- unsigned char *pre_op_sources = NULL;
- int i = 0;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int pre_op_sources_count = 0;
+ int i = 0;
- local = frame->local;
- this = frame->this;
- priv = this->private;
- pre_op_sources = local->transaction.pre_op_sources;
+ priv = this->private;
+ local = frame->local;
- /* If the fop failed on the brick, it is not a source. */
- for (i = 0; i < priv->child_count; i++)
- if (local->transaction.failed_subvols[i])
- pre_op_sources[i] = 0;
+ afr_compute_pre_op_sources(frame, this);
+ pre_op_sources_count = AFR_COUNT(local->transaction.pre_op_sources,
+ priv->child_count);
- switch (AFR_COUNT (pre_op_sources, priv->child_count)) {
- case 1:
- if (pre_op_sources[ARBITER_BRICK_INDEX])
- fop_failed = _gf_true;
- break;
- case 0:
- fop_failed = _gf_true;
- break;
- }
-
- if (fop_failed)
- return _gf_false;
-
- return _gf_true;
-}
+ /* If arbiter is the only source, do not proceed. */
+ if (pre_op_sources_count < 2 &&
+ local->transaction.pre_op_sources[ARBITER_BRICK_INDEX]) {
+ local->op_ret = -1;
+ local->op_errno = ENOTCONN;
+ for (i = 0; i < priv->child_count; i++)
+ local->transaction.failed_subvols[i] = 1;
+ }
-void
-afr_txn_arbitrate_fop (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int pre_op_sources_count = 0;
- int i = 0;
-
- priv = this->private;
- local = frame->local;
-
- afr_compute_pre_op_sources (frame, this);
- pre_op_sources_count = AFR_COUNT (local->transaction.pre_op_sources,
- priv->child_count);
-
- /* If arbiter is the only source, do not proceed. */
- if (pre_op_sources_count < 2 &&
- local->transaction.pre_op_sources[ARBITER_BRICK_INDEX]) {
- local->op_ret = -1;
- local->op_errno = ENOTCONN;
- for (i = 0; i < priv->child_count; i++)
- local->transaction.failed_subvols[i] = 1;
- afr_changelog_post_op (frame, this);/*uninherit should happen*/
- } else {
- afr_transaction_fop (frame, this);
- }
+ afr_transaction_fop(frame, this);
- return;
+ return;
}
int
-afr_transaction_perform_fop (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- int ret = 0;
- int failure_count = 0;
- struct list_head shared;
- afr_lock_t *lock = NULL;
-
- local = frame->local;
- priv = this->private;
-
- INIT_LIST_HEAD (&shared);
- if (local->transaction.type == AFR_DATA_TRANSACTION &&
- !local->transaction.inherited) {
- ret = afr_write_subvol_set (frame, this);
- if (ret) {
- /*act as if operation failed on all subvols*/
- local->op_ret = -1;
- local->op_errno = -ret;
- for (i = 0; i < priv->child_count; i++)
- local->transaction.failed_subvols[i] = 1;
- }
- }
- /* Perform fops with the lk-owner from top xlator.
- * Eg: lk-owner of posix-lk and flush should be same,
- * flush cant clear the posix-lks without that lk-owner.
- */
- afr_save_lk_owner (frame);
- frame->root->lk_owner =
- local->transaction.main_frame->root->lk_owner;
-
- if (local->pre_op_compat)
- /* old mode, pre-op was done as afr_changelog_do()
- just now, before OP */
- afr_changelog_pre_op_update (frame, this);
-
- if (!local->transaction.eager_lock_on ||
- local->transaction.inherited)
- goto fop;
- failure_count = AFR_COUNT (local->transaction.failed_subvols,
- priv->child_count);
- if (failure_count == priv->child_count) {
- afr_handle_lock_acquire_failure (local, _gf_true);
- return 0;
- } else {
- lock = &local->inode_ctx->lock[local->transaction.type];
- LOCK (&local->inode->lock);
- {
- lock->acquired = _gf_true;
- __afr_transaction_wake_shared (local, &shared);
- }
- UNLOCK (&local->inode->lock);
+afr_transaction_perform_fop(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ int ret = 0;
+ int failure_count = 0;
+ struct list_head shared;
+ afr_lock_t *lock = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ INIT_LIST_HEAD(&shared);
+ if (local->transaction.type == AFR_DATA_TRANSACTION &&
+ !local->transaction.inherited) {
+ ret = afr_write_subvol_set(frame, this);
+ if (ret) {
+ /*act as if operation failed on all subvols*/
+ local->op_ret = -1;
+ local->op_errno = -ret;
+ for (i = 0; i < priv->child_count; i++)
+ local->transaction.failed_subvols[i] = 1;
+ }
+ }
+
+ if (local->pre_op_compat)
+ /* old mode, pre-op was done as afr_changelog_do()
+ just now, before OP */
+ afr_changelog_pre_op_update(frame, this);
+
+ if (!local->transaction.eager_lock_on || local->transaction.inherited)
+ goto fop;
+ failure_count = AFR_COUNT(local->transaction.failed_subvols,
+ priv->child_count);
+ if (failure_count == priv->child_count) {
+ afr_handle_lock_acquire_failure(local);
+ return 0;
+ } else {
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ LOCK(&local->inode->lock);
+ {
+ lock->acquired = _gf_true;
+ __afr_transaction_wake_shared(local, &shared);
}
+ UNLOCK(&local->inode->lock);
+ }
fop:
- if (priv->arbiter_count == 1) {
- afr_txn_arbitrate_fop (frame, this);
- } else {
- afr_transaction_fop (frame, this);
- }
+ /* Perform fops with the lk-owner from top xlator.
+ * Eg: lk-owner of posix-lk and flush should be same,
+ * flush cant clear the posix-lks without that lk-owner.
+ */
+ afr_save_lk_owner(frame);
+ frame->root->lk_owner = local->transaction.main_frame->root->lk_owner;
- afr_lock_resume_shared (&shared);
- return 0;
+ if (priv->arbiter_count == 1) {
+ afr_txn_arbitrate_fop(frame, this);
+ } else {
+ afr_transaction_fop(frame, this);
+ }
+
+ afr_lock_resume_shared(&shared);
+ return 0;
}
int
-afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int **pending)
+afr_set_pending_dict(afr_private_t *priv, dict_t *xattr, int **pending)
{
- int i = 0;
- int ret = 0;
+ int i = 0;
+ int ret = 0;
- for (i = 0; i < priv->child_count; i++) {
+ for (i = 0; i < priv->child_count; i++) {
+ ret = dict_set_static_bin(xattr, priv->pending_key[i], pending[i],
+ AFR_NUM_CHANGE_LOGS * sizeof(int));
+ /* 3 = data+metadata+entry */
+
+ if (ret)
+ break;
+ }
- ret = dict_set_static_bin (xattr, priv->pending_key[i],
- pending[i],
- AFR_NUM_CHANGE_LOGS * sizeof (int));
- /* 3 = data+metadata+entry */
+ return ret;
+}
- if (ret)
- break;
+static void
+afr_ta_dom_lock_check_and_release(afr_ta_fop_state_t fop_state, xlator_t *this)
+{
+ afr_private_t *priv = this->private;
+ unsigned int inmem_count = 0;
+ unsigned int onwire_count = 0;
+ gf_boolean_t release = _gf_false;
+
+ LOCK(&priv->lock);
+ {
+ /*Once we get notify lock release upcall notification,
+ if any of the fop state counters are non-zero, we will
+ not release the lock.
+ */
+ onwire_count = priv->ta_on_wire_txn_count;
+ inmem_count = priv->ta_in_mem_txn_count;
+ switch (fop_state) {
+ case TA_GET_INFO_FROM_TA_FILE:
+ onwire_count = --priv->ta_on_wire_txn_count;
+ break;
+ case TA_INFO_IN_MEMORY_SUCCESS:
+ case TA_INFO_IN_MEMORY_FAILED:
+ inmem_count = --priv->ta_in_mem_txn_count;
+ break;
+ case TA_WAIT_FOR_NOTIFY_LOCK_REL:
+ GF_ASSERT(0);
+ break;
+ case TA_SUCCESS:
+ break;
}
+ release = priv->release_ta_notify_dom_lock;
+ }
+ UNLOCK(&priv->lock);
+
+ if (inmem_count != 0 || release == _gf_false || onwire_count != 0)
+ return;
- return ret;
+ afr_ta_lock_release_synctask(this);
}
-/* {{{ pending */
+static void
+afr_ta_process_onwireq(afr_ta_fop_state_t fop_state, xlator_t *this)
+{
+ afr_private_t *priv = this->private;
+ afr_local_t *entry = NULL;
+ int bad_child = AFR_CHILD_UNKNOWN;
+
+ struct list_head onwireq = {
+ 0,
+ };
+ INIT_LIST_HEAD(&onwireq);
+
+ LOCK(&priv->lock);
+ {
+ bad_child = priv->ta_bad_child_index;
+ if (bad_child == AFR_CHILD_UNKNOWN) {
+ /*The previous on-wire ta_post_op was a failure. Just dequeue
+ *one element to wind on-wire again. */
+ entry = list_entry(priv->ta_onwireq.next, afr_local_t, ta_onwireq);
+ list_del_init(&entry->ta_onwireq);
+ } else {
+ /* Prepare to process all fops based on bad_child_index. */
+ list_splice_init(&priv->ta_onwireq, &onwireq);
+ }
+ }
+ UNLOCK(&priv->lock);
+
+ if (entry) {
+ afr_ta_post_op_synctask(this, entry);
+ return;
+ } else {
+ while (!list_empty(&onwireq)) {
+ entry = list_entry(onwireq.next, afr_local_t, ta_onwireq);
+ list_del_init(&entry->ta_onwireq);
+ if (entry->ta_failed_subvol == bad_child) {
+ afr_post_op_handle_success(entry->transaction.frame, this);
+ } else {
+ afr_post_op_handle_failure(entry->transaction.frame, this, EIO);
+ }
+ }
+ }
+}
int
-afr_changelog_post_op_done (call_frame_t *frame, xlator_t *this)
+afr_changelog_post_op_done(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- int_lock = &local->internal_lock;
+ local = frame->local;
+ priv = this->private;
+ int_lock = &local->internal_lock;
- /* Fail the FOP if post-op did not succeed on quorum no. of bricks. */
- if (!afr_changelog_has_quorum (local, this)) {
- local->op_ret = -1;
- /*local->op_errno is already captured in changelog cbk*/
- }
+ if (priv->thin_arbiter_count) {
+ /*fop should not come here with TA_WAIT_FOR_NOTIFY_LOCK_REL state */
+ afr_ta_dom_lock_check_and_release(local->fop_state, this);
+ }
- if (local->transaction.resume_stub) {
- call_resume (local->transaction.resume_stub);
- local->transaction.resume_stub = NULL;
- }
+ /* Fail the FOP if post-op did not succeed on quorum no. of bricks. */
+ if (!afr_changelog_has_quorum(local, this)) {
+ local->op_ret = -1;
+ /*local->op_errno is already captured in changelog cbk*/
+ }
- int_lock->lock_cbk = afr_transaction_done;
- afr_unlock (frame, this);
+ if (local->transaction.resume_stub) {
+ call_resume(local->transaction.resume_stub);
+ local->transaction.resume_stub = NULL;
+ }
- return 0;
-}
+ int_lock->lock_cbk = afr_transaction_done;
+ afr_unlock(frame, this);
+ return 0;
+}
-unsigned char*
-afr_locked_nodes_get (afr_transaction_type type, afr_internal_lock_t *int_lock)
+static void
+afr_changelog_post_op_fail(call_frame_t *frame, xlator_t *this, int op_errno)
{
- unsigned char *locked_nodes = NULL;
- switch (type) {
- case AFR_DATA_TRANSACTION:
- case AFR_METADATA_TRANSACTION:
- locked_nodes = int_lock->locked_nodes;
- break;
+ afr_local_t *local = frame->local;
+ local->op_ret = -1;
+ local->op_errno = op_errno;
- case AFR_ENTRY_TRANSACTION:
- case AFR_ENTRY_RENAME_TRANSACTION:
- /*Because same set of subvols participate in all lockee
- * entities*/
- locked_nodes = int_lock->lockee[0].locked_nodes;
- break;
- }
- return locked_nodes;
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, AFR_MSG_THIN_ARB,
+ "Failing %s for gfid %s. Fop state is:%d", gf_fop_list[local->op],
+ uuid_utoa(local->inode->gfid), local->fop_state);
+
+ afr_changelog_post_op_done(frame, this);
}
+unsigned char *
+afr_locked_nodes_get(afr_transaction_type type, afr_internal_lock_t *int_lock)
+{
+ /*Because same set of subvols participate in all lockee
+ * entities*/
+ return int_lock->lockee[0].locked_nodes;
+}
int
-afr_changelog_call_count (afr_transaction_type type,
- unsigned char *pre_op_subvols,
- unsigned char *failed_subvols,
- unsigned int child_count)
+afr_changelog_call_count(afr_transaction_type type,
+ unsigned char *pre_op_subvols,
+ unsigned char *failed_subvols,
+ unsigned int child_count)
{
- int i = 0;
- int call_count = 0;
+ int i = 0;
+ int call_count = 0;
- for (i = 0; i < child_count; i++) {
- if (pre_op_subvols[i] && !failed_subvols[i]) {
- call_count++;
- }
+ for (i = 0; i < child_count; i++) {
+ if (pre_op_subvols[i] && !failed_subvols[i]) {
+ call_count++;
}
+ }
- if (type == AFR_ENTRY_RENAME_TRANSACTION)
- call_count *= 2;
+ if (type == AFR_ENTRY_RENAME_TRANSACTION)
+ call_count *= 2;
- return call_count;
+ return call_count;
}
-
gf_boolean_t
-afr_txn_nothing_failed (call_frame_t *frame, xlator_t *this)
+afr_txn_nothing_failed(call_frame_t *frame, xlator_t *this)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int i = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- if (priv->thin_arbiter_count) {
- /* We need to perform post-op even if 1 data brick was down
- * before the txn started.*/
- if (AFR_COUNT (local->transaction.failed_subvols,
- priv->child_count))
- return _gf_false;
- }
+ if (priv->thin_arbiter_count) {
+ /* We need to perform post-op even if 1 data brick was down
+ * before the txn started.*/
+ if (AFR_COUNT(local->transaction.failed_subvols, priv->child_count))
+ return _gf_false;
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.pre_op[i] &&
- local->transaction.failed_subvols[i])
- return _gf_false;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i] &&
+ local->transaction.failed_subvols[i])
+ return _gf_false;
+ }
- return _gf_true;
+ return _gf_true;
}
void
-afr_handle_symmetric_errors (call_frame_t *frame, xlator_t *this)
+afr_handle_symmetric_errors(call_frame_t *frame, xlator_t *this)
{
- if (afr_is_symmetric_error (frame, this))
- __mark_all_success (frame, this);
+ if (afr_is_symmetric_error(frame, this))
+ __mark_all_success(frame, this);
}
gf_boolean_t
-afr_has_quorum (unsigned char *subvols, xlator_t *this)
-{
- unsigned int quorum_count = 0;
- afr_private_t *priv = NULL;
- unsigned int up_children_count = 0;
-
- priv = this->private;
- up_children_count = AFR_COUNT (subvols, priv->child_count);
-
- if (priv->quorum_count == AFR_QUORUM_AUTO) {
- /*
- * Special case for auto-quorum with an even number of nodes.
- *
- * A replica set with even count N can only handle the same
- * number of failures as odd N-1 before losing "vanilla"
- * quorum, and the probability of more simultaneous failures is
- * actually higher. For example, with a 1% chance of failure
- * we'd have a 0.03% chance of two simultaneous failures with
- * N=3 but a 0.06% chance with N=4. However, the special case
- * is necessary for N=2 because there's no real quorum in that
- * case (i.e. can't normally survive *any* failures). In that
- * case, we treat the first node as a tie-breaker, allowing
- * quorum to be retained in some cases while still honoring the
- * all-important constraint that there can not simultaneously
- * be two partitioned sets of nodes each believing they have
- * quorum. Of two equally sized sets, the one without that
- * first node will lose.
- *
- * It turns out that the special case is beneficial for higher
- * values of N as well. Continuing the example above, the
- * probability of losing quorum with N=4 and this type of
- * quorum is (very) slightly lower than with N=3 and vanilla
- * quorum. The difference becomes even more pronounced with
- * higher N. Therefore, even though such replica counts are
- * unlikely to be seen in practice, we might as well use the
- * "special" quorum then as well.
- */
- if ((up_children_count * 2) == priv->child_count) {
- return subvols[0];
- }
- }
+afr_has_quorum(unsigned char *subvols, xlator_t *this, call_frame_t *frame)
+{
+ unsigned int quorum_count = 0;
+ afr_private_t *priv = NULL;
+ unsigned int up_children_count = 0;
- if (priv->quorum_count == AFR_QUORUM_AUTO) {
- quorum_count = priv->child_count/2 + 1;
- } else {
- quorum_count = priv->quorum_count;
+ priv = this->private;
+ up_children_count = AFR_COUNT(subvols, priv->child_count);
+
+ if (afr_lookup_has_quorum(frame, up_children_count))
+ return _gf_true;
+
+ if (priv->quorum_count == AFR_QUORUM_AUTO) {
+ /*
+ * Special case for auto-quorum with an even number of nodes.
+ *
+ * A replica set with even count N can only handle the same
+ * number of failures as odd N-1 before losing "vanilla"
+ * quorum, and the probability of more simultaneous failures is
+ * actually higher. For example, with a 1% chance of failure
+ * we'd have a 0.03% chance of two simultaneous failures with
+ * N=3 but a 0.06% chance with N=4. However, the special case
+ * is necessary for N=2 because there's no real quorum in that
+ * case (i.e. can't normally survive *any* failures). In that
+ * case, we treat the first node as a tie-breaker, allowing
+ * quorum to be retained in some cases while still honoring the
+ * all-important constraint that there can not simultaneously
+ * be two partitioned sets of nodes each believing they have
+ * quorum. Of two equally sized sets, the one without that
+ * first node will lose.
+ *
+ * It turns out that the special case is beneficial for higher
+ * values of N as well. Continuing the example above, the
+ * probability of losing quorum with N=4 and this type of
+ * quorum is (very) slightly lower than with N=3 and vanilla
+ * quorum. The difference becomes even more pronounced with
+ * higher N. Therefore, even though such replica counts are
+ * unlikely to be seen in practice, we might as well use the
+ * "special" quorum then as well.
+ */
+ if ((up_children_count * 2) == priv->child_count) {
+ return subvols[0];
}
+ }
- if (up_children_count >= quorum_count)
- return _gf_true;
+ if (priv->quorum_count == AFR_QUORUM_AUTO) {
+ quorum_count = priv->child_count / 2 + 1;
+ } else {
+ quorum_count = priv->quorum_count;
+ }
- return _gf_false;
+ if (up_children_count >= quorum_count)
+ return _gf_true;
+
+ return _gf_false;
}
static gf_boolean_t
-afr_has_fop_quorum (call_frame_t *frame)
+afr_has_fop_quorum(call_frame_t *frame)
{
- xlator_t *this = frame->this;
- afr_local_t *local = frame->local;
- unsigned char *locked_nodes = NULL;
+ xlator_t *this = frame->this;
+ afr_local_t *local = frame->local;
+ unsigned char *locked_nodes = NULL;
- locked_nodes = afr_locked_nodes_get (local->transaction.type,
- &local->internal_lock);
- return afr_has_quorum (locked_nodes, this);
+ locked_nodes = afr_locked_nodes_get(local->transaction.type,
+ &local->internal_lock);
+ return afr_has_quorum(locked_nodes, this, NULL);
}
static gf_boolean_t
-afr_has_fop_cbk_quorum (call_frame_t *frame)
+afr_has_fop_cbk_quorum(call_frame_t *frame)
{
- afr_local_t *local = frame->local;
- xlator_t *this = frame->this;
- afr_private_t *priv = this->private;
- unsigned char *success = alloca0(priv->child_count);
- int i = 0;
+ afr_local_t *local = frame->local;
+ xlator_t *this = frame->this;
+ afr_private_t *priv = this->private;
+ unsigned char *success = alloca0(priv->child_count);
+ int i = 0;
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.pre_op[i])
- if (!local->transaction.failed_subvols[i])
- success[i] = 1;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i])
+ if (!local->transaction.failed_subvols[i])
+ success[i] = 1;
+ }
- return afr_has_quorum (success, this);
+ return afr_has_quorum(success, this, NULL);
}
-void
-afr_handle_quorum (call_frame_t *frame)
+gf_boolean_t
+afr_need_dirty_marking(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- const char *file = NULL;
- uuid_t gfid = {0};
+ afr_private_t *priv = this->private;
+ afr_local_t *local = NULL;
+ gf_boolean_t need_dirty = _gf_false;
- local = frame->local;
- priv = frame->this->private;
+ local = frame->local;
- if (priv->quorum_count == 0)
- return;
+ if (!priv->quorum_count || !local->optimistic_change_log)
+ return _gf_false;
- /* If the fop already failed return right away to preserve errno */
- if (local->op_ret == -1)
- return;
+ if (local->transaction.type == AFR_DATA_TRANSACTION ||
+ local->transaction.type == AFR_METADATA_TRANSACTION)
+ return _gf_false;
- /*
- * Network split may happen just after the fops are unwound, so check
- * if the fop succeeded in a way it still follows quorum. If it doesn't,
- * mark the fop as failure, mark the changelogs so it reflects that
- * failure.
- *
- * Scenario:
- * There are 3 mounts on 3 machines(node1, node2, node3) all writing to
- * single file. Network split happened in a way that node1 can't see
- * node2, node3. Node2, node3 both of them can't see node1. Now at the
- * time of sending write all the bricks are up. Just after write fop is
- * wound on node1, network split happens. Node1 thinks write fop failed
- * on node2, node3 so marks pending changelog for those 2 extended
- * attributes on node1. Node2, node3 thinks writes failed on node1 so
- * they mark pending changelog for node1. When the network is stable
- * again the file already is in split-brain. These checks prevent
- * marking pending changelog on other subvolumes if the fop doesn't
- * succeed in a way it is still following quorum. So with this fix what
- * is happening is, node1 will have all pending changelog(FOOL) because
- * the write succeeded only on node1 but failed on node2, node3 so
- * instead of marking pending changelogs on node2, node3 it just treats
- * the fop as failure and goes into DIRTY state. Where as node2, node3
- * say they are sources and have pending changelog to node1 so there is
- * no split-brain with the fix. The problem is eliminated completely.
- */
+ if (AFR_COUNT(local->transaction.failed_subvols, priv->child_count) ==
+ priv->child_count)
+ return _gf_false;
- if (priv->arbiter_count) {
- if (afr_has_arbiter_fop_cbk_quorum (frame))
- return;
- } else if (afr_has_fop_cbk_quorum (frame)) {
- return;
- }
+ if (!afr_has_fop_cbk_quorum(frame))
+ need_dirty = _gf_true;
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.pre_op[i])
- afr_transaction_fop_failed (frame, frame->this, i);
- }
+ return need_dirty;
+}
- local->op_ret = -1;
- local->op_errno = afr_final_errno (local, priv);
- if (local->op_errno == 0)
- local->op_errno = afr_quorum_errno (priv);
+void
+afr_handle_quorum(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ const char *file = NULL;
+ uuid_t gfid = {0};
- if (local->fd) {
- gf_uuid_copy (gfid, local->fd->inode->gfid);
- file = uuid_utoa (gfid);
- } else {
- loc_path (&local->loc, local->loc.name);
- file = local->loc.path;
- }
+ local = frame->local;
+ priv = frame->this->private;
- gf_msg (frame->this->name, GF_LOG_WARNING, local->op_errno,
- AFR_MSG_QUORUM_FAIL, "%s: Failing %s as quorum is not met",
- file, gf_fop_list[local->op]);
+ if (priv->quorum_count == 0)
+ return;
- switch (local->transaction.type) {
+ /* If the fop already failed return right away to preserve errno */
+ if (local->op_ret == -1)
+ return;
+
+ /*
+ * Network split may happen just after the fops are unwound, so check
+ * if the fop succeeded in a way it still follows quorum. If it doesn't,
+ * mark the fop as failure, mark the changelogs so it reflects that
+ * failure.
+ *
+ * Scenario:
+ * There are 3 mounts on 3 machines(node1, node2, node3) all writing to
+ * single file. Network split happened in a way that node1 can't see
+ * node2, node3. Node2, node3 both of them can't see node1. Now at the
+ * time of sending write all the bricks are up. Just after write fop is
+ * wound on node1, network split happens. Node1 thinks write fop failed
+ * on node2, node3 so marks pending changelog for those 2 extended
+ * attributes on node1. Node2, node3 thinks writes failed on node1 so
+ * they mark pending changelog for node1. When the network is stable
+ * again the file already is in split-brain. These checks prevent
+ * marking pending changelog on other subvolumes if the fop doesn't
+ * succeed in a way it is still following quorum. So with this fix what
+ * is happening is, node1 will have all pending changelog(FOOL) because
+ * the write succeeded only on node1 but failed on node2, node3 so
+ * instead of marking pending changelogs on node2, node3 it just treats
+ * the fop as failure and goes into DIRTY state. Where as node2, node3
+ * say they are sources and have pending changelog to node1 so there is
+ * no split-brain with the fix. The problem is eliminated completely.
+ */
+
+ if (afr_has_fop_cbk_quorum(frame))
+ return;
+
+ if (afr_need_dirty_marking(frame, this))
+ goto set_response;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i])
+ afr_transaction_fop_failed(frame, frame->this, i);
+ }
+
+set_response:
+ local->op_ret = -1;
+ local->op_errno = afr_final_errno(local, priv);
+ if (local->op_errno == 0)
+ local->op_errno = afr_quorum_errno(priv);
+
+ if (local->fd) {
+ gf_uuid_copy(gfid, local->fd->inode->gfid);
+ file = uuid_utoa(gfid);
+ } else {
+ loc_path(&local->loc, local->loc.name);
+ file = local->loc.path;
+ }
+
+ gf_msg(frame->this->name, GF_LOG_WARNING, local->op_errno,
+ AFR_MSG_QUORUM_FAIL, "%s: Failing %s as quorum is not met", file,
+ gf_fop_list[local->op]);
+
+ switch (local->transaction.type) {
case AFR_ENTRY_TRANSACTION:
case AFR_ENTRY_RENAME_TRANSACTION:
- afr_pick_error_xdata (local, priv, local->parent,
- local->readable, local->parent2,
- local->readable2);
- break;
+ afr_pick_error_xdata(local, priv, local->parent, local->readable,
+ local->parent2, local->readable2);
+ break;
default:
- afr_pick_error_xdata (local, priv, local->inode,
- local->readable, NULL, NULL);
- break;
- }
+ afr_pick_error_xdata(local, priv, local->inode, local->readable,
+ NULL, NULL);
+ break;
+ }
}
int
-afr_fill_ta_loc (xlator_t *this, loc_t *loc)
-{
- afr_private_t *priv = NULL;
-
- priv = this->private;
- loc->parent = inode_ref (priv->root_inode);
- gf_uuid_copy (loc->pargfid, loc->parent->gfid);
- loc->name = priv->pending_key[THIN_ARBITER_BRICK_INDEX];
- gf_uuid_copy (loc->gfid, priv->ta_gfid);
- loc->inode = inode_new (loc->parent->table);
- if (!loc->inode)
- return -ENOMEM;
- return 0;
+afr_fill_ta_loc(xlator_t *this, loc_t *loc, gf_boolean_t is_gfid_based_fop)
+{
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+ loc->parent = inode_ref(priv->root_inode);
+ gf_uuid_copy(loc->pargfid, loc->parent->gfid);
+ loc->name = priv->pending_key[THIN_ARBITER_BRICK_INDEX];
+ if (is_gfid_based_fop && gf_uuid_is_null(priv->ta_gfid)) {
+ /* Except afr_ta_id_file_check() which is path based, all other gluster
+ * FOPS need gfid.*/
+ return -EINVAL;
+ }
+ gf_uuid_copy(loc->gfid, priv->ta_gfid);
+ loc->inode = inode_new(loc->parent->table);
+ if (!loc->inode) {
+ loc_wipe(loc);
+ return -ENOMEM;
+ }
+ return 0;
+}
+
+static int
+afr_ta_post_op_done(int ret, call_frame_t *frame, void *opaque)
+{
+ xlator_t *this = NULL;
+ afr_local_t *local = NULL;
+ call_frame_t *txn_frame = NULL;
+ afr_ta_fop_state_t fop_state;
+
+ local = (afr_local_t *)opaque;
+ fop_state = local->fop_state;
+ txn_frame = local->transaction.frame;
+ this = frame->this;
+
+ if (ret == 0) {
+ /*Mark pending xattrs on the up data brick.*/
+ afr_post_op_handle_success(txn_frame, this);
+ } else {
+ afr_post_op_handle_failure(txn_frame, this, -ret);
+ }
+
+ STACK_DESTROY(frame->root);
+ afr_ta_process_onwireq(fop_state, this);
+
+ return 0;
+}
+
+int **
+afr_set_changelog_xattr(afr_private_t *priv, unsigned char *pending,
+ dict_t *xattr, afr_local_t *local)
+{
+ int **changelog = NULL;
+ int idx = 0;
+ int ret = 0;
+ int i;
+
+ if (local->is_new_entry == _gf_true) {
+ changelog = afr_mark_pending_changelog(priv, pending, xattr,
+ local->cont.dir_fop.buf.ia_type);
+ } else {
+ idx = afr_index_for_transaction_type(local->transaction.type);
+ changelog = afr_matrix_create(priv->child_count, AFR_NUM_CHANGE_LOGS);
+ if (!changelog) {
+ goto out;
+ }
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.failed_subvols[i])
+ changelog[i][idx] = hton32(1);
+ }
+ ret = afr_set_pending_dict(priv, xattr, changelog);
+ if (ret < 0) {
+ afr_matrix_cleanup(changelog, priv->child_count);
+ return NULL;
+ }
+ }
+
+out:
+ return changelog;
}
-int
-afr_changelog_thin_arbiter_post_op (xlator_t *this, afr_local_t *local)
+static void
+afr_ta_locked_xattrop_validate(afr_private_t *priv, afr_local_t *local,
+ gf_boolean_t *valid)
{
- int ret = 0;
- afr_private_t *priv = NULL;
- dict_t *xattr = NULL;
- int failed_count = 0;
- struct gf_flock flock = {0, };
- loc_t loc = {0,};
- int i = 0;
-
- priv = this->private;
- if (!priv->thin_arbiter_count)
- return 0;
+ if (priv->ta_event_gen > local->ta_event_gen) {
+ /* We can't trust the ta's response anymore.*/
+ afr_ta_locked_priv_invalidate(priv);
+ *valid = _gf_false;
+ return;
+ }
+ return;
+}
+
+static int
+afr_ta_post_op_do(void *opaque)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ xlator_t *this = NULL;
+ dict_t *xattr = NULL;
+ unsigned char *pending = NULL;
+ int **changelog = NULL;
+ int failed_subvol = -1;
+ int success_subvol = -1;
+ loc_t loc = {
+ 0,
+ };
+ int i = 0;
+ int ret = 0;
+ gf_boolean_t valid = _gf_true;
+
+ local = (afr_local_t *)opaque;
+ this = local->transaction.frame->this;
+ priv = this->private;
+
+ ret = afr_fill_ta_loc(this, &loc, _gf_true);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Failed to populate loc for thin-arbiter.");
+ goto out;
+ }
+
+ xattr = dict_new();
+ if (!xattr) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ pending = alloca0(priv->child_count);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.failed_subvols[i]) {
+ pending[i] = 1;
+ failed_subvol = i;
+ } else {
+ success_subvol = i;
+ }
+ }
+
+ changelog = afr_set_changelog_xattr(priv, pending, xattr, local);
+
+ if (!changelog) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ ret = afr_ta_post_op_lock(this, &loc);
+ if (ret)
+ goto out;
+
+ ret = syncop_xattrop(priv->children[THIN_ARBITER_BRICK_INDEX], &loc,
+ GF_XATTROP_ADD_ARRAY, xattr, NULL, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
+ "Post-op on thin-arbiter id file %s failed for gfid %s.",
+ priv->pending_key[THIN_ARBITER_BRICK_INDEX],
+ uuid_utoa(local->inode->gfid));
+ }
+ LOCK(&priv->lock);
+ {
+ if (ret == 0) {
+ priv->ta_bad_child_index = failed_subvol;
+ } else if (ret == -EINVAL) {
+ priv->ta_bad_child_index = success_subvol;
+ ret = -EIO; /* TA failed the fop. Return EIO to application. */
+ }
+
+ afr_ta_locked_xattrop_validate(priv, local, &valid);
+ }
+ UNLOCK(&priv->lock);
+ if (valid == _gf_false) {
+ gf_msg(this->name, GF_LOG_ERROR, EIO, AFR_MSG_THIN_ARB,
+ "Post-op on thin-arbiter id file %s for gfid %s invalidated due "
+ "to event-gen mismatch.",
+ priv->pending_key[THIN_ARBITER_BRICK_INDEX],
+ uuid_utoa(local->inode->gfid));
+ ret = -EIO;
+ }
+
+ afr_ta_post_op_unlock(this, &loc);
+out:
+ if (xattr)
+ dict_unref(xattr);
+ if (changelog)
+ afr_matrix_cleanup(changelog, priv->child_count);
- failed_count = AFR_COUNT (local->transaction.failed_subvols,
- priv->child_count);
- if (!failed_count)
- return 0;
+ loc_wipe(&loc);
- GF_ASSERT (failed_count == 1);
- ret = afr_fill_ta_loc (this, &loc);
- if (ret)
- goto out;
+ return ret;
+}
- xattr = dict_new ();
- if (!xattr) {
- ret = -ENOMEM;
- goto out;
- }
- for (i = 0; i < priv->child_count; i++) {
- ret = dict_set_static_bin (xattr, priv->pending_key[i],
- local->pending[i],
- AFR_NUM_CHANGE_LOGS * sizeof (int));
- if (ret)
- goto out;
- }
+static int
+afr_ta_post_op_synctask(xlator_t *this, afr_local_t *local)
+{
+ call_frame_t *ta_frame = NULL;
+ int ret = 0;
- flock.l_type = F_WRLCK;
- flock.l_start = 0;
- flock.l_len = 0;
+ ta_frame = afr_ta_frame_create(this);
+ if (!ta_frame) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, AFR_MSG_THIN_ARB,
+ "Failed to create ta_frame");
+ goto err;
+ }
+ ret = synctask_new(this->ctx->env, afr_ta_post_op_do, afr_ta_post_op_done,
+ ta_frame, local);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, AFR_MSG_THIN_ARB,
+ "Failed to launch post-op on thin arbiter for gfid %s",
+ uuid_utoa(local->inode->gfid));
+ STACK_DESTROY(ta_frame->root);
+ goto err;
+ }
- /*TODO: Convert to two domain locking. */
- ret = syncop_inodelk (priv->children[THIN_ARBITER_BRICK_INDEX],
- THIN_ARBITER_DOM1, &loc, F_SETLKW, &flock,
- NULL, NULL);
- if (ret)
- goto out;
+ return ret;
+err:
+ afr_changelog_post_op_fail(local->transaction.frame, this, ENOMEM);
+ return ret;
+}
- ret = syncop_xattrop (priv->children[THIN_ARBITER_BRICK_INDEX], &loc,
- GF_XATTROP_ADD_ARRAY, xattr, NULL, NULL, NULL);
+static void
+afr_ta_set_fop_state(afr_private_t *priv, afr_local_t *local,
+ int *on_wire_count)
+{
+ LOCK(&priv->lock);
+ {
+ if (priv->release_ta_notify_dom_lock == _gf_true) {
+ /* Put the fop in waitq until notify dom lock is released.*/
+ local->fop_state = TA_WAIT_FOR_NOTIFY_LOCK_REL;
+ list_add_tail(&local->ta_waitq, &priv->ta_waitq);
+ } else if (priv->ta_bad_child_index == AFR_CHILD_UNKNOWN) {
+ /* Post-op on thin-arbiter to decide success/failure. */
+ local->fop_state = TA_GET_INFO_FROM_TA_FILE;
+ *on_wire_count = ++priv->ta_on_wire_txn_count;
+ if (*on_wire_count > 1) {
+ /*Avoid sending multiple on-wire post-ops on TA*/
+ list_add_tail(&local->ta_onwireq, &priv->ta_onwireq);
+ }
+ } else if (local->ta_failed_subvol == priv->ta_bad_child_index) {
+ /* Post-op on TA not needed as the fop failed on the in-memory bad
+ * brick. Just mark pending xattrs on the good data brick.*/
+ local->fop_state = TA_INFO_IN_MEMORY_SUCCESS;
+ priv->ta_in_mem_txn_count++;
+ } else {
+ /* Post-op on TA not needed as the fop succeeded only on the
+ * in-memory bad data brick and not the good one. Fail the fop.*/
+ local->fop_state = TA_INFO_IN_MEMORY_FAILED;
+ priv->ta_in_mem_txn_count++;
+ }
+ }
+ UNLOCK(&priv->lock);
+}
- if (ret == -EINVAL) {
- gf_msg (this->name, GF_LOG_INFO, -ret, AFR_MSG_THIN_ARB,
- "Thin-arbiter has denied post-op on %s for gfid %s.",
- priv->pending_key[THIN_ARBITER_BRICK_INDEX],
- uuid_utoa (local->inode->gfid));
+static void
+afr_ta_fill_failed_subvol(afr_private_t *priv, afr_local_t *local)
+{
+ int i = 0;
- } else if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret, AFR_MSG_THIN_ARB,
- "Post-op on thin-arbiter id file %s failed for gfid %s.",
- priv->pending_key[THIN_ARBITER_BRICK_INDEX],
- uuid_utoa (local->inode->gfid));
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.failed_subvols[i]) {
+ local->ta_failed_subvol = i;
+ break;
}
- flock.l_type = F_UNLCK;
- syncop_inodelk (priv->children[THIN_ARBITER_BRICK_INDEX],
- THIN_ARBITER_DOM1, &loc, F_SETLKW, &flock, NULL, NULL);
-out:
- if (xattr)
- dict_unref (xattr);
+ }
+}
+
+static void
+afr_post_op_handle_success(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
- return ret;
+ local = frame->local;
+ if (local->is_new_entry == _gf_true) {
+ afr_mark_new_entry_changelog(frame, this);
+ }
+ afr_changelog_post_op_do(frame, this);
+
+ return;
}
-int
-afr_changelog_post_op_now (call_frame_t *frame, xlator_t *this)
-{
- afr_private_t *priv = this->private;
- afr_local_t *local = NULL;
- dict_t *xattr = NULL;
- int i = 0;
- int ret = 0;
- int idx = 0;
- int nothing_failed = 1;
- gf_boolean_t need_undirty = _gf_false;
-
- afr_handle_quorum (frame);
- local = frame->local;
- idx = afr_index_for_transaction_type (local->transaction.type);
-
- nothing_failed = afr_txn_nothing_failed (frame, this);
-
- if (afr_changelog_pre_op_uninherit (frame, this))
- need_undirty = _gf_false;
- else
- need_undirty = _gf_true;
-
- if (local->op_ret < 0 && !nothing_failed) {
- afr_changelog_post_op_done (frame, this);
- goto out;
- }
+static void
+afr_post_op_handle_failure(call_frame_t *frame, xlator_t *this, int op_errno)
+{
+ afr_changelog_post_op_fail(frame, this, op_errno);
- if (nothing_failed && !need_undirty) {
- afr_changelog_post_op_done (frame, this);
- goto out;
- }
+ return;
+}
- if (local->transaction.in_flight_sb) {
- local->op_ret = -1;
- local->op_errno = local->transaction.in_flight_sb_errno;
- afr_changelog_post_op_done (frame, this);
- goto out;
- }
+static void
+afr_ta_decide_post_op_state(call_frame_t *frame, xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int on_wire_count = 0;
+
+ priv = this->private;
+ local = frame->local;
+
+ afr_ta_set_fop_state(priv, local, &on_wire_count);
+
+ switch (local->fop_state) {
+ case TA_GET_INFO_FROM_TA_FILE:
+ if (on_wire_count == 1)
+ afr_ta_post_op_synctask(this, local);
+ /*else, fop is queued in ta_onwireq.*/
+ break;
+ case TA_WAIT_FOR_NOTIFY_LOCK_REL:
+ /*Post releasing the notify lock, we will act on this queue*/
+ break;
+ case TA_INFO_IN_MEMORY_SUCCESS:
+ afr_post_op_handle_success(frame, this);
+ break;
+ case TA_INFO_IN_MEMORY_FAILED:
+ afr_post_op_handle_failure(frame, this, EIO);
+ break;
+ default:
+ break;
+ }
+ return;
+}
- xattr = dict_new ();
- if (!xattr) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- afr_changelog_post_op_done (frame, this);
- goto out;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.failed_subvols[i])
- local->pending[i][idx] = hton32(1);
- }
-
- ret = afr_set_pending_dict (priv, xattr, local->pending);
- if (ret < 0) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- afr_changelog_post_op_done (frame, this);
- goto out;
- }
-
- ret = afr_changelog_thin_arbiter_post_op (this, local);
- if (ret < 0) {
- local->op_ret = -1;
- local->op_errno = -ret;
- afr_changelog_post_op_done (frame, this);
- goto out;
- }
+static void
+afr_handle_failure_using_thin_arbiter(call_frame_t *frame, xlator_t *this)
+{
+ afr_private_t *priv = this->private;
+ afr_local_t *local = frame->local;
- if (need_undirty)
- local->dirty[idx] = hton32(-1);
- else
- local->dirty[idx] = hton32(0);
-
- ret = dict_set_static_bin (xattr, AFR_DIRTY, local->dirty,
- sizeof(int) * AFR_NUM_CHANGE_LOGS);
- if (ret) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- afr_changelog_post_op_done (frame, this);
- goto out;
- }
-
- afr_changelog_do (frame, this, xattr, afr_changelog_post_op_done,
- AFR_TRANSACTION_POST_OP);
+ afr_ta_fill_failed_subvol(priv, local);
+ gf_msg_debug(this->name, 0,
+ "Fop failed on data brick (%s) for gfid=%s. "
+ "ta info needed to decide fop result.",
+ priv->children[local->ta_failed_subvol]->name,
+ uuid_utoa(local->inode->gfid));
+ afr_ta_decide_post_op_state(frame, this);
+}
+
+void
+afr_changelog_post_op_do(call_frame_t *frame, xlator_t *this)
+{
+ afr_private_t *priv = this->private;
+ afr_local_t *local = NULL;
+ dict_t *xattr = NULL;
+ int i = 0;
+ int ret = 0;
+ int idx = 0;
+ int nothing_failed = 1;
+ gf_boolean_t need_undirty = _gf_false;
+
+ afr_handle_quorum(frame, this);
+ local = frame->local;
+ idx = afr_index_for_transaction_type(local->transaction.type);
+
+ xattr = dict_new();
+ if (!xattr) {
+ afr_changelog_post_op_fail(frame, this, ENOMEM);
+ goto out;
+ }
+
+ nothing_failed = afr_txn_nothing_failed(frame, this);
+
+ if (afr_changelog_pre_op_uninherit(frame, this))
+ need_undirty = _gf_false;
+ else
+ need_undirty = _gf_true;
+
+ if (local->op_ret < 0 && !nothing_failed) {
+ if (afr_need_dirty_marking(frame, this)) {
+ local->dirty[idx] = hton32(1);
+ goto set_dirty;
+ }
+
+ afr_changelog_post_op_done(frame, this);
+ goto out;
+ }
+
+ if (nothing_failed && !need_undirty) {
+ afr_changelog_post_op_done(frame, this);
+ goto out;
+ }
+
+ if (local->transaction.in_flight_sb) {
+ afr_changelog_post_op_fail(frame, this,
+ local->transaction.in_flight_sb_errno);
+ goto out;
+ }
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.failed_subvols[i])
+ local->pending[i][idx] = hton32(1);
+ }
+
+ ret = afr_set_pending_dict(priv, xattr, local->pending);
+ if (ret < 0) {
+ afr_changelog_post_op_fail(frame, this, ENOMEM);
+ goto out;
+ }
+
+ if (need_undirty)
+ local->dirty[idx] = hton32(-1);
+ else
+ local->dirty[idx] = hton32(0);
+
+set_dirty:
+ ret = dict_set_static_bin(xattr, AFR_DIRTY, local->dirty,
+ sizeof(int) * AFR_NUM_CHANGE_LOGS);
+ if (ret) {
+ afr_changelog_post_op_fail(frame, this, ENOMEM);
+ goto out;
+ }
+
+ afr_changelog_do(frame, this, xattr, afr_changelog_post_op_done,
+ AFR_TRANSACTION_POST_OP);
out:
- if (xattr)
- dict_unref (xattr);
+ if (xattr)
+ dict_unref(xattr);
- return 0;
+ return;
}
+static int
+afr_changelog_post_op_now(call_frame_t *frame, xlator_t *this)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ int failed_count = 0;
-gf_boolean_t
-afr_changelog_pre_op_uninherit (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_inode_ctx_t *ctx = NULL;
- int i = 0;
- gf_boolean_t ret = _gf_false;
- int type = 0;
-
- local = frame->local;
- priv = this->private;
- ctx = local->inode_ctx;
-
- type = afr_index_for_transaction_type (local->transaction.type);
- if (type != AFR_DATA_TRANSACTION)
- return !local->transaction.dirtied;
-
- if (local->transaction.no_uninherit)
- return _gf_false;
-
- /* This function must be idempotent. So check if we
- were called before and return the same answer again.
-
- It is important to keep this function idempotent for
- the call in afr_changelog_post_op_safe() to not have
- side effects on the call from afr_changelog_post_op_now()
- */
- if (local->transaction.uninherit_done)
- return local->transaction.uninherit_value;
-
- LOCK(&local->inode->lock);
- {
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.pre_op[i] !=
- ctx->pre_op_done[type][i]) {
- ret = !local->transaction.dirtied;
- goto unlock;
- }
- }
-
- if (ctx->inherited[type]) {
- ret = _gf_true;
- ctx->inherited[type]--;
- } else if (ctx->on_disk[type]) {
- ret = _gf_false;
- ctx->on_disk[type]--;
- } else {
- /* ASSERT */
- ret = _gf_false;
- }
-
- if (!ctx->inherited[type] && !ctx->on_disk[type]) {
- for (i = 0; i < priv->child_count; i++)
- ctx->pre_op_done[type][i] = 0;
- }
- }
-unlock:
- UNLOCK(&local->inode->lock);
+ priv = this->private;
+ local = frame->local;
- local->transaction.uninherit_done = _gf_true;
- local->transaction.uninherit_value = ret;
+ if (priv->thin_arbiter_count) {
+ failed_count = AFR_COUNT(local->transaction.failed_subvols,
+ priv->child_count);
+ if (failed_count == 1) {
+ afr_handle_failure_using_thin_arbiter(frame, this);
+ return 0;
+ } else {
+ /* Txn either succeeded or failed on both data bricks. Let
+ * post_op_do handle it as the case might be. */
+ }
+ }
- return ret;
+ afr_changelog_post_op_do(frame, this);
+ return 0;
}
-
gf_boolean_t
-afr_changelog_pre_op_inherit (call_frame_t *frame, xlator_t *this)
+afr_changelog_pre_op_uninherit(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- gf_boolean_t ret = _gf_false;
- int type = 0;
-
- local = frame->local;
- priv = this->private;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_inode_ctx_t *ctx = NULL;
+ int i = 0;
+ gf_boolean_t ret = _gf_false;
+ int type = 0;
- if (local->transaction.type != AFR_DATA_TRANSACTION)
- return _gf_false;
+ local = frame->local;
+ priv = this->private;
+ ctx = local->inode_ctx;
- type = afr_index_for_transaction_type (local->transaction.type);
+ type = afr_index_for_transaction_type(local->transaction.type);
+ if (type != AFR_DATA_TRANSACTION)
+ return !local->transaction.dirtied;
- LOCK(&local->inode->lock);
- {
- if (!local->inode_ctx->on_disk[type]) {
- /* nothing to inherit yet */
- ret = _gf_false;
- goto unlock;
- }
+ if (local->transaction.no_uninherit)
+ return _gf_false;
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.pre_op[i] !=
- local->inode_ctx->pre_op_done[type][i]) {
- /* either inherit exactly, or don't */
- ret = _gf_false;
- goto unlock;
- }
- }
+ /* This function must be idempotent. So check if we
+ were called before and return the same answer again.
- local->inode_ctx->inherited[type]++;
+ It is important to keep this function idempotent for
+ the call in afr_changelog_post_op_safe() to not have
+ side effects on the call from afr_changelog_post_op_now()
+ */
+ if (local->transaction.uninherit_done)
+ return local->transaction.uninherit_value;
- ret = _gf_true;
+ LOCK(&local->inode->lock);
+ {
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i] != ctx->pre_op_done[type][i]) {
+ ret = !local->transaction.dirtied;
+ goto unlock;
+ }
+ }
+
+ if (ctx->inherited[type]) {
+ ret = _gf_true;
+ ctx->inherited[type]--;
+ } else if (ctx->on_disk[type]) {
+ ret = _gf_false;
+ ctx->on_disk[type]--;
+ } else {
+ /* ASSERT */
+ ret = _gf_false;
+ }
- local->transaction.inherited = _gf_true;
- }
+ if (!ctx->inherited[type] && !ctx->on_disk[type]) {
+ for (i = 0; i < priv->child_count; i++)
+ ctx->pre_op_done[type][i] = 0;
+ }
+ }
unlock:
- UNLOCK(&local->inode->lock);
+ UNLOCK(&local->inode->lock);
- return ret;
-}
+ local->transaction.uninherit_done = _gf_true;
+ local->transaction.uninherit_value = ret;
+ return ret;
+}
gf_boolean_t
-afr_changelog_pre_op_update (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int i = 0;
- gf_boolean_t ret = _gf_false;
- int type = 0;
-
- local = frame->local;
- priv = this->private;
-
- if (local->transaction.type == AFR_ENTRY_TRANSACTION ||
- local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION)
- return _gf_false;
-
- if (local->transaction.inherited)
- /* was already inherited in afr_changelog_pre_op */
- return _gf_false;
-
- if (!local->transaction.dirtied)
- return _gf_false;
-
- if (!afr_txn_nothing_failed (frame, this))
- return _gf_false;
-
- type = afr_index_for_transaction_type (local->transaction.type);
-
- ret = _gf_false;
-
- LOCK(&local->inode->lock);
- {
- if (!local->inode_ctx->on_disk[type]) {
- for (i = 0; i < priv->child_count; i++)
- local->inode_ctx->pre_op_done[type][i] =
- (!local->transaction.failed_subvols[i]);
- } else {
- for (i = 0; i < priv->child_count; i++)
- if (local->inode_ctx->pre_op_done[type][i] !=
- (!local->transaction.failed_subvols[i])) {
- local->transaction.no_uninherit = 1;
- goto unlock;
- }
- }
- local->inode_ctx->on_disk[type]++;
-
- ret = _gf_true;
- }
-unlock:
- UNLOCK(&local->inode->lock);
+afr_changelog_pre_op_inherit(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ gf_boolean_t ret = _gf_false;
+ int type = 0;
- return ret;
-}
+ local = frame->local;
+ priv = this->private;
+ if (local->transaction.type != AFR_DATA_TRANSACTION)
+ return _gf_false;
-int
-afr_changelog_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
-{
- afr_local_t *local = NULL;
- int call_count = -1;
- int child_index = -1;
+ type = afr_index_for_transaction_type(local->transaction.type);
- local = frame->local;
- child_index = (long) cookie;
+ LOCK(&local->inode->lock);
+ {
+ if (!local->inode_ctx->on_disk[type]) {
+ /* nothing to inherit yet */
+ ret = _gf_false;
+ goto unlock;
+ }
- if (op_ret == -1) {
- local->op_errno = op_errno;
- afr_transaction_fop_failed (frame, this, child_index);
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.pre_op[i] !=
+ local->inode_ctx->pre_op_done[type][i]) {
+ /* either inherit exactly, or don't */
+ ret = _gf_false;
+ goto unlock;
+ }
}
- if (xattr)
- local->transaction.changelog_xdata[child_index] = dict_ref (xattr);
+ local->inode_ctx->inherited[type]++;
- call_count = afr_frame_return (frame);
+ ret = _gf_true;
- if (call_count == 0) {
- local->transaction.changelog_resume (frame, this);
- }
+ local->transaction.inherited = _gf_true;
+ }
+unlock:
+ UNLOCK(&local->inode->lock);
- return 0;
+ return ret;
}
-void
-afr_changelog_populate_xdata (call_frame_t *frame, afr_xattrop_type_t op,
- dict_t **xdata, dict_t **newloc_xdata)
-{
- int i = 0;
- int ret = 0;
- char *key = NULL;
- const char *name = NULL;
- dict_t *xdata1 = NULL;
- dict_t *xdata2 = NULL;
- xlator_t *this = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- gf_boolean_t need_entry_key_set = _gf_true;
-
- local = frame->local;
- this = THIS;
- priv = this->private;
-
- if (local->transaction.type == AFR_DATA_TRANSACTION ||
- local->transaction.type == AFR_METADATA_TRANSACTION)
- goto out;
-
- if (!priv->esh_granular)
- goto out;
-
- xdata1 = dict_new();
- if (!xdata1)
- goto out;
+gf_boolean_t
+afr_changelog_pre_op_update(call_frame_t *frame, xlator_t *this)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+ gf_boolean_t ret = _gf_false;
+ int type = 0;
- name = local->loc.name;
- if (local->op == GF_FOP_LINK)
- name = local->newloc.name;
+ local = frame->local;
+ priv = this->private;
- switch (op) {
- case AFR_TRANSACTION_PRE_OP:
- key = GF_XATTROP_ENTRY_IN_KEY;
- break;
- case AFR_TRANSACTION_POST_OP:
- if (afr_txn_nothing_failed (frame, this)) {
- key = GF_XATTROP_ENTRY_OUT_KEY;
- for (i = 0; i < priv->child_count; i++) {
- if (!local->transaction.failed_subvols[i])
- continue;
- need_entry_key_set = _gf_false;
- break;
- }
- /* If the transaction itself did not fail and there
- * are no failed subvolumes, check whether the fop
- * failed due to a symmetric error. If it did, do
- * not set the ENTRY_OUT xattr which would end up
- * deleting a name index which was created possibly by
- * an earlier entry txn that may have failed on some
- * of the sub-volumes.
- */
- if (local->op_ret)
- need_entry_key_set = _gf_false;
- } else {
- key = GF_XATTROP_ENTRY_IN_KEY;
- }
- break;
- }
+ if (local->transaction.type == AFR_ENTRY_TRANSACTION ||
+ local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION)
+ return _gf_false;
- if (need_entry_key_set) {
- ret = dict_set_str (xdata1, key, (char *)name);
- if (ret)
- gf_msg (THIS->name, GF_LOG_ERROR, 0,
- AFR_MSG_DICT_SET_FAILED,
- "%s/%s: Could not set %s key during xattrop",
- uuid_utoa (local->loc.pargfid), local->loc.name,
- key);
- if (local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
- xdata2 = dict_new ();
- if (!xdata2)
- goto out;
-
- ret = dict_set_str (xdata2, key,
- (char *)local->newloc.name);
- if (ret)
- gf_msg (THIS->name, GF_LOG_ERROR, 0,
- AFR_MSG_DICT_SET_FAILED,
- "%s/%s: Could not set %s key during "
- "xattrop",
- uuid_utoa (local->newloc.pargfid),
- local->newloc.name, key);
- }
- }
+ if (local->transaction.inherited)
+ /* was already inherited in afr_changelog_pre_op */
+ return _gf_false;
- *xdata = xdata1;
- *newloc_xdata = xdata2;
- xdata1 = xdata2 = NULL;
-out:
- if (xdata1)
- dict_unref (xdata1);
- return;
-}
+ if (!local->transaction.dirtied)
+ return _gf_false;
-int
-afr_changelog_prepare (xlator_t *this, call_frame_t *frame, int *call_count,
- afr_changelog_resume_t changelog_resume,
- afr_xattrop_type_t op, dict_t **xdata,
- dict_t **newloc_xdata)
-{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
+ if (!afr_txn_nothing_failed(frame, this))
+ return _gf_false;
- local = frame->local;
- priv = this->private;
+ type = afr_index_for_transaction_type(local->transaction.type);
- *call_count = afr_changelog_call_count (local->transaction.type,
- local->transaction.pre_op,
- local->transaction.failed_subvols,
- priv->child_count);
+ ret = _gf_false;
- if (*call_count == 0) {
- changelog_resume (frame, this);
- return -1;
+ LOCK(&local->inode->lock);
+ {
+ if (!local->inode_ctx->on_disk[type]) {
+ for (i = 0; i < priv->child_count; i++)
+ local->inode_ctx->pre_op_done[type][i] =
+ (!local->transaction.failed_subvols[i]);
+ } else {
+ for (i = 0; i < priv->child_count; i++)
+ if (local->inode_ctx->pre_op_done[type][i] !=
+ (!local->transaction.failed_subvols[i])) {
+ local->transaction.no_uninherit = 1;
+ goto unlock;
+ }
}
+ local->inode_ctx->on_disk[type]++;
- afr_changelog_populate_xdata (frame, op, xdata, newloc_xdata);
- local->call_count = *call_count;
+ ret = _gf_true;
+ }
+unlock:
+ UNLOCK(&local->inode->lock);
- local->transaction.changelog_resume = changelog_resume;
- return 0;
+ return ret;
}
int
-afr_changelog_do (call_frame_t *frame, xlator_t *this, dict_t *xattr,
- afr_changelog_resume_t changelog_resume,
- afr_xattrop_type_t op)
+afr_changelog_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, dict_t *xattr, dict_t *xdata)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- dict_t *xdata = NULL;
- dict_t *newloc_xdata = NULL;
- int i = 0;
- int call_count = 0;
- int ret = 0;
-
- local = frame->local;
- priv = this->private;
+ afr_local_t *local = NULL;
+ int call_count = -1;
+ int child_index = -1;
- for (i = 0; i < priv->child_count; i++) {
- if (local->transaction.changelog_xdata[i]) {
- dict_unref (local->transaction.changelog_xdata[i]);
- local->transaction.changelog_xdata[i] = NULL;
- }
- }
+ local = frame->local;
+ child_index = (long)cookie;
- ret = afr_changelog_prepare (this, frame, &call_count, changelog_resume,
- op, &xdata, &newloc_xdata);
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ afr_transaction_fop_failed(frame, this, child_index);
+ }
- if (ret)
- return 0;
+ if (xattr)
+ local->transaction.changelog_xdata[child_index] = dict_ref(xattr);
- for (i = 0; i < priv->child_count; i++) {
- if (!local->transaction.pre_op[i] ||
- local->transaction.failed_subvols[i])
- continue;
+ call_count = afr_frame_return(frame);
- switch (local->transaction.type) {
- case AFR_DATA_TRANSACTION:
- case AFR_METADATA_TRANSACTION:
- if (!local->fd) {
- STACK_WIND_COOKIE (frame, afr_changelog_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->xattrop,
- &local->loc,
- GF_XATTROP_ADD_ARRAY, xattr,
- xdata);
- } else {
- STACK_WIND_COOKIE (frame, afr_changelog_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->fxattrop,
- local->fd,
- GF_XATTROP_ADD_ARRAY, xattr,
- xdata);
- }
- break;
- case AFR_ENTRY_RENAME_TRANSACTION:
-
- STACK_WIND_COOKIE (frame, afr_changelog_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->xattrop,
- &local->transaction.new_parent_loc,
- GF_XATTROP_ADD_ARRAY, xattr,
- newloc_xdata);
- call_count--;
+ if (call_count == 0) {
+ local->transaction.changelog_resume(frame, this);
+ }
- /* fall through */
-
- case AFR_ENTRY_TRANSACTION:
- if (local->fd)
- STACK_WIND_COOKIE (frame, afr_changelog_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->fxattrop,
- local->fd,
- GF_XATTROP_ADD_ARRAY, xattr,
- xdata);
- else
- STACK_WIND_COOKIE (frame, afr_changelog_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->xattrop,
- &local->transaction.parent_loc,
- GF_XATTROP_ADD_ARRAY, xattr,
- xdata);
- break;
- }
-
- if (!--call_count)
- break;
- }
-
- if (xdata)
- dict_unref (xdata);
- if (newloc_xdata)
- dict_unref (newloc_xdata);
- return 0;
+ return 0;
}
-static void
-afr_init_optimistic_changelog_for_txn (xlator_t *this, afr_local_t *local)
-{
- int locked_count = 0;
- afr_private_t *priv = NULL;
+void
+afr_changelog_populate_xdata(call_frame_t *frame, afr_xattrop_type_t op,
+ dict_t **xdata, dict_t **newloc_xdata)
+{
+ int i = 0;
+ int ret = 0;
+ char *key = NULL;
+ int keylen = 0;
+ const char *name = NULL;
+ dict_t *xdata1 = NULL;
+ dict_t *xdata2 = NULL;
+ xlator_t *this = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ gf_boolean_t need_entry_key_set = _gf_true;
+
+ local = frame->local;
+ this = THIS;
+ priv = this->private;
+
+ if (local->transaction.type == AFR_DATA_TRANSACTION ||
+ local->transaction.type == AFR_METADATA_TRANSACTION)
+ goto out;
+
+ if (!priv->esh_granular)
+ goto out;
+
+ xdata1 = dict_new();
+ if (!xdata1)
+ goto out;
+
+ name = local->loc.name;
+ if (local->op == GF_FOP_LINK)
+ name = local->newloc.name;
+
+ switch (op) {
+ case AFR_TRANSACTION_PRE_OP:
+ key = GF_XATTROP_ENTRY_IN_KEY;
+ break;
+ case AFR_TRANSACTION_POST_OP:
+ if (afr_txn_nothing_failed(frame, this)) {
+ key = GF_XATTROP_ENTRY_OUT_KEY;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.failed_subvols[i])
+ continue;
+ need_entry_key_set = _gf_false;
+ break;
+ }
+ /* If the transaction itself did not fail and there
+ * are no failed subvolumes, check whether the fop
+ * failed due to a symmetric error. If it did, do
+ * not set the ENTRY_OUT xattr which would end up
+ * deleting a name index which was created possibly by
+ * an earlier entry txn that may have failed on some
+ * of the sub-volumes.
+ */
+ if (local->op_ret)
+ need_entry_key_set = _gf_false;
+ } else {
+ key = GF_XATTROP_ENTRY_IN_KEY;
+ }
+ break;
+ }
- priv = this->private;
+ if (need_entry_key_set) {
+ keylen = strlen(key);
+ ret = dict_set_strn(xdata1, key, keylen, (char *)name);
+ if (ret)
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, AFR_MSG_DICT_SET_FAILED,
+ "%s/%s: Could not set %s key during xattrop",
+ uuid_utoa(local->loc.pargfid), local->loc.name, key);
+ if (local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
+ xdata2 = dict_new();
+ if (!xdata2)
+ goto out;
- locked_count = AFR_COUNT (local->transaction.pre_op, priv->child_count);
- if (priv->optimistic_change_log && locked_count == priv->child_count)
- local->optimistic_change_log = 1;
+ ret = dict_set_strn(xdata2, key, keylen,
+ (char *)local->newloc.name);
+ if (ret)
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, AFR_MSG_DICT_SET_FAILED,
+ "%s/%s: Could not set %s key during "
+ "xattrop",
+ uuid_utoa(local->newloc.pargfid), local->newloc.name,
+ key);
+ }
+ }
- return;
+ *xdata = xdata1;
+ *newloc_xdata = xdata2;
+ xdata1 = xdata2 = NULL;
+out:
+ if (xdata1)
+ dict_unref(xdata1);
+ return;
}
int
-afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
-{
- afr_private_t * priv = this->private;
- int i = 0;
- int ret = 0;
- int call_count = 0;
- int op_errno = 0;
- afr_local_t *local = NULL;
- afr_internal_lock_t *int_lock = NULL;
- unsigned char *locked_nodes = NULL;
- int idx = -1;
- gf_boolean_t pre_nop = _gf_true;
- dict_t *xdata_req = NULL;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- idx = afr_index_for_transaction_type (local->transaction.type);
-
- locked_nodes = afr_locked_nodes_get (local->transaction.type, int_lock);
-
- for (i = 0; i < priv->child_count; i++) {
- if (locked_nodes[i]) {
- local->transaction.pre_op[i] = 1;
- call_count++;
- } else {
- local->transaction.failed_subvols[i] = 1;
- }
- }
+afr_changelog_prepare(xlator_t *this, call_frame_t *frame, int *call_count,
+ afr_changelog_resume_t changelog_resume,
+ afr_xattrop_type_t op, dict_t **xdata,
+ dict_t **newloc_xdata)
+{
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
- afr_init_optimistic_changelog_for_txn (this, local);
+ local = frame->local;
+ priv = this->private;
- if (afr_changelog_pre_op_inherit (frame, this))
- goto next;
+ *call_count = afr_changelog_call_count(
+ local->transaction.type, local->transaction.pre_op,
+ local->transaction.failed_subvols, priv->child_count);
- /* This condition should not be met with present code, as
- * transaction.done will be called if locks are not acquired on even a
- * single node.
- */
- if (call_count == 0) {
- op_errno = ENOTCONN;
- goto err;
- }
+ if (*call_count == 0) {
+ changelog_resume(frame, this);
+ return -1;
+ }
- /* Check if the fop can be performed on at least
- * quorum number of nodes.
- */
- if (priv->quorum_count && !afr_has_fop_quorum (frame)) {
- op_errno = int_lock->lock_op_errno;
- if (op_errno == 0)
- op_errno = afr_quorum_errno (priv);
- goto err;
- }
+ afr_changelog_populate_xdata(frame, op, xdata, newloc_xdata);
+ local->call_count = *call_count;
- xdata_req = dict_new();
- if (!xdata_req) {
- op_errno = ENOMEM;
- goto err;
- }
-
- if (call_count < priv->child_count)
- pre_nop = _gf_false;
+ local->transaction.changelog_resume = changelog_resume;
+ return 0;
+}
- /* Set an all-zero pending changelog so that in the cbk, we can get the
- * current on-disk values. In a replica 3 volume with arbiter enabled,
- * these values are needed to arrive at a go/ no-go of the fop phase to
- * avoid ending up in split-brain.*/
+int
+afr_changelog_do(call_frame_t *frame, xlator_t *this, dict_t *xattr,
+ afr_changelog_resume_t changelog_resume, afr_xattrop_type_t op)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ dict_t *xdata = NULL;
+ dict_t *newloc_xdata = NULL;
+ int i = 0;
+ int call_count = 0;
+ int ret = 0;
- ret = afr_set_pending_dict (priv, xdata_req, local->pending);
- if (ret < 0) {
- op_errno = ENOMEM;
- goto err;
- }
+ local = frame->local;
+ priv = this->private;
- if (afr_needs_changelog_update (local)) {
+ for (i = 0; i < priv->child_count; i++) {
+ if (local->transaction.changelog_xdata[i]) {
+ dict_unref(local->transaction.changelog_xdata[i]);
+ local->transaction.changelog_xdata[i] = NULL;
+ }
+ }
- local->dirty[idx] = hton32(1);
+ ret = afr_changelog_prepare(this, frame, &call_count, changelog_resume, op,
+ &xdata, &newloc_xdata);
- ret = dict_set_static_bin (xdata_req, AFR_DIRTY, local->dirty,
- sizeof(int) * AFR_NUM_CHANGE_LOGS);
- if (ret) {
- op_errno = ENOMEM;
- goto err;
- }
+ if (ret)
+ return 0;
- pre_nop = _gf_false;
- local->transaction.dirtied = 1;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.pre_op[i] ||
+ local->transaction.failed_subvols[i])
+ continue;
- if (pre_nop)
- goto next;
+ switch (local->transaction.type) {
+ case AFR_DATA_TRANSACTION:
+ case AFR_METADATA_TRANSACTION:
+ if (!local->fd) {
+ STACK_WIND_COOKIE(
+ frame, afr_changelog_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->xattrop,
+ &local->loc, GF_XATTROP_ADD_ARRAY, xattr, xdata);
+ } else {
+ STACK_WIND_COOKIE(
+ frame, afr_changelog_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->fxattrop,
+ local->fd, GF_XATTROP_ADD_ARRAY, xattr, xdata);
+ }
+ break;
+ case AFR_ENTRY_RENAME_TRANSACTION:
- if (!local->pre_op_compat) {
- dict_copy (xdata_req, local->xdata_req);
- goto next;
- }
+ STACK_WIND_COOKIE(frame, afr_changelog_cbk, (void *)(long)i,
+ priv->children[i],
+ priv->children[i]->fops->xattrop,
+ &local->transaction.new_parent_loc,
+ GF_XATTROP_ADD_ARRAY, xattr, newloc_xdata);
+ call_count--;
- afr_changelog_do (frame, this, xdata_req, afr_transaction_perform_fop,
- AFR_TRANSACTION_PRE_OP);
+ /* fall through */
- if (xdata_req)
- dict_unref (xdata_req);
+ case AFR_ENTRY_TRANSACTION:
+ if (local->fd)
+ STACK_WIND_COOKIE(
+ frame, afr_changelog_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->fxattrop,
+ local->fd, GF_XATTROP_ADD_ARRAY, xattr, xdata);
+ else
+ STACK_WIND_COOKIE(frame, afr_changelog_cbk, (void *)(long)i,
+ priv->children[i],
+ priv->children[i]->fops->xattrop,
+ &local->transaction.parent_loc,
+ GF_XATTROP_ADD_ARRAY, xattr, xdata);
+ break;
+ }
- return 0;
-next:
- afr_transaction_perform_fop (frame, this);
+ if (!--call_count)
+ break;
+ }
- if (xdata_req)
- dict_unref (xdata_req);
+ if (xdata)
+ dict_unref(xdata);
+ if (newloc_xdata)
+ dict_unref(newloc_xdata);
+ return 0;
+}
- return 0;
-err:
- local->internal_lock.lock_cbk = afr_transaction_done;
- local->op_ret = -1;
- local->op_errno = op_errno;
+static void
+afr_init_optimistic_changelog_for_txn(xlator_t *this, afr_local_t *local)
+{
+ int locked_count = 0;
+ afr_private_t *priv = NULL;
- afr_handle_lock_acquire_failure (local, _gf_true);
+ priv = this->private;
- if (xdata_req)
- dict_unref (xdata_req);
+ locked_count = AFR_COUNT(local->transaction.pre_op, priv->child_count);
+ if (priv->optimistic_change_log && locked_count == priv->child_count)
+ local->optimistic_change_log = 1;
- return 0;
+ return;
}
-
int
-afr_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
+afr_changelog_pre_op(call_frame_t *frame, xlator_t *this)
+{
+ afr_private_t *priv = this->private;
+ int i = 0;
+ int ret = 0;
+ int call_count = 0;
+ int op_errno = 0;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ unsigned char *locked_nodes = NULL;
+ int idx = -1;
+ gf_boolean_t pre_nop = _gf_true;
+ dict_t *xdata_req = NULL;
+
+ local = frame->local;
+ int_lock = &local->internal_lock;
+ idx = afr_index_for_transaction_type(local->transaction.type);
+
+ locked_nodes = afr_locked_nodes_get(local->transaction.type, int_lock);
+
+ for (i = 0; i < priv->child_count; i++) {
+ if (locked_nodes[i]) {
+ local->transaction.pre_op[i] = 1;
+ call_count++;
+ } else {
+ local->transaction.failed_subvols[i] = 1;
+ }
+ }
+
+ afr_init_optimistic_changelog_for_txn(this, local);
+
+ if (afr_changelog_pre_op_inherit(frame, this))
+ goto next;
+
+ /* This condition should not be met with present code, as
+ * transaction.done will be called if locks are not acquired on even a
+ * single node.
+ */
+ if (call_count == 0) {
+ op_errno = ENOTCONN;
+ goto err;
+ }
+
+ /* Check if the fop can be performed on at least
+ * quorum number of nodes.
+ */
+ if (priv->quorum_count && !afr_has_fop_quorum(frame)) {
+ op_errno = int_lock->lock_op_errno;
+ if (op_errno == 0)
+ op_errno = afr_quorum_errno(priv);
+ goto err;
+ }
+
+ xdata_req = dict_new();
+ if (!xdata_req) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ if (call_count < priv->child_count)
+ pre_nop = _gf_false;
+
+ /* Set an all-zero pending changelog so that in the cbk, we can get the
+ * current on-disk values. In a replica 3 volume with arbiter enabled,
+ * these values are needed to arrive at a go/ no-go of the fop phase to
+ * avoid ending up in split-brain.*/
+
+ ret = afr_set_pending_dict(priv, xdata_req, local->pending);
+ if (ret < 0) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ if (afr_needs_changelog_update(local)) {
+ local->dirty[idx] = hton32(1);
+
+ ret = dict_set_static_bin(xdata_req, AFR_DIRTY, local->dirty,
+ sizeof(int) * AFR_NUM_CHANGE_LOGS);
+ if (ret) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ pre_nop = _gf_false;
+ local->transaction.dirtied = 1;
+ }
+
+ if (pre_nop)
+ goto next;
+
+ if (!local->pre_op_compat) {
+ dict_copy(xdata_req, local->xdata_req);
+ goto next;
+ }
+
+ afr_changelog_do(frame, this, xdata_req, afr_transaction_perform_fop,
+ AFR_TRANSACTION_PRE_OP);
+
+ if (xdata_req)
+ dict_unref(xdata_req);
+
+ return 0;
+next:
+ afr_transaction_perform_fop(frame, this);
- local = frame->local;
- int_lock = &local->internal_lock;
+ if (xdata_req)
+ dict_unref(xdata_req);
- /* Initiate blocking locks if non-blocking has failed */
- if (int_lock->lock_op_ret < 0) {
- gf_msg_debug (this->name, 0,
- "Non blocking inodelks failed. Proceeding to blocking");
- int_lock->lock_cbk = afr_internal_lock_finish;
- afr_blocking_lock (frame, this);
- } else {
+ return 0;
+err:
+ local->internal_lock.lock_cbk = afr_transaction_done;
+ local->op_ret = -1;
+ local->op_errno = op_errno;
- gf_msg_debug (this->name, 0,
- "Non blocking inodelks done. Proceeding to FOP");
- afr_internal_lock_finish (frame, this);
- }
+ afr_handle_lock_acquire_failure(local);
- return 0;
-}
+ if (xdata_req)
+ dict_unref(xdata_req);
+ return 0;
+}
int
-afr_post_nonblocking_entrylk_cbk (call_frame_t *frame, xlator_t *this)
+afr_post_nonblocking_lock_cbk(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
- int_lock = &local->internal_lock;
-
- /* Initiate blocking locks if non-blocking has failed */
- if (int_lock->lock_op_ret < 0) {
- gf_msg_debug (this->name, 0,
- "Non blocking entrylks failed. Proceeding to blocking");
- int_lock->lock_cbk = afr_internal_lock_finish;
- afr_blocking_lock (frame, this);
- } else {
+ local = frame->local;
+ int_lock = &local->internal_lock;
- gf_msg_debug (this->name, 0,
- "Non blocking entrylks done. Proceeding to FOP");
+ /* Initiate blocking locks if non-blocking has failed */
+ if (int_lock->lock_op_ret < 0) {
+ gf_msg_debug(this->name, 0,
+ "Non blocking locks failed. Proceeding to blocking");
+ int_lock->lock_cbk = afr_internal_lock_finish;
+ afr_blocking_lock(frame, this);
+ } else {
+ gf_msg_debug(this->name, 0,
+ "Non blocking locks done. Proceeding to FOP");
- afr_internal_lock_finish (frame, this);
- }
+ afr_internal_lock_finish(frame, this);
+ }
- return 0;
+ return 0;
}
-
int
-afr_post_blocking_rename_cbk (call_frame_t *frame, xlator_t *this)
+afr_post_blocking_rename_cbk(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
- int_lock = &local->internal_lock;
+ local = frame->local;
+ int_lock = &local->internal_lock;
- if (int_lock->lock_op_ret < 0) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- AFR_MSG_BLOCKING_LKS_FAILED,
- "Blocking entrylks failed.");
-
- afr_transaction_done (frame, this);
- } else {
+ if (int_lock->lock_op_ret < 0) {
+ gf_msg(this->name, GF_LOG_INFO, 0, AFR_MSG_INTERNAL_LKS_FAILED,
+ "Blocking entrylks failed.");
- gf_msg_debug (this->name, 0,
- "Blocking entrylks done. Proceeding to FOP");
+ afr_transaction_done(frame, this);
+ } else {
+ gf_msg_debug(this->name, 0,
+ "Blocking entrylks done. Proceeding to FOP");
- afr_internal_lock_finish (frame, this);
- }
- return 0;
+ afr_internal_lock_finish(frame, this);
+ }
+ return 0;
}
int
-afr_post_lower_unlock_cbk (call_frame_t *frame, xlator_t *this)
+afr_post_lower_unlock_cbk(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
- int_lock = &local->internal_lock;
+ local = frame->local;
+ int_lock = &local->internal_lock;
- GF_ASSERT (!int_lock->higher_locked);
+ GF_ASSERT(!int_lock->higher_locked);
- int_lock->lock_cbk = afr_post_blocking_rename_cbk;
- afr_blocking_lock (frame, this);
+ int_lock->lock_cbk = afr_post_blocking_rename_cbk;
+ afr_blocking_lock(frame, this);
- return 0;
+ return 0;
}
-
int
-afr_set_transaction_flock (xlator_t *this, afr_local_t *local)
+afr_set_transaction_flock(xlator_t *this, afr_local_t *local,
+ afr_lockee_t *lockee)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
+ afr_private_t *priv = NULL;
+ struct gf_flock *flock = NULL;
- int_lock = &local->internal_lock;
- priv = this->private;
+ priv = this->private;
+ flock = &lockee->flock;
- if ((priv->arbiter_count || local->transaction.eager_lock_on ||
- priv->full_lock) &&
- local->transaction.type == AFR_DATA_TRANSACTION) {
- /*Lock entire file to avoid network split brains.*/
- int_lock->flock.l_len = 0;
- int_lock->flock.l_start = 0;
- } else {
- int_lock->flock.l_len = local->transaction.len;
- int_lock->flock.l_start = local->transaction.start;
- }
- int_lock->flock.l_type = F_WRLCK;
+ if ((priv->arbiter_count || local->transaction.eager_lock_on ||
+ priv->full_lock) &&
+ local->transaction.type == AFR_DATA_TRANSACTION) {
+ /*Lock entire file to avoid network split brains.*/
+ flock->l_len = 0;
+ flock->l_start = 0;
+ } else {
+ flock->l_len = local->transaction.len;
+ flock->l_start = local->transaction.start;
+ }
+ flock->l_type = F_WRLCK;
- return 0;
+ return 0;
}
int
-afr_lock (call_frame_t *frame, xlator_t *this)
+afr_lock(call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ int i = 0;
- local = frame->local;
- int_lock = &local->internal_lock;
+ local = frame->local;
+ int_lock = &local->internal_lock;
- int_lock->domain = this->name;
+ int_lock->lock_cbk = afr_post_nonblocking_lock_cbk;
+ int_lock->domain = this->name;
- switch (local->transaction.type) {
+ switch (local->transaction.type) {
case AFR_DATA_TRANSACTION:
case AFR_METADATA_TRANSACTION:
- afr_set_transaction_flock (this, local);
-
- int_lock->lock_cbk = afr_post_nonblocking_inodelk_cbk;
-
- afr_nonblocking_inodelk (frame, this);
- break;
-
- case AFR_ENTRY_RENAME_TRANSACTION:
+ for (i = 0; i < int_lock->lockee_count; i++) {
+ afr_set_transaction_flock(this, local, &int_lock->lockee[i]);
+ }
- int_lock->lock_cbk = afr_post_nonblocking_entrylk_cbk;
- afr_nonblocking_entrylk (frame, this);
- break;
+ break;
case AFR_ENTRY_TRANSACTION:
- int_lock->lk_basename = local->transaction.basename;
- if (local->transaction.parent_loc.path)
- int_lock->lk_loc = &local->transaction.parent_loc;
- else
- GF_ASSERT (local->fd);
-
- int_lock->lock_cbk = afr_post_nonblocking_entrylk_cbk;
- afr_nonblocking_entrylk (frame, this);
- break;
- }
+ int_lock->lk_basename = local->transaction.basename;
+ if (local->transaction.parent_loc.path)
+ int_lock->lk_loc = &local->transaction.parent_loc;
+ else
+ GF_ASSERT(local->fd);
+ break;
+ case AFR_ENTRY_RENAME_TRANSACTION:
+ break;
+ }
+ afr_lock_nonblocking(frame, this);
- return 0;
+ return 0;
}
static gf_boolean_t
-afr_locals_overlap (afr_local_t *local1, afr_local_t *local2)
+afr_locals_overlap(afr_local_t *local1, afr_local_t *local2)
{
- uint64_t start1 = local1->transaction.start;
- uint64_t start2 = local2->transaction.start;
- uint64_t end1 = 0;
- uint64_t end2 = 0;
+ uint64_t start1 = local1->transaction.start;
+ uint64_t start2 = local2->transaction.start;
+ uint64_t end1 = 0;
+ uint64_t end2 = 0;
- if (local1->transaction.len)
- end1 = start1 + local1->transaction.len - 1;
- else
- end1 = ULLONG_MAX;
+ if (local1->transaction.len)
+ end1 = start1 + local1->transaction.len - 1;
+ else
+ end1 = ULLONG_MAX;
- if (local2->transaction.len)
- end2 = start2 + local2->transaction.len - 1;
- else
- end2 = ULLONG_MAX;
+ if (local2->transaction.len)
+ end2 = start2 + local2->transaction.len - 1;
+ else
+ end2 = ULLONG_MAX;
- return ((end1 >= start2) && (end2 >= start1));
+ return ((end1 >= start2) && (end2 >= start1));
}
gf_boolean_t
-afr_has_lock_conflict (afr_local_t *local, gf_boolean_t waitlist_check)
-{
- afr_local_t *each = NULL;
- afr_lock_t *lock = NULL;
-
- lock = &local->inode_ctx->lock[local->transaction.type];
- /*
- * Once full file lock is acquired in eager-lock phase, overlapping
- * writes do not compete for inode-locks, instead are transferred to the
- * next writes. Because of this overlapping writes are not ordered.
- * This can cause inconsistencies in replication.
- * Example:
- * Two overlapping writes w1, w2 are sent in parallel on same fd
- * in two threads t1, t2.
- * Both threads can execute afr_writev_wind in the following manner.
- * t1 winds w1 on brick-0
- * t2 winds w2 on brick-0
- * t2 winds w2 on brick-1
- * t1 winds w1 on brick-1
- *
- * This check makes sure the locks are not transferred for
- * overlapping writes.
- */
- list_for_each_entry (each, &lock->owners, transaction.owner_list) {
- if (afr_locals_overlap (each, local)) {
- return _gf_true;
- }
- }
-
- if (!waitlist_check)
- return _gf_false;
- list_for_each_entry (each, &lock->waiting, transaction.wait_list) {
- if (afr_locals_overlap (each, local)) {
- return _gf_true;
- }
- }
+afr_has_lock_conflict(afr_local_t *local, gf_boolean_t waitlist_check)
+{
+ afr_local_t *each = NULL;
+ afr_lock_t *lock = NULL;
+
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ /*
+ * Once full file lock is acquired in eager-lock phase, overlapping
+ * writes do not compete for inode-locks, instead are transferred to the
+ * next writes. Because of this overlapping writes are not ordered.
+ * This can cause inconsistencies in replication.
+ * Example:
+ * Two overlapping writes w1, w2 are sent in parallel on same fd
+ * in two threads t1, t2.
+ * Both threads can execute afr_writev_wind in the following manner.
+ * t1 winds w1 on brick-0
+ * t2 winds w2 on brick-0
+ * t2 winds w2 on brick-1
+ * t1 winds w1 on brick-1
+ *
+ * This check makes sure the locks are not transferred for
+ * overlapping writes.
+ */
+ list_for_each_entry(each, &lock->owners, transaction.owner_list)
+ {
+ if (afr_locals_overlap(each, local)) {
+ return _gf_true;
+ }
+ }
+
+ if (!waitlist_check)
return _gf_false;
+ list_for_each_entry(each, &lock->waiting, transaction.wait_list)
+ {
+ if (afr_locals_overlap(each, local)) {
+ return _gf_true;
+ }
+ }
+ return _gf_false;
}
-
/* }}} */
static void
-afr_copy_inodelk_vars (afr_internal_lock_t *dst, afr_internal_lock_t *src,
- xlator_t *this)
+afr_copy_inodelk_vars(afr_internal_lock_t *dst, afr_internal_lock_t *src,
+ xlator_t *this, int lockee_num)
{
- afr_private_t *priv = this->private;
+ afr_private_t *priv = this->private;
+ afr_lockee_t *sl = &src->lockee[lockee_num];
+ afr_lockee_t *dl = &dst->lockee[lockee_num];
- dst->domain = src->domain;
- dst->flock.l_len = src->flock.l_len;
- dst->flock.l_start = src->flock.l_start;
- dst->flock.l_type = src->flock.l_type;
- dst->lock_count = src->lock_count;
- memcpy (dst->locked_nodes, src->locked_nodes,
- priv->child_count * sizeof (*dst->locked_nodes));
+ dst->domain = src->domain;
+ dl->flock.l_len = sl->flock.l_len;
+ dl->flock.l_start = sl->flock.l_start;
+ dl->flock.l_type = sl->flock.l_type;
+ dl->locked_count = sl->locked_count;
+ memcpy(dl->locked_nodes, sl->locked_nodes,
+ priv->child_count * sizeof(*dl->locked_nodes));
}
void
-__afr_transaction_wake_shared (afr_local_t *local, struct list_head *shared)
-{
- gf_boolean_t conflict = _gf_false;
- afr_local_t *each = NULL;
- afr_lock_t *lock = &local->inode_ctx->lock[local->transaction.type];
-
- while (!conflict) {
- if (list_empty (&lock->waiting))
- return;
- each = list_entry(lock->waiting.next, afr_local_t,
- transaction.wait_list);
- if (afr_has_lock_conflict (each, _gf_false)) {
- conflict = _gf_true;
- }
- if (conflict && !list_empty (&lock->owners))
- return;
- afr_copy_inodelk_vars (&each->internal_lock,
- &local->internal_lock,
- each->transaction.frame->this);
- list_move_tail (&each->transaction.wait_list, shared);
- list_add_tail(&each->transaction.owner_list, &lock->owners);
+__afr_transaction_wake_shared(afr_local_t *local, struct list_head *shared)
+{
+ gf_boolean_t conflict = _gf_false;
+ afr_local_t *each = NULL;
+ afr_lock_t *lock = &local->inode_ctx->lock[local->transaction.type];
+
+ while (!conflict) {
+ if (list_empty(&lock->waiting))
+ return;
+ each = list_entry(lock->waiting.next, afr_local_t,
+ transaction.wait_list);
+ if (afr_has_lock_conflict(each, _gf_false)) {
+ conflict = _gf_true;
}
+ if (conflict && !list_empty(&lock->owners))
+ return;
+ afr_copy_inodelk_vars(&each->internal_lock, &local->internal_lock,
+ each->transaction.frame->this, 0);
+ list_move_tail(&each->transaction.wait_list, shared);
+ list_add_tail(&each->transaction.owner_list, &lock->owners);
+ }
}
static void
-afr_lock_resume_shared (struct list_head *list)
+afr_lock_resume_shared(struct list_head *list)
{
- afr_local_t *each = NULL;
+ afr_local_t *each = NULL;
- while (!list_empty(list)) {
- each = list_entry(list->next, afr_local_t,
- transaction.wait_list);
- list_del_init(&each->transaction.wait_list);
- afr_changelog_pre_op (each->transaction.frame,
- each->transaction.frame->this);
- }
+ while (!list_empty(list)) {
+ each = list_entry(list->next, afr_local_t, transaction.wait_list);
+ list_del_init(&each->transaction.wait_list);
+ afr_changelog_pre_op(each->transaction.frame,
+ each->transaction.frame->this);
+ }
}
int
-afr_internal_lock_finish (call_frame_t *frame, xlator_t *this)
+afr_internal_lock_finish(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = frame->local;
- afr_lock_t *lock = NULL;
-
+ afr_local_t *local = frame->local;
+ afr_lock_t *lock = NULL;
- local->internal_lock.lock_cbk = NULL;
- if (!local->transaction.eager_lock_on) {
- if (local->internal_lock.lock_op_ret < 0) {
- afr_transaction_done (frame, this);
- return 0;
- }
- afr_changelog_pre_op (frame, this);
+ local->internal_lock.lock_cbk = NULL;
+ if (!local->transaction.eager_lock_on) {
+ if (local->internal_lock.lock_op_ret < 0) {
+ afr_transaction_done(frame, this);
+ return 0;
+ }
+ afr_changelog_pre_op(frame, this);
+ } else {
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ if (local->internal_lock.lock_op_ret < 0) {
+ afr_handle_lock_acquire_failure(local);
} else {
- lock = &local->inode_ctx->lock[local->transaction.type];
- if (local->internal_lock.lock_op_ret < 0) {
- afr_handle_lock_acquire_failure (local, _gf_false);
- } else {
- lock->event_generation = local->event_generation;
- afr_changelog_pre_op (frame, this);
- }
+ lock->event_generation = local->event_generation;
+ afr_changelog_pre_op(frame, this);
}
+ }
- return 0;
+ return 0;
}
gf_boolean_t
-afr_are_multiple_fds_opened (afr_local_t *local, xlator_t *this)
+afr_are_conflicting_ops_waiting(afr_local_t *local, xlator_t *this)
{
- /* Lets say mount1 has eager-lock(full-lock) and after the eager-lock
- * is taken mount2 opened the same file, it won't be able to
- * perform any data operations until mount1 releases eager-lock.
- * To avoid such scenario do not enable eager-lock for this transaction
- * if open-fd-count is > 1
- */
+ afr_lock_t *lock = NULL;
+ lock = &local->inode_ctx->lock[local->transaction.type];
- if (local->inode_ctx->open_fd_count > 1)
- return _gf_true;
+ /* Lets say mount1 has eager-lock(full-lock) and after the eager-lock
+ * is taken mount2 opened the same file, it won't be able to
+ * perform any {meta,}data operations until mount1 releases eager-lock.
+ * To avoid such scenario do not enable eager-lock for this transaction
+ * if open-fd-count is > 1 for metadata transactions and if num-inodelks > 1
+ * for data transactions
+ */
- return _gf_false;
-}
-
-
-gf_boolean_t
-afr_is_delayed_changelog_post_op_needed (call_frame_t *frame, xlator_t *this,
- int delay)
-{
- afr_local_t *local = NULL;
- afr_lock_t *lock = NULL;
- gf_boolean_t res = _gf_false;
-
- local = frame->local;
- lock = &local->inode_ctx->lock[local->transaction.type];
-
- if (!afr_txn_nothing_failed (frame, this)) {
- lock->release = _gf_true;
- goto out;
+ if (local->transaction.type == AFR_METADATA_TRANSACTION) {
+ if (local->inode_ctx->open_fd_count > 1) {
+ return _gf_true;
}
-
- if (afr_are_multiple_fds_opened (local, this)) {
- lock->release = _gf_true;
- goto out;
+ } else if (local->transaction.type == AFR_DATA_TRANSACTION) {
+ if (lock->num_inodelks > 1) {
+ return _gf_true;
}
+ }
- if (!list_empty (&lock->owners))
- goto out;
- else
- GF_ASSERT (list_empty (&lock->waiting));
-
- if (lock->release) {
- goto out;
- }
-
- if (!delay) {
- goto out;
- }
-
- if ((local->op != GF_FOP_WRITE) &&
- (local->op != GF_FOP_FXATTROP)) {
- /*Only allow writes but shard does [f]xattrops on writes, so
- * they are fine too*/
- goto out;
- }
+ return _gf_false;
+}
- res = _gf_true;
+gf_boolean_t
+afr_is_delayed_changelog_post_op_needed(call_frame_t *frame, xlator_t *this,
+ int delay)
+{
+ afr_local_t *local = NULL;
+ afr_lock_t *lock = NULL;
+ gf_boolean_t res = _gf_false;
+
+ local = frame->local;
+ lock = &local->inode_ctx->lock[local->transaction.type];
+
+ if (!afr_txn_nothing_failed(frame, this)) {
+ lock->release = _gf_true;
+ goto out;
+ }
+
+ if (afr_are_conflicting_ops_waiting(local, this)) {
+ lock->release = _gf_true;
+ goto out;
+ }
+
+ if (!list_empty(&lock->owners))
+ goto out;
+ else
+ GF_ASSERT(list_empty(&lock->waiting));
+
+ if (lock->release) {
+ goto out;
+ }
+
+ if (!delay) {
+ goto out;
+ }
+
+ if (local->transaction.disable_delayed_post_op) {
+ goto out;
+ }
+
+ if ((local->op != GF_FOP_WRITE) && (local->op != GF_FOP_FXATTROP) &&
+ (local->op != GF_FOP_FSYNC)) {
+ /*Only allow writes/fsyncs but shard does [f]xattrops on writes, so
+ * they are fine too*/
+ goto out;
+ }
+
+ res = _gf_true;
out:
- return res;
+ return res;
}
-
void
-afr_delayed_changelog_wake_up_cbk (void *data)
-{
- afr_lock_t *lock = NULL;
- afr_local_t *local = data;
- afr_local_t *timer_local = NULL;
- struct list_head shared;
-
- INIT_LIST_HEAD (&shared);
- lock = &local->inode_ctx->lock[local->transaction.type];
- LOCK (&local->inode->lock);
- {
- timer_local = list_entry(lock->post_op.next,
- afr_local_t,
- transaction.owner_list);
- if (list_empty (&lock->owners) && (local == timer_local)) {
- GF_ASSERT (list_empty (&lock->waiting));
- /*Last owner*/
- lock->release = _gf_true;
- lock->delay_timer = NULL;
- }
- }
- UNLOCK (&local->inode->lock);
- afr_changelog_post_op_now (local->transaction.frame,
- local->transaction.frame->this);
+afr_delayed_changelog_wake_up_cbk(void *data)
+{
+ afr_lock_t *lock = NULL;
+ afr_local_t *local = data;
+ afr_local_t *timer_local = NULL;
+ struct list_head shared;
+
+ INIT_LIST_HEAD(&shared);
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ LOCK(&local->inode->lock);
+ {
+ timer_local = list_entry(lock->post_op.next, afr_local_t,
+ transaction.owner_list);
+ if (list_empty(&lock->owners) && (local == timer_local)) {
+ GF_ASSERT(list_empty(&lock->waiting));
+ /*Last owner*/
+ lock->release = _gf_true;
+ lock->delay_timer = NULL;
+ }
+ }
+ UNLOCK(&local->inode->lock);
+ afr_changelog_post_op_now(local->transaction.frame,
+ local->transaction.frame->this);
}
-
/* SET operation */
int
-afr_fd_report_unstable_write (xlator_t *this, afr_local_t *local)
+afr_fd_report_unstable_write(xlator_t *this, afr_local_t *local)
{
- LOCK(&local->inode->lock);
- {
- local->inode_ctx->witnessed_unstable_write = _gf_true;
- }
- UNLOCK(&local->inode->lock);
+ LOCK(&local->inode->lock);
+ {
+ local->inode_ctx->witnessed_unstable_write = _gf_true;
+ }
+ UNLOCK(&local->inode->lock);
- return 0;
+ return 0;
}
/* TEST and CLEAR operation */
gf_boolean_t
-afr_fd_has_witnessed_unstable_write (xlator_t *this, inode_t *inode)
+afr_fd_has_witnessed_unstable_write(xlator_t *this, inode_t *inode)
{
- afr_inode_ctx_t *ctx = NULL;
- gf_boolean_t witness = _gf_false;
+ afr_inode_ctx_t *ctx = NULL;
+ gf_boolean_t witness = _gf_false;
- LOCK(&inode->lock);
- {
- (void)__afr_inode_ctx_get (this, inode, &ctx);
+ LOCK(&inode->lock);
+ {
+ (void)__afr_inode_ctx_get(this, inode, &ctx);
- if (ctx->witnessed_unstable_write) {
- witness = _gf_true;
- ctx->witnessed_unstable_write = _gf_false;
- }
+ if (ctx->witnessed_unstable_write) {
+ witness = _gf_true;
+ ctx->witnessed_unstable_write = _gf_false;
}
- UNLOCK (&inode->lock);
+ }
+ UNLOCK(&inode->lock);
- return witness;
+ return witness;
}
-
int
-afr_changelog_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *pre,
- struct iatt *post, dict_t *xdata)
+afr_changelog_fsync_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *pre,
+ struct iatt *post, dict_t *xdata)
{
- afr_private_t *priv = NULL;
- int child_index = (long) cookie;
- int call_count = -1;
- afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int child_index = (long)cookie;
+ int call_count = -1;
+ afr_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- if (op_ret != 0) {
- /* Failure of fsync() is as good as failure of previous
- write(). So treat it like one.
- */
- gf_msg (this->name, GF_LOG_WARNING,
- op_errno, AFR_MSG_FSYNC_FAILED,
- "fsync(%s) failed on subvolume %s. Transaction was %s",
- uuid_utoa (local->fd->inode->gfid),
- priv->children[child_index]->name,
- gf_fop_list[local->op]);
+ if (op_ret != 0) {
+ /* Failure of fsync() is as good as failure of previous
+ write(). So treat it like one.
+ */
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, AFR_MSG_FSYNC_FAILED,
+ "fsync(%s) failed on subvolume %s. Transaction was %s",
+ uuid_utoa(local->fd->inode->gfid),
+ priv->children[child_index]->name, gf_fop_list[local->op]);
- afr_transaction_fop_failed (frame, this, child_index);
- }
+ afr_transaction_fop_failed(frame, this, child_index);
+ }
- call_count = afr_frame_return (frame);
+ call_count = afr_frame_return(frame);
- if (call_count == 0)
- afr_changelog_post_op_now (frame, this);
+ if (call_count == 0)
+ afr_changelog_post_op_now(frame, this);
- return 0;
+ return 0;
}
-
int
-afr_changelog_fsync (call_frame_t *frame, xlator_t *this)
+afr_changelog_fsync(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- int i = 0;
- int call_count = 0;
- afr_private_t *priv = NULL;
- dict_t *xdata = NULL;
- GF_UNUSED int ret = -1;
+ afr_local_t *local = NULL;
+ int i = 0;
+ int call_count = 0;
+ afr_private_t *priv = NULL;
+ dict_t *xdata = NULL;
+ GF_UNUSED int ret = -1;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- call_count = AFR_COUNT (local->transaction.pre_op, priv->child_count);
+ call_count = AFR_COUNT(local->transaction.pre_op, priv->child_count);
- if (!call_count) {
- /* will go straight to unlock */
- afr_changelog_post_op_now (frame, this);
- return 0;
- }
+ if (!call_count) {
+ /* will go straight to unlock */
+ afr_changelog_post_op_now(frame, this);
+ return 0;
+ }
- local->call_count = call_count;
+ local->call_count = call_count;
- xdata = dict_new();
- if (xdata)
- ret = dict_set_int32 (xdata, "batch-fsync", 1);
+ xdata = dict_new();
+ if (xdata) {
+ ret = dict_set_int32_sizen(xdata, "batch-fsync", 1);
+ ret = dict_set_str(xdata, GLUSTERFS_INTERNAL_FOP_KEY, "yes");
+ }
- for (i = 0; i < priv->child_count; i++) {
- if (!local->transaction.pre_op[i])
- continue;
+ for (i = 0; i < priv->child_count; i++) {
+ if (!local->transaction.pre_op[i])
+ continue;
- STACK_WIND_COOKIE (frame, afr_changelog_fsync_cbk,
- (void *) (long) i, priv->children[i],
- priv->children[i]->fops->fsync, local->fd,
- 1, xdata);
- if (!--call_count)
- break;
- }
+ STACK_WIND_COOKIE(frame, afr_changelog_fsync_cbk, (void *)(long)i,
+ priv->children[i], priv->children[i]->fops->fsync,
+ local->fd, 1, xdata);
+ if (!--call_count)
+ break;
+ }
- if (xdata)
- dict_unref (xdata);
+ if (xdata)
+ dict_unref(xdata);
- return 0;
+ return 0;
}
-
int
-afr_changelog_post_op_safe (call_frame_t *frame, xlator_t *this)
+afr_changelog_post_op_safe(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- if (!local->fd || local->transaction.type != AFR_DATA_TRANSACTION) {
- afr_changelog_post_op_now (frame, this);
- return 0;
- }
-
- if (afr_changelog_pre_op_uninherit (frame, this) &&
- afr_txn_nothing_failed (frame, this)) {
- /* just detected that this post-op is about to
- be optimized away as a new write() has
- already piggybacked on this frame's changelog.
- */
- afr_changelog_post_op_now (frame, this);
- return 0;
- }
+ if (!local->fd || local->transaction.type != AFR_DATA_TRANSACTION) {
+ afr_changelog_post_op_now(frame, this);
+ return 0;
+ }
- /* Calling afr_changelog_post_op_now() now will result in
- issuing ->[f]xattrop().
-
- Performing a hard POST-OP (->[f]xattrop() FOP) is a more
- responsible operation that what it might appear on the surface.
-
- The changelog of a file (in the xattr of the file on the server)
- stores information (pending count) about the state of the file
- on the OTHER server. This changelog is blindly trusted, and must
- therefore be updated in such a way it remains trustworthy. This
- implies that decrementing the pending count (essentially "clearing
- the dirty flag") must be done STRICTLY after we are sure that the
- operation on the other server has reached stable storage.
-
- While the backend filesystem on that server will eventually flush
- it to stable storage, we (being in userspace) have no mechanism
- to get notified when the write became "stable".
-
- This means we need take matter into our own hands and issue an
- fsync() EVEN IF THE APPLICATION WAS PERFORMING UNSTABLE WRITES,
- and get an acknowledgement for it. And we need to wait for the
- fsync() acknowledgement before initiating the hard POST-OP.
-
- However if the FD itself was opened in O_SYNC or O_DSYNC then
- we are already guaranteed that the writes were made stable as
- part of the FOP itself. The same holds true for NFS stable
- writes which happen on an anonymous FD with O_DSYNC or O_SYNC
- flag set in the writev() @flags param. For all other write types,
- mark a flag in the fdctx whenever an unstable write is witnessed.
+ if (afr_changelog_pre_op_uninherit(frame, this) &&
+ afr_txn_nothing_failed(frame, this)) {
+ /* just detected that this post-op is about to
+ be optimized away as a new write() has
+ already piggybacked on this frame's changelog.
*/
+ afr_changelog_post_op_now(frame, this);
+ return 0;
+ }
+
+ /* Calling afr_changelog_post_op_now() now will result in
+ issuing ->[f]xattrop().
+
+ Performing a hard POST-OP (->[f]xattrop() FOP) is a more
+ responsible operation that what it might appear on the surface.
+
+ The changelog of a file (in the xattr of the file on the server)
+ stores information (pending count) about the state of the file
+ on the OTHER server. This changelog is blindly trusted, and must
+ therefore be updated in such a way it remains trustworthy. This
+ implies that decrementing the pending count (essentially "clearing
+ the dirty flag") must be done STRICTLY after we are sure that the
+ operation on the other server has reached stable storage.
+
+ While the backend filesystem on that server will eventually flush
+ it to stable storage, we (being in userspace) have no mechanism
+ to get notified when the write became "stable".
+
+ This means we need take matter into our own hands and issue an
+ fsync() EVEN IF THE APPLICATION WAS PERFORMING UNSTABLE WRITES,
+ and get an acknowledgement for it. And we need to wait for the
+ fsync() acknowledgement before initiating the hard POST-OP.
+
+ However if the FD itself was opened in O_SYNC or O_DSYNC then
+ we are already guaranteed that the writes were made stable as
+ part of the FOP itself. The same holds true for NFS stable
+ writes which happen on an anonymous FD with O_DSYNC or O_SYNC
+ flag set in the writev() @flags param. For all other write types,
+ mark a flag in the fdctx whenever an unstable write is witnessed.
+ */
+
+ if (!afr_fd_has_witnessed_unstable_write(this, local->inode)) {
+ afr_changelog_post_op_now(frame, this);
+ return 0;
+ }
- if (!afr_fd_has_witnessed_unstable_write (this, local->inode)) {
- afr_changelog_post_op_now (frame, this);
- return 0;
- }
-
- /* Check whether users want durability and perform fsync/post-op
- * accordingly.
- */
- if (priv->ensure_durability) {
- /* Time to fsync() */
- afr_changelog_fsync (frame, this);
- } else {
- afr_changelog_post_op_now (frame, this);
- }
+ /* Check whether users want durability and perform fsync/post-op
+ * accordingly.
+ */
+ if (priv->ensure_durability) {
+ /* Time to fsync() */
+ afr_changelog_fsync(frame, this);
+ } else {
+ afr_changelog_post_op_now(frame, this);
+ }
- return 0;
+ return 0;
}
void
-afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
-{
- struct timespec delta = {0, };
- afr_private_t *priv = NULL;
- afr_local_t *local = frame->local;
- afr_lock_t *lock = NULL;
- gf_boolean_t post_op = _gf_true;
- struct list_head shared;
-
- priv = this->private;
- delta.tv_sec = priv->post_op_delay_secs;
- delta.tv_nsec = 0;
-
- INIT_LIST_HEAD (&shared);
- if (!local->transaction.eager_lock_on)
- goto out;
-
- lock = &local->inode_ctx->lock[local->transaction.type];
- LOCK (&local->inode->lock);
- {
- list_del_init (&local->transaction.owner_list);
- list_add (&local->transaction.owner_list, &lock->post_op);
- __afr_transaction_wake_shared (local, &shared);
-
- if (!afr_is_delayed_changelog_post_op_needed (frame, this,
- delta.tv_sec)) {
- if (list_empty (&lock->owners))
- lock->release = _gf_true;
- goto unlock;
- }
-
- GF_ASSERT (lock->delay_timer == NULL);
- lock->delay_timer = gf_timer_call_after (this->ctx, delta,
- afr_delayed_changelog_wake_up_cbk,
- local);
- if (!lock->delay_timer) {
- lock->release = _gf_true;
- } else {
- post_op = _gf_false;
- }
+afr_changelog_post_op(call_frame_t *frame, xlator_t *this)
+{
+ struct timespec delta = {
+ 0,
+ };
+ afr_private_t *priv = NULL;
+ afr_local_t *local = frame->local;
+ afr_lock_t *lock = NULL;
+ gf_boolean_t post_op = _gf_true;
+ struct list_head shared;
+
+ priv = this->private;
+ delta.tv_sec = priv->post_op_delay_secs;
+ delta.tv_nsec = 0;
+
+ INIT_LIST_HEAD(&shared);
+ if (!local->transaction.eager_lock_on)
+ goto out;
+
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ LOCK(&local->inode->lock);
+ {
+ list_del_init(&local->transaction.owner_list);
+ list_add(&local->transaction.owner_list, &lock->post_op);
+ __afr_transaction_wake_shared(local, &shared);
+
+ if (!afr_is_delayed_changelog_post_op_needed(frame, this,
+ delta.tv_sec)) {
+ if (list_empty(&lock->owners))
+ lock->release = _gf_true;
+ goto unlock;
+ }
- }
+ GF_ASSERT(lock->delay_timer == NULL);
+ lock->delay_timer = gf_timer_call_after(
+ this->ctx, delta, afr_delayed_changelog_wake_up_cbk, local);
+ if (!lock->delay_timer) {
+ lock->release = _gf_true;
+ } else {
+ post_op = _gf_false;
+ }
+ }
unlock:
- UNLOCK (&local->inode->lock);
+ UNLOCK(&local->inode->lock);
- if (!list_empty (&shared)) {
- afr_lock_resume_shared (&shared);
- }
+ if (!list_empty(&shared)) {
+ afr_lock_resume_shared(&shared);
+ }
out:
- if (post_op) {
- if (!local->transaction.eager_lock_on || lock->release) {
- afr_changelog_post_op_safe (frame, this);
- } else {
- afr_changelog_post_op_now (frame, this);
- }
+ if (post_op) {
+ if (!local->transaction.eager_lock_on || lock->release) {
+ afr_changelog_post_op_safe(frame, this);
+ } else {
+ afr_changelog_post_op_now(frame, this);
}
+ }
}
int
-afr_transaction_resume (call_frame_t *frame, xlator_t *this)
+afr_transaction_resume(call_frame_t *frame, xlator_t *this)
{
- afr_local_t *local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- afr_restore_lk_owner (frame);
+ afr_restore_lk_owner(frame);
- afr_handle_symmetric_errors (frame, this);
+ afr_handle_symmetric_errors(frame, this);
- if (!local->pre_op_compat)
- /* new mode, pre-op was done along
- with OP */
- afr_changelog_pre_op_update (frame, this);
+ if (!local->pre_op_compat)
+ /* new mode, pre-op was done along
+ with OP */
+ afr_changelog_pre_op_update(frame, this);
- afr_changelog_post_op (frame, this);
+ afr_changelog_post_op(frame, this);
- return 0;
+ return 0;
}
-
/**
* afr_transaction_fop_failed - inform that an fop failed
*/
void
-afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this,
- int child_index)
+afr_transaction_fop_failed(call_frame_t *frame, xlator_t *this, int child_index)
{
- afr_local_t * local = NULL;
+ afr_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- local->transaction.failed_subvols[child_index] = 1;
+ local->transaction.failed_subvols[child_index] = 1;
}
static gf_boolean_t
-__need_previous_lock_unlocked (afr_local_t *local)
+__need_previous_lock_unlocked(afr_local_t *local)
{
- afr_lock_t *lock = NULL;
-
- if (!local->transaction.eager_lock_on)
- return _gf_true;
+ afr_lock_t *lock = NULL;
- lock = &local->inode_ctx->lock[local->transaction.type];
- if (!lock->acquired)
- return _gf_false;
- if (lock->acquired && lock->event_generation != local->event_generation)
- return _gf_true;
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ if (!lock->acquired)
return _gf_false;
+ if (lock->acquired && lock->event_generation != local->event_generation)
+ return _gf_true;
+ return _gf_false;
}
void
-__afr_eager_lock_handle (afr_local_t *local, gf_boolean_t *take_lock,
- gf_boolean_t *do_pre_op, afr_local_t **timer_local)
-{
- afr_lock_t *lock = NULL;
- afr_local_t *owner_local = NULL;
- xlator_t *this = local->transaction.frame->this;
-
- if (local->fd && !afr_are_multiple_fds_opened (local, this)) {
- local->transaction.eager_lock_on = _gf_true;
- }
-
- lock = &local->inode_ctx->lock[local->transaction.type];
- if (__need_previous_lock_unlocked (local)) {
- if (!list_empty (&lock->owners)) {
- lock->release = _gf_true;
- } else if (lock->delay_timer) {
- lock->release = _gf_true;
- if (gf_timer_call_cancel (this->ctx,
- lock->delay_timer)) {
- /* It will be put in frozen list
- * in the code flow below*/
- } else {
- *timer_local = list_entry(lock->post_op.next,
- afr_local_t,
- transaction.owner_list);
- lock->delay_timer = NULL;
- }
- }
- if (!local->transaction.eager_lock_on)
- goto out;
- }
-
- if (lock->release) {
- list_add_tail (&local->transaction.wait_list,
- &lock->frozen);
- *take_lock = _gf_false;
- goto out;
- }
-
- if (lock->delay_timer) {
- *take_lock = _gf_false;
- if (gf_timer_call_cancel (this->ctx,
- lock->delay_timer)) {
- list_add_tail (&local->transaction.wait_list,
- &lock->frozen);
- } else {
- *timer_local = list_entry(lock->post_op.next,
- afr_local_t,
- transaction.owner_list);
- afr_copy_inodelk_vars (&local->internal_lock,
- &(*timer_local)->internal_lock,
- this);
- lock->delay_timer = NULL;
- *do_pre_op = _gf_true;
- list_add_tail (&local->transaction.owner_list,
- &lock->owners);
- }
- goto out;
- }
-
- if (!list_empty (&lock->owners)) {
- if (!lock->acquired ||
- afr_has_lock_conflict (local, _gf_true)) {
- list_add_tail (&local->transaction.wait_list,
- &lock->waiting);
- *take_lock = _gf_false;
- goto out;
- }
- owner_local = list_entry (lock->owners.next,
- afr_local_t,
+__afr_eager_lock_handle(afr_local_t *local, gf_boolean_t *take_lock,
+ gf_boolean_t *do_pre_op, afr_local_t **timer_local)
+{
+ afr_lock_t *lock = NULL;
+ afr_local_t *owner_local = NULL;
+ xlator_t *this = local->transaction.frame->this;
+
+ local->transaction.eager_lock_on = _gf_true;
+ afr_set_lk_owner(local->transaction.frame, this, local->inode);
+
+ lock = &local->inode_ctx->lock[local->transaction.type];
+ if (__need_previous_lock_unlocked(local)) {
+ if (!list_empty(&lock->owners)) {
+ lock->release = _gf_true;
+ } else if (lock->delay_timer) {
+ lock->release = _gf_true;
+ if (gf_timer_call_cancel(this->ctx, lock->delay_timer)) {
+ /* It will be put in frozen list
+ * in the code flow below*/
+ } else {
+ *timer_local = list_entry(lock->post_op.next, afr_local_t,
transaction.owner_list);
- afr_copy_inodelk_vars (&local->internal_lock,
- &owner_local->internal_lock,
- this);
- *take_lock = _gf_false;
- *do_pre_op = _gf_true;
+ lock->delay_timer = NULL;
+ }
}
+ }
+
+ if (lock->release) {
+ list_add_tail(&local->transaction.wait_list, &lock->frozen);
+ *take_lock = _gf_false;
+ goto out;
+ }
- if (lock->acquired)
- GF_ASSERT (!(*take_lock));
- list_add_tail (&local->transaction.owner_list, &lock->owners);
+ if (lock->delay_timer) {
+ *take_lock = _gf_false;
+ if (gf_timer_call_cancel(this->ctx, lock->delay_timer)) {
+ list_add_tail(&local->transaction.wait_list, &lock->frozen);
+ } else {
+ *timer_local = list_entry(lock->post_op.next, afr_local_t,
+ transaction.owner_list);
+ afr_copy_inodelk_vars(&local->internal_lock,
+ &(*timer_local)->internal_lock, this, 0);
+ lock->delay_timer = NULL;
+ *do_pre_op = _gf_true;
+ list_add_tail(&local->transaction.owner_list, &lock->owners);
+ }
+ goto out;
+ }
+
+ if (!list_empty(&lock->owners)) {
+ if (!lock->acquired || afr_has_lock_conflict(local, _gf_true)) {
+ list_add_tail(&local->transaction.wait_list, &lock->waiting);
+ *take_lock = _gf_false;
+ goto out;
+ }
+ owner_local = list_entry(lock->owners.next, afr_local_t,
+ transaction.owner_list);
+ afr_copy_inodelk_vars(&local->internal_lock,
+ &owner_local->internal_lock, this, 0);
+ *take_lock = _gf_false;
+ *do_pre_op = _gf_true;
+ }
+
+ if (lock->acquired)
+ GF_ASSERT(!(*take_lock));
+ list_add_tail(&local->transaction.owner_list, &lock->owners);
out:
- return;
+ return;
}
void
-afr_transaction_start (afr_local_t *local, xlator_t *this)
+afr_transaction_start(afr_local_t *local, xlator_t *this)
{
- afr_private_t *priv = NULL;
- gf_boolean_t take_lock = _gf_true;
- gf_boolean_t do_pre_op = _gf_false;
- afr_local_t *timer_local = NULL;
+ afr_private_t *priv = NULL;
+ gf_boolean_t take_lock = _gf_true;
+ gf_boolean_t do_pre_op = _gf_false;
+ afr_local_t *timer_local = NULL;
- priv = this->private;
+ priv = this->private;
- if (local->transaction.type != AFR_DATA_TRANSACTION &&
- local->transaction.type != AFR_METADATA_TRANSACTION)
- goto lock_phase;
+ if (local->transaction.type != AFR_DATA_TRANSACTION &&
+ local->transaction.type != AFR_METADATA_TRANSACTION)
+ goto lock_phase;
- if (!priv->eager_lock)
- goto lock_phase;
+ if (!priv->eager_lock)
+ goto lock_phase;
- LOCK (&local->inode->lock);
- {
- __afr_eager_lock_handle (local, &take_lock, &do_pre_op,
- &timer_local);
- }
- UNLOCK (&local->inode->lock);
+ LOCK(&local->inode->lock);
+ {
+ __afr_eager_lock_handle(local, &take_lock, &do_pre_op, &timer_local);
+ }
+ UNLOCK(&local->inode->lock);
lock_phase:
- if (!local->transaction.eager_lock_on) {
- afr_set_lk_owner (local->transaction.frame, this,
- local->transaction.frame->root);
- } else {
- afr_set_lk_owner (local->transaction.frame, this, local->inode);
- }
-
-
- if (take_lock) {
- afr_lock (local->transaction.frame, this);
- } else if (do_pre_op) {
- afr_changelog_pre_op (local->transaction.frame, this);
- }
- /*Always call delayed_changelog_wake_up_cbk after calling pre-op above
- * so that any inheriting can happen*/
- if (timer_local)
- afr_delayed_changelog_wake_up_cbk (timer_local);
+ if (!local->transaction.eager_lock_on) {
+ afr_set_lk_owner(local->transaction.frame, this,
+ local->transaction.frame->root);
+ }
+
+ if (take_lock) {
+ afr_lock(local->transaction.frame, this);
+ } else if (do_pre_op) {
+ afr_changelog_pre_op(local->transaction.frame, this);
+ }
+ /*Always call delayed_changelog_wake_up_cbk after calling pre-op above
+ * so that any inheriting can happen*/
+ if (timer_local)
+ afr_delayed_changelog_wake_up_cbk(timer_local);
}
int
-afr_write_txn_refresh_done (call_frame_t *frame, xlator_t *this, int err)
+afr_write_txn_refresh_done(call_frame_t *frame, xlator_t *this, int err)
{
- afr_local_t *local = frame->local;
+ afr_local_t *local = frame->local;
- if (err) {
- AFR_SET_ERROR_AND_CHECK_SPLIT_BRAIN(-1, -err);
- goto fail;
- }
+ if (err) {
+ AFR_SET_ERROR_AND_CHECK_SPLIT_BRAIN(-1, err);
+ goto fail;
+ }
- afr_transaction_start (local, this);
- return 0;
+ afr_transaction_start(local, this);
+ return 0;
fail:
- local->transaction.unwind (frame, this);
- AFR_STACK_DESTROY (frame);
- return 0;
+ local->transaction.unwind(frame, this);
+ AFR_STACK_DESTROY(frame);
+ return 0;
}
int
-afr_transaction (call_frame_t *frame, xlator_t *this, afr_transaction_type type)
+afr_transaction_lockee_init(call_frame_t *frame)
{
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- int ret = -1;
- int event_generation = 0;
-
- local = frame->local;
- priv = this->private;
- local->transaction.frame = frame;
-
- local->transaction.type = type;
+ afr_local_t *local = frame->local;
+ afr_internal_lock_t *int_lock = &local->internal_lock;
+ afr_private_t *priv = frame->this->private;
+ int ret = 0;
- if (!afr_is_consistent_io_possible (local, priv, &ret)) {
- ret = -ret; /*op_errno to ret conversion*/
- goto out;
- }
+ switch (local->transaction.type) {
+ case AFR_DATA_TRANSACTION:
+ case AFR_METADATA_TRANSACTION:
+ ret = afr_add_inode_lockee(local, priv->child_count);
+ break;
- ret = afr_transaction_local_init (local, this);
- if (ret < 0)
+ case AFR_ENTRY_TRANSACTION:
+ case AFR_ENTRY_RENAME_TRANSACTION:
+ ret = afr_add_entry_lockee(local, &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret) {
goto out;
+ }
+ if (local->op == GF_FOP_RENAME) {
+ ret = afr_add_entry_lockee(
+ local, &local->transaction.new_parent_loc,
+ local->transaction.new_basename, priv->child_count);
+ if (ret) {
+ goto out;
+ }
+ if (local->newloc.inode &&
+ IA_ISDIR(local->newloc.inode->ia_type)) {
+ ret = afr_add_entry_lockee(local, &local->newloc, NULL,
+ priv->child_count);
+ if (ret) {
+ goto out;
+ }
+ }
+ } else if (local->op == GF_FOP_RMDIR) {
+ ret = afr_add_entry_lockee(local, &local->loc, NULL,
+ priv->child_count);
+ if (ret) {
+ goto out;
+ }
+ }
+
+ if (int_lock->lockee_count > 1) {
+ qsort(int_lock->lockee, int_lock->lockee_count,
+ sizeof(*int_lock->lockee), afr_entry_lockee_cmp);
+ }
+ break;
+ }
+out:
+ return ret;
+}
- if (type != AFR_METADATA_TRANSACTION) {
- goto txn_start;
- }
-
- ret = afr_inode_get_readable (frame, local->inode, this,
- local->readable, &event_generation, type);
- if (ret < 0 || afr_is_inode_refresh_reqd (local->inode, this,
- priv->event_generation,
- event_generation)) {
- afr_inode_refresh (frame, this, local->inode, local->loc.gfid,
- afr_write_txn_refresh_done);
- ret = 0;
- goto out;
- }
+int
+afr_transaction(call_frame_t *frame, xlator_t *this, afr_transaction_type type)
+{
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int ret = -1;
+ int event_generation = 0;
+
+ local = frame->local;
+ priv = this->private;
+ local->transaction.frame = frame;
+
+ local->transaction.type = type;
+
+ if (priv->quorum_count && !afr_has_quorum(local->child_up, this, NULL)) {
+ ret = -afr_quorum_errno(priv);
+ goto out;
+ }
+
+ if (!afr_is_consistent_io_possible(local, priv, &ret)) {
+ ret = -ret; /*op_errno to ret conversion*/
+ goto out;
+ }
+
+ if (priv->thin_arbiter_count && !afr_ta_has_quorum(priv, local)) {
+ ret = -afr_quorum_errno(priv);
+ goto out;
+ }
+
+ ret = afr_transaction_local_init(local, this);
+ if (ret < 0)
+ goto out;
+
+ ret = afr_transaction_lockee_init(frame);
+ if (ret)
+ goto out;
+
+ if (type != AFR_METADATA_TRANSACTION) {
+ goto txn_start;
+ }
+
+ ret = afr_inode_get_readable(frame, local->inode, this, local->readable,
+ &event_generation, type);
+ if (ret < 0 ||
+ afr_is_inode_refresh_reqd(local->inode, this, priv->event_generation,
+ event_generation)) {
+ afr_inode_refresh(frame, this, local->inode, local->loc.gfid,
+ afr_write_txn_refresh_done);
+ ret = 0;
+ goto out;
+ }
txn_start:
- ret = 0;
- afr_transaction_start (local, this);
+ ret = 0;
+ afr_transaction_start(local, this);
out:
- return ret;
+ return ret;
}
diff --git a/xlators/cluster/afr/src/afr-transaction.h b/xlators/cluster/afr/src/afr-transaction.h
index 629f6dd557c..beefa26f4a6 100644
--- a/xlators/cluster/afr/src/afr-transaction.h
+++ b/xlators/cluster/afr/src/afr-transaction.h
@@ -14,53 +14,62 @@
#include "afr.h"
void
-afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this,
- int child_index);
+afr_transaction_fop_failed(call_frame_t *frame, xlator_t *this,
+ int child_index);
int32_t
-afr_transaction (call_frame_t *frame, xlator_t *this, afr_transaction_type type);
+afr_transaction(call_frame_t *frame, xlator_t *this, afr_transaction_type type);
int
-afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending);
+afr_set_pending_dict(afr_private_t *priv, dict_t *xattr, int32_t **pending);
void
-afr_delayed_changelog_wake_up (xlator_t *this, fd_t *fd);
+afr_delayed_changelog_wake_up(xlator_t *this, fd_t *fd);
void
-__mark_all_success (call_frame_t *frame, xlator_t *this);
+__mark_all_success(call_frame_t *frame, xlator_t *this);
gf_boolean_t
-afr_txn_nothing_failed (call_frame_t *frame, xlator_t *this);
+afr_txn_nothing_failed(call_frame_t *frame, xlator_t *this);
-int afr_read_txn (call_frame_t *frame, xlator_t *this, inode_t *inode,
- afr_read_txn_wind_t readfn, afr_transaction_type type);
+int
+afr_read_txn(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ afr_read_txn_wind_t readfn, afr_transaction_type type);
-int afr_read_txn_continue (call_frame_t *frame, xlator_t *this, int subvol);
+int
+afr_read_txn_continue(call_frame_t *frame, xlator_t *this, int subvol);
void
-afr_pending_read_increment (afr_private_t *priv, int child_index);
+afr_pending_read_increment(afr_private_t *priv, int child_index);
void
-afr_pending_read_decrement (afr_private_t *priv, int child_index);
+afr_pending_read_decrement(afr_private_t *priv, int child_index);
-call_frame_t *afr_transaction_detach_fop_frame (call_frame_t *frame);
-gf_boolean_t afr_has_quorum (unsigned char *subvols, xlator_t *this);
-gf_boolean_t afr_needs_changelog_update (afr_local_t *local);
-void afr_zero_fill_stat (afr_local_t *local);
+call_frame_t *
+afr_transaction_detach_fop_frame(call_frame_t *frame);
+gf_boolean_t
+afr_has_quorum(unsigned char *subvols, xlator_t *this, call_frame_t *frame);
+gf_boolean_t
+afr_needs_changelog_update(afr_local_t *local);
+void
+afr_zero_fill_stat(afr_local_t *local);
void
-afr_pick_error_xdata (afr_local_t *local, afr_private_t *priv,
- inode_t *inode1, unsigned char *readable1,
- inode_t *inode2, unsigned char *readable2);
+afr_pick_error_xdata(afr_local_t *local, afr_private_t *priv, inode_t *inode1,
+ unsigned char *readable1, inode_t *inode2,
+ unsigned char *readable2);
int
-afr_transaction_resume (call_frame_t *frame, xlator_t *this);
+afr_transaction_resume(call_frame_t *frame, xlator_t *this);
int
-afr_lock (call_frame_t *frame, xlator_t *this);
+afr_lock(call_frame_t *frame, xlator_t *this);
void
-afr_delayed_changelog_wake_up_cbk (void *data);
+afr_delayed_changelog_wake_up_cbk(void *data);
+
+int
+afr_release_notify_lock_for_ta(void *opaque);
int
-afr_fill_ta_loc (xlator_t *this, loc_t *loc);
+afr_ta_lock_release_done(int ret, call_frame_t *ta_frame, void *opaque);
#endif /* __TRANSACTION_H__ */
diff --git a/xlators/cluster/afr/src/afr.c b/xlators/cluster/afr/src/afr.c
index 0d6773c96e8..df7366f0a65 100644
--- a/xlators/cluster/afr/src/afr.c
+++ b/xlators/cluster/afr/src/afr.c
@@ -21,1194 +21,1324 @@
struct volume_options options[];
static char *afr_favorite_child_policies[AFR_FAV_CHILD_POLICY_MAX + 1] = {
- [AFR_FAV_CHILD_NONE] = "none",
- [AFR_FAV_CHILD_BY_SIZE] = "size",
- [AFR_FAV_CHILD_BY_CTIME] = "ctime",
- [AFR_FAV_CHILD_BY_MTIME] = "mtime",
- [AFR_FAV_CHILD_BY_MAJORITY] = "majority",
- [AFR_FAV_CHILD_POLICY_MAX] = NULL,
+ [AFR_FAV_CHILD_NONE] = "none",
+ [AFR_FAV_CHILD_BY_SIZE] = "size",
+ [AFR_FAV_CHILD_BY_CTIME] = "ctime",
+ [AFR_FAV_CHILD_BY_MTIME] = "mtime",
+ [AFR_FAV_CHILD_BY_MAJORITY] = "majority",
+ [AFR_FAV_CHILD_POLICY_MAX] = NULL,
};
int32_t
-notify (xlator_t *this, int32_t event,
- void *data, ...)
+notify(xlator_t *this, int32_t event, void *data, ...)
{
- int ret = -1;
- va_list ap;
- void *data2 = NULL;
+ int ret = -1;
+ va_list ap;
+ void *data2 = NULL;
- va_start (ap, data);
- data2 = va_arg (ap, dict_t*);
- va_end (ap);
- ret = afr_notify (this, event, data, data2);
+ va_start(ap, data);
+ data2 = va_arg(ap, dict_t *);
+ va_end(ap);
+ ret = afr_notify(this, event, data, data2);
- return ret;
+ return ret;
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
-
- if (!this)
- return ret;
+ int ret = -1;
- ret = xlator_mem_acct_init (this, gf_afr_mt_end + 1);
+ if (!this)
+ return ret;
- if (ret != 0) {
- return ret;
- }
+ ret = xlator_mem_acct_init(this, gf_afr_mt_end + 1);
+ if (ret != 0) {
return ret;
-}
+ }
+ return ret;
+}
int
-xlator_subvolume_index (xlator_t *this, xlator_t *subvol)
+xlator_subvolume_index(xlator_t *this, xlator_t *subvol)
{
- int index = -1;
- int i = 0;
- xlator_list_t *list = NULL;
-
- list = this->children;
-
- while (list) {
- if (subvol == list->xlator ||
- strcmp (subvol->name, list->xlator->name) == 0) {
- index = i;
- break;
- }
- list = list->next;
- i++;
+ int index = -1;
+ int i = 0;
+ xlator_list_t *list = NULL;
+
+ list = this->children;
+
+ while (list) {
+ if (subvol == list->xlator ||
+ strcmp(subvol->name, list->xlator->name) == 0) {
+ index = i;
+ break;
}
+ list = list->next;
+ i++;
+ }
- return index;
+ return index;
}
static void
-fix_quorum_options (xlator_t *this, afr_private_t *priv, char *qtype,
- dict_t *options)
+fix_quorum_options(xlator_t *this, afr_private_t *priv, char *qtype,
+ dict_t *options)
{
-
- if (dict_get (options, "quorum-type") == NULL) {
- /* If user doesn't configure anything enable auto-quorum if the
- * replica has more than two subvolumes */
- if (priv->child_count > 2)
- qtype = "auto";
- }
-
- if (priv->quorum_count && strcmp (qtype, "fixed")) {
- gf_msg (this->name,GF_LOG_WARNING, 0, AFR_MSG_QUORUM_OVERRIDE,
- "quorum-type %s overriding quorum-count %u",
- qtype, priv->quorum_count);
- }
-
- if (!strcmp (qtype, "none")) {
- priv->quorum_count = 0;
- } else if (!strcmp (qtype, "auto")) {
- priv->quorum_count = AFR_QUORUM_AUTO;
- }
-
+ if (dict_get_sizen(options, "quorum-type") == NULL) {
+ /* If user doesn't configure anything enable auto-quorum if the
+ * replica has more than two subvolumes */
+ if (priv->child_count > 2)
+ qtype = "auto";
+ }
+
+ if (priv->quorum_count && strcmp(qtype, "fixed")) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_QUORUM_OVERRIDE,
+ "quorum-type %s overriding quorum-count %u", qtype,
+ priv->quorum_count);
+ }
+
+ if (!strcmp(qtype, "none")) {
+ priv->quorum_count = 0;
+ } else if (!strcmp(qtype, "auto")) {
+ priv->quorum_count = AFR_QUORUM_AUTO;
+ }
}
int
-afr_set_favorite_child_policy (afr_private_t *priv, char *policy)
+afr_set_favorite_child_policy(afr_private_t *priv, char *policy)
{
- int index = -1;
+ int index = -1;
- index = gf_get_index_by_elem (afr_favorite_child_policies, policy);
- if (index < 0 || index >= AFR_FAV_CHILD_POLICY_MAX)
- return -1;
+ index = gf_get_index_by_elem(afr_favorite_child_policies, policy);
+ if (index < 0 || index >= AFR_FAV_CHILD_POLICY_MAX)
+ return -1;
- priv->fav_child_policy = index;
+ priv->fav_child_policy = index;
- return 0;
+ return 0;
}
-int
-reconfigure (xlator_t *this, dict_t *options)
-{
- afr_private_t *priv = NULL;
- xlator_t *read_subvol = NULL;
- int read_subvol_index = -1;
- int ret = -1;
- int index = -1;
- char *qtype = NULL;
- char *fav_child_policy = NULL;
- gf_boolean_t consistent_io = _gf_false;
- gf_boolean_t choose_local_old = _gf_false;
- priv = this->private;
+static void
+set_data_self_heal_algorithm(afr_private_t *priv, char *algo)
+{
+ if (!algo) {
+ priv->data_self_heal_algorithm = AFR_SELFHEAL_DATA_DYNAMIC;
+ } else if (strcmp(algo, "full") == 0) {
+ priv->data_self_heal_algorithm = AFR_SELFHEAL_DATA_FULL;
+ } else if (strcmp(algo, "diff") == 0) {
+ priv->data_self_heal_algorithm = AFR_SELFHEAL_DATA_DIFF;
+ } else {
+ priv->data_self_heal_algorithm = AFR_SELFHEAL_DATA_DYNAMIC;
+ }
+}
- GF_OPTION_RECONF ("afr-dirty-xattr",
- priv->afr_dirty, options, str,
- out);
+void
+afr_handle_anon_inode_options(afr_private_t *priv, dict_t *options)
+{
+ char *volfile_id_str = NULL;
+ uuid_t anon_inode_gfid = {0};
+
+ /*If volume id is not present don't enable anything*/
+ if (dict_get_str(options, "volume-id", &volfile_id_str))
+ return;
+ GF_ASSERT(strlen(AFR_ANON_DIR_PREFIX) + strlen(volfile_id_str) <= NAME_MAX);
+ /*anon_inode_name is not supposed to change once assigned*/
+ if (!priv->anon_inode_name[0]) {
+ snprintf(priv->anon_inode_name, sizeof(priv->anon_inode_name), "%s-%s",
+ AFR_ANON_DIR_PREFIX, volfile_id_str);
+ gf_uuid_parse(volfile_id_str, anon_inode_gfid);
+ /*Flip a bit to make sure volfile-id and anon-gfid are not same*/
+ anon_inode_gfid[0] ^= 1;
+ uuid_utoa_r(anon_inode_gfid, priv->anon_gfid_str);
+ }
+}
- GF_OPTION_RECONF ("metadata-splitbrain-forced-heal",
- priv->metadata_splitbrain_forced_heal, options, bool,
- out);
+int
+reconfigure(xlator_t *this, dict_t *options)
+{
+ afr_private_t *priv = NULL;
+ xlator_t *read_subvol = NULL;
+ int read_subvol_index = -1;
+ int timeout_old = 0;
+ int ret = -1;
+ int index = -1;
+ char *qtype = NULL;
+ char *fav_child_policy = NULL;
+ char *data_self_heal = NULL;
+ char *data_self_heal_algorithm = NULL;
+ char *locking_scheme = NULL;
+ gf_boolean_t consistent_io = _gf_false;
+ gf_boolean_t choose_local_old = _gf_false;
+ gf_boolean_t enabled_old = _gf_false;
- GF_OPTION_RECONF ("background-self-heal-count",
- priv->background_self_heal_count, options, uint32,
- out);
+ priv = this->private;
- GF_OPTION_RECONF ("heal-wait-queue-length",
- priv->heal_wait_qlen, options, uint32, out);
+ GF_OPTION_RECONF("metadata-splitbrain-forced-heal",
+ priv->metadata_splitbrain_forced_heal, options, bool, out);
+ GF_OPTION_RECONF("background-self-heal-count",
+ priv->background_self_heal_count, options, uint32, out);
- GF_OPTION_RECONF ("metadata-self-heal",
- priv->metadata_self_heal, options, bool, out);
+ GF_OPTION_RECONF("heal-wait-queue-length", priv->heal_wait_qlen, options,
+ uint32, out);
- GF_OPTION_RECONF ("data-self-heal", priv->data_self_heal, options, str,
- out);
+ GF_OPTION_RECONF("metadata-self-heal", priv->metadata_self_heal, options,
+ bool, out);
- GF_OPTION_RECONF ("entry-self-heal", priv->entry_self_heal, options,
- bool, out);
+ GF_OPTION_RECONF("data-self-heal", data_self_heal, options, str, out);
+ if (gf_string2boolean(data_self_heal, &priv->data_self_heal) == -1)
+ goto out;
- GF_OPTION_RECONF ("data-self-heal-window-size",
- priv->data_self_heal_window_size, options,
- uint32, out);
+ GF_OPTION_RECONF("entry-self-heal", priv->entry_self_heal, options, bool,
+ out);
- GF_OPTION_RECONF ("data-self-heal-algorithm",
- priv->data_self_heal_algorithm, options, str, out);
+ GF_OPTION_RECONF("data-self-heal-window-size",
+ priv->data_self_heal_window_size, options, uint32, out);
- GF_OPTION_RECONF ("halo-enabled",
- priv->halo_enabled, options, bool,
- out);
+ GF_OPTION_RECONF("data-self-heal-algorithm", data_self_heal_algorithm,
+ options, str, out);
+ set_data_self_heal_algorithm(priv, data_self_heal_algorithm);
- GF_OPTION_RECONF ("halo-shd-max-latency",
- priv->shd.halo_max_latency_msec, options, uint32,
- out);
+ GF_OPTION_RECONF("halo-enabled", priv->halo_enabled, options, bool, out);
- GF_OPTION_RECONF ("halo-nfsd-max-latency",
- priv->nfsd.halo_max_latency_msec, options, uint32,
- out);
+ GF_OPTION_RECONF("halo-shd-max-latency", priv->shd.halo_max_latency_msec,
+ options, uint32, out);
- GF_OPTION_RECONF ("halo-max-latency", priv->halo_max_latency_msec,
- options, uint32, out);
+ GF_OPTION_RECONF("halo-nfsd-max-latency", priv->nfsd.halo_max_latency_msec,
+ options, uint32, out);
- GF_OPTION_RECONF ("halo-max-replicas", priv->halo_max_replicas, options,
- uint32, out);
+ GF_OPTION_RECONF("halo-max-latency", priv->halo_max_latency_msec, options,
+ uint32, out);
- GF_OPTION_RECONF ("halo-min-replicas", priv->halo_min_replicas, options,
- uint32, out);
+ GF_OPTION_RECONF("halo-max-replicas", priv->halo_max_replicas, options,
+ uint32, out);
- GF_OPTION_RECONF ("read-subvolume", read_subvol, options, xlator, out);
+ GF_OPTION_RECONF("halo-min-replicas", priv->halo_min_replicas, options,
+ uint32, out);
- choose_local_old = priv->choose_local;
- GF_OPTION_RECONF ("choose-local", priv->choose_local, options, bool,
- out);
+ GF_OPTION_RECONF("read-subvolume", read_subvol, options, xlator, out);
- if (choose_local_old != priv->choose_local) {
- priv->read_child = -1;
- if (choose_local_old == _gf_false)
- priv->did_discovery = _gf_false;
- }
+ choose_local_old = priv->choose_local;
+ GF_OPTION_RECONF("choose-local", priv->choose_local, options, bool, out);
- GF_OPTION_RECONF ("read-hash-mode", priv->hash_mode,
- options, uint32, out);
-
- if (read_subvol) {
- index = xlator_subvolume_index (this, read_subvol);
- if (index == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_SUBVOL, "%s not a subvolume",
- read_subvol->name);
- goto out;
- }
- priv->read_child = index;
+ if (choose_local_old != priv->choose_local) {
+ priv->read_child = -1;
+ if (choose_local_old == _gf_false)
+ priv->did_discovery = _gf_false;
+ }
+
+ GF_OPTION_RECONF("read-hash-mode", priv->hash_mode, options, uint32, out);
+
+ if (read_subvol) {
+ index = xlator_subvolume_index(this, read_subvol);
+ if (index == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_SUBVOL,
+ "%s not a subvolume", read_subvol->name);
+ goto out;
}
-
- GF_OPTION_RECONF ("read-subvolume-index",read_subvol_index, options,int32,out);
-
- if (read_subvol_index >-1) {
- index=read_subvol_index;
- if (index >= priv->child_count) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_SUBVOL,
- "%d not a subvolume-index", index);
- goto out;
- }
- priv->read_child = index;
+ priv->read_child = index;
+ }
+
+ GF_OPTION_RECONF("read-subvolume-index", read_subvol_index, options, int32,
+ out);
+
+ if (read_subvol_index > -1) {
+ index = read_subvol_index;
+ if (index >= priv->child_count) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_SUBVOL,
+ "%d not a subvolume-index", index);
+ goto out;
}
-
- GF_OPTION_RECONF ("pre-op-compat", priv->pre_op_compat, options, bool,
- out);
- GF_OPTION_RECONF ("locking-scheme", priv->locking_scheme, options, str,
- out);
- GF_OPTION_RECONF ("full-lock", priv->full_lock, options, bool, out);
- GF_OPTION_RECONF ("granular-entry-heal", priv->esh_granular, options,
- bool, out);
-
- GF_OPTION_RECONF ("eager-lock", priv->eager_lock, options, bool, out);
- GF_OPTION_RECONF ("quorum-type", qtype, options, str, out);
- GF_OPTION_RECONF ("quorum-count", priv->quorum_count, options,
- uint32, out);
- fix_quorum_options (this, priv, qtype, options);
- if (priv->quorum_count && !afr_has_quorum (priv->child_up, this))
- gf_msg (this->name, GF_LOG_WARNING, 0, AFR_MSG_QUORUM_FAIL,
- "Client-quorum is not met");
-
-
- GF_OPTION_RECONF ("post-op-delay-secs", priv->post_op_delay_secs, options,
- uint32, out);
-
- GF_OPTION_RECONF (AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size,
- options, size_uint64, out);
- /* Reset this so we re-discover in case the topology changed. */
- GF_OPTION_RECONF ("ensure-durability", priv->ensure_durability, options,
- bool, out);
-
- GF_OPTION_RECONF ("self-heal-daemon", priv->shd.enabled, options,
- bool, out);
-
- GF_OPTION_RECONF ("iam-self-heal-daemon", priv->shd.iamshd, options,
- bool, out);
-
- GF_OPTION_RECONF ("heal-timeout", priv->shd.timeout, options,
- int32, out);
-
- GF_OPTION_RECONF ("consistent-metadata", priv->consistent_metadata,
- options, bool, out);
-
- GF_OPTION_RECONF ("shd-max-threads", priv->shd.max_threads,
- options, uint32, out);
-
- GF_OPTION_RECONF ("shd-wait-qlength", priv->shd.wait_qlength,
- options, uint32, out);
-
- GF_OPTION_RECONF ("favorite-child-policy", fav_child_policy, options,
- str, out);
- if (afr_set_favorite_child_policy (priv, fav_child_policy) == -1)
- goto out;
-
- priv->did_discovery = _gf_false;
-
- GF_OPTION_RECONF ("consistent-io", consistent_io, options, bool, out);
- if (priv->quorum_count != 0)
- consistent_io = _gf_false;
- priv->consistent_io = consistent_io;
-
- ret = 0;
+ priv->read_child = index;
+ }
+
+ GF_OPTION_RECONF("pre-op-compat", priv->pre_op_compat, options, bool, out);
+ GF_OPTION_RECONF("locking-scheme", locking_scheme, options, str, out);
+ priv->granular_locks = (strcmp(locking_scheme, "granular") == 0);
+ GF_OPTION_RECONF("full-lock", priv->full_lock, options, bool, out);
+ GF_OPTION_RECONF("granular-entry-heal", priv->esh_granular, options, bool,
+ out);
+
+ GF_OPTION_RECONF("eager-lock", priv->eager_lock, options, bool, out);
+ GF_OPTION_RECONF("optimistic-change-log", priv->optimistic_change_log,
+ options, bool, out);
+ GF_OPTION_RECONF("quorum-type", qtype, options, str, out);
+ GF_OPTION_RECONF("quorum-count", priv->quorum_count, options, uint32, out);
+ fix_quorum_options(this, priv, qtype, options);
+ if (priv->quorum_count && !afr_has_quorum(priv->child_up, this, NULL))
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_QUORUM_FAIL,
+ "Client-quorum is not met");
+
+ GF_OPTION_RECONF("post-op-delay-secs", priv->post_op_delay_secs, options,
+ uint32, out);
+
+ GF_OPTION_RECONF(AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size, options,
+ size_uint64, out);
+ /* Reset this so we re-discover in case the topology changed. */
+ GF_OPTION_RECONF("ensure-durability", priv->ensure_durability, options,
+ bool, out);
+
+ enabled_old = priv->shd.enabled;
+ GF_OPTION_RECONF("self-heal-daemon", priv->shd.enabled, options, bool, out);
+
+ GF_OPTION_RECONF("iam-self-heal-daemon", priv->shd.iamshd, options, bool,
+ out);
+
+ timeout_old = priv->shd.timeout;
+ GF_OPTION_RECONF("heal-timeout", priv->shd.timeout, options, int32, out);
+
+ GF_OPTION_RECONF("consistent-metadata", priv->consistent_metadata, options,
+ bool, out);
+
+ GF_OPTION_RECONF("shd-max-threads", priv->shd.max_threads, options, uint32,
+ out);
+
+ GF_OPTION_RECONF("shd-wait-qlength", priv->shd.wait_qlength, options,
+ uint32, out);
+
+ GF_OPTION_RECONF("favorite-child-policy", fav_child_policy, options, str,
+ out);
+ if (afr_set_favorite_child_policy(priv, fav_child_policy) == -1)
+ goto out;
+
+ priv->did_discovery = _gf_false;
+
+ GF_OPTION_RECONF("consistent-io", consistent_io, options, bool, out);
+ if (priv->quorum_count != 0)
+ consistent_io = _gf_false;
+ priv->consistent_io = consistent_io;
+
+ afr_handle_anon_inode_options(priv, options);
+
+ GF_OPTION_RECONF("use-anonymous-inode", priv->use_anon_inode, options, bool,
+ out);
+ if (priv->shd.enabled) {
+ if ((priv->shd.enabled != enabled_old) ||
+ (timeout_old != priv->shd.timeout))
+ afr_selfheal_childup(this, priv);
+ }
+
+ ret = 0;
out:
- return ret;
-
+ return ret;
}
static int
-afr_pending_xattrs_init (afr_private_t *priv, xlator_t *this)
+afr_pending_xattrs_init(afr_private_t *priv, xlator_t *this)
{
- int ret = -1;
- int i = 0;
- char *ptr = NULL;
- char *ptr1 = NULL;
- char *xattrs_list = NULL;
- xlator_list_t *trav = NULL;
- int child_count = -1;
-
- trav = this->children;
- child_count = priv->child_count;
- if (priv->thin_arbiter_count) {
- /* priv->pending_key[THIN_ARBITER_BRICK_INDEX] is used as the
- * name of the thin arbiter file for persistance across add/
- * removal of DHT subvols.*/
- child_count++;
- }
-
- GF_OPTION_INIT ("afr-pending-xattr", xattrs_list, str, out);
- priv->pending_key = GF_CALLOC (sizeof (*priv->pending_key),
- child_count, gf_afr_mt_char);
- if (!priv->pending_key) {
- ret = -ENOMEM;
- goto out;
- }
- if (!xattrs_list) {
- gf_msg (this->name, GF_LOG_WARNING, 0, AFR_MSG_NO_CHANGELOG,
- "Unable to fetch afr-pending-xattr option from volfile."
- " Falling back to using client translator names. ");
-
- while (i < child_count) {
- ret = gf_asprintf (&priv->pending_key[i], "%s.%s",
- AFR_XATTR_PREFIX,
- trav->xlator->name);
- if (ret == -1) {
- ret = -ENOMEM;
- goto out;
- }
- trav = trav->next;
- i++;
- }
- ret = 0;
- goto out;
- }
+ int ret = -1;
+ int i = 0;
+ char *ptr = NULL;
+ char *ptr1 = NULL;
+ char *xattrs_list = NULL;
+ xlator_list_t *trav = NULL;
+ int child_count = -1;
+
+ trav = this->children;
+ child_count = priv->child_count;
+ if (priv->thin_arbiter_count) {
+ /* priv->pending_key[THIN_ARBITER_BRICK_INDEX] is used as the
+ * name of the thin arbiter file for persistence across add/
+ * removal of DHT subvols.*/
+ child_count++;
+ }
+
+ GF_OPTION_INIT("afr-pending-xattr", xattrs_list, str, out);
+ priv->pending_key = GF_CALLOC(sizeof(*priv->pending_key), child_count,
+ gf_afr_mt_char);
+ if (!priv->pending_key) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ if (!xattrs_list) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_NO_CHANGELOG,
+ "Unable to fetch afr-pending-xattr option from volfile."
+ " Falling back to using client translator names. ");
- ptr = ptr1 = gf_strdup (xattrs_list);
- if (!ptr) {
+ while (i < child_count) {
+ ret = gf_asprintf(&priv->pending_key[i], "%s.%s", AFR_XATTR_PREFIX,
+ trav->xlator->name);
+ if (ret == -1) {
ret = -ENOMEM;
goto out;
- }
- for (i = 0, ptr = strtok (ptr, ","); ptr; ptr = strtok (NULL, ",")) {
- ret = gf_asprintf (&priv->pending_key[i], "%s.%s",
- AFR_XATTR_PREFIX, ptr);
- if (ret == -1) {
- ret = -ENOMEM;
- goto out;
- }
- i++;
+ }
+ trav = trav->next;
+ i++;
}
ret = 0;
+ goto out;
+ }
+
+ ptr = ptr1 = gf_strdup(xattrs_list);
+ if (!ptr) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ for (i = 0, ptr = strtok(ptr, ","); ptr; ptr = strtok(NULL, ",")) {
+ ret = gf_asprintf(&priv->pending_key[i], "%s.%s", AFR_XATTR_PREFIX,
+ ptr);
+ if (ret == -1) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ i++;
+ }
+ ret = 0;
out:
- GF_FREE (ptr1);
- return ret;
+ GF_FREE(ptr1);
+ return ret;
+}
+void
+afr_ta_init(afr_private_t *priv)
+{
+ priv->thin_arbiter_count = 1;
+ priv->child_count--;
+ priv->ta_child_up = 0;
+ priv->ta_bad_child_index = AFR_CHILD_UNKNOWN;
+ priv->ta_notify_dom_lock_offset = 0;
+ priv->ta_in_mem_txn_count = 0;
+ priv->ta_on_wire_txn_count = 0;
+ priv->release_ta_notify_dom_lock = _gf_false;
+ INIT_LIST_HEAD(&priv->ta_waitq);
+ INIT_LIST_HEAD(&priv->ta_onwireq);
+ gf_uuid_clear(priv->ta_gfid);
}
int32_t
-init (xlator_t *this)
+init(xlator_t *this)
{
- afr_private_t *priv = NULL;
- int child_count = 0;
- xlator_list_t *trav = NULL;
- int i = 0;
- int ret = -1;
- GF_UNUSED int op_errno = 0;
- xlator_t *read_subvol = NULL;
- int read_subvol_index = -1;
- char *qtype = NULL;
- char *fav_child_policy = NULL;
- char *thin_arbiter = NULL;
-
- if (!this->children) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_CHILD_MISCONFIGURED,
- "replicate translator needs more than one "
- "subvolume defined.");
- return -1;
+ afr_private_t *priv = NULL;
+ int child_count = 0;
+ xlator_list_t *trav = NULL;
+ int i = 0;
+ int ret = -1;
+ GF_UNUSED int op_errno = 0;
+ xlator_t *read_subvol = NULL;
+ int read_subvol_index = -1;
+ char *qtype = NULL;
+ char *fav_child_policy = NULL;
+ char *thin_arbiter = NULL;
+ char *data_self_heal = NULL;
+ char *locking_scheme = NULL;
+ char *data_self_heal_algorithm = NULL;
+
+ if (!this->children) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_CHILD_MISCONFIGURED,
+ "replicate translator needs more than one "
+ "subvolume defined.");
+ return -1;
+ }
+
+ if (!this->parents) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_VOL_MISCONFIGURED,
+ "Volume is dangling.");
+ }
+
+ this->private = GF_CALLOC(1, sizeof(afr_private_t),
+ gf_afr_mt_afr_private_t);
+ if (!this->private)
+ goto out;
+
+ priv = this->private;
+ INIT_LIST_HEAD(&priv->saved_locks);
+ INIT_LIST_HEAD(&priv->lk_healq);
+ LOCK_INIT(&priv->lock);
+
+ child_count = xlator_subvolume_count(this);
+
+ priv->child_count = child_count;
+
+ priv->read_child = -1;
+
+ GF_OPTION_INIT("arbiter-count", priv->arbiter_count, uint32, out);
+ GF_OPTION_INIT("thin-arbiter", thin_arbiter, str, out);
+ if (thin_arbiter && strlen(thin_arbiter) > 0) {
+ afr_ta_init(priv);
+ }
+ INIT_LIST_HEAD(&priv->healing);
+ INIT_LIST_HEAD(&priv->heal_waiting);
+
+ priv->spb_choice_timeout = AFR_DEFAULT_SPB_CHOICE_TIMEOUT;
+
+ GF_OPTION_INIT("afr-dirty-xattr", priv->afr_dirty, str, out);
+
+ GF_OPTION_INIT("metadata-splitbrain-forced-heal",
+ priv->metadata_splitbrain_forced_heal, bool, out);
+
+ GF_OPTION_INIT("read-subvolume", read_subvol, xlator, out);
+ if (read_subvol) {
+ priv->read_child = xlator_subvolume_index(this, read_subvol);
+ if (priv->read_child == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_SUBVOL,
+ "%s not a subvolume", read_subvol->name);
+ goto out;
}
-
- if (!this->parents) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- AFR_MSG_VOL_MISCONFIGURED, "Volume is dangling.");
+ }
+ GF_OPTION_INIT("read-subvolume-index", read_subvol_index, int32, out);
+ if (read_subvol_index > -1) {
+ if (read_subvol_index >= priv->child_count) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_SUBVOL,
+ "%d not a subvolume-index", read_subvol_index);
+ goto out;
}
+ priv->read_child = read_subvol_index;
+ }
+ GF_OPTION_INIT("choose-local", priv->choose_local, bool, out);
- this->private = GF_CALLOC (1, sizeof (afr_private_t),
- gf_afr_mt_afr_private_t);
- if (!this->private)
- goto out;
-
- priv = this->private;
- LOCK_INIT (&priv->lock);
-
- child_count = xlator_subvolume_count (this);
-
- priv->child_count = child_count;
-
- priv->read_child = -1;
-
- GF_OPTION_INIT ("arbiter-count", priv->arbiter_count, uint32, out);
- GF_OPTION_INIT ("thin-arbiter", thin_arbiter, str, out);
- if (thin_arbiter && strlen(thin_arbiter) > 0) {
- priv->thin_arbiter_count = 1;
- priv->child_count--;
- }
- INIT_LIST_HEAD (&priv->healing);
- INIT_LIST_HEAD (&priv->heal_waiting);
-
- priv->spb_choice_timeout = AFR_DEFAULT_SPB_CHOICE_TIMEOUT;
-
- GF_OPTION_INIT ("afr-dirty-xattr", priv->afr_dirty, str, out);
-
- GF_OPTION_INIT ("metadata-splitbrain-forced-heal",
- priv->metadata_splitbrain_forced_heal, bool, out);
-
- GF_OPTION_INIT ("read-subvolume", read_subvol, xlator, out);
- if (read_subvol) {
- priv->read_child = xlator_subvolume_index (this, read_subvol);
- if (priv->read_child == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_SUBVOL, "%s not a subvolume",
- read_subvol->name);
- goto out;
- }
- }
- GF_OPTION_INIT ("read-subvolume-index",read_subvol_index,int32,out);
- if (read_subvol_index > -1) {
- if (read_subvol_index >= priv->child_count) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- AFR_MSG_INVALID_SUBVOL,
- "%d not a subvolume-index", read_subvol_index);
- goto out;
- }
- priv->read_child = read_subvol_index;
- }
- GF_OPTION_INIT ("choose-local", priv->choose_local, bool, out);
+ priv->pending_reads = GF_CALLOC(sizeof(*priv->pending_reads),
+ priv->child_count, gf_afr_mt_atomic_t);
- priv->pending_reads = GF_CALLOC (sizeof(*priv->pending_reads),
- priv->child_count, gf_afr_mt_atomic_t);
+ GF_OPTION_INIT("read-hash-mode", priv->hash_mode, uint32, out);
- GF_OPTION_INIT ("read-hash-mode", priv->hash_mode, uint32, out);
+ priv->favorite_child = -1;
- priv->favorite_child = -1;
+ GF_OPTION_INIT("favorite-child-policy", fav_child_policy, str, out);
+ if (afr_set_favorite_child_policy(priv, fav_child_policy) == -1)
+ goto out;
- GF_OPTION_INIT ("favorite-child-policy", fav_child_policy, str, out);
- if (afr_set_favorite_child_policy(priv, fav_child_policy) == -1)
- goto out;
+ GF_OPTION_INIT("shd-max-threads", priv->shd.max_threads, uint32, out);
- GF_OPTION_INIT ("shd-max-threads", priv->shd.max_threads,
- uint32, out);
+ GF_OPTION_INIT("shd-wait-qlength", priv->shd.wait_qlength, uint32, out);
- GF_OPTION_INIT ("shd-wait-qlength", priv->shd.wait_qlength,
- uint32, out);
+ GF_OPTION_INIT("background-self-heal-count",
+ priv->background_self_heal_count, uint32, out);
- GF_OPTION_INIT ("background-self-heal-count",
- priv->background_self_heal_count, uint32, out);
+ GF_OPTION_INIT("heal-wait-queue-length", priv->heal_wait_qlen, uint32, out);
- GF_OPTION_INIT ("heal-wait-queue-length",
- priv->heal_wait_qlen, uint32, out);
+ GF_OPTION_INIT("data-self-heal", data_self_heal, str, out);
+ if (gf_string2boolean(data_self_heal, &priv->data_self_heal) == -1)
+ goto out;
- GF_OPTION_INIT ("data-self-heal", priv->data_self_heal, str, out);
+ GF_OPTION_INIT("data-self-heal-algorithm", data_self_heal_algorithm, str,
+ out);
+ set_data_self_heal_algorithm(priv, data_self_heal_algorithm);
- GF_OPTION_INIT ("data-self-heal-algorithm",
- priv->data_self_heal_algorithm, str, out);
+ GF_OPTION_INIT("data-self-heal-window-size",
+ priv->data_self_heal_window_size, uint32, out);
- GF_OPTION_INIT ("data-self-heal-window-size",
- priv->data_self_heal_window_size, uint32, out);
+ GF_OPTION_INIT("metadata-self-heal", priv->metadata_self_heal, bool, out);
- GF_OPTION_INIT ("metadata-self-heal", priv->metadata_self_heal, bool,
- out);
+ GF_OPTION_INIT("entry-self-heal", priv->entry_self_heal, bool, out);
- GF_OPTION_INIT ("entry-self-heal", priv->entry_self_heal, bool, out);
+ GF_OPTION_INIT("halo-shd-max-latency", priv->shd.halo_max_latency_msec,
+ uint32, out);
- GF_OPTION_INIT ("halo-shd-max-latency", priv->shd.halo_max_latency_msec,
- uint32, out);
+ GF_OPTION_INIT("halo-max-latency", priv->halo_max_latency_msec, uint32,
+ out);
+ GF_OPTION_INIT("halo-max-replicas", priv->halo_max_replicas, uint32, out);
+ GF_OPTION_INIT("halo-min-replicas", priv->halo_min_replicas, uint32, out);
- GF_OPTION_INIT ("halo-max-latency", priv->halo_max_latency_msec,
- uint32, out);
- GF_OPTION_INIT ("halo-max-replicas", priv->halo_max_replicas, uint32,
- out);
- GF_OPTION_INIT ("halo-min-replicas", priv->halo_min_replicas, uint32,
- out);
+ GF_OPTION_INIT("halo-enabled", priv->halo_enabled, bool, out);
- GF_OPTION_INIT ("halo-enabled",
- priv->halo_enabled, bool, out);
+ GF_OPTION_INIT("halo-nfsd-max-latency", priv->nfsd.halo_max_latency_msec,
+ uint32, out);
- GF_OPTION_INIT ("halo-nfsd-max-latency",
- priv->nfsd.halo_max_latency_msec, uint32, out);
+ GF_OPTION_INIT("iam-nfs-daemon", priv->nfsd.iamnfsd, bool, out);
- GF_OPTION_INIT ("iam-nfs-daemon", priv->nfsd.iamnfsd, bool, out);
+ GF_OPTION_INIT("optimistic-change-log", priv->optimistic_change_log, bool,
+ out);
- GF_OPTION_INIT ("optimistic-change-log", priv->optimistic_change_log,
- bool, out);
+ GF_OPTION_INIT("pre-op-compat", priv->pre_op_compat, bool, out);
+ GF_OPTION_INIT("locking-scheme", locking_scheme, str, out);
+ priv->granular_locks = (strcmp(locking_scheme, "granular") == 0);
+ GF_OPTION_INIT("full-lock", priv->full_lock, bool, out);
+ GF_OPTION_INIT("granular-entry-heal", priv->esh_granular, bool, out);
- GF_OPTION_INIT ("pre-op-compat", priv->pre_op_compat, bool, out);
- GF_OPTION_INIT ("locking-scheme", priv->locking_scheme, str, out);
- GF_OPTION_INIT ("full-lock", priv->full_lock, bool, out);
- GF_OPTION_INIT ("granular-entry-heal", priv->esh_granular, bool, out);
+ GF_OPTION_INIT("eager-lock", priv->eager_lock, bool, out);
+ GF_OPTION_INIT("quorum-type", qtype, str, out);
+ GF_OPTION_INIT("quorum-count", priv->quorum_count, uint32, out);
+ GF_OPTION_INIT(AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size, size_uint64,
+ out);
+ fix_quorum_options(this, priv, qtype, this->options);
- GF_OPTION_INIT ("eager-lock", priv->eager_lock, bool, out);
- GF_OPTION_INIT ("quorum-type", qtype, str, out);
- GF_OPTION_INIT ("quorum-count", priv->quorum_count, uint32, out);
- GF_OPTION_INIT (AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size, size_uint64,
- out);
- fix_quorum_options (this, priv, qtype, this->options);
+ GF_OPTION_INIT("post-op-delay-secs", priv->post_op_delay_secs, uint32, out);
+ GF_OPTION_INIT("ensure-durability", priv->ensure_durability, bool, out);
- GF_OPTION_INIT ("post-op-delay-secs", priv->post_op_delay_secs, uint32, out);
- GF_OPTION_INIT ("ensure-durability", priv->ensure_durability, bool,
- out);
+ GF_OPTION_INIT("self-heal-daemon", priv->shd.enabled, bool, out);
- GF_OPTION_INIT ("self-heal-daemon", priv->shd.enabled, bool, out);
+ GF_OPTION_INIT("iam-self-heal-daemon", priv->shd.iamshd, bool, out);
+ GF_OPTION_INIT("heal-timeout", priv->shd.timeout, int32, out);
- GF_OPTION_INIT ("iam-self-heal-daemon", priv->shd.iamshd, bool, out);
- GF_OPTION_INIT ("heal-timeout", priv->shd.timeout, int32, out);
+ GF_OPTION_INIT("consistent-metadata", priv->consistent_metadata, bool, out);
+ GF_OPTION_INIT("consistent-io", priv->consistent_io, bool, out);
+ afr_handle_anon_inode_options(priv, this->options);
- GF_OPTION_INIT ("consistent-metadata", priv->consistent_metadata, bool,
- out);
- GF_OPTION_INIT ("consistent-io", priv->consistent_io, bool, out);
+ GF_OPTION_INIT("use-anonymous-inode", priv->use_anon_inode, bool, out);
+ if (priv->quorum_count != 0)
+ priv->consistent_io = _gf_false;
- if (priv->quorum_count != 0)
- priv->consistent_io = _gf_false;
+ priv->wait_count = 1;
- priv->wait_count = 1;
+ priv->local = GF_CALLOC(sizeof(unsigned char), child_count, gf_afr_mt_char);
+ if (!priv->local) {
+ ret = -ENOMEM;
+ goto out;
+ }
- priv->local = GF_CALLOC (sizeof (unsigned char), child_count,
+ priv->anon_inode = GF_CALLOC(sizeof(unsigned char), child_count,
gf_afr_mt_char);
- if (!priv->local) {
- ret = -ENOMEM;
- goto out;
- }
- priv->child_up = GF_CALLOC (sizeof (unsigned char), child_count,
- gf_afr_mt_char);
+ priv->child_up = GF_CALLOC(sizeof(unsigned char), child_count,
+ gf_afr_mt_char);
- priv->child_latency = GF_CALLOC (sizeof (*priv->child_latency),
- child_count,
- gf_afr_mt_child_latency_t);
+ priv->child_latency = GF_MALLOC(sizeof(*priv->child_latency) * child_count,
+ gf_afr_mt_child_latency_t);
+ priv->halo_child_up = GF_CALLOC(sizeof(unsigned char), child_count,
+ gf_afr_mt_char);
- if (!priv->child_up || !priv->child_latency) {
- ret = -ENOMEM;
- goto out;
- }
- /*Initialize to -ve ping timeout so that they are not considered
- * in child-up events until ping-event comes*/
- for (i = 0; i < child_count; i++)
- priv->child_latency[i] = -1;
-
- priv->children = GF_CALLOC (sizeof (xlator_t *), child_count,
- gf_afr_mt_xlator_t);
- if (!priv->children) {
- ret = -ENOMEM;
- goto out;
+ if (!priv->child_up || !priv->child_latency || !priv->halo_child_up ||
+ !priv->anon_inode) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ /*Initialize to -ve ping timeout so that they are not considered
+ * in child-up events until ping-event comes*/
+ for (i = 0; i < child_count; i++)
+ priv->child_latency[i] = -1;
+
+ priv->children = GF_CALLOC(sizeof(xlator_t *), child_count,
+ gf_afr_mt_xlator_t);
+ if (!priv->children) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ ret = afr_pending_xattrs_init(priv, this);
+ if (ret)
+ goto out;
+
+ trav = this->children;
+ i = 0;
+ while (i < child_count) {
+ priv->children[i] = trav->xlator;
+ trav = trav->next;
+ i++;
+ }
+
+ ret = gf_asprintf(&priv->sh_domain, AFR_SH_DATA_DOMAIN_FMT, this->name);
+ if (-1 == ret) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ priv->last_event = GF_CALLOC(child_count, sizeof(*priv->last_event),
+ gf_afr_mt_int32_t);
+ if (!priv->last_event) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ this->itable = inode_table_new(SHD_INODE_LRU_LIMIT, this);
+ if (!this->itable) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ if (priv->shd.iamshd) {
+ ret = afr_selfheal_daemon_init(this);
+ if (ret) {
+ ret = -ENOMEM;
+ goto out;
}
+ }
- ret = afr_pending_xattrs_init (priv, this);
- if (ret)
- goto out;
+ /* keep more local here as we may need them for self-heal etc */
+ this->local_pool = mem_pool_new(afr_local_t, 512);
+ if (!this->local_pool) {
+ ret = -1;
+ goto out;
+ }
- trav = this->children;
- i = 0;
- while (i < child_count) {
- priv->children[i] = trav->xlator;
- trav = trav->next;
- i++;
- }
+ priv->root_inode = NULL;
- ret = gf_asprintf (&priv->sh_domain, AFR_SH_DATA_DOMAIN_FMT,
- this->name);
- if (-1 == ret) {
- ret = -ENOMEM;
- goto out;
- }
+ ret = 0;
+out:
+ return ret;
+}
+void
+afr_destroy_healer_object(xlator_t *this, struct subvol_healer *healer)
+{
+ int ret = -1;
- priv->last_event = GF_CALLOC (child_count, sizeof (*priv->last_event),
- gf_afr_mt_int32_t);
- if (!priv->last_event) {
- ret = -ENOMEM;
- goto out;
- }
+ if (!healer)
+ return;
- ret = afr_selfheal_daemon_init (this);
- if (ret) {
- ret = -ENOMEM;
- goto out;
- }
+ if (healer->running) {
+ /*
+ * If there are any resources to cleanup, We need
+ * to do that gracefully using pthread_cleanup_push
+ */
+ ret = gf_thread_cleanup_xint(healer->thread);
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_SELF_HEAL_FAILED,
+ "Failed to clean up healer threads.");
+ healer->thread = 0;
+ }
+ pthread_cond_destroy(&healer->cond);
+ pthread_mutex_destroy(&healer->mutex);
+}
- /* keep more local here as we may need them for self-heal etc */
- this->local_pool = mem_pool_new (afr_local_t, 512);
- if (!this->local_pool) {
- ret = -1;
- goto out;
- }
+void
+afr_selfheal_daemon_fini(xlator_t *this)
+{
+ struct subvol_healer *healer = NULL;
+ afr_self_heald_t *shd = NULL;
+ afr_private_t *priv = NULL;
+ int i = 0;
+
+ priv = this->private;
+ if (!priv)
+ return;
+
+ shd = &priv->shd;
+ if (!shd->iamshd)
+ return;
+
+ for (i = 0; i < priv->child_count; i++) {
+ healer = &shd->index_healers[i];
+ afr_destroy_healer_object(this, healer);
+
+ healer = &shd->full_healers[i];
+ afr_destroy_healer_object(this, healer);
+
+ if (shd->statistics[i])
+ eh_destroy(shd->statistics[i]);
+ }
+ GF_FREE(shd->index_healers);
+ GF_FREE(shd->full_healers);
+ GF_FREE(shd->statistics);
+ if (shd->split_brain)
+ eh_destroy(shd->split_brain);
+}
+void
+fini(xlator_t *this)
+{
+ afr_private_t *priv = NULL;
- priv->root_inode = NULL;
+ priv = this->private;
- ret = 0;
-out:
- return ret;
-}
+ afr_selfheal_daemon_fini(this);
+ GF_ASSERT(list_empty(&priv->saved_locks));
+ LOCK(&priv->lock);
+ if (priv->timer != NULL) {
+ gf_timer_call_cancel(this->ctx, priv->timer);
+ priv->timer = NULL;
+ }
+ UNLOCK(&priv->lock);
-int
-fini (xlator_t *this)
-{
- afr_private_t *priv = NULL;
+ if (this->local_pool != NULL) {
+ mem_pool_destroy(this->local_pool);
+ this->local_pool = NULL;
+ }
- priv = this->private;
- LOCK (&priv->lock);
- if (priv->timer != NULL) {
- gf_timer_call_cancel(this->ctx, priv->timer);
- priv->timer = NULL;
- }
- UNLOCK (&priv->lock);
- this->private = NULL;
- afr_priv_destroy (priv);
- //if (this->itable);//I dont see any destroy func
+ this->private = NULL;
+ afr_priv_destroy(priv);
+ if (this->itable) {
+ inode_table_destroy(this->itable);
+ this->itable = NULL;
+ }
- return 0;
+ return;
}
-
struct xlator_fops fops = {
- .lookup = afr_lookup,
- .lk = afr_lk,
- .flush = afr_flush,
- .statfs = afr_statfs,
- .fsyncdir = afr_fsyncdir,
- .inodelk = afr_inodelk,
- .finodelk = afr_finodelk,
- .entrylk = afr_entrylk,
- .fentrylk = afr_fentrylk,
- .ipc = afr_ipc,
- .lease = afr_lease,
-
- /* inode read */
- .access = afr_access,
- .stat = afr_stat,
- .fstat = afr_fstat,
- .readlink = afr_readlink,
- .getxattr = afr_getxattr,
- .fgetxattr = afr_fgetxattr,
- .readv = afr_readv,
-
- /* inode write */
- .writev = afr_writev,
- .truncate = afr_truncate,
- .ftruncate = afr_ftruncate,
- .setxattr = afr_setxattr,
- .fsetxattr = afr_fsetxattr,
- .setattr = afr_setattr,
- .fsetattr = afr_fsetattr,
- .removexattr = afr_removexattr,
- .fremovexattr = afr_fremovexattr,
- .fallocate = afr_fallocate,
- .discard = afr_discard,
- .zerofill = afr_zerofill,
- .xattrop = afr_xattrop,
- .fxattrop = afr_fxattrop,
- .fsync = afr_fsync,
-
- /*inode open*/
- .opendir = afr_opendir,
- .open = afr_open,
-
- /* dir read */
- .readdir = afr_readdir,
- .readdirp = afr_readdirp,
-
- /* dir write */
- .create = afr_create,
- .mknod = afr_mknod,
- .mkdir = afr_mkdir,
- .unlink = afr_unlink,
- .rmdir = afr_rmdir,
- .link = afr_link,
- .symlink = afr_symlink,
- .rename = afr_rename,
+ .lookup = afr_lookup,
+ .lk = afr_lk,
+ .flush = afr_flush,
+ .statfs = afr_statfs,
+ .fsyncdir = afr_fsyncdir,
+ .inodelk = afr_inodelk,
+ .finodelk = afr_finodelk,
+ .entrylk = afr_entrylk,
+ .fentrylk = afr_fentrylk,
+ .ipc = afr_ipc,
+ .lease = afr_lease,
+
+ /* inode read */
+ .access = afr_access,
+ .stat = afr_stat,
+ .fstat = afr_fstat,
+ .readlink = afr_readlink,
+ .getxattr = afr_getxattr,
+ .fgetxattr = afr_fgetxattr,
+ .readv = afr_readv,
+ .seek = afr_seek,
+
+ /* inode write */
+ .writev = afr_writev,
+ .truncate = afr_truncate,
+ .ftruncate = afr_ftruncate,
+ .setxattr = afr_setxattr,
+ .fsetxattr = afr_fsetxattr,
+ .setattr = afr_setattr,
+ .fsetattr = afr_fsetattr,
+ .removexattr = afr_removexattr,
+ .fremovexattr = afr_fremovexattr,
+ .fallocate = afr_fallocate,
+ .discard = afr_discard,
+ .zerofill = afr_zerofill,
+ .xattrop = afr_xattrop,
+ .fxattrop = afr_fxattrop,
+ .fsync = afr_fsync,
+
+ /*inode open*/
+ .opendir = afr_opendir,
+ .open = afr_open,
+
+ /* dir read */
+ .readdir = afr_readdir,
+ .readdirp = afr_readdirp,
+
+ /* dir write */
+ .create = afr_create,
+ .mknod = afr_mknod,
+ .mkdir = afr_mkdir,
+ .unlink = afr_unlink,
+ .rmdir = afr_rmdir,
+ .link = afr_link,
+ .symlink = afr_symlink,
+ .rename = afr_rename,
};
-
struct xlator_dumpops dumpops = {
- .priv = afr_priv_dump,
+ .priv = afr_priv_dump,
};
-
struct xlator_cbks cbks = {
- .release = afr_release,
- .releasedir = afr_releasedir,
- .forget = afr_forget,
+ .release = afr_release,
+ .releasedir = afr_releasedir,
+ .forget = afr_forget,
};
-
struct volume_options options[] = {
- { .key = {"read-subvolume" },
- .type = GF_OPTION_TYPE_XLATOR,
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "inode-read fops happen only on one of the bricks in "
- "replicate. Afr will prefer the one specified using "
- "this option if it is not stale. Option value must be "
- "one of the xlator names of the children. "
- "Ex: <volname>-client-0 till "
- "<volname>-client-<number-of-bricks - 1>"
- },
- { .key = {"read-subvolume-index" },
- .type = GF_OPTION_TYPE_INT,
- .default_value = "-1",
- .op_version = {2},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "inode-read fops happen only on one of the bricks in "
- "replicate. AFR will prefer the one specified using "
- "this option if it is not stale. allowed options"
- " include -1 till replica-count - 1"
- },
- { .key = {"read-hash-mode" },
- .type = GF_OPTION_TYPE_INT,
- .min = 0,
- .max = 3,
- .default_value = "1",
- .op_version = {2},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "inode-read fops happen only on one of the bricks in "
- "replicate. AFR will prefer the one computed using "
- "the method specified using this option.\n"
- "0 = first readable child of AFR, starting from 1st child.\n"
- "1 = hash by GFID of file (all clients use "
- "same subvolume).\n"
- "2 = hash by GFID of file and client PID.\n"
- "3 = brick having the least outstanding read requests."
- },
- { .key = {"choose-local" },
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "true",
- .op_version = {2},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "Choose a local subvolume (i.e. Brick) to read from"
- " if read-subvolume is not explicitly set.",
- },
- { .key = {"background-self-heal-count"},
- .type = GF_OPTION_TYPE_INT,
- .min = 0,
- .max = 256,
- .default_value = "8",
- .validate = GF_OPT_VALIDATE_MIN,
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This specifies the number of per client self-heal "
- "jobs that can perform parallel heals in the "
- "background."
- },
- { .key = {"halo-shd-max-latency"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 99999,
- .default_value = "99999",
- .op_version = {GD_OP_VERSION_3_11_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate", "halo"},
- .description = "Maximum latency for shd halo replication in msec."
- },
- { .key = {"halo-enabled"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "False",
- .op_version = {GD_OP_VERSION_3_11_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate", "halo"},
- .description = "Enable Halo (geo) replication mode."
- },
- { .key = {"halo-nfsd-max-latency"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 99999,
- .default_value = "5",
- .op_version = {GD_OP_VERSION_3_11_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate", "halo"},
- .description = "Maximum latency for nfsd halo replication in msec."
- },
- { .key = {"halo-max-latency"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = AFR_HALO_MAX_LATENCY,
- .default_value = "5",
- .op_version = {GD_OP_VERSION_3_11_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate", "halo"},
- .description = "Maximum latency for halo replication in msec."
- },
- { .key = {"halo-max-replicas"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 99999,
- .default_value = "99999",
- .op_version = {GD_OP_VERSION_3_11_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate", "halo"},
- .description = "The maximum number of halo replicas; replicas"
- " beyond this value will be written asynchronously"
- "via the SHD."
- },
- { .key = {"halo-min-replicas"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 99999,
- .default_value = "2",
- .op_version = {GD_OP_VERSION_3_11_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate", "halo"},
- .description = "The minimmum number of halo replicas, before adding "
- "out of region replicas."
- },
- { .key = {"heal-wait-queue-length"},
- .type = GF_OPTION_TYPE_INT,
- .min = 0,
- .max = 10000, /*Around 100MB with sizeof(afr_local_t)= 10496 bytes*/
- .default_value = "128",
- .validate = GF_OPT_VALIDATE_MIN,
- .op_version = {GD_OP_VERSION_3_7_10},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This specifies the number of heals that can be queued"
- " for the parallel background self heal jobs."
- },
- { .key = {"data-self-heal"},
- .type = GF_OPTION_TYPE_STR,
- .value = {"1", "on", "yes", "true", "enable",
- "0", "off", "no", "false", "disable",
- "open"},
- .default_value = "on",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "Using this option we can enable/disable data "
- "self-heal on the file. \"open\" means data "
- "self-heal action will only be triggered by file "
- "open operations."
- },
- { .key = {"data-self-heal-algorithm"},
- .type = GF_OPTION_TYPE_STR,
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "Select between \"full\", \"diff\". The "
- "\"full\" algorithm copies the entire file from "
- "source to sink. The \"diff\" algorithm copies to "
- "sink only those blocks whose checksums don't match "
- "with those of source. If no option is configured "
- "the option is chosen dynamically as follows: "
- "If the file does not exist on one of the sinks "
- "or empty file exists or if the source file size is "
- "about the same as page size the entire file will "
- "be read and written i.e \"full\" algo, "
- "otherwise \"diff\" algo is chosen.",
- .value = { "diff", "full"}
- },
- { .key = {"data-self-heal-window-size"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 1024,
- .default_value = "1",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "Maximum number blocks per file for which self-heal "
- "process would be applied simultaneously."
- },
- { .key = {"metadata-self-heal"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- /*.validate_fn = validate_replica*/
- .description = "Using this option we can enable/disable metadata "
- "i.e. Permissions, ownerships, xattrs self-heal on "
- "the file/directory."
- },
- { .key = {"entry-self-heal"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- /*.validate_fn = validate_replica*/
- .description = "Using this option we can enable/disable entry "
- "self-heal on the directory."
- },
- { .key = {"data-change-log"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This option exists only for backward compatibility "
- "and configuring it doesn't have any effect"
- },
- { .key = {"metadata-change-log"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This option exists only for backward compatibility "
- "and configuring it doesn't have any effect"
- },
- { .key = {"entry-change-log"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This option exists only for backward compatibility "
- "and configuring it doesn't have any effect"
- },
- { .key = {"optimistic-change-log"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .description = "Entry/Metadata fops will not perform "
- "pre fop changelog operations in afr transaction "
- "if this option is enabled."
- },
- { .key = {"inodelk-trace"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "Enabling this option logs inode lock/unlocks"
- },
- { .key = {"entrylk-trace"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "Enabling this option logs entry lock/unlocks"
- },
- { .key = {"pre-op-compat"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .description = "Use separate pre-op xattrop() FOP rather than "
- "overloading xdata of the OP"
- },
- { .key = {"eager-lock"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "Enable/Disable eager lock for replica volume. "
- "Lock phase of a transaction has two sub-phases. "
- "First is an attempt to acquire locks in parallel by "
- "broadcasting non-blocking lock requests. If lock "
- "acquisition fails on any server, then the held locks "
- "are unlocked and we revert to a blocking locks mode "
- "sequentially on one server after another. If this "
- "option is enabled the initial broadcasting lock "
- "request attempts to acquire a full lock on the entire file. "
- "If this fails, we revert back to the sequential "
- "\"regional\" blocking locks as before. In the case "
- "where such an \"eager\" lock is granted in the "
- "non-blocking phase, it gives rise to an opportunity "
- "for optimization. i.e, if the next write transaction "
- "on the same FD arrives before the unlock phase of "
- "the first transaction, it \"takes over\" the full "
- "file lock. Similarly if yet another data transaction "
- "arrives before the unlock phase of the \"optimized\" "
- "transaction, that in turn \"takes over\" the lock as "
- "well. The actual unlock now happens at the end of "
- "the last \"optimized\" transaction."
-
- },
- { .key = {"self-heal-daemon"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
- .tags = {"replicate"},
- /*.validate_fn = validate_replica_heal_enable_disable*/
- .description = "This option applies to only self-heal-daemon. "
- "Index directory crawl and automatic healing of files "
- "will not be performed if this option is turned off."
- },
- { .key = {"iam-self-heal-daemon"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "This option differentiates if the replicate "
- "translator is running as part of self-heal-daemon "
- "or not."
- },
- { .key = {"iam-nfs-daemon"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "This option differentiates if the replicate "
- "translator is running as part of an NFS daemon "
- "or not."
- },
- { .key = {"quorum-type"},
- .type = GF_OPTION_TYPE_STR,
- .value = { "none", "auto", "fixed"},
- .default_value = "none",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- /*.option = quorum-type*/
- .description = "If value is \"fixed\" only allow writes if "
- "quorum-count bricks are present. If value is "
- "\"auto\" only allow writes if more than half of "
- "bricks, or exactly half including the first, are "
- "present.",
- },
- { .key = {"quorum-count"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = INT_MAX,
- .default_value = 0,
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- /*.option = quorum-count*/
- /*.validate_fn = validate_quorum_count*/
- .description = "If quorum-type is \"fixed\" only allow writes if "
- "this many bricks are present. Other quorum types "
- "will OVERWRITE this value.",
- },
- { .key = {"quorum-reads"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "no",
- .op_version = {GD_OP_VERSION_3_7_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This option has been removed. Reads are not allowed "
- "if quorum is not met.",
- },
- { .key = {"node-uuid"},
- .type = GF_OPTION_TYPE_STR,
- .description = "Local glusterd uuid string, used in starting "
- "self-heal-daemon so that it can crawl only on "
- "local index directories.",
- },
- { .key = {"post-op-delay-secs"},
- .type = GF_OPTION_TYPE_INT,
- .min = 0,
- .max = INT_MAX,
- .default_value = "1",
- .op_version = {2},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "Time interval induced artificially before "
- "post-operation phase of the transaction to "
- "enhance overlap of adjacent write operations.",
- },
- { .key = {AFR_SH_READDIR_SIZE_KEY},
- .type = GF_OPTION_TYPE_SIZET,
- .description = "readdirp size for performing entry self-heal",
- .min = 1024,
- .max = 131072,
- .op_version = {2},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
- .tags = {"replicate"},
- .default_value = "1KB",
- },
- { .key = {"ensure-durability"},
- .type = GF_OPTION_TYPE_BOOL,
- .op_version = {3},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "Afr performs fsyncs for transactions if this "
- "option is on to make sure the changelogs/data is "
- "written to the disk",
- .default_value = "on",
- },
- { .key = {"afr-dirty-xattr"},
- .type = GF_OPTION_TYPE_STR,
- .default_value = AFR_DIRTY_DEFAULT,
- },
- { .key = {"afr-pending-xattr"},
- .type = GF_OPTION_TYPE_STR,
- .description = "Comma separated list of xattrs that are used to "
- "capture information on pending heals."
- },
- { .key = {"metadata-splitbrain-forced-heal"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- },
- { .key = {"heal-timeout"},
- .type = GF_OPTION_TYPE_INT,
- .min = 5,
- .max = INT_MAX,
- .default_value = "600",
- .op_version = {2},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "time interval for checking the need to self-heal "
- "in self-heal-daemon"
- },
- { .key = {"consistent-metadata"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "no",
- .op_version = {GD_OP_VERSION_3_7_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "If this option is enabled, readdirp will force "
- "lookups on those entries read whose read child is "
- "not the same as that of the parent. This will "
- "guarantee that all read operations on a file serve "
- "attributes from the same subvol as long as it holds "
- " a good copy of the file/dir.",
- },
- { .key = {"arbiter-count"},
- .type = GF_OPTION_TYPE_INT,
- .description = "subset of child_count. Has to be 0 or 1."
- },
- { .key = {"thin-arbiter"},
- .type = GF_OPTION_TYPE_STR,
- .op_version = {GD_OP_VERSION_4_1_0},
- .flags = OPT_FLAG_SETTABLE,
- .tags = {"replicate"},
- .description = "contains host:path of thin abriter brick",
- },
- { .key = {"shd-max-threads"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 64,
- .default_value = "1",
- .op_version = {GD_OP_VERSION_3_7_12},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "Maximum number of parallel heals SHD can do per "
- "local brick. This can substantially lower heal times"
- ", but can also crush your bricks if you don't have "
- "the storage hardware to support this."
- },
- { .key = {"shd-wait-qlength"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 655536,
- .default_value = "1024",
- .op_version = {GD_OP_VERSION_3_7_12},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This option can be used to control number of heals"
- " that can wait in SHD per subvolume",
- },
- { .key = {"locking-scheme"},
- .type = GF_OPTION_TYPE_STR,
- .value = { "full", "granular"},
- .default_value = "full",
- .op_version = {GD_OP_VERSION_3_7_12},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "If this option is set to granular, self-heal will "
- "stop being compatible with afr-v1, which helps afr "
- "be more granular while self-healing",
- },
- { .key = {"full-lock"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "yes",
- .op_version = {GD_OP_VERSION_3_13_2},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
- .tags = {"replicate"},
- .description = "If this option is disabled, then the IOs will take "
- "range locks same as versions till 3.13.1."
- },
- { .key = {"granular-entry-heal"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "no",
- .op_version = {GD_OP_VERSION_3_8_0},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "If this option is enabled, self-heal will resort to "
- "granular way of recording changelogs and doing entry "
- "self-heal.",
- },
- { .key = {"favorite-child-policy"},
- .type = GF_OPTION_TYPE_STR,
- .value = {"none", "size", "ctime", "mtime", "majority"},
- .default_value = "none",
- .op_version = {GD_OP_VERSION_3_7_12},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This option can be used to automatically resolve "
- "split-brains using various policies without user "
- "intervention. \"size\" picks the file with the "
- "biggest size as the source. \"ctime\" and \"mtime\" "
- "pick the file with the latest ctime and mtime "
- "respectively as the source. \"majority\" picks a file"
- " with identical mtime and size in more than half the "
- "number of bricks in the replica.",
- },
- { .key = {"consistent-io"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "no",
- .description = "If this option is enabled, i/o will fail even if "
- "one of the bricks is down in the replicas",
- },
- { .key = {"use-compound-fops"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "no",
- .op_version = {GD_OP_VERSION_3_8_4},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"replicate"},
- .description = "This option exists only for backward compatibility "
- "and configuring it doesn't have any effect"
- },
- { .key = {NULL} },
+ {.key = {"read-subvolume"},
+ .type = GF_OPTION_TYPE_XLATOR,
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "inode-read fops happen only on one of the bricks in "
+ "replicate. Afr will prefer the one specified using "
+ "this option if it is not stale. Option value must be "
+ "one of the xlator names of the children. "
+ "Ex: <volname>-client-0 till "
+ "<volname>-client-<number-of-bricks - 1>"},
+ {.key = {"read-subvolume-index"},
+ .type = GF_OPTION_TYPE_INT,
+ .default_value = "-1",
+ .op_version = {2},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "inode-read fops happen only on one of the bricks in "
+ "replicate. AFR will prefer the one specified using "
+ "this option if it is not stale. allowed options"
+ " include -1 till replica-count - 1"},
+ {.key = {"read-hash-mode"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max = 5,
+ .default_value = "1",
+ .op_version = {2},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description =
+ "inode-read fops happen only on one of the bricks in "
+ "replicate. AFR will prefer the one computed using "
+ "the method specified using this option.\n"
+ "0 = first readable child of AFR, starting from 1st child.\n"
+ "1 = hash by GFID of file (all clients use "
+ "same subvolume).\n"
+ "2 = hash by GFID of file and client PID.\n"
+ "3 = brick having the least outstanding read requests.\n"
+ "4 = brick having the least network ping latency.\n"
+ "5 = Hybrid mode between 3 and 4, ie least value among "
+ "network-latency multiplied by outstanding-read-requests."},
+ {
+ .key = {"choose-local"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "true",
+ .op_version = {2},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "Choose a local subvolume (i.e. Brick) to read from"
+ " if read-subvolume is not explicitly set.",
+ },
+ {.key = {"background-self-heal-count"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max = 256,
+ .default_value = "8",
+ .validate = GF_OPT_VALIDATE_MIN,
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This specifies the number of per client self-heal "
+ "jobs that can perform parallel heals in the "
+ "background."},
+ {.key = {"halo-shd-max-latency"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 99999,
+ .default_value = "99999",
+ .op_version = {GD_OP_VERSION_3_11_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate", "halo"},
+ .description = "Maximum latency for shd halo replication in msec."},
+ {.key = {"halo-enabled"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "False",
+ .op_version = {GD_OP_VERSION_3_11_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate", "halo"},
+ .description = "Enable Halo (geo) replication mode."},
+ {.key = {"halo-nfsd-max-latency"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 99999,
+ .default_value = "5",
+ .op_version = {GD_OP_VERSION_3_11_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate", "halo"},
+ .description = "Maximum latency for nfsd halo replication in msec."},
+ {.key = {"halo-max-latency"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = AFR_HALO_MAX_LATENCY,
+ .default_value = "5",
+ .op_version = {GD_OP_VERSION_3_11_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate", "halo"},
+ .description = "Maximum latency for halo replication in msec."},
+ {.key = {"halo-max-replicas"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 99999,
+ .default_value = "99999",
+ .op_version = {GD_OP_VERSION_3_11_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate", "halo"},
+ .description = "The maximum number of halo replicas; replicas"
+ " beyond this value will be written asynchronously"
+ "via the SHD."},
+ {.key = {"halo-min-replicas"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 99999,
+ .default_value = "2",
+ .op_version = {GD_OP_VERSION_3_11_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate", "halo"},
+ .description = "The minimmum number of halo replicas, before adding "
+ "out of region replicas."},
+ {.key = {"heal-wait-queue-length"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max = 10000, /*Around 100MB with sizeof(afr_local_t)= 10496 bytes*/
+ .default_value = "128",
+ .validate = GF_OPT_VALIDATE_MIN,
+ .op_version = {GD_OP_VERSION_3_7_10},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This specifies the number of heals that can be queued"
+ " for the parallel background self heal jobs."},
+ {.key = {"data-self-heal"},
+ .type = GF_OPTION_TYPE_STR,
+ .value = {"1", "on", "yes", "true", "enable", "0", "off", "no", "false",
+ "disable", "open"},
+ .default_value = "off",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "Using this option we can enable/disable data "
+ "self-heal on the file. \"open\" means data "
+ "self-heal action will only be triggered by file "
+ "open operations."},
+ {.key = {"data-self-heal-algorithm"},
+ .type = GF_OPTION_TYPE_STR,
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "Select between \"full\", \"diff\". The "
+ "\"full\" algorithm copies the entire file from "
+ "source to sink. The \"diff\" algorithm copies to "
+ "sink only those blocks whose checksums don't match "
+ "with those of source. If no option is configured "
+ "the option is chosen dynamically as follows: "
+ "If the file does not exist on one of the sinks "
+ "or empty file exists or if the source file size is "
+ "about the same as page size the entire file will "
+ "be read and written i.e \"full\" algo, "
+ "otherwise \"diff\" algo is chosen.",
+ .value = {"diff", "full"}},
+ {.key = {"data-self-heal-window-size"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 1024,
+ .default_value = "1",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "Maximum number blocks per file for which self-heal "
+ "process would be applied simultaneously."},
+ {.key = {"metadata-self-heal"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ /*.validate_fn = validate_replica*/
+ .description = "Using this option we can enable/disable metadata "
+ "i.e. Permissions, ownerships, xattrs self-heal on "
+ "the file/directory."},
+ {.key = {"entry-self-heal"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ /*.validate_fn = validate_replica*/
+ .description = "Using this option we can enable/disable entry "
+ "self-heal on the directory."},
+ {.key = {"data-change-log"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This option exists only for backward compatibility "
+ "and configuring it doesn't have any effect"},
+ {.key = {"metadata-change-log"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This option exists only for backward compatibility "
+ "and configuring it doesn't have any effect"},
+ {.key = {"entry-change-log"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This option exists only for backward compatibility "
+ "and configuring it doesn't have any effect"},
+ {.key = {"optimistic-change-log"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "Entry/Metadata fops will not perform "
+ "pre fop changelog operations in afr transaction "
+ "if this option is enabled."},
+ {.key = {"inodelk-trace"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "Enabling this option logs inode lock/unlocks"},
+ {.key = {"entrylk-trace"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "Enabling this option logs entry lock/unlocks"},
+ {.key = {"pre-op-compat"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "Use separate pre-op xattrop() FOP rather than "
+ "overloading xdata of the OP"},
+ {.key = {"eager-lock"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description =
+ "Enable/Disable eager lock for replica volume. "
+ "Lock phase of a transaction has two sub-phases. "
+ "First is an attempt to acquire locks in parallel by "
+ "broadcasting non-blocking lock requests. If lock "
+ "acquisition fails on any server, then the held locks "
+ "are unlocked and we revert to a blocking locks mode "
+ "sequentially on one server after another. If this "
+ "option is enabled the initial broadcasting lock "
+ "request attempts to acquire a full lock on the entire file. "
+ "If this fails, we revert back to the sequential "
+ "\"regional\" blocking locks as before. In the case "
+ "where such an \"eager\" lock is granted in the "
+ "non-blocking phase, it gives rise to an opportunity "
+ "for optimization. i.e, if the next write transaction "
+ "on the same FD arrives before the unlock phase of "
+ "the first transaction, it \"takes over\" the full "
+ "file lock. Similarly if yet another data transaction "
+ "arrives before the unlock phase of the \"optimized\" "
+ "transaction, that in turn \"takes over\" the lock as "
+ "well. The actual unlock now happens at the end of "
+ "the last \"optimized\" transaction."
+
+ },
+ {.key = {"self-heal-daemon"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
+ .tags = {"replicate"},
+ /*.validate_fn = validate_replica_heal_enable_disable*/
+ .description = "This option applies to only self-heal-daemon. "
+ "Index directory crawl and automatic healing of files "
+ "will not be performed if this option is turned off."},
+ {.key = {"iam-self-heal-daemon"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "This option differentiates if the replicate "
+ "translator is running as part of self-heal-daemon "
+ "or not."},
+ {.key = {"iam-nfs-daemon"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "This option differentiates if the replicate "
+ "translator is running as part of an NFS daemon "
+ "or not."},
+ {
+ .key = {"quorum-type"},
+ .type = GF_OPTION_TYPE_STR,
+ .value = {"none", "auto", "fixed"},
+ .default_value = "none",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ /*.option = quorum-type*/
+ .description = "If value is \"fixed\" only allow writes if "
+ "quorum-count bricks are present. If value is "
+ "\"auto\" only allow writes if more than half of "
+ "bricks, or exactly half including the first, are "
+ "present.",
+ },
+ {
+ .key = {"quorum-count"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = INT_MAX,
+ .default_value = 0,
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ /*.option = quorum-count*/
+ /*.validate_fn = validate_quorum_count*/
+ .description = "If quorum-type is \"fixed\" only allow writes if "
+ "this many bricks are present. Other quorum types "
+ "will OVERWRITE this value.",
+ },
+ {
+ .key = {"quorum-reads"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "no",
+ .op_version = {GD_OP_VERSION_3_7_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This option has been removed. Reads are not allowed "
+ "if quorum is not met.",
+ },
+ {
+ .key = {"node-uuid"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Local glusterd uuid string, used in starting "
+ "self-heal-daemon so that it can crawl only on "
+ "local index directories.",
+ },
+ {
+ .key = {"post-op-delay-secs"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max = INT_MAX,
+ .default_value = "1",
+ .op_version = {2},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "Time interval induced artificially before "
+ "post-operation phase of the transaction to "
+ "enhance overlap of adjacent write operations.",
+ },
+ {
+ .key = {AFR_SH_READDIR_SIZE_KEY},
+ .type = GF_OPTION_TYPE_SIZET,
+ .description = "readdirp size for performing entry self-heal",
+ .min = 1024,
+ .max = 131072,
+ .op_version = {2},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
+ .tags = {"replicate"},
+ .default_value = "1KB",
+ },
+ {
+ .key = {"ensure-durability"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .op_version = {3},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "Afr performs fsyncs for transactions if this "
+ "option is on to make sure the changelogs/data is "
+ "written to the disk",
+ .default_value = "on",
+ },
+ {
+ .key = {"afr-dirty-xattr"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = AFR_DIRTY_DEFAULT,
+ },
+ {.key = {"afr-pending-xattr"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Comma separated list of xattrs that are used to "
+ "capture information on pending heals."},
+ {
+ .key = {"metadata-splitbrain-forced-heal"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ },
+ {.key = {"heal-timeout"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 5,
+ .max = INT_MAX,
+ .default_value = "600",
+ .op_version = {2},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "time interval for checking the need to self-heal "
+ "in self-heal-daemon"},
+ {
+ .key = {"consistent-metadata"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "no",
+ .op_version = {GD_OP_VERSION_3_7_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "If this option is enabled, readdirp will force "
+ "lookups on those entries read whose read child is "
+ "not the same as that of the parent. This will "
+ "guarantee that all read operations on a file serve "
+ "attributes from the same subvol as long as it holds "
+ " a good copy of the file/dir.",
+ },
+ {.key = {"arbiter-count"},
+ .type = GF_OPTION_TYPE_INT,
+ .description = "subset of child_count. Has to be 0 or 1."},
+ {
+ .key = {"thin-arbiter"},
+ .type = GF_OPTION_TYPE_STR,
+ .op_version = {GD_OP_VERSION_4_1_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .tags = {"replicate"},
+ .description = "contains host:path of thin abriter brick",
+ },
+ {.key = {"shd-max-threads"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 64,
+ .default_value = "1",
+ .op_version = {GD_OP_VERSION_3_7_12},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "Maximum number of parallel heals SHD can do per "
+ "local brick. This can substantially lower heal times"
+ ", but can also crush your bricks if you don't have "
+ "the storage hardware to support this."},
+ {
+ .key = {"shd-wait-qlength"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 655536,
+ .default_value = "1024",
+ .op_version = {GD_OP_VERSION_3_7_12},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This option can be used to control number of heals"
+ " that can wait in SHD per subvolume",
+ },
+ {
+ .key = {"locking-scheme"},
+ .type = GF_OPTION_TYPE_STR,
+ .value = {"full", "granular"},
+ .default_value = "full",
+ .op_version = {GD_OP_VERSION_3_7_12},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "If this option is set to granular, self-heal will "
+ "stop being compatible with afr-v1, which helps afr "
+ "be more granular while self-healing",
+ },
+ {.key = {"full-lock"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "yes",
+ .op_version = {GD_OP_VERSION_3_13_2},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
+ .tags = {"replicate"},
+ .description = "If this option is disabled, then the IOs will take "
+ "range locks same as versions till 3.13.1."},
+ {
+ .key = {"granular-entry-heal"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "no",
+ .op_version = {GD_OP_VERSION_3_8_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "If this option is enabled, self-heal will resort to "
+ "granular way of recording changelogs and doing entry "
+ "self-heal.",
+ },
+ {
+ .key = {"favorite-child-policy"},
+ .type = GF_OPTION_TYPE_STR,
+ .value = {"none", "size", "ctime", "mtime", "majority"},
+ .default_value = "none",
+ .op_version = {GD_OP_VERSION_3_7_12},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This option can be used to automatically resolve "
+ "split-brains using various policies without user "
+ "intervention. \"size\" picks the file with the "
+ "biggest size as the source. \"ctime\" and \"mtime\" "
+ "pick the file with the latest ctime and mtime "
+ "respectively as the source. \"majority\" picks a file"
+ " with identical mtime and size in more than half the "
+ "number of bricks in the replica.",
+ },
+ {
+ .key = {"consistent-io"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "no",
+ .description = "If this option is enabled, i/o will fail even if "
+ "one of the bricks is down in the replicas",
+ },
+ {.key = {"use-compound-fops"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "no",
+ .op_version = {GD_OP_VERSION_3_8_4},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"replicate"},
+ .description = "This option exists only for backward compatibility "
+ "and configuring it doesn't have any effect"},
+ {.key = {"use-anonymous-inode"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "no",
+ .op_version = {GD_OP_VERSION_8_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
+ .tags = {"replicate"},
+ .description = "Setting this option heals directory renames efficiently"},
+
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .notify = notify,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .dumpops = &dumpops,
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "replicate",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/cluster/afr/src/afr.h b/xlators/cluster/afr/src/afr.h
index 68087e0ea20..d62f9a9caf2 100644
--- a/xlators/cluster/afr/src/afr.h
+++ b/xlators/cluster/afr/src/afr.h
@@ -8,181 +8,291 @@
cases as published by the Free Software Foundation.
*/
-
#ifndef __AFR_H__
#define __AFR_H__
-#include "call-stub.h"
-#include "compat-errno.h"
+#include <glusterfs/call-stub.h>
+#include <glusterfs/compat-errno.h>
#include "afr-mem-types.h"
#include "libxlator.h"
-#include "timer.h"
-#include "syncop.h"
+#include <glusterfs/timer.h>
+#include <glusterfs/syncop.h>
#include "afr-self-heald.h"
#include "afr-messages.h"
+#define SHD_INODE_LRU_LIMIT 1
#define AFR_PATHINFO_HEADER "REPLICATE:"
#define AFR_SH_READDIR_SIZE_KEY "self-heal-readdir-size"
#define AFR_SH_DATA_DOMAIN_FMT "%s:self-heal"
#define AFR_DIRTY_DEFAULT AFR_XATTR_PREFIX ".dirty"
-#define AFR_DIRTY (((afr_private_t *) (THIS->private))->afr_dirty)
+#define AFR_DIRTY (((afr_private_t *)(THIS->private))->afr_dirty)
-#define AFR_LOCKEE_COUNT_MAX 3
-#define AFR_DOM_COUNT_MAX 3
-#define AFR_NUM_CHANGE_LOGS 3 /*data + metadata + entry*/
+#define AFR_LOCKEE_COUNT_MAX 3
+#define AFR_DOM_COUNT_MAX 3
+#define AFR_NUM_CHANGE_LOGS 3 /*data + metadata + entry*/
#define AFR_DEFAULT_SPB_CHOICE_TIMEOUT 300 /*in seconds*/
#define ARBITER_BRICK_INDEX 2
#define THIN_ARBITER_BRICK_INDEX 2
-#define THIN_ARBITER_DOM1 "afr.ta.domain-1"
+#define AFR_TA_DOM_NOTIFY "afr.ta.dom-notify"
+#define AFR_TA_DOM_MODIFY "afr.ta.dom-modify"
+
+#define AFR_LK_HEAL_DOM "afr.lock-heal.domain"
#define AFR_HALO_MAX_LATENCY 99999
-typedef int (*afr_lock_cbk_t) (call_frame_t *frame, xlator_t *this);
+#define AFR_ANON_DIR_PREFIX ".glusterfs-anonymous-inode"
+
+#define PFLAG_PENDING (1 << 0)
+#define PFLAG_SBRAIN (1 << 1)
+
+typedef int (*afr_lock_cbk_t)(call_frame_t *frame, xlator_t *this);
+
+typedef int (*afr_read_txn_wind_t)(call_frame_t *frame, xlator_t *this,
+ int subvol);
+
+typedef int (*afr_inode_refresh_cbk_t)(call_frame_t *frame, xlator_t *this,
+ int err);
+
+typedef int (*afr_changelog_resume_t)(call_frame_t *frame, xlator_t *this);
+
+#define AFR_COUNT(array, max) \
+ ({ \
+ int __i; \
+ int __res = 0; \
+ for (__i = 0; __i < max; __i++) \
+ if (array[__i]) \
+ __res++; \
+ __res; \
+ })
+#define AFR_INTERSECT(dst, src1, src2, max) \
+ ({ \
+ int __i; \
+ for (__i = 0; __i < max; __i++) \
+ dst[__i] = src1[__i] && src2[__i]; \
+ })
+#define AFR_CMP(a1, a2, len) \
+ ({ \
+ int __cmp = 0; \
+ int __i; \
+ for (__i = 0; __i < len; __i++) \
+ if (a1[__i] != a2[__i]) { \
+ __cmp = 1; \
+ break; \
+ } \
+ __cmp; \
+ })
+#define AFR_IS_ARBITER_BRICK(priv, index) \
+ ((priv->arbiter_count == 1) && (index == ARBITER_BRICK_INDEX))
-typedef int (*afr_read_txn_wind_t) (call_frame_t *frame, xlator_t *this, int subvol);
+#define AFR_SET_ERROR_AND_CHECK_SPLIT_BRAIN(ret, errnum) \
+ do { \
+ local->op_ret = ret; \
+ local->op_errno = errnum; \
+ if (local->op_errno == EIO) \
+ gf_msg(this->name, GF_LOG_ERROR, local->op_errno, \
+ AFR_MSG_SPLIT_BRAIN, \
+ "Failing %s on gfid %s: " \
+ "split-brain observed.", \
+ gf_fop_list[local->op], uuid_utoa(local->inode->gfid)); \
+ } while (0)
+
+#define AFR_ERROR_OUT_IF_FDCTX_INVALID(__fd, __this, __error, __label) \
+ do { \
+ afr_fd_ctx_t *__fd_ctx = NULL; \
+ __fd_ctx = afr_fd_ctx_get(__fd, __this); \
+ if (__fd_ctx && __fd_ctx->is_fd_bad) { \
+ __error = EBADF; \
+ goto __label; \
+ } \
+ } while (0)
-typedef int (*afr_inode_refresh_cbk_t) (call_frame_t *frame, xlator_t *this, int err);
+typedef enum {
+ AFR_READ_POLICY_FIRST_UP,
+ AFR_READ_POLICY_GFID_HASH,
+ AFR_READ_POLICY_GFID_PID_HASH,
+ AFR_READ_POLICY_LESS_LOAD,
+ AFR_READ_POLICY_LEAST_LATENCY,
+ AFR_READ_POLICY_LOAD_LATENCY_HYBRID,
+} afr_read_hash_mode_t;
-typedef int (*afr_changelog_resume_t) (call_frame_t *frame, xlator_t *this);
+typedef enum {
+ AFR_FAV_CHILD_NONE,
+ AFR_FAV_CHILD_BY_SIZE,
+ AFR_FAV_CHILD_BY_CTIME,
+ AFR_FAV_CHILD_BY_MTIME,
+ AFR_FAV_CHILD_BY_MAJORITY,
+ AFR_FAV_CHILD_POLICY_MAX,
+} afr_favorite_child_policy;
-#define AFR_COUNT(array,max) ({int __i; int __res = 0; for (__i = 0; __i < max; __i++) if (array[__i]) __res++; __res;})
-#define AFR_INTERSECT(dst,src1,src2,max) ({int __i; for (__i = 0; __i < max; __i++) dst[__i] = src1[__i] && src2[__i];})
-#define AFR_CMP(a1,a2,len) ({int __cmp = 0; int __i; for (__i = 0; __i < len; __i++) if (a1[__i] != a2[__i]) { __cmp = 1; break;} __cmp;})
-#define AFR_IS_ARBITER_BRICK(priv, index) ((priv->arbiter_count == 1) && (index == ARBITER_BRICK_INDEX))
+typedef enum {
+ AFR_SELFHEAL_DATA_FULL = 0,
+ AFR_SELFHEAL_DATA_DIFF,
+ AFR_SELFHEAL_DATA_DYNAMIC,
+} afr_data_self_heal_type_t;
-#define AFR_SET_ERROR_AND_CHECK_SPLIT_BRAIN(ret, errnum) \
- do { \
- local->op_ret = ret; \
- local->op_errno = errnum; \
- if (local->op_errno == EIO) \
- gf_msg (this->name, GF_LOG_ERROR, local->op_errno, \
- AFR_MSG_SPLIT_BRAIN, "Failing %s on gfid %s: " \
- "split-brain observed.", \
- gf_fop_list[local->op], \
- uuid_utoa (local->inode->gfid)); \
- } while (0)
+typedef enum {
+ AFR_CHILD_UNKNOWN = -1,
+ AFR_CHILD_ZERO,
+ AFR_CHILD_ONE,
+ AFR_CHILD_THIN_ARBITER,
+} afr_child_index;
typedef enum {
- AFR_FAV_CHILD_NONE,
- AFR_FAV_CHILD_BY_SIZE,
- AFR_FAV_CHILD_BY_CTIME,
- AFR_FAV_CHILD_BY_MTIME,
- AFR_FAV_CHILD_BY_MAJORITY,
- AFR_FAV_CHILD_POLICY_MAX,
-} afr_favorite_child_policy;
+ TA_WAIT_FOR_NOTIFY_LOCK_REL, /*FOP came after notify domain lock upcall
+ notification and waiting for its release.*/
+ TA_GET_INFO_FROM_TA_FILE, /*FOP needs post-op on ta file to get
+ *info about which brick is bad.*/
+ TA_INFO_IN_MEMORY_SUCCESS, /*Bad brick info is in memory and fop failed
+ *on BAD brick - Success*/
+ TA_INFO_IN_MEMORY_FAILED, /*Bad brick info is in memory and fop failed
+ *on GOOD brick - Failed*/
+ TA_SUCCESS, /*FOP succeeded on both data bricks.*/
+} afr_ta_fop_state_t;
struct afr_nfsd {
- gf_boolean_t iamnfsd;
- uint32_t halo_max_latency_msec;
+ uint32_t halo_max_latency_msec;
+ gf_boolean_t iamnfsd;
};
+typedef struct _afr_lk_heal_info {
+ fd_t *fd;
+ int32_t cmd;
+ struct gf_flock flock;
+ dict_t *xdata_req;
+ unsigned char *locked_nodes;
+ struct list_head pos;
+ gf_lkowner_t lk_owner;
+ pid_t pid;
+ int32_t *child_up_event_gen;
+ int32_t *child_down_event_gen;
+} afr_lk_heal_info_t;
+
typedef struct _afr_private {
- gf_lock_t lock; /* to guard access to child_count, etc */
- unsigned int child_count; /* total number of children */
- unsigned int arbiter_count; /*subset of child_count.
- Has to be 0 or 1.*/
- unsigned int thin_arbiter_count; /* 0 or 1 at the moment.*/
-
- xlator_t **children;
-
- inode_t *root_inode;
- uuid_t ta_gfid; /*For thin arbiter.*/
-
- unsigned char *child_up;
- int64_t *child_latency;
- unsigned char *local;
-
- char **pending_key;
-
- char *data_self_heal; /* on/off/open */
- char * data_self_heal_algorithm; /* name of algorithm */
- unsigned int data_self_heal_window_size; /* max number of pipelined
- read/writes */
-
- struct list_head heal_waiting; /*queue for files that need heal*/
- uint32_t heal_wait_qlen; /*configurable queue length for heal_waiting*/
- int32_t heal_waiters; /* No. of elements currently in wait queue.*/
-
- struct list_head healing;/* queue for files that are undergoing
- background heal*/
- uint32_t background_self_heal_count;/*configurable queue length for
- healing queue*/
- int32_t healers;/* No. of elements currently undergoing background
- heal*/
-
- gf_boolean_t metadata_self_heal; /* on/off */
- gf_boolean_t entry_self_heal; /* on/off */
-
- gf_boolean_t metadata_splitbrain_forced_heal; /* on/off */
- int read_child; /* read-subvolume */
- unsigned int hash_mode; /* for when read_child is not set */
- gf_atomic_t *pending_reads; /*No. of pending read cbks per child.*/
- int favorite_child; /* subvolume to be preferred in resolving
- split-brain cases */
-
- afr_favorite_child_policy fav_child_policy;/*Policy to use for automatic
- resolution of split-brains.*/
-
- unsigned int wait_count; /* # of servers to wait for success */
-
- gf_timer_t *timer; /* launched when parent up is received */
-
- gf_boolean_t optimistic_change_log;
- gf_boolean_t eager_lock;
- gf_boolean_t pre_op_compat; /* on/off */
- uint32_t post_op_delay_secs;
- unsigned int quorum_count;
-
- char vol_uuid[UUID_SIZE + 1];
- int32_t *last_event;
-
- /* @event_generation: Keeps count of number of events received which can
- potentially impact consistency decisions. The events are CHILD_UP
- and CHILD_DOWN, when we have to recalculate the freshness/staleness
- of copies to detect if changes had happened while the other server
- was down. CHILD_DOWN and CHILD_UP can also be received on network
- disconnect/reconnects and not necessarily server going down/up.
- Recalculating freshness/staleness on network events is equally
- important as we might have had a network split brain.
- */
- uint32_t event_generation;
-
- gf_boolean_t choose_local;
- gf_boolean_t did_discovery;
- uint64_t sh_readdir_size;
- gf_boolean_t ensure_durability;
- char *sh_domain;
- char *afr_dirty;
- gf_boolean_t halo_enabled;
-
- uint32_t halo_max_latency_msec;
- uint32_t halo_max_replicas;
- uint32_t halo_min_replicas;
-
- afr_self_heald_t shd;
- struct afr_nfsd nfsd;
-
- gf_boolean_t consistent_metadata;
- uint64_t spb_choice_timeout;
- gf_boolean_t need_heal;
-
- /* pump dependencies */
- void *pump_private;
- gf_boolean_t use_afr_in_pump;
- char *locking_scheme;
- gf_boolean_t full_lock;
- gf_boolean_t esh_granular;
- gf_boolean_t consistent_io;
+ gf_lock_t lock; /* to guard access to child_count, etc */
+ unsigned int child_count; /* total number of children */
+ unsigned int arbiter_count; /*subset of child_count.
+ Has to be 0 or 1.*/
+
+ xlator_t **children;
+
+ inode_t *root_inode;
+
+ int favorite_child; /* subvolume to be preferred in resolving
+ split-brain cases */
+ /* For thin-arbiter. */
+ uuid_t ta_gfid;
+ unsigned int thin_arbiter_count; /* 0 or 1 at the moment.*/
+ int ta_bad_child_index;
+ int ta_event_gen;
+ unsigned int ta_in_mem_txn_count;
+ unsigned int ta_on_wire_txn_count;
+ struct list_head ta_waitq;
+ struct list_head ta_onwireq;
+
+ unsigned char *anon_inode;
+ unsigned char *child_up;
+ unsigned char *halo_child_up;
+ int64_t *child_latency;
+ unsigned char *local;
+
+ char **pending_key;
+
+ afr_data_self_heal_type_t data_self_heal_algorithm;
+ unsigned int data_self_heal_window_size; /* max number of pipelined
+ read/writes */
+
+ struct list_head heal_waiting; /*queue for files that need heal*/
+ uint32_t heal_wait_qlen; /*configurable queue length for heal_waiting*/
+ int32_t heal_waiters; /* No. of elements currently in wait queue.*/
+
+ struct list_head healing; /* queue for files that are undergoing
+ background heal*/
+ uint32_t background_self_heal_count; /*configurable queue length for
+ healing queue*/
+ int32_t healers; /* No. of elements currently undergoing background
+ heal*/
+
+ gf_boolean_t release_ta_notify_dom_lock;
+
+ gf_boolean_t metadata_self_heal; /* on/off */
+ gf_boolean_t entry_self_heal; /* on/off */
+
+ gf_boolean_t metadata_splitbrain_forced_heal; /* on/off */
+ int read_child; /* read-subvolume */
+ gf_atomic_t *pending_reads; /*No. of pending read cbks per child.*/
+
+ gf_timer_t *timer; /* launched when parent up is received */
+
+ unsigned int wait_count; /* # of servers to wait for success */
+
+ unsigned char ta_child_up;
+ gf_boolean_t optimistic_change_log;
+ gf_boolean_t eager_lock;
+ gf_boolean_t pre_op_compat; /* on/off */
+ uint32_t post_op_delay_secs;
+ unsigned int quorum_count;
+
+ off_t ta_notify_dom_lock_offset;
+ afr_favorite_child_policy fav_child_policy; /*Policy to use for automatic
+ resolution of split-brains.*/
+ afr_read_hash_mode_t hash_mode; /* for when read_child is not set */
+
+ int32_t *last_event;
+
+ /* @event_generation: Keeps count of number of events received which can
+ potentially impact consistency decisions. The events are CHILD_UP
+ and CHILD_DOWN, when we have to recalculate the freshness/staleness
+ of copies to detect if changes had happened while the other server
+ was down. CHILD_DOWN and CHILD_UP can also be received on network
+ disconnect/reconnects and not necessarily server going down/up.
+ Recalculating freshness/staleness on network events is equally
+ important as we might have had a network split brain.
+ */
+ uint32_t event_generation;
+ char vol_uuid[UUID_SIZE + 1];
+
+ gf_boolean_t choose_local;
+ gf_boolean_t did_discovery;
+ gf_boolean_t ensure_durability;
+ gf_boolean_t halo_enabled;
+ gf_boolean_t consistent_metadata;
+ gf_boolean_t need_heal;
+ gf_boolean_t granular_locks;
+ uint64_t sh_readdir_size;
+ char *sh_domain;
+ char *afr_dirty;
+
+ uint64_t spb_choice_timeout;
+
+ afr_self_heald_t shd;
+ struct afr_nfsd nfsd;
+
+ uint32_t halo_max_latency_msec;
+ uint32_t halo_max_replicas;
+ uint32_t halo_min_replicas;
+
+ gf_boolean_t full_lock;
+ gf_boolean_t esh_granular;
+ gf_boolean_t consistent_io;
+ gf_boolean_t data_self_heal; /* on/off */
+ gf_boolean_t use_anon_inode;
+
+ /*For lock healing.*/
+ struct list_head saved_locks;
+ struct list_head lk_healq;
+
+ /*For anon-inode handling */
+ char anon_inode_name[NAME_MAX + 1];
+ char anon_gfid_str[UUID_SIZE + 1];
} afr_private_t;
-
typedef enum {
- AFR_DATA_TRANSACTION, /* truncate, write, ... */
- AFR_METADATA_TRANSACTION, /* chmod, chown, ... */
- AFR_ENTRY_TRANSACTION, /* create, rmdir, ... */
- AFR_ENTRY_RENAME_TRANSACTION, /* rename */
+ AFR_DATA_TRANSACTION, /* truncate, write, ... */
+ AFR_METADATA_TRANSACTION, /* chmod, chown, ... */
+ AFR_ENTRY_TRANSACTION, /* create, rmdir, ... */
+ AFR_ENTRY_RENAME_TRANSACTION, /* rename */
} afr_transaction_type;
/*
@@ -193,914 +303,940 @@ typedef enum {
*/
static inline int
-afr_index_for_transaction_type (afr_transaction_type type)
+afr_index_for_transaction_type(afr_transaction_type type)
{
- switch (type) {
-
+ switch (type) {
case AFR_DATA_TRANSACTION:
- return 0;
+ return 0;
case AFR_METADATA_TRANSACTION:
- return 1;
+ return 1;
case AFR_ENTRY_TRANSACTION:
case AFR_ENTRY_RENAME_TRANSACTION:
- return 2;
- }
+ return 2;
+ }
- return -1; /* make gcc happy */
+ return -1; /* make gcc happy */
}
static inline int
-afr_index_from_ia_type (ia_type_t type)
+afr_index_from_ia_type(ia_type_t type)
{
- switch (type) {
+ switch (type) {
case IA_IFDIR:
- return afr_index_for_transaction_type (AFR_ENTRY_TRANSACTION);
+ return afr_index_for_transaction_type(AFR_ENTRY_TRANSACTION);
case IA_IFREG:
- return afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
- default: return -1;
- }
+ return afr_index_for_transaction_type(AFR_DATA_TRANSACTION);
+ default:
+ return -1;
+ }
}
typedef struct {
- loc_t loc;
- char *basename;
- unsigned char *locked_nodes;
- int locked_count;
+ struct gf_flock flock;
+ loc_t loc;
+ fd_t *fd;
+ char *basename;
+ unsigned char *locked_nodes;
+ int locked_count;
-} afr_entry_lockee_t;
+} afr_lockee_t;
int
-afr_entry_lockee_cmp (const void *l1, const void *l2);
+afr_entry_lockee_cmp(const void *l1, const void *l2);
typedef struct {
- loc_t *lk_loc;
+ loc_t *lk_loc;
+
+ afr_lockee_t lockee[AFR_LOCKEE_COUNT_MAX];
- int lockee_count;
- afr_entry_lockee_t lockee[AFR_LOCKEE_COUNT_MAX];
+ const char *lk_basename;
+ const char *lower_basename;
+ const char *higher_basename;
- struct gf_flock flock;
- const char *lk_basename;
- const char *lower_basename;
- const char *higher_basename;
- char lower_locked;
- char higher_locked;
+ unsigned char *lower_locked_nodes;
- unsigned char *locked_nodes;
- unsigned char *lower_locked_nodes;
+ afr_lock_cbk_t lock_cbk;
- int32_t lock_count;
- int32_t entrylk_lock_count;
+ int lockee_count;
- int32_t lk_call_count;
- int32_t lk_expected_count;
- int32_t lk_attempted_count;
+ int32_t lk_call_count;
+ int32_t lk_expected_count;
+ int32_t lk_attempted_count;
- int32_t lock_op_ret;
- int32_t lock_op_errno;
- afr_lock_cbk_t lock_cbk;
- char *domain; /* Domain on which inode/entry lock/unlock in progress.*/
+ int32_t lock_op_ret;
+ int32_t lock_op_errno;
+ char *domain; /* Domain on which inode/entry lock/unlock in progress.*/
+ int32_t lock_count;
+ char lower_locked;
+ char higher_locked;
} afr_internal_lock_t;
struct afr_reply {
- int valid;
- int32_t op_ret;
- int32_t op_errno;
- dict_t *xattr;/*For xattrop*/
- dict_t *xdata;
- struct iatt poststat;
- struct iatt postparent;
- struct iatt prestat;
- struct iatt preparent;
- struct iatt preparent2;
- struct iatt postparent2;
- /* For rchecksum */
- uint8_t checksum[SHA256_DIGEST_LENGTH];
- gf_boolean_t buf_has_zeroes;
- gf_boolean_t fips_mode_rchecksum;
- /* For lookup */
- int8_t need_heal;
+ int valid;
+ int32_t op_ret;
+ dict_t *xattr; /*For xattrop*/
+ dict_t *xdata;
+ struct iatt poststat;
+ struct iatt postparent;
+ struct iatt prestat;
+ struct iatt preparent;
+ struct iatt preparent2;
+ struct iatt postparent2;
+ int32_t op_errno;
+ /* For rchecksum */
+ uint8_t checksum[SHA256_DIGEST_LENGTH];
+ gf_boolean_t buf_has_zeroes;
+ gf_boolean_t fips_mode_rchecksum;
+ /* For lookup */
+ int8_t need_heal;
};
typedef enum {
- AFR_FD_NOT_OPENED,
- AFR_FD_OPENED,
- AFR_FD_OPENING
+ AFR_FD_NOT_OPENED,
+ AFR_FD_OPENED,
+ AFR_FD_OPENING
} afr_fd_open_status_t;
typedef struct {
- afr_fd_open_status_t *opened_on; /* which subvolumes the fd is open on */
- int flags;
-
- /* the subvolume on which the latest sequence of readdirs (starting
- at offset 0) has begun. Till the next readdir request with 0 offset
- arrives, we continue to read off this subvol.
- */
- int readdir_subvol;
+ afr_fd_open_status_t *opened_on; /* which subvolumes the fd is open on */
+ int flags;
+
+ /* the subvolume on which the latest sequence of readdirs (starting
+ at offset 0) has begun. Till the next readdir request with 0 offset
+ arrives, we continue to read off this subvol.
+ */
+ int readdir_subvol;
+ /* lock-healing related members. */
+ gf_boolean_t is_fd_bad;
+ afr_lk_heal_info_t *lk_heal_info;
+
} afr_fd_ctx_t;
typedef enum {
- AFR_FOP_LOCK_PARALLEL,
- AFR_FOP_LOCK_SERIAL,
- AFR_FOP_LOCK_QUORUM_FAILED,
+ AFR_FOP_LOCK_PARALLEL,
+ AFR_FOP_LOCK_SERIAL,
+ AFR_FOP_LOCK_QUORUM_FAILED,
} afr_fop_lock_state_t;
typedef struct _afr_inode_lock_t {
- unsigned int event_generation;
- gf_boolean_t release;
- gf_boolean_t acquired;
- gf_timer_t *delay_timer;
- struct list_head owners; /*Transactions that are performing fop*/
- struct list_head post_op;/*Transactions that are done with the fop
- *So can not conflict with the fops*/
- struct list_head waiting;/*Transaction that are waiting for
- *conflicting transactions to complete*/
- struct list_head frozen;/*Transactions that need to go as part of
- * next batch of eager-lock*/
+ /* @num_inodelks:
+ Number of inodelks queried from the server, as queried through
+ xdata in FOPs. Currently, used to decide if eager-locking must be
+ temporarily disabled.
+ */
+ int32_t num_inodelks;
+ unsigned int event_generation;
+ gf_timer_t *delay_timer;
+ struct list_head owners; /*Transactions that are performing fop*/
+ struct list_head post_op; /*Transactions that are done with the fop
+ *So can not conflict with the fops*/
+ struct list_head waiting; /*Transaction that are waiting for
+ *conflicting transactions to complete*/
+ struct list_head frozen; /*Transactions that need to go as part of
+ * next batch of eager-lock*/
+ gf_boolean_t release;
+ gf_boolean_t acquired;
} afr_lock_t;
typedef struct _afr_inode_ctx {
- uint64_t read_subvol;
- uint64_t write_subvol;
- int lock_count;
- int spb_choice;
- gf_timer_t *timer;
- gf_boolean_t need_refresh;
- unsigned int *pre_op_done[AFR_NUM_CHANGE_LOGS];
- int inherited[AFR_NUM_CHANGE_LOGS];
- int on_disk[AFR_NUM_CHANGE_LOGS];
-
- /* set if any write on this fd was a non stable write
- (i.e, without O_SYNC or O_DSYNC)
- */
- gf_boolean_t witnessed_unstable_write;
-
- /* @open_fd_count:
- Number of open FDs queried from the server, as queried through
- xdata in FOPs. Currently, used to decide if eager-locking must be
- temporarily disabled.
- */
- uint32_t open_fd_count;
- /*Only 2 types of transactions support eager-locks now. DATA/METADATA*/
- afr_lock_t lock[2];
+ uint64_t read_subvol;
+ uint64_t write_subvol;
+ int lock_count;
+ int spb_choice;
+ gf_timer_t *timer;
+ unsigned int *pre_op_done[AFR_NUM_CHANGE_LOGS];
+ int inherited[AFR_NUM_CHANGE_LOGS];
+ int on_disk[AFR_NUM_CHANGE_LOGS];
+ /*Only 2 types of transactions support eager-locks now. DATA/METADATA*/
+ afr_lock_t lock[2];
+
+ /* @open_fd_count:
+ Number of open FDs queried from the server, as queried through
+ xdata in FOPs. Currently, used to decide if eager-locking must be
+ temporarily disabled.
+ */
+ uint32_t open_fd_count;
+ gf_boolean_t need_refresh;
+
+ /* set if any write on this fd was a non stable write
+ (i.e, without O_SYNC or O_DSYNC)
+ */
+ gf_boolean_t witnessed_unstable_write;
} afr_inode_ctx_t;
-
typedef struct _afr_local {
- glusterfs_fop_t op;
- unsigned int call_count;
+ glusterfs_fop_t op;
+ unsigned int call_count;
- /* @event_generation: copy of priv->event_generation taken at the
- time of starting the transaction. The copy is made so that we
- have a stable value through the various phases of the transaction.
- */
- unsigned int event_generation;
+ /* @event_generation: copy of priv->event_generation taken at the
+ time of starting the transaction. The copy is made so that we
+ have a stable value through the various phases of the transaction.
+ */
+ unsigned int event_generation;
- uint32_t open_fd_count;
- gf_boolean_t update_open_fd_count;
+ uint32_t open_fd_count;
+ int32_t num_inodelks;
- gf_lkowner_t saved_lk_owner;
+ int32_t op_ret;
+ int32_t op_errno;
- int32_t op_ret;
- int32_t op_errno;
+ int dirty[AFR_NUM_CHANGE_LOGS];
- int32_t **pending;
+ int32_t **pending;
- int dirty[AFR_NUM_CHANGE_LOGS];
+ loc_t loc;
+ loc_t newloc;
- loc_t loc;
- loc_t newloc;
+ fd_t *fd;
+ afr_fd_ctx_t *fd_ctx;
- fd_t *fd;
- afr_fd_ctx_t *fd_ctx;
+ /* @child_up: copy of priv->child_up taken at the time of transaction
+ start. The copy is taken so that we have a stable child_up array
+ through the phases of the transaction as priv->child_up[i] can keep
+ changing through time.
+ */
+ unsigned char *child_up;
- /* @child_up: copy of priv->child_up taken at the time of transaction
- start. The copy is taken so that we have a stable child_up array
- through the phases of the transaction as priv->child_up[i] can keep
- changing through time.
- */
- unsigned char *child_up;
+ /* @read_attempted:
+ array of flags representing subvolumes where read operations of
+ the read transaction have already been attempted. The array is
+ first pre-filled with down subvolumes, and as reads are performed
+ on other subvolumes, those are set as well. This way if the read
+ operation fails we do not retry on that subvolume again.
+ */
+ unsigned char *read_attempted;
- /* @read_attempted:
- array of flags representing subvolumes where read operations of
- the read transaction have already been attempted. The array is
- first pre-filled with down subvolumes, and as reads are performed
- on other subvolumes, those are set as well. This way if the read
- operation fails we do not retry on that subvolume again.
- */
- unsigned char *read_attempted;
+ /* @readfn:
- /* @readfn:
+ pointer to function which will perform the read operation on a given
+ subvolume. Used in read transactions.
+ */
- pointer to function which will perform the read operation on a given
- subvolume. Used in read transactions.
- */
+ afr_read_txn_wind_t readfn;
- afr_read_txn_wind_t readfn;
+ /* @inode:
- /* @refreshed:
+ the inode on which the read txn is performed on. ref'ed and copied
+ from either fd->inode or loc.inode
+ */
- the inode was "refreshed" (i.e, pending xattrs from all subvols
- freshly inspected and inode ctx updated accordingly) as part of
- this transaction already.
- */
- gf_boolean_t refreshed;
+ inode_t *inode;
- /* @inode:
+ /* @parent[2]:
- the inode on which the read txn is performed on. ref'ed and copied
- from either fd->inode or loc.inode
- */
+ parent inode[s] on which directory transactions are performed.
+ */
- inode_t *inode;
+ inode_t *parent;
+ inode_t *parent2;
- /* @parent[2]:
+ /* @readable:
- parent inode[s] on which directory transactions are performed.
- */
+ array of flags representing servers from which a read can be
+ performed. This is the output of afr_inode_refresh()
+ */
+ unsigned char *readable;
+ unsigned char *readable2; /*For rename transaction*/
- inode_t *parent;
- inode_t *parent2;
+ afr_inode_refresh_cbk_t refreshfn;
- /* @readable:
+ /* @refreshinode:
- array of flags representing servers from which a read can be
- performed. This is the output of afr_inode_refresh()
- */
- unsigned char *readable;
- unsigned char *readable2; /*For rename transaction*/
+ Inode currently getting refreshed.
+ */
+ inode_t *refreshinode;
- int read_subvol; /* Current read subvolume */
+ dict_t *xattr_req;
- afr_inode_refresh_cbk_t refreshfn;
+ dict_t *dict;
- /* @refreshinode:
+ int read_subvol; /* Current read subvolume */
- Inode currently getting refreshed.
- */
- inode_t *refreshinode;
+ int optimistic_change_log;
- /*To handle setattr/setxattr on yet to be linked inode from dht*/
- uuid_t refreshgfid;
+ afr_internal_lock_t internal_lock;
- /*
- @pre_op_compat:
+ /*To handle setattr/setxattr on yet to be linked inode from dht*/
+ uuid_t refreshgfid;
- compatibility mode of pre-op. send a separate pre-op and
- op operations as part of transaction, rather than combining
- */
+ /* @refreshed:
- gf_boolean_t pre_op_compat;
+ the inode was "refreshed" (i.e, pending xattrs from all subvols
+ freshly inspected and inode ctx updated accordingly) as part of
+ this transaction already.
+ */
+ gf_boolean_t refreshed;
- dict_t *xattr_req;
+ gf_boolean_t update_num_inodelks;
+ gf_boolean_t update_open_fd_count;
- afr_internal_lock_t internal_lock;
+ /*
+ @pre_op_compat:
- dict_t *dict;
+ compatibility mode of pre-op. send a separate pre-op and
+ op operations as part of transaction, rather than combining
+ */
- int optimistic_change_log;
+ gf_boolean_t pre_op_compat;
- /* Is the current writev() going to perform a stable write?
- i.e, is fd->flags or @flags writev param have O_SYNC or
- O_DSYNC?
- */
- gf_boolean_t stable_write;
+ /* Is the current writev() going to perform a stable write?
+ i.e, is fd->flags or @flags writev param have O_SYNC or
+ O_DSYNC?
+ */
+ gf_boolean_t stable_write;
- /* This write appended to the file. Nnot necessarily O_APPEND,
- just means the offset of write was at the end of file.
- */
- gf_boolean_t append_write;
+ /* This write appended to the file. Nnot necessarily O_APPEND,
+ just means the offset of write was at the end of file.
+ */
+ gf_boolean_t append_write;
- /*
- This struct contains the arguments for the "continuation"
- (scheme-like) of fops
- */
+ /*
+ This struct contains the arguments for the "continuation"
+ (scheme-like) of fops
+ */
+ struct {
struct {
- struct {
- gf_boolean_t needs_fresh_lookup;
- uuid_t gfid_req;
- } lookup;
-
- struct {
- unsigned char buf_set;
- struct statvfs buf;
- } statfs;
-
- struct {
- int32_t flags;
- fd_t *fd;
- } open;
-
- struct {
- int32_t cmd;
- struct gf_flock user_flock;
- struct gf_flock ret_flock;
- unsigned char *locked_nodes;
- } lk;
-
- /* inode read */
-
- struct {
- int32_t mask;
- int last_index; /* index of the child we tried previously */
- } access;
-
- struct {
- int last_index;
- } stat;
-
- struct {
- int last_index;
- } fstat;
-
- struct {
- size_t size;
- int last_index;
- } readlink;
-
- struct {
- char *name;
- int last_index;
- long xattr_len;
- } getxattr;
-
- struct {
- size_t size;
- off_t offset;
- int last_index;
- uint32_t flags;
- } readv;
-
- /* dir read */
-
- struct {
- int success_count;
- int32_t op_ret;
- int32_t op_errno;
-
- uint32_t *checksum;
- } opendir;
-
- struct {
- int32_t op_ret;
- int32_t op_errno;
- size_t size;
- off_t offset;
- dict_t *dict;
- gf_boolean_t failed;
- int last_index;
- } readdir;
- /* inode write */
-
- struct {
- struct iatt prebuf;
- struct iatt postbuf;
- } inode_wfop; //common structure for all inode-write-fops
-
- struct {
- int32_t op_ret;
-
- struct iovec *vector;
- struct iobref *iobref;
- int32_t count;
- off_t offset;
- uint32_t flags;
- } writev;
-
- struct {
- off_t offset;
- } truncate;
-
- struct {
- off_t offset;
- } ftruncate;
-
- struct {
- struct iatt in_buf;
- int32_t valid;
- } setattr;
-
- struct {
- struct iatt in_buf;
- int32_t valid;
- } fsetattr;
-
- struct {
- dict_t *dict;
- int32_t flags;
- } setxattr;
-
- struct {
- dict_t *dict;
- int32_t flags;
- } fsetxattr;
-
- struct {
- char *name;
- } removexattr;
-
- struct {
- dict_t *xattr;
- gf_xattrop_flags_t optype;
- } xattrop;
-
- /* dir write */
-
- struct {
- inode_t *inode;
- struct iatt buf;
- struct iatt preparent;
- struct iatt postparent;
- struct iatt prenewparent;
- struct iatt postnewparent;
- } dir_fop; //common structure for all dir fops
-
- struct {
- fd_t *fd;
- dict_t *params;
- int32_t flags;
- mode_t mode;
- } create;
-
- struct {
- dev_t dev;
- mode_t mode;
- dict_t *params;
- } mknod;
-
- struct {
- int32_t mode;
- dict_t *params;
- } mkdir;
-
- struct {
- int flags;
- } rmdir;
-
- struct {
- dict_t *params;
- char *linkpath;
- } symlink;
-
- struct {
- int32_t mode;
- off_t offset;
- size_t len;
- } fallocate;
-
- struct {
- off_t offset;
- size_t len;
- } discard;
-
- struct {
- off_t offset;
- off_t len;
- struct iatt prebuf;
- struct iatt postbuf;
- } zerofill;
-
- struct {
- char *volume;
- int32_t cmd;
- int32_t in_cmd;
- struct gf_flock in_flock;
- struct gf_flock flock;
- void *xdata;
- } inodelk;
-
- struct {
- char *volume;
- char *basename;
- entrylk_cmd in_cmd;
- entrylk_cmd cmd;
- entrylk_type type;
- void *xdata;
- } entrylk;
-
- struct {
- off_t offset;
- gf_seek_what_t what;
- } seek;
-
- struct {
- int32_t datasync;
- } fsync;
-
- struct {
- struct gf_lease user_lease;
- struct gf_lease ret_lease;
- unsigned char *locked_nodes;
- } lease;
-
-
- } cont;
+ struct statvfs buf;
+ unsigned char buf_set;
+ } statfs;
struct {
- off_t start, len;
+ fd_t *fd;
+ int32_t flags;
+ } open;
- gf_boolean_t eager_lock_on;
- gf_boolean_t do_eager_unlock;
+ struct {
+ struct gf_flock user_flock;
+ struct gf_flock ret_flock;
+ unsigned char *locked_nodes;
+ int32_t cmd;
+ /*For lock healing only.*/
+ unsigned char *dom_locked_nodes;
+ int32_t *dom_lock_op_ret;
+ int32_t *dom_lock_op_errno;
+ struct gf_flock *getlk_rsp;
+ } lk;
+
+ /* inode read */
- char *basename;
- char *new_basename;
+ struct {
+ int32_t mask;
+ int last_index; /* index of the child we tried previously */
+ } access;
- loc_t parent_loc;
- loc_t new_parent_loc;
+ struct {
+ int last_index;
+ } stat;
- afr_transaction_type type;
+ struct {
+ int last_index;
+ } fstat;
- /* stub to resume on destruction
- of the transaction frame */
- call_stub_t *resume_stub;
+ struct {
+ size_t size;
+ int last_index;
+ } readlink;
- struct list_head owner_list;
- struct list_head wait_list;
+ struct {
+ char *name;
+ long xattr_len;
+ int last_index;
+ } getxattr;
- unsigned char *pre_op;
+ struct {
+ size_t size;
+ off_t offset;
+ int last_index;
+ uint32_t flags;
+ } readv;
- /* Changelog xattr dict for [f]xattrop*/
- dict_t **changelog_xdata;
- unsigned char *pre_op_sources;
+ /* dir read */
- /* @failed_subvols: subvolumes on which a pre-op or a
- FOP failed. */
- unsigned char *failed_subvols;
+ struct {
+ uint32_t *checksum;
+ int success_count;
+ int32_t op_ret;
+ int32_t op_errno;
+ } opendir;
- /* @dirtied: flag which indicates whether we set dirty flag
- in the OP. Typically true when we are performing operation
- on more than one subvol and optimistic changelog is disabled
+ struct {
+ int32_t op_ret;
+ int32_t op_errno;
+ size_t size;
+ off_t offset;
+ dict_t *dict;
+ int last_index;
+ gf_boolean_t failed;
+ } readdir;
+ /* inode write */
- A 'true' value set in @dirtied flag means an 'undirtying'
- has to be done in POST-OP phase.
- */
- gf_boolean_t dirtied;
+ struct {
+ struct iatt prebuf;
+ struct iatt postbuf;
+ } inode_wfop; // common structure for all inode-write-fops
- /* @inherited: flag which indicates that the dirty flags
- of the previous transaction were inherited
- */
- gf_boolean_t inherited;
+ struct {
+ struct iovec *vector;
+ struct iobref *iobref;
+ off_t offset;
+ int32_t op_ret;
+ int32_t count;
+ uint32_t flags;
+ } writev;
- /*
- @no_uninherit: flag which indicates that a pre_op_uninherit()
- must _not_ be attempted (and returned as failure) always. This
- flag is set when a hard pre-op is performed, but not accounted
- for it in fd_ctx->on_disk[]. Such transactions are "isolated"
- from the pre-op piggybacking entirely and therefore uninherit
- must not be attempted.
- */
- gf_boolean_t no_uninherit;
+ struct {
+ off_t offset;
+ } truncate;
- /* @uninherit_done:
- @uninherit_value:
+ struct {
+ off_t offset;
+ } ftruncate;
- The above pair variables make pre_op_uninherit() idempotent.
- Both are FALSE initially. The first call to pre_op_uninherit
- sets @uninherit_done to TRUE and the return value to
- @uninherit_value. Further calls will check for @uninherit_done
- to be TRUE and if so will simply return @uninherit_value.
- */
- gf_boolean_t uninherit_done;
- gf_boolean_t uninherit_value;
+ struct {
+ struct iatt in_buf;
+ int32_t valid;
+ } setattr;
- gf_boolean_t in_flight_sb; /* Indicator for occurrence of
- split-brain while in the middle of
- a txn. */
- int32_t in_flight_sb_errno; /* This is where the cause of the
- failure on the last good copy of
- the file is stored.
- */
+ struct {
+ struct iatt in_buf;
+ int32_t valid;
+ } fsetattr;
- /* @changelog_resume: function to be called after changlogging
- (either pre-op or post-op) is done
- */
- afr_changelog_resume_t changelog_resume;
+ struct {
+ dict_t *dict;
+ int32_t flags;
+ } setxattr;
- call_frame_t *main_frame; /*Fop frame*/
- call_frame_t *frame; /*Transaction frame*/
+ struct {
+ dict_t *dict;
+ int32_t flags;
+ } fsetxattr;
- int (*wind) (call_frame_t *frame, xlator_t *this, int subvol);
+ struct {
+ char *name;
+ } removexattr;
- int (*unwind) (call_frame_t *frame, xlator_t *this);
+ struct {
+ dict_t *xattr;
+ gf_xattrop_flags_t optype;
+ } xattrop;
- /* post-op hook */
- } transaction;
+ /* dir write */
- syncbarrier_t barrier;
+ struct {
+ inode_t *inode;
+ struct iatt buf;
+ struct iatt preparent;
+ struct iatt postparent;
+ struct iatt prenewparent;
+ struct iatt postnewparent;
+ } dir_fop; // common structure for all dir fops
- /* extra data for fops */
- dict_t *xdata_req;
- dict_t *xdata_rsp;
+ struct {
+ fd_t *fd;
+ dict_t *params;
+ int32_t flags;
+ mode_t mode;
+ } create;
- dict_t *xattr_rsp; /*for [f]xattrop*/
+ struct {
+ dict_t *params;
+ dev_t dev;
+ mode_t mode;
+ } mknod;
- mode_t umask;
- int xflag;
- gf_boolean_t do_discovery;
- struct afr_reply *replies;
+ struct {
+ dict_t *params;
+ int32_t mode;
+ } mkdir;
- /* For client side background heals. */
- struct list_head healer;
- call_frame_t *heal_frame;
+ struct {
+ dict_t *params;
+ char *linkpath;
+ } symlink;
- gf_boolean_t need_full_crawl;
- afr_fop_lock_state_t fop_lock_state;
+ struct {
+ off_t offset;
+ size_t len;
+ int32_t mode;
+ } fallocate;
- gf_boolean_t is_read_txn;
- afr_inode_ctx_t *inode_ctx;
-} afr_local_t;
+ struct {
+ off_t offset;
+ size_t len;
+ } discard;
+
+ struct {
+ off_t offset;
+ off_t len;
+ struct iatt prebuf;
+ struct iatt postbuf;
+ } zerofill;
+
+ struct {
+ char *volume;
+ int32_t cmd;
+ int32_t in_cmd;
+ struct gf_flock in_flock;
+ struct gf_flock flock;
+ void *xdata;
+ } inodelk;
+
+ struct {
+ char *volume;
+ char *basename;
+ void *xdata;
+ entrylk_cmd in_cmd;
+ entrylk_cmd cmd;
+ entrylk_type type;
+ } entrylk;
+
+ struct {
+ off_t offset;
+ gf_seek_what_t what;
+ } seek;
+
+ struct {
+ struct gf_lease user_lease;
+ struct gf_lease ret_lease;
+ unsigned char *locked_nodes;
+ } lease;
+
+ struct {
+ int flags;
+ } rmdir;
+
+ struct {
+ int32_t datasync;
+ } fsync;
+
+ struct {
+ uuid_t gfid_req;
+ gf_boolean_t needs_fresh_lookup;
+ } lookup;
+
+ } cont;
+
+ struct {
+ char *basename;
+ char *new_basename;
+
+ loc_t parent_loc;
+ loc_t new_parent_loc;
+
+ /* stub to resume on destruction
+ of the transaction frame */
+ call_stub_t *resume_stub;
+
+ struct list_head owner_list;
+ struct list_head wait_list;
+ unsigned char *pre_op;
+
+ /* Changelog xattr dict for [f]xattrop*/
+ dict_t **changelog_xdata;
+ unsigned char *pre_op_sources;
+
+ /* @failed_subvols: subvolumes on which a pre-op or a
+ FOP failed. */
+ unsigned char *failed_subvols;
+
+ call_frame_t *main_frame; /*Fop frame*/
+ call_frame_t *frame; /*Transaction frame*/
+
+ int (*wind)(call_frame_t *frame, xlator_t *this, int subvol);
+
+ int (*unwind)(call_frame_t *frame, xlator_t *this);
+
+ off_t start, len;
+
+ afr_transaction_type type;
+
+ int32_t in_flight_sb_errno; /* This is where the cause of the
+ failure on the last good copy of
+ the file is stored.
+ */
+
+ /* @changelog_resume: function to be called after changlogging
+ (either pre-op or post-op) is done
+ */
+ afr_changelog_resume_t changelog_resume;
+
+ gf_boolean_t eager_lock_on;
+ gf_boolean_t do_eager_unlock;
+
+ /* @dirtied: flag which indicates whether we set dirty flag
+ in the OP. Typically true when we are performing operation
+ on more than one subvol and optimistic changelog is disabled
+
+ A 'true' value set in @dirtied flag means an 'undirtying'
+ has to be done in POST-OP phase.
+ */
+ gf_boolean_t dirtied;
+
+ /* @inherited: flag which indicates that the dirty flags
+ of the previous transaction were inherited
+ */
+ gf_boolean_t inherited;
+
+ /*
+ @no_uninherit: flag which indicates that a pre_op_uninherit()
+ must _not_ be attempted (and returned as failure) always. This
+ flag is set when a hard pre-op is performed, but not accounted
+ for it in fd_ctx->on_disk[]. Such transactions are "isolated"
+ from the pre-op piggybacking entirely and therefore uninherit
+ must not be attempted.
+ */
+ gf_boolean_t no_uninherit;
+
+ gf_boolean_t in_flight_sb; /* Indicator for occurrence of
+ split-brain while in the middle of
+ a txn. */
+
+ /* @uninherit_done:
+ @uninherit_value:
+
+ The above pair variables make pre_op_uninherit() idempotent.
+ Both are FALSE initially. The first call to pre_op_uninherit
+ sets @uninherit_done to TRUE and the return value to
+ @uninherit_value. Further calls will check for @uninherit_done
+ to be TRUE and if so will simply return @uninherit_value.
+ */
+ gf_boolean_t uninherit_done;
+ gf_boolean_t uninherit_value;
+
+ gf_boolean_t disable_delayed_post_op;
+ } transaction;
+
+ syncbarrier_t barrier;
+
+ /* extra data for fops */
+ dict_t *xdata_req;
+ dict_t *xdata_rsp;
+
+ dict_t *xattr_rsp; /*for [f]xattrop*/
+
+ mode_t umask;
+ int xflag;
+ struct afr_reply *replies;
+
+ /* For client side background heals. */
+ struct list_head healer;
+ call_frame_t *heal_frame;
+
+ afr_inode_ctx_t *inode_ctx;
+
+ /*For thin-arbiter transactions.*/
+ int ta_failed_subvol;
+ int ta_event_gen;
+ struct list_head ta_waitq;
+ struct list_head ta_onwireq;
+ afr_ta_fop_state_t fop_state;
+ afr_fop_lock_state_t fop_lock_state;
+ gf_lkowner_t saved_lk_owner;
+ unsigned char read_txn_query_child;
+ unsigned char ta_child_up;
+ gf_boolean_t do_discovery;
+ gf_boolean_t need_full_crawl;
+ gf_boolean_t is_read_txn;
+ gf_boolean_t is_new_entry;
+} afr_local_t;
typedef struct afr_spbc_timeout {
- call_frame_t *frame;
- gf_boolean_t d_spb;
- gf_boolean_t m_spb;
- loc_t *loc;
- int spb_child_index;
+ call_frame_t *frame;
+ loc_t *loc;
+ int spb_child_index;
+ gf_boolean_t d_spb;
+ gf_boolean_t m_spb;
} afr_spbc_timeout_t;
typedef struct afr_spb_status {
- call_frame_t *frame;
- loc_t *loc;
+ call_frame_t *frame;
+ loc_t *loc;
} afr_spb_status_t;
typedef struct afr_empty_brick_args {
- call_frame_t *frame;
- loc_t loc;
- int empty_index;
- char *op_type;
+ call_frame_t *frame;
+ char *op_type;
+ loc_t loc;
+ int empty_index;
} afr_empty_brick_args_t;
typedef struct afr_read_subvol_args {
- ia_type_t ia_type;
- uuid_t gfid;
+ ia_type_t ia_type;
+ uuid_t gfid;
} afr_read_subvol_args_t;
typedef struct afr_granular_esh_args {
- fd_t *heal_fd;
- xlator_t *xl;
- call_frame_t *frame;
- gf_boolean_t mismatch; /* flag to represent occurrence of type/gfid
- mismatch */
+ fd_t *heal_fd;
+ xlator_t *xl;
+ call_frame_t *frame;
+ gf_boolean_t mismatch; /* flag to represent occurrence of type/gfid
+ mismatch */
} afr_granular_esh_args_t;
int
-afr_inode_get_readable (call_frame_t *frame, inode_t *inode, xlator_t *this,
- unsigned char *readable, int *event_p, int type);
+afr_inode_get_readable(call_frame_t *frame, inode_t *inode, xlator_t *this,
+ unsigned char *readable, int *event_p, int type);
int
-afr_inode_read_subvol_get (inode_t *inode, xlator_t *this,
- unsigned char *data_subvols,
- unsigned char *metadata_subvols,
- int *event_generation);
+afr_inode_read_subvol_get(inode_t *inode, xlator_t *this,
+ unsigned char *data_subvols,
+ unsigned char *metadata_subvols,
+ int *event_generation);
int
-__afr_inode_read_subvol_get (inode_t *inode, xlator_t *this,
- unsigned char *data_subvols,
- unsigned char *metadata_subvols,
- int *event_generation);
+__afr_inode_read_subvol_get(inode_t *inode, xlator_t *this,
+ unsigned char *data_subvols,
+ unsigned char *metadata_subvols,
+ int *event_generation);
int
-__afr_inode_read_subvol_set (inode_t *inode, xlator_t *this,
- unsigned char *data_subvols,
- unsigned char *metadata_subvol,
- int event_generation);
+__afr_inode_read_subvol_set(inode_t *inode, xlator_t *this,
+ unsigned char *data_subvols,
+ unsigned char *metadata_subvol,
+ int event_generation);
+int
+afr_inode_read_subvol_set(inode_t *inode, xlator_t *this,
+ unsigned char *data_subvols,
+ unsigned char *metadata_subvols,
+ int event_generation);
+
int
-afr_inode_read_subvol_set (inode_t *inode, xlator_t *this,
- unsigned char *data_subvols,
- unsigned char *metadata_subvols,
- int event_generation);
+__afr_inode_need_refresh_set(inode_t *inode, xlator_t *this);
int
-afr_inode_event_gen_reset (inode_t *inode, xlator_t *this);
+afr_inode_need_refresh_set(inode_t *inode, xlator_t *this);
int
-afr_read_subvol_select_by_policy (inode_t *inode, xlator_t *this,
- unsigned char *readable,
- afr_read_subvol_args_t *args);
+afr_read_subvol_select_by_policy(inode_t *inode, xlator_t *this,
+ unsigned char *readable,
+ afr_read_subvol_args_t *args);
int
-afr_inode_read_subvol_type_get (inode_t *inode, xlator_t *this,
- unsigned char *readable, int *event_p,
- int type);
+afr_inode_read_subvol_type_get(inode_t *inode, xlator_t *this,
+ unsigned char *readable, int *event_p, int type);
int
-afr_read_subvol_get (inode_t *inode, xlator_t *this, int *subvol_p,
- unsigned char *readables,
- int *event_p, afr_transaction_type type,
- afr_read_subvol_args_t *args);
+afr_read_subvol_get(inode_t *inode, xlator_t *this, int *subvol_p,
+ unsigned char *readables, int *event_p,
+ afr_transaction_type type, afr_read_subvol_args_t *args);
-#define afr_data_subvol_get(i, t, s, r, e, a) \
- afr_read_subvol_get(i, t, s, r, e, AFR_DATA_TRANSACTION, a)
+#define afr_data_subvol_get(i, t, s, r, e, a) \
+ afr_read_subvol_get(i, t, s, r, e, AFR_DATA_TRANSACTION, a)
-#define afr_metadata_subvol_get(i, t, s, r, e, a) \
- afr_read_subvol_get(i, t, s, r, e, AFR_METADATA_TRANSACTION, a)
+#define afr_metadata_subvol_get(i, t, s, r, e, a) \
+ afr_read_subvol_get(i, t, s, r, e, AFR_METADATA_TRANSACTION, a)
int
-afr_inode_refresh (call_frame_t *frame, xlator_t *this, inode_t *inode,
- uuid_t gfid, afr_inode_refresh_cbk_t cbk);
+afr_inode_refresh(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ uuid_t gfid, afr_inode_refresh_cbk_t cbk);
int32_t
-afr_notify (xlator_t *this, int32_t event, void *data, void *data2);
+afr_notify(xlator_t *this, int32_t event, void *data, void *data2);
int
-xattr_is_equal (dict_t *this, char *key1, data_t *value1, void *data);
+xattr_is_equal(dict_t *this, char *key1, data_t *value1, void *data);
int
-afr_init_entry_lockee (afr_entry_lockee_t *lockee, afr_local_t *local,
- loc_t *loc, char *basename, int child_count);
+afr_add_entry_lockee(afr_local_t *local, loc_t *loc, char *basename,
+ int child_count);
+
+int
+afr_add_inode_lockee(afr_local_t *local, int child_count);
void
-afr_entry_lockee_cleanup (afr_internal_lock_t *int_lock);
+afr_lockees_cleanup(afr_internal_lock_t *int_lock);
int
-afr_attempt_lock_recovery (xlator_t *this, int32_t child_index);
+afr_attempt_lock_recovery(xlator_t *this, int32_t child_index);
int
-afr_mark_locked_nodes (xlator_t *this, fd_t *fd,
- unsigned char *locked_nodes);
+afr_mark_locked_nodes(xlator_t *this, fd_t *fd, unsigned char *locked_nodes);
void
-afr_set_lk_owner (call_frame_t *frame, xlator_t *this, void *lk_owner);
+afr_set_lk_owner(call_frame_t *frame, xlator_t *this, void *lk_owner);
int
-afr_set_lock_number (call_frame_t *frame, xlator_t *this);
+afr_set_lock_number(call_frame_t *frame, xlator_t *this);
int32_t
-afr_unlock (call_frame_t *frame, xlator_t *this);
-
-int
-afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this);
+afr_unlock(call_frame_t *frame, xlator_t *this);
int
-afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this);
+afr_lock_nonblocking(call_frame_t *frame, xlator_t *this);
int
-afr_blocking_lock (call_frame_t *frame, xlator_t *this);
+afr_blocking_lock(call_frame_t *frame, xlator_t *this);
int
-afr_internal_lock_finish (call_frame_t *frame, xlator_t *this);
+afr_internal_lock_finish(call_frame_t *frame, xlator_t *this);
int
-__afr_fd_ctx_set (xlator_t *this, fd_t *fd);
+__afr_fd_ctx_set(xlator_t *this, fd_t *fd);
afr_fd_ctx_t *
-afr_fd_ctx_get (fd_t *fd, xlator_t *this);
+afr_fd_ctx_get(fd_t *fd, xlator_t *this);
int
-afr_build_parent_loc (loc_t *parent, loc_t *child, int32_t *op_errno);
+afr_build_parent_loc(loc_t *parent, loc_t *child, int32_t *op_errno);
int
-afr_locked_nodes_count (unsigned char *locked_nodes, int child_count);
+afr_locked_nodes_count(unsigned char *locked_nodes, int child_count);
int
-afr_replies_interpret (call_frame_t *frame, xlator_t *this, inode_t *inode,
- gf_boolean_t *start_heal);
+afr_replies_interpret(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ gf_boolean_t *start_heal);
void
-afr_local_replies_wipe (afr_local_t *local, afr_private_t *priv);
+afr_local_replies_wipe(afr_local_t *local, afr_private_t *priv);
void
-afr_local_cleanup (afr_local_t *local, xlator_t *this);
+afr_local_cleanup(afr_local_t *local, xlator_t *this);
int
-afr_frame_return (call_frame_t *frame);
+afr_frame_return(call_frame_t *frame);
int
-afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- fd_t *fd, dict_t *xdata);
+afr_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata);
void
-afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this);
+afr_local_transaction_cleanup(afr_local_t *local, xlator_t *this);
int
-afr_cleanup_fd_ctx (xlator_t *this, fd_t *fd);
-
-#define AFR_STACK_UNWIND(fop, frame, op_ret, op_errno, params ...)\
- do { \
- afr_local_t *__local = NULL; \
- xlator_t *__this = NULL; \
- int32_t __op_ret = 0; \
- int32_t __op_errno = 0; \
- \
- __op_ret = op_ret; \
- __op_errno = op_errno; \
- if (frame) { \
- __local = frame->local; \
- __this = frame->this; \
- afr_handle_inconsistent_fop (frame, &__op_ret,\
- &__op_errno);\
- if (__local && __local->is_read_txn) \
- afr_pending_read_decrement (__this->private, __local->read_subvol); \
- frame->local = NULL; \
- } \
- \
- STACK_UNWIND_STRICT (fop, frame, __op_ret, \
- __op_errno, params); \
- if (__local) { \
- afr_local_cleanup (__local, __this); \
- mem_put (__local); \
- } \
- } while (0)
-
-#define AFR_STACK_DESTROY(frame) \
- do { \
- afr_local_t *__local = NULL; \
- xlator_t *__this = NULL; \
- __local = frame->local; \
- __this = frame->this; \
- frame->local = NULL; \
- STACK_DESTROY (frame->root); \
- if (__local) { \
- afr_local_cleanup (__local, __this); \
- mem_put (__local); \
- } \
- } while (0);
-
-#define AFR_FRAME_INIT(frame, op_errno) \
- ({frame->local = mem_get0 (THIS->local_pool); \
- if (afr_local_init (frame->local, THIS->private, &op_errno)) { \
- afr_local_cleanup (frame->local, THIS); \
- mem_put (frame->local); \
- frame->local = NULL; }; \
- frame->local; })
-
-#define AFR_STACK_RESET(frame) \
- do { \
- afr_local_t *__local = NULL; \
- xlator_t *__this = NULL; \
- __local = frame->local; \
- __this = frame->this; \
- frame->local = NULL; \
- int __opr; \
- STACK_RESET (frame->root); \
- if (__local) { \
- afr_local_cleanup (__local, __this); \
- mem_put (__local); \
- } \
- AFR_FRAME_INIT (frame, __opr); \
- } while (0)
+afr_cleanup_fd_ctx(xlator_t *this, fd_t *fd);
+
+#define AFR_STACK_UNWIND(fop, frame, op_ret, op_errno, params...) \
+ do { \
+ afr_local_t *__local = NULL; \
+ xlator_t *__this = NULL; \
+ int32_t __op_ret = 0; \
+ int32_t __op_errno = 0; \
+ \
+ __op_ret = op_ret; \
+ __op_errno = op_errno; \
+ if (frame) { \
+ __local = frame->local; \
+ __this = frame->this; \
+ afr_handle_inconsistent_fop(frame, &__op_ret, &__op_errno); \
+ if (__local && __local->is_read_txn) \
+ afr_pending_read_decrement(__this->private, \
+ __local->read_subvol); \
+ if (__local && __local->xdata_req && \
+ afr_is_lock_mode_mandatory(__local->xdata_req)) \
+ afr_dom_lock_release(frame); \
+ frame->local = NULL; \
+ } \
+ \
+ STACK_UNWIND_STRICT(fop, frame, __op_ret, __op_errno, params); \
+ if (__local) { \
+ afr_local_cleanup(__local, __this); \
+ mem_put(__local); \
+ } \
+ } while (0)
+
+#define AFR_STACK_DESTROY(frame) \
+ do { \
+ afr_local_t *__local = NULL; \
+ xlator_t *__this = NULL; \
+ __local = frame->local; \
+ __this = frame->this; \
+ frame->local = NULL; \
+ STACK_DESTROY(frame->root); \
+ if (__local) { \
+ afr_local_cleanup(__local, __this); \
+ mem_put(__local); \
+ } \
+ } while (0);
+
+#define AFR_FRAME_INIT(frame, op_errno) \
+ ({ \
+ frame->local = mem_get0(THIS->local_pool); \
+ if (afr_local_init(frame->local, frame->this->private, &op_errno)) { \
+ afr_local_cleanup(frame->local, frame->this); \
+ mem_put(frame->local); \
+ frame->local = NULL; \
+ }; \
+ frame->local; \
+ })
+
+#define AFR_STACK_RESET(frame) \
+ do { \
+ afr_local_t *__local = NULL; \
+ xlator_t *__this = NULL; \
+ __local = frame->local; \
+ __this = frame->this; \
+ frame->local = NULL; \
+ int __opr; \
+ STACK_RESET(frame->root); \
+ if (__local) { \
+ afr_local_cleanup(__local, __this); \
+ mem_put(__local); \
+ } \
+ AFR_FRAME_INIT(frame, __opr); \
+ } while (0)
/* allocate and return a string that is the basename of argument */
static inline char *
-AFR_BASENAME (const char *str)
+AFR_BASENAME(const char *str)
{
- char *__tmp_str = NULL;
- char *__basename_str = NULL;
- __tmp_str = gf_strdup (str);
- __basename_str = gf_strdup (basename (__tmp_str));
- GF_FREE (__tmp_str);
- return __basename_str;
+ char *__tmp_str = NULL;
+ char *__basename_str = NULL;
+ __tmp_str = gf_strdup(str);
+ __basename_str = gf_strdup(basename(__tmp_str));
+ GF_FREE(__tmp_str);
+ return __basename_str;
}
call_frame_t *
-afr_copy_frame (call_frame_t *base);
+afr_copy_frame(call_frame_t *base);
int
-afr_transaction_local_init (afr_local_t *local, xlator_t *this);
+afr_transaction_local_init(afr_local_t *local, xlator_t *this);
int32_t
-afr_marker_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name,afr_local_t *local, afr_private_t *priv );
+afr_marker_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, afr_local_t *local, afr_private_t *priv);
int
-afr_local_init (afr_local_t *local, afr_private_t *priv, int32_t *op_errno);
+afr_local_init(afr_local_t *local, afr_private_t *priv, int32_t *op_errno);
int
-afr_internal_lock_init (afr_internal_lock_t *lk, size_t child_count);
+afr_internal_lock_init(afr_internal_lock_t *lk, size_t child_count);
int
-afr_higher_errno (int32_t old_errno, int32_t new_errno);
+afr_higher_errno(int32_t old_errno, int32_t new_errno);
int
-afr_final_errno (afr_local_t *local, afr_private_t *priv);
+afr_final_errno(afr_local_t *local, afr_private_t *priv);
int
-afr_xattr_req_prepare (xlator_t *this, dict_t *xattr_req);
+afr_xattr_req_prepare(xlator_t *this, dict_t *xattr_req);
void
-afr_fix_open (fd_t *fd, xlator_t *this);
+afr_fix_open(fd_t *fd, xlator_t *this);
afr_fd_ctx_t *
-afr_fd_ctx_get (fd_t *fd, xlator_t *this);
+afr_fd_ctx_get(fd_t *fd, xlator_t *this);
void
-afr_set_low_priority (call_frame_t *frame);
+afr_set_low_priority(call_frame_t *frame);
int
-afr_child_fd_ctx_set (xlator_t *this, fd_t *fd, int32_t child,
- int flags);
+afr_child_fd_ctx_set(xlator_t *this, fd_t *fd, int32_t child, int flags);
void
-afr_matrix_cleanup (int32_t **pending, unsigned int m);
+afr_matrix_cleanup(int32_t **pending, unsigned int m);
-int32_t**
-afr_matrix_create (unsigned int m, unsigned int n);
+int32_t **
+afr_matrix_create(unsigned int m, unsigned int n);
-int**
-afr_mark_pending_changelog (afr_private_t *priv, unsigned char *pending,
- dict_t *xattr, ia_type_t iat);
+int **
+afr_mark_pending_changelog(afr_private_t *priv, unsigned char *pending,
+ dict_t *xattr, ia_type_t iat);
void
-afr_filter_xattrs (dict_t *xattr);
+afr_filter_xattrs(dict_t *xattr);
/*
* Special value indicating we should use the "auto" quorum method instead of
@@ -1109,124 +1245,179 @@ afr_filter_xattrs (dict_t *xattr);
#define AFR_QUORUM_AUTO INT_MAX
int
-afr_fd_report_unstable_write (xlator_t *this, afr_local_t *local);
+afr_fd_report_unstable_write(xlator_t *this, afr_local_t *local);
gf_boolean_t
-afr_fd_has_witnessed_unstable_write (xlator_t *this, inode_t *inode);
+afr_fd_has_witnessed_unstable_write(xlator_t *this, inode_t *inode);
void
-afr_reply_wipe (struct afr_reply *reply);
+afr_reply_wipe(struct afr_reply *reply);
void
-afr_replies_wipe (struct afr_reply *replies, int count);
+afr_replies_wipe(struct afr_reply *replies, int count);
gf_boolean_t
-afr_xattrs_are_equal (dict_t *dict1, dict_t *dict2);
+afr_xattrs_are_equal(dict_t *dict1, dict_t *dict2);
gf_boolean_t
-afr_is_xattr_ignorable (char *key);
+afr_is_xattr_ignorable(char *key);
int
-afr_get_heal_info (call_frame_t *frame, xlator_t *this, loc_t *loc);
+afr_get_heal_info(call_frame_t *frame, xlator_t *this, loc_t *loc);
int
afr_heal_splitbrain_file(call_frame_t *frame, xlator_t *this, loc_t *loc);
int
-afr_get_split_brain_status (void *opaque);
+afr_get_split_brain_status(void *opaque);
int
-afr_get_split_brain_status_cbk (int ret, call_frame_t *frame, void *opaque);
+afr_get_split_brain_status_cbk(int ret, call_frame_t *frame, void *opaque);
int
-afr_inode_split_brain_choice_set (inode_t *inode, xlator_t *this,
- int spb_choice);
+afr_inode_split_brain_choice_set(inode_t *inode, xlator_t *this,
+ int spb_choice);
int
-afr_inode_split_brain_choice_get (inode_t *inode, xlator_t *this,
- int *spb_choice);
+afr_split_brain_read_subvol_get(inode_t *inode, xlator_t *this,
+ call_frame_t *frame, int *spb_subvol);
int
-afr_get_child_index_from_name (xlator_t *this, char *name);
+afr_get_child_index_from_name(xlator_t *this, char *name);
int
-afr_is_split_brain (call_frame_t *frame, xlator_t *this, inode_t *inode,
- uuid_t gfid, gf_boolean_t *d_spb, gf_boolean_t *m_spb);
+afr_is_split_brain(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ uuid_t gfid, gf_boolean_t *d_spb, gf_boolean_t *m_spb);
int
-afr_spb_choice_timeout_cancel (xlator_t *this, inode_t *inode);
+afr_spb_choice_timeout_cancel(xlator_t *this, inode_t *inode);
int
-afr_set_split_brain_choice (int ret, call_frame_t *frame, void *opaque);
+afr_set_split_brain_choice(int ret, call_frame_t *frame, void *opaque);
gf_boolean_t
-afr_get_need_heal (xlator_t *this);
+afr_get_need_heal(xlator_t *this);
void
-afr_set_need_heal (xlator_t *this, afr_local_t *local);
+afr_set_need_heal(xlator_t *this, afr_local_t *local);
int
-afr_selfheal_data_open (xlator_t *this, inode_t *inode, fd_t **fd);
+afr_selfheal_data_open(xlator_t *this, inode_t *inode, fd_t **fd);
int
-afr_get_msg_id (char *op_type);
+afr_get_msg_id(char *op_type);
int
-afr_set_in_flight_sb_status (xlator_t *this, call_frame_t *frame,
- inode_t *inode);
+afr_set_in_flight_sb_status(xlator_t *this, call_frame_t *frame,
+ inode_t *inode);
int32_t
-afr_quorum_errno (afr_private_t *priv);
+afr_quorum_errno(afr_private_t *priv);
gf_boolean_t
-afr_is_consistent_io_possible (afr_local_t *local, afr_private_t *priv,
- int32_t *op_errno);
+afr_is_consistent_io_possible(afr_local_t *local, afr_private_t *priv,
+ int32_t *op_errno);
void
-afr_handle_inconsistent_fop (call_frame_t *frame, int32_t *op_ret,
- int32_t *op_errno);
+afr_handle_inconsistent_fop(call_frame_t *frame, int32_t *op_ret,
+ int32_t *op_errno);
void
-afr_inode_write_fill (call_frame_t *frame, xlator_t *this, int child_index,
+afr_inode_write_fill(call_frame_t *frame, xlator_t *this, int child_index,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
struct iatt *postbuf, dict_t *xdata);
void
-afr_process_post_writev (call_frame_t *frame, xlator_t *this);
+afr_process_post_writev(call_frame_t *frame, xlator_t *this);
void
-afr_writev_unwind (call_frame_t *frame, xlator_t *this);
+afr_writev_unwind(call_frame_t *frame, xlator_t *this);
void
-afr_writev_copy_outvars (call_frame_t *src_frame, call_frame_t *dst_frame);
+afr_writev_copy_outvars(call_frame_t *src_frame, call_frame_t *dst_frame);
void
-afr_update_uninodelk (afr_local_t *local, afr_internal_lock_t *int_lock,
- int32_t child_index);
-int
-afr_is_inodelk_transaction(afr_transaction_type type);
-
+afr_update_uninodelk(afr_local_t *local, afr_internal_lock_t *int_lock,
+ int32_t child_index);
afr_fd_ctx_t *
-__afr_fd_ctx_get (fd_t *fd, xlator_t *this);
+__afr_fd_ctx_get(fd_t *fd, xlator_t *this);
gf_boolean_t
-afr_is_inode_refresh_reqd (inode_t *inode, xlator_t *this,
- int event_gen1, int event_gen2);
+afr_is_inode_refresh_reqd(inode_t *inode, xlator_t *this, int event_gen1,
+ int event_gen2);
int
-afr_serialize_xattrs_with_delimiter (call_frame_t *frame, xlator_t *this,
- char *buf, const char *default_str,
- int32_t *serz_len, char delimiter);
+afr_serialize_xattrs_with_delimiter(call_frame_t *frame, xlator_t *this,
+ char *buf, const char *default_str,
+ int32_t *serz_len, char delimiter);
gf_boolean_t
-afr_is_symmetric_error (call_frame_t *frame, xlator_t *this);
+afr_is_symmetric_error(call_frame_t *frame, xlator_t *this);
int
-__afr_inode_ctx_get (xlator_t *this, inode_t *inode, afr_inode_ctx_t **ctx);
+__afr_inode_ctx_get(xlator_t *this, inode_t *inode, afr_inode_ctx_t **ctx);
uint64_t
-afr_write_subvol_get (call_frame_t *frame, xlator_t *this);
+afr_write_subvol_get(call_frame_t *frame, xlator_t *this);
+
+int
+afr_write_subvol_set(call_frame_t *frame, xlator_t *this);
int
-afr_write_subvol_set (call_frame_t *frame, xlator_t *this);
+afr_write_subvol_reset(call_frame_t *frame, xlator_t *this);
int
-afr_write_subvol_reset (call_frame_t *frame, xlator_t *this);
+afr_set_inode_local(xlator_t *this, afr_local_t *local, inode_t *inode);
int
-afr_set_inode_local (xlator_t *this, afr_local_t *local, inode_t *inode);
+afr_fill_ta_loc(xlator_t *this, loc_t *loc, gf_boolean_t is_gfid_based_fop);
+
+int
+afr_ta_post_op_lock(xlator_t *this, loc_t *loc);
+
+int
+afr_ta_post_op_unlock(xlator_t *this, loc_t *loc);
+
+gf_boolean_t
+afr_is_pending_set(xlator_t *this, dict_t *xdata, int type);
+
+int
+__afr_get_up_children_count(afr_private_t *priv);
+
+call_frame_t *
+afr_ta_frame_create(xlator_t *this);
+
+gf_boolean_t
+afr_ta_has_quorum(afr_private_t *priv, afr_local_t *local);
+
+void
+afr_ta_lock_release_synctask(xlator_t *this);
+
+void
+afr_ta_locked_priv_invalidate(afr_private_t *priv);
+
+gf_boolean_t
+afr_lookup_has_quorum(call_frame_t *frame,
+ const unsigned int up_children_count);
+
+void
+afr_mark_new_entry_changelog(call_frame_t *frame, xlator_t *this);
+
+void
+afr_handle_replies_quorum(call_frame_t *frame, xlator_t *this);
+
+gf_boolean_t
+afr_ta_dict_contains_pending_xattr(dict_t *dict, afr_private_t *priv,
+ int child);
+
+void
+afr_selfheal_childup(xlator_t *this, afr_private_t *priv);
+
+gf_boolean_t
+afr_is_lock_mode_mandatory(dict_t *xdata);
+
+void
+afr_dom_lock_release(call_frame_t *frame);
+
+void
+afr_fill_success_replies(afr_local_t *local, afr_private_t *priv,
+ unsigned char *replies);
+
+gf_boolean_t
+afr_is_private_directory(afr_private_t *priv, uuid_t pargfid, const char *name,
+ pid_t pid);
#endif /* __AFR_H__ */
diff --git a/xlators/cluster/dht/src/Makefile.am b/xlators/cluster/dht/src/Makefile.am
index 525a214c24a..56f1f2ad7c8 100644
--- a/xlators/cluster/dht/src/Makefile.am
+++ b/xlators/cluster/dht/src/Makefile.am
@@ -1,7 +1,4 @@
xlator_LTLIBRARIES = dht.la nufa.la switch.la
-if BUILD_GFDB
- xlator_LTLIBRARIES += tier.la
-endif
AM_CFLAGS = -Wall $(GF_CFLAGS)
@@ -16,41 +13,28 @@ dht_la_SOURCES = $(dht_common_source) dht.c
nufa_la_SOURCES = $(dht_common_source) nufa.c
switch_la_SOURCES = $(dht_common_source) switch.c
-tier_la_SOURCES = $(dht_common_source) tier.c tier-common.c
-dht_la_LDFLAGS = -module -avoid-version -export-symbols \
- $(top_srcdir)/xlators/cluster/dht/src/dht.sym
+dht_la_LDFLAGS = -module $(GF_XLATOR_DEFAULT_LDFLAGS)
dht_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-nufa_la_LDFLAGS = -module -avoid-version -export-symbols \
- $(top_srcdir)/xlators/cluster/dht/src/nufa.sym
+nufa_la_LDFLAGS = -module $(GF_XLATOR_DEFAULT_LDFLAGS)
nufa_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-switch_la_LDFLAGS = -module -avoid-version -export-symbols \
- $(top_srcdir)/xlators/cluster/dht/src/switch.sym
+switch_la_LDFLAGS = -module $(GF_XLATOR_DEFAULT_LDFLAGS)
switch_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-tier_la_LDFLAGS = -module -avoid-version -export-symbols \
- $(top_srcdir)/xlators/cluster/dht/src/tier.sym
-tier_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-
noinst_HEADERS = dht-common.h dht-mem-types.h dht-messages.h \
- dht-lock.h tier-common.h tier.h \
- $(top_builddir)/xlators/lib/src/libxlator.h
+ dht-lock.h $(top_builddir)/xlators/lib/src/libxlator.h
AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
- -I$(top_srcdir)/libglusterfs/src/gfdb \
-I$(top_srcdir)/rpc/xdr/src -I$(top_builddir)/rpc/xdr/src \
-I$(top_srcdir)/rpc/rpc-lib/src \
-I$(top_srcdir)/xlators/lib/src \
-DDATADIR=\"$(localstatedir)\" \
- -DLIBDIR=\"$(libdir)\" \
- -DLIBGFDB_VERSION=\"$(LIBGFDB_VERSION)\"
+ -DLIBDIR=\"$(libdir)\"
CLEANFILES =
-EXTRA_DIST = dht.sym nufa.sym switch.sym tier.sym
-
uninstall-local:
rm -f $(DESTDIR)$(xlatordir)/distribute.so
diff --git a/xlators/cluster/dht/src/dht-common.c b/xlators/cluster/dht/src/dht-common.c
index 47521d160ee..8ba0cc4c732 100644
--- a/xlators/cluster/dht/src/dht-common.c
+++ b/xlators/cluster/dht/src/dht-common.c
@@ -8,262 +8,209 @@
cases as published by the Free Software Foundation.
*/
-
/* TODO: add NS locking */
-#include "glusterfs.h"
-#include "xlator.h"
#include "libxlator.h"
#include "dht-common.h"
#include "dht-lock.h"
-#include "defaults.h"
-#include "byte-order.h"
-#include "quota-common-utils.h"
-#include "upcall-utils.h"
+#include <glusterfs/byte-order.h>
+#include <glusterfs/quota-common-utils.h>
+#include <glusterfs/upcall-utils.h>
+#include "glusterfs/compat-errno.h" // for ENODATA on BSD
+#include <glusterfs/common-utils.h>
#include <sys/time.h>
#include <libgen.h>
#include <signal.h>
-int run_defrag = 0;
-
-
-
-int dht_link2 (xlator_t *this, xlator_t *dst_node, call_frame_t *frame,
- int ret);
-
-int
-dht_removexattr2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame,
- int ret);
-
-int
-dht_setxattr2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame,
- int ret);
-
+static int
+dht_rmdir_readdirp_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, gf_dirent_t *entries,
+ dict_t *xdata);
-int
-dht_rmdir_readdirp_do (call_frame_t *readdirp_frame, xlator_t *this);
+static int
+dht_link2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret);
+static int
+dht_set_dir_xattr_req(xlator_t *this, loc_t *loc, dict_t *xattr_req);
-int
-dht_common_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata);
+static int
+dht_lookup_everywhere_done(call_frame_t *frame, xlator_t *this);
+static int
+dht_common_mark_mdsxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata);
-/* Sets the blocks and size values to fixed values. This is to be called
- * only for dirs. The caller is responsible for checking the type
- */
-int32_t dht_set_fixed_dir_stat (struct iatt *stat)
-{
- if (stat) {
- stat->ia_blocks = DHT_DIR_STAT_BLOCKS;
- stat->ia_size = DHT_DIR_STAT_SIZE;
- return 0;
- }
- return -1;
-}
+static int
+dht_rmdir_unlock(call_frame_t *frame, xlator_t *this);
+static const char *dht_dbg_vxattrs[] = {DHT_DBG_HASHED_SUBVOL_PATTERN, NULL};
-/* Set both DHT_IATT_IN_XDATA_KEY and DHT_MODE_IN_XDATA_KEY
- * Use DHT_MODE_IN_XDATA_KEY if available. Else fall back to
- * DHT_IATT_IN_XDATA_KEY
- */
-int dht_request_iatt_in_xdata (xlator_t *this, dict_t *xattr_req)
+/* Check the xdata to make sure EBADF has been set by client xlator */
+int32_t
+dht_check_remote_fd_failed_error(dht_local_t *local, int op_ret, int op_errno)
{
- int ret = -1;
-
- ret = dict_set_int8 (xattr_req, DHT_MODE_IN_XDATA_KEY, 1);
- ret = dict_set_int8 (xattr_req, DHT_IATT_IN_XDATA_KEY, 1);
-
- /* At least one call succeeded */
- return ret;
+ if (op_ret == -1 && (op_errno == EBADF || op_errno == EBADFD) &&
+ !(local->fd_checked)) {
+ return 1;
+ }
+ return 0;
}
-
-/* Get both DHT_IATT_IN_XDATA_KEY and DHT_MODE_IN_XDATA_KEY
- * Use DHT_MODE_IN_XDATA_KEY if available, else fall back to
- * DHT_IATT_IN_XDATA_KEY
- * This will return a dummy iatt with only the mode and type set
+/* Sets the blocks and size values to fixed values. This is to be called
+ * only for dirs. The caller is responsible for checking the type
*/
-int dht_read_iatt_from_xdata (xlator_t *this, dict_t *xdata,
- struct iatt *stbuf)
+int32_t
+dht_set_fixed_dir_stat(struct iatt *stat)
{
- int ret = -1;
- int32_t mode = 0;
-
- ret = dict_get_int32 (xdata, DHT_MODE_IN_XDATA_KEY, &mode);
-
- if (ret) {
- ret = dict_get_bin (xdata, DHT_IATT_IN_XDATA_KEY,
- (void **)&stbuf);
- } else {
- stbuf->ia_prot = ia_prot_from_st_mode (mode);
- stbuf->ia_type = ia_type_from_st_mode (mode);
- }
-
- return ret;
+ if (stat) {
+ stat->ia_blocks = DHT_DIR_STAT_BLOCKS;
+ stat->ia_size = DHT_DIR_STAT_SIZE;
+ return 0;
+ }
+ return -1;
}
-
-
-int
-dht_rmdir_unlock (call_frame_t *frame, xlator_t *this);
-
-char *xattrs_to_heal[] = {
- "user.",
- POSIX_ACL_ACCESS_XATTR,
- POSIX_ACL_DEFAULT_XATTR,
- QUOTA_LIMIT_KEY,
- QUOTA_LIMIT_OBJECTS_KEY,
- GF_SELINUX_XATTR_KEY,
- NULL
-};
-
/* Return true if key exists in array
-*/
+ */
static gf_boolean_t
-dht_match_xattr (const char *key)
+dht_match_xattr(const char *key)
{
- return gf_get_index_by_elem (xattrs_to_heal, (char *)key) >= 0;
-}
+ char **xattrs_to_heal = get_xattrs_to_heal();
-int
-dht_aggregate_quota_xattr (dict_t *dst, char *key, data_t *value)
-{
- int ret = -1;
- quota_meta_t *meta_dst = NULL;
- quota_meta_t *meta_src = NULL;
- int64_t *size = NULL;
- int64_t dst_dir_count = 0;
- int64_t src_dir_count = 0;
-
- if (value == NULL) {
- gf_msg ("dht", GF_LOG_WARNING, 0,
- DHT_MSG_DATA_NULL, "data value is NULL");
- ret = -1;
- goto out;
- }
+ return gf_get_index_by_elem(xattrs_to_heal, (char *)key) >= 0;
+}
- ret = dict_get_bin (dst, key, (void **)&meta_dst);
+static int
+dht_aggregate_quota_xattr(dict_t *dst, char *key, data_t *value)
+{
+ int ret = -1;
+ quota_meta_t *meta_dst = NULL;
+ quota_meta_t *meta_src = NULL;
+ int64_t *size = NULL;
+ int64_t dst_dir_count = 0;
+ int64_t src_dir_count = 0;
+
+ if (value == NULL) {
+ gf_msg("dht", GF_LOG_WARNING, 0, DHT_MSG_DATA_NULL,
+ "data value is NULL");
+ ret = -1;
+ goto out;
+ }
+
+ ret = dict_get_bin(dst, key, (void **)&meta_dst);
+ if (ret < 0) {
+ meta_dst = GF_CALLOC(1, sizeof(quota_meta_t), gf_common_quota_meta_t);
+ if (meta_dst == NULL) {
+ gf_msg("dht", GF_LOG_WARNING, ENOMEM, DHT_MSG_NO_MEMORY,
+ "Memory allocation failed");
+ ret = -1;
+ goto out;
+ }
+ ret = dict_set_bin(dst, key, meta_dst, sizeof(quota_meta_t));
if (ret < 0) {
- meta_dst = GF_CALLOC (1, sizeof (quota_meta_t),
- gf_common_quota_meta_t);
- if (meta_dst == NULL) {
- gf_msg ("dht", GF_LOG_WARNING, ENOMEM,
- DHT_MSG_NO_MEMORY,
- "Memory allocation failed");
- ret = -1;
- goto out;
- }
- ret = dict_set_bin (dst, key, meta_dst,
- sizeof (quota_meta_t));
- if (ret < 0) {
- gf_msg ("dht", GF_LOG_WARNING, EINVAL,
- DHT_MSG_DICT_SET_FAILED,
- "dht aggregate dict set failed");
- GF_FREE (meta_dst);
- ret = -1;
- goto out;
- }
+ gf_msg("dht", GF_LOG_WARNING, EINVAL, DHT_MSG_DICT_SET_FAILED,
+ "dht aggregate dict set failed");
+ GF_FREE(meta_dst);
+ ret = -1;
+ goto out;
}
+ }
- if (value->len > sizeof (int64_t)) {
- meta_src = data_to_bin (value);
+ if (value->len > sizeof(int64_t)) {
+ meta_src = data_to_bin(value);
- meta_dst->size = hton64 (ntoh64 (meta_dst->size) +
- ntoh64 (meta_src->size));
- meta_dst->file_count = hton64 (ntoh64 (meta_dst->file_count) +
- ntoh64 (meta_src->file_count));
+ meta_dst->size = hton64(ntoh64(meta_dst->size) +
+ ntoh64(meta_src->size));
+ meta_dst->file_count = hton64(ntoh64(meta_dst->file_count) +
+ ntoh64(meta_src->file_count));
- if (value->len > (2 * sizeof (int64_t))) {
- dst_dir_count = ntoh64 (meta_dst->dir_count);
- src_dir_count = ntoh64 (meta_src->dir_count);
+ if (value->len > (2 * sizeof(int64_t))) {
+ dst_dir_count = ntoh64(meta_dst->dir_count);
+ src_dir_count = ntoh64(meta_src->dir_count);
- if (src_dir_count > dst_dir_count)
- meta_dst->dir_count = meta_src->dir_count;
- } else {
- meta_dst->dir_count = 0;
- }
+ if (src_dir_count > dst_dir_count)
+ meta_dst->dir_count = meta_src->dir_count;
} else {
- size = data_to_bin (value);
- meta_dst->size = hton64 (ntoh64 (meta_dst->size) +
- ntoh64 (*size));
+ meta_dst->dir_count = 0;
}
+ } else {
+ size = data_to_bin(value);
+ meta_dst->size = hton64(ntoh64(meta_dst->size) + ntoh64(*size));
+ }
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
-
-int add_opt(char **optsp, const char *opt)
-{
- char *newopts = NULL;
- unsigned oldsize = 0;
- unsigned newsize = 0;
-
- if (*optsp == NULL)
- newopts = gf_strdup (opt);
- else {
- oldsize = strlen (*optsp);
- newsize = oldsize + 1 + strlen (opt) + 1;
- newopts = GF_REALLOC (*optsp, newsize);
- if (newopts)
- sprintf (newopts + oldsize, ",%s", opt);
- }
- if (newopts == NULL) {
- gf_msg ("dht", GF_LOG_WARNING, 0,
- DHT_MSG_NO_MEMORY,
- "Error to add choices in buffer in add_opt");
- return -1;
- }
- *optsp = newopts;
- return 0;
+static int
+add_opt(char **optsp, const char *opt)
+{
+ char *newopts = NULL;
+ unsigned oldsize = 0;
+ unsigned newsize = 0;
+
+ if (*optsp == NULL)
+ newopts = gf_strdup(opt);
+ else {
+ oldsize = strlen(*optsp);
+ newsize = oldsize + 1 + strlen(opt) + 1;
+ newopts = GF_REALLOC(*optsp, newsize);
+ if (newopts)
+ sprintf(newopts + oldsize, ",%s", opt);
+ }
+ if (newopts == NULL) {
+ gf_msg("dht", GF_LOG_WARNING, 0, DHT_MSG_NO_MEMORY,
+ "Error to add choices in buffer in add_opt");
+ return -1;
+ }
+ *optsp = newopts;
+ return 0;
}
/* Return Choice list from Split brain status */
static char *
-getChoices (const char *value)
+getChoices(const char *value)
{
- int i = 0;
- char *ptr = NULL;
- char *tok = NULL;
- char *result = NULL;
- char *newval = NULL;
+ int i = 0;
+ char *ptr = NULL;
+ char *tok = NULL;
+ char *result = NULL;
+ char *newval = NULL;
- ptr = strstr (value, "Choices:");
- if (!ptr) {
- result = ptr;
- goto out;
- }
+ ptr = strstr(value, "Choices:");
+ if (!ptr) {
+ result = ptr;
+ goto out;
+ }
- newval = gf_strdup (ptr);
- if (!newval) {
- result = newval;
- goto out;
- }
+ newval = gf_strdup(ptr);
+ if (!newval) {
+ result = newval;
+ goto out;
+ }
- tok = strtok (newval, ":");
- if (!tok) {
- result = tok;
- goto out;
- }
+ tok = strtok(newval, ":");
+ if (!tok) {
+ result = tok;
+ goto out;
+ }
- while (tok) {
- i++;
- if (i == 2)
- break;
- tok = strtok (NULL, ":");
- }
+ while (tok) {
+ i++;
+ if (i == 2)
+ break;
+ tok = strtok(NULL, ":");
+ }
- result = gf_strdup (tok);
+ result = gf_strdup(tok);
out:
- if (newval)
- GF_FREE (newval);
+ if (newval)
+ GF_FREE(newval);
- return result;
+ return result;
}
/* This function prepare a list of choices for key
@@ -275,245 +222,227 @@ out:
*/
-int
-dht_aggregate_split_brain_xattr (dict_t *dst, char *key, data_t *value)
-{
+static int
+dht_aggregate_split_brain_xattr(dict_t *dst, char *key, data_t *value)
+{
+ int ret = 0;
+ char *oldvalue = NULL;
+ char *old_choice = NULL;
+ char *new_choice = NULL;
+ char *full_choice = NULL;
+ char *status = NULL;
+
+ if (value == NULL) {
+ gf_msg("dht", GF_LOG_WARNING, 0, DHT_MSG_DATA_NULL,
+ "GF_AFR_SBRAIN_STATUS value is NULL");
+ ret = -1;
+ goto out;
+ }
- int ret = 0;
- char *oldvalue = NULL;
- char *old_choice = NULL;
- char *new_choice = NULL;
- char *full_choice = NULL;
- char *status = NULL;
+ ret = dict_get_str(dst, key, &oldvalue);
+ if (ret)
+ goto out;
+
+ /* skip code that is irrelevant if !oldvalue */
+ if (!oldvalue)
+ goto out;
- if (value == NULL) {
- gf_msg ("dht", GF_LOG_WARNING, 0,
- DHT_MSG_DATA_NULL,
- "GF_AFR_SBRAIN_STATUS value is NULL");
+ if (strstr(oldvalue, "not")) {
+ gf_msg_debug("dht", 0, "Need to update split-brain status in dict");
+ ret = -1;
+ goto out;
+ }
+ if (strstr(oldvalue, "metadata-split-brain:yes") &&
+ (strstr(oldvalue, "data-split-brain:no"))) {
+ if (strstr(value->data, "not")) {
+ gf_msg_debug("dht", 0, "No need to update split-brain status");
+ ret = 0;
+ goto out;
+ }
+ if (strstr(value->data, "yes") &&
+ (strncmp(oldvalue, value->data, strlen(oldvalue)))) {
+ old_choice = getChoices(oldvalue);
+ if (!old_choice) {
+ gf_msg("dht", GF_LOG_WARNING, 0, DHT_MSG_NO_MEMORY,
+ "Error to get choices");
ret = -1;
goto out;
- }
+ }
- ret = dict_get_str (dst, key, &oldvalue);
- if (ret)
+ ret = add_opt(&full_choice, old_choice);
+ if (ret) {
+ gf_msg("dht", GF_LOG_WARNING, 0, DHT_MSG_NO_MEMORY,
+ "Error to add choices");
+ ret = -1;
goto out;
+ }
- if (oldvalue && (strstr (oldvalue, "not"))) {
- gf_msg_debug ("dht", 0,
- "Need to update split-brain status in dict");
+ new_choice = getChoices(value->data);
+ if (!new_choice) {
+ gf_msg("dht", GF_LOG_WARNING, 0, DHT_MSG_NO_MEMORY,
+ "Error to get choices");
ret = -1;
goto out;
- }
- if (oldvalue && (strstr (oldvalue, "metadata-split-brain:yes"))
- && (strstr (oldvalue, "data-split-brain:no"))) {
- if (strstr (value->data, "not")) {
- gf_msg_debug ("dht", 0,
- "No need to update split-brain status");
- ret = 0;
- goto out;
- }
- if (strstr (value->data, "yes") &&
- (strncmp (oldvalue, value->data, strlen(oldvalue)))) {
- old_choice = getChoices (oldvalue);
- if (!old_choice) {
- gf_msg ("dht", GF_LOG_WARNING, 0,
- DHT_MSG_NO_MEMORY,
- "Error to get choices");
- ret = -1;
- goto out;
- }
-
- ret = add_opt (&full_choice, old_choice);
- if (ret) {
- gf_msg ("dht", GF_LOG_WARNING, 0,
- DHT_MSG_NO_MEMORY,
- "Error to add choices");
- ret = -1;
- goto out;
- }
-
- new_choice = getChoices (value->data);
- if (!new_choice) {
- gf_msg ("dht", GF_LOG_WARNING, 0,
- DHT_MSG_NO_MEMORY,
- "Error to get choices");
- ret = -1;
- goto out;
- }
+ }
- ret = add_opt (&full_choice, new_choice);
- if (ret) {
- gf_msg ("dht", GF_LOG_WARNING, 0,
- DHT_MSG_NO_MEMORY,
- "Error to add choices ");
- ret = -1;
- goto out;
- }
- ret = gf_asprintf (&status,
- "data-split-brain:%s "
- "metadata-split-brain:%s Choices:%s",
- "no", "yes", full_choice);
-
- if (-1 == ret) {
- gf_msg ("dht", GF_LOG_WARNING, 0,
- DHT_MSG_NO_MEMORY,
- "Error to prepare status ");
- goto out;
- }
- ret = dict_set_dynstr (dst, key, status);
- if (ret) {
- gf_msg ("dht", GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set full choice");
- }
- }
+ ret = add_opt(&full_choice, new_choice);
+ if (ret) {
+ gf_msg("dht", GF_LOG_WARNING, 0, DHT_MSG_NO_MEMORY,
+ "Error to add choices ");
+ ret = -1;
+ goto out;
+ }
+ ret = gf_asprintf(&status,
+ "data-split-brain:%s "
+ "metadata-split-brain:%s Choices:%s",
+ "no", "yes", full_choice);
+
+ if (-1 == ret) {
+ gf_msg("dht", GF_LOG_WARNING, 0, DHT_MSG_NO_MEMORY,
+ "Error to prepare status ");
+ goto out;
+ }
+ ret = dict_set_dynstr(dst, key, status);
+ if (ret) {
+ gf_msg("dht", GF_LOG_WARNING, 0, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set full choice");
+ }
}
+ }
out:
- if (old_choice)
- GF_FREE (old_choice);
- if (new_choice)
- GF_FREE (new_choice);
- if (full_choice)
- GF_FREE (full_choice);
+ if (old_choice)
+ GF_FREE(old_choice);
+ if (new_choice)
+ GF_FREE(new_choice);
+ if (full_choice)
+ GF_FREE(full_choice);
- return ret;
+ return ret;
}
-
-
-int
-dht_aggregate (dict_t *this, char *key, data_t *value, void *data)
+static int
+dht_aggregate(dict_t *this, char *key, data_t *value, void *data)
{
- dict_t *dst = NULL;
- int32_t ret = -1;
- data_t *dict_data = NULL;
+ dict_t *dst = NULL;
+ int32_t ret = -1;
+ data_t *dict_data = NULL;
- dst = data;
+ dst = data;
- /* compare split brain xattr only */
- if (strcmp (key, GF_AFR_SBRAIN_STATUS) == 0) {
- ret = dht_aggregate_split_brain_xattr(dst, key, value);
+ /* compare split brain xattr only */
+ if (strcmp(key, GF_AFR_SBRAIN_STATUS) == 0) {
+ ret = dht_aggregate_split_brain_xattr(dst, key, value);
+ if (!ret)
+ goto out;
+ } else if (strcmp(key, QUOTA_SIZE_KEY) == 0) {
+ ret = dht_aggregate_quota_xattr(dst, key, value);
+ if (ret) {
+ gf_msg("dht", GF_LOG_WARNING, 0,
+ DHT_MSG_AGGREGATE_QUOTA_XATTR_FAILED,
+ "Failed to aggregate quota xattr");
+ }
+ goto out;
+ } else if (fnmatch(GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0) {
+ ret = gf_get_min_stime(THIS, dst, key, value);
+ goto out;
+ } else {
+ /* compare user xattrs only */
+ if (!strncmp(key, "user.", SLEN("user."))) {
+ ret = dict_lookup(dst, key, &dict_data);
+ if (!ret && dict_data && value) {
+ ret = is_data_equal(dict_data, value);
if (!ret)
- goto out;
- } else if (strcmp (key, QUOTA_SIZE_KEY) == 0) {
- ret = dht_aggregate_quota_xattr (dst, key, value);
- if (ret) {
- gf_msg ("dht", GF_LOG_WARNING, 0,
- DHT_MSG_AGGREGATE_QUOTA_XATTR_FAILED,
- "Failed to aggregate quota xattr");
- }
- goto out;
- } else if (fnmatch (GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0) {
- ret = gf_get_min_stime (THIS, dst, key, value);
- goto out;
- } else {
- /* compare user xattrs only */
- if (!strncmp (key, "user.", strlen ("user."))) {
- ret = dict_lookup (dst, key, &dict_data);
- if (!ret && dict_data && value) {
- ret = is_data_equal (dict_data, value);
- if (!ret)
- gf_msg_debug ("dht", 0,
- "xattr mismatch for %s",
- key);
- }
- }
+ gf_msg_debug("dht", 0, "xattr mismatch for %s", key);
+ }
}
+ }
- ret = dict_set (dst, key, value);
- if (ret) {
- gf_msg ("dht", GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value: key = %s",
- key);
- }
+ ret = dict_set(dst, key, value);
+ if (ret) {
+ gf_msg("dht", GF_LOG_WARNING, 0, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary value: key = %s", key);
+ }
out:
- return ret;
+ return ret;
}
-
-void
-dht_aggregate_xattr (dict_t *dst, dict_t *src)
+static void
+dht_aggregate_xattr(dict_t *dst, dict_t *src)
{
- if ((dst == NULL) || (src == NULL)) {
- goto out;
- }
+ if ((dst == NULL) || (src == NULL)) {
+ goto out;
+ }
- dict_foreach (src, dht_aggregate, dst);
+ dict_foreach(src, dht_aggregate, dst);
out:
- return;
+ return;
}
/* Code to save hashed subvol on inode ctx as a mds subvol
-*/
+ */
int
-dht_inode_ctx_mdsvol_set (inode_t *inode, xlator_t *this, xlator_t *mds_subvol)
-{
- dht_inode_ctx_t *ctx = NULL;
- int ret = -1;
- uint64_t ctx_int = 0;
- gf_boolean_t ctx_free = _gf_false;
-
-
- LOCK (&inode->lock);
- {
- ret = __inode_ctx_get (inode, this , &ctx_int);
- if (ctx_int) {
- ctx = (dht_inode_ctx_t *)ctx_int;
- ctx->mds_subvol = mds_subvol;
- } else {
- ctx = GF_CALLOC (1, sizeof(*ctx), gf_dht_mt_inode_ctx_t);
- if (!ctx)
- goto unlock;
- ctx->mds_subvol = mds_subvol;
- ctx_free = _gf_true;
- ctx_int = (long) ctx;
- ret = __inode_ctx_set (inode, this, &ctx_int);
- }
- }
+dht_inode_ctx_mdsvol_set(inode_t *inode, xlator_t *this, xlator_t *mds_subvol)
+{
+ dht_inode_ctx_t *ctx = NULL;
+ int ret = -1;
+ uint64_t ctx_int = 0;
+ gf_boolean_t ctx_free = _gf_false;
+
+ LOCK(&inode->lock);
+ {
+ ret = __inode_ctx_get(inode, this, &ctx_int);
+ if (ctx_int) {
+ ctx = (dht_inode_ctx_t *)(uintptr_t)ctx_int;
+ ctx->mds_subvol = mds_subvol;
+ } else {
+ ctx = GF_CALLOC(1, sizeof(*ctx), gf_dht_mt_inode_ctx_t);
+ if (!ctx)
+ goto unlock;
+ ctx->mds_subvol = mds_subvol;
+ ctx_free = _gf_true;
+ ctx_int = (long)ctx;
+ ret = __inode_ctx_set(inode, this, &ctx_int);
+ }
+ }
unlock:
- UNLOCK (&inode->lock);
- if (ret && ctx_free)
- GF_FREE (ctx);
- return ret;
+ UNLOCK(&inode->lock);
+ if (ret && ctx_free)
+ GF_FREE(ctx);
+ return ret;
}
/*Code to get mds subvol from inode ctx */
int
-dht_inode_ctx_mdsvol_get (inode_t *inode, xlator_t *this, xlator_t **mdsvol)
+dht_inode_ctx_mdsvol_get(inode_t *inode, xlator_t *this, xlator_t **mdsvol)
{
- dht_inode_ctx_t *ctx = NULL;
- int ret = -1;
+ dht_inode_ctx_t *ctx = NULL;
+ int ret = -1;
- if (!mdsvol)
- return ret;
+ if (!mdsvol)
+ return ret;
- if (__is_root_gfid(inode->gfid)) {
- (*mdsvol) = FIRST_CHILD (this);
- return 0;
- }
+ if (__is_root_gfid(inode->gfid)) {
+ (*mdsvol) = FIRST_CHILD(this);
+ return 0;
+ }
- ret = dht_inode_ctx_get (inode, this, &ctx);
+ ret = dht_inode_ctx_get(inode, this, &ctx);
- if (!ret && ctx) {
- if (ctx->mds_subvol) {
- *mdsvol = ctx->mds_subvol;
- ret = 0;
- } else {
- ret = -1;
- }
+ if (!ret && ctx) {
+ if (ctx->mds_subvol) {
+ *mdsvol = ctx->mds_subvol;
+ ret = 0;
+ } else {
+ ret = -1;
}
+ }
- return ret;
+ return ret;
}
-
-
-
-
/* TODO:
- use volumename in xattr instead of "dht"
- use NS locks
@@ -521,293 +450,332 @@ dht_inode_ctx_mdsvol_get (inode_t *inode, xlator_t *this, xlator_t **mdsvol)
- complete linkfile selfheal
*/
-
-int
-dht_lookup_selfheal_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+static int
+dht_lookup_selfheal_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- dht_layout_t *layout = NULL;
- dht_conf_t *conf = NULL;
- int ret = -1;
+ dht_local_t *local = NULL;
+ dht_layout_t *layout = NULL;
+ dht_conf_t *conf = NULL;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO ("dht", frame, out);
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO("dht", frame, out);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, out);
- local = frame->local;
- conf = this->private;
- ret = op_ret;
+ local = frame->local;
+ conf = this->private;
+ ret = op_ret;
- FRAME_SU_UNDO (frame, dht_local_t);
+ FRAME_SU_UNDO(frame, dht_local_t);
- if (ret == 0) {
- layout = local->selfheal.layout;
- ret = dht_layout_set (this, local->inode, layout);
- }
+ if (ret == 0) {
+ layout = local->selfheal.layout;
+ ret = dht_layout_set(this, local->inode, layout);
+ }
- dht_inode_ctx_time_update (local->inode, this, &local->stbuf, 1);
- if (local->loc.parent) {
- dht_inode_ctx_time_update (local->loc.parent, this,
- &local->postparent, 1);
- }
+ dht_inode_ctx_time_update(local->inode, this, &local->stbuf, 1);
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update(local->loc.parent, this, &local->postparent,
+ 1);
+ }
- DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
- dht_set_fixed_dir_stat (&local->postparent);
- /* Delete mds xattr at the time of STACK UNWIND */
- GF_REMOVE_INTERNAL_XATTR (conf->mds_xattr_key, local->xattr);
+ DHT_STRIP_PHASE1_FLAGS(&local->stbuf);
+ dht_set_fixed_dir_stat(&local->postparent);
+ /* Delete mds xattr at the time of STACK UNWIND */
+ GF_REMOVE_INTERNAL_XATTR(conf->mds_xattr_key, local->xattr);
- DHT_STACK_UNWIND (lookup, frame, ret, local->op_errno, local->inode,
- &local->stbuf, local->xattr, &local->postparent);
+ DHT_STACK_UNWIND(lookup, frame, ret, local->op_errno, local->inode,
+ &local->stbuf, local->xattr, &local->postparent);
out:
- return ret;
+ return ret;
}
-int
-dht_discover_complete (xlator_t *this, call_frame_t *discover_frame)
-{
- dht_local_t *local = NULL;
- dht_local_t *heal_local = NULL;
- call_frame_t *main_frame = NULL;
- call_frame_t *heal_frame = NULL;
- int op_errno = 0;
- int ret = -1;
- dht_layout_t *layout = NULL;
- dht_conf_t *conf = NULL;
- uint32_t vol_commit_hash = 0;
- xlator_t *source = NULL;
- int heal_path = 0;
- int error_while_marking_mds = 0;
- int i = 0;
- loc_t loc = {0 };
- int8_t is_read_only = 0, layout_anomalies = 0;
- char gfid_local[GF_UUID_BUF_SIZE] = {0};
-
- local = discover_frame->local;
- layout = local->layout;
- conf = this->private;
- gf_uuid_unparse(local->gfid, gfid_local);
-
- LOCK(&discover_frame->lock);
- {
- main_frame = local->main_frame;
- local->main_frame = NULL;
- }
- UNLOCK(&discover_frame->lock);
-
- if (!main_frame)
- return 0;
-
- /* Code to update all extended attributed from
- subvol to local->xattr on that internal xattr has found
- */
- if (conf->subvolume_cnt == 1)
- local->need_xattr_heal = 0;
- if (local->need_xattr_heal && (local->mds_xattr)) {
- dht_dir_set_heal_xattr (this, local, local->xattr,
- local->mds_xattr, NULL, NULL);
- dict_unref (local->mds_xattr);
- local->mds_xattr = NULL;
- }
+static int
+dht_discover_complete(xlator_t *this, call_frame_t *discover_frame)
+{
+ dht_local_t *local = NULL;
+ dht_local_t *heal_local = NULL;
+ call_frame_t *main_frame = NULL;
+ call_frame_t *heal_frame = NULL;
+ int op_errno = 0;
+ int ret = -1;
+ dht_layout_t *layout = NULL;
+ dht_conf_t *conf = NULL;
+ uint32_t vol_commit_hash = 0;
+ xlator_t *source = NULL;
+ int heal_path = 0;
+ int error_while_marking_mds = 0;
+ int i = 0;
+ loc_t loc = {0};
+ int8_t is_read_only = 0, layout_anomalies = 0;
+ char gfid_local[GF_UUID_BUF_SIZE] = {0};
+
+ local = discover_frame->local;
+ layout = local->layout;
+ conf = this->private;
+ gf_uuid_unparse(local->gfid, gfid_local);
+
+ LOCK(&discover_frame->lock);
+ {
+ main_frame = local->main_frame;
+ local->main_frame = NULL;
+ }
+ UNLOCK(&discover_frame->lock);
+
+ if (!main_frame)
+ return 0;
+
+ /* Code to update all extended attributed from
+ subvol to local->xattr on that internal xattr has found
+ */
+ if (conf->subvolume_cnt == 1)
+ local->need_xattr_heal = 0;
+ if (local->need_xattr_heal && (local->mds_xattr)) {
+ dht_dir_set_heal_xattr(this, local, local->xattr, local->mds_xattr,
+ NULL, NULL);
+ dict_unref(local->mds_xattr);
+ local->mds_xattr = NULL;
+ }
+
+ ret = dict_get_int8(local->xattr_req, QUOTA_READ_ONLY_KEY, &is_read_only);
+ if (ret < 0)
+ gf_msg_debug(this->name, 0, "key = %s not present in dict",
+ QUOTA_READ_ONLY_KEY);
+
+ if (local->file_count && local->dir_count) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_FILE_TYPE_MISMATCH,
+ "path %s exists as a file on one subvolume "
+ "and directory on another. "
+ "Please fix it manually",
+ local->loc.path);
+ op_errno = EIO;
+ goto out;
+ }
- ret = dict_get_int8 (local->xattr_req, QUOTA_READ_ONLY_KEY,
- &is_read_only);
- if (ret < 0)
- gf_msg_debug (this->name, 0, "key = %s not present in dict",
- QUOTA_READ_ONLY_KEY);
-
- if (local->file_count && local->dir_count) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_FILE_TYPE_MISMATCH,
- "path %s exists as a file on one subvolume "
- "and directory on another. "
- "Please fix it manually",
- local->loc.path);
- op_errno = EIO;
- goto out;
+ if (local->cached_subvol) {
+ ret = dht_layout_preset(this, local->cached_subvol, local->inode);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_LAYOUT_SET_FAILED,
+ "failed to set layout for subvolume %s",
+ local->cached_subvol ? local->cached_subvol->name : "<nil>");
+ op_errno = EINVAL;
+ goto out;
+ }
+ } else {
+ ret = dht_layout_normalize(this, &local->loc, layout);
+ if ((ret < 0) || ((ret > 0) && (local->op_ret != 0))) {
+ /* either the layout is incorrect or the directory is
+ * not found even in one subvolume.
+ */
+ gf_msg_debug(this->name, 0,
+ "normalizing failed on %s "
+ "(overlaps/holes present: %s, "
+ "ENOENT errors: %d)",
+ local->loc.path, (ret < 0) ? "yes" : "no",
+ (ret > 0) ? ret : 0);
+ layout_anomalies = 1;
+ } else if (local->inode) {
+ dht_layout_set(this, local->inode, layout);
+ }
+ }
+
+ if (!conf->vch_forced) {
+ ret = dict_get_uint32(local->xattr, conf->commithash_xattr_name,
+ &vol_commit_hash);
+ if (ret == 0) {
+ conf->vol_commit_hash = vol_commit_hash;
}
+ }
- if (local->cached_subvol) {
- ret = dht_layout_preset (this, local->cached_subvol,
- local->inode);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_LAYOUT_SET_FAILED,
- "failed to set layout for subvolume %s",
- local->cached_subvol ? local->cached_subvol->name : "<nil>");
- op_errno = EINVAL;
- goto out;
- }
- } else {
- ret = dht_layout_normalize (this, &local->loc, layout);
- if ((ret < 0) || ((ret > 0) && (local->op_ret != 0))) {
- /* either the layout is incorrect or the directory is
- * not found even in one subvolume.
- */
- gf_msg_debug (this->name, 0,
- "normalizing failed on %s "
- "(overlaps/holes present: %s, "
- "ENOENT errors: %d)", local->loc.path,
- (ret < 0) ? "yes" : "no", (ret > 0) ? ret : 0);
- layout_anomalies = 1;
- } else if (local->inode) {
- dht_layout_set (this, local->inode, layout);
- }
- }
+ if (IA_ISDIR(local->stbuf.ia_type) && !is_read_only) {
+ for (i = 0; i < layout->cnt; i++) {
+ if (!source && !layout->list[i].err)
+ source = layout->list[i].xlator;
+ if (layout->list[i].err == ENOENT ||
+ layout->list[i].err == ESTALE) {
+ heal_path = 1;
+ }
- if (!conf->vch_forced) {
- ret = dict_get_uint32 (local->xattr,
- conf->commithash_xattr_name,
- &vol_commit_hash);
- if (ret == 0) {
- conf->vol_commit_hash = vol_commit_hash;
- }
+ if (source && heal_path)
+ break;
}
+ }
- if (IA_ISDIR (local->stbuf.ia_type) && !is_read_only) {
- for (i = 0; i < layout->cnt; i++) {
- if (!source && !layout->list[i].err)
- source = layout->list[i].xlator;
- if (layout->list[i].err == ENOENT ||
- layout->list[i].err == ESTALE) {
- heal_path = 1;
- }
-
- if (source && heal_path)
- break;
- }
+ if (IA_ISDIR(local->stbuf.ia_type)) {
+ /* Call function to save hashed subvol on inode ctx if
+ internal mds xattr is not present and all subvols are up
+ */
+ if (!local->op_ret && !__is_root_gfid(local->stbuf.ia_gfid))
+ (void)dht_common_mark_mdsxattr(discover_frame,
+ &error_while_marking_mds, 1);
+
+ if (local->need_xattr_heal && !heal_path) {
+ local->need_xattr_heal = 0;
+ ret = dht_dir_xattr_heal(this, local, &op_errno);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno,
+ DHT_MSG_DIR_XATTR_HEAL_FAILED,
+ "xattr heal failed for "
+ "directory gfid is %s ",
+ gfid_local);
+ }
}
+ }
- if (IA_ISDIR (local->stbuf.ia_type)) {
- /* Call function to save hashed subvol on inode ctx if
- internal mds xattr is not present and all subvols are up
- */
- if (!local->op_ret && !__is_root_gfid (local->stbuf.ia_gfid))
- (void) dht_common_mark_mdsxattr (discover_frame,
- &error_while_marking_mds, 1);
-
- if (local->need_xattr_heal && !heal_path) {
- local->need_xattr_heal = 0;
- ret = dht_dir_xattr_heal (this, local);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR,
- ret,
- DHT_MSG_DIR_XATTR_HEAL_FAILED,
- "xattr heal failed for "
- "directory gfid is %s ",
- gfid_local);
- }
+ if (source && (heal_path || layout_anomalies || error_while_marking_mds)) {
+ gf_uuid_copy(loc.gfid, local->gfid);
+ if (gf_uuid_is_null(loc.gfid)) {
+ goto done;
}
- if (source && (heal_path || layout_anomalies || error_while_marking_mds)) {
- gf_uuid_copy (loc.gfid, local->gfid);
- if (gf_uuid_is_null (loc.gfid)) {
- goto done;
- }
-
- if (local->inode)
- loc.inode = inode_ref (local->inode);
- else
- goto done;
-
- heal_frame = create_frame (this, this->ctx->pool);
- if (heal_frame) {
- heal_local = dht_local_init (heal_frame, &loc,
- NULL, 0);
- if (!heal_local)
- goto cleanup;
-
- gf_uuid_copy (heal_local->gfid, local->gfid);
- heal_frame->cookie = source;
- heal_local->xattr = dict_ref (local->xattr);
- heal_local->stbuf = local->stbuf;
- heal_local->postparent = local->postparent;
- heal_local->inode = inode_ref (loc.inode);
- heal_local->main_frame = main_frame;
- FRAME_SU_DO (heal_frame, dht_local_t);
- ret = synctask_new (this->ctx->env,
- dht_heal_full_path,
- dht_heal_full_path_done,
- heal_frame, heal_frame);
- if (!ret) {
- loc_wipe (&loc);
- return 0;
- }
- /*
- * Failed to spawn the synctask. Returning
- * with out doing heal.
- */
-cleanup:
- loc_wipe (&loc);
- DHT_STACK_DESTROY (heal_frame);
- }
-
- }
+ if (local->inode)
+ loc.inode = inode_ref(local->inode);
+ else
+ goto done;
+
+ heal_frame = create_frame(this, this->ctx->pool);
+ if (heal_frame) {
+ heal_local = dht_local_init(heal_frame, &loc, NULL, 0);
+ if (!heal_local)
+ goto cleanup;
+
+ gf_uuid_copy(heal_local->gfid, local->gfid);
+ heal_frame->cookie = source;
+ heal_local->xattr = dict_ref(local->xattr);
+ heal_local->stbuf = local->stbuf;
+ heal_local->postparent = local->postparent;
+ heal_local->inode = inode_ref(loc.inode);
+ heal_local->main_frame = main_frame;
+ FRAME_SU_DO(heal_frame, dht_local_t);
+ ret = synctask_new(this->ctx->env, dht_heal_full_path,
+ dht_heal_full_path_done, heal_frame, heal_frame);
+ if (!ret) {
+ loc_wipe(&loc);
+ return 0;
+ }
+ /*
+ * Failed to spawn the synctask. Returning
+ * with out doing heal.
+ */
+ cleanup:
+ loc_wipe(&loc);
+ DHT_STACK_DESTROY(heal_frame);
+ }
+ }
done:
- dht_set_fixed_dir_stat (&local->postparent);
- /* Delete mds xattr at the time of STACK UNWIND */
- if (local->xattr)
- GF_REMOVE_INTERNAL_XATTR (conf->mds_xattr_key, local->xattr);
+ dht_set_fixed_dir_stat(&local->postparent);
+ /* Delete mds xattr at the time of STACK UNWIND */
+ if (local->xattr)
+ GF_REMOVE_INTERNAL_XATTR(conf->mds_xattr_key, local->xattr);
- DHT_STACK_UNWIND (lookup, main_frame, local->op_ret, local->op_errno,
- local->inode, &local->stbuf, local->xattr,
- &local->postparent);
- return 0;
+ DHT_STACK_UNWIND(lookup, main_frame, local->op_ret, local->op_errno,
+ local->inode, &local->stbuf, local->xattr,
+ &local->postparent);
+ return 0;
out:
- DHT_STACK_UNWIND (lookup, main_frame, -1, op_errno, NULL, NULL, NULL,
- NULL);
+ DHT_STACK_UNWIND(lookup, main_frame, -1, op_errno, NULL, NULL, NULL, NULL);
- return ret;
+ return ret;
}
-int
-dht_common_mark_mdsxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- dict_t *xdata)
-{
- dht_local_t *local = NULL;
- xlator_t *prev = cookie;
- int ret = -1;
- dht_conf_t *conf = 0;
- dht_layout_t *layout = NULL;
+static int
+dht_common_mark_mdsxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ xlator_t *prev = cookie;
+ int ret = -1;
+ dht_conf_t *conf = 0;
+ dht_layout_t *layout = NULL;
+ int32_t mds_heal_fresh_lookup = 0;
+
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame->local, out);
+
+ local = frame->local;
+ conf = this->private;
+ layout = local->selfheal.layout;
+ mds_heal_fresh_lookup = local->mds_heal_fresh_lookup;
+
+ if (op_ret) {
+ gf_msg_debug(this->name, op_ret,
+ "Failed to set %s on the MDS %s for path %s. ",
+ conf->mds_xattr_key, prev->name, local->loc.path);
+ } else {
+ /* Save mds subvol on inode ctx */
+ ret = dht_inode_ctx_mdsvol_set(local->inode, this, prev);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_SET_INODE_CTX_FAILED,
+ "Failed to set mds subvol on inode ctx"
+ " %s for %s ",
+ prev->name, local->loc.path);
+ }
+ }
+ if (!local->mds_heal_fresh_lookup && layout) {
+ dht_selfheal_dir_setattr(frame, &local->loc, &local->stbuf, 0xffffffff,
+ layout);
+ }
+out:
+ if (mds_heal_fresh_lookup)
+ DHT_STACK_DESTROY(frame);
+ return 0;
+}
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, frame->local, out);
+static xlator_t *
+dht_inode_get_hashed_subvol(inode_t *inode, xlator_t *this, loc_t *loc)
+{
+ char *path = NULL;
+ loc_t populate_loc = {
+ 0,
+ };
+ char *name = NULL;
+ xlator_t *hash_subvol = NULL;
- local = frame->local;
- conf = this->private;
- layout = local->selfheal.layout;
+ if (!inode)
+ return hash_subvol;
- if (op_ret) {
- gf_msg_debug (this->name, op_ret,
- "Failed to set %s on the MDS %s for path %s. ",
- conf->mds_xattr_key, prev->name, local->loc.path);
- } else {
- /* Save mds subvol on inode ctx */
- ret = dht_inode_ctx_mdsvol_set (local->inode, this, prev);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_SET_INODE_CTX_FAILED,
- "Failed to set mds subvol on inode ctx"
- " %s for %s ", prev->name,
- local->loc.path);
- }
+ if (loc && loc->parent && loc->path) {
+ if (!loc->name) {
+ name = strrchr(loc->path, '/');
+ if (name) {
+ loc->name = name + 1;
+ } else {
+ goto out;
+ }
}
- if (!local->mds_heal_fresh_lookup && layout) {
- dht_selfheal_dir_setattr (frame, &local->loc, &local->stbuf,
- 0xffffffff, layout);
+ hash_subvol = dht_subvol_get_hashed(this, loc);
+ goto out;
+ }
+
+ if (!gf_uuid_is_null(inode->gfid)) {
+ populate_loc.inode = inode_ref(inode);
+ populate_loc.parent = inode_parent(populate_loc.inode, NULL, NULL);
+ inode_path(populate_loc.inode, NULL, &path);
+
+ if (!path)
+ goto out;
+
+ populate_loc.path = path;
+ if (!populate_loc.name && populate_loc.path) {
+ name = strrchr(populate_loc.path, '/');
+ if (name) {
+ populate_loc.name = name + 1;
+
+ } else {
+ goto out;
+ }
}
+ hash_subvol = dht_subvol_get_hashed(this, &populate_loc);
+ }
out:
- if (local && local->mds_heal_fresh_lookup)
- DHT_STACK_DESTROY (frame);
- return 0;
+ if (populate_loc.inode)
+ loc_wipe(&populate_loc);
+ return hash_subvol;
}
-
-
/* Common function call by revalidate/selfheal code path to populate
internal xattr if it is not present, mark_during_fresh_lookup value
determines either function is call by revalidate_cbk(discover_complete)
@@ -823,1317 +791,1510 @@ out:
penalty.
*/
int
-dht_common_mark_mdsxattr (call_frame_t *frame, int *errst, int mark_during_fresh_lookup)
-{
- dht_local_t *local = NULL;
- xlator_t *this = NULL;
- xlator_t *hashed_subvol = NULL;
- int ret = 0;
- int i = 0;
- dict_t *xattrs = NULL;
- char gfid_local[GF_UUID_BUF_SIZE] = {0,};
- int32_t zero[1] = {0};
- dht_conf_t *conf = 0;
- dht_layout_t *layout = NULL;
- dht_local_t *copy_local = NULL;
- call_frame_t *xattr_frame = NULL;
- gf_boolean_t vol_down = _gf_false;
-
- this = frame->this;
-
- GF_VALIDATE_OR_GOTO ("dht", frame, out);
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame->local, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
-
- local = frame->local;
- conf = this->private;
- layout = local->selfheal.layout;
- local->mds_heal_fresh_lookup = mark_during_fresh_lookup;
- gf_uuid_unparse(local->gfid, gfid_local);
-
- /* Code to update hashed subvol consider as a mds subvol
- and wind a setxattr call on hashed subvol to update
- internal xattr
+dht_common_mark_mdsxattr(call_frame_t *frame, int *errst,
+ int mark_during_fresh_lookup)
+{
+ dht_local_t *local = NULL;
+ xlator_t *this = NULL;
+ xlator_t *hashed_subvol = NULL;
+ int ret = 0;
+ int i = 0;
+ dict_t *xattrs = NULL;
+ char gfid_local[GF_UUID_BUF_SIZE] = {
+ 0,
+ };
+ int32_t zero[1] = {0};
+ dht_conf_t *conf = 0;
+ dht_layout_t *layout = NULL;
+ dht_local_t *copy_local = NULL;
+ call_frame_t *xattr_frame = NULL;
+ gf_boolean_t vol_down = _gf_false;
+
+ GF_VALIDATE_OR_GOTO("dht", frame, out);
+ this = frame->this;
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame->local, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
+
+ local = frame->local;
+ conf = this->private;
+ layout = local->selfheal.layout;
+ local->mds_heal_fresh_lookup = mark_during_fresh_lookup;
+
+ gf_uuid_unparse(local->gfid, gfid_local);
+
+ /* Code to update hashed subvol consider as a mds subvol
+ and wind a setxattr call on hashed subvol to update
+ internal xattr
+ */
+ if (!local->xattr || !dict_get(local->xattr, conf->mds_xattr_key)) {
+ /* It means no internal MDS xattr has been set yet
+ */
+ /* Check the status of all subvol are up while call
+ this function call by lookup code path
*/
- if (!local->xattr || !dict_get (local->xattr, conf->mds_xattr_key)) {
- /* It means no internal MDS xattr has been set yet
- */
- /* Check the status of all subvol are up while call
- this function call by lookup code path
- */
- if (mark_during_fresh_lookup) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (!conf->subvolume_status[i]) {
- vol_down = _gf_true;
- break;
- }
- }
- if (vol_down) {
- gf_msg_debug (this->name, 0,
- "subvol %s is down. Unable to "
- " save mds subvol on inode for "
- " path %s gfid is %s " ,
- conf->subvolumes[i]->name,
- local->loc.path, gfid_local);
- goto out;
- }
+ if (mark_during_fresh_lookup) {
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (!conf->subvolume_status[i]) {
+ vol_down = _gf_true;
+ break;
}
+ }
+ if (vol_down) {
+ gf_msg_debug(this->name, 0,
+ "subvol %s is down. Unable to "
+ " save mds subvol on inode for "
+ " path %s gfid is %s ",
+ conf->subvolumes[i]->name, local->loc.path,
+ gfid_local);
+ goto out;
+ }
+ }
- /* Calculate hashed subvol based on inode and parent node
- */
- hashed_subvol = dht_inode_get_hashed_subvol (local->inode, this,
- &local->loc);
- if (!hashed_subvol) {
- gf_msg (this->name, GF_LOG_DEBUG, 0,
- DHT_MSG_HASHED_SUBVOL_GET_FAILED,
- "Failed to get hashed subvol for path %s"
- "gfid is %s ",
- local->loc.path, gfid_local);
- (*errst) = 1;
- ret = -1;
- goto out;
- }
- xattrs = dict_new ();
- if (!xattrs) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_NO_MEMORY, "dict_new failed");
- ret = -1;
- goto out;
- }
- /* Add internal MDS xattr on disk for hashed subvol
- */
- ret = dht_dict_set_array (xattrs, conf->mds_xattr_key,
- zero, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary"
- " value:key = %s for "
- "path %s", conf->mds_xattr_key,
- local->loc.path);
- ret = -1;
- goto out;
- }
- /* Create a new frame to wind a call only while
- this function call by revalidate_cbk code path
- To wind a call parallel need to create a new frame
- */
- if (mark_during_fresh_lookup) {
- xattr_frame = create_frame (this, this->ctx->pool);
- if (!xattr_frame) {
- ret = -1;
- goto out;
- }
- copy_local = dht_local_init (xattr_frame, &(local->loc),
- NULL, 0);
- if (!copy_local) {
- ret = -1;
- DHT_STACK_DESTROY (xattr_frame);
- goto out;
- }
- copy_local->stbuf = local->stbuf;
- copy_local->mds_heal_fresh_lookup = mark_during_fresh_lookup;
- if (!copy_local->inode)
- copy_local->inode = inode_ref (local->inode);
- gf_uuid_copy (copy_local->loc.gfid, local->gfid);
- FRAME_SU_DO (xattr_frame, dht_local_t);
- STACK_WIND_COOKIE (xattr_frame, dht_common_mark_mdsxattr_cbk,
- hashed_subvol, hashed_subvol,
- hashed_subvol->fops->setxattr,
- &local->loc, xattrs, 0, NULL);
- } else {
- STACK_WIND_COOKIE (frame,
- dht_common_mark_mdsxattr_cbk,
- (void *)hashed_subvol,
- hashed_subvol,
- hashed_subvol->fops->setxattr,
- &local->loc, xattrs, 0,
- NULL);
- }
- } else {
- gf_msg_debug (this->name, 0,
- "internal xattr %s is present on subvol"
- "on path %s gfid is %s " , conf->mds_xattr_key,
- local->loc.path, gfid_local);
- if (!mark_during_fresh_lookup)
- dht_selfheal_dir_setattr (frame, &local->loc,
- &local->stbuf, 0xffffffff,
- layout);
+ /* Calculate hashed subvol based on inode and parent node
+ */
+ hashed_subvol = dht_inode_get_hashed_subvol(local->inode, this,
+ &local->loc);
+ if (!hashed_subvol) {
+ gf_msg(this->name, GF_LOG_DEBUG, 0,
+ DHT_MSG_HASHED_SUBVOL_GET_FAILED,
+ "Failed to get hashed subvol for path %s"
+ "gfid is %s ",
+ local->loc.path, gfid_local);
+ if (errst)
+ (*errst) = 1;
+ ret = -1;
+ goto out;
+ }
+ xattrs = dict_new();
+ if (!xattrs) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
+ "dict_new failed");
+ ret = -1;
+ goto out;
}
+ /* Add internal MDS xattr on disk for hashed subvol
+ */
+ ret = dht_dict_set_array(xattrs, conf->mds_xattr_key, zero, 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary"
+ " value:key = %s for "
+ "path %s",
+ conf->mds_xattr_key, local->loc.path);
+ ret = -1;
+ goto out;
+ }
+ /* Create a new frame to wind a call only while
+ this function call by revalidate_cbk code path
+ To wind a call parallel need to create a new frame
+ */
+ if (mark_during_fresh_lookup) {
+ xattr_frame = create_frame(this, this->ctx->pool);
+ if (!xattr_frame) {
+ ret = -1;
+ goto out;
+ }
+ copy_local = dht_local_init(xattr_frame, &(local->loc), NULL, 0);
+ if (!copy_local) {
+ ret = -1;
+ DHT_STACK_DESTROY(xattr_frame);
+ goto out;
+ }
+ copy_local->stbuf = local->stbuf;
+ copy_local->mds_heal_fresh_lookup = mark_during_fresh_lookup;
+ if (!copy_local->inode)
+ copy_local->inode = inode_ref(local->inode);
+ gf_uuid_copy(copy_local->loc.gfid, local->gfid);
+ FRAME_SU_DO(xattr_frame, dht_local_t);
+ STACK_WIND_COOKIE(xattr_frame, dht_common_mark_mdsxattr_cbk,
+ hashed_subvol, hashed_subvol,
+ hashed_subvol->fops->setxattr, &local->loc,
+ xattrs, 0, NULL);
+ } else {
+ STACK_WIND_COOKIE(frame, dht_common_mark_mdsxattr_cbk,
+ (void *)hashed_subvol, hashed_subvol,
+ hashed_subvol->fops->setxattr, &local->loc,
+ xattrs, 0, NULL);
+ }
+ } else {
+ gf_msg_debug(this->name, 0,
+ "internal xattr %s is present on subvol"
+ "on path %s gfid is %s ",
+ conf->mds_xattr_key, local->loc.path, gfid_local);
+ if (!mark_during_fresh_lookup)
+ dht_selfheal_dir_setattr(frame, &local->loc, &local->stbuf,
+ 0xffffffff, layout);
+ }
out:
- if (xattrs)
- dict_unref (xattrs);
- return ret;
+ if (xattrs)
+ dict_unref(xattrs);
+ return ret;
}
+/* Get the value of key from dict in the bytewise and save in array after
+ convert from network byte order to host byte order
+*/
+static int32_t
+dht_dict_get_array(dict_t *dict, char *key, int32_t value[], int32_t size,
+ int *errst)
+{
+ void *ptr = NULL;
+ int32_t len = -1;
+ int32_t vindex = -1;
+ int32_t err = -1;
+ int ret = 0;
+
+ if (dict == NULL) {
+ (*errst) = -1;
+ return -EINVAL;
+ }
+ err = dict_get_ptr_and_len(dict, key, &ptr, &len);
+ if (err != 0) {
+ (*errst) = -1;
+ return err;
+ }
+
+ if (len != (size * sizeof(int32_t))) {
+ (*errst) = -1;
+ return -EINVAL;
+ }
+
+ for (vindex = 0; vindex < size; vindex++) {
+ value[vindex] = ntoh32(*((int32_t *)ptr + vindex));
+ if (value[vindex] < 0)
+ ret = -1;
+ }
+
+ return ret;
+}
-int
-dht_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, dict_t *xattr,
- struct iatt *postparent)
-{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- xlator_t *prev = NULL;
- dht_layout_t *layout = NULL;
- int ret = -1;
- int is_dir = 0;
- int32_t check_mds = 0;
- int is_linkfile = 0;
- int attempt_unwind = 0;
- dht_conf_t *conf = 0;
- char gfid_local[GF_UUID_BUF_SIZE] = {0};
- char gfid_node[GF_UUID_BUF_SIZE] = {0};
- int32_t mds_xattr_val[1] = {0};
- int errst = 0;
-
-
- GF_VALIDATE_OR_GOTO ("dht", frame, out);
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
- GF_VALIDATE_OR_GOTO ("dht", this->private, out);
- GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+static int
+dht_discover_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, inode_t *inode, struct iatt *stbuf,
+ dict_t *xattr, struct iatt *postparent)
+{
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ xlator_t *prev = NULL;
+ dht_layout_t *layout = NULL;
+ int ret = -1;
+ int is_dir = 0;
+ int32_t check_mds = 0;
+ int is_linkfile = 0;
+ int attempt_unwind = 0;
+ dht_conf_t *conf = 0;
+ char gfid_local[GF_UUID_BUF_SIZE] = {0};
+ char gfid_node[GF_UUID_BUF_SIZE] = {0};
+ int32_t mds_xattr_val[1] = {0};
+ int errst = 0;
+
+ GF_VALIDATE_OR_GOTO("dht", frame, out);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO("dht", this->private, out);
+ GF_VALIDATE_OR_GOTO("dht", cookie, out);
+
+ local = frame->local;
+ prev = cookie;
+ conf = this->private;
+
+ layout = local->layout;
+
+ /* Check if the gfid is different for file from other node */
+ if (!op_ret && gf_uuid_compare(local->gfid, stbuf->ia_gfid)) {
+ gf_uuid_unparse(stbuf->ia_gfid, gfid_node);
+ gf_uuid_unparse(local->gfid, gfid_local);
- local = frame->local;
- prev = cookie;
- conf = this->private;
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_GFID_MISMATCH,
+ "%s: gfid different on %s, gfid local = %s"
+ "gfid other = %s",
+ local->loc.path, prev->name, gfid_local, gfid_node);
+ }
- layout = local->layout;
+ LOCK(&frame->lock);
+ {
+ /* TODO: assert equal mode on stbuf->st_mode and
+ local->stbuf->st_mode
+ else mkdir/chmod/chown and fix
+ */
- /* Check if the gfid is different for file from other node */
- if (!op_ret && gf_uuid_compare (local->gfid, stbuf->ia_gfid)) {
+ ret = dht_layout_merge(this, layout, prev, op_ret, op_errno, xattr);
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_LAYOUT_MERGE_FAILED,
+ "%s: failed to merge layouts for subvol %s", local->loc.path,
+ prev->name);
- gf_uuid_unparse(stbuf->ia_gfid, gfid_node);
- gf_uuid_unparse(local->gfid, gfid_local);
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ gf_msg_debug(this->name, op_errno,
+ "lookup of %s on %s returned error", local->loc.path,
+ prev->name);
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_GFID_MISMATCH,
- "%s: gfid different on %s, gfid local = %s"
- "gfid other = %s",
- local->loc.path, prev->name,
- gfid_local, gfid_node);
+ goto unlock;
}
+ is_linkfile = check_is_linkfile(inode, stbuf, xattr,
+ conf->link_xattr_name);
+ is_dir = check_is_dir(inode, stbuf, xattr);
- LOCK (&frame->lock);
- {
- /* TODO: assert equal mode on stbuf->st_mode and
- local->stbuf->st_mode
-
- else mkdir/chmod/chown and fix
- */
-
- ret = dht_layout_merge (this, layout, prev,
- op_ret, op_errno, xattr);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_LAYOUT_MERGE_FAILED,
- "%s: failed to merge layouts for subvol %s",
- local->loc.path, prev->name);
-
- if (op_ret == -1) {
- local->op_errno = op_errno;
- gf_msg_debug (this->name, op_errno,
- "lookup of %s on %s returned error",
- local->loc.path, prev->name);
-
- goto unlock;
- }
-
- is_linkfile = check_is_linkfile (inode, stbuf, xattr,
- conf->link_xattr_name);
- is_dir = check_is_dir (inode, stbuf, xattr);
-
- if (is_dir) {
- local->dir_count ++;
- } else {
- local->file_count ++;
-
- if (!is_linkfile && !local->cached_subvol) {
- /* real file */
- /* Ok, we somehow managed to find a file on
- * more than one subvol. ignore this or we
- * will end up overwriting information while a
- * a thread is potentially unwinding from
- * dht_discover_complete
- */
- local->cached_subvol = prev;
- attempt_unwind = 1;
- } else {
- goto unlock;
- }
- }
-
- local->op_ret = 0;
+ if (is_dir) {
+ local->dir_count++;
+ } else {
+ local->file_count++;
+
+ if (!is_linkfile && !local->cached_subvol) {
+ /* real file */
+ /* Ok, we somehow managed to find a file on
+ * more than one subvol. ignore this or we
+ * will end up overwriting information while a
+ * a thread is potentially unwinding from
+ * dht_discover_complete
+ */
+ local->cached_subvol = prev;
+ attempt_unwind = 1;
+ } else {
+ goto unlock;
+ }
+ }
- if (local->xattr == NULL) {
- local->xattr = dict_ref (xattr);
- } else {
- /* Don't aggregate for files. See BZ#1484113 */
- if (is_dir)
- dht_aggregate_xattr (local->xattr, xattr);
- }
+ local->op_ret = 0;
- if (local->inode == NULL)
- local->inode = inode_ref (inode);
+ if (local->xattr == NULL) {
+ local->xattr = dict_ref(xattr);
+ } else {
+ /* Don't aggregate for files. See BZ#1484709 */
+ if (is_dir)
+ dht_aggregate_xattr(local->xattr, xattr);
+ }
- dht_iatt_merge (this, &local->stbuf, stbuf, prev);
- dht_iatt_merge (this, &local->postparent, postparent,
- prev);
- if (!dict_get (xattr, conf->mds_xattr_key)) {
- goto unlock;
- } else {
- gf_msg_debug (this->name, 0,
- "internal xattr %s is present on subvol"
- "on path %s gfid is %s " ,
- conf->mds_xattr_key,
- local->loc.path, gfid_local);
- }
- check_mds = dht_dict_get_array (xattr, conf->mds_xattr_key,
- mds_xattr_val, 1, &errst);
- /* save mds subvol on inode ctx */
- ret = dht_inode_ctx_mdsvol_set (local->inode, this,
- prev);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_SET_INODE_CTX_FAILED,
- "Failed to set hashed subvol for %s vol is %s",
- local->loc.path, prev->name);
- }
+ if (local->inode == NULL)
+ local->inode = inode_ref(inode);
- if ((check_mds < 0) && !errst) {
- local->mds_xattr = dict_ref (xattr);
- gf_msg_debug (this->name, 0,
- "Value of %s is not zero on mds subvol"
- "so xattr needs to be healed on non mds"
- " path is %s and vol name is %s "
- " gfid is %s" ,
- conf->mds_xattr_key,
- local->loc.path,
- prev->name, gfid_local);
- local->need_xattr_heal = 1;
- local->mds_subvol = prev;
- }
+ dht_iatt_merge(this, &local->stbuf, stbuf);
+ dht_iatt_merge(this, &local->postparent, postparent);
- }
+ if (!dict_get(xattr, conf->mds_xattr_key)) {
+ goto unlock;
+ } else {
+ gf_msg_debug(this->name, 0,
+ "internal xattr %s is present on subvol"
+ "on path %s gfid is %s ",
+ conf->mds_xattr_key, local->loc.path, gfid_local);
+ }
+ check_mds = dht_dict_get_array(xattr, conf->mds_xattr_key,
+ mds_xattr_val, 1, &errst);
+ /* save mds subvol on inode ctx */
+ ret = dht_inode_ctx_mdsvol_set(local->inode, this, prev);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_SET_INODE_CTX_FAILED,
+ "Failed to set hashed subvol for %s vol is %s",
+ local->loc.path, prev->name);
+ }
+
+ if ((check_mds < 0) && !errst) {
+ local->mds_xattr = dict_ref(xattr);
+ gf_msg_debug(this->name, 0,
+ "Value of %s is not zero on mds subvol"
+ "so xattr needs to be healed on non mds"
+ " path is %s and vol name is %s "
+ " gfid is %s",
+ conf->mds_xattr_key, local->loc.path, prev->name,
+ gfid_local);
+ local->need_xattr_heal = 1;
+ local->mds_subvol = prev;
+ }
+ }
unlock:
- UNLOCK (&frame->lock);
+ UNLOCK(&frame->lock);
out:
- /* Make sure, the thread executing dht_discover_complete is the one
- * which calls STACK_DESTROY (frame). In the case of "attempt_unwind",
- * this makes sure that the thread don't call dht_frame_return, till
- * call to dht_discover_complete is done.
- */
- if (attempt_unwind) {
- dht_discover_complete (this, frame);
- }
+ /* Make sure, the thread executing dht_discover_complete is the one
+ * which calls STACK_DESTROY (frame). In the case of "attempt_unwind",
+ * this makes sure that the thread don't call dht_frame_return, till
+ * call to dht_discover_complete is done.
+ */
+ if (attempt_unwind) {
+ dht_discover_complete(this, frame);
+ }
- this_call_cnt = dht_frame_return (frame);
+ this_call_cnt = dht_frame_return(frame);
- if (is_last_call (this_call_cnt) && !attempt_unwind) {
- dht_discover_complete (this, frame);
- }
+ if (is_last_call(this_call_cnt) && !attempt_unwind) {
+ dht_discover_complete(this, frame);
+ }
- if (is_last_call (this_call_cnt))
- DHT_STACK_DESTROY (frame);
+ if (is_last_call(this_call_cnt))
+ DHT_STACK_DESTROY(frame);
- return 0;
+ return 0;
}
+static int
+dht_set_file_xattr_req(xlator_t *this, loc_t *loc, dict_t *xattr_req)
+{
+ int ret = -EINVAL;
+ dht_conf_t *conf = NULL;
+
+ conf = this->private;
+ if (!conf) {
+ goto err;
+ }
+
+ if (!xattr_req) {
+ goto err;
+ }
+
+ /* Used to check whether this is a linkto file.
+ */
+ ret = dict_set_uint32(xattr_req, conf->link_xattr_name, 256);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary value:key = %s for "
+ "path %s",
+ conf->link_xattr_name, loc->path);
+ goto err;
+ }
+
+ /* This is used to make sure we don't unlink linkto files
+ * which are the target of an ongoing file migration.
+ */
+ ret = dict_set_uint32(xattr_req, GLUSTERFS_OPEN_FD_COUNT, 4);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary value:key = %s for "
+ "path %s",
+ GLUSTERFS_OPEN_FD_COUNT, loc->path);
+ goto err;
+ }
+
+ ret = 0;
+err:
+ return ret;
+}
-int
-dht_discover (call_frame_t *frame, xlator_t *this, loc_t *loc)
-{
- int ret;
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- int call_cnt = 0;
- int op_errno = EINVAL;
- int i = 0;
- call_frame_t *discover_frame = NULL;
-
- conf = this->private;
- local = frame->local;
-
- ret = dict_set_uint32 (local->xattr_req, conf->xattr_name, 4 * 4);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "%s: Failed to set dictionary value:key = %s",
- loc->path, conf->xattr_name);
-
- ret = dict_set_uint32 (local->xattr_req, conf->link_xattr_name, 256);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "%s: Failed to set dictionary value:key = %s",
- loc->path, conf->link_xattr_name);
-
- if (__is_root_gfid(local->loc.gfid)) {
- ret = dict_set_uint32 (local->xattr_req,
- conf->commithash_xattr_name,
- sizeof(uint32_t));
- }
+/* This is a gfid based nameless lookup. Without a name, the hashed subvol
+ * cannot be calculated so a lookup is sent to all subvols.
+ */
+static int
+dht_do_discover(call_frame_t *frame, xlator_t *this, loc_t *loc)
+{
+ int ret;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int call_cnt = 0;
+ int op_errno = EINVAL;
+ int i = 0;
+ call_frame_t *discover_frame = NULL;
+
+ conf = this->private;
+ local = frame->local;
+
+ /* As we do not know if this is a file or directory, request
+ * both file and directory xattrs
+ */
+ ret = dht_set_file_xattr_req(this, loc, local->xattr_req);
+ if (ret) {
+ goto err;
+ }
+
+ ret = dht_set_dir_xattr_req(this, loc, local->xattr_req);
+ if (ret) {
+ goto err;
+ }
+
+ if (loc_is_root(loc)) {
+ /* Request the DHT commit hash xattr (trusted.glusterfs.dht.commithash)
+ * set on the brick root.
+ */
+ ret = dict_set_uint32(local->xattr_req, conf->commithash_xattr_name,
+ sizeof(uint32_t));
+ }
- call_cnt = conf->subvolume_cnt;
- local->call_cnt = call_cnt;
+ call_cnt = conf->subvolume_cnt;
+ local->call_cnt = call_cnt;
- local->layout = dht_layout_new (this, conf->subvolume_cnt);
+ local->layout = dht_layout_new(this, conf->subvolume_cnt);
- if (!local->layout) {
- op_errno = ENOMEM;
- goto err;
- }
+ if (!local->layout) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- gf_uuid_copy (local->gfid, loc->gfid);
+ gf_uuid_copy(local->gfid, loc->gfid);
- discover_frame = copy_frame (frame);
- if (!discover_frame) {
- op_errno = ENOMEM;
- goto err;
- }
+ discover_frame = copy_frame(frame);
+ if (!discover_frame) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- discover_frame->local = local;
- frame->local = NULL;
- local->main_frame = frame;
+ discover_frame->local = local;
+ frame->local = NULL;
+ local->main_frame = frame;
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND_COOKIE (discover_frame, dht_discover_cbk,
- conf->subvolumes[i], conf->subvolumes[i],
- conf->subvolumes[i]->fops->lookup,
- &local->loc, local->xattr_req);
- }
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND_COOKIE(discover_frame, dht_discover_cbk, conf->subvolumes[i],
+ conf->subvolumes[i],
+ conf->subvolumes[i]->fops->lookup, &local->loc,
+ local->xattr_req);
+ }
- return 0;
+ return 0;
err:
- DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL,
- NULL);
-
- return 0;
-}
-
-/* Get the value of key from dict in the bytewise and save in array after
- convert from network byte order to host byte order
-*/
-int32_t
-dht_dict_get_array (dict_t *dict, char *key, int32_t value[], int32_t size, int *errst)
-{
- void *ptr = NULL;
- int32_t len = -1;
- int32_t vindex = -1;
- int32_t err = -1;
- int ret = 0;
+ DHT_STACK_UNWIND(lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
- if (dict == NULL) {
- (*errst) = -1;
- return -EINVAL;
- }
- err = dict_get_ptr_and_len(dict, key, &ptr, &len);
- if (err != 0) {
- (*errst) = -1;
- return err;
- }
-
- if (len != (size * sizeof (int32_t))) {
- (*errst) = -1;
- return -EINVAL;
- }
-
- memset (value, 0, size * sizeof(int32_t));
- for (vindex = 0; vindex < size; vindex++) {
- value[vindex] = ntoh32(*((int32_t *)ptr + vindex));
- if (value[vindex] < 0)
- ret = -1;
- }
-
- return ret;
+ return 0;
}
-
/* Code to call syntask to heal custom xattr from hashed subvol
to non hashed subvol
*/
int
-dht_dir_xattr_heal (xlator_t *this, dht_local_t *local)
-{
- dht_local_t *copy_local = NULL;
- call_frame_t *copy = NULL;
- int ret = -1;
- char gfid_local[GF_UUID_BUF_SIZE] = {0};
-
- if (local->gfid) {
- gf_uuid_unparse(local->gfid, gfid_local);
+dht_dir_xattr_heal(xlator_t *this, dht_local_t *local, int *op_errno)
+{
+ dht_local_t *copy_local = NULL;
+ call_frame_t *copy = NULL;
+ int ret = -1;
+ char gfid_local[GF_UUID_BUF_SIZE] = {0};
+
+ if (gf_uuid_is_null(local->gfid)) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DIR_XATTR_HEAL_FAILED,
+ "No gfid exists for path %s "
+ "so healing xattr is not possible",
+ local->loc.path);
+ *op_errno = EIO;
+ goto out;
+ }
+
+ gf_uuid_unparse(local->gfid, gfid_local);
+ copy = create_frame(this, this->ctx->pool);
+ if (copy) {
+ copy_local = dht_local_init(copy, &(local->loc), NULL, 0);
+ if (!copy_local) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM,
+ DHT_MSG_DIR_XATTR_HEAL_FAILED,
+ "Memory allocation failed "
+ "for path %s gfid %s ",
+ local->loc.path, gfid_local);
+ *op_errno = ENOMEM;
+ DHT_STACK_DESTROY(copy);
} else {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DIR_XATTR_HEAL_FAILED,
- "No gfid exists for path %s "
- "so healing xattr is not possible",
- local->loc.path);
- goto out;
- }
-
- copy = create_frame (this, this->ctx->pool);
- if (copy) {
- copy_local = dht_local_init (copy, &(local->loc), NULL, 0);
- if (!copy_local) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_DIR_XATTR_HEAL_FAILED,
- "Memory allocation failed "
- "for path %s gfid %s ",
- local->loc.path, gfid_local);
- DHT_STACK_DESTROY (copy);
- } else {
- copy_local->stbuf = local->stbuf;
- gf_uuid_copy (copy_local->loc.gfid, local->gfid);
- copy_local->mds_subvol = local->mds_subvol;
- FRAME_SU_DO (copy, dht_local_t);
- ret = synctask_new (this->ctx->env, dht_dir_heal_xattrs,
- dht_dir_heal_xattrs_done,
- copy, copy);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_DIR_XATTR_HEAL_FAILED,
- "Synctask creation failed to heal xattr "
- "for path %s gfid %s ",
- local->loc.path, gfid_local);
- DHT_STACK_DESTROY (copy);
- }
- }
+ copy_local->stbuf = local->stbuf;
+ gf_uuid_copy(copy_local->loc.gfid, local->gfid);
+ copy_local->mds_subvol = local->mds_subvol;
+ FRAME_SU_DO(copy, dht_local_t);
+ ret = synctask_new(this->ctx->env, dht_dir_heal_xattrs,
+ dht_dir_heal_xattrs_done, copy, copy);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM,
+ DHT_MSG_DIR_XATTR_HEAL_FAILED,
+ "Synctask creation failed to heal xattr "
+ "for path %s gfid %s ",
+ local->loc.path, gfid_local);
+ *op_errno = ENOMEM;
+ DHT_STACK_DESTROY(copy);
+ }
}
+ }
out:
- return ret;
+ return ret;
}
+static int
+dht_needs_selfheal(call_frame_t *frame, xlator_t *this)
+{
+ dht_local_t *local = NULL;
+ dht_layout_t *layout = NULL;
+ int needs_selfheal = 0;
+ int ret = 0;
+ local = frame->local;
+ layout = local->layout;
-int
-dht_lookup_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, dict_t *xattr,
- struct iatt *postparent)
-{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- int this_call_cnt = 0;
- xlator_t *prev = NULL;
- dht_layout_t *layout = NULL;
- int ret = -1;
- int is_dir = 0;
- int32_t check_mds = 0;
- int errst = 0;
- char gfid_local[GF_UUID_BUF_SIZE] = {0};
- char gfid_node[GF_UUID_BUF_SIZE] = {0};
- int32_t mds_xattr_val[1] = {0};
-
- GF_VALIDATE_OR_GOTO ("dht", frame, out);
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
- GF_VALIDATE_OR_GOTO ("dht", this->private, out);
- GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+ if (local->need_attrheal || local->need_xattr_heal ||
+ local->need_selfheal) {
+ needs_selfheal = 1;
+ }
- local = frame->local;
- prev = cookie;
- conf = this->private;
+ ret = dht_layout_normalize(this, &local->loc, layout);
- layout = local->layout;
+ if (ret != 0) {
+ gf_msg_debug(this->name, 0, "fixing assignment on %s", local->loc.path);
+ needs_selfheal = 1;
+ }
+ return needs_selfheal;
+}
- if (!op_ret && gf_uuid_is_null (local->gfid)) {
- memcpy (local->gfid, stbuf->ia_gfid, 16);
- }
- if (local->gfid)
- gf_uuid_unparse(local->gfid, gfid_local);
+static int
+is_permission_different(ia_prot_t *prot1, ia_prot_t *prot2)
+{
+ if ((prot1->owner.read != prot2->owner.read) ||
+ (prot1->owner.write != prot2->owner.write) ||
+ (prot1->owner.exec != prot2->owner.exec) ||
+ (prot1->group.read != prot2->group.read) ||
+ (prot1->group.write != prot2->group.write) ||
+ (prot1->group.exec != prot2->group.exec) ||
+ (prot1->other.read != prot2->other.read) ||
+ (prot1->other.write != prot2->other.write) ||
+ (prot1->other.exec != prot2->other.exec) ||
+ (prot1->suid != prot2->suid) || (prot1->sgid != prot2->sgid) ||
+ (prot1->sticky != prot2->sticky)) {
+ return 1;
+ } else {
+ return 0;
+ }
+}
- /* Check if the gfid is different for file from other node */
- if (!op_ret && gf_uuid_compare (local->gfid, stbuf->ia_gfid)) {
+int
+dht_lookup_dir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode, struct iatt *stbuf,
+ dict_t *xattr, struct iatt *postparent)
+{
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int this_call_cnt = 0;
+ xlator_t *prev = NULL;
+ dht_layout_t *layout = NULL;
+ int ret = -1;
+ int is_dir = 0;
+ int32_t check_mds = 0;
+ int errst = 0;
+ char gfid_local[GF_UUID_BUF_SIZE] = {0};
+ char gfid_node[GF_UUID_BUF_SIZE] = {0};
+ int32_t mds_xattr_val[1] = {0};
+
+ GF_VALIDATE_OR_GOTO("dht", frame, out);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO("dht", this->private, out);
+ GF_VALIDATE_OR_GOTO("dht", cookie, out);
+
+ local = frame->local;
+ prev = cookie;
+ conf = this->private;
+
+ layout = local->layout;
+ gf_msg_debug(this->name, op_errno,
+ "%s: lookup on %s returned with op_ret = %d, op_errno = %d",
+ local->loc.path, prev->name, op_ret, op_errno);
+
+ /* The first successful lookup*/
+ if (!op_ret && gf_uuid_is_null(local->gfid)) {
+ memcpy(local->gfid, stbuf->ia_gfid, 16);
+ }
+ if (!gf_uuid_is_null(local->gfid)) {
+ gf_uuid_unparse(local->gfid, gfid_local);
+ }
+
+ /* Check if the gfid is different for file from other node */
+ if (!op_ret && gf_uuid_compare(local->gfid, stbuf->ia_gfid)) {
+ gf_uuid_unparse(stbuf->ia_gfid, gfid_node);
+
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_GFID_MISMATCH,
+ "%s: gfid different on %s."
+ " gfid local = %s, gfid subvol = %s",
+ local->loc.path, prev->name, gfid_local, gfid_node);
+ }
+
+ LOCK(&frame->lock);
+ {
+ /* TODO: assert equal mode on stbuf->st_mode and
+ local->stbuf->st_mode
+ else mkdir/chmod/chown and fix
+ */
+ ret = dht_layout_merge(this, layout, prev, op_ret, op_errno, xattr);
- gf_uuid_unparse(stbuf->ia_gfid, gfid_node);
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_GFID_MISMATCH,
- "%s: gfid different on %s."
- " gfid local = %s, gfid subvol = %s",
- local->loc.path, prev->name,
- gfid_local, gfid_node);
+ /* The GFID is missing on this subvol. Force a heal. */
+ if (op_errno == ENODATA) {
+ local->need_lookup_everywhere = 1;
+ }
+ goto unlock;
}
- LOCK (&frame->lock);
- {
- /* TODO: assert equal mode on stbuf->st_mode and
- local->stbuf->st_mode
-
- else mkdir/chmod/chown and fix
- */
- ret = dht_layout_merge (this, layout, prev, op_ret, op_errno,
- xattr);
-
- if (op_ret == -1) {
- local->op_errno = op_errno;
- gf_msg_debug (this->name, op_errno,
- "lookup of %s on %s returned error",
- local->loc.path, prev->name);
+ is_dir = check_is_dir(inode, stbuf, xattr);
+ if (!is_dir) {
+ gf_msg_debug(this->name, 0,
+ "%s: lookup on %s returned non dir 0%o"
+ "calling lookup_everywhere",
+ local->loc.path, prev->name, stbuf->ia_type);
- goto unlock;
- }
-
- is_dir = check_is_dir (inode, stbuf, xattr);
- if (!is_dir) {
+ local->need_lookup_everywhere = 1;
+ goto unlock;
+ }
- gf_msg_debug (this->name, 0,
- "lookup of %s on %s returned non"
- "dir 0%o"
- "calling lookup_everywhere",
- local->loc.path, prev->name,
- stbuf->ia_type);
+ local->op_ret = 0;
+ if (local->xattr == NULL) {
+ local->xattr = dict_ref(xattr);
+ } else {
+ dht_aggregate_xattr(local->xattr, xattr);
+ }
- local->need_selfheal = 1;
- goto unlock;
- }
+ if (__is_root_gfid(stbuf->ia_gfid)) {
+ ret = dht_dir_has_layout(xattr, conf->xattr_name);
+ if (ret >= 0) {
+ if (is_greater_time(local->prebuf.ia_ctime,
+ local->prebuf.ia_ctime_nsec,
+ stbuf->ia_ctime, stbuf->ia_ctime_nsec)) {
+ /* Choose source */
+ local->prebuf.ia_gid = stbuf->ia_gid;
+ local->prebuf.ia_uid = stbuf->ia_uid;
- local->op_ret = 0;
- if (local->xattr == NULL) {
- local->xattr = dict_ref (xattr);
- } else {
- dht_aggregate_xattr (local->xattr, xattr);
+ local->prebuf.ia_ctime = stbuf->ia_ctime;
+ local->prebuf.ia_ctime_nsec = stbuf->ia_ctime_nsec;
+ local->prebuf.ia_prot = stbuf->ia_prot;
}
+ }
+ }
- if (local->inode == NULL)
- local->inode = inode_ref (inode);
+ if (local->stbuf.ia_type != IA_INVAL) {
+ /* This is not the first subvol to respond
+ * Compare values to see if attrs need to be healed
+ */
+ if ((local->stbuf.ia_gid != stbuf->ia_gid) ||
+ (local->stbuf.ia_uid != stbuf->ia_uid) ||
+ (is_permission_different(&local->stbuf.ia_prot,
+ &stbuf->ia_prot))) {
+ local->need_attrheal = 1;
+ }
+ }
- dht_iatt_merge (this, &local->stbuf, stbuf, prev);
- dht_iatt_merge (this, &local->postparent, postparent, prev);
+ if (local->inode == NULL)
+ local->inode = inode_ref(inode);
- if (!dict_get (xattr, conf->mds_xattr_key)) {
- gf_msg_debug (this->name, 0,
- "Internal xattr %s is not present "
- " on path %s gfid is %s " ,
- conf->mds_xattr_key,
- local->loc.path, gfid_local);
- goto unlock;
- } else {
- /* Save mds subvol on inode ctx */
- ret = dht_inode_ctx_mdsvol_set (local->inode, this,
- prev);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_SET_INODE_CTX_FAILED,
- "Failed to set hashed subvol for %s vol is %s",
- local->loc.path, prev->name);
- }
- }
- check_mds = dht_dict_get_array (xattr, conf->mds_xattr_key,
- mds_xattr_val, 1, &errst);
- if ((check_mds < 0) && !errst) {
- local->mds_xattr = dict_ref (xattr);
- gf_msg_debug (this->name, 0,
- "Value of %s is not zero on hashed subvol "
- "so xattr needs to be heal on non hashed"
- " path is %s and vol name is %s "
- " gfid is %s" ,
- conf->mds_xattr_key,
- local->loc.path,
- prev->name, gfid_local);
- local->need_xattr_heal = 1;
- local->mds_subvol = prev;
- }
+ dht_iatt_merge(this, &local->stbuf, stbuf);
+ dht_iatt_merge(this, &local->postparent, postparent);
+ if (!dict_get(xattr, conf->mds_xattr_key)) {
+ gf_msg_debug(this->name, 0,
+ "%s: mds xattr %s is not present "
+ "on %s(gfid = %s)",
+ local->loc.path, conf->mds_xattr_key, prev->name,
+ gfid_local);
+ goto unlock;
}
-unlock:
- UNLOCK (&frame->lock);
+ /* Save the mds subvol info and stbuf. This is the value that will
+ * be used for healing
+ */
+ local->mds_subvol = prev;
+ local->mds_stbuf = *stbuf;
- this_call_cnt = dht_frame_return (frame);
+ /* Save mds subvol on inode ctx */
- if (is_last_call (this_call_cnt)) {
- /* No need to call xattr heal code if volume count is 1
- */
- if (conf->subvolume_cnt == 1)
- local->need_xattr_heal = 0;
+ ret = dht_inode_ctx_mdsvol_set(local->inode, this, prev);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_SET_INODE_CTX_FAILED,
+ "%s: Failed to set mds (%s)", local->loc.path, prev->name);
+ }
+ check_mds = dht_dict_get_array(xattr, conf->mds_xattr_key,
+ mds_xattr_val, 1, &errst);
+ if ((check_mds < 0) && !errst) {
+ /* Check if xattrs need to be healed on the directories */
+ local->mds_xattr = dict_ref(xattr);
+ gf_msg_debug(this->name, 0,
+ "%s: %s is not zero on %s. Xattrs need to be healed."
+ "(gfid = %s)",
+ local->loc.path, conf->mds_xattr_key, prev->name,
+ gfid_local);
+ local->need_xattr_heal = 1;
+ }
+ }
- /* Code to update all extended attributed from hashed subvol
- to local->xattr
- */
- if (local->need_xattr_heal && (local->mds_xattr)) {
- dht_dir_set_heal_xattr (this, local, local->xattr,
- local->mds_xattr, NULL, NULL);
- dict_unref (local->mds_xattr);
- local->mds_xattr = NULL;
- }
+unlock:
+ UNLOCK(&frame->lock);
- if (local->need_selfheal) {
- local->need_selfheal = 0;
- dht_lookup_everywhere (frame, this, &local->loc);
- return 0;
- }
+ this_call_cnt = dht_frame_return(frame);
- if (local->op_ret == 0) {
- ret = dht_layout_normalize (this, &local->loc, layout);
+ if (is_last_call(this_call_cnt)) {
+ /* If the mds subvol is not set correctly*/
+ if (!__is_root_gfid(local->gfid) &&
+ (!dict_get(local->xattr, conf->mds_xattr_key))) {
+ local->need_selfheal = 1;
+ }
- if (ret != 0) {
- gf_msg_debug (this->name, 0,
- "fixing assignment on %s",
- local->loc.path);
- goto selfheal;
- }
+ /* No need to call xattr heal code if volume count is 1
+ */
+ if (conf->subvolume_cnt == 1) {
+ local->need_xattr_heal = 0;
+ }
+
+ if (local->need_selfheal || local->need_lookup_everywhere) {
+ /* Set the gfid-req so posix will set the GFID*/
+ if (!gf_uuid_is_null(local->gfid)) {
+ /* Ok, this should _never_ happen */
+ ret = dict_set_static_bin(local->xattr_req, "gfid-req",
+ local->gfid, 16);
+ } else {
+ if (!gf_uuid_is_null(local->gfid_req))
+ ret = dict_set_static_bin(local->xattr_req, "gfid-req",
+ local->gfid_req, 16);
+ }
+ }
- dht_layout_set (this, local->inode, layout);
- if (!dict_get (local->xattr, conf->mds_xattr_key) ||
- local->need_xattr_heal)
- goto selfheal;
- }
+ if (local->need_lookup_everywhere) {
+ local->need_lookup_everywhere = 0;
+ dht_lookup_everywhere(frame, this, &local->loc);
+ return 0;
+ }
- if (local->inode) {
- dht_inode_ctx_time_update (local->inode, this,
- &local->stbuf, 1);
- }
+ if (local->op_ret == 0) {
+ if (dht_needs_selfheal(frame, this)) {
+ goto selfheal;
+ }
- if (local->loc.parent) {
- dht_inode_ctx_time_update (local->loc.parent, this,
- &local->postparent, 1);
- }
+ dht_layout_set(this, local->inode, layout);
+ if (local->inode) {
+ dht_inode_ctx_time_update(local->inode, this, &local->stbuf, 1);
+ }
- DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
- dht_set_fixed_dir_stat (&local->postparent);
- /* Delete mds xattr at the time of STACK UNWIND */
- GF_REMOVE_INTERNAL_XATTR (conf->mds_xattr_key, local->xattr);
- DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
- local->inode, &local->stbuf, local->xattr,
- &local->postparent);
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update(local->loc.parent, this,
+ &local->postparent, 1);
+ }
}
- return 0;
+ DHT_STRIP_PHASE1_FLAGS(&local->stbuf);
+ dht_set_fixed_dir_stat(&local->postparent);
+ /* Delete mds xattr at the time of STACK UNWIND */
+ if (local->xattr)
+ GF_REMOVE_INTERNAL_XATTR(conf->mds_xattr_key, local->xattr);
+
+ DHT_STACK_UNWIND(lookup, frame, local->op_ret, local->op_errno,
+ local->inode, &local->stbuf, local->xattr,
+ &local->postparent);
+ }
+
+ return 0;
selfheal:
- FRAME_SU_DO (frame, dht_local_t);
- ret = dht_selfheal_directory (frame, dht_lookup_selfheal_cbk,
- &local->loc, layout);
+ FRAME_SU_DO(frame, dht_local_t);
+ ret = dht_selfheal_directory(frame, dht_lookup_selfheal_cbk, &local->loc,
+ layout);
out:
- return ret;
+ return ret;
}
-int static
-is_permission_different (ia_prot_t *prot1, ia_prot_t *prot2)
-{
- if ((prot1->owner.read != prot2->owner.read) ||
- (prot1->owner.write != prot2->owner.write) ||
- (prot1->owner.exec != prot2->owner.exec) ||
- (prot1->group.read != prot2->group.read) ||
- (prot1->group.write != prot2->group.write) ||
- (prot1->group.exec != prot2->group.exec) ||
- (prot1->other.read != prot2->other.read) ||
- (prot1->other.write != prot2->other.write) ||
- (prot1->other.exec != prot2->other.exec) ||
- (prot1->suid != prot2->suid) ||
- (prot1->sgid != prot2->sgid) ||
- (prot1->sticky != prot2->sticky)) {
- return 1;
- } else {
- return 0;
- }
+static int
+dht_lookup_directory(call_frame_t *frame, xlator_t *this, loc_t *loc)
+{
+ int call_cnt = 0;
+ int i = 0;
+ dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ int ret = 0;
+
+ GF_VALIDATE_OR_GOTO("dht", frame, out);
+ GF_VALIDATE_OR_GOTO("dht", this, unwind);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, unwind);
+ GF_VALIDATE_OR_GOTO("dht", this->private, unwind);
+ GF_VALIDATE_OR_GOTO("dht", loc, unwind);
+
+ conf = this->private;
+ local = frame->local;
+
+ call_cnt = conf->subvolume_cnt;
+ local->call_cnt = call_cnt;
+
+ local->layout = dht_layout_new(this, conf->subvolume_cnt);
+ if (!local->layout) {
+ goto unwind;
+ }
+
+ if (local->xattr != NULL) {
+ dict_unref(local->xattr);
+ local->xattr = NULL;
+ }
+
+ if (!gf_uuid_is_null(local->gfid)) {
+ /* use this gfid in order to heal any missing ones */
+ ret = dict_set_gfuuid(local->xattr_req, "gfid-req", local->gfid, true);
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_DICT_SET_FAILED,
+ "%s: Failed to set dictionary value:"
+ " key = gfid-req",
+ local->loc.path);
+ }
+
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND_COOKIE(
+ frame, dht_lookup_dir_cbk, conf->subvolumes[i], conf->subvolumes[i],
+ conf->subvolumes[i]->fops->lookup, &local->loc, local->xattr_req);
+ }
+ return 0;
+unwind:
+ DHT_STACK_UNWIND(lookup, frame, -1, ENOMEM, NULL, NULL, NULL, NULL);
+out:
+ return 0;
}
int
-dht_revalidate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, dict_t *xattr,
- struct iatt *postparent)
-{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- xlator_t *prev = NULL;
- dht_layout_t *layout = NULL;
- dht_conf_t *conf = NULL;
- int ret = -1;
- int is_dir = 0;
- int is_linkfile = 0;
- int follow_link = 0;
- call_frame_t *copy = NULL;
- dht_local_t *copy_local = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
- uint32_t vol_commit_hash = 0;
- xlator_t *subvol = NULL;
- int32_t check_mds = 0;
- int errst = 0;
- int32_t mds_xattr_val[1] = {0};
-
- GF_VALIDATE_OR_GOTO ("dht", frame, err);
- GF_VALIDATE_OR_GOTO ("dht", this, err);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, err);
- GF_VALIDATE_OR_GOTO ("dht", cookie, err);
+dht_revalidate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode, struct iatt *stbuf,
+ dict_t *xattr, struct iatt *postparent)
+{
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ xlator_t *prev = NULL;
+ dht_layout_t *layout = NULL;
+ dht_conf_t *conf = NULL;
+ int ret = -1;
+ int is_dir = 0;
+ int is_linkfile = 0;
+ int follow_link = 0;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ uint32_t vol_commit_hash = 0;
+ xlator_t *subvol = NULL;
+ int32_t check_mds = 0;
+ int errst = 0, i = 0;
+ int32_t mds_xattr_val[1] = {0};
+
+ GF_VALIDATE_OR_GOTO("dht", frame, err);
+ GF_VALIDATE_OR_GOTO("dht", this, err);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, err);
+ GF_VALIDATE_OR_GOTO("dht", cookie, err);
+ GF_VALIDATE_OR_GOTO("dht", this->private, err);
+
+ local = frame->local;
+ prev = cookie;
+ conf = this->private;
+
+ if (!conf->vch_forced) {
+ /* Update the commithash value if available
+ */
+ ret = dict_get_uint32(xattr, conf->commithash_xattr_name,
+ &vol_commit_hash);
+ if (ret == 0) {
+ conf->vol_commit_hash = vol_commit_hash;
+ }
+ }
- local = frame->local;
- prev = cookie;
- conf = this->private;
- if (!conf)
- goto out;
+ gf_uuid_unparse(local->loc.gfid, gfid);
- if (!conf->vch_forced) {
- ret = dict_get_uint32 (xattr, conf->commithash_xattr_name,
- &vol_commit_hash);
- if (ret == 0) {
- conf->vol_commit_hash = vol_commit_hash;
- }
- }
+ gf_msg_debug(this->name, op_errno,
+ "%s: revalidate lookup on %s returned op_ret %d",
+ local->loc.path, prev->name, op_ret);
- gf_uuid_unparse (local->loc.gfid, gfid);
+ LOCK(&frame->lock);
+ {
+ if (gf_uuid_is_null(local->gfid)) {
+ memcpy(local->gfid, local->loc.gfid, 16);
+ }
- LOCK (&frame->lock);
- {
- if (gf_uuid_is_null (local->gfid)) {
- memcpy (local->gfid, local->loc.gfid, 16);
- }
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+
+ if ((op_errno != ENOTCONN) && (op_errno != ENOENT) &&
+ (op_errno != ESTALE)) {
+ gf_msg(this->name, GF_LOG_INFO, op_errno,
+ DHT_MSG_REVALIDATE_CBK_INFO,
+ "Revalidate: subvolume %s for %s "
+ "(gfid = %s) returned -1",
+ prev->name, local->loc.path, gfid);
+ }
+ if (op_errno == ESTALE) {
+ /* propagate the ESTALE to parent.
+ * setting local->return_estale would send
+ * ESTALE to parent. */
+ local->return_estale = 1;
+ }
- gf_msg_debug (this->name, op_errno,
- "revalidate lookup of %s "
- "returned with op_ret %d",
- local->loc.path, op_ret);
-
- if (op_ret == -1) {
- local->op_errno = op_errno;
-
- if ((op_errno != ENOTCONN)
- && (op_errno != ENOENT)
- && (op_errno != ESTALE)) {
- gf_msg (this->name, GF_LOG_INFO, op_errno,
- DHT_MSG_REVALIDATE_CBK_INFO,
- "Revalidate: subvolume %s for %s "
- "(gfid = %s) returned -1",
- prev->name, local->loc.path,
- gfid);
- }
- if (op_errno == ESTALE) {
- /* propagate the ESTALE to parent.
- * setting local->return_estale would send
- * ESTALE to parent. */
- local->return_estale = 1;
+ /* if it is ENOENT, we may have to do a
+ * 'lookup_everywhere()' to make sure
+ * the file is not migrated */
+ if (op_errno == ENOENT) {
+ if (IA_ISREG(local->loc.inode->ia_type)) {
+ gf_msg_debug(this->name, 0,
+ "found ENOENT for %s. "
+ "Setting "
+ "need_lookup_everywhere"
+ " flag to 1",
+ local->loc.path);
+
+ local->need_lookup_everywhere = 1;
+ } else if (IA_ISDIR(local->loc.inode->ia_type)) {
+ layout = local->layout;
+ for (i = 0; i < layout->cnt; i++) {
+ if (layout->list[i].xlator == prev) {
+ layout->list[i].err = op_errno;
+ break;
}
+ }
- /* if it is ENOENT, we may have to do a
- * 'lookup_everywhere()' to make sure
- * the file is not migrated */
- if (op_errno == ENOENT) {
- if (IA_ISREG (local->loc.inode->ia_type)) {
-
- gf_msg_debug (this->name, 0,
- "found ENOENT for %s. "
- "Setting "
- "need_lookup_everywhere"
- " flag to 1",
- local->loc.path);
-
- local->need_lookup_everywhere = 1;
- }
- }
- goto unlock;
+ local->need_selfheal = 1;
}
+ }
- if ((!IA_ISINVAL(local->inode->ia_type)) &&
- stbuf->ia_type != local->inode->ia_type) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_FILE_TYPE_MISMATCH,
- "mismatching filetypes 0%o v/s 0%o for %s,"
- " gfid = %s",
- (stbuf->ia_type), (local->inode->ia_type),
- local->loc.path, gfid);
+ /* The GFID is missing on this subvol. Lookup everywhere to force a
+ * gfid heal
+ */
+ if ((op_errno == ENODATA) &&
+ (IA_ISDIR(local->loc.inode->ia_type))) {
+ local->need_lookup_everywhere = 1;
+ }
- local->op_ret = -1;
- local->op_errno = EINVAL;
+ goto unlock;
+ }
- goto unlock;
+ if ((!IA_ISINVAL(local->inode->ia_type)) &&
+ stbuf->ia_type != local->inode->ia_type) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_FILE_TYPE_MISMATCH,
+ "mismatching filetypes 0%o v/s 0%o for %s,"
+ " gfid = %s",
+ (stbuf->ia_type), (local->inode->ia_type), local->loc.path,
+ gfid);
- }
+ local->op_ret = -1;
+ local->op_errno = EINVAL;
- layout = local->layout;
+ goto unlock;
+ }
- is_dir = check_is_dir (inode, stbuf, xattr);
- is_linkfile = check_is_linkfile (inode, stbuf, xattr,
- conf->link_xattr_name);
- if (is_linkfile) {
- follow_link = 1;
- goto unlock;
- }
- if (is_dir) {
- ret = dht_dir_has_layout (xattr, conf->xattr_name);
- if (ret >= 0) {
- if (is_greater_time(local->stbuf.ia_ctime,
- local->stbuf.ia_ctime_nsec,
- stbuf->ia_ctime,
- stbuf->ia_ctime_nsec)) {
- /* Choose source */
- local->prebuf.ia_gid = stbuf->ia_gid;
- local->prebuf.ia_uid = stbuf->ia_uid;
-
- if (__is_root_gfid (stbuf->ia_gfid))
- local->prebuf.ia_prot = stbuf->ia_prot;
- }
- }
+ layout = local->layout;
- if (local->stbuf.ia_type != IA_INVAL)
- {
- if ((local->stbuf.ia_gid != stbuf->ia_gid) ||
- (local->stbuf.ia_uid != stbuf->ia_uid) ||
- (__is_root_gfid (stbuf->ia_gfid) &&
- is_permission_different (&local->stbuf.ia_prot,
- &stbuf->ia_prot))) {
- local->need_selfheal = 1;
- }
- }
- if (!dict_get (xattr, conf->mds_xattr_key)) {
- gf_msg_debug (this->name, 0,
- "internal xattr %s is not present"
- " on path %s gfid is %s " ,
- conf->mds_xattr_key,
- local->loc.path, gfid);
- } else {
- check_mds = dht_dict_get_array (xattr, conf->mds_xattr_key,
- mds_xattr_val, 1, &errst);
- local->mds_subvol = prev;
- local->mds_stbuf.ia_gid = stbuf->ia_gid;
- local->mds_stbuf.ia_uid = stbuf->ia_uid;
- local->mds_stbuf.ia_prot = stbuf->ia_prot;
-
- /* save mds subvol on inode ctx */
- ret = dht_inode_ctx_mdsvol_set (local->inode, this,
- prev);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_SET_INODE_CTX_FAILED,
- "Failed to set MDS subvol for %s vol is %s",
- local->loc.path, prev->name);
- }
- if ((check_mds < 0) && !errst) {
- local->mds_xattr = dict_ref (xattr);
- gf_msg_debug (this->name, 0,
- "Value of %s is not zero on "
- "hashed subvol so xattr needs to"
- " be healed on non hashed"
- " path is %s and vol name is %s "
- " gfid is %s" ,
- conf->mds_xattr_key,
- local->loc.path,
- prev->name, gfid);
- local->need_xattr_heal = 1;
- }
- }
- ret = dht_layout_dir_mismatch (this, layout,
- prev, &local->loc,
- xattr);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LAYOUT_MISMATCH,
- "Mismatching layouts for %s, gfid = %s",
- local->loc.path, gfid);
-
- local->layout_mismatch = 1;
-
- goto unlock;
- }
- }
+ is_dir = check_is_dir(inode, stbuf, xattr);
+ is_linkfile = check_is_linkfile(inode, stbuf, xattr,
+ conf->link_xattr_name);
+ if (is_linkfile) {
+ follow_link = 1;
+ goto unlock;
+ }
+ if (is_dir) {
+ ret = dht_dir_has_layout(xattr, conf->xattr_name);
+ if (ret >= 0) {
+ if (is_greater_time(local->prebuf.ia_ctime,
+ local->prebuf.ia_ctime_nsec,
+ stbuf->ia_ctime, stbuf->ia_ctime_nsec)) {
+ /* Choose source */
+ local->prebuf.ia_gid = stbuf->ia_gid;
+ local->prebuf.ia_uid = stbuf->ia_uid;
+ local->prebuf.ia_ctime = stbuf->ia_ctime;
+ local->prebuf.ia_ctime_nsec = stbuf->ia_ctime_nsec;
- /* Update stbuf from the servers where layout is present. This
- * is an indication that the server is not a newly added brick.
- * Merging stbuf from newly added brick may result in the added
- * brick being the source of heal for uid/gid */
- if (!is_dir || (is_dir &&
- dht_dir_has_layout (xattr, conf->xattr_name) >= 0)
- || conf->subvolume_cnt == 1) {
+ if (__is_root_gfid(stbuf->ia_gfid))
+ local->prebuf.ia_prot = stbuf->ia_prot;
+ }
+ }
- dht_iatt_merge (this, &local->stbuf, stbuf, prev);
- dht_iatt_merge (this, &local->postparent, postparent,
- prev);
- } else {
- /* copy the gfid anyway */
- gf_uuid_copy (local->stbuf.ia_gfid, stbuf->ia_gfid);
+ if (local->stbuf.ia_type != IA_INVAL) {
+ if ((local->stbuf.ia_gid != stbuf->ia_gid) ||
+ (local->stbuf.ia_uid != stbuf->ia_uid) ||
+ is_permission_different(&local->stbuf.ia_prot,
+ &stbuf->ia_prot)) {
+ local->need_attrheal = 1;
}
+ }
- local->op_ret = 0;
+ if (!dict_get(xattr, conf->mds_xattr_key)) {
+ gf_msg_debug(this->name, 0,
+ "%s: internal xattr %s is not present"
+ " on subvol %s(gfid is %s)",
+ local->loc.path, conf->mds_xattr_key, prev->name,
+ gfid);
+ } else {
+ check_mds = dht_dict_get_array(xattr, conf->mds_xattr_key,
+ mds_xattr_val, 1, &errst);
+ local->mds_subvol = prev;
+ local->mds_stbuf.ia_gid = stbuf->ia_gid;
+ local->mds_stbuf.ia_uid = stbuf->ia_uid;
+ local->mds_stbuf.ia_prot = stbuf->ia_prot;
- if (!local->xattr) {
- local->xattr = dict_ref (xattr);
- } else if (is_dir) {
- dht_aggregate_xattr (local->xattr, xattr);
+ /* save mds subvol on inode ctx */
+ ret = dht_inode_ctx_mdsvol_set(local->inode, this, prev);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ DHT_MSG_SET_INODE_CTX_FAILED,
+ "Failed to set MDS subvol for %s vol is %s",
+ local->loc.path, prev->name);
+ }
+ if ((check_mds < 0) && !errst) {
+ /* Check if xattrs need to be healed on the directory
+ */
+ local->mds_xattr = dict_ref(xattr);
+ gf_msg_debug(this->name, 0,
+ "Value of %s is not zero on "
+ "hashed subvol so xattr needs to"
+ " be healed on non hashed"
+ " path is %s and vol name is %s "
+ " gfid is %s",
+ conf->mds_xattr_key, local->loc.path,
+ prev->name, gfid);
+ local->need_xattr_heal = 1;
}
+ }
+ ret = dht_layout_dir_mismatch(this, layout, prev, &local->loc,
+ xattr);
+ if (ret != 0) {
+ /* In memory layout does not match on-disk layout.
+ */
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_LAYOUT_MISMATCH,
+ "Mismatching layouts for %s, gfid = %s", local->loc.path,
+ gfid);
+
+ local->layout_mismatch = 1;
+
+ goto unlock;
+ }
}
+
+ gf_uuid_copy(local->stbuf.ia_gfid, stbuf->ia_gfid);
+ dht_iatt_merge(this, &local->stbuf, stbuf);
+ dht_iatt_merge(this, &local->postparent, postparent);
+
+ local->op_ret = 0;
+
+ if (!local->xattr) {
+ local->xattr = dict_ref(xattr);
+ } else if (is_dir) {
+ dht_aggregate_xattr(local->xattr, xattr);
+ }
+ }
unlock:
- UNLOCK (&frame->lock);
+ UNLOCK(&frame->lock);
- if (follow_link) {
- gf_uuid_copy (local->gfid, stbuf->ia_gfid);
+ if (follow_link) {
+ /* Found a linkto file. Follow it to see if the target file exists
+ */
+ gf_uuid_copy(local->gfid, stbuf->ia_gfid);
- subvol = dht_linkfile_subvol (this, inode, stbuf, xattr);
- if (!subvol) {
- op_errno = ESTALE;
- local->op_ret = -1;
- } else {
+ subvol = dht_linkfile_subvol(this, inode, stbuf, xattr);
+ if (!subvol) {
+ op_errno = ESTALE;
+ local->op_ret = -1;
+ } else {
+ STACK_WIND_COOKIE(frame, dht_lookup_linkfile_cbk, subvol, subvol,
+ subvol->fops->lookup, &local->loc,
+ local->xattr_req);
+ return 0;
+ }
+ }
- STACK_WIND_COOKIE (frame, dht_lookup_linkfile_cbk,
- subvol, subvol, subvol->fops->lookup,
- &local->loc, local->xattr_req);
- return 0;
+ this_call_cnt = dht_frame_return(frame);
+
+ if (is_last_call(this_call_cnt)) {
+ if (!IA_ISDIR(local->stbuf.ia_type) &&
+ (local->hashed_subvol != local->cached_subvol) &&
+ (local->stbuf.ia_nlink == 1) &&
+ (conf && conf->unhashed_sticky_bit)) {
+ local->stbuf.ia_prot.sticky = 1;
+ }
+ /* No need to call heal code if volume count is 1
+ */
+ if (conf->subvolume_cnt == 1)
+ local->need_xattr_heal = 0;
+
+ if (IA_ISDIR(local->stbuf.ia_type)) {
+ /* No mds xattr found. Trigger a heal to set it */
+ if (!__is_root_gfid(local->loc.inode->gfid) &&
+ (!dict_get(local->xattr, conf->mds_xattr_key)))
+ local->need_selfheal = 1;
+
+ if (dht_needs_selfheal(frame, this)) {
+ if (!__is_root_gfid(local->loc.inode->gfid)) {
+ if (local->mds_subvol) {
+ local->stbuf.ia_gid = local->mds_stbuf.ia_gid;
+ local->stbuf.ia_uid = local->mds_stbuf.ia_uid;
+ local->stbuf.ia_prot = local->mds_stbuf.ia_prot;
+ }
+ } else {
+ local->stbuf.ia_gid = local->prebuf.ia_gid;
+ local->stbuf.ia_uid = local->prebuf.ia_uid;
+ local->stbuf.ia_prot = local->prebuf.ia_prot;
}
+
+ layout = local->layout;
+ dht_selfheal_directory(frame, dht_lookup_selfheal_cbk,
+ &local->loc, layout);
+ return 0;
+ }
}
-out:
- this_call_cnt = dht_frame_return (frame);
-
- if (is_last_call (this_call_cnt)) {
- if (!IA_ISDIR (local->stbuf.ia_type)
- && (local->hashed_subvol != local->cached_subvol)
- && (local->stbuf.ia_nlink == 1)
- && (conf && conf->unhashed_sticky_bit)) {
- local->stbuf.ia_prot.sticky = 1;
- }
- /* No need to call heal code if volume count is 1
- */
- if (conf->subvolume_cnt == 1)
- local->need_xattr_heal = 0;
-
- if (IA_ISDIR (local->stbuf.ia_type)) {
- /* Code to update all extended attributed from hashed
- subvol to local->xattr and call heal code to heal
- custom xattr from hashed subvol to non-hashed subvol
- */
- if (local->need_xattr_heal && (local->mds_xattr)) {
- dht_dir_set_heal_xattr (this, local,
- local->xattr,
- local->mds_xattr, NULL,
- NULL);
- dict_unref (local->mds_xattr);
- local->mds_xattr = NULL;
- local->need_xattr_heal = 0;
- ret = dht_dir_xattr_heal (this, local);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR,
- ret, DHT_MSG_DIR_XATTR_HEAL_FAILED,
- "xattr heal failed for directory %s "
- " gfid %s ", local->loc.path,
- gfid);
- } else {
- /* Call function to save hashed subvol on inode
- ctx if internal mds xattr is not present and
- all subvols are up
- */
- if (inode && !__is_root_gfid (inode->gfid) &&
- (!local->op_ret))
- (void) dht_common_mark_mdsxattr (frame, NULL, 1);
- }
- }
- if (local->need_selfheal) {
- local->need_selfheal = 0;
- if (!__is_root_gfid (inode->gfid)) {
- gf_uuid_copy (local->gfid, local->mds_stbuf.ia_gfid);
- if (local->mds_stbuf.ia_gid || local->mds_stbuf.ia_uid) {
- local->stbuf.ia_gid = local->mds_stbuf.ia_gid;
- local->stbuf.ia_uid = local->mds_stbuf.ia_uid;
- }
- } else {
- gf_uuid_copy (local->gfid, local->stbuf.ia_gfid);
- local->stbuf.ia_gid = local->prebuf.ia_gid;
- local->stbuf.ia_uid = local->prebuf.ia_uid;
- local->stbuf.ia_prot = local->prebuf.ia_prot;
- }
+ if (local->layout_mismatch) {
+ /* Found layout mismatch in the directory, need to
+ fix this in the inode context */
+ dht_layout_unref(this, local->layout);
+ local->layout = NULL;
+ dht_lookup_directory(frame, this, &local->loc);
+ return 0;
+ }
- copy = create_frame (this, this->ctx->pool);
- if (copy) {
- copy_local = dht_local_init (copy, &local->loc,
- NULL, 0);
- if (!copy_local)
- goto cont;
- copy_local->stbuf = local->stbuf;
- copy_local->mds_stbuf = local->mds_stbuf;
- copy_local->mds_subvol = local->mds_subvol;
- copy->local = copy_local;
- FRAME_SU_DO (copy, dht_local_t);
- ret = synctask_new (this->ctx->env,
- dht_dir_attr_heal,
- dht_dir_attr_heal_done,
- copy, copy);
- }
- }
-cont:
- if (local->layout_mismatch) {
- /* Found layout mismatch in the directory, need to
- fix this in the inode context */
- dht_layout_unref (this, local->layout);
- local->layout = NULL;
- dht_lookup_directory (frame, this, &local->loc);
- return 0;
- }
+ if (local->need_lookup_everywhere) {
+ /* As the current layout gave ENOENT error, we would
+ need a new layout */
+ dht_layout_unref(this, local->layout);
+ local->layout = NULL;
- if (local->need_lookup_everywhere) {
- /* As the current layout gave ENOENT error, we would
- need a new layout */
- dht_layout_unref (this, local->layout);
- local->layout = NULL;
-
- /* We know that current cached subvol is no more
- valid, get the new one */
- local->cached_subvol = NULL;
- dht_lookup_everywhere (frame, this, &local->loc);
- return 0;
- }
- if (local->return_estale) {
- local->op_ret = -1;
- local->op_errno = ESTALE;
+ /* We know that current cached subvol is no longer
+ valid, get the new one */
+ local->cached_subvol = NULL;
+ if (local->xattr_req) {
+ if (!gf_uuid_is_null(local->gfid)) {
+ ret = dict_set_static_bin(local->xattr_req, "gfid-req",
+ local->gfid, 16);
}
+ }
- if (local->loc.parent) {
- dht_inode_ctx_time_update (local->loc.parent, this,
- &local->postparent, 1);
- }
+ dht_lookup_everywhere(frame, this, &local->loc);
+ return 0;
+ }
+ if (local->return_estale) {
+ local->op_ret = -1;
+ local->op_errno = ESTALE;
+ }
- DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
- dht_set_fixed_dir_stat (&local->postparent);
-
- /* local->stbuf is udpated only from subvols which have a layout
- * The reason is to avoid choosing attr heal source from newly
- * added bricks. In case e.g we have only one subvol and for
- * some reason layout is not present on it, then local->stbuf
- * will be EINVAL. This is an indication that the subvols
- * active in the cluster do not have layouts on disk.
- * Unwind with ESTALE to trigger a fresh lookup */
- if (is_dir && local->stbuf.ia_type == IA_INVAL) {
- local->op_ret = -1;
- local->op_errno = ESTALE;
- }
- /* Delete mds xattr at the time of STACK UNWIND */
- GF_REMOVE_INTERNAL_XATTR (conf->mds_xattr_key, local->xattr);
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update(local->loc.parent, this,
+ &local->postparent, 1);
+ }
+
+ DHT_STRIP_PHASE1_FLAGS(&local->stbuf);
+ dht_set_fixed_dir_stat(&local->postparent);
- DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
- local->inode, &local->stbuf, local->xattr,
- &local->postparent);
+ /* local->stbuf is updated only from subvols which have a layout
+ * The reason is to avoid choosing attr heal source from newly
+ * added bricks. In case e.g we have only one subvol and for
+ * some reason layout is not present on it, then local->stbuf
+ * will be EINVAL. This is an indication that the subvols
+ * active in the cluster do not have layouts on disk.
+ * Unwind with ESTALE to trigger a fresh lookup */
+ if (is_dir && local->stbuf.ia_type == IA_INVAL) {
+ local->op_ret = -1;
+ local->op_errno = ESTALE;
}
+ /* Delete mds xattr at the time of STACK UNWIND */
+ if (local->xattr)
+ GF_REMOVE_INTERNAL_XATTR(conf->mds_xattr_key, local->xattr);
+
+ DHT_STACK_UNWIND(lookup, frame, local->op_ret, local->op_errno,
+ local->inode, &local->stbuf, local->xattr,
+ &local->postparent);
+ }
err:
- return ret;
+ return ret;
}
+static int
+dht_lookup_linkfile_create_cbk(call_frame_t *frame, void *cooie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ xlator_t *cached_subvol = NULL;
+ dht_conf_t *conf = NULL;
+ int ret = -1;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
-int
-dht_lookup_linkfile_create_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
-{
- dht_local_t *local = NULL;
- xlator_t *cached_subvol = NULL;
- dht_conf_t *conf = NULL;
- int ret = -1;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- GF_VALIDATE_OR_GOTO ("dht", frame, out);
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
- GF_VALIDATE_OR_GOTO ("dht", this->private, out);
- GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+ GF_VALIDATE_OR_GOTO("dht", frame, out);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO("dht", this->private, out);
- local = frame->local;
- cached_subvol = local->cached_subvol;
- conf = this->private;
+ local = frame->local;
+ cached_subvol = local->cached_subvol;
+ conf = this->private;
- gf_uuid_unparse(local->loc.gfid, gfid);
+ gf_uuid_unparse(local->loc.gfid, gfid);
- ret = dht_layout_preset (this, local->cached_subvol, local->loc.inode);
- if (ret < 0) {
- gf_msg_debug (this->name, EINVAL,
- "Failed to set layout for subvolume %s, "
- "(gfid = %s)",
- cached_subvol ? cached_subvol->name : "<nil>",
- gfid);
- local->op_ret = -1;
- local->op_errno = EINVAL;
- goto unwind;
- }
+ if (local->locked)
+ dht_unlock_namespace(frame, &local->lock[0]);
- local->op_ret = 0;
- if ((local->stbuf.ia_nlink == 1)
- && (conf && conf->unhashed_sticky_bit)) {
- local->stbuf.ia_prot.sticky = 1;
- }
+ ret = dht_layout_preset(this, local->cached_subvol, local->loc.inode);
+ if (ret < 0) {
+ gf_msg_debug(this->name, EINVAL,
+ "Failed to set layout for subvolume %s, "
+ "(gfid = %s)",
+ cached_subvol ? cached_subvol->name : "<nil>", gfid);
+ local->op_ret = -1;
+ local->op_errno = EINVAL;
+ goto unwind;
+ }
- if (local->loc.parent) {
- dht_inode_ctx_time_update (local->loc.parent, this,
- postparent, 1);
- }
+ local->op_ret = 0;
+ if ((local->stbuf.ia_nlink == 1) && (conf && conf->unhashed_sticky_bit)) {
+ local->stbuf.ia_prot.sticky = 1;
+ }
-unwind:
- gf_msg_debug (this->name, 0,
- "creation of linkto on hashed subvol:%s, "
- "returned with op_ret %d and op_errno %d: %s",
- local->hashed_subvol->name,
- op_ret, op_errno, uuid_utoa (local->loc.gfid));
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update(local->loc.parent, this, postparent, 1);
+ }
- if (local->linked == _gf_true)
- dht_linkfile_attr_heal (frame, this);
+unwind:
+ gf_msg_debug(this->name, 0,
+ "creation of linkto on hashed subvol:%s, "
+ "returned with op_ret %d and op_errno %d: %s",
+ local->hashed_subvol->name, op_ret, op_errno,
+ uuid_utoa(local->loc.gfid));
+ if (local->linked == _gf_true)
+ dht_linkfile_attr_heal(frame, this);
- dht_set_fixed_dir_stat (&local->postparent);
+ dht_set_fixed_dir_stat(&local->postparent);
- DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
- DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
- local->inode, &local->stbuf, local->xattr,
- &local->postparent);
+ DHT_STRIP_PHASE1_FLAGS(&local->stbuf);
+ DHT_STACK_UNWIND(lookup, frame, local->op_ret, local->op_errno,
+ local->inode, &local->stbuf, local->xattr,
+ &local->postparent);
out:
- return ret;
+ return ret;
}
-int
-dht_lookup_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+static int
+dht_lookup_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- int this_call_cnt = 0;
- dht_local_t *local = NULL;
- const char *path = NULL;
+ int this_call_cnt = 0;
+ dht_local_t *local = NULL;
+ const char *path = NULL;
- local = (dht_local_t*)frame->local;
- path = local->loc.path;
- FRAME_SU_UNDO (frame, dht_local_t);
+ local = (dht_local_t *)frame->local;
+ path = local->loc.path;
+ FRAME_SU_UNDO(frame, dht_local_t);
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_UNLINK_LOOKUP_INFO, "lookup_unlink returned with "
- "op_ret -> %d and op-errno -> %d for %s", op_ret, op_errno,
- ((path == NULL)? "null" : path ));
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_UNLINK_LOOKUP_INFO,
+ "lookup_unlink returned with "
+ "op_ret -> %d and op-errno -> %d for %s",
+ op_ret, op_errno, ((path == NULL) ? "null" : path));
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- dht_lookup_everywhere_done (frame, this);
- }
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt)) {
+ dht_lookup_everywhere_done(frame, this);
+ }
- return 0;
+ return 0;
}
-int
-dht_lookup_unlink_of_false_linkto_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+static int
+dht_lookup_unlink_of_false_linkto_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int op_ret, int op_errno,
+ struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- int this_call_cnt = 0;
- dht_local_t *local = NULL;
- const char *path = NULL;
+ int this_call_cnt = 0;
+ dht_local_t *local = NULL;
+ const char *path = NULL;
- local = (dht_local_t*)frame->local;
- path = local->loc.path;
+ local = (dht_local_t *)frame->local;
+ path = local->loc.path;
- FRAME_SU_UNDO (frame, dht_local_t);
+ FRAME_SU_UNDO(frame, dht_local_t);
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_UNLINK_LOOKUP_INFO, "lookup_unlink returned with "
- "op_ret -> %d and op-errno -> %d for %s", op_ret, op_errno,
- ((path == NULL)? "null" : path ));
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_UNLINK_LOOKUP_INFO,
+ "lookup_unlink returned with "
+ "op_ret -> %d and op-errno -> %d for %s",
+ op_ret, op_errno, ((path == NULL) ? "null" : path));
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt)) {
+ if ((op_ret == 0) || ((op_errno != EBUSY) && (op_errno != ENOTCONN))) {
+ dht_lookup_everywhere_done(frame, this);
+ } else {
+ /*When dht_lookup_everywhere is performed, one cached
+ *and one hashed file was found and hashed file does
+ *not point to the above mentioned cached node. So it
+ *was considered as stale and an unlink was performed.
+ *But unlink fails. So may be rebalance is in progress.
+ *now ideally we have two data-files. One obtained during
+ *lookup_everywhere and one where unlink-failed. So
+ *at this point in time we cannot decide which one to
+ *choose because there are chances of first cached
+ *file is truncated after rebalance and if it is chosen
+ *as cached node, application will fail. So return EIO.*/
+
+ if (op_errno == EBUSY) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno,
+ DHT_MSG_UNLINK_FAILED,
+ "Could not unlink the linkto file as "
+ "either fd is open and/or linkto xattr "
+ "is set for %s",
+ ((path == NULL) ? "null" : path));
+ }
+ DHT_STACK_UNWIND(lookup, frame, -1, EIO, NULL, NULL, NULL, NULL);
+ }
+ }
- if (op_ret == 0) {
- dht_lookup_everywhere_done (frame, this);
- } else {
- /*When dht_lookup_everywhere is performed, one cached
- *and one hashed file was found and hashed file does
- *not point to the above mentioned cached node. So it
- *was considered as stale and an unlink was performed.
- *But unlink fails. So may be rebalance is in progress.
- *now ideally we have two data-files. One obtained during
- *lookup_everywhere and one where unlink-failed. So
- *at this point in time we cannot decide which one to
- *choose because there are chances of first cached
- *file is truncated after rebalance and if it is chosen
- *as cached node, application will fail. So return EIO.*/
-
- if (op_errno == EBUSY) {
-
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- DHT_MSG_UNLINK_FAILED,
- "Could not unlink the linkto file as "
- "either fd is open and/or linkto xattr "
- "is set for %s",
- ((path == NULL)? "null":path));
+ return 0;
+}
- }
- DHT_STACK_UNWIND (lookup, frame, -1, EIO, NULL, NULL,
- NULL, NULL);
+static int
+dht_lookup_unlink_stale_linkto_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int op_ret, int op_errno,
+ struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ const char *path = NULL;
- }
- }
+ /* NOTE:
+ * If stale file unlink fails either there is an open-fd or is not an
+ * dht-linkto-file then posix_unlink returns EBUSY, which is overwritten
+ * to ENOENT
+ */
- return 0;
+ local = frame->local;
+
+ if (local) {
+ FRAME_SU_UNDO(frame, dht_local_t);
+ if (local->loc.path)
+ path = local->loc.path;
+ }
+
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_UNLINK_LOOKUP_INFO,
+ "Returned with op_ret %d and "
+ "op_errno %d for %s",
+ op_ret, op_errno, ((path == NULL) ? "null" : path));
+
+ DHT_STACK_UNWIND(lookup, frame, -1, ENOENT, NULL, NULL, NULL, NULL);
+
+ return 0;
}
-int
-dht_lookup_unlink_stale_linkto_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+static int
+dht_fill_dict_to_avoid_unlink_of_migrating_file(dict_t *dict)
{
+ int ret = 0;
- dht_local_t *local = NULL;
- const char *path = NULL;
+ ret = dict_set_int32_sizen(dict, DHT_SKIP_NON_LINKTO_UNLINK, 1);
- /* NOTE:
- * If stale file unlink fails either there is an open-fd or is not an
- * dht-linkto-file then posix_unlink returns EBUSY, which is overwritten
- * to ENOENT
- */
+ if (ret)
+ return -1;
- local = frame->local;
+ ret = dict_set_int32_sizen(dict, DHT_SKIP_OPEN_FD_UNLINK, 1);
- if (local && local->loc.path)
- path = local->loc.path;
+ if (ret)
+ return -1;
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_UNLINK_LOOKUP_INFO,
- "Returned with op_ret %d and "
- "op_errno %d for %s", op_ret, op_errno,
- ((path==NULL)?"null":path));
+ return 0;
+}
- FRAME_SU_UNDO (frame, dht_local_t);
- DHT_STACK_UNWIND (lookup, frame, -1, ENOENT, NULL, NULL, NULL,
- NULL);
+static int32_t
+dht_linkfile_create_lookup_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ inode_t *inode, struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
+{
+ dht_local_t *local = NULL;
+ int call_cnt = 0, ret = 0;
+ xlator_t *subvol = NULL;
+ uuid_t gfid = {
+ 0,
+ };
+ char gfid_str[GF_UUID_BUF_SIZE] = {0};
- return 0;
-}
+ subvol = cookie;
+ local = frame->local;
-int
-dht_fill_dict_to_avoid_unlink_of_migrating_file (dict_t *dict) {
+ if (subvol == local->hashed_subvol) {
+ if ((op_ret == 0) || (op_errno != ENOENT))
+ local->dont_create_linkto = _gf_true;
+ } else {
+ if (gf_uuid_is_null(local->gfid))
+ gf_uuid_copy(gfid, local->loc.gfid);
+ else
+ gf_uuid_copy(gfid, local->gfid);
+
+ if ((op_ret == 0) && gf_uuid_compare(gfid, buf->ia_gfid)) {
+ gf_uuid_unparse(gfid, gfid_str);
+ gf_msg_debug(this->name, 0,
+ "gfid (%s) different on cached subvol "
+ "(%s) and looked up inode (%s), not "
+ "creating linkto",
+ uuid_utoa(buf->ia_gfid), subvol->name, gfid_str);
+ local->dont_create_linkto = _gf_true;
+ } else if (op_ret == -1) {
+ local->dont_create_linkto = _gf_true;
+ }
+ }
+
+ call_cnt = dht_frame_return(frame);
+ if (is_last_call(call_cnt)) {
+ if (local->dont_create_linkto)
+ goto no_linkto;
+ else {
+ gf_msg_debug(this->name, 0,
+ "Creating linkto file on %s(hash) to "
+ "%s on %s (gfid = %s)",
+ local->hashed_subvol->name, local->loc.path,
+ local->cached_subvol->name, gfid_str);
- int ret = 0;
- xlator_t *this = NULL;
- char *linktoskip_key = NULL;
+ ret = dht_linkfile_create(frame, dht_lookup_linkfile_create_cbk,
+ this, local->cached_subvol,
+ local->hashed_subvol, &local->loc);
- this = THIS;
- GF_VALIDATE_OR_GOTO ("dht", this, err);
+ if (ret < 0)
+ goto no_linkto;
+ }
+ }
- if (dht_is_tier_xlator (this))
- linktoskip_key = TIER_SKIP_NON_LINKTO_UNLINK;
- else
- linktoskip_key = DHT_SKIP_NON_LINKTO_UNLINK;
+ return 0;
- ret = dict_set_int32 (dict, linktoskip_key, 1);
+no_linkto:
+ gf_msg_debug(this->name, 0,
+ "skipped linkto creation (path:%s) (gfid:%s) "
+ "(hashed-subvol:%s) (cached-subvol:%s)",
+ local->loc.path, gfid_str, local->hashed_subvol->name,
+ local->cached_subvol->name);
- if (ret)
- goto err;
+ dht_lookup_linkfile_create_cbk(frame, NULL, this, 0, 0, local->loc.inode,
+ &local->stbuf, &local->preparent,
+ &local->postparent, local->xattr);
+ return 0;
+}
- ret = dict_set_int32 (dict, DHT_SKIP_OPEN_FD_UNLINK, 1);
+static int32_t
+dht_call_lookup_linkfile_create(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ int i = 0;
+ xlator_t *subvol = NULL;
- if (ret)
- goto err;
+ local = frame->local;
+ if (gf_uuid_is_null(local->gfid))
+ gf_uuid_unparse(local->loc.gfid, gfid);
+ else
+ gf_uuid_unparse(local->gfid, gfid);
+ if (op_ret < 0) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "protecting namespace failed, skipping linkto "
+ "creation (path:%s)(gfid:%s)(hashed-subvol:%s)"
+ "(cached-subvol:%s)",
+ local->loc.path, gfid, local->hashed_subvol->name,
+ local->cached_subvol->name);
+ goto err;
+ }
- return 0;
+ local->locked = _gf_true;
-err:
- return -1;
+ local->call_cnt = 2;
+
+ for (i = 0; i < 2; i++) {
+ subvol = (subvol == NULL) ? local->hashed_subvol : local->cached_subvol;
+
+ STACK_WIND_COOKIE(frame, dht_linkfile_create_lookup_cbk, subvol, subvol,
+ subvol->fops->lookup, &local->loc, NULL);
+ }
+
+ return 0;
+err:
+ dht_lookup_linkfile_create_cbk(frame, NULL, this, 0, 0, local->loc.inode,
+ &local->stbuf, &local->preparent,
+ &local->postparent, local->xattr);
+ return 0;
}
+
/* Rebalance is performed from cached_node to hashed_node. Initial cached_node
* contains a non-linkto file. After migration it is converted to linkto and
* then unlinked. And at hashed_subvolume, first a linkto file is present,
@@ -2174,753 +2335,654 @@ err:
* dht_lookup_everywhere_done takes decision based on any of the above case
*/
-int
-dht_lookup_everywhere_done (call_frame_t *frame, xlator_t *this)
-{
- int ret = 0;
- dht_local_t *local = NULL;
- xlator_t *hashed_subvol = NULL;
- xlator_t *cached_subvol = NULL;
- dht_layout_t *layout = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
- gf_boolean_t found_non_linkto_on_hashed = _gf_false;
+static int
+dht_lookup_everywhere_done(call_frame_t *frame, xlator_t *this)
+{
+ int ret = 0;
+ dht_local_t *local = NULL;
+ xlator_t *hashed_subvol = NULL;
+ xlator_t *cached_subvol = NULL;
+ dht_layout_t *layout = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ gf_boolean_t found_non_linkto_on_hashed = _gf_false;
+
+ local = frame->local;
+ hashed_subvol = local->hashed_subvol;
+ cached_subvol = local->cached_subvol;
+
+ gf_uuid_unparse(local->loc.gfid, gfid);
+
+ if (local->file_count && local->dir_count) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_FILE_TYPE_MISMATCH,
+ "path %s (gfid = %s)exists as a file on one "
+ "subvolume and directory on another. "
+ "Please fix it manually",
+ local->loc.path, gfid);
+ DHT_STACK_UNWIND(lookup, frame, -1, EIO, NULL, NULL, NULL, NULL);
+ return 0;
+ }
+ if (local->op_ret && local->gfid_missing) {
+ if (gf_uuid_is_null(local->gfid_req)) {
+ DHT_STACK_UNWIND(lookup, frame, -1, ENODATA, NULL, NULL, NULL,
+ NULL);
+ return 0;
+ }
+ /* A hack */
+ dht_lookup_directory(frame, this, &local->loc);
+ return 0;
+ }
+
+ if (local->dir_count) {
+ dht_lookup_directory(frame, this, &local->loc);
+ return 0;
+ }
+
+ gf_msg_debug(this->name, 0,
+ "STATUS: hashed_subvol %s "
+ "cached_subvol %s",
+ (hashed_subvol == NULL) ? "null" : hashed_subvol->name,
+ (cached_subvol == NULL) ? "null" : cached_subvol->name);
+
+ if (!cached_subvol) {
+ if (local->skip_unlink.handle_valid_link && hashed_subvol) {
+ /*Purpose of "DHT_SKIP_NON_LINKTO_UNLINK":
+ * If this lookup is performed by rebalance and this
+ * rebalance process detected hashed file and by
+ * the time it sends the lookup request to cached node,
+ * file got migrated and now at initial hashed_node,
+ * final migrated file is present. With current logic,
+ * because this process fails to find the cached_node,
+ * it will unlink the file at initial hashed_node.
+ *
+ * So we avoid this by setting key, and checking at the
+ * posix_unlink that unlink the file only if file is a
+ * linkto file and not a migrated_file.
+ */
+
+ ret = dht_fill_dict_to_avoid_unlink_of_migrating_file(
+ local->xattr_req);
+
+ if (ret) {
+ /* If for some reason, setting key in the dict
+ * fails, return with ENOENT, as with respect to
+ * this process, it detected only a stale link
+ * file.
+ *
+ * Next lookup will delete it.
+ *
+ * Performing deletion of stale link file when
+ * setting key in dict fails, may cause the data
+ * loss because of the above mentioned race.
+ */
- local = frame->local;
- hashed_subvol = local->hashed_subvol;
- cached_subvol = local->cached_subvol;
+ DHT_STACK_UNWIND(lookup, frame, -1, ENOENT, NULL, NULL, NULL,
+ NULL);
+ } else {
+ local->skip_unlink.handle_valid_link = _gf_false;
+
+ gf_msg_debug(this->name, 0,
+ "No Cached was found and "
+ "unlink on hashed was skipped"
+ " so performing now: %s",
+ local->loc.path);
+ FRAME_SU_DO(frame, dht_local_t);
+ STACK_WIND(frame, dht_lookup_unlink_stale_linkto_cbk,
+ hashed_subvol, hashed_subvol->fops->unlink,
+ &local->loc, 0, local->xattr_req);
+ }
- gf_uuid_unparse (local->loc.gfid, gfid);
-
- if (local->file_count && local->dir_count) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_FILE_TYPE_MISMATCH,
- "path %s (gfid = %s)exists as a file on one "
- "subvolume and directory on another. "
- "Please fix it manually",
- local->loc.path, gfid);
- DHT_STACK_UNWIND (lookup, frame, -1, EIO, NULL, NULL, NULL,
- NULL);
- return 0;
- }
+ } else {
+ gf_msg_debug(this->name, 0,
+ "There was no cached file and "
+ "unlink on hashed is not skipped %s",
+ local->loc.path);
- if (local->dir_count) {
- dht_lookup_directory (frame, this, &local->loc);
- return 0;
+ DHT_STACK_UNWIND(lookup, frame, -1, ENOENT, NULL, NULL, NULL, NULL);
}
+ return 0;
+ }
- gf_msg_debug (this->name, 0, "STATUS: hashed_subvol %s "
- "cached_subvol %s",
- (hashed_subvol == NULL)?"null":hashed_subvol->name,
- (cached_subvol == NULL)?"null":cached_subvol->name);
-
- if (!cached_subvol) {
+ /* At the time of dht_lookup, no file was found on hashed and that is
+ * why dht_lookup_everywhere is called, but by the time
+ * dht_lookup_everywhere
+ * reached to server, file might have already migrated. In that case we
+ * will find a migrated file at the hashed_node. In this case store the
+ * layout in context and return successfully.
+ */
- if (local->skip_unlink.handle_valid_link && hashed_subvol) {
-
- /*Purpose of "DHT_SKIP_NON_LINKTO_UNLINK":
- * If this lookup is performed by rebalance and this
- * rebalance process detected hashed file and by
- * the time it sends the lookup request to cached node,
- * file got migrated and now at initial hashed_node,
- * final migrated file is present. With current logic,
- * because this process fails to find the cached_node,
- * it will unlink the file at initial hashed_node.
- *
- * So we avoid this by setting key, and checking at the
- * posix_unlink that unlink the file only if file is a
- * linkto file and not a migrated_file.
- */
-
-
- ret = dht_fill_dict_to_avoid_unlink_of_migrating_file
- (local->xattr_req);
-
- if (ret) {
- /* If for some reason, setting key in the dict
- * fails, return with ENOENT, as with respect to
- * this process, it detected only a stale link
- * file.
- *
- * Next lookup will delete it.
- *
- * Performing deletion of stale link file when
- * setting key in dict fails, may cause the data
- * loss becase of the above mentioned race.
- */
-
-
- DHT_STACK_UNWIND (lookup, frame, -1, ENOENT,
- NULL, NULL, NULL, NULL);
- } else {
- local->skip_unlink.handle_valid_link = _gf_false;
-
- gf_msg_debug (this->name, 0,
- "No Cached was found and "
- "unlink on hashed was skipped"
- " so performing now: %s",
- local->loc.path);
- FRAME_SU_DO (frame, dht_local_t);
- STACK_WIND (frame,
- dht_lookup_unlink_stale_linkto_cbk,
- hashed_subvol,
- hashed_subvol->fops->unlink,
- &local->loc, 0, local->xattr_req);
- }
+ if (hashed_subvol || local->need_lookup_everywhere) {
+ if (local->need_lookup_everywhere) {
+ found_non_linkto_on_hashed = _gf_true;
- } else {
+ } else if ((local->file_count == 1) &&
+ (hashed_subvol == cached_subvol)) {
+ gf_msg_debug(this->name, 0,
+ "found cached file on hashed subvolume "
+ "so store in context and return for %s",
+ local->loc.path);
- gf_msg_debug (this->name, 0,
- "There was no cached file and "
- "unlink on hashed is not skipped %s",
- local->loc.path);
-
- DHT_STACK_UNWIND (lookup, frame, -1, ENOENT, NULL, NULL,
- NULL, NULL);
- }
- return 0;
+ found_non_linkto_on_hashed = _gf_true;
}
- /* At the time of dht_lookup, no file was found on hashed and that is
- * why dht_lookup_everywhere is called, but by the time
- * dht_lookup_everywhere
- * reached to server, file might have already migrated. In that case we
- * will find a migrated file at the hashed_node. In this case store the
- * layout in context and return successfully.
- */
-
- if (hashed_subvol || local->need_lookup_everywhere) {
-
- if (local->need_lookup_everywhere) {
-
- found_non_linkto_on_hashed = _gf_true;
+ if (found_non_linkto_on_hashed)
+ goto preset_layout;
+ }
- } else if ((local->file_count == 1) &&
- (hashed_subvol == cached_subvol)) {
+ if (hashed_subvol) {
+ if (local->skip_unlink.handle_valid_link == _gf_true) {
+ if (cached_subvol == local->skip_unlink.hash_links_to) {
+ if (gf_uuid_compare(local->skip_unlink.cached_gfid,
+ local->skip_unlink.hashed_gfid)) {
+ /*GFID different, return error*/
+ DHT_STACK_UNWIND(lookup, frame, -1, ESTALE, NULL, NULL,
+ NULL, NULL);
- gf_msg_debug (this->name, 0,
- "found cached file on hashed subvolume "
- "so store in context and return for %s",
- local->loc.path);
-
- found_non_linkto_on_hashed = _gf_true;
+ return 0;
}
- if (found_non_linkto_on_hashed)
- goto preset_layout;
-
- }
-
-
- if (hashed_subvol) {
- if (local->skip_unlink.handle_valid_link == _gf_true) {
- if (cached_subvol == local->skip_unlink.hash_links_to) {
-
- if (gf_uuid_compare (local->skip_unlink.cached_gfid,
- local->skip_unlink.hashed_gfid)){
-
- /*GFID different, return error*/
- DHT_STACK_UNWIND (lookup, frame, -1,
- ESTALE, NULL, NULL,
- NULL, NULL);
-
- return 0;
- }
-
- ret = dht_layout_preset (this, cached_subvol,
- local->loc.inode);
- if (ret) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LAYOUT_PRESET_FAILED,
- "Could not set pre-set layout "
- "for subvolume %s",
- cached_subvol->name);
- }
-
- local->op_ret = (ret == 0) ? ret : -1;
- local->op_errno = (ret == 0) ? ret : EINVAL;
-
- /* Presence of local->cached_subvol validates
- * that lookup from cached node is successful
- */
-
- if (!local->op_ret && local->loc.parent) {
- dht_inode_ctx_time_update
- (local->loc.parent, this,
- &local->postparent, 1);
- }
-
- gf_msg_debug (this->name, 0,
- "Skipped unlinking linkto file "
- "on the hashed subvolume. "
- "Returning success as it is a "
- "valid linkto file. Path:%s"
- ,local->loc.path);
-
- goto unwind_hashed_and_cached;
- } else {
-
- local->skip_unlink.handle_valid_link = _gf_false;
-
- gf_msg_debug (this->name, 0,
- "Linkto file found on hashed "
- "subvol "
- "and data file found on cached "
- "subvolume. But linkto points to "
- "different cached subvolume (%s) "
- "path %s",
- (local->skip_unlink.hash_links_to ?
- local->skip_unlink.hash_links_to->name :
- " <nil>"), local->loc.path);
-
- if (local->skip_unlink.opend_fd_count == 0) {
-
-
- ret = dht_fill_dict_to_avoid_unlink_of_migrating_file
- (local->xattr_req);
-
-
- if (ret) {
- DHT_STACK_UNWIND (lookup, frame, -1,
- EIO, NULL, NULL,
- NULL, NULL);
- } else {
- local->call_cnt = 1;
- FRAME_SU_DO (frame, dht_local_t);
- STACK_WIND (frame,
- dht_lookup_unlink_of_false_linkto_cbk,
- hashed_subvol,
- hashed_subvol->fops->unlink,
- &local->loc, 0,
- local->xattr_req);
- }
+ ret = dht_layout_preset(this, cached_subvol, local->loc.inode);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_INFO, 0,
+ DHT_MSG_LAYOUT_PRESET_FAILED,
+ "Could not set pre-set layout "
+ "for subvolume %s",
+ cached_subvol->name);
+ }
- return 0;
+ local->op_ret = (ret == 0) ? ret : -1;
+ local->op_errno = (ret == 0) ? ret : EINVAL;
- }
- }
+ /* Presence of local->cached_subvol validates
+ * that lookup from cached node is successful
+ */
+ if (!local->op_ret && local->loc.parent) {
+ dht_inode_ctx_time_update(local->loc.parent, this,
+ &local->postparent, 1);
+ }
+
+ gf_msg_debug(this->name, 0,
+ "Skipped unlinking linkto file "
+ "on the hashed subvolume. "
+ "Returning success as it is a "
+ "valid linkto file. Path:%s",
+ local->loc.path);
+
+ goto unwind_hashed_and_cached;
+ } else {
+ local->skip_unlink.handle_valid_link = _gf_false;
+
+ gf_msg_debug(this->name, 0,
+ "Linkto file found on hashed "
+ "subvol "
+ "and data file found on cached "
+ "subvolume. But linkto points to "
+ "different cached subvolume (%s) "
+ "path %s",
+ (local->skip_unlink.hash_links_to
+ ? local->skip_unlink.hash_links_to->name
+ : " <nil>"),
+ local->loc.path);
+
+ if (local->skip_unlink.opend_fd_count == 0) {
+ ret = dht_fill_dict_to_avoid_unlink_of_migrating_file(
+ local->xattr_req);
+
+ if (ret) {
+ DHT_STACK_UNWIND(lookup, frame, -1, EIO, NULL, NULL,
+ NULL, NULL);
+ } else {
+ local->call_cnt = 1;
+ FRAME_SU_DO(frame, dht_local_t);
+ STACK_WIND(frame, dht_lookup_unlink_of_false_linkto_cbk,
+ hashed_subvol, hashed_subvol->fops->unlink,
+ &local->loc, 0, local->xattr_req);
+ }
+
+ return 0;
}
+ }
}
-
+ }
preset_layout:
- if (found_non_linkto_on_hashed) {
-
- if (local->need_lookup_everywhere) {
- if (gf_uuid_compare (local->gfid, local->inode->gfid)) {
- /* GFID different, return error */
- DHT_STACK_UNWIND (lookup, frame, -1, ENOENT,
- NULL, NULL, NULL, NULL);
- return 0;
- }
- }
-
- local->op_ret = 0;
- local->op_errno = 0;
- layout = dht_layout_for_subvol (this, cached_subvol);
- if (!layout) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_SUBVOL_INFO,
- "%s: no pre-set layout for subvolume %s,"
- " gfid = %s",
- local->loc.path, (cached_subvol ?
- cached_subvol->name :
- "<nil>"), gfid);
- }
-
- ret = dht_layout_set (this, local->inode, layout);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_SUBVOL_INFO,
- "%s: failed to set layout for subvol %s, "
- "gfid = %s",
- local->loc.path, (cached_subvol ?
- cached_subvol->name :
- "<nil>"), gfid);
- }
-
- if (local->loc.parent) {
- dht_inode_ctx_time_update (local->loc.parent, this,
- &local->postparent, 1);
- }
-
- DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
- dht_set_fixed_dir_stat (&local->postparent);
- DHT_STACK_UNWIND (lookup, frame, local->op_ret,
- local->op_errno, local->inode,
- &local->stbuf, local->xattr,
- &local->postparent);
+ if (found_non_linkto_on_hashed) {
+ if (local->need_lookup_everywhere) {
+ if (gf_uuid_compare(local->gfid, local->inode->gfid)) {
+ /* GFID different, return error */
+ DHT_STACK_UNWIND(lookup, frame, -1, ENOENT, NULL, NULL, NULL,
+ NULL);
return 0;
+ }
}
- if (!hashed_subvol) {
+ local->op_ret = 0;
+ local->op_errno = 0;
+ layout = dht_layout_for_subvol(this, cached_subvol);
+ if (!layout) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_SUBVOL_INFO,
+ "%s: no pre-set layout for subvolume %s,"
+ " gfid = %s",
+ local->loc.path,
+ (cached_subvol ? cached_subvol->name : "<nil>"), gfid);
+ }
- gf_msg_debug (this->name, 0,
- "Cannot create linkfile for %s on %s: "
- "hashed subvolume cannot be found, gfid = %s.",
- local->loc.path, cached_subvol->name, gfid);
+ ret = dht_layout_set(this, local->inode, layout);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_SUBVOL_INFO,
+ "%s: failed to set layout for subvol %s, "
+ "gfid = %s",
+ local->loc.path,
+ (cached_subvol ? cached_subvol->name : "<nil>"), gfid);
+ }
- local->op_ret = 0;
- local->op_errno = 0;
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update(local->loc.parent, this,
+ &local->postparent, 1);
+ }
- ret = dht_layout_preset (frame->this, cached_subvol,
- local->inode);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LAYOUT_PRESET_FAILED,
- "Failed to set layout for subvol %s"
- ", gfid = %s",
- cached_subvol ? cached_subvol->name :
- "<nil>", gfid);
- local->op_ret = -1;
- local->op_errno = EINVAL;
- }
+ DHT_STRIP_PHASE1_FLAGS(&local->stbuf);
+ dht_set_fixed_dir_stat(&local->postparent);
+ DHT_STACK_UNWIND(lookup, frame, local->op_ret, local->op_errno,
+ local->inode, &local->stbuf, local->xattr,
+ &local->postparent);
+ return 0;
+ }
- if (local->loc.parent) {
- dht_inode_ctx_time_update (local->loc.parent, this,
- &local->postparent, 1);
- }
+ if (!hashed_subvol) {
+ gf_msg_debug(this->name, 0,
+ "Cannot create linkfile for %s on %s: "
+ "hashed subvolume cannot be found, gfid = %s.",
+ local->loc.path, cached_subvol->name, gfid);
- DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
- dht_set_fixed_dir_stat (&local->postparent);
- DHT_STACK_UNWIND (lookup, frame, local->op_ret,
- local->op_errno, local->inode,
- &local->stbuf, local->xattr,
- &local->postparent);
- return 0;
+ local->op_ret = 0;
+ local->op_errno = 0;
+
+ ret = dht_layout_preset(frame->this, cached_subvol, local->inode);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_LAYOUT_PRESET_FAILED,
+ "Failed to set layout for subvol %s"
+ ", gfid = %s",
+ cached_subvol ? cached_subvol->name : "<nil>", gfid);
+ local->op_ret = -1;
+ local->op_errno = EINVAL;
}
- gf_msg_debug (this->name, 0,
- "Creating linkto file on %s(hash) to %s on %s (gfid = %s)",
- hashed_subvol->name, local->loc.path,
- cached_subvol->name, gfid);
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update(local->loc.parent, this,
+ &local->postparent, 1);
+ }
- ret = dht_linkfile_create (frame,
- dht_lookup_linkfile_create_cbk, this,
- cached_subvol, hashed_subvol, &local->loc);
+ DHT_STRIP_PHASE1_FLAGS(&local->stbuf);
+ dht_set_fixed_dir_stat(&local->postparent);
+ DHT_STACK_UNWIND(lookup, frame, local->op_ret, local->op_errno,
+ local->inode, &local->stbuf, local->xattr,
+ &local->postparent);
+ return 0;
+ }
- return ret;
+ if (frame->root->op != GF_FOP_RENAME) {
+ local->current = &local->lock[0];
+ ret = dht_protect_namespace(frame, &local->loc, hashed_subvol,
+ &local->current->ns,
+ dht_call_lookup_linkfile_create);
+ } else {
+ gf_msg_debug(this->name, 0,
+ "Creating linkto file on %s(hash) to %s on %s "
+ "(gfid = %s)",
+ hashed_subvol->name, local->loc.path, cached_subvol->name,
+ gfid);
+
+ ret = dht_linkfile_create(frame, dht_lookup_linkfile_create_cbk, this,
+ cached_subvol, hashed_subvol, &local->loc);
+ }
+
+ return ret;
unwind_hashed_and_cached:
- DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
- dht_set_fixed_dir_stat (&local->postparent);
- DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
- local->inode, &local->stbuf, local->xattr,
- &local->postparent);
- return 0;
+ DHT_STRIP_PHASE1_FLAGS(&local->stbuf);
+ dht_set_fixed_dir_stat(&local->postparent);
+ DHT_STACK_UNWIND(lookup, frame, local->op_ret, local->op_errno,
+ local->inode, &local->stbuf, local->xattr,
+ &local->postparent);
+ return 0;
}
-int
-dht_lookup_everywhere_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf, dict_t *xattr,
- struct iatt *postparent)
-{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- xlator_t *prev = NULL;
- int is_linkfile = 0;
- int is_dir = 0;
- xlator_t *subvol = NULL;
- loc_t *loc = NULL;
- xlator_t *link_subvol = NULL;
- int ret = -1;
- int32_t fd_count = 0;
- dht_conf_t *conf = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
- dict_t *dict_req = {0};
-
- GF_VALIDATE_OR_GOTO ("dht", frame, out);
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
- GF_VALIDATE_OR_GOTO ("dht", cookie, out);
- GF_VALIDATE_OR_GOTO ("dht", this->private, out);
-
- local = frame->local;
- loc = &local->loc;
- conf = this->private;
-
- prev = cookie;
- subvol = prev;
-
- gf_msg_debug (this->name, 0,
- "returned with op_ret %d and op_errno %d (%s) "
- "from subvol %s", op_ret, op_errno, loc->path,
- subvol->name);
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- if (op_errno != ENOENT)
- local->op_errno = op_errno;
- goto unlock;
- }
+static int
+dht_lookup_everywhere_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xattr,
+ struct iatt *postparent)
+{
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ xlator_t *prev = NULL;
+ int is_linkfile = 0;
+ int is_dir = 0;
+ loc_t *loc = NULL;
+ xlator_t *link_subvol = NULL;
+ int ret = -1;
+ int32_t fd_count = 0;
+ dht_conf_t *conf = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ dict_t *dict_req = {0};
+
+ GF_VALIDATE_OR_GOTO("dht", frame, out);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO("dht", cookie, out);
+ GF_VALIDATE_OR_GOTO("dht", this->private, out);
+
+ local = frame->local;
+ loc = &local->loc;
+ conf = this->private;
+
+ prev = cookie;
+
+ gf_msg_debug(this->name, 0,
+ "returned with op_ret %d and op_errno %d (%s) "
+ "from subvol %s",
+ op_ret, op_errno, loc->path, prev->name);
+
+ LOCK(&frame->lock);
+ {
+ if (op_ret == -1) {
+ if (op_errno != ENOENT)
+ local->op_errno = op_errno;
+ if (op_errno == ENODATA)
+ local->gfid_missing = _gf_true;
+ goto unlock;
+ }
- if (gf_uuid_is_null (local->gfid))
- gf_uuid_copy (local->gfid, buf->ia_gfid);
+ if (gf_uuid_is_null(local->gfid))
+ gf_uuid_copy(local->gfid, buf->ia_gfid);
- gf_uuid_unparse(local->gfid, gfid);
+ gf_uuid_unparse(local->gfid, gfid);
- if (gf_uuid_compare (local->gfid, buf->ia_gfid)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_GFID_MISMATCH,
- "%s: gfid differs on subvolume %s,"
- " gfid local = %s, gfid node = %s",
- loc->path, prev->name, gfid,
- uuid_utoa(buf->ia_gfid));
- }
+ if (gf_uuid_compare(local->gfid, buf->ia_gfid)) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_GFID_MISMATCH,
+ "%s: gfid differs on subvolume %s,"
+ " gfid local = %s, gfid node = %s",
+ loc->path, prev->name, gfid, uuid_utoa(buf->ia_gfid));
+ }
- is_linkfile = check_is_linkfile (inode, buf, xattr,
- conf->link_xattr_name);
-
- if (is_linkfile) {
- link_subvol = dht_linkfile_subvol (this, inode, buf,
- xattr);
- gf_msg_debug (this->name, 0,
- "found on %s linkfile %s (-> %s)",
- subvol->name, loc->path,
- link_subvol ? link_subvol->name : "''");
- goto unlock;
- }
+ is_linkfile = check_is_linkfile(inode, buf, xattr,
+ conf->link_xattr_name);
- is_dir = check_is_dir (inode, buf, xattr);
+ if (is_linkfile) {
+ link_subvol = dht_linkfile_subvol(this, inode, buf, xattr);
+ gf_msg_debug(this->name, 0, "found on %s linkfile %s (-> %s)",
+ prev->name, loc->path,
+ link_subvol ? link_subvol->name : "''");
+ goto unlock;
+ }
- /* non linkfile GFID takes precedence but don't overwrite
- gfid if we have already found a cached file*/
- if (!local->cached_subvol)
- gf_uuid_copy (local->gfid, buf->ia_gfid);
+ is_dir = check_is_dir(inode, buf, xattr);
- if (is_dir) {
- local->dir_count++;
+ /* non linkfile GFID takes precedence but don't overwrite
+ gfid if we have already found a cached file*/
+ if (!local->cached_subvol)
+ gf_uuid_copy(local->gfid, buf->ia_gfid);
- gf_msg_debug (this->name, 0,
- "found on %s directory %s",
- subvol->name, loc->path);
- } else {
- local->file_count++;
-
- gf_msg_debug (this->name, 0,
- "found cached file on %s for %s",
- subvol->name, loc->path);
-
- if (!local->cached_subvol) {
- /* found one file */
- dht_iatt_merge (this, &local->stbuf, buf,
- subvol);
- local->xattr = dict_ref (xattr);
- local->cached_subvol = subvol;
-
- gf_msg_debug (this->name, 0,
- "storing cached on %s file"
- " %s", subvol->name, loc->path);
-
- dht_iatt_merge (this, &local->postparent,
- postparent, subvol);
-
- gf_uuid_copy (local->skip_unlink.cached_gfid,
- buf->ia_gfid);
- } else {
- /* This is where we need 'rename' both entries logic */
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_FILE_ON_MULT_SUBVOL,
- "multiple subvolumes (%s and %s) have "
- "file %s (preferably rename the file "
- "in the backend,and do a fresh lookup)",
- local->cached_subvol->name,
- subvol->name, local->loc.path);
- }
- }
+ if (is_dir) {
+ local->dir_count++;
+
+ gf_msg_debug(this->name, 0, "found on %s directory %s", prev->name,
+ loc->path);
+ } else {
+ local->file_count++;
+
+ gf_msg_debug(this->name, 0, "found cached file on %s for %s",
+ prev->name, loc->path);
+
+ if (!local->cached_subvol) {
+ /* found one file */
+ dht_iatt_merge(this, &local->stbuf, buf);
+
+ local->xattr = dict_ref(xattr);
+ local->cached_subvol = prev;
+
+ gf_msg_debug(this->name, 0,
+ "storing cached on %s file"
+ " %s",
+ prev->name, loc->path);
+
+ dht_iatt_merge(this, &local->postparent, postparent);
+
+ gf_uuid_copy(local->skip_unlink.cached_gfid, buf->ia_gfid);
+ } else {
+ /* This is where we need 'rename' both entries logic */
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ DHT_MSG_FILE_ON_MULT_SUBVOL,
+ "multiple subvolumes (%s and %s) have "
+ "file %s (preferably rename the file "
+ "in the backend,and do a fresh lookup)",
+ local->cached_subvol->name, prev->name, local->loc.path);
+ }
}
+ }
unlock:
- UNLOCK (&frame->lock);
+ UNLOCK(&frame->lock);
- if (is_linkfile) {
- ret = dict_get_int32 (xattr, GLUSTERFS_OPEN_FD_COUNT, &fd_count);
+ if (is_linkfile) {
+ ret = dict_get_int32(xattr, GLUSTERFS_OPEN_FD_COUNT, &fd_count);
- /* Any linkto file found on the non-hashed subvolume should
- * be unlinked (performed in the "else if" block below)
- *
- * But if a linkto file is found on hashed subvolume, it may be
- * pointing to valid cached node. So unlinking of linkto
- * file on hashed subvolume is skipped and inside
- * dht_lookup_everywhere_done, checks are performed. If this
- * linkto file is found as stale linkto file, it is deleted
- * otherwise unlink is skipped.
+ /* Any linkto file found on the non-hashed subvolume should
+ * be unlinked (performed in the "else if" block below)
+ *
+ * But if a linkto file is found on hashed subvolume, it may be
+ * pointing to valid cached node. So unlinking of linkto
+ * file on hashed subvolume is skipped and inside
+ * dht_lookup_everywhere_done, checks are performed. If this
+ * linkto file is found as stale linkto file, it is deleted
+ * otherwise unlink is skipped.
+ */
+
+ if (local->hashed_subvol && local->hashed_subvol == prev) {
+ local->skip_unlink.handle_valid_link = _gf_true;
+ local->skip_unlink.opend_fd_count = fd_count;
+ local->skip_unlink.hash_links_to = link_subvol;
+ gf_uuid_copy(local->skip_unlink.hashed_gfid, buf->ia_gfid);
+
+ gf_msg_debug(this->name, 0,
+ "Found"
+ " one linkto file on hashed subvol %s "
+ "for %s: Skipping unlinking till "
+ "everywhere_done",
+ prev->name, loc->path);
+
+ } else if (!ret && (fd_count == 0)) {
+ dict_req = dict_new();
+
+ ret = dht_fill_dict_to_avoid_unlink_of_migrating_file(dict_req);
+
+ if (ret) {
+ /* Skip unlinking for dict_failure
+ *File is found as a linkto file on non-hashed,
+ *subvolume. In the current implementation,
+ *finding a linkto-file on non-hashed does not
+ *always implies that it is stale. So deletion
+ *of file should be done only when both fd is
+ *closed and linkto-xattr is set. In case of
+ *dict_set failure, avoid skipping of file.
+ *NOTE: dht_frame_return should get called for
+ * this block.
*/
- if (local->hashed_subvol && local->hashed_subvol == subvol) {
-
- local->skip_unlink.handle_valid_link = _gf_true;
- local->skip_unlink.opend_fd_count = fd_count;
- local->skip_unlink.hash_links_to = link_subvol;
- gf_uuid_copy (local->skip_unlink.hashed_gfid,
- buf->ia_gfid);
-
- gf_msg_debug (this->name, 0, "Found"
- " one linkto file on hashed subvol %s "
- "for %s: Skipping unlinking till "
- "everywhere_done", subvol->name,
- loc->path);
-
- } else if (!ret && (fd_count == 0)) {
-
- dict_req = dict_new ();
-
- ret = dht_fill_dict_to_avoid_unlink_of_migrating_file
- (dict_req);
-
- if (ret) {
-
- /* Skip unlinking for dict_failure
- *File is found as a linkto file on non-hashed,
- *subvolume. In the current implementation,
- *finding a linkto-file on non-hashed does not
- *always implies that it is stale. So deletion
- *of file should be done only when both fd is
- *closed and linkto-xattr is set. In case of
- *dict_set failure, avoid skipping of file.
- *NOTE: dht_frame_return should get called for
- * this block.
- */
-
- dict_unref (dict_req);
-
- } else {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_SUBVOL_INFO,
- "attempting deletion of stale linkfile "
- "%s on %s (hashed subvol is %s)",
- loc->path, subvol->name,
- (local->hashed_subvol?
- local->hashed_subvol->name : "<null>"));
- /* *
- * These stale files may be created using root
- * user. Hence deletion will work only with
- * root.
- */
- FRAME_SU_DO (frame, dht_local_t);
- STACK_WIND (frame, dht_lookup_unlink_cbk,
- subvol, subvol->fops->unlink, loc,
- 0, dict_req);
-
- dict_unref (dict_req);
-
- return 0;
- }
- }
- }
+ dict_unref(dict_req);
+
+ } else {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_SUBVOL_INFO,
+ "attempting deletion of stale linkfile "
+ "%s on %s (hashed subvol is %s)",
+ loc->path, prev->name,
+ (local->hashed_subvol ? local->hashed_subvol->name
+ : "<null>"));
+ /* *
+ * These stale files may be created using root
+ * user. Hence deletion will work only with
+ * root.
+ */
+ FRAME_SU_DO(frame, dht_local_t);
+ STACK_WIND(frame, dht_lookup_unlink_cbk, prev,
+ prev->fops->unlink, loc, 0, dict_req);
+
+ dict_unref(dict_req);
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- dht_lookup_everywhere_done (frame, this);
+ return 0;
+ }
}
+ }
+
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt)) {
+ dht_lookup_everywhere_done(frame, this);
+ }
out:
- return ret;
+ return ret;
}
-
int
-dht_lookup_everywhere (call_frame_t *frame, xlator_t *this, loc_t *loc)
+dht_lookup_everywhere(call_frame_t *frame, xlator_t *this, loc_t *loc)
{
- dht_conf_t *conf = NULL;
- dht_local_t *local = NULL;
- int i = 0;
- int call_cnt = 0;
+ dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ int i = 0;
+ int call_cnt = 0;
- GF_VALIDATE_OR_GOTO ("dht", frame, err);
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
- GF_VALIDATE_OR_GOTO ("dht", this->private, out);
- GF_VALIDATE_OR_GOTO ("dht", loc, out);
+ GF_VALIDATE_OR_GOTO("dht", frame, err);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO("dht", this->private, out);
+ GF_VALIDATE_OR_GOTO("dht", loc, out);
- conf = this->private;
- local = frame->local;
+ conf = this->private;
+ local = frame->local;
- call_cnt = conf->subvolume_cnt;
- local->call_cnt = call_cnt;
+ call_cnt = conf->subvolume_cnt;
+ local->call_cnt = call_cnt;
- if (!local->inode)
- local->inode = inode_ref (loc->inode);
+ if (!local->inode)
+ local->inode = inode_ref(loc->inode);
- gf_msg_debug (this->name, 0,
- "winding lookup call to %d subvols", call_cnt);
+ gf_msg_debug(this->name, 0, "winding lookup call to %d subvols", call_cnt);
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND_COOKIE (frame, dht_lookup_everywhere_cbk,
- conf->subvolumes[i], conf->subvolumes[i],
- conf->subvolumes[i]->fops->lookup,
- loc, local->xattr_req);
- }
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND_COOKIE(frame, dht_lookup_everywhere_cbk, conf->subvolumes[i],
+ conf->subvolumes[i],
+ conf->subvolumes[i]->fops->lookup, loc,
+ local->xattr_req);
+ }
- return 0;
+ return 0;
out:
- DHT_STACK_UNWIND (lookup, frame, -1, EINVAL, NULL, NULL, NULL, NULL);
+ DHT_STACK_UNWIND(lookup, frame, -1, EINVAL, NULL, NULL, NULL, NULL);
err:
- return -1;
+ return -1;
}
-
int
-dht_lookup_linkfile_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, dict_t *xattr,
- struct iatt *postparent)
-{
- xlator_t *prev = NULL;
- dht_local_t *local = NULL;
- xlator_t *subvol = NULL;
- loc_t *loc = NULL;
- dht_conf_t *conf = NULL;
- int ret = 0;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- GF_VALIDATE_OR_GOTO ("dht", frame, out);
- GF_VALIDATE_OR_GOTO ("dht", this, unwind);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, unwind);
- GF_VALIDATE_OR_GOTO ("dht", this->private, unwind);
- GF_VALIDATE_OR_GOTO ("dht", cookie, unwind);
-
- prev = cookie;
- subvol = prev;
- conf = this->private;
- local = frame->local;
- loc = &local->loc;
-
- gf_uuid_unparse(loc->gfid, gfid);
-
- if (op_ret == -1) {
- gf_msg (this->name, GF_LOG_INFO, op_errno,
- DHT_MSG_LINK_FILE_LOOKUP_INFO,
- "Lookup of %s on %s (following linkfile) failed "
- ",gfid = %s", local->loc.path, subvol->name, gfid);
-
- /* If cached subvol returned ENOTCONN, do not do
- lookup_everywhere. We need to make sure linkfile does not get
- removed, which can take away the namespace, and subvol is
- anyways down. */
-
- if (op_errno != ENOTCONN)
- goto err;
- else
- goto unwind;
- }
-
- if (check_is_dir (inode, stbuf, xattr)) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LINK_FILE_LOOKUP_INFO,
- "Lookup of %s on %s (following linkfile) reached dir,"
- " gfid = %s", local->loc.path, subvol->name, gfid);
- goto err;
- }
-
- if (check_is_linkfile (inode, stbuf, xattr, conf->link_xattr_name)) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LINK_FILE_LOOKUP_INFO,
- "lookup of %s on %s (following linkfile) reached link,"
- "gfid = %s", local->loc.path, subvol->name, gfid);
- goto err;
- }
-
- if (gf_uuid_compare (local->gfid, stbuf->ia_gfid)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_GFID_MISMATCH,
- "%s: gfid different on data file on %s,"
- " gfid local = %s, gfid node = %s ",
- local->loc.path, subvol->name, gfid,
- uuid_utoa(stbuf->ia_gfid));
- goto err;
- }
-
- if ((stbuf->ia_nlink == 1)
- && (conf && conf->unhashed_sticky_bit)) {
- stbuf->ia_prot.sticky = 1;
- }
-
- ret = dht_layout_preset (this, prev, inode);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LAYOUT_PRESET_FAILED,
- "Failed to set layout for subvolume %s,"
- "gfid = %s", prev->name, gfid);
- op_ret = -1;
- op_errno = EINVAL;
- }
-
- if (local->loc.parent) {
- dht_inode_ctx_time_update (local->loc.parent, this,
- postparent, 1);
- }
+dht_lookup_linkfile_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *stbuf, dict_t *xattr,
+ struct iatt *postparent)
+{
+ xlator_t *prev = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+ loc_t *loc = NULL;
+ dht_conf_t *conf = NULL;
+ int ret = 0;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+
+ GF_VALIDATE_OR_GOTO("dht", frame, out);
+ GF_VALIDATE_OR_GOTO("dht", this, unwind);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, unwind);
+ GF_VALIDATE_OR_GOTO("dht", this->private, unwind);
+ GF_VALIDATE_OR_GOTO("dht", cookie, unwind);
+
+ prev = cookie;
+ subvol = prev;
+ conf = this->private;
+ local = frame->local;
+ loc = &local->loc;
+
+ gf_uuid_unparse(loc->gfid, gfid);
+
+ if (op_ret == -1) {
+ gf_msg(this->name, GF_LOG_INFO, op_errno, DHT_MSG_LINK_FILE_LOOKUP_INFO,
+ "Lookup of %s on %s (following linkfile) failed "
+ ",gfid = %s",
+ local->loc.path, subvol->name, gfid);
+
+ /* If cached subvol returned ENOTCONN, do not do
+ lookup_everywhere. We need to make sure linkfile does not get
+ removed, which can take away the namespace, and subvol is
+ anyways down. */
+
+ local->cached_subvol = NULL;
+ if (op_errno != ENOTCONN)
+ goto err;
+ else
+ goto unwind;
+ }
+
+ if (check_is_dir(inode, stbuf, xattr)) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_LINK_FILE_LOOKUP_INFO,
+ "Lookup of %s on %s (following linkfile) reached dir,"
+ " gfid = %s",
+ local->loc.path, subvol->name, gfid);
+ goto err;
+ }
+
+ if (check_is_linkfile(inode, stbuf, xattr, conf->link_xattr_name)) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_LINK_FILE_LOOKUP_INFO,
+ "lookup of %s on %s (following linkfile) reached link,"
+ "gfid = %s",
+ local->loc.path, subvol->name, gfid);
+ goto err;
+ }
+
+ if (gf_uuid_compare(local->gfid, stbuf->ia_gfid)) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_GFID_MISMATCH,
+ "%s: gfid different on data file on %s,"
+ " gfid local = %s, gfid node = %s ",
+ local->loc.path, subvol->name, gfid, uuid_utoa(stbuf->ia_gfid));
+ goto err;
+ }
+
+ if ((stbuf->ia_nlink == 1) && (conf && conf->unhashed_sticky_bit)) {
+ stbuf->ia_prot.sticky = 1;
+ }
+
+ ret = dht_layout_preset(this, prev, inode);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_LAYOUT_PRESET_FAILED,
+ "Failed to set layout for subvolume %s,"
+ "gfid = %s",
+ prev->name, gfid);
+ op_ret = -1;
+ op_errno = EINVAL;
+ }
+
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update(local->loc.parent, this, postparent, 1);
+ }
unwind:
- DHT_STRIP_PHASE1_FLAGS (stbuf);
- dht_set_fixed_dir_stat (postparent);
- DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, stbuf, xattr,
- postparent);
+ DHT_STRIP_PHASE1_FLAGS(stbuf);
+ dht_set_fixed_dir_stat(postparent);
+ DHT_STACK_UNWIND(lookup, frame, op_ret, op_errno, inode, stbuf, xattr,
+ postparent);
- return 0;
+ return 0;
err:
- dht_lookup_everywhere (frame, this, loc);
+ dht_lookup_everywhere(frame, this, loc);
out:
- return 0;
-}
-
-
-int
-dht_lookup_directory (call_frame_t *frame, xlator_t *this, loc_t *loc)
-{
- int call_cnt = 0;
- int i = 0;
- dht_conf_t *conf = NULL;
- dht_local_t *local = NULL;
- int ret = 0;
-
- GF_VALIDATE_OR_GOTO ("dht", frame, out);
- GF_VALIDATE_OR_GOTO ("dht", this, unwind);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, unwind);
- GF_VALIDATE_OR_GOTO ("dht", this->private, unwind);
- GF_VALIDATE_OR_GOTO ("dht", loc, unwind);
-
- conf = this->private;
- local = frame->local;
-
- call_cnt = conf->subvolume_cnt;
- local->call_cnt = call_cnt;
-
- local->layout = dht_layout_new (this, conf->subvolume_cnt);
- if (!local->layout) {
- goto unwind;
- }
-
- if (local->xattr != NULL) {
- dict_unref (local->xattr);
- local->xattr = NULL;
- }
-
- if (!gf_uuid_is_null (local->gfid)) {
- ret = dict_set_gfuuid (local->xattr_req, "gfid-req",
- local->gfid, true);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "%s: Failed to set dictionary value:"
- " key = gfid-req", local->loc.path);
- }
-
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND_COOKIE (frame, dht_lookup_dir_cbk,
- conf->subvolumes[i], conf->subvolumes[i],
- conf->subvolumes[i]->fops->lookup,
- &local->loc, local->xattr_req);
- }
- return 0;
-unwind:
- DHT_STACK_UNWIND (lookup, frame, -1, ENOMEM, NULL, NULL, NULL, NULL);
-out:
- return 0;
-
+ return 0;
}
/* Code to get hashed subvol based on inode and loc
@@ -2928,7382 +2990,7612 @@ out:
hashed subvol based on loc.
*/
-xlator_t *
-dht_inode_get_hashed_subvol (inode_t *inode, xlator_t *this, loc_t *loc)
-{
- char *path = NULL;
- loc_t populate_loc = {0, };
- char *name = NULL;
- xlator_t *hash_subvol = NULL;
-
- if (!inode)
- return hash_subvol;
-
- if (loc && loc->parent && loc->path) {
- if (!loc->name) {
- name = strrchr (loc->path, '/');
- if (name) {
- loc->name = name + 1;
- } else {
- goto out;
- }
- }
- hash_subvol = dht_subvol_get_hashed (this, loc);
- goto out;
+static gf_boolean_t
+dht_should_lookup_everywhere(xlator_t *this, dht_conf_t *conf, loc_t *loc)
+{
+ dht_layout_t *parent_layout = NULL;
+ int ret = 0;
+ gf_boolean_t lookup_everywhere = _gf_true;
+
+ /* lookup-optimize supersedes lookup-unhashed settings.
+ * If it is set, do not process search_unhashed
+ * If lookup-optimize if enabled, lookup everywhere if:
+ * - this is the rebalance daemon.
+ * - loc->parent is unavailable.
+ * - parent_layout is unavailable
+ * - parent_layout->commit_hash != conf->vol_commit_hash
+ */
+
+ if (conf->lookup_optimize) {
+ if (!conf->defrag && loc->parent) {
+ ret = dht_inode_ctx_layout_get(loc->parent, this, &parent_layout);
+ if (!ret && parent_layout &&
+ (parent_layout->commit_hash == conf->vol_commit_hash)) {
+ lookup_everywhere = _gf_false;
+ }
}
+ goto out;
+ } else {
+ if (conf->search_unhashed == GF_DHT_LOOKUP_UNHASHED_AUTO) {
+ if (loc->parent) {
+ ret = dht_inode_ctx_layout_get(loc->parent, this,
+ &parent_layout);
+ if (ret || !parent_layout ||
+ (!parent_layout->search_unhashed)) {
+ lookup_everywhere = _gf_false;
+ }
+ } else {
+ lookup_everywhere = _gf_false;
+ }
- if (!gf_uuid_is_null (inode->gfid)) {
- populate_loc.inode = inode_ref (inode);
- populate_loc.parent = inode_parent (populate_loc.inode,
- NULL, NULL);
- inode_path (populate_loc.inode, NULL, &path);
-
- if (!path)
- goto out;
-
- populate_loc.path = path;
- if (!populate_loc.name && populate_loc.path) {
- name = strrchr (populate_loc.path, '/');
- if (name) {
- populate_loc.name = name + 1;
-
- } else {
- goto out;
- }
- }
- hash_subvol = dht_subvol_get_hashed (this, &populate_loc);
+ goto out;
}
+ }
out:
- if (populate_loc.inode)
- loc_wipe (&populate_loc);
- return hash_subvol;
+ return lookup_everywhere;
}
-
int
-dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, dict_t *xattr,
- struct iatt *postparent)
-{
- char is_linkfile = 0;
- char is_dir = 0;
- xlator_t *subvol = NULL;
- dht_conf_t *conf = NULL;
- dht_local_t *local = NULL;
- loc_t *loc = NULL;
- xlator_t *prev = NULL;
- int ret = 0;
- dht_layout_t *parent_layout = NULL;
- uint32_t vol_commit_hash = 0;
-
- GF_VALIDATE_OR_GOTO ("dht", frame, err);
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
- GF_VALIDATE_OR_GOTO ("dht", cookie, out);
- GF_VALIDATE_OR_GOTO ("dht", this->private, out);
-
- conf = this->private;
-
- prev = cookie;
- local = frame->local;
- loc = &local->loc;
-
- /* This is required for handling stale linkfile deletion,
- * or any more call which happens from this 'loc'.
- */
- if (!op_ret && gf_uuid_is_null (local->gfid))
- memcpy (local->gfid, stbuf->ia_gfid, 16);
-
- gf_msg_debug (this->name, op_errno,
- "fresh_lookup returned for %s with op_ret %d",
- loc->path, op_ret);
-
- if (!conf->vch_forced) {
- ret = dict_get_uint32 (xattr, conf->commithash_xattr_name,
- &vol_commit_hash);
- if (ret == 0) {
- conf->vol_commit_hash = vol_commit_hash;
- }
- }
-
- if (ENTRY_MISSING (op_ret, op_errno)) {
-
- if (1 == conf->subvolume_cnt) {
- /* No need to lookup again */
- goto out;
- }
-
- gf_msg_debug (this->name, 0,
- "Entry %s missing on subvol %s",
- loc->path, prev->name);
-
- /* lookup-optimize supercedes lookup-unhashed settings,
- * - so if it is set, do not process search_unhashed
- * - except, in the case of rebalance deamon, we want to
- * force the lookup_everywhere behavior */
- if (!conf->defrag && conf->lookup_optimize && loc->parent) {
- ret = dht_inode_ctx_layout_get (loc->parent, this,
- &parent_layout);
- if (ret || !parent_layout ||
- (parent_layout->commit_hash !=
- conf->vol_commit_hash)) {
- gf_msg_debug (this->name, 0,
- "hashes don't match (ret - %d,"
- " parent_layout - %p, parent_hash - %x,"
- " vol_hash - %x), do global lookup",
- ret, parent_layout,
- (parent_layout ?
- parent_layout->commit_hash : -1),
- conf->vol_commit_hash);
- local->op_errno = ENOENT;
- dht_lookup_everywhere (frame, this, loc);
- return 0;
- }
- } else {
- if (conf->search_unhashed ==
- GF_DHT_LOOKUP_UNHASHED_ON) {
- local->op_errno = ENOENT;
- dht_lookup_everywhere (frame, this, loc);
- return 0;
- }
+dht_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+ struct iatt *postparent)
+{
+ char is_linkfile = 0;
+ char is_dir = 0;
+ xlator_t *subvol = NULL;
+ dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ loc_t *loc = NULL;
+ xlator_t *prev = NULL;
+ int ret = 0;
+ uint32_t vol_commit_hash = 0;
+
+ GF_VALIDATE_OR_GOTO("dht", frame, err);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO("dht", cookie, out);
+ GF_VALIDATE_OR_GOTO("dht", this->private, out);
+
+ conf = this->private;
+
+ prev = cookie;
+ local = frame->local;
+ loc = &local->loc;
+
+ gf_msg_debug(this->name, op_errno,
+ "%s: fresh_lookup on %s returned with op_ret %d", loc->path,
+ prev->name, op_ret);
+
+ if (op_ret == -1) {
+ if (ENTRY_MISSING(op_ret, op_errno)) {
+ if (1 == conf->subvolume_cnt) {
+ /* No need to lookup again */
+ goto out;
+ }
- if ((conf->search_unhashed ==
- GF_DHT_LOOKUP_UNHASHED_AUTO) &&
- (loc->parent)) {
- ret = dht_inode_ctx_layout_get (loc->parent,
- this,
- &parent_layout);
- if (ret || !parent_layout)
- goto out;
- if (parent_layout->search_unhashed) {
- local->op_errno = ENOENT;
- dht_lookup_everywhere (frame, this,
- loc);
- return 0;
- }
- }
- }
- }
+ gf_msg_debug(this->name, 0, "Entry %s missing on subvol %s",
+ loc->path, prev->name);
- if (op_ret == 0) {
- is_dir = check_is_dir (inode, stbuf, xattr);
- if (is_dir) {
- local->inode = inode_ref (inode);
- local->xattr = dict_ref (xattr);
- }
- }
+ if (dht_should_lookup_everywhere(this, conf, loc)) {
+ local->op_errno = ENOENT;
+ dht_lookup_everywhere(frame, this, loc);
+ return 0;
+ }
- if (is_dir || (op_ret == -1 && op_errno == ENOTCONN)) {
- dht_lookup_directory (frame, this, &local->loc);
+ } else {
+ /* posix returns ENODATA if the gfid is not set but the client and
+ * server protocol layers do not send the stbuf. We need to
+ * heal this so check if this is a directory on the other subvols.
+ */
+ if ((op_errno == ENOTCONN) || (op_errno == ENODATA)) {
+ dht_lookup_directory(frame, this, &local->loc);
return 0;
+ }
}
-
- if (op_ret == -1) {
- gf_msg_debug (this->name, op_errno,
- "Lookup of %s for subvolume"
- " %s failed", loc->path,
- prev->name);
- goto out;
+ gf_msg_debug(this->name, op_errno, "%s: Lookup on subvolume %s failed",
+ loc->path, prev->name);
+ goto out;
+ }
+
+ /* Lookup succeeded - op_ret = 0 */
+
+ /* This is required for handling stale linkfile deletion,
+ * or any more call which happens from this 'loc'.
+ */
+ if (gf_uuid_is_null(local->gfid)) {
+ /*This is set from the first successful response*/
+ memcpy(local->gfid, stbuf->ia_gfid, 16);
+ }
+
+ if (!conf->vch_forced) {
+ /* Update the commit hash in conf if it is found */
+ ret = dict_get_uint32(xattr, conf->commithash_xattr_name,
+ &vol_commit_hash);
+ if (ret == 0) {
+ conf->vol_commit_hash = vol_commit_hash;
}
+ }
- is_linkfile = check_is_linkfile (inode, stbuf, xattr,
- conf->link_xattr_name);
-
- if (!is_linkfile) {
- /* non-directory and not a linkfile */
-
- ret = dht_layout_preset (this, prev, inode);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LAYOUT_PRESET_FAILED,
- "could not set pre-set layout for subvolume %s",
- prev->name);
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
- goto out;
- }
+ is_dir = check_is_dir(inode, stbuf, xattr);
+ if (is_dir) {
+ /* A directory is present on all subvols, send the lookup to
+ * all subvols now */
+ local->inode = inode_ref(inode);
+ local->xattr = dict_ref(xattr);
+ dht_lookup_directory(frame, this, &local->loc);
+ return 0;
+ }
- subvol = dht_linkfile_subvol (this, inode, stbuf, xattr);
- if (!subvol) {
+ is_linkfile = check_is_linkfile(inode, stbuf, xattr, conf->link_xattr_name);
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_SUBVOL_INFO, "linkfile not having link "
- "subvol for %s", loc->path);
+ if (!is_linkfile) {
+ /* non-directory and not a linkto file. This is a data file
+ * Update the layout to point to the cached subvol
+ */
- gf_msg_debug (this->name, 0,
- "linkfile not having link subvolume. path=%s",
- loc->path);
- dht_lookup_everywhere (frame, this, loc);
- return 0;
+ ret = dht_layout_preset(this, prev, inode);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_LAYOUT_PRESET_FAILED,
+ "%s: could not set pre-set layout for subvolume %s",
+ loc->path, prev->name);
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto out;
}
+ goto out;
+ }
- gf_msg_debug (this->name, 0,
- "Calling lookup on linkto target %s for path %s",
- subvol->name, loc->path);
-
- STACK_WIND_COOKIE (frame, dht_lookup_linkfile_cbk, subvol,
- subvol, subvol->fops->lookup,
- &local->loc, local->xattr_req);
-
+ /* This is a linkto file. Get the value of the target subvol from the
+ * linkto xattr and lookup there to see if the file exists
+ */
+ subvol = dht_linkfile_subvol(this, inode, stbuf, xattr);
+ if (!subvol) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_SUBVOL_INFO,
+ "%s: No link subvol for linkto", loc->path);
+ dht_lookup_everywhere(frame, this, loc);
return 0;
+ }
-out:
- /*
- * FIXME: postparent->ia_size and postparent->st_blocks do not have
- * correct values. since, postparent corresponds to a directory these
- * two members should have values equal to sum of corresponding values
- * from each of the subvolume. See dht_iatt_merge for reference.
- */
+ gf_msg_debug(this->name, 0, "%s: Calling lookup on linkto target %s",
+ loc->path, subvol->name);
- if (!op_ret && local && local->loc.parent) {
- dht_inode_ctx_time_update (local->loc.parent, this,
- postparent, 1);
- }
+ STACK_WIND_COOKIE(frame, dht_lookup_linkfile_cbk, subvol, subvol,
+ subvol->fops->lookup, &local->loc, local->xattr_req);
+
+ return 0;
- DHT_STRIP_PHASE1_FLAGS (stbuf);
- dht_set_fixed_dir_stat (postparent);
- DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, stbuf, xattr,
- postparent);
+out:
+ /*
+ * FIXME: postparent->ia_size and postparent->st_blocks do not have
+ * correct values. since, postparent corresponds to a directory these
+ * two members should have values equal to sum of corresponding values
+ * from each of the subvolume. See dht_iatt_merge for reference.
+ */
+
+ if (!op_ret && local && local->loc.parent) {
+ dht_inode_ctx_time_update(local->loc.parent, this, postparent, 1);
+ }
+
+ DHT_STRIP_PHASE1_FLAGS(stbuf);
+ dht_set_fixed_dir_stat(postparent);
+ DHT_STACK_UNWIND(lookup, frame, op_ret, op_errno, inode, stbuf, xattr,
+ postparent);
err:
- return 0;
+ return 0;
}
-/* For directories, check if acl xattrs have been requested (by the acl xlator),
- * if not, request for them. These xattrs are needed for dht dir self-heal to
- * perform proper self-healing of dirs
+/* For directories, check if acl xattrs have been requested (by the acl
+ * xlator), if not, request for them. These xattrs are needed for dht dir
+ * self-heal to perform proper self-healing of dirs
*/
-void
-dht_check_and_set_acl_xattr_req (inode_t *inode, dict_t *xattr_req)
+static void
+dht_check_and_set_acl_xattr_req(xlator_t *this, dict_t *xattr_req)
{
- int ret = 0;
-
- GF_ASSERT (inode);
- GF_ASSERT (xattr_req);
+ int ret = 0;
- if (inode->ia_type != IA_IFDIR)
- return;
+ GF_ASSERT(xattr_req);
- if (!dict_get (xattr_req, POSIX_ACL_ACCESS_XATTR)) {
- ret = dict_set_int8 (xattr_req, POSIX_ACL_ACCESS_XATTR, 0);
- if (ret)
- gf_msg (THIS->name, GF_LOG_WARNING, -ret,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value:key = %s",
- POSIX_ACL_ACCESS_XATTR);
- }
+ if (!dict_get(xattr_req, POSIX_ACL_ACCESS_XATTR)) {
+ ret = dict_set_int8(xattr_req, POSIX_ACL_ACCESS_XATTR, 0);
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, -ret, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary value:key = %s",
+ POSIX_ACL_ACCESS_XATTR);
+ }
- if (!dict_get (xattr_req, POSIX_ACL_DEFAULT_XATTR)) {
- ret = dict_set_int8 (xattr_req, POSIX_ACL_DEFAULT_XATTR, 0);
- if (ret)
- gf_msg (THIS->name, GF_LOG_WARNING, -ret,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value:key = %s",
- POSIX_ACL_DEFAULT_XATTR);
- }
+ if (!dict_get(xattr_req, POSIX_ACL_DEFAULT_XATTR)) {
+ ret = dict_set_int8(xattr_req, POSIX_ACL_DEFAULT_XATTR, 0);
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, -ret, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary value:key = %s",
+ POSIX_ACL_DEFAULT_XATTR);
+ }
- return;
+ return;
}
-int
-dht_lookup (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xattr_req)
-{
- xlator_t *subvol = NULL;
- xlator_t *hashed_subvol = NULL;
- xlator_t *mds_subvol = NULL;
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- int ret = -1;
- int op_errno = -1;
- dht_layout_t *layout = NULL;
- int i = 0;
- int call_cnt = 0;
- int gen = 0;
- loc_t new_loc = {0,};
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- conf = this->private;
- if (!conf)
- goto err;
+/* for directories, we need the following info:
+ * the layout : trusted.glusterfs.dht
+ * the mds information : trusted.glusterfs.dht.mds
+ * the acl info: See above
+ */
+static int
+dht_set_dir_xattr_req(xlator_t *this, loc_t *loc, dict_t *xattr_req)
+{
+ int ret = -EINVAL;
+ dht_conf_t *conf = NULL;
+
+ conf = this->private;
+ if (!conf) {
+ goto err;
+ }
+
+ if (!xattr_req) {
+ goto err;
+ }
+
+ /* Xattr to get the layout for a directory
+ */
+ ret = dict_set_uint32(xattr_req, conf->xattr_name, 4 * 4);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary value:key = %s for "
+ "path %s",
+ conf->xattr_name, loc->path);
+ goto err;
+ }
+
+ /*Non-fatal failure */
+ ret = dict_set_uint32(xattr_req, conf->mds_xattr_key, 4);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary value:key = %s for "
+ "path %s",
+ conf->mds_xattr_key, loc->path);
+ }
+
+ dht_check_and_set_acl_xattr_req(this, xattr_req);
+ ret = 0;
+err:
+ return ret;
+}
- local = dht_local_init (frame, loc, NULL, GF_FOP_LOOKUP);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+/* If the hashed subvol is present, send the lookup to only that subvol first.
+ * If no hashed subvol, send a lookup to all subvols and proceed based on the
+ * responses.
+ */
+static int
+dht_do_fresh_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc)
+{
+ int ret = -1;
+ dht_conf_t *conf = NULL;
+ xlator_t *hashed_subvol = NULL;
+ dht_local_t *local = NULL;
+ int op_errno = -1;
+ int call_cnt = 0;
+ int i = 0;
+
+ conf = this->private;
+ if (!conf) {
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ local = frame->local;
+ if (!local) {
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ /* Since we don't know whether this is a file or a directory,
+ * request all xattrs*/
+ ret = dht_set_file_xattr_req(this, loc, local->xattr_req);
+ if (ret) {
+ op_errno = -ret;
+ goto err;
+ }
+
+ ret = dht_set_dir_xattr_req(this, loc, local->xattr_req);
+ if (ret) {
+ op_errno = -ret;
+ goto err;
+ }
+
+ /* Fuse sets a random value in gfid-req. If the gfid is missing
+ * on one or more subvols, posix will set the gfid to this value,
+ * causing GFID mismatches for directories. Remove the value fuse
+ * has sent before sending the lookup.
+ */
+ ret = dict_get_gfuuid(local->xattr_req, "gfid-req", &local->gfid_req);
+ if (ret) {
+ gf_msg_debug(this->name, 0, "%s: No gfid-req available", loc->path);
+ } else {
+ dict_del(local->xattr_req, "gfid-req");
+ }
+ /* This should have been set in dht_lookup */
+ hashed_subvol = local->hashed_subvol;
+
+ if (!hashed_subvol) {
+ gf_msg_debug(this->name, 0,
+ "%s: no subvolume in layout for path, "
+ "checking on all the subvols to see if "
+ "it is a directory",
+ loc->path);
- ret = dht_filter_loc_subvol_key (this, loc, &new_loc,
- &hashed_subvol);
- if (ret) {
- loc_wipe (&local->loc);
- ret = loc_dup (&new_loc, &local->loc);
-
- /* we no more need 'new_loc' entries */
- loc_wipe (&new_loc);
-
- /* check if loc_dup() is successful */
- if (ret == -1) {
- op_errno = errno;
- gf_msg_debug (this->name, errno,
- "copying location failed for path=%s",
- loc->path);
- goto err;
- }
- }
+ call_cnt = conf->subvolume_cnt;
+ local->call_cnt = call_cnt;
- if (xattr_req) {
- local->xattr_req = dict_ref (xattr_req);
- } else {
- local->xattr_req = dict_new ();
+ /* Allocate a layout. This will be populated and saved in
+ * the dht inode_ctx on successful lookup
+ */
+ local->layout = dht_layout_new(this, conf->subvolume_cnt);
+ if (!local->layout) {
+ op_errno = ENOMEM;
+ goto err;
}
- ret = dict_set_uint32 (local->xattr_req, conf->mds_xattr_key, 4);
+ gf_msg_debug(this->name, 0,
+ "%s: Found null hashed subvol. Calling lookup"
+ " on all nodes.",
+ loc->path);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value:key = %s for "
- "path %s", conf->mds_xattr_key, loc->path);
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND_COOKIE(frame, dht_lookup_dir_cbk, conf->subvolumes[i],
+ conf->subvolumes[i],
+ conf->subvolumes[i]->fops->lookup, &local->loc,
+ local->xattr_req);
}
+ return 0;
+ }
+ /* if the hashed_subvol is non-null, send the lookup there first so
+ * as to see whether we have a file or a directory */
+ gf_msg_debug(this->name, 0, "%s: Calling fresh lookup on %s", loc->path,
+ hashed_subvol->name);
- /* Nameless lookup */
-
- if (gf_uuid_is_null (loc->pargfid) && !gf_uuid_is_null (loc->gfid) &&
- !__is_root_gfid (loc->inode->gfid)) {
- local->cached_subvol = NULL;
- dht_discover (frame, this, loc);
- return 0;
- }
+ STACK_WIND_COOKIE(frame, dht_lookup_cbk, hashed_subvol, hashed_subvol,
+ hashed_subvol->fops->lookup, loc, local->xattr_req);
+ return 0;
+err:
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ return 0;
+}
- if (__is_root_gfid(loc->gfid)) {
- ret = dict_set_uint32 (local->xattr_req,
- conf->commithash_xattr_name,
- sizeof(uint32_t));
+static int
+dht_do_revalidate(call_frame_t *frame, xlator_t *this, loc_t *loc)
+{
+ xlator_t *subvol = NULL;
+ xlator_t *mds_subvol = NULL;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int ret = -1;
+ int op_errno = -1;
+ dht_layout_t *layout = NULL;
+ int i = 0;
+ int call_cnt = 0;
+ int gen = 0;
+
+ conf = this->private;
+ if (!conf) {
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ local = frame->local;
+ if (!local) {
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ layout = local->layout;
+ if (!layout) {
+ gf_msg_debug(this->name, 0,
+ "path = %s. No layout found in the inode ctx.", loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ /* Generation number has changed. This layout may be stale. */
+ if (layout->gen && (layout->gen < conf->gen)) {
+ gen = layout->gen;
+ dht_layout_unref(this, local->layout);
+ local->layout = NULL;
+ local->cached_subvol = NULL;
+
+ gf_msg_debug(this->name, 0,
+ "path = %s. In memory layout may be stale."
+ "(layout->gen (%d) is less than "
+ "conf->gen (%d)). Calling fresh lookup.",
+ loc->path, gen, conf->gen);
+
+ dht_do_fresh_lookup(frame, this, loc);
+ return 0;
+ }
+
+ local->inode = inode_ref(loc->inode);
+
+ /* Since we don't know whether this has changed,
+ * request all xattrs*/
+ ret = dht_set_file_xattr_req(this, loc, local->xattr_req);
+ if (ret) {
+ op_errno = -ret;
+ goto err;
+ }
+
+ ret = dht_set_dir_xattr_req(this, loc, local->xattr_req);
+ if (ret) {
+ op_errno = -ret;
+ goto err;
+ }
+
+ if (IA_ISDIR(local->inode->ia_type)) {
+ ret = dht_inode_ctx_mdsvol_get(local->inode, this, &mds_subvol);
+ if (ret || !mds_subvol) {
+ gf_msg_debug(this->name, 0, "path = %s. No mds subvol in inode ctx",
+ local->loc.path);
}
+ local->mds_subvol = mds_subvol;
+ local->call_cnt = conf->subvolume_cnt;
- if (!hashed_subvol)
- hashed_subvol = dht_subvol_get_hashed (this, loc);
- local->hashed_subvol = hashed_subvol;
-
-
- /* The entry has been looked up before and has an inode_ctx set
+ /* local->call_cnt will change as responses are processed. Always use a
+ * local copy to loop through the STACK_WIND calls
*/
- if (is_revalidate (loc)) {
- layout = local->layout;
- if (!layout) {
- gf_msg_debug (this->name, 0,
- "Revalidate lookup without cache."
- " path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
-
- if (layout->gen && (layout->gen < conf->gen)) {
- gf_msg_trace (this->name, 0,
- "incomplete layout failure for path=%s",
- loc->path);
-
- gen = layout->gen;
- dht_layout_unref (this, local->layout);
- local->layout = NULL;
- local->cached_subvol = NULL;
-
- gf_msg_debug(this->name, 0,
- "Called revalidate lookup for %s, "
- "but layout->gen (%d) is less than "
- "conf->gen (%d), calling fresh_lookup",
- loc->path, gen, conf->gen);
-
- goto do_fresh_lookup;
- }
- local->inode = inode_ref (loc->inode);
+ call_cnt = local->call_cnt;
- ret = dict_set_uint32 (local->xattr_req,
- conf->xattr_name, 4 * 4);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value:key = %s for "
- "path %s", conf->xattr_name, loc->path);
- goto err;
- }
-
- /* need it in case file is not found on cached file
- * on revalidate path and we may encounter linkto files on
- * with dht_lookup_everywhere*/
- ret = dict_set_uint32 (local->xattr_req,
- conf->link_xattr_name, 256);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value:key = %s for "
- "path %s", conf->link_xattr_name, loc->path);
- goto err;
- }
- if (IA_ISDIR (local->inode->ia_type)) {
- ret = dht_inode_ctx_mdsvol_get (local->inode, this,
- &mds_subvol);
- if (ret || !mds_subvol) {
- gf_msg_debug (this->name, 0,
- "Failed to get mds subvol for path %s",
- local->loc.path);
- }
- local->mds_subvol = mds_subvol;
- local->call_cnt = call_cnt = conf->subvolume_cnt;
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND_COOKIE (frame, dht_revalidate_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->lookup,
- loc, local->xattr_req);
- }
- return 0;
- }
-
- call_cnt = local->call_cnt = layout->cnt;
-
- /* need it for self-healing linkfiles which is
- 'in-migration' state */
- ret = dict_set_uint32 (local->xattr_req,
- GLUSTERFS_OPEN_FD_COUNT, 4);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value:key = %s for "
- "path %s", GLUSTERFS_OPEN_FD_COUNT, loc->path);
- goto err;
- }
- /* need it for dir self-heal */
- dht_check_and_set_acl_xattr_req (loc->inode, local->xattr_req);
-
- for (i = 0; i < call_cnt; i++) {
- subvol = layout->list[i].xlator;
-
- gf_msg_debug (this->name, 0, "calling "
- "revalidate lookup for %s at %s",
- loc->path, subvol->name);
-
- STACK_WIND_COOKIE (frame, dht_revalidate_cbk, subvol,
- subvol, subvol->fops->lookup,
- &local->loc, local->xattr_req);
-
- }
- } else {
- do_fresh_lookup:
- /* TODO: remove the hard-coding */
- ret = dict_set_uint32 (local->xattr_req,
- conf->xattr_name, 4 * 4);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value:key = %s for "
- "path %s", conf->xattr_name, loc->path);
- goto err;
- }
-
- ret = dict_set_uint32 (local->xattr_req,
- conf->link_xattr_name, 256);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value:key = %s for "
- "path %s", conf->link_xattr_name, loc->path);
- goto err;
- }
- /* need it for self-healing linkfiles which is
- 'in-migration' state */
- ret = dict_set_uint32 (local->xattr_req,
- GLUSTERFS_OPEN_FD_COUNT, 4);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value:key = %s for "
- "path %s", GLUSTERFS_OPEN_FD_COUNT, loc->path);
- goto err;
- }
- /* need it for dir self-heal */
- dht_check_and_set_acl_xattr_req (loc->inode, local->xattr_req);
-
- if (!hashed_subvol) {
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND_COOKIE(frame, dht_revalidate_cbk, conf->subvolumes[i],
+ conf->subvolumes[i],
+ conf->subvolumes[i]->fops->lookup, loc,
+ local->xattr_req);
+ }
+ return 0;
+ }
- gf_msg_debug (this->name, 0,
- "no subvolume in layout for path=%s, "
- "checking on all the subvols to see if "
- "it is a directory", loc->path);
+ /* If not a dir, this should be 1 */
+ local->call_cnt = layout->cnt;
+ call_cnt = local->call_cnt;
- call_cnt = conf->subvolume_cnt;
- local->call_cnt = call_cnt;
+ for (i = 0; i < call_cnt; i++) {
+ subvol = layout->list[i].xlator;
- local->layout = dht_layout_new (this,
- conf->subvolume_cnt);
- if (!local->layout) {
- op_errno = ENOMEM;
- goto err;
- }
+ gf_msg_debug(this->name, 0,
+ "path = %s. Calling "
+ "revalidate lookup on %s",
+ loc->path, subvol->name);
- gf_msg_debug (this->name, 0,
- "Found null hashed subvol. Calling lookup"
- " on all nodes.");
+ STACK_WIND_COOKIE(frame, dht_revalidate_cbk, subvol, subvol,
+ subvol->fops->lookup, &local->loc, local->xattr_req);
+ }
+ return 0;
+err:
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ return 0;
+}
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND_COOKIE (frame, dht_lookup_dir_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->lookup,
- &local->loc, local->xattr_req);
- }
- return 0;
- }
+/* Depending on the input, decide if this is a:
+ * fresh-lookup: loc->name is provided but no dht inode ctx
+ * revalidation: loc->name is provided, dht inode ctx is present
+ * discover: gfid based nameless lookup.
+ */
- gf_msg_debug (this->name, 0, "Calling fresh lookup for %s on"
- " %s", loc->path, hashed_subvol->name);
+int
+dht_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr_req)
+{
+ xlator_t *hashed_subvol = NULL;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int ret = -1;
+ int op_errno = -1;
+ loc_t new_loc = {
+ 0,
+ };
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
+
+ conf = this->private;
+ if (!conf)
+ goto err;
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_LOOKUP);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ ret = dht_filter_loc_subvol_key(this, loc, &new_loc, &hashed_subvol);
+ if (ret) {
+ loc_wipe(&local->loc);
+ ret = loc_dup(&new_loc, &local->loc);
+
+ /* we no longer need 'new_loc' entries */
+ loc_wipe(&new_loc);
+
+ /* check if loc_dup() is successful */
+ if (ret == -1) {
+ op_errno = errno;
+ gf_msg_debug(this->name, errno,
+ "copying location failed for path=%s", loc->path);
+ goto err;
+ }
+ }
+
+ if (xattr_req) {
+ local->xattr_req = dict_ref(xattr_req);
+ } else {
+ local->xattr_req = dict_new();
+ }
+
+ /* Nameless lookup */
+
+ /* This is usually sent by NFS. Lookups are done based on the gfid and
+ * no name information is available. Without the name, dht cannot calculate
+ * the hash and has to send a lookup to all subvols.
+ */
+ if (gf_uuid_is_null(loc->pargfid) && !gf_uuid_is_null(loc->gfid) &&
+ !__is_root_gfid(loc->inode->gfid)) {
+ local->cached_subvol = NULL;
+ dht_do_discover(frame, this, loc);
+ return 0;
+ }
+
+ if (loc_is_root(loc)) {
+ /* Request the DHT commit hash xattr (trusted.glusterfs.dht.commithash)
+ * set on the brick root.
+ */
+ ret = dict_set_uint32(local->xattr_req, conf->commithash_xattr_name,
+ sizeof(uint32_t));
+ }
- STACK_WIND_COOKIE (frame, dht_lookup_cbk, hashed_subvol,
- hashed_subvol, hashed_subvol->fops->lookup,
- loc, local->xattr_req);
- }
+ if (!hashed_subvol)
+ hashed_subvol = dht_subvol_get_hashed(this, loc);
+ local->hashed_subvol = hashed_subvol;
+ if (is_revalidate(loc)) {
+ /* The entry has been looked up before and has a dht inode_ctx
+ */
+ dht_do_revalidate(frame, this, loc);
+ return 0;
+ } else {
+ /* Entry has not been looked up before
+ */
+ dht_do_fresh_lookup(frame, this, loc);
return 0;
+ }
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL,
- NULL);
- return 0;
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ return 0;
}
-int
-dht_unlink_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+static int
+dht_unlink_linkfile_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
- local = frame->local;
- prev = cookie;
-
- LOCK (&frame->lock);
- {
- if ((op_ret == -1) && !((op_errno == ENOENT) ||
- (op_errno == ENOTCONN))) {
- local->op_errno = op_errno;
- gf_msg_debug (this->name, op_errno,
- "Unlink link: subvolume %s"
- " returned -1",
- prev->name);
- goto unlock;
- }
+ local = frame->local;
+ prev = cookie;
- local->op_ret = 0;
+ LOCK(&frame->lock);
+ {
+ if ((op_ret == -1) &&
+ !((op_errno == ENOENT) || (op_errno == ENOTCONN))) {
+ local->op_errno = op_errno;
+ UNLOCK(&frame->lock);
+ gf_msg_debug(this->name, op_errno,
+ "Unlink link: subvolume %s returned -1", prev->name);
+ goto post_unlock;
}
-unlock:
- UNLOCK (&frame->lock);
- dht_set_fixed_dir_stat (&local->preparent);
- dht_set_fixed_dir_stat (&local->postparent);
- DHT_STACK_UNWIND (unlink, frame, local->op_ret, local->op_errno,
- &local->preparent, &local->postparent, xdata);
+ local->op_ret = 0;
+ }
+ UNLOCK(&frame->lock);
+post_unlock:
+ dht_set_fixed_dir_stat(&local->preparent);
+ dht_set_fixed_dir_stat(&local->postparent);
+ DHT_STACK_UNWIND(unlink, frame, local->op_ret, local->op_errno,
+ &local->preparent, &local->postparent, xdata);
- return 0;
+ return 0;
}
-int
-dht_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+static int
+dht_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- xlator_t *hashed_subvol = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ xlator_t *hashed_subvol = NULL;
- local = frame->local;
- prev = cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- if (op_errno != ENOENT) {
- local->op_ret = -1;
- local->op_errno = op_errno;
- } else {
- local->op_ret = 0;
- }
- gf_msg_debug (this->name, op_errno,
- "Unlink: subvolume %s returned -1",
- prev->name);
- goto unlock;
- }
+ local = frame->local;
+ prev = cookie;
+ LOCK(&frame->lock);
+ {
+ if (op_ret == -1) {
+ if (op_errno != ENOENT) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ } else {
local->op_ret = 0;
-
- local->postparent = *postparent;
- local->preparent = *preparent;
-
- if (local->loc.parent) {
- dht_inode_ctx_time_update (local->loc.parent, this,
- &local->preparent, 0);
- dht_inode_ctx_time_update (local->loc.parent, this,
- &local->postparent, 1);
- }
+ }
+ UNLOCK(&frame->lock);
+ gf_msg_debug(this->name, op_errno,
+ "Unlink: subvolume %s returned -1", prev->name);
+ goto post_unlock;
}
-unlock:
- UNLOCK (&frame->lock);
- if (!local->op_ret) {
- hashed_subvol = dht_subvol_get_hashed (this, &local->loc);
- if (hashed_subvol && hashed_subvol != local->cached_subvol) {
- /*
- * If hashed and cached are different, then we need
- * to unlink linkfile from hashed subvol if data
- * file is deleted successfully
- */
- STACK_WIND_COOKIE (frame, dht_unlink_linkfile_cbk,
- hashed_subvol, hashed_subvol,
- hashed_subvol->fops->unlink,
- &local->loc, local->flags, xdata);
- return 0;
- }
- }
+ local->op_ret = 0;
- dht_set_fixed_dir_stat (&local->preparent);
- dht_set_fixed_dir_stat (&local->postparent);
- DHT_STACK_UNWIND (unlink, frame, local->op_ret, local->op_errno,
- &local->preparent, &local->postparent, xdata);
+ local->postparent = *postparent;
+ local->preparent = *preparent;
- return 0;
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update(local->loc.parent, this,
+ &local->preparent, 0);
+ dht_inode_ctx_time_update(local->loc.parent, this,
+ &local->postparent, 1);
+ }
+ }
+ UNLOCK(&frame->lock);
+post_unlock:
+ if (!local->op_ret) {
+ hashed_subvol = dht_subvol_get_hashed(this, &local->loc);
+ if (hashed_subvol && hashed_subvol != local->cached_subvol) {
+ /*
+ * If hashed and cached are different, then we need
+ * to unlink linkfile from hashed subvol if data
+ * file is deleted successfully
+ */
+ STACK_WIND_COOKIE(frame, dht_unlink_linkfile_cbk, hashed_subvol,
+ hashed_subvol, hashed_subvol->fops->unlink,
+ &local->loc, local->flags, xdata);
+ return 0;
+ }
+ }
+
+ dht_set_fixed_dir_stat(&local->preparent);
+ dht_set_fixed_dir_stat(&local->postparent);
+ DHT_STACK_UNWIND(unlink, frame, local->op_ret, local->op_errno,
+ &local->preparent, &local->postparent, xdata);
+
+ return 0;
}
static int
-dht_common_setxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xdata)
+dht_common_setxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- DHT_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
- return 0;
+ DHT_STACK_UNWIND(setxattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
-
-
static int
-dht_fix_layout_setxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xdata)
+dht_fix_layout_setxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- dht_layout_t *layout = NULL;
+ dht_local_t *local = NULL;
+ dht_layout_t *layout = NULL;
- if (op_ret == 0) {
+ if (op_ret == 0) {
+ /* update the layout in the inode ctx */
+ local = frame->local;
+ layout = local->selfheal.layout;
- /* update the layout in the inode ctx */
- local = frame->local;
- layout = local->selfheal.layout;
+ dht_layout_set(this, local->loc.inode, layout);
+ }
- dht_layout_set (this, local->loc.inode, layout);
- }
-
- DHT_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
- return 0;
+ DHT_STACK_UNWIND(setxattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
-
-int
-dht_err_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+static int
+dht_err_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- xlator_t *prev = NULL;
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ xlator_t *prev = NULL;
- local = frame->local;
- prev = cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->name);
- goto unlock;
- }
+ local = frame->local;
+ prev = cookie;
- local->op_ret = 0;
+ LOCK(&frame->lock);
+ {
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ UNLOCK(&frame->lock);
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->name);
+ goto post_unlock;
}
-unlock:
- UNLOCK (&frame->lock);
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- DHT_STACK_UNWIND (setxattr, frame, local->op_ret,
- local->op_errno, NULL);
- }
+ local->op_ret = 0;
+ }
+ UNLOCK(&frame->lock);
+post_unlock:
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt)) {
+ if ((local->fop == GF_FOP_SETXATTR) ||
+ (local->fop == GF_FOP_FSETXATTR)) {
+ DHT_STACK_UNWIND(setxattr, frame, local->op_ret, local->op_errno,
+ NULL);
+ /* 'local' itself may not be valid after this */
+ goto out;
+ }
+ if ((local->fop == GF_FOP_REMOVEXATTR) ||
+ (local->fop == GF_FOP_FREMOVEXATTR)) {
+ DHT_STACK_UNWIND(removexattr, frame, local->op_ret, local->op_errno,
+ NULL);
+ }
+ }
- return 0;
+out:
+ return 0;
}
/* Set the value[] of key into dict after convert from
host byte order to network byte order
*/
-int32_t dht_dict_set_array (dict_t *dict, char *key, int32_t value[],
- int32_t size)
+int32_t
+dht_dict_set_array(dict_t *dict, char *key, int32_t value[], int32_t size)
{
- int ret = -1;
- int32_t *ptr = NULL;
- int32_t vindex;
+ int ret = -1;
+ int32_t *ptr = NULL;
+ int32_t vindex;
- if (value == NULL) {
- return -EINVAL;
- }
+ if (value == NULL) {
+ return -EINVAL;
+ }
- ptr = GF_MALLOC(sizeof(int32_t) * size, gf_common_mt_char);
- if (ptr == NULL) {
- return -ENOMEM;
- }
- for (vindex = 0; vindex < size; vindex++) {
- ptr[vindex] = hton32(value[vindex]);
- }
- ret = dict_set_bin(dict, key, ptr, sizeof(int32_t) * size);
- if (ret)
- GF_FREE (ptr);
- return ret;
+ ptr = GF_MALLOC(sizeof(int32_t) * size, gf_common_mt_char);
+ if (ptr == NULL) {
+ return -ENOMEM;
+ }
+ for (vindex = 0; vindex < size; vindex++) {
+ ptr[vindex] = hton32(value[vindex]);
+ }
+ ret = dict_set_bin(dict, key, ptr, sizeof(int32_t) * size);
+ if (ret)
+ GF_FREE(ptr);
+ return ret;
}
-int
-dht_common_mds_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+static int
+dht_common_mds_xattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- call_frame_t *prev = cookie;
+ dht_local_t *local = NULL;
+ call_frame_t *prev = cookie;
- local = frame->local;
+ local = frame->local;
- if (op_ret)
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->this->name);
+ if (op_ret)
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->this->name);
- if (local->fop == GF_FOP_SETXATTR) {
- DHT_STACK_UNWIND (setxattr, frame, 0, op_errno, local->xdata);
- } else {
- DHT_STACK_UNWIND (fsetxattr, frame, 0, op_errno, local->xdata);
- }
- return 0;
+ if (local->fop == GF_FOP_SETXATTR) {
+ DHT_STACK_UNWIND(setxattr, frame, 0, op_errno, local->xdata);
+ /* 'local' itself may not be valid after this */
+ goto out;
+ }
+
+ if (local->fop == GF_FOP_FSETXATTR) {
+ DHT_STACK_UNWIND(fsetxattr, frame, 0, op_errno, local->xdata);
+ /* 'local' itself may not be valid after this */
+ goto out;
+ }
+
+ if (local->fop == GF_FOP_REMOVEXATTR) {
+ DHT_STACK_UNWIND(removexattr, frame, 0, op_errno, NULL);
+ /* 'local' itself may not be valid after this */
+ goto out;
+ }
+
+ if (local->fop == GF_FOP_FREMOVEXATTR) {
+ DHT_STACK_UNWIND(fremovexattr, frame, 0, op_errno, NULL);
+ }
+
+out:
+ return 0;
}
/* Code to wind a xattrop call to add 1 on current mds internal xattr
value
*/
-int
-dht_setxattr_non_mds_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+static int
+dht_setxattr_non_mds_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- int ret = 0;
- dict_t *xattrop = NULL;
- int32_t addone[1] = {1};
- call_frame_t *prev = NULL;
- dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ int ret = 0;
+ dict_t *xattrop = NULL;
+ int32_t addone[1] = {1};
+ call_frame_t *prev = NULL;
+ dht_conf_t *conf = NULL;
+
+ local = frame->local;
+ prev = cookie;
+ conf = this->private;
+
+ LOCK(&frame->lock);
+ {
+ if (op_ret && !local->op_ret) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ UNLOCK(&frame->lock);
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->this->name);
+ goto post_unlock;
+ }
+ }
+ UNLOCK(&frame->lock);
+post_unlock:
+ this_call_cnt = dht_frame_return(frame);
+
+ if (is_last_call(this_call_cnt)) {
+ if (!local->op_ret) {
+ xattrop = dict_new();
+ if (!xattrop) {
+ gf_msg(this->name, GF_LOG_ERROR, DHT_MSG_NO_MEMORY, 0,
+ "dictionary creation failed");
+ ret = -1;
+ goto out;
+ }
+ ret = dht_dict_set_array(xattrop, conf->mds_xattr_key, addone, 1);
+ if (ret != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "dictionary set array failed ");
+ ret = -1;
+ goto out;
+ }
+ if ((local->fop == GF_FOP_SETXATTR) ||
+ (local->fop == GF_FOP_REMOVEXATTR)) {
+ STACK_WIND(frame, dht_common_mds_xattrop_cbk, local->mds_subvol,
+ local->mds_subvol->fops->xattrop, &local->loc,
+ GF_XATTROP_ADD_ARRAY, xattrop, NULL);
+ } else {
+ STACK_WIND(frame, dht_common_mds_xattrop_cbk, local->mds_subvol,
+ local->mds_subvol->fops->fxattrop, local->fd,
+ GF_XATTROP_ADD_ARRAY, xattrop, NULL);
+ }
+ } else {
+ if (local->fop == GF_FOP_SETXATTR) {
+ DHT_STACK_UNWIND(setxattr, frame, 0, 0, local->xdata);
+ /* 'local' itself may not be valid after this */
+ goto just_return;
+ }
- local = frame->local;
- prev = cookie;
- conf = this->private;
-
- LOCK (&frame->lock);
- {
- if (op_ret && !local->op_ret) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->this->name);
- }
- }
- UNLOCK (&frame->lock);
- this_call_cnt = dht_frame_return (frame);
-
- if (is_last_call (this_call_cnt)) {
- if (!local->op_ret) {
- xattrop = dict_new ();
- if (!xattrop) {
- gf_msg (this->name, GF_LOG_ERROR,
- DHT_MSG_NO_MEMORY, 0,
- "dictionary creation failed");
- ret = -1;
- goto out;
- }
- ret = dht_dict_set_array (xattrop,
- conf->mds_xattr_key,
- addone, 1);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "dictionary set array failed ");
- ret = -1;
- goto out;
- }
- if (local->fop == GF_FOP_SETXATTR) {
- STACK_WIND (frame, dht_common_mds_xattrop_cbk,
- local->mds_subvol,
- local->mds_subvol->fops->xattrop,
- &local->loc, GF_XATTROP_ADD_ARRAY,
- xattrop, NULL);
- } else {
- STACK_WIND (frame, dht_common_mds_xattrop_cbk,
- local->mds_subvol,
- local->mds_subvol->fops->fxattrop,
- local->fd, GF_XATTROP_ADD_ARRAY,
- xattrop, NULL);
- }
- } else {
- if (local->fop == GF_FOP_SETXATTR)
- DHT_STACK_UNWIND (setxattr, frame, 0, 0, local->xdata);
- else
- DHT_STACK_UNWIND (fsetxattr, frame, 0, 0, local->xdata);
- }
+ if (local->fop == GF_FOP_FSETXATTR) {
+ DHT_STACK_UNWIND(fsetxattr, frame, 0, 0, local->xdata);
+ /* 'local' itself may not be valid after this */
+ goto just_return;
+ }
+
+ if (local->fop == GF_FOP_REMOVEXATTR) {
+ DHT_STACK_UNWIND(removexattr, frame, 0, 0, NULL);
+ /* 'local' itself may not be valid after this */
+ goto just_return;
+ }
+
+ if (local->fop == GF_FOP_FREMOVEXATTR) {
+ DHT_STACK_UNWIND(fremovexattr, frame, 0, 0, NULL);
+ /* 'local' itself may not be valid after this */
+ goto just_return;
+ }
}
+ }
out:
- if (xattrop)
- dict_unref (xattrop);
- if (ret) {
- if (local->fop == GF_FOP_SETXATTR)
- DHT_STACK_UNWIND (setxattr, frame, 0, 0, local->xdata);
- else
- DHT_STACK_UNWIND (fsetxattr, frame, 0, 0, local->xdata);
+ if (ret) {
+ if (local->fop == GF_FOP_SETXATTR) {
+ DHT_STACK_UNWIND(setxattr, frame, 0, 0, local->xdata);
+ /* 'local' itself may not be valid after this */
+ goto just_return;
}
- return 0;
-}
+ if (local->fop == GF_FOP_FSETXATTR) {
+ DHT_STACK_UNWIND(fsetxattr, frame, 0, 0, local->xdata);
+ /* 'local' itself may not be valid after this */
+ goto just_return;
+ }
-int
-dht_setxattr_mds_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+ if (local->fop == GF_FOP_REMOVEXATTR) {
+ DHT_STACK_UNWIND(removexattr, frame, 0, 0, NULL);
+ /* 'local' itself may not be valid after this */
+ goto just_return;
+ }
+
+ if (local->fop == GF_FOP_FREMOVEXATTR) {
+ DHT_STACK_UNWIND(fremovexattr, frame, 0, 0, NULL);
+ }
+ }
+just_return:
+ if (xattrop)
+ dict_unref(xattrop);
+ return 0;
+}
+
+static int
+dht_setxattr_mds_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- call_frame_t *prev = NULL;
- xlator_t *mds_subvol = NULL;
- int i = 0;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ call_frame_t *prev = NULL;
+ xlator_t *mds_subvol = NULL;
+ int i = 0;
- local = frame->local;
- prev = cookie;
- conf = this->private;
- mds_subvol = local->mds_subvol;
+ local = frame->local;
+ prev = cookie;
+ conf = this->private;
+ mds_subvol = local->mds_subvol;
- if (op_ret == -1) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->this->name);
- goto out;
+ if (op_ret == -1) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->this->name);
+ goto out;
+ }
+
+ local->op_ret = 0;
+ local->call_cnt = conf->subvolume_cnt - 1;
+ local->xdata = dict_ref(xdata);
+
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (mds_subvol && (mds_subvol == conf->subvolumes[i]))
+ continue;
+ if (local->fop == GF_FOP_SETXATTR) {
+ STACK_WIND(frame, dht_setxattr_non_mds_cbk, conf->subvolumes[i],
+ conf->subvolumes[i]->fops->setxattr, &local->loc,
+ local->xattr, local->flags, local->xattr_req);
}
- local->op_ret = 0;
- local->call_cnt = conf->subvolume_cnt - 1;
- local->xdata = dict_ref (xdata);
+ if (local->fop == GF_FOP_FSETXATTR) {
+ STACK_WIND(frame, dht_setxattr_non_mds_cbk, conf->subvolumes[i],
+ conf->subvolumes[i]->fops->fsetxattr, local->fd,
+ local->xattr, local->flags, local->xattr_req);
+ }
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (mds_subvol && (mds_subvol == conf->subvolumes[i]))
- continue;
- if (local->fop == GF_FOP_SETXATTR) {
- STACK_WIND (frame, dht_setxattr_non_mds_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->setxattr,
- &local->loc, local->xattr,
- local->flags, local->xattr_req);
- } else {
- STACK_WIND (frame, dht_setxattr_non_mds_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->fsetxattr,
- local->fd, local->xattr,
- local->flags, local->xattr_req);
- }
+ if (local->fop == GF_FOP_REMOVEXATTR) {
+ STACK_WIND(frame, dht_setxattr_non_mds_cbk, conf->subvolumes[i],
+ conf->subvolumes[i]->fops->removexattr, &local->loc,
+ local->key, local->xattr_req);
}
- return 0;
-out:
- if (local->fop == GF_FOP_SETXATTR) {
- DHT_STACK_UNWIND (setxattr, frame, local->op_ret,
- local->op_errno, xdata);
- } else {
- DHT_STACK_UNWIND (fsetxattr, frame, local->op_ret,
- local->op_errno, xdata);
+ if (local->fop == GF_FOP_FREMOVEXATTR) {
+ STACK_WIND(frame, dht_setxattr_non_mds_cbk, conf->subvolumes[i],
+ conf->subvolumes[i]->fops->fremovexattr, local->fd,
+ local->key, local->xattr_req);
}
+ }
- return 0;
+ return 0;
+out:
+ if (local->fop == GF_FOP_SETXATTR) {
+ DHT_STACK_UNWIND(setxattr, frame, local->op_ret, local->op_errno,
+ xdata);
+ /* 'local' itself may not be valid after this */
+ goto just_return;
+ }
+
+ if (local->fop == GF_FOP_FSETXATTR) {
+ DHT_STACK_UNWIND(fsetxattr, frame, local->op_ret, local->op_errno,
+ xdata);
+ /* 'local' itself may not be valid after this */
+ goto just_return;
+ }
+
+ if (local->fop == GF_FOP_REMOVEXATTR) {
+ DHT_STACK_UNWIND(removexattr, frame, local->op_ret, local->op_errno,
+ NULL);
+ /* 'local' itself may not be valid after this */
+ goto just_return;
+ }
+
+ if (local->fop == GF_FOP_FREMOVEXATTR) {
+ DHT_STACK_UNWIND(fremovexattr, frame, local->op_ret, local->op_errno,
+ NULL);
+ }
+
+just_return:
+ return 0;
}
-int
-dht_xattrop_mds_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
+static int
+dht_xattrop_mds_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
{
- dht_local_t *local = NULL;
- call_frame_t *prev = NULL;
+ dht_local_t *local = NULL;
+ call_frame_t *prev = NULL;
- local = frame->local;
- prev = cookie;
+ local = frame->local;
+ prev = cookie;
- if (op_ret == -1) {
- local->op_errno = op_errno;
- local->op_ret = op_ret;
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->this->name);
- goto out;
- }
-
- if (local->fop == GF_FOP_SETXATTR) {
- STACK_WIND (frame, dht_setxattr_mds_cbk,
- local->mds_subvol,
- local->mds_subvol->fops->setxattr,
- &local->loc, local->xattr,
- local->flags, local->xattr_req);
- } else {
- STACK_WIND (frame, dht_setxattr_mds_cbk,
- local->mds_subvol,
- local->mds_subvol->fops->fsetxattr,
- local->fd, local->xattr,
- local->flags, local->xattr_req);
- }
- return 0;
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ local->op_ret = op_ret;
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->this->name);
+ goto out;
+ }
+
+ if (local->fop == GF_FOP_SETXATTR) {
+ STACK_WIND(frame, dht_setxattr_mds_cbk, local->mds_subvol,
+ local->mds_subvol->fops->setxattr, &local->loc, local->xattr,
+ local->flags, local->xattr_req);
+ }
+
+ if (local->fop == GF_FOP_FSETXATTR) {
+ STACK_WIND(frame, dht_setxattr_mds_cbk, local->mds_subvol,
+ local->mds_subvol->fops->fsetxattr, local->fd, local->xattr,
+ local->flags, local->xattr_req);
+ }
+
+ if (local->fop == GF_FOP_REMOVEXATTR) {
+ STACK_WIND(frame, dht_setxattr_mds_cbk, local->mds_subvol,
+ local->mds_subvol->fops->removexattr, &local->loc,
+ local->key, local->xattr_req);
+ }
+
+ if (local->fop == GF_FOP_FREMOVEXATTR) {
+ STACK_WIND(frame, dht_setxattr_mds_cbk, local->mds_subvol,
+ local->mds_subvol->fops->fremovexattr, local->fd, local->key,
+ local->xattr_req);
+ }
+
+ return 0;
out:
- if (local->fop == GF_FOP_SETXATTR)
- DHT_STACK_UNWIND (setxattr, frame, local->op_ret,
- local->op_errno, xdata);
- else
- DHT_STACK_UNWIND (fsetxattr, frame, local->op_ret,
- local->op_errno, xdata);
- return 0;
+ if (local->fop == GF_FOP_SETXATTR) {
+ DHT_STACK_UNWIND(setxattr, frame, local->op_ret, local->op_errno,
+ xdata);
+ /* 'local' itself may not be valid after this */
+ goto just_return;
+ }
+
+ if (local->fop == GF_FOP_FSETXATTR) {
+ DHT_STACK_UNWIND(fsetxattr, frame, local->op_ret, local->op_errno,
+ xdata);
+ /* 'local' itself may not be valid after this */
+ goto just_return;
+ }
+
+ if (local->fop == GF_FOP_REMOVEXATTR) {
+ DHT_STACK_UNWIND(removexattr, frame, local->op_ret, local->op_errno,
+ NULL);
+ /* 'local' itself may not be valid after this */
+ goto just_return;
+ }
+
+ if (local->fop == GF_FOP_FREMOVEXATTR) {
+ DHT_STACK_UNWIND(fremovexattr, frame, local->op_ret, local->op_errno,
+ NULL);
+ }
+
+just_return:
+ return 0;
}
static void
-fill_layout_info (dht_layout_t *layout, char *buf)
+fill_layout_info(dht_layout_t *layout, char *buf)
{
- int i = 0;
- char tmp_buf[128] = {0,};
+ int i = 0;
+ char tmp_buf[128] = {
+ 0,
+ };
- for (i = 0; i < layout->cnt; i++) {
- snprintf (tmp_buf, 128, "(%s %u %u)",
- layout->list[i].xlator->name,
- layout->list[i].start,
- layout->list[i].stop);
- if (i)
- strcat (buf, " ");
- strcat (buf, tmp_buf);
- }
+ for (i = 0; i < layout->cnt; i++) {
+ snprintf(tmp_buf, sizeof(tmp_buf), "(%s %u %u)",
+ layout->list[i].xlator->name, layout->list[i].start,
+ layout->list[i].stop);
+ if (i)
+ strcat(buf, " ");
+ strcat(buf, tmp_buf);
+ }
}
-void
-dht_fill_pathinfo_xattr (xlator_t *this, dht_local_t *local,
- char *xattr_buf, int32_t alloc_len,
- int flag, char *layout_buf)
-{
- if (flag && local->xattr_val)
- snprintf (xattr_buf, alloc_len,
- "((<"DHT_PATHINFO_HEADER"%s> %s) (%s-layout %s))",
- this->name, local->xattr_val, this->name,
- layout_buf);
- else if (local->xattr_val)
- snprintf (xattr_buf, alloc_len,
- "(<"DHT_PATHINFO_HEADER"%s> %s)",
- this->name, local->xattr_val);
- else if (flag)
- snprintf (xattr_buf, alloc_len, "(%s-layout %s)",
- this->name, layout_buf);
+static void
+dht_fill_pathinfo_xattr(xlator_t *this, dht_local_t *local, char *xattr_buf,
+ int32_t alloc_len, int flag, char *layout_buf)
+{
+ if (flag) {
+ if (local->xattr_val) {
+ snprintf(xattr_buf, alloc_len,
+ "((<" DHT_PATHINFO_HEADER "%s> %s) (%s-layout %s))",
+ this->name, local->xattr_val, this->name, layout_buf);
+ } else {
+ snprintf(xattr_buf, alloc_len, "(%s-layout %s)", this->name,
+ layout_buf);
+ }
+ } else if (local->xattr_val) {
+ snprintf(xattr_buf, alloc_len, "(<" DHT_PATHINFO_HEADER "%s> %s)",
+ this->name, local->xattr_val);
+ } else {
+ xattr_buf[0] = '\0';
+ }
}
-int
-dht_vgetxattr_alloc_and_fill (dht_local_t *local, dict_t *xattr, xlator_t *this,
- int op_errno)
+static int
+dht_vgetxattr_alloc_and_fill(dht_local_t *local, dict_t *xattr, xlator_t *this,
+ int op_errno)
{
- int ret = -1;
- char *value = NULL;
- int32_t plen = 0;
+ int ret = -1;
+ char *value = NULL;
- ret = dict_get_str (xattr, local->xsel, &value);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- DHT_MSG_GET_XATTR_FAILED,
- "Subvolume %s returned -1", this->name);
- local->op_ret = -1;
- local->op_errno = op_errno;
- goto out;
- }
+ ret = dict_get_str(xattr, local->xsel, &value);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, DHT_MSG_GET_XATTR_FAILED,
+ "Subvolume %s returned -1", this->name);
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ goto out;
+ }
- local->alloc_len += strlen(value);
+ local->alloc_len += strlen(value);
+ if (!local->xattr_val) {
+ local->alloc_len += (SLEN(DHT_PATHINFO_HEADER) + 10);
+ local->xattr_val = GF_MALLOC(local->alloc_len, gf_common_mt_char);
if (!local->xattr_val) {
- local->alloc_len += (strlen (DHT_PATHINFO_HEADER) + 10);
- local->xattr_val = GF_CALLOC (local->alloc_len, sizeof (char),
- gf_common_mt_char);
- if (!local->xattr_val) {
- ret = -1;
- goto out;
- }
+ ret = -1;
+ goto out;
}
+ local->xattr_val[0] = '\0';
+ }
- if (local->xattr_val) {
- plen = strlen (local->xattr_val);
- if (plen) {
- /* extra byte(s) for \0 to be safe */
- local->alloc_len += (plen + 2);
- local->xattr_val = GF_REALLOC (local->xattr_val,
- local->alloc_len);
- if (!local->xattr_val) {
- ret = -1;
- goto out;
- }
- }
-
- (void) strcat (local->xattr_val, value);
- (void) strcat (local->xattr_val, " ");
- local->op_ret = 0;
+ int plen = strlen(local->xattr_val);
+ if (plen) {
+ /* extra byte(s) for \0 to be safe */
+ local->alloc_len += (plen + 2);
+ local->xattr_val = GF_REALLOC(local->xattr_val, local->alloc_len);
+ if (!local->xattr_val) {
+ ret = -1;
+ goto out;
}
+ }
- ret = 0;
+ (void)strcat(local->xattr_val, value);
+ (void)strcat(local->xattr_val, " ");
+ local->op_ret = 0;
- out:
- return ret;
+ ret = 0;
+
+out:
+ return ret;
}
-int
-dht_vgetxattr_fill_and_set (dht_local_t *local, dict_t **dict, xlator_t *this,
- gf_boolean_t flag)
+static int
+dht_vgetxattr_fill_and_set(dht_local_t *local, dict_t **dict, xlator_t *this,
+ gf_boolean_t flag)
{
- int ret = -1;
- char *xattr_buf = NULL;
- char layout_buf[8192] = {0,};
+ int ret = -1;
+ char *xattr_buf = NULL;
+ char layout_buf[8192] = {
+ 0,
+ };
- if (flag)
- fill_layout_info (local->layout, layout_buf);
+ if (flag)
+ fill_layout_info(local->layout, layout_buf);
- *dict = dict_new ();
- if (!*dict)
- goto out;
-
- local->xattr_val[strlen (local->xattr_val) - 1] = '\0';
+ *dict = dict_new();
+ if (!*dict)
+ goto out;
- /* we would need max this many bytes to create xattr string
- * extra 40 bytes is just an estimated amount of additional
- * space required as we include translator name and some
- * spaces, brackets etc. when forming the pathinfo string.
- *
- * For node-uuid we just don't have all the pretty formatting,
- * but since this is a generic routine for pathinfo & node-uuid
- * we dont have conditional space allocation and try to be
- * generic
- */
- local->alloc_len += (2 * strlen (this->name))
- + strlen (layout_buf)
- + 40;
- xattr_buf = GF_CALLOC (local->alloc_len, sizeof (char),
- gf_common_mt_char);
- if (!xattr_buf)
- goto out;
+ local->xattr_val[strlen(local->xattr_val) - 1] = '\0';
+
+ /* we would need max this many bytes to create xattr string
+ * extra 40 bytes is just an estimated amount of additional
+ * space required as we include translator name and some
+ * spaces, brackets etc. when forming the pathinfo string.
+ *
+ * For node-uuid we just don't have all the pretty formatting,
+ * but since this is a generic routine for pathinfo & node-uuid
+ * we don't have conditional space allocation and try to be
+ * generic
+ */
+ local->alloc_len += (2 * strlen(this->name)) + strlen(layout_buf) + 40;
+ xattr_buf = GF_MALLOC(local->alloc_len, gf_common_mt_char);
+ if (!xattr_buf)
+ goto out;
- if (XATTR_IS_PATHINFO (local->xsel)) {
- (void) dht_fill_pathinfo_xattr (this, local, xattr_buf,
- local->alloc_len, flag,
- layout_buf);
- } else if ((XATTR_IS_NODE_UUID (local->xsel))
- || (XATTR_IS_NODE_UUID_LIST (local->xsel))) {
- (void) snprintf (xattr_buf, local->alloc_len, "%s",
- local->xattr_val);
- } else {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_GET_XATTR_FAILED,
- "Unknown local->xsel (%s)", local->xsel);
- GF_FREE (xattr_buf);
- goto out;
- }
+ if (XATTR_IS_PATHINFO(local->xsel)) {
+ (void)dht_fill_pathinfo_xattr(this, local, xattr_buf, local->alloc_len,
+ flag, layout_buf);
+ } else if ((XATTR_IS_NODE_UUID(local->xsel)) ||
+ (XATTR_IS_NODE_UUID_LIST(local->xsel))) {
+ (void)snprintf(xattr_buf, local->alloc_len, "%s", local->xattr_val);
+ } else {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_GET_XATTR_FAILED,
+ "Unknown local->xsel (%s)", local->xsel);
+ GF_FREE(xattr_buf);
+ goto out;
+ }
- ret = dict_set_dynstr (*dict, local->xsel, xattr_buf);
- if (ret)
- GF_FREE (xattr_buf);
- GF_FREE (local->xattr_val);
+ ret = dict_set_dynstr(*dict, local->xsel, xattr_buf);
+ if (ret)
+ GF_FREE(xattr_buf);
+ GF_FREE(local->xattr_val);
- out:
- return ret;
+out:
+ return ret;
}
+static int
+dht_find_local_subvol_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xattr,
+ dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ xlator_t *prev = NULL;
+ int this_call_cnt = 0;
+ int ret = 0;
+ char *uuid_str = NULL;
+ char *uuid_list = NULL;
+ char *next_uuid_str = NULL;
+ char *saveptr = NULL;
+ uuid_t node_uuid = {
+ 0,
+ };
+ char *uuid_list_copy = NULL;
+ int count = 0;
+ int i = 0;
+ int index = 0;
+ int found = 0;
+ nodeuuid_info_t *tmp_ptr = NULL;
+
+ VALIDATE_OR_GOTO(frame, out);
+ VALIDATE_OR_GOTO(frame->local, out);
+
+ local = frame->local;
+ prev = cookie;
+ conf = this->private;
+
+ VALIDATE_OR_GOTO(conf->defrag, out);
+
+ gf_msg_debug(this->name, 0, "subvol %s returned", prev->name);
+
+ LOCK(&frame->lock);
+ {
+ this_call_cnt = --local->call_cnt;
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ UNLOCK(&frame->lock);
+ if (op_errno == ENODATA)
+ gf_msg_debug(this->name, 0, "failed to get node-uuid");
+ else
+ gf_msg(this->name, GF_LOG_ERROR, op_errno,
+ DHT_MSG_GET_XATTR_FAILED, "failed to get node-uuid");
+ goto post_unlock;
+ }
-int
-dht_find_local_subvol_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr,
- dict_t *xdata)
-{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- xlator_t *prev = NULL;
- int this_call_cnt = 0;
- int ret = 0;
- char *uuid_str = NULL;
- char *uuid_list = NULL;
- char *next_uuid_str = NULL;
- char *saveptr = NULL;
- uuid_t node_uuid = {0,};
- char *uuid_list_copy = NULL;
- int count = 0;
- int i = 0;
- int index = 0;
- int found = 0;
- nodeuuid_info_t *tmp_ptr = NULL;
-
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (frame->local, out);
-
- local = frame->local;
- prev = cookie;
- conf = this->private;
-
- VALIDATE_OR_GOTO (conf->defrag, out);
-
- gf_msg_debug (this->name, 0, "subvol %s returned", prev->name);
-
- LOCK (&frame->lock);
- {
- this_call_cnt = --local->call_cnt;
- if (op_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- DHT_MSG_GET_XATTR_FAILED,
- "getxattr err for dir");
- local->op_ret = -1;
- local->op_errno = op_errno;
- goto unlock;
- }
+ ret = dict_get_str(xattr, local->xsel, &uuid_list);
- ret = dict_get_str (xattr, local->xsel, &uuid_list);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_GET_FAILED,
+ "Failed to get %s", local->xsel);
+ local->op_ret = -1;
+ local->op_errno = EINVAL;
+ goto unlock;
+ }
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_GET_FAILED,
- "Failed to get %s", local->xsel);
- local->op_ret = -1;
- local->op_errno = EINVAL;
- goto unlock;
- }
+ /* As DHT will not know details of its child xlators
+ * we need to parse this twice to get the count first
+ * and allocate memory later.
+ */
+ count = 0;
+ index = conf->local_subvols_cnt;
- /* As DHT will not know details of its child xlators
- * we need to parse this twice to get the count first
- * and allocate memory later.
- */
- count = 0;
- index = conf->local_subvols_cnt;
-
- uuid_list_copy = gf_strdup (uuid_list);
-
- for (uuid_str = strtok_r (uuid_list, " ", &saveptr);
- uuid_str;
- uuid_str = next_uuid_str) {
-
- next_uuid_str = strtok_r (NULL, " ", &saveptr);
- if (gf_uuid_parse (uuid_str, node_uuid)) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_UUID_PARSE_ERROR,
- "Failed to parse uuid"
- " for %s", prev->name);
- local->op_ret = -1;
- local->op_errno = EINVAL;
- goto unlock;
- }
+ uuid_list_copy = gf_strdup(uuid_list);
+ if (!uuid_list_copy)
+ goto unlock;
- count++;
- if (gf_uuid_compare (node_uuid, conf->defrag->node_uuid)) {
- gf_msg_debug (this->name, 0, "subvol %s does not"
- "belong to this node",
- prev->name);
- } else {
-
- /* handle multiple bricks of the same replica
- * on the same node */
- if (found)
- continue;
- conf->local_subvols[(conf->local_subvols_cnt)++]
- = prev;
- found = 1;
- gf_msg_debug (this->name, 0, "subvol %s belongs to"
- " this node", prev->name);
- }
- }
+ for (uuid_str = strtok_r(uuid_list, " ", &saveptr); uuid_str;
+ uuid_str = next_uuid_str) {
+ next_uuid_str = strtok_r(NULL, " ", &saveptr);
+ if (gf_uuid_parse(uuid_str, node_uuid)) {
+ local->op_ret = -1;
+ local->op_errno = EINVAL;
+ UNLOCK(&frame->lock);
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_UUID_PARSE_ERROR,
+ "Failed to parse uuid for %s", prev->name);
+ goto post_unlock;
+ }
- if (!found) {
- local->op_ret = 0;
- goto unlock;
- }
+ count++;
+ if (gf_uuid_compare(node_uuid, conf->defrag->node_uuid)) {
+ gf_msg_debug(this->name, 0,
+ "subvol %s does not"
+ "belong to this node",
+ prev->name);
+ } else {
+ /* handle multiple bricks of the same replica
+ * on the same node */
+ if (found)
+ continue;
+ conf->local_subvols[(conf->local_subvols_cnt)++] = prev;
+ found = 1;
+ gf_msg_debug(this->name, 0,
+ "subvol %s belongs to"
+ " this node",
+ prev->name);
+ }
+ }
- conf->local_nodeuuids[index].count = count;
- conf->local_nodeuuids[index].elements
- = GF_CALLOC (count, sizeof (nodeuuid_info_t), 1);
+ if (!found) {
+ local->op_ret = 0;
+ goto unlock;
+ }
- /* The node-uuids are guaranteed to be returned in the same
- * order as the bricks
- * A null node-uuid is returned for a brick that is down.
- */
+ conf->local_nodeuuids[index].count = count;
+ conf->local_nodeuuids[index].elements = GF_CALLOC(
+ count, sizeof(nodeuuid_info_t), 1);
- saveptr = NULL;
- i = 0;
+ /* The node-uuids are guaranteed to be returned in the same
+ * order as the bricks
+ * A null node-uuid is returned for a brick that is down.
+ */
- for (uuid_str = strtok_r (uuid_list_copy, " ", &saveptr);
- uuid_str;
- uuid_str = next_uuid_str) {
+ saveptr = NULL;
+ i = 0;
- next_uuid_str = strtok_r (NULL, " ", &saveptr);
- tmp_ptr = &(conf->local_nodeuuids[index].elements[i]);
- gf_uuid_parse (uuid_str, tmp_ptr->uuid);
+ for (uuid_str = strtok_r(uuid_list_copy, " ", &saveptr); uuid_str;
+ uuid_str = next_uuid_str) {
+ next_uuid_str = strtok_r(NULL, " ", &saveptr);
+ tmp_ptr = &(conf->local_nodeuuids[index].elements[i]);
+ gf_uuid_parse(uuid_str, tmp_ptr->uuid);
- if (!gf_uuid_compare (tmp_ptr->uuid,
- conf->defrag->node_uuid)) {
- tmp_ptr->info = REBAL_NODEUUID_MINE;
- }
- i++;
- tmp_ptr = NULL;
- }
+ if (!gf_uuid_compare(tmp_ptr->uuid, conf->defrag->node_uuid)) {
+ tmp_ptr->info = REBAL_NODEUUID_MINE;
+ }
+ i++;
+ tmp_ptr = NULL;
}
+ }
- local->op_ret = 0;
- unlock:
- UNLOCK (&frame->lock);
-
- if (!is_last_call (this_call_cnt))
- goto out;
+ local->op_ret = 0;
+unlock:
+ UNLOCK(&frame->lock);
+post_unlock:
+ if (!is_last_call(this_call_cnt))
+ goto out;
- if (local->op_ret == -1) {
- goto unwind;
- }
+ if (local->op_ret == -1) {
+ goto unwind;
+ }
- DHT_STACK_UNWIND (getxattr, frame, 0, 0, xattr, xdata);
- goto out;
+ DHT_STACK_UNWIND(getxattr, frame, 0, 0, xattr, xdata);
+ goto out;
- unwind:
+unwind:
- GF_FREE (conf->local_nodeuuids[index].elements);
- conf->local_nodeuuids[index].elements = NULL;
+ GF_FREE(conf->local_nodeuuids[index].elements);
+ conf->local_nodeuuids[index].elements = NULL;
- DHT_STACK_UNWIND (getxattr, frame, -1, local->op_errno, NULL, xdata);
- out:
- GF_FREE (uuid_list_copy);
- return 0;
+ DHT_STACK_UNWIND(getxattr, frame, -1, local->op_errno, NULL, xdata);
+out:
+ GF_FREE(uuid_list_copy);
+ return 0;
}
-int
-dht_vgetxattr_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
+static int
+dht_vgetxattr_dir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
{
- int ret = 0;
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- dict_t *dict = NULL;
+ int ret = 0;
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ dict_t *dict = NULL;
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (frame->local, out);
+ VALIDATE_OR_GOTO(frame, out);
+ VALIDATE_OR_GOTO(frame->local, out);
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- this_call_cnt = --local->call_cnt;
- if (op_ret < 0) {
- if (op_errno != ENOTCONN) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- DHT_MSG_GET_XATTR_FAILED,
- "getxattr err for dir");
- local->op_ret = -1;
- local->op_errno = op_errno;
- }
+ local = frame->local;
- goto unlock;
- }
+ LOCK(&frame->lock);
+ {
+ this_call_cnt = --local->call_cnt;
+ if (op_ret < 0) {
+ if (op_errno != ENOTCONN) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ UNLOCK(&frame->lock);
+ gf_msg(this->name, GF_LOG_ERROR, op_errno,
+ DHT_MSG_GET_XATTR_FAILED, "getxattr err for dir");
+ goto post_unlock;
+ }
- ret = dht_vgetxattr_alloc_and_fill (local, xattr, this,
- op_errno);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- DHT_MSG_DICT_SET_FAILED,
- "alloc or fill failure");
+ goto unlock;
}
- unlock:
- UNLOCK (&frame->lock);
- if (!is_last_call (this_call_cnt))
- goto out;
+ ret = dht_vgetxattr_alloc_and_fill(local, xattr, this, op_errno);
+ if (ret) {
+ UNLOCK(&frame->lock);
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, DHT_MSG_DICT_SET_FAILED,
+ "alloc or fill failure");
+ goto post_unlock;
+ }
+ }
+unlock:
+ UNLOCK(&frame->lock);
+post_unlock:
+ if (!is_last_call(this_call_cnt))
+ goto out;
- /* -- last call: do patch ups -- */
+ /* -- last call: do patch ups -- */
- if (local->op_ret == -1) {
- goto unwind;
- }
+ if (local->op_ret == -1) {
+ goto unwind;
+ }
- ret = dht_vgetxattr_fill_and_set (local, &dict, this, _gf_true);
- if (ret)
- goto unwind;
+ ret = dht_vgetxattr_fill_and_set(local, &dict, this, _gf_true);
+ if (ret)
+ goto unwind;
- DHT_STACK_UNWIND (getxattr, frame, 0, 0, dict, xdata);
- goto cleanup;
+ DHT_STACK_UNWIND(getxattr, frame, 0, 0, dict, xdata);
+ goto cleanup;
- unwind:
- DHT_STACK_UNWIND (getxattr, frame, -1, local->op_errno, NULL, NULL);
- cleanup:
- if (dict)
- dict_unref (dict);
- out:
- return 0;
+unwind:
+ DHT_STACK_UNWIND(getxattr, frame, -1, local->op_errno, NULL, NULL);
+cleanup:
+ if (dict)
+ dict_unref(dict);
+out:
+ return 0;
}
-int
-dht_vgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
+static int
+dht_vgetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, dict_t *xattr, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int ret = 0;
- dict_t *dict = NULL;
- xlator_t *prev = NULL;
- gf_boolean_t flag = _gf_true;
+ dht_local_t *local = NULL;
+ int ret = 0;
+ dict_t *dict = NULL;
+ xlator_t *prev = NULL;
+ gf_boolean_t flag = _gf_true;
- local = frame->local;
- prev = cookie;
+ local = frame->local;
+ prev = cookie;
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = op_errno;
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- DHT_MSG_GET_XATTR_FAILED,
- "vgetxattr: Subvolume %s returned -1",
- prev->name);
- goto unwind;
- }
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, DHT_MSG_GET_XATTR_FAILED,
+ "vgetxattr: Subvolume %s returned -1", prev->name);
+ goto unwind;
+ }
- ret = dht_vgetxattr_alloc_and_fill (local, xattr, this,
- op_errno);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_NO_MEMORY,
- "Allocation or fill failure");
- goto unwind;
- }
+ ret = dht_vgetxattr_alloc_and_fill(local, xattr, this, op_errno);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_NO_MEMORY,
+ "Allocation or fill failure");
+ goto unwind;
+ }
- flag = (local->layout->cnt > 1) ? _gf_true : _gf_false;
+ flag = (local->layout->cnt > 1) ? _gf_true : _gf_false;
- ret = dht_vgetxattr_fill_and_set (local, &dict, this, flag);
- if (ret)
- goto unwind;
+ ret = dht_vgetxattr_fill_and_set(local, &dict, this, flag);
+ if (ret)
+ goto unwind;
- DHT_STACK_UNWIND (getxattr, frame, 0, 0, dict, xdata);
- goto cleanup;
+ DHT_STACK_UNWIND(getxattr, frame, 0, 0, dict, xdata);
+ goto cleanup;
- unwind:
- DHT_STACK_UNWIND (getxattr, frame, -1, local->op_errno,
- NULL, NULL);
- cleanup:
- if (dict)
- dict_unref (dict);
+unwind:
+ DHT_STACK_UNWIND(getxattr, frame, -1, local->op_errno, NULL, NULL);
+cleanup:
+ if (dict)
+ dict_unref(dict);
- return 0;
+ return 0;
}
-int
-dht_linkinfo_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr,
- dict_t *xdata)
+static int
+dht_linkinfo_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xattr,
+ dict_t *xdata)
{
- int ret = 0;
- char *value = NULL;
+ int ret = 0;
+ char *value = NULL;
- if (op_ret != -1) {
- ret = dict_get_str (xattr, GF_XATTR_PATHINFO_KEY, &value);
- if (!ret) {
- ret = dict_set_str (xattr, GF_XATTR_LINKINFO_KEY, value);
- if (!ret)
- gf_msg_trace (this->name, 0,
- "failed to set linkinfo");
- }
+ if (op_ret != -1) {
+ ret = dict_get_str(xattr, GF_XATTR_PATHINFO_KEY, &value);
+ if (!ret) {
+ ret = dict_set_str(xattr, GF_XATTR_LINKINFO_KEY, value);
+ if (!ret)
+ gf_msg_trace(this->name, 0, "failed to set linkinfo");
}
+ }
- DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr, xdata);
+ DHT_STACK_UNWIND(getxattr, frame, op_ret, op_errno, xattr, xdata);
- return 0;
+ return 0;
}
-
-int
-dht_mds_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
+static int
+dht_mds_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (frame->local, out);
- VALIDATE_OR_GOTO (this->private, out);
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(frame->local, err);
+ VALIDATE_OR_GOTO(this->private, err);
- conf = this->private;
- local = frame->local;
+ conf = this->private;
+ local = frame->local;
- if (!xattr || (op_ret == -1)) {
- local->op_ret = op_ret;
- goto out;
- }
- if (dict_get (xattr, conf->xattr_name)) {
- dict_del (xattr, conf->xattr_name);
- }
- local->op_ret = 0;
+ if (!xattr || (op_ret == -1)) {
+ local->op_ret = op_ret;
+ goto out;
+ }
+ dict_del(xattr, conf->xattr_name);
+ local->op_ret = 0;
- if (!local->xattr) {
- local->xattr = dict_copy_with_ref (xattr, NULL);
- }
+ if (!local->xattr) {
+ local->xattr = dict_copy_with_ref(xattr, NULL);
+ }
out:
- DHT_STACK_UNWIND (getxattr, frame, local->op_ret, op_errno,
- local->xattr, xdata);
- return 0;
+ DHT_STACK_UNWIND(getxattr, frame, local->op_ret, op_errno, local->xattr,
+ xdata);
+ return 0;
+err:
+ DHT_STACK_UNWIND(getxattr, frame, -1, EINVAL, NULL, NULL);
+ return 0;
}
-
int
-dht_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
+dht_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, dict_t *xattr, dict_t *xdata)
{
- int this_call_cnt = 0;
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
+ int this_call_cnt = 0;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int ret = 0;
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (frame->local, out);
- VALIDATE_OR_GOTO (this->private, out);
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(frame->local, err);
+ VALIDATE_OR_GOTO(this->private, err);
- conf = this->private;
- local = frame->local;
+ conf = this->private;
+ local = frame->local;
- LOCK (&frame->lock);
- {
- if (!xattr || (op_ret == -1)) {
- local->op_ret = op_ret;
- goto unlock;
- }
+ if (dht_check_remote_fd_failed_error(local, op_ret, op_errno)) {
+ ret = dht_check_and_open_fd_on_subvol(this, frame);
+ if (ret)
+ goto err;
+ return 0;
+ }
- if (dict_get (xattr, conf->xattr_name)) {
- dict_del (xattr, conf->xattr_name);
- }
+ LOCK(&frame->lock);
+ {
+ if (!xattr || (op_ret == -1)) {
+ local->op_ret = op_ret;
+ goto unlock;
+ }
- if (frame->root->pid >= 0) {
- GF_REMOVE_INTERNAL_XATTR
- ("trusted.glusterfs.quota*", xattr);
- GF_REMOVE_INTERNAL_XATTR("trusted.pgfid*", xattr);
- }
+ dict_del(xattr, conf->xattr_name);
+ dict_del(xattr, conf->mds_xattr_key);
- local->op_ret = 0;
-
- if (!local->xattr) {
- local->xattr = dict_copy_with_ref (xattr, NULL);
- } else {
- dht_aggregate_xattr (local->xattr, xattr);
- }
+ dict_del(xattr, conf->commithash_xattr_name);
+ if (frame->root->pid >= 0) {
+ GF_REMOVE_INTERNAL_XATTR("trusted.glusterfs.quota*", xattr);
+ GF_REMOVE_INTERNAL_XATTR("trusted.pgfid*", xattr);
}
-unlock:
- UNLOCK (&frame->lock);
- this_call_cnt = dht_frame_return (frame);
-out:
- if (is_last_call (this_call_cnt)) {
+ local->op_ret = 0;
- /* If we have a valid xattr received from any one of the
- * subvolume, let's return it */
- if (local->xattr) {
- local->op_ret = 0;
- }
+ if (!local->xattr) {
+ local->xattr = dict_copy_with_ref(xattr, NULL);
+ } else {
+ dht_aggregate_xattr(local->xattr, xattr);
+ }
- DHT_STACK_UNWIND (getxattr, frame, local->op_ret, op_errno,
- local->xattr, NULL);
+ if (!local->xdata) {
+ local->xdata = dict_ref(xdata);
+ } else if ((local->inode && IA_ISDIR(local->inode->ia_type)) ||
+ (local->fd && IA_ISDIR(local->fd->inode->ia_type))) {
+ dht_aggregate_xattr(local->xdata, xdata);
}
- return 0;
+ }
+unlock:
+ UNLOCK(&frame->lock);
+
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt)) {
+ /* If we have a valid xattr received from any one of the
+ * subvolume, let's return it */
+ if (local->xattr) {
+ local->op_ret = 0;
+ }
+
+ DHT_STACK_UNWIND(getxattr, frame, local->op_ret, op_errno, local->xattr,
+ local->xdata);
+ }
+ return 0;
+err:
+ DHT_STACK_UNWIND(getxattr, frame, -1, EINVAL, NULL, NULL);
+ return 0;
}
-int32_t
-dht_getxattr_unwind (call_frame_t *frame,
- int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
+static int32_t
+dht_getxattr_unwind(call_frame_t *frame, int op_ret, int op_errno, dict_t *dict,
+ dict_t *xdata)
{
- DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ DHT_STACK_UNWIND(getxattr, frame, op_ret, op_errno, dict, xdata);
+ return 0;
}
-
-int
-dht_getxattr_get_real_filename_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- dict_t *xattr, dict_t *xdata)
+static int
+dht_getxattr_get_real_filename_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int op_ret, int op_errno,
+ dict_t *xattr, dict_t *xdata)
{
- int this_call_cnt = 0;
- dht_local_t *local = NULL;
-
-
- local = frame->local;
+ int this_call_cnt = 0;
+ dht_local_t *local = NULL;
- LOCK (&frame->lock);
- {
- if (local->op_errno == ENODATA ||
- local->op_errno == EOPNOTSUPP) {
- /* Nothing to do here, we have already found
- * a subvol which does not have the get_real_filename
- * optimization. If condition is for simple logic.
- */
- goto unlock;
- }
+ local = frame->local;
- if (op_ret == -1) {
-
- if (op_errno == ENODATA || op_errno == EOPNOTSUPP) {
- /* This subvol does not have the optimization.
- * Better let the user know we don't support it.
- * Remove previous results if any.
- */
-
- if (local->xattr) {
- dict_unref (local->xattr);
- local->xattr = NULL;
- }
-
- if (local->xattr_req) {
- dict_unref (local->xattr_req);
- local->xattr_req = NULL;
- }
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_UPGRADE_BRICKS, "At least "
- "one of the bricks does not support "
- "this operation. Please upgrade all "
- "bricks.");
- goto unlock;
- }
+ LOCK(&frame->lock);
+ {
+ if (local->op_errno == EOPNOTSUPP) {
+ /* Nothing to do here, we have already found
+ * a subvol which does not have the get_real_filename
+ * optimization. If condition is for simple logic.
+ */
+ goto unlock;
+ }
- if (op_errno == ENOENT) {
- /* Do nothing, our defaults are set to this.
- */
- goto unlock;
- }
+ if (op_ret == -1) {
+ if (op_errno == EOPNOTSUPP) {
+ /* This subvol does not have the optimization.
+ * Better let the user know we don't support it.
+ * Remove previous results if any.
+ */
- /* This is a place holder for every other error
- * case. I am not sure of how to interpret
- * ENOTCONN etc. As of now, choosing to ignore
- * down subvol and return a good result(if any)
- * from other subvol.
- */
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_GET_XATTR_FAILED,
- "Failed to get real filename.");
- goto unlock;
+ if (local->xattr) {
+ dict_unref(local->xattr);
+ local->xattr = NULL;
+ }
+ if (local->xattr_req) {
+ dict_unref(local->xattr_req);
+ local->xattr_req = NULL;
}
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ UNLOCK(&frame->lock);
+ gf_msg(this->name, GF_LOG_WARNING, op_errno,
+ DHT_MSG_UPGRADE_BRICKS,
+ "At least "
+ "one of the bricks does not support "
+ "this operation. Please upgrade all "
+ "bricks.");
+ goto post_unlock;
+ }
- /* This subvol has the required file.
- * There could be other subvols which have returned
- * success already, choosing to return the latest good
- * result.
+ if (op_errno == ENOATTR) {
+ /* Do nothing, our defaults are set to this.
*/
- if (local->xattr)
- dict_unref (local->xattr);
- local->xattr = dict_ref (xattr);
+ goto unlock;
+ }
- if (local->xattr_req) {
- dict_unref (local->xattr_req);
- local->xattr_req = NULL;
- }
- if (xdata)
- local->xattr_req = dict_ref (xdata);
+ /* This is a place holder for every other error
+ * case. I am not sure of how to interpret
+ * ENOTCONN etc. As of now, choosing to ignore
+ * down subvol and return a good result(if any)
+ * from other subvol.
+ */
+ UNLOCK(&frame->lock);
+ gf_msg(this->name, GF_LOG_WARNING, op_errno,
+ DHT_MSG_GET_XATTR_FAILED, "Failed to get real filename.");
+ goto post_unlock;
+ }
+
+ /* This subvol has the required file.
+ * There could be other subvols which have returned
+ * success already, choosing to return the latest good
+ * result.
+ */
+ if (local->xattr)
+ dict_unref(local->xattr);
+ local->xattr = dict_ref(xattr);
- local->op_ret = op_ret;
- local->op_errno = 0;
- gf_msg_debug (this->name, 0, "Found a matching "
- "file.");
+ if (local->xattr_req) {
+ dict_unref(local->xattr_req);
+ local->xattr_req = NULL;
}
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ local->op_ret = op_ret;
+ local->op_errno = 0;
+ UNLOCK(&frame->lock);
+ gf_msg_debug(this->name, 0, "Found a matching file.");
+ goto post_unlock;
+ }
unlock:
- UNLOCK (&frame->lock);
+ UNLOCK(&frame->lock);
+post_unlock:
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt)) {
+ DHT_STACK_UNWIND(getxattr, frame, local->op_ret, local->op_errno,
+ local->xattr, local->xattr_req);
+ }
+ return 0;
+}
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- DHT_STACK_UNWIND (getxattr, frame, local->op_ret,
- local->op_errno, local->xattr,
- local->xattr_req);
- }
+static int
+dht_getxattr_get_real_filename(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *key, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ int i = 0;
+ dht_layout_t *layout = NULL;
+ int cnt = 0;
+ xlator_t *subvol = NULL;
- return 0;
-}
+ local = frame->local;
+ layout = local->layout;
+ cnt = local->call_cnt = layout->cnt;
-int
-dht_getxattr_get_real_filename (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *key, dict_t *xdata)
-{
- dht_local_t *local = NULL;
- int i = 0;
- dht_layout_t *layout = NULL;
- int cnt = 0;
- xlator_t *subvol = NULL;
+ local->op_ret = -1;
+ local->op_errno = ENOATTR;
+ for (i = 0; i < cnt; i++) {
+ subvol = layout->list[i].xlator;
+ STACK_WIND(frame, dht_getxattr_get_real_filename_cbk, subvol,
+ subvol->fops->getxattr, loc, key, xdata);
+ }
- local = frame->local;
- layout = local->layout;
+ return 0;
+}
- cnt = local->call_cnt = layout->cnt;
+static int
+dht_marker_populate_args(call_frame_t *frame, int type, int *gauge,
+ xlator_t **subvols)
+{
+ dht_local_t *local = NULL;
+ int i = 0;
+ dht_layout_t *layout = NULL;
- local->op_ret = -1;
- local->op_errno = ENOENT;
+ local = frame->local;
+ layout = local->layout;
- for (i = 0; i < cnt; i++) {
- subvol = layout->list[i].xlator;
- STACK_WIND (frame, dht_getxattr_get_real_filename_cbk,
- subvol, subvol->fops->getxattr,
- loc, key, xdata);
- }
+ for (i = 0; i < layout->cnt; i++)
+ subvols[i] = layout->list[i].xlator;
- return 0;
+ return layout->cnt;
}
-int
-dht_marker_populate_args (call_frame_t *frame, int type, int *gauge,
- xlator_t **subvols)
+static int
+dht_is_debug_xattr_key(const char **array, char *key)
{
- dht_local_t *local = NULL;
- int i = 0;
- dht_layout_t *layout = NULL;
-
- local = frame->local;
- layout = local->layout;
+ int i = 0;
- for (i = 0; i < layout->cnt; i++)
- subvols[i] = layout->list[i].xlator;
+ for (i = 0; array[i]; i++) {
+ if (fnmatch(array[i], key, FNM_NOESCAPE) == 0)
+ return i;
+ }
- return layout->cnt;
+ return -1;
}
-int
-dht_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *key, dict_t *xdata)
-#define DHT_IS_DIR(layout) (layout->cnt > 1)
-{
-
- xlator_t *subvol = NULL;
- xlator_t *hashed_subvol = NULL;
- xlator_t *mds_subvol = NULL;
- xlator_t *cached_subvol = NULL;
- dht_conf_t *conf = NULL;
- dht_local_t *local = NULL;
- dht_layout_t *layout = NULL;
- int op_errno = -1;
- int i = 0;
- int cnt = 0;
- char *node_uuid_key = NULL;
- int ret = -1;
-
- GF_CHECK_XATTR_KEY_AND_GOTO (key, IO_THREADS_QUEUE_SIZE_KEY,
- op_errno, err);
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (this->private, err);
-
- conf = this->private;
-
- local = dht_local_init (frame, loc, NULL, GF_FOP_GETXATTR);
- if (!local) {
- op_errno = ENOMEM;
+/* Note we already have frame->local initialised here*/
- goto err;
- }
+static int
+dht_handle_debug_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *key)
+{
+ dht_local_t *local = NULL;
+ int ret = -1;
+ int op_errno = ENODATA;
+ char *value = NULL;
+ loc_t file_loc = {0};
+ const char *name = NULL;
- layout = local->layout;
- if (!layout) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LAYOUT_NULL,
- "Layout is NULL");
- op_errno = ENOENT;
- goto err;
- }
+ local = frame->local;
- if (key) {
- local->key = gf_strdup (key);
- if (!local->key) {
- op_errno = ENOMEM;
- goto err;
- }
+ if (dht_is_debug_xattr_key(dht_dbg_vxattrs, (char *)key) == -1) {
+ goto out;
+ }
+
+ local->xattr = dict_new();
+ if (!local->xattr) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ if (strncmp(key, DHT_DBG_HASHED_SUBVOL_KEY,
+ SLEN(DHT_DBG_HASHED_SUBVOL_KEY)) == 0) {
+ name = key + strlen(DHT_DBG_HASHED_SUBVOL_KEY);
+ if (strlen(name) == 0) {
+ op_errno = EINVAL;
+ goto out;
}
- if (key &&
- (strncmp (key, conf->mds_xattr_key, strlen(key)) == 0)) {
- op_errno = ENOTSUP;
- goto err;
+ ret = dht_build_child_loc(this, &file_loc, loc, (char *)name);
+ if (ret) {
+ op_errno = ENOMEM;
+ goto out;
}
- if (key &&
- (strncmp (key, GF_XATTR_GET_REAL_FILENAME_KEY,
- strlen (GF_XATTR_GET_REAL_FILENAME_KEY)) == 0)
- && DHT_IS_DIR(layout)) {
- dht_getxattr_get_real_filename (frame, this, loc, key, xdata);
- return 0;
+ local->hashed_subvol = dht_subvol_get_hashed(this, &file_loc);
+ if (local->hashed_subvol == NULL) {
+ op_errno = ENODATA;
+ goto out;
}
- if (key && DHT_IS_DIR(layout) &&
- (!strcmp (key, GF_REBAL_FIND_LOCAL_SUBVOL))) {
- ret = gf_asprintf
- (&node_uuid_key, "%s", GF_XATTR_LIST_NODE_UUIDS_KEY);
- if (ret == -1 || !node_uuid_key) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_NO_MEMORY,
- "Failed to copy key");
- op_errno = ENOMEM;
- goto err;
- }
- (void) strncpy (local->xsel, node_uuid_key, 256);
- cnt = local->call_cnt = conf->subvolume_cnt;
- for (i = 0; i < cnt; i++) {
- STACK_WIND_COOKIE (frame, dht_find_local_subvol_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->getxattr,
- loc, node_uuid_key, xdata);
- }
- if (node_uuid_key)
- GF_FREE (node_uuid_key);
- return 0;
+ value = gf_strdup(local->hashed_subvol->name);
+ if (!value) {
+ op_errno = ENOMEM;
+ goto out;
}
- if (key && DHT_IS_DIR(layout) &&
- (!strcmp (key, GF_REBAL_OLD_FIND_LOCAL_SUBVOL))) {
- ret = gf_asprintf
- (&node_uuid_key, "%s", GF_XATTR_NODE_UUID_KEY);
- if (ret == -1 || !node_uuid_key) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_NO_MEMORY,
- "Failed to copy key");
- op_errno = ENOMEM;
- goto err;
- }
- (void) strncpy (local->xsel, node_uuid_key, 256);
- cnt = local->call_cnt = conf->subvolume_cnt;
- for (i = 0; i < cnt; i++) {
- STACK_WIND_COOKIE (frame, dht_find_local_subvol_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->getxattr,
- loc, node_uuid_key, xdata);
- }
- if (node_uuid_key)
- GF_FREE (node_uuid_key);
- return 0;
+ ret = dict_set_dynstr(local->xattr, (char *)key, value);
+ if (ret < 0) {
+ op_errno = -ret;
+ ret = -1;
+ goto out;
}
+ ret = 0;
+ goto out;
+ }
- /* for file use cached subvolume (obviously!): see if {}
- * below
- * for directory:
- * wind to all subvolumes and exclude subvolumes which
- * return ENOTCONN (in callback)
- *
- * NOTE: Don't trust inode here, as that may not be valid
- * (until inode_link() happens)
- */
+out:
+ loc_wipe(&file_loc);
+ DHT_STACK_UNWIND(getxattr, frame, ret, op_errno, local->xattr, NULL);
+ return 0;
+}
- if (key && DHT_IS_DIR(layout) &&
- (XATTR_IS_PATHINFO (key)
- || (strcmp (key, GF_XATTR_NODE_UUID_KEY) == 0)
- || (strcmp (key, GF_XATTR_LIST_NODE_UUIDS_KEY) == 0))) {
- (void) strncpy (local->xsel, key, 256);
- cnt = local->call_cnt = layout->cnt;
- for (i = 0; i < cnt; i++) {
- subvol = layout->list[i].xlator;
- STACK_WIND (frame, dht_vgetxattr_dir_cbk,
- subvol, subvol->fops->getxattr,
- loc, key, xdata);
- }
- return 0;
- }
+/* Virtual Xattr which returns 1 if all subvols are up,
+ else returns 0. Geo-rep then uses this virtual xattr
+ after a fresh mount and starts the I/O.
+*/
- /* node-uuid or pathinfo for files */
- if (key && ((strcmp (key, GF_XATTR_NODE_UUID_KEY) == 0)
- || XATTR_IS_PATHINFO (key))) {
- cached_subvol = local->cached_subvol;
- (void) strncpy (local->xsel, key, 256);
+enum dht_vxattr_subvol {
+ DHT_VXATTR_SUBVOLS_UP = 1,
+ DHT_VXATTR_SUBVOLS_DOWN = 0,
+};
- local->call_cnt = 1;
- STACK_WIND_COOKIE (frame, dht_vgetxattr_cbk, cached_subvol,
- cached_subvol, cached_subvol->fops->getxattr,
- loc, key, xdata);
+int
+dht_vgetxattr_subvol_status(call_frame_t *frame, xlator_t *this,
+ const char *key)
+{
+ dht_local_t *local = NULL;
+ int ret = -1;
+ int op_errno = ENODATA;
+ int value = DHT_VXATTR_SUBVOLS_UP;
+ int i = 0;
+ dht_conf_t *conf = NULL;
- return 0;
+ conf = this->private;
+ local = frame->local;
+
+ if (!key) {
+ op_errno = EINVAL;
+ goto out;
+ }
+ local->xattr = dict_new();
+ if (!local->xattr) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (!conf->subvolume_status[i]) {
+ value = DHT_VXATTR_SUBVOLS_DOWN;
+ gf_msg_debug(this->name, 0, "subvol %s is down ",
+ conf->subvolumes[i]->name);
+ break;
+ }
+ }
+ ret = dict_set_int8(local->xattr, (char *)key, value);
+ if (ret < 0) {
+ op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+ ret = 0;
+
+out:
+ DHT_STACK_UNWIND(getxattr, frame, ret, op_errno, local->xattr, NULL);
+ return 0;
+}
+
+int
+dht_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, const char *key,
+ dict_t *xdata)
+#define DHT_IS_DIR(layout) (layout->cnt > 1)
+{
+ xlator_t *subvol = NULL;
+ xlator_t *hashed_subvol = NULL;
+ xlator_t *mds_subvol = NULL;
+ xlator_t *cached_subvol = NULL;
+ dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ dht_layout_t *layout = NULL;
+ int op_errno = -1;
+ int i = 0;
+ int cnt = 0;
+ char *node_uuid_key = NULL;
+ int ret = -1;
+
+ GF_CHECK_XATTR_KEY_AND_GOTO(key, IO_THREADS_QUEUE_SIZE_KEY, op_errno, err);
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
+ VALIDATE_OR_GOTO(this->private, err);
+
+ conf = this->private;
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_GETXATTR);
+ if (!local) {
+ op_errno = ENOMEM;
+
+ goto err;
+ }
+
+ layout = local->layout;
+ if (!layout) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_LAYOUT_NULL,
+ "Layout is NULL");
+ op_errno = ENOENT;
+ goto err;
+ }
+
+ /* skip over code which is irrelevant without a valid key */
+ if (!key)
+ goto no_key;
+
+ local->key = gf_strdup(key);
+ if (!local->key) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ if (strncmp(key, conf->mds_xattr_key, strlen(key)) == 0) {
+ op_errno = ENOTSUP;
+ goto err;
+ }
+
+ if (strncmp(key, DHT_SUBVOL_STATUS_KEY, SLEN(DHT_SUBVOL_STATUS_KEY)) == 0) {
+ dht_vgetxattr_subvol_status(frame, this, key);
+ return 0;
+ }
+
+ /* skip over code which is irrelevant if !DHT_IS_DIR(layout) */
+ if (!DHT_IS_DIR(layout))
+ goto no_dht_is_dir;
+
+ if ((strncmp(key, GF_XATTR_GET_REAL_FILENAME_KEY,
+ SLEN(GF_XATTR_GET_REAL_FILENAME_KEY)) == 0) &&
+ DHT_IS_DIR(layout)) {
+ dht_getxattr_get_real_filename(frame, this, loc, key, xdata);
+ return 0;
+ }
+
+ if (!strcmp(key, GF_REBAL_FIND_LOCAL_SUBVOL)) {
+ ret = gf_asprintf(&node_uuid_key, "%s", GF_XATTR_LIST_NODE_UUIDS_KEY);
+ if (ret == -1 || !node_uuid_key) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_NO_MEMORY,
+ "Failed to copy node uuid key");
+ op_errno = ENOMEM;
+ goto err;
+ }
+ (void)snprintf(local->xsel, sizeof(local->xsel), "%s", node_uuid_key);
+ cnt = local->call_cnt = conf->subvolume_cnt;
+ for (i = 0; i < cnt; i++) {
+ STACK_WIND_COOKIE(frame, dht_find_local_subvol_cbk,
+ conf->subvolumes[i], conf->subvolumes[i],
+ conf->subvolumes[i]->fops->getxattr, loc,
+ node_uuid_key, xdata);
}
+ if (node_uuid_key)
+ GF_FREE(node_uuid_key);
+ return 0;
+ }
- if (key && (strcmp (key, GF_XATTR_LINKINFO_KEY) == 0)) {
+ if (!strcmp(key, GF_REBAL_OLD_FIND_LOCAL_SUBVOL)) {
+ ret = gf_asprintf(&node_uuid_key, "%s", GF_XATTR_NODE_UUID_KEY);
+ if (ret == -1 || !node_uuid_key) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_NO_MEMORY,
+ "Failed to copy node uuid key");
+ op_errno = ENOMEM;
+ goto err;
+ }
+ (void)snprintf(local->xsel, sizeof(local->xsel), "%s", node_uuid_key);
+ cnt = local->call_cnt = conf->subvolume_cnt;
+ for (i = 0; i < cnt; i++) {
+ STACK_WIND_COOKIE(frame, dht_find_local_subvol_cbk,
+ conf->subvolumes[i], conf->subvolumes[i],
+ conf->subvolumes[i]->fops->getxattr, loc,
+ node_uuid_key, xdata);
+ }
+ if (node_uuid_key)
+ GF_FREE(node_uuid_key);
+ return 0;
+ }
+
+ /* for file use cached subvolume (obviously!): see if {}
+ * below
+ * for directory:
+ * wind to all subvolumes and exclude subvolumes which
+ * return ENOTCONN (in callback)
+ *
+ * NOTE: Don't trust inode here, as that may not be valid
+ * (until inode_link() happens)
+ */
+
+ if (XATTR_IS_PATHINFO(key) || (strcmp(key, GF_XATTR_NODE_UUID_KEY) == 0) ||
+ (strcmp(key, GF_XATTR_LIST_NODE_UUIDS_KEY) == 0)) {
+ (void)snprintf(local->xsel, sizeof(local->xsel), "%s", key);
+ cnt = local->call_cnt = layout->cnt;
+ for (i = 0; i < cnt; i++) {
+ subvol = layout->list[i].xlator;
+ STACK_WIND(frame, dht_vgetxattr_dir_cbk, subvol,
+ subvol->fops->getxattr, loc, key, xdata);
+ }
+ return 0;
+ }
- hashed_subvol = dht_subvol_get_hashed (this, loc);
- if (!hashed_subvol) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_HASHED_SUBVOL_GET_FAILED,
- "Failed to get hashed subvol for %s",
- loc->path);
- op_errno = EINVAL;
- goto err;
- }
+no_dht_is_dir:
+ /* node-uuid or pathinfo for files */
+ if (XATTR_IS_PATHINFO(key) || (strcmp(key, GF_XATTR_NODE_UUID_KEY) == 0)) {
+ cached_subvol = local->cached_subvol;
+ (void)snprintf(local->xsel, sizeof(local->xsel), "%s", key);
+ local->call_cnt = 1;
+ STACK_WIND_COOKIE(frame, dht_vgetxattr_cbk, cached_subvol,
+ cached_subvol, cached_subvol->fops->getxattr, loc,
+ key, xdata);
- cached_subvol = dht_subvol_get_cached (this, loc->inode);
- if (!cached_subvol) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_CACHED_SUBVOL_GET_FAILED,
- "Failed to get cached subvol for %s",
- loc->path);
- op_errno = EINVAL;
- goto err;
- }
+ return 0;
+ }
- if (hashed_subvol == cached_subvol) {
- op_errno = ENODATA;
- goto err;
- }
+ if (strcmp(key, GF_XATTR_LINKINFO_KEY) == 0) {
+ hashed_subvol = dht_subvol_get_hashed(this, loc);
+ if (!hashed_subvol) {
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ DHT_MSG_HASHED_SUBVOL_GET_FAILED,
+ "Failed to get hashed subvol for %s", loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
- STACK_WIND (frame, dht_linkinfo_getxattr_cbk, hashed_subvol,
- hashed_subvol->fops->getxattr, loc,
- GF_XATTR_PATHINFO_KEY, xdata);
- return 0;
+ cached_subvol = dht_subvol_get_cached(this, loc->inode);
+ if (!cached_subvol) {
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ DHT_MSG_CACHED_SUBVOL_GET_FAILED,
+ "Failed to get cached subvol for %s", loc->path);
+ op_errno = EINVAL;
+ goto err;
}
- if (cluster_handle_marker_getxattr (frame, loc, key, conf->vol_uuid,
- dht_getxattr_unwind,
- dht_marker_populate_args) == 0)
- return 0;
+ if (hashed_subvol == cached_subvol) {
+ op_errno = ENODATA;
+ goto err;
+ }
- if (DHT_IS_DIR(layout)) {
- local->call_cnt = conf->subvolume_cnt;
- cnt = conf->subvolume_cnt;
- ret = dht_inode_ctx_mdsvol_get (loc->inode, this, &mds_subvol);
- if (!mds_subvol) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_HASHED_SUBVOL_GET_FAILED,
- "Cannot determine MDS, fetching xattr %s randomly"
- " from a subvol for path %s ", key, loc->path);
- } else {
- /* TODO need to handle it, As of now we are
- choosing availability instead of chossing
- consistencty, in case of mds_subvol is
- down winding a getxattr call on other subvol
- and return xattr
- */
- local->mds_subvol = mds_subvol;
- for (i = 0; i < cnt; i++) {
- if (conf->subvolumes[i] == mds_subvol) {
- if (!conf->subvolume_status[i]) {
- gf_msg (this->name,
- GF_LOG_INFO, 0,
- DHT_MSG_HASHED_SUBVOL_DOWN,
- "MDS %s is down for path"
- " path %s so fetching xattr "
- "%s randomly from a subvol ",
- local->mds_subvol->name,
- loc->path, key);
- ret = 1;
- }
- }
- }
- }
+ STACK_WIND(frame, dht_linkinfo_getxattr_cbk, hashed_subvol,
+ hashed_subvol->fops->getxattr, loc, GF_XATTR_PATHINFO_KEY,
+ xdata);
+ return 0;
+ }
- if (!ret && key && local->mds_subvol && dht_match_xattr (key)) {
- STACK_WIND (frame, dht_mds_getxattr_cbk,
- local->mds_subvol,
- local->mds_subvol->fops->getxattr,
- loc, key, xdata);
+ if (dht_is_debug_xattr_key(dht_dbg_vxattrs, (char *)key) >= 0) {
+ dht_handle_debug_getxattr(frame, this, loc, key);
+ return 0;
+ }
- return 0;
- }
+no_key:
+ if (cluster_handle_marker_getxattr(frame, loc, key, conf->vol_uuid,
+ dht_getxattr_unwind,
+ dht_marker_populate_args) == 0)
+ return 0;
+
+ if (DHT_IS_DIR(layout)) {
+ local->call_cnt = conf->subvolume_cnt;
+ cnt = conf->subvolume_cnt;
+ ret = dht_inode_ctx_mdsvol_get(loc->inode, this, &mds_subvol);
+ if (!mds_subvol) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_HASHED_SUBVOL_GET_FAILED,
+ "Cannot determine MDS, fetching xattr %s randomly"
+ " from a subvol for path %s ",
+ key, loc->path);
} else {
- cnt = local->call_cnt = 1;
+ /* TODO need to handle it, As of now we are
+ choosing availability instead of chossing
+ consistencty, in case of mds_subvol is
+ down winding a getxattr call on other subvol
+ and return xattr
+ */
+ local->mds_subvol = mds_subvol;
+ for (i = 0; i < cnt; i++) {
+ if (conf->subvolumes[i] == mds_subvol) {
+ if (!conf->subvolume_status[i]) {
+ gf_msg(this->name, GF_LOG_INFO, 0,
+ DHT_MSG_HASHED_SUBVOL_DOWN,
+ "MDS %s is down for path"
+ " path %s so fetching xattr "
+ "%s randomly from a subvol ",
+ local->mds_subvol->name, loc->path, key);
+ ret = 1;
+ }
+ }
+ }
}
- for (i = 0; i < cnt; i++) {
- subvol = layout->list[i].xlator;
- STACK_WIND (frame, dht_getxattr_cbk,
- subvol, subvol->fops->getxattr,
- loc, key, xdata);
+ if (!ret && key && local->mds_subvol && dht_match_xattr(key)) {
+ STACK_WIND(frame, dht_mds_getxattr_cbk, local->mds_subvol,
+ local->mds_subvol->fops->getxattr, loc, key, xdata);
+
+ return 0;
}
- return 0;
+ } else {
+ cnt = local->call_cnt = 1;
+ }
+
+ for (i = 0; i < cnt; i++) {
+ subvol = layout->list[i].xlator;
+ STACK_WIND(frame, dht_getxattr_cbk, subvol, subvol->fops->getxattr, loc,
+ key, xdata);
+ }
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(getxattr, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
#undef DHT_IS_DIR
int
-dht_fgetxattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, const char *key, dict_t *xdata)
-{
- xlator_t *subvol = NULL;
- dht_local_t *local = NULL;
- dht_layout_t *layout = NULL;
- int op_errno = -1;
- int i = 0;
- int cnt = 0;
- xlator_t *mds_subvol = NULL;
- int ret = -1;
- dht_conf_t *conf = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (fd->inode, err);
- VALIDATE_OR_GOTO (this->private, err);
-
- conf = this->private;
-
- local = dht_local_init (frame, NULL, fd, GF_FOP_FGETXATTR);
- if (!local) {
- op_errno = ENOMEM;
-
- goto err;
+dht_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *key,
+ dict_t *xdata)
+{
+ xlator_t *subvol = NULL;
+ dht_local_t *local = NULL;
+ dht_layout_t *layout = NULL;
+ int op_errno = -1;
+ int i = 0;
+ int cnt = 0;
+ xlator_t *mds_subvol = NULL;
+ int ret = -1;
+ dht_conf_t *conf = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+ VALIDATE_OR_GOTO(fd->inode, err);
+ VALIDATE_OR_GOTO(this->private, err);
+
+ conf = this->private;
+
+ local = dht_local_init(frame, NULL, fd, GF_FOP_FGETXATTR);
+ if (!local) {
+ op_errno = ENOMEM;
+
+ goto err;
+ }
+
+ layout = local->layout;
+ if (!layout) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_LAYOUT_NULL,
+ "Layout is NULL");
+ op_errno = ENOENT;
+ goto err;
+ }
+
+ if (key) {
+ local->key = gf_strdup(key);
+ if (!local->key) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+ }
+
+ gf_uuid_unparse(fd->inode->gfid, gfid);
+
+ if ((fd->inode->ia_type == IA_IFDIR) && key &&
+ (strncmp(key, GF_XATTR_LOCKINFO_KEY, SLEN(GF_XATTR_LOCKINFO_KEY)) !=
+ 0)) {
+ local->call_cnt = conf->subvolume_cnt;
+ cnt = conf->subvolume_cnt;
+ ret = dht_inode_ctx_mdsvol_get(fd->inode, this, &mds_subvol);
+
+ if (!mds_subvol) {
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ DHT_MSG_HASHED_SUBVOL_GET_FAILED,
+ "cannot determine MDS, fetching xattr %s "
+ " randomly from a subvol for gfid %s ",
+ key, gfid);
+ } else {
+ /* TODO need to handle it, As of now we are
+ choosing availability instead of chossing
+ consistencty, in case of hashed_subvol is
+ down winding a getxattr call on other subvol
+ and return xattr
+ */
+ local->mds_subvol = mds_subvol;
+ for (i = 0; i < cnt; i++) {
+ if (conf->subvolumes[i] == mds_subvol) {
+ if (!conf->subvolume_status[i]) {
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ DHT_MSG_HASHED_SUBVOL_DOWN,
+ "MDS subvolume %s is down"
+ " for gfid %s so fetching xattr "
+ " %s randomly from a subvol ",
+ local->mds_subvol->name, gfid, key);
+ ret = 1;
+ }
+ }
+ }
}
- layout = local->layout;
- if (!layout) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LAYOUT_NULL,
- "Layout is NULL");
- op_errno = ENOENT;
- goto err;
- }
+ if (!ret && key && local->mds_subvol && dht_match_xattr(key)) {
+ STACK_WIND(frame, dht_mds_getxattr_cbk, local->mds_subvol,
+ local->mds_subvol->fops->fgetxattr, fd, key, NULL);
- if (key) {
- local->key = gf_strdup (key);
- if (!local->key) {
- op_errno = ENOMEM;
- goto err;
- }
+ return 0;
}
- if (fd->inode)
- gf_uuid_unparse(fd->inode->gfid, gfid);
+ } else {
+ cnt = local->call_cnt = 1;
+ }
- if ((fd->inode->ia_type == IA_IFDIR)
- && key
- && (strncmp (key, GF_XATTR_LOCKINFO_KEY,
- strlen (GF_XATTR_LOCKINFO_KEY)) != 0)) {
- local->call_cnt = conf->subvolume_cnt;
- cnt = conf->subvolume_cnt;
- ret = dht_inode_ctx_mdsvol_get (fd->inode, this, &mds_subvol);
+ for (i = 0; i < cnt; i++) {
+ subvol = layout->list[i].xlator;
+ STACK_WIND(frame, dht_getxattr_cbk, subvol, subvol->fops->fgetxattr, fd,
+ key, NULL);
+ }
+ return 0;
- if (!mds_subvol) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_HASHED_SUBVOL_GET_FAILED,
- "cannot determine MDS, fetching xattr %s "
- " randomly from a subvol for gfid %s ",
- key, gfid);
- } else {
- /* TODO need to handle it, As of now we are
- choosing availability instead of chossing
- consistencty, in case of hashed_subvol is
- down winding a getxattr call on other subvol
- and return xattr
- */
- local->mds_subvol = mds_subvol;
- for (i = 0; i < cnt; i++) {
- if (conf->subvolumes[i] == mds_subvol) {
- if (!conf->subvolume_status[i]) {
- gf_msg (this->name,
- GF_LOG_WARNING, 0,
- DHT_MSG_HASHED_SUBVOL_DOWN,
- "MDS subvolume %s is down"
- " for gfid %s so fetching xattr "
- " %s randomly from a subvol ",
- local->mds_subvol->name,
- gfid, key);
- ret = 1;
- }
- }
- }
- }
+err:
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(fgetxattr, frame, -1, op_errno, NULL, NULL);
- if (!ret && key && local->mds_subvol &&
- dht_match_xattr (key)) {
- STACK_WIND (frame, dht_mds_getxattr_cbk,
- local->mds_subvol,
- local->mds_subvol->fops->fgetxattr,
- fd, key, NULL);
+ return 0;
+}
- return 0;
- }
+static int
+dht_setxattr2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
+{
+ dht_local_t *local = NULL;
+ int op_errno = EINVAL;
- } else {
- cnt = local->call_cnt = 1;
- }
+ if (!frame || !frame->local)
+ goto err;
+ local = frame->local;
+ op_errno = local->op_errno;
- for (i = 0; i < cnt; i++) {
- subvol = layout->list[i].xlator;
- STACK_WIND (frame, dht_getxattr_cbk,
- subvol, subvol->fops->fgetxattr,
- fd, key, NULL);
- }
+ if (we_are_not_migrating(ret)) {
+ /* This dht xlator is not migrating the file. Unwind and
+ * pass on the original mode bits so the higher DHT layer
+ * can handle this.
+ */
+ DHT_STACK_UNWIND(setxattr, frame, local->op_ret, local->op_errno,
+ local->rebalance.xdata);
return 0;
+ }
-err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fgetxattr, frame, -1, op_errno, NULL, NULL);
+ if (subvol == NULL)
+ goto err;
- return 0;
+ local->call_cnt = 2; /* This is the second attempt */
+
+ if (local->fop == GF_FOP_SETXATTR) {
+ STACK_WIND_COOKIE(frame, dht_file_setxattr_cbk, subvol, subvol,
+ subvol->fops->setxattr, &local->loc,
+ local->rebalance.xattr, local->rebalance.flags,
+ local->xattr_req);
+ } else {
+ STACK_WIND_COOKIE(frame, dht_file_setxattr_cbk, subvol, subvol,
+ subvol->fops->fsetxattr, local->fd,
+ local->rebalance.xattr, local->rebalance.flags,
+ local->xattr_req);
+ }
+
+ return 0;
+
+err:
+ DHT_STACK_UNWIND(setxattr, frame, (local ? local->op_ret : -1), op_errno,
+ NULL);
+ return 0;
}
int
-dht_file_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+dht_file_setxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
{
- int ret = -1;
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- struct iatt *stbuf = NULL;
- inode_t *inode = NULL;
- xlator_t *subvol1 = NULL, *subvol2 = NULL;
+ int ret = -1;
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ struct iatt *stbuf = NULL;
+ inode_t *inode = NULL;
+ xlator_t *subvol1 = NULL, *subvol2 = NULL;
- local = frame->local;
- prev = cookie;
-
- local->op_errno = op_errno;
+ local = frame->local;
+ prev = cookie;
- if ((local->fop == GF_FOP_FSETXATTR) &&
- op_ret == -1 && (op_errno == EBADF) && !(local->fd_checked)) {
- ret = dht_check_and_open_fd_on_subvol (this, frame);
- if (ret)
- goto out;
- return 0;
- }
+ local->op_errno = op_errno;
- if ((op_ret == -1) && !dht_inode_missing (op_errno)) {
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1.",
- prev->name);
- goto out;
- }
+ if ((local->fop == GF_FOP_FSETXATTR) &&
+ dht_check_remote_fd_failed_error(local, op_ret, op_errno)) {
+ ret = dht_check_and_open_fd_on_subvol(this, frame);
+ if (ret)
+ goto out;
+ return 0;
+ }
- if (local->call_cnt != 1)
- goto out;
+ if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1.",
+ prev->name);
+ goto out;
+ }
- ret = dict_get_bin (xdata, DHT_IATT_IN_XDATA_KEY, (void **) &stbuf);
+ if (local->call_cnt != 1)
+ goto out;
- if ((!op_ret) && !stbuf) {
- goto out;
- }
+ ret = dict_get_bin(xdata, DHT_IATT_IN_XDATA_KEY, (void **)&stbuf);
- local->op_ret = op_ret;
- local->rebalance.target_op_fn = dht_setxattr2;
- if (xdata)
- local->rebalance.xdata = dict_ref (xdata);
+ if ((!op_ret) && !stbuf) {
+ goto out;
+ }
- /* Phase 2 of migration */
- if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (stbuf)) {
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret)
- return 0;
- }
+ local->op_ret = op_ret;
+ local->rebalance.target_op_fn = dht_setxattr2;
+ if (xdata)
+ local->rebalance.xdata = dict_ref(xdata);
- /* Phase 1 of migration */
- if (IS_DHT_MIGRATION_PHASE1 (stbuf)) {
- inode = (local->fd) ? local->fd->inode : local->loc.inode;
+ /* Phase 2 of migration */
+ if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2(stbuf)) {
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret)
+ return 0;
+ }
- ret = dht_inode_ctx_get_mig_info (this, inode,
- &subvol1, &subvol2);
- if (!dht_mig_info_is_invalid (local->cached_subvol,
- subvol1, subvol2)) {
- dht_setxattr2 (this, subvol2, frame, 0);
- return 0;
- }
+ /* Phase 1 of migration */
+ if (IS_DHT_MIGRATION_PHASE1(stbuf)) {
+ inode = (local->fd) ? local->fd->inode : local->loc.inode;
- ret = dht_rebalance_in_progress_check (this, frame);
- if (!ret)
- return 0;
+ ret = dht_inode_ctx_get_mig_info(this, inode, &subvol1, &subvol2);
+ if (!dht_mig_info_is_invalid(local->cached_subvol, subvol1, subvol2)) {
+ dht_setxattr2(this, subvol2, frame, 0);
+ return 0;
}
+ ret = dht_rebalance_in_progress_check(this, frame);
+ if (!ret)
+ return 0;
+ }
+
out:
- if (local->fop == GF_FOP_SETXATTR) {
- DHT_STACK_UNWIND (setxattr, frame, op_ret, op_errno, NULL);
- } else {
- DHT_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, NULL);
- }
+ if (local->fop == GF_FOP_SETXATTR) {
+ DHT_STACK_UNWIND(setxattr, frame, op_ret, op_errno, xdata);
+ } else {
+ DHT_STACK_UNWIND(fsetxattr, frame, op_ret, op_errno, xdata);
+ }
- return 0;
+ return 0;
}
/* Function is call by dict_foreach_fnmatch if key is match with
user.* and set boolean flag to true
*/
static int
-dht_is_user_xattr (dict_t *this, char *key, data_t *value, void *data)
+dht_is_user_xattr(dict_t *this, char *key, data_t *value, void *data)
{
- gf_boolean_t *user_xattr_found = data;
- *user_xattr_found = _gf_true;
- return 0;
+ gf_boolean_t *user_xattr_found = data;
+ *user_xattr_found = _gf_true;
+ return 0;
}
+/* Common code to wind a (f)(set|remove)xattr call to set xattr on directory
+ */
+static int
+dht_dir_common_set_remove_xattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ fd_t *fd, dict_t *xattr, int flags,
+ dict_t *xdata, int *op_errno)
+
+{
+ dict_t *xattrop = NULL;
+ int32_t subone[1] = {-1};
+ gf_boolean_t uxattr_key_found = _gf_false;
+ xlator_t *mds_subvol = NULL;
+ xlator_t *travvol = NULL;
+ dht_conf_t *conf = NULL;
+ int ret = -1;
+ int i = 0;
+ int call_cnt = 0;
+ dht_local_t *local = NULL;
+ char gfid_local[GF_UUID_BUF_SIZE] = {0};
+ char **xattrs_to_heal;
+
+ conf = this->private;
+ local = frame->local;
+ call_cnt = conf->subvolume_cnt;
+ local->flags = flags;
+ xattrs_to_heal = get_xattrs_to_heal();
+
+ if (!gf_uuid_is_null(local->gfid)) {
+ gf_uuid_unparse(local->gfid, gfid_local);
+ }
-/* Common code to wind a (f)setxattr call to set xattr on directory
-*/
-int
-dht_dir_common_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- fd_t *fd, dict_t *xattr, int flags, dict_t *xdata,
- int *op_errno)
-
-{
- dict_t *xattrop = NULL;
- int32_t subone[1] = {-1};
- gf_boolean_t uxattr_key_found = _gf_false;
- xlator_t *mds_subvol = NULL;
- xlator_t *travvol = NULL;
- dht_conf_t *conf = NULL;
- int ret = -1;
- int i = 0;
- int call_cnt = 0;
- dht_local_t *local = NULL;
- char gfid_local[GF_UUID_BUF_SIZE] = {0};
-
- conf = this->private;
- local = frame->local;
- call_cnt = conf->subvolume_cnt;
- local->flags = flags;
-
- if (local->gfid)
- gf_uuid_unparse(local->gfid, gfid_local);
-
+ if ((local->fop == GF_FOP_SETXATTR) || (local->fop == GF_FOP_FSETXATTR)) {
/* Check if any user xattr present in xattr
- */
- dict_foreach_fnmatch (xattr, "user*", dht_is_user_xattr,
- &uxattr_key_found);
+ */
+ dict_foreach_fnmatch(xattr, "user*", dht_is_user_xattr,
+ &uxattr_key_found);
/* Check if any custom key xattr present in dict xattr
and start index from 1 because user xattr already
checked in previous line
*/
for (i = 1; xattrs_to_heal[i]; i++)
- if (dict_get (xattr, xattrs_to_heal[i]))
- uxattr_key_found = _gf_true;
+ if (dict_get(xattr, xattrs_to_heal[i]))
+ uxattr_key_found = _gf_true;
+ }
- /* If there is no custom key xattr present or gfid is root
- or call_cnt is 1 then wind a (f)setxattr call on all subvols
- */
- if (!uxattr_key_found || __is_root_gfid (local->gfid) || call_cnt == 1) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- travvol = conf->subvolumes[i];
- if (fd) {
- STACK_WIND_COOKIE (frame, dht_err_cbk,
- travvol, travvol,
- travvol->fops->fsetxattr,
- fd, xattr, flags, xdata);
- } else {
- STACK_WIND_COOKIE (frame, dht_err_cbk,
- travvol, travvol,
- travvol->fops->setxattr,
- loc, xattr, flags, xdata);
- }
- }
-
- return 0;
- }
-
- /* Calculate hash subvol based on inode and parent inode
+ if ((local->fop == GF_FOP_REMOVEXATTR) ||
+ (local->fop == GF_FOP_FREMOVEXATTR)) {
+ /* Check if any custom key xattr present in local->key
*/
- if (fd) {
- ret = dht_inode_ctx_mdsvol_get (fd->inode, this, &mds_subvol);
- } else {
- ret = dht_inode_ctx_mdsvol_get (loc->inode, this, &mds_subvol);
- }
- if (ret || !mds_subvol) {
+ for (i = 0; xattrs_to_heal[i]; i++)
+ if (strstr(local->key, xattrs_to_heal[i]))
+ uxattr_key_found = _gf_true;
+ }
+
+ /* If there is no custom key xattr present or gfid is root
+ or call_cnt is 1 then wind a (f)setxattr call on all subvols
+ */
+ if (!uxattr_key_found || __is_root_gfid(local->gfid) || call_cnt == 1) {
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ travvol = conf->subvolumes[i];
+ if ((local->fop == GF_FOP_SETXATTR) ||
+ (local->fop == GF_FOP_FSETXATTR)) {
if (fd) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_HASHED_SUBVOL_GET_FAILED,
- "Failed to get mds subvol for fd %p"
- "gfid is %s ", fd, gfid_local);
+ STACK_WIND_COOKIE(frame, dht_err_cbk, travvol, travvol,
+ travvol->fops->fsetxattr, fd, xattr,
+ flags, xdata);
} else {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_HASHED_SUBVOL_GET_FAILED,
- "Failed to get mds subvol for path %s"
- "gfid is %s ", loc->path, gfid_local);
+ STACK_WIND_COOKIE(frame, dht_err_cbk, travvol, travvol,
+ travvol->fops->setxattr, loc, xattr,
+ flags, xdata);
}
- (*op_errno) = ENOENT;
- goto err;
- }
-
- local->mds_subvol = mds_subvol;
-
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->subvolumes[i] == mds_subvol) {
- if (!conf->subvolume_status[i]) {
- gf_msg (this->name, GF_LOG_WARNING,
- 0, DHT_MSG_HASHED_SUBVOL_DOWN,
- "MDS subvol is down for path "
- " %s gfid is %s Unable to set xattr " ,
- local->loc.path, gfid_local);
- (*op_errno) = ENOTCONN;
- goto err;
- }
- }
- }
+ }
- if (uxattr_key_found) {
- xattrop = dict_new ();
- if (!xattrop) {
- gf_msg (this->name, GF_LOG_ERROR, DHT_MSG_NO_MEMORY,
- 0, "dictionary creation failed for path %s "
- "for gfid is %s ", local->loc.path, gfid_local);
- (*op_errno) = ENOMEM;
- goto err;
- }
- local->xattr = dict_ref (xattr);
- /* Subtract current MDS xattr value to -1 , value of MDS
- xattr represents no. of times xattr modification failed
- on non MDS subvols.
- */
- ret = dht_dict_set_array (xattrop, conf->mds_xattr_key, subone, 1);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
- "dictionary set array failed for path %s "
- "for gfid is %s ", local->loc.path, gfid_local);
- if (xattrop)
- dict_unref (xattrop);
- (*op_errno) = ret;
- goto err;
- }
- /* Wind a xattrop call to use ref counting approach
- update mds xattr to -1 before update xattr on
- hashed subvol and update mds xattr to +1 after update
- xattr on all non hashed subvol
- */
+ if ((local->fop == GF_FOP_REMOVEXATTR) ||
+ (local->fop == GF_FOP_FREMOVEXATTR)) {
if (fd) {
- STACK_WIND (frame, dht_xattrop_mds_cbk,
- local->mds_subvol,
- local->mds_subvol->fops->fxattrop,
- fd, GF_XATTROP_ADD_ARRAY, xattrop, NULL);
+ STACK_WIND_COOKIE(frame, dht_err_cbk, travvol, travvol,
+ travvol->fops->fremovexattr, fd,
+ local->key, local->xattr_req);
} else {
- STACK_WIND (frame, dht_xattrop_mds_cbk,
- local->mds_subvol,
- local->mds_subvol->fops->xattrop,
- loc, GF_XATTROP_ADD_ARRAY,
- xattrop, NULL);
+ STACK_WIND_COOKIE(frame, dht_err_cbk, travvol, travvol,
+ travvol->fops->removexattr, loc,
+ local->key, local->xattr_req);
}
- if (xattrop)
- dict_unref (xattrop);
+ }
}
return 0;
-err:
- return -1;
-}
-
+ }
-int
-dht_fsetxattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *xattr, int flags, dict_t *xdata)
-{
- xlator_t *subvol = NULL;
- dht_local_t *local = NULL;
- int op_errno = EINVAL;
- dht_conf_t *conf = NULL;
- dht_layout_t *layout = NULL;
- int ret = -1;
- int call_cnt = 0;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (fd->inode, err);
- VALIDATE_OR_GOTO (this->private, err);
-
- conf = this->private;
-
- if (!conf->defrag)
- GF_IF_INTERNAL_XATTR_GOTO (conf->wild_xattr_name, xattr,
- op_errno, err);
-
- local = dht_local_init (frame, NULL, fd, GF_FOP_FSETXATTR);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for fd=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
-
- layout = local->layout;
- if (!layout) {
- gf_msg_debug (this->name, 0,
- "no layout for fd=%p", fd);
- op_errno = EINVAL;
+ /* Calculate hash subvol based on inode and parent inode
+ */
+ if (fd) {
+ ret = dht_inode_ctx_mdsvol_get(fd->inode, this, &mds_subvol);
+ } else {
+ ret = dht_inode_ctx_mdsvol_get(loc->inode, this, &mds_subvol);
+ }
+ if (ret || !mds_subvol) {
+ if (fd) {
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ DHT_MSG_HASHED_SUBVOL_GET_FAILED,
+ "Failed to get mds subvol for fd %p"
+ "gfid is %s ",
+ fd, gfid_local);
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ DHT_MSG_HASHED_SUBVOL_GET_FAILED,
+ "%s: Failed to get mds subvol. (gfid is %s)", loc->path,
+ gfid_local);
+ }
+ (*op_errno) = ENOENT;
+ goto err;
+ }
+
+ local->mds_subvol = mds_subvol;
+
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->subvolumes[i] == mds_subvol) {
+ if (!conf->subvolume_status[i]) {
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ DHT_MSG_HASHED_SUBVOL_DOWN,
+ "MDS subvol is down for path "
+ " %s gfid is %s Unable to set xattr ",
+ local->loc.path, gfid_local);
+ (*op_errno) = ENOTCONN;
goto err;
+ }
}
-
- local->xattr_req = xdata ? dict_ref (xdata) : dict_new ();
- local->call_cnt = call_cnt = layout->cnt;
-
- if (IA_ISDIR (fd->inode->ia_type)) {
- local->hashed_subvol = NULL;
- ret = dht_dir_common_setxattr (frame, this, NULL, fd,
- xattr, flags, xdata, &op_errno);
- if (ret)
- goto err;
+ }
+
+ if (uxattr_key_found) {
+ xattrop = dict_new();
+ if (!xattrop) {
+ gf_msg(this->name, GF_LOG_ERROR, DHT_MSG_NO_MEMORY, 0,
+ "dictionary creation failed for path %s "
+ "for gfid is %s ",
+ local->loc.path, gfid_local);
+ (*op_errno) = ENOMEM;
+ goto err;
+ }
+ local->xattr = dict_ref(xattr);
+ /* Subtract current MDS xattr value to -1 , value of MDS
+ xattr represents no. of times xattr modification failed
+ on non MDS subvols.
+ */
+ ret = dht_dict_set_array(xattrop, conf->mds_xattr_key, subone, 1);
+ if (ret != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "dictionary set array failed for path %s "
+ "for gfid is %s ",
+ local->loc.path, gfid_local);
+ if (xattrop)
+ dict_unref(xattrop);
+ (*op_errno) = ret;
+ goto err;
+ }
+ /* Wind a xattrop call to use ref counting approach
+ update mds xattr to -1 before update xattr on
+ hashed subvol and update mds xattr to +1 after update
+ xattr on all non hashed subvol
+ */
+ if (fd) {
+ STACK_WIND(frame, dht_xattrop_mds_cbk, local->mds_subvol,
+ local->mds_subvol->fops->fxattrop, fd,
+ GF_XATTROP_ADD_ARRAY, xattrop, NULL);
} else {
-
- local->call_cnt = 1;
- local->rebalance.xattr = dict_ref (xattr);
- local->rebalance.flags = flags;
-
- ret = dict_set_int8 (local->xattr_req, DHT_IATT_IN_XDATA_KEY, 1);
- if (ret) {
- gf_msg_debug (this->name, 0,
- "Failed to set dictionary key %s for fd=%p",
- DHT_IATT_IN_XDATA_KEY, fd);
- }
-
- STACK_WIND_COOKIE (frame, dht_file_setxattr_cbk, subvol,
- subvol, subvol->fops->fsetxattr, fd, xattr,
- flags, local->xattr_req);
+ STACK_WIND(frame, dht_xattrop_mds_cbk, local->mds_subvol,
+ local->mds_subvol->fops->xattrop, loc,
+ GF_XATTROP_ADD_ARRAY, xattrop, NULL);
}
- return 0;
+ if (xattrop)
+ dict_unref(xattrop);
+ }
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fsetxattr, frame, -1, op_errno, NULL);
-
- return 0;
+ return -1;
}
-
int
-dht_checking_pathinfo_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr,
- dict_t *xdata)
-{
- int i = -1;
- int ret = -1;
- char *value = NULL;
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- xlator_t *prev = NULL;
- int this_call_cnt = 0;
-
- local = frame->local;
- prev = cookie;
- conf = this->private;
-
- if (op_ret == -1)
- goto out;
-
-
- ret = dict_get_str (xattr, GF_XATTR_PATHINFO_KEY, &value);
+dht_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xattr,
+ int flags, dict_t *xdata)
+{
+ xlator_t *subvol = NULL;
+ dht_local_t *local = NULL;
+ int op_errno = EINVAL;
+ dht_conf_t *conf = NULL;
+ dht_layout_t *layout = NULL;
+ int ret = -1;
+ int call_cnt = 0;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+ VALIDATE_OR_GOTO(fd->inode, err);
+ VALIDATE_OR_GOTO(this->private, err);
+
+ conf = this->private;
+
+ if (!conf->defrag)
+ GF_IF_INTERNAL_XATTR_GOTO(conf->wild_xattr_name, xattr, op_errno, err);
+
+ local = dht_local_init(frame, NULL, fd, GF_FOP_FSETXATTR);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ layout = local->layout;
+ if (!layout) {
+ gf_msg_debug(this->name, 0, "no layout for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ local->xattr_req = xdata ? dict_ref(xdata) : dict_new();
+ local->call_cnt = call_cnt = layout->cnt;
+
+ if (IA_ISDIR(fd->inode->ia_type)) {
+ local->hashed_subvol = NULL;
+ ret = dht_dir_common_set_remove_xattr(frame, this, NULL, fd, xattr,
+ flags, xdata, &op_errno);
if (ret)
- goto out;
-
- if (!strcmp (value, local->key)) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->subvolumes[i] == prev)
- conf->decommissioned_bricks[i] = prev;
- }
- }
+ goto err;
+ } else {
+ local->call_cnt = 1;
+ local->rebalance.xattr = dict_ref(xattr);
+ local->rebalance.flags = flags;
-out:
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- DHT_STACK_UNWIND (setxattr, frame, local->op_ret, ENOTSUP, NULL);
+ ret = dict_set_int8(local->xattr_req, DHT_IATT_IN_XDATA_KEY, 1);
+ if (ret) {
+ gf_msg_debug(this->name, 0,
+ "Failed to set dictionary key %s for fd=%p",
+ DHT_IATT_IN_XDATA_KEY, fd);
}
- return 0;
-
-}
-
-
-int
-dht_setxattr2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
-{
- dht_local_t *local = NULL;
- int op_errno = EINVAL;
- if (!frame || !frame->local)
- goto err;
+ STACK_WIND_COOKIE(frame, dht_file_setxattr_cbk, subvol, subvol,
+ subvol->fops->fsetxattr, fd, xattr, flags,
+ local->xattr_req);
+ }
+ return 0;
- local = frame->local;
- op_errno = local->op_errno;
-
- if (we_are_not_migrating (ret)) {
- /* This dht xlator is not migrating the file. Unwind and
- * pass on the original mode bits so the higher DHT layer
- * can handle this.
- */
- DHT_STACK_UNWIND (setxattr, frame, local->op_ret,
- local->op_errno, local->rebalance.xdata);
- return 0;
- }
+err:
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(fsetxattr, frame, -1, op_errno, NULL);
- if (subvol == NULL)
- goto err;
+ return 0;
+}
+static int
+dht_checking_pathinfo_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xattr,
+ dict_t *xdata)
+{
+ int i = -1;
+ int ret = -1;
+ char *value = NULL;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ xlator_t *prev = NULL;
+ int this_call_cnt = 0;
+
+ local = frame->local;
+ prev = cookie;
+ conf = this->private;
+
+ if (op_ret == -1)
+ goto out;
- local->call_cnt = 2; /* This is the second attempt */
+ ret = dict_get_str(xattr, GF_XATTR_PATHINFO_KEY, &value);
+ if (ret)
+ goto out;
- if (local->fop == GF_FOP_SETXATTR) {
- STACK_WIND_COOKIE (frame, dht_file_setxattr_cbk, subvol,
- subvol, subvol->fops->setxattr, &local->loc,
- local->rebalance.xattr,
- local->rebalance.flags, local->xattr_req);
- } else {
- STACK_WIND_COOKIE (frame, dht_file_setxattr_cbk, subvol,
- subvol, subvol->fops->fsetxattr, local->fd,
- local->rebalance.xattr,
- local->rebalance.flags, local->xattr_req);
+ if (!strcmp(value, local->key)) {
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->subvolumes[i] == prev)
+ conf->decommissioned_bricks[i] = prev;
}
+ }
- return 0;
-
-err:
- DHT_STACK_UNWIND (setxattr, frame, (local ? local->op_ret : -1),
- op_errno, NULL);
- return 0;
+out:
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt)) {
+ DHT_STACK_UNWIND(setxattr, frame, local->op_ret, ENOTSUP, NULL);
+ }
+ return 0;
}
-int
-dht_nuke_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+static int
+dht_nuke_dir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, NULL);
- return 0;
+ STACK_UNWIND_STRICT(setxattr, frame, op_ret, op_errno, NULL);
+ return 0;
}
-int
-dht_nuke_dir (call_frame_t *frame, xlator_t *this, loc_t *loc, data_t *tmp)
+static int
+dht_nuke_dir(call_frame_t *frame, xlator_t *this, loc_t *loc, data_t *tmp)
{
- if (!IA_ISDIR(loc->inode->ia_type)) {
- DHT_STACK_UNWIND (setxattr, frame, -1, ENOTSUP, NULL);
- return 0;
- }
+ if (!IA_ISDIR(loc->inode->ia_type)) {
+ DHT_STACK_UNWIND(setxattr, frame, -1, ENOTSUP, NULL);
+ return 0;
+ }
- /* Setxattr didn't need the parent, but rmdir does. */
- loc->parent = inode_parent (loc->inode, NULL, NULL);
- if (!loc->parent) {
- DHT_STACK_UNWIND (setxattr, frame, -1, ENOENT, NULL);
- return 0;
- }
- gf_uuid_copy (loc->pargfid, loc->parent->gfid);
+ /* Setxattr didn't need the parent, but rmdir does. */
+ loc->parent = inode_parent(loc->inode, NULL, NULL);
+ if (!loc->parent) {
+ DHT_STACK_UNWIND(setxattr, frame, -1, ENOENT, NULL);
+ return 0;
+ }
+ gf_uuid_copy(loc->pargfid, loc->parent->gfid);
- if (!loc->name && loc->path) {
- loc->name = strrchr (loc->path, '/');
- if (loc->name) {
- ++(loc->name);
- }
+ if (!loc->name && loc->path) {
+ loc->name = strrchr(loc->path, '/');
+ if (loc->name) {
+ ++(loc->name);
}
+ }
- /*
- * We do this instead of calling dht_rmdir_do directly for two reasons.
- * The first is that we want to reuse all of the initialization that
- * dht_rmdir does, so if it ever changes we'll just follow along. The
- * second (i.e. why we don't use STACK_WIND_TAIL) is so that we don't
- * obscure the fact that we came in via this path instead of a genuine
- * rmdir. That makes debugging just a tiny bit easier.
- */
- STACK_WIND (frame, dht_nuke_dir_cbk, this, this->fops->rmdir,
- loc, 1, NULL);
+ /*
+ * We do this instead of calling dht_rmdir_do directly for two reasons.
+ * The first is that we want to reuse all of the initialization that
+ * dht_rmdir does, so if it ever changes we'll just follow along. The
+ * second (i.e. why we don't use STACK_WIND_TAIL) is so that we don't
+ * obscure the fact that we came in via this path instead of a genuine
+ * rmdir. That makes debugging just a tiny bit easier.
+ */
+ STACK_WIND(frame, dht_nuke_dir_cbk, this, this->fops->rmdir, loc, 1, NULL);
- return 0;
+ return 0;
}
-
int
-dht_setxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xattr, int flags, dict_t *xdata)
-{
- xlator_t *subvol = NULL;
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- dht_methods_t *methods = NULL;
- dht_layout_t *layout = NULL;
- int i = 0;
- int op_errno = EINVAL;
- int ret = -1;
- data_t *tmp = NULL;
- uint32_t dir_spread = 0;
- char value[4096] = {0,};
- gf_dht_migrate_data_type_t forced_rebalance = GF_DHT_MIGRATE_DATA;
- int call_cnt = 0;
- uint32_t new_hash = 0;
-
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- conf = this->private;
- GF_VALIDATE_OR_GOTO (this->name, conf, err);
-
- methods = &(conf->methods);
-
- /* Rebalance daemon is allowed to set internal keys */
- if (!conf->defrag)
- GF_IF_INTERNAL_XATTR_GOTO (conf->wild_xattr_name, xattr,
- op_errno, err);
-
- local = dht_local_init (frame, loc, NULL, GF_FOP_SETXATTR);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for path=%s",
- loc->path);
- op_errno = EINVAL;
- goto err;
- }
-
- layout = local->layout;
- if (!layout) {
- gf_msg_debug (this->name, 0,
- "no layout for path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
-
- local->call_cnt = call_cnt = layout->cnt;
- tmp = dict_get (xattr, conf->mds_xattr_key);
- if (tmp) {
- op_errno = ENOTSUP;
- goto err;
- }
-
- tmp = dict_get (xattr, GF_XATTR_FILE_MIGRATE_KEY);
- if (tmp) {
-
- if (IA_ISDIR (loc->inode->ia_type)) {
- op_errno = ENOTSUP;
- goto err;
- }
-
- /* TODO: need to interpret the 'value' for more meaning
- (ie, 'target' subvolume given there, etc) */
- memcpy (value, tmp->data, tmp->len);
- if (strcmp (value, "force") == 0)
- forced_rebalance =
- GF_DHT_MIGRATE_DATA_EVEN_IF_LINK_EXISTS;
-
- if (conf->decommission_in_progress)
- forced_rebalance = GF_DHT_MIGRATE_HARDLINK;
-
- if (!loc->path) {
- op_errno = EINVAL;
- goto err;
- }
-
- if (!local->loc.name)
- local->loc.name = strrchr (local->loc.path, '/')+1;
-
- if (!local->loc.parent)
- local->loc.parent =
- inode_parent(local->loc.inode, NULL, NULL);
-
- if ((!local->loc.name) || (!local->loc.parent)) {
- op_errno = EINVAL;
- goto err;
- }
-
- if (gf_uuid_is_null (local->loc.pargfid))
- gf_uuid_copy (local->loc.pargfid, local->loc.parent->gfid);
-
- methods->migration_get_dst_subvol(this, local);
-
- if (!local->rebalance.target_node) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_HASHED_SUBVOL_GET_FAILED,
- "Failed to get hashed subvol for %s",
- loc->path);
- op_errno = EINVAL;
- goto err;
- }
-
- local->rebalance.from_subvol = local->cached_subvol;
-
- if (local->rebalance.target_node == local->rebalance.from_subvol) {
- op_errno = EEXIST;
- goto err;
- }
- if (local->rebalance.target_node) {
- local->flags = forced_rebalance;
-
- /* Flag to suggest its a tiering migration
- * The reason for this dic key-value is that
- * promotions and demotions are multithreaded
- * so the original frame from gf_defrag_start()
- * is not carried. A new frame will be created when
- * we do syncop_setxattr(). This doesnot have the
- * frame->root->pid of the original frame. So we pass
- * this dic key-value when we do syncop_setxattr() to do
- * data migration and set the frame->root->pid to
- * GF_CLIENT_PID_TIER_DEFRAG in dht_setxattr() just before
- * calling dht_start_rebalance_task() */
- tmp = dict_get (xattr, TIERING_MIGRATION_KEY);
- if (tmp)
- frame->root->pid = GF_CLIENT_PID_TIER_DEFRAG;
- else
- frame->root->pid = GF_CLIENT_PID_DEFRAG;
-
- ret = dht_start_rebalance_task (this, frame);
- if (!ret)
- return 0;
-
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_REBALANCE_START_FAILED,
- "%s: failed to create a new rebalance synctask",
- loc->path);
- }
- op_errno = EINVAL;
- goto err;
-
+dht_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr,
+ int flags, dict_t *xdata)
+{
+ xlator_t *subvol = NULL;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ dht_methods_t *methods = NULL;
+ dht_layout_t *layout = NULL;
+ int i = 0;
+ int op_errno = EINVAL;
+ int ret = -1;
+ data_t *tmp = NULL;
+ uint32_t dir_spread = 0;
+ char value[4096] = {
+ 0,
+ };
+ gf_dht_migrate_data_type_t forced_rebalance = GF_DHT_MIGRATE_DATA;
+ int call_cnt = 0;
+ uint32_t new_hash = 0;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
+
+ conf = this->private;
+ GF_VALIDATE_OR_GOTO(this->name, conf, err);
+
+ methods = &(conf->methods);
+
+ /* Rebalance daemon is allowed to set internal keys */
+ if (!conf->defrag)
+ GF_IF_INTERNAL_XATTR_GOTO(conf->wild_xattr_name, xattr, op_errno, err);
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_SETXATTR);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for path=%s",
+ loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ layout = local->layout;
+ if (!layout) {
+ gf_msg_debug(this->name, 0, "no layout for path=%s", loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ local->call_cnt = call_cnt = layout->cnt;
+ tmp = dict_get(xattr, conf->mds_xattr_key);
+ if (tmp) {
+ op_errno = ENOTSUP;
+ goto err;
+ }
+
+ tmp = dict_get(xattr, GF_XATTR_FILE_MIGRATE_KEY);
+ if (tmp) {
+ if (IA_ISDIR(loc->inode->ia_type)) {
+ op_errno = ENOTSUP;
+ goto err;
+ }
+
+ /* TODO: need to interpret the 'value' for more meaning
+ (ie, 'target' subvolume given there, etc) */
+ memcpy(value, tmp->data, tmp->len);
+ if (strcmp(value, "force") == 0)
+ forced_rebalance = GF_DHT_MIGRATE_DATA_EVEN_IF_LINK_EXISTS;
+
+ if (conf->decommission_in_progress)
+ forced_rebalance = GF_DHT_MIGRATE_HARDLINK;
+
+ if (!loc->path) {
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ if (!local->loc.name)
+ local->loc.name = strrchr(local->loc.path, '/') + 1;
+
+ if (!local->loc.parent)
+ local->loc.parent = inode_parent(local->loc.inode, NULL, NULL);
+
+ if ((!local->loc.name) || (!local->loc.parent)) {
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ if (gf_uuid_is_null(local->loc.pargfid))
+ gf_uuid_copy(local->loc.pargfid, local->loc.parent->gfid);
+
+ methods->migration_get_dst_subvol(this, local);
+
+ if (!local->rebalance.target_node) {
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ DHT_MSG_HASHED_SUBVOL_GET_FAILED,
+ "Failed to get hashed subvol for %s", loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ local->rebalance.from_subvol = local->cached_subvol;
+
+ if (local->rebalance.target_node == local->rebalance.from_subvol) {
+ op_errno = EEXIST;
+ goto err;
}
+ if (local->rebalance.target_node) {
+ local->flags = forced_rebalance;
- tmp = dict_get (xattr, "decommission-brick");
- if (tmp) {
- /* This operation should happen only on '/' */
- if (!__is_root_gfid (loc->inode->gfid)) {
- op_errno = ENOTSUP;
- goto err;
- }
-
- memcpy (value, tmp->data, ((tmp->len < 4095) ? tmp->len : 4095));
- local->key = gf_strdup (value);
- local->call_cnt = conf->subvolume_cnt;
+ frame->root->pid = GF_CLIENT_PID_DEFRAG;
- for (i = 0 ; i < conf->subvolume_cnt; i++) {
- /* Get the pathinfo, and then compare */
- STACK_WIND_COOKIE (frame, dht_checking_pathinfo_cbk,
- conf->subvolumes[i], conf->subvolumes[i],
- conf->subvolumes[i]->fops->getxattr,
- loc, GF_XATTR_PATHINFO_KEY, NULL);
- }
+ ret = dht_start_rebalance_task(this, frame);
+ if (!ret)
return 0;
- }
-
- tmp = dict_get (xattr, GF_XATTR_FIX_LAYOUT_KEY);
- if (tmp) {
- ret = dict_get_uint32(xattr, "new-commit-hash", &new_hash);
- if (ret == 0) {
- gf_msg_debug (this->name, 0,
- "updating commit hash for %s from %u to %u",
- uuid_utoa(loc->gfid),
- layout->commit_hash, new_hash);
- layout->commit_hash = new_hash;
-
- ret = dht_update_commit_hash_for_layout (frame);
- if (ret) {
- op_errno = ENOTCONN;
- goto err;
- }
- return ret;
- }
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_FIX_LAYOUT_INFO,
- "fixing the layout of %s", loc->path);
-
- ret = dht_fix_directory_layout (frame,
- dht_fix_layout_setxattr_cbk,
- layout);
- if (ret) {
- op_errno = ENOTCONN;
- goto err;
- }
- return ret;
- }
-
- tmp = dict_get (xattr, "distribute.directory-spread-count");
- if (tmp) {
- /* Setxattr value is packed as 'binary', not string */
- memcpy (value, tmp->data, ((tmp->len < 4095)?tmp->len:4095));
- ret = gf_string2uint32 (value, &dir_spread);
- if (!ret && ((dir_spread <= conf->subvolume_cnt) &&
- (dir_spread > 0))) {
- layout->spread_cnt = dir_spread;
-
- ret = dht_fix_directory_layout (frame,
- dht_common_setxattr_cbk,
- layout);
- if (ret) {
- op_errno = ENOTCONN;
- goto err;
- }
- return ret;
- }
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_OPERATION_NOT_SUP,
- "wrong 'directory-spread-count' value (%s)", value);
- op_errno = ENOTSUP;
- goto err;
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_REBALANCE_START_FAILED,
+ "%s: failed to create a new rebalance synctask", loc->path);
}
+ op_errno = EINVAL;
+ goto err;
+ }
- tmp = dict_get (xattr, "glusterfs.dht.nuke");
- if (tmp) {
- return dht_nuke_dir (frame, this, loc, tmp);
+ tmp = dict_get(xattr, "decommission-brick");
+ if (tmp) {
+ /* This operation should happen only on '/' */
+ if (!__is_root_gfid(loc->inode->gfid)) {
+ op_errno = ENOTSUP;
+ goto err;
}
- local->xattr_req = xdata ? dict_ref (xdata) : dict_new ();
- if (IA_ISDIR (loc->inode->ia_type)) {
- local->hashed_subvol = NULL;
- ret = dht_dir_common_setxattr (frame, this, loc, NULL,
- xattr, flags, xdata, &op_errno);
- if (ret)
- goto err;
- } else {
-
- local->rebalance.xattr = dict_ref (xattr);
- local->rebalance.flags = flags;
- local->call_cnt = 1;
-
- ret = dict_set_int8 (local->xattr_req, DHT_IATT_IN_XDATA_KEY, 1);
+ memcpy(value, tmp->data, min(tmp->len, 4095));
+ local->key = gf_strdup(value);
+ local->call_cnt = conf->subvolume_cnt;
- STACK_WIND_COOKIE (frame, dht_file_setxattr_cbk, subvol,
- subvol, subvol->fops->setxattr, loc, xattr,
- flags, local->xattr_req);
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ /* Get the pathinfo, and then compare */
+ STACK_WIND_COOKIE(frame, dht_checking_pathinfo_cbk,
+ conf->subvolumes[i], conf->subvolumes[i],
+ conf->subvolumes[i]->fops->getxattr, loc,
+ GF_XATTR_PATHINFO_KEY, NULL);
}
-
return 0;
+ }
-err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
-
- return 0;
-}
-
-
-
-
-int
-dht_file_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
-{
- int ret = -1;
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- struct iatt *stbuf = NULL;
- inode_t *inode = NULL;
- xlator_t *subvol1 = NULL, *subvol2 = NULL;
-
- local = frame->local;
- prev = cookie;
-
- local->op_errno = op_errno;
-
- if ((local->fop == GF_FOP_FREMOVEXATTR) &&
- (op_ret == -1) && (op_errno == EBADF) && !(local->fd_checked)) {
- ret = dht_check_and_open_fd_on_subvol (this, frame);
- if (ret)
- goto out;
- return 0;
- }
-
- if ((op_ret == -1) && !dht_inode_missing (op_errno)) {
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->name);
- goto out;
- }
-
- if (local->call_cnt != 1)
- goto out;
-
- ret = dict_get_bin (xdata, DHT_IATT_IN_XDATA_KEY, (void **) &stbuf);
-
- if ((!op_ret) && !stbuf) {
- goto out;
+ tmp = dict_get(xattr, GF_XATTR_FIX_LAYOUT_KEY);
+ if (tmp) {
+ ret = dict_get_uint32(xattr, "new-commit-hash", &new_hash);
+ if (ret == 0) {
+ gf_msg_debug(this->name, 0,
+ "updating commit hash for %s from %u to %u",
+ uuid_utoa(loc->gfid), layout->commit_hash, new_hash);
+ layout->commit_hash = new_hash;
+
+ ret = dht_update_commit_hash_for_layout(frame);
+ if (ret) {
+ op_errno = ENOTCONN;
+ goto err;
+ }
+ return ret;
}
- local->op_ret = 0;
-
- local->rebalance.target_op_fn = dht_removexattr2;
- if (xdata)
- local->rebalance.xdata = dict_ref (xdata);
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_FIX_LAYOUT_INFO,
+ "fixing the layout of %s", loc->path);
- /* Phase 2 of migration */
- if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (stbuf)) {
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret)
- return 0;
+ ret = dht_fix_directory_layout(frame, dht_fix_layout_setxattr_cbk,
+ layout);
+ if (ret) {
+ op_errno = ENOTCONN;
+ goto err;
}
+ return ret;
+ }
+
+ tmp = dict_get(xattr, "distribute.directory-spread-count");
+ if (tmp) {
+ /* Setxattr value is packed as 'binary', not string */
+ memcpy(value, tmp->data, min(tmp->len, 4095));
+ ret = gf_string2uint32(value, &dir_spread);
+ if (!ret && ((dir_spread <= conf->subvolume_cnt) && (dir_spread > 0))) {
+ layout->spread_cnt = dir_spread;
+
+ ret = dht_fix_directory_layout(frame, dht_common_setxattr_cbk,
+ layout);
+ if (ret) {
+ op_errno = ENOTCONN;
+ goto err;
+ }
+ return ret;
+ }
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_OPERATION_NOT_SUP,
+ "wrong 'directory-spread-count' value (%s)", value);
+ op_errno = ENOTSUP;
+ goto err;
+ }
+
+ tmp = dict_get(xattr, "glusterfs.dht.nuke");
+ if (tmp) {
+ return dht_nuke_dir(frame, this, loc, tmp);
+ }
+ local->xattr_req = xdata ? dict_ref(xdata) : dict_new();
+
+ if (IA_ISDIR(loc->inode->ia_type)) {
+ local->hashed_subvol = NULL;
+ ret = dht_dir_common_set_remove_xattr(frame, this, loc, NULL, xattr,
+ flags, xdata, &op_errno);
+ if (ret)
+ goto err;
+ } else {
+ local->rebalance.xattr = dict_ref(xattr);
+ local->rebalance.flags = flags;
+ local->call_cnt = 1;
- /* Phase 1 of migration */
- if (IS_DHT_MIGRATION_PHASE1 (stbuf)) {
- inode = (local->fd) ? local->fd->inode : local->loc.inode;
+ ret = dict_set_int8(local->xattr_req, DHT_IATT_IN_XDATA_KEY, 1);
- ret = dht_inode_ctx_get_mig_info (this, inode,
- &subvol1, &subvol2);
- if (!dht_mig_info_is_invalid (local->cached_subvol,
- subvol1, subvol2)) {
- dht_removexattr2 (this, subvol2, frame, 0);
- return 0;
- }
+ STACK_WIND_COOKIE(frame, dht_file_setxattr_cbk, subvol, subvol,
+ subvol->fops->setxattr, loc, xattr, flags,
+ local->xattr_req);
+ }
- ret = dht_rebalance_in_progress_check (this, frame);
- if (!ret)
- return 0;
- }
+ return 0;
-out:
- if (local->fop == GF_FOP_REMOVEXATTR) {
- DHT_STACK_UNWIND (removexattr, frame, op_ret, op_errno, NULL);
- } else {
- DHT_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, NULL);
- }
- return 0;
+err:
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(setxattr, frame, -1, op_errno, NULL);
+ return 0;
}
-int
-dht_removexattr2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame,
- int ret)
+static int
+dht_removexattr2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
{
- dht_local_t *local = NULL;
- int op_errno = EINVAL;
-
- if (!frame || !frame->local)
- goto err;
+ dht_local_t *local = NULL;
+ int op_errno = EINVAL;
- local = frame->local;
- op_errno = local->op_errno;
+ if (!frame || !frame->local)
+ goto err;
- local->call_cnt = 2; /* This is the second attempt */
+ local = frame->local;
+ op_errno = local->op_errno;
- if (we_are_not_migrating (ret)) {
+ local->call_cnt = 2; /* This is the second attempt */
- /* This dht xlator is not migrating the file. Unwind and
- * pass on the original mode bits so the higher DHT layer
- * can handle this.
- */
- DHT_STACK_UNWIND (removexattr, frame, local->op_ret,
- local->op_errno, local->rebalance.xdata);
- return 0;
- }
+ if (we_are_not_migrating(ret)) {
+ /* This dht xlator is not migrating the file. Unwind and
+ * pass on the original mode bits so the higher DHT layer
+ * can handle this.
+ */
+ DHT_STACK_UNWIND(removexattr, frame, local->op_ret, local->op_errno,
+ local->rebalance.xdata);
+ return 0;
+ }
- if (subvol == NULL)
- goto err;
+ if (subvol == NULL)
+ goto err;
- if (local->fop == GF_FOP_REMOVEXATTR) {
- STACK_WIND_COOKIE (frame, dht_file_removexattr_cbk, subvol,
- subvol, subvol->fops->removexattr,
- &local->loc, local->key, local->xattr_req);
- } else {
- STACK_WIND_COOKIE (frame, dht_file_removexattr_cbk, subvol,
- subvol, subvol->fops->fremovexattr,
- local->fd, local->key, local->xattr_req);
- }
+ if (local->fop == GF_FOP_REMOVEXATTR) {
+ STACK_WIND_COOKIE(frame, dht_file_removexattr_cbk, subvol, subvol,
+ subvol->fops->removexattr, &local->loc, local->key,
+ local->xattr_req);
+ } else {
+ STACK_WIND_COOKIE(frame, dht_file_removexattr_cbk, subvol, subvol,
+ subvol->fops->fremovexattr, local->fd, local->key,
+ local->xattr_req);
+ }
- return 0;
+ return 0;
err:
- DHT_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
- return 0;
+ DHT_STACK_UNWIND(removexattr, frame, -1, op_errno, NULL);
+ return 0;
}
-
int
-dht_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+dht_file_removexattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- xlator_t *prev = NULL;
-
- local = frame->local;
- prev = cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->name);
- goto unlock;
- }
+ int ret = -1;
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ struct iatt *stbuf = NULL;
+ inode_t *inode = NULL;
+ xlator_t *subvol1 = NULL, *subvol2 = NULL;
- local->op_ret = 0;
- }
-unlock:
- UNLOCK (&frame->lock);
+ local = frame->local;
+ prev = cookie;
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- DHT_STACK_UNWIND (removexattr, frame, local->op_ret,
- local->op_errno, NULL);
- }
+ local->op_errno = op_errno;
+ if ((local->fop == GF_FOP_FREMOVEXATTR) &&
+ dht_check_remote_fd_failed_error(local, op_ret, op_errno)) {
+ ret = dht_check_and_open_fd_on_subvol(this, frame);
+ if (ret)
+ goto out;
return 0;
-}
+ }
+ if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->name);
+ goto out;
+ }
-int
-dht_removexattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *key, dict_t *xdata)
-{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
- dht_layout_t *layout = NULL;
- int call_cnt = 0;
- dht_conf_t *conf = NULL;
- int i;
- int ret = 0;
-
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (this->private, err);
-
- conf = this->private;
+ if (local->call_cnt != 1)
+ goto out;
- GF_IF_NATIVE_XATTR_GOTO (conf->wild_xattr_name, key, op_errno, err);
+ ret = dict_get_bin(xdata, DHT_IATT_IN_XDATA_KEY, (void **)&stbuf);
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
+ if ((!op_ret) && !stbuf) {
+ goto out;
+ }
- local = dht_local_init (frame, loc, NULL, GF_FOP_REMOVEXATTR);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+ local->op_ret = 0;
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
+ local->rebalance.target_op_fn = dht_removexattr2;
+ if (xdata)
+ local->rebalance.xdata = dict_ref(xdata);
- layout = local->layout;
- if (!local->layout) {
- gf_msg_debug (this->name, 0,
- "no layout for path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
- local->xattr_req = (xdata) ? dict_ref (xdata) : dict_new ();
+ /* Phase 2 of migration */
+ if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2(stbuf)) {
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret)
+ return 0;
+ }
- local->call_cnt = call_cnt = layout->cnt;
- local->key = gf_strdup (key);
+ /* Phase 1 of migration */
+ if (IS_DHT_MIGRATION_PHASE1(stbuf)) {
+ inode = (local->fd) ? local->fd->inode : local->loc.inode;
- if (key &&
- (strncmp (key, conf->mds_xattr_key, strlen(key)) == 0)) {
- op_errno = ENOTSUP;
- goto err;
+ ret = dht_inode_ctx_get_mig_info(this, inode, &subvol1, &subvol2);
+ if (!dht_mig_info_is_invalid(local->cached_subvol, subvol1, subvol2)) {
+ dht_removexattr2(this, subvol2, frame, 0);
+ return 0;
}
- if (IA_ISDIR (loc->inode->ia_type)) {
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND_COOKIE (frame, dht_removexattr_cbk,
- layout->list[i].xlator,
- layout->list[i].xlator,
- layout->list[i].xlator->fops->removexattr,
- loc, key, local->xattr_req);
- }
+ ret = dht_rebalance_in_progress_check(this, frame);
+ if (!ret)
+ return 0;
+ }
- } else {
+out:
+ if (local->fop == GF_FOP_REMOVEXATTR) {
+ DHT_STACK_UNWIND(removexattr, frame, op_ret, op_errno, xdata);
+ } else {
+ DHT_STACK_UNWIND(fremovexattr, frame, op_ret, op_errno, xdata);
+ }
+ return 0;
+}
- local->call_cnt = 1;
- ret = dict_set_int8 (local->xattr_req, DHT_IATT_IN_XDATA_KEY, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_DICT_SET_FAILED, "Failed to "
- "set dictionary key %s for %s",
- DHT_IATT_IN_XDATA_KEY, loc->path);
- }
+int
+dht_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *key, dict_t *xdata)
+{
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+ dht_layout_t *layout = NULL;
+ int call_cnt = 0;
+ dht_conf_t *conf = NULL;
+ int ret = 0;
+
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(this->private, err);
+
+ conf = this->private;
+
+ GF_IF_NATIVE_XATTR_GOTO(conf->wild_xattr_name, key, op_errno, err);
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_REMOVEXATTR);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for path=%s",
+ loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ layout = local->layout;
+ if (!local->layout) {
+ gf_msg_debug(this->name, 0, "no layout for path=%s", loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
+ local->xattr_req = (xdata) ? dict_ref(xdata) : dict_new();
+
+ local->call_cnt = call_cnt = layout->cnt;
+ local->key = gf_strdup(key);
+
+ if (key && (strncmp(key, conf->mds_xattr_key, strlen(key)) == 0)) {
+ op_errno = ENOTSUP;
+ goto err;
+ }
+
+ if (IA_ISDIR(loc->inode->ia_type)) {
+ local->hashed_subvol = NULL;
+ ret = dht_dir_common_set_remove_xattr(frame, this, loc, NULL, NULL, 0,
+ local->xattr_req, &op_errno);
+ if (ret)
+ goto err;
- STACK_WIND_COOKIE (frame, dht_file_removexattr_cbk, subvol,
- subvol, subvol->fops->removexattr, loc, key,
- local->xattr_req);
+ } else {
+ local->call_cnt = 1;
+ ret = dict_set_int8(local->xattr_req, DHT_IATT_IN_XDATA_KEY, 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_DICT_SET_FAILED,
+ "Failed to "
+ "set dictionary key %s for %s",
+ DHT_IATT_IN_XDATA_KEY, loc->path);
}
- return 0;
+ STACK_WIND_COOKIE(frame, dht_file_removexattr_cbk, subvol, subvol,
+ subvol->fops->removexattr, loc, key,
+ local->xattr_req);
+ }
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(removexattr, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
int
-dht_fremovexattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, const char *key, dict_t *xdata)
+dht_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *key,
+ dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
- dht_layout_t *layout = NULL;
- int call_cnt = 0;
- dht_conf_t *conf = 0;
- int ret = 0;
-
- int i;
-
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (this->private, err);
-
- conf = this->private;
-
- GF_IF_NATIVE_XATTR_GOTO (conf->wild_xattr_name, key, op_errno, err);
-
- VALIDATE_OR_GOTO (frame, err);
-
- local = dht_local_init (frame, NULL, fd, GF_FOP_FREMOVEXATTR);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for inode=%s",
- uuid_utoa (fd->inode->gfid));
- op_errno = EINVAL;
- goto err;
- }
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+ dht_layout_t *layout = NULL;
+ int call_cnt = 0;
+ dht_conf_t *conf = 0;
+ int ret = 0;
+
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(this->private, err);
+
+ conf = this->private;
+
+ GF_IF_NATIVE_XATTR_GOTO(conf->wild_xattr_name, key, op_errno, err);
+
+ VALIDATE_OR_GOTO(frame, err);
+
+ local = dht_local_init(frame, NULL, fd, GF_FOP_FREMOVEXATTR);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for inode=%s",
+ uuid_utoa(fd->inode->gfid));
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ layout = local->layout;
+ if (!local->layout) {
+ gf_msg_debug(this->name, 0, "no layout for inode=%s",
+ uuid_utoa(fd->inode->gfid));
+ op_errno = EINVAL;
+ goto err;
+ }
+ local->xattr_req = xdata ? dict_ref(xdata) : dict_new();
+
+ local->call_cnt = call_cnt = layout->cnt;
+ local->key = gf_strdup(key);
+
+ if (IA_ISDIR(fd->inode->ia_type)) {
+ local->hashed_subvol = NULL;
+ ret = dht_dir_common_set_remove_xattr(frame, this, NULL, fd, NULL, 0,
+ local->xattr_req, &op_errno);
+ if (ret)
+ goto err;
- layout = local->layout;
- if (!local->layout) {
- gf_msg_debug (this->name, 0,
- "no layout for inode=%s",
- uuid_utoa (fd->inode->gfid));
- op_errno = EINVAL;
- goto err;
+ } else {
+ local->call_cnt = 1;
+ ret = dict_set_int8(local->xattr_req, DHT_IATT_IN_XDATA_KEY, 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_DICT_SET_FAILED,
+ "Failed to "
+ "set dictionary key %s for fd=%p",
+ DHT_IATT_IN_XDATA_KEY, fd);
}
- local->xattr_req = xdata ? dict_ref (xdata) : dict_new ();
-
- local->call_cnt = call_cnt = layout->cnt;
- local->key = gf_strdup (key);
- if (IA_ISDIR (fd->inode->ia_type)) {
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND_COOKIE (frame, dht_removexattr_cbk,
- layout->list[i].xlator,
- layout->list[i].xlator,
- layout->list[i].xlator->fops->fremovexattr,
- fd, key, local->xattr_req);
- }
+ STACK_WIND_COOKIE(frame, dht_file_removexattr_cbk, subvol, subvol,
+ subvol->fops->fremovexattr, fd, key,
+ local->xattr_req);
+ }
- } else {
-
- local->call_cnt = 1;
- ret = dict_set_int8 (local->xattr_req, DHT_IATT_IN_XDATA_KEY, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_DICT_SET_FAILED, "Failed to "
- "set dictionary key %s for fd=%p",
- DHT_IATT_IN_XDATA_KEY, fd);
- }
-
- STACK_WIND_COOKIE (frame, dht_file_removexattr_cbk, subvol,
- subvol, subvol->fops->fremovexattr, fd, key,
- local->xattr_req);
- }
-
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fremovexattr, frame, -1, op_errno, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(fremovexattr, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
-
int
-dht_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
+dht_fd_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, fd_t *fd, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- xlator_t *prev = NULL;
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ xlator_t *prev = NULL;
- local = frame->local;
- prev = cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->name);
- goto unlock;
- }
+ local = frame->local;
+ prev = cookie;
- local->op_ret = 0;
+ LOCK(&frame->lock);
+ {
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ UNLOCK(&frame->lock);
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->name);
+ goto post_unlock;
}
-unlock:
- UNLOCK (&frame->lock);
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt))
- DHT_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
- local->fd, NULL);
+ local->op_ret = 0;
+ }
+ UNLOCK(&frame->lock);
+post_unlock:
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt))
+ DHT_STACK_UNWIND(open, frame, local->op_ret, local->op_errno, local->fd,
+ NULL);
- return 0;
+ return 0;
}
/*
* dht_normalize_stats -
*/
-void
-dht_normalize_stats (struct statvfs *buf, unsigned long bsize,
- unsigned long frsize)
+static void
+dht_normalize_stats(struct statvfs *buf, unsigned long bsize,
+ unsigned long frsize)
{
- double factor = 0;
-
- if (buf->f_bsize != bsize) {
- buf->f_bsize = bsize;
- }
+ double factor = 0;
- if (buf->f_frsize != frsize) {
- factor = ((double) buf->f_frsize) / frsize;
- buf->f_frsize = frsize;
- buf->f_blocks = (fsblkcnt_t) (factor * buf->f_blocks);
- buf->f_bfree = (fsblkcnt_t) (factor * buf->f_bfree);
- buf->f_bavail = (fsblkcnt_t) (factor * buf->f_bavail);
+ if (buf->f_bsize != bsize) {
+ buf->f_bsize = bsize;
+ }
- }
+ if (buf->f_frsize != frsize) {
+ factor = ((double)buf->f_frsize) / frsize;
+ buf->f_frsize = frsize;
+ buf->f_blocks = (fsblkcnt_t)(factor * buf->f_blocks);
+ buf->f_bfree = (fsblkcnt_t)(factor * buf->f_bfree);
+ buf->f_bavail = (fsblkcnt_t)(factor * buf->f_bavail);
+ }
}
-int
-dht_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct statvfs *statvfs,
- dict_t *xdata)
-{
-
- gf_boolean_t event = _gf_false;
- qdstatfs_action_t action = qdstatfs_action_OFF;
- dht_local_t * local = NULL;
- int this_call_cnt = 0;
- int bsize = 0;
- int frsize = 0;
- GF_UNUSED int ret = 0;
- unsigned long new_usage = 0;
- unsigned long cur_usage = 0;
-
- local = frame->local;
- GF_ASSERT (local);
-
- if (xdata)
- ret = dict_get_int8 (xdata, "quota-deem-statfs",
- (int8_t *)&event);
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- goto unlock;
- }
- if (!statvfs) {
- op_errno = EINVAL;
- local->op_ret = -1;
- goto unlock;
- }
- local->op_ret = 0;
-
- if (local->quota_deem_statfs) {
- if (event == _gf_true) {
- action = qdstatfs_action_COMPARE;
- } else {
- action = qdstatfs_action_NEGLECT;
- }
- } else {
- if (event == _gf_true) {
- action = qdstatfs_action_REPLACE;
- local->quota_deem_statfs = _gf_true;
- }
- }
-
- if (local->quota_deem_statfs) {
- switch (action) {
- case qdstatfs_action_NEGLECT:
- goto unlock;
-
- case qdstatfs_action_REPLACE:
- local->statvfs = *statvfs;
- goto unlock;
-
- case qdstatfs_action_COMPARE:
- new_usage = statvfs->f_blocks -
- statvfs->f_bfree;
- cur_usage = local->statvfs.f_blocks -
- local->statvfs.f_bfree;
+static int
+dht_statfs_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct statvfs *statvfs, dict_t *xdata)
+{
+ gf_boolean_t event = _gf_false;
+ qdstatfs_action_t action = qdstatfs_action_OFF;
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ int bsize = 0;
+ int frsize = 0;
+ GF_UNUSED int ret = 0;
+ unsigned long new_usage = 0;
+ unsigned long cur_usage = 0;
+
+ local = frame->local;
+ GF_ASSERT(local);
+
+ if (xdata)
+ ret = dict_get_int8(xdata, "quota-deem-statfs", (int8_t *)&event);
+
+ LOCK(&frame->lock);
+ {
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ goto unlock;
+ }
+ if (!statvfs) {
+ op_errno = EINVAL;
+ local->op_ret = -1;
+ goto unlock;
+ }
+ local->op_ret = 0;
- /* Take the max of the usage from subvols */
- if (new_usage >= cur_usage)
- local->statvfs = *statvfs;
- goto unlock;
+ if (local->quota_deem_statfs) {
+ if (event == _gf_true) {
+ action = qdstatfs_action_COMPARE;
+ } else {
+ action = qdstatfs_action_NEGLECT;
+ }
+ } else {
+ if (event == _gf_true) {
+ action = qdstatfs_action_REPLACE;
+ local->quota_deem_statfs = _gf_true;
+ }
+ }
- default:
- break;
- }
- }
+ if (local->quota_deem_statfs) {
+ switch (action) {
+ case qdstatfs_action_NEGLECT:
+ goto unlock;
- if (local->statvfs.f_bsize != 0) {
- bsize = max(local->statvfs.f_bsize, statvfs->f_bsize);
- frsize = max(local->statvfs.f_frsize, statvfs->f_frsize);
- dht_normalize_stats(&local->statvfs, bsize, frsize);
- dht_normalize_stats(statvfs, bsize, frsize);
- } else {
- local->statvfs.f_bsize = statvfs->f_bsize;
- local->statvfs.f_frsize = statvfs->f_frsize;
- }
+ case qdstatfs_action_REPLACE:
+ local->statvfs = *statvfs;
+ goto unlock;
- local->statvfs.f_blocks += statvfs->f_blocks;
- local->statvfs.f_bfree += statvfs->f_bfree;
- local->statvfs.f_bavail += statvfs->f_bavail;
- local->statvfs.f_files += statvfs->f_files;
- local->statvfs.f_ffree += statvfs->f_ffree;
- local->statvfs.f_favail += statvfs->f_favail;
- local->statvfs.f_fsid = statvfs->f_fsid;
- local->statvfs.f_flag = statvfs->f_flag;
- local->statvfs.f_namemax = statvfs->f_namemax;
+ case qdstatfs_action_COMPARE:
+ new_usage = statvfs->f_blocks - statvfs->f_bfree;
+ cur_usage = local->statvfs.f_blocks -
+ local->statvfs.f_bfree;
+ /* Take the max of the usage from subvols */
+ if (new_usage >= cur_usage)
+ local->statvfs = *statvfs;
+ goto unlock;
+ default:
+ break;
+ }
}
-unlock:
- UNLOCK (&frame->lock);
+ if (local->statvfs.f_bsize != 0) {
+ bsize = max(local->statvfs.f_bsize, statvfs->f_bsize);
+ frsize = max(local->statvfs.f_frsize, statvfs->f_frsize);
+ dht_normalize_stats(&local->statvfs, bsize, frsize);
+ dht_normalize_stats(statvfs, bsize, frsize);
+ } else {
+ local->statvfs.f_bsize = statvfs->f_bsize;
+ local->statvfs.f_frsize = statvfs->f_frsize;
+ }
+
+ local->statvfs.f_blocks += statvfs->f_blocks;
+ local->statvfs.f_bfree += statvfs->f_bfree;
+ local->statvfs.f_bavail += statvfs->f_bavail;
+ local->statvfs.f_files += statvfs->f_files;
+ local->statvfs.f_ffree += statvfs->f_ffree;
+ local->statvfs.f_favail += statvfs->f_favail;
+ local->statvfs.f_fsid = statvfs->f_fsid;
+ local->statvfs.f_flag = statvfs->f_flag;
+ local->statvfs.f_namemax = statvfs->f_namemax;
+ }
+unlock:
+ UNLOCK(&frame->lock);
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt))
- DHT_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno,
- &local->statvfs, xdata);
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt))
+ DHT_STACK_UNWIND(statfs, frame, local->op_ret, local->op_errno,
+ &local->statvfs, xdata);
- return 0;
+ return 0;
}
-
int
-dht_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
-{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- int op_errno = -1;
- int i = -1;
- inode_t *inode = NULL;
- inode_table_t *itable = NULL;
- uuid_t root_gfid = {0, };
- loc_t newloc = {0, };
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (this->private, err);
-
- conf = this->private;
-
- local = dht_local_init (frame, NULL, NULL, GF_FOP_STATFS);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+dht_statfs(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int op_errno = -1;
+ int i = -1;
+ inode_t *inode = NULL;
+ inode_table_t *itable = NULL;
+ static uuid_t root_gfid = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1};
+ loc_t newloc = {
+ 0,
+ };
- if (loc->inode && !IA_ISDIR (loc->inode->ia_type)) {
- itable = loc->inode->table;
- if (!itable) {
- op_errno = EINVAL;
- goto err;
- }
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(this->private, err);
- loc = &local->loc2;
- root_gfid[15] = 1;
+ conf = this->private;
- inode = inode_find (itable, root_gfid);
- if (!inode) {
- op_errno = EINVAL;
- goto err;
- }
+ local = dht_local_init(frame, NULL, NULL, GF_FOP_STATFS);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- dht_build_root_loc (inode, &newloc);
- loc = &newloc;
+ if (loc->inode && !IA_ISDIR(loc->inode->ia_type)) {
+ itable = loc->inode->table;
+ if (!itable) {
+ op_errno = EINVAL;
+ goto err;
}
- local->call_cnt = conf->subvolume_cnt;
+ loc = &local->loc2;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- STACK_WIND (frame, dht_statfs_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->statfs, loc,
- xdata);
+ inode = inode_find(itable, root_gfid);
+ if (!inode) {
+ op_errno = EINVAL;
+ goto err;
}
- return 0;
+
+ dht_build_root_loc(inode, &newloc);
+ loc = &newloc;
+ }
+
+ local->call_cnt = conf->subvolume_cnt;
+
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ STACK_WIND(frame, dht_statfs_cbk, conf->subvolumes[i],
+ conf->subvolumes[i]->fops->statfs, loc, xdata);
+ }
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(statfs, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
-
int
-dht_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
- dict_t *xdata)
+dht_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- int op_errno = -1;
- int i = -1;
- int ret = 0;
- gf_boolean_t new_xdata = _gf_false;
- xlator_t **subvolumes = NULL;
- int call_count = 0;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (this->private, err);
-
- conf = this->private;
-
- local = dht_local_init (frame, loc, fd, GF_FOP_OPENDIR);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->first_up_subvol = dht_first_up_subvol (this);
-
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int op_errno = -1;
+ int i = -1;
+ int ret = 0;
+ gf_boolean_t new_xdata = _gf_false;
+ xlator_t **subvolumes = NULL;
+ int call_count = 0;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+ VALIDATE_OR_GOTO(this->private, err);
+
+ conf = this->private;
+
+ local = dht_local_init(frame, loc, fd, GF_FOP_OPENDIR);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+ local->first_up_subvol = dht_first_up_subvol(this);
+
+ if (!xdata) {
+ xdata = dict_new();
if (!xdata) {
- xdata = dict_new ();
- if (!xdata) {
- op_errno = ENOMEM;
- goto err;
- }
- new_xdata = _gf_true;
+ op_errno = ENOMEM;
+ goto err;
+ }
+ new_xdata = _gf_true;
+ }
+
+ ret = dict_set_uint32(xdata, conf->link_xattr_name, 256);
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary value : key = %s",
+ conf->link_xattr_name);
+
+ /* dht_readdirp will wind to all subvols so open has to be sent to
+ * all subvols whether or not conf->local_subvols is set */
+
+ call_count = local->call_cnt = conf->subvolume_cnt;
+ subvolumes = conf->subvolumes;
+
+ /* In case of parallel-readdir, the readdir-ahead will be loaded
+ * below dht, in this case, if we want to enable or disable SKIP_DIRs
+ * it has to be done in opendir, so that prefetching logic in
+ * readdir-ahead, honors it */
+ for (i = 0; i < call_count; i++) {
+ if (conf->readdir_optimize == _gf_true) {
+ if (subvolumes[i] != local->first_up_subvol) {
+ ret = dict_set_int32(xdata, GF_READDIR_SKIP_DIRS, 1);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary"
+ " value :key = %s, ret:%d",
+ GF_READDIR_SKIP_DIRS, ret);
+ }
}
- ret = dict_set_uint32 (xdata, conf->link_xattr_name, 256);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0, DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value : key = %s",
- conf->link_xattr_name);
-
- /* dht_readdirp will wind to all subvols so open has to be sent to
- * all subvols whether or not conf->local_subvols is set */
-
- call_count = local->call_cnt = conf->subvolume_cnt;
- subvolumes = conf->subvolumes;
-
- /* In case of parallel-readdir, the readdir-ahead will be loaded
- * below dht, in this case, if we want to enable or disable SKIP_DIRs
- * it has to be done in opendir, so that prefetching logic in
- * readdir-ahead, honors it */
- for (i = 0; i < call_count; i++) {
- if (conf->readdir_optimize == _gf_true) {
- if (subvolumes[i] != local->first_up_subvol) {
- ret = dict_set_int32 (xdata,
- GF_READDIR_SKIP_DIRS, 1);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary"
- " value :key = %s, ret:%d",
- GF_READDIR_SKIP_DIRS, ret);
- }
- }
-
- STACK_WIND_COOKIE (frame, dht_fd_cbk,
- subvolumes[i],
- subvolumes[i],
- subvolumes[i]->fops->opendir,
- loc, fd, xdata);
- dict_del (xdata, GF_READDIR_SKIP_DIRS);
- }
+ STACK_WIND_COOKIE(frame, dht_fd_cbk, subvolumes[i], subvolumes[i],
+ subvolumes[i]->fops->opendir, loc, fd, xdata);
+ dict_del(xdata, GF_READDIR_SKIP_DIRS);
+ }
- if (new_xdata)
- dict_unref (xdata);
+ if (new_xdata)
+ dict_unref(xdata);
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (opendir, frame, -1, op_errno, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(opendir, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
-
/* dht_readdirp_cbk creates a new dentry and dentry->inode is not assigned.
- This functions assigns an inode if all of the following conditions are true:
+ This functions assigns an inode if all of the following conditions are
+ true:
- * DHT has only one child. In this case the entire layout is present on this
- single child and hence we can set complete layout in inode.
- * backend has complete layout and there are no anomalies in it and from this
- information layout can be constructed and set in inode.
+ * DHT has only one child. In this case the entire layout is present on
+ this single child and hence we can set complete layout in inode.
+ * backend has complete layout and there are no anomalies in it and from
+ this information layout can be constructed and set in inode.
*/
-void
-dht_populate_inode_for_dentry (xlator_t *this, xlator_t *subvol,
- gf_dirent_t *entry, gf_dirent_t *orig_entry)
+static void
+dht_populate_inode_for_dentry(xlator_t *this, xlator_t *subvol,
+ gf_dirent_t *entry, gf_dirent_t *orig_entry)
{
- dht_layout_t *layout = NULL;
- int ret = 0;
- loc_t loc = {0, };
-
- if (gf_uuid_is_null (orig_entry->d_stat.ia_gfid)) {
- /* this skips the '..' entry for the root of the volume */
- return;
- }
+ dht_layout_t *layout = NULL;
+ int ret = 0;
+ loc_t loc = {
+ 0,
+ };
- gf_uuid_copy (loc.gfid, orig_entry->d_stat.ia_gfid);
- loc.inode = inode_ref (orig_entry->inode);
+ if (gf_uuid_is_null(orig_entry->d_stat.ia_gfid)) {
+ /* this skips the '..' entry for the root of the volume */
+ return;
+ }
- if (is_revalidate (&loc)) {
- goto out;
- }
+ gf_uuid_copy(loc.gfid, orig_entry->d_stat.ia_gfid);
+ loc.inode = inode_ref(orig_entry->inode);
- layout = dht_layout_new (this, 1);
- if (!layout)
- goto out;
+ if (is_revalidate(&loc)) {
+ goto out;
+ }
- ret = dht_layout_merge (this, layout, subvol, 0, 0, orig_entry->dict);
- if (!ret) {
- ret = dht_layout_normalize (this, &loc, layout);
- if (ret == 0) {
- dht_layout_set (this, orig_entry->inode, layout);
- entry->inode = inode_ref (orig_entry->inode);
- layout = NULL;
- }
+ layout = dht_layout_new(this, 1);
+ if (!layout)
+ goto out;
+ ret = dht_layout_merge(this, layout, subvol, 0, 0, orig_entry->dict);
+ if (!ret) {
+ ret = dht_layout_normalize(this, &loc, layout);
+ if (ret == 0) {
+ dht_layout_set(this, orig_entry->inode, layout);
+ entry->inode = inode_ref(orig_entry->inode);
+ layout = NULL;
}
+ }
- if (layout)
- dht_layout_unref (this, layout);
+ if (layout)
+ dht_layout_unref(this, layout);
out:
- loc_wipe (&loc);
- return;
+ loc_wipe(&loc);
+ return;
}
-int
-dht_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
- int op_errno, gf_dirent_t *orig_entries, dict_t *xdata)
-{
- dht_local_t *local = NULL;
- gf_dirent_t *orig_entry = NULL;
- gf_dirent_t *entry = NULL;
- xlator_t *prev = NULL;
- xlator_t *next_subvol = NULL;
- off_t next_offset = 0;
- int count = 0;
- dht_layout_t *layout = NULL;
- dht_conf_t *conf = NULL;
- dht_methods_t *methods = NULL;
- xlator_t *subvol = 0;
- xlator_t *hashed_subvol = 0;
- int ret = 0;
- int readdir_optimize = 0;
- inode_table_t *itable = NULL;
- inode_t *inode = NULL;
-
- prev = cookie;
- local = frame->local;
- itable = local->fd ? local->fd->inode->table : NULL;
+/* Posix returns op_errno = ENOENT to indicate that there are no more
+ * entries
+ */
+static int
+dht_readdirp_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, gf_dirent_t *orig_entries, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ gf_dirent_t entries;
+ gf_dirent_t *orig_entry = NULL;
+ gf_dirent_t *entry = NULL;
+ xlator_t *prev = NULL;
+ xlator_t *next_subvol = NULL;
+ off_t next_offset = 0;
+ int count = 0;
+ dht_layout_t *layout = NULL;
+ dht_conf_t *conf = NULL;
+ dht_methods_t *methods = NULL;
+ xlator_t *subvol = 0;
+ xlator_t *hashed_subvol = 0;
+ int ret = 0;
+ int readdir_optimize = 0;
+ inode_table_t *itable = NULL;
+ inode_t *inode = NULL;
+ gf_boolean_t skip_hashed_check = _gf_false;
+
+ INIT_LIST_HEAD(&entries.list);
+
+ prev = cookie;
+ local = frame->local;
+ GF_VALIDATE_OR_GOTO(this->name, local->fd, unwind);
+
+ itable = local->fd->inode->table;
+
+ conf = this->private;
+ GF_VALIDATE_OR_GOTO(this->name, conf, unwind);
+
+ methods = &(conf->methods);
+
+ if (op_ret <= 0) {
+ goto done;
+ }
+
+ /* Why aren't we skipping DHT entirely in case of a single subvol?
+ * Because if this was a larger volume earlier and all but one subvol
+ * was removed, there might be stale linkto files on the subvol.
+ */
+ if (conf->subvolume_cnt == 1) {
+ /* return all directory and file entries except
+ * linkto files for a single child DHT
+ */
+ skip_hashed_check = _gf_true;
+ }
- conf = this->private;
- GF_VALIDATE_OR_GOTO(this->name, conf, unwind);
+ if (!local->layout)
+ local->layout = dht_layout_get(this, local->fd->inode);
- methods = &(conf->methods);
+ layout = local->layout;
- local->op_errno = op_errno;
+ /* This will skip the entries on the subvol without a layout,
+ * hence preventing the crash but rmdir might fail with
+ * "directory not empty" errors*/
- if (op_ret < 0)
- goto done;
+ if (layout == NULL)
+ goto done;
- if (local->op_ret < 0)
- local->op_ret = 0;
+ if (conf->readdir_optimize == _gf_true)
+ readdir_optimize = 1;
- if (!local->layout)
- local->layout = dht_layout_get (this, local->fd->inode);
+ gf_msg_debug(this->name, 0, "Processing entries from %s", prev->name);
- layout = local->layout;
+ list_for_each_entry(orig_entry, (&orig_entries->list), list)
+ {
+ next_offset = orig_entry->d_off;
- /* We have seen crashes in while running "rm -rf" on tier volumes
- when the layout was NULL on the hot tier. This will skip the
- entries on the subvol without a layout, hence preventing the crash
- but rmdir might fail with "directory not empty" errors*/
+ gf_msg_debug(this->name, 0, "%s: entry = %s, type = %d", prev->name,
+ orig_entry->d_name, orig_entry->d_type);
- if (layout == NULL)
- goto done;
-
- if (conf->readdir_optimize == _gf_true)
- readdir_optimize = 1;
+ if (IA_ISINVAL(orig_entry->d_stat.ia_type)) {
+ /*stat failed somewhere- display this entry but the data may
+ * be inaccurate.
+ */
+ gf_msg_debug(this->name, EINVAL, "Invalid stat for %s (gfid %s)",
+ orig_entry->d_name,
+ uuid_utoa(orig_entry->d_stat.ia_gfid));
+ }
- gf_msg_debug (this->name, 0, "Processing entries from %s",
- prev->name);
+ if (check_is_linkfile(NULL, (&orig_entry->d_stat), orig_entry->dict,
+ conf->link_xattr_name)) {
+ gf_msg_debug(this->name, 0, "%s: %s is a linkto file", prev->name,
+ orig_entry->d_name);
+ continue;
+ }
- list_for_each_entry (orig_entry, (&orig_entries->list), list) {
- next_offset = orig_entry->d_off;
+ if (skip_hashed_check) {
+ goto list;
+ }
- gf_msg_debug (this->name, 0, "%s: entry = %s, type = %d",
- prev->name, orig_entry->d_name,
- orig_entry->d_type);
+ if (check_is_dir(NULL, (&orig_entry->d_stat), NULL)) {
+ /*Directory entries filtering :
+ * a) If rebalance is running, pick from first_up_subvol
+ * b) (rebalance not running)hashed subvolume is NULL or
+ * down then filter in first_up_subvolume. Other wise the
+ * corresponding hashed subvolume will take care of the
+ * directory entry.
+ */
+ if (readdir_optimize) {
+ if (prev == local->first_up_subvol)
+ goto list;
+ else
+ continue;
+ }
- if (IA_ISINVAL(orig_entry->d_stat.ia_type)) {
- /*stat failed somewhere- ignore this entry*/
- gf_msg_debug (this->name, EINVAL,
- "Invalid stat, ignoring entry "
- "%s gfid %s", orig_entry->d_name,
- uuid_utoa (orig_entry->d_stat.ia_gfid));
- continue;
- }
+ hashed_subvol = methods->layout_search(this, layout,
+ orig_entry->d_name);
- if (check_is_dir (NULL, (&orig_entry->d_stat), NULL)) {
+ if (prev == hashed_subvol)
+ goto list;
+ if ((hashed_subvol && dht_subvol_status(conf, hashed_subvol)) ||
+ (prev != local->first_up_subvol))
+ continue;
- /*Directory entries filtering :
- * a) If rebalance is running, pick from first_up_subvol
- * b) (rebalance not running)hashed subvolume is NULL or
- * down then filter in first_up_subvolume. Other wise the
- * corresponding hashed subvolume will take care of the
- * directory entry.
- */
- if (readdir_optimize) {
- if (prev == local->first_up_subvol)
- goto list;
- else
- continue;
+ goto list;
+ }
- }
+ list:
+ entry = gf_dirent_for_name(orig_entry->d_name);
+ if (!entry) {
+ goto unwind;
+ }
- hashed_subvol = methods->layout_search (this, layout,
- orig_entry->d_name);
+ /* Do this if conf->search_unhashed is set to "auto" */
+ if (conf->search_unhashed == GF_DHT_LOOKUP_UNHASHED_AUTO) {
+ subvol = methods->layout_search(this, layout, orig_entry->d_name);
+ if (!subvol || (subvol != prev)) {
+ /* TODO: Count the number of entries which need
+ linkfile to prove its existence in fs */
+ layout->search_unhashed++;
+ }
+ }
- if (prev == hashed_subvol)
- goto list;
- if ((hashed_subvol
- && dht_subvol_status (conf, hashed_subvol))
- || (prev != local->first_up_subvol))
- continue;
+ entry->d_off = orig_entry->d_off;
+ entry->d_stat = orig_entry->d_stat;
+ entry->d_ino = orig_entry->d_ino;
+ entry->d_type = orig_entry->d_type;
+ entry->d_len = orig_entry->d_len;
- goto list;
- }
+ if (orig_entry->dict)
+ entry->dict = dict_ref(orig_entry->dict);
- if (check_is_linkfile (NULL, (&orig_entry->d_stat),
- orig_entry->dict,
- conf->link_xattr_name)) {
- gf_msg_debug (this->name, 0, "%s: %s is a linkto file",
- prev->name, orig_entry->d_name);
- continue;
- }
+ /* making sure we set the inode ctx right with layout,
+ currently possible only for non-directories, so for
+ directories don't set entry inodes */
+ if (IA_ISDIR(entry->d_stat.ia_type)) {
+ entry->d_stat.ia_blocks = DHT_DIR_STAT_BLOCKS;
+ entry->d_stat.ia_size = DHT_DIR_STAT_SIZE;
+ if (orig_entry->inode) {
+ dht_inode_ctx_time_update(orig_entry->inode, this,
+ &entry->d_stat, 1);
-list:
- entry = gf_dirent_for_name (orig_entry->d_name);
- if (!entry) {
- goto unwind;
+ if (conf->subvolume_cnt == 1) {
+ dht_populate_inode_for_dentry(this, prev, entry,
+ orig_entry);
}
+ }
+ } else {
+ if (orig_entry->dict &&
+ dict_get(orig_entry->dict, conf->link_xattr_name)) {
+ /* Strip out the S and T flags set by rebalance*/
+ DHT_STRIP_PHASE1_FLAGS(&entry->d_stat);
+ }
- /* Do this if conf->search_unhashed is set to "auto" */
- if (conf->search_unhashed == GF_DHT_LOOKUP_UNHASHED_AUTO) {
- subvol = methods->layout_search (this, layout,
- orig_entry->d_name);
- if (!subvol || (subvol != prev)) {
- /* TODO: Count the number of entries which need
- linkfile to prove its existence in fs */
- layout->search_unhashed++;
- }
+ if (orig_entry->inode) {
+ ret = dht_layout_preset(this, prev, orig_entry->inode);
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ DHT_MSG_LAYOUT_SET_FAILED,
+ "failed to link the layout "
+ "in inode for %s",
+ orig_entry->d_name);
+
+ entry->inode = inode_ref(orig_entry->inode);
+ } else if (itable) {
+ /*
+ * orig_entry->inode might be null if any upper
+ * layer xlators below client set to null, to
+ * force a lookup on the inode even if the inode
+ * is present in the inode table. In that case
+ * we just update the ctx to make sure we didn't
+ * missed anything.
+ */
+ inode = inode_find(itable, orig_entry->d_stat.ia_gfid);
+ if (inode) {
+ ret = dht_layout_preset(this, prev, inode);
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ DHT_MSG_LAYOUT_SET_FAILED,
+ "failed to link the layout"
+ " in inode for %s",
+ orig_entry->d_name);
+ inode_unref(inode);
+ inode = NULL;
}
+ }
+ }
- entry->d_off = orig_entry->d_off;
- entry->d_stat = orig_entry->d_stat;
- entry->d_ino = orig_entry->d_ino;
- entry->d_type = orig_entry->d_type;
- entry->d_len = orig_entry->d_len;
-
- if (orig_entry->dict)
- entry->dict = dict_ref (orig_entry->dict);
-
- /* making sure we set the inode ctx right with layout,
- currently possible only for non-directories, so for
- directories don't set entry inodes */
- if (IA_ISDIR(entry->d_stat.ia_type)) {
- entry->d_stat.ia_blocks = DHT_DIR_STAT_BLOCKS;
- entry->d_stat.ia_size = DHT_DIR_STAT_SIZE;
- if (orig_entry->inode) {
- dht_inode_ctx_time_update (orig_entry->inode,
- this, &entry->d_stat,
- 1);
-
- if (conf->subvolume_cnt == 1) {
- dht_populate_inode_for_dentry (this,
- prev,
- entry,
- orig_entry);
- }
+ gf_msg_debug(this->name, 0, "%s: Adding entry = %s", prev->name,
+ entry->d_name);
- }
- } else {
- if (orig_entry->inode) {
- ret = dht_layout_preset (this, prev,
- orig_entry->inode);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_LAYOUT_SET_FAILED,
- "failed to link the layout "
- "in inode for %s",
- orig_entry->d_name);
-
- entry->inode = inode_ref (orig_entry->inode);
- } else if (itable) {
- /*
- * orig_entry->inode might be null if any upper
- * layer xlators below client set to null, to
- * force a lookup on the inode even if the inode
- * is present in the inode table. In that case
- * we just update the ctx to make sure we didn't
- * missed anything.
- */
- inode = inode_find (itable,
- orig_entry->d_stat.ia_gfid);
- if (inode) {
- ret = dht_layout_preset
- (this, prev,
- inode);
- if (ret)
- gf_msg (this->name,
- GF_LOG_WARNING, 0,
- DHT_MSG_LAYOUT_SET_FAILED,
- "failed to link the layout"
- " in inode for %s",
- orig_entry->d_name);
- inode_unref (inode);
- inode = NULL;
- }
- }
- }
+ list_add_tail(&entry->list, &entries.list);
+ count++;
+ }
- gf_msg_debug (this->name, 0, "%s: Adding entry = %s",
- prev->name, entry->d_name);
+done:
- list_add_tail (&entry->list, &local->entries.list);
- local->filled += gf_dirent_size (entry->d_name);
- count++;
- local->op_ret++;
+ /* We need to ensure that only the last subvolume's end-of-directory
+ * notification is respected so that directory reading does not stop
+ * before all subvolumes have been read. That could happen because the
+ * posix for each subvolume sends a ENOENT on end-of-directory but in
+ * distribute we're not concerned only with a posix's view of the
+ * directory but the aggregated namespace' view of the directory.
+ * Possible values:
+ * op_ret == 0 and op_errno != 0
+ * if op_errno != ENOENT : Error.Unwind.
+ * if op_errno == ENOENT : There are no more entries on this subvol.
+ * Move to the next one.
+ * op_ret > 0 and count == 0 :
+ * The subvol returned entries to dht but all were stripped out.
+ * For example, if they were linkto files or dirs where
+ * hashed_subvol != prev. Try to get some entries by winding
+ * to the next subvol. This can be dangerous if parallel readdir
+ * is enabled as it grows the stack.
+ *
+ * op_ret > 0 and count > 0:
+ * We found some entries. Unwind even if the buffer is not full.
+ *
+ */
+
+ op_ret = count;
+ if (count == 0) {
+ /* non-zero next_offset means that
+ * EOF is not yet hit on the current subvol
+ */
+ if ((next_offset == 0) || (op_errno == ENOENT)) {
+ next_offset = 0;
+ next_subvol = dht_subvol_next(this, prev);
+ } else {
+ next_subvol = prev;
}
-done:
- if ((op_ret == 0) && op_errno != ENOENT) {
- /* remaining buffer size is not enough to hold even one
- * dentry
- */
- goto unwind;
+ if (!next_subvol) {
+ goto unwind;
}
- if ((count == 0) || (local && (local->filled < local->size))) {
- if ((next_offset == 0) || (op_errno == ENOENT)) {
- next_offset = 0;
- next_subvol = dht_subvol_next (this, prev);
- } else {
- next_subvol = prev;
- }
-
- if (!next_subvol) {
- goto unwind;
- }
-
- if (conf->readdir_optimize == _gf_true) {
- if (next_subvol != local->first_up_subvol) {
- ret = dict_set_int32 (local->xattr,
- GF_READDIR_SKIP_DIRS, 1);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value"
- ":key = %s",
- GF_READDIR_SKIP_DIRS );
- } else {
- dict_del (local->xattr,
- GF_READDIR_SKIP_DIRS);
- }
- }
-
- STACK_WIND_COOKIE (frame, dht_readdirp_cbk, next_subvol,
- next_subvol, next_subvol->fops->readdirp,
- local->fd, (local->size - local->filled),
- next_offset, local->xattr);
- return 0;
+ if (conf->readdir_optimize == _gf_true) {
+ if (next_subvol != local->first_up_subvol) {
+ ret = dict_set_int32(local->xattr, GF_READDIR_SKIP_DIRS, 1);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary value"
+ ":key = %s",
+ GF_READDIR_SKIP_DIRS);
+ } else {
+ dict_del(local->xattr, GF_READDIR_SKIP_DIRS);
+ }
}
+ STACK_WIND_COOKIE(frame, dht_readdirp_cbk, next_subvol, next_subvol,
+ next_subvol->fops->readdirp, local->fd, local->size,
+ next_offset, local->xattr);
+ return 0;
+ }
+
unwind:
- /* We need to ensure that only the last subvolume's end-of-directory
- * notification is respected so that directory reading does not stop
- * before all subvolumes have been read. That could happen because the
- * posix for each subvolume sends a ENOENT on end-of-directory but in
- * distribute we're not concerned only with a posix's view of the
- * directory but the aggregated namespace' view of the directory.
- */
- if ((local->op_ret >= 0) && (prev != dht_last_up_subvol (this)))
- local->op_errno = 0;
+ /* We need to ensure that only the last subvolume's end-of-directory
+ * notification is respected so that directory reading does not stop
+ * before all subvolumes have been read. That could happen because the
+ * posix for each subvolume sends a ENOENT on end-of-directory but in
+ * distribute we're not concerned only with a posix's view of the
+ * directory but the aggregated namespace' view of the directory.
+ */
+ if (op_ret < 0)
+ op_ret = 0;
+ if (prev != dht_last_up_subvol(this))
+ op_errno = 0;
- DHT_STACK_UNWIND (readdirp, frame, local->op_ret, local->op_errno,
- &local->entries, NULL);
+ DHT_STACK_UNWIND(readdirp, frame, op_ret, op_errno, &entries, NULL);
- return 0;
+ gf_dirent_free(&entries);
+ return 0;
}
-
-int
-dht_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, gf_dirent_t *orig_entries,
- dict_t *xdata)
+static int
+dht_readdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, gf_dirent_t *orig_entries, dict_t *xdata)
{
- dht_local_t *local = NULL;
- gf_dirent_t *orig_entry = NULL;
- gf_dirent_t *entry = NULL;
- xlator_t *prev = NULL;
- xlator_t *next_subvol = NULL;
- off_t next_offset = 0;
- int count = 0;
- dht_layout_t *layout = 0;
- xlator_t *subvol = 0;
- dht_conf_t *conf = NULL;
- dht_methods_t *methods = NULL;
-
- prev = cookie;
- local = frame->local;
+ dht_local_t *local = NULL;
+ gf_dirent_t entries;
+ gf_dirent_t *orig_entry = NULL;
+ gf_dirent_t *entry = NULL;
+ xlator_t *prev = NULL;
+ xlator_t *next_subvol = NULL;
+ off_t next_offset = 0;
+ int count = 0;
+ dht_layout_t *layout = 0;
+ xlator_t *subvol = 0;
+ dht_conf_t *conf = NULL;
+ dht_methods_t *methods = NULL;
+ gf_boolean_t skip_hashed_check = _gf_false;
- conf = this->private;
- GF_VALIDATE_OR_GOTO (this->name, conf, done);
+ INIT_LIST_HEAD(&entries.list);
- methods = &(conf->methods);
+ prev = cookie;
+ local = frame->local;
- local->op_errno = op_errno;
+ conf = this->private;
+ GF_VALIDATE_OR_GOTO(this->name, conf, done);
- if (op_ret < 0) {
- goto done;
- }
+ methods = &(conf->methods);
- if (local->op_ret < 0)
- local->op_ret = 0;
+ if (op_ret <= 0)
+ goto done;
- if (!local->layout)
- local->layout = dht_layout_get (this, local->fd->inode);
+ if (!local->layout)
+ local->layout = dht_layout_get(this, local->fd->inode);
- layout = local->layout;
+ layout = local->layout;
- gf_msg_debug (this->name, 0, "Processing entries from %s",
- prev->name);
+ gf_msg_debug(this->name, 0, "Processing entries from %s", prev->name);
- list_for_each_entry (orig_entry, (&orig_entries->list), list) {
- next_offset = orig_entry->d_off;
+ if (conf->subvolume_cnt == 1) {
+ /*return everything*/
+ skip_hashed_check = _gf_true;
+ count = op_ret;
+ goto done;
+ }
- gf_msg_debug (this->name, 0, "%s: entry = %s, type = %d",
- prev->name, orig_entry->d_name,
- orig_entry->d_type);
+ list_for_each_entry(orig_entry, (&orig_entries->list), list)
+ {
+ next_offset = orig_entry->d_off;
- subvol = methods->layout_search (this, layout,
- orig_entry->d_name);
+ gf_msg_debug(this->name, 0, "%s: entry = %s, type = %d", prev->name,
+ orig_entry->d_name, orig_entry->d_type);
- if (!subvol || (subvol == prev)) {
- entry = gf_dirent_for_name (orig_entry->d_name);
- if (!entry) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_NO_MEMORY,
- "Memory allocation failed ");
- goto unwind;
- }
+ subvol = methods->layout_search(this, layout, orig_entry->d_name);
- entry->d_off = orig_entry->d_off;
- entry->d_ino = orig_entry->d_ino;
- entry->d_type = orig_entry->d_type;
- entry->d_len = orig_entry->d_len;
+ if (!subvol || (subvol == prev)) {
+ entry = gf_dirent_for_name(orig_entry->d_name);
+ if (!entry) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
+ "Memory allocation failed ");
+ goto unwind;
+ }
- gf_msg_debug (this->name, 0, "%s: Adding = entry %s",
- prev->name, entry->d_name);
+ entry->d_off = orig_entry->d_off;
+ entry->d_ino = orig_entry->d_ino;
+ entry->d_type = orig_entry->d_type;
+ entry->d_len = orig_entry->d_len;
- list_add_tail (&entry->list, &local->entries.list);
- count++;
- local->filled += gf_dirent_size (entry->d_name);
- local->op_ret++;
- }
- }
+ gf_msg_debug(this->name, 0, "%s: Adding = entry %s", prev->name,
+ entry->d_name);
+ list_add_tail(&entry->list, &entries.list);
+ count++;
+ }
+ }
done:
- if ((op_ret == 0) && op_errno != ENOENT) {
- /* remaining buffer size is not enough to hold even one
- * dentry
- */
- goto unwind;
+ op_ret = count;
+ /* We need to ensure that only the last subvolume's end-of-directory
+ * notification is respected so that directory reading does not stop
+ * before all subvolumes have been read. That could happen because the
+ * posix for each subvolume sends a ENOENT on end-of-directory but in
+ * distribute we're not concerned only with a posix's view of the
+ * directory but the aggregated namespace' view of the directory.
+ */
+ if (count == 0) {
+ if ((next_offset == 0) || (op_errno == ENOENT)) {
+ next_offset = 0;
+ next_subvol = dht_subvol_next(this, prev);
+ } else {
+ next_subvol = prev;
}
- if ((count == 0) || (local && (local->filled < local->size))) {
- if ((op_ret <= 0) || (op_errno == ENOENT)) {
- next_subvol = dht_subvol_next (this, prev);
- } else {
- next_subvol = prev;
- }
-
- if (!next_subvol) {
- goto unwind;
- }
-
- STACK_WIND_COOKIE (frame, dht_readdir_cbk, next_subvol,
- next_subvol, next_subvol->fops->readdir,
- local->fd, (local->size - local->filled),
- next_offset, NULL);
- return 0;
+ if (!next_subvol) {
+ goto unwind;
}
-unwind:
- /* We need to ensure that only the last subvolume's end-of-directory
- * notification is respected so that directory reading does not stop
- * before all subvolumes have been read. That could happen because the
- * posix for each subvolume sends a ENOENT on end-of-directory but in
- * distribute we're not concerned only with a posix's view of the
- * directory but the aggregated namespace' view of the directory.
- */
- if ((local->op_ret >= 0) && (prev != dht_last_up_subvol (this)))
- local->op_errno = 0;
-
-
- DHT_STACK_UNWIND (readdir, frame, local->op_ret, local->op_errno,
- &local->entries, NULL);
-
+ STACK_WIND_COOKIE(frame, dht_readdir_cbk, next_subvol, next_subvol,
+ next_subvol->fops->readdir, local->fd, local->size,
+ next_offset, NULL);
return 0;
-}
-
-
-int
-dht_do_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t yoff, int whichop, dict_t *dict)
-{
- dht_local_t *local = NULL;
- int op_errno = -1;
- xlator_t *xvol = NULL;
- int ret = 0;
- dht_conf_t *conf = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (this->private, err);
+ }
- conf = this->private;
-
- local = dht_local_init (frame, NULL, NULL, whichop);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local->fd = fd_ref (fd);
- local->size = size;
- local->xattr_req = (dict)? dict_ref (dict) : NULL;
- local->first_up_subvol = dht_first_up_subvol (this);
- local->op_ret = -1;
-
- dht_deitransform (this, yoff, &xvol);
-
- /* TODO: do proper readdir */
- if (whichop == GF_FOP_READDIRP) {
- if (dict)
- local->xattr = dict_ref (dict);
- else
- local->xattr = dict_new ();
+unwind:
+ /* We need to ensure that only the last subvolume's end-of-directory
+ * notification is respected so that directory reading does not stop
+ * before all subvolumes have been read. That could happen because the
+ * posix for each subvolume sends a ENOENT on end-of-directory but in
+ * distribute we're not concerned only with a posix's view of the
+ * directory but the aggregated namespace' view of the directory.
+ */
- if (local->xattr) {
- ret = dict_set_uint32 (local->xattr,
- conf->link_xattr_name, 256);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value"
- " : key = %s",
- conf->link_xattr_name);
+ if (prev != dht_last_up_subvol(this))
+ op_errno = 0;
- if (conf->readdir_optimize == _gf_true) {
- if (xvol != local->first_up_subvol) {
- ret = dict_set_int32 (local->xattr,
- GF_READDIR_SKIP_DIRS, 1);
- if (ret)
- gf_msg (this->name,
- GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set "
- "dictionary value: "
- "key = %s",
- GF_READDIR_SKIP_DIRS);
- } else {
- dict_del (local->xattr,
- GF_READDIR_SKIP_DIRS);
- }
- }
-
- if (conf->subvolume_cnt == 1) {
- ret = dict_set_uint32 (local->xattr,
- conf->xattr_name, 4 * 4);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING,
- ENOMEM,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary "
- "value:key = %s ",
- conf->xattr_name);
- }
- }
+ if (!skip_hashed_check) {
+ DHT_STACK_UNWIND(readdir, frame, op_ret, op_errno, &entries, NULL);
+ gf_dirent_free(&entries);
+ } else {
+ DHT_STACK_UNWIND(readdir, frame, op_ret, op_errno, orig_entries, NULL);
+ }
+ return 0;
+}
+static int
+dht_do_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t yoff, int whichop, dict_t *dict)
+{
+ dht_local_t *local = NULL;
+ int op_errno = -1;
+ xlator_t *xvol = NULL;
+ int ret = 0;
+ dht_conf_t *conf = NULL;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+ VALIDATE_OR_GOTO(this->private, err);
+
+ conf = this->private;
+
+ local = dht_local_init(frame, NULL, NULL, whichop);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->fd = fd_ref(fd);
+ local->size = size;
+ local->xattr_req = (dict) ? dict_ref(dict) : NULL;
+ local->first_up_subvol = dht_first_up_subvol(this);
+ local->op_ret = -1;
+
+ dht_deitransform(this, yoff, &xvol);
+
+ /* TODO: do proper readdir */
+ if (whichop == GF_FOP_READDIRP) {
+ if (dict)
+ local->xattr = dict_ref(dict);
+ else
+ local->xattr = dict_new();
+
+ if (local->xattr) {
+ ret = dict_set_uint32(local->xattr, conf->link_xattr_name, 256);
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary value"
+ " : key = %s",
+ conf->link_xattr_name);
+
+ if (conf->readdir_optimize == _gf_true) {
+ if (xvol != local->first_up_subvol) {
+ ret = dict_set_int32(local->xattr, GF_READDIR_SKIP_DIRS, 1);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ DHT_MSG_DICT_SET_FAILED,
+ "Failed to set "
+ "dictionary value: "
+ "key = %s",
+ GF_READDIR_SKIP_DIRS);
+ } else {
+ dict_del(local->xattr, GF_READDIR_SKIP_DIRS);
}
+ }
- STACK_WIND_COOKIE (frame, dht_readdirp_cbk, xvol, xvol,
- xvol->fops->readdirp, fd, size, yoff,
- local->xattr);
- } else {
- STACK_WIND_COOKIE (frame, dht_readdir_cbk, xvol, xvol,
- xvol->fops->readdir, fd, size, yoff,
- local->xattr);
+ if (conf->subvolume_cnt == 1) {
+ ret = dict_set_uint32(local->xattr, conf->xattr_name, 4 * 4);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM,
+ DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary "
+ "value:key = %s ",
+ conf->xattr_name);
+ }
+ }
}
- return 0;
+ STACK_WIND_COOKIE(frame, dht_readdirp_cbk, xvol, xvol,
+ xvol->fops->readdirp, fd, size, yoff, local->xattr);
+ } else {
+ STACK_WIND_COOKIE(frame, dht_readdir_cbk, xvol, xvol,
+ xvol->fops->readdir, fd, size, yoff, local->xattr);
+ }
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(readdir, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
-
int
-dht_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t yoff, dict_t *xdata)
+dht_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t yoff, dict_t *xdata)
{
- int op = GF_FOP_READDIR;
- dht_conf_t *conf = NULL;
- int i = 0;
+ int op = GF_FOP_READDIR;
+ dht_conf_t *conf = NULL;
+ int i = 0;
- conf = this->private;
- if (!conf)
- goto out;
+ conf = this->private;
+ if (!conf)
+ goto out;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (!conf->subvolume_status[i]) {
- op = GF_FOP_READDIRP;
- break;
- }
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (!conf->subvolume_status[i]) {
+ op = GF_FOP_READDIRP;
+ break;
}
+ }
- if (conf->use_readdirp)
- op = GF_FOP_READDIRP;
+ if (conf->use_readdirp)
+ op = GF_FOP_READDIRP;
out:
- dht_do_readdir (frame, this, fd, size, yoff, op, 0);
- return 0;
+ dht_do_readdir(frame, this, fd, size, yoff, op, 0);
+ return 0;
}
int
-dht_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t yoff, dict_t *dict)
+dht_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t yoff, dict_t *dict)
{
- dht_do_readdir (frame, this, fd, size, yoff, GF_FOP_READDIRP, dict);
- return 0;
+ dht_do_readdir(frame, this, fd, size, yoff, GF_FOP_READDIRP, dict);
+ return 0;
}
-
-
-int
-dht_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+static int
+dht_fsyncdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
-
-
- local = frame->local;
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
- LOCK (&frame->lock);
- {
- if (op_ret == -1)
- local->op_errno = op_errno;
+ local = frame->local;
- if (op_ret == 0)
- local->op_ret = 0;
- }
- UNLOCK (&frame->lock);
-
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt))
- DHT_STACK_UNWIND (fsyncdir, frame, local->op_ret,
- local->op_errno, xdata);
+ LOCK(&frame->lock);
+ {
+ if (op_ret == -1)
+ local->op_errno = op_errno;
+ else if (op_ret == 0)
+ local->op_ret = 0;
+ }
+ UNLOCK(&frame->lock);
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt))
+ DHT_STACK_UNWIND(fsyncdir, frame, local->op_ret, local->op_errno,
+ xdata);
- return 0;
+ return 0;
}
-
int
-dht_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int datasync, dict_t *xdata)
+dht_fsyncdir(call_frame_t *frame, xlator_t *this, fd_t *fd, int datasync,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- int op_errno = -1;
- int i = -1;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int op_errno = -1;
+ int i = -1;
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (this->private, err);
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+ VALIDATE_OR_GOTO(this->private, err);
- conf = this->private;
+ conf = this->private;
- local = dht_local_init (frame, NULL, NULL, GF_FOP_FSYNCDIR);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+ local = dht_local_init(frame, NULL, NULL, GF_FOP_FSYNCDIR);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- local->fd = fd_ref (fd);
- local->call_cnt = conf->subvolume_cnt;
+ local->fd = fd_ref(fd);
+ local->call_cnt = conf->subvolume_cnt;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- STACK_WIND (frame, dht_fsyncdir_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->fsyncdir,
- fd, datasync, xdata);
- }
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ STACK_WIND(frame, dht_fsyncdir_cbk, conf->subvolumes[i],
+ conf->subvolumes[i]->fops->fsyncdir, fd, datasync, xdata);
+ }
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fsyncdir, frame, -1, op_errno, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(fsyncdir, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
-
int
-dht_newfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+dht_newfile_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, inode_t *inode, struct iatt *stbuf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
{
- xlator_t *prev = NULL;
- int ret = -1;
- dht_local_t *local = NULL;
-
-
- if (op_ret == -1)
- goto out;
-
- local = frame->local;
- if (!local) {
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
+ xlator_t *prev = NULL;
+ int ret = -1;
+ dht_local_t *local = NULL;
- prev = cookie;
-
- if (local->loc.parent) {
-
- dht_inode_ctx_time_update (local->loc.parent, this,
- preparent, 0);
- dht_inode_ctx_time_update (local->loc.parent, this,
- postparent, 1);
- }
+ if (op_ret == -1)
+ goto out;
- ret = dht_layout_preset (this, prev, inode);
- if (ret < 0) {
- gf_msg_debug (this->name, EINVAL,
- "could not set pre-set layout for subvolume %s",
- prev? prev->name: NULL);
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
- if (local->linked == _gf_true)
- dht_linkfile_attr_heal (frame, this);
+ local = frame->local;
+ if (!local) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ prev = cookie;
+
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update(local->loc.parent, this, preparent, 0);
+ dht_inode_ctx_time_update(local->loc.parent, this, postparent, 1);
+ }
+
+ ret = dht_layout_preset(this, prev, inode);
+ if (ret < 0) {
+ gf_msg_debug(this->name, EINVAL,
+ "could not set pre-set layout for subvolume %s",
+ prev ? prev->name : NULL);
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto out;
+ }
+ if (local->linked == _gf_true)
+ dht_linkfile_attr_heal(frame, this);
out:
- /*
- * FIXME: ia_size and st_blocks of preparent and postparent do not have
- * correct values. since, preparent and postparent buffers correspond
- * to a directory these two members should have values equal to sum of
- * corresponding values from each of the subvolume.
- * See dht_iatt_merge for reference.
- */
- DHT_STRIP_PHASE1_FLAGS (stbuf);
- dht_set_fixed_dir_stat (postparent);
- dht_set_fixed_dir_stat (preparent);
-
- if (local && local->lock[0].layout.parent_layout.locks) {
- /* store op_errno for failure case*/
- local->op_errno = op_errno;
- local->refresh_layout_unlock (frame, this, op_ret, 1);
+ /*
+ * FIXME: ia_size and st_blocks of preparent and postparent do not have
+ * correct values. since, preparent and postparent buffers correspond
+ * to a directory these two members should have values equal to sum of
+ * corresponding values from each of the subvolume.
+ * See dht_iatt_merge for reference.
+ */
+ DHT_STRIP_PHASE1_FLAGS(stbuf);
+ dht_set_fixed_dir_stat(postparent);
+ dht_set_fixed_dir_stat(preparent);
+
+ if (local && local->lock[0].layout.parent_layout.locks) {
+ /* store op_errno for failure case*/
+ local->op_errno = op_errno;
+ local->refresh_layout_unlock(frame, this, op_ret, 1);
- if (op_ret == 0) {
- DHT_STACK_UNWIND (mknod, frame, op_ret, op_errno,
- inode, stbuf, preparent, postparent,
- xdata);
- }
- } else {
- DHT_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode,
- stbuf, preparent, postparent, xdata);
+ if (op_ret == 0) {
+ DHT_STACK_UNWIND(mknod, frame, op_ret, op_errno, inode, stbuf,
+ preparent, postparent, xdata);
}
+ } else {
+ DHT_STACK_UNWIND(mknod, frame, op_ret, op_errno, inode, stbuf,
+ preparent, postparent, xdata);
+ }
- return 0;
+ return 0;
}
-int
-dht_mknod_linkfile_create_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+static int
+dht_mknod_linkfile_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *cached_subvol = NULL;
- dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *cached_subvol = NULL;
+ dht_conf_t *conf = NULL;
- local = frame->local;
+ local = frame->local;
- if (!local || !local->cached_subvol) {
- op_errno = EINVAL;
- goto err;
- }
+ if (!local || !local->cached_subvol) {
+ op_errno = EINVAL;
+ goto err;
+ }
- if (op_ret == -1) {
- local->op_errno = op_errno;
- goto err;
- }
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ goto err;
+ }
- conf = this->private;
- if (!conf) {
- local->op_errno = EINVAL;
- op_errno = EINVAL;
- goto err;
- }
+ conf = this->private;
+ if (!conf) {
+ local->op_errno = EINVAL;
+ op_errno = EINVAL;
+ goto err;
+ }
- cached_subvol = local->cached_subvol;
+ cached_subvol = local->cached_subvol;
- if (local->params) {
- dict_del (local->params, conf->link_xattr_name);
- dict_del (local->params, GLUSTERFS_INTERNAL_FOP_KEY);
- }
+ if (local->params) {
+ dict_del(local->params, conf->link_xattr_name);
+ dict_del(local->params, GLUSTERFS_INTERNAL_FOP_KEY);
+ }
- STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)cached_subvol,
- cached_subvol, cached_subvol->fops->mknod,
- &local->loc, local->mode, local->rdev, local->umask,
- local->params);
+ STACK_WIND_COOKIE(frame, dht_newfile_cbk, (void *)cached_subvol,
+ cached_subvol, cached_subvol->fops->mknod, &local->loc,
+ local->mode, local->rdev, local->umask, local->params);
- return 0;
+ return 0;
err:
- if (local && local->lock[0].layout.parent_layout.locks) {
- local->refresh_layout_unlock (frame, this, -1, 1);
- } else {
- DHT_STACK_UNWIND (mknod, frame, -1,
- op_errno, NULL, NULL, NULL,
- NULL, NULL);
- }
- return 0;
+ if (local && local->lock[0].layout.parent_layout.locks) {
+ local->refresh_layout_unlock(frame, this, -1, 1);
+ } else {
+ DHT_STACK_UNWIND(mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
+ }
+ return 0;
}
-int
-dht_mknod_wind_to_avail_subvol (call_frame_t *frame, xlator_t *this,
- xlator_t *subvol, loc_t *loc, dev_t rdev,
- mode_t mode, mode_t umask, dict_t *params)
+static int
+dht_mknod_wind_to_avail_subvol(call_frame_t *frame, xlator_t *this,
+ xlator_t *subvol, loc_t *loc, dev_t rdev,
+ mode_t mode, mode_t umask, dict_t *params)
{
- dht_local_t *local = NULL;
- xlator_t *avail_subvol = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *avail_subvol = NULL;
- local = frame->local;
+ local = frame->local;
- if (!dht_is_subvol_filled (this, subvol)) {
- gf_msg_debug (this->name, 0,
- "creating %s on %s", loc->path,
- subvol->name);
+ if (!dht_is_subvol_filled(this, subvol)) {
+ gf_msg_debug(this->name, 0, "creating %s on %s", loc->path,
+ subvol->name);
- STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol,
- subvol, subvol->fops->mknod, loc, mode,
- rdev, umask, params);
- } else {
- avail_subvol = dht_free_disk_available_subvol (this, subvol, local);
+ STACK_WIND_COOKIE(frame, dht_newfile_cbk, (void *)subvol, subvol,
+ subvol->fops->mknod, loc, mode, rdev, umask, params);
+ } else {
+ avail_subvol = dht_free_disk_available_subvol(this, subvol, local);
- if (avail_subvol != subvol) {
- local->params = dict_ref (params);
- local->rdev = rdev;
- local->mode = mode;
- local->umask = umask;
- local->cached_subvol = avail_subvol;
- local->hashed_subvol = subvol;
+ if (avail_subvol != subvol) {
+ local->params = dict_ref(params);
+ local->rdev = rdev;
+ local->mode = mode;
+ local->umask = umask;
+ local->cached_subvol = avail_subvol;
+ local->hashed_subvol = subvol;
- gf_msg_debug (this->name, 0,
- "creating %s on %s (link at %s)", loc->path,
- avail_subvol->name, subvol->name);
+ gf_msg_debug(this->name, 0, "creating %s on %s (link at %s)",
+ loc->path, avail_subvol->name, subvol->name);
- dht_linkfile_create (frame,
- dht_mknod_linkfile_create_cbk,
- this, avail_subvol, subvol, loc);
+ dht_linkfile_create(frame, dht_mknod_linkfile_create_cbk, this,
+ avail_subvol, subvol, loc);
- goto out;
- }
-
- gf_msg_debug (this->name, 0,
- "creating %s on %s", loc->path, subvol->name);
+ goto out;
+ }
- STACK_WIND_COOKIE (frame, dht_newfile_cbk,
- (void *)subvol, subvol,
- subvol->fops->mknod, loc, mode,
- rdev, umask, params);
+ gf_msg_debug(this->name, 0, "creating %s on %s", loc->path,
+ subvol->name);
- }
+ STACK_WIND_COOKIE(frame, dht_newfile_cbk, (void *)subvol, subvol,
+ subvol->fops->mknod, loc, mode, rdev, umask, params);
+ }
out:
- return 0;
+ return 0;
}
-int32_t
-dht_mknod_do (call_frame_t *frame)
+static int32_t
+dht_mknod_do(call_frame_t *frame)
{
- dht_local_t *local = NULL;
- dht_layout_t *refreshed = NULL;
- xlator_t *subvol = NULL;
- xlator_t *this = NULL;
- dht_conf_t *conf = NULL;
- dht_methods_t *methods = NULL;
+ dht_local_t *local = NULL;
+ dht_layout_t *refreshed = NULL;
+ xlator_t *subvol = NULL;
+ xlator_t *this = NULL;
+ dht_conf_t *conf = NULL;
+ dht_methods_t *methods = NULL;
- local = frame->local;
+ local = frame->local;
- this = THIS;
+ this = THIS;
- conf = this->private;
+ conf = this->private;
- GF_VALIDATE_OR_GOTO (this->name, conf, err);
+ GF_VALIDATE_OR_GOTO(this->name, conf, err);
- methods = &(conf->methods);
+ methods = &(conf->methods);
- /* We don't need parent_loc anymore */
- loc_wipe (&local->loc);
+ /* We don't need parent_loc anymore */
+ loc_wipe(&local->loc);
- loc_copy (&local->loc, &local->loc2);
+ loc_copy(&local->loc, &local->loc2);
- loc_wipe (&local->loc2);
+ loc_wipe(&local->loc2);
- refreshed = local->selfheal.refreshed_layout;
+ refreshed = local->selfheal.refreshed_layout;
- subvol = methods->layout_search (this, refreshed, local->loc.name);
+ subvol = methods->layout_search(this, refreshed, local->loc.name);
- if (!subvol) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_HASHED_SUBVOL_GET_FAILED, "no subvolume in "
- "layout for path=%s", local->loc.path);
- local->op_errno = ENOENT;
- goto err;
- }
+ if (!subvol) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_HASHED_SUBVOL_GET_FAILED,
+ "no subvolume in "
+ "layout for path=%s",
+ local->loc.path);
+ local->op_errno = ENOENT;
+ goto err;
+ }
- dht_mknod_wind_to_avail_subvol (frame, this, subvol, &local->loc,
- local->rdev, local->mode,
- local->umask, local->params);
- return 0;
+ dht_mknod_wind_to_avail_subvol(frame, this, subvol, &local->loc,
+ local->rdev, local->mode, local->umask,
+ local->params);
+ return 0;
err:
- local->refresh_layout_unlock (frame, this, -1, 1);
+ local->refresh_layout_unlock(frame, this, -1, 1);
- return 0;
+ return 0;
}
-
-int32_t
-dht_mknod_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+static int32_t
+dht_mknod_unlock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- DHT_STACK_DESTROY (frame);
- return 0;
+ DHT_STACK_DESTROY(frame);
+ return 0;
}
-int32_t
-dht_mknod_finish (call_frame_t *frame, xlator_t *this, int op_ret,
- int invoke_cbk)
+static int32_t
+dht_mknod_finish(call_frame_t *frame, xlator_t *this, int op_ret,
+ int invoke_cbk)
{
- dht_local_t *local = NULL, *lock_local = NULL;
- call_frame_t *lock_frame = NULL;
- int lock_count = 0;
+ dht_local_t *local = NULL, *lock_local = NULL;
+ call_frame_t *lock_frame = NULL;
+ int lock_count = 0;
- local = frame->local;
- lock_count = dht_lock_count (local->lock[0].layout.parent_layout.locks,
- local->lock[0].layout.parent_layout.lk_count);
- if (lock_count == 0)
- goto done;
+ local = frame->local;
+ lock_count = dht_lock_count(local->lock[0].layout.parent_layout.locks,
+ local->lock[0].layout.parent_layout.lk_count);
+ if (lock_count == 0)
+ goto done;
- lock_frame = copy_frame (frame);
- if (lock_frame == NULL) {
- goto done;
- }
+ lock_frame = copy_frame(frame);
+ if (lock_frame == NULL) {
+ goto done;
+ }
- lock_local = dht_local_init (lock_frame, &local->loc, NULL,
- lock_frame->root->op);
- if (lock_local == NULL) {
- goto done;
- }
+ lock_local = dht_local_init(lock_frame, &local->loc, NULL,
+ lock_frame->root->op);
+ if (lock_local == NULL) {
+ goto done;
+ }
- lock_local->lock[0].layout.parent_layout.locks = local->lock[0].layout.parent_layout.locks;
- lock_local->lock[0].layout.parent_layout.lk_count = local->lock[0].layout.parent_layout.lk_count;
+ lock_local->lock[0]
+ .layout.parent_layout.locks = local->lock[0].layout.parent_layout.locks;
+ lock_local->lock[0].layout.parent_layout.lk_count =
+ local->lock[0].layout.parent_layout.lk_count;
- local->lock[0].layout.parent_layout.locks = NULL;
- local->lock[0].layout.parent_layout.lk_count = 0;
+ local->lock[0].layout.parent_layout.locks = NULL;
+ local->lock[0].layout.parent_layout.lk_count = 0;
- dht_unlock_inodelk (lock_frame,
- lock_local->lock[0].layout.parent_layout.locks,
- lock_local->lock[0].layout.parent_layout.lk_count,
- dht_mknod_unlock_cbk);
- lock_frame = NULL;
+ dht_unlock_inodelk(lock_frame,
+ lock_local->lock[0].layout.parent_layout.locks,
+ lock_local->lock[0].layout.parent_layout.lk_count,
+ dht_mknod_unlock_cbk);
+ lock_frame = NULL;
done:
- if (lock_frame != NULL) {
- DHT_STACK_DESTROY (lock_frame);
- }
-
- if (op_ret == 0)
- return 0;
+ if (lock_frame != NULL) {
+ DHT_STACK_DESTROY(lock_frame);
+ }
- DHT_STACK_UNWIND (mknod, frame, op_ret, local->op_errno, NULL, NULL,
- NULL, NULL, NULL);
+ if (op_ret == 0)
return 0;
+
+ DHT_STACK_UNWIND(mknod, frame, op_ret, local->op_errno, NULL, NULL, NULL,
+ NULL, NULL);
+ return 0;
}
-int32_t
-dht_mknod_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+static int32_t
+dht_mknod_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
+ dht_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (!local) {
- goto err;
- }
+ if (!local) {
+ goto err;
+ }
- if (op_ret < 0) {
- gf_msg ("DHT", GF_LOG_ERROR, 0, DHT_MSG_INODE_LK_ERROR,
- "mknod lock failed for file: %s", local->loc2.name);
+ if (op_ret < 0) {
+ gf_msg("DHT", GF_LOG_ERROR, 0, DHT_MSG_INODE_LK_ERROR,
+ "mknod lock failed for file: %s", local->loc2.name);
- local->op_errno = op_errno;
+ local->op_errno = op_errno;
- goto err;
- }
+ goto err;
+ }
- local->refresh_layout_unlock = dht_mknod_finish;
+ local->refresh_layout_unlock = dht_mknod_finish;
- local->refresh_layout_done = dht_mknod_do;
+ local->refresh_layout_done = dht_mknod_do;
- dht_refresh_layout (frame);
+ dht_refresh_layout(frame);
- return 0;
+ return 0;
err:
- dht_mknod_finish (frame, this, -1, 0);
- return 0;
+ if (local)
+ dht_mknod_finish(frame, this, -1, 0);
+ else
+ DHT_STACK_UNWIND(mknod, frame, -1, EINVAL, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
-int32_t
-dht_mknod_lock (call_frame_t *frame, xlator_t *subvol)
+static int32_t
+dht_mknod_lock(call_frame_t *frame, xlator_t *subvol)
{
- dht_local_t *local = NULL;
- int count = 1, ret = -1;
- dht_lock_t **lk_array = NULL;
+ dht_local_t *local = NULL;
+ int count = 1, ret = -1;
+ dht_lock_t **lk_array = NULL;
- GF_VALIDATE_OR_GOTO ("dht", frame, err);
- GF_VALIDATE_OR_GOTO (frame->this->name, frame->local, err);
+ GF_VALIDATE_OR_GOTO("dht", frame, err);
+ GF_VALIDATE_OR_GOTO(frame->this->name, frame->local, err);
- local = frame->local;
+ local = frame->local;
- lk_array = GF_CALLOC (count, sizeof (*lk_array), gf_common_mt_pointer);
+ lk_array = GF_CALLOC(count, sizeof(*lk_array), gf_common_mt_pointer);
- if (lk_array == NULL)
- goto err;
+ if (lk_array == NULL)
+ goto err;
- lk_array[0] = dht_lock_new (frame->this, subvol, &local->loc, F_RDLCK,
- DHT_LAYOUT_HEAL_DOMAIN, NULL,
- IGNORE_ENOENT_ESTALE);
+ lk_array[0] = dht_lock_new(frame->this, subvol, &local->loc, F_RDLCK,
+ DHT_LAYOUT_HEAL_DOMAIN, NULL,
+ IGNORE_ENOENT_ESTALE);
- if (lk_array[0] == NULL)
- goto err;
+ if (lk_array[0] == NULL)
+ goto err;
- local->lock[0].layout.parent_layout.locks = lk_array;
- local->lock[0].layout.parent_layout.lk_count = count;
+ local->lock[0].layout.parent_layout.locks = lk_array;
+ local->lock[0].layout.parent_layout.lk_count = count;
- ret = dht_blocking_inodelk (frame, lk_array, count, dht_mknod_lock_cbk);
+ ret = dht_blocking_inodelk(frame, lk_array, count, dht_mknod_lock_cbk);
- if (ret < 0) {
- local->lock[0].layout.parent_layout.locks = NULL;
- local->lock[0].layout.parent_layout.lk_count = 0;
- goto err;
- }
+ if (ret < 0) {
+ local->lock[0].layout.parent_layout.locks = NULL;
+ local->lock[0].layout.parent_layout.lk_count = 0;
+ goto err;
+ }
- return 0;
+ return 0;
err:
- if (lk_array != NULL) {
- dht_lock_array_free (lk_array, count);
- GF_FREE (lk_array);
- }
+ if (lk_array != NULL) {
+ dht_lock_array_free(lk_array, count);
+ GF_FREE(lk_array);
+ }
- return -1;
+ return -1;
}
-int
-dht_refresh_parent_layout_resume (call_frame_t *frame, xlator_t *this, int ret,
- int invoke_cbk)
+static int
+dht_refresh_parent_layout_resume(call_frame_t *frame, xlator_t *this, int ret,
+ int invoke_cbk)
{
- dht_local_t *local = NULL, *parent_local = NULL;
- call_stub_t *stub = NULL;
- call_frame_t *parent_frame = NULL;
+ dht_local_t *local = NULL, *parent_local = NULL;
+ call_stub_t *stub = NULL;
+ call_frame_t *parent_frame = NULL;
- local = frame->local;
+ local = frame->local;
- stub = local->stub;
- local->stub = NULL;
+ stub = local->stub;
+ local->stub = NULL;
- parent_frame = stub->frame;
- parent_local = parent_frame->local;
+ parent_frame = stub->frame;
+ parent_local = parent_frame->local;
- if (ret < 0) {
- parent_local->op_ret = -1;
- parent_local->op_errno = local->op_errno
- ? local->op_errno : EIO;
- } else {
- parent_local->op_ret = 0;
- }
+ if (ret < 0) {
+ parent_local->op_ret = -1;
+ parent_local->op_errno = local->op_errno ? local->op_errno : EIO;
+ } else {
+ parent_local->op_ret = 0;
+ }
- call_resume (stub);
+ call_resume(stub);
- DHT_STACK_DESTROY (frame);
+ DHT_STACK_DESTROY(frame);
- return 0;
+ return 0;
}
-
-int
-dht_refresh_parent_layout_done (call_frame_t *frame)
+static int
+dht_refresh_parent_layout_done(call_frame_t *frame)
{
- dht_local_t *local = NULL;
- int ret = 0;
+ dht_local_t *local = NULL;
+ int ret = 0;
- local = frame->local;
+ local = frame->local;
- if (local->op_ret < 0) {
- ret = -1;
- goto resume;
- }
+ if (local->op_ret < 0) {
+ ret = -1;
+ goto resume;
+ }
- dht_layout_set (frame->this, local->loc.inode,
- local->selfheal.refreshed_layout);
+ dht_layout_set(frame->this, local->loc.inode,
+ local->selfheal.refreshed_layout);
resume:
- dht_refresh_parent_layout_resume (frame, frame->this, ret, 1);
- return 0;
+ dht_refresh_parent_layout_resume(frame, frame->this, ret, 1);
+ return 0;
}
-
-int
-dht_handle_parent_layout_change (xlator_t *this, call_stub_t *stub)
+static int
+dht_handle_parent_layout_change(xlator_t *this, call_stub_t *stub)
{
- call_frame_t *refresh_frame = NULL, *frame = NULL;
- dht_local_t *refresh_local = NULL, *local = NULL;
+ call_frame_t *refresh_frame = NULL, *frame = NULL;
+ dht_local_t *refresh_local = NULL, *local = NULL;
- frame = stub->frame;
- local = frame->local;
+ frame = stub->frame;
+ local = frame->local;
+
+ refresh_frame = copy_frame(frame);
+ if (!refresh_frame) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
+ "mem allocation failed for refresh_frame");
+ return -1;
+ }
- refresh_frame = copy_frame (frame);
- refresh_local = dht_local_init (refresh_frame, NULL, NULL,
- stub->fop);
+ refresh_local = dht_local_init(refresh_frame, NULL, NULL, stub->fop);
+ if (!refresh_local) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
+ "mem allocation failed for refresh_local");
+ return -1;
+ }
- refresh_local->loc.inode = inode_ref (local->loc.parent);
- gf_uuid_copy (refresh_local->loc.gfid, local->loc.parent->gfid);
+ refresh_local->loc.inode = inode_ref(local->loc.parent);
+ gf_uuid_copy(refresh_local->loc.gfid, local->loc.parent->gfid);
- refresh_local->stub = stub;
+ refresh_local->stub = stub;
- refresh_local->refresh_layout_unlock = dht_refresh_parent_layout_resume;
- refresh_local->refresh_layout_done = dht_refresh_parent_layout_done;
+ refresh_local->refresh_layout_unlock = dht_refresh_parent_layout_resume;
+ refresh_local->refresh_layout_done = dht_refresh_parent_layout_done;
- dht_refresh_layout (refresh_frame);
- return 0;
+ dht_refresh_layout(refresh_frame);
+ return 0;
}
-int32_t
-dht_call_mkdir_stub (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+static int32_t
+dht_call_mkdir_stub(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- call_stub_t *stub = NULL;
+ dht_local_t *local = NULL;
+ call_stub_t *stub = NULL;
- local = frame->local;
- stub = local->stub;
- local->stub = NULL;
+ local = frame->local;
+ stub = local->stub;
+ local->stub = NULL;
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = op_errno;
- } else {
- local->op_ret = 0;
- }
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ } else {
+ local->op_ret = 0;
+ }
- call_resume (stub);
+ call_resume(stub);
- return 0;
+ return 0;
}
-int32_t
-dht_guard_parent_layout_and_namespace (xlator_t *subvol, call_stub_t *stub)
+static int32_t
+dht_guard_parent_layout_and_namespace(xlator_t *subvol, call_stub_t *stub)
{
- dht_local_t *local = NULL;
- int ret = -1;
- loc_t *loc = NULL;
- xlator_t *hashed_subvol = NULL, *this = NULL;;
- call_frame_t *frame = NULL;
- char pgfid[GF_UUID_BUF_SIZE] = {0};
- int32_t *parent_disk_layout = NULL;
- dht_layout_t *parent_layout = NULL;
- dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ int ret = -1;
+ loc_t *loc = NULL;
+ xlator_t *hashed_subvol = NULL, *this = NULL;
+ ;
+ call_frame_t *frame = NULL;
+ char pgfid[GF_UUID_BUF_SIZE] = {0};
+ int32_t *parent_disk_layout = NULL;
+ dht_layout_t *parent_layout = NULL;
+ dht_conf_t *conf = NULL;
- GF_VALIDATE_OR_GOTO ("dht", stub, err);
+ GF_VALIDATE_OR_GOTO("dht", stub, err);
- frame = stub->frame;
- this = frame->this;
+ frame = stub->frame;
+ this = frame->this;
- conf = this->private;
+ conf = this->private;
- local = frame->local;
+ local = frame->local;
- local->stub = stub;
+ local->stub = stub;
- /* TODO: recheck whether we should lock on src or dst if we do similar
- * stale layout checks for rename.
- */
- loc = &stub->args.loc;
+ /* TODO: recheck whether we should lock on src or dst if we do similar
+ * stale layout checks for rename.
+ */
+ loc = &stub->args.loc;
- gf_uuid_unparse (loc->parent->gfid, pgfid);
+ gf_uuid_unparse(loc->parent->gfid, pgfid);
+ if (local->params == NULL) {
+ local->params = dict_new();
if (local->params == NULL) {
- local->params = dict_new ();
- if (local->params == NULL) {
- local->op_errno = ENOMEM;
- gf_msg (this->name, GF_LOG_WARNING, local->op_errno,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "%s (%s/%s) (path: %s): "
- "dict allocation failed",
- gf_fop_list[stub->fop],
- pgfid, loc->name, loc->path);
- goto err;
- }
- }
-
- hashed_subvol = dht_subvol_get_hashed (this, loc);
- if (hashed_subvol == NULL) {
- local->op_errno = EINVAL;
-
- gf_msg (this->name, GF_LOG_WARNING, local->op_errno,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "%s (%s/%s) (path: %s): "
- "hashed subvolume not found", gf_fop_list[stub->fop],
- pgfid, loc->name, loc->path);
- goto err;
- }
-
- parent_layout = dht_layout_get (this, loc->parent);
-
- ret = dht_disk_layout_extract_for_subvol (this, parent_layout,
- hashed_subvol,
- &parent_disk_layout);
- if (ret == -1) {
- local->op_errno = EINVAL;
- gf_msg (this->name, GF_LOG_WARNING, local->op_errno,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "%s (%s/%s) (path: %s): "
- "extracting in-memory layout of parent failed. ",
- gf_fop_list[stub->fop], pgfid, loc->name, loc->path);
- goto err;
- }
-
- memcpy ((void *)local->parent_disk_layout, (void *)parent_disk_layout,
- sizeof (local->parent_disk_layout));
-
- dht_layout_unref (this, parent_layout);
- parent_layout = NULL;
-
- ret = dict_set_str (local->params, GF_PREOP_PARENT_KEY,
- conf->xattr_name);
- if (ret < 0) {
- local->op_errno = -ret;
- gf_msg (this->name, GF_LOG_WARNING, local->op_errno,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "%s (%s/%s) (path: %s): "
- "setting %s key in params dictionary failed. ",
- gf_fop_list[stub->fop], pgfid, loc->name, loc->path,
- GF_PREOP_PARENT_KEY);
- goto err;
- }
-
- ret = dict_set_bin (local->params, conf->xattr_name, parent_disk_layout,
- 4 * 4);
- if (ret < 0) {
- local->op_errno = -ret;
- gf_msg (this->name, GF_LOG_WARNING, local->op_errno,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "%s (%s/%s) (path: %s): "
- "setting parent-layout in params dictionary failed. ",
- gf_fop_list[stub->fop], pgfid, loc->name, loc->path);
- goto err;
- }
-
- parent_disk_layout = NULL;
- local->hashed_subvol = hashed_subvol;
-
- local->current = &local->lock[0];
- ret = dht_protect_namespace (frame, loc, hashed_subvol,
- &local->current->ns, dht_call_mkdir_stub);
- if (ret < 0)
- goto err;
-
- return 0;
+ local->op_errno = ENOMEM;
+ gf_msg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "%s (%s/%s) (path: %s): "
+ "dict allocation failed",
+ gf_fop_list[stub->fop], pgfid, loc->name, loc->path);
+ goto err;
+ }
+ }
+
+ hashed_subvol = dht_subvol_get_hashed(this, loc);
+ if (hashed_subvol == NULL) {
+ local->op_errno = EINVAL;
+
+ gf_msg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "%s (%s/%s) (path: %s): "
+ "hashed subvolume not found",
+ gf_fop_list[stub->fop], pgfid, loc->name, loc->path);
+ goto err;
+ }
+
+ parent_layout = dht_layout_get(this, loc->parent);
+
+ ret = dht_disk_layout_extract_for_subvol(this, parent_layout, hashed_subvol,
+ &parent_disk_layout);
+ if (ret == -1) {
+ local->op_errno = EINVAL;
+ gf_msg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "%s (%s/%s) (path: %s): "
+ "extracting in-memory layout of parent failed. ",
+ gf_fop_list[stub->fop], pgfid, loc->name, loc->path);
+ goto err;
+ }
+
+ memcpy((void *)local->parent_disk_layout, (void *)parent_disk_layout,
+ sizeof(local->parent_disk_layout));
+
+ dht_layout_unref(this, parent_layout);
+ parent_layout = NULL;
+
+ ret = dict_set_str(local->params, GF_PREOP_PARENT_KEY, conf->xattr_name);
+ if (ret < 0) {
+ local->op_errno = -ret;
+ gf_msg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "%s (%s/%s) (path: %s): "
+ "setting %s key in params dictionary failed. ",
+ gf_fop_list[stub->fop], pgfid, loc->name, loc->path,
+ GF_PREOP_PARENT_KEY);
+ goto err;
+ }
+
+ ret = dict_set_bin(local->params, conf->xattr_name, parent_disk_layout,
+ 4 * 4);
+ if (ret < 0) {
+ local->op_errno = -ret;
+ gf_msg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "%s (%s/%s) (path: %s): "
+ "setting parent-layout in params dictionary failed. ",
+ gf_fop_list[stub->fop], pgfid, loc->name, loc->path);
+ goto err;
+ }
+
+ parent_disk_layout = NULL;
+ local->hashed_subvol = hashed_subvol;
+
+ local->current = &local->lock[0];
+ ret = dht_protect_namespace(frame, loc, hashed_subvol, &local->current->ns,
+ dht_call_mkdir_stub);
+ if (ret < 0)
+ goto err;
+
+ return 0;
err:
- if (parent_disk_layout != NULL)
- GF_FREE (parent_disk_layout);
+ if (parent_disk_layout != NULL)
+ GF_FREE(parent_disk_layout);
- if (parent_layout != NULL)
- dht_layout_unref (this, parent_layout);
+ if (parent_layout != NULL)
+ dht_layout_unref(this, parent_layout);
- return -1;
+ return -1;
}
int
-dht_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t rdev, mode_t umask, dict_t *params)
-{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- int i = 0;
- int ret = 0;
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
-
- conf = this->private;
-
- dht_get_du_info (frame, this, loc);
-
- local = dht_local_init (frame, loc, NULL, GF_FOP_MKNOD);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- subvol = dht_subvol_get_hashed (this, loc);
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no subvolume in layout for path=%s",
- loc->path);
- op_errno = EIO;
- goto err;
- }
-
- /* Post remove-brick, the client layout may not be in sync with
- * disk layout because of lack of lookup. Hence,a mknod call
- * may fall on the decommissioned brick. Hence, if the
- * hashed_subvol is part of decommissioned bricks list, do a
- * lookup on parent dir. If a fix-layout is already done by the
- * remove-brick process, the parent directory layout will be in
- * sync with that of the disk. If fix-layout is still ending
- * on the parent directory, we can let the file get created on
- * the decommissioned brick which will be eventually migrated to
- * non-decommissioned brick based on the new layout.
- */
+dht_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *params)
+{
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ int i = 0;
+ int ret = 0;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+
+ conf = this->private;
+
+ dht_get_du_info(frame, this, loc);
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_MKNOD);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ subvol = dht_subvol_get_hashed(this, loc);
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no subvolume in layout for path=%s",
+ loc->path);
+ op_errno = EIO;
+ goto err;
+ }
+
+ /* Post remove-brick, the client layout may not be in sync with
+ * disk layout because of lack of lookup. Hence,a mknod call
+ * may fall on the decommissioned brick. Hence, if the
+ * hashed_subvol is part of decommissioned bricks list, do a
+ * lookup on parent dir. If a fix-layout is already done by the
+ * remove-brick process, the parent directory layout will be in
+ * sync with that of the disk. If fix-layout is still ending
+ * on the parent directory, we can let the file get created on
+ * the decommissioned brick which will be eventually migrated to
+ * non-decommissioned brick based on the new layout.
+ */
+
+ if (conf->decommission_subvols_cnt) {
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->decommissioned_bricks[i] &&
+ conf->decommissioned_bricks[i] == subvol) {
+ gf_msg_debug(this->name, 0,
+ "hashed subvol:%s is "
+ "part of decommission brick list for "
+ "file: %s",
+ subvol->name, loc->path);
+
+ /* dht_refresh_layout needs directory info in
+ * local->loc. Hence, storing the parent_loc in
+ * local->loc and storing the create context in
+ * local->loc2. We will restore this information
+ * in dht_creation do */
+
+ ret = loc_copy(&local->loc2, &local->loc);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
+ "loc_copy failed %s", loc->path);
- if (conf->decommission_subvols_cnt) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->decommissioned_bricks[i] &&
- conf->decommissioned_bricks[i] == subvol) {
-
- gf_msg_debug (this->name, 0, "hashed subvol:%s is "
- "part of decommission brick list for "
- "file: %s", subvol->name, loc->path);
-
- /* dht_refresh_layout needs directory info in
- * local->loc. Hence, storing the parent_loc in
- * local->loc and storing the create context in
- * local->loc2. We will restore this information
- * in dht_creation do */
-
- ret = loc_copy (&local->loc2, &local->loc);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_NO_MEMORY,
- "loc_copy failed %s", loc->path);
-
- goto err;
- }
+ goto err;
+ }
- local->params = dict_ref (params);
- local->rdev = rdev;
- local->mode = mode;
- local->umask = umask;
+ local->params = dict_ref(params);
+ local->rdev = rdev;
+ local->mode = mode;
+ local->umask = umask;
- loc_wipe (&local->loc);
+ loc_wipe(&local->loc);
- ret = dht_build_parent_loc (this, &local->loc, loc,
- &op_errno);
+ ret = dht_build_parent_loc(this, &local->loc, loc, &op_errno);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_LOC_FAILED,
- "parent loc build failed");
- goto err;
- }
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_LOC_FAILED,
+ "parent loc build failed");
+ goto err;
+ }
- ret = dht_mknod_lock (frame, subvol);
+ ret = dht_mknod_lock(frame, subvol);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_INODE_LK_ERROR,
- "locking parent failed");
- goto err;
- }
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_INODE_LK_ERROR,
+ "locking parent failed");
+ goto err;
+ }
- goto done;
- }
+ goto done;
}
}
+ }
- dht_mknod_wind_to_avail_subvol (frame, this, subvol, loc, rdev, mode,
- umask, params);
+ dht_mknod_wind_to_avail_subvol(frame, this, subvol, loc, rdev, mode, umask,
+ params);
done:
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
- NULL, NULL, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
int
-dht_symlink (call_frame_t *frame, xlator_t *this,
- const char *linkname, loc_t *loc, mode_t umask, dict_t *params)
+dht_symlink(call_frame_t *frame, xlator_t *this, const char *linkname,
+ loc_t *loc, mode_t umask, dict_t *params)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
- local = dht_local_init (frame, loc, NULL, GF_FOP_SYMLINK);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+ local = dht_local_init(frame, loc, NULL, GF_FOP_SYMLINK);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- subvol = dht_subvol_get_hashed (this, loc);
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no subvolume in layout for path=%s",
- loc->path);
- op_errno = EIO;
- goto err;
- }
+ subvol = dht_subvol_get_hashed(this, loc);
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no subvolume in layout for path=%s",
+ loc->path);
+ op_errno = EIO;
+ goto err;
+ }
- gf_msg_trace (this->name, 0,
- "creating %s on %s", loc->path, subvol->name);
+ gf_msg_trace(this->name, 0, "creating %s on %s", loc->path, subvol->name);
- STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol, subvol,
- subvol->fops->symlink, linkname, loc, umask,
- params);
+ STACK_WIND_COOKIE(frame, dht_newfile_cbk, (void *)subvol, subvol,
+ subvol->fops->symlink, linkname, loc, umask, params);
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (link, frame, -1, op_errno,
- NULL, NULL, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
int
-dht_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
- dict_t *xdata)
+dht_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
{
- xlator_t *cached_subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
+ xlator_t *cached_subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
- local = dht_local_init (frame, loc, NULL, GF_FOP_UNLINK);
- if (!local) {
- op_errno = ENOMEM;
+ local = dht_local_init(frame, loc, NULL, GF_FOP_UNLINK);
+ if (!local) {
+ op_errno = ENOMEM;
- goto err;
- }
+ goto err;
+ }
- cached_subvol = local->cached_subvol;
- if (!cached_subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
+ cached_subvol = local->cached_subvol;
+ if (!cached_subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for path=%s",
+ loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
- local->flags = xflag;
- STACK_WIND_COOKIE (frame, dht_unlink_cbk, cached_subvol, cached_subvol,
- cached_subvol->fops->unlink, loc, xflag, xdata);
+ local->flags = xflag;
+ STACK_WIND_COOKIE(frame, dht_unlink_cbk, cached_subvol, cached_subvol,
+ cached_subvol->fops->unlink, loc, xflag, xdata);
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(unlink, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-int
-dht_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- dht_local_t *local = NULL;
- int ret = -1;
- gf_boolean_t stbuf_merged = _gf_false;
- xlator_t *subvol = NULL;
- call_frame_t *cleanup_frame = NULL;
- dht_local_t *cleanup_local = NULL;
+static int
+dht_remove_stale_linkto_cbk(int ret, call_frame_t *sync_frame, void *data)
+{
+ DHT_STACK_DESTROY(sync_frame);
+ return 0;
+}
- local = frame->local;
+static int
+dht_remove_stale_linkto(void *data)
+{
+ call_frame_t *frame = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *this = NULL;
+ dict_t *xdata_in = NULL;
+ int ret = 0;
- if (op_ret == -1) {
- /* Remove the linkto if exists */
- if (local->linked) {
- cleanup_frame = create_frame (this, this->ctx->pool);
- if (cleanup_frame) {
- cleanup_local = dht_local_init (cleanup_frame,
- &local->loc2,
- NULL, 0);
- if (!cleanup_local || !local->link_subvol) {
- DHT_STACK_DESTROY (cleanup_frame);
- goto out;
- }
- cleanup_local->link_subvol = local->link_subvol;
- FRAME_SU_DO (cleanup_frame, dht_local_t);
- ret = synctask_new (this->ctx->env,
- dht_remove_stale_linkto,
- dht_remove_stale_linkto_cbk,
- cleanup_frame,
- cleanup_frame);
- }
- }
- /* No continuation on DHT inode missing errors, as we should
- * then have a good stbuf that states P2 happened. We would
- * get inode missing if, the file completed migrated between
- * the lookup and the link call */
- goto out;
- }
+ GF_VALIDATE_OR_GOTO("dht", data, out);
- /* Update parent on success, even if P1/2 checks are positve.
- * The second call on success will further update the parent */
- if (local->loc.parent) {
- dht_inode_ctx_time_update (local->loc.parent, this,
- preparent, 0);
- dht_inode_ctx_time_update (local->loc.parent, this,
- postparent, 1);
- }
-
- /* Update linkto attrs, if this is the first call and non-P2,
- * if we detect P2 then we need to trust the attrs from the
- * second call, not the first */
- if (local->linked == _gf_true &&
- ((local->call_cnt == 1 && !IS_DHT_MIGRATION_PHASE2 (stbuf))
- || (local->call_cnt != 1 &&
- IS_DHT_MIGRATION_PHASE2 (&local->stbuf)))) {
- dht_iatt_merge (this, &local->stbuf, stbuf, NULL);
- stbuf_merged = _gf_true;
- dht_linkfile_attr_heal (frame, this);
- }
-
- /* No further P1/2 checks if we are in the second iteration of
- * the call */
- if (local->call_cnt != 1) {
- goto out;
- } else {
- /* Preserve the return values, in case the migration decides
- * to recreate the link on the same subvol that the current
- * hased for the link was created on. */
- dht_iatt_merge (this, &local->preparent,
- preparent, NULL);
- dht_iatt_merge (this, &local->postparent,
- postparent, NULL);
- if (!stbuf_merged) {
- dht_iatt_merge (this, &local->stbuf,
- stbuf, NULL);
- stbuf_merged = _gf_true;
- }
+ frame = data;
+ local = frame->local;
+ this = frame->this;
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", local, out);
+ GF_VALIDATE_OR_GOTO("dht", local->link_subvol, out);
- local->inode = inode_ref (inode);
- }
+ xdata_in = dict_new();
+ if (!xdata_in)
+ goto out;
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- local->rebalance.target_op_fn = dht_link2;
- dht_set_local_rebalance (this, local, stbuf, preparent,
- postparent, xdata);
-
- /* Check if the rebalance phase2 is true */
- if (IS_DHT_MIGRATION_PHASE2 (stbuf)) {
- ret = dht_inode_ctx_get_mig_info (this, local->loc.inode, NULL,
- &subvol);
- if (!subvol) {
- /* Phase 2 of migration */
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret)
- return 0;
- } else {
- dht_link2 (this, subvol, frame, 0);
- return 0;
- }
- }
+ ret = dht_fill_dict_to_avoid_unlink_of_migrating_file(xdata_in);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, 0,
+ "Failed to set keys for stale linkto"
+ "deletion on path %s",
+ local->loc.path);
+ goto out;
+ }
+
+ ret = syncop_unlink(local->link_subvol, &local->loc, xdata_in, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, 0,
+ "Removal of linkto failed"
+ " on path %s at subvol %s",
+ local->loc.path, local->link_subvol->name);
+ }
+out:
+ if (xdata_in)
+ dict_unref(xdata_in);
+ return ret;
+}
- /* Check if the rebalance phase1 is true */
- if (IS_DHT_MIGRATION_PHASE1 (stbuf)) {
- ret = dht_inode_ctx_get_mig_info (this, local->loc.inode, NULL,
- &subvol);
- if (subvol) {
- dht_link2 (this, subvol, frame, 0);
- return 0;
- }
- ret = dht_rebalance_in_progress_check (this, frame);
- if (!ret)
- return 0;
+static int
+dht_link_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, inode_t *inode, struct iatt *stbuf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ int ret = -1;
+ gf_boolean_t stbuf_merged = _gf_false;
+ xlator_t *subvol = NULL;
+ call_frame_t *cleanup_frame = NULL;
+ dht_local_t *cleanup_local = NULL;
+
+ local = frame->local;
+
+ if (op_ret == -1) {
+ /* Remove the linkto if exists */
+ if (local->linked) {
+ cleanup_frame = create_frame(this, this->ctx->pool);
+ if (cleanup_frame) {
+ cleanup_local = dht_local_init(cleanup_frame, &local->loc2,
+ NULL, 0);
+ if (!cleanup_local || !local->link_subvol) {
+ DHT_STACK_DESTROY(cleanup_frame);
+ goto out;
+ }
+ cleanup_local->link_subvol = local->link_subvol;
+ FRAME_SU_DO(cleanup_frame, dht_local_t);
+ ret = synctask_new(this->ctx->env, dht_remove_stale_linkto,
+ dht_remove_stale_linkto_cbk, cleanup_frame,
+ cleanup_frame);
+ }
}
+ /* No continuation on DHT inode missing errors, as we should
+ * then have a good stbuf that states P2 happened. We would
+ * get inode missing if, the file completed migrated between
+ * the lookup and the link call */
+ goto out;
+ }
+
+ /* Update parent on success, even if P1/2 checks are positive.
+ * The second call on success will further update the parent */
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update(local->loc.parent, this, preparent, 0);
+ dht_inode_ctx_time_update(local->loc.parent, this, postparent, 1);
+ }
+
+ /* Update linkto attrs, if this is the first call and non-P2,
+ * if we detect P2 then we need to trust the attrs from the
+ * second call, not the first */
+ if (local->linked == _gf_true &&
+ ((local->call_cnt == 1 && !IS_DHT_MIGRATION_PHASE2(stbuf)) ||
+ (local->call_cnt != 1 && IS_DHT_MIGRATION_PHASE2(&local->stbuf)))) {
+ dht_iatt_merge(this, &local->stbuf, stbuf);
+ stbuf_merged = _gf_true;
+ dht_linkfile_attr_heal(frame, this);
+ }
+
+ /* No further P1/2 checks if we are in the second iteration of
+ * the call */
+ if (local->call_cnt != 1) {
+ goto out;
+ } else {
+ /* Preserve the return values, in case the migration decides
+ * to recreate the link on the same subvol that the current
+ * hased for the link was created on. */
+ dht_iatt_merge(this, &local->preparent, preparent);
+ dht_iatt_merge(this, &local->postparent, postparent);
+ if (!stbuf_merged) {
+ dht_iatt_merge(this, &local->stbuf, stbuf);
+ stbuf_merged = _gf_true;
+ }
+
+ local->inode = inode_ref(inode);
+ }
+
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ local->rebalance.target_op_fn = dht_link2;
+ dht_set_local_rebalance(this, local, stbuf, preparent, postparent, xdata);
+
+ /* Check if the rebalance phase2 is true */
+ if (IS_DHT_MIGRATION_PHASE2(stbuf)) {
+ ret = dht_inode_ctx_get_mig_info(this, local->loc.inode, NULL, &subvol);
+ if (!subvol) {
+ /* Phase 2 of migration */
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret)
+ return 0;
+ } else {
+ dht_link2(this, subvol, frame, 0);
+ return 0;
+ }
+ }
+
+ /* Check if the rebalance phase1 is true */
+ if (IS_DHT_MIGRATION_PHASE1(stbuf)) {
+ ret = dht_inode_ctx_get_mig_info(this, local->loc.inode, NULL, &subvol);
+ if (subvol) {
+ dht_link2(this, subvol, frame, 0);
+ return 0;
+ }
+ ret = dht_rebalance_in_progress_check(this, frame);
+ if (!ret)
+ return 0;
+ }
out:
- DHT_STRIP_PHASE1_FLAGS (stbuf);
+ DHT_STRIP_PHASE1_FLAGS(stbuf);
- dht_set_fixed_dir_stat (preparent);
- dht_set_fixed_dir_stat (postparent);
- DHT_STACK_UNWIND (link, frame, op_ret, op_errno, inode, stbuf,
- preparent, postparent, NULL);
+ dht_set_fixed_dir_stat(preparent);
+ dht_set_fixed_dir_stat(postparent);
+ DHT_STACK_UNWIND(link, frame, op_ret, op_errno, inode, stbuf, preparent,
+ postparent, NULL);
- return 0;
+ return 0;
}
-
-int
-dht_link2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
+static int
+dht_link2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
{
- dht_local_t *local = NULL;
- int op_errno = EINVAL;
+ dht_local_t *local = NULL;
+ int op_errno = EINVAL;
- local = frame->local;
- if (!local)
- goto err;
+ local = frame->local;
+ if (!local)
+ goto err;
- op_errno = local->op_errno;
+ op_errno = local->op_errno;
- if (we_are_not_migrating (ret)) {
- /* This dht xlator is not migrating the file. Unwind and
- * pass on the original mode bits so the higher DHT layer
- * can handle this.
- */
- dht_set_fixed_dir_stat (&local->preparent);
- dht_set_fixed_dir_stat (&local->postparent);
+ if (we_are_not_migrating(ret)) {
+ /* This dht xlator is not migrating the file. Unwind and
+ * pass on the original mode bits so the higher DHT layer
+ * can handle this.
+ */
+ dht_set_fixed_dir_stat(&local->preparent);
+ dht_set_fixed_dir_stat(&local->postparent);
- DHT_STACK_UNWIND (link, frame, local->op_ret, op_errno,
- local->inode,
- &local->stbuf, &local->preparent,
- &local->postparent, NULL);
- return 0;
- }
+ DHT_STACK_UNWIND(link, frame, local->op_ret, op_errno, local->inode,
+ &local->stbuf, &local->preparent, &local->postparent,
+ NULL);
+ return 0;
+ }
- if (subvol == NULL) {
- op_errno = EINVAL;
- goto err;
- }
+ if (subvol == NULL) {
+ op_errno = EINVAL;
+ goto err;
+ }
- /* Second call to create link file could result in EEXIST as the
- * first call created the linkto in the currently
- * migrating subvol, which could be the new hashed subvol */
- if (local->link_subvol == subvol) {
- DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
- dht_set_fixed_dir_stat (&local->preparent);
- dht_set_fixed_dir_stat (&local->postparent);
- DHT_STACK_UNWIND (link, frame, 0, 0, local->inode,
- &local->stbuf, &local->preparent,
- &local->postparent, NULL);
+ /* Second call to create link file could result in EEXIST as the
+ * first call created the linkto in the currently
+ * migrating subvol, which could be the new hashed subvol */
+ if (local->link_subvol == subvol) {
+ DHT_STRIP_PHASE1_FLAGS(&local->stbuf);
+ dht_set_fixed_dir_stat(&local->preparent);
+ dht_set_fixed_dir_stat(&local->postparent);
+ DHT_STACK_UNWIND(link, frame, 0, 0, local->inode, &local->stbuf,
+ &local->preparent, &local->postparent, NULL);
- return 0;
- }
+ return 0;
+ }
- local->call_cnt = 2;
+ local->call_cnt = 2;
- STACK_WIND (frame, dht_link_cbk, subvol, subvol->fops->link,
- &local->loc, &local->loc2, local->xattr_req);
+ STACK_WIND(frame, dht_link_cbk, subvol, subvol->fops->link, &local->loc,
+ &local->loc2, local->xattr_req);
- return 0;
+ return 0;
err:
- DHT_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL,
- NULL, NULL);
+ DHT_STACK_UNWIND(link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-int
-dht_link_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+static int
+dht_link_linkfile_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *srcvol = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *srcvol = NULL;
- if (op_ret == -1)
- goto err;
+ if (op_ret == -1)
+ goto err;
- local = frame->local;
- srcvol = local->linkfile.srcvol;
+ local = frame->local;
+ srcvol = local->linkfile.srcvol;
- STACK_WIND (frame, dht_link_cbk, srcvol, srcvol->fops->link,
- &local->loc, &local->loc2, local->xattr_req);
+ STACK_WIND(frame, dht_link_cbk, srcvol, srcvol->fops->link, &local->loc,
+ &local->loc2, local->xattr_req);
- return 0;
+ return 0;
err:
- DHT_STRIP_PHASE1_FLAGS (stbuf);
- dht_set_fixed_dir_stat (preparent);
- dht_set_fixed_dir_stat (postparent);
- DHT_STACK_UNWIND (link, frame, op_ret, op_errno, inode, stbuf, preparent,
- postparent, xdata);
+ DHT_STRIP_PHASE1_FLAGS(stbuf);
+ dht_set_fixed_dir_stat(preparent);
+ dht_set_fixed_dir_stat(postparent);
+ DHT_STACK_UNWIND(link, frame, op_ret, op_errno, inode, stbuf, preparent,
+ postparent, xdata);
- return 0;
+ return 0;
}
-
int
-dht_link (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
-{
- xlator_t *cached_subvol = NULL;
- xlator_t *hashed_subvol = NULL;
- int op_errno = -1;
- int ret = -1;
- dht_local_t *local = NULL;
+dht_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
+{
+ xlator_t *cached_subvol = NULL;
+ xlator_t *hashed_subvol = NULL;
+ int op_errno = -1;
+ int ret = -1;
+ dht_local_t *local = NULL;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(oldloc, err);
+ VALIDATE_OR_GOTO(newloc, err);
+
+ local = dht_local_init(frame, oldloc, NULL, GF_FOP_LINK);
+ if (!local) {
+ op_errno = ENOMEM;
+
+ goto err;
+ }
+ local->call_cnt = 1;
+
+ cached_subvol = local->cached_subvol;
+ if (!cached_subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for path=%s",
+ oldloc->path);
+ op_errno = ENOENT;
+ goto err;
+ }
+
+ hashed_subvol = dht_subvol_get_hashed(this, newloc);
+ if (!hashed_subvol) {
+ gf_msg_debug(this->name, 0, "no subvolume in layout for path=%s",
+ newloc->path);
+ op_errno = EIO;
+ goto err;
+ }
+
+ ret = loc_copy(&local->loc2, newloc);
+ if (ret == -1) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ if (hashed_subvol != cached_subvol) {
+ gf_uuid_copy(local->gfid, oldloc->inode->gfid);
+ dht_linkfile_create(frame, dht_link_linkfile_cbk, this, cached_subvol,
+ hashed_subvol, newloc);
+ } else {
+ STACK_WIND(frame, dht_link_cbk, cached_subvol,
+ cached_subvol->fops->link, oldloc, newloc, xdata);
+ }
+
+ return 0;
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (oldloc, err);
- VALIDATE_OR_GOTO (newloc, err);
+err:
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
- local = dht_local_init (frame, oldloc, NULL, GF_FOP_LINK);
- if (!local) {
- op_errno = ENOMEM;
+ return 0;
+}
- goto err;
- }
- local->call_cnt = 1;
+int
+dht_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, fd_t *fd, inode_t *inode, struct iatt *stbuf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
+{
+ xlator_t *prev = NULL;
+ int ret = -1;
+ dht_local_t *local = NULL;
+ gf_boolean_t parent_layout_changed = _gf_false;
+ char pgfid[GF_UUID_BUF_SIZE] = {0};
+ xlator_t *subvol = NULL;
- cached_subvol = local->cached_subvol;
- if (!cached_subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for path=%s", oldloc->path);
- op_errno = ENOENT;
- goto err;
- }
+ local = frame->local;
- hashed_subvol = dht_subvol_get_hashed (this, newloc);
- if (!hashed_subvol) {
- gf_msg_debug (this->name, 0,
- "no subvolume in layout for path=%s",
- newloc->path);
- op_errno = EIO;
- goto err;
- }
+ local = frame->local;
+ if (!local) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto out;
+ }
- ret = loc_copy (&local->loc2, newloc);
- if (ret == -1) {
- op_errno = ENOMEM;
- goto err;
- }
- if (xdata)
- local->xattr_req = dict_ref (xdata);
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ parent_layout_changed = (xdata &&
+ dict_get(xdata, GF_PREOP_CHECK_FAILED))
+ ? _gf_true
+ : _gf_false;
+
+ if (parent_layout_changed) {
+ if (local && local->lock[0].layout.parent_layout.locks) {
+ /* Returning failure as the layout could not be fixed even under
+ * the lock */
+ goto out;
+ }
- if (hashed_subvol != cached_subvol) {
- gf_uuid_copy (local->gfid, oldloc->inode->gfid);
- dht_linkfile_create (frame, dht_link_linkfile_cbk, this,
- cached_subvol, hashed_subvol, newloc);
- } else {
- STACK_WIND (frame, dht_link_cbk,
- cached_subvol, cached_subvol->fops->link,
- oldloc, newloc, xdata);
- }
+ gf_uuid_unparse(local->loc.parent->gfid, pgfid);
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "create (%s/%s) (path: %s): parent layout "
+ "changed. Attempting a layout refresh and then a "
+ "retry",
+ pgfid, local->loc.name, local->loc.path);
- return 0;
+ /*
+ dht_refresh_layout needs directory info in local->loc.Hence,
+ storing the parent_loc in local->loc and storing the create
+ context in local->loc2. We will restore this information in
+ dht_creation_do.
+ */
-err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+ loc_wipe(&local->loc2);
- return 0;
-}
+ ret = loc_copy(&local->loc2, &local->loc);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
+ "loc_copy failed %s", local->loc.path);
+ goto out;
+ }
-int
-dht_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- fd_t *fd, inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
-{
- xlator_t *prev = NULL;
- int ret = -1;
- dht_local_t *local = NULL;
+ loc_wipe(&local->loc);
- local = frame->local;
- if (!local) {
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
+ ret = dht_build_parent_loc(this, &local->loc, &local->loc2,
+ &op_errno);
- if (op_ret == -1)
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_LOC_FAILED,
+ "parent loc build failed");
goto out;
+ }
- prev = cookie;
-
- if (local->loc.parent) {
- dht_inode_ctx_time_update (local->loc.parent, this,
- preparent, 0);
+ subvol = dht_subvol_get_hashed(this, &local->loc2);
- dht_inode_ctx_time_update (local->loc.parent, this,
- postparent, 1);
- }
+ ret = dht_create_lock(frame, subvol);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_INODE_LK_ERROR,
+ "locking parent failed");
+ goto out;
+ }
- ret = dht_fd_ctx_set (this, fd, prev);
- if (ret != 0) {
- gf_msg_debug (this->name, 0, "Possible fd leak. "
- "Could not set fd ctx for subvol %s",
- prev->name);
+ return 0;
}
+ goto out;
+ }
+
+ prev = cookie;
+
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update(local->loc.parent, this, preparent, 0);
+
+ dht_inode_ctx_time_update(local->loc.parent, this, postparent, 1);
+ }
+
+ ret = dht_fd_ctx_set(this, fd, prev);
+ if (ret != 0) {
+ gf_msg_debug(this->name, 0,
+ "Possible fd leak. "
+ "Could not set fd ctx for subvol %s",
+ prev->name);
+ }
+
+ ret = dht_layout_preset(this, prev, inode);
+ if (ret != 0) {
+ gf_msg_debug(this->name, 0, "could not set preset layout for subvol %s",
+ prev->name);
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto out;
+ }
- ret = dht_layout_preset (this, prev, inode);
- if (ret != 0) {
- gf_msg_debug (this->name, 0,
- "could not set preset layout for subvol %s",
- prev->name);
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
-
- local->op_errno = op_errno;
+ local->op_errno = op_errno;
- if (local->linked == _gf_true) {
- local->stbuf = *stbuf;
- dht_linkfile_attr_heal (frame, this);
- }
+ if (local->linked == _gf_true) {
+ local->stbuf = *stbuf;
+ dht_linkfile_attr_heal(frame, this);
+ }
out:
- DHT_STRIP_PHASE1_FLAGS (stbuf);
- dht_set_fixed_dir_stat (preparent);
- dht_set_fixed_dir_stat (postparent);
+ DHT_STRIP_PHASE1_FLAGS(stbuf);
+ dht_set_fixed_dir_stat(preparent);
+ dht_set_fixed_dir_stat(postparent);
- if (local && local->lock[0].layout.parent_layout.locks) {
- /* store op_errno for failure case*/
- local->op_errno = op_errno;
- local->refresh_layout_unlock (frame, this, op_ret, 1);
+ if (local && local->lock[0].layout.parent_layout.locks) {
+ /* store op_errno for failure case*/
+ local->op_errno = op_errno;
+ local->refresh_layout_unlock(frame, this, op_ret, 1);
- if (op_ret == 0) {
- DHT_STACK_UNWIND (create, frame, op_ret, op_errno, fd,
- inode, stbuf, preparent, postparent,
- xdata);
- }
- } else {
- DHT_STACK_UNWIND (create, frame, op_ret, op_errno, fd, inode,
- stbuf, preparent, postparent, xdata);
+ if (op_ret == 0) {
+ DHT_STACK_UNWIND(create, frame, op_ret, op_errno, fd, inode, stbuf,
+ preparent, postparent, xdata);
}
- return 0;
+ } else {
+ DHT_STACK_UNWIND(create, frame, op_ret, op_errno, fd, inode, stbuf,
+ preparent, postparent, xdata);
+ }
+ return 0;
}
-int
-dht_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+static int
+dht_create_linkfile_create_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ inode_t *inode, struct iatt *stbuf,
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *cached_subvol = NULL;
- dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *cached_subvol = NULL;
+ dht_conf_t *conf = NULL;
- local = frame->local;
- if (!local) {
- op_errno = EINVAL;
- goto err;
- }
+ local = frame->local;
+ if (!local) {
+ op_errno = EINVAL;
+ goto err;
+ }
- if (op_ret == -1) {
- local->op_errno = op_errno;
- goto err;
- }
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ goto err;
+ }
- conf = this->private;
- if (!conf) {
- local->op_errno = EINVAL;
- op_errno = EINVAL;
- goto err;
- }
+ conf = this->private;
+ if (!conf) {
+ local->op_errno = EINVAL;
+ op_errno = EINVAL;
+ goto err;
+ }
- cached_subvol = local->cached_subvol;
+ cached_subvol = local->cached_subvol;
- if (local->params) {
- dict_del (local->params, conf->link_xattr_name);
- dict_del (local->params, GLUSTERFS_INTERNAL_FOP_KEY);
- }
+ if (local->params) {
+ dict_del(local->params, conf->link_xattr_name);
+ dict_del(local->params, GLUSTERFS_INTERNAL_FOP_KEY);
+ }
- STACK_WIND_COOKIE (frame, dht_create_cbk, cached_subvol,
- cached_subvol, cached_subvol->fops->create,
- &local->loc, local->flags, local->mode,
- local->umask, local->fd, local->params);
+ STACK_WIND_COOKIE(frame, dht_create_cbk, cached_subvol, cached_subvol,
+ cached_subvol->fops->create, &local->loc, local->flags,
+ local->mode, local->umask, local->fd, local->params);
- return 0;
+ return 0;
err:
- if (local && local->lock[0].layout.parent_layout.locks) {
- local->refresh_layout_unlock (frame, this, -1, 1);
- } else {
- DHT_STACK_UNWIND (create, frame, -1,
- op_errno, NULL, NULL, NULL,
- NULL, NULL, NULL);
- }
- return 0;
+ if (local && local->lock[0].layout.parent_layout.locks) {
+ local->refresh_layout_unlock(frame, this, -1, 1);
+ } else {
+ DHT_STACK_UNWIND(create, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL, NULL);
+ }
+ return 0;
}
-int
-dht_create_wind_to_avail_subvol (call_frame_t *frame, xlator_t *this,
- xlator_t *subvol, loc_t *loc, int32_t flags,
- mode_t mode, mode_t umask, fd_t *fd,
- dict_t *params)
+static int
+dht_create_wind_to_avail_subvol(call_frame_t *frame, xlator_t *this,
+ xlator_t *subvol, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd,
+ dict_t *params)
{
- dht_local_t *local = NULL;
- xlator_t *avail_subvol = NULL;
-
- local = frame->local;
+ dht_local_t *local = NULL;
+ xlator_t *avail_subvol = NULL;
- if (!dht_is_subvol_filled (this, subvol)) {
- gf_msg_debug (this->name, 0,
- "creating %s on %s", loc->path,
- subvol->name);
+ local = frame->local;
- STACK_WIND_COOKIE (frame, dht_create_cbk, subvol,
- subvol, subvol->fops->create,
- loc, flags, mode, umask, fd, params);
+ if (!dht_is_subvol_filled(this, subvol)) {
+ gf_msg_debug(this->name, 0, "creating %s on %s", loc->path,
+ subvol->name);
- } else {
- avail_subvol = dht_free_disk_available_subvol (this, subvol, local);
+ dht_set_parent_layout_in_dict(loc, this, local);
- if (avail_subvol != subvol) {
- local->params = dict_ref (params);
- local->flags = flags;
- local->mode = mode;
- local->umask = umask;
- local->cached_subvol = avail_subvol;
- local->hashed_subvol = subvol;
+ STACK_WIND_COOKIE(frame, dht_create_cbk, subvol, subvol,
+ subvol->fops->create, loc, flags, mode, umask, fd,
+ params);
- gf_msg_debug (this->name, 0,
- "creating %s on %s (link at %s)", loc->path,
- avail_subvol->name, subvol->name);
+ } else {
+ avail_subvol = dht_free_disk_available_subvol(this, subvol, local);
- dht_linkfile_create (frame, dht_create_linkfile_create_cbk,
- this, avail_subvol, subvol, loc);
+ if (avail_subvol != subvol) {
+ local->cached_subvol = avail_subvol;
+ local->hashed_subvol = subvol;
- goto out;
- }
+ gf_msg_debug(this->name, 0, "creating %s on %s (link at %s)",
+ loc->path, avail_subvol->name, subvol->name);
- gf_msg_debug (this->name, 0,
- "creating %s on %s", loc->path, subvol->name);
+ dht_linkfile_create(frame, dht_create_linkfile_create_cbk, this,
+ avail_subvol, subvol, loc);
- STACK_WIND_COOKIE (frame, dht_create_cbk, subvol,
- subvol, subvol->fops->create,
- loc, flags, mode, umask, fd, params);
+ goto out;
}
+
+ gf_msg_debug(this->name, 0, "creating %s on %s", loc->path,
+ subvol->name);
+
+ dht_set_parent_layout_in_dict(loc, this, local);
+
+ STACK_WIND_COOKIE(frame, dht_create_cbk, subvol, subvol,
+ subvol->fops->create, loc, flags, mode, umask, fd,
+ params);
+ }
out:
- return 0;
+ return 0;
}
int
-dht_build_parent_loc (xlator_t *this, loc_t *parent, loc_t *child,
- int32_t *op_errno)
+dht_build_parent_loc(xlator_t *this, loc_t *parent, loc_t *child,
+ int32_t *op_errno)
{
- inode_table_t *table = NULL;
- int ret = -1;
+ inode_table_t *table = NULL;
+ int ret = -1;
- if (!parent || !child) {
- if (op_errno)
- *op_errno = EINVAL;
- goto out;
- }
+ if (!parent || !child) {
+ if (op_errno)
+ *op_errno = EINVAL;
+ goto out;
+ }
- if (child->parent) {
- parent->inode = inode_ref (child->parent);
- if (!parent->inode) {
- if (op_errno)
- *op_errno = EINVAL;
- goto out;
- }
+ if (child->parent) {
+ parent->inode = inode_ref(child->parent);
+ if (!parent->inode) {
+ if (op_errno)
+ *op_errno = EINVAL;
+ goto out;
+ }
- gf_uuid_copy (parent->gfid, child->pargfid);
+ gf_uuid_copy(parent->gfid, child->pargfid);
- ret = 0;
+ ret = 0;
- goto out;
- } else {
- if (gf_uuid_is_null (child->pargfid)) {
- if (op_errno)
- *op_errno = EINVAL;
- goto out;
- }
+ goto out;
+ } else {
+ if (gf_uuid_is_null(child->pargfid)) {
+ if (op_errno)
+ *op_errno = EINVAL;
+ goto out;
+ }
- table = this->itable;
+ table = this->itable;
- if (!table) {
- if (op_errno) {
- *op_errno = EINVAL;
- goto out;
- }
- }
+ if (!table) {
+ if (op_errno) {
+ *op_errno = EINVAL;
+ goto out;
+ }
+ }
- parent->inode = inode_find (table, child->pargfid);
+ parent->inode = inode_find(table, child->pargfid);
- if (!parent->inode) {
- if (op_errno) {
- *op_errno = EINVAL;
- goto out;
- }
- }
+ if (!parent->inode) {
+ if (op_errno) {
+ *op_errno = EINVAL;
+ goto out;
+ }
+ }
- gf_uuid_copy (parent->gfid, child->pargfid);
+ gf_uuid_copy(parent->gfid, child->pargfid);
- ret = 0;
- }
+ ret = 0;
+ }
out:
- return ret;
+ return ret;
}
-
-int32_t
-dht_create_do (call_frame_t *frame)
+static int32_t
+dht_create_do(call_frame_t *frame)
{
- dht_local_t *local = NULL;
- dht_layout_t *refreshed = NULL;
- xlator_t *subvol = NULL;
- xlator_t *this = NULL;
- dht_conf_t *conf = NULL;
- dht_methods_t *methods = NULL;
+ dht_local_t *local = NULL;
+ dht_layout_t *refreshed = NULL;
+ xlator_t *subvol = NULL;
+ xlator_t *this = NULL;
+ dht_conf_t *conf = NULL;
+ dht_methods_t *methods = NULL;
- local = frame->local;
+ local = frame->local;
- this = THIS;
+ this = THIS;
- conf = this->private;
+ conf = this->private;
- GF_VALIDATE_OR_GOTO (this->name, conf, err);
+ GF_VALIDATE_OR_GOTO(this->name, conf, err);
- methods = &(conf->methods);
+ methods = &(conf->methods);
- /* We don't need parent_loc anymore */
- loc_wipe (&local->loc);
+ /* We don't need parent_loc anymore */
+ loc_wipe(&local->loc);
- loc_copy (&local->loc, &local->loc2);
+ loc_copy(&local->loc, &local->loc2);
- loc_wipe (&local->loc2);
+ loc_wipe(&local->loc2);
- refreshed = local->selfheal.refreshed_layout;
+ refreshed = local->selfheal.refreshed_layout;
- subvol = methods->layout_search (this, refreshed, local->loc.name);
+ subvol = methods->layout_search(this, refreshed, local->loc.name);
- if (!subvol) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_HASHED_SUBVOL_GET_FAILED, "no subvolume in "
- "layout for path=%s", local->loc.path);
- local->op_errno = ENOENT;
- goto err;
- }
+ if (!subvol) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_HASHED_SUBVOL_GET_FAILED,
+ "no subvolume in "
+ "layout for path=%s",
+ local->loc.path);
+ local->op_errno = ENOENT;
+ goto err;
+ }
- dht_create_wind_to_avail_subvol (frame, this, subvol, &local->loc,
- local->flags, local->mode,
- local->umask, local->fd, local->params);
- return 0;
+ dht_create_wind_to_avail_subvol(frame, this, subvol, &local->loc,
+ local->flags, local->mode, local->umask,
+ local->fd, local->params);
+ return 0;
err:
- local->refresh_layout_unlock (frame, this, -1, 1);
+ local->refresh_layout_unlock(frame, this, -1, 1);
- return 0;
+ return 0;
}
-int32_t
-dht_create_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+static int32_t
+dht_create_unlock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- DHT_STACK_DESTROY (frame);
- return 0;
+ DHT_STACK_DESTROY(frame);
+ return 0;
}
-int32_t
-dht_create_finish (call_frame_t *frame, xlator_t *this, int op_ret,
- int invoke_cbk)
+static int32_t
+dht_create_finish(call_frame_t *frame, xlator_t *this, int op_ret,
+ int invoke_cbk)
{
- dht_local_t *local = NULL, *lock_local = NULL;
- call_frame_t *lock_frame = NULL;
- int lock_count = 0;
-
- local = frame->local;
- lock_count = dht_lock_count (local->lock[0].layout.parent_layout.locks,
- local->lock[0].layout.parent_layout.lk_count);
- if (lock_count == 0)
- goto done;
-
- lock_frame = copy_frame (frame);
- if (lock_frame == NULL) {
- goto done;
- }
-
- lock_local = dht_local_init (lock_frame, &local->loc, NULL,
- lock_frame->root->op);
- if (lock_local == NULL) {
- goto done;
- }
-
- lock_local->lock[0].layout.parent_layout.locks = local->lock[0].layout.parent_layout.locks;
- lock_local->lock[0].layout.parent_layout.lk_count = local->lock[0].layout.parent_layout.lk_count;
-
- local->lock[0].layout.parent_layout.locks = NULL;
- local->lock[0].layout.parent_layout.lk_count = 0;
-
- dht_unlock_inodelk (lock_frame,
- lock_local->lock[0].layout.parent_layout.locks,
- lock_local->lock[0].layout.parent_layout.lk_count,
- dht_create_unlock_cbk);
- lock_frame = NULL;
+ dht_local_t *local = NULL, *lock_local = NULL;
+ call_frame_t *lock_frame = NULL;
+ int lock_count = 0;
+
+ local = frame->local;
+ lock_count = dht_lock_count(local->lock[0].layout.parent_layout.locks,
+ local->lock[0].layout.parent_layout.lk_count);
+ if (lock_count == 0)
+ goto done;
+
+ lock_frame = copy_frame(frame);
+ if (lock_frame == NULL) {
+ goto done;
+ }
+
+ lock_local = dht_local_init(lock_frame, &local->loc, NULL,
+ lock_frame->root->op);
+ if (lock_local == NULL) {
+ goto done;
+ }
+
+ lock_local->lock[0]
+ .layout.parent_layout.locks = local->lock[0].layout.parent_layout.locks;
+ lock_local->lock[0].layout.parent_layout.lk_count =
+ local->lock[0].layout.parent_layout.lk_count;
+
+ local->lock[0].layout.parent_layout.locks = NULL;
+ local->lock[0].layout.parent_layout.lk_count = 0;
+
+ dht_unlock_inodelk(lock_frame,
+ lock_local->lock[0].layout.parent_layout.locks,
+ lock_local->lock[0].layout.parent_layout.lk_count,
+ dht_create_unlock_cbk);
+ lock_frame = NULL;
done:
- if (lock_frame != NULL) {
- DHT_STACK_DESTROY (lock_frame);
- }
+ if (lock_frame != NULL) {
+ DHT_STACK_DESTROY(lock_frame);
+ }
- if (op_ret == 0)
- return 0;
-
- DHT_STACK_UNWIND (create, frame, op_ret, local->op_errno, NULL, NULL,
- NULL, NULL, NULL, NULL);
+ if (op_ret == 0)
return 0;
+
+ DHT_STACK_UNWIND(create, frame, op_ret, local->op_errno, NULL, NULL, NULL,
+ NULL, NULL, NULL);
+ return 0;
}
-int32_t
-dht_create_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+static int32_t
+dht_create_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
+ dht_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (!local) {
- goto err;
- }
+ if (!local) {
+ goto err;
+ }
- if (op_ret < 0) {
- gf_msg ("DHT", GF_LOG_ERROR, 0, DHT_MSG_INODE_LK_ERROR,
- "Create lock failed for file: %s", local->loc2.name);
+ if (op_ret < 0) {
+ gf_msg("DHT", GF_LOG_ERROR, 0, DHT_MSG_INODE_LK_ERROR,
+ "Create lock failed for file: %s", local->loc2.name);
- local->op_errno = op_errno;
+ local->op_errno = op_errno;
- goto err;
- }
+ goto err;
+ }
- local->refresh_layout_unlock = dht_create_finish;
+ local->refresh_layout_unlock = dht_create_finish;
- local->refresh_layout_done = dht_create_do;
+ local->refresh_layout_done = dht_create_do;
- dht_refresh_layout (frame);
+ dht_refresh_layout(frame);
- return 0;
+ return 0;
err:
- dht_create_finish (frame, this, -1, 0);
- return 0;
+ if (local)
+ dht_create_finish(frame, this, -1, 0);
+ else
+ DHT_STACK_UNWIND(create, frame, -1, EINVAL, NULL, NULL, NULL, NULL,
+ NULL, NULL);
+ return 0;
}
int32_t
-dht_create_lock (call_frame_t *frame, xlator_t *subvol)
+dht_create_lock(call_frame_t *frame, xlator_t *subvol)
{
- dht_local_t *local = NULL;
- int count = 1, ret = -1;
- dht_lock_t **lk_array = NULL;
+ dht_local_t *local = NULL;
+ int count = 1, ret = -1;
+ dht_lock_t **lk_array = NULL;
- GF_VALIDATE_OR_GOTO ("dht", frame, err);
- GF_VALIDATE_OR_GOTO (frame->this->name, frame->local, err);
+ GF_VALIDATE_OR_GOTO("dht", frame, err);
+ GF_VALIDATE_OR_GOTO(frame->this->name, frame->local, err);
- local = frame->local;
+ local = frame->local;
- lk_array = GF_CALLOC (count, sizeof (*lk_array), gf_common_mt_pointer);
+ lk_array = GF_CALLOC(count, sizeof(*lk_array), gf_common_mt_pointer);
- if (lk_array == NULL)
- goto err;
+ if (lk_array == NULL)
+ goto err;
- lk_array[0] = dht_lock_new (frame->this, subvol, &local->loc, F_RDLCK,
- DHT_LAYOUT_HEAL_DOMAIN, NULL,
- IGNORE_ENOENT_ESTALE);
+ lk_array[0] = dht_lock_new(frame->this, subvol, &local->loc, F_RDLCK,
+ DHT_LAYOUT_HEAL_DOMAIN, NULL,
+ IGNORE_ENOENT_ESTALE);
- if (lk_array[0] == NULL)
- goto err;
+ if (lk_array[0] == NULL)
+ goto err;
- local->lock[0].layout.parent_layout.locks = lk_array;
- local->lock[0].layout.parent_layout.lk_count = count;
+ local->lock[0].layout.parent_layout.locks = lk_array;
+ local->lock[0].layout.parent_layout.lk_count = count;
- ret = dht_blocking_inodelk (frame, lk_array, count,
- dht_create_lock_cbk);
+ ret = dht_blocking_inodelk(frame, lk_array, count, dht_create_lock_cbk);
- if (ret < 0) {
- local->lock[0].layout.parent_layout.locks = NULL;
- local->lock[0].layout.parent_layout.lk_count = 0;
- goto err;
- }
+ if (ret < 0) {
+ local->lock[0].layout.parent_layout.locks = NULL;
+ local->lock[0].layout.parent_layout.lk_count = 0;
+ goto err;
+ }
- return 0;
+ return 0;
err:
- if (lk_array != NULL) {
- dht_lock_array_free (lk_array, count);
- GF_FREE (lk_array);
- }
+ if (lk_array != NULL) {
+ dht_lock_array_free(lk_array, count);
+ GF_FREE(lk_array);
+ }
- return -1;
+ return -1;
}
int
-dht_create (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, mode_t mode,
- mode_t umask, fd_t *fd, dict_t *params)
-{
- int op_errno = -1;
- xlator_t *subvol = NULL;
- dht_local_t *local = NULL;
- int i = 0;
- dht_conf_t *conf = NULL;
- int ret = 0;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
-
- conf = this->private;
+dht_set_parent_layout_in_dict(loc_t *loc, xlator_t *this, dht_local_t *local)
+{
+ dht_conf_t *conf = this->private;
+ dht_layout_t *parent_layout = NULL;
+ int *parent_disk_layout = NULL;
+ xlator_t *hashed_subvol = NULL;
+ char pgfid[GF_UUID_BUF_SIZE] = {0};
+ int ret = 0;
+
+ gf_uuid_unparse(loc->parent->gfid, pgfid);
+
+ parent_layout = dht_layout_get(this, loc->parent);
+ hashed_subvol = dht_subvol_get_hashed(this, loc);
+
+ ret = dht_disk_layout_extract_for_subvol(this, parent_layout, hashed_subvol,
+ &parent_disk_layout);
+ if (ret == -1) {
+ gf_msg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "%s (%s/%s) (path: %s): "
+ "extracting in-memory layout of parent failed. ",
+ gf_fop_list[local->fop], pgfid, loc->name, loc->path);
+ goto err;
+ }
+
+ ret = dict_set_str_sizen(local->params, GF_PREOP_PARENT_KEY,
+ conf->xattr_name);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "%s (%s/%s) (path: %s): "
+ "setting %s key in params dictionary failed. ",
+ gf_fop_list[local->fop], pgfid, loc->name, loc->path,
+ GF_PREOP_PARENT_KEY);
+ goto err;
+ }
+
+ ret = dict_set_bin(local->params, conf->xattr_name, parent_disk_layout,
+ 4 * 4);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "%s (%s/%s) (path: %s): "
+ "setting parent-layout in params dictionary failed. ",
+ gf_fop_list[local->fop], pgfid, loc->name, loc->path);
+ goto err;
+ }
- dht_get_du_info (frame, this, loc);
-
- local = dht_local_init (frame, loc, fd, GF_FOP_CREATE);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+err:
+ dht_layout_unref(this, parent_layout);
+ return ret;
+}
- if (dht_filter_loc_subvol_key (this, loc, &local->loc,
- &subvol)) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_SUBVOL_INFO,
- "creating %s on %s (got create on %s)",
- local->loc.path, subvol->name, loc->path);
- STACK_WIND_COOKIE (frame, dht_create_cbk, subvol,
- subvol, subvol->fops->create, &local->loc,
- flags, mode, umask, fd, params);
- goto done;
- }
+int
+dht_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *params)
+{
+ int op_errno = -1;
+ xlator_t *subvol = NULL;
+ xlator_t *hashed_subvol = NULL;
+ dht_local_t *local = NULL;
+ int i = 0;
+ dht_conf_t *conf = NULL;
+ int ret = 0;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+
+ conf = this->private;
+
+ dht_get_du_info(frame, this, loc);
+
+ local = dht_local_init(frame, loc, fd, GF_FOP_CREATE);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->params = dict_ref(params);
+ local->flags = flags;
+ local->mode = mode;
+ local->umask = umask;
+
+ if (dht_filter_loc_subvol_key(this, loc, &local->loc, &subvol)) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_SUBVOL_INFO,
+ "creating %s on %s (got create on %s)", local->loc.path,
+ subvol->name, loc->path);
+
+ /* Since lookup-optimize is enabled by default, we need
+ * to create the linkto file if required.
+ * Note this does not check for decommisioned bricks
+ * and min-free-disk limits as this is a debugging tool
+ * and not expected to be used in production.
+ */
+ hashed_subvol = dht_subvol_get_hashed(this, &local->loc);
- subvol = dht_subvol_get_hashed (this, loc);
- if (!subvol) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_HASHED_SUBVOL_GET_FAILED,
- "no subvolume in layout for path=%s",
- loc->path);
+ if (hashed_subvol && (hashed_subvol != subvol)) {
+ /* Create the linkto file and then the data file */
+ local->cached_subvol = subvol;
+ local->hashed_subvol = hashed_subvol;
- op_errno = EIO;
- goto err;
+ dht_linkfile_create(frame, dht_create_linkfile_create_cbk, this,
+ subvol, hashed_subvol, &local->loc);
+ goto done;
}
+ /* We either don't have a hashed subvol or the hashed subvol is
+ * the same as the one specified. No need to create the linkto
+ * file as we expect a lookup everywhere if there are problems
+ * with the parent layout
+ */
- /* Post remove-brick, the client layout may not be in sync with
- * disk layout because of lack of lookup. Hence,a create call
- * may fall on the decommissioned brick. Hence, if the
- * hashed_subvol is part of decommissioned bricks list, do a
- * lookup on parent dir. If a fix-layout is already done by the
- * remove-brick process, the parent directory layout will be in
- * sync with that of the disk. If fix-layout is still ending
- * on the parent directory, we can let the file get created on
- * the decommissioned brick which will be eventually migrated to
- * non-decommissioned brick based on the new layout.
- */
-
- if (conf->decommission_subvols_cnt) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->decommissioned_bricks[i] &&
- conf->decommissioned_bricks[i] == subvol) {
-
- gf_msg_debug (this->name, 0, "hashed subvol:%s is "
- "part of decommission brick list for "
- "file: %s", subvol->name, loc->path);
-
- /* dht_refresh_layout needs directory info in
- * local->loc. Hence, storing the parent_loc in
- * local->loc and storing the create context in
- * local->loc2. We will restore this information
- * in dht_creation do */
-
- ret = loc_copy (&local->loc2, &local->loc);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_NO_MEMORY,
- "loc_copy failed %s", loc->path);
-
- goto err;
- }
+ dht_set_parent_layout_in_dict(loc, this, local);
+
+ STACK_WIND_COOKIE(frame, dht_create_cbk, subvol, subvol,
+ subvol->fops->create, &local->loc, flags, mode, umask,
+ fd, params);
+ goto done;
+ }
+
+ subvol = dht_subvol_get_hashed(this, loc);
+ if (!subvol) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_HASHED_SUBVOL_GET_FAILED,
+ "no subvolume in layout for path=%s", loc->path);
+
+ op_errno = EIO;
+ goto err;
+ }
+
+ /* Post remove-brick, the client layout may not be in sync with
+ * disk layout because of lack of lookup. Hence,a create call
+ * may fall on the decommissioned brick. Hence, if the
+ * hashed_subvol is part of decommissioned bricks list, do a
+ * lookup on parent dir. If a fix-layout is already done by the
+ * remove-brick process, the parent directory layout will be in
+ * sync with that of the disk. If fix-layout is still ending
+ * on the parent directory, we can let the file get created on
+ * the decommissioned brick which will be eventually migrated to
+ * non-decommissioned brick based on the new layout.
+ */
+
+ if (conf->decommission_subvols_cnt) {
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->decommissioned_bricks[i] &&
+ conf->decommissioned_bricks[i] == subvol) {
+ gf_msg_debug(this->name, 0,
+ "hashed subvol:%s is "
+ "part of decommission brick list for "
+ "file: %s",
+ subvol->name, loc->path);
+
+ /* dht_refresh_layout needs directory info in
+ * local->loc. Hence, storing the parent_loc in
+ * local->loc and storing the create context in
+ * local->loc2. We will restore this information
+ * in dht_creation do */
+
+ ret = loc_copy(&local->loc2, &local->loc);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
+ "loc_copy failed %s", loc->path);
- local->params = dict_ref (params);
- local->flags = flags;
- local->mode = mode;
- local->umask = umask;
+ goto err;
+ }
- loc_wipe (&local->loc);
+ loc_wipe(&local->loc);
- ret = dht_build_parent_loc (this, &local->loc, loc,
- &op_errno);
+ ret = dht_build_parent_loc(this, &local->loc, loc, &op_errno);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_LOC_FAILED,
- "parent loc build failed");
- goto err;
- }
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_LOC_FAILED,
+ "parent loc build failed");
+ goto err;
+ }
- ret = dht_create_lock (frame, subvol);
+ ret = dht_create_lock(frame, subvol);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_INODE_LK_ERROR,
- "locking parent failed");
- goto err;
- }
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_INODE_LK_ERROR,
+ "locking parent failed");
+ goto err;
+ }
- goto done;
- }
+ goto done;
}
}
+ }
-
- dht_create_wind_to_avail_subvol (frame, this, subvol, loc, flags, mode,
- umask, fd, params);
+ dht_create_wind_to_avail_subvol(frame, this, subvol, loc, flags, mode,
+ umask, fd, params);
done:
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL,
- NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(create, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL,
+ NULL);
- return 0;
+ return 0;
}
-
-int
-dht_mkdir_selfheal_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+static int
+dht_mkdir_selfheal_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- dht_layout_t *layout = NULL;
+ dht_local_t *local = NULL;
+ dht_layout_t *layout = NULL;
- local = frame->local;
- layout = local->selfheal.layout;
+ local = frame->local;
+ layout = local->selfheal.layout;
- FRAME_SU_UNDO (frame, dht_local_t);
- dht_set_fixed_dir_stat (&local->preparent);
- dht_set_fixed_dir_stat (&local->postparent);
+ FRAME_SU_UNDO(frame, dht_local_t);
+ dht_set_fixed_dir_stat(&local->preparent);
+ dht_set_fixed_dir_stat(&local->postparent);
- if (op_ret == 0) {
- dht_layout_set (this, local->inode, layout);
+ if (op_ret == 0) {
+ dht_layout_set(this, local->inode, layout);
- dht_inode_ctx_time_update (local->inode, this,
- &local->stbuf, 1);
- if (local->loc.parent) {
- dht_inode_ctx_time_update (local->loc.parent, this,
- &local->preparent, 0);
+ dht_inode_ctx_time_update(local->inode, this, &local->stbuf, 1);
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update(local->loc.parent, this,
+ &local->preparent, 0);
- dht_inode_ctx_time_update (local->loc.parent, this,
- &local->postparent, 1);
- }
+ dht_inode_ctx_time_update(local->loc.parent, this,
+ &local->postparent, 1);
}
+ }
- DHT_STACK_UNWIND (mkdir, frame, op_ret, op_errno,
- local->inode, &local->stbuf, &local->preparent,
- &local->postparent, NULL);
+ DHT_STACK_UNWIND(mkdir, frame, op_ret, op_errno, local->inode,
+ &local->stbuf, &local->preparent, &local->postparent,
+ NULL);
- return 0;
+ return 0;
}
-int
-dht_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
-{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- int ret = -1;
- gf_boolean_t subvol_filled = _gf_false;
- gf_boolean_t dir_exists = _gf_false;
- xlator_t *prev = NULL;
- dht_layout_t *layout = NULL;
-
- local = frame->local;
- prev = cookie;
- layout = local->layout;
-
- subvol_filled = dht_is_subvol_filled (this, prev);
-
- LOCK (&frame->lock);
- {
- if (subvol_filled && (op_ret != -1)) {
- ret = dht_layout_merge (this, layout, prev,
- -1, ENOSPC, NULL);
- } else {
- if (op_ret == -1 && op_errno == EEXIST) {
- /* Very likely just a race between mkdir and
- self-heal (from lookup of a concurrent mkdir
- attempt).
- Ignore error for now. layout setting will
- anyways fail if this was a different (old)
- pre-existing different directory.
- */
- op_ret = 0;
- dir_exists = _gf_true;
- }
- ret = dht_layout_merge (this, layout, prev,
- op_ret, op_errno, NULL);
- }
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_LAYOUT_MERGE_FAILED,
- "%s: failed to merge layouts for subvol %s",
- local->loc.path, prev->name);
-
- if (op_ret == -1) {
- local->op_errno = op_errno;
- goto unlock;
- }
-
- if (dir_exists)
- goto unlock;
-
- dht_iatt_merge (this, &local->stbuf, stbuf, prev);
- dht_iatt_merge (this, &local->preparent, preparent, prev);
- dht_iatt_merge (this, &local->postparent, postparent, prev);
+static int
+dht_mkdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, inode_t *inode, struct iatt *stbuf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ int ret = -1;
+ gf_boolean_t subvol_filled = _gf_false;
+ gf_boolean_t dir_exists = _gf_false;
+ xlator_t *prev = NULL;
+ dht_layout_t *layout = NULL;
+
+ local = frame->local;
+ prev = cookie;
+ layout = local->layout;
+
+ subvol_filled = dht_is_subvol_filled(this, prev);
+
+ LOCK(&frame->lock);
+ {
+ if (subvol_filled && (op_ret != -1)) {
+ ret = dht_layout_merge(this, layout, prev, -1, ENOSPC, NULL);
+ } else {
+ if (op_ret == -1 && op_errno == EEXIST) {
+ /* Very likely just a race between mkdir and
+ self-heal (from lookup of a concurrent mkdir
+ attempt).
+ Ignore error for now. layout setting will
+ anyways fail if this was a different (old)
+ pre-existing different directory.
+ */
+ op_ret = 0;
+ dir_exists = _gf_true;
+ }
+ ret = dht_layout_merge(this, layout, prev, op_ret, op_errno, NULL);
}
-unlock:
- UNLOCK (&frame->lock);
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_LAYOUT_MERGE_FAILED,
+ "%s: failed to merge layouts for subvol %s", local->loc.path,
+ prev->name);
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- /*Unlock entrylk and inodelk once mkdir is done on all subvols*/
- dht_unlock_namespace (frame, &local->lock[0]);
- FRAME_SU_DO (frame, dht_local_t);
- dht_selfheal_new_directory (frame, dht_mkdir_selfheal_cbk,
- layout);
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ goto unlock;
}
- return 0;
-}
+ if (dir_exists)
+ goto unlock;
-int
-dht_mkdir_hashed_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata);
+ dht_iatt_merge(this, &local->stbuf, stbuf);
+ dht_iatt_merge(this, &local->preparent, preparent);
+ dht_iatt_merge(this, &local->postparent, postparent);
+ }
+unlock:
+ UNLOCK(&frame->lock);
-int
-dht_mkdir_helper (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, mode_t umask, dict_t *params)
-{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- int op_errno = -1, ret = -1;
- xlator_t *hashed_subvol = NULL;
- int32_t *parent_disk_layout = NULL;
- dht_layout_t *parent_layout = NULL;
- char pgfid[GF_UUID_BUF_SIZE] = {0};
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
- VALIDATE_OR_GOTO (this->private, err);
-
- gf_uuid_unparse (loc->parent->gfid, pgfid);
-
- conf = this->private;
- local = frame->local;
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt)) {
+ /*Unlock entrylk and inodelk once mkdir is done on all subvols*/
+ dht_unlock_namespace(frame, &local->lock[0]);
+ FRAME_SU_DO(frame, dht_local_t);
+ dht_selfheal_new_directory(frame, dht_mkdir_selfheal_cbk, layout);
+ }
- if (local->op_ret == -1) {
- gf_msg (this->name, GF_LOG_WARNING, local->op_errno,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "mkdir (%s/%s) (path: %s): refreshing parent layout "
- "failed.", pgfid, loc->name,
- loc->path);
+ return 0;
+}
- op_errno = local->op_errno;
- goto err;
- }
+static int
+dht_mkdir_hashed_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata);
- local->op_ret = -1;
+static int
+dht_mkdir_helper(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *params)
+{
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int op_errno = -1, ret = -1;
+ xlator_t *hashed_subvol = NULL;
+ int32_t *parent_disk_layout = NULL;
+ dht_layout_t *parent_layout = NULL;
+ char pgfid[GF_UUID_BUF_SIZE] = {0};
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
+ VALIDATE_OR_GOTO(loc->path, err);
+ VALIDATE_OR_GOTO(this->private, err);
+
+ gf_uuid_unparse(loc->parent->gfid, pgfid);
+
+ conf = this->private;
+ local = frame->local;
+
+ if (local->op_ret == -1) {
+ gf_msg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "mkdir (%s/%s) (path: %s): refreshing parent layout "
+ "failed.",
+ pgfid, loc->name, loc->path);
- hashed_subvol = dht_subvol_get_hashed (this, loc);
- if (hashed_subvol == NULL) {
- gf_msg_debug (this->name, 0,
- "mkdir (%s/%s) (path: %s): hashed subvol not "
- "found", pgfid, loc->name, loc->path);
- op_errno = ENOENT;
- goto err;
- }
+ op_errno = local->op_errno;
+ goto err;
+ }
+
+ local->op_ret = -1;
+
+ hashed_subvol = dht_subvol_get_hashed(this, loc);
+ if (hashed_subvol == NULL) {
+ gf_msg_debug(this->name, 0,
+ "mkdir (%s/%s) (path: %s): hashed subvol not "
+ "found",
+ pgfid, loc->name, loc->path);
+ op_errno = ENOENT;
+ goto err;
+ }
+
+ local->hashed_subvol = hashed_subvol;
+
+ parent_layout = dht_layout_get(this, loc->parent);
+
+ ret = dht_disk_layout_extract_for_subvol(this, parent_layout, hashed_subvol,
+ &parent_disk_layout);
+ if (ret == -1) {
+ gf_msg(this->name, GF_LOG_WARNING, EIO, DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "mkdir (%s/%s) (path: %s): "
+ "extracting in-memory layout of parent failed. ",
+ pgfid, loc->name, loc->path);
+ goto err;
+ }
+
+ if (memcmp(local->parent_disk_layout, parent_disk_layout,
+ sizeof(local->parent_disk_layout)) == 0) {
+ gf_msg(this->name, GF_LOG_WARNING, EIO, DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "mkdir (%s/%s) (path: %s): loop detected. "
+ "parent layout didn't change even though "
+ "previous attempt of mkdir failed because of "
+ "in-memory layout not matching with that on disk.",
+ pgfid, loc->name, loc->path);
+ op_errno = EIO;
+ goto err;
+ }
+
+ memcpy((void *)local->parent_disk_layout, (void *)parent_disk_layout,
+ sizeof(local->parent_disk_layout));
+
+ dht_layout_unref(this, parent_layout);
+ parent_layout = NULL;
+
+ ret = dict_set_str(params, GF_PREOP_PARENT_KEY, conf->xattr_name);
+ if (ret < 0) {
+ local->op_errno = -ret;
+ gf_msg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "mkdir (%s/%s) (path: %s): "
+ "setting %s key in params dictionary failed. ",
+ pgfid, loc->name, loc->path, GF_PREOP_PARENT_KEY);
+ goto err;
+ }
+
+ ret = dict_set_bin(params, conf->xattr_name, parent_disk_layout, 4 * 4);
+ if (ret < 0) {
+ local->op_errno = -ret;
+ gf_msg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "setting parent-layout in params dictionary failed. "
+ "mkdir (%s/%s) (path: %s)",
+ pgfid, loc->name, loc->path);
+ goto err;
+ }
+
+ parent_disk_layout = NULL;
+
+ STACK_WIND_COOKIE(frame, dht_mkdir_hashed_cbk, hashed_subvol, hashed_subvol,
+ hashed_subvol->fops->mkdir, loc, mode, umask, params);
+
+ return 0;
- local->hashed_subvol = hashed_subvol;
+err:
+ dht_unlock_namespace(frame, &local->lock[0]);
- parent_layout = dht_layout_get (this, loc->parent);
+ op_errno = local ? local->op_errno : op_errno;
+ DHT_STACK_UNWIND(mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
- ret = dht_disk_layout_extract_for_subvol (this, parent_layout,
- hashed_subvol,
- &parent_disk_layout);
- if (ret == -1) {
- gf_msg (this->name, GF_LOG_WARNING, EIO,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "mkdir (%s/%s) (path: %s): "
- "extracting in-memory layout of parent failed. ",
- pgfid, loc->name, loc->path);
- goto err;
- }
+ if (parent_disk_layout != NULL)
+ GF_FREE(parent_disk_layout);
- if (memcmp (local->parent_disk_layout, parent_disk_layout,
- sizeof (local->parent_disk_layout)) == 0) {
- gf_msg (this->name, GF_LOG_WARNING, EIO,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "mkdir (%s/%s) (path: %s): loop detected. "
- "parent layout didn't change even though "
- "previous attempt of mkdir failed because of "
- "in-memory layout not matching with that on disk.",
- pgfid, loc->name, loc->path);
- op_errno = EIO;
- goto err;
- }
+ if (parent_layout != NULL)
+ dht_layout_unref(this, parent_layout);
- memcpy ((void *)local->parent_disk_layout, (void *)parent_disk_layout,
- sizeof (local->parent_disk_layout));
+ return 0;
+}
- dht_layout_unref (this, parent_layout);
- parent_layout = NULL;
+static int
+dht_mkdir_hashed_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ int ret = -1;
+ xlator_t *prev = NULL;
+ dht_layout_t *layout = NULL;
+ dht_conf_t *conf = NULL;
+ int i = 0;
+ xlator_t *hashed_subvol = NULL;
+ char pgfid[GF_UUID_BUF_SIZE] = {0};
+ gf_boolean_t parent_layout_changed = _gf_false;
+ call_stub_t *stub = NULL;
+
+ local = frame->local;
+ prev = cookie;
+ layout = local->layout;
+ conf = this->private;
+ hashed_subvol = local->hashed_subvol;
+
+ gf_uuid_unparse(local->loc.parent->gfid, pgfid);
+
+ if (gf_uuid_is_null(local->loc.gfid) && !op_ret)
+ gf_uuid_copy(local->loc.gfid, stbuf->ia_gfid);
+
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
- ret = dict_set_str (params, GF_PREOP_PARENT_KEY, conf->xattr_name);
- if (ret < 0) {
- local->op_errno = -ret;
- gf_msg (this->name, GF_LOG_WARNING, local->op_errno,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "mkdir (%s/%s) (path: %s): "
- "setting %s key in params dictionary failed. ",
- pgfid, loc->name, loc->path, GF_PREOP_PARENT_KEY);
+ parent_layout_changed = (xdata &&
+ dict_get(xdata, GF_PREOP_CHECK_FAILED))
+ ? 1
+ : 0;
+ if (parent_layout_changed) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "mkdir (%s/%s) (path: %s): parent layout "
+ "changed. Attempting a refresh and then a "
+ "retry",
+ pgfid, local->loc.name, local->loc.path);
+
+ stub = fop_mkdir_stub(frame, dht_mkdir_helper, &local->loc,
+ local->mode, local->umask, local->params);
+ if (stub == NULL) {
goto err;
- }
+ }
- ret = dict_set_bin (params, conf->xattr_name, parent_disk_layout,
- 4 * 4);
- if (ret < 0) {
- local->op_errno = -ret;
- gf_msg (this->name, GF_LOG_WARNING, local->op_errno,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "setting parent-layout in params dictionary failed. "
- "mkdir (%s/%s) (path: %s)", pgfid, loc->name,
- loc->path);
+ ret = dht_handle_parent_layout_change(this, stub);
+ if (ret) {
goto err;
- }
-
- parent_disk_layout = NULL;
-
- STACK_WIND_COOKIE (frame, dht_mkdir_hashed_cbk, hashed_subvol,
- hashed_subvol, hashed_subvol->fops->mkdir,
- loc, mode, umask, params);
-
- return 0;
+ }
+ stub = NULL;
+
+ return 0;
+ }
+
+ goto err;
+ }
+
+ dict_del(local->params, GF_PREOP_PARENT_KEY);
+ dict_del(local->params, conf->xattr_name);
+
+ if (dht_is_subvol_filled(this, hashed_subvol))
+ ret = dht_layout_merge(this, layout, prev, -1, ENOSPC, NULL);
+ else
+ ret = dht_layout_merge(this, layout, prev, op_ret, op_errno, NULL);
+
+ /* TODO: we may have to return from the function
+ if layout merge fails. For now, lets just log an error */
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_LAYOUT_MERGE_FAILED,
+ "%s: failed to merge layouts for subvol %s", local->loc.path,
+ prev->name);
+
+ local->op_ret = 0;
+
+ dht_iatt_merge(this, &local->stbuf, stbuf);
+ dht_iatt_merge(this, &local->preparent, preparent);
+ dht_iatt_merge(this, &local->postparent, postparent);
+
+ local->call_cnt = conf->subvolume_cnt - 1;
+ /* Delete internal mds xattr from params dict to avoid store
+ internal mds xattr on other subvols
+ */
+ dict_del(local->params, conf->mds_xattr_key);
+
+ if (gf_uuid_is_null(local->loc.gfid))
+ gf_uuid_copy(local->loc.gfid, stbuf->ia_gfid);
+
+ /* Set hashed subvol as a mds subvol on inode ctx */
+ /*if (!local->inode)
+ local->inode = inode_ref (inode);
+ */
+ ret = dht_inode_ctx_mdsvol_set(local->inode, this, hashed_subvol);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_SET_INODE_CTX_FAILED,
+ "Failed to set hashed subvol for %s on inode vol is %s",
+ local->loc.path, hashed_subvol->name);
+ }
+
+ if (local->call_cnt == 0) {
+ /*Unlock namespace lock once mkdir is done on all subvols*/
+ dht_unlock_namespace(frame, &local->lock[0]);
+ FRAME_SU_DO(frame, dht_local_t);
+ dht_selfheal_directory(frame, dht_mkdir_selfheal_cbk, &local->loc,
+ layout);
+ return 0;
+ }
+
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->subvolumes[i] == hashed_subvol)
+ continue;
+ STACK_WIND_COOKIE(frame, dht_mkdir_cbk, conf->subvolumes[i],
+ conf->subvolumes[i], conf->subvolumes[i]->fops->mkdir,
+ &local->loc, local->mode, local->umask,
+ local->params);
+ }
+
+ return 0;
err:
- dht_unlock_namespace (frame, &local->lock[0]);
+ if (local->op_ret != 0) {
+ dht_unlock_namespace(frame, &local->lock[0]);
+ }
- op_errno = local ? local->op_errno : op_errno;
- DHT_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL,
- NULL, NULL);
+ DHT_STACK_UNWIND(mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
- if (parent_disk_layout != NULL)
- GF_FREE (parent_disk_layout);
+ return 0;
+}
- if (parent_layout != NULL)
- dht_layout_unref (this, parent_layout);
+static int
+dht_mkdir_guard_parent_layout_cbk(call_frame_t *frame, xlator_t *this,
+ loc_t *loc, mode_t mode, mode_t umask,
+ dict_t *params)
+{
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = 0;
+ char pgfid[GF_UUID_BUF_SIZE] = {0};
+ int ret = -1;
+ int32_t zero[1] = {0};
+
+ local = frame->local;
+ conf = this->private;
+
+ gf_uuid_unparse(loc->parent->gfid, pgfid);
+
+ if (local->op_ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "mkdir (%s/%s) (path: %s): "
+ "Acquiring lock on parent to guard against "
+ "layout-change failed.",
+ pgfid, loc->name, loc->path);
+ goto err;
+ }
+
+ local->op_ret = -1;
+ /* Add internal MDS xattr on disk for hashed subvol
+ */
+ ret = dht_dict_set_array(params, conf->mds_xattr_key, zero, 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary value:key = %s for "
+ "path %s",
+ conf->mds_xattr_key, loc->path);
+ }
+
+ STACK_WIND_COOKIE(frame, dht_mkdir_hashed_cbk, local->hashed_subvol,
+ local->hashed_subvol, local->hashed_subvol->fops->mkdir,
+ loc, mode, umask, params);
+
+ return 0;
+err:
+ DHT_STACK_UNWIND(mkdir, frame, -1, local->op_errno, NULL, NULL, NULL, NULL,
+ NULL);
- return 0;
+ return 0;
}
int
-dht_mkdir_hashed_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
-{
- dht_local_t *local = NULL;
- int ret = -1;
- xlator_t *prev = NULL;
- dht_layout_t *layout = NULL;
- dht_conf_t *conf = NULL;
- int i = 0;
- xlator_t *hashed_subvol = NULL;
- char pgfid[GF_UUID_BUF_SIZE] = {0};
- gf_boolean_t parent_layout_changed = _gf_false;
- call_stub_t *stub = NULL;
-
- VALIDATE_OR_GOTO (this->private, err);
-
- local = frame->local;
- prev = cookie;
- layout = local->layout;
- conf = this->private;
- hashed_subvol = local->hashed_subvol;
-
- gf_uuid_unparse (local->loc.parent->gfid, pgfid);
-
- if (gf_uuid_is_null (local->loc.gfid) && !op_ret)
- gf_uuid_copy (local->loc.gfid, stbuf->ia_gfid);
+dht_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *params)
+{
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int op_errno = EINVAL, ret = -1;
+ xlator_t *hashed_subvol = NULL;
+ char pgfid[GF_UUID_BUF_SIZE] = {0};
+ call_stub_t *stub = NULL;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
+ VALIDATE_OR_GOTO(loc->path, err);
+ VALIDATE_OR_GOTO(this->private, err);
+
+ gf_uuid_unparse(loc->parent->gfid, pgfid);
+
+ conf = this->private;
+
+ if (!params || !dict_get(params, "gfid-req")) {
+ op_errno = EPERM;
+ gf_msg_callingfn(this->name, GF_LOG_WARNING, op_errno,
+ DHT_MSG_GFID_NULL,
+ "mkdir: %s is received "
+ "without gfid-req %p",
+ loc->path, params);
+ goto err;
+ }
+
+ dht_get_du_info(frame, this, loc);
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_MKDIR);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ hashed_subvol = dht_subvol_get_hashed(this, loc);
+ if (hashed_subvol == NULL) {
+ gf_msg_debug(this->name, 0, "hashed subvol not found for %s",
+ loc->path);
+ local->op_errno = EIO;
+ goto err;
+ }
+
+ local->hashed_subvol = hashed_subvol;
+ local->mode = mode;
+ local->umask = umask;
+ if (params)
+ local->params = dict_ref(params);
+
+ local->inode = inode_ref(loc->inode);
+
+ local->layout = dht_layout_new(this, conf->subvolume_cnt);
+ if (!local->layout) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ /* set the newly created directory hash to the commit hash
+ * if the configuration option is set. If configuration option
+ * is not set, the older clients may still be connecting to the
+ * volume and hence we need to preserve the 1 in disk[0] part of the
+ * layout xattr */
+ if (conf->lookup_optimize)
+ local->layout->commit_hash = conf->vol_commit_hash;
+ else
+ local->layout->commit_hash = DHT_LAYOUT_HASH_INVALID;
+
+ stub = fop_mkdir_stub(frame, dht_mkdir_guard_parent_layout_cbk, loc, mode,
+ umask, params);
+ if (stub == NULL) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM,
+ DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "mkdir (%s/%s) (path: %s): "
+ "creating stub failed.",
+ pgfid, loc->name, loc->path);
+ local->op_errno = ENOMEM;
+ goto err;
+ }
+
+ ret = dht_guard_parent_layout_and_namespace(this, stub);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_PARENT_LAYOUT_CHANGED,
+ "mkdir (%s/%s) (path: %s) cannot wind lock request to "
+ "guard parent layout",
+ pgfid, loc->name, loc->path);
+ goto err;
+ }
+
+ return 0;
- if (op_ret == -1) {
- local->op_errno = op_errno;
-
- parent_layout_changed = (xdata && dict_get (xdata, GF_PREOP_CHECK_FAILED))
- ? 1 : 0;
- if (parent_layout_changed) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "mkdir (%s/%s) (path: %s): parent layout "
- "changed. Attempting a refresh and then a "
- "retry", pgfid, local->loc.name,
- local->loc.path);
-
- stub = fop_mkdir_stub (frame, dht_mkdir_helper,
- &local->loc, local->mode,
- local->umask, local->params);
- if (stub == NULL) {
- goto err;
- }
-
- dht_handle_parent_layout_change (this, stub);
- stub = NULL;
-
- return 0;
- }
-
- goto err;
- }
-
- dict_del (local->params, GF_PREOP_PARENT_KEY);
- dict_del (local->params, conf->xattr_name);
-
- if (dht_is_subvol_filled (this, hashed_subvol))
- ret = dht_layout_merge (this, layout, prev,
- -1, ENOSPC, NULL);
- else
- ret = dht_layout_merge (this, layout, prev,
- op_ret, op_errno, NULL);
-
- /* TODO: we may have to return from the function
- if layout merge fails. For now, lets just log an error */
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_LAYOUT_MERGE_FAILED,
- "%s: failed to merge layouts for subvol %s",
- local->loc.path, prev->name);
-
- local->op_ret = 0;
-
- dht_iatt_merge (this, &local->stbuf, stbuf, prev);
- dht_iatt_merge (this, &local->preparent, preparent, prev);
- dht_iatt_merge (this, &local->postparent, postparent, prev);
-
- local->call_cnt = conf->subvolume_cnt - 1;
- /* Delete internal mds xattr from params dict to avoid store
- internal mds xattr on other subvols
- */
- dict_del (local->params, conf->mds_xattr_key);
-
- if (gf_uuid_is_null (local->loc.gfid))
- gf_uuid_copy (local->loc.gfid, stbuf->ia_gfid);
+err:
+ op_errno = local ? local->op_errno : op_errno;
+ DHT_STACK_UNWIND(mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
- /* Set hashed subvol as a mds subvol on inode ctx */
- /*if (!local->inode)
- local->inode = inode_ref (inode);
- */
- ret = dht_inode_ctx_mdsvol_set (local->inode, this, hashed_subvol);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, DHT_MSG_SET_INODE_CTX_FAILED,
- "Failed to set hashed subvol for %s on inode vol is %s",
- local->loc.path, hashed_subvol->name);
- }
+ return 0;
+}
- if (local->call_cnt == 0) {
- /*Unlock namespace lock once mkdir is done on all subvols*/
- dht_unlock_namespace (frame, &local->lock[0]);
- FRAME_SU_DO (frame, dht_local_t);
- dht_selfheal_directory (frame, dht_mkdir_selfheal_cbk,
- &local->loc, layout);
- return 0;
- }
+static int
+dht_rmdir_selfheal_cbk(call_frame_t *heal_frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ dht_local_t *heal_local = NULL;
+ call_frame_t *main_frame = NULL;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->subvolumes[i] == hashed_subvol)
- continue;
- STACK_WIND_COOKIE (frame, dht_mkdir_cbk, conf->subvolumes[i],
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->mkdir,
- &local->loc, local->mode, local->umask,
- local->params);
- }
+ heal_local = heal_frame->local;
+ main_frame = heal_local->main_frame;
+ local = main_frame->local;
- return 0;
-err:
- if (local->op_ret != 0) {
- dht_unlock_namespace (frame, &local->lock[0]);
- }
+ DHT_STACK_DESTROY(heal_frame);
+ dht_set_fixed_dir_stat(&local->preparent);
+ dht_set_fixed_dir_stat(&local->postparent);
- DHT_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL,
- NULL, NULL);
+ DHT_STACK_UNWIND(rmdir, main_frame, local->op_ret, local->op_errno,
+ &local->preparent, &local->postparent, NULL);
- return 0;
+ return 0;
}
-int
-dht_mkdir_guard_parent_layout_cbk (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, mode_t umask,
- dict_t *params)
-{
- dht_local_t *local = NULL;
- dht_conf_t *conf = 0;
- char pgfid[GF_UUID_BUF_SIZE] = {0};
- int ret = -1;
- int32_t zero[1] = {0};
-
- local = frame->local;
- conf = this->private;
-
- gf_uuid_unparse (loc->parent->gfid, pgfid);
+static int
+dht_rmdir_hashed_subvol_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ dht_local_t *heal_local = NULL;
+ call_frame_t *heal_frame = NULL;
+ dht_conf_t *conf = NULL;
+ int this_call_cnt = 0;
+ xlator_t *prev = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+
+ local = frame->local;
+ prev = cookie;
+ conf = this->private;
+
+ gf_uuid_unparse(local->loc.gfid, gfid);
+
+ LOCK(&frame->lock);
+ {
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ local->op_ret = -1;
+ if (conf->subvolume_cnt != 1) {
+ if (op_errno != ENOENT && op_errno != EACCES &&
+ op_errno != ESTALE) {
+ local->need_selfheal = 1;
+ }
+ }
- if (local->op_ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, local->op_errno,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "mkdir (%s/%s) (path: %s): "
- "Acquiring lock on parent to guard against "
- "layout-change failed.", pgfid, loc->name, loc->path);
- goto err;
+ gf_msg_debug(this->name, op_errno,
+ "rmdir on %s for %s failed "
+ "(gfid = %s)",
+ prev->name, local->loc.path, gfid);
+ goto unlock;
}
- local->op_ret = -1;
- /* Add internal MDS xattr on disk for hashed subvol
- */
- ret = dht_dict_set_array (params, conf->mds_xattr_key, zero, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value:key = %s for "
- "path %s", conf->mds_xattr_key, loc->path);
- }
+ dht_iatt_merge(this, &local->preparent, preparent);
+ dht_iatt_merge(this, &local->postparent, postparent);
+ }
+unlock:
+ UNLOCK(&frame->lock);
- STACK_WIND_COOKIE (frame, dht_mkdir_hashed_cbk, local->hashed_subvol,
- local->hashed_subvol,
- local->hashed_subvol->fops->mkdir,
- loc, mode, umask, params);
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt)) {
+ if (local->need_selfheal) {
+ dht_rmdir_unlock(frame, this);
+ local->layout = dht_layout_get(this, local->loc.inode);
- return 0;
-err:
- DHT_STACK_UNWIND (mkdir, frame, -1, local->op_errno, NULL, NULL, NULL,
- NULL, NULL);
+ /* TODO: neater interface needed below */
+ local->stbuf.ia_type = local->loc.inode->ia_type;
- return 0;
-}
+ gf_uuid_copy(local->gfid, local->loc.inode->gfid);
-int
-dht_mkdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, mode_t umask, dict_t *params)
-{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- int op_errno = -1, ret = -1;
- xlator_t *hashed_subvol = NULL;
- char pgfid[GF_UUID_BUF_SIZE] = {0};
- call_stub_t *stub = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
- VALIDATE_OR_GOTO (this->private, err);
-
- gf_uuid_unparse (loc->parent->gfid, pgfid);
-
- conf = this->private;
-
- if (!params || !dict_get (params, "gfid-req")) {
- op_errno = EPERM;
- gf_msg_callingfn (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_GFID_NULL, "mkdir: %s is received "
- "without gfid-req %p", loc->path, params);
- goto err;
- }
+ /* Use a different frame or else the rmdir op_ret is
+ * overwritten by that of the selfheal */
- dht_get_du_info (frame, this, loc);
+ heal_frame = copy_frame(frame);
- local = dht_local_init (frame, loc, NULL, GF_FOP_MKDIR);
- if (!local) {
- op_errno = ENOMEM;
+ if (heal_frame == NULL) {
goto err;
- }
+ }
- hashed_subvol = dht_subvol_get_hashed (this, loc);
- if (hashed_subvol == NULL) {
- gf_msg_debug (this->name, 0,
- "hashed subvol not found for %s",
- loc->path);
- local->op_errno = EIO;
+ heal_local = dht_local_init(heal_frame, &local->loc, NULL, 0);
+ if (!heal_local) {
+ DHT_STACK_DESTROY(heal_frame);
goto err;
- }
-
-
- local->hashed_subvol = hashed_subvol;
- local->mode = mode;
- local->umask = umask;
- if (params)
- local->params = dict_ref (params);
-
- local->inode = inode_ref (loc->inode);
+ }
- local->layout = dht_layout_new (this, conf->subvolume_cnt);
- if (!local->layout) {
- op_errno = ENOMEM;
- goto err;
- }
+ heal_local->inode = inode_ref(local->loc.inode);
+ heal_local->main_frame = frame;
+ gf_uuid_copy(heal_local->gfid, local->loc.inode->gfid);
- /* set the newly created directory hash to the commit hash
- * if the configuration option is set. If configuration option
- * is not set, the older clients may still be connecting to the
- * volume and hence we need to preserve the 1 in disk[0] part of the
- * layout xattr */
- if (conf->lookup_optimize)
- local->layout->commit_hash = conf->vol_commit_hash;
- else
- local->layout->commit_hash = DHT_LAYOUT_HASH_INVALID;
+ dht_selfheal_restore(heal_frame, dht_rmdir_selfheal_cbk,
+ &heal_local->loc, heal_local->layout);
+ return 0;
+ } else {
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update(local->loc.parent, this,
+ &local->preparent, 0);
+ dht_inode_ctx_time_update(local->loc.parent, this,
+ &local->postparent, 1);
+ }
- stub = fop_mkdir_stub (frame, dht_mkdir_guard_parent_layout_cbk, loc,
- mode, umask, params);
- if (stub == NULL) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "mkdir (%s/%s) (path: %s): "
- "creating stub failed.", pgfid, loc->name, loc->path);
- local->op_errno = ENOMEM;
- goto err;
- }
+ dht_set_fixed_dir_stat(&local->preparent);
+ dht_set_fixed_dir_stat(&local->postparent);
- ret = dht_guard_parent_layout_and_namespace (this, stub);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "mkdir (%s/%s) (path: %s) cannot wind lock request to "
- "guard parent layout", pgfid, loc->name, loc->path);
- goto err;
+ dht_rmdir_unlock(frame, this);
+ DHT_STACK_UNWIND(rmdir, frame, local->op_ret, local->op_errno,
+ &local->preparent, &local->postparent, NULL);
}
+ }
- return 0;
+ return 0;
err:
- op_errno = local ? local->op_errno : op_errno;
- DHT_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL,
- NULL, NULL);
-
- return 0;
+ DHT_STACK_UNWIND(rmdir, frame, local->op_ret, local->op_errno, NULL, NULL,
+ NULL);
+ return 0;
}
-
-int
-dht_rmdir_selfheal_cbk (call_frame_t *heal_frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+static int
+dht_rmdir_unlock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- dht_local_t *heal_local = NULL;
- call_frame_t *main_frame = NULL;
-
- heal_local = heal_frame->local;
- main_frame = heal_local->main_frame;
- local = main_frame->local;
-
- DHT_STACK_DESTROY (heal_frame);
- dht_set_fixed_dir_stat (&local->preparent);
- dht_set_fixed_dir_stat (&local->postparent);
-
- DHT_STACK_UNWIND (rmdir, main_frame, local->op_ret, local->op_errno,
- &local->preparent, &local->postparent, NULL);
-
- return 0;
+ DHT_STACK_DESTROY(frame);
+ return 0;
}
+static int
+dht_rmdir_unlock(call_frame_t *frame, xlator_t *this)
+{
+ dht_local_t *local = NULL, *lock_local = NULL;
+ call_frame_t *lock_frame = NULL;
+ int lock_count = 0;
-int
-dht_rmdir_hashed_subvol_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- dht_local_t *local = NULL;
- dht_local_t *heal_local = NULL;
- call_frame_t *heal_frame = NULL;
- dht_conf_t *conf = NULL;
- int this_call_cnt = 0;
- xlator_t *prev = NULL;
- char gfid[GF_UUID_BUF_SIZE] ={0};
-
- local = frame->local;
- prev = cookie;
- conf = this->private;
-
- gf_uuid_unparse(local->loc.gfid, gfid);
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- local->op_ret = -1;
- if (conf->subvolume_cnt != 1) {
- if (op_errno != ENOENT && op_errno != EACCES
- && op_errno != ESTALE) {
- local->need_selfheal = 1;
- }
- }
-
- gf_msg_debug (this->name, op_errno,
- "rmdir on %s for %s failed "
- "(gfid = %s)",
- prev->name, local->loc.path,
- gfid);
- goto unlock;
- }
-
- dht_iatt_merge (this, &local->preparent, preparent, prev);
- dht_iatt_merge (this, &local->postparent, postparent, prev);
-
- }
-unlock:
- UNLOCK (&frame->lock);
-
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- if (local->need_selfheal) {
- dht_rmdir_unlock (frame, this);
- local->layout =
- dht_layout_get (this, local->loc.inode);
-
- /* TODO: neater interface needed below */
- local->stbuf.ia_type = local->loc.inode->ia_type;
-
- gf_uuid_copy (local->gfid, local->loc.inode->gfid);
+ local = frame->local;
- /* Use a different frame or else the rmdir op_ret is
- * overwritten by that of the selfheal */
+ /* Unlock entrylk */
+ dht_unlock_entrylk_wrapper(frame, &local->lock[0].ns.directory_ns);
- heal_frame = copy_frame (frame);
+ /* Unlock inodelk */
+ lock_count = dht_lock_count(local->lock[0].ns.parent_layout.locks,
+ local->lock[0].ns.parent_layout.lk_count);
- if (heal_frame == NULL) {
- goto err;
- }
+ if (lock_count == 0)
+ goto done;
- heal_local = dht_local_init (heal_frame,
- &local->loc,
- NULL, 0);
- if (!heal_local) {
- DHT_STACK_DESTROY (heal_frame);
- goto err;
- }
+ lock_frame = copy_frame(frame);
+ if (lock_frame == NULL)
+ goto done;
- heal_local->inode = inode_ref (local->loc.inode);
- heal_local->main_frame = frame;
- gf_uuid_copy (heal_local->gfid, local->loc.inode->gfid);
-
- dht_selfheal_restore (heal_frame,
- dht_rmdir_selfheal_cbk,
- &heal_local->loc,
- heal_local->layout);
- return 0;
- } else {
-
- if (local->loc.parent) {
- dht_inode_ctx_time_update (local->loc.parent,
- this,
- &local->preparent,
- 0);
-
- dht_inode_ctx_time_update (local->loc.parent,
- this,
- &local->postparent,
- 1);
- }
+ lock_local = dht_local_init(lock_frame, &local->loc, NULL,
+ lock_frame->root->op);
+ if (lock_local == NULL)
+ goto done;
- dht_set_fixed_dir_stat (&local->preparent);
- dht_set_fixed_dir_stat (&local->postparent);
+ lock_local->lock[0].ns.parent_layout.locks = local->lock[0]
+ .ns.parent_layout.locks;
+ lock_local->lock[0]
+ .ns.parent_layout.lk_count = local->lock[0].ns.parent_layout.lk_count;
- dht_rmdir_unlock (frame, this);
- DHT_STACK_UNWIND (rmdir, frame, local->op_ret,
- local->op_errno, &local->preparent,
- &local->postparent, NULL);
- }
- }
+ local->lock[0].ns.parent_layout.locks = NULL;
+ local->lock[0].ns.parent_layout.lk_count = 0;
+ dht_unlock_inodelk(lock_frame, lock_local->lock[0].ns.parent_layout.locks,
+ lock_local->lock[0].ns.parent_layout.lk_count,
+ dht_rmdir_unlock_cbk);
+ lock_frame = NULL;
- return 0;
-
-err:
- DHT_STACK_UNWIND (rmdir, frame, local->op_ret,
- local->op_errno, NULL, NULL, NULL);
- return 0;
+done:
+ if (lock_frame != NULL) {
+ DHT_STACK_DESTROY(lock_frame);
+ }
+ return 0;
}
+static int
+dht_rmdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ xlator_t *prev = NULL;
+ int done = 0;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ dht_local_t *heal_local = NULL;
+ call_frame_t *heal_frame = NULL;
+ int ret = -1;
+
+ local = frame->local;
+ prev = cookie;
+
+ LOCK(&frame->lock);
+ {
+ if (op_ret == -1) {
+ if ((op_errno != ENOENT) && (op_errno != ESTALE)) {
+ local->op_errno = op_errno;
+ local->op_ret = -1;
-int
-dht_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- xlator_t *prev = NULL;
- int done = 0;
- char gfid[GF_UUID_BUF_SIZE] ={0};
- dht_local_t *heal_local = NULL;
- call_frame_t *heal_frame = NULL;
- int ret = -1;
-
- local = frame->local;
- prev = cookie;
-
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- if ((op_errno != ENOENT) && (op_errno != ESTALE)) {
- local->op_errno = op_errno;
- local->op_ret = -1;
-
- if (op_errno != EACCES)
- local->need_selfheal = 1;
- }
-
- gf_uuid_unparse(local->loc.gfid, gfid);
-
- gf_msg_debug (this->name, op_errno,
- "rmdir on %s for %s failed."
- "(gfid = %s)",
- prev->name, local->loc.path,
- gfid);
- goto unlock;
- }
-
- /* Track if rmdir succeeded on atleast one subvol*/
- local->fop_succeeded = 1;
- dht_iatt_merge (this, &local->preparent, preparent, prev);
- dht_iatt_merge (this, &local->postparent, postparent, prev);
- }
-unlock:
- UNLOCK (&frame->lock);
-
+ if (op_errno != EACCES)
+ local->need_selfheal = 1;
+ }
- this_call_cnt = dht_frame_return (frame);
+ gf_uuid_unparse(local->loc.gfid, gfid);
- /* if local->hashed_subvol, we are yet to wind to hashed_subvol. */
- if (local->hashed_subvol && (this_call_cnt == 1)) {
- done = 1;
- } else if (!local->hashed_subvol && !this_call_cnt) {
- done = 1;
+ gf_msg_debug(this->name, op_errno,
+ "rmdir on %s for %s failed."
+ "(gfid = %s)",
+ prev->name, local->loc.path, gfid);
+ goto unlock;
}
+ /* Track if rmdir succeeded on at least one subvol*/
+ local->fop_succeeded = 1;
+ dht_iatt_merge(this, &local->preparent, preparent);
+ dht_iatt_merge(this, &local->postparent, postparent);
+ }
+unlock:
+ UNLOCK(&frame->lock);
- if (done) {
- if (local->need_selfheal && local->fop_succeeded) {
- dht_rmdir_unlock (frame, this);
- local->layout =
- dht_layout_get (this, local->loc.inode);
+ this_call_cnt = dht_frame_return(frame);
- /* TODO: neater interface needed below */
- local->stbuf.ia_type = local->loc.inode->ia_type;
+ /* if local->hashed_subvol, we are yet to wind to hashed_subvol. */
+ if (local->hashed_subvol && (this_call_cnt == 1)) {
+ done = 1;
+ } else if (!local->hashed_subvol && !this_call_cnt) {
+ done = 1;
+ }
- gf_uuid_copy (local->gfid, local->loc.inode->gfid);
- heal_frame = copy_frame (frame);
- if (heal_frame == NULL) {
- goto err;
- }
+ if (done) {
+ if (local->need_selfheal && local->fop_succeeded) {
+ dht_rmdir_unlock(frame, this);
+ local->layout = dht_layout_get(this, local->loc.inode);
- heal_local = dht_local_init (heal_frame, &local->loc,
- NULL, 0);
- if (!heal_local) {
- DHT_STACK_DESTROY (heal_frame);
- goto err;
- }
+ /* TODO: neater interface needed below */
+ local->stbuf.ia_type = local->loc.inode->ia_type;
- heal_local->inode = inode_ref (local->loc.inode);
- heal_local->main_frame = frame;
- gf_uuid_copy (heal_local->gfid, local->loc.inode->gfid);
- ret = dht_selfheal_restore (heal_frame,
- dht_rmdir_selfheal_cbk,
- &heal_local->loc,
- heal_local->layout);
- if (ret) {
- DHT_STACK_DESTROY (heal_frame);
- goto err;
- }
+ gf_uuid_copy(local->gfid, local->loc.inode->gfid);
+ heal_frame = copy_frame(frame);
+ if (heal_frame == NULL) {
+ goto err;
+ }
- } else if (this_call_cnt) {
- /* If non-hashed subvol's have responded, proceed */
- if (local->op_ret == 0) {
- /* Delete the dir from the hashed subvol if:
- * The fop succeeded on at least one subvol
- * and did not fail on any
- * or
- * The fop failed with ENOENT/ESTALE on
- * all subvols */
-
- STACK_WIND_COOKIE (frame, dht_rmdir_hashed_subvol_cbk,
- local->hashed_subvol,
- local->hashed_subvol,
- local->hashed_subvol->fops->rmdir,
- &local->loc, local->flags, NULL);
- } else {
- /* hashed-subvol was non-NULL and rmdir failed on
- * all non hashed-subvols. Unwind rmdir with
- * local->op_ret and local->op_errno. */
- dht_rmdir_unlock (frame, this);
- DHT_STACK_UNWIND (rmdir, frame, local->op_ret,
- local->op_errno, &local->preparent,
- &local->postparent, NULL);
-
- return 0;
+ heal_local = dht_local_init(heal_frame, &local->loc, NULL, 0);
+ if (!heal_local) {
+ DHT_STACK_DESTROY(heal_frame);
+ goto err;
+ }
- }
- } else if (!this_call_cnt) {
- /* All subvol's have responded, proceed */
+ heal_local->inode = inode_ref(local->loc.inode);
+ heal_local->main_frame = frame;
+ gf_uuid_copy(heal_local->gfid, local->loc.inode->gfid);
+ ret = dht_selfheal_restore(heal_frame, dht_rmdir_selfheal_cbk,
+ &heal_local->loc, heal_local->layout);
+ if (ret) {
+ DHT_STACK_DESTROY(heal_frame);
+ goto err;
+ }
- if (local->loc.parent) {
+ } else if (this_call_cnt) {
+ /* If non-hashed subvol's have responded, proceed */
+ if (local->op_ret == 0) {
+ /* Delete the dir from the hashed subvol if:
+ * The fop succeeded on at least one subvol
+ * and did not fail on any
+ * or
+ * The fop failed with ENOENT/ESTALE on
+ * all subvols */
+
+ STACK_WIND_COOKIE(frame, dht_rmdir_hashed_subvol_cbk,
+ local->hashed_subvol, local->hashed_subvol,
+ local->hashed_subvol->fops->rmdir,
+ &local->loc, local->flags, NULL);
+ } else {
+ /* hashed-subvol was non-NULL and rmdir failed on
+ * all non hashed-subvols. Unwind rmdir with
+ * local->op_ret and local->op_errno. */
+ dht_rmdir_unlock(frame, this);
+ DHT_STACK_UNWIND(rmdir, frame, local->op_ret, local->op_errno,
+ &local->preparent, &local->postparent, NULL);
- dht_inode_ctx_time_update (local->loc.parent,
- this,
- &local->preparent,
- 0);
+ return 0;
+ }
+ } else if (!this_call_cnt) {
+ /* All subvol's have responded, proceed */
- dht_inode_ctx_time_update (local->loc.parent,
- this,
- &local->postparent,
- 1);
+ if (local->loc.parent) {
+ dht_inode_ctx_time_update(local->loc.parent, this,
+ &local->preparent, 0);
- }
+ dht_inode_ctx_time_update(local->loc.parent, this,
+ &local->postparent, 1);
+ }
- dht_set_fixed_dir_stat (&local->preparent);
- dht_set_fixed_dir_stat (&local->postparent);
+ dht_set_fixed_dir_stat(&local->preparent);
+ dht_set_fixed_dir_stat(&local->postparent);
- dht_rmdir_unlock (frame, this);
- DHT_STACK_UNWIND (rmdir, frame, local->op_ret,
- local->op_errno, &local->preparent,
- &local->postparent, NULL);
- }
+ dht_rmdir_unlock(frame, this);
+ DHT_STACK_UNWIND(rmdir, frame, local->op_ret, local->op_errno,
+ &local->preparent, &local->postparent, NULL);
}
+ }
- return 0;
+ return 0;
err:
- DHT_STACK_UNWIND (rmdir, frame, -1, local->op_errno, NULL, NULL, NULL);
- return 0;
-
+ DHT_STACK_UNWIND(rmdir, frame, -1, local->op_errno, NULL, NULL, NULL);
+ return 0;
}
-
-int
-dht_rmdir_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- DHT_STACK_DESTROY (frame);
- return 0;
-}
-
-
-int
-dht_rmdir_unlock (call_frame_t *frame, xlator_t *this)
+static int
+dht_rmdir_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL, *lock_local = NULL;
- call_frame_t *lock_frame = NULL;
- int lock_count = 0;
-
- local = frame->local;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int i = 0;
+ xlator_t *hashed_subvol;
- /* Unlock entrylk */
- dht_unlock_entrylk_wrapper (frame, &local->lock[0].ns.directory_ns);
+ conf = this->private;
+ local = frame->local;
- /* Unlock inodelk */
- lock_count = dht_lock_count (local->lock[0].ns.parent_layout.locks,
- local->lock[0].ns.parent_layout.lk_count);
+ if (op_ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, DHT_MSG_INODE_LK_ERROR,
+ "acquiring entrylk after inodelk failed rmdir for %s)",
+ local->loc.path);
- if (lock_count == 0)
- goto done;
-
- lock_frame = copy_frame (frame);
- if (lock_frame == NULL)
- goto done;
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ goto err;
+ }
- lock_local = dht_local_init (lock_frame, &local->loc, NULL,
- lock_frame->root->op);
- if (lock_local == NULL)
- goto done;
+ hashed_subvol = local->hashed_subvol;
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (hashed_subvol && (hashed_subvol == conf->subvolumes[i]))
+ continue;
- lock_local->lock[0].ns.parent_layout.locks = local->lock[0].ns.parent_layout.locks;
- lock_local->lock[0].ns.parent_layout.lk_count = local->lock[0].ns.parent_layout.lk_count;
+ STACK_WIND_COOKIE(frame, dht_rmdir_cbk, conf->subvolumes[i],
+ conf->subvolumes[i], conf->subvolumes[i]->fops->rmdir,
+ &local->loc, local->flags, NULL);
+ }
- local->lock[0].ns.parent_layout.locks = NULL;
- local->lock[0].ns.parent_layout.lk_count = 0;
- dht_unlock_inodelk (lock_frame,
- lock_local->lock[0].ns.parent_layout.locks,
- lock_local->lock[0].ns.parent_layout.lk_count,
- dht_rmdir_unlock_cbk);
- lock_frame = NULL;
+ return 0;
-done:
- if (lock_frame != NULL) {
- DHT_STACK_DESTROY (lock_frame);
- }
+err:
+ DHT_STACK_UNWIND(rmdir, frame, local->op_ret, local->op_errno,
+ &local->preparent, &local->postparent, NULL);
- return 0;
+ return 0;
}
-
-int
-dht_rmdir_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+static int
+dht_rmdir_do(call_frame_t *frame, xlator_t *this)
{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- int i = 0;
- xlator_t *hashed_subvol;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int ret = -1;
+ xlator_t *hashed_subvol = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- VALIDATE_OR_GOTO (this->private, err);
+ VALIDATE_OR_GOTO(frame->local, err);
+ local = frame->local;
+ VALIDATE_OR_GOTO(this->private, out);
+ conf = this->private;
- conf = this->private;
- local = frame->local;
+ if (local->op_ret == -1)
+ goto out;
- if (op_ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_INODE_LK_ERROR,
- "acquiring entrylk after inodelk failed rmdir for %s)",
- local->loc.path);
+ local->call_cnt = conf->subvolume_cnt;
- local->op_ret = -1;
- local->op_errno = op_errno;
- goto err;
- }
+ /* first remove from non-hashed_subvol */
+ hashed_subvol = dht_subvol_get_hashed(this, &local->loc);
- hashed_subvol = local->hashed_subvol;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (hashed_subvol &&
- (hashed_subvol == conf->subvolumes[i]))
- continue;
+ if (!hashed_subvol) {
+ gf_uuid_unparse(local->loc.gfid, gfid);
- STACK_WIND_COOKIE (frame, dht_rmdir_cbk, conf->subvolumes[i],
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->rmdir,
- &local->loc, local->flags, NULL);
- }
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_HASHED_SUBVOL_GET_FAILED,
+ "Failed to get hashed subvol for %s (gfid = %s)",
+ local->loc.path, gfid);
+ } else {
+ local->hashed_subvol = hashed_subvol;
+ }
+ /* When DHT has only 1 child */
+ if (conf->subvolume_cnt == 1) {
+ STACK_WIND_COOKIE(frame, dht_rmdir_hashed_subvol_cbk,
+ conf->subvolumes[0], conf->subvolumes[0],
+ conf->subvolumes[0]->fops->rmdir, &local->loc,
+ local->flags, NULL);
return 0;
+ }
-err:
- DHT_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
- &local->preparent, &local->postparent, NULL);
+ local->current = &local->lock[0];
+ ret = dht_protect_namespace(frame, &local->loc, local->hashed_subvol,
+ &local->current->ns, dht_rmdir_lock_cbk);
+ if (ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = errno ? errno : EINVAL;
+ goto out;
+ }
- return 0;
-}
+ return 0;
+out:
+ dht_set_fixed_dir_stat(&local->preparent);
+ dht_set_fixed_dir_stat(&local->postparent);
-int
-dht_rmdir_do (call_frame_t *frame, xlator_t *this)
-{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- int ret = -1;
- xlator_t *hashed_subvol = NULL;
- char gfid[GF_UUID_BUF_SIZE] ={0};
+ DHT_STACK_UNWIND(rmdir, frame, local->op_ret, local->op_errno,
+ &local->preparent, &local->postparent, NULL);
+ return 0;
+err:
+ DHT_STACK_UNWIND(rmdir, frame, -1, EINVAL, NULL, NULL, NULL);
+ return 0;
+}
- VALIDATE_OR_GOTO (this->private, err);
+static void
+dht_rmdir_readdirp_done(call_frame_t *readdirp_frame, xlator_t *this)
+{
+ call_frame_t *main_frame = NULL;
+ dht_local_t *main_local = NULL;
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
- conf = this->private;
- local = frame->local;
+ local = readdirp_frame->local;
+ main_frame = local->main_frame;
+ main_local = main_frame->local;
- if (local->op_ret == -1)
- goto err;
+ /* At least one readdirp failed.
+ * This is a bit hit or miss - if readdirp failed on more than
+ * one subvol, we don't know which error is returned.
+ */
+ if (local->op_ret == -1) {
+ main_local->op_ret = local->op_ret;
+ main_local->op_errno = local->op_errno;
+ }
- local->call_cnt = conf->subvolume_cnt;
+ this_call_cnt = dht_frame_return(main_frame);
- /* first remove from non-hashed_subvol */
- hashed_subvol = dht_subvol_get_hashed (this, &local->loc);
+ if (is_last_call(this_call_cnt))
+ dht_rmdir_do(main_frame, this);
- if (!hashed_subvol) {
- gf_uuid_unparse(local->loc.gfid, gfid);
+ DHT_STACK_DESTROY(readdirp_frame);
+}
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_HASHED_SUBVOL_GET_FAILED,
- "Failed to get hashed subvol for %s (gfid = %s)",
- local->loc.path, gfid);
- } else {
- local->hashed_subvol = hashed_subvol;
- }
+/* Keep sending readdirp on the subvol until it returns no more entries
+ * It is possible that not all entries will fit in a single readdirp in
+ * which case the rmdir will keep failing with ENOTEMPTY
+ */
- /* When DHT has only 1 child */
- if (conf->subvolume_cnt == 1) {
- STACK_WIND_COOKIE (frame, dht_rmdir_hashed_subvol_cbk,
- conf->subvolumes[0], conf->subvolumes[0],
- conf->subvolumes[0]->fops->rmdir,
- &local->loc, local->flags, NULL);
- return 0;
- }
+static int
+dht_rmdir_readdirp_do(call_frame_t *readdirp_frame, xlator_t *this)
+{
+ dht_local_t *local = NULL;
- local->current = &local->lock[0];
- ret = dht_protect_namespace (frame, &local->loc, local->hashed_subvol,
- &local->current->ns, dht_rmdir_lock_cbk);
- if (ret < 0) {
- local->op_ret = -1;
- local->op_errno = errno ? errno : EINVAL;
- goto err;
- }
+ local = readdirp_frame->local;
+ if (local->op_ret == -1) {
+ /* there is no point doing another readdirp on this
+ * subvol . */
+ dht_rmdir_readdirp_done(readdirp_frame, this);
return 0;
+ }
-err:
- dht_set_fixed_dir_stat (&local->preparent);
- dht_set_fixed_dir_stat (&local->postparent);
+ STACK_WIND_COOKIE(readdirp_frame, dht_rmdir_readdirp_cbk,
+ local->hashed_subvol, local->hashed_subvol,
+ local->hashed_subvol->fops->readdirp, local->fd, 4096, 0,
+ local->xattr);
- DHT_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
- &local->preparent, &local->postparent, NULL);
- return 0;
+ return 0;
}
-
-int
-dht_rmdir_linkfile_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+static int
+dht_rmdir_linkfile_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- xlator_t *src = NULL;
- call_frame_t *readdirp_frame = NULL;
- dht_local_t *readdirp_local = NULL;
- int this_call_cnt = 0;
- char gfid[GF_UUID_BUF_SIZE] ={0};
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ xlator_t *src = NULL;
+ call_frame_t *readdirp_frame = NULL;
+ dht_local_t *readdirp_local = NULL;
+ int this_call_cnt = 0;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ local = frame->local;
+ prev = cookie;
+ src = prev;
- local = frame->local;
- prev = cookie;
- src = prev;
+ readdirp_frame = local->main_frame;
+ readdirp_local = readdirp_frame->local;
+ gf_uuid_unparse(local->loc.gfid, gfid);
- readdirp_frame = local->main_frame;
- readdirp_local = readdirp_frame->local;
-
- gf_uuid_unparse(local->loc.gfid, gfid);
-
- if (op_ret == 0) {
- gf_msg_trace (this->name, 0,
- "Unlinked linkfile %s on %s, gfid = %s",
- local->loc.path, src->name, gfid);
- } else {
- if (op_errno != ENOENT) {
- readdirp_local->op_ret = -1;
- readdirp_local->op_errno = op_errno;
- }
- gf_msg_debug (this->name, op_errno,
- "Unlink of %s on %s failed. (gfid = %s)",
- local->loc.path, src->name, gfid);
+ if (op_ret == 0) {
+ gf_msg_trace(this->name, 0, "Unlinked linkfile %s on %s, gfid = %s",
+ local->loc.path, src->name, gfid);
+ } else {
+ if (op_errno != ENOENT) {
+ readdirp_local->op_ret = -1;
+ readdirp_local->op_errno = op_errno;
}
+ gf_msg_debug(this->name, op_errno,
+ "Unlink of %s on %s failed. (gfid = %s)", local->loc.path,
+ src->name, gfid);
+ }
- this_call_cnt = dht_frame_return (readdirp_frame);
+ this_call_cnt = dht_frame_return(readdirp_frame);
- if (is_last_call (this_call_cnt))
- dht_rmdir_readdirp_do (readdirp_frame, this);
+ if (is_last_call(this_call_cnt))
+ dht_rmdir_readdirp_do(readdirp_frame, this);
- DHT_STACK_DESTROY (frame);
- return 0;
+ DHT_STACK_DESTROY(frame);
+ return 0;
}
+static int
+dht_rmdir_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *stbuf, dict_t *xattr, struct iatt *parent)
+{
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ xlator_t *src = NULL;
+ call_frame_t *readdirp_frame = NULL;
+ dht_local_t *readdirp_local = NULL;
+ int this_call_cnt = 0;
+ dht_conf_t *conf = this->private;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+
+ local = frame->local;
+ prev = cookie;
+ src = prev;
+
+ gf_msg_debug(this->name, 0, "dht_rmdir_lookup_cbk %s", local->loc.path);
+
+ readdirp_frame = local->main_frame;
+ readdirp_local = readdirp_frame->local;
+
+ if (op_ret != 0) {
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, DHT_MSG_FILE_LOOKUP_FAILED,
+ "lookup failed for %s on %s", local->loc.path, src->name);
+ goto err;
+ }
+
+ if (!check_is_linkfile(inode, stbuf, xattr, conf->link_xattr_name)) {
+ readdirp_local->op_ret = -1;
+ readdirp_local->op_errno = ENOTEMPTY;
-int
-dht_rmdir_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode,
- struct iatt *stbuf, dict_t *xattr, struct iatt *parent)
-{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- xlator_t *src = NULL;
- call_frame_t *readdirp_frame = NULL;
- dht_local_t *readdirp_local = NULL;
- int this_call_cnt = 0;
- dht_conf_t *conf = this->private;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- local = frame->local;
- prev = cookie;
- src = prev;
-
- gf_msg_debug (this->name, 0, "dht_rmdir_lookup_cbk %s",
- local->loc.path);
-
- readdirp_frame = local->main_frame;
- readdirp_local = readdirp_frame->local;
-
- if (op_ret != 0) {
-
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_FILE_LOOKUP_FAILED,
- "lookup failed for %s on %s (type=0%o)",
- local->loc.path, src->name, stbuf->ia_type);
- goto err;
- }
-
- if (!check_is_linkfile (inode, stbuf, xattr, conf->link_xattr_name)) {
- readdirp_local->op_ret = -1;
- readdirp_local->op_errno = ENOTEMPTY;
-
- gf_uuid_unparse(local->loc.gfid, gfid);
+ gf_uuid_unparse(local->loc.gfid, gfid);
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_NOT_LINK_FILE_ERROR,
- "%s on %s is not a linkfile (type=0%o, gfid = %s)",
- local->loc.path, src->name, stbuf->ia_type, gfid);
- goto err;
- }
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_NOT_LINK_FILE_ERROR,
+ "%s on %s is not a linkfile (type=0%o, gfid = %s)",
+ local->loc.path, src->name, stbuf->ia_type, gfid);
+ goto err;
+ }
- STACK_WIND_COOKIE (frame, dht_rmdir_linkfile_unlink_cbk, src,
- src, src->fops->unlink, &local->loc, 0, NULL);
- return 0;
+ STACK_WIND_COOKIE(frame, dht_rmdir_linkfile_unlink_cbk, src, src,
+ src->fops->unlink, &local->loc, 0, NULL);
+ return 0;
err:
- this_call_cnt = dht_frame_return (readdirp_frame);
- if (is_last_call (this_call_cnt)) {
- dht_rmdir_readdirp_do (readdirp_frame, this);
- }
+ this_call_cnt = dht_frame_return(readdirp_frame);
+ if (is_last_call(this_call_cnt)) {
+ dht_rmdir_readdirp_do(readdirp_frame, this);
+ }
- DHT_STACK_DESTROY (frame);
- return 0;
+ DHT_STACK_DESTROY(frame);
+ return 0;
}
-
-int
-dht_rmdir_cached_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode,
- struct iatt *stbuf, dict_t *xattr,
- struct iatt *parent)
-{
- dht_local_t *local = NULL;
- xlator_t *src = NULL;
- call_frame_t *readdirp_frame = NULL;
- dht_local_t *readdirp_local = NULL;
- int this_call_cnt = 0;
- dht_conf_t *conf = this->private;
- dict_t *xattrs = NULL;
- int ret = 0;
-
- local = frame->local;
- src = local->hashed_subvol;
-
-
- /* main_frame here is the readdirp_frame */
-
- readdirp_frame = local->main_frame;
- readdirp_local = readdirp_frame->local;
-
- gf_msg_debug (this->name, 0, "returning for %s ",
- local->loc.path);
-
- if (op_ret == 0) {
- readdirp_local->op_ret = -1;
- readdirp_local->op_errno = ENOTEMPTY;
-
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_SUBVOL_ERROR,
- "%s found on cached subvol %s",
- local->loc.path, src->name);
- goto err;
- } else if (op_errno != ENOENT) {
- readdirp_local->op_ret = -1;
- readdirp_local->op_errno = op_errno;
-
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_SUBVOL_ERROR,
- "%s not found on cached subvol %s",
- local->loc.path, src->name);
- goto err;
- }
-
- xattrs = dict_new ();
- if (!xattrs) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_NO_MEMORY, "dict_new failed");
- goto err;
- }
-
- ret = dict_set_uint32 (xattrs, conf->link_xattr_name, 256);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value: key = %s",
- conf->link_xattr_name);
- if (xattrs)
- dict_unref (xattrs);
- goto err;
- }
- STACK_WIND_COOKIE (frame, dht_rmdir_lookup_cbk, src, src,
- src->fops->lookup, &local->loc, xattrs);
+static int
+dht_rmdir_cached_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *stbuf, dict_t *xattr,
+ struct iatt *parent)
+{
+ dht_local_t *local = NULL;
+ xlator_t *src = NULL;
+ call_frame_t *readdirp_frame = NULL;
+ dht_local_t *readdirp_local = NULL;
+ int this_call_cnt = 0;
+ dht_conf_t *conf = this->private;
+ dict_t *xattrs = NULL;
+ int ret = 0;
+
+ local = frame->local;
+ src = local->hashed_subvol;
+
+ /* main_frame here is the readdirp_frame */
+
+ readdirp_frame = local->main_frame;
+ readdirp_local = readdirp_frame->local;
+
+ gf_msg_debug(this->name, 0, "returning for %s ", local->loc.path);
+
+ if (op_ret == 0) {
+ readdirp_local->op_ret = -1;
+ readdirp_local->op_errno = ENOTEMPTY;
+
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_SUBVOL_ERROR,
+ "%s found on cached subvol %s", local->loc.path, src->name);
+ goto err;
+ } else if (op_errno != ENOENT) {
+ readdirp_local->op_ret = -1;
+ readdirp_local->op_errno = op_errno;
+
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, DHT_MSG_SUBVOL_ERROR,
+ "%s not found on cached subvol %s", local->loc.path, src->name);
+ goto err;
+ }
+
+ xattrs = dict_new();
+ if (!xattrs) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
+ "dict_new failed");
+ goto err;
+ }
+
+ ret = dict_set_uint32(xattrs, conf->link_xattr_name, 256);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary value: key = %s",
+ conf->link_xattr_name);
if (xattrs)
- dict_unref (xattrs);
-
- return 0;
+ dict_unref(xattrs);
+ goto err;
+ }
+ STACK_WIND_COOKIE(frame, dht_rmdir_lookup_cbk, src, src, src->fops->lookup,
+ &local->loc, xattrs);
+ if (xattrs)
+ dict_unref(xattrs);
+
+ return 0;
err:
- this_call_cnt = dht_frame_return (readdirp_frame);
+ this_call_cnt = dht_frame_return(readdirp_frame);
- /* Once all the lookups/unlinks etc have returned, proceed to wind
- * readdirp on the subvol again until no entries are returned.
- * This is required if there are more entries than can be returned
- * in a single readdirp call.
- */
+ /* Once all the lookups/unlinks etc have returned, proceed to wind
+ * readdirp on the subvol again until no entries are returned.
+ * This is required if there are more entries than can be returned
+ * in a single readdirp call.
+ */
- if (is_last_call (this_call_cnt))
- dht_rmdir_readdirp_do (readdirp_frame, this);
+ if (is_last_call(this_call_cnt))
+ dht_rmdir_readdirp_do(readdirp_frame, this);
- DHT_STACK_DESTROY (frame);
- return 0;
+ DHT_STACK_DESTROY(frame);
+ return 0;
}
+static int
+dht_rmdir_is_subvol_empty(call_frame_t *frame, xlator_t *this,
+ gf_dirent_t *entries, xlator_t *src)
+{
+ int ret = 0;
+ int build_ret = 0;
+ gf_dirent_t *trav = NULL;
+ call_frame_t *lookup_frame = NULL;
+ dht_local_t *lookup_local = NULL;
+ dht_local_t *local = NULL;
+ dict_t *xattrs = NULL;
+ dht_conf_t *conf = this->private;
+ xlator_t *subvol = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ int count = 0;
+ gf_boolean_t unwind = _gf_false;
+
+ local = frame->local;
+
+ list_for_each_entry(trav, &entries->list, list)
+ {
+ if (strcmp(trav->d_name, ".") == 0)
+ continue;
+ if (strcmp(trav->d_name, "..") == 0)
+ continue;
+ if (check_is_linkfile(NULL, (&trav->d_stat), trav->dict,
+ conf->link_xattr_name)) {
+ count++;
+ continue;
+ }
+
+ /* this entry is either a directory which is neither "." nor "..",
+ or a non directory which is not a linkfile. the directory is to
+ be treated as non-empty
+ */
+ return 0;
+ }
-int
-dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
- gf_dirent_t *entries, xlator_t *src)
-{
- int ret = 0;
- int build_ret = 0;
- gf_dirent_t *trav = NULL;
- call_frame_t *lookup_frame = NULL;
- dht_local_t *lookup_local = NULL;
- dht_local_t *local = NULL;
- dict_t *xattrs = NULL;
- dht_conf_t *conf = this->private;
- xlator_t *subvol = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
- int count = 0;
- gf_boolean_t unwind = _gf_false;
-
+ xattrs = dict_new();
+ if (!xattrs) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
+ "dict_new failed");
+ return -1;
+ }
- local = frame->local;
+ ret = dict_set_uint32(xattrs, conf->link_xattr_name, 256);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary value: key = %s",
+ conf->link_xattr_name);
- list_for_each_entry (trav, &entries->list, list) {
- if (strcmp (trav->d_name, ".") == 0)
- continue;
- if (strcmp (trav->d_name, "..") == 0)
- continue;
- if (check_is_linkfile (NULL, (&trav->d_stat), trav->dict,
- conf->link_xattr_name)) {
- count++;
- continue;
- }
+ if (xattrs)
+ dict_unref(xattrs);
+ return -1;
+ }
- /* this entry is either a directory which is neither "." nor "..",
- or a non directory which is not a linkfile. the directory is to
- be treated as non-empty
- */
- return 0;
- }
+ local->call_cnt = count;
+ ret = 0;
- xattrs = dict_new ();
- if (!xattrs) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_NO_MEMORY, "dict_new failed");
- return -1;
- }
+ list_for_each_entry(trav, &entries->list, list)
+ {
+ if (strcmp(trav->d_name, ".") == 0)
+ continue;
+ if (strcmp(trav->d_name, "..") == 0)
+ continue;
- ret = dict_set_uint32 (xattrs, conf->link_xattr_name, 256);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value: key = %s",
- conf->link_xattr_name);
+ lookup_frame = copy_frame(frame);
- if (xattrs)
- dict_unref (xattrs);
- return -1;
+ if (!lookup_frame) {
+ /* out of memory, let the rmdir fail
+ (as non-empty, unfortunately) */
+ goto err;
}
- local->call_cnt = count;
- ret = 0;
-
- list_for_each_entry (trav, &entries->list, list) {
- if (strcmp (trav->d_name, ".") == 0)
- continue;
- if (strcmp (trav->d_name, "..") == 0)
- continue;
-
- lookup_frame = copy_frame (frame);
-
- if (!lookup_frame) {
- /* out of memory, let the rmdir fail
- (as non-empty, unfortunately) */
- goto err;
- }
-
- lookup_local = dht_local_init (lookup_frame, NULL, NULL,
- GF_FOP_LOOKUP);
- if (!lookup_local) {
- goto err;
- }
-
- lookup_frame->local = lookup_local;
- lookup_local->main_frame = frame;
- lookup_local->hashed_subvol = src;
+ lookup_local = dht_local_init(lookup_frame, NULL, NULL, GF_FOP_LOOKUP);
+ if (!lookup_local) {
+ goto err;
+ }
- build_ret = dht_build_child_loc (this, &lookup_local->loc,
- &local->loc, trav->d_name);
- if (build_ret != 0)
- goto err;
+ lookup_frame->local = lookup_local;
+ lookup_local->main_frame = frame;
+ lookup_local->hashed_subvol = src;
- gf_uuid_copy (lookup_local->loc.gfid, trav->d_stat.ia_gfid);
+ build_ret = dht_build_child_loc(this, &lookup_local->loc, &local->loc,
+ trav->d_name);
+ if (build_ret != 0)
+ goto err;
- gf_uuid_unparse(lookup_local->loc.gfid, gfid);
+ gf_uuid_copy(lookup_local->loc.gfid, trav->d_stat.ia_gfid);
- gf_msg_trace (this->name, 0,
- "looking up %s on subvolume %s, gfid = %s",
- lookup_local->loc.path, src->name, gfid);
+ gf_uuid_unparse(lookup_local->loc.gfid, gfid);
- subvol = dht_linkfile_subvol (this, NULL, &trav->d_stat,
- trav->dict);
- if (!subvol) {
+ gf_msg_trace(this->name, 0, "looking up %s on subvolume %s, gfid = %s",
+ lookup_local->loc.path, src->name, gfid);
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_INVALID_LINKFILE,
- "Linkfile does not have link subvolume. "
- "path = %s, gfid = %s",
- lookup_local->loc.path, gfid);
+ subvol = dht_linkfile_subvol(this, NULL, &trav->d_stat, trav->dict);
+ if (!subvol || (subvol == src)) {
+ /* we need to delete the linkto file if it does not have a
+ * valid subvol or it points to itself.
+ */
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_INVALID_LINKFILE,
+ "Linkfile does not have link subvolume. "
+ "path = %s, gfid = %s",
+ lookup_local->loc.path, gfid);
- gf_msg_debug (this->name, 0,
- "looking up %s on subvol %s, gfid = %s",
- lookup_local->loc.path, src->name, gfid);
+ gf_msg_debug(this->name, 0, "looking up %s on subvol %s, gfid = %s",
+ lookup_local->loc.path, src->name, gfid);
- STACK_WIND_COOKIE (lookup_frame, dht_rmdir_lookup_cbk,
- src, src, src->fops->lookup,
- &lookup_local->loc, xattrs);
- } else {
- gf_msg_debug (this->name, 0,
- "Looking up linkfile target %s on "
- " subvol %s, gfid = %s",
- lookup_local->loc.path, subvol->name,
- gfid);
-
- STACK_WIND (lookup_frame, dht_rmdir_cached_lookup_cbk,
- subvol, subvol->fops->lookup,
- &lookup_local->loc, xattrs);
- }
- ret++;
+ STACK_WIND_COOKIE(lookup_frame, dht_rmdir_lookup_cbk, src, src,
+ src->fops->lookup, &lookup_local->loc, xattrs);
+ } else {
+ gf_msg_debug(this->name, 0,
+ "Looking up linkfile target %s on "
+ " subvol %s, gfid = %s",
+ lookup_local->loc.path, subvol->name, gfid);
- lookup_frame = NULL;
- lookup_local = NULL;
+ STACK_WIND(lookup_frame, dht_rmdir_cached_lookup_cbk, subvol,
+ subvol->fops->lookup, &lookup_local->loc, xattrs);
}
+ ret++;
- if (xattrs)
- dict_unref (xattrs);
+ lookup_frame = NULL;
+ lookup_local = NULL;
+ }
- return ret;
+ if (xattrs)
+ dict_unref(xattrs);
+
+ return ret;
err:
- if (xattrs)
- dict_unref (xattrs);
+ if (xattrs)
+ dict_unref(xattrs);
- if (lookup_frame)
- DHT_STACK_DESTROY (lookup_frame);
+ if (lookup_frame)
+ DHT_STACK_DESTROY(lookup_frame);
- /* Handle the case where the wound calls have unwound before the
- * loop processing is done
- */
+ /* Handle the case where the wound calls have unwound before the
+ * loop processing is done
+ */
- LOCK (&frame->lock);
- {
- local->op_ret = -1;
- local->op_errno = ENOTEMPTY;
+ LOCK(&frame->lock);
+ {
+ local->op_ret = -1;
+ local->op_errno = ENOTEMPTY;
- local->call_cnt -= (count - ret);
- if (!local->call_cnt)
- unwind = _gf_true;
- }
- UNLOCK (&frame->lock);
+ local->call_cnt -= (count - ret);
+ if (!local->call_cnt)
+ unwind = _gf_true;
+ }
+ UNLOCK(&frame->lock);
- if (!unwind) {
- return ret;
- }
- return 0;
+ if (!unwind) {
+ return ret;
+ }
+ return 0;
}
-
-
/*
* No more entries on this subvol. Proceed to the actual rmdir operation.
*/
-void
-dht_rmdir_readdirp_done (call_frame_t *readdirp_frame, xlator_t *this)
+static int
+dht_rmdir_readdirp_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, gf_dirent_t *entries,
+ dict_t *xdata)
{
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ xlator_t *src = NULL;
+ int ret = 0;
+ char *path = NULL;
- call_frame_t *main_frame = NULL;
- dht_local_t *main_local = NULL;
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
-
+ local = frame->local;
+ prev = cookie;
+ src = prev;
- local = readdirp_frame->local;
- main_frame = local->main_frame;
- main_local = main_frame->local;
-
- /* At least one readdirp failed.
- * This is a bit hit or miss - if readdirp failed on more than
- * one subvol, we don't know which error is returned.
+ if (op_ret > 2) {
+ /* dht_rmdir_is_subvol_empty() may free the frame,
+ * copy path for logging.
*/
- if (local->op_ret == -1) {
- main_local->op_ret = local->op_ret;
- main_local->op_errno = local->op_errno;
- }
-
- this_call_cnt = dht_frame_return (main_frame);
-
- if (is_last_call (this_call_cnt))
- dht_rmdir_do (main_frame, this);
-
-
- DHT_STACK_DESTROY (readdirp_frame);
-}
-
-
-
-int
-dht_rmdir_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, gf_dirent_t *entries,
- dict_t *xdata)
-{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- xlator_t *src = NULL;
- int ret = 0;
-
+ path = gf_strdup(local->loc.path);
- local = frame->local;
- prev = cookie;
- src = prev;
-
- if (op_ret > 2) {
- ret = dht_rmdir_is_subvol_empty (frame, this, entries, src);
-
- switch (ret) {
- case 0: /* non linkfiles exist */
- gf_msg_trace (this->name, 0,
- "readdir on %s for %s returned %d "
- "entries", prev->name,
- local->loc.path, op_ret);
- local->op_ret = -1;
- local->op_errno = ENOTEMPTY;
- goto done;
- default:
- /* @ret number of linkfiles are getting unlinked */
- gf_msg_trace (this->name, 0,
- "readdir on %s for %s found %d "
- "linkfiles", prev->name,
- local->loc.path, ret);
- break;
- }
-
- }
+ ret = dht_rmdir_is_subvol_empty(frame, this, entries, src);
-
- if (ret) {
- return 0;
+ switch (ret) {
+ case 0: /* non linkfiles exist */
+ gf_msg_trace(this->name, 0,
+ "readdir on %s for %s returned %d "
+ "entries",
+ prev->name, local->loc.path, op_ret);
+ local->op_ret = -1;
+ local->op_errno = ENOTEMPTY;
+ break;
+ default:
+ /* @ret number of linkfiles are getting unlinked */
+ gf_msg_trace(this->name, 0,
+ "readdir on %s for %s found %d "
+ "linkfiles",
+ prev->name, path, ret);
+ break;
}
+ }
-done:
- /* readdirp failed or no linkto files were found on this subvol */
+ /* readdirp failed or no linkto files were found on this subvol */
+ if (!ret)
+ dht_rmdir_readdirp_done(frame, this);
- dht_rmdir_readdirp_done (frame, this);
- return 0;
+ GF_FREE(path);
+ return 0;
}
-/* Keep sending readdirp on the subvol until it returns no more entries
- * It is possible that not all entries will fit in a single readdirp in which
- * case the rmdir will keep failing with ENOTEMPTY
- */
-
-int
-dht_rmdir_readdirp_do (call_frame_t *readdirp_frame, xlator_t *this)
-{
- dht_local_t *local = NULL;
-
- local = readdirp_frame->local;
+static int
+dht_rmdir_opendir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ int this_call_cnt = -1;
+ xlator_t *prev = NULL;
+ int ret = 0;
+ dht_conf_t *conf = this->private;
+ dict_t *dict = NULL;
+ int i = 0;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ dht_local_t *readdirp_local = NULL;
+ call_frame_t *readdirp_frame = NULL;
+ int cnt = 0;
+
+ local = frame->local;
+ prev = cookie;
+
+ this_call_cnt = dht_frame_return(frame);
+ if (op_ret == -1) {
+ gf_uuid_unparse(local->loc.gfid, gfid);
- if (local->op_ret == -1) {
- /* there is no point doing another readdirp on this
- * subvol . */
- dht_rmdir_readdirp_done (readdirp_frame, this);
- return 0;
+ gf_msg_debug(this->name, op_errno,
+ "opendir on %s for %s failed, "
+ "gfid = %s,",
+ prev->name, local->loc.path, gfid);
+ if ((op_errno != ENOENT) && (op_errno != ESTALE)) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
}
+ goto err;
+ }
- STACK_WIND_COOKIE (readdirp_frame, dht_rmdir_readdirp_cbk,
- local->hashed_subvol,
- local->hashed_subvol,
- local->hashed_subvol->fops->readdirp,
- local->fd, 4096, 0, local->xattr);
-
-
+ if (!is_last_call(this_call_cnt))
return 0;
-}
-
+ if (local->op_ret == -1)
+ goto err;
-int
-dht_rmdir_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
-{
- dht_local_t *local = NULL;
- int this_call_cnt = -1;
- xlator_t *prev = NULL;
- int ret = 0;
- dht_conf_t *conf = this->private;
- dict_t *dict = NULL;
- int i = 0;
- char gfid[GF_UUID_BUF_SIZE] = {0};
- dht_local_t *readdirp_local = NULL;
- call_frame_t *readdirp_frame = NULL;
- int cnt = 0;
-
- local = frame->local;
- prev = cookie;
+ fd_bind(fd);
+ dict = dict_new();
+ if (!dict) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto err;
+ }
- this_call_cnt = dht_frame_return (frame);
- if (op_ret == -1) {
- gf_uuid_unparse(local->loc.gfid, gfid);
-
- gf_msg_debug (this->name, op_errno,
- "opendir on %s for %s failed, "
- "gfid = %s,",
- prev->name, local->loc.path, gfid);
- if ((op_errno != ENOENT) && (op_errno != ESTALE)) {
- local->op_ret = -1;
- local->op_errno = op_errno;
- }
- goto err;
- }
+ ret = dict_set_uint32(dict, conf->link_xattr_name, 256);
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_DICT_SET_FAILED,
+ "%s: Failed to set dictionary value:key = %s", local->loc.path,
+ conf->link_xattr_name);
- if (!is_last_call (this_call_cnt))
- return 0;
+ cnt = local->call_cnt = conf->subvolume_cnt;
- if (local->op_ret == -1)
- goto err;
+ /* Create a separate frame per subvol as we might need
+ * to resend readdirp multiple times to get all the
+ * entries.
+ */
- fd_bind (fd);
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ readdirp_frame = copy_frame(frame);
- dict = dict_new ();
- if (!dict) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto err;
+ if (!readdirp_frame) {
+ cnt--;
+ /* Reduce the local->call_cnt as well */
+ (void)dht_frame_return(frame);
+ continue;
}
- ret = dict_set_uint32 (dict, conf->link_xattr_name, 256);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "%s: Failed to set dictionary value:key = %s",
- local->loc.path, conf->link_xattr_name);
-
- cnt = local->call_cnt = conf->subvolume_cnt;
-
-
- /* Create a separate frame per subvol as we might need
- * to resend readdirp multiple times to get all the
- * entries.
- */
-
- for (i = 0; i < conf->subvolume_cnt; i++) {
-
- readdirp_frame = copy_frame (frame);
-
- if (!readdirp_frame) {
- cnt--;
- /* Reduce the local->call_cnt as well */
- (void) dht_frame_return (frame);
- continue;
- }
-
- readdirp_local = dht_local_init (readdirp_frame, &local->loc,
- local->fd, 0);
-
- if (!readdirp_local) {
- DHT_STACK_DESTROY (readdirp_frame);
- cnt--;
- /* Reduce the local->call_cnt as well */
- dht_frame_return (frame);
- continue;
- }
- readdirp_local->main_frame = frame;
- readdirp_local->op_ret = 0;
- readdirp_local->xattr = dict_ref (dict);
- /* overload this field to save the subvol info */
- readdirp_local->hashed_subvol = conf->subvolumes[i];
+ readdirp_local = dht_local_init(readdirp_frame, &local->loc, local->fd,
+ 0);
- STACK_WIND_COOKIE (readdirp_frame, dht_rmdir_readdirp_cbk,
- conf->subvolumes[i], conf->subvolumes[i],
- conf->subvolumes[i]->fops->readdirp,
- readdirp_local->fd, 4096, 0,
- readdirp_local->xattr);
+ if (!readdirp_local) {
+ DHT_STACK_DESTROY(readdirp_frame);
+ cnt--;
+ /* Reduce the local->call_cnt as well */
+ dht_frame_return(frame);
+ continue;
}
+ readdirp_local->main_frame = frame;
+ readdirp_local->op_ret = 0;
+ readdirp_local->xattr = dict_ref(dict);
+ /* overload this field to save the subvol info */
+ readdirp_local->hashed_subvol = conf->subvolumes[i];
- if (dict)
- dict_unref (dict);
+ STACK_WIND_COOKIE(readdirp_frame, dht_rmdir_readdirp_cbk,
+ conf->subvolumes[i], conf->subvolumes[i],
+ conf->subvolumes[i]->fops->readdirp,
+ readdirp_local->fd, 4096, 0, readdirp_local->xattr);
+ }
- /* Could not wind readdirp to any subvol */
+ if (dict)
+ dict_unref(dict);
- if (!cnt)
- goto err;
+ /* Could not wind readdirp to any subvol */
- return 0;
+ if (!cnt)
+ goto err;
+
+ return 0;
err:
- if (is_last_call (this_call_cnt)) {
- dht_rmdir_do (frame, this);
- }
+ if (is_last_call(this_call_cnt)) {
+ dht_rmdir_do(frame, this);
+ }
- return 0;
+ return 0;
}
-
int
-dht_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
- dict_t *xdata)
-{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- int op_errno = -1;
- int i = -1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
- VALIDATE_OR_GOTO (this->private, err);
-
- conf = this->private;
-
- local = dht_local_init (frame, loc, NULL, GF_FOP_RMDIR);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local->call_cnt = conf->subvolume_cnt;
- local->op_ret = 0;
- local->fop_succeeded = 0;
-
- local->flags = flags;
-
- local->fd = fd_create (local->loc.inode, frame->root->pid);
- if (!local->fd) {
-
- op_errno = ENOMEM;
- goto err;
- }
-
- if (flags) {
- return dht_rmdir_do (frame, this);
- }
-
- for (i = 0; i < conf->subvolume_cnt; i++) {
- STACK_WIND_COOKIE (frame, dht_rmdir_opendir_cbk,
- conf->subvolumes[i], conf->subvolumes[i],
- conf->subvolumes[i]->fops->opendir,
- loc, local->fd, NULL);
- }
-
- return 0;
+dht_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int op_errno = -1;
+ int i = -1;
+ int ret = -1;
+ dict_t *xattr_req = NULL;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
+ VALIDATE_OR_GOTO(loc->path, err);
+ VALIDATE_OR_GOTO(this->private, err);
+
+ conf = this->private;
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_RMDIR);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->call_cnt = conf->subvolume_cnt;
+ local->op_ret = 0;
+ local->fop_succeeded = 0;
+
+ local->flags = flags;
+
+ local->fd = fd_create(local->loc.inode, frame->root->pid);
+ if (!local->fd) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ if (flags) {
+ return dht_rmdir_do(frame, this);
+ }
+ if (xdata) {
+ xattr_req = dict_ref(xdata);
+ } else {
+ xattr_req = dict_new();
+ }
+ if (xattr_req) {
+ ret = dict_set_uint32(xattr_req, conf->link_xattr_name, 256);
+ /* If parallel-readdir is enabled, this is required
+ * to handle stale linkto files in the directory
+ * being deleted. If this fails, log an error but
+ * do not prevent the operation.
+ */
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "%s: failed to set key %s",
+ loc->path, conf->link_xattr_name);
+ }
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "%s: failed to set key %s",
+ loc->path, conf->link_xattr_name);
+ }
+
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ STACK_WIND_COOKIE(frame, dht_rmdir_opendir_cbk, conf->subvolumes[i],
+ conf->subvolumes[i],
+ conf->subvolumes[i]->fops->opendir, loc, local->fd,
+ xattr_req);
+ }
+
+ if (xattr_req) {
+ dict_unref(xattr_req);
+ }
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (rmdir, frame, -1, op_errno,
- NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(rmdir, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-int
-dht_entrylk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
+static int
+dht_entrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- DHT_STACK_UNWIND (entrylk, frame, op_ret, op_errno, xdata);
- return 0;
+ DHT_STACK_UNWIND(entrylk, frame, op_ret, op_errno, xdata);
+ return 0;
}
/* TODO
@@ -10311,769 +10603,789 @@ dht_entrylk_cbk (call_frame_t *frame, void *cookie,
* as described in the bug 1311002.
*/
int
-dht_entrylk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, const char *basename,
- entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
-{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- local = dht_local_init (frame, loc, NULL, GF_FOP_ENTRYLK);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+dht_entrylk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata)
+{
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_uuid_unparse(loc->gfid, gfid);
+ local = dht_local_init(frame, loc, NULL, GF_FOP_ENTRYLK);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- gf_msg_debug (this->name, 0,
- "no cached subvolume for path=%s, "
- "gfid = %s", loc->path, gfid);
- op_errno = EINVAL;
- goto err;
- }
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_uuid_unparse(loc->gfid, gfid);
- local->call_cnt = 1;
+ gf_msg_debug(this->name, 0,
+ "no cached subvolume for path=%s, "
+ "gfid = %s",
+ loc->path, gfid);
+ op_errno = EINVAL;
+ goto err;
+ }
- STACK_WIND (frame, dht_entrylk_cbk,
- subvol, subvol->fops->entrylk,
- volume, loc, basename, cmd, type, xdata);
+ local->call_cnt = 1;
- return 0;
+ STACK_WIND(frame, dht_entrylk_cbk, subvol, subvol->fops->entrylk, volume,
+ loc, basename, cmd, type, xdata);
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (entrylk, frame, -1, op_errno, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(entrylk, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
-
-int
-dht_fentrylk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
+static int
+dht_fentrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- DHT_STACK_UNWIND (fentrylk, frame, op_ret, op_errno, NULL);
- return 0;
+ DHT_STACK_UNWIND(fentrylk, frame, op_ret, op_errno, NULL);
+ return 0;
}
-
int
-dht_fentrylk (call_frame_t *frame, xlator_t *this,
- const char *volume, fd_t *fd, const char *basename,
- entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
+dht_fentrylk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- char gfid[GF_UUID_BUF_SIZE] = {0};
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+ VALIDATE_OR_GOTO(fd->inode, err);
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO(fd->inode, err);
+ gf_uuid_unparse(fd->inode->gfid, gfid);
- gf_uuid_unparse(fd->inode->gfid, gfid);
+ subvol = dht_subvol_get_cached(this, fd->inode);
+ if (!subvol) {
+ gf_msg_debug(this->name, 0,
+ "No cached subvolume for fd=%p,"
+ " gfid = %s",
+ fd, gfid);
+ op_errno = EINVAL;
+ goto err;
+ }
- subvol = dht_subvol_get_cached (this, fd->inode);
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "No cached subvolume for fd=%p,"
- " gfid = %s", fd, gfid);
- op_errno = EINVAL;
- goto err;
- }
-
- STACK_WIND (frame, dht_fentrylk_cbk,
- subvol, subvol->fops->fentrylk,
- volume, fd, basename, cmd, type, xdata);
+ STACK_WIND(frame, dht_fentrylk_cbk, subvol, subvol->fops->fentrylk, volume,
+ fd, basename, cmd, type, xdata);
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fentrylk, frame, -1, op_errno, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(fentrylk, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
-
-int32_t
-dht_ipc_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+static int32_t
+dht_ipc_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
- GF_VALIDATE_OR_GOTO ("dht", frame, out);
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO("dht", frame, out);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, out);
- local = frame->local;
+ local = frame->local;
- LOCK (&frame->lock);
- {
- if (op_ret < 0 && op_errno != ENOTCONN) {
- local->op_errno = op_errno;
- goto unlock;
- }
- local->op_ret = 0;
+ LOCK(&frame->lock);
+ {
+ if (op_ret < 0 && op_errno != ENOTCONN) {
+ local->op_errno = op_errno;
+ goto unlock;
}
+ local->op_ret = 0;
+ }
unlock:
- UNLOCK (&frame->lock);
+ UNLOCK(&frame->lock);
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- DHT_STACK_UNWIND (ipc, frame, local->op_ret, local->op_errno,
- NULL);
- }
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt)) {
+ DHT_STACK_UNWIND(ipc, frame, local->op_ret, local->op_errno, NULL);
+ }
out:
- return 0;
+ return 0;
}
-
int32_t
-dht_ipc (call_frame_t *frame, xlator_t *this, int32_t op, dict_t *xdata)
+dht_ipc(call_frame_t *frame, xlator_t *this, int32_t op, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int op_errno = EINVAL;
- dht_conf_t *conf = NULL;
- int call_cnt = 0;
- int i = 0;
+ dht_local_t *local = NULL;
+ int op_errno = EINVAL;
+ dht_conf_t *conf = NULL;
+ int call_cnt = 0;
+ int i = 0;
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
- if (op != GF_IPC_TARGET_UPCALL)
- goto wind_default;
+ if (op != GF_IPC_TARGET_UPCALL)
+ goto wind_default;
- VALIDATE_OR_GOTO (this->private, err);
- conf = this->private;
+ VALIDATE_OR_GOTO(this->private, err);
+ conf = this->private;
- local = dht_local_init (frame, NULL, NULL, GF_FOP_IPC);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+ local = dht_local_init(frame, NULL, NULL, GF_FOP_IPC);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- call_cnt = conf->subvolume_cnt;
- local->call_cnt = call_cnt;
+ call_cnt = conf->subvolume_cnt;
+ local->call_cnt = call_cnt;
- if (xdata) {
- if (dict_set_int8 (xdata, conf->xattr_name, 0) < 0)
- goto err;
- }
+ if (xdata) {
+ if (dict_set_int8(xdata, conf->xattr_name, 0) < 0)
+ goto err;
+ }
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND (frame, dht_ipc_cbk, conf->subvolumes[i],
- conf->subvolumes[i]->fops->ipc, op, xdata);
- }
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND(frame, dht_ipc_cbk, conf->subvolumes[i],
+ conf->subvolumes[i]->fops->ipc, op, xdata);
+ }
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (ipc, frame, -1, op_errno, NULL);
+ DHT_STACK_UNWIND(ipc, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
wind_default:
- STACK_WIND (frame, default_ipc_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->ipc, op, xdata);
- return 0;
+ STACK_WIND(frame, default_ipc_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ipc, op, xdata);
+ return 0;
}
-
int
-dht_forget (xlator_t *this, inode_t *inode)
+dht_forget(xlator_t *this, inode_t *inode)
{
- uint64_t ctx_int = 0;
- dht_inode_ctx_t *ctx = NULL;
- dht_layout_t *layout = NULL;
+ uint64_t ctx_int = 0;
+ dht_inode_ctx_t *ctx = NULL;
+ dht_layout_t *layout = NULL;
- inode_ctx_del (inode, this, &ctx_int);
+ inode_ctx_del(inode, this, &ctx_int);
- if (!ctx_int)
- return 0;
+ if (!ctx_int)
+ return 0;
- ctx = (dht_inode_ctx_t *) (long) ctx_int;
+ ctx = (dht_inode_ctx_t *)(long)ctx_int;
- layout = ctx->layout;
- ctx->layout = NULL;
- dht_layout_unref (this, layout);
- GF_FREE (ctx);
+ layout = ctx->layout;
+ ctx->layout = NULL;
+ dht_layout_unref(this, layout);
+ GF_FREE(ctx);
- return 0;
+ return 0;
}
-
int
-dht_notify (xlator_t *this, int event, void *data, ...)
-{
- xlator_t *subvol = NULL;
- int cnt = -1;
- int i = -1;
- dht_conf_t *conf = NULL;
- int ret = -1;
- int propagate = 0;
-
- int had_heard_from_all = 0;
- int have_heard_from_all = 0;
- struct timeval time = {0,};
- gf_defrag_info_t *defrag = NULL;
- dict_t *dict = NULL;
- gf_defrag_type cmd = 0;
- dict_t *output = NULL;
- va_list ap;
- dht_methods_t *methods = NULL;
- struct gf_upcall *up_data = NULL;
- struct gf_upcall_cache_invalidation *up_ci = NULL;
-
- conf = this->private;
- GF_VALIDATE_OR_GOTO (this->name, conf, out);
-
- methods = &(conf->methods);
-
- /* had all subvolumes reported status once till now? */
- had_heard_from_all = 1;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (!conf->last_event[i]) {
- had_heard_from_all = 0;
- }
- }
-
- switch (event) {
+dht_notify(xlator_t *this, int event, void *data, ...)
+{
+ xlator_t *subvol = NULL;
+ int cnt = -1;
+ int i = -1;
+ dht_conf_t *conf = NULL;
+ int ret = -1;
+ int propagate = 0;
+
+ int had_heard_from_all = 0;
+ int have_heard_from_all = 0;
+ gf_defrag_info_t *defrag = NULL;
+ dict_t *dict = NULL;
+ gf_defrag_type cmd = 0;
+ dict_t *output = NULL;
+ va_list ap;
+ struct gf_upcall *up_data = NULL;
+ struct gf_upcall_cache_invalidation *up_ci = NULL;
+
+ conf = this->private;
+ GF_VALIDATE_OR_GOTO(this->name, conf, out);
+
+ /* had all subvolumes reported status once till now? */
+ had_heard_from_all = 1;
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (!conf->last_event[i]) {
+ had_heard_from_all = 0;
+ }
+ }
+
+ switch (event) {
case GF_EVENT_CHILD_UP:
- subvol = data;
+ subvol = data;
- conf->gen++;
+ conf->gen++;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (subvol == conf->subvolumes[i]) {
- cnt = i;
- break;
- }
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (subvol == conf->subvolumes[i]) {
+ cnt = i;
+ break;
}
+ }
- if (cnt == -1) {
- gf_msg_debug (this->name, 0,
- "got GF_EVENT_CHILD_UP bad "
- "subvolume %s",
- subvol->name);
- break;
- }
+ if (cnt == -1) {
+ gf_msg_debug(this->name, 0,
+ "got GF_EVENT_CHILD_UP bad "
+ "subvolume %s",
+ subvol->name);
+ break;
+ }
- gettimeofday (&time, NULL);
- LOCK (&conf->subvolume_lock);
- {
- conf->subvolume_status[cnt] = 1;
- conf->last_event[cnt] = event;
- conf->subvol_up_time[cnt] = time.tv_sec;
- }
- UNLOCK (&conf->subvolume_lock);
+ LOCK(&conf->subvolume_lock);
+ {
+ conf->subvolume_status[cnt] = 1;
+ conf->last_event[cnt] = event;
+ conf->subvol_up_time[cnt] = gf_time();
+ }
+ UNLOCK(&conf->subvolume_lock);
- /* one of the node came back up, do a stat update */
- dht_get_du_info_for_subvol (this, cnt);
+ /* one of the node came back up, do a stat update */
+ dht_get_du_info_for_subvol(this, cnt);
- break;
+ break;
case GF_EVENT_SOME_DESCENDENT_UP:
- subvol = data;
- conf->gen++;
- propagate = 1;
+ subvol = data;
+ conf->gen++;
+ propagate = 1;
- break;
+ break;
case GF_EVENT_SOME_DESCENDENT_DOWN:
- subvol = data;
- propagate = 1;
+ subvol = data;
+ propagate = 1;
- break;
+ break;
case GF_EVENT_CHILD_DOWN:
- subvol = data;
-
- if (conf->assert_no_child_down) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_CHILD_DOWN,
- "Received CHILD_DOWN. Exiting");
- if (conf->defrag) {
- gf_defrag_stop (conf,
- GF_DEFRAG_STATUS_FAILED, NULL);
- } else {
- kill (getpid(), SIGTERM);
- }
- }
+ subvol = data;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (subvol == conf->subvolumes[i]) {
- cnt = i;
- break;
- }
- }
-
- if (cnt == -1) {
- gf_msg_debug (this->name, 0,
- "got GF_EVENT_CHILD_DOWN bad "
- "subvolume %s", subvol->name);
- break;
+ if (conf->assert_no_child_down) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_CHILD_DOWN,
+ "Received CHILD_DOWN. Exiting");
+ if (conf->defrag) {
+ gf_defrag_stop(conf, GF_DEFRAG_STATUS_FAILED, NULL);
+ } else {
+ kill(getpid(), SIGTERM);
}
+ }
- LOCK (&conf->subvolume_lock);
- {
- conf->subvolume_status[cnt] = 0;
- conf->last_event[cnt] = event;
- conf->subvol_up_time[cnt] = 0;
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (subvol == conf->subvolumes[i]) {
+ cnt = i;
+ break;
}
- UNLOCK (&conf->subvolume_lock);
+ }
- for (i = 0; i < conf->subvolume_cnt; i++)
- if (conf->last_event[i] != event)
- event = GF_EVENT_SOME_DESCENDENT_DOWN;
+ if (cnt == -1) {
+ gf_msg_debug(this->name, 0,
+ "got GF_EVENT_CHILD_DOWN bad "
+ "subvolume %s",
+ subvol->name);
break;
+ }
- case GF_EVENT_CHILD_CONNECTING:
- subvol = data;
+ LOCK(&conf->subvolume_lock);
+ {
+ conf->subvolume_status[cnt] = 0;
+ conf->last_event[cnt] = event;
+ conf->subvol_up_time[cnt] = 0;
+ }
+ UNLOCK(&conf->subvolume_lock);
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (subvol == conf->subvolumes[i]) {
- cnt = i;
- break;
- }
- }
+ for (i = 0; i < conf->subvolume_cnt; i++)
+ if (conf->last_event[i] != event)
+ event = GF_EVENT_SOME_DESCENDENT_DOWN;
+ break;
- if (cnt == -1) {
- gf_msg_debug (this->name, 0,
- "got GF_EVENT_CHILD_CONNECTING"
- " bad subvolume %s",
- subvol->name);
- break;
- }
+ case GF_EVENT_CHILD_CONNECTING:
+ subvol = data;
- LOCK (&conf->subvolume_lock);
- {
- conf->last_event[cnt] = event;
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (subvol == conf->subvolumes[i]) {
+ cnt = i;
+ break;
}
- UNLOCK (&conf->subvolume_lock);
+ }
+ if (cnt == -1) {
+ gf_msg_debug(this->name, 0,
+ "got GF_EVENT_CHILD_CONNECTING"
+ " bad subvolume %s",
+ subvol->name);
break;
- case GF_EVENT_VOLUME_DEFRAG:
- {
- if (!conf->defrag) {
- return ret;
- }
- defrag = conf->defrag;
+ }
- dict = data;
- va_start (ap, data);
- output = va_arg (ap, dict_t*);
+ LOCK(&conf->subvolume_lock);
+ {
+ conf->last_event[cnt] = event;
+ }
+ UNLOCK(&conf->subvolume_lock);
- ret = dict_get_int32 (dict, "rebalance-command",
- (int32_t*)&cmd);
- if (ret) {
- va_end (ap);
- return ret;
- }
- LOCK (&defrag->lock);
- {
- if (defrag->is_exiting)
- goto unlock;
- if ((cmd == GF_DEFRAG_CMD_STATUS) ||
- (cmd == GF_DEFRAG_CMD_STATUS_TIER) ||
- (cmd == GF_DEFRAG_CMD_DETACH_STATUS))
- gf_defrag_status_get (conf, output);
- else if (cmd == GF_DEFRAG_CMD_START_DETACH_TIER)
- gf_defrag_start_detach_tier(defrag);
- else if (cmd == GF_DEFRAG_CMD_DETACH_START)
- defrag->cmd = GF_DEFRAG_CMD_DETACH_START;
- else if (cmd == GF_DEFRAG_CMD_STOP ||
- cmd == GF_DEFRAG_CMD_STOP_DETACH_TIER ||
- cmd == GF_DEFRAG_CMD_DETACH_STOP)
- gf_defrag_stop (conf,
- GF_DEFRAG_STATUS_STOPPED, output);
- else if (cmd == GF_DEFRAG_CMD_PAUSE_TIER)
- ret = gf_defrag_pause_tier (this, defrag);
- else if (cmd == GF_DEFRAG_CMD_RESUME_TIER)
- ret = gf_defrag_resume_tier (this, defrag);
- }
-unlock:
- UNLOCK (&defrag->lock);
- va_end (ap);
+ break;
+ case GF_EVENT_VOLUME_DEFRAG: {
+ if (!conf->defrag) {
return ret;
- break;
+ }
+ defrag = conf->defrag;
+
+ dict = data;
+ va_start(ap, data);
+ output = va_arg(ap, dict_t *);
+
+ ret = dict_get_int32(dict, "rebalance-command", (int32_t *)&cmd);
+ if (ret) {
+ va_end(ap);
+ return ret;
+ }
+ LOCK(&defrag->lock);
+ {
+ if (defrag->is_exiting)
+ goto unlock;
+ if ((cmd == GF_DEFRAG_CMD_STATUS) ||
+ (cmd == GF_DEFRAG_CMD_DETACH_STATUS))
+ gf_defrag_status_get(conf, output);
+ else if (cmd == GF_DEFRAG_CMD_DETACH_START)
+ defrag->cmd = GF_DEFRAG_CMD_DETACH_START;
+ else if (cmd == GF_DEFRAG_CMD_STOP ||
+ cmd == GF_DEFRAG_CMD_DETACH_STOP)
+ gf_defrag_stop(conf, GF_DEFRAG_STATUS_STOPPED, output);
+ }
+ unlock:
+ UNLOCK(&defrag->lock);
+ va_end(ap);
+ return ret;
+ break;
}
case GF_EVENT_UPCALL:
- up_data = (struct gf_upcall *)data;
- if (up_data->event_type != GF_UPCALL_CACHE_INVALIDATION)
- break;
- up_ci = (struct gf_upcall_cache_invalidation *)up_data->data;
-
- /* Since md-cache will be aggressively filtering lookups,
- * the stale layout issue will be more pronounced. Hence
- * when a layout xattr is changed by the rebalance process
- * notify all the md-cache clients to invalidate the existing
- * stat cache and send the lookup next time*/
- if (up_ci->dict && dict_get (up_ci->dict, conf->xattr_name))
- up_ci->flags |= UP_EXPLICIT_LOOKUP;
-
- /* TODO: Instead of invalidating iatt, update the new
- * hashed/cached subvolume in dht inode_ctx */
- if (IS_DHT_LINKFILE_MODE (&up_ci->stat))
- up_ci->flags |= UP_EXPLICIT_LOOKUP;
-
- propagate = 1;
+ up_data = (struct gf_upcall *)data;
+ if (up_data->event_type != GF_UPCALL_CACHE_INVALIDATION)
break;
+ up_ci = (struct gf_upcall_cache_invalidation *)up_data->data;
+
+ /* Since md-cache will be aggressively filtering lookups,
+ * the stale layout issue will be more pronounced. Hence
+ * when a layout xattr is changed by the rebalance process
+ * notify all the md-cache clients to invalidate the existing
+ * stat cache and send the lookup next time*/
+ if (up_ci->dict && dict_get(up_ci->dict, conf->xattr_name))
+ up_ci->flags |= UP_EXPLICIT_LOOKUP;
+
+ /* TODO: Instead of invalidating iatt, update the new
+ * hashed/cached subvolume in dht inode_ctx */
+ if (IS_DHT_LINKFILE_MODE(&up_ci->stat))
+ up_ci->flags |= UP_EXPLICIT_LOOKUP;
+
+ propagate = 1;
+ break;
default:
- propagate = 1;
- break;
- }
-
+ propagate = 1;
+ break;
+ }
+
+ /* have all subvolumes reported status once by now? */
+ have_heard_from_all = 1;
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (!conf->last_event[i])
+ have_heard_from_all = 0;
+ }
+
+ /* if all subvols have reported status, no need to hide anything
+ or wait for anything else. Just propagate blindly */
+ if (have_heard_from_all) {
+ propagate = 1;
+ }
+
+ if (!had_heard_from_all && have_heard_from_all) {
+ static int run_defrag = 0;
+ /* This is the first event which completes aggregation
+ of events from all subvolumes. If at least one subvol
+ had come up, propagate CHILD_UP, but only this time
+ */
+ event = GF_EVENT_CHILD_DOWN;
- /* have all subvolumes reported status once by now? */
- have_heard_from_all = 1;
for (i = 0; i < conf->subvolume_cnt; i++) {
- if (!conf->last_event[i])
- have_heard_from_all = 0;
- }
-
- /* if all subvols have reported status, no need to hide anything
- or wait for anything else. Just propagate blindly */
- if (have_heard_from_all) {
- propagate = 1;
+ if (conf->last_event[i] == GF_EVENT_CHILD_UP) {
+ event = GF_EVENT_CHILD_UP;
+ break;
+ }
+ if (conf->last_event[i] == GF_EVENT_CHILD_CONNECTING) {
+ event = GF_EVENT_CHILD_CONNECTING;
+ /* continue to check other events for CHILD_UP */
+ }
}
-
- if (!had_heard_from_all && have_heard_from_all) {
- /* This is the first event which completes aggregation
- of events from all subvolumes. If at least one subvol
- had come up, propagate CHILD_UP, but only this time
- */
- event = GF_EVENT_CHILD_DOWN;
-
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->last_event[i] == GF_EVENT_CHILD_UP) {
- event = GF_EVENT_CHILD_UP;
- break;
- }
-
- if (conf->last_event[i] == GF_EVENT_CHILD_CONNECTING) {
- event = GF_EVENT_CHILD_CONNECTING;
- /* continue to check other events for CHILD_UP */
- }
- }
-
- /* Rebalance is started with assert_no_child_down. So we do
- * not need to handle CHILD_DOWN event here.
- *
- * If there is a graph switch, we should not restart the
- * rebalance daemon. Use 'run_defrag' to indicate if the
- * thread has already started.
- */
- if (conf->defrag && !run_defrag) {
- if (methods->migration_needed(this)) {
- run_defrag = 1;
- ret = gf_thread_create(&conf->defrag->th,
- NULL,
- gf_defrag_start, this,
- "dhtdg");
- if (ret) {
- GF_FREE (conf->defrag);
- conf->defrag = NULL;
- kill (getpid(), SIGTERM);
- }
- }
- }
+ /* Rebalance is started with assert_no_child_down. So we do
+ * not need to handle CHILD_DOWN event here.
+ *
+ * If there is a graph switch, we should not restart the
+ * rebalance daemon. Use 'run_defrag' to indicate if the
+ * thread has already started.
+ */
+ if (conf->defrag && !run_defrag) {
+ run_defrag = 1;
+ ret = gf_thread_create(&conf->defrag->th, NULL, gf_defrag_start,
+ this, "dhtdg");
+ if (ret) {
+ GF_FREE(conf->defrag);
+ conf->defrag = NULL;
+ kill(getpid(), SIGTERM);
+ }
}
+ }
- ret = 0;
- if (propagate)
- ret = default_notify (this, event, data);
+ ret = 0;
+ if (propagate)
+ ret = default_notify(this, event, data);
out:
- return ret;
+ return ret;
}
int
-dht_inode_ctx_layout_get (inode_t *inode, xlator_t *this, dht_layout_t **layout)
+dht_inode_ctx_layout_get(inode_t *inode, xlator_t *this, dht_layout_t **layout)
{
- dht_inode_ctx_t *ctx = NULL;
- int ret = -1;
+ dht_inode_ctx_t *ctx = NULL;
+ int ret = -1;
- ret = dht_inode_ctx_get (inode, this, &ctx);
+ ret = dht_inode_ctx_get(inode, this, &ctx);
- if (!ret && ctx) {
- if (ctx->layout) {
- if (layout)
- *layout = ctx->layout;
- ret = 0;
- } else {
- ret = -1;
- }
+ if (!ret && ctx) {
+ if (ctx->layout) {
+ if (layout)
+ *layout = ctx->layout;
+ ret = 0;
+ } else {
+ ret = -1;
}
+ }
- return ret;
+ return ret;
}
void
-dht_log_new_layout_for_dir_selfheal (xlator_t *this, loc_t *loc,
- dht_layout_t *layout)
-{
-
- char string[2048] = {0};
- char *output_string = NULL;
- int len = 0;
- int off = 0;
- int i = 0;
- gf_loglevel_t log_level = gf_log_get_loglevel();
- int ret = 0;
- int max_string_len = 0;
+dht_log_new_layout_for_dir_selfheal(xlator_t *this, loc_t *loc,
+ dht_layout_t *layout)
+{
+ char string[2048] = {0};
+ char *output_string = NULL;
+ int len = 0;
+ int off = 0;
+ int i = 0;
+ gf_loglevel_t log_level = gf_log_get_loglevel();
+ int ret = 0;
+
+ if (log_level < GF_LOG_INFO)
+ return;
- if (log_level < GF_LOG_INFO)
- return;
+ if (!layout)
+ return;
- if (!layout)
- return;
+ if (!layout->cnt)
+ return;
- if (!layout->cnt)
- return;
+ if (!loc)
+ return;
- if (!loc)
- return;
+ if (!loc->path)
+ return;
- if (!loc->path)
- return;
+ ret = snprintf(string, sizeof(string), "Setting layout of %s with ",
+ loc->path);
- max_string_len = sizeof (string);
+ if (ret < 0)
+ return;
- ret = snprintf (string, max_string_len, "Setting layout of %s with ",
- loc->path);
+ len += ret;
+
+ /* Calculation of total length of the string required to calloc
+ * output_string. Log includes subvolume-name, start-range, end-range
+ * and err value.
+ *
+ * This log will help to debug cases where:
+ * a) Different processes set different layout of a directory.
+ * b) Error captured in lookup, which will be filled in layout->err
+ * (like ENOENT, ESTALE etc)
+ */
+
+ for (i = 0; i < layout->cnt; i++) {
+ ret = snprintf(string, sizeof(string),
+ "[Subvol_name: %s, Err: %d , Start: "
+ "0x%x, Stop: 0x%x, Hash: 0x%x], ",
+ layout->list[i].xlator->name, layout->list[i].err,
+ layout->list[i].start, layout->list[i].stop,
+ layout->list[i].commit_hash);
if (ret < 0)
- return;
+ return;
len += ret;
+ }
- /* Calculation of total length of the string required to calloc
- * output_string. Log includes subvolume-name, start-range, end-range and
- * err value.
- *
- * This log will help to debug cases where:
- * a) Different processes set different layout of a directory.
- * b) Error captured in lookup, which will be filled in layout->err
- * (like ENOENT, ESTALE etc)
- */
-
- for (i = 0; i < layout->cnt; i++) {
+ len++;
- ret = snprintf (string, max_string_len,
- "[Subvol_name: %s, Err: %d , Start: "
- "%"PRIu32 " , Stop: %"PRIu32 " , Hash: %"
- PRIu32 " ], ",
- layout->list[i].xlator->name,
- layout->list[i].err, layout->list[i].start,
- layout->list[i].stop,
- layout->list[i].commit_hash);
+ output_string = GF_MALLOC(len + 1, gf_common_mt_char);
- if (ret < 0)
- return;
-
- len += ret;
-
- }
+ if (!output_string)
+ return;
- len++;
+ ret = snprintf(output_string, len + 1, "Setting layout of %s with ",
+ loc->path);
- output_string = GF_CALLOC (len, sizeof (char), gf_common_mt_char);
+ if (ret < 0)
+ goto err;
- if (!output_string)
- return;
+ off += ret;
- ret = snprintf (output_string, len, "Setting layout of %s with ",
- loc->path);
+ for (i = 0; i < layout->cnt; i++) {
+ ret = snprintf(output_string + off, len - off,
+ "[Subvol_name: %s, Err: %d , Start: "
+ "0x%x, Stop: 0x%x, Hash: 0x%x], ",
+ layout->list[i].xlator->name, layout->list[i].err,
+ layout->list[i].start, layout->list[i].stop,
+ layout->list[i].commit_hash);
if (ret < 0)
- goto err;
+ goto err;
off += ret;
+ }
-
- for (i = 0; i < layout->cnt; i++) {
-
- ret = snprintf (output_string + off, len - off,
- "[Subvol_name: %s, Err: %d , Start: "
- "%"PRIu32 " , Stop: %"PRIu32 " , Hash: %"
- PRIu32 " ], ",
- layout->list[i].xlator->name,
- layout->list[i].err, layout->list[i].start,
- layout->list[i].stop,
- layout->list[i].commit_hash);
-
- if (ret < 0)
- goto err;
-
- off += ret;
-
- }
-
- gf_msg (this->name, GF_LOG_DEBUG, 0, DHT_MSG_LOG_FIXED_LAYOUT,
- "%s", output_string);
+ gf_msg(this->name, GF_LOG_DEBUG, 0, DHT_MSG_LOG_FIXED_LAYOUT, "%s",
+ output_string);
err:
- GF_FREE (output_string);
+ GF_FREE(output_string);
}
-int32_t dht_migration_get_dst_subvol(xlator_t *this, dht_local_t *local)
+int32_t
+dht_migration_get_dst_subvol(xlator_t *this, dht_local_t *local)
{
- int ret = -1;
+ int ret = -1;
- if (!local)
- goto out;
+ if (!local)
+ goto out;
- local->rebalance.target_node =
- dht_subvol_get_hashed (this, &local->loc);
+ local->rebalance.target_node = dht_subvol_get_hashed(this, &local->loc);
- if (local->rebalance.target_node)
- ret = 0;
+ if (local->rebalance.target_node)
+ ret = 0;
out:
- return ret;
+ return ret;
}
-int32_t dht_migration_needed(xlator_t *this)
+/*
+This function should not be called more then once during a FOP
+handling path. It is valid only for for ops on files
+*/
+int32_t
+dht_set_local_rebalance(xlator_t *this, dht_local_t *local, struct iatt *stbuf,
+ struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata)
{
- gf_defrag_info_t *defrag = NULL;
- dht_conf_t *conf = NULL;
- int ret = 0;
+ if (!local)
+ return -1;
- conf = this->private;
+ if (local->rebalance.set) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_REBAL_STRUCT_SET,
+ "local->rebalance already set");
+ }
- GF_VALIDATE_OR_GOTO ("dht", conf, out);
- GF_VALIDATE_OR_GOTO ("dht", conf->defrag, out);
+ if (stbuf)
+ memcpy(&local->rebalance.stbuf, stbuf, sizeof(struct iatt));
- defrag = conf->defrag;
+ if (prebuf)
+ memcpy(&local->rebalance.prebuf, prebuf, sizeof(struct iatt));
- if ((defrag->cmd != GF_DEFRAG_CMD_START_TIER) &&
- (defrag->cmd != GF_DEFRAG_CMD_START_DETACH_TIER))
- ret = 1;
+ if (postbuf)
+ memcpy(&local->rebalance.postbuf, postbuf, sizeof(struct iatt));
-out:
- return ret;
-}
+ if (xdata)
+ local->rebalance.xdata = dict_ref(xdata);
+ local->rebalance.set = 1;
+ return 0;
+}
-/*
-This function should not be called more then once during a FOP
-handling path. It is valid only for for ops on files
-*/
-int32_t dht_set_local_rebalance (xlator_t *this, dht_local_t *local,
- struct iatt *stbuf,
- struct iatt *prebuf, struct iatt *postbuf,
- dict_t *xdata)
+int32_t
+dht_release(xlator_t *this, fd_t *fd)
{
+ return dht_fd_ctx_destroy(this, fd);
+}
- if (!local)
- return -1;
-
- if (local->rebalance.set) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_REBAL_STRUCT_SET,
- "local->rebalance already set");
- }
+static int
+dht_pt_mkdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, inode_t *inode, struct iatt *stbuf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ local = frame->local;
- if (stbuf)
- memcpy (&local->rebalance.stbuf, stbuf, sizeof (struct iatt));
+ if (!op_ret) {
+ dht_layout_set(this, inode, local->layout);
+ }
- if (prebuf)
- memcpy (&local->rebalance.prebuf, prebuf, sizeof (struct iatt));
+ DHT_STACK_UNWIND(mkdir, frame, op_ret, op_errno, inode, stbuf, preparent,
+ postparent, NULL);
- if (postbuf)
- memcpy (&local->rebalance.postbuf, postbuf,
- sizeof (struct iatt));
+ return 0;
+}
- if (xdata)
- local->rebalance.xdata = dict_ref (xdata);
+int32_t
+dht_pt_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
+{
+ dht_layout_t *layout = NULL;
+ dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ bool free_xdata = false;
+ int ret = 0;
+ int op_errno = 0;
+ int32_t *disk_layout_p = NULL;
+
+ conf = this->private;
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_MKDIR);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ layout = dht_layout_new(this, conf->subvolume_cnt);
+ if (!layout)
+ goto wind;
+
+ local->layout = layout;
+
+ if (!xdata) {
+ xdata = dict_new();
+ if (!xdata)
+ goto wind;
+ free_xdata = true;
+ }
+
+ /*Set the xlator or the following will crash*/
+ layout->list[0].xlator = conf->subvolumes[0];
+
+ dht_selfheal_layout_new_directory(frame, loc, layout);
+
+ dht_disk_layout_extract(this, layout, 0, &disk_layout_p);
+
+ ret = dict_set_bin(xdata, conf->xattr_name, disk_layout_p, 4 * 4);
+ if (ret) {
+ gf_msg("dht", GF_LOG_DEBUG, EINVAL, DHT_MSG_DICT_SET_FAILED,
+ "dht layout dict set failed");
+ }
+wind:
+ STACK_WIND(frame, dht_pt_mkdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
+ if (free_xdata)
+ dict_unref(xdata);
+ return 0;
- local->rebalance.set = 1;
+err:
+ op_errno = local ? local->op_errno : op_errno;
+ DHT_STACK_UNWIND(mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-gf_boolean_t
-dht_is_tier_xlator (xlator_t *this)
+static int
+dht_pt_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
{
+ dht_conf_t *conf = NULL;
- if (strcmp (this->type, "cluster/tier") == 0)
- return _gf_true;
- return _gf_false;
-}
+ conf = this->private;
+ dict_del(xattr, conf->xattr_name);
+ dict_del(xattr, conf->mds_xattr_key);
+ dict_del(xattr, conf->commithash_xattr_name);
-int32_t
-dht_release (xlator_t *this, fd_t *fd)
-{
- return dht_fd_ctx_destroy (this, fd);
+ if (frame->root->pid >= 0) {
+ GF_REMOVE_INTERNAL_XATTR("trusted.glusterfs.quota*", xattr);
+ GF_REMOVE_INTERNAL_XATTR("trusted.pgfid*", xattr);
+ }
+
+ DHT_STACK_UNWIND(getxattr, frame, op_ret, op_errno, xattr, xdata);
+ return 0;
}
int
-dht_remove_stale_linkto (void *data)
+dht_pt_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *key, dict_t *xdata)
{
- call_frame_t *frame = NULL;
- dht_local_t *local = NULL;
- xlator_t *this = NULL;
- dict_t *xdata_in = NULL;
- int ret = 0;
-
- GF_VALIDATE_OR_GOTO ("dht", data, out);
+ STACK_WIND(frame, dht_pt_getxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->getxattr, loc, key, xdata);
+ return 0;
+}
- frame = data;
- local = frame->local;
- this = frame->this;
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", local, out);
- GF_VALIDATE_OR_GOTO ("dht", local->link_subvol, out);
+static int
+dht_pt_fgetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
+{
+ dht_conf_t *conf = NULL;
- xdata_in = dict_new ();
- if (!xdata_in)
- goto out;
+ conf = this->private;
+ dict_del(xattr, conf->xattr_name);
- ret = dht_fill_dict_to_avoid_unlink_of_migrating_file (xdata_in);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, -ret, 0,
- "Failed to set keys for stale linkto"
- "deletion on path %s", local->loc.path);
- goto out;
- }
+ if (frame->root->pid >= 0) {
+ GF_REMOVE_INTERNAL_XATTR("trusted.glusterfs.quota*", xattr);
+ GF_REMOVE_INTERNAL_XATTR("trusted.pgfid*", xattr);
+ }
- ret = syncop_unlink (local->link_subvol, &local->loc, xdata_in, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, -ret, 0,
- "Removal of linkto failed"
- " on path %s at subvol %s",
- local->loc.path, local->link_subvol->name);
+ DHT_STACK_UNWIND(fgetxattr, frame, op_ret, op_errno, xattr, xdata);
+ return 0;
+}
- }
-out:
- if (xdata_in)
- dict_unref (xdata_in);
- return ret;
+int
+dht_pt_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *key,
+ dict_t *xdata)
+{
+ STACK_WIND(frame, dht_pt_fgetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fgetxattr, fd, key, xdata);
+ return 0;
}
+/* The job of this function is to check if all the xlators have updated
+ * error in the layout. */
int
-dht_remove_stale_linkto_cbk (int ret, call_frame_t *sync_frame, void *data)
+dht_dir_layout_error_check(xlator_t *this, inode_t *inode)
{
- DHT_STACK_DESTROY (sync_frame);
- return 0;
+ dht_layout_t *layout = NULL;
+ int i = 0;
+
+ layout = dht_layout_get(this, inode);
+ for (i = 0; i < layout->cnt; i++) {
+ if (layout->list[i].err == 0) {
+ return 0;
+ }
+ }
+
+ /* Returning the first xlator error as all xlators have errors */
+ return layout->list[0].err;
}
diff --git a/xlators/cluster/dht/src/dht-common.h b/xlators/cluster/dht/src/dht-common.h
index 86c810b16ff..fe0dc3db34a 100644
--- a/xlators/cluster/dht/src/dht-common.h
+++ b/xlators/cluster/dht/src/dht-common.h
@@ -9,150 +9,148 @@
*/
#include <regex.h>
-#include <signal.h>
-#include <fnmatch.h>
#include "dht-mem-types.h"
#include "dht-messages.h"
-#include "call-stub.h"
+#include <glusterfs/call-stub.h>
#include "libxlator.h"
-#include "syncop.h"
-#include "refcount.h"
-#include "timer.h"
+#include <glusterfs/syncop.h>
+#include <glusterfs/refcount.h>
+#include <glusterfs/timer.h>
#include "protocol-common.h"
-#include "glusterfs-acl.h"
+#include <glusterfs/glusterfs-acl.h>
#ifndef _DHT_H
#define _DHT_H
-#define GF_XATTR_FIX_LAYOUT_KEY "distribute.fix.layout"
-#define GF_XATTR_TIER_LAYOUT_FIXED_KEY "trusted.tier.fix.layout.complete"
-#define GF_XATTR_FILE_MIGRATE_KEY "trusted.distribute.migrate-data"
-#define DHT_MDS_STR "mds"
-#define GF_DHT_LOOKUP_UNHASHED_OFF 0
-#define GF_DHT_LOOKUP_UNHASHED_ON 1
-#define GF_DHT_LOOKUP_UNHASHED_AUTO 2
-#define DHT_PATHINFO_HEADER "DISTRIBUTE:"
-#define DHT_FILE_MIGRATE_DOMAIN "dht.file.migrate"
+#define GF_XATTR_FIX_LAYOUT_KEY "distribute.fix.layout"
+#define GF_XATTR_FILE_MIGRATE_KEY "trusted.distribute.migrate-data"
+#define DHT_MDS_STR "mds"
+#define GF_DHT_LOOKUP_UNHASHED_OFF 0
+#define GF_DHT_LOOKUP_UNHASHED_ON 1
+#define GF_DHT_LOOKUP_UNHASHED_AUTO 2
+#define DHT_PATHINFO_HEADER "DISTRIBUTE:"
+#define DHT_FILE_MIGRATE_DOMAIN "dht.file.migrate"
/* Layout synchronization */
-#define DHT_LAYOUT_HEAL_DOMAIN "dht.layout.heal"
+#define DHT_LAYOUT_HEAL_DOMAIN "dht.layout.heal"
/* Namespace synchronization */
-#define DHT_ENTRY_SYNC_DOMAIN "dht.entry.sync"
-#define TIERING_MIGRATION_KEY "tiering.migration"
-#define DHT_LAYOUT_HASH_INVALID 1
-#define MAX_REBAL_THREADS sysconf(_SC_NPROCESSORS_ONLN)
+#define DHT_ENTRY_SYNC_DOMAIN "dht.entry.sync"
+#define DHT_LAYOUT_HASH_INVALID 1
+#define MAX_REBAL_THREADS sysconf(_SC_NPROCESSORS_ONLN)
-#define DHT_DIR_STAT_BLOCKS 8
-#define DHT_DIR_STAT_SIZE 4096
+#define DHT_DIR_STAT_BLOCKS 8
+#define DHT_DIR_STAT_SIZE 4096
-/* Array to hold custom xattr keys
-*/
-extern char *xattrs_to_heal[];
+/* Virtual xattr for subvols status */
-/* Rebalance nodeuuid flags */
-#define REBAL_NODEUUID_MINE 0x01
+#define DHT_SUBVOL_STATUS_KEY "dht.subvol.status"
+
+/* Virtual xattrs for debugging */
+#define DHT_DBG_HASHED_SUBVOL_PATTERN "dht.file.hashed-subvol.*"
+#define DHT_DBG_HASHED_SUBVOL_KEY "dht.file.hashed-subvol."
+
+/* Rebalance nodeuuid flags */
+#define REBAL_NODEUUID_MINE 0x01
-typedef int (*dht_selfheal_dir_cbk_t) (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- dict_t *xdata);
-typedef int (*dht_defrag_cbk_fn_t) (xlator_t *this, xlator_t *dst_node,
- call_frame_t *frame, int ret);
+typedef int (*dht_selfheal_dir_cbk_t)(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata);
+typedef int (*dht_defrag_cbk_fn_t)(xlator_t *this, xlator_t *dst_node,
+ call_frame_t *frame, int ret);
-typedef int (*dht_refresh_layout_unlock) (call_frame_t *frame, xlator_t *this,
+typedef int (*dht_refresh_layout_unlock)(call_frame_t *frame, xlator_t *this,
int op_ret, int invoke_cbk);
-typedef int (*dht_refresh_layout_done_handle) (call_frame_t *frame);
+typedef int (*dht_refresh_layout_done_handle)(call_frame_t *frame);
struct dht_layout {
- int spread_cnt; /* layout spread count per directory,
- is controlled by 'setxattr()' with
- special key */
- int cnt;
- int preset;
- /*
- * The last *configuration* state for which this directory was known
- * to be in balance. The corresponding vol_commit_hash changes
- * whenever bricks are added or removed. This value changes when a
- * (full) rebalance is complete. If they match, it's safe to assume
- * that every file is where it should be and there's no need to do
- * lookups for files elsewhere. If they don't, then we have to do a
- * global lookup to be sure.
- */
- uint32_t commit_hash;
- /*
- * The *runtime* state of the volume, changes when connections to
- * bricks are made or lost.
- */
- int gen;
- int type;
- gf_atomic_t ref; /* use with dht_conf_t->layout_lock */
- uint32_t search_unhashed;
- struct {
- int err; /* 0 = normal
- -1 = dir exists and no xattr
- >0 = dir lookup failed with errno
- */
- uint32_t start;
- uint32_t stop;
- uint32_t commit_hash;
- xlator_t *xlator;
- } list[];
+ int spread_cnt; /* layout spread count per directory,
+ is controlled by 'setxattr()' with
+ special key */
+ int cnt;
+ int preset;
+ /*
+ * The last *configuration* state for which this directory was known
+ * to be in balance. The corresponding vol_commit_hash changes
+ * whenever bricks are added or removed. This value changes when a
+ * (full) rebalance is complete. If they match, it's safe to assume
+ * that every file is where it should be and there's no need to do
+ * lookups for files elsewhere. If they don't, then we have to do a
+ * global lookup to be sure.
+ */
+ uint32_t commit_hash;
+ /*
+ * The *runtime* state of the volume, changes when connections to
+ * bricks are made or lost.
+ */
+ int gen;
+ int type;
+ gf_atomic_t ref; /* use with dht_conf_t->layout_lock */
+ uint32_t search_unhashed;
+ struct {
+ int err; /* 0 = normal
+ -1 = dir exists and no xattr
+ >0 = dir lookup failed with errno
+ */
+ uint32_t start;
+ uint32_t stop;
+ uint32_t commit_hash;
+ xlator_t *xlator;
+ } list[];
};
-typedef struct dht_layout dht_layout_t;
+typedef struct dht_layout dht_layout_t;
struct dht_stat_time {
- uint32_t atime;
- uint32_t atime_nsec;
- uint32_t ctime;
- uint32_t ctime_nsec;
- uint32_t mtime;
- uint32_t mtime_nsec;
+ uint32_t atime;
+ uint32_t atime_nsec;
+ uint32_t ctime;
+ uint32_t ctime_nsec;
+ uint32_t mtime;
+ uint32_t mtime_nsec;
};
typedef struct dht_stat_time dht_stat_time_t;
struct dht_inode_ctx {
- dht_layout_t *layout;
- dht_stat_time_t time;
- xlator_t *lock_subvol;
- xlator_t *mds_subvol; /* This is only used for directories */
+ dht_layout_t *layout;
+ dht_stat_time_t time;
+ xlator_t *lock_subvol;
+ xlator_t *mds_subvol; /* This is only used for directories */
};
typedef struct dht_inode_ctx dht_inode_ctx_t;
-
typedef enum {
- DHT_HASH_TYPE_DM,
- DHT_HASH_TYPE_DM_USER,
+ DHT_HASH_TYPE_DM,
+ DHT_HASH_TYPE_DM_USER,
} dht_hashfn_type_t;
typedef enum {
- DHT_INODELK,
- DHT_ENTRYLK,
+ DHT_INODELK,
+ DHT_ENTRYLK,
} dht_lock_type_t;
/* rebalance related */
struct dht_rebalance_ {
- xlator_t *from_subvol;
- xlator_t *target_node;
- off_t offset;
- size_t size;
- int32_t flags;
- int count;
- struct iobref *iobref;
- struct iovec *vector;
- struct iatt stbuf;
- struct iatt prebuf;
- struct iatt postbuf;
- dht_defrag_cbk_fn_t target_op_fn;
- dict_t *xdata;
- dict_t *xattr;
- dict_t *dict;
- int32_t set;
- struct gf_flock flock;
- int lock_cmd;
+ xlator_t *from_subvol;
+ xlator_t *target_node;
+ off_t offset;
+ size_t size;
+ int32_t flags;
+ int count;
+ struct iobref *iobref;
+ struct iovec *vector;
+ struct iatt stbuf;
+ struct iatt prebuf;
+ struct iatt postbuf;
+ dht_defrag_cbk_fn_t target_op_fn;
+ dict_t *xdata;
+ dict_t *xattr;
+ dict_t *dict;
+ struct gf_flock flock;
+ int32_t set;
+ int lock_cmd;
};
/**
@@ -160,62 +158,62 @@ struct dht_rebalance_ {
* events
**/
typedef enum {
- qdstatfs_action_OFF = 0,
- qdstatfs_action_REPLACE,
- qdstatfs_action_NEGLECT,
- qdstatfs_action_COMPARE,
+ qdstatfs_action_OFF = 0,
+ qdstatfs_action_REPLACE,
+ qdstatfs_action_NEGLECT,
+ qdstatfs_action_COMPARE,
} qdstatfs_action_t;
typedef enum {
- REACTION_INVALID,
- FAIL_ON_ANY_ERROR,
- IGNORE_ENOENT_ESTALE
+ REACTION_INVALID,
+ FAIL_ON_ANY_ERROR,
+ IGNORE_ENOENT_ESTALE,
+ IGNORE_ENOENT_ESTALE_EIO,
} dht_reaction_type_t;
struct dht_skip_linkto_unlink {
-
- gf_boolean_t handle_valid_link;
- int opend_fd_count;
- xlator_t *hash_links_to;
- uuid_t cached_gfid;
- uuid_t hashed_gfid;
+ xlator_t *hash_links_to;
+ uuid_t cached_gfid;
+ uuid_t hashed_gfid;
+ int opend_fd_count;
+ gf_boolean_t handle_valid_link;
};
typedef struct {
- xlator_t *xl;
- loc_t loc; /* contains/points to inode to lock on. */
- short type; /* read/write lock. */
- char *domain; /* Only locks within a single domain
- * contend with each other
- */
- char *basename; /* Required for entrylk */
- gf_lkowner_t lk_owner;
- gf_boolean_t locked;
- dht_reaction_type_t do_on_failure;
+ xlator_t *xl;
+ loc_t loc; /* contains/points to inode to lock on. */
+ char *domain; /* Only locks within a single domain
+ * contend with each other
+ */
+ char *basename; /* Required for entrylk */
+ gf_boolean_t locked;
+ dht_reaction_type_t do_on_failure;
+ short type; /* read/write lock. */
+ gf_lkowner_t lk_owner;
} dht_lock_t;
/* The lock structure represents inodelk. */
typedef struct {
- fop_inodelk_cbk_t inodelk_cbk;
- dht_lock_t **locks;
- int lk_count;
- dht_reaction_type_t reaction;
-
- /* whether locking failed on _any_ of the "locks" above */
- int op_ret;
- int op_errno;
+ fop_inodelk_cbk_t inodelk_cbk;
+ dht_lock_t **locks;
+ int lk_count;
+ dht_reaction_type_t reaction;
+
+ /* whether locking failed on _any_ of the "locks" above */
+ int op_ret;
+ int op_errno;
} dht_ilock_wrap_t;
/* The lock structure represents entrylk. */
typedef struct {
- fop_entrylk_cbk_t entrylk_cbk;
- dht_lock_t **locks;
- int lk_count;
- dht_reaction_type_t reaction;
-
- /* whether locking failed on _any_ of the "locks" above */
- int op_ret;
- int op_errno;
+ fop_entrylk_cbk_t entrylk_cbk;
+ dht_lock_t **locks;
+ int lk_count;
+ dht_reaction_type_t reaction;
+
+ /* whether locking failed on _any_ of the "locks" above */
+ int op_ret;
+ int op_errno;
} dht_elock_wrap_t;
/* The first member of dht_dir_transaction_t should be of type dht_ilock_wrap_t.
@@ -224,1257 +222,1163 @@ typedef struct {
* lock[0].ns.parent_layout (like in dht_local_wipe).
*/
typedef union {
- union {
- dht_ilock_wrap_t my_layout;
- dht_ilock_wrap_t parent_layout;
- } layout;
- struct dht_namespace {
- dht_ilock_wrap_t parent_layout;
- dht_elock_wrap_t directory_ns;
- fop_entrylk_cbk_t ns_cbk;
- } ns;
+ union {
+ dht_ilock_wrap_t my_layout;
+ dht_ilock_wrap_t parent_layout;
+ } layout;
+ struct dht_namespace {
+ dht_ilock_wrap_t parent_layout;
+ dht_elock_wrap_t directory_ns;
+ fop_entrylk_cbk_t ns_cbk;
+ } ns;
} dht_dir_transaction_t;
-typedef
-int (*dht_selfheal_layout_t)(call_frame_t *frame, loc_t *loc,
- dht_layout_t *layout);
-
-typedef
-gf_boolean_t (*dht_need_heal_t)(call_frame_t *frame, dht_layout_t **inmem,
- dht_layout_t **ondisk);
+typedef int (*dht_selfheal_layout_t)(call_frame_t *frame, loc_t *loc,
+ dht_layout_t *layout);
-typedef struct {
- uint64_t blocks_used;
- uint64_t pblocks_used;
- uint64_t files_used;
- uint64_t pfiles_used;
- uint64_t unhashed_blocks_used;
- uint64_t unhashed_pblocks_used;
- uint64_t unhashed_files_used;
- uint64_t unhashed_pfiles_used;
- uint64_t unhashed_fsid;
- uint64_t hashed_fsid;
-} tier_statvfs_t;
+typedef gf_boolean_t (*dht_need_heal_t)(call_frame_t *frame,
+ dht_layout_t **inmem,
+ dht_layout_t **ondisk);
struct dht_local {
- int call_cnt;
- loc_t loc;
- loc_t loc2;
- int op_ret;
- int op_errno;
- int layout_mismatch;
- /* Use stbuf as the postbuf, when we require both
- * pre and post attrs */
- struct iatt stbuf;
- struct iatt mds_stbuf;
- struct iatt prebuf;
- struct iatt preoldparent;
- struct iatt postoldparent;
- struct iatt preparent;
- struct iatt postparent;
- struct statvfs statvfs;
- tier_statvfs_t tier_statvfs;
- fd_t *fd;
- inode_t *inode;
- dict_t *params;
- dict_t *xattr;
- dict_t *mds_xattr;
- dict_t *xdata; /* dict used to save xdata response by xattr fop */
- dict_t *xattr_req;
- dht_layout_t *layout;
- size_t size;
- ino_t ia_ino;
- xlator_t *src_hashed, *src_cached;
- xlator_t *dst_hashed, *dst_cached;
- xlator_t *cached_subvol;
- xlator_t *hashed_subvol;
- xlator_t *mds_subvol; /* This is use for dir only */
- char need_selfheal;
- char need_xattr_heal;
- int file_count;
- int dir_count;
- call_frame_t *main_frame;
- int fop_succeeded;
- struct {
- fop_mknod_cbk_t linkfile_cbk;
- struct iatt stbuf;
- loc_t loc;
- inode_t *inode;
- dict_t *xattr;
- xlator_t *srcvol;
- } linkfile;
- struct {
- uint32_t hole_cnt;
- uint32_t overlaps_cnt;
- uint32_t down;
- uint32_t misc;
- dht_selfheal_dir_cbk_t dir_cbk;
- dht_selfheal_layout_t healer;
- dht_need_heal_t should_heal;
- gf_boolean_t force_mkdir;
- dht_layout_t *layout, *refreshed_layout;
- } selfheal;
-
- dht_refresh_layout_unlock refresh_layout_unlock;
- dht_refresh_layout_done_handle refresh_layout_done;
-
- uint32_t uid;
- uint32_t gid;
-
- /* needed by nufa */
- int32_t flags;
- mode_t mode;
- dev_t rdev;
- mode_t umask;
-
- /* need for file-info */
- char *xattr_val;
- char *key;
-
- /* which xattr request? */
- char xsel[256];
- int32_t alloc_len;
-
- /* gfid related */
- uuid_t gfid;
-
- /* flag used to make sure we need to return estale in
- {lookup,revalidate}_cbk */
- char return_estale;
- char need_lookup_everywhere;
-
- glusterfs_fop_t fop;
-
- gf_boolean_t linked;
- xlator_t *link_subvol;
-
- struct dht_rebalance_ rebalance;
- xlator_t *first_up_subvol;
-
- gf_boolean_t quota_deem_statfs;
-
- gf_boolean_t added_link;
- gf_boolean_t is_linkfile;
-
- struct dht_skip_linkto_unlink skip_unlink;
-
- dht_dir_transaction_t lock[2], *current;
-
- short lock_type;
-
- call_stub_t *stub;
- int32_t parent_disk_layout[4];
-
- /* To hold dentries of readdir spawning across subvols */
- gf_dirent_t entries;
- size_t filled;
-
- /* rename rollback */
- int *ret_cache ;
-
- /* fd open check */
- gf_boolean_t fd_checked;
- /* This is use only for directory operation */
- int32_t valid;
- gf_boolean_t heal_layout;
- int32_t mds_heal_fresh_lookup;
+ loc_t loc;
+ loc_t loc2;
+ int call_cnt;
+ int op_ret;
+ int op_errno;
+ int layout_mismatch;
+ /* Use stbuf as the postbuf, when we require both
+ * pre and post attrs */
+ struct iatt stbuf;
+ struct iatt mds_stbuf;
+ struct iatt prebuf;
+ struct iatt preoldparent;
+ struct iatt postoldparent;
+ struct iatt preparent;
+ struct iatt postparent;
+ struct statvfs statvfs;
+ fd_t *fd;
+ inode_t *inode;
+ dict_t *params;
+ dict_t *xattr;
+ dict_t *mds_xattr;
+ dict_t *xdata; /* dict used to save xdata response by xattr fop */
+ dict_t *xattr_req;
+ dht_layout_t *layout;
+ size_t size;
+ ino_t ia_ino;
+ xlator_t *src_hashed, *src_cached;
+ xlator_t *dst_hashed, *dst_cached;
+ xlator_t *cached_subvol;
+ xlator_t *hashed_subvol;
+ xlator_t *mds_subvol; /* This is use for dir only */
+ int file_count;
+ int dir_count;
+ call_frame_t *main_frame;
+ int fop_succeeded;
+ struct {
+ fop_mknod_cbk_t linkfile_cbk;
+ struct iatt stbuf;
+ loc_t loc;
+ inode_t *inode;
+ dict_t *xattr;
+ xlator_t *srcvol;
+ } linkfile;
+ struct {
+ uint32_t hole_cnt;
+ uint32_t overlaps_cnt;
+ uint32_t down;
+ uint32_t misc;
+ dht_selfheal_dir_cbk_t dir_cbk;
+ dht_selfheal_layout_t healer;
+ dht_need_heal_t should_heal;
+ dht_layout_t *layout, *refreshed_layout;
+ uint32_t missing_cnt;
+ gf_boolean_t force_mkdir;
+ } selfheal;
+
+ dht_refresh_layout_unlock refresh_layout_unlock;
+ dht_refresh_layout_done_handle refresh_layout_done;
+
+ uint32_t uid;
+ uint32_t gid;
+ pid_t pid;
+
+ glusterfs_fop_t fop;
+
+ /* need for file-info */
+ char *xattr_val;
+ char *key;
+
+ /* needed by nufa */
+ int32_t flags;
+ mode_t mode;
+ dev_t rdev;
+ mode_t umask;
+
+ /* which xattr request? */
+ char xsel[256];
+ int32_t alloc_len;
+
+ /* gfid related */
+ uuid_t gfid;
+ uuid_t gfid_req;
+
+ xlator_t *link_subvol;
+
+ struct dht_rebalance_ rebalance;
+ xlator_t *first_up_subvol;
+
+ struct dht_skip_linkto_unlink skip_unlink;
+
+ dht_dir_transaction_t lock[2], *current;
+
+ /* inodelks during filerename for backward compatibility */
+ dht_lock_t **rename_inodelk_backward_compatible;
+
+ call_stub_t *stub;
+ int32_t parent_disk_layout[4];
+
+ /* rename rollback */
+ int *ret_cache;
+
+ loc_t loc2_copy;
+
+ int rename_inodelk_bc_count;
+ /* This is use only for directory operation */
+ int32_t valid;
+ int32_t mds_heal_fresh_lookup;
+ short lock_type;
+ char need_selfheal;
+ char need_xattr_heal;
+ char need_attrheal;
+ /* flag used to make sure we need to return estale in
+ {lookup,revalidate}_cbk */
+ char return_estale;
+ char need_lookup_everywhere;
+ /* fd open check */
+ gf_boolean_t fd_checked;
+ gf_boolean_t linked;
+ gf_boolean_t added_link;
+ gf_boolean_t is_linkfile;
+ gf_boolean_t quota_deem_statfs;
+ gf_boolean_t heal_layout;
+ gf_boolean_t locked;
+ gf_boolean_t dont_create_linkto;
+ gf_boolean_t gfid_missing;
};
typedef struct dht_local dht_local_t;
/* du - disk-usage */
struct dht_du {
- double avail_percent;
- double avail_inodes;
- uint64_t avail_space;
- uint32_t log;
- uint32_t chunks;
- uint32_t total_blocks;
- uint32_t avail_blocks;
- uint32_t frsize; /*fragment size*/
+ double avail_percent;
+ double avail_inodes;
+ uint64_t avail_space;
+ uint32_t log;
+ uint32_t chunks;
+ uint32_t total_blocks;
+ uint32_t avail_blocks;
+ uint32_t frsize; /*fragment size*/
};
typedef struct dht_du dht_du_t;
enum gf_defrag_type {
- GF_DEFRAG_CMD_NONE = 0,
- GF_DEFRAG_CMD_START = 1,
- GF_DEFRAG_CMD_STOP = 1 + 1,
- GF_DEFRAG_CMD_STATUS = 1 + 2,
- GF_DEFRAG_CMD_START_LAYOUT_FIX = 1 + 3,
- GF_DEFRAG_CMD_START_FORCE = 1 + 4,
- GF_DEFRAG_CMD_START_TIER = 1 + 5,
- GF_DEFRAG_CMD_STATUS_TIER = 1 + 6,
- GF_DEFRAG_CMD_START_DETACH_TIER = 1 + 7,
- GF_DEFRAG_CMD_STOP_DETACH_TIER = 1 + 8,
- GF_DEFRAG_CMD_PAUSE_TIER = 1 + 9,
- GF_DEFRAG_CMD_RESUME_TIER = 1 + 10,
- GF_DEFRAG_CMD_DETACH_STATUS = 1 + 11,
- GF_DEFRAG_CMD_STOP_TIER = 1 + 12,
- GF_DEFRAG_CMD_DETACH_START = 1 + 13,
- GF_DEFRAG_CMD_DETACH_COMMIT = 1 + 14,
- GF_DEFRAG_CMD_DETACH_COMMIT_FORCE = 1 + 15,
- GF_DEFRAG_CMD_DETACH_STOP = 1 + 16,
- /* new labels are used so it will help
- * while removing old labels by easily differentiating.
- * A few labels are added so that the count remains same
- * between this enum and the ones on the xdr file.
- * different values for the same enum cause errors and
- * confusion.
- */
+ GF_DEFRAG_CMD_NONE = 0,
+ GF_DEFRAG_CMD_START = 1,
+ GF_DEFRAG_CMD_STOP = 1 + 1,
+ GF_DEFRAG_CMD_STATUS = 1 + 2,
+ GF_DEFRAG_CMD_START_LAYOUT_FIX = 1 + 3,
+ GF_DEFRAG_CMD_START_FORCE = 1 + 4,
+ GF_DEFRAG_CMD_DETACH_STATUS = 1 + 11,
+ GF_DEFRAG_CMD_DETACH_START = 1 + 13,
+ GF_DEFRAG_CMD_DETACH_COMMIT = 1 + 14,
+ GF_DEFRAG_CMD_DETACH_COMMIT_FORCE = 1 + 15,
+ GF_DEFRAG_CMD_DETACH_STOP = 1 + 16,
+ /* new labels are used so it will help
+ * while removing old labels by easily differentiating.
+ * A few labels are added so that the count remains same
+ * between this enum and the ones on the xdr file.
+ * different values for the same enum cause errors and
+ * confusion.
+ */
};
typedef enum gf_defrag_type gf_defrag_type;
enum gf_defrag_status_t {
- GF_DEFRAG_STATUS_NOT_STARTED,
- GF_DEFRAG_STATUS_STARTED,
- GF_DEFRAG_STATUS_STOPPED,
- GF_DEFRAG_STATUS_COMPLETE,
- GF_DEFRAG_STATUS_FAILED,
- GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED,
- GF_DEFRAG_STATUS_LAYOUT_FIX_STOPPED,
- GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE,
- GF_DEFRAG_STATUS_LAYOUT_FIX_FAILED,
+ GF_DEFRAG_STATUS_NOT_STARTED,
+ GF_DEFRAG_STATUS_STARTED,
+ GF_DEFRAG_STATUS_STOPPED,
+ GF_DEFRAG_STATUS_COMPLETE,
+ GF_DEFRAG_STATUS_FAILED,
+ GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED,
+ GF_DEFRAG_STATUS_LAYOUT_FIX_STOPPED,
+ GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE,
+ GF_DEFRAG_STATUS_LAYOUT_FIX_FAILED,
};
typedef enum gf_defrag_status_t gf_defrag_status_t;
typedef struct gf_defrag_pattern_list gf_defrag_pattern_list_t;
struct gf_defrag_pattern_list {
- char path_pattern[256];
- uint64_t size;
- gf_defrag_pattern_list_t *next;
+ char path_pattern[256];
+ uint64_t size;
+ gf_defrag_pattern_list_t *next;
};
struct dht_container {
- union {
- struct list_head list;
- struct {
- struct _gf_dirent_t *next;
- struct _gf_dirent_t *prev;
- };
+ union {
+ struct list_head list;
+ struct {
+ struct _gf_dirent_t *next;
+ struct _gf_dirent_t *prev;
};
- gf_dirent_t *df_entry;
- xlator_t *this;
- loc_t *parent_loc;
- dict_t *migrate_data;
- int local_subvol_index;
+ };
+ gf_dirent_t *df_entry;
+ xlator_t *this;
+ loc_t *parent_loc;
+ dict_t *migrate_data;
+ int local_subvol_index;
};
-typedef enum tier_mode_ {
- TIER_MODE_NONE = 0,
- TIER_MODE_TEST,
- TIER_MODE_WM
-} tier_mode_t;
-
-typedef enum tier_pause_state_ {
- TIER_RUNNING = 0,
- TIER_REQUEST_PAUSE,
- TIER_PAUSED
-} tier_pause_state_t;
-
-/* This Structure is only used in tiering fixlayout */
-typedef struct gf_tier_fix_layout_arg {
- xlator_t *this;
- dict_t *fix_layout;
- pthread_t thread_id;
-} gf_tier_fix_layout_arg_t;
-
-typedef struct gf_tier_conf {
- int is_tier;
- int watermark_hi;
- int watermark_low;
- int watermark_last;
- unsigned long block_size;
- fsblkcnt_t blocks_total;
- fsblkcnt_t blocks_used;
- int percent_full;
- uint64_t max_migrate_bytes;
- int max_migrate_files;
- int query_limit;
- tier_mode_t mode;
- /* These flags are only used for tier-compact */
- gf_boolean_t compact_active;
- /* These 3 flags are set to true when the client changes the */
- /* compaction mode on the command line. */
- /* When they are set, the daemon will trigger compaction as */
- /* soon as possible to activate or deactivate compaction. */
- /* If in the middle of a compaction, then the switches take */
- /* effect on the next compaction, not the current one. */
- /* If the user switches it off, we want to avoid needless */
- /* compactions. */
- /* If the user switches it on, they want to compact as soon */
- /* as possible. */
- gf_boolean_t compact_mode_switched;
- gf_boolean_t compact_mode_switched_hot;
- gf_boolean_t compact_mode_switched_cold;
- int tier_max_promote_size;
- int tier_promote_frequency;
- int tier_demote_frequency;
- int tier_compact_hot_frequency;
- int tier_compact_cold_frequency;
- uint64_t st_last_promoted_size;
- uint64_t st_last_demoted_size;
- tier_pause_state_t pause_state;
- struct synctask *pause_synctask;
- gf_timer_t *pause_timer;
- pthread_mutex_t pause_mutex;
- int promote_in_progress;
- int demote_in_progress;
- /* This Structure is only used in tiering fixlayout */
- gf_tier_fix_layout_arg_t tier_fix_layout_arg;
- /* Indicates the index of the first queryfile picked
- * in the last cycle of promote or demote */
- int32_t last_promote_qfile_index;
- int32_t last_demote_qfile_index;
- char volname[GD_VOLUME_NAME_MAX + 1];
-} gf_tier_conf_t;
-
typedef struct nodeuuid_info {
- char info; /* Set to 1 is this is my node's uuid*/
- uuid_t uuid; /* Store the nodeuuid as well for debugging*/
+ char info; /* Set to 1 is this is my node's uuid*/
+ uuid_t uuid; /* Store the nodeuuid as well for debugging*/
} nodeuuid_info_t;
typedef struct subvol_nodeuuids_info {
- nodeuuid_info_t *elements;
- int count;
+ nodeuuid_info_t *elements;
+ int count;
} subvol_nodeuuids_info_t;
-
struct gf_defrag_info_ {
- uint64_t total_files;
- uint64_t total_data;
- uint64_t num_files_lookedup;
- uint64_t total_failures;
- uint64_t skipped;
- uint64_t num_dirs_processed;
- uint64_t size_processed;
- gf_lock_t lock;
- int cmd;
- pthread_t th;
- gf_defrag_status_t defrag_status;
- struct rpc_clnt *rpc;
- uint32_t connected;
- uint32_t is_exiting;
- pid_t pid;
- inode_t *root_inode;
- uuid_t node_uuid;
- struct timeval start_time;
- gf_boolean_t stats;
- uint32_t new_commit_hash;
- gf_defrag_pattern_list_t *defrag_pattern;
- gf_tier_conf_t tier_conf;
-
- /*Data Tiering params for scanner*/
- uint64_t total_files_promoted;
- uint64_t total_files_demoted;
- int write_freq_threshold;
- int read_freq_threshold;
-
- pthread_cond_t parallel_migration_cond;
- pthread_mutex_t dfq_mutex;
- pthread_cond_t rebalance_crawler_alarm;
- int32_t q_entry_count;
- int32_t global_error;
- struct dht_container *queue;
- int32_t crawl_done;
- int32_t abort;
- int32_t wakeup_crawler;
-
- /*Throttle params*/
- /*stands for reconfigured thread count*/
- int32_t recon_thread_count;
- /*stands for current running thread count*/
- int32_t current_thread_count;
- pthread_cond_t df_wakeup_thread;
-
- /* lock migration flag */
- gf_boolean_t lock_migration_enabled;
-
- /* backpointer to make it easier to write functions for rebalance */
- xlator_t *this;
-
- pthread_cond_t fc_wakeup_cond;
- pthread_mutex_t fc_mutex;
+ uint64_t total_files;
+ uint64_t total_data;
+ uint64_t num_files_lookedup;
+ uint64_t total_failures;
+ uint64_t skipped;
+ uint64_t num_dirs_processed;
+ uint64_t size_processed;
+ gf_lock_t lock;
+ pthread_t th;
+ struct rpc_clnt *rpc;
+ uint32_t connected;
+ uint32_t is_exiting;
+ pid_t pid;
+ int cmd;
+ inode_t *root_inode;
+ uuid_t node_uuid;
+ time_t start_time;
+ uint32_t new_commit_hash;
+ gf_defrag_status_t defrag_status;
+ gf_defrag_pattern_list_t *defrag_pattern;
+
+ pthread_cond_t parallel_migration_cond;
+ pthread_mutex_t dfq_mutex;
+ pthread_cond_t rebalance_crawler_alarm;
+ int32_t q_entry_count;
+ int32_t global_error;
+ struct dht_container *queue;
+ int32_t crawl_done;
+ int32_t abort;
+ int32_t wakeup_crawler;
+
+ /*Throttle params*/
+ /*stands for reconfigured thread count*/
+ int32_t recon_thread_count;
+ pthread_cond_t df_wakeup_thread;
+
+ /* backpointer to make it easier to write functions for rebalance */
+ xlator_t *this;
+
+ pthread_cond_t fc_wakeup_cond;
+ pthread_mutex_t fc_mutex;
+
+ /*stands for current running thread count*/
+ int32_t current_thread_count;
+
+ gf_boolean_t stats;
+ /* lock migration flag */
+ gf_boolean_t lock_migration_enabled;
};
typedef struct gf_defrag_info_ gf_defrag_info_t;
struct dht_methods_s {
- int32_t (*migration_get_dst_subvol)(xlator_t *this,
- dht_local_t *local);
- int32_t (*migration_other)(xlator_t *this,
- gf_defrag_info_t *defrag);
- int32_t (*migration_needed)(xlator_t *this);
- xlator_t* (*layout_search)(xlator_t *this,
- dht_layout_t *layout,
- const char *name);
+ int32_t (*migration_get_dst_subvol)(xlator_t *this, dht_local_t *local);
+ int32_t (*migration_other)(xlator_t *this, gf_defrag_info_t *defrag);
+ xlator_t *(*layout_search)(xlator_t *this, dht_layout_t *layout,
+ const char *name);
};
typedef struct dht_methods_s dht_methods_t;
struct dht_conf {
- gf_lock_t subvolume_lock;
- int subvolume_cnt;
- xlator_t **subvolumes;
- char *subvolume_status;
- int *last_event;
- dht_layout_t **file_layouts;
- dht_layout_t **dir_layouts;
- unsigned int search_unhashed;
- gf_boolean_t lookup_optimize;
- int gen;
- dht_du_t *du_stats;
- double min_free_disk;
- double min_free_inodes;
- char disk_unit;
- int32_t refresh_interval;
- gf_boolean_t unhashed_sticky_bit;
- struct timeval last_stat_fetch;
- gf_lock_t layout_lock;
- dict_t *leaf_to_subvol;
- void *private; /* Can be used by wrapper xlators over
- dht */
- gf_boolean_t use_readdirp;
- char vol_uuid[UUID_SIZE + 1];
- gf_boolean_t assert_no_child_down;
- time_t *subvol_up_time;
-
- /* This is the count used as the distribute layout for a directory */
- /* Will be a global flag to control the layout spread count */
- uint32_t dir_spread_cnt;
-
- /* to keep track of nodes which are decommissioned */
- xlator_t **decommissioned_bricks;
- int decommission_in_progress;
- int decommission_subvols_cnt;
-
- /* defrag related */
- gf_defrag_info_t *defrag;
-
- /* Request to filter directory entries in readdir request */
-
- gf_boolean_t readdir_optimize;
-
- /* Support regex-based name reinterpretation. */
- regex_t rsync_regex;
- gf_boolean_t rsync_regex_valid;
- regex_t extra_regex;
- gf_boolean_t extra_regex_valid;
-
- /* Support variable xattr names. */
- char *xattr_name;
- char *mds_xattr_key;
- char *link_xattr_name;
- char *commithash_xattr_name;
- char *wild_xattr_name;
-
- /* Support size-weighted rebalancing (heterogeneous bricks). */
- gf_boolean_t do_weighting;
- gf_boolean_t randomize_by_gfid;
- int dthrottle;
-
- dht_methods_t methods;
-
- struct mem_pool *lock_pool;
-
- /*local subvol storage for rebalance*/
- xlator_t **local_subvols;
- subvol_nodeuuids_info_t *local_nodeuuids;
- int32_t local_subvols_cnt;
-
- /*
- * "Commit hash" for this volume topology. Changed whenever bricks
- * are added or removed.
- */
- uint32_t vol_commit_hash;
- gf_boolean_t vch_forced;
-
- /* lock migration */
-
- gf_boolean_t lock_migration_enabled;
- gf_lock_t lock;
-
- /* Hard link handle requirement for migration triggered from client*/
- synclock_t link_lock;
-
- gf_boolean_t use_fallocate;
-
- gf_boolean_t force_migration;
+ xlator_t **subvolumes;
+ char *subvolume_status;
+ int *last_event;
+ dht_layout_t **file_layouts;
+ dht_layout_t **dir_layouts;
+ unsigned int search_unhashed;
+ int gen;
+ dht_du_t *du_stats;
+ double min_free_disk;
+ double min_free_inodes;
+ int subvolume_cnt;
+ int32_t refresh_interval;
+ gf_lock_t subvolume_lock;
+ time_t last_stat_fetch;
+ gf_lock_t layout_lock;
+ dict_t *leaf_to_subvol;
+ void *private; /* Can be used by wrapper xlators over
+ dht */
+ time_t *subvol_up_time;
+
+ /* to keep track of nodes which are decommissioned */
+ xlator_t **decommissioned_bricks;
+ int decommission_in_progress;
+ int decommission_subvols_cnt;
+
+ /* defrag related */
+ gf_defrag_info_t *defrag;
+
+ /* Support regex-based name reinterpretation. */
+ regex_t rsync_regex;
+ regex_t extra_regex;
+
+ /* Support variable xattr names. */
+ char *xattr_name;
+ char *mds_xattr_key;
+ char *link_xattr_name;
+ char *commithash_xattr_name;
+ char *wild_xattr_name;
+
+ dht_methods_t methods;
+
+ struct mem_pool *lock_pool;
+
+ /*local subvol storage for rebalance*/
+ xlator_t **local_subvols;
+ subvol_nodeuuids_info_t *local_nodeuuids;
+ int32_t local_subvols_cnt;
+
+ int dthrottle;
+
+ /* Hard link handle requirement for migration triggered from client*/
+ synclock_t link_lock;
+
+ /* lock migration */
+ gf_lock_t lock;
+
+ /* This is the count used as the distribute layout for a directory */
+ /* Will be a global flag to control the layout spread count */
+ uint32_t dir_spread_cnt;
+
+ /*
+ * "Commit hash" for this volume topology. Changed whenever bricks
+ * are added or removed.
+ */
+ uint32_t vol_commit_hash;
+
+ char vol_uuid[UUID_SIZE + 1];
+
+ char disk_unit;
+
+ gf_boolean_t lock_migration_enabled;
+
+ gf_boolean_t vch_forced;
+
+ gf_boolean_t use_fallocate;
+
+ gf_boolean_t force_migration;
+
+ gf_boolean_t lookup_optimize;
+
+ gf_boolean_t unhashed_sticky_bit;
+
+ gf_boolean_t assert_no_child_down;
+
+ gf_boolean_t use_readdirp;
+
+ /* Request to filter directory entries in readdir request */
+ gf_boolean_t readdir_optimize;
+
+ gf_boolean_t rsync_regex_valid;
+
+ gf_boolean_t extra_regex_valid;
+
+ /* Support size-weighted rebalancing (heterogeneous bricks). */
+ gf_boolean_t do_weighting;
+
+ gf_boolean_t randomize_by_gfid;
};
typedef struct dht_conf dht_conf_t;
struct dht_dfoffset_ctx {
- xlator_t *this;
- off_t offset;
- int32_t readdir_done;
+ xlator_t *this;
+ off_t offset;
+ int32_t readdir_done;
};
typedef struct dht_dfoffset_ctx dht_dfoffset_ctx_t;
struct dht_disk_layout {
- uint32_t cnt;
- uint32_t type;
- struct {
- uint32_t start;
- uint32_t stop;
- } list[1];
+ uint32_t cnt;
+ uint32_t type;
+ struct {
+ uint32_t start;
+ uint32_t stop;
+ } list[1];
};
typedef struct dht_disk_layout dht_disk_layout_t;
typedef enum {
- GF_DHT_MIGRATE_DATA,
- GF_DHT_MIGRATE_DATA_EVEN_IF_LINK_EXISTS,
- GF_DHT_MIGRATE_HARDLINK,
- GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS
+ GF_DHT_MIGRATE_DATA,
+ GF_DHT_MIGRATE_DATA_EVEN_IF_LINK_EXISTS,
+ GF_DHT_MIGRATE_HARDLINK,
+ GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS
} gf_dht_migrate_data_type_t;
typedef enum {
- GF_DHT_EQUAL_DISTRIBUTION,
- GF_DHT_WEIGHTED_DISTRIBUTION
+ GF_DHT_EQUAL_DISTRIBUTION,
+ GF_DHT_WEIGHTED_DISTRIBUTION
} dht_distribution_type_t;
struct dir_dfmeta {
- gf_dirent_t *equeue;
- dht_dfoffset_ctx_t *offset_var;
- struct list_head **head;
- struct list_head **iterator;
- int *fetch_entries;
+ gf_dirent_t *equeue;
+ dht_dfoffset_ctx_t *offset_var;
+ struct list_head **head;
+ struct list_head **iterator;
+ int *fetch_entries;
+ /* fds corresponding to local subvols only */
+ fd_t **lfd;
};
typedef struct dht_migrate_info {
- xlator_t *src_subvol;
- xlator_t *dst_subvol;
- GF_REF_DECL;
+ xlator_t *src_subvol;
+ xlator_t *dst_subvol;
+ GF_REF_DECL;
} dht_migrate_info_t;
-
-
typedef struct dht_fd_ctx {
- uint64_t opened_on_dst;
- GF_REF_DECL;
+ uint64_t opened_on_dst;
+ GF_REF_DECL;
} dht_fd_ctx_t;
-
#define ENTRY_MISSING(op_ret, op_errno) (op_ret == -1 && op_errno == ENOENT)
-#define is_revalidate(loc) (dht_inode_ctx_layout_get ((loc)->inode, this, NULL) == 0)
+#define is_revalidate(loc) \
+ (dht_inode_ctx_layout_get((loc)->inode, this, NULL) == 0)
#define is_last_call(cnt) (cnt == 0)
#define DHT_MIGRATION_IN_PROGRESS 1
-#define DHT_MIGRATION_COMPLETED 2
+#define DHT_MIGRATION_COMPLETED 2
-#define check_is_linkfile(i,s,x,n) (IS_DHT_LINKFILE_MODE (s) && dict_get (x, n))
+#define check_is_linkfile(i, s, x, n) \
+ (IS_DHT_LINKFILE_MODE(s) && dict_get(x, n))
-#define IS_DHT_MIGRATION_PHASE2(buf) ( \
- IA_ISREG ((buf)->ia_type) && \
- ((st_mode_from_ia ((buf)->ia_prot, (buf)->ia_type) & \
- ~S_IFMT) == DHT_LINKFILE_MODE))
+#define IS_DHT_MIGRATION_PHASE2(buf) \
+ (IA_ISREG((buf)->ia_type) && \
+ ((st_mode_from_ia((buf)->ia_prot, (buf)->ia_type) & ~S_IFMT) == \
+ DHT_LINKFILE_MODE))
-#define IS_DHT_MIGRATION_PHASE1(buf) ( \
- IA_ISREG ((buf)->ia_type) && \
- ((buf)->ia_prot.sticky == 1) && \
- ((buf)->ia_prot.sgid == 1))
+#define IS_DHT_MIGRATION_PHASE1(buf) \
+ (IA_ISREG((buf)->ia_type) && ((buf)->ia_prot.sticky == 1) && \
+ ((buf)->ia_prot.sgid == 1))
-#define DHT_STRIP_PHASE1_FLAGS(buf) do { \
- if ((buf) && IS_DHT_MIGRATION_PHASE1(buf)) { \
- (buf)->ia_prot.sticky = 0; \
- (buf)->ia_prot.sgid = 0; \
- } \
- } while (0)
+#define DHT_STRIP_PHASE1_FLAGS(buf) \
+ do { \
+ if ((buf) && IS_DHT_MIGRATION_PHASE1(buf)) { \
+ (buf)->ia_prot.sticky = 0; \
+ (buf)->ia_prot.sgid = 0; \
+ } \
+ } while (0)
#define dht_inode_missing(op_errno) (op_errno == ENOENT || op_errno == ESTALE)
-#define check_is_dir(i,s,x) (IA_ISDIR(s->ia_type))
+#define check_is_dir(i, s, x) (IA_ISDIR(s->ia_type))
#define layout_is_sane(layout) ((layout) && (layout->cnt > 0))
-#define we_are_not_migrating(x) ((x) == 1)
-
-#define DHT_STACK_UNWIND(fop, frame, params ...) do { \
- dht_local_t *__local = NULL; \
- xlator_t *__xl = NULL; \
- if (frame) { \
- __xl = frame->this; \
- __local = frame->local; \
- frame->local = NULL; \
- } \
- STACK_UNWIND_STRICT (fop, frame, params); \
- dht_local_wipe (__xl, __local); \
- } while (0)
-
-#define DHT_STACK_DESTROY(frame) do { \
- dht_local_t *__local = NULL; \
- xlator_t *__xl = NULL; \
- __xl = frame->this; \
- __local = frame->local; \
- frame->local = NULL; \
- STACK_DESTROY (frame->root); \
- dht_local_wipe (__xl, __local); \
- } while (0)
-
-#define DHT_UPDATE_TIME(ctx_sec, ctx_nsec, new_sec, new_nsec, inode, post) do {\
- LOCK (&inode->lock); \
- { \
- if (ctx_sec == new_sec) \
- new_nsec = max (new_nsec, ctx_nsec); \
- else if (ctx_sec > new_sec) { \
- new_sec = ctx_sec; \
- new_nsec = ctx_nsec; \
- } \
- if (post) { \
- ctx_sec = new_sec; \
- ctx_nsec = new_nsec; \
- } \
- } \
- UNLOCK (&inode->lock); \
- } while (0)
-
-#define is_greater_time(a, an, b, bn) (((a) < (b)) || (((a) == (b)) && ((an) < (bn))))
-
-#define DHT_MARK_FOP_INTERNAL(xattr) do { \
- int tmp = -1; \
- if (!xattr) { \
- xattr = dict_new (); \
- if (!xattr) \
- break; \
- } \
- tmp = dict_set_str (xattr, GLUSTERFS_INTERNAL_FOP_KEY, "yes"); \
- if (tmp) { \
- gf_msg (this->name, GF_LOG_ERROR, 0, \
- DHT_MSG_DICT_SET_FAILED, \
- "Failed to set dictionary value: key = %s," \
- " path = %s", GLUSTERFS_INTERNAL_FOP_KEY, \
- local->loc.path); \
- } \
- } while (0)
-
-dht_layout_t *dht_layout_new (xlator_t *this, int cnt);
-dht_layout_t *dht_layout_get (xlator_t *this, inode_t *inode);
-dht_layout_t *dht_layout_for_subvol (xlator_t *this, xlator_t *subvol);
-xlator_t *dht_layout_search (xlator_t *this, dht_layout_t *layout,
- const char *name);
+#define we_are_not_migrating(x) ((x) == 1)
+
+#define DHT_STACK_UNWIND(fop, frame, params...) \
+ do { \
+ dht_local_t *__local = NULL; \
+ xlator_t *__xl = NULL; \
+ if (frame) { \
+ __xl = frame->this; \
+ __local = frame->local; \
+ frame->local = NULL; \
+ } \
+ STACK_UNWIND_STRICT(fop, frame, params); \
+ dht_local_wipe(__xl, __local); \
+ } while (0)
+
+#define DHT_STACK_DESTROY(frame) \
+ do { \
+ dht_local_t *__local = NULL; \
+ xlator_t *__xl = NULL; \
+ __xl = frame->this; \
+ __local = frame->local; \
+ frame->local = NULL; \
+ STACK_DESTROY(frame->root); \
+ dht_local_wipe(__xl, __local); \
+ } while (0)
+
+#define DHT_UPDATE_TIME(ctx_sec, ctx_nsec, new_sec, new_nsec, post) \
+ do { \
+ if (ctx_sec == new_sec) \
+ new_nsec = max(new_nsec, ctx_nsec); \
+ else if (ctx_sec > new_sec) { \
+ new_sec = ctx_sec; \
+ new_nsec = ctx_nsec; \
+ } \
+ if (post) { \
+ ctx_sec = new_sec; \
+ ctx_nsec = new_nsec; \
+ } \
+ } while (0)
+
+#define is_greater_time(a, an, b, bn) \
+ (((a) < (b)) || (((a) == (b)) && ((an) < (bn))))
+
+#define DHT_MARK_FOP_INTERNAL(xattr) \
+ do { \
+ int tmp = -1; \
+ if (!xattr) { \
+ xattr = dict_new(); \
+ if (!xattr) \
+ break; \
+ } \
+ tmp = dict_set_str(xattr, GLUSTERFS_INTERNAL_FOP_KEY, "yes"); \
+ if (tmp) { \
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED, \
+ "Failed to set dictionary value: key = %s," \
+ " path = %s", \
+ GLUSTERFS_INTERNAL_FOP_KEY, local->loc.path); \
+ } \
+ } while (0)
+
+dht_layout_t *
+dht_layout_new(xlator_t *this, int cnt);
+dht_layout_t *
+dht_layout_get(xlator_t *this, inode_t *inode);
+dht_layout_t *
+dht_layout_for_subvol(xlator_t *this, xlator_t *subvol);
+xlator_t *
+dht_layout_search(xlator_t *this, dht_layout_t *layout, const char *name);
int32_t
-dht_migration_get_dst_subvol(xlator_t *this, dht_local_t *local);
+dht_migration_get_dst_subvol(xlator_t *this, dht_local_t *local);
int32_t
dht_migration_needed(xlator_t *this);
-int dht_layout_normalize (xlator_t *this, loc_t *loc, dht_layout_t *layout);
-int dht_layout_anomalies (xlator_t *this, loc_t *loc, dht_layout_t *layout,
- uint32_t *holes_p, uint32_t *overlaps_p,
- uint32_t *missing_p, uint32_t *down_p,
- uint32_t *misc_p, uint32_t *no_space_p);
-int dht_layout_dir_mismatch (xlator_t *this, dht_layout_t *layout,
- xlator_t *subvol, loc_t *loc, dict_t *xattr);
-
-xlator_t *dht_linkfile_subvol (xlator_t *this, inode_t *inode,
- struct iatt *buf, dict_t *xattr);
-int dht_linkfile_unlink (call_frame_t *frame, xlator_t *this,
- xlator_t *subvol, loc_t *loc);
-
-int dht_layouts_init (xlator_t *this, dht_conf_t *conf);
-int dht_layout_merge (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
- int op_ret, int op_errno, dict_t *xattr);
-
-int dht_disk_layout_extract (xlator_t *this, dht_layout_t *layout,
- int pos, int32_t **disk_layout_p);
-int dht_disk_layout_merge (xlator_t *this, dht_layout_t *layout,
- int pos, void *disk_layout_raw, int disk_layout_len);
int
-dht_disk_layout_extract_for_subvol (xlator_t *this, dht_layout_t *layout,
- xlator_t *subvol, int32_t **disk_layout_p);
+dht_layout_normalize(xlator_t *this, loc_t *loc, dht_layout_t *layout);
+void
+dht_layout_anomalies(xlator_t *this, loc_t *loc, dht_layout_t *layout,
+ uint32_t *holes_p, uint32_t *overlaps_p,
+ uint32_t *missing_p, uint32_t *down_p, uint32_t *misc_p,
+ uint32_t *no_space_p);
+int
+dht_layout_dir_mismatch(xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
+ loc_t *loc, dict_t *xattr);
+xlator_t *
+dht_linkfile_subvol(xlator_t *this, inode_t *inode, struct iatt *buf,
+ dict_t *xattr);
+int
+dht_linkfile_unlink(call_frame_t *frame, xlator_t *this, xlator_t *subvol,
+ loc_t *loc);
-int dht_frame_return (call_frame_t *frame);
+int
+dht_layouts_init(xlator_t *this, dht_conf_t *conf);
+int
+dht_layout_merge(xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
+ int op_ret, int op_errno, dict_t *xattr);
-int dht_deitransform (xlator_t *this, uint64_t y, xlator_t **subvol);
+int
+dht_disk_layout_extract(xlator_t *this, dht_layout_t *layout, int pos,
+ int32_t **disk_layout_p);
+int
+dht_disk_layout_extract_for_subvol(xlator_t *this, dht_layout_t *layout,
+ xlator_t *subvol, int32_t **disk_layout_p);
-void dht_local_wipe (xlator_t *this, dht_local_t *local);
-dht_local_t *dht_local_init (call_frame_t *frame, loc_t *loc, fd_t *fd,
- glusterfs_fop_t fop);
-int dht_iatt_merge (xlator_t *this, struct iatt *to, struct iatt *from,
- xlator_t *subvol);
+int
+dht_frame_return(call_frame_t *frame);
-xlator_t *dht_subvol_get_hashed (xlator_t *this, loc_t *loc);
-xlator_t *dht_subvol_get_cached (xlator_t *this, inode_t *inode);
-xlator_t *dht_subvol_next (xlator_t *this, xlator_t *prev);
-xlator_t *dht_subvol_next_available (xlator_t *this, xlator_t *prev);
-int dht_subvol_cnt (xlator_t *this, xlator_t *subvol);
+int
+dht_deitransform(xlator_t *this, uint64_t y, xlator_t **subvol);
-int dht_hash_compute (xlator_t *this, int type, const char *name, uint32_t *hash_p);
+void
+dht_local_wipe(xlator_t *this, dht_local_t *local);
+dht_local_t *
+dht_local_init(call_frame_t *frame, loc_t *loc, fd_t *fd, glusterfs_fop_t fop);
+int
+dht_iatt_merge(xlator_t *this, struct iatt *to, struct iatt *from);
-int dht_linkfile_create (call_frame_t *frame, fop_mknod_cbk_t linkfile_cbk,
- xlator_t *this, xlator_t *tovol,
- xlator_t *fromvol, loc_t *loc);
-int dht_lookup_directory (call_frame_t *frame, xlator_t *this, loc_t *loc);
-int dht_lookup_everywhere (call_frame_t *frame, xlator_t *this, loc_t *loc);
+xlator_t *
+dht_subvol_get_hashed(xlator_t *this, loc_t *loc);
+xlator_t *
+dht_subvol_get_cached(xlator_t *this, inode_t *inode);
+xlator_t *
+dht_subvol_next(xlator_t *this, xlator_t *prev);
+xlator_t *
+dht_subvol_next_available(xlator_t *this, xlator_t *prev);
int
-dht_selfheal_directory (call_frame_t *frame, dht_selfheal_dir_cbk_t cbk,
- loc_t *loc, dht_layout_t *layout);
+dht_subvol_cnt(xlator_t *this, xlator_t *subvol);
int
-dht_selfheal_directory_for_nameless_lookup (call_frame_t *frame,
- dht_selfheal_dir_cbk_t cbk,
- loc_t *loc, dht_layout_t *layout);
+dht_hash_compute(xlator_t *this, int type, const char *name, uint32_t *hash_p);
int
-dht_selfheal_new_directory (call_frame_t *frame, dht_selfheal_dir_cbk_t cbk,
- dht_layout_t *layout);
+dht_linkfile_create(call_frame_t *frame, fop_mknod_cbk_t linkfile_cbk,
+ xlator_t *this, xlator_t *tovol, xlator_t *fromvol,
+ loc_t *loc);
+int
+dht_lookup_everywhere(call_frame_t *frame, xlator_t *this, loc_t *loc);
int
-dht_selfheal_restore (call_frame_t *frame, dht_selfheal_dir_cbk_t cbk,
- loc_t *loc, dht_layout_t *layout);
+dht_selfheal_directory(call_frame_t *frame, dht_selfheal_dir_cbk_t cbk,
+ loc_t *loc, dht_layout_t *layout);
int
-dht_layout_sort_volname (dht_layout_t *layout);
+dht_selfheal_new_directory(call_frame_t *frame, dht_selfheal_dir_cbk_t cbk,
+ dht_layout_t *layout);
+int
+dht_selfheal_restore(call_frame_t *frame, dht_selfheal_dir_cbk_t cbk,
+ loc_t *loc, dht_layout_t *layout);
+void
+dht_layout_sort_volname(dht_layout_t *layout);
-int dht_get_du_info (call_frame_t *frame, xlator_t *this, loc_t *loc);
+int
+dht_get_du_info(call_frame_t *frame, xlator_t *this, loc_t *loc);
-gf_boolean_t dht_is_subvol_filled (xlator_t *this, xlator_t *subvol);
-xlator_t *dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol,
- dht_local_t *layout);
-int dht_get_du_info_for_subvol (xlator_t *this, int subvol_idx);
+gf_boolean_t
+dht_is_subvol_filled(xlator_t *this, xlator_t *subvol);
+xlator_t *
+dht_free_disk_available_subvol(xlator_t *this, xlator_t *subvol,
+ dht_local_t *layout);
+int
+dht_get_du_info_for_subvol(xlator_t *this, int subvol_idx);
-int dht_layout_preset (xlator_t *this, xlator_t *subvol, inode_t *inode);
-int dht_layout_index_for_subvol (dht_layout_t *layout, xlator_t *subvol);
-int dht_layout_set (xlator_t *this, inode_t *inode, dht_layout_t *layout);;
-void dht_layout_unref (xlator_t *this, dht_layout_t *layout);
-dht_layout_t *dht_layout_ref (xlator_t *this, dht_layout_t *layout);
-xlator_t *dht_first_up_subvol (xlator_t *this);
-xlator_t *dht_last_up_subvol (xlator_t *this);
+int
+dht_layout_preset(xlator_t *this, xlator_t *subvol, inode_t *inode);
+int
+dht_layout_set(xlator_t *this, inode_t *inode, dht_layout_t *layout);
+;
+void
+dht_layout_unref(xlator_t *this, dht_layout_t *layout);
+dht_layout_t *
+dht_layout_ref(xlator_t *this, dht_layout_t *layout);
+int
+dht_layout_index_for_subvol(dht_layout_t *layout, xlator_t *subvol);
+xlator_t *
+dht_first_up_subvol(xlator_t *this);
+xlator_t *
+dht_last_up_subvol(xlator_t *this);
-int dht_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name);
+int
+dht_build_child_loc(xlator_t *this, loc_t *child, loc_t *parent, char *name);
-int dht_filter_loc_subvol_key (xlator_t *this, loc_t *loc, loc_t *new_loc,
- xlator_t **subvol);
+int
+dht_filter_loc_subvol_key(xlator_t *this, loc_t *loc, loc_t *new_loc,
+ xlator_t **subvol);
-int dht_rename_cleanup (call_frame_t *frame);
-int dht_rename_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata);
+int
+dht_rename_cleanup(call_frame_t *frame);
+int
+dht_rename_link_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata);
-int dht_update_commit_hash_for_layout (call_frame_t *frame);
-int dht_fix_directory_layout (call_frame_t *frame,
- dht_selfheal_dir_cbk_t dir_cbk,
- dht_layout_t *layout);
+int
+dht_update_commit_hash_for_layout(call_frame_t *frame);
+int
+dht_fix_directory_layout(call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
+ dht_layout_t *layout);
-int dht_init_subvolumes (xlator_t *this, dht_conf_t *conf);
+int
+dht_init_subvolumes(xlator_t *this, dht_conf_t *conf);
/* migration/rebalance */
-int dht_start_rebalance_task (xlator_t *this, call_frame_t *frame);
+int
+dht_start_rebalance_task(xlator_t *this, call_frame_t *frame);
-int dht_rebalance_in_progress_check (xlator_t *this, call_frame_t *frame);
-int dht_rebalance_complete_check (xlator_t *this, call_frame_t *frame);
+int
+dht_rebalance_in_progress_check(xlator_t *this, call_frame_t *frame);
+int
+dht_rebalance_complete_check(xlator_t *this, call_frame_t *frame);
int
-dht_init_local_subvolumes (xlator_t *this, dht_conf_t *conf);
+dht_init_local_subvolumes(xlator_t *this, dht_conf_t *conf);
/* FOPS */
-int32_t dht_lookup (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- dict_t *xattr_req);
-
-int32_t dht_stat (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc, dict_t *xdata);
-
-int32_t dht_fstat (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd, dict_t *xdata);
-
-int32_t dht_truncate (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- off_t offset, dict_t *xdata);
-
-int32_t dht_ftruncate (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- off_t offset, dict_t *xdata);
-
-int32_t dht_access (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- int32_t mask, dict_t *xdata);
-
-int32_t dht_readlink (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- size_t size, dict_t *xdata);
-
-int32_t dht_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
- mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata);
-
-int32_t dht_mkdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata);
-
-int32_t dht_unlink (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc, int xflag, dict_t *xdata);
-
-int32_t dht_rmdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int flags, dict_t *xdata);
-
-int32_t dht_symlink (call_frame_t *frame, xlator_t *this,
- const char *linkpath, loc_t *loc, mode_t umask,
- dict_t *xdata);
-
-int32_t dht_rename (call_frame_t *frame,
- xlator_t *this,
- loc_t *oldloc,
- loc_t *newloc, dict_t *xdata);
-
-int32_t dht_link (call_frame_t *frame,
- xlator_t *this,
- loc_t *oldloc,
- loc_t *newloc, dict_t *xdata);
-
-int32_t dht_create (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, mode_t mode,
- mode_t umask, fd_t *fd, dict_t *params);
-
-int32_t dht_open (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- int32_t flags, fd_t *fd, dict_t *xdata);
-
-int32_t dht_readv (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- size_t size,
- off_t offset, uint32_t flags, dict_t *xdata);
-
-int32_t dht_writev (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- struct iovec *vector,
- int32_t count,
- off_t offset,
- uint32_t flags,
- struct iobref *iobref, dict_t *xdata);
-
-int32_t dht_flush (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd, dict_t *xdata);
-
-int32_t dht_fsync (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- int32_t datasync, dict_t *xdata);
-
-int32_t dht_opendir (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc, fd_t *fd, dict_t *xdata);
-
-int32_t dht_fsyncdir (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- int32_t datasync, dict_t *xdata);
-
-int32_t dht_statfs (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc, dict_t *xdata);
-
-int32_t dht_setxattr (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- dict_t *dict,
- int32_t flags, dict_t *xdata);
-
-int32_t dht_getxattr (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- const char *name, dict_t *xdata);
-
-int32_t dht_fsetxattr (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- dict_t *dict,
- int32_t flags, dict_t *xdata);
-
-int32_t dht_fgetxattr (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- const char *name, dict_t *xdata);
-
-int32_t dht_removexattr (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- const char *name, dict_t *xdata);
-int32_t dht_fremovexattr (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- const char *name, dict_t *xdata);
-
-int32_t dht_lk (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- int32_t cmd,
- struct gf_flock *flock, dict_t *xdata);
-
-int32_t dht_lease (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- struct gf_lease *lease, dict_t *xdata);
-
-int32_t dht_inodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, int32_t cmd,
- struct gf_flock *flock, dict_t *xdata);
-
-int32_t dht_finodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, fd_t *fd, int32_t cmd,
- struct gf_flock *flock, dict_t *xdata);
-
-int32_t dht_entrylk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, const char *basename,
- entrylk_cmd cmd, entrylk_type type, dict_t *xdata);
-
-int32_t dht_fentrylk (call_frame_t *frame, xlator_t *this,
- const char *volume, fd_t *fd, const char *basename,
- entrylk_cmd cmd, entrylk_type type, dict_t *xdata);
-
-int32_t dht_readdir (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- size_t size, off_t off, dict_t *xdata);
-
-int32_t dht_readdirp (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- size_t size, off_t off, dict_t *dict);
-
-int32_t dht_xattrop (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- gf_xattrop_flags_t flags,
- dict_t *dict, dict_t *xdata);
-
-int32_t dht_fxattrop (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- gf_xattrop_flags_t flags,
- dict_t *dict, dict_t *xdata);
-
-int32_t dht_forget (xlator_t *this, inode_t *inode);
-int32_t dht_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata);
-int32_t dht_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iatt *stbuf, int32_t valid, dict_t *xdata);
-int32_t dht_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd,
- int32_t mode, off_t offset, size_t len, dict_t *xdata);
-int32_t dht_discard(call_frame_t *frame, xlator_t *this, fd_t *fd,
- off_t offset, size_t len, dict_t *xdata);
-int32_t dht_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd,
- off_t offset, off_t len, dict_t *xdata);
-int32_t dht_ipc (call_frame_t *frame, xlator_t *this, int32_t op,
- dict_t *xdata);
+int32_t
+dht_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr_req);
+
+int32_t
+dht_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata);
+
+int32_t
+dht_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata);
+
+int32_t
+dht_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata);
+
+int32_t
+dht_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata);
+
+int32_t
+dht_access(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+ dict_t *xdata);
+
+int32_t
+dht_readlink(call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+ dict_t *xdata);
+
+int32_t
+dht_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *xdata);
+
+int32_t
+dht_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata);
+
+int32_t
+dht_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata);
+
+int32_t
+dht_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata);
+
+int32_t
+dht_symlink(call_frame_t *frame, xlator_t *this, const char *linkpath,
+ loc_t *loc, mode_t umask, dict_t *xdata);
+
+int32_t
+dht_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata);
+
+int32_t
+dht_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata);
+
+int32_t
+dht_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *params);
+
+int32_t
+dht_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata);
+
+int32_t
+dht_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata);
+
+int32_t
+dht_writev(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
+ int32_t count, off_t offset, uint32_t flags, struct iobref *iobref,
+ dict_t *xdata);
+
+int32_t
+dht_flush(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata);
+
+int32_t
+dht_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+ dict_t *xdata);
+
+int32_t
+dht_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata);
+
+int32_t
+dht_fsyncdir(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+ dict_t *xdata);
+
+int32_t
+dht_statfs(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata);
+
+int32_t
+dht_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata);
+
+int32_t
+dht_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, const char *name,
+ dict_t *xdata);
+
+int32_t
+dht_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata);
+
+int32_t
+dht_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
+ dict_t *xdata);
+
+int32_t
+dht_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata);
+int32_t
+dht_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata);
+
+int32_t
+dht_lk(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
+ struct gf_flock *flock, dict_t *xdata);
+
+int32_t
+dht_lease(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct gf_lease *lease, dict_t *xdata);
+
+int32_t
+dht_inodelk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ int32_t cmd, struct gf_flock *flock, dict_t *xdata);
+
+int32_t
+dht_finodelk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ int32_t cmd, struct gf_flock *flock, dict_t *xdata);
+
+int32_t
+dht_entrylk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata);
+
+int32_t
+dht_fentrylk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata);
+
+int32_t
+dht_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *xdata);
+
+int32_t
+dht_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *dict);
+
+int32_t
+dht_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata);
+
+int32_t
+dht_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata);
+
+int32_t
+dht_forget(xlator_t *this, inode_t *inode);
+int32_t
+dht_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *stbuf,
+ int32_t valid, dict_t *xdata);
+int32_t
+dht_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
+ int32_t valid, dict_t *xdata);
+int32_t
+dht_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+ off_t offset, size_t len, dict_t *xdata);
+int32_t
+dht_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata);
+int32_t
+dht_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ off_t len, dict_t *xdata);
+int32_t
+dht_ipc(call_frame_t *frame, xlator_t *this, int32_t op, dict_t *xdata);
int
dht_set_subvol_range(xlator_t *this);
-int32_t dht_init (xlator_t *this);
-void dht_fini (xlator_t *this);
-int dht_reconfigure (xlator_t *this, dict_t *options);
-int32_t dht_notify (xlator_t *this, int32_t event, void *data, ...);
+int32_t
+dht_init(xlator_t *this);
+void
+dht_fini(xlator_t *this);
+int
+dht_reconfigure(xlator_t *this, dict_t *options);
+int32_t
+dht_notify(xlator_t *this, int32_t event, void *data, ...);
/* definitions for nufa/switch */
-int dht_revalidate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode,
- struct iatt *stbuf, dict_t *xattr,
- struct iatt *postparent);
-int dht_lookup_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+int
+dht_revalidate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode, struct iatt *stbuf,
+ dict_t *xattr, struct iatt *postparent);
+int
+dht_lookup_dir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode, struct iatt *stbuf,
+ dict_t *xattr, struct iatt *postparent);
+int
+dht_lookup_linkfile_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int op_ret, int op_errno, inode_t *inode,
struct iatt *stbuf, dict_t *xattr,
struct iatt *postparent);
-int dht_lookup_linkfile_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, dict_t *xattr,
- struct iatt *postparent);
-int dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, dict_t *xattr,
- struct iatt *postparent);
-int dht_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- fd_t *fd, inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata);
-int dht_newfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata);
-
-int
-gf_defrag_status_get (dht_conf_t *conf, dict_t *dict);
-
-void
-gf_defrag_set_pause_state (gf_tier_conf_t *tier_conf, tier_pause_state_t state);
-
-tier_pause_state_t
-gf_defrag_get_pause_state (gf_tier_conf_t *tier_conf);
-
int
-gf_defrag_pause_tier (xlator_t *this, gf_defrag_info_t *defrag);
+dht_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+ struct iatt *postparent);
+int
+dht_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, fd_t *fd, inode_t *inode, struct iatt *stbuf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata);
+int
+dht_newfile_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, inode_t *inode, struct iatt *stbuf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata);
-tier_pause_state_t
-gf_defrag_check_pause_tier (gf_tier_conf_t *defrag);
+int
+dht_finodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata);
int
-gf_defrag_resume_tier (xlator_t *this, gf_defrag_info_t *defrag);
+dht_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, dict_t *xattr, dict_t *xdata);
int
-gf_defrag_start_detach_tier (gf_defrag_info_t *defrag);
+dht_common_xattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata);
+int
+gf_defrag_status_get(dht_conf_t *conf, dict_t *dict);
int
-gf_defrag_stop (dht_conf_t *conf, gf_defrag_status_t status,
- dict_t *output);
+gf_defrag_stop(dht_conf_t *conf, gf_defrag_status_t status, dict_t *output);
-void*
-gf_defrag_start (void *this);
+void *
+gf_defrag_start(void *this);
int32_t
-gf_defrag_handle_hardlink (xlator_t *this, loc_t *loc, int *fop_errno);
+gf_defrag_handle_hardlink(xlator_t *this, loc_t *loc, int *fop_errno);
int
-dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
+dht_migrate_file(xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
int flag, int *fop_errno);
int
-dht_inode_ctx_layout_get (inode_t *inode, xlator_t *this,
- dht_layout_t **layout_int);
+dht_inode_ctx_layout_get(inode_t *inode, xlator_t *this,
+ dht_layout_t **layout_int);
int
-dht_inode_ctx_layout_set (inode_t *inode, xlator_t *this,
- dht_layout_t* layout_int);
+dht_inode_ctx_layout_set(inode_t *inode, xlator_t *this,
+ dht_layout_t *layout_int);
int
-dht_inode_ctx_time_update (inode_t *inode, xlator_t *this, struct iatt *stat,
- int32_t update_ctx);
-void dht_inode_ctx_time_set (inode_t *inode, xlator_t *this, struct iatt *stat);
+dht_inode_ctx_time_update(inode_t *inode, xlator_t *this, struct iatt *stat,
+ int32_t update_ctx);
+void
+dht_inode_ctx_time_set(inode_t *inode, xlator_t *this, struct iatt *stat);
-int dht_inode_ctx_get (inode_t *inode, xlator_t *this, dht_inode_ctx_t **ctx);
-int dht_inode_ctx_set (inode_t *inode, xlator_t *this, dht_inode_ctx_t *ctx);
int
-dht_dir_attr_heal (void *data);
+dht_inode_ctx_get(inode_t *inode, xlator_t *this, dht_inode_ctx_t **ctx);
int
-dht_dir_attr_heal_done (int ret, call_frame_t *sync_frame, void *data);
+dht_inode_ctx_set(inode_t *inode, xlator_t *this, dht_inode_ctx_t *ctx);
int
-dht_dir_has_layout (dict_t *xattr, char *name);
-gf_boolean_t
-dht_is_subvol_in_layout (dht_layout_t *layout, xlator_t *xlator);
+dht_dir_attr_heal(void *data);
+int
+dht_dir_attr_heal_done(int ret, call_frame_t *sync_frame, void *data);
xlator_t *
-dht_subvol_with_free_space_inodes (xlator_t *this, xlator_t *subvol, xlator_t *ignore,
- dht_layout_t *layout, uint64_t filesize);
+dht_subvol_with_free_space_inodes(xlator_t *this, xlator_t *subvol,
+ xlator_t *ignore, dht_layout_t *layout,
+ uint64_t filesize);
xlator_t *
-dht_subvol_maxspace_nonzeroinode (xlator_t *this, xlator_t *subvol,
- dht_layout_t *layout);
+dht_subvol_maxspace_nonzeroinode(xlator_t *this, xlator_t *subvol,
+ dht_layout_t *layout);
int
-dht_linkfile_attr_heal (call_frame_t *frame, xlator_t *this);
+dht_dir_has_layout(dict_t *xattr, char *name);
+int
+dht_linkfile_attr_heal(call_frame_t *frame, xlator_t *this);
-void
-dht_layout_dump (dht_layout_t *layout, const char *prefix);
int32_t
-dht_priv_dump (xlator_t *this);
+dht_priv_dump(xlator_t *this);
int32_t
-dht_inodectx_dump (xlator_t *this, inode_t *inode);
+dht_inodectx_dump(xlator_t *this, inode_t *inode);
-int
-dht_inode_ctx_get_mig_info (xlator_t *this, inode_t *inode,
- xlator_t **src_subvol, xlator_t **dst_subvol);
gf_boolean_t
-dht_mig_info_is_invalid (xlator_t *current, xlator_t *src_subvol,
- xlator_t *dst_subvol);
+dht_is_subvol_in_layout(dht_layout_t *layout, xlator_t *xlator);
int
-dht_subvol_status (dht_conf_t *conf, xlator_t *subvol);
+dht_inode_ctx_get_mig_info(xlator_t *this, inode_t *inode,
+ xlator_t **src_subvol, xlator_t **dst_subvol);
+gf_boolean_t
+dht_mig_info_is_invalid(xlator_t *current, xlator_t *src_subvol,
+ xlator_t *dst_subvol);
-void
-dht_log_new_layout_for_dir_selfheal (xlator_t *this, loc_t *loc,
- dht_layout_t *layout);
int
-dht_lookup_everywhere_done (call_frame_t *frame, xlator_t *this);
+dht_subvol_status(dht_conf_t *conf, xlator_t *subvol);
-int
-dht_fill_dict_to_avoid_unlink_of_migrating_file (dict_t *dict);
+void
+dht_log_new_layout_for_dir_selfheal(xlator_t *this, loc_t *loc,
+ dht_layout_t *layout);
int
-dht_layout_sort (dht_layout_t *layout);
+dht_layout_sort(dht_layout_t *layout);
int
-dht_heal_full_path (void *data);
+dht_heal_full_path(void *data);
int
-dht_heal_full_path_done (int op_ret, call_frame_t *frame, void *data);
+dht_heal_full_path_done(int op_ret, call_frame_t *frame, void *data);
int
-dht_layout_missing_dirs (dht_layout_t *layout);
+dht_layout_missing_dirs(dht_layout_t *layout);
int
-dht_refresh_layout (call_frame_t *frame);
-
-gf_boolean_t
-dht_is_tier_xlator (xlator_t *this);
+dht_refresh_layout(call_frame_t *frame);
int
-dht_build_parent_loc (xlator_t *this, loc_t *parent, loc_t *child,
- int32_t *op_errno);
+dht_build_parent_loc(xlator_t *this, loc_t *parent, loc_t *child,
+ int32_t *op_errno);
int32_t
-dht_set_local_rebalance (xlator_t *this, dht_local_t *local,
- struct iatt *stbuf,
- struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata);
+dht_set_local_rebalance(xlator_t *this, dht_local_t *local, struct iatt *stbuf,
+ struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata);
void
-dht_build_root_loc (inode_t *inode, loc_t *loc);
+dht_build_root_loc(inode_t *inode, loc_t *loc);
gf_boolean_t
-dht_fd_open_on_dst (xlator_t *this, fd_t *fd, xlator_t *dst);
+dht_fd_open_on_dst(xlator_t *this, fd_t *fd, xlator_t *dst);
int32_t
-dht_fd_ctx_destroy (xlator_t *this, fd_t *fd);
+dht_fd_ctx_destroy(xlator_t *this, fd_t *fd);
int32_t
-dht_release (xlator_t *this, fd_t *fd);
-
+dht_release(xlator_t *this, fd_t *fd);
int32_t
-dht_set_fixed_dir_stat (struct iatt *stat);
+dht_set_fixed_dir_stat(struct iatt *stat);
-xlator_t*
-dht_get_lock_subvolume (xlator_t *this, struct gf_flock *lock,
- dht_local_t *local);
+xlator_t *
+dht_get_lock_subvolume(xlator_t *this, struct gf_flock *lock,
+ dht_local_t *local);
int
-dht_lk_inode_unref (call_frame_t *frame, int32_t op_ret);
+dht_lk_inode_unref(call_frame_t *frame, int32_t op_ret);
-void
-dht_normalize_stats (struct statvfs *buf, unsigned long bsize,
- unsigned long frsize);
+int
+dht_fd_ctx_set(xlator_t *this, fd_t *fd, xlator_t *subvol);
int
-add_opt(char **optsp, const char *opt);
+dht_check_and_open_fd_on_subvol(xlator_t *this, call_frame_t *frame);
+
+/* FD fop callbacks */
int
-dht_aggregate_split_brain_xattr (dict_t *dst, char *key, data_t *value);
+dht_writev_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata);
int
-dht_remove_stale_linkto (void *data);
+dht_flush_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, dict_t *xdata);
int
-dht_remove_stale_linkto_cbk (int ret, call_frame_t *sync_frame, void *data);
+dht_file_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata);
int
-dht_fd_ctx_set (xlator_t *this, fd_t *fd, xlator_t *subvol);
+dht_zerofill_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata);
int
-dht_check_and_open_fd_on_subvol (xlator_t *this, call_frame_t *frame);
+dht_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata);
-/* FD fop callbacks */
+int
+dht_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata);
int
-dht_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata);
+dht_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata);
int
-dht_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata);
+dht_fsync_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata);
int
-dht_file_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata);
+dht_readv_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iovec *vector, int count, struct iatt *stbuf,
+ struct iobref *iobref, dict_t *xdata);
int
-dht_zerofill_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata);
+dht_file_attr_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *stbuf, dict_t *xdata);
int
-dht_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata);
+dht_file_removexattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata);
int
-dht_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata);
+dht_file_setxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata);
+/* All custom xattr heal functions */
int
-dht_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata);
+dht_dir_heal_xattrs(void *data);
int
-dht_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
- int op_errno, struct iatt *prebuf, struct iatt *postbuf,
- dict_t *xdata);
+dht_dir_heal_xattrs_done(int ret, call_frame_t *sync_frame, void *data);
+int32_t
+dht_dict_set_array(dict_t *dict, char *key, int32_t value[], int32_t size);
int
-dht_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- struct iovec *vector, int count, struct iatt *stbuf,
- struct iobref *iobref, dict_t *xdata);
+dht_set_user_xattr(dict_t *dict, char *k, data_t *v, void *data);
-int
-dht_file_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *stbuf, dict_t *xdata);
+void
+dht_dir_set_heal_xattr(xlator_t *this, dht_local_t *local, dict_t *dst,
+ dict_t *src, int *uret, int *uflag);
int
-dht_file_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata);
+dht_dir_xattr_heal(xlator_t *this, dht_local_t *local, int *op_errno);
int
-dht_file_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata);
+dht_common_mark_mdsxattr(call_frame_t *frame, int *errst, int flag);
-/* All custom xattr heal functions */
int
-dht_dir_heal_xattrs (void *data);
+dht_inode_ctx_mdsvol_get(inode_t *inode, xlator_t *this, xlator_t **mdsvol);
int
-dht_dir_heal_xattrs_done (int ret, call_frame_t *sync_frame, void *data);
+dht_selfheal_dir_setattr(call_frame_t *frame, loc_t *loc, struct iatt *stbuf,
+ int32_t valid, dht_layout_t *layout);
+
+/* Abstract out the DHT-IATT-IN-DICT */
void
-dht_aggregate_xattr (dict_t *dst, dict_t *src);
+dht_selfheal_layout_new_directory(call_frame_t *frame, loc_t *loc,
+ dht_layout_t *new_layout);
-int32_t
-dht_dict_set_array(dict_t *dict, char *key, int32_t value[], int32_t size);
+int
+dht_pt_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *key,
+ dict_t *xdata);
int
-dht_set_user_xattr (dict_t *dict, char *k, data_t *v, void *data);
+dht_pt_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *key, dict_t *xdata);
-void
-dht_dir_set_heal_xattr (xlator_t *this, dht_local_t *local, dict_t *dst,
- dict_t *src, int *uret, int *uflag);
+int32_t
+dht_pt_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata);
int
-dht_dir_xattr_heal (xlator_t *this, dht_local_t *local);
+dht_pt_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata);
int32_t
-dht_dict_get_array (dict_t *dict, char *key, int32_t value[], int32_t size, int *errst);
-
-xlator_t *
-dht_inode_get_hashed_subvol (inode_t *inode, xlator_t *this, loc_t *loc);
+dht_check_remote_fd_failed_error(dht_local_t *local, int op_ret, int op_errno);
int
-dht_common_mark_mdsxattr (call_frame_t *frame, int *errst, int flag);
+dht_common_xattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata);
-int
-dht_common_mark_mdsxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- dict_t *xdata);
+int32_t
+dht_create_lock(call_frame_t *frame, xlator_t *subvol);
int
-dht_inode_ctx_mdsvol_set (inode_t *inode, xlator_t *this,
- xlator_t *mds_subvol);
-int
-dht_inode_ctx_mdsvol_get (inode_t *inode, xlator_t *this,
- xlator_t **mdsvol);
+dht_set_parent_layout_in_dict(loc_t *loc, xlator_t *this, dht_local_t *local);
int
-dht_selfheal_dir_setattr (call_frame_t *frame, loc_t *loc, struct iatt *stbuf,
- int32_t valid, dht_layout_t *layout);
-
-/* Abstract out the DHT-IATT-IN-DICT */
+dht_dir_layout_error_check(xlator_t *this, inode_t *inode);
-
-int dht_request_iatt_in_xdata (xlator_t *this, dict_t *xattr_req);
-
-int dht_read_iatt_from_xdata (xlator_t *this, dict_t *xdata,
- struct iatt *stbuf);
-
-#endif/* _DHT_H */
+int
+dht_inode_ctx_mdsvol_set(inode_t *inode, xlator_t *this, xlator_t *mds_subvol);
+#endif /* _DHT_H */
diff --git a/xlators/cluster/dht/src/dht-diskusage.c b/xlators/cluster/dht/src/dht-diskusage.c
index c2b9b2e37c4..c0588828fdb 100644
--- a/xlators/cluster/dht/src/dht-diskusage.c
+++ b/xlators/cluster/dht/src/dht-diskusage.c
@@ -8,505 +8,480 @@
cases as published by the Free Software Foundation.
*/
-
/* TODO: add NS locking */
-#include "glusterfs.h"
-#include "xlator.h"
#include "dht-common.h"
-#include "dht-messages.h"
-#include "defaults.h"
#include <sys/time.h>
-#include "events.h"
-
+#include <glusterfs/events.h>
int
-dht_du_info_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct statvfs *statvfs,
- dict_t *xdata)
+dht_du_info_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct statvfs *statvfs, dict_t *xdata)
{
- dht_conf_t *conf = NULL;
- xlator_t *prev = NULL;
- int this_call_cnt = 0;
- int i = 0;
- double percent = 0;
- double percent_inodes = 0;
- uint64_t bytes = 0;
- uint32_t bpc; /* blocks per chunk */
- uint32_t chunks = 0;
-
- conf = this->private;
- prev = cookie;
-
- if (op_ret == -1 || !statvfs) {
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_GET_DISK_INFO_ERROR,
- "failed to get disk info from %s", prev->name);
- goto out;
- }
-
- if (statvfs->f_blocks) {
- percent = (statvfs->f_bavail * 100) / statvfs->f_blocks;
- bytes = (statvfs->f_bavail * statvfs->f_frsize);
- /*
- * A 32-bit count of 1MB chunks allows a maximum brick size of
- * ~4PB. It's possible that we could see a single local FS
- * bigger than that some day, but this code is likely to be
- * irrelevant by then. Meanwhile, it's more important to keep
- * the chunk size small so the layout-calculation code that
- * uses this value can be tested on normal machines.
- */
- bpc = (1 << 20) / statvfs->f_bsize;
- chunks = (statvfs->f_blocks + bpc - 1) / bpc;
- }
-
- if (statvfs->f_files) {
- percent_inodes = (statvfs->f_ffree * 100) / statvfs->f_files;
- } else {
- /*
- * Set percent inodes to 100 for dynamically allocated inode
- * filesystems. The rationale is that distribute need not
- * worry about total inodes; rather, let the 'create()' be
- * scheduled on the hashed subvol regardless of the total
- * inodes.
- */
- percent_inodes = 100;
- }
-
- LOCK (&conf->subvolume_lock);
- {
- for (i = 0; i < conf->subvolume_cnt; i++)
- if (prev == conf->subvolumes[i]) {
- conf->du_stats[i].avail_percent = percent;
- conf->du_stats[i].avail_space = bytes;
- conf->du_stats[i].avail_inodes = percent_inodes;
- conf->du_stats[i].chunks = chunks;
- conf->du_stats[i].total_blocks = statvfs->f_blocks;
- conf->du_stats[i].avail_blocks = statvfs->f_bavail;
- conf->du_stats[i].frsize = statvfs->f_frsize;
-
- gf_msg_debug (this->name, 0,
- "subvolume '%s': avail_percent "
- "is: %.2f and avail_space "
- "is: %" PRIu64" and avail_inodes"
- " is: %.2f",
- prev->name,
- conf->du_stats[i].avail_percent,
- conf->du_stats[i].avail_space,
- conf->du_stats[i].avail_inodes);
- break; /* no point in looping further */
- }
- }
- UNLOCK (&conf->subvolume_lock);
+ dht_conf_t *conf = NULL;
+ xlator_t *prev = NULL;
+ int this_call_cnt = 0;
+ int i = 0;
+ double percent = 0;
+ double percent_inodes = 0;
+ uint64_t bytes = 0;
+ uint32_t bpc; /* blocks per chunk */
+ uint32_t chunks = 0;
+
+ conf = this->private;
+ prev = cookie;
+
+ if (op_ret == -1 || !statvfs) {
+ gf_msg(this->name, GF_LOG_WARNING, op_errno,
+ DHT_MSG_GET_DISK_INFO_ERROR, "failed to get disk info from %s",
+ prev->name);
+ goto out;
+ }
+
+ if (statvfs->f_blocks) {
+ percent = (statvfs->f_bavail * 100) / statvfs->f_blocks;
+ bytes = (statvfs->f_bavail * statvfs->f_frsize);
+ /*
+ * A 32-bit count of 1MB chunks allows a maximum brick size of
+ * ~4PB. It's possible that we could see a single local FS
+ * bigger than that some day, but this code is likely to be
+ * irrelevant by then. Meanwhile, it's more important to keep
+ * the chunk size small so the layout-calculation code that
+ * uses this value can be tested on normal machines.
+ */
+ bpc = (1 << 20) / statvfs->f_bsize;
+ chunks = (statvfs->f_blocks + bpc - 1) / bpc;
+ }
+
+ if (statvfs->f_files) {
+ percent_inodes = (statvfs->f_ffree * 100) / statvfs->f_files;
+ } else {
+ /*
+ * Set percent inodes to 100 for dynamically allocated inode
+ * filesystems. The rationale is that distribute need not
+ * worry about total inodes; rather, let the 'create()' be
+ * scheduled on the hashed subvol regardless of the total
+ * inodes.
+ */
+ percent_inodes = 100;
+ }
+
+ LOCK(&conf->subvolume_lock);
+ {
+ for (i = 0; i < conf->subvolume_cnt; i++)
+ if (prev == conf->subvolumes[i]) {
+ conf->du_stats[i].avail_percent = percent;
+ conf->du_stats[i].avail_space = bytes;
+ conf->du_stats[i].avail_inodes = percent_inodes;
+ conf->du_stats[i].chunks = chunks;
+ conf->du_stats[i].total_blocks = statvfs->f_blocks;
+ conf->du_stats[i].avail_blocks = statvfs->f_bavail;
+ conf->du_stats[i].frsize = statvfs->f_frsize;
+
+ gf_msg_debug(this->name, 0,
+ "subvolume '%s': avail_percent "
+ "is: %.2f and avail_space "
+ "is: %" PRIu64
+ " and avail_inodes"
+ " is: %.2f",
+ prev->name, conf->du_stats[i].avail_percent,
+ conf->du_stats[i].avail_space,
+ conf->du_stats[i].avail_inodes);
+ break; /* no point in looping further */
+ }
+ }
+ UNLOCK(&conf->subvolume_lock);
out:
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt))
- DHT_STACK_DESTROY (frame);
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt))
+ DHT_STACK_DESTROY(frame);
- return 0;
+ return 0;
}
int
-dht_get_du_info_for_subvol (xlator_t *this, int subvol_idx)
+dht_get_du_info_for_subvol(xlator_t *this, int subvol_idx)
{
- dht_conf_t *conf = NULL;
- call_frame_t *statfs_frame = NULL;
- dht_local_t *statfs_local = NULL;
- call_pool_t *pool = NULL;
- loc_t tmp_loc = {0,};
-
- conf = this->private;
- pool = this->ctx->pool;
-
- statfs_frame = create_frame (this, pool);
- if (!statfs_frame) {
- goto err;
- }
-
- /* local->fop value is not used in this case */
- statfs_local = dht_local_init (statfs_frame, NULL, NULL,
- GF_FOP_MAXVALUE);
- if (!statfs_local) {
- goto err;
- }
-
- /* make it root gfid, should be enough to get the proper info back */
- tmp_loc.gfid[15] = 1;
-
- statfs_local->call_cnt = 1;
- STACK_WIND_COOKIE (statfs_frame, dht_du_info_cbk,
- conf->subvolumes[subvol_idx],
- conf->subvolumes[subvol_idx],
- conf->subvolumes[subvol_idx]->fops->statfs,
- &tmp_loc, NULL);
-
- return 0;
+ dht_conf_t *conf = NULL;
+ call_frame_t *statfs_frame = NULL;
+ dht_local_t *statfs_local = NULL;
+ call_pool_t *pool = NULL;
+ loc_t tmp_loc = {
+ 0,
+ };
+
+ conf = this->private;
+ pool = this->ctx->pool;
+
+ statfs_frame = create_frame(this, pool);
+ if (!statfs_frame) {
+ goto err;
+ }
+
+ /* local->fop value is not used in this case */
+ statfs_local = dht_local_init(statfs_frame, NULL, NULL, GF_FOP_MAXVALUE);
+ if (!statfs_local) {
+ goto err;
+ }
+
+ /* make it root gfid, should be enough to get the proper info back */
+ tmp_loc.gfid[15] = 1;
+
+ statfs_local->call_cnt = 1;
+ STACK_WIND_COOKIE(
+ statfs_frame, dht_du_info_cbk, conf->subvolumes[subvol_idx],
+ conf->subvolumes[subvol_idx],
+ conf->subvolumes[subvol_idx]->fops->statfs, &tmp_loc, NULL);
+
+ return 0;
err:
- if (statfs_frame)
- DHT_STACK_DESTROY (statfs_frame);
+ if (statfs_frame)
+ DHT_STACK_DESTROY(statfs_frame);
- return -1;
+ return -1;
}
int
-dht_get_du_info (call_frame_t *frame, xlator_t *this, loc_t *loc)
+dht_get_du_info(call_frame_t *frame, xlator_t *this, loc_t *loc)
{
- int i = 0;
- int ret = -1;
- dht_conf_t *conf = NULL;
- call_frame_t *statfs_frame = NULL;
- dht_local_t *statfs_local = NULL;
- struct timeval tv = {0,};
- loc_t tmp_loc = {0,};
-
- conf = this->private;
-
- gettimeofday (&tv, NULL);
-
- /* make it root gfid, should be enough to get the proper
- info back */
- tmp_loc.gfid[15] = 1;
-
- if (tv.tv_sec > (conf->refresh_interval
- + conf->last_stat_fetch.tv_sec)) {
-
- statfs_frame = copy_frame (frame);
- if (!statfs_frame) {
- goto err;
- }
-
- /* In this case, 'local->fop' is not used */
- statfs_local = dht_local_init (statfs_frame, loc, NULL,
- GF_FOP_MAXVALUE);
- if (!statfs_local) {
- goto err;
- }
-
- statfs_local->params = dict_new ();
- if (!statfs_local->params)
- goto err;
-
- ret = dict_set_int8 (statfs_local->params,
- GF_INTERNAL_IGNORE_DEEM_STATFS, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set "
- GF_INTERNAL_IGNORE_DEEM_STATFS" in dict");
- goto err;
- }
+ int i = 0;
+ int ret = -1;
+ dht_conf_t *conf = NULL;
+ call_frame_t *statfs_frame = NULL;
+ dht_local_t *statfs_local = NULL;
+ loc_t tmp_loc = {
+ 0,
+ };
+ time_t now;
+
+ conf = this->private;
+ now = gf_time();
+ /* make it root gfid, should be enough to get the proper
+ info back */
+ tmp_loc.gfid[15] = 1;
+
+ if (now > (conf->refresh_interval + conf->last_stat_fetch)) {
+ statfs_frame = copy_frame(frame);
+ if (!statfs_frame) {
+ goto err;
+ }
+
+ /* In this case, 'local->fop' is not used */
+ statfs_local = dht_local_init(statfs_frame, loc, NULL, GF_FOP_MAXVALUE);
+ if (!statfs_local) {
+ goto err;
+ }
+
+ statfs_local->params = dict_new();
+ if (!statfs_local->params)
+ goto err;
+
+ ret = dict_set_int8(statfs_local->params,
+ GF_INTERNAL_IGNORE_DEEM_STATFS, 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set " GF_INTERNAL_IGNORE_DEEM_STATFS " in dict");
+ goto err;
+ }
+
+ statfs_local->call_cnt = conf->subvolume_cnt;
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ STACK_WIND_COOKIE(statfs_frame, dht_du_info_cbk,
+ conf->subvolumes[i], conf->subvolumes[i],
+ conf->subvolumes[i]->fops->statfs, &tmp_loc,
+ statfs_local->params);
+ }
- statfs_local->call_cnt = conf->subvolume_cnt;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- STACK_WIND_COOKIE (statfs_frame, dht_du_info_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->statfs,
- &tmp_loc, statfs_local->params);
- }
-
- conf->last_stat_fetch.tv_sec = tv.tv_sec;
- }
- return 0;
+ conf->last_stat_fetch = now;
+ }
+ return 0;
err:
- if (statfs_frame)
- DHT_STACK_DESTROY (statfs_frame);
+ if (statfs_frame)
+ DHT_STACK_DESTROY(statfs_frame);
- return -1;
+ return -1;
}
-
gf_boolean_t
-dht_is_subvol_filled (xlator_t *this, xlator_t *subvol)
+dht_is_subvol_filled(xlator_t *this, xlator_t *subvol)
{
- int i = 0;
- char vol_name[256];
- dht_conf_t *conf = NULL;
- gf_boolean_t subvol_filled_inodes = _gf_false;
- gf_boolean_t subvol_filled_space = _gf_false;
- gf_boolean_t is_subvol_filled = _gf_false;
- double usage = 0;
-
- conf = this->private;
-
- /* Check for values above specified percent or free disk */
- LOCK (&conf->subvolume_lock);
- {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (subvol == conf->subvolumes[i]) {
- if (conf->disk_unit == 'p') {
- if (conf->du_stats[i].avail_percent <
- conf->min_free_disk) {
- subvol_filled_space = _gf_true;
- break;
- }
-
- } else {
- if (conf->du_stats[i].avail_space <
- conf->min_free_disk) {
- subvol_filled_space = _gf_true;
- break;
- }
- }
- if (conf->du_stats[i].avail_inodes <
- conf->min_free_inodes) {
- subvol_filled_inodes = _gf_true;
- break;
- }
- }
- }
- }
- UNLOCK (&conf->subvolume_lock);
-
- if (subvol_filled_space && conf->subvolume_status[i]) {
- if (!(conf->du_stats[i].log++ % (GF_UNIVERSAL_ANSWER * 10))) {
- usage = 100 - conf->du_stats[i].avail_percent;
-
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_SUBVOL_INSUFF_SPACE,
- "disk space on subvolume '%s' is getting "
- "full (%.2f %%), consider adding more bricks",
- subvol->name, usage);
-
- strncpy(vol_name, this->name, sizeof(vol_name));
- vol_name[(strlen(this->name)-4)] = '\0';
-
- gf_event(EVENT_DHT_DISK_USAGE,
- "volume=%s;subvol=%s;usage=%.2f %%",
- vol_name, subvol->name, usage);
- }
- }
-
- if (subvol_filled_inodes && conf->subvolume_status[i]) {
- if (!(conf->du_stats[i].log++ % (GF_UNIVERSAL_ANSWER * 10))) {
- usage = 100 - conf->du_stats[i].avail_inodes;
- gf_msg (this->name, GF_LOG_CRITICAL, 0,
- DHT_MSG_SUBVOL_INSUFF_INODES,
- "inodes on subvolume '%s' are at "
- "(%.2f %%), consider adding more bricks",
- subvol->name, usage);
-
- strncpy(vol_name, this->name, sizeof(vol_name));
- vol_name[(strlen(this->name)-4)] = '\0';
-
- gf_event(EVENT_DHT_INODES_USAGE,
- "volume=%s;subvol=%s;usage=%.2f %%",
- vol_name, subvol->name, usage);
- }
- }
-
- is_subvol_filled = (subvol_filled_space || subvol_filled_inodes);
-
- return is_subvol_filled;
-}
+ int i = 0;
+ char vol_name[256];
+ dht_conf_t *conf = NULL;
+ gf_boolean_t subvol_filled_inodes = _gf_false;
+ gf_boolean_t subvol_filled_space = _gf_false;
+ gf_boolean_t is_subvol_filled = _gf_false;
+ double usage = 0;
+
+ conf = this->private;
+
+ /* Check for values above specified percent or free disk */
+ LOCK(&conf->subvolume_lock);
+ {
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (subvol == conf->subvolumes[i]) {
+ if (conf->disk_unit == 'p') {
+ if (conf->du_stats[i].avail_percent < conf->min_free_disk) {
+ subvol_filled_space = _gf_true;
+ break;
+ }
+
+ } else {
+ if (conf->du_stats[i].avail_space < conf->min_free_disk) {
+ subvol_filled_space = _gf_true;
+ break;
+ }
+ }
+ if (conf->du_stats[i].avail_inodes < conf->min_free_inodes) {
+ subvol_filled_inodes = _gf_true;
+ break;
+ }
+ }
+ }
+ }
+ UNLOCK(&conf->subvolume_lock);
+ if (subvol_filled_space && conf->subvolume_status[i]) {
+ if (!(conf->du_stats[i].log++ % (GF_UNIVERSAL_ANSWER * 10))) {
+ usage = 100 - conf->du_stats[i].avail_percent;
-/*Get the best subvolume to create the file in*/
-xlator_t *
-dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol,
- dht_local_t *local)
-{
- xlator_t *avail_subvol = NULL;
- dht_conf_t *conf = NULL;
- dht_layout_t *layout = NULL;
- loc_t *loc = NULL;
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_SUBVOL_INSUFF_SPACE,
+ "disk space on subvolume '%s' is getting "
+ "full (%.2f %%), consider adding more bricks",
+ subvol->name, usage);
- conf = this->private;
- if (!local)
- goto out;
- loc = &local->loc;
- if (!local->layout) {
- layout = dht_layout_get (this, loc->parent);
-
- if (!layout) {
- gf_msg_debug (this->name, 0,
- "Missing layout. path=%s,"
- " parent gfid = %s", loc->path,
- uuid_utoa (loc->parent->gfid));
- goto out;
- }
- } else {
- layout = dht_layout_ref (this, local->layout);
- }
+ (void)snprintf(vol_name, sizeof(vol_name), "%s", this->name);
+ vol_name[(strlen(this->name) - 4)] = '\0';
- LOCK (&conf->subvolume_lock);
- {
- avail_subvol = dht_subvol_with_free_space_inodes(this, subvol, NULL,
- layout, 0);
- if(!avail_subvol)
- {
- avail_subvol = dht_subvol_maxspace_nonzeroinode(this,
- subvol,
- layout);
- }
+ gf_event(EVENT_DHT_DISK_USAGE, "volume=%s;subvol=%s;usage=%.2f %%",
+ vol_name, subvol->name, usage);
+ }
+ }
+
+ if (subvol_filled_inodes && conf->subvolume_status[i]) {
+ if (!(conf->du_stats[i].log++ % (GF_UNIVERSAL_ANSWER * 10))) {
+ usage = 100 - conf->du_stats[i].avail_inodes;
+ gf_msg(this->name, GF_LOG_CRITICAL, 0, DHT_MSG_SUBVOL_INSUFF_INODES,
+ "inodes on subvolume '%s' are at "
+ "(%.2f %%), consider adding more bricks",
+ subvol->name, usage);
+
+ (void)snprintf(vol_name, sizeof(vol_name), "%s", this->name);
+ vol_name[(strlen(this->name) - 4)] = '\0';
+
+ gf_event(EVENT_DHT_INODES_USAGE,
+ "volume=%s;subvol=%s;usage=%.2f %%", vol_name,
+ subvol->name, usage);
+ }
+ }
- }
- UNLOCK (&conf->subvolume_lock);
-out:
- if (!avail_subvol) {
- gf_msg_debug (this->name, 0,
- "No subvolume has enough free space \
- and/or inodes to create");
- avail_subvol = subvol;
- }
+ is_subvol_filled = (subvol_filled_space || subvol_filled_inodes);
- if (layout)
- dht_layout_unref (this, layout);
- return avail_subvol;
+ return is_subvol_filled;
}
-static inline
-int32_t dht_subvol_has_err (dht_conf_t *conf, xlator_t *this, xlator_t *ignore,
- dht_layout_t *layout)
+/*Get the best subvolume to create the file in*/
+xlator_t *
+dht_free_disk_available_subvol(xlator_t *this, xlator_t *subvol,
+ dht_local_t *local)
{
- int ret = -1;
- int i = 0;
-
- if (!this || !layout)
- goto out;
-
- /* this check is meant for rebalance process. The source of the file
- * should be ignored for space check */
- if (this == ignore) {
- goto out;
+ xlator_t *avail_subvol = NULL;
+ dht_conf_t *conf = NULL;
+ dht_layout_t *layout = NULL;
+ loc_t *loc = NULL;
+
+ conf = this->private;
+ if (!local)
+ goto out;
+ loc = &local->loc;
+ if (!local->layout) {
+ layout = dht_layout_get(this, loc->parent);
+
+ if (!layout) {
+ gf_msg_debug(this->name, 0,
+ "Missing layout. path=%s,"
+ " parent gfid = %s",
+ loc->path, uuid_utoa(loc->parent->gfid));
+ goto out;
+ }
+ } else {
+ layout = dht_layout_ref(this, local->layout);
+ }
+
+ LOCK(&conf->subvolume_lock);
+ {
+ avail_subvol = dht_subvol_with_free_space_inodes(this, subvol, NULL,
+ layout, 0);
+ if (!avail_subvol) {
+ avail_subvol = dht_subvol_maxspace_nonzeroinode(this, subvol,
+ layout);
}
+ }
+ UNLOCK(&conf->subvolume_lock);
+out:
+ if (!avail_subvol) {
+ gf_msg_debug(this->name, 0,
+ "No subvolume has enough free space \
+ and/or inodes to create");
+ avail_subvol = subvol;
+ }
+ if (layout)
+ dht_layout_unref(this, layout);
+ return avail_subvol;
+}
- /* check if subvol has layout errors, before selecting it */
- for (i = 0; i < layout->cnt; i++) {
- if (!strcmp (layout->list[i].xlator->name, this->name) &&
- (layout->list[i].err != 0)) {
- ret = -1;
- goto out;
- }
+static inline int32_t
+dht_subvol_has_err(dht_conf_t *conf, xlator_t *this, xlator_t *ignore,
+ dht_layout_t *layout)
+{
+ int ret = -1;
+ int i = 0;
+
+ if (!this || !layout)
+ goto out;
+
+ /* this check is meant for rebalance process. The source of the file
+ * should be ignored for space check */
+ if (this == ignore) {
+ goto out;
+ }
+
+ /* check if subvol has layout errors, before selecting it */
+ for (i = 0; i < layout->cnt; i++) {
+ if (!strcmp(layout->list[i].xlator->name, this->name) &&
+ (layout->list[i].err != 0)) {
+ ret = -1;
+ goto out;
}
+ }
- /* discard decommissioned subvol */
- if (conf->decommission_subvols_cnt) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->decommissioned_bricks[i] &&
- conf->decommissioned_bricks[i] == this) {
- ret = -1;
- goto out;
- }
- }
+ /* discard decommissioned subvol */
+ if (conf->decommission_subvols_cnt) {
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->decommissioned_bricks[i] &&
+ conf->decommissioned_bricks[i] == this) {
+ ret = -1;
+ goto out;
+ }
}
+ }
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
/*Get subvolume which has both space and inodes more than the min criteria*/
xlator_t *
-dht_subvol_with_free_space_inodes(xlator_t *this, xlator_t *subvol, xlator_t *ignore,
- dht_layout_t *layout, uint64_t filesize)
+dht_subvol_with_free_space_inodes(xlator_t *this, xlator_t *subvol,
+ xlator_t *ignore, dht_layout_t *layout,
+ uint64_t filesize)
{
- int i = 0;
- double max = 0;
- double max_inodes = 0;
- int ignore_subvol = 0;
- uint64_t total_blocks = 0;
- uint64_t avail_blocks = 0;
- uint64_t frsize = 0;
- double post_availspace = 0;
- double post_percent = 0;
-
- xlator_t *avail_subvol = NULL;
- dht_conf_t *conf = NULL;
-
- conf = this->private;
-
- for(i=0; i < conf->subvolume_cnt; i++) {
- /* check if subvol has layout errors and also it is not a
- * decommissioned brick, before selecting it */
- ignore_subvol = dht_subvol_has_err (conf, conf->subvolumes[i],
- ignore, layout);
- if (ignore_subvol)
- continue;
-
- if ((conf->disk_unit == 'p') &&
- (conf->du_stats[i].avail_percent > conf->min_free_disk) &&
- (conf->du_stats[i].avail_inodes > conf->min_free_inodes)) {
- if ((conf->du_stats[i].avail_inodes > max_inodes) ||
- (conf->du_stats[i].avail_percent > max)) {
- max = conf->du_stats[i].avail_percent;
- max_inodes = conf->du_stats[i].avail_inodes;
- avail_subvol = conf->subvolumes[i];
- total_blocks = conf->du_stats[i].total_blocks;
- avail_blocks = conf->du_stats[i].avail_blocks;
- frsize = conf->du_stats[i].frsize;
- }
- }
-
- if ((conf->disk_unit != 'p') &&
- (conf->du_stats[i].avail_space > conf->min_free_disk) &&
- (conf->du_stats[i].avail_inodes > conf->min_free_inodes)) {
- if ((conf->du_stats[i].avail_inodes > max_inodes) ||
- (conf->du_stats[i].avail_space > max)) {
- max = conf->du_stats[i].avail_space;
- max_inodes = conf->du_stats[i].avail_inodes;
- avail_subvol = conf->subvolumes[i];
- }
- }
+ int i = 0;
+ double max = 0;
+ double max_inodes = 0;
+ int ignore_subvol = 0;
+ uint64_t total_blocks = 0;
+ uint64_t avail_blocks = 0;
+ uint64_t frsize = 0;
+ double post_availspace = 0;
+ double post_percent = 0;
+
+ xlator_t *avail_subvol = NULL;
+ dht_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ /* check if subvol has layout errors and also it is not a
+ * decommissioned brick, before selecting it */
+ ignore_subvol = dht_subvol_has_err(conf, conf->subvolumes[i], ignore,
+ layout);
+ if (ignore_subvol)
+ continue;
+
+ if ((conf->disk_unit == 'p') &&
+ (conf->du_stats[i].avail_percent > conf->min_free_disk) &&
+ (conf->du_stats[i].avail_inodes > conf->min_free_inodes)) {
+ if ((conf->du_stats[i].avail_inodes > max_inodes) ||
+ (conf->du_stats[i].avail_percent > max)) {
+ max = conf->du_stats[i].avail_percent;
+ max_inodes = conf->du_stats[i].avail_inodes;
+ avail_subvol = conf->subvolumes[i];
+ total_blocks = conf->du_stats[i].total_blocks;
+ avail_blocks = conf->du_stats[i].avail_blocks;
+ frsize = conf->du_stats[i].frsize;
+ }
}
- if (avail_subvol) {
- if (conf->disk_unit == 'p') {
- post_availspace = (avail_blocks * frsize) - filesize;
- post_percent = (post_availspace * 100) / (total_blocks * frsize);
- if (post_percent < conf->min_free_disk)
- avail_subvol = NULL;
- }
- if (conf->disk_unit != 'p') {
- if ((max - filesize) < conf->min_free_disk)
- avail_subvol = NULL;
- }
+ if ((conf->disk_unit != 'p') &&
+ (conf->du_stats[i].avail_space > conf->min_free_disk) &&
+ (conf->du_stats[i].avail_inodes > conf->min_free_inodes)) {
+ if ((conf->du_stats[i].avail_inodes > max_inodes) ||
+ (conf->du_stats[i].avail_space > max)) {
+ max = conf->du_stats[i].avail_space;
+ max_inodes = conf->du_stats[i].avail_inodes;
+ avail_subvol = conf->subvolumes[i];
+ }
}
+ }
+
+ if (avail_subvol) {
+ if (conf->disk_unit == 'p') {
+ post_availspace = (avail_blocks * frsize) - filesize;
+ post_percent = (post_availspace * 100) / (total_blocks * frsize);
+ if (post_percent < conf->min_free_disk)
+ avail_subvol = NULL;
+ }
+ if (conf->disk_unit != 'p') {
+ if ((max - filesize) < conf->min_free_disk)
+ avail_subvol = NULL;
+ }
+ }
- return avail_subvol;
+ return avail_subvol;
}
-
-/* Get subvol which has atleast one inode and maximum space */
+/* Get subvol which has at least one inode and maximum space */
xlator_t *
-dht_subvol_maxspace_nonzeroinode (xlator_t *this, xlator_t *subvol,
- dht_layout_t *layout)
+dht_subvol_maxspace_nonzeroinode(xlator_t *this, xlator_t *subvol,
+ dht_layout_t *layout)
{
- int i = 0;
- double max = 0;
- int ignore_subvol = 0;
-
- xlator_t *avail_subvol = NULL;
- dht_conf_t *conf = NULL;
-
- conf = this->private;
-
- for (i = 0; i < conf->subvolume_cnt; i++) {
- /* check if subvol has layout errors and also it is not a
- * decommissioned brick, before selecting it*/
-
- ignore_subvol = dht_subvol_has_err (conf, conf->subvolumes[i], NULL,
- layout);
- if (ignore_subvol)
- continue;
-
- if (conf->disk_unit == 'p') {
- if ((conf->du_stats[i].avail_percent > max)
- && (conf->du_stats[i].avail_inodes > 0 )) {
- max = conf->du_stats[i].avail_percent;
- avail_subvol = conf->subvolumes[i];
- }
- } else {
- if ((conf->du_stats[i].avail_space > max)
- && (conf->du_stats[i].avail_inodes > 0)) {
- max = conf->du_stats[i].avail_space;
- avail_subvol = conf->subvolumes[i];
- }
- }
+ int i = 0;
+ double max = 0;
+ int ignore_subvol = 0;
+
+ xlator_t *avail_subvol = NULL;
+ dht_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ /* check if subvol has layout errors and also it is not a
+ * decommissioned brick, before selecting it*/
+
+ ignore_subvol = dht_subvol_has_err(conf, conf->subvolumes[i], NULL,
+ layout);
+ if (ignore_subvol)
+ continue;
+
+ if (conf->disk_unit == 'p') {
+ if ((conf->du_stats[i].avail_percent > max) &&
+ (conf->du_stats[i].avail_inodes > 0)) {
+ max = conf->du_stats[i].avail_percent;
+ avail_subvol = conf->subvolumes[i];
+ }
+ } else {
+ if ((conf->du_stats[i].avail_space > max) &&
+ (conf->du_stats[i].avail_inodes > 0)) {
+ max = conf->du_stats[i].avail_space;
+ avail_subvol = conf->subvolumes[i];
+ }
}
+ }
- return avail_subvol;
+ return avail_subvol;
}
diff --git a/xlators/cluster/dht/src/dht-hashfn.c b/xlators/cluster/dht/src/dht-hashfn.c
index f8e614a40aa..acda67c312a 100644
--- a/xlators/cluster/dht/src/dht-hashfn.c
+++ b/xlators/cluster/dht/src/dht-hashfn.c
@@ -8,104 +8,103 @@
cases as published by the Free Software Foundation.
*/
-
-#include "glusterfs.h"
-#include "xlator.h"
#include "dht-common.h"
-#include "hashfn.h"
-
+#include <glusterfs/hashfn.h>
-int
-dht_hash_compute_internal (int type, const char *name, uint32_t *hash_p)
+static int
+dht_hash_compute_internal(int type, const char *name, const int len,
+ uint32_t *hash_p)
{
- int ret = 0;
- uint32_t hash = 0;
+ int ret = 0;
+ uint32_t hash = 0;
- switch (type) {
+ switch (type) {
case DHT_HASH_TYPE_DM:
case DHT_HASH_TYPE_DM_USER:
- hash = gf_dm_hashfn (name, strlen (name));
- break;
+ hash = gf_dm_hashfn(name, len);
+ break;
default:
- ret = -1;
- break;
- }
+ ret = -1;
+ break;
+ }
- if (ret == 0) {
- *hash_p = hash;
- }
+ if (ret == 0) {
+ *hash_p = hash;
+ }
- return ret;
+ return ret;
}
-
-static
-gf_boolean_t
-dht_munge_name (const char *original, char *modified,
- size_t len, regex_t *re)
+/* The function returns:
+ * 0 : in case no munge took place
+ * >0 : the length (inc. terminating NULL!) of the newly modified string,
+ * if it was munged.
+ */
+static int
+dht_munge_name(const char *original, char *modified, size_t len, regex_t *re)
{
- regmatch_t matches[2] = {{0}, };
- size_t new_len = 0;
- int ret = 0;
-
- ret = regexec(re, original, 2, matches, 0);
-
- if (ret != REG_NOMATCH) {
- if (matches[1].rm_so != -1) {
- new_len = matches[1].rm_eo - matches[1].rm_so;
- /* Equal would fail due to the NUL at the end. */
- if (new_len < len) {
- memcpy (modified,original+matches[1].rm_so,
- new_len);
- modified[new_len] = '\0';
- return _gf_true;
- }
- }
+ regmatch_t matches[2] = {
+ {0},
+ };
+ size_t new_len = 0;
+ int ret = 0;
+
+ ret = regexec(re, original, 2, matches, 0);
+
+ if (ret != REG_NOMATCH) {
+ if (matches[1].rm_so != -1) {
+ new_len = matches[1].rm_eo - matches[1].rm_so;
+ /* Equal would fail due to the NUL at the end. */
+ if (new_len < len) {
+ memcpy(modified, original + matches[1].rm_so, new_len);
+ modified[new_len] = '\0';
+ return new_len + 1; /* +1 for the terminating NULL */
+ }
}
+ }
- /* This is guaranteed safe because of how the dest was allocated. */
- strcpy(modified, original);
- return _gf_false;
+ /* This is guaranteed safe because of how the dest was allocated. */
+ strcpy(modified, original);
+ return 0;
}
int
-dht_hash_compute (xlator_t *this, int type, const char *name, uint32_t *hash_p)
+dht_hash_compute(xlator_t *this, int type, const char *name, uint32_t *hash_p)
{
- char *rsync_friendly_name = NULL;
- dht_conf_t *priv = NULL;
- size_t len = 0;
- gf_boolean_t munged = _gf_false;
-
- priv = this->private;
-
- LOCK (&priv->lock);
- {
- if (priv->extra_regex_valid) {
- len = strlen(name) + 1;
- rsync_friendly_name = alloca(len);
- munged = dht_munge_name (name, rsync_friendly_name, len,
- &priv->extra_regex);
- }
-
- if (!munged && priv->rsync_regex_valid) {
- len = strlen(name) + 1;
- rsync_friendly_name = alloca(len);
- gf_msg_trace (this->name, 0, "trying regex for %s",
- name);
- munged = dht_munge_name (name, rsync_friendly_name, len,
- &priv->rsync_regex);
- if (munged) {
- gf_msg_debug (this->name, 0,
- "munged down to %s",
- rsync_friendly_name);
- }
- }
- }
- UNLOCK (&priv->lock);
+ char *rsync_friendly_name = NULL;
+ dht_conf_t *priv = NULL;
+ size_t len = 0;
+ int munged = 0;
+
+ priv = this->private;
+
+ if (name == NULL)
+ return -1;
- if (!munged) {
- rsync_friendly_name = (char *)name;
+ len = strlen(name) + 1;
+ rsync_friendly_name = alloca(len);
+
+ LOCK(&priv->lock);
+ {
+ if (priv->extra_regex_valid) {
+ munged = dht_munge_name(name, rsync_friendly_name, len,
+ &priv->extra_regex);
}
- return dht_hash_compute_internal (type, rsync_friendly_name, hash_p);
+ if (!munged && priv->rsync_regex_valid) {
+ gf_msg_trace(this->name, 0, "trying regex for %s", name);
+ munged = dht_munge_name(name, rsync_friendly_name, len,
+ &priv->rsync_regex);
+ }
+ }
+ UNLOCK(&priv->lock);
+ if (munged) {
+ gf_msg_debug(this->name, 0, "munged down to %s", rsync_friendly_name);
+ len = munged;
+ } else {
+ rsync_friendly_name = (char *)name;
+ }
+
+ return dht_hash_compute_internal(type, rsync_friendly_name, len - 1,
+ hash_p);
}
diff --git a/xlators/cluster/dht/src/dht-helper.c b/xlators/cluster/dht/src/dht-helper.c
index 381856f2455..3f2fe43d5f3 100644
--- a/xlators/cluster/dht/src/dht-helper.c
+++ b/xlators/cluster/dht/src/dht-helper.c
@@ -8,266 +8,249 @@
cases as published by the Free Software Foundation.
*/
-
-#include "glusterfs.h"
-#include "xlator.h"
#include "dht-common.h"
#include "dht-lock.h"
+#include "glusterfs/compat-errno.h" // for ENODATA on BSD
static void
-dht_free_fd_ctx (dht_fd_ctx_t *fd_ctx)
+dht_free_fd_ctx(dht_fd_ctx_t *fd_ctx)
{
- GF_FREE (fd_ctx);
+ GF_FREE(fd_ctx);
}
-
int32_t
-dht_fd_ctx_destroy (xlator_t *this, fd_t *fd)
+dht_fd_ctx_destroy(xlator_t *this, fd_t *fd)
{
- dht_fd_ctx_t *fd_ctx = NULL;
- uint64_t value = 0;
- int32_t ret = -1;
+ dht_fd_ctx_t *fd_ctx = NULL;
+ uint64_t value = 0;
+ int32_t ret = -1;
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
- ret = fd_ctx_del (fd, this, &value);
- if (ret) {
- goto out;
- }
+ ret = fd_ctx_del(fd, this, &value);
+ if (ret) {
+ goto out;
+ }
- fd_ctx = (dht_fd_ctx_t *)value;
- if (fd_ctx) {
- GF_REF_PUT (fd_ctx);
- }
+ fd_ctx = (dht_fd_ctx_t *)(uintptr_t)value;
+ if (fd_ctx) {
+ GF_REF_PUT(fd_ctx);
+ }
out:
- return ret;
+ return ret;
}
-
static int
-__dht_fd_ctx_set (xlator_t *this, fd_t *fd, xlator_t *dst)
+__dht_fd_ctx_set(xlator_t *this, fd_t *fd, xlator_t *dst)
{
- dht_fd_ctx_t *fd_ctx = NULL;
- uint64_t value = 0;
- int ret = -1;
+ dht_fd_ctx_t *fd_ctx = NULL;
+ uint64_t value = 0;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
- fd_ctx = GF_CALLOC (1, sizeof (*fd_ctx), gf_dht_mt_fd_ctx_t);
+ fd_ctx = GF_CALLOC(1, sizeof(*fd_ctx), gf_dht_mt_fd_ctx_t);
- if (!fd_ctx) {
- goto out;
- }
+ if (!fd_ctx) {
+ goto out;
+ }
- fd_ctx->opened_on_dst = (uint64_t) dst;
- GF_REF_INIT (fd_ctx, dht_free_fd_ctx);
+ fd_ctx->opened_on_dst = (uint64_t)(uintptr_t)dst;
+ GF_REF_INIT(fd_ctx, dht_free_fd_ctx);
- value = (uint64_t) fd_ctx;
+ value = (uint64_t)(uintptr_t)fd_ctx;
- ret = __fd_ctx_set (fd, this, value);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_FD_CTX_SET_FAILED,
- "Failed to set fd ctx in fd=0x%p", fd);
- GF_REF_PUT (fd_ctx);
- }
+ ret = __fd_ctx_set(fd, this, value);
+ if (ret < 0) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_FD_CTX_SET_FAILED,
+ "fd=0x%p", fd, NULL);
+ GF_REF_PUT(fd_ctx);
+ }
out:
- return ret;
+ return ret;
}
-
-
int
-dht_fd_ctx_set (xlator_t *this, fd_t *fd, xlator_t *dst)
+dht_fd_ctx_set(xlator_t *this, fd_t *fd, xlator_t *dst)
{
- dht_fd_ctx_t *fd_ctx = NULL;
- uint64_t value = 0;
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
-
- LOCK (&fd->lock);
- {
- ret = __fd_ctx_get (fd, this, &value);
- if (ret && value) {
-
- fd_ctx = (dht_fd_ctx_t *) value;
- if (fd_ctx->opened_on_dst == (uint64_t) dst) {
- /* This could happen due to racing
- * check_progress tasks*/
- goto unlock;
- } else {
- /* This would be a big problem*/
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_INVALID_VALUE,
- "Different dst found in the fd ctx");
-
- /* Overwrite and hope for the best*/
- fd_ctx->opened_on_dst = (uint64_t)dst;
- goto unlock;
- }
+ dht_fd_ctx_t *fd_ctx = NULL;
+ uint64_t value = 0;
+ int ret = -1;
+
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
+
+ LOCK(&fd->lock);
+ {
+ ret = __fd_ctx_get(fd, this, &value);
+ if (ret && value) {
+ fd_ctx = (dht_fd_ctx_t *)(uintptr_t)value;
+ if (fd_ctx->opened_on_dst == (uint64_t)(uintptr_t)dst) {
+ /* This could happen due to racing
+ * check_progress tasks*/
+ goto unlock;
+ } else {
+ /* This would be a big problem*/
+ /* Overwrite and hope for the best*/
+ fd_ctx->opened_on_dst = (uint64_t)(uintptr_t)dst;
+ UNLOCK(&fd->lock);
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_INVALID_VALUE,
+ NULL);
- }
- ret = __dht_fd_ctx_set (this, fd, dst);
+ goto out;
+ }
}
+ ret = __dht_fd_ctx_set(this, fd, dst);
+ }
unlock:
- UNLOCK (&fd->lock);
+ UNLOCK(&fd->lock);
out:
- return ret;
+ return ret;
}
-
-
-static
-dht_fd_ctx_t *
-dht_fd_ctx_get (xlator_t *this, fd_t *fd)
+static dht_fd_ctx_t *
+dht_fd_ctx_get(xlator_t *this, fd_t *fd)
{
- dht_fd_ctx_t *fd_ctx = NULL;
- int ret = -1;
- uint64_t tmp_val = 0;
-
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
-
- LOCK (&fd->lock);
- {
- ret = __fd_ctx_get (fd, this, &tmp_val);
- if ((ret < 0) || (tmp_val == 0)) {
- UNLOCK (&fd->lock);
- goto out;
- }
+ dht_fd_ctx_t *fd_ctx = NULL;
+ int ret = -1;
+ uint64_t tmp_val = 0;
- fd_ctx = (dht_fd_ctx_t *)tmp_val;
- GF_REF_GET (fd_ctx);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
+
+ LOCK(&fd->lock);
+ {
+ ret = __fd_ctx_get(fd, this, &tmp_val);
+ if ((ret < 0) || (tmp_val == 0)) {
+ goto unlock;
}
- UNLOCK (&fd->lock);
+
+ fd_ctx = (dht_fd_ctx_t *)(uintptr_t)tmp_val;
+ GF_REF_GET(fd_ctx);
+ }
+unlock:
+ UNLOCK(&fd->lock);
out:
- return fd_ctx;
+ return fd_ctx;
}
gf_boolean_t
-dht_fd_open_on_dst (xlator_t *this, fd_t *fd, xlator_t *dst)
+dht_fd_open_on_dst(xlator_t *this, fd_t *fd, xlator_t *dst)
{
- dht_fd_ctx_t *fd_ctx = NULL;
- gf_boolean_t opened = _gf_false;
+ dht_fd_ctx_t *fd_ctx = NULL;
+ gf_boolean_t opened = _gf_false;
- fd_ctx = dht_fd_ctx_get (this, fd);
+ fd_ctx = dht_fd_ctx_get(this, fd);
- if (fd_ctx) {
- if (fd_ctx->opened_on_dst == (uint64_t) dst) {
- opened = _gf_true;
- }
- GF_REF_PUT (fd_ctx);
+ if (fd_ctx) {
+ if (fd_ctx->opened_on_dst == (uint64_t)(uintptr_t)dst) {
+ opened = _gf_true;
}
+ GF_REF_PUT(fd_ctx);
+ }
- return opened;
+ return opened;
}
-
void
-dht_free_mig_info (void *data)
+dht_free_mig_info(void *data)
{
- dht_migrate_info_t *miginfo = NULL;
+ dht_migrate_info_t *miginfo = NULL;
- miginfo = data;
- GF_FREE (miginfo);
+ miginfo = data;
+ GF_FREE(miginfo);
- return;
+ return;
}
static int
-dht_inode_ctx_set_mig_info (xlator_t *this, inode_t *inode,
- xlator_t *src_subvol, xlator_t *dst_subvol)
+dht_inode_ctx_set_mig_info(xlator_t *this, inode_t *inode, xlator_t *src_subvol,
+ xlator_t *dst_subvol)
{
- dht_migrate_info_t *miginfo = NULL;
- uint64_t value = 0;
- int ret = -1;
+ dht_migrate_info_t *miginfo = NULL;
+ uint64_t value = 0;
+ int ret = -1;
- miginfo = GF_CALLOC (1, sizeof (*miginfo), gf_dht_mt_miginfo_t);
- if (miginfo == NULL)
- goto out;
+ miginfo = GF_CALLOC(1, sizeof(*miginfo), gf_dht_mt_miginfo_t);
+ if (miginfo == NULL)
+ goto out;
- miginfo->src_subvol = src_subvol;
- miginfo->dst_subvol = dst_subvol;
- GF_REF_INIT (miginfo, dht_free_mig_info);
+ miginfo->src_subvol = src_subvol;
+ miginfo->dst_subvol = dst_subvol;
+ GF_REF_INIT(miginfo, dht_free_mig_info);
- value = (uint64_t) miginfo;
+ value = (uint64_t)(uintptr_t)miginfo;
- ret = inode_ctx_set1 (inode, this, &value);
- if (ret < 0) {
- GF_REF_PUT (miginfo);
- }
+ ret = inode_ctx_set1(inode, this, &value);
+ if (ret < 0) {
+ GF_REF_PUT(miginfo);
+ }
out:
- return ret;
+ return ret;
}
-
int
-dht_inode_ctx_get_mig_info (xlator_t *this, inode_t *inode,
- xlator_t **src_subvol, xlator_t **dst_subvol)
+dht_inode_ctx_get_mig_info(xlator_t *this, inode_t *inode,
+ xlator_t **src_subvol, xlator_t **dst_subvol)
{
- int ret = -1;
- uint64_t tmp_miginfo = 0;
- dht_migrate_info_t *miginfo = NULL;
-
- LOCK (&inode->lock);
- {
- ret = __inode_ctx_get1 (inode, this, &tmp_miginfo);
- if ((ret < 0) || (tmp_miginfo == 0)) {
- UNLOCK (&inode->lock);
- goto out;
- }
+ int ret = -1;
+ uint64_t tmp_miginfo = 0;
+ dht_migrate_info_t *miginfo = NULL;
- miginfo = (dht_migrate_info_t *)tmp_miginfo;
- GF_REF_GET (miginfo);
+ LOCK(&inode->lock);
+ {
+ ret = __inode_ctx_get1(inode, this, &tmp_miginfo);
+ if ((ret < 0) || (tmp_miginfo == 0)) {
+ UNLOCK(&inode->lock);
+ goto out;
}
- UNLOCK (&inode->lock);
- if (src_subvol)
- *src_subvol = miginfo->src_subvol;
+ miginfo = (dht_migrate_info_t *)(uintptr_t)tmp_miginfo;
+ GF_REF_GET(miginfo);
+ }
+ UNLOCK(&inode->lock);
+
+ if (src_subvol)
+ *src_subvol = miginfo->src_subvol;
- if (dst_subvol)
- *dst_subvol = miginfo->dst_subvol;
+ if (dst_subvol)
+ *dst_subvol = miginfo->dst_subvol;
- GF_REF_PUT (miginfo);
+ GF_REF_PUT(miginfo);
out:
- return ret;
+ return ret;
}
gf_boolean_t
-dht_mig_info_is_invalid (xlator_t *current, xlator_t *src_subvol,
- xlator_t *dst_subvol)
+dht_mig_info_is_invalid(xlator_t *current, xlator_t *src_subvol,
+ xlator_t *dst_subvol)
{
-
-/* Not set
- */
- if (!src_subvol || !dst_subvol)
- return _gf_true;
-
-/* Invalid scenarios:
- * The src_subvol does not match the subvol on which the current op was sent
- * so the cached subvol has changed between the last mig_info_set and now.
- * src_subvol == dst_subvol. The file was migrated without any FOP detecting
- * a P2 so the old dst is now the current subvol.
- *
- * There is still one scenario where the info could be outdated - if
- * file has undergone multiple migrations and ends up on the same src_subvol
- * on which the mig_info was first set.
- */
- if ((current == dst_subvol) || (current != src_subvol))
- return _gf_true;
-
- return _gf_false;
+ /* Not set
+ */
+ if (!src_subvol || !dst_subvol)
+ return _gf_true;
+
+ /* Invalid scenarios:
+ * The src_subvol does not match the subvol on which the current op was sent
+ * so the cached subvol has changed between the last mig_info_set and now.
+ * src_subvol == dst_subvol. The file was migrated without any FOP detecting
+ * a P2 so the old dst is now the current subvol.
+ *
+ * There is still one scenario where the info could be outdated - if
+ * file has undergone multiple migrations and ends up on the same src_subvol
+ * on which the mig_info was first set.
+ */
+ if ((current == dst_subvol) || (current != src_subvol))
+ return _gf_true;
+
+ return _gf_false;
}
-
-
/* Used to check if fd fops have the fd opened on the cached subvol
* This is required when:
* 1. an fd is opened on FILE1 on subvol1
@@ -279,984 +262,974 @@ dht_mig_info_is_invalid (xlator_t *current, xlator_t *src_subvol,
*
*/
-
int
-dht_check_and_open_fd_on_subvol_complete (int ret, call_frame_t *frame,
- void *data)
+dht_check_and_open_fd_on_subvol_complete(int ret, call_frame_t *frame,
+ void *data)
{
- glusterfs_fop_t fop = 0;
- dht_local_t *local = NULL;
- xlator_t *subvol = NULL;
- xlator_t *this = NULL;
- fd_t *fd = NULL;
- int op_errno = -1;
-
- local = frame->local;
- this = frame->this;
- fop = local->fop;
- subvol = local->cached_subvol;
- fd = local->fd;
-
- if (ret) {
- op_errno = local->op_errno;
- goto handle_err;
- }
-
- switch (fop) {
-
+ glusterfs_fop_t fop = 0;
+ dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+ xlator_t *this = NULL;
+ fd_t *fd = NULL;
+ int op_errno = -1;
+
+ local = frame->local;
+ this = frame->this;
+ fop = local->fop;
+ subvol = local->cached_subvol;
+ fd = local->fd;
+
+ if (ret) {
+ op_errno = local->op_errno;
+ goto handle_err;
+ }
+
+ switch (fop) {
case GF_FOP_WRITE:
- STACK_WIND_COOKIE (frame, dht_writev_cbk, subvol, subvol,
- subvol->fops->writev, fd,
- local->rebalance.vector,
- local->rebalance.count,
- local->rebalance.offset,
- local->rebalance.flags,
- local->rebalance.iobref, local->xattr_req);
- break;
+ STACK_WIND_COOKIE(frame, dht_writev_cbk, subvol, subvol,
+ subvol->fops->writev, fd, local->rebalance.vector,
+ local->rebalance.count, local->rebalance.offset,
+ local->rebalance.flags, local->rebalance.iobref,
+ local->xattr_req);
+ break;
case GF_FOP_FLUSH:
- STACK_WIND (frame, dht_flush_cbk, subvol,
- subvol->fops->flush, fd, local->xattr_req);
- break;
+ STACK_WIND(frame, dht_flush_cbk, subvol, subvol->fops->flush, fd,
+ local->xattr_req);
+ break;
case GF_FOP_FSETATTR:
- STACK_WIND_COOKIE (frame, dht_file_setattr_cbk, subvol,
- subvol, subvol->fops->fsetattr, fd,
- &local->rebalance.stbuf,
- local->rebalance.flags,
- local->xattr_req);
- break;
+ STACK_WIND_COOKIE(frame, dht_file_setattr_cbk, subvol, subvol,
+ subvol->fops->fsetattr, fd,
+ &local->rebalance.stbuf, local->rebalance.flags,
+ local->xattr_req);
+ break;
case GF_FOP_ZEROFILL:
- STACK_WIND_COOKIE (frame, dht_zerofill_cbk, subvol, subvol,
- subvol->fops->zerofill, fd,
- local->rebalance.offset,
- local->rebalance.size, local->xattr_req);
+ STACK_WIND_COOKIE(frame, dht_zerofill_cbk, subvol, subvol,
+ subvol->fops->zerofill, fd,
+ local->rebalance.offset, local->rebalance.size,
+ local->xattr_req);
- break;
+ break;
case GF_FOP_DISCARD:
- STACK_WIND_COOKIE (frame, dht_discard_cbk, subvol, subvol,
- subvol->fops->discard, local->fd,
- local->rebalance.offset,
- local->rebalance.size,
- local->xattr_req);
- break;
+ STACK_WIND_COOKIE(frame, dht_discard_cbk, subvol, subvol,
+ subvol->fops->discard, local->fd,
+ local->rebalance.offset, local->rebalance.size,
+ local->xattr_req);
+ break;
case GF_FOP_FALLOCATE:
- STACK_WIND_COOKIE (frame, dht_fallocate_cbk, subvol, subvol,
- subvol->fops->fallocate, fd,
- local->rebalance.flags,
- local->rebalance.offset,
- local->rebalance.size,
- local->xattr_req);
- break;
+ STACK_WIND_COOKIE(frame, dht_fallocate_cbk, subvol, subvol,
+ subvol->fops->fallocate, fd,
+ local->rebalance.flags, local->rebalance.offset,
+ local->rebalance.size, local->xattr_req);
+ break;
case GF_FOP_FTRUNCATE:
- STACK_WIND_COOKIE (frame, dht_truncate_cbk, subvol, subvol,
- subvol->fops->ftruncate, fd,
- local->rebalance.offset, local->xattr_req);
- break;
+ STACK_WIND_COOKIE(frame, dht_truncate_cbk, subvol, subvol,
+ subvol->fops->ftruncate, fd,
+ local->rebalance.offset, local->xattr_req);
+ break;
case GF_FOP_FSYNC:
- STACK_WIND_COOKIE (frame, dht_fsync_cbk, subvol, subvol,
- subvol->fops->fsync, local->fd,
- local->rebalance.flags, local->xattr_req);
- break;
+ STACK_WIND_COOKIE(frame, dht_fsync_cbk, subvol, subvol,
+ subvol->fops->fsync, local->fd,
+ local->rebalance.flags, local->xattr_req);
+ break;
case GF_FOP_READ:
- STACK_WIND (frame, dht_readv_cbk, subvol, subvol->fops->readv,
- local->fd, local->rebalance.size,
- local->rebalance.offset,
- local->rebalance.flags, local->xattr_req);
- break;
+ STACK_WIND(frame, dht_readv_cbk, subvol, subvol->fops->readv,
+ local->fd, local->rebalance.size,
+ local->rebalance.offset, local->rebalance.flags,
+ local->xattr_req);
+ break;
case GF_FOP_FSTAT:
- STACK_WIND_COOKIE (frame, dht_file_attr_cbk, subvol,
- subvol, subvol->fops->fstat, fd,
- local->xattr_req);
- break;
+ STACK_WIND_COOKIE(frame, dht_file_attr_cbk, subvol, subvol,
+ subvol->fops->fstat, fd, local->xattr_req);
+ break;
case GF_FOP_FSETXATTR:
- STACK_WIND_COOKIE (frame, dht_file_setxattr_cbk, subvol,
- subvol, subvol->fops->fsetxattr, local->fd,
- local->rebalance.xattr,
- local->rebalance.flags, local->xattr_req);
- break;
+ STACK_WIND_COOKIE(frame, dht_file_setxattr_cbk, subvol, subvol,
+ subvol->fops->fsetxattr, local->fd,
+ local->rebalance.xattr, local->rebalance.flags,
+ local->xattr_req);
+ break;
case GF_FOP_FREMOVEXATTR:
- STACK_WIND_COOKIE (frame, dht_file_removexattr_cbk, subvol,
- subvol, subvol->fops->fremovexattr,
- local->fd, local->key, local->xattr_req);
-
- break;
-
+ STACK_WIND_COOKIE(frame, dht_file_removexattr_cbk, subvol, subvol,
+ subvol->fops->fremovexattr, local->fd, local->key,
+ local->xattr_req);
+
+ break;
+
+ case GF_FOP_FXATTROP:
+ STACK_WIND(frame, dht_common_xattrop_cbk, subvol,
+ subvol->fops->fxattrop, local->fd,
+ local->rebalance.flags, local->rebalance.xattr,
+ local->xattr_req);
+ break;
+
+ case GF_FOP_FGETXATTR:
+ STACK_WIND(frame, dht_getxattr_cbk, subvol, subvol->fops->fgetxattr,
+ local->fd, local->key, NULL);
+ break;
+
+ case GF_FOP_FINODELK:
+ STACK_WIND(frame, dht_finodelk_cbk, subvol, subvol->fops->finodelk,
+ local->key, local->fd, local->rebalance.lock_cmd,
+ &local->rebalance.flock, local->xattr_req);
+ break;
default:
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_UNKNOWN_FOP,
- "Unknown FOP on fd (%p) on file %s @ %s",
- fd, uuid_utoa (fd->inode->gfid),
- subvol->name);
- break;
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_UNKNOWN_FOP, "fd=%p",
+ fd, "gfid=%s", uuid_utoa(fd->inode->gfid), "name=%s",
+ subvol->name, NULL);
+ break;
+ }
- }
+ goto out;
- goto out;
-
- /* Could not open the fd on the dst. Unwind */
+ /* Could not open the fd on the dst. Unwind */
handle_err:
- switch (fop) {
-
+ switch (fop) {
case GF_FOP_WRITE:
- DHT_STACK_UNWIND (writev, frame, -1,
- op_errno, NULL, NULL, NULL);
- break;
+ DHT_STACK_UNWIND(writev, frame, -1, op_errno, NULL, NULL, NULL);
+ break;
case GF_FOP_FLUSH:
- DHT_STACK_UNWIND (flush, frame, -1, op_errno, NULL);
- break;
+ DHT_STACK_UNWIND(flush, frame, -1, op_errno, NULL);
+ break;
case GF_FOP_FSETATTR:
- DHT_STACK_UNWIND (fsetattr, frame, -1, op_errno,
- NULL, NULL, NULL);
- break;
+ DHT_STACK_UNWIND(fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
+ break;
case GF_FOP_ZEROFILL:
- DHT_STACK_UNWIND (zerofill, frame, -1, op_errno,
- NULL, NULL, NULL);
- break;
+ DHT_STACK_UNWIND(zerofill, frame, -1, op_errno, NULL, NULL, NULL);
+ break;
case GF_FOP_DISCARD:
- DHT_STACK_UNWIND (discard, frame, -1, op_errno,
- NULL, NULL, NULL);
- break;
+ DHT_STACK_UNWIND(discard, frame, -1, op_errno, NULL, NULL, NULL);
+ break;
case GF_FOP_FALLOCATE:
- DHT_STACK_UNWIND (fallocate, frame, -1, op_errno,
- NULL, NULL, NULL);
- break;
+ DHT_STACK_UNWIND(fallocate, frame, -1, op_errno, NULL, NULL, NULL);
+ break;
case GF_FOP_FTRUNCATE:
- DHT_STACK_UNWIND (ftruncate, frame, -1, op_errno,
- NULL, NULL, NULL);
- break;
+ DHT_STACK_UNWIND(ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
+ break;
case GF_FOP_FSYNC:
- DHT_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL, NULL);
- break;
+ DHT_STACK_UNWIND(fsync, frame, -1, op_errno, NULL, NULL, NULL);
+ break;
case GF_FOP_READ:
- DHT_STACK_UNWIND (readv, frame, -1, op_errno, NULL,
- 0, NULL, NULL, NULL);
- break;
+ DHT_STACK_UNWIND(readv, frame, -1, op_errno, NULL, 0, NULL, NULL,
+ NULL);
+ break;
case GF_FOP_FSTAT:
- DHT_STACK_UNWIND (fstat, frame, -1, op_errno, NULL, NULL);
- break;
+ DHT_STACK_UNWIND(fstat, frame, -1, op_errno, NULL, NULL);
+ break;
case GF_FOP_FSETXATTR:
- DHT_STACK_UNWIND (fsetxattr, frame, -1, op_errno, NULL);
- break;
+ DHT_STACK_UNWIND(fsetxattr, frame, -1, op_errno, NULL);
+ break;
case GF_FOP_FREMOVEXATTR:
- DHT_STACK_UNWIND (fremovexattr, frame, -1, op_errno, NULL);
- break;
+ DHT_STACK_UNWIND(fremovexattr, frame, -1, op_errno, NULL);
+ break;
+
+ case GF_FOP_FXATTROP:
+ DHT_STACK_UNWIND(fxattrop, frame, -1, op_errno, NULL, NULL);
+ break;
+
+ case GF_FOP_FGETXATTR:
+ DHT_STACK_UNWIND(fgetxattr, frame, -1, op_errno, NULL, NULL);
+ break;
+
+ case GF_FOP_FINODELK:
+ DHT_STACK_UNWIND(finodelk, frame, -1, op_errno, NULL);
+ break;
default:
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_UNKNOWN_FOP,
- "Unknown FOP on fd (%p) on file %s @ %s",
- fd, uuid_utoa (fd->inode->gfid),
- subvol->name);
- break;
- }
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_UNKNOWN_FOP, "fd=%p",
+ fd, "gfid=%s", uuid_utoa(fd->inode->gfid), "name=%s",
+ subvol->name, NULL);
+ break;
+ }
out:
- return 0;
-
+ return 0;
}
-
/* Check once again if the fd has been opened on the cached subvol.
* If not, open and update the fd_ctx.
*/
int
-dht_check_and_open_fd_on_subvol_task (void *data)
+dht_check_and_open_fd_on_subvol_task(void *data)
{
- loc_t loc = {0,};
- int ret = -1;
- call_frame_t *frame = NULL;
- dht_local_t *local = NULL;
- fd_t *fd = NULL;
- xlator_t *this = NULL;
- xlator_t *subvol = NULL;
-
-
- frame = data;
- local = frame->local;
- this = THIS;
- fd = local->fd;
- subvol = local->cached_subvol;
-
- local->fd_checked = _gf_true;
-
- if (fd_is_anonymous (fd) || dht_fd_open_on_dst (this, fd, subvol)) {
- ret = 0;
- goto out;
- }
-
- gf_msg_debug (this->name, 0,
- "Opening fd (%p, flags=0%o) on file %s @ %s",
- fd, fd->flags, uuid_utoa (fd->inode->gfid),
- subvol->name);
-
-
- loc.inode = inode_ref (fd->inode);
- gf_uuid_copy (loc.gfid, fd->inode->gfid);
+ loc_t loc = {
+ 0,
+ };
+ int ret = -1;
+ call_frame_t *frame = NULL;
+ dht_local_t *local = NULL;
+ fd_t *fd = NULL;
+ xlator_t *this = NULL;
+ xlator_t *subvol = NULL;
+
+ frame = data;
+ local = frame->local;
+ this = THIS;
+ fd = local->fd;
+ subvol = local->cached_subvol;
+
+ local->fd_checked = _gf_true;
+
+ if (fd_is_anonymous(fd) || dht_fd_open_on_dst(this, fd, subvol)) {
+ ret = 0;
+ goto out;
+ }
- /* Open this on the dst subvol */
+ gf_msg_debug(this->name, 0, "Opening fd (%p, flags=0%o) on file %s @ %s",
+ fd, fd->flags, uuid_utoa(fd->inode->gfid), subvol->name);
- SYNCTASK_SETID(0, 0);
+ loc.inode = inode_ref(fd->inode);
+ gf_uuid_copy(loc.gfid, fd->inode->gfid);
- ret = syncop_open (subvol, &loc,
- (fd->flags & ~(O_CREAT | O_EXCL | O_TRUNC)),
- fd, NULL, NULL);
+ /* Open this on the dst subvol */
- if (ret < 0) {
+ SYNCTASK_SETID(0, 0);
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_OPEN_FD_ON_DST_FAILED,
- "Failed to open the fd"
- " (%p, flags=0%o) on file %s @ %s",
- fd, fd->flags, uuid_utoa (fd->inode->gfid),
- subvol->name);
- /* This can happen if the cached subvol was updated in the
- * inode_ctx and the fd was opened on the new cached suvol
- * after this fop was wound on the old cached subvol.
- * As we do not close the fd on the old subvol (a leak)
- * don't treat ENOENT as an error and allow the phase1/phase2
- * checks to handle it.
- */
-
- if ((-ret != ENOENT) && (-ret != ESTALE)) {
- local->op_errno = -ret;
- ret = -1;
- } else {
- ret = 0;
- }
+ ret = syncop_open(subvol, &loc, (fd->flags & ~(O_CREAT | O_EXCL | O_TRUNC)),
+ fd, NULL, NULL);
- local->op_errno = -ret;
- ret = -1;
+ if (ret < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_OPEN_FD_ON_DST_FAILED,
+ "fd=%p", fd, "flags=0%o", fd->flags, "gfid=%s",
+ uuid_utoa(fd->inode->gfid), "name=%s", subvol->name, NULL);
+ /* This can happen if the cached subvol was updated in the
+ * inode_ctx and the fd was opened on the new cached suvol
+ * after this fop was wound on the old cached subvol.
+ * As we do not close the fd on the old subvol (a leak)
+ * don't treat ENOENT as an error and allow the phase1/phase2
+ * checks to handle it.
+ */
+ if ((-ret != ENOENT) && (-ret != ESTALE)) {
+ local->op_errno = -ret;
+ ret = -1;
} else {
- dht_fd_ctx_set (this, fd, subvol);
+ ret = 0;
}
- SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+ local->op_errno = -ret;
+ ret = -1;
+
+ } else {
+ dht_fd_ctx_set(this, fd, subvol);
+ }
+
+ SYNCTASK_SETID(frame->root->uid, frame->root->gid);
out:
- loc_wipe (&loc);
+ loc_wipe(&loc);
- return ret;
+ return ret;
}
-
int
-dht_check_and_open_fd_on_subvol (xlator_t *this, call_frame_t *frame)
+dht_check_and_open_fd_on_subvol(xlator_t *this, call_frame_t *frame)
{
- int ret = -1;
- dht_local_t *local = NULL;
+ int ret = -1;
+ dht_local_t *local = NULL;
-/*
- if (dht_fd_open_on_dst (this, fd, subvol))
- goto out;
-*/
- local = frame->local;
+ /*
+ if (dht_fd_open_on_dst (this, fd, subvol))
+ goto out;
+ */
+ local = frame->local;
- ret = synctask_new (this->ctx->env,
- dht_check_and_open_fd_on_subvol_task,
- dht_check_and_open_fd_on_subvol_complete,
- frame, frame);
+ ret = synctask_new(this->ctx->env, dht_check_and_open_fd_on_subvol_task,
+ dht_check_and_open_fd_on_subvol_complete, frame, frame);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0,
- "Failed to create synctask"
- " to check and open fd=%p", local->fd);
- }
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_SYNCTASK_CREATE_FAILED,
+ "to-check-and-open fd=%p", local->fd, NULL);
+ }
- return ret;
+ return ret;
}
-
-
int
-dht_frame_return (call_frame_t *frame)
+dht_frame_return(call_frame_t *frame)
{
- dht_local_t *local = NULL;
- int this_call_cnt = -1;
+ dht_local_t *local = NULL;
+ int this_call_cnt = -1;
- if (!frame)
- return -1;
+ if (!frame)
+ return -1;
- local = frame->local;
+ local = frame->local;
- LOCK (&frame->lock);
- {
- this_call_cnt = --local->call_cnt;
- }
- UNLOCK (&frame->lock);
+ LOCK(&frame->lock);
+ {
+ this_call_cnt = --local->call_cnt;
+ }
+ UNLOCK(&frame->lock);
- return this_call_cnt;
+ return this_call_cnt;
}
+/*
+ * Use this function to specify which subvol you want the file created
+ * on - this need not be the hashed subvol.
+ * Format: <filename>@<this->name>:<subvol-name>
+ * Eg: file-1@vol1-dht:vol1-client-0
+ * where vol1 is a pure distribute volume
+ * will create file-1 on vol1-client-0
+ */
int
-dht_filter_loc_subvol_key (xlator_t *this, loc_t *loc, loc_t *new_loc,
- xlator_t **subvol)
+dht_filter_loc_subvol_key(xlator_t *this, loc_t *loc, loc_t *new_loc,
+ xlator_t **subvol)
{
- char *new_name = NULL;
- char *new_path = NULL;
- xlator_list_t *trav = NULL;
- char key[1024] = {0,};
- int ret = 0; /* not found */
-
- /* Why do other tasks if first required 'char' itself is not there */
- if (!new_loc || !loc || !loc->name || !strchr (loc->name, '@')) {
- /* Skip the GF_FREE checks here */
- return ret;
- }
-
- trav = this->children;
- while (trav) {
- snprintf (key, 1024, "*@%s:%s", this->name, trav->xlator->name);
- if (fnmatch (key, loc->name, FNM_NOESCAPE) == 0) {
- new_name = GF_CALLOC(strlen (loc->name),
- sizeof (char),
- gf_common_mt_char);
- if (!new_name)
- goto out;
- if (fnmatch (key, loc->path, FNM_NOESCAPE) == 0) {
- new_path = GF_CALLOC(strlen (loc->path),
- sizeof (char),
- gf_common_mt_char);
- if (!new_path)
- goto out;
- strncpy (new_path, loc->path, (strlen (loc->path) -
- strlen (key) + 1));
- }
- strncpy (new_name, loc->name, (strlen (loc->name) -
- strlen (key) + 1));
-
- if (new_loc) {
- new_loc->path = ((new_path) ? new_path:
- gf_strdup (loc->path));
- new_loc->name = new_name;
- new_loc->inode = inode_ref (loc->inode);
- new_loc->parent = inode_ref (loc->parent);
- }
- *subvol = trav->xlator;
- ret = 1; /* success */
- goto out;
- }
- trav = trav->next;
- }
-out:
- if (!ret) {
- /* !success */
- GF_FREE (new_path);
- GF_FREE (new_name);
- }
+ char *new_name = NULL;
+ char *new_path = NULL;
+ xlator_list_t *trav = NULL;
+ char key[1024] = {
+ 0,
+ };
+ int ret = 0; /* not found */
+ int keylen = 0;
+ int name_len = 0;
+ int path_len = 0;
+
+ /* Why do other tasks if first required 'char' itself is not there */
+ if (!new_loc || !loc || !loc->name || !strchr(loc->name, '@')) {
+ /* Skip the GF_FREE checks here */
return ret;
+ }
+
+ trav = this->children;
+ while (trav) {
+ keylen = snprintf(key, sizeof(key), "*@%s:%s", this->name,
+ trav->xlator->name);
+ /* Ignore '*' */
+ keylen = keylen - 1;
+ if (fnmatch(key, loc->name, FNM_NOESCAPE) == 0) {
+ name_len = strlen(loc->name) - keylen;
+ new_name = GF_MALLOC(name_len + 1, gf_common_mt_char);
+ if (!new_name)
+ goto out;
+ if (fnmatch(key, loc->path, FNM_NOESCAPE) == 0) {
+ path_len = strlen(loc->path) - keylen;
+ new_path = GF_MALLOC(path_len + 1, gf_common_mt_char);
+ if (!new_path)
+ goto out;
+ snprintf(new_path, path_len + 1, "%s", loc->path);
+ }
+ snprintf(new_name, name_len + 1, "%s", loc->name);
+
+ if (new_loc) {
+ new_loc->path = ((new_path) ? new_path : gf_strdup(loc->path));
+ new_loc->name = new_name;
+ new_loc->inode = inode_ref(loc->inode);
+ new_loc->parent = inode_ref(loc->parent);
+ }
+ *subvol = trav->xlator;
+ ret = 1; /* success */
+ goto out;
+ }
+ trav = trav->next;
+ }
+out:
+ if (!ret) {
+ /* !success */
+ GF_FREE(new_path);
+ GF_FREE(new_name);
+ }
+ return ret;
}
static xlator_t *
dht_get_subvol_from_id(xlator_t *this, int client_id)
{
- xlator_t *xl = NULL;
- dht_conf_t *conf = NULL;
- char *sid = NULL;
- int32_t ret = -1;
-
- conf = this->private;
-
- ret = gf_asprintf(&sid, "%d", client_id);
- if (ret == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_ASPRINTF_FAILED, "asprintf failed while "
- "fetching subvol from the id");
- goto out;
- }
+ xlator_t *xl = NULL;
+ dht_conf_t *conf = NULL;
+ char *sid = NULL;
+ int32_t ret = -1;
- if (dict_get_ptr(conf->leaf_to_subvol, sid, (void **) &xl))
- xl = NULL;
+ conf = this->private;
- GF_FREE (sid);
+ ret = gf_asprintf(&sid, "%d", client_id);
+ if (ret == -1) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_ASPRINTF_FAILED, NULL);
+ goto out;
+ }
-out:
- return xl;
+ if (dict_get_ptr(conf->leaf_to_subvol, sid, (void **)&xl))
+ xl = NULL;
+
+ GF_FREE(sid);
+out:
+ return xl;
}
int
-dht_deitransform (xlator_t *this, uint64_t y, xlator_t **subvol_p)
+dht_deitransform(xlator_t *this, uint64_t y, xlator_t **subvol_p)
{
- int client_id = 0;
- xlator_t *subvol = 0;
- dht_conf_t *conf = NULL;
+ int client_id = 0;
+ xlator_t *subvol = 0;
+ dht_conf_t *conf = NULL;
- if (!this->private)
- return -1;
+ if (!this->private)
+ return -1;
- conf = this->private;
+ conf = this->private;
- client_id = gf_deitransform(this, y);
+ client_id = gf_deitransform(this, y);
- subvol = dht_get_subvol_from_id(this, client_id);
+ subvol = dht_get_subvol_from_id(this, client_id);
- if (!subvol)
- subvol = conf->subvolumes[0];
+ if (!subvol)
+ subvol = conf->subvolumes[0];
- if (subvol_p)
- *subvol_p = subvol;
+ if (subvol_p)
+ *subvol_p = subvol;
- return 0;
+ return 0;
}
void
-dht_local_wipe (xlator_t *this, dht_local_t *local)
+dht_local_wipe(xlator_t *this, dht_local_t *local)
{
- int i = 0;
+ int i = 0;
- if (!local)
- return;
+ if (!local)
+ return;
- loc_wipe (&local->loc);
- loc_wipe (&local->loc2);
+ loc_wipe(&local->loc);
+ loc_wipe(&local->loc2);
+ loc_wipe(&local->loc2_copy);
- if (local->xattr)
- dict_unref (local->xattr);
+ if (local->xattr)
+ dict_unref(local->xattr);
- if (local->inode)
- inode_unref (local->inode);
+ if (local->inode)
+ inode_unref(local->inode);
- if (local->layout) {
- dht_layout_unref (this, local->layout);
- local->layout = NULL;
- }
+ if (local->layout) {
+ dht_layout_unref(this, local->layout);
+ local->layout = NULL;
+ }
- loc_wipe (&local->linkfile.loc);
+ loc_wipe(&local->linkfile.loc);
- if (local->linkfile.xattr)
- dict_unref (local->linkfile.xattr);
+ if (local->linkfile.xattr)
+ dict_unref(local->linkfile.xattr);
- if (local->linkfile.inode)
- inode_unref (local->linkfile.inode);
+ if (local->linkfile.inode)
+ inode_unref(local->linkfile.inode);
- if (local->fd) {
- fd_unref (local->fd);
- local->fd = NULL;
- }
+ if (local->fd) {
+ fd_unref(local->fd);
+ local->fd = NULL;
+ }
- if (local->params) {
- dict_unref (local->params);
- local->params = NULL;
- }
+ if (local->params) {
+ dict_unref(local->params);
+ local->params = NULL;
+ }
- if (local->xattr_req)
- dict_unref (local->xattr_req);
- if (local->mds_xattr)
- dict_unref (local->mds_xattr);
- if (local->xdata)
- dict_unref (local->xdata);
+ if (local->xattr_req)
+ dict_unref(local->xattr_req);
+ if (local->mds_xattr)
+ dict_unref(local->mds_xattr);
+ if (local->xdata)
+ dict_unref(local->xdata);
- if (local->selfheal.layout) {
- dht_layout_unref (this, local->selfheal.layout);
- local->selfheal.layout = NULL;
- }
+ if (local->selfheal.layout) {
+ dht_layout_unref(this, local->selfheal.layout);
+ local->selfheal.layout = NULL;
+ }
- if (local->selfheal.refreshed_layout) {
- dht_layout_unref (this, local->selfheal.refreshed_layout);
- local->selfheal.refreshed_layout = NULL;
- }
-
- for (i = 0; i < 2; i++) {
- dht_lock_array_free (local->lock[i].ns.parent_layout.locks,
- local->lock[i].ns.parent_layout.lk_count);
+ if (local->selfheal.refreshed_layout) {
+ dht_layout_unref(this, local->selfheal.refreshed_layout);
+ local->selfheal.refreshed_layout = NULL;
+ }
- GF_FREE (local->lock[i].ns.parent_layout.locks);
+ for (i = 0; i < 2; i++) {
+ dht_lock_array_free(local->lock[i].ns.parent_layout.locks,
+ local->lock[i].ns.parent_layout.lk_count);
- dht_lock_array_free (local->lock[i].ns.directory_ns.locks,
- local->lock[i].ns.directory_ns.lk_count);
- GF_FREE (local->lock[i].ns.directory_ns.locks);
- }
+ GF_FREE(local->lock[i].ns.parent_layout.locks);
- GF_FREE (local->key);
+ dht_lock_array_free(local->lock[i].ns.directory_ns.locks,
+ local->lock[i].ns.directory_ns.lk_count);
+ GF_FREE(local->lock[i].ns.directory_ns.locks);
+ }
- if (local->rebalance.xdata)
- dict_unref (local->rebalance.xdata);
+ GF_FREE(local->key);
- if (local->rebalance.xattr)
- dict_unref (local->rebalance.xattr);
+ if (local->rebalance.xdata)
+ dict_unref(local->rebalance.xdata);
- if (local->rebalance.dict)
- dict_unref (local->rebalance.dict);
+ if (local->rebalance.xattr)
+ dict_unref(local->rebalance.xattr);
- GF_FREE (local->rebalance.vector);
+ if (local->rebalance.dict)
+ dict_unref(local->rebalance.dict);
- if (local->rebalance.iobref)
- iobref_unref (local->rebalance.iobref);
+ GF_FREE(local->rebalance.vector);
- if (local->stub) {
- call_stub_destroy (local->stub);
- local->stub = NULL;
- }
+ if (local->rebalance.iobref)
+ iobref_unref(local->rebalance.iobref);
- if (local->ret_cache)
- GF_FREE (local->ret_cache);
+ if (local->stub) {
+ call_stub_destroy(local->stub);
+ local->stub = NULL;
+ }
- gf_dirent_free (&local->entries);
+ if (local->ret_cache)
+ GF_FREE(local->ret_cache);
- mem_put (local);
+ mem_put(local);
}
-
dht_local_t *
-dht_local_init (call_frame_t *frame, loc_t *loc, fd_t *fd, glusterfs_fop_t fop)
+dht_local_init(call_frame_t *frame, loc_t *loc, fd_t *fd, glusterfs_fop_t fop)
{
- dht_local_t *local = NULL;
- inode_t *inode = NULL;
- int ret = 0;
+ dht_local_t *local = NULL;
+ inode_t *inode = NULL;
+ int ret = 0;
- local = mem_get0 (THIS->local_pool);
- if (!local)
- goto out;
+ local = mem_get0(THIS->local_pool);
+ if (!local)
+ goto out;
- if (loc) {
- ret = loc_copy (&local->loc, loc);
- if (ret)
- goto out;
+ if (loc) {
+ ret = loc_copy(&local->loc, loc);
+ if (ret)
+ goto out;
- inode = loc->inode;
- }
+ inode = loc->inode;
+ }
- if (fd) {
- local->fd = fd_ref (fd);
- if (!inode)
- inode = fd->inode;
- }
+ if (fd) {
+ local->fd = fd_ref(fd);
+ if (!inode)
+ inode = fd->inode;
+ }
- local->op_ret = -1;
- local->op_errno = EUCLEAN;
- local->fop = fop;
+ local->op_ret = -1;
+ local->op_errno = EUCLEAN;
+ local->fop = fop;
- if (inode) {
- local->layout = dht_layout_get (frame->this, inode);
- local->cached_subvol = dht_subvol_get_cached (frame->this,
- inode);
- }
+ if (inode) {
+ local->layout = dht_layout_get(frame->this, inode);
+ local->cached_subvol = dht_subvol_get_cached(frame->this, inode);
+ }
- INIT_LIST_HEAD (&local->entries.list);
- frame->local = local;
+ frame->local = local;
out:
- if (ret) {
- if (local)
- mem_put (local);
- local = NULL;
- }
- return local;
+ if (ret) {
+ if (local)
+ mem_put(local);
+ local = NULL;
+ }
+ return local;
}
xlator_t *
-dht_first_up_subvol (xlator_t *this)
+dht_first_up_subvol(xlator_t *this)
{
- dht_conf_t *conf = NULL;
- xlator_t *child = NULL;
- int i = 0;
- time_t time = 0;
+ dht_conf_t *conf = NULL;
+ xlator_t *child = NULL;
+ int i = 0;
+ time_t time = 0;
- conf = this->private;
- if (!conf)
- goto out;
+ conf = this->private;
+ if (!conf)
+ goto out;
- LOCK (&conf->subvolume_lock);
- {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->subvol_up_time[i]) {
- if (!time) {
- time = conf->subvol_up_time[i];
- child = conf->subvolumes[i];
- } else if (time > conf->subvol_up_time[i]) {
- time = conf->subvol_up_time[i];
- child = conf->subvolumes[i];
- }
- }
+ LOCK(&conf->subvolume_lock);
+ {
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->subvol_up_time[i]) {
+ if (!time) {
+ time = conf->subvol_up_time[i];
+ child = conf->subvolumes[i];
+ } else if (time > conf->subvol_up_time[i]) {
+ time = conf->subvol_up_time[i];
+ child = conf->subvolumes[i];
}
+ }
}
- UNLOCK (&conf->subvolume_lock);
+ }
+ UNLOCK(&conf->subvolume_lock);
out:
- return child;
+ return child;
}
xlator_t *
-dht_last_up_subvol (xlator_t *this)
+dht_last_up_subvol(xlator_t *this)
{
- dht_conf_t *conf = NULL;
- xlator_t *child = NULL;
- int i = 0;
+ dht_conf_t *conf = NULL;
+ xlator_t *child = NULL;
+ int i = 0;
- conf = this->private;
- if (!conf)
- goto out;
+ conf = this->private;
+ if (!conf)
+ goto out;
- LOCK (&conf->subvolume_lock);
- {
- for (i = conf->subvolume_cnt-1; i >= 0; i--) {
- if (conf->subvolume_status[i]) {
- child = conf->subvolumes[i];
- break;
- }
- }
+ LOCK(&conf->subvolume_lock);
+ {
+ for (i = conf->subvolume_cnt - 1; i >= 0; i--) {
+ if (conf->subvolume_status[i]) {
+ child = conf->subvolumes[i];
+ break;
+ }
}
- UNLOCK (&conf->subvolume_lock);
+ }
+ UNLOCK(&conf->subvolume_lock);
out:
- return child;
+ return child;
}
xlator_t *
-dht_subvol_get_hashed (xlator_t *this, loc_t *loc)
+dht_subvol_get_hashed(xlator_t *this, loc_t *loc)
{
- dht_layout_t *layout = NULL;
- xlator_t *subvol = NULL;
- dht_conf_t *conf = NULL;
- dht_methods_t *methods = NULL;
+ dht_layout_t *layout = NULL;
+ xlator_t *subvol = NULL;
+ dht_conf_t *conf = NULL;
+ dht_methods_t *methods = NULL;
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
- conf = this->private;
- GF_VALIDATE_OR_GOTO (this->name, conf, out);
+ conf = this->private;
+ GF_VALIDATE_OR_GOTO(this->name, conf, out);
- methods = &(conf->methods);
+ methods = &(conf->methods);
- if (__is_root_gfid (loc->gfid)) {
- subvol = dht_first_up_subvol (this);
- goto out;
- }
+ if (__is_root_gfid(loc->gfid)) {
+ subvol = dht_first_up_subvol(this);
+ goto out;
+ }
- GF_VALIDATE_OR_GOTO (this->name, loc->parent, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->name, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->parent, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->name, out);
- layout = dht_layout_get (this, loc->parent);
+ layout = dht_layout_get(this, loc->parent);
- if (!layout) {
- gf_msg_debug (this->name, 0,
- "Missing layout. path=%s, parent gfid =%s",
- loc->path, uuid_utoa (loc->parent->gfid));
- goto out;
- }
+ if (!layout) {
+ gf_msg_debug(this->name, 0, "Missing layout. path=%s, parent gfid =%s",
+ loc->path, uuid_utoa(loc->parent->gfid));
+ goto out;
+ }
- subvol = methods->layout_search (this, layout, loc->name);
+ subvol = methods->layout_search(this, layout, loc->name);
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "No hashed subvolume for path=%s",
- loc->path);
- goto out;
- }
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "No hashed subvolume for path=%s",
+ loc->path);
+ goto out;
+ }
out:
- if (layout) {
- dht_layout_unref (this, layout);
- }
+ if (layout) {
+ dht_layout_unref(this, layout);
+ }
- return subvol;
+ return subvol;
}
-
xlator_t *
-dht_subvol_get_cached (xlator_t *this, inode_t *inode)
+dht_subvol_get_cached(xlator_t *this, inode_t *inode)
{
- dht_layout_t *layout = NULL;
- xlator_t *subvol = NULL;
+ dht_layout_t *layout = NULL;
+ xlator_t *subvol = NULL;
- GF_VALIDATE_OR_GOTO (this->name, this, out);
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- layout = dht_layout_get (this, inode);
+ layout = dht_layout_get(this, inode);
- if (!layout) {
- goto out;
- }
+ if (!layout) {
+ goto out;
+ }
- subvol = layout->list[0].xlator;
+ subvol = layout->list[0].xlator;
out:
- if (layout) {
- dht_layout_unref (this, layout);
- }
+ if (layout) {
+ dht_layout_unref(this, layout);
+ }
- return subvol;
+ return subvol;
}
-
xlator_t *
-dht_subvol_next (xlator_t *this, xlator_t *prev)
+dht_subvol_next(xlator_t *this, xlator_t *prev)
{
- dht_conf_t *conf = NULL;
- int i = 0;
- xlator_t *next = NULL;
+ dht_conf_t *conf = NULL;
+ int i = 0;
+ xlator_t *next = NULL;
- conf = this->private;
- if (!conf)
- goto out;
+ conf = this->private;
+ if (!conf)
+ goto out;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->subvolumes[i] == prev) {
- if ((i + 1) < conf->subvolume_cnt)
- next = conf->subvolumes[i + 1];
- break;
- }
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->subvolumes[i] == prev) {
+ if ((i + 1) < conf->subvolume_cnt)
+ next = conf->subvolumes[i + 1];
+ break;
}
+ }
out:
- return next;
+ return next;
}
/* This func wraps around, if prev is actually the last subvol.
*/
xlator_t *
-dht_subvol_next_available (xlator_t *this, xlator_t *prev)
+dht_subvol_next_available(xlator_t *this, xlator_t *prev)
{
- dht_conf_t *conf = NULL;
- int i = 0;
- xlator_t *next = NULL;
+ dht_conf_t *conf = NULL;
+ int i = 0;
+ xlator_t *next = NULL;
- conf = this->private;
- if (!conf)
- goto out;
+ conf = this->private;
+ if (!conf)
+ goto out;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->subvolumes[i] == prev) {
- /* if prev is last in conf->subvolumes, then wrap
- * around.
- */
- if ((i + 1) < conf->subvolume_cnt) {
- next = conf->subvolumes[i + 1];
- } else {
- next = conf->subvolumes[0];
- }
- break;
- }
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->subvolumes[i] == prev) {
+ /* if prev is last in conf->subvolumes, then wrap
+ * around.
+ */
+ if ((i + 1) < conf->subvolume_cnt) {
+ next = conf->subvolumes[i + 1];
+ } else {
+ next = conf->subvolumes[0];
+ }
+ break;
}
+ }
out:
- return next;
+ return next;
}
int
-dht_subvol_cnt (xlator_t *this, xlator_t *subvol)
+dht_subvol_cnt(xlator_t *this, xlator_t *subvol)
{
- int i = 0;
- int ret = -1;
- dht_conf_t *conf = NULL;
+ int i = 0;
+ int ret = -1;
+ dht_conf_t *conf = NULL;
- conf = this->private;
- if (!conf)
- goto out;
+ conf = this->private;
+ if (!conf)
+ goto out;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (subvol == conf->subvolumes[i]) {
- ret = i;
- break;
- }
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (subvol == conf->subvolumes[i]) {
+ ret = i;
+ break;
}
+ }
out:
- return ret;
+ return ret;
}
+#define set_if_greater(a, b) \
+ do { \
+ if ((a) < (b)) \
+ (a) = (b); \
+ } while (0)
-#define set_if_greater(a, b) do { \
- if ((a) < (b)) \
- (a) = (b); \
- } while (0)
-
-
-#define set_if_greater_time(a, an, b, bn) do { \
- if (((a) < (b)) || (((a) == (b)) && ((an) < (bn)))){ \
- (a) = (b); \
- (an) = (bn); \
- } \
- } while (0) \
-
+#define set_if_greater_time(a, an, b, bn) \
+ do { \
+ if (((a) < (b)) || (((a) == (b)) && ((an) < (bn)))) { \
+ (a) = (b); \
+ (an) = (bn); \
+ } \
+ } while (0)
int
-dht_iatt_merge (xlator_t *this, struct iatt *to,
- struct iatt *from, xlator_t *subvol)
+dht_iatt_merge(xlator_t *this, struct iatt *to, struct iatt *from)
{
- if (!from || !to)
- return 0;
-
- to->ia_dev = from->ia_dev;
-
- gf_uuid_copy (to->ia_gfid, from->ia_gfid);
-
- to->ia_ino = from->ia_ino;
- to->ia_prot = from->ia_prot;
- to->ia_type = from->ia_type;
- to->ia_nlink = from->ia_nlink;
- to->ia_rdev = from->ia_rdev;
- to->ia_size += from->ia_size;
- to->ia_blksize = from->ia_blksize;
- to->ia_blocks += from->ia_blocks;
-
- if (IA_ISDIR (from->ia_type)) {
- to->ia_blocks = DHT_DIR_STAT_BLOCKS;
- to->ia_size = DHT_DIR_STAT_SIZE;
- }
- set_if_greater (to->ia_uid, from->ia_uid);
- set_if_greater (to->ia_gid, from->ia_gid);
-
- set_if_greater_time(to->ia_atime, to->ia_atime_nsec,
- from->ia_atime, from->ia_atime_nsec);
- set_if_greater_time (to->ia_mtime, to->ia_mtime_nsec,
- from->ia_mtime, from->ia_mtime_nsec);
- set_if_greater_time (to->ia_ctime, to->ia_ctime_nsec,
- from->ia_ctime, from->ia_ctime_nsec);
-
+ if (!from || !to)
return 0;
+
+ to->ia_dev = from->ia_dev;
+
+ gf_uuid_copy(to->ia_gfid, from->ia_gfid);
+
+ to->ia_ino = from->ia_ino;
+ to->ia_prot = from->ia_prot;
+ to->ia_type = from->ia_type;
+ to->ia_nlink = from->ia_nlink;
+ to->ia_rdev = from->ia_rdev;
+ to->ia_size += from->ia_size;
+ to->ia_blksize = from->ia_blksize;
+ to->ia_blocks += from->ia_blocks;
+
+ if (IA_ISDIR(from->ia_type)) {
+ to->ia_blocks = DHT_DIR_STAT_BLOCKS;
+ to->ia_size = DHT_DIR_STAT_SIZE;
+ }
+ set_if_greater(to->ia_uid, from->ia_uid);
+ set_if_greater(to->ia_gid, from->ia_gid);
+
+ set_if_greater_time(to->ia_atime, to->ia_atime_nsec, from->ia_atime,
+ from->ia_atime_nsec);
+ set_if_greater_time(to->ia_mtime, to->ia_mtime_nsec, from->ia_mtime,
+ from->ia_mtime_nsec);
+ set_if_greater_time(to->ia_ctime, to->ia_ctime_nsec, from->ia_ctime,
+ from->ia_ctime_nsec);
+
+ return 0;
}
int
-dht_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name)
+dht_build_child_loc(xlator_t *this, loc_t *child, loc_t *parent, char *name)
{
- if (!child) {
- goto err;
- }
+ if (!child) {
+ goto err;
+ }
- if (strcmp (parent->path, "/") == 0)
- gf_asprintf ((char **)&child->path, "/%s", name);
- else
- gf_asprintf ((char **)&child->path, "%s/%s", parent->path, name);
+ if (strcmp(parent->path, "/") == 0)
+ gf_asprintf((char **)&child->path, "/%s", name);
+ else
+ gf_asprintf((char **)&child->path, "%s/%s", parent->path, name);
- if (!child->path) {
- goto err;
- }
+ if (!child->path) {
+ goto err;
+ }
- child->name = strrchr (child->path, '/');
- if (child->name)
- child->name++;
+ child->name = strrchr(child->path, '/');
+ if (child->name)
+ child->name++;
- child->parent = inode_ref (parent->inode);
- child->inode = inode_new (parent->inode->table);
+ child->parent = inode_ref(parent->inode);
+ child->inode = inode_new(parent->inode->table);
- if (!child->inode) {
- goto err;
- }
+ if (!child->inode) {
+ goto err;
+ }
- return 0;
+ return 0;
err:
- if (child) {
- loc_wipe (child);
- }
- return -1;
+ if (child) {
+ loc_wipe(child);
+ }
+ return -1;
}
int
-dht_init_local_subvolumes (xlator_t *this, dht_conf_t *conf)
+dht_init_local_subvolumes(xlator_t *this, dht_conf_t *conf)
{
- xlator_list_t *subvols = NULL;
- int cnt = 0;
+ xlator_list_t *subvols = NULL;
+ int cnt = 0;
- if (!conf)
- return -1;
+ if (!conf)
+ return -1;
- for (subvols = this->children; subvols; subvols = subvols->next)
- cnt++;
+ for (subvols = this->children; subvols; subvols = subvols->next)
+ cnt++;
- conf->local_subvols = GF_CALLOC (cnt, sizeof (xlator_t *),
- gf_dht_mt_xlator_t);
+ conf->local_subvols = GF_CALLOC(cnt, sizeof(xlator_t *),
+ gf_dht_mt_xlator_t);
- /* FIX FIX : do this dynamically*/
- conf->local_nodeuuids = GF_CALLOC (cnt,
- sizeof (subvol_nodeuuids_info_t),
- gf_dht_nodeuuids_t);
+ /* FIX FIX : do this dynamically*/
+ conf->local_nodeuuids = GF_CALLOC(cnt, sizeof(subvol_nodeuuids_info_t),
+ gf_dht_nodeuuids_t);
- if (!conf->local_subvols || !conf->local_nodeuuids) {
- return -1;
- }
+ if (!conf->local_subvols || !conf->local_nodeuuids) {
+ return -1;
+ }
- conf->local_subvols_cnt = 0;
+ conf->local_subvols_cnt = 0;
- return 0;
+ return 0;
}
int
-dht_init_subvolumes (xlator_t *this, dht_conf_t *conf)
+dht_init_subvolumes(xlator_t *this, dht_conf_t *conf)
{
- xlator_list_t *subvols = NULL;
- int cnt = 0;
+ xlator_list_t *subvols = NULL;
+ int cnt = 0;
- if (!conf)
- return -1;
+ if (!conf)
+ return -1;
- for (subvols = this->children; subvols; subvols = subvols->next)
- cnt++;
+ for (subvols = this->children; subvols; subvols = subvols->next)
+ cnt++;
- conf->subvolumes = GF_CALLOC (cnt, sizeof (xlator_t *),
- gf_dht_mt_xlator_t);
- if (!conf->subvolumes) {
- return -1;
- }
- conf->subvolume_cnt = cnt;
+ conf->subvolumes = GF_CALLOC(cnt, sizeof(xlator_t *), gf_dht_mt_xlator_t);
+ if (!conf->subvolumes) {
+ return -1;
+ }
+ conf->subvolume_cnt = cnt;
+ /* Doesn't make sense to do any dht layer tasks
+ if the subvol count is 1. Set it as pass_through */
+ if (cnt == 1)
+ this->pass_through = _gf_true;
- conf->local_subvols_cnt = 0;
+ conf->local_subvols_cnt = 0;
- dht_set_subvol_range(this);
+ dht_set_subvol_range(this);
- cnt = 0;
- for (subvols = this->children; subvols; subvols = subvols->next)
- conf->subvolumes[cnt++] = subvols->xlator;
+ cnt = 0;
+ for (subvols = this->children; subvols; subvols = subvols->next)
+ conf->subvolumes[cnt++] = subvols->xlator;
- conf->subvolume_status = GF_CALLOC (cnt, sizeof (char),
- gf_dht_mt_char);
- if (!conf->subvolume_status) {
- return -1;
- }
+ conf->subvolume_status = GF_CALLOC(cnt, sizeof(char), gf_dht_mt_char);
+ if (!conf->subvolume_status) {
+ return -1;
+ }
- conf->last_event = GF_CALLOC (cnt, sizeof (int),
- gf_dht_mt_char);
- if (!conf->last_event) {
- return -1;
- }
+ conf->last_event = GF_CALLOC(cnt, sizeof(int), gf_dht_mt_char);
+ if (!conf->last_event) {
+ return -1;
+ }
- conf->subvol_up_time = GF_CALLOC (cnt, sizeof (time_t),
- gf_dht_mt_subvol_time);
- if (!conf->subvol_up_time) {
- return -1;
- }
+ conf->subvol_up_time = GF_CALLOC(cnt, sizeof(time_t),
+ gf_dht_mt_subvol_time);
+ if (!conf->subvol_up_time) {
+ return -1;
+ }
- conf->du_stats = GF_CALLOC (conf->subvolume_cnt, sizeof (dht_du_t),
- gf_dht_mt_dht_du_t);
- if (!conf->du_stats) {
- return -1;
- }
+ conf->du_stats = GF_CALLOC(conf->subvolume_cnt, sizeof(dht_du_t),
+ gf_dht_mt_dht_du_t);
+ if (!conf->du_stats) {
+ return -1;
+ }
- conf->decommissioned_bricks = GF_CALLOC (cnt, sizeof (xlator_t *),
- gf_dht_mt_xlator_t);
- if (!conf->decommissioned_bricks) {
- return -1;
- }
+ conf->decommissioned_bricks = GF_CALLOC(cnt, sizeof(xlator_t *),
+ gf_dht_mt_xlator_t);
+ if (!conf->decommissioned_bricks) {
+ return -1;
+ }
- return 0;
+ return 0;
}
-
/*
op_ret values :
0 : Success.
@@ -1265,270 +1238,274 @@ dht_init_subvolumes (xlator_t *this, dht_conf_t *conf)
*/
static int
-dht_migration_complete_check_done (int op_ret, call_frame_t *frame, void *data)
+dht_migration_complete_check_done(int op_ret, call_frame_t *frame, void *data)
{
- dht_local_t *local = NULL;
- xlator_t *subvol = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret != 0)
- goto out;
+ if (op_ret != 0)
+ goto out;
- if (local->cached_subvol == NULL) {
- local->op_errno = EINVAL;
- goto out;
- }
+ if (local->cached_subvol == NULL) {
+ local->op_errno = EINVAL;
+ goto out;
+ }
- subvol = local->cached_subvol;
+ subvol = local->cached_subvol;
out:
- local->rebalance.target_op_fn (THIS, subvol, frame, op_ret);
+ local->rebalance.target_op_fn(THIS, subvol, frame, op_ret);
- return 0;
+ return 0;
}
-
int
-dht_migration_complete_check_task (void *data)
+dht_migration_complete_check_task(void *data)
{
- int ret = -1;
- xlator_t *src_node = NULL;
- xlator_t *dst_node = NULL, *linkto_target = NULL;
- dht_local_t *local = NULL;
- dict_t *dict = NULL;
- struct iatt stbuf = {0,};
- xlator_t *this = NULL;
- call_frame_t *frame = NULL;
- loc_t tmp_loc = {0,};
- char *path = NULL;
- dht_conf_t *conf = NULL;
- inode_t *inode = NULL;
- fd_t *iter_fd = NULL;
- fd_t *tmp = NULL;
- uint64_t tmp_miginfo = 0;
- dht_migrate_info_t *miginfo = NULL;
- int open_failed = 0;
-
- this = THIS;
- frame = data;
- local = frame->local;
- conf = this->private;
-
- src_node = local->cached_subvol;
-
- if (!local->loc.inode && !local->fd) {
- local->op_errno = EINVAL;
- goto out;
- }
-
- inode = (!local->fd) ? local->loc.inode : local->fd->inode;
-
- /* getxattr on cached_subvol for 'linkto' value. Do path based getxattr
- * as root:root. If a fd is already open, access check wont be done*/
-
- if (!local->loc.inode) {
- ret = syncop_fgetxattr (src_node, local->fd, &dict,
- conf->link_xattr_name, NULL, NULL);
- } else {
- SYNCTASK_SETID (0, 0);
- ret = syncop_getxattr (src_node, &local->loc, &dict,
- conf->link_xattr_name, NULL, NULL);
- SYNCTASK_SETID (frame->root->uid, frame->root->gid);
- }
-
-
- /*
- * Each DHT xlator layer has its own name for the linkto xattr.
- * If the file mode bits indicate the the file is being migrated but
- * this layer's linkto xattr is not set, it means that another
- * DHT layer is migrating the file. In this case, return 1 so
- * the mode bits can be passed on to the higher layer for appropriate
- * action.
- */
- if (-ret == ENODATA) {
- /* This DHT translator is not migrating this file */
-
- ret = inode_ctx_reset1 (inode, this, &tmp_miginfo);
- if (tmp_miginfo) {
-
- /* This can be a problem if the file was
- * migrated by two different layers. Raise
- * a warning here.
- */
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_HAS_MIGINFO,
- "%s: Found miginfo in the inode ctx",
- tmp_loc.path ? tmp_loc.path :
- uuid_utoa (tmp_loc.gfid));
-
- miginfo = (void *)tmp_miginfo;
- GF_REF_PUT (miginfo);
- }
- ret = 1;
- goto out;
- }
+ int ret = -1;
+ xlator_t *src_node = NULL;
+ xlator_t *dst_node = NULL, *linkto_target = NULL;
+ dht_local_t *local = NULL;
+ dict_t *dict = NULL;
+ struct iatt stbuf = {
+ 0,
+ };
+ xlator_t *this = NULL;
+ call_frame_t *frame = NULL;
+ loc_t tmp_loc = {
+ 0,
+ };
+ char *path = NULL;
+ dht_conf_t *conf = NULL;
+ inode_t *inode = NULL;
+ fd_t *iter_fd = NULL;
+ fd_t *tmp = NULL;
+ uint64_t tmp_miginfo = 0;
+ dht_migrate_info_t *miginfo = NULL;
+ gf_boolean_t skip_open = _gf_false;
+ int open_failed = 0;
+
+ this = THIS;
+ frame = data;
+ local = frame->local;
+ conf = this->private;
+
+ src_node = local->cached_subvol;
+
+ if (!local->loc.inode && !local->fd) {
+ local->op_errno = EINVAL;
+ goto out;
+ }
- if (!ret)
- linkto_target = dht_linkfile_subvol (this, NULL, NULL, dict);
+ inode = (!local->fd) ? local->loc.inode : local->fd->inode;
- if (local->loc.inode) {
- loc_copy (&tmp_loc, &local->loc);
- } else {
- tmp_loc.inode = inode_ref (inode);
- gf_uuid_copy (tmp_loc.gfid, inode->gfid);
- }
+ /* getxattr on cached_subvol for 'linkto' value. Do path based getxattr
+ * as root:root. If a fd is already open, access check won't be done*/
- ret = syncop_lookup (this, &tmp_loc, &stbuf, 0, 0, 0);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_FILE_LOOKUP_FAILED,
- "%s: failed to lookup the file on %s",
- tmp_loc.path ? tmp_loc.path : uuid_utoa (tmp_loc.gfid),
- this->name);
- local->op_errno = -ret;
- ret = -1;
- goto out;
- }
+ if (!local->loc.inode) {
+ ret = syncop_fgetxattr(src_node, local->fd, &dict,
+ conf->link_xattr_name, NULL, NULL);
+ } else {
+ SYNCTASK_SETID(0, 0);
+ ret = syncop_getxattr(src_node, &local->loc, &dict,
+ conf->link_xattr_name, NULL, NULL);
+ SYNCTASK_SETID(frame->root->uid, frame->root->gid);
+ }
+
+ /*
+ * Each DHT xlator layer has its own name for the linkto xattr.
+ * If the file mode bits indicate the the file is being migrated but
+ * this layer's linkto xattr is not set, it means that another
+ * DHT layer is migrating the file. In this case, return 1 so
+ * the mode bits can be passed on to the higher layer for appropriate
+ * action.
+ */
+ if (-ret == ENODATA) {
+ /* This DHT translator is not migrating this file */
+
+ ret = inode_ctx_reset1(inode, this, &tmp_miginfo);
+ if (tmp_miginfo) {
+ /* This can be a problem if the file was
+ * migrated by two different layers. Raise
+ * a warning here.
+ */
+ gf_smsg(
+ this->name, GF_LOG_WARNING, 0, DHT_MSG_HAS_MIGINFO, "tmp=%s",
+ tmp_loc.path ? tmp_loc.path : uuid_utoa(tmp_loc.gfid), NULL);
+
+ miginfo = (void *)(uintptr_t)tmp_miginfo;
+ GF_REF_PUT(miginfo);
+ }
+ ret = 1;
+ goto out;
+ }
+
+ if (!ret)
+ linkto_target = dht_linkfile_subvol(this, NULL, NULL, dict);
+
+ if (local->loc.inode) {
+ loc_copy(&tmp_loc, &local->loc);
+ } else {
+ tmp_loc.inode = inode_ref(inode);
+ gf_uuid_copy(tmp_loc.gfid, inode->gfid);
+ }
+
+ ret = syncop_lookup(this, &tmp_loc, &stbuf, 0, 0, 0);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_FILE_LOOKUP_FAILED,
+ "tmp=%s", tmp_loc.path ? tmp_loc.path : uuid_utoa(tmp_loc.gfid),
+ "name=%s", this->name, NULL);
+ local->op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ dst_node = dht_subvol_get_cached(this, tmp_loc.inode);
+ if (linkto_target && dst_node != linkto_target) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_INVALID_LINKFILE,
+ "linkto_target_name=%s", linkto_target->name, "dst_name=%s",
+ dst_node->name, NULL);
+ }
+
+ if (gf_uuid_compare(stbuf.ia_gfid, tmp_loc.inode->gfid)) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_GFID_MISMATCH, "tmp=%s",
+ tmp_loc.path ? tmp_loc.path : uuid_utoa(tmp_loc.gfid),
+ "dst_name=%s", dst_node->name, NULL);
+ ret = -1;
+ local->op_errno = EIO;
+ goto out;
+ }
- dst_node = dht_subvol_get_cached (this, tmp_loc.inode);
- if (linkto_target && dst_node != linkto_target) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_INVALID_LINKFILE,
- "linkto target (%s) is "
- "different from cached-subvol (%s). Treating %s as "
- "destination subvol", linkto_target->name,
- dst_node->name, dst_node->name);
- }
+ /* update local. A layout is set in inode-ctx in lookup already */
- if (gf_uuid_compare (stbuf.ia_gfid, tmp_loc.inode->gfid)) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_GFID_MISMATCH,
- "%s: gfid different on the target file on %s",
- tmp_loc.path ? tmp_loc.path :
- uuid_utoa (tmp_loc.gfid), dst_node->name);
- ret = -1;
- local->op_errno = EIO;
- goto out;
- }
+ dht_layout_unref(this, local->layout);
- /* update local. A layout is set in inode-ctx in lookup already */
+ local->layout = dht_layout_get(frame->this, inode);
+ local->cached_subvol = dst_node;
- dht_layout_unref (this, local->layout);
+ ret = 0;
- local->layout = dht_layout_get (frame->this, inode);
- local->cached_subvol = dst_node;
+ /* once we detect the migration complete, the inode-ctx2 is no more
+ required.. delete the ctx and also, it means, open() already
+ done on all the fd of inode */
+ ret = inode_ctx_reset1(inode, this, &tmp_miginfo);
+ if (tmp_miginfo) {
+ miginfo = (void *)(uintptr_t)tmp_miginfo;
+ GF_REF_PUT(miginfo);
+ goto out;
+ }
+
+ /* perform 'open()' on all the fd's present on the inode */
+ if (tmp_loc.path == NULL) {
+ inode_path(inode, NULL, &path);
+ if (path)
+ tmp_loc.path = path;
+ }
+
+ LOCK(&inode->lock);
+
+ if (list_empty(&inode->fd_list))
+ goto unlock;
+
+ /* perform open as root:root. There is window between linkfile
+ * creation(root:root) and setattr with the correct uid/gid
+ */
+ SYNCTASK_SETID(0, 0);
+
+ /* It's possible that we are the last user of iter_fd after each
+ * iteration. In this case the fd_unref() of iter_fd at the end of
+ * the loop will cause the destruction of the fd. So we need to
+ * iterate the list safely because iter_fd cannot be trusted.
+ */
+ iter_fd = list_entry((&inode->fd_list)->next, typeof(*iter_fd), inode_list);
+ while (&iter_fd->inode_list != (&inode->fd_list)) {
+ if (fd_is_anonymous(iter_fd) ||
+ (dht_fd_open_on_dst(this, iter_fd, dst_node))) {
+ if (!tmp) {
+ iter_fd = list_entry(iter_fd->inode_list.next, typeof(*iter_fd),
+ inode_list);
+ continue;
+ }
+ skip_open = _gf_true;
+ }
+ /* We need to release the inode->lock before calling
+ * syncop_open() to avoid possible deadlocks. However this
+ * can cause the iter_fd to be released by other threads.
+ * To avoid this, we take a reference before releasing the
+ * lock.
+ */
+ fd_ref(iter_fd);
- ret = 0;
+ UNLOCK(&inode->lock);
- /* once we detect the migration complete, the inode-ctx2 is no more
- required.. delete the ctx and also, it means, open() already
- done on all the fd of inode */
- ret = inode_ctx_reset1 (inode, this, &tmp_miginfo);
- if (tmp_miginfo) {
- miginfo = (void *)tmp_miginfo;
- GF_REF_PUT (miginfo);
- goto out;
+ if (tmp) {
+ fd_unref(tmp);
+ tmp = NULL;
}
+ if (skip_open)
+ goto next;
- /* perform 'open()' on all the fd's present on the inode */
- if (tmp_loc.path == NULL) {
- inode_path (inode, NULL, &path);
- if (path)
- tmp_loc.path = path;
+ /* flags for open are stripped down to allow following the
+ * new location of the file, otherwise we can get EEXIST or
+ * truncate the file again as rebalance is moving the data */
+ ret = syncop_open(dst_node, &tmp_loc,
+ (iter_fd->flags & ~(O_CREAT | O_EXCL | O_TRUNC)),
+ iter_fd, NULL, NULL);
+ if (ret < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret,
+ DHT_MSG_OPEN_FD_ON_DST_FAILED, "id=%p", iter_fd,
+ "flags=0%o", iter_fd->flags, "path=%s", path, "name=%s",
+ dst_node->name, NULL);
+
+ open_failed = 1;
+ local->op_errno = -ret;
+ ret = -1;
+ } else {
+ dht_fd_ctx_set(this, iter_fd, dst_node);
}
+ next:
LOCK(&inode->lock);
+ skip_open = _gf_false;
+ tmp = iter_fd;
+ iter_fd = list_entry(tmp->inode_list.next, typeof(*tmp), inode_list);
+ }
- if (list_empty (&inode->fd_list))
- goto unlock;
-
- /* perform open as root:root. There is window between linkfile
- * creation(root:root) and setattr with the correct uid/gid
- */
- SYNCTASK_SETID(0, 0);
-
- /* It's possible that we are the last user of iter_fd after each
- * iteration. In this case the fd_unref() of iter_fd at the end of
- * the loop will cause the destruction of the fd. So we need to
- * iterate the list safely because iter_fd cannot be trusted.
- */
- list_for_each_entry_safe (iter_fd, tmp, &inode->fd_list, inode_list) {
-
- if (fd_is_anonymous (iter_fd))
- continue;
-
- if (dht_fd_open_on_dst (this, iter_fd, dst_node))
- continue;
-
- /* We need to release the inode->lock before calling
- * syncop_open() to avoid possible deadlocks. However this
- * can cause the iter_fd to be released by other threads.
- * To avoid this, we take a reference before releasing the
- * lock.
- */
- __fd_ref(iter_fd);
-
- UNLOCK(&inode->lock);
-
- /* flags for open are stripped down to allow following the
- * new location of the file, otherwise we can get EEXIST or
- * truncate the file again as rebalance is moving the data */
- ret = syncop_open (dst_node, &tmp_loc,
- (iter_fd->flags &
- ~(O_CREAT | O_EXCL | O_TRUNC)),
- iter_fd, NULL, NULL);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_OPEN_FD_ON_DST_FAILED, "failed"
- " to open the fd"
- " (%p, flags=0%o) on file %s @ %s",
- iter_fd, iter_fd->flags, path,
- dst_node->name);
-
- open_failed = 1;
- local->op_errno = -ret;
- ret = -1;
- } else {
- dht_fd_ctx_set (this, iter_fd, dst_node);
- }
+ SYNCTASK_SETID(frame->root->uid, frame->root->gid);
- fd_unref(iter_fd);
-
- LOCK(&inode->lock);
- }
-
- SYNCTASK_SETID (frame->root->uid, frame->root->gid);
-
- if (open_failed) {
- ret = -1;
- goto unlock;
- }
- ret = 0;
+ if (open_failed) {
+ ret = -1;
+ goto unlock;
+ }
+ ret = 0;
unlock:
- UNLOCK(&inode->lock);
+ UNLOCK(&inode->lock);
+ if (tmp) {
+ fd_unref(tmp);
+ tmp = NULL;
+ }
out:
- if (dict) {
- dict_unref (dict);
- }
+ if (dict) {
+ dict_unref(dict);
+ }
- loc_wipe (&tmp_loc);
+ loc_wipe(&tmp_loc);
- return ret;
+ return ret;
}
int
-dht_rebalance_complete_check (xlator_t *this, call_frame_t *frame)
+dht_rebalance_complete_check(xlator_t *this, call_frame_t *frame)
{
- int ret = -1;
-
- ret = synctask_new (this->ctx->env, dht_migration_complete_check_task,
- dht_migration_complete_check_done,
- frame, frame);
- return ret;
+ int ret = -1;
+ ret = synctask_new(this->ctx->env, dht_migration_complete_check_task,
+ dht_migration_complete_check_done, frame, frame);
+ return ret;
}
/* During 'in-progress' state, both nodes should have the file */
@@ -1539,785 +1516,789 @@ dht_rebalance_complete_check (xlator_t *this, call_frame_t *frame)
1 : File is being migrated but not by this DHT layer.
*/
static int
-dht_inprogress_check_done (int op_ret, call_frame_t *frame, void *data)
+dht_inprogress_check_done(int op_ret, call_frame_t *frame, void *data)
{
- dht_local_t *local = NULL;
- xlator_t *dst_subvol = NULL, *src_subvol = NULL;
- inode_t *inode = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *dst_subvol = NULL, *src_subvol = NULL;
+ inode_t *inode = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret != 0)
- goto out;
+ if (op_ret != 0)
+ goto out;
- inode = local->loc.inode ? local->loc.inode : local->fd->inode;
+ inode = local->loc.inode ? local->loc.inode : local->fd->inode;
- dht_inode_ctx_get_mig_info (THIS, inode, &src_subvol, &dst_subvol);
- if (dht_mig_info_is_invalid (local->cached_subvol,
- src_subvol, dst_subvol)) {
- dst_subvol = dht_subvol_get_cached (THIS, inode);
- if (!dst_subvol) {
- local->op_errno = EINVAL;
- goto out;
- }
+ dht_inode_ctx_get_mig_info(THIS, inode, &src_subvol, &dst_subvol);
+ if (dht_mig_info_is_invalid(local->cached_subvol, src_subvol, dst_subvol)) {
+ dst_subvol = dht_subvol_get_cached(THIS, inode);
+ if (!dst_subvol) {
+ local->op_errno = EINVAL;
+ goto out;
}
+ }
out:
- local->rebalance.target_op_fn (THIS, dst_subvol, frame, op_ret);
+ local->rebalance.target_op_fn(THIS, dst_subvol, frame, op_ret);
- return 0;
+ return 0;
}
static int
-dht_rebalance_inprogress_task (void *data)
+dht_rebalance_inprogress_task(void *data)
{
- int ret = -1;
- xlator_t *src_node = NULL;
- xlator_t *dst_node = NULL;
- dht_local_t *local = NULL;
- dict_t *dict = NULL;
- call_frame_t *frame = NULL;
- xlator_t *this = NULL;
- char *path = NULL;
- struct iatt stbuf = {0,};
- loc_t tmp_loc = {0,};
- dht_conf_t *conf = NULL;
- inode_t *inode = NULL;
- fd_t *iter_fd = NULL;
- fd_t *tmp = NULL;
- int open_failed = 0;
- uint64_t tmp_miginfo = 0;
- dht_migrate_info_t *miginfo = NULL;
-
-
- this = THIS;
- frame = data;
- local = frame->local;
- conf = this->private;
-
- src_node = local->cached_subvol;
-
- if (!local->loc.inode && !local->fd)
- goto out;
+ int ret = -1;
+ xlator_t *src_node = NULL;
+ xlator_t *dst_node = NULL;
+ dht_local_t *local = NULL;
+ dict_t *dict = NULL;
+ call_frame_t *frame = NULL;
+ xlator_t *this = NULL;
+ char *path = NULL;
+ struct iatt stbuf = {
+ 0,
+ };
+ loc_t tmp_loc = {
+ 0,
+ };
+ dht_conf_t *conf = NULL;
+ inode_t *inode = NULL;
+ fd_t *iter_fd = NULL;
+ fd_t *tmp = NULL;
+ int open_failed = 0;
+ uint64_t tmp_miginfo = 0;
+ dht_migrate_info_t *miginfo = NULL;
+ gf_boolean_t skip_open = _gf_false;
+
+ this = THIS;
+ frame = data;
+ local = frame->local;
+ conf = this->private;
+
+ src_node = local->cached_subvol;
+
+ if (!local->loc.inode && !local->fd)
+ goto out;
- inode = (!local->fd) ? local->loc.inode : local->fd->inode;
+ inode = (!local->fd) ? local->loc.inode : local->fd->inode;
- /* getxattr on cached_subvol for 'linkto' value. Do path based getxattr
- * as root:root. If a fd is already open, access check wont be done*/
- if (local->loc.inode) {
- SYNCTASK_SETID (0, 0);
- ret = syncop_getxattr (src_node, &local->loc, &dict,
- conf->link_xattr_name, NULL, NULL);
- SYNCTASK_SETID (frame->root->uid, frame->root->gid);
- } else {
- ret = syncop_fgetxattr (src_node, local->fd, &dict,
- conf->link_xattr_name, NULL, NULL);
- }
+ /* getxattr on cached_subvol for 'linkto' value. Do path based getxattr
+ * as root:root. If a fd is already open, access check won't be done*/
+ if (local->loc.inode) {
+ SYNCTASK_SETID(0, 0);
+ ret = syncop_getxattr(src_node, &local->loc, &dict,
+ conf->link_xattr_name, NULL, NULL);
+ SYNCTASK_SETID(frame->root->uid, frame->root->gid);
+ } else {
+ ret = syncop_fgetxattr(src_node, local->fd, &dict,
+ conf->link_xattr_name, NULL, NULL);
+ }
+
+ /*
+ * Each DHT xlator layer has its own name for the linkto xattr.
+ * If the file mode bits indicate the the file is being migrated but
+ * this layer's linkto xattr is not present, it means that another
+ * DHT layer is migrating the file. In this case, return 1 so
+ * the mode bits can be passed on to the higher layer for appropriate
+ * action.
+ */
+
+ if (-ret == ENODATA) {
+ /* This DHT layer is not migrating this file */
+ ret = inode_ctx_reset1(inode, this, &tmp_miginfo);
+ if (tmp_miginfo) {
+ /* This can be a problem if the file was
+ * migrated by two different layers. Raise
+ * a warning here.
+ */
+ gf_smsg(
+ this->name, GF_LOG_WARNING, 0, DHT_MSG_HAS_MIGINFO, "tmp=%s",
+ tmp_loc.path ? tmp_loc.path : uuid_utoa(tmp_loc.gfid), NULL);
+ miginfo = (void *)(uintptr_t)tmp_miginfo;
+ GF_REF_PUT(miginfo);
+ }
+ ret = 1;
+ goto out;
+ }
- /*
- * Each DHT xlator layer has its own name for the linkto xattr.
- * If the file mode bits indicate the the file is being migrated but
- * this layer's linkto xattr is not present, it means that another
- * DHT layer is migrating the file. In this case, return 1 so
- * the mode bits can be passed on to the higher layer for appropriate
- * action.
- */
+ if (ret < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_GET_XATTR_FAILED,
+ "path=%s", local->loc.path, NULL);
+ ret = -1;
+ goto out;
+ }
- if (-ret == ENODATA) {
- /* This DHT layer is not migrating this file */
- ret = inode_ctx_reset1 (inode, this, &tmp_miginfo);
- if (tmp_miginfo) {
- /* This can be a problem if the file was
- * migrated by two different layers. Raise
- * a warning here.
- */
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_HAS_MIGINFO,
- "%s: Found miginfo in the inode ctx",
- tmp_loc.path ? tmp_loc.path :
- uuid_utoa (tmp_loc.gfid));
- miginfo = (void *)tmp_miginfo;
- GF_REF_PUT (miginfo);
- }
- ret = 1;
- goto out;
- }
+ dst_node = dht_linkfile_subvol(this, NULL, NULL, dict);
+ if (!dst_node) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_GET_XATTR_FAILED,
+ "path=%s", local->loc.path, NULL);
+ ret = -1;
+ goto out;
+ }
+
+ local->rebalance.target_node = dst_node;
+
+ if (local->loc.inode) {
+ loc_copy(&tmp_loc, &local->loc);
+ } else {
+ tmp_loc.inode = inode_ref(inode);
+ gf_uuid_copy(tmp_loc.gfid, inode->gfid);
+ }
+
+ /* lookup on dst */
+ ret = syncop_lookup(dst_node, &tmp_loc, &stbuf, NULL, NULL, NULL);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_FILE_LOOKUP_FAILED,
+ "tmp=%s", tmp_loc.path ? tmp_loc.path : uuid_utoa(tmp_loc.gfid),
+ "name=%s", dst_node->name, NULL);
+ ret = -1;
+ goto out;
+ }
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_GET_XATTR_FAILED,
- "%s: failed to get the 'linkto' xattr",
- local->loc.path);
- ret = -1;
- goto out;
- }
+ if (gf_uuid_compare(stbuf.ia_gfid, tmp_loc.inode->gfid)) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_GFID_MISMATCH, "tmp=%s",
+ tmp_loc.path ? tmp_loc.path : uuid_utoa(tmp_loc.gfid),
+ "name=%s", dst_node->name, NULL);
+ ret = -1;
+ goto out;
+ }
+ ret = 0;
+
+ if (tmp_loc.path == NULL) {
+ inode_path(inode, NULL, &path);
+ if (path)
+ tmp_loc.path = path;
+ }
+
+ LOCK(&inode->lock);
+
+ if (list_empty(&inode->fd_list))
+ goto unlock;
+
+ /* perform open as root:root. There is window between linkfile
+ * creation(root:root) and setattr with the correct uid/gid
+ */
+ SYNCTASK_SETID(0, 0);
+
+ /* It's possible that we are the last user of iter_fd after each
+ * iteration. In this case the fd_unref() of iter_fd at the end of
+ * the loop will cause the destruction of the fd. So we need to
+ * iterate the list safely because iter_fd cannot be trusted.
+ */
+ iter_fd = list_entry((&inode->fd_list)->next, typeof(*iter_fd), inode_list);
+ while (&iter_fd->inode_list != (&inode->fd_list)) {
+ /* We need to release the inode->lock before calling
+ * syncop_open() to avoid possible deadlocks. However this
+ * can cause the iter_fd to be released by other threads.
+ * To avoid this, we take a reference before releasing the
+ * lock.
+ */
- dst_node = dht_linkfile_subvol (this, NULL, NULL, dict);
- if (!dst_node) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_SUBVOL_NOT_FOUND,
- "%s: failed to get the 'linkto' xattr from dict",
- local->loc.path);
- ret = -1;
- goto out;
+ if (fd_is_anonymous(iter_fd) ||
+ (dht_fd_open_on_dst(this, iter_fd, dst_node))) {
+ if (!tmp) {
+ iter_fd = list_entry(iter_fd->inode_list.next, typeof(*iter_fd),
+ inode_list);
+ continue;
+ }
+ skip_open = _gf_true;
}
- local->rebalance.target_node = dst_node;
-
- if (local->loc.inode) {
- loc_copy (&tmp_loc, &local->loc);
- } else {
- tmp_loc.inode = inode_ref (inode);
- gf_uuid_copy (tmp_loc.gfid, inode->gfid);
- }
+ /* Yes, this is ugly but there isn't a cleaner way to do this
+ * the fd_ref is an atomic increment so not too bad. We want to
+ * reduce the number of inode locks and unlocks.
+ */
- /* lookup on dst */
- ret = syncop_lookup (dst_node, &tmp_loc, &stbuf, NULL,
- NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_FILE_LOOKUP_ON_DST_FAILED,
- "%s: failed to lookup the file on %s",
- tmp_loc.path ? tmp_loc.path : uuid_utoa (tmp_loc.gfid),
- dst_node->name);
- ret = -1;
- goto out;
- }
+ fd_ref(iter_fd);
+ UNLOCK(&inode->lock);
- if (gf_uuid_compare (stbuf.ia_gfid, tmp_loc.inode->gfid)) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_GFID_MISMATCH,
- "%s: gfid different on the target file on %s",
- tmp_loc.path ? tmp_loc.path : uuid_utoa (tmp_loc.gfid),
- dst_node->name);
- ret = -1;
- goto out;
+ if (tmp) {
+ fd_unref(tmp);
+ tmp = NULL;
}
- ret = 0;
+ if (skip_open)
+ goto next;
- if (tmp_loc.path == NULL) {
- inode_path (inode, NULL, &path);
- if (path)
- tmp_loc.path = path;
+ /* flags for open are stripped down to allow following the
+ * new location of the file, otherwise we can get EEXIST or
+ * truncate the file again as rebalance is moving the data */
+ ret = syncop_open(dst_node, &tmp_loc,
+ (iter_fd->flags & ~(O_CREAT | O_EXCL | O_TRUNC)),
+ iter_fd, NULL, NULL);
+ if (ret < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret,
+ DHT_MSG_OPEN_FD_ON_DST_FAILED, "fd=%p", iter_fd,
+ "flags=0%o", iter_fd->flags, "path=%s", path, "name=%s",
+ dst_node->name, NULL);
+ ret = -1;
+ open_failed = 1;
+ } else {
+ /* Potential fd leak if this fails here as it will be
+ reopened at the next Phase1/2 check */
+ dht_fd_ctx_set(this, iter_fd, dst_node);
}
+ next:
LOCK(&inode->lock);
+ skip_open = _gf_false;
+ tmp = iter_fd;
+ iter_fd = list_entry(tmp->inode_list.next, typeof(*tmp), inode_list);
+ }
- if (list_empty (&inode->fd_list))
- goto unlock;
-
- /* perform open as root:root. There is window between linkfile
- * creation(root:root) and setattr with the correct uid/gid
- */
- SYNCTASK_SETID (0, 0);
-
- /* It's possible that we are the last user of iter_fd after each
- * iteration. In this case the fd_unref() of iter_fd at the end of
- * the loop will cause the destruction of the fd. So we need to
- * iterate the list safely because iter_fd cannot be trusted.
- */
- list_for_each_entry_safe (iter_fd, tmp, &inode->fd_list, inode_list) {
- if (fd_is_anonymous (iter_fd))
- continue;
-
- if (dht_fd_open_on_dst (this, iter_fd, dst_node))
- continue;
-
- /* We need to release the inode->lock before calling
- * syncop_open() to avoid possible deadlocks. However this
- * can cause the iter_fd to be released by other threads.
- * To avoid this, we take a reference before releasing the
- * lock.
- */
- __fd_ref(iter_fd);
-
- UNLOCK(&inode->lock);
-
- /* flags for open are stripped down to allow following the
- * new location of the file, otherwise we can get EEXIST or
- * truncate the file again as rebalance is moving the data */
- ret = syncop_open (dst_node, &tmp_loc,
- (iter_fd->flags &
- ~(O_CREAT | O_EXCL | O_TRUNC)),
- iter_fd, NULL, NULL);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_OPEN_FD_ON_DST_FAILED,
- "failed to send open "
- "the fd (%p, flags=0%o) on file %s @ %s",
- iter_fd, iter_fd->flags, path,
- dst_node->name);
- ret = -1;
- open_failed = 1;
- } else {
- /* Potential fd leak if this fails here as it will be
- reopened at the next Phase1/2 check */
- dht_fd_ctx_set (this, iter_fd, dst_node);
- }
-
- fd_unref(iter_fd);
-
- LOCK(&inode->lock);
- }
-
- SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+ SYNCTASK_SETID(frame->root->uid, frame->root->gid);
unlock:
- UNLOCK(&inode->lock);
-
- if (open_failed) {
- ret = -1;
- goto out;
- }
+ UNLOCK(&inode->lock);
+
+ if (tmp) {
+ fd_unref(tmp);
+ tmp = NULL;
+ }
+ if (open_failed) {
+ ret = -1;
+ goto out;
+ }
- ret = dht_inode_ctx_set_mig_info (this, inode, src_node, dst_node);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_SET_INODE_CTX_FAILED,
- "%s: failed to set inode-ctx target file at %s",
- local->loc.path, dst_node->name);
- goto out;
- }
+ ret = dht_inode_ctx_set_mig_info(this, inode, src_node, dst_node);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_SET_INODE_CTX_FAILED,
+ "path=%s", local->loc.path, "name=%s", dst_node->name, NULL);
+ goto out;
+ }
- ret = 0;
+ ret = 0;
out:
- if (dict) {
- dict_unref (dict);
- }
+ if (dict) {
+ dict_unref(dict);
+ }
- loc_wipe (&tmp_loc);
- return ret;
+ loc_wipe(&tmp_loc);
+ return ret;
}
int
-dht_rebalance_in_progress_check (xlator_t *this, call_frame_t *frame)
+dht_rebalance_in_progress_check(xlator_t *this, call_frame_t *frame)
{
+ int ret = -1;
- int ret = -1;
-
- ret = synctask_new (this->ctx->env, dht_rebalance_inprogress_task,
- dht_inprogress_check_done,
- frame, frame);
- return ret;
+ ret = synctask_new(this->ctx->env, dht_rebalance_inprogress_task,
+ dht_inprogress_check_done, frame, frame);
+ return ret;
}
int
-dht_inode_ctx_layout_set (inode_t *inode, xlator_t *this,
- dht_layout_t *layout_int)
+dht_inode_ctx_layout_set(inode_t *inode, xlator_t *this,
+ dht_layout_t *layout_int)
{
- dht_inode_ctx_t *ctx = NULL;
- int ret = -1;
-
- ret = dht_inode_ctx_get (inode, this, &ctx);
- if (!ret && ctx) {
- ctx->layout = layout_int;
- } else {
- ctx = GF_CALLOC (1, sizeof (*ctx), gf_dht_mt_inode_ctx_t);
- if (!ctx)
- return ret;
- ctx->layout = layout_int;
- }
-
- ret = dht_inode_ctx_set (inode, this, ctx);
-
- return ret;
+ dht_inode_ctx_t *ctx = NULL;
+ int ret = -1;
+
+ ret = dht_inode_ctx_get(inode, this, &ctx);
+ if (!ret && ctx) {
+ ctx->layout = layout_int;
+ } else {
+ ctx = GF_CALLOC(1, sizeof(*ctx), gf_dht_mt_inode_ctx_t);
+ if (!ctx)
+ return ret;
+ ctx->layout = layout_int;
+ }
+
+ ret = dht_inode_ctx_set(inode, this, ctx);
+
+ return ret;
}
-
void
-dht_inode_ctx_time_set (inode_t *inode, xlator_t *this, struct iatt *stat)
+dht_inode_ctx_time_set(inode_t *inode, xlator_t *this, struct iatt *stat)
{
- dht_inode_ctx_t *ctx = NULL;
- dht_stat_time_t *time = 0;
- int ret = -1;
+ dht_inode_ctx_t *ctx = NULL;
+ dht_stat_time_t *time = 0;
+ int ret = -1;
- ret = dht_inode_ctx_get (inode, this, &ctx);
+ ret = dht_inode_ctx_get(inode, this, &ctx);
- if (ret)
- return;
+ if (ret)
+ return;
- time = &ctx->time;
+ time = &ctx->time;
- time->mtime = stat->ia_mtime;
- time->mtime_nsec = stat->ia_mtime_nsec;
+ time->mtime = stat->ia_mtime;
+ time->mtime_nsec = stat->ia_mtime_nsec;
- time->ctime = stat->ia_ctime;
- time->ctime_nsec = stat->ia_ctime_nsec;
+ time->ctime = stat->ia_ctime;
+ time->ctime_nsec = stat->ia_ctime_nsec;
- time->atime = stat->ia_atime;
- time->atime_nsec = stat->ia_atime_nsec;
+ time->atime = stat->ia_atime;
+ time->atime_nsec = stat->ia_atime_nsec;
- return;
+ return;
}
-
int
-dht_inode_ctx_time_update (inode_t *inode, xlator_t *this, struct iatt *stat,
- int32_t post)
+dht_inode_ctx_time_update(inode_t *inode, xlator_t *this, struct iatt *stat,
+ int32_t post)
{
- dht_inode_ctx_t *ctx = NULL;
- dht_stat_time_t *time = 0;
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO (this->name, stat, out);
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
-
- ret = dht_inode_ctx_get (inode, this, &ctx);
-
- if (ret) {
- ctx = GF_CALLOC (1, sizeof (*ctx), gf_dht_mt_inode_ctx_t);
- if (!ctx)
- return -1;
- }
-
- time = &ctx->time;
-
- DHT_UPDATE_TIME(time->mtime, time->mtime_nsec,
- stat->ia_mtime, stat->ia_mtime_nsec, inode, post);
- DHT_UPDATE_TIME(time->ctime, time->ctime_nsec,
- stat->ia_ctime, stat->ia_ctime_nsec, inode, post);
- DHT_UPDATE_TIME(time->atime, time->atime_nsec,
- stat->ia_atime, stat->ia_atime_nsec, inode, post);
-
- ret = dht_inode_ctx_set (inode, this, ctx);
+ dht_inode_ctx_t *ctx = NULL;
+ dht_stat_time_t *time = 0;
+ int ret = -1;
+
+ GF_VALIDATE_OR_GOTO(this->name, stat, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
+
+ ret = dht_inode_ctx_get(inode, this, &ctx);
+
+ if (ret) {
+ ctx = GF_CALLOC(1, sizeof(*ctx), gf_dht_mt_inode_ctx_t);
+ if (!ctx)
+ return -1;
+ }
+
+ time = &ctx->time;
+
+ LOCK(&inode->lock);
+ {
+ DHT_UPDATE_TIME(time->mtime, time->mtime_nsec, stat->ia_mtime,
+ stat->ia_mtime_nsec, post);
+ DHT_UPDATE_TIME(time->ctime, time->ctime_nsec, stat->ia_ctime,
+ stat->ia_ctime_nsec, post);
+ DHT_UPDATE_TIME(time->atime, time->atime_nsec, stat->ia_atime,
+ stat->ia_atime_nsec, post);
+ }
+ UNLOCK(&inode->lock);
+
+ ret = dht_inode_ctx_set(inode, this, ctx);
out:
- return 0;
+ return 0;
}
int
-dht_inode_ctx_get (inode_t *inode, xlator_t *this, dht_inode_ctx_t **ctx)
+dht_inode_ctx_get(inode_t *inode, xlator_t *this, dht_inode_ctx_t **ctx)
{
- int ret = -1;
- uint64_t ctx_int = 0;
+ int ret = -1;
+ uint64_t ctx_int = 0;
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- ret = inode_ctx_get (inode, this, &ctx_int);
+ ret = inode_ctx_get(inode, this, &ctx_int);
- if (ret)
- return ret;
+ if (ret)
+ return ret;
- if (ctx)
- *ctx = (dht_inode_ctx_t *) ctx_int;
+ if (ctx)
+ *ctx = (dht_inode_ctx_t *)(uintptr_t)ctx_int;
out:
- return ret;
+ return ret;
}
-int dht_inode_ctx_set (inode_t *inode, xlator_t *this, dht_inode_ctx_t *ctx)
+int
+dht_inode_ctx_set(inode_t *inode, xlator_t *this, dht_inode_ctx_t *ctx)
{
- int ret = -1;
- uint64_t ctx_int = 0;
+ int ret = -1;
+ uint64_t ctx_int = 0;
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
- GF_VALIDATE_OR_GOTO (this->name, ctx, out);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, ctx, out);
- ctx_int = (long)ctx;
- ret = inode_ctx_set (inode, this, &ctx_int);
+ ctx_int = (long)ctx;
+ ret = inode_ctx_set(inode, this, &ctx_int);
out:
- return ret;
+ return ret;
}
int
-dht_subvol_status (dht_conf_t *conf, xlator_t *subvol)
+dht_subvol_status(dht_conf_t *conf, xlator_t *subvol)
{
- int i;
+ int i;
- for (i=0 ; i < conf->subvolume_cnt; i++) {
- if (conf->subvolumes[i] == subvol) {
- return conf->subvolume_status[i];
- }
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->subvolumes[i] == subvol) {
+ return conf->subvolume_status[i];
}
- return 0;
+ }
+ return 0;
}
-inode_t*
-dht_heal_path (xlator_t *this, char *path, inode_table_t *itable)
+inode_t *
+dht_heal_path(xlator_t *this, char *path, inode_table_t *itable)
{
- int ret = -1;
- struct iatt iatt = {0, };
- inode_t *linked_inode = NULL;
- loc_t loc = {0, };
- char *bname = NULL;
- char *save_ptr = NULL;
- uuid_t gfid = {0, };
- char *tmp_path = NULL;
-
-
- tmp_path = gf_strdup (path);
- if (!tmp_path) {
- goto out;
- }
-
- memset (gfid, 0, 16);
- gfid[15] = 1;
+ int ret = -1;
+ struct iatt iatt = {
+ 0,
+ };
+ inode_t *linked_inode = NULL;
+ loc_t loc = {
+ 0,
+ };
+ char *bname = NULL;
+ char *save_ptr = NULL;
+ static uuid_t gfid = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1};
+ char *tmp_path = NULL;
+
+ tmp_path = gf_strdup(path);
+ if (!tmp_path) {
+ goto out;
+ }
- gf_uuid_copy (loc.pargfid, gfid);
- loc.parent = inode_ref (itable->root);
+ gf_uuid_copy(loc.pargfid, gfid);
+ loc.parent = inode_ref(itable->root);
- bname = strtok_r (tmp_path, "/", &save_ptr);
+ bname = strtok_r(tmp_path, "/", &save_ptr);
- /* sending a lookup on parent directory,
- * Eg: if path is like /a/b/c/d/e/f/g/
- * then we will send a lookup on a first and then b,c,d,etc
- */
+ /* sending a lookup on parent directory,
+ * Eg: if path is like /a/b/c/d/e/f/g/
+ * then we will send a lookup on a first and then b,c,d,etc
+ */
- while (bname) {
- linked_inode = NULL;
- loc.inode = inode_grep (itable, loc.parent, bname);
- if (loc.inode == NULL) {
- loc.inode = inode_new (itable);
- if (loc.inode == NULL) {
- ret = -ENOMEM;
- goto out;
- }
- } else {
- /*
- * Inode is already populated in the inode table.
- * Which means we already looked up the inde and
- * linked with a dentry. So that we will skip
- * lookup on this entry, and proceed to next.
- */
- bname = strtok_r (NULL, "/", &save_ptr);
- inode_unref (loc.parent);
- loc.parent = loc.inode;
- gf_uuid_copy (loc.pargfid, loc.inode->gfid);
- loc.inode = NULL;
- continue;
- }
+ while (bname) {
+ linked_inode = NULL;
+ loc.inode = inode_grep(itable, loc.parent, bname);
+ if (loc.inode == NULL) {
+ loc.inode = inode_new(itable);
+ if (loc.inode == NULL) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ } else {
+ /*
+ * Inode is already populated in the inode table.
+ * Which means we already looked up the inode and
+ * linked with a dentry. So that we will skip
+ * lookup on this entry, and proceed to next.
+ */
+ linked_inode = loc.inode;
+ bname = strtok_r(NULL, "/", &save_ptr);
+ if (!bname) {
+ goto out;
+ }
+ inode_unref(loc.parent);
+ loc.parent = loc.inode;
+ gf_uuid_copy(loc.pargfid, loc.inode->gfid);
+ loc.inode = NULL;
+ continue;
+ }
- loc.name = bname;
- ret = loc_path (&loc, bname);
+ loc.name = bname;
+ ret = loc_path(&loc, bname);
- ret = syncop_lookup (this, &loc, &iatt, NULL, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_INFO, -ret,
- DHT_MSG_DIR_SELFHEAL_FAILED,
- "Healing of path %s failed on subvolume %s for "
- "directory %s", path, this->name, bname);
- goto out;
- }
+ ret = syncop_lookup(this, &loc, &iatt, NULL, NULL, NULL);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_INFO, -ret, DHT_MSG_DIR_SELFHEAL_FAILED,
+ "path=%s", path, "subvolume=%s", this->name, "bname=%s",
+ bname, NULL);
+ goto out;
+ }
- linked_inode = inode_link (loc.inode, loc.parent, bname, &iatt);
- if (!linked_inode)
- goto out;
+ linked_inode = inode_link(loc.inode, loc.parent, bname, &iatt);
+ if (!linked_inode)
+ goto out;
- loc_wipe (&loc);
- gf_uuid_copy (loc.pargfid, linked_inode->gfid);
- loc.inode = NULL;
- loc.parent = linked_inode;
+ loc_wipe(&loc);
+ gf_uuid_copy(loc.pargfid, linked_inode->gfid);
+ loc.inode = NULL;
- bname = strtok_r (NULL, "/", &save_ptr);
- }
+ bname = strtok_r(NULL, "/", &save_ptr);
+ if (bname)
+ loc.parent = linked_inode;
+ }
out:
- inode_ref (linked_inode);
- loc_wipe (&loc);
- GF_FREE (tmp_path);
+ inode_ref(linked_inode);
+ loc_wipe(&loc);
+ GF_FREE(tmp_path);
- return linked_inode;
+ return linked_inode;
}
-
int
-dht_heal_full_path (void *data)
+dht_heal_full_path(void *data)
{
- call_frame_t *heal_frame = data;
- dht_local_t *local = NULL;
- loc_t loc = {0, };
- dict_t *dict = NULL;
- char *path = NULL;
- int ret = -1;
- xlator_t *source = NULL;
- xlator_t *this = NULL;
- inode_table_t *itable = NULL;
- inode_t *inode = NULL;
- inode_t *tmp_inode = NULL;
-
- GF_VALIDATE_OR_GOTO ("DHT", heal_frame, out);
-
- local = heal_frame->local;
- this = heal_frame->this;
- source = heal_frame->cookie;
- heal_frame->cookie = NULL;
- gf_uuid_copy (loc.gfid, local->gfid);
-
- if (local->loc.inode)
- loc.inode = inode_ref (local->loc.inode);
- else
- goto out;
-
- itable = loc.inode->table;
- ret = syncop_getxattr (source, &loc, &dict,
- GET_ANCESTRY_PATH_KEY, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_INFO, -ret,
- DHT_MSG_DIR_SELFHEAL_FAILED,
- "Failed to get path from subvol %s. Aborting "
- "directory healing.", source->name);
- goto out;
- }
+ call_frame_t *heal_frame = data;
+ dht_local_t *local = NULL;
+ loc_t loc = {
+ 0,
+ };
+ dict_t *dict = NULL;
+ char *path = NULL;
+ int ret = -1;
+ xlator_t *source = NULL;
+ xlator_t *this = NULL;
+ inode_table_t *itable = NULL;
+ inode_t *inode = NULL;
+ inode_t *tmp_inode = NULL;
+
+ GF_VALIDATE_OR_GOTO("DHT", heal_frame, out);
+
+ local = heal_frame->local;
+ this = heal_frame->this;
+ source = heal_frame->cookie;
+ heal_frame->cookie = NULL;
+ gf_uuid_copy(loc.gfid, local->gfid);
+
+ if (local->loc.inode)
+ loc.inode = inode_ref(local->loc.inode);
+ else
+ goto out;
- ret = dict_get_str (dict, GET_ANCESTRY_PATH_KEY, &path);
- if (path) {
- inode = dht_heal_path (this, path, itable);
- if (inode && inode != local->inode) {
- /*
- * if inode returned by heal function is different
- * from what we passed, which means a racing thread
- * already linked a different inode for dentry.
- * So we will update our local->inode, so that we can
- * retrurn proper inode.
- */
- tmp_inode = local->inode;
- local->inode = inode;
- inode_unref (tmp_inode);
- tmp_inode = NULL;
- } else {
- inode_unref (inode);
- }
+ itable = loc.inode->table;
+ ret = syncop_getxattr(source, &loc, &dict, GET_ANCESTRY_PATH_KEY, NULL,
+ NULL);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_INFO, -ret, DHT_MSG_DIR_HEAL_ABORT,
+ "subvol=%s", source->name, NULL);
+ goto out;
+ }
+
+ ret = dict_get_str(dict, GET_ANCESTRY_PATH_KEY, &path);
+ if (path) {
+ inode = dht_heal_path(this, path, itable);
+ if (inode && inode != local->inode) {
+ /*
+ * if inode returned by heal function is different
+ * from what we passed, which means a racing thread
+ * already linked a different inode for dentry.
+ * So we will update our local->inode, so that we can
+ * retrurn proper inode.
+ */
+ tmp_inode = local->inode;
+ local->inode = inode;
+ inode_unref(tmp_inode);
+ tmp_inode = NULL;
+ } else {
+ inode_unref(inode);
}
+ }
out:
- loc_wipe (&loc);
- if (dict)
- dict_unref (dict);
- return 0;
+ loc_wipe(&loc);
+ if (dict)
+ dict_unref(dict);
+ return 0;
}
int
-dht_heal_full_path_done (int op_ret, call_frame_t *heal_frame, void *data)
+dht_heal_full_path_done(int op_ret, call_frame_t *heal_frame, void *data)
{
-
- call_frame_t *main_frame = NULL;
- dht_local_t *local = NULL;
- xlator_t *this = NULL;
- int ret = -1;
-
- local = heal_frame->local;
- main_frame = local->main_frame;
- local->main_frame = NULL;
- this = heal_frame->this;
-
- dht_set_fixed_dir_stat (&local->postparent);
- if (local->need_xattr_heal) {
- local->need_xattr_heal = 0;
- ret = dht_dir_xattr_heal (this, local);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, ret,
- DHT_MSG_DIR_XATTR_HEAL_FAILED,
- "xattr heal failed for directory %s ",
- local->loc.path);
+ call_frame_t *main_frame = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *this = NULL;
+ int ret = -1;
+ int op_errno = 0;
+
+ local = heal_frame->local;
+ main_frame = local->main_frame;
+ local->main_frame = NULL;
+ this = heal_frame->this;
+
+ dht_set_fixed_dir_stat(&local->postparent);
+ if (local->need_xattr_heal) {
+ local->need_xattr_heal = 0;
+ ret = dht_dir_xattr_heal(this, local, &op_errno);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno,
+ DHT_MSG_DIR_XATTR_HEAL_FAILED, "path=%s", local->loc.path,
+ NULL);
}
+ }
- DHT_STACK_UNWIND (lookup, main_frame, 0, 0,
- local->inode, &local->stbuf, local->xattr,
- &local->postparent);
+ DHT_STACK_UNWIND(lookup, main_frame, 0, 0, local->inode, &local->stbuf,
+ local->xattr, &local->postparent);
- DHT_STACK_DESTROY (heal_frame);
- return 0;
+ DHT_STACK_DESTROY(heal_frame);
+ return 0;
}
/* This function must be called inside an inode lock */
int
-__dht_lock_subvol_set (inode_t *inode, xlator_t *this,
- xlator_t *lock_subvol)
+__dht_lock_subvol_set(inode_t *inode, xlator_t *this, xlator_t *lock_subvol)
{
- dht_inode_ctx_t *ctx = NULL;
- int ret = -1;
- uint64_t value = 0;
+ dht_inode_ctx_t *ctx = NULL;
+ int ret = -1;
+ uint64_t value = 0;
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- ret = __inode_ctx_get0 (inode, this, &value);
- if (ret || !value) {
- return -1;
- }
+ ret = __inode_ctx_get0(inode, this, &value);
+ if (ret || !value) {
+ return -1;
+ }
- ctx = (dht_inode_ctx_t *) value;
- ctx->lock_subvol = lock_subvol;
+ ctx = (dht_inode_ctx_t *)(uintptr_t)value;
+ ctx->lock_subvol = lock_subvol;
out:
- return ret;
+ return ret;
}
-xlator_t*
-dht_get_lock_subvolume (xlator_t *this, struct gf_flock *lock,
- dht_local_t *local)
+xlator_t *
+dht_get_lock_subvolume(xlator_t *this, struct gf_flock *lock,
+ dht_local_t *local)
{
- xlator_t *subvol = NULL;
- inode_t *inode = NULL;
- int32_t ret = -1;
- uint64_t value = 0;
- xlator_t *cached_subvol = NULL;
- dht_inode_ctx_t *ctx = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- GF_VALIDATE_OR_GOTO (this->name, lock, out);
- GF_VALIDATE_OR_GOTO (this->name, local, out);
+ xlator_t *subvol = NULL;
+ inode_t *inode = NULL;
+ int32_t ret = -1;
+ uint64_t value = 0;
+ xlator_t *cached_subvol = NULL;
+ dht_inode_ctx_t *ctx = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- cached_subvol = local->cached_subvol;
+ GF_VALIDATE_OR_GOTO(this->name, lock, out);
+ GF_VALIDATE_OR_GOTO(this->name, local, out);
- if (local->loc.inode || local->fd) {
- inode = local->loc.inode ? local->loc.inode : local->fd->inode;
- }
+ cached_subvol = local->cached_subvol;
- if (!inode)
- goto out;
+ if (local->loc.inode || local->fd) {
+ inode = local->loc.inode ? local->loc.inode : local->fd->inode;
+ }
- if (!(IA_ISDIR (inode->ia_type) || IA_ISINVAL (inode->ia_type))) {
- /*
- * We may get non-linked inode for directories as part
- * of the selfheal code path. So checking for IA_INVAL
- * type also. This will only happen for directory.
- */
- subvol = local->cached_subvol;
- goto out;
- }
+ if (!inode)
+ goto out;
- if (lock->l_type != F_UNLCK) {
- /*
- * inode purging might happen on NFS between a lk
- * and unlk. Due to this lk and unlk might be sent
- * to different subvols.
- * So during a lock request, taking a ref on inode
- * to prevent inode purging. inode unref will happen
- * in unlock cbk code path.
- */
- inode_ref (inode);
- }
+ if (!(IA_ISDIR(inode->ia_type) || IA_ISINVAL(inode->ia_type))) {
+ /*
+ * We may get non-linked inode for directories as part
+ * of the selfheal code path. So checking for IA_INVAL
+ * type also. This will only happen for directory.
+ */
+ subvol = local->cached_subvol;
+ goto out;
+ }
- LOCK (&inode->lock);
- ret = __inode_ctx_get0 (inode, this, &value);
- if (!ret && value) {
- ctx = (dht_inode_ctx_t *) value;
- subvol = ctx->lock_subvol;
- }
- if (!subvol && lock->l_type != F_UNLCK && cached_subvol) {
- ret = __dht_lock_subvol_set (inode, this,
- cached_subvol);
- if (ret) {
- gf_uuid_unparse(inode->gfid, gfid);
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_SET_INODE_CTX_FAILED,
- "Failed to set lock_subvol in "
- "inode ctx for gfid %s",
- gfid);
- goto unlock;
- }
- subvol = cached_subvol;
- }
-unlock:
- UNLOCK (&inode->lock);
- if (!subvol && inode && lock->l_type != F_UNLCK) {
- inode_unref (inode);
- }
+ if (lock->l_type != F_UNLCK) {
+ /*
+ * inode purging might happen on NFS between a lk
+ * and unlk. Due to this lk and unlk might be sent
+ * to different subvols.
+ * So during a lock request, taking a ref on inode
+ * to prevent inode purging. inode unref will happen
+ * in unlock cbk code path.
+ */
+ inode_ref(inode);
+ }
+
+ LOCK(&inode->lock);
+ ret = __inode_ctx_get0(inode, this, &value);
+ if (!ret && value) {
+ ctx = (dht_inode_ctx_t *)(uintptr_t)value;
+ subvol = ctx->lock_subvol;
+ }
+ if (!subvol && lock->l_type != F_UNLCK && cached_subvol) {
+ ret = __dht_lock_subvol_set(inode, this, cached_subvol);
+ if (ret) {
+ gf_uuid_unparse(inode->gfid, gfid);
+ UNLOCK(&inode->lock);
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_SET_INODE_CTX_FAILED,
+ "lock_subvol gfid=%s", gfid, NULL);
+ goto post_unlock;
+ }
+ subvol = cached_subvol;
+ }
+ UNLOCK(&inode->lock);
+post_unlock:
+ if (!subvol && inode && lock->l_type != F_UNLCK) {
+ inode_unref(inode);
+ }
out:
- return subvol;
+ return subvol;
}
int
-dht_lk_inode_unref (call_frame_t *frame, int32_t op_ret)
+dht_lk_inode_unref(call_frame_t *frame, int32_t op_ret)
{
- int ret = -1;
- dht_local_t *local = NULL;
- inode_t *inode = NULL;
- xlator_t *this = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
+ int ret = -1;
+ dht_local_t *local = NULL;
+ inode_t *inode = NULL;
+ xlator_t *this = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- local = frame->local;
- this = frame->this;
+ local = frame->local;
+ this = frame->this;
- if (local->loc.inode || local->fd) {
- inode = local->loc.inode ? local->loc.inode : local->fd->inode;
- }
- if (!inode) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_LOCK_INODE_UNREF_FAILED,
- "Found a NULL inode. Failed to unref the inode");
- goto out;
- }
+ if (local->loc.inode || local->fd) {
+ inode = local->loc.inode ? local->loc.inode : local->fd->inode;
+ }
+ if (!inode) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_LOCK_INODE_UNREF_FAILED,
+ NULL);
+ goto out;
+ }
- if (!(IA_ISDIR (inode->ia_type) || IA_ISINVAL (inode->ia_type))) {
- ret = 0;
- goto out;
- }
+ if (!(IA_ISDIR(inode->ia_type) || IA_ISINVAL(inode->ia_type))) {
+ ret = 0;
+ goto out;
+ }
- switch (local->lock_type) {
+ switch (local->lock_type) {
case F_RDLCK:
case F_WRLCK:
- if (op_ret) {
- gf_uuid_unparse(inode->gfid, gfid);
- gf_msg_debug (this->name, 0,
- "lock request failed for gfid %s", gfid);
- inode_unref (inode);
- goto out;
- }
- break;
+ if (op_ret) {
+ gf_uuid_unparse(inode->gfid, gfid);
+ gf_msg_debug(this->name, 0, "lock request failed for gfid %s",
+ gfid);
+ inode_unref(inode);
+ goto out;
+ }
+ break;
case F_UNLCK:
- if (!op_ret) {
- inode_unref (inode);
- } else {
- gf_uuid_unparse(inode->gfid, gfid);
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_LOCK_INODE_UNREF_FAILED,
- "Unlock request failed for gfid %s."
- "Failed to unref the inode", gfid);
- goto out;
- }
+ if (!op_ret) {
+ inode_unref(inode);
+ } else {
+ gf_uuid_unparse(inode->gfid, gfid);
+ gf_smsg(this->name, GF_LOG_WARNING, 0,
+ DHT_MSG_LOCK_INODE_UNREF_FAILED, "gfid=%s", gfid, NULL);
+ goto out;
+ }
default:
- break;
- }
- ret = 0;
+ break;
+ }
+ ret = 0;
out:
- return ret;
+ return ret;
}
/* Code to update custom extended attributes from src dict to dst dict
-*/
+ */
void
-dht_dir_set_heal_xattr (xlator_t *this, dht_local_t *local, dict_t *dst,
- dict_t *src, int *uret, int *uflag)
+dht_dir_set_heal_xattr(xlator_t *this, dht_local_t *local, dict_t *dst,
+ dict_t *src, int *uret, int *uflag)
{
- int ret = -1;
- data_t *keyval = NULL;
- int luret = -1;
- int luflag = -1;
- int i = 0;
-
- if (!src || !dst) {
- gf_msg (this->name, GF_LOG_WARNING, EINVAL,
- DHT_MSG_DICT_SET_FAILED,
- "src or dst is NULL. Failed to set "
- " dictionary value for path %s",
- local->loc.path);
- return;
- }
- /* Check if any user xattr present in src dict and set
- it to dst dict
- */
- luret = dict_foreach_fnmatch (src, "user.*",
- dht_set_user_xattr, dst);
- /* Check if any other custom xattr present in src dict
- and set it to dst dict, here index start from 1 because
- user xattr already checked in previous statement
- */
- for (i = 1; xattrs_to_heal[i]; i++) {
- keyval = dict_get (src, xattrs_to_heal[i]);
- if (keyval) {
- luflag = 1;
- ret = dict_set (dst, xattrs_to_heal[i], keyval);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value:key = %s for "
- "path %s", xattrs_to_heal[i],
- local->loc.path);
- keyval = NULL;
- }
- }
- if (uret)
- (*uret) = luret;
- if (uflag)
- (*uflag) = luflag;
+ int ret = -1;
+ data_t *keyval = NULL;
+ int luret = -1;
+ int luflag = -1;
+ int i = 0;
+ char **xattrs_to_heal;
+
+ if (!src || !dst) {
+ gf_smsg(this->name, GF_LOG_WARNING, EINVAL, DHT_MSG_DST_NULL_SET_FAILED,
+ "path=%s", local->loc.path, NULL);
+ return;
+ }
+ /* Check if any user xattr present in src dict and set
+ it to dst dict
+ */
+ luret = dict_foreach_fnmatch(src, "user.*", dht_set_user_xattr, dst);
+ /* Check if any other custom xattr present in src dict
+ and set it to dst dict, here index start from 1 because
+ user xattr already checked in previous statement
+ */
+
+ xattrs_to_heal = get_xattrs_to_heal();
+
+ for (i = 1; xattrs_to_heal[i]; i++) {
+ keyval = dict_get(src, xattrs_to_heal[i]);
+ if (keyval) {
+ luflag = 1;
+ ret = dict_set(dst, xattrs_to_heal[i], keyval);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_WARNING, ENOMEM,
+ DHT_MSG_DICT_SET_FAILED, "key=%s", xattrs_to_heal[i],
+ "path=%s", local->loc.path, NULL);
+ keyval = NULL;
+ }
+ }
+ if (uret)
+ (*uret) = luret;
+ if (uflag)
+ (*uflag) = luflag;
}
diff --git a/xlators/cluster/dht/src/dht-inode-read.c b/xlators/cluster/dht/src/dht-inode-read.c
index 7e4aef4ccc6..dbb8070b0da 100644
--- a/xlators/cluster/dht/src/dht-inode-read.c
+++ b/xlators/cluster/dht/src/dht-inode-read.c
@@ -10,1661 +10,1649 @@
#include "dht-common.h"
-int dht_access2 (xlator_t *this, xlator_t *dst_node,
- call_frame_t *frame, int ret);
-int dht_readv2 (xlator_t *this, xlator_t *dst_node,
- call_frame_t *frame, int ret);
-int dht_attr2 (xlator_t *this, xlator_t *dst_node,
- call_frame_t *frame, int ret);
-int dht_open2 (xlator_t *this, xlator_t *dst_node,
- call_frame_t *frame, int ret);
-int dht_flush2 (xlator_t *this, xlator_t *dst_node,
- call_frame_t *frame, int ret);
-int dht_lk2 (xlator_t *this, xlator_t *dst_node,
- call_frame_t *frame, int ret);
-int dht_fsync2 (xlator_t *this, xlator_t *dst_node,
- call_frame_t *frame, int ret);
-int
-dht_common_xattrop2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame,
- int ret);
-
-int
-dht_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
+static int
+dht_access2(xlator_t *this, xlator_t *dst_node, call_frame_t *frame, int ret);
+static int
+dht_readv2(xlator_t *this, xlator_t *dst_node, call_frame_t *frame, int ret);
+static int
+dht_attr2(xlator_t *this, xlator_t *dst_node, call_frame_t *frame, int ret);
+static int
+dht_open2(xlator_t *this, xlator_t *dst_node, call_frame_t *frame, int ret);
+static int
+dht_flush2(xlator_t *this, xlator_t *dst_node, call_frame_t *frame, int ret);
+static int
+dht_lk2(xlator_t *this, xlator_t *dst_node, call_frame_t *frame, int ret);
+static int
+dht_fsync2(xlator_t *this, xlator_t *dst_node, call_frame_t *frame, int ret);
+static int
+dht_common_xattrop2(xlator_t *this, xlator_t *subvol, call_frame_t *frame,
+ int ret);
+
+static int
+dht_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, fd_t *fd, dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- int ret = 0;
-
- local = frame->local;
- prev = cookie;
-
- local->op_errno = op_errno;
- if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->name);
- goto out;
- }
-
- /* Update ctx if the fd has been opened on the target*/
- if (!op_ret && (local->call_cnt == 1)) {
- dht_fd_ctx_set (this, fd, prev);
- goto out;
- }
-
- if (!op_ret || (local->call_cnt != 1))
- goto out;
-
- /* rebalance would have happened */
- local->rebalance.target_op_fn = dht_open2;
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret)
- return 0;
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ int ret = 0;
+
+ local = frame->local;
+ prev = cookie;
+
+ local->op_errno = op_errno;
+ if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->name);
+ goto out;
+ }
+
+ /* Update ctx if the fd has been opened on the target*/
+ if (!op_ret && (local->call_cnt == 1)) {
+ dht_fd_ctx_set(this, fd, prev);
+ goto out;
+ }
+
+ if (!op_ret || (local->call_cnt != 1))
+ goto out;
+
+ /* rebalance would have happened */
+ local->rebalance.target_op_fn = dht_open2;
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret)
+ return 0;
out:
- DHT_STACK_UNWIND (open, frame, op_ret, op_errno, local->fd, xdata);
+ DHT_STACK_UNWIND(open, frame, op_ret, op_errno, local->fd, xdata);
- return 0;
+ return 0;
}
-int
-dht_open2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
+static int
+dht_open2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
{
- dht_local_t *local = NULL;
- int op_errno = EINVAL;
+ dht_local_t *local = NULL;
+ int op_errno = EINVAL;
- if (!frame || !frame->local)
- goto out;
+ if (!frame || !frame->local)
+ goto out;
- local = frame->local;
- op_errno = local->op_errno;
+ local = frame->local;
+ op_errno = local->op_errno;
- if (we_are_not_migrating (ret)) {
- /* This DHT layer is not migrating the file */
- DHT_STACK_UNWIND (open, frame, -1, local->op_errno,
- NULL, local->rebalance.xdata);
- return 0;
+ if (we_are_not_migrating(ret)) {
+ /* This DHT layer is not migrating the file */
+ DHT_STACK_UNWIND(open, frame, -1, local->op_errno, NULL,
+ local->rebalance.xdata);
+ return 0;
+ }
- }
+ if (subvol == NULL)
+ goto out;
- if (subvol == NULL)
- goto out;
+ local->call_cnt = 2;
- local->call_cnt = 2;
-
- STACK_WIND_COOKIE (frame, dht_open_cbk, subvol, subvol,
- subvol->fops->open, &local->loc,
- local->rebalance.flags, local->fd, local->xattr_req);
- return 0;
+ STACK_WIND_COOKIE(frame, dht_open_cbk, subvol, subvol, subvol->fops->open,
+ &local->loc, local->rebalance.flags, local->fd,
+ local->xattr_req);
+ return 0;
out:
- DHT_STACK_UNWIND (open, frame, -1, op_errno, NULL, NULL);
- return 0;
+ DHT_STACK_UNWIND(open, frame, -1, op_errno, NULL, NULL);
+ return 0;
}
-
int
-dht_open (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int flags, fd_t *fd, dict_t *xdata)
+dht_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags, fd_t *fd,
+ dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- local = dht_local_init (frame, loc, fd, GF_FOP_OPEN);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for fd=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
- if (xdata)
- local->xattr_req = dict_ref (xdata);
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
- local->rebalance.flags = flags;
- local->call_cnt = 1;
+ local = dht_local_init(frame, loc, fd, GF_FOP_OPEN);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- STACK_WIND_COOKIE (frame, dht_open_cbk, subvol, subvol,
- subvol->fops->open, loc, flags, fd, xdata);
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
- return 0;
+ local->rebalance.flags = flags;
+ local->call_cnt = 1;
+
+ STACK_WIND_COOKIE(frame, dht_open_cbk, subvol, subvol, subvol->fops->open,
+ loc, flags, fd, xdata);
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (open, frame, -1, op_errno, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(open, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
int
-dht_file_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *stbuf, dict_t *xdata)
+dht_file_attr_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *stbuf, dict_t *xdata)
{
- xlator_t *subvol1 = 0;
- xlator_t *subvol2 = 0;
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- int ret = -1;
- inode_t *inode = NULL;
-
- GF_VALIDATE_OR_GOTO ("dht", frame, err);
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
- GF_VALIDATE_OR_GOTO ("dht", cookie, out);
-
- local = frame->local;
- prev = cookie;
-
- if ((local->fop == GF_FOP_FSTAT) && (op_ret == -1)
- && (op_errno == EBADF) && !(local->fd_checked)) {
-
- ret = dht_check_and_open_fd_on_subvol (this, frame);
- if (ret)
- goto out;
- return 0;
- }
-
- if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
- local->op_errno = op_errno;
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->name);
- goto out;
- }
-
- if (local->call_cnt != 1)
- goto out;
+ xlator_t *subvol1 = 0;
+ xlator_t *subvol2 = 0;
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ int ret = -1;
+ inode_t *inode = NULL;
+
+ GF_VALIDATE_OR_GOTO("dht", frame, err);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO("dht", cookie, out);
+
+ local = frame->local;
+ prev = cookie;
+
+ if ((local->fop == GF_FOP_FSTAT) &&
+ dht_check_remote_fd_failed_error(local, op_ret, op_errno)) {
+ ret = dht_check_and_open_fd_on_subvol(this, frame);
+ if (ret)
+ goto out;
+ return 0;
+ }
+ if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
local->op_errno = op_errno;
- local->op_ret = op_ret;
-
- /* Check if the rebalance phase2 is true */
- if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (stbuf)) {
-
- local->rebalance.target_op_fn = dht_attr2;
- dht_set_local_rebalance (this, local, NULL, NULL,
- stbuf, xdata);
- inode = (local->fd) ? local->fd->inode : local->loc.inode;
-
- dht_inode_ctx_get_mig_info (this, inode, &subvol1, &subvol2);
- if (dht_mig_info_is_invalid (local->cached_subvol,
- subvol1, subvol2)){
- /* Phase 2 of migration */
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret)
- return 0;
- } else {
- /* it is a non-fd op or it is an fd based Fop and
- opened on the dst.*/
- if (local->fd &&
- !dht_fd_open_on_dst (this, local->fd, subvol2)) {
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret)
- return 0;
- } else {
- dht_attr2 (this, subvol2, frame, 0);
- return 0;
- }
- }
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->name);
+ goto out;
+ }
+
+ if (local->call_cnt != 1)
+ goto out;
+
+ local->op_errno = op_errno;
+ local->op_ret = op_ret;
+
+ /* Check if the rebalance phase2 is true */
+ if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2(stbuf)) {
+ local->rebalance.target_op_fn = dht_attr2;
+ dht_set_local_rebalance(this, local, NULL, NULL, stbuf, xdata);
+ inode = (local->fd) ? local->fd->inode : local->loc.inode;
+
+ dht_inode_ctx_get_mig_info(this, inode, &subvol1, &subvol2);
+ if (dht_mig_info_is_invalid(local->cached_subvol, subvol1, subvol2)) {
+ /* Phase 2 of migration */
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret)
+ return 0;
+ } else {
+ /* it is a non-fd op or it is an fd based Fop and
+ opened on the dst.*/
+ if (local->fd && !dht_fd_open_on_dst(this, local->fd, subvol2)) {
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret)
+ return 0;
+ } else {
+ dht_attr2(this, subvol2, frame, 0);
+ return 0;
+ }
}
+ }
out:
- DHT_STRIP_PHASE1_FLAGS (stbuf);
- DHT_STACK_UNWIND (stat, frame, op_ret, op_errno, stbuf, xdata);
+ DHT_STRIP_PHASE1_FLAGS(stbuf);
+ DHT_STACK_UNWIND(stat, frame, op_ret, op_errno, stbuf, xdata);
err:
- return 0;
+ return 0;
}
-int
-dht_attr2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
+static int
+dht_attr2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
{
- dht_local_t *local = NULL;
- int op_errno = EINVAL;
-
- local = frame->local;
- if (!local)
- goto out;
-
- op_errno = local->op_errno;
-
- if (we_are_not_migrating (ret)) {
- /* This dht xlator is not migrating the file. Unwind and
- * pass on the original mode bits so the higher DHT layer
- * can handle this.
- */
- DHT_STACK_UNWIND (stat, frame, local->op_ret, op_errno,
- &local->rebalance.postbuf,
- local->rebalance.xdata);
- return 0;
- }
-
+ dht_local_t *local = NULL;
+ int op_errno = EINVAL;
+
+ local = frame->local;
+ if (!local)
+ goto out;
+
+ op_errno = local->op_errno;
+
+ if (we_are_not_migrating(ret)) {
+ /* This dht xlator is not migrating the file. Unwind and
+ * pass on the original mode bits so the higher DHT layer
+ * can handle this.
+ */
+ DHT_STACK_UNWIND(stat, frame, local->op_ret, op_errno,
+ &local->rebalance.postbuf, local->rebalance.xdata);
+ return 0;
+ }
- if (subvol == NULL)
- goto out;
+ if (subvol == NULL)
+ goto out;
- local->call_cnt = 2;
+ local->call_cnt = 2;
- if (local->fop == GF_FOP_FSTAT) {
- STACK_WIND_COOKIE (frame, dht_file_attr_cbk, subvol, subvol,
- subvol->fops->fstat, local->fd,
- local->xattr_req);
- } else {
- STACK_WIND_COOKIE (frame, dht_file_attr_cbk, subvol, subvol,
- subvol->fops->stat, &local->loc,
- local->xattr_req);
- }
+ if (local->fop == GF_FOP_FSTAT) {
+ STACK_WIND_COOKIE(frame, dht_file_attr_cbk, subvol, subvol,
+ subvol->fops->fstat, local->fd, local->xattr_req);
+ } else {
+ STACK_WIND_COOKIE(frame, dht_file_attr_cbk, subvol, subvol,
+ subvol->fops->stat, &local->loc, local->xattr_req);
+ }
- return 0;
+ return 0;
out:
- DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
- return 0;
+ DHT_STACK_UNWIND(stat, frame, -1, op_errno, NULL, NULL);
+ return 0;
}
-int
-dht_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *stbuf, dict_t *xdata)
+static int
+dht_attr_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *stbuf, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- xlator_t *prev = NULL;
-
- GF_VALIDATE_OR_GOTO ("dht", frame, err);
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
- GF_VALIDATE_OR_GOTO ("dht", cookie, out);
-
- local = frame->local;
- prev = cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->name);
-
- goto unlock;
- }
-
- dht_iatt_merge (this, &local->stbuf, stbuf, prev);
-
- local->op_ret = 0;
- }
-unlock:
- UNLOCK (&frame->lock);
-out:
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- DHT_STACK_UNWIND (stat, frame, local->op_ret, local->op_errno,
- &local->stbuf, xdata);
- }
-err:
- return 0;
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ xlator_t *prev = NULL;
+ local = frame->local;
+ prev = cookie;
+
+ LOCK(&frame->lock);
+ {
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ UNLOCK(&frame->lock);
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->name);
+
+ goto post_unlock;
+ }
+
+ dht_iatt_merge(this, &local->stbuf, stbuf);
+
+ local->op_ret = 0;
+ }
+ UNLOCK(&frame->lock);
+post_unlock:
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt)) {
+ DHT_STACK_UNWIND(stat, frame, local->op_ret, local->op_errno,
+ &local->stbuf, xdata);
+ }
+
+ return 0;
}
int
-dht_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+dht_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
- dht_layout_t *layout = NULL;
- int i = 0;
- int call_cnt = 0;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
-
-
- local = dht_local_init (frame, loc, NULL, GF_FOP_STAT);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- layout = local->layout;
- if (!layout) {
- gf_msg_debug (this->name, 0,
- "no layout for path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
- if (xdata)
- local->xattr_req = dict_ref (xdata);
-
- if (IA_ISREG (loc->inode->ia_type)) {
- local->call_cnt = 1;
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+ dht_layout_t *layout = NULL;
+ int i = 0;
+ int call_cnt = 0;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
+ VALIDATE_OR_GOTO(loc->path, err);
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_STAT);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ layout = local->layout;
+ if (!layout) {
+ gf_msg_debug(this->name, 0, "no layout for path=%s", loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ if (IA_ISREG(loc->inode->ia_type)) {
+ local->call_cnt = 1;
- subvol = local->cached_subvol;
+ subvol = local->cached_subvol;
- STACK_WIND_COOKIE (frame, dht_file_attr_cbk, subvol, subvol,
- subvol->fops->stat, loc, xdata);
+ STACK_WIND_COOKIE(frame, dht_file_attr_cbk, subvol, subvol,
+ subvol->fops->stat, loc, xdata);
- return 0;
- }
+ return 0;
+ }
- local->call_cnt = call_cnt = layout->cnt;
+ local->call_cnt = call_cnt = layout->cnt;
- for (i = 0; i < call_cnt; i++) {
- subvol = layout->list[i].xlator;
+ for (i = 0; i < call_cnt; i++) {
+ subvol = layout->list[i].xlator;
- STACK_WIND_COOKIE (frame, dht_attr_cbk, subvol, subvol,
- subvol->fops->stat, loc, xdata);
- }
+ STACK_WIND_COOKIE(frame, dht_attr_cbk, subvol, subvol,
+ subvol->fops->stat, loc, xdata);
+ }
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(stat, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
-
int
-dht_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+dht_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
- dht_layout_t *layout = NULL;
- int i = 0;
- int call_cnt = 0;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- local = dht_local_init (frame, NULL, fd, GF_FOP_FSTAT);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- layout = local->layout;
- if (!layout) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0,
- "no layout for fd=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
- if (xdata)
- local->xattr_req = dict_ref (xdata);
-
- if (IA_ISREG (fd->inode->ia_type)) {
- local->call_cnt = 1;
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+ dht_layout_t *layout = NULL;
+ int i = 0;
+ int call_cnt = 0;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+
+ local = dht_local_init(frame, NULL, fd, GF_FOP_FSTAT);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ layout = local->layout;
+ if (!layout) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "no layout for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ if (IA_ISREG(fd->inode->ia_type)) {
+ local->call_cnt = 1;
- subvol = local->cached_subvol;
+ subvol = local->cached_subvol;
- STACK_WIND_COOKIE (frame, dht_file_attr_cbk, subvol,
- subvol, subvol->fops->fstat, fd,
- xdata);
- return 0;
- }
+ STACK_WIND_COOKIE(frame, dht_file_attr_cbk, subvol, subvol,
+ subvol->fops->fstat, fd, xdata);
+ return 0;
+ }
- local->call_cnt = call_cnt = layout->cnt;
+ local->call_cnt = call_cnt = layout->cnt;
- for (i = 0; i < call_cnt; i++) {
- subvol = layout->list[i].xlator;
- STACK_WIND_COOKIE (frame, dht_attr_cbk, subvol, subvol,
- subvol->fops->fstat, fd, xdata);
- }
+ for (i = 0; i < call_cnt; i++) {
+ subvol = layout->list[i].xlator;
+ STACK_WIND_COOKIE(frame, dht_attr_cbk, subvol, subvol,
+ subvol->fops->fstat, fd, xdata);
+ }
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fstat, frame, -1, op_errno, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(fstat, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
int
-dht_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- struct iovec *vector, int count, struct iatt *stbuf,
- struct iobref *iobref, dict_t *xdata)
+dht_readv_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iovec *vector, int count, struct iatt *stbuf,
+ struct iobref *iobref, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int ret = 0;
- xlator_t *src_subvol = 0;
- xlator_t *dst_subvol = 0;
-
- local = frame->local;
- if (!local) {
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
+ dht_local_t *local = NULL;
+ int ret = 0;
+ xlator_t *src_subvol = 0;
+ xlator_t *dst_subvol = 0;
+
+ local = frame->local;
+ if (!local) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ /* This is already second try, no need for re-check */
+ if (local->call_cnt != 1)
+ goto out;
+
+ if (dht_check_remote_fd_failed_error(local, op_ret, op_errno)) {
+ ret = dht_check_and_open_fd_on_subvol(this, frame);
+ if (ret)
+ goto out;
+ return 0;
+ }
- /* This is already second try, no need for re-check */
- if (local->call_cnt != 1)
- goto out;
+ if ((op_ret == -1) && !dht_inode_missing(op_errno))
+ goto out;
- if (op_ret == -1 && (op_errno == EBADF) && !(local->fd_checked)) {
- ret = dht_check_and_open_fd_on_subvol (this, frame);
- if (ret)
- goto out;
+ local->op_errno = op_errno;
+ if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2(stbuf)) {
+ local->op_ret = op_ret;
+ local->rebalance.target_op_fn = dht_readv2;
+ dht_set_local_rebalance(this, local, NULL, NULL, stbuf, xdata);
+ /* File would be migrated to other node */
+ ret = dht_inode_ctx_get_mig_info(this, local->fd->inode, &src_subvol,
+ &dst_subvol);
+
+ if (dht_mig_info_is_invalid(local->cached_subvol, src_subvol,
+ dst_subvol) ||
+ !dht_fd_open_on_dst(this, local->fd, dst_subvol)) {
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret)
return 0;
+ } else {
+ /* value is already set in fd_ctx, that means no need
+ to check for whether its complete or not. */
+ dht_readv2(this, dst_subvol, frame, 0);
+ return 0;
}
-
- if ((op_ret == -1) && !dht_inode_missing(op_errno))
- goto out;
-
-
- local->op_errno = op_errno;
- if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (stbuf)) {
-
- local->op_ret = op_ret;
- local->rebalance.target_op_fn = dht_readv2;
- dht_set_local_rebalance (this, local, NULL, NULL,
- stbuf, xdata);
- /* File would be migrated to other node */
- ret = dht_inode_ctx_get_mig_info (this, local->fd->inode,
- &src_subvol,
- &dst_subvol);
-
- if (dht_mig_info_is_invalid (local->cached_subvol,
- src_subvol, dst_subvol)
- || !dht_fd_open_on_dst(this, local->fd, dst_subvol)) {
-
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret)
- return 0;
- } else {
- /* value is already set in fd_ctx, that means no need
- to check for whether its complete or not. */
- dht_readv2 (this, dst_subvol, frame, 0);
- return 0;
- }
- }
+ }
out:
- DHT_STRIP_PHASE1_FLAGS (stbuf);
+ DHT_STRIP_PHASE1_FLAGS(stbuf);
- DHT_STACK_UNWIND (readv, frame, op_ret, op_errno, vector, count, stbuf,
- iobref, xdata);
+ DHT_STACK_UNWIND(readv, frame, op_ret, op_errno, vector, count, stbuf,
+ iobref, xdata);
- return 0;
+ return 0;
}
-int
-dht_readv2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
+static int
+dht_readv2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
{
- dht_local_t *local = NULL;
- int op_errno = EINVAL;
-
- local = frame->local;
- if (!local)
- goto out;
-
- op_errno = local->op_errno;
-
- if (we_are_not_migrating (ret)) {
- /* This dht xlator is not migrating the file. Unwind and
- * pass on the original mode bits so the higher DHT layer
- * can handle this.
- */
- DHT_STACK_UNWIND (readv, frame, local->op_ret, op_errno,
- NULL, 0, &local->rebalance.postbuf,
- NULL, local->rebalance.xdata);
- return 0;
- }
+ dht_local_t *local = NULL;
+ int op_errno = EINVAL;
+
+ local = frame->local;
+ if (!local)
+ goto out;
+
+ op_errno = local->op_errno;
+
+ if (we_are_not_migrating(ret)) {
+ /* This dht xlator is not migrating the file. Unwind and
+ * pass on the original mode bits so the higher DHT layer
+ * can handle this.
+ */
+ DHT_STACK_UNWIND(readv, frame, local->op_ret, op_errno, NULL, 0,
+ &local->rebalance.postbuf, NULL,
+ local->rebalance.xdata);
+ return 0;
+ }
- if (subvol == NULL)
- goto out;
+ if (subvol == NULL)
+ goto out;
- local->call_cnt = 2;
+ local->call_cnt = 2;
- STACK_WIND (frame, dht_readv_cbk, subvol, subvol->fops->readv,
- local->fd, local->rebalance.size, local->rebalance.offset,
- local->rebalance.flags, local->xattr_req);
+ STACK_WIND(frame, dht_readv_cbk, subvol, subvol->fops->readv, local->fd,
+ local->rebalance.size, local->rebalance.offset,
+ local->rebalance.flags, local->xattr_req);
- return 0;
+ return 0;
out:
- DHT_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, NULL, NULL);
- return 0;
+ DHT_STACK_UNWIND(readv, frame, -1, op_errno, NULL, 0, NULL, NULL, NULL);
+ return 0;
}
-
int
-dht_readv (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t off, uint32_t flags, dict_t *xdata)
+dht_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size, off_t off,
+ uint32_t flags, dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- local = dht_local_init (frame, NULL, fd, GF_FOP_READ);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for fd=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
- if (xdata)
- local->xattr_req = dict_ref (xdata);
+ local = dht_local_init(frame, NULL, fd, GF_FOP_READ);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- local->rebalance.offset = off;
- local->rebalance.size = size;
- local->rebalance.flags = flags;
- local->call_cnt = 1;
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
- STACK_WIND (frame, dht_readv_cbk, subvol, subvol->fops->readv,
- local->fd, local->rebalance.size,
- local->rebalance.offset,
- local->rebalance.flags, local->xattr_req);
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
- return 0;
+ local->rebalance.offset = off;
+ local->rebalance.size = size;
+ local->rebalance.flags = flags;
+ local->call_cnt = 1;
+
+ STACK_WIND(frame, dht_readv_cbk, subvol, subvol->fops->readv, local->fd,
+ local->rebalance.size, local->rebalance.offset,
+ local->rebalance.flags, local->xattr_req);
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(readv, frame, -1, op_errno, NULL, 0, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-int
-dht_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+static int
+dht_access_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, dict_t *xdata)
{
- int ret = -1;
- dht_local_t *local = NULL;
- xlator_t *subvol = NULL;
- xlator_t *prev = NULL;
-
- local = frame->local;
- prev = cookie;
-
- if (!prev)
- goto out;
- if (local->call_cnt != 1)
- goto out;
- if ((op_ret == -1) && ((op_errno == ENOTCONN) ||
- dht_inode_missing(op_errno)) &&
- IA_ISDIR(local->loc.inode->ia_type)) {
- subvol = dht_subvol_next_available (this, prev);
- if (!subvol)
- goto out;
-
- /* check if we are done with visiting every node */
- if (subvol == local->cached_subvol) {
- goto out;
- }
-
- STACK_WIND_COOKIE (frame, dht_access_cbk, subvol, subvol,
- subvol->fops->access, &local->loc,
- local->rebalance.flags, NULL);
- return 0;
- }
- if ((op_ret == -1) && dht_inode_missing(op_errno) &&
- !(IA_ISDIR(local->loc.inode->ia_type))) {
- /* File would be migrated to other node */
- local->op_errno = op_errno;
- local->rebalance.target_op_fn = dht_access2;
- ret = dht_rebalance_complete_check (frame->this, frame);
- if (!ret)
- return 0;
- }
+ int ret = -1;
+ dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+ xlator_t *prev = NULL;
+
+ local = frame->local;
+ prev = cookie;
+
+ if (!prev)
+ goto out;
+ if (local->call_cnt != 1)
+ goto out;
+ if ((op_ret == -1) &&
+ ((op_errno == ENOTCONN) || dht_inode_missing(op_errno)) &&
+ IA_ISDIR(local->loc.inode->ia_type)) {
+ subvol = dht_subvol_next_available(this, prev);
+ if (!subvol)
+ goto out;
+
+ /* check if we are done with visiting every node */
+ if (subvol == local->cached_subvol) {
+ goto out;
+ }
+
+ STACK_WIND_COOKIE(frame, dht_access_cbk, subvol, subvol,
+ subvol->fops->access, &local->loc,
+ local->rebalance.flags, NULL);
+ return 0;
+ }
+ if ((op_ret == -1) && dht_inode_missing(op_errno) &&
+ !(IA_ISDIR(local->loc.inode->ia_type))) {
+ /* File would be migrated to other node */
+ local->op_errno = op_errno;
+ local->rebalance.target_op_fn = dht_access2;
+ ret = dht_rebalance_complete_check(frame->this, frame);
+ if (!ret)
+ return 0;
+ }
out:
- DHT_STACK_UNWIND (access, frame, op_ret, op_errno, xdata);
- return 0;
+ DHT_STACK_UNWIND(access, frame, op_ret, op_errno, xdata);
+ return 0;
}
-int
-dht_access2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
+static int
+dht_access2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
{
- dht_local_t *local = NULL;
- int op_errno = EINVAL;
+ dht_local_t *local = NULL;
+ int op_errno = EINVAL;
- local = frame->local;
- if (!local)
- goto out;
+ local = frame->local;
+ if (!local)
+ goto out;
- op_errno = local->op_errno;
+ op_errno = local->op_errno;
- if (we_are_not_migrating (ret)) {
- /* This dht xlator is not migrating the file. Unwind and
- * pass on the original mode bits so the higher DHT layer
- * can handle this.
- */
+ if (we_are_not_migrating(ret)) {
+ /* This dht xlator is not migrating the file. Unwind and
+ * pass on the original mode bits so the higher DHT layer
+ * can handle this.
+ */
- DHT_STACK_UNWIND (access, frame, -1, op_errno, NULL);
- return 0;
- }
+ DHT_STACK_UNWIND(access, frame, -1, op_errno, NULL);
+ return 0;
+ }
- if (subvol == NULL)
- goto out;
+ if (subvol == NULL)
+ goto out;
- local->call_cnt = 2;
+ local->call_cnt = 2;
- STACK_WIND_COOKIE (frame, dht_access_cbk, subvol, subvol,
- subvol->fops->access, &local->loc,
- local->rebalance.flags, local->xattr_req);
+ STACK_WIND_COOKIE(frame, dht_access_cbk, subvol, subvol,
+ subvol->fops->access, &local->loc, local->rebalance.flags,
+ local->xattr_req);
- return 0;
+ return 0;
out:
- DHT_STACK_UNWIND (access, frame, -1, op_errno, NULL);
- return 0;
+ DHT_STACK_UNWIND(access, frame, -1, op_errno, NULL);
+ return 0;
}
-
int
-dht_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
- dict_t *xdata)
+dht_access(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+ dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
-
- local = dht_local_init (frame, loc, NULL, GF_FOP_ACCESS);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local->rebalance.flags = mask;
- local->call_cnt = 1;
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
- if (xdata)
- local->xattr_req = dict_ref (xdata);
-
- STACK_WIND_COOKIE (frame, dht_access_cbk, subvol, subvol,
- subvol->fops->access, loc, mask, xdata);
-
- return 0;
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
+ VALIDATE_OR_GOTO(loc->path, err);
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_ACCESS);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->rebalance.flags = mask;
+ local->call_cnt = 1;
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for path=%s",
+ loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ STACK_WIND_COOKIE(frame, dht_access_cbk, subvol, subvol,
+ subvol->fops->access, loc, mask, xdata);
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (access, frame, -1, op_errno, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(access, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
-
int
-dht_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+dht_flush_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *subvol = 0;
- int ret = 0;
+ dht_local_t *local = NULL;
+ xlator_t *subvol = 0;
+ int ret = 0;
- local = frame->local;
+ local = frame->local;
- local->op_errno = op_errno;
+ local->op_errno = op_errno;
- if (local->call_cnt != 1)
- goto out;
+ if (local->call_cnt != 1)
+ goto out;
- if (op_ret == -1 && (op_errno == EBADF) && !(local->fd_checked)) {
- ret = dht_check_and_open_fd_on_subvol (this, frame);
- if (ret)
- goto out;
- return 0;
- }
+ if (dht_check_remote_fd_failed_error(local, op_ret, op_errno)) {
+ ret = dht_check_and_open_fd_on_subvol(this, frame);
+ if (ret)
+ goto out;
+ return 0;
+ }
- local->rebalance.target_op_fn = dht_flush2;
+ local->rebalance.target_op_fn = dht_flush2;
- local->op_ret = op_ret;
- local->op_errno = op_errno;
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
- /* If context is set, then send flush() it to the destination */
- dht_inode_ctx_get_mig_info (this, local->fd->inode, NULL, &subvol);
- if (subvol && dht_fd_open_on_dst (this, local->fd, subvol)) {
- dht_flush2 (this, subvol, frame, 0);
- return 0;
- }
+ /* If context is set, then send flush() it to the destination */
+ dht_inode_ctx_get_mig_info(this, local->fd->inode, NULL, &subvol);
+ if (subvol && dht_fd_open_on_dst(this, local->fd, subvol)) {
+ dht_flush2(this, subvol, frame, 0);
+ return 0;
+ }
- if (op_errno == EREMOTE) {
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret) {
- return 0;
- }
+ if (op_errno == EREMOTE) {
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret) {
+ return 0;
}
+ }
out:
- DHT_STACK_UNWIND (flush, frame, op_ret, op_errno, xdata);
+ DHT_STACK_UNWIND(flush, frame, op_ret, op_errno, xdata);
- return 0;
+ return 0;
}
-int
-dht_flush2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
+static int
+dht_flush2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
{
- dht_local_t *local = NULL;
- int32_t op_errno = EINVAL;
+ dht_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
- if ((frame == NULL) || (frame->local == NULL))
- goto out;
+ if ((frame == NULL) || (frame->local == NULL))
+ goto out;
- local = frame->local;
+ local = frame->local;
- op_errno = local->op_errno;
+ op_errno = local->op_errno;
- if (subvol == NULL)
- goto out;
+ if (subvol == NULL)
+ goto out;
- local->call_cnt = 2; /* This is the second attempt */
+ local->call_cnt = 2; /* This is the second attempt */
- STACK_WIND (frame, dht_flush_cbk,
- subvol, subvol->fops->flush, local->fd,
- local->xattr_req);
+ STACK_WIND(frame, dht_flush_cbk, subvol, subvol->fops->flush, local->fd,
+ local->xattr_req);
- return 0;
+ return 0;
out:
- DHT_STACK_UNWIND (flush, frame, -1, op_errno, NULL);
- return 0;
+ DHT_STACK_UNWIND(flush, frame, -1, op_errno, NULL);
+ return 0;
}
-
int
-dht_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+dht_flush(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- local = dht_local_init (frame, NULL, fd, GF_FOP_FLUSH);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for fd=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
- if (xdata)
- local->xattr_req = dict_ref (xdata);
+ local = dht_local_init(frame, NULL, fd, GF_FOP_FLUSH);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- local->call_cnt = 1;
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
- STACK_WIND (frame, dht_flush_cbk,
- subvol, subvol->fops->flush, fd, local->xattr_req);
- return 0;
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ local->call_cnt = 1;
+
+ STACK_WIND(frame, dht_flush_cbk, subvol, subvol->fops->flush, fd,
+ local->xattr_req);
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (flush, frame, -1, op_errno, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(flush, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
-
int
-dht_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
- int op_errno, struct iatt *prebuf, struct iatt *postbuf,
- dict_t *xdata)
+dht_fsync_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- int ret = -1;
- inode_t *inode = NULL;
- xlator_t *src_subvol = 0;
- xlator_t *dst_subvol = 0;
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ int ret = -1;
+ inode_t *inode = NULL;
+ xlator_t *src_subvol = 0;
+ xlator_t *dst_subvol = 0;
- local = frame->local;
- prev = cookie;
+ local = frame->local;
+ prev = cookie;
- local->op_errno = op_errno;
+ local->op_errno = op_errno;
- if (op_ret == -1 && (op_errno == EBADF) && !(local->fd_checked)) {
- ret = dht_check_and_open_fd_on_subvol (this, frame);
- if (ret)
- goto out;
- return 0;
- }
+ if (dht_check_remote_fd_failed_error(local, op_ret, op_errno)) {
+ ret = dht_check_and_open_fd_on_subvol(this, frame);
+ if (ret)
+ goto out;
+ return 0;
+ }
- if (op_ret == -1 && !dht_inode_missing(op_errno)) {
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->name);
- goto out;
- }
+ if (op_ret == -1 && !dht_inode_missing(op_errno)) {
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->name);
+ goto out;
+ }
- if (local->call_cnt != 1) {
- if (local->stbuf.ia_blocks) {
- dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
- dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
- }
- goto out;
+ if (local->call_cnt != 1) {
+ if (local->stbuf.ia_blocks) {
+ dht_iatt_merge(this, postbuf, &local->stbuf);
+ dht_iatt_merge(this, prebuf, &local->prebuf);
}
+ goto out;
+ }
- local->op_ret = op_ret;
- inode = local->fd->inode;
-
- local->rebalance.target_op_fn = dht_fsync2;
- dht_set_local_rebalance (this, local, NULL, prebuf,
- postbuf, xdata);
-
- if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret)
- return 0;
- }
+ local->op_ret = op_ret;
+ inode = local->fd->inode;
- /* Check if the rebalance phase1 is true */
- if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
+ local->rebalance.target_op_fn = dht_fsync2;
+ dht_set_local_rebalance(this, local, NULL, prebuf, postbuf, xdata);
- dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
- dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
+ if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2(postbuf)) {
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret)
+ return 0;
+ }
- dht_inode_ctx_get_mig_info (this, inode, &src_subvol, &dst_subvol);
+ /* Check if the rebalance phase1 is true */
+ if (IS_DHT_MIGRATION_PHASE1(postbuf)) {
+ dht_iatt_merge(this, &local->stbuf, postbuf);
+ dht_iatt_merge(this, &local->prebuf, prebuf);
- if (dht_mig_info_is_invalid (local->cached_subvol, src_subvol,
- dst_subvol) ||
- !dht_fd_open_on_dst (this, local->fd, dst_subvol)) {
+ dht_inode_ctx_get_mig_info(this, inode, &src_subvol, &dst_subvol);
- ret = dht_rebalance_in_progress_check (this, frame);
- if (!ret)
- return 0;
- } else {
- dht_fsync2 (this, dst_subvol, frame, 0);
- return 0;
- }
+ if (dht_mig_info_is_invalid(local->cached_subvol, src_subvol,
+ dst_subvol) ||
+ !dht_fd_open_on_dst(this, local->fd, dst_subvol)) {
+ ret = dht_rebalance_in_progress_check(this, frame);
+ if (!ret)
+ return 0;
+ } else {
+ dht_fsync2(this, dst_subvol, frame, 0);
+ return 0;
}
-
+ }
out:
- DHT_STRIP_PHASE1_FLAGS (postbuf);
- DHT_STRIP_PHASE1_FLAGS (prebuf);
+ DHT_STRIP_PHASE1_FLAGS(postbuf);
+ DHT_STRIP_PHASE1_FLAGS(prebuf);
- DHT_STACK_UNWIND (fsync, frame, op_ret, op_errno,
- prebuf, postbuf, xdata);
+ DHT_STACK_UNWIND(fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
- return 0;
+ return 0;
}
-int
-dht_fsync2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
+static int
+dht_fsync2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
{
- dht_local_t *local = NULL;
- int32_t op_errno = EINVAL;
-
- if ((frame == NULL) || (frame->local == NULL))
- goto out;
-
- local = frame->local;
- op_errno = local->op_errno;
-
- if (we_are_not_migrating (ret)) {
- /* This dht xlator is not migrating the file. Unwind and
- * pass on the original mode bits so the higher DHT layer
- * can handle this.
- */
- DHT_STACK_UNWIND (fsync, frame, local->op_ret,
- op_errno, &local->rebalance.prebuf,
- &local->rebalance.postbuf,
- local->rebalance.xdata);
- return 0;
- }
+ dht_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
+
+ if ((frame == NULL) || (frame->local == NULL))
+ goto out;
+
+ local = frame->local;
+ op_errno = local->op_errno;
+
+ if (we_are_not_migrating(ret)) {
+ /* This dht xlator is not migrating the file. Unwind and
+ * pass on the original mode bits so the higher DHT layer
+ * can handle this.
+ */
+ DHT_STACK_UNWIND(fsync, frame, local->op_ret, op_errno,
+ &local->rebalance.prebuf, &local->rebalance.postbuf,
+ local->rebalance.xdata);
+ return 0;
+ }
- if (subvol == NULL)
- goto out;
+ if (subvol == NULL)
+ goto out;
- local->call_cnt = 2; /* This is the second attempt */
+ local->call_cnt = 2; /* This is the second attempt */
- STACK_WIND_COOKIE (frame, dht_fsync_cbk, subvol, subvol,
- subvol->fops->fsync, local->fd,
- local->rebalance.flags, local->xattr_req);
+ STACK_WIND_COOKIE(frame, dht_fsync_cbk, subvol, subvol, subvol->fops->fsync,
+ local->fd, local->rebalance.flags, local->xattr_req);
- return 0;
+ return 0;
out:
- DHT_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ DHT_STACK_UNWIND(fsync, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
int
-dht_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int datasync,
- dict_t *xdata)
+dht_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int datasync,
+ dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
- local = dht_local_init (frame, NULL, fd, GF_FOP_FSYNC);
- if (!local) {
- op_errno = ENOMEM;
+ local = dht_local_init(frame, NULL, fd, GF_FOP_FSYNC);
+ if (!local) {
+ op_errno = ENOMEM;
- goto err;
- }
- if (xdata)
- local->xattr_req = dict_ref (xdata);
+ goto err;
+ }
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
- local->call_cnt = 1;
- local->rebalance.flags = datasync;
+ local->call_cnt = 1;
+ local->rebalance.flags = datasync;
- subvol = local->cached_subvol;
+ subvol = local->cached_subvol;
- STACK_WIND_COOKIE (frame, dht_fsync_cbk, subvol, subvol,
- subvol->fops->fsync, local->fd,
- local->rebalance.flags, local->xattr_req);
- return 0;
+ STACK_WIND_COOKIE(frame, dht_fsync_cbk, subvol, subvol, subvol->fops->fsync,
+ local->fd, local->rebalance.flags, local->xattr_req);
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(fsync, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
/* TODO: for 'lk()' call, we need some other special error, may be ESTALE to
indicate that lock migration happened on the fd, so we can consider it as
phase 2 of migration */
-int
-dht_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct gf_flock *flock, dict_t *xdata)
+static int
+dht_lk_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct gf_flock *flock, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int ret = -1;
- xlator_t *subvol = NULL;
+ dht_local_t *local = NULL;
+ int ret = -1;
+ xlator_t *subvol = NULL;
- local = frame->local;
+ local = frame->local;
- if (!local) {
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
+ if (!local) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto out;
+ }
- if (local->call_cnt != 1)
- goto out;
+ if (local->call_cnt != 1)
+ goto out;
- local->rebalance.target_op_fn = dht_lk2;
+ local->rebalance.target_op_fn = dht_lk2;
- local->op_ret = op_ret;
- local->op_errno = op_errno;
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
- if (xdata)
- local->rebalance.xdata = dict_ref (xdata);
-
- if (op_errno == EREMOTE) {
- dht_inode_ctx_get_mig_info (this, local->fd->inode,
- NULL, &subvol);
- if (subvol && dht_fd_open_on_dst (this, local->fd, subvol)) {
- dht_lk2 (this, subvol, frame, 0);
- return 0;
- } else {
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret) {
- return 0;
- }
- }
+ if (xdata)
+ local->rebalance.xdata = dict_ref(xdata);
+
+ if (op_errno == EREMOTE) {
+ dht_inode_ctx_get_mig_info(this, local->fd->inode, NULL, &subvol);
+ if (subvol && dht_fd_open_on_dst(this, local->fd, subvol)) {
+ dht_lk2(this, subvol, frame, 0);
+ return 0;
+ } else {
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret) {
+ return 0;
+ }
}
+ }
out:
- dht_lk_inode_unref (frame, op_ret);
- DHT_STACK_UNWIND (lk, frame, op_ret, op_errno, flock, xdata);
+ dht_lk_inode_unref(frame, op_ret);
+ DHT_STACK_UNWIND(lk, frame, op_ret, op_errno, flock, xdata);
- return 0;
+ return 0;
}
-int
-dht_lk2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
+static int
+dht_lk2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
{
- dht_local_t *local = NULL;
- int32_t op_errno = EINVAL;
+ dht_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
- if ((frame == NULL) || (frame->local == NULL))
- goto out;
+ if ((frame == NULL) || (frame->local == NULL))
+ goto out;
- local = frame->local;
+ local = frame->local;
- op_errno = local->op_errno;
+ op_errno = local->op_errno;
- if (subvol == NULL)
- goto out;
+ if (subvol == NULL)
+ goto out;
- local->call_cnt = 2; /* This is the second attempt */
+ local->call_cnt = 2; /* This is the second attempt */
- STACK_WIND (frame, dht_lk_cbk, subvol, subvol->fops->lk, local->fd,
- local->rebalance.lock_cmd, &local->rebalance.flock,
- local->xattr_req);
+ STACK_WIND(frame, dht_lk_cbk, subvol, subvol->fops->lk, local->fd,
+ local->rebalance.lock_cmd, &local->rebalance.flock,
+ local->xattr_req);
- return 0;
+ return 0;
out:
- DHT_STACK_UNWIND (lk, frame, -1, op_errno, NULL, NULL);
- return 0;
+ DHT_STACK_UNWIND(lk, frame, -1, op_errno, NULL, NULL);
+ return 0;
}
int
-dht_lk (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int cmd, struct gf_flock *flock, dict_t *xdata)
+dht_lk(call_frame_t *frame, xlator_t *this, fd_t *fd, int cmd,
+ struct gf_flock *flock, dict_t *xdata)
{
- xlator_t *lock_subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- local = dht_local_init (frame, NULL, fd, GF_FOP_LK);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local->lock_type = flock->l_type;
- lock_subvol = dht_get_lock_subvolume (this, flock, local);
- if (!lock_subvol) {
- gf_msg_debug (this->name, 0,
- "no lock subvolume for path=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
-
-/*
- local->cached_subvol = lock_subvol;
- ret = dht_check_and_open_fd_on_subvol (this, frame);
- if (ret)
- goto err;
-*/
- if (xdata)
- local->xattr_req = dict_ref (xdata);
-
- local->rebalance.flock = *flock;
- local->rebalance.lock_cmd = cmd;
-
- local->call_cnt = 1;
-
- STACK_WIND (frame, dht_lk_cbk, lock_subvol, lock_subvol->fops->lk, fd,
- cmd, flock, xdata);
-
- return 0;
+ xlator_t *lock_subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+
+ local = dht_local_init(frame, NULL, fd, GF_FOP_LK);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->lock_type = flock->l_type;
+ lock_subvol = dht_get_lock_subvolume(this, flock, local);
+ if (!lock_subvol) {
+ gf_msg_debug(this->name, 0, "no lock subvolume for path=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ /*
+ local->cached_subvol = lock_subvol;
+ ret = dht_check_and_open_fd_on_subvol (this, frame);
+ if (ret)
+ goto err;
+ */
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ local->rebalance.flock = *flock;
+ local->rebalance.lock_cmd = cmd;
+
+ local->call_cnt = 1;
+
+ STACK_WIND(frame, dht_lk_cbk, lock_subvol, lock_subvol->fops->lk, fd, cmd,
+ flock, xdata);
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (lk, frame, -1, op_errno, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(lk, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
-int
-dht_lease_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct gf_lease *lease, dict_t *xdata)
+static int
+dht_lease_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct gf_lease *lease, dict_t *xdata)
{
- DHT_STACK_UNWIND (lease, frame, op_ret, op_errno, lease, xdata);
+ DHT_STACK_UNWIND(lease, frame, op_ret, op_errno, lease, xdata);
- return 0;
+ return 0;
}
int
-dht_lease (call_frame_t *frame, xlator_t *this,
- loc_t *loc, struct gf_lease *lease, dict_t *xdata)
+dht_lease(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct gf_lease *lease, dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
-
- subvol = dht_subvol_get_cached (this, loc->inode);
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
- /* TODO: for rebalance, we need to preserve the fop arguments */
- STACK_WIND (frame, dht_lease_cbk, subvol, subvol->fops->lease,
- loc, lease, xdata);
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
- return 0;
+ subvol = dht_subvol_get_cached(this, loc->inode);
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for path=%s",
+ loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ /* TODO: for rebalance, we need to preserve the fop arguments */
+ STACK_WIND(frame, dht_lease_cbk, subvol, subvol->fops->lease, loc, lease,
+ xdata);
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (lease, frame, -1, op_errno, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(lease, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
/* Symlinks are currently not migrated, so no need for any check here */
-int
-dht_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, const char *path,
- struct iatt *stbuf, dict_t *xdata)
+static int
+dht_readlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, const char *path, struct iatt *stbuf,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
+ dht_local_t *local = NULL;
- local = frame->local;
- if (op_ret == -1)
- goto err;
+ local = frame->local;
+ if (op_ret == -1)
+ goto err;
- if (!local) {
- op_ret = -1;
- op_errno = EINVAL;
- }
+ if (!local) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ }
err:
- DHT_STRIP_PHASE1_FLAGS (stbuf);
- DHT_STACK_UNWIND (readlink, frame, op_ret, op_errno, path, stbuf, xdata);
+ DHT_STRIP_PHASE1_FLAGS(stbuf);
+ DHT_STACK_UNWIND(readlink, frame, op_ret, op_errno, path, stbuf, xdata);
- return 0;
+ return 0;
}
-
int
-dht_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
- dict_t *xdata)
+dht_readlink(call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+ dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
-
- local = dht_local_init (frame, loc, NULL, GF_FOP_READLINK);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
-
- STACK_WIND (frame, dht_readlink_cbk,
- subvol, subvol->fops->readlink,
- loc, size, xdata);
-
- return 0;
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
+ VALIDATE_OR_GOTO(loc->path, err);
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_READLINK);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for path=%s",
+ loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ STACK_WIND(frame, dht_readlink_cbk, subvol, subvol->fops->readlink, loc,
+ size, xdata);
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (readlink, frame, -1, op_errno, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(readlink, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
-int
-dht_common_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
+/* Get both DHT_IATT_IN_XDATA_KEY and DHT_MODE_IN_XDATA_KEY
+ * Use DHT_MODE_IN_XDATA_KEY if available, else fall back to
+ * DHT_IATT_IN_XDATA_KEY
+ * This will return a dummy iatt with only the mode and type set
+ */
+static int
+dht_read_iatt_from_xdata(dict_t *xdata, struct iatt *stbuf)
{
- dht_local_t *local = NULL;
- call_frame_t *call_frame = NULL;
- xlator_t *prev = NULL;
- xlator_t *src_subvol = NULL;
- xlator_t *dst_subvol = NULL;
- struct iatt stbuf = {0,};
- int ret = -1;
- inode_t *inode = NULL;
-
- local = frame->local;
- call_frame = cookie;
- prev = call_frame->this;
+ int ret = -1;
+ int32_t mode = 0;
- local->op_errno = op_errno;
-
- if ((op_ret == -1) && !dht_inode_missing (op_errno)) {
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1.",
- prev->name);
- goto out;
- }
+ ret = dict_get_int32(xdata, DHT_MODE_IN_XDATA_KEY, &mode);
- if (local->call_cnt != 1)
- goto out;
+ if (ret) {
+ ret = dict_get_bin(xdata, DHT_IATT_IN_XDATA_KEY, (void **)&stbuf);
+ } else {
+ stbuf->ia_prot = ia_prot_from_st_mode(mode);
+ stbuf->ia_type = ia_type_from_st_mode(mode);
+ }
- ret = dht_read_iatt_from_xdata (this, xdata, &stbuf);
-
- if ((!op_ret) && (ret)) {
- /* This is a potential problem and can cause corruption
- * with sharding.
- * Oh well. We tried.
- */
- goto out;
- }
-
- local->op_ret = op_ret;
- local->rebalance.target_op_fn = dht_common_xattrop2;
- if (xdata)
- local->rebalance.xdata = dict_ref (xdata);
+ return ret;
+}
- if (dict)
- local->rebalance.dict = dict_ref (dict);
+int
+dht_common_xattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ call_frame_t *call_frame = NULL;
+ xlator_t *prev = NULL;
+ xlator_t *src_subvol = NULL;
+ xlator_t *dst_subvol = NULL;
+ struct iatt stbuf = {
+ 0,
+ };
+ int ret = -1;
+ inode_t *inode = NULL;
+
+ local = frame->local;
+ call_frame = cookie;
+ prev = call_frame->this;
+
+ local->op_errno = op_errno;
+
+ if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1.",
+ prev->name);
+ goto out;
+ }
+
+ if (local->call_cnt != 1)
+ goto out;
+
+ if (dht_check_remote_fd_failed_error(local, op_ret, op_errno)) {
+ ret = dht_check_and_open_fd_on_subvol(this, frame);
+ if (ret)
+ goto out;
+ return 0;
+ }
- /* Phase 2 of migration */
- if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (&stbuf)) {
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret)
- return 0;
- }
+ ret = dht_read_iatt_from_xdata(xdata, &stbuf);
- /* Check if the rebalance phase1 is true */
- if (IS_DHT_MIGRATION_PHASE1 (&stbuf)) {
+ if ((!op_ret) && (ret)) {
+ /* This is a potential problem and can cause corruption
+ * with sharding.
+ * Oh well. We tried.
+ */
+ goto out;
+ }
- inode = local->loc.inode ? local->loc.inode : local->fd->inode;
- dht_inode_ctx_get_mig_info (this, inode, &src_subvol,
- &dst_subvol);
+ local->op_ret = op_ret;
+ local->rebalance.target_op_fn = dht_common_xattrop2;
+ if (xdata)
+ local->rebalance.xdata = dict_ref(xdata);
- if (dht_mig_info_is_invalid (local->cached_subvol, src_subvol,
- dst_subvol) ||
- !dht_fd_open_on_dst (this, local->fd, dst_subvol)) {
+ if (dict)
+ local->rebalance.dict = dict_ref(dict);
- ret = dht_rebalance_in_progress_check (this, frame);
- if (!ret)
- return 0;
- } else {
- dht_common_xattrop2 (this, dst_subvol, frame, 0);
- return 0;
- }
+ /* Phase 2 of migration */
+ if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2(&stbuf)) {
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret)
+ return 0;
+ }
+
+ /* Check if the rebalance phase1 is true */
+ if (IS_DHT_MIGRATION_PHASE1(&stbuf)) {
+ inode = local->loc.inode ? local->loc.inode : local->fd->inode;
+ dht_inode_ctx_get_mig_info(this, inode, &src_subvol, &dst_subvol);
+
+ if (dht_mig_info_is_invalid(local->cached_subvol, src_subvol,
+ dst_subvol) ||
+ !dht_fd_open_on_dst(this, local->fd, dst_subvol)) {
+ ret = dht_rebalance_in_progress_check(this, frame);
+ if (!ret)
+ return 0;
+ } else {
+ dht_common_xattrop2(this, dst_subvol, frame, 0);
+ return 0;
}
-
+ }
out:
- if (local->fop == GF_FOP_XATTROP) {
- DHT_STACK_UNWIND (xattrop, frame, op_ret, op_errno,
- dict, xdata);
- } else {
- DHT_STACK_UNWIND (fxattrop, frame, op_ret, op_errno,
- dict, xdata);
- }
+ if (local->fop == GF_FOP_XATTROP) {
+ DHT_STACK_UNWIND(xattrop, frame, op_ret, op_errno, dict, xdata);
+ } else {
+ DHT_STACK_UNWIND(fxattrop, frame, op_ret, op_errno, dict, xdata);
+ }
- return 0;
+ return 0;
}
-
-int
-dht_common_xattrop2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame,
- int ret)
+static int
+dht_common_xattrop2(xlator_t *this, xlator_t *subvol, call_frame_t *frame,
+ int ret)
{
- dht_local_t *local = NULL;
- int32_t op_errno = EINVAL;
-
- if ((frame == NULL) || (frame->local == NULL))
- goto out;
-
- local = frame->local;
- op_errno = local->op_errno;
-
- if (we_are_not_migrating (ret)) {
- /* This dht xlator is not migrating the file. Unwind and
- * pass on the original mode bits so the higher DHT layer
- * can handle this.
- */
- if (local->fop == GF_FOP_XATTROP) {
- DHT_STACK_UNWIND (xattrop, frame, local->op_ret,
- op_errno, local->rebalance.dict,
- local->rebalance.xdata);
- } else {
- DHT_STACK_UNWIND (fxattrop, frame, local->op_ret,
- op_errno, local->rebalance.dict,
- local->rebalance.xdata);
- }
-
- return 0;
- }
+ dht_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
- if (subvol == NULL)
- goto out;
+ if ((frame == NULL) || (frame->local == NULL))
+ goto out;
- local->call_cnt = 2; /* This is the second attempt */
+ local = frame->local;
+ op_errno = local->op_errno;
+ if (we_are_not_migrating(ret)) {
+ /* This dht xlator is not migrating the file. Unwind and
+ * pass on the original mode bits so the higher DHT layer
+ * can handle this.
+ */
if (local->fop == GF_FOP_XATTROP) {
- STACK_WIND (frame, dht_common_xattrop_cbk, subvol,
- subvol->fops->xattrop, &local->loc,
- local->rebalance.flags, local->rebalance.xattr,
- local->xattr_req);
+ DHT_STACK_UNWIND(xattrop, frame, local->op_ret, op_errno,
+ local->rebalance.dict, local->rebalance.xdata);
} else {
- STACK_WIND (frame, dht_common_xattrop_cbk, subvol,
- subvol->fops->fxattrop, local->fd,
- local->rebalance.flags, local->rebalance.xattr,
- local->xattr_req);
+ DHT_STACK_UNWIND(fxattrop, frame, local->op_ret, op_errno,
+ local->rebalance.dict, local->rebalance.xdata);
}
return 0;
+ }
+
+ if (subvol == NULL)
+ goto out;
+
+ local->call_cnt = 2; /* This is the second attempt */
+
+ if (local->fop == GF_FOP_XATTROP) {
+ STACK_WIND(frame, dht_common_xattrop_cbk, subvol, subvol->fops->xattrop,
+ &local->loc, local->rebalance.flags, local->rebalance.xattr,
+ local->xattr_req);
+ } else {
+ STACK_WIND(frame, dht_common_xattrop_cbk, subvol,
+ subvol->fops->fxattrop, local->fd, local->rebalance.flags,
+ local->rebalance.xattr, local->xattr_req);
+ }
+
+ return 0;
out:
- /* If local is unavailable we could be unwinding the wrong
- * function here */
+ /* If local is unavailable we could be unwinding the wrong
+ * function here */
- if (local && (local->fop == GF_FOP_XATTROP)) {
- DHT_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL, NULL);
- } else {
- DHT_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL, NULL);
- }
- return 0;
+ if (local && (local->fop == GF_FOP_XATTROP)) {
+ DHT_STACK_UNWIND(xattrop, frame, -1, op_errno, NULL, NULL);
+ } else {
+ DHT_STACK_UNWIND(fxattrop, frame, -1, op_errno, NULL, NULL);
+ }
+ return 0;
}
-
-int
-dht_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+static int
+dht_xattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- DHT_STACK_UNWIND (xattrop, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ DHT_STACK_UNWIND(xattrop, frame, op_ret, op_errno, dict, xdata);
+ return 0;
}
-
-int
-dht_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+/* Set both DHT_IATT_IN_XDATA_KEY and DHT_MODE_IN_XDATA_KEY
+ * Use DHT_MODE_IN_XDATA_KEY if available. Else fall back to
+ * DHT_IATT_IN_XDATA_KEY
+ */
+static int
+dht_request_iatt_in_xdata(dict_t *xattr_req)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
- int ret = -1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- local = dht_local_init (frame, loc, NULL, GF_FOP_XATTROP);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+ int ret = -1;
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for gfid=%s",
- uuid_utoa (loc->inode->gfid));
- op_errno = EINVAL;
- goto err;
- }
-
- /* Todo : Handle dirs as well. At the moment the only xlator above dht
- * that uses xattrop is sharding and that is only for files */
-
- if (IA_ISDIR (loc->inode->ia_type)) {
- STACK_WIND (frame, dht_xattrop_cbk, subvol,
- subvol->fops->xattrop, loc, flags, dict, xdata);
+ ret = dict_set_int8(xattr_req, DHT_MODE_IN_XDATA_KEY, 1);
+ ret = dict_set_int8(xattr_req, DHT_IATT_IN_XDATA_KEY, 1);
- } else {
- local->xattr_req = xdata ? dict_ref(xdata) : dict_new ();
- local->call_cnt = 1;
+ /* At least one call succeeded */
+ return ret;
+}
- local->rebalance.xattr = dict_ref (dict);
- local->rebalance.flags = flags;
+int
+dht_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+{
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+ int ret = -1;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_XATTROP);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for gfid=%s",
+ uuid_utoa(loc->inode->gfid));
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ /* Todo : Handle dirs as well. At the moment the only xlator above dht
+ * that uses xattrop is sharding and that is only for files */
+
+ if (IA_ISDIR(loc->inode->ia_type)) {
+ STACK_WIND(frame, dht_xattrop_cbk, subvol, subvol->fops->xattrop, loc,
+ flags, dict, xdata);
+
+ } else {
+ local->xattr_req = xdata ? dict_ref(xdata) : dict_new();
+ local->call_cnt = 1;
- ret = dht_request_iatt_in_xdata (this, local->xattr_req);
+ local->rebalance.xattr = dict_ref(dict);
+ local->rebalance.flags = flags;
- if (ret) {
- gf_msg_debug (this->name, 0,
- "Failed to set dictionary key %s file=%s",
- DHT_IATT_IN_XDATA_KEY, loc->path);
- }
+ ret = dht_request_iatt_in_xdata(local->xattr_req);
- STACK_WIND (frame, dht_common_xattrop_cbk, subvol,
- subvol->fops->xattrop, loc,
- local->rebalance.flags, local->rebalance.xattr,
- local->xattr_req);
+ if (ret) {
+ gf_msg_debug(this->name, 0,
+ "Failed to set dictionary key %s file=%s",
+ DHT_IATT_IN_XDATA_KEY, loc->path);
}
- return 0;
+ STACK_WIND(frame, dht_common_xattrop_cbk, subvol, subvol->fops->xattrop,
+ loc, local->rebalance.flags, local->rebalance.xattr,
+ local->xattr_req);
+ }
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(xattrop, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
-
-int
-dht_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+static int
+dht_fxattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- DHT_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ DHT_STACK_UNWIND(fxattrop, frame, op_ret, op_errno, dict, xdata);
+ return 0;
}
-
int
-dht_fxattrop (call_frame_t *frame, xlator_t *this,
- fd_t *fd, gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+dht_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
- int ret = -1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- subvol = dht_subvol_get_cached (this, fd->inode);
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for fd=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
-
- local = dht_local_init (frame, NULL, fd, GF_FOP_FXATTROP);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- /* Todo : Handle dirs as well. At the moment the only xlator above dht
- * that uses xattrop is sharding and that is only for files */
-
- if (IA_ISDIR (fd->inode->ia_type)) {
- STACK_WIND (frame, dht_fxattrop_cbk, subvol,
- subvol->fops->fxattrop, fd, flags, dict, xdata);
-
- } else {
- local->xattr_req = xdata ? dict_ref(xdata) : dict_new ();
- local->call_cnt = 1;
-
- local->rebalance.xattr = dict_ref (dict);
- local->rebalance.flags = flags;
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+ int ret = -1;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+
+ subvol = dht_subvol_get_cached(this, fd->inode);
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ local = dht_local_init(frame, NULL, fd, GF_FOP_FXATTROP);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ /* Todo : Handle dirs as well. At the moment the only xlator above dht
+ * that uses xattrop is sharding and that is only for files */
+
+ if (IA_ISDIR(fd->inode->ia_type)) {
+ STACK_WIND(frame, dht_fxattrop_cbk, subvol, subvol->fops->fxattrop, fd,
+ flags, dict, xdata);
+
+ } else {
+ local->xattr_req = xdata ? dict_ref(xdata) : dict_new();
+ local->call_cnt = 1;
- ret = dht_request_iatt_in_xdata (this, local->xattr_req);
+ local->rebalance.xattr = dict_ref(dict);
+ local->rebalance.flags = flags;
- if (ret) {
- gf_msg_debug (this->name, 0,
- "Failed to set dictionary key %s fd=%p",
- DHT_IATT_IN_XDATA_KEY, fd);
- }
+ ret = dht_request_iatt_in_xdata(local->xattr_req);
- STACK_WIND (frame, dht_common_xattrop_cbk, subvol,
- subvol->fops->fxattrop, fd,
- local->rebalance.flags, local->rebalance.xattr,
- local->xattr_req);
+ if (ret) {
+ gf_msg_debug(this->name, 0, "Failed to set dictionary key %s fd=%p",
+ DHT_IATT_IN_XDATA_KEY, fd);
}
- return 0;
+ STACK_WIND(frame, dht_common_xattrop_cbk, subvol,
+ subvol->fops->fxattrop, fd, local->rebalance.flags,
+ local->rebalance.xattr, local->xattr_req);
+ }
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(fxattrop, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
/* Currently no translators on top of 'distribute' will be using
* below fops, hence not implementing 'migration' related checks
*/
-int
-dht_inodelk_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
+static int
+dht_inodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_lk_inode_unref (frame, op_ret);
- DHT_STACK_UNWIND (inodelk, frame, op_ret, op_errno, xdata);
- return 0;
+ dht_lk_inode_unref(frame, op_ret);
+ DHT_STACK_UNWIND(inodelk, frame, op_ret, op_errno, xdata);
+ return 0;
}
-
int32_t
-dht_inodelk (call_frame_t *frame, xlator_t *this, const char *volume,
- loc_t *loc, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
+dht_inodelk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ int32_t cmd, struct gf_flock *lock, dict_t *xdata)
{
- xlator_t *lock_subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
+ xlator_t *lock_subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
+ local = dht_local_init(frame, loc, NULL, GF_FOP_INODELK);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- local = dht_local_init (frame, loc, NULL, GF_FOP_INODELK);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+ local->lock_type = lock->l_type;
+ lock_subvol = dht_get_lock_subvolume(this, lock, local);
+ if (!lock_subvol) {
+ gf_msg_debug(this->name, 0, "no lock subvolume for path=%s", loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
- local->lock_type = lock->l_type;
- lock_subvol = dht_get_lock_subvolume (this, lock, local);
- if (!lock_subvol) {
- gf_msg_debug (this->name, 0,
- "no lock subvolume for path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
+ local->call_cnt = 1;
- local->call_cnt = 1;
+ STACK_WIND(frame, dht_inodelk_cbk, lock_subvol, lock_subvol->fops->inodelk,
+ volume, loc, cmd, lock, xdata);
- STACK_WIND (frame, dht_inodelk_cbk,
- lock_subvol, lock_subvol->fops->inodelk,
- volume, loc, cmd, lock, xdata);
-
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (inodelk, frame, -1, op_errno, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(inodelk, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
int
-dht_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+dht_finodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_lk_inode_unref (frame, op_ret);
- DHT_STACK_UNWIND (finodelk, frame, op_ret, op_errno, xdata);
- return 0;
-}
+ dht_local_t *local = NULL;
+ int ret = 0;
+ GF_VALIDATE_OR_GOTO("dht", frame, out);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, out);
-int
-dht_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
- fd_t *fd, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
-{
- xlator_t *lock_subvol = NULL;
- dht_local_t *local = NULL;
- int op_errno = -1;
-
+ local = frame->local;
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- local = dht_local_init (frame, NULL, fd, GF_FOP_INODELK);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+ if (dht_check_remote_fd_failed_error(local, op_ret, op_errno)) {
+ ret = dht_check_and_open_fd_on_subvol(this, frame);
+ if (ret)
+ goto out;
+ return 0;
+ }
- local->call_cnt = 1;
- local->lock_type = lock->l_type;
-
- lock_subvol = dht_get_lock_subvolume (this, lock, local);
- if (!lock_subvol) {
- gf_msg_debug (this->name, 0,
- "no lock subvolume for fd=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
+out:
+ dht_lk_inode_unref(frame, op_ret);
+ DHT_STACK_UNWIND(finodelk, frame, op_ret, op_errno, xdata);
-/*
- local->cached_subvol = lock_subvol;
- ret = dht_check_and_open_fd_on_subvol (this, frame);
- if (ret)
- goto err;
-*/
- STACK_WIND (frame, dht_finodelk_cbk, lock_subvol,
- lock_subvol->fops->finodelk,
- volume, fd, cmd, lock, xdata);
+ return 0;
+}
- return 0;
+int
+dht_finodelk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ int32_t cmd, struct gf_flock *lock, dict_t *xdata)
+{
+ xlator_t *lock_subvol = NULL;
+ dht_local_t *local = NULL;
+ int op_errno = -1;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+
+ local = dht_local_init(frame, NULL, fd, GF_FOP_INODELK);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->call_cnt = 1;
+ local->lock_type = lock->l_type;
+
+ lock_subvol = dht_get_lock_subvolume(this, lock, local);
+ if (!lock_subvol) {
+ gf_msg_debug(this->name, 0, "no lock subvolume for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ /*
+ local->cached_subvol = lock_subvol;
+ ret = dht_check_and_open_fd_on_subvol (this, frame);
+ if (ret)
+ goto err;
+ */
+ local->rebalance.flock = *lock;
+ local->rebalance.lock_cmd = cmd;
+ local->key = gf_strdup(volume);
+
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ STACK_WIND(frame, dht_finodelk_cbk, lock_subvol,
+ lock_subvol->fops->finodelk, volume, fd, cmd, lock, xdata);
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (finodelk, frame, -1, op_errno, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(finodelk, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
diff --git a/xlators/cluster/dht/src/dht-inode-write.c b/xlators/cluster/dht/src/dht-inode-write.c
index 226ee95e8b3..2f23ce90fbd 100644
--- a/xlators/cluster/dht/src/dht-inode-write.c
+++ b/xlators/cluster/dht/src/dht-inode-write.c
@@ -8,1487 +8,1397 @@
cases as published by the Free Software Foundation.
*/
-
#include "dht-common.h"
-int dht_writev2 (xlator_t *this, xlator_t *subvol,
- call_frame_t *frame, int ret);
-int dht_truncate2 (xlator_t *this, xlator_t *subvol,
- call_frame_t *frame, int ret);
-int dht_setattr2 (xlator_t *this, xlator_t *subvol,
- call_frame_t *frame, int ret);
-int dht_fallocate2 (xlator_t *this, xlator_t *subvol,
- call_frame_t *frame, int ret);
-int dht_discard2 (xlator_t *this, xlator_t *subvol,
- call_frame_t *frame, int ret);
-int dht_zerofill2 (xlator_t *this, xlator_t *subvol,
- call_frame_t *frame, int ret);
+static int
+dht_writev2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret);
+static int
+dht_truncate2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret);
+static int
+dht_setattr2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret);
+static int
+dht_fallocate2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret);
+static int
+dht_discard2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret);
+static int
+dht_zerofill2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret);
int
-dht_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+dht_writev_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- int ret = -1;
- xlator_t *subvol1 = NULL;
- xlator_t *subvol2 = NULL;
-
- local = frame->local;
- prev = cookie;
-
- if (!local) {
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
-
- /* writev fails with EBADF if dht has not yet opened the fd
- * on the cached subvol. This could happen if the file was migrated
- * and a lookup updated the cached subvol in the inode ctx.
- * We only check once as this could be a valid bad fd error.
- */
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ int ret = -1;
+ xlator_t *subvol1 = NULL;
+ xlator_t *subvol2 = NULL;
+
+ local = frame->local;
+ prev = cookie;
+
+ if (!local) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ /* writev fails with EBADF if dht has not yet opened the fd
+ * on the cached subvol. This could happen if the file was migrated
+ * and a lookup updated the cached subvol in the inode ctx.
+ * We only check once as this could be a valid bad fd error.
+ */
+
+ if (dht_check_remote_fd_failed_error(local, op_ret, op_errno)) {
+ ret = dht_check_and_open_fd_on_subvol(this, frame);
+ if (ret)
+ goto out;
+ return 0;
+ }
- if (op_ret == -1 && (op_errno == EBADF) && !(local->fd_checked)) {
- ret = dht_check_and_open_fd_on_subvol (this, frame);
- if (ret)
- goto out;
- return 0;
+ if (op_ret == -1 && !dht_inode_missing(op_errno)) {
+ local->op_errno = op_errno;
+ local->op_ret = -1;
+ gf_msg_debug(this->name, 0, "subvolume %s returned -1 (%s)", prev->name,
+ strerror(op_errno));
+ goto out;
+ }
+
+ if (local->call_cnt != 1) {
+ /* preserve the modes of source */
+ if (local->stbuf.ia_blocks) {
+ dht_iatt_merge(this, postbuf, &local->stbuf);
+ dht_iatt_merge(this, prebuf, &local->prebuf);
}
-
- if (op_ret == -1 && !dht_inode_missing(op_errno)) {
- local->op_errno = op_errno;
+ goto out;
+ }
+
+ local->rebalance.target_op_fn = dht_writev2;
+
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+
+ /* We might need to pass the stbuf information to the higher DHT
+ * layer for appropriate handling.
+ */
+
+ dht_set_local_rebalance(this, local, NULL, prebuf, postbuf, xdata);
+
+ /* Phase 2 of migration */
+ if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2(postbuf)) {
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret)
+ return 0;
+ }
+
+ /* Check if the rebalance phase1 is true */
+ if (IS_DHT_MIGRATION_PHASE1(postbuf)) {
+ if (!local->xattr_req) {
+ local->xattr_req = dict_new();
+ if (!local->xattr_req) {
+ gf_msg(this->name, GF_LOG_ERROR, DHT_MSG_NO_MEMORY, ENOMEM,
+ "insufficient memory");
+ local->op_errno = ENOMEM;
local->op_ret = -1;
- gf_msg_debug (this->name, 0,
- "subvolume %s returned -1 (%s)",
- prev->name, strerror (op_errno));
goto out;
+ }
}
- if (local->call_cnt != 1) {
- /* preserve the modes of source */
- if (local->stbuf.ia_blocks) {
- dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
- dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
- }
- goto out;
- }
-
- local->rebalance.target_op_fn = dht_writev2;
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
-
- /* We might need to pass the stbuf information to the higher DHT
- * layer for appropriate handling.
- */
-
- dht_set_local_rebalance (this, local, NULL, prebuf, postbuf, xdata);
-
- /* Phase 2 of migration */
- if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret)
- return 0;
+ ret = dict_set_uint32(local->xattr_req, GF_PROTECT_FROM_EXTERNAL_WRITES,
+ 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, DHT_MSG_DICT_SET_FAILED, 0,
+ "Failed to set key %s in dictionary",
+ GF_PROTECT_FROM_EXTERNAL_WRITES);
+ local->op_errno = ENOMEM;
+ local->op_ret = -1;
+ goto out;
}
- /* Check if the rebalance phase1 is true */
- if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
- if (!dht_is_tier_xlator (this)) {
- if (!local->xattr_req) {
- local->xattr_req = dict_new ();
- if (!local->xattr_req) {
- gf_msg (this->name, GF_LOG_ERROR,
- DHT_MSG_NO_MEMORY,
- ENOMEM, "insufficient memory");
- local->op_errno = ENOMEM;
- local->op_ret = -1;
- goto out;
- }
- }
-
- ret = dict_set_uint32 (local->xattr_req,
- GF_PROTECT_FROM_EXTERNAL_WRITES,
- 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- DHT_MSG_DICT_SET_FAILED, 0,
- "Failed to set key %s in dictionary",
- GF_PROTECT_FROM_EXTERNAL_WRITES);
- local->op_errno = ENOMEM;
- local->op_ret = -1;
- goto out;
- }
- }
+ dht_iatt_merge(this, &local->stbuf, postbuf);
+ dht_iatt_merge(this, &local->prebuf, prebuf);
- dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
- dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
-
- ret = dht_inode_ctx_get_mig_info (this, local->fd->inode,
- &subvol1, &subvol2);
- if (!dht_mig_info_is_invalid (local->cached_subvol,
- subvol1, subvol2)) {
- if (dht_fd_open_on_dst (this, local->fd, subvol2)) {
- dht_writev2 (this, subvol2, frame, 0);
- return 0;
- }
- }
- ret = dht_rebalance_in_progress_check (this, frame);
- if (!ret)
- return 0;
+ ret = dht_inode_ctx_get_mig_info(this, local->fd->inode, &subvol1,
+ &subvol2);
+ if (!dht_mig_info_is_invalid(local->cached_subvol, subvol1, subvol2)) {
+ if (dht_fd_open_on_dst(this, local->fd, subvol2)) {
+ dht_writev2(this, subvol2, frame, 0);
+ return 0;
+ }
}
+ ret = dht_rebalance_in_progress_check(this, frame);
+ if (!ret)
+ return 0;
+ }
out:
- DHT_STRIP_PHASE1_FLAGS (postbuf);
- DHT_STRIP_PHASE1_FLAGS (prebuf);
+ DHT_STRIP_PHASE1_FLAGS(postbuf);
+ DHT_STRIP_PHASE1_FLAGS(prebuf);
- DHT_STACK_UNWIND (writev, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
+ DHT_STACK_UNWIND(writev, frame, op_ret, op_errno, prebuf, postbuf, xdata);
- return 0;
+ return 0;
}
-int
-dht_writev2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
+static int
+dht_writev2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
{
- dht_local_t *local = NULL;
- int32_t op_errno = EINVAL;
+ dht_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
- if ((frame == NULL) || (frame->local == NULL))
- goto out;
+ if ((frame == NULL) || (frame->local == NULL))
+ goto out;
- local = frame->local;
- op_errno = local->op_errno;
-
- if (we_are_not_migrating (ret)) {
- /* This dht xlator is not migrating the file. Unwind and
- * pass on the original mode bits so the higher DHT layer
- * can handle this.
- */
- DHT_STACK_UNWIND (writev, frame, local->op_ret,
- local->op_errno, &local->rebalance.prebuf,
- &local->rebalance.postbuf,
- local->rebalance.xdata);
- return 0;
- }
+ local = frame->local;
+ op_errno = local->op_errno;
- if (subvol == NULL)
- goto out;
+ if (we_are_not_migrating(ret)) {
+ /* This dht xlator is not migrating the file. Unwind and
+ * pass on the original mode bits so the higher DHT layer
+ * can handle this.
+ */
+ DHT_STACK_UNWIND(writev, frame, local->op_ret, local->op_errno,
+ &local->rebalance.prebuf, &local->rebalance.postbuf,
+ local->rebalance.xdata);
+ return 0;
+ }
- local->call_cnt = 2; /* This is the second attempt */
+ if (subvol == NULL)
+ goto out;
- STACK_WIND_COOKIE (frame, dht_writev_cbk, subvol,
- subvol, subvol->fops->writev,
- local->fd, local->rebalance.vector,
- local->rebalance.count,
- local->rebalance.offset, local->rebalance.flags,
- local->rebalance.iobref, local->xattr_req);
+ local->call_cnt = 2; /* This is the second attempt */
- return 0;
+ STACK_WIND_COOKIE(frame, dht_writev_cbk, subvol, subvol,
+ subvol->fops->writev, local->fd, local->rebalance.vector,
+ local->rebalance.count, local->rebalance.offset,
+ local->rebalance.flags, local->rebalance.iobref,
+ local->xattr_req);
+
+ return 0;
out:
- DHT_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
+ DHT_STACK_UNWIND(writev, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
int
-dht_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int count, off_t off, uint32_t flags,
- struct iobref *iobref, dict_t *xdata)
+dht_writev(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
+ int count, off_t off, uint32_t flags, struct iobref *iobref,
+ dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- local = dht_local_init (frame, NULL, fd, GF_FOP_WRITE);
- if (!local) {
-
- op_errno = ENOMEM;
- goto err;
- }
-
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for fd=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
-
- if (xdata)
- local->xattr_req = dict_ref (xdata);
-
- local->rebalance.vector = iov_dup (vector, count);
- local->rebalance.offset = off;
- local->rebalance.count = count;
- local->rebalance.flags = flags;
- local->rebalance.iobref = iobref_ref (iobref);
- local->call_cnt = 1;
-
- STACK_WIND_COOKIE (frame, dht_writev_cbk, subvol, subvol,
- subvol->fops->writev, fd,
- local->rebalance.vector,
- local->rebalance.count,
- local->rebalance.offset,
- local->rebalance.flags,
- local->rebalance.iobref, local->xattr_req);
-
- return 0;
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+
+ local = dht_local_init(frame, NULL, fd, GF_FOP_WRITE);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ local->rebalance.vector = iov_dup(vector, count);
+ local->rebalance.offset = off;
+ local->rebalance.count = count;
+ local->rebalance.flags = flags;
+ local->rebalance.iobref = iobref_ref(iobref);
+ local->call_cnt = 1;
+
+ STACK_WIND_COOKIE(frame, dht_writev_cbk, subvol, subvol,
+ subvol->fops->writev, fd, local->rebalance.vector,
+ local->rebalance.count, local->rebalance.offset,
+ local->rebalance.flags, local->rebalance.iobref,
+ local->xattr_req);
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(writev, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
-
int
-dht_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+dht_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- int ret = -1;
- xlator_t *src_subvol = NULL;
- xlator_t *dst_subvol = NULL;
- inode_t *inode = NULL;
-
-
- GF_VALIDATE_OR_GOTO ("dht", frame, err);
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
- GF_VALIDATE_OR_GOTO ("dht", cookie, out);
-
- local = frame->local;
- prev = cookie;
-
- /* Needs to be checked only for ftruncate.
- * ftruncate fails with EBADF/EINVAL if dht has not yet opened the fd
- * on the cached subvol. This could happen if the file was migrated
- * and a lookup updated the cached subvol in the inode ctx.
- * We only check once as this could actually be a valid error.
- */
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ int ret = -1;
+ xlator_t *src_subvol = NULL;
+ xlator_t *dst_subvol = NULL;
+ inode_t *inode = NULL;
+
+ GF_VALIDATE_OR_GOTO("dht", frame, err);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO("dht", cookie, out);
+
+ local = frame->local;
+ prev = cookie;
+
+ /* Needs to be checked only for ftruncate.
+ * ftruncate fails with EBADF/EINVAL if dht has not yet opened the fd
+ * on the cached subvol. This could happen if the file was migrated
+ * and a lookup updated the cached subvol in the inode ctx.
+ * We only check once as this could actually be a valid error.
+ */
+
+ if ((local->fop == GF_FOP_FTRUNCATE) &&
+ dht_check_remote_fd_failed_error(local, op_ret, op_errno)) {
+ ret = dht_check_and_open_fd_on_subvol(this, frame);
+ if (ret)
+ goto out;
+ return 0;
+ }
- if ((local->fop == GF_FOP_FTRUNCATE) && (op_ret == -1)
- && ((op_errno == EBADF) || (op_errno == EINVAL))
- && !(local->fd_checked)) {
- ret = dht_check_and_open_fd_on_subvol (this, frame);
- if (ret)
- goto out;
- return 0;
- }
+ if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
+ local->op_errno = op_errno;
+ local->op_ret = -1;
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->name);
- if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
- local->op_errno = op_errno;
- local->op_ret = -1;
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->name);
+ goto out;
+ }
- goto out;
+ if (local->call_cnt != 1) {
+ if (local->stbuf.ia_blocks) {
+ dht_iatt_merge(this, postbuf, &local->stbuf);
+ dht_iatt_merge(this, prebuf, &local->prebuf);
}
+ goto out;
+ }
+ local->rebalance.target_op_fn = dht_truncate2;
- if (local->call_cnt != 1) {
- if (local->stbuf.ia_blocks) {
- dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
- dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
- }
- goto out;
- }
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
- local->rebalance.target_op_fn = dht_truncate2;
+ /* We might need to pass the stbuf information to the higher DHT
+ * layer for appropriate handling.
+ */
- local->op_ret = op_ret;
- local->op_errno = op_errno;
+ dht_set_local_rebalance(this, local, NULL, prebuf, postbuf, xdata);
- /* We might need to pass the stbuf information to the higher DHT
- * layer for appropriate handling.
- */
+ /* Phase 2 of migration */
+ if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2(postbuf)) {
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret)
+ return 0;
+ }
- dht_set_local_rebalance (this, local, NULL, prebuf, postbuf, xdata);
+ /* Check if the rebalance phase1 is true */
+ if (IS_DHT_MIGRATION_PHASE1(postbuf)) {
+ dht_iatt_merge(this, &local->stbuf, postbuf);
+ dht_iatt_merge(this, &local->prebuf, prebuf);
- /* Phase 2 of migration */
- if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret)
- return 0;
- }
+ inode = (local->fd) ? local->fd->inode : local->loc.inode;
- /* Check if the rebalance phase1 is true */
- if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
- dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
- dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
-
- inode = (local->fd) ? local->fd->inode : local->loc.inode;
-
- dht_inode_ctx_get_mig_info (this, inode, &src_subvol,
- &dst_subvol);
- if (!dht_mig_info_is_invalid (local->cached_subvol,
- src_subvol, dst_subvol)) {
- if ((!local->fd) || ((local->fd) &&
- dht_fd_open_on_dst (this, local->fd, dst_subvol))) {
- dht_truncate2 (this, dst_subvol, frame, 0);
- return 0;
- }
- }
- ret = dht_rebalance_in_progress_check (this, frame);
- if (!ret)
- return 0;
+ dht_inode_ctx_get_mig_info(this, inode, &src_subvol, &dst_subvol);
+ if (!dht_mig_info_is_invalid(local->cached_subvol, src_subvol,
+ dst_subvol)) {
+ if ((!local->fd) ||
+ ((local->fd) &&
+ dht_fd_open_on_dst(this, local->fd, dst_subvol))) {
+ dht_truncate2(this, dst_subvol, frame, 0);
+ return 0;
+ }
}
+ ret = dht_rebalance_in_progress_check(this, frame);
+ if (!ret)
+ return 0;
+ }
out:
- DHT_STRIP_PHASE1_FLAGS (postbuf);
- DHT_STRIP_PHASE1_FLAGS (prebuf);
+ DHT_STRIP_PHASE1_FLAGS(postbuf);
+ DHT_STRIP_PHASE1_FLAGS(prebuf);
- DHT_STACK_UNWIND (truncate, frame, op_ret, op_errno,
- prebuf, postbuf, xdata);
+ DHT_STACK_UNWIND(truncate, frame, op_ret, op_errno, prebuf, postbuf, xdata);
err:
- return 0;
+ return 0;
}
-
-int
-dht_truncate2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
+static int
+dht_truncate2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
{
- dht_local_t *local = NULL;
- int32_t op_errno = EINVAL;
+ dht_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
- if (!frame || !frame->local)
- goto out;
-
- local = frame->local;
- op_errno = local->op_errno;
+ if (!frame || !frame->local)
+ goto out;
- /* This dht xlator is not migrating the file */
- if (we_are_not_migrating (ret)) {
+ local = frame->local;
+ op_errno = local->op_errno;
- DHT_STACK_UNWIND (truncate, frame, local->op_ret,
- local->op_errno, &local->rebalance.prebuf,
- &local->rebalance.postbuf,
- local->rebalance.xdata);
- return 0;
- }
+ /* This dht xlator is not migrating the file */
+ if (we_are_not_migrating(ret)) {
+ DHT_STACK_UNWIND(truncate, frame, local->op_ret, local->op_errno,
+ &local->rebalance.prebuf, &local->rebalance.postbuf,
+ local->rebalance.xdata);
+ return 0;
+ }
- if (subvol == NULL)
- goto out;
+ if (subvol == NULL)
+ goto out;
- local->call_cnt = 2; /* This is the second attempt */
+ local->call_cnt = 2; /* This is the second attempt */
- if (local->fop == GF_FOP_TRUNCATE) {
- STACK_WIND_COOKIE (frame, dht_truncate_cbk, subvol, subvol,
- subvol->fops->truncate, &local->loc,
- local->rebalance.offset, local->xattr_req);
- } else {
- STACK_WIND_COOKIE (frame, dht_truncate_cbk, subvol, subvol,
- subvol->fops->ftruncate, local->fd,
- local->rebalance.offset, local->xattr_req);
- }
+ if (local->fop == GF_FOP_TRUNCATE) {
+ STACK_WIND_COOKIE(frame, dht_truncate_cbk, subvol, subvol,
+ subvol->fops->truncate, &local->loc,
+ local->rebalance.offset, local->xattr_req);
+ } else {
+ STACK_WIND_COOKIE(frame, dht_truncate_cbk, subvol, subvol,
+ subvol->fops->ftruncate, local->fd,
+ local->rebalance.offset, local->xattr_req);
+ }
- return 0;
+ return 0;
out:
- DHT_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ DHT_STACK_UNWIND(truncate, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
int
-dht_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
- dict_t *xdata)
+dht_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- local = dht_local_init (frame, loc, NULL, GF_FOP_TRUNCATE);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local->rebalance.offset = offset;
- local->call_cnt = 1;
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for gfid=%s",
- uuid_utoa (loc->inode->gfid));
- op_errno = EINVAL;
- goto err;
- }
-
- if (xdata)
- local->xattr_req = dict_ref (xdata);
-
- STACK_WIND_COOKIE (frame, dht_truncate_cbk, subvol, subvol,
- subvol->fops->truncate, loc, offset, xdata);
-
- return 0;
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_TRUNCATE);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->rebalance.offset = offset;
+ local->call_cnt = 1;
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for gfid=%s",
+ uuid_utoa(loc->inode->gfid));
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ STACK_WIND_COOKIE(frame, dht_truncate_cbk, subvol, subvol,
+ subvol->fops->truncate, loc, offset, xdata);
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(truncate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
int
-dht_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- dict_t *xdata)
+dht_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
-
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- local = dht_local_init (frame, NULL, fd, GF_FOP_FTRUNCATE);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local->rebalance.offset = offset;
- local->call_cnt = 1;
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for fd=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
-
- if (xdata)
- local->xattr_req = dict_ref (xdata);
-
- STACK_WIND_COOKIE (frame, dht_truncate_cbk, subvol, subvol,
- subvol->fops->ftruncate, fd,
- local->rebalance.offset, local->xattr_req);
- return 0;
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+
+ local = dht_local_init(frame, NULL, fd, GF_FOP_FTRUNCATE);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->rebalance.offset = offset;
+ local->call_cnt = 1;
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ STACK_WIND_COOKIE(frame, dht_truncate_cbk, subvol, subvol,
+ subvol->fops->ftruncate, fd, local->rebalance.offset,
+ local->xattr_req);
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
int
-dht_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+dht_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- int ret = -1;
- xlator_t *src_subvol = NULL;
- xlator_t *dst_subvol = NULL;
-
- GF_VALIDATE_OR_GOTO ("dht", frame, err);
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
- GF_VALIDATE_OR_GOTO ("dht", cookie, out);
-
- local = frame->local;
- prev = cookie;
-
- /* fallocate fails with EBADF if dht has not yet opened the fd
- * on the cached subvol. This could happen if the file was migrated
- * and a lookup updated the cached subvol in the inode ctx.
- * We only check once as this could actually be a valid error.
- */
-
- if ((op_ret == -1) && (op_errno == EBADF)
- && !(local->fd_checked)) {
- ret = dht_check_and_open_fd_on_subvol (this, frame);
- if (ret)
- goto out;
- return 0;
- }
-
- if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
- local->op_errno = op_errno;
- local->op_ret = -1;
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->name);
-
- goto out;
- }
-
- if (local->call_cnt != 1) {
- if (local->stbuf.ia_blocks) {
- dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
- dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
- }
- goto out;
- }
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ int ret = -1;
+ xlator_t *src_subvol = NULL;
+ xlator_t *dst_subvol = NULL;
+
+ GF_VALIDATE_OR_GOTO("dht", frame, err);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO("dht", cookie, out);
+
+ local = frame->local;
+ prev = cookie;
+
+ /* fallocate fails with EBADF if dht has not yet opened the fd
+ * on the cached subvol. This could happen if the file was migrated
+ * and a lookup updated the cached subvol in the inode ctx.
+ * We only check once as this could actually be a valid error.
+ */
+
+ if (dht_check_remote_fd_failed_error(local, op_ret, op_errno)) {
+ ret = dht_check_and_open_fd_on_subvol(this, frame);
+ if (ret)
+ goto out;
+ return 0;
+ }
- local->op_ret = op_ret;
+ if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
local->op_errno = op_errno;
- local->rebalance.target_op_fn = dht_fallocate2;
+ local->op_ret = -1;
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->name);
- dht_set_local_rebalance (this, local, NULL, prebuf, postbuf, xdata);
+ goto out;
+ }
- /* Phase 2 of migration */
- if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret)
- return 0;
+ if (local->call_cnt != 1) {
+ if (local->stbuf.ia_blocks) {
+ dht_iatt_merge(this, postbuf, &local->stbuf);
+ dht_iatt_merge(this, prebuf, &local->prebuf);
}
-
- /* Check if the rebalance phase1 is true */
- if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
- dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
- dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
-
- dht_inode_ctx_get_mig_info (this, local->fd->inode, &src_subvol,
- &dst_subvol);
- if (!dht_mig_info_is_invalid (local->cached_subvol,
- src_subvol, dst_subvol)) {
- if (dht_fd_open_on_dst (this, local->fd, dst_subvol)) {
- dht_fallocate2 (this, dst_subvol, frame, 0);
- return 0;
- }
- }
- ret = dht_rebalance_in_progress_check (this, frame);
- if (!ret)
- return 0;
+ goto out;
+ }
+
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ local->rebalance.target_op_fn = dht_fallocate2;
+
+ dht_set_local_rebalance(this, local, NULL, prebuf, postbuf, xdata);
+
+ /* Phase 2 of migration */
+ if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2(postbuf)) {
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret)
+ return 0;
+ }
+
+ /* Check if the rebalance phase1 is true */
+ if (IS_DHT_MIGRATION_PHASE1(postbuf)) {
+ dht_iatt_merge(this, &local->stbuf, postbuf);
+ dht_iatt_merge(this, &local->prebuf, prebuf);
+
+ dht_inode_ctx_get_mig_info(this, local->fd->inode, &src_subvol,
+ &dst_subvol);
+ if (!dht_mig_info_is_invalid(local->cached_subvol, src_subvol,
+ dst_subvol)) {
+ if (dht_fd_open_on_dst(this, local->fd, dst_subvol)) {
+ dht_fallocate2(this, dst_subvol, frame, 0);
+ return 0;
+ }
}
+ ret = dht_rebalance_in_progress_check(this, frame);
+ if (!ret)
+ return 0;
+ }
out:
- DHT_STRIP_PHASE1_FLAGS (postbuf);
- DHT_STRIP_PHASE1_FLAGS (prebuf);
+ DHT_STRIP_PHASE1_FLAGS(postbuf);
+ DHT_STRIP_PHASE1_FLAGS(prebuf);
- DHT_STACK_UNWIND (fallocate, frame, op_ret, op_errno,
- prebuf, postbuf, xdata);
+ DHT_STACK_UNWIND(fallocate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
err:
- return 0;
+ return 0;
}
-int
-dht_fallocate2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
+static int
+dht_fallocate2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
{
- dht_local_t *local = NULL;
- int32_t op_errno = EINVAL;
+ dht_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
- if (!frame || !frame->local)
- goto out;
+ if (!frame || !frame->local)
+ goto out;
- local = frame->local;
- op_errno = local->op_errno;
-
- if (we_are_not_migrating (ret)) {
- /* This dht xlator is not migrating the file. Unwind and
- * pass on the original mode bits so the higher DHT layer
- * can handle this.
- */
- DHT_STACK_UNWIND (fallocate, frame, local->op_ret,
- local->op_errno,
- &local->rebalance.prebuf,
- &local->rebalance.postbuf,
- local->rebalance.xdata);
- return 0;
- }
+ local = frame->local;
+ op_errno = local->op_errno;
- if (subvol == NULL)
- goto out;
+ if (we_are_not_migrating(ret)) {
+ /* This dht xlator is not migrating the file. Unwind and
+ * pass on the original mode bits so the higher DHT layer
+ * can handle this.
+ */
+ DHT_STACK_UNWIND(fallocate, frame, local->op_ret, local->op_errno,
+ &local->rebalance.prebuf, &local->rebalance.postbuf,
+ local->rebalance.xdata);
+ return 0;
+ }
- local->call_cnt = 2; /* This is the second attempt */
+ if (subvol == NULL)
+ goto out;
- STACK_WIND_COOKIE (frame, dht_fallocate_cbk, subvol, subvol,
- subvol->fops->fallocate, local->fd,
- local->rebalance.flags, local->rebalance.offset,
- local->rebalance.size, local->xattr_req);
+ local->call_cnt = 2; /* This is the second attempt */
- return 0;
+ STACK_WIND_COOKIE(frame, dht_fallocate_cbk, subvol, subvol,
+ subvol->fops->fallocate, local->fd,
+ local->rebalance.flags, local->rebalance.offset,
+ local->rebalance.size, local->xattr_req);
+
+ return 0;
out:
- DHT_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ DHT_STACK_UNWIND(fallocate, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
int
-dht_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
- off_t offset, size_t len, dict_t *xdata)
+dht_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+ off_t offset, size_t len, dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- local = dht_local_init (frame, NULL, fd, GF_FOP_FALLOCATE);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local->rebalance.flags = mode;
- local->rebalance.offset = offset;
- local->rebalance.size = len;
-
- local->call_cnt = 1;
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for fd=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
-
- if (xdata)
- local->xattr_req = dict_ref (xdata);
-
- STACK_WIND_COOKIE (frame, dht_fallocate_cbk, subvol, subvol,
- subvol->fops->fallocate, fd,
- local->rebalance.flags,
- local->rebalance.offset,
- local->rebalance.size,
- local->xattr_req);
-
- return 0;
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+
+ local = dht_local_init(frame, NULL, fd, GF_FOP_FALLOCATE);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->rebalance.flags = mode;
+ local->rebalance.offset = offset;
+ local->rebalance.size = len;
+
+ local->call_cnt = 1;
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ STACK_WIND_COOKIE(frame, dht_fallocate_cbk, subvol, subvol,
+ subvol->fops->fallocate, fd, local->rebalance.flags,
+ local->rebalance.offset, local->rebalance.size,
+ local->xattr_req);
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(fallocate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
int
-dht_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+dht_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- int ret = -1;
- xlator_t *src_subvol = NULL;
- xlator_t *dst_subvol = NULL;
-
- GF_VALIDATE_OR_GOTO ("dht", frame, err);
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
- GF_VALIDATE_OR_GOTO ("dht", cookie, out);
-
- local = frame->local;
- prev = cookie;
-
-
- /* discard fails with EBADF if dht has not yet opened the fd
- * on the cached subvol. This could happen if the file was migrated
- * and a lookup updated the cached subvol in the inode ctx.
- * We only check once as this could actually be a valid error.
- */
- if ((op_ret == -1) && (op_errno == EBADF)
- && !(local->fd_checked)) {
- ret = dht_check_and_open_fd_on_subvol (this, frame);
- if (ret)
- goto out;
- return 0;
- }
-
- if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
- local->op_errno = op_errno;
- local->op_ret = -1;
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->name);
-
- goto out;
- }
-
- if (local->call_cnt != 1) {
- if (local->stbuf.ia_blocks) {
- dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
- dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
- }
- goto out;
- }
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ int ret = -1;
+ xlator_t *src_subvol = NULL;
+ xlator_t *dst_subvol = NULL;
+
+ GF_VALIDATE_OR_GOTO("dht", frame, err);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO("dht", cookie, out);
+
+ local = frame->local;
+ prev = cookie;
+
+ /* discard fails with EBADF if dht has not yet opened the fd
+ * on the cached subvol. This could happen if the file was migrated
+ * and a lookup updated the cached subvol in the inode ctx.
+ * We only check once as this could actually be a valid error.
+ */
+ if (dht_check_remote_fd_failed_error(local, op_ret, op_errno)) {
+ ret = dht_check_and_open_fd_on_subvol(this, frame);
+ if (ret)
+ goto out;
+ return 0;
+ }
- local->rebalance.target_op_fn = dht_discard2;
- local->op_ret = op_ret;
+ if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
local->op_errno = op_errno;
+ local->op_ret = -1;
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->name);
- dht_set_local_rebalance (this, local, NULL, prebuf, postbuf, xdata);
+ goto out;
+ }
- /* Phase 2 of migration */
- if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret)
- return 0;
+ if (local->call_cnt != 1) {
+ if (local->stbuf.ia_blocks) {
+ dht_iatt_merge(this, postbuf, &local->stbuf);
+ dht_iatt_merge(this, prebuf, &local->prebuf);
}
-
- /* Check if the rebalance phase1 is true */
- if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
- dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
- dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
-
- dht_inode_ctx_get_mig_info (this, local->fd->inode, &src_subvol,
- &dst_subvol);
- if (!dht_mig_info_is_invalid(local->cached_subvol,
- src_subvol, dst_subvol)) {
- if (dht_fd_open_on_dst (this, local->fd, dst_subvol)) {
- dht_discard2 (this, dst_subvol, frame, 0);
- return 0;
- }
- }
- ret = dht_rebalance_in_progress_check (this, frame);
- if (!ret)
- return 0;
+ goto out;
+ }
+
+ local->rebalance.target_op_fn = dht_discard2;
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+
+ dht_set_local_rebalance(this, local, NULL, prebuf, postbuf, xdata);
+
+ /* Phase 2 of migration */
+ if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2(postbuf)) {
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret)
+ return 0;
+ }
+
+ /* Check if the rebalance phase1 is true */
+ if (IS_DHT_MIGRATION_PHASE1(postbuf)) {
+ dht_iatt_merge(this, &local->stbuf, postbuf);
+ dht_iatt_merge(this, &local->prebuf, prebuf);
+
+ dht_inode_ctx_get_mig_info(this, local->fd->inode, &src_subvol,
+ &dst_subvol);
+ if (!dht_mig_info_is_invalid(local->cached_subvol, src_subvol,
+ dst_subvol)) {
+ if (dht_fd_open_on_dst(this, local->fd, dst_subvol)) {
+ dht_discard2(this, dst_subvol, frame, 0);
+ return 0;
+ }
}
+ ret = dht_rebalance_in_progress_check(this, frame);
+ if (!ret)
+ return 0;
+ }
out:
- DHT_STRIP_PHASE1_FLAGS (postbuf);
- DHT_STRIP_PHASE1_FLAGS (prebuf);
+ DHT_STRIP_PHASE1_FLAGS(postbuf);
+ DHT_STRIP_PHASE1_FLAGS(prebuf);
- DHT_STACK_UNWIND (discard, frame, op_ret, op_errno,
- prebuf, postbuf, xdata);
+ DHT_STACK_UNWIND(discard, frame, op_ret, op_errno, prebuf, postbuf, xdata);
err:
- return 0;
+ return 0;
}
-int
-dht_discard2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
+static int
+dht_discard2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
{
- dht_local_t *local = NULL;
- int32_t op_errno = EINVAL;
+ dht_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
- if (!frame || !frame->local)
- goto out;
+ if (!frame || !frame->local)
+ goto out;
- local = frame->local;
- op_errno = local->op_errno;
-
- if (we_are_not_migrating (ret)) {
- /* This dht xlator is not migrating the file. Unwind and
- * pass on the original mode bits so the higher DHT layer
- * can handle this.
- */
- DHT_STACK_UNWIND (discard, frame, local->op_ret,
- local->op_errno,
- &local->rebalance.prebuf,
- &local->rebalance.postbuf,
- local->rebalance.xdata);
- return 0;
- }
+ local = frame->local;
+ op_errno = local->op_errno;
- if (subvol == NULL)
- goto out;
+ if (we_are_not_migrating(ret)) {
+ /* This dht xlator is not migrating the file. Unwind and
+ * pass on the original mode bits so the higher DHT layer
+ * can handle this.
+ */
+ DHT_STACK_UNWIND(discard, frame, local->op_ret, local->op_errno,
+ &local->rebalance.prebuf, &local->rebalance.postbuf,
+ local->rebalance.xdata);
+ return 0;
+ }
- local->call_cnt = 2; /* This is the second attempt */
+ if (subvol == NULL)
+ goto out;
- STACK_WIND_COOKIE (frame, dht_discard_cbk, subvol, subvol,
- subvol->fops->discard, local->fd,
- local->rebalance.offset, local->rebalance.size,
- local->xattr_req);
+ local->call_cnt = 2; /* This is the second attempt */
- return 0;
+ STACK_WIND_COOKIE(frame, dht_discard_cbk, subvol, subvol,
+ subvol->fops->discard, local->fd, local->rebalance.offset,
+ local->rebalance.size, local->xattr_req);
+
+ return 0;
out:
- DHT_STACK_UNWIND (discard, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ DHT_STACK_UNWIND(discard, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
int
-dht_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- size_t len, dict_t *xdata)
+dht_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- local = dht_local_init (frame, NULL, fd, GF_FOP_DISCARD);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
- local->rebalance.offset = offset;
- local->rebalance.size = len;
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
- local->call_cnt = 1;
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for fd=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
+ local = dht_local_init(frame, NULL, fd, GF_FOP_DISCARD);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- if (xdata)
- local->xattr_req = dict_ref (xdata);
+ local->rebalance.offset = offset;
+ local->rebalance.size = len;
- STACK_WIND_COOKIE (frame, dht_discard_cbk, subvol, subvol,
- subvol->fops->discard, fd,
- local->rebalance.offset,
- local->rebalance.size,
- local->xattr_req);
+ local->call_cnt = 1;
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
- return 0;
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ STACK_WIND_COOKIE(frame, dht_discard_cbk, subvol, subvol,
+ subvol->fops->discard, fd, local->rebalance.offset,
+ local->rebalance.size, local->xattr_req);
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (discard, frame, -1, op_errno, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(discard, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
int
-dht_zerofill_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+dht_zerofill_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- int ret = -1;
- xlator_t *subvol1 = NULL, *subvol2 = NULL;
-
- GF_VALIDATE_OR_GOTO ("dht", frame, err);
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
- GF_VALIDATE_OR_GOTO ("dht", cookie, out);
-
- local = frame->local;
- prev = cookie;
-
- /* zerofill fails with EBADF if dht has not yet opened the fd
- * on the cached subvol. This could happen if the file was migrated
- * and a lookup updated the cached subvol in the inode ctx.
- * We only check once as this could actually be a valid error.
- */
- if ((op_ret == -1) && (op_errno == EBADF)
- && !(local->fd_checked)) {
- ret = dht_check_and_open_fd_on_subvol (this, frame);
- if (ret)
- goto out;
- return 0;
- }
-
- if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
- local->op_errno = op_errno;
- local->op_ret = -1;
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->name);
- goto out;
- }
-
- if (local->call_cnt != 1) {
- if (local->stbuf.ia_blocks) {
- dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
- dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
- }
- goto out;
- }
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ int ret = -1;
+ xlator_t *subvol1 = NULL, *subvol2 = NULL;
+
+ GF_VALIDATE_OR_GOTO("dht", frame, err);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, out);
+ GF_VALIDATE_OR_GOTO("dht", cookie, out);
+
+ local = frame->local;
+ prev = cookie;
+
+ /* zerofill fails with EBADF if dht has not yet opened the fd
+ * on the cached subvol. This could happen if the file was migrated
+ * and a lookup updated the cached subvol in the inode ctx.
+ * We only check once as this could actually be a valid error.
+ */
+ if (dht_check_remote_fd_failed_error(local, op_ret, op_errno)) {
+ ret = dht_check_and_open_fd_on_subvol(this, frame);
+ if (ret)
+ goto out;
+ return 0;
+ }
- local->rebalance.target_op_fn = dht_zerofill2;
- local->op_ret = op_ret;
+ if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
local->op_errno = op_errno;
-
- dht_set_local_rebalance (this, local, NULL, prebuf, postbuf, xdata);
-
- /* Phase 2 of migration */
- if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret)
- return 0;
+ local->op_ret = -1;
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->name);
+ goto out;
+ }
+
+ if (local->call_cnt != 1) {
+ if (local->stbuf.ia_blocks) {
+ dht_iatt_merge(this, postbuf, &local->stbuf);
+ dht_iatt_merge(this, prebuf, &local->prebuf);
}
-
- /* Check if the rebalance phase1 is true */
- if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
- dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
- dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
-
- ret = dht_inode_ctx_get_mig_info (this, local->fd->inode,
- &subvol1, &subvol2);
- if (!dht_mig_info_is_invalid (local->cached_subvol,
- subvol1, subvol2)) {
- if (dht_fd_open_on_dst (this, local->fd, subvol2)) {
- dht_zerofill2 (this, subvol2, frame, 0);
- return 0;
- }
- }
-
- ret = dht_rebalance_in_progress_check (this, frame);
- if (!ret)
- return 0;
+ goto out;
+ }
+
+ local->rebalance.target_op_fn = dht_zerofill2;
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+
+ dht_set_local_rebalance(this, local, NULL, prebuf, postbuf, xdata);
+
+ /* Phase 2 of migration */
+ if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2(postbuf)) {
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret)
+ return 0;
+ }
+
+ /* Check if the rebalance phase1 is true */
+ if (IS_DHT_MIGRATION_PHASE1(postbuf)) {
+ dht_iatt_merge(this, &local->stbuf, postbuf);
+ dht_iatt_merge(this, &local->prebuf, prebuf);
+
+ ret = dht_inode_ctx_get_mig_info(this, local->fd->inode, &subvol1,
+ &subvol2);
+ if (!dht_mig_info_is_invalid(local->cached_subvol, subvol1, subvol2)) {
+ if (dht_fd_open_on_dst(this, local->fd, subvol2)) {
+ dht_zerofill2(this, subvol2, frame, 0);
+ return 0;
+ }
}
+ ret = dht_rebalance_in_progress_check(this, frame);
+ if (!ret)
+ return 0;
+ }
+
out:
- DHT_STRIP_PHASE1_FLAGS (postbuf);
- DHT_STRIP_PHASE1_FLAGS (prebuf);
+ DHT_STRIP_PHASE1_FLAGS(postbuf);
+ DHT_STRIP_PHASE1_FLAGS(prebuf);
- DHT_STACK_UNWIND (zerofill, frame, op_ret, op_errno,
- prebuf, postbuf, xdata);
+ DHT_STACK_UNWIND(zerofill, frame, op_ret, op_errno, prebuf, postbuf, xdata);
err:
- return 0;
+ return 0;
}
-int
-dht_zerofill2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
+static int
+dht_zerofill2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
{
- dht_local_t *local = NULL;
- int32_t op_errno = EINVAL;
+ dht_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
- if (!frame || !frame->local)
- goto out;
+ if (!frame || !frame->local)
+ goto out;
- local = frame->local;
+ local = frame->local;
- op_errno = local->op_errno;
+ op_errno = local->op_errno;
- if (we_are_not_migrating (ret)) {
- /* This dht xlator is not migrating the file. Unwind and
- * pass on the original mode bits so the higher DHT layer
- * can handle this.
- */
- DHT_STACK_UNWIND (zerofill, frame, local->op_ret,
- local->op_errno,
- &local->rebalance.prebuf,
- &local->rebalance.postbuf,
- local->rebalance.xdata);
+ if (we_are_not_migrating(ret)) {
+ /* This dht xlator is not migrating the file. Unwind and
+ * pass on the original mode bits so the higher DHT layer
+ * can handle this.
+ */
+ DHT_STACK_UNWIND(zerofill, frame, local->op_ret, local->op_errno,
+ &local->rebalance.prebuf, &local->rebalance.postbuf,
+ local->rebalance.xdata);
- return 0;
- }
+ return 0;
+ }
- if (subvol == NULL)
- goto out;
+ if (subvol == NULL)
+ goto out;
- local->call_cnt = 2; /* This is the second attempt */
+ local->call_cnt = 2; /* This is the second attempt */
- STACK_WIND_COOKIE (frame, dht_zerofill_cbk, subvol, subvol,
- subvol->fops->zerofill,
- local->fd, local->rebalance.offset,
- local->rebalance.size, local->xattr_req);
+ STACK_WIND_COOKIE(frame, dht_zerofill_cbk, subvol, subvol,
+ subvol->fops->zerofill, local->fd,
+ local->rebalance.offset, local->rebalance.size,
+ local->xattr_req);
- return 0;
+ return 0;
out:
- DHT_STACK_UNWIND (zerofill, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ DHT_STACK_UNWIND(zerofill, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
int
-dht_zerofill (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- off_t len, dict_t *xdata)
+dht_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ off_t len, dict_t *xdata)
{
- xlator_t *subvol = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- local = dht_local_init (frame, NULL, fd, GF_FOP_ZEROFILL);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+ xlator_t *subvol = NULL;
+ int op_errno = -1;
+ dht_local_t *local = NULL;
- local->rebalance.offset = offset;
- local->rebalance.size = len;
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
- local->call_cnt = 1;
- subvol = local->cached_subvol;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for fd=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
+ local = dht_local_init(frame, NULL, fd, GF_FOP_ZEROFILL);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- if (xdata)
- local->xattr_req = dict_ref (xdata);
+ local->rebalance.offset = offset;
+ local->rebalance.size = len;
- STACK_WIND_COOKIE (frame, dht_zerofill_cbk, subvol, subvol,
- subvol->fops->zerofill, fd,
- local->rebalance.offset,
- local->rebalance.size, local->xattr_req);
+ local->call_cnt = 1;
+ subvol = local->cached_subvol;
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no cached subvolume for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
- return 0;
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
-err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (zerofill, frame, -1, op_errno, NULL, NULL, NULL);
+ STACK_WIND_COOKIE(frame, dht_zerofill_cbk, subvol, subvol,
+ subvol->fops->zerofill, fd, local->rebalance.offset,
+ local->rebalance.size, local->xattr_req);
- return 0;
-}
+ return 0;
+err:
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(zerofill, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
+}
/* handle cases of migration here for 'setattr()' calls */
int
-dht_file_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+dht_file_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- int ret = -1;
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ int ret = -1;
- local = frame->local;
- prev = cookie;
+ local = frame->local;
+ prev = cookie;
- local->op_errno = op_errno;
+ local->op_errno = op_errno;
- if ((local->fop == GF_FOP_FSETATTR) &&
- (op_ret == -1) && (op_errno == EBADF) && !(local->fd_checked)) {
- ret = dht_check_and_open_fd_on_subvol (this, frame);
- if (ret)
- goto out;
- return 0;
- }
-
- if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->name);
- goto out;
- }
-
- if (local->call_cnt != 1)
- goto out;
+ if ((local->fop == GF_FOP_FSETATTR) &&
+ dht_check_remote_fd_failed_error(local, op_ret, op_errno)) {
+ ret = dht_check_and_open_fd_on_subvol(this, frame);
+ if (ret)
+ goto out;
+ return 0;
+ }
- local->op_ret = op_ret;
- local->op_errno = op_errno;
+ if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->name);
+ goto out;
+ }
- local->rebalance.target_op_fn = dht_setattr2;
+ if (local->call_cnt != 1)
+ goto out;
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
- /* Phase 2 of migration */
- if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
+ local->rebalance.target_op_fn = dht_setattr2;
- dht_set_local_rebalance (this, local, NULL, prebuf,
- postbuf, xdata);
+ /* Phase 2 of migration */
+ if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2(postbuf)) {
+ dht_set_local_rebalance(this, local, NULL, prebuf, postbuf, xdata);
- ret = dht_rebalance_complete_check (this, frame);
- if (!ret)
- return 0;
- }
+ ret = dht_rebalance_complete_check(this, frame);
+ if (!ret)
+ return 0;
+ }
- /* At the end of the migration process, whatever 'attr' we
- have on source file will be migrated to destination file
- in one shot, hence we don't need to check for in progress
- state here (ie, PHASE1) */
+ /* At the end of the migration process, whatever 'attr' we
+ have on source file will be migrated to destination file
+ in one shot, hence we don't need to check for in progress
+ state here (ie, PHASE1) */
out:
- DHT_STRIP_PHASE1_FLAGS (postbuf);
- DHT_STRIP_PHASE1_FLAGS (prebuf);
+ DHT_STRIP_PHASE1_FLAGS(postbuf);
+ DHT_STRIP_PHASE1_FLAGS(prebuf);
- DHT_STACK_UNWIND (setattr, frame, op_ret, op_errno,
- prebuf, postbuf, xdata);
+ DHT_STACK_UNWIND(setattr, frame, op_ret, op_errno, prebuf, postbuf, xdata);
- return 0;
+ return 0;
}
-int
-dht_setattr2 (xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
+static int
+dht_setattr2(xlator_t *this, xlator_t *subvol, call_frame_t *frame, int ret)
{
- dht_local_t *local = NULL;
- int32_t op_errno = EINVAL;
+ dht_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
- if (!frame || !frame->local)
- goto out;
+ if (!frame || !frame->local)
+ goto out;
- local = frame->local;
- op_errno = local->op_errno;
-
- if (we_are_not_migrating (ret)) {
- /* This dht xlator is not migrating the file. Unwind and
- * pass on the original mode bits so the higher DHT layer
- * can handle this.
- */
- DHT_STACK_UNWIND (setattr, frame, local->op_ret,
- local->op_errno,
- &local->rebalance.prebuf,
- &local->rebalance.postbuf,
- local->rebalance.xdata);
- return 0;
- }
+ local = frame->local;
+ op_errno = local->op_errno;
- if (subvol == NULL)
- goto out;
+ if (we_are_not_migrating(ret)) {
+ /* This dht xlator is not migrating the file. Unwind and
+ * pass on the original mode bits so the higher DHT layer
+ * can handle this.
+ */
+ DHT_STACK_UNWIND(setattr, frame, local->op_ret, local->op_errno,
+ &local->rebalance.prebuf, &local->rebalance.postbuf,
+ local->rebalance.xdata);
+ return 0;
+ }
- local->call_cnt = 2; /* This is the second attempt */
-
- if (local->fop == GF_FOP_SETATTR) {
- STACK_WIND_COOKIE (frame, dht_file_setattr_cbk, subvol,
- subvol, subvol->fops->setattr, &local->loc,
- &local->rebalance.stbuf, local->rebalance.flags,
- local->xattr_req);
- } else {
- STACK_WIND_COOKIE (frame, dht_file_setattr_cbk, subvol,
- subvol, subvol->fops->fsetattr, local->fd,
- &local->rebalance.stbuf, local->rebalance.flags,
- local->xattr_req);
- }
+ if (subvol == NULL)
+ goto out;
- return 0;
+ local->call_cnt = 2; /* This is the second attempt */
+
+ if (local->fop == GF_FOP_SETATTR) {
+ STACK_WIND_COOKIE(frame, dht_file_setattr_cbk, subvol, subvol,
+ subvol->fops->setattr, &local->loc,
+ &local->rebalance.stbuf, local->rebalance.flags,
+ local->xattr_req);
+ } else {
+ STACK_WIND_COOKIE(frame, dht_file_setattr_cbk, subvol, subvol,
+ subvol->fops->fsetattr, local->fd,
+ &local->rebalance.stbuf, local->rebalance.flags,
+ local->xattr_req);
+ }
+
+ return 0;
out:
- DHT_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ DHT_STACK_UNWIND(setattr, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
-
/* Keep the existing code same for all the cases other than regular file */
int
-dht_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *statpre,
- struct iatt *statpost, dict_t *xdata)
+dht_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *statpre, struct iatt *statpost,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- xlator_t *prev = NULL;
-
-
- local = frame->local;
- prev = cookie;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->name);
- goto unlock;
- }
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ xlator_t *prev = NULL;
- dht_iatt_merge (this, &local->prebuf, statpre, prev);
- dht_iatt_merge (this, &local->stbuf, statpost, prev);
+ local = frame->local;
+ prev = cookie;
- local->op_ret = 0;
- local->op_errno = 0;
+ LOCK(&frame->lock);
+ {
+ if (op_ret == -1) {
+ local->op_errno = op_errno;
+ UNLOCK(&frame->lock);
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->name);
+ goto post_unlock;
}
-unlock:
- UNLOCK (&frame->lock);
-
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- if (local->op_ret == 0)
- dht_inode_ctx_time_set (local->loc.inode, this,
- &local->stbuf);
- DHT_STACK_UNWIND (setattr, frame, local->op_ret, local->op_errno,
- &local->prebuf, &local->stbuf, xdata);
- }
- return 0;
-}
+ dht_iatt_merge(this, &local->prebuf, statpre);
+ dht_iatt_merge(this, &local->stbuf, statpost);
+ local->op_ret = 0;
+ local->op_errno = 0;
+ }
+ UNLOCK(&frame->lock);
+post_unlock:
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt)) {
+ if (local->op_ret == 0)
+ dht_inode_ctx_time_set(local->loc.inode, this, &local->stbuf);
+ DHT_STACK_UNWIND(setattr, frame, local->op_ret, local->op_errno,
+ &local->prebuf, &local->stbuf, xdata);
+ }
+
+ return 0;
+}
/* Keep the existing code same for all the cases other than regular file */
int
-dht_non_mds_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *statpre,
- struct iatt *statpost, dict_t *xdata)
+dht_non_mds_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- xlator_t *prev = NULL;
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ xlator_t *prev = NULL;
+ local = frame->local;
+ prev = cookie;
- local = frame->local;
- prev = cookie;
+ if (op_ret == -1) {
+ gf_msg(this->name, op_errno, 0, 0, "subvolume %s returned -1",
+ prev->name);
+ goto post_unlock;
+ }
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- gf_msg (this->name, op_errno, 0,
- 0, "subvolume %s returned -1",
- prev->name);
+ LOCK(&frame->lock);
+ {
+ dht_iatt_merge(this, &local->prebuf, statpre);
+ dht_iatt_merge(this, &local->stbuf, statpost);
- goto unlock;
- }
-
- dht_iatt_merge (this, &local->prebuf, statpre, prev);
- dht_iatt_merge (this, &local->stbuf, statpost, prev);
-
- local->op_ret = 0;
- local->op_errno = 0;
- }
-unlock:
- UNLOCK (&frame->lock);
-
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- dht_inode_ctx_time_set (local->loc.inode, this, &local->stbuf);
- DHT_STACK_UNWIND (setattr, frame, 0, 0,
- &local->prebuf, &local->stbuf, xdata);
- }
-
- return 0;
+ local->op_ret = 0;
+ local->op_errno = 0;
+ }
+ UNLOCK(&frame->lock);
+post_unlock:
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt)) {
+ dht_inode_ctx_time_set(local->loc.inode, this, &local->stbuf);
+ DHT_STACK_UNWIND(setattr, frame, 0, 0, &local->prebuf, &local->stbuf,
+ xdata);
+ }
+
+ return 0;
}
-
-
-
-
int
-dht_mds_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *statpre,
- struct iatt *statpost, dict_t *xdata)
+dht_mds_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- xlator_t *prev = NULL;
- xlator_t *mds_subvol = NULL;
- struct iatt loc_stbuf = {0,};
- int i = 0;
-
- local = frame->local;
- prev = cookie;
- conf = this->private;
- mds_subvol = local->mds_subvol;
-
- if (op_ret == -1) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- gf_msg_debug (this->name, op_errno,
- "subvolume %s returned -1",
- prev->name);
- goto out;
- }
-
- local->op_ret = 0;
- loc_stbuf = local->stbuf;
- dht_iatt_merge (this, &local->prebuf, statpre, prev);
- dht_iatt_merge (this, &local->stbuf, statpost, prev);
-
- local->call_cnt = conf->subvolume_cnt - 1;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (mds_subvol == conf->subvolumes[i])
- continue;
- STACK_WIND_COOKIE (frame, dht_non_mds_setattr_cbk,
- conf->subvolumes[i], conf->subvolumes[i],
- conf->subvolumes[i]->fops->setattr,
- &local->loc, &loc_stbuf,
- local->valid, local->xattr_req);
- }
-
- return 0;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ xlator_t *prev = NULL;
+ xlator_t *mds_subvol = NULL;
+ struct iatt loc_stbuf = {
+ 0,
+ };
+ int i = 0;
+
+ local = frame->local;
+ prev = cookie;
+ conf = this->private;
+ mds_subvol = local->mds_subvol;
+
+ if (op_ret == -1) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ gf_msg_debug(this->name, op_errno, "subvolume %s returned -1",
+ prev->name);
+ goto out;
+ }
+
+ local->op_ret = 0;
+ loc_stbuf = local->stbuf;
+ dht_iatt_merge(this, &local->prebuf, statpre);
+ dht_iatt_merge(this, &local->stbuf, statpost);
+
+ local->call_cnt = conf->subvolume_cnt - 1;
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (mds_subvol == conf->subvolumes[i])
+ continue;
+ STACK_WIND_COOKIE(frame, dht_non_mds_setattr_cbk, conf->subvolumes[i],
+ conf->subvolumes[i],
+ conf->subvolumes[i]->fops->setattr, &local->loc,
+ &loc_stbuf, local->valid, local->xattr_req);
+ }
+
+ return 0;
out:
- DHT_STACK_UNWIND (setattr, frame, local->op_ret, local->op_errno,
- &local->prebuf, &local->stbuf, xdata);
+ DHT_STACK_UNWIND(setattr, frame, local->op_ret, local->op_errno,
+ &local->prebuf, &local->stbuf, xdata);
- return 0;
+ return 0;
}
int
-dht_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+dht_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *stbuf,
+ int32_t valid, dict_t *xdata)
{
- xlator_t *subvol = NULL;
- xlator_t *mds_subvol = NULL;
- dht_layout_t *layout = NULL;
- dht_local_t *local = NULL;
- int op_errno = -1;
- int i = -1;
- int ret = -1;
- int call_cnt = 0;
- dht_conf_t *conf = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
-
- conf = this->private;
- local = dht_local_init (frame, loc, NULL, GF_FOP_SETATTR);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- layout = local->layout;
- if (!layout) {
- gf_msg_debug (this->name, 0,
- "no layout for path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
-
- if (!layout_is_sane (layout)) {
- gf_msg_debug (this->name, 0,
- "layout is not sane for path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
- if (xdata)
- local->xattr_req = dict_ref (xdata);
-
- if (IA_ISREG (loc->inode->ia_type)) {
- /* in the regular file _cbk(), we need to check for
- migration possibilities */
- local->rebalance.stbuf = *stbuf;
- local->rebalance.flags = valid;
- local->call_cnt = 1;
- subvol = local->cached_subvol;
+ xlator_t *subvol = NULL;
+ xlator_t *mds_subvol = NULL;
+ dht_layout_t *layout = NULL;
+ dht_local_t *local = NULL;
+ int op_errno = -1;
+ int i = -1;
+ int ret = -1;
+ int call_cnt = 0;
+ dht_conf_t *conf = NULL;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
+ VALIDATE_OR_GOTO(loc->path, err);
+
+ conf = this->private;
+ local = dht_local_init(frame, loc, NULL, GF_FOP_SETATTR);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ layout = local->layout;
+ if (!layout) {
+ gf_msg_debug(this->name, 0, "no layout for path=%s", loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ if (!layout_is_sane(layout)) {
+ gf_msg_debug(this->name, 0, "layout is not sane for path=%s",
+ loc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ if (IA_ISREG(loc->inode->ia_type)) {
+ /* in the regular file _cbk(), we need to check for
+ migration possibilities */
+ local->rebalance.stbuf = *stbuf;
+ local->rebalance.flags = valid;
+ local->call_cnt = 1;
+ subvol = local->cached_subvol;
- STACK_WIND_COOKIE (frame, dht_file_setattr_cbk, subvol,
- subvol, subvol->fops->setattr, loc, stbuf,
- valid, xdata);
+ STACK_WIND_COOKIE(frame, dht_file_setattr_cbk, subvol, subvol,
+ subvol->fops->setattr, loc, stbuf, valid, xdata);
- return 0;
+ return 0;
+ }
+
+ local->call_cnt = call_cnt = layout->cnt;
+
+ if (IA_ISDIR(loc->inode->ia_type) && !__is_root_gfid(loc->inode->gfid) &&
+ call_cnt != 1) {
+ ret = dht_inode_ctx_mdsvol_get(loc->inode, this, &mds_subvol);
+ if (ret || !mds_subvol) {
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ DHT_MSG_HASHED_SUBVOL_GET_FAILED,
+ "Failed to get mds subvol for path %s", local->loc.path);
+ op_errno = EINVAL;
+ goto err;
}
- local->call_cnt = call_cnt = layout->cnt;
-
- if (IA_ISDIR (loc->inode->ia_type) &&
- !__is_root_gfid (loc->inode->gfid) && call_cnt != 1) {
- ret = dht_inode_ctx_mdsvol_get (loc->inode, this, &mds_subvol);
- if (ret || !mds_subvol) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_HASHED_SUBVOL_GET_FAILED,
- "Failed to get mds subvol for path %s",
- local->loc.path);
- op_errno = EINVAL;
- goto err;
- }
-
- local->mds_subvol = mds_subvol;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->subvolumes[i] == mds_subvol) {
- if (!conf->subvolume_status[i]) {
- gf_msg (this->name, GF_LOG_WARNING,
- layout->list[i].err,
- DHT_MSG_HASHED_SUBVOL_DOWN,
- "MDS subvol is down for path "
- " %s Unable to set attr " ,
- local->loc.path);
- op_errno = ENOTCONN;
- goto err;
- }
- }
- }
- local->valid = valid;
- local->stbuf = *stbuf;
-
- STACK_WIND_COOKIE (frame, dht_mds_setattr_cbk,
- local->mds_subvol,
- local->mds_subvol,
- local->mds_subvol->fops->setattr,
- loc, stbuf, valid, xdata);
- return 0;
- } else {
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND_COOKIE (frame, dht_setattr_cbk,
- layout->list[i].xlator,
- layout->list[i].xlator,
- layout->list[i].xlator->fops->setattr,
- loc, stbuf, valid, xdata);
+ local->mds_subvol = mds_subvol;
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->subvolumes[i] == mds_subvol) {
+ if (!conf->subvolume_status[i]) {
+ gf_msg(this->name, GF_LOG_WARNING, layout->list[i].err,
+ DHT_MSG_HASHED_SUBVOL_DOWN,
+ "MDS subvol is down for path "
+ " %s Unable to set attr ",
+ local->loc.path);
+ op_errno = ENOTCONN;
+ goto err;
}
+ }
}
+ local->valid = valid;
+ local->stbuf = *stbuf;
+ STACK_WIND_COOKIE(frame, dht_mds_setattr_cbk, local->mds_subvol,
+ local->mds_subvol, local->mds_subvol->fops->setattr,
+ loc, stbuf, valid, xdata);
return 0;
+ } else {
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND_COOKIE(frame, dht_setattr_cbk, layout->list[i].xlator,
+ layout->list[i].xlator,
+ layout->list[i].xlator->fops->setattr, loc, stbuf,
+ valid, xdata);
+ }
+ }
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(setattr, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
int
-dht_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
- int32_t valid, dict_t *xdata)
+dht_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
+ int32_t valid, dict_t *xdata)
{
- xlator_t *subvol = NULL;
- dht_layout_t *layout = NULL;
- dht_local_t *local = NULL;
- int op_errno = -1;
- int i = -1;
- int call_cnt = 0;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- local = dht_local_init (frame, NULL, fd, GF_FOP_FSETATTR);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- layout = local->layout;
- if (!layout) {
- gf_msg_debug (this->name, 0,
- "no layout for fd=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
+ xlator_t *subvol = NULL;
+ dht_layout_t *layout = NULL;
+ dht_local_t *local = NULL;
+ int op_errno = -1;
+ int i = -1;
+ int call_cnt = 0;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+
+ local = dht_local_init(frame, NULL, fd, GF_FOP_FSETATTR);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ layout = local->layout;
+ if (!layout) {
+ gf_msg_debug(this->name, 0, "no layout for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ if (!layout_is_sane(layout)) {
+ gf_msg_debug(this->name, 0, "layout is not sane for fd=%p", fd);
+ op_errno = EINVAL;
+ goto err;
+ }
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ if (IA_ISREG(fd->inode->ia_type)) {
+ /* in the regular file _cbk(), we need to check for
+ migration possibilities */
+ local->rebalance.stbuf = *stbuf;
+ local->rebalance.flags = valid;
+ local->call_cnt = 1;
+ subvol = local->cached_subvol;
- if (!layout_is_sane (layout)) {
- gf_msg_debug (this->name, 0,
- "layout is not sane for fd=%p", fd);
- op_errno = EINVAL;
- goto err;
- }
- if (xdata)
- local->xattr_req = dict_ref (xdata);
-
- if (IA_ISREG (fd->inode->ia_type)) {
- /* in the regular file _cbk(), we need to check for
- migration possibilities */
- local->rebalance.stbuf = *stbuf;
- local->rebalance.flags = valid;
- local->call_cnt = 1;
- subvol = local->cached_subvol;
-
- STACK_WIND_COOKIE (frame, dht_file_setattr_cbk, subvol,
- subvol, subvol->fops->fsetattr, fd,
- &local->rebalance.stbuf,
- local->rebalance.flags,
- local->xattr_req);
- return 0;
- }
+ STACK_WIND_COOKIE(frame, dht_file_setattr_cbk, subvol, subvol,
+ subvol->fops->fsetattr, fd, &local->rebalance.stbuf,
+ local->rebalance.flags, local->xattr_req);
+ return 0;
+ }
- local->call_cnt = call_cnt = layout->cnt;
+ local->call_cnt = call_cnt = layout->cnt;
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND_COOKIE (frame, dht_setattr_cbk,
- layout->list[i].xlator,
- layout->list[i].xlator,
- layout->list[i].xlator->fops->fsetattr,
- fd, stbuf, valid, xdata);
- }
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND_COOKIE(frame, dht_setattr_cbk, layout->list[i].xlator,
+ layout->list[i].xlator,
+ layout->list[i].xlator->fops->fsetattr, fd, stbuf,
+ valid, xdata);
+ }
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
diff --git a/xlators/cluster/dht/src/dht-layout.c b/xlators/cluster/dht/src/dht-layout.c
index 97b98e01451..fda904c92c9 100644
--- a/xlators/cluster/dht/src/dht-layout.c
+++ b/xlators/cluster/dht/src/dht-layout.c
@@ -8,870 +8,801 @@
cases as published by the Free Software Foundation.
*/
-
-#include "glusterfs.h"
-#include "xlator.h"
#include "dht-common.h"
-#include "byte-order.h"
-#include "dht-messages.h"
+#include <glusterfs/byte-order.h>
#include "unittest/unittest.h"
+#define layout_base_size (sizeof(dht_layout_t))
-#define layout_base_size (sizeof (dht_layout_t))
-
-#define layout_entry_size (sizeof ((dht_layout_t *)NULL)->list[0])
+#define layout_entry_size (sizeof((dht_layout_t *)NULL)->list[0])
#define layout_size(cnt) (layout_base_size + (cnt * layout_entry_size))
dht_layout_t *
-dht_layout_new (xlator_t *this, int cnt)
+dht_layout_new(xlator_t *this, int cnt)
{
- dht_layout_t *layout = NULL;
- dht_conf_t *conf = NULL;
+ dht_layout_t *layout = NULL;
+ dht_conf_t *conf = NULL;
- REQUIRE(NULL != this);
- REQUIRE(cnt >= 0);
+ REQUIRE(NULL != this);
+ REQUIRE(cnt >= 0);
- conf = this->private;
+ conf = this->private;
- layout = GF_CALLOC (1, layout_size (cnt),
- gf_dht_mt_dht_layout_t);
- if (!layout) {
- goto out;
- }
+ layout = GF_CALLOC(1, layout_size(cnt), gf_dht_mt_dht_layout_t);
+ if (!layout) {
+ goto out;
+ }
- layout->type = DHT_HASH_TYPE_DM;
- layout->cnt = cnt;
+ layout->type = DHT_HASH_TYPE_DM;
+ layout->cnt = cnt;
- if (conf) {
- layout->spread_cnt = conf->dir_spread_cnt;
- layout->gen = conf->gen;
- }
+ if (conf) {
+ layout->spread_cnt = conf->dir_spread_cnt;
+ layout->gen = conf->gen;
+ }
- GF_ATOMIC_INIT (layout->ref, 1);
+ GF_ATOMIC_INIT(layout->ref, 1);
- ENSURE(NULL != layout);
- ENSURE(layout->type == DHT_HASH_TYPE_DM);
- ENSURE(layout->cnt == cnt);
- ENSURE(GF_ATOMIC_GET (layout->ref) == 1);
+ ENSURE(NULL != layout);
+ ENSURE(layout->type == DHT_HASH_TYPE_DM);
+ ENSURE(layout->cnt == cnt);
+ ENSURE(GF_ATOMIC_GET(layout->ref) == 1);
out:
- return layout;
+ return layout;
}
-
dht_layout_t *
-dht_layout_get (xlator_t *this, inode_t *inode)
+dht_layout_get(xlator_t *this, inode_t *inode)
{
- dht_layout_t *layout = NULL;
- int ret = 0;
-
- ret = dht_inode_ctx_layout_get (inode, this, &layout);
- if ((!ret) && layout) {
- GF_ATOMIC_INC (layout->ref);
- }
- return layout;
+ dht_layout_t *layout = NULL;
+ int ret = 0;
+
+ ret = dht_inode_ctx_layout_get(inode, this, &layout);
+ if ((!ret) && layout) {
+ GF_ATOMIC_INC(layout->ref);
+ }
+ return layout;
}
-
int
-dht_layout_set (xlator_t *this, inode_t *inode, dht_layout_t *layout)
+dht_layout_set(xlator_t *this, inode_t *inode, dht_layout_t *layout)
{
- dht_conf_t *conf = NULL;
- int oldret = -1;
- int ret = -1;
- dht_layout_t *old_layout;
-
- conf = this->private;
- if (!conf || !layout)
- goto out;
-
- LOCK (&conf->layout_lock);
- {
- oldret = dht_inode_ctx_layout_get (inode, this, &old_layout);
- if (layout)
- GF_ATOMIC_INC (layout->ref);
- ret = dht_inode_ctx_layout_set (inode, this, layout);
- }
- UNLOCK (&conf->layout_lock);
-
- if (!oldret) {
- dht_layout_unref (this, old_layout);
- }
- if (ret)
- GF_ATOMIC_DEC (layout->ref);
+ dht_conf_t *conf = NULL;
+ int oldret = -1;
+ int ret = -1;
+ dht_layout_t *old_layout;
+
+ conf = this->private;
+ if (!conf || !layout)
+ goto out;
+
+ LOCK(&conf->layout_lock);
+ {
+ oldret = dht_inode_ctx_layout_get(inode, this, &old_layout);
+ if (layout)
+ GF_ATOMIC_INC(layout->ref);
+ ret = dht_inode_ctx_layout_set(inode, this, layout);
+ }
+ UNLOCK(&conf->layout_lock);
+
+ if (!oldret) {
+ dht_layout_unref(this, old_layout);
+ }
+ if (ret)
+ GF_ATOMIC_DEC(layout->ref);
out:
- return ret;
+ return ret;
}
-
void
-dht_layout_unref (xlator_t *this, dht_layout_t *layout)
+dht_layout_unref(xlator_t *this, dht_layout_t *layout)
{
- int ref = 0;
+ int ref = 0;
- if (!layout || layout->preset || !this->private)
- return;
+ if (!layout || layout->preset || !this->private)
+ return;
- ref = GF_ATOMIC_DEC (layout->ref);
+ ref = GF_ATOMIC_DEC(layout->ref);
- if (!ref)
- GF_FREE (layout);
+ if (!ref)
+ GF_FREE(layout);
}
-
dht_layout_t *
-dht_layout_ref (xlator_t *this, dht_layout_t *layout)
+dht_layout_ref(xlator_t *this, dht_layout_t *layout)
{
- if (layout->preset || !this->private)
- return layout;
+ if (layout->preset || !this->private)
+ return layout;
- GF_ATOMIC_INC (layout->ref);
+ GF_ATOMIC_INC(layout->ref);
- return layout;
+ return layout;
}
-
xlator_t *
-dht_layout_search (xlator_t *this, dht_layout_t *layout, const char *name)
+dht_layout_search(xlator_t *this, dht_layout_t *layout, const char *name)
{
- uint32_t hash = 0;
- xlator_t *subvol = NULL;
- int i = 0;
- int ret = 0;
-
- ret = dht_hash_compute (this, layout->type, name, &hash);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_COMPUTE_HASH_FAILED,
- "hash computation failed for type=%d name=%s",
- layout->type, name);
- goto out;
- }
-
- for (i = 0; i < layout->cnt; i++) {
- if (layout->list[i].start <= hash
- && layout->list[i].stop >= hash) {
- subvol = layout->list[i].xlator;
- break;
- }
- }
-
- if (!subvol) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_HASHED_SUBVOL_GET_FAILED,
- "no subvolume for hash (value) = %u", hash);
- }
+ uint32_t hash = 0;
+ xlator_t *subvol = NULL;
+ int i = 0;
+ int ret = 0;
+
+ ret = dht_hash_compute(this, layout->type, name, &hash);
+ if (ret != 0) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_COMPUTE_HASH_FAILED,
+ "type=%d", layout->type, "name=%s", name, NULL);
+ goto out;
+ }
+
+ for (i = 0; i < layout->cnt; i++) {
+ if (layout->list[i].start <= hash && layout->list[i].stop >= hash) {
+ subvol = layout->list[i].xlator;
+ break;
+ }
+ }
+
+ if (!subvol) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_HASHED_SUBVOL_GET_FAILED,
+ "hash-value=0x%x", hash, NULL);
+ }
out:
- return subvol;
+ return subvol;
}
-
dht_layout_t *
-dht_layout_for_subvol (xlator_t *this, xlator_t *subvol)
+dht_layout_for_subvol(xlator_t *this, xlator_t *subvol)
{
- dht_conf_t *conf = NULL;
- dht_layout_t *layout = NULL;
- int i = 0;
-
- conf = this->private;
- if (!conf)
- goto out;
-
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->subvolumes[i] == subvol) {
- layout = conf->file_layouts[i];
- break;
- }
+ dht_conf_t *conf = NULL;
+ dht_layout_t *layout = NULL;
+ int i = 0;
+
+ conf = this->private;
+ if (!conf)
+ goto out;
+
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->subvolumes[i] == subvol) {
+ layout = conf->file_layouts[i];
+ break;
}
+ }
out:
- return layout;
+ return layout;
}
-
int
-dht_layouts_init (xlator_t *this, dht_conf_t *conf)
+dht_layouts_init(xlator_t *this, dht_conf_t *conf)
{
- dht_layout_t *layout = NULL;
- int i = 0;
- int ret = -1;
-
- if (!conf)
- goto out;
-
- conf->file_layouts = GF_CALLOC (conf->subvolume_cnt,
- sizeof (dht_layout_t *),
- gf_dht_mt_dht_layout_t);
- if (!conf->file_layouts) {
- goto out;
- }
+ dht_layout_t *layout = NULL;
+ int i = 0;
+ int ret = -1;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- layout = dht_layout_new (this, 1);
+ if (!conf)
+ goto out;
- if (!layout) {
- goto out;
- }
+ conf->file_layouts = GF_CALLOC(conf->subvolume_cnt, sizeof(dht_layout_t *),
+ gf_dht_mt_dht_layout_t);
+ if (!conf->file_layouts) {
+ goto out;
+ }
- layout->preset = 1;
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ layout = dht_layout_new(this, 1);
- layout->list[0].xlator = conf->subvolumes[i];
-
- conf->file_layouts[i] = layout;
+ if (!layout) {
+ goto out;
}
- ret = 0;
+ layout->preset = 1;
+
+ layout->list[0].xlator = conf->subvolumes[i];
+
+ conf->file_layouts[i] = layout;
+ }
+
+ ret = 0;
out:
- return ret;
+ return ret;
}
-
int
-dht_disk_layout_extract (xlator_t *this, dht_layout_t *layout,
- int pos, int32_t **disk_layout_p)
+dht_disk_layout_extract(xlator_t *this, dht_layout_t *layout, int pos,
+ int32_t **disk_layout_p)
{
- int ret = -1;
- int32_t *disk_layout = NULL;
+ int ret = -1;
+ int32_t *disk_layout = NULL;
- disk_layout = GF_CALLOC (5, sizeof (int),
- gf_dht_mt_int32_t);
- if (!disk_layout) {
- goto out;
- }
+ disk_layout = GF_CALLOC(5, sizeof(int), gf_dht_mt_int32_t);
+ if (!disk_layout) {
+ goto out;
+ }
- disk_layout[0] = hton32 (layout->list[pos].commit_hash);
- disk_layout[1] = hton32 (layout->type);
- disk_layout[2] = hton32 (layout->list[pos].start);
- disk_layout[3] = hton32 (layout->list[pos].stop);
+ disk_layout[0] = hton32(layout->list[pos].commit_hash);
+ disk_layout[1] = hton32(layout->type);
+ disk_layout[2] = hton32(layout->list[pos].start);
+ disk_layout[3] = hton32(layout->list[pos].stop);
- if (disk_layout_p)
- *disk_layout_p = disk_layout;
- else
- GF_FREE (disk_layout);
+ if (disk_layout_p)
+ *disk_layout_p = disk_layout;
+ else
+ GF_FREE(disk_layout);
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-dht_disk_layout_extract_for_subvol (xlator_t *this, dht_layout_t *layout,
- xlator_t *subvol, int32_t **disk_layout_p)
+dht_disk_layout_extract_for_subvol(xlator_t *this, dht_layout_t *layout,
+ xlator_t *subvol, int32_t **disk_layout_p)
{
- int i = 0;
+ int i = 0;
- for (i = 0; i < layout->cnt; i++) {
- if (layout->list[i].xlator == subvol)
- break;
- }
+ for (i = 0; i < layout->cnt; i++) {
+ if (layout->list[i].xlator == subvol)
+ break;
+ }
- if (i == layout->cnt)
- return -1;
+ if (i == layout->cnt)
+ return -1;
- return dht_disk_layout_extract (this, layout, i, disk_layout_p);
+ return dht_disk_layout_extract(this, layout, i, disk_layout_p);
}
-int
-dht_disk_layout_merge (xlator_t *this, dht_layout_t *layout,
- int pos, void *disk_layout_raw, int disk_layout_len)
+static int
+dht_disk_layout_merge(xlator_t *this, dht_layout_t *layout, int pos,
+ void *disk_layout_raw, int disk_layout_len)
{
- int type = 0;
- int start_off = 0;
- int stop_off = 0;
- int commit_hash = 0;
- int disk_layout[4];
+ int type = 0;
+ int start_off = 0;
+ int stop_off = 0;
+ int commit_hash = 0;
+ int disk_layout[4];
- if (!disk_layout_raw) {
- gf_msg (this->name, GF_LOG_CRITICAL, 0,
- DHT_MSG_LAYOUT_MERGE_FAILED,
- "error no layout on disk for merge");
- return -1;
- }
+ if (!disk_layout_raw) {
+ gf_smsg(this->name, GF_LOG_CRITICAL, 0, DHT_MSG_LAYOUT_MERGE_FAILED,
+ NULL);
+ return -1;
+ }
- GF_ASSERT (disk_layout_len == sizeof (disk_layout));
+ GF_ASSERT(disk_layout_len == sizeof(disk_layout));
- memcpy (disk_layout, disk_layout_raw, disk_layout_len);
+ memcpy(disk_layout, disk_layout_raw, disk_layout_len);
- type = ntoh32 (disk_layout[1]);
- switch (type) {
+ type = ntoh32(disk_layout[1]);
+ switch (type) {
case DHT_HASH_TYPE_DM_USER:
- gf_msg_debug (this->name, 0, "found user-set layout");
- layout->type = type;
- /* Fall through. */
- case DHT_HASH_TYPE_DM:
- break;
+ gf_msg_debug(this->name, 0, "found user-set layout");
+ layout->type = type;
+ /* Fall through. */
+ case DHT_HASH_TYPE_DM:
+ break;
default:
- gf_msg (this->name, GF_LOG_CRITICAL, 0,
- DHT_MSG_INVALID_DISK_LAYOUT,
- "Invalid disk layout: "
- "Catastrophic error layout with unknown type found %d",
- disk_layout[1]);
- return -1;
- }
-
- commit_hash = ntoh32 (disk_layout[0]);
- start_off = ntoh32 (disk_layout[2]);
- stop_off = ntoh32 (disk_layout[3]);
-
- layout->list[pos].commit_hash = commit_hash;
- layout->list[pos].start = start_off;
- layout->list[pos].stop = stop_off;
-
- gf_msg_trace (this->name, 0,
- "merged to layout: %u - %u (type %d, hash %d) from %s",
- start_off, stop_off, commit_hash, type,
- layout->list[pos].xlator->name);
-
- return 0;
+ gf_smsg(this->name, GF_LOG_CRITICAL, 0, DHT_MSG_INVALID_DISK_LAYOUT,
+ "layout=%d", disk_layout[1], NULL);
+ return -1;
+ }
+
+ commit_hash = ntoh32(disk_layout[0]);
+ start_off = ntoh32(disk_layout[2]);
+ stop_off = ntoh32(disk_layout[3]);
+
+ layout->list[pos].commit_hash = commit_hash;
+ layout->list[pos].start = start_off;
+ layout->list[pos].stop = stop_off;
+
+ gf_msg_trace(this->name, 0,
+ "merged to layout: 0x%x - 0x%x (hash 0x%x, type %d) from %s",
+ start_off, stop_off, commit_hash, type,
+ layout->list[pos].xlator->name);
+
+ return 0;
}
int
-dht_layout_merge (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
- int op_ret, int op_errno, dict_t *xattr)
+dht_layout_merge(xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
+ int op_ret, int op_errno, dict_t *xattr)
{
- int i = 0;
- int ret = -1;
- int err = -1;
- void *disk_layout_raw = NULL;
- int disk_layout_len = 0;
- dht_conf_t *conf = this->private;
-
- if (op_ret != 0) {
- err = op_errno;
- }
+ int i = 0;
+ int ret = -1;
+ int err = -1;
+ void *disk_layout_raw = NULL;
+ int disk_layout_len = 0;
+ dht_conf_t *conf = this->private;
- if (!layout)
- goto out;
+ if (op_ret != 0) {
+ err = op_errno;
+ }
- for (i = 0; i < layout->cnt; i++) {
- if (layout->list[i].xlator == NULL) {
- layout->list[i].err = err;
- layout->list[i].xlator = subvol;
- break;
- }
- }
+ if (!layout)
+ goto out;
- if (op_ret != 0) {
- ret = 0;
- goto out;
+ for (i = 0; i < layout->cnt; i++) {
+ if (layout->list[i].xlator == NULL) {
+ layout->list[i].err = err;
+ layout->list[i].xlator = subvol;
+ break;
}
+ }
- if (xattr) {
- /* during lookup and not mkdir */
- ret = dict_get_ptr_and_len (xattr, conf->xattr_name,
- &disk_layout_raw, &disk_layout_len);
- }
+ if (op_ret != 0) {
+ ret = 0;
+ goto out;
+ }
- if (ret != 0) {
- layout->list[i].err = 0;
- gf_msg_trace (this->name, 0,
- "Missing disk layout on %s. err = %d",
- subvol->name, err);
- ret = 0;
- goto out;
- }
+ if (xattr) {
+ /* during lookup and not mkdir */
+ ret = dict_get_ptr_and_len(xattr, conf->xattr_name, &disk_layout_raw,
+ &disk_layout_len);
+ }
- ret = dht_disk_layout_merge (this, layout, i, disk_layout_raw,
- disk_layout_len);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_LAYOUT_MERGE_FAILED,
- "layout merge from subvolume %s failed",
- subvol->name);
- goto out;
- }
+ if (ret != 0) {
+ layout->list[i].err = 0;
+ gf_msg_trace(this->name, 0, "Missing disk layout on %s. err = %d",
+ subvol->name, err);
+ ret = 0;
+ goto out;
+ }
- if (layout->commit_hash == 0) {
- layout->commit_hash = layout->list[i].commit_hash;
- } else if (layout->commit_hash != layout->list[i].commit_hash) {
- layout->commit_hash = DHT_LAYOUT_HASH_INVALID;
- }
+ ret = dht_disk_layout_merge(this, layout, i, disk_layout_raw,
+ disk_layout_len);
+ if (ret != 0) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_LAYOUT_MERGE_FAILED,
+ "subvolume=%s", subvol->name, NULL);
+ goto out;
+ }
- layout->list[i].err = 0;
+ if (layout->commit_hash == 0) {
+ layout->commit_hash = layout->list[i].commit_hash;
+ } else if (layout->commit_hash != layout->list[i].commit_hash) {
+ layout->commit_hash = DHT_LAYOUT_HASH_INVALID;
+ }
+
+ layout->list[i].err = 0;
out:
- return ret;
+ return ret;
}
-
void
-dht_layout_entry_swap (dht_layout_t *layout, int i, int j)
+dht_layout_entry_swap(dht_layout_t *layout, int i, int j)
{
- uint32_t start_swap = 0;
- uint32_t stop_swap = 0;
- uint32_t commit_hash_swap = 0;
- xlator_t *xlator_swap = 0;
- int err_swap = 0;
-
- start_swap = layout->list[i].start;
- stop_swap = layout->list[i].stop;
- xlator_swap = layout->list[i].xlator;
- err_swap = layout->list[i].err;
- commit_hash_swap = layout->list[i].commit_hash;
-
- layout->list[i].start = layout->list[j].start;
- layout->list[i].stop = layout->list[j].stop;
- layout->list[i].xlator = layout->list[j].xlator;
- layout->list[i].err = layout->list[j].err;
- layout->list[i].commit_hash = layout->list[j].commit_hash;
-
- layout->list[j].start = start_swap;
- layout->list[j].stop = stop_swap;
- layout->list[j].xlator = xlator_swap;
- layout->list[j].err = err_swap;
- layout->list[j].commit_hash = commit_hash_swap;
+ uint32_t start_swap = 0;
+ uint32_t stop_swap = 0;
+ uint32_t commit_hash_swap = 0;
+ xlator_t *xlator_swap = 0;
+ int err_swap = 0;
+
+ start_swap = layout->list[i].start;
+ stop_swap = layout->list[i].stop;
+ xlator_swap = layout->list[i].xlator;
+ err_swap = layout->list[i].err;
+ commit_hash_swap = layout->list[i].commit_hash;
+
+ layout->list[i].start = layout->list[j].start;
+ layout->list[i].stop = layout->list[j].stop;
+ layout->list[i].xlator = layout->list[j].xlator;
+ layout->list[i].err = layout->list[j].err;
+ layout->list[i].commit_hash = layout->list[j].commit_hash;
+
+ layout->list[j].start = start_swap;
+ layout->list[j].stop = stop_swap;
+ layout->list[j].xlator = xlator_swap;
+ layout->list[j].err = err_swap;
+ layout->list[j].commit_hash = commit_hash_swap;
}
void
-dht_layout_range_swap (dht_layout_t *layout, int i, int j)
+dht_layout_range_swap(dht_layout_t *layout, int i, int j)
{
- uint32_t start_swap = 0;
- uint32_t stop_swap = 0;
+ uint32_t start_swap = 0;
+ uint32_t stop_swap = 0;
- start_swap = layout->list[i].start;
- stop_swap = layout->list[i].stop;
+ start_swap = layout->list[i].start;
+ stop_swap = layout->list[i].stop;
- layout->list[i].start = layout->list[j].start;
- layout->list[i].stop = layout->list[j].stop;
+ layout->list[i].start = layout->list[j].start;
+ layout->list[i].stop = layout->list[j].stop;
- layout->list[j].start = start_swap;
- layout->list[j].stop = stop_swap;
+ layout->list[j].start = start_swap;
+ layout->list[j].stop = stop_swap;
}
-
-int64_t
-dht_layout_entry_cmp_volname (dht_layout_t *layout, int i, int j)
+static int64_t
+dht_layout_entry_cmp_volname(dht_layout_t *layout, int i, int j)
{
- return (strcmp (layout->list[i].xlator->name,
- layout->list[j].xlator->name));
+ return (strcmp(layout->list[i].xlator->name, layout->list[j].xlator->name));
}
gf_boolean_t
-dht_is_subvol_in_layout (dht_layout_t *layout, xlator_t *xlator)
+dht_is_subvol_in_layout(dht_layout_t *layout, xlator_t *xlator)
{
- int i = 0;
-
- for (i = 0; i < layout->cnt; i++) {
- /* Check if xlator is already part of layout, and layout is
- * non-zero. */
- if (!strcmp (layout->list[i].xlator->name, xlator->name)) {
- if (layout->list[i].start != layout->list[i].stop)
- return _gf_true;
- break;
- }
- }
- return _gf_false;
+ int i = 0;
+
+ for (i = 0; i < layout->cnt; i++) {
+ /* Check if xlator is already part of layout, and layout is
+ * non-zero. */
+ if (!strcmp(layout->list[i].xlator->name, xlator->name)) {
+ if (layout->list[i].start != layout->list[i].stop)
+ return _gf_true;
+ break;
+ }
+ }
+ return _gf_false;
}
-int64_t
-dht_layout_entry_cmp (dht_layout_t *layout, int i, int j)
+static int64_t
+dht_layout_entry_cmp(dht_layout_t *layout, int i, int j)
{
- int64_t diff = 0;
+ int64_t diff = 0;
- /* swap zero'ed out layouts to front, if needed */
- if (!layout->list[j].start && !layout->list[j].stop) {
- diff = (int64_t) layout->list[i].stop
- - (int64_t) layout->list[j].stop;
- goto out;
- }
- diff = (int64_t) layout->list[i].start
- - (int64_t) layout->list[j].start;
+ /* swap zero'ed out layouts to front, if needed */
+ if (!layout->list[j].start && !layout->list[j].stop) {
+ diff = (int64_t)layout->list[i].stop - (int64_t)layout->list[j].stop;
+ goto out;
+ }
+ diff = (int64_t)layout->list[i].start - (int64_t)layout->list[j].start;
out:
- return diff;
+ return diff;
}
-
int
-dht_layout_sort (dht_layout_t *layout)
+dht_layout_sort(dht_layout_t *layout)
{
- int i = 0;
- int j = 0;
- int64_t ret = 0;
+ int i = 0;
+ int j = 0;
+ int64_t ret = 0;
- /* TODO: O(n^2) -- bad bad */
+ /* TODO: O(n^2) -- bad bad */
- for (i = 0; i < layout->cnt - 1; i++) {
- for (j = i + 1; j < layout->cnt; j++) {
- ret = dht_layout_entry_cmp (layout, i, j);
- if (ret > 0)
- dht_layout_entry_swap (layout, i, j);
- }
+ for (i = 0; i < layout->cnt - 1; i++) {
+ for (j = i + 1; j < layout->cnt; j++) {
+ ret = dht_layout_entry_cmp(layout, i, j);
+ if (ret > 0)
+ dht_layout_entry_swap(layout, i, j);
}
+ }
- return 0;
+ return 0;
}
-int
-dht_layout_sort_volname (dht_layout_t *layout)
+void
+dht_layout_sort_volname(dht_layout_t *layout)
{
- int i = 0;
- int j = 0;
- int64_t ret = 0;
+ int i = 0;
+ int j = 0;
+ int64_t ret = 0;
- /* TODO: O(n^2) -- bad bad */
+ /* TODO: O(n^2) -- bad bad */
- for (i = 0; i < layout->cnt - 1; i++) {
- for (j = i + 1; j < layout->cnt; j++) {
- ret = dht_layout_entry_cmp_volname (layout, i, j);
- if (ret > 0)
- dht_layout_entry_swap (layout, i, j);
- }
+ for (i = 0; i < layout->cnt - 1; i++) {
+ for (j = i + 1; j < layout->cnt; j++) {
+ ret = dht_layout_entry_cmp_volname(layout, i, j);
+ if (ret > 0)
+ dht_layout_entry_swap(layout, i, j);
}
-
- return 0;
+ }
}
-
-int
-dht_layout_anomalies (xlator_t *this, loc_t *loc, dht_layout_t *layout,
- uint32_t *holes_p, uint32_t *overlaps_p,
- uint32_t *missing_p, uint32_t *down_p, uint32_t *misc_p,
- uint32_t *no_space_p)
+void
+dht_layout_anomalies(xlator_t *this, loc_t *loc, dht_layout_t *layout,
+ uint32_t *holes_p, uint32_t *overlaps_p,
+ uint32_t *missing_p, uint32_t *down_p, uint32_t *misc_p,
+ uint32_t *no_space_p)
{
- uint32_t overlaps = 0;
- uint32_t missing = 0;
- uint32_t down = 0;
- uint32_t misc = 0;
- uint32_t hole_cnt = 0;
- uint32_t overlap_cnt = 0;
- int i = 0;
- int ret = 0;
- uint32_t prev_stop = 0;
- uint32_t last_stop = 0;
- char is_virgin = 1;
- uint32_t no_space = 0;
-
- /* This function scans through the layout spread of a directory to
- check if there are any anomalies. Prior to calling this function
- the layout entries should be sorted in the ascending order.
-
- If the layout entry has err != 0
- then increment the corresponding anomaly.
- else
- if (start of the current layout entry > stop + 1 of previous
- non erroneous layout entry)
- then it indicates a hole in the layout
- if (start of the current layout entry < stop + 1 of previous
- non erroneous layout entry)
- then it indicates an overlap in the layout
- */
- last_stop = layout->list[0].start - 1;
- prev_stop = last_stop;
-
- for (i = 0; i < layout->cnt; i++) {
- switch (layout->list[i].err) {
- case -1:
- case ENOENT:
- case ESTALE:
- missing++;
- continue;
- case ENOTCONN:
- down++;
- continue;
- case ENOSPC:
- no_space++;
- continue;
- case 0:
- /* if err == 0 and start == stop, then it is a non misc++;
- * participating subvolume(spread-cnt). Then, do not
- * check for anomalies. If start != stop, then treat it
- * as misc err */
- if (layout->list[i].start == layout->list[i].stop) {
- continue;
- }
- break;
- default:
- misc++;
- continue;
- }
-
- is_virgin = 0;
-
- if ((prev_stop + 1) < layout->list[i].start) {
- hole_cnt++;
+ uint32_t overlaps = 0;
+ uint32_t missing = 0;
+ uint32_t down = 0;
+ uint32_t misc = 0;
+ uint32_t hole_cnt = 0;
+ uint32_t overlap_cnt = 0;
+ int i = 0;
+ uint32_t prev_stop = 0;
+ uint32_t last_stop = 0;
+ char is_virgin = 1;
+ uint32_t no_space = 0;
+
+ /* This function scans through the layout spread of a directory to
+ check if there are any anomalies. Prior to calling this function
+ the layout entries should be sorted in the ascending order.
+
+ If the layout entry has err != 0
+ then increment the corresponding anomaly.
+ else
+ if (start of the current layout entry > stop + 1 of previous
+ non erroneous layout entry)
+ then it indicates a hole in the layout
+ if (start of the current layout entry < stop + 1 of previous
+ non erroneous layout entry)
+ then it indicates an overlap in the layout
+ */
+ last_stop = layout->list[0].start - 1;
+ prev_stop = last_stop;
+
+ for (i = 0; i < layout->cnt; i++) {
+ switch (layout->list[i].err) {
+ case -1:
+ case ENOENT:
+ case ESTALE:
+ missing++;
+ continue;
+ case ENOTCONN:
+ down++;
+ continue;
+ case ENOSPC:
+ no_space++;
+ continue;
+ case 0:
+ /* if err == 0 and start == stop, then it is a non misc++;
+ * participating subvolume(spread-cnt). Then, do not
+ * check for anomalies. If start != stop, then treat it
+ * as misc err */
+ if (layout->list[i].start == layout->list[i].stop) {
+ continue;
}
+ break;
+ default:
+ misc++;
+ continue;
+ }
- if ((prev_stop + 1) > layout->list[i].start) {
- overlap_cnt++;
- overlaps += ((prev_stop + 1) - layout->list[i].start);
- }
- prev_stop = layout->list[i].stop;
+ is_virgin = 0;
+
+ if ((prev_stop + 1) < layout->list[i].start) {
+ hole_cnt++;
}
- if ((last_stop - prev_stop) || is_virgin)
- hole_cnt++;
+ if ((prev_stop + 1) > layout->list[i].start) {
+ overlap_cnt++;
+ overlaps += ((prev_stop + 1) - layout->list[i].start);
+ }
+ prev_stop = layout->list[i].stop;
+ }
- if (holes_p)
- *holes_p = hole_cnt;
+ if ((last_stop - prev_stop) || is_virgin)
+ hole_cnt++;
- if (overlaps_p)
- *overlaps_p = overlap_cnt;
+ if (holes_p)
+ *holes_p = hole_cnt;
- if (missing_p)
- *missing_p = missing;
+ if (overlaps_p)
+ *overlaps_p = overlap_cnt;
- if (down_p)
- *down_p = down;
+ if (missing_p)
+ *missing_p = missing;
- if (misc_p)
- *misc_p = misc;
+ if (down_p)
+ *down_p = down;
- if (no_space_p)
- *no_space_p = no_space;
+ if (misc_p)
+ *misc_p = misc;
- return ret;
+ if (no_space_p)
+ *no_space_p = no_space;
}
-
int
-dht_layout_missing_dirs (dht_layout_t *layout)
+dht_layout_missing_dirs(dht_layout_t *layout)
{
- int i = 0, missing = 0;
+ int i = 0, missing = 0;
- if (layout == NULL)
- goto out;
+ if (layout == NULL)
+ goto out;
- for (i = 0; i < layout->cnt; i++) {
- if ((layout->list[i].err == ENOENT)
- || ((layout->list[i].err == -1)
- && (layout->list[i].start == 0)
- && (layout->list[i].stop == 0))) {
- missing++;
- }
+ for (i = 0; i < layout->cnt; i++) {
+ if ((layout->list[i].err == ENOENT) ||
+ ((layout->list[i].err == -1) && (layout->list[i].start == 0) &&
+ (layout->list[i].stop == 0))) {
+ missing++;
}
+ }
out:
- return missing;
+ return missing;
}
-
int
-dht_layout_normalize (xlator_t *this, loc_t *loc, dht_layout_t *layout)
+dht_layout_normalize(xlator_t *this, loc_t *loc, dht_layout_t *layout)
{
- int ret = 0;
- uint32_t holes = 0;
- uint32_t overlaps = 0;
- uint32_t missing = 0;
- uint32_t down = 0;
- uint32_t misc = 0, missing_dirs = 0;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- ret = dht_layout_sort (layout);
- if (ret == -1) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_LAYOUT_SORT_FAILED,
- "sort failed?! how the ....");
- goto out;
- }
-
- gf_uuid_unparse(loc->gfid, gfid);
-
- ret = dht_layout_anomalies (this, loc, layout,
- &holes, &overlaps,
- &missing, &down, &misc, NULL);
- if (ret == -1) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_FIND_LAYOUT_ANOMALIES_ERROR,
- "Error finding anomalies in %s, gfid = %s",
- loc->path, gfid);
- goto out;
- }
-
- if (holes || overlaps) {
- if (missing == layout->cnt) {
- gf_msg_debug (this->name, 0,
- "Directory %s looked up first time"
- " gfid = %s", loc->path, gfid);
- } else {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_ANOMALIES_INFO,
- "Found anomalies in %s (gfid = %s). "
- "Holes=%d overlaps=%d",
- loc->path, gfid, holes, overlaps );
- }
- ret = -1;
- }
-
- if (ret >= 0) {
- missing_dirs = dht_layout_missing_dirs (layout);
- /* TODO During DHT selfheal rewrite (almost) find a better place
- * to detect this - probably in dht_layout_anomalies()
- */
- if (missing_dirs > 0)
- ret += missing_dirs;
- }
+ int ret = 0;
+ uint32_t holes = 0;
+ uint32_t overlaps = 0;
+ uint32_t missing = 0;
+ uint32_t down = 0;
+ uint32_t misc = 0, missing_dirs = 0;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+
+ ret = dht_layout_sort(layout);
+ if (ret == -1) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_LAYOUT_SORT_FAILED,
+ NULL);
+ goto out;
+ }
+
+ gf_uuid_unparse(loc->gfid, gfid);
+
+ dht_layout_anomalies(this, loc, layout, &holes, &overlaps, &missing, &down,
+ &misc, NULL);
+
+ if (holes || overlaps) {
+ if (missing == layout->cnt) {
+ gf_msg_debug(this->name, 0,
+ "Directory %s looked up first time"
+ " gfid = %s",
+ loc->path, gfid);
+ } else {
+ gf_smsg(this->name, GF_LOG_INFO, 0, DHT_MSG_ANOMALIES_INFO,
+ "path=%s", loc->path, "gfid=%s", gfid, "holes=%d", holes,
+ "overlaps=%d", overlaps, NULL);
+ }
+ ret = -1;
+ }
+
+ if (ret >= 0) {
+ missing_dirs = dht_layout_missing_dirs(layout);
+ /* TODO During DHT selfheal rewrite (almost) find a better place
+ * to detect this - probably in dht_layout_anomalies()
+ */
+ if (missing_dirs > 0)
+ ret += missing_dirs;
+ }
out:
- return ret;
+ return ret;
}
int
-dht_dir_has_layout (dict_t *xattr, char *name)
+dht_dir_has_layout(dict_t *xattr, char *name)
{
+ void *disk_layout_raw = NULL;
- void *disk_layout_raw = NULL;
-
- return dict_get_ptr (xattr, name, &disk_layout_raw);
+ return dict_get_ptr(xattr, name, &disk_layout_raw);
}
int
-dht_layout_dir_mismatch (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
- loc_t *loc, dict_t *xattr)
+dht_layout_dir_mismatch(xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
+ loc_t *loc, dict_t *xattr)
{
- int idx = 0;
- int pos = -1;
- int ret = 0;
- int err = 0;
- int dict_ret = 0;
- int32_t disk_layout[4];
- void *disk_layout_raw = NULL;
- uint32_t start_off = -1;
- uint32_t stop_off = -1;
- uint32_t commit_hash = -1;
- dht_conf_t *conf = this->private;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- if(loc && loc->inode)
- gf_uuid_unparse(loc->inode->gfid, gfid);
-
- for (idx = 0; idx < layout->cnt; idx++) {
- if (layout->list[idx].xlator == subvol) {
- pos = idx;
- break;
- }
- }
-
- if (pos == -1) {
- if (loc) {
- gf_msg_debug (this->name, 0,
- "%s - no layout info for subvolume %s",
- loc ? loc->path : "path not found",
- subvol->name);
- }
- ret = 1;
- goto out;
- }
-
- err = layout->list[pos].err;
-
- if (!xattr) {
- if (err == 0) {
- if (loc) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_DICT_GET_FAILED,
- "%s: xattr dictionary is NULL",
- loc->path);
- } else {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_DICT_GET_FAILED,
- "path not found: "
- "xattr dictionary is NULL");
- }
- ret = -1;
- }
- goto out;
- }
-
- dict_ret = dict_get_ptr (xattr, conf->xattr_name,
- &disk_layout_raw);
-
- if (dict_ret < 0) {
- if (err == 0 && layout->list[pos].stop) {
- if (loc) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_DISK_LAYOUT_MISSING,
- "%s: Disk layout missing, gfid = %s",
- loc->path, gfid);
- } else {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_DISK_LAYOUT_MISSING,
- "path not found: "
- "Disk layout missing, gfid = %s",
- gfid);
- }
- ret = -1;
- }
- goto out;
- }
-
- memcpy (disk_layout, disk_layout_raw, sizeof (disk_layout));
-
- start_off = ntoh32 (disk_layout[2]);
- stop_off = ntoh32 (disk_layout[3]);
- commit_hash = ntoh32 (disk_layout[0]);
-
- if ((layout->list[pos].start != start_off)
- || (layout->list[pos].stop != stop_off)
- || (layout->list[pos].commit_hash != commit_hash)) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LAYOUT_INFO,
- "subvol: %s; inode layout - %"PRIu32" - %"PRIu32
- " - %"PRIu32"; "
- "disk layout - %"PRIu32" - %"PRIu32" - %"PRIu32,
- layout->list[pos].xlator->name,
- layout->list[pos].start, layout->list[pos].stop,
- layout->list[pos].commit_hash,
- start_off, stop_off, commit_hash);
- ret = 1;
- } else {
- ret = 0;
- }
+ int idx = 0;
+ int pos = -1;
+ int ret = 0;
+ int err = 0;
+ int dict_ret = 0;
+ int32_t disk_layout[4];
+ void *disk_layout_raw = NULL;
+ uint32_t start_off = -1;
+ uint32_t stop_off = -1;
+ uint32_t commit_hash = -1;
+ dht_conf_t *conf = this->private;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+
+ if (loc && loc->inode)
+ gf_uuid_unparse(loc->inode->gfid, gfid);
+
+ for (idx = 0; idx < layout->cnt; idx++) {
+ if (layout->list[idx].xlator == subvol) {
+ pos = idx;
+ break;
+ }
+ }
+
+ if (pos == -1) {
+ if (loc) {
+ gf_msg_debug(this->name, 0, "%s - no layout info for subvolume %s",
+ loc ? loc->path : "path not found", subvol->name);
+ }
+ ret = 1;
+ goto out;
+ }
+
+ err = layout->list[pos].err;
+
+ if (!xattr) {
+ if (err == 0) {
+ if (loc) {
+ gf_smsg(this->name, GF_LOG_INFO, 0, DHT_MSG_XATTR_DICT_NULL,
+ "path=%s", loc->path, NULL);
+ } else {
+ gf_smsg(this->name, GF_LOG_INFO, 0, DHT_MSG_XATTR_DICT_NULL,
+ "path not found", NULL);
+ }
+ ret = -1;
+ }
+ goto out;
+ }
+
+ dict_ret = dict_get_ptr(xattr, conf->xattr_name, &disk_layout_raw);
+
+ if (dict_ret < 0) {
+ if (err == 0 && layout->list[pos].stop) {
+ if (loc) {
+ gf_smsg(this->name, GF_LOG_INFO, 0, DHT_MSG_DISK_LAYOUT_MISSING,
+ "path=%s", loc->path, "gfid=%s", gfid, NULL);
+ } else {
+ gf_smsg(this->name, GF_LOG_INFO, 0, DHT_MSG_DISK_LAYOUT_MISSING,
+ "path not found"
+ "gfid=%s",
+ gfid, NULL);
+ }
+ ret = -1;
+ }
+ goto out;
+ }
+
+ memcpy(disk_layout, disk_layout_raw, sizeof(disk_layout));
+
+ start_off = ntoh32(disk_layout[2]);
+ stop_off = ntoh32(disk_layout[3]);
+ commit_hash = ntoh32(disk_layout[0]);
+
+ if ((layout->list[pos].start != start_off) ||
+ (layout->list[pos].stop != stop_off) ||
+ (layout->list[pos].commit_hash != commit_hash)) {
+ gf_smsg(this->name, GF_LOG_INFO, 0, DHT_MSG_LAYOUT_INFO, "subvol=%s",
+ layout->list[pos].xlator->name, "inode-layout:start=0x%x",
+ layout->list[pos].start, "inode-layout:stop=0x%x",
+ layout->list[pos].stop, "layout-commit-hash=0x%x; ",
+ layout->list[pos].commit_hash, "disk-layout:start-off=0x%x",
+ start_off, "disk-layout:top-off=0x%x", stop_off,
+ "commit-hash=0x%x", commit_hash, NULL);
+ ret = 1;
+ } else {
+ ret = 0;
+ }
out:
- return ret;
+ return ret;
}
-
int
-dht_layout_preset (xlator_t *this, xlator_t *subvol, inode_t *inode)
+dht_layout_preset(xlator_t *this, xlator_t *subvol, inode_t *inode)
{
- dht_layout_t *layout = NULL;
- int ret = -1;
- dht_conf_t *conf = NULL;
+ dht_layout_t *layout = NULL;
+ int ret = -1;
+ dht_conf_t *conf = NULL;
- conf = this->private;
- if (!conf)
- goto out;
+ conf = this->private;
+ if (!conf)
+ goto out;
+ layout = dht_layout_for_subvol(this, subvol);
+ if (!layout) {
+ gf_smsg(this->name, GF_LOG_INFO, 0, DHT_MSG_SUBVOL_NO_LAYOUT_INFO,
+ "subvolume=%s", subvol ? subvol->name : "<nil>", NULL);
+ ret = -1;
+ goto out;
+ }
- layout = dht_layout_for_subvol (this, subvol);
- if (!layout) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_SUBVOL_NO_LAYOUT_INFO,
- "no pre-set layout for subvolume %s",
- subvol ? subvol->name : "<nil>");
- ret = -1;
- goto out;
- }
+ gf_msg_debug(this->name, 0, "file = %s, subvol = %s",
+ uuid_utoa(inode->gfid), subvol ? subvol->name : "<nil>");
- gf_msg_debug (this->name, 0, "file = %s, subvol = %s",
- uuid_utoa (inode->gfid), subvol ? subvol->name : "<nil>");
+ LOCK(&conf->layout_lock);
+ {
+ dht_inode_ctx_layout_set(inode, this, layout);
+ }
- LOCK (&conf->layout_lock);
- {
- dht_inode_ctx_layout_set (inode, this, layout);
- }
+ UNLOCK(&conf->layout_lock);
- UNLOCK (&conf->layout_lock);
-
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-dht_layout_index_for_subvol (dht_layout_t *layout, xlator_t *subvol)
+dht_layout_index_for_subvol(dht_layout_t *layout, xlator_t *subvol)
{
- int i = 0, ret = -1;
+ int i = 0, ret = -1;
- for (i = 0; i < layout->cnt; i++) {
- if (layout->list[i].xlator == subvol) {
- ret = i;
- break;
- }
+ for (i = 0; i < layout->cnt; i++) {
+ if (layout->list[i].xlator == subvol) {
+ ret = i;
+ break;
}
+ }
- return ret;
+ return ret;
}
diff --git a/xlators/cluster/dht/src/dht-linkfile.c b/xlators/cluster/dht/src/dht-linkfile.c
index a6d994b4157..89ec6cca56e 100644
--- a/xlators/cluster/dht/src/dht-linkfile.c
+++ b/xlators/cluster/dht/src/dht-linkfile.c
@@ -8,351 +8,321 @@
cases as published by the Free Software Foundation.
*/
-
-#include "glusterfs.h"
-#include "xlator.h"
-#include "compat.h"
+#include <glusterfs/compat.h>
#include "dht-common.h"
-#include "dht-messages.h"
-int
-dht_linkfile_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, dict_t *xattr,
- struct iatt *postparent)
+static int
+dht_linkfile_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *stbuf, dict_t *xattr,
+ struct iatt *postparent)
{
- char is_linkfile = 0;
- dht_conf_t *conf = NULL;
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- local = frame->local;
- prev = cookie;
- conf = this->private;
-
- if (op_ret)
- goto out;
-
- gf_uuid_unparse(local->loc.gfid, gfid);
-
- is_linkfile = check_is_linkfile (inode, stbuf, xattr,
- conf->link_xattr_name);
- if (!is_linkfile)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_NOT_LINK_FILE_ERROR,
- "got non-linkfile %s:%s, gfid = %s",
- prev->name, local->loc.path, gfid);
+ char is_linkfile = 0;
+ dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+
+ local = frame->local;
+ prev = cookie;
+ conf = this->private;
+
+ if (op_ret)
+ goto out;
+
+ gf_uuid_unparse(local->loc.gfid, gfid);
+
+ is_linkfile = check_is_linkfile(inode, stbuf, xattr, conf->link_xattr_name);
+ if (!is_linkfile)
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_NOT_LINK_FILE_ERROR,
+ "name=%s", prev->name, "path=%s", local->loc.path, "gfid=%s",
+ gfid, NULL);
out:
- local->linkfile.linkfile_cbk (frame, cookie, this, op_ret, op_errno,
- inode, stbuf, postparent, postparent,
- xattr);
- return 0;
+ local->linkfile.linkfile_cbk(frame, cookie, this, op_ret, op_errno, inode,
+ stbuf, postparent, postparent, xattr);
+ return 0;
}
-#define is_equal(a, b) ((a) == (b))
-int
-dht_linkfile_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode,
- struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+static int
+dht_linkfile_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *subvol = NULL;
- dict_t *xattrs = NULL;
- dht_conf_t *conf = NULL;
- int ret = -1;
-
- local = frame->local;
-
- if (!op_ret)
- local->linked = _gf_true;
-
- FRAME_SU_UNDO (frame, dht_local_t);
-
- if (op_ret && (op_errno == EEXIST)) {
- conf = this->private;
- subvol = cookie;
- if (!subvol)
- goto out;
- xattrs = dict_new ();
- if (!xattrs)
- goto out;
- ret = dict_set_uint32 (xattrs, conf->link_xattr_name, 256);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value. key : %s",
- conf->link_xattr_name);
- goto out;
- }
-
- STACK_WIND_COOKIE (frame, dht_linkfile_lookup_cbk, subvol,
- subvol, subvol->fops->lookup, &local->linkfile.loc,
- xattrs);
- if (xattrs)
- dict_unref (xattrs);
- return 0;
+ dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+ dict_t *xattrs = NULL;
+ dht_conf_t *conf = NULL;
+ int ret = -1;
+
+ local = frame->local;
+
+ if (!op_ret)
+ local->linked = _gf_true;
+
+ FRAME_SU_UNDO(frame, dht_local_t);
+
+ if (op_ret && (op_errno == EEXIST)) {
+ conf = this->private;
+ subvol = cookie;
+ if (!subvol)
+ goto out;
+ xattrs = dict_new();
+ if (!xattrs)
+ goto out;
+ ret = dict_set_uint32(xattrs, conf->link_xattr_name, 256);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "mame=%s", conf->link_xattr_name, NULL);
+ goto out;
}
-out:
- local->linkfile.linkfile_cbk (frame, cookie, this, op_ret, op_errno,
- inode, stbuf, preparent, postparent,
- xdata);
+
+ STACK_WIND_COOKIE(frame, dht_linkfile_lookup_cbk, subvol, subvol,
+ subvol->fops->lookup, &local->linkfile.loc, xattrs);
if (xattrs)
- dict_unref (xattrs);
+ dict_unref(xattrs);
return 0;
+ }
+out:
+ local->linkfile.linkfile_cbk(frame, cookie, this, op_ret, op_errno, inode,
+ stbuf, preparent, postparent, xdata);
+ if (xattrs)
+ dict_unref(xattrs);
+ return 0;
}
-
int
-dht_linkfile_create (call_frame_t *frame, fop_mknod_cbk_t linkfile_cbk,
- xlator_t *this,
- xlator_t *tovol, xlator_t *fromvol, loc_t *loc)
+dht_linkfile_create(call_frame_t *frame, fop_mknod_cbk_t linkfile_cbk,
+ xlator_t *this, xlator_t *tovol, xlator_t *fromvol,
+ loc_t *loc)
{
- dht_local_t *local = NULL;
- dict_t *dict = NULL;
- int need_unref = 0;
- int ret = 0;
- dht_conf_t *conf = this->private;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- local = frame->local;
- local->linkfile.linkfile_cbk = linkfile_cbk;
- local->linkfile.srcvol = tovol;
- loc_copy (&local->linkfile.loc, loc);
-
- local->linked = _gf_false;
-
- dict = local->params;
- if (!dict) {
- dict = dict_new ();
- if (!dict)
- goto out;
- need_unref = 1;
- }
-
-
- if (!gf_uuid_is_null (local->gfid)) {
- gf_uuid_unparse(local->gfid, gfid);
-
- ret = dict_set_gfuuid (dict, "gfid-req", local->gfid, true);
- if (ret)
- gf_msg ("dht-linkfile", GF_LOG_INFO, 0,
- DHT_MSG_DICT_SET_FAILED,
- "%s: Failed to set dictionary value: "
- "key = gfid-req, gfid = %s ", loc->path, gfid);
- } else {
- gf_uuid_unparse(loc->gfid, gfid);
- }
-
- ret = dict_set_str (dict, GLUSTERFS_INTERNAL_FOP_KEY, "yes");
+ dht_local_t *local = NULL;
+ dict_t *dict = NULL;
+ int need_unref = 0;
+ int ret = 0;
+ dht_conf_t *conf = this->private;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+
+ local = frame->local;
+ local->linkfile.linkfile_cbk = linkfile_cbk;
+ local->linkfile.srcvol = tovol;
+ loc_copy(&local->linkfile.loc, loc);
+
+ local->linked = _gf_false;
+
+ dict = local->params;
+ if (!dict) {
+ dict = dict_new();
+ if (!dict)
+ goto out;
+ need_unref = 1;
+ }
+
+ if (!gf_uuid_is_null(local->gfid)) {
+ gf_uuid_unparse(local->gfid, gfid);
+
+ ret = dict_set_gfuuid(dict, "gfid-req", local->gfid, true);
if (ret)
- gf_msg ("dht-linkfile", GF_LOG_INFO, 0,
- DHT_MSG_DICT_SET_FAILED,
- "%s: Failed to set dictionary value: key = %s,"
- " gfid = %s", loc->path,
- GLUSTERFS_INTERNAL_FOP_KEY, gfid);
-
- ret = dict_set_str (dict, conf->link_xattr_name, tovol->name);
-
- if (ret < 0) {
- gf_msg (frame->this->name, GF_LOG_INFO, 0,
- DHT_MSG_CREATE_LINK_FAILED,
- "%s: failed to initialize linkfile data, gfid = %s",
- loc->path, gfid);
- goto out;
- }
-
- local->link_subvol = fromvol;
- /* Always create as root:root. dht_linkfile_attr_heal fixes the
- * ownsership */
- FRAME_SU_DO (frame, dht_local_t);
- STACK_WIND_COOKIE (frame, dht_linkfile_create_cbk, fromvol, fromvol,
- fromvol->fops->mknod, loc,
- S_IFREG | DHT_LINKFILE_MODE, 0, 0, dict);
-
- if (need_unref && dict)
- dict_unref (dict);
-
- return 0;
+ gf_smsg("dht-linkfile", GF_LOG_INFO, 0, DHT_MSG_DICT_SET_FAILED,
+ "path=%s", loc->path, "gfid=%s", gfid, NULL);
+ } else {
+ gf_uuid_unparse(loc->gfid, gfid);
+ }
+
+ ret = dict_set_str(dict, GLUSTERFS_INTERNAL_FOP_KEY, "yes");
+ if (ret)
+ gf_smsg("dht-linkfile", GF_LOG_INFO, 0, DHT_MSG_DICT_SET_FAILED,
+ "path=%s", loc->path, "key=%s", GLUSTERFS_INTERNAL_FOP_KEY,
+ "gfid=%s", gfid, NULL);
+
+ ret = dict_set_str(dict, conf->link_xattr_name, tovol->name);
+
+ if (ret < 0) {
+ gf_smsg(frame->this->name, GF_LOG_INFO, 0, DHT_MSG_CREATE_LINK_FAILED,
+ "path=%s", loc->path, "gfid=%s", gfid, NULL);
+ goto out;
+ }
+
+ local->link_subvol = fromvol;
+ /* Always create as root:root. dht_linkfile_attr_heal fixes the
+ * ownsership */
+ FRAME_SU_DO(frame, dht_local_t);
+ STACK_WIND_COOKIE(frame, dht_linkfile_create_cbk, fromvol, fromvol,
+ fromvol->fops->mknod, loc, S_IFREG | DHT_LINKFILE_MODE, 0,
+ 0, dict);
+
+ if (need_unref && dict)
+ dict_unref(dict);
+
+ return 0;
out:
- local->linkfile.linkfile_cbk (frame, frame->this, frame->this, -1, ENOMEM,
- loc->inode, NULL, NULL, NULL, NULL);
+ local->linkfile.linkfile_cbk(frame, frame->this, frame->this, -1, ENOMEM,
+ loc->inode, NULL, NULL, NULL, NULL);
- if (need_unref && dict)
- dict_unref (dict);
+ if (need_unref && dict)
+ dict_unref(dict);
- return 0;
+ return 0;
}
-
int
-dht_linkfile_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+dht_linkfile_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *subvol = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- local = frame->local;
- subvol = cookie;
+ dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ local = frame->local;
+ subvol = cookie;
- if (op_ret == -1) {
-
- gf_uuid_unparse(local->loc.gfid, gfid);
- gf_msg (this->name, GF_LOG_INFO, op_errno,
- DHT_MSG_UNLINK_FAILED,
- "Unlinking linkfile %s (gfid = %s)on "
- "subvolume %s failed ",
- local->loc.path, gfid, subvol->name);
- }
+ if (op_ret == -1) {
+ gf_uuid_unparse(local->loc.gfid, gfid);
+ gf_smsg(this->name, GF_LOG_INFO, op_errno, DHT_MSG_UNLINK_FAILED,
+ "path=%s", local->loc.path, "gfid=%s", gfid, "subvolume=%s",
+ subvol->name, NULL);
+ }
- DHT_STACK_DESTROY (frame);
+ DHT_STACK_DESTROY(frame);
- return 0;
+ return 0;
}
-
int
-dht_linkfile_unlink (call_frame_t *frame, xlator_t *this,
- xlator_t *subvol, loc_t *loc)
+dht_linkfile_unlink(call_frame_t *frame, xlator_t *this, xlator_t *subvol,
+ loc_t *loc)
{
- call_frame_t *unlink_frame = NULL;
- dht_local_t *unlink_local = NULL;
+ call_frame_t *unlink_frame = NULL;
+ dht_local_t *unlink_local = NULL;
- unlink_frame = copy_frame (frame);
- if (!unlink_frame) {
- goto err;
- }
+ unlink_frame = copy_frame(frame);
+ if (!unlink_frame) {
+ goto err;
+ }
- /* Using non-fop value here, as anyways, 'local->fop' is not used in
- this particular case */
- unlink_local = dht_local_init (unlink_frame, loc, NULL,
- GF_FOP_MAXVALUE);
- if (!unlink_local) {
- goto err;
- }
+ /* Using non-fop value here, as anyways, 'local->fop' is not used in
+ this particular case */
+ unlink_local = dht_local_init(unlink_frame, loc, NULL, GF_FOP_MAXVALUE);
+ if (!unlink_local) {
+ goto err;
+ }
- STACK_WIND_COOKIE (unlink_frame, dht_linkfile_unlink_cbk, subvol,
- subvol, subvol->fops->unlink,
- &unlink_local->loc, 0, NULL);
+ STACK_WIND_COOKIE(unlink_frame, dht_linkfile_unlink_cbk, subvol, subvol,
+ subvol->fops->unlink, &unlink_local->loc, 0, NULL);
- return 0;
+ return 0;
err:
- if (unlink_frame)
- DHT_STACK_DESTROY (unlink_frame);
+ if (unlink_frame)
+ DHT_STACK_DESTROY(unlink_frame);
- return -1;
+ return -1;
}
-
xlator_t *
-dht_linkfile_subvol (xlator_t *this, inode_t *inode, struct iatt *stbuf,
- dict_t *xattr)
+dht_linkfile_subvol(xlator_t *this, inode_t *inode, struct iatt *stbuf,
+ dict_t *xattr)
{
- dht_conf_t *conf = NULL;
- xlator_t *subvol = NULL;
- void *volname = NULL;
- int i = 0, ret = 0;
+ dht_conf_t *conf = NULL;
+ xlator_t *subvol = NULL;
+ void *volname = NULL;
+ int i = 0, ret = 0;
- conf = this->private;
+ conf = this->private;
- if (!xattr)
- goto out;
+ if (!xattr)
+ goto out;
- ret = dict_get_ptr (xattr, conf->link_xattr_name, &volname);
+ ret = dict_get_ptr(xattr, conf->link_xattr_name, &volname);
- if ((-1 == ret) || !volname)
- goto out;
+ if ((-1 == ret) || !volname)
+ goto out;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (strcmp (conf->subvolumes[i]->name, (char *)volname) == 0) {
- subvol = conf->subvolumes[i];
- break;
- }
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (strcmp(conf->subvolumes[i]->name, (char *)volname) == 0) {
+ subvol = conf->subvolumes[i];
+ break;
}
+ }
out:
- return subvol;
+ return subvol;
}
-int
-dht_linkfile_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *statpre,
- struct iatt *statpost, dict_t *xdata)
+static int
+dht_linkfile_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
{
- dht_local_t *local = NULL;
- loc_t *loc = NULL;
+ dht_local_t *local = NULL;
+ loc_t *loc = NULL;
- local = frame->local;
- loc = &local->loc;
+ local = frame->local;
+ loc = &local->loc;
- if (op_ret)
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- DHT_MSG_SETATTR_FAILED,
- "Failed to set attr uid/gid on %s"
- " :<gfid:%s> ",
- (loc->path? loc->path: "NULL"),
- uuid_utoa(local->gfid));
+ if (op_ret)
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno, DHT_MSG_SETATTR_FAILED,
+ "path=%s", (loc->path ? loc->path : "NULL"), "gfid=%s",
+ uuid_utoa(local->gfid), NULL);
- DHT_STACK_DESTROY (frame);
+ DHT_STACK_DESTROY(frame);
- return 0;
+ return 0;
}
int
-dht_linkfile_attr_heal (call_frame_t *frame, xlator_t *this)
+dht_linkfile_attr_heal(call_frame_t *frame, xlator_t *this)
{
- int ret = -1;
- call_frame_t *copy = NULL;
- dht_local_t *local = NULL;
- dht_local_t *copy_local = NULL;
- xlator_t *subvol = NULL;
- struct iatt stbuf = {0,};
- dict_t *xattr = NULL;
-
- local = frame->local;
-
- GF_VALIDATE_OR_GOTO ("dht", local, out);
- GF_VALIDATE_OR_GOTO ("dht", local->link_subvol, out);
-
- if (local->stbuf.ia_type == IA_INVAL)
- return 0;
+ int ret = -1;
+ call_frame_t *copy = NULL;
+ dht_local_t *local = NULL;
+ dht_local_t *copy_local = NULL;
+ xlator_t *subvol = NULL;
+ struct iatt stbuf = {
+ 0,
+ };
+ dict_t *xattr = NULL;
+
+ local = frame->local;
+
+ GF_VALIDATE_OR_GOTO("dht", local, out);
+ GF_VALIDATE_OR_GOTO("dht", local->link_subvol, out);
+
+ if (local->stbuf.ia_type == IA_INVAL)
+ return 0;
- DHT_MARK_FOP_INTERNAL (xattr);
+ DHT_MARK_FOP_INTERNAL(xattr);
- gf_uuid_copy (local->loc.gfid, local->stbuf.ia_gfid);
+ gf_uuid_copy(local->loc.gfid, local->stbuf.ia_gfid);
- copy = copy_frame (frame);
+ copy = copy_frame(frame);
- if (!copy)
- goto out;
+ if (!copy)
+ goto out;
- copy_local = dht_local_init (copy, &local->loc, NULL, 0);
+ copy_local = dht_local_init(copy, &local->loc, NULL, 0);
- if (!copy_local)
- goto out;
+ if (!copy_local)
+ goto out;
- stbuf = local->stbuf;
- subvol = local->link_subvol;
+ stbuf = local->stbuf;
+ subvol = local->link_subvol;
- copy->local = copy_local;
+ copy->local = copy_local;
- FRAME_SU_DO (copy, dht_local_t);
+ FRAME_SU_DO(copy, dht_local_t);
- STACK_WIND (copy, dht_linkfile_setattr_cbk, subvol,
- subvol->fops->setattr, &copy_local->loc,
- &stbuf, (GF_SET_ATTR_UID | GF_SET_ATTR_GID), xattr);
- ret = 0;
+ STACK_WIND(copy, dht_linkfile_setattr_cbk, subvol, subvol->fops->setattr,
+ &copy_local->loc, &stbuf, (GF_SET_ATTR_UID | GF_SET_ATTR_GID),
+ xattr);
+ ret = 0;
out:
- if ((ret < 0) && (copy))
- DHT_STACK_DESTROY (copy);
+ if ((ret < 0) && (copy))
+ DHT_STACK_DESTROY(copy);
- if (xattr)
- dict_unref (xattr);
+ if (xattr)
+ dict_unref(xattr);
- return ret;
+ return ret;
}
diff --git a/xlators/cluster/dht/src/dht-lock.c b/xlators/cluster/dht/src/dht-lock.c
index 3e82c98bafd..638821ccee5 100644
--- a/xlators/cluster/dht/src/dht-lock.c
+++ b/xlators/cluster/dht/src/dht-lock.c
@@ -11,1235 +11,1250 @@
#include "dht-lock.h"
static char *
-dht_lock_asprintf (dht_lock_t *lock)
+dht_lock_asprintf(dht_lock_t *lock)
{
- char *lk_buf = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0, };
+ char *lk_buf = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {
+ 0,
+ };
- if (lock == NULL)
- goto out;
+ if (lock == NULL)
+ goto out;
- uuid_utoa_r (lock->loc.gfid, gfid);
+ uuid_utoa_r(lock->loc.gfid, gfid);
- gf_asprintf (&lk_buf, "%s:%s", lock->xl->name, gfid);
+ gf_asprintf(&lk_buf, "%s:%s", lock->xl->name, gfid);
out:
- return lk_buf;
+ return lk_buf;
}
static void
-dht_log_lk_array (char *name, gf_loglevel_t log_level, dht_lock_t **lk_array,
- int count)
+dht_log_lk_array(char *name, gf_loglevel_t log_level, dht_lock_t **lk_array,
+ int count)
{
- int i = 0;
- char *lk_buf = NULL;
+ int i = 0;
+ char *lk_buf = NULL;
- if ((lk_array == NULL) || (count == 0))
- goto out;
+ if ((lk_array == NULL) || (count == 0))
+ goto out;
- for (i = 0; i < count; i++) {
- lk_buf = dht_lock_asprintf (lk_array[i]);
- if (!lk_buf)
- goto out;
+ for (i = 0; i < count; i++) {
+ lk_buf = dht_lock_asprintf(lk_array[i]);
+ if (!lk_buf)
+ goto out;
- gf_msg (name, log_level, 0, DHT_MSG_LK_ARRAY_INFO,
- "%d. %s", i, lk_buf);
- GF_FREE (lk_buf);
- }
+ gf_smsg(name, log_level, 0, DHT_MSG_LK_ARRAY_INFO, "index=%d", i,
+ "lk_buf=%s", lk_buf, NULL);
+ GF_FREE(lk_buf);
+ }
out:
- return;
+ return;
}
static void
-dht_lock_stack_destroy (call_frame_t *lock_frame, dht_lock_type_t lk)
+dht_lock_stack_destroy(call_frame_t *lock_frame, dht_lock_type_t lk)
{
- dht_local_t *local = NULL;
+ dht_local_t *local = NULL;
- local = lock_frame->local;
+ local = lock_frame->local;
- if (lk == DHT_INODELK) {
- local->lock[0].layout.my_layout.locks = NULL;
- local->lock[0].layout.my_layout.lk_count = 0;
- } else {
- local->lock[0].ns.directory_ns.locks = NULL;
- local->lock[0].ns.directory_ns.lk_count = 0;
- }
+ if (lk == DHT_INODELK) {
+ local->lock[0].layout.my_layout.locks = NULL;
+ local->lock[0].layout.my_layout.lk_count = 0;
+ } else {
+ local->lock[0].ns.directory_ns.locks = NULL;
+ local->lock[0].ns.directory_ns.lk_count = 0;
+ }
- DHT_STACK_DESTROY (lock_frame);
- return;
+ DHT_STACK_DESTROY(lock_frame);
+ return;
}
static void
-dht_lock_free (dht_lock_t *lock)
+dht_lock_free(dht_lock_t *lock)
{
- if (lock == NULL)
- goto out;
+ if (lock == NULL)
+ goto out;
- loc_wipe (&lock->loc);
- GF_FREE (lock->domain);
- GF_FREE (lock->basename);
- mem_put (lock);
+ loc_wipe(&lock->loc);
+ GF_FREE(lock->domain);
+ GF_FREE(lock->basename);
+ mem_put(lock);
out:
- return;
+ return;
}
static void
-dht_set_lkowner (dht_lock_t **lk_array, int count, gf_lkowner_t *lkowner)
+dht_set_lkowner(dht_lock_t **lk_array, int count, gf_lkowner_t *lkowner)
{
- int i = 0;
+ int i = 0;
- if (!lk_array || !lkowner)
- goto out;
+ if (!lk_array || !lkowner)
+ goto out;
- for (i = 0; i < count; i++) {
- lk_array[i]->lk_owner = *lkowner;
- }
+ for (i = 0; i < count; i++) {
+ lk_array[i]->lk_owner = *lkowner;
+ }
out:
- return;
+ return;
}
static int
-dht_lock_request_cmp (const void *val1, const void *val2)
+dht_lock_request_cmp(const void *val1, const void *val2)
{
- dht_lock_t *lock1 = NULL;
- dht_lock_t *lock2 = NULL;
- int ret = -1;
+ dht_lock_t *lock1 = NULL;
+ dht_lock_t *lock2 = NULL;
+ int ret = -1;
- lock1 = *(dht_lock_t **)val1;
- lock2 = *(dht_lock_t **)val2;
+ lock1 = *(dht_lock_t **)val1;
+ lock2 = *(dht_lock_t **)val2;
- GF_VALIDATE_OR_GOTO ("dht-locks", lock1, out);
- GF_VALIDATE_OR_GOTO ("dht-locks", lock2, out);
+ GF_VALIDATE_OR_GOTO("dht-locks", lock1, out);
+ GF_VALIDATE_OR_GOTO("dht-locks", lock2, out);
- ret = strcmp (lock1->xl->name, lock2->xl->name);
+ ret = strcmp(lock1->xl->name, lock2->xl->name);
- if (ret == 0) {
- ret = gf_uuid_compare (lock1->loc.gfid, lock2->loc.gfid);
- }
+ if (ret == 0) {
+ ret = gf_uuid_compare(lock1->loc.gfid, lock2->loc.gfid);
+ }
out:
- return ret;
+ return ret;
}
static int
-dht_lock_order_requests (dht_lock_t **locks, int count)
+dht_lock_order_requests(dht_lock_t **locks, int count)
{
- int ret = -1;
+ int ret = -1;
- if (!locks || !count)
- goto out;
+ if (!locks || !count)
+ goto out;
- qsort (locks, count, sizeof (*locks), dht_lock_request_cmp);
- ret = 0;
+ qsort(locks, count, sizeof(*locks), dht_lock_request_cmp);
+ ret = 0;
out:
- return ret;
+ return ret;
}
void
-dht_lock_array_free (dht_lock_t **lk_array, int count)
+dht_lock_array_free(dht_lock_t **lk_array, int count)
{
- int i = 0;
- dht_lock_t *lock = NULL;
+ int i = 0;
+ dht_lock_t *lock = NULL;
- if (lk_array == NULL)
- goto out;
+ if (lk_array == NULL)
+ goto out;
- for (i = 0; i < count; i++) {
- lock = lk_array[i];
- lk_array[i] = NULL;
- dht_lock_free (lock);
- }
+ for (i = 0; i < count; i++) {
+ lock = lk_array[i];
+ lk_array[i] = NULL;
+ dht_lock_free(lock);
+ }
out:
- return;
+ return;
}
int32_t
-dht_lock_count (dht_lock_t **lk_array, int lk_count)
+dht_lock_count(dht_lock_t **lk_array, int lk_count)
{
- int i = 0, locked = 0;
+ int i = 0, locked = 0;
- if ((lk_array == NULL) || (lk_count == 0))
- goto out;
+ if ((lk_array == NULL) || (lk_count == 0))
+ goto out;
- for (i = 0; i < lk_count; i++) {
- if (lk_array[i]->locked)
- locked++;
- }
+ for (i = 0; i < lk_count; i++) {
+ if (lk_array[i]->locked)
+ locked++;
+ }
out:
- return locked;
+ return locked;
}
static call_frame_t *
-dht_lock_frame (call_frame_t *parent_frame)
+dht_lock_frame(call_frame_t *parent_frame)
{
- call_frame_t *lock_frame = NULL;
+ call_frame_t *lock_frame = NULL;
- lock_frame = copy_frame (parent_frame);
- if (lock_frame == NULL)
- goto out;
+ lock_frame = copy_frame(parent_frame);
+ if (lock_frame == NULL)
+ goto out;
- set_lk_owner_from_ptr (&lock_frame->root->lk_owner, parent_frame->root);
+ set_lk_owner_from_ptr(&lock_frame->root->lk_owner, parent_frame->root);
out:
- return lock_frame;
+ return lock_frame;
}
dht_lock_t *
-dht_lock_new (xlator_t *this, xlator_t *xl, loc_t *loc, short type,
- const char *domain, const char *basename,
- dht_reaction_type_t do_on_failure)
+dht_lock_new(xlator_t *this, xlator_t *xl, loc_t *loc, short type,
+ const char *domain, const char *basename,
+ dht_reaction_type_t do_on_failure)
{
- dht_conf_t *conf = NULL;
- dht_lock_t *lock = NULL;
-
- conf = this->private;
-
- lock = mem_get0 (conf->lock_pool);
- if (lock == NULL)
- goto out;
-
- lock->xl = xl;
- lock->type = type;
- lock->do_on_failure = do_on_failure;
-
- lock->domain = gf_strdup (domain);
- if (lock->domain == NULL) {
- dht_lock_free (lock);
- lock = NULL;
- goto out;
- }
-
- if (basename) {
- lock->basename = gf_strdup (basename);
- if (lock->basename == NULL) {
- dht_lock_free (lock);
- lock = NULL;
- goto out;
- }
+ dht_conf_t *conf = NULL;
+ dht_lock_t *lock = NULL;
+
+ conf = this->private;
+
+ lock = mem_get0(conf->lock_pool);
+ if (lock == NULL)
+ goto out;
+
+ lock->xl = xl;
+ lock->type = type;
+ lock->do_on_failure = do_on_failure;
+
+ lock->domain = gf_strdup(domain);
+ if (lock->domain == NULL) {
+ dht_lock_free(lock);
+ lock = NULL;
+ goto out;
+ }
+
+ if (basename) {
+ lock->basename = gf_strdup(basename);
+ if (lock->basename == NULL) {
+ dht_lock_free(lock);
+ lock = NULL;
+ goto out;
}
-
- /* Fill only inode and gfid.
- posix and protocol/server give preference to pargfid/basename over
- gfid/inode for resolution if all the three parameters of loc_t are
- present. I want to avoid the following hypothetical situation:
-
- 1. rebalance did a lookup on a dentry and got a gfid.
- 2. rebalance acquires lock on loc_t which was filled with gfid and
- path (pargfid/bname) from step 1.
- 3. somebody deleted and recreated the same file
- 4. rename on the same path acquires lock on loc_t which now points
- to a different inode (and hence gets the lock).
- 5. rebalance continues to migrate file (note that not all fops done
- by rebalance during migration are inode/gfid based Eg., unlink)
- 6. rename continues.
- */
- lock->loc.inode = inode_ref (loc->inode);
- loc_gfid (loc, lock->loc.gfid);
+ }
+
+ /* Fill only inode and gfid.
+ posix and protocol/server give preference to pargfid/basename over
+ gfid/inode for resolution if all the three parameters of loc_t are
+ present. I want to avoid the following hypothetical situation:
+
+ 1. rebalance did a lookup on a dentry and got a gfid.
+ 2. rebalance acquires lock on loc_t which was filled with gfid and
+ path (pargfid/bname) from step 1.
+ 3. somebody deleted and recreated the same file
+ 4. rename on the same path acquires lock on loc_t which now points
+ to a different inode (and hence gets the lock).
+ 5. rebalance continues to migrate file (note that not all fops done
+ by rebalance during migration are inode/gfid based Eg., unlink)
+ 6. rename continues.
+ */
+ lock->loc.inode = inode_ref(loc->inode);
+ loc_gfid(loc, lock->loc.gfid);
out:
- return lock;
+ return lock;
}
static int
-dht_local_entrylk_init (call_frame_t *frame, dht_lock_t **lk_array,
- int lk_count, fop_entrylk_cbk_t entrylk_cbk)
+dht_local_entrylk_init(call_frame_t *frame, dht_lock_t **lk_array, int lk_count,
+ fop_entrylk_cbk_t entrylk_cbk)
{
- int ret = -1;
- dht_local_t *local = NULL;
+ int ret = -1;
+ dht_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local == NULL) {
- local = dht_local_init (frame, NULL, NULL, 0);
- }
+ if (local == NULL) {
+ local = dht_local_init(frame, NULL, NULL, 0);
+ }
- if (local == NULL) {
- goto out;
- }
+ if (local == NULL) {
+ goto out;
+ }
- local->lock[0].ns.directory_ns.entrylk_cbk = entrylk_cbk;
- local->lock[0].ns.directory_ns.locks = lk_array;
- local->lock[0].ns.directory_ns.lk_count = lk_count;
+ local->lock[0].ns.directory_ns.entrylk_cbk = entrylk_cbk;
+ local->lock[0].ns.directory_ns.locks = lk_array;
+ local->lock[0].ns.directory_ns.lk_count = lk_count;
- ret = dht_lock_order_requests (local->lock[0].ns.directory_ns.locks,
- local->lock[0].ns.directory_ns.lk_count);
- if (ret < 0)
- goto out;
+ ret = dht_lock_order_requests(local->lock[0].ns.directory_ns.locks,
+ local->lock[0].ns.directory_ns.lk_count);
+ if (ret < 0)
+ goto out;
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
static void
-dht_entrylk_done (call_frame_t *lock_frame)
+dht_entrylk_done(call_frame_t *lock_frame)
{
- fop_entrylk_cbk_t entrylk_cbk = NULL;
- call_frame_t *main_frame = NULL;
- dht_local_t *local = NULL;
+ fop_entrylk_cbk_t entrylk_cbk = NULL;
+ call_frame_t *main_frame = NULL;
+ dht_local_t *local = NULL;
- local = lock_frame->local;
- main_frame = local->main_frame;
+ local = lock_frame->local;
+ main_frame = local->main_frame;
- local->lock[0].ns.directory_ns.locks = NULL;
- local->lock[0].ns.directory_ns.lk_count = 0;
+ local->lock[0].ns.directory_ns.locks = NULL;
+ local->lock[0].ns.directory_ns.lk_count = 0;
- entrylk_cbk = local->lock[0].ns.directory_ns.entrylk_cbk;
- local->lock[0].ns.directory_ns.entrylk_cbk = NULL;
+ entrylk_cbk = local->lock[0].ns.directory_ns.entrylk_cbk;
+ local->lock[0].ns.directory_ns.entrylk_cbk = NULL;
- entrylk_cbk (main_frame, NULL, main_frame->this,
- local->lock[0].ns.directory_ns.op_ret,
- local->lock[0].ns.directory_ns.op_errno, NULL);
+ entrylk_cbk(main_frame, NULL, main_frame->this,
+ local->lock[0].ns.directory_ns.op_ret,
+ local->lock[0].ns.directory_ns.op_errno, NULL);
- dht_lock_stack_destroy (lock_frame, DHT_ENTRYLK);
- return;
+ dht_lock_stack_destroy(lock_frame, DHT_ENTRYLK);
+ return;
}
static int32_t
-dht_unlock_entrylk_done (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+dht_unlock_entrylk_done(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
+ dht_local_t *local = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- local = frame->local;
- gf_uuid_unparse (local->lock[0].ns.directory_ns.locks[0]->loc.inode->gfid, gfid);
+ local = frame->local;
+ gf_uuid_unparse(local->lock[0].ns.directory_ns.locks[0]->loc.inode->gfid,
+ gfid);
- if (op_ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "unlock failed on gfid: %s, stale lock might be left "
- "in DHT_LAYOUT_HEAL_DOMAIN", gfid);
- }
+ if (op_ret < 0) {
+ gf_smsg(this->name, GF_LOG_WARNING, op_errno,
+ DHT_MSG_UNLOCK_GFID_FAILED, "gfid=%s", gfid,
+ "DHT_LAYOUT_HEAL_DOMAIN", NULL);
+ }
- DHT_STACK_DESTROY (frame);
- return 0;
+ DHT_STACK_DESTROY(frame);
+ return 0;
}
static int32_t
-dht_unlock_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+dht_unlock_entrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int lk_index = 0, call_cnt = 0;
- char gfid[GF_UUID_BUF_SIZE] = {0};
+ dht_local_t *local = NULL;
+ int lk_index = 0, call_cnt = 0;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- lk_index = (long) cookie;
+ lk_index = (long)cookie;
- local = frame->local;
+ local = frame->local;
- uuid_utoa_r (local->lock[0].ns.directory_ns.locks[lk_index]->loc.gfid, gfid);
+ uuid_utoa_r(local->lock[0].ns.directory_ns.locks[lk_index]->loc.gfid, gfid);
- if (op_ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_UNLOCKING_FAILED,
- "unlocking failed on %s:%s",
- local->lock[0].ns.directory_ns.locks[lk_index]->xl->name,
- gfid);
- } else {
- local->lock[0].ns.directory_ns.locks[lk_index]->locked = 0;
- }
+ if (op_ret < 0) {
+ gf_smsg(this->name, GF_LOG_WARNING, op_errno, DHT_MSG_UNLOCKING_FAILED,
+ "name=%s",
+ local->lock[0].ns.directory_ns.locks[lk_index]->xl->name,
+ "gfid=%s", gfid, NULL);
+ } else {
+ local->lock[0].ns.directory_ns.locks[lk_index]->locked = 0;
+ }
- call_cnt = dht_frame_return (frame);
- if (is_last_call (call_cnt)) {
- dht_entrylk_done (frame);
- }
+ call_cnt = dht_frame_return(frame);
+ if (is_last_call(call_cnt)) {
+ dht_entrylk_done(frame);
+ }
- return 0;
+ return 0;
}
static int32_t
-dht_unlock_entrylk (call_frame_t *frame, dht_lock_t **lk_array, int lk_count,
- fop_entrylk_cbk_t entrylk_cbk)
+dht_unlock_entrylk(call_frame_t *frame, dht_lock_t **lk_array, int lk_count,
+ fop_entrylk_cbk_t entrylk_cbk)
{
- dht_local_t *local = NULL;
- int ret = -1 , i = 0;
- call_frame_t *lock_frame = NULL;
- int call_cnt = 0;
-
- GF_VALIDATE_OR_GOTO ("dht-locks", frame, done);
- GF_VALIDATE_OR_GOTO (frame->this->name, lk_array, done);
- GF_VALIDATE_OR_GOTO (frame->this->name, entrylk_cbk, done);
-
- call_cnt = dht_lock_count (lk_array, lk_count);
- if (call_cnt == 0) {
- ret = 0;
- goto done;
- }
-
- lock_frame = dht_lock_frame (frame);
- if (lock_frame == NULL) {
- gf_msg (frame->this->name, GF_LOG_WARNING, 0,
- DHT_MSG_UNLOCKING_FAILED,
- "cannot allocate a frame, not unlocking following "
- "entrylks:");
-
- dht_log_lk_array (frame->this->name, GF_LOG_WARNING, lk_array,
- lk_count);
- goto done;
- }
+ dht_local_t *local = NULL;
+ int ret = -1, i = 0;
+ call_frame_t *lock_frame = NULL;
+ int call_cnt = 0;
- ret = dht_local_entrylk_init (lock_frame, lk_array, lk_count,
- entrylk_cbk);
- if (ret < 0) {
- gf_msg (frame->this->name, GF_LOG_WARNING, 0,
- DHT_MSG_UNLOCKING_FAILED,
- "storing locks in local failed, not unlocking "
- "following entrylks:");
-
- dht_log_lk_array (frame->this->name, GF_LOG_WARNING, lk_array,
- lk_count);
-
- goto done;
- }
-
- local = lock_frame->local;
- local->main_frame = frame;
- local->call_cnt = call_cnt;
-
- for (i = 0; i < local->lock[0].ns.directory_ns.lk_count; i++) {
- if (!local->lock[0].ns.directory_ns.locks[i]->locked)
- continue;
-
- lock_frame->root->lk_owner = local->lock[0].ns.directory_ns.locks[i]->lk_owner;
- STACK_WIND_COOKIE (lock_frame, dht_unlock_entrylk_cbk,
- (void *)(long)i,
- local->lock[0].ns.directory_ns.locks[i]->xl,
- local->lock[0].ns.directory_ns.locks[i]->xl->fops->entrylk,
- local->lock[0].ns.directory_ns.locks[i]->domain,
- &local->lock[0].ns.directory_ns.locks[i]->loc,
- local->lock[0].ns.directory_ns.locks[i]->basename,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK, NULL);
- if (!--call_cnt)
- break;
- }
+ GF_VALIDATE_OR_GOTO("dht-locks", frame, done);
+ GF_VALIDATE_OR_GOTO(frame->this->name, lk_array, done);
+ GF_VALIDATE_OR_GOTO(frame->this->name, entrylk_cbk, done);
- return 0;
+ call_cnt = dht_lock_count(lk_array, lk_count);
+ if (call_cnt == 0) {
+ ret = 0;
+ goto done;
+ }
+
+ lock_frame = dht_lock_frame(frame);
+ if (lock_frame == NULL) {
+ gf_smsg(frame->this->name, GF_LOG_WARNING, 0,
+ DHT_MSG_ALLOC_FRAME_FAILED_NOT_UNLOCKING_FOLLOWING_ENTRYLKS,
+ NULL);
+
+ dht_log_lk_array(frame->this->name, GF_LOG_WARNING, lk_array, lk_count);
+ goto done;
+ }
+
+ ret = dht_local_entrylk_init(lock_frame, lk_array, lk_count, entrylk_cbk);
+ if (ret < 0) {
+ gf_smsg(frame->this->name, GF_LOG_WARNING, 0,
+ DHT_MSG_LOCAL_LOCKS_STORE_FAILED_UNLOCKING_FOLLOWING_ENTRYLK,
+ NULL);
+
+ dht_log_lk_array(frame->this->name, GF_LOG_WARNING, lk_array, lk_count);
+
+ goto done;
+ }
+
+ local = lock_frame->local;
+ local->main_frame = frame;
+ local->call_cnt = call_cnt;
+
+ for (i = 0; i < local->lock[0].ns.directory_ns.lk_count; i++) {
+ if (!local->lock[0].ns.directory_ns.locks[i]->locked)
+ continue;
+
+ lock_frame->root
+ ->lk_owner = local->lock[0].ns.directory_ns.locks[i]->lk_owner;
+ STACK_WIND_COOKIE(
+ lock_frame, dht_unlock_entrylk_cbk, (void *)(long)i,
+ local->lock[0].ns.directory_ns.locks[i]->xl,
+ local->lock[0].ns.directory_ns.locks[i]->xl->fops->entrylk,
+ local->lock[0].ns.directory_ns.locks[i]->domain,
+ &local->lock[0].ns.directory_ns.locks[i]->loc,
+ local->lock[0].ns.directory_ns.locks[i]->basename, ENTRYLK_UNLOCK,
+ ENTRYLK_WRLCK, NULL);
+ if (!--call_cnt)
+ break;
+ }
+
+ return 0;
done:
- if (lock_frame)
- dht_lock_stack_destroy (lock_frame, DHT_ENTRYLK);
+ if (lock_frame)
+ dht_lock_stack_destroy(lock_frame, DHT_ENTRYLK);
- /* no locks acquired, invoke entrylk_cbk */
- if (ret == 0)
- entrylk_cbk (frame, NULL, frame->this, 0, 0, NULL);
+ /* no locks acquired, invoke entrylk_cbk */
+ if (ret == 0)
+ entrylk_cbk(frame, NULL, frame->this, 0, 0, NULL);
- return ret;
+ return ret;
}
int32_t
-dht_unlock_entrylk_wrapper (call_frame_t *frame, dht_elock_wrap_t *entrylk)
+dht_unlock_entrylk_wrapper(call_frame_t *frame, dht_elock_wrap_t *entrylk)
{
- dht_local_t *local = NULL, *lock_local = NULL;
- call_frame_t *lock_frame = NULL;
- char pgfid[GF_UUID_BUF_SIZE] = {0};
- int ret = 0;
-
- local = frame->local;
-
- if (!entrylk || !entrylk->locks)
- goto out;
-
- gf_uuid_unparse (local->loc.parent->gfid, pgfid);
-
- lock_frame = copy_frame (frame);
- if (lock_frame == NULL) {
- gf_msg (frame->this->name, GF_LOG_WARNING, ENOMEM,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "mkdir (%s/%s) (path: %s): "
- "copy frame failed", pgfid, local->loc.name,
- local->loc.path);
- goto done;
- }
+ dht_local_t *local = NULL, *lock_local = NULL;
+ call_frame_t *lock_frame = NULL;
+ char pgfid[GF_UUID_BUF_SIZE] = {0};
+ int ret = 0;
- lock_local = dht_local_init (lock_frame, NULL, NULL, 0);
- if (lock_local == NULL) {
- gf_msg (frame->this->name, GF_LOG_WARNING, ENOMEM,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "mkdir (%s/%s) (path: %s): "
- "local creation failed", pgfid, local->loc.name,
- local->loc.path);
- goto done;
- }
+ local = frame->local;
- lock_frame->local = lock_local;
+ if (!entrylk || !entrylk->locks)
+ goto out;
- lock_local->lock[0].ns.directory_ns.locks = entrylk->locks;
- lock_local->lock[0].ns.directory_ns.lk_count = entrylk->lk_count;
- entrylk->locks = NULL;
- entrylk->lk_count = 0;
+ gf_uuid_unparse(local->loc.parent->gfid, pgfid);
+
+ lock_frame = copy_frame(frame);
+ if (lock_frame == NULL) {
+ gf_smsg(frame->this->name, GF_LOG_WARNING, ENOMEM,
+ DHT_MSG_COPY_FRAME_FAILED, "pgfid=%s", pgfid, "name=%s",
+ local->loc.name, "path=%s", local->loc.path, NULL);
+ goto done;
+ }
- ret = dht_unlock_entrylk (lock_frame,
- lock_local->lock[0].ns.directory_ns.locks,
- lock_local->lock[0].ns.directory_ns.lk_count,
- dht_unlock_entrylk_done);
- if (ret)
- goto done;
+ lock_local = dht_local_init(lock_frame, NULL, NULL, 0);
+ if (lock_local == NULL) {
+ gf_smsg(frame->this->name, GF_LOG_WARNING, ENOMEM,
+ DHT_MSG_CREATE_FAILED, "local", "pgfid=%s", pgfid, "name=%s",
+ local->loc.name, "path=%s", local->loc.path, NULL);
+ goto done;
+ }
- lock_frame = NULL;
+ lock_frame->local = lock_local;
+
+ lock_local->lock[0].ns.directory_ns.locks = entrylk->locks;
+ lock_local->lock[0].ns.directory_ns.lk_count = entrylk->lk_count;
+ entrylk->locks = NULL;
+ entrylk->lk_count = 0;
+
+ ret = dht_unlock_entrylk(
+ lock_frame, lock_local->lock[0].ns.directory_ns.locks,
+ lock_local->lock[0].ns.directory_ns.lk_count, dht_unlock_entrylk_done);
+ if (ret)
+ goto done;
+
+ lock_frame = NULL;
done:
- if (lock_frame != NULL) {
- DHT_STACK_DESTROY (lock_frame);
- }
+ if (lock_frame != NULL) {
+ DHT_STACK_DESTROY(lock_frame);
+ }
out:
- return 0;
+ return 0;
}
static int
-dht_entrylk_cleanup_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xdata)
+dht_entrylk_cleanup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_entrylk_done (frame);
- return 0;
+ dht_entrylk_done(frame);
+ return 0;
}
static void
-dht_entrylk_cleanup (call_frame_t *lock_frame)
+dht_entrylk_cleanup(call_frame_t *lock_frame)
{
- dht_lock_t **lk_array = NULL;
- int lk_count = 0, lk_acquired = 0;
- dht_local_t *local = NULL;
+ dht_lock_t **lk_array = NULL;
+ int lk_count = 0, lk_acquired = 0;
+ dht_local_t *local = NULL;
- local = lock_frame->local;
+ local = lock_frame->local;
- lk_array = local->lock[0].ns.directory_ns.locks;
- lk_count = local->lock[0].ns.directory_ns.lk_count;
+ lk_array = local->lock[0].ns.directory_ns.locks;
+ lk_count = local->lock[0].ns.directory_ns.lk_count;
- lk_acquired = dht_lock_count (lk_array, lk_count);
- if (lk_acquired != 0) {
- dht_unlock_entrylk (lock_frame, lk_array, lk_count,
- dht_entrylk_cleanup_cbk);
- } else {
- dht_entrylk_done (lock_frame);
- }
+ lk_acquired = dht_lock_count(lk_array, lk_count);
+ if (lk_acquired != 0) {
+ dht_unlock_entrylk(lock_frame, lk_array, lk_count,
+ dht_entrylk_cleanup_cbk);
+ } else {
+ dht_entrylk_done(lock_frame);
+ }
- return;
+ return;
}
-
static int32_t
-dht_blocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+dht_blocking_entrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- int lk_index = 0;
- int i = 0;
- dht_local_t *local = NULL;
-
- lk_index = (long) cookie;
-
- local = frame->local;
- if (op_ret == 0) {
- local->lock[0].ns.directory_ns.locks[lk_index]->locked = _gf_true;
- } else {
- switch (op_errno) {
- case ESTALE:
- case ENOENT:
- if (local->lock[0].ns.directory_ns.locks[lk_index]->do_on_failure != IGNORE_ENOENT_ESTALE) {
- local->lock[0].ns.directory_ns.op_ret = -1;
- local->lock[0].ns.directory_ns.op_errno = op_errno;
- goto cleanup;
- }
- break;
- default:
- local->lock[0].ns.directory_ns.op_ret = -1;
- local->lock[0].ns.directory_ns.op_errno = op_errno;
- goto cleanup;
+ int lk_index = 0;
+ int i = 0;
+ dht_local_t *local = NULL;
+
+ lk_index = (long)cookie;
+
+ local = frame->local;
+ if (op_ret == 0) {
+ local->lock[0].ns.directory_ns.locks[lk_index]->locked = _gf_true;
+ } else {
+ switch (op_errno) {
+ case ESTALE:
+ case ENOENT:
+ if (local->lock[0]
+ .ns.directory_ns.locks[lk_index]
+ ->do_on_failure != IGNORE_ENOENT_ESTALE) {
+ local->lock[0].ns.directory_ns.op_ret = -1;
+ local->lock[0].ns.directory_ns.op_errno = op_errno;
+ goto cleanup;
}
+ break;
+ default:
+ local->lock[0].ns.directory_ns.op_ret = -1;
+ local->lock[0].ns.directory_ns.op_errno = op_errno;
+ goto cleanup;
}
+ }
- if (lk_index == (local->lock[0].ns.directory_ns.lk_count - 1)) {
- for (i = 0; (i < local->lock[0].ns.directory_ns.lk_count) &&
- (!local->lock[0].ns.directory_ns.locks[i]->locked); i++)
- ;
-
- if (i == local->lock[0].ns.directory_ns.lk_count) {
- local->lock[0].ns.directory_ns.op_ret = -1;
- local->lock[0].ns.directory_ns.op_errno = op_errno;
- }
+ if (lk_index == (local->lock[0].ns.directory_ns.lk_count - 1)) {
+ for (i = 0; (i < local->lock[0].ns.directory_ns.lk_count) &&
+ (!local->lock[0].ns.directory_ns.locks[i]->locked);
+ i++)
+ ;
- dht_entrylk_done (frame);
- } else {
- dht_blocking_entrylk_rec (frame, ++lk_index);
+ if (i == local->lock[0].ns.directory_ns.lk_count) {
+ local->lock[0].ns.directory_ns.op_ret = -1;
+ local->lock[0].ns.directory_ns.op_errno = op_errno;
}
- return 0;
+ dht_entrylk_done(frame);
+ } else {
+ dht_blocking_entrylk_rec(frame, ++lk_index);
+ }
+
+ return 0;
cleanup:
- dht_entrylk_cleanup (frame);
+ dht_entrylk_cleanup(frame);
- return 0;
+ return 0;
}
void
-dht_blocking_entrylk_rec (call_frame_t *frame, int i)
+dht_blocking_entrylk_rec(call_frame_t *frame, int i)
{
- dht_local_t *local = NULL;
+ dht_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- STACK_WIND_COOKIE (frame, dht_blocking_entrylk_cbk,
- (void *) (long) i,
- local->lock[0].ns.directory_ns.locks[i]->xl,
- local->lock[0].ns.directory_ns.locks[i]->xl->fops->entrylk,
- local->lock[0].ns.directory_ns.locks[i]->domain,
- &local->lock[0].ns.directory_ns.locks[i]->loc,
- local->lock[0].ns.directory_ns.locks[i]->basename,
- ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
+ STACK_WIND_COOKIE(
+ frame, dht_blocking_entrylk_cbk, (void *)(long)i,
+ local->lock[0].ns.directory_ns.locks[i]->xl,
+ local->lock[0].ns.directory_ns.locks[i]->xl->fops->entrylk,
+ local->lock[0].ns.directory_ns.locks[i]->domain,
+ &local->lock[0].ns.directory_ns.locks[i]->loc,
+ local->lock[0].ns.directory_ns.locks[i]->basename, ENTRYLK_LOCK,
+ ENTRYLK_WRLCK, NULL);
- return;
+ return;
}
int
-dht_blocking_entrylk (call_frame_t *frame, dht_lock_t **lk_array,
- int lk_count, fop_entrylk_cbk_t entrylk_cbk)
+dht_blocking_entrylk(call_frame_t *frame, dht_lock_t **lk_array, int lk_count,
+ fop_entrylk_cbk_t entrylk_cbk)
{
- int ret = -1;
- call_frame_t *lock_frame = NULL;
- dht_local_t *local = NULL;
-
- GF_VALIDATE_OR_GOTO ("dht-locks", frame, out);
- GF_VALIDATE_OR_GOTO (frame->this->name, lk_array, out);
- GF_VALIDATE_OR_GOTO (frame->this->name, entrylk_cbk, out);
-
- lock_frame = dht_lock_frame (frame);
- if (lock_frame == NULL)
- goto out;
-
- ret = dht_local_entrylk_init (lock_frame, lk_array, lk_count,
- entrylk_cbk);
- if (ret < 0) {
- goto out;
- }
+ int ret = -1;
+ call_frame_t *lock_frame = NULL;
+ dht_local_t *local = NULL;
+
+ GF_VALIDATE_OR_GOTO("dht-locks", frame, out);
+ GF_VALIDATE_OR_GOTO(frame->this->name, lk_array, out);
+ GF_VALIDATE_OR_GOTO(frame->this->name, entrylk_cbk, out);
+
+ lock_frame = dht_lock_frame(frame);
+ if (lock_frame == NULL)
+ goto out;
- dht_set_lkowner (lk_array, lk_count, &lock_frame->root->lk_owner);
+ ret = dht_local_entrylk_init(lock_frame, lk_array, lk_count, entrylk_cbk);
+ if (ret < 0) {
+ goto out;
+ }
- local = lock_frame->local;
- local->main_frame = frame;
+ dht_set_lkowner(lk_array, lk_count, &lock_frame->root->lk_owner);
- dht_blocking_entrylk_rec (lock_frame, 0);
+ local = lock_frame->local;
+ local->main_frame = frame;
- return 0;
+ dht_blocking_entrylk_rec(lock_frame, 0);
+
+ return 0;
out:
- if (lock_frame)
- dht_lock_stack_destroy (lock_frame, DHT_ENTRYLK);
+ if (lock_frame)
+ dht_lock_stack_destroy(lock_frame, DHT_ENTRYLK);
- return -1;
+ return -1;
}
static int
-dht_local_inodelk_init (call_frame_t *frame, dht_lock_t **lk_array,
- int lk_count, fop_inodelk_cbk_t inodelk_cbk)
+dht_local_inodelk_init(call_frame_t *frame, dht_lock_t **lk_array, int lk_count,
+ fop_inodelk_cbk_t inodelk_cbk)
{
- int ret = -1;
- dht_local_t *local = NULL;
+ int ret = -1;
+ dht_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local == NULL) {
- local = dht_local_init (frame, NULL, NULL, 0);
- }
+ if (local == NULL) {
+ local = dht_local_init(frame, NULL, NULL, 0);
+ }
- if (local == NULL) {
- goto out;
- }
+ if (local == NULL) {
+ goto out;
+ }
- local->lock[0].layout.my_layout.inodelk_cbk = inodelk_cbk;
- local->lock[0].layout.my_layout.locks = lk_array;
- local->lock[0].layout.my_layout.lk_count = lk_count;
+ local->lock[0].layout.my_layout.inodelk_cbk = inodelk_cbk;
+ local->lock[0].layout.my_layout.locks = lk_array;
+ local->lock[0].layout.my_layout.lk_count = lk_count;
- ret = dht_lock_order_requests (local->lock[0].layout.my_layout.locks,
- local->lock[0].layout.my_layout.lk_count);
- if (ret < 0)
- goto out;
+ ret = dht_lock_order_requests(local->lock[0].layout.my_layout.locks,
+ local->lock[0].layout.my_layout.lk_count);
+ if (ret < 0)
+ goto out;
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
static void
-dht_inodelk_done (call_frame_t *lock_frame)
+dht_inodelk_done(call_frame_t *lock_frame)
{
- fop_inodelk_cbk_t inodelk_cbk = NULL;
- call_frame_t *main_frame = NULL;
- dht_local_t *local = NULL;
+ fop_inodelk_cbk_t inodelk_cbk = NULL;
+ call_frame_t *main_frame = NULL;
+ dht_local_t *local = NULL;
- local = lock_frame->local;
- main_frame = local->main_frame;
+ local = lock_frame->local;
+ main_frame = local->main_frame;
- local->lock[0].layout.my_layout.locks = NULL;
- local->lock[0].layout.my_layout.lk_count = 0;
+ local->lock[0].layout.my_layout.locks = NULL;
+ local->lock[0].layout.my_layout.lk_count = 0;
- inodelk_cbk = local->lock[0].layout.my_layout.inodelk_cbk;
- local->lock[0].layout.my_layout.inodelk_cbk = NULL;
+ inodelk_cbk = local->lock[0].layout.my_layout.inodelk_cbk;
+ local->lock[0].layout.my_layout.inodelk_cbk = NULL;
- inodelk_cbk (main_frame, NULL, main_frame->this,
- local->lock[0].layout.my_layout.op_ret,
- local->lock[0].layout.my_layout.op_errno, NULL);
+ inodelk_cbk(main_frame, NULL, main_frame->this,
+ local->lock[0].layout.my_layout.op_ret,
+ local->lock[0].layout.my_layout.op_errno, NULL);
- dht_lock_stack_destroy (lock_frame, DHT_INODELK);
- return;
+ dht_lock_stack_destroy(lock_frame, DHT_INODELK);
+ return;
}
static int32_t
-dht_unlock_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+dht_unlock_inodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int lk_index = 0, call_cnt = 0;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- lk_index = (long) cookie;
-
- local = frame->local;
- if (op_ret < 0) {
- uuid_utoa_r (local->lock[0].layout.my_layout.locks[lk_index]->loc.gfid,
- gfid);
-
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_UNLOCKING_FAILED,
- "unlocking failed on %s:%s",
- local->lock[0].layout.my_layout.locks[lk_index]->xl->name,
- gfid);
- } else {
- local->lock[0].layout.my_layout.locks[lk_index]->locked = 0;
- }
-
- call_cnt = dht_frame_return (frame);
- if (is_last_call (call_cnt)) {
- dht_inodelk_done (frame);
- }
-
- return 0;
+ dht_local_t *local = NULL;
+ int lk_index = 0, call_cnt = 0;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+
+ lk_index = (long)cookie;
+
+ local = frame->local;
+ if (op_ret < 0) {
+ uuid_utoa_r(local->lock[0].layout.my_layout.locks[lk_index]->loc.gfid,
+ gfid);
+
+ gf_smsg(this->name, GF_LOG_WARNING, op_errno, DHT_MSG_UNLOCKING_FAILED,
+ "name=%s",
+ local->lock[0].layout.my_layout.locks[lk_index]->xl->name,
+ "gfid=%s", gfid, NULL);
+ } else {
+ local->lock[0].layout.my_layout.locks[lk_index]->locked = 0;
+ }
+
+ call_cnt = dht_frame_return(frame);
+ if (is_last_call(call_cnt)) {
+ dht_inodelk_done(frame);
+ }
+
+ return 0;
}
static int32_t
-dht_unlock_inodelk_done (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+dht_unlock_inodelk_done(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
+ dht_local_t *local = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- local = frame->local;
- gf_uuid_unparse (local->lock[0].layout.my_layout.locks[0]->loc.inode->gfid, gfid);
+ local = frame->local;
+ gf_uuid_unparse(local->lock[0].layout.my_layout.locks[0]->loc.inode->gfid,
+ gfid);
- if (op_ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "unlock failed on gfid: %s, stale lock might be left "
- "in DHT_LAYOUT_HEAL_DOMAIN", gfid);
- }
+ if (op_ret < 0) {
+ gf_smsg(this->name, GF_LOG_WARNING, op_errno,
+ DHT_MSG_UNLOCK_GFID_FAILED, "DHT_LAYOUT_HEAL_DOMAIN gfid=%s",
+ gfid, NULL);
+ }
- DHT_STACK_DESTROY (frame);
- return 0;
+ DHT_STACK_DESTROY(frame);
+ return 0;
}
int32_t
-dht_unlock_inodelk (call_frame_t *frame, dht_lock_t **lk_array, int lk_count,
- fop_inodelk_cbk_t inodelk_cbk)
+dht_unlock_inodelk(call_frame_t *frame, dht_lock_t **lk_array, int lk_count,
+ fop_inodelk_cbk_t inodelk_cbk)
{
- dht_local_t *local = NULL;
- struct gf_flock flock = {0,};
- int ret = -1 , i = 0;
- call_frame_t *lock_frame = NULL;
- int call_cnt = 0;
-
- GF_VALIDATE_OR_GOTO ("dht-locks", frame, done);
- GF_VALIDATE_OR_GOTO (frame->this->name, lk_array, done);
- GF_VALIDATE_OR_GOTO (frame->this->name, inodelk_cbk, done);
-
- call_cnt = dht_lock_count (lk_array, lk_count);
- if (call_cnt == 0) {
- ret = 0;
- goto done;
- }
-
- lock_frame = dht_lock_frame (frame);
- if (lock_frame == NULL) {
- gf_msg (frame->this->name, GF_LOG_WARNING, 0,
- DHT_MSG_UNLOCKING_FAILED,
- "cannot allocate a frame, not unlocking following "
- "locks:");
-
- dht_log_lk_array (frame->this->name, GF_LOG_WARNING, lk_array,
- lk_count);
- goto done;
- }
-
- ret = dht_local_inodelk_init (lock_frame, lk_array, lk_count,
- inodelk_cbk);
- if (ret < 0) {
- gf_msg (frame->this->name, GF_LOG_WARNING, 0,
- DHT_MSG_UNLOCKING_FAILED,
- "storing locks in local failed, not unlocking "
- "following locks:");
-
- dht_log_lk_array (frame->this->name, GF_LOG_WARNING, lk_array,
- lk_count);
-
- goto done;
- }
-
- local = lock_frame->local;
- local->main_frame = frame;
- local->call_cnt = call_cnt;
-
- flock.l_type = F_UNLCK;
-
- for (i = 0; i < local->lock[0].layout.my_layout.lk_count; i++) {
- if (!local->lock[0].layout.my_layout.locks[i]->locked)
- continue;
-
- lock_frame->root->lk_owner = local->lock[0].layout.my_layout.locks[i]->lk_owner;
- STACK_WIND_COOKIE (lock_frame, dht_unlock_inodelk_cbk,
- (void *)(long)i,
- local->lock[0].layout.my_layout.locks[i]->xl,
- local->lock[0].layout.my_layout.locks[i]->xl->fops->inodelk,
- local->lock[0].layout.my_layout.locks[i]->domain,
- &local->lock[0].layout.my_layout.locks[i]->loc, F_SETLK,
- &flock, NULL);
- if (!--call_cnt)
- break;
- }
-
- return 0;
+ dht_local_t *local = NULL;
+ struct gf_flock flock = {
+ 0,
+ };
+ int ret = -1, i = 0;
+ call_frame_t *lock_frame = NULL;
+ int call_cnt = 0;
+
+ GF_VALIDATE_OR_GOTO("dht-locks", frame, done);
+ GF_VALIDATE_OR_GOTO(frame->this->name, lk_array, done);
+ GF_VALIDATE_OR_GOTO(frame->this->name, inodelk_cbk, done);
+
+ call_cnt = dht_lock_count(lk_array, lk_count);
+ if (call_cnt == 0) {
+ ret = 0;
+ goto done;
+ }
+
+ lock_frame = dht_lock_frame(frame);
+ if (lock_frame == NULL) {
+ gf_smsg(frame->this->name, GF_LOG_WARNING, 0,
+ DHT_MSG_ALLOC_FRAME_FAILED_NOT_UNLOCKING_FOLLOWING_ENTRYLKS,
+ NULL);
+
+ dht_log_lk_array(frame->this->name, GF_LOG_WARNING, lk_array, lk_count);
+ goto done;
+ }
+
+ ret = dht_local_inodelk_init(lock_frame, lk_array, lk_count, inodelk_cbk);
+ if (ret < 0) {
+ gf_smsg(frame->this->name, GF_LOG_WARNING, 0,
+ DHT_MSG_LOCAL_LOCKS_STORE_FAILED_UNLOCKING_FOLLOWING_ENTRYLK,
+ NULL);
+
+ dht_log_lk_array(frame->this->name, GF_LOG_WARNING, lk_array, lk_count);
+
+ goto done;
+ }
+
+ local = lock_frame->local;
+ local->main_frame = frame;
+ local->call_cnt = call_cnt;
+
+ flock.l_type = F_UNLCK;
+
+ for (i = 0; i < local->lock[0].layout.my_layout.lk_count; i++) {
+ if (!local->lock[0].layout.my_layout.locks[i]->locked)
+ continue;
+
+ lock_frame->root
+ ->lk_owner = local->lock[0].layout.my_layout.locks[i]->lk_owner;
+ STACK_WIND_COOKIE(
+ lock_frame, dht_unlock_inodelk_cbk, (void *)(long)i,
+ local->lock[0].layout.my_layout.locks[i]->xl,
+ local->lock[0].layout.my_layout.locks[i]->xl->fops->inodelk,
+ local->lock[0].layout.my_layout.locks[i]->domain,
+ &local->lock[0].layout.my_layout.locks[i]->loc, F_SETLK, &flock,
+ NULL);
+ if (!--call_cnt)
+ break;
+ }
+
+ return 0;
done:
- if (lock_frame)
- dht_lock_stack_destroy (lock_frame, DHT_INODELK);
+ if (lock_frame)
+ dht_lock_stack_destroy(lock_frame, DHT_INODELK);
- /* no locks acquired, invoke inodelk_cbk */
- if (ret == 0)
- inodelk_cbk (frame, NULL, frame->this, 0, 0, NULL);
+ /* no locks acquired, invoke inodelk_cbk */
+ if (ret == 0)
+ inodelk_cbk(frame, NULL, frame->this, 0, 0, NULL);
- return ret;
+ return ret;
}
int32_t
-dht_unlock_inodelk_wrapper (call_frame_t *frame, dht_ilock_wrap_t *inodelk)
+dht_unlock_inodelk_wrapper(call_frame_t *frame, dht_ilock_wrap_t *inodelk)
{
- dht_local_t *local = NULL, *lock_local = NULL;
- call_frame_t *lock_frame = NULL;
- char pgfid[GF_UUID_BUF_SIZE] = {0};
- int ret = 0;
-
- local = frame->local;
-
- if (!inodelk || !inodelk->locks)
- goto out;
-
- gf_uuid_unparse (local->loc.parent->gfid, pgfid);
-
- lock_frame = copy_frame (frame);
- if (lock_frame == NULL) {
- gf_msg (frame->this->name, GF_LOG_WARNING, ENOMEM,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "mkdir (%s/%s) (path: %s): "
- "copy frame failed", pgfid, local->loc.name,
- local->loc.path);
- goto done;
- }
+ dht_local_t *local = NULL, *lock_local = NULL;
+ call_frame_t *lock_frame = NULL;
+ char pgfid[GF_UUID_BUF_SIZE] = {0};
+ int ret = 0;
- lock_local = dht_local_init (lock_frame, NULL, NULL, 0);
- if (lock_local == NULL) {
- gf_msg (frame->this->name, GF_LOG_WARNING, ENOMEM,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- "mkdir (%s/%s) (path: %s): "
- "local creation failed", pgfid, local->loc.name,
- local->loc.path);
- goto done;
- }
+ local = frame->local;
- lock_frame->local = lock_local;
+ if (!inodelk || !inodelk->locks)
+ goto out;
- lock_local->lock[0].layout.my_layout.locks = inodelk->locks;
- lock_local->lock[0].layout.my_layout.lk_count = inodelk->lk_count;
- inodelk->locks = NULL;
- inodelk->lk_count = 0;
+ gf_uuid_unparse(local->loc.parent->gfid, pgfid);
+
+ lock_frame = copy_frame(frame);
+ if (lock_frame == NULL) {
+ gf_smsg(frame->this->name, GF_LOG_WARNING, ENOMEM,
+ DHT_MSG_COPY_FRAME_FAILED, "pgfid=%s", pgfid, "name=%s",
+ local->loc.name, "path=%s", local->loc.path, NULL);
+ goto done;
+ }
- ret = dht_unlock_inodelk (lock_frame,
- lock_local->lock[0].layout.my_layout.locks,
- lock_local->lock[0].layout.my_layout.lk_count,
- dht_unlock_inodelk_done);
+ lock_local = dht_local_init(lock_frame, NULL, NULL, 0);
+ if (lock_local == NULL) {
+ gf_smsg(frame->this->name, GF_LOG_WARNING, ENOMEM,
+ DHT_MSG_CREATE_FAILED, "local", "gfid=%s", pgfid, "name=%s",
+ local->loc.name, "path=%s", local->loc.path, NULL);
+ goto done;
+ }
- if (ret)
- goto done;
+ lock_frame->local = lock_local;
- lock_frame = NULL;
+ lock_local->lock[0].layout.my_layout.locks = inodelk->locks;
+ lock_local->lock[0].layout.my_layout.lk_count = inodelk->lk_count;
+ inodelk->locks = NULL;
+ inodelk->lk_count = 0;
+
+ ret = dht_unlock_inodelk(
+ lock_frame, lock_local->lock[0].layout.my_layout.locks,
+ lock_local->lock[0].layout.my_layout.lk_count, dht_unlock_inodelk_done);
+
+ if (ret)
+ goto done;
+
+ lock_frame = NULL;
done:
- if (lock_frame != NULL) {
- DHT_STACK_DESTROY (lock_frame);
- }
+ if (lock_frame != NULL) {
+ DHT_STACK_DESTROY(lock_frame);
+ }
out:
- return 0;
+ return 0;
}
static int
-dht_inodelk_cleanup_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xdata)
+dht_inodelk_cleanup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_inodelk_done (frame);
- return 0;
+ dht_inodelk_done(frame);
+ return 0;
}
static void
-dht_inodelk_cleanup (call_frame_t *lock_frame)
+dht_inodelk_cleanup(call_frame_t *lock_frame)
{
- dht_lock_t **lk_array = NULL;
- int lk_count = 0, lk_acquired = 0;
- dht_local_t *local = NULL;
+ dht_lock_t **lk_array = NULL;
+ int lk_count = 0, lk_acquired = 0;
+ dht_local_t *local = NULL;
- local = lock_frame->local;
+ local = lock_frame->local;
- lk_array = local->lock[0].layout.my_layout.locks;
- lk_count = local->lock[0].layout.my_layout.lk_count;
+ lk_array = local->lock[0].layout.my_layout.locks;
+ lk_count = local->lock[0].layout.my_layout.lk_count;
- lk_acquired = dht_lock_count (lk_array, lk_count);
- if (lk_acquired != 0) {
- dht_unlock_inodelk (lock_frame, lk_array, lk_count,
- dht_inodelk_cleanup_cbk);
- } else {
- dht_inodelk_done (lock_frame);
- }
+ lk_acquired = dht_lock_count(lk_array, lk_count);
+ if (lk_acquired != 0) {
+ dht_unlock_inodelk(lock_frame, lk_array, lk_count,
+ dht_inodelk_cleanup_cbk);
+ } else {
+ dht_inodelk_done(lock_frame);
+ }
- return;
+ return;
}
static int32_t
-dht_nonblocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+dht_nonblocking_inodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int lk_index = 0, call_cnt = 0;
- char gfid[GF_UUID_BUF_SIZE] = {0};
+ dht_local_t *local = NULL;
+ int lk_index = 0, call_cnt = 0;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- local = frame->local;
- lk_index = (long) cookie;
+ local = frame->local;
+ lk_index = (long)cookie;
- if (op_ret == -1) {
- local->lock[0].layout.my_layout.op_ret = -1;
- local->lock[0].layout.my_layout.op_errno = op_errno;
+ if (op_ret == -1) {
+ local->lock[0].layout.my_layout.op_ret = -1;
+ local->lock[0].layout.my_layout.op_errno = op_errno;
- if (local && local->lock[0].layout.my_layout.locks[lk_index]) {
- uuid_utoa_r (local->lock[0].layout.my_layout.locks[lk_index]->loc.inode->gfid,
- gfid);
-
- gf_msg_debug (this->name, op_errno,
- "inodelk failed on gfid: %s "
- "subvolume: %s", gfid,
- local->lock[0].layout.my_layout.locks[lk_index]->xl->name);
- }
+ if (local && local->lock[0].layout.my_layout.locks[lk_index]) {
+ uuid_utoa_r(local->lock[0]
+ .layout.my_layout.locks[lk_index]
+ ->loc.inode->gfid,
+ gfid);
- goto out;
+ gf_msg_debug(
+ this->name, op_errno,
+ "inodelk failed on gfid: %s "
+ "subvolume: %s",
+ gfid,
+ local->lock[0].layout.my_layout.locks[lk_index]->xl->name);
}
- local->lock[0].layout.my_layout.locks[lk_index]->locked = _gf_true;
+ goto out;
+ }
-out:
- call_cnt = dht_frame_return (frame);
- if (is_last_call (call_cnt)) {
- if (local->lock[0].layout.my_layout.op_ret < 0) {
- dht_inodelk_cleanup (frame);
- return 0;
- }
+ local->lock[0].layout.my_layout.locks[lk_index]->locked = _gf_true;
- dht_inodelk_done (frame);
+out:
+ call_cnt = dht_frame_return(frame);
+ if (is_last_call(call_cnt)) {
+ if (local->lock[0].layout.my_layout.op_ret < 0) {
+ dht_inodelk_cleanup(frame);
+ return 0;
}
- return 0;
+ dht_inodelk_done(frame);
+ }
+
+ return 0;
}
int
-dht_nonblocking_inodelk (call_frame_t *frame, dht_lock_t **lk_array,
- int lk_count, fop_inodelk_cbk_t inodelk_cbk)
+dht_nonblocking_inodelk(call_frame_t *frame, dht_lock_t **lk_array,
+ int lk_count, fop_inodelk_cbk_t inodelk_cbk)
{
- struct gf_flock flock = {0,};
- int i = 0, ret = 0;
- dht_local_t *local = NULL;
- call_frame_t *lock_frame = NULL;
-
- GF_VALIDATE_OR_GOTO ("dht-locks", frame, out);
- GF_VALIDATE_OR_GOTO (frame->this->name, lk_array, out);
- GF_VALIDATE_OR_GOTO (frame->this->name, inodelk_cbk, out);
-
- lock_frame = dht_lock_frame (frame);
- if (lock_frame == NULL)
- goto out;
-
- ret = dht_local_inodelk_init (lock_frame, lk_array, lk_count,
- inodelk_cbk);
- if (ret < 0) {
- goto out;
- }
+ struct gf_flock flock = {
+ 0,
+ };
+ int i = 0, ret = 0;
+ dht_local_t *local = NULL;
+ call_frame_t *lock_frame = NULL;
- dht_set_lkowner (lk_array, lk_count, &lock_frame->root->lk_owner);
+ GF_VALIDATE_OR_GOTO("dht-locks", frame, out);
+ GF_VALIDATE_OR_GOTO(frame->this->name, lk_array, out);
+ GF_VALIDATE_OR_GOTO(frame->this->name, inodelk_cbk, out);
- local = lock_frame->local;
- local->main_frame = frame;
+ lock_frame = dht_lock_frame(frame);
+ if (lock_frame == NULL)
+ goto out;
- local->call_cnt = lk_count;
+ ret = dht_local_inodelk_init(lock_frame, lk_array, lk_count, inodelk_cbk);
+ if (ret < 0) {
+ goto out;
+ }
- for (i = 0; i < lk_count; i++) {
- flock.l_type = local->lock[0].layout.my_layout.locks[i]->type;
+ dht_set_lkowner(lk_array, lk_count, &lock_frame->root->lk_owner);
- STACK_WIND_COOKIE (lock_frame, dht_nonblocking_inodelk_cbk,
- (void *) (long) i,
- local->lock[0].layout.my_layout.locks[i]->xl,
- local->lock[0].layout.my_layout.locks[i]->xl->fops->inodelk,
- local->lock[0].layout.my_layout.locks[i]->domain,
- &local->lock[0].layout.my_layout.locks[i]->loc,
- F_SETLK,
- &flock, NULL);
- }
+ local = lock_frame->local;
+ local->main_frame = frame;
+
+ local->call_cnt = lk_count;
+
+ for (i = 0; i < lk_count; i++) {
+ flock.l_type = local->lock[0].layout.my_layout.locks[i]->type;
+
+ STACK_WIND_COOKIE(
+ lock_frame, dht_nonblocking_inodelk_cbk, (void *)(long)i,
+ local->lock[0].layout.my_layout.locks[i]->xl,
+ local->lock[0].layout.my_layout.locks[i]->xl->fops->inodelk,
+ local->lock[0].layout.my_layout.locks[i]->domain,
+ &local->lock[0].layout.my_layout.locks[i]->loc, F_SETLK, &flock,
+ NULL);
+ }
- return 0;
+ return 0;
out:
- if (lock_frame)
- dht_lock_stack_destroy (lock_frame, DHT_INODELK);
+ if (lock_frame)
+ dht_lock_stack_destroy(lock_frame, DHT_INODELK);
- return -1;
+ return -1;
}
static int32_t
-dht_blocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+dht_blocking_inodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- int lk_index = 0;
- int i = 0;
- dht_local_t *local = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0,};
-
- lk_index = (long) cookie;
-
- local = frame->local;
- if (op_ret == 0) {
- local->lock[0].layout.my_layout.locks[lk_index]->locked = _gf_true;
- } else {
- switch (op_errno) {
- case ESTALE:
- case ENOENT:
- if (local->lock[0].layout.my_layout.locks[lk_index]->do_on_failure
- != IGNORE_ENOENT_ESTALE) {
- gf_uuid_unparse (local->lock[0].layout.my_layout.locks[lk_index]->loc.gfid, gfid);
- local->lock[0].layout.my_layout.op_ret = -1;
- local->lock[0].layout.my_layout.op_errno = op_errno;
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- DHT_MSG_INODELK_FAILED,
- "inodelk failed on subvol %s. gfid:%s",
- local->lock[0].layout.my_layout.locks[lk_index]->xl->name,
- gfid);
- goto cleanup;
- }
- break;
- default:
- gf_uuid_unparse (local->lock[0].layout.my_layout.locks[lk_index]->loc.gfid, gfid);
- local->lock[0].layout.my_layout.op_ret = -1;
- local->lock[0].layout.my_layout.op_errno = op_errno;
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- DHT_MSG_INODELK_FAILED,
- "inodelk failed on subvol %s, gfid:%s",
- local->lock[0].layout.my_layout.locks[lk_index]->xl->name, gfid);
- goto cleanup;
+ int lk_index = 0;
+ int i = 0;
+ dht_local_t *local = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {
+ 0,
+ };
+ dht_reaction_type_t reaction = 0;
+
+ lk_index = (long)cookie;
+
+ local = frame->local;
+ if (op_ret == 0) {
+ local->lock[0].layout.my_layout.locks[lk_index]->locked = _gf_true;
+ } else {
+ switch (op_errno) {
+ case ESTALE:
+ case ENOENT:
+ reaction = local->lock[0]
+ .layout.my_layout.locks[lk_index]
+ ->do_on_failure;
+ if ((reaction != IGNORE_ENOENT_ESTALE) &&
+ (reaction != IGNORE_ENOENT_ESTALE_EIO)) {
+ gf_uuid_unparse(local->lock[0]
+ .layout.my_layout.locks[lk_index]
+ ->loc.gfid,
+ gfid);
+ local->lock[0].layout.my_layout.op_ret = -1;
+ local->lock[0].layout.my_layout.op_errno = op_errno;
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno,
+ DHT_MSG_INODELK_FAILED, "subvol=%s",
+ local->lock[0]
+ .layout.my_layout.locks[lk_index]
+ ->xl->name,
+ "gfid=%s", gfid, NULL);
+ goto cleanup;
}
- }
+ break;
+ case EIO:
+ reaction = local->lock[0]
+ .layout.my_layout.locks[lk_index]
+ ->do_on_failure;
+ if (reaction != IGNORE_ENOENT_ESTALE_EIO) {
+ gf_uuid_unparse(local->lock[0]
+ .layout.my_layout.locks[lk_index]
+ ->loc.gfid,
+ gfid);
+ local->lock[0].layout.my_layout.op_ret = -1;
+ local->lock[0].layout.my_layout.op_errno = op_errno;
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno,
+ DHT_MSG_INODELK_FAILED, "subvol=%s",
+ local->lock[0]
+ .layout.my_layout.locks[lk_index]
+ ->xl->name,
+ "gfid=%s", gfid, NULL);
+ goto cleanup;
+ }
+ break;
- if (lk_index == (local->lock[0].layout.my_layout.lk_count - 1)) {
- for (i = 0; (i < local->lock[0].layout.my_layout.lk_count) &&
- (!local->lock[0].layout.my_layout.locks[i]->locked); i++)
- ;
+ default:
+ gf_uuid_unparse(
+ local->lock[0].layout.my_layout.locks[lk_index]->loc.gfid,
+ gfid);
+ local->lock[0].layout.my_layout.op_ret = -1;
+ local->lock[0].layout.my_layout.op_errno = op_errno;
+ gf_smsg(
+ this->name, GF_LOG_ERROR, op_errno, DHT_MSG_INODELK_FAILED,
+ "subvol=%s",
+ local->lock[0].layout.my_layout.locks[lk_index]->xl->name,
+ "gfid=%s", gfid, NULL);
+ goto cleanup;
+ }
+ }
- if (i == local->lock[0].layout.my_layout.lk_count) {
- local->lock[0].layout.my_layout.op_ret = -1;
- local->lock[0].layout.my_layout.op_errno = op_errno;
- }
+ if (lk_index == (local->lock[0].layout.my_layout.lk_count - 1)) {
+ for (i = 0; (i < local->lock[0].layout.my_layout.lk_count) &&
+ (!local->lock[0].layout.my_layout.locks[i]->locked);
+ i++)
+ ;
- dht_inodelk_done (frame);
- } else {
- dht_blocking_inodelk_rec (frame, ++lk_index);
+ if (i == local->lock[0].layout.my_layout.lk_count) {
+ local->lock[0].layout.my_layout.op_ret = -1;
+ local->lock[0].layout.my_layout.op_errno = op_errno;
}
- return 0;
+ dht_inodelk_done(frame);
+ } else {
+ dht_blocking_inodelk_rec(frame, ++lk_index);
+ }
+
+ return 0;
cleanup:
- dht_inodelk_cleanup (frame);
+ dht_inodelk_cleanup(frame);
- return 0;
+ return 0;
}
void
-dht_blocking_inodelk_rec (call_frame_t *frame, int i)
+dht_blocking_inodelk_rec(call_frame_t *frame, int i)
{
- dht_local_t *local = NULL;
- struct gf_flock flock = {0,};
+ dht_local_t *local = NULL;
+ struct gf_flock flock = {
+ 0,
+ };
- local = frame->local;
+ local = frame->local;
- flock.l_type = local->lock[0].layout.my_layout.locks[i]->type;
+ flock.l_type = local->lock[0].layout.my_layout.locks[i]->type;
- STACK_WIND_COOKIE (frame, dht_blocking_inodelk_cbk,
- (void *) (long) i,
- local->lock[0].layout.my_layout.locks[i]->xl,
- local->lock[0].layout.my_layout.locks[i]->xl->fops->inodelk,
- local->lock[0].layout.my_layout.locks[i]->domain,
- &local->lock[0].layout.my_layout.locks[i]->loc,
- F_SETLKW,
- &flock, NULL);
+ STACK_WIND_COOKIE(
+ frame, dht_blocking_inodelk_cbk, (void *)(long)i,
+ local->lock[0].layout.my_layout.locks[i]->xl,
+ local->lock[0].layout.my_layout.locks[i]->xl->fops->inodelk,
+ local->lock[0].layout.my_layout.locks[i]->domain,
+ &local->lock[0].layout.my_layout.locks[i]->loc, F_SETLKW, &flock, NULL);
- return;
+ return;
}
int
-dht_blocking_inodelk (call_frame_t *frame, dht_lock_t **lk_array,
- int lk_count, fop_inodelk_cbk_t inodelk_cbk)
+dht_blocking_inodelk(call_frame_t *frame, dht_lock_t **lk_array, int lk_count,
+ fop_inodelk_cbk_t inodelk_cbk)
{
- int ret = -1;
- call_frame_t *lock_frame = NULL;
- dht_local_t *local = NULL;
- dht_local_t *tmp_local = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0,};
-
- GF_VALIDATE_OR_GOTO ("dht-locks", frame, out);
- GF_VALIDATE_OR_GOTO (frame->this->name, lk_array, out);
- GF_VALIDATE_OR_GOTO (frame->this->name, inodelk_cbk, out);
-
- tmp_local = frame->local;
-
- lock_frame = dht_lock_frame (frame);
- if (lock_frame == NULL) {
- gf_uuid_unparse (tmp_local->loc.gfid, gfid);
- gf_msg ("dht", GF_LOG_ERROR, ENOMEM,
- DHT_MSG_LOCK_FRAME_FAILED,
- "memory allocation failed for lock_frame. gfid:%s"
- " path:%s", gfid, tmp_local->loc.path);
- goto out;
- }
-
- ret = dht_local_inodelk_init (lock_frame, lk_array, lk_count,
- inodelk_cbk);
- if (ret < 0) {
- gf_uuid_unparse (tmp_local->loc.gfid, gfid);
- gf_msg ("dht", GF_LOG_ERROR, ENOMEM,
- DHT_MSG_LOCAL_LOCK_INIT_FAILED,
- "dht_local_lock_init failed, gfid: %s path:%s", gfid,
- tmp_local->loc.path);
- goto out;
- }
-
- dht_set_lkowner (lk_array, lk_count, &lock_frame->root->lk_owner);
-
- local = lock_frame->local;
- local->main_frame = frame;
-
- dht_blocking_inodelk_rec (lock_frame, 0);
-
- return 0;
+ int ret = -1;
+ call_frame_t *lock_frame = NULL;
+ dht_local_t *local = NULL;
+ dht_local_t *tmp_local = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("dht-locks", frame, out);
+ GF_VALIDATE_OR_GOTO(frame->this->name, lk_array, out);
+ GF_VALIDATE_OR_GOTO(frame->this->name, inodelk_cbk, out);
+
+ tmp_local = frame->local;
+
+ lock_frame = dht_lock_frame(frame);
+ if (lock_frame == NULL) {
+ gf_uuid_unparse(tmp_local->loc.gfid, gfid);
+ gf_smsg("dht", GF_LOG_ERROR, ENOMEM, DHT_MSG_LOCK_FRAME_FAILED,
+ "gfid=%s", gfid, "path=%s", tmp_local->loc.path, NULL);
+ goto out;
+ }
+
+ ret = dht_local_inodelk_init(lock_frame, lk_array, lk_count, inodelk_cbk);
+ if (ret < 0) {
+ gf_uuid_unparse(tmp_local->loc.gfid, gfid);
+ gf_smsg("dht", GF_LOG_ERROR, ENOMEM, DHT_MSG_LOCAL_LOCK_INIT_FAILED,
+ "gfid=%s", gfid, "path=%s", tmp_local->loc.path, NULL);
+ goto out;
+ }
+
+ dht_set_lkowner(lk_array, lk_count, &lock_frame->root->lk_owner);
+
+ local = lock_frame->local;
+ local->main_frame = frame;
+
+ dht_blocking_inodelk_rec(lock_frame, 0);
+
+ return 0;
out:
- if (lock_frame)
- dht_lock_stack_destroy (lock_frame, DHT_INODELK);
+ if (lock_frame)
+ dht_lock_stack_destroy(lock_frame, DHT_INODELK);
- return -1;
+ return -1;
}
void
-dht_unlock_namespace (call_frame_t *frame, dht_dir_transaction_t *lock)
+dht_unlock_namespace(call_frame_t *frame, dht_dir_transaction_t *lock)
{
- GF_VALIDATE_OR_GOTO ("dht-locks", frame, out);
- GF_VALIDATE_OR_GOTO (frame->this->name, lock, out);
+ GF_VALIDATE_OR_GOTO("dht-locks", frame, out);
+ GF_VALIDATE_OR_GOTO(frame->this->name, lock, out);
- dht_unlock_entrylk_wrapper (frame, &lock->ns.directory_ns);
- dht_unlock_inodelk_wrapper (frame, &lock->ns.parent_layout);
+ dht_unlock_entrylk_wrapper(frame, &lock->ns.directory_ns);
+ dht_unlock_inodelk_wrapper(frame, &lock->ns.parent_layout);
out:
- return;
+ return;
}
static int32_t
-dht_protect_namespace_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+dht_protect_namespace_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
+ dht_local_t *local = NULL;
- local = frame->local;
- if (op_ret != 0)
- dht_unlock_inodelk_wrapper (frame,
- &local->current->ns.parent_layout);
+ local = frame->local;
+ if (op_ret != 0)
+ dht_unlock_inodelk_wrapper(frame, &local->current->ns.parent_layout);
- local->current->ns.ns_cbk (frame, cookie, this, op_ret, op_errno,
- xdata);
- return 0;
+ local->current->ns.ns_cbk(frame, cookie, this, op_ret, op_errno, xdata);
+ return 0;
}
int32_t
-dht_blocking_entrylk_after_inodelk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
+dht_blocking_entrylk_after_inodelk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int ret = -1;
- loc_t *loc = NULL;
- dht_lock_t **lk_array = NULL;
- char pgfid[GF_UUID_BUF_SIZE] = {0};
- int count = 0;
- dht_elock_wrap_t *entrylk = NULL;
-
- local = frame->local;
- entrylk = &local->current->ns.directory_ns;
-
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = op_errno;
- goto err;
- }
-
- loc = &entrylk->locks[0]->loc;
- gf_uuid_unparse (loc->gfid, pgfid);
-
- local->op_ret = 0;
- lk_array = entrylk->locks;
- count = entrylk->lk_count;
-
- ret = dht_blocking_entrylk (frame, lk_array, count,
- dht_protect_namespace_cbk);
-
- if (ret < 0) {
- local->op_ret = -1;
- local->op_errno = EIO;
- gf_msg (this->name, GF_LOG_WARNING, local->op_errno,
- DHT_MSG_ENTRYLK_ERROR,
- "%s (%s/%s): "
- "dht_blocking_entrylk failed after taking inodelk",
- gf_fop_list[local->fop], pgfid,
- entrylk->locks[0]->basename);
- goto err;
- }
-
- return 0;
+ dht_local_t *local = NULL;
+ int ret = -1;
+ loc_t *loc = NULL;
+ dht_lock_t **lk_array = NULL;
+ char pgfid[GF_UUID_BUF_SIZE] = {0};
+ int count = 0;
+ dht_elock_wrap_t *entrylk = NULL;
+
+ local = frame->local;
+ entrylk = &local->current->ns.directory_ns;
+
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ goto err;
+ }
+
+ loc = &entrylk->locks[0]->loc;
+ gf_uuid_unparse(loc->gfid, pgfid);
+
+ local->op_ret = 0;
+ lk_array = entrylk->locks;
+ count = entrylk->lk_count;
+
+ ret = dht_blocking_entrylk(frame, lk_array, count,
+ dht_protect_namespace_cbk);
+
+ if (ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = EIO;
+ gf_smsg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_ENTRYLK_FAILED_AFT_INODELK, "fop=%s",
+ gf_fop_list[local->fop], "pgfid=%s", pgfid, "basename=%s",
+ entrylk->locks[0]->basename, NULL);
+ goto err;
+ }
+
+ return 0;
err:
- if (lk_array != NULL) {
- dht_lock_array_free (lk_array, count);
- GF_FREE (lk_array);
- entrylk->locks = NULL;
- entrylk->lk_count = 0;
- }
-
- /* Unlock inodelk. No harm calling unlock twice */
- dht_unlock_inodelk_wrapper (frame, &local->current->ns.parent_layout);
- /* Call ns_cbk. It will take care of unwinding */
- local->current->ns.ns_cbk (frame, NULL, this, local->op_ret,
- local->op_errno, NULL);
- return 0;
+ if (lk_array != NULL) {
+ dht_lock_array_free(lk_array, count);
+ GF_FREE(lk_array);
+ entrylk->locks = NULL;
+ entrylk->lk_count = 0;
+ }
+
+ /* Unlock inodelk. No harm calling unlock twice */
+ dht_unlock_inodelk_wrapper(frame, &local->current->ns.parent_layout);
+ /* Call ns_cbk. It will take care of unwinding */
+ local->current->ns.ns_cbk(frame, NULL, this, local->op_ret, local->op_errno,
+ NULL);
+ return 0;
}
/* Given the loc and the subvol, this routine takes the inodelk on
@@ -1248,137 +1263,130 @@ err:
* and then entrylk serially.
*/
int
-dht_protect_namespace (call_frame_t *frame, loc_t *loc,
- xlator_t *subvol,
- struct dht_namespace *ns,
- fop_entrylk_cbk_t ns_cbk)
+dht_protect_namespace(call_frame_t *frame, loc_t *loc, xlator_t *subvol,
+ struct dht_namespace *ns, fop_entrylk_cbk_t ns_cbk)
{
- dht_ilock_wrap_t *inodelk = NULL;
- dht_elock_wrap_t *entrylk = NULL;
- dht_lock_t **lk_array = NULL;
- dht_local_t *local = NULL;
- xlator_t *this = NULL;
- loc_t parent = {0,};
- int ret = -1;
- char pgfid[GF_UUID_BUF_SIZE] = {0};
- int32_t op_errno = 0;
- int count = 1;
-
- GF_VALIDATE_OR_GOTO ("dht-locks", frame, out);
- GF_VALIDATE_OR_GOTO (frame->this->name, loc, out);
- GF_VALIDATE_OR_GOTO (frame->this->name, loc->parent, out);
- GF_VALIDATE_OR_GOTO (frame->this->name, subvol, out);
-
- local = frame->local;
- this = frame->this;
-
- inodelk = &ns->parent_layout;
- entrylk = &ns->directory_ns;
-
- /* Initialize entrylk_cbk and parent loc */
- ns->ns_cbk = ns_cbk;
-
- ret = dht_build_parent_loc (this, &parent, loc, &op_errno);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- DHT_MSG_LOC_FAILED, "gfid:%s (name:%s) (path: %s): "
- "parent loc build failed", loc->gfid, loc->name,
- loc->path);
- goto out;
- }
- gf_uuid_unparse (parent.gfid, pgfid);
-
- /* Alloc inodelk */
- inodelk->locks = GF_CALLOC (count, sizeof (*lk_array),
- gf_common_mt_pointer);
- if (inodelk->locks == NULL) {
- local->op_errno = ENOMEM;
- gf_msg (this->name, GF_LOG_WARNING, local->op_errno,
- DHT_MSG_NO_MEMORY,
- "%s (%s/%s) (path: %s): "
- "calloc failure",
- gf_fop_list[local->fop], pgfid, loc->name, loc->path);
- goto out;
- }
-
- inodelk->locks[0] = dht_lock_new (this, subvol, &parent, F_RDLCK,
- DHT_LAYOUT_HEAL_DOMAIN, NULL,
- FAIL_ON_ANY_ERROR);
- if (inodelk->locks[0] == NULL) {
- local->op_errno = ENOMEM;
- gf_msg (this->name, GF_LOG_WARNING, local->op_errno,
- DHT_MSG_NO_MEMORY,
- "%s (%s/%s) (path: %s): "
- "inodelk: lock allocation failed",
- gf_fop_list[local->fop], pgfid, loc->name, loc->path);
- goto err;
- }
- inodelk->lk_count = count;
-
- /* Allock entrylk */
- entrylk->locks = GF_CALLOC (count, sizeof (*lk_array),
- gf_common_mt_pointer);
- if (entrylk->locks == NULL) {
- local->op_errno = ENOMEM;
- gf_msg (this->name, GF_LOG_WARNING, local->op_errno,
- DHT_MSG_NO_MEMORY,
- "%s (%s/%s) (path: %s): "
- "entrylk: calloc failure",
- gf_fop_list[local->fop], pgfid, loc->name, loc->path);
-
- goto err;
- }
-
- entrylk->locks[0] = dht_lock_new (this, subvol, &parent, F_WRLCK,
- DHT_ENTRY_SYNC_DOMAIN, loc->name,
- FAIL_ON_ANY_ERROR);
- if (entrylk->locks[0] == NULL) {
- local->op_errno = ENOMEM;
- gf_msg (this->name, GF_LOG_WARNING, local->op_errno,
- DHT_MSG_NO_MEMORY,
- "%s (%s/%s) (path: %s): "
- "entrylk: lock allocation failed",
- gf_fop_list[local->fop], pgfid, loc->name, loc->path);
-
- goto err;
- }
- entrylk->lk_count = count;
-
- /* Take read inodelk on parent. If it is successful, take write entrylk
- * on name in cbk.
- */
- lk_array = inodelk->locks;
- ret = dht_blocking_inodelk (frame, lk_array, count,
- dht_blocking_entrylk_after_inodelk);
- if (ret < 0) {
- local->op_errno = EIO;
- gf_msg (this->name, GF_LOG_WARNING, local->op_errno,
- DHT_MSG_INODELK_ERROR,
- "%s (%s/%s) (path: %s): "
- "dht_blocking_inodelk failed",
- gf_fop_list[local->fop], pgfid, loc->name, loc->path);
- goto err;
- }
-
- loc_wipe (&parent);
-
- return 0;
+ dht_ilock_wrap_t *inodelk = NULL;
+ dht_elock_wrap_t *entrylk = NULL;
+ dht_lock_t **lk_array = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *this = NULL;
+ loc_t parent = {
+ 0,
+ };
+ int ret = -1;
+ char pgfid[GF_UUID_BUF_SIZE] = {0};
+ int32_t op_errno = 0;
+ int count = 1;
+
+ GF_VALIDATE_OR_GOTO("dht-locks", frame, out);
+ GF_VALIDATE_OR_GOTO(frame->this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(frame->this->name, loc->parent, out);
+ GF_VALIDATE_OR_GOTO(frame->this->name, subvol, out);
+
+ local = frame->local;
+ this = frame->this;
+
+ inodelk = &ns->parent_layout;
+ entrylk = &ns->directory_ns;
+
+ /* Initialize entrylk_cbk and parent loc */
+ ns->ns_cbk = ns_cbk;
+
+ ret = dht_build_parent_loc(this, &parent, loc, &op_errno);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno, DHT_MSG_LOC_FAILED,
+ "gfid=%s", loc->gfid, "name=%s", loc->name, "path=%s",
+ loc->path, NULL);
+ goto out;
+ }
+ gf_uuid_unparse(parent.gfid, pgfid);
+
+ /* Alloc inodelk */
+ inodelk->locks = GF_CALLOC(count, sizeof(*lk_array), gf_common_mt_pointer);
+ if (inodelk->locks == NULL) {
+ local->op_errno = ENOMEM;
+ gf_smsg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_CALLOC_FAILED, "fop=%s", gf_fop_list[local->fop],
+ "pgfid=%s", pgfid, "name=%s", loc->name, "path=%s", loc->path,
+ NULL);
+ goto out;
+ }
+
+ inodelk->locks[0] = dht_lock_new(this, subvol, &parent, F_RDLCK,
+ DHT_LAYOUT_HEAL_DOMAIN, NULL,
+ FAIL_ON_ANY_ERROR);
+ if (inodelk->locks[0] == NULL) {
+ local->op_errno = ENOMEM;
+ gf_smsg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_LOCK_ALLOC_FAILED, "inodelk-fop=%s",
+ gf_fop_list[local->fop], "pgfid=%s", pgfid, "name=%s",
+ loc->name, "path=%s", loc->path, NULL);
+ goto err;
+ }
+ inodelk->lk_count = count;
+
+ /* Allock entrylk */
+ entrylk->locks = GF_CALLOC(count, sizeof(*lk_array), gf_common_mt_pointer);
+ if (entrylk->locks == NULL) {
+ local->op_errno = ENOMEM;
+ gf_smsg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_CALLOC_FAILED, "entrylk-fop=%s",
+ gf_fop_list[local->fop], "pgfid=%s", pgfid, "name=%s",
+ loc->name, "path=%s", loc->path, NULL);
+
+ goto err;
+ }
+
+ entrylk->locks[0] = dht_lock_new(this, subvol, &parent, F_WRLCK,
+ DHT_ENTRY_SYNC_DOMAIN, loc->name,
+ FAIL_ON_ANY_ERROR);
+ if (entrylk->locks[0] == NULL) {
+ local->op_errno = ENOMEM;
+ gf_smsg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_LOCK_ALLOC_FAILED, "entrylk-fop=%s",
+ gf_fop_list[local->fop], "pgfid=%s", pgfid, "name=%s",
+ loc->name, "path=%s", loc->path, NULL);
+
+ goto err;
+ }
+ entrylk->lk_count = count;
+
+ /* Take read inodelk on parent. If it is successful, take write entrylk
+ * on name in cbk.
+ */
+ lk_array = inodelk->locks;
+ ret = dht_blocking_inodelk(frame, lk_array, count,
+ dht_blocking_entrylk_after_inodelk);
+ if (ret < 0) {
+ local->op_errno = EIO;
+ gf_smsg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_BLOCK_INODELK_FAILED, "fop=%s", gf_fop_list[local->fop],
+ "pgfid=%s", pgfid, "name=%s", loc->name, "path=%s", loc->path,
+ NULL);
+
+ goto err;
+ }
+
+ loc_wipe(&parent);
+
+ return 0;
err:
- if (entrylk->locks != NULL) {
- dht_lock_array_free (entrylk->locks, count);
- GF_FREE (entrylk->locks);
- entrylk->locks = NULL;
- entrylk->lk_count = 0;
- }
+ if (entrylk->locks != NULL) {
+ dht_lock_array_free(entrylk->locks, count);
+ GF_FREE(entrylk->locks);
+ entrylk->locks = NULL;
+ entrylk->lk_count = 0;
+ }
- if (inodelk->locks != NULL) {
- dht_lock_array_free (inodelk->locks, count);
- GF_FREE (inodelk->locks);
- inodelk->locks = NULL;
- inodelk->lk_count = 0;
- }
+ if (inodelk->locks != NULL) {
+ dht_lock_array_free(inodelk->locks, count);
+ GF_FREE(inodelk->locks);
+ inodelk->locks = NULL;
+ inodelk->lk_count = 0;
+ }
- loc_wipe (&parent);
+ loc_wipe(&parent);
out:
- return -1;
+ return -1;
}
diff --git a/xlators/cluster/dht/src/dht-lock.h b/xlators/cluster/dht/src/dht-lock.h
index f36bf3fd685..6485c03fb6e 100644
--- a/xlators/cluster/dht/src/dht-lock.h
+++ b/xlators/cluster/dht/src/dht-lock.h
@@ -11,36 +11,35 @@
#ifndef _DHT_LOCK_H
#define _DHT_LOCK_H
-#include "xlator.h"
#include "dht-common.h"
void
-dht_lock_array_free (dht_lock_t **lk_array, int count);
+dht_lock_array_free(dht_lock_t **lk_array, int count);
int32_t
-dht_lock_count (dht_lock_t **lk_array, int lk_count);
+dht_lock_count(dht_lock_t **lk_array, int lk_count);
dht_lock_t *
-dht_lock_new (xlator_t *this, xlator_t *xl, loc_t *loc, short type,
- const char *domain, const char *basename,
- dht_reaction_type_t do_on_failure);
+dht_lock_new(xlator_t *this, xlator_t *xl, loc_t *loc, short type,
+ const char *domain, const char *basename,
+ dht_reaction_type_t do_on_failure);
int32_t
-dht_unlock_entrylk_wrapper (call_frame_t *, dht_elock_wrap_t *);
+dht_unlock_entrylk_wrapper(call_frame_t *, dht_elock_wrap_t *);
void
-dht_blocking_entrylk_rec (call_frame_t *frame, int i);
+dht_blocking_entrylk_rec(call_frame_t *frame, int i);
int
-dht_blocking_entrylk (call_frame_t *frame, dht_lock_t **lk_array,
- int lk_count, fop_inodelk_cbk_t entrylk_cbk);
+dht_blocking_entrylk(call_frame_t *frame, dht_lock_t **lk_array, int lk_count,
+ fop_inodelk_cbk_t entrylk_cbk);
int32_t
-dht_unlock_inodelk (call_frame_t *frame, dht_lock_t **lk_array, int lk_count,
- fop_inodelk_cbk_t inodelk_cbk);
+dht_unlock_inodelk(call_frame_t *frame, dht_lock_t **lk_array, int lk_count,
+ fop_inodelk_cbk_t inodelk_cbk);
int32_t
-dht_unlock_inodelk_wrapper (call_frame_t *, dht_ilock_wrap_t *);
+dht_unlock_inodelk_wrapper(call_frame_t *, dht_ilock_wrap_t *);
/* Acquire non-blocking inodelk on a list of xlators.
*
@@ -57,11 +56,11 @@ dht_unlock_inodelk_wrapper (call_frame_t *, dht_ilock_wrap_t *);
*/
int
-dht_nonblocking_inodelk (call_frame_t *frame, dht_lock_t **lk_array,
- int lk_count, fop_inodelk_cbk_t inodelk_cbk);
+dht_nonblocking_inodelk(call_frame_t *frame, dht_lock_t **lk_array,
+ int lk_count, fop_inodelk_cbk_t inodelk_cbk);
void
-dht_blocking_inodelk_rec (call_frame_t *frame, int i);
+dht_blocking_inodelk_rec(call_frame_t *frame, int i);
/* same as dht_nonblocking_inodelk, but issues sequential blocking locks on
* @lk_array directly. locks are issued on some order which remains same
@@ -69,25 +68,24 @@ dht_blocking_inodelk_rec (call_frame_t *frame, int i);
*/
int
-dht_blocking_inodelk (call_frame_t *frame, dht_lock_t **lk_array,
- int lk_count, fop_inodelk_cbk_t inodelk_cbk);
+dht_blocking_inodelk(call_frame_t *frame, dht_lock_t **lk_array, int lk_count,
+ fop_inodelk_cbk_t inodelk_cbk);
int32_t
-dht_blocking_entrylk_after_inodelk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xdata);
+dht_blocking_entrylk_after_inodelk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata);
int32_t
-dht_blocking_entrylk_after_inodelk_rename (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xdata);
+dht_blocking_entrylk_after_inodelk_rename(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata);
void
-dht_unlock_namespace (call_frame_t *, dht_dir_transaction_t *);
+dht_unlock_namespace(call_frame_t *, dht_dir_transaction_t *);
int
-dht_protect_namespace (call_frame_t *frame, loc_t *loc, xlator_t *subvol,
- struct dht_namespace *ns,
- fop_entrylk_cbk_t ns_cbk);
+dht_protect_namespace(call_frame_t *frame, loc_t *loc, xlator_t *subvol,
+ struct dht_namespace *ns, fop_entrylk_cbk_t ns_cbk);
-#endif /* _DHT_LOCK_H */
+#endif /* _DHT_LOCK_H */
diff --git a/xlators/cluster/dht/src/dht-mem-types.h b/xlators/cluster/dht/src/dht-mem-types.h
index 19cccef537b..e3c4471334a 100644
--- a/xlators/cluster/dht/src/dht-mem-types.h
+++ b/xlators/cluster/dht/src/dht-mem-types.h
@@ -8,38 +8,31 @@
cases as published by the Free Software Foundation.
*/
-
#ifndef __DHT_MEM_TYPES_H__
#define __DHT_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_dht_mem_types_ {
- gf_dht_mt_dht_du_t = gf_common_mt_end + 1,
- gf_dht_mt_dht_conf_t,
- gf_dht_mt_char,
- gf_dht_mt_int32_t,
- gf_dht_mt_xlator_t,
- gf_dht_mt_dht_layout_t,
- gf_switch_mt_dht_conf_t,
- gf_switch_mt_dht_du_t,
- gf_switch_mt_switch_sched_array,
- gf_switch_mt_switch_struct,
- gf_dht_mt_subvol_time,
- gf_dht_mt_loc_t,
- gf_defrag_info_mt,
- gf_dht_mt_inode_ctx_t,
- gf_dht_mt_ctx_stat_time_t,
- gf_dht_mt_dirent_t,
- gf_dht_mt_container_t,
- gf_dht_mt_octx_t,
- gf_dht_mt_miginfo_t,
- gf_tier_mt_bricklist_t,
- gf_tier_mt_ipc_ctr_params_t,
- gf_dht_mt_fd_ctx_t,
- gf_tier_mt_qfile_array_t,
- gf_dht_ret_cache_t,
- gf_dht_nodeuuids_t,
- gf_dht_mt_end
+ gf_dht_mt_dht_du_t = gf_common_mt_end + 1,
+ gf_dht_mt_dht_conf_t,
+ gf_dht_mt_char,
+ gf_dht_mt_int32_t,
+ gf_dht_mt_xlator_t,
+ gf_dht_mt_dht_layout_t,
+ gf_switch_mt_switch_sched_array,
+ gf_switch_mt_switch_struct,
+ gf_dht_mt_subvol_time,
+ gf_dht_mt_loc_t,
+ gf_defrag_info_mt,
+ gf_dht_mt_inode_ctx_t,
+ gf_dht_mt_dirent_t,
+ gf_dht_mt_container_t,
+ gf_dht_mt_octx_t,
+ gf_dht_mt_miginfo_t,
+ gf_dht_mt_fd_ctx_t,
+ gf_dht_ret_cache_t,
+ gf_dht_nodeuuids_t,
+ gf_dht_mt_end
};
#endif
diff --git a/xlators/cluster/dht/src/dht-messages.h b/xlators/cluster/dht/src/dht-messages.h
index bdea5a092b7..601f8dad78b 100644
--- a/xlators/cluster/dht/src/dht-messages.h
+++ b/xlators/cluster/dht/src/dht-messages.h
@@ -10,7 +10,7 @@
#ifndef _DHT_MESSAGES_H_
#define _DHT_MESSAGES_H_
-#include "glfs-message-id.h"
+#include <glusterfs/glfs-message-id.h>
/* To add new message IDs, append new identifiers at the end of the list.
*
@@ -22,136 +22,365 @@
* glfs-message-id.h.
*/
-GLFS_MSGID(DHT,
- DHT_MSG_CACHED_SUBVOL_GET_FAILED,
- DHT_MSG_CREATE_LINK_FAILED,
- DHT_MSG_DICT_SET_FAILED,
- DHT_MSG_DIR_ATTR_HEAL_FAILED,
- DHT_MSG_DIR_SELFHEAL_FAILED,
- DHT_MSG_DIR_SELFHEAL_XATTR_FAILED,
- DHT_MSG_FILE_ON_MULT_SUBVOL,
- DHT_MSG_FILE_TYPE_MISMATCH,
- DHT_MSG_GFID_MISMATCH,
- DHT_MSG_GFID_NULL,
- DHT_MSG_HASHED_SUBVOL_GET_FAILED,
- DHT_MSG_INIT_FAILED,
- DHT_MSG_INVALID_CONFIGURATION,
- DHT_MSG_INVALID_DISK_LAYOUT,
- DHT_MSG_INVALID_OPTION,
- DHT_MSG_LAYOUT_FIX_FAILED,
- DHT_MSG_LAYOUT_MERGE_FAILED,
- DHT_MSG_LAYOUT_MISMATCH,
- DHT_MSG_LAYOUT_NULL,
- DHT_MSG_MIGRATE_DATA_COMPLETE,
- DHT_MSG_MIGRATE_DATA_FAILED,
- DHT_MSG_MIGRATE_FILE_COMPLETE,
- DHT_MSG_MIGRATE_FILE_FAILED,
- DHT_MSG_NO_MEMORY,
- DHT_MSG_OPENDIR_FAILED,
- DHT_MSG_REBALANCE_FAILED,
- DHT_MSG_REBALANCE_START_FAILED,
- DHT_MSG_REBALANCE_STATUS,
- DHT_MSG_REBALANCE_STOPPED,
- DHT_MSG_RENAME_FAILED,
- DHT_MSG_SETATTR_FAILED,
- DHT_MSG_SUBVOL_INSUFF_INODES,
- DHT_MSG_SUBVOL_INSUFF_SPACE,
- DHT_MSG_UNLINK_FAILED,
- DHT_MSG_LAYOUT_SET_FAILED,
- DHT_MSG_LOG_FIXED_LAYOUT,
- DHT_MSG_LOG_TIER_ERROR,
- DHT_MSG_LOG_TIER_STATUS,
- DHT_MSG_GET_XATTR_FAILED,
- DHT_MSG_FILE_LOOKUP_FAILED,
- DHT_MSG_OPEN_FD_FAILED,
- DHT_MSG_SET_INODE_CTX_FAILED,
- DHT_MSG_UNLOCKING_FAILED,
- DHT_MSG_DISK_LAYOUT_NULL,
- DHT_MSG_SUBVOL_INFO,
- DHT_MSG_CHUNK_SIZE_INFO,
- DHT_MSG_LAYOUT_FORM_FAILED,
- DHT_MSG_SUBVOL_ERROR,
- DHT_MSG_LAYOUT_SORT_FAILED,
- DHT_MSG_REGEX_INFO,
- DHT_MSG_FOPEN_FAILED,
- DHT_MSG_SET_HOSTNAME_FAILED,
- DHT_MSG_BRICK_ERROR,
- DHT_MSG_SYNCOP_FAILED,
- DHT_MSG_MIGRATE_INFO,
- DHT_MSG_SOCKET_ERROR,
- DHT_MSG_CREATE_FD_FAILED,
- DHT_MSG_READDIR_ERROR,
- DHT_MSG_CHILD_LOC_BUILD_FAILED,
- DHT_MSG_SET_SWITCH_PATTERN_ERROR,
- DHT_MSG_COMPUTE_HASH_FAILED,
- DHT_MSG_FIND_LAYOUT_ANOMALIES_ERROR,
- DHT_MSG_ANOMALIES_INFO,
- DHT_MSG_LAYOUT_INFO,
- DHT_MSG_INODE_LK_ERROR,
- DHT_MSG_RENAME_INFO,
- DHT_MSG_DATA_NULL,
- DHT_MSG_AGGREGATE_QUOTA_XATTR_FAILED,
- DHT_MSG_UNLINK_LOOKUP_INFO,
- DHT_MSG_LINK_FILE_LOOKUP_INFO,
- DHT_MSG_OPERATION_NOT_SUP,
- DHT_MSG_NOT_LINK_FILE_ERROR,
- DHT_MSG_CHILD_DOWN,
- DHT_MSG_UUID_PARSE_ERROR,
- DHT_MSG_GET_DISK_INFO_ERROR,
- DHT_MSG_INVALID_VALUE,
- DHT_MSG_SWITCH_PATTERN_INFO,
- DHT_MSG_SUBVOL_OP_FAILED,
- DHT_MSG_LAYOUT_PRESET_FAILED,
- DHT_MSG_INVALID_LINKFILE,
- DHT_MSG_FIX_LAYOUT_INFO,
- DHT_MSG_GET_HOSTNAME_FAILED,
- DHT_MSG_WRITE_FAILED,
- DHT_MSG_MIGRATE_HARDLINK_FILE_FAILED,
- DHT_MSG_FSYNC_FAILED,
- DHT_MSG_SUBVOL_DECOMMISSION_INFO,
- DHT_MSG_BRICK_QUERY_FAILED,
- DHT_MSG_SUBVOL_NO_LAYOUT_INFO,
- DHT_MSG_OPEN_FD_ON_DST_FAILED,
- DHT_MSG_SUBVOL_NOT_FOUND,
- DHT_MSG_FILE_LOOKUP_ON_DST_FAILED,
- DHT_MSG_DISK_LAYOUT_MISSING,
- DHT_MSG_DICT_GET_FAILED,
- DHT_MSG_REVALIDATE_CBK_INFO,
- DHT_MSG_UPGRADE_BRICKS,
- DHT_MSG_LK_ARRAY_INFO,
- DHT_MSG_RENAME_NOT_LOCAL,
- DHT_MSG_RECONFIGURE_INFO,
- DHT_MSG_INIT_LOCAL_SUBVOL_FAILED,
- DHT_MSG_SYS_CALL_GET_TIME_FAILED,
- DHT_MSG_NO_DISK_USAGE_STATUS,
- DHT_MSG_SUBVOL_DOWN_ERROR,
- DHT_MSG_REBAL_THROTTLE_INFO,
- DHT_MSG_COMMIT_HASH_INFO,
- DHT_MSG_REBAL_STRUCT_SET,
- DHT_MSG_HAS_MIGINFO,
- DHT_MSG_LOG_IPC_TIER_ERROR,
- DHT_MSG_TIER_PAUSED,
- DHT_MSG_TIER_RESUME,
- DHT_MSG_SETTLE_HASH_FAILED,
- DHT_MSG_DEFRAG_PROCESS_DIR_FAILED,
- DHT_MSG_FD_CTX_SET_FAILED,
- DHT_MSG_STALE_LOOKUP,
- DHT_MSG_PARENT_LAYOUT_CHANGED,
- DHT_MSG_LOCK_MIGRATION_FAILED,
- DHT_MSG_LOCK_INODE_UNREF_FAILED,
- DHT_MSG_ASPRINTF_FAILED,
- DHT_MSG_DIR_LOOKUP_FAILED,
- DHT_MSG_INODELK_FAILED,
- DHT_MSG_LOCK_FRAME_FAILED,
- DHT_MSG_LOCAL_LOCK_INIT_FAILED,
- DHT_MSG_ENTRYLK_ERROR,
- DHT_MSG_INODELK_ERROR,
- DHT_MSG_LOC_FAILED,
- DHT_MSG_UNKNOWN_FOP,
- DHT_MSG_MIGRATE_FILE_SKIPPED,
- DHT_MSG_DIR_XATTR_HEAL_FAILED,
- DHT_MSG_HASHED_SUBVOL_DOWN,
- DHT_MSG_NON_HASHED_SUBVOL_DOWN
-);
+GLFS_MSGID(
+ DHT, DHT_MSG_CACHED_SUBVOL_GET_FAILED, DHT_MSG_CREATE_LINK_FAILED,
+ DHT_MSG_DICT_SET_FAILED, DHT_MSG_DIR_ATTR_HEAL_FAILED,
+ DHT_MSG_DIR_SELFHEAL_FAILED, DHT_MSG_DIR_SELFHEAL_XATTR_FAILED,
+ DHT_MSG_FILE_ON_MULT_SUBVOL, DHT_MSG_FILE_TYPE_MISMATCH,
+ DHT_MSG_GFID_MISMATCH, DHT_MSG_GFID_NULL, DHT_MSG_HASHED_SUBVOL_GET_FAILED,
+ DHT_MSG_INIT_FAILED, DHT_MSG_INVALID_CONFIGURATION,
+ DHT_MSG_INVALID_DISK_LAYOUT, DHT_MSG_INVALID_OPTION,
+ DHT_MSG_LAYOUT_FIX_FAILED, DHT_MSG_LAYOUT_MERGE_FAILED,
+ DHT_MSG_LAYOUT_MISMATCH, DHT_MSG_LAYOUT_NULL, DHT_MSG_MIGRATE_DATA_COMPLETE,
+ DHT_MSG_MIGRATE_DATA_FAILED, DHT_MSG_MIGRATE_FILE_COMPLETE,
+ DHT_MSG_MIGRATE_FILE_FAILED, DHT_MSG_NO_MEMORY, DHT_MSG_OPENDIR_FAILED,
+ DHT_MSG_REBALANCE_FAILED, DHT_MSG_REBALANCE_START_FAILED,
+ DHT_MSG_REBALANCE_STATUS, DHT_MSG_REBALANCE_STOPPED, DHT_MSG_RENAME_FAILED,
+ DHT_MSG_SETATTR_FAILED, DHT_MSG_SUBVOL_INSUFF_INODES,
+ DHT_MSG_SUBVOL_INSUFF_SPACE, DHT_MSG_UNLINK_FAILED,
+ DHT_MSG_LAYOUT_SET_FAILED, DHT_MSG_LOG_FIXED_LAYOUT,
+ DHT_MSG_GET_XATTR_FAILED, DHT_MSG_FILE_LOOKUP_FAILED,
+ DHT_MSG_OPEN_FD_FAILED, DHT_MSG_SET_INODE_CTX_FAILED,
+ DHT_MSG_UNLOCKING_FAILED, DHT_MSG_DISK_LAYOUT_NULL, DHT_MSG_SUBVOL_INFO,
+ DHT_MSG_CHUNK_SIZE_INFO, DHT_MSG_LAYOUT_FORM_FAILED, DHT_MSG_SUBVOL_ERROR,
+ DHT_MSG_LAYOUT_SORT_FAILED, DHT_MSG_REGEX_INFO, DHT_MSG_FOPEN_FAILED,
+ DHT_MSG_SET_HOSTNAME_FAILED, DHT_MSG_BRICK_ERROR, DHT_MSG_SYNCOP_FAILED,
+ DHT_MSG_MIGRATE_INFO, DHT_MSG_SOCKET_ERROR, DHT_MSG_CREATE_FD_FAILED,
+ DHT_MSG_READDIR_ERROR, DHT_MSG_CHILD_LOC_BUILD_FAILED,
+ DHT_MSG_SET_SWITCH_PATTERN_ERROR, DHT_MSG_COMPUTE_HASH_FAILED,
+ DHT_MSG_FIND_LAYOUT_ANOMALIES_ERROR, DHT_MSG_ANOMALIES_INFO,
+ DHT_MSG_LAYOUT_INFO, DHT_MSG_INODE_LK_ERROR, DHT_MSG_RENAME_INFO,
+ DHT_MSG_DATA_NULL, DHT_MSG_AGGREGATE_QUOTA_XATTR_FAILED,
+ DHT_MSG_UNLINK_LOOKUP_INFO, DHT_MSG_LINK_FILE_LOOKUP_INFO,
+ DHT_MSG_OPERATION_NOT_SUP, DHT_MSG_NOT_LINK_FILE_ERROR, DHT_MSG_CHILD_DOWN,
+ DHT_MSG_UUID_PARSE_ERROR, DHT_MSG_GET_DISK_INFO_ERROR,
+ DHT_MSG_INVALID_VALUE, DHT_MSG_SWITCH_PATTERN_INFO,
+ DHT_MSG_SUBVOL_OP_FAILED, DHT_MSG_LAYOUT_PRESET_FAILED,
+ DHT_MSG_INVALID_LINKFILE, DHT_MSG_FIX_LAYOUT_INFO,
+ DHT_MSG_GET_HOSTNAME_FAILED, DHT_MSG_WRITE_FAILED,
+ DHT_MSG_MIGRATE_HARDLINK_FILE_FAILED, DHT_MSG_FSYNC_FAILED,
+ DHT_MSG_SUBVOL_DECOMMISSION_INFO, DHT_MSG_BRICK_QUERY_FAILED,
+ DHT_MSG_SUBVOL_NO_LAYOUT_INFO, DHT_MSG_OPEN_FD_ON_DST_FAILED,
+ DHT_MSG_SUBVOL_NOT_FOUND, DHT_MSG_FILE_LOOKUP_ON_DST_FAILED,
+ DHT_MSG_DISK_LAYOUT_MISSING, DHT_MSG_DICT_GET_FAILED,
+ DHT_MSG_REVALIDATE_CBK_INFO, DHT_MSG_UPGRADE_BRICKS, DHT_MSG_LK_ARRAY_INFO,
+ DHT_MSG_RENAME_NOT_LOCAL, DHT_MSG_RECONFIGURE_INFO,
+ DHT_MSG_INIT_LOCAL_SUBVOL_FAILED, DHT_MSG_SYS_CALL_GET_TIME_FAILED,
+ DHT_MSG_NO_DISK_USAGE_STATUS, DHT_MSG_SUBVOL_DOWN_ERROR,
+ DHT_MSG_REBAL_THROTTLE_INFO, DHT_MSG_COMMIT_HASH_INFO,
+ DHT_MSG_REBAL_STRUCT_SET, DHT_MSG_HAS_MIGINFO, DHT_MSG_SETTLE_HASH_FAILED,
+ DHT_MSG_DEFRAG_PROCESS_DIR_FAILED, DHT_MSG_FD_CTX_SET_FAILED,
+ DHT_MSG_STALE_LOOKUP, DHT_MSG_PARENT_LAYOUT_CHANGED,
+ DHT_MSG_LOCK_MIGRATION_FAILED, DHT_MSG_LOCK_INODE_UNREF_FAILED,
+ DHT_MSG_ASPRINTF_FAILED, DHT_MSG_DIR_LOOKUP_FAILED, DHT_MSG_INODELK_FAILED,
+ DHT_MSG_LOCK_FRAME_FAILED, DHT_MSG_LOCAL_LOCK_INIT_FAILED,
+ DHT_MSG_ENTRYLK_ERROR, DHT_MSG_INODELK_ERROR, DHT_MSG_LOC_FAILED,
+ DHT_MSG_UNKNOWN_FOP, DHT_MSG_MIGRATE_FILE_SKIPPED,
+ DHT_MSG_DIR_XATTR_HEAL_FAILED, DHT_MSG_HASHED_SUBVOL_DOWN,
+ DHT_MSG_NON_HASHED_SUBVOL_DOWN, DHT_MSG_SYNCTASK_CREATE_FAILED,
+ DHT_MSG_DIR_HEAL_ABORT, DHT_MSG_MIGRATE_SKIP, DHT_MSG_FD_CREATE_FAILED,
+ DHT_MSG_DICT_NEW_FAILED, DHT_MSG_FAILED_TO_OPEN, DHT_MSG_CREATE_FAILED,
+ DHT_MSG_FILE_NOT_EXIST, DHT_MSG_CHOWN_FAILED, DHT_MSG_FALLOCATE_FAILED,
+ DHT_MSG_FTRUNCATE_FAILED, DHT_MSG_STATFS_FAILED, DHT_MSG_WRITE_CROSS,
+ DHT_MSG_NEW_TARGET_FOUND, DHT_MSG_INSUFF_MEMORY, DHT_MSG_SET_XATTR_FAILED,
+ DHT_MSG_SET_MODE_FAILED, DHT_MSG_FILE_EXISTS_IN_DEST,
+ DHT_MSG_SYMLINK_FAILED, DHT_MSG_LINKFILE_DEL_FAILED, DHT_MSG_MKNOD_FAILED,
+ DHT_MSG_MIGRATE_CLEANUP_FAILED, DHT_MSG_LOCK_MIGRATE,
+ DHT_MSG_PARENT_BUILD_FAILED, DHT_MSG_HASHED_SUBVOL_NOT_FOUND,
+ DHT_MSG_ACQUIRE_ENTRYLK_FAILED, DHT_MSG_CREATE_DST_FAILED,
+ DHT_MSG_MIGRATION_EXIT, DHT_MSG_CHANGED_DST, DHT_MSG_TRACE_FAILED,
+ DHT_MSG_WRITE_LOCK_FAILED, DHT_MSG_GETACTIVELK_FAILED, DHT_MSG_STAT_FAILED,
+ DHT_MSG_UNLINK_PERFORM_FAILED, DHT_MSG_CLANUP_SOURCE_FILE_FAILED,
+ DHT_MSG_UNLOCK_FILE_FAILED, DHT_MSG_REMOVE_XATTR_FAILED,
+ DHT_MSG_DATA_MIGRATE_ABORT, DHT_MSG_DEFRAG_NULL, DHT_MSG_PARENT_NULL,
+ DHT_MSG_GFID_NOT_PRESENT, DHT_MSG_CHILD_LOC_FAILED,
+ DHT_MSG_SET_LOOKUP_FAILED, DHT_MSG_DIR_REMOVED, DHT_MSG_FIX_NOT_COMP,
+ DHT_MSG_SUBVOL_DETER_FAILED, DHT_MSG_LOCAL_SUBVOL, DHT_MSG_NODE_UUID,
+ DHT_MSG_SIZE_FILE, DHT_MSG_GET_DATA_SIZE_FAILED,
+ DHT_MSG_PTHREAD_JOIN_FAILED, DHT_MSG_COUNTER_THREAD_CREATE_FAILED,
+ DHT_MSG_MIGRATION_INIT_QUEUE_FAILED, DHT_MSG_PAUSED_TIMEOUT, DHT_MSG_WOKE,
+ DHT_MSG_ABORT_REBALANCE, DHT_MSG_CREATE_TASK_REBAL_FAILED,
+ DHT_MSG_REBAL_ESTIMATE_NOT_AVAIL, DHT_MSG_ADD_CHOICES_ERROR,
+ DHT_MSG_GET_CHOICES_ERROR, DHT_MSG_PREPARE_STATUS_ERROR,
+ DHT_MSG_SET_CHOICE_FAILED, DHT_MSG_SET_HASHED_SUBVOL_FAILED,
+ DHT_MSG_XATTR_HEAL_NOT_POSS, DHT_MSG_LINKTO_FILE_FAILED,
+ DHT_MSG_STALE_LINKFILE_DELETE, DHT_MSG_NO_SUBVOL_FOR_LINKTO,
+ DHT_MSG_SUBVOL_RETURNED, DHT_MSG_UNKNOWN_LOCAL_XSEL, DHT_MSG_GET_XATTR_ERR,
+ DHT_MSG_ALLOC_OR_FILL_FAILED, DHT_MSG_GET_REAL_NAME_FAILED,
+ DHT_MSG_COPY_UUID_FAILED, DHT_MSG_MDS_DETER_FAILED,
+ DHT_MSG_CREATE_REBAL_FAILED, DHT_MSG_LINK_LAYOUT_FAILED,
+ DHT_MSG_NO_SUBVOL_IN_LAYOUT, DHT_MSG_MEM_ALLOC_FAILED,
+ DHT_MSG_SET_IN_PARAMS_DICT_FAILED, DHT_MSG_LOC_COPY_FAILED,
+ DHT_MSG_PARENT_LOC_FAILED, DHT_MSG_CREATE_LOCK_FAILED,
+ DHT_MSG_PREV_ATTEMPT_FAILED, DHT_MSG_REFRESH_ATTEMPT,
+ DHT_MSG_ACQUIRE_LOCK_FAILED, DHT_MSG_CREATE_STUB_FAILED,
+ DHT_MSG_WIND_LOCK_REQ_FAILED, DHT_MSG_REFRESH_FAILED,
+ DHT_MSG_CACHED_SUBVOL_ERROR, DHT_MSG_NO_LINK_SUBVOL, DHT_MSG_SET_KEY_FAILED,
+ DHT_MSG_REMOVE_LINKTO_FAILED, DHT_MSG_LAYOUT_DICT_SET_FAILED,
+ DHT_MSG_XATTR_DICT_NULL, DHT_MSG_DUMMY_ALLOC_FAILED, DHT_MSG_DICT_IS_NULL,
+ DHT_MSG_LINK_INODE_FAILED, DHT_MSG_SELFHEAL_FAILED, DHT_MSG_NO_MDS_SUBVOL,
+ DHT_MSG_LIST_XATTRS_FAILED, DHT_MSG_RESET_INTER_XATTR_FAILED,
+ DHT_MSG_MDS_DOWN_UNABLE_TO_SET, DHT_MSG_WIND_UNLOCK_FAILED,
+ DHT_MSG_COMMIT_HASH_FAILED, DHT_MSG_UNLOCK_GFID_FAILED,
+ DHT_MSG_UNLOCK_FOLLOW_ENTRYLK, DHT_MSG_COPY_FRAME_FAILED,
+ DHT_MSG_UNLOCK_FOLLOW_LOCKS, DHT_MSG_ENTRYLK_FAILED_AFT_INODELK,
+ DHT_MSG_CALLOC_FAILED, DHT_MSG_LOCK_ALLOC_FAILED,
+ DHT_MSG_BLOCK_INODELK_FAILED,
+ DHT_MSG_LOCAL_LOCKS_STORE_FAILED_UNLOCKING_FOLLOWING_ENTRYLK,
+ DHT_MSG_ALLOC_FRAME_FAILED_NOT_UNLOCKING_FOLLOWING_ENTRYLKS,
+ DHT_MSG_DST_NULL_SET_FAILED);
+
+#define DHT_MSG_FD_CTX_SET_FAILED_STR "Failed to set fd ctx"
+#define DHT_MSG_INVALID_VALUE_STR "Different dst found in the fd ctx"
+#define DHT_MSG_UNKNOWN_FOP_STR "Unknown FOP on file"
+#define DHT_MSG_OPEN_FD_ON_DST_FAILED_STR "Failed to open the fd on file"
+#define DHT_MSG_SYNCTASK_CREATE_FAILED_STR "Failed to create synctask"
+#define DHT_MSG_ASPRINTF_FAILED_STR \
+ "asprintf failed while fetching subvol from the id"
+#define DHT_MSG_HAS_MIGINFO_STR "Found miginfo in the inode ctx"
+#define DHT_MSG_FILE_LOOKUP_FAILED_STR "failed to lookup the file"
+#define DHT_MSG_INVALID_LINKFILE_STR \
+ "linkto target is different from cached-subvol. treating as destination " \
+ "subvol"
+#define DHT_MSG_GFID_MISMATCH_STR "gfid different on the target file"
+#define DHT_MSG_GET_XATTR_FAILED_STR "failed to get 'linkto' xattr"
+#define DHT_MSG_SET_INODE_CTX_FAILED_STR "failed to set inode-ctx target file"
+#define DHT_MSG_DIR_SELFHEAL_FAILED_STR "Healing of path failed"
+#define DHT_MSG_DIR_HEAL_ABORT_STR \
+ "Failed to get path from subvol. Aborting directory healing"
+#define DHT_MSG_DIR_XATTR_HEAL_FAILED_STR "xattr heal failed for directory"
+#define DHT_MSG_LOCK_INODE_UNREF_FAILED_STR \
+ "Found a NULL inode. Failed to unref the inode"
+#define DHT_MSG_DICT_SET_FAILED_STR "Failed to set dictionary value"
+#define DHT_MSG_NOT_LINK_FILE_ERROR_STR "got non-linkfile"
+#define DHT_MSG_CREATE_LINK_FAILED_STR "failed to initialize linkfile data"
+#define DHT_MSG_UNLINK_FAILED_STR "Unlinking linkfile on subvolume failed"
+#define DHT_MSG_MIGRATE_FILE_FAILED_STR "Migrate file failed"
+#define DHT_MSG_NO_MEMORY_STR "could not allocate memory for dict"
+#define DHT_MSG_SUBVOL_ERROR_STR "Failed to get linkto subvol"
+#define DHT_MSG_MIGRATE_HARDLINK_FILE_FAILED_STR "link failed on subvol"
+#define DHT_MSG_MIGRATE_FILE_SKIPPED_STR "Migration skipped"
+#define DHT_MSG_FD_CREATE_FAILED_STR "fd create failed"
+#define DHT_MSG_DICT_NEW_FAILED_STR "dict_new failed"
+#define DHT_MSG_FAILED_TO_OPEN_STR "failed to open"
+#define DHT_MSG_CREATE_FAILED_STR "failed to create"
+#define DHT_MSG_FILE_NOT_EXIST_STR "file does not exist"
+#define DHT_MSG_CHOWN_FAILED_STR "chown failed"
+#define DHT_MSG_FALLOCATE_FAILED_STR "fallocate failed"
+#define DHT_MSG_FTRUNCATE_FAILED_STR "ftruncate failed"
+#define DHT_MSG_STATFS_FAILED_STR "failed to get statfs"
+#define DHT_MSG_WRITE_CROSS_STR \
+ "write will cross min-fre-disk for file on subvol. looking for new subvol"
+#define DHT_MSG_SUBVOL_INSUFF_SPACE_STR \
+ "Could not find any subvol with space accommodating the file. Cosider " \
+ "adding bricks"
+#define DHT_MSG_NEW_TARGET_FOUND_STR "New target found for file"
+#define DHT_MSG_INSUFF_MEMORY_STR "insufficient memory"
+#define DHT_MSG_SET_XATTR_FAILED_STR "failed to set xattr"
+#define DHT_MSG_SET_MODE_FAILED_STR "failed to set mode"
+#define DHT_MSG_FILE_EXISTS_IN_DEST_STR "file exists in destination"
+#define DHT_MSG_LINKFILE_DEL_FAILED_STR "failed to delete the linkfile"
+#define DHT_MSG_SYMLINK_FAILED_STR "symlink failed"
+#define DHT_MSG_MKNOD_FAILED_STR "mknod failed"
+#define DHT_MSG_SETATTR_FAILED_STR "failed to perform setattr"
+#define DHT_MSG_MIGRATE_CLEANUP_FAILED_STR \
+ "Migrate file cleanup failed: failed to fstat file"
+#define DHT_MSG_LOCK_MIGRATE_STR "locks will be migrated for file"
+#define DHT_MSG_PARENT_BUILD_FAILED_STR \
+ "failed to build parent loc, which is needed to acquire entrylk to " \
+ "synchronize with renames on this path. Skipping migration"
+#define DHT_MSG_HASHED_SUBVOL_NOT_FOUND_STR \
+ "cannot find hashed subvol which is needed to synchronize with renames " \
+ "on this path. Skipping migration"
+#define DHT_MSG_ACQUIRE_ENTRYLK_FAILED_STR "failed to acquire entrylk on subvol"
+#define DHT_MSG_CREATE_DST_FAILED_STR "create dst failed for file"
+#define DHT_MSG_MIGRATION_EXIT_STR "Exiting migration"
+#define DHT_MSG_CHANGED_DST_STR "destination changed fo file"
+#define DHT_MSG_TRACE_FAILED_STR "Trace failed"
+#define DHT_MSG_WRITE_LOCK_FAILED_STR "write lock failed"
+#define DHT_MSG_GETACTIVELK_FAILED_STR "getactivelk failed for file"
+#define DHT_MSG_STAT_FAILED_STR "failed to do a stat"
+#define DHT_MSG_UNLINK_PERFORM_FAILED_STR "failed to perform unlink"
+#define DHT_MSG_MIGRATE_FILE_COMPLETE_STR "completed migration"
+#define DHT_MSG_CLANUP_SOURCE_FILE_FAILED_STR "failed to cleanup source file"
+#define DHT_MSG_UNLOCK_FILE_FAILED_STR "failed to unlock file"
+#define DHT_MSG_REMOVE_XATTR_FAILED_STR "remove xattr failed"
+#define DHT_MSG_SOCKET_ERROR_STR "Failed to unlink listener socket"
+#define DHT_MSG_HASHED_SUBVOL_GET_FAILED_STR "Failed to get hashed subvolume"
+#define DHT_MSG_CACHED_SUBVOL_GET_FAILED_STR "Failed to get cached subvolume"
+#define DHT_MSG_MIGRATE_DATA_FAILED_STR "migrate-data failed"
+#define DHT_MSG_DEFRAG_NULL_STR "defrag is NULL"
+#define DHT_MSG_DATA_MIGRATE_ABORT_STR \
+ "Readdirp failed. Aborting data migration for dict"
+#define DHT_MSG_LAYOUT_FIX_FAILED_STR "fix layout failed"
+#define DHT_MSG_PARENT_NULL_STR "parent is NULL"
+#define DHT_MSG_GFID_NOT_PRESENT_STR "gfid not present"
+#define DHT_MSG_CHILD_LOC_FAILED_STR "Child loc build failed"
+#define DHT_MSG_SET_LOOKUP_FAILED_STR "Failed to set lookup"
+#define DHT_MSG_DIR_LOOKUP_FAILED_STR "lookup failed"
+#define DHT_MSG_DIR_REMOVED_STR "Dir renamed or removed. Skipping"
+#define DHT_MSG_READDIR_ERROR_STR "readdir failed, Aborting fix-layout"
+#define DHT_MSG_SETTLE_HASH_FAILED_STR "Settle hash failed"
+#define DHT_MSG_DEFRAG_PROCESS_DIR_FAILED_STR "gf_defrag_process_dir failed"
+#define DHT_MSG_FIX_NOT_COMP_STR \
+ "Unable to retrieve fixlayout xattr. Assume background fix layout not " \
+ "complete"
+#define DHT_MSG_SUBVOL_DETER_FAILED_STR \
+ "local subvolume determination failed with error"
+#define DHT_MSG_LOCAL_SUBVOL_STR "local subvol"
+#define DHT_MSG_NODE_UUID_STR "node uuid"
+#define DHT_MSG_SIZE_FILE_STR "Total size files"
+#define DHT_MSG_GET_DATA_SIZE_FAILED_STR \
+ "Failed to get the total data size. Unable to estimate time to complete " \
+ "rebalance"
+#define DHT_MSG_PTHREAD_JOIN_FAILED_STR \
+ "file_counter_thread: pthread_join failed"
+#define DHT_MSG_COUNTER_THREAD_CREATE_FAILED_STR \
+ "Failed to create the file counter thread"
+#define DHT_MSG_MIGRATION_INIT_QUEUE_FAILED_STR \
+ "Failed to initialise migration queue"
+#define DHT_MSG_REBALANCE_STOPPED_STR "Received stop command on rebalance"
+#define DHT_MSG_PAUSED_TIMEOUT_STR "Request pause timer timeout"
+#define DHT_MSG_WOKE_STR "woken"
+#define DHT_MSG_ABORT_REBALANCE_STR "Aborting rebalance"
+#define DHT_MSG_REBALANCE_START_FAILED_STR \
+ "Failed to start rebalance: look up on / failed"
+#define DHT_MSG_CREATE_TASK_REBAL_FAILED_STR \
+ "Could not create task for rebalance"
+#define DHT_MSG_REBAL_ESTIMATE_NOT_AVAIL_STR \
+ "Rebalance estimates will not be available"
+#define DHT_MSG_REBALANCE_STATUS_STR "Rebalance status"
+#define DHT_MSG_DATA_NULL_STR "data value is NULL"
+#define DHT_MSG_ADD_CHOICES_ERROR_STR "Error to add choices in buffer"
+#define DHT_MSG_GET_CHOICES_ERROR_STR "Error to get choices"
+#define DHT_MSG_PREPARE_STATUS_ERROR_STR "Error to prepare status"
+#define DHT_MSG_SET_CHOICE_FAILED_STR "Failed to set full choice"
+#define DHT_MSG_AGGREGATE_QUOTA_XATTR_FAILED_STR \
+ "Failed to aggregate quota xattr"
+#define DHT_MSG_FILE_TYPE_MISMATCH_STR \
+ "path exists as a file on one subvolume and directory on another. Please " \
+ "fix it manually"
+#define DHT_MSG_LAYOUT_SET_FAILED_STR "failed to set layout for subvolume"
+#define DHT_MSG_LAYOUT_MERGE_FAILED_STR "failed to merge layouts for subvolume"
+#define DHT_MSG_SET_HASHED_SUBVOL_FAILED_STR "Failed to set hashed subvolume"
+#define DHT_MSG_XATTR_HEAL_NOT_POSS_STR \
+ "No gfid exists for path. so healing xattr is not possible"
+#define DHT_MSG_REVALIDATE_CBK_INFO_STR "Revalidate: subvolume returned -1"
+#define DHT_MSG_LAYOUT_MISMATCH_STR "Mismatching layouts"
+#define DHT_MSG_UNLINK_LOOKUP_INFO_STR "lookup_unlink retuened"
+#define DHT_MSG_LINKTO_FILE_FAILED_STR \
+ "Could not unlink the linkto file as either fd is open and/or linkto " \
+ "xattr is set"
+#define DHT_MSG_LAYOUT_PRESET_FAILED_STR \
+ "Could not set pre-set layout for subvolume"
+#define DHT_MSG_FILE_ON_MULT_SUBVOL_STR \
+ "multiple subvolumes have file (preferably rename the file in the " \
+ "backend, and do a fresh lookup"
+#define DHT_MSG_STALE_LINKFILE_DELETE_STR \
+ "attempting deletion of stale linkfile"
+#define DHT_MSG_LINK_FILE_LOOKUP_INFO_STR "Lookup on following linkfile"
+#define DHT_MSG_NO_SUBVOL_FOR_LINKTO_STR "No link subvolume for linkto"
+#define DHT_MSG_SUBVOL_RETURNED_STR "Subvolume returned -1"
+#define DHT_MSG_UNKNOWN_LOCAL_XSEL_STR "Unknown local->xsel"
+#define DHT_MSG_DICT_GET_FAILED_STR "Failed to get"
+#define DHT_MSG_UUID_PARSE_ERROR_STR "Failed to parse uuid"
+#define DHT_MSG_GET_XATTR_ERR_STR "getxattr err for dir"
+#define DHT_MSG_ALLOC_OR_FILL_FAILED_STR "alloc or fill failed"
+#define DHT_MSG_UPGRADE_BRICKS_STR \
+ "At least one of the bricks does not support this operation. Please " \
+ "upgrade all bricks"
+#define DHT_MSG_GET_REAL_NAME_FAILED_STR "Failed to get real filename"
+#define DHT_MSG_LAYOUT_NULL_STR "Layout is NULL"
+#define DHT_MSG_COPY_UUID_FAILED_STR "Failed to copy node uuid key"
+#define DHT_MSG_MDS_DETER_FAILED_STR \
+ "Cannot determine MDS, fetching xattr randomly from a subvol"
+#define DHT_MSG_HASHED_SUBVOL_DOWN_STR \
+ "MDS is down for path, so fetching xattr randomly from subvol"
+#define DHT_MSG_CREATE_REBAL_FAILED_STR \
+ "failed to create a new rebalance synctask"
+#define DHT_MSG_FIX_LAYOUT_INFO_STR "fixing the layout"
+#define DHT_MSG_OPERATION_NOT_SUP_STR "wrong directory-spread-count value"
+#define DHT_MSG_LINK_LAYOUT_FAILED_STR "failed to link the layout in inode"
+#define DHT_MSG_NO_SUBVOL_IN_LAYOUT_STR "no subvolume in layout for path"
+#define DHT_MSG_INODE_LK_ERROR_STR "mknod lock failed for file"
+#define DHT_MSG_MEM_ALLOC_FAILED_STR "mem allocation failed"
+#define DHT_MSG_PARENT_LAYOUT_CHANGED_STR \
+ "extracting in-memory layout of parent failed"
+#define DHT_MSG_SET_IN_PARAMS_DICT_FAILED_STR \
+ "setting in params dictionary failed"
+#define DHT_MSG_LOC_COPY_FAILED_STR "loc_copy failed"
+#define DHT_MSG_LOC_FAILED_STR "parent loc build failed"
+#define DHT_MSG_PARENT_LOC_FAILED_STR "locking parent failed"
+#define DHT_MSG_CREATE_LOCK_FAILED_STR "Create lock failed"
+#define DHT_MSG_PREV_ATTEMPT_FAILED_STR \
+ "mkdir loop detected. parent layout didn't change even though previous " \
+ "attempt of mkdir failed because of in-memory layout not matching with " \
+ "that on disk."
+#define DHT_MSG_REFRESH_ATTEMPT_STR \
+ "mkdir parent layout changed. Attempting a refresh and then a retry"
+#define DHT_MSG_ACQUIRE_LOCK_FAILED_STR \
+ "Acquiring lock on parent to guard against layout-change failed"
+#define DHT_MSG_CREATE_STUB_FAILED_STR "creating stub failed"
+#define DHT_MSG_WIND_LOCK_REQ_FAILED_STR \
+ "cannot wind lock request to guard parent layout"
+#define DHT_MSG_REFRESH_FAILED_STR "refreshing parent layout failed."
+#define DHT_MSG_CACHED_SUBVOL_ERROR_STR "On cached subvol"
+#define DHT_MSG_NO_LINK_SUBVOL_STR "Linkfile does not have link subvolume"
+#define DHT_MSG_SET_KEY_FAILED_STR "failed to set key"
+#define DHT_MSG_CHILD_DOWN_STR "Received CHILD_DOWN. Exiting"
+#define DHT_MSG_LOG_FIXED_LAYOUT_STR "log layout fixed"
+#define DHT_MSG_REBAL_STRUCT_SET_STR "local->rebalance already set"
+#define DHT_MSG_REMOVE_LINKTO_FAILED_STR "Removal of linkto failed at subvol"
+#define DHT_MSG_LAYOUT_DICT_SET_FAILED_STR "dht layout dict set failed"
+#define DHT_MSG_SUBVOL_INFO_STR "creating subvolume"
+#define DHT_MSG_COMPUTE_HASH_FAILED_STR "hash computation failed"
+#define DHT_MSG_INVALID_DISK_LAYOUT_STR \
+ "Invalid disk layout: Catastrophic error layout with unknown type found"
+#define DHT_MSG_LAYOUT_SORT_FAILED_STR "layout sort failed"
+#define DHT_MSG_ANOMALIES_INFO_STR "Found anomalies"
+#define DHT_MSG_XATTR_DICT_NULL_STR "xattr dictionary is NULL"
+#define DHT_MSG_DISK_LAYOUT_MISSING_STR "Disk layout missing"
+#define DHT_MSG_LAYOUT_INFO_STR "layout info"
+#define DHT_MSG_SUBVOL_NO_LAYOUT_INFO_STR "no pre-set layout for subvol"
+#define DHT_MSG_SELFHEAL_XATTR_FAILED_STR "layout setxattr failed"
+#define DHT_MSG_DIR_SELFHEAL_XATTR_FAILED_STR "Directory self heal xattr failed"
+#define DHT_MSG_DUMMY_ALLOC_FAILED_STR "failed to allocate dummy layout"
+#define DHT_MSG_DICT_IS_NULL_STR \
+ "dict is NULL, need to make sure gfids are same"
+#define DHT_MSG_ENTRYLK_ERROR_STR "acquiring entrylk after inodelk failed"
+#define DHT_MSG_NO_DISK_USAGE_STATUS_STR "no du stats"
+#define DHT_MSG_LINK_INODE_FAILED_STR "linking inode failed"
+#define DHT_MSG_SELFHEAL_FAILED_STR "Directory selfheal failed"
+#define DHT_MSG_NO_MDS_SUBVOL_STR "No mds subvol"
+#define DHT_MSG_LIST_XATTRS_FAILED_STR "failed to list xattrs"
+#define DHT_MSG_RESET_INTER_XATTR_FAILED_STR "Failed to reset internal xattr"
+#define DHT_MSG_MDS_DOWN_UNABLE_TO_SET_STR \
+ "mds subvol is down, unable to set xattr"
+#define DHT_MSG_DIR_ATTR_HEAL_FAILED_STR \
+ "Directory attr heal failed. Failed to set uid/gid"
+#define DHT_MSG_WIND_UNLOCK_FAILED_STR \
+ "Winding unlock failed: stale locks left on brick"
+#define DHT_MSG_COMMIT_HASH_FAILED_STR "Directory commit hash updaten failed"
+#define DHT_MSG_LK_ARRAY_INFO_STR "lk info"
+#define DHT_MSG_UNLOCK_GFID_FAILED_STR \
+ "unlock failed on gfid: stale lock might be left"
+#define DHT_MSG_UNLOCKING_FAILED_STR "unlocking failed"
+#define DHT_MSG_UNLOCK_FOLLOW_ENTRYLK_STR "not unlocking following entrylks"
+#define DHT_MSG_COPY_FRAME_FAILED_STR "copy frame failed"
+#define DHT_MSG_UNLOCK_FOLLOW_LOCKS_STR "not unlocking following locks"
+#define DHT_MSG_INODELK_FAILED_STR "inodelk failed on subvol"
+#define DHT_MSG_LOCK_FRAME_FAILED_STR "memory allocation failed for lock_frame"
+#define DHT_MSG_LOCAL_LOCK_INIT_FAILED_STR "dht_local_lock_init failed"
+#define DHT_MSG_ENTRYLK_FAILED_AFT_INODELK_STR \
+ "dht_blocking_entrylk failed after taking inodelk"
+#define DHT_MSG_BLOCK_INODELK_FAILED_STR "dht_blocking_inodelk failed"
+#define DHT_MSG_CALLOC_FAILED_STR "calloc failed"
+#define DHT_MSG_LOCK_ALLOC_FAILED_STR "lock allocation failed"
+#define DHT_MSG_ALLOC_FRAME_FAILED_NOT_UNLOCKING_FOLLOWING_ENTRYLKS_STR \
+ "cannot allocate a frame, not unlocking following entrylks"
+#define DHT_MSG_LOCAL_LOCKS_STORE_FAILED_UNLOCKING_FOLLOWING_ENTRYLK_STR \
+ "storing locks in local failed, not unlocking following entrylks"
+#define DHT_MSG_DST_NULL_SET_FAILED_STR \
+ "src or dst is NULL, Failed to set dictionary value"
#endif /* _DHT_MESSAGES_H_ */
diff --git a/xlators/cluster/dht/src/dht-rebalance.c b/xlators/cluster/dht/src/dht-rebalance.c
index e5ac9d0d653..8ba8082bd86 100644
--- a/xlators/cluster/dht/src/dht-rebalance.c
+++ b/xlators/cluster/dht/src/dht-rebalance.c
@@ -8,254 +8,143 @@
cases as published by the Free Software Foundation.
*/
-
-#include "tier.h"
#include "dht-common.h"
-#include "xlator.h"
-#include "syscall.h"
-#include <signal.h>
+#include <glusterfs/syscall.h>
#include <fnmatch.h>
#include <signal.h>
-#include "events.h"
-
-#define GF_DISK_SECTOR_SIZE 512
-#define DHT_REBALANCE_PID 4242 /* Change it if required */
-#define DHT_REBALANCE_BLKSIZE (1024 * 1024) /* 1 MB */
-#define MAX_MIGRATE_QUEUE_COUNT 500
-#define MIN_MIGRATE_QUEUE_COUNT 200
-#define MAX_REBAL_TYPE_SIZE 16
-#define FILE_CNT_INTERVAL 600 /* 10 mins */
-#define ESTIMATE_START_INTERVAL 600 /* 10 mins */
-#define HARDLINK_MIG_INPROGRESS -2
-#define SKIP_MIGRATION_FD_POSITIVE -3
+#include <glusterfs/events.h>
+#include "glusterfs/compat-errno.h" // for ENODATA on BSD
+
+#define GF_DISK_SECTOR_SIZE 512
+#define DHT_REBALANCE_PID 4242 /* Change it if required */
+#define DHT_REBALANCE_BLKSIZE 1048576 /* 1 MB */
+#define MAX_MIGRATE_QUEUE_COUNT 500
+#define MIN_MIGRATE_QUEUE_COUNT 200
+#define MAX_REBAL_TYPE_SIZE 16
+#define FILE_CNT_INTERVAL 600 /* 10 mins */
+#define ESTIMATE_START_INTERVAL 600 /* 10 mins */
+#define HARDLINK_MIG_INPROGRESS -2
+#define SKIP_MIGRATION_FD_POSITIVE -3
#ifndef MAX
-#define MAX(a, b) (((a) > (b))?(a):(b))
+#define MAX(a, b) (((a) > (b)) ? (a) : (b))
#endif
-
-#define GF_CRAWL_INDEX_MOVE(idx, sv_cnt) { \
- idx++; \
- idx %= sv_cnt; \
- }
+#define GF_CRAWL_INDEX_MOVE(idx, sv_cnt) \
+ { \
+ idx++; \
+ idx %= sv_cnt; \
+ }
uint64_t g_totalfiles = 0;
uint64_t g_totalsize = 0;
-
void
-gf_defrag_free_dir_dfmeta (struct dir_dfmeta *meta, int local_subvols_cnt)
+gf_defrag_free_dir_dfmeta(struct dir_dfmeta *meta, int local_subvols_cnt)
{
- int i = 0;
-
- if (meta) {
- for (i = 0; i < local_subvols_cnt; i++) {
- gf_dirent_free (&meta->equeue[i]);
- }
-
- GF_FREE (meta->equeue);
- GF_FREE (meta->head);
- GF_FREE (meta->iterator);
- GF_FREE (meta->offset_var);
- GF_FREE (meta->fetch_entries);
- GF_FREE (meta);
- }
+ int i = 0;
+
+ if (meta) {
+ for (i = 0; i < local_subvols_cnt; i++) {
+ if (meta->equeue)
+ gf_dirent_free(&meta->equeue[i]);
+ if (meta->lfd && meta->lfd[i])
+ fd_unref(meta->lfd[i]);
+ }
+
+ GF_FREE(meta->equeue);
+ GF_FREE(meta->head);
+ GF_FREE(meta->iterator);
+ GF_FREE(meta->offset_var);
+ GF_FREE(meta->fetch_entries);
+ GF_FREE(meta->lfd);
+ GF_FREE(meta);
+ }
}
void
-gf_defrag_free_container (struct dht_container *container)
+gf_defrag_free_container(struct dht_container *container)
{
- if (container) {
- gf_dirent_entry_free (container->df_entry);
+ if (container) {
+ gf_dirent_entry_free(container->df_entry);
- if (container->parent_loc) {
- loc_wipe (container->parent_loc);
- }
+ if (container->parent_loc) {
+ loc_wipe(container->parent_loc);
+ }
- GF_FREE (container->parent_loc);
+ GF_FREE(container->parent_loc);
- GF_FREE (container);
- }
+ GF_FREE(container);
+ }
}
void
-dht_set_global_defrag_error (gf_defrag_info_t *defrag, int ret)
+dht_set_global_defrag_error(gf_defrag_info_t *defrag, int ret)
{
- LOCK (&defrag->lock);
- {
- defrag->global_error = ret;
- }
- UNLOCK (&defrag->lock);
- return;
+ LOCK(&defrag->lock);
+ {
+ defrag->global_error = ret;
+ }
+ UNLOCK(&defrag->lock);
+ return;
}
-
-static gf_boolean_t
-dht_is_tier_command (int cmd) {
-
- gf_boolean_t is_tier = _gf_false;
-
- switch (cmd) {
- case GF_DEFRAG_CMD_START_TIER:
- case GF_DEFRAG_CMD_STATUS_TIER:
- case GF_DEFRAG_CMD_START_DETACH_TIER:
- case GF_DEFRAG_CMD_STOP_DETACH_TIER:
- case GF_DEFRAG_CMD_PAUSE_TIER:
- case GF_DEFRAG_CMD_RESUME_TIER:
- is_tier = _gf_true;
- break;
- default:
- break;
- }
- return is_tier;
-
-}
-
-
static int
-dht_send_rebalance_event (xlator_t *this, int cmd, gf_defrag_status_t status)
+dht_send_rebalance_event(xlator_t *this, int cmd, gf_defrag_status_t status)
{
- int ret = -1;
- char *volname = NULL;
- char *tmpstr = NULL;
- char *ptr = NULL;
- char *suffix = "-dht";
- dht_conf_t *conf = NULL;
- gf_defrag_info_t *defrag = NULL;
- int len = 0;
-
- eventtypes_t event = EVENT_LAST;
-
- switch (status) {
+ int ret = -1;
+ char *volname = NULL;
+ char *tmpstr = NULL;
+ char *ptr = NULL;
+ char *suffix = "-dht";
+ int len = 0;
+
+ eventtypes_t event = EVENT_LAST;
+
+ switch (status) {
case GF_DEFRAG_STATUS_COMPLETE:
- event = EVENT_VOLUME_REBALANCE_COMPLETE;
- break;
+ event = EVENT_VOLUME_REBALANCE_COMPLETE;
+ break;
case GF_DEFRAG_STATUS_FAILED:
- event = EVENT_VOLUME_REBALANCE_FAILED;
- break;
+ event = EVENT_VOLUME_REBALANCE_FAILED;
+ break;
case GF_DEFRAG_STATUS_STOPPED:
- event = EVENT_VOLUME_REBALANCE_STOP;
- break;
+ event = EVENT_VOLUME_REBALANCE_STOP;
+ break;
default:
- break;
-
- }
-
- if (dht_is_tier_command (cmd)) {
- /* We should have the tier volume name*/
- conf = this->private;
- defrag = conf->defrag;
- volname = defrag->tier_conf.volname;
- } else {
- /* DHT volume */
- len = strlen (this->name);
- tmpstr = gf_strdup (this->name);
- if (tmpstr) {
- ptr = tmpstr + (len - strlen (suffix));
- if (!strcmp (ptr, suffix)) {
- tmpstr[len - strlen (suffix)] = '\0';
- volname = tmpstr;
- }
- }
- }
-
- if (!volname) {
- /* Better than nothing */
- volname = this->name;
- }
-
- if (event != EVENT_LAST) {
- gf_event (event, "volume=%s", volname);
- }
-
- GF_FREE (tmpstr);
- return ret;
+ break;
+ }
+
+ /* DHT volume */
+ len = strlen(this->name) - strlen(suffix);
+ tmpstr = gf_strdup(this->name);
+ if (tmpstr) {
+ ptr = tmpstr + len;
+ if (!strcmp(ptr, suffix)) {
+ tmpstr[len] = '\0';
+ volname = tmpstr;
+ }
+ }
+
+ if (!volname) {
+ /* Better than nothing */
+ volname = this->name;
+ }
+
+ if (event != EVENT_LAST) {
+ gf_event(event, "volume=%s", volname);
+ }
+
+ GF_FREE(tmpstr);
+ return ret;
}
-
static void
-dht_strip_out_acls (dict_t *dict)
-{
- if (dict) {
- dict_del (dict, "trusted.SGI_ACL_FILE");
- dict_del (dict, POSIX_ACL_ACCESS_XATTR);
- }
-}
-
-
-
-static int
-dht_write_with_holes (xlator_t *to, fd_t *fd, struct iovec *vec, int count,
- int32_t size, off_t offset, struct iobref *iobref,
- int *fop_errno)
+dht_strip_out_acls(dict_t *dict)
{
- int i = 0;
- int ret = -1;
- int start_idx = 0;
- int tmp_offset = 0;
- int write_needed = 0;
- int buf_len = 0;
- int size_pending = 0;
- char *buf = NULL;
-
- /* loop through each vector */
- for (i = 0; i < count; i++) {
- buf = vec[i].iov_base;
- buf_len = vec[i].iov_len;
-
- for (start_idx = 0; (start_idx + GF_DISK_SECTOR_SIZE) <= buf_len;
- start_idx += GF_DISK_SECTOR_SIZE) {
-
- if (mem_0filled (buf + start_idx, GF_DISK_SECTOR_SIZE) != 0) {
- write_needed = 1;
- continue;
- }
-
- if (write_needed) {
- ret = syncop_write (to, fd, (buf + tmp_offset),
- (start_idx - tmp_offset),
- (offset + tmp_offset),
- iobref, 0, NULL, NULL);
- /* 'path' will be logged in calling function */
- if (ret < 0) {
- gf_log (THIS->name, GF_LOG_WARNING,
- "failed to write (%s)",
- strerror (-ret));
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
-
- write_needed = 0;
- }
- tmp_offset = start_idx + GF_DISK_SECTOR_SIZE;
- }
-
- if ((start_idx < buf_len) || write_needed) {
- /* This means, last chunk is not yet written.. write it */
- ret = syncop_write (to, fd, (buf + tmp_offset),
- (buf_len - tmp_offset),
- (offset + tmp_offset), iobref, 0,
- NULL, NULL);
- if (ret < 0) {
- /* 'path' will be logged in calling function */
- gf_log (THIS->name, GF_LOG_WARNING,
- "failed to write (%s)",
- strerror (-ret));
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
- }
-
- size_pending = (size - buf_len);
- if (!size_pending)
- break;
- }
-
- ret = size;
-out:
- return ret;
-
+ if (dict) {
+ dict_del(dict, "trusted.SGI_ACL_FILE");
+ dict_del(dict, POSIX_ACL_ACCESS_XATTR);
+ }
}
/*
@@ -294,323 +183,318 @@ be converted to "0" in dht_migrate_file.
*/
int32_t
-gf_defrag_handle_hardlink (xlator_t *this, loc_t *loc, int *fop_errno)
+gf_defrag_handle_hardlink(xlator_t *this, loc_t *loc, int *fop_errno)
{
- int32_t ret = -1;
- xlator_t *cached_subvol = NULL;
- xlator_t *hashed_subvol = NULL;
- xlator_t *linkto_subvol = NULL;
- data_t *data = NULL;
- struct iatt iatt = {0,};
- int32_t op_errno = 0;
- dht_conf_t *conf = NULL;
- gf_loglevel_t loglevel = 0;
- dict_t *link_xattr = NULL;
- dict_t *dict = NULL;
- dict_t *xattr_rsp = NULL;
- struct iatt stbuf = {0,};
-
+ int32_t ret = -1;
+ xlator_t *cached_subvol = NULL;
+ xlator_t *hashed_subvol = NULL;
+ xlator_t *linkto_subvol = NULL;
+ data_t *data = NULL;
+ struct iatt iatt = {
+ 0,
+ };
+ int32_t op_errno = 0;
+ dht_conf_t *conf = NULL;
+ gf_loglevel_t loglevel = 0;
+ dict_t *link_xattr = NULL;
+ dict_t *dict = NULL;
+ dict_t *xattr_rsp = NULL;
+ struct iatt stbuf = {
+ 0,
+ };
+
+ *fop_errno = EINVAL;
+
+ GF_VALIDATE_OR_GOTO("defrag", loc, out);
+ GF_VALIDATE_OR_GOTO("defrag", loc->name, out);
+ GF_VALIDATE_OR_GOTO("defrag", this, out);
+ GF_VALIDATE_OR_GOTO("defrag", this->private, out);
+
+ conf = this->private;
+
+ if (gf_uuid_is_null(loc->pargfid)) {
+ gf_msg("", GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed :"
+ "loc->pargfid is NULL for %s",
+ loc->path);
*fop_errno = EINVAL;
+ ret = -1;
+ goto out;
+ }
+
+ if (gf_uuid_is_null(loc->gfid)) {
+ gf_msg("", GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed :"
+ "loc->gfid is NULL for %s",
+ loc->path);
+ *fop_errno = EINVAL;
+ ret = -1;
+ goto out;
+ }
- GF_VALIDATE_OR_GOTO ("defrag", loc, out);
- GF_VALIDATE_OR_GOTO ("defrag", loc->name, out);
- GF_VALIDATE_OR_GOTO ("defrag", this, out);
- GF_VALIDATE_OR_GOTO ("defrag", this->private, out);
-
- conf = this->private;
-
- if (gf_uuid_is_null (loc->pargfid)) {
- gf_msg ("", GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed :"
- "loc->pargfid is NULL for %s", loc->path);
- *fop_errno = EINVAL;
- ret = -1;
- goto out;
- }
-
- if (gf_uuid_is_null (loc->gfid)) {
- gf_msg ("", GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed :"
- "loc->gfid is NULL for %s", loc->path);
- *fop_errno = EINVAL;
- ret = -1;
- goto out;
- }
-
- link_xattr = dict_new ();
- if (!link_xattr) {
- ret = -1;
- *fop_errno = ENOMEM;
- goto out;
- }
-
- /*
- Parallel migration can lead to migration of the hard link multiple
- times which can lead to data loss. Hence, adding a fresh lookup to
- decide whether migration is required or not.
-
- Elaborating the scenario for let say 10 hardlinks [link{1..10}]:
- Let say the first hard link "link1" does the setxattr of the
- new hashed subvolume info on the cached file. As there are multiple
- threads working, we might have already all the links created on the
- new hashed by the time we reach hardlink let say link5. Now the
- number of links on hashed is equal to that of cached. Hence, file
- migration will happen for link6.
-
- Cached Hashed
- --------T link6 rwxrwxrwx link6
-
- Now post above state all the link file on the cached will be zero
- byte linkto files. Hence, if we still do migration for the following
- files link{7..10}, we will end up migrating 0 data leading to data
- loss.
- Hence, a lookup can make sure whether we need to migrate the
- file or not.
- */
-
- dict = dict_new ();
- if (!dict) {
- ret = -1;
- *fop_errno = ENOMEM;
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
- "could not allocate memory for dict");
- goto out;
- }
-
- ret = dict_set_int32 (dict, conf->link_xattr_name, 256);
+ link_xattr = dict_new();
+ if (!link_xattr) {
+ ret = -1;
+ *fop_errno = ENOMEM;
+ goto out;
+ }
+
+ /*
+ Parallel migration can lead to migration of the hard link multiple
+ times which can lead to data loss. Hence, adding a fresh lookup to
+ decide whether migration is required or not.
+
+ Elaborating the scenario for let say 10 hardlinks [link{1..10}]:
+ Let say the first hard link "link1" does the setxattr of the
+ new hashed subvolume info on the cached file. As there are multiple
+ threads working, we might have already all the links created on the
+ new hashed by the time we reach hardlink let say link5. Now the
+ number of links on hashed is equal to that of cached. Hence, file
+ migration will happen for link6.
+
+ Cached Hashed
+ --------T link6 rwxrwxrwx link6
+
+ Now post above state all the link file on the cached will be zero
+ byte linkto files. Hence, if we still do migration for the following
+ files link{7..10}, we will end up migrating 0 data leading to data
+ loss.
+ Hence, a lookup can make sure whether we need to migrate the
+ file or not.
+ */
+
+ dict = dict_new();
+ if (!dict) {
+ ret = -1;
+ *fop_errno = ENOMEM;
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
+ "could not allocate memory for dict");
+ goto out;
+ }
+
+ ret = dict_set_int32(dict, conf->link_xattr_name, 256);
+ if (ret) {
+ *fop_errno = ENOMEM;
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed:"
+ "%s: failed to set 'linkto' key in dict",
+ loc->path);
+ goto out;
+ }
+
+ ret = syncop_lookup(this, loc, &stbuf, NULL, dict, &xattr_rsp);
+ if (ret) {
+ /*Ignore ENOENT and ESTALE as file might have been
+ migrated already*/
+ if (-ret == ENOENT || -ret == ESTALE) {
+ ret = -2;
+ goto out;
+ }
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed:%s lookup failed with ret = %d", loc->path,
+ ret);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ cached_subvol = dht_subvol_get_cached(this, loc->inode);
+ if (!cached_subvol) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed :"
+ "Failed to get cached subvol"
+ " for %s on %s",
+ loc->name, this->name);
+ *fop_errno = EINVAL;
+ ret = -1;
+ goto out;
+ }
+
+ hashed_subvol = dht_subvol_get_hashed(this, loc);
+ if (!hashed_subvol) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed :"
+ "Failed to get hashed subvol"
+ " for %s on %s",
+ loc->name, this->name);
+ *fop_errno = EINVAL;
+ ret = -1;
+ goto out;
+ }
+
+ /* Hardlink migration happens only with remove-brick. So this condition will
+ * be true only when the migration has happened. In case hardlinks are
+ * migrated for rebalance case, remove this check. Having this check here
+ * avoid redundant calls below*/
+ if (hashed_subvol == cached_subvol) {
+ ret = -2;
+ goto out;
+ }
+
+ gf_log(this->name, GF_LOG_INFO,
+ "Attempting to migrate hardlink %s "
+ "with gfid %s from %s -> %s",
+ loc->name, uuid_utoa(loc->gfid), cached_subvol->name,
+ hashed_subvol->name);
+
+ data = dict_get(xattr_rsp, conf->link_xattr_name);
+ /* set linkto on cached -> hashed if not present, else link it */
+ if (!data) {
+ ret = dict_set_str(link_xattr, conf->link_xattr_name,
+ hashed_subvol->name);
if (ret) {
- *fop_errno = ENOMEM;
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed:"
- "%s: failed to set 'linkto' key in dict", loc->path);
- goto out;
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed :"
+ "Failed to set dictionary value:"
+ " key = %s for %s",
+ conf->link_xattr_name, loc->name);
+ *fop_errno = ENOMEM;
+ ret = -1;
+ goto out;
}
- ret = syncop_lookup (this, loc, &stbuf, NULL, dict, &xattr_rsp);
+ ret = syncop_setxattr(cached_subvol, loc, link_xattr, 0, NULL, NULL);
if (ret) {
- /*Ignore ENOENT and ESTALE as file might have been
- migrated already*/
- if (-ret == ENOENT || -ret == ESTALE) {
- ret = -2;
- goto out;
- }
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed:%s lookup failed with ret = %d",
- loc->path, ret);
- *fop_errno = -ret;
- ret = -1;
- goto out;
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed :"
+ "Linkto setxattr failed %s -> %s",
+ cached_subvol->name, loc->name);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
}
- cached_subvol = dht_subvol_get_cached (this, loc->inode);
- if (!cached_subvol) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed :"
- "Failed to get cached subvol"
- " for %s on %s", loc->name, this->name);
- *fop_errno = EINVAL;
- ret = -1;
- goto out;
- }
-
- hashed_subvol = dht_subvol_get_hashed (this, loc);
- if (!hashed_subvol) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed :"
- "Failed to get hashed subvol"
- " for %s on %s", loc->name, this->name);
- *fop_errno = EINVAL;
- ret = -1;
- goto out;
- }
-
- /* Hardlink migration happens only with remove-brick. So this condition will
- * be true only when the migration has happened. In case hardlinks are migrated
- * for rebalance case, remove this check. Having this check here avoid redundant
- * calls below*/
- if (hashed_subvol == cached_subvol) {
- ret = -2;
- goto out;
- }
+ gf_msg_debug(this->name, 0,
+ "hardlink target subvol created on %s "
+ ",cached %s, file %s",
+ hashed_subvol->name, cached_subvol->name, loc->path);
- gf_log (this->name, GF_LOG_INFO, "Attempting to migrate hardlink %s "
- "with gfid %s from %s -> %s", loc->name, uuid_utoa (loc->gfid),
- cached_subvol->name, hashed_subvol->name);
-
- data = dict_get (xattr_rsp, conf->link_xattr_name);
- /* set linkto on cached -> hashed if not present, else link it */
- if (!data) {
- ret = dict_set_str (link_xattr, conf->link_xattr_name,
- hashed_subvol->name);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed :"
- "Failed to set dictionary value:"
- " key = %s for %s",
- conf->link_xattr_name, loc->name);
- *fop_errno = ENOMEM;
- ret = -1;
- goto out;
- }
-
- ret = syncop_setxattr (cached_subvol, loc, link_xattr, 0, NULL,
- NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed :"
- "Linkto setxattr failed %s -> %s",
- cached_subvol->name,
- loc->name);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
-
- gf_msg_debug (this->name, 0, "hardlink target subvol created on %s "
- ",cached %s, file %s",
- hashed_subvol->name, cached_subvol->name, loc->path);
-
- ret = -2;
- goto out;
+ ret = -2;
+ goto out;
+ } else {
+ linkto_subvol = dht_linkfile_subvol(this, NULL, NULL, xattr_rsp);
+ if (!linkto_subvol) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_SUBVOL_ERROR,
+ "Failed to get "
+ "linkto subvol for %s",
+ loc->name);
} else {
- linkto_subvol = dht_linkfile_subvol (this, NULL, NULL, xattr_rsp);
- if (!linkto_subvol) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_SUBVOL_ERROR,
- "Failed to get "
- "linkto subvol for %s", loc->name);
- } else {
- hashed_subvol = linkto_subvol;
- }
-
- ret = syncop_link (hashed_subvol, loc, loc, &iatt, NULL, NULL);
- if (ret) {
- op_errno = -ret;
- ret = -1;
-
- loglevel = (op_errno == EEXIST) ? GF_LOG_DEBUG : \
- GF_LOG_ERROR;
- gf_msg (this->name, loglevel, op_errno,
- DHT_MSG_MIGRATE_HARDLINK_FILE_FAILED,
- "link of %s -> %s"
- " failed on subvol %s", loc->name,
- uuid_utoa(loc->gfid),
- hashed_subvol->name);
- if (op_errno != EEXIST) {
- *fop_errno = op_errno;
- goto out;
- }
- } else {
- gf_msg_debug (this->name, 0, "syncop_link successful for"
- " hardlink %s on subvol %s, cached %s", loc->path,
- hashed_subvol->name, cached_subvol->name);
-
- }
+ hashed_subvol = linkto_subvol;
}
- ret = syncop_lookup (hashed_subvol, loc, &iatt, NULL, NULL, NULL);
+ ret = syncop_link(hashed_subvol, loc, loc, &iatt, NULL, NULL);
if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed :Failed lookup %s on %s ",
- loc->name, hashed_subvol->name);
-
- *fop_errno = -ret;
- ret = -1;
+ op_errno = -ret;
+ ret = -1;
+
+ loglevel = (op_errno == EEXIST) ? GF_LOG_DEBUG : GF_LOG_ERROR;
+ gf_msg(this->name, loglevel, op_errno,
+ DHT_MSG_MIGRATE_HARDLINK_FILE_FAILED,
+ "link of %s -> %s"
+ " failed on subvol %s",
+ loc->name, uuid_utoa(loc->gfid), hashed_subvol->name);
+ if (op_errno != EEXIST) {
+ *fop_errno = op_errno;
goto out;
+ }
+ } else {
+ gf_msg_debug(this->name, 0,
+ "syncop_link successful for"
+ " hardlink %s on subvol %s, cached %s",
+ loc->path, hashed_subvol->name, cached_subvol->name);
}
+ }
- /* There is a race where on the target subvol for the hardlink
- * (note: hash subvol for the hardlink might differ from this), some
- * other client(non-rebalance) would have created a linkto file for that
- * hardlink as part of lookup. So let say there are 10 hardlinks, on the
- * 5th hardlink it self the hardlinks might have migrated. Now for
- * (6..10th) hardlinks the cached and target would be same as the file
- * has already migrated. Hence this check is needed */
- if (cached_subvol == hashed_subvol) {
- gf_msg_debug (this->name, 0, "source %s and destination %s "
- "for hardlink %s are same", cached_subvol->name,
- hashed_subvol->name, loc->path);
- ret = -2;
- goto out;
- }
+ ret = syncop_lookup(hashed_subvol, loc, &iatt, NULL, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed :Failed lookup %s on %s ", loc->name,
+ hashed_subvol->name);
- if (iatt.ia_nlink == stbuf.ia_nlink) {
- ret = dht_migrate_file (this, loc, cached_subvol, hashed_subvol,
- GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS,
- fop_errno);
- if (ret) {
- goto out;
- }
- }
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ /* There is a race where on the target subvol for the hardlink
+ * (note: hash subvol for the hardlink might differ from this), some
+ * other client(non-rebalance) would have created a linkto file for that
+ * hardlink as part of lookup. So let say there are 10 hardlinks, on the
+ * 5th hardlink it self the hardlinks might have migrated. Now for
+ * (6..10th) hardlinks the cached and target would be same as the file
+ * has already migrated. Hence this check is needed */
+ if (cached_subvol == hashed_subvol) {
+ gf_msg_debug(this->name, 0,
+ "source %s and destination %s "
+ "for hardlink %s are same",
+ cached_subvol->name, hashed_subvol->name, loc->path);
ret = -2;
+ goto out;
+ }
+
+ if (iatt.ia_nlink == stbuf.ia_nlink) {
+ ret = dht_migrate_file(this, loc, cached_subvol, hashed_subvol,
+ GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS, fop_errno);
+ if (ret) {
+ goto out;
+ }
+ }
+ ret = -2;
out:
- if (link_xattr)
- dict_unref (link_xattr);
+ if (link_xattr)
+ dict_unref(link_xattr);
- if (xattr_rsp)
- dict_unref (xattr_rsp);
+ if (xattr_rsp)
+ dict_unref(xattr_rsp);
- if (dict)
- dict_unref (dict);
+ if (dict)
+ dict_unref(dict);
- return ret;
+ return ret;
}
-
-
static int
-__check_file_has_hardlink (xlator_t *this, loc_t *loc,
- struct iatt *stbuf, dict_t *xattrs, int flags,
- gf_defrag_info_t *defrag, dht_conf_t *conf, int *fop_errno)
+__check_file_has_hardlink(xlator_t *this, loc_t *loc, struct iatt *stbuf,
+ dict_t *xattrs, int flags, gf_defrag_info_t *defrag,
+ dht_conf_t *conf, int *fop_errno)
{
- int ret = 0;
+ int ret = 0;
- if (flags == GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS) {
- ret = 0;
- return ret;
- }
- if (stbuf->ia_nlink > 1) {
- /* support for decomission */
- if (flags == GF_DHT_MIGRATE_HARDLINK) {
- synclock_lock (&conf->link_lock);
- ret = gf_defrag_handle_hardlink
- (this, loc, fop_errno);
- synclock_unlock (&conf->link_lock);
- /*
- Returning zero will force the file to be remigrated.
- Checkout gf_defrag_handle_hardlink for more information.
- */
- if (ret && ret != -2) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed:"
- "%s: failed to migrate file with link",
- loc->path);
- }
- } else {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migration skipped for:"
- "%s: file has hardlinks", loc->path);
- *fop_errno = ENOTSUP;
- ret = 1;
- }
- }
+ if (flags == GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS) {
+ ret = 0;
+ return ret;
+ }
+ if (stbuf->ia_nlink > 1) {
+ /* support for decomission */
+ if (flags == GF_DHT_MIGRATE_HARDLINK) {
+ synclock_lock(&conf->link_lock);
+ ret = gf_defrag_handle_hardlink(this, loc, fop_errno);
+ synclock_unlock(&conf->link_lock);
+ /*
+ Returning zero will force the file to be remigrated.
+ Checkout gf_defrag_handle_hardlink for more information.
+ */
+ if (ret && ret != -2) {
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed:"
+ "%s: failed to migrate file with link",
+ loc->path);
+ }
+ } else {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migration skipped for:"
+ "%s: file has hardlinks",
+ loc->path);
+ *fop_errno = ENOTSUP;
+ ret = 1;
+ }
+ }
- return ret;
+ return ret;
}
-
/*
return values
0 : File will be migrated
@@ -620,961 +504,963 @@ __check_file_has_hardlink (xlator_t *this, loc_t *loc,
-1 : failure
*/
static int
-__is_file_migratable (xlator_t *this, loc_t *loc,
- struct iatt *stbuf, dict_t *xattrs, int flags,
- gf_defrag_info_t *defrag, dht_conf_t *conf,
- int *fop_errno)
+__is_file_migratable(xlator_t *this, loc_t *loc, struct iatt *stbuf,
+ dict_t *xattrs, int flags, gf_defrag_info_t *defrag,
+ dht_conf_t *conf, int *fop_errno)
{
- int ret = -1;
- int lock_count = 0;
-
- if (IA_ISDIR (stbuf->ia_type)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed:"
- "%s: migrate-file called on directory", loc->path);
- *fop_errno = EISDIR;
- ret = -1;
- goto out;
- }
-
- if (!conf->lock_migration_enabled) {
- ret = dict_get_int32 (xattrs, GLUSTERFS_POSIXLK_COUNT,
- &lock_count);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed:"
- "%s: Unable to get lock count for file",
- loc->path);
- *fop_errno = EINVAL;
- ret = -1;
- goto out;
- }
-
- if (lock_count) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed: %s: File has locks."
- " Skipping file migration", loc->path);
- *fop_errno = ENOTSUP;
- ret = 1;
- goto out;
- }
- }
+ int ret = -1;
+ int lock_count = 0;
+
+ if (IA_ISDIR(stbuf->ia_type)) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed:"
+ "%s: migrate-file called on directory",
+ loc->path);
+ *fop_errno = EISDIR;
+ ret = -1;
+ goto out;
+ }
- /* Check if file has hardlink*/
- ret = __check_file_has_hardlink (this, loc, stbuf, xattrs,
- flags, defrag, conf, fop_errno);
+ if (!conf->lock_migration_enabled) {
+ ret = dict_get_int32(xattrs, GLUSTERFS_POSIXLK_COUNT, &lock_count);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed:"
+ "%s: Unable to get lock count for file",
+ loc->path);
+ *fop_errno = EINVAL;
+ ret = -1;
+ goto out;
+ }
+
+ if (lock_count) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed: %s: File has locks."
+ " Skipping file migration",
+ loc->path);
+ *fop_errno = ENOTSUP;
+ ret = 1;
+ goto out;
+ }
+ }
+
+ /* Check if file has hardlink*/
+ ret = __check_file_has_hardlink(this, loc, stbuf, xattrs, flags, defrag,
+ conf, fop_errno);
out:
- return ret;
+ return ret;
}
-
static int
-__dht_rebalance_create_dst_file (xlator_t *this, xlator_t *to, xlator_t *from,
- loc_t *loc, struct iatt *stbuf, fd_t **dst_fd,
- int *fop_errno)
+__dht_rebalance_create_dst_file(xlator_t *this, xlator_t *to, xlator_t *from,
+ loc_t *loc, struct iatt *stbuf, fd_t **dst_fd,
+ int *fop_errno, int file_has_holes)
{
- int ret = -1;
- int ret2 = -1;
- fd_t *fd = NULL;
- struct iatt new_stbuf = {0,};
- struct iatt check_stbuf= {0,};
- dht_conf_t *conf = NULL;
- dict_t *dict = NULL;
- dict_t *xdata = NULL;
-
- conf = this->private;
-
- dict = dict_new ();
- if (!dict) {
- *fop_errno = ENOMEM;
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_NO_MEMORY, "dictionary allocation failed for"
- "path:%s", loc->path);
- goto out;
- }
- ret = dict_set_gfuuid (dict, "gfid-req", stbuf->ia_gfid, true);
- if (ret) {
- *fop_errno = ENOMEM;
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "%s: failed to set dictionary value: key = gfid-req",
- loc->path);
- goto out;
- }
-
- ret = dict_set_str (dict, conf->link_xattr_name, from->name);
- if (ret) {
- *fop_errno = ENOMEM;
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "%s: failed to set dictionary value: key = %s ",
- loc->path, conf->link_xattr_name);
- goto out;
- }
-
- fd = fd_create (loc->inode, DHT_REBALANCE_PID);
- if (!fd) {
- *fop_errno = ENOMEM;
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: fd create failed (destination)",
- loc->path);
- goto out;
- }
-
- if (!!dht_is_tier_xlator (this)) {
- xdata = dict_new ();
- if (!xdata) {
- *fop_errno = ENOMEM;
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: dict_new failed)",
- loc->path);
- goto out;
- }
-
- ret = dict_set_int32 (xdata, GF_CLEAN_WRITE_PROTECTION, 1);
- if (ret) {
- *fop_errno = ENOMEM;
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "%s: failed to set dictionary value: key = %s ",
- loc->path, GF_CLEAN_WRITE_PROTECTION);
- goto out;
- }
- }
-
- ret = syncop_lookup (to, loc, &new_stbuf, NULL, xdata, NULL);
- if (!ret) {
- /* File exits in the destination, check if gfid matches */
- if (gf_uuid_compare (stbuf->ia_gfid, new_stbuf.ia_gfid) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_GFID_MISMATCH,
- "file %s exists in %s with different gfid",
- loc->path, to->name);
- *fop_errno = EINVAL;
- ret = -1;
- goto out;
- }
- }
- if ((ret < 0) && (-ret != ENOENT)) {
- /* File exists in destination, but not accessible */
- gf_msg (THIS->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: failed to lookup file",
- loc->path);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
-
- /* Create the destination with LINKFILE mode, and linkto xattr,
- if the linkfile already exists, just open the file */
- if (!ret) {
- /*
- * File already present, just open the file.
- */
- ret = syncop_open (to, loc, O_RDWR, fd, NULL, NULL);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "failed to open %s on %s",
- loc->path, to->name);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
- } else {
- ret = syncop_create (to, loc, O_RDWR, DHT_LINKFILE_MODE, fd,
- &new_stbuf, dict, NULL);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "failed to create %s on %s",
- loc->path, to->name);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
-
- }
-
- fd_bind (fd);
+ int ret = -1;
+ int ret2 = -1;
+ fd_t *fd = NULL;
+ struct iatt new_stbuf = {
+ 0,
+ };
+ struct iatt check_stbuf = {
+ 0,
+ };
+ dht_conf_t *conf = NULL;
+ dict_t *dict = NULL;
+ dict_t *xdata = NULL;
+
+ conf = this->private;
+
+ dict = dict_new();
+ if (!dict) {
+ *fop_errno = ENOMEM;
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
+ "dictionary allocation failed for"
+ "path:%s",
+ loc->path);
+ goto out;
+ }
+ ret = dict_set_gfuuid(dict, "gfid-req", stbuf->ia_gfid, true);
+ if (ret) {
+ *fop_errno = ENOMEM;
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "%s: failed to set dictionary value: key = gfid-req", loc->path);
+ goto out;
+ }
+
+ ret = dict_set_str(dict, conf->link_xattr_name, from->name);
+ if (ret) {
+ *fop_errno = ENOMEM;
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "%s: failed to set dictionary value: key = %s ", loc->path,
+ conf->link_xattr_name);
+ goto out;
+ }
+
+ fd = fd_create(loc->inode, DHT_REBALANCE_PID);
+ if (!fd) {
+ *fop_errno = ENOMEM;
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: fd create failed (destination)", loc->path);
+ goto out;
+ }
+
+ xdata = dict_new();
+ if (!xdata) {
+ *fop_errno = ENOMEM;
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: dict_new failed)", loc->path);
+ goto out;
+ }
+
+ ret = dict_set_int32_sizen(xdata, GF_CLEAN_WRITE_PROTECTION, 1);
+ if (ret) {
+ *fop_errno = ENOMEM;
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "%s: failed to set dictionary value: key = %s ", loc->path,
+ GF_CLEAN_WRITE_PROTECTION);
+ goto out;
+ }
+
+ ret = syncop_lookup(to, loc, &new_stbuf, NULL, xdata, NULL);
+ if (!ret) {
+ /* File exits in the destination, check if gfid matches */
+ if (gf_uuid_compare(stbuf->ia_gfid, new_stbuf.ia_gfid) != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_GFID_MISMATCH,
+ "file %s exists in %s with different gfid", loc->path,
+ to->name);
+ *fop_errno = EINVAL;
+ ret = -1;
+ goto out;
+ }
+ }
+ if ((ret < 0) && (-ret != ENOENT)) {
+ /* File exists in destination, but not accessible */
+ gf_msg(THIS->name, GF_LOG_WARNING, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: failed to lookup file", loc->path);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
- /*Reason of doing lookup after create again:
- *In the create, there is some time-gap between opening fd at the
- *server (posix_layer) and binding it in server (incrementing fd count),
- *so if in that time-gap, if other process sends unlink considering it
- *as a linkto file, because inode->fd count will be 0, so file will be
- *unlinked at the backend. And because furthur operations are performed
- *on fd, so though migration will be done but will end with no file
- *at the backend.
+ /* Create the destination with LINKFILE mode, and linkto xattr,
+ if the linkfile already exists, just open the file */
+ if (!ret) {
+ /*
+ * File already present, just open the file.
*/
-
- ret = syncop_lookup (to, loc, &check_stbuf, NULL, NULL, NULL);
- if (!ret) {
-
- if (gf_uuid_compare (stbuf->ia_gfid, check_stbuf.ia_gfid) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_GFID_MISMATCH,
- "file %s exists in %s with different gfid,"
- "found in lookup after create",
- loc->path, to->name);
- *fop_errno = EINVAL;
- ret = -1;
- goto out;
- }
-
- }
-
- if (-ret == ENOENT) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED, "%s: file does not exist"
- "on %s", loc->path, to->name);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
-
- ret = syncop_fsetattr (to, fd, stbuf,
- (GF_SET_ATTR_UID | GF_SET_ATTR_GID),
- NULL, NULL, NULL, NULL);
+ ret = syncop_open(to, loc, O_RDWR, fd, NULL, NULL);
if (ret < 0) {
- *fop_errno = -ret;
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "chown failed for %s on %s",
- loc->path, to->name);
- }
-
- /* No need to bother about 0 byte size files */
- if (stbuf->ia_size > 0) {
- if (conf->use_fallocate) {
- ret = syncop_fallocate (to, fd, 0, 0, stbuf->ia_size,
- NULL, NULL);
- if (ret < 0) {
- if (ret == -EOPNOTSUPP || ret == -EINVAL ||
- ret == -ENOSYS) {
- conf->use_fallocate = _gf_false;
- } else {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "fallocate failed for %s on %s",
- loc->path, to->name);
-
- *fop_errno = -ret;
-
- /* fallocate does not release the space
- * in some cases
- */
- ret2 = syncop_ftruncate (to, fd, 0,
- NULL, NULL);
- if (ret2 < 0) {
- gf_msg (this->name,
- GF_LOG_WARNING, -ret2,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "ftruncate failed for "
- "%s on %s",
- loc->path, to->name);
- }
- goto out;
- }
- }
- }
-
- if (!conf->use_fallocate) {
- ret = syncop_ftruncate (to, fd, stbuf->ia_size, NULL, NULL);
- if (ret < 0) {
- *fop_errno = -ret;
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "ftruncate failed for %s on %s",
- loc->path, to->name);
- }
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "failed to open %s on %s", loc->path, to->name);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+ } else {
+ ret = syncop_create(to, loc, O_RDWR, DHT_LINKFILE_MODE, fd, &new_stbuf,
+ dict, NULL);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "failed to create %s on %s", loc->path, to->name);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+ }
+
+ fd_bind(fd);
+
+ /*Reason of doing lookup after create again:
+ *In the create, there is some time-gap between opening fd at the
+ *server (posix_layer) and binding it in server (incrementing fd count),
+ *so if in that time-gap, if other process sends unlink considering it
+ *as a linkto file, because inode->fd count will be 0, so file will be
+ *unlinked at the backend. And because further operations are performed
+ *on fd, so though migration will be done but will end with no file
+ *at the backend.
+ */
+
+ ret = syncop_lookup(to, loc, &check_stbuf, NULL, NULL, NULL);
+ if (!ret) {
+ if (gf_uuid_compare(stbuf->ia_gfid, check_stbuf.ia_gfid) != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_GFID_MISMATCH,
+ "file %s exists in %s with different gfid,"
+ "found in lookup after create",
+ loc->path, to->name);
+ *fop_errno = EINVAL;
+ ret = -1;
+ goto out;
+ }
+ }
+
+ if (-ret == ENOENT) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: file does not exist"
+ "on %s",
+ loc->path, to->name);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ ret = syncop_fsetattr(to, fd, stbuf, (GF_SET_ATTR_UID | GF_SET_ATTR_GID),
+ NULL, NULL, NULL, NULL);
+ if (ret < 0) {
+ *fop_errno = -ret;
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "chown failed for %s on %s", loc->path, to->name);
+ }
+
+ /* No need to bother about 0 byte size files */
+ if (stbuf->ia_size > 0) {
+ if (conf->use_fallocate && !file_has_holes) {
+ ret = syncop_fallocate(to, fd, 0, 0, stbuf->ia_size, NULL, NULL);
+ if (ret < 0) {
+ if (ret == -EOPNOTSUPP || ret == -EINVAL || ret == -ENOSYS) {
+ conf->use_fallocate = _gf_false;
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, -ret,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "fallocate failed for %s on %s", loc->path,
+ to->name);
+
+ *fop_errno = -ret;
+
+ /* fallocate does not release the space
+ * in some cases
+ */
+ ret2 = syncop_ftruncate(to, fd, 0, NULL, NULL, NULL, NULL);
+ if (ret2 < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret2,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "ftruncate failed for "
+ "%s on %s",
+ loc->path, to->name);
+ }
+ goto out;
}
+ }
+ } else {
+ ret = syncop_ftruncate(to, fd, stbuf->ia_size, NULL, NULL, NULL,
+ NULL);
+ if (ret < 0) {
+ *fop_errno = -ret;
+ gf_msg(this->name, GF_LOG_WARNING, -ret,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "ftruncate failed for %s on %s", loc->path, to->name);
+ }
}
+ }
- /* success */
- ret = 0;
+ /* success */
+ ret = 0;
- if (dst_fd)
- *dst_fd = fd;
+ if (dst_fd)
+ *dst_fd = fd;
out:
- if (ret) {
- if (fd) {
- fd_unref (fd);
- }
+ if (ret) {
+ if (fd) {
+ fd_unref(fd);
}
- if (dict)
- dict_unref (dict);
-
- if (xdata)
- dict_unref (dict);
+ }
+ if (dict)
+ dict_unref(dict);
+ if (xdata)
+ dict_unref(xdata);
- return ret;
+ return ret;
}
static int
-__dht_check_free_space (xlator_t *this, xlator_t *to, xlator_t *from,
- loc_t *loc, struct iatt *stbuf, int flag,
- dht_conf_t *conf, gf_boolean_t *target_changed,
- xlator_t **new_subvol, int *fop_errno)
+__dht_check_free_space(xlator_t *this, xlator_t *to, xlator_t *from, loc_t *loc,
+ struct iatt *stbuf, int flag, dht_conf_t *conf,
+ gf_boolean_t *target_changed, xlator_t **new_subvol,
+ int *fop_errno)
{
- struct statvfs src_statfs = {0,};
- struct statvfs dst_statfs = {0,};
- int ret = -1;
- dict_t *xdata = NULL;
- dht_layout_t *layout = NULL;
- uint64_t src_statfs_blocks = 1;
- uint64_t dst_statfs_blocks = 1;
- double dst_post_availspacepercent = 0;
- double src_post_availspacepercent = 0;
- uint64_t file_blocks = 0;
- uint64_t src_total_blocks = 0;
- uint64_t dst_total_blocks = 0;
-
- xdata = dict_new ();
- if (!xdata) {
- *fop_errno = ENOMEM;
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_NO_MEMORY,
- "failed to allocate dictionary");
- goto out;
- }
-
- ret = dict_set_int8 (xdata, GF_INTERNAL_IGNORE_DEEM_STATFS, 1);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to set "
- GF_INTERNAL_IGNORE_DEEM_STATFS" in dict");
- ret = -1;
- *fop_errno = ENOMEM;
- goto out;
- }
-
- ret = syncop_statfs (from, loc, &src_statfs, xdata, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "failed to get statfs of %s on %s",
- loc->path, from->name);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
-
- ret = syncop_statfs (to, loc, &dst_statfs, xdata, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "failed to get statfs of %s on %s",
- loc->path, to->name);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
-
- gf_msg_debug (this->name, 0, "min_free_disk - %f , block available - "
- "%lu , block size - %lu ", conf->min_free_disk,
- dst_statfs.f_bavail, dst_statfs.f_bsize);
-
- dst_statfs_blocks = dst_statfs.f_bavail *
- (dst_statfs.f_frsize /
- GF_DISK_SECTOR_SIZE);
-
- src_statfs_blocks = src_statfs.f_bavail *
- (src_statfs.f_frsize /
- GF_DISK_SECTOR_SIZE);
-
- dst_total_blocks = dst_statfs.f_blocks *
- (dst_statfs.f_frsize /
- GF_DISK_SECTOR_SIZE);
-
- src_total_blocks = src_statfs.f_blocks *
- (src_statfs.f_frsize /
- GF_DISK_SECTOR_SIZE);
-
- /* if force option is given, do not check for space @ dst.
- * Check only if space is avail for the file */
- if (flag != GF_DHT_MIGRATE_DATA)
- goto check_avail_space;
-
- /* Check:
- During rebalance `migrate-data` - Destination subvol experiences
- a `reduction` in 'blocks' of free space, at the same time source
- subvol gains certain 'blocks' of free space. A valid check is
- necessary here to avoid errorneous move to destination where
- the space could be scantily available.
- With heterogenous brick support, an actual space comparison could
- prevent any files being migrated to newly added bricks if they are
- smaller then the free space available on the existing bricks.
- */
- if (stbuf) {
- if (!conf->use_fallocate) {
- file_blocks = stbuf->ia_size + GF_DISK_SECTOR_SIZE - 1;
- file_blocks /= GF_DISK_SECTOR_SIZE;
-
- if (file_blocks >= dst_statfs_blocks) {
- dst_statfs_blocks = 0;
- } else {
- dst_statfs_blocks -= file_blocks;
- }
- }
-
- src_post_availspacepercent =
- ((src_statfs_blocks + file_blocks) * 100) / src_total_blocks;
-
- dst_post_availspacepercent =
- (dst_statfs_blocks * 100) / dst_total_blocks;
-
- if (dst_post_availspacepercent < src_post_availspacepercent) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "data movement of file "
- "{blocks:%"PRIu64" name:(%s)} would result in "
- "dst node (%s:%"PRIu64") having lower disk "
- "space than the source node (%s:%"PRIu64")"
- ".Skipping file.", stbuf->ia_blocks, loc->path,
- to->name, dst_statfs_blocks, from->name,
- src_statfs_blocks);
-
- /* this is not a 'failure', but we don't want to
- consider this as 'success' too :-/ */
- *fop_errno = ENOSPC;
- ret = 1;
- goto out;
- }
- }
+ struct statvfs src_statfs = {
+ 0,
+ };
+ struct statvfs dst_statfs = {
+ 0,
+ };
+ int ret = -1;
+ dict_t *xdata = NULL;
+ dht_layout_t *layout = NULL;
+ uint64_t src_statfs_blocks = 1;
+ uint64_t dst_statfs_blocks = 1;
+ double dst_post_availspacepercent = 0;
+ double src_post_availspacepercent = 0;
+ uint64_t file_blocks = 0;
+ uint64_t src_total_blocks = 0;
+ uint64_t dst_total_blocks = 0;
+
+ xdata = dict_new();
+ if (!xdata) {
+ *fop_errno = ENOMEM;
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
+ "failed to allocate dictionary");
+ goto out;
+ }
+
+ ret = dict_set_int8(xdata, GF_INTERNAL_IGNORE_DEEM_STATFS, 1);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Failed to set " GF_INTERNAL_IGNORE_DEEM_STATFS " in dict");
+ ret = -1;
+ *fop_errno = ENOMEM;
+ goto out;
+ }
+
+ ret = syncop_statfs(from, loc, &src_statfs, xdata, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "failed to get statfs of %s on %s", loc->path, from->name);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ ret = syncop_statfs(to, loc, &dst_statfs, xdata, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "failed to get statfs of %s on %s", loc->path, to->name);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ gf_msg_debug(this->name, 0,
+ "min_free_disk - %f , block available - %" PRId64
+ ", block size - %lu",
+ conf->min_free_disk, dst_statfs.f_bavail, dst_statfs.f_bsize);
+
+ dst_statfs_blocks = dst_statfs.f_bavail *
+ (dst_statfs.f_frsize / GF_DISK_SECTOR_SIZE);
+
+ src_statfs_blocks = src_statfs.f_bavail *
+ (src_statfs.f_frsize / GF_DISK_SECTOR_SIZE);
+
+ dst_total_blocks = dst_statfs.f_blocks *
+ (dst_statfs.f_frsize / GF_DISK_SECTOR_SIZE);
+
+ src_total_blocks = src_statfs.f_blocks *
+ (src_statfs.f_frsize / GF_DISK_SECTOR_SIZE);
+
+ /* if force option is given, do not check for space @ dst.
+ * Check only if space is avail for the file */
+ if (flag != GF_DHT_MIGRATE_DATA)
+ goto check_avail_space;
+
+ /* Check:
+ During rebalance `migrate-data` - Destination subvol experiences
+ a `reduction` in 'blocks' of free space, at the same time source
+ subvol gains certain 'blocks' of free space. A valid check is
+ necessary here to avoid erroneous move to destination where
+ the space could be scantily available.
+ With heterogeneous brick support, an actual space comparison could
+ prevent any files being migrated to newly added bricks if they are
+ smaller then the free space available on the existing bricks.
+ */
+ if (!conf->use_fallocate) {
+ file_blocks = stbuf->ia_size + GF_DISK_SECTOR_SIZE - 1;
+ file_blocks /= GF_DISK_SECTOR_SIZE;
+
+ if (file_blocks >= dst_statfs_blocks) {
+ dst_statfs_blocks = 0;
+ } else {
+ dst_statfs_blocks -= file_blocks;
+ }
+ }
+
+ src_post_availspacepercent = ((src_statfs_blocks + file_blocks) * 100) /
+ src_total_blocks;
+
+ dst_post_availspacepercent = (dst_statfs_blocks * 100) / dst_total_blocks;
+
+ if (dst_post_availspacepercent < src_post_availspacepercent) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "data movement of file "
+ "{blocks:%" PRIu64
+ " name:(%s)} would result in "
+ "dst node (%s:%" PRIu64
+ ") having lower disk "
+ "space than the source node (%s:%" PRIu64
+ ")"
+ ".Skipping file.",
+ stbuf->ia_blocks, loc->path, to->name, dst_statfs_blocks,
+ from->name, src_statfs_blocks);
+
+ /* this is not a 'failure', but we don't want to
+ consider this as 'success' too :-/ */
+ *fop_errno = ENOSPC;
+ ret = 1;
+ goto out;
+ }
check_avail_space:
- if (conf->disk_unit == 'p' && dst_statfs.f_blocks) {
- dst_post_availspacepercent =
- (dst_statfs_blocks * 100) / dst_total_blocks;
-
- gf_msg_debug (this->name, 0, "file : %s, post_availspacepercent"
- " : %lf f_bavail : %lu min-free-disk: %lf",
- loc->path, dst_post_availspacepercent,
- dst_statfs.f_bavail, conf->min_free_disk);
-
- if (dst_post_availspacepercent < conf->min_free_disk) {
- gf_msg (this->name, GF_LOG_WARNING, 0, 0,
- "Write will cross min-free-disk for "
- "file - %s on subvol - %s. Looking "
- "for new subvol", loc->path, to->name);
-
- goto find_new_subvol;
- } else {
- ret = 0;
- goto out;
- }
+ if (conf->disk_unit == 'p' && dst_statfs.f_blocks) {
+ dst_post_availspacepercent = (dst_statfs_blocks * 100) /
+ dst_total_blocks;
+
+ gf_msg_debug(this->name, 0,
+ "file : %s, post_availspacepercent"
+ " : %lf f_bavail : %" PRIu64 " min-free-disk: %lf",
+ loc->path, dst_post_availspacepercent, dst_statfs.f_bavail,
+ conf->min_free_disk);
+
+ if (dst_post_availspacepercent < conf->min_free_disk) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, 0,
+ "Write will cross min-free-disk for "
+ "file - %s on subvol - %s. Looking "
+ "for new subvol",
+ loc->path, to->name);
+
+ goto find_new_subvol;
+ } else {
+ ret = 0;
+ goto out;
}
+ }
- if (conf->disk_unit != 'p') {
- if ((dst_statfs_blocks * GF_DISK_SECTOR_SIZE) <
- conf->min_free_disk) {
- gf_msg_debug (this->name, 0, "file : %s, destination "
- "frsize: %lu f_bavail : %lu "
- "min-free-disk: %lf", loc->path,
- dst_statfs.f_frsize, dst_statfs.f_bavail,
- conf->min_free_disk);
+ if (conf->disk_unit != 'p') {
+ if ((dst_statfs_blocks * GF_DISK_SECTOR_SIZE) < conf->min_free_disk) {
+ gf_msg_debug(this->name, 0,
+ "file : %s, destination frsize: %lu "
+ "f_bavail : %" PRIu64 " min-free-disk: %lf",
+ loc->path, dst_statfs.f_frsize, dst_statfs.f_bavail,
+ conf->min_free_disk);
- gf_msg (this->name, GF_LOG_WARNING, 0, 0, "write will"
- " cross min-free-disk for file - %s on subvol -"
- " %s. looking for new subvol", loc->path,
- to->name);
+ gf_msg(this->name, GF_LOG_WARNING, 0, 0,
+ "write will"
+ " cross min-free-disk for file - %s on subvol -"
+ " %s. looking for new subvol",
+ loc->path, to->name);
- goto find_new_subvol;
+ goto find_new_subvol;
- } else {
- ret = 0;
- goto out;
- }
+ } else {
+ ret = 0;
+ goto out;
}
+ }
find_new_subvol:
- layout = dht_layout_get (this, loc->parent);
- if (!layout) {
- gf_log (this->name, GF_LOG_ERROR, "Layout is NULL");
- *fop_errno = EINVAL;
- ret = -1;
- goto out;
- }
-
- *new_subvol = dht_subvol_with_free_space_inodes (this, to, from, layout,
- stbuf->ia_size);
- if ((!(*new_subvol)) || (*new_subvol == from)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_SUBVOL_INSUFF_SPACE, "Could not find any subvol"
- " with space accomodating the file - %s. Consider "
- "adding bricks", loc->path);
-
- *target_changed = _gf_false;
- *fop_errno = ENOSPC;
- ret = -1;
- } else {
- gf_msg (this->name, GF_LOG_INFO, 0, 0, "new target found - %s"
- " for file - %s", (*new_subvol)->name, loc->path);
- *target_changed = _gf_true;
- ret = 0;
- }
+ layout = dht_layout_get(this, loc->parent);
+ if (!layout) {
+ gf_log(this->name, GF_LOG_ERROR, "Layout is NULL");
+ *fop_errno = EINVAL;
+ ret = -1;
+ goto out;
+ }
+
+ *new_subvol = dht_subvol_with_free_space_inodes(this, to, from, layout,
+ stbuf->ia_size);
+ if ((!(*new_subvol)) || (*new_subvol == from)) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_SUBVOL_INSUFF_SPACE,
+ "Could not find any subvol"
+ " with space accommodating the file - %s. Consider "
+ "adding bricks",
+ loc->path);
+
+ *target_changed = _gf_false;
+ *fop_errno = ENOSPC;
+ ret = -1;
+ } else {
+ gf_msg(this->name, GF_LOG_INFO, 0, 0,
+ "new target found - %s"
+ " for file - %s",
+ (*new_subvol)->name, loc->path);
+ *target_changed = _gf_true;
+ ret = 0;
+ }
out:
- if (xdata)
- dict_unref (xdata);
- return ret;
+ if (xdata)
+ dict_unref(xdata);
+ return ret;
}
static int
-__dht_rebalance_migrate_data (xlator_t *this, gf_defrag_info_t *defrag,
- xlator_t *from, xlator_t *to, fd_t *src,
- fd_t *dst, uint64_t ia_size, int hole_exists,
- int *fop_errno)
+__dht_rebalance_migrate_data(xlator_t *this, gf_defrag_info_t *defrag,
+ xlator_t *from, xlator_t *to, fd_t *src, fd_t *dst,
+ uint64_t ia_size, int hole_exists, int *fop_errno)
{
- int ret = 0;
- int count = 0;
- off_t offset = 0;
- struct iovec *vector = NULL;
- struct iobref *iobref = NULL;
- uint64_t total = 0;
- size_t read_size = 0;
- dict_t *xdata = NULL;
- dht_conf_t *conf = NULL;
-
- conf = this->private;
- /* if file size is '0', no need to enter this loop */
- while (total < ia_size) {
- read_size = (((ia_size - total) > DHT_REBALANCE_BLKSIZE) ?
- DHT_REBALANCE_BLKSIZE : (ia_size - total));
-
- ret = syncop_readv (from, src, read_size,
- offset, 0, &vector, &count, &iobref, NULL,
- NULL);
- if (!ret || (ret < 0)) {
- *fop_errno = -ret;
- break;
- }
-
- if (hole_exists) {
- ret = dht_write_with_holes (to, dst, vector, count,
- ret, offset, iobref,
- fop_errno);
- } else {
- if (!conf->force_migration &&
- !dht_is_tier_xlator (this)) {
- xdata = dict_new ();
- if (!xdata) {
- gf_msg ("dht", GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "insufficient memory");
- ret = -1;
- *fop_errno = ENOMEM;
- break;
- }
-
- /* Fail this write and abort rebalance if we
- * detect a write from client since migration of
- * this file started. This is done to avoid
- * potential data corruption due to out of order
- * writes from rebalance and client to the same
- * region (as compared between src and dst
- * files). See
- * https://github.com/gluster/glusterfs/issues/308
- * for more details.
- */
- ret = dict_set_int32 (xdata,
- GF_AVOID_OVERWRITE, 1);
- if (ret) {
- gf_msg ("dht", GF_LOG_ERROR, 0,
- ENOMEM, "failed to set dict");
- ret = -1;
- *fop_errno = ENOMEM;
- break;
- }
-
- }
-
- ret = syncop_writev (to, dst, vector, count,
- offset, iobref, 0, xdata, NULL);
- if (ret < 0) {
- *fop_errno = -ret;
- }
- }
+ int ret = 0;
+ int count = 0;
+ off_t offset = 0;
+ off_t data_offset = 0;
+ off_t hole_offset = 0;
+ struct iovec *vector = NULL;
+ struct iobref *iobref = NULL;
+ uint64_t total = 0;
+ size_t read_size = 0;
+ size_t data_block_size = 0;
+ dict_t *xdata = NULL;
+ dht_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ /* if file size is '0', no need to enter this loop */
+ while (total < ia_size) {
+ /* This is a regular file - read it sequentially */
+ if (!hole_exists) {
+ read_size = (((ia_size - total) > DHT_REBALANCE_BLKSIZE)
+ ? DHT_REBALANCE_BLKSIZE
+ : (ia_size - total));
+ } else {
+ /* This is a sparse file - read only the data segments in the file
+ */
+
+ /* If the previous data block is fully copied, find the next data
+ * segment
+ * starting at the offset of the last read and written byte, */
+ if (data_block_size <= 0) {
+ ret = syncop_seek(from, src, offset, GF_SEEK_DATA, NULL,
+ &data_offset);
+ if (ret) {
+ if (ret == -ENXIO)
+ ret = 0; /* No more data segments */
+ else
+ *fop_errno = -ret; /* Error occurred */
- if ((defrag && defrag->cmd == GF_DEFRAG_CMD_START_TIER) &&
- (gf_defrag_get_pause_state (&defrag->tier_conf) != TIER_RUNNING)) {
- gf_msg ("tier", GF_LOG_INFO, 0,
- DHT_MSG_TIER_PAUSED,
- "Migrate file paused");
- ret = -1;
+ break;
}
- if (ret < 0) {
+ /* If the position of the current data segment is greater than
+ * the position of the next hole, find the next hole in order to
+ * calculate the length of the new data segment */
+ if (data_offset > hole_offset) {
+ /* Starting at the offset of the last data segment, find the
+ * next hole */
+ ret = syncop_seek(from, src, data_offset, GF_SEEK_HOLE,
+ NULL, &hole_offset);
+ if (ret) {
+ /* If an error occurred here it's a real error because
+ * if the seek for a data segment was successful then
+ * necessarily another hole must exist (EOF is a hole)
+ */
+ *fop_errno = -ret;
break;
- }
-
- offset += ret;
- total += ret;
-
- GF_FREE (vector);
- if (iobref)
- iobref_unref (iobref);
- iobref = NULL;
- vector = NULL;
- }
- if (iobref)
- iobref_unref (iobref);
- GF_FREE (vector);
+ }
- if (ret >= 0)
- ret = 0;
- else
+ /* Calculate the total size of the current data block */
+ data_block_size = hole_offset - data_offset;
+ }
+ } else {
+ /* There is still data in the current segment, move the
+ * data_offset to the position of the last written byte */
+ data_offset = offset;
+ }
+
+ /* Calculate how much data needs to be read and written. If the data
+ * segment's length is bigger than DHT_REBALANCE_BLKSIZE, read and
+ * write DHT_REBALANCE_BLKSIZE data length and the rest in the
+ * next iteration(s) */
+ read_size = ((data_block_size > DHT_REBALANCE_BLKSIZE)
+ ? DHT_REBALANCE_BLKSIZE
+ : data_block_size);
+
+ /* Calculate the remaining size of the data block - maybe there's no
+ * need to seek for data in the next iteration */
+ data_block_size -= read_size;
+
+ /* Set offset to the offset of the data segment so read and write
+ * will have the correct position */
+ offset = data_offset;
+ }
+
+ ret = syncop_readv(from, src, read_size, offset, 0, &vector, &count,
+ &iobref, NULL, NULL, NULL);
+
+ if (!ret || (ret < 0)) {
+ if (!ret) {
+ /* File was probably truncated*/
ret = -1;
-
- if (xdata) {
- dict_unref (xdata);
+ *fop_errno = ENOSPC;
+ } else {
+ *fop_errno = -ret;
+ }
+ break;
}
- return ret;
-}
-
-
-static int
-__dht_rebalance_open_src_file (xlator_t *this, xlator_t *from, xlator_t *to, loc_t *loc,
- struct iatt *stbuf, fd_t **src_fd,
- gf_boolean_t *clean_src, int *fop_errno)
-{
-
- int ret = 0;
- fd_t *fd = NULL;
- dict_t *dict = NULL;
- struct iatt iatt = {0,};
- dht_conf_t *conf = NULL;
-
- conf = this->private;
-
- *clean_src = _gf_false;
+ if (!conf->force_migration) {
+ if (!xdata) {
+ xdata = dict_new();
+ if (!xdata) {
+ gf_msg("dht", GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "insufficient memory");
+ ret = -1;
+ *fop_errno = ENOMEM;
+ break;
+ }
- fd = fd_create (loc->inode, DHT_REBALANCE_PID);
- if (!fd) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: fd create failed (source)", loc->path);
- *fop_errno = ENOMEM;
- ret = -1;
- goto out;
+ /* Fail this write and abort rebalance if we
+ * detect a write from client since migration of
+ * this file started. This is done to avoid
+ * potential data corruption due to out of order
+ * writes from rebalance and client to the same
+ * region (as compared between src and dst
+ * files). See
+ * https://github.com/gluster/glusterfs/issues/308
+ * for more details.
+ */
+ ret = dict_set_int32_sizen(xdata, GF_AVOID_OVERWRITE, 1);
+ if (ret) {
+ gf_msg("dht", GF_LOG_ERROR, 0, ENOMEM,
+ "failed to set dict");
+ ret = -1;
+ *fop_errno = ENOMEM;
+ break;
+ }
+ }
}
- ret = syncop_open (from, loc, O_RDWR, fd, NULL, NULL);
+ ret = syncop_writev(to, dst, vector, count, offset, iobref, 0, NULL,
+ NULL, xdata, NULL);
if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "failed to open file %s on %s",
- loc->path, from->name);
- *fop_errno = -ret;
- ret = -1;
- goto out;
+ *fop_errno = -ret;
+ break;
}
- fd_bind (fd);
-
- if (src_fd)
- *src_fd = fd;
+ offset += ret;
+ total += ret;
+ GF_FREE(vector);
+ if (iobref)
+ iobref_unref(iobref);
+ iobref = NULL;
+ vector = NULL;
+ }
+ if (iobref)
+ iobref_unref(iobref);
+ GF_FREE(vector);
+
+ if (ret >= 0)
+ ret = 0;
+ else
ret = -1;
- dict = dict_new ();
- if (!dict) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: Could not allocate memory for dict", loc->path);
- *fop_errno = ENOMEM;
- ret = -1;
- goto out;
- }
- ret = dict_set_str (dict, conf->link_xattr_name, to->name);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "failed to set xattr in dict for %s (linkto:%s)",
- loc->path, to->name);
- *fop_errno = ENOMEM;
- ret = -1;
- goto out;
- }
+ if (xdata) {
+ dict_unref(xdata);
+ }
- /* Once the migration starts, the source should have 'linkto' key set
- to show which is the target, so other clients can work around it */
- ret = syncop_setxattr (from, loc, dict, 0, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "failed to set xattr on %s in %s",
- loc->path, from->name);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
+ return ret;
+}
- /* Reset source mode/xattr if migration fails*/
- *clean_src = _gf_true;
+static int
+__dht_rebalance_open_src_file(xlator_t *this, xlator_t *from, xlator_t *to,
+ loc_t *loc, struct iatt *stbuf, fd_t **src_fd,
+ gf_boolean_t *clean_src, int *fop_errno)
+{
+ int ret = 0;
+ fd_t *fd = NULL;
+ dict_t *dict = NULL;
+ struct iatt iatt = {
+ 0,
+ };
+ dht_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ *clean_src = _gf_false;
+
+ fd = fd_create(loc->inode, DHT_REBALANCE_PID);
+ if (!fd) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: fd create failed (source)", loc->path);
+ *fop_errno = ENOMEM;
+ ret = -1;
+ goto out;
+ }
+
+ ret = syncop_open(from, loc, O_RDWR, fd, NULL, NULL);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "failed to open file %s on %s", loc->path, from->name);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
- /* mode should be (+S+T) to indicate migration is in progress */
- iatt.ia_prot = stbuf->ia_prot;
- iatt.ia_type = stbuf->ia_type;
- iatt.ia_prot.sticky = 1;
- iatt.ia_prot.sgid = 1;
+ fd_bind(fd);
- ret = syncop_setattr (from, loc, &iatt, GF_SET_ATTR_MODE, NULL, NULL,
- NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "failed to set mode on %s in %s",
- loc->path, from->name);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
+ if (src_fd)
+ *src_fd = fd;
- /* success */
- ret = 0;
+ ret = -1;
+ dict = dict_new();
+ if (!dict) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: Could not allocate memory for dict", loc->path);
+ *fop_errno = ENOMEM;
+ ret = -1;
+ goto out;
+ }
+
+ ret = dict_set_str(dict, conf->link_xattr_name, to->name);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to set xattr in dict for %s (linkto:%s)", loc->path,
+ to->name);
+ *fop_errno = ENOMEM;
+ ret = -1;
+ goto out;
+ }
+
+ /* Once the migration starts, the source should have 'linkto' key set
+ to show which is the target, so other clients can work around it */
+ ret = syncop_setxattr(from, loc, dict, 0, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "failed to set xattr on %s in %s", loc->path, from->name);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ /* Reset source mode/xattr if migration fails*/
+ *clean_src = _gf_true;
+
+ /* mode should be (+S+T) to indicate migration is in progress */
+ iatt.ia_prot = stbuf->ia_prot;
+ iatt.ia_type = stbuf->ia_type;
+ iatt.ia_prot.sticky = 1;
+ iatt.ia_prot.sgid = 1;
+
+ ret = syncop_setattr(from, loc, &iatt, GF_SET_ATTR_MODE, NULL, NULL, NULL,
+ NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "failed to set mode on %s in %s", loc->path, from->name);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ /* success */
+ ret = 0;
out:
- if (dict)
- dict_unref (dict);
+ if (dict)
+ dict_unref(dict);
- return ret;
+ return ret;
}
int
-migrate_special_files (xlator_t *this, xlator_t *from, xlator_t *to, loc_t *loc,
- struct iatt *buf, int *fop_errno)
+migrate_special_files(xlator_t *this, xlator_t *from, xlator_t *to, loc_t *loc,
+ struct iatt *buf, int *fop_errno)
{
- int ret = -1;
- dict_t *rsp_dict = NULL;
- dict_t *dict = NULL;
- char *link = NULL;
- struct iatt stbuf = {0,};
- dht_conf_t *conf = this->private;
-
- dict = dict_new ();
- if (!dict) {
- *fop_errno = ENOMEM;
- ret = -1;
- goto out;
- }
- ret = dict_set_int32 (dict, conf->link_xattr_name, 256);
- if (ret) {
- *fop_errno = ENOMEM;
- ret = -1;
- gf_log (this->name, GF_LOG_ERROR,
- "%s: failed to set 'linkto' key in dict", loc->path);
- goto out;
- }
-
- /* check in the destination if the file is link file */
- ret = syncop_lookup (to, loc, &stbuf, NULL, dict, &rsp_dict);
- if ((ret < 0) && (-ret != ENOENT)) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: lookup failed",
- loc->path);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
-
- /* we no more require this key */
- dict_del (dict, conf->link_xattr_name);
-
- /* file exists in target node, only if it is 'linkfile' its valid,
- otherwise, error out */
- if (!ret) {
- if (!check_is_linkfile (loc->inode, &stbuf, rsp_dict,
- conf->link_xattr_name)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: file exists in destination", loc->path);
- *fop_errno = EINVAL;
- ret = -1;
- goto out;
- }
-
- /* as file is linkfile, delete it */
- ret = syncop_unlink (to, loc, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: failed to delete the linkfile",
- loc->path);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
- }
-
- /* Set the gfid of the source file in dict */
- ret = dict_set_gfuuid (dict, "gfid-req", buf->ia_gfid, true);
+ int ret = -1;
+ dict_t *rsp_dict = NULL;
+ dict_t *dict = NULL;
+ char *link = NULL;
+ struct iatt stbuf = {
+ 0,
+ };
+ dht_conf_t *conf = this->private;
+
+ dict = dict_new();
+ if (!dict) {
+ *fop_errno = ENOMEM;
+ ret = -1;
+ goto out;
+ }
+ ret = dict_set_int32(dict, conf->link_xattr_name, 256);
+ if (ret) {
+ *fop_errno = ENOMEM;
+ ret = -1;
+ gf_log(this->name, GF_LOG_ERROR,
+ "%s: failed to set 'linkto' key in dict", loc->path);
+ goto out;
+ }
+
+ /* check in the destination if the file is link file */
+ ret = syncop_lookup(to, loc, &stbuf, NULL, dict, &rsp_dict);
+ if ((ret < 0) && (-ret != ENOENT)) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: lookup failed", loc->path);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ /* we no more require this key */
+ dict_del(dict, conf->link_xattr_name);
+
+ /* file exists in target node, only if it is 'linkfile' its valid,
+ otherwise, error out */
+ if (!ret) {
+ if (!check_is_linkfile(loc->inode, &stbuf, rsp_dict,
+ conf->link_xattr_name)) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: file exists in destination", loc->path);
+ *fop_errno = EINVAL;
+ ret = -1;
+ goto out;
+ }
+
+ /* as file is linkfile, delete it */
+ ret = syncop_unlink(to, loc, NULL, NULL);
if (ret) {
- *fop_errno = ENOMEM;
- ret = -1;
- gf_log (this->name, GF_LOG_ERROR,
- "%s: failed to set gfid in dict for create", loc->path);
- goto out;
- }
-
- /* Create the file in target */
- if (IA_ISLNK (buf->ia_type)) {
- /* Handle symlinks separately */
- ret = syncop_readlink (from, loc, &link, buf->ia_size, NULL,
- NULL);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: readlink on symlink failed",
- loc->path);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
-
- ret = syncop_symlink (to, loc, link, 0, dict, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: creating symlink failed",
- loc->path);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
-
- goto done;
+ gf_msg(this->name, GF_LOG_WARNING, -ret,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: failed to delete the linkfile", loc->path);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+ }
+
+ /* Set the gfid of the source file in dict */
+ ret = dict_set_gfuuid(dict, "gfid-req", buf->ia_gfid, true);
+ if (ret) {
+ *fop_errno = ENOMEM;
+ ret = -1;
+ gf_log(this->name, GF_LOG_ERROR,
+ "%s: failed to set gfid in dict for create", loc->path);
+ goto out;
+ }
+
+ /* Create the file in target */
+ if (IA_ISLNK(buf->ia_type)) {
+ /* Handle symlinks separately */
+ ret = syncop_readlink(from, loc, &link, buf->ia_size, NULL, NULL);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: readlink on symlink failed", loc->path);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
}
- ret = syncop_mknod (to, loc, st_mode_from_ia (buf->ia_prot,
- buf->ia_type),
- makedev (ia_major (buf->ia_rdev),
- ia_minor (buf->ia_rdev)), 0, dict, NULL);
+ ret = syncop_symlink(to, loc, link, 0, dict, NULL);
if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: mknod failed",
- loc->path);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
+ gf_msg(this->name, GF_LOG_WARNING, -ret,
+ DHT_MSG_MIGRATE_FILE_FAILED, "%s: creating symlink failed",
+ loc->path);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ goto done;
+ }
+
+ ret = syncop_mknod(to, loc, st_mode_from_ia(buf->ia_prot, buf->ia_type),
+ makedev(ia_major(buf->ia_rdev), ia_minor(buf->ia_rdev)),
+ 0, dict, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: mknod failed", loc->path);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
done:
- ret = syncop_setattr (to, loc, buf,
- (GF_SET_ATTR_MTIME |
- GF_SET_ATTR_UID | GF_SET_ATTR_GID |
- GF_SET_ATTR_MODE), NULL, NULL, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: failed to perform setattr on %s",
- loc->path, to->name);
- *fop_errno = -ret;
- }
-
- ret = syncop_unlink (from, loc, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: unlink failed",
- loc->path);
- *fop_errno = -ret;
- ret = -1;
- }
+ ret = syncop_setattr(to, loc, buf,
+ (GF_SET_ATTR_MTIME | GF_SET_ATTR_UID |
+ GF_SET_ATTR_GID | GF_SET_ATTR_MODE),
+ NULL, NULL, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: failed to perform setattr on %s", loc->path, to->name);
+ *fop_errno = -ret;
+ }
+
+ ret = syncop_unlink(from, loc, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: unlink failed", loc->path);
+ *fop_errno = -ret;
+ ret = -1;
+ }
out:
- GF_FREE (link);
- if (dict)
- dict_unref (dict);
+ GF_FREE(link);
+ if (dict)
+ dict_unref(dict);
- if (rsp_dict)
- dict_unref (rsp_dict);
+ if (rsp_dict)
+ dict_unref(rsp_dict);
- return ret;
+ return ret;
}
static int
-__dht_migration_cleanup_src_file (xlator_t *this, loc_t *loc, fd_t *fd,
- xlator_t *from, ia_prot_t *src_ia_prot)
+__dht_migration_cleanup_src_file(xlator_t *this, loc_t *loc, fd_t *fd,
+ xlator_t *from, ia_prot_t *src_ia_prot)
{
- int ret = -1;
- dht_conf_t *conf = NULL;
- struct iatt new_stbuf = {0,};
-
- if (!this || !fd || !from || !src_ia_prot) {
- goto out;
- }
-
- conf = this->private;
-
- /*Revert source mode and xattr changes*/
- ret = syncop_fstat (from, fd, &new_stbuf, NULL, NULL);
- if (ret < 0) {
- /* Failed to get the stat info */
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file cleanup failed: failed to fstat "
- "file %s on %s ", loc->path, from->name);
- ret = -1;
- goto out;
- }
-
-
- /* Remove the sticky bit and sgid bit set, reset it to 0*/
- if (!src_ia_prot->sticky)
- new_stbuf.ia_prot.sticky = 0;
+ int ret = -1;
+ dht_conf_t *conf = NULL;
+ struct iatt new_stbuf = {
+ 0,
+ };
+
+ if (!this || !fd || !from || !src_ia_prot) {
+ goto out;
+ }
+
+ conf = this->private;
+
+ /*Revert source mode and xattr changes*/
+ ret = syncop_fstat(from, fd, &new_stbuf, NULL, NULL);
+ if (ret < 0) {
+ /* Failed to get the stat info */
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file cleanup failed: failed to fstat "
+ "file %s on %s ",
+ loc->path, from->name);
+ ret = -1;
+ goto out;
+ }
- if (!src_ia_prot->sgid)
- new_stbuf.ia_prot.sgid = 0;
+ /* Remove the sticky bit and sgid bit set, reset it to 0*/
+ if (!src_ia_prot->sticky)
+ new_stbuf.ia_prot.sticky = 0;
- ret = syncop_fsetattr (from, fd, &new_stbuf,
- (GF_SET_ATTR_GID | GF_SET_ATTR_MODE),
- NULL, NULL, NULL, NULL);
+ if (!src_ia_prot->sgid)
+ new_stbuf.ia_prot.sgid = 0;
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file cleanup failed:"
- "%s: failed to perform fsetattr on %s ",
- loc->path, from->name);
- ret = -1;
- goto out;
- }
+ ret = syncop_fsetattr(from, fd, &new_stbuf,
+ (GF_SET_ATTR_GID | GF_SET_ATTR_MODE), NULL, NULL,
+ NULL, NULL);
- ret = syncop_fremovexattr (from, fd, conf->link_xattr_name, 0, NULL);
- if (ret) {
- gf_log (this->name, GF_LOG_WARNING,
- "%s: failed to remove linkto xattr on %s (%s)",
- loc->path, from->name, strerror (-ret));
- ret = -1;
- goto out;
- }
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file cleanup failed:"
+ "%s: failed to perform fsetattr on %s ",
+ loc->path, from->name);
+ ret = -1;
+ goto out;
+ }
+
+ ret = syncop_fremovexattr(from, fd, conf->link_xattr_name, 0, NULL);
+ if (ret) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "%s: failed to remove linkto xattr on %s (%s)", loc->path,
+ from->name, strerror(-ret));
+ ret = -1;
+ goto out;
+ }
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
-
-
/*
return values:
@@ -1583,1041 +1469,1066 @@ out:
1 : not a failure, but we can't migrate data as of now
*/
int
-dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
- int flag, int *fop_errno)
+dht_migrate_file(xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
+ int flag, int *fop_errno)
{
- int ret = -1;
- struct iatt new_stbuf = {0,};
- struct iatt stbuf = {0,};
- struct iatt empty_iatt = {0,};
- ia_prot_t src_ia_prot = {0,};
- fd_t *src_fd = NULL;
- fd_t *dst_fd = NULL;
- dict_t *dict = NULL;
- dict_t *xattr = NULL;
- dict_t *xattr_rsp = NULL;
- int file_has_holes = 0;
- dht_conf_t *conf = this->private;
- int rcvd_enoent_from_src = 0;
- struct gf_flock flock = {0, };
- struct gf_flock plock = {0, };
- loc_t tmp_loc = {0, };
- gf_boolean_t locked = _gf_false;
- gf_boolean_t p_locked = _gf_false;
- int lk_ret = -1;
- gf_defrag_info_t *defrag = NULL;
- gf_boolean_t clean_src = _gf_false;
- gf_boolean_t clean_dst = _gf_false;
- int log_level = GF_LOG_INFO;
- gf_boolean_t delete_src_linkto = _gf_true;
- lock_migration_info_t locklist;
- dict_t *meta_dict = NULL;
- gf_boolean_t meta_locked = _gf_false;
- gf_boolean_t target_changed = _gf_false;
- xlator_t *new_target = NULL;
- xlator_t *old_target = NULL;
- fd_t *linkto_fd = NULL;
-
-
- if (from == to) {
- gf_msg_debug (this->name, 0, "destination and source are same. file %s"
- " might have migrated already", loc->path);
- ret = 0;
- goto out;
- }
-
- /* If defrag is NULL, it should be assumed that migration is triggered
- * from client */
- defrag = conf->defrag;
-
- /* migration of files from clients is restricted to non-tiered clients
- * for now */
- if (!defrag && dht_is_tier_xlator (this)) {
- ret = ENOTSUP;
- goto out;
- }
-
- if (defrag && defrag->tier_conf.is_tier)
- log_level = GF_LOG_TRACE;
+ int ret = -1;
+ struct iatt new_stbuf = {
+ 0,
+ };
+ struct iatt stbuf = {
+ 0,
+ };
+ struct iatt empty_iatt = {
+ 0,
+ };
+ ia_prot_t src_ia_prot = {
+ 0,
+ };
+ fd_t *src_fd = NULL;
+ fd_t *dst_fd = NULL;
+ dict_t *dict = NULL;
+ dict_t *xattr = NULL;
+ dict_t *xattr_rsp = NULL;
+ int file_has_holes = 0;
+ dht_conf_t *conf = this->private;
+ int rcvd_enoent_from_src = 0;
+ struct gf_flock flock = {
+ 0,
+ };
+ struct gf_flock plock = {
+ 0,
+ };
+ loc_t tmp_loc = {
+ 0,
+ };
+ loc_t parent_loc = {
+ 0,
+ };
+ gf_boolean_t inodelk_locked = _gf_false;
+ gf_boolean_t entrylk_locked = _gf_false;
+ gf_boolean_t p_locked = _gf_false;
+ int lk_ret = -1;
+ gf_defrag_info_t *defrag = NULL;
+ gf_boolean_t clean_src = _gf_false;
+ gf_boolean_t clean_dst = _gf_false;
+ int log_level = GF_LOG_INFO;
+ gf_boolean_t delete_src_linkto = _gf_true;
+ lock_migration_info_t locklist;
+ dict_t *meta_dict = NULL;
+ gf_boolean_t meta_locked = _gf_false;
+ gf_boolean_t target_changed = _gf_false;
+ xlator_t *new_target = NULL;
+ xlator_t *old_target = NULL;
+ xlator_t *hashed_subvol = NULL;
+ fd_t *linkto_fd = NULL;
+ dict_t *xdata = NULL;
+
+ if (from == to) {
+ gf_msg_debug(this->name, 0,
+ "destination and source are same. file %s"
+ " might have migrated already",
+ loc->path);
+ ret = 0;
+ goto out;
+ }
- gf_log (this->name,
- log_level, "%s: attempting to move from %s to %s",
- loc->path, from->name, to->name);
+ gf_log(this->name, log_level, "%s: attempting to move from %s to %s",
+ loc->path, from->name, to->name);
- dict = dict_new ();
- if (!dict) {
- ret = -1;
- *fop_errno = ENOMEM;
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
- "Could not allocate memory for dict");
- goto out;
- }
- ret = dict_set_int32 (dict, conf->link_xattr_name, 256);
+ dict = dict_new();
+ if (!dict) {
+ ret = -1;
+ *fop_errno = ENOMEM;
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
+ "Could not allocate memory for dict");
+ goto out;
+ }
+ ret = dict_set_int32(dict, conf->link_xattr_name, 256);
+ if (ret) {
+ *fop_errno = ENOMEM;
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed:"
+ "%s: failed to set 'linkto' key in dict",
+ loc->path);
+ goto out;
+ }
+
+ /* Do not migrate file in case lock migration is not enabled on the
+ * volume*/
+ if (!conf->lock_migration_enabled) {
+ ret = dict_set_int32(dict, GLUSTERFS_POSIXLK_COUNT, sizeof(int32_t));
if (ret) {
- *fop_errno = ENOMEM;
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed:"
- "%s: failed to set 'linkto' key in dict", loc->path);
- goto out;
- }
-
- /* Do not migrate file in case lock migration is not enabled on the
- * volume*/
- if (!conf->lock_migration_enabled) {
- ret = dict_set_int32 (dict,
- GLUSTERFS_POSIXLK_COUNT, sizeof(int32_t));
- if (ret) {
- *fop_errno = ENOMEM;
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed: %s: failed to "
- "set "GLUSTERFS_POSIXLK_COUNT" key in dict",
- loc->path);
- goto out;
- }
- } else {
- gf_msg (this->name, GF_LOG_INFO, 0, 0, "locks will be migrated"
- " for file: %s", loc->path);
- }
-
- flock.l_type = F_WRLCK;
-
- tmp_loc.inode = inode_ref (loc->inode);
- gf_uuid_copy (tmp_loc.gfid, loc->gfid);
- tmp_loc.path = gf_strdup(loc->path);
-
- /* this inodelk happens with flock.owner being zero. But to synchronize
- * hardlink migration we need to have different lkowner for each migration
- * Filed a bug here: https://bugzilla.redhat.com/show_bug.cgi?id=1468202 to
- * track the fix for this. Currently synclock takes care of synchronizing
- * hardlink migration. Once this bug is fixed we can avoid taking synclock */
- ret = syncop_inodelk (from, DHT_FILE_MIGRATE_DOMAIN, &tmp_loc, F_SETLKW,
- &flock, NULL, NULL);
- if (ret < 0) {
- *fop_errno = -ret;
- ret = -1;
- gf_msg (this->name, GF_LOG_WARNING, *fop_errno,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "migrate file failed: "
- "%s: failed to lock file on %s",
- loc->path, from->name);
- goto out;
- }
-
- locked = _gf_true;
-
- /* Phase 1 - Data migration is in progress from now on */
- ret = syncop_lookup (from, loc, &stbuf, NULL, dict, &xattr_rsp);
+ *fop_errno = ENOMEM;
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed: %s: failed to "
+ "set " GLUSTERFS_POSIXLK_COUNT " key in dict",
+ loc->path);
+ goto out;
+ }
+ } else {
+ gf_msg(this->name, GF_LOG_INFO, 0, 0,
+ "locks will be migrated"
+ " for file: %s",
+ loc->path);
+ }
+
+ /* The file is locked to prevent a rename during a migration. Renames
+ * and migrations on the file at the same time can lead to data loss.
+ */
+
+ ret = dht_build_parent_loc(this, &parent_loc, loc, fop_errno);
+ if (ret < 0) {
+ ret = -1;
+ gf_msg(this->name, GF_LOG_WARNING, *fop_errno,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: failed to build parent loc, which is needed to "
+ "acquire entrylk to synchronize with renames on this "
+ "path. Skipping migration",
+ loc->path);
+ goto out;
+ }
+
+ hashed_subvol = dht_subvol_get_hashed(this, loc);
+ if (hashed_subvol == NULL) {
+ ret = -1;
+ gf_msg(this->name, GF_LOG_WARNING, EINVAL, DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: cannot find hashed subvol which is needed to "
+ "synchronize with renames on this path. "
+ "Skipping migration",
+ loc->path);
+ goto out;
+ }
+
+ flock.l_type = F_WRLCK;
+
+ tmp_loc.inode = inode_ref(loc->inode);
+ gf_uuid_copy(tmp_loc.gfid, loc->gfid);
+ tmp_loc.path = gf_strdup(loc->path);
+
+ /* this inodelk happens with flock.owner being zero. But to synchronize
+ * hardlink migration we need to have different lkowner for each migration
+ * Filed a bug here: https://bugzilla.redhat.com/show_bug.cgi?id=1468202 to
+ * track the fix for this. Currently synclock takes care of synchronizing
+ * hardlink migration. Once this bug is fixed we can avoid taking synclock
+ */
+ ret = syncop_inodelk(from, DHT_FILE_MIGRATE_DOMAIN, &tmp_loc, F_SETLKW,
+ &flock, NULL, NULL);
+ if (ret < 0) {
+ *fop_errno = -ret;
+ ret = -1;
+ gf_msg(this->name, GF_LOG_WARNING, *fop_errno,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "migrate file failed: "
+ "%s: failed to lock file on %s",
+ loc->path, from->name);
+ goto out;
+ }
+
+ inodelk_locked = _gf_true;
+
+ /* dht_rename has changed to use entrylk on hashed subvol for
+ * synchronization. So, rebalance too has to acquire an entrylk on
+ * hashed subvol.
+ */
+ ret = syncop_entrylk(hashed_subvol, DHT_ENTRY_SYNC_DOMAIN, &parent_loc,
+ loc->name, ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL, NULL);
+ if (ret < 0) {
+ *fop_errno = -ret;
+ ret = -1;
+ gf_msg(this->name, GF_LOG_WARNING, *fop_errno,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: failed to acquire entrylk on subvol %s", loc->path,
+ hashed_subvol->name);
+ goto out;
+ }
+
+ entrylk_locked = _gf_true;
+
+ /* Phase 1 - Data migration is in progress from now on */
+ ret = syncop_lookup(from, loc, &stbuf, NULL, dict, &xattr_rsp);
+ if (ret) {
+ *fop_errno = -ret;
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, *fop_errno,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed:"
+ "%s: lookup failed on %s",
+ loc->path, from->name);
+ goto out;
+ }
+
+ /* preserve source mode, so set the same to the destination */
+ src_ia_prot = stbuf.ia_prot;
+
+ /* Check if file can be migrated */
+ ret = __is_file_migratable(this, loc, &stbuf, xattr_rsp, flag, defrag, conf,
+ fop_errno);
+ if (ret) {
+ if (ret == HARDLINK_MIG_INPROGRESS)
+ ret = 0;
+ goto out;
+ }
+
+ /* Take care of the special files */
+ if (!IA_ISREG(stbuf.ia_type)) {
+ /* Special files */
+ ret = migrate_special_files(this, from, to, loc, &stbuf, fop_errno);
+ goto out;
+ }
+
+ /* Try to preserve 'holes' while migrating data */
+ if (stbuf.ia_size > (stbuf.ia_blocks * GF_DISK_SECTOR_SIZE))
+ file_has_holes = 1;
+
+ /* create the destination, with required modes/xattr */
+ ret = __dht_rebalance_create_dst_file(this, to, from, loc, &stbuf, &dst_fd,
+ fop_errno, file_has_holes);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "Create dst failed"
+ " on - %s for file - %s",
+ to->name, loc->path);
+ goto out;
+ }
+
+ clean_dst = _gf_true;
+
+ ret = __dht_check_free_space(this, to, from, loc, &stbuf, flag, conf,
+ &target_changed, &new_target, fop_errno);
+ if (target_changed) {
+ /* Can't handle for hardlinks. Marking this as failure */
+ if (flag == GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS || stbuf.ia_nlink > 1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_SUBVOL_INSUFF_SPACE,
+ "Exiting migration for"
+ " file - %s. flag - %d, stbuf.ia_nlink - %d",
+ loc->path, flag, stbuf.ia_nlink);
+ ret = -1;
+ goto out;
+ }
+
+ ret = syncop_ftruncate(to, dst_fd, 0, NULL, NULL, NULL, NULL);
if (ret) {
- *fop_errno = -ret;
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, *fop_errno,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed:"
- "%s: lookup failed on %s",
- loc->path, from->name);
- goto out;
+ gf_log(this->name, GF_LOG_WARNING,
+ "%s: failed to perform truncate on %s (%s)", loc->path,
+ to->name, strerror(-ret));
}
- /* preserve source mode, so set the same to the destination */
- src_ia_prot = stbuf.ia_prot;
+ syncop_close(dst_fd);
+ dst_fd = NULL;
- /* Check if file can be migrated */
- ret = __is_file_migratable (this, loc, &stbuf, xattr_rsp, flag, defrag, conf,
- fop_errno);
- if (ret) {
- if (ret == HARDLINK_MIG_INPROGRESS)
- ret = 0;
- goto out;
- }
+ old_target = to;
+ to = new_target;
- /* Take care of the special files */
- if (!IA_ISREG (stbuf.ia_type)) {
- /* Special files */
- ret = migrate_special_files (this, from, to, loc, &stbuf,
- fop_errno);
- goto out;
- }
+ clean_dst = _gf_false;
- /* create the destination, with required modes/xattr */
- ret = __dht_rebalance_create_dst_file (this, to, from, loc, &stbuf,
- &dst_fd, fop_errno);
+ /* if the file migration is successful to this new target, then
+ * update the xattr on the old destination to point the new
+ * destination. We need to do update this only post migration
+ * as in case of failure the linkto needs to point to the source
+ * subvol */
+ ret = __dht_rebalance_create_dst_file(
+ this, to, from, loc, &stbuf, &dst_fd, fop_errno, file_has_holes);
if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "Create dst failed"
- " on - %s for file - %s", to->name, loc->path);
- goto out;
- }
-
- clean_dst = _gf_true;
-
- ret = __dht_check_free_space (this, to, from, loc, &stbuf, flag, conf,
- &target_changed, &new_target, fop_errno);
- if (target_changed) {
- /* Can't handle for hardlinks. Marking this as failure */
- if (flag == GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS || stbuf.ia_nlink > 1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_SUBVOL_INSUFF_SPACE, "Exiting migration for"
- " file - %s. flag - %d, stbuf.ia_nlink - %d",
- loc->path, flag, stbuf.ia_nlink);
- ret = -1;
- goto out;
- }
+ gf_log(this->name, GF_LOG_ERROR,
+ "Create dst failed"
+ " on - %s for file - %s",
+ to->name, loc->path);
+ goto out;
+ } else {
+ gf_msg(this->name, GF_LOG_INFO, 0, 0,
+ "destination for file "
+ "- %s is changed to - %s",
+ loc->path, to->name);
+ clean_dst = _gf_true;
+ }
+ }
+
+ if (ret) {
+ goto out;
+ }
+
+ /* Open the source, and also update mode/xattr */
+ ret = __dht_rebalance_open_src_file(this, from, to, loc, &stbuf, &src_fd,
+ &clean_src, fop_errno);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed: failed to open %s on %s", loc->path,
+ from->name);
+ goto out;
+ }
+
+ /* TODO: move all xattr related operations to fd based operations */
+ ret = syncop_listxattr(from, loc, &xattr, NULL, NULL);
+ if (ret < 0) {
+ *fop_errno = -ret;
+ gf_msg(this->name, GF_LOG_WARNING, *fop_errno,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed:"
+ "%s: failed to get xattr from %s",
+ loc->path, from->name);
+ ret = -1;
+ goto out;
+ }
+
+ /* Copying posix acls to the linkto file messes up the permissions*/
+ dht_strip_out_acls(xattr);
+
+ /* Remove the linkto xattr as we don't want to overwrite the value
+ * set on the dst.
+ */
+ dict_del(xattr, conf->link_xattr_name);
+
+ /* We need to error out if this fails as having the wrong shard xattrs
+ * set on the dst could cause data corruption
+ */
+ ret = syncop_fsetxattr(to, dst_fd, xattr, 0, NULL, NULL);
+ if (ret < 0) {
+ *fop_errno = -ret;
+ gf_msg(this->name, GF_LOG_WARNING, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: failed to set xattr on %s", loc->path, to->name);
+ ret = -1;
+ goto out;
+ }
+ if (xattr_rsp) {
+ /* we no more require this key */
+ dict_del(dict, conf->link_xattr_name);
+ dict_unref(xattr_rsp);
+ }
+
+ ret = syncop_fstat(from, src_fd, &stbuf, dict, &xattr_rsp);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed:failed to lookup %s on %s ", loc->path,
+ from->name);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ /* Check again if file has hardlink */
+ ret = __check_file_has_hardlink(this, loc, &stbuf, xattr_rsp, flag, defrag,
+ conf, fop_errno);
+ if (ret) {
+ if (ret == HARDLINK_MIG_INPROGRESS)
+ ret = 0;
+ goto out;
+ }
+
+ ret = __dht_rebalance_migrate_data(this, defrag, from, to, src_fd, dst_fd,
+ stbuf.ia_size, file_has_holes,
+ fop_errno);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed: %s: failed to migrate data", loc->path);
- ret = syncop_ftruncate (to, dst_fd, 0, NULL, NULL);
- if (ret) {
- gf_log (this->name, GF_LOG_WARNING,
- "%s: failed to perform truncate on %s (%s)",
- loc->path, to->name, strerror (-ret));
- }
+ ret = -1;
+ goto out;
+ }
+
+ /* TODO: Sync the locks */
+
+ xdata = dict_new();
+ if (!xdata || dict_set_int8(xdata, "last-fsync", 1)) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "%s: failed to set last-fsync flag on "
+ "%s (%s)",
+ loc->path, to->name, strerror(ENOMEM));
+ }
+
+ ret = syncop_fsync(to, dst_fd, 0, NULL, NULL, xdata, NULL);
+ if (ret) {
+ gf_log(this->name, GF_LOG_WARNING, "%s: failed to fsync on %s (%s)",
+ loc->path, to->name, strerror(-ret));
+ *fop_errno = -ret;
+ }
+
+ /* Phase 2 - Data-Migration Complete, Housekeeping updates pending */
+
+ ret = syncop_fstat(from, src_fd, &new_stbuf, NULL, NULL);
+ if (ret < 0) {
+ /* Failed to get the stat info */
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed: failed to fstat file %s on %s ", loc->path,
+ from->name);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ }
- syncop_close (dst_fd);
- dst_fd = NULL;
+ /* Lock the entire source file to prevent clients from taking a
+ lock on it as dht_lk does not handle file migration.
- old_target = to;
- to = new_target;
+ This still leaves a small window where conflicting locks can
+ be granted to different clients. If client1 requests a blocking
+ lock on the src file, it will be granted after the migrating
+ process releases its lock. If client2 requests a lock on the dst
+ data file, it will also be granted, but all FOPs will be redirected
+ to the dst data file.
+ */
- clean_dst = _gf_false;
+ /* Take meta lock */
+ if (conf->lock_migration_enabled) {
+ meta_dict = dict_new();
+ if (!meta_dict) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "dict_new failed");
- /* if the file migration is successful to this new target, then
- * update the xattr on the old destination to point the new
- * destination. We need to do update this only post migration
- * as in case of failure the linkto needs to point to the source
- * subvol */
- ret = __dht_rebalance_create_dst_file (this, to, from, loc, &stbuf,
- &dst_fd, fop_errno);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "Create dst failed"
- " on - %s for file - %s", to->name, loc->path);
- goto out;
- } else {
- gf_msg (this->name, GF_LOG_INFO, 0, 0, "destination for file "
- "- %s is changed to - %s", loc->path, to->name);
- clean_dst = _gf_true;
- }
+ *fop_errno = ENOMEM;
+ ret = -1;
+ goto out;
}
+ ret = dict_set_str(meta_dict, GLUSTERFS_INTERNAL_FOP_KEY, "yes");
if (ret) {
- goto out;
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary value: key = %s,"
+ " path = %s",
+ GLUSTERFS_INTERNAL_FOP_KEY, loc->path);
+ *fop_errno = ENOMEM;
+ ret = -1;
+ goto out;
}
- /* Open the source, and also update mode/xattr */
- ret = __dht_rebalance_open_src_file (this, from, to, loc, &stbuf, &src_fd,
- &clean_src, fop_errno);
+ ret = dict_set_int32(meta_dict, GF_META_LOCK_KEY, 1);
if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed: failed to open %s on %s",
- loc->path, from->name);
- goto out;
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Trace dict_set failed");
+ *fop_errno = ENOMEM;
+ ret = -1;
+ goto out;
}
- /* TODO: move all xattr related operations to fd based operations */
- ret = syncop_listxattr (from, loc, &xattr, NULL, NULL);
- if (ret < 0) {
- *fop_errno = -ret;
- gf_msg (this->name, GF_LOG_WARNING, *fop_errno,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed:"
- "%s: failed to get xattr from %s",
- loc->path, from->name);
- ret = -1;
- goto out;
- }
-
- /* Copying posix acls to the linkto file messes up the permissions*/
- dht_strip_out_acls (xattr);
-
- /* Remove the linkto xattr as we don't want to overwrite the value
- * set on the dst.
- */
- dict_del (xattr, conf->link_xattr_name);
-
- /* We need to error out if this fails as having the wrong shard xattrs
- * set on the dst could cause data corruption
- */
- ret = syncop_fsetxattr (to, dst_fd, xattr, 0, NULL, NULL);
- if (ret < 0) {
- *fop_errno = -ret;
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: failed to set xattr on %s",
- loc->path, to->name);
- ret = -1;
- goto out;
- }
-
- if (xattr_rsp) {
- /* we no more require this key */
- dict_del (dict, conf->link_xattr_name);
- dict_unref (xattr_rsp);
- }
-
- ret = syncop_fstat (from, src_fd, &stbuf, dict, &xattr_rsp);
+ ret = syncop_setxattr(from, loc, meta_dict, 0, NULL, NULL);
if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed:failed to lookup %s on %s ",
- loc->path, from->name);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Trace syncop_setxattr metalock failed");
- /* Check again if file has hardlink */
- ret = __check_file_has_hardlink (this, loc, &stbuf, xattr_rsp,
- flag, defrag, conf, fop_errno);
- if (ret) {
- if (ret == HARDLINK_MIG_INPROGRESS)
- ret = 0;
- goto out;
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ } else {
+ meta_locked = _gf_true;
}
- /* Try to preserve 'holes' while migrating data */
- if (stbuf.ia_size > (stbuf.ia_blocks * GF_DISK_SECTOR_SIZE))
- file_has_holes = 1;
+ }
+ if (!conf->lock_migration_enabled) {
+ plock.l_type = F_WRLCK;
+ plock.l_start = 0;
+ plock.l_len = 0;
+ plock.l_whence = SEEK_SET;
- ret = __dht_rebalance_migrate_data (this, defrag, from, to,
- src_fd, dst_fd, stbuf.ia_size,
- file_has_holes, fop_errno);
+ ret = syncop_lk(from, src_fd, F_SETLK, &plock, NULL, NULL);
if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed: %s: failed to migrate data",
- loc->path);
-
- ret = -1;
- goto out;
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed:"
+ "%s: Failed to lock on %s",
+ loc->path, from->name);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
}
- /* TODO: Sync the locks */
+ p_locked = _gf_true;
- ret = syncop_fsync (to, dst_fd, 0, NULL, NULL);
- if (ret) {
- gf_log (this->name, GF_LOG_WARNING,
- "%s: failed to fsync on %s (%s)",
- loc->path, to->name, strerror (-ret));
- *fop_errno = -ret;
- }
+ } else {
+ INIT_LIST_HEAD(&locklist.list);
+ ret = syncop_getactivelk(from, loc, &locklist, NULL, NULL);
+ if (ret == 0) {
+ gf_log(this->name, GF_LOG_INFO, "No active locks on:%s", loc->path);
- /* Phase 2 - Data-Migration Complete, Housekeeping updates pending */
+ } else if (ret > 0) {
+ ret = syncop_setactivelk(to, loc, &locklist, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret,
+ DHT_MSG_LOCK_MIGRATION_FAILED, "write lock failed on:%s",
+ loc->path);
- ret = syncop_fstat (from, src_fd, &new_stbuf, NULL, NULL);
- if (ret < 0) {
- /* Failed to get the stat info */
- gf_msg ( this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed: failed to fstat file %s on %s ",
- loc->path, from->name);
*fop_errno = -ret;
ret = -1;
- goto out;
- }
-
- /* Lock the entire source file to prevent clients from taking a
- lock on it as dht_lk does not handle file migration.
-
- This still leaves a small window where conflicting locks can
- be granted to different clients. If client1 requests a blocking
- lock on the src file, it will be granted after the migrating
- process releases its lock. If client2 requests a lock on the dst
- data file, it will also be granted, but all FOPs will be redirected
- to the dst data file.
- */
-
- /* Take meta lock */
-
- if (conf->lock_migration_enabled) {
- meta_dict = dict_new ();
- if (!meta_dict) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "dict_new failed");
-
- *fop_errno = ENOMEM;
- ret = -1;
- goto out;
- }
-
- ret = dict_set_str (meta_dict, GLUSTERFS_INTERNAL_FOP_KEY, "yes");
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value: key = %s,"
- " path = %s", GLUSTERFS_INTERNAL_FOP_KEY,
- loc->path);
- *fop_errno = ENOMEM;
- ret = -1;
- goto out;
- }
-
- ret = dict_set_int32 (meta_dict, GF_META_LOCK_KEY, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Trace dict_set failed");
- *fop_errno = ENOMEM;
- ret = -1;
- goto out;
- }
-
- ret = syncop_setxattr (from, loc, meta_dict, 0, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Trace syncop_setxattr metalock failed");
-
- *fop_errno = -ret;
- ret = -1;
- goto out;
- } else {
- meta_locked = _gf_true;
- }
- }
-
- if (!conf->lock_migration_enabled) {
- plock.l_type = F_WRLCK;
- plock.l_start = 0;
- plock.l_len = 0;
- plock.l_whence = SEEK_SET;
-
- ret = syncop_lk (from, src_fd, F_SETLK, &plock, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed:"
- "%s: Failed to lock on %s",
- loc->path, from->name);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
-
- p_locked = _gf_true;
-
+ goto metaunlock;
+ }
} else {
-
- INIT_LIST_HEAD (&locklist.list);
-
- ret = syncop_getactivelk (from, loc, &locklist, NULL, NULL);
- if (ret == 0) {
- gf_log (this->name, GF_LOG_INFO, "No active locks on:%s"
- , loc->path);
-
- } else if (ret > 0) {
-
- ret = syncop_setactivelk (to, loc, &locklist, NULL,
- NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_LOCK_MIGRATION_FAILED,
- "write lock failed on:%s", loc->path);
-
- *fop_errno = -ret;
- ret = -1;
- goto metaunlock;
- }
- } else {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_LOCK_MIGRATION_FAILED,
- "getactivelk failed for file: %s", loc->path);
- *fop_errno = -ret;
- }
- }
-
-
- /* source would have both sticky bit and sgid bit set, reset it to 0,
- and set the source permission on destination, if it was not set
- prior to setting rebalance-modes in source */
- if (!src_ia_prot.sticky)
- new_stbuf.ia_prot.sticky = 0;
-
- if (!src_ia_prot.sgid)
- new_stbuf.ia_prot.sgid = 0;
-
- /* TODO: if the source actually had sticky bit, or sgid bit set,
- we are not handling it */
-
- ret = syncop_fsetattr (to, dst_fd, &new_stbuf,
- (GF_SET_ATTR_UID | GF_SET_ATTR_GID |
- GF_SET_ATTR_MODE), NULL, NULL, NULL, NULL);
+ gf_msg(this->name, GF_LOG_ERROR, -ret,
+ DHT_MSG_LOCK_MIGRATION_FAILED,
+ "getactivelk failed for file: %s", loc->path);
+ *fop_errno = -ret;
+ }
+ }
+
+ /* source would have both sticky bit and sgid bit set, reset it to 0,
+ and set the source permission on destination, if it was not set
+ prior to setting rebalance-modes in source */
+ if (!src_ia_prot.sticky)
+ new_stbuf.ia_prot.sticky = 0;
+
+ if (!src_ia_prot.sgid)
+ new_stbuf.ia_prot.sgid = 0;
+
+ /* TODO: if the source actually had sticky bit, or sgid bit set,
+ we are not handling it */
+
+ ret = syncop_fsetattr(
+ to, dst_fd, &new_stbuf,
+ (GF_SET_ATTR_UID | GF_SET_ATTR_GID | GF_SET_ATTR_MODE), NULL, NULL,
+ NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed:"
+ "%s: failed to perform setattr on %s ",
+ loc->path, to->name);
+ *fop_errno = -ret;
+ ret = -1;
+ goto metaunlock;
+ }
+
+ /* Because 'futimes' is not portable */
+ ret = syncop_setattr(to, loc, &new_stbuf,
+ (GF_SET_ATTR_MTIME | GF_SET_ATTR_ATIME), NULL, NULL,
+ NULL, NULL);
+ if (ret) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "%s: failed to perform setattr on %s ", loc->path, to->name);
+ *fop_errno = -ret;
+ }
+
+ if (target_changed) {
+ dict_del(dict, GLUSTERFS_POSIXLK_COUNT);
+ ret = dict_set_str(dict, conf->link_xattr_name, to->name);
if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed:"
- "%s: failed to perform setattr on %s ",
- loc->path, to->name);
- *fop_errno = -ret;
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to set xattr in dict for %s (linkto:%s)", loc->path,
+ to->name);
+ *fop_errno = ENOMEM;
+ ret = -1;
+ goto out;
+ }
+
+ ret = syncop_setxattr(old_target, loc, dict, 0, NULL, NULL);
+ if (ret && -ret != ESTALE && -ret != ENOENT) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "failed to set xattr on %s in %s", loc->path,
+ old_target->name);
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
+ } else if (-ret == ESTALE || -ret == ENOENT) {
+ /* The failure ESTALE indicates that the linkto
+ * file on the hashed subvol might have been deleted.
+ * In this case will create a linkto file with new target
+ * as linkto xattr value*/
+ linkto_fd = fd_create(loc->inode, DHT_REBALANCE_PID);
+ if (!linkto_fd) {
+ gf_msg(this->name, GF_LOG_ERROR, errno,
+ DHT_MSG_MIGRATE_FILE_FAILED, "%s: fd create failed",
+ loc->path);
+ *fop_errno = ENOMEM;
ret = -1;
- goto metaunlock;
- }
-
- /* Because 'futimes' is not portable */
- ret = syncop_setattr (to, loc, &new_stbuf,
- (GF_SET_ATTR_MTIME | GF_SET_ATTR_ATIME),
- NULL, NULL, NULL, NULL);
- if (ret) {
- gf_log (this->name, GF_LOG_WARNING,
- "%s: failed to perform setattr on %s ",
- loc->path, to->name);
+ goto out;
+ }
+ ret = syncop_create(old_target, loc, O_RDWR, DHT_LINKFILE_MODE,
+ linkto_fd, NULL, dict, NULL);
+ if (ret != 0 && -ret != EEXIST && -ret != ESTALE) {
*fop_errno = -ret;
- }
-
- if (target_changed) {
- dict_del (dict, conf->link_xattr_name);
- dict_del (dict, GLUSTERFS_POSIXLK_COUNT);
- ret = dict_set_str (dict, conf->link_xattr_name, to->name);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "failed to set xattr in dict for %s (linkto:%s)",
- loc->path, to->name);
- *fop_errno = ENOMEM;
- ret = -1;
- goto out;
- }
-
- ret = syncop_setxattr (old_target, loc, dict, 0, NULL, NULL);
- if (ret && -ret != ESTALE && -ret != ENOENT) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "failed to set xattr on %s in %s",
- loc->path, old_target->name);
- *fop_errno = -ret;
- ret = -1;
- goto out;
- } else if (-ret == ESTALE || -ret == ENOENT) {
- /* The failure ESTALE indicates that the linkto
- * file on the hashed subvol might have been deleted.
- * In this case will create a linkto file with new target
- * as linkto xattr value*/
- linkto_fd = fd_create (loc->inode, DHT_REBALANCE_PID);
- if (!linkto_fd) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: fd create failed", loc->path);
- *fop_errno = ENOMEM;
- ret = -1;
- goto out;
- }
- ret = syncop_create (old_target, loc, O_RDWR,
- DHT_LINKFILE_MODE, linkto_fd,
- NULL, dict, NULL);
- if (ret != 0 && -ret != EEXIST && -ret != ESTALE) {
- *fop_errno = -ret;
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "failed to create linkto file on %s in %s",
- loc->path, old_target->name);
- goto out;
- } else if (ret == 0) {
- ret = syncop_fsetattr (old_target, linkto_fd, &stbuf,
- (GF_SET_ATTR_UID | GF_SET_ATTR_GID),
- NULL, NULL, NULL, NULL);
- if (ret < 0) {
- *fop_errno = -ret;
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "chown failed for %s on %s",
- loc->path, old_target->name);
- }
- }
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, -ret,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "failed to create linkto file on %s in %s", loc->path,
+ old_target->name);
+ goto out;
+ } else if (ret == 0) {
+ ret = syncop_fsetattr(old_target, linkto_fd, &stbuf,
+ (GF_SET_ATTR_UID | GF_SET_ATTR_GID), NULL,
+ NULL, NULL, NULL);
+ if (ret < 0) {
+ *fop_errno = -ret;
+ gf_msg(this->name, GF_LOG_ERROR, -ret,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "chown failed for %s on %s", loc->path,
+ old_target->name);
}
+ }
+ }
+ }
+
+ clean_dst = _gf_false;
+
+ /* Posix acls are not set on DHT linkto files as part of the initial
+ * initial xattrs set on the dst file, so these need
+ * to be set on the dst file after the linkto attrs are removed.
+ * TODO: Optimize this.
+ */
+ if (xattr) {
+ dict_unref(xattr);
+ xattr = NULL;
+ }
+
+ /* Set only the Posix ACLs this time */
+ ret = syncop_getxattr(from, loc, &xattr, POSIX_ACL_ACCESS_XATTR, NULL,
+ NULL);
+ if (ret < 0) {
+ if ((-ret != ENODATA) && (-ret != ENOATTR)) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed:"
+ "%s: failed to get xattr from %s",
+ loc->path, from->name);
+ *fop_errno = -ret;
+ }
+ } else {
+ ret = syncop_setxattr(to, loc, xattr, 0, NULL, NULL);
+ if (ret < 0) {
+ /* Potential problem here where Posix ACLs will
+ * not be set on the target file */
+
+ gf_msg(this->name, GF_LOG_WARNING, -ret,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed:"
+ "%s: failed to set xattr on %s",
+ loc->path, to->name);
+ *fop_errno = -ret;
+ }
+ }
+
+ /* The src file is being unlinked after this so we don't need
+ to clean it up */
+ clean_src = _gf_false;
+
+ /* Make the source as a linkfile first before deleting it */
+ empty_iatt.ia_prot.sticky = 1;
+ ret = syncop_fsetattr(from, src_fd, &empty_iatt, GF_SET_ATTR_MODE, NULL,
+ NULL, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed:"
+ "%s: failed to perform setattr on %s ",
+ loc->path, from->name);
+ *fop_errno = -ret;
+ ret = -1;
+ goto metaunlock;
+ }
+
+ /* Free up the data blocks on the source node, as the whole
+ file is migrated */
+ ret = syncop_ftruncate(from, src_fd, 0, NULL, NULL, NULL, NULL);
+ if (ret) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "%s: failed to perform truncate on %s (%s)", loc->path,
+ from->name, strerror(-ret));
+ *fop_errno = -ret;
+ }
+
+ /* remove the 'linkto' xattr from the destination */
+ ret = syncop_fremovexattr(to, dst_fd, conf->link_xattr_name, 0, NULL);
+ if (ret) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "%s: failed to perform removexattr on %s (%s)", loc->path,
+ to->name, strerror(-ret));
+ *fop_errno = -ret;
+ }
+
+ /* Do a stat and check the gfid before unlink */
+
+ /*
+ * Cached file changes its state from non-linkto to linkto file after
+ * migrating data. If lookup from any other mount-point is performed,
+ * converted-linkto-cached file will be treated as a stale and will be
+ * unlinked. But by this time, file is already migrated. So further
+ * failure because of ENOENT should not be treated as error
+ */
+
+ ret = syncop_stat(from, loc, &empty_iatt, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: failed to do a stat on %s", loc->path, from->name);
+
+ if (-ret != ENOENT) {
+ *fop_errno = -ret;
+ ret = -1;
+ goto metaunlock;
+ }
+
+ rcvd_enoent_from_src = 1;
+ }
+
+ if ((gf_uuid_compare(empty_iatt.ia_gfid, loc->gfid) == 0) &&
+ (!rcvd_enoent_from_src) && delete_src_linkto) {
+ /* take out the source from namespace */
+ ret = syncop_unlink(from, loc, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: failed to perform unlink on %s", loc->path, from->name);
+ *fop_errno = -ret;
+ ret = -1;
+ goto metaunlock;
}
+ }
- clean_dst = _gf_false;
-
- /* Posix acls are not set on DHT linkto files as part of the initial
- * initial xattrs set on the dst file, so these need
- * to be set on the dst file after the linkto attrs are removed.
- * TODO: Optimize this.
- */
- if (xattr) {
- dict_unref (xattr);
- xattr = NULL;
- }
+ ret = syncop_lookup(this, loc, NULL, NULL, NULL, NULL);
+ if (ret) {
+ gf_msg_debug(this->name, -ret,
+ "%s: failed to lookup the file on subvolumes", loc->path);
+ *fop_errno = -ret;
+ }
- /* Set only the Posix ACLs this time */
- ret = syncop_getxattr (from, loc, &xattr, POSIX_ACL_ACCESS_XATTR,
- NULL, NULL);
- if (ret < 0) {
- if ((-ret != ENODATA) && (-ret != ENOATTR)) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed:"
- "%s: failed to get xattr from %s",
- loc->path, from->name);
- *fop_errno = -ret;
- }
- } else {
- ret = syncop_setxattr (to, loc, xattr, 0, NULL, NULL);
- if (ret < 0) {
- /* Potential problem here where Posix ACLs will
- * not be set on the target file */
-
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed:"
- "%s: failed to set xattr on %s",
- loc->path, to->name);
- *fop_errno = -ret;
- }
- }
+ gf_msg(this->name, log_level, 0, DHT_MSG_MIGRATE_FILE_COMPLETE,
+ "completed migration of %s from subvolume %s to %s", loc->path,
+ from->name, to->name);
- /* store size of previous migrated file */
- if (defrag && defrag->tier_conf.is_tier) {
- if (from != TIER_HASHED_SUBVOL) {
- defrag->tier_conf.st_last_promoted_size = stbuf.ia_size;
- } else {
- /* Don't delete the linkto file on the hashed subvol */
- delete_src_linkto = _gf_false;
- defrag->tier_conf.st_last_demoted_size = stbuf.ia_size;
- }
- }
+ ret = 0;
- /* The src file is being unlinked after this so we don't need
- to clean it up */
- clean_src = _gf_false;
+metaunlock:
- /* Make the source as a linkfile first before deleting it */
- empty_iatt.ia_prot.sticky = 1;
- ret = syncop_fsetattr (from, src_fd, &empty_iatt,
- GF_SET_ATTR_MODE, NULL, NULL, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed:"
- "%s: failed to perform setattr on %s ",
- loc->path, from->name);
- *fop_errno = -ret;
- ret = -1;
- goto metaunlock;
- }
+ if (conf->lock_migration_enabled && meta_locked) {
+ dict_del(meta_dict, GF_META_LOCK_KEY);
- /* Free up the data blocks on the source node, as the whole
- file is migrated */
- ret = syncop_ftruncate (from, src_fd, 0, NULL, NULL);
+ ret = dict_set_int32(meta_dict, GF_META_UNLOCK_KEY, 1);
if (ret) {
- gf_log (this->name, GF_LOG_WARNING,
- "%s: failed to perform truncate on %s (%s)",
- loc->path, from->name, strerror (-ret));
- *fop_errno = -ret;
- }
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Trace dict_set failed");
- /* remove the 'linkto' xattr from the destination */
- ret = syncop_fremovexattr (to, dst_fd, conf->link_xattr_name, 0, NULL);
- if (ret) {
- gf_log (this->name, GF_LOG_WARNING,
- "%s: failed to perform removexattr on %s (%s)",
- loc->path, to->name, strerror (-ret));
- *fop_errno = -ret;
+ *fop_errno = ENOMEM;
+ ret = -1;
+ goto out;
}
- /* Do a stat and check the gfid before unlink */
-
- /*
- * Cached file changes its state from non-linkto to linkto file after
- * migrating data. If lookup from any other mount-point is performed,
- * converted-linkto-cached file will be treated as a stale and will be
- * unlinked. But by this time, file is already migrated. So further
- * failure because of ENOENT should not be treated as error
- */
+ if (clean_dst == _gf_false)
+ ret = dict_set_int32(meta_dict, "status", 1);
+ else
+ ret = dict_set_int32(meta_dict, "status", 0);
- ret = syncop_stat (from, loc, &empty_iatt, NULL, NULL);
if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: failed to do a stat on %s",
- loc->path, from->name);
-
- if (-ret != ENOENT) {
- *fop_errno = -ret;
- ret = -1;
- goto metaunlock;
- }
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Trace dict_set failed");
- rcvd_enoent_from_src = 1;
+ *fop_errno = ENOMEM;
+ ret = -1;
+ goto out;
}
-
- if ((gf_uuid_compare (empty_iatt.ia_gfid, loc->gfid) == 0 ) &&
- (!rcvd_enoent_from_src) && delete_src_linkto) {
- /* take out the source from namespace */
- ret = syncop_unlink (from, loc, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: failed to perform unlink on %s",
- loc->path, from->name);
- *fop_errno = -ret;
- ret = -1;
- goto metaunlock;
- }
- }
-
- ret = syncop_lookup (this, loc, NULL, NULL, NULL, NULL);
+ ret = syncop_setxattr(from, loc, meta_dict, 0, NULL, NULL);
if (ret) {
- gf_msg_debug (this->name, -ret,
- "%s: failed to lookup the file on subvolumes",
- loc->path);
- *fop_errno = -ret;
- }
-
- gf_msg (this->name, log_level, 0,
- DHT_MSG_MIGRATE_FILE_COMPLETE,
- "completed migration of %s from subvolume %s to %s",
- loc->path, from->name, to->name);
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Trace syncop_setxattr meta unlock failed");
- ret = 0;
-
-metaunlock:
-
- if (conf->lock_migration_enabled && meta_locked) {
-
- dict_del (meta_dict, GF_META_LOCK_KEY);
-
- ret = dict_set_int32 (meta_dict, GF_META_UNLOCK_KEY, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Trace dict_set failed");
-
- *fop_errno = ENOMEM;
- ret = -1;
- goto out;
- }
-
- if (clean_dst == _gf_false)
- ret = dict_set_int32 (meta_dict, "status", 1);
- else
- ret = dict_set_int32 (meta_dict, "status", 0);
-
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Trace dict_set failed");
-
- *fop_errno = ENOMEM;
- ret = -1;
- goto out;
- }
-
- ret = syncop_setxattr (from, loc, meta_dict, 0, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Trace syncop_setxattr meta unlock failed");
-
- *fop_errno = -ret;
- ret = -1;
- goto out;
- }
+ *fop_errno = -ret;
+ ret = -1;
+ goto out;
}
+ }
out:
- if (clean_src) {
- /* Revert source mode and xattr changes*/
- lk_ret = __dht_migration_cleanup_src_file (this, loc, src_fd,
- from, &src_ia_prot);
- if (lk_ret) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: failed to cleanup source file on %s",
- loc->path, from->name);
- }
- }
-
- /* reset the destination back to 0 */
- if (clean_dst) {
- lk_ret = syncop_ftruncate (to, dst_fd, 0, NULL, NULL);
- if (lk_ret) {
- gf_msg (this->name, GF_LOG_ERROR, -lk_ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed: "
- "%s: failed to reset target size back to 0",
- loc->path);
- }
- }
-
- if (locked) {
- flock.l_type = F_UNLCK;
-
- lk_ret = syncop_inodelk (from, DHT_FILE_MIGRATE_DOMAIN,
- &tmp_loc, F_SETLK, &flock, NULL, NULL);
- if (lk_ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, -lk_ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: failed to unlock file on %s",
- loc->path, from->name);
- }
- }
-
- if (p_locked) {
- plock.l_type = F_UNLCK;
- lk_ret = syncop_lk (from, src_fd, F_SETLK, &plock, NULL, NULL);
-
- if (lk_ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, -lk_ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "%s: failed to unlock file on %s",
- loc->path, from->name);
- }
- }
-
- if (!dht_is_tier_xlator (this)) {
- lk_ret = syncop_removexattr (to, loc,
- GF_PROTECT_FROM_EXTERNAL_WRITES,
- NULL, NULL);
- if (lk_ret && (lk_ret != -ENODATA) && (lk_ret != -ENOATTR)) {
- gf_msg (this->name, GF_LOG_WARNING, -lk_ret, 0,
- "%s: removexattr failed key %s", loc->path,
- GF_PROTECT_FROM_EXTERNAL_WRITES);
- }
- }
-
- if (dict)
- dict_unref (dict);
-
- if (xattr)
- dict_unref (xattr);
- if (xattr_rsp)
- dict_unref (xattr_rsp);
-
- if (dst_fd)
- syncop_close (dst_fd);
- if (src_fd)
- syncop_close (src_fd);
- if (linkto_fd)
- syncop_close (linkto_fd);
-
- loc_wipe (&tmp_loc);
-
- return ret;
+ if (clean_src) {
+ /* Revert source mode and xattr changes*/
+ lk_ret = __dht_migration_cleanup_src_file(this, loc, src_fd, from,
+ &src_ia_prot);
+ if (lk_ret) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: failed to cleanup source file on %s", loc->path,
+ from->name);
+ }
+ }
+
+ /* reset the destination back to 0 */
+ if (clean_dst) {
+ lk_ret = syncop_ftruncate(to, dst_fd, 0, NULL, NULL, NULL, NULL);
+ if (lk_ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -lk_ret,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed: "
+ "%s: failed to reset target size back to 0",
+ loc->path);
+ }
+ }
+
+ if (inodelk_locked) {
+ flock.l_type = F_UNLCK;
+
+ lk_ret = syncop_inodelk(from, DHT_FILE_MIGRATE_DOMAIN, &tmp_loc,
+ F_SETLK, &flock, NULL, NULL);
+ if (lk_ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, -lk_ret,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: failed to unlock file on %s", loc->path, from->name);
+ }
+ }
+
+ if (entrylk_locked) {
+ lk_ret = syncop_entrylk(hashed_subvol, DHT_ENTRY_SYNC_DOMAIN,
+ &parent_loc, loc->name, ENTRYLK_UNLOCK,
+ ENTRYLK_UNLOCK, NULL, NULL);
+ if (lk_ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, -lk_ret,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: failed to unlock entrylk on %s", loc->path,
+ hashed_subvol->name);
+ }
+ }
+
+ if (p_locked) {
+ plock.l_type = F_UNLCK;
+ lk_ret = syncop_lk(from, src_fd, F_SETLK, &plock, NULL, NULL);
+
+ if (lk_ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, -lk_ret,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "%s: failed to unlock file on %s", loc->path, from->name);
+ }
+ }
+
+ lk_ret = syncop_removexattr(to, loc, GF_PROTECT_FROM_EXTERNAL_WRITES, NULL,
+ NULL);
+ if (lk_ret && (lk_ret != -ENODATA) && (lk_ret != -ENOATTR)) {
+ gf_msg(this->name, GF_LOG_WARNING, -lk_ret, 0,
+ "%s: removexattr failed key %s", loc->path,
+ GF_PROTECT_FROM_EXTERNAL_WRITES);
+ }
+
+ if (dict)
+ dict_unref(dict);
+
+ if (xattr)
+ dict_unref(xattr);
+ if (xattr_rsp)
+ dict_unref(xattr_rsp);
+
+ if (dst_fd)
+ syncop_close(dst_fd);
+
+ if (src_fd)
+ syncop_close(src_fd);
+ if (linkto_fd)
+ syncop_close(linkto_fd);
+
+ if (xdata)
+ dict_unref(xdata);
+
+ loc_wipe(&tmp_loc);
+ loc_wipe(&parent_loc);
+
+ return ret;
}
static int
-rebalance_task (void *data)
+rebalance_task(void *data)
{
- int ret = -1;
- dht_local_t *local = NULL;
- call_frame_t *frame = NULL;
- int fop_errno = 0;
+ int ret = -1;
+ dht_local_t *local = NULL;
+ call_frame_t *frame = NULL;
+ int fop_errno = 0;
- frame = data;
+ frame = data;
- local = frame->local;
+ local = frame->local;
- /* This function is 'synchrounous', hence if it returns,
- we are done with the task */
- ret = dht_migrate_file (THIS, &local->loc, local->rebalance.from_subvol,
- local->rebalance.target_node, local->flags,
- &fop_errno);
+ /* This function is 'synchrounous', hence if it returns,
+ we are done with the task */
+ ret = dht_migrate_file(THIS, &local->loc, local->rebalance.from_subvol,
+ local->rebalance.target_node, local->flags,
+ &fop_errno);
- return ret;
+ return ret;
}
static int
-rebalance_task_completion (int op_ret, call_frame_t *sync_frame, void *data)
+rebalance_task_completion(int op_ret, call_frame_t *sync_frame, void *data)
{
- int32_t op_errno = EINVAL;
-
- if (op_ret == -1) {
- /* Failure of migration process, mostly due to write process.
- as we can't preserve the exact errno, lets say there was
- no space to migrate-data
- */
- op_errno = ENOSPC;
- } else if (op_ret == 1) {
- /* migration didn't happen, but is not a failure, let the user
- understand that he doesn't have permission to migrate the
- file.
- */
- op_ret = -1;
- op_errno = EPERM;
- } else if (op_ret != 0) {
- op_errno = -op_ret;
- op_ret = -1;
- }
+ int32_t op_errno = EINVAL;
- DHT_STACK_UNWIND (setxattr, sync_frame, op_ret, op_errno, NULL);
- return 0;
+ if (op_ret == -1) {
+ /* Failure of migration process, mostly due to write process.
+ as we can't preserve the exact errno, lets say there was
+ no space to migrate-data
+ */
+ op_errno = ENOSPC;
+ } else if (op_ret == 1) {
+ /* migration didn't happen, but is not a failure, let the user
+ understand that he doesn't have permission to migrate the
+ file.
+ */
+ op_ret = -1;
+ op_errno = EPERM;
+ } else if (op_ret != 0) {
+ op_errno = -op_ret;
+ op_ret = -1;
+ }
+
+ DHT_STACK_UNWIND(setxattr, sync_frame, op_ret, op_errno, NULL);
+ return 0;
}
int
-dht_start_rebalance_task (xlator_t *this, call_frame_t *frame)
+dht_start_rebalance_task(xlator_t *this, call_frame_t *frame)
{
- int ret = -1;
+ int ret = -1;
- ret = synctask_new (this->ctx->env, rebalance_task,
- rebalance_task_completion,
- frame, frame);
- return ret;
+ ret = synctask_new(this->ctx->env, rebalance_task,
+ rebalance_task_completion, frame, frame);
+ return ret;
}
-
int
-gf_listener_stop (xlator_t *this)
+gf_listener_stop(xlator_t *this)
{
- glusterfs_ctx_t *ctx = NULL;
- cmd_args_t *cmd_args = NULL;
- int ret = 0;
-
- ctx = this->ctx;
- GF_ASSERT (ctx);
- cmd_args = &ctx->cmd_args;
- if (cmd_args->sock_file) {
- ret = sys_unlink (cmd_args->sock_file);
- if (ret && (ENOENT == errno)) {
- ret = 0;
- }
- }
-
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- DHT_MSG_SOCKET_ERROR,
- "Failed to unlink listener "
- "socket %s", cmd_args->sock_file);
- }
- return ret;
+ glusterfs_ctx_t *ctx = NULL;
+ cmd_args_t *cmd_args = NULL;
+ int ret = 0;
+
+ ctx = this->ctx;
+ GF_ASSERT(ctx);
+ cmd_args = &ctx->cmd_args;
+ if (cmd_args->sock_file) {
+ ret = sys_unlink(cmd_args->sock_file);
+ if (ret && (ENOENT == errno)) {
+ ret = 0;
+ }
+ }
+
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, errno, DHT_MSG_SOCKET_ERROR,
+ "Failed to unlink listener "
+ "socket %s",
+ cmd_args->sock_file);
+ }
+ return ret;
}
void
-dht_build_root_inode (xlator_t *this, inode_t **inode)
+dht_build_root_inode(xlator_t *this, inode_t **inode)
{
- inode_table_t *itable = NULL;
- uuid_t root_gfid = {0, };
+ inode_table_t *itable = NULL;
+ static uuid_t root_gfid = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1};
- itable = inode_table_new (0, this);
- if (!itable)
- return;
+ itable = inode_table_new(0, this);
+ if (!itable)
+ return;
- root_gfid[15] = 1;
- *inode = inode_find (itable, root_gfid);
+ *inode = inode_find(itable, root_gfid);
}
void
-dht_build_root_loc (inode_t *inode, loc_t *loc)
+dht_build_root_loc(inode_t *inode, loc_t *loc)
{
- loc->path = "/";
- loc->inode = inode;
- loc->inode->ia_type = IA_IFDIR;
- memset (loc->gfid, 0, 16);
- loc->gfid[15] = 1;
+ loc->path = "/";
+ loc->inode = inode;
+ loc->inode->ia_type = IA_IFDIR;
+ memset(loc->gfid, 0, 16);
+ loc->gfid[15] = 1;
}
/* return values: 1 -> error, bug ignore and continue
0 -> proceed
-1 -> error, handle it */
int32_t
-gf_defrag_handle_migrate_error (int32_t op_errno, gf_defrag_info_t *defrag)
+gf_defrag_handle_migrate_error(int32_t op_errno, gf_defrag_info_t *defrag)
{
- int ret = 0;
- /* if errno is not ENOTCONN, we can still continue
- with rebalance process */
- if (op_errno != ENOTCONN) {
- ret = 1;
- goto out;
- }
-
- if (op_errno == ENOTCONN) {
- /* Most probably mount point went missing (mostly due
- to a brick down), say rebalance failure to user,
- let him restart it if everything is fine */
- defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
- ret = -1;
- goto out;
- }
+ int ret = 0;
+ /* if errno is not ENOTCONN, we can still continue
+ with rebalance process */
+ if (op_errno != ENOTCONN) {
+ ret = 1;
+ goto out;
+ }
+
+ if (op_errno == ENOTCONN) {
+ /* Most probably mount point went missing (mostly due
+ to a brick down), say rebalance failure to user,
+ let him restart it if everything is fine */
+ defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
+ ret = -1;
+ goto out;
+ }
out:
- return ret;
+ return ret;
}
static gf_boolean_t
-gf_defrag_pattern_match (gf_defrag_info_t *defrag, char *name, uint64_t size)
+gf_defrag_pattern_match(gf_defrag_info_t *defrag, char *name, uint64_t size)
{
- gf_defrag_pattern_list_t *trav = NULL;
- gf_boolean_t match = _gf_false;
- gf_boolean_t ret = _gf_false;
+ gf_defrag_pattern_list_t *trav = NULL;
+ gf_boolean_t match = _gf_false;
+ gf_boolean_t ret = _gf_false;
- GF_VALIDATE_OR_GOTO ("dht", defrag, out);
+ GF_VALIDATE_OR_GOTO("dht", defrag, out);
- trav = defrag->defrag_pattern;
- while (trav) {
- if (!fnmatch (trav->path_pattern, name, FNM_NOESCAPE)) {
- match = _gf_true;
- break;
- }
- trav = trav->next;
+ trav = defrag->defrag_pattern;
+ while (trav) {
+ if (!fnmatch(trav->path_pattern, name, FNM_NOESCAPE)) {
+ match = _gf_true;
+ break;
}
+ trav = trav->next;
+ }
- if ((match == _gf_true) && (size >= trav->size))
- ret = _gf_true;
+ if ((match == _gf_true) && (size >= trav->size))
+ ret = _gf_true;
- out:
- return ret;
+out:
+ return ret;
}
-int dht_dfreaddirp_done (dht_dfoffset_ctx_t *offset_var, int cnt) {
-
- int i;
- int result = 1;
+int
+dht_dfreaddirp_done(dht_dfoffset_ctx_t *offset_var, int cnt)
+{
+ int i;
+ int result = 1;
- for (i = 0; i < cnt; i++) {
- if (offset_var[i].readdir_done == 0) {
- result = 0;
- break;
- }
+ for (i = 0; i < cnt; i++) {
+ if (offset_var[i].readdir_done == 0) {
+ result = 0;
+ break;
}
- return result;
+ }
+ return result;
}
-int static
-gf_defrag_ctx_subvols_init (dht_dfoffset_ctx_t *offset_var, xlator_t *this) {
+int static gf_defrag_ctx_subvols_init(dht_dfoffset_ctx_t *offset_var,
+ xlator_t *this)
+{
+ int i;
+ dht_conf_t *conf = NULL;
- int i;
- dht_conf_t *conf = NULL;
+ conf = this->private;
- conf = this->private;
+ if (!conf)
+ return -1;
- if (!conf)
- return -1;
+ for (i = 0; i < conf->local_subvols_cnt; i++) {
+ offset_var[i].this = conf->local_subvols[i];
+ offset_var[i].offset = (off_t)0;
+ offset_var[i].readdir_done = 0;
+ }
- for (i = 0; i < conf->local_subvols_cnt; i++) {
- offset_var[i].this = conf->local_subvols[i];
- offset_var[i].offset = (off_t) 0;
- offset_var[i].readdir_done = 0;
- }
-
- return 0;
+ return 0;
}
-
static int
-dht_get_first_non_null_index (subvol_nodeuuids_info_t *entry)
+dht_get_first_non_null_index(subvol_nodeuuids_info_t *entry)
{
- int i = 0;
- int index = 0;
+ int i = 0;
+ int index = 0;
- for (i = 0; i < entry->count; i++) {
- if (!gf_uuid_is_null (entry->elements[i].uuid)) {
- index = i;
- goto out;
- }
+ for (i = 0; i < entry->count; i++) {
+ if (!gf_uuid_is_null(entry->elements[i].uuid)) {
+ index = i;
+ goto out;
}
+ }
- if (i == entry->count) {
- index = -1;
- }
+ if (i == entry->count) {
+ index = -1;
+ }
out:
- return index;
+ return index;
}
-
/* Return value
* 0 : this node does not migrate the file
* 1 : this node migrates the file
@@ -2627,2779 +2538,2165 @@ out:
* all hardlinks.
*/
-int
-gf_defrag_should_i_migrate (xlator_t *this, int local_subvol_index, uuid_t gfid)
+gf_boolean_t
+gf_defrag_should_i_migrate(xlator_t *this, int local_subvol_index, uuid_t gfid)
{
- int ret = 0;
- int i = local_subvol_index;
- char *str = NULL;
- uint32_t hashval = 0;
- int32_t index = 0;
- dht_conf_t *conf = NULL;
- char buf[UUID_CANONICAL_FORM_LEN + 1] = {0, };
- subvol_nodeuuids_info_t *entry = NULL;
-
-
- conf = this->private;
-
- /* Pure distribute. A subvol in this case
- will be handled by only one node */
-
- entry = &(conf->local_nodeuuids[i]);
- if (entry->count == 1) {
- return 1;
- }
-
- str = uuid_utoa_r (gfid, buf);
- ret = dht_hash_compute (this, 0, str, &hashval);
- if (ret == 0) {
- index = (hashval % entry->count);
- if (entry->elements[index].info
- == REBAL_NODEUUID_MINE) {
- /* Index matches this node's nodeuuid.*/
- ret = 1;
- goto out;
- }
-
- /* Brick down - some other node has to migrate these files*/
- if (gf_uuid_is_null (entry->elements[index].uuid)) {
- /* Fall back to the first non-null index */
- index = dht_get_first_non_null_index (entry);
-
- if (index == -1) {
- /* None of the bricks in the subvol are up.
- * CHILD_DOWN will kill the process soon */
-
- return 0;
- }
-
- if (entry->elements[index].info == REBAL_NODEUUID_MINE) {
- /* Index matches this node's nodeuuid.*/
- ret = 1;
- goto out;
- }
- }
+ gf_boolean_t ret = _gf_false;
+ int i = local_subvol_index;
+ char *str = NULL;
+ uint32_t hashval = 0;
+ int32_t index = 0;
+ dht_conf_t *conf = NULL;
+ char buf[UUID_CANONICAL_FORM_LEN + 1] = {
+ 0,
+ };
+ subvol_nodeuuids_info_t *entry = NULL;
+
+ conf = this->private;
+
+ /* Pure distribute. A subvol in this case
+ will be handled by only one node */
+
+ entry = &(conf->local_nodeuuids[i]);
+ if (entry->count == 1) {
+ return 1;
+ }
+
+ str = uuid_utoa_r(gfid, buf);
+ if (dht_hash_compute(this, 0, str, &hashval) == 0) {
+ index = (hashval % entry->count);
+ if (entry->elements[index].info == REBAL_NODEUUID_MINE) {
+ /* Index matches this node's nodeuuid.*/
+ ret = _gf_true;
+ goto out;
+ }
+
+ /* Brick down - some other node has to migrate these files*/
+ if (gf_uuid_is_null(entry->elements[index].uuid)) {
+ /* Fall back to the first non-null index */
+ index = dht_get_first_non_null_index(entry);
+
+ if (index == -1) {
+ /* None of the bricks in the subvol are up.
+ * CHILD_DOWN will kill the process soon */
+
+ return _gf_false;
+ }
+
+ if (entry->elements[index].info == REBAL_NODEUUID_MINE) {
+ /* Index matches this node's nodeuuid.*/
+ ret = _gf_true;
+ goto out;
+ }
}
+ }
out:
- return ret;
+ return ret;
}
-
int
-gf_defrag_migrate_single_file (void *opaque)
+gf_defrag_migrate_single_file(void *opaque)
{
- xlator_t *this = NULL;
- dht_conf_t *conf = NULL;
- gf_defrag_info_t *defrag = NULL;
- int ret = 0;
- gf_dirent_t *entry = NULL;
- struct timeval start = {0,};
- loc_t entry_loc = {0,};
- loc_t *loc = NULL;
- struct iatt iatt = {0,};
- dict_t *migrate_data = NULL;
- struct timeval end = {0,};
- double elapsed = {0,};
- struct dht_container *rebal_entry = NULL;
- inode_t *inode = NULL;
- xlator_t *hashed_subvol = NULL;
- xlator_t *cached_subvol = NULL;
- call_frame_t *statfs_frame = NULL;
- xlator_t *old_THIS = NULL;
- data_t *tmp = NULL;
- int fop_errno = 0;
- gf_dht_migrate_data_type_t rebal_type = GF_DHT_MIGRATE_DATA;
- char value[MAX_REBAL_TYPE_SIZE] = {0,};
- struct iatt *iatt_ptr = NULL;
- gf_boolean_t update_skippedcount = _gf_true;
- int i = 0;
-
- rebal_entry = (struct dht_container *)opaque;
- if (!rebal_entry) {
- gf_log ("DHT", GF_LOG_ERROR, "rebal_entry is NULL");
- ret = -1;
- goto out;
- }
+ xlator_t *this = NULL;
+ dht_conf_t *conf = NULL;
+ gf_defrag_info_t *defrag = NULL;
+ int ret = 0;
+ gf_dirent_t *entry = NULL;
+ struct timeval start = {
+ 0,
+ };
+ loc_t entry_loc = {
+ 0,
+ };
+ loc_t *loc = NULL;
+ struct iatt iatt = {
+ 0,
+ };
+ dict_t *migrate_data = NULL;
+ struct timeval end = {
+ 0,
+ };
+ double elapsed = {
+ 0,
+ };
+ struct dht_container *rebal_entry = NULL;
+ inode_t *inode = NULL;
+ xlator_t *hashed_subvol = NULL;
+ xlator_t *cached_subvol = NULL;
+ call_frame_t *statfs_frame = NULL;
+ xlator_t *old_THIS = NULL;
+ data_t *tmp = NULL;
+ int fop_errno = 0;
+ gf_dht_migrate_data_type_t rebal_type = GF_DHT_MIGRATE_DATA;
+ char value[MAX_REBAL_TYPE_SIZE] = {
+ 0,
+ };
+ struct iatt *iatt_ptr = NULL;
+ gf_boolean_t update_skippedcount = _gf_true;
+ int i = 0;
+ gf_boolean_t should_i_migrate = 0;
+
+ rebal_entry = (struct dht_container *)opaque;
+ if (!rebal_entry) {
+ gf_log("DHT", GF_LOG_ERROR, "rebal_entry is NULL");
+ ret = -1;
+ goto out;
+ }
- this = rebal_entry->this;
+ this = rebal_entry->this;
- conf = this->private;
+ conf = this->private;
- defrag = conf->defrag;
+ defrag = conf->defrag;
- loc = rebal_entry->parent_loc;
+ loc = rebal_entry->parent_loc;
- migrate_data = rebal_entry->migrate_data;
+ migrate_data = rebal_entry->migrate_data;
- entry = rebal_entry->df_entry;
- iatt_ptr = &entry->d_stat;
+ entry = rebal_entry->df_entry;
+ iatt_ptr = &entry->d_stat;
- if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
- ret = -1;
- goto out;
- }
-
- if (defrag->stats == _gf_true) {
- gettimeofday (&start, NULL);
- }
-
- if (defrag->defrag_pattern &&
- (gf_defrag_pattern_match (defrag, entry->d_name,
- entry->d_stat.ia_size) == _gf_false)) {
- gf_log (this->name, GF_LOG_ERROR, "pattern_match failed");
- goto out;
- }
-
- memset (&entry_loc, 0, sizeof (entry_loc));
+ if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
+ ret = -1;
+ goto out;
+ }
- ret = dht_build_child_loc (this, &entry_loc, loc, entry->d_name);
- if (ret) {
- LOCK (&defrag->lock);
- {
- defrag->total_failures += 1;
- }
- UNLOCK (&defrag->lock);
+ if (defrag->stats == _gf_true) {
+ gettimeofday(&start, NULL);
+ }
- ret = 0;
+ if (defrag->defrag_pattern &&
+ (gf_defrag_pattern_match(defrag, entry->d_name,
+ entry->d_stat.ia_size) == _gf_false)) {
+ gf_log(this->name, GF_LOG_ERROR, "pattern_match failed");
+ goto out;
+ }
- gf_log (this->name, GF_LOG_ERROR, "Child loc build failed");
+ memset(&entry_loc, 0, sizeof(entry_loc));
- goto out;
+ ret = dht_build_child_loc(this, &entry_loc, loc, entry->d_name);
+ if (ret) {
+ LOCK(&defrag->lock);
+ {
+ defrag->total_failures += 1;
}
+ UNLOCK(&defrag->lock);
- if (!gf_defrag_should_i_migrate (this, rebal_entry->local_subvol_index,
- entry->d_stat.ia_gfid)) {
- gf_msg_debug (this->name, 0, "Don't migrate %s ",
- entry_loc.path);
- goto out;
- }
+ ret = 0;
- gf_uuid_copy (entry_loc.gfid, entry->d_stat.ia_gfid);
+ gf_log(this->name, GF_LOG_ERROR, "Child loc build failed");
- gf_uuid_copy (entry_loc.pargfid, loc->gfid);
+ goto out;
+ }
- ret = syncop_lookup (this, &entry_loc, &iatt, NULL, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "Migrate file failed: %s lookup failed",
- entry_loc.path);
- ret = 0;
- goto out;
- }
-
- iatt_ptr = &iatt;
-
- hashed_subvol = dht_subvol_get_hashed (this, &entry_loc);
- if (!hashed_subvol) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_HASHED_SUBVOL_GET_FAILED,
- "Failed to get hashed subvol for %s",
- entry_loc.path);
- ret = 0;
- goto out;
- }
+ should_i_migrate = gf_defrag_should_i_migrate(
+ this, rebal_entry->local_subvol_index, entry->d_stat.ia_gfid);
- cached_subvol = dht_subvol_get_cached (this, entry_loc.inode);
- if (!cached_subvol) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_CACHED_SUBVOL_GET_FAILED,
- "Failed to get cached subvol for %s",
- entry_loc.path);
+ gf_uuid_copy(entry_loc.gfid, entry->d_stat.ia_gfid);
- ret = 0;
- goto out;
- }
+ gf_uuid_copy(entry_loc.pargfid, loc->gfid);
- if (hashed_subvol == cached_subvol) {
- ret = 0;
- goto out;
- }
+ ret = syncop_lookup(this, &entry_loc, &iatt, NULL, NULL, NULL);
- inode = inode_link (entry_loc.inode, entry_loc.parent, entry->d_name, &iatt);
- inode_unref (entry_loc.inode);
- /* use the inode returned by inode_link */
- entry_loc.inode = inode;
-
- old_THIS = THIS;
- THIS = this;
- statfs_frame = create_frame (this, this->ctx->pool);
- if (!statfs_frame) {
- gf_msg (this->name, GF_LOG_ERROR, DHT_MSG_NO_MEMORY, ENOMEM,
- "Insufficient memory. Frame creation failed");
- ret = -1;
- goto out;
+ if (!should_i_migrate) {
+ /* this node isn't supposed to migrate the file. suppressing any
+ * potential error from lookup as this file is under migration by
+ * another node */
+ if (ret) {
+ gf_msg_debug(this->name, -ret,
+ "Ignoring lookup failure: node isn't migrating %s",
+ entry_loc.path);
+ ret = 0;
+ }
+ gf_msg_debug(this->name, 0, "Don't migrate %s ", entry_loc.path);
+ goto out;
+ }
+
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_MIGRATE_FILE_FAILED,
+ "Migrate file failed: %s lookup failed", entry_loc.path);
+
+ /* Increase failure count only for remove-brick op, so that
+ * user is warned to check the removed-brick for any files left
+ * unmigrated
+ */
+ if (conf->decommission_subvols_cnt) {
+ LOCK(&defrag->lock);
+ {
+ defrag->total_failures += 1;
+ }
+ UNLOCK(&defrag->lock);
}
- /* async statfs information for honoring min-free-disk */
- dht_get_du_info (statfs_frame, this, loc);
- THIS = old_THIS;
+ ret = 0;
+ goto out;
+ }
- tmp = dict_get (migrate_data, GF_XATTR_FILE_MIGRATE_KEY);
- if (tmp) {
- memcpy (value, tmp->data, tmp->len);
- if (strcmp (value, "force") == 0)
- rebal_type = GF_DHT_MIGRATE_DATA_EVEN_IF_LINK_EXISTS;
+ iatt_ptr = &iatt;
- if (conf->decommission_in_progress)
- rebal_type = GF_DHT_MIGRATE_HARDLINK;
- }
+ hashed_subvol = dht_subvol_get_hashed(this, &entry_loc);
+ if (!hashed_subvol) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_HASHED_SUBVOL_GET_FAILED,
+ "Failed to get hashed subvol for %s", entry_loc.path);
+ ret = 0;
+ goto out;
+ }
- ret = dht_migrate_file (this, &entry_loc, cached_subvol,
- hashed_subvol, rebal_type, &fop_errno);
- if (ret == 1) {
- if (fop_errno == ENOSPC) {
- gf_msg_debug (this->name, 0, "migrate-data skipped for"
- " %s due to space constraints",
- entry_loc.path);
-
- /* For remove-brick case if the source is not one of the
- * removed-brick, do not mark the error as failure */
- if (conf->decommission_subvols_cnt) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->decommissioned_bricks[i] == cached_subvol) {
- LOCK (&defrag->lock);
- {
- defrag->total_failures += 1;
- update_skippedcount = _gf_false;
- }
- UNLOCK (&defrag->lock);
-
- break;
- }
- }
- }
+ cached_subvol = dht_subvol_get_cached(this, entry_loc.inode);
+ if (!cached_subvol) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_CACHED_SUBVOL_GET_FAILED,
+ "Failed to get cached subvol for %s", entry_loc.path);
- if (update_skippedcount) {
- LOCK (&defrag->lock);
- {
- defrag->skipped += 1;
- }
- UNLOCK (&defrag->lock);
-
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_MIGRATE_FILE_SKIPPED,
- "File migration skipped for %s.",
- entry_loc.path);
- }
+ ret = 0;
+ goto out;
+ }
- } else if (fop_errno == ENOTSUP) {
- gf_msg_debug (this->name, 0, "migrate-data skipped for"
- " hardlink %s ", entry_loc.path);
- LOCK (&defrag->lock);
+ if (hashed_subvol == cached_subvol) {
+ ret = 0;
+ goto out;
+ }
+
+ inode = inode_link(entry_loc.inode, entry_loc.parent, entry->d_name, &iatt);
+ inode_unref(entry_loc.inode);
+ /* use the inode returned by inode_link */
+ entry_loc.inode = inode;
+
+ old_THIS = THIS;
+ THIS = this;
+ statfs_frame = create_frame(this, this->ctx->pool);
+ if (!statfs_frame) {
+ gf_msg(this->name, GF_LOG_ERROR, DHT_MSG_NO_MEMORY, ENOMEM,
+ "Insufficient memory. Frame creation failed");
+ ret = -1;
+ goto out;
+ }
+
+ /* async statfs information for honoring min-free-disk */
+ dht_get_du_info(statfs_frame, this, loc);
+ THIS = old_THIS;
+
+ tmp = dict_get(migrate_data, GF_XATTR_FILE_MIGRATE_KEY);
+ if (tmp) {
+ memcpy(value, tmp->data, tmp->len);
+ if (strcmp(value, "force") == 0)
+ rebal_type = GF_DHT_MIGRATE_DATA_EVEN_IF_LINK_EXISTS;
+
+ if (conf->decommission_in_progress)
+ rebal_type = GF_DHT_MIGRATE_HARDLINK;
+ }
+
+ ret = dht_migrate_file(this, &entry_loc, cached_subvol, hashed_subvol,
+ rebal_type, &fop_errno);
+ if (ret == 1) {
+ if (fop_errno == ENOSPC) {
+ gf_msg_debug(this->name, 0,
+ "migrate-data skipped for"
+ " %s due to space constraints",
+ entry_loc.path);
+
+ /* For remove-brick case if the source is not one of the
+ * removed-brick, do not mark the error as failure */
+ if (conf->decommission_subvols_cnt) {
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->decommissioned_bricks[i] == cached_subvol) {
+ LOCK(&defrag->lock);
{
- defrag->skipped += 1;
+ defrag->total_failures += 1;
+ update_skippedcount = _gf_false;
}
- UNLOCK (&defrag->lock);
+ UNLOCK(&defrag->lock);
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_MIGRATE_FILE_SKIPPED,
- "File migration skipped for %s.",
- entry_loc.path);
+ break;
+ }
}
+ }
- ret = 0;
-
- } else if (ret < 0) {
- if (fop_errno != EEXIST) {
- gf_msg (this->name, GF_LOG_ERROR, fop_errno,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "migrate-data failed for %s", entry_loc.path);
-
- LOCK (&defrag->lock);
- {
- defrag->total_failures += 1;
- }
- UNLOCK (&defrag->lock);
-
+ if (update_skippedcount) {
+ LOCK(&defrag->lock);
+ {
+ defrag->skipped += 1;
}
+ UNLOCK(&defrag->lock);
- ret = gf_defrag_handle_migrate_error (fop_errno, defrag);
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_MIGRATE_FILE_SKIPPED,
+ "File migration skipped for %s.", entry_loc.path);
+ }
- if (!ret) {
- gf_msg(this->name, GF_LOG_ERROR, fop_errno,
- DHT_MSG_MIGRATE_FILE_FAILED,
- "migrate-data on %s failed:", entry_loc.path);
- } else if (ret == 1) {
- ret = 0;
- }
+ } else if (fop_errno == ENOTSUP) {
+ gf_msg_debug(this->name, 0,
+ "migrate-data skipped for"
+ " hardlink %s ",
+ entry_loc.path);
+ LOCK(&defrag->lock);
+ {
+ defrag->skipped += 1;
+ }
+ UNLOCK(&defrag->lock);
- goto out;
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_MIGRATE_FILE_SKIPPED,
+ "File migration skipped for %s.", entry_loc.path);
}
- LOCK (&defrag->lock);
- {
- defrag->total_files += 1;
- defrag->total_data += iatt.ia_size;
- }
- UNLOCK (&defrag->lock);
-
- if (defrag->stats == _gf_true) {
- gettimeofday (&end, NULL);
- elapsed = (end.tv_sec - start.tv_sec) * 1e6 +
- (end.tv_usec - start.tv_usec);
- gf_log (this->name, GF_LOG_INFO, "Migration of "
- "file:%s size:%"PRIu64" bytes took %.2f"
- "secs and ret: %d", entry_loc.name,
- iatt.ia_size, elapsed/1e6, ret);
+ ret = 0;
+ goto out;
+ } else if (ret < 0) {
+ if (fop_errno != EEXIST) {
+ gf_msg(this->name, GF_LOG_ERROR, fop_errno,
+ DHT_MSG_MIGRATE_FILE_FAILED, "migrate-data failed for %s",
+ entry_loc.path);
+
+ LOCK(&defrag->lock);
+ {
+ defrag->total_failures += 1;
+ }
+ UNLOCK(&defrag->lock);
}
+ ret = gf_defrag_handle_migrate_error(fop_errno, defrag);
+
+ if (!ret) {
+ gf_msg(this->name, GF_LOG_ERROR, fop_errno,
+ DHT_MSG_MIGRATE_FILE_FAILED,
+ "migrate-data on %s failed:", entry_loc.path);
+ } else if (ret == 1) {
+ ret = 0;
+ }
+
+ goto out;
+ }
+
+ LOCK(&defrag->lock);
+ {
+ defrag->total_files += 1;
+ defrag->total_data += iatt.ia_size;
+ }
+ UNLOCK(&defrag->lock);
+
+ if (defrag->stats == _gf_true) {
+ gettimeofday(&end, NULL);
+ elapsed = gf_tvdiff(&start, &end);
+ gf_log(this->name, GF_LOG_INFO,
+ "Migration of "
+ "file:%s size:%" PRIu64
+ " bytes took %.2f"
+ "secs and ret: %d",
+ entry_loc.name, iatt.ia_size, elapsed / 1e6, ret);
+ }
+
out:
- if (statfs_frame) {
- STACK_DESTROY (statfs_frame->root);
- }
+ if (statfs_frame) {
+ STACK_DESTROY(statfs_frame->root);
+ }
- if (iatt_ptr) {
- LOCK (&defrag->lock);
- {
- defrag->size_processed += iatt_ptr->ia_size;
- }
- UNLOCK (&defrag->lock);
+ if (iatt_ptr) {
+ LOCK(&defrag->lock);
+ {
+ defrag->size_processed += iatt_ptr->ia_size;
}
- loc_wipe (&entry_loc);
-
- return ret;
+ UNLOCK(&defrag->lock);
+ }
+ loc_wipe(&entry_loc);
+ return ret;
}
void *
-gf_defrag_task (void *opaque)
+gf_defrag_task(void *opaque)
{
- struct list_head *q_head = NULL;
- struct dht_container *iterator = NULL;
- gf_defrag_info_t *defrag = NULL;
- int ret = 0;
- pid_t pid = GF_CLIENT_PID_DEFRAG;
-
- defrag = (gf_defrag_info_t *)opaque;
- if (!defrag) {
- gf_msg ("dht", GF_LOG_ERROR, 0, 0, "defrag is NULL");
- goto out;
+ struct list_head *q_head = NULL;
+ struct dht_container *iterator = NULL;
+ gf_defrag_info_t *defrag = NULL;
+ int ret = 0;
+ pid_t pid = GF_CLIENT_PID_DEFRAG;
+
+ defrag = (gf_defrag_info_t *)opaque;
+ if (!defrag) {
+ gf_msg("dht", GF_LOG_ERROR, 0, 0, "defrag is NULL");
+ goto out;
+ }
+
+ syncopctx_setfspid(&pid);
+
+ q_head = &(defrag->queue[0].list);
+
+ /* The following while loop will dequeue one entry from the defrag->queue
+ under lock. We will update the defrag->global_error only when there
+ is an error which is critical to stop the rebalance process. The stop
+ message will be intimated to other migrator threads by setting the
+ defrag->defrag_status to GF_DEFRAG_STATUS_FAILED.
+
+ In defrag->queue, a low watermark (MIN_MIGRATE_QUEUE_COUNT) is
+ maintained so that crawler does not starve the file migration
+ workers and a high watermark (MAX_MIGRATE_QUEUE_COUNT) so that
+ crawler does not go far ahead in filling up the queue.
+ */
+
+ while (_gf_true) {
+ if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
+ pthread_cond_broadcast(&defrag->rebalance_crawler_alarm);
+ pthread_cond_broadcast(&defrag->parallel_migration_cond);
+ goto out;
}
- syncopctx_setfspid (&pid);
+ pthread_mutex_lock(&defrag->dfq_mutex);
+ {
+ /*Throttle down:
+ If the reconfigured count is less than current thread
+ count, then the current thread will sleep */
+
+ /*TODO: Need to refactor the following block to work
+ *under defrag->lock. For now access
+ * defrag->current_thread_count and rthcount under
+ * dfq_mutex lock */
+ while (!defrag->crawl_done && (defrag->recon_thread_count <
+ defrag->current_thread_count)) {
+ defrag->current_thread_count--;
+ gf_msg_debug("DHT", 0,
+ "Thread sleeping. "
+ "current thread count: %d",
+ defrag->current_thread_count);
+
+ pthread_cond_wait(&defrag->df_wakeup_thread,
+ &defrag->dfq_mutex);
+
+ defrag->current_thread_count++;
+ gf_msg_debug("DHT", 0,
+ "Thread wokeup. "
+ "current thread count: %d",
+ defrag->current_thread_count);
+ }
+
+ if (defrag->q_entry_count) {
+ iterator = list_entry(q_head->next, typeof(*iterator), list);
+
+ gf_msg_debug("DHT", 0,
+ "picking entry "
+ "%s",
+ iterator->df_entry->d_name);
+
+ list_del_init(&(iterator->list));
+
+ defrag->q_entry_count--;
+
+ if ((defrag->q_entry_count < MIN_MIGRATE_QUEUE_COUNT) &&
+ defrag->wakeup_crawler) {
+ pthread_cond_broadcast(&defrag->rebalance_crawler_alarm);
+ }
+ pthread_mutex_unlock(&defrag->dfq_mutex);
+ ret = gf_defrag_migrate_single_file((void *)iterator);
- q_head = &(defrag->queue[0].list);
+ /*Critical errors: ENOTCONN and ENOSPACE*/
+ if (ret) {
+ dht_set_global_defrag_error(defrag, ret);
- /* The following while loop will dequeue one entry from the defrag->queue
- under lock. We will update the defrag->global_error only when there
- is an error which is critical to stop the rebalance process. The stop
- message will be intimated to other migrator threads by setting the
- defrag->defrag_status to GF_DEFRAG_STATUS_FAILED.
+ defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
- In defrag->queue, a low watermark (MIN_MIGRATE_QUEUE_COUNT) is
- maintained so that crawler does not starve the file migration
- workers and a high watermark (MAX_MIGRATE_QUEUE_COUNT) so that
- crawler does not go far ahead in filling up the queue.
- */
+ pthread_cond_broadcast(&defrag->rebalance_crawler_alarm);
- while (_gf_true) {
+ pthread_cond_broadcast(&defrag->parallel_migration_cond);
- if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
- pthread_cond_broadcast (
- &defrag->rebalance_crawler_alarm);
- pthread_cond_broadcast (
- &defrag->parallel_migration_cond);
- goto out;
+ goto out;
}
- pthread_mutex_lock (&defrag->dfq_mutex);
- {
-
- /*Throttle down:
- If the reconfigured count is less than current thread
- count, then the current thread will sleep */
-
- /*TODO: Need to refactor the following block to work
- *under defrag->lock. For now access
- * defrag->current_thread_count and rthcount under
- * dfq_mutex lock */
- while (!defrag->crawl_done &&
- (defrag->recon_thread_count <
- defrag->current_thread_count)) {
- defrag->current_thread_count--;
- gf_msg_debug ("DHT", 0, "Thread sleeping. "
- "current thread count: %d",
- defrag->current_thread_count);
-
- pthread_cond_wait (
- &defrag->df_wakeup_thread,
- &defrag->dfq_mutex);
-
- defrag->current_thread_count++;
- gf_msg_debug ("DHT", 0, "Thread wokeup. "
- "current thread count: %d",
- defrag->current_thread_count);
-
- }
+ gf_defrag_free_container(iterator);
+
+ continue;
+ } else {
+ /* defrag->crawl_done flag is set means crawling
+ file system is done and hence a list_empty when
+ the above flag is set indicates there are no more
+ entries to be added to the queue and rebalance is
+ finished */
+
+ if (!defrag->crawl_done) {
+ defrag->current_thread_count--;
+ gf_msg_debug("DHT", 0,
+ "Thread "
+ "sleeping while waiting "
+ "for migration entries. "
+ "current thread count:%d",
+ defrag->current_thread_count);
+
+ pthread_cond_wait(&defrag->parallel_migration_cond,
+ &defrag->dfq_mutex);
+ }
- if (defrag->q_entry_count) {
- iterator = list_entry (q_head->next,
- typeof(*iterator), list);
-
- gf_msg_debug ("DHT", 0, "picking entry "
- "%s", iterator->df_entry->d_name);
-
- list_del_init (&(iterator->list));
-
- defrag->q_entry_count--;
-
- if ((defrag->q_entry_count <
- MIN_MIGRATE_QUEUE_COUNT) &&
- defrag->wakeup_crawler) {
- pthread_cond_broadcast (
- &defrag->rebalance_crawler_alarm);
- }
- pthread_mutex_unlock (&defrag->dfq_mutex);
- ret = gf_defrag_migrate_single_file
- ((void *)iterator);
-
- /*Critical errors: ENOTCONN and ENOSPACE*/
- if (ret) {
- dht_set_global_defrag_error
- (defrag, ret);
-
- defrag->defrag_status =
- GF_DEFRAG_STATUS_FAILED;
-
- pthread_cond_broadcast (
- &defrag->rebalance_crawler_alarm);
-
- pthread_cond_broadcast (
- &defrag->parallel_migration_cond);
-
- goto out;
- }
-
- gf_defrag_free_container (iterator);
-
- continue;
- } else {
-
- /* defrag->crawl_done flag is set means crawling
- file system is done and hence a list_empty when
- the above flag is set indicates there are no more
- entries to be added to the queue and rebalance is
- finished */
-
- if (!defrag->crawl_done) {
-
- defrag->current_thread_count--;
- gf_msg_debug ("DHT", 0, "Thread "
- "sleeping while waiting "
- "for migration entries. "
- "current thread count:%d",
- defrag->current_thread_count);
-
- pthread_cond_wait (
- &defrag->parallel_migration_cond,
- &defrag->dfq_mutex);
- }
-
- if (defrag->crawl_done &&
- !defrag->q_entry_count) {
- defrag->current_thread_count++;
- gf_msg_debug ("DHT", 0, "Exiting thread");
-
- pthread_cond_broadcast (
- &defrag->parallel_migration_cond);
- goto unlock;
- } else {
- defrag->current_thread_count++;
- gf_msg_debug ("DHT", 0, "Thread woke up"
- " as found migrating entries. "
- "current thread count:%d",
- defrag->current_thread_count);
-
- pthread_mutex_unlock
- (&defrag->dfq_mutex);
- continue;
- }
- }
+ if (defrag->crawl_done && !defrag->q_entry_count) {
+ defrag->current_thread_count++;
+ gf_msg_debug("DHT", 0, "Exiting thread");
+ pthread_cond_broadcast(&defrag->parallel_migration_cond);
+ goto unlock;
+ } else {
+ defrag->current_thread_count++;
+ gf_msg_debug("DHT", 0,
+ "Thread woke up"
+ " as found migrating entries. "
+ "current thread count:%d",
+ defrag->current_thread_count);
+
+ pthread_mutex_unlock(&defrag->dfq_mutex);
+ continue;
}
-unlock:
- pthread_mutex_unlock (&defrag->dfq_mutex);
- break;
+ }
}
+ unlock:
+ pthread_mutex_unlock(&defrag->dfq_mutex);
+ break;
+ }
out:
- return NULL;
+ return NULL;
}
-int static
-gf_defrag_get_entry (xlator_t *this, int i, struct dht_container **container,
- loc_t *loc, dht_conf_t *conf, gf_defrag_info_t *defrag,
- fd_t *fd, dict_t *migrate_data,
- struct dir_dfmeta *dir_dfmeta, dict_t *xattr_req,
- int *should_commit_hash, int *perrno)
+int static gf_defrag_get_entry(xlator_t *this, int i,
+ struct dht_container **container, loc_t *loc,
+ dht_conf_t *conf, gf_defrag_info_t *defrag,
+ fd_t *fd, dict_t *migrate_data,
+ struct dir_dfmeta *dir_dfmeta, dict_t *xattr_req,
+ int *perrno)
{
- int ret = -1;
- char is_linkfile = 0;
- gf_dirent_t *df_entry = NULL;
- struct dht_container *tmp_container = NULL;
-
- if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
- ret = -1;
- goto out;
- }
-
- if (dir_dfmeta->offset_var[i].readdir_done == 1) {
- ret = 0;
- goto out;
- }
+ int ret = 0;
+ char is_linkfile = 0;
+ gf_dirent_t *df_entry = NULL;
+ struct dht_container *tmp_container = NULL;
- if (dir_dfmeta->fetch_entries[i] == 1) {
- ret = syncop_readdirp (conf->local_subvols[i], fd, 131072,
- dir_dfmeta->offset_var[i].offset,
- &(dir_dfmeta->equeue[i]),
- xattr_req, NULL);
- if (ret == 0) {
- dir_dfmeta->offset_var[i].readdir_done = 1;
- ret = 0;
- goto out;
- }
+ if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
+ ret = -1;
+ goto out;
+ }
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_DATA_FAILED,
- "Readdirp failed. Aborting data migration for "
- "directory: %s", loc->path);
- *perrno = -ret;
- ret = -1;
- goto out;
- }
+ if (dir_dfmeta->offset_var[i].readdir_done == 1) {
+ ret = 0;
+ goto out;
+ }
- if (list_empty (&(dir_dfmeta->equeue[i].list))) {
- dir_dfmeta->offset_var[i].readdir_done = 1;
- ret = 0;
- goto out;
- }
+ if (dir_dfmeta->fetch_entries[i] == 1) {
+ if (!fd) {
+ dir_dfmeta->fetch_entries[i] = 0;
+ dir_dfmeta->offset_var[i].readdir_done = 1;
+ ret = 0;
+ goto out;
+ }
- dir_dfmeta->fetch_entries[i] = 0;
+ ret = syncop_readdirp(conf->local_subvols[i], fd, 131072,
+ dir_dfmeta->offset_var[i].offset,
+ &(dir_dfmeta->equeue[i]), xattr_req, NULL);
+ if (ret == 0) {
+ dir_dfmeta->offset_var[i].readdir_done = 1;
+ ret = 0;
+ goto out;
}
- while (1) {
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret,
+ DHT_MSG_MIGRATE_DATA_FAILED,
+ "Readdirp failed. Aborting data migration for "
+ "directory: %s",
+ loc->path);
+ *perrno = -ret;
+ ret = -1;
+ goto out;
+ }
- if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
- ret = -1;
- goto out;
- }
+ if (list_empty(&(dir_dfmeta->equeue[i].list))) {
+ dir_dfmeta->offset_var[i].readdir_done = 1;
+ ret = 0;
+ goto out;
+ }
- df_entry = list_entry (dir_dfmeta->iterator[i]->next,
- typeof (*df_entry), list);
+ dir_dfmeta->fetch_entries[i] = 0;
+ }
- if (&df_entry->list == dir_dfmeta->head[i]) {
- gf_dirent_free (&(dir_dfmeta->equeue[i]));
- INIT_LIST_HEAD (&(dir_dfmeta->equeue[i].list));
- dir_dfmeta->fetch_entries[i] = 1;
- dir_dfmeta->iterator[i] = dir_dfmeta->head[i];
- ret = 0;
- goto out;
- }
+ while (1) {
+ if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
+ ret = -1;
+ goto out;
+ }
- dir_dfmeta->iterator[i] = dir_dfmeta->iterator[i]->next;
+ df_entry = list_entry(dir_dfmeta->iterator[i]->next, typeof(*df_entry),
+ list);
- dir_dfmeta->offset_var[i].offset = df_entry->d_off;
- if (!strcmp (df_entry->d_name, ".") ||
- !strcmp (df_entry->d_name, ".."))
- continue;
+ if (&df_entry->list == dir_dfmeta->head[i]) {
+ gf_dirent_free(&(dir_dfmeta->equeue[i]));
+ INIT_LIST_HEAD(&(dir_dfmeta->equeue[i].list));
+ dir_dfmeta->fetch_entries[i] = 1;
+ dir_dfmeta->iterator[i] = dir_dfmeta->head[i];
+ ret = 0;
+ goto out;
+ }
- if (IA_ISDIR (df_entry->d_stat.ia_type)) {
- defrag->size_processed += df_entry->d_stat.ia_size;
- continue;
- }
+ dir_dfmeta->iterator[i] = dir_dfmeta->iterator[i]->next;
- defrag->num_files_lookedup++;
+ dir_dfmeta->offset_var[i].offset = df_entry->d_off;
+ if (!strcmp(df_entry->d_name, ".") || !strcmp(df_entry->d_name, ".."))
+ continue;
- if (defrag->defrag_pattern &&
- (gf_defrag_pattern_match (defrag, df_entry->d_name,
- df_entry->d_stat.ia_size)
- == _gf_false)) {
- defrag->size_processed += df_entry->d_stat.ia_size;
- continue;
- }
+ if (IA_ISDIR(df_entry->d_stat.ia_type)) {
+ defrag->size_processed += df_entry->d_stat.ia_size;
+ continue;
+ }
- is_linkfile = check_is_linkfile (NULL, &df_entry->d_stat,
- df_entry->dict,
- conf->link_xattr_name);
+ defrag->num_files_lookedup++;
- if (is_linkfile) {
- /* No need to add linkto file to the queue for
- migration. Only the actual data file need to
- be checked for migration criteria.
- */
+ if (defrag->defrag_pattern &&
+ (gf_defrag_pattern_match(defrag, df_entry->d_name,
+ df_entry->d_stat.ia_size) == _gf_false)) {
+ defrag->size_processed += df_entry->d_stat.ia_size;
+ continue;
+ }
- gf_msg_debug (this->name, 0, "Skipping linkfile"
- " %s on subvol: %s", df_entry->d_name,
- conf->local_subvols[i]->name);
- continue;
- }
+ is_linkfile = check_is_linkfile(NULL, &df_entry->d_stat, df_entry->dict,
+ conf->link_xattr_name);
- /*Build Container Structure */
+ if (is_linkfile) {
+ /* No need to add linkto file to the queue for
+ migration. Only the actual data file need to
+ be checked for migration criteria.
+ */
- tmp_container = GF_CALLOC (1, sizeof(struct dht_container),
- gf_dht_mt_container_t);
- if (!tmp_container) {
- gf_log (this->name, GF_LOG_ERROR, "Failed to allocate "
- "memory for container");
- ret = -1;
- goto out;
- }
- tmp_container->df_entry = gf_dirent_for_name (df_entry->d_name);
- if (!tmp_container->df_entry) {
- gf_log (this->name, GF_LOG_ERROR, "Failed to allocate "
- "memory for df_entry");
- ret = -1;
- goto out;
- }
+ gf_msg_debug(this->name, 0,
+ "Skipping linkfile"
+ " %s on subvol: %s",
+ df_entry->d_name, conf->local_subvols[i]->name);
+ continue;
+ }
- tmp_container->local_subvol_index = i;
+ /*Build Container Structure */
- tmp_container->df_entry->d_stat = df_entry->d_stat;
+ tmp_container = GF_CALLOC(1, sizeof(struct dht_container),
+ gf_dht_mt_container_t);
+ if (!tmp_container) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Failed to allocate "
+ "memory for container");
+ ret = -1;
+ goto out;
+ }
+ tmp_container->df_entry = gf_dirent_for_name(df_entry->d_name);
+ if (!tmp_container->df_entry) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Failed to allocate "
+ "memory for df_entry");
+ ret = -1;
+ goto out;
+ }
- tmp_container->df_entry->d_ino = df_entry->d_ino;
+ tmp_container->local_subvol_index = i;
- tmp_container->df_entry->d_type = df_entry->d_type;
+ tmp_container->df_entry->d_stat = df_entry->d_stat;
- tmp_container->df_entry->d_len = df_entry->d_len;
+ tmp_container->df_entry->d_ino = df_entry->d_ino;
- tmp_container->parent_loc = GF_CALLOC(1, sizeof(*loc),
- gf_dht_mt_loc_t);
- if (!tmp_container->parent_loc) {
- gf_log (this->name, GF_LOG_ERROR, "Failed to allocate "
- "memory for loc");
- ret = -1;
- goto out;
- }
+ tmp_container->df_entry->d_type = df_entry->d_type;
+ tmp_container->df_entry->d_len = df_entry->d_len;
- ret = loc_copy (tmp_container->parent_loc, loc);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "loc_copy failed");
- ret = -1;
- goto out;
- }
+ tmp_container->parent_loc = GF_CALLOC(1, sizeof(*loc), gf_dht_mt_loc_t);
+ if (!tmp_container->parent_loc) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Failed to allocate "
+ "memory for loc");
+ ret = -1;
+ goto out;
+ }
- tmp_container->migrate_data = migrate_data;
+ ret = loc_copy(tmp_container->parent_loc, loc);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR, "loc_copy failed");
+ ret = -1;
+ goto out;
+ }
- tmp_container->this = this;
+ tmp_container->migrate_data = migrate_data;
- if (df_entry->dict)
- tmp_container->df_entry->dict =
- dict_ref (df_entry->dict);
+ tmp_container->this = this;
- /*Build Container Structue >> END*/
+ if (df_entry->dict)
+ tmp_container->df_entry->dict = dict_ref(df_entry->dict);
- ret = 0;
- goto out;
+ /*Build Container Structure >> END*/
- }
+ ret = 0;
+ goto out;
+ }
out:
- if (ret == 0) {
- *container = tmp_container;
- } else {
- if (tmp_container) {
- gf_defrag_free_container (tmp_container);
- }
+ if (ret == 0) {
+ *container = tmp_container;
+ } else {
+ if (tmp_container) {
+ gf_defrag_free_container(tmp_container);
}
+ }
- return ret;
+ return ret;
}
int
-gf_defrag_process_dir (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
- dict_t *migrate_data, int *perrno)
+gf_defrag_process_dir(xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
+ dict_t *migrate_data, int *perrno)
{
- int ret = -1;
- fd_t *fd = NULL;
- dht_conf_t *conf = NULL;
- gf_dirent_t entries;
- dict_t *xattr_req = NULL;
- struct timeval dir_start = {0,};
- struct timeval end = {0,};
- double elapsed = {0,};
- int local_subvols_cnt = 0;
- int i = 0;
- int j = 0;
- struct dht_container *container = NULL;
- int ldfq_count = 0;
- int dfc_index = 0;
- int throttle_up = 0;
- struct dir_dfmeta *dir_dfmeta = NULL;
- int should_commit_hash = 1;
-
- gf_log (this->name, GF_LOG_INFO, "migrate data called on %s",
- loc->path);
- gettimeofday (&dir_start, NULL);
-
- conf = this->private;
- local_subvols_cnt = conf->local_subvols_cnt;
-
- if (!local_subvols_cnt) {
- ret = 0;
- goto out;
- }
-
- fd = fd_create (loc->inode, defrag->pid);
- if (!fd) {
- gf_log (this->name, GF_LOG_ERROR, "Failed to create fd");
- ret = -1;
- goto out;
- }
-
- ret = syncop_opendir (this, loc, fd, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_MIGRATE_DATA_FAILED,
- "Migrate data failed: Failed to open dir %s",
- loc->path);
- *perrno = -ret;
- ret = -1;
- goto out;
- }
-
- fd_bind (fd);
- dir_dfmeta = GF_CALLOC (1, sizeof (*dir_dfmeta),
- gf_common_mt_pointer);
- if (!dir_dfmeta) {
- gf_log (this->name, GF_LOG_ERROR, "dir_dfmeta is NULL");
- ret = -1;
- goto out;
- }
-
-
- dir_dfmeta->head = GF_CALLOC (local_subvols_cnt,
- sizeof (*(dir_dfmeta->head)),
- gf_common_mt_pointer);
- if (!dir_dfmeta->head) {
- gf_log (this->name, GF_LOG_ERROR, "dir_dfmeta->head is NULL");
- ret = -1;
- goto out;
- }
-
- dir_dfmeta->iterator = GF_CALLOC (local_subvols_cnt,
- sizeof (*(dir_dfmeta->iterator)),
- gf_common_mt_pointer);
- if (!dir_dfmeta->iterator) {
- gf_log (this->name, GF_LOG_ERROR,
- "dir_dfmeta->iterator is NULL");
- ret = -1;
- goto out;
- }
+ int ret = -1;
+ dht_conf_t *conf = NULL;
+ gf_dirent_t entries;
+ dict_t *xattr_req = NULL;
+ struct timeval dir_start = {
+ 0,
+ };
+ struct timeval end = {
+ 0,
+ };
+ double elapsed = {
+ 0,
+ };
+ int local_subvols_cnt = 0;
+ int i = 0;
+ int j = 0;
+ struct dht_container *container = NULL;
+ int ldfq_count = 0;
+ int dfc_index = 0;
+ int throttle_up = 0;
+ struct dir_dfmeta *dir_dfmeta = NULL;
+ xlator_t *old_THIS = NULL;
+
+ gf_log(this->name, GF_LOG_INFO, "migrate data called on %s", loc->path);
+ gettimeofday(&dir_start, NULL);
+
+ conf = this->private;
+ local_subvols_cnt = conf->local_subvols_cnt;
+
+ if (!local_subvols_cnt) {
+ ret = 0;
+ goto out;
+ }
- dir_dfmeta->equeue = GF_CALLOC (local_subvols_cnt, sizeof (entries),
- gf_dht_mt_dirent_t);
- if (!dir_dfmeta->equeue) {
- gf_log (this->name, GF_LOG_ERROR, "dir_dfmeta->equeue is NULL");
- ret = -1;
- goto out;
- }
+ old_THIS = THIS;
+ THIS = this;
- dir_dfmeta->offset_var = GF_CALLOC (local_subvols_cnt,
- sizeof (dht_dfoffset_ctx_t),
- gf_dht_mt_octx_t);
- if (!dir_dfmeta->offset_var) {
- gf_log (this->name, GF_LOG_ERROR,
- "dir_dfmeta->offset_var is NULL");
- ret = -1;
- goto out;
- }
- ret = gf_defrag_ctx_subvols_init (dir_dfmeta->offset_var, this);
+ dir_dfmeta = GF_CALLOC(1, sizeof(*dir_dfmeta), gf_common_mt_pointer);
+ if (!dir_dfmeta) {
+ gf_log(this->name, GF_LOG_ERROR, "dir_dfmeta is NULL");
+ ret = -1;
+ goto out;
+ }
+
+ dir_dfmeta->lfd = GF_CALLOC(local_subvols_cnt, sizeof(fd_t *),
+ gf_common_mt_pointer);
+ if (!dir_dfmeta->lfd) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_INSUFF_MEMORY,
+ "for dir_dfmeta", NULL);
+ ret = -1;
+ *perrno = ENOMEM;
+ goto out;
+ }
+
+ for (i = 0; i < local_subvols_cnt; i++) {
+ dir_dfmeta->lfd[i] = fd_create(loc->inode, defrag->pid);
+ if (!dir_dfmeta->lfd[i]) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_FD_CREATE_FAILED,
+ NULL);
+ *perrno = ENOMEM;
+ ret = -1;
+ goto out;
+ }
+
+ ret = syncop_opendir(conf->local_subvols[i], loc, dir_dfmeta->lfd[i],
+ NULL, NULL);
if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "dht_dfoffset_ctx_t"
- "initialization failed");
- ret = -1;
- goto out;
- }
+ fd_unref(dir_dfmeta->lfd[i]);
+ dir_dfmeta->lfd[i] = NULL;
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_FAILED_TO_OPEN,
+ "dir: %s", loc->path, "subvol: %s",
+ conf->local_subvols[i]->name, NULL);
- dir_dfmeta->fetch_entries = GF_CALLOC (local_subvols_cnt,
- sizeof (int), gf_common_mt_int);
- if (!dir_dfmeta->fetch_entries) {
- gf_log (this->name, GF_LOG_ERROR,
- "dir_dfmeta->fetch_entries is NULL");
+ if (conf->decommission_in_progress) {
+ *perrno = -ret;
ret = -1;
goto out;
+ }
+ } else {
+ fd_bind(dir_dfmeta->lfd[i]);
}
+ }
- for (i = 0; i < local_subvols_cnt ; i++) {
- INIT_LIST_HEAD (&(dir_dfmeta->equeue[i].list));
- dir_dfmeta->head[i] = &(dir_dfmeta->equeue[i].list);
- dir_dfmeta->iterator[i] = dir_dfmeta->head[i];
- dir_dfmeta->fetch_entries[i] = 1;
- }
-
- xattr_req = dict_new ();
- if (!xattr_req) {
- gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
- ret = -1;
- goto out;
- }
-
- ret = dict_set_uint32 (xattr_req,
- conf->link_xattr_name, 256);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "failed to set dict for "
- "key: %s", conf->link_xattr_name);
- ret = -1;
- goto out;
- }
-
- /*
- Job: Read entries from each local subvol and store the entries
- in equeue array of linked list. Now pick one entry from the
- equeue array in a round robin basis and add them to defrag Queue.
- */
-
- while (!dht_dfreaddirp_done(dir_dfmeta->offset_var,
- local_subvols_cnt)) {
-
- pthread_mutex_lock (&defrag->dfq_mutex);
- {
-
- /*Throttle up: If reconfigured count is higher than
- current thread count, wake up the sleeping threads
- TODO: Need to refactor this. Instead of making the
- thread sleep and wake, we should terminate and spawn
- threads on-demand*/
-
- if (defrag->recon_thread_count >
- defrag->current_thread_count) {
- throttle_up =
- (defrag->recon_thread_count -
- defrag->current_thread_count);
- for (j = 0; j < throttle_up; j++) {
- pthread_cond_signal (
- &defrag->df_wakeup_thread);
- }
-
- }
+ dir_dfmeta->head = GF_CALLOC(local_subvols_cnt, sizeof(*(dir_dfmeta->head)),
+ gf_common_mt_pointer);
+ if (!dir_dfmeta->head) {
+ gf_log(this->name, GF_LOG_ERROR, "dir_dfmeta->head is NULL");
+ ret = -1;
+ goto out;
+ }
+
+ dir_dfmeta->iterator = GF_CALLOC(local_subvols_cnt,
+ sizeof(*(dir_dfmeta->iterator)),
+ gf_common_mt_pointer);
+ if (!dir_dfmeta->iterator) {
+ gf_log(this->name, GF_LOG_ERROR, "dir_dfmeta->iterator is NULL");
+ ret = -1;
+ goto out;
+ }
- while (defrag->q_entry_count >
- MAX_MIGRATE_QUEUE_COUNT) {
- defrag->wakeup_crawler = 1;
- pthread_cond_wait (
- &defrag->rebalance_crawler_alarm,
- &defrag->dfq_mutex);
- }
+ dir_dfmeta->equeue = GF_CALLOC(local_subvols_cnt, sizeof(entries),
+ gf_dht_mt_dirent_t);
+ if (!dir_dfmeta->equeue) {
+ gf_log(this->name, GF_LOG_ERROR, "dir_dfmeta->equeue is NULL");
+ ret = -1;
+ goto out;
+ }
- ldfq_count = defrag->q_entry_count;
+ dir_dfmeta->offset_var = GF_CALLOC(
+ local_subvols_cnt, sizeof(dht_dfoffset_ctx_t), gf_dht_mt_octx_t);
+ if (!dir_dfmeta->offset_var) {
+ gf_log(this->name, GF_LOG_ERROR, "dir_dfmeta->offset_var is NULL");
+ ret = -1;
+ goto out;
+ }
+
+ ret = gf_defrag_ctx_subvols_init(dir_dfmeta->offset_var, this);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "dht_dfoffset_ctx_t"
+ "initialization failed");
+ ret = -1;
+ goto out;
+ }
+
+ dir_dfmeta->fetch_entries = GF_CALLOC(local_subvols_cnt, sizeof(int),
+ gf_common_mt_int);
+ if (!dir_dfmeta->fetch_entries) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_INSUFF_MEMORY,
+ "for dir_dfmeta->fetch_entries", NULL);
+ ret = -1;
+ goto out;
+ }
+
+ for (i = 0; i < local_subvols_cnt; i++) {
+ INIT_LIST_HEAD(&(dir_dfmeta->equeue[i].list));
+ dir_dfmeta->head[i] = &(dir_dfmeta->equeue[i].list);
+ dir_dfmeta->iterator[i] = dir_dfmeta->head[i];
+ dir_dfmeta->fetch_entries[i] = 1;
+ }
+
+ xattr_req = dict_new();
+ if (!xattr_req) {
+ gf_log(this->name, GF_LOG_ERROR, "dict_new failed");
+ ret = -1;
+ goto out;
+ }
+
+ ret = dict_set_uint32(xattr_req, conf->link_xattr_name, 256);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to set dict for "
+ "key: %s",
+ conf->link_xattr_name);
+ ret = -1;
+ goto out;
+ }
- if (defrag->wakeup_crawler) {
- defrag->wakeup_crawler = 0;
- }
+ /*
+ Job: Read entries from each local subvol and store the entries
+ in equeue array of linked list. Now pick one entry from the
+ equeue array in a round robin basis and add them to defrag Queue.
+ */
+ while (!dht_dfreaddirp_done(dir_dfmeta->offset_var, local_subvols_cnt)) {
+ pthread_mutex_lock(&defrag->dfq_mutex);
+ {
+ /*Throttle up: If reconfigured count is higher than
+ current thread count, wake up the sleeping threads
+ TODO: Need to refactor this. Instead of making the
+ thread sleep and wake, we should terminate and spawn
+ threads on-demand*/
+
+ if (defrag->recon_thread_count > defrag->current_thread_count) {
+ throttle_up = (defrag->recon_thread_count -
+ defrag->current_thread_count);
+ for (j = 0; j < throttle_up; j++) {
+ pthread_cond_signal(&defrag->df_wakeup_thread);
}
- pthread_mutex_unlock (&defrag->dfq_mutex);
-
- while (ldfq_count <= MAX_MIGRATE_QUEUE_COUNT &&
- !dht_dfreaddirp_done(dir_dfmeta->offset_var,
- local_subvols_cnt)) {
-
- ret = gf_defrag_get_entry (this, dfc_index, &container,
- loc, conf, defrag, fd,
- migrate_data, dir_dfmeta,
- xattr_req,
- &should_commit_hash, perrno);
+ }
- if (ret) {
- gf_log (this->name, GF_LOG_WARNING, "Found "
- "error from gf_defrag_get_entry");
+ while (defrag->q_entry_count > MAX_MIGRATE_QUEUE_COUNT) {
+ defrag->wakeup_crawler = 1;
+ pthread_cond_wait(&defrag->rebalance_crawler_alarm,
+ &defrag->dfq_mutex);
+ }
- ret = -1;
- goto out;
- }
-
- /* Check if we got an entry, else we need to move the
- index to the next subvol */
- if (!container) {
- GF_CRAWL_INDEX_MOVE(dfc_index,
- local_subvols_cnt);
- continue;
- }
+ ldfq_count = defrag->q_entry_count;
- /* Q this entry in the dfq */
- pthread_mutex_lock (&defrag->dfq_mutex);
- {
- list_add_tail (&container->list,
- &(defrag->queue[0].list));
- defrag->q_entry_count++;
- ldfq_count = defrag->q_entry_count;
-
- gf_msg_debug (this->name, 0, "added "
- "file:%s parent:%s to the queue ",
- container->df_entry->d_name,
- container->parent_loc->path);
-
- pthread_cond_signal (
- &defrag->parallel_migration_cond);
- }
- pthread_mutex_unlock (&defrag->dfq_mutex);
-
- GF_CRAWL_INDEX_MOVE(dfc_index, local_subvols_cnt);
- }
+ if (defrag->wakeup_crawler) {
+ defrag->wakeup_crawler = 0;
+ }
}
+ pthread_mutex_unlock(&defrag->dfq_mutex);
- gettimeofday (&end, NULL);
- elapsed = (end.tv_sec - dir_start.tv_sec) * 1e6 +
- (end.tv_usec - dir_start.tv_usec);
- gf_log (this->name, GF_LOG_INFO, "Migration operation on dir %s took "
- "%.2f secs", loc->path, elapsed/1e6);
- ret = 0;
-out:
-
- gf_defrag_free_dir_dfmeta (dir_dfmeta, local_subvols_cnt);
+ while (
+ ldfq_count <= MAX_MIGRATE_QUEUE_COUNT &&
+ !dht_dfreaddirp_done(dir_dfmeta->offset_var, local_subvols_cnt)) {
+ ret = gf_defrag_get_entry(this, dfc_index, &container, loc, conf,
+ defrag, dir_dfmeta->lfd[dfc_index],
+ migrate_data, dir_dfmeta, xattr_req,
+ perrno);
- if (xattr_req)
- dict_unref(xattr_req);
-
- if (fd)
- fd_unref (fd);
+ if (defrag->defrag_status == GF_DEFRAG_STATUS_STOPPED) {
+ goto out;
+ }
- if (ret == 0 && should_commit_hash == 0) {
- ret = 2;
- }
+ if (ret) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "Found "
+ "error from gf_defrag_get_entry");
- /* It does not matter if it errored out - this number is
- * used to calculate rebalance estimated time to complete.
- * No locking required as dirs are processed by a single thread.
- */
- defrag->num_dirs_processed++;
- return ret;
+ ret = -1;
+ goto out;
+ }
+
+ /* Check if we got an entry, else we need to move the
+ index to the next subvol */
+ if (!container) {
+ GF_CRAWL_INDEX_MOVE(dfc_index, local_subvols_cnt);
+ continue;
+ }
+
+ /* Q this entry in the dfq */
+ pthread_mutex_lock(&defrag->dfq_mutex);
+ {
+ list_add_tail(&container->list, &(defrag->queue[0].list));
+ defrag->q_entry_count++;
+ ldfq_count = defrag->q_entry_count;
+
+ gf_msg_debug(this->name, 0,
+ "added "
+ "file:%s parent:%s to the queue ",
+ container->df_entry->d_name,
+ container->parent_loc->path);
+
+ pthread_cond_signal(&defrag->parallel_migration_cond);
+ }
+ pthread_mutex_unlock(&defrag->dfq_mutex);
+
+ GF_CRAWL_INDEX_MOVE(dfc_index, local_subvols_cnt);
+ }
+ }
+
+ gettimeofday(&end, NULL);
+ elapsed = gf_tvdiff(&dir_start, &end);
+ gf_log(this->name, GF_LOG_INFO,
+ "Migration operation on dir %s took "
+ "%.2f secs",
+ loc->path, elapsed / 1e6);
+ ret = 0;
+out:
+ THIS = old_THIS;
+ gf_defrag_free_dir_dfmeta(dir_dfmeta, local_subvols_cnt);
+
+ if (xattr_req)
+ dict_unref(xattr_req);
+
+ /* It does not matter if it errored out - this number is
+ * used to calculate rebalance estimated time to complete.
+ * No locking required as dirs are processed by a single thread.
+ */
+ defrag->num_dirs_processed++;
+ return ret;
}
+
int
-gf_defrag_settle_hash (xlator_t *this, gf_defrag_info_t *defrag,
- loc_t *loc, dict_t *fix_layout)
+gf_defrag_settle_hash(xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
+ dict_t *fix_layout)
{
- int ret;
- dht_conf_t *conf = NULL;
- /*
- * Now we're ready to update the directory commit hash for the volume
- * root, so that hash miscompares and broadcast lookups can stop.
- * However, we want to skip that if fix-layout is all we did. In
- * that case, we want the miscompares etc. to continue until a real
- * rebalance is complete.
- */
- if (defrag->cmd == GF_DEFRAG_CMD_START_LAYOUT_FIX
- || defrag->cmd == GF_DEFRAG_CMD_START_DETACH_TIER ||
- defrag->cmd == GF_DEFRAG_CMD_DETACH_START) {
- return 0;
- }
-
- conf = this->private;
- if (!conf) {
- /*Uh oh
- */
- return -1;
- }
+ int ret;
+ dht_conf_t *conf = NULL;
+ /*
+ * Now we're ready to update the directory commit hash for the volume
+ * root, so that hash miscompares and broadcast lookups can stop.
+ * However, we want to skip that if fix-layout is all we did. In
+ * that case, we want the miscompares etc. to continue until a real
+ * rebalance is complete.
+ */
+ if (defrag->cmd == GF_DEFRAG_CMD_START_LAYOUT_FIX ||
+ defrag->cmd == GF_DEFRAG_CMD_DETACH_START) {
+ return 0;
+ }
- if (conf->local_subvols_cnt == 0 || !conf->lookup_optimize) {
- /* Commit hash updates are only done on local subvolumes and
- * only when lookup optmization is needed (for older client
- * support)
- */
- return 0;
- }
+ conf = this->private;
+ if (!conf) {
+ /*Uh oh
+ */
+ return -1;
+ }
- ret = dict_set_uint32 (fix_layout, "new-commit-hash",
- defrag->new_commit_hash);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to set new-commit-hash");
- return -1;
- }
+ if (conf->local_subvols_cnt == 0 || !conf->lookup_optimize) {
+ /* Commit hash updates are only done on local subvolumes and
+ * only when lookup optimization is needed (for older client
+ * support)
+ */
+ return 0;
+ }
- ret = syncop_setxattr (this, loc, fix_layout, 0, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_LAYOUT_FIX_FAILED,
- "fix layout on %s failed", loc->path);
+ ret = dict_set_uint32(fix_layout, "new-commit-hash",
+ defrag->new_commit_hash);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR, "Failed to set new-commit-hash");
+ return -1;
+ }
- if (-ret == ENOENT || -ret == ESTALE) {
- /* Dir most likely is deleted */
- return 0;
- }
+ ret = syncop_setxattr(this, loc, fix_layout, 0, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_LAYOUT_FIX_FAILED,
+ "fix layout on %s failed", loc->path);
- return -1;
+ if (-ret == ENOENT || -ret == ESTALE) {
+ /* Dir most likely is deleted */
+ return 0;
}
- /* TBD: find more efficient solution than adding/deleting every time */
- dict_del(fix_layout, "new-commit-hash");
-
- return 0;
-}
+ return -1;
+ }
+ /* TBD: find more efficient solution than adding/deleting every time */
+ dict_del(fix_layout, "new-commit-hash");
+ return 0;
+}
-/* Function for doing a named lookup on file inodes during an attach tier
- * So that a hardlink lookup heal i.e gfid to parent gfid lookup heal
- * happens on pre-existing data. This is required so that the ctr database has
- * hardlinks of all the exisitng file in the volume. CTR xlator on the
- * brick/server side does db update/insert of the hardlink on a namelookup.
- * Currently the namedlookup is done synchronous to the fixlayout that is
- * triggered by attach tier. This is not performant, adding more time to
- * fixlayout. The performant approach is record the hardlinks on a compressed
- * datastore and then do the namelookup asynchronously later, giving the ctr db
- * eventual consistency
- * */
int
-gf_fix_layout_tier_attach_lookup (xlator_t *this,
- loc_t *parent_loc,
- gf_dirent_t *file_dentry)
+gf_defrag_fix_layout(xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
+ dict_t *fix_layout, dict_t *migrate_data)
{
- int ret = -1;
- dict_t *lookup_xdata = NULL;
- dht_conf_t *conf = NULL;
- loc_t file_loc = {0,};
- struct iatt iatt = {0,};
-
- GF_VALIDATE_OR_GOTO ("tier", this, out);
-
- GF_VALIDATE_OR_GOTO (this->name, parent_loc, out);
-
- GF_VALIDATE_OR_GOTO (this->name, file_dentry, out);
+ int ret = -1;
+ loc_t entry_loc = {
+ 0,
+ };
+ fd_t *fd = NULL;
+ gf_dirent_t entries;
+ gf_dirent_t *tmp = NULL;
+ gf_dirent_t *entry = NULL;
+ gf_boolean_t free_entries = _gf_false;
+ off_t offset = 0;
+ struct iatt iatt = {
+ 0,
+ };
+ inode_t *linked_inode = NULL, *inode = NULL;
+ dht_conf_t *conf = NULL;
+ int perrno = 0;
+
+ conf = this->private;
+ if (!conf) {
+ ret = -1;
+ goto out;
+ }
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
+ ret = syncop_lookup(this, loc, &iatt, NULL, NULL, NULL);
+ if (ret) {
+ if (strcmp(loc->path, "/") == 0) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_DIR_LOOKUP_FAILED,
+ "lookup failed for:%s", loc->path);
- if (!parent_loc->inode) {
- gf_msg (this->name, GF_LOG_ERROR, 0, DHT_MSG_LOG_TIER_ERROR,
- "%s/%s parent is NULL", parent_loc->path,
- file_dentry->d_name);
- goto out;
+ defrag->total_failures++;
+ ret = -1;
+ goto out;
}
+ if (-ret == ENOENT || -ret == ESTALE) {
+ gf_msg(this->name, GF_LOG_INFO, -ret, DHT_MSG_DIR_LOOKUP_FAILED,
+ "Dir:%s renamed or removed. Skipping", loc->path);
+ if (conf->decommission_subvols_cnt) {
+ defrag->total_failures++;
+ }
+ ret = 0;
+ goto out;
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_DIR_LOOKUP_FAILED,
+ "lookup failed for:%s", loc->path);
- conf = this->private;
-
- loc_wipe (&file_loc);
-
- if (gf_uuid_is_null (file_dentry->d_stat.ia_gfid)) {
- gf_msg (this->name, GF_LOG_ERROR, 0, DHT_MSG_LOG_TIER_ERROR,
- "%s/%s gfid not present", parent_loc->path,
- file_dentry->d_name);
- goto out;
+ defrag->total_failures++;
+ goto out;
}
+ }
- gf_uuid_copy (file_loc.gfid, file_dentry->d_stat.ia_gfid);
+ fd = fd_create(loc->inode, defrag->pid);
+ if (!fd) {
+ gf_log(this->name, GF_LOG_ERROR, "Failed to create fd");
+ ret = -1;
+ goto out;
+ }
- if (gf_uuid_is_null (parent_loc->gfid)) {
- gf_msg (this->name, GF_LOG_ERROR, 0, DHT_MSG_LOG_TIER_ERROR,
- "%s/%s"
- " gfid not present", parent_loc->path,
- file_dentry->d_name);
- goto out;
+ ret = syncop_opendir(this, loc, fd, NULL, NULL);
+ if (ret) {
+ if (-ret == ENOENT || -ret == ESTALE) {
+ if (conf->decommission_subvols_cnt) {
+ defrag->total_failures++;
+ }
+ ret = 0;
+ goto out;
}
- gf_uuid_copy (file_loc.pargfid, parent_loc->gfid);
-
+ gf_log(this->name, GF_LOG_ERROR,
+ "Failed to open dir %s, "
+ "err:%d",
+ loc->path, -ret);
- ret = dht_build_child_loc (this, &file_loc, parent_loc,
- file_dentry->d_name);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, DHT_MSG_LOG_TIER_ERROR,
- "Child loc build failed");
- ret = -1;
- goto out;
- }
+ ret = -1;
+ goto out;
+ }
- lookup_xdata = dict_new ();
- if (!lookup_xdata) {
- gf_msg (this->name, GF_LOG_ERROR, 0, DHT_MSG_LOG_TIER_ERROR,
- "Failed creating lookup dict for %s",
- file_dentry->d_name);
- goto out;
- }
+ fd_bind(fd);
+ INIT_LIST_HEAD(&entries.list);
- ret = dict_set_int32 (lookup_xdata, CTR_ATTACH_TIER_LOOKUP, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, DHT_MSG_LOG_TIER_ERROR,
- "Failed to set lookup flag");
+ while ((ret = syncop_readdirp(this, fd, 131072, offset, &entries, NULL,
+ NULL)) != 0) {
+ if (ret < 0) {
+ if (-ret == ENOENT || -ret == ESTALE) {
+ if (conf->decommission_subvols_cnt) {
+ defrag->total_failures++;
+ }
+ ret = 0;
goto out;
- }
+ }
- gf_uuid_copy (file_loc.parent->gfid, parent_loc->gfid);
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_READDIR_ERROR,
+ "readdirp failed for "
+ "path %s. Aborting fix-layout",
+ loc->path);
- /* Sending lookup to cold tier only */
- ret = syncop_lookup (conf->subvolumes[0], &file_loc, &iatt,
- NULL, lookup_xdata, NULL);
- if (ret) {
- /* If the file does not exist on the cold tier than it must */
- /* have been discovered on the hot tier. This is not an error. */
- gf_msg (this->name, GF_LOG_INFO, 0, DHT_MSG_LOG_TIER_STATUS,
- "%s lookup to cold tier on attach heal failed", file_loc.path);
- goto out;
+ ret = -1;
+ goto out;
}
- ret = 0;
+ if (list_empty(&entries.list))
+ break;
-out:
+ free_entries = _gf_true;
- loc_wipe (&file_loc);
+ list_for_each_entry_safe(entry, tmp, &entries.list, list)
+ {
+ if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
+ ret = 1;
+ goto out;
+ }
- if (lookup_xdata)
- dict_unref (lookup_xdata);
+ offset = entry->d_off;
- return ret;
-}
-
-int
-gf_defrag_fix_layout (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
- dict_t *fix_layout, dict_t *migrate_data)
-{
- int ret = -1;
- loc_t entry_loc = {0,};
- fd_t *fd = NULL;
- gf_dirent_t entries;
- gf_dirent_t *tmp = NULL;
- gf_dirent_t *entry = NULL;
- gf_boolean_t free_entries = _gf_false;
- off_t offset = 0;
- struct iatt iatt = {0,};
- inode_t *linked_inode = NULL, *inode = NULL;
- dht_conf_t *conf = NULL;
- int should_commit_hash = 1;
- int perrno = 0;
-
- conf = this->private;
- if (!conf) {
- ret = -1;
- goto out;
- }
+ if (!strcmp(entry->d_name, ".") || !strcmp(entry->d_name, ".."))
+ continue;
+ if (!IA_ISDIR(entry->d_stat.ia_type)) {
+ continue;
+ }
+ loc_wipe(&entry_loc);
- ret = syncop_lookup (this, loc, &iatt, NULL, NULL, NULL);
- if (ret) {
- if (strcmp (loc->path, "/") == 0) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_DIR_LOOKUP_FAILED,
- "lookup failed for:%s", loc->path);
+ ret = dht_build_child_loc(this, &entry_loc, loc, entry->d_name);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Child loc"
+ " build failed for entry: %s",
+ entry->d_name);
- defrag->total_failures++;
- ret = -1;
- goto out;
- }
+ if (conf->decommission_in_progress) {
+ defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
- if (-ret == ENOENT || -ret == ESTALE) {
- gf_msg (this->name, GF_LOG_INFO, -ret,
- DHT_MSG_DIR_LOOKUP_FAILED,
- "Dir:%s renamed or removed. Skipping",
- loc->path);
- ret = 0;
- goto out;
+ goto out;
} else {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_DIR_LOOKUP_FAILED,
- "lookup failed for:%s", loc->path);
-
- defrag->total_failures++;
- goto out;
+ continue;
}
- }
-
- fd = fd_create (loc->inode, defrag->pid);
- if (!fd) {
- gf_log (this->name, GF_LOG_ERROR, "Failed to create fd");
- ret = -1;
- goto out;
- }
-
- ret = syncop_opendir (this, loc, fd, NULL, NULL);
- if (ret) {
+ }
+
+ if (gf_uuid_is_null(entry->d_stat.ia_gfid)) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "%s/%s"
+ " gfid not present",
+ loc->path, entry->d_name);
+ continue;
+ }
+
+ gf_uuid_copy(entry_loc.gfid, entry->d_stat.ia_gfid);
+
+ /*In case the gfid stored in the inode by inode_link
+ * and the gfid obtained in the lookup differs, then
+ * client3_3_lookup_cbk will return ESTALE and proper
+ * error will be captured
+ */
+
+ linked_inode = inode_link(entry_loc.inode, loc->inode,
+ entry->d_name, &entry->d_stat);
+
+ inode = entry_loc.inode;
+ entry_loc.inode = linked_inode;
+ inode_unref(inode);
+
+ if (gf_uuid_is_null(loc->gfid)) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "%s/%s"
+ " gfid not present",
+ loc->path, entry->d_name);
+ continue;
+ }
+
+ gf_uuid_copy(entry_loc.pargfid, loc->gfid);
+
+ ret = syncop_lookup(this, &entry_loc, &iatt, NULL, NULL, NULL);
+ if (ret) {
if (-ret == ENOENT || -ret == ESTALE) {
- ret = 0;
- goto out;
- }
-
- gf_log (this->name, GF_LOG_ERROR, "Failed to open dir %s, "
- "err:%d", loc->path, -ret);
-
- ret = -1;
- goto out;
- }
-
- fd_bind (fd);
- INIT_LIST_HEAD (&entries.list);
-
- while ((ret = syncop_readdirp (this, fd, 131072, offset, &entries,
- NULL, NULL)) != 0)
- {
- if (ret < 0) {
- if (-ret == ENOENT || -ret == ESTALE) {
- ret = 0;
- goto out;
- }
+ gf_msg(this->name, GF_LOG_INFO, -ret,
+ DHT_MSG_DIR_LOOKUP_FAILED,
+ "Dir:%s renamed or removed. "
+ "Skipping",
+ loc->path);
+ ret = 0;
+ if (conf->decommission_subvols_cnt) {
+ defrag->total_failures++;
+ }
+ continue;
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, -ret,
+ DHT_MSG_DIR_LOOKUP_FAILED, "lookup failed for:%s",
+ entry_loc.path);
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_READDIR_ERROR, "readdirp failed for "
- "path %s. Aborting fix-layout", loc->path);
+ defrag->total_failures++;
+ if (conf->decommission_in_progress) {
+ defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
ret = -1;
goto out;
+ } else {
+ continue;
+ }
}
+ }
- if (list_empty (&entries.list))
- break;
-
- free_entries = _gf_true;
-
- list_for_each_entry_safe (entry, tmp, &entries.list, list) {
- if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
- ret = 1;
- goto out;
- }
-
- offset = entry->d_off;
-
- if (!strcmp (entry->d_name, ".") ||
- !strcmp (entry->d_name, ".."))
- continue;
- if (!IA_ISDIR (entry->d_stat.ia_type)) {
-
- /* If its a fix layout during the attach
- * tier operation do lookups on files
- * on cold subvolume so that there is a
- * CTR DB Lookup Heal triggered on existing
- * data.
- * */
- if (defrag->cmd == GF_DEFRAG_CMD_START_TIER) {
- gf_fix_layout_tier_attach_lookup
- (this, loc, entry);
- }
-
- continue;
- }
- loc_wipe (&entry_loc);
-
- ret = dht_build_child_loc (this, &entry_loc, loc,
- entry->d_name);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "Child loc"
- " build failed for entry: %s",
- entry->d_name);
-
- if (conf->decommission_in_progress) {
- defrag->defrag_status =
- GF_DEFRAG_STATUS_FAILED;
-
- goto out;
- } else {
- should_commit_hash = 0;
-
- continue;
- }
- }
-
- if (gf_uuid_is_null (entry->d_stat.ia_gfid)) {
- gf_log (this->name, GF_LOG_ERROR, "%s/%s"
- " gfid not present", loc->path,
- entry->d_name);
- continue;
- }
+ /* A return value of 2 means, either process_dir or
+ * lookup of a dir failed. Hence, don't commit hash
+ * for the current directory*/
+ ret = gf_defrag_fix_layout(this, defrag, &entry_loc, fix_layout,
+ migrate_data);
- gf_uuid_copy (entry_loc.gfid, entry->d_stat.ia_gfid);
-
- /*In case the gfid stored in the inode by inode_link
- * and the gfid obtained in the lookup differs, then
- * client3_3_lookup_cbk will return ESTALE and proper
- * error will be captured
- */
-
- linked_inode = inode_link (entry_loc.inode, loc->inode,
- entry->d_name,
- &entry->d_stat);
-
- inode = entry_loc.inode;
- entry_loc.inode = linked_inode;
- inode_unref (inode);
-
- if (gf_uuid_is_null (loc->gfid)) {
- gf_log (this->name, GF_LOG_ERROR, "%s/%s"
- " gfid not present", loc->path,
- entry->d_name);
- continue;
- }
-
- gf_uuid_copy (entry_loc.pargfid, loc->gfid);
-
- ret = syncop_lookup (this, &entry_loc, &iatt, NULL,
- NULL, NULL);
- if (ret) {
- if (-ret == ENOENT || -ret == ESTALE) {
- gf_msg (this->name, GF_LOG_INFO, -ret,
- DHT_MSG_DIR_LOOKUP_FAILED,
- "Dir:%s renamed or removed. "
- "Skipping", loc->path);
- ret = 0;
- continue;
- } else {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_DIR_LOOKUP_FAILED,
- "lookup failed for:%s",
- entry_loc.path);
-
- defrag->total_failures++;
-
- if (conf->decommission_in_progress) {
- defrag->defrag_status =
- GF_DEFRAG_STATUS_FAILED;
- ret = -1;
- goto out;
- } else {
- should_commit_hash = 0;
- continue;
- }
- }
- }
-
- /* A return value of 2 means, either process_dir or
- * lookup of a dir failed. Hence, don't commit hash
- * for the current directory*/
-
- ret = gf_defrag_fix_layout (this, defrag, &entry_loc,
- fix_layout, migrate_data);
-
- if (ret && ret != 2) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LAYOUT_FIX_FAILED,
- "Fix layout failed for %s",
- entry_loc.path);
-
- defrag->total_failures++;
+ if (defrag->defrag_status == GF_DEFRAG_STATUS_STOPPED ||
+ defrag->defrag_status == GF_DEFRAG_STATUS_FAILED) {
+ goto out;
+ }
- if (conf->decommission_in_progress) {
- defrag->defrag_status =
- GF_DEFRAG_STATUS_FAILED;
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_LAYOUT_FIX_FAILED,
+ "Fix layout failed for %s", entry_loc.path);
- goto out;
- } else {
- /* Let's not commit-hash if
- * gf_defrag_fix_layout failed*/
- continue;
- }
- }
- }
+ defrag->total_failures++;
- gf_dirent_free (&entries);
- free_entries = _gf_false;
- INIT_LIST_HEAD (&entries.list);
- }
+ if (conf->decommission_in_progress) {
+ defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
- ret = syncop_setxattr (this, loc, fix_layout, 0, NULL, NULL);
- if (ret) {
- if (-ret == ENOENT || -ret == ESTALE) {
- gf_msg (this->name, GF_LOG_INFO, -ret,
- DHT_MSG_LAYOUT_FIX_FAILED,
- "Setxattr failed. Dir %s "
- "renamed or removed",
- loc->path);
- ret = 0;
+ goto out;
} else {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_LAYOUT_FIX_FAILED,
- "Setxattr failed for %s",
- loc->path);
-
- defrag->total_failures++;
-
- if (conf->decommission_in_progress) {
- defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
- ret = -1;
- goto out;
- }
- }
- }
-
- if ((defrag->cmd != GF_DEFRAG_CMD_START_TIER) &&
- (defrag->cmd != GF_DEFRAG_CMD_START_LAYOUT_FIX)) {
- ret = gf_defrag_process_dir (this, defrag, loc, migrate_data,
- &perrno);
-
- if (ret && (ret != 2)) {
- if (perrno == ENOENT || perrno == ESTALE) {
- ret = 0;
- goto out;
- } else {
-
- defrag->total_failures++;
-
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DEFRAG_PROCESS_DIR_FAILED,
- "gf_defrag_process_dir failed for "
- "directory: %s", loc->path);
-
- if (conf->decommission_in_progress) {
- goto out;
- }
-
- should_commit_hash = 0;
- }
- } else if (ret == 2) {
- should_commit_hash = 0;
+ /* Let's not commit-hash if
+ * gf_defrag_fix_layout failed*/
+ continue;
}
- }
-
- gf_msg_trace (this->name, 0, "fix layout called on %s", loc->path);
-
- if (should_commit_hash &&
- gf_defrag_settle_hash (this, defrag, loc, fix_layout) != 0) {
-
+ }
+ }
+
+ gf_dirent_free(&entries);
+ free_entries = _gf_false;
+ INIT_LIST_HEAD(&entries.list);
+ }
+
+ /* A directory layout is fixed only after its subdirs are healed to
+ * any newly added bricks. If the layout is fixed before subdirs are
+ * healed, the newly added brick will get a non-null layout.
+ * Any subdirs which hash to that layout will no longer show up
+ * in a directory listing until they are healed.
+ */
+
+ ret = syncop_setxattr(this, loc, fix_layout, 0, NULL, NULL);
+
+ /* In case of a race where the directory is deleted just before
+ * layout setxattr, the errors are updated in the layout structure.
+ * We can use this information to make a decision whether the directory
+ * is deleted entirely.
+ */
+ if (ret == 0) {
+ ret = dht_dir_layout_error_check(this, loc->inode);
+ ret = -ret;
+ }
+
+ if (ret) {
+ if (-ret == ENOENT || -ret == ESTALE) {
+ gf_msg(this->name, GF_LOG_INFO, -ret, DHT_MSG_LAYOUT_FIX_FAILED,
+ "Setxattr failed. Dir %s "
+ "renamed or removed",
+ loc->path);
+ if (conf->decommission_subvols_cnt) {
defrag->total_failures++;
+ }
+ ret = 0;
+ goto out;
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_LAYOUT_FIX_FAILED,
+ "Setxattr failed for %s", loc->path);
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_SETTLE_HASH_FAILED,
- "Settle hash failed for %s",
- loc->path);
+ defrag->total_failures++;
+ if (conf->decommission_in_progress) {
+ defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
ret = -1;
-
- if (conf->decommission_in_progress) {
- defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
- goto out;
- }
- }
-
- ret = 0;
-out:
- if (free_entries)
- gf_dirent_free (&entries);
-
- loc_wipe (&entry_loc);
-
- if (fd)
- fd_unref (fd);
-
- if (ret == 0 && should_commit_hash == 0) {
- ret = 2;
+ goto out;
+ }
}
+ }
- return ret;
+ if (defrag->cmd != GF_DEFRAG_CMD_START_LAYOUT_FIX) {
+ ret = gf_defrag_process_dir(this, defrag, loc, migrate_data, &perrno);
-}
-
-
-
-/******************************************************************************
- * Tier background Fix layout functions
- ******************************************************************************/
-/* This is the background tier fixlayout thread */
-void *
-gf_tier_do_fix_layout (void *args)
-{
- gf_tier_fix_layout_arg_t *tier_fix_layout_arg = args;
- int ret = -1;
- xlator_t *this = NULL;
- dht_conf_t *conf = NULL;
- gf_defrag_info_t *defrag = NULL;
- dict_t *dict = NULL;
- loc_t loc = {0,};
- struct iatt iatt = {0,};
- struct iatt parent = {0,};
-
- GF_VALIDATE_OR_GOTO ("tier", tier_fix_layout_arg, out);
- GF_VALIDATE_OR_GOTO ("tier", tier_fix_layout_arg->this, out);
- this = tier_fix_layout_arg->this;
-
- conf = this->private;
- GF_VALIDATE_OR_GOTO (this->name, conf, out);
-
- defrag = conf->defrag;
- GF_VALIDATE_OR_GOTO (this->name, defrag, out);
- GF_VALIDATE_OR_GOTO (this->name, defrag->root_inode, out);
-
- GF_VALIDATE_OR_GOTO (this->name, tier_fix_layout_arg->fix_layout, out);
-
-
- /* Get Root loc_t */
- dht_build_root_loc (defrag->root_inode, &loc);
- ret = syncop_lookup (this, &loc, &iatt, &parent, NULL, NULL);
if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_REBALANCE_START_FAILED,
- "Lookup on root failed.");
- ret = -1;
+ if (perrno == ENOENT || perrno == ESTALE) {
+ ret = 0;
goto out;
- }
-
+ } else {
+ defrag->total_failures++;
- /* Start the crawl */
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LOG_TIER_STATUS, "Tiering Fixlayout started");
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ DHT_MSG_DEFRAG_PROCESS_DIR_FAILED,
+ "gf_defrag_process_dir failed for "
+ "directory: %s",
+ loc->path);
- ret = gf_defrag_fix_layout (this, defrag, &loc,
- tier_fix_layout_arg->fix_layout, NULL);
- if (ret && ret != 2) {
- gf_msg (this->name, GF_LOG_ERROR, 0, DHT_MSG_REBALANCE_FAILED,
- "Tiering fixlayout failed.");
- ret = -1;
- goto out;
+ if (conf->decommission_in_progress) {
+ goto out;
+ }
+ }
}
+ }
- if (ret != 2 && gf_defrag_settle_hash
- (this, defrag, &loc,
- tier_fix_layout_arg->fix_layout) != 0) {
- defrag->total_failures++;
- ret = -1;
- goto out;
- }
+ gf_msg_trace(this->name, 0, "fix layout called on %s", loc->path);
- dict = dict_new ();
- if (!dict) {
- ret = -1;
- goto out;
- }
+ if (gf_defrag_settle_hash(this, defrag, loc, fix_layout) != 0) {
+ defrag->total_failures++;
- ret = dict_set_str (dict, GF_XATTR_TIER_LAYOUT_FIXED_KEY, "yes");
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_REBALANCE_FAILED,
- "Failed to set dictionary value: key = %s",
- GF_XATTR_TIER_LAYOUT_FIXED_KEY);
- ret = -1;
- goto out;
- }
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_SETTLE_HASH_FAILED,
+ "Settle hash failed for %s", loc->path);
- /* Marking the completion of tiering fix layout via a xattr on root */
- ret = syncop_setxattr (this, &loc, dict, 0, NULL, NULL);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "Failed to set tiering fix "
- "layout completed xattr on %s", loc.path);
- ret = -1;
- goto out;
+ ret = -1;
+
+ if (conf->decommission_in_progress) {
+ defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
+ goto out;
}
+ }
- ret = 0;
+ ret = 0;
out:
- if (ret && defrag)
- defrag->total_failures++;
+ if (free_entries)
+ gf_dirent_free(&entries);
- if (dict)
- dict_unref (dict);
+ loc_wipe(&entry_loc);
- return NULL;
+ if (fd)
+ fd_unref(fd);
+
+ return ret;
}
int
-gf_tier_start_fix_layout (xlator_t *this,
- loc_t *loc,
- gf_defrag_info_t *defrag,
- dict_t *fix_layout)
+dht_init_local_subvols_and_nodeuuids(xlator_t *this, dht_conf_t *conf,
+ loc_t *loc)
{
- int ret = -1;
- dict_t *tier_dict = NULL;
- gf_tier_fix_layout_arg_t *tier_fix_layout_arg = NULL;
-
- tier_dict = dict_new ();
- if (!tier_dict) {
- gf_log ("tier", GF_LOG_ERROR, "Tier fix layout failed :"
- "Creation of tier_dict failed");
- ret = -1;
- goto out;
- }
-
- /* Check if layout is fixed already */
- ret = syncop_getxattr (this, loc, &tier_dict,
- GF_XATTR_TIER_LAYOUT_FIXED_KEY,
- NULL, NULL);
- if (ret != 0) {
-
- tier_fix_layout_arg = &defrag->tier_conf.tier_fix_layout_arg;
-
- /*Fill crawl arguments */
- tier_fix_layout_arg->this = this;
- tier_fix_layout_arg->fix_layout = fix_layout;
+ dict_t *dict = NULL;
+ uuid_t *uuid_ptr = NULL;
+ int ret = -1;
+ int i = 0;
+ int j = 0;
+
+ /* Find local subvolumes */
+ ret = syncop_getxattr(this, loc, &dict, GF_REBAL_FIND_LOCAL_SUBVOL, NULL,
+ NULL);
+ if (ret && (ret != -ENODATA)) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, 0,
+ "local "
+ "subvolume determination failed with error: %d",
+ -ret);
+ ret = -1;
+ goto out;
+ }
+
+ if (!ret)
+ goto out;
+
+ ret = syncop_getxattr(this, loc, &dict, GF_REBAL_OLD_FIND_LOCAL_SUBVOL,
+ NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, 0,
+ "local "
+ "subvolume determination failed with error: %d",
+ -ret);
+ ret = -1;
+ goto out;
+ }
+ ret = 0;
- /* Spawn the fix layout thread so that its done in the
- * background */
- ret = gf_thread_create (&tier_fix_layout_arg->thread_id, NULL,
- gf_tier_do_fix_layout,
- tier_fix_layout_arg, "tierfixl");
- if (ret) {
- gf_log ("tier", GF_LOG_ERROR, "Thread creation failed. "
- "Background fix layout for tiering will not "
- "work.");
- defrag->total_failures++;
- goto out;
- }
- }
- ret = 0;
out:
- if (tier_dict)
- dict_unref (tier_dict);
-
+ if (ret) {
return ret;
-}
+ }
-void
-gf_tier_clear_fix_layout (xlator_t *this, loc_t *loc, gf_defrag_info_t *defrag)
-{
- int ret = -1;
- dict_t *dict = NULL;
-
- GF_VALIDATE_OR_GOTO ("tier", this, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, defrag, out);
-
- /* Check if background fixlayout is completed. This is not
- * multi-process safe i.e there is a possibility that by the time
- * we move to remove the xattr there it might have been cleared by some
- * other detach process from other node. We ignore the error if such
- * a thing happens */
- ret = syncop_getxattr (this, loc, &dict,
- GF_XATTR_TIER_LAYOUT_FIXED_KEY, NULL, NULL);
- if (ret) {
- /* Background fixlayout not complete - nothing to clear*/
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_LOG_TIER_STATUS,
- "Unable to retrieve fixlayout xattr."
- "Assume background fix layout not complete");
- goto out;
- }
+ for (i = 0; i < conf->local_subvols_cnt; i++) {
+ gf_msg(this->name, GF_LOG_INFO, 0, 0,
+ "local subvol: "
+ "%s",
+ conf->local_subvols[i]->name);
- ret = syncop_removexattr (this, loc, GF_XATTR_TIER_LAYOUT_FIXED_KEY,
- NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_LOG_TIER_STATUS,
- "Failed removing tier fix layout "
- "xattr from %s", loc->path);
- goto out;
+ for (j = 0; j < conf->local_nodeuuids[i].count; j++) {
+ uuid_ptr = &(conf->local_nodeuuids[i].elements[j].uuid);
+ gf_msg(this->name, GF_LOG_INFO, 0, 0, "node uuid : %s",
+ uuid_utoa(*uuid_ptr));
}
- ret = 0;
-out:
- if (dict)
- dict_unref (dict);
-}
+ }
-void
-gf_tier_wait_fix_lookup (gf_defrag_info_t *defrag) {
- if (defrag->tier_conf.tier_fix_layout_arg.thread_id) {
- pthread_join (defrag->tier_conf.tier_fix_layout_arg.thread_id,
- NULL);
- }
+ return ret;
}
-/******************Tier background Fix layout functions END********************/
+/* Functions for the rebalance estimates feature */
uint64_t
-gf_defrag_subvol_file_size (xlator_t *this, loc_t *root_loc)
+gf_defrag_subvol_file_size(xlator_t *this, loc_t *root_loc)
{
- int ret = -1;
- struct statvfs buf = {0,};
-
- if (!this)
- return 0;
-
- ret = syncop_statfs (this, root_loc, &buf, NULL, NULL);
- if (ret) {
- /* Aargh! */
- return 0;
- }
- return ((buf.f_blocks - buf.f_bfree) * buf.f_frsize);
+ int ret = -1;
+ struct statvfs buf = {
+ 0,
+ };
+
+ ret = syncop_statfs(this, root_loc, &buf, NULL, NULL);
+ if (ret) {
+ /* Aargh! */
+ return 0;
+ }
+ return ((buf.f_blocks - buf.f_bfree) * buf.f_frsize);
}
uint64_t
-gf_defrag_subvol_file_cnt (xlator_t *this, loc_t *root_loc)
+gf_defrag_total_file_size(xlator_t *this, loc_t *root_loc)
{
- int ret = -1;
- struct statvfs buf = {0,};
+ dht_conf_t *conf = NULL;
+ int i = 0;
+ uint64_t size_files = 0;
+ uint64_t total_size = 0;
- if (!this)
- return 0;
-
- ret = syncop_statfs (this, root_loc, &buf, NULL, NULL);
- if (ret) {
- /* Aargh! */
- return 0;
- }
- return (buf.f_files - buf.f_ffree);
+ conf = this->private;
+ if (!conf) {
+ return 0;
+ }
+
+ for (i = 0; i < conf->local_subvols_cnt; i++) {
+ size_files = gf_defrag_subvol_file_size(conf->local_subvols[i],
+ root_loc);
+ total_size += size_files;
+ gf_msg(this->name, GF_LOG_INFO, 0, 0,
+ "local subvol: %s,"
+ "cnt = %" PRIu64,
+ conf->local_subvols[i]->name, size_files);
+ }
+
+ gf_msg(this->name, GF_LOG_INFO, 0, 0, "Total size files = %" PRIu64,
+ total_size);
+
+ return total_size;
}
-
-uint64_t
-gf_defrag_total_file_size (xlator_t *this, loc_t *root_loc)
+static void *
+dht_file_counter_thread(void *args)
{
- dht_conf_t *conf = NULL;
- int i = 0;
- uint64_t size_files = 0;
- uint64_t total_size = 0;
-
- conf = this->private;
- if (!conf) {
- return 0;
- }
-
- for (i = 0 ; i < conf->local_subvols_cnt; i++) {
- size_files = gf_defrag_subvol_file_size (conf->local_subvols[i],
- root_loc);
- total_size += size_files;
- gf_msg (this->name, GF_LOG_INFO, 0, 0, "local subvol: %s,"
- "cnt = %"PRIu64, conf->local_subvols[i]->name,
- size_files);
- }
+ gf_defrag_info_t *defrag = NULL;
+ loc_t root_loc = {
+ 0,
+ };
+ struct timespec time_to_wait = {
+ 0,
+ };
+ uint64_t tmp_size = 0;
+
+ if (!args)
+ return NULL;
- gf_msg (this->name, GF_LOG_INFO, 0, 0,
- "Total size files = %"PRIu64, total_size);
+ defrag = (gf_defrag_info_t *)args;
+ dht_build_root_loc(defrag->root_inode, &root_loc);
- return total_size;
-}
+ while (defrag->defrag_status == GF_DEFRAG_STATUS_STARTED) {
+ timespec_now(&time_to_wait);
+ time_to_wait.tv_sec += 600;
+ pthread_mutex_lock(&defrag->fc_mutex);
+ pthread_cond_timedwait(&defrag->fc_wakeup_cond, &defrag->fc_mutex,
+ &time_to_wait);
-uint64_t
-gf_defrag_total_file_cnt (xlator_t *this, loc_t *root_loc)
-{
- dht_conf_t *conf = NULL;
- int i = 0;
- uint64_t num_files = 0;
- uint64_t total_entries = 0;
-
- conf = this->private;
- if (!conf) {
- return 0;
- }
+ pthread_mutex_unlock(&defrag->fc_mutex);
- for (i = 0 ; i < conf->local_subvols_cnt; i++) {
- num_files = gf_defrag_subvol_file_cnt (conf->local_subvols[i],
- root_loc);
- total_entries += num_files;
- gf_msg (this->name, GF_LOG_INFO, 0, 0, "local subvol: %s,"
- "cnt = %"PRIu64, conf->local_subvols[i]->name,
- num_files);
- }
+ if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED)
+ break;
- /* FIXFIXFIX: halve the number of files to negate .glusterfs contents
- We need a better way to figure this out */
+ tmp_size = gf_defrag_total_file_size(defrag->this, &root_loc);
- total_entries = total_entries/2;
- if (total_entries > 20000)
- total_entries += 10000;
+ gf_log("dht", GF_LOG_INFO, "tmp data size =%" PRIu64, tmp_size);
- gf_msg (this->name, GF_LOG_INFO, 0, 0,
- "Total number of files = %"PRIu64, total_entries);
+ if (!tmp_size) {
+ gf_msg("dht", GF_LOG_ERROR, 0, 0,
+ "Failed to get "
+ "the total data size. Unable to estimate "
+ "time to complete rebalance.");
+ } else {
+ g_totalsize = tmp_size;
+ gf_msg_debug("dht", 0, "total data size =%" PRIu64, g_totalsize);
+ }
+ }
- return total_entries;
+ return NULL;
}
+int
+gf_defrag_estimates_cleanup(xlator_t *this, gf_defrag_info_t *defrag,
+ pthread_t filecnt_thread)
+{
+ int ret = -1;
+
+ /* Wake up the filecounter thread.
+ * By now the defrag status will no longer be
+ * GF_DEFRAG_STATUS_STARTED so the thread will exit the loop.
+ */
+ pthread_mutex_lock(&defrag->fc_mutex);
+ {
+ pthread_cond_broadcast(&defrag->fc_wakeup_cond);
+ }
+ pthread_mutex_unlock(&defrag->fc_mutex);
+
+ ret = pthread_join(filecnt_thread, NULL);
+ if (ret) {
+ gf_msg("dht", GF_LOG_ERROR, ret, 0,
+ "file_counter_thread: pthread_join failed.");
+ ret = -1;
+ }
+ return ret;
+}
+int
+gf_defrag_estimates_init(xlator_t *this, loc_t *loc, pthread_t *filecnt_thread)
+{
+ int ret = -1;
+ dht_conf_t *conf = NULL;
+ gf_defrag_info_t *defrag = NULL;
+
+ conf = this->private;
+ defrag = conf->defrag;
+
+ g_totalsize = gf_defrag_total_file_size(this, loc);
+ if (!g_totalsize) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "Failed to get "
+ "the total data size. Unable to estimate "
+ "time to complete rebalance.");
+ goto out;
+ }
+
+ ret = gf_thread_create(filecnt_thread, NULL, dht_file_counter_thread,
+ (void *)defrag, "dhtfcnt");
+
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, ret, 0,
+ "Failed to "
+ "create the file counter thread ");
+ ret = -1;
+ goto out;
+ }
+ ret = 0;
+out:
+ return ret;
+}
+/* Init and cleanup functions for parallel file migration*/
int
-dht_get_local_subvols_and_nodeuuids (xlator_t *this, dht_conf_t *conf,
- loc_t *loc)
+gf_defrag_parallel_migration_init(xlator_t *this, gf_defrag_info_t *defrag,
+ pthread_t **tid_array, int *thread_index)
{
+ int ret = -1;
+ int thread_spawn_count = 0;
+ int index = 0;
+ pthread_t *tid = NULL;
- dict_t *dict = NULL;
- gf_defrag_info_t *defrag = NULL;
- int ret = -1;
+ if (!defrag)
+ goto out;
- defrag = conf->defrag;
+ /* Initialize global entry queue */
+ defrag->queue = GF_CALLOC(1, sizeof(struct dht_container),
+ gf_dht_mt_container_t);
- if (defrag->cmd != GF_DEFRAG_CMD_START_TIER) {
- /* Find local subvolumes */
- ret = syncop_getxattr (this, loc, &dict,
- GF_REBAL_FIND_LOCAL_SUBVOL,
- NULL, NULL);
- if (ret && (ret != -ENODATA)) {
+ if (!defrag->queue) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, 0,
+ "Failed to initialise migration queue");
+ ret = -1;
+ goto out;
+ }
- gf_msg (this->name, GF_LOG_ERROR, -ret, 0, "local "
- "subvolume determination failed with error: %d",
- -ret);
- ret = -1;
- goto out;
- }
+ INIT_LIST_HEAD(&(defrag->queue[0].list));
- if (!ret)
- goto out;
- }
+ thread_spawn_count = MAX(MAX_REBAL_THREADS, 4);
- ret = syncop_getxattr (this, loc, &dict,
- GF_REBAL_OLD_FIND_LOCAL_SUBVOL,
- NULL, NULL);
- if (ret) {
+ gf_msg_debug(this->name, 0, "thread_spawn_count: %d", thread_spawn_count);
- gf_msg (this->name, GF_LOG_ERROR, -ret, 0, "local "
- "subvolume determination failed with error: %d",
- -ret);
- ret = -1;
- goto out;
+ tid = GF_CALLOC(thread_spawn_count, sizeof(pthread_t),
+ gf_common_mt_pthread_t);
+ if (!tid) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, 0,
+ "Failed to create migration threads");
+ ret = -1;
+ goto out;
+ }
+ defrag->current_thread_count = thread_spawn_count;
+
+ /*Spawn Threads Here*/
+ while (index < thread_spawn_count) {
+ ret = gf_thread_create(&(tid[index]), NULL, gf_defrag_task,
+ (void *)defrag, "dhtmig%d", (index + 1) & 0x3ff);
+ if (ret != 0) {
+ gf_msg("DHT", GF_LOG_ERROR, ret, 0, "Thread[%d] creation failed. ",
+ index);
+ ret = -1;
+ goto out;
+ } else {
+ gf_log("DHT", GF_LOG_INFO,
+ "Thread[%d] "
+ "creation successful",
+ index);
}
- ret = 0;
+ index++;
+ }
+
+ ret = 0;
out:
- return ret;
+ *thread_index = index;
+ *tid_array = tid;
+
+ return ret;
}
-static void*
-dht_file_counter_thread (void *args)
+int
+gf_defrag_parallel_migration_cleanup(gf_defrag_info_t *defrag,
+ pthread_t *tid_array, int thread_index)
{
- gf_defrag_info_t *defrag = NULL;
- loc_t root_loc = {0,};
- struct timespec time_to_wait = {0,};
- struct timeval now = {0,};
- uint64_t tmp_size = 0;
-
-
- if (!args)
- return NULL;
+ int ret = -1;
+ int i = 0;
- defrag = (gf_defrag_info_t *) args;
- dht_build_root_loc (defrag->root_inode, &root_loc);
+ if (!defrag)
+ goto out;
- while (defrag->defrag_status == GF_DEFRAG_STATUS_STARTED) {
+ /* Wake up all migration threads */
+ pthread_mutex_lock(&defrag->dfq_mutex);
+ {
+ defrag->crawl_done = 1;
- gettimeofday (&now, NULL);
- time_to_wait.tv_sec = now.tv_sec + 600;
- time_to_wait.tv_nsec = 0;
+ pthread_cond_broadcast(&defrag->parallel_migration_cond);
+ pthread_cond_broadcast(&defrag->df_wakeup_thread);
+ }
+ pthread_mutex_unlock(&defrag->dfq_mutex);
+ /*Wait for all the threads to complete their task*/
+ for (i = 0; i < thread_index; i++) {
+ pthread_join(tid_array[i], NULL);
+ }
- pthread_mutex_lock (&defrag->fc_mutex);
- pthread_cond_timedwait (&defrag->fc_wakeup_cond,
- &defrag->fc_mutex,
- &time_to_wait);
+ GF_FREE(tid_array);
- pthread_mutex_unlock (&defrag->fc_mutex);
-
-
- if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED)
- break;
+ /* Cleanup the migration queue */
+ if (defrag->queue) {
+ gf_dirent_free(defrag->queue[0].df_entry);
+ INIT_LIST_HEAD(&(defrag->queue[0].list));
+ }
- tmp_size = gf_defrag_total_file_size (defrag->this,
- &root_loc);
+ GF_FREE(defrag->queue);
- gf_log ("dht", GF_LOG_INFO,
- "tmp data size =%"PRIu64,
- tmp_size);
-
- if (!tmp_size) {
- gf_msg ("dht", GF_LOG_ERROR, 0, 0, "Failed to get "
- "the total data size. Unable to estimate "
- "time to complete rebalance.");
- } else {
- g_totalsize = tmp_size;
- gf_msg_debug ("dht", 0,
- "total data size =%"PRIu64,
- g_totalsize);
- }
- }
-
- return NULL;
+ ret = 0;
+out:
+ return ret;
}
-
-
int
-gf_defrag_start_crawl (void *data)
+gf_defrag_start_crawl(void *data)
{
- xlator_t *this = NULL;
- dht_conf_t *conf = NULL;
- gf_defrag_info_t *defrag = NULL;
- int ret = -1;
- loc_t loc = {0,};
- struct iatt iatt = {0,};
- struct iatt parent = {0,};
- dict_t *fix_layout = NULL;
- dict_t *migrate_data = NULL;
- dict_t *status = NULL;
- glusterfs_ctx_t *ctx = NULL;
- dht_methods_t *methods = NULL;
- int i = 0;
- int thread_index = 0;
- int err = 0;
- int thread_spawn_count = 0;
- pthread_t *tid = NULL;
- char thread_name[GF_THREAD_NAMEMAX] = {0,};
- pthread_t filecnt_thread;
- gf_boolean_t is_tier_detach = _gf_false;
- call_frame_t *statfs_frame = NULL;
- xlator_t *old_THIS = NULL;
- int j = 0;
- gf_boolean_t fc_thread_started = _gf_false;
- uuid_t *uuid_ptr = NULL;
-
- this = data;
- if (!this)
- goto exit;
-
- ctx = this->ctx;
- if (!ctx)
- goto exit;
-
- conf = this->private;
- if (!conf)
- goto exit;
-
- defrag = conf->defrag;
- if (!defrag)
- goto exit;
-
- gettimeofday (&defrag->start_time, NULL);
- dht_build_root_inode (this, &defrag->root_inode);
- if (!defrag->root_inode)
- goto out;
+ xlator_t *this = NULL;
+ dht_conf_t *conf = NULL;
+ gf_defrag_info_t *defrag = NULL;
+ dict_t *fix_layout = NULL;
+ dict_t *migrate_data = NULL;
+ dict_t *status = NULL;
+ glusterfs_ctx_t *ctx = NULL;
+ call_frame_t *statfs_frame = NULL;
+ xlator_t *old_THIS = NULL;
+ int ret = -1;
+ loc_t loc = {
+ 0,
+ };
+ struct iatt iatt = {
+ 0,
+ };
+ struct iatt parent = {
+ 0,
+ };
+ int thread_index = 0;
+ pthread_t *tid = NULL;
+ pthread_t filecnt_thread;
+ gf_boolean_t fc_thread_started = _gf_false;
+
+ this = data;
+ if (!this)
+ goto exit;
+
+ ctx = this->ctx;
+ if (!ctx)
+ goto exit;
+
+ conf = this->private;
+ if (!conf)
+ goto exit;
+
+ defrag = conf->defrag;
+ if (!defrag)
+ goto exit;
+
+ defrag->start_time = gf_time();
+
+ dht_build_root_inode(this, &defrag->root_inode);
+ if (!defrag->root_inode)
+ goto out;
+
+ dht_build_root_loc(defrag->root_inode, &loc);
+
+ /* fix-layout on '/' first */
+
+ ret = syncop_lookup(this, &loc, &iatt, &parent, NULL, NULL);
+
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_REBALANCE_START_FAILED,
+ "Failed to start rebalance: look up on / failed");
+ ret = -1;
+ goto out;
+ }
- dht_build_root_loc (defrag->root_inode, &loc);
+ old_THIS = THIS;
+ THIS = this;
- /* fix-layout on '/' first */
+ statfs_frame = create_frame(this, this->ctx->pool);
+ if (!statfs_frame) {
+ gf_msg(this->name, GF_LOG_ERROR, DHT_MSG_NO_MEMORY, ENOMEM,
+ "Insufficient memory. Frame creation failed");
+ ret = -1;
+ goto out;
+ }
- ret = syncop_lookup (this, &loc, &iatt, &parent, NULL, NULL);
+ /* async statfs update for honoring min-free-disk */
+ dht_get_du_info(statfs_frame, this, &loc);
+ THIS = old_THIS;
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_REBALANCE_START_FAILED,
- "Failed to start rebalance: look up on / failed");
- ret = -1;
- goto out;
- }
+ fix_layout = dict_new();
+ if (!fix_layout) {
+ ret = -1;
+ goto out;
+ }
+
+ /*
+ * Unfortunately, we can't do special xattrs (like fix.layout) and
+ * real ones in the same call currently, and changing it seems
+ * riskier than just doing two calls.
+ */
+
+ gf_log(this->name, GF_LOG_INFO, "%s using commit hash %u", __func__,
+ conf->vol_commit_hash);
+
+ ret = dict_set_uint32(fix_layout, conf->commithash_xattr_name,
+ conf->vol_commit_hash);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR, "Failed to set %s",
+ conf->commithash_xattr_name);
+ defrag->total_failures++;
+ ret = -1;
+ goto out;
+ }
+
+ ret = syncop_setxattr(this, &loc, fix_layout, 0, NULL, NULL);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Failed to set commit hash on %s. "
+ "Rebalance cannot proceed.",
+ loc.path);
+ defrag->total_failures++;
+ ret = -1;
+ goto out;
+ }
+
+ /* We now return to our regularly scheduled program. */
+
+ ret = dict_set_str(fix_layout, GF_XATTR_FIX_LAYOUT_KEY, "yes");
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_REBALANCE_START_FAILED,
+ "Failed to start rebalance:"
+ "Failed to set dictionary value: key = %s",
+ GF_XATTR_FIX_LAYOUT_KEY);
+ defrag->total_failures++;
+ ret = -1;
+ goto out;
+ }
- old_THIS = THIS;
- THIS = this;
+ defrag->new_commit_hash = conf->vol_commit_hash;
- statfs_frame = create_frame (this, this->ctx->pool);
- if (!statfs_frame) {
- gf_msg (this->name, GF_LOG_ERROR, DHT_MSG_NO_MEMORY, ENOMEM,
- "Insufficient memory. Frame creation failed");
- ret = -1;
- goto out;
- }
+ ret = syncop_setxattr(this, &loc, fix_layout, 0, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_REBALANCE_FAILED,
+ "fix layout on %s failed", loc.path);
+ defrag->total_failures++;
+ ret = -1;
+ goto out;
+ }
- /* async statfs update for honoring min-free-disk */
- dht_get_du_info (statfs_frame, this, &loc);
- THIS = old_THIS;
+ if (defrag->cmd != GF_DEFRAG_CMD_START_LAYOUT_FIX) {
+ /* We need to migrate files */
- fix_layout = dict_new ();
- if (!fix_layout) {
- ret = -1;
- goto out;
+ migrate_data = dict_new();
+ if (!migrate_data) {
+ defrag->total_failures++;
+ ret = -1;
+ goto out;
}
-
- /*
- * Unfortunately, we can't do special xattrs (like fix.layout) and
- * real ones in the same call currently, and changing it seems
- * riskier than just doing two calls.
- */
-
- gf_log (this->name, GF_LOG_INFO, "%s using commit hash %u",
- __func__, conf->vol_commit_hash);
-
- ret = dict_set_uint32 (fix_layout, conf->commithash_xattr_name,
- conf->vol_commit_hash);
+ ret = dict_set_str(
+ migrate_data, GF_XATTR_FILE_MIGRATE_KEY,
+ (defrag->cmd == GF_DEFRAG_CMD_START_FORCE) ? "force" : "non-force");
if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to set %s", conf->commithash_xattr_name);
- defrag->total_failures++;
- ret = -1;
- goto out;
+ defrag->total_failures++;
+ ret = -1;
+ goto out;
}
- ret = syncop_setxattr (this, &loc, fix_layout, 0, NULL, NULL);
+ ret = dht_init_local_subvols_and_nodeuuids(this, conf, &loc);
if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to set commit hash on %s. "
- "Rebalance cannot proceed.",
- loc.path);
- defrag->total_failures++;
- ret = -1;
- goto out;
+ ret = -1;
+ goto out;
}
- /* We now return to our regularly scheduled program. */
-
- ret = dict_set_str (fix_layout, GF_XATTR_FIX_LAYOUT_KEY, "yes");
+ /* Initialise the structures required for parallel migration */
+ ret = gf_defrag_parallel_migration_init(this, defrag, &tid,
+ &thread_index);
if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_REBALANCE_START_FAILED,
- "Failed to start rebalance:"
- "Failed to set dictionary value: key = %s",
- GF_XATTR_FIX_LAYOUT_KEY);
- defrag->total_failures++;
- ret = -1;
- goto out;
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "Aborting rebalance.");
+ goto out;
}
- defrag->new_commit_hash = conf->vol_commit_hash;
-
- ret = syncop_setxattr (this, &loc, fix_layout, 0, NULL, NULL);
+ ret = gf_defrag_estimates_init(this, &loc, &filecnt_thread);
if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_REBALANCE_FAILED,
- "fix layout on %s failed",
- loc.path);
- defrag->total_failures++;
- ret = -1;
- goto out;
- }
-
- if (defrag->cmd != GF_DEFRAG_CMD_START_LAYOUT_FIX) {
- migrate_data = dict_new ();
- if (!migrate_data) {
- defrag->total_failures++;
- ret = -1;
- goto out;
- }
- ret = dict_set_str (migrate_data, GF_XATTR_FILE_MIGRATE_KEY,
- (defrag->cmd == GF_DEFRAG_CMD_START_FORCE)
- ? "force" : "non-force");
- if (ret) {
- defrag->total_failures++;
- ret = -1;
- goto out;
- }
-
- ret = dht_get_local_subvols_and_nodeuuids (this, conf, &loc);
- if (ret) {
- ret = -1;
- goto out;
- }
-
- for (i = 0 ; i < conf->local_subvols_cnt; i++) {
- gf_msg (this->name, GF_LOG_INFO, 0, 0, "local subvols "
- "are %s", conf->local_subvols[i]->name);
-
- for (j = 0; j < conf->local_nodeuuids[i].count; j++) {
- uuid_ptr = &(conf->local_nodeuuids[i].elements[j].uuid);
- gf_msg (this->name, GF_LOG_INFO, 0, 0,
- "node uuids are %s",
- uuid_utoa(*uuid_ptr));
- }
- }
-
- g_totalsize = gf_defrag_total_file_size (this, &loc);
- if (!g_totalsize) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "Failed to get "
- "the total data size. Unable to estimate "
- "time to complete rebalance.");
- }
-
- g_totalfiles = gf_defrag_total_file_cnt (this, &loc);
- if (!g_totalfiles) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "Failed to get "
- "the total number of files. Unable to estimate "
- "time to complete rebalance.");
- }
-
- ret = gf_thread_create (&filecnt_thread, NULL,
- &dht_file_counter_thread,
- (void *)defrag, "dhtfcnt");
-
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, ret, 0, "Failed to "
- "create the file counter thread ");
- ret = 0;
- } else {
- fc_thread_started = _gf_true;
- }
-
-
- /* Initialize global entry queue */
- defrag->queue = GF_CALLOC (1, sizeof (struct dht_container),
- gf_dht_mt_container_t);
-
- if (!defrag->queue) {
- gf_log (this->name, GF_LOG_ERROR, "No memory for "
- "queue");
- ret = -1;
- goto out;
- }
-
- INIT_LIST_HEAD (&(defrag->queue[0].list));
-
- thread_spawn_count = MAX (MAX_REBAL_THREADS, 4);
-
- gf_msg_debug (this->name, 0, "thread_spawn_count: %d",
- thread_spawn_count);
-
- tid = GF_CALLOC (thread_spawn_count, sizeof (pthread_t),
- gf_common_mt_pthread_t);
- if (!tid) {
- gf_log (this->name, GF_LOG_ERROR, "Insufficient memory "
- "for tid");
- ret = -1;
- goto out;
- }
-
- defrag->current_thread_count = thread_spawn_count;
-
- /*Spawn Threads Here*/
- while (thread_index < thread_spawn_count) {
- snprintf (thread_name, sizeof(thread_name),
- "%s%d", "dhtdf", thread_index + 1);
- err = gf_thread_create (&(tid[thread_index]), NULL,
- &gf_defrag_task, (void *)defrag,
- thread_name);
- if (err != 0) {
- gf_log ("DHT", GF_LOG_ERROR,
- "Thread[%d] creation failed. "
- "Aborting Rebalance",
- thread_index);
- ret = -1;
- goto out;
- } else {
- gf_log ("DHT", GF_LOG_INFO, "Thread[%d] "
- "creation successful", thread_index);
- }
- thread_index++;
- }
- }
-
- if (defrag->cmd == GF_DEFRAG_CMD_START_TIER) {
- /* Fix layout for attach tier */
- ret = gf_tier_start_fix_layout (this, &loc, defrag, fix_layout);
- if (ret) {
- goto out;
- }
-
- methods = &(conf->methods);
-
- /* Calling tier_start of tier.c */
- methods->migration_other(this, defrag);
- if (defrag->cmd == GF_DEFRAG_CMD_START_DETACH_TIER ||
- defrag->cmd == GF_DEFRAG_CMD_DETACH_START) {
-
- ret = dict_set_str (migrate_data,
- GF_XATTR_FILE_MIGRATE_KEY,
- "force");
- if (ret)
- goto out;
-
- }
+ /* Not a fatal error. Allow the rebalance to proceed*/
+ ret = 0;
} else {
- ret = gf_defrag_fix_layout (this, defrag, &loc, fix_layout,
- migrate_data);
- if (ret && ret != 2) {
- defrag->total_failures++;
- ret = -1;
- goto out;
- }
-
- if (ret != 2 && gf_defrag_settle_hash
- (this, defrag, &loc, fix_layout) != 0) {
- defrag->total_failures++;
- ret = -1;
- goto out;
- }
-
- if (defrag->cmd == GF_DEFRAG_CMD_START_DETACH_TIER ||
- defrag->cmd == GF_DEFRAG_CMD_DETACH_START)
- is_tier_detach = _gf_true;
-
- }
-
- gf_log ("DHT", GF_LOG_INFO, "crawling file-system completed");
-out:
-
- /* We are here means crawling the entire file system is done
- or something failed. Set defrag->crawl_done flag to intimate
- the migrator threads to exhaust the defrag->queue and terminate*/
-
- if (ret) {
- defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
+ fc_thread_started = _gf_true;
}
+ }
- pthread_mutex_lock (&defrag->dfq_mutex);
- {
- defrag->crawl_done = 1;
-
- pthread_cond_broadcast (
- &defrag->parallel_migration_cond);
- pthread_cond_broadcast (
- &defrag->df_wakeup_thread);
- }
- pthread_mutex_unlock (&defrag->dfq_mutex);
-
- /*Wait for all the threads to complete their task*/
- for (i = 0; i < thread_index; i++) {
- pthread_join (tid[i], NULL);
- }
-
- GF_FREE (tid);
+ ret = gf_defrag_fix_layout(this, defrag, &loc, fix_layout, migrate_data);
+ if (ret) {
+ defrag->total_failures++;
+ ret = -1;
+ goto out;
+ }
- if (defrag->cmd == GF_DEFRAG_CMD_START_TIER) {
- /* Wait for the tier fixlayout to
- * complete if its was started.*/
- gf_tier_wait_fix_lookup (defrag);
- }
+ if (gf_defrag_settle_hash(this, defrag, &loc, fix_layout) != 0) {
+ defrag->total_failures++;
+ ret = -1;
+ goto out;
+ }
- if (is_tier_detach && ret == 0) {
- /* If it was a detach remove the tier fix-layout
- * xattr on root. Ignoring the failure, as nothing has to be
- * done, logging is done in gf_tier_clear_fix_layout */
- gf_tier_clear_fix_layout (this, &loc, defrag);
- }
+ gf_log("DHT", GF_LOG_INFO, "crawling file-system completed");
+out:
- if (defrag->queue) {
- gf_dirent_free (defrag->queue[0].df_entry);
- INIT_LIST_HEAD (&(defrag->queue[0].list));
- }
+ /* We are here means crawling the entire file system is done
+ or something failed. Set defrag->crawl_done flag to intimate
+ the migrator threads to exhaust the defrag->queue and terminate*/
- if ((defrag->defrag_status != GF_DEFRAG_STATUS_STOPPED) &&
- (defrag->defrag_status != GF_DEFRAG_STATUS_FAILED)) {
- defrag->defrag_status = GF_DEFRAG_STATUS_COMPLETE;
- }
+ if (ret) {
+ defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
+ }
- if (fc_thread_started) {
- pthread_mutex_lock (&defrag->fc_mutex);
- {
- pthread_cond_broadcast (&defrag->fc_wakeup_cond);
- }
- pthread_mutex_unlock (&defrag->fc_mutex);
+ gf_defrag_parallel_migration_cleanup(defrag, tid, thread_index);
- pthread_join (filecnt_thread, NULL);
- }
+ if ((defrag->defrag_status != GF_DEFRAG_STATUS_STOPPED) &&
+ (defrag->defrag_status != GF_DEFRAG_STATUS_FAILED)) {
+ defrag->defrag_status = GF_DEFRAG_STATUS_COMPLETE;
+ }
- dht_send_rebalance_event (this, defrag->cmd, defrag->defrag_status);
+ if (fc_thread_started) {
+ gf_defrag_estimates_cleanup(this, defrag, filecnt_thread);
+ }
- LOCK (&defrag->lock);
- {
- status = dict_new ();
- gf_defrag_status_get (conf, status);
- if (ctx && ctx->notify)
- ctx->notify (GF_EN_DEFRAG_STATUS, status);
- if (status)
- dict_unref (status);
- defrag->is_exiting = 1;
- }
- UNLOCK (&defrag->lock);
+ dht_send_rebalance_event(this, defrag->cmd, defrag->defrag_status);
- GF_FREE (defrag->queue);
+ status = dict_new();
+ LOCK(&defrag->lock);
+ {
+ gf_defrag_status_get(conf, status);
+ if (ctx && ctx->notify)
+ ctx->notify(GF_EN_DEFRAG_STATUS, status);
+ if (status)
+ dict_unref(status);
+ defrag->is_exiting = 1;
+ }
+ UNLOCK(&defrag->lock);
- GF_FREE (defrag);
- conf->defrag = NULL;
+ GF_FREE(defrag);
+ conf->defrag = NULL;
- if (migrate_data)
- dict_unref (migrate_data);
+ if (migrate_data)
+ dict_unref(migrate_data);
- if (statfs_frame) {
- STACK_DESTROY (statfs_frame->root);
- }
+ if (statfs_frame) {
+ STACK_DESTROY(statfs_frame->root);
+ }
exit:
- return ret;
+ return ret;
}
-
-
static int
-gf_defrag_done (int ret, call_frame_t *sync_frame, void *data)
+gf_defrag_done(int ret, call_frame_t *sync_frame, void *data)
{
- gf_listener_stop (sync_frame->this);
+ gf_listener_stop(sync_frame->this);
- STACK_DESTROY (sync_frame->root);
- kill (getpid(), SIGTERM);
- return 0;
+ STACK_DESTROY(sync_frame->root);
+ kill(getpid(), SIGTERM);
+ return 0;
}
void *
-gf_defrag_start (void *data)
+gf_defrag_start(void *data)
{
- int ret = -1;
- call_frame_t *frame = NULL;
- dht_conf_t *conf = NULL;
- gf_defrag_info_t *defrag = NULL;
- xlator_t *this = NULL;
- xlator_t *old_THIS = NULL;
-
- this = data;
- conf = this->private;
- if (!conf)
- goto out;
+ int ret = -1;
+ call_frame_t *frame = NULL;
+ dht_conf_t *conf = NULL;
+ gf_defrag_info_t *defrag = NULL;
+ xlator_t *this = NULL;
+ xlator_t *old_THIS = NULL;
- defrag = conf->defrag;
- if (!defrag)
- goto out;
+ this = data;
+ conf = this->private;
+ if (!conf)
+ goto out;
- frame = create_frame (this, this->ctx->pool);
- if (!frame)
- goto out;
+ defrag = conf->defrag;
+ if (!defrag)
+ goto out;
- frame->root->pid = GF_CLIENT_PID_DEFRAG;
+ frame = create_frame(this, this->ctx->pool);
+ if (!frame)
+ goto out;
- defrag->pid = frame->root->pid;
+ frame->root->pid = GF_CLIENT_PID_DEFRAG;
- defrag->defrag_status = GF_DEFRAG_STATUS_STARTED;
+ defrag->pid = frame->root->pid;
- old_THIS = THIS;
- THIS = this;
- ret = synctask_new (this->ctx->env, gf_defrag_start_crawl,
- gf_defrag_done, frame, this);
+ defrag->defrag_status = GF_DEFRAG_STATUS_STARTED;
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_REBALANCE_START_FAILED,
- "Could not create task for rebalance");
- THIS = old_THIS;
-out:
- return NULL;
-}
-
-
-uint64_t
-gf_defrag_get_estimates_based_on_size (dht_conf_t *conf)
-{
- gf_defrag_info_t *defrag = NULL;
- double rate_processed = 0;
- uint64_t total_processed = 0;
- uint64_t tmp_count = 0;
- uint64_t time_to_complete = 0;
- struct timeval now = {0,};
- double elapsed = 0;
-
- defrag = conf->defrag;
-
- if (!g_totalsize)
- goto out;
-
- gettimeofday (&now, NULL);
- elapsed = now.tv_sec - defrag->start_time.tv_sec;
-
- /* Don't calculate the estimates for the first 10 minutes.
- * It is unlikely to be accurate and estimates are not required
- * if the process finishes in less than 10 mins.
- */
-
- if (elapsed < ESTIMATE_START_INTERVAL) {
- gf_msg (THIS->name, GF_LOG_INFO, 0, 0,
- "Rebalance estimates will not be available for the "
- "first %d seconds.", ESTIMATE_START_INTERVAL);
-
- goto out;
- }
-
- total_processed = defrag->size_processed;
-
- /* rate at which files processed */
- rate_processed = (total_processed)/elapsed;
-
- tmp_count = g_totalsize;
-
- if (rate_processed) {
- time_to_complete = (tmp_count)/rate_processed;
-
- } else {
- gf_msg (THIS->name, GF_LOG_ERROR, 0, 0,
- "Unable to calculate estimated time for rebalance");
- }
-
- gf_log (THIS->name, GF_LOG_INFO,
- "TIME: (size) total_processed=%"PRIu64" tmp_cnt = %"PRIu64","
- "rate_processed=%f, elapsed = %f", total_processed, tmp_count,
- rate_processed, elapsed);
+ old_THIS = THIS;
+ THIS = this;
+ ret = synctask_new(this->ctx->env, gf_defrag_start_crawl, gf_defrag_done,
+ frame, this);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_REBALANCE_START_FAILED,
+ "Could not create task for rebalance");
+ THIS = old_THIS;
out:
- return time_to_complete;
+ return NULL;
}
-
-
uint64_t
-gf_defrag_get_estimates (dht_conf_t *conf)
+gf_defrag_get_estimates_based_on_size(dht_conf_t *conf)
{
- gf_defrag_info_t *defrag = NULL;
- loc_t loc = {0,};
- double rate_lookedup = 0;
- uint64_t dirs_processed = 0;
- uint64_t files_processed = 0;
- uint64_t total_processed = 0;
- uint64_t tmp_count = 0;
- uint64_t time_to_complete = 0;
- struct timeval now = {0,};
- double elapsed = 0;
+ gf_defrag_info_t *defrag = NULL;
+ double rate_processed = 0;
+ uint64_t total_processed = 0;
+ uint64_t tmp_count = 0;
+ uint64_t time_to_complete = 0;
+ double elapsed = 0;
+ defrag = conf->defrag;
- defrag = conf->defrag;
+ if (!g_totalsize)
+ goto out;
- if (!g_totalfiles)
- goto out;
+ elapsed = gf_time() - defrag->start_time;
- gettimeofday (&now, NULL);
- elapsed = now.tv_sec - defrag->start_time.tv_sec;
+ /* Don't calculate the estimates for the first 10 minutes.
+ * It is unlikely to be accurate and estimates are not required
+ * if the process finishes in less than 10 mins.
+ */
- /* I tried locking before accessing num_files_lookedup and
- * num_dirs_processed but the status function
- * never seemed to get the lock, causing the status cli to
- * hang.
- */
+ if (elapsed < ESTIMATE_START_INTERVAL) {
+ gf_msg(THIS->name, GF_LOG_INFO, 0, 0,
+ "Rebalance estimates will not be available for the "
+ "first %d seconds.",
+ ESTIMATE_START_INTERVAL);
- dirs_processed = defrag->num_dirs_processed;
- files_processed = defrag->num_files_lookedup;
+ goto out;
+ }
- total_processed = files_processed + dirs_processed;
+ total_processed = defrag->size_processed;
- if (total_processed > g_totalfiles) {
- /* lookup the number of files again
- * The problem here is that not all the newly added files
- * might need to be processed. So this need not work
- * in some cases
- */
- dht_build_root_loc (defrag->root_inode, &loc);
- g_totalfiles = gf_defrag_total_file_cnt (defrag->this, &loc);
- if (!g_totalfiles)
- goto out;
- }
+ /* rate at which files processed */
+ rate_processed = (total_processed) / elapsed;
- /* rate at which files looked up */
- rate_lookedup = (total_processed)/elapsed;
+ tmp_count = g_totalsize;
- /* We initially sum up dirs across all local subvols because we get the
- * file count from the inodes on each subvol.
- * The same directories will be counted for each subvol but
- * we want them to be counted once.
- */
+ if (rate_processed) {
+ time_to_complete = (tmp_count) / rate_processed;
- tmp_count = g_totalfiles
- - (dirs_processed * (conf->local_subvols_cnt - 1));
+ } else {
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, 0,
+ "Unable to calculate estimated time for rebalance");
+ }
- if (rate_lookedup) {
- time_to_complete = (tmp_count)/rate_lookedup;
-
- } else {
-
- gf_msg (THIS->name, GF_LOG_ERROR, 0, 0,
- "Unable to calculate estimated time for rebalance");
- }
-
- gf_log (THIS->name, GF_LOG_INFO,
- "TIME: (count) total_processed=%"PRIu64" tmp_cnt = %"PRIu64","
- "rate_lookedup=%f", total_processed, tmp_count,
- rate_lookedup);
+ gf_log(THIS->name, GF_LOG_INFO,
+ "TIME: (size) total_processed=%" PRIu64 " tmp_cnt = %" PRIu64
+ ","
+ "rate_processed=%f, elapsed = %f",
+ total_processed, tmp_count, rate_processed, elapsed);
out:
- return time_to_complete;
+ return time_to_complete;
}
-
int
-gf_defrag_status_get (dht_conf_t *conf, dict_t *dict)
+gf_defrag_status_get(dht_conf_t *conf, dict_t *dict)
{
- int ret = 0;
- uint64_t files = 0;
- uint64_t size = 0;
- uint64_t lookup = 0;
- uint64_t failures = 0;
- uint64_t skipped = 0;
- uint64_t promoted = 0;
- uint64_t demoted = 0;
- char *status = "";
- double elapsed = 0;
- struct timeval end = {0,};
- uint64_t time_to_complete = 0;
- uint64_t time_left = 0;
- gf_defrag_info_t *defrag = conf->defrag;
-
- if (!defrag)
- goto out;
-
- ret = 0;
- if (defrag->defrag_status == GF_DEFRAG_STATUS_NOT_STARTED)
- goto out;
-
- files = defrag->total_files;
- size = defrag->total_data;
- lookup = defrag->num_files_lookedup;
- failures = defrag->total_failures;
- skipped = defrag->skipped;
- promoted = defrag->total_files_promoted;
- demoted = defrag->total_files_demoted;
-
- gettimeofday (&end, NULL);
-
- elapsed = end.tv_sec - defrag->start_time.tv_sec;
-
-
- /* The rebalance is still in progress */
-
- if ((defrag->cmd != GF_DEFRAG_CMD_START_TIER)
- && (defrag->defrag_status == GF_DEFRAG_STATUS_STARTED)) {
-
-/*
- time_to_complete = gf_defrag_get_estimates (conf);
-
- if (time_to_complete && (time_to_complete > elapsed))
- time_left = time_to_complete - elapsed;
-
- gf_log (THIS->name, GF_LOG_INFO,
- "TIME: Estimated total time to complete based on"
- " count = %"PRIu64 " seconds, seconds left = %"PRIu64"",
- time_to_complete, time_left);
-*/
-
- time_to_complete = gf_defrag_get_estimates_based_on_size (conf);
-
- if (time_to_complete && (time_to_complete > elapsed))
- time_left = time_to_complete - elapsed;
-
- gf_log (THIS->name, GF_LOG_INFO,
- "TIME: Estimated total time to complete (size)= %"PRIu64
- " seconds, seconds left = %"PRIu64"",
- time_to_complete, time_left);
- }
-
- if (!dict)
- goto log;
-
- ret = dict_set_uint64 (dict, "promoted", promoted);
- if (ret)
- gf_log (THIS->name, GF_LOG_WARNING,
- "failed to set promoted count");
-
- ret = dict_set_uint64 (dict, "demoted", demoted);
- if (ret)
- gf_log (THIS->name, GF_LOG_WARNING,
- "failed to set demoted count");
-
- ret = dict_set_uint64 (dict, "files", files);
- if (ret)
- gf_log (THIS->name, GF_LOG_WARNING,
- "failed to set file count");
-
- ret = dict_set_uint64 (dict, "size", size);
- if (ret)
- gf_log (THIS->name, GF_LOG_WARNING,
- "failed to set size of xfer");
-
- ret = dict_set_uint64 (dict, "lookups", lookup);
- if (ret)
- gf_log (THIS->name, GF_LOG_WARNING,
- "failed to set lookedup file count");
-
-
- ret = dict_set_int32 (dict, "status", defrag->defrag_status);
- if (ret)
- gf_log (THIS->name, GF_LOG_WARNING,
- "failed to set status");
-
- ret = dict_set_double (dict, "run-time", elapsed);
- if (ret)
- gf_log (THIS->name, GF_LOG_WARNING,
- "failed to set run-time");
-
- ret = dict_set_uint64 (dict, "failures", failures);
- if (ret)
- gf_log (THIS->name, GF_LOG_WARNING,
- "failed to set failure count");
-
- ret = dict_set_uint64 (dict, "skipped", skipped);
- if (ret)
- gf_log (THIS->name, GF_LOG_WARNING,
- "failed to set skipped file count");
-
- ret = dict_set_uint64 (dict, "time-left", time_left);
- if (ret)
- gf_log (THIS->name, GF_LOG_WARNING,
- "failed to set time-left");
+ int ret = 0;
+ uint64_t files = 0;
+ uint64_t size = 0;
+ uint64_t lookup = 0;
+ uint64_t failures = 0;
+ uint64_t skipped = 0;
+ char *status = "";
+ double elapsed = 0;
+ uint64_t time_to_complete = 0;
+ uint64_t time_left = 0;
+ gf_defrag_info_t *defrag = conf->defrag;
+
+ if (!defrag)
+ goto out;
+
+ ret = 0;
+ if (defrag->defrag_status == GF_DEFRAG_STATUS_NOT_STARTED)
+ goto out;
+
+ files = defrag->total_files;
+ size = defrag->total_data;
+ lookup = defrag->num_files_lookedup;
+ failures = defrag->total_failures;
+ skipped = defrag->skipped;
+
+ elapsed = gf_time() - defrag->start_time;
+
+ /* The rebalance is still in progress */
+
+ if (defrag->defrag_status == GF_DEFRAG_STATUS_STARTED) {
+ time_to_complete = gf_defrag_get_estimates_based_on_size(conf);
+
+ if (time_to_complete && (time_to_complete > elapsed))
+ time_left = time_to_complete - elapsed;
+
+ gf_log(THIS->name, GF_LOG_INFO,
+ "TIME: Estimated total time to complete (size)= %" PRIu64
+ " seconds, seconds left = %" PRIu64 "",
+ time_to_complete, time_left);
+ }
+
+ if (!dict)
+ goto log;
+
+ ret = dict_set_uint64(dict, "files", files);
+ if (ret)
+ gf_log(THIS->name, GF_LOG_WARNING, "failed to set file count");
+
+ ret = dict_set_uint64(dict, "size", size);
+ if (ret)
+ gf_log(THIS->name, GF_LOG_WARNING, "failed to set size of xfer");
+
+ ret = dict_set_uint64(dict, "lookups", lookup);
+ if (ret)
+ gf_log(THIS->name, GF_LOG_WARNING, "failed to set lookedup file count");
+
+ ret = dict_set_int32(dict, "status", defrag->defrag_status);
+ if (ret)
+ gf_log(THIS->name, GF_LOG_WARNING, "failed to set status");
+
+ ret = dict_set_double(dict, "run-time", elapsed);
+ if (ret)
+ gf_log(THIS->name, GF_LOG_WARNING, "failed to set run-time");
+
+ ret = dict_set_uint64(dict, "failures", failures);
+ if (ret)
+ gf_log(THIS->name, GF_LOG_WARNING, "failed to set failure count");
+
+ ret = dict_set_uint64(dict, "skipped", skipped);
+ if (ret)
+ gf_log(THIS->name, GF_LOG_WARNING, "failed to set skipped file count");
+
+ ret = dict_set_uint64(dict, "time-left", time_left);
+ if (ret)
+ gf_log(THIS->name, GF_LOG_WARNING, "failed to set time-left");
log:
- switch (defrag->defrag_status) {
+ switch (defrag->defrag_status) {
case GF_DEFRAG_STATUS_NOT_STARTED:
- status = "not started";
- break;
+ status = "not started";
+ break;
case GF_DEFRAG_STATUS_STARTED:
- status = "in progress";
- break;
+ status = "in progress";
+ break;
case GF_DEFRAG_STATUS_STOPPED:
- status = "stopped";
- break;
+ status = "stopped";
+ break;
case GF_DEFRAG_STATUS_COMPLETE:
- status = "completed";
- break;
+ status = "completed";
+ break;
case GF_DEFRAG_STATUS_FAILED:
- status = "failed";
- break;
+ status = "failed";
+ break;
default:
- break;
- }
-
- gf_msg (THIS->name, GF_LOG_INFO, 0, DHT_MSG_REBALANCE_STATUS,
- "Rebalance is %s. Time taken is %.2f secs",
- status, elapsed);
- gf_msg (THIS->name, GF_LOG_INFO, 0, DHT_MSG_REBALANCE_STATUS,
- "Files migrated: %"PRIu64", size: %"
- PRIu64", lookups: %"PRIu64", failures: %"PRIu64", skipped: "
- "%"PRIu64, files, size, lookup, failures, skipped);
+ break;
+ }
+
+ gf_msg(THIS->name, GF_LOG_INFO, 0, DHT_MSG_REBALANCE_STATUS,
+ "Rebalance is %s. Time taken is %.2f secs", status, elapsed);
+ gf_msg(THIS->name, GF_LOG_INFO, 0, DHT_MSG_REBALANCE_STATUS,
+ "Files migrated: %" PRIu64 ", size: %" PRIu64 ", lookups: %" PRIu64
+ ", failures: %" PRIu64
+ ", skipped: "
+ "%" PRIu64,
+ files, size, lookup, failures, skipped);
out:
- return 0;
-}
-
-void
-gf_defrag_set_pause_state (gf_tier_conf_t *tier_conf, tier_pause_state_t state)
-{
- pthread_mutex_lock (&tier_conf->pause_mutex);
- tier_conf->pause_state = state;
- pthread_mutex_unlock (&tier_conf->pause_mutex);
-}
-
-
-tier_pause_state_t
-gf_defrag_get_pause_state (gf_tier_conf_t *tier_conf)
-{
- int state;
-
- pthread_mutex_lock (&tier_conf->pause_mutex);
- state = tier_conf->pause_state;
- pthread_mutex_unlock (&tier_conf->pause_mutex);
-
- return state;
-}
-
-tier_pause_state_t
-gf_defrag_check_pause_tier (gf_tier_conf_t *tier_conf)
-{
- int woke = 0;
- int state = -1;
-
- pthread_mutex_lock (&tier_conf->pause_mutex);
-
- if (tier_conf->pause_state == TIER_RUNNING)
- goto out;
-
- if (tier_conf->pause_state == TIER_PAUSED)
- goto out;
-
- if (tier_conf->promote_in_progress ||
- tier_conf->demote_in_progress)
- goto out;
-
- tier_conf->pause_state = TIER_PAUSED;
-
- if (tier_conf->pause_synctask) {
- synctask_wake (tier_conf->pause_synctask);
- tier_conf->pause_synctask = 0;
- woke = 1;
- }
-
- gf_msg ("tier", GF_LOG_DEBUG, 0,
- DHT_MSG_TIER_PAUSED,
- "woken %d", woke);
-
- gf_event (EVENT_TIER_PAUSE, "vol=%s", tier_conf->volname);
-out:
- state = tier_conf->pause_state;
-
- pthread_mutex_unlock (&tier_conf->pause_mutex);
-
- return state;
-}
-
-void
-gf_defrag_pause_tier_timeout (void *data)
-{
- xlator_t *this = NULL;
- dht_conf_t *conf = NULL;
- gf_defrag_info_t *defrag = NULL;
-
- this = (xlator_t *) data;
- GF_VALIDATE_OR_GOTO ("tier", this, out);
-
- conf = this->private;
- GF_VALIDATE_OR_GOTO (this->name, conf, out);
-
- defrag = conf->defrag;
- GF_VALIDATE_OR_GOTO (this->name, defrag, out);
-
- gf_msg (this->name, GF_LOG_DEBUG, 0,
- DHT_MSG_TIER_PAUSED,
- "Request pause timer timeout");
-
- gf_defrag_check_pause_tier (&defrag->tier_conf);
-
-out:
- return;
-}
-
-int
-gf_defrag_pause_tier (xlator_t *this, gf_defrag_info_t *defrag)
-{
- int ret = 0;
- struct timespec delta = {0,};
- int delay = 2;
-
- if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED)
- goto out;
-
- /*
- * Set flag requesting to pause tiering. Wait 'delay' seconds for
- * tiering to actually stop as indicated by the pause state
- * before returning success or failure.
- */
- gf_defrag_set_pause_state (&defrag->tier_conf, TIER_REQUEST_PAUSE);
-
- /*
- * If migration is not underway, can pause immediately.
- */
- gf_defrag_check_pause_tier (&defrag->tier_conf);
- if (gf_defrag_get_pause_state (&defrag->tier_conf) == TIER_PAUSED)
- goto out;
-
- gf_msg (this->name, GF_LOG_DEBUG, 0,
- DHT_MSG_TIER_PAUSED,
- "Request pause tier");
-
- defrag->tier_conf.pause_synctask = synctask_get ();
- delta.tv_sec = delay;
- delta.tv_nsec = 0;
- defrag->tier_conf.pause_timer =
- gf_timer_call_after (this->ctx, delta,
- gf_defrag_pause_tier_timeout,
- this);
-
- synctask_yield (defrag->tier_conf.pause_synctask);
-
- if (gf_defrag_get_pause_state (&defrag->tier_conf) == TIER_PAUSED)
- goto out;
-
- gf_defrag_set_pause_state (&defrag->tier_conf, TIER_RUNNING);
-
- ret = -1;
-out:
-
- gf_msg (this->name, GF_LOG_DEBUG, 0,
- DHT_MSG_TIER_PAUSED,
- "Pause tiering ret=%d", ret);
-
- return ret;
-}
-
-int
-gf_defrag_resume_tier (xlator_t *this, gf_defrag_info_t *defrag)
-{
- gf_msg (this->name, GF_LOG_DEBUG, 0,
- DHT_MSG_TIER_RESUME,
- "Pause end. Resume tiering");
-
- gf_defrag_set_pause_state (&defrag->tier_conf, TIER_RUNNING);
-
- gf_event (EVENT_TIER_RESUME, "vol=%s", defrag->tier_conf.volname);
-
- return 0;
+ return 0;
}
int
-gf_defrag_start_detach_tier (gf_defrag_info_t *defrag)
+gf_defrag_stop(dht_conf_t *conf, gf_defrag_status_t status, dict_t *output)
{
- defrag->cmd = GF_DEFRAG_CMD_START_DETACH_TIER;
+ /* TODO: set a variable 'stop_defrag' here, it should be checked
+ in defrag loop */
+ int ret = -1;
+ gf_defrag_info_t *defrag = conf->defrag;
- return 0;
-}
+ GF_ASSERT(defrag);
-int
-gf_defrag_stop (dht_conf_t *conf, gf_defrag_status_t status,
- dict_t *output)
-{
- /* TODO: set a variable 'stop_defrag' here, it should be checked
- in defrag loop */
- int ret = -1;
- gf_defrag_info_t *defrag = conf->defrag;
+ if (defrag->defrag_status == GF_DEFRAG_STATUS_NOT_STARTED) {
+ goto out;
+ }
- GF_ASSERT (defrag);
+ gf_msg("", GF_LOG_INFO, 0, DHT_MSG_REBALANCE_STOPPED,
+ "Received stop command on rebalance");
+ defrag->defrag_status = status;
- if (defrag->defrag_status == GF_DEFRAG_STATUS_NOT_STARTED) {
- goto out;
- }
-
- gf_msg ("", GF_LOG_INFO, 0, DHT_MSG_REBALANCE_STOPPED,
- "Received stop command on rebalance");
- defrag->defrag_status = status;
-
- if (output)
- gf_defrag_status_get (conf, output);
- ret = 0;
+ if (output)
+ gf_defrag_status_get(conf, output);
+ ret = 0;
out:
- gf_msg_debug ("", 0, "Returning %d", ret);
- return ret;
+ gf_msg_debug("", 0, "Returning %d", ret);
+ return ret;
}
diff --git a/xlators/cluster/dht/src/dht-rename.c b/xlators/cluster/dht/src/dht-rename.c
index 3dc042e3ce5..d9dbf50492f 100644
--- a/xlators/cluster/dht/src/dht-rename.c
+++ b/xlators/cluster/dht/src/dht-rename.c
@@ -11,486 +11,454 @@
/* TODO: link(oldpath, newpath) fails if newpath already exists. DHT should
* delete the newpath if it gets EEXISTS from link() call.
*/
-#include "glusterfs.h"
-#include "xlator.h"
#include "dht-common.h"
#include "dht-lock.h"
-#include "defaults.h"
+#include <glusterfs/defaults.h>
-int dht_rename_unlock (call_frame_t *frame, xlator_t *this);
+int
+dht_rename_unlock(call_frame_t *frame, xlator_t *this);
+int32_t
+dht_rename_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata);
int
-dht_rename_unlock_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xdata)
+dht_rename_unlock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
+ dht_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- dht_set_fixed_dir_stat (&local->preoldparent);
- dht_set_fixed_dir_stat (&local->postoldparent);
- dht_set_fixed_dir_stat (&local->preparent);
- dht_set_fixed_dir_stat (&local->postparent);
+ dht_set_fixed_dir_stat(&local->preoldparent);
+ dht_set_fixed_dir_stat(&local->postoldparent);
+ dht_set_fixed_dir_stat(&local->preparent);
+ dht_set_fixed_dir_stat(&local->postparent);
- if (IA_ISREG (local->stbuf.ia_type))
- DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+ if (IA_ISREG(local->stbuf.ia_type))
+ DHT_STRIP_PHASE1_FLAGS(&local->stbuf);
- DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
- &local->stbuf, &local->preoldparent,
- &local->postoldparent, &local->preparent,
- &local->postparent, local->xattr);
- return 0;
+ DHT_STACK_UNWIND(rename, frame, local->op_ret, local->op_errno,
+ &local->stbuf, &local->preoldparent, &local->postoldparent,
+ &local->preparent, &local->postparent, local->xattr);
+ return 0;
}
static void
-dht_rename_unlock_src (call_frame_t *frame, xlator_t *this)
+dht_rename_dir_unlock_src(call_frame_t *frame, xlator_t *this)
{
- dht_local_t *local = NULL;
+ dht_local_t *local = NULL;
- local = frame->local;
- dht_unlock_namespace (frame, &local->lock[0]);
- return;
+ local = frame->local;
+ dht_unlock_namespace(frame, &local->lock[0]);
+ return;
}
static void
-dht_rename_unlock_dst (call_frame_t *frame, xlator_t *this)
+dht_rename_dir_unlock_dst(call_frame_t *frame, xlator_t *this)
{
- dht_local_t *local = NULL;
- int op_ret = -1;
- char src_gfid[GF_UUID_BUF_SIZE] = {0};
- char dst_gfid[GF_UUID_BUF_SIZE] = {0};
-
- local = frame->local;
-
- /* Unlock entrylk */
- dht_unlock_entrylk_wrapper (frame, &local->lock[1].ns.directory_ns);
-
- /* Unlock inodelk */
- op_ret = dht_unlock_inodelk (frame,
- local->lock[1].ns.parent_layout.locks,
- local->lock[1].ns.parent_layout.lk_count,
- dht_rename_unlock_cbk);
- if (op_ret < 0) {
- uuid_utoa_r (local->loc.inode->gfid, src_gfid);
-
- if (local->loc2.inode)
- uuid_utoa_r (local->loc2.inode->gfid, dst_gfid);
-
- if (IA_ISREG (local->stbuf.ia_type))
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_UNLOCKING_FAILED,
- "winding unlock inodelk failed "
- "rename (%s:%s:%s %s:%s:%s), "
- "stale locks left on bricks",
- local->loc.path, src_gfid,
- local->src_cached->name,
- local->loc2.path, dst_gfid,
- local->dst_cached ?
- local->dst_cached->name : NULL);
- else
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_UNLOCKING_FAILED,
- "winding unlock inodelk failed "
- "rename (%s:%s %s:%s), "
- "stale locks left on bricks",
- local->loc.path, src_gfid,
- local->loc2.path, dst_gfid);
-
- dht_rename_unlock_cbk (frame, NULL, this, 0, 0, NULL);
- }
+ dht_local_t *local = NULL;
+ int op_ret = -1;
+ char src_gfid[GF_UUID_BUF_SIZE] = {0};
+ char dst_gfid[GF_UUID_BUF_SIZE] = {0};
+
+ local = frame->local;
+
+ /* Unlock entrylk */
+ dht_unlock_entrylk_wrapper(frame, &local->lock[1].ns.directory_ns);
+
+ /* Unlock inodelk */
+ op_ret = dht_unlock_inodelk(frame, local->lock[1].ns.parent_layout.locks,
+ local->lock[1].ns.parent_layout.lk_count,
+ dht_rename_unlock_cbk);
+ if (op_ret < 0) {
+ uuid_utoa_r(local->loc.inode->gfid, src_gfid);
+
+ if (local->loc2.inode)
+ uuid_utoa_r(local->loc2.inode->gfid, dst_gfid);
+
+ if (IA_ISREG(local->stbuf.ia_type))
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_UNLOCKING_FAILED,
+ "winding unlock inodelk failed "
+ "rename (%s:%s:%s %s:%s:%s), "
+ "stale locks left on bricks",
+ local->loc.path, src_gfid, local->src_cached->name,
+ local->loc2.path, dst_gfid,
+ local->dst_cached ? local->dst_cached->name : NULL);
+ else
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_UNLOCKING_FAILED,
+ "winding unlock inodelk failed "
+ "rename (%s:%s %s:%s), "
+ "stale locks left on bricks",
+ local->loc.path, src_gfid, local->loc2.path, dst_gfid);
+
+ dht_rename_unlock_cbk(frame, NULL, this, 0, 0, NULL);
+ }
- return;
+ return;
}
static int
-dht_rename_dir_unlock (call_frame_t *frame, xlator_t *this)
+dht_rename_dir_unlock(call_frame_t *frame, xlator_t *this)
{
-
- dht_rename_unlock_src (frame, this);
- dht_rename_unlock_dst (frame, this);
- return 0;
+ dht_rename_dir_unlock_src(frame, this);
+ dht_rename_dir_unlock_dst(frame, this);
+ return 0;
}
int
-dht_rename_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent,
- dict_t *xdata)
+dht_rename_dir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
+ struct iatt *preoldparent, struct iatt *postoldparent,
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
{
- dht_conf_t *conf = NULL;
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- xlator_t *prev = NULL;
- int i = 0;
- char gfid[GF_UUID_BUF_SIZE] = {0};
- int subvol_cnt = -1;
-
- conf = this->private;
- local = frame->local;
- prev = cookie;
- subvol_cnt = dht_subvol_cnt (this, prev);
- local->ret_cache[subvol_cnt] = op_ret;
-
- if (op_ret == -1) {
- gf_uuid_unparse(local->loc.inode->gfid, gfid);
-
- gf_msg (this->name, GF_LOG_INFO, op_errno,
- DHT_MSG_RENAME_FAILED,
- "Rename %s -> %s on %s failed, (gfid = %s)",
- local->loc.path, local->loc2.path,
- prev->name, gfid);
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto unwind;
- }
- /* TODO: construct proper stbuf for dir */
- /*
- * FIXME: is this the correct way to build stbuf and
- * parent bufs?
- */
- dht_iatt_merge (this, &local->stbuf, stbuf, prev);
- dht_iatt_merge (this, &local->preoldparent, preoldparent, prev);
- dht_iatt_merge (this, &local->postoldparent, postoldparent, prev);
- dht_iatt_merge (this, &local->preparent, prenewparent, prev);
- dht_iatt_merge (this, &local->postparent, postnewparent, prev);
+ dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ xlator_t *prev = NULL;
+ int i = 0;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ int subvol_cnt = -1;
+
+ conf = this->private;
+ local = frame->local;
+ prev = cookie;
+ subvol_cnt = dht_subvol_cnt(this, prev);
+ local->ret_cache[subvol_cnt] = op_ret;
+
+ if (op_ret == -1) {
+ gf_uuid_unparse(local->loc.inode->gfid, gfid);
+
+ gf_msg(this->name, GF_LOG_INFO, op_errno, DHT_MSG_RENAME_FAILED,
+ "Rename %s -> %s on %s failed, (gfid = %s)", local->loc.path,
+ local->loc2.path, prev->name, gfid);
+
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto unwind;
+ }
+ /* TODO: construct proper stbuf for dir */
+ /*
+ * FIXME: is this the correct way to build stbuf and
+ * parent bufs?
+ */
+ dht_iatt_merge(this, &local->stbuf, stbuf);
+ dht_iatt_merge(this, &local->preoldparent, preoldparent);
+ dht_iatt_merge(this, &local->postoldparent, postoldparent);
+ dht_iatt_merge(this, &local->preparent, prenewparent);
+ dht_iatt_merge(this, &local->postparent, postnewparent);
unwind:
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- /* We get here with local->call_cnt == 0. Which means
- * we are the only one executing this code, there is
- * no contention. Therefore it's safe to manipulate or
- * deref local->call_cnt directly (without locking).
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt)) {
+ /* We get here with local->call_cnt == 0. Which means
+ * we are the only one executing this code, there is
+ * no contention. Therefore it's safe to manipulate or
+ * deref local->call_cnt directly (without locking).
+ */
+ if (local->ret_cache[conf->subvolume_cnt] == 0) {
+ /* count errant subvols in last field of ret_cache */
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (local->ret_cache[i] != 0)
+ ++local->ret_cache[conf->subvolume_cnt];
+ }
+ if (local->ret_cache[conf->subvolume_cnt]) {
+ /* undoing the damage:
+ * for all subvolumes, where rename
+ * succeeded, we perform the reverse operation
*/
- if (local->ret_cache[conf->subvolume_cnt] == 0) {
- /* count errant subvols in last field of ret_cache */
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (local->ret_cache[i] != 0)
- ++local->ret_cache[conf->subvolume_cnt];
- }
- if (local->ret_cache[conf->subvolume_cnt]) {
- /* undoing the damage:
- * for all subvolumes, where rename
- * succeeded, we perform the reverse operation
- */
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (local->ret_cache[i] == 0)
- ++local->call_cnt;
- }
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (local->ret_cache[i])
- continue;
-
- STACK_WIND (frame,
- dht_rename_dir_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->rename,
- &local->loc2, &local->loc,
- NULL);
- }
-
- return 0;
- }
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (local->ret_cache[i] == 0)
+ ++local->call_cnt;
}
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (local->ret_cache[i])
+ continue;
- WIPE (&local->preoldparent);
- WIPE (&local->postoldparent);
- WIPE (&local->preparent);
- WIPE (&local->postparent);
+ STACK_WIND(frame, dht_rename_dir_cbk, conf->subvolumes[i],
+ conf->subvolumes[i]->fops->rename, &local->loc2,
+ &local->loc, NULL);
+ }
- dht_rename_dir_unlock (frame, this);
+ return 0;
+ }
}
- return 0;
-}
+ WIPE(&local->preoldparent);
+ WIPE(&local->postoldparent);
+ WIPE(&local->preparent);
+ WIPE(&local->postparent);
+
+ dht_rename_dir_unlock(frame, this);
+ }
+ return 0;
+}
int
-dht_rename_hashed_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
- struct iatt *preoldparent,
- struct iatt *postoldparent,
- struct iatt *prenewparent,
- struct iatt *postnewparent, dict_t *xdata)
+dht_rename_hashed_dir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
+ struct iatt *preoldparent, struct iatt *postoldparent,
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
{
- dht_conf_t *conf = NULL;
- dht_local_t *local = NULL;
- int call_cnt = 0;
- xlator_t *prev = NULL;
- int i = 0;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- conf = this->private;
- local = frame->local;
- prev = cookie;
-
-
- if (op_ret == -1) {
- gf_uuid_unparse(local->loc.inode->gfid, gfid);
-
- gf_msg (this->name, GF_LOG_INFO, op_errno,
- DHT_MSG_RENAME_FAILED,
- "rename %s -> %s on %s failed, (gfid = %s) ",
- local->loc.path, local->loc2.path,
- prev->name, gfid);
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto unwind;
- }
- /* TODO: construct proper stbuf for dir */
- /*
- * FIXME: is this the correct way to build stbuf and
- * parent bufs?
- */
- dht_iatt_merge (this, &local->stbuf, stbuf, prev);
- dht_iatt_merge (this, &local->preoldparent, preoldparent, prev);
- dht_iatt_merge (this, &local->postoldparent, postoldparent, prev);
- dht_iatt_merge (this, &local->preparent, prenewparent, prev);
- dht_iatt_merge (this, &local->postparent, postnewparent, prev);
-
- call_cnt = local->call_cnt = conf->subvolume_cnt - 1;
-
- if (!local->call_cnt)
- goto unwind;
-
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->subvolumes[i] == local->dst_hashed)
- continue;
- STACK_WIND_COOKIE (frame, dht_rename_dir_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->rename,
- &local->loc, &local->loc2, NULL);
- if (!--call_cnt)
- break;
- }
-
+ dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ int call_cnt = 0;
+ xlator_t *prev = NULL;
+ int i = 0;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+
+ conf = this->private;
+ local = frame->local;
+ prev = cookie;
+
+ if (op_ret == -1) {
+ gf_uuid_unparse(local->loc.inode->gfid, gfid);
- return 0;
+ gf_msg(this->name, GF_LOG_INFO, op_errno, DHT_MSG_RENAME_FAILED,
+ "rename %s -> %s on %s failed, (gfid = %s) ", local->loc.path,
+ local->loc2.path, prev->name, gfid);
+
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto unwind;
+ }
+ /* TODO: construct proper stbuf for dir */
+ /*
+ * FIXME: is this the correct way to build stbuf and
+ * parent bufs?
+ */
+ dht_iatt_merge(this, &local->stbuf, stbuf);
+ dht_iatt_merge(this, &local->preoldparent, preoldparent);
+ dht_iatt_merge(this, &local->postoldparent, postoldparent);
+ dht_iatt_merge(this, &local->preparent, prenewparent);
+ dht_iatt_merge(this, &local->postparent, postnewparent);
+
+ call_cnt = local->call_cnt = conf->subvolume_cnt - 1;
+
+ if (!local->call_cnt)
+ goto unwind;
+
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->subvolumes[i] == local->dst_hashed)
+ continue;
+ STACK_WIND_COOKIE(
+ frame, dht_rename_dir_cbk, conf->subvolumes[i], conf->subvolumes[i],
+ conf->subvolumes[i]->fops->rename, &local->loc, &local->loc2, NULL);
+ if (!--call_cnt)
+ break;
+ }
+
+ return 0;
unwind:
- WIPE (&local->preoldparent);
- WIPE (&local->postoldparent);
- WIPE (&local->preparent);
- WIPE (&local->postparent);
+ WIPE(&local->preoldparent);
+ WIPE(&local->postoldparent);
+ WIPE(&local->preparent);
+ WIPE(&local->postparent);
- dht_rename_dir_unlock (frame, this);
- return 0;
+ dht_rename_dir_unlock(frame, this);
+ return 0;
}
-
int
-dht_rename_dir_do (call_frame_t *frame, xlator_t *this)
+dht_rename_dir_do(call_frame_t *frame, xlator_t *this)
{
- dht_local_t *local = NULL;
+ dht_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->op_ret == -1)
- goto err;
+ if (local->op_ret == -1)
+ goto err;
- local->op_ret = 0;
+ local->op_ret = 0;
- STACK_WIND_COOKIE (frame, dht_rename_hashed_dir_cbk, local->dst_hashed,
- local->dst_hashed,
- local->dst_hashed->fops->rename,
- &local->loc, &local->loc2, NULL);
- return 0;
+ STACK_WIND_COOKIE(frame, dht_rename_hashed_dir_cbk, local->dst_hashed,
+ local->dst_hashed, local->dst_hashed->fops->rename,
+ &local->loc, &local->loc2, NULL);
+ return 0;
err:
- dht_rename_dir_unlock (frame, this);
- return 0;
+ dht_rename_dir_unlock(frame, this);
+ return 0;
}
-
int
-dht_rename_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, gf_dirent_t *entries,
- dict_t *xdata)
+dht_rename_readdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, gf_dirent_t *entries,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- int this_call_cnt = -1;
- xlator_t *prev = NULL;
-
- local = frame->local;
- prev = cookie;
-
- if (op_ret > 2) {
- gf_msg_trace (this->name, 0,
- "readdir on %s for %s returned %d entries",
- prev->name, local->loc.path, op_ret);
- local->op_ret = -1;
- local->op_errno = ENOTEMPTY;
- }
+ dht_local_t *local = NULL;
+ int this_call_cnt = -1;
+ xlator_t *prev = NULL;
- this_call_cnt = dht_frame_return (frame);
+ local = frame->local;
+ prev = cookie;
- if (is_last_call (this_call_cnt)) {
- dht_rename_dir_do (frame, this);
- }
+ if (op_ret > 2) {
+ gf_msg_trace(this->name, 0, "readdir on %s for %s returned %d entries",
+ prev->name, local->loc.path, op_ret);
+ local->op_ret = -1;
+ local->op_errno = ENOTEMPTY;
+ }
- return 0;
-}
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt)) {
+ dht_rename_dir_do(frame, this);
+ }
+
+ return 0;
+}
int
-dht_rename_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
+dht_rename_opendir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int this_call_cnt = -1;
- xlator_t *prev = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
+ dht_local_t *local = NULL;
+ int this_call_cnt = -1;
+ xlator_t *prev = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- local = frame->local;
- prev = cookie;
+ local = frame->local;
+ prev = cookie;
+ if (op_ret == -1) {
+ gf_uuid_unparse(local->loc.inode->gfid, gfid);
+ gf_msg(this->name, GF_LOG_INFO, op_errno, DHT_MSG_OPENDIR_FAILED,
+ "opendir on %s for %s failed,(gfid = %s) ", prev->name,
+ local->loc.path, gfid);
+ goto err;
+ }
- if (op_ret == -1) {
-
- gf_uuid_unparse(local->loc.inode->gfid, gfid);
- gf_msg (this->name, GF_LOG_INFO, op_errno,
- DHT_MSG_OPENDIR_FAILED,
- "opendir on %s for %s failed,(gfid = %s) ",
- prev->name, local->loc.path, gfid);
- goto err;
- }
-
- fd_bind (fd);
- STACK_WIND_COOKIE (frame, dht_rename_readdir_cbk, prev, prev,
- prev->fops->readdir, local->fd, 4096, 0, NULL);
+ fd_bind(fd);
+ STACK_WIND_COOKIE(frame, dht_rename_readdir_cbk, prev, prev,
+ prev->fops->readdir, local->fd, 4096, 0, NULL);
- return 0;
+ return 0;
err:
- this_call_cnt = dht_frame_return (frame);
+ this_call_cnt = dht_frame_return(frame);
- if (is_last_call (this_call_cnt)) {
- dht_rename_dir_do (frame, this);
- }
+ if (is_last_call(this_call_cnt)) {
+ dht_rename_dir_do(frame, this);
+ }
- return 0;
+ return 0;
}
int
-dht_rename_dir_lock2_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+dht_rename_dir_lock2_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- char src_gfid[GF_UUID_BUF_SIZE] = {0};
- char dst_gfid[GF_UUID_BUF_SIZE] = {0};
- dht_conf_t *conf = NULL;
- int i = 0;
-
- local = frame->local;
- conf = this->private;
-
- if (op_ret < 0) {
- uuid_utoa_r (local->loc.inode->gfid, src_gfid);
-
- if (local->loc2.inode)
- uuid_utoa_r (local->loc2.inode->gfid, dst_gfid);
-
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_INODE_LK_ERROR,
- "acquiring entrylk after inodelk failed"
- "rename (%s:%s:%s %s:%s:%s)",
- local->loc.path, src_gfid, local->src_cached->name,
- local->loc2.path, dst_gfid,
- local->dst_cached ? local->dst_cached->name : NULL);
-
- local->op_ret = -1;
- local->op_errno = op_errno;
- goto err;
- }
+ dht_local_t *local = NULL;
+ char src_gfid[GF_UUID_BUF_SIZE] = {0};
+ char dst_gfid[GF_UUID_BUF_SIZE] = {0};
+ dht_conf_t *conf = NULL;
+ int i = 0;
+
+ local = frame->local;
+ conf = this->private;
+
+ if (op_ret < 0) {
+ uuid_utoa_r(local->loc.inode->gfid, src_gfid);
+
+ if (local->loc2.inode)
+ uuid_utoa_r(local->loc2.inode->gfid, dst_gfid);
+
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, DHT_MSG_INODE_LK_ERROR,
+ "acquiring entrylk after inodelk failed"
+ "rename (%s:%s:%s %s:%s:%s)",
+ local->loc.path, src_gfid, local->src_cached->name,
+ local->loc2.path, dst_gfid,
+ local->dst_cached ? local->dst_cached->name : NULL);
+
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ goto err;
+ }
+
+ local->fd = fd_create(local->loc.inode, frame->root->pid);
+ if (!local->fd) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->op_ret = 0;
+
+ if (!local->dst_cached) {
+ dht_rename_dir_do(frame, this);
+ return 0;
+ }
- local->fd = fd_create (local->loc.inode, frame->root->pid);
- if (!local->fd) {
- op_errno = ENOMEM;
- goto err;
- }
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ STACK_WIND_COOKIE(frame, dht_rename_opendir_cbk, conf->subvolumes[i],
+ conf->subvolumes[i],
+ conf->subvolumes[i]->fops->opendir, &local->loc2,
+ local->fd, NULL);
+ }
- local->op_ret = 0;
-
- if (!local->dst_cached) {
- dht_rename_dir_do (frame, this);
- return 0;
- }
-
- for (i = 0; i < conf->subvolume_cnt; i++) {
- STACK_WIND_COOKIE (frame, dht_rename_opendir_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->opendir,
- &local->loc2, local->fd, NULL);
- }
-
- return 0;
+ return 0;
err:
- /* No harm in calling an extra unlock */
- dht_rename_dir_unlock (frame, this);
- return 0;
+ /* No harm in calling an extra unlock */
+ dht_rename_dir_unlock(frame, this);
+ return 0;
}
int
-dht_rename_dir_lock1_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+dht_rename_dir_lock1_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- char src_gfid[GF_UUID_BUF_SIZE] = {0};
- char dst_gfid[GF_UUID_BUF_SIZE] = {0};
- int ret = 0;
- loc_t *loc = NULL;
- xlator_t *subvol = NULL;
-
- local = frame->local;
-
- if (op_ret < 0) {
- uuid_utoa_r (local->loc.inode->gfid, src_gfid);
-
- if (local->loc2.inode)
- uuid_utoa_r (local->loc2.inode->gfid, dst_gfid);
-
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_INODE_LK_ERROR,
- "acquiring entrylk after inodelk failed"
- "rename (%s:%s:%s %s:%s:%s)",
- local->loc.path, src_gfid, local->src_cached->name,
- local->loc2.path, dst_gfid,
- local->dst_cached ? local->dst_cached->name : NULL);
-
- local->op_ret = -1;
- local->op_errno = op_errno;
- goto err;
- }
-
- if (local->current == &local->lock[0]) {
- loc = &local->loc2;
- subvol = local->dst_hashed;
- local->current = &local->lock[1];
- } else {
- loc = &local->loc;
- subvol = local->src_hashed;
- local->current = &local->lock[0];
- }
- ret = dht_protect_namespace (frame, loc, subvol, &local->current->ns,
- dht_rename_dir_lock2_cbk);
- if (ret < 0) {
- op_errno = EINVAL;
- goto err;
- }
-
- return 0;
+ dht_local_t *local = NULL;
+ char src_gfid[GF_UUID_BUF_SIZE] = {0};
+ char dst_gfid[GF_UUID_BUF_SIZE] = {0};
+ int ret = 0;
+ loc_t *loc = NULL;
+ xlator_t *subvol = NULL;
+
+ local = frame->local;
+
+ if (op_ret < 0) {
+ uuid_utoa_r(local->loc.inode->gfid, src_gfid);
+
+ if (local->loc2.inode)
+ uuid_utoa_r(local->loc2.inode->gfid, dst_gfid);
+
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, DHT_MSG_INODE_LK_ERROR,
+ "acquiring entrylk after inodelk failed"
+ "rename (%s:%s:%s %s:%s:%s)",
+ local->loc.path, src_gfid, local->src_cached->name,
+ local->loc2.path, dst_gfid,
+ local->dst_cached ? local->dst_cached->name : NULL);
+
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ goto err;
+ }
+
+ if (local->current == &local->lock[0]) {
+ loc = &local->loc2;
+ subvol = local->dst_hashed;
+ local->current = &local->lock[1];
+ } else {
+ loc = &local->loc;
+ subvol = local->src_hashed;
+ local->current = &local->lock[0];
+ }
+ ret = dht_protect_namespace(frame, loc, subvol, &local->current->ns,
+ dht_rename_dir_lock2_cbk);
+ if (ret < 0) {
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ return 0;
err:
- /* No harm in calling an extra unlock */
- dht_rename_dir_unlock (frame, this);
- return 0;
+ /* No harm in calling an extra unlock */
+ dht_rename_dir_unlock(frame, this);
+ return 0;
}
-
/*
* If the hashed subvolumes of both source and dst are the different,
* lock in dictionary order of hashed subvol->name. This is important
@@ -502,1298 +470,1528 @@ err:
* dirs are the same.
*
*/
-static void
-dht_order_rename_lock (call_frame_t *frame, loc_t **loc, xlator_t **subvol)
+static int
+dht_order_rename_lock(call_frame_t *frame, loc_t **loc, xlator_t **subvol)
{
- int ret = 0;
- dht_local_t *local = NULL;
- char src[GF_UUID_BNAME_BUF_SIZE] = {0};
- char dst[GF_UUID_BNAME_BUF_SIZE] = {0};
-
+ int ret = 0;
+ int op_ret = 0;
+ dht_local_t *local = NULL;
+ char *src = NULL;
+ char *dst = NULL;
+
+ local = frame->local;
+
+ if (local->src_hashed->name == local->dst_hashed->name) {
+ ret = 0;
+ } else {
+ ret = strcmp(local->src_hashed->name, local->dst_hashed->name);
+ }
+
+ if (ret == 0) {
+ /* hashed subvols are the same for src and dst */
+ /* Entrylks need to be ordered*/
+
+ src = alloca(GF_UUID_BNAME_BUF_SIZE + strlen(local->loc.name) + 1);
+ if (!src) {
+ gf_msg(frame->this->name, GF_LOG_ERROR, ENOMEM, 0,
+ "Insufficient memory for src");
+ op_ret = -1;
+ goto out;
+ }
+
+ if (!gf_uuid_is_null(local->loc.pargfid))
+ uuid_utoa_r(local->loc.pargfid, src);
+ else if (local->loc.parent)
+ uuid_utoa_r(local->loc.parent->gfid, src);
+ else
+ src[0] = '\0';
- local = frame->local;
+ strcat(src, local->loc.name);
- if (local->src_hashed->name == local->dst_hashed->name) {
- ret = 0;
- } else {
- ret = strcmp (local->src_hashed->name, local->dst_hashed->name);
+ dst = alloca(GF_UUID_BNAME_BUF_SIZE + strlen(local->loc2.name) + 1);
+ if (!dst) {
+ gf_msg(frame->this->name, GF_LOG_ERROR, ENOMEM, 0,
+ "Insufficient memory for dst");
+ op_ret = -1;
+ goto out;
}
- if (ret == 0) {
+ if (!gf_uuid_is_null(local->loc2.pargfid))
+ uuid_utoa_r(local->loc2.pargfid, dst);
+ else if (local->loc2.parent)
+ uuid_utoa_r(local->loc2.parent->gfid, dst);
+ else
+ dst[0] = '\0';
- /* hashed subvols are the same for src and dst */
- /* Entrylks need to be ordered*/
- if (local->loc.pargfid)
- uuid_utoa_r (local->loc.pargfid, src);
- else if (local->loc.parent)
- uuid_utoa_r (local->loc.parent->gfid, src);
+ strcat(dst, local->loc2.name);
+ ret = strcmp(src, dst);
+ }
- strcat (src, local->loc.name);
+ if (ret <= 0) {
+ /*inodelk in dictionary order of hashed subvol names*/
+ /*entrylk in dictionary order of gfid/basename */
+ local->current = &local->lock[0];
+ *loc = &local->loc;
+ *subvol = local->src_hashed;
- if (local->loc2.pargfid)
- uuid_utoa_r (local->loc2.pargfid, dst);
- else if (local->loc2.parent)
- uuid_utoa_r (local->loc2.parent->gfid, dst);
+ } else {
+ local->current = &local->lock[1];
+ *loc = &local->loc2;
+ *subvol = local->dst_hashed;
+ }
- strcat (dst, local->loc2.name);
- ret = strcmp (src, dst);
- }
+ op_ret = 0;
- if (ret <= 0) {
- /*inodelk in dictionary order of hashed subvol names*/
- /*entrylk in dictionary order of gfid/basename */
- local->current = &local->lock[0];
- *loc = &local->loc;
- *subvol = local->src_hashed;
-
- } else {
- local->current = &local->lock[1];
- *loc = &local->loc2;
- *subvol = local->dst_hashed;
- }
-
- return;
+out:
+ return op_ret;
}
int
-dht_rename_dir (call_frame_t *frame, xlator_t *this)
+dht_rename_dir(call_frame_t *frame, xlator_t *this)
{
- dht_conf_t *conf = NULL;
- dht_local_t *local = NULL;
- loc_t *loc = NULL;
- xlator_t *subvol = NULL;
- int i = 0;
- int ret = 0;
- int op_errno = -1;
-
- conf = frame->this->private;
- local = frame->local;
-
- local->ret_cache = GF_CALLOC (conf->subvolume_cnt + 1, sizeof (int),
- gf_dht_ret_cache_t);
-
- if (local->ret_cache == NULL) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local->call_cnt = conf->subvolume_cnt;
-
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (!conf->subvolume_status[i]) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_RENAME_FAILED,
- "Rename dir failed: subvolume down (%s)",
- conf->subvolumes[i]->name);
- op_errno = ENOTCONN;
- goto err;
- }
- }
-
-
- /* Locks on src and dst needs to ordered which otherwise might cause
- * deadlocks when rename (src, dst) and rename (dst, src) is done from
- * two different clients
- */
- dht_order_rename_lock (frame, &loc, &subvol);
-
- /* Rename must take locks on src to avoid lookup selfheal from
- * recreating src on those subvols where the rename was successful.
- * The locks can't be issued parallel as two different clients might
- * attempt same rename command and be in dead lock.
- */
- ret = dht_protect_namespace (frame, loc, subvol,
- &local->current->ns,
- dht_rename_dir_lock1_cbk);
- if (ret < 0) {
- op_errno = EINVAL;
- goto err;
- }
-
- return 0;
+ dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ loc_t *loc = NULL;
+ xlator_t *subvol = NULL;
+ int i = 0;
+ int ret = 0;
+ int op_errno = -1;
+
+ conf = frame->this->private;
+ local = frame->local;
+
+ local->ret_cache = GF_CALLOC(conf->subvolume_cnt + 1, sizeof(int),
+ gf_dht_ret_cache_t);
+
+ if (local->ret_cache == NULL) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->call_cnt = conf->subvolume_cnt;
+
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (!conf->subvolume_status[i]) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_RENAME_FAILED,
+ "Rename dir failed: subvolume down (%s)",
+ conf->subvolumes[i]->name);
+ op_errno = ENOTCONN;
+ goto err;
+ }
+ }
+
+ /* Locks on src and dst needs to ordered which otherwise might cause
+ * deadlocks when rename (src, dst) and rename (dst, src) is done from
+ * two different clients
+ */
+ ret = dht_order_rename_lock(frame, &loc, &subvol);
+ if (ret) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ /* Rename must take locks on src to avoid lookup selfheal from
+ * recreating src on those subvols where the rename was successful.
+ * The locks can't be issued parallel as two different clients might
+ * attempt same rename command and be in dead lock.
+ */
+ ret = dht_protect_namespace(frame, loc, subvol, &local->current->ns,
+ dht_rename_dir_lock1_cbk);
+ if (ret < 0) {
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
- NULL, NULL);
- return 0;
+ DHT_STACK_UNWIND(rename, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
static int
-dht_rename_track_for_changelog (xlator_t *this, dict_t *xattr,
- loc_t *oldloc, loc_t *newloc)
+dht_rename_track_for_changelog(xlator_t *this, dict_t *xattr, loc_t *oldloc,
+ loc_t *newloc)
{
- int ret = -1;
- dht_changelog_rename_info_t *info = NULL;
- char *name = NULL;
- int len1 = 0;
- int len2 = 0;
- int size = 0;
-
- if (!xattr || !oldloc || !newloc || !this)
- return ret;
-
- len1 = strlen (oldloc->name) + 1;
- len2 = strlen (newloc->name) + 1;
- size = sizeof (dht_changelog_rename_info_t) + len1 + len2;
-
- info = GF_CALLOC (size, sizeof(char), gf_common_mt_char);
- if (!info) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to calloc memory");
- return ret;
- }
+ int ret = -1;
+ dht_changelog_rename_info_t *info = NULL;
+ char *name = NULL;
+ int len1 = 0;
+ int len2 = 0;
+ int size = 0;
+
+ if (!xattr || !oldloc || !newloc || !this)
+ return ret;
- gf_uuid_copy (info->old_pargfid, oldloc->pargfid);
- gf_uuid_copy (info->new_pargfid, newloc->pargfid);
-
- info->oldname_len = len1;
- info->newname_len = len2;
- strncpy (info->buffer, oldloc->name, len1);
- name = info->buffer + len1;
- strncpy (name, newloc->name, len2);
-
- ret = dict_set_bin (xattr, DHT_CHANGELOG_RENAME_OP_KEY,
- info, size);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value: key = %s,"
- " path = %s", DHT_CHANGELOG_RENAME_OP_KEY,
- oldloc->name);
- GF_FREE (info);
- }
+ len1 = strlen(oldloc->name) + 1;
+ len2 = strlen(newloc->name) + 1;
+ size = sizeof(dht_changelog_rename_info_t) + len1 + len2;
+ info = GF_CALLOC(size, sizeof(char), gf_common_mt_char);
+ if (!info) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "Failed to calloc memory");
return ret;
+ }
+
+ gf_uuid_copy(info->old_pargfid, oldloc->pargfid);
+ gf_uuid_copy(info->new_pargfid, newloc->pargfid);
+
+ info->oldname_len = len1;
+ info->newname_len = len2;
+ strncpy(info->buffer, oldloc->name, len1);
+ name = info->buffer + len1;
+ strncpy(name, newloc->name, len2);
+
+ ret = dict_set_bin(xattr, DHT_CHANGELOG_RENAME_OP_KEY, info, size);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dictionary value: key = %s,"
+ " path = %s",
+ DHT_CHANGELOG_RENAME_OP_KEY, oldloc->name);
+ GF_FREE(info);
+ }
+
+ return ret;
}
-
-
-#define DHT_MARKER_DONT_ACCOUNT(xattr) do { \
- int tmp = -1; \
- if (!xattr) { \
- xattr = dict_new (); \
- if (!xattr) \
- break; \
- } \
- tmp = dict_set_str (xattr, GLUSTERFS_MARKER_DONT_ACCOUNT_KEY, \
- "yes"); \
- if (tmp) { \
- gf_msg (this->name, GF_LOG_ERROR, 0, \
- DHT_MSG_DICT_SET_FAILED, \
- "Failed to set dictionary value: key = %s," \
- " path = %s",GLUSTERFS_MARKER_DONT_ACCOUNT_KEY, \
- local->loc.path); \
- } \
- }while (0)
-
-
-#define DHT_CHANGELOG_TRACK_AS_RENAME(xattr, oldloc, newloc) do { \
- int tmp = -1; \
- if (!xattr) { \
- xattr = dict_new (); \
- if (!xattr) { \
- gf_msg (this->name, GF_LOG_ERROR, 0, \
- DHT_MSG_DICT_SET_FAILED, \
- "Failed to create dictionary to " \
- "track rename"); \
- break; \
- } \
- } \
- \
- tmp = dht_rename_track_for_changelog (this, xattr, \
- oldloc, newloc); \
- \
- if (tmp) { \
- gf_msg (this->name, GF_LOG_ERROR, 0, \
- DHT_MSG_DICT_SET_FAILED, \
- "Failed to set dictionary value: key = %s," \
- " path = %s", DHT_CHANGELOG_RENAME_OP_KEY, \
- (oldloc)->path); \
- } \
- } while (0)
-
+#define DHT_MARKER_DONT_ACCOUNT(xattr) \
+ do { \
+ int tmp = -1; \
+ if (!xattr) { \
+ xattr = dict_new(); \
+ if (!xattr) \
+ break; \
+ } \
+ tmp = dict_set_str(xattr, GLUSTERFS_MARKER_DONT_ACCOUNT_KEY, "yes"); \
+ if (tmp) { \
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED, \
+ "Failed to set dictionary value: key = %s," \
+ " path = %s", \
+ GLUSTERFS_MARKER_DONT_ACCOUNT_KEY, local->loc.path); \
+ } \
+ } while (0)
+
+#define DHT_CHANGELOG_TRACK_AS_RENAME(xattr, oldloc, newloc) \
+ do { \
+ int tmp = -1; \
+ if (!xattr) { \
+ xattr = dict_new(); \
+ if (!xattr) { \
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED, \
+ "Failed to create dictionary to " \
+ "track rename"); \
+ break; \
+ } \
+ } \
+ \
+ tmp = dht_rename_track_for_changelog(this, xattr, oldloc, newloc); \
+ \
+ if (tmp) { \
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED, \
+ "Failed to set dictionary value: key = %s," \
+ " path = %s", \
+ DHT_CHANGELOG_RENAME_OP_KEY, (oldloc)->path); \
+ } \
+ } while (0)
int
-dht_rename_unlock (call_frame_t *frame, xlator_t *this)
+dht_rename_unlock(call_frame_t *frame, xlator_t *this)
{
- dht_local_t *local = NULL;
- int op_ret = -1;
- char src_gfid[GF_UUID_BUF_SIZE] = {0};
- char dst_gfid[GF_UUID_BUF_SIZE] = {0};
-
- local = frame->local;
- op_ret = dht_unlock_inodelk (frame,
- local->lock[0].layout.parent_layout.locks,
- local->lock[0].layout.parent_layout.lk_count,
- dht_rename_unlock_cbk);
- if (op_ret < 0) {
- uuid_utoa_r (local->loc.inode->gfid, src_gfid);
-
- if (local->loc2.inode)
- uuid_utoa_r (local->loc2.inode->gfid, dst_gfid);
-
- if (IA_ISREG (local->stbuf.ia_type))
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_UNLOCKING_FAILED,
- "winding unlock inodelk failed "
- "rename (%s:%s:%s %s:%s:%s), "
- "stale locks left on bricks",
- local->loc.path, src_gfid,
- local->src_cached->name,
- local->loc2.path, dst_gfid,
- local->dst_cached ?
- local->dst_cached->name : NULL);
- else
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_UNLOCKING_FAILED,
- "winding unlock inodelk failed "
- "rename (%s:%s %s:%s), "
- "stale locks left on bricks",
- local->loc.path, src_gfid,
- local->loc2.path, dst_gfid);
-
- dht_rename_unlock_cbk (frame, NULL, this, 0, 0, NULL);
- }
-
- return 0;
+ dht_local_t *local = NULL;
+ int op_ret = -1;
+ char src_gfid[GF_UUID_BUF_SIZE] = {0};
+ char dst_gfid[GF_UUID_BUF_SIZE] = {0};
+ dht_ilock_wrap_t inodelk_wrapper = {
+ 0,
+ };
+
+ local = frame->local;
+ inodelk_wrapper.locks = local->rename_inodelk_backward_compatible;
+ inodelk_wrapper.lk_count = local->rename_inodelk_bc_count;
+
+ op_ret = dht_unlock_inodelk_wrapper(frame, &inodelk_wrapper);
+ if (op_ret < 0) {
+ uuid_utoa_r(local->loc.inode->gfid, src_gfid);
+
+ if (local->loc2.inode)
+ uuid_utoa_r(local->loc2.inode->gfid, dst_gfid);
+
+ if (IA_ISREG(local->stbuf.ia_type))
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_UNLOCKING_FAILED,
+ "winding unlock inodelk failed "
+ "rename (%s:%s:%s %s:%s:%s), "
+ "stale locks left on bricks",
+ local->loc.path, src_gfid, local->src_cached->name,
+ local->loc2.path, dst_gfid,
+ local->dst_cached ? local->dst_cached->name : NULL);
+ else
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_UNLOCKING_FAILED,
+ "winding unlock inodelk failed "
+ "rename (%s:%s %s:%s), "
+ "stale locks left on bricks",
+ local->loc.path, src_gfid, local->loc2.path, dst_gfid);
+ }
+
+ dht_unlock_namespace(frame, &local->lock[0]);
+ dht_unlock_namespace(frame, &local->lock[1]);
+
+ dht_rename_unlock_cbk(frame, NULL, this, local->op_ret, local->op_errno,
+ NULL);
+ return 0;
}
int
-dht_rename_done (call_frame_t *frame, xlator_t *this)
+dht_rename_done(call_frame_t *frame, xlator_t *this)
{
- dht_local_t *local = NULL;
+ dht_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->linked == _gf_true) {
- local->linked = _gf_false;
- dht_linkfile_attr_heal (frame, this);
- }
+ if (local->linked == _gf_true) {
+ local->linked = _gf_false;
+ dht_linkfile_attr_heal(frame, this);
+ }
- dht_rename_unlock (frame, this);
- return 0;
+ dht_rename_unlock(frame, this);
+ return 0;
}
int
-dht_rename_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+dht_rename_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- int this_call_cnt = 0;
-
- local = frame->local;
- prev = cookie;
-
- FRAME_SU_UNDO (frame, dht_local_t);
- if (!local) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_INVALID_VALUE,
- "!local, should not happen");
- goto out;
- }
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ int this_call_cnt = 0;
- this_call_cnt = dht_frame_return (frame);
+ local = frame->local;
+ prev = cookie;
- if (op_ret == -1) {
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_UNLINK_FAILED,
- "%s: Rename: unlink on %s failed ",
- local->loc.path, prev->name);
- }
+ FRAME_SU_UNDO(frame, dht_local_t);
+ if (!local) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_INVALID_VALUE,
+ "!local, should not happen");
+ goto out;
+ }
- WIPE (&local->preoldparent);
- WIPE (&local->postoldparent);
- WIPE (&local->preparent);
- WIPE (&local->postparent);
+ this_call_cnt = dht_frame_return(frame);
- if (is_last_call (this_call_cnt)) {
- dht_rename_done (frame, this);
- }
+ if (op_ret == -1) {
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, DHT_MSG_UNLINK_FAILED,
+ "%s: Rename: unlink on %s failed ", local->loc.path, prev->name);
+ }
+
+ WIPE(&local->preoldparent);
+ WIPE(&local->postoldparent);
+ WIPE(&local->preparent);
+ WIPE(&local->postparent);
+
+ if (is_last_call(this_call_cnt)) {
+ dht_rename_done(frame, this);
+ }
out:
- return 0;
+ return 0;
}
-
int
-dht_rename_cleanup (call_frame_t *frame)
+dht_rename_cleanup(call_frame_t *frame)
{
- dht_local_t *local = NULL;
- xlator_t *this = NULL;
- xlator_t *src_hashed = NULL;
- xlator_t *src_cached = NULL;
- xlator_t *dst_hashed = NULL;
- xlator_t *dst_cached = NULL;
- int call_cnt = 0;
- dict_t *xattr = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- local = frame->local;
- this = frame->this;
-
- src_hashed = local->src_hashed;
- src_cached = local->src_cached;
- dst_hashed = local->dst_hashed;
- dst_cached = local->dst_cached;
-
- if (src_cached == dst_cached)
- goto nolinks;
-
- if (local->linked && (dst_hashed != src_hashed )&&
- (dst_hashed != src_cached)) {
- call_cnt++;
- }
+ dht_local_t *local = NULL;
+ xlator_t *this = NULL;
+ xlator_t *src_hashed = NULL;
+ xlator_t *src_cached = NULL;
+ xlator_t *dst_hashed = NULL;
+ xlator_t *dst_cached = NULL;
+ int call_cnt = 0;
+ dict_t *xattr = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- if (local->added_link && (src_cached != dst_hashed)) {
- call_cnt++;
- }
+ local = frame->local;
+ this = frame->this;
- local->call_cnt = call_cnt;
+ src_hashed = local->src_hashed;
+ src_cached = local->src_cached;
+ dst_hashed = local->dst_hashed;
+ dst_cached = local->dst_cached;
- if (!call_cnt)
- goto nolinks;
+ if (src_cached == dst_cached)
+ goto nolinks;
- DHT_MARK_FOP_INTERNAL (xattr);
+ if (local->linked && (dst_hashed != src_hashed) &&
+ (dst_hashed != src_cached)) {
+ call_cnt++;
+ }
- gf_uuid_unparse(local->loc.inode->gfid, gfid);
+ if (local->added_link && (src_cached != dst_hashed)) {
+ call_cnt++;
+ }
- if (local->linked && (dst_hashed != src_hashed) &&
- (dst_hashed != src_cached)) {
- dict_t *xattr_new = NULL;
+ local->call_cnt = call_cnt;
- gf_msg_trace (this->name, 0,
- "unlinking linkfile %s @ %s => %s, (gfid = %s)",
- local->loc.path, dst_hashed->name,
- src_cached->name, gfid);
+ if (!call_cnt)
+ goto nolinks;
- xattr_new = dict_copy_with_ref (xattr, NULL);
+ DHT_MARK_FOP_INTERNAL(xattr);
+ gf_uuid_unparse(local->loc.inode->gfid, gfid);
- DHT_MARKER_DONT_ACCOUNT(xattr_new);
+ if (local->linked && (dst_hashed != src_hashed) &&
+ (dst_hashed != src_cached)) {
+ dict_t *xattr_new = NULL;
- FRAME_SU_DO (frame, dht_local_t);
- STACK_WIND_COOKIE (frame, dht_rename_unlink_cbk, dst_hashed,
- dst_hashed, dst_hashed->fops->unlink,
- &local->loc, 0, xattr_new);
+ gf_msg_trace(this->name, 0,
+ "unlinking linkfile %s @ %s => %s, (gfid = %s)",
+ local->loc.path, dst_hashed->name, src_cached->name, gfid);
- dict_unref (xattr_new);
- xattr_new = NULL;
- }
+ xattr_new = dict_copy_with_ref(xattr, NULL);
- if (local->added_link && (src_cached != dst_hashed)) {
- dict_t *xattr_new = NULL;
+ DHT_MARKER_DONT_ACCOUNT(xattr_new);
- gf_msg_trace (this->name, 0,
- "unlinking link %s => %s (%s), (gfid = %s)",
- local->loc.path, local->loc2.path,
- src_cached->name, gfid);
+ FRAME_SU_DO(frame, dht_local_t);
+ STACK_WIND_COOKIE(frame, dht_rename_unlink_cbk, dst_hashed, dst_hashed,
+ dst_hashed->fops->unlink, &local->loc, 0, xattr_new);
- xattr_new = dict_copy_with_ref (xattr, NULL);
+ dict_unref(xattr_new);
+ xattr_new = NULL;
+ }
- if (gf_uuid_compare (local->loc.pargfid,
- local->loc2.pargfid) == 0) {
- DHT_MARKER_DONT_ACCOUNT(xattr_new);
- }
- /* *
- * The link to file is created using root permission.
- * Hence deletion should happen using root. Otherwise
- * it will fail.
- */
- FRAME_SU_DO (frame, dht_local_t);
- STACK_WIND_COOKIE (frame, dht_rename_unlink_cbk, src_cached,
- src_cached, src_cached->fops->unlink,
- &local->loc2, 0, xattr_new);
+ if (local->added_link && (src_cached != dst_hashed)) {
+ dict_t *xattr_new = NULL;
- dict_unref (xattr_new);
- xattr_new = NULL;
- }
-
- if (xattr)
- dict_unref (xattr);
+ gf_msg_trace(this->name, 0, "unlinking link %s => %s (%s), (gfid = %s)",
+ local->loc.path, local->loc2.path, src_cached->name, gfid);
- return 0;
-
-nolinks:
- WIPE (&local->preoldparent);
- WIPE (&local->postoldparent);
- WIPE (&local->preparent);
- WIPE (&local->postparent);
-
- dht_rename_unlock (frame, this);
- return 0;
-}
+ xattr_new = dict_copy_with_ref(xattr, NULL);
+ if (gf_uuid_compare(local->loc.pargfid, local->loc2.pargfid) == 0) {
+ DHT_MARKER_DONT_ACCOUNT(xattr_new);
+ }
+ /* *
+ * The link to file is created using root permission.
+ * Hence deletion should happen using root. Otherwise
+ * it will fail.
+ */
+ FRAME_SU_DO(frame, dht_local_t);
+ STACK_WIND_COOKIE(frame, dht_rename_unlink_cbk, src_cached, src_cached,
+ src_cached->fops->unlink, &local->loc2, 0, xattr_new);
-int
-dht_rename_links_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
-{
- xlator_t *prev = NULL;
- dht_local_t *local = NULL;
+ dict_unref(xattr_new);
+ xattr_new = NULL;
+ }
- prev = cookie;
- local = frame->local;
+ if (xattr)
+ dict_unref(xattr);
- if (op_ret == -1) {
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_CREATE_LINK_FAILED,
- "link/file %s on %s failed",
- local->loc.path, prev->name);
- }
+ return 0;
- if (local->linked == _gf_true) {
- local->linked = _gf_false;
- dht_linkfile_attr_heal (frame, this);
- }
- DHT_STACK_DESTROY (frame);
+nolinks:
+ WIPE(&local->preoldparent);
+ WIPE(&local->postoldparent);
+ WIPE(&local->preparent);
+ WIPE(&local->postparent);
- return 0;
+ dht_rename_unlock(frame, this);
+ return 0;
}
-
int
-dht_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent,
- dict_t *xdata)
+dht_rename_unlink(call_frame_t *frame, xlator_t *this)
{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- xlator_t *src_hashed = NULL;
- xlator_t *src_cached = NULL;
- xlator_t *dst_hashed = NULL;
- xlator_t *dst_cached = NULL;
- xlator_t *rename_subvol = NULL;
- call_frame_t *link_frame = NULL;
- dht_local_t *link_local = NULL;
- dict_t *xattr = NULL;
-
- local = frame->local;
- prev = cookie;
-
- src_hashed = local->src_hashed;
- src_cached = local->src_cached;
- dst_hashed = local->dst_hashed;
- dst_cached = local->dst_cached;
-
- if (local->linked == _gf_true)
- FRAME_SU_UNDO (frame, dht_local_t);
-
- /* It is a critical failure iff we fail to rename the cached file
- * if the rename of the linkto failed, it is not a critical failure,
- * and we do not want to lose the created hard link for the new
- * name as that could have been read by other clients.
- *
- * NOTE: If another client is attempting the same oldname -> newname
- * rename, and finds both file names as existing, and are hard links
- * to each other, then FUSE would send in an unlink for oldname. In
- * this time duration if we treat the linkto as a critical error and
- * unlink the newname we created, we would have effectively lost the
- * file to rename operations.
- *
- * Repercussions of treating this as a non-critical error is that
- * we could leave behind a stale linkto file and/or not create the new
- * linkto file, the second case would be rectified by a subsequent
- * lookup, the first case by a rebalance, like for all stale linkto
- * files */
-
- if (op_ret == -1) {
- /* Critical failure: unable to rename the cached file */
- if (prev == src_cached) {
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_RENAME_FAILED,
- "%s: Rename on %s failed, (gfid = %s) ",
- local->loc.path, prev->name,
- local->loc.inode ?
- uuid_utoa(local->loc.inode->gfid):"");
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto cleanup;
- } else {
- /* Non-critical failure, unable to rename the linkto
- * file
- */
- gf_msg (this->name, GF_LOG_INFO, op_errno,
- DHT_MSG_RENAME_FAILED,
- "%s: Rename (linkto file) on %s failed, "
- "(gfid = %s) ",
- local->loc.path, prev->name,
- local->loc.inode ?
- uuid_utoa(local->loc.inode->gfid):"");
- }
- }
- if (xdata) {
- if (!local->xattr)
- local->xattr = dict_ref (xdata);
- else
- local->xattr = dict_copy_with_ref (xdata, local->xattr);
- }
+ dht_local_t *local = NULL;
+ xlator_t *src_hashed = NULL;
+ xlator_t *src_cached = NULL;
+ xlator_t *dst_hashed = NULL;
+ xlator_t *dst_cached = NULL;
+ xlator_t *rename_subvol = NULL;
+ dict_t *xattr = NULL;
- if ((src_cached == dst_cached) && (dst_hashed != dst_cached)) {
- link_frame = copy_frame (frame);
- if (!link_frame) {
- goto err;
- }
+ local = frame->local;
- /* fop value sent as maxvalue because it is not used
- anywhere in this case */
- link_local = dht_local_init (link_frame, &local->loc2, NULL,
- GF_FOP_MAXVALUE);
- if (!link_local) {
- goto err;
- }
+ src_hashed = local->src_hashed;
+ src_cached = local->src_cached;
+ dst_hashed = local->dst_hashed;
+ dst_cached = local->dst_cached;
- if (link_local->loc.inode)
- inode_unref (link_local->loc.inode);
- link_local->loc.inode = inode_ref (local->loc.inode);
- gf_uuid_copy (link_local->gfid, local->loc.inode->gfid);
+ local->call_cnt = 0;
- dht_linkfile_create (link_frame, dht_rename_links_create_cbk,
- this, src_cached, dst_hashed,
- &link_local->loc);
- }
+ /* NOTE: rename_subvol is the same subvolume from which dht_rename_cbk
+ * is called. since rename has already happened on rename_subvol,
+ * unlink shouldn't be sent for oldpath (either linkfile or cached-file)
+ * on rename_subvol. */
+ if (src_cached == dst_cached)
+ rename_subvol = src_cached;
+ else
+ rename_subvol = dst_hashed;
-err:
- /* Merge attrs only from src_cached. In case there of src_cached !=
- * dst_hashed, this ignores linkfile attrs. */
- if (prev == src_cached) {
- dht_iatt_merge (this, &local->stbuf, stbuf, prev);
- dht_iatt_merge (this, &local->preoldparent, preoldparent,
- prev);
- dht_iatt_merge (this, &local->postoldparent, postoldparent,
- prev);
- dht_iatt_merge (this, &local->preparent, prenewparent, prev);
- dht_iatt_merge (this, &local->postparent, postnewparent, prev);
- }
-
-
- /* NOTE: rename_subvol is the same subvolume from which dht_rename_cbk
- * is called. since rename has already happened on rename_subvol,
- * unlink should not be sent for oldpath (either linkfile or cached-file)
- * on rename_subvol. */
- if (src_cached == dst_cached)
- rename_subvol = src_cached;
- else
- rename_subvol = dst_hashed;
+ /* TODO: delete files in background */
- /* TODO: delete files in background */
+ if (src_cached != dst_hashed && src_cached != dst_cached)
+ local->call_cnt++;
- if (src_cached != dst_hashed && src_cached != dst_cached)
- local->call_cnt++;
+ if (src_hashed != rename_subvol && src_hashed != src_cached)
+ local->call_cnt++;
- if (src_hashed != rename_subvol && src_hashed != src_cached)
- local->call_cnt++;
+ if (dst_cached && dst_cached != dst_hashed && dst_cached != src_cached)
+ local->call_cnt++;
- if (dst_cached && dst_cached != dst_hashed && dst_cached != src_cached)
- local->call_cnt++;
+ if (local->call_cnt == 0)
+ goto unwind;
- if (local->call_cnt == 0)
- goto unwind;
+ DHT_MARK_FOP_INTERNAL(xattr);
- DHT_MARK_FOP_INTERNAL (xattr);
+ if (src_cached != dst_hashed && src_cached != dst_cached) {
+ dict_t *xattr_new = NULL;
- if (src_cached != dst_hashed && src_cached != dst_cached) {
- dict_t *xattr_new = NULL;
+ xattr_new = dict_copy_with_ref(xattr, NULL);
- xattr_new = dict_copy_with_ref (xattr, NULL);
+ gf_msg_trace(this->name, 0, "deleting old src datafile %s @ %s",
+ local->loc.path, src_cached->name);
- gf_msg_trace (this->name, 0,
- "deleting old src datafile %s @ %s",
- local->loc.path, src_cached->name);
-
- if (gf_uuid_compare (local->loc.pargfid,
- local->loc2.pargfid) == 0) {
- DHT_MARKER_DONT_ACCOUNT(xattr_new);
- }
-
- DHT_CHANGELOG_TRACK_AS_RENAME(xattr_new, &local->loc,
- &local->loc2);
- STACK_WIND_COOKIE (frame, dht_rename_unlink_cbk, src_cached,
- src_cached, src_cached->fops->unlink,
- &local->loc, 0, xattr_new);
-
- dict_unref (xattr_new);
- xattr_new = NULL;
+ if (gf_uuid_compare(local->loc.pargfid, local->loc2.pargfid) == 0) {
+ DHT_MARKER_DONT_ACCOUNT(xattr_new);
}
- if (src_hashed != rename_subvol && src_hashed != src_cached) {
- dict_t *xattr_new = NULL;
+ DHT_CHANGELOG_TRACK_AS_RENAME(xattr_new, &local->loc, &local->loc2);
+ STACK_WIND_COOKIE(frame, dht_rename_unlink_cbk, src_cached, src_cached,
+ src_cached->fops->unlink, &local->loc, 0, xattr_new);
- xattr_new = dict_copy_with_ref (xattr, NULL);
+ dict_unref(xattr_new);
+ xattr_new = NULL;
+ }
- gf_msg_trace (this->name, 0,
- "deleting old src linkfile %s @ %s",
- local->loc.path, src_hashed->name);
+ if (src_hashed != rename_subvol && src_hashed != src_cached) {
+ dict_t *xattr_new = NULL;
- DHT_MARKER_DONT_ACCOUNT(xattr_new);
+ xattr_new = dict_copy_with_ref(xattr, NULL);
- STACK_WIND_COOKIE (frame, dht_rename_unlink_cbk, src_hashed,
- src_hashed, src_hashed->fops->unlink,
- &local->loc, 0, xattr_new);
+ gf_msg_trace(this->name, 0, "deleting old src linkfile %s @ %s",
+ local->loc.path, src_hashed->name);
- dict_unref (xattr_new);
- xattr_new = NULL;
- }
+ DHT_MARKER_DONT_ACCOUNT(xattr_new);
- if (dst_cached
- && (dst_cached != dst_hashed)
- && (dst_cached != src_cached)) {
- gf_msg_trace (this->name, 0,
- "deleting old dst datafile %s @ %s",
- local->loc2.path, dst_cached->name);
+ STACK_WIND_COOKIE(frame, dht_rename_unlink_cbk, src_hashed, src_hashed,
+ src_hashed->fops->unlink, &local->loc, 0, xattr_new);
- STACK_WIND_COOKIE (frame, dht_rename_unlink_cbk, dst_cached,
- dst_cached, dst_cached->fops->unlink,
- &local->loc2, 0, xattr);
- }
- if (xattr)
- dict_unref (xattr);
- return 0;
+ dict_unref(xattr_new);
+ xattr_new = NULL;
+ }
-unwind:
- WIPE (&local->preoldparent);
- WIPE (&local->postoldparent);
- WIPE (&local->preparent);
- WIPE (&local->postparent);
+ if (dst_cached && (dst_cached != dst_hashed) &&
+ (dst_cached != src_cached)) {
+ gf_msg_trace(this->name, 0, "deleting old dst datafile %s @ %s",
+ local->loc2.path, dst_cached->name);
- dht_rename_done (frame, this);
+ STACK_WIND_COOKIE(frame, dht_rename_unlink_cbk, dst_cached, dst_cached,
+ dst_cached->fops->unlink, &local->loc2, 0, xattr);
+ }
+ if (xattr)
+ dict_unref(xattr);
+ return 0;
- return 0;
+unwind:
+ WIPE(&local->preoldparent);
+ WIPE(&local->postoldparent);
+ WIPE(&local->preparent);
+ WIPE(&local->postparent);
-cleanup:
- dht_rename_cleanup (frame);
+ dht_rename_done(frame, this);
- return 0;
+ return 0;
}
+int
+dht_rename_links_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ xlator_t *prev = NULL;
+ dht_local_t *local = NULL;
+ call_frame_t *main_frame = NULL;
+
+ prev = cookie;
+ local = frame->local;
+ main_frame = local->main_frame;
+
+ /* TODO: Handle this case in lookup-optimize */
+ if (op_ret == -1) {
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, DHT_MSG_CREATE_LINK_FAILED,
+ "link/file %s on %s failed", local->loc.path, prev->name);
+ }
+
+ if (local->linked == _gf_true) {
+ local->linked = _gf_false;
+ dht_linkfile_attr_heal(frame, this);
+ }
+
+ dht_rename_unlink(main_frame, this);
+ DHT_STACK_DESTROY(frame);
+ return 0;
+}
int
-dht_do_rename (call_frame_t *frame)
+dht_rename_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
+ struct iatt *preoldparent, struct iatt *postoldparent,
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *dst_hashed = NULL;
- xlator_t *src_cached = NULL;
- xlator_t *dst_cached = NULL;
- xlator_t *this = NULL;
- xlator_t *rename_subvol = NULL;
-
- local = frame->local;
- this = frame->this;
-
- dst_hashed = local->dst_hashed;
- dst_cached = local->dst_cached;
- src_cached = local->src_cached;
-
- if (src_cached == dst_cached)
- rename_subvol = src_cached;
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ xlator_t *src_cached = NULL;
+ xlator_t *dst_hashed = NULL;
+ xlator_t *dst_cached = NULL;
+ call_frame_t *link_frame = NULL;
+ dht_local_t *link_local = NULL;
+
+ local = frame->local;
+ prev = cookie;
+
+ src_cached = local->src_cached;
+ dst_hashed = local->dst_hashed;
+ dst_cached = local->dst_cached;
+
+ if (local->linked == _gf_true)
+ FRAME_SU_UNDO(frame, dht_local_t);
+
+ /* It is a critical failure iff we fail to rename the cached file
+ * if the rename of the linkto failed, it is not a critical failure,
+ * and we do not want to lose the created hard link for the new
+ * name as that could have been read by other clients.
+ *
+ * NOTE: If another client is attempting the same oldname -> newname
+ * rename, and finds both file names as existing, and are hard links
+ * to each other, then FUSE would send in an unlink for oldname. In
+ * this time duration if we treat the linkto as a critical error and
+ * unlink the newname we created, we would have effectively lost the
+ * file to rename operations.
+ *
+ * Repercussions of treating this as a non-critical error is that
+ * we could leave behind a stale linkto file and/or not create the new
+ * linkto file, the second case would be rectified by a subsequent
+ * lookup, the first case by a rebalance, like for all stale linkto
+ * files */
+
+ if (op_ret == -1) {
+ /* Critical failure: unable to rename the cached file */
+ if (prev == src_cached) {
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, DHT_MSG_RENAME_FAILED,
+ "%s: Rename on %s failed, (gfid = %s) ", local->loc.path,
+ prev->name,
+ local->loc.inode ? uuid_utoa(local->loc.inode->gfid) : "");
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto cleanup;
+ } else {
+ /* Non-critical failure, unable to rename the linkto
+ * file
+ */
+ gf_msg(this->name, GF_LOG_INFO, op_errno, DHT_MSG_RENAME_FAILED,
+ "%s: Rename (linkto file) on %s failed, "
+ "(gfid = %s) ",
+ local->loc.path, prev->name,
+ local->loc.inode ? uuid_utoa(local->loc.inode->gfid) : "");
+ }
+ }
+ if (xdata) {
+ if (!local->xattr)
+ local->xattr = dict_ref(xdata);
else
- rename_subvol = dst_hashed;
+ local->xattr = dict_copy_with_ref(xdata, local->xattr);
+ }
+
+ /* Merge attrs only from src_cached. In case there of src_cached !=
+ * dst_hashed, this ignores linkfile attrs. */
+ if (prev == src_cached) {
+ dht_iatt_merge(this, &local->stbuf, stbuf);
+ dht_iatt_merge(this, &local->preoldparent, preoldparent);
+ dht_iatt_merge(this, &local->postoldparent, postoldparent);
+ dht_iatt_merge(this, &local->preparent, prenewparent);
+ dht_iatt_merge(this, &local->postparent, postnewparent);
+ }
+
+ /* Create the linkto file for the dst file */
+ if ((src_cached == dst_cached) && (dst_hashed != dst_cached)) {
+ link_frame = copy_frame(frame);
+ if (!link_frame) {
+ goto unlink;
+ }
+
+ /* fop value sent as maxvalue because it is not used
+ * anywhere in this case */
+ link_local = dht_local_init(link_frame, &local->loc2, NULL,
+ GF_FOP_MAXVALUE);
+ if (!link_local) {
+ goto unlink;
+ }
+
+ if (link_local->loc.inode)
+ inode_unref(link_local->loc.inode);
+ link_local->loc.inode = inode_ref(local->loc.inode);
+ link_local->main_frame = frame;
+ link_local->stbuf = local->stbuf;
+ gf_uuid_copy(link_local->gfid, local->loc.inode->gfid);
+
+ dht_linkfile_create(link_frame, dht_rename_links_create_cbk, this,
+ src_cached, dst_hashed, &link_local->loc);
+ return 0;
+ }
- if ((src_cached != dst_hashed) && (rename_subvol == dst_hashed)) {
- DHT_MARKER_DONT_ACCOUNT(local->xattr_req);
- }
+unlink:
- if (rename_subvol == src_cached) {
- DHT_CHANGELOG_TRACK_AS_RENAME(local->xattr_req, &local->loc,
- &local->loc2);
- }
+ if (link_frame) {
+ DHT_STACK_DESTROY(link_frame);
+ }
+ dht_rename_unlink(frame, this);
+ return 0;
- gf_msg_trace (this->name, 0,
- "renaming %s => %s (%s)",
- local->loc.path, local->loc2.path, rename_subvol->name);
+cleanup:
+ dht_rename_cleanup(frame);
- if (local->linked == _gf_true)
- FRAME_SU_DO (frame, dht_local_t);
- STACK_WIND_COOKIE (frame, dht_rename_cbk, rename_subvol, rename_subvol,
- rename_subvol->fops->rename, &local->loc,
- &local->loc2, local->xattr_req);
- return 0;
+ return 0;
}
int
-dht_rename_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+dht_do_rename(call_frame_t *frame)
{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *dst_hashed = NULL;
+ xlator_t *src_cached = NULL;
+ xlator_t *dst_cached = NULL;
+ xlator_t *this = NULL;
+ xlator_t *rename_subvol = NULL;
+
+ local = frame->local;
+ this = frame->this;
+
+ dst_hashed = local->dst_hashed;
+ dst_cached = local->dst_cached;
+ src_cached = local->src_cached;
+
+ if (src_cached == dst_cached)
+ rename_subvol = src_cached;
+ else
+ rename_subvol = dst_hashed;
+
+ if ((src_cached != dst_hashed) && (rename_subvol == dst_hashed)) {
+ DHT_MARKER_DONT_ACCOUNT(local->xattr_req);
+ }
+
+ if (rename_subvol == src_cached) {
+ DHT_CHANGELOG_TRACK_AS_RENAME(local->xattr_req, &local->loc,
+ &local->loc2);
+ }
+
+ gf_msg_trace(this->name, 0, "renaming %s => %s (%s)", local->loc.path,
+ local->loc2.path, rename_subvol->name);
+
+ if (local->linked == _gf_true)
+ FRAME_SU_DO(frame, dht_local_t);
+ STACK_WIND_COOKIE(frame, dht_rename_cbk, rename_subvol, rename_subvol,
+ rename_subvol->fops->rename, &local->loc, &local->loc2,
+ local->xattr_req);
+ return 0;
+}
- local = frame->local;
- prev = cookie;
+int
+dht_rename_link_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
- if (op_ret == -1) {
- gf_msg_debug (this->name, 0,
- "link/file on %s failed (%s)",
- prev->name, strerror (op_errno));
- local->op_ret = -1;
- local->op_errno = op_errno;
- local->added_link = _gf_false;
- } else
- dht_iatt_merge (this, &local->stbuf, stbuf, prev);
+ local = frame->local;
+ prev = cookie;
- if (local->op_ret == -1)
- goto cleanup;
+ if (op_ret == -1) {
+ gf_msg_debug(this->name, 0, "link/file on %s failed (%s)", prev->name,
+ strerror(op_errno));
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ local->added_link = _gf_false;
+ } else
+ dht_iatt_merge(this, &local->stbuf, stbuf);
- dht_do_rename (frame);
+ if (local->op_ret == -1)
+ goto cleanup;
- return 0;
+ dht_do_rename(frame);
+
+ return 0;
cleanup:
- dht_rename_cleanup (frame);
+ dht_rename_cleanup(frame);
- return 0;
+ return 0;
}
int
-dht_rename_linkto_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+dht_rename_linkto_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- xlator_t *src_cached = NULL;
- dict_t *xattr = NULL;
-
- local = frame->local;
- DHT_MARK_FOP_INTERNAL (xattr);
- prev = cookie;
- src_cached = local->src_cached;
-
- if (op_ret == -1) {
- gf_msg_debug (this->name, 0,
- "link/file on %s failed (%s)",
- prev->name, strerror (op_errno));
- local->op_ret = -1;
- local->op_errno = op_errno;
- }
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ xlator_t *src_cached = NULL;
+ dict_t *xattr = NULL;
- /* If linkto creation failed move to failure cleanup code,
- * instead of continuing with creating the link file */
- if (local->op_ret != 0) {
- goto cleanup;
- }
+ local = frame->local;
+ DHT_MARK_FOP_INTERNAL(xattr);
+ prev = cookie;
+ src_cached = local->src_cached;
- gf_msg_trace (this->name, 0,
- "link %s => %s (%s)", local->loc.path,
- local->loc2.path, src_cached->name);
- if (gf_uuid_compare (local->loc.pargfid,
- local->loc2.pargfid) == 0) {
- DHT_MARKER_DONT_ACCOUNT(xattr);
- }
+ if (op_ret == -1) {
+ gf_msg_debug(this->name, 0, "link/file on %s failed (%s)", prev->name,
+ strerror(op_errno));
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ }
- local->added_link = _gf_true;
+ /* If linkto creation failed move to failure cleanup code,
+ * instead of continuing with creating the link file */
+ if (local->op_ret != 0) {
+ goto cleanup;
+ }
- STACK_WIND_COOKIE (frame, dht_rename_link_cbk, src_cached, src_cached,
- src_cached->fops->link, &local->loc, &local->loc2,
- xattr);
+ gf_msg_trace(this->name, 0, "link %s => %s (%s)", local->loc.path,
+ local->loc2.path, src_cached->name);
+ if (gf_uuid_compare(local->loc.pargfid, local->loc2.pargfid) == 0) {
+ DHT_MARKER_DONT_ACCOUNT(xattr);
+ }
- if (xattr)
- dict_unref (xattr);
+ local->added_link = _gf_true;
- return 0;
+ STACK_WIND_COOKIE(frame, dht_rename_link_cbk, src_cached, src_cached,
+ src_cached->fops->link, &local->loc, &local->loc2, xattr);
+
+ if (xattr)
+ dict_unref(xattr);
+
+ return 0;
cleanup:
- dht_rename_cleanup (frame);
+ dht_rename_cleanup(frame);
- if (xattr)
- dict_unref (xattr);
+ if (xattr)
+ dict_unref(xattr);
- return 0;
+ return 0;
}
int
-dht_rename_unlink_links_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+dht_rename_unlink_links_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *prev = NULL;
+ local = frame->local;
+ prev = cookie;
- local = frame->local;
- prev = cookie;
+ if ((op_ret == -1) && (op_errno != ENOENT)) {
+ gf_msg_debug(this->name, 0, "unlink of %s on %s failed (%s)",
+ local->loc2.path, prev->name, strerror(op_errno));
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ }
- if ((op_ret == -1) && (op_errno != ENOENT)) {
- gf_msg_debug (this->name, 0,
- "unlink of %s on %s failed (%s)",
- local->loc2.path, prev->name,
- strerror (op_errno));
- local->op_ret = -1;
- local->op_errno = op_errno;
- }
+ if (local->op_ret == -1)
+ goto cleanup;
- if (local->op_ret == -1)
- goto cleanup;
+ dht_do_rename(frame);
- dht_do_rename (frame);
-
- return 0;
+ return 0;
cleanup:
- dht_rename_cleanup (frame);
+ dht_rename_cleanup(frame);
- return 0;
+ return 0;
}
-
int
-dht_rename_create_links (call_frame_t *frame)
+dht_rename_create_links(call_frame_t *frame)
{
- dht_local_t *local = NULL;
- xlator_t *this = NULL;
- xlator_t *src_hashed = NULL;
- xlator_t *src_cached = NULL;
- xlator_t *dst_hashed = NULL;
- xlator_t *dst_cached = NULL;
- int call_cnt = 0;
- dict_t *xattr = NULL;
-
-
- local = frame->local;
- this = frame->this;
+ dht_local_t *local = NULL;
+ xlator_t *this = NULL;
+ xlator_t *src_hashed = NULL;
+ xlator_t *src_cached = NULL;
+ xlator_t *dst_hashed = NULL;
+ xlator_t *dst_cached = NULL;
+ int call_cnt = 0;
+ dict_t *xattr = NULL;
- src_hashed = local->src_hashed;
- src_cached = local->src_cached;
- dst_hashed = local->dst_hashed;
- dst_cached = local->dst_cached;
+ local = frame->local;
+ this = frame->this;
- DHT_MARK_FOP_INTERNAL (xattr);
+ src_hashed = local->src_hashed;
+ src_cached = local->src_cached;
+ dst_hashed = local->dst_hashed;
+ dst_cached = local->dst_cached;
- if (src_cached == dst_cached) {
- dict_t *xattr_new = NULL;
+ DHT_MARK_FOP_INTERNAL(xattr);
- if (dst_hashed == dst_cached)
- goto nolinks;
+ if (src_cached == dst_cached) {
+ dict_t *xattr_new = NULL;
- xattr_new = dict_copy_with_ref (xattr, NULL);
+ if (dst_hashed == dst_cached)
+ goto nolinks;
- gf_msg_trace (this->name, 0,
- "unlinking dst linkfile %s @ %s",
- local->loc2.path, dst_hashed->name);
+ xattr_new = dict_copy_with_ref(xattr, NULL);
- DHT_MARKER_DONT_ACCOUNT(xattr_new);
+ gf_msg_trace(this->name, 0, "unlinking dst linkfile %s @ %s",
+ local->loc2.path, dst_hashed->name);
- STACK_WIND_COOKIE (frame, dht_rename_unlink_links_cbk,
- dst_hashed, dst_hashed,
- dst_hashed->fops->unlink, &local->loc2, 0,
- xattr_new);
+ DHT_MARKER_DONT_ACCOUNT(xattr_new);
- dict_unref (xattr_new);
- if (xattr)
- dict_unref (xattr);
+ STACK_WIND_COOKIE(frame, dht_rename_unlink_links_cbk, dst_hashed,
+ dst_hashed, dst_hashed->fops->unlink, &local->loc2, 0,
+ xattr_new);
- return 0;
- }
+ dict_unref(xattr_new);
+ if (xattr)
+ dict_unref(xattr);
- if (src_cached != dst_hashed) {
- /* needed to create the link file */
- call_cnt++;
- if (dst_hashed != src_hashed)
- /* needed to create the linkto file */
- call_cnt ++;
+ return 0;
+ }
+
+ if (src_cached != dst_hashed) {
+ /* needed to create the link file */
+ call_cnt++;
+ if (dst_hashed != src_hashed)
+ /* needed to create the linkto file */
+ call_cnt++;
+ }
+
+ /* We should not have any failures post the link creation, as this
+ * introduces the newname into the namespace. Clients could have cached
+ * the existence of the newname and may start taking actions based on
+ * the same. Hence create the linkto first, and then attempt the link.
+ *
+ * NOTE: If another client is attempting the same oldname -> newname
+ * rename, and finds both file names as existing, and are hard links
+ * to each other, then FUSE would send in an unlink for oldname. In
+ * this time duration if we treat the linkto as a critical error and
+ * unlink the newname we created, we would have effectively lost the
+ * file to rename operations. */
+ if (dst_hashed != src_hashed && src_cached != dst_hashed) {
+ gf_msg_trace(this->name, 0, "linkfile %s @ %s => %s", local->loc.path,
+ dst_hashed->name, src_cached->name);
+
+ memcpy(local->gfid, local->loc.inode->gfid, 16);
+ dht_linkfile_create(frame, dht_rename_linkto_cbk, this, src_cached,
+ dst_hashed, &local->loc);
+ } else if (src_cached != dst_hashed) {
+ dict_t *xattr_new = NULL;
+
+ xattr_new = dict_copy_with_ref(xattr, NULL);
+
+ gf_msg_trace(this->name, 0, "link %s => %s (%s)", local->loc.path,
+ local->loc2.path, src_cached->name);
+ if (gf_uuid_compare(local->loc.pargfid, local->loc2.pargfid) == 0) {
+ DHT_MARKER_DONT_ACCOUNT(xattr_new);
}
- /* We should not have any failures post the link creation, as this
- * introduces the newname into the namespace. Clients could have cached
- * the existence of the newname and may start taking actions based on
- * the same. Hence create the linkto first, and then attempt the link.
- *
- * NOTE: If another client is attempting the same oldname -> newname
- * rename, and finds both file names as existing, and are hard links
- * to each other, then FUSE would send in an unlink for oldname. In
- * this time duration if we treat the linkto as a critical error and
- * unlink the newname we created, we would have effectively lost the
- * file to rename operations. */
- if (dst_hashed != src_hashed && src_cached != dst_hashed) {
- gf_msg_trace (this->name, 0,
- "linkfile %s @ %s => %s",
- local->loc.path, dst_hashed->name,
- src_cached->name);
-
- memcpy (local->gfid, local->loc.inode->gfid, 16);
- dht_linkfile_create (frame, dht_rename_linkto_cbk, this,
- src_cached, dst_hashed, &local->loc);
- } else if (src_cached != dst_hashed) {
- dict_t *xattr_new = NULL;
-
- xattr_new = dict_copy_with_ref (xattr, NULL);
-
- gf_msg_trace (this->name, 0,
- "link %s => %s (%s)", local->loc.path,
- local->loc2.path, src_cached->name);
- if (gf_uuid_compare (local->loc.pargfid,
- local->loc2.pargfid) == 0) {
- DHT_MARKER_DONT_ACCOUNT(xattr_new);
- }
-
- local->added_link = _gf_true;
+ local->added_link = _gf_true;
- STACK_WIND_COOKIE (frame, dht_rename_link_cbk, src_cached,
- src_cached, src_cached->fops->link,
- &local->loc, &local->loc2, xattr_new);
+ STACK_WIND_COOKIE(frame, dht_rename_link_cbk, src_cached, src_cached,
+ src_cached->fops->link, &local->loc, &local->loc2,
+ xattr_new);
- dict_unref (xattr_new);
- }
+ dict_unref(xattr_new);
+ }
nolinks:
- if (!call_cnt) {
- /* skip to next step */
- dht_do_rename (frame);
- }
- if (xattr)
- dict_unref (xattr);
-
- return 0;
+ if (!call_cnt) {
+ /* skip to next step */
+ dht_do_rename(frame);
+ }
+ if (xattr)
+ dict_unref(xattr);
+
+ return 0;
}
int
-dht_rename_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, dict_t *xattr,
- struct iatt *postparent)
+dht_rename_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *stbuf, dict_t *xattr,
+ struct iatt *postparent)
{
- dht_local_t *local = NULL;
- int call_cnt = 0;
- dht_conf_t *conf = NULL;
- char gfid_local[GF_UUID_BUF_SIZE] = {0};
- char gfid_server[GF_UUID_BUF_SIZE] = {0};
- int child_index = -1;
-
-
- child_index = (long)cookie;
-
- local = frame->local;
- conf = this->private;
-
- if (op_ret < 0) {
- /* The meaning of is_linkfile is overloaded here. For locking
- * to work properly both rebalance and rename should acquire
- * lock on datafile. The reason for sending this lookup is to
- * find out whether we've acquired a lock on data file.
- * Between the lookup before rename and this rename, the
- * file could be migrated by a rebalance process and now this
- * file this might be a linkto file. We verify that by sending
- * this lookup. However, if this lookup fails we cannot really
- * say whether we've acquired lock on a datafile or linkto file.
- * So, we act conservatively and _assume_
- * that this is a linkfile and fail the rename operation.
- */
- local->is_linkfile = _gf_true;
- local->op_errno = op_errno;
- } else if (xattr && check_is_linkfile (inode, stbuf, xattr,
- conf->link_xattr_name)) {
- local->is_linkfile = _gf_true;
- /* Found linkto file instead of data file, passdown ENOENT
- * based on the above comment */
- local->op_errno = ENOENT;
- }
-
- if (!local->is_linkfile &&
- gf_uuid_compare (local->lock[0].layout.parent_layout.locks[child_index]->loc.gfid,
- stbuf->ia_gfid)) {
- gf_uuid_unparse (local->lock[0].layout.parent_layout.locks[child_index]->loc.gfid,
- gfid_local);
- gf_uuid_unparse (stbuf->ia_gfid, gfid_server);
-
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_GFID_MISMATCH,
- "path:%s, received a different gfid, local_gfid= %s"
- " server_gfid: %s",
- local->loc.path, gfid_local, gfid_server);
-
- /* Will passdown ENOENT anyway since the file we sent on
- * rename is replaced with a different file */
- local->op_errno = ENOENT;
- /* Since local->is_linkfile is used here to detect failure,
- * marking this to true */
- local->is_linkfile = _gf_true;
- }
-
- call_cnt = dht_frame_return (frame);
- if (is_last_call (call_cnt)) {
- if (local->is_linkfile) {
- local->op_ret = -1;
- goto fail;
+ dht_local_t *local = NULL;
+ int call_cnt = 0;
+ dht_conf_t *conf = NULL;
+ char gfid_local[GF_UUID_BUF_SIZE] = {0};
+ char gfid_server[GF_UUID_BUF_SIZE] = {0};
+ int child_index = -1;
+ gf_boolean_t is_src = _gf_false;
+ loc_t *loc = NULL;
+
+ child_index = (long)cookie;
+
+ local = frame->local;
+ conf = this->private;
+
+ is_src = (child_index == 0);
+ if (is_src)
+ loc = &local->loc;
+ else
+ loc = &local->loc2;
+
+ if (op_ret >= 0) {
+ if (is_src)
+ local->src_cached = dht_subvol_get_cached(this, local->loc.inode);
+ else {
+ if (loc->inode)
+ gf_uuid_unparse(loc->inode->gfid, gfid_local);
+
+ gf_msg_debug(this->name, 0,
+ "dst_cached before lookup: %s, "
+ "(path:%s)(gfid:%s),",
+ local->loc2.path,
+ local->dst_cached ? local->dst_cached->name : NULL,
+ local->dst_cached ? gfid_local : NULL);
+
+ local->dst_cached = dht_subvol_get_cached(this,
+ local->loc2_copy.inode);
+
+ gf_uuid_unparse(stbuf->ia_gfid, gfid_local);
+
+ gf_msg_debug(this->name, GF_LOG_WARNING,
+ "dst_cached after lookup: %s, "
+ "(path:%s)(gfid:%s)",
+ local->loc2.path,
+ local->dst_cached ? local->dst_cached->name : NULL,
+ local->dst_cached ? gfid_local : NULL);
+
+ if ((local->loc2.inode == NULL) ||
+ gf_uuid_compare(stbuf->ia_gfid, local->loc2.inode->gfid)) {
+ if (local->loc2.inode != NULL) {
+ inode_unlink(local->loc2.inode, local->loc2.parent,
+ local->loc2.name);
+ inode_unref(local->loc2.inode);
}
- dht_rename_create_links (frame);
- }
-
- return 0;
+ local->loc2.inode = inode_link(local->loc2_copy.inode,
+ local->loc2_copy.parent,
+ local->loc2_copy.name, stbuf);
+ gf_uuid_copy(local->loc2.gfid, stbuf->ia_gfid);
+ }
+ }
+ }
+
+ if (op_ret < 0) {
+ if (is_src) {
+ /* The meaning of is_linkfile is overloaded here. For locking
+ * to work properly both rebalance and rename should acquire
+ * lock on datafile. The reason for sending this lookup is to
+ * find out whether we've acquired a lock on data file.
+ * Between the lookup before rename and this rename, the
+ * file could be migrated by a rebalance process and now this
+ * file this might be a linkto file. We verify that by sending
+ * this lookup. However, if this lookup fails we cannot really
+ * say whether we've acquired lock on a datafile or linkto file.
+ * So, we act conservatively and _assume_
+ * that this is a linkfile and fail the rename operation.
+ */
+ local->is_linkfile = _gf_true;
+ local->op_errno = op_errno;
+ } else {
+ if (local->dst_cached)
+ gf_msg_debug(this->name, op_errno,
+ "file %s (gfid:%s) was present "
+ "(hashed-subvol=%s, "
+ "cached-subvol=%s) before rename,"
+ " but lookup failed",
+ local->loc2.path,
+ uuid_utoa(local->loc2.inode->gfid),
+ local->dst_hashed->name, local->dst_cached->name);
+ if (dht_inode_missing(op_errno))
+ local->dst_cached = NULL;
+ }
+ } else if (is_src && xattr &&
+ check_is_linkfile(inode, stbuf, xattr, conf->link_xattr_name)) {
+ local->is_linkfile = _gf_true;
+ /* Found linkto file instead of data file, passdown ENOENT
+ * based on the above comment */
+ local->op_errno = ENOENT;
+ }
+
+ if (!local->is_linkfile && (op_ret >= 0) &&
+ gf_uuid_compare(loc->gfid, stbuf->ia_gfid)) {
+ gf_uuid_unparse(loc->gfid, gfid_local);
+ gf_uuid_unparse(stbuf->ia_gfid, gfid_server);
+
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_GFID_MISMATCH,
+ "path:%s, received a different gfid, local_gfid= %s"
+ " server_gfid: %s",
+ local->loc.path, gfid_local, gfid_server);
+
+ /* Will passdown ENOENT anyway since the file we sent on
+ * rename is replaced with a different file */
+ local->op_errno = ENOENT;
+ /* Since local->is_linkfile is used here to detect failure,
+ * marking this to true */
+ local->is_linkfile = _gf_true;
+ }
+
+ call_cnt = dht_frame_return(frame);
+ if (is_last_call(call_cnt)) {
+ if (local->is_linkfile) {
+ local->op_ret = -1;
+ goto fail;
+ }
+
+ dht_rename_create_links(frame);
+ }
+
+ return 0;
fail:
- dht_rename_unlock (frame, this);
- return 0;
+ dht_rename_unlock(frame, this);
+ return 0;
}
-int32_t
-dht_rename_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+int
+dht_rename_file_lock1_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- char src_gfid[GF_UUID_BUF_SIZE] = {0};
- char dst_gfid[GF_UUID_BUF_SIZE] = {0};
- dict_t *xattr_req = NULL;
- dht_conf_t *conf = NULL;
- int i = 0;
- int count = 0;
-
-
- local = frame->local;
- conf = this->private;
-
- if (op_ret < 0) {
- uuid_utoa_r (local->loc.inode->gfid, src_gfid);
-
- if (local->loc2.inode)
- uuid_utoa_r (local->loc2.inode->gfid, dst_gfid);
-
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_INODE_LK_ERROR,
- "acquiring inodelk failed "
- "rename (%s:%s:%s %s:%s:%s)",
- local->loc.path, src_gfid, local->src_cached->name,
- local->loc2.path, dst_gfid,
- local->dst_cached ? local->dst_cached->name : NULL);
+ dht_local_t *local = NULL;
+ char src_gfid[GF_UUID_BUF_SIZE] = {0};
+ char dst_gfid[GF_UUID_BUF_SIZE] = {0};
+ int ret = 0;
+ loc_t *loc = NULL;
+ xlator_t *subvol = NULL;
+
+ local = frame->local;
+
+ if (op_ret < 0) {
+ uuid_utoa_r(local->loc.inode->gfid, src_gfid);
+
+ if (local->loc2.inode)
+ uuid_utoa_r(local->loc2.inode->gfid, dst_gfid);
+
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, DHT_MSG_INODE_LK_ERROR,
+ "protecting namespace of %s failed"
+ "rename (%s:%s:%s %s:%s:%s)",
+ local->current == &local->lock[0] ? local->loc.path
+ : local->loc2.path,
+ local->loc.path, src_gfid, local->src_hashed->name,
+ local->loc2.path, dst_gfid,
+ local->dst_hashed ? local->dst_hashed->name : NULL);
+
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+ goto err;
+ }
+
+ if (local->current == &local->lock[0]) {
+ loc = &local->loc2;
+ subvol = local->dst_hashed;
+ local->current = &local->lock[1];
+ } else {
+ loc = &local->loc;
+ subvol = local->src_hashed;
+ local->current = &local->lock[0];
+ }
+
+ ret = dht_protect_namespace(frame, loc, subvol, &local->current->ns,
+ dht_rename_lock_cbk);
+ if (ret < 0) {
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ return 0;
+err:
+ /* No harm in calling an extra unlock */
+ dht_rename_unlock(frame, this);
+ return 0;
+}
- local->op_ret = -1;
- local->op_errno = op_errno;
+int32_t
+dht_rename_file_protect_namespace(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ char src_gfid[GF_UUID_BUF_SIZE] = {0};
+ char dst_gfid[GF_UUID_BUF_SIZE] = {0};
+ int ret = 0;
+ loc_t *loc = NULL;
+ xlator_t *subvol = NULL;
+
+ local = frame->local;
+
+ if (op_ret < 0) {
+ uuid_utoa_r(local->loc.inode->gfid, src_gfid);
+
+ if (local->loc2.inode)
+ uuid_utoa_r(local->loc2.inode->gfid, dst_gfid);
+
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, DHT_MSG_INODE_LK_ERROR,
+ "acquiring inodelk failed "
+ "rename (%s:%s:%s %s:%s:%s)",
+ local->loc.path, src_gfid, local->src_cached->name,
+ local->loc2.path, dst_gfid,
+ local->dst_cached ? local->dst_cached->name : NULL);
+
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+
+ goto err;
+ }
+
+ /* Locks on src and dst needs to ordered which otherwise might cause
+ * deadlocks when rename (src, dst) and rename (dst, src) is done from
+ * two different clients
+ */
+ ret = dht_order_rename_lock(frame, &loc, &subvol);
+ if (ret) {
+ local->op_errno = ENOMEM;
+ goto err;
+ }
+
+ ret = dht_protect_namespace(frame, loc, subvol, &local->current->ns,
+ dht_rename_file_lock1_cbk);
+ if (ret < 0) {
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ return 0;
- goto done;
- }
+err:
+ /* Its fine to call unlock even when no locks are acquired, as we check
+ * for lock->locked before winding a unlock call.
+ */
+ dht_rename_unlock(frame, this);
- xattr_req = dict_new ();
- if (xattr_req == NULL) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto done;
- }
+ return 0;
+}
- op_ret = dict_set_uint32 (xattr_req,
- conf->link_xattr_name, 256);
- if (op_ret < 0) {
- local->op_ret = -1;
- local->op_errno = -op_ret;
- goto done;
+int32_t
+dht_rename_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+ dht_local_t *local = NULL;
+ char src_gfid[GF_UUID_BUF_SIZE] = {0};
+ char dst_gfid[GF_UUID_BUF_SIZE] = {0};
+ dict_t *xattr_req = NULL;
+ dht_conf_t *conf = NULL;
+ int i = 0;
+ xlator_t *subvol = NULL;
+ dht_lock_t *lock = NULL;
+
+ local = frame->local;
+ conf = this->private;
+
+ if (op_ret < 0) {
+ uuid_utoa_r(local->loc.inode->gfid, src_gfid);
+
+ if (local->loc2.inode)
+ uuid_utoa_r(local->loc2.inode->gfid, dst_gfid);
+
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, DHT_MSG_INODE_LK_ERROR,
+ "protecting namespace of %s failed. "
+ "rename (%s:%s:%s %s:%s:%s)",
+ local->current == &local->lock[0] ? local->loc.path
+ : local->loc2.path,
+ local->loc.path, src_gfid, local->src_hashed->name,
+ local->loc2.path, dst_gfid,
+ local->dst_hashed ? local->dst_hashed->name : NULL);
+
+ local->op_ret = -1;
+ local->op_errno = op_errno;
+
+ goto done;
+ }
+
+ xattr_req = dict_new();
+ if (xattr_req == NULL) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto done;
+ }
+
+ op_ret = dict_set_uint32(xattr_req, conf->link_xattr_name, 256);
+ if (op_ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -op_ret;
+ goto done;
+ }
+
+ /* dst_cached might've changed. This normally happens for two reasons:
+ * 1. rebalance migrated dst
+ * 2. Another parallel rename was done overwriting dst
+ *
+ * Doing a lookup on local->loc2 when dst exists, but is associated
+ * with a different gfid will result in an ESTALE error. So, do a fresh
+ * lookup with a new inode on dst-path and handle change of dst-cached
+ * in the cbk. Also, to identify dst-cached changes we do a lookup on
+ * "this" rather than the subvol.
+ */
+ loc_copy(&local->loc2_copy, &local->loc2);
+ inode_unref(local->loc2_copy.inode);
+ local->loc2_copy.inode = inode_new(local->loc.inode->table);
+
+ /* Why not use local->lock.locks[?].loc for lookup post lock phase
+ * ---------------------------------------------------------------
+ * "layout.parent_layout.locks[?].loc" does not have the name and pargfid
+ * populated.
+ * Reason: If we had populated the name and pargfid, server might
+ * resolve to a successful lookup even if there is a file with same name
+ * with a different gfid(unlink & create) as server does name based
+ * resolution on first priority. And this can result in operating on a
+ * different inode entirely.
+ *
+ * Now consider a scenario where source file was renamed by some other
+ * client to a new name just before this lock was granted. So if a
+ * lookup would be done on local->lock[0].layout.parent_layout.locks[?].loc,
+ * server will send success even if the entry was renamed (since server will
+ * do a gfid based resolution). So once a lock is granted, make sure the
+ * file exists with the name that the client requested with.
+ * */
+
+ local->call_cnt = 2;
+ for (i = 0; i < 2; i++) {
+ if (i == 0) {
+ lock = local->rename_inodelk_backward_compatible[0];
+ if (gf_uuid_compare(local->loc.gfid, lock->loc.gfid) == 0)
+ subvol = lock->xl;
+ else {
+ lock = local->rename_inodelk_backward_compatible[1];
+ subvol = lock->xl;
+ }
+ } else {
+ subvol = this;
}
- count = local->call_cnt = local->lock[0].layout.parent_layout.lk_count;
-
- /* Why not use local->lock.locks[?].loc for lookup post lock phase
- * ---------------------------------------------------------------
- * "layout.parent_layout.locks[?].loc" does not have the name and pargfid
- * populated.
- * Reason: If we had populated the name and pargfid, server might
- * resolve to a successful lookup even if there is a file with same name
- * with a different gfid(unlink & create) as server does name based
- * resolution on first priority. And this can result in operating on a
- * different inode entirely.
- *
- * Now consider a scenario where source file was renamed by some other
- * client to a new name just before this lock was granted. So if a
- * lookup would be done on local->lock[0].layout.parent_layout.locks[?].loc,
- * server will send success even if the entry was renamed (since server will
- * do a gfid based resolution). So once a lock is granted, make sure the file
- * exists with the name that the client requested with.
- * */
-
- for (i = 0; i < count; i++) {
- STACK_WIND_COOKIE (frame, dht_rename_lookup_cbk, (void *)(long)i,
- local->lock[0].layout.parent_layout.locks[i]->xl,
- local->lock[0].layout.parent_layout.locks[i]->xl->fops->lookup,
- ((gf_uuid_compare (local->loc.gfid, \
- local->lock[0].layout.parent_layout.locks[i]->loc.gfid) == 0) ?
- &local->loc : &local->loc2), xattr_req);
- }
+ STACK_WIND_COOKIE(frame, dht_rename_lookup_cbk, (void *)(long)i, subvol,
+ subvol->fops->lookup,
+ (i == 0) ? &local->loc : &local->loc2_copy,
+ xattr_req);
+ }
- dict_unref (xattr_req);
- return 0;
+ dict_unref(xattr_req);
+ return 0;
done:
- /* Its fine to call unlock even when no locks are acquired, as we check
- * for lock->locked before winding a unlock call.
- */
- dht_rename_unlock (frame, this);
+ /* Its fine to call unlock even when no locks are acquired, as we check
+ * for lock->locked before winding a unlock call.
+ */
+ dht_rename_unlock(frame, this);
- if (xattr_req)
- dict_unref (xattr_req);
+ if (xattr_req)
+ dict_unref(xattr_req);
- return 0;
+ return 0;
}
int
-dht_rename_lock (call_frame_t *frame)
+dht_rename_lock(call_frame_t *frame)
{
- dht_local_t *local = NULL;
- int count = 1, ret = -1;
- dht_lock_t **lk_array = NULL;
-
- local = frame->local;
-
- if (local->dst_cached)
- count++;
-
- lk_array = GF_CALLOC (count, sizeof (*lk_array), gf_common_mt_pointer);
- if (lk_array == NULL)
- goto err;
-
- lk_array[0] = dht_lock_new (frame->this, local->src_cached, &local->loc,
- F_WRLCK, DHT_FILE_MIGRATE_DOMAIN, NULL,
- FAIL_ON_ANY_ERROR);
- if (lk_array[0] == NULL)
- goto err;
-
- if (local->dst_cached) {
- lk_array[1] = dht_lock_new (frame->this, local->dst_cached,
- &local->loc2, F_WRLCK,
- DHT_FILE_MIGRATE_DOMAIN, NULL,
- FAIL_ON_ANY_ERROR);
- if (lk_array[1] == NULL)
- goto err;
- }
-
- local->lock[0].layout.parent_layout.locks = lk_array;
- local->lock[0].layout.parent_layout.lk_count = count;
-
- ret = dht_blocking_inodelk (frame, lk_array, count,
- dht_rename_lock_cbk);
- if (ret < 0) {
- local->lock[0].layout.parent_layout.locks = NULL;
- local->lock[0].layout.parent_layout.lk_count = 0;
- goto err;
- }
-
- return 0;
+ dht_local_t *local = NULL;
+ int count = 1, ret = -1;
+ dht_lock_t **lk_array = NULL;
+
+ local = frame->local;
+
+ if (local->dst_cached)
+ count++;
+
+ lk_array = GF_CALLOC(count, sizeof(*lk_array), gf_common_mt_pointer);
+ if (lk_array == NULL)
+ goto err;
+
+ lk_array[0] = dht_lock_new(frame->this, local->src_cached, &local->loc,
+ F_WRLCK, DHT_FILE_MIGRATE_DOMAIN, NULL,
+ FAIL_ON_ANY_ERROR);
+ if (lk_array[0] == NULL)
+ goto err;
+
+ if (local->dst_cached) {
+ /* dst might be removed by the time inodelk reaches bricks,
+ * which can result in ESTALE errors. POSIX imposes no
+ * restriction for dst to be present for renames to be
+ * successful. So, we'll ignore ESTALE errors. As far as
+ * synchronization on dst goes, we'll achieve the same by
+ * holding entrylk on parent directory of dst in the namespace
+ * of basename(dst). Also, there might not be quorum in cluster
+ * xlators like EC/disperse on errno, in which case they return
+ * EIO. For eg., in a disperse (4 + 2), 3 might return success
+ * and three might return ESTALE. Disperse, having no Quorum
+ * unwinds inodelk with EIO. So, ignore EIO too.
+ */
+ lk_array[1] = dht_lock_new(frame->this, local->dst_cached, &local->loc2,
+ F_WRLCK, DHT_FILE_MIGRATE_DOMAIN, NULL,
+ IGNORE_ENOENT_ESTALE_EIO);
+ if (lk_array[1] == NULL)
+ goto err;
+ }
+
+ local->rename_inodelk_backward_compatible = lk_array;
+ local->rename_inodelk_bc_count = count;
+
+ /* retaining inodelks for the sake of backward compatibility. Please
+ * make sure to remove this inodelk once all of 3.10, 3.12 and 3.13
+ * reach EOL. Better way of getting synchronization would be to acquire
+ * entrylks on src and dst parent directories in the namespace of
+ * basenames of src and dst
+ */
+ ret = dht_blocking_inodelk(frame, lk_array, count,
+ dht_rename_file_protect_namespace);
+ if (ret < 0) {
+ local->rename_inodelk_backward_compatible = NULL;
+ local->rename_inodelk_bc_count = 0;
+ goto err;
+ }
+
+ return 0;
err:
- if (lk_array != NULL) {
- int tmp_count = 0, i = 0;
+ if (lk_array != NULL) {
+ int tmp_count = 0, i = 0;
- for (i = 0; (i < count) && (lk_array[i]); i++, tmp_count++);
+ for (i = 0; (i < count) && (lk_array[i]); i++, tmp_count++)
+ ;
- dht_lock_array_free (lk_array, tmp_count);
- GF_FREE (lk_array);
- }
+ dht_lock_array_free(lk_array, tmp_count);
+ GF_FREE(lk_array);
+ }
- return -1;
+ return -1;
}
int
-dht_rename (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+dht_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- xlator_t *src_cached = NULL;
- xlator_t *src_hashed = NULL;
- xlator_t *dst_cached = NULL;
- xlator_t *dst_hashed = NULL;
- int op_errno = -1;
- int ret = -1;
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (oldloc, err);
- VALIDATE_OR_GOTO (newloc, err);
-
- conf = this->private;
-
- if (conf->subvolume_cnt == 1) {
- default_rename (frame, this, oldloc, newloc, xdata);
- return 0;
- }
-
- gf_uuid_unparse(oldloc->inode->gfid, gfid);
-
- src_hashed = dht_subvol_get_hashed (this, oldloc);
- if (!src_hashed) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_RENAME_FAILED,
- "No hashed subvolume in layout for path=%s,"
- "(gfid = %s)", oldloc->path, gfid);
- op_errno = EINVAL;
- goto err;
- }
-
- src_cached = dht_subvol_get_cached (this, oldloc->inode);
- if (!src_cached) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_RENAME_FAILED,
- "No cached subvolume for path = %s,"
- "(gfid = %s)", oldloc->path, gfid);
-
- op_errno = EINVAL;
- goto err;
- }
-
- dst_hashed = dht_subvol_get_hashed (this, newloc);
- if (!dst_hashed) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_RENAME_FAILED,
- "No hashed subvolume in layout for path=%s",
- newloc->path);
- op_errno = EINVAL;
- goto err;
- }
-
- if (newloc->inode)
- dst_cached = dht_subvol_get_cached (this, newloc->inode);
-
- local = dht_local_init (frame, oldloc, NULL, GF_FOP_RENAME);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- /* cached_subvol will be set from dht_local_init, reset it to NULL,
- as the logic of handling rename is different */
- local->cached_subvol = NULL;
-
- ret = loc_copy (&local->loc2, newloc);
- if (ret == -1) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local->src_hashed = src_hashed;
- local->src_cached = src_cached;
- local->dst_hashed = dst_hashed;
- local->dst_cached = dst_cached;
- if (xdata)
- local->xattr_req = dict_ref (xdata);
-
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_RENAME_INFO,
- "renaming %s (hash=%s/cache=%s) => %s (hash=%s/cache=%s)",
- oldloc->path, src_hashed->name, src_cached->name,
- newloc->path, dst_hashed->name,
- dst_cached ? dst_cached->name : "<nul>");
-
- if (IA_ISDIR (oldloc->inode->ia_type)) {
- dht_rename_dir (frame, this);
- } else {
- local->op_ret = 0;
- ret = dht_rename_lock (frame);
- if (ret < 0)
- goto err;
+ xlator_t *src_cached = NULL;
+ xlator_t *src_hashed = NULL;
+ xlator_t *dst_cached = NULL;
+ xlator_t *dst_hashed = NULL;
+ int op_errno = -1;
+ int ret = -1;
+ dht_local_t *local = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ char newgfid[GF_UUID_BUF_SIZE] = {0};
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(oldloc, err);
+ VALIDATE_OR_GOTO(newloc, err);
+
+ gf_uuid_unparse(oldloc->inode->gfid, gfid);
+
+ src_hashed = dht_subvol_get_hashed(this, oldloc);
+ if (!src_hashed) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_RENAME_FAILED,
+ "No hashed subvolume in layout for path=%s,"
+ "(gfid = %s)",
+ oldloc->path, gfid);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ src_cached = dht_subvol_get_cached(this, oldloc->inode);
+ if (!src_cached) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_RENAME_FAILED,
+ "No cached subvolume for path = %s,"
+ "(gfid = %s)",
+ oldloc->path, gfid);
+
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ dst_hashed = dht_subvol_get_hashed(this, newloc);
+ if (!dst_hashed) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_RENAME_FAILED,
+ "No hashed subvolume in layout for path=%s", newloc->path);
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ if (newloc->inode)
+ dst_cached = dht_subvol_get_cached(this, newloc->inode);
+
+ local = dht_local_init(frame, oldloc, NULL, GF_FOP_RENAME);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+ /* cached_subvol will be set from dht_local_init, reset it to NULL,
+ as the logic of handling rename is different */
+ local->cached_subvol = NULL;
+
+ ret = loc_copy(&local->loc2, newloc);
+ if (ret == -1) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->src_hashed = src_hashed;
+ local->src_cached = src_cached;
+ local->dst_hashed = dst_hashed;
+ local->dst_cached = dst_cached;
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ if (newloc->inode)
+ gf_uuid_unparse(newloc->inode->gfid, newgfid);
+
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_RENAME_INFO,
+ "renaming %s (%s) (hash=%s/cache=%s) => %s (%s) "
+ "(hash=%s/cache=%s) ",
+ oldloc->path, gfid, src_hashed->name, src_cached->name, newloc->path,
+ newloc->inode ? newgfid : NULL, dst_hashed->name,
+ dst_cached ? dst_cached->name : "<nul>");
+
+ if (IA_ISDIR(oldloc->inode->ia_type)) {
+ dht_rename_dir(frame, this);
+ } else {
+ local->op_ret = 0;
+ ret = dht_rename_lock(frame);
+ if (ret < 0) {
+ op_errno = ENOMEM;
+ goto err;
}
+ }
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
- NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(rename, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL,
+ NULL);
- return 0;
+ return 0;
+}
+
+int
+dht_pt_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
+{
+ gf_boolean_t free_xdata = _gf_false;
+
+ /* Just a pass through */
+ if (!IA_ISDIR(oldloc->inode->ia_type)) {
+ if (!xdata) {
+ free_xdata = _gf_true;
+ }
+ DHT_CHANGELOG_TRACK_AS_RENAME(xdata, oldloc, newloc);
+ }
+ default_rename(frame, this, oldloc, newloc, xdata);
+ if (free_xdata && xdata) {
+ dict_unref(xdata);
+ xdata = NULL;
+ }
+ return 0;
}
diff --git a/xlators/cluster/dht/src/dht-selfheal.c b/xlators/cluster/dht/src/dht-selfheal.c
index 49540a04c9e..3e24065227c 100644
--- a/xlators/cluster/dht/src/dht-selfheal.c
+++ b/xlators/cluster/dht/src/dht-selfheal.c
@@ -8,2923 +8,2531 @@
cases as published by the Free Software Foundation.
*/
-
-#include "glusterfs.h"
-#include "xlator.h"
-#include "dht-common.h"
-#include "dht-messages.h"
#include "dht-lock.h"
-#include "glusterfs-acl.h"
-
-#define DHT_SET_LAYOUT_RANGE(layout,i,srt,chunk,path) do { \
- layout->list[i].start = srt; \
- layout->list[i].stop = srt + chunk - 1; \
- layout->list[i].commit_hash = layout->commit_hash; \
- \
- gf_msg_trace (this->name, 0, \
- "gave fix: %u - %u, with commit-hash %u" \
- " on %s for %s", \
- layout->list[i].start, \
- layout->list[i].stop, \
- layout->list[i].commit_hash, \
- layout->list[i].xlator->name, path); \
- } while (0)
-
-#define DHT_RESET_LAYOUT_RANGE(layout) do { \
- int cnt = 0; \
- for (cnt = 0; cnt < layout->cnt; cnt++ ) { \
- layout->list[cnt].start = 0; \
- layout->list[cnt].stop = 0; \
- } \
- } while (0)
-int
-dht_selfheal_layout_lock (call_frame_t *frame, dht_layout_t *layout,
- gf_boolean_t newdir,
- dht_selfheal_layout_t healer,
- dht_need_heal_t should_heal);
+#define DHT_SET_LAYOUT_RANGE(layout, i, srt, chunk, path) \
+ do { \
+ layout->list[i].start = srt; \
+ layout->list[i].stop = srt + chunk - 1; \
+ layout->list[i].commit_hash = layout->commit_hash; \
+ \
+ gf_msg_trace(this->name, 0, \
+ "gave fix: 0x%x - 0x%x, with commit-hash 0x%x" \
+ " on %s for %s", \
+ layout->list[i].start, layout->list[i].stop, \
+ layout->list[i].commit_hash, \
+ layout->list[i].xlator->name, path); \
+ } while (0)
+
+#define DHT_RESET_LAYOUT_RANGE(layout) \
+ do { \
+ int cnt = 0; \
+ for (cnt = 0; cnt < layout->cnt; cnt++) { \
+ layout->list[cnt].start = 0; \
+ layout->list[cnt].stop = 0; \
+ } \
+ } while (0)
+
+static int
+dht_selfheal_layout_lock(call_frame_t *frame, dht_layout_t *layout,
+ gf_boolean_t newdir, dht_selfheal_layout_t healer,
+ dht_need_heal_t should_heal);
static uint32_t
-dht_overlap_calc (dht_layout_t *old, int o, dht_layout_t *new, int n)
+dht_overlap_calc(dht_layout_t *old, int o, dht_layout_t *new, int n)
{
- if (o >= old->cnt || n >= new->cnt)
- return 0;
+ if (o >= old->cnt || n >= new->cnt)
+ return 0;
- if (old->list[o].err > 0 || new->list[n].err > 0)
- return 0;
+ if (old->list[o].err > 0 || new->list[n].err > 0)
+ return 0;
- if (old->list[o].start == old->list[o].stop) {
- return 0;
- }
+ if (old->list[o].start == old->list[o].stop) {
+ return 0;
+ }
- if (new->list[n].start == new->list[n].stop) {
- return 0;
- }
+ if (new->list[n].start == new->list[n].stop) {
+ return 0;
+ }
- if ((old->list[o].start > new->list[n].stop) ||
- (old->list[o].stop < new->list[n].start))
- return 0;
+ if ((old->list[o].start > new->list[n].stop) ||
+ (old->list[o].stop < new->list[n].start))
+ return 0;
- return min (old->list[o].stop, new->list[n].stop) -
- max (old->list[o].start, new->list[n].start) + 1;
+ return min(old->list[o].stop, new->list[n].stop) -
+ max(old->list[o].start, new->list[n].start) + 1;
}
int
-dht_selfheal_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+dht_selfheal_unlock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- DHT_STACK_DESTROY (frame);
- return 0;
+ DHT_STACK_DESTROY(frame);
+ return 0;
}
int
-dht_selfheal_dir_finish (call_frame_t *frame, xlator_t *this, int ret,
- int invoke_cbk)
+dht_selfheal_dir_finish(call_frame_t *frame, xlator_t *this, int ret,
+ int invoke_cbk)
{
- dht_local_t *local = NULL, *lock_local = NULL;
- call_frame_t *lock_frame = NULL;
- int lock_count = 0;
+ dht_local_t *local = NULL, *lock_local = NULL;
+ call_frame_t *lock_frame = NULL;
+ int lock_count = 0;
- local = frame->local;
+ local = frame->local;
- /* Unlock entrylk */
- dht_unlock_entrylk_wrapper (frame, &local->lock[0].ns.directory_ns);
+ /* Unlock entrylk */
+ dht_unlock_entrylk_wrapper(frame, &local->lock[0].ns.directory_ns);
- /* Unlock inodelk */
- lock_count = dht_lock_count (local->lock[0].ns.parent_layout.locks,
- local->lock[0].ns.parent_layout.lk_count);
- if (lock_count == 0)
- goto done;
+ /* Unlock inodelk */
+ lock_count = dht_lock_count(local->lock[0].ns.parent_layout.locks,
+ local->lock[0].ns.parent_layout.lk_count);
+ if (lock_count == 0)
+ goto done;
- lock_frame = copy_frame (frame);
- if (lock_frame == NULL) {
- goto done;
- }
+ lock_frame = copy_frame(frame);
+ if (lock_frame == NULL) {
+ goto done;
+ }
- lock_local = dht_local_init (lock_frame, &local->loc, NULL,
- lock_frame->root->op);
- if (lock_local == NULL) {
- goto done;
- }
+ lock_local = dht_local_init(lock_frame, &local->loc, NULL,
+ lock_frame->root->op);
+ if (lock_local == NULL) {
+ goto done;
+ }
- lock_local->lock[0].ns.parent_layout.locks = local->lock[0].ns.parent_layout.locks;
- lock_local->lock[0].ns.parent_layout.lk_count = local->lock[0].ns.parent_layout.lk_count;
+ lock_local->lock[0].ns.parent_layout.locks = local->lock[0]
+ .ns.parent_layout.locks;
+ lock_local->lock[0]
+ .ns.parent_layout.lk_count = local->lock[0].ns.parent_layout.lk_count;
- local->lock[0].ns.parent_layout.locks = NULL;
- local->lock[0].ns.parent_layout.lk_count = 0;
+ local->lock[0].ns.parent_layout.locks = NULL;
+ local->lock[0].ns.parent_layout.lk_count = 0;
- dht_unlock_inodelk (lock_frame,
- lock_local->lock[0].ns.parent_layout.locks,
- lock_local->lock[0].ns.parent_layout.lk_count,
- dht_selfheal_unlock_cbk);
- lock_frame = NULL;
+ dht_unlock_inodelk(lock_frame, lock_local->lock[0].ns.parent_layout.locks,
+ lock_local->lock[0].ns.parent_layout.lk_count,
+ dht_selfheal_unlock_cbk);
+ lock_frame = NULL;
done:
- if (invoke_cbk)
- local->selfheal.dir_cbk (frame, NULL, frame->this, ret,
- local->op_errno, NULL);
- if (lock_frame != NULL) {
- DHT_STACK_DESTROY (lock_frame);
- }
+ if (invoke_cbk)
+ local->selfheal.dir_cbk(frame, NULL, frame->this, ret, local->op_errno,
+ NULL);
+ if (lock_frame != NULL) {
+ DHT_STACK_DESTROY(lock_frame);
+ }
- return 0;
+ return 0;
}
int
-dht_refresh_layout_done (call_frame_t *frame)
+dht_refresh_layout_done(call_frame_t *frame)
{
- int ret = -1;
- dht_layout_t *refreshed = NULL, *heal = NULL;
- dht_local_t *local = NULL;
- dht_need_heal_t should_heal = NULL;
- dht_selfheal_layout_t healer = NULL;
+ int ret = -1;
+ dht_layout_t *refreshed = NULL, *heal = NULL;
+ dht_local_t *local = NULL;
+ dht_need_heal_t should_heal = NULL;
+ dht_selfheal_layout_t healer = NULL;
- local = frame->local;
+ local = frame->local;
- refreshed = local->selfheal.refreshed_layout;
- heal = local->selfheal.layout;
+ refreshed = local->selfheal.refreshed_layout;
+ heal = local->selfheal.layout;
- healer = local->selfheal.healer;
- should_heal = local->selfheal.should_heal;
+ healer = local->selfheal.healer;
+ should_heal = local->selfheal.should_heal;
- ret = dht_layout_sort (refreshed);
- if (ret == -1) {
- gf_msg (frame->this->name, GF_LOG_WARNING, 0,
- DHT_MSG_LAYOUT_SORT_FAILED,
- "sorting the layout failed");
- goto err;
- }
+ ret = dht_layout_sort(refreshed);
+ if (ret == -1) {
+ gf_smsg(frame->this->name, GF_LOG_WARNING, 0,
+ DHT_MSG_LAYOUT_SORT_FAILED, NULL);
+ goto err;
+ }
- if (should_heal (frame, &heal, &refreshed)) {
- healer (frame, &local->loc, heal);
- } else {
- local->selfheal.layout = NULL;
- local->selfheal.refreshed_layout = NULL;
- local->selfheal.layout = refreshed;
+ if (should_heal(frame, &heal, &refreshed)) {
+ healer(frame, &local->loc, heal);
+ } else {
+ local->selfheal.layout = NULL;
+ local->selfheal.refreshed_layout = NULL;
+ local->selfheal.layout = refreshed;
- dht_layout_unref (frame->this, heal);
+ dht_layout_unref(frame->this, heal);
- dht_selfheal_dir_finish (frame, frame->this, 0, 1);
- }
+ dht_selfheal_dir_finish(frame, frame->this, 0, 1);
+ }
- return 0;
+ return 0;
err:
- dht_selfheal_dir_finish (frame, frame->this, -1, 1);
- return 0;
+ dht_selfheal_dir_finish(frame, frame->this, -1, 1);
+ return 0;
}
int
-dht_refresh_layout_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode,
- struct iatt *stbuf, dict_t *xattr,
- struct iatt *postparent)
+dht_refresh_layout_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *stbuf, dict_t *xattr,
+ struct iatt *postparent)
{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- xlator_t *prev = NULL;
- dht_layout_t *layout = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0,};
-
- GF_VALIDATE_OR_GOTO ("dht", frame, err);
- GF_VALIDATE_OR_GOTO ("dht", this, err);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, err);
- GF_VALIDATE_OR_GOTO ("dht", this->private, err);
-
- local = frame->local;
- prev = cookie;
-
- layout = local->selfheal.refreshed_layout;
-
- LOCK (&frame->lock);
- {
- op_ret = dht_layout_merge (this, layout, prev,
- op_ret, op_errno, xattr);
-
- dht_iatt_merge (this, &local->stbuf, stbuf, prev);
-
- if (op_ret == -1) {
- gf_uuid_unparse (local->loc.gfid, gfid);
- local->op_errno = op_errno;
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- DHT_MSG_FILE_LOOKUP_FAILED,
- "lookup of %s on %s returned error, gfid: %s",
- local->loc.path, prev->name, gfid);
-
- goto unlock;
- }
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
+ xlator_t *prev = NULL;
+ dht_layout_t *layout = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("dht", frame, err);
+ GF_VALIDATE_OR_GOTO("dht", this, err);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, err);
+ GF_VALIDATE_OR_GOTO("dht", this->private, err);
+
+ local = frame->local;
+ prev = cookie;
+
+ layout = local->selfheal.refreshed_layout;
- local->op_ret = 0;
+ LOCK(&frame->lock);
+ {
+ op_ret = dht_layout_merge(this, layout, prev, op_ret, op_errno, xattr);
+
+ dht_iatt_merge(this, &local->stbuf, stbuf);
+
+ if (op_ret == -1) {
+ gf_uuid_unparse(local->loc.gfid, gfid);
+ local->op_errno = op_errno;
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno,
+ DHT_MSG_FILE_LOOKUP_FAILED, "path=%s", local->loc.path,
+ "name=%s", prev->name, "gfid=%s", gfid, NULL);
+
+ goto unlock;
}
-unlock:
- UNLOCK (&frame->lock);
- this_call_cnt = dht_frame_return (frame);
+ local->op_ret = 0;
+ }
+unlock:
+ UNLOCK(&frame->lock);
- if (is_last_call (this_call_cnt)) {
- if (local->op_ret == 0) {
- local->refresh_layout_done (frame);
- } else {
- goto err;
- }
+ this_call_cnt = dht_frame_return(frame);
+ if (is_last_call(this_call_cnt)) {
+ if (local->op_ret == 0) {
+ local->refresh_layout_done(frame);
+ } else {
+ goto err;
}
+ }
- return 0;
+ return 0;
err:
- if (local) {
- local->refresh_layout_unlock (frame, this, -1, 1);
- }
- return 0;
+ if (local) {
+ local->refresh_layout_unlock(frame, this, -1, 1);
+ }
+ return 0;
}
int
-dht_refresh_layout (call_frame_t *frame)
+dht_refresh_layout(call_frame_t *frame)
{
- int call_cnt = 0;
- int i = 0, ret = -1;
- dht_conf_t *conf = NULL;
- dht_local_t *local = NULL;
- xlator_t *this = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0,};
-
- GF_VALIDATE_OR_GOTO ("dht", frame, out);
- GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
-
- this = frame->this;
- conf = this->private;
- local = frame->local;
-
- call_cnt = conf->subvolume_cnt;
- local->call_cnt = call_cnt;
- local->op_ret = -1;
-
- if (local->selfheal.refreshed_layout) {
- dht_layout_unref (this, local->selfheal.refreshed_layout);
- local->selfheal.refreshed_layout = NULL;
- }
-
- local->selfheal.refreshed_layout = dht_layout_new (this,
- conf->subvolume_cnt);
- if (!local->selfheal.refreshed_layout) {
- gf_uuid_unparse (local->loc.gfid, gfid);
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
- "mem allocation for layout failed, path:%s gfid:%s",
- local->loc.path, gfid);
- goto out;
- }
+ int call_cnt = 0;
+ int i = 0, ret = -1;
+ dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *this = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("dht", frame, out);
+ GF_VALIDATE_OR_GOTO("dht", frame->local, out);
+
+ this = frame->this;
+ conf = this->private;
+ local = frame->local;
+
+ call_cnt = conf->subvolume_cnt;
+ local->call_cnt = call_cnt;
+ local->op_ret = -1;
+
+ if (local->selfheal.refreshed_layout) {
+ dht_layout_unref(this, local->selfheal.refreshed_layout);
+ local->selfheal.refreshed_layout = NULL;
+ }
+
+ local->selfheal.refreshed_layout = dht_layout_new(this,
+ conf->subvolume_cnt);
+ if (!local->selfheal.refreshed_layout) {
+ gf_uuid_unparse(local->loc.gfid, gfid);
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_MEM_ALLOC_FAILED,
+ "path=%s", local->loc.path, "gfid=%s", gfid, NULL);
+ goto out;
+ }
- if (local->xattr != NULL) {
- dict_del (local->xattr, conf->xattr_name);
- }
+ if (local->xattr != NULL) {
+ dict_del(local->xattr, conf->xattr_name);
+ }
+ if (local->xattr_req == NULL) {
+ gf_uuid_unparse(local->loc.gfid, gfid);
+ local->xattr_req = dict_new();
if (local->xattr_req == NULL) {
- gf_uuid_unparse (local->loc.gfid, gfid);
- local->xattr_req = dict_new ();
- if (local->xattr_req == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_NO_MEMORY,
- "dict mem allocation failed, path:%s gfid:%s",
- local->loc.path, gfid);
- goto out;
- }
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
+ "path=%s", local->loc.path, "gfid=%s", gfid, NULL);
+ goto out;
}
+ }
- if (dict_get (local->xattr_req, conf->xattr_name) == 0) {
- ret = dict_set_uint32 (local->xattr_req, conf->xattr_name,
- 4 * 4);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "%s: Failed to set dictionary value:key = %s",
- local->loc.path, conf->xattr_name);
- }
+ if (dict_get(local->xattr_req, conf->xattr_name) == 0) {
+ ret = dict_set_uint32(local->xattr_req, conf->xattr_name, 4 * 4);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_DICT_SET_FAILED,
+ "path=%s", local->loc.path, "key=%s", conf->xattr_name,
+ NULL);
+ }
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND_COOKIE (frame, dht_refresh_layout_cbk,
- conf->subvolumes[i], conf->subvolumes[i],
- conf->subvolumes[i]->fops->lookup,
- &local->loc, local->xattr_req);
- }
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND_COOKIE(frame, dht_refresh_layout_cbk, conf->subvolumes[i],
+ conf->subvolumes[i],
+ conf->subvolumes[i]->fops->lookup, &local->loc,
+ local->xattr_req);
+ }
- return 0;
+ return 0;
out:
- if (local) {
- local->refresh_layout_unlock (frame, this, -1, 1);
- }
- return 0;
+ if (local) {
+ local->refresh_layout_unlock(frame, this, -1, 1);
+ }
+ return 0;
}
-
int32_t
-dht_selfheal_layout_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+dht_selfheal_layout_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
+ dht_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (!local) {
- goto err;
- }
+ if (!local) {
+ goto err;
+ }
- if (op_ret < 0) {
- local->op_errno = op_errno;
- goto err;
- }
+ if (op_ret < 0) {
+ local->op_errno = op_errno;
+ goto err;
+ }
- local->refresh_layout_unlock = dht_selfheal_dir_finish;
- local->refresh_layout_done = dht_refresh_layout_done;
+ local->refresh_layout_unlock = dht_selfheal_dir_finish;
+ local->refresh_layout_done = dht_refresh_layout_done;
- dht_refresh_layout (frame);
- return 0;
+ dht_refresh_layout(frame);
+ return 0;
err:
- dht_selfheal_dir_finish (frame, this, -1, 1);
- return 0;
+ dht_selfheal_dir_finish(frame, this, -1, 1);
+ return 0;
}
-
gf_boolean_t
-dht_should_heal_layout (call_frame_t *frame, dht_layout_t **heal,
- dht_layout_t **ondisk)
+dht_should_heal_layout(call_frame_t *frame, dht_layout_t **heal,
+ dht_layout_t **ondisk)
{
- gf_boolean_t fixit = _gf_true;
- dht_local_t *local = NULL;
- int ret = -1, heal_missing_dirs = 0;
-
- local = frame->local;
-
- if ((heal == NULL) || (*heal == NULL) || (ondisk == NULL)
- || (*ondisk == NULL))
- goto out;
-
- ret = dht_layout_anomalies (frame->this, &local->loc, *ondisk,
- &local->selfheal.hole_cnt,
- &local->selfheal.overlaps_cnt,
- NULL, &local->selfheal.down,
- &local->selfheal.misc, NULL);
-
- if (ret < 0)
- goto out;
-
- /* Directories might've been created as part of this self-heal. We've to
- * sync non-layout xattrs and set range 0-0 on new directories
+ gf_boolean_t fixit = _gf_true;
+ dht_local_t *local = NULL;
+ int heal_missing_dirs = 0;
+
+ local = frame->local;
+
+ if ((heal == NULL) || (*heal == NULL) || (ondisk == NULL) ||
+ (*ondisk == NULL))
+ goto out;
+
+ dht_layout_anomalies(
+ frame->this, &local->loc, *ondisk, &local->selfheal.hole_cnt,
+ &local->selfheal.overlaps_cnt, &local->selfheal.missing_cnt,
+ &local->selfheal.down, &local->selfheal.misc, NULL);
+
+ /* Directories might've been created as part of this self-heal. We've to
+ * sync non-layout xattrs and set range 0-0 on new directories
+ */
+ heal_missing_dirs = local->selfheal.force_mkdir
+ ? local->selfheal.force_mkdir
+ : dht_layout_missing_dirs(*heal);
+
+ if ((local->selfheal.hole_cnt == 0) &&
+ (local->selfheal.overlaps_cnt == 0) && heal_missing_dirs) {
+ dht_layout_t *tmp = NULL;
+
+ /* Just added a brick and need to set 0-0 range on this brick.
+ * But ondisk layout is well-formed. So, swap layouts "heal" and
+ * "ondisk". Now "ondisk" layout will be used for healing
+ * xattrs. If there are any non-participating subvols in
+ * "ondisk" layout, dht_selfheal_dir_xattr_persubvol will set
+ * 0-0 and non-layout xattrs. This way we won't end up in
+ * "corrupting" already set and well-formed "ondisk" layout.
*/
- heal_missing_dirs = local->selfheal.force_mkdir
- ? local->selfheal.force_mkdir : dht_layout_missing_dirs (*heal);
-
- if ((local->selfheal.hole_cnt == 0)
- && (local->selfheal.overlaps_cnt == 0) && heal_missing_dirs) {
- dht_layout_t *tmp = NULL;
-
- /* Just added a brick and need to set 0-0 range on this brick.
- * But ondisk layout is well-formed. So, swap layouts "heal" and
- * "ondisk". Now "ondisk" layout will be used for healing
- * xattrs. If there are any non-participating subvols in
- * "ondisk" layout, dht_selfheal_dir_xattr_persubvol will set
- * 0-0 and non-layout xattrs. This way we won't end up in
- * "corrupting" already set and well-formed "ondisk" layout.
- */
- tmp = *heal;
- *heal = *ondisk;
- *ondisk = tmp;
-
- /* Current selfheal code, heals non-layout xattrs only after
- * an add-brick. In fact non-layout xattrs are considered as
- * secondary citizens which are healed only if layout xattrs
- * need to be healed. This is wrong, since for eg., quota can be
- * set when layout is well-formed, but a node is down. Also,
- * just for healing non-layout xattrs, we don't need locking.
- * This issue is _NOT FIXED_ by this patch.
- */
- }
+ tmp = *heal;
+ *heal = *ondisk;
+ *ondisk = tmp;
+
+ /* Current selfheal code, heals non-layout xattrs only after
+ * an add-brick. In fact non-layout xattrs are considered as
+ * secondary citizens which are healed only if layout xattrs
+ * need to be healed. This is wrong, since for eg., quota can be
+ * set when layout is well-formed, but a node is down. Also,
+ * just for healing non-layout xattrs, we don't need locking.
+ * This issue is _NOT FIXED_ by this patch.
+ */
+ }
- fixit = (local->selfheal.hole_cnt || local->selfheal.overlaps_cnt
- || heal_missing_dirs);
+ fixit = (local->selfheal.hole_cnt || local->selfheal.overlaps_cnt ||
+ heal_missing_dirs);
out:
- return fixit;
+ return fixit;
}
int
-dht_layout_span (dht_layout_t *layout)
+dht_layout_span(dht_layout_t *layout)
{
- int i = 0, count = 0;
+ int i = 0, count = 0;
- for (i = 0; i < layout->cnt; i++) {
- if (layout->list[i].err)
- continue;
+ for (i = 0; i < layout->cnt; i++) {
+ if (layout->list[i].err)
+ continue;
- if (layout->list[i].start != layout->list[i].stop)
- count++;
- }
+ if (layout->list[i].start != layout->list[i].stop)
+ count++;
+ }
- return count;
+ return count;
}
int
-dht_decommissioned_bricks_in_layout (xlator_t *this, dht_layout_t *layout)
+dht_decommissioned_bricks_in_layout(xlator_t *this, dht_layout_t *layout)
{
- dht_conf_t *conf = NULL;
- int count = 0, i = 0, j = 0;
+ dht_conf_t *conf = NULL;
+ int count = 0, i = 0, j = 0;
- if ((this == NULL) || (layout == NULL))
- goto out;
+ if ((this == NULL) || (layout == NULL))
+ goto out;
- conf = this->private;
+ conf = this->private;
- for (i = 0; i < layout->cnt; i++) {
- for (j = 0; j < conf->subvolume_cnt; j++) {
- if (conf->decommissioned_bricks[j] &&
- conf->decommissioned_bricks[j]
- == layout->list[i].xlator) {
- count++;
- }
- }
+ for (i = 0; i < layout->cnt; i++) {
+ for (j = 0; j < conf->subvolume_cnt; j++) {
+ if (conf->decommissioned_bricks[j] &&
+ conf->decommissioned_bricks[j] == layout->list[i].xlator) {
+ count++;
+ }
}
+ }
out:
- return count;
+ return count;
}
dht_distribution_type_t
-dht_distribution_type (xlator_t *this, dht_layout_t *layout)
+dht_distribution_type(xlator_t *this, dht_layout_t *layout)
{
- dht_distribution_type_t type = GF_DHT_EQUAL_DISTRIBUTION;
- int i = 0;
- uint32_t start_range = 0, range = 0, diff = 0;
+ dht_distribution_type_t type = GF_DHT_EQUAL_DISTRIBUTION;
+ int i = 0;
+ uint32_t start_range = 0, range = 0, diff = 0;
- if ((this == NULL) || (layout == NULL) || (layout->cnt < 1)) {
- goto out;
- }
+ if ((this == NULL) || (layout == NULL) || (layout->cnt < 1)) {
+ goto out;
+ }
- for (i = 0; i < layout->cnt; i++) {
- if (start_range == 0) {
- start_range = layout->list[i].stop
- - layout->list[i].start;
- continue;
- }
+ for (i = 0; i < layout->cnt; i++) {
+ if (start_range == 0) {
+ start_range = layout->list[i].stop - layout->list[i].start;
+ continue;
+ }
- range = layout->list[i].stop - layout->list[i].start;
- diff = (range >= start_range)
- ? range - start_range
- : start_range - range;
+ range = layout->list[i].stop - layout->list[i].start;
+ diff = (range >= start_range) ? range - start_range
+ : start_range - range;
- if ((range != 0) && (diff > layout->cnt)) {
- type = GF_DHT_WEIGHTED_DISTRIBUTION;
- break;
- }
+ if ((range != 0) && (diff > layout->cnt)) {
+ type = GF_DHT_WEIGHTED_DISTRIBUTION;
+ break;
}
+ }
out:
- return type;
+ return type;
}
gf_boolean_t
-dht_should_fix_layout (call_frame_t *frame, dht_layout_t **inmem,
- dht_layout_t **ondisk)
+dht_should_fix_layout(call_frame_t *frame, dht_layout_t **inmem,
+ dht_layout_t **ondisk)
{
- gf_boolean_t fixit = _gf_true;
+ gf_boolean_t fixit = _gf_true;
- dht_local_t *local = NULL;
- int layout_span = 0;
- int decommissioned_bricks = 0;
- int ret = 0;
- dht_conf_t *conf = NULL;
- dht_distribution_type_t inmem_dist_type = 0;
- dht_distribution_type_t ondisk_dist_type = 0;
+ dht_local_t *local = NULL;
+ int layout_span = 0;
+ int decommissioned_bricks = 0;
+ dht_conf_t *conf = NULL;
+ dht_distribution_type_t inmem_dist_type = 0;
+ dht_distribution_type_t ondisk_dist_type = 0;
- conf = frame->this->private;
+ conf = frame->this->private;
- local = frame->local;
+ local = frame->local;
- if ((inmem == NULL) || (*inmem == NULL) || (ondisk == NULL)
- || (*ondisk == NULL))
- goto out;
+ if ((inmem == NULL) || (*inmem == NULL) || (ondisk == NULL) ||
+ (*ondisk == NULL))
+ goto out;
- ret = dht_layout_anomalies (frame->this, &local->loc, *ondisk,
- &local->selfheal.hole_cnt,
- &local->selfheal.overlaps_cnt, NULL,
- &local->selfheal.down,
- &local->selfheal.misc, NULL);
- if (ret < 0) {
- fixit = _gf_false;
- goto out;
- }
+ dht_layout_anomalies(frame->this, &local->loc, *ondisk,
+ &local->selfheal.hole_cnt,
+ &local->selfheal.overlaps_cnt, NULL,
+ &local->selfheal.down, &local->selfheal.misc, NULL);
- if (local->selfheal.down || local->selfheal.misc) {
- fixit = _gf_false;
- goto out;
- }
+ if (local->selfheal.down || local->selfheal.misc) {
+ fixit = _gf_false;
+ goto out;
+ }
- if (local->selfheal.hole_cnt || local->selfheal.overlaps_cnt)
- goto out;
+ if (local->selfheal.hole_cnt || local->selfheal.overlaps_cnt)
+ goto out;
- /* If commit hashes are being updated, let it through */
- if ((*inmem)->commit_hash != (*ondisk)->commit_hash)
- goto out;
+ /* If commit hashes are being updated, let it through */
+ if ((*inmem)->commit_hash != (*ondisk)->commit_hash)
+ goto out;
- layout_span = dht_layout_span (*ondisk);
+ layout_span = dht_layout_span(*ondisk);
- decommissioned_bricks
- = dht_decommissioned_bricks_in_layout (frame->this,
- *ondisk);
- inmem_dist_type = dht_distribution_type (frame->this, *inmem);
- ondisk_dist_type = dht_distribution_type (frame->this, *ondisk);
+ decommissioned_bricks = dht_decommissioned_bricks_in_layout(frame->this,
+ *ondisk);
+ inmem_dist_type = dht_distribution_type(frame->this, *inmem);
+ ondisk_dist_type = dht_distribution_type(frame->this, *ondisk);
- if ((decommissioned_bricks == 0)
- && (layout_span == (conf->subvolume_cnt
- - conf->decommission_subvols_cnt))
- && (inmem_dist_type == ondisk_dist_type))
- fixit = _gf_false;
+ if ((decommissioned_bricks == 0) &&
+ (layout_span ==
+ (conf->subvolume_cnt - conf->decommission_subvols_cnt)) &&
+ (inmem_dist_type == ondisk_dist_type))
+ fixit = _gf_false;
out:
- return fixit;
+ return fixit;
}
-int
-dht_selfheal_layout_lock (call_frame_t *frame, dht_layout_t *layout,
- gf_boolean_t newdir,
- dht_selfheal_layout_t healer,
- dht_need_heal_t should_heal)
+static int
+dht_selfheal_layout_lock(call_frame_t *frame, dht_layout_t *layout,
+ gf_boolean_t newdir, dht_selfheal_layout_t healer,
+ dht_need_heal_t should_heal)
{
- dht_local_t *local = NULL;
- int count = 1, ret = -1, i = 0;
- dht_lock_t **lk_array = NULL;
- dht_conf_t *conf = NULL;
- dht_layout_t *tmp = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- GF_VALIDATE_OR_GOTO ("dht", frame, err);
- GF_VALIDATE_OR_GOTO (frame->this->name, frame->local, err);
-
- local = frame->local;
-
- conf = frame->this->private;
-
- local->selfheal.healer = healer;
- local->selfheal.should_heal = should_heal;
-
- tmp = local->selfheal.layout;
- local->selfheal.layout = dht_layout_ref (frame->this, layout);
- dht_layout_unref (frame->this, tmp);
-
- if (!newdir) {
- count = conf->subvolume_cnt;
-
- lk_array = GF_CALLOC (count, sizeof (*lk_array),
- gf_common_mt_char);
- if (lk_array == NULL) {
- gf_uuid_unparse (local->stbuf.ia_gfid, gfid);
- gf_msg ("dht", GF_LOG_ERROR, ENOMEM,
- DHT_MSG_NO_MEMORY, "mem allocation failed for "
- "lk_array, gfid:%s path: %s", gfid,
- local->loc.path);
- goto err;
- }
+ dht_local_t *local = NULL;
+ int count = 1, ret = -1, i = 0;
+ dht_lock_t **lk_array = NULL;
+ dht_conf_t *conf = NULL;
+ dht_layout_t *tmp = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- for (i = 0; i < count; i++) {
- lk_array[i] = dht_lock_new (frame->this,
- conf->subvolumes[i],
- &local->loc, F_WRLCK,
- DHT_LAYOUT_HEAL_DOMAIN,
- NULL,
- FAIL_ON_ANY_ERROR);
- if (lk_array[i] == NULL) {
- gf_uuid_unparse (local->stbuf.ia_gfid, gfid);
- gf_msg (THIS->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_NO_MEMORY, "mem allocation "
- "failed for lk_array, gfid:%s path:%s",
- gfid, local->loc.path);
- goto err;
- }
- }
- } else {
- count = 1;
- lk_array = GF_CALLOC (count, sizeof (*lk_array),
- gf_common_mt_char);
- if (lk_array == NULL) {
- gf_uuid_unparse (local->stbuf.ia_gfid, gfid);
- gf_msg (THIS->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_NO_MEMORY, "mem allocation failed for "
- "lk_array, gfid:%s path:%s",
- gfid, local->loc.path);
- goto err;
- }
+ GF_VALIDATE_OR_GOTO("dht", frame, err);
+ GF_VALIDATE_OR_GOTO(frame->this->name, frame->local, err);
- lk_array[0] = dht_lock_new (frame->this, local->hashed_subvol,
- &local->loc, F_WRLCK,
- DHT_LAYOUT_HEAL_DOMAIN, NULL,
- FAIL_ON_ANY_ERROR);
- if (lk_array[0] == NULL) {
- gf_uuid_unparse (local->stbuf.ia_gfid, gfid);
- gf_msg (THIS->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_NO_MEMORY, "mem allocation failed for "
- "lk_array, gfid:%s path:%s", gfid,
- local->loc.path);
- goto err;
- }
- }
+ local = frame->local;
- local->lock[0].layout.my_layout.locks = lk_array;
- local->lock[0].layout.my_layout.lk_count = count;
+ conf = frame->this->private;
- ret = dht_blocking_inodelk (frame, lk_array, count,
- dht_selfheal_layout_lock_cbk);
- if (ret < 0) {
- local->lock[0].layout.my_layout.locks = NULL;
- local->lock[0].layout.my_layout.lk_count = 0;
- goto err;
+ local->selfheal.healer = healer;
+ local->selfheal.should_heal = should_heal;
+
+ tmp = local->selfheal.layout;
+ local->selfheal.layout = dht_layout_ref(frame->this, layout);
+ dht_layout_unref(frame->this, tmp);
+
+ if (!newdir) {
+ count = conf->subvolume_cnt;
+
+ lk_array = GF_CALLOC(count, sizeof(*lk_array), gf_common_mt_char);
+ if (lk_array == NULL) {
+ gf_uuid_unparse(local->stbuf.ia_gfid, gfid);
+ gf_smsg("dht", GF_LOG_ERROR, ENOMEM, DHT_MSG_MEM_ALLOC_FAILED,
+ "lk_array-gfid=%s", gfid, "path=%s", local->loc.path, NULL);
+ goto err;
}
- return 0;
+ for (i = 0; i < count; i++) {
+ lk_array[i] = dht_lock_new(
+ frame->this, conf->subvolumes[i], &local->loc, F_WRLCK,
+ DHT_LAYOUT_HEAL_DOMAIN, NULL, FAIL_ON_ANY_ERROR);
+ if (lk_array[i] == NULL) {
+ gf_uuid_unparse(local->stbuf.ia_gfid, gfid);
+ gf_smsg(THIS->name, GF_LOG_ERROR, ENOMEM,
+ DHT_MSG_MEM_ALLOC_FAILED, "lk_array-gfid=%s", gfid,
+ "path=%s", local->loc.path, NULL);
+ goto err;
+ }
+ }
+ } else {
+ count = 1;
+ lk_array = GF_CALLOC(count, sizeof(*lk_array), gf_common_mt_char);
+ if (lk_array == NULL) {
+ gf_uuid_unparse(local->stbuf.ia_gfid, gfid);
+ gf_smsg(THIS->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_MEM_ALLOC_FAILED,
+ "lk_array-gfid=%s", gfid, "path=%s", local->loc.path, NULL);
+ goto err;
+ }
+
+ lk_array[0] = dht_lock_new(frame->this, local->hashed_subvol,
+ &local->loc, F_WRLCK, DHT_LAYOUT_HEAL_DOMAIN,
+ NULL, FAIL_ON_ANY_ERROR);
+ if (lk_array[0] == NULL) {
+ gf_uuid_unparse(local->stbuf.ia_gfid, gfid);
+ gf_smsg(THIS->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_MEM_ALLOC_FAILED,
+ "lk_array-gfid=%s", gfid, "path=%s", local->loc.path, NULL);
+ goto err;
+ }
+ }
+
+ local->lock[0].layout.my_layout.locks = lk_array;
+ local->lock[0].layout.my_layout.lk_count = count;
+
+ ret = dht_blocking_inodelk(frame, lk_array, count,
+ dht_selfheal_layout_lock_cbk);
+ if (ret < 0) {
+ local->lock[0].layout.my_layout.locks = NULL;
+ local->lock[0].layout.my_layout.lk_count = 0;
+ goto err;
+ }
+
+ return 0;
err:
- if (lk_array != NULL) {
- dht_lock_array_free (lk_array, count);
- GF_FREE (lk_array);
- }
+ if (lk_array != NULL) {
+ dht_lock_array_free(lk_array, count);
+ GF_FREE(lk_array);
+ }
- return -1;
+ return -1;
}
-int
-dht_selfheal_dir_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+static int
+dht_selfheal_dir_xattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- xlator_t *subvol = NULL;
- struct iatt *stbuf = NULL;
- int i = 0;
- int ret = 0;
- dht_layout_t *layout = NULL;
- int err = 0;
- int this_call_cnt = 0;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- local = frame->local;
- layout = local->selfheal.layout;
- subvol = cookie;
-
- if (op_ret == 0) {
- err = 0;
- } else {
- gf_uuid_unparse (local->loc.gfid, gfid);
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- DHT_MSG_DIR_SELFHEAL_XATTR_FAILED,
- "layout setxattr failed on %s, path:%s gfid:%s",
- subvol->name, local->loc.path, gfid);
- err = op_errno;
- }
-
- ret = dict_get_bin (xdata, DHT_IATT_IN_XDATA_KEY, (void **) &stbuf);
- if (ret < 0) {
- gf_uuid_unparse (local->loc.gfid, gfid);
- gf_msg_debug (this->name, 0, "key = %s not present in dict"
- ", path:%s gfid:%s", DHT_IATT_IN_XDATA_KEY,
- local->loc.path, gfid);
- }
+ dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+ struct iatt *stbuf = NULL;
+ int i = 0;
+ int ret = 0;
+ dht_layout_t *layout = NULL;
+ int err = 0;
+ int this_call_cnt = 0;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+
+ local = frame->local;
+ layout = local->selfheal.layout;
+ subvol = cookie;
+
+ if (op_ret == 0) {
+ err = 0;
+ } else {
+ gf_uuid_unparse(local->loc.gfid, gfid);
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno,
+ DHT_MSG_DIR_SELFHEAL_XATTR_FAILED, "name=%s", subvol->name,
+ "path=%s", local->loc.path, "gfid=%s", gfid, NULL);
+ err = op_errno;
+ }
+
+ ret = dict_get_bin(xdata, DHT_IATT_IN_XDATA_KEY, (void **)&stbuf);
+ if (ret < 0) {
+ gf_uuid_unparse(local->loc.gfid, gfid);
+ gf_msg_debug(this->name, 0,
+ "key = %s not present in dict"
+ ", path:%s gfid:%s",
+ DHT_IATT_IN_XDATA_KEY, local->loc.path, gfid);
+ }
- for (i = 0; i < layout->cnt; i++) {
- if (layout->list[i].xlator == subvol) {
- layout->list[i].err = err;
- break;
- }
+ for (i = 0; i < layout->cnt; i++) {
+ if (layout->list[i].xlator == subvol) {
+ layout->list[i].err = err;
+ break;
}
+ }
- LOCK (&frame->lock);
- {
- dht_iatt_merge (this, &local->stbuf, stbuf, subvol);
- }
- UNLOCK (&frame->lock);
+ LOCK(&frame->lock);
+ {
+ dht_iatt_merge(this, &local->stbuf, stbuf);
+ }
+ UNLOCK(&frame->lock);
- this_call_cnt = dht_frame_return (frame);
+ this_call_cnt = dht_frame_return(frame);
- if (is_last_call (this_call_cnt)) {
- dht_selfheal_dir_finish (frame, this, 0, 1);
- }
+ if (is_last_call(this_call_cnt)) {
+ dht_selfheal_dir_finish(frame, this, 0, 1);
+ }
- return 0;
+ return 0;
}
/* Code is required to set user xattr to local->xattr
-*/
+ */
int
-dht_set_user_xattr (dict_t *dict, char *k, data_t *v, void *data)
+dht_set_user_xattr(dict_t *dict, char *k, data_t *v, void *data)
{
- dict_t *set_xattr = data;
- int ret = -1;
+ dict_t *set_xattr = data;
+ int ret = -1;
- ret = dict_set (set_xattr, k, v);
- return ret;
+ ret = dict_set(set_xattr, k, v);
+ return ret;
}
-
-int
-dht_selfheal_dir_xattr_persubvol (call_frame_t *frame, loc_t *loc,
- dht_layout_t *layout, int i,
- xlator_t *req_subvol)
+static int
+dht_selfheal_dir_xattr_persubvol(call_frame_t *frame, loc_t *loc,
+ dht_layout_t *layout, int i,
+ xlator_t *req_subvol)
{
- xlator_t *subvol = NULL;
- dict_t *xattr = NULL;
- dict_t *xdata = NULL;
- int ret = 0;
- xlator_t *this = NULL;
- int32_t *disk_layout = NULL;
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- data_t *data = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- local = frame->local;
- if (req_subvol)
- subvol = req_subvol;
- else
- subvol = layout->list[i].xlator;
- this = frame->this;
-
- GF_VALIDATE_OR_GOTO ("", this, err);
- GF_VALIDATE_OR_GOTO (this->name, layout, err);
- GF_VALIDATE_OR_GOTO (this->name, local, err);
- GF_VALIDATE_OR_GOTO (this->name, subvol, err);
- VALIDATE_OR_GOTO (this->private, err);
-
- conf = this->private;
-
- xattr = dict_new ();
- if (!xattr) {
- goto err;
- }
-
- xdata = dict_new ();
- if (!xdata)
- goto err;
-
- ret = dict_set_str (xdata, GLUSTERFS_INTERNAL_FOP_KEY, "yes");
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, 0, DHT_MSG_DICT_SET_FAILED,
- "%s: Failed to set dictionary value: key = %s,"
- " gfid = %s", loc->path,
- GLUSTERFS_INTERNAL_FOP_KEY, gfid);
- goto err;
- }
-
- ret = dict_set_int8 (xdata, DHT_IATT_IN_XDATA_KEY, 1);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, 0, DHT_MSG_DICT_SET_FAILED,
- "%s: Failed to set dictionary value: key = %s,"
- " gfid = %s", loc->path,
- DHT_IATT_IN_XDATA_KEY, gfid);
- goto err;
- }
-
- gf_uuid_unparse(loc->inode->gfid, gfid);
-
- ret = dht_disk_layout_extract (this, layout, i, &disk_layout);
- if (ret == -1) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DIR_SELFHEAL_XATTR_FAILED,
- "Directory self heal xattr failed:"
- " %s: (subvol %s) Failed to extract disk layout,"
- " gfid = %s", loc->path, subvol->name, gfid);
- goto err;
- }
-
- ret = dict_set_bin (xattr, conf->xattr_name, disk_layout, 4 * 4);
- if (ret == -1) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DIR_SELFHEAL_XATTR_FAILED,
- "Directory self heal xattr failed:"
- "%s: (subvol %s) Failed to set xattr dictionary,"
- " gfid = %s", loc->path, subvol->name, gfid);
- goto err;
- }
- disk_layout = NULL;
-
- gf_msg_trace (this->name, 0,
- "setting hash range %u - %u (type %d) on subvolume %s"
- " for %s", layout->list[i].start, layout->list[i].stop,
- layout->type, subvol->name, loc->path);
-
- if (local->xattr) {
- data = dict_get (local->xattr, QUOTA_LIMIT_KEY);
- if (data) {
- ret = dict_add (xattr, QUOTA_LIMIT_KEY, data);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "%s: Failed to set dictionary value:"
- " key = %s",
- loc->path, QUOTA_LIMIT_KEY);
- }
- }
- data = dict_get (local->xattr, QUOTA_LIMIT_OBJECTS_KEY);
- if (data) {
- ret = dict_add (xattr, QUOTA_LIMIT_OBJECTS_KEY, data);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "%s: Failed to set dictionary value:"
- " key = %s",
- loc->path, QUOTA_LIMIT_OBJECTS_KEY);
- }
- }
- }
-
- if (!gf_uuid_is_null (local->gfid))
- gf_uuid_copy (loc->gfid, local->gfid);
-
- STACK_WIND_COOKIE (frame, dht_selfheal_dir_xattr_cbk,
- (void *) subvol, subvol, subvol->fops->setxattr,
- loc, xattr, 0, xdata);
-
- dict_unref (xattr);
- dict_unref (xdata);
-
- return 0;
+ xlator_t *subvol = NULL;
+ dict_t *xattr = NULL;
+ dict_t *xdata = NULL;
+ int ret = 0;
+ xlator_t *this = NULL;
+ int32_t *disk_layout = NULL;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ data_t *data = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+
+ local = frame->local;
+ if (req_subvol)
+ subvol = req_subvol;
+ else
+ subvol = layout->list[i].xlator;
+ this = frame->this;
+
+ GF_VALIDATE_OR_GOTO("", this, err);
+ GF_VALIDATE_OR_GOTO(this->name, layout, err);
+ GF_VALIDATE_OR_GOTO(this->name, local, err);
+ GF_VALIDATE_OR_GOTO(this->name, subvol, err);
+ VALIDATE_OR_GOTO(this->private, err);
+
+ conf = this->private;
+
+ xattr = dict_new();
+ if (!xattr) {
+ goto err;
+ }
+
+ xdata = dict_new();
+ if (!xdata)
+ goto err;
+
+ ret = dict_set_str(xdata, GLUSTERFS_INTERNAL_FOP_KEY, "yes");
+ if (ret < 0) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_DICT_SET_FAILED,
+ "path=%s", loc->path, "key=%s", GLUSTERFS_INTERNAL_FOP_KEY,
+ "gfid=%s", gfid, NULL);
+ goto err;
+ }
+
+ ret = dict_set_int8(xdata, DHT_IATT_IN_XDATA_KEY, 1);
+ if (ret < 0) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_DICT_SET_FAILED,
+ "path=%s", loc->path, "key=%s", DHT_IATT_IN_XDATA_KEY,
+ "gfid=%s", gfid, NULL);
+ goto err;
+ }
+
+ gf_uuid_unparse(loc->inode->gfid, gfid);
+
+ ret = dht_disk_layout_extract(this, layout, i, &disk_layout);
+ if (ret == -1) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0,
+ DHT_MSG_DIR_SELFHEAL_XATTR_FAILED,
+ "extract-disk-layout-failed, path=%s", loc->path, "subvol=%s",
+ subvol->name, "gfid=%s", gfid, NULL);
+ goto err;
+ }
+
+ ret = dict_set_bin(xattr, conf->xattr_name, disk_layout, 4 * 4);
+ if (ret == -1) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0,
+ DHT_MSG_DIR_SELFHEAL_XATTR_FAILED, "path=%s", loc->path,
+ "subvol=%s", subvol->name,
+ "set-xattr-dictionary-failed"
+ "gfid=%s",
+ gfid, NULL);
+ goto err;
+ }
+ disk_layout = NULL;
+
+ gf_msg_trace(this->name, 0,
+ "setting hash range 0x%x - 0x%x (type %d) on subvolume %s"
+ " for %s",
+ layout->list[i].start, layout->list[i].stop, layout->type,
+ subvol->name, loc->path);
+
+ if (local->xattr) {
+ data = dict_get(local->xattr, QUOTA_LIMIT_KEY);
+ if (data) {
+ ret = dict_add(xattr, QUOTA_LIMIT_KEY, data);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "path=%s", loc->path, "key=%s", QUOTA_LIMIT_KEY, NULL);
+ }
+ }
+ data = dict_get(local->xattr, QUOTA_LIMIT_OBJECTS_KEY);
+ if (data) {
+ ret = dict_add(xattr, QUOTA_LIMIT_OBJECTS_KEY, data);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "path=%s", loc->path, "key=%s", QUOTA_LIMIT_OBJECTS_KEY,
+ NULL);
+ }
+ }
+ }
+
+ if (!gf_uuid_is_null(local->gfid))
+ gf_uuid_copy(loc->gfid, local->gfid);
+
+ STACK_WIND_COOKIE(frame, dht_selfheal_dir_xattr_cbk, (void *)subvol, subvol,
+ subvol->fops->setxattr, loc, xattr, 0, xdata);
+
+ dict_unref(xattr);
+ dict_unref(xdata);
+
+ return 0;
err:
- if (xattr)
- dict_unref (xattr);
- if (xdata)
- dict_unref (xdata);
+ if (xattr)
+ dict_unref(xattr);
+ if (xdata)
+ dict_unref(xdata);
- GF_FREE (disk_layout);
+ GF_FREE(disk_layout);
- dht_selfheal_dir_xattr_cbk (frame, (void *) subvol, frame->this,
- -1, ENOMEM, NULL);
- return 0;
+ dht_selfheal_dir_xattr_cbk(frame, (void *)subvol, frame->this, -1, ENOMEM,
+ NULL);
+ return 0;
}
-int
-dht_fix_dir_xattr (call_frame_t *frame, loc_t *loc, dht_layout_t *layout)
+static int
+dht_fix_dir_xattr(call_frame_t *frame, loc_t *loc, dht_layout_t *layout)
{
- dht_local_t *local = NULL;
- int i = 0;
- int count = 0;
- xlator_t *this = NULL;
- dht_conf_t *conf = NULL;
- dht_layout_t *dummy = NULL;
-
- local = frame->local;
- this = frame->this;
- conf = this->private;
-
- gf_msg_debug (this->name, 0,
- "%s: Writing the new range for all subvolumes",
- loc->path);
-
- local->call_cnt = count = conf->subvolume_cnt;
-
- if (gf_log_get_loglevel () >= GF_LOG_DEBUG)
- dht_log_new_layout_for_dir_selfheal (this, loc, layout);
-
- for (i = 0; i < layout->cnt; i++) {
- dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i, NULL);
-
- if (--count == 0)
- goto out;
- }
- /* if we are here, subvolcount > layout_count. subvols-per-directory
- * option might be set here. We need to clear out layout from the
- * non-participating subvolumes, else it will result in overlaps */
- dummy = dht_layout_new (this, 1);
- if (!dummy)
- goto out;
- dummy->commit_hash = layout->commit_hash;
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (_gf_false ==
- dht_is_subvol_in_layout (layout, conf->subvolumes[i])) {
- dht_selfheal_dir_xattr_persubvol (frame, loc, dummy, 0,
- conf->subvolumes[i]);
- if (--count == 0)
- break;
- }
- }
-
- dht_layout_unref (this, dummy);
+ dht_local_t *local = NULL;
+ int i = 0;
+ int count = 0;
+ xlator_t *this = NULL;
+ dht_conf_t *conf = NULL;
+ dht_layout_t *dummy = NULL;
+
+ local = frame->local;
+ this = frame->this;
+ conf = this->private;
+
+ gf_msg_debug(this->name, 0, "%s: Writing the new range for all subvolumes",
+ loc->path);
+
+ local->call_cnt = count = conf->subvolume_cnt;
+
+ if (gf_log_get_loglevel() >= GF_LOG_DEBUG)
+ dht_log_new_layout_for_dir_selfheal(this, loc, layout);
+
+ for (i = 0; i < layout->cnt; i++) {
+ dht_selfheal_dir_xattr_persubvol(frame, loc, layout, i, NULL);
+
+ if (--count == 0)
+ goto out;
+ }
+ /* if we are here, subvolcount > layout_count. subvols-per-directory
+ * option might be set here. We need to clear out layout from the
+ * non-participating subvolumes, else it will result in overlaps */
+ dummy = dht_layout_new(this, 1);
+ if (!dummy)
+ goto out;
+ dummy->commit_hash = layout->commit_hash;
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (_gf_false == dht_is_subvol_in_layout(layout, conf->subvolumes[i])) {
+ dht_selfheal_dir_xattr_persubvol(frame, loc, dummy, 0,
+ conf->subvolumes[i]);
+ if (--count == 0)
+ break;
+ }
+ }
+
+ dht_layout_unref(this, dummy);
out:
- return 0;
+ return 0;
}
-int
-dht_selfheal_dir_xattr (call_frame_t *frame, loc_t *loc, dht_layout_t *layout)
+static int
+dht_selfheal_dir_xattr(call_frame_t *frame, loc_t *loc, dht_layout_t *layout)
{
- dht_local_t *local = NULL;
- int missing_xattr = 0;
- int i = 0;
- xlator_t *this = NULL;
- dht_conf_t *conf = NULL;
- dht_layout_t *dummy = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0,};
-
- local = frame->local;
- this = frame->this;
- conf = this->private;
-
- for (i = 0; i < layout->cnt; i++) {
- if (layout->list[i].err != -1 || !layout->list[i].stop) {
- /* err != -1 would mean xattr present on the directory
- * or the directory is non existent.
- * !layout->list[i].stop would mean layout absent
- */
-
- continue;
- }
- missing_xattr++;
- }
- /* Also account for subvolumes with no-layout. Used for zero'ing out
- * the layouts and for setting quota key's if present */
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (_gf_false ==
- dht_is_subvol_in_layout (layout, conf->subvolumes[i])) {
- missing_xattr++;
- }
- }
- gf_msg_trace (this->name, 0,
- "%d subvolumes missing xattr for %s",
- missing_xattr, loc->path);
-
- if (missing_xattr == 0) {
- dht_selfheal_dir_finish (frame, this, 0, 1);
- return 0;
- }
-
- local->call_cnt = missing_xattr;
-
- if (gf_log_get_loglevel () >= GF_LOG_DEBUG)
- dht_log_new_layout_for_dir_selfheal (this, loc, layout);
-
- for (i = 0; i < layout->cnt; i++) {
- if (layout->list[i].err != -1 || !layout->list[i].stop)
- continue;
-
- dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i, NULL);
-
- if (--missing_xattr == 0)
- break;
- }
- dummy = dht_layout_new (this, 1);
- if (!dummy) {
- gf_uuid_unparse (loc->gfid, gfid);
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
- "failed to allocate dummy layout, path:%s gfid:%s",
- loc->path, gfid);
- goto out;
- }
- for (i = 0; i < conf->subvolume_cnt && missing_xattr; i++) {
- if (_gf_false ==
- dht_is_subvol_in_layout (layout, conf->subvolumes[i])) {
- dht_selfheal_dir_xattr_persubvol (frame, loc, dummy, 0,
- conf->subvolumes[i]);
- missing_xattr--;
- }
- }
-
- dht_layout_unref (this, dummy);
-out:
+ dht_local_t *local = NULL;
+ int missing_xattr = 0;
+ int i = 0;
+ xlator_t *this = NULL;
+ dht_conf_t *conf = NULL;
+ dht_layout_t *dummy = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {
+ 0,
+ };
+
+ local = frame->local;
+ this = frame->this;
+ conf = this->private;
+
+ for (i = 0; i < layout->cnt; i++) {
+ if (layout->list[i].err != -1 || !layout->list[i].stop) {
+ /* err != -1 would mean xattr present on the directory
+ * or the directory is non existent.
+ * !layout->list[i].stop would mean layout absent
+ */
+
+ continue;
+ }
+ missing_xattr++;
+ }
+ /* Also account for subvolumes with no-layout. Used for zero'ing out
+ * the layouts and for setting quota key's if present */
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (_gf_false == dht_is_subvol_in_layout(layout, conf->subvolumes[i])) {
+ missing_xattr++;
+ }
+ }
+ gf_msg_trace(this->name, 0, "%d subvolumes missing xattr for %s",
+ missing_xattr, loc->path);
+
+ if (missing_xattr == 0) {
+ dht_selfheal_dir_finish(frame, this, 0, 1);
return 0;
-}
-
-gf_boolean_t
-dht_is_subvol_part_of_layout (dht_layout_t *layout, xlator_t *xlator)
-{
- int i = 0;
- gf_boolean_t ret = _gf_false;
-
- for (i = 0; i < layout->cnt; i++) {
- if (!strcmp (layout->list[i].xlator->name, xlator->name)) {
- ret = _gf_true;
- break;
-
- }
- }
-
- return ret;
+ }
+
+ local->call_cnt = missing_xattr;
+
+ if (gf_log_get_loglevel() >= GF_LOG_DEBUG)
+ dht_log_new_layout_for_dir_selfheal(this, loc, layout);
+
+ for (i = 0; i < layout->cnt; i++) {
+ if (layout->list[i].err != -1 || !layout->list[i].stop)
+ continue;
+
+ dht_selfheal_dir_xattr_persubvol(frame, loc, layout, i, NULL);
+
+ if (--missing_xattr == 0)
+ break;
+ }
+ dummy = dht_layout_new(this, 1);
+ if (!dummy) {
+ gf_uuid_unparse(loc->gfid, gfid);
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_DUMMY_ALLOC_FAILED,
+ "path=%s", loc->path, "gfid=%s", gfid, NULL);
+ goto out;
+ }
+ for (i = 0; i < conf->subvolume_cnt && missing_xattr; i++) {
+ if (_gf_false == dht_is_subvol_in_layout(layout, conf->subvolumes[i])) {
+ dht_selfheal_dir_xattr_persubvol(frame, loc, dummy, 0,
+ conf->subvolumes[i]);
+ missing_xattr--;
+ }
+ }
+
+ dht_layout_unref(this, dummy);
+out:
+ return 0;
}
int
-dht_layout_index_from_conf (dht_layout_t *layout, xlator_t *xlator)
+dht_selfheal_dir_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
{
- int i = -1;
- int j = 0;
+ dht_local_t *local = NULL;
+ dht_layout_t *layout = NULL;
+ int this_call_cnt = 0, ret = -1;
- for (j = 0; j < layout->cnt; j++) {
- if (!strcmp (layout->list[j].xlator->name, xlator->name)) {
- i = j;
- break;
- }
- }
-
- return i;
-}
-
-
-static int
-dht_selfheal_dir_xattr_for_nameless_lookup (call_frame_t *frame, loc_t *loc,
- dht_layout_t *layout)
-{
- dht_local_t *local = NULL;
- int missing_xattr = 0;
- int i = 0;
- xlator_t *this = NULL;
- dht_conf_t *conf = NULL;
- dht_layout_t *dummy = NULL;
- int j = 0;
-
- local = frame->local;
- this = frame->this;
- conf = this->private;
-
- for (i = 0; i < layout->cnt; i++) {
- if (layout->list[i].err != -1 || !layout->list[i].stop) {
- /* err != -1 would mean xattr present on the directory
- or the directory is non existent.
- !layout->list[i].stop would mean layout absent
- */
+ local = frame->local;
+ layout = local->selfheal.layout;
- continue;
- }
- missing_xattr++;
- }
-
- /* Also account for subvolumes with no-layout. Used for zero'ing out
- the layouts and for setting quota key's if present */
-
- /* Send where either the subvol is not part of layout,
- * or it is part of the layout but error is non-zero but error
- * is not equal to -1 or ENOENT.
- */
+ this_call_cnt = dht_frame_return(frame);
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (dht_is_subvol_part_of_layout (layout, conf->subvolumes[i])
- == _gf_false) {
- missing_xattr++;
- continue;
- }
-
- j = dht_layout_index_from_conf (layout, conf->subvolumes[i]);
-
- if ((j != -1) && (layout->list[j].err != -1) &&
- (layout->list[j].err != 0) &&
- (layout->list[j].err != ENOENT)) {
- missing_xattr++;
- }
+ if (is_last_call(this_call_cnt)) {
+ if (!local->heal_layout) {
+ gf_msg_trace(this->name, 0, "Skip heal layout for %s gfid = %s ",
+ local->loc.path, uuid_utoa(local->gfid));
+ dht_selfheal_dir_finish(frame, this, 0, 1);
+ return 0;
}
+ ret = dht_selfheal_layout_lock(frame, layout, _gf_false,
+ dht_selfheal_dir_xattr,
+ dht_should_heal_layout);
-
- gf_msg_trace (this->name, 0,
- "%d subvolumes missing xattr for %s",
- missing_xattr, loc->path);
-
- if (missing_xattr == 0) {
- dht_selfheal_dir_finish (frame, this, 0, 1);
- return 0;
- }
-
- local->call_cnt = missing_xattr;
-
- if (gf_log_get_loglevel () >= GF_LOG_DEBUG)
- dht_log_new_layout_for_dir_selfheal (this, loc, layout);
-
- for (i = 0; i < layout->cnt; i++) {
- if (layout->list[i].err != -1 || !layout->list[i].stop)
- continue;
-
- dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i, NULL);
-
- if (--missing_xattr == 0)
- break;
- }
-
- dummy = dht_layout_new (this, 1);
- if (!dummy)
- goto out;
-
- for (i = 0; i < conf->subvolume_cnt && missing_xattr; i++) {
- if (dht_is_subvol_part_of_layout (layout, conf->subvolumes[i])
- == _gf_false) {
- dht_selfheal_dir_xattr_persubvol (frame, loc, dummy, 0,
- conf->subvolumes[i]);
- missing_xattr--;
- continue;
- }
-
- j = dht_layout_index_from_conf (layout, conf->subvolumes[i]);
-
- if ((j != -1) && (layout->list[j].err != -1) &&
- (layout->list[j].err != ENOENT) &&
- (layout->list[j].err != 0)) {
- dht_selfheal_dir_xattr_persubvol (frame, loc, dummy, 0,
- conf->subvolumes[i]);
- missing_xattr--;
- }
+ if (ret < 0) {
+ dht_selfheal_dir_finish(frame, this, -1, 1);
}
+ }
- dht_layout_unref (this, dummy);
-out:
- return 0;
-
+ return 0;
}
int
-dht_selfheal_dir_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *statpre,
- struct iatt *statpost, dict_t *xdata)
+dht_selfheal_dir_setattr(call_frame_t *frame, loc_t *loc, struct iatt *stbuf,
+ int32_t valid, dht_layout_t *layout)
{
- dht_local_t *local = NULL;
- dht_layout_t *layout = NULL;
- int this_call_cnt = 0, ret = -1;
-
- local = frame->local;
- layout = local->selfheal.layout;
+ int missing_attr = 0;
+ int i = 0, ret = -1;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ xlator_t *this = NULL;
+ int cnt = 0;
+
+ local = frame->local;
+ this = frame->this;
+ conf = this->private;
+
+ /* We need to heal the attrs if:
+ * 1. Any directories were missing - the newly created dirs will need
+ * to have the correct attrs set
+ * 2. An existing dir does not have the correct permissions -they may
+ * have been changed when a brick was down.
+ */
+
+ for (i = 0; i < layout->cnt; i++) {
+ if (layout->list[i].err == -1)
+ missing_attr++;
+ }
+
+ if ((missing_attr == 0) && (local->need_attrheal == 0)) {
+ if (!local->heal_layout) {
+ gf_msg_trace(this->name, 0, "Skip heal layout for %s gfid = %s ",
+ loc->path, uuid_utoa(loc->gfid));
+ dht_selfheal_dir_finish(frame, this, 0, 1);
+ return 0;
+ }
+ ret = dht_selfheal_layout_lock(frame, layout, _gf_false,
+ dht_selfheal_dir_xattr,
+ dht_should_heal_layout);
- this_call_cnt = dht_frame_return (frame);
-
- if (is_last_call (this_call_cnt)) {
- if (!local->heal_layout) {
- gf_msg_trace (this->name, 0,
- "Skip heal layout for %s gfid = %s ",
- local->loc.path, uuid_utoa(local->gfid));
-
- dht_selfheal_dir_finish (frame, this, 0, 1);
- return 0;
- }
- ret = dht_selfheal_layout_lock (frame, layout, _gf_false,
- dht_selfheal_dir_xattr,
- dht_should_heal_layout);
-
- if (ret < 0) {
- dht_selfheal_dir_finish (frame, this, -1, 1);
- }
+ if (ret < 0) {
+ dht_selfheal_dir_finish(frame, this, -1, 1);
}
return 0;
-}
-
-
-int
-dht_selfheal_dir_setattr (call_frame_t *frame, loc_t *loc, struct iatt *stbuf,
- int32_t valid, dht_layout_t *layout)
-{
- int missing_attr = 0;
- int i = 0, ret = -1;
- dht_local_t *local = NULL;
- xlator_t *this = NULL;
- int cnt = 0;
-
- local = frame->local;
- this = frame->this;
-
- for (i = 0; i < layout->cnt; i++) {
- if (layout->list[i].err == -1)
- missing_attr++;
- }
+ }
- if (missing_attr == 0) {
- if (!local->heal_layout) {
- gf_msg_trace (this->name, 0,
- "Skip heal layout for %s gfid = %s ",
- loc->path, uuid_utoa(loc->gfid));
- dht_selfheal_dir_finish (frame, this, 0, 1);
- return 0;
- }
- ret = dht_selfheal_layout_lock (frame, layout, _gf_false,
- dht_selfheal_dir_xattr,
- dht_should_heal_layout);
-
- if (ret < 0) {
- dht_selfheal_dir_finish (frame, this, -1, 1);
- }
+ cnt = local->call_cnt = conf->subvolume_cnt;
- return 0;
- }
+ for (i = 0; i < cnt; i++) {
+ STACK_WIND(frame, dht_selfheal_dir_setattr_cbk, layout->list[i].xlator,
+ layout->list[i].xlator->fops->setattr, loc, stbuf, valid,
+ NULL);
+ }
- local->call_cnt = missing_attr;
- cnt = layout->cnt;
-
- for (i = 0; i < cnt; i++) {
- if (layout->list[i].err == -1) {
- gf_msg_trace (this->name, 0,
- "%s: setattr on subvol %s, gfid = %s",
- loc->path, layout->list[i].xlator->name,
- uuid_utoa(loc->gfid));
-
- STACK_WIND (frame, dht_selfheal_dir_setattr_cbk,
- layout->list[i].xlator,
- layout->list[i].xlator->fops->setattr,
- loc, stbuf, valid, NULL);
- }
- }
-
- return 0;
+ return 0;
}
-int
-dht_selfheal_dir_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+static int
+dht_selfheal_dir_mkdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- dht_local_t *local = NULL;
- dht_layout_t *layout = NULL;
- xlator_t *prev = NULL;
- xlator_t *subvol = NULL;
- int i = 0, ret = -1;
- int this_call_cnt = 0;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- local = frame->local;
- layout = local->selfheal.layout;
- prev = cookie;
- subvol = prev;
-
- if ((op_ret == 0) || ((op_ret == -1) && (op_errno == EEXIST))) {
- for (i = 0; i < layout->cnt; i++) {
- if (layout->list[i].xlator == subvol) {
- layout->list[i].err = -1;
- break;
- }
- }
+ dht_local_t *local = NULL;
+ dht_layout_t *layout = NULL;
+ xlator_t *prev = NULL;
+ xlator_t *subvol = NULL;
+ int i = 0, ret = -1;
+ int this_call_cnt = 0;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+
+ local = frame->local;
+ layout = local->selfheal.layout;
+ prev = cookie;
+ subvol = prev;
+
+ if ((op_ret == 0) || ((op_ret == -1) && (op_errno == EEXIST))) {
+ for (i = 0; i < layout->cnt; i++) {
+ if (layout->list[i].xlator == subvol) {
+ layout->list[i].err = -1;
+ break;
+ }
}
+ }
- if (op_ret) {
- gf_uuid_unparse(local->loc.gfid, gfid);
- gf_msg (this->name, ((op_errno == EEXIST) ? GF_LOG_DEBUG :
- GF_LOG_WARNING),
- op_errno, DHT_MSG_DIR_SELFHEAL_FAILED,
- "Directory selfheal failed: path = %s, gfid = %s",
- local->loc.path, gfid );
- goto out;
- }
- dht_iatt_merge (this, &local->preparent, preparent, prev);
- dht_iatt_merge (this, &local->postparent, postparent, prev);
- ret = 0;
+ if (op_ret) {
+ gf_uuid_unparse(local->loc.gfid, gfid);
+ gf_smsg(this->name,
+ ((op_errno == EEXIST) ? GF_LOG_DEBUG : GF_LOG_WARNING),
+ op_errno, DHT_MSG_DIR_SELFHEAL_FAILED, "path=%s",
+ local->loc.path, "gfid=%s", gfid, NULL);
+ goto out;
+ }
+ dht_iatt_merge(this, &local->preparent, preparent);
+ dht_iatt_merge(this, &local->postparent, postparent);
+ ret = 0;
out:
- this_call_cnt = dht_frame_return (frame);
+ this_call_cnt = dht_frame_return(frame);
- if (is_last_call (this_call_cnt)) {
- dht_selfheal_dir_finish (frame, this, ret, 0);
- dht_selfheal_dir_setattr (frame, &local->loc, &local->stbuf, 0xffffff, layout);
- }
+ if (is_last_call(this_call_cnt)) {
+ dht_selfheal_dir_finish(frame, this, ret, 0);
+ dht_selfheal_dir_setattr(frame, &local->loc, &local->stbuf, 0xffffff,
+ layout);
+ }
- return 0;
+ return 0;
}
-void
-dht_selfheal_dir_mkdir_setacl (dict_t *xattr, dict_t *dict)
+static int
+dht_selfheal_dir_mkdir_lookup_done(call_frame_t *frame, xlator_t *this)
{
- data_t *acl_default = NULL;
- data_t *acl_access = NULL;
- xlator_t *this = NULL;
- int ret = -1;
-
- GF_ASSERT (xattr);
- GF_ASSERT (dict);
-
- this = THIS;
- GF_ASSERT (this);
-
- acl_default = dict_get (xattr, POSIX_ACL_DEFAULT_XATTR);
-
- if (!acl_default) {
- gf_msg_debug (this->name, 0,
- "ACL_DEFAULT xattr not present");
- goto cont;
- }
- ret = dict_set (dict, POSIX_ACL_DEFAULT_XATTR, acl_default);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value.key = %s",
- POSIX_ACL_DEFAULT_XATTR);
-cont:
- acl_access = dict_get (xattr, POSIX_ACL_ACCESS_XATTR);
- if (!acl_access) {
- gf_msg_debug (this->name, 0,
- "ACL_ACCESS xattr not present");
- goto out;
- }
- ret = dict_set (dict, POSIX_ACL_ACCESS_XATTR, acl_access);
+ dht_local_t *local = NULL;
+ int i = 0;
+ dict_t *dict = NULL;
+ dht_layout_t *layout = NULL;
+ loc_t *loc = NULL;
+ int cnt = 0;
+ int ret = -1;
+
+ VALIDATE_OR_GOTO(this->private, err);
+
+ local = frame->local;
+ layout = local->layout;
+ loc = &local->loc;
+
+ if (!gf_uuid_is_null(local->gfid)) {
+ dict = dict_new();
+ if (!dict)
+ return -1;
+
+ ret = dict_set_gfuuid(dict, "gfid-req", local->gfid, true);
if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value.key = %s",
- POSIX_ACL_ACCESS_XATTR);
-
-out:
- return;
-}
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_DICT_SET_FAILED,
+ "path=%s", loc->path, "key=gfid-req", NULL);
+ } else if (local->params) {
+ /* Send the dictionary from higher layers directly */
+
+ dict = dict_ref(local->params);
+ }
+ /* Code to update all extended attributed from local->xattr
+ to dict
+ */
+ dht_dir_set_heal_xattr(this, local, dict, local->xattr, NULL, NULL);
+
+ if (!dict) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_DICT_IS_NULL, NULL);
+ dict = dict_new();
+ if (!dict)
+ return -1;
+ }
+ ret = dict_set_flag(dict, GF_INTERNAL_CTX_KEY, GF_DHT_HEAL_DIR);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED, "key=%s",
+ GF_INTERNAL_CTX_KEY, "path=%s", loc->path, NULL);
+ /* We can still continue. As heal can still happen
+ * unless quota limits have reached for the dir.
+ */
+ }
+ cnt = layout->cnt;
+ for (i = 0; i < cnt; i++) {
+ if (layout->list[i].err == ESTALE || layout->list[i].err == ENOENT ||
+ local->selfheal.force_mkdir) {
+ gf_msg_debug(this->name, 0, "Creating directory %s on subvol %s",
+ loc->path, layout->list[i].xlator->name);
-void
-dht_selfheal_dir_mkdir_setquota (dict_t *src, dict_t *dst)
-{
- data_t *quota_limit_key = NULL;
- data_t *quota_limit_obj_key = NULL;
- xlator_t *this = NULL;
- int ret = -1;
-
- GF_ASSERT (src);
- GF_ASSERT (dst);
-
- this = THIS;
- GF_ASSERT (this);
-
- quota_limit_key = dict_get (src, QUOTA_LIMIT_KEY);
- if (!quota_limit_key) {
- gf_msg_debug (this->name, 0,
- "QUOTA_LIMIT_KEY xattr not present");
- goto cont;
+ STACK_WIND_COOKIE(
+ frame, dht_selfheal_dir_mkdir_cbk, layout->list[i].xlator,
+ layout->list[i].xlator, layout->list[i].xlator->fops->mkdir,
+ loc,
+ st_mode_from_ia(local->stbuf.ia_prot, local->stbuf.ia_type), 0,
+ dict);
}
- ret = dict_set(dst, QUOTA_LIMIT_KEY, quota_limit_key);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value.key = %s",
- QUOTA_LIMIT_KEY);
-
-cont:
- quota_limit_obj_key = dict_get (src, QUOTA_LIMIT_OBJECTS_KEY);
- if (!quota_limit_obj_key) {
- gf_msg_debug (this->name, 0,
- "QUOTA_LIMIT_OBJECTS_KEY xattr not present");
- goto out;
- }
- ret = dict_set (dst, QUOTA_LIMIT_OBJECTS_KEY, quota_limit_obj_key);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value.key = %s",
- QUOTA_LIMIT_OBJECTS_KEY);
-
-out:
- return;
-}
-
+ }
+ if (dict)
+ dict_unref(dict);
+ return 0;
+err:
+ dht_selfheal_dir_finish(frame, this, -1, 1);
+ return 0;
+}
-int
-dht_selfheal_dir_mkdir_lookup_done (call_frame_t *frame, xlator_t *this)
+static int
+dht_selfheal_dir_mkdir_lookup_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int op_ret, int op_errno,
+ inode_t *inode, struct iatt *stbuf,
+ dict_t *xattr, struct iatt *postparent)
{
- dht_local_t *local = NULL;
- int i = 0;
- dict_t *dict = NULL;
- dht_layout_t *layout = NULL;
- loc_t *loc = NULL;
- int cnt = 0;
- int ret = -1;
-
- VALIDATE_OR_GOTO (this->private, err);
-
- local = frame->local;
- layout = local->layout;
- loc = &local->loc;
-
- if (!gf_uuid_is_null (local->gfid)) {
- dict = dict_new ();
- if (!dict)
- return -1;
-
- ret = dict_set_gfuuid (dict, "gfid-req", local->gfid, true);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "%s: Failed to set dictionary value:"
- " key = gfid-req", loc->path);
- } else if (local->params) {
- /* Send the dictionary from higher layers directly */
-
- dict = dict_ref (local->params);
- }
- /* Code to update all extended attributed from local->xattr
- to dict
- */
- dht_dir_set_heal_xattr (this, local, dict, local->xattr, NULL,
- NULL);
-
- if (!dict) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "dict is NULL, need to make sure gfids are same");
- dict = dict_new ();
- if (!dict)
- return -1;
- }
- ret = dict_set_flag (dict, GF_INTERNAL_CTX_KEY, GF_DHT_HEAL_DIR);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value for"
- " key = %s at path: %s",
- GF_INTERNAL_CTX_KEY, loc->path);
- /* We can still continue. As heal can still happen
- * unless quota limits have reached for the dir.
- */
- }
+ dht_local_t *local = NULL;
+ int i = 0;
+ int this_call_cnt = 0;
+ int missing_dirs = 0;
+ dht_layout_t *layout = NULL;
+ xlator_t *prev = 0;
+ loc_t *loc = NULL;
+ char gfid_local[GF_UUID_BUF_SIZE] = {0};
+ int index = -1;
+
+ VALIDATE_OR_GOTO(this->private, err);
+
+ local = frame->local;
+ layout = local->layout;
+ loc = &local->loc;
+ prev = cookie;
+
+ if (!gf_uuid_is_null(local->gfid))
+ gf_uuid_unparse(local->gfid, gfid_local);
+
+ LOCK(&frame->lock);
+ {
+ index = dht_layout_index_for_subvol(layout, prev);
+ if ((op_ret < 0) && (op_errno == ENOENT || op_errno == ESTALE)) {
+ local->selfheal.hole_cnt = !local->selfheal.hole_cnt
+ ? 1
+ : local->selfheal.hole_cnt + 1;
+ /* the status might have changed. Update the layout with the
+ * new status
+ */
+ if (index >= 0) {
+ layout->list[index].err = op_errno;
+ }
+ }
+
+ if (!op_ret) {
+ dht_iatt_merge(this, &local->stbuf, stbuf);
+ if (prev == local->mds_subvol) {
+ dict_unref(local->xattr);
+ local->xattr = dict_ref(xattr);
+ }
+ /* the status might have changed. Update the layout with the
+ * new status
+ */
+ if (index >= 0) {
+ layout->list[index].err = -1;
+ }
+ }
+ }
+ UNLOCK(&frame->lock);
+
+ this_call_cnt = dht_frame_return(frame);
+
+ if (is_last_call(this_call_cnt)) {
+ if (local->selfheal.hole_cnt == layout->cnt) {
+ gf_msg_debug(this->name, op_errno,
+ "Lookup failed, an rmdir could have "
+ "deleted this entry %s",
+ loc->name);
+ local->op_errno = op_errno;
+ goto err;
+ } else {
+ for (i = 0; i < layout->cnt; i++) {
+ if (layout->list[i].err == ENOENT ||
+ layout->list[i].err == ESTALE ||
+ local->selfheal.force_mkdir)
+ missing_dirs++;
+ }
+
+ if (missing_dirs == 0) {
+ dht_selfheal_dir_finish(frame, this, 0, 0);
+ dht_selfheal_dir_setattr(frame, loc, &local->stbuf, 0xffffffff,
+ layout);
+ return 0;
+ }
- cnt = layout->cnt;
- for (i = 0; i < cnt; i++) {
- if (layout->list[i].err == ESTALE ||
- layout->list[i].err == ENOENT ||
- local->selfheal.force_mkdir) {
- gf_msg_debug (this->name, 0,
- "Creating directory %s on subvol %s",
- loc->path, layout->list[i].xlator->name);
-
- STACK_WIND_COOKIE (frame, dht_selfheal_dir_mkdir_cbk,
- layout->list[i].xlator,
- layout->list[i].xlator,
- layout->list[i].xlator->fops->mkdir,
- loc,
- st_mode_from_ia (local->stbuf.ia_prot,
- local->stbuf.ia_type),
- 0, dict);
- }
+ local->call_cnt = missing_dirs;
+ dht_selfheal_dir_mkdir_lookup_done(frame, this);
}
+ }
- if (dict)
- dict_unref (dict);
-
- return 0;
+ return 0;
err:
- dht_selfheal_dir_finish (frame, this, -1, 1);
- return 0;
+ dht_selfheal_dir_finish(frame, this, -1, 1);
+ return 0;
}
-int
-dht_selfheal_dir_mkdir_lookup_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf,
- dict_t *xattr, struct iatt *postparent)
+static int
+dht_selfheal_dir_mkdir_lock_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int i = 0;
- int this_call_cnt = 0;
- int missing_dirs = 0;
- dht_layout_t *layout = NULL;
- dht_conf_t *conf = 0;
- loc_t *loc = NULL;
- xlator_t *prev = NULL;
- int check_mds = 0;
- int errst = 0;
- int32_t mds_xattr_val[1] = {0};
- char gfid_local[GF_UUID_BUF_SIZE] = {0};
-
- VALIDATE_OR_GOTO (this->private, err);
-
- local = frame->local;
- layout = local->layout;
- loc = &local->loc;
- prev = cookie;
- conf = this->private;
-
- if (local->gfid)
- gf_uuid_unparse(local->gfid, gfid_local);
-
- this_call_cnt = dht_frame_return (frame);
-
- LOCK (&frame->lock);
- {
- if ((op_ret < 0) &&
- (op_errno == ENOENT || op_errno == ESTALE)) {
- local->selfheal.hole_cnt = !local->selfheal.hole_cnt ? 1
- : local->selfheal.hole_cnt + 1;
- }
-
- if (!op_ret) {
- dht_iatt_merge (this, &local->stbuf, stbuf, prev);
- }
- check_mds = dht_dict_get_array (xattr, conf->mds_xattr_key,
- mds_xattr_val, 1, &errst);
- if (dict_get (xattr, conf->mds_xattr_key) && check_mds && !errst) {
- dict_unref (local->xattr);
- local->xattr = dict_ref (xattr);
- }
-
- }
- UNLOCK (&frame->lock);
-
- if (is_last_call (this_call_cnt)) {
- if (local->selfheal.hole_cnt == layout->cnt) {
- gf_msg_debug (this->name, op_errno,
- "Lookup failed, an rmdir could have "
- "deleted this entry %s", loc->name);
- local->op_errno = op_errno;
- goto err;
- } else {
- for (i = 0; i < layout->cnt; i++) {
- if (layout->list[i].err == ENOENT ||
- layout->list[i].err == ESTALE ||
- local->selfheal.force_mkdir)
- missing_dirs++;
- }
-
- if (missing_dirs == 0) {
- dht_selfheal_dir_finish (frame, this, 0, 0);
- dht_selfheal_dir_setattr (frame, loc,
- &local->stbuf,
- 0xffffffff, layout);
- return 0;
- }
-
- local->call_cnt = missing_dirs;
- dht_selfheal_dir_mkdir_lookup_done (frame, this);
- }
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int i = 0;
+ int ret = -1;
+ xlator_t *mds_subvol = NULL;
+
+ VALIDATE_OR_GOTO(this->private, err);
+
+ conf = this->private;
+ local = frame->local;
+ mds_subvol = local->mds_subvol;
+
+ local->call_cnt = conf->subvolume_cnt;
+
+ if (op_ret < 0) {
+ if (op_errno == EINVAL) {
+ local->call_cnt = 1;
+ dht_selfheal_dir_mkdir_lookup_done(frame, this);
+ return 0;
+ }
+
+ gf_smsg(this->name, GF_LOG_WARNING, op_errno, DHT_MSG_ENTRYLK_ERROR,
+ "path=%s", local->loc.path, NULL);
+
+ local->op_errno = op_errno;
+ goto err;
+ }
+
+ /* After getting locks, perform lookup again to ensure that the
+ directory was not deleted by a racing rmdir
+ */
+ if (!local->xattr_req)
+ local->xattr_req = dict_new();
+
+ ret = dict_set_int32(local->xattr_req, "list-xattr", 1);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED, "path=%s",
+ local->loc.path, NULL);
+
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (mds_subvol && conf->subvolumes[i] == mds_subvol) {
+ STACK_WIND_COOKIE(frame, dht_selfheal_dir_mkdir_lookup_cbk,
+ conf->subvolumes[i], conf->subvolumes[i],
+ conf->subvolumes[i]->fops->lookup, &local->loc,
+ local->xattr_req);
+ } else {
+ STACK_WIND_COOKIE(frame, dht_selfheal_dir_mkdir_lookup_cbk,
+ conf->subvolumes[i], conf->subvolumes[i],
+ conf->subvolumes[i]->fops->lookup, &local->loc,
+ NULL);
}
+ }
- return 0;
+ return 0;
err:
- dht_selfheal_dir_finish (frame, this, -1, 1);
- return 0;
+ dht_selfheal_dir_finish(frame, this, -1, 1);
+ return 0;
}
-
-int
-dht_selfheal_dir_mkdir_lock_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
+static int
+dht_selfheal_dir_mkdir(call_frame_t *frame, loc_t *loc, dht_layout_t *layout,
+ int force)
{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- int i = 0;
- int ret = -1;
- xlator_t *mds_subvol = NULL;
-
- VALIDATE_OR_GOTO (this->private, err);
-
- conf = this->private;
- local = frame->local;
- mds_subvol = local->mds_subvol;
-
- local->call_cnt = conf->subvolume_cnt;
-
- if (op_ret < 0) {
-
- /* We get this error when the directory entry was not created
- * on a newky attatched tier subvol. Hence proceed and do mkdir
- * on the tier subvol.
- */
- if (op_errno == EINVAL) {
- local->call_cnt = 1;
- dht_selfheal_dir_mkdir_lookup_done (frame, this);
- return 0;
+ int missing_dirs = 0;
+ int i = 0;
+ int op_errno = 0;
+ int ret = -1;
+ dht_local_t *local = NULL;
+ xlator_t *this = NULL;
+ dht_conf_t *conf = NULL;
+
+ local = frame->local;
+ this = frame->this;
+ conf = this->private;
+
+ local->selfheal.force_mkdir = force;
+ local->selfheal.hole_cnt = 0;
+
+ for (i = 0; i < layout->cnt; i++) {
+ if (layout->list[i].err == ENOENT || force)
+ missing_dirs++;
+ }
+
+ if (missing_dirs == 0) {
+ /* We don't need to create any directories. Proceed to heal the
+ * attrs and xattrs
+ */
+ if (!__is_root_gfid(local->stbuf.ia_gfid)) {
+ if (local->need_xattr_heal) {
+ local->need_xattr_heal = 0;
+ ret = dht_dir_xattr_heal(this, local, &op_errno);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno,
+ DHT_MSG_DIR_XATTR_HEAL_FAILED, "path=%s",
+ local->loc.path, "gfid=%s", local->gfid, NULL);
}
+ } else {
+ if (!gf_uuid_is_null(local->gfid))
+ gf_uuid_copy(loc->gfid, local->gfid);
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- DHT_MSG_ENTRYLK_ERROR,
- "acquiring entrylk after inodelk failed for %s",
- local->loc.path);
-
- local->op_errno = op_errno;
- goto err;
- }
-
- /* After getting locks, perform lookup again to ensure that the
- directory was not deleted by a racing rmdir
- */
- if (!local->xattr_req)
- local->xattr_req = dict_new ();
+ ret = dht_common_mark_mdsxattr(frame, NULL, 0);
+ if (!ret)
+ return 0;
- ret = dict_set_int32 (local->xattr_req, "list-xattr", 1);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary key list-xattr value "
- " for path %s ", local->loc.path);
-
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (mds_subvol && conf->subvolumes[i] == mds_subvol) {
- STACK_WIND_COOKIE (frame,
- dht_selfheal_dir_mkdir_lookup_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->lookup,
- &local->loc, local->xattr_req);
- } else {
- STACK_WIND_COOKIE (frame,
- dht_selfheal_dir_mkdir_lookup_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->lookup,
- &local->loc, NULL);
- }
+ gf_smsg(this->name, GF_LOG_INFO, 0, DHT_MSG_SET_XATTR_FAILED,
+ "path=%s", local->loc.path, "gfid=%s", local->gfid,
+ NULL);
+ }
}
-
+ dht_selfheal_dir_setattr(frame, loc, &local->stbuf, 0xffffffff, layout);
return 0;
-
-err:
- dht_selfheal_dir_finish (frame, this, -1, 1);
- return 0;
-}
-
-int
-dht_selfheal_dir_mkdir (call_frame_t *frame, loc_t *loc,
- dht_layout_t *layout, int force)
-{
- int missing_dirs = 0;
- int i = 0;
- int ret = -1;
- dht_local_t *local = NULL;
- xlator_t *this = NULL;
-
- local = frame->local;
- this = frame->this;
-
- local->selfheal.force_mkdir = force;
- local->selfheal.hole_cnt = 0;
-
- for (i = 0; i < layout->cnt; i++) {
- if (layout->list[i].err == ENOENT || force)
- missing_dirs++;
+ }
+
+ /* MDS xattr is populated only while DHT is having more than one
+ subvol.In case of graph switch while adding more dht subvols need to
+ consider hash subvol as a MDS to avoid MDS check failure at the time
+ of running fop on directory
+ */
+ if (!dict_get(local->xattr, conf->mds_xattr_key) &&
+ (conf->subvolume_cnt > 1)) {
+ if (local->hashed_subvol == NULL) {
+ local->hashed_subvol = dht_subvol_get_hashed(this, loc);
+ if (local->hashed_subvol == NULL) {
+ local->op_errno = EINVAL;
+ gf_smsg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_HASHED_SUBVOL_GET_FAILED, "gfid=%s",
+ loc->pargfid, "name=%s", loc->name, "path=%s",
+ loc->path, NULL);
+ goto err;
+ }
}
-
- if (missing_dirs == 0) {
- if (!__is_root_gfid (local->stbuf.ia_gfid)) {
- if (local->need_xattr_heal) {
- local->need_xattr_heal = 0;
- ret = dht_dir_xattr_heal (this, local);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR,
- ret,
- DHT_MSG_DIR_XATTR_HEAL_FAILED,
- "xattr heal failed for "
- "directory %s gfid %s ",
- local->loc.path,
- local->gfid);
- } else {
- if (!gf_uuid_is_null (local->gfid))
- gf_uuid_copy (loc->gfid, local->gfid);
-
- ret = dht_common_mark_mdsxattr (frame, NULL, 0);
- if (!ret)
- return 0;
-
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_DIR_XATTR_HEAL_FAILED,
- "Failed to set mds xattr "
- "for directory %s gfid %s ",
- local->loc.path, local->gfid);
- }
- }
- dht_selfheal_dir_setattr (frame, loc, &local->stbuf,
- 0xffffffff, layout);
- return 0;
+ ret = dht_inode_ctx_mdsvol_set(local->inode, this,
+ local->hashed_subvol);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_SET_INODE_CTX_FAILED,
+ "Failed to set hashed subvol for %s on inode vol is %s",
+ local->loc.path,
+ local->hashed_subvol ? local->hashed_subvol->name : "NULL");
+ goto err;
}
+ }
- if (local->hashed_subvol == NULL)
- local->hashed_subvol = dht_subvol_get_hashed (this, loc);
-
+ if (local->hashed_subvol == NULL) {
+ local->hashed_subvol = dht_subvol_get_hashed(this, loc);
if (local->hashed_subvol == NULL) {
- local->op_errno = EINVAL;
- gf_msg (this->name, GF_LOG_WARNING, local->op_errno,
- DHT_MSG_HASHED_SUBVOL_GET_FAILED,
- "(%s/%s) (path: %s): "
- "hashed subvolume not found", loc->pargfid, loc->name,
- loc->path);
- goto err;
+ local->op_errno = EINVAL;
+ gf_smsg(this->name, GF_LOG_WARNING, local->op_errno,
+ DHT_MSG_HASHED_SUBVOL_GET_FAILED, "gfid=%s", loc->pargfid,
+ "name=%s", loc->name, "path=%s", loc->path, NULL);
+ goto err;
}
+ }
- local->current = &local->lock[0];
- ret = dht_protect_namespace (frame, loc, local->hashed_subvol,
- &local->current->ns,
- dht_selfheal_dir_mkdir_lock_cbk);
+ local->current = &local->lock[0];
+ ret = dht_protect_namespace(frame, loc, local->hashed_subvol,
+ &local->current->ns,
+ dht_selfheal_dir_mkdir_lock_cbk);
- if (ret < 0)
- goto err;
+ if (ret < 0)
+ goto err;
- return 0;
+ return 0;
err:
- return -1;
+ return -1;
}
-int
-dht_selfheal_layout_alloc_start (xlator_t *this, loc_t *loc,
- dht_layout_t *layout)
+static int
+dht_selfheal_layout_alloc_start(xlator_t *this, loc_t *loc,
+ dht_layout_t *layout)
{
- int start = 0;
- uint32_t hashval = 0;
- int ret = 0;
- const char *str = NULL;
- dht_conf_t *conf = NULL;
- char buf[UUID_CANONICAL_FORM_LEN + 1] = {0, };
-
- conf = this->private;
-
- if (conf->randomize_by_gfid) {
- str = uuid_utoa_r (loc->gfid, buf);
- } else {
- str = loc->path;
- }
-
- ret = dht_hash_compute (this, layout->type, str, &hashval);
- if (ret == 0) {
- start = (hashval % layout->cnt);
- }
-
- return start;
+ int start = 0;
+ uint32_t hashval = 0;
+ int ret = 0;
+ const char *str = NULL;
+ dht_conf_t *conf = NULL;
+ char buf[UUID_CANONICAL_FORM_LEN + 1] = {
+ 0,
+ };
+
+ conf = this->private;
+
+ if (conf->randomize_by_gfid) {
+ str = uuid_utoa_r(loc->gfid, buf);
+ } else {
+ str = loc->path;
+ }
+
+ ret = dht_hash_compute(this, layout->type, str, &hashval);
+ if (ret == 0) {
+ start = (hashval % layout->cnt);
+ }
+
+ return start;
}
static int
-dht_get_layout_count (xlator_t *this, dht_layout_t *layout, int new_layout)
+dht_get_layout_count(xlator_t *this, dht_layout_t *layout, int new_layout)
{
- int i = 0;
- int j = 0;
- int err = 0;
- int count = 0;
- dht_conf_t *conf = NULL;
-
- /* Gets in use only for replace-brick, remove-brick */
- conf = this->private;
- for (i = 0; i < layout->cnt; i++) {
- for (j = 0; j < conf->subvolume_cnt; j++) {
- if (conf->decommissioned_bricks[j] &&
- conf->decommissioned_bricks[j] == layout->list[i].xlator) {
- layout->list[i].err = EINVAL;
- break;
- }
- }
- }
-
+ int i = 0;
+ int j = 0;
+ int err = 0;
+ int count = 0;
+ dht_conf_t *conf = NULL;
+
+ /* Gets in use only for replace-brick, remove-brick */
+ conf = this->private;
+ for (i = 0; i < layout->cnt; i++) {
+ for (j = 0; j < conf->subvolume_cnt; j++) {
+ if (conf->decommissioned_bricks[j] &&
+ conf->decommissioned_bricks[j] == layout->list[i].xlator) {
+ layout->list[i].err = EINVAL;
+ break;
+ }
+ }
+ }
+
+ for (i = 0; i < layout->cnt; i++) {
+ err = layout->list[i].err;
+ if (err == -1 || err == 0 || err == ENOENT) {
+ /* Take this with a pinch of salt. The behaviour seems
+ * to be slightly different when this function is
+ * invoked from mkdir codepath. For eg., err == 0 in
+ * mkdir codepath means directory created but xattr
+ * is not set yet.
+ */
+
+ /* Setting list[i].err = -1 is an indication for
+ dht_selfheal_layout_new_directory() to assign
+ a range. We set it to -1 based on any one of
+ the three criteria:
+
+ - err == -1 already, which means directory
+ existed but layout was not set on it.
+
+ - err == 0, which means directory exists and
+ has an old layout piece which will be
+ overwritten now.
+
+ - err == ENOENT, which means directory does
+ not exist (possibly racing with mkdir or
+ finishing half done mkdir). The missing
+ directory will be attempted to be recreated.
+ */
+ count++;
+ if (!err)
+ layout->list[i].err = -1;
+ }
+ }
+
+ /* no subvolume has enough space, but can't stop directory creation */
+ if (!count || !new_layout) {
for (i = 0; i < layout->cnt; i++) {
- err = layout->list[i].err;
- if (err == -1 || err == 0 || err == ENOENT) {
- /* Take this with a pinch of salt. The behaviour seems
- * to be slightly different when this function is
- * invoked from mkdir codepath. For eg., err == 0 in
- * mkdir codepath means directory created but xattr
- * is not set yet.
- */
-
- /* Setting list[i].err = -1 is an indication for
- dht_selfheal_layout_new_directory() to assign
- a range. We set it to -1 based on any one of
- the three criteria:
-
- - err == -1 already, which means directory
- existed but layout was not set on it.
-
- - err == 0, which means directory exists and
- has an old layout piece which will be
- overwritten now.
-
- - err == ENOENT, which means directory does
- not exist (possibly racing with mkdir or
- finishing half done mkdir). The missing
- directory will be attempted to be recreated.
- */
- count++;
- if (!err)
- layout->list[i].err = -1;
- }
- }
-
- /* no subvolume has enough space, but can't stop directory creation */
- if (!count || !new_layout) {
- for (i = 0; i < layout->cnt; i++) {
- err = layout->list[i].err;
- if (err == ENOSPC) {
- layout->list[i].err = -1;
- count++;
- }
- }
- }
-
- /* if layout->spread_cnt is set, check if it is <= available
- * subvolumes (down brick and decommissioned bricks are considered
- * un-availbale). Else return count (available up bricks) */
- count = ((layout->spread_cnt &&
- (layout->spread_cnt <= count)) ?
- layout->spread_cnt : ((count) ? count : 1));
-
- return count;
+ err = layout->list[i].err;
+ if (err == ENOSPC) {
+ layout->list[i].err = -1;
+ count++;
+ }
+ }
+ }
+
+ /* if layout->spread_cnt is set, check if it is <= available
+ * subvolumes (down brick and decommissioned bricks are considered
+ * un-available). Else return count (available up bricks) */
+ count = ((layout->spread_cnt && (layout->spread_cnt <= count))
+ ? layout->spread_cnt
+ : ((count) ? count : 1));
+
+ return count;
}
+void
+dht_selfheal_layout_new_directory(call_frame_t *frame, loc_t *loc,
+ dht_layout_t *new_layout);
-void dht_selfheal_layout_new_directory (call_frame_t *frame, loc_t *loc,
- dht_layout_t *new_layout);
-
-void dht_layout_entry_swap (dht_layout_t *layout, int i, int j);
-void dht_layout_range_swap (dht_layout_t *layout, int i, int j);
+void
+dht_layout_range_swap(dht_layout_t *layout, int i, int j);
/*
* It's a bit icky using local variables in a macro, but it makes the rest
* of the code a lot clearer.
*/
-#define OV_ENTRY(x,y) table[x*new->cnt+y]
+#define OV_ENTRY(x, y) table[x * new->cnt + y]
-void
-dht_selfheal_layout_maximize_overlap (call_frame_t *frame, loc_t *loc,
- dht_layout_t *new, dht_layout_t *old)
+static void
+dht_selfheal_layout_maximize_overlap(call_frame_t *frame, loc_t *loc,
+ dht_layout_t *new, dht_layout_t *old)
{
- int i = 0;
- int j = 0;
- uint32_t curr_overlap = 0;
- uint32_t max_overlap = 0;
- int max_overlap_idx = -1;
- uint32_t overlap = 0;
- uint32_t *table = NULL;
-
- dht_layout_sort_volname (old);
- /* Now both old_layout->list[] and new_layout->list[]
- are match the same xlators/subvolumes. i.e,
- old_layout->[i] and new_layout->[i] are referring
- to the same subvolumes
- */
-
- /* Build a table of overlaps between new[i] and old[j]. */
- table = alloca(sizeof(overlap)*old->cnt*new->cnt);
- if (!table) {
- return;
- }
- memset(table,0,sizeof(overlap)*old->cnt*new->cnt);
- for (i = 0; i < new->cnt; ++i) {
- for (j = 0; j < old->cnt; ++j) {
- OV_ENTRY(i,j) = dht_overlap_calc(old,j,new,i);
- }
- }
-
- for (i = 0; i < new->cnt; i++) {
- if (new->list[i].err > 0) {
- /* Subvol might be marked for decommission
- with EINVAL, or some other serious error
- marked with positive errno.
- */
- continue;
- }
-
- max_overlap = 0;
- max_overlap_idx = i;
- for (j = (i + 1); j < new->cnt; ++j) {
- if (new->list[j].err > 0) {
- /* Subvol might be marked for decommission
- with EINVAL, or some other serious error
- marked with positive errno.
- */
- continue;
- }
- /* Calculate the overlap now. */
- curr_overlap = OV_ENTRY(i,i) + OV_ENTRY(j,j);
- /* Calculate the overlap after the proposed swap. */
- overlap = OV_ENTRY(i,j) + OV_ENTRY(j,i);
- /* Are we better than status quo? */
- if (overlap > curr_overlap) {
- overlap -= curr_overlap;
- /* Are we better than the previous choice? */
- if (overlap > max_overlap) {
- max_overlap = overlap;
- max_overlap_idx = j;
- }
- }
- }
-
- if (max_overlap_idx != i) {
- dht_layout_range_swap (new, i, max_overlap_idx);
- /* Need to swap the table values too. */
- for (j = 0; j < old->cnt; ++j) {
- overlap = OV_ENTRY(i,j);
- OV_ENTRY(i,j) = OV_ENTRY(max_overlap_idx,j);
- OV_ENTRY(max_overlap_idx,j) = overlap;
- }
- }
- }
+ int i = 0;
+ int j = 0;
+ uint32_t curr_overlap = 0;
+ uint32_t max_overlap = 0;
+ int max_overlap_idx = -1;
+ uint32_t overlap = 0;
+ uint32_t *table = NULL;
+
+ dht_layout_sort_volname(old);
+ /* Now both old_layout->list[] and new_layout->list[]
+ are match the same xlators/subvolumes. i.e,
+ old_layout->[i] and new_layout->[i] are referring
+ to the same subvolumes
+ */
+
+ /* Build a table of overlaps between new[i] and old[j]. */
+ table = alloca(sizeof(overlap) * old->cnt * new->cnt);
+ if (!table) {
+ return;
+ }
+ memset(table, 0, sizeof(overlap) * old->cnt * new->cnt);
+ for (i = 0; i < new->cnt; ++i) {
+ for (j = 0; j < old->cnt; ++j) {
+ OV_ENTRY(i, j) = dht_overlap_calc(old, j, new, i);
+ }
+ }
+
+ for (i = 0; i < new->cnt; i++) {
+ if (new->list[i].err > 0) {
+ /* Subvol might be marked for decommission
+ with EINVAL, or some other serious error
+ marked with positive errno.
+ */
+ continue;
+ }
+
+ max_overlap = 0;
+ max_overlap_idx = i;
+ for (j = (i + 1); j < new->cnt; ++j) {
+ if (new->list[j].err > 0) {
+ /* Subvol might be marked for decommission
+ with EINVAL, or some other serious error
+ marked with positive errno.
+ */
+ continue;
+ }
+ /* Calculate the overlap now. */
+ curr_overlap = OV_ENTRY(i, i) + OV_ENTRY(j, j);
+ /* Calculate the overlap after the proposed swap. */
+ overlap = OV_ENTRY(i, j) + OV_ENTRY(j, i);
+ /* Are we better than status quo? */
+ if (overlap > curr_overlap) {
+ overlap -= curr_overlap;
+ /* Are we better than the previous choice? */
+ if (overlap > max_overlap) {
+ max_overlap = overlap;
+ max_overlap_idx = j;
+ }
+ }
+ }
+
+ if (max_overlap_idx != i) {
+ dht_layout_range_swap(new, i, max_overlap_idx);
+ /* Need to swap the table values too. */
+ for (j = 0; j < old->cnt; ++j) {
+ overlap = OV_ENTRY(i, j);
+ OV_ENTRY(i, j) = OV_ENTRY(max_overlap_idx, j);
+ OV_ENTRY(max_overlap_idx, j) = overlap;
+ }
+ }
+ }
}
-
-dht_layout_t *
-dht_fix_layout_of_directory (call_frame_t *frame, loc_t *loc,
- dht_layout_t *layout)
+static dht_layout_t *
+dht_fix_layout_of_directory(call_frame_t *frame, loc_t *loc,
+ dht_layout_t *layout)
{
- int i = 0;
- xlator_t *this = NULL;
- dht_layout_t *new_layout = NULL;
- dht_conf_t *priv = NULL;
- dht_local_t *local = NULL;
- uint32_t subvol_down = 0;
- int ret = 0;
- gf_boolean_t maximize_overlap = _gf_true;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
- this = frame->this;
- priv = this->private;
- local = frame->local;
-
- if (layout->type == DHT_HASH_TYPE_DM_USER) {
- gf_msg_debug (THIS->name, 0, "leaving %s alone",
- loc->path);
- goto done;
- }
-
- new_layout = dht_layout_new (this, priv->subvolume_cnt);
- if (!new_layout) {
- gf_uuid_unparse (loc->gfid, gfid);
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
- "mem allocation failed for new_layout, path:%s gfid:%s",
- loc->path, gfid);
- goto done;
- }
-
- /* If a subvolume is down, do not re-write the layout. */
- ret = dht_layout_anomalies (this, loc, layout, NULL, NULL, NULL,
- &subvol_down, NULL, NULL);
-
- if (subvol_down || (ret == -1)) {
- gf_uuid_unparse (loc->gfid, gfid);
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_LAYOUT_FIX_FAILED,
- "Layout fix failed: %u subvolume(s) are down"
- ". Skipping fix layout. path:%s gfid:%s", subvol_down,
- loc->path, gfid);
- GF_FREE (new_layout);
- return NULL;
- }
-
- for (i = 0; i < new_layout->cnt; i++) {
- if (layout->list[i].err != ENOSPC)
- new_layout->list[i].err = layout->list[i].err;
- else
- new_layout->list[i].err = -1;
-
- new_layout->list[i].xlator = layout->list[i].xlator;
- }
-
- new_layout->commit_hash = layout->commit_hash;
-
- if (priv->du_stats) {
- for (i = 0; i < priv->subvolume_cnt; ++i) {
- gf_msg (this->name, GF_LOG_DEBUG, 0,
- DHT_MSG_SUBVOL_INFO,
- "subvolume %d (%s): %u chunks, path:%s", i,
- priv->subvolumes[i]->name,
- priv->du_stats[i].chunks, loc->path);
-
- /* Maximize overlap if the bricks are all the same
- * size.
- * This is probably not going to be very common on
- * live setups but will benefit our regression tests
- */
- if (i && (priv->du_stats[i].chunks
- != priv->du_stats[0].chunks)) {
- maximize_overlap = _gf_false;
- }
- }
- } else {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_NO_DISK_USAGE_STATUS, "no du stats ?!?");
- }
-
- /* First give it a layout as though it is a new directory. This
- ensures rotation to kick in */
- dht_layout_sort_volname (new_layout);
- dht_selfheal_layout_new_directory (frame, loc, new_layout);
-
-
- /* Maximize overlap if weighted-rebalance is disabled */
- if (!priv->do_weighting)
- maximize_overlap = _gf_true;
-
- /* Now selectively re-assign ranges only when it helps */
- if (maximize_overlap) {
- dht_selfheal_layout_maximize_overlap (frame, loc, new_layout,
- layout);
- }
+ int i = 0;
+ xlator_t *this = NULL;
+ dht_layout_t *new_layout = NULL;
+ dht_conf_t *priv = NULL;
+ dht_local_t *local = NULL;
+ uint32_t subvol_down = 0;
+ gf_boolean_t maximize_overlap = _gf_true;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+
+ this = frame->this;
+ priv = this->private;
+ local = frame->local;
+
+ if (layout->type == DHT_HASH_TYPE_DM_USER) {
+ gf_msg_debug(THIS->name, 0, "leaving %s alone", loc->path);
+ goto done;
+ }
+
+ new_layout = dht_layout_new(this, priv->subvolume_cnt);
+ if (!new_layout) {
+ gf_uuid_unparse(loc->gfid, gfid);
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_MEM_ALLOC_FAILED,
+ "new_layout, path=%s", loc->path, "gfid=%s", gfid, NULL);
+ goto done;
+ }
+
+ /* If a subvolume is down, do not re-write the layout. */
+ dht_layout_anomalies(this, loc, layout, NULL, NULL, NULL, &subvol_down,
+ NULL, NULL);
+
+ if (subvol_down) {
+ gf_uuid_unparse(loc->gfid, gfid);
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_LAYOUT_FIX_FAILED,
+ "subvol-down=%u", subvol_down, "Skipping-fix-layout", "path=%s",
+ loc->path, "gfid=%s", gfid, NULL);
+ GF_FREE(new_layout);
+ return NULL;
+ }
+
+ for (i = 0; i < new_layout->cnt; i++) {
+ if (layout->list[i].err != ENOSPC)
+ new_layout->list[i].err = layout->list[i].err;
+ else
+ new_layout->list[i].err = -1;
+
+ new_layout->list[i].xlator = layout->list[i].xlator;
+ }
+
+ new_layout->commit_hash = layout->commit_hash;
+
+ if (priv->du_stats) {
+ for (i = 0; i < priv->subvolume_cnt; ++i) {
+ gf_smsg(this->name, GF_LOG_DEBUG, 0, DHT_MSG_SUBVOL_INFO,
+ "index=%d", i, "name=%s", priv->subvolumes[i]->name,
+ "chunks=%u", priv->du_stats[i].chunks, "path=%s", loc->path,
+ NULL);
+
+ /* Maximize overlap if the bricks are all the same
+ * size.
+ * This is probably not going to be very common on
+ * live setups but will benefit our regression tests
+ */
+ if (i && (priv->du_stats[i].chunks != priv->du_stats[0].chunks)) {
+ maximize_overlap = _gf_false;
+ }
+ }
+ } else {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_NO_DISK_USAGE_STATUS,
+ NULL);
+ }
+
+ /* First give it a layout as though it is a new directory. This
+ ensures rotation to kick in */
+ dht_layout_sort_volname(new_layout);
+ dht_selfheal_layout_new_directory(frame, loc, new_layout);
+
+ /* Maximize overlap if weighted-rebalance is disabled */
+ if (!priv->do_weighting)
+ maximize_overlap = _gf_true;
+
+ /* Now selectively re-assign ranges only when it helps */
+ if (maximize_overlap) {
+ dht_selfheal_layout_maximize_overlap(frame, loc, new_layout, layout);
+ }
done:
- if (new_layout) {
+ if (new_layout) {
+ /* Make sure the extra 'ref' for existing layout is removed */
+ dht_layout_unref(this, local->layout);
- /* Make sure the extra 'ref' for existing layout is removed */
- dht_layout_unref (this, local->layout);
-
- local->layout = new_layout;
- }
+ local->layout = new_layout;
+ }
- return local->layout;
+ return local->layout;
}
-
/*
* Having to call this 2x for each entry in the layout is pretty horrible, but
* that's what all of this layout-sorting nonsense gets us.
*/
-uint32_t
-dht_get_chunks_from_xl (xlator_t *parent, xlator_t *child)
+static uint32_t
+dht_get_chunks_from_xl(xlator_t *parent, xlator_t *child)
{
- dht_conf_t *priv = parent->private;
- xlator_list_t *trav;
- uint32_t index = 0;
+ dht_conf_t *priv = parent->private;
+ xlator_list_t *trav;
+ uint32_t index = 0;
- if (!priv->du_stats) {
- return 0;
- }
+ if (!priv->du_stats) {
+ return 0;
+ }
- for (trav = parent->children; trav; trav = trav->next) {
- if (trav->xlator == child) {
- return priv->du_stats[index].chunks;
- }
- ++index;
+ for (trav = parent->children; trav; trav = trav->next) {
+ if (trav->xlator == child) {
+ return priv->du_stats[index].chunks;
}
+ ++index;
+ }
- return 0;
+ return 0;
}
-
void
-dht_selfheal_layout_new_directory (call_frame_t *frame, loc_t *loc,
- dht_layout_t *layout)
+dht_selfheal_layout_new_directory(call_frame_t *frame, loc_t *loc,
+ dht_layout_t *layout)
{
- xlator_t *this = NULL;
- double chunk = 0;
- int i = 0;
- uint32_t start = 0;
- int bricks_to_use = 0;
- int err = 0;
- int start_subvol = 0;
- uint32_t curr_size;
- uint32_t range_size;
- uint64_t total_size = 0;
- int real_i;
- dht_conf_t *priv;
- gf_boolean_t weight_by_size;
- int bricks_used = 0;
-
- this = frame->this;
- priv = this->private;
- weight_by_size = priv->do_weighting;
-
- bricks_to_use = dht_get_layout_count (this, layout, 1);
- GF_ASSERT (bricks_to_use > 0);
-
- bricks_used = 0;
- for (i = 0; i < layout->cnt; ++i) {
- err = layout->list[i].err;
- if ((err != -1) && (err != ENOENT)) {
- continue;
- }
- curr_size = dht_get_chunks_from_xl (this,
- layout->list[i].xlator);
- if (!curr_size) {
- weight_by_size = _gf_false;
- break;
- }
- total_size += curr_size;
- if (++bricks_used >= bricks_to_use) {
- break;
- }
- }
-
- if (weight_by_size && total_size) {
- /* We know total_size is not zero. */
- chunk = ((double) 0xffffffff) / ((double) total_size);
- gf_msg_debug (this->name, 0,
- "chunk size = 0xffffffff / %lu = %f",
- total_size, chunk);
- }
- else {
- weight_by_size = _gf_false;
- chunk = ((unsigned long) 0xffffffff) / bricks_to_use;
+ xlator_t *this = NULL;
+ double chunk = 0;
+ int i = 0;
+ uint32_t start = 0;
+ int bricks_to_use = 0;
+ int err = 0;
+ int start_subvol = 0;
+ uint32_t curr_size;
+ uint32_t range_size;
+ uint64_t total_size = 0;
+ int real_i;
+ dht_conf_t *priv;
+ gf_boolean_t weight_by_size;
+ int bricks_used = 0;
+
+ this = frame->this;
+ priv = this->private;
+ weight_by_size = priv->do_weighting;
+
+ bricks_to_use = dht_get_layout_count(this, layout, 1);
+ GF_ASSERT(bricks_to_use > 0);
+
+ bricks_used = 0;
+ for (i = 0; i < layout->cnt; ++i) {
+ err = layout->list[i].err;
+ if ((err != -1) && (err != ENOENT)) {
+ continue;
+ }
+ curr_size = dht_get_chunks_from_xl(this, layout->list[i].xlator);
+ if (!curr_size) {
+ weight_by_size = _gf_false;
+ break;
+ }
+ total_size += curr_size;
+ if (++bricks_used >= bricks_to_use) {
+ break;
+ }
+ }
+
+ if (weight_by_size && total_size) {
+ /* We know total_size is not zero. */
+ chunk = ((double)0xffffffff) / ((double)total_size);
+ gf_msg_debug(this->name, 0,
+ "chunk size = 0xffffffff / %" PRIu64 " = %f", total_size,
+ chunk);
+ } else {
+ weight_by_size = _gf_false;
+ chunk = ((unsigned long)0xffffffff) / bricks_to_use;
+ }
+
+ start_subvol = dht_selfheal_layout_alloc_start(this, loc, layout);
+
+ /* clear out the range, as we are re-computing here */
+ DHT_RESET_LAYOUT_RANGE(layout);
+
+ /*
+ * OK, what's this "real_i" stuff about? This used to be two loops -
+ * from start_subvol to layout->cnt-1, then from 0 to start_subvol-1.
+ * That way is practically an open invitation to bugs when only one
+ * of the loops is updated. Using real_i and modulo operators to make
+ * it one loop avoids this problem. Remember, folks: it's everyone's
+ * responsibility to help stamp out copy/paste abuse.
+ */
+ bricks_used = 0;
+ for (real_i = 0; real_i < layout->cnt; real_i++) {
+ i = (real_i + start_subvol) % layout->cnt;
+ err = layout->list[i].err;
+ if ((err != -1) && (err != ENOENT)) {
+ continue;
+ }
+ if (weight_by_size) {
+ curr_size = dht_get_chunks_from_xl(this, layout->list[i].xlator);
+ if (!curr_size) {
+ continue;
+ }
+ } else {
+ curr_size = 1;
}
-
- start_subvol = dht_selfheal_layout_alloc_start (this, loc, layout);
-
- /* clear out the range, as we are re-computing here */
- DHT_RESET_LAYOUT_RANGE (layout);
-
- /*
- * OK, what's this "real_i" stuff about? This used to be two loops -
- * from start_subvol to layout->cnt-1, then from 0 to start_subvol-1.
- * That way is practically an open invitation to bugs when only one
- * of the loops is updated. Using real_i and modulo operators to make
- * it one loop avoids this problem. Remember, folks: it's everyone's
- * responsibility to help stamp out copy/paste abuse.
- */
- bricks_used = 0;
- for (real_i = 0; real_i < layout->cnt; real_i++) {
- i = (real_i + start_subvol) % layout->cnt;
- err = layout->list[i].err;
- if ((err != -1) && (err != ENOENT)) {
- continue;
- }
- if (weight_by_size) {
- curr_size = dht_get_chunks_from_xl (this,
- layout->list[i].xlator);
- if (!curr_size) {
- continue;
- }
- }
- else {
- curr_size = 1;
- }
- range_size = chunk * curr_size;
- gf_msg_debug (this->name, 0,
- "assigning range size 0x%x to %s",
- range_size,
- layout->list[i].xlator->name);
- DHT_SET_LAYOUT_RANGE(layout, i, start, range_size,
- loc->path);
- if (++bricks_used >= bricks_to_use) {
- layout->list[i].stop = 0xffffffff;
- goto done;
- }
- start += range_size;
+ range_size = chunk * curr_size;
+ gf_msg_debug(this->name, 0, "assigning range size 0x%x to %s",
+ range_size, layout->list[i].xlator->name);
+ DHT_SET_LAYOUT_RANGE(layout, i, start, range_size, loc->path);
+ if (++bricks_used >= bricks_to_use) {
+ layout->list[i].stop = 0xffffffff;
+ goto done;
}
+ start += range_size;
+ }
done:
- return;
+ return;
}
-int
-dht_selfheal_dir_getafix (call_frame_t *frame, loc_t *loc,
- dht_layout_t *layout)
+static int
+dht_selfheal_dir_getafix(call_frame_t *frame, loc_t *loc, dht_layout_t *layout)
{
- dht_local_t *local = NULL;
- uint32_t holes = 0;
- int ret = -1;
- int i = -1;
- uint32_t overlaps = 0;
-
- local = frame->local;
-
- holes = local->selfheal.hole_cnt;
- overlaps = local->selfheal.overlaps_cnt;
-
- if (holes || overlaps) {
- /* If the layout has anomolies which would change the hash
- * ranges, then we need to reset the commit_hash for this
- * directory, as the layout would change and things may not
- * be in place as expected */
- layout->commit_hash = DHT_LAYOUT_HASH_INVALID;
- dht_selfheal_layout_new_directory (frame, loc, layout);
- ret = 0;
- }
+ dht_local_t *local = NULL;
+ uint32_t holes = 0;
+ int ret = -1;
+ int i = -1;
+ uint32_t overlaps = 0;
+
+ local = frame->local;
+
+ holes = local->selfheal.hole_cnt;
+ overlaps = local->selfheal.overlaps_cnt;
+
+ if (holes || overlaps) {
+ /* If the layout has anomalies which would change the hash
+ * ranges, then we need to reset the commit_hash for this
+ * directory, as the layout would change and things may not
+ * be in place as expected */
+ layout->commit_hash = DHT_LAYOUT_HASH_INVALID;
+ dht_selfheal_layout_new_directory(frame, loc, layout);
+ ret = 0;
+ }
- for (i = 0; i < layout->cnt; i++) {
- /* directory not present */
- if (layout->list[i].err == ENOENT) {
- ret = 0;
- break;
- }
+ for (i = 0; i < layout->cnt; i++) {
+ /* directory not present */
+ if (layout->list[i].err == ENOENT) {
+ ret = 0;
+ break;
}
+ }
- /* TODO: give a fix to these non-virgins */
+ /* TODO: give a fix to these non-virgins */
- return ret;
+ return ret;
}
int
-dht_selfheal_new_directory (call_frame_t *frame,
- dht_selfheal_dir_cbk_t dir_cbk,
- dht_layout_t *layout)
+dht_selfheal_new_directory(call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
+ dht_layout_t *layout)
{
- dht_local_t *local = NULL;
- int ret = 0;
- inode_t *linked_inode = NULL, *inode = NULL;
- loc_t *loc = NULL;
- char pgfid[GF_UUID_BUF_SIZE] = {0};
- char gfid[GF_UUID_BUF_SIZE] = {0};
- int32_t op_errno = EIO;
+ dht_local_t *local = NULL;
+ int ret = 0;
+ inode_t *linked_inode = NULL, *inode = NULL;
+ loc_t *loc = NULL;
+ char pgfid[GF_UUID_BUF_SIZE] = {0};
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ int32_t op_errno = EIO;
- local = frame->local;
+ local = frame->local;
- loc = &local->loc;
+ loc = &local->loc;
- gf_uuid_unparse(local->stbuf.ia_gfid, gfid);
- gf_uuid_unparse(loc->parent->gfid, pgfid);
+ gf_uuid_unparse(local->stbuf.ia_gfid, gfid);
+ gf_uuid_unparse(loc->parent->gfid, pgfid);
- linked_inode = inode_link (loc->inode, loc->parent, loc->name,
- &local->stbuf);
- if (!linked_inode) {
- gf_msg (frame->this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DIR_SELFHEAL_FAILED,
- "linking inode failed (%s/%s) => %s",
- pgfid, loc->name, gfid);
- ret = -1;
- goto out;
- }
+ linked_inode = inode_link(loc->inode, loc->parent, loc->name,
+ &local->stbuf);
+ if (!linked_inode) {
+ gf_smsg(frame->this->name, GF_LOG_WARNING, 0, DHT_MSG_LINK_INODE_FAILED,
+ "pgfid=%s", pgfid, "name=%s", loc->name, "gfid=%s", gfid, NULL);
+ ret = -1;
+ goto out;
+ }
- inode = loc->inode;
- loc->inode = linked_inode;
- inode_unref (inode);
+ inode = loc->inode;
+ loc->inode = linked_inode;
+ inode_unref(inode);
- local->selfheal.dir_cbk = dir_cbk;
- local->selfheal.layout = dht_layout_ref (frame->this, layout);
+ local->selfheal.dir_cbk = dir_cbk;
+ local->selfheal.layout = dht_layout_ref(frame->this, layout);
- dht_layout_sort_volname (layout);
- dht_selfheal_layout_new_directory (frame, &local->loc, layout);
+ dht_layout_sort_volname(layout);
+ dht_selfheal_layout_new_directory(frame, &local->loc, layout);
- op_errno = ENOMEM;
- ret = dht_selfheal_layout_lock (frame, layout, _gf_true,
- dht_selfheal_dir_xattr,
- dht_should_heal_layout);
+ op_errno = ENOMEM;
+ ret = dht_selfheal_layout_lock(frame, layout, _gf_true,
+ dht_selfheal_dir_xattr,
+ dht_should_heal_layout);
out:
- if (ret < 0) {
- dir_cbk (frame, NULL, frame->this, -1, op_errno, NULL);
- }
+ if (ret < 0) {
+ dir_cbk(frame, NULL, frame->this, -1, op_errno, NULL);
+ }
- return 0;
+ return 0;
}
int
-dht_fix_directory_layout (call_frame_t *frame,
- dht_selfheal_dir_cbk_t dir_cbk,
- dht_layout_t *layout)
+dht_fix_directory_layout(call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
+ dht_layout_t *layout)
{
- dht_local_t *local = NULL;
- dht_layout_t *tmp_layout = NULL;
- int ret = 0;
+ dht_local_t *local = NULL;
+ dht_layout_t *tmp_layout = NULL;
+ int ret = 0;
- local = frame->local;
+ local = frame->local;
- local->selfheal.dir_cbk = dir_cbk;
- local->selfheal.layout = dht_layout_ref (frame->this, layout);
+ local->selfheal.dir_cbk = dir_cbk;
+ local->selfheal.layout = dht_layout_ref(frame->this, layout);
- /* No layout sorting required here */
- tmp_layout = dht_fix_layout_of_directory (frame, &local->loc, layout);
- if (!tmp_layout) {
- return -1;
- }
+ /* No layout sorting required here */
+ tmp_layout = dht_fix_layout_of_directory(frame, &local->loc, layout);
+ if (!tmp_layout) {
+ return -1;
+ }
- ret = dht_selfheal_layout_lock (frame, tmp_layout, _gf_false,
- dht_fix_dir_xattr,
- dht_should_fix_layout);
+ ret = dht_selfheal_layout_lock(frame, tmp_layout, _gf_false,
+ dht_fix_dir_xattr, dht_should_fix_layout);
- return ret;
+ return ret;
}
-
int
-dht_selfheal_directory (call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
- loc_t *loc, dht_layout_t *layout)
+dht_selfheal_directory(call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
+ loc_t *loc, dht_layout_t *layout)
{
- dht_local_t *local = NULL;
- uint32_t down = 0;
- uint32_t misc = 0;
- int ret = 0;
- xlator_t *this = NULL;
- char pgfid[GF_UUID_BUF_SIZE] = {0};
- char gfid[GF_UUID_BUF_SIZE] = {0};
- inode_t *linked_inode = NULL, *inode = NULL;
-
- local = frame->local;
- this = frame->this;
-
- local->selfheal.dir_cbk = dir_cbk;
- local->selfheal.layout = dht_layout_ref (this, layout);
-
- if (!__is_root_gfid (local->stbuf.ia_gfid)) {
- gf_uuid_unparse(local->stbuf.ia_gfid, gfid);
- gf_uuid_unparse(loc->parent->gfid, pgfid);
-
- linked_inode = inode_link (loc->inode, loc->parent, loc->name,
- &local->stbuf);
- if (!linked_inode) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DIR_SELFHEAL_FAILED,
- "linking inode failed (%s/%s) => %s",
- pgfid, loc->name, gfid);
- ret = 0;
- goto sorry_no_fix;
- }
-
- inode = loc->inode;
- loc->inode = linked_inode;
- inode_unref (inode);
- }
-
- dht_layout_anomalies (this, loc, layout,
- &local->selfheal.hole_cnt,
- &local->selfheal.overlaps_cnt,
- NULL, &local->selfheal.down,
- &local->selfheal.misc, NULL);
-
- down = local->selfheal.down;
- misc = local->selfheal.misc;
-
- if (down) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DIR_SELFHEAL_FAILED,
- "Directory selfheal failed: %d subvolumes down."
- "Not fixing. path = %s, gfid = %s",
- down, loc->path, gfid);
- ret = 0;
- goto sorry_no_fix;
- }
+ dht_local_t *local = NULL;
+ xlator_t *this = NULL;
+ uint32_t down = 0;
+ uint32_t misc = 0;
+ int ret = 0;
+ char pgfid[GF_UUID_BUF_SIZE] = {0};
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ inode_t *linked_inode = NULL, *inode = NULL;
- if (misc) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DIR_SELFHEAL_FAILED,
- "Directory selfheal failed : %d subvolumes "
- "have unrecoverable errors. path = %s, gfid = %s",
- misc, loc->path, gfid);
+ local = frame->local;
+ this = frame->this;
- ret = 0;
- goto sorry_no_fix;
- }
+ local->selfheal.dir_cbk = dir_cbk;
+ local->selfheal.layout = dht_layout_ref(this, layout);
- dht_layout_sort_volname (layout);
- local->heal_layout = _gf_true;
- ret = dht_selfheal_dir_getafix (frame, loc, layout);
+ if (local->need_attrheal) {
+ if (__is_root_gfid(local->stbuf.ia_gfid)) {
+ local->stbuf.ia_gid = local->prebuf.ia_gid;
+ local->stbuf.ia_uid = local->prebuf.ia_uid;
- if (ret == -1) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_DIR_SELFHEAL_FAILED,
- "Directory selfheal failed: "
- "Unable to form layout for directory %s",
- loc->path);
- local->heal_layout = _gf_false;
- }
+ local->stbuf.ia_ctime = local->prebuf.ia_ctime;
+ local->stbuf.ia_ctime_nsec = local->prebuf.ia_ctime_nsec;
+ local->stbuf.ia_prot = local->prebuf.ia_prot;
- ret = dht_selfheal_dir_mkdir (frame, loc, layout, 0);
- if (ret < 0) {
- ret = 0;
- goto sorry_no_fix;
+ } else if (!IA_ISINVAL(local->mds_stbuf.ia_type)) {
+ local->stbuf = local->mds_stbuf;
}
+ }
- return 0;
-
-sorry_no_fix:
- /* TODO: need to put appropriate local->op_errno */
- dht_selfheal_dir_finish (frame, this, ret, 1);
-
- return 0;
-}
+ if (!__is_root_gfid(local->stbuf.ia_gfid)) {
+ gf_uuid_unparse(local->stbuf.ia_gfid, gfid);
+ gf_uuid_unparse(loc->parent->gfid, pgfid);
-int
-dht_selfheal_directory_for_nameless_lookup (call_frame_t *frame,
- dht_selfheal_dir_cbk_t dir_cbk,
- loc_t *loc, dht_layout_t *layout)
-{
- dht_local_t *local = NULL;
- uint32_t down = 0;
- uint32_t misc = 0;
- int ret = 0;
- xlator_t *this = NULL;
-
- local = frame->local;
- this = frame->this;
- dht_layout_anomalies (this, loc, layout,
- &local->selfheal.hole_cnt,
- &local->selfheal.overlaps_cnt,
- NULL, &local->selfheal.down,
- &local->selfheal.misc, NULL);
-
- down = local->selfheal.down;
- misc = local->selfheal.misc;
-
- local->selfheal.dir_cbk = dir_cbk;
- local->selfheal.layout = dht_layout_ref (this, layout);
-
- if (down) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_SUBVOL_DOWN_ERROR,
- "%d subvolumes down -- not fixing", down);
- ret = 0;
- goto sorry_no_fix;
+ linked_inode = inode_link(loc->inode, loc->parent, loc->name,
+ &local->stbuf);
+ if (!linked_inode) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_LINK_INODE_FAILED,
+ "pgfid=%s", pgfid, "name=%s", loc->name, "gfid=%s", gfid,
+ NULL);
+ ret = 0;
+ goto sorry_no_fix;
}
- if (misc) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_SUBVOL_ERROR,
- "%d subvolumes have unrecoverable errors", misc);
- ret = 0;
- goto sorry_no_fix;
- }
+ inode = loc->inode;
+ loc->inode = linked_inode;
+ inode_unref(inode);
+ }
+
+ if (local->need_xattr_heal && (local->mds_xattr)) {
+ dht_dir_set_heal_xattr(this, local, local->xattr, local->mds_xattr,
+ NULL, NULL);
+ dict_unref(local->mds_xattr);
+ local->mds_xattr = NULL;
+ }
+
+ dht_layout_anomalies(this, loc, layout, &local->selfheal.hole_cnt,
+ &local->selfheal.overlaps_cnt,
+ &local->selfheal.missing_cnt, &local->selfheal.down,
+ &local->selfheal.misc, NULL);
+
+ down = local->selfheal.down;
+ misc = local->selfheal.misc;
+
+ if (down) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_SELFHEAL_FAILED,
+ "path=%s", loc->path, "subvol-down=%d", down, "Not-fixing",
+ "gfid=%s", gfid, NULL);
+ ret = 0;
+ goto sorry_no_fix;
+ }
- dht_layout_sort_volname (layout);
- ret = dht_selfheal_dir_getafix (frame, loc, layout);
+ if (misc) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_SELFHEAL_FAILED,
+ "path=%s", loc->path, "misc=%d", misc, "unrecoverable-errors",
+ "gfid=%s", gfid, NULL);
- if (ret == -1) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_LAYOUT_FORM_FAILED,
- "not able to form layout for the directory");
- goto sorry_no_fix;
- }
+ ret = 0;
+ goto sorry_no_fix;
+ }
- ret = dht_selfheal_layout_lock (frame, layout, _gf_false,
- dht_selfheal_dir_xattr_for_nameless_lookup,
- dht_should_heal_layout);
+ dht_layout_sort_volname(layout);
+ local->heal_layout = _gf_true;
- if (ret < 0) {
- goto sorry_no_fix;
- }
+ /* Ignore return value as it can be inferred from result of
+ * dht_layout_anomalies
+ */
+ dht_selfheal_dir_getafix(frame, loc, layout);
- return 0;
+ if (!(local->selfheal.hole_cnt || local->selfheal.overlaps_cnt ||
+ local->selfheal.missing_cnt)) {
+ local->heal_layout = _gf_false;
+ }
-sorry_no_fix:
- /* TODO: need to put appropriate local->op_errno */
- dht_selfheal_dir_finish (frame, this, ret, 1);
+ ret = dht_selfheal_dir_mkdir(frame, loc, layout, 0);
+ if (ret < 0) {
+ ret = 0;
+ goto sorry_no_fix;
+ }
- return 0;
+ return 0;
+sorry_no_fix:
+ /* TODO: need to put appropriate local->op_errno */
+ dht_selfheal_dir_finish(frame, this, ret, 1);
+ return 0;
}
int
-dht_selfheal_restore (call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
- loc_t *loc, dht_layout_t *layout)
+dht_selfheal_restore(call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
+ loc_t *loc, dht_layout_t *layout)
{
- int ret = 0;
- dht_local_t *local = NULL;
+ int ret = 0;
+ dht_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- local->selfheal.dir_cbk = dir_cbk;
- local->selfheal.layout = dht_layout_ref (frame->this, layout);
+ local->selfheal.dir_cbk = dir_cbk;
+ local->selfheal.layout = dht_layout_ref(frame->this, layout);
- ret = dht_selfheal_dir_mkdir (frame, loc, layout, 1);
+ ret = dht_selfheal_dir_mkdir(frame, loc, layout, 1);
- return ret;
+ return ret;
}
int
-dht_dir_heal_xattrs (void *data)
+dht_dir_heal_xattrs(void *data)
{
- call_frame_t *frame = NULL;
- dht_local_t *local = NULL;
- xlator_t *subvol = NULL;
- xlator_t *mds_subvol = NULL;
- xlator_t *this = NULL;
- dht_conf_t *conf = NULL;
- dict_t *user_xattr = NULL;
- dict_t *internal_xattr = NULL;
- dict_t *mds_xattr = NULL;
- dict_t *xdata = NULL;
- int call_cnt = 0;
- int ret = -1;
- int uret = 0;
- int uflag = 0;
- int i = 0;
- int xattr_hashed = 0;
- char gfid[GF_UUID_BUF_SIZE] = {0};
- int32_t allzero[1] = {0};
-
- GF_VALIDATE_OR_GOTO ("dht", data, out);
-
- frame = data;
- local = frame->local;
- this = frame->this;
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO (this->name, local, out);
- mds_subvol = local->mds_subvol;
- conf = this->private;
- GF_VALIDATE_OR_GOTO (this->name, conf, out);
- gf_uuid_unparse(local->loc.gfid, gfid);
-
- if (!mds_subvol) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DIR_XATTR_HEAL_FAILED,
- "No mds subvol for %s gfid = %s",
- local->loc.path, gfid);
- goto out;
- }
-
- if ((local->loc.inode && gf_uuid_is_null (local->loc.inode->gfid)) ||
- gf_uuid_is_null (local->loc.gfid)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DIR_XATTR_HEAL_FAILED,
- "No gfid present so skip heal for path %s gfid = %s",
- local->loc.path, gfid);
- goto out;
- }
-
- internal_xattr = dict_new ();
- if (!internal_xattr) {
- gf_msg (this->name, GF_LOG_ERROR, DHT_MSG_NO_MEMORY, 0,
- "dictionary creation failed");
- goto out;
- }
- xdata = dict_new ();
- if (!xdata) {
- gf_msg (this->name, GF_LOG_ERROR, DHT_MSG_NO_MEMORY, 0,
- "dictionary creation failed");
- goto out;
- }
-
- call_cnt = conf->subvolume_cnt;
-
- user_xattr = dict_new ();
- if (!user_xattr) {
- gf_msg (this->name, GF_LOG_ERROR, DHT_MSG_NO_MEMORY, 0,
- "dictionary creation failed");
+ call_frame_t *frame = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+ xlator_t *mds_subvol = NULL;
+ xlator_t *this = NULL;
+ dht_conf_t *conf = NULL;
+ dict_t *user_xattr = NULL;
+ dict_t *internal_xattr = NULL;
+ dict_t *mds_xattr = NULL;
+ dict_t *xdata = NULL;
+ int call_cnt = 0;
+ int ret = -1;
+ int uret = 0;
+ int uflag = 0;
+ int i = 0;
+ int xattr_hashed = 0;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ int32_t allzero[1] = {0};
+
+ GF_VALIDATE_OR_GOTO("dht", data, out);
+
+ frame = data;
+ local = frame->local;
+ this = frame->this;
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, local, out);
+ mds_subvol = local->mds_subvol;
+ conf = this->private;
+ GF_VALIDATE_OR_GOTO(this->name, conf, out);
+ gf_uuid_unparse(local->loc.gfid, gfid);
+
+ if (!mds_subvol) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_NO_MDS_SUBVOL, "path=%s",
+ local->loc.path, "gfid=%s", gfid, NULL);
+ goto out;
+ }
+
+ if ((local->loc.inode && gf_uuid_is_null(local->loc.inode->gfid)) ||
+ gf_uuid_is_null(local->loc.gfid)) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_GFID_NOT_PRESENT,
+ "skip-heal path=%s", local->loc.path, "gfid=%s", gfid, NULL);
+ goto out;
+ }
+
+ internal_xattr = dict_new();
+ if (!internal_xattr) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_CREATE_FAILED,
+ "dictionary", NULL);
+ goto out;
+ }
+ xdata = dict_new();
+ if (!xdata) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_CREATE_FAILED,
+ "dictionary", NULL);
+ goto out;
+ }
+
+ call_cnt = conf->subvolume_cnt;
+
+ user_xattr = dict_new();
+ if (!user_xattr) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_CREATE_FAILED,
+ "dictionary", NULL);
+ goto out;
+ }
+
+ ret = syncop_listxattr(local->mds_subvol, &local->loc, &mds_xattr, NULL,
+ NULL);
+ if (ret < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret, DHT_MSG_LIST_XATTRS_FAILED,
+ "path=%s", local->loc.path, "name=%s", local->mds_subvol->name,
+ NULL);
+ }
+
+ if (!mds_xattr)
+ goto out;
+
+ dht_dir_set_heal_xattr(this, local, user_xattr, mds_xattr, &uret, &uflag);
+
+ /* To set quota related xattr need to set GLUSTERFS_INTERNAL_FOP_KEY
+ * key value to 1
+ */
+ if (dict_get(user_xattr, QUOTA_LIMIT_KEY) ||
+ dict_get(user_xattr, QUOTA_LIMIT_OBJECTS_KEY)) {
+ ret = dict_set_int32(xdata, GLUSTERFS_INTERNAL_FOP_KEY, 1);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "key=%s", GLUSTERFS_INTERNAL_FOP_KEY, "path=%s",
+ local->loc.path, NULL);
+ goto out;
+ }
+ }
+ if (uret <= 0 && !uflag)
+ goto out;
+
+ for (i = 0; i < call_cnt; i++) {
+ subvol = conf->subvolumes[i];
+ if (subvol == mds_subvol)
+ continue;
+ if (uret || uflag) {
+ /* Custom xattr heal is required - let posix handle it */
+ ret = dict_set_int8(xdata, "sync_backend_xattrs", _gf_true);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_DICT_SET_FAILED,
+ "path=%s", local->loc.path, "key=%s",
+ "sync_backend_xattrs", NULL);
goto out;
- }
+ }
- ret = syncop_listxattr (local->mds_subvol, &local->loc,
- &mds_xattr, NULL, NULL);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
+ ret = syncop_setxattr(subvol, &local->loc, user_xattr, 0, xdata,
+ NULL);
+ if (ret) {
+ xattr_hashed = 1;
+ gf_smsg(this->name, GF_LOG_ERROR, -ret,
DHT_MSG_DIR_XATTR_HEAL_FAILED,
- "failed to list xattrs for "
- "%s: on %s ",
- local->loc.path, local->mds_subvol->name);
- }
-
- if (!mds_xattr)
- goto out;
-
- dht_dir_set_heal_xattr (this, local, user_xattr, mds_xattr,
- &uret, &uflag);
-
- /* To set quota related xattr need to set GLUSTERFS_INTERNAL_FOP_KEY
- * key value to 1
- */
- if (dict_get (user_xattr, QUOTA_LIMIT_KEY) ||
- dict_get (user_xattr, QUOTA_LIMIT_OBJECTS_KEY)) {
- ret = dict_set_int32 (xdata, GLUSTERFS_INTERNAL_FOP_KEY, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value: key = %s,"
- " path = %s", GLUSTERFS_INTERNAL_FOP_KEY,
- local->loc.path);
- goto out;
- }
- }
- if (uret <= 0 && !uflag)
- goto out;
-
- for (i = 0; i < call_cnt; i++) {
- subvol = conf->subvolumes[i];
- if (subvol == mds_subvol)
- continue;
- if (uret || uflag) {
- ret = syncop_setxattr (subvol, &local->loc, user_xattr,
- 0, xdata, NULL);
- if (ret) {
- xattr_hashed = 1;
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_DIR_XATTR_HEAL_FAILED,
- "Directory xattr heal failed. Failed to set"
- "user xattr on path %s on "
- "subvol %s, gfid = %s ",
- local->loc.path, subvol->name, gfid);
- }
- }
+ "set-user-xattr-failed path=%s", local->loc.path,
+ "subvol=%s", subvol->name, "gfid=%s", gfid, NULL);
+ } else {
+ dict_del(xdata, "sync_backend_xattrs");
+ }
+ }
+ }
+ /* After heal all custom xattr reset internal MDS xattr to 0 */
+ if (!xattr_hashed) {
+ ret = dht_dict_set_array(internal_xattr, conf->mds_xattr_key, allzero,
+ 1);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_WARNING, ENOMEM, DHT_MSG_DICT_SET_FAILED,
+ "key=%s", conf->mds_xattr_key, "path=%s", local->loc.path,
+ NULL);
+ goto out;
}
- /* After heal all custom xattr reset internal MDS xattr to 0 */
- if (!xattr_hashed) {
- ret = dht_dict_set_array (internal_xattr,
- conf->mds_xattr_key,
- allzero, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value:key = %s for "
- "path %s", conf->mds_xattr_key,
- local->loc.path);
- goto out;
- }
- ret = syncop_setxattr (mds_subvol, &local->loc, internal_xattr,
- 0, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_DIR_XATTR_HEAL_FAILED,
- "Failed to reset internal xattr "
- "on path %s on subvol %s"
- "gfid = %s ", local->loc.path,
- mds_subvol->name, gfid);
- }
+ ret = syncop_setxattr(mds_subvol, &local->loc, internal_xattr, 0, NULL,
+ NULL);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret,
+ DHT_MSG_DIR_XATTR_HEAL_FAILED, "path=%s", local->loc.path,
+ "subvol=%s", mds_subvol->name, "gfid=%s", gfid, NULL);
}
+ }
out:
- if (user_xattr)
- dict_unref (user_xattr);
- if (mds_xattr)
- dict_unref (mds_xattr);
- if (internal_xattr)
- dict_unref (internal_xattr);
- if (xdata)
- dict_unref (xdata);
- return 0;
+ if (user_xattr)
+ dict_unref(user_xattr);
+ if (mds_xattr)
+ dict_unref(mds_xattr);
+ if (internal_xattr)
+ dict_unref(internal_xattr);
+ if (xdata)
+ dict_unref(xdata);
+ return 0;
}
-
int
-dht_dir_heal_xattrs_done (int ret, call_frame_t *sync_frame, void *data)
+dht_dir_heal_xattrs_done(int ret, call_frame_t *sync_frame, void *data)
{
- DHT_STACK_DESTROY (sync_frame);
- return 0;
+ DHT_STACK_DESTROY(sync_frame);
+ return 0;
}
-
int
-dht_dir_attr_heal (void *data)
+dht_dir_attr_heal(void *data)
{
- call_frame_t *frame = NULL;
- dht_local_t *local = NULL;
- xlator_t *subvol = NULL;
- xlator_t *mds_subvol = NULL;
- xlator_t *this = NULL;
- dht_conf_t *conf = NULL;
- int call_cnt = 0;
- int ret = -1;
- int i = 0;
- char gfid[GF_UUID_BUF_SIZE] = {0};
-
-
- GF_VALIDATE_OR_GOTO ("dht", data, out);
-
- frame = data;
- local = frame->local;
- mds_subvol = local->mds_subvol;
- this = frame->this;
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", local, out);
- conf = this->private;
- GF_VALIDATE_OR_GOTO ("dht", conf, out);
-
- call_cnt = conf->subvolume_cnt;
-
- if (!__is_root_gfid (local->stbuf.ia_gfid) && (!mds_subvol)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DIR_ATTR_HEAL_FAILED,
- "No mds subvol for %s gfid = %s",
- local->loc.path, gfid);
- goto out;
- }
-
- if (!__is_root_gfid (local->stbuf.ia_gfid)) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->subvolumes[i] == mds_subvol) {
- if (!conf->subvolume_status[i]) {
- gf_msg (this->name, GF_LOG_ERROR,
- 0, DHT_MSG_HASHED_SUBVOL_DOWN,
- "mds subvol is down for path "
- " %s gfid is %s Unable to set xattr " ,
- local->loc.path, gfid);
- goto out;
- }
- }
- }
+ call_frame_t *frame = NULL;
+ dht_local_t *local = NULL;
+ xlator_t *subvol = NULL;
+ xlator_t *mds_subvol = NULL;
+ xlator_t *this = NULL;
+ dht_conf_t *conf = NULL;
+ int call_cnt = 0;
+ int ret = -1;
+ int i = 0;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+
+ GF_VALIDATE_OR_GOTO("dht", data, out);
+
+ frame = data;
+ local = frame->local;
+ this = frame->this;
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", local, out);
+ conf = this->private;
+ GF_VALIDATE_OR_GOTO("dht", conf, out);
+
+ mds_subvol = local->mds_subvol;
+ call_cnt = conf->subvolume_cnt;
+
+ if (!__is_root_gfid(local->stbuf.ia_gfid) && (!mds_subvol)) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_NO_MDS_SUBVOL, "path=%s",
+ local->loc.path, "gfid=%s", gfid, NULL);
+ goto out;
+ }
+
+ if (!__is_root_gfid(local->stbuf.ia_gfid)) {
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->subvolumes[i] == mds_subvol) {
+ if (!conf->subvolume_status[i]) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ DHT_MSG_MDS_DOWN_UNABLE_TO_SET, "path=%s",
+ local->loc.path, "gfid=%s", gfid, NULL);
+ goto out;
+ }
+ }
+ }
+ }
+
+ for (i = 0; i < call_cnt; i++) {
+ subvol = conf->subvolumes[i];
+ if (!subvol || subvol == mds_subvol)
+ continue;
+ if (__is_root_gfid(local->stbuf.ia_gfid)) {
+ ret = syncop_setattr(
+ subvol, &local->loc, &local->stbuf,
+ (GF_SET_ATTR_UID | GF_SET_ATTR_GID | GF_SET_ATTR_MODE), NULL,
+ NULL, NULL, NULL);
+ } else {
+ ret = syncop_setattr(
+ subvol, &local->loc, &local->mds_stbuf,
+ (GF_SET_ATTR_UID | GF_SET_ATTR_GID | GF_SET_ATTR_MODE), NULL,
+ NULL, NULL, NULL);
}
- for (i = 0; i < call_cnt; i++) {
- subvol = conf->subvolumes[i];
- if (!subvol || subvol == mds_subvol)
- continue;
- if (__is_root_gfid (local->stbuf.ia_gfid)) {
- ret = syncop_setattr (subvol, &local->loc, &local->stbuf,
- (GF_SET_ATTR_UID | GF_SET_ATTR_GID | GF_SET_ATTR_MODE),
- NULL, NULL, NULL, NULL);
- } else {
- ret = syncop_setattr (subvol, &local->loc, &local->mds_stbuf,
- (GF_SET_ATTR_UID | GF_SET_ATTR_GID),
- NULL, NULL, NULL, NULL);
- }
-
- if (ret) {
- gf_uuid_unparse(local->loc.gfid, gfid);
+ if (ret) {
+ gf_uuid_unparse(local->loc.gfid, gfid);
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_DIR_ATTR_HEAL_FAILED,
- "Directory attr heal failed. Failed to set"
- " uid/gid on path %s on subvol %s, gfid = %s ",
- local->loc.path, subvol->name, gfid);
- }
+ gf_smsg(this->name, GF_LOG_ERROR, -ret,
+ DHT_MSG_DIR_ATTR_HEAL_FAILED, "path=%s", local->loc.path,
+ "subvol=%s", subvol->name, "gfid=%s", gfid, NULL);
}
+ }
out:
- return 0;
+ return 0;
}
int
-dht_dir_attr_heal_done (int ret, call_frame_t *sync_frame, void *data)
+dht_dir_attr_heal_done(int ret, call_frame_t *sync_frame, void *data)
{
- DHT_STACK_DESTROY (sync_frame);
- return 0;
+ DHT_STACK_DESTROY(sync_frame);
+ return 0;
}
/* EXIT: dht_update_commit_hash_for_layout */
-int
-dht_update_commit_hash_for_layout_done (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xdata)
+static int
+dht_update_commit_hash_for_layout_done(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
+ dht_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- /* preserve oldest error */
- if (op_ret && !local->op_ret) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- }
+ /* preserve oldest error */
+ if (op_ret && !local->op_ret) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ }
- DHT_STACK_UNWIND (setxattr, frame, local->op_ret,
- local->op_errno, NULL);
+ DHT_STACK_UNWIND(setxattr, frame, local->op_ret, local->op_errno, NULL);
- return 0;
+ return 0;
}
-int
-dht_update_commit_hash_for_layout_unlock (call_frame_t *frame, xlator_t *this)
+static int
+dht_update_commit_hash_for_layout_unlock(call_frame_t *frame, xlator_t *this)
{
- dht_local_t *local = NULL;
- int ret = 0;
+ dht_local_t *local = NULL;
+ int ret = 0;
- local = frame->local;
+ local = frame->local;
- ret = dht_unlock_inodelk (frame, local->lock[0].layout.my_layout.locks,
- local->lock[0].layout.my_layout.lk_count,
- dht_update_commit_hash_for_layout_done);
- if (ret < 0) {
- /* preserve oldest error, just ... */
- if (!local->op_ret) {
- local->op_errno = errno;
- local->op_ret = -1;
- }
+ ret = dht_unlock_inodelk(frame, local->lock[0].layout.my_layout.locks,
+ local->lock[0].layout.my_layout.lk_count,
+ dht_update_commit_hash_for_layout_done);
+ if (ret < 0) {
+ /* preserve oldest error, just ... */
+ if (!local->op_ret) {
+ local->op_errno = errno;
+ local->op_ret = -1;
+ }
- gf_msg (this->name, GF_LOG_WARNING, errno,
- DHT_MSG_DIR_SELFHEAL_XATTR_FAILED,
- "Winding unlock failed: stale locks left on brick"
- " %s", local->loc.path);
+ gf_smsg(this->name, GF_LOG_WARNING, errno, DHT_MSG_WIND_UNLOCK_FAILED,
+ "path=%s", local->loc.path, NULL);
- dht_update_commit_hash_for_layout_done (frame, NULL, this,
- 0, 0, NULL);
- }
+ dht_update_commit_hash_for_layout_done(frame, NULL, this, 0, 0, NULL);
+ }
- return 0;
+ return 0;
}
-int
-dht_update_commit_hash_for_layout_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret,
- int op_errno, dict_t *xdata)
+static int
+dht_update_commit_hash_for_layout_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int op_ret, int op_errno,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
+ dht_local_t *local = NULL;
+ int this_call_cnt = 0;
- local = frame->local;
+ local = frame->local;
- LOCK (&frame->lock);
- /* store first failure, just because */
- if (op_ret && !local->op_ret) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- }
- UNLOCK (&frame->lock);
+ LOCK(&frame->lock);
+ /* store first failure, just because */
+ if (op_ret && !local->op_ret) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ }
+ UNLOCK(&frame->lock);
- this_call_cnt = dht_frame_return (frame);
+ this_call_cnt = dht_frame_return(frame);
- if (is_last_call (this_call_cnt)) {
- dht_update_commit_hash_for_layout_unlock (frame, this);
- }
+ if (is_last_call(this_call_cnt)) {
+ dht_update_commit_hash_for_layout_unlock(frame, this);
+ }
- return 0;
+ return 0;
}
-int
-dht_update_commit_hash_for_layout_resume (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
+static int
+dht_update_commit_hash_for_layout_resume(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
- dht_local_t *local = NULL;
- int count = 1, ret = -1, i = 0, j = 0;
- dht_conf_t *conf = NULL;
- dht_layout_t *layout = NULL;
- int32_t *disk_layout = NULL;
- dict_t **xattr = NULL;
-
- local = frame->local;
- conf = frame->this->private;
- count = conf->local_subvols_cnt;
- layout = local->layout;
-
- if (op_ret < 0) {
- goto err_done;
- }
-
- /* We precreate the xattr list as we cannot change call count post the
- * first wind as we may never continue from there. So we finish prep
- * work before winding the setxattrs */
- xattr = GF_CALLOC (count, sizeof (*xattr), gf_common_mt_char);
- if (!xattr) {
- local->op_errno = errno;
+ dht_local_t *local = NULL;
+ int count = 1, ret = -1, i = 0, j = 0;
+ dht_conf_t *conf = NULL;
+ dht_layout_t *layout = NULL;
+ int32_t *disk_layout = NULL;
+ dict_t **xattr = NULL;
+
+ local = frame->local;
+ conf = frame->this->private;
+ count = conf->local_subvols_cnt;
+ layout = local->layout;
+
+ if (op_ret < 0) {
+ goto err_done;
+ }
+
+ /* We precreate the xattr list as we cannot change call count post the
+ * first wind as we may never continue from there. So we finish prep
+ * work before winding the setxattrs */
+ xattr = GF_CALLOC(count, sizeof(*xattr), gf_common_mt_char);
+ if (!xattr) {
+ local->op_errno = errno;
+
+ gf_smsg(this->name, GF_LOG_WARNING, errno, DHT_MSG_COMMIT_HASH_FAILED,
+ "allocation-failed path=%s", local->loc.path, NULL);
+
+ goto err;
+ }
+
+ for (i = 0; i < count; i++) {
+ /* find the layout index for the subvolume */
+ ret = dht_layout_index_for_subvol(layout, conf->local_subvols[i]);
+ if (ret < 0) {
+ local->op_errno = ENOENT;
- gf_msg (this->name, GF_LOG_WARNING, errno,
- DHT_MSG_DIR_SELFHEAL_XATTR_FAILED,
- "Directory commit hash update failed:"
- " %s: Allocation failed", local->loc.path);
+ gf_smsg(this->name, GF_LOG_WARNING, 0, DHT_MSG_COMMIT_HASH_FAILED,
+ "path=%s", local->loc.path, "subvol=%s",
+ conf->local_subvols[i]->name, "find-disk-layout-failed",
+ NULL);
- goto err;
+ goto err;
}
+ j = ret;
- for (i = 0; i < count; i++) {
- /* find the layout index for the subvolume */
- ret = dht_layout_index_for_subvol (layout,
- conf->local_subvols[i]);
- if (ret < 0) {
- local->op_errno = ENOENT;
-
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DIR_SELFHEAL_XATTR_FAILED,
- "Directory commit hash update failed:"
- " %s: (subvol %s) Failed to find disk layout",
- local->loc.path, conf->local_subvols[i]->name);
-
- goto err;
- }
- j = ret;
-
- /* update the commit hash for the layout */
- layout->list[j].commit_hash = layout->commit_hash;
+ /* update the commit hash for the layout */
+ layout->list[j].commit_hash = layout->commit_hash;
- /* extract the current layout */
- ret = dht_disk_layout_extract (this, layout, j, &disk_layout);
- if (ret == -1) {
- local->op_errno = errno;
-
- gf_msg (this->name, GF_LOG_WARNING, errno,
- DHT_MSG_DIR_SELFHEAL_XATTR_FAILED,
- "Directory commit hash update failed:"
- " %s: (subvol %s) Failed to extract disk"
- " layout", local->loc.path,
- conf->local_subvols[i]->name);
+ /* extract the current layout */
+ ret = dht_disk_layout_extract(this, layout, j, &disk_layout);
+ if (ret == -1) {
+ local->op_errno = errno;
- goto err;
- }
+ gf_smsg(this->name, GF_LOG_WARNING, errno,
+ DHT_MSG_COMMIT_HASH_FAILED, "path=%s", local->loc.path,
+ "subvol=%s", conf->local_subvols[i]->name,
+ "extract-disk-layout-failed", NULL);
- xattr[i] = dict_new ();
- if (!xattr[i]) {
- local->op_errno = errno;
+ goto err;
+ }
- gf_msg (this->name, GF_LOG_WARNING, errno,
- DHT_MSG_DIR_SELFHEAL_XATTR_FAILED,
- "Directory commit hash update failed:"
- " %s: Allocation failed", local->loc.path);
+ xattr[i] = dict_new();
+ if (!xattr[i]) {
+ local->op_errno = errno;
- goto err;
- }
+ gf_smsg(this->name, GF_LOG_WARNING, errno,
+ DHT_MSG_COMMIT_HASH_FAILED, "path=%s Allocation-failed",
+ local->loc.path, NULL);
- ret = dict_set_bin (xattr[i], conf->xattr_name,
- disk_layout, 4 * 4);
- if (ret != 0) {
- local->op_errno = ENOMEM;
+ goto err;
+ }
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DIR_SELFHEAL_XATTR_FAILED,
- "Directory self heal xattr failed:"
- "%s: (subvol %s) Failed to set xattr"
- " dictionary,", local->loc.path,
- conf->local_subvols[i]->name);
+ ret = dict_set_bin(xattr[i], conf->xattr_name, disk_layout, 4 * 4);
+ if (ret != 0) {
+ local->op_errno = ENOMEM;
- goto err;
- }
- disk_layout = NULL;
+ gf_smsg(this->name, GF_LOG_WARNING, 0,
+ DHT_MSG_DIR_SELFHEAL_XATTR_FAILED, "path=%s",
+ local->loc.path, "subvol=%s", conf->local_subvols[i]->name,
+ "set-xattr-failed", NULL);
- gf_msg_trace (this->name, 0,
- "setting commit hash %u on subvolume %s"
- " for %s", layout->list[j].commit_hash,
- conf->local_subvols[i]->name, local->loc.path);
+ goto err;
}
+ disk_layout = NULL;
- /* wind the setting of the commit hash across the local subvols */
- local->call_cnt = count;
- local->op_ret = 0;
- local->op_errno = 0;
+ gf_msg_trace(this->name, 0,
+ "setting commit hash %u on subvolume %s"
+ " for %s",
+ layout->list[j].commit_hash, conf->local_subvols[i]->name,
+ local->loc.path);
+ }
+
+ /* wind the setting of the commit hash across the local subvols */
+ local->call_cnt = count;
+ local->op_ret = 0;
+ local->op_errno = 0;
+ for (i = 0; i < count; i++) {
+ STACK_WIND(frame, dht_update_commit_hash_for_layout_cbk,
+ conf->local_subvols[i],
+ conf->local_subvols[i]->fops->setxattr, &local->loc,
+ xattr[i], 0, NULL);
+ }
+ for (i = 0; i < count; i++)
+ dict_unref(xattr[i]);
+ GF_FREE(xattr);
+
+ return 0;
+err:
+ if (xattr) {
for (i = 0; i < count; i++) {
- STACK_WIND (frame, dht_update_commit_hash_for_layout_cbk,
- conf->local_subvols[i],
- conf->local_subvols[i]->fops->setxattr,
- &local->loc, xattr[i], 0, NULL);
-
+ if (xattr[i])
+ dict_unref(xattr[i]);
}
- for (i = 0; i < count; i++)
- dict_unref (xattr[i]);
- GF_FREE (xattr);
-
- return 0;
-err:
- if (xattr) {
- for (i = 0; i < count; i++) {
- if (xattr[i])
- dict_unref (xattr[i]);
- }
- GF_FREE (xattr);
- }
+ GF_FREE(xattr);
+ }
- GF_FREE (disk_layout);
+ GF_FREE(disk_layout);
- local->op_ret = -1;
+ local->op_ret = -1;
- dht_update_commit_hash_for_layout_unlock (frame, this);
+ dht_update_commit_hash_for_layout_unlock(frame, this);
- return 0;
+ return 0;
err_done:
- local->op_ret = -1;
+ local->op_ret = -1;
- dht_update_commit_hash_for_layout_done (frame, NULL, this, 0, 0, NULL);
+ dht_update_commit_hash_for_layout_done(frame, NULL, this, 0, 0, NULL);
- return 0;
+ return 0;
}
/* ENTER: dht_update_commit_hash_for_layout (see EXIT above)
@@ -2941,55 +2549,52 @@ err_done:
* - Unlock and return.
*/
int
-dht_update_commit_hash_for_layout (call_frame_t *frame)
+dht_update_commit_hash_for_layout(call_frame_t *frame)
{
- dht_local_t *local = NULL;
- int count = 1, ret = -1, i = 0;
- dht_lock_t **lk_array = NULL;
- dht_conf_t *conf = NULL;
-
- GF_VALIDATE_OR_GOTO ("dht", frame, err);
- GF_VALIDATE_OR_GOTO (frame->this->name, frame->local, err);
-
- local = frame->local;
- conf = frame->this->private;
-
- if (!conf->defrag)
- goto err;
-
- count = conf->local_subvols_cnt;
- lk_array = GF_CALLOC (count, sizeof (*lk_array),
- gf_common_mt_char);
- if (lk_array == NULL)
- goto err;
-
- for (i = 0; i < count; i++) {
- lk_array[i] = dht_lock_new (frame->this,
- conf->local_subvols[i],
- &local->loc, F_WRLCK,
- DHT_LAYOUT_HEAL_DOMAIN, NULL,
- FAIL_ON_ANY_ERROR);
- if (lk_array[i] == NULL)
- goto err;
- }
-
- local->lock[0].layout.my_layout.locks = lk_array;
- local->lock[0].layout.my_layout.lk_count = count;
-
- ret = dht_blocking_inodelk (frame, lk_array, count,
- dht_update_commit_hash_for_layout_resume);
- if (ret < 0) {
- local->lock[0].layout.my_layout.locks = NULL;
- local->lock[0].layout.my_layout.lk_count = 0;
- goto err;
- }
-
- return 0;
+ dht_local_t *local = NULL;
+ int count = 1, ret = -1, i = 0;
+ dht_lock_t **lk_array = NULL;
+ dht_conf_t *conf = NULL;
+
+ GF_VALIDATE_OR_GOTO("dht", frame, err);
+ GF_VALIDATE_OR_GOTO(frame->this->name, frame->local, err);
+
+ local = frame->local;
+ conf = frame->this->private;
+
+ if (!conf->defrag)
+ goto err;
+
+ count = conf->local_subvols_cnt;
+ lk_array = GF_CALLOC(count, sizeof(*lk_array), gf_common_mt_char);
+ if (lk_array == NULL)
+ goto err;
+
+ for (i = 0; i < count; i++) {
+ lk_array[i] = dht_lock_new(frame->this, conf->local_subvols[i],
+ &local->loc, F_WRLCK, DHT_LAYOUT_HEAL_DOMAIN,
+ NULL, FAIL_ON_ANY_ERROR);
+ if (lk_array[i] == NULL)
+ goto err;
+ }
+
+ local->lock[0].layout.my_layout.locks = lk_array;
+ local->lock[0].layout.my_layout.lk_count = count;
+
+ ret = dht_blocking_inodelk(frame, lk_array, count,
+ dht_update_commit_hash_for_layout_resume);
+ if (ret < 0) {
+ local->lock[0].layout.my_layout.locks = NULL;
+ local->lock[0].layout.my_layout.lk_count = 0;
+ goto err;
+ }
+
+ return 0;
err:
- if (lk_array != NULL) {
- dht_lock_array_free (lk_array, count);
- GF_FREE (lk_array);
- }
+ if (lk_array != NULL) {
+ dht_lock_array_free(lk_array, count);
+ GF_FREE(lk_array);
+ }
- return -1;
+ return -1;
}
diff --git a/xlators/cluster/dht/src/dht-shared.c b/xlators/cluster/dht/src/dht-shared.c
index b5f790b111b..bb72b0ffbb5 100644
--- a/xlators/cluster/dht/src/dht-shared.c
+++ b/xlators/cluster/dht/src/dht-shared.c
@@ -8,1231 +8,1097 @@
cases as published by the Free Software Foundation.
*/
-
/* TODO: add NS locking */
-#include "statedump.h"
+#include <glusterfs/statedump.h>
#include "dht-common.h"
#include "dht-messages.h"
#ifndef MAX
-#define MAX(a, b) (((a) > (b))?(a):(b))
+#define MAX(a, b) (((a) > (b)) ? (a) : (b))
#endif
-#define GF_DECIDE_DEFRAG_THROTTLE_COUNT(throttle_count, conf) { \
- \
- pthread_mutex_lock (&conf->defrag->dfq_mutex); \
- \
- if (!strcasecmp (conf->dthrottle, "lazy")) \
- conf->defrag->recon_thread_count = 1; \
- \
- throttle_count = \
- MAX ((sysconf(_SC_NPROCESSORS_ONLN) - 4), 4); \
- \
- if (!strcasecmp (conf->dthrottle, "normal")) \
- conf->defrag->recon_thread_count = \
- (throttle_count / 2); \
- \
- if (!strcasecmp (conf->dthrottle, "aggressive")) \
- conf->defrag->recon_thread_count = \
- throttle_count; \
- \
- pthread_mutex_unlock (&conf->defrag->dfq_mutex); \
- } \
-
/* TODO:
- use volumename in xattr instead of "dht"
- use NS locks
- handle all cases in self heal layout reconstruction
- complete linkfile selfheal
*/
-struct volume_options options[];
-
-extern dht_methods_t dht_methods;
-void
-dht_layout_dump (dht_layout_t *layout, const char *prefix)
+static void
+dht_layout_dump(dht_layout_t *layout, const char *prefix)
{
-
- char key[GF_DUMP_MAX_BUF_LEN];
- int i = 0;
-
- if (!layout)
- goto out;
- if (!prefix)
- goto out;
-
- gf_proc_dump_build_key(key, prefix, "cnt");
- gf_proc_dump_write(key, "%d", layout->cnt);
- gf_proc_dump_build_key(key, prefix, "preset");
- gf_proc_dump_write(key, "%d", layout->preset);
- gf_proc_dump_build_key(key, prefix, "gen");
- gf_proc_dump_write(key, "%d", layout->gen);
- if (layout->type != IA_INVAL) {
- gf_proc_dump_build_key(key, prefix, "inode type");
- gf_proc_dump_write(key, "%d", layout->type);
- }
-
- if (!IA_ISDIR (layout->type))
- goto out;
-
- for (i = 0; i < layout->cnt; i++) {
- gf_proc_dump_build_key(key, prefix,"list[%d].err", i);
- gf_proc_dump_write(key, "%d", layout->list[i].err);
- gf_proc_dump_build_key(key, prefix,"list[%d].start", i);
- gf_proc_dump_write(key, "%u", layout->list[i].start);
- gf_proc_dump_build_key(key, prefix,"list[%d].stop", i);
- gf_proc_dump_write(key, "%u", layout->list[i].stop);
- if (layout->list[i].xlator) {
- gf_proc_dump_build_key(key, prefix,
- "list[%d].xlator.type", i);
- gf_proc_dump_write(key, "%s",
- layout->list[i].xlator->type);
- gf_proc_dump_build_key(key, prefix,
- "list[%d].xlator.name", i);
- gf_proc_dump_write(key, "%s",
- layout->list[i].xlator->name);
- }
+ char key[GF_DUMP_MAX_BUF_LEN];
+ int i = 0;
+
+ if (!layout)
+ goto out;
+
+ gf_proc_dump_build_key(key, prefix, "cnt");
+ gf_proc_dump_write(key, "%d", layout->cnt);
+ gf_proc_dump_build_key(key, prefix, "preset");
+ gf_proc_dump_write(key, "%d", layout->preset);
+ gf_proc_dump_build_key(key, prefix, "gen");
+ gf_proc_dump_write(key, "%d", layout->gen);
+ if (layout->type != IA_INVAL) {
+ gf_proc_dump_build_key(key, prefix, "inode type");
+ gf_proc_dump_write(key, "%d", layout->type);
+ }
+
+ if (!IA_ISDIR(layout->type))
+ goto out;
+
+ for (i = 0; i < layout->cnt; i++) {
+ gf_proc_dump_build_key(key, prefix, "list[%d].err", i);
+ gf_proc_dump_write(key, "%d", layout->list[i].err);
+ gf_proc_dump_build_key(key, prefix, "list[%d].start", i);
+ gf_proc_dump_write(key, "0x%x", layout->list[i].start);
+ gf_proc_dump_build_key(key, prefix, "list[%d].stop", i);
+ gf_proc_dump_write(key, "0x%x", layout->list[i].stop);
+ if (layout->list[i].xlator) {
+ gf_proc_dump_build_key(key, prefix, "list[%d].xlator.type", i);
+ gf_proc_dump_write(key, "%s", layout->list[i].xlator->type);
+ gf_proc_dump_build_key(key, prefix, "list[%d].xlator.name", i);
+ gf_proc_dump_write(key, "%s", layout->list[i].xlator->name);
}
+ }
out:
- return;
+ return;
}
-
int32_t
-dht_priv_dump (xlator_t *this)
+dht_priv_dump(xlator_t *this)
{
- char key_prefix[GF_DUMP_MAX_BUF_LEN];
- char key[GF_DUMP_MAX_BUF_LEN];
- int i = 0;
- dht_conf_t *conf = NULL;
- int ret = -1;
+ char key_prefix[GF_DUMP_MAX_BUF_LEN];
+ char key[GF_DUMP_MAX_BUF_LEN];
+ int i = 0;
+ dht_conf_t *conf = NULL;
+ int ret = -1;
- if (!this)
- goto out;
+ if (!this)
+ goto out;
- conf = this->private;
- if (!conf)
- goto out;
+ conf = this->private;
+ if (!conf)
+ goto out;
- ret = TRY_LOCK(&conf->subvolume_lock);
- if (ret != 0) {
- return ret;
+ ret = TRY_LOCK(&conf->subvolume_lock);
+ if (ret != 0) {
+ return ret;
+ }
+
+ gf_proc_dump_add_section("xlator.cluster.dht.%s.priv", this->name);
+ gf_proc_dump_build_key(key_prefix, "xlator.cluster.dht", "%s.priv",
+ this->name);
+ gf_proc_dump_write("subvol_cnt", "%d", conf->subvolume_cnt);
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ snprintf(key, sizeof(key), "subvolumes[%d]", i);
+ gf_proc_dump_write(key, "%s.%s", conf->subvolumes[i]->type,
+ conf->subvolumes[i]->name);
+ if (conf->file_layouts && conf->file_layouts[i]) {
+ snprintf(key, sizeof(key), "file_layouts[%d]", i);
+ dht_layout_dump(conf->file_layouts[i], key);
}
-
- gf_proc_dump_add_section("xlator.cluster.dht.%s.priv", this->name);
- gf_proc_dump_build_key(key_prefix,"xlator.cluster.dht","%s.priv",
- this->name);
- gf_proc_dump_write("subvol_cnt","%d", conf->subvolume_cnt);
+ if (conf->dir_layouts && conf->dir_layouts[i]) {
+ snprintf(key, sizeof(key), "dir_layouts[%d]", i);
+ dht_layout_dump(conf->dir_layouts[i], key);
+ }
+ if (conf->subvolume_status) {
+ snprintf(key, sizeof(key), "subvolume_status[%d]", i);
+ gf_proc_dump_write(key, "%d", (int)conf->subvolume_status[i]);
+ }
+ }
+
+ gf_proc_dump_write("search_unhashed", "%d", conf->search_unhashed);
+ gf_proc_dump_write("gen", "%d", conf->gen);
+ gf_proc_dump_write("min_free_disk", "%lf", conf->min_free_disk);
+ gf_proc_dump_write("min_free_inodes", "%lf", conf->min_free_inodes);
+ gf_proc_dump_write("disk_unit", "%c", conf->disk_unit);
+ gf_proc_dump_write("refresh_interval", "%d", conf->refresh_interval);
+ gf_proc_dump_write("unhashed_sticky_bit", "%d", conf->unhashed_sticky_bit);
+ gf_proc_dump_write("use-readdirp", "%d", conf->use_readdirp);
+
+ if (conf->du_stats && conf->subvolume_status) {
for (i = 0; i < conf->subvolume_cnt; i++) {
- snprintf (key, sizeof (key), "subvolumes[%d]", i);
- gf_proc_dump_write(key, "%s.%s", conf->subvolumes[i]->type,
- conf->subvolumes[i]->name);
- if (conf->file_layouts && conf->file_layouts[i]){
- snprintf (key, sizeof (key), "file_layouts[%d]", i);
- dht_layout_dump(conf->file_layouts[i], key);
- }
- if (conf->dir_layouts && conf->dir_layouts[i]) {
- snprintf (key, sizeof (key), "dir_layouts[%d]", i);
- dht_layout_dump(conf->dir_layouts[i], key);
- }
- if (conf->subvolume_status) {
-
- snprintf (key, sizeof (key), "subvolume_status[%d]", i);
- gf_proc_dump_write(key, "%d",
- (int)conf->subvolume_status[i]);
- }
+ if (!conf->subvolume_status[i])
+ continue;
- }
+ snprintf(key, sizeof(key), "subvolumes[%d]", i);
+ gf_proc_dump_write(key, "%s", conf->subvolumes[i]->name);
+
+ snprintf(key, sizeof(key), "du_stats[%d].avail_percent", i);
+ gf_proc_dump_write(key, "%lf", conf->du_stats[i].avail_percent);
+
+ snprintf(key, sizeof(key), "du_stats[%d].avail_space", i);
+ gf_proc_dump_write(key, "%" PRIu64, conf->du_stats[i].avail_space);
- gf_proc_dump_write("search_unhashed", "%d", conf->search_unhashed);
- gf_proc_dump_write("gen", "%d", conf->gen);
- gf_proc_dump_write("min_free_disk", "%lf", conf->min_free_disk);
- gf_proc_dump_write("min_free_inodes", "%lf", conf->min_free_inodes);
- gf_proc_dump_write("disk_unit", "%c", conf->disk_unit);
- gf_proc_dump_write("refresh_interval", "%d", conf->refresh_interval);
- gf_proc_dump_write("unhashed_sticky_bit", "%d", conf->unhashed_sticky_bit);
- gf_proc_dump_write("use-readdirp", "%d", conf->use_readdirp);
-
- if (conf->du_stats && conf->subvolume_status) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (!conf->subvolume_status[i])
- continue;
-
- snprintf (key, sizeof (key), "subvolumes[%d]", i);
- gf_proc_dump_write (key, "%s",
- conf->subvolumes[i]->name);
-
- snprintf (key, sizeof (key),
- "du_stats[%d].avail_percent", i);
- gf_proc_dump_write (key, "%lf",
- conf->du_stats[i].avail_percent);
-
- snprintf (key, sizeof (key), "du_stats[%d].avail_space",
- i);
- gf_proc_dump_write (key, "%lu",
- conf->du_stats[i].avail_space);
-
- snprintf (key, sizeof (key),
- "du_stats[%d].avail_inodes", i);
- gf_proc_dump_write (key, "%lf",
- conf->du_stats[i].avail_inodes);
-
- snprintf (key, sizeof (key), "du_stats[%d].log", i);
- gf_proc_dump_write (key, "%lu",
- conf->du_stats[i].log);
- }
+ snprintf(key, sizeof(key), "du_stats[%d].avail_inodes", i);
+ gf_proc_dump_write(key, "%lf", conf->du_stats[i].avail_inodes);
+
+ snprintf(key, sizeof(key), "du_stats[%d].log", i);
+ gf_proc_dump_write(key, "%" PRIu32, conf->du_stats[i].log);
}
+ }
- if (conf->last_stat_fetch.tv_sec)
- gf_proc_dump_write("last_stat_fetch", "%s",
- ctime(&conf->last_stat_fetch.tv_sec));
+ if (conf->last_stat_fetch)
+ gf_proc_dump_write("last_stat_fetch", "%s",
+ ctime(&conf->last_stat_fetch));
- UNLOCK(&conf->subvolume_lock);
+ UNLOCK(&conf->subvolume_lock);
out:
- return ret;
+ return ret;
}
int32_t
-dht_inodectx_dump (xlator_t *this, inode_t *inode)
+dht_inodectx_dump(xlator_t *this, inode_t *inode)
{
- int ret = -1;
- dht_layout_t *layout = NULL;
+ int ret = -1;
+ dht_layout_t *layout = NULL;
- if (!this)
- goto out;
- if (!inode)
- goto out;
+ if (!this)
+ goto out;
+ if (!inode)
+ goto out;
- ret = dht_inode_ctx_layout_get (inode, this, &layout);
+ ret = dht_inode_ctx_layout_get(inode, this, &layout);
- if ((ret != 0) || !layout)
- return ret;
+ if ((ret != 0) || !layout)
+ return ret;
- gf_proc_dump_add_section("xlator.cluster.dht.%s.inode", this->name);
- dht_layout_dump(layout, "layout");
+ gf_proc_dump_add_section("xlator.cluster.dht.%s.inode", this->name);
+ dht_layout_dump(layout, "layout");
out:
- return ret;
+ return ret;
}
void
-dht_fini (xlator_t *this)
+dht_fini(xlator_t *this)
{
- int i = 0;
- dht_conf_t *conf = NULL;
-
- GF_VALIDATE_OR_GOTO ("dht", this, out);
-
- conf = this->private;
- this->private = NULL;
- if (conf) {
- if (conf->file_layouts) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- GF_FREE (conf->file_layouts[i]);
- }
- GF_FREE (conf->file_layouts);
- }
-
- dict_unref(conf->leaf_to_subvol);
-
- /* allocated in dht_init_subvolumes() */
- GF_FREE (conf->subvolumes);
- GF_FREE (conf->subvolume_status);
- GF_FREE (conf->last_event);
- GF_FREE (conf->subvol_up_time);
- GF_FREE (conf->du_stats);
- GF_FREE (conf->decommissioned_bricks);
-
- /* allocated in dht_init() */
- GF_FREE (conf->mds_xattr_key);
- GF_FREE (conf->link_xattr_name);
- GF_FREE (conf->commithash_xattr_name);
- GF_FREE (conf->wild_xattr_name);
-
- /* allocated in dht_init_regex() */
- if (conf->rsync_regex_valid)
- regfree (&conf->rsync_regex);
- if (conf->extra_regex_valid)
- regfree (&conf->extra_regex);
-
- synclock_destroy (&conf->link_lock);
-
- if (conf->lock_pool)
- mem_pool_destroy (conf->lock_pool);
-
- GF_FREE (conf);
+ int i = 0;
+ dht_conf_t *conf = NULL;
+
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+
+ conf = this->private;
+ this->private = NULL;
+ if (conf) {
+ if (conf->file_layouts) {
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ GF_FREE(conf->file_layouts[i]);
+ }
+ GF_FREE(conf->file_layouts);
}
+
+ dict_unref(conf->leaf_to_subvol);
+
+ /* allocated in dht_init_subvolumes() */
+ GF_FREE(conf->subvolumes);
+ GF_FREE(conf->subvolume_status);
+ GF_FREE(conf->last_event);
+ GF_FREE(conf->subvol_up_time);
+ GF_FREE(conf->du_stats);
+ GF_FREE(conf->decommissioned_bricks);
+
+ /* allocated in dht_init() */
+ GF_FREE(conf->mds_xattr_key);
+ GF_FREE(conf->link_xattr_name);
+ GF_FREE(conf->commithash_xattr_name);
+ GF_FREE(conf->wild_xattr_name);
+
+ /* allocated in dht_init_regex() */
+ if (conf->rsync_regex_valid)
+ regfree(&conf->rsync_regex);
+ if (conf->extra_regex_valid)
+ regfree(&conf->extra_regex);
+
+ synclock_destroy(&conf->link_lock);
+
+ if (conf->lock_pool)
+ mem_pool_destroy(conf->lock_pool);
+
+ GF_FREE(conf);
+ }
out:
- return;
+ return;
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO ("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
- ret = xlator_mem_acct_init (this, gf_dht_mt_end + 1);
+ ret = xlator_mem_acct_init(this, gf_dht_mt_end + 1);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_NO_MEMORY,
- "Memory accounting init failed");
- return ret;
- }
-out:
+ if (ret != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_NO_MEMORY,
+ "Memory accounting init failed");
return ret;
+ }
+out:
+ return ret;
}
-
-int
-dht_parse_decommissioned_bricks (xlator_t *this, dht_conf_t *conf,
- const char *bricks)
+static int
+dht_parse_decommissioned_bricks(xlator_t *this, dht_conf_t *conf,
+ const char *bricks)
{
- int i = 0;
- int ret = -1;
- char *tmpstr = NULL;
- char *dup_brick = NULL;
- char *node = NULL;
-
- if (!conf || !bricks)
- goto out;
-
- dup_brick = gf_strdup (bricks);
- node = strtok_r (dup_brick, ",", &tmpstr);
- while (node) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (!strcmp (conf->subvolumes[i]->name, node)) {
- conf->decommissioned_bricks[i] =
- conf->subvolumes[i];
- conf->decommission_subvols_cnt++;
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_SUBVOL_DECOMMISSION_INFO,
- "decommissioning subvolume %s",
- conf->subvolumes[i]->name);
- break;
- }
- }
- if (i == conf->subvolume_cnt) {
- /* Wrong node given. */
- goto out;
- }
- node = strtok_r (NULL, ",", &tmpstr);
+ int i = 0;
+ int ret = -1;
+ char *tmpstr = NULL;
+ char *dup_brick = NULL;
+ char *node = NULL;
+
+ if (!conf || !bricks)
+ goto out;
+
+ dup_brick = gf_strdup(bricks);
+ if (dup_brick == NULL) {
+ goto out;
+ }
+
+ node = strtok_r(dup_brick, ",", &tmpstr);
+ while (node) {
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (!strcmp(conf->subvolumes[i]->name, node)) {
+ conf->decommissioned_bricks[i] = conf->subvolumes[i];
+ conf->decommission_subvols_cnt++;
+ gf_msg(this->name, GF_LOG_INFO, 0,
+ DHT_MSG_SUBVOL_DECOMMISSION_INFO,
+ "decommissioning subvolume %s",
+ conf->subvolumes[i]->name);
+ break;
+ }
+ }
+ if (i == conf->subvolume_cnt) {
+ /* Wrong node given. */
+ goto out;
}
+ node = strtok_r(NULL, ",", &tmpstr);
+ }
- ret = 0;
- conf->decommission_in_progress = 1;
+ ret = 0;
+ conf->decommission_in_progress = 1;
out:
- GF_FREE (dup_brick);
+ GF_FREE(dup_brick);
- return ret;
+ return ret;
}
-int
-dht_decommissioned_remove (xlator_t *this, dht_conf_t *conf)
+static void
+dht_decommissioned_remove(xlator_t *this, dht_conf_t *conf)
{
- int i = 0;
- int ret = -1;
+ int i = 0;
- if (!conf)
- goto out;
-
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->decommissioned_bricks[i]) {
- conf->decommissioned_bricks[i] = NULL;
- conf->decommission_subvols_cnt--;
- }
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ if (conf->decommissioned_bricks[i]) {
+ conf->decommissioned_bricks[i] = NULL;
+ conf->decommission_subvols_cnt--;
}
-
- ret = 0;
-out:
-
- return ret;
+ }
}
-void
-dht_init_regex (xlator_t *this, dict_t *odict, char *name,
- regex_t *re, gf_boolean_t *re_valid, dht_conf_t *conf)
+
+static void
+dht_init_regex(xlator_t *this, dict_t *odict, char *name, regex_t *re,
+ gf_boolean_t *re_valid, dht_conf_t *conf)
{
- char *temp_str = NULL;
+ char *temp_str = NULL;
- if (dict_get_str (odict, name, &temp_str) != 0) {
- if (strcmp(name,"rsync-hash-regex")) {
- return;
- }
- temp_str = "^\\.(.+)\\.[^.]+$";
+ if (dict_get_str(odict, name, &temp_str) != 0) {
+ if (strcmp(name, "rsync-hash-regex")) {
+ return;
+ }
+ temp_str = "^\\.(.+)\\.[^.]+$";
+ }
+
+ LOCK(&conf->lock);
+ {
+ if (*re_valid) {
+ regfree(re);
+ *re_valid = _gf_false;
+ }
+
+ if (!strcmp(temp_str, "none")) {
+ goto unlock;
}
- LOCK (&conf->lock);
- {
- if (*re_valid) {
- regfree(re);
- *re_valid = _gf_false;
- }
-
- if (!strcmp(temp_str, "none")) {
- goto unlock;
- }
-
- if (regcomp(re, temp_str, REG_EXTENDED) == 0) {
- gf_msg_debug (this->name, 0,
- "using regex %s = %s", name, temp_str);
- *re_valid = _gf_true;
- } else {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_REGEX_INFO,
- "compiling regex %s failed", temp_str);
- }
+ if (regcomp(re, temp_str, REG_EXTENDED) == 0) {
+ gf_msg_debug(this->name, 0, "using regex %s = %s", name, temp_str);
+ *re_valid = _gf_true;
+ } else {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_REGEX_INFO,
+ "compiling regex %s failed", temp_str);
}
+ }
unlock:
- UNLOCK (&conf->lock);
+ UNLOCK(&conf->lock);
}
int
dht_set_subvol_range(xlator_t *this)
{
- int ret = -1;
- dht_conf_t *conf = NULL;
+ int ret = -1;
+ dht_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf)
- goto out;
+ if (!conf)
+ goto out;
- conf->leaf_to_subvol = dict_new();
- if (!conf->leaf_to_subvol)
- goto out;
+ conf->leaf_to_subvol = dict_new();
+ if (!conf->leaf_to_subvol)
+ goto out;
- ret = glusterfs_reachable_leaves(this, conf->leaf_to_subvol);
+ ret = glusterfs_reachable_leaves(this, conf->leaf_to_subvol);
out:
- return ret;
+ return ret;
}
-int
-dht_configure_throttle (xlator_t *this, dht_conf_t *conf, char *temp_str)
+static int
+dht_configure_throttle(xlator_t *this, dht_conf_t *conf, char *temp_str)
{
- int rebal_thread_count = 0;
- int ret = 0;
-
- pthread_mutex_lock (&conf->defrag->dfq_mutex);
- {
- if (!strcasecmp (temp_str, "lazy")) {
- conf->defrag->recon_thread_count = 1;
- } else if (!strcasecmp (temp_str, "normal")) {
- conf->defrag->recon_thread_count = 2;
- } else if (!strcasecmp (temp_str, "aggressive")) {
- conf->defrag->recon_thread_count = MAX (MAX_REBAL_THREADS - 4, 4);
- } else if ((gf_string2int (temp_str, &rebal_thread_count) == 0)) {
- if ((rebal_thread_count > 0) && (rebal_thread_count <= MAX_REBAL_THREADS)) {
- gf_msg (this->name, GF_LOG_INFO, 0, 0,
- "rebal thread count configured to %d",
- rebal_thread_count);
- conf->defrag->recon_thread_count = rebal_thread_count;
- } else {
- gf_msg(this->name, GF_LOG_ERROR, 0,
- DHT_MSG_INVALID_OPTION,
- "Invalid option: Reconfigure: "
- "rebal-throttle should be "
- "within range of 0 and maximum number of"
- " cores available");
- ret = -1;
- pthread_mutex_unlock (&conf->defrag->dfq_mutex);
- goto out;
- }
+ int rebal_thread_count = 0;
+ int ret = 0;
+
+ pthread_mutex_lock(&conf->defrag->dfq_mutex);
+ {
+ if (!strcasecmp(temp_str, "lazy")) {
+ conf->defrag->recon_thread_count = 1;
+ } else if (!strcasecmp(temp_str, "normal")) {
+ conf->defrag->recon_thread_count = 2;
+ } else if (!strcasecmp(temp_str, "aggressive")) {
+ conf->defrag->recon_thread_count = MAX(MAX_REBAL_THREADS - 4, 4);
+ } else if ((gf_string2int(temp_str, &rebal_thread_count) == 0)) {
+ if ((rebal_thread_count > 0) &&
+ (rebal_thread_count <= MAX_REBAL_THREADS)) {
+ conf->defrag->recon_thread_count = rebal_thread_count;
+ pthread_mutex_unlock(&conf->defrag->dfq_mutex);
+ gf_msg(this->name, GF_LOG_INFO, 0, 0,
+ "rebal thread count configured to %d",
+ rebal_thread_count);
+ goto out;
+ } else {
+ pthread_mutex_unlock(&conf->defrag->dfq_mutex);
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_INVALID_OPTION,
+ "Invalid option: Reconfigure: "
+ "rebal-throttle should be "
+ "within range of 0 and maximum number of"
+ " cores available");
+ ret = -1;
+ goto out;
+ }
} else {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_INVALID_OPTION,
- "Invalid option: Reconfigure: "
- "rebal-throttle should be {lazy|normal|aggressive}"
- " or a number up to the number of cores available,"
- " not (%s), defaulting to (%d)",
- temp_str, conf->dthrottle);
- ret = -1;
- }
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_INVALID_OPTION,
+ "Invalid option: Reconfigure: "
+ "rebal-throttle should be {lazy|normal|aggressive}"
+ " or a number up to the number of cores available,"
+ " not (%s), defaulting to (%d)",
+ temp_str, conf->dthrottle);
+ ret = -1;
}
- pthread_mutex_unlock (&conf->defrag->dfq_mutex);
+ }
+ pthread_mutex_unlock(&conf->defrag->dfq_mutex);
out:
- return ret;
+ return ret;
}
int
-dht_reconfigure (xlator_t *this, dict_t *options)
+dht_reconfigure(xlator_t *this, dict_t *options)
{
- dht_conf_t *conf = NULL;
- char *temp_str = NULL;
- gf_boolean_t search_unhashed;
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO ("dht", this, out);
- GF_VALIDATE_OR_GOTO ("dht", options, out);
-
- conf = this->private;
- if (!conf)
- return 0;
-
- if (dict_get_str (options, "lookup-unhashed", &temp_str) == 0) {
- /* If option is not "auto", other options _should_ be boolean*/
- if (strcasecmp (temp_str, "auto")) {
- if (!gf_string2boolean (temp_str, &search_unhashed)) {
- gf_msg_debug(this->name, 0, "Reconfigure: "
- "lookup-unhashed reconfigured(%s)",
- temp_str);
- conf->search_unhashed = search_unhashed;
- } else {
- gf_msg(this->name, GF_LOG_ERROR, 0,
- DHT_MSG_INVALID_OPTION,
- "Invalid option: Reconfigure: "
- "lookup-unhashed should be boolean,"
- " not (%s), defaulting to (%d)",
- temp_str, conf->search_unhashed);
- ret = -1;
- goto out;
- }
- } else {
- gf_msg_debug(this->name, 0, "Reconfigure:"
- " lookup-unhashed reconfigured auto ");
- conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
- }
- }
+ dht_conf_t *conf = NULL;
+ char *temp_str = NULL;
+ gf_boolean_t search_unhashed;
+ int ret = -1;
- GF_OPTION_RECONF ("lookup-optimize", conf->lookup_optimize, options,
- bool, out);
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+ GF_VALIDATE_OR_GOTO("dht", options, out);
- GF_OPTION_RECONF ("min-free-disk", conf->min_free_disk, options,
- percent_or_size, out);
- /* option can be any one of percent or bytes */
- conf->disk_unit = 0;
- if (conf->min_free_disk < 100.0)
- conf->disk_unit = 'p';
+ conf = this->private;
+ if (!conf)
+ return 0;
- GF_OPTION_RECONF ("min-free-inodes", conf->min_free_inodes, options,
- percent, out);
+ if (dict_get_str(options, "lookup-unhashed", &temp_str) == 0) {
+ /* If option is not "auto", other options _should_ be boolean*/
+ if (strcasecmp(temp_str, "auto")) {
+ if (!gf_string2boolean(temp_str, &search_unhashed)) {
+ gf_msg_debug(this->name, 0,
+ "Reconfigure: "
+ "lookup-unhashed reconfigured(%s)",
+ temp_str);
+ conf->search_unhashed = search_unhashed;
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_INVALID_OPTION,
+ "Invalid option: Reconfigure: "
+ "lookup-unhashed should be boolean,"
+ " not (%s), defaulting to (%d)",
+ temp_str, conf->search_unhashed);
+ ret = -1;
+ goto out;
+ }
+ } else {
+ gf_msg_debug(this->name, 0,
+ "Reconfigure:"
+ " lookup-unhashed reconfigured auto ");
+ conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
+ }
+ }
- GF_OPTION_RECONF ("directory-layout-spread", conf->dir_spread_cnt,
- options, uint32, out);
+ GF_OPTION_RECONF("lookup-optimize", conf->lookup_optimize, options, bool,
+ out);
- GF_OPTION_RECONF ("readdir-optimize", conf->readdir_optimize, options,
- bool, out);
- GF_OPTION_RECONF ("randomize-hash-range-by-gfid",
- conf->randomize_by_gfid,
- options, bool, out);
+ GF_OPTION_RECONF("min-free-disk", conf->min_free_disk, options,
+ percent_or_size, out);
+ /* option can be any one of percent or bytes */
+ conf->disk_unit = 0;
+ if (conf->min_free_disk < 100.0)
+ conf->disk_unit = 'p';
- GF_OPTION_RECONF ("lock-migration", conf->lock_migration_enabled,
- options, bool, out);
+ GF_OPTION_RECONF("min-free-inodes", conf->min_free_inodes, options, percent,
+ out);
- GF_OPTION_RECONF ("force-migration", conf->force_migration,
- options, bool, out);
+ GF_OPTION_RECONF("directory-layout-spread", conf->dir_spread_cnt, options,
+ uint32, out);
+ GF_OPTION_RECONF("readdir-optimize", conf->readdir_optimize, options, bool,
+ out);
+ GF_OPTION_RECONF("randomize-hash-range-by-gfid", conf->randomize_by_gfid,
+ options, bool, out);
- if (conf->defrag) {
- if (dict_get_str (options, "rebal-throttle", &temp_str) == 0) {
- ret = dht_configure_throttle (this, conf, temp_str);
- if (ret == -1)
- goto out;
- }
- }
+ GF_OPTION_RECONF("lock-migration", conf->lock_migration_enabled, options,
+ bool, out);
- if (conf->defrag) {
- conf->defrag->lock_migration_enabled =
- conf->lock_migration_enabled;
- }
+ GF_OPTION_RECONF("force-migration", conf->force_migration, options, bool,
+ out);
- if (conf->defrag) {
- GF_OPTION_RECONF ("rebalance-stats", conf->defrag->stats,
- options, bool, out);
+ if (conf->defrag) {
+ if (dict_get_str(options, "rebal-throttle", &temp_str) == 0) {
+ ret = dht_configure_throttle(this, conf, temp_str);
+ if (ret == -1)
+ goto out;
}
+ }
+
+ if (conf->defrag) {
+ conf->defrag->lock_migration_enabled = conf->lock_migration_enabled;
+ }
+
+ if (conf->defrag) {
+ GF_OPTION_RECONF("rebalance-stats", conf->defrag->stats, options, bool,
+ out);
+ }
+
+ if (dict_get_str(options, "decommissioned-bricks", &temp_str) == 0) {
+ ret = dht_parse_decommissioned_bricks(this, conf, temp_str);
+ if (ret == -1)
+ goto out;
+ } else {
+ dht_decommissioned_remove(this, conf);
+ }
+
+ dht_init_regex(this, options, "rsync-hash-regex", &conf->rsync_regex,
+ &conf->rsync_regex_valid, conf);
+ dht_init_regex(this, options, "extra-hash-regex", &conf->extra_regex,
+ &conf->extra_regex_valid, conf);
+
+ GF_OPTION_RECONF("weighted-rebalance", conf->do_weighting, options, bool,
+ out);
+
+ GF_OPTION_RECONF("use-readdirp", conf->use_readdirp, options, bool, out);
+ ret = 0;
+out:
+ return ret;
+}
- if (dict_get_str (options, "decommissioned-bricks", &temp_str) == 0) {
- ret = dht_parse_decommissioned_bricks (this, conf, temp_str);
- if (ret == -1)
- goto out;
- } else {
- ret = dht_decommissioned_remove (this, conf);
- if (ret == -1)
- goto out;
+static int
+gf_defrag_pattern_list_fill(xlator_t *this, gf_defrag_info_t *defrag,
+ char *data)
+{
+ int ret = -1;
+ char *tmp_str = NULL;
+ char *tmp_str1 = NULL;
+ char *dup_str = NULL;
+ char *num = NULL;
+ char *pattern_str = NULL;
+ char *pattern = NULL;
+ gf_defrag_pattern_list_t *temp_list = NULL;
+ gf_defrag_pattern_list_t *pattern_list = NULL;
+
+ if (!this || !defrag || !data)
+ goto out;
+
+ /* Get the pattern for pattern list. "pattern:<optional-size>"
+ * eg: *avi, *pdf:10MB, *:1TB
+ */
+ pattern_str = strtok_r(data, ",", &tmp_str);
+ while (pattern_str) {
+ dup_str = gf_strdup(pattern_str);
+ if (!dup_str)
+ goto out;
+ pattern_list = GF_CALLOC(1, sizeof(gf_defrag_pattern_list_t), 1);
+ if (!pattern_list) {
+ goto out;
}
+ pattern = strtok_r(dup_str, ":", &tmp_str1);
+ num = strtok_r(NULL, ":", &tmp_str1);
+ if (!pattern)
+ goto out;
+ if (!num) {
+ if (gf_string2bytesize_uint64(pattern, &pattern_list->size) == 0) {
+ pattern = "*";
+ }
+ } else if (gf_string2bytesize_uint64(num, &pattern_list->size) != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_INVALID_OPTION,
+ "Invalid option. Defrag pattern:"
+ " Invalid number format \"%s\"",
+ num);
+ goto out;
+ }
+ memcpy(pattern_list->path_pattern, pattern, strlen(dup_str));
- dht_init_regex (this, options, "rsync-hash-regex",
- &conf->rsync_regex, &conf->rsync_regex_valid, conf);
- dht_init_regex (this, options, "extra-hash-regex",
- &conf->extra_regex, &conf->extra_regex_valid, conf);
+ if (!defrag->defrag_pattern)
+ temp_list = NULL;
+ else
+ temp_list = defrag->defrag_pattern;
- GF_OPTION_RECONF ("weighted-rebalance", conf->do_weighting, options,
- bool, out);
+ pattern_list->next = temp_list;
- GF_OPTION_RECONF ("use-readdirp", conf->use_readdirp, options,
- bool, out);
- ret = 0;
-out:
- return ret;
-}
+ defrag->defrag_pattern = pattern_list;
+ pattern_list = NULL;
-static int
-gf_defrag_pattern_list_fill (xlator_t *this, gf_defrag_info_t *defrag, char *data)
-{
- int ret = -1;
- char *tmp_str = NULL;
- char *tmp_str1 = NULL;
- char *dup_str = NULL;
- char *num = NULL;
- char *pattern_str = NULL;
- char *pattern = NULL;
- gf_defrag_pattern_list_t *temp_list = NULL;
- gf_defrag_pattern_list_t *pattern_list = NULL;
-
- if (!this || !defrag || !data)
- goto out;
+ GF_FREE(dup_str);
+ dup_str = NULL;
- /* Get the pattern for pattern list. "pattern:<optional-size>"
- * eg: *avi, *pdf:10MB, *:1TB
- */
- pattern_str = strtok_r (data, ",", &tmp_str);
- while (pattern_str) {
- dup_str = gf_strdup (pattern_str);
- pattern_list = GF_CALLOC (1, sizeof (gf_defrag_pattern_list_t),
- 1);
- if (!pattern_list) {
- goto out;
- }
- pattern = strtok_r (dup_str, ":", &tmp_str1);
- num = strtok_r (NULL, ":", &tmp_str1);
- if (!pattern)
- goto out;
- if (!num) {
- if (gf_string2bytesize_uint64(pattern, &pattern_list->size)
- == 0) {
- pattern = "*";
- }
- } else if (gf_string2bytesize_uint64 (num, &pattern_list->size) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_INVALID_OPTION,
- "Invalid option. Defrag pattern:"
- " Invalid number format \"%s\"", num);
- goto out;
- }
- memcpy (pattern_list->path_pattern, pattern, strlen (dup_str));
-
- if (!defrag->defrag_pattern)
- temp_list = NULL;
- else
- temp_list = defrag->defrag_pattern;
-
- pattern_list->next = temp_list;
-
- defrag->defrag_pattern = pattern_list;
- pattern_list = NULL;
-
- GF_FREE (dup_str);
- dup_str = NULL;
-
- pattern_str = strtok_r (NULL, ",", &tmp_str);
- }
+ pattern_str = strtok_r(NULL, ",", &tmp_str);
+ }
- ret = 0;
+ ret = 0;
out:
- if (ret)
- GF_FREE (pattern_list);
- GF_FREE (dup_str);
+ if (ret)
+ GF_FREE(pattern_list);
+ GF_FREE(dup_str);
- return ret;
+ return ret;
}
-
-
-int
-dht_init_methods (xlator_t *this)
+static int
+dht_init_methods(xlator_t *this)
{
- int ret = -1;
- dht_conf_t *conf = NULL;
- dht_methods_t *methods = NULL;
+ int ret = -1;
+ dht_conf_t *conf = NULL;
+ dht_methods_t *methods = NULL;
- GF_VALIDATE_OR_GOTO ("dht", this, err);
+ GF_VALIDATE_OR_GOTO("dht", this, err);
- conf = this->private;
- methods = &(conf->methods);
+ conf = this->private;
+ methods = &(conf->methods);
- methods->migration_get_dst_subvol = dht_migration_get_dst_subvol;
- methods->migration_needed = dht_migration_needed;
- methods->migration_other = NULL;
- methods->layout_search = dht_layout_search;
+ methods->migration_get_dst_subvol = dht_migration_get_dst_subvol;
+ methods->migration_other = NULL;
+ methods->layout_search = dht_layout_search;
- ret = 0;
+ ret = 0;
err:
- return ret;
+ return ret;
}
int
-dht_init (xlator_t *this)
+dht_init(xlator_t *this)
{
- dht_conf_t *conf = NULL;
- char *temp_str = NULL;
- int ret = -1;
- int i = 0;
- gf_defrag_info_t *defrag = NULL;
- int cmd = 0;
- char *node_uuid = NULL;
- uint32_t commit_hash = 0;
-
- GF_VALIDATE_OR_GOTO ("dht", this, err);
-
- if (!this->children) {
- gf_msg (this->name, GF_LOG_CRITICAL, 0,
- DHT_MSG_INVALID_CONFIGURATION,
- "Distribute needs more than one subvolume");
- return -1;
- }
-
- if (!this->parents) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_INVALID_CONFIGURATION,
- "dangling volume. check volfile");
- }
-
- conf = GF_CALLOC (1, sizeof (*conf), gf_dht_mt_dht_conf_t);
- if (!conf) {
- goto err;
- }
-
- LOCK_INIT (&conf->subvolume_lock);
- LOCK_INIT (&conf->layout_lock);
- LOCK_INIT (&conf->lock);
- synclock_init (&conf->link_lock, SYNC_LOCK_DEFAULT);
-
- /* We get the commit-hash to set only for rebalance process */
- if (dict_get_uint32 (this->options,
- "commit-hash", &commit_hash) == 0) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_COMMIT_HASH_INFO, "%s using commit hash %u",
- __func__, commit_hash);
- conf->vol_commit_hash = commit_hash;
- conf->vch_forced = _gf_true;
- }
-
- ret = dict_get_int32 (this->options, "rebalance-cmd", &cmd);
-
- if (cmd) {
- defrag = GF_CALLOC (1, sizeof (gf_defrag_info_t),
- gf_defrag_info_mt);
-
- GF_VALIDATE_OR_GOTO (this->name, defrag, err);
-
- LOCK_INIT (&defrag->lock);
-
- defrag->is_exiting = 0;
-
- conf->defrag = defrag;
- defrag->this = this;
-
- ret = dict_get_str (this->options, "node-uuid", &node_uuid);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_INVALID_CONFIGURATION,
- "Invalid volume configuration: "
- "node-uuid not specified");
- goto err;
- }
+ dht_conf_t *conf = NULL;
+ char *temp_str = NULL;
+ int ret = -1;
+ int i = 0;
+ gf_defrag_info_t *defrag = NULL;
+ int cmd = 0;
+ char *node_uuid = NULL;
+ uint32_t commit_hash = 0;
+
+ GF_VALIDATE_OR_GOTO("dht", this, err);
+
+ if (!this->children) {
+ gf_msg(this->name, GF_LOG_CRITICAL, 0, DHT_MSG_INVALID_CONFIGURATION,
+ "Distribute needs more than one subvolume");
+ return -1;
+ }
- if (gf_uuid_parse (node_uuid, defrag->node_uuid)) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_INVALID_OPTION, "Invalid option:"
- " Cannot parse glusterd node uuid");
- goto err;
- }
+ if (!this->parents) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_INVALID_CONFIGURATION,
+ "dangling volume. check volfile");
+ }
- defrag->cmd = cmd;
+ conf = GF_CALLOC(1, sizeof(*conf), gf_dht_mt_dht_conf_t);
+ if (!conf) {
+ goto err;
+ }
- defrag->stats = _gf_false;
+ LOCK_INIT(&conf->subvolume_lock);
+ LOCK_INIT(&conf->layout_lock);
+ LOCK_INIT(&conf->lock);
+ synclock_init(&conf->link_lock, SYNC_LOCK_DEFAULT);
- defrag->queue = NULL;
+ /* We get the commit-hash to set only for rebalance process */
+ if (dict_get_uint32(this->options, "commit-hash", &commit_hash) == 0) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_COMMIT_HASH_INFO,
+ "%s using commit hash %u", __func__, commit_hash);
+ conf->vol_commit_hash = commit_hash;
+ conf->vch_forced = _gf_true;
+ }
- defrag->crawl_done = 0;
+ ret = dict_get_int32(this->options, "rebalance-cmd", &cmd);
- defrag->global_error = 0;
+ if (cmd) {
+ defrag = GF_CALLOC(1, sizeof(gf_defrag_info_t), gf_defrag_info_mt);
- defrag->q_entry_count = 0;
+ GF_VALIDATE_OR_GOTO(this->name, defrag, err);
- defrag->wakeup_crawler = 0;
+ LOCK_INIT(&defrag->lock);
- pthread_mutex_init (&defrag->dfq_mutex, 0);
- pthread_cond_init (&defrag->parallel_migration_cond, 0);
- pthread_cond_init (&defrag->rebalance_crawler_alarm, 0);
- pthread_cond_init (&defrag->df_wakeup_thread, 0);
+ defrag->is_exiting = 0;
- pthread_mutex_init (&defrag->fc_mutex, 0);
- pthread_cond_init (&defrag->fc_wakeup_cond, 0);
+ conf->defrag = defrag;
+ defrag->this = this;
- defrag->global_error = 0;
+ ret = dict_get_str(this->options, "node-uuid", &node_uuid);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_INVALID_CONFIGURATION,
+ "Invalid volume configuration: "
+ "node-uuid not specified");
+ goto err;
}
- conf->use_fallocate = 1;
-
- conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_ON;
- if (dict_get_str (this->options, "lookup-unhashed", &temp_str) == 0) {
- /* If option is not "auto", other options _should_ be boolean */
- if (strcasecmp (temp_str, "auto")) {
- gf_boolean_t search_unhashed_bool;
- ret = gf_string2boolean (temp_str, &search_unhashed_bool);
- if (ret == -1) {
- goto err;
- }
- conf->search_unhashed = search_unhashed_bool
- ? GF_DHT_LOOKUP_UNHASHED_ON
- : GF_DHT_LOOKUP_UNHASHED_OFF;
- }
- else {
- conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
- }
+ if (gf_uuid_parse(node_uuid, defrag->node_uuid)) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_INVALID_OPTION,
+ "Invalid option:"
+ " Cannot parse glusterd node uuid");
+ goto err;
}
- GF_OPTION_INIT ("lookup-optimize", conf->lookup_optimize, bool, err);
+ defrag->cmd = cmd;
- GF_OPTION_INIT ("unhashed-sticky-bit", conf->unhashed_sticky_bit, bool,
- err);
+ defrag->stats = _gf_false;
- GF_OPTION_INIT ("use-readdirp", conf->use_readdirp, bool, err);
+ defrag->queue = NULL;
- GF_OPTION_INIT ("min-free-disk", conf->min_free_disk, percent_or_size,
- err);
+ defrag->crawl_done = 0;
- GF_OPTION_INIT ("min-free-inodes", conf->min_free_inodes, percent,
- err);
+ defrag->global_error = 0;
- conf->dir_spread_cnt = conf->subvolume_cnt;
- GF_OPTION_INIT ("directory-layout-spread", conf->dir_spread_cnt,
- uint32, err);
+ defrag->q_entry_count = 0;
- GF_OPTION_INIT ("assert-no-child-down", conf->assert_no_child_down,
- bool, err);
+ defrag->wakeup_crawler = 0;
- GF_OPTION_INIT ("readdir-optimize", conf->readdir_optimize, bool, err);
+ pthread_mutex_init(&defrag->dfq_mutex, 0);
+ pthread_cond_init(&defrag->parallel_migration_cond, 0);
+ pthread_cond_init(&defrag->rebalance_crawler_alarm, 0);
+ pthread_cond_init(&defrag->df_wakeup_thread, 0);
+ pthread_mutex_init(&defrag->fc_mutex, 0);
+ pthread_cond_init(&defrag->fc_wakeup_cond, 0);
- GF_OPTION_INIT ("lock-migration", conf->lock_migration_enabled,
- bool, err);
+ defrag->global_error = 0;
+ }
- GF_OPTION_INIT ("force-migration", conf->force_migration,
- bool, err);
+ conf->use_fallocate = 1;
+ conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_ON;
+ if (dict_get_str(this->options, "lookup-unhashed", &temp_str) == 0) {
+ /* If option is not "auto", other options _should_ be boolean */
+ if (strcasecmp(temp_str, "auto")) {
+ gf_boolean_t search_unhashed_bool;
+ ret = gf_string2boolean(temp_str, &search_unhashed_bool);
+ if (ret == -1) {
+ goto err;
+ }
+ conf->search_unhashed = search_unhashed_bool
+ ? GF_DHT_LOOKUP_UNHASHED_ON
+ : GF_DHT_LOOKUP_UNHASHED_OFF;
+ } else {
+ conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
+ }
+ }
- if (defrag) {
- defrag->lock_migration_enabled = conf->lock_migration_enabled;
+ GF_OPTION_INIT("lookup-optimize", conf->lookup_optimize, bool, err);
- GF_OPTION_INIT ("rebalance-stats", defrag->stats, bool, err);
- if (dict_get_str (this->options, "rebalance-filter", &temp_str)
- == 0) {
- if (gf_defrag_pattern_list_fill (this, defrag, temp_str)
- == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_INVALID_OPTION,
- "Invalid option:"
- " Cannot parse rebalance-filter (%s)",
- temp_str);
+ GF_OPTION_INIT("unhashed-sticky-bit", conf->unhashed_sticky_bit, bool, err);
- goto err;
- }
- }
- }
+ GF_OPTION_INIT("use-readdirp", conf->use_readdirp, bool, err);
- /* option can be any one of percent or bytes */
- conf->disk_unit = 0;
- if (conf->min_free_disk < 100)
- conf->disk_unit = 'p';
+ GF_OPTION_INIT("min-free-disk", conf->min_free_disk, percent_or_size, err);
- ret = dht_init_subvolumes (this, conf);
- if (ret == -1) {
- goto err;
- }
+ GF_OPTION_INIT("min-free-inodes", conf->min_free_inodes, percent, err);
- if (cmd) {
- ret = dht_init_local_subvolumes (this, conf);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_INIT_LOCAL_SUBVOL_FAILED,
- "dht_init_local_subvolumes failed");
- goto err;
- }
- }
+ conf->dir_spread_cnt = conf->subvolume_cnt;
+ GF_OPTION_INIT("directory-layout-spread", conf->dir_spread_cnt, uint32,
+ err);
- if (dict_get_str (this->options, "decommissioned-bricks", &temp_str) == 0) {
- ret = dht_parse_decommissioned_bricks (this, conf, temp_str);
- if (ret == -1)
- goto err;
- }
+ GF_OPTION_INIT("assert-no-child-down", conf->assert_no_child_down, bool,
+ err);
- dht_init_regex (this, this->options, "rsync-hash-regex",
- &conf->rsync_regex, &conf->rsync_regex_valid, conf);
- dht_init_regex (this, this->options, "extra-hash-regex",
- &conf->extra_regex, &conf->extra_regex_valid, conf);
+ GF_OPTION_INIT("readdir-optimize", conf->readdir_optimize, bool, err);
- ret = dht_layouts_init (this, conf);
- if (ret == -1) {
- goto err;
- }
+ GF_OPTION_INIT("lock-migration", conf->lock_migration_enabled, bool, err);
+
+ GF_OPTION_INIT("force-migration", conf->force_migration, bool, err);
+ if (defrag) {
+ defrag->lock_migration_enabled = conf->lock_migration_enabled;
- conf->gen = 1;
+ GF_OPTION_INIT("rebalance-stats", defrag->stats, bool, err);
+ if (dict_get_str(this->options, "rebalance-filter", &temp_str) == 0) {
+ if (gf_defrag_pattern_list_fill(this, defrag, temp_str) == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_INVALID_OPTION,
+ "Invalid option:"
+ " Cannot parse rebalance-filter (%s)",
+ temp_str);
- this->local_pool = mem_pool_new (dht_local_t, 512);
- if (!this->local_pool) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_NO_MEMORY,
- " DHT initialisation failed. "
- "failed to create local_t's memory pool");
goto err;
+ }
}
-
- GF_OPTION_INIT ("randomize-hash-range-by-gfid",
- conf->randomize_by_gfid, bool, err);
-
- if (defrag) {
- GF_OPTION_INIT ("rebal-throttle", temp_str, str, err);
- if (temp_str) {
- ret = dht_configure_throttle (this, conf, temp_str);
- if (ret == -1)
- goto err;
- }
+ }
+
+ /* option can be any one of percent or bytes */
+ conf->disk_unit = 0;
+ if (conf->min_free_disk < 100)
+ conf->disk_unit = 'p';
+
+ ret = dht_init_subvolumes(this, conf);
+ if (ret == -1) {
+ goto err;
+ }
+
+ if (cmd) {
+ ret = dht_init_local_subvolumes(this, conf);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ DHT_MSG_INIT_LOCAL_SUBVOL_FAILED,
+ "dht_init_local_subvolumes failed");
+ goto err;
}
-
- GF_OPTION_INIT ("xattr-name", conf->xattr_name, str, err);
- gf_asprintf (&conf->mds_xattr_key, "%s."DHT_MDS_STR, conf->xattr_name);
- gf_asprintf (&conf->link_xattr_name, "%s."DHT_LINKFILE_STR,
- conf->xattr_name);
- gf_asprintf (&conf->commithash_xattr_name, "%s."DHT_COMMITHASH_STR,
- conf->xattr_name);
- gf_asprintf (&conf->wild_xattr_name, "%s*", conf->xattr_name);
- if (!conf->link_xattr_name || !conf->wild_xattr_name) {
+ }
+
+ if (dict_get_str(this->options, "decommissioned-bricks", &temp_str) == 0) {
+ ret = dht_parse_decommissioned_bricks(this, conf, temp_str);
+ if (ret == -1)
+ goto err;
+ }
+
+ dht_init_regex(this, this->options, "rsync-hash-regex", &conf->rsync_regex,
+ &conf->rsync_regex_valid, conf);
+ dht_init_regex(this, this->options, "extra-hash-regex", &conf->extra_regex,
+ &conf->extra_regex_valid, conf);
+
+ ret = dht_layouts_init(this, conf);
+ if (ret == -1) {
+ goto err;
+ }
+
+ conf->gen = 1;
+
+ this->local_pool = mem_pool_new(dht_local_t, 512);
+ if (!this->local_pool) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, DHT_MSG_NO_MEMORY,
+ " DHT initialisation failed. "
+ "failed to create local_t's memory pool");
+ goto err;
+ }
+
+ GF_OPTION_INIT("randomize-hash-range-by-gfid", conf->randomize_by_gfid,
+ bool, err);
+
+ if (defrag) {
+ GF_OPTION_INIT("rebal-throttle", temp_str, str, err);
+ if (temp_str) {
+ ret = dht_configure_throttle(this, conf, temp_str);
+ if (ret == -1)
goto err;
}
+ }
- GF_OPTION_INIT ("weighted-rebalance", conf->do_weighting, bool, err);
+ GF_OPTION_INIT("xattr-name", conf->xattr_name, str, err);
+ gf_asprintf(&conf->mds_xattr_key, "%s." DHT_MDS_STR, conf->xattr_name);
+ gf_asprintf(&conf->link_xattr_name, "%s." DHT_LINKFILE_STR,
+ conf->xattr_name);
+ gf_asprintf(&conf->commithash_xattr_name, "%s." DHT_COMMITHASH_STR,
+ conf->xattr_name);
+ gf_asprintf(&conf->wild_xattr_name, "%s*", conf->xattr_name);
+ if (!conf->link_xattr_name || !conf->wild_xattr_name) {
+ goto err;
+ }
- conf->lock_pool = mem_pool_new (dht_lock_t, 512);
- if (!conf->lock_pool) {
- gf_msg (this->name, GF_LOG_ERROR, 0, DHT_MSG_INIT_FAILED,
- "failed to create lock mem_pool, failing "
- "initialization");
- goto err;
- }
+ GF_OPTION_INIT("weighted-rebalance", conf->do_weighting, bool, err);
- this->private = conf;
+ conf->lock_pool = mem_pool_new(dht_lock_t, 512);
+ if (!conf->lock_pool) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_INIT_FAILED,
+ "failed to create lock mem_pool, failing "
+ "initialization");
+ goto err;
+ }
- if (dht_set_subvol_range(this))
- goto err;
+ this->private = conf;
- if (dht_init_methods (this))
- goto err;
+ if (dht_set_subvol_range(this))
+ goto err;
- return 0;
+ if (dht_init_methods(this))
+ goto err;
+
+ return 0;
err:
- if (conf) {
- if (conf->file_layouts) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- GF_FREE (conf->file_layouts[i]);
- }
- GF_FREE (conf->file_layouts);
- }
+ if (conf) {
+ if (conf->file_layouts) {
+ for (i = 0; i < conf->subvolume_cnt; i++) {
+ GF_FREE(conf->file_layouts[i]);
+ }
+ GF_FREE(conf->file_layouts);
+ }
- GF_FREE (conf->subvolumes);
+ GF_FREE(conf->subvolumes);
- GF_FREE (conf->subvolume_status);
+ GF_FREE(conf->subvolume_status);
- GF_FREE (conf->du_stats);
+ GF_FREE(conf->du_stats);
- GF_FREE (conf->defrag);
+ GF_FREE(conf->defrag);
- GF_FREE (conf->xattr_name);
- GF_FREE (conf->link_xattr_name);
- GF_FREE (conf->wild_xattr_name);
- GF_FREE (conf->mds_xattr_key);
+ GF_FREE(conf->xattr_name);
+ GF_FREE(conf->link_xattr_name);
+ GF_FREE(conf->wild_xattr_name);
+ GF_FREE(conf->mds_xattr_key);
- if (conf->lock_pool)
- mem_pool_destroy (conf->lock_pool);
+ if (conf->lock_pool)
+ mem_pool_destroy(conf->lock_pool);
- GF_FREE (conf);
- }
+ GF_FREE(conf);
+ }
- return -1;
+ return -1;
}
-
-struct volume_options options[] = {
- { .key = {"lookup-unhashed"},
- .value = {"auto", "yes", "no", "enable", "disable", "1", "0",
- "on", "off"},
- .type = GF_OPTION_TYPE_STR,
- .default_value = "on",
- .description = "This option if set to ON, does a lookup through "
- "all the sub-volumes, in case a lookup didn't return any result "
- "from the hash subvolume. If set to OFF, it does not do a lookup "
- "on the remaining subvolumes.",
- .op_version = {1},
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
- .level = OPT_STATUS_BASIC,
- },
- { .key = {"lookup-optimize"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .description = "This option if set to ON enables the optimization "
- "of -ve lookups, by not doing a lookup on non-hashed subvolumes for "
- "files, in case the hashed subvolume does not return any result. "
- "This option disregards the lookup-unhashed setting, when enabled.",
- .op_version = {GD_OP_VERSION_3_7_2},
- .level = OPT_STATUS_ADVANCED,
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC
- },
- { .key = {"min-free-disk"},
- .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
- .default_value = "10%",
- .description = "Percentage/Size of disk space, after which the "
- "process starts balancing out the cluster, and logs will appear "
- "in log files",
- .op_version = {1},
- .level = OPT_STATUS_BASIC,
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC
- },
- { .key = {"min-free-inodes"},
- .type = GF_OPTION_TYPE_PERCENT,
- .default_value = "5%",
- .description = "after system has only N% of inodes, warnings "
- "starts to appear in log files",
- .op_version = {1},
- .level = OPT_STATUS_BASIC,
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC
- },
- { .key = {"unhashed-sticky-bit"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- },
- { .key = {"use-readdirp"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .description = "This option if set to ON, forces the use of "
- "readdirp, and hence also displays the stats of the files.",
- .level = OPT_STATUS_ADVANCED,
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC
- },
- { .key = {"assert-no-child-down"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "This option if set to ON, in the event of "
- "CHILD_DOWN, will call exit."
- },
- { .key = {"directory-layout-spread"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .validate = GF_OPT_VALIDATE_MIN,
- .description = "Specifies the directory layout spread. Takes number "
- "of subvolumes as default value.",
-
- .op_version = {2},
- },
- { .key = {"decommissioned-bricks"},
- .type = GF_OPTION_TYPE_ANY,
- .description = "This option if set to ON, decommissions "
- "the brick, so that no new data is allowed to be created "
- "on that brick.",
- .level = OPT_STATUS_ADVANCED,
- },
- { .key = {"rebalance-cmd"},
- .type = GF_OPTION_TYPE_INT,
- },
- { .key = {"commit-hash"},
- .type = GF_OPTION_TYPE_INT,
- },
- { .key = {"node-uuid"},
- .type = GF_OPTION_TYPE_STR,
- },
- { .key = {"rebalance-stats"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "This option if set to ON displays and logs the "
- " time taken for migration of each file, during the rebalance "
- "process. If set to OFF, the rebalance logs will only display the "
- "time spent in each directory.",
- .op_version = {2},
- .level = OPT_STATUS_BASIC,
- },
- { .key = {"readdir-optimize"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "This option if set to ON enables the optimization "
- "that allows DHT to requests non-first subvolumes to filter out "
- "directory entries.",
- .op_version = {1},
- .level = OPT_STATUS_ADVANCED,
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC
- },
- { .key = {"rsync-hash-regex"},
- .type = GF_OPTION_TYPE_STR,
- /* Setting a default here doesn't work. See dht_init_regex. */
- .description = "Regular expression for stripping temporary-file "
- "suffix and prefix used by rsync, to prevent relocation when the "
- "file is renamed.",
- .op_version = {3},
- .level = OPT_STATUS_BASIC,
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC
- },
- { .key = {"extra-hash-regex"},
- .type = GF_OPTION_TYPE_STR,
- /* Setting a default here doesn't work. See dht_init_regex. */
- .description = "Regular expression for stripping temporary-file "
- "suffix and prefix used by an application, to prevent relocation when "
- "the file is renamed.",
- .op_version = {3},
- .level = OPT_STATUS_BASIC,
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC
- },
- { .key = {"rebalance-filter"},
- .type = GF_OPTION_TYPE_STR,
- },
-
- { .key = {"xattr-name"},
- .type = GF_OPTION_TYPE_STR,
- .default_value = "trusted.glusterfs.dht",
- .description = "Base for extended attributes used by this "
- "translator instance, to avoid conflicts with others above or "
- "below it.",
- .op_version = {3},
- },
-
- { .key = {"weighted-rebalance"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .description = "When enabled, files will be allocated to bricks "
- "with a probability proportional to their size. Otherwise, all "
- "bricks will have the same probability (legacy behavior).",
- .op_version = {GD_OP_VERSION_3_6_0},
- .level = OPT_STATUS_BASIC,
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC
- },
-
- /* NUFA option */
- { .key = {"local-volume-name"},
- .type = GF_OPTION_TYPE_XLATOR
- },
-
- /* tier options */
- { .key = {"tier-pause"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- },
-
- { .key = {"tier-promote-frequency"},
- .type = GF_OPTION_TYPE_INT,
- .default_value = "120",
- },
-
- { .key = {"tier-demote-frequency"},
- .type = GF_OPTION_TYPE_INT,
- .default_value = "3600",
- },
-
- { .key = {"write-freq-threshold"},
- .type = GF_OPTION_TYPE_INT,
- .default_value = "0",
- },
-
- { .key = {"read-freq-threshold"},
- .type = GF_OPTION_TYPE_INT,
- .default_value = "0",
- },
- { .key = {"watermark-hi"},
- .type = GF_OPTION_TYPE_PERCENT,
- .default_value = "90",
- },
- { .key = {"watermark-low"},
- .type = GF_OPTION_TYPE_PERCENT,
- .default_value = "75",
- },
- { .key = {"tier-mode"},
- .type = GF_OPTION_TYPE_STR,
- .default_value = "test",
- },
- { .key = {"tier-compact"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- },
- { .key = {"tier-hot-compact-frequency"},
- .type = GF_OPTION_TYPE_INT,
- .default_value = "604800",
- .description = "Frequency to compact DBs on hot tier in system"
- },
- { .key = {"tier-cold-compact-frequency"},
- .type = GF_OPTION_TYPE_INT,
- .default_value = "604800",
- .description = "Frequency to compact DBs on cold tier in system"
- },
- { .key = {"tier-max-mb"},
- .type = GF_OPTION_TYPE_INT,
- .default_value = "4000",
- },
- { .key = {"tier-max-promote-file-size"},
- .type = GF_OPTION_TYPE_INT,
- .default_value = "0",
- },
- { .key = {"tier-max-files"},
- .type = GF_OPTION_TYPE_INT,
- .default_value = "10000",
- },
- { .key = {"tier-query-limit"},
- .type = GF_OPTION_TYPE_INT,
- .default_value = "100",
- },
- /* switch option */
- { .key = {"pattern.switch.case"},
- .type = GF_OPTION_TYPE_ANY
- },
-
- { .key = {"randomize-hash-range-by-gfid"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "Use gfid of directory to determine the subvolume "
- "from which hash ranges are allocated starting with 0. "
- "Note that we still use a directory/file's name to determine the "
- "subvolume to which it hashes",
- .op_version = {GD_OP_VERSION_3_6_0},
- },
-
- { .key = {"rebal-throttle"},
- .type = GF_OPTION_TYPE_STR,
- .default_value = "normal",
- .description = " Sets the maximum number of parallel file migrations "
- "allowed on a node during the rebalance operation. The"
- " default value is normal and allows a max of "
- "[($(processing units) - 4) / 2), 2] files to be "
- "migrated at a time. Lazy will allow only one file to "
- "be migrated at a time and aggressive will allow "
- "max of [($(processing units) - 4) / 2), 4]",
- .op_version = {GD_OP_VERSION_3_7_0},
- .level = OPT_STATUS_BASIC,
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC
-
- },
-
- { .key = {"lock-migration"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = " If enabled this feature will migrate the posix locks"
- " associated with a file during rebalance",
- .op_version = {GD_OP_VERSION_3_8_0},
- .level = OPT_STATUS_ADVANCED,
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC
- },
-
- { .key = {"force-migration"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "If disabled, rebalance will not migrate files that "
- "are being written to by an application",
- .op_version = {GD_OP_VERSION_4_0_0},
- .level = OPT_STATUS_ADVANCED,
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC
- },
-
- { .key = {NULL} },
+struct volume_options dht_options[] = {
+ {
+ .key = {"lookup-unhashed"},
+ .value = {"auto", "yes", "no", "enable", "disable", "1", "0", "on",
+ "off"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = "on",
+ .description =
+ "This option if set to ON, does a lookup through "
+ "all the sub-volumes, in case a lookup didn't return any result "
+ "from the hash subvolume. If set to OFF, it does not do a lookup "
+ "on the remaining subvolumes.",
+ .op_version = {1},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
+ .level = OPT_STATUS_BASIC,
+ },
+ {.key = {"lookup-optimize"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description =
+ "This option if set to ON enables the optimization "
+ "of -ve lookups, by not doing a lookup on non-hashed subvolumes for "
+ "files, in case the hashed subvolume does not return any result. "
+ "This option disregards the lookup-unhashed setting, when enabled.",
+ .op_version = {GD_OP_VERSION_3_7_2},
+ .level = OPT_STATUS_ADVANCED,
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC},
+ {.key = {"min-free-disk"},
+ .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
+ .default_value = "10%",
+ .description =
+ "Percentage/Size of disk space, after which the "
+ "process starts balancing out the cluster, and logs will appear "
+ "in log files",
+ .op_version = {1},
+ .level = OPT_STATUS_BASIC,
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC},
+ {.key = {"min-free-inodes"},
+ .type = GF_OPTION_TYPE_PERCENT,
+ .default_value = "5%",
+ .description = "after system has only N% of inodes, warnings "
+ "starts to appear in log files",
+ .op_version = {1},
+ .level = OPT_STATUS_BASIC,
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC},
+ {
+ .key = {"unhashed-sticky-bit"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ },
+ {.key = {"use-readdirp"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "This option if set to ON, forces the use of "
+ "readdirp, and hence also displays the stats of the files.",
+ .level = OPT_STATUS_ADVANCED,
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC},
+ {.key = {"assert-no-child-down"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "This option if set to ON, in the event of "
+ "CHILD_DOWN, will call exit."},
+ {
+ .key = {"directory-layout-spread"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .validate = GF_OPT_VALIDATE_MIN,
+ .description = "Specifies the directory layout spread. Takes number "
+ "of subvolumes as default value.",
+
+ .op_version = {2},
+ },
+ {
+ .key = {"decommissioned-bricks"},
+ .type = GF_OPTION_TYPE_ANY,
+ .description =
+ "This option if set to ON, decommissions "
+ "the brick, so that no new data is allowed to be created "
+ "on that brick.",
+ .level = OPT_STATUS_ADVANCED,
+ },
+ {
+ .key = {"rebalance-cmd"},
+ .type = GF_OPTION_TYPE_INT,
+ },
+ {
+ .key = {"commit-hash"},
+ .type = GF_OPTION_TYPE_INT,
+ },
+ {
+ .key = {"node-uuid"},
+ .type = GF_OPTION_TYPE_STR,
+ },
+ {
+ .key = {"rebalance-stats"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description =
+ "This option if set to ON displays and logs the "
+ " time taken for migration of each file, during the rebalance "
+ "process. If set to OFF, the rebalance logs will only display the "
+ "time spent in each directory.",
+ .op_version = {2},
+ .level = OPT_STATUS_BASIC,
+ },
+ {.key = {"readdir-optimize"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description =
+ "This option if set to ON enables the optimization "
+ "that allows DHT to requests non-first subvolumes to filter out "
+ "directory entries.",
+ .op_version = {1},
+ .level = OPT_STATUS_ADVANCED,
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC},
+ {.key = {"rsync-hash-regex"},
+ .type = GF_OPTION_TYPE_STR,
+ /* Setting a default here doesn't work. See dht_init_regex. */
+ .description =
+ "Regular expression for stripping temporary-file "
+ "suffix and prefix used by rsync, to prevent relocation when the "
+ "file is renamed.",
+ .op_version = {3},
+ .level = OPT_STATUS_BASIC,
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC},
+ {.key = {"extra-hash-regex"},
+ .type = GF_OPTION_TYPE_STR,
+ /* Setting a default here doesn't work. See dht_init_regex. */
+ .description =
+ "Regular expression for stripping temporary-file "
+ "suffix and prefix used by an application, to prevent relocation when "
+ "the file is renamed.",
+ .op_version = {3},
+ .level = OPT_STATUS_BASIC,
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC},
+ {
+ .key = {"rebalance-filter"},
+ .type = GF_OPTION_TYPE_STR,
+ },
+
+ {
+ .key = {"xattr-name"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = "trusted.glusterfs.dht",
+ .description =
+ "Base for extended attributes used by this "
+ "translator instance, to avoid conflicts with others above or "
+ "below it.",
+ .op_version = {3},
+ },
+
+ {.key = {"weighted-rebalance"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description =
+ "When enabled, files will be allocated to bricks "
+ "with a probability proportional to their size. Otherwise, all "
+ "bricks will have the same probability (legacy behavior).",
+ .op_version = {GD_OP_VERSION_3_6_0},
+ .level = OPT_STATUS_BASIC,
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC},
+
+ /* NUFA option */
+ {.key = {"local-volume-name"}, .type = GF_OPTION_TYPE_XLATOR},
+
+ /* switch option */
+ {.key = {"pattern.switch.case"}, .type = GF_OPTION_TYPE_ANY},
+
+ {
+ .key = {"randomize-hash-range-by-gfid"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description =
+ "Use gfid of directory to determine the subvolume "
+ "from which hash ranges are allocated starting with 0. "
+ "Note that we still use a directory/file's name to determine the "
+ "subvolume to which it hashes",
+ .op_version = {GD_OP_VERSION_3_6_0},
+ },
+
+ {.key = {"rebal-throttle"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = "normal",
+ .description = " Sets the maximum number of parallel file migrations "
+ "allowed on a node during the rebalance operation. The"
+ " default value is normal and allows a max of "
+ "[($(processing units) - 4) / 2), 2] files to be "
+ "migrated at a time. Lazy will allow only one file to "
+ "be migrated at a time and aggressive will allow "
+ "max of [($(processing units) - 4) / 2), 4]",
+ .op_version = {GD_OP_VERSION_3_7_0},
+ .level = OPT_STATUS_BASIC,
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC
+
+ },
+
+ {.key = {"lock-migration"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = " If enabled this feature will migrate the posix locks"
+ " associated with a file during rebalance",
+ .op_version = {GD_OP_VERSION_3_8_0},
+ .level = OPT_STATUS_ADVANCED,
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC},
+
+ {.key = {"force-migration"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "If disabled, rebalance will not migrate files that "
+ "are being written to by an application",
+ .op_version = {GD_OP_VERSION_4_0_0},
+ .level = OPT_STATUS_ADVANCED,
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC},
+
+ {.key = {NULL}},
};
+
+#define NUM_DHT_OPTIONS (sizeof(dht_options) / sizeof(dht_options[0]))
+
+extern struct volume_options options[NUM_DHT_OPTIONS]
+ __attribute__((alias("dht_options")));
diff --git a/xlators/cluster/dht/src/dht.c b/xlators/cluster/dht/src/dht.c
index 90962636d18..53de8292704 100644
--- a/xlators/cluster/dht/src/dht.c
+++ b/xlators/cluster/dht/src/dht.c
@@ -8,78 +8,116 @@
cases as published by the Free Software Foundation.
*/
-
-#include "statedump.h"
#include "dht-common.h"
-class_methods_t class_methods = {
- .init = dht_init,
- .fini = dht_fini,
- .reconfigure = dht_reconfigure,
- .notify = dht_notify
+struct xlator_fops dht_pt_fops = {
+ /* we need to keep mkdir to make sure we
+ have layout on new directory */
+ .mkdir = dht_pt_mkdir,
+ .getxattr = dht_pt_getxattr,
+ .fgetxattr = dht_pt_fgetxattr,
+
+ /* required to trace fop properly in changelog */
+ .rename = dht_pt_rename,
+
+ /* FIXME: commenting the '.lookup()' below made some of
+ the failing tests to pass. I would remove the below
+ line, but keeping it here as a reminder for people
+ to check for issues if they find concerns with DHT
+ pass-through logic */
+ /*
+ .lookup = dht_lookup,
+ .readdir = dht_readdir,
+ .readdirp = dht_readdirp,
+ */
+ /* Keeping above as commented, mainly to support the
+ usecase of a gluster volume getting to 1x(anytype),
+ due to remove-brick (shrinking) exercise. In that case,
+ we would need above fops to be available, so we can
+ handle the case of dangling linkto files (if any) */
};
struct xlator_fops fops = {
- .ipc = dht_ipc,
- .lookup = dht_lookup,
- .mknod = dht_mknod,
- .create = dht_create,
+ .ipc = dht_ipc,
+ .lookup = dht_lookup,
+ .mknod = dht_mknod,
+ .create = dht_create,
- .open = dht_open,
- .statfs = dht_statfs,
- .opendir = dht_opendir,
- .readdir = dht_readdir,
- .readdirp = dht_readdirp,
- .fsyncdir = dht_fsyncdir,
- .symlink = dht_symlink,
- .unlink = dht_unlink,
- .link = dht_link,
- .mkdir = dht_mkdir,
- .rmdir = dht_rmdir,
- .rename = dht_rename,
- .entrylk = dht_entrylk,
- .fentrylk = dht_fentrylk,
+ .open = dht_open,
+ .statfs = dht_statfs,
+ .opendir = dht_opendir,
+ .readdir = dht_readdir,
+ .readdirp = dht_readdirp,
+ .fsyncdir = dht_fsyncdir,
+ .symlink = dht_symlink,
+ .unlink = dht_unlink,
+ .link = dht_link,
+ .mkdir = dht_mkdir,
+ .rmdir = dht_rmdir,
+ .rename = dht_rename,
+ .entrylk = dht_entrylk,
+ .fentrylk = dht_fentrylk,
- /* Inode read operations */
- .stat = dht_stat,
- .fstat = dht_fstat,
- .access = dht_access,
- .readlink = dht_readlink,
- .getxattr = dht_getxattr,
- .fgetxattr = dht_fgetxattr,
- .readv = dht_readv,
- .flush = dht_flush,
- .fsync = dht_fsync,
- .inodelk = dht_inodelk,
- .finodelk = dht_finodelk,
- .lk = dht_lk,
- .lease = dht_lease,
+ /* Inode read operations */
+ .stat = dht_stat,
+ .fstat = dht_fstat,
+ .access = dht_access,
+ .readlink = dht_readlink,
+ .getxattr = dht_getxattr,
+ .fgetxattr = dht_fgetxattr,
+ .readv = dht_readv,
+ .flush = dht_flush,
+ .fsync = dht_fsync,
+ .inodelk = dht_inodelk,
+ .finodelk = dht_finodelk,
+ .lk = dht_lk,
+ .lease = dht_lease,
- /* Inode write operations */
- .fremovexattr = dht_fremovexattr,
- .removexattr = dht_removexattr,
- .setxattr = dht_setxattr,
- .fsetxattr = dht_fsetxattr,
- .truncate = dht_truncate,
- .ftruncate = dht_ftruncate,
- .writev = dht_writev,
- .xattrop = dht_xattrop,
- .fxattrop = dht_fxattrop,
- .setattr = dht_setattr,
- .fsetattr = dht_fsetattr,
- .fallocate = dht_fallocate,
- .discard = dht_discard,
- .zerofill = dht_zerofill,
+ /* Inode write operations */
+ .fremovexattr = dht_fremovexattr,
+ .removexattr = dht_removexattr,
+ .setxattr = dht_setxattr,
+ .fsetxattr = dht_fsetxattr,
+ .truncate = dht_truncate,
+ .ftruncate = dht_ftruncate,
+ .writev = dht_writev,
+ .xattrop = dht_xattrop,
+ .fxattrop = dht_fxattrop,
+ .setattr = dht_setattr,
+ .fsetattr = dht_fsetattr,
+ .fallocate = dht_fallocate,
+ .discard = dht_discard,
+ .zerofill = dht_zerofill,
};
struct xlator_dumpops dumpops = {
- .priv = dht_priv_dump,
- .inodectx = dht_inodectx_dump,
+ .priv = dht_priv_dump,
+ .inodectx = dht_inodectx_dump,
};
-
struct xlator_cbks cbks = {
- .release = dht_release,
-// .releasedir = dht_releasedir,
- .forget = dht_forget
+ .release = dht_release,
+ // .releasedir = dht_releasedir,
+ .forget = dht_forget,
+};
+
+extern int32_t
+mem_acct_init(xlator_t *this);
+
+extern struct volume_options dht_options[];
+
+xlator_api_t xlator_api = {
+ .init = dht_init,
+ .fini = dht_fini,
+ .notify = dht_notify,
+ .reconfigure = dht_reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .dumpops = &dumpops,
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = dht_options,
+ .identifier = "distribute",
+ .pass_through_fops = &dht_pt_fops,
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/cluster/dht/src/dht.sym b/xlators/cluster/dht/src/dht.sym
deleted file mode 100644
index 780b5fc0387..00000000000
--- a/xlators/cluster/dht/src/dht.sym
+++ /dev/null
@@ -1,8 +0,0 @@
-fops
-cbks
-class_methods
-dht_methods
-options
-mem_acct_init
-reconfigure
-dumpops
diff --git a/xlators/cluster/dht/src/nufa.c b/xlators/cluster/dht/src/nufa.c
index afa446584ba..3648a564840 100644
--- a/xlators/cluster/dht/src/nufa.c
+++ b/xlators/cluster/dht/src/nufa.c
@@ -8,681 +8,650 @@
cases as published by the Free Software Foundation.
*/
-
#include "dht-common.h"
/* TODO: all 'TODO's in dht.c holds good */
-extern struct volume_options options[];
+extern struct volume_options dht_options[];
int
-nufa_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, dict_t *xattr,
- struct iatt *postparent)
+nufa_local_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *stbuf, dict_t *xattr,
+ struct iatt *postparent)
{
- xlator_t *subvol = NULL;
- char is_linkfile = 0;
- char is_dir = 0;
- dht_conf_t *conf = NULL;
- dht_local_t *local = NULL;
- loc_t *loc = NULL;
- int i = 0;
- xlator_t *prev = NULL;
- int call_cnt = 0;
- int ret = 0;
-
- conf = this->private;
-
- prev = cookie;
- local = frame->local;
- loc = &local->loc;
-
- if (ENTRY_MISSING (op_ret, op_errno)) {
- if (conf->search_unhashed) {
- local->op_errno = ENOENT;
- dht_lookup_everywhere (frame, this, loc);
- return 0;
- }
+ xlator_t *subvol = NULL;
+ char is_linkfile = 0;
+ char is_dir = 0;
+ dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ loc_t *loc = NULL;
+ int i = 0;
+ xlator_t *prev = NULL;
+ int call_cnt = 0;
+ int ret = 0;
+
+ conf = this->private;
+
+ prev = cookie;
+ local = frame->local;
+ loc = &local->loc;
+
+ if (ENTRY_MISSING(op_ret, op_errno)) {
+ if (conf->search_unhashed) {
+ local->op_errno = ENOENT;
+ dht_lookup_everywhere(frame, this, loc);
+ return 0;
}
-
- if (op_ret == -1)
- goto out;
-
- is_linkfile = check_is_linkfile (inode, stbuf, xattr,
- conf->link_xattr_name);
- is_dir = check_is_dir (inode, stbuf, xattr);
-
- if (!is_dir && !is_linkfile) {
- /* non-directory and not a linkfile */
- ret = dht_layout_preset (this, prev, inode);
- if (ret < 0) {
- gf_msg_debug (this->name, 0,
- "could not set pre-set layout for subvol"
- " %s", prev->name);
- op_ret = -1;
- op_errno = EINVAL;
- goto err;
- }
-
- goto out;
+ }
+
+ if (op_ret == -1)
+ goto out;
+
+ is_linkfile = check_is_linkfile(inode, stbuf, xattr, conf->link_xattr_name);
+ is_dir = check_is_dir(inode, stbuf, xattr);
+
+ if (!is_dir && !is_linkfile) {
+ /* non-directory and not a linkfile */
+ ret = dht_layout_preset(this, prev, inode);
+ if (ret < 0) {
+ gf_msg_debug(this->name, 0,
+ "could not set pre-set layout for subvol"
+ " %s",
+ prev->name);
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto err;
}
- if (is_dir) {
- call_cnt = conf->subvolume_cnt;
- local->call_cnt = call_cnt;
-
- local->inode = inode_ref (inode);
- local->xattr = dict_ref (xattr);
-
- local->op_ret = 0;
- local->op_errno = 0;
-
- local->layout = dht_layout_new (this, conf->subvolume_cnt);
- if (!local->layout) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto err;
- }
-
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND_COOKIE (frame, dht_lookup_dir_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->lookup,
- &local->loc, local->xattr_req);
- }
- }
+ goto out;
+ }
- if (is_linkfile) {
- subvol = dht_linkfile_subvol (this, inode, stbuf, xattr);
+ if (is_dir) {
+ call_cnt = conf->subvolume_cnt;
+ local->call_cnt = call_cnt;
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "linkfile has no link subvolume. path=%s",
- loc->path);
- dht_lookup_everywhere (frame, this, loc);
- return 0;
- }
+ local->inode = inode_ref(inode);
+ local->xattr = dict_ref(xattr);
- STACK_WIND_COOKIE (frame, dht_lookup_linkfile_cbk, subvol,
- subvol, subvol->fops->lookup,
- &local->loc, local->xattr_req);
+ local->op_ret = 0;
+ local->op_errno = 0;
+
+ local->layout = dht_layout_new(this, conf->subvolume_cnt);
+ if (!local->layout) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto err;
}
- return 0;
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND_COOKIE(frame, dht_lookup_dir_cbk, conf->subvolumes[i],
+ conf->subvolumes[i],
+ conf->subvolumes[i]->fops->lookup, &local->loc,
+ local->xattr_req);
+ }
+ }
-out:
- if (!local->hashed_subvol) {
- gf_msg_debug (this->name, 0,
- "no subvolume in layout for path=%s",
- local->loc.path);
- local->op_errno = ENOENT;
- dht_lookup_everywhere (frame, this, loc);
- return 0;
+ if (is_linkfile) {
+ subvol = dht_linkfile_subvol(this, inode, stbuf, xattr);
+
+ if (!subvol) {
+ gf_msg_debug(this->name, 0,
+ "linkfile has no link subvolume. path=%s", loc->path);
+ dht_lookup_everywhere(frame, this, loc);
+ return 0;
}
- STACK_WIND_COOKIE (frame, dht_lookup_cbk, local->hashed_subvol,
- local->hashed_subvol,
- local->hashed_subvol->fops->lookup,
- &local->loc, local->xattr_req);
+ STACK_WIND_COOKIE(frame, dht_lookup_linkfile_cbk, subvol, subvol,
+ subvol->fops->lookup, &local->loc, local->xattr_req);
+ }
+ return 0;
+
+out:
+ if (!local->hashed_subvol) {
+ gf_msg_debug(this->name, 0, "no subvolume in layout for path=%s",
+ local->loc.path);
+ local->op_errno = ENOENT;
+ dht_lookup_everywhere(frame, this, loc);
return 0;
+ }
+
+ STACK_WIND_COOKIE(frame, dht_lookup_cbk, local->hashed_subvol,
+ local->hashed_subvol, local->hashed_subvol->fops->lookup,
+ &local->loc, local->xattr_req);
+
+ return 0;
err:
- DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno,
- inode, stbuf, xattr, postparent);
- return 0;
+ DHT_STACK_UNWIND(lookup, frame, op_ret, op_errno, inode, stbuf, xattr,
+ postparent);
+ return 0;
}
int
-nufa_lookup (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xattr_req)
+nufa_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr_req)
{
- xlator_t *hashed_subvol = NULL;
- xlator_t *subvol = NULL;
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- int ret = -1;
- int op_errno = -1;
- dht_layout_t *layout = NULL;
- int i = 0;
- int call_cnt = 0;
-
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
-
- conf = this->private;
-
- local = dht_local_init (frame, loc, NULL, GF_FOP_LOOKUP);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
+ xlator_t *hashed_subvol = NULL;
+ xlator_t *subvol = NULL;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int ret = -1;
+ int op_errno = -1;
+ dht_layout_t *layout = NULL;
+ int i = 0;
+ int call_cnt = 0;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
+ VALIDATE_OR_GOTO(loc->path, err);
+
+ conf = this->private;
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_LOOKUP);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ if (xattr_req) {
+ local->xattr_req = dict_ref(xattr_req);
+ } else {
+ local->xattr_req = dict_new();
+ }
+
+ hashed_subvol = dht_subvol_get_hashed(this, &local->loc);
+
+ local->hashed_subvol = hashed_subvol;
+
+ if (is_revalidate(loc)) {
+ layout = local->layout;
+ if (!layout) {
+ gf_msg_debug(this->name, 0,
+ "revalidate lookup without cache. "
+ "path=%s",
+ loc->path);
+ op_errno = EINVAL;
+ goto err;
}
- if (xattr_req) {
- local->xattr_req = dict_ref (xattr_req);
- } else {
- local->xattr_req = dict_new ();
+ if (layout->gen && (layout->gen < conf->gen)) {
+ gf_msg_debug(this->name, 0, "incomplete layout failure for path=%s",
+ loc->path);
+ dht_layout_unref(this, local->layout);
+ goto do_fresh_lookup;
}
- hashed_subvol = dht_subvol_get_hashed (this, &local->loc);
-
- local->hashed_subvol = hashed_subvol;
-
- if (is_revalidate (loc)) {
- layout = local->layout;
- if (!layout) {
- gf_msg_debug (this->name, 0,
- "revalidate lookup without cache. "
- "path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
-
- if (layout->gen && (layout->gen < conf->gen)) {
- gf_msg_debug (this->name, 0,
- "incomplete layout failure for path=%s",
- loc->path);
- dht_layout_unref (this, local->layout);
- goto do_fresh_lookup;
- }
-
- local->inode = inode_ref (loc->inode);
-
- local->call_cnt = layout->cnt;
- call_cnt = local->call_cnt;
-
- /* NOTE: we don't require 'trusted.glusterfs.dht.linkto' attribute,
- * revalidates directly go to the cached-subvolume.
- */
- ret = dict_set_uint32 (local->xattr_req,
- conf->xattr_name, 4 * 4);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dict value.");
- op_errno = -1;
- goto err;
- }
-
- for (i = 0; i < layout->cnt; i++) {
- subvol = layout->list[i].xlator;
-
- STACK_WIND_COOKIE (frame, dht_revalidate_cbk, subvol,
- subvol, subvol->fops->lookup,
- loc, local->xattr_req);
-
- if (!--call_cnt)
- break;
- }
- } else {
- do_fresh_lookup:
- ret = dict_set_uint32 (local->xattr_req,
- conf->xattr_name, 4 * 4);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dict value.");
- op_errno = -1;
- goto err;
- }
-
- ret = dict_set_uint32 (local->xattr_req,
- conf->link_xattr_name, 256);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dict value.");
- op_errno = -1;
- goto err;
- }
-
- /* Send it to only local volume */
- STACK_WIND_COOKIE (frame, nufa_local_lookup_cbk,
- ((xlator_t *)conf->private),
- ((xlator_t *)conf->private),
- ((xlator_t *)conf->private)->fops->lookup,
- loc, local->xattr_req);
- }
+ local->inode = inode_ref(loc->inode);
- return 0;
+ local->call_cnt = layout->cnt;
+ call_cnt = local->call_cnt;
-err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL,
- NULL);
- return 0;
-}
+ /* NOTE: we don't require 'trusted.glusterfs.dht.linkto' attribute,
+ * revalidates directly go to the cached-subvolume.
+ */
+ ret = dict_set_uint32(local->xattr_req, conf->xattr_name, 4 * 4);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dict value.");
+ op_errno = -1;
+ goto err;
+ }
-int
-nufa_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- dht_local_t *local = NULL;
+ for (i = 0; i < layout->cnt; i++) {
+ subvol = layout->list[i].xlator;
- local = frame->local;
+ STACK_WIND_COOKIE(frame, dht_revalidate_cbk, subvol, subvol,
+ subvol->fops->lookup, loc, local->xattr_req);
- if (op_ret == -1)
- goto err;
+ if (!--call_cnt)
+ break;
+ }
+ } else {
+ do_fresh_lookup:
+ ret = dict_set_uint32(local->xattr_req, conf->xattr_name, 4 * 4);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dict value.");
+ op_errno = -1;
+ goto err;
+ }
+
+ ret = dict_set_uint32(local->xattr_req, conf->link_xattr_name, 256);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_DICT_SET_FAILED,
+ "Failed to set dict value.");
+ op_errno = -1;
+ goto err;
+ }
- STACK_WIND_COOKIE (frame, dht_create_cbk, local->cached_subvol,
- local->cached_subvol, local->cached_subvol->fops->create,
- &local->loc, local->flags, local->mode, local->umask,
- local->fd, local->params);
+ /* Send it to only local volume */
+ STACK_WIND_COOKIE(
+ frame, nufa_local_lookup_cbk, ((xlator_t *)conf->private),
+ ((xlator_t *)conf->private),
+ ((xlator_t *)conf->private)->fops->lookup, loc, local->xattr_req);
+ }
- return 0;
+ return 0;
err:
- DHT_STACK_UNWIND (create, frame, -1, op_errno,
- NULL, NULL, NULL, NULL, NULL, NULL);
- return 0;
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ return 0;
}
int
-nufa_create (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, mode_t mode,
- mode_t umask, fd_t *fd, dict_t *params)
+nufa_create_linkfile_create_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int op_ret, int op_errno,
+ inode_t *inode, struct iatt *stbuf,
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- xlator_t *subvol = NULL;
- xlator_t *avail_subvol = NULL;
- int op_errno = -1;
+ dht_local_t *local = NULL;
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
+ local = frame->local;
- conf = this->private;
+ if (op_ret == -1)
+ goto err;
- dht_get_du_info (frame, this, loc);
+ STACK_WIND_COOKIE(frame, dht_create_cbk, local->cached_subvol,
+ local->cached_subvol, local->cached_subvol->fops->create,
+ &local->loc, local->flags, local->mode, local->umask,
+ local->fd, local->params);
- local = dht_local_init (frame, loc, fd, GF_FOP_CREATE);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+ return 0;
- subvol = dht_subvol_get_hashed (this, loc);
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no subvolume in layout for path=%s",
- loc->path);
- op_errno = ENOENT;
- goto err;
- }
-
- avail_subvol = conf->private;
- if (dht_is_subvol_filled (this, (xlator_t *)conf->private)) {
- avail_subvol =
- dht_free_disk_available_subvol (this,
- (xlator_t *)conf->private,
- local);
- }
+err:
+ DHT_STACK_UNWIND(create, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
+}
- if (subvol != avail_subvol) {
- /* create a link file instead of actual file */
- local->params = dict_ref (params);
- local->mode = mode;
- local->flags = flags;
- local->umask = umask;
- local->cached_subvol = avail_subvol;
- dht_linkfile_create (frame, nufa_create_linkfile_create_cbk,
- this, avail_subvol, subvol, loc);
- return 0;
- }
+int
+nufa_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *params)
+{
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ xlator_t *subvol = NULL;
+ xlator_t *avail_subvol = NULL;
+ int op_errno = -1;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+
+ conf = this->private;
+
+ dht_get_du_info(frame, this, loc);
+
+ local = dht_local_init(frame, loc, fd, GF_FOP_CREATE);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ subvol = dht_subvol_get_hashed(this, loc);
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no subvolume in layout for path=%s",
+ loc->path);
+ op_errno = ENOENT;
+ goto err;
+ }
+
+ avail_subvol = conf->private;
+ if (dht_is_subvol_filled(this, (xlator_t *)conf->private)) {
+ avail_subvol = dht_free_disk_available_subvol(
+ this, (xlator_t *)conf->private, local);
+ }
+
+ if (subvol != avail_subvol) {
+ /* create a link file instead of actual file */
+ local->params = dict_ref(params);
+ local->mode = mode;
+ local->flags = flags;
+ local->umask = umask;
+ local->cached_subvol = avail_subvol;
+ dht_linkfile_create(frame, nufa_create_linkfile_create_cbk, this,
+ avail_subvol, subvol, loc);
+ return 0;
+ }
- gf_msg_trace (this->name, 0,
- "creating %s on %s", loc->path, subvol->name);
+ gf_msg_trace(this->name, 0, "creating %s on %s", loc->path, subvol->name);
- STACK_WIND_COOKIE (frame, dht_create_cbk, subvol,
- subvol, subvol->fops->create,
- loc, flags, mode, umask, fd, params);
+ STACK_WIND_COOKIE(frame, dht_create_cbk, subvol, subvol,
+ subvol->fops->create, loc, flags, mode, umask, fd,
+ params);
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (create, frame, -1, op_errno,
- NULL, NULL, NULL, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(create, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL,
+ NULL);
- return 0;
+ return 0;
}
int
-nufa_mknod_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode,
- struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+nufa_mknod_linkfile_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- dht_local_t *local = NULL;
+ dht_local_t *local = NULL;
- local = frame->local;
- if (!local || !local->cached_subvol) {
- op_errno = EINVAL;
- op_ret = -1;
- goto err;
- }
+ local = frame->local;
+ if (!local || !local->cached_subvol) {
+ op_errno = EINVAL;
+ op_ret = -1;
+ goto err;
+ }
- if (op_ret >= 0) {
- STACK_WIND_COOKIE (frame, dht_newfile_cbk,
- (void *)local->cached_subvol, local->cached_subvol,
- local->cached_subvol->fops->mknod,
- &local->loc, local->mode, local->rdev,
- local->umask, local->params);
+ if (op_ret >= 0) {
+ STACK_WIND_COOKIE(
+ frame, dht_newfile_cbk, (void *)local->cached_subvol,
+ local->cached_subvol, local->cached_subvol->fops->mknod,
+ &local->loc, local->mode, local->rdev, local->umask, local->params);
- return 0;
- }
+ return 0;
+ }
err:
- WIPE (postparent);
- WIPE (preparent);
+ WIPE(postparent);
+ WIPE(preparent);
- DHT_STACK_UNWIND (link, frame, op_ret, op_errno,
- inode, stbuf, preparent, postparent, xdata);
- return 0;
+ DHT_STACK_UNWIND(link, frame, op_ret, op_errno, inode, stbuf, preparent,
+ postparent, xdata);
+ return 0;
}
-
int
-nufa_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t rdev, mode_t umask, dict_t *params)
+nufa_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *params)
{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- xlator_t *subvol = NULL;
- xlator_t *avail_subvol = NULL;
- int op_errno = -1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
-
- conf = this->private;
-
- dht_get_du_info (frame, this, loc);
-
- local = dht_local_init (frame, loc, NULL, GF_FOP_MKNOD);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- subvol = dht_subvol_get_hashed (this, loc);
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no subvolume in layout for path=%s",
- loc->path);
- op_errno = ENOENT;
- goto err;
- }
-
- /* Consider the disksize in consideration */
- avail_subvol = conf->private;
- if (dht_is_subvol_filled (this, (xlator_t *)conf->private)) {
- avail_subvol =
- dht_free_disk_available_subvol (this,
- (xlator_t *)conf->private,
- local);
- }
-
- if (avail_subvol != subvol) {
- /* Create linkfile first */
-
- local->params = dict_ref (params);
- local->mode = mode;
- local->umask = umask;
- local->rdev = rdev;
- local->cached_subvol = avail_subvol;
-
- dht_linkfile_create (frame, nufa_mknod_linkfile_cbk, this,
- avail_subvol, subvol, loc);
- return 0;
- }
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ xlator_t *subvol = NULL;
+ xlator_t *avail_subvol = NULL;
+ int op_errno = -1;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+
+ conf = this->private;
+
+ dht_get_du_info(frame, this, loc);
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_MKNOD);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ subvol = dht_subvol_get_hashed(this, loc);
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no subvolume in layout for path=%s",
+ loc->path);
+ op_errno = ENOENT;
+ goto err;
+ }
+
+ /* Consider the disksize in consideration */
+ avail_subvol = conf->private;
+ if (dht_is_subvol_filled(this, (xlator_t *)conf->private)) {
+ avail_subvol = dht_free_disk_available_subvol(
+ this, (xlator_t *)conf->private, local);
+ }
+
+ if (avail_subvol != subvol) {
+ /* Create linkfile first */
+
+ local->params = dict_ref(params);
+ local->mode = mode;
+ local->umask = umask;
+ local->rdev = rdev;
+ local->cached_subvol = avail_subvol;
+
+ dht_linkfile_create(frame, nufa_mknod_linkfile_cbk, this, avail_subvol,
+ subvol, loc);
+ return 0;
+ }
- gf_msg_trace (this->name, 0,
- "creating %s on %s", loc->path, subvol->name);
+ gf_msg_trace(this->name, 0, "creating %s on %s", loc->path, subvol->name);
- STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol, subvol,
- subvol->fops->mknod, loc, mode, rdev, umask,
- params);
+ STACK_WIND_COOKIE(frame, dht_newfile_cbk, (void *)subvol, subvol,
+ subvol->fops->mknod, loc, mode, rdev, umask, params);
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
- NULL, NULL, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
gf_boolean_t
-same_first_part (char *str1, char term1, char *str2, char term2)
+same_first_part(char *str1, char term1, char *str2, char term2)
{
- gf_boolean_t ended1;
- gf_boolean_t ended2;
-
- for (;;) {
- ended1 = ((*str1 == '\0') || (*str1 == term1));
- ended2 = ((*str2 == '\0') || (*str2 == term2));
- if (ended1 && ended2) {
- return _gf_true;
- }
- if (ended1 || ended2 || (*str1 != *str2)) {
- return _gf_false;
- }
- ++str1;
- ++str2;
+ gf_boolean_t ended1;
+ gf_boolean_t ended2;
+
+ for (;;) {
+ ended1 = ((*str1 == '\0') || (*str1 == term1));
+ ended2 = ((*str2 == '\0') || (*str2 == term2));
+ if (ended1 && ended2) {
+ return _gf_true;
}
+ if (ended1 || ended2 || (*str1 != *str2)) {
+ return _gf_false;
+ }
+ ++str1;
+ ++str2;
+ }
}
typedef struct nufa_args {
- xlator_t *this;
- char *volname;
- gf_boolean_t addr_match;
+ xlator_t *this;
+ char *volname;
+ gf_boolean_t addr_match;
} nufa_args_t;
static void
-nufa_find_local_brick (xlator_t *xl, void *data)
+nufa_find_local_brick(xlator_t *xl, void *data)
{
- nufa_args_t *args = data;
- xlator_t *this = args->this;
- char *local_volname = args->volname;
- gf_boolean_t addr_match = args->addr_match;
- char *brick_host = NULL;
- dht_conf_t *conf = this->private;
- int ret = -1;
-
- /*This means a local subvol was already found. We pick the first brick
- * that is local*/
- if (conf->private)
- return;
-
- if (strcmp (xl->name, local_volname) == 0) {
- conf->private = xl;
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_SUBVOL_INFO,
- "Using specified subvol %s",
- local_volname);
- return;
- }
-
- if (!addr_match)
- return;
-
- ret = dict_get_str (xl->options, "remote-host", &brick_host);
- if ((ret == 0) &&
- (gf_is_same_address (local_volname, brick_host) ||
- gf_is_local_addr (brick_host))) {
- conf->private = xl;
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_SUBVOL_INFO, "Using the first local "
- "subvol %s", xl->name);
- return;
- }
-
+ nufa_args_t *args = data;
+ xlator_t *this = args->this;
+ char *local_volname = args->volname;
+ gf_boolean_t addr_match = args->addr_match;
+ char *brick_host = NULL;
+ dht_conf_t *conf = this->private;
+ int ret = -1;
+
+ /*This means a local subvol was already found. We pick the first brick
+ * that is local*/
+ if (conf->private)
+ return;
+
+ if (strcmp(xl->name, local_volname) == 0) {
+ conf->private = xl;
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_SUBVOL_INFO,
+ "Using specified subvol %s", local_volname);
+ return;
+ }
+
+ if (!addr_match)
+ return;
+
+ ret = dict_get_str(xl->options, "remote-host", &brick_host);
+ if ((ret == 0) && (gf_is_same_address(local_volname, brick_host) ||
+ gf_is_local_addr(brick_host))) {
+ conf->private = xl;
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_SUBVOL_INFO,
+ "Using the first local "
+ "subvol %s",
+ xl->name);
+ return;
+ }
}
static void
-nufa_to_dht (xlator_t *this)
+nufa_to_dht(xlator_t *this)
{
- GF_ASSERT (this);
- GF_ASSERT (this->fops);
+ GF_ASSERT(this);
+ GF_ASSERT(this->fops);
- this->fops->lookup = dht_lookup;
- this->fops->create = dht_create;
- this->fops->mknod = dht_mknod;
+ this->fops->lookup = dht_lookup;
+ this->fops->create = dht_create;
+ this->fops->mknod = dht_mknod;
}
int
-nufa_find_local_subvol (xlator_t *this,
- void (*fn) (xlator_t *each, void* data), void *data)
+nufa_find_local_subvol(xlator_t *this, void (*fn)(xlator_t *each, void *data),
+ void *data)
{
- int ret = -1;
- dht_conf_t *conf = this->private;
- xlator_list_t *trav = NULL;
- xlator_t *parent = NULL;
- xlator_t *candidate = NULL;
-
- xlator_foreach_depth_first (this, fn, data);
- if (!conf->private) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_BRICK_ERROR, "Couldn't find a local "
- "brick");
- return -1;
+ int ret = -1;
+ dht_conf_t *conf = this->private;
+ xlator_list_t *trav = NULL;
+ xlator_t *parent = NULL;
+ xlator_t *candidate = NULL;
+
+ xlator_foreach_depth_first(this, fn, data);
+ if (!conf->private) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_BRICK_ERROR,
+ "Couldn't find a local "
+ "brick");
+ return -1;
+ }
+
+ candidate = conf->private;
+ trav = candidate->parents;
+ while (trav) {
+ parent = trav->xlator;
+ if (strcmp(parent->type, "cluster/nufa") == 0) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_SUBVOL_INFO,
+ "Found local subvol, "
+ "%s",
+ candidate->name);
+ ret = 0;
+ conf->private = candidate;
+ break;
}
- candidate = conf->private;
- trav = candidate->parents;
- while (trav) {
-
- parent = trav->xlator;
- if (strcmp (parent->type, "cluster/nufa") == 0) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_SUBVOL_INFO, "Found local subvol, "
- "%s", candidate->name);
- ret = 0;
- conf->private = candidate;
- break;
- }
-
- candidate = parent;
- trav = parent->parents;
- }
+ candidate = parent;
+ trav = parent->parents;
+ }
- return ret;
+ return ret;
}
int
-nufa_init (xlator_t *this)
+nufa_init(xlator_t *this)
{
- data_t *data = NULL;
- char *local_volname = NULL;
- int ret = -1;
- char my_hostname[256];
- gf_boolean_t addr_match = _gf_false;
- nufa_args_t args = {0, };
-
- ret = dht_init(this);
- if (ret) {
- return ret;
- }
-
- if ((data = dict_get (this->options, "local-volume-name"))) {
- local_volname = data->data;
-
- } else {
- addr_match = _gf_true;
- local_volname = "localhost";
- ret = gethostname (my_hostname, 256);
- if (ret == 0)
- local_volname = my_hostname;
-
- else
- gf_msg (this->name, GF_LOG_WARNING, errno,
- DHT_MSG_GET_HOSTNAME_FAILED,
- "could not find hostname");
-
- }
-
- args.this = this;
- args.volname = local_volname;
- args.addr_match = addr_match;
- ret = nufa_find_local_subvol (this, nufa_find_local_brick, &args);
- if (ret) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_SUBVOL_INFO,
- "Unable to find local subvolume, switching "
- "to dht mode");
- nufa_to_dht (this);
- }
- return 0;
+ data_t *data = NULL;
+ char *local_volname = NULL;
+ int ret = -1;
+ char my_hostname[256];
+ gf_boolean_t addr_match = _gf_false;
+ nufa_args_t args = {
+ 0,
+ };
+
+ ret = dht_init(this);
+ if (ret) {
+ return ret;
+ }
+
+ if ((data = dict_get(this->options, "local-volume-name"))) {
+ local_volname = data->data;
+
+ } else {
+ addr_match = _gf_true;
+ local_volname = "localhost";
+ ret = gethostname(my_hostname, 256);
+ if (ret == 0)
+ local_volname = my_hostname;
+
+ else
+ gf_msg(this->name, GF_LOG_WARNING, errno,
+ DHT_MSG_GET_HOSTNAME_FAILED, "could not find hostname");
+ }
+
+ args.this = this;
+ args.volname = local_volname;
+ args.addr_match = addr_match;
+ ret = nufa_find_local_subvol(this, nufa_find_local_brick, &args);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_INFO, 0, DHT_MSG_SUBVOL_INFO,
+ "Unable to find local subvolume, switching "
+ "to dht mode");
+ nufa_to_dht(this);
+ }
+ return 0;
}
dht_methods_t dht_methods = {
- .migration_get_dst_subvol = dht_migration_get_dst_subvol,
- .migration_needed = dht_migration_needed,
- .layout_search = dht_layout_search,
-};
-
-class_methods_t class_methods = {
- .init = nufa_init,
- .fini = dht_fini,
- .reconfigure = dht_reconfigure,
- .notify = dht_notify
+ .migration_get_dst_subvol = dht_migration_get_dst_subvol,
+ .layout_search = dht_layout_search,
};
-
struct xlator_fops fops = {
- .lookup = nufa_lookup,
- .create = nufa_create,
- .mknod = nufa_mknod,
-
- .stat = dht_stat,
- .fstat = dht_fstat,
- .truncate = dht_truncate,
- .ftruncate = dht_ftruncate,
- .access = dht_access,
- .readlink = dht_readlink,
- .setxattr = dht_setxattr,
- .getxattr = dht_getxattr,
- .removexattr = dht_removexattr,
- .open = dht_open,
- .readv = dht_readv,
- .writev = dht_writev,
- .flush = dht_flush,
- .fsync = dht_fsync,
- .statfs = dht_statfs,
- .lk = dht_lk,
- .opendir = dht_opendir,
- .readdir = dht_readdir,
- .readdirp = dht_readdirp,
- .fsyncdir = dht_fsyncdir,
- .symlink = dht_symlink,
- .unlink = dht_unlink,
- .link = dht_link,
- .mkdir = dht_mkdir,
- .rmdir = dht_rmdir,
- .rename = dht_rename,
- .inodelk = dht_inodelk,
- .finodelk = dht_finodelk,
- .entrylk = dht_entrylk,
- .fentrylk = dht_fentrylk,
- .xattrop = dht_xattrop,
- .fxattrop = dht_fxattrop,
- .setattr = dht_setattr,
+ .lookup = nufa_lookup,
+ .create = nufa_create,
+ .mknod = nufa_mknod,
+
+ .stat = dht_stat,
+ .fstat = dht_fstat,
+ .truncate = dht_truncate,
+ .ftruncate = dht_ftruncate,
+ .access = dht_access,
+ .readlink = dht_readlink,
+ .setxattr = dht_setxattr,
+ .getxattr = dht_getxattr,
+ .removexattr = dht_removexattr,
+ .open = dht_open,
+ .readv = dht_readv,
+ .writev = dht_writev,
+ .flush = dht_flush,
+ .fsync = dht_fsync,
+ .statfs = dht_statfs,
+ .lk = dht_lk,
+ .opendir = dht_opendir,
+ .readdir = dht_readdir,
+ .readdirp = dht_readdirp,
+ .fsyncdir = dht_fsyncdir,
+ .symlink = dht_symlink,
+ .unlink = dht_unlink,
+ .link = dht_link,
+ .mkdir = dht_mkdir,
+ .rmdir = dht_rmdir,
+ .rename = dht_rename,
+ .inodelk = dht_inodelk,
+ .finodelk = dht_finodelk,
+ .entrylk = dht_entrylk,
+ .fentrylk = dht_fentrylk,
+ .xattrop = dht_xattrop,
+ .fxattrop = dht_fxattrop,
+ .setattr = dht_setattr,
};
-
-struct xlator_cbks cbks = {
- .forget = dht_forget
+struct xlator_cbks cbks = {.forget = dht_forget};
+extern int32_t
+mem_acct_init(xlator_t *this);
+
+xlator_api_t xlator_api = {
+ .init = nufa_init,
+ .fini = dht_fini,
+ .notify = dht_notify,
+ .reconfigure = dht_reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = dht_options,
+ .identifier = "nufa",
+ .category = GF_TECH_PREVIEW,
};
diff --git a/xlators/cluster/dht/src/nufa.sym b/xlators/cluster/dht/src/nufa.sym
deleted file mode 100644
index 780b5fc0387..00000000000
--- a/xlators/cluster/dht/src/nufa.sym
+++ /dev/null
@@ -1,8 +0,0 @@
-fops
-cbks
-class_methods
-dht_methods
-options
-mem_acct_init
-reconfigure
-dumpops
diff --git a/xlators/cluster/dht/src/switch.c b/xlators/cluster/dht/src/switch.c
index 9795ff4af4d..207d109a025 100644
--- a/xlators/cluster/dht/src/switch.c
+++ b/xlators/cluster/dht/src/switch.c
@@ -8,7 +8,6 @@
cases as published by the Free Software Foundation.
*/
-
#include "dht-common.h"
#include "dht-mem-types.h"
@@ -17,895 +16,876 @@
#include <fnmatch.h>
#include <string.h>
-extern struct volume_options options[];
+extern struct volume_options dht_options[];
struct switch_sched_array {
- xlator_t *xl;
- int32_t eligible;
- int32_t considered;
+ xlator_t *xl;
+ int32_t eligible;
+ int32_t considered;
};
/* Select one of this struct based on the path's pattern match */
struct switch_struct {
- struct switch_struct *next;
- struct switch_sched_array *array;
- int32_t node_index; /* Index of the node in
- this pattern. */
- int32_t num_child; /* Total num of child nodes
- with this pattern. */
- char path_pattern[256];
+ struct switch_struct *next;
+ struct switch_sched_array *array;
+ int32_t node_index; /* Index of the node in
+ this pattern. */
+ int32_t num_child; /* Total num of child nodes
+ with this pattern. */
+ char path_pattern[256];
};
/* TODO: all 'TODO's in dht.c holds good */
/* This function should return child node as '*:subvolumes' is inserterd */
static int32_t
-gf_switch_valid_child (xlator_t *this, const char *child)
+gf_switch_valid_child(xlator_t *this, const char *child)
{
- xlator_list_t *children = NULL;
- int32_t ret = 0;
-
- children = this->children;
- while (children) {
- if (!strcmp (child, children->xlator->name)) {
- ret = 1;
- break;
- }
- children = children->next;
+ xlator_list_t *children = NULL;
+ int32_t ret = 0;
+
+ children = this->children;
+ while (children) {
+ if (!strcmp(child, children->xlator->name)) {
+ ret = 1;
+ break;
}
+ children = children->next;
+ }
- return ret;
+ return ret;
}
static xlator_t *
-get_switch_matching_subvol (const char *path, dht_conf_t *conf,
- xlator_t *hashed_subvol)
+get_switch_matching_subvol(const char *path, dht_conf_t *conf,
+ xlator_t *hashed_subvol)
{
- struct switch_struct *cond = NULL;
- struct switch_struct *trav = NULL;
- char *pathname = NULL;
- int idx = 0;
- xlator_t *subvol = NULL;
-
- cond = conf->private;
- subvol = hashed_subvol;
- if (!cond)
- goto out;
-
- pathname = gf_strdup (path);
- if (!pathname)
- goto out;
-
- trav = cond;
- while (trav) {
- if (fnmatch (trav->path_pattern,
- pathname, FNM_NOESCAPE) == 0) {
- for (idx = 0; idx < trav->num_child; idx++) {
- if (trav->array[idx].xl == hashed_subvol)
- goto out;
- }
- idx = trav->node_index++;
- trav->node_index %= trav->num_child;
- subvol = trav->array[idx].xl;
- goto out;
- }
- trav = trav->next;
+ struct switch_struct *cond = NULL;
+ struct switch_struct *trav = NULL;
+ char *pathname = NULL;
+ int idx = 0;
+ xlator_t *subvol = NULL;
+
+ cond = conf->private;
+ subvol = hashed_subvol;
+ if (!cond)
+ goto out;
+
+ pathname = gf_strdup(path);
+ if (!pathname)
+ goto out;
+
+ trav = cond;
+ while (trav) {
+ if (fnmatch(trav->path_pattern, pathname, FNM_NOESCAPE) == 0) {
+ for (idx = 0; idx < trav->num_child; idx++) {
+ if (trav->array[idx].xl == hashed_subvol)
+ goto out;
+ }
+ idx = trav->node_index++;
+ trav->node_index %= trav->num_child;
+ subvol = trav->array[idx].xl;
+ goto out;
}
+ trav = trav->next;
+ }
out:
- GF_FREE (pathname);
+ GF_FREE(pathname);
- return subvol;
+ return subvol;
}
-
int
-switch_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, dict_t *xattr,
- struct iatt *postparent)
+switch_local_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *stbuf, dict_t *xattr,
+ struct iatt *postparent)
{
- xlator_t *subvol = NULL;
- char is_linkfile = 0;
- char is_dir = 0;
- dht_conf_t *conf = NULL;
- dht_local_t *local = NULL;
- loc_t *loc = NULL;
- int i = 0;
- xlator_t *prev = NULL;
- int call_cnt = 0;
- int ret = 0;
-
- conf = this->private;
-
- prev = cookie;
- local = frame->local;
- loc = &local->loc;
-
- if (ENTRY_MISSING (op_ret, op_errno)) {
- if (conf->search_unhashed) {
- local->op_errno = ENOENT;
- dht_lookup_everywhere (frame, this, loc);
- return 0;
- }
+ xlator_t *subvol = NULL;
+ char is_linkfile = 0;
+ char is_dir = 0;
+ dht_conf_t *conf = NULL;
+ dht_local_t *local = NULL;
+ loc_t *loc = NULL;
+ int i = 0;
+ xlator_t *prev = NULL;
+ int call_cnt = 0;
+ int ret = 0;
+
+ conf = this->private;
+
+ prev = cookie;
+ local = frame->local;
+ loc = &local->loc;
+
+ if (ENTRY_MISSING(op_ret, op_errno)) {
+ if (conf->search_unhashed) {
+ local->op_errno = ENOENT;
+ dht_lookup_everywhere(frame, this, loc);
+ return 0;
}
-
- if (op_ret == -1)
- goto out;
-
- is_linkfile = check_is_linkfile (inode, stbuf, xattr,
- conf->link_xattr_name);
- is_dir = check_is_dir (inode, stbuf, xattr);
-
- if (!is_dir && !is_linkfile) {
- /* non-directory and not a linkfile */
-
- ret = dht_layout_preset (this, prev, inode);
- if (ret < 0) {
- gf_msg_debug (this->name, 0,
- "could not set pre-set layout "
- "for subvol %s",
- prev->name);
- op_ret = -1;
- op_errno = EINVAL;
- goto err;
- }
-
- goto out;
+ }
+
+ if (op_ret == -1)
+ goto out;
+
+ is_linkfile = check_is_linkfile(inode, stbuf, xattr, conf->link_xattr_name);
+ is_dir = check_is_dir(inode, stbuf, xattr);
+
+ if (!is_dir && !is_linkfile) {
+ /* non-directory and not a linkfile */
+
+ ret = dht_layout_preset(this, prev, inode);
+ if (ret < 0) {
+ gf_msg_debug(this->name, 0,
+ "could not set pre-set layout "
+ "for subvol %s",
+ prev->name);
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto err;
}
- if (is_dir) {
- call_cnt = conf->subvolume_cnt;
- local->call_cnt = call_cnt;
+ goto out;
+ }
- local->inode = inode_ref (inode);
- local->xattr = dict_ref (xattr);
+ if (is_dir) {
+ call_cnt = conf->subvolume_cnt;
+ local->call_cnt = call_cnt;
- local->op_ret = 0;
- local->op_errno = 0;
+ local->inode = inode_ref(inode);
+ local->xattr = dict_ref(xattr);
- local->layout = dht_layout_new (this, conf->subvolume_cnt);
- if (!local->layout) {
- op_ret = -1;
- op_errno = ENOMEM;
- gf_msg_debug (this->name, 0,
- "memory allocation failed :(");
- goto err;
- }
+ local->op_ret = 0;
+ local->op_errno = 0;
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND_COOKIE (frame, dht_lookup_dir_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->lookup,
- &local->loc, local->xattr_req);
- }
+ local->layout = dht_layout_new(this, conf->subvolume_cnt);
+ if (!local->layout) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ gf_msg_debug(this->name, 0, "memory allocation failed :(");
+ goto err;
}
- if (is_linkfile) {
- subvol = dht_linkfile_subvol (this, inode, stbuf, xattr);
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND_COOKIE(frame, dht_lookup_dir_cbk, conf->subvolumes[i],
+ conf->subvolumes[i],
+ conf->subvolumes[i]->fops->lookup, &local->loc,
+ local->xattr_req);
+ }
+ }
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "linkfile has no link subvolume.path=%s",
- loc->path);
- dht_lookup_everywhere (frame, this, loc);
- return 0;
- }
+ if (is_linkfile) {
+ subvol = dht_linkfile_subvol(this, inode, stbuf, xattr);
- STACK_WIND_COOKIE (frame, dht_lookup_linkfile_cbk, subvol,
- subvol, subvol->fops->lookup,
- &local->loc, local->xattr_req);
+ if (!subvol) {
+ gf_msg_debug(this->name, 0,
+ "linkfile has no link subvolume.path=%s", loc->path);
+ dht_lookup_everywhere(frame, this, loc);
+ return 0;
}
- return 0;
+ STACK_WIND_COOKIE(frame, dht_lookup_linkfile_cbk, subvol, subvol,
+ subvol->fops->lookup, &local->loc, local->xattr_req);
+ }
+
+ return 0;
out:
- if (!local->hashed_subvol) {
- gf_msg_debug (this->name, 0,
- "no subvolume in layout for path=%s",
- local->loc.path);
- local->op_errno = ENOENT;
- dht_lookup_everywhere (frame, this, loc);
- return 0;
- }
+ if (!local->hashed_subvol) {
+ gf_msg_debug(this->name, 0, "no subvolume in layout for path=%s",
+ local->loc.path);
+ local->op_errno = ENOENT;
+ dht_lookup_everywhere(frame, this, loc);
+ return 0;
+ }
- STACK_WIND_COOKIE (frame, dht_lookup_cbk, local->hashed_subvol,
- local->hashed_subvol,
- local->hashed_subvol->fops->lookup,
- &local->loc, local->xattr_req);
+ STACK_WIND_COOKIE(frame, dht_lookup_cbk, local->hashed_subvol,
+ local->hashed_subvol, local->hashed_subvol->fops->lookup,
+ &local->loc, local->xattr_req);
- return 0;
+ return 0;
err:
- DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno,
- inode, stbuf, xattr, NULL);
- return 0;
+ DHT_STACK_UNWIND(lookup, frame, op_ret, op_errno, inode, stbuf, xattr,
+ NULL);
+ return 0;
}
int
-switch_lookup (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xattr_req)
+switch_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ dict_t *xattr_req)
{
- xlator_t *hashed_subvol = NULL;
- xlator_t *cached_subvol = NULL;
- xlator_t *subvol = NULL;
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- int ret = -1;
- int op_errno = -1;
- dht_layout_t *layout = NULL;
- int i = 0;
- int call_cnt = 0;
-
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
- VALIDATE_OR_GOTO (loc->path, err);
-
- conf = this->private;
-
- local = dht_local_init (frame, loc, NULL, GF_FOP_LOOKUP);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
+ xlator_t *hashed_subvol = NULL;
+ xlator_t *cached_subvol = NULL;
+ xlator_t *subvol = NULL;
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ int ret = -1;
+ int op_errno = -1;
+ dht_layout_t *layout = NULL;
+ int i = 0;
+ int call_cnt = 0;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+ VALIDATE_OR_GOTO(loc->inode, err);
+ VALIDATE_OR_GOTO(loc->path, err);
+
+ conf = this->private;
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_LOOKUP);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ if (xattr_req) {
+ local->xattr_req = dict_ref(xattr_req);
+ } else {
+ local->xattr_req = dict_new();
+ }
+
+ hashed_subvol = dht_subvol_get_hashed(this, &local->loc);
+ cached_subvol = local->cached_subvol;
+
+ local->hashed_subvol = hashed_subvol;
+
+ if (is_revalidate(loc)) {
+ layout = local->layout;
+ if (!layout) {
+ gf_msg_debug(this->name, 0,
+ "revalidate lookup without cache. path=%s", loc->path);
+ op_errno = EINVAL;
+ goto err;
}
- if (xattr_req) {
- local->xattr_req = dict_ref (xattr_req);
- } else {
- local->xattr_req = dict_new ();
+ if (layout->gen && (layout->gen < conf->gen)) {
+ gf_msg_debug(this->name, 0, "incomplete layout failure for path=%s",
+ loc->path);
+ dht_layout_unref(this, local->layout);
+ goto do_fresh_lookup;
}
- hashed_subvol = dht_subvol_get_hashed (this, &local->loc);
- cached_subvol = local->cached_subvol;
+ local->inode = inode_ref(loc->inode);
- local->hashed_subvol = hashed_subvol;
+ local->call_cnt = layout->cnt;
+ call_cnt = local->call_cnt;
- if (is_revalidate (loc)) {
- layout = local->layout;
- if (!layout) {
- gf_msg_debug(this->name, 0,
- "revalidate lookup without cache. path=%s",
- loc->path);
- op_errno = EINVAL;
- goto err;
- }
-
- if (layout->gen && (layout->gen < conf->gen)) {
- gf_msg_debug (this->name, 0,
- "incomplete layout failure for path=%s",
- loc->path);
- dht_layout_unref (this, local->layout);
- goto do_fresh_lookup;
- }
-
- local->inode = inode_ref (loc->inode);
-
- local->call_cnt = layout->cnt;
- call_cnt = local->call_cnt;
+ /* NOTE: we don't require 'trusted.glusterfs.dht.linkto'
+ * attribute, revalidates directly go to the cached-subvolume.
+ */
+ ret = dict_set_uint32(local->xattr_req, conf->xattr_name, 4 * 4);
+ if (ret < 0)
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_DICT_SET_FAILED,
+ "failed to set dict value for %s", conf->xattr_name);
- /* NOTE: we don't require 'trusted.glusterfs.dht.linkto'
- * attribute, revalidates directly go to the cached-subvolume.
- */
- ret = dict_set_uint32 (local->xattr_req,
- conf->xattr_name, 4 * 4);
- if (ret < 0)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "failed to set dict value for %s",
- conf->xattr_name);
+ for (i = 0; i < layout->cnt; i++) {
+ subvol = layout->list[i].xlator;
- for (i = 0; i < layout->cnt; i++) {
- subvol = layout->list[i].xlator;
+ STACK_WIND_COOKIE(frame, dht_revalidate_cbk, subvol, subvol,
+ subvol->fops->lookup, loc, local->xattr_req);
- STACK_WIND_COOKIE (frame, dht_revalidate_cbk, subvol,
- subvol, subvol->fops->lookup,
- loc, local->xattr_req);
+ if (!--call_cnt)
+ break;
+ }
+ } else {
+ do_fresh_lookup:
+ ret = dict_set_uint32(local->xattr_req, conf->xattr_name, 4 * 4);
+ if (ret < 0)
+ gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_DICT_SET_FAILED,
+ "failed to set dict value for %s", conf->xattr_name);
+
+ ret = dict_set_uint32(local->xattr_req, conf->link_xattr_name, 256);
+ if (ret < 0)
+ gf_msg(this->name, GF_LOG_WARNING, EINVAL, DHT_MSG_DICT_SET_FAILED,
+ "failed to set dict value for %s", conf->link_xattr_name);
+
+ if (!hashed_subvol) {
+ gf_msg_debug(this->name, 0,
+ "no subvolume in layout for path=%s, "
+ "checking on all the subvols to see if "
+ "it is a directory",
+ loc->path);
+ call_cnt = conf->subvolume_cnt;
+ local->call_cnt = call_cnt;
+
+ local->layout = dht_layout_new(this, conf->subvolume_cnt);
+ if (!local->layout) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND_COOKIE(frame, dht_lookup_dir_cbk,
+ conf->subvolumes[i], conf->subvolumes[i],
+ conf->subvolumes[i]->fops->lookup,
+ &local->loc, local->xattr_req);
+ }
+ return 0;
+ }
- if (!--call_cnt)
- break;
- }
+ /* */
+ cached_subvol = get_switch_matching_subvol(loc->path, conf,
+ hashed_subvol);
+ if (cached_subvol == hashed_subvol) {
+ STACK_WIND_COOKIE(frame, dht_lookup_cbk, hashed_subvol,
+ hashed_subvol, hashed_subvol->fops->lookup, loc,
+ local->xattr_req);
} else {
- do_fresh_lookup:
- ret = dict_set_uint32 (local->xattr_req,
- conf->xattr_name, 4 * 4);
- if (ret < 0)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "failed to set dict value for %s",
- conf->xattr_name);
-
- ret = dict_set_uint32 (local->xattr_req,
- conf->link_xattr_name, 256);
- if (ret < 0)
- gf_msg (this->name, GF_LOG_WARNING, EINVAL,
- DHT_MSG_DICT_SET_FAILED,
- "failed to set dict value for %s",
- conf->link_xattr_name);
-
- if (!hashed_subvol) {
- gf_msg_debug (this->name, 0,
- "no subvolume in layout for path=%s, "
- "checking on all the subvols to see if "
- "it is a directory", loc->path);
- call_cnt = conf->subvolume_cnt;
- local->call_cnt = call_cnt;
-
- local->layout = dht_layout_new (this,
- conf->subvolume_cnt);
- if (!local->layout) {
- op_errno = ENOMEM;
- goto err;
- }
-
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND_COOKIE (frame, dht_lookup_dir_cbk,
- conf->subvolumes[i],
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->lookup,
- &local->loc, local->xattr_req);
- }
- return 0;
- }
-
- /* */
- cached_subvol = get_switch_matching_subvol (loc->path, conf,
- hashed_subvol);
- if (cached_subvol == hashed_subvol) {
- STACK_WIND_COOKIE (frame, dht_lookup_cbk, hashed_subvol,
- hashed_subvol,
- hashed_subvol->fops->lookup,
- loc, local->xattr_req);
- } else {
- STACK_WIND_COOKIE (frame, switch_local_lookup_cbk,
- cached_subvol, cached_subvol,
- cached_subvol->fops->lookup,
- loc, local->xattr_req);
- }
+ STACK_WIND_COOKIE(frame, switch_local_lookup_cbk, cached_subvol,
+ cached_subvol, cached_subvol->fops->lookup, loc,
+ local->xattr_req);
}
+ }
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (lookup, frame, -1, op_errno,
- NULL, NULL, NULL, NULL);
- return 0;
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+ return 0;
}
int
-switch_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+switch_create_linkfile_create_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int op_ret, int op_errno,
+ inode_t *inode, struct iatt *stbuf,
+ struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- dht_local_t *local = NULL;
+ dht_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret == -1)
- goto err;
+ if (op_ret == -1)
+ goto err;
- STACK_WIND_COOKIE (frame, dht_create_cbk, local->cached_subvol,
- local->cached_subvol, local->cached_subvol->fops->create,
- &local->loc, local->flags, local->mode, local->umask,
- local->fd, local->params);
+ STACK_WIND_COOKIE(frame, dht_create_cbk, local->cached_subvol,
+ local->cached_subvol, local->cached_subvol->fops->create,
+ &local->loc, local->flags, local->mode, local->umask,
+ local->fd, local->params);
- return 0;
+ return 0;
err:
- DHT_STACK_UNWIND (create, frame, -1, op_errno,
- NULL, NULL, NULL, NULL, NULL, NULL);
- return 0;
+ DHT_STACK_UNWIND(create, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
int
-switch_create (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, mode_t mode,
- mode_t umask, fd_t *fd, dict_t *params)
+switch_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *params)
{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- xlator_t *subvol = NULL;
- xlator_t *avail_subvol = NULL;
- int op_errno = -1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
-
- conf = this->private;
-
- dht_get_du_info (frame, this, loc);
-
- local = dht_local_init (frame, loc, fd, GF_FOP_CREATE);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- subvol = dht_subvol_get_hashed (this, loc);
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no subvolume in layout for path=%s",
- loc->path);
- op_errno = ENOENT;
- goto err;
- }
-
- avail_subvol = get_switch_matching_subvol (loc->path, conf, subvol);
- if (dht_is_subvol_filled (this, avail_subvol)) {
- avail_subvol =
- dht_free_disk_available_subvol (this, avail_subvol,
- local);
- }
-
- if (subvol != avail_subvol) {
- /* create a link file instead of actual file */
- local->mode = mode;
- local->flags = flags;
- local->umask = umask;
- local->cached_subvol = avail_subvol;
- dht_linkfile_create (frame, switch_create_linkfile_create_cbk,
- this, avail_subvol, subvol, loc);
- return 0;
- }
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ xlator_t *subvol = NULL;
+ xlator_t *avail_subvol = NULL;
+ int op_errno = -1;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+
+ conf = this->private;
+
+ dht_get_du_info(frame, this, loc);
+
+ local = dht_local_init(frame, loc, fd, GF_FOP_CREATE);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ subvol = dht_subvol_get_hashed(this, loc);
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no subvolume in layout for path=%s",
+ loc->path);
+ op_errno = ENOENT;
+ goto err;
+ }
+
+ avail_subvol = get_switch_matching_subvol(loc->path, conf, subvol);
+ if (dht_is_subvol_filled(this, avail_subvol)) {
+ avail_subvol = dht_free_disk_available_subvol(this, avail_subvol,
+ local);
+ }
+
+ if (subvol != avail_subvol) {
+ /* create a link file instead of actual file */
+ local->mode = mode;
+ local->flags = flags;
+ local->umask = umask;
+ local->cached_subvol = avail_subvol;
+ dht_linkfile_create(frame, switch_create_linkfile_create_cbk, this,
+ avail_subvol, subvol, loc);
+ return 0;
+ }
- gf_msg_trace (this->name, 0,
- "creating %s on %s", loc->path, subvol->name);
+ gf_msg_trace(this->name, 0, "creating %s on %s", loc->path, subvol->name);
- STACK_WIND_COOKIE (frame, dht_create_cbk, subvol, subvol,
- subvol->fops->create, loc, flags, mode, umask, fd,
- params);
+ STACK_WIND_COOKIE(frame, dht_create_cbk, subvol, subvol,
+ subvol->fops->create, loc, flags, mode, umask, fd,
+ params);
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (create, frame, -1, op_errno,
- NULL, NULL, NULL, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(create, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL,
+ NULL);
- return 0;
+ return 0;
}
int
-switch_mknod_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode,
- struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+switch_mknod_linkfile_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- dht_local_t *local = NULL;
+ dht_local_t *local = NULL;
- local = frame->local;
- if (!local || !local->cached_subvol) {
- op_errno = EINVAL;
- op_ret = -1;
- goto err;
- }
+ local = frame->local;
+ if (!local || !local->cached_subvol) {
+ op_errno = EINVAL;
+ op_ret = -1;
+ goto err;
+ }
- if (op_ret >= 0) {
- STACK_WIND_COOKIE (frame, dht_newfile_cbk,
- (void *)local->cached_subvol, local->cached_subvol,
- local->cached_subvol->fops->mknod,
- &local->loc, local->mode, local->rdev,
- local->umask, local->params);
+ if (op_ret >= 0) {
+ STACK_WIND_COOKIE(
+ frame, dht_newfile_cbk, (void *)local->cached_subvol,
+ local->cached_subvol, local->cached_subvol->fops->mknod,
+ &local->loc, local->mode, local->rdev, local->umask, local->params);
- return 0;
- }
-err:
- DHT_STACK_UNWIND (link, frame, op_ret, op_errno,
- inode, stbuf, preparent, postparent, xdata);
return 0;
+ }
+err:
+ DHT_STACK_UNWIND(link, frame, op_ret, op_errno, inode, stbuf, preparent,
+ postparent, xdata);
+ return 0;
}
-
int
-switch_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- dev_t rdev, mode_t umask, dict_t *params)
+switch_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *params)
{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- xlator_t *subvol = NULL;
- xlator_t *avail_subvol = NULL;
- int op_errno = -1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
-
- conf = this->private;
-
- dht_get_du_info (frame, this, loc);
-
- local = dht_local_init (frame, loc, NULL, GF_FOP_MKNOD);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- subvol = dht_subvol_get_hashed (this, loc);
- if (!subvol) {
- gf_msg_debug (this->name, 0,
- "no subvolume in layout for path=%s",
- loc->path);
- op_errno = ENOENT;
- goto err;
- }
-
- /* Consider the disksize in consideration */
- avail_subvol = get_switch_matching_subvol (loc->path, conf, subvol);
- if (dht_is_subvol_filled (this, avail_subvol)) {
- avail_subvol =
- dht_free_disk_available_subvol (this, avail_subvol,
- local);
- }
-
- if (avail_subvol != subvol) {
- /* Create linkfile first */
-
- local->params = dict_ref (params);
- local->mode = mode;
- local->umask = umask;
- local->rdev = rdev;
- local->cached_subvol = avail_subvol;
-
- dht_linkfile_create (frame, switch_mknod_linkfile_cbk,
- this, avail_subvol, subvol, loc);
- return 0;
- }
+ dht_local_t *local = NULL;
+ dht_conf_t *conf = NULL;
+ xlator_t *subvol = NULL;
+ xlator_t *avail_subvol = NULL;
+ int op_errno = -1;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+
+ conf = this->private;
+
+ dht_get_du_info(frame, this, loc);
+
+ local = dht_local_init(frame, loc, NULL, GF_FOP_MKNOD);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ subvol = dht_subvol_get_hashed(this, loc);
+ if (!subvol) {
+ gf_msg_debug(this->name, 0, "no subvolume in layout for path=%s",
+ loc->path);
+ op_errno = ENOENT;
+ goto err;
+ }
+
+ /* Consider the disksize in consideration */
+ avail_subvol = get_switch_matching_subvol(loc->path, conf, subvol);
+ if (dht_is_subvol_filled(this, avail_subvol)) {
+ avail_subvol = dht_free_disk_available_subvol(this, avail_subvol,
+ local);
+ }
+
+ if (avail_subvol != subvol) {
+ /* Create linkfile first */
+
+ local->params = dict_ref(params);
+ local->mode = mode;
+ local->umask = umask;
+ local->rdev = rdev;
+ local->cached_subvol = avail_subvol;
+
+ dht_linkfile_create(frame, switch_mknod_linkfile_cbk, this,
+ avail_subvol, subvol, loc);
+ return 0;
+ }
- gf_msg_trace (this->name, 0,
- "creating %s on %s", loc->path, subvol->name);
+ gf_msg_trace(this->name, 0, "creating %s on %s", loc->path, subvol->name);
- STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol, subvol,
- subvol->fops->mknod, loc, mode, rdev, umask,
- params);
+ STACK_WIND_COOKIE(frame, dht_newfile_cbk, (void *)subvol, subvol,
+ subvol->fops->mknod, loc, mode, rdev, umask, params);
- return 0;
+ return 0;
err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
- NULL, NULL, NULL, NULL, NULL);
+ op_errno = (op_errno == -1) ? errno : op_errno;
+ DHT_STACK_UNWIND(mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
void
-switch_fini (xlator_t *this)
+switch_fini(xlator_t *this)
{
- dht_conf_t *conf = NULL;
- struct switch_struct *trav = NULL;
- struct switch_struct *prev = NULL;
-
- conf = this->private;
-
- if (conf) {
- trav = (struct switch_struct *)conf->private;
- conf->private = NULL;
- while (trav) {
- GF_FREE (trav->array);
- prev = trav;
- trav = trav->next;
- GF_FREE (prev);
- }
+ dht_conf_t *conf = NULL;
+ struct switch_struct *trav = NULL;
+ struct switch_struct *prev = NULL;
+
+ conf = this->private;
+
+ if (conf) {
+ trav = (struct switch_struct *)conf->private;
+ conf->private = NULL;
+ while (trav) {
+ GF_FREE(trav->array);
+ prev = trav;
+ trav = trav->next;
+ GF_FREE(prev);
}
+ }
- dht_fini(this);
+ dht_fini(this);
}
int
-set_switch_pattern (xlator_t *this, dht_conf_t *conf,
- const char *pattern_str)
+set_switch_pattern(xlator_t *this, dht_conf_t *conf, const char *pattern_str)
{
- int flag = 0;
- int idx = 0;
- int index = 0;
- int child_count = 0;
- char *tmp = NULL;
- char *tmp1 = NULL;
- char *child = NULL;
- char *tmp_str = NULL;
- char *tmp_str1 = NULL;
- char *dup_str = NULL;
- char *dup_childs = NULL;
- char *switch_str = NULL;
- char *pattern = NULL;
- char *childs = NULL;
- char *option_string = NULL;
- struct switch_struct *switch_buf = NULL;
- struct switch_struct *switch_opt = NULL;
- struct switch_struct *trav = NULL;
- struct switch_sched_array *switch_buf_array = NULL;
- xlator_list_t *trav_xl = NULL;
-
- trav_xl = this->children;
- while (trav_xl) {
- index++;
- trav_xl = trav_xl->next;
+ int flag = 0;
+ int idx = 0;
+ int index = 0;
+ int child_count = 0;
+ char *tmp = NULL;
+ char *tmp1 = NULL;
+ char *child = NULL;
+ char *tmp_str = NULL;
+ char *tmp_str1 = NULL;
+ char *dup_str = NULL;
+ char *dup_childs = NULL;
+ char *switch_str = NULL;
+ char *pattern = NULL;
+ char *childs = NULL;
+ char *option_string = NULL;
+ size_t pattern_length;
+ struct switch_struct *switch_buf = NULL;
+ struct switch_struct *switch_opt = NULL;
+ struct switch_struct *trav = NULL;
+ struct switch_sched_array *switch_buf_array = NULL;
+ xlator_list_t *trav_xl = NULL;
+
+ trav_xl = this->children;
+ while (trav_xl) {
+ index++;
+ trav_xl = trav_xl->next;
+ }
+ child_count = index;
+ switch_buf_array = GF_CALLOC((index + 1), sizeof(struct switch_sched_array),
+ gf_switch_mt_switch_sched_array);
+ if (!switch_buf_array)
+ goto err;
+
+ trav_xl = this->children;
+ index = 0;
+
+ while (trav_xl) {
+ switch_buf_array[index].xl = trav_xl->xlator;
+ switch_buf_array[index].eligible = 1;
+ trav_xl = trav_xl->next;
+ index++;
+ }
+
+ /* *jpg:child1,child2;*mpg:child3;*:child4,child5,child6 */
+
+ /* Get the pattern for considering switch case.
+ "option block-size *avi:10MB" etc */
+ option_string = gf_strdup(pattern_str);
+ if (option_string == NULL) {
+ goto err;
+ }
+ switch_str = strtok_r(option_string, ";", &tmp_str);
+ while (switch_str) {
+ dup_str = gf_strdup(switch_str);
+ if (dup_str == NULL) {
+ goto err;
}
- child_count = index;
- switch_buf_array = GF_CALLOC ((index + 1),
- sizeof (struct switch_sched_array),
- gf_switch_mt_switch_sched_array);
- if (!switch_buf_array)
- goto err;
-
- trav_xl = this->children;
- index = 0;
-
- while (trav_xl) {
- switch_buf_array[index].xl = trav_xl->xlator;
- switch_buf_array[index].eligible = 1;
- trav_xl = trav_xl->next;
- index++;
+ switch_opt = GF_CALLOC(1, sizeof(struct switch_struct),
+ gf_switch_mt_switch_struct);
+ if (!switch_opt) {
+ GF_FREE(dup_str);
+ goto err;
}
- /* *jpg:child1,child2;*mpg:child3;*:child4,child5,child6 */
-
- /* Get the pattern for considering switch case.
- "option block-size *avi:10MB" etc */
- option_string = gf_strdup (pattern_str);
- switch_str = strtok_r (option_string, ";", &tmp_str);
- while (switch_str) {
- dup_str = gf_strdup (switch_str);
- switch_opt = GF_CALLOC (1, sizeof (struct switch_struct),
- gf_switch_mt_switch_struct);
- if (!switch_opt) {
- GF_FREE (dup_str);
- goto err;
- }
+ pattern = strtok_r(dup_str, ":", &tmp_str1);
+ childs = strtok_r(NULL, ":", &tmp_str1);
+ if (strncmp(pattern, "*", 2) == 0) {
+ gf_msg("switch", GF_LOG_INFO, 0, DHT_MSG_SWITCH_PATTERN_INFO,
+ "'*' pattern will be taken by default "
+ "for all the unconfigured child nodes,"
+ " hence neglecting current option");
+ switch_str = strtok_r(NULL, ";", &tmp_str);
+ GF_FREE(switch_opt);
+ switch_opt = NULL;
+ GF_FREE(dup_str);
+ continue;
+ }
+ GF_FREE(dup_str);
+
+ pattern_length = strlen(pattern);
+ if (pattern_length >= (sizeof(switch_opt->path_pattern))) {
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ DHT_MSG_SET_SWITCH_PATTERN_ERROR, "Pattern (%s) too long",
+ pattern);
+ goto err;
+ }
+ memcpy(switch_opt->path_pattern, pattern, pattern_length);
+ switch_opt->path_pattern[pattern_length] = '\0';
- pattern = strtok_r (dup_str, ":", &tmp_str1);
- childs = strtok_r (NULL, ":", &tmp_str1);
- if (strncmp (pattern, "*", 2) == 0) {
- gf_msg ("switch", GF_LOG_INFO, 0,
- DHT_MSG_SWITCH_PATTERN_INFO,
- "'*' pattern will be taken by default "
- "for all the unconfigured child nodes,"
- " hence neglecting current option");
- switch_str = strtok_r (NULL, ";", &tmp_str);
- GF_FREE (switch_opt);
- GF_FREE (dup_str);
- continue;
- }
- GF_FREE (dup_str);
- memcpy (switch_opt->path_pattern, pattern, strlen (pattern));
- if (childs) {
- dup_childs = gf_strdup (childs);
- child = strtok_r (dup_childs, ",", &tmp);
- while (child) {
- if (gf_switch_valid_child (this, child)) {
- idx++;
- child = strtok_r (NULL, ",", &tmp);
- } else {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_SUBVOL_ERROR,
- "%s is not a subvolume of %s. "
- "pattern can only be scheduled "
- "only to a subvolume of %s",
- child, this->name, this->name);
- goto err;
- }
- }
- GF_FREE (dup_childs);
- child = strtok_r (childs, ",", &tmp1);
- switch_opt->num_child = idx;
- switch_opt->array = GF_CALLOC (1, (idx *
- sizeof (struct switch_sched_array)),
- gf_switch_mt_switch_sched_array);
- if (!switch_opt->array)
- goto err;
- idx = 0;
- while (child) {
- for (index = 0; index < child_count; index++) {
- if (strcmp (switch_buf_array[index].xl->name,
- child) == 0) {
- gf_msg_debug ("switch", 0,
- "'%s' pattern will be "
- "scheduled to \"%s\"",
- switch_opt->path_pattern, child);
- /*
- if (switch_buf_array[index-1].considered) {
- gf_msg_debug ("switch", 0,
- "ambiguity found, exiting");
- return -1;
- }
- */
- switch_opt->array[idx].xl = switch_buf_array[index].xl;
- switch_buf_array[index].considered = 1;
- idx++;
- break;
- }
- }
- child = strtok_r (NULL, ",", &tmp1);
- }
+ if (childs) {
+ dup_childs = gf_strdup(childs);
+ if (dup_childs == NULL) {
+ goto err;
+ }
+ child = strtok_r(dup_childs, ",", &tmp);
+ while (child) {
+ if (gf_switch_valid_child(this, child)) {
+ idx++;
+ child = strtok_r(NULL, ",", &tmp);
} else {
- /* error */
- gf_msg ("switch", GF_LOG_ERROR, 0,
- DHT_MSG_SET_SWITCH_PATTERN_ERROR,
- "Check \"scheduler.switch.case\" "
- "option in unify volume. Exiting");
- goto err;
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_SUBVOL_ERROR,
+ "%s is not a subvolume of %s. "
+ "pattern can only be scheduled "
+ "only to a subvolume of %s",
+ child, this->name, this->name);
+ GF_FREE(dup_childs);
+ goto err;
}
-
- /* Link it to the main structure */
- if (switch_buf) {
- /* there are already few entries */
- trav = switch_buf;
- while (trav->next)
- trav = trav->next;
- trav->next = switch_opt;
- } else {
- /* First entry */
- switch_buf = switch_opt;
+ }
+ GF_FREE(dup_childs);
+ child = strtok_r(childs, ",", &tmp1);
+ switch_opt->num_child = idx;
+ switch_opt->array = GF_CALLOC(
+ 1, (idx * sizeof(struct switch_sched_array)),
+ gf_switch_mt_switch_sched_array);
+ if (!switch_opt->array)
+ goto err;
+ idx = 0;
+ while (child) {
+ for (index = 0; index < child_count; index++) {
+ if (strcmp(switch_buf_array[index].xl->name, child) == 0) {
+ gf_msg_debug("switch", 0,
+ "'%s' pattern will be "
+ "scheduled to \"%s\"",
+ switch_opt->path_pattern, child);
+ /*
+ if (switch_buf_array[index-1].considered) {
+ gf_msg_debug ("switch", 0,
+ "ambiguity found, exiting");
+ return -1;
+ }
+ */
+ switch_opt->array[idx].xl = switch_buf_array[index].xl;
+ switch_buf_array[index].considered = 1;
+ idx++;
+ break;
+ }
}
- switch_opt = NULL;
- switch_str = strtok_r (NULL, ";", &tmp_str);
+ child = strtok_r(NULL, ",", &tmp1);
+ }
+ } else {
+ /* error */
+ gf_msg("switch", GF_LOG_ERROR, 0, DHT_MSG_SET_SWITCH_PATTERN_ERROR,
+ "Check \"scheduler.switch.case\" "
+ "option in unify volume. Exiting");
+ goto err;
}
- /* Now, all the pattern based considerations done, so for all the
- * remaining pattern, '*' to all the remaining child nodes
- */
- {
- for (index=0; index < child_count; index++) {
- /* check for considered flag */
- if (switch_buf_array[index].considered)
- continue;
- flag++;
- }
- if (!flag) {
- gf_msg ("switch", GF_LOG_ERROR, 0,
- DHT_MSG_SET_SWITCH_PATTERN_ERROR,
- "No nodes left for pattern '*'. Exiting");
- goto err;
- }
- switch_opt = GF_CALLOC (1, sizeof (struct switch_struct),
- gf_switch_mt_switch_struct);
- if (!switch_opt)
- goto err;
-
- /* Add the '*' pattern to the array */
- memcpy (switch_opt->path_pattern, "*", 2);
- switch_opt->num_child = flag;
- switch_opt->array =
- GF_CALLOC (1,
- flag * sizeof (struct switch_sched_array),
- gf_switch_mt_switch_sched_array);
- if (!switch_opt->array)
- goto err;
- flag = 0;
- for (index=0; index < child_count; index++) {
- /* check for considered flag */
- if (switch_buf_array[index].considered)
- continue;
- gf_msg_debug ("switch", 0, "'%s'"
- " pattern will be scheduled to \"%s\"",
- switch_opt->path_pattern,
- switch_buf_array[index].xl->name);
-
- switch_opt->array[flag].xl =
- switch_buf_array[index].xl;
- switch_buf_array[index].considered = 1;
- flag++;
- }
- if (switch_buf) {
- /* there are already few entries */
- trav = switch_buf;
- while (trav->next)
- trav = trav->next;
- trav->next = switch_opt;
- } else {
- /* First entry */
- switch_buf = switch_opt;
- }
- switch_opt = NULL;
+ /* Link it to the main structure */
+ if (switch_buf) {
+ /* there are already few entries */
+ trav = switch_buf;
+ while (trav->next)
+ trav = trav->next;
+ trav->next = switch_opt;
+ } else {
+ /* First entry */
+ switch_buf = switch_opt;
+ }
+ switch_opt = NULL;
+ switch_str = strtok_r(NULL, ";", &tmp_str);
+ }
+
+ /* Now, all the pattern based considerations done, so for all the
+ * remaining pattern, '*' to all the remaining child nodes
+ */
+ {
+ for (index = 0; index < child_count; index++) {
+ /* check for considered flag */
+ if (switch_buf_array[index].considered)
+ continue;
+ flag++;
+ }
+ if (!flag) {
+ gf_msg("switch", GF_LOG_ERROR, 0, DHT_MSG_SET_SWITCH_PATTERN_ERROR,
+ "No nodes left for pattern '*'. Exiting");
+ goto err;
+ }
+ switch_opt = GF_CALLOC(1, sizeof(struct switch_struct),
+ gf_switch_mt_switch_struct);
+ if (!switch_opt)
+ goto err;
+
+ /* Add the '*' pattern to the array */
+ memcpy(switch_opt->path_pattern, "*", 2);
+ switch_opt->num_child = flag;
+ switch_opt->array = GF_CALLOC(1,
+ flag * sizeof(struct switch_sched_array),
+ gf_switch_mt_switch_sched_array);
+ if (!switch_opt->array)
+ goto err;
+ flag = 0;
+ for (index = 0; index < child_count; index++) {
+ /* check for considered flag */
+ if (switch_buf_array[index].considered)
+ continue;
+ gf_msg_debug("switch", 0,
+ "'%s'"
+ " pattern will be scheduled to \"%s\"",
+ switch_opt->path_pattern,
+ switch_buf_array[index].xl->name);
+
+ switch_opt->array[flag].xl = switch_buf_array[index].xl;
+ switch_buf_array[index].considered = 1;
+ flag++;
+ }
+ if (switch_buf) {
+ /* there are already few entries */
+ trav = switch_buf;
+ while (trav->next)
+ trav = trav->next;
+ trav->next = switch_opt;
+ } else {
+ /* First entry */
+ switch_buf = switch_opt;
}
- /* */
- conf->private = switch_buf;
+ switch_opt = NULL;
+ }
+ /* */
+ conf->private = switch_buf;
- return 0;
+ GF_FREE(option_string);
+ return 0;
err:
- GF_FREE (switch_buf_array);
- GF_FREE (switch_opt);
+ GF_FREE(switch_buf_array);
+ GF_FREE(switch_opt);
+ GF_FREE(option_string);
- if (switch_buf) {
- trav = switch_buf;
- while (trav) {
- GF_FREE (trav->array);
- switch_opt = trav;
- trav = trav->next;
- GF_FREE (switch_opt);
- }
+ if (switch_buf) {
+ trav = switch_buf;
+ while (trav) {
+ GF_FREE(trav->array);
+ switch_opt = trav;
+ trav = trav->next;
+ GF_FREE(switch_opt);
}
- return -1;
+ }
+ return -1;
}
-
int32_t
-switch_init (xlator_t *this)
+switch_init(xlator_t *this)
{
- dht_conf_t *conf = NULL;
- data_t *data = NULL;
- int ret = -1;
+ dht_conf_t *conf = NULL;
+ data_t *data = NULL;
+ int ret = -1;
+
+ ret = dht_init(this);
+ if (ret) {
+ return ret;
+ }
+ conf = this->private;
- ret = dht_init(this);
+ data = dict_get(this->options, "pattern.switch.case");
+ if (data) {
+ /* TODO: */
+ ret = set_switch_pattern(this, conf, data->data);
if (ret) {
- return ret;
- }
- conf = this->private;
-
- data = dict_get (this->options, "pattern.switch.case");
- if (data) {
- /* TODO: */
- ret = set_switch_pattern (this, conf, data->data);
- if (ret) {
- goto err;
- }
+ goto err;
}
+ }
- this->private = conf;
- return 0;
+ this->private = conf;
+ return 0;
err:
- dht_fini(this);
- return -1;
+ dht_fini(this);
+ return -1;
}
-
-class_methods_t class_methods = {
- .init = switch_init,
- .fini = switch_fini,
- .reconfigure = dht_reconfigure,
- .notify = dht_notify
-};
-
-
struct xlator_fops fops = {
- .lookup = switch_lookup,
- .create = switch_create,
- .mknod = switch_mknod,
-
- .stat = dht_stat,
- .fstat = dht_fstat,
- .truncate = dht_truncate,
- .ftruncate = dht_ftruncate,
- .access = dht_access,
- .readlink = dht_readlink,
- .setxattr = dht_setxattr,
- .getxattr = dht_getxattr,
- .removexattr = dht_removexattr,
- .open = dht_open,
- .readv = dht_readv,
- .writev = dht_writev,
- .flush = dht_flush,
- .fsync = dht_fsync,
- .statfs = dht_statfs,
- .lk = dht_lk,
- .opendir = dht_opendir,
- .readdir = dht_readdir,
- .readdirp = dht_readdirp,
- .fsyncdir = dht_fsyncdir,
- .symlink = dht_symlink,
- .unlink = dht_unlink,
- .link = dht_link,
- .mkdir = dht_mkdir,
- .rmdir = dht_rmdir,
- .rename = dht_rename,
- .inodelk = dht_inodelk,
- .finodelk = dht_finodelk,
- .entrylk = dht_entrylk,
- .fentrylk = dht_fentrylk,
- .xattrop = dht_xattrop,
- .fxattrop = dht_fxattrop,
- .setattr = dht_setattr,
+ .lookup = switch_lookup,
+ .create = switch_create,
+ .mknod = switch_mknod,
+
+ .stat = dht_stat,
+ .fstat = dht_fstat,
+ .truncate = dht_truncate,
+ .ftruncate = dht_ftruncate,
+ .access = dht_access,
+ .readlink = dht_readlink,
+ .setxattr = dht_setxattr,
+ .getxattr = dht_getxattr,
+ .removexattr = dht_removexattr,
+ .open = dht_open,
+ .readv = dht_readv,
+ .writev = dht_writev,
+ .flush = dht_flush,
+ .fsync = dht_fsync,
+ .statfs = dht_statfs,
+ .lk = dht_lk,
+ .opendir = dht_opendir,
+ .readdir = dht_readdir,
+ .readdirp = dht_readdirp,
+ .fsyncdir = dht_fsyncdir,
+ .symlink = dht_symlink,
+ .unlink = dht_unlink,
+ .link = dht_link,
+ .mkdir = dht_mkdir,
+ .rmdir = dht_rmdir,
+ .rename = dht_rename,
+ .inodelk = dht_inodelk,
+ .finodelk = dht_finodelk,
+ .entrylk = dht_entrylk,
+ .fentrylk = dht_fentrylk,
+ .xattrop = dht_xattrop,
+ .fxattrop = dht_fxattrop,
+ .setattr = dht_setattr,
};
-
-struct xlator_cbks cbks = {
- .forget = dht_forget
+struct xlator_cbks cbks = {.forget = dht_forget};
+extern int32_t
+mem_acct_init(xlator_t *this);
+
+xlator_api_t xlator_api = {
+ .init = switch_init,
+ .fini = switch_fini,
+ .notify = dht_notify,
+ .reconfigure = dht_reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = dht_options,
+ .identifier = "switch",
+ .category = GF_TECH_PREVIEW,
};
diff --git a/xlators/cluster/dht/src/switch.sym b/xlators/cluster/dht/src/switch.sym
deleted file mode 100644
index 780b5fc0387..00000000000
--- a/xlators/cluster/dht/src/switch.sym
+++ /dev/null
@@ -1,8 +0,0 @@
-fops
-cbks
-class_methods
-dht_methods
-options
-mem_acct_init
-reconfigure
-dumpops
diff --git a/xlators/cluster/dht/src/tier-common.c b/xlators/cluster/dht/src/tier-common.c
deleted file mode 100644
index c6b54d21d8e..00000000000
--- a/xlators/cluster/dht/src/tier-common.c
+++ /dev/null
@@ -1,1249 +0,0 @@
-/*
- Copyright (c) 2015 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#include "glusterfs.h"
-#include "xlator.h"
-#include "libxlator.h"
-#include "dht-common.h"
-#include "defaults.h"
-#include "tier-common.h"
-#include "tier.h"
-
-int
-dht_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata);
-
-
-int
-tier_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- dht_local_t *local = NULL;
- loc_t *oldloc = NULL;
- loc_t *newloc = NULL;
-
- local = frame->local;
-
- oldloc = &local->loc;
- newloc = &local->loc2;
-
- if (op_ret == -1) {
- /* No continuation on DHT inode missing errors, as we should
- * then have a good stbuf that states P2 happened. We would
- * get inode missing if, the file completed migrated between
- * the lookup and the link call */
- goto out;
- }
-
- if (local->call_cnt != 1) {
- goto out;
- }
-
- local->call_cnt = 2;
-
- /* Do this on the hot tier now */
-
- STACK_WIND (frame, tier_link_cbk, local->cached_subvol,
- local->cached_subvol->fops->link,
- oldloc, newloc, xdata);
-
- return 0;
-
-out:
- DHT_STRIP_PHASE1_FLAGS (stbuf);
-
- DHT_STACK_UNWIND (link, frame, op_ret, op_errno, inode, stbuf,
- preparent, postparent, NULL);
-
- return 0;
-}
-
-
-int
-tier_link (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
-{
- xlator_t *cached_subvol = NULL;
- xlator_t *hashed_subvol = NULL;
- int op_errno = -1;
- int ret = -1;
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
-
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (oldloc, err);
- VALIDATE_OR_GOTO (newloc, err);
-
- conf = this->private;
-
- local = dht_local_init (frame, oldloc, NULL, GF_FOP_LINK);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->call_cnt = 1;
-
- cached_subvol = local->cached_subvol;
-
- if (!cached_subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for path=%s", oldloc->path);
- op_errno = ENOENT;
- goto err;
- }
-
- hashed_subvol = TIER_HASHED_SUBVOL;
-
- ret = loc_copy (&local->loc2, newloc);
- if (ret == -1) {
- op_errno = ENOMEM;
- goto err;
- }
-
- if (hashed_subvol == cached_subvol) {
- STACK_WIND (frame, dht_link_cbk,
- cached_subvol, cached_subvol->fops->link,
- oldloc, newloc, xdata);
- return 0;
- }
-
-
- /* Create hardlinks to both the data file on the hot tier
- and the linkto file on the cold tier */
-
- gf_uuid_copy (local->gfid, oldloc->inode->gfid);
-
- STACK_WIND (frame, tier_link_cbk,
- hashed_subvol, hashed_subvol->fops->link,
- oldloc, newloc, xdata);
-
- return 0;
-err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL,
- NULL);
- return 0;
-}
-
-
-
-int
-tier_create_unlink_stale_linkto_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
-
- dht_local_t *local = NULL;
-
- local = frame->local;
-
- if (local->params) {
- dict_del (local->params, GLUSTERFS_INTERNAL_FOP_KEY);
- }
-
- DHT_STACK_UNWIND (create, frame, -1, local->op_errno,
- NULL, NULL, NULL, NULL, NULL, NULL);
-
- return 0;
-}
-
-int
-tier_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- fd_t *fd, inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
-{
- xlator_t *prev = NULL;
- int ret = -1;
- dht_local_t *local = NULL;
- xlator_t *hashed_subvol = NULL;
- dht_conf_t *conf = NULL;
-
- local = frame->local;
- conf = this->private;
-
- hashed_subvol = TIER_HASHED_SUBVOL;
-
- if (!local) {
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
-
- if (op_ret == -1) {
- if (local->linked == _gf_true && local->xattr_req) {
- local->op_errno = op_errno;
- local->op_ret = op_ret;
- ret = dht_fill_dict_to_avoid_unlink_of_migrating_file
- (local->xattr_req);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value to "
- "unlink of migrating file");
- goto out;
- }
-
- STACK_WIND (frame,
- tier_create_unlink_stale_linkto_cbk,
- hashed_subvol,
- hashed_subvol->fops->unlink,
- &local->loc, 0, local->xattr_req);
- return 0;
- }
- goto out;
- }
-
- prev = cookie;
-
- if (local->loc.parent) {
- dht_inode_ctx_time_update (local->loc.parent, this,
- preparent, 0);
-
- dht_inode_ctx_time_update (local->loc.parent, this,
- postparent, 1);
- }
-
- ret = dht_layout_preset (this, prev, inode);
- if (ret != 0) {
- gf_msg_debug (this->name, 0,
- "could not set preset layout for subvol %s",
- prev->name);
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
-
- local->op_errno = op_errno;
-
- if (local->linked == _gf_true) {
- local->stbuf = *stbuf;
- dht_linkfile_attr_heal (frame, this);
- }
-out:
- if (local) {
- if (local->xattr_req) {
- dict_del (local->xattr_req, TIER_LINKFILE_GFID);
- }
- }
-
- DHT_STRIP_PHASE1_FLAGS (stbuf);
-
- DHT_STACK_UNWIND (create, frame, op_ret, op_errno, fd, inode,
- stbuf, preparent, postparent, xdata);
-
- return 0;
-}
-
-int
-tier_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
-{
- dht_local_t *local = NULL;
- xlator_t *cached_subvol = NULL;
- dht_conf_t *conf = NULL;
- int ret = -1;
- unsigned char *gfid = NULL;
-
- local = frame->local;
- if (!local) {
- op_errno = EINVAL;
- goto err;
- }
-
- if (op_ret == -1) {
- local->op_errno = op_errno;
- goto err;
- }
-
- conf = this->private;
- if (!conf) {
- local->op_errno = EINVAL;
- op_errno = EINVAL;
- goto err;
- }
-
- cached_subvol = TIER_UNHASHED_SUBVOL;
-
- if (local->params) {
- dict_del (local->params, conf->link_xattr_name);
- dict_del (local->params, GLUSTERFS_INTERNAL_FOP_KEY);
- }
-
- /*
- * We will delete the linkfile if data file creation fails.
- * When deleting this stale linkfile, there is a possibility
- * for a race between this linkfile deletion and a stale
- * linkfile deletion triggered by another lookup from different
- * client.
- *
- * For eg:
- *
- * Client 1 Client 2
- *
- * 1 linkfile created for foo
- *
- * 2 data file creation failed
- *
- * 3 creating a file with same name
- *
- * 4 lookup before creation deleted
- * the linkfile created by client1
- * considering as a stale linkfile.
- *
- * 5 New linkfile created for foo
- * with different gfid.
- *
- * 6 Trigger linkfile deletion as
- * data file creation failed.
- *
- * 7 Linkfile deleted which is
- * created by client2.
- *
- * 8 Data file created.
- *
- * With this race, we will end up having a file in a non-hashed subvol
- * without a linkfile in hashed subvol.
- *
- * To avoid this, we store the gfid of linkfile created by client, So
- * If we delete the linkfile , we validate gfid of existing file with
- * stored value from posix layer.
- *
- * Storing this value in local->xattr_req as local->params was also used
- * to create the data file. During the linkfile deletion we will use
- * local->xattr_req dictionary.
- */
- if (!local->xattr_req) {
- local->xattr_req = dict_new ();
- if (!local->xattr_req) {
- local->op_errno = ENOMEM;
- op_errno = ENOMEM;
- goto err;
- }
- }
-
- gfid = GF_CALLOC (1, sizeof (uuid_t), gf_common_mt_char);
- if (!gfid) {
- local->op_errno = ENOMEM;
- op_errno = ENOMEM;
- goto err;
- }
-
- gf_uuid_copy (gfid, stbuf->ia_gfid);
- ret = dict_set_dynptr (local->xattr_req, TIER_LINKFILE_GFID,
- gfid, sizeof (uuid_t));
- if (ret) {
- GF_FREE (gfid);
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value"
- " : key = %s", TIER_LINKFILE_GFID);
- }
-
- STACK_WIND_COOKIE (frame, tier_create_cbk, cached_subvol,
- cached_subvol, cached_subvol->fops->create,
- &local->loc, local->flags, local->mode,
- local->umask, local->fd, local->params);
-
- return 0;
-err:
- DHT_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL,
- NULL, NULL, NULL);
- return 0;
-}
-
-gf_boolean_t
-tier_is_hot_tier_decommissioned (xlator_t *this)
-{
- dht_conf_t *conf = NULL;
- xlator_t *hot_tier = NULL;
- int i = 0;
-
- conf = this->private;
- hot_tier = conf->subvolumes[1];
-
- if (conf->decommission_subvols_cnt) {
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (conf->decommissioned_bricks[i] &&
- conf->decommissioned_bricks[i] == hot_tier)
- return _gf_true;
- }
- }
-
- return _gf_false;
-}
-
-int
-tier_create (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, mode_t mode,
- mode_t umask, fd_t *fd, dict_t *params)
-{
- int op_errno = -1;
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- xlator_t *hot_subvol = NULL;
- xlator_t *cold_subvol = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
-
- conf = this->private;
-
- dht_get_du_info (frame, this, loc);
-
- local = dht_local_init (frame, loc, fd, GF_FOP_CREATE);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
-
- cold_subvol = TIER_HASHED_SUBVOL;
- hot_subvol = TIER_UNHASHED_SUBVOL;
-
- if (conf->subvolumes[0] != cold_subvol) {
- hot_subvol = conf->subvolumes[0];
- }
- /*
- * if hot tier full, write to cold.
- * Also if hot tier is full, create in cold
- */
- if (dht_is_subvol_filled (this, hot_subvol) ||
- tier_is_hot_tier_decommissioned (this)) {
- gf_msg_debug (this->name, 0,
- "creating %s on %s", loc->path,
- cold_subvol->name);
-
- STACK_WIND_COOKIE (frame, tier_create_cbk, cold_subvol,
- cold_subvol, cold_subvol->fops->create,
- loc, flags, mode, umask, fd, params);
- } else {
- local->params = dict_ref (params);
- local->flags = flags;
- local->mode = mode;
- local->umask = umask;
- local->cached_subvol = hot_subvol;
- local->hashed_subvol = cold_subvol;
-
- gf_msg_debug (this->name, 0,
- "creating %s on %s (link at %s)", loc->path,
- hot_subvol->name, cold_subvol->name);
-
- dht_linkfile_create (frame, tier_create_linkfile_create_cbk,
- this, hot_subvol, cold_subvol, loc);
-
- goto out;
- }
-out:
- return 0;
-
-err:
-
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL,
- NULL, NULL, NULL);
-
- return 0;
-}
-
-int
-tier_unlink_nonhashed_linkfile_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
-
- local = frame->local;
- prev = cookie;
-
- LOCK (&frame->lock);
- {
- if ((op_ret == -1) && (op_errno != ENOENT)) {
- local->op_errno = op_errno;
- local->op_ret = op_ret;
- gf_msg_debug (this->name, op_errno,
- "Unlink link: subvolume %s"
- " returned -1",
- prev->name);
- goto unlock;
- }
-
- local->op_ret = 0;
- }
-unlock:
- UNLOCK (&frame->lock);
-
- if (local->op_ret == -1)
- goto err;
- DHT_STACK_UNWIND (unlink, frame, local->op_ret, local->op_errno,
- &local->preparent, &local->postparent, NULL);
-
-
- return 0;
-
-err:
- DHT_STACK_UNWIND (unlink, frame, -1, local->op_errno,
- NULL, NULL, NULL);
- return 0;
-}
-
-int
-tier_unlink_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode,
- struct iatt *preparent, dict_t *xdata,
- struct iatt *postparent)
-{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- dht_conf_t *conf = NULL;
- xlator_t *hot_subvol = NULL;
-
- local = frame->local;
- prev = cookie;
- conf = this->private;
- hot_subvol = TIER_UNHASHED_SUBVOL;
-
- if (!op_ret) {
- /*
- * linkfile present on hot tier. unlinking the linkfile
- */
- STACK_WIND_COOKIE (frame, tier_unlink_nonhashed_linkfile_cbk,
- hot_subvol, hot_subvol, hot_subvol->fops->unlink,
- &local->loc, local->flags, NULL);
- return 0;
- }
-
- LOCK (&frame->lock);
- {
- if (op_errno == ENOENT) {
- local->op_ret = 0;
- local->op_errno = op_errno;
- } else {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- }
- gf_msg_debug (this->name, op_errno,
- "Lookup : subvolume %s returned -1",
- prev->name);
- }
-
- UNLOCK (&frame->lock);
-
- DHT_STACK_UNWIND (unlink, frame, local->op_ret, local->op_errno,
- &local->preparent, &local->postparent, xdata);
-
- return 0;
-}
-
-int
-tier_unlink_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
-
- local = frame->local;
- prev = cookie;
-
- LOCK (&frame->lock);
- {
- /* Ignore EINVAL for tier to ignore error when the file
- does not exist on the other tier */
- if ((op_ret == -1) && !((op_errno == ENOENT) ||
- (op_errno == EINVAL))) {
- local->op_errno = op_errno;
- local->op_ret = op_ret;
- gf_msg_debug (this->name, op_errno,
- "Unlink link: subvolume %s"
- " returned -1",
- prev->name);
- goto unlock;
- }
-
- local->op_ret = 0;
- }
-unlock:
- UNLOCK (&frame->lock);
-
- if (local->op_ret == -1)
- goto err;
-
- DHT_STACK_UNWIND (unlink, frame, local->op_ret, local->op_errno,
- &local->preparent, &local->postparent, xdata);
-
- return 0;
-
-err:
- DHT_STACK_UNWIND (unlink, frame, -1, local->op_errno,
- NULL, NULL, NULL);
- return 0;
-}
-
-int32_t
-tier_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- dht_local_t *local = NULL;
- xlator_t *prev = NULL;
- struct iatt *stbuf = NULL;
- dht_conf_t *conf = NULL;
- int ret = -1;
- xlator_t *hot_tier = NULL;
- xlator_t *cold_tier = NULL;
-
- local = frame->local;
- prev = cookie;
- conf = this->private;
-
- cold_tier = TIER_HASHED_SUBVOL;
- hot_tier = TIER_UNHASHED_SUBVOL;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- if (op_errno == ENOENT) {
- local->op_ret = 0;
- } else {
- local->op_ret = -1;
- local->op_errno = op_errno;
- }
- gf_msg_debug (this->name, op_errno,
- "Unlink: subvolume %s returned -1"
- " with errno = %d",
- prev->name, op_errno);
- goto unlock;
- }
-
- local->op_ret = 0;
-
- local->postparent = *postparent;
- local->preparent = *preparent;
-
- if (local->loc.parent) {
- dht_inode_ctx_time_update (local->loc.parent, this,
- &local->preparent, 0);
- dht_inode_ctx_time_update (local->loc.parent, this,
- &local->postparent, 1);
- }
- }
-unlock:
- UNLOCK (&frame->lock);
-
- if (local->op_ret)
- goto out;
-
- if (cold_tier != local->cached_subvol) {
- /*
- * File is present in hot tier, so there will be
- * a link file on cold tier, deleting the linkfile
- * from cold tier
- */
- STACK_WIND_COOKIE (frame, tier_unlink_linkfile_cbk, cold_tier,
- cold_tier, cold_tier->fops->unlink,
- &local->loc,
- local->flags, xdata);
- return 0;
- }
-
- ret = dict_get_bin (xdata, DHT_IATT_IN_XDATA_KEY, (void **) &stbuf);
- if (!ret && stbuf && ((IS_DHT_MIGRATION_PHASE2 (stbuf)) ||
- IS_DHT_MIGRATION_PHASE1 (stbuf))) {
- /*
- * File is migrating from cold to hot tier.
- * Delete the destination linkfile.
- */
- STACK_WIND_COOKIE (frame, tier_unlink_lookup_cbk, hot_tier,
- hot_tier, hot_tier->fops->lookup,
- &local->loc, NULL);
- return 0;
-
- }
-
-out:
- DHT_STACK_UNWIND (unlink, frame, local->op_ret, local->op_errno,
- &local->preparent, &local->postparent, xdata);
-
- return 0;
-}
-
-int
-tier_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
- dict_t *xdata)
-{
- xlator_t *cached_subvol = NULL;
- xlator_t *hashed_subvol = NULL;
- dht_conf_t *conf = NULL;
- int op_errno = -1;
- dht_local_t *local = NULL;
- int ret = -1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
-
- conf = this->private;
-
- local = dht_local_init (frame, loc, NULL, GF_FOP_UNLINK);
- if (!local) {
- op_errno = ENOMEM;
-
- goto err;
- }
-
- hashed_subvol = TIER_HASHED_SUBVOL;
-
- cached_subvol = local->cached_subvol;
- if (!cached_subvol) {
- gf_msg_debug (this->name, 0,
- "no cached subvolume for path=%s", loc->path);
- op_errno = EINVAL;
- goto err;
- }
-
- local->flags = xflag;
- if (IA_ISREG (loc->inode->ia_type) &&
- (hashed_subvol == cached_subvol)) {
- /*
- * File resides in cold tier. We need to stat
- * the file to see if it is being promoted.
- * If yes we need to delete the destination
- * file as well.
- *
- * Currently we are doing this check only for
- * regular files.
- */
- xdata = xdata ? dict_ref (xdata) : dict_new ();
- if (xdata) {
- ret = dict_set_int8 (xdata, DHT_IATT_IN_XDATA_KEY, 1);
- if (ret) {
- gf_msg_debug (this->name, 0,
- "Failed to set dictionary key %s",
- DHT_IATT_IN_XDATA_KEY);
- }
- }
- }
-
- /*
- * File is on hot tier, delete the data file first, then
- * linkfile from cold.
- */
- STACK_WIND_COOKIE (frame, tier_unlink_cbk, cached_subvol,
- cached_subvol, cached_subvol->fops->unlink, loc,
- xflag, xdata);
- if (xdata)
- dict_unref (xdata);
- return 0;
-err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
-
- return 0;
-}
-
-int
-tier_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, gf_dirent_t *orig_entries,
- dict_t *xdata)
-{
- gf_dirent_t entries;
- gf_dirent_t *orig_entry = NULL;
- gf_dirent_t *entry = NULL;
- int count = 0;
-
- INIT_LIST_HEAD (&entries.list);
-
- if (op_ret < 0)
- goto unwind;
-
- list_for_each_entry (orig_entry, (&orig_entries->list), list) {
- entry = gf_dirent_for_name (orig_entry->d_name);
- if (!entry) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_NO_MEMORY,
- "Memory allocation failed ");
- goto unwind;
- }
-
- entry->d_off = orig_entry->d_off;
- entry->d_ino = orig_entry->d_ino;
- entry->d_type = orig_entry->d_type;
- entry->d_len = orig_entry->d_len;
-
- list_add_tail (&entry->list, &entries.list);
- count++;
- }
- op_ret = count;
-
-unwind:
- if (op_ret < 0)
- op_ret = 0;
-
- DHT_STACK_UNWIND (readdir, frame, op_ret, op_errno, &entries, NULL);
-
- gf_dirent_free (&entries);
-
- return 0;
-}
-
-int
-tier_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
- int op_errno, gf_dirent_t *orig_entries, dict_t *xdata)
-{
- dht_local_t *local = NULL;
- gf_dirent_t entries;
- gf_dirent_t *orig_entry = NULL;
- gf_dirent_t *entry = NULL;
- xlator_t *prev = NULL;
- xlator_t *next_subvol = NULL;
- off_t next_offset = 0;
- int count = 0;
- dht_conf_t *conf = NULL;
- int ret = 0;
- inode_table_t *itable = NULL;
- inode_t *inode = NULL;
-
- INIT_LIST_HEAD (&entries.list);
- prev = cookie;
- local = frame->local;
- itable = local->fd ? local->fd->inode->table : NULL;
-
- conf = this->private;
- GF_VALIDATE_OR_GOTO(this->name, conf, unwind);
-
- if (op_ret < 0)
- goto done;
-
- list_for_each_entry (orig_entry, (&orig_entries->list), list) {
- next_offset = orig_entry->d_off;
-
- if (IA_ISINVAL(orig_entry->d_stat.ia_type)) {
- /*stat failed somewhere- ignore this entry*/
- continue;
- }
-
- entry = gf_dirent_for_name (orig_entry->d_name);
- if (!entry) {
-
- goto unwind;
- }
-
- entry->d_off = orig_entry->d_off;
- entry->d_stat = orig_entry->d_stat;
- entry->d_ino = orig_entry->d_ino;
- entry->d_type = orig_entry->d_type;
- entry->d_len = orig_entry->d_len;
-
- if (orig_entry->dict)
- entry->dict = dict_ref (orig_entry->dict);
-
- if (check_is_linkfile (NULL, (&orig_entry->d_stat),
- orig_entry->dict,
- conf->link_xattr_name)) {
- goto entries;
-
- } else if (IA_ISDIR(entry->d_stat.ia_type)) {
- if (orig_entry->inode) {
- dht_inode_ctx_time_update (orig_entry->inode,
- this, &entry->d_stat,
- 1);
- }
- } else {
- if (orig_entry->inode) {
- ret = dht_layout_preset (this, prev,
- orig_entry->inode);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_LAYOUT_SET_FAILED,
- "failed to link the layout "
- "in inode");
-
- entry->inode = inode_ref (orig_entry->inode);
- } else if (itable) {
- /*
- * orig_entry->inode might be null if any upper
- * layer xlators below client set to null, to
- * force a lookup on the inode even if the inode
- * is present in the inode table. In that case
- * we just update the ctx to make sure we didn't
- * missed anything.
- */
- inode = inode_find (itable,
- orig_entry->d_stat.ia_gfid);
- if (inode) {
- ret = dht_layout_preset
- (this, TIER_HASHED_SUBVOL,
- inode);
- if (ret)
- gf_msg (this->name,
- GF_LOG_WARNING, 0,
- DHT_MSG_LAYOUT_SET_FAILED,
- "failed to link the layout"
- " in inode");
- inode_unref (inode);
- inode = NULL;
- }
- }
- }
-
-entries:
- list_add_tail (&entry->list, &entries.list);
- count++;
- }
- op_ret = count;
-
-done:
- if (count == 0) {
- /* non-zero next_offset means that
- EOF is not yet hit on the current subvol
- */
- if (next_offset != 0) {
- next_subvol = prev;
- } else {
- goto unwind;
- }
-
- STACK_WIND_COOKIE (frame, tier_readdirp_cbk, next_subvol,
- next_subvol, next_subvol->fops->readdirp,
- local->fd, local->size, next_offset,
- local->xattr);
- return 0;
- }
-
-unwind:
- if (op_ret < 0)
- op_ret = 0;
-
- DHT_STACK_UNWIND (readdirp, frame, op_ret, op_errno, &entries, NULL);
-
- gf_dirent_free (&entries);
-
- return 0;
-}
-
-int
-tier_do_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t yoff, int whichop, dict_t *dict)
-{
- dht_local_t *local = NULL;
- int op_errno = -1;
- xlator_t *hashed_subvol = NULL;
- int ret = 0;
- dht_conf_t *conf = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (this->private, err);
-
- conf = this->private;
-
- local = dht_local_init (frame, NULL, NULL, whichop);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local->fd = fd_ref (fd);
- local->size = size;
- local->xattr_req = (dict) ? dict_ref (dict) : NULL;
-
- hashed_subvol = TIER_HASHED_SUBVOL;
-
-
- /* TODO: do proper readdir */
- if (whichop == GF_FOP_READDIRP) {
- if (dict)
- local->xattr = dict_ref (dict);
- else
- local->xattr = dict_new ();
-
- if (local->xattr) {
- ret = dict_set_uint32 (local->xattr,
- conf->link_xattr_name, 256);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set dictionary value"
- " : key = %s",
- conf->link_xattr_name);
-
- }
-
- STACK_WIND_COOKIE (frame, tier_readdirp_cbk, hashed_subvol,
- hashed_subvol, hashed_subvol->fops->readdirp,
- fd, size, yoff, local->xattr);
-
- } else {
- STACK_WIND_COOKIE (frame, tier_readdir_cbk, hashed_subvol,
- hashed_subvol, hashed_subvol->fops->readdir,
- fd, size, yoff, local->xattr);
- }
-
- return 0;
-
-err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
-
- return 0;
-}
-
-int
-tier_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t yoff, dict_t *xdata)
-{
- int op = GF_FOP_READDIR;
- dht_conf_t *conf = NULL;
- int i = 0;
-
- conf = this->private;
- if (!conf)
- goto out;
-
- for (i = 0; i < conf->subvolume_cnt; i++) {
- if (!conf->subvolume_status[i]) {
- op = GF_FOP_READDIRP;
- break;
- }
- }
-
- if (conf->use_readdirp)
- op = GF_FOP_READDIRP;
-
-out:
- tier_do_readdir (frame, this, fd, size, yoff, op, 0);
- return 0;
-}
-
-int
-tier_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t yoff, dict_t *dict)
-{
- tier_do_readdir (frame, this, fd, size, yoff, GF_FOP_READDIRP, dict);
- return 0;
-}
-
-int
-tier_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct statvfs *statvfs,
- dict_t *xdata)
-{
- gf_boolean_t event = _gf_false;
- qdstatfs_action_t action = qdstatfs_action_OFF;
- dht_local_t *local = NULL;
- int this_call_cnt = 0;
- int bsize = 0;
- int frsize = 0;
- GF_UNUSED int ret = 0;
- unsigned long new_usage = 0;
- unsigned long cur_usage = 0;
- xlator_t *prev = NULL;
- dht_conf_t *conf = NULL;
- tier_statvfs_t *tier_stat = NULL;
-
- prev = cookie;
- local = frame->local;
- GF_ASSERT (local);
-
- conf = this->private;
-
- if (xdata)
- ret = dict_get_int8 (xdata, "quota-deem-statfs",
- (int8_t *)&event);
-
- tier_stat = &local->tier_statvfs;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- local->op_errno = op_errno;
- goto unlock;
- }
- if (!statvfs) {
- op_errno = EINVAL;
- local->op_ret = -1;
- goto unlock;
- }
- local->op_ret = 0;
-
- if (local->quota_deem_statfs) {
- if (event == _gf_true) {
- action = qdstatfs_action_COMPARE;
- } else {
- action = qdstatfs_action_NEGLECT;
- }
- } else {
-
- if (event == _gf_true) {
- action = qdstatfs_action_REPLACE;
- local->quota_deem_statfs = _gf_true;
- }
- }
-
- if (local->quota_deem_statfs) {
- switch (action) {
- case qdstatfs_action_NEGLECT:
- goto unlock;
-
- case qdstatfs_action_REPLACE:
- local->statvfs = *statvfs;
- goto unlock;
-
- case qdstatfs_action_COMPARE:
- new_usage = statvfs->f_blocks -
- statvfs->f_bfree;
- cur_usage = local->statvfs.f_blocks -
- local->statvfs.f_bfree;
-
- /* Take the max of the usage from subvols */
- if (new_usage >= cur_usage)
- local->statvfs = *statvfs;
- goto unlock;
-
- default:
- break;
- }
- }
-
- if (local->statvfs.f_bsize != 0) {
- bsize = max(local->statvfs.f_bsize, statvfs->f_bsize);
- frsize = max(local->statvfs.f_frsize, statvfs->f_frsize);
- dht_normalize_stats(&local->statvfs, bsize, frsize);
- dht_normalize_stats(statvfs, bsize, frsize);
- } else {
- local->statvfs.f_bsize = statvfs->f_bsize;
- local->statvfs.f_frsize = statvfs->f_frsize;
- }
-
- if (prev == TIER_HASHED_SUBVOL) {
- local->statvfs.f_blocks = statvfs->f_blocks;
- local->statvfs.f_files = statvfs->f_files;
- local->statvfs.f_fsid = statvfs->f_fsid;
- local->statvfs.f_flag = statvfs->f_flag;
- local->statvfs.f_namemax = statvfs->f_namemax;
- tier_stat->blocks_used = (statvfs->f_blocks - statvfs->f_bfree);
- tier_stat->pblocks_used = (statvfs->f_blocks - statvfs->f_bavail);
- tier_stat->files_used = (statvfs->f_files - statvfs->f_ffree);
- tier_stat->pfiles_used = (statvfs->f_files - statvfs->f_favail);
- tier_stat->hashed_fsid = statvfs->f_fsid;
- } else {
- tier_stat->unhashed_fsid = statvfs->f_fsid;
- tier_stat->unhashed_blocks_used = (statvfs->f_blocks - statvfs->f_bfree);
- tier_stat->unhashed_pblocks_used = (statvfs->f_blocks - statvfs->f_bavail);
- tier_stat->unhashed_files_used = (statvfs->f_files - statvfs->f_ffree);
- tier_stat->unhashed_pfiles_used = (statvfs->f_files - statvfs->f_favail);
- }
-
- }
-unlock:
- UNLOCK (&frame->lock);
-
- this_call_cnt = dht_frame_return (frame);
- if (is_last_call (this_call_cnt)) {
- if (tier_stat->unhashed_fsid != tier_stat->hashed_fsid) {
- tier_stat->blocks_used += tier_stat->unhashed_blocks_used;
- tier_stat->pblocks_used += tier_stat->unhashed_pblocks_used;
- tier_stat->files_used += tier_stat->unhashed_files_used;
- tier_stat->pfiles_used += tier_stat->unhashed_pfiles_used;
- }
- local->statvfs.f_bfree = local->statvfs.f_blocks -
- tier_stat->blocks_used;
- local->statvfs.f_bavail = local->statvfs.f_blocks -
- tier_stat->pblocks_used;
- local->statvfs.f_ffree = local->statvfs.f_files -
- tier_stat->files_used;
- local->statvfs.f_favail = local->statvfs.f_files -
- tier_stat->pfiles_used;
- DHT_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno,
- &local->statvfs, xdata);
- }
-
- return 0;
-}
-
-
-int
-tier_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
-{
- dht_local_t *local = NULL;
- dht_conf_t *conf = NULL;
- int op_errno = -1;
- int i = -1;
- inode_t *inode = NULL;
- inode_table_t *itable = NULL;
- uuid_t root_gfid = {0, };
- loc_t newloc = {0, };
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (this->private, err);
-
- conf = this->private;
-
- local = dht_local_init (frame, NULL, NULL, GF_FOP_STATFS);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- if (loc->inode && !IA_ISDIR (loc->inode->ia_type)) {
- itable = loc->inode->table;
- if (!itable) {
- op_errno = EINVAL;
- goto err;
- }
-
- loc = &local->loc2;
- root_gfid[15] = 1;
-
- inode = inode_find (itable, root_gfid);
- if (!inode) {
- op_errno = EINVAL;
- goto err;
- }
-
- dht_build_root_loc (inode, &newloc);
- loc = &newloc;
- }
-
- local->call_cnt = conf->subvolume_cnt;
-
- for (i = 0; i < conf->subvolume_cnt; i++) {
- STACK_WIND_COOKIE (frame, tier_statfs_cbk, conf->subvolumes[i],
- conf->subvolumes[i],
- conf->subvolumes[i]->fops->statfs, loc,
- xdata);
- }
-
- return 0;
-
-err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- DHT_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
-
- return 0;
-}
diff --git a/xlators/cluster/dht/src/tier-common.h b/xlators/cluster/dht/src/tier-common.h
deleted file mode 100644
index 26e01d7f141..00000000000
--- a/xlators/cluster/dht/src/tier-common.h
+++ /dev/null
@@ -1,68 +0,0 @@
-/*
- Copyright (c) 2015 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#ifndef _TIER_COMMON_H_
-#define _TIER_COMMON_H_
-/* Function definitions */
-int
-tier_create_unlink_stale_linkto_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata);
-
-int
-tier_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- fd_t *fd, inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata);
-
-int
-tier_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent,
- struct iatt *postparent,
- dict_t *xdata);
-
-int
-tier_create (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, mode_t mode,
- mode_t umask, fd_t *fd, dict_t *params);
-
-int32_t
-tier_unlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int xflag, dict_t *xdata);
-
-int32_t
-tier_readdirp (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- size_t size, off_t off, dict_t *dict);
-
-int
-tier_readdir (call_frame_t *frame,
- xlator_t *this, fd_t *fd, size_t size,
- off_t yoff, dict_t *xdata);
-
-
-
-int
-tier_link (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata);
-
-
-int
-tier_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata);
-
-
-#endif
-
diff --git a/xlators/cluster/dht/src/tier.c b/xlators/cluster/dht/src/tier.c
deleted file mode 100644
index 5fa50a26d80..00000000000
--- a/xlators/cluster/dht/src/tier.c
+++ /dev/null
@@ -1,3326 +0,0 @@
-/*
- Copyright (c) 2015 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#include <dlfcn.h>
-
-#include "dht-common.h"
-#include "tier.h"
-#include "tier-common.h"
-#include "syscall.h"
-#include "events.h"
-#include "tier-ctr-interface.h"
-
-/*Hard coded DB info*/
-static gfdb_db_type_t dht_tier_db_type = GFDB_SQLITE3;
-/*Hard coded DB info*/
-
-/*Mutex for updating the data movement stats*/
-static pthread_mutex_t dm_stat_mutex = PTHREAD_MUTEX_INITIALIZER;
-
-/* Stores the path location of promotion query files */
-static char *promotion_qfile;
-/* Stores the path location of demotion query files */
-static char *demotion_qfile;
-
-static void *libhandle;
-static gfdb_methods_t gfdb_methods;
-
-#define DB_QUERY_RECORD_SIZE 4096
-
-/*
- * Closes all the fds and frees the qfile_array
- * */
-static void
-qfile_array_free (tier_qfile_array_t *qfile_array)
-{
- ssize_t i = 0;
-
- if (qfile_array) {
- if (qfile_array->fd_array) {
- for (i = 0; i < qfile_array->array_size; i++) {
- if (qfile_array->fd_array[i] != -1) {
- sys_close (qfile_array->fd_array[i]);
- }
- }
- }
- GF_FREE (qfile_array->fd_array);
- }
- GF_FREE (qfile_array);
-}
-
-
-/* Create a new query file list with given size */
-static tier_qfile_array_t *
-qfile_array_new (ssize_t array_size)
-{
- int ret = -1;
- tier_qfile_array_t *qfile_array = NULL;
- ssize_t i = 0;
-
- GF_VALIDATE_OR_GOTO ("tier", (array_size > 0), out);
-
- qfile_array = GF_CALLOC (1, sizeof (tier_qfile_array_t),
- gf_tier_mt_qfile_array_t);
- if (!qfile_array) {
- gf_msg ("tier", GF_LOG_ERROR, 0, DHT_MSG_LOG_TIER_ERROR,
- "Failed to allocate memory for tier_qfile_array_t");
- goto out;
- }
-
- qfile_array->fd_array = GF_CALLOC (array_size, sizeof (int),
- gf_dht_mt_int32_t);
- if (!qfile_array->fd_array) {
- gf_msg ("tier", GF_LOG_ERROR, 0, DHT_MSG_LOG_TIER_ERROR,
- "Failed to allocate memory for "
- "tier_qfile_array_t->fd_array");
- goto out;
- }
-
- /* Init all the fds to -1 */
- for (i = 0; i < array_size; i++) {
- qfile_array->fd_array[i] = -1;
- }
-
- qfile_array->array_size = array_size;
- qfile_array->next_index = 0;
-
- /* Set exhausted count to list size as the list is empty */
- qfile_array->exhausted_count = qfile_array->array_size;
-
- ret = 0;
-out:
- if (ret) {
- qfile_array_free (qfile_array);
- qfile_array = NULL;
- }
- return qfile_array;
-}
-
-
-/* Checks if the query file list is empty or totally exhausted. */
-static gf_boolean_t
-is_qfile_array_empty (tier_qfile_array_t *qfile_array)
-{
- return (qfile_array->exhausted_count == qfile_array->array_size) ?
- _gf_true : _gf_false;
-}
-
-
-/* Shifts the next_fd pointer to the next available fd in the list */
-static void
-shift_next_index (tier_qfile_array_t *qfile_array)
-{
- int qfile_fd = 0;
- int spin_count = 0;
-
- if (is_qfile_array_empty (qfile_array)) {
- return;
- }
-
- do {
- /* change next_index in a rotional manner */
- (qfile_array->next_index == (qfile_array->array_size - 1)) ?
- qfile_array->next_index = 0 : qfile_array->next_index++;
-
- qfile_fd = (qfile_array->fd_array[qfile_array->next_index]);
-
- spin_count++;
-
- } while ((qfile_fd == -1) && (spin_count < qfile_array->array_size));
-
-}
-
-/*
- * This is a non-thread safe function to read query records
- * from a list of query files in a Round-Robin manner.
- * As in when the query files get exhuasted they are closed.
- * Returns:
- * 0 if all the query records in all the query files of the list are
- * exhausted.
- * > 0 if a query record is successfully read. Indicates the size of the query
- * record read.
- * < 0 if there was failure
- * */
-static int
-read_query_record_list (tier_qfile_array_t *qfile_array,
- gfdb_query_record_t **query_record)
-{
- int ret = -1;
- int qfile_fd = 0;
-
- GF_VALIDATE_OR_GOTO ("tier", qfile_array, out);
- GF_VALIDATE_OR_GOTO ("tier", qfile_array->fd_array, out);
-
- do {
- if (is_qfile_array_empty (qfile_array)) {
- ret = 0;
- break;
- }
-
- qfile_fd = qfile_array->fd_array[qfile_array->next_index];
- ret = gfdb_methods.gfdb_read_query_record
- (qfile_fd, query_record);
- if (ret <= 0) {
- /*The qfile_fd has reached EOF or
- * there was an error.
- * 1. Close the exhausted fd
- * 2. increment the exhausted count
- * 3. shift next_qfile to next qfile
- **/
- sys_close (qfile_fd);
- qfile_array->fd_array[qfile_array->next_index] = -1;
- qfile_array->exhausted_count++;
- /* shift next_qfile to next qfile */
- shift_next_index (qfile_array);
- continue;
- } else {
- /* shift next_qfile to next qfile */
- shift_next_index (qfile_array);
- break;
- }
- } while (1);
-out:
- return ret;
-}
-
-
-/* Check and update the watermark every WM_INTERVAL seconds */
-#define WM_INTERVAL 5
-#define WM_INTERVAL_EMERG 1
-
-static int
-tier_check_same_node (xlator_t *this, loc_t *loc, gf_defrag_info_t *defrag)
-{
- int ret = -1;
- dict_t *dict = NULL;
- char *uuid_str = NULL;
- uuid_t node_uuid = {0,};
-
- GF_VALIDATE_OR_GOTO ("tier", this, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, defrag, out);
-
- if (syncop_getxattr (this, loc, &dict, GF_XATTR_NODE_UUID_KEY,
- NULL, NULL)) {
- gf_msg (this->name, GF_LOG_ERROR, 0, DHT_MSG_LOG_TIER_ERROR,
- "Unable to get NODE_UUID_KEY %s %s\n",
- loc->name, loc->path);
- goto out;
- }
-
- if (dict_get_str (dict, GF_XATTR_NODE_UUID_KEY, &uuid_str) < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0, DHT_MSG_LOG_TIER_ERROR,
- "Failed to get node-uuids for %s", loc->path);
- goto out;
- }
-
-
- if (gf_uuid_parse (uuid_str, node_uuid)) {
- gf_msg (this->name, GF_LOG_ERROR, 0, DHT_MSG_LOG_TIER_ERROR,
- "uuid_parse failed for %s", loc->path);
- goto out;
- }
-
- if (gf_uuid_compare (node_uuid, defrag->node_uuid)) {
- gf_msg_debug (this->name, 0,
- "%s does not belong to this node", loc->path);
- ret = 1;
- goto out;
- }
-
- ret = 0;
-out:
- if (dict)
- dict_unref(dict);
-
- return ret;
-}
-
-int
-tier_get_fs_stat (xlator_t *this, loc_t *root_loc)
-{
- int ret = 0;
- gf_defrag_info_t *defrag = NULL;
- dht_conf_t *conf = NULL;
- dict_t *xdata = NULL;
- struct statvfs statfs = {0, };
- gf_tier_conf_t *tier_conf = NULL;
-
-
- conf = this->private;
- if (!conf) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "conf is NULL");
- ret = -1;
- goto exit;
- }
-
- defrag = conf->defrag;
- if (!defrag) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "defrag is NULL");
- ret = -1;
- goto exit;
- }
-
- tier_conf = &defrag->tier_conf;
-
- xdata = dict_new ();
- if (!xdata) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- DHT_MSG_NO_MEMORY,
- "failed to allocate dictionary");
- ret = -1;
- goto exit;
- }
-
- ret = dict_set_int8 (xdata, GF_INTERNAL_IGNORE_DEEM_STATFS, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_DICT_SET_FAILED,
- "Failed to set "
- GF_INTERNAL_IGNORE_DEEM_STATFS" in dict");
- ret = -1;
- goto exit;
- }
-
- /* Find how much free space is on the hot subvolume.
- * Then see if that value */
- /* is less than or greater than user defined watermarks.
- * Stash results in */
- /* the tier_conf data structure. */
-
- ret = syncop_statfs (conf->subvolumes[1], root_loc, &statfs,
- xdata, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_LOG_TIER_STATUS,
- "Unable to obtain statfs.");
- goto exit;
- }
-
- pthread_mutex_lock (&dm_stat_mutex);
-
- tier_conf->block_size = statfs.f_bsize;
- tier_conf->blocks_total = statfs.f_blocks;
- tier_conf->blocks_used = statfs.f_blocks - statfs.f_bfree;
-
- tier_conf->percent_full = GF_PERCENTAGE(tier_conf->blocks_used,
- statfs.f_blocks);
- pthread_mutex_unlock (&dm_stat_mutex);
-
-exit:
- if (xdata)
- dict_unref (xdata);
- return ret;
-}
-
-static void
-tier_send_watermark_event (const char *volname,
- tier_watermark_op_t old_wm,
- tier_watermark_op_t new_wm)
-{
- if (old_wm == TIER_WM_LOW || old_wm == TIER_WM_NONE) {
- if (new_wm == TIER_WM_MID) {
- gf_event (EVENT_TIER_WATERMARK_RAISED_TO_MID,
- "vol=%s", volname);
- } else if (new_wm == TIER_WM_HI) {
- gf_event (EVENT_TIER_WATERMARK_HI, "vol=%s", volname);
- }
- } else if (old_wm == TIER_WM_MID) {
- if (new_wm == TIER_WM_LOW) {
- gf_event (EVENT_TIER_WATERMARK_DROPPED_TO_LOW,
- "vol=%s", volname);
- } else if (new_wm == TIER_WM_HI) {
- gf_event (EVENT_TIER_WATERMARK_HI, "vol=%s", volname);
- }
- } else if (old_wm == TIER_WM_HI) {
- if (new_wm == TIER_WM_MID) {
- gf_event (EVENT_TIER_WATERMARK_DROPPED_TO_MID,
- "vol=%s", volname);
- } else if (new_wm == TIER_WM_LOW) {
- gf_event (EVENT_TIER_WATERMARK_DROPPED_TO_LOW,
- "vol=%s", volname);
- }
- }
-}
-
-int
-tier_check_watermark (xlator_t *this)
-{
- int ret = -1;
- gf_defrag_info_t *defrag = NULL;
- dht_conf_t *conf = NULL;
- gf_tier_conf_t *tier_conf = NULL;
- tier_watermark_op_t wm = TIER_WM_NONE;
-
- conf = this->private;
- if (!conf)
- goto exit;
-
- defrag = conf->defrag;
- if (!defrag)
- goto exit;
-
- tier_conf = &defrag->tier_conf;
-
- if (tier_conf->percent_full < tier_conf->watermark_low) {
- wm = TIER_WM_LOW;
-
- } else if (tier_conf->percent_full < tier_conf->watermark_hi) {
- wm = TIER_WM_MID;
-
- } else {
- wm = TIER_WM_HI;
- }
-
- if (wm != tier_conf->watermark_last) {
-
- tier_send_watermark_event (tier_conf->volname,
- tier_conf->watermark_last,
- wm);
-
- tier_conf->watermark_last = wm;
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "Tier watermark now %d", wm);
- }
-
- ret = 0;
-
-exit:
- return ret;
-}
-
-
-static gf_boolean_t
-is_hot_tier_full (gf_tier_conf_t *tier_conf)
-{
- if (tier_conf && (tier_conf->mode == TIER_MODE_WM) &&
- (tier_conf->watermark_last == TIER_WM_HI))
- return _gf_true;
-
- return _gf_false;
-}
-
-int
-tier_do_migration (xlator_t *this, int promote)
-{
- gf_defrag_info_t *defrag = NULL;
- dht_conf_t *conf = NULL;
- long rand = 0;
- int migrate = 0;
- gf_tier_conf_t *tier_conf = NULL;
-
- conf = this->private;
- if (!conf)
- goto exit;
-
- defrag = conf->defrag;
- if (!defrag)
- goto exit;
-
- if (tier_check_watermark (this) != 0) {
- gf_msg (this->name, GF_LOG_CRITICAL, errno,
- DHT_MSG_LOG_TIER_ERROR,
- "Failed to get watermark");
- goto exit;
- }
-
- tier_conf = &defrag->tier_conf;
-
- switch (tier_conf->watermark_last) {
- case TIER_WM_LOW:
- migrate = promote ? 1 : 0;
- break;
- case TIER_WM_HI:
- migrate = promote ? 0 : 1;
- break;
- case TIER_WM_MID:
- /* coverity[DC.WEAK_CRYPTO] */
- rand = random() % 100;
- if (promote) {
- migrate = (rand > tier_conf->percent_full);
- } else {
- migrate = (rand <= tier_conf->percent_full);
- }
- break;
- }
-
-exit:
- return migrate;
-}
-
-int
-tier_migrate (xlator_t *this, int is_promotion, dict_t *migrate_data,
- loc_t *loc, gf_tier_conf_t *tier_conf)
-{
- int ret = -1;
-
- pthread_mutex_lock (&tier_conf->pause_mutex);
- if (is_promotion)
- tier_conf->promote_in_progress = 1;
- else
- tier_conf->demote_in_progress = 1;
- pthread_mutex_unlock (&tier_conf->pause_mutex);
-
- /* Data migration */
- ret = syncop_setxattr (this, loc, migrate_data, 0,
- NULL, NULL);
-
- pthread_mutex_lock (&tier_conf->pause_mutex);
- if (is_promotion)
- tier_conf->promote_in_progress = 0;
- else
- tier_conf->demote_in_progress = 0;
- pthread_mutex_unlock (&tier_conf->pause_mutex);
-
- return ret;
-}
-
-/* returns _gf_true: if file can be promoted
- * returns _gf_false: if file cannot be promoted
- */
-static gf_boolean_t
-tier_can_promote_file (xlator_t *this, char const *file_name,
- struct iatt *current, gf_defrag_info_t *defrag)
-{
- gf_boolean_t ret = _gf_false;
- fsblkcnt_t estimated_usage = 0;
-
- if (defrag->tier_conf.tier_max_promote_size &&
- (current->ia_size > defrag->tier_conf.tier_max_promote_size)) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "File %s (gfid:%s) with size (%lu) exceeds maxsize "
- "(%d) for promotion. File will not be promoted.",
- file_name,
- uuid_utoa(current->ia_gfid),
- current->ia_size,
- defrag->tier_conf.tier_max_promote_size);
- goto err;
- }
-
- /* bypass further validations for TEST mode */
- if (defrag->tier_conf.mode != TIER_MODE_WM) {
- ret = _gf_true;
- goto err;
- }
-
- /* convert the file size to blocks as per the block size of the
- * destination tier
- * NOTE: add (block_size - 1) to get the correct block size when
- * there is a remainder after a modulo
- */
- estimated_usage = ((current->ia_size + defrag->tier_conf.block_size - 1) /
- defrag->tier_conf.block_size) +
- defrag->tier_conf.blocks_used;
-
- /* test if the estimated block usage goes above HI watermark */
- if (GF_PERCENTAGE (estimated_usage, defrag->tier_conf.blocks_total) >=
- defrag->tier_conf.watermark_hi) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "Estimated block count consumption on "
- "hot tier (%lu) exceeds hi watermark (%d%%). "
- "File will not be promoted.",
- estimated_usage,
- defrag->tier_conf.watermark_hi);
- goto err;
- }
- ret = _gf_true;
-err:
- return ret;
-}
-
-static int
-tier_set_migrate_data (dict_t *migrate_data)
-{
- int failed = 1;
-
-
- failed = dict_set_str (migrate_data, GF_XATTR_FILE_MIGRATE_KEY, "force");
- if (failed) {
- goto bail_out;
- }
-
- /* Flag to suggest the xattr call is from migrator */
- failed = dict_set_str (migrate_data, "from.migrator", "yes");
- if (failed) {
- goto bail_out;
- }
-
- /* Flag to suggest its a tiering migration
- * The reason for this dic key-value is that
- * promotions and demotions are multithreaded
- * so the original frame from gf_defrag_start()
- * is not carried. A new frame will be created when
- * we do syncop_setxattr(). This doesnot have the
- * frame->root->pid of the original frame. So we pass
- * this dic key-value when we do syncop_setxattr() to do
- * data migration and set the frame->root->pid to
- * GF_CLIENT_PID_TIER_DEFRAG in dht_setxattr() just before
- * calling dht_start_rebalance_task() */
- failed = dict_set_str (migrate_data, TIERING_MIGRATION_KEY, "yes");
- if (failed) {
- goto bail_out;
- }
-
- failed = 0;
-
-bail_out:
- return failed;
-}
-
-static char *
-tier_get_parent_path (xlator_t *this,
- loc_t *p_loc,
- struct iatt *par_stbuf,
- int *per_link_status)
-{
- int ret = -1;
- char *parent_path = NULL;
- dict_t *xdata_request = NULL;
- dict_t *xdata_response = NULL;
-
-
- xdata_request = dict_new ();
- if (!xdata_request) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Failed to create xdata_request dict");
- goto err;
- }
- ret = dict_set_int32 (xdata_request,
- GET_ANCESTRY_PATH_KEY, 42);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Failed to set value to dict : key %s \n",
- GET_ANCESTRY_PATH_KEY);
- goto err;
- }
-
- ret = syncop_lookup (this, p_loc, par_stbuf, NULL,
- xdata_request, &xdata_response);
- /* When the parent gfid is a stale entry, the lookup
- * will fail and stop the demotion process.
- * The parent gfid can be stale when a huge folder is
- * deleted while the files within it are being migrated
- */
- if (ret == -ESTALE) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_STALE_LOOKUP,
- "Stale entry in parent lookup for %s",
- uuid_utoa (p_loc->gfid));
- *per_link_status = 1;
- goto err;
- } else if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_LOG_TIER_ERROR,
- "Error in parent lookup for %s",
- uuid_utoa (p_loc->gfid));
- *per_link_status = -1;
- goto err;
- }
- ret = dict_get_str (xdata_response, GET_ANCESTRY_PATH_KEY, &parent_path);
- if (ret || !parent_path) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Failed to get parent path for %s",
- uuid_utoa (p_loc->gfid));
- *per_link_status = -1;
- goto err;
- }
-
-err:
- if (xdata_request) {
- dict_unref (xdata_request);
- }
-
- if (xdata_response) {
- dict_unref (xdata_response);
- xdata_response = NULL;
- }
-
- return parent_path;
-}
-
-static int
-tier_get_file_name_and_path (xlator_t *this,
- uuid_t gfid,
- gfdb_link_info_t *link_info,
- char const *parent_path,
- loc_t *loc,
- int *per_link_status)
-{
- int ret = -1;
-
- loc->name = gf_strdup (link_info->file_name);
- if (!loc->name) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR, "Memory "
- "allocation failed for %s",
- uuid_utoa (gfid));
- *per_link_status = -1;
- goto err;
- }
- ret = gf_asprintf((char **)&(loc->path), "%s/%s", parent_path, loc->name);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR, "Failed to "
- "construct file path for %s %s\n",
- parent_path, loc->name);
- *per_link_status = -1;
- goto err;
- }
-
- ret = 0;
-
-err:
- return ret;
-}
-
-static int
-tier_lookup_file (xlator_t *this,
- loc_t *p_loc,
- loc_t *loc,
- struct iatt *current,
- int *per_link_status)
-{
- int ret = -1;
-
- ret = syncop_lookup (this, loc, current, NULL, NULL, NULL);
-
- /* The file may be deleted even when the parent
- * is available and the lookup will
- * return a stale entry which would stop the
- * migration. so if its a stale entry, then skip
- * the file and keep migrating.
- */
- if (ret == -ESTALE) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- DHT_MSG_STALE_LOOKUP,
- "Stale lookup for %s",
- uuid_utoa (p_loc->gfid));
- *per_link_status = 1;
- goto err;
- } else if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_LOG_TIER_ERROR, "Failed to "
- "lookup file %s\n", loc->name);
- *per_link_status = -1;
- goto err;
- }
- ret = 0;
-
-err:
- return ret;
-}
-
-static gf_boolean_t
-tier_is_file_already_at_destination (xlator_t *src_subvol,
- query_cbk_args_t *query_cbk_args,
- dht_conf_t *conf,
- int *per_link_status)
-{
- gf_boolean_t at_destination = _gf_true;
-
- if (src_subvol == NULL) {
- *per_link_status = 1;
- goto err;
- }
- if (query_cbk_args->is_promotion &&
- src_subvol == conf->subvolumes[1]) {
- *per_link_status = 1;
- goto err;
- }
-
- if (!query_cbk_args->is_promotion &&
- src_subvol == conf->subvolumes[0]) {
- *per_link_status = 1;
- goto err;
- }
- at_destination = _gf_false;
-
-err:
- return at_destination;
-}
-
-static void
-tier_update_migration_counters (query_cbk_args_t *query_cbk_args,
- gf_defrag_info_t *defrag,
- uint64_t *total_migrated_bytes,
- int *total_files)
-{
- if (query_cbk_args->is_promotion) {
- defrag->total_files_promoted++;
- *total_migrated_bytes +=
- defrag->tier_conf.st_last_promoted_size;
- pthread_mutex_lock (&dm_stat_mutex);
- defrag->tier_conf.blocks_used +=
- defrag->tier_conf.st_last_promoted_size;
- pthread_mutex_unlock (&dm_stat_mutex);
- } else {
- defrag->total_files_demoted++;
- *total_migrated_bytes +=
- defrag->tier_conf.st_last_demoted_size;
- pthread_mutex_lock (&dm_stat_mutex);
- defrag->tier_conf.blocks_used -=
- defrag->tier_conf.st_last_demoted_size;
- pthread_mutex_unlock (&dm_stat_mutex);
- }
- if (defrag->tier_conf.blocks_total) {
- pthread_mutex_lock (&dm_stat_mutex);
- defrag->tier_conf.percent_full =
- GF_PERCENTAGE (defrag->tier_conf.blocks_used,
- defrag->tier_conf.blocks_total);
- pthread_mutex_unlock (&dm_stat_mutex);
- }
-
- (*total_files)++;
-}
-
-static int
-tier_migrate_link (xlator_t *this,
- dht_conf_t *conf,
- uuid_t gfid,
- gfdb_link_info_t *link_info,
- gf_defrag_info_t *defrag,
- query_cbk_args_t *query_cbk_args,
- dict_t *migrate_data,
- int *per_link_status,
- int *total_files,
- uint64_t *total_migrated_bytes)
-{
- int ret = -1;
- struct iatt current = {0,};
- struct iatt par_stbuf = {0,};
- loc_t p_loc = {0,};
- loc_t loc = {0,};
- xlator_t *src_subvol = NULL;
- inode_t *linked_inode = NULL;
- char *parent_path = NULL;
-
-
- /* Lookup for parent and get the path of parent */
- gf_uuid_copy (p_loc.gfid, link_info->pargfid);
- p_loc.inode = inode_new (defrag->root_inode->table);
- if (!p_loc.inode) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Failed to create reference to inode"
- " for %s", uuid_utoa (p_loc.gfid));
-
- *per_link_status = -1;
- goto err;
- }
-
- parent_path = tier_get_parent_path (this, &p_loc, &par_stbuf,
- per_link_status);
- if (!parent_path) {
- goto err;
- }
-
- linked_inode = inode_link (p_loc.inode, NULL, NULL, &par_stbuf);
- inode_unref (p_loc.inode);
- p_loc.inode = linked_inode;
-
-
- /* Preparing File Inode */
- gf_uuid_copy (loc.gfid, gfid);
- loc.inode = inode_new (defrag->root_inode->table);
- gf_uuid_copy (loc.pargfid, link_info->pargfid);
- loc.parent = inode_ref (p_loc.inode);
-
- /* Get filename and Construct file path */
- if (tier_get_file_name_and_path (this, gfid, link_info,
- parent_path, &loc, per_link_status) != 0) {
- goto err;
- }
- gf_uuid_copy (loc.parent->gfid, link_info->pargfid);
-
- /* lookup file inode */
- if (tier_lookup_file (this, &p_loc, &loc,
- &current, per_link_status) != 0) {
- goto err;
- }
-
- if (query_cbk_args->is_promotion) {
- if (!tier_can_promote_file (this,
- link_info->file_name,
- &current,
- defrag)) {
- *per_link_status = 1;
- goto err;
- }
- }
-
- linked_inode = inode_link (loc.inode, NULL, NULL, &current);
- inode_unref (loc.inode);
- loc.inode = linked_inode;
-
-
- /*
- * Do not promote/demote if file already is where it
- * should be. It means another brick moved the file
- * so is not an error. So we set per_link_status = 1
- * so that we ignore counting this.
- */
- src_subvol = dht_subvol_get_cached (this, loc.inode);
-
- if (tier_is_file_already_at_destination (src_subvol, query_cbk_args,
- conf, per_link_status)) {
- goto err;
- }
-
- gf_msg_debug (this->name, 0, "Tier %s: src_subvol %s file %s",
- (query_cbk_args->is_promotion ? "promote" : "demote"),
- src_subvol->name,
- loc.path);
-
-
- ret = tier_check_same_node (this, &loc, defrag);
- if (ret != 0) {
- if (ret < 0) {
- *per_link_status = -1;
- goto err;
- }
- ret = 0;
- /* By setting per_link_status to 1 we are
- * ignoring this status and will not be counting
- * this file for migration */
- *per_link_status = 1;
- goto err;
- }
-
- gf_uuid_copy (loc.gfid, loc.inode->gfid);
-
- if (gf_defrag_get_pause_state (&defrag->tier_conf) != TIER_RUNNING) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "Tiering paused. "
- "Exiting tier_migrate_link");
- goto err;
- }
-
- ret = tier_migrate (this, query_cbk_args->is_promotion,
- migrate_data, &loc, &defrag->tier_conf);
-
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- DHT_MSG_LOG_TIER_ERROR, "Failed to "
- "migrate %s ", loc.path);
- *per_link_status = -1;
- goto err;
- }
-
- tier_update_migration_counters (query_cbk_args, defrag,
- total_migrated_bytes, total_files);
-
- ret = 0;
-
-err:
- GF_FREE ((char *) loc.name);
- loc.name = NULL;
- loc_wipe (&loc);
- loc_wipe (&p_loc);
-
- if ((*total_files >= defrag->tier_conf.max_migrate_files)
- || (*total_migrated_bytes > defrag->tier_conf.max_migrate_bytes)) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "Reached cycle migration limit."
- "migrated bytes %"PRId64" files %d",
- *total_migrated_bytes,
- *total_files);
- ret = -1;
- }
-
- return ret;
-}
-
-
-static int
-tier_migrate_using_query_file (void *_args)
-{
- int ret = -1;
- query_cbk_args_t *query_cbk_args = (query_cbk_args_t *) _args;
- xlator_t *this = NULL;
- gf_defrag_info_t *defrag = NULL;
- gfdb_query_record_t *query_record = NULL;
- gfdb_link_info_t *link_info = NULL;
- dict_t *migrate_data = NULL;
- /*
- * per_file_status and per_link_status
- * 0 : success
- * -1 : failure
- * 1 : ignore the status and dont count for migration
- * */
- int per_file_status = 0;
- int per_link_status = 0;
- int total_status = 0;
- dht_conf_t *conf = NULL;
- uint64_t total_migrated_bytes = 0;
- int total_files = 0;
- loc_t root_loc = { 0 };
- gfdb_time_t start_time = { 0 };
- gfdb_time_t current_time = { 0 };
- int total_time = 0;
- int max_time = 0;
- gf_boolean_t emergency_demote_mode = _gf_false;
-
-
- GF_VALIDATE_OR_GOTO ("tier", query_cbk_args, out);
- GF_VALIDATE_OR_GOTO ("tier", query_cbk_args->this, out);
- this = query_cbk_args->this;
- GF_VALIDATE_OR_GOTO (this->name, query_cbk_args->defrag, out);
- GF_VALIDATE_OR_GOTO (this->name, query_cbk_args->qfile_array, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
-
- conf = this->private;
-
- defrag = query_cbk_args->defrag;
- migrate_data = dict_new ();
- if (!migrate_data)
- goto out;
-
- emergency_demote_mode = (!query_cbk_args->is_promotion &&
- is_hot_tier_full(&defrag->tier_conf));
-
- if (tier_set_migrate_data (migrate_data) != 0) {
- goto out;
- }
-
- dht_build_root_loc (defrag->root_inode, &root_loc);
-
- ret = gettimeofday (&start_time, NULL);
- if (query_cbk_args->is_promotion) {
- max_time = defrag->tier_conf.tier_promote_frequency;
- } else {
- max_time = defrag->tier_conf.tier_demote_frequency;
- }
-
- /* Per file */
- while ((ret = read_query_record_list (query_cbk_args->qfile_array,
- &query_record)) != 0) {
-
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Failed to fetch query record "
- "from query file");
- goto out;
- }
-
- if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Exiting tier migration as"
- "defrag status is not started");
- goto out;
- }
-
- ret = gettimeofday (&current_time, NULL);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Could not get current time.");
- goto out;
- }
-
- total_time = current_time.tv_sec - start_time.tv_sec;
- if (total_time > max_time) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "Max cycle time reached. Exiting migration.");
- goto out;
- }
-
- per_file_status = 0;
- per_link_status = 0;
-
- if (gf_defrag_get_pause_state (&defrag->tier_conf)
- != TIER_RUNNING) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "Tiering paused. "
- "Exiting tier_migrate_using_query_file");
- break;
- }
-
- if (defrag->tier_conf.mode == TIER_MODE_WM) {
- ret = tier_get_fs_stat (this, &root_loc);
- if (ret != 0) {
- gfdb_methods.gfdb_query_record_free (query_record);
- query_record = NULL;
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "tier_get_fs_stat() FAILED ... "
- "skipping file migrations until next cycle");
- break;
- }
-
- if (!tier_do_migration (this, query_cbk_args->is_promotion)) {
- gfdb_methods.gfdb_query_record_free (query_record);
- query_record = NULL;
-
- /* We have crossed the high watermark. Stop processing
- * files if this is a promotion cycle so demotion gets
- * a chance to start if not already running*/
-
- if (query_cbk_args->is_promotion &&
- is_hot_tier_full (&defrag->tier_conf)) {
-
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "High watermark crossed during "
- "promotion. Exiting "
- "tier_migrate_using_query_file");
- break;
- }
- continue;
- }
- }
-
- per_link_status = 0;
-
- /* For now we only support single link migration. And we will
- * ignore other hard links in the link info list of query record
- * TODO: Multiple hard links migration */
- if (!list_empty (&query_record->link_list)) {
- link_info = list_first_entry (&query_record->link_list,
- gfdb_link_info_t, list);
- }
- if (link_info != NULL) {
- if (tier_migrate_link (this,
- conf,
- query_record->gfid,
- link_info,
- defrag,
- query_cbk_args,
- migrate_data,
- &per_link_status,
- &total_files,
- &total_migrated_bytes) != 0) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "%s failed for %s(gfid:%s)",
- (query_cbk_args->is_promotion ?
- "Promotion" : "Demotion"),
- link_info->file_name,
- uuid_utoa (query_record->gfid));
- }
- }
- per_file_status = per_link_status;
-
- if (per_file_status < 0) {/* Failure */
- pthread_mutex_lock (&dm_stat_mutex);
- defrag->total_failures++;
- pthread_mutex_unlock (&dm_stat_mutex);
- } else if (per_file_status == 0) {/* Success */
- pthread_mutex_lock (&dm_stat_mutex);
- defrag->total_files++;
- pthread_mutex_unlock (&dm_stat_mutex);
- } else if (per_file_status == 1) {/* Ignore */
- per_file_status = 0;
- /* Since this attempt was ignored we
- * decrement the lookup count*/
- pthread_mutex_lock (&dm_stat_mutex);
- defrag->num_files_lookedup--;
- pthread_mutex_unlock (&dm_stat_mutex);
- }
- total_status = total_status + per_file_status;
- per_link_status = 0;
- per_file_status = 0;
-
- gfdb_methods.gfdb_query_record_free (query_record);
- query_record = NULL;
-
- /* If we are demoting and the entry watermark was HI, then
- * we are done with emergency demotions if the current
- * watermark has fallen below hi-watermark level
- */
- if (emergency_demote_mode) {
- if (tier_check_watermark (this) == 0) {
- if (!is_hot_tier_full (&defrag->tier_conf)) {
- break;
- }
- }
- }
- }
-
-out:
- if (migrate_data)
- dict_unref (migrate_data);
-
- gfdb_methods.gfdb_query_record_free (query_record);
- query_record = NULL;
-
- return total_status;
-}
-
-
-/* This is the call back function per record/file from data base */
-static int
-tier_gf_query_callback (gfdb_query_record_t *gfdb_query_record,
- void *_args) {
- int ret = -1;
- query_cbk_args_t *query_cbk_args = _args;
-
- GF_VALIDATE_OR_GOTO ("tier", query_cbk_args, out);
- GF_VALIDATE_OR_GOTO ("tier", query_cbk_args->defrag, out);
- GF_VALIDATE_OR_GOTO ("tier", (query_cbk_args->query_fd > 0), out);
-
- ret = gfdb_methods.gfdb_write_query_record (query_cbk_args->query_fd,
- gfdb_query_record);
- if (ret) {
- gf_msg ("tier", GF_LOG_ERROR, 0, DHT_MSG_LOG_TIER_ERROR,
- "Failed writing query record to query file");
- goto out;
- }
-
- pthread_mutex_lock (&dm_stat_mutex);
- query_cbk_args->defrag->num_files_lookedup++;
- pthread_mutex_unlock (&dm_stat_mutex);
-
- ret = 0;
-out:
- return ret;
-}
-
-
-
-
-/* Create query file in tier process */
-static int
-tier_process_self_query (tier_brick_list_t *local_brick, void *args)
-{
- int ret = -1;
- char *db_path = NULL;
- query_cbk_args_t *query_cbk_args = NULL;
- xlator_t *this = NULL;
- gfdb_conn_node_t *conn_node = NULL;
- dict_t *params_dict = NULL;
- dict_t *ctr_ipc_dict = NULL;
- gfdb_brick_info_t *gfdb_brick_info = args;
-
- /*Init of all the essentials*/
- GF_VALIDATE_OR_GOTO ("tier", gfdb_brick_info , out);
- query_cbk_args = gfdb_brick_info->_query_cbk_args;
-
- GF_VALIDATE_OR_GOTO ("tier", query_cbk_args->this, out);
- this = query_cbk_args->this;
-
- GF_VALIDATE_OR_GOTO (this->name,
- gfdb_brick_info->_query_cbk_args, out);
-
- GF_VALIDATE_OR_GOTO (this->name, local_brick, out);
-
- GF_VALIDATE_OR_GOTO (this->name, local_brick->xlator, out);
-
- GF_VALIDATE_OR_GOTO (this->name, local_brick->brick_db_path, out);
-
- db_path = local_brick->brick_db_path;
-
- /*Preparing DB parameters before init_db i.e getting db connection*/
- params_dict = dict_new ();
- if (!params_dict) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "DB Params cannot initialized");
- goto out;
- }
- SET_DB_PARAM_TO_DICT(this->name, params_dict,
- (char *) gfdb_methods.get_db_path_key(),
- db_path, ret, out);
-
- /*Get the db connection*/
- conn_node = gfdb_methods.init_db ((void *)params_dict, dht_tier_db_type);
- if (!conn_node) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "FATAL: Failed initializing db operations");
- goto out;
- }
-
- /* Query for eligible files from db */
- query_cbk_args->query_fd = open (local_brick->qfile_path,
- O_WRONLY | O_CREAT | O_APPEND,
- S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
- if (query_cbk_args->query_fd < 0) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- DHT_MSG_LOG_TIER_ERROR,
- "Failed to open query file %s",
- local_brick->qfile_path);
- goto out;
- }
- if (!gfdb_brick_info->_gfdb_promote) {
- if (query_cbk_args->defrag->tier_conf.watermark_last ==
- TIER_WM_HI) {
- /* emergency demotion mode */
- ret = gfdb_methods.find_all (conn_node,
- tier_gf_query_callback,
- (void *)query_cbk_args,
- query_cbk_args->defrag->tier_conf.
- query_limit);
- } else {
- if (query_cbk_args->defrag->write_freq_threshold == 0 &&
- query_cbk_args->defrag->read_freq_threshold == 0) {
- ret = gfdb_methods.find_unchanged_for_time (
- conn_node,
- tier_gf_query_callback,
- (void *)query_cbk_args,
- gfdb_brick_info->time_stamp);
- } else {
- ret = gfdb_methods.find_unchanged_for_time_freq (
- conn_node,
- tier_gf_query_callback,
- (void *)query_cbk_args,
- gfdb_brick_info->time_stamp,
- query_cbk_args->defrag->
- write_freq_threshold,
- query_cbk_args->defrag->
- read_freq_threshold,
- _gf_false);
- }
- }
- } else {
- if (query_cbk_args->defrag->write_freq_threshold == 0 &&
- query_cbk_args->defrag->read_freq_threshold == 0) {
- ret = gfdb_methods.find_recently_changed_files (
- conn_node,
- tier_gf_query_callback,
- (void *)query_cbk_args,
- gfdb_brick_info->time_stamp);
- } else {
- ret = gfdb_methods.find_recently_changed_files_freq (
- conn_node,
- tier_gf_query_callback,
- (void *)query_cbk_args,
- gfdb_brick_info->time_stamp,
- query_cbk_args->defrag->write_freq_threshold,
- query_cbk_args->defrag->read_freq_threshold,
- _gf_false);
- }
- }
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "FATAL: query from db failed");
- goto out;
- }
-
- /*Clear the heat on the DB entries*/
- /*Preparing ctr_ipc_dict*/
- ctr_ipc_dict = dict_new ();
- if (!ctr_ipc_dict) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "ctr_ipc_dict cannot initialized");
- goto out;
- }
-
- SET_DB_PARAM_TO_DICT(this->name, ctr_ipc_dict,
- GFDB_IPC_CTR_KEY, GFDB_IPC_CTR_CLEAR_OPS,
- ret, out);
-
- ret = syncop_ipc (local_brick->xlator, GF_IPC_TARGET_CTR, ctr_ipc_dict,
- NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR, "Failed clearing the heat "
- "on db %s error %d", local_brick->brick_db_path, ret);
- goto out;
- }
-
- ret = 0;
-out:
- if (params_dict) {
- dict_unref (params_dict);
- params_dict = NULL;
- }
-
- if (ctr_ipc_dict) {
- dict_unref (ctr_ipc_dict);
- ctr_ipc_dict = NULL;
- }
-
- if (query_cbk_args && query_cbk_args->query_fd >= 0) {
- sys_close (query_cbk_args->query_fd);
- query_cbk_args->query_fd = -1;
- }
- gfdb_methods.fini_db (conn_node);
-
- return ret;
-}
-
-
-
-
-
-/*Ask CTR to create the query file*/
-static int
-tier_process_ctr_query (tier_brick_list_t *local_brick, void *args)
-{
- int ret = -1;
- query_cbk_args_t *query_cbk_args = NULL;
- xlator_t *this = NULL;
- dict_t *ctr_ipc_in_dict = NULL;
- dict_t *ctr_ipc_out_dict = NULL;
- gfdb_brick_info_t *gfdb_brick_info = args;
- gfdb_ipc_ctr_params_t *ipc_ctr_params = NULL;
- int count = 0;
-
- /*Init of all the essentials*/
- GF_VALIDATE_OR_GOTO ("tier", gfdb_brick_info , out);
- query_cbk_args = gfdb_brick_info->_query_cbk_args;
-
- GF_VALIDATE_OR_GOTO ("tier", query_cbk_args->this, out);
- this = query_cbk_args->this;
-
- GF_VALIDATE_OR_GOTO (this->name,
- gfdb_brick_info->_query_cbk_args, out);
-
- GF_VALIDATE_OR_GOTO (this->name, local_brick, out);
-
- GF_VALIDATE_OR_GOTO (this->name, local_brick->xlator, out);
-
- GF_VALIDATE_OR_GOTO (this->name, local_brick->brick_db_path, out);
-
-
- /*Preparing ctr_ipc_in_dict*/
- ctr_ipc_in_dict = dict_new ();
- if (!ctr_ipc_in_dict) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "ctr_ipc_in_dict cannot initialized");
- goto out;
- }
-
- ipc_ctr_params = GF_CALLOC (1, sizeof (gfdb_ipc_ctr_params_t),
- gf_tier_mt_ipc_ctr_params_t);
- if (!ipc_ctr_params) {
- goto out;
- }
-
- /* set all the query params*/
- ipc_ctr_params->is_promote = gfdb_brick_info->_gfdb_promote;
-
- ipc_ctr_params->write_freq_threshold =
- query_cbk_args->defrag->write_freq_threshold;
-
- ipc_ctr_params->read_freq_threshold =
- query_cbk_args->defrag->read_freq_threshold;
-
- ipc_ctr_params->query_limit =
- query_cbk_args->defrag->tier_conf.query_limit;
-
- ipc_ctr_params->emergency_demote =
- (!gfdb_brick_info->_gfdb_promote &&
- query_cbk_args->defrag->tier_conf.watermark_last ==
- TIER_WM_HI);
-
- memcpy (&ipc_ctr_params->time_stamp,
- gfdb_brick_info->time_stamp,
- sizeof (gfdb_time_t));
-
- SET_DB_PARAM_TO_DICT (this->name, ctr_ipc_in_dict,
- GFDB_IPC_CTR_KEY, GFDB_IPC_CTR_QUERY_OPS,
- ret, out);
-
-
- SET_DB_PARAM_TO_DICT (this->name, ctr_ipc_in_dict,
- GFDB_IPC_CTR_GET_QFILE_PATH,
- local_brick->qfile_path,
- ret, out);
-
- ret = dict_set_bin (ctr_ipc_in_dict, GFDB_IPC_CTR_GET_QUERY_PARAMS,
- ipc_ctr_params, sizeof (*ipc_ctr_params));
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, LG_MSG_SET_PARAM_FAILED,
- "Failed setting %s to params dictionary",
- GFDB_IPC_CTR_GET_QUERY_PARAMS);
- GF_FREE (ipc_ctr_params);
- goto out;
- }
- ipc_ctr_params = NULL;
-
- ret = syncop_ipc (local_brick->xlator, GF_IPC_TARGET_CTR,
- ctr_ipc_in_dict, &ctr_ipc_out_dict);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_IPC_TIER_ERROR, "Failed query on %s ret %d",
- local_brick->brick_db_path, ret);
- goto out;
- }
-
- ret = dict_get_int32(ctr_ipc_out_dict, GFDB_IPC_CTR_RET_QUERY_COUNT,
- &count);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR, "Failed getting count "
- "of records on %s",
- local_brick->brick_db_path);
- goto out;
- }
-
- if (count < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR, "Failed query on %s",
- local_brick->brick_db_path);
- ret = -1;
- goto out;
- }
-
- pthread_mutex_lock (&dm_stat_mutex);
- query_cbk_args->defrag->num_files_lookedup = count;
- pthread_mutex_unlock (&dm_stat_mutex);
-
- ret = 0;
-out:
-
- if (ctr_ipc_in_dict) {
- dict_unref(ctr_ipc_in_dict);
- ctr_ipc_in_dict = NULL;
- }
-
- if (ctr_ipc_out_dict) {
- dict_unref(ctr_ipc_out_dict);
- ctr_ipc_out_dict = NULL;
- }
-
- GF_FREE (ipc_ctr_params);
-
- return ret;
-}
-
-
-
-
-/* This is the call back function for each brick from hot/cold bricklist
- * It picks up each bricks db and queries for eligible files for migration.
- * The list of eligible files are populated in appropriate query files*/
-static int
-tier_process_brick (tier_brick_list_t *local_brick, void *args) {
- int ret = -1;
- dict_t *ctr_ipc_in_dict = NULL;
- dict_t *ctr_ipc_out_dict = NULL;
- char *strval = NULL;
-
- GF_VALIDATE_OR_GOTO ("tier", local_brick, out);
-
- GF_VALIDATE_OR_GOTO ("tier", local_brick->xlator, out);
-
- if (dht_tier_db_type == GFDB_SQLITE3) {
-
- /*Preparing ctr_ipc_in_dict*/
- ctr_ipc_in_dict = dict_new ();
- if (!ctr_ipc_in_dict) {
- gf_msg ("tier", GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "ctr_ipc_in_dict cannot initialized");
- goto out;
- }
-
- ret = dict_set_str (ctr_ipc_in_dict, GFDB_IPC_CTR_KEY,
- GFDB_IPC_CTR_GET_DB_PARAM_OPS);
- if (ret) {
- gf_msg ("tier", GF_LOG_ERROR, 0,\
- LG_MSG_SET_PARAM_FAILED, "Failed to set %s "
- "to params dictionary", GFDB_IPC_CTR_KEY);
- goto out;
- }
-
- ret = dict_set_str (ctr_ipc_in_dict,
- GFDB_IPC_CTR_GET_DB_PARAM_OPS, "");
- if (ret) {
- gf_msg ("tier", GF_LOG_ERROR, 0,\
- LG_MSG_SET_PARAM_FAILED, "Failed to set %s "
- "to params dictionary",
- GFDB_IPC_CTR_GET_DB_PARAM_OPS);
- goto out;
- }
-
- ret = dict_set_str (ctr_ipc_in_dict,
- GFDB_IPC_CTR_GET_DB_KEY, "journal_mode");
- if (ret) {
- gf_msg ("tier", GF_LOG_ERROR, 0,
- LG_MSG_SET_PARAM_FAILED, "Failed to set %s "
- "to params dictionary",
- GFDB_IPC_CTR_GET_DB_KEY);
- goto out;
- }
-
-
-
- ret = syncop_ipc (local_brick->xlator, GF_IPC_TARGET_CTR,
- ctr_ipc_in_dict, &ctr_ipc_out_dict);
- if (ret || ctr_ipc_out_dict == NULL) {
- gf_msg ("tier", GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR, "Failed to get "
- "journal_mode of sql db %s",
- local_brick->brick_db_path);
- goto out;
- }
-
- ret = dict_get_str (ctr_ipc_out_dict, "journal_mode", &strval);
- if (ret) {
- gf_msg ("tier", GF_LOG_ERROR, 0,
- LG_MSG_GET_PARAM_FAILED, "Failed to get %s "
- "from params dictionary"
- "journal_mode", strval);
- goto out;
- }
-
- if (strval && (strncmp(strval, "wal", strlen ("wal")) == 0)) {
- ret = tier_process_self_query (local_brick, args);
- if (ret) {
- goto out;
- }
- } else {
- ret = tier_process_ctr_query (local_brick, args);
- if (ret) {
- goto out;
- }
- }
- ret = 0;
-
- } else {
- ret = tier_process_self_query (local_brick, args);
- if (ret) {
- goto out;
- }
- }
-
- ret = 0;
-out:
- if (ctr_ipc_in_dict)
- dict_unref (ctr_ipc_in_dict);
-
- if (ctr_ipc_out_dict)
- dict_unref (ctr_ipc_out_dict);
-
- return ret;
-}
-
-
-
-
-static int
-tier_build_migration_qfile (migration_args_t *args,
- query_cbk_args_t *query_cbk_args,
- gf_boolean_t is_promotion)
-{
- gfdb_time_t current_time;
- gfdb_brick_info_t gfdb_brick_info;
- gfdb_time_t time_in_past;
- int ret = -1;
- tier_brick_list_t *local_brick = NULL;
- int i = 0;
- time_in_past.tv_sec = args->freq_time;
- time_in_past.tv_usec = 0;
-
- ret = gettimeofday (&current_time, NULL);
- if (ret == -1) {
- gf_msg (args->this->name, GF_LOG_ERROR, errno,
- DHT_MSG_SYS_CALL_GET_TIME_FAILED,
- "Failed to get current time");
- goto out;
- }
- time_in_past.tv_sec = current_time.tv_sec - time_in_past.tv_sec;
-
- /* The migration daemon may run a varying numberof usec after the */
- /* sleep call triggers. A file may be registered in CTR some number */
- /* of usec X after the daemon started and missed in the subsequent */
- /* cycle if the daemon starts Y usec after the period in seconds */
- /* where Y>X. Normalize away this problem by always setting usec */
- /* to 0. */
- time_in_past.tv_usec = 0;
-
- gfdb_brick_info.time_stamp = &time_in_past;
- gfdb_brick_info._gfdb_promote = is_promotion;
- gfdb_brick_info._query_cbk_args = query_cbk_args;
-
- list_for_each_entry (local_brick, args->brick_list, list) {
-
- /* Construct query file path for this brick
- * i.e
- * /var/run/gluster/xlator_name/
- * {promote/demote}-brickname-indexinbricklist
- * So that no two query files will have same path even
- * bricks have the same name
- * */
- snprintf (local_brick->qfile_path, PATH_MAX , "%s-%s-%d",
- GET_QFILE_PATH (gfdb_brick_info._gfdb_promote),
- local_brick->brick_name, i);
-
- /* Delete any old query files for this brick */
- sys_unlink (local_brick->qfile_path);
-
- ret = tier_process_brick (local_brick,
- &gfdb_brick_info);
- if (ret) {
- gf_msg (args->this->name, GF_LOG_ERROR, 0,
- DHT_MSG_BRICK_QUERY_FAILED,
- "Brick %s query failed\n",
- local_brick->brick_db_path);
- }
- i++;
- }
- ret = 0;
-out:
- return ret;
-}
-
-static int
-tier_migrate_files_using_qfile (migration_args_t *comp,
- query_cbk_args_t *query_cbk_args)
-{
- int ret = -1;
- tier_brick_list_t *local_brick = NULL;
- tier_brick_list_t *temp = NULL;
- char query_file_path_err[PATH_MAX+128] = {0,};
- struct tm tm = {0,};
- gfdb_time_t current_time = {0,};
- char time_str[128] = {0,};
- ssize_t qfile_array_size = 0;
- int count = 0;
- int temp_fd = 0;
- gf_tier_conf_t *tier_conf = NULL;
-
- tier_conf = &(query_cbk_args->defrag->tier_conf);
-
- /* Time format for error query files */
- gettimeofday (&current_time, NULL);
- gmtime_r (&current_time.tv_sec, &tm);
- strftime (time_str, sizeof (time_str), "%F-%T", &tm);
-
- /* Build the qfile list */
- list_for_each_entry_safe (local_brick, temp, comp->brick_list, list) {
- qfile_array_size++;
- }
- query_cbk_args->qfile_array = qfile_array_new (qfile_array_size);
- if (!query_cbk_args->qfile_array) {
- gf_msg ("tier", GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR, "Failed to create new "
- "qfile_array");
- goto out;
- }
-
- /*Open all qfiles*/
- count = 0;
- query_cbk_args->qfile_array->exhausted_count = 0;
- list_for_each_entry_safe (local_brick, temp, comp->brick_list, list) {
- temp_fd = query_cbk_args->qfile_array->fd_array[count];
- temp_fd = open (local_brick->qfile_path, O_RDONLY,
- S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
- if (temp_fd < 0) {
- gf_msg ("tier", GF_LOG_ERROR, errno,
- DHT_MSG_LOG_TIER_ERROR, "Failed to open "
- "%s to the query file",
- local_brick->qfile_path);
- query_cbk_args->qfile_array->exhausted_count++;
- }
- query_cbk_args->qfile_array->fd_array[count] = temp_fd;
- count++;
- }
-
- /* Moving the query file index to the next, so that we won't the same
- * query file every cycle as the first one */
- query_cbk_args->qfile_array->next_index =
- (query_cbk_args->is_promotion) ?
- tier_conf->last_promote_qfile_index :
- tier_conf->last_demote_qfile_index;
- shift_next_index (query_cbk_args->qfile_array);
- if (query_cbk_args->is_promotion) {
- tier_conf->last_promote_qfile_index =
- query_cbk_args->qfile_array->next_index;
- } else {
- tier_conf->last_demote_qfile_index =
- query_cbk_args->qfile_array->next_index;
- }
-
- /* Migrate files using query file list */
- ret = tier_migrate_using_query_file ((void *)query_cbk_args);
-out:
- qfile_array_free (query_cbk_args->qfile_array);
-
- /* If there is an error rename all the query files to .err files
- * with a timestamp for better debugging */
- if (ret) {
- list_for_each_entry_safe (local_brick, temp, comp->brick_list,
- list) {
- /* rename error qfile*/
- snprintf (query_file_path_err,
- sizeof (query_file_path_err),
- "%s-%s.err", local_brick->qfile_path,
- time_str);
- if (sys_rename (local_brick->qfile_path,
- query_file_path_err) == -1)
- gf_msg_debug ("tier", 0, "rename "
- "failed");
- }
- }
-
- query_cbk_args->qfile_array = NULL;
-
- return ret;
-}
-
-
-
-int
-tier_demote (migration_args_t *demotion_args)
-{
- query_cbk_args_t query_cbk_args;
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO ("tier", demotion_args, out);
- GF_VALIDATE_OR_GOTO ("tier", demotion_args->this, out);
- GF_VALIDATE_OR_GOTO (demotion_args->this->name,
- demotion_args->brick_list, out);
- GF_VALIDATE_OR_GOTO (demotion_args->this->name,
- demotion_args->defrag, out);
-
- THIS = demotion_args->this;
-
- query_cbk_args.this = demotion_args->this;
- query_cbk_args.defrag = demotion_args->defrag;
- query_cbk_args.is_promotion = 0;
-
- /*Build the query file using bricklist*/
- ret = tier_build_migration_qfile (demotion_args, &query_cbk_args,
- _gf_false);
- if (ret)
- goto out;
-
- /* Migrate files using the query file */
- ret = tier_migrate_files_using_qfile (demotion_args,
- &query_cbk_args);
- if (ret)
- goto out;
-
-out:
- demotion_args->return_value = ret;
- return ret;
-}
-
-
-int
-tier_promote (migration_args_t *promotion_args)
-{
- int ret = -1;
- query_cbk_args_t query_cbk_args;
-
- GF_VALIDATE_OR_GOTO ("tier", promotion_args->this, out);
- GF_VALIDATE_OR_GOTO (promotion_args->this->name,
- promotion_args->brick_list, out);
- GF_VALIDATE_OR_GOTO (promotion_args->this->name,
- promotion_args->defrag, out);
-
- THIS = promotion_args->this;
-
- query_cbk_args.this = promotion_args->this;
- query_cbk_args.defrag = promotion_args->defrag;
- query_cbk_args.is_promotion = 1;
-
- /*Build the query file using bricklist*/
- ret = tier_build_migration_qfile (promotion_args, &query_cbk_args,
- _gf_true);
- if (ret)
- goto out;
-
- /* Migrate files using the query file */
- ret = tier_migrate_files_using_qfile (promotion_args, &query_cbk_args);
- if (ret)
- goto out;
-
-out:
- promotion_args->return_value = ret;
- return ret;
-}
-
-
-/*
- * Command the CTR on a brick to compact the local database using an IPC
- */
-static int
-tier_process_self_compact (tier_brick_list_t *local_brick, void *args)
-{
- int ret = -1;
- char *db_path = NULL;
- query_cbk_args_t *query_cbk_args = NULL;
- xlator_t *this = NULL;
- gfdb_conn_node_t *conn_node = NULL;
- dict_t *params_dict = NULL;
- dict_t *ctr_ipc_dict = NULL;
- gfdb_brick_info_t *gfdb_brick_info = args;
-
- /*Init of all the essentials*/
- GF_VALIDATE_OR_GOTO ("tier", gfdb_brick_info , out);
- query_cbk_args = gfdb_brick_info->_query_cbk_args;
-
- GF_VALIDATE_OR_GOTO ("tier", query_cbk_args->this, out);
- this = query_cbk_args->this;
-
- GF_VALIDATE_OR_GOTO (this->name,
- gfdb_brick_info->_query_cbk_args, out);
-
- GF_VALIDATE_OR_GOTO (this->name, local_brick, out);
-
- GF_VALIDATE_OR_GOTO (this->name, local_brick->xlator, out);
-
- GF_VALIDATE_OR_GOTO (this->name, local_brick->brick_db_path, out);
-
- db_path = local_brick->brick_db_path;
-
- /*Preparing DB parameters before init_db i.e getting db connection*/
- params_dict = dict_new ();
- if (!params_dict) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "DB Params cannot initialized");
- goto out;
- }
- SET_DB_PARAM_TO_DICT (this->name, params_dict,
- (char *) gfdb_methods.get_db_path_key(), db_path,
- ret, out);
-
- /*Get the db connection*/
- conn_node = gfdb_methods.init_db ((void *)params_dict,
- dht_tier_db_type);
- if (!conn_node) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "FATAL: Failed initializing db operations");
- goto out;
- }
-
- ret = 0;
-
- /*Preparing ctr_ipc_dict*/
- ctr_ipc_dict = dict_new ();
- if (!ctr_ipc_dict) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "ctr_ipc_dict cannot initialized");
- goto out;
- }
-
- ret = dict_set_int32 (ctr_ipc_dict, "compact_active",
- query_cbk_args->defrag->
- tier_conf.compact_active);
-
- if (ret) {
- gf_msg ("tier", GF_LOG_ERROR, 0,
- LG_MSG_SET_PARAM_FAILED, "Failed to set %s "
- "to params dictionary",
- "compact_active");
- goto out;
- }
-
- ret = dict_set_int32 (ctr_ipc_dict, "compact_mode_switched",
- query_cbk_args->defrag->
- tier_conf.compact_mode_switched);
-
- if (ret) {
- gf_msg ("tier", GF_LOG_ERROR, 0,
- LG_MSG_SET_PARAM_FAILED, "Failed to set %s "
- "to params dictionary",
- "compact_mode_switched");
- goto out;
- }
-
- SET_DB_PARAM_TO_DICT(this->name, ctr_ipc_dict,
- GFDB_IPC_CTR_KEY, GFDB_IPC_CTR_SET_COMPACT_PRAGMA,
- ret, out);
-
- gf_msg (this->name, GF_LOG_TRACE, 0, DHT_MSG_LOG_TIER_STATUS,
- "Starting Compaction IPC");
-
- ret = syncop_ipc (local_brick->xlator, GF_IPC_TARGET_CTR, ctr_ipc_dict,
- NULL);
-
- gf_msg (this->name, GF_LOG_TRACE, 0, DHT_MSG_LOG_TIER_STATUS,
- "Ending Compaction IPC");
-
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR, "Failed compaction "
- "on db %s error %d", local_brick->brick_db_path, ret);
- goto out;
- }
-
- gf_msg (this->name, GF_LOG_TRACE, 0, DHT_MSG_LOG_TIER_STATUS,
- "SUCCESS: %s Compaction", local_brick->brick_name);
-
- ret = 0;
-out:
- if (params_dict) {
- dict_unref (params_dict);
- params_dict = NULL;
- }
-
- if (ctr_ipc_dict) {
- dict_unref (ctr_ipc_dict);
- ctr_ipc_dict = NULL;
- }
-
- gfdb_methods.fini_db (conn_node);
-
- return ret;
-}
-
-/*
- * This is the call back function for each brick from hot/cold bricklist.
- * It determines the database type on each brick and calls the corresponding
- * function to prepare the compaction IPC.
- */
-static int
-tier_compact_db_brick (tier_brick_list_t *local_brick, void *args)
-{
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO ("tier", local_brick, out);
-
- GF_VALIDATE_OR_GOTO ("tier", local_brick->xlator, out);
-
- ret = tier_process_self_compact (local_brick, args);
- if (ret) {
- gf_msg ("tier", GF_LOG_INFO, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "Brick %s did not compact",
- local_brick->brick_name);
- goto out;
- }
-
- ret = 0;
-
-out:
-
- return ret;
-}
-
-static int
-tier_send_compact (migration_args_t *args,
- query_cbk_args_t *query_cbk_args)
-{
- gfdb_time_t current_time;
- gfdb_brick_info_t gfdb_brick_info;
- gfdb_time_t time_in_past;
- int ret = -1;
- tier_brick_list_t *local_brick = NULL;
-
- time_in_past.tv_sec = args->freq_time;
- time_in_past.tv_usec = 0;
-
- ret = gettimeofday (&current_time, NULL);
- if (ret == -1) {
- gf_msg (args->this->name, GF_LOG_ERROR, errno,
- DHT_MSG_SYS_CALL_GET_TIME_FAILED,
- "Failed to get current time");
- goto out;
- }
- time_in_past.tv_sec = current_time.tv_sec - time_in_past.tv_sec;
-
- /* The migration daemon may run a varying numberof usec after the sleep
- call triggers. A file may be registered in CTR some number of usec X
- after the daemon started and missed in the subsequent cycle if the
- daemon starts Y usec after the period in seconds where Y>X. Normalize
- away this problem by always setting usec to 0. */
- time_in_past.tv_usec = 0;
-
- gfdb_brick_info.time_stamp = &time_in_past;
-
- /* This is meant to say we are always compacting at this point */
- /* We simply borrow the promotion flag to do this */
- gfdb_brick_info._gfdb_promote = 1;
-
- gfdb_brick_info._query_cbk_args = query_cbk_args;
-
- list_for_each_entry (local_brick, args->brick_list, list) {
-
- gf_msg (args->this->name, GF_LOG_TRACE, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "Start compaction for %s",
- local_brick->brick_name);
-
- ret = tier_compact_db_brick (local_brick,
- &gfdb_brick_info);
- if (ret) {
- gf_msg (args->this->name, GF_LOG_ERROR, 0,
- DHT_MSG_BRICK_QUERY_FAILED,
- "Brick %s compaction failed\n",
- local_brick->brick_db_path);
- }
-
- gf_msg (args->this->name, GF_LOG_TRACE, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "End compaction for %s",
- local_brick->brick_name);
-
- }
- ret = 0;
-out:
- return ret;
-}
-
-static int
-tier_compact (void *args)
-{
- int ret = -1;
- query_cbk_args_t query_cbk_args;
- migration_args_t *compaction_args = args;
-
- GF_VALIDATE_OR_GOTO ("tier", compaction_args->this, out);
- GF_VALIDATE_OR_GOTO (compaction_args->this->name,
- compaction_args->brick_list, out);
- GF_VALIDATE_OR_GOTO (compaction_args->this->name,
- compaction_args->defrag, out);
-
- THIS = compaction_args->this;
-
- query_cbk_args.this = compaction_args->this;
- query_cbk_args.defrag = compaction_args->defrag;
- query_cbk_args.is_compaction = 1;
-
- /* Send the compaction pragma out to all the bricks on the bricklist. */
- /* tier_get_bricklist ensures all bricks on the list are local to */
- /* this node. */
- ret = tier_send_compact (compaction_args, &query_cbk_args);
- if (ret)
- goto out;
-
- ret = 0;
-out:
- compaction_args->return_value = ret;
- return ret;
- }
-
-static int
-tier_get_bricklist (xlator_t *xl, struct list_head *local_bricklist_head)
-{
- xlator_list_t *child = NULL;
- char *rv = NULL;
- char *rh = NULL;
- char *brickname = NULL;
- char db_name[PATH_MAX] = "";
- int ret = 0;
- tier_brick_list_t *local_brick = NULL;
-
- GF_VALIDATE_OR_GOTO ("tier", xl, out);
- GF_VALIDATE_OR_GOTO ("tier", local_bricklist_head, out);
-
- /*
- * This function obtains remote subvolumes and filters out only
- * those running on the same node as the tier daemon.
- */
- if (strcmp(xl->type, "protocol/client") == 0) {
- ret = dict_get_str (xl->options, "remote-host", &rh);
- if (ret < 0)
- goto out;
-
- if (gf_is_local_addr (rh)) {
-
- local_brick = GF_CALLOC (1, sizeof(tier_brick_list_t),
- gf_tier_mt_bricklist_t);
- if (!local_brick) {
- goto out;
- }
-
- ret = dict_get_str (xl->options, "remote-subvolume",
- &rv);
- if (ret < 0)
- goto out;
-
- brickname = strrchr(rv, '/') + 1;
- snprintf(db_name, sizeof(db_name), "%s.db",
- brickname);
-
- local_brick->brick_db_path =
- GF_CALLOC (PATH_MAX, 1, gf_common_mt_char);
- if (!local_brick->brick_db_path) {
- gf_msg ("tier", GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "Failed to allocate memory for"
- " bricklist.");
- goto out;
- }
-
- snprintf(local_brick->brick_db_path,
- PATH_MAX, "%s/%s/%s", rv,
- GF_HIDDEN_PATH, db_name);
-
- local_brick->xlator = xl;
-
- snprintf (local_brick->brick_name,
- NAME_MAX, "%s", brickname);
-
- list_add_tail (&(local_brick->list),
- local_bricklist_head);
-
- ret = 0;
- goto out;
- }
- }
-
- for (child = xl->children; child; child = child->next) {
- ret = tier_get_bricklist (child->xlator, local_bricklist_head);
- if (ret) {
- goto out;
- }
- }
-
- ret = 0;
-out:
-
- if (ret) {
- if (local_brick) {
- GF_FREE (local_brick->brick_db_path);
- }
- GF_FREE (local_brick);
- }
-
- return ret;
-}
-
-int
-tier_get_freq_demote (gf_tier_conf_t *tier_conf)
-{
- if ((tier_conf->mode == TIER_MODE_WM) &&
- (tier_conf->watermark_last == TIER_WM_HI))
- return DEFAULT_DEMOTE_DEGRADED;
- else
- return tier_conf->tier_demote_frequency;
-}
-
-int
-tier_get_freq_promote (gf_tier_conf_t *tier_conf)
-{
- return tier_conf->tier_promote_frequency;
-}
-
-int
-tier_get_freq_compact_hot (gf_tier_conf_t *tier_conf)
-{
- return tier_conf->tier_compact_hot_frequency;
-}
-
-int
-tier_get_freq_compact_cold (gf_tier_conf_t *tier_conf)
-{
- return tier_conf->tier_compact_cold_frequency;
-}
-
-static int
-tier_check_demote (gfdb_time_t current_time, int freq)
-{
- return ((current_time.tv_sec % freq) == 0) ?
- _gf_true : _gf_false;
-}
-
-static gf_boolean_t
-tier_check_promote (gf_tier_conf_t *tier_conf,
- gfdb_time_t current_time,
- int freq)
-{
- if ((tier_conf->mode == TIER_MODE_WM) &&
- (tier_conf->watermark_last == TIER_WM_HI))
- return _gf_false;
-
- else
- return ((current_time.tv_sec % freq) == 0) ?
- _gf_true : _gf_false;
-}
-
-static gf_boolean_t
-tier_check_compact (gf_tier_conf_t *tier_conf,
- gfdb_time_t current_time,
- int freq_compact)
-{
-
- if (!(tier_conf->compact_active ||
- tier_conf->compact_mode_switched))
- return _gf_false;
-
-
- return ((current_time.tv_sec % freq_compact) == 0) ?
- _gf_true : _gf_false;
-}
-
-
-void
-clear_bricklist (struct list_head *brick_list)
-{
- tier_brick_list_t *local_brick = NULL;
- tier_brick_list_t *temp = NULL;
-
- if (list_empty(brick_list)) {
- return;
- }
-
- list_for_each_entry_safe (local_brick, temp, brick_list, list) {
- list_del (&local_brick->list);
- GF_FREE (local_brick->brick_db_path);
- GF_FREE (local_brick);
- }
-}
-
-
-static void
-set_brick_list_qpath (struct list_head *brick_list, gf_boolean_t is_cold)
-{
-
- tier_brick_list_t *local_brick = NULL;
- int i = 0;
-
- GF_VALIDATE_OR_GOTO ("tier", brick_list, out);
-
- list_for_each_entry (local_brick, brick_list, list) {
-
- /* Construct query file path for this brick
- * i.e
- * /var/run/gluster/xlator_name/
- * {promote/demote}-brickname-indexinbricklist
- * So that no two query files will have same path even
- * bricks have the same name
- * */
- snprintf (local_brick->qfile_path, PATH_MAX , "%s-%s-%d",
- GET_QFILE_PATH (is_cold),
- local_brick->brick_name, i);
- i++;
- }
-out:
- return;
-}
-
-static int
-tier_prepare_compact (migration_args_t *args, gfdb_time_t current_time)
-{
- xlator_t *this = NULL;
- dht_conf_t *conf = NULL;
- gf_defrag_info_t *defrag = NULL;
- gf_tier_conf_t *tier_conf = NULL;
- gf_boolean_t is_hot_tier = args->is_hot_tier;
- int freq = 0;
- int ret = -1;
- const char *tier_type = is_hot_tier ? "hot" : "cold";
-
- this = args->this;
-
- conf = this->private;
-
- defrag = conf->defrag;
-
- tier_conf = &defrag->tier_conf;
-
- freq = is_hot_tier ? tier_get_freq_compact_hot (tier_conf) :
- tier_get_freq_compact_cold (tier_conf);
-
- defrag->tier_conf.compact_mode_switched = is_hot_tier ?
- defrag->tier_conf.compact_mode_switched_hot :
- defrag->tier_conf.compact_mode_switched_cold;
-
- gf_msg(this->name, GF_LOG_TRACE, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "Compact mode %i",
- defrag->tier_conf.compact_mode_switched);
-
- if (tier_check_compact (tier_conf, current_time,
- freq)) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "Start compaction on %s tier",
- tier_type);
-
- args->freq_time = freq;
- ret = tier_compact (args);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR, "Compaction failed on "
- "%s tier", tier_type);
- goto out;
- }
-
- gf_msg (this->name, GF_LOG_INFO, 0, DHT_MSG_LOG_TIER_STATUS,
- "End compaction on %s tier", tier_type);
-
- if (is_hot_tier) {
- defrag->tier_conf.compact_mode_switched_hot =
- _gf_false;
- } else {
- defrag->tier_conf.compact_mode_switched_cold =
- _gf_false;
- }
- }
-
-out:
- return ret;
-}
-
-static int
-tier_get_wm_interval(tier_mode_t mode, tier_watermark_op_t wm)
-{
- if (mode == TIER_MODE_WM && wm == TIER_WM_HI)
- return WM_INTERVAL_EMERG;
-
- return WM_INTERVAL;
-}
-
-/*
- * Main tiering loop. This is called from the promotion and the
- * demotion threads spawned in tier_start().
- *
- * Every second, wake from sleep to perform tasks.
- * 1. Check trigger to migrate data.
- * 2. Check for state changes (pause, unpause, stop).
- */
-static void
-*tier_run (void *in_args)
-{
- dht_conf_t *conf = NULL;
- gfdb_time_t current_time = { 0 };
- int freq = 0;
- int ret = 0;
- xlator_t *any = NULL;
- xlator_t *xlator = NULL;
- gf_tier_conf_t *tier_conf = NULL;
- loc_t root_loc = { 0 };
- int check_watermark = 0;
- gf_defrag_info_t *defrag = NULL;
- xlator_t *this = NULL;
- migration_args_t *args = in_args;
- GF_VALIDATE_OR_GOTO ("tier", args, out);
- GF_VALIDATE_OR_GOTO ("tier", args->brick_list, out);
-
- this = args->this;
- GF_VALIDATE_OR_GOTO ("tier", this, out);
-
- conf = this->private;
- GF_VALIDATE_OR_GOTO ("tier", conf, out);
-
- defrag = conf->defrag;
- GF_VALIDATE_OR_GOTO ("tier", defrag, out);
-
- if (list_empty (args->brick_list)) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Brick list for tier is empty. Exiting.");
- goto out;
- }
-
- defrag->defrag_status = GF_DEFRAG_STATUS_STARTED;
- tier_conf = &defrag->tier_conf;
-
- dht_build_root_loc (defrag->root_inode, &root_loc);
-
- while (1) {
-
- /*
- * Check if a graph switch occured. If so, stop migration
- * thread. It will need to be restarted manually.
- */
- any = THIS->ctx->active->first;
- xlator = xlator_search_by_name (any, this->name);
-
- if (xlator != this) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "Detected graph switch. Exiting migration "
- "daemon.");
- goto out;
- }
-
- gf_defrag_check_pause_tier (tier_conf);
-
- sleep(1);
-
- if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
- ret = 1;
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "defrag->defrag_status != "
- "GF_DEFRAG_STATUS_STARTED");
- goto out;
- }
-
- if (defrag->cmd == GF_DEFRAG_CMD_START_DETACH_TIER ||
- defrag->cmd == GF_DEFRAG_CMD_DETACH_START) {
- ret = 0;
- defrag->defrag_status =
- GF_DEFRAG_STATUS_COMPLETE;
- gf_msg (this->name, GF_LOG_DEBUG, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "defrag->defrag_cmd == "
- "GF_DEFRAG_CMD_START_DETACH_TIER");
- goto out;
- }
-
- if (gf_defrag_get_pause_state (&defrag->tier_conf) !=
- TIER_RUNNING)
- continue;
-
- /* To have proper synchronization amongst all
- * brick holding nodes, so that promotion and demotions
- * start atomicly w.r.t promotion/demotion frequency
- * period, all nodes should have thier system time
- * in-sync with each other either manually set or
- * using a NTP server*/
- ret = gettimeofday (&current_time, NULL);
- if (ret == -1) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- DHT_MSG_SYS_CALL_GET_TIME_FAILED,
- "Failed to get current time");
- goto out;
- }
-
- check_watermark++;
-
- /* emergency demotion requires frequent watermark monitoring */
- if (check_watermark >=
- tier_get_wm_interval(tier_conf->mode,
- tier_conf->watermark_last)) {
- check_watermark = 0;
- if (tier_conf->mode == TIER_MODE_WM) {
- ret = tier_get_fs_stat (this, &root_loc);
- if (ret != 0) {
- continue;
- }
- ret = tier_check_watermark (this);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_CRITICAL, errno,
- DHT_MSG_LOG_TIER_ERROR,
- "Failed to get watermark");
- continue;
- }
- }
- }
-
- if (args->is_promotion) {
- freq = tier_get_freq_promote (tier_conf);
-
- if (tier_check_promote (tier_conf, current_time, freq)) {
- args->freq_time = freq;
- ret = tier_promote (args);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Promotion failed");
- }
- }
- } else if (args->is_compaction) {
- tier_prepare_compact (args, current_time);
- } else {
- freq = tier_get_freq_demote (tier_conf);
-
- if (tier_check_demote (current_time, freq)) {
- args->freq_time = freq;
- ret = tier_demote (args);
- if (ret) {
- gf_msg (this->name,
- GF_LOG_ERROR,
- 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Demotion failed");
- }
- }
- }
-
- /* Check the statfs immediately after the processing threads
- return */
- check_watermark = WM_INTERVAL;
- }
-
- ret = 0;
-out:
-
- args->return_value = ret;
-
- return NULL;
-}
-
-int
-tier_start (xlator_t *this, gf_defrag_info_t *defrag)
-{
- pthread_t promote_thread;
- pthread_t demote_thread;
- pthread_t hot_compact_thread;
- pthread_t cold_compact_thread;
- int ret = -1;
- struct list_head bricklist_hot = { 0 };
- struct list_head bricklist_cold = { 0 };
- migration_args_t promotion_args = { 0 };
- migration_args_t demotion_args = { 0 };
- migration_args_t hot_compaction_args = { 0 };
- migration_args_t cold_compaction_args = { 0 };
- dht_conf_t *conf = NULL;
-
- INIT_LIST_HEAD ((&bricklist_hot));
- INIT_LIST_HEAD ((&bricklist_cold));
-
- conf = this->private;
-
- tier_get_bricklist (conf->subvolumes[1], &bricklist_hot);
- set_brick_list_qpath (&bricklist_hot, _gf_false);
-
- demotion_args.this = this;
- demotion_args.brick_list = &bricklist_hot;
- demotion_args.defrag = defrag;
- demotion_args.is_promotion = _gf_false;
- demotion_args.is_compaction = _gf_false;
-
- ret = gf_thread_create (&demote_thread,
- NULL, &tier_run, &demotion_args, "tierdem");
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Failed to start demotion thread.");
- defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
- goto cleanup;
- }
-
- tier_get_bricklist (conf->subvolumes[0], &bricklist_cold);
- set_brick_list_qpath (&bricklist_cold, _gf_true);
-
- promotion_args.this = this;
- promotion_args.brick_list = &bricklist_cold;
- promotion_args.defrag = defrag;
- promotion_args.is_promotion = _gf_true;
-
- ret = gf_thread_create (&promote_thread, NULL, &tier_run,
- &promotion_args, "tierpro");
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Failed to start promotion thread.");
- defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
- goto waitforspawned;
- }
-
- hot_compaction_args.this = this;
- hot_compaction_args.brick_list = &bricklist_hot;
- hot_compaction_args.defrag = defrag;
- hot_compaction_args.is_promotion = _gf_false;
- hot_compaction_args.is_compaction = _gf_true;
- hot_compaction_args.is_hot_tier = _gf_true;
-
- ret = gf_thread_create (&hot_compact_thread, NULL, &tier_run,
- &hot_compaction_args, "tierhcom");
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Failed to start compaction thread.");
- defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
- goto waitforspawnedpromote;
- }
-
- cold_compaction_args.this = this;
- cold_compaction_args.brick_list = &bricklist_cold;
- cold_compaction_args.defrag = defrag;
- cold_compaction_args.is_promotion = _gf_false;
- cold_compaction_args.is_compaction = _gf_true;
- cold_compaction_args.is_hot_tier = _gf_false;
-
- ret = gf_thread_create (&cold_compact_thread, NULL, &tier_run,
- &cold_compaction_args, "tierccom");
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Failed to start compaction thread.");
- defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
- goto waitforspawnedhotcompact;
- }
- pthread_join (cold_compact_thread, NULL);
-
-waitforspawnedhotcompact:
- pthread_join (hot_compact_thread, NULL);
-
-waitforspawnedpromote:
- pthread_join (promote_thread, NULL);
-
-waitforspawned:
- pthread_join (demote_thread, NULL);
-
-cleanup:
- clear_bricklist (&bricklist_cold);
- clear_bricklist (&bricklist_hot);
- return ret;
-}
-
-int32_t
-tier_migration_needed (xlator_t *this)
-{
- gf_defrag_info_t *defrag = NULL;
- dht_conf_t *conf = NULL;
- int ret = 0;
-
- conf = this->private;
-
- GF_VALIDATE_OR_GOTO (this->name, conf, out);
- GF_VALIDATE_OR_GOTO (this->name, conf->defrag, out);
-
- defrag = conf->defrag;
-
- if ((defrag->cmd == GF_DEFRAG_CMD_START_TIER) ||
- (defrag->cmd == GF_DEFRAG_CMD_START_DETACH_TIER))
- ret = 1;
-out:
- return ret;
-}
-
-int32_t
-tier_migration_get_dst (xlator_t *this, dht_local_t *local)
-{
- dht_conf_t *conf = NULL;
- int32_t ret = -1;
- gf_defrag_info_t *defrag = NULL;
-
- GF_VALIDATE_OR_GOTO ("tier", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
-
- conf = this->private;
-
- defrag = conf->defrag;
-
- if (defrag && defrag->cmd == GF_DEFRAG_CMD_START_DETACH_TIER) {
- local->rebalance.target_node = conf->subvolumes[0];
-
- } else if (conf->subvolumes[0] == local->cached_subvol)
- local->rebalance.target_node =
- conf->subvolumes[1];
- else
- local->rebalance.target_node =
- conf->subvolumes[0];
-
- if (local->rebalance.target_node)
- ret = 0;
-
-out:
- return ret;
-}
-
-xlator_t *
-tier_search (xlator_t *this, dht_layout_t *layout, const char *name)
-{
- xlator_t *subvol = NULL;
- dht_conf_t *conf = NULL;
-
- GF_VALIDATE_OR_GOTO ("tier", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
-
- conf = this->private;
-
- subvol = TIER_HASHED_SUBVOL;
-
- out:
- return subvol;
-}
-
-
-static int
-tier_load_externals (xlator_t *this)
-{
- int ret = -1;
- char *libpathfull = (LIBDIR "/libgfdb.so.0");
- get_gfdb_methods_t get_gfdb_methods;
-
- GF_VALIDATE_OR_GOTO ("this", this, out);
-
- libhandle = dlopen (libpathfull, RTLD_NOW);
- if (!libhandle) {
- gf_msg(this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Error loading libgfdb.so %s\n", dlerror());
- ret = -1;
- goto out;
- }
-
- get_gfdb_methods = dlsym (libhandle, "get_gfdb_methods");
- if (!get_gfdb_methods) {
- gf_msg(this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Error loading get_gfdb_methods()");
- ret = -1;
- goto out;
- }
-
- get_gfdb_methods (&gfdb_methods);
-
- ret = 0;
-
-out:
- if (ret && libhandle)
- dlclose (libhandle);
-
- return ret;
-}
-
-static tier_mode_t
-tier_validate_mode (char *mode)
-{
- int ret = -1;
-
- if (strcmp (mode, "test") == 0) {
- ret = TIER_MODE_TEST;
- } else {
- ret = TIER_MODE_WM;
- }
-
- return ret;
-}
-
-static gf_boolean_t
-tier_validate_compact_mode (char *mode)
-{
- gf_boolean_t ret = _gf_false;
-
- gf_msg ("tier", GF_LOG_INFO, 0, DHT_MSG_LOG_TIER_STATUS,
- "tier_validate_compact_mode: mode = %s", mode);
-
- if (!strcmp (mode, "on")) {
- ret = _gf_true;
- } else {
- ret = _gf_false;
- }
-
- gf_msg ("tier", GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "tier_validate_compact_mode: ret = %i", ret);
-
- return ret;
-}
-
-int
-tier_init_methods (xlator_t *this)
-{
- int ret = -1;
- dht_conf_t *conf = NULL;
- dht_methods_t *methods = NULL;
-
- GF_VALIDATE_OR_GOTO ("tier", this, err);
-
- conf = this->private;
-
- methods = &(conf->methods);
-
- methods->migration_get_dst_subvol = tier_migration_get_dst;
- methods->migration_other = tier_start;
- methods->migration_needed = tier_migration_needed;
- methods->layout_search = tier_search;
-
- ret = 0;
-err:
- return ret;
-}
-
-
-static void
-tier_save_vol_name (xlator_t *this)
-{
- dht_conf_t *conf = NULL;
- gf_defrag_info_t *defrag = NULL;
- char *suffix = NULL;
- int name_len = 0;
-
-
- conf = this->private;
- defrag = conf->defrag;
-
- suffix = strstr (this->name, "-tier-dht");
-
- if (suffix)
- name_len = suffix - this->name;
- else
- name_len = strlen (this->name);
-
- if (name_len > GD_VOLUME_NAME_MAX)
- name_len = GD_VOLUME_NAME_MAX;
-
- strncpy (defrag->tier_conf.volname, this->name, name_len);
- defrag->tier_conf.volname[name_len] = 0;
-}
-
-int
-tier_init (xlator_t *this)
-{
- int ret = -1;
- int freq = 0;
- int maxsize = 0;
- dht_conf_t *conf = NULL;
- gf_defrag_info_t *defrag = NULL;
- char *voldir = NULL;
- char *mode = NULL;
- char *paused = NULL;
- tier_mode_t tier_mode = DEFAULT_TIER_MODE;
- gf_boolean_t compact_mode = _gf_false;
-
- ret = dht_init (this);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "tier_init failed");
- goto out;
- }
-
- conf = this->private;
-
- ret = tier_init_methods (this);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "tier_init_methods failed");
- goto out;
- }
-
- if (conf->subvolume_cnt != 2) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Invalid number of subvolumes %d", conf->subvolume_cnt);
- goto out;
- }
-
- /* if instatiated from client side initialization is complete. */
- if (!conf->defrag) {
- ret = 0;
- goto out;
- }
-
- /* if instatiated from server side, load db libraries */
- ret = tier_load_externals (this);
- if (ret) {
- gf_msg(this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "Could not load externals. Aborting");
- goto out;
- }
-
- defrag = conf->defrag;
-
- defrag->tier_conf.last_demote_qfile_index = 0;
- defrag->tier_conf.last_promote_qfile_index = 0;
-
- defrag->tier_conf.is_tier = 1;
- defrag->this = this;
-
- ret = dict_get_int32 (this->options,
- "tier-max-promote-file-size", &maxsize);
- if (ret) {
- maxsize = 0;
- }
-
- defrag->tier_conf.tier_max_promote_size = maxsize;
-
- ret = dict_get_int32 (this->options,
- "tier-promote-frequency", &freq);
- if (ret) {
- freq = DEFAULT_PROMOTE_FREQ_SEC;
- }
-
- defrag->tier_conf.tier_promote_frequency = freq;
-
- ret = dict_get_int32 (this->options,
- "tier-demote-frequency", &freq);
- if (ret) {
- freq = DEFAULT_DEMOTE_FREQ_SEC;
- }
-
- defrag->tier_conf.tier_demote_frequency = freq;
-
- ret = dict_get_int32 (this->options,
- "tier-hot-compact-frequency", &freq);
- if (ret) {
- freq = DEFAULT_HOT_COMPACT_FREQ_SEC;
- }
-
- defrag->tier_conf.tier_compact_hot_frequency = freq;
-
- ret = dict_get_int32 (this->options,
- "tier-cold-compact-frequency", &freq);
- if (ret) {
- freq = DEFAULT_COLD_COMPACT_FREQ_SEC;
- }
-
- defrag->tier_conf.tier_compact_cold_frequency = freq;
-
- ret = dict_get_int32 (this->options,
- "watermark-hi", &freq);
- if (ret) {
- freq = DEFAULT_WM_HI;
- }
-
- defrag->tier_conf.watermark_hi = freq;
-
- ret = dict_get_int32 (this->options,
- "watermark-low", &freq);
- if (ret) {
- freq = DEFAULT_WM_LOW;
- }
-
- defrag->tier_conf.watermark_low = freq;
-
- ret = dict_get_int32 (this->options,
- "write-freq-threshold", &freq);
- if (ret) {
- freq = DEFAULT_WRITE_FREQ_SEC;
- }
-
- defrag->write_freq_threshold = freq;
-
- ret = dict_get_int32 (this->options,
- "read-freq-threshold", &freq);
- if (ret) {
- freq = DEFAULT_READ_FREQ_SEC;
- }
-
- defrag->read_freq_threshold = freq;
-
- ret = dict_get_int32 (this->options,
- "tier-max-mb", &freq);
- if (ret) {
- freq = DEFAULT_TIER_MAX_MIGRATE_MB;
- }
-
- defrag->tier_conf.max_migrate_bytes = (uint64_t) freq * 1024 * 1024;
-
- ret = dict_get_int32 (this->options,
- "tier-max-files", &freq);
- if (ret) {
- freq = DEFAULT_TIER_MAX_MIGRATE_FILES;
- }
-
- defrag->tier_conf.max_migrate_files = freq;
-
-
- ret = dict_get_int32 (this->options,
- "tier-query-limit",
- &(defrag->tier_conf.query_limit));
- if (ret) {
- defrag->tier_conf.query_limit =
- DEFAULT_TIER_QUERY_LIMIT;
- }
-
- ret = dict_get_str (this->options,
- "tier-compact", &mode);
-
- if (ret) {
- defrag->tier_conf.compact_active = DEFAULT_COMP_MODE;
- } else {
- compact_mode = tier_validate_compact_mode (mode);
- /* If compaction is now active, we need to inform the bricks on
- the hot and cold tier of this. See dht-common.h for more. */
- defrag->tier_conf.compact_active = compact_mode;
- if (compact_mode) {
- defrag->tier_conf.compact_mode_switched_hot = _gf_true;
- defrag->tier_conf.compact_mode_switched_cold = _gf_true;
- }
- }
-
- ret = dict_get_str (this->options,
- "tier-mode", &mode);
- if (ret) {
- defrag->tier_conf.mode = DEFAULT_TIER_MODE;
- } else {
- tier_mode = tier_validate_mode (mode);
- defrag->tier_conf.mode = tier_mode;
- }
-
- pthread_mutex_init (&defrag->tier_conf.pause_mutex, 0);
-
- gf_defrag_set_pause_state (&defrag->tier_conf, TIER_RUNNING);
-
- ret = dict_get_str (this->options,
- "tier-pause", &paused);
-
- if (paused && strcmp (paused, "on") == 0)
- gf_defrag_set_pause_state (&defrag->tier_conf,
- TIER_REQUEST_PAUSE);
-
- ret = gf_asprintf(&voldir, "%s/%s",
- DEFAULT_VAR_RUN_DIRECTORY,
- this->name);
- if (ret < 0)
- goto out;
-
- ret = mkdir_p(voldir, 0777, _gf_true);
- if (ret == -1 && errno != EEXIST) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "tier_init failed");
-
- GF_FREE(voldir);
- goto out;
- }
-
- GF_FREE(voldir);
-
- ret = gf_asprintf (&promotion_qfile, "%s/%s/promote",
- DEFAULT_VAR_RUN_DIRECTORY,
- this->name);
- if (ret < 0)
- goto out;
-
- ret = gf_asprintf (&demotion_qfile, "%s/%s/demote",
- DEFAULT_VAR_RUN_DIRECTORY,
- this->name);
- if (ret < 0) {
- GF_FREE (promotion_qfile);
- goto out;
- }
-
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "Promote/demote frequency %d/%d "
- "Write/Read freq thresholds %d/%d",
- defrag->tier_conf.tier_promote_frequency,
- defrag->tier_conf.tier_demote_frequency,
- defrag->write_freq_threshold,
- defrag->read_freq_threshold);
-
- tier_save_vol_name (this);
-
- ret = 0;
-
-out:
-
- return ret;
-}
-
-int
-tier_cli_pause_done (int op_ret, call_frame_t *sync_frame, void *data)
-{
- gf_msg ("tier", GF_LOG_INFO, 0,
- DHT_MSG_TIER_PAUSED,
- "Migrate file paused with op_ret %d", op_ret);
-
- return op_ret;
-}
-
-int
-tier_cli_pause (void *data)
-{
- gf_defrag_info_t *defrag = NULL;
- xlator_t *this = NULL;
- dht_conf_t *conf = NULL;
- int ret = -1;
-
- this = data;
-
- conf = this->private;
- GF_VALIDATE_OR_GOTO (this->name, conf, exit);
-
- defrag = conf->defrag;
- GF_VALIDATE_OR_GOTO (this->name, defrag, exit);
-
- gf_defrag_pause_tier (this, defrag);
-
- ret = 0;
-exit:
- return ret;
-}
-
-int
-tier_reconfigure (xlator_t *this, dict_t *options)
-{
- dht_conf_t *conf = NULL;
- gf_defrag_info_t *defrag = NULL;
- char *mode = NULL;
- int migrate_mb = 0;
- gf_boolean_t req_pause = _gf_false;
- int ret = 0;
- call_frame_t *frame = NULL;
- gf_boolean_t last_compact_setting = _gf_false;
-
- conf = this->private;
-
- if (conf->defrag) {
- defrag = conf->defrag;
- GF_OPTION_RECONF ("tier-max-promote-file-size",
- defrag->tier_conf.tier_max_promote_size,
- options, int32, out);
-
- GF_OPTION_RECONF ("tier-promote-frequency",
- defrag->tier_conf.tier_promote_frequency,
- options, int32, out);
-
- GF_OPTION_RECONF ("tier-demote-frequency",
- defrag->tier_conf.tier_demote_frequency,
- options, int32, out);
-
- GF_OPTION_RECONF ("write-freq-threshold",
- defrag->write_freq_threshold, options,
- int32, out);
-
- GF_OPTION_RECONF ("read-freq-threshold",
- defrag->read_freq_threshold, options,
- int32, out);
-
- GF_OPTION_RECONF ("watermark-hi",
- defrag->tier_conf.watermark_hi, options,
- int32, out);
-
- GF_OPTION_RECONF ("watermark-low",
- defrag->tier_conf.watermark_low, options,
- int32, out);
-
- last_compact_setting = defrag->tier_conf.compact_active;
-
- GF_OPTION_RECONF ("tier-compact",
- defrag->tier_conf.compact_active, options,
- bool, out);
-
- if (last_compact_setting != defrag->tier_conf.compact_active) {
- defrag->tier_conf.compact_mode_switched_hot = _gf_true;
- defrag->tier_conf.compact_mode_switched_cold = _gf_true;
- gf_msg (this->name, GF_LOG_INFO, 0,
- DHT_MSG_LOG_TIER_STATUS,
- "compact mode switched");
- }
-
- GF_OPTION_RECONF ("tier-hot-compact-frequency",
- defrag->tier_conf.tier_compact_hot_frequency,
- options, int32, out);
-
- GF_OPTION_RECONF ("tier-cold-compact-frequency",
- defrag->tier_conf.tier_compact_cold_frequency,
- options, int32, out);
-
- GF_OPTION_RECONF ("tier-mode",
- mode, options,
- str, out);
- defrag->tier_conf.mode = tier_validate_mode (mode);
-
- GF_OPTION_RECONF ("tier-max-mb",
- migrate_mb, options,
- int32, out);
- defrag->tier_conf.max_migrate_bytes = (uint64_t) migrate_mb *
- 1024 * 1024;
-
- GF_OPTION_RECONF ("tier-max-files",
- defrag->tier_conf.max_migrate_files, options,
- int32, out);
-
- GF_OPTION_RECONF ("tier-query-limit",
- defrag->tier_conf.query_limit,
- options, int32, out);
-
- GF_OPTION_RECONF ("tier-pause",
- req_pause, options,
- bool, out);
-
- if (req_pause == _gf_true) {
-
- frame = create_frame (this, this->ctx->pool);
- if (!frame)
- goto out;
-
- frame->root->pid = GF_CLIENT_PID_DEFRAG;
-
- ret = synctask_new (this->ctx->env, tier_cli_pause,
- tier_cli_pause_done, frame, this);
-
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "pause tier failed on reconfigure");
- }
- } else {
- ret = gf_defrag_resume_tier (this, defrag);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- DHT_MSG_LOG_TIER_ERROR,
- "resume tier failed on reconfigure");
- }
- }
-
- }
-
-out:
- return dht_reconfigure (this, options);
-}
-
-void
-tier_fini (xlator_t *this)
-{
- if (libhandle)
- dlclose (libhandle);
-
- GF_FREE (demotion_qfile);
- GF_FREE (promotion_qfile);
-
- dht_fini(this);
-}
-
-class_methods_t class_methods = {
- .init = tier_init,
- .fini = tier_fini,
- .reconfigure = tier_reconfigure,
- .notify = dht_notify
-};
-
-struct xlator_fops fops = {
-
- .lookup = dht_lookup,
- .create = tier_create,
- .mknod = dht_mknod,
-
- .open = dht_open,
- .statfs = tier_statfs,
- .opendir = dht_opendir,
- .readdir = tier_readdir,
- .readdirp = tier_readdirp,
- .fsyncdir = dht_fsyncdir,
- .symlink = dht_symlink,
- .unlink = tier_unlink,
- .link = tier_link,
- .mkdir = dht_mkdir,
- .rmdir = dht_rmdir,
- .rename = dht_rename,
- .entrylk = dht_entrylk,
- .fentrylk = dht_fentrylk,
-
- /* Inode read operations */
- .stat = dht_stat,
- .fstat = dht_fstat,
- .access = dht_access,
- .readlink = dht_readlink,
- .getxattr = dht_getxattr,
- .fgetxattr = dht_fgetxattr,
- .readv = dht_readv,
- .flush = dht_flush,
- .fsync = dht_fsync,
- .inodelk = dht_inodelk,
- .finodelk = dht_finodelk,
- .lk = dht_lk,
-
- /* Inode write operations */
- .fremovexattr = dht_fremovexattr,
- .removexattr = dht_removexattr,
- .setxattr = dht_setxattr,
- .fsetxattr = dht_fsetxattr,
- .truncate = dht_truncate,
- .ftruncate = dht_ftruncate,
- .writev = dht_writev,
- .xattrop = dht_xattrop,
- .fxattrop = dht_fxattrop,
- .setattr = dht_setattr,
- .fsetattr = dht_fsetattr,
- .fallocate = dht_fallocate,
- .discard = dht_discard,
- .zerofill = dht_zerofill,
-};
-
-struct xlator_cbks cbks = {
- .release = dht_release,
- .forget = dht_forget
-};
diff --git a/xlators/cluster/dht/src/tier.h b/xlators/cluster/dht/src/tier.h
deleted file mode 100644
index 764860e6884..00000000000
--- a/xlators/cluster/dht/src/tier.h
+++ /dev/null
@@ -1,112 +0,0 @@
-/*
- Copyright (c) 2015 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#ifndef _TIER_H_
-#define _TIER_H_
-
-
-/******************************************************************************/
-/* This is from dht-rebalancer.c as we dont have dht-rebalancer.h */
-#include "dht-common.h"
-#include "xlator.h"
-#include <signal.h>
-#include <fnmatch.h>
-#include <signal.h>
-
-/*
- * Size of timer wheel. We would not promote or demote less
- * frequently than this number.
- */
-#define TIMER_SECS 3600
-
-#include "gfdb_data_store.h"
-#include <ctype.h>
-#include <sys/stat.h>
-
-#define PROMOTION_QFILE "promotequeryfile"
-#define DEMOTION_QFILE "demotequeryfile"
-
-#define TIER_HASHED_SUBVOL conf->subvolumes[0]
-#define TIER_UNHASHED_SUBVOL conf->subvolumes[1]
-
-#define GET_QFILE_PATH(is_promotion)\
- (is_promotion) ? promotion_qfile : demotion_qfile
-
-typedef struct tier_qfile_array {
- int *fd_array;
- ssize_t array_size;
- ssize_t next_index;
- /* Indicate the number of exhuasted FDs*/
- ssize_t exhausted_count;
-} tier_qfile_array_t;
-
-
-typedef struct _query_cbk_args {
- xlator_t *this;
- gf_defrag_info_t *defrag;
- /* This is write */
- int query_fd;
- int is_promotion;
- int is_compaction;
- /* This is for read */
- tier_qfile_array_t *qfile_array;
-} query_cbk_args_t;
-
-int
-gf_run_tier(xlator_t *this, gf_defrag_info_t *defrag);
-
-typedef struct gfdb_brick_info {
- gfdb_time_t *time_stamp;
- gf_boolean_t _gfdb_promote;
- query_cbk_args_t *_query_cbk_args;
-} gfdb_brick_info_t;
-
-typedef struct brick_list {
- xlator_t *xlator;
- char *brick_db_path;
- char brick_name[NAME_MAX];
- char qfile_path[PATH_MAX];
- struct list_head list;
-} tier_brick_list_t;
-
-typedef struct _dm_thread_args {
- xlator_t *this;
- gf_defrag_info_t *defrag;
- struct list_head *brick_list;
- int freq_time;
- int return_value;
- int is_promotion;
- int is_compaction;
- gf_boolean_t is_hot_tier;
-} migration_args_t;
-
-typedef enum tier_watermark_op_ {
- TIER_WM_NONE = 0,
- TIER_WM_LOW,
- TIER_WM_HI,
- TIER_WM_MID
-} tier_watermark_op_t;
-
-#define DEFAULT_PROMOTE_FREQ_SEC 120
-#define DEFAULT_DEMOTE_FREQ_SEC 120
-#define DEFAULT_HOT_COMPACT_FREQ_SEC 604800
-#define DEFAULT_COLD_COMPACT_FREQ_SEC 604800
-#define DEFAULT_DEMOTE_DEGRADED 1
-#define DEFAULT_WRITE_FREQ_SEC 0
-#define DEFAULT_READ_FREQ_SEC 0
-#define DEFAULT_WM_LOW 75
-#define DEFAULT_WM_HI 90
-#define DEFAULT_TIER_MODE TIER_MODE_TEST
-#define DEFAULT_COMP_MODE _gf_true
-#define DEFAULT_TIER_MAX_MIGRATE_MB 1000
-#define DEFAULT_TIER_MAX_MIGRATE_FILES 5000
-#define DEFAULT_TIER_QUERY_LIMIT 100
-
-#endif
diff --git a/xlators/cluster/dht/src/tier.sym b/xlators/cluster/dht/src/tier.sym
deleted file mode 100644
index 60205d145b6..00000000000
--- a/xlators/cluster/dht/src/tier.sym
+++ /dev/null
@@ -1,9 +0,0 @@
-fops
-cbks
-class_methods
-dht_methods
-tier_methods
-options
-mem_acct_init
-reconfigure
-dumpops
diff --git a/xlators/cluster/dht/src/unittest/dht_layout_mock.c b/xlators/cluster/dht/src/unittest/dht_layout_mock.c
index 6544f4208f5..771452963d1 100644
--- a/xlators/cluster/dht/src/unittest/dht_layout_mock.c
+++ b/xlators/cluster/dht/src/unittest/dht_layout_mock.c
@@ -7,66 +7,67 @@
later), or the GNU General Public License, version 2 (GPLv2), in all
cases as published by the Free Software Foundation.
*/
-#include "glusterfs.h"
-#include "xlator.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/xlator.h>
#include "dht-common.h"
-#include "byte-order.h"
+#include <glusterfs/byte-order.h>
int
-dht_hash_compute (xlator_t *this, int type, const char *name, uint32_t *hash_p)
+dht_hash_compute(xlator_t *this, int type, const char *name, uint32_t *hash_p)
{
return 0;
}
int
-dht_inode_ctx_layout_get (inode_t *inode, xlator_t *this, dht_layout_t **layout)
+dht_inode_ctx_layout_get(inode_t *inode, xlator_t *this, dht_layout_t **layout)
{
return 0;
}
int
-dht_inode_ctx_layout_set (inode_t *inode, xlator_t *this,
- dht_layout_t *layout_int)
+dht_inode_ctx_layout_set(inode_t *inode, xlator_t *this,
+ dht_layout_t *layout_int)
{
return 0;
}
int
-dict_get_ptr (dict_t *this, char *key, void **ptr)
+dict_get_ptr(dict_t *this, char *key, void **ptr)
{
return 0;
}
int
-dict_get_ptr_and_len (dict_t *this, char *key, void **ptr, int *len)
+dict_get_ptr_and_len(dict_t *this, char *key, void **ptr, int *len)
{
return 0;
}
-int _gf_log (const char *domain, const char *file,
- const char *function, int32_t line, gf_loglevel_t level,
- const char *fmt, ...)
+int
+_gf_log(const char *domain, const char *file, const char *function,
+ int32_t line, gf_loglevel_t level, const char *fmt, ...)
{
return 0;
}
-int _gf_log_callingfn (const char *domain, const char *file,
- const char *function, int32_t line, gf_loglevel_t level,
- const char *fmt, ...)
+int
+_gf_log_callingfn(const char *domain, const char *file, const char *function,
+ int32_t line, gf_loglevel_t level, const char *fmt, ...)
{
return 0;
}
-void gf_uuid_unparse(const uuid_t uu, char *out)
+void
+gf_uuid_unparse(const uuid_t uu, char *out)
{
// could call a will-return function here
// to place the correct data in *out
}
int
-_gf_msg (const char *domain, const char *file, const char *function,
- int32_t line, gf_loglevel_t level, int errnum, int trace,
- uint64_t msgid, const char *fmt, ...)
+_gf_msg(const char *domain, const char *file, const char *function,
+ int32_t line, gf_loglevel_t level, int errnum, int trace,
+ uint64_t msgid, const char *fmt, ...)
{
return 0;
}
diff --git a/xlators/cluster/dht/src/unittest/dht_layout_unittest.c b/xlators/cluster/dht/src/unittest/dht_layout_unittest.c
index fb6445ce1f3..c94a1d0a2e1 100644
--- a/xlators/cluster/dht/src/unittest/dht_layout_unittest.c
+++ b/xlators/cluster/dht/src/unittest/dht_layout_unittest.c
@@ -9,8 +9,8 @@
*/
#include "dht-common.h"
-#include "logging.h"
-#include "xlator.h"
+#include <glusterfs/logging.h>
+#include <glusterfs/xlator.h>
#include <inttypes.h>
#include <stdarg.h>
@@ -34,16 +34,16 @@ helper_xlator_init(uint32_t num_types)
xl = test_calloc(1, sizeof(xlator_t));
assert_non_null(xl);
xl->mem_acct->num_types = num_types;
- xl->mem_acct = test_calloc (sizeof(struct mem_acct)
- + sizeof(struct mem_acct_rec) + num_types);
+ xl->mem_acct = test_calloc(sizeof(struct mem_acct) +
+ sizeof(struct mem_acct_rec) + num_types);
assert_non_null(xl->mem_acct);
xl->ctx = test_calloc(1, sizeof(glusterfs_ctx_t));
assert_non_null(xl->ctx);
for (i = 0; i < num_types; i++) {
- ret = LOCK_INIT(&(xl->mem_acct.rec[i].lock));
- assert_false(ret);
+ ret = LOCK_INIT(&(xl->mem_acct.rec[i].lock));
+ assert_false(ret);
}
ENSURE(num_types == xl->mem_acct.num_types);
@@ -58,8 +58,8 @@ helper_xlator_destroy(xlator_t *xl)
int i, ret;
for (i = 0; i < xl->mem_acct.num_types; i++) {
- ret = LOCK_DESTROY(&(xl->mem_acct.rec[i].lock));
- assert_int_equal(ret, 0);
+ ret = LOCK_DESTROY(&(xl->mem_acct.rec[i].lock));
+ assert_int_equal(ret, 0);
}
free(xl->mem_acct.rec);
@@ -76,7 +76,7 @@ test_dht_layout_new(void **state)
{
xlator_t *xl;
dht_layout_t *layout;
- dht_conf_t *conf;
+ dht_conf_t *conf;
int cnt;
expect_assert_failure(dht_layout_new(NULL, 0));
@@ -90,7 +90,7 @@ test_dht_layout_new(void **state)
assert_non_null(layout);
assert_int_equal(layout->type, DHT_HASH_TYPE_DM);
assert_int_equal(layout->cnt, cnt);
- assert_int_equal(GF_ATOMIC_GET (layout->ref), 1);
+ assert_int_equal(GF_ATOMIC_GET(layout->ref), 1);
assert_int_equal(layout->gen, 0);
assert_int_equal(layout->spread_cnt, 0);
free(layout);
@@ -107,7 +107,7 @@ test_dht_layout_new(void **state)
assert_non_null(layout);
assert_int_equal(layout->type, DHT_HASH_TYPE_DM);
assert_int_equal(layout->cnt, cnt);
- assert_int_equal(GF_ATOMIC_GET (layout->ref), 1);
+ assert_int_equal(GF_ATOMIC_GET(layout->ref), 1);
assert_int_equal(layout->gen, conf->gen);
assert_int_equal(layout->spread_cnt, conf->dir_spread_cnt);
free(layout);
@@ -116,7 +116,9 @@ test_dht_layout_new(void **state)
helper_xlator_destroy(xl);
}
-int main(void) {
+int
+main(void)
+{
const struct CMUnitTest xlator_dht_layout_tests[] = {
unit_test(test_dht_layout_new),
};
diff --git a/xlators/cluster/ec/src/ec-code-avx.c b/xlators/cluster/ec/src/ec-code-avx.c
index 92bd3e83c5e..70afaa00f54 100644
--- a/xlators/cluster/ec/src/ec-code-avx.c
+++ b/xlators/cluster/ec/src/ec-code-avx.c
@@ -34,10 +34,9 @@ ec_code_avx_load(ec_code_builder_t *builder, uint32_t dst, uint32_t idx,
uint32_t bit)
{
if (builder->linear) {
- ec_code_intel_op_mov_m2avx(builder, REG_SI, REG_DX, 1,
- idx * builder->width * builder->bits +
- bit * builder->width,
- dst);
+ ec_code_intel_op_mov_m2avx(
+ builder, REG_SI, REG_DX, 1,
+ idx * builder->width * builder->bits + bit * builder->width, dst);
} else {
if (builder->base != idx) {
ec_code_intel_op_mov_m2r(builder, REG_SI, REG_NULL, 0, idx * 8,
@@ -81,10 +80,9 @@ ec_code_avx_xorm(ec_code_builder_t *builder, uint32_t dst, uint32_t idx,
uint32_t bit)
{
if (builder->linear) {
- ec_code_intel_op_xor_m2avx(builder, REG_SI, REG_DX, 1,
- idx * builder->width * builder->bits +
- bit * builder->width,
- dst);
+ ec_code_intel_op_xor_m2avx(
+ builder, REG_SI, REG_DX, 1,
+ idx * builder->width * builder->bits + bit * builder->width, dst);
} else {
if (builder->base != idx) {
ec_code_intel_op_mov_m2r(builder, REG_SI, REG_NULL, 0, idx * 8,
@@ -96,21 +94,16 @@ ec_code_avx_xorm(ec_code_builder_t *builder, uint32_t dst, uint32_t idx,
}
}
-static char *ec_code_avx_needed_flags[] = {
- "avx2",
- NULL
-};
+static char *ec_code_avx_needed_flags[] = {"avx2", NULL};
-ec_code_gen_t ec_code_gen_avx = {
- .name = "avx",
- .flags = ec_code_avx_needed_flags,
- .width = 32,
- .prolog = ec_code_avx_prolog,
- .epilog = ec_code_avx_epilog,
- .load = ec_code_avx_load,
- .store = ec_code_avx_store,
- .copy = ec_code_avx_copy,
- .xor2 = ec_code_avx_xor2,
- .xor3 = ec_code_avx_xor3,
- .xorm = ec_code_avx_xorm
-};
+ec_code_gen_t ec_code_gen_avx = {.name = "avx",
+ .flags = ec_code_avx_needed_flags,
+ .width = 32,
+ .prolog = ec_code_avx_prolog,
+ .epilog = ec_code_avx_epilog,
+ .load = ec_code_avx_load,
+ .store = ec_code_avx_store,
+ .copy = ec_code_avx_copy,
+ .xor2 = ec_code_avx_xor2,
+ .xor3 = ec_code_avx_xor3,
+ .xorm = ec_code_avx_xorm};
diff --git a/xlators/cluster/ec/src/ec-code-c.c b/xlators/cluster/ec/src/ec-code-c.c
index 7387f3ea435..acdc665c2cf 100644
--- a/xlators/cluster/ec/src/ec-code-c.c
+++ b/xlators/cluster/ec/src/ec-code-c.c
@@ -16,12 +16,14 @@
#define WIDTH (EC_METHOD_WORD_SIZE / sizeof(uint64_t))
-static void gf8_muladd_00(void *out, void *in)
+static void
+gf8_muladd_00(void *out, void *in)
{
memcpy(out, in, EC_METHOD_WORD_SIZE * 8);
}
-static void gf8_muladd_01(void *out, void *in)
+static void
+gf8_muladd_01(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -42,7 +44,8 @@ static void gf8_muladd_01(void *out, void *in)
}
}
-static void gf8_muladd_02(void *out, void *in)
+static void
+gf8_muladd_02(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -83,7 +86,8 @@ static void gf8_muladd_02(void *out, void *in)
}
}
-static void gf8_muladd_03(void *out, void *in)
+static void
+gf8_muladd_03(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -126,7 +130,8 @@ static void gf8_muladd_03(void *out, void *in)
}
}
-static void gf8_muladd_04(void *out, void *in)
+static void
+gf8_muladd_04(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -169,7 +174,8 @@ static void gf8_muladd_04(void *out, void *in)
}
}
-static void gf8_muladd_05(void *out, void *in)
+static void
+gf8_muladd_05(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -210,7 +216,8 @@ static void gf8_muladd_05(void *out, void *in)
}
}
-static void gf8_muladd_06(void *out, void *in)
+static void
+gf8_muladd_06(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -253,7 +260,8 @@ static void gf8_muladd_06(void *out, void *in)
}
}
-static void gf8_muladd_07(void *out, void *in)
+static void
+gf8_muladd_07(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -299,7 +307,8 @@ static void gf8_muladd_07(void *out, void *in)
}
}
-static void gf8_muladd_08(void *out, void *in)
+static void
+gf8_muladd_08(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -340,7 +349,8 @@ static void gf8_muladd_08(void *out, void *in)
}
}
-static void gf8_muladd_09(void *out, void *in)
+static void
+gf8_muladd_09(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -383,7 +393,8 @@ static void gf8_muladd_09(void *out, void *in)
}
}
-static void gf8_muladd_0A(void *out, void *in)
+static void
+gf8_muladd_0A(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -424,7 +435,8 @@ static void gf8_muladd_0A(void *out, void *in)
}
}
-static void gf8_muladd_0B(void *out, void *in)
+static void
+gf8_muladd_0B(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -469,7 +481,8 @@ static void gf8_muladd_0B(void *out, void *in)
}
}
-static void gf8_muladd_0C(void *out, void *in)
+static void
+gf8_muladd_0C(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -513,7 +526,8 @@ static void gf8_muladd_0C(void *out, void *in)
}
}
-static void gf8_muladd_0D(void *out, void *in)
+static void
+gf8_muladd_0D(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -558,7 +572,8 @@ static void gf8_muladd_0D(void *out, void *in)
}
}
-static void gf8_muladd_0E(void *out, void *in)
+static void
+gf8_muladd_0E(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -604,7 +619,8 @@ static void gf8_muladd_0E(void *out, void *in)
}
}
-static void gf8_muladd_0F(void *out, void *in)
+static void
+gf8_muladd_0F(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -650,7 +666,8 @@ static void gf8_muladd_0F(void *out, void *in)
}
}
-static void gf8_muladd_10(void *out, void *in)
+static void
+gf8_muladd_10(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -694,7 +711,8 @@ static void gf8_muladd_10(void *out, void *in)
}
}
-static void gf8_muladd_11(void *out, void *in)
+static void
+gf8_muladd_11(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -735,7 +753,8 @@ static void gf8_muladd_11(void *out, void *in)
}
}
-static void gf8_muladd_12(void *out, void *in)
+static void
+gf8_muladd_12(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -779,7 +798,8 @@ static void gf8_muladd_12(void *out, void *in)
}
}
-static void gf8_muladd_13(void *out, void *in)
+static void
+gf8_muladd_13(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -823,7 +843,8 @@ static void gf8_muladd_13(void *out, void *in)
}
}
-static void gf8_muladd_14(void *out, void *in)
+static void
+gf8_muladd_14(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -867,7 +888,8 @@ static void gf8_muladd_14(void *out, void *in)
}
}
-static void gf8_muladd_15(void *out, void *in)
+static void
+gf8_muladd_15(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -910,7 +932,8 @@ static void gf8_muladd_15(void *out, void *in)
}
}
-static void gf8_muladd_16(void *out, void *in)
+static void
+gf8_muladd_16(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -956,7 +979,8 @@ static void gf8_muladd_16(void *out, void *in)
}
}
-static void gf8_muladd_17(void *out, void *in)
+static void
+gf8_muladd_17(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1002,7 +1026,8 @@ static void gf8_muladd_17(void *out, void *in)
}
}
-static void gf8_muladd_18(void *out, void *in)
+static void
+gf8_muladd_18(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1046,7 +1071,8 @@ static void gf8_muladd_18(void *out, void *in)
}
}
-static void gf8_muladd_19(void *out, void *in)
+static void
+gf8_muladd_19(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1090,7 +1116,8 @@ static void gf8_muladd_19(void *out, void *in)
}
}
-static void gf8_muladd_1A(void *out, void *in)
+static void
+gf8_muladd_1A(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1136,7 +1163,8 @@ static void gf8_muladd_1A(void *out, void *in)
}
}
-static void gf8_muladd_1B(void *out, void *in)
+static void
+gf8_muladd_1B(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1183,7 +1211,8 @@ static void gf8_muladd_1B(void *out, void *in)
}
}
-static void gf8_muladd_1C(void *out, void *in)
+static void
+gf8_muladd_1C(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1230,7 +1259,8 @@ static void gf8_muladd_1C(void *out, void *in)
}
}
-static void gf8_muladd_1D(void *out, void *in)
+static void
+gf8_muladd_1D(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1277,7 +1307,8 @@ static void gf8_muladd_1D(void *out, void *in)
}
}
-static void gf8_muladd_1E(void *out, void *in)
+static void
+gf8_muladd_1E(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1323,7 +1354,8 @@ static void gf8_muladd_1E(void *out, void *in)
}
}
-static void gf8_muladd_1F(void *out, void *in)
+static void
+gf8_muladd_1F(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1367,7 +1399,8 @@ static void gf8_muladd_1F(void *out, void *in)
}
}
-static void gf8_muladd_20(void *out, void *in)
+static void
+gf8_muladd_20(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1411,7 +1444,8 @@ static void gf8_muladd_20(void *out, void *in)
}
}
-static void gf8_muladd_21(void *out, void *in)
+static void
+gf8_muladd_21(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1454,7 +1488,8 @@ static void gf8_muladd_21(void *out, void *in)
}
}
-static void gf8_muladd_22(void *out, void *in)
+static void
+gf8_muladd_22(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1495,7 +1530,8 @@ static void gf8_muladd_22(void *out, void *in)
}
}
-static void gf8_muladd_23(void *out, void *in)
+static void
+gf8_muladd_23(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1536,7 +1572,8 @@ static void gf8_muladd_23(void *out, void *in)
}
}
-static void gf8_muladd_24(void *out, void *in)
+static void
+gf8_muladd_24(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1580,7 +1617,8 @@ static void gf8_muladd_24(void *out, void *in)
}
}
-static void gf8_muladd_25(void *out, void *in)
+static void
+gf8_muladd_25(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1623,7 +1661,8 @@ static void gf8_muladd_25(void *out, void *in)
}
}
-static void gf8_muladd_26(void *out, void *in)
+static void
+gf8_muladd_26(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1668,7 +1707,8 @@ static void gf8_muladd_26(void *out, void *in)
}
}
-static void gf8_muladd_27(void *out, void *in)
+static void
+gf8_muladd_27(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1709,7 +1749,8 @@ static void gf8_muladd_27(void *out, void *in)
}
}
-static void gf8_muladd_28(void *out, void *in)
+static void
+gf8_muladd_28(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1754,7 +1795,8 @@ static void gf8_muladd_28(void *out, void *in)
}
}
-static void gf8_muladd_29(void *out, void *in)
+static void
+gf8_muladd_29(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1799,7 +1841,8 @@ static void gf8_muladd_29(void *out, void *in)
}
}
-static void gf8_muladd_2A(void *out, void *in)
+static void
+gf8_muladd_2A(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1843,7 +1886,8 @@ static void gf8_muladd_2A(void *out, void *in)
}
}
-static void gf8_muladd_2B(void *out, void *in)
+static void
+gf8_muladd_2B(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1887,7 +1931,8 @@ static void gf8_muladd_2B(void *out, void *in)
}
}
-static void gf8_muladd_2C(void *out, void *in)
+static void
+gf8_muladd_2C(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1933,7 +1978,8 @@ static void gf8_muladd_2C(void *out, void *in)
}
}
-static void gf8_muladd_2D(void *out, void *in)
+static void
+gf8_muladd_2D(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -1979,7 +2025,8 @@ static void gf8_muladd_2D(void *out, void *in)
}
}
-static void gf8_muladd_2E(void *out, void *in)
+static void
+gf8_muladd_2E(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2024,7 +2071,8 @@ static void gf8_muladd_2E(void *out, void *in)
}
}
-static void gf8_muladd_2F(void *out, void *in)
+static void
+gf8_muladd_2F(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2069,7 +2117,8 @@ static void gf8_muladd_2F(void *out, void *in)
}
}
-static void gf8_muladd_30(void *out, void *in)
+static void
+gf8_muladd_30(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2113,7 +2162,8 @@ static void gf8_muladd_30(void *out, void *in)
}
}
-static void gf8_muladd_31(void *out, void *in)
+static void
+gf8_muladd_31(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2158,7 +2208,8 @@ static void gf8_muladd_31(void *out, void *in)
}
}
-static void gf8_muladd_32(void *out, void *in)
+static void
+gf8_muladd_32(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2202,7 +2253,8 @@ static void gf8_muladd_32(void *out, void *in)
}
}
-static void gf8_muladd_33(void *out, void *in)
+static void
+gf8_muladd_33(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2249,7 +2301,8 @@ static void gf8_muladd_33(void *out, void *in)
}
}
-static void gf8_muladd_34(void *out, void *in)
+static void
+gf8_muladd_34(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2296,7 +2349,8 @@ static void gf8_muladd_34(void *out, void *in)
}
}
-static void gf8_muladd_35(void *out, void *in)
+static void
+gf8_muladd_35(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2341,7 +2395,8 @@ static void gf8_muladd_35(void *out, void *in)
}
}
-static void gf8_muladd_36(void *out, void *in)
+static void
+gf8_muladd_36(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2385,7 +2440,8 @@ static void gf8_muladd_36(void *out, void *in)
}
}
-static void gf8_muladd_37(void *out, void *in)
+static void
+gf8_muladd_37(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2431,7 +2487,8 @@ static void gf8_muladd_37(void *out, void *in)
}
}
-static void gf8_muladd_38(void *out, void *in)
+static void
+gf8_muladd_38(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2476,7 +2533,8 @@ static void gf8_muladd_38(void *out, void *in)
}
}
-static void gf8_muladd_39(void *out, void *in)
+static void
+gf8_muladd_39(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2521,7 +2579,8 @@ static void gf8_muladd_39(void *out, void *in)
}
}
-static void gf8_muladd_3A(void *out, void *in)
+static void
+gf8_muladd_3A(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2569,7 +2628,8 @@ static void gf8_muladd_3A(void *out, void *in)
}
}
-static void gf8_muladd_3B(void *out, void *in)
+static void
+gf8_muladd_3B(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2614,7 +2674,8 @@ static void gf8_muladd_3B(void *out, void *in)
}
}
-static void gf8_muladd_3C(void *out, void *in)
+static void
+gf8_muladd_3C(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2659,7 +2720,8 @@ static void gf8_muladd_3C(void *out, void *in)
}
}
-static void gf8_muladd_3D(void *out, void *in)
+static void
+gf8_muladd_3D(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2704,7 +2766,8 @@ static void gf8_muladd_3D(void *out, void *in)
}
}
-static void gf8_muladd_3E(void *out, void *in)
+static void
+gf8_muladd_3E(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2748,7 +2811,8 @@ static void gf8_muladd_3E(void *out, void *in)
}
}
-static void gf8_muladd_3F(void *out, void *in)
+static void
+gf8_muladd_3F(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2793,7 +2857,8 @@ static void gf8_muladd_3F(void *out, void *in)
}
}
-static void gf8_muladd_40(void *out, void *in)
+static void
+gf8_muladd_40(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2837,7 +2902,8 @@ static void gf8_muladd_40(void *out, void *in)
}
}
-static void gf8_muladd_41(void *out, void *in)
+static void
+gf8_muladd_41(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2881,7 +2947,8 @@ static void gf8_muladd_41(void *out, void *in)
}
}
-static void gf8_muladd_42(void *out, void *in)
+static void
+gf8_muladd_42(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2922,7 +2989,8 @@ static void gf8_muladd_42(void *out, void *in)
}
}
-static void gf8_muladd_43(void *out, void *in)
+static void
+gf8_muladd_43(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -2963,7 +3031,8 @@ static void gf8_muladd_43(void *out, void *in)
}
}
-static void gf8_muladd_44(void *out, void *in)
+static void
+gf8_muladd_44(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3006,7 +3075,8 @@ static void gf8_muladd_44(void *out, void *in)
}
}
-static void gf8_muladd_45(void *out, void *in)
+static void
+gf8_muladd_45(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3047,7 +3117,8 @@ static void gf8_muladd_45(void *out, void *in)
}
}
-static void gf8_muladd_46(void *out, void *in)
+static void
+gf8_muladd_46(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3088,7 +3159,8 @@ static void gf8_muladd_46(void *out, void *in)
}
}
-static void gf8_muladd_47(void *out, void *in)
+static void
+gf8_muladd_47(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3131,7 +3203,8 @@ static void gf8_muladd_47(void *out, void *in)
}
}
-static void gf8_muladd_48(void *out, void *in)
+static void
+gf8_muladd_48(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3175,7 +3248,8 @@ static void gf8_muladd_48(void *out, void *in)
}
}
-static void gf8_muladd_49(void *out, void *in)
+static void
+gf8_muladd_49(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3219,7 +3293,8 @@ static void gf8_muladd_49(void *out, void *in)
}
}
-static void gf8_muladd_4A(void *out, void *in)
+static void
+gf8_muladd_4A(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3263,7 +3338,8 @@ static void gf8_muladd_4A(void *out, void *in)
}
}
-static void gf8_muladd_4B(void *out, void *in)
+static void
+gf8_muladd_4B(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3309,7 +3385,8 @@ static void gf8_muladd_4B(void *out, void *in)
}
}
-static void gf8_muladd_4C(void *out, void *in)
+static void
+gf8_muladd_4C(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3354,7 +3431,8 @@ static void gf8_muladd_4C(void *out, void *in)
}
}
-static void gf8_muladd_4D(void *out, void *in)
+static void
+gf8_muladd_4D(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3399,7 +3477,8 @@ static void gf8_muladd_4D(void *out, void *in)
}
}
-static void gf8_muladd_4E(void *out, void *in)
+static void
+gf8_muladd_4E(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3440,7 +3519,8 @@ static void gf8_muladd_4E(void *out, void *in)
}
}
-static void gf8_muladd_4F(void *out, void *in)
+static void
+gf8_muladd_4F(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3481,7 +3561,8 @@ static void gf8_muladd_4F(void *out, void *in)
}
}
-static void gf8_muladd_50(void *out, void *in)
+static void
+gf8_muladd_50(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3526,7 +3607,8 @@ static void gf8_muladd_50(void *out, void *in)
}
}
-static void gf8_muladd_51(void *out, void *in)
+static void
+gf8_muladd_51(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3567,7 +3649,8 @@ static void gf8_muladd_51(void *out, void *in)
}
}
-static void gf8_muladd_52(void *out, void *in)
+static void
+gf8_muladd_52(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3613,7 +3696,8 @@ static void gf8_muladd_52(void *out, void *in)
}
}
-static void gf8_muladd_53(void *out, void *in)
+static void
+gf8_muladd_53(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3654,7 +3738,8 @@ static void gf8_muladd_53(void *out, void *in)
}
}
-static void gf8_muladd_54(void *out, void *in)
+static void
+gf8_muladd_54(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3700,7 +3785,8 @@ static void gf8_muladd_54(void *out, void *in)
}
}
-static void gf8_muladd_55(void *out, void *in)
+static void
+gf8_muladd_55(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3745,7 +3831,8 @@ static void gf8_muladd_55(void *out, void *in)
}
}
-static void gf8_muladd_56(void *out, void *in)
+static void
+gf8_muladd_56(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3788,7 +3875,8 @@ static void gf8_muladd_56(void *out, void *in)
}
}
-static void gf8_muladd_57(void *out, void *in)
+static void
+gf8_muladd_57(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3832,7 +3920,8 @@ static void gf8_muladd_57(void *out, void *in)
}
}
-static void gf8_muladd_58(void *out, void *in)
+static void
+gf8_muladd_58(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3876,7 +3965,8 @@ static void gf8_muladd_58(void *out, void *in)
}
}
-static void gf8_muladd_59(void *out, void *in)
+static void
+gf8_muladd_59(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3920,7 +4010,8 @@ static void gf8_muladd_59(void *out, void *in)
}
}
-static void gf8_muladd_5A(void *out, void *in)
+static void
+gf8_muladd_5A(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -3965,7 +4056,8 @@ static void gf8_muladd_5A(void *out, void *in)
}
}
-static void gf8_muladd_5B(void *out, void *in)
+static void
+gf8_muladd_5B(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4012,7 +4104,8 @@ static void gf8_muladd_5B(void *out, void *in)
}
}
-static void gf8_muladd_5C(void *out, void *in)
+static void
+gf8_muladd_5C(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4057,7 +4150,8 @@ static void gf8_muladd_5C(void *out, void *in)
}
}
-static void gf8_muladd_5D(void *out, void *in)
+static void
+gf8_muladd_5D(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4104,7 +4198,8 @@ static void gf8_muladd_5D(void *out, void *in)
}
}
-static void gf8_muladd_5E(void *out, void *in)
+static void
+gf8_muladd_5E(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4151,7 +4246,8 @@ static void gf8_muladd_5E(void *out, void *in)
}
}
-static void gf8_muladd_5F(void *out, void *in)
+static void
+gf8_muladd_5F(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4197,7 +4293,8 @@ static void gf8_muladd_5F(void *out, void *in)
}
}
-static void gf8_muladd_60(void *out, void *in)
+static void
+gf8_muladd_60(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4241,7 +4338,8 @@ static void gf8_muladd_60(void *out, void *in)
}
}
-static void gf8_muladd_61(void *out, void *in)
+static void
+gf8_muladd_61(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4285,7 +4383,8 @@ static void gf8_muladd_61(void *out, void *in)
}
}
-static void gf8_muladd_62(void *out, void *in)
+static void
+gf8_muladd_62(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4331,7 +4430,8 @@ static void gf8_muladd_62(void *out, void *in)
}
}
-static void gf8_muladd_63(void *out, void *in)
+static void
+gf8_muladd_63(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4378,7 +4478,8 @@ static void gf8_muladd_63(void *out, void *in)
}
}
-static void gf8_muladd_64(void *out, void *in)
+static void
+gf8_muladd_64(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4422,7 +4523,8 @@ static void gf8_muladd_64(void *out, void *in)
}
}
-static void gf8_muladd_65(void *out, void *in)
+static void
+gf8_muladd_65(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4468,7 +4570,8 @@ static void gf8_muladd_65(void *out, void *in)
}
}
-static void gf8_muladd_66(void *out, void *in)
+static void
+gf8_muladd_66(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4515,7 +4618,8 @@ static void gf8_muladd_66(void *out, void *in)
}
}
-static void gf8_muladd_67(void *out, void *in)
+static void
+gf8_muladd_67(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4561,7 +4665,8 @@ static void gf8_muladd_67(void *out, void *in)
}
}
-static void gf8_muladd_68(void *out, void *in)
+static void
+gf8_muladd_68(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4607,7 +4712,8 @@ static void gf8_muladd_68(void *out, void *in)
}
}
-static void gf8_muladd_69(void *out, void *in)
+static void
+gf8_muladd_69(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4650,7 +4756,8 @@ static void gf8_muladd_69(void *out, void *in)
}
}
-static void gf8_muladd_6A(void *out, void *in)
+static void
+gf8_muladd_6A(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4695,7 +4802,8 @@ static void gf8_muladd_6A(void *out, void *in)
}
}
-static void gf8_muladd_6B(void *out, void *in)
+static void
+gf8_muladd_6B(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4739,7 +4847,8 @@ static void gf8_muladd_6B(void *out, void *in)
}
}
-static void gf8_muladd_6C(void *out, void *in)
+static void
+gf8_muladd_6C(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4783,7 +4892,8 @@ static void gf8_muladd_6C(void *out, void *in)
}
}
-static void gf8_muladd_6D(void *out, void *in)
+static void
+gf8_muladd_6D(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4827,7 +4937,8 @@ static void gf8_muladd_6D(void *out, void *in)
}
}
-static void gf8_muladd_6E(void *out, void *in)
+static void
+gf8_muladd_6E(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4872,7 +4983,8 @@ static void gf8_muladd_6E(void *out, void *in)
}
}
-static void gf8_muladd_6F(void *out, void *in)
+static void
+gf8_muladd_6F(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4917,7 +5029,8 @@ static void gf8_muladd_6F(void *out, void *in)
}
}
-static void gf8_muladd_70(void *out, void *in)
+static void
+gf8_muladd_70(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -4962,7 +5075,8 @@ static void gf8_muladd_70(void *out, void *in)
}
}
-static void gf8_muladd_71(void *out, void *in)
+static void
+gf8_muladd_71(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5007,7 +5121,8 @@ static void gf8_muladd_71(void *out, void *in)
}
}
-static void gf8_muladd_72(void *out, void *in)
+static void
+gf8_muladd_72(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5052,7 +5167,8 @@ static void gf8_muladd_72(void *out, void *in)
}
}
-static void gf8_muladd_73(void *out, void *in)
+static void
+gf8_muladd_73(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5093,7 +5209,8 @@ static void gf8_muladd_73(void *out, void *in)
}
}
-static void gf8_muladd_74(void *out, void *in)
+static void
+gf8_muladd_74(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5137,7 +5254,8 @@ static void gf8_muladd_74(void *out, void *in)
}
}
-static void gf8_muladd_75(void *out, void *in)
+static void
+gf8_muladd_75(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5182,7 +5300,8 @@ static void gf8_muladd_75(void *out, void *in)
}
}
-static void gf8_muladd_76(void *out, void *in)
+static void
+gf8_muladd_76(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5228,7 +5347,8 @@ static void gf8_muladd_76(void *out, void *in)
}
}
-static void gf8_muladd_77(void *out, void *in)
+static void
+gf8_muladd_77(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5274,7 +5394,8 @@ static void gf8_muladd_77(void *out, void *in)
}
}
-static void gf8_muladd_78(void *out, void *in)
+static void
+gf8_muladd_78(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5319,7 +5440,8 @@ static void gf8_muladd_78(void *out, void *in)
}
}
-static void gf8_muladd_79(void *out, void *in)
+static void
+gf8_muladd_79(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5365,7 +5487,8 @@ static void gf8_muladd_79(void *out, void *in)
}
}
-static void gf8_muladd_7A(void *out, void *in)
+static void
+gf8_muladd_7A(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5409,7 +5532,8 @@ static void gf8_muladd_7A(void *out, void *in)
}
}
-static void gf8_muladd_7B(void *out, void *in)
+static void
+gf8_muladd_7B(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5454,7 +5578,8 @@ static void gf8_muladd_7B(void *out, void *in)
}
}
-static void gf8_muladd_7C(void *out, void *in)
+static void
+gf8_muladd_7C(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5498,7 +5623,8 @@ static void gf8_muladd_7C(void *out, void *in)
}
}
-static void gf8_muladd_7D(void *out, void *in)
+static void
+gf8_muladd_7D(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5544,7 +5670,8 @@ static void gf8_muladd_7D(void *out, void *in)
}
}
-static void gf8_muladd_7E(void *out, void *in)
+static void
+gf8_muladd_7E(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5589,7 +5716,8 @@ static void gf8_muladd_7E(void *out, void *in)
}
}
-static void gf8_muladd_7F(void *out, void *in)
+static void
+gf8_muladd_7F(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5635,7 +5763,8 @@ static void gf8_muladd_7F(void *out, void *in)
}
}
-static void gf8_muladd_80(void *out, void *in)
+static void
+gf8_muladd_80(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5680,7 +5809,8 @@ static void gf8_muladd_80(void *out, void *in)
}
}
-static void gf8_muladd_81(void *out, void *in)
+static void
+gf8_muladd_81(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5724,7 +5854,8 @@ static void gf8_muladd_81(void *out, void *in)
}
}
-static void gf8_muladd_82(void *out, void *in)
+static void
+gf8_muladd_82(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5767,7 +5898,8 @@ static void gf8_muladd_82(void *out, void *in)
}
}
-static void gf8_muladd_83(void *out, void *in)
+static void
+gf8_muladd_83(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5814,7 +5946,8 @@ static void gf8_muladd_83(void *out, void *in)
}
}
-static void gf8_muladd_84(void *out, void *in)
+static void
+gf8_muladd_84(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5855,7 +5988,8 @@ static void gf8_muladd_84(void *out, void *in)
}
}
-static void gf8_muladd_85(void *out, void *in)
+static void
+gf8_muladd_85(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5901,7 +6035,8 @@ static void gf8_muladd_85(void *out, void *in)
}
}
-static void gf8_muladd_86(void *out, void *in)
+static void
+gf8_muladd_86(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5942,7 +6077,8 @@ static void gf8_muladd_86(void *out, void *in)
}
}
-static void gf8_muladd_87(void *out, void *in)
+static void
+gf8_muladd_87(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -5986,7 +6122,8 @@ static void gf8_muladd_87(void *out, void *in)
}
}
-static void gf8_muladd_88(void *out, void *in)
+static void
+gf8_muladd_88(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6030,7 +6167,8 @@ static void gf8_muladd_88(void *out, void *in)
}
}
-static void gf8_muladd_89(void *out, void *in)
+static void
+gf8_muladd_89(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6075,7 +6213,8 @@ static void gf8_muladd_89(void *out, void *in)
}
}
-static void gf8_muladd_8A(void *out, void *in)
+static void
+gf8_muladd_8A(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6116,7 +6255,8 @@ static void gf8_muladd_8A(void *out, void *in)
}
}
-static void gf8_muladd_8B(void *out, void *in)
+static void
+gf8_muladd_8B(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6163,7 +6303,8 @@ static void gf8_muladd_8B(void *out, void *in)
}
}
-static void gf8_muladd_8C(void *out, void *in)
+static void
+gf8_muladd_8C(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6204,7 +6345,8 @@ static void gf8_muladd_8C(void *out, void *in)
}
}
-static void gf8_muladd_8D(void *out, void *in)
+static void
+gf8_muladd_8D(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6247,7 +6389,8 @@ static void gf8_muladd_8D(void *out, void *in)
}
}
-static void gf8_muladd_8E(void *out, void *in)
+static void
+gf8_muladd_8E(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6288,7 +6431,8 @@ static void gf8_muladd_8E(void *out, void *in)
}
}
-static void gf8_muladd_8F(void *out, void *in)
+static void
+gf8_muladd_8F(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6331,7 +6475,8 @@ static void gf8_muladd_8F(void *out, void *in)
}
}
-static void gf8_muladd_90(void *out, void *in)
+static void
+gf8_muladd_90(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6376,7 +6521,8 @@ static void gf8_muladd_90(void *out, void *in)
}
}
-static void gf8_muladd_91(void *out, void *in)
+static void
+gf8_muladd_91(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6422,7 +6568,8 @@ static void gf8_muladd_91(void *out, void *in)
}
}
-static void gf8_muladd_92(void *out, void *in)
+static void
+gf8_muladd_92(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6466,7 +6613,8 @@ static void gf8_muladd_92(void *out, void *in)
}
}
-static void gf8_muladd_93(void *out, void *in)
+static void
+gf8_muladd_93(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6511,7 +6659,8 @@ static void gf8_muladd_93(void *out, void *in)
}
}
-static void gf8_muladd_94(void *out, void *in)
+static void
+gf8_muladd_94(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6554,7 +6703,8 @@ static void gf8_muladd_94(void *out, void *in)
}
}
-static void gf8_muladd_95(void *out, void *in)
+static void
+gf8_muladd_95(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6602,7 +6752,8 @@ static void gf8_muladd_95(void *out, void *in)
}
}
-static void gf8_muladd_96(void *out, void *in)
+static void
+gf8_muladd_96(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6647,7 +6798,8 @@ static void gf8_muladd_96(void *out, void *in)
}
}
-static void gf8_muladd_97(void *out, void *in)
+static void
+gf8_muladd_97(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6693,7 +6845,8 @@ static void gf8_muladd_97(void *out, void *in)
}
}
-static void gf8_muladd_98(void *out, void *in)
+static void
+gf8_muladd_98(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6737,7 +6890,8 @@ static void gf8_muladd_98(void *out, void *in)
}
}
-static void gf8_muladd_99(void *out, void *in)
+static void
+gf8_muladd_99(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6782,7 +6936,8 @@ static void gf8_muladd_99(void *out, void *in)
}
}
-static void gf8_muladd_9A(void *out, void *in)
+static void
+gf8_muladd_9A(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6827,7 +6982,8 @@ static void gf8_muladd_9A(void *out, void *in)
}
}
-static void gf8_muladd_9B(void *out, void *in)
+static void
+gf8_muladd_9B(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6870,7 +7026,8 @@ static void gf8_muladd_9B(void *out, void *in)
}
}
-static void gf8_muladd_9C(void *out, void *in)
+static void
+gf8_muladd_9C(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6913,7 +7070,8 @@ static void gf8_muladd_9C(void *out, void *in)
}
}
-static void gf8_muladd_9D(void *out, void *in)
+static void
+gf8_muladd_9D(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6956,7 +7114,8 @@ static void gf8_muladd_9D(void *out, void *in)
}
}
-static void gf8_muladd_9E(void *out, void *in)
+static void
+gf8_muladd_9E(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -6999,7 +7158,8 @@ static void gf8_muladd_9E(void *out, void *in)
}
}
-static void gf8_muladd_9F(void *out, void *in)
+static void
+gf8_muladd_9F(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7042,7 +7202,8 @@ static void gf8_muladd_9F(void *out, void *in)
}
}
-static void gf8_muladd_A0(void *out, void *in)
+static void
+gf8_muladd_A0(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7088,7 +7249,8 @@ static void gf8_muladd_A0(void *out, void *in)
}
}
-static void gf8_muladd_A1(void *out, void *in)
+static void
+gf8_muladd_A1(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7133,7 +7295,8 @@ static void gf8_muladd_A1(void *out, void *in)
}
}
-static void gf8_muladd_A2(void *out, void *in)
+static void
+gf8_muladd_A2(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7176,7 +7339,8 @@ static void gf8_muladd_A2(void *out, void *in)
}
}
-static void gf8_muladd_A3(void *out, void *in)
+static void
+gf8_muladd_A3(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7220,7 +7384,8 @@ static void gf8_muladd_A3(void *out, void *in)
}
}
-static void gf8_muladd_A4(void *out, void *in)
+static void
+gf8_muladd_A4(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7267,7 +7432,8 @@ static void gf8_muladd_A4(void *out, void *in)
}
}
-static void gf8_muladd_A5(void *out, void *in)
+static void
+gf8_muladd_A5(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7312,7 +7478,8 @@ static void gf8_muladd_A5(void *out, void *in)
}
}
-static void gf8_muladd_A6(void *out, void *in)
+static void
+gf8_muladd_A6(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7353,7 +7520,8 @@ static void gf8_muladd_A6(void *out, void *in)
}
}
-static void gf8_muladd_A7(void *out, void *in)
+static void
+gf8_muladd_A7(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7394,7 +7562,8 @@ static void gf8_muladd_A7(void *out, void *in)
}
}
-static void gf8_muladd_A8(void *out, void *in)
+static void
+gf8_muladd_A8(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7439,7 +7608,8 @@ static void gf8_muladd_A8(void *out, void *in)
}
}
-static void gf8_muladd_A9(void *out, void *in)
+static void
+gf8_muladd_A9(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7480,7 +7650,8 @@ static void gf8_muladd_A9(void *out, void *in)
}
}
-static void gf8_muladd_AA(void *out, void *in)
+static void
+gf8_muladd_AA(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7525,7 +7696,8 @@ static void gf8_muladd_AA(void *out, void *in)
}
}
-static void gf8_muladd_AB(void *out, void *in)
+static void
+gf8_muladd_AB(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7569,7 +7741,8 @@ static void gf8_muladd_AB(void *out, void *in)
}
}
-static void gf8_muladd_AC(void *out, void *in)
+static void
+gf8_muladd_AC(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7612,7 +7785,8 @@ static void gf8_muladd_AC(void *out, void *in)
}
}
-static void gf8_muladd_AD(void *out, void *in)
+static void
+gf8_muladd_AD(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7653,7 +7827,8 @@ static void gf8_muladd_AD(void *out, void *in)
}
}
-static void gf8_muladd_AE(void *out, void *in)
+static void
+gf8_muladd_AE(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7698,7 +7873,8 @@ static void gf8_muladd_AE(void *out, void *in)
}
}
-static void gf8_muladd_AF(void *out, void *in)
+static void
+gf8_muladd_AF(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7741,7 +7917,8 @@ static void gf8_muladd_AF(void *out, void *in)
}
}
-static void gf8_muladd_B0(void *out, void *in)
+static void
+gf8_muladd_B0(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7787,7 +7964,8 @@ static void gf8_muladd_B0(void *out, void *in)
}
}
-static void gf8_muladd_B1(void *out, void *in)
+static void
+gf8_muladd_B1(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7831,7 +8009,8 @@ static void gf8_muladd_B1(void *out, void *in)
}
}
-static void gf8_muladd_B2(void *out, void *in)
+static void
+gf8_muladd_B2(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7878,7 +8057,8 @@ static void gf8_muladd_B2(void *out, void *in)
}
}
-static void gf8_muladd_B3(void *out, void *in)
+static void
+gf8_muladd_B3(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7923,7 +8103,8 @@ static void gf8_muladd_B3(void *out, void *in)
}
}
-static void gf8_muladd_B4(void *out, void *in)
+static void
+gf8_muladd_B4(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -7966,7 +8147,8 @@ static void gf8_muladd_B4(void *out, void *in)
}
}
-static void gf8_muladd_B5(void *out, void *in)
+static void
+gf8_muladd_B5(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8011,7 +8193,8 @@ static void gf8_muladd_B5(void *out, void *in)
}
}
-static void gf8_muladd_B6(void *out, void *in)
+static void
+gf8_muladd_B6(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8057,7 +8240,8 @@ static void gf8_muladd_B6(void *out, void *in)
}
}
-static void gf8_muladd_B7(void *out, void *in)
+static void
+gf8_muladd_B7(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8101,7 +8285,8 @@ static void gf8_muladd_B7(void *out, void *in)
}
}
-static void gf8_muladd_B8(void *out, void *in)
+static void
+gf8_muladd_B8(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8146,7 +8331,8 @@ static void gf8_muladd_B8(void *out, void *in)
}
}
-static void gf8_muladd_B9(void *out, void *in)
+static void
+gf8_muladd_B9(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8191,7 +8377,8 @@ static void gf8_muladd_B9(void *out, void *in)
}
}
-static void gf8_muladd_BA(void *out, void *in)
+static void
+gf8_muladd_BA(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8236,7 +8423,8 @@ static void gf8_muladd_BA(void *out, void *in)
}
}
-static void gf8_muladd_BB(void *out, void *in)
+static void
+gf8_muladd_BB(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8280,7 +8468,8 @@ static void gf8_muladd_BB(void *out, void *in)
}
}
-static void gf8_muladd_BC(void *out, void *in)
+static void
+gf8_muladd_BC(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8325,7 +8514,8 @@ static void gf8_muladd_BC(void *out, void *in)
}
}
-static void gf8_muladd_BD(void *out, void *in)
+static void
+gf8_muladd_BD(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8370,7 +8560,8 @@ static void gf8_muladd_BD(void *out, void *in)
}
}
-static void gf8_muladd_BE(void *out, void *in)
+static void
+gf8_muladd_BE(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8413,7 +8604,8 @@ static void gf8_muladd_BE(void *out, void *in)
}
}
-static void gf8_muladd_BF(void *out, void *in)
+static void
+gf8_muladd_BF(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8459,7 +8651,8 @@ static void gf8_muladd_BF(void *out, void *in)
}
}
-static void gf8_muladd_C0(void *out, void *in)
+static void
+gf8_muladd_C0(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8503,7 +8696,8 @@ static void gf8_muladd_C0(void *out, void *in)
}
}
-static void gf8_muladd_C1(void *out, void *in)
+static void
+gf8_muladd_C1(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8547,7 +8741,8 @@ static void gf8_muladd_C1(void *out, void *in)
}
}
-static void gf8_muladd_C2(void *out, void *in)
+static void
+gf8_muladd_C2(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8591,7 +8786,8 @@ static void gf8_muladd_C2(void *out, void *in)
}
}
-static void gf8_muladd_C3(void *out, void *in)
+static void
+gf8_muladd_C3(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8636,7 +8832,8 @@ static void gf8_muladd_C3(void *out, void *in)
}
}
-static void gf8_muladd_C4(void *out, void *in)
+static void
+gf8_muladd_C4(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8680,7 +8877,8 @@ static void gf8_muladd_C4(void *out, void *in)
}
}
-static void gf8_muladd_C5(void *out, void *in)
+static void
+gf8_muladd_C5(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8723,7 +8921,8 @@ static void gf8_muladd_C5(void *out, void *in)
}
}
-static void gf8_muladd_C6(void *out, void *in)
+static void
+gf8_muladd_C6(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8771,7 +8970,8 @@ static void gf8_muladd_C6(void *out, void *in)
}
}
-static void gf8_muladd_C7(void *out, void *in)
+static void
+gf8_muladd_C7(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8815,7 +9015,8 @@ static void gf8_muladd_C7(void *out, void *in)
}
}
-static void gf8_muladd_C8(void *out, void *in)
+static void
+gf8_muladd_C8(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8859,7 +9060,8 @@ static void gf8_muladd_C8(void *out, void *in)
}
}
-static void gf8_muladd_C9(void *out, void *in)
+static void
+gf8_muladd_C9(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8902,7 +9104,8 @@ static void gf8_muladd_C9(void *out, void *in)
}
}
-static void gf8_muladd_CA(void *out, void *in)
+static void
+gf8_muladd_CA(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8948,7 +9151,8 @@ static void gf8_muladd_CA(void *out, void *in)
}
}
-static void gf8_muladd_CB(void *out, void *in)
+static void
+gf8_muladd_CB(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -8993,7 +9197,8 @@ static void gf8_muladd_CB(void *out, void *in)
}
}
-static void gf8_muladd_CC(void *out, void *in)
+static void
+gf8_muladd_CC(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9039,7 +9244,8 @@ static void gf8_muladd_CC(void *out, void *in)
}
}
-static void gf8_muladd_CD(void *out, void *in)
+static void
+gf8_muladd_CD(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9083,7 +9289,8 @@ static void gf8_muladd_CD(void *out, void *in)
}
}
-static void gf8_muladd_CE(void *out, void *in)
+static void
+gf8_muladd_CE(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9128,7 +9335,8 @@ static void gf8_muladd_CE(void *out, void *in)
}
}
-static void gf8_muladd_CF(void *out, void *in)
+static void
+gf8_muladd_CF(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9173,7 +9381,8 @@ static void gf8_muladd_CF(void *out, void *in)
}
}
-static void gf8_muladd_D0(void *out, void *in)
+static void
+gf8_muladd_D0(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9220,7 +9429,8 @@ static void gf8_muladd_D0(void *out, void *in)
}
}
-static void gf8_muladd_D1(void *out, void *in)
+static void
+gf8_muladd_D1(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9265,7 +9475,8 @@ static void gf8_muladd_D1(void *out, void *in)
}
}
-static void gf8_muladd_D2(void *out, void *in)
+static void
+gf8_muladd_D2(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9308,7 +9519,8 @@ static void gf8_muladd_D2(void *out, void *in)
}
}
-static void gf8_muladd_D3(void *out, void *in)
+static void
+gf8_muladd_D3(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9354,7 +9566,8 @@ static void gf8_muladd_D3(void *out, void *in)
}
}
-static void gf8_muladd_D4(void *out, void *in)
+static void
+gf8_muladd_D4(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9399,7 +9612,8 @@ static void gf8_muladd_D4(void *out, void *in)
}
}
-static void gf8_muladd_D5(void *out, void *in)
+static void
+gf8_muladd_D5(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9444,7 +9658,8 @@ static void gf8_muladd_D5(void *out, void *in)
}
}
-static void gf8_muladd_D6(void *out, void *in)
+static void
+gf8_muladd_D6(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9488,7 +9703,8 @@ static void gf8_muladd_D6(void *out, void *in)
}
}
-static void gf8_muladd_D7(void *out, void *in)
+static void
+gf8_muladd_D7(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9533,7 +9749,8 @@ static void gf8_muladd_D7(void *out, void *in)
}
}
-static void gf8_muladd_D8(void *out, void *in)
+static void
+gf8_muladd_D8(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9577,7 +9794,8 @@ static void gf8_muladd_D8(void *out, void *in)
}
}
-static void gf8_muladd_D9(void *out, void *in)
+static void
+gf8_muladd_D9(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9618,7 +9836,8 @@ static void gf8_muladd_D9(void *out, void *in)
}
}
-static void gf8_muladd_DA(void *out, void *in)
+static void
+gf8_muladd_DA(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9662,7 +9881,8 @@ static void gf8_muladd_DA(void *out, void *in)
}
}
-static void gf8_muladd_DB(void *out, void *in)
+static void
+gf8_muladd_DB(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9709,7 +9929,8 @@ static void gf8_muladd_DB(void *out, void *in)
}
}
-static void gf8_muladd_DC(void *out, void *in)
+static void
+gf8_muladd_DC(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9755,7 +9976,8 @@ static void gf8_muladd_DC(void *out, void *in)
}
}
-static void gf8_muladd_DD(void *out, void *in)
+static void
+gf8_muladd_DD(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9796,7 +10018,8 @@ static void gf8_muladd_DD(void *out, void *in)
}
}
-static void gf8_muladd_DE(void *out, void *in)
+static void
+gf8_muladd_DE(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9840,7 +10063,8 @@ static void gf8_muladd_DE(void *out, void *in)
}
}
-static void gf8_muladd_DF(void *out, void *in)
+static void
+gf8_muladd_DF(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9885,7 +10109,8 @@ static void gf8_muladd_DF(void *out, void *in)
}
}
-static void gf8_muladd_E0(void *out, void *in)
+static void
+gf8_muladd_E0(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9929,7 +10154,8 @@ static void gf8_muladd_E0(void *out, void *in)
}
}
-static void gf8_muladd_E1(void *out, void *in)
+static void
+gf8_muladd_E1(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -9975,7 +10201,8 @@ static void gf8_muladd_E1(void *out, void *in)
}
}
-static void gf8_muladd_E2(void *out, void *in)
+static void
+gf8_muladd_E2(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10016,7 +10243,8 @@ static void gf8_muladd_E2(void *out, void *in)
}
}
-static void gf8_muladd_E3(void *out, void *in)
+static void
+gf8_muladd_E3(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10063,7 +10291,8 @@ static void gf8_muladd_E3(void *out, void *in)
}
}
-static void gf8_muladd_E4(void *out, void *in)
+static void
+gf8_muladd_E4(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10108,7 +10337,8 @@ static void gf8_muladd_E4(void *out, void *in)
}
}
-static void gf8_muladd_E5(void *out, void *in)
+static void
+gf8_muladd_E5(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10153,7 +10383,8 @@ static void gf8_muladd_E5(void *out, void *in)
}
}
-static void gf8_muladd_E6(void *out, void *in)
+static void
+gf8_muladd_E6(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10194,7 +10425,8 @@ static void gf8_muladd_E6(void *out, void *in)
}
}
-static void gf8_muladd_E7(void *out, void *in)
+static void
+gf8_muladd_E7(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10241,7 +10473,8 @@ static void gf8_muladd_E7(void *out, void *in)
}
}
-static void gf8_muladd_E8(void *out, void *in)
+static void
+gf8_muladd_E8(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10287,7 +10520,8 @@ static void gf8_muladd_E8(void *out, void *in)
}
}
-static void gf8_muladd_E9(void *out, void *in)
+static void
+gf8_muladd_E9(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10332,7 +10566,8 @@ static void gf8_muladd_E9(void *out, void *in)
}
}
-static void gf8_muladd_EA(void *out, void *in)
+static void
+gf8_muladd_EA(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10373,7 +10608,8 @@ static void gf8_muladd_EA(void *out, void *in)
}
}
-static void gf8_muladd_EB(void *out, void *in)
+static void
+gf8_muladd_EB(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10417,7 +10653,8 @@ static void gf8_muladd_EB(void *out, void *in)
}
}
-static void gf8_muladd_EC(void *out, void *in)
+static void
+gf8_muladd_EC(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10458,7 +10695,8 @@ static void gf8_muladd_EC(void *out, void *in)
}
}
-static void gf8_muladd_ED(void *out, void *in)
+static void
+gf8_muladd_ED(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10502,7 +10740,8 @@ static void gf8_muladd_ED(void *out, void *in)
}
}
-static void gf8_muladd_EE(void *out, void *in)
+static void
+gf8_muladd_EE(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10548,7 +10787,8 @@ static void gf8_muladd_EE(void *out, void *in)
}
}
-static void gf8_muladd_EF(void *out, void *in)
+static void
+gf8_muladd_EF(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10592,7 +10832,8 @@ static void gf8_muladd_EF(void *out, void *in)
}
}
-static void gf8_muladd_F0(void *out, void *in)
+static void
+gf8_muladd_F0(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10638,7 +10879,8 @@ static void gf8_muladd_F0(void *out, void *in)
}
}
-static void gf8_muladd_F1(void *out, void *in)
+static void
+gf8_muladd_F1(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10684,7 +10926,8 @@ static void gf8_muladd_F1(void *out, void *in)
}
}
-static void gf8_muladd_F2(void *out, void *in)
+static void
+gf8_muladd_F2(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10730,7 +10973,8 @@ static void gf8_muladd_F2(void *out, void *in)
}
}
-static void gf8_muladd_F3(void *out, void *in)
+static void
+gf8_muladd_F3(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10774,7 +11018,8 @@ static void gf8_muladd_F3(void *out, void *in)
}
}
-static void gf8_muladd_F4(void *out, void *in)
+static void
+gf8_muladd_F4(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10817,7 +11062,8 @@ static void gf8_muladd_F4(void *out, void *in)
}
}
-static void gf8_muladd_F5(void *out, void *in)
+static void
+gf8_muladd_F5(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10860,7 +11106,8 @@ static void gf8_muladd_F5(void *out, void *in)
}
}
-static void gf8_muladd_F6(void *out, void *in)
+static void
+gf8_muladd_F6(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10904,7 +11151,8 @@ static void gf8_muladd_F6(void *out, void *in)
}
}
-static void gf8_muladd_F7(void *out, void *in)
+static void
+gf8_muladd_F7(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10947,7 +11195,8 @@ static void gf8_muladd_F7(void *out, void *in)
}
}
-static void gf8_muladd_F8(void *out, void *in)
+static void
+gf8_muladd_F8(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -10992,7 +11241,8 @@ static void gf8_muladd_F8(void *out, void *in)
}
}
-static void gf8_muladd_F9(void *out, void *in)
+static void
+gf8_muladd_F9(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -11039,7 +11289,8 @@ static void gf8_muladd_F9(void *out, void *in)
}
}
-static void gf8_muladd_FA(void *out, void *in)
+static void
+gf8_muladd_FA(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -11085,7 +11336,8 @@ static void gf8_muladd_FA(void *out, void *in)
}
}
-static void gf8_muladd_FB(void *out, void *in)
+static void
+gf8_muladd_FB(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -11129,7 +11381,8 @@ static void gf8_muladd_FB(void *out, void *in)
}
}
-static void gf8_muladd_FC(void *out, void *in)
+static void
+gf8_muladd_FC(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -11175,7 +11428,8 @@ static void gf8_muladd_FC(void *out, void *in)
}
}
-static void gf8_muladd_FD(void *out, void *in)
+static void
+gf8_muladd_FD(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -11221,7 +11475,8 @@ static void gf8_muladd_FD(void *out, void *in)
}
}
-static void gf8_muladd_FE(void *out, void *in)
+static void
+gf8_muladd_FE(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -11268,7 +11523,8 @@ static void gf8_muladd_FE(void *out, void *in)
}
}
-static void gf8_muladd_FF(void *out, void *in)
+static void
+gf8_muladd_FF(void *out, void *in)
{
unsigned int i;
uint64_t *in_ptr = (uint64_t *)in;
@@ -11315,75 +11571,65 @@ static void gf8_muladd_FF(void *out, void *in)
}
static void (*gf8_muladd[])(void *out, void *in) = {
- gf8_muladd_00, gf8_muladd_01, gf8_muladd_02, gf8_muladd_03,
- gf8_muladd_04, gf8_muladd_05, gf8_muladd_06, gf8_muladd_07,
- gf8_muladd_08, gf8_muladd_09, gf8_muladd_0A, gf8_muladd_0B,
- gf8_muladd_0C, gf8_muladd_0D, gf8_muladd_0E, gf8_muladd_0F,
- gf8_muladd_10, gf8_muladd_11, gf8_muladd_12, gf8_muladd_13,
- gf8_muladd_14, gf8_muladd_15, gf8_muladd_16, gf8_muladd_17,
- gf8_muladd_18, gf8_muladd_19, gf8_muladd_1A, gf8_muladd_1B,
- gf8_muladd_1C, gf8_muladd_1D, gf8_muladd_1E, gf8_muladd_1F,
- gf8_muladd_20, gf8_muladd_21, gf8_muladd_22, gf8_muladd_23,
- gf8_muladd_24, gf8_muladd_25, gf8_muladd_26, gf8_muladd_27,
- gf8_muladd_28, gf8_muladd_29, gf8_muladd_2A, gf8_muladd_2B,
- gf8_muladd_2C, gf8_muladd_2D, gf8_muladd_2E, gf8_muladd_2F,
- gf8_muladd_30, gf8_muladd_31, gf8_muladd_32, gf8_muladd_33,
- gf8_muladd_34, gf8_muladd_35, gf8_muladd_36, gf8_muladd_37,
- gf8_muladd_38, gf8_muladd_39, gf8_muladd_3A, gf8_muladd_3B,
- gf8_muladd_3C, gf8_muladd_3D, gf8_muladd_3E, gf8_muladd_3F,
- gf8_muladd_40, gf8_muladd_41, gf8_muladd_42, gf8_muladd_43,
- gf8_muladd_44, gf8_muladd_45, gf8_muladd_46, gf8_muladd_47,
- gf8_muladd_48, gf8_muladd_49, gf8_muladd_4A, gf8_muladd_4B,
- gf8_muladd_4C, gf8_muladd_4D, gf8_muladd_4E, gf8_muladd_4F,
- gf8_muladd_50, gf8_muladd_51, gf8_muladd_52, gf8_muladd_53,
- gf8_muladd_54, gf8_muladd_55, gf8_muladd_56, gf8_muladd_57,
- gf8_muladd_58, gf8_muladd_59, gf8_muladd_5A, gf8_muladd_5B,
- gf8_muladd_5C, gf8_muladd_5D, gf8_muladd_5E, gf8_muladd_5F,
- gf8_muladd_60, gf8_muladd_61, gf8_muladd_62, gf8_muladd_63,
- gf8_muladd_64, gf8_muladd_65, gf8_muladd_66, gf8_muladd_67,
- gf8_muladd_68, gf8_muladd_69, gf8_muladd_6A, gf8_muladd_6B,
- gf8_muladd_6C, gf8_muladd_6D, gf8_muladd_6E, gf8_muladd_6F,
- gf8_muladd_70, gf8_muladd_71, gf8_muladd_72, gf8_muladd_73,
- gf8_muladd_74, gf8_muladd_75, gf8_muladd_76, gf8_muladd_77,
- gf8_muladd_78, gf8_muladd_79, gf8_muladd_7A, gf8_muladd_7B,
- gf8_muladd_7C, gf8_muladd_7D, gf8_muladd_7E, gf8_muladd_7F,
- gf8_muladd_80, gf8_muladd_81, gf8_muladd_82, gf8_muladd_83,
- gf8_muladd_84, gf8_muladd_85, gf8_muladd_86, gf8_muladd_87,
- gf8_muladd_88, gf8_muladd_89, gf8_muladd_8A, gf8_muladd_8B,
- gf8_muladd_8C, gf8_muladd_8D, gf8_muladd_8E, gf8_muladd_8F,
- gf8_muladd_90, gf8_muladd_91, gf8_muladd_92, gf8_muladd_93,
- gf8_muladd_94, gf8_muladd_95, gf8_muladd_96, gf8_muladd_97,
- gf8_muladd_98, gf8_muladd_99, gf8_muladd_9A, gf8_muladd_9B,
- gf8_muladd_9C, gf8_muladd_9D, gf8_muladd_9E, gf8_muladd_9F,
- gf8_muladd_A0, gf8_muladd_A1, gf8_muladd_A2, gf8_muladd_A3,
- gf8_muladd_A4, gf8_muladd_A5, gf8_muladd_A6, gf8_muladd_A7,
- gf8_muladd_A8, gf8_muladd_A9, gf8_muladd_AA, gf8_muladd_AB,
- gf8_muladd_AC, gf8_muladd_AD, gf8_muladd_AE, gf8_muladd_AF,
- gf8_muladd_B0, gf8_muladd_B1, gf8_muladd_B2, gf8_muladd_B3,
- gf8_muladd_B4, gf8_muladd_B5, gf8_muladd_B6, gf8_muladd_B7,
- gf8_muladd_B8, gf8_muladd_B9, gf8_muladd_BA, gf8_muladd_BB,
- gf8_muladd_BC, gf8_muladd_BD, gf8_muladd_BE, gf8_muladd_BF,
- gf8_muladd_C0, gf8_muladd_C1, gf8_muladd_C2, gf8_muladd_C3,
- gf8_muladd_C4, gf8_muladd_C5, gf8_muladd_C6, gf8_muladd_C7,
- gf8_muladd_C8, gf8_muladd_C9, gf8_muladd_CA, gf8_muladd_CB,
- gf8_muladd_CC, gf8_muladd_CD, gf8_muladd_CE, gf8_muladd_CF,
- gf8_muladd_D0, gf8_muladd_D1, gf8_muladd_D2, gf8_muladd_D3,
- gf8_muladd_D4, gf8_muladd_D5, gf8_muladd_D6, gf8_muladd_D7,
- gf8_muladd_D8, gf8_muladd_D9, gf8_muladd_DA, gf8_muladd_DB,
- gf8_muladd_DC, gf8_muladd_DD, gf8_muladd_DE, gf8_muladd_DF,
- gf8_muladd_E0, gf8_muladd_E1, gf8_muladd_E2, gf8_muladd_E3,
- gf8_muladd_E4, gf8_muladd_E5, gf8_muladd_E6, gf8_muladd_E7,
- gf8_muladd_E8, gf8_muladd_E9, gf8_muladd_EA, gf8_muladd_EB,
- gf8_muladd_EC, gf8_muladd_ED, gf8_muladd_EE, gf8_muladd_EF,
- gf8_muladd_F0, gf8_muladd_F1, gf8_muladd_F2, gf8_muladd_F3,
- gf8_muladd_F4, gf8_muladd_F5, gf8_muladd_F6, gf8_muladd_F7,
- gf8_muladd_F8, gf8_muladd_F9, gf8_muladd_FA, gf8_muladd_FB,
- gf8_muladd_FC, gf8_muladd_FD, gf8_muladd_FE, gf8_muladd_FF
+ gf8_muladd_00, gf8_muladd_01, gf8_muladd_02, gf8_muladd_03, gf8_muladd_04,
+ gf8_muladd_05, gf8_muladd_06, gf8_muladd_07, gf8_muladd_08, gf8_muladd_09,
+ gf8_muladd_0A, gf8_muladd_0B, gf8_muladd_0C, gf8_muladd_0D, gf8_muladd_0E,
+ gf8_muladd_0F, gf8_muladd_10, gf8_muladd_11, gf8_muladd_12, gf8_muladd_13,
+ gf8_muladd_14, gf8_muladd_15, gf8_muladd_16, gf8_muladd_17, gf8_muladd_18,
+ gf8_muladd_19, gf8_muladd_1A, gf8_muladd_1B, gf8_muladd_1C, gf8_muladd_1D,
+ gf8_muladd_1E, gf8_muladd_1F, gf8_muladd_20, gf8_muladd_21, gf8_muladd_22,
+ gf8_muladd_23, gf8_muladd_24, gf8_muladd_25, gf8_muladd_26, gf8_muladd_27,
+ gf8_muladd_28, gf8_muladd_29, gf8_muladd_2A, gf8_muladd_2B, gf8_muladd_2C,
+ gf8_muladd_2D, gf8_muladd_2E, gf8_muladd_2F, gf8_muladd_30, gf8_muladd_31,
+ gf8_muladd_32, gf8_muladd_33, gf8_muladd_34, gf8_muladd_35, gf8_muladd_36,
+ gf8_muladd_37, gf8_muladd_38, gf8_muladd_39, gf8_muladd_3A, gf8_muladd_3B,
+ gf8_muladd_3C, gf8_muladd_3D, gf8_muladd_3E, gf8_muladd_3F, gf8_muladd_40,
+ gf8_muladd_41, gf8_muladd_42, gf8_muladd_43, gf8_muladd_44, gf8_muladd_45,
+ gf8_muladd_46, gf8_muladd_47, gf8_muladd_48, gf8_muladd_49, gf8_muladd_4A,
+ gf8_muladd_4B, gf8_muladd_4C, gf8_muladd_4D, gf8_muladd_4E, gf8_muladd_4F,
+ gf8_muladd_50, gf8_muladd_51, gf8_muladd_52, gf8_muladd_53, gf8_muladd_54,
+ gf8_muladd_55, gf8_muladd_56, gf8_muladd_57, gf8_muladd_58, gf8_muladd_59,
+ gf8_muladd_5A, gf8_muladd_5B, gf8_muladd_5C, gf8_muladd_5D, gf8_muladd_5E,
+ gf8_muladd_5F, gf8_muladd_60, gf8_muladd_61, gf8_muladd_62, gf8_muladd_63,
+ gf8_muladd_64, gf8_muladd_65, gf8_muladd_66, gf8_muladd_67, gf8_muladd_68,
+ gf8_muladd_69, gf8_muladd_6A, gf8_muladd_6B, gf8_muladd_6C, gf8_muladd_6D,
+ gf8_muladd_6E, gf8_muladd_6F, gf8_muladd_70, gf8_muladd_71, gf8_muladd_72,
+ gf8_muladd_73, gf8_muladd_74, gf8_muladd_75, gf8_muladd_76, gf8_muladd_77,
+ gf8_muladd_78, gf8_muladd_79, gf8_muladd_7A, gf8_muladd_7B, gf8_muladd_7C,
+ gf8_muladd_7D, gf8_muladd_7E, gf8_muladd_7F, gf8_muladd_80, gf8_muladd_81,
+ gf8_muladd_82, gf8_muladd_83, gf8_muladd_84, gf8_muladd_85, gf8_muladd_86,
+ gf8_muladd_87, gf8_muladd_88, gf8_muladd_89, gf8_muladd_8A, gf8_muladd_8B,
+ gf8_muladd_8C, gf8_muladd_8D, gf8_muladd_8E, gf8_muladd_8F, gf8_muladd_90,
+ gf8_muladd_91, gf8_muladd_92, gf8_muladd_93, gf8_muladd_94, gf8_muladd_95,
+ gf8_muladd_96, gf8_muladd_97, gf8_muladd_98, gf8_muladd_99, gf8_muladd_9A,
+ gf8_muladd_9B, gf8_muladd_9C, gf8_muladd_9D, gf8_muladd_9E, gf8_muladd_9F,
+ gf8_muladd_A0, gf8_muladd_A1, gf8_muladd_A2, gf8_muladd_A3, gf8_muladd_A4,
+ gf8_muladd_A5, gf8_muladd_A6, gf8_muladd_A7, gf8_muladd_A8, gf8_muladd_A9,
+ gf8_muladd_AA, gf8_muladd_AB, gf8_muladd_AC, gf8_muladd_AD, gf8_muladd_AE,
+ gf8_muladd_AF, gf8_muladd_B0, gf8_muladd_B1, gf8_muladd_B2, gf8_muladd_B3,
+ gf8_muladd_B4, gf8_muladd_B5, gf8_muladd_B6, gf8_muladd_B7, gf8_muladd_B8,
+ gf8_muladd_B9, gf8_muladd_BA, gf8_muladd_BB, gf8_muladd_BC, gf8_muladd_BD,
+ gf8_muladd_BE, gf8_muladd_BF, gf8_muladd_C0, gf8_muladd_C1, gf8_muladd_C2,
+ gf8_muladd_C3, gf8_muladd_C4, gf8_muladd_C5, gf8_muladd_C6, gf8_muladd_C7,
+ gf8_muladd_C8, gf8_muladd_C9, gf8_muladd_CA, gf8_muladd_CB, gf8_muladd_CC,
+ gf8_muladd_CD, gf8_muladd_CE, gf8_muladd_CF, gf8_muladd_D0, gf8_muladd_D1,
+ gf8_muladd_D2, gf8_muladd_D3, gf8_muladd_D4, gf8_muladd_D5, gf8_muladd_D6,
+ gf8_muladd_D7, gf8_muladd_D8, gf8_muladd_D9, gf8_muladd_DA, gf8_muladd_DB,
+ gf8_muladd_DC, gf8_muladd_DD, gf8_muladd_DE, gf8_muladd_DF, gf8_muladd_E0,
+ gf8_muladd_E1, gf8_muladd_E2, gf8_muladd_E3, gf8_muladd_E4, gf8_muladd_E5,
+ gf8_muladd_E6, gf8_muladd_E7, gf8_muladd_E8, gf8_muladd_E9, gf8_muladd_EA,
+ gf8_muladd_EB, gf8_muladd_EC, gf8_muladd_ED, gf8_muladd_EE, gf8_muladd_EF,
+ gf8_muladd_F0, gf8_muladd_F1, gf8_muladd_F2, gf8_muladd_F3, gf8_muladd_F4,
+ gf8_muladd_F5, gf8_muladd_F6, gf8_muladd_F7, gf8_muladd_F8, gf8_muladd_F9,
+ gf8_muladd_FA, gf8_muladd_FB, gf8_muladd_FC, gf8_muladd_FD, gf8_muladd_FE,
+ gf8_muladd_FF};
+
+static uint64_t zero[EC_METHOD_WORD_SIZE * 8] = {
+ 0,
};
-static uint64_t zero[EC_METHOD_WORD_SIZE * 8] = {0, };
-
-void ec_code_c_prepare(ec_gf_t *gf, uint32_t *values, uint32_t count)
+void
+ec_code_c_prepare(ec_gf_t *gf, uint32_t *values, uint32_t count)
{
uint32_t i, last, tmp;
@@ -11397,8 +11643,9 @@ void ec_code_c_prepare(ec_gf_t *gf, uint32_t *values, uint32_t count)
}
}
-void ec_code_c_linear(void *dst, void *src, uint64_t offset, uint32_t *values,
- uint32_t count)
+void
+ec_code_c_linear(void *dst, void *src, uint64_t offset, uint32_t *values,
+ uint32_t count)
{
src += offset;
gf8_muladd_00(dst, src);
@@ -11409,8 +11656,9 @@ void ec_code_c_linear(void *dst, void *src, uint64_t offset, uint32_t *values,
}
}
-void ec_code_c_interleaved(void *dst, void **src, uint64_t offset,
- uint32_t *values, uint32_t count)
+void
+ec_code_c_interleaved(void *dst, void **src, uint64_t offset, uint32_t *values,
+ uint32_t count)
{
uint32_t i, last, tmp;
diff --git a/xlators/cluster/ec/src/ec-code-c.h b/xlators/cluster/ec/src/ec-code-c.h
index 92e8070e514..42b5a064eb8 100644
--- a/xlators/cluster/ec/src/ec-code-c.h
+++ b/xlators/cluster/ec/src/ec-code-c.h
@@ -13,12 +13,15 @@
#include "ec-types.h"
-void ec_code_c_prepare(ec_gf_t *gf, uint32_t *values, uint32_t count);
+void
+ec_code_c_prepare(ec_gf_t *gf, uint32_t *values, uint32_t count);
-void ec_code_c_linear(void *dst, void *src, uint64_t offset, uint32_t *values,
- uint32_t count);
+void
+ec_code_c_linear(void *dst, void *src, uint64_t offset, uint32_t *values,
+ uint32_t count);
-void ec_code_c_interleaved(void *dst, void **src, uint64_t offset,
- uint32_t *values, uint32_t count);
+void
+ec_code_c_interleaved(void *dst, void **src, uint64_t offset, uint32_t *values,
+ uint32_t count);
#endif /* __EC_CODE_C_H__ */
diff --git a/xlators/cluster/ec/src/ec-code-intel.c b/xlators/cluster/ec/src/ec-code-intel.c
index b9fdcad4421..f1c4e13e321 100644
--- a/xlators/cluster/ec/src/ec-code-intel.c
+++ b/xlators/cluster/ec/src/ec-code-intel.c
@@ -71,16 +71,15 @@ ec_code_intel_vex(ec_code_intel_t *intel, gf_boolean_t w, gf_boolean_t l,
uint32_t reg)
{
ec_code_intel_rex(intel, w);
- if (((intel->rex.w == 1) ||
- (intel->rex.x == 0) ||
- (intel->rex.b == 0)) ||
+ if (((intel->rex.w == 1) || (intel->rex.x == 0) || (intel->rex.b == 0)) ||
((opcode != VEX_OPCODE_NONE) && (opcode != VEX_OPCODE_0F))) {
intel->rex.present = _gf_false;
intel->vex.bytes = 3;
intel->vex.data[0] = 0xC4;
intel->vex.data[1] = ((intel->rex.r << 7) | (intel->rex.x << 6) |
- (intel->rex.b << 5) | opcode) ^ 0xE0;
+ (intel->rex.b << 5) | opcode) ^
+ 0xE0;
intel->vex.data[2] = (intel->rex.w << 7) | ((~reg & 0x0F) << 3) |
(l ? 0x04 : 0x00) | prefix;
} else {
@@ -214,22 +213,17 @@ ec_code_intel_emit(ec_code_builder_t *builder, ec_code_intel_t *intel)
insn[count++] = intel->vex.data[i];
}
if (intel->rex.present) {
- insn[count++] = 0x40 |
- (intel->rex.w << 3) |
- (intel->rex.r << 2) |
- (intel->rex.x << 1) |
- (intel->rex.b << 0);
+ insn[count++] = 0x40 | (intel->rex.w << 3) | (intel->rex.r << 2) |
+ (intel->rex.x << 1) | (intel->rex.b << 0);
}
for (i = 0; i < intel->opcode.bytes; i++) {
insn[count++] = intel->opcode.data[i];
}
if (intel->modrm.present) {
- insn[count++] = (intel->modrm.mod << 6) |
- (intel->modrm.reg << 3) |
+ insn[count++] = (intel->modrm.mod << 6) | (intel->modrm.reg << 3) |
(intel->modrm.rm << 0);
if (intel->sib.present) {
- insn[count++] = (intel->sib.scale << 6) |
- (intel->sib.index << 3) |
+ insn[count++] = (intel->sib.scale << 6) | (intel->sib.index << 3) |
(intel->sib.base << 0);
}
}
@@ -467,9 +461,9 @@ ec_code_intel_op_mov_sse2m(ec_code_builder_t *builder, uint32_t src,
}
void
-ec_code_intel_op_mov_m2sse(ec_code_builder_t *builder,
- ec_code_intel_reg_t base, ec_code_intel_reg_t index,
- uint32_t scale, int32_t offset, uint32_t dst)
+ec_code_intel_op_mov_m2sse(ec_code_builder_t *builder, ec_code_intel_reg_t base,
+ ec_code_intel_reg_t index, uint32_t scale,
+ int32_t offset, uint32_t dst)
{
ec_code_intel_t intel;
@@ -500,9 +494,9 @@ ec_code_intel_op_xor_sse2sse(ec_code_builder_t *builder, uint32_t src,
}
void
-ec_code_intel_op_xor_m2sse(ec_code_builder_t *builder,
- ec_code_intel_reg_t base, ec_code_intel_reg_t index,
- uint32_t scale, int32_t offset, uint32_t dst)
+ec_code_intel_op_xor_m2sse(ec_code_builder_t *builder, ec_code_intel_reg_t base,
+ ec_code_intel_reg_t index, uint32_t scale,
+ int32_t offset, uint32_t dst)
{
ec_code_intel_t intel;
@@ -526,8 +520,8 @@ ec_code_intel_op_mov_avx2avx(ec_code_builder_t *builder, uint32_t src,
ec_code_intel_modrm_reg(&intel, src, dst);
ec_code_intel_op_1(&intel, 0x6F, 0);
- ec_code_intel_vex(&intel, _gf_false, _gf_true, VEX_OPCODE_0F,
- VEX_PREFIX_66, VEX_REG_NONE);
+ ec_code_intel_vex(&intel, _gf_false, _gf_true, VEX_OPCODE_0F, VEX_PREFIX_66,
+ VEX_REG_NONE);
ec_code_intel_emit(builder, &intel);
}
@@ -543,16 +537,16 @@ ec_code_intel_op_mov_avx2m(ec_code_builder_t *builder, uint32_t src,
ec_code_intel_modrm_mem(&intel, src, base, index, scale, offset);
ec_code_intel_op_1(&intel, 0x7F, 0);
- ec_code_intel_vex(&intel, _gf_false, _gf_true, VEX_OPCODE_0F,
- VEX_PREFIX_66, VEX_REG_NONE);
+ ec_code_intel_vex(&intel, _gf_false, _gf_true, VEX_OPCODE_0F, VEX_PREFIX_66,
+ VEX_REG_NONE);
ec_code_intel_emit(builder, &intel);
}
void
-ec_code_intel_op_mov_m2avx(ec_code_builder_t *builder,
- ec_code_intel_reg_t base, ec_code_intel_reg_t index,
- uint32_t scale, int32_t offset, uint32_t dst)
+ec_code_intel_op_mov_m2avx(ec_code_builder_t *builder, ec_code_intel_reg_t base,
+ ec_code_intel_reg_t index, uint32_t scale,
+ int32_t offset, uint32_t dst)
{
ec_code_intel_t intel;
@@ -560,8 +554,8 @@ ec_code_intel_op_mov_m2avx(ec_code_builder_t *builder,
ec_code_intel_modrm_mem(&intel, dst, base, index, scale, offset);
ec_code_intel_op_1(&intel, 0x6F, 0);
- ec_code_intel_vex(&intel, _gf_false, _gf_true, VEX_OPCODE_0F,
- VEX_PREFIX_66, VEX_REG_NONE);
+ ec_code_intel_vex(&intel, _gf_false, _gf_true, VEX_OPCODE_0F, VEX_PREFIX_66,
+ VEX_REG_NONE);
ec_code_intel_emit(builder, &intel);
}
@@ -576,16 +570,16 @@ ec_code_intel_op_xor_avx2avx(ec_code_builder_t *builder, uint32_t src,
ec_code_intel_modrm_reg(&intel, src, dst);
ec_code_intel_op_1(&intel, 0xEF, 0);
- ec_code_intel_vex(&intel, _gf_false, _gf_true, VEX_OPCODE_0F,
- VEX_PREFIX_66, dst);
+ ec_code_intel_vex(&intel, _gf_false, _gf_true, VEX_OPCODE_0F, VEX_PREFIX_66,
+ dst);
ec_code_intel_emit(builder, &intel);
}
void
-ec_code_intel_op_xor_m2avx(ec_code_builder_t *builder,
- ec_code_intel_reg_t base, ec_code_intel_reg_t index,
- uint32_t scale, int32_t offset, uint32_t dst)
+ec_code_intel_op_xor_m2avx(ec_code_builder_t *builder, ec_code_intel_reg_t base,
+ ec_code_intel_reg_t index, uint32_t scale,
+ int32_t offset, uint32_t dst)
{
ec_code_intel_t intel;
@@ -593,8 +587,8 @@ ec_code_intel_op_xor_m2avx(ec_code_builder_t *builder,
ec_code_intel_modrm_mem(&intel, dst, base, index, scale, offset);
ec_code_intel_op_1(&intel, 0xEF, 0);
- ec_code_intel_vex(&intel, _gf_false, _gf_true, VEX_OPCODE_0F,
- VEX_PREFIX_66, dst);
+ ec_code_intel_vex(&intel, _gf_false, _gf_true, VEX_OPCODE_0F, VEX_PREFIX_66,
+ dst);
ec_code_intel_emit(builder, &intel);
}
diff --git a/xlators/cluster/ec/src/ec-code-intel.h b/xlators/cluster/ec/src/ec-code-intel.h
index 903d023f962..3fa4a174765 100644
--- a/xlators/cluster/ec/src/ec-code-intel.h
+++ b/xlators/cluster/ec/src/ec-code-intel.h
@@ -76,109 +76,116 @@ enum _ec_code_vex_opcode {
struct _ec_code_intel_buffer {
uint32_t bytes;
union {
- uint8_t data[4];
+ uint8_t data[4];
uint32_t value;
};
};
struct _ec_code_intel_sib {
gf_boolean_t present;
- uint32_t base;
- uint32_t index;
- uint32_t scale;
+ uint32_t base;
+ uint32_t index;
+ uint32_t scale;
};
struct _ec_code_intel_modrm {
gf_boolean_t present;
- uint32_t mod;
- uint32_t rm;
- uint32_t reg;
+ uint32_t mod;
+ uint32_t rm;
+ uint32_t reg;
};
struct _ec_code_intel_rex {
gf_boolean_t present;
- uint32_t w;
- uint32_t r;
- uint32_t x;
- uint32_t b;
+ uint32_t w;
+ uint32_t r;
+ uint32_t x;
+ uint32_t b;
};
struct _ec_code_intel {
- gf_boolean_t invalid;
+ gf_boolean_t invalid;
ec_code_intel_buffer_t prefix;
ec_code_intel_buffer_t opcode;
ec_code_intel_buffer_t offset;
ec_code_intel_buffer_t immediate;
ec_code_intel_buffer_t vex;
- ec_code_intel_rex_t rex;
- ec_code_intel_modrm_t modrm;
- ec_code_intel_sib_t sib;
- uint32_t reg;
+ ec_code_intel_rex_t rex;
+ ec_code_intel_modrm_t modrm;
+ ec_code_intel_sib_t sib;
+ uint32_t reg;
};
-void ec_code_intel_op_push_r(ec_code_builder_t *builder,
- ec_code_intel_reg_t reg);
-void ec_code_intel_op_pop_r(ec_code_builder_t *builder,
- ec_code_intel_reg_t reg);
-void ec_code_intel_op_ret(ec_code_builder_t *builder, uint32_t size);
-
-void ec_code_intel_op_mov_r2r(ec_code_builder_t *builder,
- ec_code_intel_reg_t src,
- ec_code_intel_reg_t dst);
-void ec_code_intel_op_mov_r2m(ec_code_builder_t *builder,
- ec_code_intel_reg_t src,
- ec_code_intel_reg_t base,
- ec_code_intel_reg_t index, uint32_t scale,
- int32_t offset);
-void ec_code_intel_op_mov_m2r(ec_code_builder_t *builder,
- ec_code_intel_reg_t base,
- ec_code_intel_reg_t index, uint32_t scale,
- int32_t offset, ec_code_intel_reg_t dst);
-void ec_code_intel_op_xor_r2r(ec_code_builder_t *builder,
- ec_code_intel_reg_t src,
- ec_code_intel_reg_t dst);
-void ec_code_intel_op_xor_m2r(ec_code_builder_t *builder,
- ec_code_intel_reg_t base,
- ec_code_intel_reg_t index, uint32_t scale,
- int32_t offset, ec_code_intel_reg_t dst);
-void ec_code_intel_op_add_i2r(ec_code_builder_t *builder, int32_t value,
- ec_code_intel_reg_t reg);
-void ec_code_intel_op_test_i2r(ec_code_builder_t *builder, uint32_t value,
- ec_code_intel_reg_t reg);
-void ec_code_intel_op_jne(ec_code_builder_t *builder, uint32_t address);
-
-void ec_code_intel_op_mov_sse2sse(ec_code_builder_t *builder, uint32_t src,
- uint32_t dst);
-void ec_code_intel_op_mov_sse2m(ec_code_builder_t *builder, uint32_t src,
- ec_code_intel_reg_t base,
- ec_code_intel_reg_t index, uint32_t scale,
- int32_t offset);
-void ec_code_intel_op_mov_m2sse(ec_code_builder_t *builder,
- ec_code_intel_reg_t base,
- ec_code_intel_reg_t index, uint32_t scale,
- int32_t offset, uint32_t dst);
-void ec_code_intel_op_xor_sse2sse(ec_code_builder_t *builder, uint32_t src,
- uint32_t dst);
-void ec_code_intel_op_xor_m2sse(ec_code_builder_t *builder,
- ec_code_intel_reg_t base,
- ec_code_intel_reg_t index, uint32_t scale,
- int32_t offset, uint32_t dst);
-
-void ec_code_intel_op_mov_avx2avx(ec_code_builder_t *builder, uint32_t src,
- uint32_t dst);
-void ec_code_intel_op_mov_avx2m(ec_code_builder_t *builder, uint32_t src,
- ec_code_intel_reg_t base,
- ec_code_intel_reg_t index, uint32_t scale,
- int32_t offset);
-void ec_code_intel_op_mov_m2avx(ec_code_builder_t *builder,
- ec_code_intel_reg_t base,
- ec_code_intel_reg_t index, uint32_t scale,
- int32_t offset, uint32_t dst);
-void ec_code_intel_op_xor_avx2avx(ec_code_builder_t *builder, uint32_t src,
- uint32_t dst);
-void ec_code_intel_op_xor_m2avx(ec_code_builder_t *builder,
- ec_code_intel_reg_t base,
- ec_code_intel_reg_t index, uint32_t scale,
- int32_t offset, uint32_t dst);
+void
+ec_code_intel_op_push_r(ec_code_builder_t *builder, ec_code_intel_reg_t reg);
+void
+ec_code_intel_op_pop_r(ec_code_builder_t *builder, ec_code_intel_reg_t reg);
+void
+ec_code_intel_op_ret(ec_code_builder_t *builder, uint32_t size);
+
+void
+ec_code_intel_op_mov_r2r(ec_code_builder_t *builder, ec_code_intel_reg_t src,
+ ec_code_intel_reg_t dst);
+void
+ec_code_intel_op_mov_r2m(ec_code_builder_t *builder, ec_code_intel_reg_t src,
+ ec_code_intel_reg_t base, ec_code_intel_reg_t index,
+ uint32_t scale, int32_t offset);
+void
+ec_code_intel_op_mov_m2r(ec_code_builder_t *builder, ec_code_intel_reg_t base,
+ ec_code_intel_reg_t index, uint32_t scale,
+ int32_t offset, ec_code_intel_reg_t dst);
+void
+ec_code_intel_op_xor_r2r(ec_code_builder_t *builder, ec_code_intel_reg_t src,
+ ec_code_intel_reg_t dst);
+void
+ec_code_intel_op_xor_m2r(ec_code_builder_t *builder, ec_code_intel_reg_t base,
+ ec_code_intel_reg_t index, uint32_t scale,
+ int32_t offset, ec_code_intel_reg_t dst);
+void
+ec_code_intel_op_add_i2r(ec_code_builder_t *builder, int32_t value,
+ ec_code_intel_reg_t reg);
+void
+ec_code_intel_op_test_i2r(ec_code_builder_t *builder, uint32_t value,
+ ec_code_intel_reg_t reg);
+void
+ec_code_intel_op_jne(ec_code_builder_t *builder, uint32_t address);
+
+void
+ec_code_intel_op_mov_sse2sse(ec_code_builder_t *builder, uint32_t src,
+ uint32_t dst);
+void
+ec_code_intel_op_mov_sse2m(ec_code_builder_t *builder, uint32_t src,
+ ec_code_intel_reg_t base, ec_code_intel_reg_t index,
+ uint32_t scale, int32_t offset);
+void
+ec_code_intel_op_mov_m2sse(ec_code_builder_t *builder, ec_code_intel_reg_t base,
+ ec_code_intel_reg_t index, uint32_t scale,
+ int32_t offset, uint32_t dst);
+void
+ec_code_intel_op_xor_sse2sse(ec_code_builder_t *builder, uint32_t src,
+ uint32_t dst);
+void
+ec_code_intel_op_xor_m2sse(ec_code_builder_t *builder, ec_code_intel_reg_t base,
+ ec_code_intel_reg_t index, uint32_t scale,
+ int32_t offset, uint32_t dst);
+
+void
+ec_code_intel_op_mov_avx2avx(ec_code_builder_t *builder, uint32_t src,
+ uint32_t dst);
+void
+ec_code_intel_op_mov_avx2m(ec_code_builder_t *builder, uint32_t src,
+ ec_code_intel_reg_t base, ec_code_intel_reg_t index,
+ uint32_t scale, int32_t offset);
+void
+ec_code_intel_op_mov_m2avx(ec_code_builder_t *builder, ec_code_intel_reg_t base,
+ ec_code_intel_reg_t index, uint32_t scale,
+ int32_t offset, uint32_t dst);
+void
+ec_code_intel_op_xor_avx2avx(ec_code_builder_t *builder, uint32_t src,
+ uint32_t dst);
+void
+ec_code_intel_op_xor_m2avx(ec_code_builder_t *builder, ec_code_intel_reg_t base,
+ ec_code_intel_reg_t index, uint32_t scale,
+ int32_t offset, uint32_t dst);
#endif /* __EC_CODE_INTEL_H__ */
diff --git a/xlators/cluster/ec/src/ec-code-sse.c b/xlators/cluster/ec/src/ec-code-sse.c
index 6f2c6fa593f..e11e7ff8400 100644
--- a/xlators/cluster/ec/src/ec-code-sse.c
+++ b/xlators/cluster/ec/src/ec-code-sse.c
@@ -34,10 +34,9 @@ ec_code_sse_load(ec_code_builder_t *builder, uint32_t dst, uint32_t idx,
uint32_t bit)
{
if (builder->linear) {
- ec_code_intel_op_mov_m2sse(builder, REG_SI, REG_DX, 1,
- idx * builder->width * builder->bits +
- bit * builder->width,
- dst);
+ ec_code_intel_op_mov_m2sse(
+ builder, REG_SI, REG_DX, 1,
+ idx * builder->width * builder->bits + bit * builder->width, dst);
} else {
if (builder->base != idx) {
ec_code_intel_op_mov_m2r(builder, REG_SI, REG_NULL, 0, idx * 8,
@@ -73,10 +72,9 @@ ec_code_sse_xorm(ec_code_builder_t *builder, uint32_t dst, uint32_t idx,
uint32_t bit)
{
if (builder->linear) {
- ec_code_intel_op_xor_m2sse(builder, REG_SI, REG_DX, 1,
- idx * builder->width * builder->bits +
- bit * builder->width,
- dst);
+ ec_code_intel_op_xor_m2sse(
+ builder, REG_SI, REG_DX, 1,
+ idx * builder->width * builder->bits + bit * builder->width, dst);
} else {
if (builder->base != idx) {
ec_code_intel_op_mov_m2r(builder, REG_SI, REG_NULL, 0, idx * 8,
@@ -88,21 +86,16 @@ ec_code_sse_xorm(ec_code_builder_t *builder, uint32_t dst, uint32_t idx,
}
}
-static char *ec_code_sse_needed_flags[] = {
- "sse2",
- NULL
-};
+static char *ec_code_sse_needed_flags[] = {"sse2", NULL};
-ec_code_gen_t ec_code_gen_sse = {
- .name = "sse",
- .flags = ec_code_sse_needed_flags,
- .width = 16,
- .prolog = ec_code_sse_prolog,
- .epilog = ec_code_sse_epilog,
- .load = ec_code_sse_load,
- .store = ec_code_sse_store,
- .copy = ec_code_sse_copy,
- .xor2 = ec_code_sse_xor2,
- .xor3 = NULL,
- .xorm = ec_code_sse_xorm
-};
+ec_code_gen_t ec_code_gen_sse = {.name = "sse",
+ .flags = ec_code_sse_needed_flags,
+ .width = 16,
+ .prolog = ec_code_sse_prolog,
+ .epilog = ec_code_sse_epilog,
+ .load = ec_code_sse_load,
+ .store = ec_code_sse_store,
+ .copy = ec_code_sse_copy,
+ .xor2 = ec_code_sse_xor2,
+ .xor3 = NULL,
+ .xorm = ec_code_sse_xorm};
diff --git a/xlators/cluster/ec/src/ec-code-x64.c b/xlators/cluster/ec/src/ec-code-x64.c
index e94ddd4b155..26565b4493f 100644
--- a/xlators/cluster/ec/src/ec-code-x64.c
+++ b/xlators/cluster/ec/src/ec-code-x64.c
@@ -14,8 +14,7 @@
static ec_code_intel_reg_t ec_code_x64_regmap[] = {
REG_AX, REG_CX, REG_BP, REG_8, REG_9, REG_10,
- REG_11, REG_12, REG_13, REG_14, REG_15
-};
+ REG_11, REG_12, REG_13, REG_14, REG_15};
static void
ec_code_x64_prolog(ec_code_builder_t *builder)
@@ -49,6 +48,7 @@ ec_code_x64_epilog(ec_code_builder_t *builder)
if (builder->regs > 11) {
ec_code_error(builder, EINVAL);
+ return;
}
for (i = builder->regs; i > 7; i--) {
ec_code_intel_op_pop_r(builder, ec_code_x64_regmap[i - 1]);
@@ -67,10 +67,9 @@ ec_code_x64_load(ec_code_builder_t *builder, uint32_t dst, uint32_t idx,
dst = ec_code_x64_regmap[dst];
if (builder->linear) {
- ec_code_intel_op_mov_m2r(builder, REG_SI, REG_DX, 1,
- idx * builder->width * builder->bits +
- bit * builder->width,
- dst);
+ ec_code_intel_op_mov_m2r(
+ builder, REG_SI, REG_DX, 1,
+ idx * builder->width * builder->bits + bit * builder->width, dst);
} else {
if (builder->base != idx) {
ec_code_intel_op_mov_m2r(builder, REG_SI, REG_NULL, 0, idx * 8,
@@ -116,10 +115,9 @@ ec_code_x64_xorm(ec_code_builder_t *builder, uint32_t dst, uint32_t idx,
dst = ec_code_x64_regmap[dst];
if (builder->linear) {
- ec_code_intel_op_xor_m2r(builder, REG_SI, REG_DX, 1,
- idx * builder->width * builder->bits +
- bit * builder->width,
- dst);
+ ec_code_intel_op_xor_m2r(
+ builder, REG_SI, REG_DX, 1,
+ idx * builder->width * builder->bits + bit * builder->width, dst);
} else {
if (builder->base != idx) {
ec_code_intel_op_mov_m2r(builder, REG_SI, REG_NULL, 0, idx * 8,
@@ -131,20 +129,16 @@ ec_code_x64_xorm(ec_code_builder_t *builder, uint32_t dst, uint32_t idx,
}
}
-static char *ec_code_x64_needed_flags[] = {
- NULL
-};
-
-ec_code_gen_t ec_code_gen_x64 = {
- .name = "x64",
- .flags = ec_code_x64_needed_flags,
- .width = sizeof(uint64_t),
- .prolog = ec_code_x64_prolog,
- .epilog = ec_code_x64_epilog,
- .load = ec_code_x64_load,
- .store = ec_code_x64_store,
- .copy = ec_code_x64_copy,
- .xor2 = ec_code_x64_xor2,
- .xor3 = NULL,
- .xorm = ec_code_x64_xorm
-};
+static char *ec_code_x64_needed_flags[] = {NULL};
+
+ec_code_gen_t ec_code_gen_x64 = {.name = "x64",
+ .flags = ec_code_x64_needed_flags,
+ .width = sizeof(uint64_t),
+ .prolog = ec_code_x64_prolog,
+ .epilog = ec_code_x64_epilog,
+ .load = ec_code_x64_load,
+ .store = ec_code_x64_store,
+ .copy = ec_code_x64_copy,
+ .xor2 = ec_code_x64_xor2,
+ .xor3 = NULL,
+ .xorm = ec_code_x64_xorm};
diff --git a/xlators/cluster/ec/src/ec-code.c b/xlators/cluster/ec/src/ec-code.c
index 44957dd788f..03162ae05a9 100644
--- a/xlators/cluster/ec/src/ec-code.c
+++ b/xlators/cluster/ec/src/ec-code.c
@@ -14,7 +14,7 @@
#include <sys/stat.h>
#include <ctype.h>
-#include "syscall.h"
+#include <glusterfs/syscall.h>
#include "ec-mem-types.h"
#include "ec-code.h"
@@ -47,13 +47,13 @@ struct _ec_code_proc;
typedef struct _ec_code_proc ec_code_proc_t;
struct _ec_code_proc {
- int32_t fd;
+ int32_t fd;
gf_boolean_t eof;
gf_boolean_t error;
gf_boolean_t skip;
- ssize_t size;
- ssize_t pos;
- char buffer[EC_PROC_BUFFER_SIZE];
+ ssize_t size;
+ ssize_t pos;
+ char buffer[EC_PROC_BUFFER_SIZE];
};
static ec_code_gen_t *ec_code_gen_table[] = {
@@ -66,8 +66,7 @@ static ec_code_gen_t *ec_code_gen_table[] = {
#ifdef USE_EC_DYNAMIC_X64
&ec_code_gen_x64,
#endif
- NULL
-};
+ NULL};
static void
ec_code_arg_set(ec_code_arg_t *arg, uint32_t value)
@@ -84,7 +83,6 @@ ec_code_arg_assign(ec_code_builder_t *builder, ec_code_op_t *op,
if (builder->regs <= reg) {
builder->regs = reg + 1;
}
-
}
static void
@@ -202,17 +200,17 @@ static void
ec_code_dup(ec_code_builder_t *builder, ec_gf_op_t *op)
{
switch (op->op) {
- case EC_GF_OP_COPY:
- ec_code_copy(builder, op->arg1, op->arg2);
- break;
- case EC_GF_OP_XOR2:
- ec_code_xor2(builder, op->arg1, op->arg2);
- break;
- case EC_GF_OP_XOR3:
- ec_code_xor3(builder, op->arg1, op->arg2, op->arg3);
- break;
- default:
- break;
+ case EC_GF_OP_COPY:
+ ec_code_copy(builder, op->arg1, op->arg2);
+ break;
+ case EC_GF_OP_XOR2:
+ ec_code_xor2(builder, op->arg1, op->arg2);
+ break;
+ case EC_GF_OP_XOR3:
+ ec_code_xor3(builder, op->arg1, op->arg2, op->arg3);
+ break;
+ default:
+ break;
}
}
@@ -285,8 +283,9 @@ ec_code_prepare(ec_code_t *code, uint32_t count, uint32_t width,
count *= code->gf->bits + code->gf->max_ops;
count += code->gf->bits;
- builder = GF_MALLOC(sizeof(ec_code_builder_t) +
- sizeof(ec_code_op_t) * count, ec_mt_ec_code_builder_t);
+ builder = GF_MALLOC(
+ sizeof(ec_code_builder_t) + sizeof(ec_code_op_t) * count,
+ ec_mt_ec_code_builder_t);
if (builder == NULL) {
return EC_ERR(ENOMEM);
}
@@ -331,15 +330,15 @@ ec_code_chunk_from_space(ec_code_space_t *space)
static void *
ec_code_to_executable(ec_code_space_t *space, void *addr)
{
- return (void *)((uintptr_t)addr - (uintptr_t)space
- + (uintptr_t)space->exec);
+ return (void *)((uintptr_t)addr - (uintptr_t)space +
+ (uintptr_t)space->exec);
}
static void *
ec_code_from_executable(ec_code_space_t *space, void *addr)
{
- return (void *)((uintptr_t)addr - (uintptr_t)space->exec
- + (uintptr_t)space);
+ return (void *)((uintptr_t)addr - (uintptr_t)space->exec +
+ (uintptr_t)space);
}
static void *
@@ -395,104 +394,105 @@ ec_code_chunk_touch(ec_code_chunk_t *prev, ec_code_chunk_t *next)
static ec_code_space_t *
ec_code_space_create(ec_code_t *code, size_t size)
{
- char path[] = GLUSTERFS_LIBEXECDIR "/ec-code-dynamic.XXXXXX";
- ec_code_space_t *space;
- void *exec;
- int32_t fd, err;
-
- /* We need to create memory areas to store the generated dynamic code.
- * Obviously these areas need to be written to be able to create the
- * code and they also need to be executable to execute it.
- *
- * However it's a bad practice to have a memory region that is both
- * writable *and* executable. In fact, selinux forbids this and causes
- * attempts to do so to fail (unless specifically configured).
- *
- * To solve the problem we'll use two distinct memory areas mapped to
- * the same physical storage. One of the memory areas will have write
- * permission, and the other will have execute permission. Both areas
- * will have the same contents. The physical storage will be a regular
- * file that will be mmapped to both areas.
- */
-
- /* We need to create a temporary file as the backend storage for the
- * memory mapped areas. */
- fd = mkstemp(path);
- if (fd < 0) {
- err = errno;
- gf_msg(THIS->name, GF_LOG_ERROR, err, EC_MSG_DYN_CREATE_FAILED,
- "Unable to create a temporary file for the ec dynamic "
- "code");
- space = EC_ERR(err);
- goto done;
- }
- /* Once created we don't need to keep it in the file system. It will
- * still exist until we close the last file descriptor or unmap the
- * memory areas bound to the file. */
- sys_unlink(path);
-
- size = (size + EC_CODE_ALIGN - 1) & ~(EC_CODE_ALIGN - 1);
- if (sys_ftruncate(fd, size) < 0) {
- err = errno;
- gf_msg(THIS->name, GF_LOG_ERROR, err, EC_MSG_DYN_CREATE_FAILED,
- "Unable to resize the file for the ec dynamic code");
- space = EC_ERR(err);
- goto done_close;
- }
+ char path[] = GLUSTERFS_LIBEXECDIR "/ec-code-dynamic.XXXXXX";
+ ec_code_space_t *space;
+ void *exec;
+ int32_t fd, err;
+
+ /* We need to create memory areas to store the generated dynamic code.
+ * Obviously these areas need to be written to be able to create the
+ * code and they also need to be executable to execute it.
+ *
+ * However it's a bad practice to have a memory region that is both
+ * writable *and* executable. In fact, selinux forbids this and causes
+ * attempts to do so to fail (unless specifically configured).
+ *
+ * To solve the problem we'll use two distinct memory areas mapped to
+ * the same physical storage. One of the memory areas will have write
+ * permission, and the other will have execute permission. Both areas
+ * will have the same contents. The physical storage will be a regular
+ * file that will be mmapped to both areas.
+ */
+
+ /* We need to create a temporary file as the backend storage for the
+ * memory mapped areas. */
+ /* coverity[secure_temp] mkstemp uses 0600 as the mode and is safe */
+ fd = mkstemp(path);
+ if (fd < 0) {
+ err = errno;
+ gf_msg(THIS->name, GF_LOG_ERROR, err, EC_MSG_DYN_CREATE_FAILED,
+ "Unable to create a temporary file for the ec dynamic "
+ "code");
+ space = EC_ERR(err);
+ goto done;
+ }
+ /* Once created we don't need to keep it in the file system. It will
+ * still exist until we close the last file descriptor or unmap the
+ * memory areas bound to the file. */
+ sys_unlink(path);
+
+ size = (size + EC_CODE_ALIGN - 1) & ~(EC_CODE_ALIGN - 1);
+ if (sys_ftruncate(fd, size) < 0) {
+ err = errno;
+ gf_msg(THIS->name, GF_LOG_ERROR, err, EC_MSG_DYN_CREATE_FAILED,
+ "Unable to resize the file for the ec dynamic code");
+ space = EC_ERR(err);
+ goto done_close;
+ }
- /* This creates an executable memory area to be able to run the
- * generated fragments of code. */
- exec = mmap(NULL, size, PROT_READ | PROT_EXEC, MAP_SHARED, fd, 0);
- if (exec == MAP_FAILED) {
- err = errno;
- gf_msg(THIS->name, GF_LOG_ERROR, err, EC_MSG_DYN_CREATE_FAILED,
- "Unable to map the executable area for the ec dynamic "
- "code");
- space = EC_ERR(err);
- goto done_close;
- }
- /* It's not important to check the return value of mlock(). If it fails
- * everything will continue to work normally. */
- mlock(exec, size);
-
- /* This maps a read/write memory area to be able to create the dynamici
- * code. */
- space = mmap(NULL, size, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
- if (space == MAP_FAILED) {
- err = errno;
- gf_msg(THIS->name, GF_LOG_ERROR, err, EC_MSG_DYN_CREATE_FAILED,
- "Unable to map the writable area for the ec dynamic "
- "code");
- space = EC_ERR(err);
-
- munmap(exec, size);
-
- goto done_close;
- }
+ /* This creates an executable memory area to be able to run the
+ * generated fragments of code. */
+ exec = mmap(NULL, size, PROT_READ | PROT_EXEC, MAP_SHARED, fd, 0);
+ if (exec == MAP_FAILED) {
+ err = errno;
+ gf_msg(THIS->name, GF_LOG_ERROR, err, EC_MSG_DYN_CREATE_FAILED,
+ "Unable to map the executable area for the ec dynamic "
+ "code");
+ space = EC_ERR(err);
+ goto done_close;
+ }
+ /* It's not important to check the return value of mlock(). If it fails
+ * everything will continue to work normally. */
+ mlock(exec, size);
+
+ /* This maps a read/write memory area to be able to create the dynamici
+ * code. */
+ space = mmap(NULL, size, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
+ if (space == MAP_FAILED) {
+ err = errno;
+ gf_msg(THIS->name, GF_LOG_ERROR, err, EC_MSG_DYN_CREATE_FAILED,
+ "Unable to map the writable area for the ec dynamic "
+ "code");
+ space = EC_ERR(err);
+
+ munmap(exec, size);
+
+ goto done_close;
+ }
- space->exec = exec;
- space->size = size;
- space->code = code;
- list_add_tail(&space->list, &code->spaces);
- INIT_LIST_HEAD(&space->chunks);
+ space->exec = exec;
+ space->size = size;
+ space->code = code;
+ list_add_tail(&space->list, &code->spaces);
+ INIT_LIST_HEAD(&space->chunks);
done_close:
- /* If everything has succeeded, we already have the memory areas
- * mapped. We don't need the file descriptor anymore because the
- * backend storage will be there until the mmaped regions are
- * unmapped. */
- sys_close(fd);
+ /* If everything has succeeded, we already have the memory areas
+ * mapped. We don't need the file descriptor anymore because the
+ * backend storage will be there until the mmap()'d regions are
+ * unmapped. */
+ sys_close(fd);
done:
- return space;
+ return space;
}
static void
ec_code_space_destroy(ec_code_space_t *space)
{
- list_del_init(&space->list);
+ list_del_init(&space->list);
- munmap(space->exec, space->size);
- munmap(space, space->size);
+ munmap(space->exec, space->size);
+ munmap(space, space->size);
}
static void
@@ -500,7 +500,8 @@ ec_code_chunk_merge(ec_code_chunk_t *chunk)
{
ec_code_chunk_t *item, *tmp;
- list_for_each_entry_safe(item, tmp, &chunk->space->chunks, list) {
+ list_for_each_entry_safe(item, tmp, &chunk->space->chunks, list)
+ {
if ((uintptr_t)item > (uintptr_t)chunk) {
list_add_tail(&chunk->list, &item->list);
if (ec_code_chunk_touch(chunk, item)) {
@@ -519,8 +520,8 @@ ec_code_chunk_merge(ec_code_chunk_t *chunk)
list_add_tail(&chunk->list, &chunk->space->chunks);
check:
- if (chunk->size == chunk->space->size - ec_code_space_size() -
- ec_code_chunk_size()) {
+ if (chunk->size ==
+ chunk->space->size - ec_code_space_size() - ec_code_chunk_size()) {
ec_code_space_destroy(chunk->space);
}
}
@@ -535,9 +536,12 @@ ec_code_space_alloc(ec_code_t *code, size_t size)
/* To minimize fragmentation, we only allocate chunks of sizes multiples
* of EC_CODE_CHUNK_MIN_SIZE. */
size = ((size + ec_code_chunk_size() + EC_CODE_CHUNK_MIN_SIZE - 1) &
- ~(EC_CODE_CHUNK_MIN_SIZE - 1)) - ec_code_chunk_size();
- list_for_each_entry(space, &code->spaces, list) {
- list_for_each_entry(chunk, &space->chunks, list) {
+ ~(EC_CODE_CHUNK_MIN_SIZE - 1)) -
+ ec_code_chunk_size();
+ list_for_each_entry(space, &code->spaces, list)
+ {
+ list_for_each_entry(chunk, &space->chunks, list)
+ {
if (chunk->size >= size) {
goto out;
}
@@ -607,26 +611,29 @@ ec_code_write(ec_code_builder_t *builder)
for (i = 0; i < builder->count; i++) {
op = &builder->ops[i];
switch (op->op) {
- case EC_GF_OP_LOAD:
- gen->load(builder, op->arg1.value, op->arg2.value, op->arg3.value);
- break;
- case EC_GF_OP_STORE:
- gen->store(builder, op->arg1.value, op->arg3.value);
- break;
- case EC_GF_OP_COPY:
- gen->copy(builder, op->arg1.value, op->arg2.value);
- break;
- case EC_GF_OP_XOR2:
- gen->xor2(builder, op->arg1.value, op->arg2.value);
- break;
- case EC_GF_OP_XOR3:
- gen->xor3(builder, op->arg1.value, op->arg2.value, op->arg3.value);
- break;
- case EC_GF_OP_XORM:
- gen->xorm(builder, op->arg1.value, op->arg2.value, op->arg3.value);
- break;
- default:
- break;
+ case EC_GF_OP_LOAD:
+ gen->load(builder, op->arg1.value, op->arg2.value,
+ op->arg3.value);
+ break;
+ case EC_GF_OP_STORE:
+ gen->store(builder, op->arg1.value, op->arg3.value);
+ break;
+ case EC_GF_OP_COPY:
+ gen->copy(builder, op->arg1.value, op->arg2.value);
+ break;
+ case EC_GF_OP_XOR2:
+ gen->xor2(builder, op->arg1.value, op->arg2.value);
+ break;
+ case EC_GF_OP_XOR3:
+ gen->xor3(builder, op->arg1.value, op->arg2.value,
+ op->arg3.value);
+ break;
+ case EC_GF_OP_XORM:
+ gen->xorm(builder, op->arg1.value, op->arg2.value,
+ op->arg3.value);
+ break;
+ default:
+ break;
}
}
gen->epilog(builder);
@@ -715,67 +722,65 @@ static void *
ec_code_build_dynamic(ec_code_t *code, uint32_t width, uint32_t *values,
uint32_t count, gf_boolean_t linear)
{
- ec_code_builder_t *builder;
- uint32_t offset, val, next;
+ ec_code_builder_t *builder;
+ uint32_t offset, val, next;
- builder = ec_code_prepare(code, count, width, linear);
- if (EC_IS_ERR(builder)) {
- return builder;
- }
+ builder = ec_code_prepare(code, count, width, linear);
+ if (EC_IS_ERR(builder)) {
+ return builder;
+ }
- offset = -1;
- next = ec_code_value_next(values, count, &offset);
- if (next != 0) {
- ec_code_gf_load(builder, offset);
- do {
- val = next;
- next = ec_code_value_next(values, count, &offset);
- if (next != 0) {
- ec_code_gf_mul(builder, ec_gf_div(code->gf,
- val, next));
- ec_code_gf_load_xor(builder, offset);
- }
- } while (next != 0);
- ec_code_gf_mul(builder, val);
- ec_code_gf_store(builder);
- } else {
- ec_code_gf_clear(builder);
- }
+ offset = -1;
+ next = ec_code_value_next(values, count, &offset);
+ if (next != 0) {
+ ec_code_gf_load(builder, offset);
+ do {
+ val = next;
+ next = ec_code_value_next(values, count, &offset);
+ if (next != 0) {
+ ec_code_gf_mul(builder, ec_gf_div(code->gf, val, next));
+ ec_code_gf_load_xor(builder, offset);
+ }
+ } while (next != 0);
+ ec_code_gf_mul(builder, val);
+ ec_code_gf_store(builder);
+ } else {
+ ec_code_gf_clear(builder);
+ }
- return ec_code_compile(builder);
+ return ec_code_compile(builder);
}
static void *
-ec_code_build(ec_code_t *code, uint32_t width, uint32_t *values,
- uint32_t count, gf_boolean_t linear)
+ec_code_build(ec_code_t *code, uint32_t width, uint32_t *values, uint32_t count,
+ gf_boolean_t linear)
{
- void *func;
+ void *func;
- if (code->gen != NULL) {
- func = ec_code_build_dynamic(code, width, values, count,
- linear);
- if (!EC_IS_ERR(func)) {
- return func;
- }
+ if (code->gen != NULL) {
+ func = ec_code_build_dynamic(code, width, values, count, linear);
+ if (!EC_IS_ERR(func)) {
+ return func;
+ }
- gf_msg_debug(THIS->name, GF_LOG_DEBUG,
- "Unable to generate dynamic code. Falling back "
- "to precompiled code");
+ gf_msg_debug(THIS->name, GF_LOG_DEBUG,
+ "Unable to generate dynamic code. Falling back "
+ "to precompiled code");
- /* The dynamic code generation shouldn't fail in normal
- * conditions, but if it fails at some point, it's very
- * probable that it will fail again, so we completely disable
- * dynamic code generation. */
- code->gen = NULL;
- }
+ /* The dynamic code generation shouldn't fail in normal
+ * conditions, but if it fails at some point, it's very
+ * probable that it will fail again, so we completely disable
+ * dynamic code generation. */
+ code->gen = NULL;
+ }
- ec_code_c_prepare(code->gf, values, count);
+ ec_code_c_prepare(code->gf, values, count);
- if (linear) {
- return ec_code_c_linear;
- }
+ if (linear) {
+ return ec_code_c_linear;
+ }
- return ec_code_c_interleaved;
+ return ec_code_c_interleaved;
}
ec_code_func_linear_t
@@ -790,17 +795,17 @@ ec_code_func_interleaved_t
ec_code_build_interleaved(ec_code_t *code, uint32_t width, uint32_t *values,
uint32_t count)
{
- return (ec_code_func_interleaved_t)ec_code_build(code, width, values,
- count, _gf_false);
+ return (ec_code_func_interleaved_t)ec_code_build(code, width, values, count,
+ _gf_false);
}
void
ec_code_release(ec_code_t *code, ec_code_func_t *func)
{
- if ((func->linear != ec_code_c_linear) &&
- (func->interleaved != ec_code_c_interleaved)) {
- ec_code_free(ec_code_chunk_from_func(func->linear));
- }
+ if ((func->linear != ec_code_c_linear) &&
+ (func->interleaved != ec_code_c_interleaved)) {
+ ec_code_free(ec_code_chunk_from_func(func->linear));
+ }
}
void
@@ -939,7 +944,7 @@ ec_code_cpu_check(uint32_t idx, char *list, uint32_t count)
{
ec_code_gen_t *gen;
char **ptr;
- char *table[count];
+ char *table[count + 1];
uint32_t i;
for (i = 0; i < count; i++) {
@@ -1002,7 +1007,8 @@ ec_code_detect(xlator_t *xl, const char *def)
if (ec_code_gen_table[select] == NULL) {
gf_msg(xl->name, GF_LOG_WARNING, EINVAL, EC_MSG_EXTENSION_UNKNOWN,
"CPU extension '%s' is not known. Not using any cpu "
- "extensions", def);
+ "extensions",
+ def);
return NULL;
}
diff --git a/xlators/cluster/ec/src/ec-code.h b/xlators/cluster/ec/src/ec-code.h
index 355209c3944..75fb35d93e3 100644
--- a/xlators/cluster/ec/src/ec-code.h
+++ b/xlators/cluster/ec/src/ec-code.h
@@ -11,8 +11,8 @@
#ifndef __EC_CODE_H__
#define __EC_CODE_H__
-#include "xlator.h"
-#include "list.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/list.h>
#include "ec-types.h"
#include "ec-galois.h"
diff --git a/xlators/cluster/ec/src/ec-combine.c b/xlators/cluster/ec/src/ec-combine.c
index b51ec87a934..703a30e2485 100644
--- a/xlators/cluster/ec/src/ec-combine.c
+++ b/xlators/cluster/ec/src/ec-combine.c
@@ -11,14 +11,14 @@
#include <fnmatch.h>
#include "libxlator.h"
-#include "byte-order.h"
+#include <glusterfs/byte-order.h>
#include "ec-types.h"
#include "ec-helpers.h"
#include "ec-common.h"
#include "ec-combine.h"
#include "ec-messages.h"
-#include "quota-common-utils.h"
+#include <glusterfs/quota-common-utils.h>
#define EC_QUOTA_PREFIX "trusted.glusterfs.quota."
@@ -30,41 +30,38 @@ typedef struct _ec_dict_info ec_dict_info_t;
struct _ec_dict_combine;
typedef struct _ec_dict_combine ec_dict_combine_t;
-struct _ec_dict_info
-{
- dict_t * dict;
- int32_t count;
+struct _ec_dict_info {
+ dict_t *dict;
+ int32_t count;
};
-struct _ec_dict_combine
-{
- ec_cbk_data_t * cbk;
- int32_t which;
+struct _ec_dict_combine {
+ ec_cbk_data_t *cbk;
+ int32_t which;
};
int32_t
-ec_combine_write (ec_fop_data_t *fop, ec_cbk_data_t *dst,
- ec_cbk_data_t *src)
+ec_combine_write(ec_fop_data_t *fop, ec_cbk_data_t *dst, ec_cbk_data_t *src)
{
- int valid = 0;
+ int valid = 0;
- if (!fop || !dst || !src)
- return 0;
+ if (!fop || !dst || !src)
+ return 0;
- switch (fop->id) {
+ switch (fop->id) {
case GF_FOP_REMOVEXATTR:
case GF_FOP_FREMOVEXATTR:
case GF_FOP_SETXATTR:
case GF_FOP_FSETXATTR:
- return 1;
+ return 1;
case GF_FOP_SYMLINK:
case GF_FOP_LINK:
case GF_FOP_CREATE:
case GF_FOP_MKNOD:
case GF_FOP_MKDIR:
- valid = 3;
- break;
+ valid = 3;
+ break;
case GF_FOP_UNLINK:
case GF_FOP_RMDIR:
case GF_FOP_SETATTR:
@@ -75,42 +72,40 @@ ec_combine_write (ec_fop_data_t *fop, ec_cbk_data_t *dst,
case GF_FOP_FALLOCATE:
case GF_FOP_DISCARD:
case GF_FOP_ZEROFILL:
- valid = 2;
- break;
+ valid = 2;
+ break;
case GF_FOP_RENAME:
- valid = 5;
- break;
+ valid = 5;
+ break;
default:
- gf_msg_callingfn (fop->xl->name, GF_LOG_WARNING, EINVAL,
- EC_MSG_INVALID_FOP,
- "Invalid fop %d", fop->id);
- return 0;
- break;
- }
+ gf_msg_callingfn(fop->xl->name, GF_LOG_WARNING, EINVAL,
+ EC_MSG_INVALID_FOP, "Invalid fop %d", fop->id);
+ return 0;
+ break;
+ }
- if (!ec_iatt_combine(fop, dst->iatt, src->iatt, valid)) {
- gf_msg (fop->xl->name, GF_LOG_NOTICE, 0,
- EC_MSG_IATT_MISMATCH,
- "Mismatching iatt in "
- "answers of '%s'", gf_fop_list[fop->id]);
- return 0;
- }
- return 1;
+ if (!ec_iatt_combine(fop, dst->iatt, src->iatt, valid)) {
+ gf_msg(fop->xl->name, GF_LOG_NOTICE, 0, EC_MSG_IATT_MISMATCH,
+ "Mismatching iatt in "
+ "answers of '%s'",
+ gf_fop_list[fop->id]);
+ return 0;
+ }
+ return 1;
}
-void ec_iatt_time_merge(int64_t *dst_sec, uint32_t *dst_nsec,
- int64_t src_sec, uint32_t src_nsec)
+void
+ec_iatt_time_merge(int64_t *dst_sec, uint32_t *dst_nsec, int64_t src_sec,
+ uint32_t src_nsec)
{
if ((*dst_sec < src_sec) ||
- ((*dst_sec == src_sec) && (*dst_nsec < src_nsec)))
- {
+ ((*dst_sec == src_sec) && (*dst_nsec < src_nsec))) {
*dst_sec = src_sec;
*dst_nsec = src_nsec;
}
}
-static
-gf_boolean_t
+static gf_boolean_t
ec_iatt_is_trusted(ec_fop_data_t *fop, struct iatt *iatt)
{
uint64_t ino;
@@ -141,14 +136,14 @@ ec_iatt_is_trusted(ec_fop_data_t *fop, struct iatt *iatt)
return _gf_false;
}
-int32_t ec_iatt_combine(ec_fop_data_t *fop, struct iatt *dst, struct iatt *src,
- int32_t count)
+int32_t
+ec_iatt_combine(ec_fop_data_t *fop, struct iatt *dst, struct iatt *src,
+ int32_t count)
{
int32_t i;
gf_boolean_t failed = _gf_false;
- for (i = 0; i < count; i++)
- {
+ for (i = 0; i < count; i++) {
/* Check for basic fields. These fields must be equal always, even if
* the inode is not locked because in these cases the parent inode
* will be locked and differences in these fields require changes in
@@ -172,33 +167,34 @@ int32_t ec_iatt_combine(ec_fop_data_t *fop, struct iatt *dst, struct iatt *src,
* data is returned. */
failed = _gf_true;
} else {
- gf_msg_debug (fop->xl->name, 0,
- "Ignoring iatt differences because inode is not "
- "locked");
+ gf_msg_debug(fop->xl->name, 0,
+ "Ignoring iatt differences because inode is not "
+ "locked");
}
}
if (failed) {
- gf_msg (fop->xl->name, GF_LOG_WARNING, 0,
- EC_MSG_IATT_COMBINE_FAIL,
- "Failed to combine iatt (inode: %lu-%lu, links: %u-%u, "
- "uid: %u-%u, gid: %u-%u, rdev: %lu-%lu, size: %lu-%lu, "
- "mode: %o-%o)",
- dst[i].ia_ino, src[i].ia_ino, dst[i].ia_nlink,
- src[i].ia_nlink, dst[i].ia_uid, src[i].ia_uid,
- dst[i].ia_gid, src[i].ia_gid, dst[i].ia_rdev,
- src[i].ia_rdev, dst[i].ia_size, src[i].ia_size,
- st_mode_from_ia(dst[i].ia_prot, dst[i].ia_type),
- st_mode_from_ia(src[i].ia_prot, dst[i].ia_type));
+ gf_msg(fop->xl->name, GF_LOG_WARNING, 0, EC_MSG_IATT_COMBINE_FAIL,
+ "Failed to combine iatt (inode: %" PRIu64 "-%" PRIu64
+ ", "
+ "links: %u-%u, uid: %u-%u, gid: %u-%u, "
+ "rdev: %" PRIu64 "-%" PRIu64 ", size: %" PRIu64 "-%" PRIu64
+ ", "
+ "mode: %o-%o), %s",
+ dst[i].ia_ino, src[i].ia_ino, dst[i].ia_nlink,
+ src[i].ia_nlink, dst[i].ia_uid, src[i].ia_uid, dst[i].ia_gid,
+ src[i].ia_gid, dst[i].ia_rdev, src[i].ia_rdev,
+ dst[i].ia_size, src[i].ia_size,
+ st_mode_from_ia(dst[i].ia_prot, dst[i].ia_type),
+ st_mode_from_ia(src[i].ia_prot, dst[i].ia_type),
+ ec_msg_str(fop));
return 0;
}
}
- while (count-- > 0)
- {
+ while (count-- > 0) {
dst[count].ia_blocks += src[count].ia_blocks;
- if (dst[count].ia_blksize < src[count].ia_blksize)
- {
+ if (dst[count].ia_blksize < src[count].ia_blksize) {
dst[count].ia_blksize = src[count].ia_blksize;
}
@@ -213,13 +209,12 @@ int32_t ec_iatt_combine(ec_fop_data_t *fop, struct iatt *dst, struct iatt *src,
return 1;
}
-void ec_iatt_rebuild(ec_t * ec, struct iatt * iatt, int32_t count,
- int32_t answers)
+void
+ec_iatt_rebuild(ec_t *ec, struct iatt *iatt, int32_t count, int32_t answers)
{
uint64_t blocks;
- while (count-- > 0)
- {
+ while (count-- > 0) {
blocks = iatt[count].ia_blocks * ec->fragments + answers - 1;
blocks /= answers;
iatt[count].ia_blocks = blocks;
@@ -227,92 +222,93 @@ void ec_iatt_rebuild(ec_t * ec, struct iatt * iatt, int32_t count,
}
gf_boolean_t
-ec_xattr_match (dict_t *dict, char *key, data_t *value, void *arg)
+ec_xattr_match(dict_t *dict, char *key, data_t *value, void *arg)
{
- if ((fnmatch(GF_XATTR_STIME_PATTERN, key, 0) == 0) ||
- (strcmp(key, GET_LINK_COUNT) == 0) ||
- (strcmp(key, GLUSTERFS_INODELK_COUNT) == 0) ||
- (strcmp(key, GLUSTERFS_ENTRYLK_COUNT) == 0) ||
- (strcmp(key, GLUSTERFS_OPEN_FD_COUNT) == 0)) {
- return _gf_false;
- }
+ if ((fnmatch(GF_XATTR_STIME_PATTERN, key, 0) == 0) ||
+ (strcmp(key, GET_LINK_COUNT) == 0) ||
+ (strcmp(key, GLUSTERFS_INODELK_COUNT) == 0) ||
+ (strcmp(key, GLUSTERFS_ENTRYLK_COUNT) == 0) ||
+ (strcmp(key, GLUSTERFS_OPEN_FD_COUNT) == 0)) {
+ return _gf_false;
+ }
- return _gf_true;
+ return _gf_true;
}
gf_boolean_t
-ec_value_ignore (char *key)
+ec_value_ignore(char *key)
{
- if ((strcmp(key, GF_CONTENT_KEY) == 0) ||
- (strcmp(key, GF_XATTR_PATHINFO_KEY) == 0) ||
- (strcmp(key, GF_XATTR_USER_PATHINFO_KEY) == 0) ||
- (strcmp(key, GF_XATTR_LOCKINFO_KEY) == 0) ||
- (strcmp(key, GLUSTERFS_OPEN_FD_COUNT) == 0) ||
- (strcmp(key, GLUSTERFS_INODELK_COUNT) == 0) ||
- (strcmp(key, GLUSTERFS_ENTRYLK_COUNT) == 0) ||
- (strncmp(key, GF_XATTR_CLRLK_CMD,
- strlen (GF_XATTR_CLRLK_CMD)) == 0) ||
- (strcmp(key, DHT_IATT_IN_XDATA_KEY) == 0) ||
- (strncmp(key, EC_QUOTA_PREFIX, strlen(EC_QUOTA_PREFIX)) == 0) ||
- (fnmatch(MARKER_XATTR_PREFIX ".*." XTIME, key, 0) == 0) ||
- (fnmatch(GF_XATTR_MARKER_KEY ".*", key, 0) == 0) ||
- (XATTR_IS_NODE_UUID(key))) {
- return _gf_true;
- }
- return _gf_false;
+ if ((strcmp(key, GF_CONTENT_KEY) == 0) ||
+ (strcmp(key, GF_XATTR_PATHINFO_KEY) == 0) ||
+ (strcmp(key, GF_XATTR_USER_PATHINFO_KEY) == 0) ||
+ (strcmp(key, GF_XATTR_LOCKINFO_KEY) == 0) ||
+ (strcmp(key, GLUSTERFS_OPEN_FD_COUNT) == 0) ||
+ (strcmp(key, GLUSTERFS_INODELK_COUNT) == 0) ||
+ (strcmp(key, GLUSTERFS_ENTRYLK_COUNT) == 0) ||
+ (strncmp(key, GF_XATTR_CLRLK_CMD, SLEN(GF_XATTR_CLRLK_CMD)) == 0) ||
+ (strcmp(key, DHT_IATT_IN_XDATA_KEY) == 0) ||
+ (strncmp(key, EC_QUOTA_PREFIX, SLEN(EC_QUOTA_PREFIX)) == 0) ||
+ (fnmatch(MARKER_XATTR_PREFIX ".*." XTIME, key, 0) == 0) ||
+ (fnmatch(GF_XATTR_MARKER_KEY ".*", key, 0) == 0) ||
+ (XATTR_IS_NODE_UUID(key))) {
+ return _gf_true;
+ }
+
+ return _gf_false;
}
int32_t
-ec_dict_compare (dict_t *dict1, dict_t *dict2)
+ec_dict_compare(dict_t *dict1, dict_t *dict2)
{
- if (are_dicts_equal (dict1, dict2, ec_xattr_match, ec_value_ignore))
- return 1;
- return 0;
+ if (are_dicts_equal(dict1, dict2, ec_xattr_match, ec_value_ignore))
+ return 1;
+ return 0;
}
static uint32_t
ec_dict_list(data_t **list, ec_cbk_data_t *cbk, int32_t which, char *key,
gf_boolean_t global)
{
- ec_t *ec = cbk->fop->xl->private;
- ec_cbk_data_t *ans = NULL;
- dict_t *dict = NULL;
- data_t *data;
- uint32_t count;
- int32_t i;
-
- for (i = 0; i < ec->nodes; i++) {
- /* We initialize the list with EC_MISSING_DATA if we are
- * returning a global list or the current subvolume belongs
- * to the group of the accepted answer. Note that if some
- * subvolume is known to be down before issuing the request,
- * we won't have any answer from it, so we set here the
- * appropriate default value. */
- if (global || ((cbk->mask & (1ULL << i)) != 0)) {
- list[i] = EC_MISSING_DATA;
- } else {
- list[i] = NULL;
- }
+ ec_t *ec = cbk->fop->xl->private;
+ ec_cbk_data_t *ans = NULL;
+ dict_t *dict = NULL;
+ data_t *data;
+ uint32_t count;
+ int32_t i;
+
+ for (i = 0; i < ec->nodes; i++) {
+ /* We initialize the list with EC_MISSING_DATA if we are
+ * returning a global list or the current subvolume belongs
+ * to the group of the accepted answer. Note that if some
+ * subvolume is known to be down before issuing the request,
+ * we won't have any answer from it, so we set here the
+ * appropriate default value. */
+ if (global || ((cbk->mask & (1ULL << i)) != 0)) {
+ list[i] = EC_MISSING_DATA;
+ } else {
+ list[i] = NULL;
}
+ }
- count = 0;
- list_for_each_entry(ans, &cbk->fop->answer_list, answer_list) {
- if (global || ((cbk->mask & ans->mask) != 0)) {
- dict = (which == EC_COMBINE_XDATA) ? ans->xdata
- : ans->dict;
- data = dict_get(dict, key);
- if (data != NULL) {
- list[ans->idx] = data;
- count++;
- }
- }
+ count = 0;
+ list_for_each_entry(ans, &cbk->fop->answer_list, answer_list)
+ {
+ if (global || ((cbk->mask & ans->mask) != 0)) {
+ dict = (which == EC_COMBINE_XDATA) ? ans->xdata : ans->dict;
+ data = dict_get(dict, key);
+ if (data != NULL) {
+ list[ans->idx] = data;
+ count++;
+ }
}
+ }
- return count;
+ return count;
}
-int32_t ec_concat_prepare(xlator_t *xl, char **str, char **sep, char **post,
- const char *fmt, va_list args)
+int32_t
+ec_concat_prepare(xlator_t *xl, char **str, char **sep, char **post,
+ const char *fmt, va_list args)
{
char *tmp;
int32_t len;
@@ -338,9 +334,8 @@ int32_t ec_concat_prepare(xlator_t *xl, char **str, char **sep, char **post,
return 0;
out:
- gf_msg (xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_INVALID_FORMAT,
- "Invalid concat format");
+ gf_msg(xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_INVALID_FORMAT,
+ "Invalid concat format");
GF_FREE(*str);
@@ -348,9 +343,8 @@ out:
}
static int32_t
-ec_dict_data_concat(const char *fmt, ec_cbk_data_t *cbk, int32_t which,
- char *key, char *new_key, const char *def,
- gf_boolean_t global, ...)
+ec_dict_data_concat(ec_cbk_data_t *cbk, int32_t which, char *key, char *new_key,
+ const char *def, gf_boolean_t global, const char *fmt, ...)
{
ec_t *ec = cbk->fop->xl->private;
data_t *data[ec->nodes];
@@ -362,7 +356,7 @@ ec_dict_data_concat(const char *fmt, ec_cbk_data_t *cbk, int32_t which,
ec_dict_list(data, cbk, which, key, global);
- va_start(args, global);
+ va_start(args, fmt);
err = ec_concat_prepare(cbk->fop->xl, &pre, &sep, &post, fmt, args);
va_end(args);
@@ -448,7 +442,8 @@ out:
return err;
}
-int32_t ec_dict_data_merge(ec_cbk_data_t *cbk, int32_t which, char *key)
+int32_t
+ec_dict_data_merge(ec_cbk_data_t *cbk, int32_t which, char *key)
{
ec_t *ec = cbk->fop->xl->private;
data_t *data[ec->nodes];
@@ -457,7 +452,6 @@ int32_t ec_dict_data_merge(ec_cbk_data_t *cbk, int32_t which, char *key)
int32_t i, len;
int32_t err;
-
ec_dict_list(data, cbk, which, key, _gf_false);
lockinfo = dict_new();
@@ -491,22 +485,12 @@ int32_t ec_dict_data_merge(ec_cbk_data_t *cbk, int32_t which, char *key)
tmp = NULL;
- len = dict_serialized_length(lockinfo);
- if (len < 0) {
- err = len;
-
- goto out;
- }
- ptr = GF_MALLOC(len, gf_common_mt_char);
- if (ptr == NULL) {
- err = -ENOMEM;
-
- goto out;
- }
- err = dict_serialize(lockinfo, ptr);
+ err = dict_allocate_and_serialize(lockinfo, (char **)&ptr,
+ (unsigned int *)&len);
if (err != 0) {
goto out;
}
+
dict = (which == EC_COMBINE_XDATA) ? cbk->xdata : cbk->dict;
err = dict_set_dynptr(dict, key, ptr, len);
if (err != 0) {
@@ -525,11 +509,12 @@ out:
return err;
}
-int32_t ec_dict_data_uuid(ec_cbk_data_t * cbk, int32_t which, char * key)
+int32_t
+ec_dict_data_uuid(ec_cbk_data_t *cbk, int32_t which, char *key)
{
- ec_cbk_data_t * ans, * min;
- dict_t * src, * dst;
- data_t * data;
+ ec_cbk_data_t *ans, *min;
+ dict_t *src, *dst;
+ data_t *data;
min = cbk;
for (ans = cbk->next; ans != NULL; ans = ans->next) {
@@ -554,7 +539,67 @@ int32_t ec_dict_data_uuid(ec_cbk_data_t * cbk, int32_t which, char * key)
return 0;
}
-int32_t ec_dict_data_max32(ec_cbk_data_t *cbk, int32_t which, char *key)
+int32_t
+ec_dict_data_iatt(ec_cbk_data_t *cbk, int32_t which, char *key)
+{
+ ec_t *ec = cbk->fop->xl->private;
+ data_t *data[ec->nodes];
+ dict_t *dict;
+ struct iatt *stbuf, *tmp;
+ int32_t i, ret;
+
+ ec_dict_list(data, cbk, which, key, _gf_false);
+
+ stbuf = NULL;
+ for (i = 0; i < ec->nodes; i++) {
+ if ((data[i] == NULL) || (data[i] == EC_MISSING_DATA)) {
+ continue;
+ }
+ tmp = data_to_iatt(data[i], key);
+ if (tmp == NULL) {
+ ret = -EINVAL;
+ goto out;
+ }
+ if (stbuf == NULL) {
+ stbuf = GF_MALLOC(sizeof(struct iatt), gf_common_mt_char);
+ if (stbuf == NULL) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ *stbuf = *tmp;
+ } else {
+ if (!ec_iatt_combine(cbk->fop, stbuf, tmp, 1)) {
+ ret = -EINVAL;
+ goto out;
+ }
+ }
+ }
+
+ if ((stbuf != NULL) && (stbuf->ia_type == IA_IFREG)) {
+ ec_iatt_rebuild(ec, stbuf, 1, cbk->count);
+ /* TODO: not sure if an iatt could come in xdata from a fop that takes
+ * no locks. */
+ if (!ec_get_inode_size(cbk->fop, cbk->fop->locks[0].lock->loc.inode,
+ &stbuf->ia_size)) {
+ ret = -EINVAL;
+ goto out;
+ }
+ }
+
+ dict = (which == EC_COMBINE_XDATA) ? cbk->xdata : cbk->dict;
+ ret = dict_set_iatt(dict, key, stbuf, false);
+ if (ret >= 0) {
+ stbuf = NULL;
+ }
+
+out:
+ GF_FREE(stbuf);
+
+ return ret;
+}
+
+int32_t
+ec_dict_data_max32(ec_cbk_data_t *cbk, int32_t which, char *key)
{
ec_t *ec = cbk->fop->xl->private;
data_t *data[ec->nodes];
@@ -580,7 +625,8 @@ int32_t ec_dict_data_max32(ec_cbk_data_t *cbk, int32_t which, char *key)
return dict_set_uint32(dict, key, max);
}
-int32_t ec_dict_data_max64(ec_cbk_data_t *cbk, int32_t which, char *key)
+int32_t
+ec_dict_data_max64(ec_cbk_data_t *cbk, int32_t which, char *key)
{
ec_t *ec = cbk->fop->xl->private;
data_t *data[ec->nodes];
@@ -606,14 +652,19 @@ int32_t ec_dict_data_max64(ec_cbk_data_t *cbk, int32_t which, char *key)
return dict_set_uint64(dict, key, max);
}
-int32_t ec_dict_data_quota(ec_cbk_data_t *cbk, int32_t which, char *key)
+int32_t
+ec_dict_data_quota(ec_cbk_data_t *cbk, int32_t which, char *key)
{
- ec_t *ec = cbk->fop->xl->private;
- data_t *data[ec->nodes];
- dict_t *dict = NULL;
- int32_t i = 0;
- quota_meta_t size = {0, };
- quota_meta_t max_size = {0, };
+ ec_t *ec = cbk->fop->xl->private;
+ data_t *data[ec->nodes];
+ dict_t *dict = NULL;
+ int32_t i = 0;
+ quota_meta_t size = {
+ 0,
+ };
+ quota_meta_t max_size = {
+ 0,
+ };
if (ec_dict_list(data, cbk, which, key, _gf_false) == 0) {
return 0;
@@ -626,25 +677,26 @@ int32_t ec_dict_data_quota(ec_cbk_data_t *cbk, int32_t which, char *key)
*/
for (i = 0; i < ec->nodes; i++) {
if ((data[i] == NULL) || (data[i] == EC_MISSING_DATA) ||
- (quota_data_to_meta (data[i], QUOTA_SIZE_KEY, &size) < 0)) {
- continue;
+ (quota_data_to_meta(data[i], &size) < 0)) {
+ continue;
}
if (size.size > max_size.size)
- max_size.size = size.size;
+ max_size.size = size.size;
if (size.file_count > max_size.file_count)
- max_size.file_count = size.file_count;
+ max_size.file_count = size.file_count;
if (size.dir_count > max_size.dir_count)
- max_size.dir_count = size.dir_count;
+ max_size.dir_count = size.dir_count;
}
max_size.size *= ec->fragments;
dict = (which == EC_COMBINE_XDATA) ? cbk->xdata : cbk->dict;
- return quota_dict_set_meta (dict, key, &max_size, IA_IFDIR);
+ return quota_dict_set_meta(dict, key, &max_size, IA_IFDIR);
}
-int32_t ec_dict_data_stime(ec_cbk_data_t * cbk, int32_t which, char * key)
+int32_t
+ec_dict_data_stime(ec_cbk_data_t *cbk, int32_t which, char *key)
{
ec_t *ec = cbk->fop->xl->private;
data_t *data[ec->nodes];
@@ -660,8 +712,8 @@ int32_t ec_dict_data_stime(ec_cbk_data_t * cbk, int32_t which, char * key)
}
err = gf_get_max_stime(cbk->fop->xl, dict, key, data[i]);
if (err != 0) {
- gf_msg (cbk->fop->xl->name, GF_LOG_ERROR, -err,
- EC_MSG_STIME_COMBINE_FAIL, "STIME combination failed");
+ gf_msg(cbk->fop->xl->name, GF_LOG_ERROR, -err,
+ EC_MSG_STIME_COMBINE_FAIL, "STIME combination failed");
return err;
}
@@ -670,28 +722,24 @@ int32_t ec_dict_data_stime(ec_cbk_data_t * cbk, int32_t which, char * key)
return 0;
}
-int32_t ec_dict_data_combine(dict_t * dict, char * key, data_t * value,
- void * arg)
+int32_t
+ec_dict_data_combine(dict_t *dict, char *key, data_t *value, void *arg)
{
- ec_dict_combine_t * data = arg;
+ ec_dict_combine_t *data = arg;
if ((strcmp(key, GF_XATTR_PATHINFO_KEY) == 0) ||
- (strcmp(key, GF_XATTR_USER_PATHINFO_KEY) == 0))
- {
- return ec_dict_data_concat("(<EC:%s> { })", data->cbk, data->which,
- key, NULL, NULL, _gf_false,
+ (strcmp(key, GF_XATTR_USER_PATHINFO_KEY) == 0)) {
+ return ec_dict_data_concat(data->cbk, data->which, key, NULL, NULL,
+ _gf_false, _gf_false, "(<EC:%s> { })",
data->cbk->fop->xl->name);
}
- if (strncmp(key, GF_XATTR_CLRLK_CMD, strlen(GF_XATTR_CLRLK_CMD)) == 0)
- {
- return ec_dict_data_concat("{\n}", data->cbk, data->which, key, NULL,
- NULL, _gf_false);
+ if (strncmp(key, GF_XATTR_CLRLK_CMD, SLEN(GF_XATTR_CLRLK_CMD)) == 0) {
+ return ec_dict_data_concat(data->cbk, data->which, key, NULL, NULL,
+ _gf_false, "{\n}");
}
- if (strncmp(key, GF_XATTR_LOCKINFO_KEY,
- strlen(GF_XATTR_LOCKINFO_KEY)) == 0)
- {
+ if (strncmp(key, GF_XATTR_LOCKINFO_KEY, SLEN(GF_XATTR_LOCKINFO_KEY)) == 0) {
return ec_dict_data_merge(data->cbk, data->which, key);
}
@@ -699,8 +747,7 @@ int32_t ec_dict_data_combine(dict_t * dict, char * key, data_t * value,
return ec_dict_data_max32(data->cbk, data->which, key);
}
- if (strcmp(key, GLUSTERFS_OPEN_FD_COUNT) == 0)
- {
+ if (strcmp(key, GLUSTERFS_OPEN_FD_COUNT) == 0) {
return ec_dict_data_max32(data->cbk, data->which, key);
}
if ((strcmp(key, GLUSTERFS_INODELK_COUNT) == 0) ||
@@ -712,23 +759,22 @@ int32_t ec_dict_data_combine(dict_t * dict, char * key, data_t * value,
return ec_dict_data_quota(data->cbk, data->which, key);
}
/* Ignore all other quota attributes */
- if (strncmp(key, EC_QUOTA_PREFIX, strlen(EC_QUOTA_PREFIX)) == 0) {
+ if (strncmp(key, EC_QUOTA_PREFIX, SLEN(EC_QUOTA_PREFIX)) == 0) {
return 0;
}
if (XATTR_IS_NODE_UUID(key)) {
if (data->cbk->fop->int32) {
- /* List of node uuid is requested */
- return ec_dict_data_concat("{ }", data->cbk, data->which, key,
- GF_XATTR_LIST_NODE_UUIDS_KEY,
- UUID0_STR, _gf_true);
+ /* List of node uuid is requested */
+ return ec_dict_data_concat(data->cbk, data->which, key,
+ GF_XATTR_LIST_NODE_UUIDS_KEY, UUID0_STR,
+ _gf_true, "{ }");
} else {
- return ec_dict_data_uuid(data->cbk, data->which, key);
+ return ec_dict_data_uuid(data->cbk, data->which, key);
}
}
- if (fnmatch(GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0)
- {
+ if (fnmatch(GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0) {
return ec_dict_data_stime(data->cbk, data->which, key);
}
@@ -736,10 +782,15 @@ int32_t ec_dict_data_combine(dict_t * dict, char * key, data_t * value,
return ec_dict_data_max64(data->cbk, data->which, key);
}
+ if (strcmp(key, GF_PRESTAT) == 0 || strcmp(key, GF_POSTSTAT) == 0) {
+ return ec_dict_data_iatt(data->cbk, data->which, key);
+ }
+
return 0;
}
-int32_t ec_dict_combine(ec_cbk_data_t * cbk, int32_t which)
+int32_t
+ec_dict_combine(ec_cbk_data_t *cbk, int32_t which)
{
dict_t *dict = NULL;
ec_dict_combine_t data;
@@ -752,9 +803,8 @@ int32_t ec_dict_combine(ec_cbk_data_t * cbk, int32_t which)
if (dict != NULL) {
err = dict_foreach(dict, ec_dict_data_combine, &data);
if (err != 0) {
- gf_msg (cbk->fop->xl->name, GF_LOG_ERROR, -err,
- EC_MSG_DICT_COMBINE_FAIL,
- "Dictionary combination failed");
+ gf_msg(cbk->fop->xl->name, GF_LOG_ERROR, -err,
+ EC_MSG_DICT_COMBINE_FAIL, "Dictionary combination failed");
return err;
}
@@ -763,47 +813,43 @@ int32_t ec_dict_combine(ec_cbk_data_t * cbk, int32_t which)
return 0;
}
-int32_t ec_vector_compare(struct iovec * dst_vector, int32_t dst_count,
- struct iovec * src_vector, int32_t src_count)
+int32_t
+ec_vector_compare(struct iovec *dst_vector, int32_t dst_count,
+ struct iovec *src_vector, int32_t src_count)
{
int32_t dst_size = 0, src_size = 0;
- if (dst_count > 0)
- {
+ if (dst_count > 0) {
dst_size = iov_length(dst_vector, dst_count);
}
- if (src_count > 0)
- {
+ if (src_count > 0) {
src_size = iov_length(src_vector, src_count);
}
return (dst_size == src_size);
}
-int32_t ec_flock_compare(struct gf_flock * dst, struct gf_flock * src)
+int32_t
+ec_flock_compare(struct gf_flock *dst, struct gf_flock *src)
{
- if ((dst->l_type != src->l_type) ||
- (dst->l_whence != src->l_whence) ||
- (dst->l_start != src->l_start) ||
- (dst->l_len != src->l_len) ||
+ if ((dst->l_type != src->l_type) || (dst->l_whence != src->l_whence) ||
+ (dst->l_start != src->l_start) || (dst->l_len != src->l_len) ||
(dst->l_pid != src->l_pid) ||
- !is_same_lkowner(&dst->l_owner, &src->l_owner))
- {
+ !is_same_lkowner(&dst->l_owner, &src->l_owner)) {
return 0;
}
return 1;
}
-void ec_statvfs_combine(struct statvfs * dst, struct statvfs * src)
+void
+ec_statvfs_combine(struct statvfs *dst, struct statvfs *src)
{
- if (dst->f_bsize < src->f_bsize)
- {
+ if (dst->f_bsize < src->f_bsize) {
dst->f_bsize = src->f_bsize;
}
- if (dst->f_frsize < src->f_frsize)
- {
+ if (dst->f_frsize < src->f_frsize) {
dst->f_blocks *= dst->f_frsize;
dst->f_blocks /= src->f_frsize;
@@ -814,9 +860,7 @@ void ec_statvfs_combine(struct statvfs * dst, struct statvfs * src)
dst->f_bavail /= src->f_frsize;
dst->f_frsize = src->f_frsize;
- }
- else if (dst->f_frsize > src->f_frsize)
- {
+ } else if (dst->f_frsize > src->f_frsize) {
src->f_blocks *= src->f_frsize;
src->f_blocks /= dst->f_frsize;
@@ -826,90 +870,80 @@ void ec_statvfs_combine(struct statvfs * dst, struct statvfs * src)
src->f_bavail *= src->f_frsize;
src->f_bavail /= dst->f_frsize;
}
- if (dst->f_blocks > src->f_blocks)
- {
+ if (dst->f_blocks > src->f_blocks) {
dst->f_blocks = src->f_blocks;
}
- if (dst->f_bfree > src->f_bfree)
- {
+ if (dst->f_bfree > src->f_bfree) {
dst->f_bfree = src->f_bfree;
}
- if (dst->f_bavail > src->f_bavail)
- {
+ if (dst->f_bavail > src->f_bavail) {
dst->f_bavail = src->f_bavail;
}
- if (dst->f_files < src->f_files)
- {
+ if (dst->f_files < src->f_files) {
dst->f_files = src->f_files;
}
- if (dst->f_ffree > src->f_ffree)
- {
+ if (dst->f_ffree > src->f_ffree) {
dst->f_ffree = src->f_ffree;
}
- if (dst->f_favail > src->f_favail)
- {
+ if (dst->f_favail > src->f_favail) {
dst->f_favail = src->f_favail;
}
- if (dst->f_namemax > src->f_namemax)
- {
+ if (dst->f_namemax > src->f_namemax) {
dst->f_namemax = src->f_namemax;
}
- if (dst->f_flag != src->f_flag)
- {
- gf_msg_debug (THIS->name, 0,
- "Mismatching file system flags "
- "(%lX, %lX)",
- dst->f_flag, src->f_flag);
+ if (dst->f_flag != src->f_flag) {
+ gf_msg_debug(THIS->name, 0,
+ "Mismatching file system flags "
+ "(%lX, %lX)",
+ dst->f_flag, src->f_flag);
}
dst->f_flag &= src->f_flag;
}
-int32_t ec_combine_check(ec_cbk_data_t * dst, ec_cbk_data_t * src,
- ec_combine_f combine)
+int32_t
+ec_combine_check(ec_cbk_data_t *dst, ec_cbk_data_t *src, ec_combine_f combine)
{
- ec_fop_data_t * fop = dst->fop;
+ ec_fop_data_t *fop = dst->fop;
- if (dst->op_ret != src->op_ret)
- {
- gf_msg_debug (fop->xl->name, 0, "Mismatching return code in "
- "answers of '%s': %d <-> %d",
- ec_fop_name(fop->id), dst->op_ret, src->op_ret);
+ if (dst->op_ret != src->op_ret) {
+ gf_msg_debug(fop->xl->name, 0,
+ "Mismatching return code in "
+ "answers of '%s': %d <-> %d",
+ ec_fop_name(fop->id), dst->op_ret, src->op_ret);
return 0;
}
- if (dst->op_ret < 0)
- {
- if (dst->op_errno != src->op_errno)
- {
- gf_msg_debug (fop->xl->name, 0, "Mismatching errno code in "
- "answers of '%s': %d <-> %d",
- ec_fop_name(fop->id), dst->op_errno, src->op_errno);
+ if (dst->op_ret < 0) {
+ if (dst->op_errno != src->op_errno) {
+ gf_msg_debug(fop->xl->name, 0,
+ "Mismatching errno code in "
+ "answers of '%s': %d <-> %d",
+ ec_fop_name(fop->id), dst->op_errno, src->op_errno);
return 0;
}
}
- if (!ec_dict_compare(dst->xdata, src->xdata))
- {
- gf_msg (fop->xl->name, GF_LOG_DEBUG, 0,
- EC_MSG_XDATA_MISMATCH,
- "Mismatching xdata in answers "
- "of '%s'", ec_fop_name(fop->id));
+ if (!ec_dict_compare(dst->xdata, src->xdata)) {
+ gf_msg(fop->xl->name, GF_LOG_DEBUG, 0, EC_MSG_XDATA_MISMATCH,
+ "Mismatching xdata in answers "
+ "of '%s'",
+ ec_fop_name(fop->id));
return 0;
}
- if ((dst->op_ret >= 0) && (combine != NULL))
- {
+ if ((dst->op_ret >= 0) && (combine != NULL)) {
return combine(fop, dst, src);
}
return 1;
}
-void ec_combine (ec_cbk_data_t *newcbk, ec_combine_f combine)
+void
+ec_combine(ec_cbk_data_t *newcbk, ec_combine_f combine)
{
ec_fop_data_t *fop = newcbk->fop;
ec_cbk_data_t *cbk = NULL, *tmp = NULL;
@@ -924,17 +958,14 @@ void ec_combine (ec_cbk_data_t *newcbk, ec_combine_f combine)
item = fop->cbk_list.prev;
list_for_each_entry(cbk, &fop->cbk_list, list)
{
- if (ec_combine_check(newcbk, cbk, combine))
- {
+ if (ec_combine_check(newcbk, cbk, combine)) {
newcbk->count += cbk->count;
newcbk->mask |= cbk->mask;
item = cbk->list.prev;
- while (item != &fop->cbk_list)
- {
+ while (item != &fop->cbk_list) {
tmp = list_entry(item, ec_cbk_data_t, list);
- if (tmp->count >= newcbk->count)
- {
+ if (tmp->count >= newcbk->count) {
break;
}
item = item->prev;
diff --git a/xlators/cluster/ec/src/ec-combine.h b/xlators/cluster/ec/src/ec-combine.h
index 19a42ded706..1010cc3be26 100644
--- a/xlators/cluster/ec/src/ec-combine.h
+++ b/xlators/cluster/ec/src/ec-combine.h
@@ -11,28 +11,34 @@
#ifndef __EC_COMBINE_H__
#define __EC_COMBINE_H__
-#define EC_COMBINE_DICT 0
+#define EC_COMBINE_DICT 0
#define EC_COMBINE_XDATA 1
-typedef int32_t (* ec_combine_f)(ec_fop_data_t * fop, ec_cbk_data_t * dst,
- ec_cbk_data_t * src);
+typedef int32_t (*ec_combine_f)(ec_fop_data_t *fop, ec_cbk_data_t *dst,
+ ec_cbk_data_t *src);
-void ec_iatt_rebuild(ec_t * ec, struct iatt * iatt, int32_t count,
- int32_t answers);
+void
+ec_iatt_rebuild(ec_t *ec, struct iatt *iatt, int32_t count, int32_t answers);
-int32_t ec_iatt_combine(ec_fop_data_t *fop, struct iatt *dst, struct iatt *src,
- int32_t count);
-int32_t ec_dict_compare(dict_t * dict1, dict_t * dict2);
-int32_t ec_vector_compare(struct iovec * dst_vector, int32_t dst_count,
- struct iovec * src_vector, int32_t src_count);
-int32_t ec_flock_compare(struct gf_flock * dst, struct gf_flock * src);
-void ec_statvfs_combine(struct statvfs * dst, struct statvfs * src);
+int32_t
+ec_iatt_combine(ec_fop_data_t *fop, struct iatt *dst, struct iatt *src,
+ int32_t count);
+int32_t
+ec_dict_compare(dict_t *dict1, dict_t *dict2);
+int32_t
+ec_vector_compare(struct iovec *dst_vector, int32_t dst_count,
+ struct iovec *src_vector, int32_t src_count);
+int32_t
+ec_flock_compare(struct gf_flock *dst, struct gf_flock *src);
+void
+ec_statvfs_combine(struct statvfs *dst, struct statvfs *src);
-int32_t ec_dict_combine(ec_cbk_data_t * cbk, int32_t which);
+int32_t
+ec_dict_combine(ec_cbk_data_t *cbk, int32_t which);
-void ec_combine(ec_cbk_data_t * cbk, ec_combine_f combine);
+void
+ec_combine(ec_cbk_data_t *cbk, ec_combine_f combine);
int32_t
-ec_combine_write (ec_fop_data_t *fop, ec_cbk_data_t *dst,
- ec_cbk_data_t *src);
+ec_combine_write(ec_fop_data_t *fop, ec_cbk_data_t *dst, ec_cbk_data_t *src);
#endif /* __EC_COMBINE_H__ */
diff --git a/xlators/cluster/ec/src/ec-common.c b/xlators/cluster/ec/src/ec-common.c
index e3e34811395..b955efd8c2d 100644
--- a/xlators/cluster/ec/src/ec-common.c
+++ b/xlators/cluster/ec/src/ec-common.c
@@ -8,8 +8,8 @@
cases as published by the Free Software Foundation.
*/
-#include "byte-order.h"
-#include "hashfn.h"
+#include <glusterfs/byte-order.h>
+#include <glusterfs/hashfn.h>
#include "ec-mem-types.h"
#include "ec-types.h"
@@ -24,196 +24,202 @@
#define EC_INVALID_INDEX UINT32_MAX
void
-ec_update_fd_status (fd_t *fd, xlator_t *xl, int idx,
- int32_t ret_status)
+ec_update_fd_status(fd_t *fd, xlator_t *xl, int idx, int32_t ret_status)
{
- ec_fd_t *fd_ctx;
+ ec_fd_t *fd_ctx;
- if (fd == NULL)
- return;
+ if (fd == NULL)
+ return;
- LOCK (&fd->lock);
- {
- fd_ctx = __ec_fd_get(fd, xl);
- if (fd_ctx) {
- if (ret_status >= 0)
- fd_ctx->fd_status[idx] = EC_FD_OPENED;
- else
- fd_ctx->fd_status[idx] = EC_FD_NOT_OPENED;
- }
+ LOCK(&fd->lock);
+ {
+ fd_ctx = __ec_fd_get(fd, xl);
+ if (fd_ctx) {
+ if (ret_status >= 0)
+ fd_ctx->fd_status[idx] = EC_FD_OPENED;
+ else
+ fd_ctx->fd_status[idx] = EC_FD_NOT_OPENED;
}
- UNLOCK (&fd->lock);
+ }
+ UNLOCK(&fd->lock);
}
-static int
-ec_fd_ctx_need_open (fd_t *fd, xlator_t *this, uintptr_t *need_open)
+static uintptr_t
+ec_fd_ctx_need_open(fd_t *fd, xlator_t *this, uintptr_t mask)
{
int i = 0;
int count = 0;
ec_t *ec = NULL;
ec_fd_t *fd_ctx = NULL;
+ uintptr_t need_open = 0;
ec = this->private;
- *need_open = 0;
- fd_ctx = ec_fd_get (fd, this);
+ fd_ctx = ec_fd_get(fd, this);
if (!fd_ctx)
return count;
- LOCK (&fd->lock);
+ LOCK(&fd->lock);
{
for (i = 0; i < ec->nodes; i++) {
- if ((fd_ctx->fd_status[i] == EC_FD_NOT_OPENED) &&
- (ec->xl_up & (1<<i))) {
- fd_ctx->fd_status[i] = EC_FD_OPENING;
- *need_open |= (1<<i);
- count++;
- }
+ if ((fd_ctx->fd_status[i] == EC_FD_NOT_OPENED) &&
+ ((ec->xl_up & (1 << i)) != 0) && ((mask & (1 << i)) != 0)) {
+ fd_ctx->fd_status[i] = EC_FD_OPENING;
+ need_open |= (1 << i);
+ count++;
+ }
}
}
- UNLOCK (&fd->lock);
+ UNLOCK(&fd->lock);
/* If fd needs to open on minimum number of nodes
* then ignore fixing the fd as it has been
* requested from heal operation.
*/
- if (count >= ec->fragments)
- count = 0;
+ if (count >= ec->fragments) {
+ need_open = 0;
+ }
- return count;
+ return need_open;
}
static gf_boolean_t
-ec_is_fd_fixable (fd_t *fd)
+ec_is_fd_fixable(fd_t *fd)
{
if (!fd || !fd->inode)
return _gf_false;
- else if (fd_is_anonymous (fd))
+ else if (fd_is_anonymous(fd))
return _gf_false;
- else if (gf_uuid_is_null (fd->inode->gfid))
+ else if (gf_uuid_is_null(fd->inode->gfid))
return _gf_false;
return _gf_true;
}
static void
-ec_fix_open (ec_fop_data_t *fop)
+ec_fix_open(ec_fop_data_t *fop, uintptr_t mask)
{
- int call_count = 0;
- uintptr_t need_open = 0;
- int ret = 0;
- loc_t loc = {0, };
+ uintptr_t need_open = 0;
+ int ret = 0;
+ int32_t flags = 0;
+ loc_t loc = {
+ 0,
+ };
- if (!ec_is_fd_fixable (fop->fd))
+ if (!ec_is_fd_fixable(fop->fd))
goto out;
/* Evaluate how many remote fd's to be opened */
- call_count = ec_fd_ctx_need_open (fop->fd, fop->xl, &need_open);
- if (!call_count)
+ need_open = ec_fd_ctx_need_open(fop->fd, fop->xl, mask);
+ if (need_open == 0) {
goto out;
+ }
- loc.inode = inode_ref (fop->fd->inode);
- gf_uuid_copy (loc.gfid, fop->fd->inode->gfid);
- ret = loc_path (&loc, NULL);
+ loc.inode = inode_ref(fop->fd->inode);
+ gf_uuid_copy(loc.gfid, fop->fd->inode->gfid);
+ ret = loc_path(&loc, NULL);
if (ret < 0) {
goto out;
}
+ flags = fop->fd->flags & (~(O_TRUNC | O_APPEND | O_CREAT | O_EXCL));
if (IA_IFDIR == fop->fd->inode->ia_type) {
- ec_opendir(fop->frame, fop->xl, need_open, EC_MINIMUM_ONE,
- NULL, NULL, &fop->loc[0], fop->fd, NULL);
- } else{
- ec_open(fop->frame, fop->xl, need_open, EC_MINIMUM_ONE,
- NULL, NULL, &loc, fop->fd->flags, fop->fd, NULL);
+ ec_opendir(fop->frame, fop->xl, need_open,
+ EC_MINIMUM_ONE | EC_FOP_NO_PROPAGATE_ERROR, NULL, NULL,
+ &fop->loc[0], fop->fd, NULL);
+ } else {
+ ec_open(fop->frame, fop->xl, need_open,
+ EC_MINIMUM_ONE | EC_FOP_NO_PROPAGATE_ERROR, NULL, NULL, &loc,
+ flags, fop->fd, NULL);
}
out:
- loc_wipe (&loc);
+ loc_wipe(&loc);
}
-off_t
-ec_range_end_get (off_t fl_start, size_t fl_size)
+static off_t
+ec_range_end_get(off_t fl_start, uint64_t fl_size)
{
- off_t fl_end = 0;
- switch (fl_size) {
- case 0:
- return fl_start;
- case LLONG_MAX: /*Infinity*/
- return LLONG_MAX;
- default:
- fl_end = fl_start + fl_size - 1;
- if (fl_end < 0) /*over-flow*/
- return LLONG_MAX;
- else
- return fl_end;
+ if (fl_size > 0) {
+ if (fl_size >= EC_RANGE_FULL) {
+ /* Infinity */
+ fl_start = LLONG_MAX;
+ } else {
+ fl_start += fl_size - 1;
+ if (fl_start < 0) {
+ /* Overflow */
+ fl_start = LLONG_MAX;
+ }
}
+ }
+
+ return fl_start;
}
static gf_boolean_t
-ec_is_range_conflict (ec_lock_link_t *l1, ec_lock_link_t *l2)
+ec_is_range_conflict(ec_lock_link_t *l1, ec_lock_link_t *l2)
{
- return ((l1->fl_end >= l2->fl_start) && (l2->fl_end >= l1->fl_start));
+ return ((l1->fl_end >= l2->fl_start) && (l2->fl_end >= l1->fl_start));
}
static gf_boolean_t
-ec_lock_conflict (ec_lock_link_t *l1, ec_lock_link_t *l2)
+ec_lock_conflict(ec_lock_link_t *l1, ec_lock_link_t *l2)
{
- ec_t *ec = l1->fop->xl->private;
+ ec_t *ec = l1->fop->xl->private;
- /* Fops like access/stat won't have to worry what the other fops are
- * modifying as the fop is wound only to one brick. So it can be
- * executed in parallel*/
- if (l1->fop->minimum == EC_MINIMUM_ONE ||
- l2->fop->minimum == EC_MINIMUM_ONE)
- return _gf_false;
+ /* Fops like access/stat won't have to worry what the other fops are
+ * modifying as the fop is wound only to one brick. So it can be
+ * executed in parallel*/
+ if (l1->fop->minimum == EC_MINIMUM_ONE ||
+ l2->fop->minimum == EC_MINIMUM_ONE)
+ return _gf_false;
- if ((l1->fop->flags & EC_FLAG_LOCK_SHARED) &&
- (l2->fop->flags & EC_FLAG_LOCK_SHARED))
- return _gf_false;
+ if ((l1->fop->flags & EC_FLAG_LOCK_SHARED) &&
+ (l2->fop->flags & EC_FLAG_LOCK_SHARED))
+ return _gf_false;
- if (!ec->parallel_writes) {
- return _gf_true;
- }
+ if (!ec->parallel_writes) {
+ return _gf_true;
+ }
- return ec_is_range_conflict (l1, l2);
+ return ec_is_range_conflict(l1, l2);
}
uint32_t
-ec_select_first_by_read_policy (ec_t *ec, ec_fop_data_t *fop)
-{
- if (ec->read_policy == EC_ROUND_ROBIN) {
- return ec->idx;
- } else if (ec->read_policy == EC_GFID_HASH) {
- if (fop->use_fd) {
- return SuperFastHash((char *)fop->fd->inode->gfid,
- sizeof(fop->fd->inode->gfid)) % ec->nodes;
- } else {
- if (gf_uuid_is_null (fop->loc[0].gfid))
- loc_gfid (&fop->loc[0], fop->loc[0].gfid);
- return SuperFastHash((char *)fop->loc[0].gfid,
- sizeof(fop->loc[0].gfid)) % ec->nodes;
- }
+ec_select_first_by_read_policy(ec_t *ec, ec_fop_data_t *fop)
+{
+ if (ec->read_policy == EC_ROUND_ROBIN) {
+ return ec->idx;
+ } else if (ec->read_policy == EC_GFID_HASH) {
+ if (fop->use_fd) {
+ return SuperFastHash((char *)fop->fd->inode->gfid,
+ sizeof(fop->fd->inode->gfid)) %
+ ec->nodes;
+ } else {
+ if (gf_uuid_is_null(fop->loc[0].gfid))
+ loc_gfid(&fop->loc[0], fop->loc[0].gfid);
+ return SuperFastHash((char *)fop->loc[0].gfid,
+ sizeof(fop->loc[0].gfid)) %
+ ec->nodes;
}
- return 0;
+ }
+ return 0;
}
-static
-gf_boolean_t ec_child_valid(ec_t * ec, ec_fop_data_t * fop, uint32_t idx)
+static gf_boolean_t
+ec_child_valid(ec_t *ec, ec_fop_data_t *fop, uint32_t idx)
{
return (idx < ec->nodes) && (((fop->remaining >> idx) & 1) == 1);
}
-static
-uint32_t ec_child_next(ec_t * ec, ec_fop_data_t * fop, uint32_t idx)
+static uint32_t
+ec_child_next(ec_t *ec, ec_fop_data_t *fop, uint32_t idx)
{
- while (!ec_child_valid(ec, fop, idx))
- {
- if (++idx >= ec->nodes)
- {
+ while (!ec_child_valid(ec, fop, idx)) {
+ if (++idx >= ec->nodes) {
idx = 0;
}
- if (idx == fop->first)
- {
+ if (idx == fop->first) {
return EC_INVALID_INDEX;
}
}
@@ -221,20 +227,21 @@ uint32_t ec_child_next(ec_t * ec, ec_fop_data_t * fop, uint32_t idx)
return idx;
}
-int32_t ec_heal_report(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, uintptr_t mask,
- uintptr_t good, uintptr_t bad, dict_t * xdata)
+int32_t
+ec_heal_report(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, uintptr_t mask, uintptr_t good,
+ uintptr_t bad, uint32_t pending, dict_t *xdata)
{
if (op_ret < 0) {
- gf_msg (this->name, GF_LOG_DEBUG, op_errno,
- EC_MSG_HEAL_FAIL, "Heal failed");
+ gf_msg(this->name, GF_LOG_DEBUG, op_errno, EC_MSG_HEAL_FAIL,
+ "Heal failed");
} else {
if ((mask & ~good) != 0) {
- gf_msg (this->name, GF_LOG_DEBUG, 0,
- EC_MSG_HEAL_SUCCESS, "Heal succeeded on %d/%d "
- "subvolumes",
- gf_bits_count(mask & ~(good | bad)),
- gf_bits_count(mask & ~good));
+ gf_msg(this->name, GF_LOG_DEBUG, 0, EC_MSG_HEAL_SUCCESS,
+ "Heal succeeded on %d/%d "
+ "subvolumes",
+ gf_bits_count(mask & ~(good | bad)),
+ gf_bits_count(mask & ~good));
}
}
@@ -242,103 +249,104 @@ int32_t ec_heal_report(call_frame_t * frame, void * cookie, xlator_t * this,
}
static uintptr_t
-ec_fop_needs_name_heal (ec_fop_data_t *fop)
+ec_fop_needs_name_heal(ec_fop_data_t *fop)
{
- ec_t *ec = NULL;
- ec_cbk_data_t *cbk = NULL;
- ec_cbk_data_t *enoent_cbk = NULL;
+ ec_t *ec = NULL;
+ ec_cbk_data_t *cbk = NULL;
+ ec_cbk_data_t *enoent_cbk = NULL;
- ec = fop->xl->private;
- if (fop->id != GF_FOP_LOOKUP)
- return 0;
+ ec = fop->xl->private;
+ if (fop->id != GF_FOP_LOOKUP)
+ return 0;
- if (!fop->loc[0].name || strlen (fop->loc[0].name) == 0)
- return 0;
+ if (!fop->loc[0].name || strlen(fop->loc[0].name) == 0)
+ return 0;
- list_for_each_entry(cbk, &fop->cbk_list, list)
- {
- if (cbk->op_ret < 0 && cbk->op_errno == ENOENT) {
- enoent_cbk = cbk;
- break;
- }
+ list_for_each_entry(cbk, &fop->cbk_list, list)
+ {
+ if (cbk->op_ret < 0 && cbk->op_errno == ENOENT) {
+ enoent_cbk = cbk;
+ break;
}
+ }
- if (!enoent_cbk)
- return 0;
+ if (!enoent_cbk)
+ return 0;
- return ec->xl_up & ~enoent_cbk->mask;
+ return ec->xl_up & ~enoent_cbk->mask;
}
-int32_t ec_fop_needs_heal(ec_fop_data_t *fop)
+int32_t
+ec_fop_needs_heal(ec_fop_data_t *fop)
{
ec_t *ec = fop->xl->private;
if (fop->lock_count == 0) {
- /*
- * if fop->lock_count is zero that means it saw version mismatch
- * without any locks so it can't be trusted. If we launch a heal
- * based on this it will lead to INODELKs which will affect I/O
- * performance. Considering self-heal-daemon and operations on
- * the inode from client which take locks can still trigger the
- * heal we can choose to not attempt a heal when fop->lock_count
- * is zero.
- */
- return 0;
+ /*
+ * if fop->lock_count is zero that means it saw version mismatch
+ * without any locks so it can't be trusted. If we launch a heal
+ * based on this it will lead to INODELKs which will affect I/O
+ * performance. Considering self-heal-daemon and operations on
+ * the inode from client which take locks can still trigger the
+ * heal we can choose to not attempt a heal when fop->lock_count
+ * is zero.
+ */
+ return 0;
}
return (ec->xl_up & ~(fop->remaining | fop->good)) != 0;
}
-void ec_check_status(ec_fop_data_t * fop)
+void
+ec_check_status(ec_fop_data_t *fop)
{
- ec_t * ec = fop->xl->private;
+ ec_t *ec = fop->xl->private;
int32_t partial = 0;
char str1[32], str2[32], str3[32], str4[32], str5[32];
- if (!ec_fop_needs_name_heal (fop) && !ec_fop_needs_heal(fop)) {
+ if (!ec_fop_needs_name_heal(fop) && !ec_fop_needs_heal(fop)) {
return;
}
if (fop->answer && fop->answer->op_ret >= 0) {
- if ((fop->id == GF_FOP_LOOKUP) ||
- (fop->id == GF_FOP_STAT) || (fop->id == GF_FOP_FSTAT)) {
+ if ((fop->id == GF_FOP_LOOKUP) || (fop->id == GF_FOP_STAT) ||
+ (fop->id == GF_FOP_FSTAT)) {
partial = fop->answer->iatt[0].ia_type == IA_IFDIR;
} else if (fop->id == GF_FOP_OPENDIR) {
partial = 1;
}
}
- gf_msg (fop->xl->name, GF_LOG_WARNING, 0,
- EC_MSG_OP_FAIL_ON_SUBVOLS,
- "Operation failed on %d of %d subvolumes.(up=%s, mask=%s, "
- "remaining=%s, good=%s, bad=%s)",
- gf_bits_count(ec->xl_up & ~(fop->remaining | fop->good)), ec->nodes,
- ec_bin(str1, sizeof(str1), ec->xl_up, ec->nodes),
- ec_bin(str2, sizeof(str2), fop->mask, ec->nodes),
- ec_bin(str3, sizeof(str3), fop->remaining, ec->nodes),
- ec_bin(str4, sizeof(str4), fop->good, ec->nodes),
- ec_bin(str5, sizeof(str5),
- ec->xl_up & ~(fop->remaining | fop->good), ec->nodes));
- if (fop->use_fd)
- {
+ gf_msg(
+ fop->xl->name, GF_LOG_WARNING, 0, EC_MSG_OP_FAIL_ON_SUBVOLS,
+ "Operation failed on %d of %d subvolumes.(up=%s, mask=%s, "
+ "remaining=%s, good=%s, bad=%s,"
+ "(Least significant bit represents first client/brick of subvol), %s)",
+ gf_bits_count(ec->xl_up & ~(fop->remaining | fop->good)), ec->nodes,
+ ec_bin(str1, sizeof(str1), ec->xl_up, ec->nodes),
+ ec_bin(str2, sizeof(str2), fop->mask, ec->nodes),
+ ec_bin(str3, sizeof(str3), fop->remaining, ec->nodes),
+ ec_bin(str4, sizeof(str4), fop->good, ec->nodes),
+ ec_bin(str5, sizeof(str5), ec->xl_up & ~(fop->remaining | fop->good),
+ ec->nodes),
+ ec_msg_str(fop));
+ if (fop->use_fd) {
if (fop->fd != NULL) {
ec_fheal(NULL, fop->xl, -1, EC_MINIMUM_ONE, ec_heal_report, NULL,
fop->fd, partial, NULL);
}
- }
- else
- {
+ } else {
ec_heal(NULL, fop->xl, -1, EC_MINIMUM_ONE, ec_heal_report, NULL,
&fop->loc[0], partial, NULL);
- if (fop->loc[1].inode != NULL)
- {
+ if (fop->loc[1].inode != NULL) {
ec_heal(NULL, fop->xl, -1, EC_MINIMUM_ONE, ec_heal_report, NULL,
&fop->loc[1], partial, NULL);
}
}
}
-void ec_update_good(ec_fop_data_t *fop, uintptr_t good)
+void
+ec_update_good(ec_fop_data_t *fop, uintptr_t good)
{
fop->good = good;
@@ -349,7 +357,8 @@ void ec_update_good(ec_fop_data_t *fop, uintptr_t good)
}
}
-void ec_lock_update_good(ec_lock_t *lock, ec_fop_data_t *fop)
+void
+ec_lock_update_good(ec_lock_t *lock, ec_fop_data_t *fop)
{
/* Fops that are executed only on one brick do not have enough information
* to update the global mask of good bricks. */
@@ -365,15 +374,16 @@ void ec_lock_update_good(ec_lock_t *lock, ec_fop_data_t *fop)
lock->good_mask &= fop->good | fop->remaining;
}
-void __ec_fop_set_error(ec_fop_data_t * fop, int32_t error)
+void
+__ec_fop_set_error(ec_fop_data_t *fop, int32_t error)
{
- if ((error != 0) && (fop->error == 0))
- {
+ if ((error != 0) && (fop->error == 0)) {
fop->error = error;
}
}
-void ec_fop_set_error(ec_fop_data_t * fop, int32_t error)
+void
+ec_fop_set_error(ec_fop_data_t *fop, int32_t error)
{
LOCK(&fop->lock);
@@ -425,18 +435,20 @@ ec_fop_prepare_answer(ec_fop_data_t *fop, gf_boolean_t ro)
return cbk;
}
-void ec_sleep(ec_fop_data_t *fop)
+void
+ec_sleep(ec_fop_data_t *fop)
{
LOCK(&fop->lock);
- GF_ASSERT (fop->refs > 0);
+ GF_ASSERT(fop->refs > 0);
fop->refs++;
fop->jobs++;
UNLOCK(&fop->lock);
}
-int32_t ec_check_complete(ec_fop_data_t * fop, ec_resume_f resume)
+int32_t
+ec_check_complete(ec_fop_data_t *fop, ec_resume_f resume)
{
int32_t error = -1;
@@ -444,14 +456,11 @@ int32_t ec_check_complete(ec_fop_data_t * fop, ec_resume_f resume)
GF_ASSERT(fop->resume == NULL);
- if (--fop->jobs != 0)
- {
+ if (--fop->jobs != 0) {
ec_trace("WAIT", fop, "resume=%p", resume);
fop->resume = resume;
- }
- else
- {
+ } else {
error = fop->error;
fop->error = 0;
}
@@ -461,7 +470,8 @@ int32_t ec_check_complete(ec_fop_data_t * fop, ec_resume_f resume)
return error;
}
-void ec_resume(ec_fop_data_t * fop, int32_t error)
+void
+ec_resume(ec_fop_data_t *fop, int32_t error)
{
ec_resume_f resume = NULL;
@@ -469,16 +479,13 @@ void ec_resume(ec_fop_data_t * fop, int32_t error)
__ec_fop_set_error(fop, error);
- if (--fop->jobs == 0)
- {
+ if (--fop->jobs == 0) {
resume = fop->resume;
fop->resume = NULL;
- if (resume != NULL)
- {
+ if (resume != NULL) {
ec_trace("RESUME", fop, "error=%d", error);
- if (fop->error != 0)
- {
+ if (fop->error != 0) {
error = fop->error;
}
fop->error = 0;
@@ -487,21 +494,24 @@ void ec_resume(ec_fop_data_t * fop, int32_t error)
UNLOCK(&fop->lock);
- if (resume != NULL)
- {
+ if (resume != NULL) {
resume(fop, error);
}
ec_fop_data_release(fop);
}
-void ec_resume_parent(ec_fop_data_t * fop, int32_t error)
+void
+ec_resume_parent(ec_fop_data_t *fop)
{
- ec_fop_data_t * parent;
+ ec_fop_data_t *parent;
+ int32_t error = 0;
parent = fop->parent;
- if (parent != NULL)
- {
+ if (parent != NULL) {
+ if ((fop->fop_flags & EC_FOP_NO_PROPAGATE_ERROR) == 0) {
+ error = fop->error;
+ }
ec_trace("RESUME_PARENT", fop, "error=%u", error);
fop->parent = NULL;
ec_resume(parent, error);
@@ -509,22 +519,23 @@ void ec_resume_parent(ec_fop_data_t * fop, int32_t error)
}
gf_boolean_t
-ec_is_recoverable_error (int32_t op_errno)
+ec_is_recoverable_error(int32_t op_errno)
{
- switch (op_errno) {
+ switch (op_errno) {
case ENOTCONN:
case ESTALE:
case ENOENT:
- case EBADFD:/*Opened fd but brick is disconnected*/
- case EIO:/*Backend-fs crash like XFS/ext4 etc*/
- return _gf_true;
- }
- return _gf_false;
+ case EBADFD: /*Opened fd but brick is disconnected*/
+ case EIO: /*Backend-fs crash like XFS/ext4 etc*/
+ return _gf_true;
+ }
+ return _gf_false;
}
-void ec_complete(ec_fop_data_t * fop)
+void
+ec_complete(ec_fop_data_t *fop)
{
- ec_cbk_data_t * cbk = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t resume = 0, update = 0;
int healing_count = 0;
@@ -536,9 +547,9 @@ void ec_complete(ec_fop_data_t * fop)
if (fop->answer == NULL) {
if (!list_empty(&fop->cbk_list)) {
cbk = list_entry(fop->cbk_list.next, ec_cbk_data_t, list);
- healing_count = gf_bits_count (cbk->mask & fop->healing);
- /* fop shouldn't be treated as success if it is not
- * successful on at least fop->minimum good copies*/
+ healing_count = gf_bits_count(cbk->mask & fop->healing);
+ /* fop shouldn't be treated as success if it is not
+ * successful on at least fop->minimum good copies*/
if ((cbk->count - healing_count) >= fop->minimum) {
fop->answer = cbk;
@@ -560,8 +571,7 @@ void ec_complete(ec_fop_data_t * fop)
ec_update_good(fop, cbk->mask);
}
- if (resume)
- {
+ if (resume) {
ec_resume(fop, 0);
}
@@ -571,36 +581,95 @@ void ec_complete(ec_fop_data_t * fop)
/* There could be already granted locks sitting on the bricks, unlock for which
* must be wound at all costs*/
static gf_boolean_t
-ec_must_wind (ec_fop_data_t *fop)
+ec_must_wind(ec_fop_data_t *fop)
{
- if ((fop->id == GF_FOP_INODELK) || (fop->id == GF_FOP_FINODELK) ||
- (fop->id == GF_FOP_LK)) {
- if (fop->flock.l_type == F_UNLCK)
- return _gf_true;
- } else if ((fop->id == GF_FOP_ENTRYLK) ||
- (fop->id == GF_FOP_FENTRYLK)) {
- if (fop->entrylk_cmd == ENTRYLK_UNLOCK)
- return _gf_true;
- }
+ if ((fop->id == GF_FOP_INODELK) || (fop->id == GF_FOP_FINODELK) ||
+ (fop->id == GF_FOP_LK)) {
+ if (fop->flock.l_type == F_UNLCK)
+ return _gf_true;
+ } else if ((fop->id == GF_FOP_ENTRYLK) || (fop->id == GF_FOP_FENTRYLK)) {
+ if (fop->entrylk_cmd == ENTRYLK_UNLOCK)
+ return _gf_true;
+ }
- return _gf_false;
+ return _gf_false;
}
static gf_boolean_t
-ec_internal_op (ec_fop_data_t *fop)
-{
- if (ec_must_wind (fop))
- return _gf_true;
- if (fop->id == GF_FOP_XATTROP)
- return _gf_true;
- if (fop->id == GF_FOP_FXATTROP)
- return _gf_true;
- return _gf_false;
+ec_internal_op(ec_fop_data_t *fop)
+{
+ if (ec_must_wind(fop))
+ return _gf_true;
+ if (fop->id == GF_FOP_XATTROP)
+ return _gf_true;
+ if (fop->id == GF_FOP_FXATTROP)
+ return _gf_true;
+ if (fop->id == GF_FOP_OPEN)
+ return _gf_true;
+ return _gf_false;
+}
+
+char *
+ec_msg_str(ec_fop_data_t *fop)
+{
+ loc_t *loc1 = NULL;
+ loc_t *loc2 = NULL;
+ char gfid1[64] = {0};
+ char gfid2[64] = {0};
+ ec_fop_data_t *parent = fop->parent;
+
+ if (fop->errstr)
+ return fop->errstr;
+ if (!fop->use_fd) {
+ loc1 = &fop->loc[0];
+ loc2 = &fop->loc[1];
+
+ if (fop->id == GF_FOP_RENAME) {
+ gf_asprintf(&fop->errstr,
+ "FOP : '%s' failed on '%s' and '%s' with gfids "
+ "%s and %s respectively. Parent FOP: %s",
+ ec_fop_name(fop->id), loc1->path, loc2->path,
+ uuid_utoa_r(loc1->gfid, gfid1),
+ uuid_utoa_r(loc2->gfid, gfid2),
+ parent ? ec_fop_name(parent->id) : "No Parent");
+ } else {
+ gf_asprintf(
+ &fop->errstr,
+ "FOP : '%s' failed on '%s' with gfid %s. Parent FOP: %s",
+ ec_fop_name(fop->id), loc1->path,
+ uuid_utoa_r(loc1->gfid, gfid1),
+ parent ? ec_fop_name(parent->id) : "No Parent");
+ }
+ } else {
+ gf_asprintf(
+ &fop->errstr, "FOP : '%s' failed on gfid %s. Parent FOP: %s",
+ ec_fop_name(fop->id), uuid_utoa_r(fop->fd->inode->gfid, gfid1),
+ parent ? ec_fop_name(parent->id) : "No Parent");
+ }
+ return fop->errstr;
+}
+
+static void
+ec_log_insufficient_vol(ec_fop_data_t *fop, int32_t have, uint32_t need,
+ int32_t loglevel)
+{
+ ec_t *ec = fop->xl->private;
+ char str1[32], str2[32], str3[32];
+
+ gf_msg(ec->xl->name, loglevel, 0, EC_MSG_CHILDS_INSUFFICIENT,
+ "Insufficient available children for this request: "
+ "Have : %d, Need : %u : Child UP : %s "
+ "Mask: %s, Healing : %s : %s ",
+ have, need, ec_bin(str1, sizeof(str1), ec->xl_up, ec->nodes),
+ ec_bin(str2, sizeof(str2), fop->mask, ec->nodes),
+ ec_bin(str3, sizeof(str3), fop->healing, ec->nodes),
+ ec_msg_str(fop));
}
-int32_t ec_child_select(ec_fop_data_t * fop)
+static int32_t
+ec_child_select(ec_fop_data_t *fop)
{
- ec_t * ec = fop->xl->private;
+ ec_t *ec = fop->xl->private;
int32_t first = 0, num = 0;
ec_fop_cleanup(fop);
@@ -609,27 +678,25 @@ int32_t ec_child_select(ec_fop_data_t * fop)
/* Wind the fop on same subvols as parent for any internal extra fops like
* head/tail read in case of writev fop. Unlocks shouldn't do this because
* unlock should go on all subvols where lock is performed*/
- if (fop->parent && !ec_internal_op (fop)) {
- fop->mask &= (fop->parent->mask & ~fop->parent->healing);
+ if (fop->parent && !ec_internal_op(fop)) {
+ fop->mask &= (fop->parent->mask & ~fop->parent->healing);
+ if (ec_is_data_fop(fop->id)) {
+ fop->healing |= fop->parent->healing;
+ }
}
- if ((fop->mask & ~ec->xl_up) != 0)
- {
- gf_msg (fop->xl->name, GF_LOG_WARNING, 0,
- EC_MSG_OP_EXEC_UNAVAIL,
- "Executing operation with "
- "some subvolumes unavailable "
- "(%lX)", fop->mask & ~ec->xl_up);
-
+ if ((fop->mask & ~ec->xl_up) != 0) {
+ gf_msg(fop->xl->name, GF_LOG_WARNING, 0, EC_MSG_OP_EXEC_UNAVAIL,
+ "Executing operation with "
+ "some subvolumes unavailable. (%" PRIXPTR "). %s ",
+ fop->mask & ~ec->xl_up, ec_msg_str(fop));
fop->mask &= ec->xl_up;
}
- switch (fop->minimum)
- {
+ switch (fop->minimum) {
case EC_MINIMUM_ALL:
fop->minimum = gf_bits_count(fop->mask);
- if (fop->minimum >= ec->fragments)
- {
+ if (fop->minimum >= ec->fragments) {
break;
}
case EC_MINIMUM_MIN:
@@ -640,11 +707,11 @@ int32_t ec_child_select(ec_fop_data_t * fop)
}
if (ec->read_policy == EC_ROUND_ROBIN) {
- first = ec->idx;
- if (++first >= ec->nodes) {
- first = 0;
- }
- ec->idx = first;
+ first = ec->idx;
+ if (++first >= ec->nodes) {
+ first = 0;
+ }
+ ec->idx = first;
}
num = gf_bits_count(fop->mask);
@@ -655,26 +722,28 @@ int32_t ec_child_select(ec_fop_data_t * fop)
ec_trace("SELECT", fop, "");
- if ((num < fop->minimum) && (num < ec->fragments))
- {
- gf_msg (ec->xl->name, GF_LOG_ERROR, 0,
- EC_MSG_CHILDS_INSUFFICIENT,
- "Insufficient available children "
- "for this request (have %d, need "
- "%d)", num, fop->minimum);
-
+ if ((num < fop->minimum) && (num < ec->fragments)) {
+ ec_log_insufficient_vol(fop, num, fop->minimum, GF_LOG_ERROR);
return 0;
}
- ec_sleep(fop);
+ if (!fop->parent && fop->lock_count &&
+ (fop->locks[0].update[EC_DATA_TXN] ||
+ fop->locks[0].update[EC_METADATA_TXN])) {
+ if (ec->quorum_count && (num < ec->quorum_count)) {
+ ec_log_insufficient_vol(fop, num, ec->quorum_count, GF_LOG_ERROR);
+ return 0;
+ }
+ }
return 1;
}
-void ec_dispatch_next(ec_fop_data_t * fop, uint32_t idx)
+void
+ec_dispatch_next(ec_fop_data_t *fop, uint32_t idx)
{
uint32_t i = EC_INVALID_INDEX;
- ec_t * ec = fop->xl->private;
+ ec_t *ec = fop->xl->private;
LOCK(&fop->lock);
@@ -692,15 +761,15 @@ void ec_dispatch_next(ec_fop_data_t * fop, uint32_t idx)
UNLOCK(&fop->lock);
- if (i < EC_MAX_NODES)
- {
+ if (i < EC_MAX_NODES) {
fop->wind(ec, fop, idx);
}
}
-void ec_dispatch_mask(ec_fop_data_t * fop, uintptr_t mask)
+void
+ec_dispatch_mask(ec_fop_data_t *fop, uintptr_t mask)
{
- ec_t * ec = fop->xl->private;
+ ec_t *ec = fop->xl->private;
int32_t count, idx;
count = gf_bits_count(mask);
@@ -717,10 +786,8 @@ void ec_dispatch_mask(ec_fop_data_t * fop, uintptr_t mask)
UNLOCK(&fop->lock);
idx = 0;
- while (mask != 0)
- {
- if ((mask & 1) != 0)
- {
+ while (mask != 0) {
+ if ((mask & 1) != 0) {
fop->wind(ec, fop, idx);
}
idx++;
@@ -728,27 +795,29 @@ void ec_dispatch_mask(ec_fop_data_t * fop, uintptr_t mask)
}
}
-void ec_dispatch_start(ec_fop_data_t * fop)
+void
+ec_dispatch_start(ec_fop_data_t *fop)
{
fop->answer = NULL;
fop->good = 0;
INIT_LIST_HEAD(&fop->cbk_list);
- if (fop->lock_count > 0)
- {
+ if (fop->lock_count > 0) {
ec_owner_copy(fop->frame, &fop->req_frame->root->lk_owner);
}
}
-void ec_dispatch_one(ec_fop_data_t * fop)
+void
+ec_dispatch_one(ec_fop_data_t *fop)
{
ec_dispatch_start(fop);
- if (ec_child_select(fop))
- {
+ if (ec_child_select(fop)) {
+ ec_sleep(fop);
+
fop->expected = 1;
- fop->first = ec_select_first_by_read_policy (fop->xl->private, fop);
+ fop->first = ec_select_first_by_read_policy(fop->xl->private, fop);
ec_dispatch_next(fop, fop->first);
}
@@ -764,8 +833,8 @@ ec_dispatch_one_retry(ec_fop_data_t *fop, ec_cbk_data_t **cbk)
*cbk = tmp;
}
if ((tmp != NULL) && (tmp->op_ret < 0) &&
- ec_is_recoverable_error (tmp->op_errno)) {
- GF_ASSERT (fop->mask & (1ULL << tmp->idx));
+ ec_is_recoverable_error(tmp->op_errno)) {
+ GF_ASSERT(fop->mask & (1ULL << tmp->idx));
fop->mask ^= (1ULL << tmp->idx);
if (fop->mask) {
return _gf_true;
@@ -775,12 +844,14 @@ ec_dispatch_one_retry(ec_fop_data_t *fop, ec_cbk_data_t **cbk)
return _gf_false;
}
-void ec_dispatch_inc(ec_fop_data_t * fop)
+void
+ec_dispatch_inc(ec_fop_data_t *fop)
{
ec_dispatch_start(fop);
- if (ec_child_select(fop))
- {
+ if (ec_child_select(fop)) {
+ ec_sleep(fop);
+
fop->expected = gf_bits_count(fop->remaining);
fop->first = 0;
@@ -789,35 +860,38 @@ void ec_dispatch_inc(ec_fop_data_t * fop)
}
void
-ec_dispatch_all (ec_fop_data_t *fop)
+ec_dispatch_all(ec_fop_data_t *fop)
{
- ec_dispatch_start(fop);
+ ec_dispatch_start(fop);
- if (ec_child_select(fop)) {
- fop->expected = gf_bits_count(fop->remaining);
- fop->first = 0;
+ if (ec_child_select(fop)) {
+ ec_sleep(fop);
- ec_dispatch_mask(fop, fop->remaining);
- }
+ fop->expected = gf_bits_count(fop->remaining);
+ fop->first = 0;
+
+ ec_dispatch_mask(fop, fop->remaining);
+ }
}
-void ec_dispatch_min(ec_fop_data_t * fop)
+void
+ec_dispatch_min(ec_fop_data_t *fop)
{
- ec_t * ec = fop->xl->private;
+ ec_t *ec = fop->xl->private;
uintptr_t mask;
uint32_t idx;
int32_t count;
ec_dispatch_start(fop);
- if (ec_child_select(fop))
- {
+ if (ec_child_select(fop)) {
+ ec_sleep(fop);
+
fop->expected = count = ec->fragments;
- fop->first = ec_select_first_by_read_policy (fop->xl->private, fop);
+ fop->first = ec_select_first_by_read_policy(fop->xl->private, fop);
idx = fop->first - 1;
mask = 0;
- while (count-- > 0)
- {
+ while (count-- > 0) {
idx = ec_child_next(ec, fop, idx + 1);
if (idx < EC_MAX_NODES)
mask |= 1ULL << idx;
@@ -827,19 +901,35 @@ void ec_dispatch_min(ec_fop_data_t * fop)
}
}
-ec_lock_t *ec_lock_allocate(ec_fop_data_t *fop, loc_t *loc)
+void
+ec_succeed_all(ec_fop_data_t *fop)
+{
+ ec_dispatch_start(fop);
+
+ if (ec_child_select(fop)) {
+ fop->expected = gf_bits_count(fop->remaining);
+ fop->first = 0;
+
+ /* Simulate a successful execution on all bricks */
+ ec_trace("SUCCEED", fop, "");
+
+ fop->good = fop->remaining;
+ fop->remaining = 0;
+ }
+}
+
+ec_lock_t *
+ec_lock_allocate(ec_fop_data_t *fop, loc_t *loc)
{
ec_t *ec = fop->xl->private;
- ec_lock_t * lock;
+ ec_lock_t *lock;
int32_t err;
if ((loc->inode == NULL) ||
- (gf_uuid_is_null(loc->gfid) && gf_uuid_is_null(loc->inode->gfid)))
- {
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_INVALID_INODE,
- "Trying to lock based on an invalid "
- "inode");
+ (gf_uuid_is_null(loc->gfid) && gf_uuid_is_null(loc->inode->gfid))) {
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_INVALID_INODE,
+ "Trying to lock based on an invalid "
+ "inode");
__ec_fop_set_error(fop, EINVAL);
@@ -847,9 +937,8 @@ ec_lock_t *ec_lock_allocate(ec_fop_data_t *fop, loc_t *loc)
}
lock = mem_get0(ec->lock_pool);
- if (lock != NULL)
- {
- lock->good_mask = -1ULL;
+ if (lock != NULL) {
+ lock->good_mask = UINTPTR_MAX;
INIT_LIST_HEAD(&lock->owners);
INIT_LIST_HEAD(&lock->waiting);
INIT_LIST_HEAD(&lock->frozen);
@@ -865,7 +954,8 @@ ec_lock_t *ec_lock_allocate(ec_fop_data_t *fop, loc_t *loc)
return lock;
}
-void ec_lock_destroy(ec_lock_t * lock)
+void
+ec_lock_destroy(ec_lock_t *lock)
{
loc_wipe(&lock->loc);
if (lock->fd != NULL) {
@@ -875,13 +965,15 @@ void ec_lock_destroy(ec_lock_t * lock)
mem_put(lock);
}
-int32_t ec_lock_compare(ec_lock_t * lock1, ec_lock_t * lock2)
+int32_t
+ec_lock_compare(ec_lock_t *lock1, ec_lock_t *lock2)
{
return gf_uuid_compare(lock1->loc.gfid, lock2->loc.gfid);
}
-void ec_lock_insert(ec_fop_data_t *fop, ec_lock_t *lock, uint32_t flags,
- loc_t *base, off_t fl_start, size_t fl_size)
+static void
+ec_lock_insert(ec_fop_data_t *fop, ec_lock_t *lock, uint32_t flags, loc_t *base,
+ off_t fl_start, uint64_t fl_size)
{
ec_lock_link_t *link;
@@ -916,14 +1008,14 @@ void ec_lock_insert(ec_fop_data_t *fop, ec_lock_t *lock, uint32_t flags,
link->update[EC_METADATA_TXN] = (flags & EC_UPDATE_META) != 0;
link->base = base;
link->fl_start = fl_start;
- link->fl_end = ec_range_end_get (fl_start, fl_size);
+ link->fl_end = ec_range_end_get(fl_start, fl_size);
lock->refs_pending++;
}
-void ec_lock_prepare_inode_internal(ec_fop_data_t *fop, loc_t *loc,
- uint32_t flags, loc_t *base,
- off_t fl_start, size_t fl_size)
+static void
+ec_lock_prepare_inode_internal(ec_fop_data_t *fop, loc_t *loc, uint32_t flags,
+ loc_t *base, off_t fl_start, uint64_t fl_size)
{
ec_lock_t *lock = NULL;
ec_inode_t *ctx;
@@ -952,8 +1044,8 @@ void ec_lock_prepare_inode_internal(ec_fop_data_t *fop, loc_t *loc,
if ((fop->lock_count > 0) && (fop->locks[0].lock == lock)) {
/* Combine data/meta updates */
fop->locks[0].update[EC_DATA_TXN] |= (flags & EC_UPDATE_DATA) != 0;
- fop->locks[0].update[EC_METADATA_TXN] |=
- (flags & EC_UPDATE_META) != 0;
+ fop->locks[0].update[EC_METADATA_TXN] |= (flags & EC_UPDATE_META) !=
+ 0;
/* Only one base inode is allowed per fop, so there shouldn't be
* overwrites here. */
@@ -964,8 +1056,10 @@ void ec_lock_prepare_inode_internal(ec_fop_data_t *fop, loc_t *loc,
goto update_query;
}
- ec_trace("LOCK_INODELK", fop, "lock=%p, inode=%p. Lock already "
- "acquired", lock, loc->inode);
+ ec_trace("LOCK_INODELK", fop,
+ "lock=%p, inode=%p. Lock already "
+ "acquired",
+ lock, loc->inode);
goto insert;
}
@@ -991,14 +1085,16 @@ unlock:
UNLOCK(&loc->inode->lock);
}
-void ec_lock_prepare_inode(ec_fop_data_t *fop, loc_t *loc, uint32_t flags,
- off_t fl_start, size_t fl_size)
+void
+ec_lock_prepare_inode(ec_fop_data_t *fop, loc_t *loc, uint32_t flags,
+ off_t fl_start, uint64_t fl_size)
{
ec_lock_prepare_inode_internal(fop, loc, flags, NULL, fl_start, fl_size);
}
-void ec_lock_prepare_parent_inode(ec_fop_data_t *fop, loc_t *loc, loc_t *base,
- uint32_t flags)
+void
+ec_lock_prepare_parent_inode(ec_fop_data_t *fop, loc_t *loc, loc_t *base,
+ uint32_t flags)
{
loc_t tmp;
int32_t err;
@@ -1017,16 +1113,17 @@ void ec_lock_prepare_parent_inode(ec_fop_data_t *fop, loc_t *loc, loc_t *base,
if ((flags & EC_INODE_SIZE) != 0) {
flags ^= EC_INODE_SIZE;
} else {
- base = NULL;
+ base = NULL;
}
- ec_lock_prepare_inode_internal(fop, &tmp, flags, base, 0, LLONG_MAX);
+ ec_lock_prepare_inode_internal(fop, &tmp, flags, base, 0, EC_RANGE_FULL);
loc_wipe(&tmp);
}
-void ec_lock_prepare_fd(ec_fop_data_t *fop, fd_t *fd, uint32_t flags,
- off_t fl_start, size_t fl_size)
+void
+ec_lock_prepare_fd(ec_fop_data_t *fop, fd_t *fd, uint32_t flags, off_t fl_start,
+ uint64_t fl_size)
{
loc_t loc;
int32_t err;
@@ -1048,15 +1145,14 @@ void ec_lock_prepare_fd(ec_fop_data_t *fop, fd_t *fd, uint32_t flags,
}
gf_boolean_t
-ec_config_check (xlator_t *xl, ec_config_t *config)
+ec_config_check(xlator_t *xl, ec_config_t *config)
{
ec_t *ec;
ec = xl->private;
if ((config->version != EC_CONFIG_VERSION) ||
(config->algorithm != EC_CONFIG_ALGORITHM) ||
- (config->gf_word_size != EC_GF_BITS) ||
- (config->bricks != ec->nodes) ||
+ (config->gf_word_size != EC_GF_BITS) || (config->bricks != ec->nodes) ||
(config->redundancy != ec->redundancy) ||
(config->chunk_size != EC_METHOD_CHUNK_SIZE)) {
uint32_t data_bricks;
@@ -1075,20 +1171,17 @@ ec_config_check (xlator_t *xl, ec_config_t *config)
if ((config->redundancy < 1) ||
(config->redundancy * 2 >= config->bricks) ||
!ec_is_power_of_2(config->gf_word_size) ||
- ((config->chunk_size * 8) % (config->gf_word_size * data_bricks)
- != 0)) {
- gf_msg (xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_INVALID_CONFIG,
- "Invalid or corrupted config");
+ ((config->chunk_size * 8) % (config->gf_word_size * data_bricks) !=
+ 0)) {
+ gf_msg(xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_INVALID_CONFIG,
+ "Invalid or corrupted config");
} else {
- gf_msg (xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_INVALID_CONFIG,
- "Unsupported config "
- "(V=%u, A=%u, W=%u, "
- "N=%u, R=%u, S=%u)",
- config->version, config->algorithm,
- config->gf_word_size, config->bricks,
- config->redundancy, config->chunk_size);
+ gf_msg(xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_INVALID_CONFIG,
+ "Unsupported config "
+ "(V=%u, A=%u, W=%u, "
+ "N=%u, R=%u, S=%u)",
+ config->version, config->algorithm, config->gf_word_size,
+ config->bricks, config->redundancy, config->chunk_size);
}
return _gf_false;
@@ -1098,20 +1191,18 @@ ec_config_check (xlator_t *xl, ec_config_t *config)
}
gf_boolean_t
-ec_set_dirty_flag (ec_lock_link_t *link, ec_inode_t *ctx,
- uint64_t *dirty)
+ec_set_dirty_flag(ec_lock_link_t *link, ec_inode_t *ctx, uint64_t *dirty)
{
-
gf_boolean_t set_dirty = _gf_false;
if (link->update[EC_DATA_TXN] && !ctx->dirty[EC_DATA_TXN]) {
- if (!link->optimistic_changelog)
- dirty[EC_DATA_TXN] = 1;
+ if (!link->optimistic_changelog)
+ dirty[EC_DATA_TXN] = 1;
}
if (link->update[EC_METADATA_TXN] && !ctx->dirty[EC_METADATA_TXN]) {
- if (!link->optimistic_changelog)
- dirty[EC_METADATA_TXN] = 1;
+ if (!link->optimistic_changelog)
+ dirty[EC_METADATA_TXN] = 1;
}
if (dirty[EC_METADATA_TXN] || dirty[EC_DATA_TXN]) {
@@ -1122,9 +1213,9 @@ ec_set_dirty_flag (ec_lock_link_t *link, ec_inode_t *ctx,
}
int32_t
-ec_prepare_update_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
+ec_prepare_update_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
struct list_head list;
ec_fop_data_t *fop = cookie, *parent, *tmp;
@@ -1135,7 +1226,6 @@ ec_prepare_update_cbk (call_frame_t *frame, void *cookie,
gf_boolean_t release = _gf_false;
uint64_t provided_flags = 0;
uint64_t dirty[EC_VERSION_SIZE] = {0, 0};
-
lock = parent_link->lock;
parent = parent_link->fop;
ctx = lock->ctx;
@@ -1145,92 +1235,87 @@ ec_prepare_update_cbk (call_frame_t *frame, void *cookie,
LOCK(&lock->loc.inode->lock);
- list_for_each_entry(link, &lock->owners, owner_list) {
+ list_for_each_entry(link, &lock->owners, owner_list)
+ {
if ((link->waiting_flags & provided_flags) != 0) {
link->waiting_flags ^= (link->waiting_flags & provided_flags);
if (EC_NEEDED_FLAGS(link->waiting_flags) == 0)
- list_add_tail(&link->fop->cbk_list, &list);
+ list_add_tail(&link->fop->cbk_list, &list);
}
}
-
if (op_ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- EC_MSG_SIZE_VERS_GET_FAIL,
- "Failed to get size and version");
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, EC_MSG_SIZE_VERS_GET_FAIL,
+ "Failed to get size and version : %s", ec_msg_str(fop));
goto unlock;
}
if (EC_FLAGS_HAVE(provided_flags, EC_FLAG_XATTROP)) {
- op_errno = -ec_dict_del_array(dict, EC_XATTR_VERSION,
- ctx->pre_version,
- EC_VERSION_SIZE);
+ op_errno = -ec_dict_del_array(dict, EC_XATTR_VERSION, ctx->pre_version,
+ EC_VERSION_SIZE);
+ if (op_errno != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno,
+ EC_MSG_VER_XATTR_GET_FAIL, "Unable to get version xattr. %s",
+ ec_msg_str(fop));
+ goto unlock;
+ }
+ ctx->post_version[0] += ctx->pre_version[0];
+ ctx->post_version[1] += ctx->pre_version[1];
+
+ ctx->have_version = _gf_true;
+
+ if (lock->loc.inode->ia_type == IA_IFREG ||
+ lock->loc.inode->ia_type == IA_INVAL) {
+ op_errno = -ec_dict_del_number(dict, EC_XATTR_SIZE, &ctx->pre_size);
if (op_errno != 0) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- EC_MSG_VER_XATTR_GET_FAIL,
- "Unable to get version xattr");
- goto unlock;
+ if (lock->loc.inode->ia_type == IA_IFREG) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno,
+ EC_MSG_SIZE_XATTR_GET_FAIL,
+ "Unable to get size xattr. %s", ec_msg_str(fop));
+ goto unlock;
+ }
+ } else {
+ ctx->post_size = ctx->pre_size;
+
+ ctx->have_size = _gf_true;
}
- ctx->post_version[0] += ctx->pre_version[0];
- ctx->post_version[1] += ctx->pre_version[1];
- ctx->have_version = _gf_true;
+ op_errno = -ec_dict_del_config(dict, EC_XATTR_CONFIG, &ctx->config);
+ if (op_errno != 0) {
+ if ((lock->loc.inode->ia_type == IA_IFREG) ||
+ (op_errno != ENODATA)) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno,
+ EC_MSG_CONFIG_XATTR_GET_FAIL,
+ "Unable to get config xattr. %s", ec_msg_str(fop));
- if (lock->loc.inode->ia_type == IA_IFREG ||
- lock->loc.inode->ia_type == IA_INVAL) {
- op_errno = -ec_dict_del_number(dict, EC_XATTR_SIZE,
- &ctx->pre_size);
- if (op_errno != 0) {
- if (lock->loc.inode->ia_type == IA_IFREG) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- EC_MSG_SIZE_XATTR_GET_FAIL,
- "Unable to get size xattr");
- goto unlock;
- }
- } else {
- ctx->post_size = ctx->pre_size;
-
- ctx->have_size = _gf_true;
+ goto unlock;
}
+ } else {
+ if (!ec_config_check(parent->xl, &ctx->config)) {
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL,
+ EC_MSG_CONFIG_XATTR_INVALID, "Invalid config xattr");
+
+ op_errno = EINVAL;
- op_errno = -ec_dict_del_config(dict, EC_XATTR_CONFIG,
- &ctx->config);
- if (op_errno != 0) {
- if ((lock->loc.inode->ia_type == IA_IFREG) ||
- (op_errno != ENODATA)) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- EC_MSG_CONFIG_XATTR_GET_FAIL,
- "Unable to get config xattr");
-
- goto unlock;
- }
- } else {
- if (!ec_config_check(parent->xl, &ctx->config)) {
- gf_msg (this->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_CONFIG_XATTR_INVALID,
- "Invalid config xattr");
-
- op_errno = EINVAL;
-
- goto unlock;
- }
- ctx->have_config = _gf_true;
+ goto unlock;
}
+ ctx->have_config = _gf_true;
}
- ctx->have_info = _gf_true;
+ }
+ ctx->have_info = _gf_true;
}
- ec_set_dirty_flag (fop->data, ctx, dirty);
+ ec_set_dirty_flag(fop->data, ctx, dirty);
if (dirty[EC_METADATA_TXN] &&
(EC_FLAGS_HAVE(provided_flags, EC_FLAG_METADATA_DIRTY))) {
- GF_ASSERT (!ctx->dirty[EC_METADATA_TXN]);
- ctx->dirty[EC_METADATA_TXN] = 1;
+ GF_ASSERT(!ctx->dirty[EC_METADATA_TXN]);
+ ctx->dirty[EC_METADATA_TXN] = 1;
}
if (dirty[EC_DATA_TXN] &&
(EC_FLAGS_HAVE(provided_flags, EC_FLAG_DATA_DIRTY))) {
- GF_ASSERT (!ctx->dirty[EC_DATA_TXN]);
- ctx->dirty[EC_DATA_TXN] = 1;
+ GF_ASSERT(!ctx->dirty[EC_DATA_TXN]);
+ ctx->dirty[EC_DATA_TXN] = 1;
}
op_errno = 0;
unlock:
@@ -1242,20 +1327,20 @@ unlock:
* it dirty and update versions right away if dirty was not set before.
*/
if (lock->good_mask & ~(fop->good | fop->remaining)) {
- release = _gf_true;
+ release = _gf_true;
}
if (parent_link->update[0] && !parent_link->dirty[0]) {
- lock->release |= release;
+ lock->release |= release;
}
if (parent_link->update[1] && !parent_link->dirty[1]) {
- lock->release |= release;
+ lock->release |= release;
}
/* We don't allow the main fop to be executed on bricks that have not
* succeeded the initial xattrop. */
- ec_lock_update_good (lock, fop);
+ ec_lock_update_good(lock, fop);
/*As of now only data healing marks bricks as healing*/
lock->healing |= fop->healing;
@@ -1271,7 +1356,7 @@ unlock:
tmp->mask &= fop->good;
/*As of now only data healing marks bricks as healing*/
- if (ec_is_data_fop (tmp->id)) {
+ if (ec_is_data_fop(tmp->id)) {
tmp->healing |= fop->healing;
}
}
@@ -1285,52 +1370,53 @@ unlock:
static gf_boolean_t
ec_set_needed_flag(ec_lock_t *lock, ec_lock_link_t *link, uint64_t flag)
{
- uint64_t current;
+ uint64_t current;
- link->waiting_flags |= EC_FLAG_NEEDS(flag);
+ link->waiting_flags |= EC_FLAG_NEEDS(flag);
- current = EC_NEEDED_FLAGS(lock->waiting_flags);
- if (!EC_FLAGS_HAVE(current, flag)) {
- lock->waiting_flags |= EC_FLAG_NEEDS(flag);
- link->waiting_flags |= EC_FLAG_PROVIDES(flag);
+ current = EC_NEEDED_FLAGS(lock->waiting_flags);
+ if (!EC_FLAGS_HAVE(current, flag)) {
+ lock->waiting_flags |= EC_FLAG_NEEDS(flag);
+ link->waiting_flags |= EC_FLAG_PROVIDES(flag);
- return _gf_true;
- }
+ return _gf_true;
+ }
- return _gf_false;
+ return _gf_false;
}
static uint64_t
-ec_set_xattrop_flags_and_params (ec_lock_t *lock, ec_lock_link_t *link,
- uint64_t *dirty)
+ec_set_xattrop_flags_and_params(ec_lock_t *lock, ec_lock_link_t *link,
+ uint64_t *dirty)
{
- uint64_t oldflags = 0;
- uint64_t newflags = 0;
- ec_inode_t *ctx = lock->ctx;
+ uint64_t oldflags = 0;
+ uint64_t newflags = 0;
+ ec_inode_t *ctx = lock->ctx;
- oldflags = EC_NEEDED_FLAGS(lock->waiting_flags);
+ oldflags = EC_NEEDED_FLAGS(lock->waiting_flags);
- if (lock->query && !ctx->have_info) {
- ec_set_needed_flag(lock, link, EC_FLAG_XATTROP);
- }
+ if (lock->query && !ctx->have_info) {
+ ec_set_needed_flag(lock, link, EC_FLAG_XATTROP);
+ }
- if (dirty[EC_DATA_TXN]) {
- if (!ec_set_needed_flag(lock, link, EC_FLAG_DATA_DIRTY)) {
- dirty[EC_DATA_TXN] = 0;
- }
+ if (dirty[EC_DATA_TXN]) {
+ if (!ec_set_needed_flag(lock, link, EC_FLAG_DATA_DIRTY)) {
+ dirty[EC_DATA_TXN] = 0;
}
+ }
- if (dirty[EC_METADATA_TXN]) {
- if (!ec_set_needed_flag(lock, link, EC_FLAG_METADATA_DIRTY)) {
- dirty[EC_METADATA_TXN] = 0;
- }
+ if (dirty[EC_METADATA_TXN]) {
+ if (!ec_set_needed_flag(lock, link, EC_FLAG_METADATA_DIRTY)) {
+ dirty[EC_METADATA_TXN] = 0;
}
- newflags = EC_NEEDED_FLAGS(lock->waiting_flags);
+ }
+ newflags = EC_NEEDED_FLAGS(lock->waiting_flags);
- return oldflags ^ newflags;
+ return oldflags ^ newflags;
}
-void ec_get_size_version(ec_lock_link_t *link)
+void
+ec_get_size_version(ec_lock_link_t *link)
{
loc_t loc;
ec_lock_t *lock;
@@ -1338,7 +1424,7 @@ void ec_get_size_version(ec_lock_link_t *link)
ec_fop_data_t *fop;
dict_t *dict = NULL;
dict_t *xdata = NULL;
- ec_t *ec = NULL;
+ ec_t *ec = NULL;
int32_t error = 0;
gf_boolean_t set_dirty = _gf_false;
uint64_t allzero[EC_VERSION_SIZE] = {0, 0};
@@ -1346,45 +1432,46 @@ void ec_get_size_version(ec_lock_link_t *link)
lock = link->lock;
ctx = lock->ctx;
fop = link->fop;
- ec = fop->xl->private;
+ ec = fop->xl->private;
uint64_t changed_flags = 0;
- if (ec->optimistic_changelog &&
- !(ec->node_mask & ~link->lock->good_mask) && !ec_is_data_fop (fop->id))
- link->optimistic_changelog = _gf_true;
+ if (ec->optimistic_changelog && !(ec->node_mask & ~link->lock->good_mask) &&
+ !ec_is_data_fop(fop->id))
+ link->optimistic_changelog = _gf_true;
- set_dirty = ec_set_dirty_flag (link, ctx, dirty);
+ memset(&loc, 0, sizeof(loc));
+
+ LOCK(&lock->loc.inode->lock);
+
+ set_dirty = ec_set_dirty_flag(link, ctx, dirty);
/* If ec metadata has already been retrieved, do not try again. */
- if (ctx->have_info && (!set_dirty)) {
- if (ec_is_data_fop (fop->id)) {
+ if (ctx->have_info) {
+ if (ec_is_data_fop(fop->id)) {
fop->healing |= lock->healing;
}
- return;
+ if (!set_dirty)
+ goto unlock;
}
/* Determine if there's something we need to retrieve for the current
* operation. */
- if (!set_dirty && !lock->query &&
- (lock->loc.inode->ia_type != IA_IFREG) &&
+ if (!set_dirty && !lock->query && (lock->loc.inode->ia_type != IA_IFREG) &&
(lock->loc.inode->ia_type != IA_INVAL)) {
- return;
+ goto unlock;
}
- memset(&loc, 0, sizeof(loc));
-
- LOCK(&lock->loc.inode->lock);
-
- changed_flags = ec_set_xattrop_flags_and_params (lock, link, dirty);
+ changed_flags = ec_set_xattrop_flags_and_params(lock, link, dirty);
if (link->waiting_flags) {
- /* This fop needs to wait until all its flags are cleared which
- * potentially can be cleared by other xattrops that are already
- * wound*/
- ec_sleep(fop);
+ /* This fop needs to wait until all its flags are cleared which
+ * potentially can be cleared by other xattrops that are already
+ * wound*/
+ ec_sleep(fop);
} else {
- GF_ASSERT (!changed_flags);
+ GF_ASSERT(!changed_flags);
}
+unlock:
UNLOCK(&lock->loc.inode->lock);
if (!changed_flags)
@@ -1397,40 +1484,38 @@ void ec_get_size_version(ec_lock_link_t *link)
}
if (EC_FLAGS_HAVE(changed_flags, EC_FLAG_XATTROP)) {
- /* Once we know that an xattrop will be needed,
- * we try to get all available information in a
- * single call. */
- error = ec_dict_set_array(dict, EC_XATTR_VERSION, allzero,
- EC_VERSION_SIZE);
+ /* Once we know that an xattrop will be needed,
+ * we try to get all available information in a
+ * single call. */
+ error = ec_dict_set_array(dict, EC_XATTR_VERSION, allzero,
+ EC_VERSION_SIZE);
+ if (error != 0) {
+ goto out;
+ }
+
+ if (lock->loc.inode->ia_type == IA_IFREG ||
+ lock->loc.inode->ia_type == IA_INVAL) {
+ error = ec_dict_set_number(dict, EC_XATTR_SIZE, 0);
+ if (error == 0) {
+ error = ec_dict_set_number(dict, EC_XATTR_CONFIG, 0);
+ }
if (error != 0) {
goto out;
}
- if (lock->loc.inode->ia_type == IA_IFREG ||
- lock->loc.inode->ia_type == IA_INVAL) {
- error = ec_dict_set_number(dict, EC_XATTR_SIZE, 0);
- if (error == 0) {
- error = ec_dict_set_number(dict, EC_XATTR_CONFIG, 0);
- }
- if (error != 0) {
- goto out;
- }
-
- xdata = dict_new();
- if (xdata == NULL || dict_set_int32 (xdata, GF_GET_SIZE, 1)) {
- error = -ENOMEM;
- goto out;
- }
-
+ xdata = dict_new();
+ if (xdata == NULL || dict_set_int32(xdata, GF_GET_SIZE, 1)) {
+ error = -ENOMEM;
+ goto out;
}
+ }
}
- if (memcmp (allzero, dirty, sizeof (allzero))) {
- error = ec_dict_set_array(dict, EC_XATTR_DIRTY, dirty,
- EC_VERSION_SIZE);
- if (error != 0) {
- goto out;
- }
+ if (memcmp(allzero, dirty, sizeof(allzero))) {
+ error = ec_dict_set_array(dict, EC_XATTR_DIRTY, dirty, EC_VERSION_SIZE);
+ if (error != 0) {
+ goto out;
+ }
}
fop->frame->root->uid = 0;
@@ -1458,13 +1543,13 @@ void ec_get_size_version(ec_lock_link_t *link)
loc.name = NULL;
}
- ec_xattrop (fop->frame, fop->xl, fop->mask, fop->minimum,
- ec_prepare_update_cbk, link, &loc,
- GF_XATTROP_ADD_ARRAY64, dict, xdata);
+ ec_xattrop(fop->frame, fop->xl, fop->mask, fop->minimum,
+ ec_prepare_update_cbk, link, &loc, GF_XATTROP_ADD_ARRAY64,
+ dict, xdata);
} else {
ec_fxattrop(fop->frame, fop->xl, fop->mask, fop->minimum,
- ec_prepare_update_cbk, link, lock->fd,
- GF_XATTROP_ADD_ARRAY64, dict, xdata);
+ ec_prepare_update_cbk, link, lock->fd,
+ GF_XATTROP_ADD_ARRAY64, dict, xdata);
}
error = 0;
@@ -1489,8 +1574,7 @@ out:
}
gf_boolean_t
-__ec_get_inode_size(ec_fop_data_t *fop, inode_t *inode,
- uint64_t *size)
+__ec_get_inode_size(ec_fop_data_t *fop, inode_t *inode, uint64_t *size)
{
ec_inode_t *ctx;
gf_boolean_t found = _gf_false;
@@ -1510,14 +1594,13 @@ out:
}
gf_boolean_t
-ec_get_inode_size(ec_fop_data_t *fop, inode_t *inode,
- uint64_t *size)
+ec_get_inode_size(ec_fop_data_t *fop, inode_t *inode, uint64_t *size)
{
gf_boolean_t found = _gf_false;
LOCK(&inode->lock);
{
- found = __ec_get_inode_size (fop, inode, size);
+ found = __ec_get_inode_size(fop, inode, size);
}
UNLOCK(&inode->lock);
@@ -1525,8 +1608,7 @@ ec_get_inode_size(ec_fop_data_t *fop, inode_t *inode,
}
gf_boolean_t
-__ec_set_inode_size(ec_fop_data_t *fop, inode_t *inode,
- uint64_t size)
+__ec_set_inode_size(ec_fop_data_t *fop, inode_t *inode, uint64_t size)
{
ec_inode_t *ctx;
gf_boolean_t found = _gf_false;
@@ -1553,38 +1635,37 @@ out:
}
gf_boolean_t
-ec_set_inode_size(ec_fop_data_t *fop, inode_t *inode,
- uint64_t size)
+ec_set_inode_size(ec_fop_data_t *fop, inode_t *inode, uint64_t size)
{
gf_boolean_t found = _gf_false;
- LOCK (&inode->lock);
+ LOCK(&inode->lock);
{
- found = __ec_set_inode_size (fop, inode, size);
+ found = __ec_set_inode_size(fop, inode, size);
}
- UNLOCK (&inode->lock);
+ UNLOCK(&inode->lock);
return found;
}
static void
-ec_release_stripe_cache (ec_inode_t *ctx)
+ec_release_stripe_cache(ec_inode_t *ctx)
{
- ec_stripe_list_t *stripe_cache = NULL;
- ec_stripe_t *stripe = NULL;
+ ec_stripe_list_t *stripe_cache = NULL;
+ ec_stripe_t *stripe = NULL;
- stripe_cache = &ctx->stripe_cache;
- while (!list_empty (&stripe_cache->lru)) {
- stripe = list_first_entry (&stripe_cache->lru, ec_stripe_t,
- lru);
- list_del (&stripe->lru);
- GF_FREE (stripe);
- }
- stripe_cache->count = 0;
- stripe_cache->max = 0;
+ stripe_cache = &ctx->stripe_cache;
+ while (!list_empty(&stripe_cache->lru)) {
+ stripe = list_first_entry(&stripe_cache->lru, ec_stripe_t, lru);
+ list_del(&stripe->lru);
+ GF_FREE(stripe);
+ }
+ stripe_cache->count = 0;
+ stripe_cache->max = 0;
}
-void ec_clear_inode_info(ec_fop_data_t *fop, inode_t *inode)
+void
+ec_clear_inode_info(ec_fop_data_t *fop, inode_t *inode)
{
ec_inode_t *ctx;
@@ -1595,7 +1676,7 @@ void ec_clear_inode_info(ec_fop_data_t *fop, inode_t *inode)
goto unlock;
}
- ec_release_stripe_cache (ctx);
+ ec_release_stripe_cache(ctx);
ctx->have_info = _gf_false;
ctx->have_config = _gf_false;
ctx->have_version = _gf_false;
@@ -1611,10 +1692,10 @@ unlock:
UNLOCK(&inode->lock);
}
-int32_t ec_get_real_size_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata,
- struct iatt *postparent)
+int32_t
+ec_get_real_size_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata, struct iatt *postparent)
{
ec_fop_data_t *fop = cookie;
ec_lock_link_t *link;
@@ -1631,14 +1712,15 @@ int32_t ec_get_real_size_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
}
/* This function is used to get the trusted.ec.size xattr from a file when
- * no lock is needed on the inode. This is only required to maintan iatt
+ * no lock is needed on the inode. This is only required to maintain iatt
* structs on fops that manipulate directory entries but do not operate
* directly on the inode, like link, rename, ...
*
* Any error processing this request is ignored. In the worst case, an invalid
* or not up to date value in the iatt could cause some cache invalidation.
*/
-void ec_get_real_size(ec_lock_link_t *link)
+void
+ec_get_real_size(ec_lock_link_t *link)
{
ec_fop_data_t *fop;
dict_t *xdata;
@@ -1688,24 +1770,26 @@ ec_lock_update_fd(ec_lock_t *lock, ec_fop_data_t *fop)
}
static gf_boolean_t
-ec_link_has_lock_conflict (ec_lock_link_t *link, gf_boolean_t waitlist_check)
+ec_link_has_lock_conflict(ec_lock_link_t *link, gf_boolean_t waitlist_check)
{
- ec_lock_link_t *trav_link = NULL;
+ ec_lock_link_t *trav_link = NULL;
- list_for_each_entry (trav_link, &link->lock->owners, owner_list) {
- if (ec_lock_conflict (trav_link, link))
- return _gf_true;
- }
+ list_for_each_entry(trav_link, &link->lock->owners, owner_list)
+ {
+ if (ec_lock_conflict(trav_link, link))
+ return _gf_true;
+ }
- if (!waitlist_check)
- return _gf_false;
+ if (!waitlist_check)
+ return _gf_false;
- list_for_each_entry (trav_link, &link->lock->waiting, wait_list) {
- if (ec_lock_conflict (trav_link, link))
- return _gf_true;
- }
+ list_for_each_entry(trav_link, &link->lock->waiting, wait_list)
+ {
+ if (ec_lock_conflict(trav_link, link))
+ return _gf_true;
+ }
- return _gf_false;
+ return _gf_false;
}
static void
@@ -1726,7 +1810,7 @@ ec_lock_wake_shared(ec_lock_t *lock, struct list_head *list)
/* If the fop is not shareable, only this fop can be assigned as owner.
* Other fops will need to wait until this one finishes. */
- if (ec_link_has_lock_conflict (link, _gf_false)) {
+ if (ec_link_has_lock_conflict(link, _gf_false)) {
conflict = _gf_true;
}
@@ -1757,7 +1841,8 @@ ec_lock_apply(ec_lock_link_t *link)
ec_get_real_size(link);
}
-gf_boolean_t ec_lock_acquire(ec_lock_link_t *link);
+gf_boolean_t
+ec_lock_acquire(ec_lock_link_t *link);
static void
ec_lock_resume_shared(struct list_head *list)
@@ -1781,7 +1866,8 @@ ec_lock_resume_shared(struct list_head *list)
}
}
-void ec_lock_acquired(ec_lock_link_t *link)
+void
+ec_lock_acquired(ec_lock_link_t *link)
{
struct list_head list;
ec_lock_t *lock;
@@ -1797,6 +1883,10 @@ void ec_lock_acquired(ec_lock_link_t *link)
LOCK(&lock->loc.inode->lock);
lock->acquired = _gf_true;
+ if (lock->contention) {
+ lock->release = _gf_true;
+ lock->contention = _gf_false;
+ }
ec_lock_update_fd(lock, fop);
ec_lock_wake_shared(lock, &list);
@@ -1807,37 +1897,44 @@ void ec_lock_acquired(ec_lock_link_t *link)
if (fop->use_fd &&
(link->update[EC_DATA_TXN] || link->update[EC_METADATA_TXN])) {
- ec_fix_open(fop);
+ /* Try to reopen closed fd's only if lock has succeeded. */
+ ec_fix_open(fop, lock->mask);
}
ec_lock_resume_shared(&list);
}
-int32_t ec_locked(call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+int32_t
+ec_locked(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
ec_fop_data_t *fop = cookie;
ec_lock_link_t *link = NULL;
ec_lock_t *lock = NULL;
+ link = fop->data;
+ lock = link->lock;
if (op_ret >= 0) {
- link = fop->data;
- lock = link->lock;
lock->mask = lock->good_mask = fop->good;
lock->healing = 0;
ec_lock_acquired(link);
ec_lock(fop->parent);
} else {
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- EC_MSG_PREOP_LOCK_FAILED,
- "Failed to complete preop lock");
+ LOCK(&lock->loc.inode->lock);
+ {
+ lock->contention = _gf_false;
+ }
+ UNLOCK(&lock->loc.inode->lock);
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, EC_MSG_PREOP_LOCK_FAILED,
+ "Failed to complete preop lock");
}
return 0;
}
-gf_boolean_t ec_lock_acquire(ec_lock_link_t *link)
+gf_boolean_t
+ec_lock_acquire(ec_lock_link_t *link)
{
ec_lock_t *lock;
ec_fop_data_t *fop;
@@ -1870,62 +1967,62 @@ gf_boolean_t ec_lock_acquire(ec_lock_link_t *link)
static ec_lock_link_t *
ec_lock_timer_cancel(xlator_t *xl, ec_lock_t *lock)
{
- ec_lock_link_t *timer_link;
-
- /* If we don't have any timer, there's nothing to cancel. */
- if (lock->timer == NULL) {
- return NULL;
- }
-
- /* We are trying to access a lock that has an unlock timer active.
- * This means that the lock must be idle, i.e. no fop can be in the
- * owner, waiting or frozen lists. It also means that the lock cannot
- * have been marked as being released (this is done without timers).
- * There should only be one owner reference, but it's possible that
- * some fops are being prepared to use this lock. */
- GF_ASSERT ((lock->refs_owners == 1) &&
- list_empty(&lock->owners) && list_empty(&lock->waiting));
-
- /* We take the timer_link before cancelling the timer, since a
- * successful cancellation will destroy it. It must not be NULL
- * because it references the fop responsible for the delayed unlock
- * that we are currently trying to cancel. */
- timer_link = lock->timer->data;
- GF_ASSERT(timer_link != NULL);
-
- if (gf_timer_call_cancel(xl->ctx, lock->timer) < 0) {
- /* It's too late to avoid the execution of the timer callback.
- * Since we need to be sure that the callback has access to all
- * needed resources, we cannot resume the execution of the
- * timer fop now. This will be done in the callback. */
- timer_link = NULL;
- } else {
- /* The timer has been cancelled. The fop referenced by
- * timer_link holds the last reference. The caller is
- * responsible to release it when not needed anymore. */
- ec_trace("UNLOCK_CANCELLED", timer_link->fop, "lock=%p", lock);
- }
+ ec_lock_link_t *timer_link;
- /* We have two options here:
- *
- * 1. The timer has been successfully cancelled.
- *
- * This is the easiest case and we can continue with the currently
- * acquired lock.
- *
- * 2. The timer callback has already been fired.
- *
- * In this case we have not been able to cancel the timer before
- * the timer callback has been fired, but we also know that
- * lock->timer != NULL. This means that the timer callback is still
- * trying to acquire the inode mutex that we currently own. We are
- * safe until we release it. In this case we can safely clear
- * lock->timer. This will cause that the timer callback does nothing
- * once it acquires the mutex.
- */
- lock->timer = NULL;
+ /* If we don't have any timer, there's nothing to cancel. */
+ if (lock->timer == NULL) {
+ return NULL;
+ }
+
+ /* We are trying to access a lock that has an unlock timer active.
+ * This means that the lock must be idle, i.e. no fop can be in the
+ * owner, waiting or frozen lists. It also means that the lock cannot
+ * have been marked as being released (this is done without timers).
+ * There should only be one owner reference, but it's possible that
+ * some fops are being prepared to use this lock. */
+ GF_ASSERT((lock->refs_owners == 1) && list_empty(&lock->owners) &&
+ list_empty(&lock->waiting));
+
+ /* We take the timer_link before cancelling the timer, since a
+ * successful cancellation will destroy it. It must not be NULL
+ * because it references the fop responsible for the delayed unlock
+ * that we are currently trying to cancel. */
+ timer_link = lock->timer->data;
+ GF_ASSERT(timer_link != NULL);
+
+ if (gf_timer_call_cancel(xl->ctx, lock->timer) < 0) {
+ /* It's too late to avoid the execution of the timer callback.
+ * Since we need to be sure that the callback has access to all
+ * needed resources, we cannot resume the execution of the
+ * timer fop now. This will be done in the callback. */
+ timer_link = NULL;
+ } else {
+ /* The timer has been cancelled. The fop referenced by
+ * timer_link holds the last reference. The caller is
+ * responsible to release it when not needed anymore. */
+ ec_trace("UNLOCK_CANCELLED", timer_link->fop, "lock=%p", lock);
+ }
+
+ /* We have two options here:
+ *
+ * 1. The timer has been successfully cancelled.
+ *
+ * This is the easiest case and we can continue with the currently
+ * acquired lock.
+ *
+ * 2. The timer callback has already been fired.
+ *
+ * In this case we have not been able to cancel the timer before
+ * the timer callback has been fired, but we also know that
+ * lock->timer != NULL. This means that the timer callback is still
+ * trying to acquire the inode mutex that we currently own. We are
+ * safe until we release it. In this case we can safely clear
+ * lock->timer. This will cause that the timer callback does nothing
+ * once it acquires the mutex.
+ */
+ lock->timer = NULL;
- return timer_link;
+ return timer_link;
}
static gf_boolean_t
@@ -1947,7 +2044,7 @@ ec_lock_assign_owner(ec_lock_link_t *link)
/* Since the link has just been prepared but it's not active yet, the
* refs_pending must be one at least (the ref owned by this link). */
- GF_ASSERT (lock->refs_pending > 0);
+ GF_ASSERT(lock->refs_pending > 0);
/* The link is not pending any more. It will be assigned to the owner,
* waiting or frozen list. */
lock->refs_pending--;
@@ -1980,7 +2077,7 @@ ec_lock_assign_owner(ec_lock_link_t *link)
* owners, or waiters(to prevent starvation).
* Otherwise we need to wait.
*/
- if (!lock->acquired || ec_link_has_lock_conflict (link, _gf_true)) {
+ if (!lock->acquired || ec_link_has_lock_conflict(link, _gf_true)) {
ec_trace("LOCK_QUEUE_WAIT", fop, "lock=%p", lock);
list_add_tail(&link->wait_list, &lock->waiting);
@@ -1995,7 +2092,7 @@ ec_lock_assign_owner(ec_lock_link_t *link)
* reference assigned to the timer fop. In this case we simply reuse it.
* Otherwise we need to increase the number of owners. */
if (timer_link == NULL) {
- lock->refs_owners++;
+ lock->refs_owners++;
}
assigned = _gf_true;
@@ -2053,14 +2150,14 @@ ec_lock_next_owner(ec_lock_link_t *link, ec_cbk_data_t *cbk,
/* If the fop fails on any of the good bricks, it is important to mark
* it dirty and update versions right away. */
if (link->update[0] || link->update[1]) {
- if (lock->good_mask & ~(fop->good | fop->remaining)) {
- lock->release = _gf_true;
- }
+ if (lock->good_mask & ~(fop->good | fop->remaining)) {
+ lock->release = _gf_true;
+ }
}
}
if (fop->healing) {
- lock->healing = fop->healing & (fop->good | fop->remaining);
+ lock->healing = fop->healing & (fop->good | fop->remaining);
}
ec_lock_update_good(lock, fop);
@@ -2071,7 +2168,8 @@ ec_lock_next_owner(ec_lock_link_t *link, ec_cbk_data_t *cbk,
ec_lock_resume_shared(&list);
}
-void ec_lock(ec_fop_data_t *fop)
+void
+ec_lock(ec_fop_data_t *fop)
{
ec_lock_link_t *link;
@@ -2079,7 +2177,7 @@ void ec_lock(ec_fop_data_t *fop)
* Which can result in refs == 0 for fop leading to use after free in this
* function when it calls ec_sleep so do ec_sleep at start and ec_resume at
* the end of this function.*/
- ec_sleep (fop);
+ ec_sleep(fop);
while (fop->locked < fop->lock_count) {
/* Since there are only up to 2 locks per fop, this xor will change
@@ -2149,17 +2247,16 @@ ec_lock_unfreeze(ec_lock_link_t *link)
}
}
-int32_t ec_unlocked(call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+int32_t
+ec_unlocked(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
ec_fop_data_t *fop = cookie;
ec_lock_link_t *link = fop->data;
if (op_ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- EC_MSG_UNLOCK_FAILED,
- "entry/inode unlocking failed (%s)",
- ec_fop_name(link->fop->id));
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, EC_MSG_UNLOCK_FAILED,
+ "entry/inode unlocking failed :(%s)", ec_msg_str(link->fop));
} else {
ec_trace("UNLOCKED", link->fop, "lock=%p", link->lock);
}
@@ -2169,7 +2266,8 @@ int32_t ec_unlocked(call_frame_t *frame, void *cookie, xlator_t *this,
return 0;
}
-void ec_unlock_lock(ec_lock_link_t *link)
+void
+ec_unlock_lock(ec_lock_link_t *link)
{
ec_lock_t *lock;
ec_fop_data_t *fop;
@@ -2195,10 +2293,27 @@ void ec_unlock_lock(ec_lock_link_t *link)
}
}
-int32_t ec_update_size_version_done(call_frame_t * frame, void * cookie,
- xlator_t * this, int32_t op_ret,
- int32_t op_errno, dict_t * xattr,
- dict_t * xdata)
+void
+ec_inode_bad_inc(inode_t *inode, xlator_t *xl)
+{
+ ec_inode_t *ctx = NULL;
+
+ LOCK(&inode->lock);
+ {
+ ctx = __ec_inode_get(inode, xl);
+ if (ctx == NULL) {
+ goto unlock;
+ }
+ ctx->bad_version++;
+ }
+unlock:
+ UNLOCK(&inode->lock);
+}
+
+int32_t
+ec_update_size_version_done(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
{
ec_fop_data_t *fop = cookie;
ec_lock_link_t *link;
@@ -2210,9 +2325,15 @@ int32_t ec_update_size_version_done(call_frame_t * frame, void * cookie,
ctx = lock->ctx;
if (op_ret < 0) {
- gf_msg(fop->xl->name, fop_log_level (fop->id, op_errno), op_errno,
+ if (link->lock->fd == NULL) {
+ ec_inode_bad_inc(link->lock->loc.inode, this);
+ } else {
+ ec_inode_bad_inc(link->lock->fd->inode, this);
+ }
+
+ gf_msg(fop->xl->name, fop_log_level(fop->id, op_errno), op_errno,
EC_MSG_SIZE_VERS_UPDATE_FAIL,
- "Failed to update version and size");
+ "Failed to update version and size. %s", ec_msg_str(fop));
} else {
fop->parent->good &= fop->good;
@@ -2249,15 +2370,14 @@ int32_t ec_update_size_version_done(call_frame_t * frame, void * cookie,
}
void
-ec_update_size_version(ec_lock_link_t *link, uint64_t *version,
- uint64_t size, uint64_t *dirty)
+ec_update_size_version(ec_lock_link_t *link, uint64_t *version, uint64_t size,
+ uint64_t *dirty)
{
ec_fop_data_t *fop;
ec_lock_t *lock;
ec_inode_t *ctx;
dict_t *dict = NULL;
- uintptr_t update_on = 0;
-
+ uintptr_t update_on = 0;
int32_t err = -ENOMEM;
fop = link->fop;
@@ -2294,11 +2414,10 @@ ec_update_size_version(ec_lock_link_t *link, uint64_t *version,
}
if (dirty[0] || dirty[1]) {
- err = ec_dict_set_array(dict, EC_XATTR_DIRTY,
- dirty, EC_VERSION_SIZE);
- if (err != 0) {
- goto out;
- }
+ err = ec_dict_set_array(dict, EC_XATTR_DIRTY, dirty, EC_VERSION_SIZE);
+ if (err != 0) {
+ goto out;
+ }
}
/* If config information is not known, we request it now. */
@@ -2314,13 +2433,13 @@ ec_update_size_version(ec_lock_link_t *link, uint64_t *version,
update_on = lock->good_mask | lock->healing;
if (link->lock->fd == NULL) {
- ec_xattrop(fop->frame, fop->xl, update_on, EC_MINIMUM_MIN,
- ec_update_size_version_done, link, &link->lock->loc,
- GF_XATTROP_ADD_ARRAY64, dict, NULL);
+ ec_xattrop(fop->frame, fop->xl, update_on, EC_MINIMUM_MIN,
+ ec_update_size_version_done, link, &link->lock->loc,
+ GF_XATTROP_ADD_ARRAY64, dict, NULL);
} else {
- ec_fxattrop(fop->frame, fop->xl, update_on, EC_MINIMUM_MIN,
- ec_update_size_version_done, link, link->lock->fd,
- GF_XATTROP_ADD_ARRAY64, dict, NULL);
+ ec_fxattrop(fop->frame, fop->xl, update_on, EC_MINIMUM_MIN,
+ ec_update_size_version_done, link, link->lock->fd,
+ GF_XATTROP_ADD_ARRAY64, dict, NULL);
}
fop->frame->root->uid = fop->uid;
@@ -2337,13 +2456,12 @@ out:
ec_fop_set_error(fop, -err);
- gf_msg (fop->xl->name, GF_LOG_ERROR, -err, EC_MSG_SIZE_VERS_UPDATE_FAIL,
- "Unable to update version and size");
+ gf_msg(fop->xl->name, GF_LOG_ERROR, -err, EC_MSG_SIZE_VERS_UPDATE_FAIL,
+ "Unable to update version and size. %s", ec_msg_str(fop));
if (lock->unlock_now) {
ec_unlock_lock(fop->data);
}
-
}
gf_boolean_t
@@ -2355,38 +2473,57 @@ ec_update_info(ec_lock_link_t *link)
uint64_t dirty[2] = {0, 0};
uint64_t size;
ec_t *ec = NULL;
+ uintptr_t mask;
lock = link->lock;
ctx = lock->ctx;
ec = link->fop->xl->private;
/* pre_version[*] will be 0 if have_version is false */
- version[0] = ctx->post_version[0] - ctx->pre_version[0];
- version[1] = ctx->post_version[1] - ctx->pre_version[1];
+ version[EC_DATA_TXN] = ctx->post_version[EC_DATA_TXN] -
+ ctx->pre_version[EC_DATA_TXN];
+ version[EC_METADATA_TXN] = ctx->post_version[EC_METADATA_TXN] -
+ ctx->pre_version[EC_METADATA_TXN];
size = ctx->post_size - ctx->pre_size;
/* If we set the dirty flag for update fop, we have to unset it.
* If fop has failed on some bricks, leave the dirty as marked. */
+
if (lock->unlock_now) {
- /* Ensure that nodes are up while doing final
- * metadata update.*/
- if (!(ec->node_mask & ~lock->good_mask) &&
- !(ec->node_mask & ~ec->xl_up)) {
- if (ctx->dirty[0] != 0) {
- dirty[0] = -1;
- }
- if (ctx->dirty[1] != 0) {
- dirty[1] = -1;
- }
- } else {
- link->optimistic_changelog = _gf_false;
- ec_set_dirty_flag (link, ctx, dirty);
+ if (version[EC_DATA_TXN]) {
+ /*A data fop will have difference in post and pre version
+ *and for data fop we send writes on healing bricks also */
+ mask = lock->good_mask | lock->healing;
+ } else {
+ mask = lock->good_mask;
+ }
+ /* Ensure that nodes are up while doing final
+ * metadata update.*/
+ if (!(ec->node_mask & ~(mask)) && !(ec->node_mask & ~ec->xl_up)) {
+ if (ctx->dirty[EC_DATA_TXN] != 0) {
+ dirty[EC_DATA_TXN] = -1;
+ }
+ if (ctx->dirty[EC_METADATA_TXN] != 0) {
+ dirty[EC_METADATA_TXN] = -1;
+ }
+ /*If everything is fine and we already
+ *have version xattr set on entry, there
+ *is no need to update version again*/
+ if (ctx->pre_version[EC_DATA_TXN]) {
+ version[EC_DATA_TXN] = 0;
}
- memset(ctx->dirty, 0, sizeof(ctx->dirty));
+ if (ctx->pre_version[EC_METADATA_TXN]) {
+ version[EC_METADATA_TXN] = 0;
+ }
+ } else {
+ link->optimistic_changelog = _gf_false;
+ ec_set_dirty_flag(link, ctx, dirty);
+ }
+ memset(ctx->dirty, 0, sizeof(ctx->dirty));
}
- if ((version[0] != 0) || (version[1] != 0) ||
- (dirty[0] != 0) || (dirty[1] != 0)) {
+ if ((version[EC_DATA_TXN] != 0) || (version[EC_METADATA_TXN] != 0) ||
+ (dirty[EC_DATA_TXN] != 0) || (dirty[EC_METADATA_TXN] != 0)) {
ec_update_size_version(link, version, size, dirty);
return _gf_true;
}
@@ -2417,185 +2554,192 @@ ec_unlock_now(ec_lock_link_t *link)
void
ec_lock_release(ec_t *ec, inode_t *inode)
{
- ec_lock_t *lock;
- ec_inode_t *ctx;
- ec_lock_link_t *timer_link = NULL;
+ ec_lock_t *lock;
+ ec_inode_t *ctx;
+ ec_lock_link_t *timer_link = NULL;
- LOCK(&inode->lock);
+ LOCK(&inode->lock);
- ctx = __ec_inode_get(inode, ec->xl);
- if (ctx == NULL) {
- goto done;
- }
- lock = ctx->inode_lock;
- if ((lock == NULL) || !lock->acquired || lock->release) {
- goto done;
- }
+ ctx = __ec_inode_get(inode, ec->xl);
+ if (ctx == NULL) {
+ goto done;
+ }
+ lock = ctx->inode_lock;
+ if ((lock == NULL) || lock->release) {
+ goto done;
+ }
- gf_msg_debug(ec->xl->name, 0,
- "Releasing inode %p due to lock contention", inode);
+ gf_msg_debug(ec->xl->name, 0, "Releasing inode %p due to lock contention",
+ inode);
- /* The lock is not marked to be released, so the frozen list should be
- * empty. */
- GF_ASSERT(list_empty(&lock->frozen));
+ if (!lock->acquired) {
+ /* This happens if some bricks already got the lock while inodelk is in
+ * progress. Set release to true after lock is acquired*/
+ lock->contention = _gf_true;
+ goto done;
+ }
- timer_link = ec_lock_timer_cancel(ec->xl, lock);
+ /* The lock is not marked to be released, so the frozen list should be
+ * empty. */
+ GF_ASSERT(list_empty(&lock->frozen));
- /* We mark the lock to be released as soon as possible. */
- lock->release = _gf_true;
+ timer_link = ec_lock_timer_cancel(ec->xl, lock);
+
+ /* We mark the lock to be released as soon as possible. */
+ lock->release = _gf_true;
done:
- UNLOCK(&inode->lock);
+ UNLOCK(&inode->lock);
- /* If we have cancelled the timer, we need to start the unlock of the
- * inode. If there was a timer but we have been unable to cancel it
- * because it was just triggered, the timer callback will take care
- * of releasing the inode. */
- if (timer_link != NULL) {
- ec_unlock_now(timer_link);
- }
+ /* If we have cancelled the timer, we need to start the unlock of the
+ * inode. If there was a timer but we have been unable to cancel it
+ * because it was just triggered, the timer callback will take care
+ * of releasing the inode. */
+ if (timer_link != NULL) {
+ ec_unlock_now(timer_link);
+ }
}
-void ec_unlock_timer_add(ec_lock_link_t *link);
+void
+ec_unlock_timer_add(ec_lock_link_t *link);
void
ec_unlock_timer_del(ec_lock_link_t *link)
{
- ec_lock_t *lock;
- inode_t *inode;
- gf_boolean_t now = _gf_false;
+ ec_lock_t *lock;
+ inode_t *inode;
+ gf_boolean_t now = _gf_false;
+
+ /* If we are here, it means that the timer has expired before having
+ * been cancelled. This guarantees that 'link' is still valid because
+ * the fop that contains it must be pending (if timer cancellation in
+ * ec_lock_assign_owner() fails, the fop is left sleeping).
+ *
+ * At the same time, the fop still has a reference to the lock, so
+ * it must also be valid.
+ */
+ lock = link->lock;
+
+ /* 'lock' must have a valid inode since it can only be destroyed
+ * when the lock itself is destroyed, but we have a reference to the
+ * lock to avoid this.
+ */
+ inode = lock->loc.inode;
+
+ LOCK(&inode->lock);
+
+ if (lock->timer != NULL) {
+ ec_trace("UNLOCK_DELAYED", link->fop, "lock=%p", lock);
+
+ /* The unlock timer has expired without anyone cancelling it.
+ * This means that it shouldn't have any owner, and the waiting
+ * and frozen lists should be empty. It must have only one
+ * owner reference, but there can be fops being prepared
+ * though.
+ * */
+ GF_ASSERT(!lock->release && (lock->refs_owners == 1) &&
+ list_empty(&lock->owners) && list_empty(&lock->waiting) &&
+ list_empty(&lock->frozen));
+
+ gf_timer_call_cancel(link->fop->xl->ctx, lock->timer);
+ lock->timer = NULL;
+
+ /* Any fop being processed from now on, will need to wait
+ * until the next unlock/lock cycle. */
+ lock->release = now = _gf_true;
+ }
+
+ UNLOCK(&inode->lock);
- /* If we are here, it means that the timer has expired before having
- * been cancelled. This guarantees that 'link' is still valid because
- * the fop that contains it must be pending (if timer cancellation in
- * ec_lock_assign_owner() fails, the fop is left sleeping).
+ if (now) {
+ ec_unlock_now(link);
+ } else {
+ /* The timer has been cancelled just after firing it but before
+ * getting here. This means that another fop has used the lock
+ * and everything should be handled as if this callback were
+ * have not been executed. However we still have an owner
+ * reference.
+ *
+ * We need to release our reference. If this is not the last
+ * reference (the most common case because another fop has
+ * taken another ref) we only need to decrement the counter.
+ * Otherwise we have been delayed enough so that the other fop
+ * has had time to acquire the reference, do its operation and
+ * release it. At the time of releasing it, the fop did found
+ * that the ref counter was > 1 (our reference), so the delayed
+ * unlock timer wasn't started. We need to start it again if we
+ * are the last reference.
*
- * At the same time, the fop still has a reference to the lock, so
- * it must also be valid.
+ * ec_unlock_timer_add() handles both cases.
*/
- lock = link->lock;
+ ec_unlock_timer_add(link);
- /* 'lock' must have a valid inode since it can only be destroyed
- * when the lock itself is destroyed, but we have a reference to the
- * lock to avoid this.
+ /* We need to resume the fop that was waiting for the delayed
+ * unlock.
*/
- inode = lock->loc.inode;
-
- LOCK(&inode->lock);
-
- if (lock->timer != NULL) {
- ec_trace("UNLOCK_DELAYED", link->fop, "lock=%p", lock);
-
- /* The unlock timer has expired without anyone cancelling it.
- * This means that it shouldn't have any owner, and the waiting
- * and frozen lists should be empty. It must have only one
- * owner reference, but there can be fops being prepared
- * though.
- * */
- GF_ASSERT(!lock->release &&
- (lock->refs_owners == 1) &&
- list_empty(&lock->owners) &&
- list_empty(&lock->waiting) &&
- list_empty(&lock->frozen));
-
- gf_timer_call_cancel(link->fop->xl->ctx, lock->timer);
- lock->timer = NULL;
-
- /* Any fop being processed from now on, will need to wait
- * until the next unlock/lock cycle. */
- lock->release = now = _gf_true;
- }
-
- UNLOCK(&inode->lock);
+ ec_resume(link->fop, 0);
+ }
+}
- if (now) {
- ec_unlock_now(link);
- } else {
- /* The timer has been cancelled just after firing it but before
- * getting here. This means that another fop has used the lock
- * and everything should be handled as if this callback were
- * have not been executed. However we still have an owner
- * reference.
- *
- * We need to release our reference. If this is not the last
- * reference (the most common case because another fop has
- * taken another ref) we only need to decrement the counter.
- * Otherwise we have been delayed enough so that the other fop
- * has had time to acquire the reference, do its operation and
- * release it. At the time of releasing it, the fop did found
- * that the ref counter was > 1 (our reference), so the delayed
- * unlock timer wasn't started. We need to start it again if we
- * are the last reference.
- *
- * ec_unlock_timer_add() handles both cases.
- */
- ec_unlock_timer_add(link);
-
- /* We need to resume the fop that was waiting for the delayed
- * unlock.
- */
- ec_resume(link->fop, 0);
- }
-}
-
-void ec_unlock_timer_cbk(void *data)
-{
- ec_unlock_timer_del(data);
+void
+ec_unlock_timer_cbk(void *data)
+{
+ ec_unlock_timer_del(data);
}
static gf_boolean_t
ec_eager_lock_used(ec_t *ec, ec_fop_data_t *fop)
{
- /* Fops with no locks at this point mean that they are sent as sub-fops
- * of other higher level fops. In this case we simply assume that the
- * parent fop will take correct care of the eager lock. */
- if (fop->lock_count == 0) {
- return _gf_true;
- }
+ /* Fops with no locks at this point mean that they are sent as sub-fops
+ * of other higher level fops. In this case we simply assume that the
+ * parent fop will take correct care of the eager lock. */
+ if (fop->lock_count == 0) {
+ return _gf_true;
+ }
- /* We may have more than one lock, but this only happens in the rename
- * fop, and both locks will reference an inode of the same type (a
- * directory in this case), so we only need to check the first lock. */
- if (fop->locks[0].lock->loc.inode->ia_type == IA_IFREG) {
- return ec->eager_lock;
- }
+ /* We may have more than one lock, but this only happens in the rename
+ * fop, and both locks will reference an inode of the same type (a
+ * directory in this case), so we only need to check the first lock. */
+ if (fop->locks[0].lock->loc.inode->ia_type == IA_IFREG) {
+ return ec->eager_lock;
+ }
- return ec->other_eager_lock;
+ return ec->other_eager_lock;
}
static uint32_t
ec_eager_lock_timeout(ec_t *ec, ec_lock_t *lock)
{
- if (lock->loc.inode->ia_type == IA_IFREG) {
- return ec->eager_lock_timeout;
- }
+ if (lock->loc.inode->ia_type == IA_IFREG) {
+ return ec->eager_lock_timeout;
+ }
- return ec->other_eager_lock_timeout;
+ return ec->other_eager_lock_timeout;
}
static gf_boolean_t
ec_lock_delay_create(ec_lock_link_t *link)
{
- struct timespec delay;
- ec_fop_data_t *fop = link->fop;
- ec_lock_t *lock = link->lock;
+ struct timespec delay;
+ ec_fop_data_t *fop = link->fop;
+ ec_lock_t *lock = link->lock;
- delay.tv_sec = ec_eager_lock_timeout(fop->xl->private, lock);
- delay.tv_nsec = 0;
- lock->timer = gf_timer_call_after(fop->xl->ctx, delay,
- ec_unlock_timer_cbk, link);
- if (lock->timer == NULL) {
- gf_msg(fop->xl->name, GF_LOG_WARNING, ENOMEM,
- EC_MSG_UNLOCK_DELAY_FAILED,
- "Unable to delay an unlock");
+ delay.tv_sec = ec_eager_lock_timeout(fop->xl->private, lock);
+ delay.tv_nsec = 0;
+ lock->timer = gf_timer_call_after(fop->xl->ctx, delay, ec_unlock_timer_cbk,
+ link);
+ if (lock->timer == NULL) {
+ gf_msg(fop->xl->name, GF_LOG_WARNING, ENOMEM,
+ EC_MSG_UNLOCK_DELAY_FAILED, "Unable to delay an unlock");
- return _gf_false;
- }
+ return _gf_false;
+ }
- return _gf_true;
+ return _gf_true;
}
-void ec_unlock_timer_add(ec_lock_link_t *link)
+void
+ec_unlock_timer_add(ec_lock_link_t *link)
{
ec_fop_data_t *fop = link->fop;
ec_lock_t *lock = link->lock;
@@ -2685,7 +2829,8 @@ void ec_unlock_timer_add(ec_lock_link_t *link)
}
}
-void ec_unlock(ec_fop_data_t *fop)
+void
+ec_unlock(ec_fop_data_t *fop)
{
int32_t i;
@@ -2694,7 +2839,8 @@ void ec_unlock(ec_fop_data_t *fop)
}
}
-void ec_flush_size_version(ec_fop_data_t * fop)
+void
+ec_flush_size_version(ec_fop_data_t *fop)
{
GF_ASSERT(fop->lock_count == 1);
ec_update_info(&fop->locks[0]);
@@ -2704,99 +2850,98 @@ static void
ec_update_stripe(ec_t *ec, ec_stripe_list_t *stripe_cache, ec_stripe_t *stripe,
ec_fop_data_t *fop)
{
- off_t base;
-
- /* On write fops, we only update existing fragments if the write has
- * succeeded. Otherwise, we remove them from the cache. */
- if ((fop->id == GF_FOP_WRITE) && (fop->answer != NULL) &&
- (fop->answer->op_ret >= 0)) {
- base = stripe->frag_offset - fop->frag_range.first;
- base *= ec->fragments;
-
- /* We check if the stripe offset falls inside the real region
- * modified by the write fop (a write request is allowed,
- * though uncommon, to write less bytes than requested). The
- * current write fop implementation doesn't allow partial
- * writes of fragments, so if there's no error, we are sure
- * that a full stripe has been completely modified or not
- * touched at all. The value of op_ret may not be a multiple
- * of the stripe size because it depends on the requested
- * size by the user, so we update the stripe if the write has
- * modified at least one byte (meaning ec has written the full
- * stripe). */
- if (base < fop->answer->op_ret + fop->head) {
- memcpy(stripe->data, fop->vector[0].iov_base + base,
- ec->stripe_size);
- list_move_tail(&stripe->lru, &stripe_cache->lru);
-
- GF_ATOMIC_INC(ec->stats.stripe_cache.updates);
- }
- } else {
- stripe->frag_offset = -1;
- list_move (&stripe->lru, &stripe_cache->lru);
-
- GF_ATOMIC_INC(ec->stats.stripe_cache.invals);
+ off_t base;
+
+ /* On write fops, we only update existing fragments if the write has
+ * succeeded. Otherwise, we remove them from the cache. */
+ if ((fop->id == GF_FOP_WRITE) && (fop->answer != NULL) &&
+ (fop->answer->op_ret >= 0)) {
+ base = stripe->frag_offset - fop->frag_range.first;
+ base *= ec->fragments;
+
+ /* We check if the stripe offset falls inside the real region
+ * modified by the write fop (a write request is allowed,
+ * though uncommon, to write less bytes than requested). The
+ * current write fop implementation doesn't allow partial
+ * writes of fragments, so if there's no error, we are sure
+ * that a full stripe has been completely modified or not
+ * touched at all. The value of op_ret may not be a multiple
+ * of the stripe size because it depends on the requested
+ * size by the user, so we update the stripe if the write has
+ * modified at least one byte (meaning ec has written the full
+ * stripe). */
+ if (base < fop->answer->op_ret + fop->head) {
+ memcpy(stripe->data, fop->vector[0].iov_base + base,
+ ec->stripe_size);
+ list_move_tail(&stripe->lru, &stripe_cache->lru);
+
+ GF_ATOMIC_INC(ec->stats.stripe_cache.updates);
}
+ } else {
+ stripe->frag_offset = -1;
+ list_move(&stripe->lru, &stripe_cache->lru);
+
+ GF_ATOMIC_INC(ec->stats.stripe_cache.invals);
+ }
}
static void
-ec_update_cached_stripes (ec_fop_data_t *fop)
-{
- uint64_t first;
- uint64_t last;
- ec_stripe_t *stripe = NULL;
- ec_inode_t *ctx = NULL;
- ec_stripe_list_t *stripe_cache = NULL;
- inode_t *inode = NULL;
- struct list_head *temp;
- struct list_head sentinel;
-
- first = fop->frag_range.first;
- /* 'last' represents the first stripe not touched by the operation */
- last = fop->frag_range.last;
-
- /* If there are no modified stripes, we don't need to do anything
- * else. */
- if (last <= first) {
- return;
- }
+ec_update_cached_stripes(ec_fop_data_t *fop)
+{
+ uint64_t first;
+ uint64_t last;
+ ec_stripe_t *stripe = NULL;
+ ec_inode_t *ctx = NULL;
+ ec_stripe_list_t *stripe_cache = NULL;
+ inode_t *inode = NULL;
+ struct list_head *temp;
+ struct list_head sentinel;
+
+ first = fop->frag_range.first;
+ /* 'last' represents the first stripe not touched by the operation */
+ last = fop->frag_range.last;
+
+ /* If there are no modified stripes, we don't need to do anything
+ * else. */
+ if (last <= first) {
+ return;
+ }
- if (!fop->use_fd) {
- inode = fop->loc[0].inode;
- } else {
- inode = fop->fd->inode;
- }
+ if (!fop->use_fd) {
+ inode = fop->loc[0].inode;
+ } else {
+ inode = fop->fd->inode;
+ }
- LOCK(&inode->lock);
+ LOCK(&inode->lock);
- ctx = __ec_inode_get (inode, fop->xl);
- if (ctx == NULL) {
- goto out;
- }
- stripe_cache = &ctx->stripe_cache;
-
- /* Since we'll be moving elements of the list to the tail, we might
- * end in an infinite loop. To avoid it, we insert a sentinel element
- * into the list, so that it will be used to detect when we have
- * traversed all existing elements once. */
- list_add_tail(&sentinel, &stripe_cache->lru);
- temp = stripe_cache->lru.next;
- while (temp != &sentinel) {
- stripe = list_entry(temp, ec_stripe_t, lru);
- temp = temp->next;
- if ((first <= stripe->frag_offset) &&
- (stripe->frag_offset < last)) {
- ec_update_stripe (fop->xl->private, stripe_cache,
- stripe, fop);
- }
+ ctx = __ec_inode_get(inode, fop->xl);
+ if (ctx == NULL) {
+ goto out;
+ }
+ stripe_cache = &ctx->stripe_cache;
+
+ /* Since we'll be moving elements of the list to the tail, we might
+ * end in an infinite loop. To avoid it, we insert a sentinel element
+ * into the list, so that it will be used to detect when we have
+ * traversed all existing elements once. */
+ list_add_tail(&sentinel, &stripe_cache->lru);
+ temp = stripe_cache->lru.next;
+ while (temp != &sentinel) {
+ stripe = list_entry(temp, ec_stripe_t, lru);
+ temp = temp->next;
+ if ((first <= stripe->frag_offset) && (stripe->frag_offset < last)) {
+ ec_update_stripe(fop->xl->private, stripe_cache, stripe, fop);
}
- list_del(&sentinel);
+ }
+ list_del(&sentinel);
out:
- UNLOCK(&inode->lock);
+ UNLOCK(&inode->lock);
}
-void ec_lock_reuse(ec_fop_data_t *fop)
+void
+ec_lock_reuse(ec_fop_data_t *fop)
{
ec_cbk_data_t *cbk;
ec_t *ec = NULL;
@@ -2807,13 +2952,13 @@ void ec_lock_reuse(ec_fop_data_t *fop)
if (ec_eager_lock_used(ec, fop) && cbk != NULL) {
if (cbk->xdata != NULL) {
- if ((dict_get_int32(cbk->xdata, GLUSTERFS_INODELK_COUNT,
- &count) == 0) && (count > 1)) {
+ if ((dict_get_int32(cbk->xdata, GLUSTERFS_INODELK_COUNT, &count) ==
+ 0) &&
+ (count > 1)) {
release = _gf_true;
}
if (release) {
- gf_msg_debug (fop->xl->name, 0,
- "Lock contention detected");
+ gf_msg_debug(fop->xl->name, 0, "Lock contention detected");
}
}
} else {
@@ -2822,24 +2967,25 @@ void ec_lock_reuse(ec_fop_data_t *fop)
* the lock. */
release = _gf_true;
}
- ec_update_cached_stripes (fop);
+ ec_update_cached_stripes(fop);
for (i = 0; i < fop->lock_count; i++) {
ec_lock_next_owner(&fop->locks[i], cbk, release);
}
}
-void __ec_manager(ec_fop_data_t * fop, int32_t error)
+void
+__ec_manager(ec_fop_data_t *fop, int32_t error)
{
ec_t *ec = fop->xl->private;
do {
ec_trace("MANAGER", fop, "error=%d", error);
- if (!ec_must_wind (fop)) {
- if (ec->xl_up_count < ec->fragments) {
- error = ENOTCONN;
- }
+ if (!ec_must_wind(fop)) {
+ if (ec->xl_up_count < ec->fragments) {
+ error = ENOTCONN;
+ }
}
if (error != 0) {
@@ -2865,22 +3011,32 @@ void __ec_manager(ec_fop_data_t * fop, int32_t error)
fop->jobs = 1;
fop->state = fop->handler(fop, fop->state);
- GF_ASSERT (fop->state >= 0);
+ GF_ASSERT(fop->state >= 0);
error = ec_check_complete(fop, __ec_manager);
} while (error >= 0);
}
-void ec_manager(ec_fop_data_t * fop, int32_t error)
+void
+ec_manager(ec_fop_data_t *fop, int32_t error)
{
GF_ASSERT(fop->jobs == 0);
GF_ASSERT(fop->winds == 0);
GF_ASSERT(fop->error == 0);
- if (fop->state == EC_STATE_START)
- {
+ if (fop->state == EC_STATE_START) {
fop->state = EC_STATE_INIT;
}
__ec_manager(fop, error);
}
+
+gf_boolean_t
+__ec_is_last_fop(ec_t *ec)
+{
+ if ((list_empty(&ec->pending_fops)) &&
+ (GF_ATOMIC_GET(ec->async_fop_count) == 0)) {
+ return _gf_true;
+ }
+ return _gf_false;
+}
diff --git a/xlators/cluster/ec/src/ec-common.h b/xlators/cluster/ec/src/ec-common.h
index 372be52470c..51493612ac6 100644
--- a/xlators/cluster/ec/src/ec-common.h
+++ b/xlators/cluster/ec/src/ec-common.h
@@ -11,37 +11,57 @@
#ifndef __EC_COMMON_H__
#define __EC_COMMON_H__
-#include "xlator.h"
-
+#include "glusterfs/compat-errno.h" // for ENODATA on BSD
#include "ec-data.h"
-typedef enum {
- EC_DATA_TXN,
- EC_METADATA_TXN
-} ec_txn_t;
+typedef enum { EC_DATA_TXN, EC_METADATA_TXN } ec_txn_t;
-#define EC_FOP_HEAL -1
-#define EC_FOP_FHEAL -2
+#define EC_FOP_HEAL -1
+#define EC_FOP_FHEAL -2
#define EC_CONFIG_VERSION 0
#define EC_CONFIG_ALGORITHM 0
-#define EC_FLAG_LOCK_SHARED 0x0001
+#define EC_FLAG_LOCK_SHARED 0x0001
+
+#define QUORUM_CBK(fn, fop, frame, cookie, this, op_ret, op_errno, params...) \
+ do { \
+ ec_t *__ec = fop->xl->private; \
+ int32_t __op_ret = 0; \
+ int32_t __op_errno = 0; \
+ int32_t __success_count = gf_bits_count(fop->good); \
+ \
+ __op_ret = op_ret; \
+ __op_errno = op_errno; \
+ if (!fop->parent && frame && \
+ (GF_CLIENT_PID_SELF_HEALD != frame->root->pid) && \
+ __ec->quorum_count && (__success_count < __ec->quorum_count) && \
+ op_ret >= 0) { \
+ __op_ret = -1; \
+ __op_errno = EIO; \
+ gf_msg(__ec->xl->name, GF_LOG_ERROR, 0, \
+ EC_MSG_CHILDS_INSUFFICIENT, \
+ "Insufficient available children for this request " \
+ "(have %d, need %d). %s", \
+ __success_count, __ec->quorum_count, ec_msg_str(fop)); \
+ } \
+ fn(frame, cookie, this, __op_ret, __op_errno, params); \
+ } while (0)
enum _ec_xattrop_flags {
- EC_FLAG_XATTROP,
- EC_FLAG_DATA_DIRTY,
- EC_FLAG_METADATA_DIRTY,
+ EC_FLAG_XATTROP,
+ EC_FLAG_DATA_DIRTY,
+ EC_FLAG_METADATA_DIRTY,
- /* Add any new flag here, before EC_FLAG_MAX. The maximum number of
- * flags that can be defined is 16. */
+ /* Add any new flag here, before EC_FLAG_MAX. The maximum number of
+ * flags that can be defined is 16. */
- EC_FLAG_MAX
+ EC_FLAG_MAX
};
/* We keep two sets of flags. One to determine what's really providing the
- * currect xattrop and the other to know what the parent fop of the xattrop
+ * current xattrop and the other to know what the parent fop of the xattrop
* needs to proceed. It might happen that a fop needs some information that
* is being already requested by a previous fop. The two sets are stored
* contiguously. */
@@ -57,57 +77,69 @@ enum _ec_xattrop_flags {
#define EC_SELFHEAL_BIT 62
-#define EC_MINIMUM_ONE -1
-#define EC_MINIMUM_MIN -2
-#define EC_MINIMUM_ALL -3
-
-#define EC_UPDATE_DATA 1
-#define EC_UPDATE_META 2
-#define EC_QUERY_INFO 4
-#define EC_INODE_SIZE 8
-
-#define EC_STATE_START 0
-#define EC_STATE_END 0
-#define EC_STATE_INIT 1
-#define EC_STATE_LOCK 2
-#define EC_STATE_DISPATCH 3
-#define EC_STATE_PREPARE_ANSWER 4
-#define EC_STATE_REPORT 5
-#define EC_STATE_LOCK_REUSE 6
-#define EC_STATE_UNLOCK 7
-
-#define EC_STATE_DELAYED_START 100
-
-#define EC_STATE_HEAL_ENTRY_LOOKUP 200
-#define EC_STATE_HEAL_ENTRY_PREPARE 201
-#define EC_STATE_HEAL_PRE_INODELK_LOCK 202
-#define EC_STATE_HEAL_PRE_INODE_LOOKUP 203
-#define EC_STATE_HEAL_XATTRIBUTES_REMOVE 204
-#define EC_STATE_HEAL_XATTRIBUTES_SET 205
-#define EC_STATE_HEAL_ATTRIBUTES 206
-#define EC_STATE_HEAL_OPEN 207
-#define EC_STATE_HEAL_REOPEN_FD 208
-#define EC_STATE_HEAL_UNLOCK 209
-#define EC_STATE_HEAL_UNLOCK_ENTRY 210
-#define EC_STATE_HEAL_DATA_LOCK 211
-#define EC_STATE_HEAL_DATA_COPY 212
-#define EC_STATE_HEAL_DATA_UNLOCK 213
-#define EC_STATE_HEAL_POST_INODELK_LOCK 214
-#define EC_STATE_HEAL_POST_INODE_LOOKUP 215
-#define EC_STATE_HEAL_SETATTR 216
-#define EC_STATE_HEAL_POST_INODELK_UNLOCK 217
-#define EC_STATE_HEAL_DISPATCH 218
-
-gf_boolean_t ec_dispatch_one_retry (ec_fop_data_t *fop, ec_cbk_data_t **cbk);
-void ec_dispatch_next(ec_fop_data_t * fop, uint32_t idx);
-
-void ec_complete(ec_fop_data_t *fop);
-
-void ec_update_good(ec_fop_data_t *fop, uintptr_t good);
-
-void ec_fop_set_error(ec_fop_data_t *fop, int32_t error);
-
-void __ec_fop_set_error(ec_fop_data_t *fop, int32_t error);
+#define EC_MINIMUM_ONE (1 << 6)
+#define EC_MINIMUM_MIN (2 << 6)
+#define EC_MINIMUM_ALL (3 << 6)
+#define EC_FOP_NO_PROPAGATE_ERROR (1 << 8)
+#define EC_FOP_MINIMUM(_flags) ((_flags)&255)
+#define EC_FOP_FLAGS(_flags) ((_flags) & ~255)
+
+#define EC_UPDATE_DATA 1
+#define EC_UPDATE_META 2
+#define EC_QUERY_INFO 4
+#define EC_INODE_SIZE 8
+
+#define EC_STATE_START 0
+#define EC_STATE_END 0
+#define EC_STATE_INIT 1
+#define EC_STATE_LOCK 2
+#define EC_STATE_DISPATCH 3
+#define EC_STATE_PREPARE_ANSWER 4
+#define EC_STATE_REPORT 5
+#define EC_STATE_LOCK_REUSE 6
+#define EC_STATE_UNLOCK 7
+
+#define EC_STATE_DELAYED_START 100
+
+#define EC_STATE_HEAL_ENTRY_LOOKUP 200
+#define EC_STATE_HEAL_ENTRY_PREPARE 201
+#define EC_STATE_HEAL_PRE_INODELK_LOCK 202
+#define EC_STATE_HEAL_PRE_INODE_LOOKUP 203
+#define EC_STATE_HEAL_XATTRIBUTES_REMOVE 204
+#define EC_STATE_HEAL_XATTRIBUTES_SET 205
+#define EC_STATE_HEAL_ATTRIBUTES 206
+#define EC_STATE_HEAL_OPEN 207
+#define EC_STATE_HEAL_REOPEN_FD 208
+#define EC_STATE_HEAL_UNLOCK 209
+#define EC_STATE_HEAL_UNLOCK_ENTRY 210
+#define EC_STATE_HEAL_DATA_LOCK 211
+#define EC_STATE_HEAL_DATA_COPY 212
+#define EC_STATE_HEAL_DATA_UNLOCK 213
+#define EC_STATE_HEAL_POST_INODELK_LOCK 214
+#define EC_STATE_HEAL_POST_INODE_LOOKUP 215
+#define EC_STATE_HEAL_SETATTR 216
+#define EC_STATE_HEAL_POST_INODELK_UNLOCK 217
+#define EC_STATE_HEAL_DISPATCH 218
+
+/* Value to cover the full range of a file */
+#define EC_RANGE_FULL ((uint64_t)LLONG_MAX + 1)
+
+gf_boolean_t
+ec_dispatch_one_retry(ec_fop_data_t *fop, ec_cbk_data_t **cbk);
+void
+ec_dispatch_next(ec_fop_data_t *fop, uint32_t idx);
+
+void
+ec_complete(ec_fop_data_t *fop);
+
+void
+ec_update_good(ec_fop_data_t *fop, uintptr_t good);
+
+void
+ec_fop_set_error(ec_fop_data_t *fop, int32_t error);
+
+void
+__ec_fop_set_error(ec_fop_data_t *fop, int32_t error);
ec_cbk_data_t *
ec_fop_prepare_answer(ec_fop_data_t *fop, gf_boolean_t ro);
@@ -115,55 +147,88 @@ ec_fop_prepare_answer(ec_fop_data_t *fop, gf_boolean_t ro);
gf_boolean_t
ec_cbk_set_error(ec_cbk_data_t *cbk, int32_t error, gf_boolean_t ro);
-void ec_lock_prepare_inode(ec_fop_data_t *fop, loc_t *loc, uint32_t flags,
- off_t fl_start, size_t fl_size);
-void ec_lock_prepare_parent_inode(ec_fop_data_t *fop, loc_t *loc, loc_t *base,
- uint32_t flags);
-void ec_lock_prepare_fd(ec_fop_data_t *fop, fd_t *fd, uint32_t flags,
- off_t fl_start, size_t fl_size);
-void ec_lock(ec_fop_data_t * fop);
-void ec_lock_reuse(ec_fop_data_t *fop);
-void ec_unlock(ec_fop_data_t * fop);
-void ec_lock_release(ec_t *ec, inode_t *inode);
-
-gf_boolean_t ec_get_inode_size(ec_fop_data_t *fop, inode_t *inode,
- uint64_t *size);
-gf_boolean_t __ec_get_inode_size(ec_fop_data_t *fop, inode_t *inode,
- uint64_t *size);
-gf_boolean_t ec_set_inode_size(ec_fop_data_t *fop, inode_t *inode,
- uint64_t size);
-gf_boolean_t __ec_set_inode_size(ec_fop_data_t *fop, inode_t *inode,
- uint64_t size);
-void ec_clear_inode_info(ec_fop_data_t *fop, inode_t *inode);
-
-void ec_flush_size_version(ec_fop_data_t * fop);
-
-void ec_dispatch_all(ec_fop_data_t * fop);
-void ec_dispatch_inc(ec_fop_data_t * fop);
-void ec_dispatch_min(ec_fop_data_t * fop);
-void ec_dispatch_one(ec_fop_data_t * fop);
-
-void ec_sleep(ec_fop_data_t *fop);
-void ec_resume(ec_fop_data_t * fop, int32_t error);
-void ec_resume_parent(ec_fop_data_t * fop, int32_t error);
-
-void ec_manager(ec_fop_data_t * fop, int32_t error);
-gf_boolean_t ec_is_recoverable_error (int32_t op_errno);
-void ec_handle_healers_done (ec_fop_data_t *fop);
+void
+ec_lock_prepare_inode(ec_fop_data_t *fop, loc_t *loc, uint32_t flags,
+ off_t fl_start, uint64_t fl_size);
+void
+ec_lock_prepare_parent_inode(ec_fop_data_t *fop, loc_t *loc, loc_t *base,
+ uint32_t flags);
+void
+ec_lock_prepare_fd(ec_fop_data_t *fop, fd_t *fd, uint32_t flags, off_t fl_start,
+ uint64_t fl_size);
+void
+ec_lock(ec_fop_data_t *fop);
+void
+ec_lock_reuse(ec_fop_data_t *fop);
+void
+ec_unlock(ec_fop_data_t *fop);
+void
+ec_lock_release(ec_t *ec, inode_t *inode);
+
+gf_boolean_t
+ec_get_inode_size(ec_fop_data_t *fop, inode_t *inode, uint64_t *size);
+gf_boolean_t
+__ec_get_inode_size(ec_fop_data_t *fop, inode_t *inode, uint64_t *size);
+gf_boolean_t
+ec_set_inode_size(ec_fop_data_t *fop, inode_t *inode, uint64_t size);
+gf_boolean_t
+__ec_set_inode_size(ec_fop_data_t *fop, inode_t *inode, uint64_t size);
+void
+ec_clear_inode_info(ec_fop_data_t *fop, inode_t *inode);
+
+void
+ec_flush_size_version(ec_fop_data_t *fop);
+
+void
+ec_dispatch_all(ec_fop_data_t *fop);
+void
+ec_dispatch_inc(ec_fop_data_t *fop);
+void
+ec_dispatch_min(ec_fop_data_t *fop);
+void
+ec_dispatch_one(ec_fop_data_t *fop);
+
+void
+ec_succeed_all(ec_fop_data_t *fop);
+
+void
+ec_sleep(ec_fop_data_t *fop);
+void
+ec_resume(ec_fop_data_t *fop, int32_t error);
+void
+ec_resume_parent(ec_fop_data_t *fop);
+
+void
+ec_manager(ec_fop_data_t *fop, int32_t error);
+gf_boolean_t
+ec_is_recoverable_error(int32_t op_errno);
+void
+ec_handle_healers_done(ec_fop_data_t *fop);
int32_t
-ec_heal_inspect (call_frame_t *frame, ec_t *ec,
- inode_t *inode, unsigned char *locked_on,
- gf_boolean_t self_locked, gf_boolean_t thorough,
- ec_heal_need_t *need_heal);
+ec_heal_inspect(call_frame_t *frame, ec_t *ec, inode_t *inode,
+ unsigned char *locked_on, gf_boolean_t self_locked,
+ gf_boolean_t thorough, ec_heal_need_t *need_heal);
int32_t
-ec_get_heal_info (xlator_t *this, loc_t *loc, dict_t **dict);
+ec_get_heal_info(xlator_t *this, loc_t *loc, dict_t **dict);
-int32_t ec_lock_unlocked(call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xdata);
+int32_t
+ec_lock_unlocked(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata);
void
-ec_update_fd_status (fd_t *fd, xlator_t *xl,
- int child_index, int32_t ret_status);
+ec_update_fd_status(fd_t *fd, xlator_t *xl, int child_index,
+ int32_t ret_status);
+gf_boolean_t
+ec_is_entry_healing(ec_fop_data_t *fop);
+void
+ec_set_entry_healing(ec_fop_data_t *fop);
+void
+ec_reset_entry_healing(ec_fop_data_t *fop);
+char *
+ec_msg_str(ec_fop_data_t *fop);
+gf_boolean_t
+__ec_is_last_fop(ec_t *ec);
+void
+ec_lock_update_good(ec_lock_t *lock, ec_fop_data_t *fop);
#endif /* __EC_COMMON_H__ */
diff --git a/xlators/cluster/ec/src/ec-data.c b/xlators/cluster/ec/src/ec-data.c
index a869735794e..06388833546 100644
--- a/xlators/cluster/ec/src/ec-data.c
+++ b/xlators/cluster/ec/src/ec-data.c
@@ -8,53 +8,48 @@
cases as published by the Free Software Foundation.
*/
-#include "ec-mem-types.h"
#include "ec-helpers.h"
#include "ec-common.h"
#include "ec-data.h"
#include "ec-messages.h"
-ec_cbk_data_t * ec_cbk_data_allocate(call_frame_t * frame, xlator_t * this,
- ec_fop_data_t * fop, int32_t id,
- int32_t idx, int32_t op_ret,
- int32_t op_errno)
+ec_cbk_data_t *
+ec_cbk_data_allocate(call_frame_t *frame, xlator_t *this, ec_fop_data_t *fop,
+ int32_t id, int32_t idx, int32_t op_ret, int32_t op_errno)
{
- ec_cbk_data_t * cbk;
- ec_t * ec = this->private;
+ ec_cbk_data_t *cbk;
+ ec_t *ec = this->private;
- if (fop->xl != this)
- {
- gf_msg (this->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_XLATOR_MISMATCH, "Mismatching xlators between request "
- "and answer (req=%s, ans=%s).", fop->xl->name, this->name);
+ if (fop->xl != this) {
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL, EC_MSG_XLATOR_MISMATCH,
+ "Mismatching xlators between request "
+ "and answer (req=%s, ans=%s).",
+ fop->xl->name, this->name);
return NULL;
}
- if (fop->frame != frame)
- {
- gf_msg (this->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_FRAME_MISMATCH, "Mismatching frames between request "
- "and answer (req=%p, ans=%p).",
- fop->frame, frame);
+ if (fop->frame != frame) {
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL, EC_MSG_FRAME_MISMATCH,
+ "Mismatching frames between request "
+ "and answer (req=%p, ans=%p).",
+ fop->frame, frame);
return NULL;
}
- if (fop->id != id)
- {
- gf_msg (this->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_FOP_MISMATCH, "Mismatching fops between request "
- "and answer (req=%d, ans=%d).",
- fop->id, id);
+ if (fop->id != id) {
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL, EC_MSG_FOP_MISMATCH,
+ "Mismatching fops between request "
+ "and answer (req=%d, ans=%d).",
+ fop->id, id);
return NULL;
}
cbk = mem_get0(ec->cbk_pool);
- if (cbk == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY, "Failed to allocate memory for an "
- "answer.");
+ if (cbk == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to allocate memory for an "
+ "answer.");
return NULL;
}
@@ -64,7 +59,7 @@ ec_cbk_data_t * ec_cbk_data_allocate(call_frame_t * frame, xlator_t * this,
cbk->count = 1;
cbk->op_ret = op_ret;
cbk->op_errno = op_errno;
- INIT_LIST_HEAD (&cbk->entries.list);
+ INIT_LIST_HEAD(&cbk->entries.list);
LOCK(&fop->lock);
@@ -75,50 +70,45 @@ ec_cbk_data_t * ec_cbk_data_allocate(call_frame_t * frame, xlator_t * this,
return cbk;
}
-void ec_cbk_data_destroy(ec_cbk_data_t * cbk)
+void
+ec_cbk_data_destroy(ec_cbk_data_t *cbk)
{
- if (cbk->xdata != NULL)
- {
+ if (cbk->xdata != NULL) {
dict_unref(cbk->xdata);
}
- if (cbk->dict != NULL)
- {
+ if (cbk->dict != NULL) {
dict_unref(cbk->dict);
}
- if (cbk->inode != NULL)
- {
+ if (cbk->inode != NULL) {
inode_unref(cbk->inode);
}
- if (cbk->fd != NULL)
- {
+ if (cbk->fd != NULL) {
fd_unref(cbk->fd);
}
- if (cbk->buffers != NULL)
- {
+ if (cbk->buffers != NULL) {
iobref_unref(cbk->buffers);
}
GF_FREE(cbk->vector);
- gf_dirent_free (&cbk->entries);
- GF_FREE (cbk->str);
+ gf_dirent_free(&cbk->entries);
+ GF_FREE(cbk->str);
mem_put(cbk);
}
-ec_fop_data_t * ec_fop_data_allocate(call_frame_t * frame, xlator_t * this,
- int32_t id, uint32_t flags,
- uintptr_t target, int32_t minimum,
- ec_wind_f wind, ec_handler_f handler,
- ec_cbk_t cbks, void * data)
+ec_fop_data_t *
+ec_fop_data_allocate(call_frame_t *frame, xlator_t *this, int32_t id,
+ uint32_t flags, uintptr_t target, uint32_t fop_flags,
+ ec_wind_f wind, ec_handler_f handler, ec_cbk_t cbks,
+ void *data)
{
- ec_fop_data_t * fop, * parent;
- ec_t * ec = this->private;
+ ec_fop_data_t *fop, *parent;
+ ec_t *ec = this->private;
fop = mem_get0(ec->fop_pool);
- if (fop == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY, "Failed to allocate memory for a "
- "request.");
+ if (fop == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to allocate memory for a "
+ "request.");
return NULL;
}
@@ -142,19 +132,15 @@ ec_fop_data_t * ec_fop_data_allocate(call_frame_t * frame, xlator_t * this,
* TODO: minimize usage of private frames. Reuse req_frame as much as
* possible.
*/
- if (frame != NULL)
- {
+ if (frame != NULL) {
fop->frame = copy_frame(frame);
- }
- else
- {
+ } else {
fop->frame = create_frame(this, this->ctx->pool);
}
- if (fop->frame == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY, "Failed to create a private frame "
- "for a request");
+ if (fop->frame == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to create a private frame "
+ "for a request");
mem_put(fop);
@@ -164,7 +150,8 @@ ec_fop_data_t * ec_fop_data_allocate(call_frame_t * frame, xlator_t * this,
fop->refs = 1;
fop->flags = flags;
- fop->minimum = minimum;
+ fop->minimum = EC_FOP_MINIMUM(fop_flags);
+ fop->fop_flags = EC_FOP_FLAGS(fop_flags);
fop->mask = target;
fop->wind = wind;
@@ -179,11 +166,9 @@ ec_fop_data_t * ec_fop_data_allocate(call_frame_t * frame, xlator_t * this,
fop->frame->local = fop;
- if (frame != NULL)
- {
+ if (frame != NULL) {
parent = frame->local;
- if (parent != NULL)
- {
+ if (parent != NULL) {
ec_sleep(parent);
}
@@ -199,7 +184,8 @@ ec_fop_data_t * ec_fop_data_allocate(call_frame_t * frame, xlator_t * this,
return fop;
}
-void ec_fop_data_acquire(ec_fop_data_t * fop)
+void
+ec_fop_data_acquire(ec_fop_data_t *fop)
{
LOCK(&fop->lock);
@@ -211,36 +197,40 @@ void ec_fop_data_acquire(ec_fop_data_t * fop)
}
static void
-ec_handle_last_pending_fop_completion (ec_fop_data_t *fop, gf_boolean_t *notify)
+ec_handle_last_pending_fop_completion(ec_fop_data_t *fop, gf_boolean_t *notify)
{
- ec_t *ec = fop->xl->private;
-
- if (!list_empty (&fop->pending_list)) {
- LOCK(&ec->lock);
- {
- list_del_init (&fop->pending_list);
- *notify = list_empty (&ec->pending_fops);
- }
- UNLOCK(&ec->lock);
+ ec_t *ec = fop->xl->private;
+
+ *notify = _gf_false;
+
+ if (!list_empty(&fop->pending_list)) {
+ LOCK(&ec->lock);
+ {
+ list_del_init(&fop->pending_list);
+ *notify = __ec_is_last_fop(ec);
}
+ UNLOCK(&ec->lock);
+ }
}
void
ec_fop_cleanup(ec_fop_data_t *fop)
{
- ec_cbk_data_t *cbk, *tmp;
+ ec_cbk_data_t *cbk, *tmp;
- list_for_each_entry_safe(cbk, tmp, &fop->answer_list, answer_list) {
- list_del_init(&cbk->answer_list);
+ list_for_each_entry_safe(cbk, tmp, &fop->answer_list, answer_list)
+ {
+ list_del_init(&cbk->answer_list);
- ec_cbk_data_destroy(cbk);
- }
- INIT_LIST_HEAD(&fop->cbk_list);
+ ec_cbk_data_destroy(cbk);
+ }
+ INIT_LIST_HEAD(&fop->cbk_list);
- fop->answer = NULL;
+ fop->answer = NULL;
}
-void ec_fop_data_release(ec_fop_data_t * fop)
+void
+ec_fop_data_release(ec_fop_data_t *fop)
{
ec_t *ec = NULL;
int32_t refs;
@@ -250,36 +240,30 @@ void ec_fop_data_release(ec_fop_data_t * fop)
ec_trace("RELEASE", fop, "");
- GF_ASSERT (fop->refs > 0);
+ GF_ASSERT(fop->refs > 0);
refs = --fop->refs;
UNLOCK(&fop->lock);
- if (refs == 0)
- {
+ if (refs == 0) {
fop->frame->local = NULL;
STACK_DESTROY(fop->frame->root);
LOCK_DESTROY(&fop->lock);
- if (fop->xdata != NULL)
- {
+ if (fop->xdata != NULL) {
dict_unref(fop->xdata);
}
- if (fop->dict != NULL)
- {
+ if (fop->dict != NULL) {
dict_unref(fop->dict);
}
- if (fop->inode != NULL)
- {
+ if (fop->inode != NULL) {
inode_unref(fop->inode);
}
- if (fop->fd != NULL)
- {
+ if (fop->fd != NULL) {
fd_unref(fop->fd);
}
- if (fop->buffers != NULL)
- {
+ if (fop->buffers != NULL) {
iobref_unref(fop->buffers);
}
GF_FREE(fop->vector);
@@ -287,14 +271,15 @@ void ec_fop_data_release(ec_fop_data_t * fop)
GF_FREE(fop->str[1]);
loc_wipe(&fop->loc[0]);
loc_wipe(&fop->loc[1]);
+ GF_FREE(fop->errstr);
- ec_resume_parent(fop, fop->error);
+ ec_resume_parent(fop);
ec_fop_cleanup(fop);
ec = fop->xl->private;
- ec_handle_last_pending_fop_completion (fop, &notify);
- ec_handle_healers_done (fop);
+ ec_handle_last_pending_fop_completion(fop, &notify);
+ ec_handle_healers_done(fop);
mem_put(fop);
if (notify) {
ec_pending_fops_completed(ec);
diff --git a/xlators/cluster/ec/src/ec-data.h b/xlators/cluster/ec/src/ec-data.h
index 965bc1e25ef..c8a74ffe1ed 100644
--- a/xlators/cluster/ec/src/ec-data.h
+++ b/xlators/cluster/ec/src/ec-data.h
@@ -13,20 +13,23 @@
#include "ec-types.h"
-ec_cbk_data_t * ec_cbk_data_allocate(call_frame_t * frame, xlator_t * this,
- ec_fop_data_t * fop, int32_t id,
- int32_t idx, int32_t op_ret,
- int32_t op_errno);
-ec_fop_data_t * ec_fop_data_allocate(call_frame_t * frame, xlator_t * this,
- int32_t id, uint32_t flags,
- uintptr_t target, int32_t minimum,
- ec_wind_f wind, ec_handler_f handler,
- ec_cbk_t cbks, void * data);
-void ec_fop_data_acquire(ec_fop_data_t * fop);
-void ec_fop_data_release(ec_fop_data_t * fop);
+ec_cbk_data_t *
+ec_cbk_data_allocate(call_frame_t *frame, xlator_t *this, ec_fop_data_t *fop,
+ int32_t id, int32_t idx, int32_t op_ret, int32_t op_errno);
+ec_fop_data_t *
+ec_fop_data_allocate(call_frame_t *frame, xlator_t *this, int32_t id,
+ uint32_t flags, uintptr_t target, uint32_t fop_flags,
+ ec_wind_f wind, ec_handler_f handler, ec_cbk_t cbks,
+ void *data);
+void
+ec_fop_data_acquire(ec_fop_data_t *fop);
+void
+ec_fop_data_release(ec_fop_data_t *fop);
-void ec_fop_cleanup(ec_fop_data_t *fop);
+void
+ec_fop_cleanup(ec_fop_data_t *fop);
-void ec_pending_fops_completed(ec_t *ec);
+void
+ec_pending_fops_completed(ec_t *ec);
#endif /* __EC_DATA_H__ */
diff --git a/xlators/cluster/ec/src/ec-dir-read.c b/xlators/cluster/ec/src/ec-dir-read.c
index b44bb4239b1..f71dcfac293 100644
--- a/xlators/cluster/ec/src/ec-dir-read.c
+++ b/xlators/cluster/ec/src/ec-dir-read.c
@@ -8,15 +8,11 @@
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "defaults.h"
-
#include "ec.h"
#include "ec-messages.h"
#include "ec-helpers.h"
#include "ec-common.h"
#include "ec-combine.h"
-#include "ec-method.h"
#include "ec-fops.h"
/****************************************************************
@@ -25,15 +21,14 @@
*
***************************************************************/
-int32_t ec_combine_opendir(ec_fop_data_t * fop, ec_cbk_data_t * dst,
- ec_cbk_data_t * src)
+int32_t
+ec_combine_opendir(ec_fop_data_t *fop, ec_cbk_data_t *dst, ec_cbk_data_t *src)
{
- if (dst->fd != src->fd)
- {
- gf_msg (fop->xl->name, GF_LOG_NOTICE, 0,
- EC_MSG_FD_MISMATCH, "Mismatching fd in answers "
- "of 'GF_FOP_OPENDIR': %p <-> %p",
- dst->fd, src->fd);
+ if (dst->fd != src->fd) {
+ gf_msg(fop->xl->name, GF_LOG_NOTICE, 0, EC_MSG_FD_MISMATCH,
+ "Mismatching fd in answers "
+ "of 'GF_FOP_OPENDIR': %p <-> %p",
+ dst->fd, src->fd);
return 0;
}
@@ -41,12 +36,12 @@ int32_t ec_combine_opendir(ec_fop_data_t * fop, ec_cbk_data_t * dst,
return 1;
}
-int32_t ec_opendir_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, fd_t * fd,
- dict_t * xdata)
+int32_t
+ec_opendir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -56,36 +51,31 @@ int32_t ec_opendir_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_OPENDIR, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (op_ret >= 0)
- {
- if (fd != NULL)
- {
+ if (cbk != NULL) {
+ if (op_ret >= 0) {
+ if (fd != NULL) {
cbk->fd = fd_ref(fd);
- if (cbk->fd == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL, "Failed to reference a "
- "file descriptor.");
+ if (cbk->fd == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
}
}
- if (xdata != NULL)
- {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -93,19 +83,19 @@ int32_t ec_opendir_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
ec_combine(cbk, ec_combine_opendir);
- ec_update_fd_status (fd, this, idx, op_ret);
+ ec_update_fd_status(fd, this, idx, op_ret);
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_opendir(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_opendir(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -114,14 +104,14 @@ void ec_wind_opendir(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
&fop->loc[0], fop->fd, fop->xdata);
}
-int32_t ec_manager_opendir(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_opendir(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
ec_fd_t *ctx;
int32_t err;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
LOCK(&fop->fd->lock);
@@ -133,13 +123,15 @@ int32_t ec_manager_opendir(ec_fop_data_t * fop, int32_t state)
return EC_STATE_REPORT;
}
- err = ec_loc_from_loc(fop->xl, &ctx->loc, &fop->loc[0]);
- if (err != 0) {
- UNLOCK(&fop->fd->lock);
+ if (!ctx->loc.inode) {
+ err = ec_loc_from_loc(fop->xl, &ctx->loc, &fop->loc[0]);
+ if (err != 0) {
+ UNLOCK(&fop->fd->lock);
- fop->error = -err;
+ fop->error = -err;
- return EC_STATE_REPORT;
+ return EC_STATE_REPORT;
+ }
}
UNLOCK(&fop->fd->lock);
@@ -148,7 +140,7 @@ int32_t ec_manager_opendir(ec_fop_data_t * fop, int32_t state)
case EC_STATE_LOCK:
ec_lock_prepare_inode(fop, &fop->loc[0], EC_QUERY_INFO, 0,
- LLONG_MAX);
+ EC_RANGE_FULL);
ec_lock(fop);
return EC_STATE_DISPATCH;
@@ -182,8 +174,7 @@ int32_t ec_manager_opendir(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.opendir != NULL)
- {
+ if (fop->cbks.opendir != NULL) {
fop->cbks.opendir(fop->req_frame, fop, fop->xl, cbk->op_ret,
cbk->op_errno, cbk->fd, cbk->xdata);
}
@@ -197,8 +188,7 @@ int32_t ec_manager_opendir(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.opendir != NULL)
- {
+ if (fop->cbks.opendir != NULL) {
fop->cbks.opendir(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL, NULL);
}
@@ -218,40 +208,39 @@ int32_t ec_manager_opendir(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_opendir(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_opendir_cbk_t func, void * data,
- loc_t * loc, fd_t * fd, dict_t * xdata)
+void
+ec_opendir(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_opendir_cbk_t func, void *data, loc_t *loc,
+ fd_t *fd, dict_t *xdata)
{
- ec_cbk_t callback = { .opendir = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.opendir = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(OPENDIR) %p", frame);
+ gf_msg_trace("ec", 0, "EC(OPENDIR) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_OPENDIR,
- EC_FLAG_LOCK_SHARED, target, minimum,
- ec_wind_opendir, ec_manager_opendir, callback,
- data);
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_OPENDIR, EC_FLAG_LOCK_SHARED,
+ target, fop_flags, ec_wind_opendir,
+ ec_manager_opendir, callback, data);
if (fop == NULL) {
goto out;
}
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
@@ -259,9 +248,9 @@ void ec_opendir(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL, "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -269,9 +258,9 @@ void ec_opendir(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -289,50 +278,49 @@ out:
/* Returns -1 if client_id is invalid else index of child subvol in xl_list */
int
-ec_deitransform (xlator_t *this, off_t offset)
+ec_deitransform(xlator_t *this, off_t offset)
{
- int idx = -1;
- int client_id = -1;
- ec_t *ec = this->private;
- char id[32] = {0};
- int err;
-
- client_id = gf_deitransform (this, offset);
- sprintf (id, "%d", client_id);
- err = dict_get_int32 (ec->leaf_to_subvolid, id, &idx);
- if (err < 0) {
- idx = err;
- goto out;
- }
+ int idx = -1;
+ int client_id = -1;
+ ec_t *ec = this->private;
+ char id[32] = {0};
+ int err;
+
+ client_id = gf_deitransform(this, offset);
+ sprintf(id, "%d", client_id);
+ err = dict_get_int32(ec->leaf_to_subvolid, id, &idx);
+ if (err < 0) {
+ idx = err;
+ goto out;
+ }
out:
- if (idx < 0) {
- gf_msg (this->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_INVALID_REQUEST,
- "Invalid index %d in readdirp request", client_id);
- idx = -EINVAL;
- }
- return idx;
+ if (idx < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL, EC_MSG_INVALID_REQUEST,
+ "Invalid index %d in readdirp request", client_id);
+ idx = -EINVAL;
+ }
+ return idx;
}
/* FOP: readdir */
-void ec_adjust_readdirp (ec_t *ec, int32_t idx, gf_dirent_t *entries)
+void
+ec_adjust_readdirp(ec_t *ec, int32_t idx, gf_dirent_t *entries)
{
- gf_dirent_t * entry;
+ gf_dirent_t *entry;
list_for_each_entry(entry, &entries->list, list)
{
if (!entry->inode)
- continue;
+ continue;
- if (entry->d_stat.ia_type == IA_IFREG)
- {
+ if (entry->d_stat.ia_type == IA_IFREG) {
if ((entry->dict == NULL) ||
(ec_dict_del_number(entry->dict, EC_XATTR_SIZE,
&entry->d_stat.ia_size) != 0)) {
- inode_unref (entry->inode);
- entry->inode = NULL;
+ inode_unref(entry->inode);
+ entry->inode = NULL;
} else {
ec_iatt_rebuild(ec, &entry->d_stat, 1, 1);
}
@@ -341,9 +329,9 @@ void ec_adjust_readdirp (ec_t *ec, int32_t idx, gf_dirent_t *entries)
}
int32_t
-ec_common_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- gf_dirent_t *entries, dict_t *xdata)
+ec_common_readdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+ dict_t *xdata)
{
ec_fop_data_t *fop = NULL;
ec_cbk_data_t *cbk = NULL;
@@ -356,30 +344,29 @@ ec_common_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
- cbk = ec_cbk_data_allocate (frame, this, fop, fop->id,
- idx, op_ret, op_errno);
+ cbk = ec_cbk_data_allocate(frame, this, fop, fop->id, idx, op_ret,
+ op_errno);
if (cbk) {
if (xdata)
- cbk->xdata = dict_ref (xdata);
+ cbk->xdata = dict_ref(xdata);
if (cbk->op_ret >= 0)
- list_splice_init (&entries->list,
- &cbk->entries.list);
- ec_combine (cbk, NULL);
+ list_splice_init(&entries->list, &cbk->entries.list);
+ ec_combine(cbk, NULL);
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_readdir(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_readdir(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -388,60 +375,66 @@ void ec_wind_readdir(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->fd, fop->size, fop->offset, fop->xdata);
}
-int32_t ec_manager_readdir(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_readdir(ec_fop_data_t *fop, int32_t state)
{
ec_fd_t *ctx = NULL;
ec_cbk_data_t *cbk = NULL;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
/* Return error if opendir has not been successfully called on
* any subvolume. */
ctx = ec_fd_get(fop->fd, fop->xl);
- if ((ctx == NULL) || (ctx->open == 0)) {
- fop->error = EINVAL;
+ if (ctx == NULL) {
+ fop->error = ENOMEM;
+ } else if (ctx->open == 0) {
+ fop->error = EBADFD;
+ }
+ if (fop->error) {
+ gf_msg(fop->xl->name, GF_LOG_ERROR, fop->error,
+ EC_MSG_INVALID_REQUEST, "EC is not winding readdir: %s",
+ ec_msg_str(fop));
return EC_STATE_REPORT;
}
if (fop->id == GF_FOP_READDIRP) {
- int32_t err;
+ int32_t err;
+ if (fop->xdata == NULL) {
+ fop->xdata = dict_new();
if (fop->xdata == NULL) {
- fop->xdata = dict_new();
- if (fop->xdata == NULL) {
- fop->error = ENOMEM;
+ fop->error = ENOMEM;
- return EC_STATE_REPORT;
- }
+ return EC_STATE_REPORT;
}
+ }
- err = dict_set_uint64(fop->xdata, EC_XATTR_SIZE, 0);
- if (err != 0) {
- fop->error = -err;
+ err = dict_set_uint64(fop->xdata, EC_XATTR_SIZE, 0);
+ if (err != 0) {
+ fop->error = -err;
- return EC_STATE_REPORT;
- }
+ return EC_STATE_REPORT;
+ }
}
- if (fop->offset != 0)
- {
- /* Non-zero offset is irrecoverable error as the offset may not be
- * valid on other bricks*/
+ if (fop->offset != 0) {
+ /* Non-zero offset is irrecoverable error as the offset may not
+ * be valid on other bricks*/
int32_t idx = -1;
- idx = ec_deitransform (fop->xl, fop->offset);
+ idx = ec_deitransform(fop->xl, fop->offset);
if (idx < 0) {
- fop->error = -idx;
- return EC_STATE_REPORT;
+ fop->error = -idx;
+ return EC_STATE_REPORT;
}
fop->mask &= 1ULL << idx;
} else {
- ec_lock_prepare_fd(fop, fop->fd, EC_QUERY_INFO, 0,
- LLONG_MAX);
- ec_lock(fop);
+ ec_lock_prepare_fd(fop, fop->fd, EC_QUERY_INFO, 0,
+ EC_RANGE_FULL);
+ ec_lock(fop);
}
return EC_STATE_DISPATCH;
@@ -458,14 +451,14 @@ int32_t ec_manager_readdir(ec_fop_data_t * fop, int32_t state)
if ((cbk != NULL) && (cbk->op_ret > 0) &&
(fop->id == GF_FOP_READDIRP)) {
- ec_adjust_readdirp (fop->xl->private, cbk->idx, &cbk->entries);
+ ec_adjust_readdirp(fop->xl->private, cbk->idx, &cbk->entries);
}
return EC_STATE_REPORT;
case EC_STATE_REPORT:
cbk = fop->answer;
- GF_ASSERT (cbk);
+ GF_ASSERT(cbk);
if (fop->id == GF_FOP_READDIR) {
if (fop->cbks.readdir != NULL) {
fop->cbks.readdir(fop->req_frame, fop, fop->xl, cbk->op_ret,
@@ -479,9 +472,9 @@ int32_t ec_manager_readdir(ec_fop_data_t * fop, int32_t state)
}
}
if (fop->offset == 0)
- return EC_STATE_LOCK_REUSE;
+ return EC_STATE_LOCK_REUSE;
else
- return EC_STATE_END;
+ return EC_STATE_END;
case -EC_STATE_INIT:
case -EC_STATE_LOCK:
@@ -500,50 +493,49 @@ int32_t ec_manager_readdir(ec_fop_data_t * fop, int32_t state)
}
}
if (fop->offset == 0)
- return EC_STATE_LOCK_REUSE;
+ return EC_STATE_LOCK_REUSE;
else
- return EC_STATE_END;
+ return EC_STATE_END;
case -EC_STATE_LOCK_REUSE:
case EC_STATE_LOCK_REUSE:
- GF_ASSERT (fop->offset == 0);
+ GF_ASSERT(fop->offset == 0);
ec_lock_reuse(fop);
return EC_STATE_UNLOCK;
case -EC_STATE_UNLOCK:
case EC_STATE_UNLOCK:
- GF_ASSERT (fop->offset == 0);
+ GF_ASSERT(fop->offset == 0);
ec_unlock(fop);
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_readdir(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_readdir_cbk_t func, void * data,
- fd_t * fd, size_t size, off_t offset, dict_t * xdata)
+void
+ec_readdir(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_readdir_cbk_t func, void *data, fd_t *fd,
+ size_t size, off_t offset, dict_t *xdata)
{
- ec_cbk_t callback = { .readdir = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.readdir = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(READDIR) %p", frame);
+ gf_msg_trace("ec", 0, "EC(READDIR) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_READDIR,
- EC_FLAG_LOCK_SHARED, target, minimum,
- ec_wind_readdir, ec_manager_readdir, callback,
- data);
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_READDIR, EC_FLAG_LOCK_SHARED,
+ target, fop_flags, ec_wind_readdir,
+ ec_manager_readdir, callback, data);
if (fop == NULL) {
goto out;
}
@@ -556,9 +548,9 @@ void ec_readdir(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL, "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -566,9 +558,9 @@ void ec_readdir(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -586,7 +578,8 @@ out:
/* FOP: readdirp */
-void ec_wind_readdirp(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_readdirp(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -595,24 +588,24 @@ void ec_wind_readdirp(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->fd, fop->size, fop->offset, fop->xdata);
}
-void ec_readdirp(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_readdirp_cbk_t func, void * data,
- fd_t * fd, size_t size, off_t offset, dict_t * xdata)
+void
+ec_readdirp(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_readdirp_cbk_t func, void *data, fd_t *fd,
+ size_t size, off_t offset, dict_t *xdata)
{
- ec_cbk_t callback = { .readdirp = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.readdirp = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(READDIRP) %p", frame);
+ gf_msg_trace("ec", 0, "EC(READDIRP) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_READDIRP,
- EC_FLAG_LOCK_SHARED, target, minimum,
- ec_wind_readdirp, ec_manager_readdir, callback,
- data);
+ fop = ec_fop_data_allocate(
+ frame, this, GF_FOP_READDIRP, EC_FLAG_LOCK_SHARED, target, fop_flags,
+ ec_wind_readdirp, ec_manager_readdir, callback, data);
if (fop == NULL) {
goto out;
}
@@ -625,9 +618,9 @@ void ec_readdirp(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL, "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -635,9 +628,9 @@ void ec_readdirp(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
diff --git a/xlators/cluster/ec/src/ec-dir-write.c b/xlators/cluster/ec/src/ec-dir-write.c
index 7779d4849f3..53d27d895c3 100644
--- a/xlators/cluster/ec/src/ec-dir-write.c
+++ b/xlators/cluster/ec/src/ec-dir-write.c
@@ -8,9 +8,6 @@
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "defaults.h"
-
#include "ec.h"
#include "ec-messages.h"
#include "ec-helpers.h"
@@ -20,76 +17,75 @@
#include "ec-fops.h"
int
-ec_dir_write_cbk (call_frame_t *frame, xlator_t *this,
- void *cookie, int op_ret, int op_errno,
- struct iatt *poststat, struct iatt *preparent,
- struct iatt *postparent, struct iatt *preparent2,
- struct iatt *postparent2, dict_t *xdata)
+ec_dir_write_cbk(call_frame_t *frame, xlator_t *this, void *cookie, int op_ret,
+ int op_errno, struct iatt *poststat, struct iatt *preparent,
+ struct iatt *postparent, struct iatt *preparent2,
+ struct iatt *postparent2, dict_t *xdata)
{
- ec_fop_data_t *fop = NULL;
- ec_cbk_data_t *cbk = NULL;
- int i = 0;
- int idx = 0;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
+ int i = 0;
+ int idx = 0;
- VALIDATE_OR_GOTO (this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, frame->local, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
+ VALIDATE_OR_GOTO(this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame->local, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = frame->local;
- idx = (long) cookie;
+ fop = frame->local;
+ idx = (long)cookie;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
- cbk = ec_cbk_data_allocate (frame, this, fop, fop->id, idx, op_ret,
- op_errno);
- if (!cbk)
- goto out;
+ cbk = ec_cbk_data_allocate(frame, this, fop, fop->id, idx, op_ret,
+ op_errno);
+ if (!cbk)
+ goto out;
- if (xdata)
- cbk->xdata = dict_ref (xdata);
+ if (xdata)
+ cbk->xdata = dict_ref(xdata);
- if (op_ret < 0)
- goto out;
+ if (op_ret < 0)
+ goto out;
- if (poststat)
- cbk->iatt[i++] = *poststat;
+ if (poststat)
+ cbk->iatt[i++] = *poststat;
- if (preparent)
- cbk->iatt[i++] = *preparent;
+ if (preparent)
+ cbk->iatt[i++] = *preparent;
- if (postparent)
- cbk->iatt[i++] = *postparent;
+ if (postparent)
+ cbk->iatt[i++] = *postparent;
- if (preparent2)
- cbk->iatt[i++] = *preparent2;
+ if (preparent2)
+ cbk->iatt[i++] = *preparent2;
- if (postparent2)
- cbk->iatt[i++] = *postparent2;
+ if (postparent2)
+ cbk->iatt[i++] = *postparent2;
out:
- if (cbk)
- ec_combine (cbk, ec_combine_write);
+ if (cbk)
+ ec_combine(cbk, ec_combine_write);
- if (fop)
- ec_complete (fop);
- return 0;
+ if (fop)
+ ec_complete(fop);
+ return 0;
}
/* FOP: create */
-int32_t ec_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd,
- inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+int32_t
+ec_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, fd_t *fd, inode_t *inode, struct iatt *buf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
{
- return ec_dir_write_cbk (frame, this, cookie, op_ret, op_errno,
- buf, preparent, postparent, NULL, NULL, xdata);
+ return ec_dir_write_cbk(frame, this, cookie, op_ret, op_errno, buf,
+ preparent, postparent, NULL, NULL, xdata);
}
-void ec_wind_create(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_create(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -99,7 +95,8 @@ void ec_wind_create(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->fd, fop->xdata);
}
-int32_t ec_manager_create(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_create(ec_fop_data_t *fop, int32_t state)
{
ec_config_t config;
ec_t *ec;
@@ -108,8 +105,7 @@ int32_t ec_manager_create(ec_fop_data_t * fop, int32_t state)
uint64_t version[2] = {0, 0};
int32_t err;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
LOCK(&fop->fd->lock);
@@ -176,7 +172,7 @@ int32_t ec_manager_create(ec_fop_data_t * fop, int32_t state)
* need to remove O_APPEND from flags (if present) */
fop->int32 &= ~O_APPEND;
- /* Fall through */
+ /* Fall through */
case EC_STATE_LOCK:
ec_lock_prepare_parent_inode(fop, &fop->loc[0], NULL,
@@ -218,12 +214,11 @@ int32_t ec_manager_create(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.create != NULL)
- {
- fop->cbks.create (fop->req_frame, fop, fop->xl, cbk->op_ret,
- cbk->op_errno, fop->fd, fop->loc[0].inode,
- &cbk->iatt[0], &cbk->iatt[1], &cbk->iatt[2],
- cbk->xdata);
+ if (fop->cbks.create != NULL) {
+ QUORUM_CBK(fop->cbks.create, fop, fop->req_frame, fop, fop->xl,
+ cbk->op_ret, cbk->op_errno, fop->fd,
+ fop->loc[0].inode, &cbk->iatt[0], &cbk->iatt[1],
+ &cbk->iatt[2], cbk->xdata);
}
return EC_STATE_LOCK_REUSE;
@@ -235,8 +230,7 @@ int32_t ec_manager_create(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.create != NULL)
- {
+ if (fop->cbks.create != NULL) {
fop->cbks.create(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL, NULL, NULL, NULL, NULL, NULL);
}
@@ -256,30 +250,29 @@ int32_t ec_manager_create(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_create(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_create_cbk_t func, void * data,
- loc_t * loc, int32_t flags, mode_t mode, mode_t umask,
- fd_t * fd, dict_t * xdata)
+void
+ec_create(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_create_cbk_t func, void *data, loc_t *loc,
+ int32_t flags, mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- ec_cbk_t callback = { .create = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.create = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(CREATE) %p", frame);
+ gf_msg_trace("ec", 0, "EC(CREATE) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_CREATE, 0, target, minimum,
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_CREATE, 0, target, fop_flags,
ec_wind_create, ec_manager_create, callback,
data);
if (fop == NULL) {
@@ -292,8 +285,8 @@ void ec_create(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
@@ -301,19 +294,19 @@ void ec_create(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL, "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
}
if (xdata != NULL) {
- fop->xdata = dict_copy_with_ref (xdata, NULL);
+ fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -331,16 +324,17 @@ out:
/* FOP: link */
-int32_t ec_link_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, inode_t * inode,
- struct iatt * buf, struct iatt * preparent,
- struct iatt * postparent, dict_t * xdata)
+int32_t
+ec_link_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode, struct iatt *buf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
{
- return ec_dir_write_cbk (frame, this, cookie, op_ret, op_errno,
- buf, preparent, postparent, NULL, NULL, xdata);
+ return ec_dir_write_cbk(frame, this, cookie, op_ret, op_errno, buf,
+ preparent, postparent, NULL, NULL, xdata);
}
-void ec_wind_link(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_link(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -349,17 +343,17 @@ void ec_wind_link(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
&fop->loc[0], &fop->loc[1], fop->xdata);
}
-int32_t ec_manager_link(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_link(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_LOCK:
- ec_lock_prepare_parent_inode(fop, &fop->loc[1], &fop->loc[0],
- EC_UPDATE_DATA | EC_UPDATE_META |
- EC_INODE_SIZE);
+ ec_lock_prepare_parent_inode(
+ fop, &fop->loc[1], &fop->loc[0],
+ EC_UPDATE_DATA | EC_UPDATE_META | EC_INODE_SIZE);
ec_lock(fop);
return EC_STATE_DISPATCH;
@@ -392,11 +386,11 @@ int32_t ec_manager_link(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.link != NULL)
- {
- fop->cbks.link(fop->req_frame, fop, fop->xl, cbk->op_ret,
- cbk->op_errno, fop->loc[0].inode, &cbk->iatt[0],
- &cbk->iatt[1], &cbk->iatt[2], cbk->xdata);
+ if (fop->cbks.link != NULL) {
+ QUORUM_CBK(fop->cbks.link, fop, fop->req_frame, fop, fop->xl,
+ cbk->op_ret, cbk->op_errno, fop->loc[0].inode,
+ &cbk->iatt[0], &cbk->iatt[1], &cbk->iatt[2],
+ cbk->xdata);
}
return EC_STATE_LOCK_REUSE;
@@ -408,8 +402,7 @@ int32_t ec_manager_link(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.link != NULL)
- {
+ if (fop->cbks.link != NULL) {
fop->cbks.link(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL, NULL, NULL, NULL, NULL);
}
@@ -429,29 +422,29 @@ int32_t ec_manager_link(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_link(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_link_cbk_t func, void * data, loc_t * oldloc,
- loc_t * newloc, dict_t * xdata)
+void
+ec_link(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_link_cbk_t func, void *data, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata)
{
- ec_cbk_t callback = { .link = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.link = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(LINK) %p", frame);
+ gf_msg_trace("ec", 0, "EC(LINK) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_LINK, 0, target, minimum,
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_LINK, 0, target, fop_flags,
ec_wind_link, ec_manager_link, callback, data);
if (fop == NULL) {
goto out;
@@ -459,26 +452,26 @@ void ec_link(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (oldloc != NULL) {
if (loc_copy(&fop->loc[0], oldloc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
}
if (newloc != NULL) {
if (loc_copy(&fop->loc[1], newloc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
}
if (xdata != NULL) {
- fop->xdata = dict_copy_with_ref (xdata, NULL);
+ fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -496,16 +489,17 @@ out:
/* FOP: mkdir */
-int32_t ec_mkdir_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, inode_t * inode,
- struct iatt * buf, struct iatt * preparent,
- struct iatt * postparent, dict_t * xdata)
+int32_t
+ec_mkdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode, struct iatt *buf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
{
- return ec_dir_write_cbk (frame, this, cookie, op_ret, op_errno,
- buf, preparent, postparent, NULL, NULL, xdata);
+ return ec_dir_write_cbk(frame, this, cookie, op_ret, op_errno, buf,
+ preparent, postparent, NULL, NULL, xdata);
}
-void ec_wind_mkdir(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_mkdir(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -514,14 +508,14 @@ void ec_wind_mkdir(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
&fop->loc[0], fop->mode[0], fop->mode[1], fop->xdata);
}
-int32_t ec_manager_mkdir(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_mkdir(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
uint64_t version[2] = {0, 0};
int32_t err;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
if (fop->xdata == NULL) {
fop->xdata = dict_new();
@@ -539,7 +533,7 @@ int32_t ec_manager_mkdir(ec_fop_data_t * fop, int32_t state)
return EC_STATE_REPORT;
}
- /* Fall through */
+ /* Fall through */
case EC_STATE_LOCK:
ec_lock_prepare_parent_inode(fop, &fop->loc[0], NULL,
@@ -572,11 +566,11 @@ int32_t ec_manager_mkdir(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.mkdir != NULL)
- {
- fop->cbks.mkdir(fop->req_frame, fop, fop->xl, cbk->op_ret,
- cbk->op_errno, fop->loc[0].inode, &cbk->iatt[0],
- &cbk->iatt[1], &cbk->iatt[2], cbk->xdata);
+ if (fop->cbks.mkdir != NULL) {
+ QUORUM_CBK(fop->cbks.mkdir, fop, fop->req_frame, fop, fop->xl,
+ cbk->op_ret, cbk->op_errno, fop->loc[0].inode,
+ &cbk->iatt[0], &cbk->iatt[1], &cbk->iatt[2],
+ cbk->xdata);
}
return EC_STATE_LOCK_REUSE;
@@ -589,8 +583,7 @@ int32_t ec_manager_mkdir(ec_fop_data_t * fop, int32_t state)
cbk = fop->answer;
GF_ASSERT(fop->error != 0);
- if (fop->cbks.mkdir != NULL)
- {
+ if (fop->cbks.mkdir != NULL) {
fop->cbks.mkdir(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL, NULL, NULL, NULL,
((cbk) ? cbk->xdata : NULL));
@@ -611,31 +604,30 @@ int32_t ec_manager_mkdir(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_mkdir(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_mkdir_cbk_t func, void * data, loc_t * loc,
- mode_t mode, mode_t umask, dict_t * xdata)
+void
+ec_mkdir(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_mkdir_cbk_t func, void *data, loc_t *loc,
+ mode_t mode, mode_t umask, dict_t *xdata)
{
- ec_cbk_t callback = { .mkdir = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.mkdir = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(MKDIR) %p", frame);
+ gf_msg_trace("ec", 0, "EC(MKDIR) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_MKDIR, 0, target, minimum,
- ec_wind_mkdir, ec_manager_mkdir, callback,
- data);
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_MKDIR, 0, target, fop_flags,
+ ec_wind_mkdir, ec_manager_mkdir, callback, data);
if (fop == NULL) {
goto out;
}
@@ -645,18 +637,18 @@ void ec_mkdir(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
}
if (xdata != NULL) {
- fop->xdata = dict_copy_with_ref (xdata, NULL);
+ fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -674,16 +666,17 @@ out:
/* FOP: mknod */
-int32_t ec_mknod_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, inode_t * inode,
- struct iatt * buf, struct iatt * preparent,
- struct iatt * postparent, dict_t * xdata)
+int32_t
+ec_mknod_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode, struct iatt *buf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
{
- return ec_dir_write_cbk (frame, this, cookie, op_ret, op_errno,
- buf, preparent, postparent, NULL, NULL, xdata);
+ return ec_dir_write_cbk(frame, this, cookie, op_ret, op_errno, buf,
+ preparent, postparent, NULL, NULL, xdata);
}
-void ec_wind_mknod(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_mknod(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -693,15 +686,15 @@ void ec_wind_mknod(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->xdata);
}
-int32_t ec_manager_mknod(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_mknod(ec_fop_data_t *fop, int32_t state)
{
ec_config_t config;
ec_t *ec;
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
uint64_t version[2] = {0, 0};
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
if (S_ISREG(fop->mode[0])) {
int32_t err;
@@ -745,7 +738,7 @@ int32_t ec_manager_mknod(ec_fop_data_t * fop, int32_t state)
}
}
- /* Fall through */
+ /* Fall through */
case EC_STATE_LOCK:
ec_lock_prepare_parent_inode(fop, &fop->loc[0], NULL,
@@ -778,11 +771,11 @@ int32_t ec_manager_mknod(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.mknod != NULL)
- {
- fop->cbks.mknod(fop->req_frame, fop, fop->xl, cbk->op_ret,
- cbk->op_errno, fop->loc[0].inode, &cbk->iatt[0],
- &cbk->iatt[1], &cbk->iatt[2], cbk->xdata);
+ if (fop->cbks.mknod != NULL) {
+ QUORUM_CBK(fop->cbks.mknod, fop, fop->req_frame, fop, fop->xl,
+ cbk->op_ret, cbk->op_errno, fop->loc[0].inode,
+ &cbk->iatt[0], &cbk->iatt[1], &cbk->iatt[2],
+ cbk->xdata);
}
return EC_STATE_LOCK_REUSE;
@@ -794,8 +787,7 @@ int32_t ec_manager_mknod(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.mknod != NULL)
- {
+ if (fop->cbks.mknod != NULL) {
fop->cbks.mknod(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL, NULL, NULL, NULL, NULL);
}
@@ -815,31 +807,30 @@ int32_t ec_manager_mknod(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_mknod(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_mknod_cbk_t func, void * data, loc_t * loc,
- mode_t mode, dev_t rdev, mode_t umask, dict_t * xdata)
+void
+ec_mknod(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_mknod_cbk_t func, void *data, loc_t *loc,
+ mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata)
{
- ec_cbk_t callback = { .mknod = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.mknod = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(MKNOD) %p", frame);
+ gf_msg_trace("ec", 0, "EC(MKNOD) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_MKNOD, 0, target, minimum,
- ec_wind_mknod, ec_manager_mknod, callback,
- data);
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_MKNOD, 0, target, fop_flags,
+ ec_wind_mknod, ec_manager_mknod, callback, data);
if (fop == NULL) {
goto out;
}
@@ -850,18 +841,18 @@ void ec_mknod(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
}
if (xdata != NULL) {
- fop->xdata = dict_copy_with_ref (xdata, NULL);
+ fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -879,18 +870,19 @@ out:
/* FOP: rename */
-int32_t ec_rename_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, struct iatt * buf,
- struct iatt * preoldparent, struct iatt * postoldparent,
- struct iatt * prenewparent, struct iatt * postnewparent,
- dict_t * xdata)
+int32_t
+ec_rename_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *buf, struct iatt *preoldparent,
+ struct iatt *postoldparent, struct iatt *prenewparent,
+ struct iatt *postnewparent, dict_t *xdata)
{
- return ec_dir_write_cbk (frame, this, cookie, op_ret, op_errno,
- buf, preoldparent, postoldparent, prenewparent,
- postnewparent, xdata);
+ return ec_dir_write_cbk(frame, this, cookie, op_ret, op_errno, buf,
+ preoldparent, postoldparent, prenewparent,
+ postnewparent, xdata);
}
-void ec_wind_rename(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_rename(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -899,17 +891,17 @@ void ec_wind_rename(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
&fop->loc[0], &fop->loc[1], fop->xdata);
}
-int32_t ec_manager_rename(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_rename(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_LOCK:
- ec_lock_prepare_parent_inode(fop, &fop->loc[0], &fop->loc[0],
- EC_UPDATE_DATA | EC_UPDATE_META |
- EC_INODE_SIZE);
+ ec_lock_prepare_parent_inode(
+ fop, &fop->loc[0], &fop->loc[0],
+ EC_UPDATE_DATA | EC_UPDATE_META | EC_INODE_SIZE);
ec_lock_prepare_parent_inode(fop, &fop->loc[1], NULL,
EC_UPDATE_DATA | EC_UPDATE_META);
ec_lock(fop);
@@ -938,12 +930,11 @@ int32_t ec_manager_rename(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.rename != NULL)
- {
- fop->cbks.rename(fop->req_frame, fop, fop->xl, cbk->op_ret,
- cbk->op_errno, &cbk->iatt[0], &cbk->iatt[1],
- &cbk->iatt[2], &cbk->iatt[3], &cbk->iatt[4],
- cbk->xdata);
+ if (fop->cbks.rename != NULL) {
+ QUORUM_CBK(fop->cbks.rename, fop, fop->req_frame, fop, fop->xl,
+ cbk->op_ret, cbk->op_errno, &cbk->iatt[0],
+ &cbk->iatt[1], &cbk->iatt[2], &cbk->iatt[3],
+ &cbk->iatt[4], cbk->xdata);
}
return EC_STATE_LOCK_REUSE;
@@ -955,8 +946,7 @@ int32_t ec_manager_rename(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.rename != NULL)
- {
+ if (fop->cbks.rename != NULL) {
fop->cbks.rename(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL, NULL, NULL, NULL, NULL, NULL);
}
@@ -976,29 +966,29 @@ int32_t ec_manager_rename(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_rename(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_rename_cbk_t func, void * data,
- loc_t * oldloc, loc_t * newloc, dict_t * xdata)
+void
+ec_rename(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_rename_cbk_t func, void *data, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata)
{
- ec_cbk_t callback = { .rename = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.rename = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(RENAME) %p", frame);
+ gf_msg_trace("ec", 0, "EC(RENAME) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_RENAME, 0, target, minimum,
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_RENAME, 0, target, fop_flags,
ec_wind_rename, ec_manager_rename, callback,
data);
if (fop == NULL) {
@@ -1007,26 +997,26 @@ void ec_rename(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (oldloc != NULL) {
if (loc_copy(&fop->loc[0], oldloc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
}
if (newloc != NULL) {
if (loc_copy(&fop->loc[1], newloc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
}
if (xdata != NULL) {
- fop->xdata = dict_copy_with_ref (xdata, NULL);
+ fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -1044,15 +1034,17 @@ out:
/* FOP: rmdir */
-int32_t ec_rmdir_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, struct iatt * preparent,
- struct iatt * postparent, dict_t * xdata)
+int32_t
+ec_rmdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- return ec_dir_write_cbk (frame, this, cookie, op_ret, op_errno, NULL,
- preparent, postparent, NULL, NULL, xdata);
+ return ec_dir_write_cbk(frame, this, cookie, op_ret, op_errno, NULL,
+ preparent, postparent, NULL, NULL, xdata);
}
-void ec_wind_rmdir(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_rmdir(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -1061,12 +1053,12 @@ void ec_wind_rmdir(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
&fop->loc[0], fop->int32, fop->xdata);
}
-int32_t ec_manager_rmdir(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_rmdir(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_LOCK:
ec_lock_prepare_parent_inode(fop, &fop->loc[0], NULL,
@@ -1090,11 +1082,10 @@ int32_t ec_manager_rmdir(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.rmdir != NULL)
- {
- fop->cbks.rmdir(fop->req_frame, fop, fop->xl, cbk->op_ret,
- cbk->op_errno, &cbk->iatt[0], &cbk->iatt[1],
- cbk->xdata);
+ if (fop->cbks.rmdir != NULL) {
+ QUORUM_CBK(fop->cbks.rmdir, fop, fop->req_frame, fop, fop->xl,
+ cbk->op_ret, cbk->op_errno, &cbk->iatt[0],
+ &cbk->iatt[1], cbk->xdata);
}
return EC_STATE_LOCK_REUSE;
@@ -1106,8 +1097,7 @@ int32_t ec_manager_rmdir(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.rmdir != NULL)
- {
+ if (fop->cbks.rmdir != NULL) {
fop->cbks.rmdir(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL, NULL, NULL);
}
@@ -1127,31 +1117,30 @@ int32_t ec_manager_rmdir(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_rmdir(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_rmdir_cbk_t func, void * data, loc_t * loc,
- int xflags, dict_t * xdata)
+void
+ec_rmdir(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_rmdir_cbk_t func, void *data, loc_t *loc,
+ int xflags, dict_t *xdata)
{
- ec_cbk_t callback = { .rmdir = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.rmdir = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(RMDIR) %p", frame);
+ gf_msg_trace("ec", 0, "EC(RMDIR) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_RMDIR, 0, target, minimum,
- ec_wind_rmdir, ec_manager_rmdir, callback,
- data);
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_RMDIR, 0, target, fop_flags,
+ ec_wind_rmdir, ec_manager_rmdir, callback, data);
if (fop == NULL) {
goto out;
}
@@ -1160,18 +1149,18 @@ void ec_rmdir(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
}
if (xdata != NULL) {
- fop->xdata = dict_copy_with_ref (xdata, NULL);
+ fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -1189,16 +1178,18 @@ out:
/* FOP: symlink */
-int32_t ec_symlink_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, inode_t * inode,
- struct iatt * buf, struct iatt * preparent,
- struct iatt * postparent, dict_t * xdata)
+int32_t
+ec_symlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- return ec_dir_write_cbk (frame, this, cookie, op_ret, op_errno,
- buf, preparent, postparent, NULL, NULL, xdata);
+ return ec_dir_write_cbk(frame, this, cookie, op_ret, op_errno, buf,
+ preparent, postparent, NULL, NULL, xdata);
}
-void ec_wind_symlink(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_symlink(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -1207,12 +1198,12 @@ void ec_wind_symlink(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->str[0], &fop->loc[0], fop->mode[0], fop->xdata);
}
-int32_t ec_manager_symlink(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_symlink(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_LOCK:
ec_lock_prepare_parent_inode(fop, &fop->loc[0], NULL,
@@ -1245,12 +1236,11 @@ int32_t ec_manager_symlink(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.symlink != NULL)
- {
- fop->cbks.symlink(fop->req_frame, fop, fop->xl, cbk->op_ret,
- cbk->op_errno, fop->loc[0].inode,
- &cbk->iatt[0], &cbk->iatt[1], &cbk->iatt[2],
- cbk->xdata);
+ if (fop->cbks.symlink != NULL) {
+ QUORUM_CBK(fop->cbks.symlink, fop, fop->req_frame, fop, fop->xl,
+ cbk->op_ret, cbk->op_errno, fop->loc[0].inode,
+ &cbk->iatt[0], &cbk->iatt[1], &cbk->iatt[2],
+ cbk->xdata);
}
return EC_STATE_LOCK_REUSE;
@@ -1262,8 +1252,7 @@ int32_t ec_manager_symlink(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.symlink != NULL)
- {
+ if (fop->cbks.symlink != NULL) {
fop->cbks.symlink(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL, NULL, NULL, NULL, NULL);
}
@@ -1283,32 +1272,31 @@ int32_t ec_manager_symlink(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_symlink(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_symlink_cbk_t func, void * data,
- const char * linkname, loc_t * loc, mode_t umask,
- dict_t * xdata)
+void
+ec_symlink(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_symlink_cbk_t func, void *data,
+ const char *linkname, loc_t *loc, mode_t umask, dict_t *xdata)
{
- ec_cbk_t callback = { .symlink = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.symlink = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(SYMLINK) %p", frame);
+ gf_msg_trace("ec", 0, "EC(SYMLINK) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_SYMLINK, 0, target, minimum,
- ec_wind_symlink, ec_manager_symlink, callback,
- data);
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_SYMLINK, 0, target,
+ fop_flags, ec_wind_symlink, ec_manager_symlink,
+ callback, data);
if (fop == NULL) {
goto out;
}
@@ -1318,16 +1306,16 @@ void ec_symlink(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (linkname != NULL) {
fop->str[0] = gf_strdup(linkname);
if (fop->str[0] == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY, "Failed to duplicate a string.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to duplicate a string.");
goto out;
}
}
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
@@ -1335,9 +1323,9 @@ void ec_symlink(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -1355,16 +1343,17 @@ out:
/* FOP: unlink */
-int32_t ec_unlink_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno,
- struct iatt * preparent, struct iatt * postparent,
- dict_t * xdata)
+int32_t
+ec_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- return ec_dir_write_cbk (frame, this, cookie, op_ret, op_errno, NULL,
- preparent, postparent, NULL, NULL, xdata);
+ return ec_dir_write_cbk(frame, this, cookie, op_ret, op_errno, NULL,
+ preparent, postparent, NULL, NULL, xdata);
}
-void ec_wind_unlink(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_unlink(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -1373,12 +1362,12 @@ void ec_wind_unlink(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
&fop->loc[0], fop->int32, fop->xdata);
}
-int32_t ec_manager_unlink(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_unlink(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_LOCK:
ec_lock_prepare_parent_inode(fop, &fop->loc[0], NULL,
@@ -1402,11 +1391,10 @@ int32_t ec_manager_unlink(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.unlink != NULL)
- {
- fop->cbks.unlink(fop->req_frame, fop, fop->xl, cbk->op_ret,
- cbk->op_errno, &cbk->iatt[0], &cbk->iatt[1],
- cbk->xdata);
+ if (fop->cbks.unlink != NULL) {
+ QUORUM_CBK(fop->cbks.unlink, fop, fop->req_frame, fop, fop->xl,
+ cbk->op_ret, cbk->op_errno, &cbk->iatt[0],
+ &cbk->iatt[1], cbk->xdata);
}
return EC_STATE_LOCK_REUSE;
@@ -1418,8 +1406,7 @@ int32_t ec_manager_unlink(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.unlink != NULL)
- {
+ if (fop->cbks.unlink != NULL) {
fop->cbks.unlink(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL, NULL, NULL);
}
@@ -1439,29 +1426,29 @@ int32_t ec_manager_unlink(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_unlink(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_unlink_cbk_t func, void * data,
- loc_t * loc, int xflags, dict_t * xdata)
+void
+ec_unlink(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_unlink_cbk_t func, void *data, loc_t *loc,
+ int xflags, dict_t *xdata)
{
- ec_cbk_t callback = { .unlink = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.unlink = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(UNLINK) %p", frame);
+ gf_msg_trace("ec", 0, "EC(UNLINK) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_UNLINK, 0, target, minimum,
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_UNLINK, 0, target, fop_flags,
ec_wind_unlink, ec_manager_unlink, callback,
data);
if (fop == NULL) {
@@ -1472,18 +1459,18 @@ void ec_unlink(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
}
if (xdata != NULL) {
- fop->xdata = dict_copy_with_ref (xdata, NULL);
+ fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
diff --git a/xlators/cluster/ec/src/ec-fops.h b/xlators/cluster/ec/src/ec-fops.h
index 4a926cf4802..07edf8a7fec 100644
--- a/xlators/cluster/ec/src/ec-fops.h
+++ b/xlators/cluster/ec/src/ec-fops.h
@@ -11,204 +11,244 @@
#ifndef __EC_FOPS_H__
#define __EC_FOPS_H__
-#include "xlator.h"
+#include <glusterfs/xlator.h>
#include "ec-types.h"
#include "ec-common.h"
-void ec_access(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_access_cbk_t func, void *data, loc_t * loc,
- int32_t mask, dict_t * xdata);
-
-void ec_create(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_create_cbk_t func, void *data, loc_t * loc,
- int32_t flags, mode_t mode, mode_t umask, fd_t * fd,
- dict_t * xdata);
-
-void ec_entrylk(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_entrylk_cbk_t func, void *data,
- const char * volume, loc_t * loc, const char * basename,
- entrylk_cmd cmd, entrylk_type type, dict_t * xdata);
-
-void ec_fentrylk(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fentrylk_cbk_t func, void *data,
- const char * volume, fd_t * fd, const char * basename,
- entrylk_cmd cmd, entrylk_type type, dict_t * xdata);
-
-void ec_flush(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_flush_cbk_t func, void *data, fd_t * fd,
- dict_t * xdata);
-
-void ec_fsync(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fsync_cbk_t func, void *data, fd_t * fd,
- int32_t datasync, dict_t * xdata);
-
-void ec_fsyncdir(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fsyncdir_cbk_t func, void *data,
- fd_t * fd, int32_t datasync, dict_t * xdata);
-
-void ec_getxattr(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_getxattr_cbk_t func, void *data,
- loc_t * loc, const char * name, dict_t * xdata);
-
-void ec_fgetxattr(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fgetxattr_cbk_t func, void *data,
- fd_t * fd, const char * name, dict_t * xdata);
-
-void ec_heal(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_heal_cbk_t func, void *data, loc_t * loc,
- int32_t partial, dict_t *xdata);
-
-void ec_fheal(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fheal_cbk_t func, void *data, fd_t * fd,
- int32_t partial, dict_t *xdata);
-
-void ec_inodelk (call_frame_t *frame, xlator_t *this, gf_lkowner_t *owner,
- uintptr_t target, int32_t minimum, fop_inodelk_cbk_t func,
- void *data, const char *volume, loc_t *loc, int32_t cmd,
- struct gf_flock * flock, dict_t * xdata);
-
-void ec_finodelk(call_frame_t *frame, xlator_t *this, gf_lkowner_t *owner,
- uintptr_t target, int32_t minimum, fop_finodelk_cbk_t func,
- void *data, const char *volume, fd_t *fd, int32_t cmd,
- struct gf_flock *flock, dict_t *xdata);
-
-void ec_link(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_link_cbk_t func, void *data, loc_t * oldloc,
- loc_t * newloc, dict_t * xdata);
-
-void ec_lk(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_lk_cbk_t func, void *data, fd_t * fd,
- int32_t cmd, struct gf_flock * flock, dict_t * xdata);
-
-void ec_lookup(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_lookup_cbk_t func, void *data, loc_t * loc,
- dict_t * xdata);
-
-void ec_mkdir(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_mkdir_cbk_t func, void *data, loc_t * loc,
- mode_t mode, mode_t umask, dict_t * xdata);
-
-void ec_mknod(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_mknod_cbk_t func, void *data, loc_t * loc,
- mode_t mode, dev_t rdev, mode_t umask, dict_t * xdata);
-
-void ec_open(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_open_cbk_t func, void *data, loc_t * loc,
- int32_t flags, fd_t * fd, dict_t * xdata);
-
-void ec_opendir(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_opendir_cbk_t func, void *data,
- loc_t * loc, fd_t * fd, dict_t * xdata);
-
-void ec_readdir(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_readdir_cbk_t func, void *data, fd_t * fd,
- size_t size, off_t offset, dict_t * xdata);
-
-void ec_readdirp(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_readdirp_cbk_t func, void *data,
- fd_t * fd, size_t size, off_t offset, dict_t * xdata);
-
-void ec_readlink(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_readlink_cbk_t func, void *data,
- loc_t * loc, size_t size, dict_t * xdata);
-
-void ec_readv(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_readv_cbk_t func, void *data, fd_t * fd,
- size_t size, off_t offset, uint32_t flags, dict_t * xdata);
-
-void ec_removexattr(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_removexattr_cbk_t func, void *data,
- loc_t * loc, const char * name, dict_t * xdata);
-
-void ec_fremovexattr(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fremovexattr_cbk_t func, void *data,
- fd_t * fd, const char * name, dict_t * xdata);
-
-void ec_rename(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_rename_cbk_t func, void *data,
- loc_t * oldloc, loc_t * newloc, dict_t * xdata);
-
-void ec_rmdir(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_rmdir_cbk_t func, void *data, loc_t * loc,
- int xflags, dict_t * xdata);
-
-void ec_setattr(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_setattr_cbk_t func, void *data,
- loc_t * loc, struct iatt * stbuf, int32_t valid,
- dict_t * xdata);
-
-void ec_fsetattr(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fsetattr_cbk_t func, void *data,
- fd_t * fd, struct iatt * stbuf, int32_t valid,
- dict_t * xdata);
-
-void ec_setxattr(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_setxattr_cbk_t func, void *data,
- loc_t * loc, dict_t * dict, int32_t flags, dict_t * xdata);
-
-void ec_fsetxattr(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fsetxattr_cbk_t func, void *data,
- fd_t * fd, dict_t * dict, int32_t flags, dict_t * xdata);
-
-void ec_stat(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_stat_cbk_t func, void *data, loc_t * loc,
- dict_t * xdata);
-
-void ec_fstat(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fstat_cbk_t func, void *data, fd_t * fd,
- dict_t * xdata);
-
-void ec_statfs(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_statfs_cbk_t func, void *data, loc_t * loc,
- dict_t * xdata);
-
-void ec_symlink(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_symlink_cbk_t func, void *data,
- const char * linkname, loc_t * loc, mode_t umask,
- dict_t * xdata);
-
-void ec_fallocate(call_frame_t *frame, xlator_t *this, uintptr_t target,
- int32_t minimum, fop_fallocate_cbk_t func, void *data, fd_t *fd,
- int32_t mode, off_t offset, size_t len, dict_t *xdata);
-
-void ec_discard(call_frame_t *frame, xlator_t *this, uintptr_t target,
- int32_t minimum, fop_discard_cbk_t func, void *data, fd_t *fd,
- off_t offset, size_t len, dict_t *xdata);
-
-void ec_truncate(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_truncate_cbk_t func, void *data,
- loc_t * loc, off_t offset, dict_t * xdata);
-
-void ec_ftruncate(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_ftruncate_cbk_t func, void *data,
- fd_t * fd, off_t offset, dict_t * xdata);
-
-void ec_unlink(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_unlink_cbk_t func, void *data, loc_t * loc,
- int xflags, dict_t * xdata);
-
-void ec_writev(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_writev_cbk_t func, void *data, fd_t * fd,
- struct iovec * vector, int32_t count, off_t offset,
- uint32_t flags, struct iobref * iobref, dict_t * xdata);
-
-void ec_xattrop(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_xattrop_cbk_t func, void *data,
- loc_t * loc, gf_xattrop_flags_t optype, dict_t * xattr,
- dict_t * xdata);
-
-void ec_fxattrop(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fxattrop_cbk_t func, void *data,
- fd_t * fd, gf_xattrop_flags_t optype, dict_t * xattr,
- dict_t * xdata);
-
-void ec_seek(call_frame_t *frame, xlator_t *this, uintptr_t target,
- int32_t minimum, fop_seek_cbk_t func, void *data, fd_t *fd,
- off_t offset, gf_seek_what_t what, dict_t *xdata);
-
-void ec_ipc(call_frame_t *frame, xlator_t *this, uintptr_t target,
- int32_t minimum, fop_ipc_cbk_t func, void *data, int32_t op,
- dict_t *xdata);
+void
+ec_access(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_access_cbk_t func, void *data, loc_t *loc,
+ int32_t mask, dict_t *xdata);
+
+void
+ec_create(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_create_cbk_t func, void *data, loc_t *loc,
+ int32_t flags, mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata);
+
+void
+ec_entrylk(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_entrylk_cbk_t func, void *data,
+ const char *volume, loc_t *loc, const char *basename,
+ entrylk_cmd cmd, entrylk_type type, dict_t *xdata);
+
+void
+ec_fentrylk(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fentrylk_cbk_t func, void *data,
+ const char *volume, fd_t *fd, const char *basename, entrylk_cmd cmd,
+ entrylk_type type, dict_t *xdata);
+
+void
+ec_flush(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_flush_cbk_t func, void *data, fd_t *fd,
+ dict_t *xdata);
+
+void
+ec_fsync(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fsync_cbk_t func, void *data, fd_t *fd,
+ int32_t datasync, dict_t *xdata);
+
+void
+ec_fsyncdir(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fsyncdir_cbk_t func, void *data, fd_t *fd,
+ int32_t datasync, dict_t *xdata);
+
+void
+ec_getxattr(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_getxattr_cbk_t func, void *data, loc_t *loc,
+ const char *name, dict_t *xdata);
+
+void
+ec_fgetxattr(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fgetxattr_cbk_t func, void *data, fd_t *fd,
+ const char *name, dict_t *xdata);
+
+void
+ec_heal(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_heal_cbk_t func, void *data, loc_t *loc,
+ int32_t partial, dict_t *xdata);
+
+void
+ec_fheal(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fheal_cbk_t func, void *data, fd_t *fd,
+ int32_t partial, dict_t *xdata);
+
+void
+ec_inodelk(call_frame_t *frame, xlator_t *this, gf_lkowner_t *owner,
+ uintptr_t target, uint32_t fop_flags, fop_inodelk_cbk_t func,
+ void *data, const char *volume, loc_t *loc, int32_t cmd,
+ struct gf_flock *flock, dict_t *xdata);
+
+void
+ec_finodelk(call_frame_t *frame, xlator_t *this, gf_lkowner_t *owner,
+ uintptr_t target, uint32_t fop_flags, fop_finodelk_cbk_t func,
+ void *data, const char *volume, fd_t *fd, int32_t cmd,
+ struct gf_flock *flock, dict_t *xdata);
+
+void
+ec_link(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_link_cbk_t func, void *data, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata);
+
+void
+ec_lk(call_frame_t *frame, xlator_t *this, uintptr_t target, uint32_t fop_flags,
+ fop_lk_cbk_t func, void *data, fd_t *fd, int32_t cmd,
+ struct gf_flock *flock, dict_t *xdata);
+
+void
+ec_lookup(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_lookup_cbk_t func, void *data, loc_t *loc,
+ dict_t *xdata);
+
+void
+ec_mkdir(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_mkdir_cbk_t func, void *data, loc_t *loc,
+ mode_t mode, mode_t umask, dict_t *xdata);
+
+void
+ec_mknod(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_mknod_cbk_t func, void *data, loc_t *loc,
+ mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata);
+
+void
+ec_open(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_open_cbk_t func, void *data, loc_t *loc,
+ int32_t flags, fd_t *fd, dict_t *xdata);
+
+void
+ec_opendir(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_opendir_cbk_t func, void *data, loc_t *loc,
+ fd_t *fd, dict_t *xdata);
+
+void
+ec_readdir(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_readdir_cbk_t func, void *data, fd_t *fd,
+ size_t size, off_t offset, dict_t *xdata);
+
+void
+ec_readdirp(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_readdirp_cbk_t func, void *data, fd_t *fd,
+ size_t size, off_t offset, dict_t *xdata);
+
+void
+ec_readlink(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_readlink_cbk_t func, void *data, loc_t *loc,
+ size_t size, dict_t *xdata);
+
+void
+ec_readv(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_readv_cbk_t func, void *data, fd_t *fd,
+ size_t size, off_t offset, uint32_t flags, dict_t *xdata);
+
+void
+ec_removexattr(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_removexattr_cbk_t func, void *data,
+ loc_t *loc, const char *name, dict_t *xdata);
+
+void
+ec_fremovexattr(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fremovexattr_cbk_t func, void *data,
+ fd_t *fd, const char *name, dict_t *xdata);
+
+void
+ec_rename(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_rename_cbk_t func, void *data, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata);
+
+void
+ec_rmdir(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_rmdir_cbk_t func, void *data, loc_t *loc,
+ int xflags, dict_t *xdata);
+
+void
+ec_setattr(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_setattr_cbk_t func, void *data, loc_t *loc,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata);
+
+void
+ec_fsetattr(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fsetattr_cbk_t func, void *data, fd_t *fd,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata);
+
+void
+ec_setxattr(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_setxattr_cbk_t func, void *data, loc_t *loc,
+ dict_t *dict, int32_t flags, dict_t *xdata);
+
+void
+ec_fsetxattr(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fsetxattr_cbk_t func, void *data, fd_t *fd,
+ dict_t *dict, int32_t flags, dict_t *xdata);
+
+void
+ec_stat(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_stat_cbk_t func, void *data, loc_t *loc,
+ dict_t *xdata);
+
+void
+ec_fstat(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fstat_cbk_t func, void *data, fd_t *fd,
+ dict_t *xdata);
+
+void
+ec_statfs(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_statfs_cbk_t func, void *data, loc_t *loc,
+ dict_t *xdata);
+
+void
+ec_symlink(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_symlink_cbk_t func, void *data,
+ const char *linkname, loc_t *loc, mode_t umask, dict_t *xdata);
+
+void
+ec_fallocate(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fallocate_cbk_t func, void *data, fd_t *fd,
+ int32_t mode, off_t offset, size_t len, dict_t *xdata);
+
+void
+ec_discard(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_discard_cbk_t func, void *data, fd_t *fd,
+ off_t offset, size_t len, dict_t *xdata);
+
+void
+ec_truncate(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_truncate_cbk_t func, void *data, loc_t *loc,
+ off_t offset, dict_t *xdata);
+
+void
+ec_ftruncate(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_ftruncate_cbk_t func, void *data, fd_t *fd,
+ off_t offset, dict_t *xdata);
+
+void
+ec_unlink(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_unlink_cbk_t func, void *data, loc_t *loc,
+ int xflags, dict_t *xdata);
+
+void
+ec_writev(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_writev_cbk_t func, void *data, fd_t *fd,
+ struct iovec *vector, int32_t count, off_t offset, uint32_t flags,
+ struct iobref *iobref, dict_t *xdata);
+
+void
+ec_xattrop(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_xattrop_cbk_t func, void *data, loc_t *loc,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata);
+
+void
+ec_fxattrop(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fxattrop_cbk_t func, void *data, fd_t *fd,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata);
+
+void
+ec_seek(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_seek_cbk_t func, void *data, fd_t *fd,
+ off_t offset, gf_seek_what_t what, dict_t *xdata);
+
+void
+ec_ipc(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_ipc_cbk_t func, void *data, int32_t op,
+ dict_t *xdata);
#endif /* __EC_FOPS_H__ */
diff --git a/xlators/cluster/ec/src/ec-galois.c b/xlators/cluster/ec/src/ec-galois.c
index 8cb4dc2e4e3..6e4990c71f5 100644
--- a/xlators/cluster/ec/src/ec-galois.c
+++ b/xlators/cluster/ec/src/ec-galois.c
@@ -10,9 +10,6 @@
#include <string.h>
-#include "mem-pool.h"
-#include "list.h"
-
#include "ec-mem-types.h"
#include "ec-gf8.h"
#include "ec-helpers.h"
diff --git a/xlators/cluster/ec/src/ec-galois.h b/xlators/cluster/ec/src/ec-galois.h
index 02e6b6c1bc2..ed55d53e419 100644
--- a/xlators/cluster/ec/src/ec-galois.h
+++ b/xlators/cluster/ec/src/ec-galois.h
@@ -15,12 +15,18 @@
#include "ec-types.h"
-ec_gf_t *ec_gf_prepare(uint32_t bits, uint32_t mod);
-void ec_gf_destroy(ec_gf_t *gf);
+ec_gf_t *
+ec_gf_prepare(uint32_t bits, uint32_t mod);
+void
+ec_gf_destroy(ec_gf_t *gf);
-uint32_t ec_gf_add(ec_gf_t *gf, uint32_t a, uint32_t b);
-uint32_t ec_gf_mul(ec_gf_t *gf, uint32_t a, uint32_t b);
-uint32_t ec_gf_div(ec_gf_t *gf, uint32_t a, uint32_t b);
-uint32_t ec_gf_exp(ec_gf_t *gf, uint32_t a, uint32_t b);
+uint32_t
+ec_gf_add(ec_gf_t *gf, uint32_t a, uint32_t b);
+uint32_t
+ec_gf_mul(ec_gf_t *gf, uint32_t a, uint32_t b);
+uint32_t
+ec_gf_div(ec_gf_t *gf, uint32_t a, uint32_t b);
+uint32_t
+ec_gf_exp(ec_gf_t *gf, uint32_t a, uint32_t b);
#endif /* __EC_GALOIS_H__ */
diff --git a/xlators/cluster/ec/src/ec-generic.c b/xlators/cluster/ec/src/ec-generic.c
index a5f986e74f4..884deb93669 100644
--- a/xlators/cluster/ec/src/ec-generic.c
+++ b/xlators/cluster/ec/src/ec-generic.c
@@ -8,25 +8,23 @@
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "defaults.h"
-#include "byte-order.h"
+#include <glusterfs/byte-order.h>
#include "ec.h"
#include "ec-messages.h"
#include "ec-helpers.h"
#include "ec-common.h"
#include "ec-combine.h"
-#include "ec-method.h"
#include "ec-fops.h"
/* FOP: flush */
-int32_t ec_flush_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, dict_t * xdata)
+int32_t
+ec_flush_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -36,21 +34,18 @@ int32_t ec_flush_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_FLUSH, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (xdata != NULL)
- {
+ if (cbk != NULL) {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -60,15 +55,15 @@ int32_t ec_flush_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_flush(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_flush(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -77,15 +72,15 @@ void ec_wind_flush(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->xdata);
}
-int32_t ec_manager_flush(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_flush(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_LOCK:
- ec_lock_prepare_fd(fop, fop->fd, 0, 0, LLONG_MAX);
+ ec_lock_prepare_fd(fop, fop->fd, 0, 0, EC_RANGE_FULL);
ec_lock(fop);
return EC_STATE_DISPATCH;
@@ -110,8 +105,7 @@ int32_t ec_manager_flush(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.flush != NULL)
- {
+ if (fop->cbks.flush != NULL) {
fop->cbks.flush(fop->req_frame, fop, fop->xl, cbk->op_ret,
cbk->op_errno, cbk->xdata);
}
@@ -126,8 +120,7 @@ int32_t ec_manager_flush(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.flush != NULL)
- {
+ if (fop->cbks.flush != NULL) {
fop->cbks.flush(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL);
}
@@ -147,31 +140,71 @@ int32_t ec_manager_flush(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_flush(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_flush_cbk_t func, void * data, fd_t * fd,
- dict_t * xdata)
+static int32_t
+ec_validate_fd(fd_t *fd, xlator_t *xl)
{
- ec_cbk_t callback = { .flush = func };
- ec_fop_data_t * fop = NULL;
+ uint64_t iversion = 0;
+ uint64_t fversion = 0;
+ ec_inode_t *inode_ctx = NULL;
+ ec_fd_t *fd_ctx = NULL;
+
+ LOCK(&fd->lock);
+ {
+ fd_ctx = __ec_fd_get(fd, xl);
+ if (fd_ctx) {
+ fversion = fd_ctx->bad_version;
+ }
+ }
+ UNLOCK(&fd->lock);
+
+ LOCK(&fd->inode->lock);
+ {
+ inode_ctx = __ec_inode_get(fd->inode, xl);
+ if (inode_ctx) {
+ iversion = inode_ctx->bad_version;
+ }
+ }
+ UNLOCK(&fd->inode->lock);
+ if (fversion < iversion) {
+ return EBADF;
+ }
+ return 0;
+}
+
+void
+ec_flush(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_flush_cbk_t func, void *data, fd_t *fd,
+ dict_t *xdata)
+{
+ ec_cbk_t callback = {.flush = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(FLUSH) %p", frame);
+ gf_msg_trace("ec", 0, "EC(FLUSH) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_FLUSH, 0, target, minimum,
- ec_wind_flush, ec_manager_flush, callback,
- data);
+ if (fd) {
+ error = ec_validate_fd(fd, this);
+ if (error) {
+ gf_msg(this->name, GF_LOG_ERROR, EBADF, EC_MSG_FD_BAD,
+ "Failing %s on %s", gf_fop_list[GF_FOP_FLUSH],
+ fd->inode ? uuid_utoa(fd->inode->gfid) : "");
+ goto out;
+ }
+ }
+
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_FLUSH, 0, target, fop_flags,
+ ec_wind_flush, ec_manager_flush, callback, data);
if (fop == NULL) {
goto out;
}
@@ -181,9 +214,9 @@ void ec_flush(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL, "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -191,9 +224,9 @@ void ec_flush(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -211,13 +244,13 @@ out:
/* FOP: fsync */
-int32_t ec_combine_fsync(ec_fop_data_t * fop, ec_cbk_data_t * dst,
- ec_cbk_data_t * src)
+int32_t
+ec_combine_fsync(ec_fop_data_t *fop, ec_cbk_data_t *dst, ec_cbk_data_t *src)
{
if (!ec_iatt_combine(fop, dst->iatt, src->iatt, 2)) {
- gf_msg (fop->xl->name, GF_LOG_NOTICE, 0,
- EC_MSG_IATT_MISMATCH, "Mismatching iatt in "
- "answers of 'GF_FOP_FSYNC'");
+ gf_msg(fop->xl->name, GF_LOG_NOTICE, 0, EC_MSG_IATT_MISMATCH,
+ "Mismatching iatt in "
+ "answers of 'GF_FOP_FSYNC'");
return 0;
}
@@ -225,12 +258,13 @@ int32_t ec_combine_fsync(ec_fop_data_t * fop, ec_cbk_data_t * dst,
return 1;
}
-int32_t ec_fsync_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, struct iatt * prebuf,
- struct iatt * postbuf, dict_t * xdata)
+int32_t
+ec_fsync_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -240,32 +274,26 @@ int32_t ec_fsync_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_FSYNC, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (op_ret >= 0)
- {
- if (prebuf != NULL)
- {
+ if (cbk != NULL) {
+ if (op_ret >= 0) {
+ if (prebuf != NULL) {
cbk->iatt[0] = *prebuf;
}
- if (postbuf != NULL)
- {
+ if (postbuf != NULL) {
cbk->iatt[1] = *postbuf;
}
}
- if (xdata != NULL)
- {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -275,15 +303,15 @@ int32_t ec_fsync_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_fsync(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_fsync(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -292,15 +320,15 @@ void ec_wind_fsync(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->int32, fop->xdata);
}
-int32_t ec_manager_fsync(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_fsync(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_LOCK:
- ec_lock_prepare_fd(fop, fop->fd, EC_QUERY_INFO, 0, LLONG_MAX);
+ ec_lock_prepare_fd(fop, fop->fd, EC_QUERY_INFO, 0, EC_RANGE_FULL);
ec_lock(fop);
return EC_STATE_DISPATCH;
@@ -318,8 +346,7 @@ int32_t ec_manager_fsync(ec_fop_data_t * fop, int32_t state)
case EC_STATE_PREPARE_ANSWER:
cbk = ec_fop_prepare_answer(fop, _gf_false);
if (cbk != NULL) {
- ec_iatt_rebuild(fop->xl->private, cbk->iatt, 2,
- cbk->count);
+ ec_iatt_rebuild(fop->xl->private, cbk->iatt, 2, cbk->count);
/* This shouldn't fail because we have the inode locked. */
GF_ASSERT(ec_get_inode_size(fop, fop->fd->inode,
@@ -334,8 +361,7 @@ int32_t ec_manager_fsync(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.fsync != NULL)
- {
+ if (fop->cbks.fsync != NULL) {
fop->cbks.fsync(fop->req_frame, fop, fop->xl, cbk->op_ret,
cbk->op_errno, &cbk->iatt[0], &cbk->iatt[1],
cbk->xdata);
@@ -351,8 +377,7 @@ int32_t ec_manager_fsync(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_DELAYED_START:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.fsync != NULL)
- {
+ if (fop->cbks.fsync != NULL) {
fop->cbks.fsync(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL, NULL, NULL);
}
@@ -372,31 +397,40 @@ int32_t ec_manager_fsync(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_fsync(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fsync_cbk_t func, void * data, fd_t * fd,
- int32_t datasync, dict_t * xdata)
+void
+ec_fsync(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fsync_cbk_t func, void *data, fd_t *fd,
+ int32_t datasync, dict_t *xdata)
{
- ec_cbk_t callback = { .fsync = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.fsync = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(FSYNC) %p", frame);
+ gf_msg_trace("ec", 0, "EC(FSYNC) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_FSYNC, 0, target, minimum,
- ec_wind_fsync, ec_manager_fsync, callback,
- data);
+ if (fd) {
+ error = ec_validate_fd(fd, this);
+ if (error) {
+ gf_msg(this->name, GF_LOG_ERROR, EBADF, EC_MSG_FD_BAD,
+ "Failing %s on %s", gf_fop_list[GF_FOP_FSYNC],
+ fd->inode ? uuid_utoa(fd->inode->gfid) : "");
+ goto out;
+ }
+ }
+
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_FSYNC, 0, target, fop_flags,
+ ec_wind_fsync, ec_manager_fsync, callback, data);
if (fop == NULL) {
goto out;
}
@@ -408,9 +442,9 @@ void ec_fsync(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL, "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -418,9 +452,9 @@ void ec_fsync(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -438,11 +472,12 @@ out:
/* FOP: fsyncdir */
-int32_t ec_fsyncdir_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, dict_t * xdata)
+int32_t
+ec_fsyncdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -452,21 +487,18 @@ int32_t ec_fsyncdir_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_FSYNCDIR, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (xdata != NULL)
- {
+ if (cbk != NULL) {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -476,15 +508,15 @@ int32_t ec_fsyncdir_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_fsyncdir(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_fsyncdir(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -493,15 +525,15 @@ void ec_wind_fsyncdir(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->fd, fop->int32, fop->xdata);
}
-int32_t ec_manager_fsyncdir(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_fsyncdir(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_LOCK:
- ec_lock_prepare_fd(fop, fop->fd, 0, 0, LLONG_MAX);
+ ec_lock_prepare_fd(fop, fop->fd, 0, 0, EC_RANGE_FULL);
ec_lock(fop);
return EC_STATE_DISPATCH;
@@ -526,8 +558,7 @@ int32_t ec_manager_fsyncdir(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.fsyncdir != NULL)
- {
+ if (fop->cbks.fsyncdir != NULL) {
fop->cbks.fsyncdir(fop->req_frame, fop, fop->xl, cbk->op_ret,
cbk->op_errno, cbk->xdata);
}
@@ -542,10 +573,9 @@ int32_t ec_manager_fsyncdir(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_DELAYED_START:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.fsyncdir != NULL)
- {
- fop->cbks.fsyncdir(fop->req_frame, fop, fop->xl, -1,
- fop->error, NULL);
+ if (fop->cbks.fsyncdir != NULL) {
+ fop->cbks.fsyncdir(fop->req_frame, fop, fop->xl, -1, fop->error,
+ NULL);
}
return EC_STATE_LOCK_REUSE;
@@ -563,30 +593,30 @@ int32_t ec_manager_fsyncdir(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_fsyncdir(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fsyncdir_cbk_t func, void * data,
- fd_t * fd, int32_t datasync, dict_t * xdata)
+void
+ec_fsyncdir(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fsyncdir_cbk_t func, void *data, fd_t *fd,
+ int32_t datasync, dict_t *xdata)
{
- ec_cbk_t callback = { .fsyncdir = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.fsyncdir = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(FSYNCDIR) %p", frame);
+ gf_msg_trace("ec", 0, "EC(FSYNCDIR) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_FSYNCDIR, 0, target,
- minimum, ec_wind_fsyncdir, ec_manager_fsyncdir,
+ fop_flags, ec_wind_fsyncdir, ec_manager_fsyncdir,
callback, data);
if (fop == NULL) {
goto out;
@@ -599,9 +629,9 @@ void ec_fsyncdir(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL, "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -609,9 +639,9 @@ void ec_fsyncdir(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -629,9 +659,10 @@ out:
/* FOP: lookup */
-void ec_lookup_rebuild(ec_t * ec, ec_fop_data_t * fop, ec_cbk_data_t * cbk)
+void
+ec_lookup_rebuild(ec_t *ec, ec_fop_data_t *fop, ec_cbk_data_t *cbk)
{
- ec_inode_t * ctx = NULL;
+ ec_inode_t *ctx = NULL;
uint64_t size = 0;
int32_t have_size = 0, err;
@@ -650,8 +681,7 @@ void ec_lookup_rebuild(ec_t * ec, ec_fop_data_t * fop, ec_cbk_data_t * cbk)
LOCK(&cbk->inode->lock);
ctx = __ec_inode_get(cbk->inode, fop->xl);
- if (ctx != NULL)
- {
+ if (ctx != NULL) {
if (ctx->have_version) {
cbk->version[0] = ctx->post_version[0];
cbk->version[1] = ctx->post_version[1];
@@ -664,24 +694,22 @@ void ec_lookup_rebuild(ec_t * ec, ec_fop_data_t * fop, ec_cbk_data_t * cbk)
UNLOCK(&cbk->inode->lock);
- if (cbk->iatt[0].ia_type == IA_IFREG)
- {
+ if (cbk->iatt[0].ia_type == IA_IFREG) {
cbk->size = cbk->iatt[0].ia_size;
ec_dict_del_number(cbk->xdata, EC_XATTR_SIZE, &cbk->iatt[0].ia_size);
- if (have_size)
- {
+ if (have_size) {
cbk->iatt[0].ia_size = size;
}
}
}
-int32_t ec_combine_lookup(ec_fop_data_t * fop, ec_cbk_data_t * dst,
- ec_cbk_data_t * src)
+int32_t
+ec_combine_lookup(ec_fop_data_t *fop, ec_cbk_data_t *dst, ec_cbk_data_t *src)
{
if (!ec_iatt_combine(fop, dst->iatt, src->iatt, 2)) {
- gf_msg (fop->xl->name, GF_LOG_NOTICE, 0,
- EC_MSG_IATT_MISMATCH, "Mismatching iatt in "
- "answers of 'GF_FOP_LOOKUP'");
+ gf_msg(fop->xl->name, GF_LOG_DEBUG, 0, EC_MSG_IATT_MISMATCH,
+ "Mismatching iatt in "
+ "answers of 'GF_FOP_LOOKUP'");
return 0;
}
@@ -689,15 +717,15 @@ int32_t ec_combine_lookup(ec_fop_data_t * fop, ec_cbk_data_t * dst,
return 1;
}
-int32_t ec_lookup_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, inode_t * inode,
- struct iatt * buf, dict_t * xdata,
- struct iatt * postparent)
+int32_t
+ec_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode, struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
- uint64_t dirty[2] = {0};
+ uint64_t dirty[2] = {0};
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
@@ -706,63 +734,54 @@ int32_t ec_lookup_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_LOOKUP, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (op_ret >= 0)
- {
- if (inode != NULL)
- {
+ if (cbk != NULL) {
+ if (op_ret >= 0) {
+ if (inode != NULL) {
cbk->inode = inode_ref(inode);
- if (cbk->inode == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_INODE_REF_FAIL,
- "Failed to reference an inode.");
+ if (cbk->inode == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_INODE_REF_FAIL,
+ "Failed to reference an inode.");
goto out;
}
}
- if (buf != NULL)
- {
+ if (buf != NULL) {
cbk->iatt[0] = *buf;
}
- if (postparent != NULL)
- {
+ if (postparent != NULL) {
cbk->iatt[1] = *postparent;
}
}
- if (xdata != NULL)
- {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
- ec_dict_del_array (xdata, EC_XATTR_DIRTY, dirty, EC_VERSION_SIZE);
+ ec_dict_del_array(xdata, EC_XATTR_DIRTY, dirty, EC_VERSION_SIZE);
}
ec_combine(cbk, ec_combine_lookup);
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_lookup(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_lookup(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -771,20 +790,21 @@ void ec_wind_lookup(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
&fop->loc[0], fop->xdata);
}
-int32_t ec_manager_lookup(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_lookup(ec_fop_data_t *fop, int32_t state)
{
ec_cbk_data_t *cbk;
int32_t err;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
if (fop->xdata == NULL) {
fop->xdata = dict_new();
if (fop->xdata == NULL) {
- gf_msg (fop->xl->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOOKUP_REQ_PREP_FAIL, "Unable to prepare "
- "lookup request");
+ gf_msg(fop->xl->name, GF_LOG_ERROR, ENOMEM,
+ EC_MSG_LOOKUP_REQ_PREP_FAIL,
+ "Unable to prepare "
+ "lookup request");
fop->error = ENOMEM;
@@ -792,7 +812,7 @@ int32_t ec_manager_lookup(ec_fop_data_t * fop, int32_t state)
}
} else {
/*TODO: To be handled once we have 'syndromes' */
- dict_del (fop->xdata, GF_CONTENT_KEY);
+ dict_del(fop->xdata, GF_CONTENT_KEY);
}
err = dict_set_uint64(fop->xdata, EC_XATTR_SIZE, 0);
if (err == 0) {
@@ -802,16 +822,17 @@ int32_t ec_manager_lookup(ec_fop_data_t * fop, int32_t state)
err = dict_set_uint64(fop->xdata, EC_XATTR_DIRTY, 0);
}
if (err != 0) {
- gf_msg (fop->xl->name, GF_LOG_ERROR, -err,
- EC_MSG_LOOKUP_REQ_PREP_FAIL, "Unable to prepare lookup "
- "request");
+ gf_msg(fop->xl->name, GF_LOG_ERROR, -err,
+ EC_MSG_LOOKUP_REQ_PREP_FAIL,
+ "Unable to prepare lookup "
+ "request");
fop->error = -err;
return EC_STATE_REPORT;
}
- /* Fall through */
+ /* Fall through */
case EC_STATE_DISPATCH:
ec_dispatch_all(fop);
@@ -827,8 +848,8 @@ int32_t ec_manager_lookup(ec_fop_data_t * fop, int32_t state)
*/
if (!fop->answer && !list_empty(&fop->cbk_list)) {
- fop->answer = list_entry (fop->cbk_list.next, ec_cbk_data_t,
- list);
+ fop->answer = list_entry(fop->cbk_list.next, ec_cbk_data_t,
+ list);
}
cbk = ec_fop_prepare_answer(fop, _gf_true);
@@ -845,8 +866,7 @@ int32_t ec_manager_lookup(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.lookup != NULL)
- {
+ if (fop->cbks.lookup != NULL) {
fop->cbks.lookup(fop->req_frame, fop, fop->xl, cbk->op_ret,
cbk->op_errno, cbk->inode, &cbk->iatt[0],
cbk->xdata, &cbk->iatt[1]);
@@ -860,8 +880,7 @@ int32_t ec_manager_lookup(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.lookup != NULL)
- {
+ if (fop->cbks.lookup != NULL) {
fop->cbks.lookup(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL, NULL, NULL, NULL);
}
@@ -869,30 +888,30 @@ int32_t ec_manager_lookup(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_lookup(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_lookup_cbk_t func, void * data,
- loc_t * loc, dict_t * xdata)
+void
+ec_lookup(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_lookup_cbk_t func, void *data, loc_t *loc,
+ dict_t *xdata)
{
- ec_cbk_t callback = { .lookup = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.lookup = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(LOOKUP) %p", frame);
+ gf_msg_trace("ec", 0, "EC(LOOKUP) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_LOOKUP, EC_FLAG_LOCK_SHARED,
- target, minimum, ec_wind_lookup,
+ target, fop_flags, ec_wind_lookup,
ec_manager_lookup, callback, data);
if (fop == NULL) {
goto out;
@@ -900,14 +919,14 @@ void ec_lookup(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
}
if (xdata != NULL) {
- fop->xdata = dict_copy_with_ref (xdata, NULL);
+ fop->xdata = dict_copy_with_ref(xdata, NULL);
/* Do not log failures here as a memory problem would have already
* been logged by the corresponding alloc functions */
if (fop->xdata == NULL)
@@ -926,20 +945,20 @@ out:
/* FOP: statfs */
-int32_t ec_combine_statfs(ec_fop_data_t * fop, ec_cbk_data_t * dst,
- ec_cbk_data_t * src)
+int32_t
+ec_combine_statfs(ec_fop_data_t *fop, ec_cbk_data_t *dst, ec_cbk_data_t *src)
{
ec_statvfs_combine(&dst->statvfs, &src->statvfs);
return 1;
}
-int32_t ec_statfs_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, struct statvfs * buf,
- dict_t * xdata)
+int32_t
+ec_statfs_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct statvfs *buf, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -949,28 +968,23 @@ int32_t ec_statfs_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_STATFS, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (op_ret >= 0)
- {
- if (buf != NULL)
- {
+ if (cbk != NULL) {
+ if (op_ret >= 0) {
+ if (buf != NULL) {
cbk->statvfs = *buf;
}
}
- if (xdata != NULL)
- {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -980,15 +994,15 @@ int32_t ec_statfs_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_statfs(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_statfs(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -997,14 +1011,14 @@ void ec_wind_statfs(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
&fop->loc[0], fop->xdata);
}
-int32_t ec_manager_statfs(ec_fop_data_t *fop, int32_t state)
+int32_t
+ec_manager_statfs(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t *cbk = NULL;
- gf_boolean_t deem_statfs_enabled = _gf_false;
- int32_t err = 0;
+ ec_cbk_data_t *cbk = NULL;
+ gf_boolean_t deem_statfs_enabled = _gf_false;
+ int32_t err = 0;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_DISPATCH:
ec_dispatch_all(fop);
@@ -1017,8 +1031,8 @@ int32_t ec_manager_statfs(ec_fop_data_t *fop, int32_t state)
ec_t *ec = fop->xl->private;
if (cbk->xdata) {
- err = dict_get_int8 (cbk->xdata, "quota-deem-statfs",
- (int8_t *)&deem_statfs_enabled);
+ err = dict_get_int8(cbk->xdata, "quota-deem-statfs",
+ (int8_t *)&deem_statfs_enabled);
if (err != -ENOENT) {
ec_cbk_set_error(cbk, -err, _gf_true);
}
@@ -1038,8 +1052,7 @@ int32_t ec_manager_statfs(ec_fop_data_t *fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.statfs != NULL)
- {
+ if (fop->cbks.statfs != NULL) {
fop->cbks.statfs(fop->req_frame, fop, fop->xl, cbk->op_ret,
cbk->op_errno, &cbk->statvfs, cbk->xdata);
}
@@ -1052,8 +1065,7 @@ int32_t ec_manager_statfs(ec_fop_data_t *fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.statfs != NULL)
- {
+ if (fop->cbks.statfs != NULL) {
fop->cbks.statfs(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL, NULL);
}
@@ -1061,30 +1073,30 @@ int32_t ec_manager_statfs(ec_fop_data_t *fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_statfs(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_statfs_cbk_t func, void * data,
- loc_t * loc, dict_t * xdata)
+void
+ec_statfs(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_statfs_cbk_t func, void *data, loc_t *loc,
+ dict_t *xdata)
{
- ec_cbk_t callback = { .statfs = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.statfs = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(STATFS) %p", frame);
+ gf_msg_trace("ec", 0, "EC(STATFS) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_STATFS, EC_FLAG_LOCK_SHARED,
- target, minimum, ec_wind_statfs,
+ target, fop_flags, ec_wind_statfs,
ec_manager_statfs, callback, data);
if (fop == NULL) {
goto out;
@@ -1092,8 +1104,8 @@ void ec_statfs(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
@@ -1101,9 +1113,9 @@ void ec_statfs(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -1121,14 +1133,13 @@ out:
/* FOP: xattrop */
-int32_t ec_combine_xattrop(ec_fop_data_t *fop, ec_cbk_data_t *dst,
- ec_cbk_data_t *src)
+int32_t
+ec_combine_xattrop(ec_fop_data_t *fop, ec_cbk_data_t *dst, ec_cbk_data_t *src)
{
- if (!ec_dict_compare(dst->dict, src->dict))
- {
- gf_msg (fop->xl->name, GF_LOG_DEBUG, 0,
- EC_MSG_DICT_MISMATCH, "Mismatching dictionary in "
- "answers of 'GF_FOP_XATTROP'");
+ if (!ec_dict_compare(dst->dict, src->dict)) {
+ gf_msg(fop->xl->name, GF_LOG_DEBUG, 0, EC_MSG_DICT_MISMATCH,
+ "Mismatching dictionary in "
+ "answers of 'GF_FOP_XATTROP'");
return 0;
}
@@ -1137,72 +1148,71 @@ int32_t ec_combine_xattrop(ec_fop_data_t *fop, ec_cbk_data_t *dst,
}
int32_t
-ec_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xattr,
- dict_t *xdata)
+ec_xattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xattr, dict_t *xdata)
{
- ec_fop_data_t *fop = NULL;
- ec_lock_link_t *link = NULL;
- ec_cbk_data_t *cbk = NULL;
- uint64_t dirty[2] = {0};
- data_t *data;
- uint64_t *version;
- int32_t idx = (int32_t)(uintptr_t)cookie;
-
- VALIDATE_OR_GOTO (this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, frame->local, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
-
- fop = frame->local;
-
- ec_trace ("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
-
- cbk = ec_cbk_data_allocate (frame, this, fop, fop->id, idx, op_ret,
- op_errno);
- if (!cbk)
- goto out;
+ ec_fop_data_t *fop = NULL;
+ ec_lock_link_t *link = NULL;
+ ec_cbk_data_t *cbk = NULL;
+ uint64_t dirty[2] = {0};
+ data_t *data;
+ uint64_t *version;
+ int32_t idx = (int32_t)(uintptr_t)cookie;
- if (op_ret >= 0) {
- cbk->dict = dict_ref (xattr);
+ VALIDATE_OR_GOTO(this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame->local, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- data = dict_get(cbk->dict, EC_XATTR_VERSION);
- if ((data != NULL) && (data->len >= sizeof(uint64_t))) {
- version = (uint64_t *)data->data;
+ fop = frame->local;
- if (((ntoh64(version[0]) >> EC_SELFHEAL_BIT) & 1) != 0) {
- LOCK(&fop->lock);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
- fop->healing |= 1ULL << idx;
+ cbk = ec_cbk_data_allocate(frame, this, fop, fop->id, idx, op_ret,
+ op_errno);
+ if (!cbk)
+ goto out;
- UNLOCK(&fop->lock);
- }
- }
+ if (op_ret >= 0) {
+ cbk->dict = dict_ref(xattr);
- ec_dict_del_array (xattr, EC_XATTR_DIRTY, dirty,
- EC_VERSION_SIZE);
- link = fop->data;
- if (link) {
- /*Keep a note of if the dirty is already set or not*/
- link->dirty[0] |= (dirty[0] != 0);
- link->dirty[1] |= (dirty[1] != 0);
- }
+ data = dict_get(cbk->dict, EC_XATTR_VERSION);
+ if ((data != NULL) && (data->len >= sizeof(uint64_t))) {
+ version = (uint64_t *)data->data;
+
+ if (((ntoh64(version[0]) >> EC_SELFHEAL_BIT) & 1) != 0) {
+ LOCK(&fop->lock);
+
+ fop->healing |= 1ULL << idx;
+
+ UNLOCK(&fop->lock);
+ }
}
- if (xdata)
- cbk->xdata = dict_ref(xdata);
+ ec_dict_del_array(xattr, EC_XATTR_DIRTY, dirty, EC_VERSION_SIZE);
+ link = fop->data;
+ if (link) {
+ /*Keep a note of if the dirty is already set or not*/
+ link->dirty[0] |= (dirty[0] != 0);
+ link->dirty[1] |= (dirty[1] != 0);
+ }
+ }
+
+ if (xdata)
+ cbk->xdata = dict_ref(xdata);
- ec_combine (cbk, ec_combine_xattrop);
+ ec_combine(cbk, ec_combine_xattrop);
out:
- if (fop)
- ec_complete(fop);
+ if (fop)
+ ec_complete(fop);
- return 0;
+ return 0;
}
-void ec_wind_xattrop(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_xattrop(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -1211,20 +1221,20 @@ void ec_wind_xattrop(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
&fop->loc[0], fop->xattrop_flags, fop->dict, fop->xdata);
}
-int32_t ec_manager_xattrop(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_xattrop(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_LOCK:
if (fop->fd == NULL) {
ec_lock_prepare_inode(fop, &fop->loc[0], EC_UPDATE_META, 0,
- LLONG_MAX);
+ EC_RANGE_FULL);
} else {
ec_lock_prepare_fd(fop, fop->fd, EC_UPDATE_META, 0,
- LLONG_MAX);
+ EC_RANGE_FULL);
}
ec_lock(fop);
@@ -1251,19 +1261,13 @@ int32_t ec_manager_xattrop(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->id == GF_FOP_XATTROP)
- {
- if (fop->cbks.xattrop != NULL)
- {
- fop->cbks.xattrop(fop->req_frame, fop, fop->xl,
- cbk->op_ret, cbk->op_errno, cbk->dict,
- cbk->xdata);
+ if (fop->id == GF_FOP_XATTROP) {
+ if (fop->cbks.xattrop != NULL) {
+ fop->cbks.xattrop(fop->req_frame, fop, fop->xl, cbk->op_ret,
+ cbk->op_errno, cbk->dict, cbk->xdata);
}
- }
- else
- {
- if (fop->cbks.fxattrop != NULL)
- {
+ } else {
+ if (fop->cbks.fxattrop != NULL) {
fop->cbks.fxattrop(fop->req_frame, fop, fop->xl,
cbk->op_ret, cbk->op_errno, cbk->dict,
cbk->xdata);
@@ -1279,18 +1283,13 @@ int32_t ec_manager_xattrop(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->id == GF_FOP_XATTROP)
- {
- if (fop->cbks.xattrop != NULL)
- {
+ if (fop->id == GF_FOP_XATTROP) {
+ if (fop->cbks.xattrop != NULL) {
fop->cbks.xattrop(fop->req_frame, fop, fop->xl, -1,
fop->error, NULL, NULL);
}
- }
- else
- {
- if (fop->cbks.fxattrop != NULL)
- {
+ } else {
+ if (fop->cbks.fxattrop != NULL) {
fop->cbks.fxattrop(fop->req_frame, fop, fop->xl, -1,
fop->error, NULL, NULL);
}
@@ -1311,32 +1310,31 @@ int32_t ec_manager_xattrop(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_xattrop(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_xattrop_cbk_t func, void * data,
- loc_t * loc, gf_xattrop_flags_t optype, dict_t * xattr,
- dict_t * xdata)
+void
+ec_xattrop(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_xattrop_cbk_t func, void *data, loc_t *loc,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
- ec_cbk_t callback = { .xattrop = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.xattrop = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(XATTROP) %p", frame);
+ gf_msg_trace("ec", 0, "EC(XATTROP) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_XATTROP, 0, target, minimum,
- ec_wind_xattrop, ec_manager_xattrop, callback,
- data);
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_XATTROP, 0, target,
+ fop_flags, ec_wind_xattrop, ec_manager_xattrop,
+ callback, data);
if (fop == NULL) {
goto out;
}
@@ -1345,8 +1343,8 @@ void ec_xattrop(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
@@ -1354,9 +1352,9 @@ void ec_xattrop(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xattr != NULL) {
fop->dict = dict_ref(xattr);
if (fop->dict == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -1364,9 +1362,9 @@ void ec_xattrop(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -1382,7 +1380,8 @@ out:
}
}
-void ec_wind_fxattrop(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_fxattrop(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -1391,23 +1390,23 @@ void ec_wind_fxattrop(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->fd, fop->xattrop_flags, fop->dict, fop->xdata);
}
-void ec_fxattrop(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fxattrop_cbk_t func, void * data,
- fd_t * fd, gf_xattrop_flags_t optype, dict_t * xattr,
- dict_t * xdata)
+void
+ec_fxattrop(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fxattrop_cbk_t func, void *data, fd_t *fd,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
- ec_cbk_t callback = { .fxattrop = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.fxattrop = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(FXATTROP) %p", frame);
+ gf_msg_trace("ec", 0, "EC(FXATTROP) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_FXATTROP, 0, target,
- minimum, ec_wind_fxattrop, ec_manager_xattrop,
+ fop_flags, ec_wind_fxattrop, ec_manager_xattrop,
callback, data);
if (fop == NULL) {
goto out;
@@ -1420,9 +1419,9 @@ void ec_fxattrop(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL, "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -1430,9 +1429,9 @@ void ec_fxattrop(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xattr != NULL) {
fop->dict = dict_ref(xattr);
if (fop->dict == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -1440,9 +1439,9 @@ void ec_fxattrop(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -1460,11 +1459,12 @@ out:
/* FOP: IPC */
-int32_t ec_ipc_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, dict_t * xdata)
+int32_t
+ec_ipc_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -1474,16 +1474,14 @@ int32_t ec_ipc_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_IPC, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (xdata != NULL)
- {
+ if (cbk != NULL) {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
}
@@ -1491,29 +1489,29 @@ int32_t ec_ipc_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_ipc(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_ipc(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
STACK_WIND_COOKIE(fop->frame, ec_ipc_cbk, (void *)(uintptr_t)idx,
- ec->xl_list[idx], ec->xl_list[idx]->fops->ipc,
- fop->int32, fop->xdata);
+ ec->xl_list[idx], ec->xl_list[idx]->fops->ipc, fop->int32,
+ fop->xdata);
}
-int32_t ec_manager_ipc(ec_fop_data_t *fop, int32_t state)
+int32_t
+ec_manager_ipc(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_DISPATCH:
ec_dispatch_all(fop);
@@ -1529,10 +1527,9 @@ int32_t ec_manager_ipc(ec_fop_data_t *fop, int32_t state)
cbk = fop->answer;
GF_ASSERT(cbk != NULL);
- if (fop->cbks.ipc != NULL)
- {
+ if (fop->cbks.ipc != NULL) {
fop->cbks.ipc(fop->req_frame, fop, fop->xl, cbk->op_ret,
- cbk->op_errno, cbk->xdata);
+ cbk->op_errno, cbk->xdata);
}
return EC_STATE_END;
@@ -1543,8 +1540,7 @@ int32_t ec_manager_ipc(ec_fop_data_t *fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.ipc != NULL)
- {
+ if (fop->cbks.ipc != NULL) {
fop->cbks.ipc(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL);
}
@@ -1552,29 +1548,29 @@ int32_t ec_manager_ipc(ec_fop_data_t *fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_ipc(call_frame_t *frame, xlator_t *this, uintptr_t target,
- int32_t minimum, fop_ipc_cbk_t func, void *data, int32_t op,
- dict_t *xdata)
+void
+ec_ipc(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_ipc_cbk_t func, void *data, int32_t op,
+ dict_t *xdata)
{
- ec_cbk_t callback = { .ipc = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.ipc = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(IPC) %p", frame);
+ gf_msg_trace("ec", 0, "EC(IPC) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_IPC, 0, target, minimum,
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_IPC, 0, target, fop_flags,
ec_wind_ipc, ec_manager_ipc, callback, data);
if (fop == NULL) {
goto out;
diff --git a/xlators/cluster/ec/src/ec-gf8.c b/xlators/cluster/ec/src/ec-gf8.c
index 2665632706b..039adae5929 100644
--- a/xlators/cluster/ec/src/ec-gf8.c
+++ b/xlators/cluster/ec/src/ec-gf8.c
@@ -10,5886 +10,5810 @@
#include "ec-gf8.h"
-static ec_gf_op_t ec_gf8_mul_00_ops[] = {
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_00 = {
- 0,
- { 0, },
- ec_gf8_mul_00_ops
-};
-
-static ec_gf_op_t ec_gf8_mul_01_ops[] = {
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_01 = {
- 8,
- { 0, 1, 2, 3, 4, 5, 6, 7, },
- ec_gf8_mul_01_ops
-};
-
-static ec_gf_op_t ec_gf8_mul_02_ops[] = {
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_02 = {
- 8,
- { 7, 0, 1, 2, 3, 4, 5, 6, },
- ec_gf8_mul_02_ops
-};
+static ec_gf_op_t ec_gf8_mul_00_ops[] = {{EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_00 = {0,
+ {
+ 0,
+ },
+ ec_gf8_mul_00_ops};
+
+static ec_gf_op_t ec_gf8_mul_01_ops[] = {{EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_01 = {8,
+ {
+ 0,
+ 1,
+ 2,
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ },
+ ec_gf8_mul_01_ops};
+
+static ec_gf_op_t ec_gf8_mul_02_ops[] = {{EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_02 = {8,
+ {
+ 7,
+ 0,
+ 1,
+ 2,
+ 3,
+ 4,
+ 5,
+ 6,
+ },
+ ec_gf8_mul_02_ops};
static ec_gf_op_t ec_gf8_mul_03_ops[] = {
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_COPY, 8, 3, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 4, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_03 = {
- 9,
- { 0, 1, 2, 3, 4, 5, 6, 7, 8, },
- ec_gf8_mul_03_ops
-};
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_COPY, 8, 3, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 4, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_03 = {9,
+ {
+ 0,
+ 1,
+ 2,
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ 8,
+ },
+ ec_gf8_mul_03_ops};
static ec_gf_op_t ec_gf8_mul_04_ops[] = {
- { EC_GF_OP_XOR3, 8, 6, 7 },
- { EC_GF_OP_XOR2, 2, 8, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_04 = {
- 9,
- { 6, 7, 0, 1, 2, 3, 4, 5, 8, },
- ec_gf8_mul_04_ops
-};
+ {EC_GF_OP_XOR3, 8, 6, 7}, {EC_GF_OP_XOR2, 2, 8, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 1, 8, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_04 = {9,
+ {
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ 3,
+ 4,
+ 5,
+ 8,
+ },
+ ec_gf8_mul_04_ops};
static ec_gf_op_t ec_gf8_mul_05_ops[] = {
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_05 = {
- 8,
- { 0, 1, 2, 6, 7, 3, 4, 5, },
- ec_gf8_mul_05_ops
-};
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_05 = {8,
+ {
+ 0,
+ 1,
+ 2,
+ 6,
+ 7,
+ 3,
+ 4,
+ 5,
+ },
+ ec_gf8_mul_05_ops};
static ec_gf_op_t ec_gf8_mul_06_ops[] = {
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_COPY, 8, 2, 0 },
- { EC_GF_OP_XOR2, 8, 3, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_06 = {
- 9,
- { 7, 0, 1, 2, 8, 3, 4, 5, 6, },
- ec_gf8_mul_06_ops
-};
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_COPY, 8, 2, 0},
+ {EC_GF_OP_XOR2, 8, 3, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 1, 0, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_06 = {9,
+ {
+ 7,
+ 0,
+ 1,
+ 2,
+ 8,
+ 3,
+ 4,
+ 5,
+ 6,
+ },
+ ec_gf8_mul_06_ops};
static ec_gf_op_t ec_gf8_mul_07_ops[] = {
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_07 = {
- 8,
- { 6, 0, 1, 3, 2, 4, 5, 7, },
- ec_gf8_mul_07_ops
-};
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_07 = {8,
+ {
+ 6,
+ 0,
+ 1,
+ 3,
+ 2,
+ 4,
+ 5,
+ 7,
+ },
+ ec_gf8_mul_07_ops};
static ec_gf_op_t ec_gf8_mul_08_ops[] = {
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR3, 8, 6, 7 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 2, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_08 = {
- 9,
- { 5, 6, 7, 0, 1, 2, 3, 4, 8, },
- ec_gf8_mul_08_ops
-};
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR3, 8, 6, 7},
+ {EC_GF_OP_XOR2, 1, 8, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_XOR2, 2, 8, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_08 = {9,
+ {
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ 3,
+ 4,
+ 8,
+ },
+ ec_gf8_mul_08_ops};
static ec_gf_op_t ec_gf8_mul_09_ops[] = {
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_09 = {
- 8,
- { 0, 1, 2, 3, 5, 6, 7, 4, },
- ec_gf8_mul_09_ops
-};
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_09 = {8,
+ {
+ 0,
+ 1,
+ 2,
+ 3,
+ 5,
+ 6,
+ 7,
+ 4,
+ },
+ ec_gf8_mul_09_ops};
static ec_gf_op_t ec_gf8_mul_0A_ops[] = {
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_0A = {
- 8,
- { 5, 0, 1, 2, 6, 7, 3, 4, },
- ec_gf8_mul_0A_ops
-};
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_0A = {8,
+ {
+ 5,
+ 0,
+ 1,
+ 2,
+ 6,
+ 7,
+ 3,
+ 4,
+ },
+ ec_gf8_mul_0A_ops};
static ec_gf_op_t ec_gf8_mul_0B_ops[] = {
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_COPY, 9, 3, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_COPY, 8, 5, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR3, 3, 8, 6 },
- { EC_GF_OP_XOR2, 1, 9, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_0B = {
- 10,
- { 7, 1, 5, 2, 4, 3, 0, 6, 8, 9, },
- ec_gf8_mul_0B_ops
-};
+ {EC_GF_OP_XOR2, 4, 7, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_COPY, 9, 3, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_COPY, 8, 5, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 1, 0, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR3, 3, 8, 6}, {EC_GF_OP_XOR2, 1, 9, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_0B = {10,
+ {
+ 7,
+ 1,
+ 5,
+ 2,
+ 4,
+ 3,
+ 0,
+ 6,
+ 8,
+ 9,
+ },
+ ec_gf8_mul_0B_ops};
static ec_gf_op_t ec_gf8_mul_0C_ops[] = {
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_COPY, 8, 1, 0 },
- { EC_GF_OP_XOR2, 8, 2, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_0C = {
- 9,
- { 5, 7, 0, 1, 8, 2, 3, 4, 6, },
- ec_gf8_mul_0C_ops
-};
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_COPY, 8, 1, 0},
+ {EC_GF_OP_XOR2, 8, 2, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 1, 0, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_0C = {9,
+ {
+ 5,
+ 7,
+ 0,
+ 1,
+ 8,
+ 2,
+ 3,
+ 4,
+ 6,
+ },
+ ec_gf8_mul_0C_ops};
static ec_gf_op_t ec_gf8_mul_0D_ops[] = {
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR3, 8, 2, 4 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR3, 2, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_0D = {
- 9,
- { 5, 6, 7, 3, 1, 0, 2, 4, 8, },
- ec_gf8_mul_0D_ops
-};
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_XOR3, 8, 2, 4}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 1, 8, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR3, 2, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_0D = {9,
+ {
+ 5,
+ 6,
+ 7,
+ 3,
+ 1,
+ 0,
+ 2,
+ 4,
+ 8,
+ },
+ ec_gf8_mul_0D_ops};
static ec_gf_op_t ec_gf8_mul_0E_ops[] = {
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_0E = {
- 8,
- { 7, 0, 6, 1, 3, 2, 4, 5, },
- ec_gf8_mul_0E_ops
-};
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_0E = {8,
+ {
+ 7,
+ 0,
+ 6,
+ 1,
+ 3,
+ 2,
+ 4,
+ 5,
+ },
+ ec_gf8_mul_0E_ops};
static ec_gf_op_t ec_gf8_mul_0F_ops[] = {
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_0F = {
- 8,
- { 1, 0, 5, 6, 7, 2, 3, 4, },
- ec_gf8_mul_0F_ops
-};
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_0F = {8,
+ {
+ 1,
+ 0,
+ 5,
+ 6,
+ 7,
+ 2,
+ 3,
+ 4,
+ },
+ ec_gf8_mul_0F_ops};
static ec_gf_op_t ec_gf8_mul_10_ops[] = {
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_10 = {
- 8,
- { 4, 5, 6, 7, 0, 1, 2, 3, },
- ec_gf8_mul_10_ops
-};
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_10 = {8,
+ {
+ 4,
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ 3,
+ },
+ ec_gf8_mul_10_ops};
static ec_gf_op_t ec_gf8_mul_11_ops[] = {
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_11 = {
- 8,
- { 4, 1, 2, 6, 0, 5, 7, 3, },
- ec_gf8_mul_11_ops
-};
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_11 = {8,
+ {
+ 4,
+ 1,
+ 2,
+ 6,
+ 0,
+ 5,
+ 7,
+ 3,
+ },
+ ec_gf8_mul_11_ops};
static ec_gf_op_t ec_gf8_mul_12_ops[] = {
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_12 = {
- 8,
- { 7, 0, 1, 2, 3, 5, 6, 4, },
- ec_gf8_mul_12_ops
-};
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_12 = {8,
+ {
+ 7,
+ 0,
+ 1,
+ 2,
+ 3,
+ 5,
+ 6,
+ 4,
+ },
+ ec_gf8_mul_12_ops};
static ec_gf_op_t ec_gf8_mul_13_ops[] = {
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR3, 8, 3, 7 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 6, 8, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 0, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_13 = {
- 9,
- { 4, 5, 2, 6, 0, 1, 7, 3, 8, },
- ec_gf8_mul_13_ops
-};
+ {EC_GF_OP_XOR2, 4, 7, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR3, 8, 3, 7},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 6, 8, 0},
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 0, 8, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_13 = {9,
+ {
+ 4,
+ 5,
+ 2,
+ 6,
+ 0,
+ 1,
+ 7,
+ 3,
+ 8,
+ },
+ ec_gf8_mul_13_ops};
static ec_gf_op_t ec_gf8_mul_14_ops[] = {
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_14 = {
- 8,
- { 6, 7, 0, 1, 2, 4, 5, 3, },
- ec_gf8_mul_14_ops
-};
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_14 = {8,
+ {
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ 4,
+ 5,
+ 3,
+ },
+ ec_gf8_mul_14_ops};
static ec_gf_op_t ec_gf8_mul_15_ops[] = {
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR3, 5, 8, 7 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_15 = {
- 9,
- { 0, 1, 2, 4, 7, 6, 5, 3, 8, },
- ec_gf8_mul_15_ops
-};
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR3, 5, 8, 7},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_15 = {9,
+ {
+ 0,
+ 1,
+ 2,
+ 4,
+ 7,
+ 6,
+ 5,
+ 3,
+ 8,
+ },
+ ec_gf8_mul_15_ops};
static ec_gf_op_t ec_gf8_mul_16_ops[] = {
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_16 = {
- 8,
- { 6, 7, 4, 1, 2, 3, 5, 0, },
- ec_gf8_mul_16_ops
-};
+ {EC_GF_OP_XOR2, 4, 7, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_16 = {8,
+ {
+ 6,
+ 7,
+ 4,
+ 1,
+ 2,
+ 3,
+ 5,
+ 0,
+ },
+ ec_gf8_mul_16_ops};
static ec_gf_op_t ec_gf8_mul_17_ops[] = {
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_17 = {
- 8,
- { 5, 7, 0, 1, 3, 2, 4, 6, },
- ec_gf8_mul_17_ops
-};
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_17 = {8,
+ {
+ 5,
+ 7,
+ 0,
+ 1,
+ 3,
+ 2,
+ 4,
+ 6,
+ },
+ ec_gf8_mul_17_ops};
static ec_gf_op_t ec_gf8_mul_18_ops[] = {
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 6, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_18 = {
- 9,
- { 4, 5, 7, 6, 0, 1, 2, 3, 8, },
- ec_gf8_mul_18_ops
-};
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 6, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_18 = {9,
+ {
+ 4,
+ 5,
+ 7,
+ 6,
+ 0,
+ 1,
+ 2,
+ 3,
+ 8,
+ },
+ ec_gf8_mul_18_ops};
static ec_gf_op_t ec_gf8_mul_19_ops[] = {
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_19 = {
- 8,
- { 0, 5, 2, 6, 7, 1, 3, 4, },
- ec_gf8_mul_19_ops
-};
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_19 = {8,
+ {
+ 0,
+ 5,
+ 2,
+ 6,
+ 7,
+ 1,
+ 3,
+ 4,
+ },
+ ec_gf8_mul_19_ops};
static ec_gf_op_t ec_gf8_mul_1A_ops[] = {
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_1A = {
- 8,
- { 7, 0, 4, 5, 3, 1, 2, 6, },
- ec_gf8_mul_1A_ops
-};
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_1A = {8,
+ {
+ 7,
+ 0,
+ 4,
+ 5,
+ 3,
+ 1,
+ 2,
+ 6,
+ },
+ ec_gf8_mul_1A_ops};
static ec_gf_op_t ec_gf8_mul_1B_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_1B = {
- 8,
- { 7, 4, 5, 6, 3, 1, 2, 0, },
- ec_gf8_mul_1B_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_1B = {8,
+ {
+ 7,
+ 4,
+ 5,
+ 6,
+ 3,
+ 1,
+ 2,
+ 0,
+ },
+ ec_gf8_mul_1B_ops};
static ec_gf_op_t ec_gf8_mul_1C_ops[] = {
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_1C = {
- 8,
- { 5, 4, 3, 0, 1, 7, 2, 6, },
- ec_gf8_mul_1C_ops
-};
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_1C = {8,
+ {
+ 5,
+ 4,
+ 3,
+ 0,
+ 1,
+ 7,
+ 2,
+ 6,
+ },
+ ec_gf8_mul_1C_ops};
static ec_gf_op_t ec_gf8_mul_1D_ops[] = {
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR3, 8, 4, 2 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 8, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_1D = {
- 9,
- { 0, 7, 5, 8, 2, 3, 4, 1, 6, },
- ec_gf8_mul_1D_ops
-};
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 1, 0, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR3, 8, 4, 2},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 5, 8, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_1D = {9,
+ {
+ 0,
+ 7,
+ 5,
+ 8,
+ 2,
+ 3,
+ 4,
+ 1,
+ 6,
+ },
+ ec_gf8_mul_1D_ops};
static ec_gf_op_t ec_gf8_mul_1E_ops[] = {
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_1E = {
- 8,
- { 4, 7, 5, 1, 6, 0, 2, 3, },
- ec_gf8_mul_1E_ops
-};
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_1E = {8,
+ {
+ 4,
+ 7,
+ 5,
+ 1,
+ 6,
+ 0,
+ 2,
+ 3,
+ },
+ ec_gf8_mul_1E_ops};
static ec_gf_op_t ec_gf8_mul_1F_ops[] = {
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR3, 8, 3, 7 },
- { EC_GF_OP_XOR2, 4, 8, 0 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_1F = {
- 9,
- { 1, 4, 5, 6, 7, 0, 3, 2, 8, },
- ec_gf8_mul_1F_ops
-};
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR3, 8, 3, 7},
+ {EC_GF_OP_XOR2, 4, 8, 0}, {EC_GF_OP_XOR2, 1, 8, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_1F = {9,
+ {
+ 1,
+ 4,
+ 5,
+ 6,
+ 7,
+ 0,
+ 3,
+ 2,
+ 8,
+ },
+ ec_gf8_mul_1F_ops};
static ec_gf_op_t ec_gf8_mul_20_ops[] = {
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_20 = {
- 8,
- { 7, 4, 5, 6, 3, 0, 1, 2, },
- ec_gf8_mul_20_ops
-};
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_20 = {8,
+ {
+ 7,
+ 4,
+ 5,
+ 6,
+ 3,
+ 0,
+ 1,
+ 2,
+ },
+ ec_gf8_mul_20_ops};
static ec_gf_op_t ec_gf8_mul_21_ops[] = {
- { EC_GF_OP_COPY, 9, 0, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR3, 8, 7, 5 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 3, 8, 0 },
- { EC_GF_OP_XOR2, 2, 8, 0 },
- { EC_GF_OP_XOR2, 4, 9, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_21 = {
- 10,
- { 0, 1, 2, 7, 5, 4, 3, 6, 8, 9, },
- ec_gf8_mul_21_ops
-};
+ {EC_GF_OP_COPY, 9, 0, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR3, 8, 7, 5}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_XOR2, 3, 8, 0}, {EC_GF_OP_XOR2, 2, 8, 0},
+ {EC_GF_OP_XOR2, 4, 9, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_21 = {10,
+ {
+ 0,
+ 1,
+ 2,
+ 7,
+ 5,
+ 4,
+ 3,
+ 6,
+ 8,
+ 9,
+ },
+ ec_gf8_mul_21_ops};
static ec_gf_op_t ec_gf8_mul_22_ops[] = {
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_22 = {
- 8,
- { 3, 0, 5, 2, 6, 4, 1, 7, },
- ec_gf8_mul_22_ops
-};
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_22 = {8,
+ {
+ 3,
+ 0,
+ 5,
+ 2,
+ 6,
+ 4,
+ 1,
+ 7,
+ },
+ ec_gf8_mul_22_ops};
static ec_gf_op_t ec_gf8_mul_23_ops[] = {
- { EC_GF_OP_COPY, 8, 2, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 3, 8, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_23 = {
- 9,
- { 0, 4, 3, 2, 5, 6, 1, 8, 7, },
- ec_gf8_mul_23_ops
-};
+ {EC_GF_OP_COPY, 8, 2, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 3, 8, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_23 = {9,
+ {
+ 0,
+ 4,
+ 3,
+ 2,
+ 5,
+ 6,
+ 1,
+ 8,
+ 7,
+ },
+ ec_gf8_mul_23_ops};
static ec_gf_op_t ec_gf8_mul_24_ops[] = {
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_24 = {
- 8,
- { 6, 7, 0, 1, 2, 4, 5, 3, },
- ec_gf8_mul_24_ops
-};
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_24 = {8,
+ {
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ 4,
+ 5,
+ 3,
+ },
+ ec_gf8_mul_24_ops};
static ec_gf_op_t ec_gf8_mul_25_ops[] = {
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_25 = {
- 8,
- { 2, 7, 0, 1, 3, 4, 5, 6, },
- ec_gf8_mul_25_ops
-};
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_25 = {8,
+ {
+ 2,
+ 7,
+ 0,
+ 1,
+ 3,
+ 4,
+ 5,
+ 6,
+ },
+ ec_gf8_mul_25_ops};
static ec_gf_op_t ec_gf8_mul_26_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_26 = {
- 8,
- { 3, 4, 1, 2, 0, 5, 6, 7, },
- ec_gf8_mul_26_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_26 = {8,
+ {
+ 3,
+ 4,
+ 1,
+ 2,
+ 0,
+ 5,
+ 6,
+ 7,
+ },
+ ec_gf8_mul_26_ops};
static ec_gf_op_t ec_gf8_mul_27_ops[] = {
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_27 = {
- 8,
- { 3, 0, 1, 2, 6, 7, 4, 5, },
- ec_gf8_mul_27_ops
-};
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_27 = {8,
+ {
+ 3,
+ 0,
+ 1,
+ 2,
+ 6,
+ 7,
+ 4,
+ 5,
+ },
+ ec_gf8_mul_27_ops};
static ec_gf_op_t ec_gf8_mul_28_ops[] = {
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_28 = {
- 8,
- { 5, 6, 3, 0, 1, 2, 4, 7, },
- ec_gf8_mul_28_ops
-};
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_28 = {8,
+ {
+ 5,
+ 6,
+ 3,
+ 0,
+ 1,
+ 2,
+ 4,
+ 7,
+ },
+ ec_gf8_mul_28_ops};
static ec_gf_op_t ec_gf8_mul_29_ops[] = {
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_29 = {
- 8,
- { 4, 6, 3, 5, 7, 0, 1, 2, },
- ec_gf8_mul_29_ops
-};
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_29 = {8,
+ {
+ 4,
+ 6,
+ 3,
+ 5,
+ 7,
+ 0,
+ 1,
+ 2,
+ },
+ ec_gf8_mul_29_ops};
static ec_gf_op_t ec_gf8_mul_2A_ops[] = {
- { EC_GF_OP_COPY, 8, 1, 0 },
- { EC_GF_OP_XOR2, 8, 0, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR3, 6, 8, 4 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_2A = {
- 9,
- { 3, 4, 7, 2, 6, 5, 1, 0, 8, },
- ec_gf8_mul_2A_ops
-};
+ {EC_GF_OP_COPY, 8, 1, 0}, {EC_GF_OP_XOR2, 8, 0, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR3, 6, 8, 4}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_2A = {9,
+ {
+ 3,
+ 4,
+ 7,
+ 2,
+ 6,
+ 5,
+ 1,
+ 0,
+ 8,
+ },
+ ec_gf8_mul_2A_ops};
static ec_gf_op_t ec_gf8_mul_2B_ops[] = {
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_2B = {
- 8,
- { 3, 4, 7, 5, 6, 0, 1, 2, },
- ec_gf8_mul_2B_ops
-};
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_2B = {8,
+ {
+ 3,
+ 4,
+ 7,
+ 5,
+ 6,
+ 0,
+ 1,
+ 2,
+ },
+ ec_gf8_mul_2B_ops};
static ec_gf_op_t ec_gf8_mul_2C_ops[] = {
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_2C = {
- 8,
- { 5, 6, 7, 0, 2, 3, 4, 1, },
- ec_gf8_mul_2C_ops
-};
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_2C = {8,
+ {
+ 5,
+ 6,
+ 7,
+ 0,
+ 2,
+ 3,
+ 4,
+ 1,
+ },
+ ec_gf8_mul_2C_ops};
static ec_gf_op_t ec_gf8_mul_2D_ops[] = {
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR3, 8, 4, 6 },
- { EC_GF_OP_XOR2, 5, 8, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_2D = {
- 9,
- { 7, 0, 3, 5, 1, 4, 2, 6, 8, },
- ec_gf8_mul_2D_ops
-};
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR3, 8, 4, 6},
+ {EC_GF_OP_XOR2, 5, 8, 0}, {EC_GF_OP_XOR2, 7, 8, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_2D = {9,
+ {
+ 7,
+ 0,
+ 3,
+ 5,
+ 1,
+ 4,
+ 2,
+ 6,
+ 8,
+ },
+ ec_gf8_mul_2D_ops};
static ec_gf_op_t ec_gf8_mul_2E_ops[] = {
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_COPY, 8, 4, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 8, 7, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 2, 8, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 6, 8, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_2E = {
- 9,
- { 5, 0, 7, 3, 2, 6, 4, 1, 8, },
- ec_gf8_mul_2E_ops
-};
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_COPY, 8, 4, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 8, 7, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 2, 8, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 6, 8, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_2E = {9,
+ {
+ 5,
+ 0,
+ 7,
+ 3,
+ 2,
+ 6,
+ 4,
+ 1,
+ 8,
+ },
+ ec_gf8_mul_2E_ops};
static ec_gf_op_t ec_gf8_mul_2F_ops[] = {
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR3, 8, 7, 6 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 3, 8, 0 },
- { EC_GF_OP_XOR2, 2, 8, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_2F = {
- 9,
- { 6, 3, 2, 5, 7, 0, 1, 4, 8, },
- ec_gf8_mul_2F_ops
-};
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR3, 8, 7, 6}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 3, 8, 0},
+ {EC_GF_OP_XOR2, 2, 8, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_2F = {9,
+ {
+ 6,
+ 3,
+ 2,
+ 5,
+ 7,
+ 0,
+ 1,
+ 4,
+ 8,
+ },
+ ec_gf8_mul_2F_ops};
static ec_gf_op_t ec_gf8_mul_30_ops[] = {
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 8, 1, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR3, 6, 8, 7 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_30 = {
- 9,
- { 3, 4, 7, 5, 0, 6, 1, 2, 8, },
- ec_gf8_mul_30_ops
-};
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 8, 1, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR3, 6, 8, 7},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_30 = {9,
+ {
+ 3,
+ 4,
+ 7,
+ 5,
+ 0,
+ 6,
+ 1,
+ 2,
+ 8,
+ },
+ ec_gf8_mul_30_ops};
static ec_gf_op_t ec_gf8_mul_31_ops[] = {
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_31 = {
- 8,
- { 7, 1, 4, 5, 6, 0, 2, 3, },
- ec_gf8_mul_31_ops
-};
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_31 = {8,
+ {
+ 7,
+ 1,
+ 4,
+ 5,
+ 6,
+ 0,
+ 2,
+ 3,
+ },
+ ec_gf8_mul_31_ops};
static ec_gf_op_t ec_gf8_mul_32_ops[] = {
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_32 = {
- 8,
- { 3, 4, 6, 7, 5, 0, 1, 2, },
- ec_gf8_mul_32_ops
-};
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_32 = {8,
+ {
+ 3,
+ 4,
+ 6,
+ 7,
+ 5,
+ 0,
+ 1,
+ 2,
+ },
+ ec_gf8_mul_32_ops};
static ec_gf_op_t ec_gf8_mul_33_ops[] = {
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_33 = {
- 8,
- { 5, 4, 3, 0, 2, 1, 6, 7, },
- ec_gf8_mul_33_ops
-};
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_33 = {8,
+ {
+ 5,
+ 4,
+ 3,
+ 0,
+ 2,
+ 1,
+ 6,
+ 7,
+ },
+ ec_gf8_mul_33_ops};
static ec_gf_op_t ec_gf8_mul_34_ops[] = {
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_34 = {
- 8,
- { 7, 5, 3, 0, 2, 4, 1, 6, },
- ec_gf8_mul_34_ops
-};
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_34 = {8,
+ {
+ 7,
+ 5,
+ 3,
+ 0,
+ 2,
+ 4,
+ 1,
+ 6,
+ },
+ ec_gf8_mul_34_ops};
static ec_gf_op_t ec_gf8_mul_35_ops[] = {
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_35 = {
- 8,
- { 6, 7, 5, 4, 2, 0, 1, 3, },
- ec_gf8_mul_35_ops
-};
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_35 = {8,
+ {
+ 6,
+ 7,
+ 5,
+ 4,
+ 2,
+ 0,
+ 1,
+ 3,
+ },
+ ec_gf8_mul_35_ops};
static ec_gf_op_t ec_gf8_mul_36_ops[] = {
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_36 = {
- 8,
- { 6, 7, 4, 1, 2, 3, 0, 5, },
- ec_gf8_mul_36_ops
-};
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_36 = {8,
+ {
+ 6,
+ 7,
+ 4,
+ 1,
+ 2,
+ 3,
+ 0,
+ 5,
+ },
+ ec_gf8_mul_36_ops};
static ec_gf_op_t ec_gf8_mul_37_ops[] = {
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR3, 8, 0, 1 },
- { EC_GF_OP_XOR2, 3, 8, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_37 = {
- 9,
- { 6, 7, 2, 1, 0, 3, 4, 5, 8, },
- ec_gf8_mul_37_ops
-};
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR3, 8, 0, 1},
+ {EC_GF_OP_XOR2, 3, 8, 0}, {EC_GF_OP_XOR2, 7, 8, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_37 = {9,
+ {
+ 6,
+ 7,
+ 2,
+ 1,
+ 0,
+ 3,
+ 4,
+ 5,
+ 8,
+ },
+ ec_gf8_mul_37_ops};
static ec_gf_op_t ec_gf8_mul_38_ops[] = {
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR3, 8, 6, 7 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 0, 8, 0 },
- { EC_GF_OP_XOR2, 4, 8, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_38 = {
- 9,
- { 4, 5, 6, 3, 0, 1, 7, 2, 8, },
- ec_gf8_mul_38_ops
-};
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR3, 8, 6, 7},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 0, 8, 0}, {EC_GF_OP_XOR2, 4, 8, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_38 = {9,
+ {
+ 4,
+ 5,
+ 6,
+ 3,
+ 0,
+ 1,
+ 7,
+ 2,
+ 8,
+ },
+ ec_gf8_mul_38_ops};
static ec_gf_op_t ec_gf8_mul_39_ops[] = {
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_39 = {
- 8,
- { 1, 6, 3, 0, 5, 2, 4, 7, },
- ec_gf8_mul_39_ops
-};
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_39 = {8,
+ {
+ 1,
+ 6,
+ 3,
+ 0,
+ 5,
+ 2,
+ 4,
+ 7,
+ },
+ ec_gf8_mul_39_ops};
static ec_gf_op_t ec_gf8_mul_3A_ops[] = {
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_3A = {
- 8,
- { 3, 4, 7, 0, 5, 6, 1, 2, },
- ec_gf8_mul_3A_ops
-};
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_3A = {8,
+ {
+ 3,
+ 4,
+ 7,
+ 0,
+ 5,
+ 6,
+ 1,
+ 2,
+ },
+ ec_gf8_mul_3A_ops};
static ec_gf_op_t ec_gf8_mul_3B_ops[] = {
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR3, 8, 7, 3 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_3B = {
- 9,
- { 3, 0, 1, 7, 6, 2, 4, 8, 5, },
- ec_gf8_mul_3B_ops
-};
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR3, 8, 7, 3}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 1, 8, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_3B = {9,
+ {
+ 3,
+ 0,
+ 1,
+ 7,
+ 6,
+ 2,
+ 4,
+ 8,
+ 5,
+ },
+ ec_gf8_mul_3B_ops};
static ec_gf_op_t ec_gf8_mul_3C_ops[] = {
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_3C = {
- 8,
- { 3, 6, 4, 1, 7, 2, 0, 5, },
- ec_gf8_mul_3C_ops
-};
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_3C = {8,
+ {
+ 3,
+ 6,
+ 4,
+ 1,
+ 7,
+ 2,
+ 0,
+ 5,
+ },
+ ec_gf8_mul_3C_ops};
static ec_gf_op_t ec_gf8_mul_3D_ops[] = {
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_3D = {
- 8,
- { 2, 3, 4, 5, 6, 7, 0, 1, },
- ec_gf8_mul_3D_ops
-};
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 1, 0, 0},
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_3D = {8,
+ {
+ 2,
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ },
+ ec_gf8_mul_3D_ops};
static ec_gf_op_t ec_gf8_mul_3E_ops[] = {
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_3E = {
- 8,
- { 6, 1, 2, 7, 0, 3, 5, 4, },
- ec_gf8_mul_3E_ops
-};
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_3E = {8,
+ {
+ 6,
+ 1,
+ 2,
+ 7,
+ 0,
+ 3,
+ 5,
+ 4,
+ },
+ ec_gf8_mul_3E_ops};
static ec_gf_op_t ec_gf8_mul_3F_ops[] = {
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_COPY, 10, 4, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_COPY, 9, 2, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR3, 4, 9, 7 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 3, 10, 0 },
- { EC_GF_OP_XOR2, 5, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_3F = {
- 11,
- { 1, 7, 6, 2, 4, 3, 5, 0, 8, 9, 10, },
- ec_gf8_mul_3F_ops
-};
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_COPY, 10, 4, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_COPY, 9, 2, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR3, 4, 9, 7},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 3, 10, 0}, {EC_GF_OP_XOR2, 5, 8, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_3F = {11,
+ {
+ 1,
+ 7,
+ 6,
+ 2,
+ 4,
+ 3,
+ 5,
+ 0,
+ 8,
+ 9,
+ 10,
+ },
+ ec_gf8_mul_3F_ops};
static ec_gf_op_t ec_gf8_mul_40_ops[] = {
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR3, 8, 7, 6 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_XOR2, 5, 8, 0 },
- { EC_GF_OP_XOR2, 4, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_40 = {
- 9,
- { 5, 7, 4, 6, 2, 3, 0, 1, 8, },
- ec_gf8_mul_40_ops
-};
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR3, 8, 7, 6},
+ {EC_GF_OP_XOR2, 1, 8, 0}, {EC_GF_OP_XOR2, 5, 8, 0},
+ {EC_GF_OP_XOR2, 4, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_40 = {9,
+ {
+ 5,
+ 7,
+ 4,
+ 6,
+ 2,
+ 3,
+ 0,
+ 1,
+ 8,
+ },
+ ec_gf8_mul_40_ops};
static ec_gf_op_t ec_gf8_mul_41_ops[] = {
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 8, 4, 0 },
- { EC_GF_OP_XOR2, 8, 5, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_41 = {
- 9,
- { 0, 7, 6, 5, 3, 4, 8, 1, 2, },
- ec_gf8_mul_41_ops
-};
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 8, 4, 0},
+ {EC_GF_OP_XOR2, 8, 5, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_41 = {9,
+ {
+ 0,
+ 7,
+ 6,
+ 5,
+ 3,
+ 4,
+ 8,
+ 1,
+ 2,
+ },
+ ec_gf8_mul_41_ops};
static ec_gf_op_t ec_gf8_mul_42_ops[] = {
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 8, 3, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_42 = {
- 9,
- { 2, 7, 1, 6, 4, 3, 0, 5, 8, },
- ec_gf8_mul_42_ops
-};
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 8, 3, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 7, 8, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_42 = {9,
+ {
+ 2,
+ 7,
+ 1,
+ 6,
+ 4,
+ 3,
+ 0,
+ 5,
+ 8,
+ },
+ ec_gf8_mul_42_ops};
static ec_gf_op_t ec_gf8_mul_43_ops[] = {
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_43 = {
- 8,
- { 2, 6, 4, 1, 7, 3, 0, 5, },
- ec_gf8_mul_43_ops
-};
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_43 = {8,
+ {
+ 2,
+ 6,
+ 4,
+ 1,
+ 7,
+ 3,
+ 0,
+ 5,
+ },
+ ec_gf8_mul_43_ops};
static ec_gf_op_t ec_gf8_mul_44_ops[] = {
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_44 = {
- 8,
- { 2, 3, 4, 1, 6, 5, 0, 7, },
- ec_gf8_mul_44_ops
-};
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_44 = {8,
+ {
+ 2,
+ 3,
+ 4,
+ 1,
+ 6,
+ 5,
+ 0,
+ 7,
+ },
+ ec_gf8_mul_44_ops};
static ec_gf_op_t ec_gf8_mul_45_ops[] = {
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_45 = {
- 8,
- { 2, 3, 0, 1, 7, 4, 5, 6, },
- ec_gf8_mul_45_ops
-};
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_45 = {8,
+ {
+ 2,
+ 3,
+ 0,
+ 1,
+ 7,
+ 4,
+ 5,
+ 6,
+ },
+ ec_gf8_mul_45_ops};
static ec_gf_op_t ec_gf8_mul_46_ops[] = {
- { EC_GF_OP_XOR3, 8, 2, 4 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 8, 0, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_46 = {
- 9,
- { 2, 0, 1, 3, 4, 5, 6, 7, 8, },
- ec_gf8_mul_46_ops
-};
+ {EC_GF_OP_XOR3, 8, 2, 4}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 8, 0, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 1, 8, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_46 = {9,
+ {
+ 2,
+ 0,
+ 1,
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ 8,
+ },
+ ec_gf8_mul_46_ops};
static ec_gf_op_t ec_gf8_mul_47_ops[] = {
- { EC_GF_OP_XOR3, 8, 0, 1 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 3, 8, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 4, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_47 = {
- 9,
- { 2, 3, 4, 5, 6, 7, 0, 1, 8, },
- ec_gf8_mul_47_ops
-};
+ {EC_GF_OP_XOR3, 8, 0, 1}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 3, 8, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 4, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_47 = {9,
+ {
+ 2,
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ 8,
+ },
+ ec_gf8_mul_47_ops};
static ec_gf_op_t ec_gf8_mul_48_ops[] = {
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_48 = {
- 8,
- { 4, 5, 6, 0, 1, 3, 7, 2, },
- ec_gf8_mul_48_ops
-};
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_48 = {8,
+ {
+ 4,
+ 5,
+ 6,
+ 0,
+ 1,
+ 3,
+ 7,
+ 2,
+ },
+ ec_gf8_mul_48_ops};
static ec_gf_op_t ec_gf8_mul_49_ops[] = {
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR3, 8, 0, 6 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR3, 1, 8, 5 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_49 = {
- 9,
- { 7, 2, 4, 0, 3, 5, 1, 6, 8, },
- ec_gf8_mul_49_ops
-};
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR3, 8, 0, 6},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 7, 8, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR3, 1, 8, 5},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_49 = {9,
+ {
+ 7,
+ 2,
+ 4,
+ 0,
+ 3,
+ 5,
+ 1,
+ 6,
+ 8,
+ },
+ ec_gf8_mul_49_ops};
static ec_gf_op_t ec_gf8_mul_4A_ops[] = {
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_4A = {
- 8,
- { 5, 6, 7, 0, 1, 3, 4, 2, },
- ec_gf8_mul_4A_ops
-};
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_4A = {8,
+ {
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ 3,
+ 4,
+ 2,
+ },
+ ec_gf8_mul_4A_ops};
static ec_gf_op_t ec_gf8_mul_4B_ops[] = {
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR3, 8, 3, 7 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 4, 8, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 5, 8, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_4B = {
- 9,
- { 5, 3, 6, 7, 0, 2, 4, 1, 8, },
- ec_gf8_mul_4B_ops
-};
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR3, 8, 3, 7}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 4, 8, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 5, 8, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_4B = {9,
+ {
+ 5,
+ 3,
+ 6,
+ 7,
+ 0,
+ 2,
+ 4,
+ 1,
+ 8,
+ },
+ ec_gf8_mul_4B_ops};
static ec_gf_op_t ec_gf8_mul_4C_ops[] = {
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_4C = {
- 8,
- { 5, 3, 4, 7, 0, 6, 2, 1, },
- ec_gf8_mul_4C_ops
-};
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_4C = {8,
+ {
+ 5,
+ 3,
+ 4,
+ 7,
+ 0,
+ 6,
+ 2,
+ 1,
+ },
+ ec_gf8_mul_4C_ops};
static ec_gf_op_t ec_gf8_mul_4D_ops[] = {
- { EC_GF_OP_COPY, 8, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR3, 9, 3, 1 },
- { EC_GF_OP_XOR2, 5, 9, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR3, 0, 8, 2 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_4D = {
- 10,
- { 0, 9, 3, 5, 6, 4, 7, 1, 2, 8, },
- ec_gf8_mul_4D_ops
-};
+ {EC_GF_OP_COPY, 8, 3, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR3, 9, 3, 1},
+ {EC_GF_OP_XOR2, 5, 9, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR3, 0, 8, 2},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_4D = {10,
+ {
+ 0,
+ 9,
+ 3,
+ 5,
+ 6,
+ 4,
+ 7,
+ 1,
+ 2,
+ 8,
+ },
+ ec_gf8_mul_4D_ops};
static ec_gf_op_t ec_gf8_mul_4E_ops[] = {
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_4E = {
- 8,
- { 2, 3, 0, 1, 5, 6, 7, 4, },
- ec_gf8_mul_4E_ops
-};
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_4E = {8,
+ {
+ 2,
+ 3,
+ 0,
+ 1,
+ 5,
+ 6,
+ 7,
+ 4,
+ },
+ ec_gf8_mul_4E_ops};
static ec_gf_op_t ec_gf8_mul_4F_ops[] = {
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_4F = {
- 8,
- { 0, 3, 5, 6, 1, 2, 7, 4, },
- ec_gf8_mul_4F_ops
-};
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 1, 0, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_4F = {8,
+ {
+ 0,
+ 3,
+ 5,
+ 6,
+ 1,
+ 2,
+ 7,
+ 4,
+ },
+ ec_gf8_mul_4F_ops};
static ec_gf_op_t ec_gf8_mul_50_ops[] = {
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_50 = {
- 8,
- { 4, 5, 7, 3, 0, 1, 2, 6, },
- ec_gf8_mul_50_ops
-};
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 4, 7, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_50 = {8,
+ {
+ 4,
+ 5,
+ 7,
+ 3,
+ 0,
+ 1,
+ 2,
+ 6,
+ },
+ ec_gf8_mul_50_ops};
static ec_gf_op_t ec_gf8_mul_51_ops[] = {
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_51 = {
- 8,
- { 0, 1, 7, 2, 3, 4, 5, 6, },
- ec_gf8_mul_51_ops
-};
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_51 = {8,
+ {
+ 0,
+ 1,
+ 7,
+ 2,
+ 3,
+ 4,
+ 5,
+ 6,
+ },
+ ec_gf8_mul_51_ops};
static ec_gf_op_t ec_gf8_mul_52_ops[] = {
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_COPY, 9, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR3, 3, 5, 8 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 2, 9, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_52 = {
- 10,
- { 2, 3, 1, 4, 6, 7, 0, 5, 8, 9, },
- ec_gf8_mul_52_ops
-};
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_COPY, 9, 4, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR3, 3, 5, 8},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 2, 9, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_52 = {10,
+ {
+ 2,
+ 3,
+ 1,
+ 4,
+ 6,
+ 7,
+ 0,
+ 5,
+ 8,
+ 9,
+ },
+ ec_gf8_mul_52_ops};
static ec_gf_op_t ec_gf8_mul_53_ops[] = {
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_53 = {
- 8,
- { 2, 0, 1, 4, 5, 6, 7, 3, },
- ec_gf8_mul_53_ops
-};
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_53 = {8,
+ {
+ 2,
+ 0,
+ 1,
+ 4,
+ 5,
+ 6,
+ 7,
+ 3,
+ },
+ ec_gf8_mul_53_ops};
static ec_gf_op_t ec_gf8_mul_54_ops[] = {
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_54 = {
- 8,
- { 7, 3, 0, 4, 2, 6, 5, 1, },
- ec_gf8_mul_54_ops
-};
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_54 = {8,
+ {
+ 7,
+ 3,
+ 0,
+ 4,
+ 2,
+ 6,
+ 5,
+ 1,
+ },
+ ec_gf8_mul_54_ops};
static ec_gf_op_t ec_gf8_mul_55_ops[] = {
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_55 = {
- 8,
- { 1, 5, 6, 4, 3, 7, 2, 0, },
- ec_gf8_mul_55_ops
-};
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_55 = {8,
+ {
+ 1,
+ 5,
+ 6,
+ 4,
+ 3,
+ 7,
+ 2,
+ 0,
+ },
+ ec_gf8_mul_55_ops};
static ec_gf_op_t ec_gf8_mul_56_ops[] = {
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_56 = {
- 8,
- { 2, 3, 0, 4, 5, 6, 7, 1, },
- ec_gf8_mul_56_ops
-};
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_56 = {8,
+ {
+ 2,
+ 3,
+ 0,
+ 4,
+ 5,
+ 6,
+ 7,
+ 1,
+ },
+ ec_gf8_mul_56_ops};
static ec_gf_op_t ec_gf8_mul_57_ops[] = {
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_57 = {
- 8,
- { 2, 3, 0, 1, 4, 5, 6, 7, },
- ec_gf8_mul_57_ops
-};
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_57 = {8,
+ {
+ 2,
+ 3,
+ 0,
+ 1,
+ 4,
+ 5,
+ 6,
+ 7,
+ },
+ ec_gf8_mul_57_ops};
static ec_gf_op_t ec_gf8_mul_58_ops[] = {
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_58 = {
- 8,
- { 4, 3, 2, 7, 0, 1, 5, 6, },
- ec_gf8_mul_58_ops
-};
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_58 = {8,
+ {
+ 4,
+ 3,
+ 2,
+ 7,
+ 0,
+ 1,
+ 5,
+ 6,
+ },
+ ec_gf8_mul_58_ops};
static ec_gf_op_t ec_gf8_mul_59_ops[] = {
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_59 = {
- 8,
- { 7, 3, 5, 6, 1, 2, 0, 4, },
- ec_gf8_mul_59_ops
-};
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_59 = {8,
+ {
+ 7,
+ 3,
+ 5,
+ 6,
+ 1,
+ 2,
+ 0,
+ 4,
+ },
+ ec_gf8_mul_59_ops};
static ec_gf_op_t ec_gf8_mul_5A_ops[] = {
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_5A = {
- 8,
- { 6, 7, 0, 1, 2, 3, 5, 4, },
- ec_gf8_mul_5A_ops
-};
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_5A = {8,
+ {
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ 3,
+ 5,
+ 4,
+ },
+ ec_gf8_mul_5A_ops};
static ec_gf_op_t ec_gf8_mul_5B_ops[] = {
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_5B = {
- 8,
- { 6, 0, 7, 5, 2, 1, 3, 4, },
- ec_gf8_mul_5B_ops
-};
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_5B = {8,
+ {
+ 6,
+ 0,
+ 7,
+ 5,
+ 2,
+ 1,
+ 3,
+ 4,
+ },
+ ec_gf8_mul_5B_ops};
static ec_gf_op_t ec_gf8_mul_5C_ops[] = {
- { EC_GF_OP_COPY, 8, 3, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 2, 8, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_5C = {
- 9,
- { 7, 5, 2, 4, 1, 0, 6, 3, 8, },
- ec_gf8_mul_5C_ops
-};
+ {EC_GF_OP_COPY, 8, 3, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_XOR2, 2, 8, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_5C = {9,
+ {
+ 7,
+ 5,
+ 2,
+ 4,
+ 1,
+ 0,
+ 6,
+ 3,
+ 8,
+ },
+ ec_gf8_mul_5C_ops};
static ec_gf_op_t ec_gf8_mul_5D_ops[] = {
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_5D = {
- 8,
- { 1, 3, 5, 4, 6, 7, 2, 0, },
- ec_gf8_mul_5D_ops
-};
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 1, 0, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_5D = {8,
+ {
+ 1,
+ 3,
+ 5,
+ 4,
+ 6,
+ 7,
+ 2,
+ 0,
+ },
+ ec_gf8_mul_5D_ops};
static ec_gf_op_t ec_gf8_mul_5E_ops[] = {
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_5E = {
- 8,
- { 4, 3, 6, 2, 5, 7, 0, 1, },
- ec_gf8_mul_5E_ops
-};
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_5E = {8,
+ {
+ 4,
+ 3,
+ 6,
+ 2,
+ 5,
+ 7,
+ 0,
+ 1,
+ },
+ ec_gf8_mul_5E_ops};
static ec_gf_op_t ec_gf8_mul_5F_ops[] = {
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_5F = {
- 8,
- { 6, 1, 3, 4, 5, 7, 2, 0, },
- ec_gf8_mul_5F_ops
-};
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 1, 0, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_5F = {8,
+ {
+ 6,
+ 1,
+ 3,
+ 4,
+ 5,
+ 7,
+ 2,
+ 0,
+ },
+ ec_gf8_mul_5F_ops};
static ec_gf_op_t ec_gf8_mul_60_ops[] = {
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_60 = {
- 8,
- { 2, 3, 4, 7, 5, 6, 0, 1, },
- ec_gf8_mul_60_ops
-};
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_60 = {8,
+ {
+ 2,
+ 3,
+ 4,
+ 7,
+ 5,
+ 6,
+ 0,
+ 1,
+ },
+ ec_gf8_mul_60_ops};
static ec_gf_op_t ec_gf8_mul_61_ops[] = {
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_61 = {
- 8,
- { 0, 5, 6, 7, 4, 2, 1, 3, },
- ec_gf8_mul_61_ops
-};
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_61 = {8,
+ {
+ 0,
+ 5,
+ 6,
+ 7,
+ 4,
+ 2,
+ 1,
+ 3,
+ },
+ ec_gf8_mul_61_ops};
static ec_gf_op_t ec_gf8_mul_62_ops[] = {
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_62 = {
- 8,
- { 2, 0, 3, 4, 5, 6, 7, 1, },
- ec_gf8_mul_62_ops
-};
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_62 = {8,
+ {
+ 2,
+ 0,
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ 1,
+ },
+ ec_gf8_mul_62_ops};
static ec_gf_op_t ec_gf8_mul_63_ops[] = {
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_63 = {
- 8,
- { 3, 4, 6, 5, 7, 0, 1, 2, },
- ec_gf8_mul_63_ops
-};
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_63 = {8,
+ {
+ 3,
+ 4,
+ 6,
+ 5,
+ 7,
+ 0,
+ 1,
+ 2,
+ },
+ ec_gf8_mul_63_ops};
static ec_gf_op_t ec_gf8_mul_64_ops[] = {
- { EC_GF_OP_COPY, 8, 1, 0 },
- { EC_GF_OP_XOR2, 8, 0, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 8, 7, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_64 = {
- 9,
- { 2, 3, 4, 6, 5, 7, 8, 1, 0, },
- ec_gf8_mul_64_ops
-};
+ {EC_GF_OP_COPY, 8, 1, 0}, {EC_GF_OP_XOR2, 8, 0, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 8, 7, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_64 = {9,
+ {
+ 2,
+ 3,
+ 4,
+ 6,
+ 5,
+ 7,
+ 8,
+ 1,
+ 0,
+ },
+ ec_gf8_mul_64_ops};
static ec_gf_op_t ec_gf8_mul_65_ops[] = {
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_65 = {
- 8,
- { 2, 5, 1, 3, 4, 0, 6, 7, },
- ec_gf8_mul_65_ops
-};
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_65 = {8,
+ {
+ 2,
+ 5,
+ 1,
+ 3,
+ 4,
+ 0,
+ 6,
+ 7,
+ },
+ ec_gf8_mul_65_ops};
static ec_gf_op_t ec_gf8_mul_66_ops[] = {
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_66 = {
- 8,
- { 2, 3, 1, 4, 5, 7, 0, 6, },
- ec_gf8_mul_66_ops
-};
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_66 = {8,
+ {
+ 2,
+ 3,
+ 1,
+ 4,
+ 5,
+ 7,
+ 0,
+ 6,
+ },
+ ec_gf8_mul_66_ops};
static ec_gf_op_t ec_gf8_mul_67_ops[] = {
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_67 = {
- 8,
- { 2, 4, 5, 6, 7, 3, 1, 0, },
- ec_gf8_mul_67_ops
-};
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_67 = {8,
+ {
+ 2,
+ 4,
+ 5,
+ 6,
+ 7,
+ 3,
+ 1,
+ 0,
+ },
+ ec_gf8_mul_67_ops};
static ec_gf_op_t ec_gf8_mul_68_ops[] = {
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_68 = {
- 8,
- { 5, 7, 2, 3, 0, 6, 4, 1, },
- ec_gf8_mul_68_ops
-};
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_68 = {8,
+ {
+ 5,
+ 7,
+ 2,
+ 3,
+ 0,
+ 6,
+ 4,
+ 1,
+ },
+ ec_gf8_mul_68_ops};
static ec_gf_op_t ec_gf8_mul_69_ops[] = {
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_69 = {
- 8,
- { 0, 1, 3, 2, 4, 5, 7, 6, },
- ec_gf8_mul_69_ops
-};
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_69 = {8,
+ {
+ 0,
+ 1,
+ 3,
+ 2,
+ 4,
+ 5,
+ 7,
+ 6,
+ },
+ ec_gf8_mul_69_ops};
static ec_gf_op_t ec_gf8_mul_6A_ops[] = {
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_6A = {
- 8,
- { 5, 7, 4, 6, 1, 2, 0, 3, },
- ec_gf8_mul_6A_ops
-};
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_6A = {8,
+ {
+ 5,
+ 7,
+ 4,
+ 6,
+ 1,
+ 2,
+ 0,
+ 3,
+ },
+ ec_gf8_mul_6A_ops};
static ec_gf_op_t ec_gf8_mul_6B_ops[] = {
- { EC_GF_OP_COPY, 8, 1, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 0, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_6B = {
- 9,
- { 6, 7, 2, 0, 3, 1, 5, 4, 8, },
- ec_gf8_mul_6B_ops
-};
+ {EC_GF_OP_COPY, 8, 1, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 0, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_6B = {9,
+ {
+ 6,
+ 7,
+ 2,
+ 0,
+ 3,
+ 1,
+ 5,
+ 4,
+ 8,
+ },
+ ec_gf8_mul_6B_ops};
static ec_gf_op_t ec_gf8_mul_6C_ops[] = {
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_6C = {
- 8,
- { 5, 6, 7, 0, 1, 2, 3, 4, },
- ec_gf8_mul_6C_ops
-};
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_6C = {8,
+ {
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ 3,
+ 4,
+ },
+ ec_gf8_mul_6C_ops};
static ec_gf_op_t ec_gf8_mul_6D_ops[] = {
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR3, 8, 3, 4 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 0, 8, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 6, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_6D = {
- 9,
- { 3, 6, 7, 0, 4, 5, 1, 2, 8, },
- ec_gf8_mul_6D_ops
-};
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 1, 0, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR3, 8, 3, 4}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 0, 8, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_XOR2, 6, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_6D = {9,
+ {
+ 3,
+ 6,
+ 7,
+ 0,
+ 4,
+ 5,
+ 1,
+ 2,
+ 8,
+ },
+ ec_gf8_mul_6D_ops};
static ec_gf_op_t ec_gf8_mul_6E_ops[] = {
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_6E = {
- 8,
- { 5, 6, 3, 1, 7, 2, 0, 4, },
- ec_gf8_mul_6E_ops
-};
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_6E = {8,
+ {
+ 5,
+ 6,
+ 3,
+ 1,
+ 7,
+ 2,
+ 0,
+ 4,
+ },
+ ec_gf8_mul_6E_ops};
static ec_gf_op_t ec_gf8_mul_6F_ops[] = {
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR3, 0, 8, 7 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_6F = {
- 9,
- { 2, 6, 3, 7, 0, 1, 4, 5, 8, },
- ec_gf8_mul_6F_ops
-};
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR3, 0, 8, 7}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_6F = {9,
+ {
+ 2,
+ 6,
+ 3,
+ 7,
+ 0,
+ 1,
+ 4,
+ 5,
+ 8,
+ },
+ ec_gf8_mul_6F_ops};
static ec_gf_op_t ec_gf8_mul_70_ops[] = {
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_70 = {
- 8,
- { 3, 4, 5, 2, 6, 0, 1, 7, },
- ec_gf8_mul_70_ops
-};
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_70 = {8,
+ {
+ 3,
+ 4,
+ 5,
+ 2,
+ 6,
+ 0,
+ 1,
+ 7,
+ },
+ ec_gf8_mul_70_ops};
static ec_gf_op_t ec_gf8_mul_71_ops[] = {
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_71 = {
- 8,
- { 4, 7, 5, 3, 6, 0, 2, 1, },
- ec_gf8_mul_71_ops
-};
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_71 = {8,
+ {
+ 4,
+ 7,
+ 5,
+ 3,
+ 6,
+ 0,
+ 2,
+ 1,
+ },
+ ec_gf8_mul_71_ops};
static ec_gf_op_t ec_gf8_mul_72_ops[] = {
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_72 = {
- 8,
- { 0, 5, 2, 7, 4, 1, 3, 6, },
- ec_gf8_mul_72_ops
-};
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_72 = {8,
+ {
+ 0,
+ 5,
+ 2,
+ 7,
+ 4,
+ 1,
+ 3,
+ 6,
+ },
+ ec_gf8_mul_72_ops};
static ec_gf_op_t ec_gf8_mul_73_ops[] = {
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_73 = {
- 8,
- { 6, 0, 1, 7, 4, 5, 2, 3, },
- ec_gf8_mul_73_ops
-};
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_73 = {8,
+ {
+ 6,
+ 0,
+ 1,
+ 7,
+ 4,
+ 5,
+ 2,
+ 3,
+ },
+ ec_gf8_mul_73_ops};
static ec_gf_op_t ec_gf8_mul_74_ops[] = {
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_74 = {
- 8,
- { 3, 2, 1, 0, 4, 5, 6, 7, },
- ec_gf8_mul_74_ops
-};
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_74 = {8,
+ {
+ 3,
+ 2,
+ 1,
+ 0,
+ 4,
+ 5,
+ 6,
+ 7,
+ },
+ ec_gf8_mul_74_ops};
static ec_gf_op_t ec_gf8_mul_75_ops[] = {
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_75 = {
- 8,
- { 4, 5, 6, 7, 0, 1, 2, 3, },
- ec_gf8_mul_75_ops
-};
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 1, 0, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_75 = {8,
+ {
+ 4,
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ 3,
+ },
+ ec_gf8_mul_75_ops};
static ec_gf_op_t ec_gf8_mul_76_ops[] = {
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR3, 8, 6, 2 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 0, 8, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_76 = {
- 9,
- { 2, 3, 0, 6, 5, 1, 7, 8, 4, },
- ec_gf8_mul_76_ops
-};
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR3, 8, 6, 2},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 0, 8, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_76 = {9,
+ {
+ 2,
+ 3,
+ 0,
+ 6,
+ 5,
+ 1,
+ 7,
+ 8,
+ 4,
+ },
+ ec_gf8_mul_76_ops};
static ec_gf_op_t ec_gf8_mul_77_ops[] = {
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_77 = {
- 8,
- { 7, 4, 3, 6, 0, 1, 5, 2, },
- ec_gf8_mul_77_ops
-};
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_77 = {8,
+ {
+ 7,
+ 4,
+ 3,
+ 6,
+ 0,
+ 1,
+ 5,
+ 2,
+ },
+ ec_gf8_mul_77_ops};
static ec_gf_op_t ec_gf8_mul_78_ops[] = {
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR3, 8, 0, 2 },
- { EC_GF_OP_XOR2, 4, 8, 0 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_78 = {
- 9,
- { 4, 7, 3, 2, 5, 1, 6, 0, 8, },
- ec_gf8_mul_78_ops
-};
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR3, 8, 0, 2},
+ {EC_GF_OP_XOR2, 4, 8, 0}, {EC_GF_OP_XOR2, 1, 8, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_78 = {9,
+ {
+ 4,
+ 7,
+ 3,
+ 2,
+ 5,
+ 1,
+ 6,
+ 0,
+ 8,
+ },
+ ec_gf8_mul_78_ops};
static ec_gf_op_t ec_gf8_mul_79_ops[] = {
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR3, 8, 4, 7 },
- { EC_GF_OP_XOR2, 0, 8, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 6, 8, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_79 = {
- 9,
- { 4, 5, 7, 3, 1, 6, 2, 0, 8, },
- ec_gf8_mul_79_ops
-};
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR3, 8, 4, 7},
+ {EC_GF_OP_XOR2, 0, 8, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 6, 8, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_79 = {9,
+ {
+ 4,
+ 5,
+ 7,
+ 3,
+ 1,
+ 6,
+ 2,
+ 0,
+ 8,
+ },
+ ec_gf8_mul_79_ops};
static ec_gf_op_t ec_gf8_mul_7A_ops[] = {
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_7A = {
- 8,
- { 1, 2, 3, 4, 5, 6, 7, 0, },
- ec_gf8_mul_7A_ops
-};
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR2, 1, 0, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_7A = {8,
+ {
+ 1,
+ 2,
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ 0,
+ },
+ ec_gf8_mul_7A_ops};
static ec_gf_op_t ec_gf8_mul_7B_ops[] = {
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR3, 8, 5, 3 },
- { EC_GF_OP_XOR2, 8, 0, 0 },
- { EC_GF_OP_COPY, 9, 4, 0 },
- { EC_GF_OP_XOR2, 8, 2, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 4, 8, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR3, 4, 1, 9 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_7B = {
- 10,
- { 1, 2, 3, 4, 8, 5, 6, 0, 7, 9, },
- ec_gf8_mul_7B_ops
-};
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR3, 8, 5, 3},
+ {EC_GF_OP_XOR2, 8, 0, 0}, {EC_GF_OP_COPY, 9, 4, 0},
+ {EC_GF_OP_XOR2, 8, 2, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 4, 8, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR3, 4, 1, 9},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_7B = {10,
+ {
+ 1,
+ 2,
+ 3,
+ 4,
+ 8,
+ 5,
+ 6,
+ 0,
+ 7,
+ 9,
+ },
+ ec_gf8_mul_7B_ops};
static ec_gf_op_t ec_gf8_mul_7C_ops[] = {
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_7C = {
- 8,
- { 2, 4, 1, 6, 3, 5, 7, 0, },
- ec_gf8_mul_7C_ops
-};
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_7C = {8,
+ {
+ 2,
+ 4,
+ 1,
+ 6,
+ 3,
+ 5,
+ 7,
+ 0,
+ },
+ ec_gf8_mul_7C_ops};
static ec_gf_op_t ec_gf8_mul_7D_ops[] = {
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_7D = {
- 8,
- { 1, 0, 3, 5, 6, 7, 2, 4, },
- ec_gf8_mul_7D_ops
-};
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 1, 0, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_7D = {8,
+ {
+ 1,
+ 0,
+ 3,
+ 5,
+ 6,
+ 7,
+ 2,
+ 4,
+ },
+ ec_gf8_mul_7D_ops};
static ec_gf_op_t ec_gf8_mul_7E_ops[] = {
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR3, 6, 2, 7 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 6, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_7E = {
- 9,
- { 5, 1, 2, 0, 7, 3, 4, 6, 8, },
- ec_gf8_mul_7E_ops
-};
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_COPY, 8, 0, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR3, 6, 2, 7},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 6, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_7E = {9,
+ {
+ 5,
+ 1,
+ 2,
+ 0,
+ 7,
+ 3,
+ 4,
+ 6,
+ 8,
+ },
+ ec_gf8_mul_7E_ops};
static ec_gf_op_t ec_gf8_mul_7F_ops[] = {
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR3, 9, 7, 5 },
- { EC_GF_OP_XOR2, 2, 9, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 6, 9, 0 },
- { EC_GF_OP_XOR3, 9, 6, 4 },
- { EC_GF_OP_XOR2, 7, 9, 0 },
- { EC_GF_OP_XOR2, 3, 9, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_7F = {
- 10,
- { 4, 1, 0, 5, 6, 7, 2, 3, 8, 9, },
- ec_gf8_mul_7F_ops
-};
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR3, 9, 7, 5}, {EC_GF_OP_XOR2, 2, 9, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 6, 9, 0},
+ {EC_GF_OP_XOR3, 9, 6, 4}, {EC_GF_OP_XOR2, 7, 9, 0},
+ {EC_GF_OP_XOR2, 3, 9, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 7, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_7F = {10,
+ {
+ 4,
+ 1,
+ 0,
+ 5,
+ 6,
+ 7,
+ 2,
+ 3,
+ 8,
+ 9,
+ },
+ ec_gf8_mul_7F_ops};
static ec_gf_op_t ec_gf8_mul_80_ops[] = {
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_80 = {
- 8,
- { 7, 5, 6, 4, 1, 2, 3, 0, },
- ec_gf8_mul_80_ops
-};
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_80 = {8,
+ {
+ 7,
+ 5,
+ 6,
+ 4,
+ 1,
+ 2,
+ 3,
+ 0,
+ },
+ ec_gf8_mul_80_ops};
static ec_gf_op_t ec_gf8_mul_81_ops[] = {
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_81 = {
- 8,
- { 2, 7, 4, 1, 5, 6, 3, 0, },
- ec_gf8_mul_81_ops
-};
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_81 = {8,
+ {
+ 2,
+ 7,
+ 4,
+ 1,
+ 5,
+ 6,
+ 3,
+ 0,
+ },
+ ec_gf8_mul_81_ops};
static ec_gf_op_t ec_gf8_mul_82_ops[] = {
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_COPY, 8, 6, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR3, 5, 8, 7 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_82 = {
- 9,
- { 6, 2, 7, 5, 1, 3, 4, 0, 8, },
- ec_gf8_mul_82_ops
-};
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_COPY, 8, 6, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR3, 5, 8, 7}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_82 = {9,
+ {
+ 6,
+ 2,
+ 7,
+ 5,
+ 1,
+ 3,
+ 4,
+ 0,
+ 8,
+ },
+ ec_gf8_mul_82_ops};
static ec_gf_op_t ec_gf8_mul_83_ops[] = {
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_83 = {
- 8,
- { 3, 5, 6, 7, 1, 2, 4, 0, },
- ec_gf8_mul_83_ops
-};
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_83 = {8,
+ {
+ 3,
+ 5,
+ 6,
+ 7,
+ 1,
+ 2,
+ 4,
+ 0,
+ },
+ ec_gf8_mul_83_ops};
static ec_gf_op_t ec_gf8_mul_84_ops[] = {
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_84 = {
- 8,
- { 7, 6, 0, 4, 1, 5, 3, 2, },
- ec_gf8_mul_84_ops
-};
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_84 = {8,
+ {
+ 7,
+ 6,
+ 0,
+ 4,
+ 1,
+ 5,
+ 3,
+ 2,
+ },
+ ec_gf8_mul_84_ops};
static ec_gf_op_t ec_gf8_mul_85_ops[] = {
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_85 = {
- 8,
- { 7, 6, 0, 3, 2, 4, 5, 1, },
- ec_gf8_mul_85_ops
-};
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_85 = {8,
+ {
+ 7,
+ 6,
+ 0,
+ 3,
+ 2,
+ 4,
+ 5,
+ 1,
+ },
+ ec_gf8_mul_85_ops};
static ec_gf_op_t ec_gf8_mul_86_ops[] = {
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_86 = {
- 8,
- { 1, 2, 6, 4, 5, 7, 3, 0, },
- ec_gf8_mul_86_ops
-};
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_86 = {8,
+ {
+ 1,
+ 2,
+ 6,
+ 4,
+ 5,
+ 7,
+ 3,
+ 0,
+ },
+ ec_gf8_mul_86_ops};
static ec_gf_op_t ec_gf8_mul_87_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_COPY, 8, 1, 0 },
- { EC_GF_OP_XOR2, 8, 6, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR3, 5, 8, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 2, 8, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_87 = {
- 9,
- { 1, 2, 3, 4, 5, 7, 6, 0, 8, },
- ec_gf8_mul_87_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_COPY, 8, 1, 0},
+ {EC_GF_OP_XOR2, 8, 6, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR3, 5, 8, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 2, 8, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_87 = {9,
+ {
+ 1,
+ 2,
+ 3,
+ 4,
+ 5,
+ 7,
+ 6,
+ 0,
+ 8,
+ },
+ ec_gf8_mul_87_ops};
static ec_gf_op_t ec_gf8_mul_88_ops[] = {
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_88 = {
- 8,
- { 6, 7, 3, 1, 2, 4, 5, 0, },
- ec_gf8_mul_88_ops
-};
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 1, 0, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_88 = {8,
+ {
+ 6,
+ 7,
+ 3,
+ 1,
+ 2,
+ 4,
+ 5,
+ 0,
+ },
+ ec_gf8_mul_88_ops};
static ec_gf_op_t ec_gf8_mul_89_ops[] = {
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR3, 8, 5, 2 },
- { EC_GF_OP_XOR2, 4, 8, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 0, 8, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_89 = {
- 9,
- { 2, 1, 6, 5, 7, 3, 4, 0, 8, },
- ec_gf8_mul_89_ops
-};
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR3, 8, 5, 2},
+ {EC_GF_OP_XOR2, 4, 8, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 0, 8, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_89 = {9,
+ {
+ 2,
+ 1,
+ 6,
+ 5,
+ 7,
+ 3,
+ 4,
+ 0,
+ 8,
+ },
+ ec_gf8_mul_89_ops};
static ec_gf_op_t ec_gf8_mul_8A_ops[] = {
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_8A = {
- 8,
- { 1, 2, 3, 0, 6, 7, 4, 5, },
- ec_gf8_mul_8A_ops
-};
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_8A = {8,
+ {
+ 1,
+ 2,
+ 3,
+ 0,
+ 6,
+ 7,
+ 4,
+ 5,
+ },
+ ec_gf8_mul_8A_ops};
static ec_gf_op_t ec_gf8_mul_8B_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_8B = {
- 8,
- { 6, 1, 2, 3, 5, 7, 4, 0, },
- ec_gf8_mul_8B_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_8B = {8,
+ {
+ 6,
+ 1,
+ 2,
+ 3,
+ 5,
+ 7,
+ 4,
+ 0,
+ },
+ ec_gf8_mul_8B_ops};
static ec_gf_op_t ec_gf8_mul_8C_ops[] = {
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_8C = {
- 8,
- { 1, 2, 0, 7, 3, 4, 5, 6, },
- ec_gf8_mul_8C_ops
-};
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_8C = {8,
+ {
+ 1,
+ 2,
+ 0,
+ 7,
+ 3,
+ 4,
+ 5,
+ 6,
+ },
+ ec_gf8_mul_8C_ops};
static ec_gf_op_t ec_gf8_mul_8D_ops[] = {
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_8D = {
- 8,
- { 7, 1, 3, 2, 4, 5, 0, 6, },
- ec_gf8_mul_8D_ops
-};
-
-static ec_gf_op_t ec_gf8_mul_8E_ops[] = {
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_8E = {
- 8,
- { 1, 2, 3, 4, 5, 6, 7, 0, },
- ec_gf8_mul_8E_ops
-};
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_8D = {8,
+ {
+ 7,
+ 1,
+ 3,
+ 2,
+ 4,
+ 5,
+ 0,
+ 6,
+ },
+ ec_gf8_mul_8D_ops};
+
+static ec_gf_op_t ec_gf8_mul_8E_ops[] = {{EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_8E = {8,
+ {
+ 1,
+ 2,
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ 0,
+ },
+ ec_gf8_mul_8E_ops};
static ec_gf_op_t ec_gf8_mul_8F_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_8F = {
- 8,
- { 1, 2, 3, 4, 5, 6, 7, 0, },
- ec_gf8_mul_8F_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_8F = {8,
+ {
+ 1,
+ 2,
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ 0,
+ },
+ ec_gf8_mul_8F_ops};
static ec_gf_op_t ec_gf8_mul_90_ops[] = {
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_90 = {
- 8,
- { 4, 5, 6, 7, 0, 1, 3, 2, },
- ec_gf8_mul_90_ops
-};
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_90 = {8,
+ {
+ 4,
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ 3,
+ 2,
+ },
+ ec_gf8_mul_90_ops};
static ec_gf_op_t ec_gf8_mul_91_ops[] = {
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_COPY, 9, 1, 0 },
- { EC_GF_OP_COPY, 8, 3, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 7, 9, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR3, 5, 8, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_91 = {
- 10,
- { 2, 3, 1, 4, 0, 6, 7, 5, 8, 9, },
- ec_gf8_mul_91_ops
-};
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_COPY, 9, 1, 0}, {EC_GF_OP_COPY, 8, 3, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 7, 9, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR3, 5, 8, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_91 = {10,
+ {
+ 2,
+ 3,
+ 1,
+ 4,
+ 0,
+ 6,
+ 7,
+ 5,
+ 8,
+ 9,
+ },
+ ec_gf8_mul_91_ops};
static ec_gf_op_t ec_gf8_mul_92_ops[] = {
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_92 = {
- 8,
- { 6, 7, 0, 1, 2, 3, 5, 4, },
- ec_gf8_mul_92_ops
-};
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_92 = {8,
+ {
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ 3,
+ 5,
+ 4,
+ },
+ ec_gf8_mul_92_ops};
static ec_gf_op_t ec_gf8_mul_93_ops[] = {
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_93 = {
- 8,
- { 6, 4, 5, 1, 7, 2, 3, 0, },
- ec_gf8_mul_93_ops
-};
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_93 = {8,
+ {
+ 6,
+ 4,
+ 5,
+ 1,
+ 7,
+ 2,
+ 3,
+ 0,
+ },
+ ec_gf8_mul_93_ops};
static ec_gf_op_t ec_gf8_mul_94_ops[] = {
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_94 = {
- 8,
- { 7, 5, 0, 2, 6, 1, 3, 4, },
- ec_gf8_mul_94_ops
-};
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_94 = {8,
+ {
+ 7,
+ 5,
+ 0,
+ 2,
+ 6,
+ 1,
+ 3,
+ 4,
+ },
+ ec_gf8_mul_94_ops};
static ec_gf_op_t ec_gf8_mul_95_ops[] = {
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_95 = {
- 8,
- { 7, 6, 1, 3, 0, 4, 5, 2, },
- ec_gf8_mul_95_ops
-};
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_95 = {8,
+ {
+ 7,
+ 6,
+ 1,
+ 3,
+ 0,
+ 4,
+ 5,
+ 2,
+ },
+ ec_gf8_mul_95_ops};
static ec_gf_op_t ec_gf8_mul_96_ops[] = {
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR3, 8, 0, 4 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 8, 3, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 5, 8, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_96 = {
- 9,
- { 4, 0, 1, 6, 7, 2, 3, 5, 8, },
- ec_gf8_mul_96_ops
-};
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR3, 8, 0, 4}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 7, 8, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 8, 3, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 1, 8, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 5, 8, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_96 = {9,
+ {
+ 4,
+ 0,
+ 1,
+ 6,
+ 7,
+ 2,
+ 3,
+ 5,
+ 8,
+ },
+ ec_gf8_mul_96_ops};
static ec_gf_op_t ec_gf8_mul_97_ops[] = {
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_COPY, 8, 2, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 8, 6, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 5, 8, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_97 = {
- 9,
- { 4, 5, 3, 6, 7, 1, 2, 0, 8, },
- ec_gf8_mul_97_ops
-};
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_COPY, 8, 2, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 8, 6, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 1, 8, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 5, 8, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_97 = {9,
+ {
+ 4,
+ 5,
+ 3,
+ 6,
+ 7,
+ 1,
+ 2,
+ 0,
+ 8,
+ },
+ ec_gf8_mul_97_ops};
static ec_gf_op_t ec_gf8_mul_98_ops[] = {
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_98 = {
- 8,
- { 4, 2, 3, 6, 7, 5, 1, 0, },
- ec_gf8_mul_98_ops
-};
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_98 = {8,
+ {
+ 4,
+ 2,
+ 3,
+ 6,
+ 7,
+ 5,
+ 1,
+ 0,
+ },
+ ec_gf8_mul_98_ops};
static ec_gf_op_t ec_gf8_mul_99_ops[] = {
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_99 = {
- 8,
- { 6, 5, 3, 7, 0, 1, 4, 2, },
- ec_gf8_mul_99_ops
-};
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_99 = {8,
+ {
+ 6,
+ 5,
+ 3,
+ 7,
+ 0,
+ 1,
+ 4,
+ 2,
+ },
+ ec_gf8_mul_99_ops};
static ec_gf_op_t ec_gf8_mul_9A_ops[] = {
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR3, 8, 4, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_9A = {
- 9,
- { 6, 3, 4, 0, 5, 1, 2, 7, 8, },
- ec_gf8_mul_9A_ops
-};
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR3, 8, 4, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 7, 8, 0}, {EC_GF_OP_XOR2, 1, 8, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_9A = {9,
+ {
+ 6,
+ 3,
+ 4,
+ 0,
+ 5,
+ 1,
+ 2,
+ 7,
+ 8,
+ },
+ ec_gf8_mul_9A_ops};
static ec_gf_op_t ec_gf8_mul_9B_ops[] = {
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_COPY, 9, 5, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR3, 8, 3, 2 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 3, 9, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_9B = {
- 10,
- { 4, 5, 8, 6, 7, 1, 2, 0, 3, 9, },
- ec_gf8_mul_9B_ops
-};
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_COPY, 9, 5, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR3, 8, 3, 2}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 3, 9, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_9B = {10,
+ {
+ 4,
+ 5,
+ 8,
+ 6,
+ 7,
+ 1,
+ 2,
+ 0,
+ 3,
+ 9,
+ },
+ ec_gf8_mul_9B_ops};
static ec_gf_op_t ec_gf8_mul_9C_ops[] = {
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_9C = {
- 8,
- { 3, 2, 1, 0, 4, 5, 6, 7, },
- ec_gf8_mul_9C_ops
-};
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_9C = {8,
+ {
+ 3,
+ 2,
+ 1,
+ 0,
+ 4,
+ 5,
+ 6,
+ 7,
+ },
+ ec_gf8_mul_9C_ops};
static ec_gf_op_t ec_gf8_mul_9D_ops[] = {
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_9D = {
- 8,
- { 0, 1, 2, 3, 7, 4, 5, 6, },
- ec_gf8_mul_9D_ops
-};
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_9D = {8,
+ {
+ 0,
+ 1,
+ 2,
+ 3,
+ 7,
+ 4,
+ 5,
+ 6,
+ },
+ ec_gf8_mul_9D_ops};
static ec_gf_op_t ec_gf8_mul_9E_ops[] = {
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_COPY, 8, 7, 0 },
- { EC_GF_OP_XOR2, 8, 5, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 0, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_9E = {
- 9,
- { 4, 5, 3, 8, 6, 0, 2, 7, 1, },
- ec_gf8_mul_9E_ops
-};
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_COPY, 8, 7, 0},
+ {EC_GF_OP_XOR2, 8, 5, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 0, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_9E = {9,
+ {
+ 4,
+ 5,
+ 3,
+ 8,
+ 6,
+ 0,
+ 2,
+ 7,
+ 1,
+ },
+ ec_gf8_mul_9E_ops};
static ec_gf_op_t ec_gf8_mul_9F_ops[] = {
- { EC_GF_OP_XOR3, 8, 1, 2 },
- { EC_GF_OP_XOR2, 8, 3, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 6, 8, 0 },
- { EC_GF_OP_XOR2, 5, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_9F = {
- 9,
- { 4, 5, 6, 7, 0, 1, 2, 3, 8, },
- ec_gf8_mul_9F_ops
-};
+ {EC_GF_OP_XOR3, 8, 1, 2}, {EC_GF_OP_XOR2, 8, 3, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 6, 8, 0}, {EC_GF_OP_XOR2, 5, 8, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_9F = {9,
+ {
+ 4,
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ 3,
+ 8,
+ },
+ ec_gf8_mul_9F_ops};
static ec_gf_op_t ec_gf8_mul_A0_ops[] = {
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_A0 = {
- 8,
- { 3, 1, 6, 7, 5, 2, 4, 0, },
- ec_gf8_mul_A0_ops
-};
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_A0 = {8,
+ {
+ 3,
+ 1,
+ 6,
+ 7,
+ 5,
+ 2,
+ 4,
+ 0,
+ },
+ ec_gf8_mul_A0_ops};
static ec_gf_op_t ec_gf8_mul_A1_ops[] = {
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR3, 8, 0, 6 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_XOR2, 3, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_A1 = {
- 9,
- { 7, 4, 1, 5, 6, 0, 2, 3, 8, },
- ec_gf8_mul_A1_ops
-};
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR3, 8, 0, 6},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 7, 8, 0},
+ {EC_GF_OP_XOR2, 3, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_A1 = {9,
+ {
+ 7,
+ 4,
+ 1,
+ 5,
+ 6,
+ 0,
+ 2,
+ 3,
+ 8,
+ },
+ ec_gf8_mul_A1_ops};
static ec_gf_op_t ec_gf8_mul_A2_ops[] = {
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_A2 = {
- 8,
- { 7, 0, 6, 3, 2, 1, 4, 5, },
- ec_gf8_mul_A2_ops
-};
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 4, 7, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_A2 = {8,
+ {
+ 7,
+ 0,
+ 6,
+ 3,
+ 2,
+ 1,
+ 4,
+ 5,
+ },
+ ec_gf8_mul_A2_ops};
static ec_gf_op_t ec_gf8_mul_A3_ops[] = {
- { EC_GF_OP_COPY, 8, 2, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 3, 8, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_A3 = {
- 9,
- { 3, 7, 2, 6, 1, 4, 0, 5, 8, },
- ec_gf8_mul_A3_ops
-};
+ {EC_GF_OP_COPY, 8, 2, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 3, 8, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_A3 = {9,
+ {
+ 3,
+ 7,
+ 2,
+ 6,
+ 1,
+ 4,
+ 0,
+ 5,
+ 8,
+ },
+ ec_gf8_mul_A3_ops};
static ec_gf_op_t ec_gf8_mul_A4_ops[] = {
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_A4 = {
- 8,
- { 5, 6, 7, 2, 4, 3, 0, 1, },
- ec_gf8_mul_A4_ops
-};
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_A4 = {8,
+ {
+ 5,
+ 6,
+ 7,
+ 2,
+ 4,
+ 3,
+ 0,
+ 1,
+ },
+ ec_gf8_mul_A4_ops};
static ec_gf_op_t ec_gf8_mul_A5_ops[] = {
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR3, 8, 5, 6 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 4, 8, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_A5 = {
- 9,
- { 1, 4, 2, 5, 6, 7, 3, 0, 8, },
- ec_gf8_mul_A5_ops
-};
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR3, 8, 5, 6}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 4, 8, 0}, {EC_GF_OP_XOR2, 7, 8, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_A5 = {9,
+ {
+ 1,
+ 4,
+ 2,
+ 5,
+ 6,
+ 7,
+ 3,
+ 0,
+ 8,
+ },
+ ec_gf8_mul_A5_ops};
static ec_gf_op_t ec_gf8_mul_A6_ops[] = {
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_A6 = {
- 8,
- { 1, 2, 0, 3, 4, 5, 6, 7, },
- ec_gf8_mul_A6_ops
-};
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_A6 = {8,
+ {
+ 1,
+ 2,
+ 0,
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ },
+ ec_gf8_mul_A6_ops};
static ec_gf_op_t ec_gf8_mul_A7_ops[] = {
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_A7 = {
- 8,
- { 0, 1, 2, 5, 6, 7, 3, 4, },
- ec_gf8_mul_A7_ops
-};
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_A7 = {8,
+ {
+ 0,
+ 1,
+ 2,
+ 5,
+ 6,
+ 7,
+ 3,
+ 4,
+ },
+ ec_gf8_mul_A7_ops};
static ec_gf_op_t ec_gf8_mul_A8_ops[] = {
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 8, 1, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_COPY, 9, 4, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 8, 3, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 9, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_A8 = {
- 10,
- { 1, 7, 5, 8, 6, 3, 4, 0, 2, 9, },
- ec_gf8_mul_A8_ops
-};
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 8, 1, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_COPY, 9, 4, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 8, 3, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 2, 9, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_A8 = {10,
+ {
+ 1,
+ 7,
+ 5,
+ 8,
+ 6,
+ 3,
+ 4,
+ 0,
+ 2,
+ 9,
+ },
+ ec_gf8_mul_A8_ops};
static ec_gf_op_t ec_gf8_mul_A9_ops[] = {
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_A9 = {
- 8,
- { 3, 7, 6, 1, 2, 0, 4, 5, },
- ec_gf8_mul_A9_ops
-};
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 1, 0, 0},
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_A9 = {8,
+ {
+ 3,
+ 7,
+ 6,
+ 1,
+ 2,
+ 0,
+ 4,
+ 5,
+ },
+ ec_gf8_mul_A9_ops};
static ec_gf_op_t ec_gf8_mul_AA_ops[] = {
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_AA = {
- 8,
- { 0, 4, 5, 3, 6, 7, 1, 2, },
- ec_gf8_mul_AA_ops
-};
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_AA = {8,
+ {
+ 0,
+ 4,
+ 5,
+ 3,
+ 6,
+ 7,
+ 1,
+ 2,
+ },
+ ec_gf8_mul_AA_ops};
static ec_gf_op_t ec_gf8_mul_AB_ops[] = {
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_COPY, 9, 6, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 8, 7, 0 },
- { EC_GF_OP_XOR2, 3, 8, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR3, 3, 9, 7 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_AB = {
- 10,
- { 2, 3, 8, 0, 5, 6, 1, 4, 7, 9, },
- ec_gf8_mul_AB_ops
-};
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_COPY, 9, 6, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_XOR2, 8, 7, 0}, {EC_GF_OP_XOR2, 3, 8, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR3, 3, 9, 7},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_AB = {10,
+ {
+ 2,
+ 3,
+ 8,
+ 0,
+ 5,
+ 6,
+ 1,
+ 4,
+ 7,
+ 9,
+ },
+ ec_gf8_mul_AB_ops};
static ec_gf_op_t ec_gf8_mul_AC_ops[] = {
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_AC = {
- 8,
- { 3, 2, 1, 0, 4, 5, 6, 7, },
- ec_gf8_mul_AC_ops
-};
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_AC = {8,
+ {
+ 3,
+ 2,
+ 1,
+ 0,
+ 4,
+ 5,
+ 6,
+ 7,
+ },
+ ec_gf8_mul_AC_ops};
static ec_gf_op_t ec_gf8_mul_AD_ops[] = {
- { EC_GF_OP_XOR3, 8, 1, 2 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 4, 8, 0 },
- { EC_GF_OP_XOR2, 5, 8, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_AD = {
- 9,
- { 3, 4, 5, 6, 7, 0, 1, 2, 8, },
- ec_gf8_mul_AD_ops
-};
+ {EC_GF_OP_XOR3, 8, 1, 2}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 4, 8, 0},
+ {EC_GF_OP_XOR2, 5, 8, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_AD = {9,
+ {
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ 8,
+ },
+ ec_gf8_mul_AD_ops};
static ec_gf_op_t ec_gf8_mul_AE_ops[] = {
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_COPY, 8, 5, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 4, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_AE = {
- 9,
- { 7, 0, 5, 6, 3, 4, 1, 2, 8, },
- ec_gf8_mul_AE_ops
-};
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_COPY, 8, 5, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 4, 8, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_AE = {9,
+ {
+ 7,
+ 0,
+ 5,
+ 6,
+ 3,
+ 4,
+ 1,
+ 2,
+ 8,
+ },
+ ec_gf8_mul_AE_ops};
static ec_gf_op_t ec_gf8_mul_AF_ops[] = {
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_AF = {
- 8,
- { 0, 1, 2, 7, 3, 4, 5, 6, },
- ec_gf8_mul_AF_ops
-};
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_AF = {8,
+ {
+ 0,
+ 1,
+ 2,
+ 7,
+ 3,
+ 4,
+ 5,
+ 6,
+ },
+ ec_gf8_mul_AF_ops};
static ec_gf_op_t ec_gf8_mul_B0_ops[] = {
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_B0 = {
- 8,
- { 4, 0, 7, 2, 3, 1, 6, 5, },
- ec_gf8_mul_B0_ops
-};
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 1, 0, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_B0 = {8,
+ {
+ 4,
+ 0,
+ 7,
+ 2,
+ 3,
+ 1,
+ 6,
+ 5,
+ },
+ ec_gf8_mul_B0_ops};
static ec_gf_op_t ec_gf8_mul_B1_ops[] = {
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_COPY, 8, 4, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR3, 5, 8, 1 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_B1 = {
- 9,
- { 2, 6, 4, 7, 0, 1, 3, 5, 8, },
- ec_gf8_mul_B1_ops
-};
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_COPY, 8, 4, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR3, 5, 8, 1}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_B1 = {9,
+ {
+ 2,
+ 6,
+ 4,
+ 7,
+ 0,
+ 1,
+ 3,
+ 5,
+ 8,
+ },
+ ec_gf8_mul_B1_ops};
static ec_gf_op_t ec_gf8_mul_B2_ops[] = {
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR3, 8, 4, 5 },
- { EC_GF_OP_XOR2, 2, 8, 0 },
- { EC_GF_OP_XOR2, 8, 1, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_XOR2, 3, 8, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_B2 = {
- 9,
- { 0, 7, 4, 5, 6, 1, 2, 3, 8, },
- ec_gf8_mul_B2_ops
-};
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR3, 8, 4, 5},
+ {EC_GF_OP_XOR2, 2, 8, 0}, {EC_GF_OP_XOR2, 8, 1, 0},
+ {EC_GF_OP_XOR2, 7, 8, 0}, {EC_GF_OP_XOR2, 3, 8, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_B2 = {9,
+ {
+ 0,
+ 7,
+ 4,
+ 5,
+ 6,
+ 1,
+ 2,
+ 3,
+ 8,
+ },
+ ec_gf8_mul_B2_ops};
static ec_gf_op_t ec_gf8_mul_B3_ops[] = {
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_COPY, 9, 5, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR3, 8, 6, 4 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 8, 5, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_XOR2, 0, 8, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR3, 1, 9, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_B3 = {
- 10,
- { 2, 3, 4, 5, 1, 6, 0, 7, 8, 9, },
- ec_gf8_mul_B3_ops
-};
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_COPY, 9, 5, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR3, 8, 6, 4},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 8, 5, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 7, 8, 0},
+ {EC_GF_OP_XOR2, 0, 8, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_XOR3, 1, 9, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_B3 = {10,
+ {
+ 2,
+ 3,
+ 4,
+ 5,
+ 1,
+ 6,
+ 0,
+ 7,
+ 8,
+ 9,
+ },
+ ec_gf8_mul_B3_ops};
static ec_gf_op_t ec_gf8_mul_B4_ops[] = {
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_B4 = {
- 8,
- { 5, 6, 7, 0, 1, 2, 3, 4, },
- ec_gf8_mul_B4_ops
-};
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_B4 = {8,
+ {
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ 3,
+ 4,
+ },
+ ec_gf8_mul_B4_ops};
static ec_gf_op_t ec_gf8_mul_B5_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_COPY, 8, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR3, 4, 8, 3 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_B5 = {
- 9,
- { 3, 4, 0, 7, 1, 5, 6, 2, 8, },
- ec_gf8_mul_B5_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_COPY, 8, 6, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR3, 4, 8, 3}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_B5 = {9,
+ {
+ 3,
+ 4,
+ 0,
+ 7,
+ 1,
+ 5,
+ 6,
+ 2,
+ 8,
+ },
+ ec_gf8_mul_B5_ops};
static ec_gf_op_t ec_gf8_mul_B6_ops[] = {
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_B6 = {
- 8,
- { 5, 3, 6, 4, 7, 0, 1, 2, },
- ec_gf8_mul_B6_ops
-};
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_B6 = {8,
+ {
+ 5,
+ 3,
+ 6,
+ 4,
+ 7,
+ 0,
+ 1,
+ 2,
+ },
+ ec_gf8_mul_B6_ops};
static ec_gf_op_t ec_gf8_mul_B7_ops[] = {
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_B7 = {
- 8,
- { 5, 0, 1, 4, 2, 6, 7, 3, },
- ec_gf8_mul_B7_ops
-};
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_B7 = {8,
+ {
+ 5,
+ 0,
+ 1,
+ 4,
+ 2,
+ 6,
+ 7,
+ 3,
+ },
+ ec_gf8_mul_B7_ops};
static ec_gf_op_t ec_gf8_mul_B8_ops[] = {
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_B8 = {
- 8,
- { 6, 4, 5, 1, 2, 0, 7, 3, },
- ec_gf8_mul_B8_ops
-};
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 4, 7, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_B8 = {8,
+ {
+ 6,
+ 4,
+ 5,
+ 1,
+ 2,
+ 0,
+ 7,
+ 3,
+ },
+ ec_gf8_mul_B8_ops};
static ec_gf_op_t ec_gf8_mul_B9_ops[] = {
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR3, 0, 8, 2 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_B9 = {
- 9,
- { 6, 7, 0, 2, 1, 4, 5, 3, 8, },
- ec_gf8_mul_B9_ops
-};
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR3, 0, 8, 2}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_B9 = {9,
+ {
+ 6,
+ 7,
+ 0,
+ 2,
+ 1,
+ 4,
+ 5,
+ 3,
+ 8,
+ },
+ ec_gf8_mul_B9_ops};
static ec_gf_op_t ec_gf8_mul_BA_ops[] = {
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_BA = {
- 8,
- { 1, 2, 4, 3, 5, 6, 0, 7, },
- ec_gf8_mul_BA_ops
-};
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_BA = {8,
+ {
+ 1,
+ 2,
+ 4,
+ 3,
+ 5,
+ 6,
+ 0,
+ 7,
+ },
+ ec_gf8_mul_BA_ops};
static ec_gf_op_t ec_gf8_mul_BB_ops[] = {
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_COPY, 8, 3, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 8, 5, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 8, 7, 0 },
- { EC_GF_OP_XOR2, 2, 8, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_BB = {
- 9,
- { 7, 2, 1, 8, 3, 5, 6, 4, 0, },
- ec_gf8_mul_BB_ops
-};
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_COPY, 8, 3, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 8, 5, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 8, 7, 0}, {EC_GF_OP_XOR2, 2, 8, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_BB = {9,
+ {
+ 7,
+ 2,
+ 1,
+ 8,
+ 3,
+ 5,
+ 6,
+ 4,
+ 0,
+ },
+ ec_gf8_mul_BB_ops};
static ec_gf_op_t ec_gf8_mul_BC_ops[] = {
- { EC_GF_OP_COPY, 8, 1, 0 },
- { EC_GF_OP_XOR2, 8, 2, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_XOR3, 2, 8, 4 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_BC = {
- 9,
- { 2, 6, 3, 4, 5, 1, 7, 0, 8, },
- ec_gf8_mul_BC_ops
-};
+ {EC_GF_OP_COPY, 8, 1, 0}, {EC_GF_OP_XOR2, 8, 2, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 7, 8, 0}, {EC_GF_OP_XOR3, 2, 8, 4},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_BC = {9,
+ {
+ 2,
+ 6,
+ 3,
+ 4,
+ 5,
+ 1,
+ 7,
+ 0,
+ 8,
+ },
+ ec_gf8_mul_BC_ops};
static ec_gf_op_t ec_gf8_mul_BD_ops[] = {
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_BD = {
- 8,
- { 4, 5, 0, 2, 7, 1, 6, 3, },
- ec_gf8_mul_BD_ops
-};
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_BD = {8,
+ {
+ 4,
+ 5,
+ 0,
+ 2,
+ 7,
+ 1,
+ 6,
+ 3,
+ },
+ ec_gf8_mul_BD_ops};
static ec_gf_op_t ec_gf8_mul_BE_ops[] = {
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_BE = {
- 8,
- { 0, 6, 7, 4, 5, 1, 3, 2, },
- ec_gf8_mul_BE_ops
-};
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_BE = {8,
+ {
+ 0,
+ 6,
+ 7,
+ 4,
+ 5,
+ 1,
+ 3,
+ 2,
+ },
+ ec_gf8_mul_BE_ops};
static ec_gf_op_t ec_gf8_mul_BF_ops[] = {
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_BF = {
- 8,
- { 5, 6, 1, 7, 3, 0, 2, 4, },
- ec_gf8_mul_BF_ops
-};
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_BF = {8,
+ {
+ 5,
+ 6,
+ 1,
+ 7,
+ 3,
+ 0,
+ 2,
+ 4,
+ },
+ ec_gf8_mul_BF_ops};
static ec_gf_op_t ec_gf8_mul_C0_ops[] = {
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_C0 = {
- 8,
- { 1, 2, 3, 4, 7, 5, 6, 0, },
- ec_gf8_mul_C0_ops
-};
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_C0 = {8,
+ {
+ 1,
+ 2,
+ 3,
+ 4,
+ 7,
+ 5,
+ 6,
+ 0,
+ },
+ ec_gf8_mul_C0_ops};
static ec_gf_op_t ec_gf8_mul_C1_ops[] = {
- { EC_GF_OP_XOR3, 8, 1, 2 },
- { EC_GF_OP_XOR2, 8, 3, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 6, 8, 0 },
- { EC_GF_OP_XOR2, 5, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_C1 = {
- 9,
- { 5, 6, 7, 4, 1, 2, 3, 0, 8, },
- ec_gf8_mul_C1_ops
-};
+ {EC_GF_OP_XOR3, 8, 1, 2}, {EC_GF_OP_XOR2, 8, 3, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 6, 8, 0}, {EC_GF_OP_XOR2, 5, 8, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_C1 = {9,
+ {
+ 5,
+ 6,
+ 7,
+ 4,
+ 1,
+ 2,
+ 3,
+ 0,
+ 8,
+ },
+ ec_gf8_mul_C1_ops};
static ec_gf_op_t ec_gf8_mul_C2_ops[] = {
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_C2 = {
- 8,
- { 7, 6, 3, 0, 1, 4, 5, 2, },
- ec_gf8_mul_C2_ops
-};
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_C2 = {8,
+ {
+ 7,
+ 6,
+ 3,
+ 0,
+ 1,
+ 4,
+ 5,
+ 2,
+ },
+ ec_gf8_mul_C2_ops};
static ec_gf_op_t ec_gf8_mul_C3_ops[] = {
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR3, 0, 2, 6 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR3, 9, 1, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 7, 9, 0 },
- { EC_GF_OP_XOR2, 3, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_C3 = {
- 10,
- { 5, 6, 4, 7, 1, 2, 3, 0, 8, 9, },
- ec_gf8_mul_C3_ops
-};
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_XOR3, 0, 2, 6}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR3, 9, 1, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 7, 9, 0},
+ {EC_GF_OP_XOR2, 3, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_C3 = {10,
+ {
+ 5,
+ 6,
+ 4,
+ 7,
+ 1,
+ 2,
+ 3,
+ 0,
+ 8,
+ 9,
+ },
+ ec_gf8_mul_C3_ops};
static ec_gf_op_t ec_gf8_mul_C4_ops[] = {
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_C4 = {
- 8,
- { 0, 2, 1, 3, 4, 5, 6, 7, },
- ec_gf8_mul_C4_ops
-};
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 1, 0, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_C4 = {8,
+ {
+ 0,
+ 2,
+ 1,
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ },
+ ec_gf8_mul_C4_ops};
static ec_gf_op_t ec_gf8_mul_C5_ops[] = {
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_C5 = {
- 8,
- { 4, 3, 5, 7, 6, 2, 0, 1, },
- ec_gf8_mul_C5_ops
-};
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_C5 = {8,
+ {
+ 4,
+ 3,
+ 5,
+ 7,
+ 6,
+ 2,
+ 0,
+ 1,
+ },
+ ec_gf8_mul_C5_ops};
static ec_gf_op_t ec_gf8_mul_C6_ops[] = {
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_COPY, 8, 4, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR3, 9, 5, 4 },
- { EC_GF_OP_XOR2, 6, 9, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 7, 9, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 6, 8, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_C6 = {
- 10,
- { 6, 3, 0, 4, 5, 7, 2, 1, 8, 9, },
- ec_gf8_mul_C6_ops
-};
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_COPY, 8, 4, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR3, 9, 5, 4},
+ {EC_GF_OP_XOR2, 6, 9, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 7, 9, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 6, 8, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_C6 = {10,
+ {
+ 6,
+ 3,
+ 0,
+ 4,
+ 5,
+ 7,
+ 2,
+ 1,
+ 8,
+ 9,
+ },
+ ec_gf8_mul_C6_ops};
static ec_gf_op_t ec_gf8_mul_C7_ops[] = {
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_C7 = {
- 8,
- { 7, 0, 6, 2, 5, 3, 4, 1, },
- ec_gf8_mul_C7_ops
-};
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_C7 = {8,
+ {
+ 7,
+ 0,
+ 6,
+ 2,
+ 5,
+ 3,
+ 4,
+ 1,
+ },
+ ec_gf8_mul_C7_ops};
static ec_gf_op_t ec_gf8_mul_C8_ops[] = {
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_C8 = {
- 8,
- { 1, 3, 2, 4, 6, 7, 5, 0, },
- ec_gf8_mul_C8_ops
-};
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_C8 = {8,
+ {
+ 1,
+ 3,
+ 2,
+ 4,
+ 6,
+ 7,
+ 5,
+ 0,
+ },
+ ec_gf8_mul_C8_ops};
static ec_gf_op_t ec_gf8_mul_C9_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_C9 = {
- 8,
- { 2, 3, 4, 5, 6, 7, 0, 1, },
- ec_gf8_mul_C9_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_C9 = {8,
+ {
+ 2,
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ },
+ ec_gf8_mul_C9_ops};
static ec_gf_op_t ec_gf8_mul_CA_ops[] = {
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_CA = {
- 8,
- { 1, 2, 5, 7, 3, 4, 0, 6, },
- ec_gf8_mul_CA_ops
-};
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 4, 5, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_CA = {8,
+ {
+ 1,
+ 2,
+ 5,
+ 7,
+ 3,
+ 4,
+ 0,
+ 6,
+ },
+ ec_gf8_mul_CA_ops};
static ec_gf_op_t ec_gf8_mul_CB_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_CB = {
- 8,
- { 2, 3, 4, 5, 7, 6, 0, 1, },
- ec_gf8_mul_CB_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_CB = {8,
+ {
+ 2,
+ 3,
+ 4,
+ 5,
+ 7,
+ 6,
+ 0,
+ 1,
+ },
+ ec_gf8_mul_CB_ops};
static ec_gf_op_t ec_gf8_mul_CC_ops[] = {
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_CC = {
- 8,
- { 2, 7, 1, 0, 5, 6, 3, 4, },
- ec_gf8_mul_CC_ops
-};
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_CC = {8,
+ {
+ 2,
+ 7,
+ 1,
+ 0,
+ 5,
+ 6,
+ 3,
+ 4,
+ },
+ ec_gf8_mul_CC_ops};
static ec_gf_op_t ec_gf8_mul_CD_ops[] = {
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_CD = {
- 8,
- { 0, 6, 1, 2, 7, 3, 4, 5, },
- ec_gf8_mul_CD_ops
-};
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 4, 7, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_CD = {8,
+ {
+ 0,
+ 6,
+ 1,
+ 2,
+ 7,
+ 3,
+ 4,
+ 5,
+ },
+ ec_gf8_mul_CD_ops};
static ec_gf_op_t ec_gf8_mul_CE_ops[] = {
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_COPY, 8, 7, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR3, 3, 6, 8 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR3, 8, 2, 3 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_XOR2, 4, 8, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_CE = {
- 9,
- { 5, 7, 3, 0, 2, 6, 4, 1, 8, },
- ec_gf8_mul_CE_ops
-};
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_COPY, 8, 7, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR3, 3, 6, 8},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR3, 8, 2, 3},
+ {EC_GF_OP_XOR2, 1, 8, 0}, {EC_GF_OP_XOR2, 4, 8, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_CE = {9,
+ {
+ 5,
+ 7,
+ 3,
+ 0,
+ 2,
+ 6,
+ 4,
+ 1,
+ 8,
+ },
+ ec_gf8_mul_CE_ops};
static ec_gf_op_t ec_gf8_mul_CF_ops[] = {
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_CF = {
- 8,
- { 3, 6, 7, 0, 2, 4, 5, 1, },
- ec_gf8_mul_CF_ops
-};
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_CF = {8,
+ {
+ 3,
+ 6,
+ 7,
+ 0,
+ 2,
+ 4,
+ 5,
+ 1,
+ },
+ ec_gf8_mul_CF_ops};
static ec_gf_op_t ec_gf8_mul_D0_ops[] = {
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_D0 = {
- 8,
- { 5, 6, 7, 2, 0, 3, 1, 4, },
- ec_gf8_mul_D0_ops
-};
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_D0 = {8,
+ {
+ 5,
+ 6,
+ 7,
+ 2,
+ 0,
+ 3,
+ 1,
+ 4,
+ },
+ ec_gf8_mul_D0_ops};
static ec_gf_op_t ec_gf8_mul_D1_ops[] = {
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR3, 8, 6, 0 },
- { EC_GF_OP_XOR2, 4, 8, 0 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_D1 = {
- 9,
- { 5, 6, 3, 2, 0, 7, 4, 1, 8, },
- ec_gf8_mul_D1_ops
-};
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR3, 8, 6, 0},
+ {EC_GF_OP_XOR2, 4, 8, 0}, {EC_GF_OP_XOR2, 1, 8, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_D1 = {9,
+ {
+ 5,
+ 6,
+ 3,
+ 2,
+ 0,
+ 7,
+ 4,
+ 1,
+ 8,
+ },
+ ec_gf8_mul_D1_ops};
static ec_gf_op_t ec_gf8_mul_D2_ops[] = {
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_D2 = {
- 8,
- { 7, 0, 2, 1, 3, 4, 6, 5, },
- ec_gf8_mul_D2_ops
-};
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_D2 = {8,
+ {
+ 7,
+ 0,
+ 2,
+ 1,
+ 3,
+ 4,
+ 6,
+ 5,
+ },
+ ec_gf8_mul_D2_ops};
static ec_gf_op_t ec_gf8_mul_D3_ops[] = {
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_COPY, 8, 4, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 8, 6, 0 },
- { EC_GF_OP_XOR2, 3, 8, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_D3 = {
- 9,
- { 0, 3, 2, 8, 4, 6, 7, 1, 5, },
- ec_gf8_mul_D3_ops
-};
+ {EC_GF_OP_XOR2, 4, 7, 0}, {EC_GF_OP_COPY, 8, 4, 0},
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 8, 6, 0}, {EC_GF_OP_XOR2, 3, 8, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 1, 3, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 4, 7, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_D3 = {9,
+ {
+ 0,
+ 3,
+ 2,
+ 8,
+ 4,
+ 6,
+ 7,
+ 1,
+ 5,
+ },
+ ec_gf8_mul_D3_ops};
static ec_gf_op_t ec_gf8_mul_D4_ops[] = {
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_COPY, 8, 1, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR3, 1, 7, 8 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_D4 = {
- 9,
- { 4, 1, 7, 5, 0, 6, 3, 2, 8, },
- ec_gf8_mul_D4_ops
-};
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_COPY, 8, 1, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR3, 1, 7, 8},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_D4 = {9,
+ {
+ 4,
+ 1,
+ 7,
+ 5,
+ 0,
+ 6,
+ 3,
+ 2,
+ 8,
+ },
+ ec_gf8_mul_D4_ops};
static ec_gf_op_t ec_gf8_mul_D5_ops[] = {
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_D5 = {
- 8,
- { 6, 7, 4, 5, 2, 3, 1, 0, },
- ec_gf8_mul_D5_ops
-};
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_D5 = {8,
+ {
+ 6,
+ 7,
+ 4,
+ 5,
+ 2,
+ 3,
+ 1,
+ 0,
+ },
+ ec_gf8_mul_D5_ops};
static ec_gf_op_t ec_gf8_mul_D6_ops[] = {
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_D6 = {
- 9,
- { 0, 6, 2, 7, 1, 3, 4, 5, 8, },
- ec_gf8_mul_D6_ops
-};
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 7, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_D6 = {9,
+ {
+ 0,
+ 6,
+ 2,
+ 7,
+ 1,
+ 3,
+ 4,
+ 5,
+ 8,
+ },
+ ec_gf8_mul_D6_ops};
static ec_gf_op_t ec_gf8_mul_D7_ops[] = {
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR3, 8, 3, 5 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 0, 8, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR3, 6, 7, 8 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_D7 = {
- 9,
- { 3, 4, 6, 5, 0, 7, 1, 2, 8, },
- ec_gf8_mul_D7_ops
-};
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR3, 8, 3, 5}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 0, 8, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR3, 6, 7, 8}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_D7 = {9,
+ {
+ 3,
+ 4,
+ 6,
+ 5,
+ 0,
+ 7,
+ 1,
+ 2,
+ 8,
+ },
+ ec_gf8_mul_D7_ops};
static ec_gf_op_t ec_gf8_mul_D8_ops[] = {
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_D8 = {
- 8,
- { 4, 5, 6, 7, 0, 1, 2, 3, },
- ec_gf8_mul_D8_ops
-};
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_D8 = {8,
+ {
+ 4,
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ 3,
+ },
+ ec_gf8_mul_D8_ops};
static ec_gf_op_t ec_gf8_mul_D9_ops[] = {
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_D9 = {
- 8,
- { 1, 2, 6, 7, 4, 5, 0, 3, },
- ec_gf8_mul_D9_ops
-};
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 7, 0, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_D9 = {8,
+ {
+ 1,
+ 2,
+ 6,
+ 7,
+ 4,
+ 5,
+ 0,
+ 3,
+ },
+ ec_gf8_mul_D9_ops};
static ec_gf_op_t ec_gf8_mul_DA_ops[] = {
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR3, 8, 2, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_XOR2, 5, 8, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_DA = {
- 9,
- { 2, 5, 7, 1, 0, 4, 3, 6, 8, },
- ec_gf8_mul_DA_ops
-};
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR3, 8, 2, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 1, 8, 0},
+ {EC_GF_OP_XOR2, 5, 8, 0}, {EC_GF_OP_XOR2, 2, 4, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_DA = {9,
+ {
+ 2,
+ 5,
+ 7,
+ 1,
+ 0,
+ 4,
+ 3,
+ 6,
+ 8,
+ },
+ ec_gf8_mul_DA_ops};
static ec_gf_op_t ec_gf8_mul_DB_ops[] = {
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 8, 4, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 3, 8, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_DB = {
- 9,
- { 7, 5, 6, 2, 3, 4, 1, 0, 8, },
- ec_gf8_mul_DB_ops
-};
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 8, 4, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 6, 3, 0}, {EC_GF_OP_XOR2, 3, 8, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_DB = {9,
+ {
+ 7,
+ 5,
+ 6,
+ 2,
+ 3,
+ 4,
+ 1,
+ 0,
+ 8,
+ },
+ ec_gf8_mul_DB_ops};
static ec_gf_op_t ec_gf8_mul_DC_ops[] = {
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_DC = {
- 8,
- { 4, 5, 2, 6, 7, 1, 0, 3, },
- ec_gf8_mul_DC_ops
-};
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_DC = {8,
+ {
+ 4,
+ 5,
+ 2,
+ 6,
+ 7,
+ 1,
+ 0,
+ 3,
+ },
+ ec_gf8_mul_DC_ops};
static ec_gf_op_t ec_gf8_mul_DD_ops[] = {
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_DD = {
- 8,
- { 1, 2, 3, 6, 7, 0, 4, 5, },
- ec_gf8_mul_DD_ops
-};
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_DD = {8,
+ {
+ 1,
+ 2,
+ 3,
+ 6,
+ 7,
+ 0,
+ 4,
+ 5,
+ },
+ ec_gf8_mul_DD_ops};
static ec_gf_op_t ec_gf8_mul_DE_ops[] = {
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_DE = {
- 8,
- { 0, 5, 2, 6, 7, 1, 3, 4, },
- ec_gf8_mul_DE_ops
-};
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_DE = {8,
+ {
+ 0,
+ 5,
+ 2,
+ 6,
+ 7,
+ 1,
+ 3,
+ 4,
+ },
+ ec_gf8_mul_DE_ops};
static ec_gf_op_t ec_gf8_mul_DF_ops[] = {
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 8, 3, 0 },
- { EC_GF_OP_COPY, 9, 0, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 8, 7, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 5, 8, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR3, 1, 9, 2 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_DF = {
- 10,
- { 7, 2, 8, 4, 3, 1, 0, 6, 5, 9, },
- ec_gf8_mul_DF_ops
-};
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 8, 3, 0},
+ {EC_GF_OP_COPY, 9, 0, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 8, 7, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 4, 7, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR2, 5, 8, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR3, 1, 9, 2}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_DF = {10,
+ {
+ 7,
+ 2,
+ 8,
+ 4,
+ 3,
+ 1,
+ 0,
+ 6,
+ 5,
+ 9,
+ },
+ ec_gf8_mul_DF_ops};
static ec_gf_op_t ec_gf8_mul_E0_ops[] = {
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_E0 = {
- 8,
- { 2, 3, 4, 7, 5, 6, 0, 1, },
- ec_gf8_mul_E0_ops
-};
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 4, 1, 0}, {EC_GF_OP_XOR2, 7, 1, 0},
+ {EC_GF_OP_XOR2, 5, 7, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_E0 = {8,
+ {
+ 2,
+ 3,
+ 4,
+ 7,
+ 5,
+ 6,
+ 0,
+ 1,
+ },
+ ec_gf8_mul_E0_ops};
static ec_gf_op_t ec_gf8_mul_E1_ops[] = {
- { EC_GF_OP_COPY, 8, 1, 0 },
- { EC_GF_OP_XOR2, 8, 7, 0 },
- { EC_GF_OP_XOR2, 3, 8, 0 },
- { EC_GF_OP_XOR3, 9, 5, 3 },
- { EC_GF_OP_XOR2, 0, 9, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 4, 9, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 2, 8, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_E1 = {
- 10,
- { 0, 7, 1, 3, 4, 5, 6, 2, 8, 9, },
- ec_gf8_mul_E1_ops
-};
+ {EC_GF_OP_COPY, 8, 1, 0}, {EC_GF_OP_XOR2, 8, 7, 0},
+ {EC_GF_OP_XOR2, 3, 8, 0}, {EC_GF_OP_XOR3, 9, 5, 3},
+ {EC_GF_OP_XOR2, 0, 9, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 4, 9, 0}, {EC_GF_OP_XOR2, 0, 2, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 2, 8, 0},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_E1 = {10,
+ {
+ 0,
+ 7,
+ 1,
+ 3,
+ 4,
+ 5,
+ 6,
+ 2,
+ 8,
+ 9,
+ },
+ ec_gf8_mul_E1_ops};
static ec_gf_op_t ec_gf8_mul_E2_ops[] = {
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_E2 = {
- 8,
- { 2, 3, 7, 1, 5, 6, 0, 4, },
- ec_gf8_mul_E2_ops
-};
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_E2 = {8,
+ {
+ 2,
+ 3,
+ 7,
+ 1,
+ 5,
+ 6,
+ 0,
+ 4,
+ },
+ ec_gf8_mul_E2_ops};
static ec_gf_op_t ec_gf8_mul_E3_ops[] = {
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR3, 8, 2, 7 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 0, 8, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR3, 6, 8, 4 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_E3 = {
- 9,
- { 5, 4, 7, 2, 1, 3, 6, 0, 8, },
- ec_gf8_mul_E3_ops
-};
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 3, 1, 0},
+ {EC_GF_OP_XOR3, 8, 2, 7}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 0, 1, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_XOR2, 0, 8, 0}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR3, 6, 8, 4},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_E3 = {9,
+ {
+ 5,
+ 4,
+ 7,
+ 2,
+ 1,
+ 3,
+ 6,
+ 0,
+ 8,
+ },
+ ec_gf8_mul_E3_ops};
static ec_gf_op_t ec_gf8_mul_E4_ops[] = {
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 4, 5, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_E4 = {
- 8,
- { 7, 0, 1, 6, 3, 4, 2, 5, },
- ec_gf8_mul_E4_ops
-};
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 4, 5, 0},
+ {EC_GF_OP_XOR2, 3, 4, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_E4 = {8,
+ {
+ 7,
+ 0,
+ 1,
+ 6,
+ 3,
+ 4,
+ 2,
+ 5,
+ },
+ ec_gf8_mul_E4_ops};
static ec_gf_op_t ec_gf8_mul_E5_ops[] = {
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 3, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_E5 = {
- 9,
- { 4, 5, 3, 6, 7, 1, 0, 2, 8, },
- ec_gf8_mul_E5_ops
-};
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_COPY, 8, 0, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 3, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_E5 = {9,
+ {
+ 4,
+ 5,
+ 3,
+ 6,
+ 7,
+ 1,
+ 0,
+ 2,
+ 8,
+ },
+ ec_gf8_mul_E5_ops};
static ec_gf_op_t ec_gf8_mul_E6_ops[] = {
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_E6 = {
- 8,
- { 5, 4, 3, 6, 7, 0, 1, 2, },
- ec_gf8_mul_E6_ops
-};
+ {EC_GF_OP_XOR2, 6, 2, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 1, 4, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_E6 = {8,
+ {
+ 5,
+ 4,
+ 3,
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ },
+ ec_gf8_mul_E6_ops};
static ec_gf_op_t ec_gf8_mul_E7_ops[] = {
- { EC_GF_OP_COPY, 8, 6, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR3, 9, 0, 6 },
- { EC_GF_OP_XOR2, 4, 9, 0 },
- { EC_GF_OP_XOR2, 5, 9, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_E7 = {
- 10,
- { 1, 4, 3, 6, 7, 5, 2, 0, 8, 9, },
- ec_gf8_mul_E7_ops
-};
+ {EC_GF_OP_COPY, 8, 6, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR3, 9, 0, 6}, {EC_GF_OP_XOR2, 4, 9, 0},
+ {EC_GF_OP_XOR2, 5, 9, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 7, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_E7 = {10,
+ {
+ 1,
+ 4,
+ 3,
+ 6,
+ 7,
+ 5,
+ 2,
+ 0,
+ 8,
+ 9,
+ },
+ ec_gf8_mul_E7_ops};
static ec_gf_op_t ec_gf8_mul_E8_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 1, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_E8 = {
- 8,
- { 1, 4, 2, 7, 3, 0, 5, 6, },
- ec_gf8_mul_E8_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 2, 5, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 1, 4, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_E8 = {8,
+ {
+ 1,
+ 4,
+ 2,
+ 7,
+ 3,
+ 0,
+ 5,
+ 6,
+ },
+ ec_gf8_mul_E8_ops};
static ec_gf_op_t ec_gf8_mul_E9_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_COPY, 8, 1, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 6, 3, 0 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR3, 1, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_E9 = {
- 9,
- { 6, 2, 0, 3, 4, 1, 5, 7, 8, },
- ec_gf8_mul_E9_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_COPY, 8, 1, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 6, 3, 0},
+ {EC_GF_OP_XOR2, 4, 6, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR3, 1, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_E9 = {9,
+ {
+ 6,
+ 2,
+ 0,
+ 3,
+ 4,
+ 1,
+ 5,
+ 7,
+ 8,
+ },
+ ec_gf8_mul_E9_ops};
static ec_gf_op_t ec_gf8_mul_EA_ops[] = {
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_EA = {
- 8,
- { 3, 4, 5, 6, 7, 0, 1, 2, },
- ec_gf8_mul_EA_ops
-};
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_EA = {8,
+ {
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ },
+ ec_gf8_mul_EA_ops};
static ec_gf_op_t ec_gf8_mul_EB_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_EB = {
- 8,
- { 3, 4, 5, 6, 7, 0, 1, 2, },
- ec_gf8_mul_EB_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_EB = {8,
+ {
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ },
+ ec_gf8_mul_EB_ops};
static ec_gf_op_t ec_gf8_mul_EC_ops[] = {
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR3, 8, 4, 0 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 3, 8, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_EC = {
- 9,
- { 7, 4, 3, 0, 2, 5, 1, 6, 8, },
- ec_gf8_mul_EC_ops
-};
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR3, 8, 4, 0}, {EC_GF_OP_XOR2, 1, 8, 0},
+ {EC_GF_OP_XOR2, 7, 3, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 3, 8, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_EC = {9,
+ {
+ 7,
+ 4,
+ 3,
+ 0,
+ 2,
+ 5,
+ 1,
+ 6,
+ 8,
+ },
+ ec_gf8_mul_EC_ops};
static ec_gf_op_t ec_gf8_mul_ED_ops[] = {
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_ED = {
- 8,
- { 5, 6, 7, 0, 1, 4, 3, 2, },
- ec_gf8_mul_ED_ops
-};
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 4, 0, 0},
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 3, 6, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 5, 2, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_ED = {8,
+ {
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ 4,
+ 3,
+ 2,
+ },
+ ec_gf8_mul_ED_ops};
static ec_gf_op_t ec_gf8_mul_EE_ops[] = {
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR3, 8, 2, 3 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 4, 8, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 8, 5, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_EE = {
- 9,
- { 6, 4, 5, 7, 2, 3, 0, 1, 8, },
- ec_gf8_mul_EE_ops
-};
+ {EC_GF_OP_XOR2, 5, 3, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR3, 8, 2, 3},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_XOR2, 4, 8, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 8, 5, 0},
+ {EC_GF_OP_XOR2, 4, 7, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 1, 8, 0}, {EC_GF_OP_XOR2, 7, 8, 0},
+ {EC_GF_OP_XOR2, 6, 0, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_EE = {9,
+ {
+ 6,
+ 4,
+ 5,
+ 7,
+ 2,
+ 3,
+ 0,
+ 1,
+ 8,
+ },
+ ec_gf8_mul_EE_ops};
static ec_gf_op_t ec_gf8_mul_EF_ops[] = {
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_COPY, 8, 0, 0 },
- { EC_GF_OP_XOR2, 8, 2, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 6, 8, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_EF = {
- 9,
- { 6, 4, 5, 7, 2, 0, 3, 1, 8, },
- ec_gf8_mul_EF_ops
-};
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_COPY, 8, 0, 0},
+ {EC_GF_OP_XOR2, 8, 2, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 7, 8, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 6, 8, 0},
+ {EC_GF_OP_XOR2, 4, 7, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 7, 5, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_EF = {9,
+ {
+ 6,
+ 4,
+ 5,
+ 7,
+ 2,
+ 0,
+ 3,
+ 1,
+ 8,
+ },
+ ec_gf8_mul_EF_ops};
static ec_gf_op_t ec_gf8_mul_F0_ops[] = {
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR3, 8, 3, 6 },
- { EC_GF_OP_XOR2, 5, 8, 0 },
- { EC_GF_OP_XOR2, 8, 4, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 1, 8, 0 },
- { EC_GF_OP_XOR2, 0, 2, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_F0 = {
- 9,
- { 3, 4, 6, 1, 2, 0, 5, 7, 8, },
- ec_gf8_mul_F0_ops
-};
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR3, 8, 3, 6},
+ {EC_GF_OP_XOR2, 5, 8, 0}, {EC_GF_OP_XOR2, 8, 4, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 7, 8, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 1, 8, 0},
+ {EC_GF_OP_XOR2, 0, 2, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_F0 = {9,
+ {
+ 3,
+ 4,
+ 6,
+ 1,
+ 2,
+ 0,
+ 5,
+ 7,
+ 8,
+ },
+ ec_gf8_mul_F0_ops};
static ec_gf_op_t ec_gf8_mul_F1_ops[] = {
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_COPY, 8, 3, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_COPY, 9, 2, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 9, 0, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 5, 2, 0 },
- { EC_GF_OP_XOR2, 7, 9, 0 },
- { EC_GF_OP_XOR2, 4, 9, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR3, 9, 8, 7 },
- { EC_GF_OP_XOR2, 1, 9, 0 },
- { EC_GF_OP_XOR2, 5, 9, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_F1 = {
- 10,
- { 7, 2, 6, 3, 5, 1, 4, 0, 8, 9, },
- ec_gf8_mul_F1_ops
-};
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_COPY, 8, 3, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 2, 3, 0}, {EC_GF_OP_COPY, 9, 2, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 9, 0, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 5, 2, 0},
+ {EC_GF_OP_XOR2, 7, 9, 0}, {EC_GF_OP_XOR2, 4, 9, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR3, 9, 8, 7},
+ {EC_GF_OP_XOR2, 1, 9, 0}, {EC_GF_OP_XOR2, 5, 9, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_F1 = {10,
+ {
+ 7,
+ 2,
+ 6,
+ 3,
+ 5,
+ 1,
+ 4,
+ 0,
+ 8,
+ 9,
+ },
+ ec_gf8_mul_F1_ops};
static ec_gf_op_t ec_gf8_mul_F2_ops[] = {
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 2, 3, 0 },
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_XOR3, 8, 6, 4 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 3, 8, 0 },
- { EC_GF_OP_XOR2, 5, 8, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_F2 = {
- 9,
- { 1, 0, 6, 7, 4, 5, 2, 3, 8, },
- ec_gf8_mul_F2_ops
-};
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 7, 2, 0},
+ {EC_GF_OP_XOR2, 0, 6, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 2, 3, 0},
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_XOR3, 8, 6, 4},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 3, 8, 0}, {EC_GF_OP_XOR2, 5, 8, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_F2 = {9,
+ {
+ 1,
+ 0,
+ 6,
+ 7,
+ 4,
+ 5,
+ 2,
+ 3,
+ 8,
+ },
+ ec_gf8_mul_F2_ops};
static ec_gf_op_t ec_gf8_mul_F3_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_F3 = {
- 8,
- { 5, 6, 7, 0, 1, 2, 3, 4, },
- ec_gf8_mul_F3_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 6, 5, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_F3 = {8,
+ {
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ 3,
+ 4,
+ },
+ ec_gf8_mul_F3_ops};
static ec_gf_op_t ec_gf8_mul_F4_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_F4 = {
- 8,
- { 0, 1, 2, 3, 4, 5, 6, 7, },
- ec_gf8_mul_F4_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 1, 7, 0}, {EC_GF_OP_XOR2, 3, 7, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_F4 = {8,
+ {
+ 0,
+ 1,
+ 2,
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ },
+ ec_gf8_mul_F4_ops};
static ec_gf_op_t ec_gf8_mul_F5_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_F5 = {
- 8,
- { 7, 0, 1, 2, 3, 4, 5, 6, },
- ec_gf8_mul_F5_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_F5 = {8,
+ {
+ 7,
+ 0,
+ 1,
+ 2,
+ 3,
+ 4,
+ 5,
+ 6,
+ },
+ ec_gf8_mul_F5_ops};
static ec_gf_op_t ec_gf8_mul_F6_ops[] = {
- { EC_GF_OP_XOR2, 3, 1, 0 },
- { EC_GF_OP_COPY, 8, 3, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_COPY, 9, 3, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 9, 4, 0 },
- { EC_GF_OP_XOR2, 4, 1, 0 },
- { EC_GF_OP_XOR2, 6, 9, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 5, 7, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR3, 7, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_F6 = {
- 10,
- { 0, 6, 2, 7, 4, 3, 5, 9, 1, 8, },
- ec_gf8_mul_F6_ops
-};
+ {EC_GF_OP_XOR2, 3, 1, 0}, {EC_GF_OP_COPY, 8, 3, 0},
+ {EC_GF_OP_XOR2, 3, 5, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_COPY, 9, 3, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 2, 7, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 9, 4, 0}, {EC_GF_OP_XOR2, 4, 1, 0},
+ {EC_GF_OP_XOR2, 6, 9, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 5, 7, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR3, 7, 8, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_F6 = {10,
+ {
+ 0,
+ 6,
+ 2,
+ 7,
+ 4,
+ 3,
+ 5,
+ 9,
+ 1,
+ 8,
+ },
+ ec_gf8_mul_F6_ops};
static ec_gf_op_t ec_gf8_mul_F7_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_F7 = {
- 8,
- { 6, 7, 0, 1, 2, 3, 4, 5, },
- ec_gf8_mul_F7_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 3, 2, 0}, {EC_GF_OP_XOR2, 4, 3, 0},
+ {EC_GF_OP_XOR2, 5, 4, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_F7 = {8,
+ {
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ 3,
+ 4,
+ 5,
+ },
+ ec_gf8_mul_F7_ops};
static ec_gf_op_t ec_gf8_mul_F8_ops[] = {
- { EC_GF_OP_XOR2, 4, 0, 0 },
- { EC_GF_OP_XOR2, 3, 5, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_F8 = {
- 8,
- { 6, 2, 0, 1, 4, 5, 3, 7, },
- ec_gf8_mul_F8_ops
-};
+ {EC_GF_OP_XOR2, 4, 0, 0}, {EC_GF_OP_XOR2, 3, 5, 0},
+ {EC_GF_OP_XOR2, 6, 4, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 1, 6, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 5, 1, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 6, 7, 0},
+ {EC_GF_OP_XOR2, 0, 3, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_F8 = {8,
+ {
+ 6,
+ 2,
+ 0,
+ 1,
+ 4,
+ 5,
+ 3,
+ 7,
+ },
+ ec_gf8_mul_F8_ops};
static ec_gf_op_t ec_gf8_mul_F9_ops[] = {
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 6, 4, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR3, 8, 7, 1 },
- { EC_GF_OP_XOR2, 1, 3, 0 },
- { EC_GF_OP_XOR2, 4, 8, 0 },
- { EC_GF_OP_XOR2, 5, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_F9 = {
- 9,
- { 4, 1, 7, 6, 0, 3, 5, 2, 8, },
- ec_gf8_mul_F9_ops
-};
+ {EC_GF_OP_XOR2, 1, 5, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 0, 5, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 6, 4, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR3, 8, 7, 1}, {EC_GF_OP_XOR2, 1, 3, 0},
+ {EC_GF_OP_XOR2, 4, 8, 0}, {EC_GF_OP_XOR2, 5, 8, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_F9 = {9,
+ {
+ 4,
+ 1,
+ 7,
+ 6,
+ 0,
+ 3,
+ 5,
+ 2,
+ 8,
+ },
+ ec_gf8_mul_F9_ops};
static ec_gf_op_t ec_gf8_mul_FA_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 7, 2, 0 },
- { EC_GF_OP_XOR2, 1, 5, 0 },
- { EC_GF_OP_XOR2, 3, 7, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 0, 3, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_FA = {
- 8,
- { 0, 1, 2, 4, 5, 6, 7, 3, },
- ec_gf8_mul_FA_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 2, 1, 0}, {EC_GF_OP_XOR2, 0, 7, 0},
+ {EC_GF_OP_XOR2, 7, 2, 0}, {EC_GF_OP_XOR2, 1, 5, 0},
+ {EC_GF_OP_XOR2, 3, 7, 0}, {EC_GF_OP_XOR2, 5, 0, 0},
+ {EC_GF_OP_XOR2, 7, 6, 0}, {EC_GF_OP_XOR2, 0, 3, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 2, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_FA = {8,
+ {
+ 0,
+ 1,
+ 2,
+ 4,
+ 5,
+ 6,
+ 7,
+ 3,
+ },
+ ec_gf8_mul_FA_ops};
static ec_gf_op_t ec_gf8_mul_FB_ops[] = {
- { EC_GF_OP_XOR2, 1, 0, 0 },
- { EC_GF_OP_XOR2, 2, 1, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 3, 2, 0 },
- { EC_GF_OP_XOR2, 0, 7, 0 },
- { EC_GF_OP_XOR2, 2, 7, 0 },
- { EC_GF_OP_XOR2, 1, 6, 0 },
- { EC_GF_OP_XOR2, 7, 6, 0 },
- { EC_GF_OP_XOR2, 4, 3, 0 },
- { EC_GF_OP_XOR2, 6, 5, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 5, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_FB = {
- 8,
- { 4, 5, 6, 7, 0, 1, 2, 3, },
- ec_gf8_mul_FB_ops
-};
+ {EC_GF_OP_XOR2, 1, 0, 0}, {EC_GF_OP_XOR2, 2, 1, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 3, 2, 0},
+ {EC_GF_OP_XOR2, 0, 7, 0}, {EC_GF_OP_XOR2, 2, 7, 0},
+ {EC_GF_OP_XOR2, 1, 6, 0}, {EC_GF_OP_XOR2, 7, 6, 0},
+ {EC_GF_OP_XOR2, 4, 3, 0}, {EC_GF_OP_XOR2, 6, 5, 0},
+ {EC_GF_OP_XOR2, 7, 4, 0}, {EC_GF_OP_XOR2, 5, 4, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_FB = {8,
+ {
+ 4,
+ 5,
+ 6,
+ 7,
+ 0,
+ 1,
+ 2,
+ 3,
+ },
+ ec_gf8_mul_FB_ops};
static ec_gf_op_t ec_gf8_mul_FC_ops[] = {
- { EC_GF_OP_XOR2, 7, 0, 0 },
- { EC_GF_OP_XOR2, 7, 4, 0 },
- { EC_GF_OP_XOR2, 5, 1, 0 },
- { EC_GF_OP_COPY, 9, 3, 0 },
- { EC_GF_OP_XOR3, 8, 5, 7 },
- { EC_GF_OP_XOR2, 3, 6, 0 },
- { EC_GF_OP_XOR2, 8, 3, 0 },
- { EC_GF_OP_XOR2, 2, 8, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 3, 4, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 6, 0, 0 },
- { EC_GF_OP_XOR3, 0, 9, 2 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_FC = {
- 10,
- { 5, 6, 3, 7, 1, 8, 0, 4, 2, 9, },
- ec_gf8_mul_FC_ops
-};
+ {EC_GF_OP_XOR2, 7, 0, 0}, {EC_GF_OP_XOR2, 7, 4, 0},
+ {EC_GF_OP_XOR2, 5, 1, 0}, {EC_GF_OP_COPY, 9, 3, 0},
+ {EC_GF_OP_XOR3, 8, 5, 7}, {EC_GF_OP_XOR2, 3, 6, 0},
+ {EC_GF_OP_XOR2, 8, 3, 0}, {EC_GF_OP_XOR2, 2, 8, 0},
+ {EC_GF_OP_XOR2, 1, 2, 0}, {EC_GF_OP_XOR2, 4, 2, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 3, 4, 0},
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_XOR2, 6, 0, 0},
+ {EC_GF_OP_XOR3, 0, 9, 2}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_FC = {10,
+ {
+ 5,
+ 6,
+ 3,
+ 7,
+ 1,
+ 8,
+ 0,
+ 4,
+ 2,
+ 9,
+ },
+ ec_gf8_mul_FC_ops};
static ec_gf_op_t ec_gf8_mul_FD_ops[] = {
- { EC_GF_OP_XOR2, 7, 1, 0 },
- { EC_GF_OP_COPY, 8, 7, 0 },
- { EC_GF_OP_XOR2, 5, 0, 0 },
- { EC_GF_OP_XOR2, 7, 5, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 2, 5, 0 },
- { EC_GF_OP_XOR2, 1, 2, 0 },
- { EC_GF_OP_XOR2, 0, 1, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR3, 1, 8, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_FD = {
- 9,
- { 5, 3, 7, 6, 1, 2, 4, 0, 8, },
- ec_gf8_mul_FD_ops
-};
+ {EC_GF_OP_XOR2, 7, 1, 0}, {EC_GF_OP_COPY, 8, 7, 0},
+ {EC_GF_OP_XOR2, 5, 0, 0}, {EC_GF_OP_XOR2, 7, 5, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 4, 7, 0},
+ {EC_GF_OP_XOR2, 5, 6, 0}, {EC_GF_OP_XOR2, 0, 4, 0},
+ {EC_GF_OP_XOR2, 3, 0, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 2, 5, 0}, {EC_GF_OP_XOR2, 1, 2, 0},
+ {EC_GF_OP_XOR2, 0, 1, 0}, {EC_GF_OP_XOR2, 6, 1, 0},
+ {EC_GF_OP_XOR3, 1, 8, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_FD = {9,
+ {
+ 5,
+ 3,
+ 7,
+ 6,
+ 1,
+ 2,
+ 4,
+ 0,
+ 8,
+ },
+ ec_gf8_mul_FD_ops};
static ec_gf_op_t ec_gf8_mul_FE_ops[] = {
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_COPY, 8, 2, 0 },
- { EC_GF_OP_XOR2, 2, 4, 0 },
- { EC_GF_OP_XOR2, 6, 2, 0 },
- { EC_GF_OP_XOR2, 8, 5, 0 },
- { EC_GF_OP_XOR2, 5, 6, 0 },
- { EC_GF_OP_XOR2, 6, 1, 0 },
- { EC_GF_OP_XOR2, 0, 6, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 7, 8, 0 },
- { EC_GF_OP_XOR2, 3, 0, 0 },
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR2, 0, 4, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_FE = {
- 9,
- { 3, 4, 8, 2, 5, 0, 6, 1, 7, },
- ec_gf8_mul_FE_ops
-};
+ {EC_GF_OP_XOR2, 2, 0, 0}, {EC_GF_OP_COPY, 8, 2, 0},
+ {EC_GF_OP_XOR2, 2, 4, 0}, {EC_GF_OP_XOR2, 6, 2, 0},
+ {EC_GF_OP_XOR2, 8, 5, 0}, {EC_GF_OP_XOR2, 5, 6, 0},
+ {EC_GF_OP_XOR2, 6, 1, 0}, {EC_GF_OP_XOR2, 0, 6, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 7, 8, 0}, {EC_GF_OP_XOR2, 3, 0, 0},
+ {EC_GF_OP_XOR2, 4, 7, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR2, 0, 4, 0}, {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_FE = {9,
+ {
+ 3,
+ 4,
+ 8,
+ 2,
+ 5,
+ 0,
+ 6,
+ 1,
+ 7,
+ },
+ ec_gf8_mul_FE_ops};
static ec_gf_op_t ec_gf8_mul_FF_ops[] = {
- { EC_GF_OP_XOR2, 4, 7, 0 },
- { EC_GF_OP_COPY, 9, 0, 0 },
- { EC_GF_OP_COPY, 8, 4, 0 },
- { EC_GF_OP_XOR2, 9, 1, 0 },
- { EC_GF_OP_XOR2, 4, 2, 0 },
- { EC_GF_OP_XOR2, 9, 4, 0 },
- { EC_GF_OP_XOR2, 0, 5, 0 },
- { EC_GF_OP_XOR2, 2, 0, 0 },
- { EC_GF_OP_XOR2, 3, 9, 0 },
- { EC_GF_OP_XOR2, 7, 3, 0 },
- { EC_GF_OP_XOR2, 2, 6, 0 },
- { EC_GF_OP_XOR2, 5, 3, 0 },
- { EC_GF_OP_XOR2, 6, 7, 0 },
- { EC_GF_OP_XOR2, 1, 7, 0 },
- { EC_GF_OP_XOR3, 3, 8, 5 },
- { EC_GF_OP_XOR2, 4, 6, 0 },
- { EC_GF_OP_END, 0, 0, 0 }
-};
-
-static ec_gf_mul_t ec_gf8_mul_FF = {
- 10,
- { 6, 5, 0, 1, 2, 4, 9, 3, 7, 8, },
- ec_gf8_mul_FF_ops
-};
+ {EC_GF_OP_XOR2, 4, 7, 0}, {EC_GF_OP_COPY, 9, 0, 0},
+ {EC_GF_OP_COPY, 8, 4, 0}, {EC_GF_OP_XOR2, 9, 1, 0},
+ {EC_GF_OP_XOR2, 4, 2, 0}, {EC_GF_OP_XOR2, 9, 4, 0},
+ {EC_GF_OP_XOR2, 0, 5, 0}, {EC_GF_OP_XOR2, 2, 0, 0},
+ {EC_GF_OP_XOR2, 3, 9, 0}, {EC_GF_OP_XOR2, 7, 3, 0},
+ {EC_GF_OP_XOR2, 2, 6, 0}, {EC_GF_OP_XOR2, 5, 3, 0},
+ {EC_GF_OP_XOR2, 6, 7, 0}, {EC_GF_OP_XOR2, 1, 7, 0},
+ {EC_GF_OP_XOR3, 3, 8, 5}, {EC_GF_OP_XOR2, 4, 6, 0},
+ {EC_GF_OP_END, 0, 0, 0}};
+
+static ec_gf_mul_t ec_gf8_mul_FF = {10,
+ {
+ 6,
+ 5,
+ 0,
+ 1,
+ 2,
+ 4,
+ 9,
+ 3,
+ 7,
+ 8,
+ },
+ ec_gf8_mul_FF_ops};
ec_gf_mul_t *ec_gf8_mul[] = {
&ec_gf8_mul_00, &ec_gf8_mul_01, &ec_gf8_mul_02, &ec_gf8_mul_03,
@@ -5955,5 +5879,4 @@ ec_gf_mul_t *ec_gf8_mul[] = {
&ec_gf8_mul_F0, &ec_gf8_mul_F1, &ec_gf8_mul_F2, &ec_gf8_mul_F3,
&ec_gf8_mul_F4, &ec_gf8_mul_F5, &ec_gf8_mul_F6, &ec_gf8_mul_F7,
&ec_gf8_mul_F8, &ec_gf8_mul_F9, &ec_gf8_mul_FA, &ec_gf8_mul_FB,
- &ec_gf8_mul_FC, &ec_gf8_mul_FD, &ec_gf8_mul_FE, &ec_gf8_mul_FF
-};
+ &ec_gf8_mul_FC, &ec_gf8_mul_FD, &ec_gf8_mul_FE, &ec_gf8_mul_FF};
diff --git a/xlators/cluster/ec/src/ec-heal.c b/xlators/cluster/ec/src/ec-heal.c
index 2ff713b7bbd..7d991f04aac 100644
--- a/xlators/cluster/ec/src/ec-heal.c
+++ b/xlators/cluster/ec/src/ec-heal.c
@@ -8,16 +8,14 @@
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "defaults.h"
-#include "compat-errno.h"
-#include "byte-order.h"
-#include "syncop.h"
-#include "syncop-utils.h"
-#include "cluster-syncop.h"
+#include <glusterfs/defaults.h>
+#include <glusterfs/compat-errno.h>
+#include <glusterfs/byte-order.h>
+#include <glusterfs/syncop.h>
+#include <glusterfs/syncop-utils.h>
+#include <glusterfs/cluster-syncop.h>
#include "ec.h"
-#include "ec-mem-types.h"
#include "ec-types.h"
#include "ec-messages.h"
#include "ec-helpers.h"
@@ -27,85 +25,147 @@
#include "ec-fops.h"
#include "ec-heald.h"
-#define alloca0(size) ({void *__ptr; __ptr = alloca(size); memset(__ptr, 0, size); __ptr; })
-#define EC_COUNT(array, max) ({int __i; int __res = 0; for (__i = 0; __i < max; __i++) if (array[__i]) __res++; __res; })
-#define EC_INTERSECT(dst, src1, src2, max) ({int __i; for (__i = 0; __i < max; __i++) dst[__i] = src1[__i] && src2[__i]; })
-#define EC_ADJUST_SOURCE(source, sources, max) ({int __i; if (sources[source] == 0) {source = -1; for (__i = 0; __i < max; __i++) if (sources[__i]) source = __i; } })
-#define IA_EQUAL(f, s, field) (memcmp (&(f.ia_##field), &(s.ia_##field), sizeof (s.ia_##field)) == 0)
-#define EC_REPLIES_ALLOC(replies, numsubvols) do { \
- int __i = 0; \
- replies = alloca0(numsubvols * sizeof (*replies)); \
- for (__i = 0; __i < numsubvols; __i++) \
- INIT_LIST_HEAD (&replies[__i].entries.list); \
- } while (0)
-
+#define EC_COUNT(array, max) \
+ ({ \
+ int __i; \
+ int __res = 0; \
+ for (__i = 0; __i < max; __i++) \
+ if (array[__i]) \
+ __res++; \
+ __res; \
+ })
+#define EC_INTERSECT(dst, src1, src2, max) \
+ ({ \
+ int __i; \
+ for (__i = 0; __i < max; __i++) \
+ dst[__i] = src1[__i] && src2[__i]; \
+ })
+#define EC_ADJUST_SOURCE(source, sources, max) \
+ ({ \
+ int __i; \
+ if (sources[source] == 0) { \
+ source = -1; \
+ for (__i = 0; __i < max; __i++) \
+ if (sources[__i]) \
+ source = __i; \
+ } \
+ })
+#define IA_EQUAL(f, s, field) \
+ (memcmp(&(f.ia_##field), &(s.ia_##field), sizeof(s.ia_##field)) == 0)
+#define EC_REPLIES_ALLOC(replies, numsubvols) \
+ do { \
+ int __i = 0; \
+ replies = alloca0(numsubvols * sizeof(*replies)); \
+ for (__i = 0; __i < numsubvols; __i++) \
+ INIT_LIST_HEAD(&replies[__i].entries.list); \
+ } while (0)
struct ec_name_data {
- call_frame_t *frame;
- unsigned char *participants;
- unsigned char *failed_on;
- unsigned char *gfidless;
- unsigned char *enoent;
- unsigned char *same;
- char *name;
- inode_t *parent;
- default_args_cbk_t *replies;
+ call_frame_t *frame;
+ unsigned char *participants;
+ unsigned char *failed_on;
+ unsigned char *gfidless;
+ unsigned char *enoent;
+ unsigned char *same;
+ char *name;
+ inode_t *parent;
+ default_args_cbk_t *replies;
+ uint32_t heal_pending;
};
-static char *ec_ignore_xattrs[] = {
- GF_SELINUX_XATTR_KEY,
- QUOTA_SIZE_KEY,
- NULL
-};
+static char *ec_ignore_xattrs[] = {GF_SELINUX_XATTR_KEY, QUOTA_SIZE_KEY, NULL};
static gf_boolean_t
-ec_ignorable_key_match (dict_t *dict, char *key, data_t *val, void *mdata)
+ec_ignorable_key_match(dict_t *dict, char *key, data_t *val, void *mdata)
{
- int i = 0;
+ int i = 0;
- if (!key)
- goto out;
+ if (!key)
+ goto out;
- if (strncmp (key, EC_XATTR_PREFIX, strlen (EC_XATTR_PREFIX)) == 0)
- return _gf_true;
+ if (strncmp(key, EC_XATTR_PREFIX, SLEN(EC_XATTR_PREFIX)) == 0)
+ return _gf_true;
- for (i = 0; ec_ignore_xattrs[i]; i++) {
- if (!strcmp (key, ec_ignore_xattrs[i]))
- return _gf_true;
- }
+ for (i = 0; ec_ignore_xattrs[i]; i++) {
+ if (!strcmp(key, ec_ignore_xattrs[i]))
+ return _gf_true;
+ }
out:
- return _gf_false;
+ return _gf_false;
}
static gf_boolean_t
-ec_sh_key_match (dict_t *dict, char *key, data_t *val, void *mdata)
+ec_sh_key_match(dict_t *dict, char *key, data_t *val, void *mdata)
{
- return !ec_ignorable_key_match (dict, key, val, mdata);
+ return !ec_ignorable_key_match(dict, key, val, mdata);
}
/* FOP: heal */
-uintptr_t ec_heal_check(ec_fop_data_t * fop, uintptr_t * pgood)
+void
+ec_set_entry_healing(ec_fop_data_t *fop)
+{
+ ec_inode_t *ctx = NULL;
+ loc_t *loc = NULL;
+
+ if (!fop)
+ return;
+
+ loc = &fop->loc[0];
+ LOCK(&loc->inode->lock);
+ {
+ ctx = __ec_inode_get(loc->inode, fop->xl);
+ if (ctx) {
+ ctx->heal_count += 1;
+ }
+ }
+ UNLOCK(&loc->inode->lock);
+}
+
+void
+ec_reset_entry_healing(ec_fop_data_t *fop)
+{
+ ec_inode_t *ctx = NULL;
+ loc_t *loc = NULL;
+ int32_t heal_count = 0;
+ if (!fop)
+ return;
+
+ loc = &fop->loc[0];
+ LOCK(&loc->inode->lock);
+ {
+ ctx = __ec_inode_get(loc->inode, fop->xl);
+ if (ctx) {
+ ctx->heal_count += -1;
+ heal_count = ctx->heal_count;
+ }
+ }
+ UNLOCK(&loc->inode->lock);
+ GF_ASSERT(heal_count >= 0);
+}
+
+uintptr_t
+ec_heal_check(ec_fop_data_t *fop, uintptr_t *pgood)
{
- ec_cbk_data_t * cbk;
- uintptr_t mask[2] = { 0, 0 };
+ ec_cbk_data_t *cbk;
+ uintptr_t mask[2] = {0, 0};
list_for_each_entry(cbk, &fop->cbk_list, list)
{
mask[cbk->op_ret >= 0] |= cbk->mask;
}
- if (pgood != NULL)
- {
+ if (pgood != NULL) {
*pgood = mask[1];
}
return mask[0];
}
-void ec_heal_update(ec_fop_data_t * fop, int32_t is_open)
+void
+ec_heal_update(ec_fop_data_t *fop, int32_t is_open)
{
- ec_heal_t * heal = fop->data;
+ ec_heal_t *heal = fop->data;
uintptr_t good, bad;
bad = ec_heal_check(fop, &good);
@@ -113,8 +173,7 @@ void ec_heal_update(ec_fop_data_t * fop, int32_t is_open)
LOCK(&heal->lock);
heal->bad &= ~bad;
- if (is_open)
- {
+ if (is_open) {
heal->open |= good;
}
@@ -123,9 +182,10 @@ void ec_heal_update(ec_fop_data_t * fop, int32_t is_open)
fop->error = 0;
}
-void ec_heal_avoid(ec_fop_data_t * fop)
+void
+ec_heal_avoid(ec_fop_data_t *fop)
{
- ec_heal_t * heal = fop->data;
+ ec_heal_t *heal = fop->data;
uintptr_t bad;
bad = ec_heal_check(fop, NULL);
@@ -137,22 +197,24 @@ void ec_heal_avoid(ec_fop_data_t * fop)
UNLOCK(&heal->lock);
}
-int32_t ec_heal_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+int32_t
+ec_heal_lock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
ec_fop_data_t *fop = cookie;
ec_heal_t *heal = fop->data;
if (op_ret >= 0) {
- GF_ASSERT(ec_set_inode_size(heal->fop, heal->fd->inode,
- heal->total_size));
+ GF_ASSERT(
+ ec_set_inode_size(heal->fop, heal->fd->inode, heal->total_size));
}
return 0;
}
-void ec_heal_lock(ec_heal_t *heal, int32_t type, fd_t *fd, loc_t *loc,
- off_t offset, size_t size)
+void
+ec_heal_lock(ec_heal_t *heal, int32_t type, fd_t *fd, loc_t *loc, off_t offset,
+ size_t size)
{
struct gf_flock flock;
fop_inodelk_cbk_t cbk = NULL;
@@ -177,15 +239,12 @@ void ec_heal_lock(ec_heal_t *heal, int32_t type, fd_t *fd, loc_t *loc,
cbk = ec_heal_lock_cbk;
}
- if (fd != NULL)
- {
+ if (fd != NULL) {
ec_finodelk(heal->fop->frame, heal->xl,
&heal->fop->frame->root->lk_owner, heal->fop->mask,
EC_MINIMUM_ALL, cbk, heal, heal->xl->name, fd, F_SETLKW,
&flock, NULL);
- }
- else
- {
+ } else {
ec_inodelk(heal->fop->frame, heal->xl,
&heal->fop->frame->root->lk_owner, heal->fop->mask,
EC_MINIMUM_ALL, cbk, heal, heal->xl->name, loc, F_SETLKW,
@@ -193,28 +252,28 @@ void ec_heal_lock(ec_heal_t *heal, int32_t type, fd_t *fd, loc_t *loc,
}
}
-void ec_heal_inodelk(ec_heal_t *heal, int32_t type, int32_t use_fd,
- off_t offset, size_t size)
+void
+ec_heal_inodelk(ec_heal_t *heal, int32_t type, int32_t use_fd, off_t offset,
+ size_t size)
{
ec_heal_lock(heal, type, use_fd ? heal->fd : NULL, &heal->loc, offset,
size);
}
int32_t
-ec_heal_xattr_clean (dict_t *dict, char *key, data_t *data,
- void *arg)
+ec_heal_xattr_clean(dict_t *dict, char *key, data_t *data, void *arg)
{
- dict_t *base = arg;
+ dict_t *base = arg;
- if (ec_ignorable_key_match (NULL, key, NULL, NULL)) {
- dict_del (dict, key);
- return 0;
- }
+ if (ec_ignorable_key_match(NULL, key, NULL, NULL)) {
+ dict_del(dict, key);
+ return 0;
+ }
- if (dict_get (base, key) != NULL)
- dict_del (dict, key);
+ if (dict_get(base, key) != NULL)
+ dict_del(dict, key);
- return 0;
+ return 0;
}
/********************************************************************
@@ -224,68 +283,66 @@ ec_heal_xattr_clean (dict_t *dict, char *key, data_t *data,
*
*******************************************************************/
void
-ec_wind_xattrop_parallel (call_frame_t *frame, xlator_t *subvol,
- int child_index, loc_t *loc,
- gf_xattrop_flags_t flags, dict_t **dict,
- dict_t *xdata)
+ec_wind_xattrop_parallel(call_frame_t *frame, xlator_t *subvol, int child_index,
+ loc_t *loc, gf_xattrop_flags_t flags, dict_t **dict,
+ dict_t *xdata)
{
- gf_msg_debug ("EC", 0, "WIND: on child %d ", child_index);
- STACK_WIND_COOKIE (frame, cluster_xattrop_cbk,
- (void *)(uintptr_t) child_index,
- subvol, subvol->fops->xattrop, loc,
- flags, dict[child_index], xdata);
+ gf_msg_debug("EC", 0, "WIND: on child %d ", child_index);
+ STACK_WIND_COOKIE(
+ frame, cluster_xattrop_cbk, (void *)(uintptr_t)child_index, subvol,
+ subvol->fops->xattrop, loc, flags, dict[child_index], xdata);
}
int32_t
-ec_heal_writev_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct iatt *prebuf, struct iatt *postbuf,
- dict_t *xdata)
+ec_heal_writev_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
ec_fop_data_t *fop = cookie;
ec_heal_t *heal = fop->data;
ec_trace("WRITE_CBK", cookie, "ret=%d, errno=%d", op_ret, op_errno);
- gf_msg_debug (fop->xl->name, 0, "%s: write op_ret %d, op_errno %s"
- " at %"PRIu64, uuid_utoa (heal->fd->inode->gfid), op_ret,
- strerror (op_errno), heal->offset);
+ gf_msg_debug(fop->xl->name, 0,
+ "%s: write op_ret %d, op_errno %s"
+ " at %" PRIu64,
+ uuid_utoa(heal->fd->inode->gfid), op_ret, strerror(op_errno),
+ heal->offset);
ec_heal_update(cookie, 0);
return 0;
}
-int32_t ec_heal_readv_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno,
- struct iovec * vector, int32_t count,
- struct iatt * stbuf, struct iobref * iobref,
- dict_t * xdata)
+int32_t
+ec_heal_readv_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iovec *vector,
+ int32_t count, struct iatt *stbuf, struct iobref *iobref,
+ dict_t *xdata)
{
- ec_fop_data_t * fop = cookie;
- ec_heal_t * heal = fop->data;
+ ec_fop_data_t *fop = cookie;
+ ec_heal_t *heal = fop->data;
ec_trace("READ_CBK", fop, "ret=%d, errno=%d", op_ret, op_errno);
ec_heal_avoid(fop);
- if (op_ret > 0)
- {
- gf_msg_debug (fop->xl->name, 0, "%s: read succeeded, proceeding "
- "to write at %"PRIu64, uuid_utoa (heal->fd->inode->gfid),
- heal->offset);
+ if (op_ret > 0) {
+ gf_msg_debug(fop->xl->name, 0,
+ "%s: read succeeded, proceeding "
+ "to write at %" PRIu64,
+ uuid_utoa(heal->fd->inode->gfid), heal->offset);
ec_writev(heal->fop->frame, heal->xl, heal->bad, EC_MINIMUM_ONE,
ec_heal_writev_cbk, heal, heal->fd, vector, count,
heal->offset, 0, iobref, NULL);
- }
- else
- {
+ } else {
if (op_ret < 0) {
- gf_msg_debug (fop->xl->name, 0, "%s: read failed %s, failing "
- "to heal block at %"PRIu64,
- uuid_utoa (heal->fd->inode->gfid), strerror (op_errno),
- heal->offset);
- heal->bad = 0;
+ gf_msg_debug(fop->xl->name, 0,
+ "%s: read failed %s, failing "
+ "to heal block at %" PRIu64,
+ uuid_utoa(heal->fd->inode->gfid), strerror(op_errno),
+ heal->offset);
+ heal->bad = 0;
}
heal->done = 1;
}
@@ -293,1652 +350,1657 @@ int32_t ec_heal_readv_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
return 0;
}
-void ec_heal_data_block(ec_heal_t *heal)
+void
+ec_heal_data_block(ec_heal_t *heal)
{
ec_trace("DATA", heal->fop, "good=%lX, bad=%lX", heal->good, heal->bad);
if ((heal->good != 0) && (heal->bad != 0) &&
- (heal->iatt.ia_type == IA_IFREG))
- {
+ (heal->iatt.ia_type == IA_IFREG)) {
ec_readv(heal->fop->frame, heal->xl, heal->good, EC_MINIMUM_MIN,
- ec_heal_readv_cbk, heal, heal->fd, heal->size, heal->offset,
- 0, NULL);
+ ec_heal_readv_cbk, heal, heal->fd, heal->size, heal->offset, 0,
+ NULL);
}
}
/* FOP: fheal */
-void ec_fheal(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fheal_cbk_t func, void * data, fd_t * fd,
- int32_t partial, dict_t *xdata)
+void
+ec_fheal(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fheal_cbk_t func, void *data, fd_t *fd,
+ int32_t partial, dict_t *xdata)
{
- ec_fd_t * ctx = ec_fd_get(fd, this);
+ ec_fd_t *ctx = ec_fd_get(fd, this);
- if (ctx != NULL)
- {
- gf_msg_trace ("ec", 0, "FHEAL ctx: flags=%X, open=%lX", ctx->flags,
- ctx->open);
- ec_heal(frame, this, target, minimum, func, data, &ctx->loc, partial,
+ if (ctx != NULL) {
+ gf_msg_trace("ec", 0, "FHEAL ctx: flags=%X, open=%" PRIXPTR, ctx->flags,
+ ctx->open);
+ ec_heal(frame, this, target, fop_flags, func, data, &ctx->loc, partial,
xdata);
}
}
/* Common heal code */
void
-ec_mask_to_char_array (uintptr_t mask, unsigned char *array, int numsubvols)
+ec_mask_to_char_array(uintptr_t mask, unsigned char *array, int numsubvols)
{
- int i = 0;
+ int i = 0;
- for (i = 0; i < numsubvols; i++)
- array[i] = ((mask >> i) & 1);
+ for (i = 0; i < numsubvols; i++)
+ array[i] = ((mask >> i) & 1);
}
uintptr_t
-ec_char_array_to_mask (unsigned char *array, int numsubvols)
+ec_char_array_to_mask(unsigned char *array, int numsubvols)
{
- int i = 0;
- uintptr_t mask = 0;
+ int i = 0;
+ uintptr_t mask = 0;
- for (i = 0; i < numsubvols; i++)
- if (array[i])
- mask |= (1ULL<<i);
- return mask;
+ if (array == NULL)
+ goto out;
+
+ for (i = 0; i < numsubvols; i++)
+ if (array[i])
+ mask |= (1ULL << i);
+out:
+ return mask;
}
int
-ec_heal_entry_find_direction (ec_t *ec, default_args_cbk_t *replies,
- uint64_t *versions, uint64_t *dirty,
- unsigned char *sources, unsigned char *healed_sinks)
+ec_heal_entry_find_direction(ec_t *ec, default_args_cbk_t *replies,
+ uint64_t *versions, uint64_t *dirty,
+ unsigned char *sources,
+ unsigned char *healed_sinks)
{
- uint64_t xattr[EC_VERSION_SIZE] = {0};
- int source = -1;
- uint64_t max_version = 0;
- int ret = 0;
- int i = 0;
+ uint64_t xattr[EC_VERSION_SIZE] = {0};
+ int source = -1;
+ uint64_t max_version = 0;
+ int ret = 0;
+ int i = 0;
- for (i = 0; i < ec->nodes; i++) {
- if (!replies[i].valid)
- continue;
-
- if (replies[i].op_ret == -1)
- continue;
-
- if (source == -1)
- source = i;
-
- ret = ec_dict_get_array (replies[i].xdata, EC_XATTR_VERSION,
- xattr, EC_VERSION_SIZE);
- if (ret == 0) {
- versions[i] = xattr[EC_DATA_TXN];
- if (max_version < versions[i]) {
- max_version = versions[i];
- source = i;
- }
- }
+ for (i = 0; i < ec->nodes; i++) {
+ if (!replies[i].valid)
+ continue;
- memset (xattr, 0, sizeof(xattr));
- ret = ec_dict_get_array (replies[i].xdata, EC_XATTR_DIRTY,
- xattr, EC_VERSION_SIZE);
- if (ret == 0) {
- dirty[i] = xattr[EC_DATA_TXN];
- }
+ if (replies[i].op_ret == -1)
+ continue;
+
+ if (source == -1)
+ source = i;
+
+ ret = ec_dict_get_array(replies[i].xdata, EC_XATTR_VERSION, xattr,
+ EC_VERSION_SIZE);
+ if (ret == 0) {
+ versions[i] = xattr[EC_DATA_TXN];
+ if (max_version < versions[i]) {
+ max_version = versions[i];
+ source = i;
+ }
}
- if (source < 0)
- goto out;
+ memset(xattr, 0, sizeof(xattr));
+ ret = ec_dict_get_array(replies[i].xdata, EC_XATTR_DIRTY, xattr,
+ EC_VERSION_SIZE);
+ if (ret == 0) {
+ dirty[i] = xattr[EC_DATA_TXN];
+ }
+ }
- for (i = 0; i < ec->nodes; i++) {
- if (!replies[i].valid)
- continue;
+ if (source < 0)
+ goto out;
- if (replies[i].op_ret == -1)
- continue;
+ for (i = 0; i < ec->nodes; i++) {
+ if (!replies[i].valid)
+ continue;
- if (versions[i] == versions[source])
- sources[i] = 1;
- else
- healed_sinks[i] = 1;
- }
+ if (replies[i].op_ret == -1)
+ continue;
+
+ if (versions[i] == versions[source])
+ sources[i] = 1;
+ else
+ healed_sinks[i] = 1;
+ }
out:
- return source;
+ return source;
}
int
-ec_adjust_versions (call_frame_t *frame, ec_t *ec, ec_txn_t type,
- inode_t *inode, int source, unsigned char *sources,
- unsigned char *healed_sinks, uint64_t *versions,
- uint64_t *dirty)
-{
- int i = 0;
- int ret = 0;
- int call_count = 0;
- dict_t **xattr = NULL;
- int op_ret = 0;
- loc_t loc = {0};
- gf_boolean_t erase_dirty = _gf_false;
- uint64_t *versions_xattr = NULL;
- uint64_t *dirty_xattr = NULL;
- uint64_t allzero[2] = {0};
- unsigned char *on = NULL;
- unsigned char *output = NULL;
- default_args_cbk_t *replies = NULL;
-
- /* Allocate the required memory */
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
- on = alloca0 (ec->nodes);
- output = alloca0 (ec->nodes);
- EC_REPLIES_ALLOC (replies, ec->nodes);
- xattr = GF_CALLOC (ec->nodes, sizeof (*xattr), gf_common_mt_pointer);
- if (!xattr) {
- op_ret = -ENOMEM;
- goto out;
- }
- for (i = 0; i < ec->nodes; i++) {
- xattr[i] = dict_new ();
- if (!xattr[i]) {
- op_ret = -ENOMEM;
- goto out;
- }
+ec_adjust_versions(call_frame_t *frame, ec_t *ec, ec_txn_t type, inode_t *inode,
+ int source, unsigned char *sources,
+ unsigned char *healed_sinks, uint64_t *versions,
+ uint64_t *dirty)
+{
+ int i = 0;
+ int ret = 0;
+ int call_count = 0;
+ dict_t **xattr = NULL;
+ int op_ret = 0;
+ loc_t loc = {0};
+ gf_boolean_t erase_dirty = _gf_false;
+ uint64_t *versions_xattr = NULL;
+ uint64_t *dirty_xattr = NULL;
+ uint64_t allzero[2] = {0};
+ unsigned char *on = NULL;
+ unsigned char *output = NULL;
+ default_args_cbk_t *replies = NULL;
+
+ /* Allocate the required memory */
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+ on = alloca0(ec->nodes);
+ output = alloca0(ec->nodes);
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ xattr = GF_CALLOC(ec->nodes, sizeof(*xattr), gf_common_mt_pointer);
+ if (!xattr) {
+ op_ret = -ENOMEM;
+ goto out;
+ }
+ for (i = 0; i < ec->nodes; i++) {
+ xattr[i] = dict_new();
+ if (!xattr[i]) {
+ op_ret = -ENOMEM;
+ goto out;
}
+ }
- /* dirty xattr represents if the file/dir needs heal. Unless all the
- * copies are healed, don't erase it */
- if (EC_COUNT (sources, ec->nodes) +
- EC_COUNT (healed_sinks, ec->nodes) == ec->nodes)
- erase_dirty = _gf_true;
- else
- op_ret = -ENOTCONN;
-
- /* Populate the xattr array */
- for (i = 0; i < ec->nodes; i++) {
- if (!sources[i] && !healed_sinks[i])
- continue;
- versions_xattr = GF_CALLOC (EC_VERSION_SIZE,
- sizeof(*versions_xattr),
- gf_common_mt_pointer);
- if (!versions_xattr) {
- op_ret = -ENOMEM;
- continue;
- }
-
- versions_xattr[type] = hton64(versions[source] - versions[i]);
- ret = dict_set_bin (xattr[i], EC_XATTR_VERSION,
- versions_xattr,
- (sizeof (*versions_xattr) * EC_VERSION_SIZE)
- );
- if (ret < 0) {
- op_ret = -ENOMEM;
- continue;
- }
-
- if (erase_dirty) {
- dirty_xattr = GF_CALLOC (EC_VERSION_SIZE,
- sizeof(*dirty_xattr),
- gf_common_mt_pointer);
- if (!dirty_xattr) {
- op_ret = -ENOMEM;
- continue;
- }
-
- dirty_xattr[type] = hton64(-dirty[i]);
- ret = dict_set_bin (xattr[i], EC_XATTR_DIRTY,
- dirty_xattr,
- (sizeof(*dirty_xattr) *
- EC_VERSION_SIZE)
- );
- if (ret < 0) {
- op_ret = -ENOMEM;
- continue;
- }
- }
+ /* dirty xattr represents if the file/dir needs heal. Unless all the
+ * copies are healed, don't erase it */
+ if (EC_COUNT(sources, ec->nodes) + EC_COUNT(healed_sinks, ec->nodes) ==
+ ec->nodes)
+ erase_dirty = _gf_true;
+ else
+ op_ret = -ENOTCONN;
+
+ /* Populate the xattr array */
+ for (i = 0; i < ec->nodes; i++) {
+ if (!sources[i] && !healed_sinks[i])
+ continue;
+ versions_xattr = GF_CALLOC(EC_VERSION_SIZE, sizeof(*versions_xattr),
+ gf_common_mt_pointer);
+ if (!versions_xattr) {
+ op_ret = -ENOMEM;
+ continue;
+ }
+
+ versions_xattr[type] = hton64(versions[source] - versions[i]);
+ ret = dict_set_bin(xattr[i], EC_XATTR_VERSION, versions_xattr,
+ (sizeof(*versions_xattr) * EC_VERSION_SIZE));
+ if (ret < 0) {
+ op_ret = -ENOMEM;
+ continue;
+ }
- if (memcmp (versions_xattr, allzero,
- (sizeof(*versions_xattr) * EC_VERSION_SIZE)) == 0) {
+ if (erase_dirty) {
+ dirty_xattr = GF_CALLOC(EC_VERSION_SIZE, sizeof(*dirty_xattr),
+ gf_common_mt_pointer);
+ if (!dirty_xattr) {
+ op_ret = -ENOMEM;
+ continue;
+ }
- if (!erase_dirty) {
- continue;
- }
+ dirty_xattr[type] = hton64(-dirty[i]);
+ ret = dict_set_bin(xattr[i], EC_XATTR_DIRTY, dirty_xattr,
+ (sizeof(*dirty_xattr) * EC_VERSION_SIZE));
+ if (ret < 0) {
+ op_ret = -ENOMEM;
+ continue;
+ }
+ }
- if (memcmp (dirty_xattr, allzero, (sizeof (*dirty_xattr)
- * EC_VERSION_SIZE)) == 0) {
- continue;
- }
- }
+ if (memcmp(versions_xattr, allzero,
+ (sizeof(*versions_xattr) * EC_VERSION_SIZE)) == 0) {
+ if (!erase_dirty) {
+ continue;
+ }
- on[i] = 1;
- call_count++;
+ if (memcmp(dirty_xattr, allzero,
+ (sizeof(*dirty_xattr) * EC_VERSION_SIZE)) == 0) {
+ continue;
+ }
}
- /* Update the bricks with xattr */
- if (call_count) {
- PARALLEL_FOP_ONLIST (ec->xl_list, on, ec->nodes, replies,
- frame, ec_wind_xattrop_parallel,
- &loc, GF_XATTROP_ADD_ARRAY64, xattr, NULL);
- ret = cluster_fop_success_fill (replies, ec->nodes, output);
- }
+ on[i] = 1;
+ call_count++;
+ }
- if (ret < call_count) {
- op_ret = -ENOTCONN;
- goto out;
- }
+ /* Update the bricks with xattr */
+ if (call_count) {
+ PARALLEL_FOP_ONLIST(ec->xl_list, on, ec->nodes, replies, frame,
+ ec_wind_xattrop_parallel, &loc,
+ GF_XATTROP_ADD_ARRAY64, xattr, NULL);
+ ret = cluster_fop_success_fill(replies, ec->nodes, output);
+ }
+
+ if (ret < call_count) {
+ op_ret = -ENOTCONN;
+ goto out;
+ }
out:
- /* Cleanup */
- if (xattr) {
- for (i = 0; i < ec->nodes; i++) {
- if (xattr[i])
- dict_unref (xattr[i]);
- }
- GF_FREE (xattr);
+ /* Cleanup */
+ if (xattr) {
+ for (i = 0; i < ec->nodes; i++) {
+ if (xattr[i])
+ dict_unref(xattr[i]);
}
- cluster_replies_wipe (replies, ec->nodes);
- loc_wipe (&loc);
- return op_ret;
+ GF_FREE(xattr);
+ }
+ cluster_replies_wipe(replies, ec->nodes);
+ loc_wipe(&loc);
+ return op_ret;
}
int
-ec_heal_metadata_find_direction (ec_t *ec, default_args_cbk_t *replies,
- uint64_t *versions, uint64_t *dirty,
- unsigned char *sources, unsigned char *healed_sinks)
-{
- uint64_t xattr[EC_VERSION_SIZE] = {0};
- uint64_t max_version = 0;
- int same_count = 0;
- int max_same_count = 0;
- int same_source = -1;
- int ret = 0;
- int i = 0;
- int j = 0;
- int *groups = NULL;
- struct iatt source_ia = {0};
- struct iatt child_ia = {0};
-
- groups = alloca0 (ec->nodes * sizeof(*groups));
- for (i = 0; i < ec->nodes; i++)
- groups[i] = -1;
-
- for (i = 0; i < ec->nodes; i++) {
- if (!replies[i].valid)
- continue;
- if (replies[i].op_ret < 0)
- continue;
- ret = ec_dict_get_array (replies[i].xdata, EC_XATTR_VERSION,
- xattr, EC_VERSION_SIZE);
- if (ret == 0) {
- versions[i] = xattr[EC_METADATA_TXN];
- }
-
- memset (xattr, 0, sizeof (xattr));
- ret = ec_dict_get_array (replies[i].xdata, EC_XATTR_DIRTY,
- xattr, EC_VERSION_SIZE);
- if (ret == 0) {
- dirty[i] = xattr[EC_METADATA_TXN];
- }
- if (groups[i] >= 0) /*Already part of group*/
- continue;
- groups[i] = i;
- same_count = 1;
- source_ia = replies[i].stat;
- for (j = i + 1; j < ec->nodes; j++) {
- if (!replies[j].valid || replies[j].op_ret < 0)
- continue;
- child_ia = replies[j].stat;
- if (!IA_EQUAL(source_ia, child_ia, gfid) ||
- !IA_EQUAL(source_ia, child_ia, type) ||
- !IA_EQUAL(source_ia, child_ia, prot) ||
- !IA_EQUAL(source_ia, child_ia, uid) ||
- !IA_EQUAL(source_ia, child_ia, gid))
- continue;
- if (!are_dicts_equal(replies[i].xdata, replies[j].xdata,
- ec_sh_key_match, NULL))
- continue;
- groups[j] = i;
- same_count++;
- }
-
- if (max_same_count < same_count) {
- max_same_count = same_count;
- same_source = i;
- }
+ec_heal_metadata_find_direction(ec_t *ec, default_args_cbk_t *replies,
+ uint64_t *versions, uint64_t *dirty,
+ unsigned char *sources,
+ unsigned char *healed_sinks)
+{
+ uint64_t xattr[EC_VERSION_SIZE] = {0};
+ uint64_t max_version = 0;
+ int same_count = 0;
+ int max_same_count = 0;
+ int same_source = -1;
+ int ret = 0;
+ int i = 0;
+ int j = 0;
+ int *groups = NULL;
+ struct iatt source_ia = {0};
+ struct iatt child_ia = {0};
+
+ groups = alloca0(ec->nodes * sizeof(*groups));
+ for (i = 0; i < ec->nodes; i++)
+ groups[i] = -1;
+
+ for (i = 0; i < ec->nodes; i++) {
+ if (!replies[i].valid)
+ continue;
+ if (replies[i].op_ret < 0)
+ continue;
+ ret = ec_dict_get_array(replies[i].xdata, EC_XATTR_VERSION, xattr,
+ EC_VERSION_SIZE);
+ if (ret == 0) {
+ versions[i] = xattr[EC_METADATA_TXN];
}
- if (max_same_count < ec->fragments) {
- ret = -EIO;
- goto out;
+ memset(xattr, 0, sizeof(xattr));
+ ret = ec_dict_get_array(replies[i].xdata, EC_XATTR_DIRTY, xattr,
+ EC_VERSION_SIZE);
+ if (ret == 0) {
+ dirty[i] = xattr[EC_METADATA_TXN];
+ }
+ if (groups[i] >= 0) /*Already part of group*/
+ continue;
+ groups[i] = i;
+ same_count = 1;
+ source_ia = replies[i].stat;
+ for (j = i + 1; j < ec->nodes; j++) {
+ if (!replies[j].valid || replies[j].op_ret < 0)
+ continue;
+ child_ia = replies[j].stat;
+ if (!IA_EQUAL(source_ia, child_ia, gfid) ||
+ !IA_EQUAL(source_ia, child_ia, type) ||
+ !IA_EQUAL(source_ia, child_ia, prot) ||
+ !IA_EQUAL(source_ia, child_ia, uid) ||
+ !IA_EQUAL(source_ia, child_ia, gid))
+ continue;
+ if (!are_dicts_equal(replies[i].xdata, replies[j].xdata,
+ ec_sh_key_match, NULL))
+ continue;
+ groups[j] = i;
+ same_count++;
+ }
+
+ if (max_same_count < same_count) {
+ max_same_count = same_count;
+ same_source = i;
}
+ }
- for (i = 0; i < ec->nodes; i++) {
- if (groups[i] == groups[same_source])
- sources[i] = 1;
- else if (replies[i].valid && replies[i].op_ret >= 0)
- healed_sinks[i] = 1;
- }
- for (i = 0; i < ec->nodes; i++) {
- if (sources[i] && (versions[i] > max_version)) {
- same_source = i;
- max_version = versions[i];
- }
+ if (max_same_count < ec->fragments) {
+ ret = -EIO;
+ goto out;
+ }
+
+ for (i = 0; i < ec->nodes; i++) {
+ if (groups[i] == groups[same_source])
+ sources[i] = 1;
+ else if (replies[i].valid && replies[i].op_ret >= 0)
+ healed_sinks[i] = 1;
+ }
+ for (i = 0; i < ec->nodes; i++) {
+ if (sources[i] && (versions[i] > max_version)) {
+ same_source = i;
+ max_version = versions[i];
}
- ret = same_source;
+ }
+ ret = same_source;
out:
- return ret;
+ return ret;
}
int
-__ec_heal_metadata_prepare (call_frame_t *frame, ec_t *ec, inode_t *inode,
- unsigned char *locked_on, default_args_cbk_t *replies,
- uint64_t *versions, uint64_t *dirty, unsigned char *sources,
- unsigned char *healed_sinks)
-{
- loc_t loc = {0};
- unsigned char *output = NULL;
- unsigned char *lookup_on = NULL;
- int ret = 0;
- int source = 0;
- default_args_cbk_t *greplies = NULL;
- int i = 0;
- EC_REPLIES_ALLOC (greplies, ec->nodes);
-
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
- output = alloca0 (ec->nodes);
- lookup_on = alloca0 (ec->nodes);
- ret = cluster_lookup (ec->xl_list, locked_on, ec->nodes, replies,
- output, frame, ec->xl, &loc, NULL);
- if (ret <= ec->fragments) {
- ret = -ENOTCONN;
- goto out;
- }
+__ec_heal_metadata_prepare(call_frame_t *frame, ec_t *ec, inode_t *inode,
+ unsigned char *locked_on,
+ default_args_cbk_t *replies, uint64_t *versions,
+ uint64_t *dirty, unsigned char *sources,
+ unsigned char *healed_sinks)
+{
+ loc_t loc = {0};
+ unsigned char *output = NULL;
+ unsigned char *lookup_on = NULL;
+ int ret = 0;
+ int source = 0;
+ default_args_cbk_t *greplies = NULL;
+ int i = 0;
+ EC_REPLIES_ALLOC(greplies, ec->nodes);
+
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+ output = alloca0(ec->nodes);
+ lookup_on = alloca0(ec->nodes);
+ ret = cluster_lookup(ec->xl_list, locked_on, ec->nodes, replies, output,
+ frame, ec->xl, &loc, NULL);
+ if (ret <= ec->fragments) {
+ ret = -ENOTCONN;
+ goto out;
+ }
- memcpy (lookup_on, output, ec->nodes);
- /*Use getxattr to get the filtered xattrs which filter internal xattrs*/
- ret = cluster_getxattr (ec->xl_list, lookup_on, ec->nodes, greplies,
- output, frame, ec->xl, &loc, NULL, NULL);
- for (i = 0; i < ec->nodes; i++) {
- if (lookup_on[i] && !output[i]) {
- replies[i].valid = 0;
- continue;
- }
- if (replies[i].xdata) {
- dict_unref (replies[i].xdata);
- replies[i].xdata = NULL;
- if (greplies[i].xattr)
- replies[i].xdata = dict_ref (greplies[i].xattr);
- }
+ memcpy(lookup_on, output, ec->nodes);
+ /*Use getxattr to get the filtered xattrs which filter internal xattrs*/
+ ret = cluster_getxattr(ec->xl_list, lookup_on, ec->nodes, greplies, output,
+ frame, ec->xl, &loc, NULL, NULL);
+ for (i = 0; i < ec->nodes; i++) {
+ if (lookup_on[i] && !output[i]) {
+ replies[i].valid = 0;
+ continue;
+ }
+ if (replies[i].xdata) {
+ dict_unref(replies[i].xdata);
+ replies[i].xdata = NULL;
+ if (greplies[i].xattr)
+ replies[i].xdata = dict_ref(greplies[i].xattr);
}
+ }
- source = ec_heal_metadata_find_direction (ec, replies, versions,
- dirty, sources, healed_sinks);
- if (source < 0) {
- ret = -EIO;
- goto out;
- }
- ret = source;
+ source = ec_heal_metadata_find_direction(ec, replies, versions, dirty,
+ sources, healed_sinks);
+ if (source < 0) {
+ ret = -EIO;
+ goto out;
+ }
+ ret = source;
out:
- cluster_replies_wipe (greplies, ec->nodes);
- loc_wipe (&loc);
- return ret;
+ cluster_replies_wipe(greplies, ec->nodes);
+ loc_wipe(&loc);
+ return ret;
}
/* Metadata heal */
int
-__ec_removexattr_sinks (call_frame_t *frame, ec_t *ec, inode_t *inode,
- int source, unsigned char *sources,
- unsigned char *healed_sinks,
- default_args_cbk_t *replies)
-{
- int i = 0;
- int ret = 0;
- loc_t loc = {0};
-
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
-
- for (i = 0; i < ec->nodes; i++) {
- if (i == source)
- continue;
- if (!sources[i] && !healed_sinks[i])
- continue;
- ret = dict_foreach (replies[i].xdata, ec_heal_xattr_clean,
- replies[source].xdata);
- if (ret < 0) {
- sources[i] = 0;
- healed_sinks[i] = 0;
- continue;
- }
-
- if (replies[i].xdata->count == 0) {
- continue;
- } else if (sources[i]) {
- /* This can happen if setxattr/removexattr succeeds on
- * the bricks but fails to update the version. This
- * will make sure that the xattrs are made equal after
- * heal*/
- sources[i] = 0;
- healed_sinks[i] = 1;
- }
+__ec_removexattr_sinks(call_frame_t *frame, ec_t *ec, inode_t *inode,
+ int source, unsigned char *sources,
+ unsigned char *healed_sinks, default_args_cbk_t *replies)
+{
+ int i = 0;
+ int ret = 0;
+ loc_t loc = {0};
+
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+
+ for (i = 0; i < ec->nodes; i++) {
+ if (i == source)
+ continue;
+ if (!sources[i] && !healed_sinks[i])
+ continue;
+ ret = dict_foreach(replies[i].xdata, ec_heal_xattr_clean,
+ replies[source].xdata);
+ if (ret < 0) {
+ sources[i] = 0;
+ healed_sinks[i] = 0;
+ continue;
+ }
- ret = syncop_removexattr (ec->xl_list[i], &loc, "",
- replies[i].xdata, NULL);
- if (ret < 0)
- healed_sinks[i] = 0;
+ if (replies[i].xdata->count == 0) {
+ continue;
+ } else if (sources[i]) {
+ /* This can happen if setxattr/removexattr succeeds on
+ * the bricks but fails to update the version. This
+ * will make sure that the xattrs are made equal after
+ * heal*/
+ sources[i] = 0;
+ healed_sinks[i] = 1;
}
- loc_wipe (&loc);
- if (EC_COUNT (healed_sinks, ec->nodes) == 0)
- return -ENOTCONN;
- return 0;
+ ret = syncop_removexattr(ec->xl_list[i], &loc, "", replies[i].xdata,
+ NULL);
+ if (ret < 0)
+ healed_sinks[i] = 0;
+ }
+
+ loc_wipe(&loc);
+ if (EC_COUNT(healed_sinks, ec->nodes) == 0)
+ return -ENOTCONN;
+ return 0;
}
int
-__ec_heal_metadata (call_frame_t *frame, ec_t *ec, inode_t *inode,
- unsigned char *locked_on, unsigned char *sources,
- unsigned char *healed_sinks)
-{
- loc_t loc = {0};
- int ret = 0;
- int source = 0;
- default_args_cbk_t *replies = NULL;
- default_args_cbk_t *sreplies = NULL;
- uint64_t *versions = NULL;
- uint64_t *dirty = NULL;
- unsigned char *output = NULL;
- dict_t *source_dict = NULL;
- struct iatt source_buf = {0};
-
- EC_REPLIES_ALLOC (replies, ec->nodes);
- EC_REPLIES_ALLOC (sreplies, ec->nodes);
-
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
- output = alloca0 (ec->nodes);
- versions = alloca0 (ec->nodes * sizeof (*versions));
- dirty = alloca0 (ec->nodes * sizeof (*dirty));
- source = __ec_heal_metadata_prepare (frame, ec, inode, locked_on, replies,
- versions, dirty, sources, healed_sinks);
- if (source < 0) {
- ret = -EIO;
- goto out;
- }
+__ec_heal_metadata(call_frame_t *frame, ec_t *ec, inode_t *inode,
+ unsigned char *locked_on, unsigned char *sources,
+ unsigned char *healed_sinks)
+{
+ loc_t loc = {0};
+ int ret = 0;
+ int source = 0;
+ default_args_cbk_t *replies = NULL;
+ default_args_cbk_t *sreplies = NULL;
+ uint64_t *versions = NULL;
+ uint64_t *dirty = NULL;
+ unsigned char *output = NULL;
+ dict_t *source_dict = NULL;
+ struct iatt source_buf = {0};
+
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ EC_REPLIES_ALLOC(sreplies, ec->nodes);
+
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+ output = alloca0(ec->nodes);
+ versions = alloca0(ec->nodes * sizeof(*versions));
+ dirty = alloca0(ec->nodes * sizeof(*dirty));
+ source = __ec_heal_metadata_prepare(frame, ec, inode, locked_on, replies,
+ versions, dirty, sources, healed_sinks);
+ if (source < 0) {
+ ret = -EIO;
+ goto out;
+ }
- if ((EC_COUNT (sources, ec->nodes) == ec->nodes) ||
- (EC_COUNT (healed_sinks, ec->nodes) == 0)) {
- ret = 0;
- goto erase_dirty;
- }
+ if ((EC_COUNT(sources, ec->nodes) == ec->nodes) ||
+ (EC_COUNT(healed_sinks, ec->nodes) == 0)) {
+ ret = 0;
+ goto erase_dirty;
+ }
- source_buf = replies[source].stat;
- ret = cluster_setattr (ec->xl_list, healed_sinks, ec->nodes, sreplies,
- output, frame, ec->xl, &loc,
- &source_buf, GF_SET_ATTR_MODE |
- GF_SET_ATTR_UID | GF_SET_ATTR_GID, NULL);
- /*In case the operation fails on some of the subvols*/
- memcpy (healed_sinks, output, ec->nodes);
- if (EC_COUNT (healed_sinks, ec->nodes) == 0) {
- ret = -ENOTCONN;
- goto out;
- }
+ source_buf = replies[source].stat;
+ ret = cluster_setattr(ec->xl_list, healed_sinks, ec->nodes, sreplies,
+ output, frame, ec->xl, &loc, &source_buf,
+ GF_SET_ATTR_MODE | GF_SET_ATTR_UID | GF_SET_ATTR_GID,
+ NULL);
+ /*In case the operation fails on some of the subvols*/
+ memcpy(healed_sinks, output, ec->nodes);
+ if (EC_COUNT(healed_sinks, ec->nodes) == 0) {
+ ret = -ENOTCONN;
+ goto out;
+ }
- ret = __ec_removexattr_sinks (frame, ec, inode, source, sources,
- healed_sinks, replies);
- if (ret < 0)
- goto out;
+ ret = __ec_removexattr_sinks(frame, ec, inode, source, sources,
+ healed_sinks, replies);
+ if (ret < 0)
+ goto out;
- source_dict = dict_ref (replies[source].xdata);
- if (dict_foreach_match (source_dict, ec_ignorable_key_match, NULL,
- dict_remove_foreach_fn, NULL) == -1) {
- ret = -ENOMEM;
- goto out;
- }
+ source_dict = dict_ref(replies[source].xdata);
+ if (dict_foreach_match(source_dict, ec_ignorable_key_match, NULL,
+ dict_remove_foreach_fn, NULL) == -1) {
+ ret = -ENOMEM;
+ goto out;
+ }
- ret = cluster_setxattr (ec->xl_list, healed_sinks, ec->nodes,
- replies, output, frame, ec->xl, &loc,
- source_dict, 0, NULL);
+ ret = cluster_setxattr(ec->xl_list, healed_sinks, ec->nodes, replies,
+ output, frame, ec->xl, &loc, source_dict, 0, NULL);
- EC_INTERSECT (healed_sinks, healed_sinks, output, ec->nodes);
- if (EC_COUNT (healed_sinks, ec->nodes) == 0) {
- ret = -ENOTCONN;
- goto out;
- }
+ EC_INTERSECT(healed_sinks, healed_sinks, output, ec->nodes);
+ if (EC_COUNT(healed_sinks, ec->nodes) == 0) {
+ ret = -ENOTCONN;
+ goto out;
+ }
erase_dirty:
- ret = ec_adjust_versions (frame, ec, EC_METADATA_TXN, inode, source,
- sources, healed_sinks, versions, dirty);
+ ret = ec_adjust_versions(frame, ec, EC_METADATA_TXN, inode, source, sources,
+ healed_sinks, versions, dirty);
out:
- if (source_dict)
- dict_unref (source_dict);
+ if (source_dict)
+ dict_unref(source_dict);
- loc_wipe (&loc);
- cluster_replies_wipe (replies, ec->nodes);
- cluster_replies_wipe (sreplies, ec->nodes);
- return ret;
+ loc_wipe(&loc);
+ cluster_replies_wipe(replies, ec->nodes);
+ cluster_replies_wipe(sreplies, ec->nodes);
+ return ret;
}
int
-ec_heal_metadata (call_frame_t *frame, ec_t *ec, inode_t *inode,
- unsigned char *sources, unsigned char *healed_sinks)
-{
- unsigned char *locked_on = NULL;
- unsigned char *up_subvols = NULL;
- unsigned char *output = NULL;
- int ret = 0;
- default_args_cbk_t *replies = NULL;
-
- EC_REPLIES_ALLOC (replies, ec->nodes);
- locked_on = alloca0(ec->nodes);
- output = alloca0(ec->nodes);
- up_subvols = alloca0(ec->nodes);
- ec_mask_to_char_array (ec->xl_up, up_subvols, ec->nodes);
- ret = cluster_inodelk (ec->xl_list, up_subvols, ec->nodes, replies,
- locked_on, frame, ec->xl, ec->xl->name, inode, 0,
- 0);
- {
- if (ret <= ec->fragments) {
- gf_msg_debug (ec->xl->name, 0, "%s: Skipping heal "
- "as only %d number of subvolumes could "
- "be locked", uuid_utoa (inode->gfid), ret);
- ret = -ENOTCONN;
- goto unlock;
- }
- ret = __ec_heal_metadata (frame, ec, inode, locked_on, sources,
- healed_sinks);
- }
+ec_heal_metadata(call_frame_t *frame, ec_t *ec, inode_t *inode,
+ unsigned char *sources, unsigned char *healed_sinks)
+{
+ unsigned char *locked_on = NULL;
+ unsigned char *up_subvols = NULL;
+ unsigned char *output = NULL;
+ int ret = 0;
+ default_args_cbk_t *replies = NULL;
+
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ locked_on = alloca0(ec->nodes);
+ output = alloca0(ec->nodes);
+ up_subvols = alloca0(ec->nodes);
+ ec_mask_to_char_array(ec->xl_up, up_subvols, ec->nodes);
+ ret = cluster_inodelk(ec->xl_list, up_subvols, ec->nodes, replies,
+ locked_on, frame, ec->xl, ec->xl->name, inode, 0, 0);
+ {
+ if (ret <= ec->fragments) {
+ gf_msg_debug(ec->xl->name, 0,
+ "%s: Skipping heal "
+ "as only %d number of subvolumes could "
+ "be locked",
+ uuid_utoa(inode->gfid), ret);
+ ret = -ENOTCONN;
+ goto unlock;
+ }
+ ret = __ec_heal_metadata(frame, ec, inode, locked_on, sources,
+ healed_sinks);
+ }
unlock:
- cluster_uninodelk (ec->xl_list, locked_on, ec->nodes, replies, output,
- frame, ec->xl, ec->xl->name, inode, 0, 0);
- cluster_replies_wipe (replies, ec->nodes);
- return ret;
+ cluster_uninodelk(ec->xl_list, locked_on, ec->nodes, replies, output, frame,
+ ec->xl, ec->xl->name, inode, 0, 0);
+ cluster_replies_wipe(replies, ec->nodes);
+ return ret;
}
/*entry heal*/
int
-__ec_heal_entry_prepare (call_frame_t *frame, ec_t *ec, inode_t *inode,
- unsigned char *locked_on, uint64_t *versions,
- uint64_t *dirty, unsigned char *sources,
- unsigned char *healed_sinks)
-{
- loc_t loc = {0};
- int source = 0;
- int ret = 0;
- default_args_cbk_t *replies = NULL;
- unsigned char *output = NULL;
- dict_t *xdata = NULL;
-
- EC_REPLIES_ALLOC (replies, ec->nodes);
-
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
- xdata = dict_new ();
- if (!xdata) {
- ret = -ENOMEM;
- goto out;
- }
+__ec_heal_entry_prepare(call_frame_t *frame, ec_t *ec, inode_t *inode,
+ unsigned char *locked_on, uint64_t *versions,
+ uint64_t *dirty, unsigned char *sources,
+ unsigned char *healed_sinks)
+{
+ loc_t loc = {0};
+ int source = 0;
+ int ret = 0;
+ default_args_cbk_t *replies = NULL;
+ unsigned char *output = NULL;
+ dict_t *xdata = NULL;
+
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+ xdata = dict_new();
+ if (!xdata) {
+ ret = -ENOMEM;
+ goto out;
+ }
- if (dict_set_uint64(xdata, EC_XATTR_VERSION, 0) ||
- dict_set_uint64(xdata, EC_XATTR_DIRTY, 0)) {
- ret = -ENOMEM;
- goto out;
- }
+ if (dict_set_uint64(xdata, EC_XATTR_VERSION, 0) ||
+ dict_set_uint64(xdata, EC_XATTR_DIRTY, 0)) {
+ ret = -ENOMEM;
+ goto out;
+ }
- output = alloca0 (ec->nodes);
- ret = cluster_lookup (ec->xl_list, locked_on, ec->nodes, replies,
- output, frame, ec->xl, &loc, xdata);
- if (ret <= ec->fragments) {
- ret = -ENOTCONN;
- goto out;
- }
+ output = alloca0(ec->nodes);
+ ret = cluster_lookup(ec->xl_list, locked_on, ec->nodes, replies, output,
+ frame, ec->xl, &loc, xdata);
+ if (ret <= ec->fragments) {
+ ret = -ENOTCONN;
+ goto out;
+ }
- source = ec_heal_entry_find_direction (ec, replies, versions,
- dirty, sources, healed_sinks);
- if (source < 0) {
- ret = -EIO;
- goto out;
- }
- ret = source;
+ source = ec_heal_entry_find_direction(ec, replies, versions, dirty, sources,
+ healed_sinks);
+ if (source < 0) {
+ ret = -EIO;
+ goto out;
+ }
+ ret = source;
out:
- if (xdata)
- dict_unref (xdata);
- loc_wipe (&loc);
- cluster_replies_wipe (replies, ec->nodes);
- return ret;
+ if (xdata)
+ dict_unref(xdata);
+ loc_wipe(&loc);
+ cluster_replies_wipe(replies, ec->nodes);
+ return ret;
}
int32_t
-ec_set_new_entry_dirty (ec_t *ec, loc_t *loc, struct iatt *ia,
- call_frame_t *frame, xlator_t *this, unsigned char *on)
-{
- dict_t *xattr = NULL;
- int32_t ret = -1;
- default_args_cbk_t *replies = NULL;
- unsigned char *output = NULL;
- uint64_t dirty[EC_VERSION_SIZE] = {1, 1};
- loc_t newloc = {0};
-
- /*Symlinks don't have any data to be healed*/
- if (ia->ia_type == IA_IFLNK)
- dirty[EC_DATA_TXN] = 0;
-
- newloc.inode = inode_ref (loc->inode);
- gf_uuid_copy (newloc.gfid, ia->ia_gfid);
- EC_REPLIES_ALLOC (replies, ec->nodes);
- output = alloca0 (ec->nodes);
- xattr = dict_new();
- if (!xattr) {
- ret = -ENOMEM;
- goto out;
- }
+ec_set_new_entry_dirty(ec_t *ec, loc_t *loc, struct iatt *ia,
+ call_frame_t *frame, xlator_t *this, unsigned char *on)
+{
+ dict_t *xattr = NULL;
+ int32_t ret = -1;
+ default_args_cbk_t *replies = NULL;
+ unsigned char *output = NULL;
+ uint64_t dirty[EC_VERSION_SIZE] = {1, 1};
+ loc_t newloc = {0};
+
+ /*Symlinks don't have any data to be healed*/
+ if (ia->ia_type == IA_IFLNK)
+ dirty[EC_DATA_TXN] = 0;
+
+ newloc.inode = inode_ref(loc->inode);
+ gf_uuid_copy(newloc.gfid, ia->ia_gfid);
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ output = alloca0(ec->nodes);
+ xattr = dict_new();
+ if (!xattr) {
+ ret = -ENOMEM;
+ goto out;
+ }
- ret = ec_dict_set_array (xattr, EC_XATTR_DIRTY, dirty,
- EC_VERSION_SIZE);
- if (ret)
- goto out;
+ ret = ec_dict_set_array(xattr, EC_XATTR_DIRTY, dirty, EC_VERSION_SIZE);
+ if (ret)
+ goto out;
- ret = cluster_xattrop (ec->xl_list, on, ec->nodes, replies, output,
- frame, ec->xl, &newloc,
- GF_XATTROP_ADD_ARRAY64, xattr, NULL);
+ ret = cluster_xattrop(ec->xl_list, on, ec->nodes, replies, output, frame,
+ ec->xl, &newloc, GF_XATTROP_ADD_ARRAY64, xattr, NULL);
+
+ if (ret < ec->fragments) {
+ ret = -ENOTCONN;
+ goto out;
+ }
- if (ret < ec->fragments) {
- ret = -ENOTCONN;
- goto out;
- }
out:
- if (xattr)
- dict_unref (xattr);
- cluster_replies_wipe (replies, ec->nodes);
- loc_wipe (&newloc);
- return ret;
+ if (xattr)
+ dict_unref(xattr);
+ cluster_replies_wipe(replies, ec->nodes);
+ loc_wipe(&newloc);
+ return ret;
}
/*Name heal*/
int
-ec_delete_stale_name (dict_t *gfid_db, char *key, data_t *d, void *data)
-{
- struct ec_name_data *name_data = data;
- struct iatt *ia = NULL;
- ec_t *ec = NULL;
- loc_t loc = {0};
- unsigned char *same = data_to_bin (d);
- default_args_cbk_t *replies = NULL;
- unsigned char *output = NULL;
- int ret = 0;
- int estale_count = 0;
- int i = 0;
- call_frame_t *frame = name_data->frame;
-
- ec = name_data->frame->this->private;
- EC_REPLIES_ALLOC (replies, ec->nodes);
- if (EC_COUNT (same, ec->nodes) >= ec->fragments) {
- ret = 0;
- goto out;
- }
+ec_delete_stale_name(dict_t *gfid_db, char *key, data_t *d, void *data)
+{
+ struct ec_name_data *name_data = data;
+ struct iatt *ia = NULL;
+ ec_t *ec = NULL;
+ loc_t loc = {0};
+ unsigned char *same = data_to_bin(d);
+ default_args_cbk_t *replies = NULL;
+ unsigned char *output = NULL;
+ int ret = 0;
+ int estale_count = 0;
+ int i = 0;
+ call_frame_t *frame = name_data->frame;
+ uuid_t gfid;
+
+ ec = name_data->frame->this->private;
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ if (EC_COUNT(same, ec->nodes) >= ec->fragments) {
+ ret = 0;
+ goto out;
+ }
- loc.inode = inode_new (name_data->parent->table);
- if (!loc.inode) {
- ret = -ENOMEM;
- goto out;
- }
- gf_uuid_parse (key, loc.gfid);
- output = alloca0(ec->nodes);
- ret = cluster_lookup (ec->xl_list, name_data->participants, ec->nodes,
- replies, output, name_data->frame, ec->xl, &loc,
- NULL);
+ loc.parent = inode_ref(name_data->parent);
+ loc.inode = inode_new(name_data->parent->table);
+ if (!loc.inode) {
+ ret = -ENOMEM;
+ goto out;
+ }
- for (i = 0; i < ec->nodes; i++) {
- if (!replies[i].valid)
- continue;
- if (replies[i].op_ret == -1) {
- if (replies[i].op_errno == ESTALE ||
- replies[i].op_errno == ENOENT)
- estale_count++;
- else
- name_data->participants[i] = 0;
- }
+ gf_uuid_parse(key, gfid);
+ gf_uuid_copy(loc.pargfid, name_data->parent->gfid);
+ loc.name = name_data->name;
+ output = alloca0(ec->nodes);
+ ret = cluster_lookup(ec->xl_list, name_data->participants, ec->nodes,
+ replies, output, name_data->frame, ec->xl, &loc, NULL);
+
+ for (i = 0; i < ec->nodes; i++) {
+ if (!replies[i].valid)
+ continue;
+ if (replies[i].op_ret == -1) {
+ if (replies[i].op_errno == ESTALE || replies[i].op_errno == ENOENT)
+ estale_count++;
+ else
+ name_data->participants[i] = 0;
+ } else if (gf_uuid_compare(gfid, replies[i].stat.ia_gfid)) {
+ estale_count++;
+ gf_msg_debug(ec->xl->name, 0, "%s/%s: different gfid as %s",
+ uuid_utoa(name_data->parent->gfid), name_data->name,
+ key);
}
+ }
- if (estale_count <= ec->redundancy) {
- /* We have at least ec->fragments number of fragments, so the
- * file is recoverable, so don't delete it*/
+ if (estale_count <= ec->redundancy) {
+ /* We have at least ec->fragments number of fragments, so the
+ * file is recoverable, so don't delete it*/
- /* Please note that the lookup call above could fail with
- * ENOTCONN on all subvoumes and still this branch will be
- * true, but in those cases conservatively we decide to not
- * delete the file until we are sure*/
- ret = 0;
- goto out;
- }
+ /* Please note that the lookup call above could fail with
+ * ENOTCONN on all subvoumes and still this branch will be
+ * true, but in those cases conservatively we decide to not
+ * delete the file until we are sure*/
+ ret = 0;
+ goto out;
+ }
- /*Noway to recover, delete the name*/
- loc_wipe (&loc);
- loc.parent = inode_ref (name_data->parent);
- gf_uuid_copy (loc.pargfid, loc.parent->gfid);
- loc.name = name_data->name;
- for (i = 0; i < ec->nodes; i++) {
- if (same[i] && replies[i].valid && (replies[i].op_ret == 0)) {
- ia = &replies[i].stat;
- break;
- }
+ /*Noway to recover, delete the name*/
+ loc_wipe(&loc);
+ loc.parent = inode_ref(name_data->parent);
+ gf_uuid_copy(loc.pargfid, loc.parent->gfid);
+ loc.name = name_data->name;
+ for (i = 0; i < ec->nodes; i++) {
+ if (same[i] && replies[i].valid && (replies[i].op_ret == 0)) {
+ ia = &replies[i].stat;
+ break;
}
+ }
- if (!ia) {
- ret = -ENOTCONN;
- goto out;
- }
+ if (!ia) {
+ ret = -ENOTCONN;
+ goto out;
+ }
- if (IA_ISDIR (ia->ia_type)) {
- ret = cluster_rmdir (ec->xl_list, same, ec->nodes, replies,
- output, frame, ec->xl, &loc, 1, NULL);
- gf_msg_debug (ec->xl->name, 0, "cluster rmdir succeeded on %d "
- "nodes", ret);
- } else {
- ret = cluster_unlink (ec->xl_list, same, ec->nodes, replies,
- output, frame, ec->xl, &loc, 0, NULL);
- gf_msg_debug (ec->xl->name, 0, "cluster unlink succeeded on %d "
- "nodes", ret);
- }
+ if (IA_ISDIR(ia->ia_type)) {
+ ret = cluster_rmdir(ec->xl_list, same, ec->nodes, replies, output,
+ frame, ec->xl, &loc, 1, NULL);
+ gf_msg_debug(ec->xl->name, 0,
+ "cluster rmdir succeeded on %d "
+ "nodes",
+ ret);
+ } else {
+ ret = cluster_unlink(ec->xl_list, same, ec->nodes, replies, output,
+ frame, ec->xl, &loc, 0, NULL);
+ gf_msg_debug(ec->xl->name, 0,
+ "cluster unlink succeeded on %d "
+ "nodes",
+ ret);
+ }
- for (i = 0; i < ec->nodes; i++) {
- if (output[i]) {
- same[i] = 0;
- name_data->enoent[i] = 1;
- } else {
- /*op failed*/
- if (same[i])
- name_data->participants[i] = 0;
- }
+ for (i = 0; i < ec->nodes; i++) {
+ if (output[i]) {
+ same[i] = 0;
+ name_data->enoent[i] = 1;
+ } else {
+ /*op failed*/
+ if (same[i])
+ name_data->participants[i] = 0;
}
- ret = 0;
- /*This will help in making decisions about creating names*/
- dict_del (gfid_db, key);
+ }
+ ret = 0;
+ /*This will help in making decisions about creating names*/
+ dict_del(gfid_db, key);
out:
- if (ret < 0) {
- gf_msg_debug (ec->xl->name, 0, "%s/%s: heal failed %s",
- uuid_utoa (name_data->parent->gfid), name_data->name,
- strerror (-ret));
- }
- cluster_replies_wipe (replies, ec->nodes);
- loc_wipe (&loc);
- return ret;
+ if (ret < 0) {
+ gf_msg_debug(ec->xl->name, 0, "%s/%s: heal failed %s",
+ uuid_utoa(name_data->parent->gfid), name_data->name,
+ strerror(-ret));
+ }
+ cluster_replies_wipe(replies, ec->nodes);
+ loc_wipe(&loc);
+ return ret;
}
int
-ec_delete_stale_names (call_frame_t *frame, ec_t *ec, inode_t *parent,
- char *name, default_args_cbk_t *replies, dict_t *gfid_db,
- unsigned char *enoent, unsigned char *gfidless,
- unsigned char *participants)
+ec_delete_stale_names(call_frame_t *frame, ec_t *ec, inode_t *parent,
+ char *name, default_args_cbk_t *replies, dict_t *gfid_db,
+ unsigned char *enoent, unsigned char *gfidless,
+ unsigned char *participants)
{
- struct ec_name_data name_data = {0};
+ struct ec_name_data name_data = {0};
- name_data.enoent = enoent;
- name_data.gfidless = gfidless;
- name_data.participants = participants;
- name_data.name = name;
- name_data.parent = parent;
- name_data.frame = frame;
- name_data.replies = replies;
- return dict_foreach (gfid_db, ec_delete_stale_name, &name_data);
+ name_data.enoent = enoent;
+ name_data.gfidless = gfidless;
+ name_data.participants = participants;
+ name_data.name = name;
+ name_data.parent = parent;
+ name_data.frame = frame;
+ name_data.replies = replies;
+ return dict_foreach(gfid_db, ec_delete_stale_name, &name_data);
}
int
-_assign_same (dict_t *dict, char *key, data_t *value, void *data)
+_assign_same(dict_t *dict, char *key, data_t *value, void *data)
{
- struct ec_name_data *name_data = data;
+ struct ec_name_data *name_data = data;
- name_data->same = data_to_bin (value);
- return 0;
+ name_data->same = data_to_bin(value);
+ return 0;
}
int
-ec_create_name (call_frame_t *frame, ec_t *ec, inode_t *parent, char *name,
- default_args_cbk_t *lookup_replies, dict_t *gfid_db,
- unsigned char *enoent, unsigned char *participants)
-{
- int ret = 0;
- int i = 0;
- struct ec_name_data name_data = {0};
- struct iatt *ia = NULL;
- unsigned char *output = 0;
- unsigned char *output1 = 0;
- unsigned char *on = NULL;
- default_args_cbk_t *replies = NULL;
- loc_t loc = {0};
- loc_t srcloc = {0};
- unsigned char *link = NULL;
- unsigned char *create = NULL;
- dict_t *xdata = NULL;
- char *linkname = NULL;
- ec_config_t config;
- /* There should be just one gfid key */
- EC_REPLIES_ALLOC (replies, ec->nodes);
- if (gfid_db->count != 1) {
- ret = -EINVAL;
- goto out;
- }
-
- ret = dict_foreach (gfid_db, _assign_same, &name_data);
- if (ret < 0)
- goto out;
- /*There should at least be one valid success reply with gfid*/
- for (i = 0; i < ec->nodes; i++)
- if (name_data.same[i])
- break;
+ec_create_name(call_frame_t *frame, ec_t *ec, inode_t *parent, char *name,
+ default_args_cbk_t *lookup_replies, dict_t *gfid_db,
+ unsigned char *enoent, unsigned char *participants)
+{
+ int ret = 0;
+ int i = 0;
+ struct ec_name_data name_data = {0};
+ struct iatt *ia = NULL;
+ unsigned char *output = 0;
+ unsigned char *output1 = 0;
+ unsigned char *on = NULL;
+ default_args_cbk_t *replies = NULL;
+ loc_t loc = {0};
+ loc_t srcloc = {0};
+ unsigned char *link = NULL;
+ unsigned char *create = NULL;
+ dict_t *xdata = NULL;
+ char *linkname = NULL;
+ ec_config_t config;
+
+ /* There should be just one gfid key */
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ if (gfid_db->count != 1) {
+ ret = -EINVAL;
+ goto out;
+ }
- if (i == ec->nodes) {
- ret = -EINVAL;
- goto out;
- }
+ ret = dict_foreach(gfid_db, _assign_same, &name_data);
+ if (ret < 0)
+ goto out;
+ /*There should at least be one valid success reply with gfid*/
+ for (i = 0; i < ec->nodes; i++)
+ if (name_data.same[i])
+ break;
- ia = &lookup_replies[i].stat;
- xdata = dict_new ();
- loc.parent = inode_ref (parent);
- gf_uuid_copy (loc.pargfid, parent->gfid);
- loc.inode = inode_new (parent->table);
- if (loc.inode)
- srcloc.inode = inode_ref (loc.inode);
- gf_uuid_copy (srcloc.gfid, ia->ia_gfid);
- if (!loc.inode || !xdata || dict_set_static_bin (xdata, "gfid-req",
- ia->ia_gfid,
- sizeof (ia->ia_gfid))) {
- ret = -ENOMEM;
- goto out;
- }
- loc.name = name;
- link = alloca0 (ec->nodes);
- create = alloca0 (ec->nodes);
- on = alloca0 (ec->nodes);
- output = alloca0 (ec->nodes);
- output1 = alloca0 (ec->nodes);
+ if (i == ec->nodes) {
+ ret = -EINVAL;
+ goto out;
+ }
- for (i = 0; i < ec->nodes; i++) {
- if (!lookup_replies[i].valid)
- continue;
- if (lookup_replies[i].op_ret)
- continue;
- on[i] = 1;
- }
- switch (ia->ia_type) {
+ ia = &lookup_replies[i].stat;
+ xdata = dict_new();
+ loc.parent = inode_ref(parent);
+ gf_uuid_copy(loc.pargfid, parent->gfid);
+ loc.inode = inode_new(parent->table);
+ if (loc.inode)
+ srcloc.inode = inode_ref(loc.inode);
+ gf_uuid_copy(srcloc.gfid, ia->ia_gfid);
+ if (!loc.inode || !xdata ||
+ dict_set_static_bin(xdata, "gfid-req", ia->ia_gfid,
+ sizeof(ia->ia_gfid))) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ loc.name = name;
+ link = alloca0(ec->nodes);
+ create = alloca0(ec->nodes);
+ on = alloca0(ec->nodes);
+ output = alloca0(ec->nodes);
+ output1 = alloca0(ec->nodes);
+
+ for (i = 0; i < ec->nodes; i++) {
+ if (!lookup_replies[i].valid)
+ continue;
+ if (lookup_replies[i].op_ret)
+ continue;
+ on[i] = 1;
+ }
+ switch (ia->ia_type) {
case IA_IFDIR:
- ec_set_new_entry_dirty (ec, &loc, ia, frame, ec->xl, on);
- (void) cluster_mkdir (ec->xl_list, enoent, ec->nodes,
- replies, output, frame, ec->xl, &loc,
- st_mode_from_ia (ia->ia_prot,
- ia->ia_type), 0, xdata);
- break;
+ ec_set_new_entry_dirty(ec, &loc, ia, frame, ec->xl, on);
+ (void)cluster_mkdir(
+ ec->xl_list, enoent, ec->nodes, replies, output, frame, ec->xl,
+ &loc, st_mode_from_ia(ia->ia_prot, ia->ia_type), 0, xdata);
+ break;
case IA_IFLNK:
- /*Check for hard links and create/link*/
- ret = cluster_lookup (ec->xl_list, enoent, ec->nodes,
- replies, output, frame, ec->xl,
- &srcloc, NULL);
- for (i = 0; i < ec->nodes; i++) {
- if (output[i]) {
- link[i] = 1;
- } else {
- if (replies[i].op_errno == ENOENT ||
- replies[i].op_errno == ESTALE) {
- create[i] = 1;
- }
- }
+ /*Check for hard links and create/link*/
+ ret = cluster_lookup(ec->xl_list, enoent, ec->nodes, replies,
+ output, frame, ec->xl, &srcloc, NULL);
+ for (i = 0; i < ec->nodes; i++) {
+ if (output[i]) {
+ link[i] = 1;
+ } else {
+ if (replies[i].op_errno == ENOENT ||
+ replies[i].op_errno == ESTALE) {
+ create[i] = 1;
+ }
}
-
- if (EC_COUNT (link, ec->nodes)) {
- cluster_link (ec->xl_list, link, ec->nodes,
- replies, output1, frame, ec->xl,
- &srcloc, &loc, NULL);
+ }
+
+ if (EC_COUNT(link, ec->nodes)) {
+ cluster_link(ec->xl_list, link, ec->nodes, replies, output1,
+ frame, ec->xl, &srcloc, &loc, NULL);
+ }
+
+ if (EC_COUNT(create, ec->nodes)) {
+ cluster_readlink(ec->xl_list, name_data.same, ec->nodes,
+ replies, output, frame, ec->xl, &srcloc, 4096,
+ NULL);
+ if (EC_COUNT(output, ec->nodes) == 0) {
+ ret = -ENOTCONN;
+ goto out;
}
- if (EC_COUNT (create, ec->nodes)) {
- cluster_readlink (ec->xl_list, name_data.same,
- ec->nodes, replies, output,
- frame, ec->xl, &srcloc, 4096,
- NULL);
- if (EC_COUNT (output, ec->nodes) == 0) {
- ret = -ENOTCONN;
- goto out;
- }
-
- for (i = 0; i < ec->nodes; i++) {
- if (output[i])
- break;
- }
- linkname = alloca0 (strlen(replies[i].buf) + 1);
- strcpy (linkname, replies[i].buf);
- ec_set_new_entry_dirty (ec, &loc, ia, frame,
- ec->xl, on);
- cluster_symlink (ec->xl_list, create, ec->nodes,
- replies, output, frame, ec->xl,
- linkname, &loc, 0, xdata);
+ for (i = 0; i < ec->nodes; i++) {
+ if (output[i])
+ break;
}
- for (i = 0; i < ec->nodes; i++)
- if (output1[i])
- output[i] = 1;
- break;
+ linkname = alloca0(strlen(replies[i].buf) + 1);
+ strcpy(linkname, replies[i].buf);
+ ec_set_new_entry_dirty(ec, &loc, ia, frame, ec->xl, on);
+ cluster_symlink(ec->xl_list, create, ec->nodes, replies, output,
+ frame, ec->xl, linkname, &loc, 0, xdata);
+ }
+ for (i = 0; i < ec->nodes; i++)
+ if (output1[i])
+ output[i] = 1;
+ break;
case IA_IFREG:
- ec_set_new_entry_dirty (ec, &loc, ia,
- frame, ec->xl, on);
- config.version = EC_CONFIG_VERSION;
- config.algorithm = EC_CONFIG_ALGORITHM;
- config.gf_word_size = EC_GF_BITS;
- config.bricks = ec->nodes;
- config.redundancy = ec->redundancy;
- config.chunk_size = EC_METHOD_CHUNK_SIZE;
-
- ret = ec_dict_set_config(xdata, EC_XATTR_CONFIG, &config);
- if (ret != 0) {
- goto out;
- }
+ ec_set_new_entry_dirty(ec, &loc, ia, frame, ec->xl, on);
+ config.version = EC_CONFIG_VERSION;
+ config.algorithm = EC_CONFIG_ALGORITHM;
+ config.gf_word_size = EC_GF_BITS;
+ config.bricks = ec->nodes;
+ config.redundancy = ec->redundancy;
+ config.chunk_size = EC_METHOD_CHUNK_SIZE;
- /* Fall through */
+ ret = ec_dict_set_config(xdata, EC_XATTR_CONFIG, &config);
+ if (ret != 0) {
+ goto out;
+ }
+
+ /* Fall through */
default:
- ret = dict_set_int32 (xdata, GLUSTERFS_INTERNAL_FOP_KEY,
- 1);
- if (ret)
- goto out;
- ret = cluster_mknod (ec->xl_list, enoent, ec->nodes,
- replies, output, frame, ec->xl,
- &loc, st_mode_from_ia (ia->ia_prot,
- ia->ia_type),
- makedev(ia_major(ia->ia_rdev), ia_minor(ia->ia_rdev)),
- 0, xdata);
- break;
- }
+ ret = dict_set_int32(xdata, GLUSTERFS_INTERNAL_FOP_KEY, 1);
+ if (ret)
+ goto out;
+ ret = cluster_mknod(
+ ec->xl_list, enoent, ec->nodes, replies, output, frame, ec->xl,
+ &loc, st_mode_from_ia(ia->ia_prot, ia->ia_type),
+ makedev(ia_major(ia->ia_rdev), ia_minor(ia->ia_rdev)), 0,
+ xdata);
+ break;
+ }
- for (i = 0; i < ec->nodes; i++) {
- if (enoent[i] && !output[i])
- participants[i] = 0;
- }
+ for (i = 0; i < ec->nodes; i++) {
+ if (enoent[i] && !output[i])
+ participants[i] = 0;
+ }
- ret = 0;
+ ret = 0;
out:
- if (ret < 0)
- gf_msg_debug (ec->xl->name, 0, "%s/%s: heal failed %s",
- uuid_utoa (parent->gfid), name, strerror (-ret));
- cluster_replies_wipe (replies, ec->nodes);
- loc_wipe (&loc);
- loc_wipe (&srcloc);
- if (xdata)
- dict_unref (xdata);
- return ret;
+ if (ret < 0)
+ gf_msg_debug(ec->xl->name, 0, "%s/%s: heal failed %s",
+ uuid_utoa(parent->gfid), name, strerror(-ret));
+ cluster_replies_wipe(replies, ec->nodes);
+ loc_wipe(&loc);
+ loc_wipe(&srcloc);
+ if (xdata)
+ dict_unref(xdata);
+ return ret;
}
int
-__ec_heal_name (call_frame_t *frame, ec_t *ec, inode_t *parent, char *name,
- unsigned char *participants)
-{
- unsigned char *output = NULL;
- unsigned char *enoent = NULL;
- default_args_cbk_t *replies = NULL;
- dict_t *xdata = NULL;
- dict_t *gfid_db = NULL;
- int ret = 0;
- loc_t loc = {0};
- int i = 0;
- struct iatt *ia = NULL;
- char gfid[64] = {0};
- unsigned char *same = NULL;
- unsigned char *gfidless = NULL;
-
- EC_REPLIES_ALLOC (replies, ec->nodes);
- loc.parent = inode_ref (parent);
- loc.inode = inode_new (parent->table);
- gf_uuid_copy (loc.pargfid, parent->gfid);
- loc.name = name;
- xdata = dict_new ();
- gfid_db = dict_new ();
- if (!xdata || !gfid_db || !loc.inode) {
- ret = -ENOMEM;
- goto out;
- }
+__ec_heal_name(call_frame_t *frame, ec_t *ec, inode_t *parent, char *name,
+ unsigned char *participants)
+{
+ unsigned char *output = NULL;
+ unsigned char *enoent = NULL;
+ default_args_cbk_t *replies = NULL;
+ dict_t *xdata = NULL;
+ dict_t *gfid_db = NULL;
+ int ret = 0;
+ loc_t loc = {0};
+ int i = 0;
+ struct iatt *ia = NULL;
+ char gfid[64] = {0};
+ unsigned char *same = NULL;
+ unsigned char *gfidless = NULL;
+
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ loc.parent = inode_ref(parent);
+ loc.inode = inode_new(parent->table);
+ gf_uuid_copy(loc.pargfid, parent->gfid);
+ loc.name = name;
+ xdata = dict_new();
+ gfid_db = dict_new();
+ if (!xdata || !gfid_db || !loc.inode) {
+ ret = -ENOMEM;
+ goto out;
+ }
- ret = dict_set_int32 (xdata, GF_GFIDLESS_LOOKUP, 1);
- if (ret) {
- ret = -ENOMEM;
- goto out;
- }
+ ret = dict_set_int32(xdata, GF_GFIDLESS_LOOKUP, 1);
+ if (ret) {
+ ret = -ENOMEM;
+ goto out;
+ }
- output = alloca0 (ec->nodes);
- gfidless = alloca0 (ec->nodes);
- enoent = alloca0 (ec->nodes);
- ret = cluster_lookup (ec->xl_list, participants, ec->nodes, replies,
- output, frame, ec->xl, &loc, NULL);
- for (i = 0; i < ec->nodes; i++) {
- if (!replies[i].valid)
- continue;
-
- if (replies[i].op_ret == -1) {
- /*If ESTALE comes here, that means parent dir is not
- * present, nothing to do there, so reset participants
- * for that brick*/
- if (replies[i].op_errno == ENOENT)
- enoent[i] = 1;
- else
- participants[i] = 0;
- continue;
- }
- ia = &replies[i].stat;
- if (gf_uuid_is_null (ia->ia_gfid)) {
- if (IA_ISDIR (ia->ia_type) || ia->ia_size == 0)
- gfidless[i] = 1;
- else
- participants[i] = 0;
- } else {
- uuid_utoa_r (ia->ia_gfid, gfid);
- ret = dict_get_bin (gfid_db, gfid, (void **)&same);
- if (ret < 0) {
- same = alloca0(ec->nodes);
- }
- same[i] = 1;
- if (ret < 0) {
- ret = dict_set_static_bin (gfid_db, gfid, same,
- ec->nodes);
- }
- if (ret < 0)
- goto out;
- }
+ output = alloca0(ec->nodes);
+ gfidless = alloca0(ec->nodes);
+ enoent = alloca0(ec->nodes);
+ ret = cluster_lookup(ec->xl_list, participants, ec->nodes, replies, output,
+ frame, ec->xl, &loc, NULL);
+ for (i = 0; i < ec->nodes; i++) {
+ if (!replies[i].valid)
+ continue;
+
+ if (replies[i].op_ret == -1) {
+ /*If ESTALE comes here, that means parent dir is not
+ * present, nothing to do there, so reset participants
+ * for that brick*/
+ if (replies[i].op_errno == ENOENT)
+ enoent[i] = 1;
+ else
+ participants[i] = 0;
+ continue;
+ }
+ ia = &replies[i].stat;
+ if (gf_uuid_is_null(ia->ia_gfid)) {
+ if (IA_ISDIR(ia->ia_type) || ia->ia_size == 0)
+ gfidless[i] = 1;
+ else
+ participants[i] = 0;
+ } else {
+ uuid_utoa_r(ia->ia_gfid, gfid);
+ ret = dict_get_bin(gfid_db, gfid, (void **)&same);
+ if (ret < 0) {
+ same = alloca0(ec->nodes);
+ }
+ same[i] = 1;
+ if (ret < 0) {
+ ret = dict_set_static_bin(gfid_db, gfid, same, ec->nodes);
+ }
+ if (ret < 0)
+ goto out;
}
+ }
- ret = ec_delete_stale_names (frame, ec, parent, name, replies, gfid_db,
- enoent, gfidless, participants);
+ ret = ec_delete_stale_names(frame, ec, parent, name, replies, gfid_db,
+ enoent, gfidless, participants);
- if (gfid_db->count == 0) {
- /* All entries seem to be stale entries and deleted,
- * nothing more to do.*/
- goto out;
- }
+ if (gfid_db->count == 0) {
+ /* All entries seem to be stale entries and deleted,
+ * nothing more to do.*/
+ goto out;
+ }
- if (gfid_db->count > 1) {
- gf_msg (ec->xl->name, GF_LOG_INFO, 0,
- EC_MSG_HEAL_FAIL, "%s/%s: Not able to heal",
- uuid_utoa (parent->gfid), name);
- memset (participants, 0, ec->nodes);
- goto out;
- }
+ if (gfid_db->count > 1) {
+ gf_msg(ec->xl->name, GF_LOG_INFO, 0, EC_MSG_HEAL_FAIL,
+ "%s/%s: Not able to heal", uuid_utoa(parent->gfid), name);
+ memset(participants, 0, ec->nodes);
+ goto out;
+ }
- EC_INTERSECT (enoent, enoent, participants, ec->nodes);
- if (EC_COUNT (enoent, ec->nodes) == 0) {
- ret = 0;
- goto out;
- }
+ EC_INTERSECT(enoent, enoent, participants, ec->nodes);
+ if (EC_COUNT(enoent, ec->nodes) == 0) {
+ ret = 0;
+ goto out;
+ }
- ret = ec_create_name (frame, ec, parent, name, replies, gfid_db, enoent,
- participants);
+ ret = ec_create_name(frame, ec, parent, name, replies, gfid_db, enoent,
+ participants);
+ if (ret >= 0) {
+ /* If ec_create_name() succeeded we return 1 to indicate that a new
+ * file has been created and it will need to be healed. */
+ ret = 1;
+ }
out:
- cluster_replies_wipe (replies, ec->nodes);
- loc_wipe (&loc);
- if (xdata)
- dict_unref (xdata);
- if (gfid_db)
- dict_unref (gfid_db);
- return ret;
+ cluster_replies_wipe(replies, ec->nodes);
+ loc_wipe(&loc);
+ if (xdata)
+ dict_unref(xdata);
+ if (gfid_db)
+ dict_unref(gfid_db);
+ return ret;
}
int
-ec_heal_name (call_frame_t *frame, ec_t *ec, inode_t *parent, char *name,
- unsigned char *participants)
-{
- int ret = 0;
- default_args_cbk_t *replies = NULL;
- unsigned char *output = NULL;
- unsigned char *locked_on = NULL;
- loc_t loc = {0};
-
- loc.parent = inode_ref (parent);
- loc.name = name;
- loc.inode = inode_new (parent->table);
- if (!loc.inode) {
- ret = -ENOMEM;
- goto out;
- }
+ec_heal_name(call_frame_t *frame, ec_t *ec, inode_t *parent, char *name,
+ unsigned char *participants)
+{
+ int ret = 0;
+ default_args_cbk_t *replies = NULL;
+ unsigned char *output = NULL;
+ unsigned char *locked_on = NULL;
+ loc_t loc = {0};
+
+ loc.parent = inode_ref(parent);
+ loc.name = name;
+ loc.inode = inode_new(parent->table);
+ if (!loc.inode) {
+ ret = -ENOMEM;
+ goto out;
+ }
- EC_REPLIES_ALLOC (replies, ec->nodes);
- output = alloca0 (ec->nodes);
- locked_on = alloca0 (ec->nodes);
- ret = cluster_inodelk (ec->xl_list, participants, ec->nodes, replies,
- locked_on, frame, ec->xl, ec->xl->name, parent,
- 0, 0);
- {
- if (ret <= ec->fragments) {
- gf_msg_debug (ec->xl->name, 0, "%s/%s: Skipping "
- "heal as only %d number of subvolumes could "
- "be locked", uuid_utoa (parent->gfid), name,
- ret);
- ret = -ENOTCONN;
- goto unlock;
- }
- EC_INTERSECT (participants, participants, locked_on, ec->nodes);
- ret = __ec_heal_name (frame, ec, parent, name, participants);
- }
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ output = alloca0(ec->nodes);
+ locked_on = alloca0(ec->nodes);
+ ret = cluster_inodelk(ec->xl_list, participants, ec->nodes, replies,
+ locked_on, frame, ec->xl, ec->xl->name, parent, 0, 0);
+ {
+ if (ret <= ec->fragments) {
+ gf_msg_debug(ec->xl->name, 0,
+ "%s/%s: Skipping "
+ "heal as only %d number of subvolumes could "
+ "be locked",
+ uuid_utoa(parent->gfid), name, ret);
+ ret = -ENOTCONN;
+ goto unlock;
+ }
+ EC_INTERSECT(participants, participants, locked_on, ec->nodes);
+ ret = __ec_heal_name(frame, ec, parent, name, participants);
+ }
unlock:
- cluster_uninodelk (ec->xl_list, locked_on, ec->nodes, replies, output,
- frame, ec->xl, ec->xl->name, parent, 0, 0);
+ cluster_uninodelk(ec->xl_list, locked_on, ec->nodes, replies, output, frame,
+ ec->xl, ec->xl->name, parent, 0, 0);
out:
- cluster_replies_wipe (replies, ec->nodes);
- loc_wipe (&loc);
- return ret;
+ cluster_replies_wipe(replies, ec->nodes);
+ loc_wipe(&loc);
+ return ret;
}
int
-ec_name_heal_handler (xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
- void *data)
-{
- struct ec_name_data *name_data = data;
- xlator_t *this = THIS;
- ec_t *ec = this->private;
- unsigned char *name_on = alloca0 (ec->nodes);
- int i = 0;
- int ret = 0;
+ec_name_heal_handler(xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
+ void *data)
+{
+ struct ec_name_data *name_data = data;
+ xlator_t *this = THIS;
+ ec_t *ec = this->private;
+ unsigned char *name_on = alloca0(ec->nodes);
+ int i = 0;
+ int ret = 0;
+
+ if (ec->shutdown) {
+ gf_msg_debug(this->name, 0,
+ "Cancelling directory heal "
+ "because EC is stopping.");
+ return -ENOTCONN;
+ }
- if (ec->shutdown) {
- gf_msg_debug(this->name, 0, "Cancelling directory heal "
- "because EC is stopping.");
- return -ENOTCONN;
- }
+ memcpy(name_on, name_data->participants, ec->nodes);
+ ret = ec_heal_name(name_data->frame, ec, parent->inode, entry->d_name,
+ name_on);
- memcpy (name_on, name_data->participants, ec->nodes);
- ret = ec_heal_name (name_data->frame, ec, parent->inode,
- entry->d_name, name_on);
+ if (ret < 0) {
+ memset(name_on, 0, ec->nodes);
+ } else {
+ name_data->heal_pending += ret;
+ }
- if (ret < 0)
- memset (name_on, 0, ec->nodes);
+ for (i = 0; i < ec->nodes; i++)
+ if (name_data->participants[i] && !name_on[i])
+ name_data->failed_on[i] = 1;
- for (i = 0; i < ec->nodes; i++)
- if (name_data->participants[i] && !name_on[i])
- name_data->failed_on[i] = 1;
- return 0;
+ return 0;
}
int
-ec_heal_names (call_frame_t *frame, ec_t *ec, inode_t *inode,
- unsigned char *participants)
-{
- int i = 0;
- int j = 0;
- loc_t loc = {0};
- struct ec_name_data name_data = {0};
- int ret = 0;
-
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
- name_data.frame = frame;
- name_data.participants = participants;
- name_data.failed_on = alloca0(ec->nodes);;
-
- for (i = 0; i < ec->nodes; i++) {
- if (!participants[i])
- continue;
- ret = syncop_dir_scan (ec->xl_list[i], &loc,
- GF_CLIENT_PID_SELF_HEALD, &name_data,
- ec_name_heal_handler);
- if (ret < 0) {
- break;
- }
- for (j = 0; j < ec->nodes; j++)
- if (name_data.failed_on[j])
- participants[j] = 0;
+ec_heal_names(call_frame_t *frame, ec_t *ec, inode_t *inode,
+ unsigned char *participants, uint32_t *pending)
+{
+ int i = 0;
+ int j = 0;
+ loc_t loc = {0};
+ struct ec_name_data name_data = {0};
+ int ret = 0;
+
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+ name_data.frame = frame;
+ name_data.participants = participants;
+ name_data.failed_on = alloca0(ec->nodes);
+ name_data.heal_pending = 0;
+
+ for (i = 0; i < ec->nodes; i++) {
+ if (!participants[i])
+ continue;
+ ret = syncop_dir_scan(ec->xl_list[i], &loc, GF_CLIENT_PID_SELF_HEALD,
+ &name_data, ec_name_heal_handler);
+ if (ret < 0) {
+ break;
+ }
+ for (j = 0; j < ec->nodes; j++)
+ if (name_data.failed_on[j])
+ participants[j] = 0;
- if (EC_COUNT (participants, ec->nodes) <= ec->fragments) {
- ret = -ENOTCONN;
- break;
- }
+ if (EC_COUNT(participants, ec->nodes) <= ec->fragments) {
+ ret = -ENOTCONN;
+ break;
}
- loc_wipe (&loc);
- return ret;
+ }
+ *pending += name_data.heal_pending;
+
+ loc_wipe(&loc);
+ return ret;
}
int
-__ec_heal_entry (call_frame_t *frame, ec_t *ec, inode_t *inode,
- unsigned char *heal_on, unsigned char *sources,
- unsigned char *healed_sinks)
-{
- unsigned char *locked_on = NULL;
- unsigned char *output = NULL;
- uint64_t *versions = NULL;
- uint64_t *dirty = NULL;
- unsigned char *participants = NULL;
- default_args_cbk_t *replies = NULL;
- int ret = 0;
- int source = 0;
- int i = 0;
-
- locked_on = alloca0(ec->nodes);
- output = alloca0(ec->nodes);
- versions = alloca0 (ec->nodes * sizeof (*versions));
- dirty = alloca0 (ec->nodes * sizeof (*dirty));
-
- EC_REPLIES_ALLOC (replies, ec->nodes);
- ret = cluster_inodelk (ec->xl_list, heal_on, ec->nodes, replies,
- locked_on, frame, ec->xl, ec->xl->name, inode,
- 0, 0);
- {
- if (ret <= ec->fragments) {
- gf_msg_debug (ec->xl->name, 0, "%s: Skipping heal "
- "as only %d number of subvolumes could "
- "be locked", uuid_utoa (inode->gfid), ret);
- ret = -ENOTCONN;
- goto unlock;
- }
- ret = __ec_heal_entry_prepare (frame, ec, inode, locked_on,
- versions, dirty, sources,
- healed_sinks);
- source = ret;
- }
+__ec_heal_entry(call_frame_t *frame, ec_t *ec, inode_t *inode,
+ unsigned char *heal_on, unsigned char *sources,
+ unsigned char *healed_sinks, uint32_t *pending)
+{
+ unsigned char *locked_on = NULL;
+ unsigned char *output = NULL;
+ uint64_t *versions = NULL;
+ uint64_t *dirty = NULL;
+ unsigned char *participants = NULL;
+ default_args_cbk_t *replies = NULL;
+ int ret = 0;
+ int source = 0;
+ int i = 0;
+
+ locked_on = alloca0(ec->nodes);
+ output = alloca0(ec->nodes);
+ versions = alloca0(ec->nodes * sizeof(*versions));
+ dirty = alloca0(ec->nodes * sizeof(*dirty));
+
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ ret = cluster_inodelk(ec->xl_list, heal_on, ec->nodes, replies, locked_on,
+ frame, ec->xl, ec->xl->name, inode, 0, 0);
+ {
+ if (ret <= ec->fragments) {
+ gf_msg_debug(ec->xl->name, 0,
+ "%s: Skipping heal "
+ "as only %d number of subvolumes could "
+ "be locked",
+ uuid_utoa(inode->gfid), ret);
+ ret = -ENOTCONN;
+ goto unlock;
+ }
+ ret = __ec_heal_entry_prepare(frame, ec, inode, locked_on, versions,
+ dirty, sources, healed_sinks);
+ source = ret;
+ }
unlock:
- cluster_uninodelk (ec->xl_list, locked_on, ec->nodes, replies, output,
- frame, ec->xl, ec->xl->name, inode, 0, 0);
- if (ret < 0)
- goto out;
+ cluster_uninodelk(ec->xl_list, locked_on, ec->nodes, replies, output, frame,
+ ec->xl, ec->xl->name, inode, 0, 0);
+ if (ret < 0)
+ goto out;
- participants = alloca0 (ec->nodes);
- for (i = 0; i < ec->nodes; i++) {
- if (sources[i] || healed_sinks[i])
- participants[i] = 1;
- }
- ret = ec_heal_names (frame, ec, inode, participants);
+ participants = alloca0(ec->nodes);
+ for (i = 0; i < ec->nodes; i++) {
+ if (sources[i] || healed_sinks[i])
+ participants[i] = 1;
+ }
+ ret = ec_heal_names(frame, ec, inode, participants, pending);
- if (EC_COUNT (participants, ec->nodes) <= ec->fragments)
- goto out;
+ if (EC_COUNT(participants, ec->nodes) <= ec->fragments)
+ goto out;
- for (i = 0; i < ec->nodes; i++) {
- if (!participants[i]) {
- sources[i] = 0;
- healed_sinks[i] = 0;
- }
+ for (i = 0; i < ec->nodes; i++) {
+ if (!participants[i]) {
+ sources[i] = 0;
+ healed_sinks[i] = 0;
}
+ }
- ec_adjust_versions (frame, ec, EC_DATA_TXN, inode, source,
- sources, healed_sinks, versions, dirty);
+ ec_adjust_versions(frame, ec, EC_DATA_TXN, inode, source, sources,
+ healed_sinks, versions, dirty);
out:
- cluster_replies_wipe (replies, ec->nodes);
- return ret;
+ cluster_replies_wipe(replies, ec->nodes);
+ return ret;
}
int
-ec_heal_entry (call_frame_t *frame, ec_t *ec, inode_t *inode,
- unsigned char *sources, unsigned char *healed_sinks)
-{
- unsigned char *locked_on = NULL;
- unsigned char *up_subvols = NULL;
- unsigned char *output = NULL;
- char selfheal_domain[1024] = {0};
- int ret = 0;
- default_args_cbk_t *replies = NULL;
-
- EC_REPLIES_ALLOC (replies, ec->nodes);
- locked_on = alloca0(ec->nodes);
- output = alloca0(ec->nodes);
- up_subvols = alloca0(ec->nodes);
-
- sprintf (selfheal_domain, "%s:self-heal", ec->xl->name);
- ec_mask_to_char_array (ec->xl_up, up_subvols, ec->nodes);
- /*If other processes are already doing the heal, don't block*/
- ret = cluster_tiebreaker_inodelk (ec->xl_list, up_subvols, ec->nodes,
- replies, locked_on, frame, ec->xl,
- selfheal_domain, inode, 0, 0);
- {
- if (ret <= ec->fragments) {
- gf_msg_debug (ec->xl->name, 0, "%s: Skipping heal "
- "as only %d number of subvolumes could "
- "be locked", uuid_utoa (inode->gfid), ret);
- ret = -ENOTCONN;
- goto unlock;
- }
- ret = __ec_heal_entry (frame, ec, inode, locked_on,
- sources, healed_sinks);
- }
+ec_heal_entry(call_frame_t *frame, ec_t *ec, inode_t *inode,
+ unsigned char *sources, unsigned char *healed_sinks,
+ uint32_t *pending)
+{
+ unsigned char *locked_on = NULL;
+ unsigned char *up_subvols = NULL;
+ unsigned char *output = NULL;
+ char selfheal_domain[1024] = {0};
+ int ret = 0;
+ default_args_cbk_t *replies = NULL;
+
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ locked_on = alloca0(ec->nodes);
+ output = alloca0(ec->nodes);
+ up_subvols = alloca0(ec->nodes);
+
+ sprintf(selfheal_domain, "%s:self-heal", ec->xl->name);
+ ec_mask_to_char_array(ec->xl_up, up_subvols, ec->nodes);
+ /*If other processes are already doing the heal, don't block*/
+ ret = cluster_tiebreaker_inodelk(ec->xl_list, up_subvols, ec->nodes,
+ replies, locked_on, frame, ec->xl,
+ selfheal_domain, inode, 0, 0);
+ {
+ if (ret <= ec->fragments) {
+ gf_msg_debug(ec->xl->name, 0,
+ "%s: Skipping heal "
+ "as only %d number of subvolumes could "
+ "be locked",
+ uuid_utoa(inode->gfid), ret);
+ ret = -ENOTCONN;
+ goto unlock;
+ }
+ ret = __ec_heal_entry(frame, ec, inode, locked_on, sources,
+ healed_sinks, pending);
+ }
unlock:
- cluster_uninodelk (ec->xl_list, locked_on, ec->nodes, replies, output,
- frame, ec->xl, selfheal_domain, inode, 0, 0);
- cluster_replies_wipe (replies, ec->nodes);
- return ret;
+ cluster_uninodelk(ec->xl_list, locked_on, ec->nodes, replies, output, frame,
+ ec->xl, selfheal_domain, inode, 0, 0);
+ cluster_replies_wipe(replies, ec->nodes);
+ return ret;
}
/*Find direction for data heal and heal info*/
int
-ec_heal_data_find_direction (ec_t *ec, default_args_cbk_t *replies,
- uint64_t *data_versions,
- uint64_t *dirty, uint64_t *size, unsigned char *sources,
- unsigned char *healed_sinks,
- gf_boolean_t check_ondisksize, int which)
-{
- uint64_t xattr[EC_VERSION_SIZE] = {0};
- char version_size[128] = {0};
- dict_t *version_size_db = NULL;
- unsigned char *same = NULL;
- int max_same_count = 0;
- int source = 0;
- int i = 0;
- int ret = 0;
- dict_t *dict = NULL;
- uint64_t source_size = 0;
-
- version_size_db = dict_new ();
- if (!version_size_db) {
- ret = -ENOMEM;
- goto out;
+ec_heal_data_find_direction(ec_t *ec, default_args_cbk_t *replies,
+ uint64_t *data_versions, uint64_t *dirty,
+ uint64_t *size, unsigned char *sources,
+ unsigned char *healed_sinks,
+ gf_boolean_t check_ondisksize, int which)
+{
+ uint64_t xattr[EC_VERSION_SIZE] = {0};
+ char version_size[128] = {0};
+ dict_t *version_size_db = NULL;
+ unsigned char *same = NULL;
+ int max_same_count = 0;
+ int source = 0;
+ int i = 0;
+ int ret = 0;
+ dict_t *dict = NULL;
+ uint64_t source_size = 0;
+
+ version_size_db = dict_new();
+ if (!version_size_db) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ for (i = 0; i < ec->nodes; i++) {
+ if (!replies[i].valid)
+ continue;
+ if (replies[i].op_ret < 0)
+ continue;
+ dict = (which == EC_COMBINE_XDATA) ? replies[i].xdata
+ : replies[i].xattr;
+
+ ret = ec_dict_get_array(dict, EC_XATTR_VERSION, xattr, EC_VERSION_SIZE);
+ if (ret == 0) {
+ data_versions[i] = xattr[EC_DATA_TXN];
}
- for (i = 0; i < ec->nodes; i++) {
- if (!replies[i].valid)
- continue;
- if (replies[i].op_ret < 0)
- continue;
- dict = (which == EC_COMBINE_XDATA) ? replies[i].xdata :
- replies[i].xattr;
-
- ret = ec_dict_get_array (dict, EC_XATTR_VERSION,
- xattr, EC_VERSION_SIZE);
- if (ret == 0) {
- data_versions[i] = xattr[EC_DATA_TXN];
- }
+ memset(xattr, 0, sizeof(xattr));
+ ret = ec_dict_get_array(dict, EC_XATTR_DIRTY, xattr, EC_VERSION_SIZE);
+ if (ret == 0) {
+ dirty[i] = xattr[EC_DATA_TXN];
+ }
+ ret = ec_dict_del_number(dict, EC_XATTR_SIZE, &size[i]);
+ /*Build a db of same metadata and data version and size*/
+ snprintf(version_size, sizeof(version_size), "%" PRIu64 "-%" PRIu64,
+ data_versions[i], size[i]);
- memset (xattr, 0, sizeof (xattr));
- ret = ec_dict_get_array (dict, EC_XATTR_DIRTY,
- xattr, EC_VERSION_SIZE);
- if (ret == 0) {
- dirty[i] = xattr[EC_DATA_TXN];
- }
- ret = ec_dict_del_number (dict, EC_XATTR_SIZE,
- &size[i]);
- /*Build a db of same metadata and data version and size*/
- snprintf (version_size, sizeof (version_size),
- "%"PRIu64"-%"PRIu64, data_versions[i], size[i]);
-
- ret = dict_get_bin (version_size_db, version_size,
- (void **)&same);
- if (ret < 0) {
- same = alloca0 (ec->nodes);
- }
+ ret = dict_get_bin(version_size_db, version_size, (void **)&same);
+ if (ret < 0) {
+ same = alloca0(ec->nodes);
+ }
- same[i] = 1;
- if (max_same_count < EC_COUNT (same, ec->nodes)) {
- max_same_count = EC_COUNT (same, ec->nodes);
- source = i;
- }
+ same[i] = 1;
+ if (max_same_count < EC_COUNT(same, ec->nodes)) {
+ max_same_count = EC_COUNT(same, ec->nodes);
+ source = i;
+ }
- if (ret < 0) {
- ret = dict_set_static_bin (version_size_db,
- version_size, same, ec->nodes);
- }
+ if (ret < 0) {
+ ret = dict_set_static_bin(version_size_db, version_size, same,
+ ec->nodes);
+ }
- if (ret < 0) {
- ret = -ENOMEM;
- goto out;
- }
+ if (ret < 0) {
+ ret = -ENOMEM;
+ goto out;
}
- /* If we don't have ec->fragments number of same version,size it is not
- * recoverable*/
- if (max_same_count < ec->fragments) {
- ret = -EIO;
- goto out;
- } else {
- snprintf (version_size, sizeof (version_size),
- "%"PRIu64"-%"PRIu64, data_versions[source],
- size[source]);
-
- ret = dict_get_bin (version_size_db, version_size,
- (void **)&same);
- if (ret < 0)
- goto out;
- memcpy (sources, same, ec->nodes);
- for (i = 0; i < ec->nodes; i++) {
- if (replies[i].valid && (replies[i].op_ret == 0) &&
- !sources[i])
- healed_sinks[i] = 1;
- }
+ }
+ /* If we don't have ec->fragments number of same version,size it is not
+ * recoverable*/
+ if (max_same_count < ec->fragments) {
+ ret = -EIO;
+ goto out;
+ } else {
+ snprintf(version_size, sizeof(version_size), "%" PRIu64 "-%" PRIu64,
+ data_versions[source], size[source]);
+
+ ret = dict_get_bin(version_size_db, version_size, (void **)&same);
+ if (ret < 0)
+ goto out;
+ memcpy(sources, same, ec->nodes);
+ for (i = 0; i < ec->nodes; i++) {
+ if (replies[i].valid && (replies[i].op_ret == 0) && !sources[i])
+ healed_sinks[i] = 1;
}
+ }
- /* There could be files with versions, size same but on disk ia_size
- * could be different because of disk crashes, mark them as sinks as
- * well*/
+ /* There could be files with versions, size same but on disk ia_size
+ * could be different because of disk crashes, mark them as sinks as
+ * well*/
- if (check_ondisksize) {
- source_size = size[source];
- ec_adjust_size_up (ec, &source_size, _gf_true);
+ if (check_ondisksize) {
+ source_size = size[source];
+ ec_adjust_size_up(ec, &source_size, _gf_true);
- for (i = 0; i < ec->nodes; i++) {
- if (sources[i]) {
- if (replies[i].stat.ia_size != source_size) {
- sources[i] = 0;
- healed_sinks[i] = 1;
- max_same_count--;
- } else {
- source = i;
- }
- }
- }
- if (max_same_count < ec->fragments) {
- ret = -EIO;
- goto out;
+ for (i = 0; i < ec->nodes; i++) {
+ if (sources[i]) {
+ if (replies[i].stat.ia_size != source_size) {
+ sources[i] = 0;
+ healed_sinks[i] = 1;
+ max_same_count--;
+ } else {
+ source = i;
}
+ }
+ }
+ if (max_same_count < ec->fragments) {
+ ret = -EIO;
+ goto out;
}
+ }
- ret = source;
+ ret = source;
out:
- if (version_size_db)
- dict_unref (version_size_db);
- return ret;
+ if (version_size_db)
+ dict_unref(version_size_db);
+ return ret;
}
int
-__ec_heal_data_prepare (call_frame_t *frame, ec_t *ec, fd_t *fd,
- unsigned char *locked_on, uint64_t *versions,
- uint64_t *dirty, uint64_t *size, unsigned char *sources,
- unsigned char *healed_sinks, unsigned char *trim,
- struct iatt *stbuf)
-{
- default_args_cbk_t *replies = NULL;
- default_args_cbk_t *fstat_replies = NULL;
- unsigned char *output = NULL;
- unsigned char *fstat_output = NULL;
- dict_t *xattrs = NULL;
- uint64_t zero_array[2] = {0};
- int source = 0;
- int ret = 0;
- uint64_t zero_value = 0;
- int i = 0;
-
- EC_REPLIES_ALLOC (replies, ec->nodes);
- EC_REPLIES_ALLOC (fstat_replies, ec->nodes);
- output = alloca0(ec->nodes);
- fstat_output = alloca0(ec->nodes);
- xattrs = dict_new ();
- if (!xattrs ||
- dict_set_static_bin (xattrs, EC_XATTR_VERSION, zero_array,
- sizeof (zero_array)) ||
- dict_set_static_bin (xattrs, EC_XATTR_DIRTY, zero_array,
- sizeof (zero_array)) ||
- dict_set_static_bin (xattrs, EC_XATTR_SIZE, &zero_value,
- sizeof (zero_value))) {
- ret = -ENOMEM;
- goto out;
- }
+__ec_heal_data_prepare(call_frame_t *frame, ec_t *ec, fd_t *fd,
+ unsigned char *locked_on, uint64_t *versions,
+ uint64_t *dirty, uint64_t *size, unsigned char *sources,
+ unsigned char *healed_sinks, unsigned char *trim,
+ struct iatt *stbuf)
+{
+ default_args_cbk_t *replies = NULL;
+ default_args_cbk_t *fstat_replies = NULL;
+ unsigned char *output = NULL;
+ unsigned char *fstat_output = NULL;
+ dict_t *xattrs = NULL;
+ uint64_t zero_array[2] = {0};
+ int source = 0;
+ int ret = 0;
+ uint64_t zero_value = 0;
+ int i = 0;
+
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ EC_REPLIES_ALLOC(fstat_replies, ec->nodes);
+ output = alloca0(ec->nodes);
+ fstat_output = alloca0(ec->nodes);
+ xattrs = dict_new();
+ if (!xattrs ||
+ dict_set_static_bin(xattrs, EC_XATTR_VERSION, zero_array,
+ sizeof(zero_array)) ||
+ dict_set_static_bin(xattrs, EC_XATTR_DIRTY, zero_array,
+ sizeof(zero_array)) ||
+ dict_set_static_bin(xattrs, EC_XATTR_SIZE, &zero_value,
+ sizeof(zero_value))) {
+ ret = -ENOMEM;
+ goto out;
+ }
- ret = cluster_fxattrop (ec->xl_list, locked_on, ec->nodes,
- replies, output, frame, ec->xl, fd,
- GF_XATTROP_ADD_ARRAY64, xattrs, NULL);
+ ret = cluster_fxattrop(ec->xl_list, locked_on, ec->nodes, replies, output,
+ frame, ec->xl, fd, GF_XATTROP_ADD_ARRAY64, xattrs,
+ NULL);
- ret = cluster_fstat (ec->xl_list, locked_on, ec->nodes, fstat_replies,
- fstat_output, frame, ec->xl, fd, NULL);
+ ret = cluster_fstat(ec->xl_list, locked_on, ec->nodes, fstat_replies,
+ fstat_output, frame, ec->xl, fd, NULL);
- for (i = 0; i < ec->nodes; i++) {
- output[i] = output[i] && fstat_output[i];
- replies[i].valid = output[i];
- if (output[i])
- replies[i].stat = fstat_replies[i].stat;
- }
+ for (i = 0; i < ec->nodes; i++) {
+ output[i] = output[i] && fstat_output[i];
+ replies[i].valid = output[i];
+ if (output[i])
+ replies[i].stat = fstat_replies[i].stat;
+ }
- if (EC_COUNT (output, ec->nodes) <= ec->fragments) {
- ret = -ENOTCONN;
- goto out;
- }
+ if (EC_COUNT(output, ec->nodes) <= ec->fragments) {
+ ret = -ENOTCONN;
+ goto out;
+ }
- source = ec_heal_data_find_direction (ec, replies, versions,
- dirty, size, sources,
- healed_sinks, _gf_true,
- EC_COMBINE_DICT);
- ret = source;
- if (ret < 0)
- goto out;
+ source = ec_heal_data_find_direction(ec, replies, versions, dirty, size,
+ sources, healed_sinks, _gf_true,
+ EC_COMBINE_DICT);
+ ret = source;
+ if (ret < 0)
+ goto out;
- if (stbuf)
- *stbuf = replies[source].stat;
+ if (stbuf)
+ *stbuf = replies[source].stat;
- for (i = 0; i < ec->nodes; i++) {
- if (healed_sinks[i]) {
- if (replies[i].stat.ia_size)
- trim[i] = 1;
- }
+ for (i = 0; i < ec->nodes; i++) {
+ if (healed_sinks[i]) {
+ if (replies[i].stat.ia_size)
+ trim[i] = 1;
}
+ }
- if (EC_COUNT(sources, ec->nodes) < ec->fragments) {
- ret = -ENOTCONN;
- goto out;
- }
+ if (EC_COUNT(sources, ec->nodes) < ec->fragments) {
+ ret = -ENOTCONN;
+ goto out;
+ }
- ret = source;
+ ret = source;
out:
- if (xattrs)
- dict_unref (xattrs);
- cluster_replies_wipe (replies, ec->nodes);
- cluster_replies_wipe (fstat_replies, ec->nodes);
- if (ret < 0) {
- gf_msg_debug (ec->xl->name, 0, "%s: heal failed %s",
- uuid_utoa (fd->inode->gfid), strerror (-ret));
- } else {
- gf_msg_debug (ec->xl->name, 0, "%s: sources: %d, sinks: "
- "%d", uuid_utoa (fd->inode->gfid),
- EC_COUNT (sources, ec->nodes),
- EC_COUNT (healed_sinks, ec->nodes));
- }
- return ret;
+ if (xattrs)
+ dict_unref(xattrs);
+ cluster_replies_wipe(replies, ec->nodes);
+ cluster_replies_wipe(fstat_replies, ec->nodes);
+ if (ret < 0) {
+ gf_msg_debug(ec->xl->name, 0, "%s: heal failed %s",
+ uuid_utoa(fd->inode->gfid), strerror(-ret));
+ } else {
+ gf_msg_debug(ec->xl->name, 0,
+ "%s: sources: %d, sinks: "
+ "%d",
+ uuid_utoa(fd->inode->gfid), EC_COUNT(sources, ec->nodes),
+ EC_COUNT(healed_sinks, ec->nodes));
+ }
+ return ret;
}
int
-__ec_heal_mark_sinks (call_frame_t *frame, ec_t *ec, fd_t *fd,
- uint64_t *versions, unsigned char *healed_sinks)
-{
- int i = 0;
- int ret = 0;
- unsigned char *mark = NULL;
- dict_t *xattrs = NULL;
- default_args_cbk_t *replies = NULL;
- unsigned char *output = NULL;
- uint64_t versions_xattr[2] = {0};
-
- EC_REPLIES_ALLOC (replies, ec->nodes);
- xattrs = dict_new ();
- if (!xattrs) {
- ret = -ENOMEM;
- goto out;
- }
+__ec_heal_mark_sinks(call_frame_t *frame, ec_t *ec, fd_t *fd,
+ uint64_t *versions, unsigned char *healed_sinks)
+{
+ int i = 0;
+ int ret = 0;
+ unsigned char *mark = NULL;
+ dict_t *xattrs = NULL;
+ default_args_cbk_t *replies = NULL;
+ unsigned char *output = NULL;
+ uint64_t versions_xattr[2] = {0};
+
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ xattrs = dict_new();
+ if (!xattrs) {
+ ret = -ENOMEM;
+ goto out;
+ }
- mark = alloca0 (ec->nodes);
- for (i = 0; i < ec->nodes; i++) {
- if (!healed_sinks[i])
- continue;
- if ((versions[i] >> EC_SELFHEAL_BIT) & 1)
- continue;
- mark[i] = 1;
- }
+ mark = alloca0(ec->nodes);
+ for (i = 0; i < ec->nodes; i++) {
+ if (!healed_sinks[i])
+ continue;
+ if ((versions[i] >> EC_SELFHEAL_BIT) & 1)
+ continue;
+ mark[i] = 1;
+ }
- if (EC_COUNT (mark, ec->nodes) == 0)
- return 0;
+ if (EC_COUNT(mark, ec->nodes) == 0)
+ return 0;
- versions_xattr[EC_DATA_TXN] = hton64(1ULL<<EC_SELFHEAL_BIT);
- if (dict_set_static_bin (xattrs, EC_XATTR_VERSION, versions_xattr,
- sizeof (versions_xattr))) {
- ret = -ENOMEM;
- goto out;
- }
+ versions_xattr[EC_DATA_TXN] = hton64(1ULL << EC_SELFHEAL_BIT);
+ if (dict_set_static_bin(xattrs, EC_XATTR_VERSION, versions_xattr,
+ sizeof(versions_xattr))) {
+ ret = -ENOMEM;
+ goto out;
+ }
- output = alloca0 (ec->nodes);
- ret = cluster_fxattrop (ec->xl_list, mark, ec->nodes,
- replies, output, frame, ec->xl, fd,
- GF_XATTROP_ADD_ARRAY64, xattrs, NULL);
- for (i = 0; i < ec->nodes; i++) {
- if (!output[i]) {
- if (mark[i])
- healed_sinks[i] = 0;
- continue;
- }
- versions[i] |= (1ULL<<EC_SELFHEAL_BIT);
+ output = alloca0(ec->nodes);
+ ret = cluster_fxattrop(ec->xl_list, mark, ec->nodes, replies, output, frame,
+ ec->xl, fd, GF_XATTROP_ADD_ARRAY64, xattrs, NULL);
+ for (i = 0; i < ec->nodes; i++) {
+ if (!output[i]) {
+ if (mark[i])
+ healed_sinks[i] = 0;
+ continue;
}
+ versions[i] |= (1ULL << EC_SELFHEAL_BIT);
+ }
- if (EC_COUNT (healed_sinks, ec->nodes) == 0) {
- ret = -ENOTCONN;
- goto out;
- }
- ret = 0;
+ if (EC_COUNT(healed_sinks, ec->nodes) == 0) {
+ ret = -ENOTCONN;
+ goto out;
+ }
+ ret = 0;
out:
- cluster_replies_wipe (replies, ec->nodes);
- if (xattrs)
- dict_unref (xattrs);
- if (ret < 0)
- gf_msg_debug (ec->xl->name, 0, "%s: heal failed %s",
- uuid_utoa (fd->inode->gfid), strerror (-ret));
- return ret;
+ cluster_replies_wipe(replies, ec->nodes);
+ if (xattrs)
+ dict_unref(xattrs);
+ if (ret < 0)
+ gf_msg_debug(ec->xl->name, 0, "%s: heal failed %s",
+ uuid_utoa(fd->inode->gfid), strerror(-ret));
+ return ret;
}
int32_t
-ec_manager_heal_block (ec_fop_data_t *fop, int32_t state)
+ec_manager_heal_block(ec_fop_data_t *fop, int32_t state)
{
ec_heal_t *heal = fop->data;
heal->fop = fop;
switch (state) {
- case EC_STATE_INIT:
- ec_owner_set(fop->frame, fop->frame->root);
+ case EC_STATE_INIT:
+ ec_owner_set(fop->frame, fop->frame->root);
- ec_heal_inodelk(heal, F_WRLCK, 1, 0, 0);
+ ec_heal_inodelk(heal, F_WRLCK, 1, 0, 0);
- return EC_STATE_HEAL_DATA_COPY;
+ return EC_STATE_HEAL_DATA_COPY;
- case EC_STATE_HEAL_DATA_COPY:
- gf_msg_debug (fop->xl->name, 0, "%s: read/write starting",
- uuid_utoa (heal->fd->inode->gfid));
- ec_heal_data_block (heal);
+ case EC_STATE_HEAL_DATA_COPY:
+ gf_msg_debug(fop->xl->name, 0, "%s: read/write starting",
+ uuid_utoa(heal->fd->inode->gfid));
+ ec_heal_data_block(heal);
- return EC_STATE_HEAL_DATA_UNLOCK;
+ return EC_STATE_HEAL_DATA_UNLOCK;
- case -EC_STATE_HEAL_DATA_COPY:
- case -EC_STATE_HEAL_DATA_UNLOCK:
- case EC_STATE_HEAL_DATA_UNLOCK:
- ec_heal_inodelk(heal, F_UNLCK, 1, 0, 0);
+ case -EC_STATE_HEAL_DATA_COPY:
+ case -EC_STATE_HEAL_DATA_UNLOCK:
+ case EC_STATE_HEAL_DATA_UNLOCK:
+ ec_heal_inodelk(heal, F_UNLCK, 1, 0, 0);
- return EC_STATE_REPORT;
+ return EC_STATE_REPORT;
- case EC_STATE_REPORT:
- if (fop->cbks.heal) {
- fop->cbks.heal (fop->req_frame, fop, fop->xl, 0,
- 0, (heal->good | heal->bad),
- heal->good, heal->bad, NULL);
- }
+ case EC_STATE_REPORT:
+ if (fop->cbks.heal) {
+ fop->cbks.heal(fop->req_frame, fop->data, fop->xl, 0, 0,
+ (heal->good | heal->bad), heal->good, heal->bad,
+ 0, NULL);
+ }
- return EC_STATE_END;
- case -EC_STATE_REPORT:
- if (fop->cbks.heal) {
- fop->cbks.heal (fop->req_frame, fop, fop->xl, -1,
- fop->error, 0, 0, 0, NULL);
- }
+ return EC_STATE_END;
+ case -EC_STATE_REPORT:
+ if (fop->cbks.heal) {
+ fop->cbks.heal(fop->req_frame, fop->data, fop->xl, -1,
+ fop->error, 0, 0, 0, 0, NULL);
+ }
- return EC_STATE_END;
- default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, 0,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ return EC_STATE_END;
+ default:
+ gf_msg(fop->xl->name, GF_LOG_ERROR, 0, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
- return EC_STATE_END;
+ return EC_STATE_END;
}
}
/*Takes lock */
void
-ec_heal_block (call_frame_t *frame, xlator_t *this, uintptr_t target,
- int32_t minimum, fop_heal_cbk_t func, ec_heal_t *heal)
+ec_heal_block(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_heal_cbk_t func, ec_heal_t *heal)
{
- ec_cbk_t callback = { .heal = func };
+ ec_cbk_t callback = {.heal = func};
ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
@@ -1947,9 +2009,8 @@ ec_heal_block (call_frame_t *frame, xlator_t *this, uintptr_t target,
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate (frame, this, EC_FOP_HEAL, 0, target, minimum,
- NULL, ec_manager_heal_block, callback,
- heal);
+ fop = ec_fop_data_allocate(frame, this, EC_FOP_HEAL, 0, target, fop_flags,
+ NULL, ec_manager_heal_block, callback, heal);
if (fop == NULL)
goto out;
@@ -1959,766 +2020,878 @@ out:
if (fop != NULL) {
ec_manager(fop, error);
} else {
- func(frame, NULL, this, -1, error, 0, 0, 0, NULL);
+ func(frame, heal, this, -1, error, 0, 0, 0, 0, NULL);
}
}
int32_t
-ec_heal_block_done (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, uintptr_t mask,
- uintptr_t good, uintptr_t bad, dict_t *xdata)
+ec_heal_block_done(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, uintptr_t mask,
+ uintptr_t good, uintptr_t bad, uint32_t pending,
+ dict_t *xdata)
{
- ec_fop_data_t *fop = cookie;
- ec_heal_t *heal = fop->data;
+ ec_heal_t *heal = cookie;
- fop->heal = NULL;
- heal->fop = NULL;
- heal->error = op_ret < 0 ? op_errno : 0;
- syncbarrier_wake (heal->data);
- return 0;
+ if (heal->fop) {
+ heal->fop->heal = NULL;
+ }
+ heal->fop = NULL;
+ heal->error = op_ret < 0 ? op_errno : 0;
+ syncbarrier_wake(heal->data);
+ return 0;
}
int
-ec_sync_heal_block (call_frame_t *frame, xlator_t *this, ec_heal_t *heal)
+ec_sync_heal_block(call_frame_t *frame, xlator_t *this, ec_heal_t *heal)
{
- ec_heal_block (frame, this, heal->bad|heal->good, EC_MINIMUM_ONE,
- ec_heal_block_done, heal);
- syncbarrier_wait (heal->data, 1);
- if (heal->error != 0) {
- return -heal->error;
- }
- if (heal->bad == 0)
- return -ENOTCONN;
- return 0;
+ ec_heal_block(frame, this, heal->bad | heal->good, EC_MINIMUM_ONE,
+ ec_heal_block_done, heal);
+ syncbarrier_wait(heal->data, 1);
+ if (heal->error != 0) {
+ return -heal->error;
+ }
+ if (heal->bad == 0)
+ return -ENOTCONN;
+ return 0;
}
int
-ec_rebuild_data (call_frame_t *frame, ec_t *ec, fd_t *fd, uint64_t size,
- unsigned char *sources, unsigned char *healed_sinks)
+ec_rebuild_data(call_frame_t *frame, ec_t *ec, fd_t *fd, uint64_t size,
+ unsigned char *sources, unsigned char *healed_sinks)
{
- ec_heal_t *heal = NULL;
- int ret = 0;
- syncbarrier_t barrier;
-
- if (syncbarrier_init (&barrier))
- return -ENOMEM;
-
- heal = alloca0(sizeof (*heal));
- heal->fd = fd_ref (fd);
- heal->xl = ec->xl;
- heal->data = &barrier;
- ec_adjust_size_up (ec, &size, _gf_false);
- heal->total_size = size;
- heal->size = (128 * GF_UNIT_KB * (ec->self_heal_window_size));
- /* We need to adjust the size to a multiple of the stripe size of the
- * volume. Otherwise writes would need to fill gaps (head and/or tail)
- * with existent data from the bad bricks. This could be garbage on a
- * damaged file or it could fail if there aren't enough bricks. */
- heal->size -= heal->size % ec->stripe_size;
- heal->bad = ec_char_array_to_mask (healed_sinks, ec->nodes);
- heal->good = ec_char_array_to_mask (sources, ec->nodes);
- heal->iatt.ia_type = IA_IFREG;
- LOCK_INIT(&heal->lock);
-
- for (heal->offset = 0; (heal->offset < size) && !heal->done;
- heal->offset += heal->size) {
- /* We immediately abort any heal if a shutdown request has been
- * received to avoid delays. The healing of this file will be
- * restarted by another SHD or other client that accesses the
- * file. */
- if (ec->shutdown) {
- gf_msg_debug(ec->xl->name, 0, "Cancelling heal because "
- "EC is stopping.");
- ret = -ENOTCONN;
- break;
- }
-
- gf_msg_debug (ec->xl->name, 0, "%s: sources: %d, sinks: "
- "%d, offset: %"PRIu64" bsize: %"PRIu64,
- uuid_utoa (fd->inode->gfid),
- EC_COUNT (sources, ec->nodes),
- EC_COUNT (healed_sinks, ec->nodes), heal->offset,
- heal->size);
- ret = ec_sync_heal_block (frame, ec->xl, heal);
- if (ret < 0)
- break;
-
- }
- memset (healed_sinks, 0, ec->nodes);
- ec_mask_to_char_array (heal->bad, healed_sinks, ec->nodes);
- fd_unref (heal->fd);
- LOCK_DESTROY (&heal->lock);
- syncbarrier_destroy (heal->data);
+ ec_heal_t *heal = NULL;
+ int ret = 0;
+ syncbarrier_t barrier;
+
+ if (syncbarrier_init(&barrier))
+ return -ENOMEM;
+
+ heal = alloca0(sizeof(*heal));
+ heal->fd = fd_ref(fd);
+ heal->xl = ec->xl;
+ heal->data = &barrier;
+ ec_adjust_size_up(ec, &size, _gf_false);
+ heal->total_size = size;
+ heal->size = (128 * GF_UNIT_KB * (ec->self_heal_window_size));
+ /* We need to adjust the size to a multiple of the stripe size of the
+ * volume. Otherwise writes would need to fill gaps (head and/or tail)
+ * with existent data from the bad bricks. This could be garbage on a
+ * damaged file or it could fail if there aren't enough bricks. */
+ heal->size -= heal->size % ec->stripe_size;
+ heal->bad = ec_char_array_to_mask(healed_sinks, ec->nodes);
+ heal->good = ec_char_array_to_mask(sources, ec->nodes);
+ heal->iatt.ia_type = IA_IFREG;
+ LOCK_INIT(&heal->lock);
+
+ for (heal->offset = 0; (heal->offset < size) && !heal->done;
+ heal->offset += heal->size) {
+ /* We immediately abort any heal if a shutdown request has been
+ * received to avoid delays. The healing of this file will be
+ * restarted by another SHD or other client that accesses the
+ * file. */
+ if (ec->shutdown) {
+ gf_msg_debug(ec->xl->name, 0,
+ "Cancelling heal because "
+ "EC is stopping.");
+ ret = -ENOTCONN;
+ break;
+ }
+
+ gf_msg_debug(ec->xl->name, 0,
+ "%s: sources: %d, sinks: "
+ "%d, offset: %" PRIu64 " bsize: %" PRIu64,
+ uuid_utoa(fd->inode->gfid), EC_COUNT(sources, ec->nodes),
+ EC_COUNT(healed_sinks, ec->nodes), heal->offset,
+ heal->size);
+ ret = ec_sync_heal_block(frame, ec->xl, heal);
if (ret < 0)
- gf_msg_debug (ec->xl->name, 0, "%s: heal failed %s",
- uuid_utoa (fd->inode->gfid), strerror (-ret));
- return ret;
+ break;
+ }
+ memset(healed_sinks, 0, ec->nodes);
+ ec_mask_to_char_array(heal->bad, healed_sinks, ec->nodes);
+ fd_unref(heal->fd);
+ LOCK_DESTROY(&heal->lock);
+ syncbarrier_destroy(heal->data);
+ if (ret < 0)
+ gf_msg_debug(ec->xl->name, 0, "%s: heal failed %s",
+ uuid_utoa(fd->inode->gfid), strerror(-ret));
+ return ret;
}
int
-__ec_heal_trim_sinks (call_frame_t *frame, ec_t *ec,
- fd_t *fd, unsigned char *healed_sinks,
- unsigned char *trim, uint64_t size)
+__ec_heal_trim_sinks(call_frame_t *frame, ec_t *ec, fd_t *fd,
+ unsigned char *healed_sinks, unsigned char *trim,
+ uint64_t size)
{
- default_args_cbk_t *replies = NULL;
- unsigned char *output = NULL;
- int ret = 0;
- int i = 0;
- off_t trim_offset = 0;
+ default_args_cbk_t *replies = NULL;
+ unsigned char *output = NULL;
+ int ret = 0;
+ int i = 0;
+ off_t trim_offset = 0;
- EC_REPLIES_ALLOC (replies, ec->nodes);
- output = alloca0 (ec->nodes);
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ output = alloca0(ec->nodes);
- if (EC_COUNT (trim, ec->nodes) == 0) {
- ret = 0;
- goto out;
- }
- trim_offset = size;
- ec_adjust_offset_up (ec, &trim_offset, _gf_true);
- ret = cluster_ftruncate (ec->xl_list, trim, ec->nodes, replies, output,
- frame, ec->xl, fd, trim_offset, NULL);
- for (i = 0; i < ec->nodes; i++) {
- if (!output[i] && trim[i])
- healed_sinks[i] = 0;
- }
+ if (EC_COUNT(trim, ec->nodes) == 0) {
+ ret = 0;
+ goto out;
+ }
+ trim_offset = size;
+ ec_adjust_offset_up(ec, &trim_offset, _gf_true);
+ ret = cluster_ftruncate(ec->xl_list, trim, ec->nodes, replies, output,
+ frame, ec->xl, fd, trim_offset, NULL);
+ for (i = 0; i < ec->nodes; i++) {
+ if (!output[i] && trim[i])
+ healed_sinks[i] = 0;
+ }
- if (EC_COUNT (healed_sinks, ec->nodes) == 0) {
- ret = -ENOTCONN;
- goto out;
- }
+ if (EC_COUNT(healed_sinks, ec->nodes) == 0) {
+ ret = -ENOTCONN;
+ goto out;
+ }
out:
- cluster_replies_wipe (replies, ec->nodes);
- if (ret < 0)
- gf_msg_debug (ec->xl->name, 0, "%s: heal failed %s",
- uuid_utoa (fd->inode->gfid), strerror (-ret));
- return ret;
+ cluster_replies_wipe(replies, ec->nodes);
+ if (ret < 0)
+ gf_msg_debug(ec->xl->name, 0, "%s: heal failed %s",
+ uuid_utoa(fd->inode->gfid), strerror(-ret));
+ return ret;
}
int
-ec_data_undo_pending (call_frame_t *frame, ec_t *ec, fd_t *fd, dict_t *xattr,
- uint64_t *versions, uint64_t *dirty, uint64_t *size,
- int source, gf_boolean_t erase_dirty, int idx)
-{
- uint64_t versions_xattr[2] = {0};
- uint64_t dirty_xattr[2] = {0};
- uint64_t allzero[2] = {0};
- uint64_t size_xattr = 0;
- int ret = 0;
-
- versions_xattr[EC_DATA_TXN] = hton64(versions[source] - versions[idx]);
- ret = dict_set_static_bin (xattr, EC_XATTR_VERSION,
- versions_xattr,
- sizeof (versions_xattr));
- if (ret < 0)
- goto out;
+ec_data_undo_pending(call_frame_t *frame, ec_t *ec, fd_t *fd, dict_t *xattr,
+ uint64_t *versions, uint64_t *dirty, uint64_t *size,
+ int source, gf_boolean_t erase_dirty, int idx)
+{
+ uint64_t versions_xattr[2] = {0};
+ uint64_t dirty_xattr[2] = {0};
+ uint64_t allzero[2] = {0};
+ uint64_t size_xattr = 0;
+ int ret = 0;
+
+ versions_xattr[EC_DATA_TXN] = hton64(versions[source] - versions[idx]);
+ ret = dict_set_static_bin(xattr, EC_XATTR_VERSION, versions_xattr,
+ sizeof(versions_xattr));
+ if (ret < 0)
+ goto out;
- size_xattr = hton64(size[source] - size[idx]);
- ret = dict_set_static_bin (xattr, EC_XATTR_SIZE,
- &size_xattr, sizeof (size_xattr));
+ size_xattr = hton64(size[source] - size[idx]);
+ ret = dict_set_static_bin(xattr, EC_XATTR_SIZE, &size_xattr,
+ sizeof(size_xattr));
+ if (ret < 0)
+ goto out;
+
+ if (erase_dirty) {
+ dirty_xattr[EC_DATA_TXN] = hton64(-dirty[idx]);
+ ret = dict_set_static_bin(xattr, EC_XATTR_DIRTY, dirty_xattr,
+ sizeof(dirty_xattr));
if (ret < 0)
- goto out;
+ goto out;
+ }
- if (erase_dirty) {
- dirty_xattr[EC_DATA_TXN] = hton64(-dirty[idx]);
- ret = dict_set_static_bin (xattr, EC_XATTR_DIRTY,
- dirty_xattr,
- sizeof (dirty_xattr));
- if (ret < 0)
- goto out;
- }
-
- if ((memcmp (versions_xattr, allzero, sizeof (allzero)) == 0) &&
- (memcmp (dirty_xattr, allzero, sizeof (allzero)) == 0) &&
- (size_xattr == 0)) {
- ret = 0;
- goto out;
- }
+ if ((memcmp(versions_xattr, allzero, sizeof(allzero)) == 0) &&
+ (memcmp(dirty_xattr, allzero, sizeof(allzero)) == 0) &&
+ (size_xattr == 0)) {
+ ret = 0;
+ goto out;
+ }
- ret = syncop_fxattrop (ec->xl_list[idx], fd,
- GF_XATTROP_ADD_ARRAY64, xattr, NULL, NULL, NULL);
+ ret = syncop_fxattrop(ec->xl_list[idx], fd, GF_XATTROP_ADD_ARRAY64, xattr,
+ NULL, NULL, NULL);
out:
- return ret;
+ return ret;
}
int
-__ec_fd_data_adjust_versions (call_frame_t *frame, ec_t *ec, fd_t *fd,
- unsigned char *sources, unsigned char *healed_sinks,
- uint64_t *versions, uint64_t *dirty, uint64_t *size)
-{
- dict_t *xattr = NULL;
- int i = 0;
- int ret = 0;
- int op_ret = 0;
- int source = -1;
- gf_boolean_t erase_dirty = _gf_false;
-
- xattr = dict_new ();
- if (!xattr) {
- op_ret = -ENOMEM;
- goto out;
- }
+__ec_fd_data_adjust_versions(call_frame_t *frame, ec_t *ec, fd_t *fd,
+ unsigned char *sources,
+ unsigned char *healed_sinks, uint64_t *versions,
+ uint64_t *dirty, uint64_t *size)
+{
+ dict_t *xattr = NULL;
+ int i = 0;
+ int ret = 0;
+ int op_ret = 0;
+ int source = -1;
+ gf_boolean_t erase_dirty = _gf_false;
+
+ xattr = dict_new();
+ if (!xattr) {
+ op_ret = -ENOMEM;
+ goto out;
+ }
- /* dirty xattr represents if the file needs heal. Unless all the
- * copies are healed, don't erase it */
- if (EC_COUNT (sources, ec->nodes) +
- EC_COUNT (healed_sinks, ec->nodes) == ec->nodes)
- erase_dirty = _gf_true;
+ /* dirty xattr represents if the file needs heal. Unless all the
+ * copies are healed, don't erase it */
+ if (EC_COUNT(sources, ec->nodes) + EC_COUNT(healed_sinks, ec->nodes) ==
+ ec->nodes)
+ erase_dirty = _gf_true;
- for (i = 0; i < ec->nodes; i++) {
- if (sources[i]) {
- source = i;
- break;
- }
- }
-
- if (source == -1) {
- op_ret = -ENOTCONN;
- goto out;
+ for (i = 0; i < ec->nodes; i++) {
+ if (sources[i]) {
+ source = i;
+ break;
}
+ }
- for (i = 0; i < ec->nodes; i++) {
- if (healed_sinks[i]) {
- ret = ec_data_undo_pending (frame, ec, fd, xattr,
- versions, dirty, size,
- source, erase_dirty, i);
- if (ret < 0)
- goto out;
- }
-
- }
+ if (source == -1) {
+ op_ret = -ENOTCONN;
+ goto out;
+ }
- if (!erase_dirty)
+ for (i = 0; i < ec->nodes; i++) {
+ if (healed_sinks[i]) {
+ ret = ec_data_undo_pending(frame, ec, fd, xattr, versions, dirty,
+ size, source, erase_dirty, i);
+ if (ret < 0)
goto out;
+ }
+ }
- for (i = 0; i < ec->nodes; i++) {
- if (sources[i]) {
- ret = ec_data_undo_pending (frame, ec, fd, xattr,
- versions, dirty, size,
- source, erase_dirty, i);
- if (ret < 0)
- continue;
- }
+ if (!erase_dirty)
+ goto out;
+ for (i = 0; i < ec->nodes; i++) {
+ if (sources[i]) {
+ ret = ec_data_undo_pending(frame, ec, fd, xattr, versions, dirty,
+ size, source, erase_dirty, i);
+ if (ret < 0)
+ continue;
}
+ }
out:
- if (xattr)
- dict_unref (xattr);
- return op_ret;
+ if (xattr)
+ dict_unref(xattr);
+ return op_ret;
}
int
-ec_restore_time_and_adjust_versions (call_frame_t *frame, ec_t *ec, fd_t *fd,
- unsigned char *sources,
- unsigned char *healed_sinks,
- uint64_t *versions, uint64_t *dirty,
- uint64_t *size)
-{
- unsigned char *locked_on = NULL;
- unsigned char *participants = NULL;
- unsigned char *output = NULL;
- default_args_cbk_t *replies = NULL;
- unsigned char *postsh_sources = NULL;
- unsigned char *postsh_healed_sinks = NULL;
- unsigned char *postsh_trim = NULL;
- uint64_t *postsh_versions = NULL;
- uint64_t *postsh_dirty = NULL;
- uint64_t *postsh_size = NULL;
- int ret = 0;
- int i = 0;
- struct iatt source_buf = {0};
- loc_t loc = {0};
-
- locked_on = alloca0(ec->nodes);
- output = alloca0(ec->nodes);
- participants = alloca0(ec->nodes);
- postsh_sources = alloca0(ec->nodes);
- postsh_healed_sinks = alloca0(ec->nodes);
- postsh_trim = alloca0(ec->nodes);
- postsh_versions = alloca0(ec->nodes * sizeof (*postsh_versions));
- postsh_dirty = alloca0(ec->nodes * sizeof (*postsh_dirty));
- postsh_size = alloca0(ec->nodes * sizeof (*postsh_size));
-
- for (i = 0; i < ec->nodes; i++) {
- if (healed_sinks[i] || sources[i])
- participants[i] = 1;
- }
-
- EC_REPLIES_ALLOC (replies, ec->nodes);
- ret = cluster_inodelk (ec->xl_list, participants, ec->nodes, replies,
- locked_on, frame, ec->xl, ec->xl->name,
- fd->inode, 0, 0);
- {
- if (ret <= ec->fragments) {
- gf_msg_debug (ec->xl->name, 0, "%s: Skipping heal "
- "as only %d number of subvolumes could "
- "be locked", uuid_utoa (fd->inode->gfid), ret);
- ret = -ENOTCONN;
- goto unlock;
- }
+ec_restore_time_and_adjust_versions(call_frame_t *frame, ec_t *ec, fd_t *fd,
+ unsigned char *sources,
+ unsigned char *healed_sinks,
+ uint64_t *versions, uint64_t *dirty,
+ uint64_t *size)
+{
+ unsigned char *locked_on = NULL;
+ unsigned char *participants = NULL;
+ unsigned char *output = NULL;
+ default_args_cbk_t *replies = NULL;
+ unsigned char *postsh_sources = NULL;
+ unsigned char *postsh_healed_sinks = NULL;
+ unsigned char *postsh_trim = NULL;
+ uint64_t *postsh_versions = NULL;
+ uint64_t *postsh_dirty = NULL;
+ uint64_t *postsh_size = NULL;
+ int ret = 0;
+ int i = 0;
+ struct iatt source_buf = {0};
+ loc_t loc = {0};
+
+ locked_on = alloca0(ec->nodes);
+ output = alloca0(ec->nodes);
+ participants = alloca0(ec->nodes);
+ postsh_sources = alloca0(ec->nodes);
+ postsh_healed_sinks = alloca0(ec->nodes);
+ postsh_trim = alloca0(ec->nodes);
+ postsh_versions = alloca0(ec->nodes * sizeof(*postsh_versions));
+ postsh_dirty = alloca0(ec->nodes * sizeof(*postsh_dirty));
+ postsh_size = alloca0(ec->nodes * sizeof(*postsh_size));
+
+ for (i = 0; i < ec->nodes; i++) {
+ if (healed_sinks[i] || sources[i])
+ participants[i] = 1;
+ }
- ret = __ec_heal_data_prepare (frame, ec, fd, locked_on,
- postsh_versions, postsh_dirty,
- postsh_size, postsh_sources,
- postsh_healed_sinks, postsh_trim,
- &source_buf);
- if (ret < 0)
- goto unlock;
-
- loc.inode = inode_ref (fd->inode);
- gf_uuid_copy (loc.gfid, fd->inode->gfid);
- ret = cluster_setattr (ec->xl_list, healed_sinks, ec->nodes,
- replies, output, frame, ec->xl, &loc,
- &source_buf,
- GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME,
- NULL);
- EC_INTERSECT (healed_sinks, healed_sinks, output, ec->nodes);
- if (EC_COUNT (healed_sinks, ec->nodes) == 0) {
- ret = -ENOTCONN;
- goto unlock;
- }
- ret = __ec_fd_data_adjust_versions (frame, ec, fd, sources,
- healed_sinks, versions, dirty, size);
- }
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ ret = cluster_inodelk(ec->xl_list, participants, ec->nodes, replies,
+ locked_on, frame, ec->xl, ec->xl->name, fd->inode, 0,
+ 0);
+ {
+ if (ret <= ec->fragments) {
+ gf_msg_debug(ec->xl->name, 0,
+ "%s: Skipping heal "
+ "as only %d number of subvolumes could "
+ "be locked",
+ uuid_utoa(fd->inode->gfid), ret);
+ ret = -ENOTCONN;
+ goto unlock;
+ }
+
+ ret = __ec_heal_data_prepare(frame, ec, fd, locked_on, postsh_versions,
+ postsh_dirty, postsh_size, postsh_sources,
+ postsh_healed_sinks, postsh_trim,
+ &source_buf);
+ if (ret < 0)
+ goto unlock;
+
+ loc.inode = inode_ref(fd->inode);
+ gf_uuid_copy(loc.gfid, fd->inode->gfid);
+ ret = cluster_setattr(
+ ec->xl_list, healed_sinks, ec->nodes, replies, output, frame,
+ ec->xl, &loc, &source_buf,
+ GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME | GF_SET_ATTR_CTIME, NULL);
+ EC_INTERSECT(healed_sinks, healed_sinks, output, ec->nodes);
+ if (EC_COUNT(healed_sinks, ec->nodes) == 0) {
+ ret = -ENOTCONN;
+ goto unlock;
+ }
+ ret = __ec_fd_data_adjust_versions(frame, ec, fd, sources, healed_sinks,
+ versions, dirty, size);
+ }
unlock:
- cluster_uninodelk (ec->xl_list, locked_on, ec->nodes, replies, output,
- frame, ec->xl, ec->xl->name, fd->inode, 0, 0);
- cluster_replies_wipe (replies, ec->nodes);
- loc_wipe (&loc);
- return ret;
+ cluster_uninodelk(ec->xl_list, locked_on, ec->nodes, replies, output, frame,
+ ec->xl, ec->xl->name, fd->inode, 0, 0);
+ cluster_replies_wipe(replies, ec->nodes);
+ loc_wipe(&loc);
+ return ret;
}
int
-__ec_heal_data (call_frame_t *frame, ec_t *ec, fd_t *fd, unsigned char *heal_on,
- unsigned char *sources, unsigned char *healed_sinks)
+__ec_heal_data(call_frame_t *frame, ec_t *ec, fd_t *fd, unsigned char *heal_on,
+ unsigned char *sources, unsigned char *healed_sinks)
{
- unsigned char *locked_on = NULL;
- unsigned char *output = NULL;
- uint64_t *versions = NULL;
- uint64_t *dirty = NULL;
- uint64_t *size = NULL;
- unsigned char *trim = NULL;
- default_args_cbk_t *replies = NULL;
- int ret = 0;
- int source = 0;
-
- locked_on = alloca0(ec->nodes);
- output = alloca0(ec->nodes);
- trim = alloca0 (ec->nodes);
- versions = alloca0 (ec->nodes * sizeof (*versions));
- dirty = alloca0 (ec->nodes * sizeof (*dirty));
- size = alloca0 (ec->nodes * sizeof (*size));
-
- EC_REPLIES_ALLOC (replies, ec->nodes);
- ret = cluster_inodelk (ec->xl_list, heal_on, ec->nodes, replies,
- locked_on, frame, ec->xl, ec->xl->name,
- fd->inode, 0, 0);
- {
- if (ret <= ec->fragments) {
- gf_msg_debug (ec->xl->name, 0, "%s: Skipping heal "
- "as only %d number of subvolumes could "
- "be locked", uuid_utoa (fd->inode->gfid), ret);
- ret = -ENOTCONN;
- goto unlock;
- }
-
- ret = __ec_heal_data_prepare (frame, ec, fd, locked_on,
- versions, dirty, size, sources,
- healed_sinks, trim, NULL);
- if (ret < 0)
- goto unlock;
-
- if (EC_COUNT(healed_sinks, ec->nodes) == 0) {
- ret = __ec_fd_data_adjust_versions (frame, ec, fd,
- sources,
- healed_sinks, versions, dirty, size);
- goto unlock;
- }
+ unsigned char *locked_on = NULL;
+ unsigned char *output = NULL;
+ uint64_t *versions = NULL;
+ uint64_t *dirty = NULL;
+ uint64_t *size = NULL;
+ unsigned char *trim = NULL;
+ default_args_cbk_t *replies = NULL;
+ int ret = 0;
+ int source = 0;
+
+ locked_on = alloca0(ec->nodes);
+ output = alloca0(ec->nodes);
+ trim = alloca0(ec->nodes);
+ versions = alloca0(ec->nodes * sizeof(*versions));
+ dirty = alloca0(ec->nodes * sizeof(*dirty));
+ size = alloca0(ec->nodes * sizeof(*size));
+
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ ret = cluster_inodelk(ec->xl_list, heal_on, ec->nodes, replies, locked_on,
+ frame, ec->xl, ec->xl->name, fd->inode, 0, 0);
+ {
+ if (ret <= ec->fragments) {
+ gf_msg_debug(ec->xl->name, 0,
+ "%s: Skipping heal "
+ "as only %d number of subvolumes could "
+ "be locked",
+ uuid_utoa(fd->inode->gfid), ret);
+ ret = -ENOTCONN;
+ goto unlock;
+ }
- source = ret;
- ret = __ec_heal_mark_sinks (frame, ec, fd, versions,
- healed_sinks);
- if (ret < 0)
- goto unlock;
+ ret = __ec_heal_data_prepare(frame, ec, fd, locked_on, versions, dirty,
+ size, sources, healed_sinks, trim, NULL);
+ if (ret < 0)
+ goto unlock;
- ret = __ec_heal_trim_sinks (frame, ec, fd, healed_sinks,
- trim, size[source]);
+ if (EC_COUNT(healed_sinks, ec->nodes) == 0) {
+ ret = __ec_fd_data_adjust_versions(
+ frame, ec, fd, sources, healed_sinks, versions, dirty, size);
+ goto unlock;
}
-unlock:
- cluster_uninodelk (ec->xl_list, locked_on, ec->nodes, replies, output,
- frame, ec->xl, ec->xl->name, fd->inode, 0, 0);
+
+ source = ret;
+ ret = __ec_heal_mark_sinks(frame, ec, fd, versions, healed_sinks);
if (ret < 0)
- goto out;
+ goto unlock;
- if (EC_COUNT(healed_sinks, ec->nodes) == 0)
- goto out;
+ ret = __ec_heal_trim_sinks(frame, ec, fd, healed_sinks, trim,
+ size[source]);
+ }
+unlock:
+ cluster_uninodelk(ec->xl_list, locked_on, ec->nodes, replies, output, frame,
+ ec->xl, ec->xl->name, fd->inode, 0, 0);
+ if (ret < 0)
+ goto out;
- gf_msg_debug (ec->xl->name, 0, "%s: sources: %d, sinks: "
- "%d", uuid_utoa (fd->inode->gfid),
- EC_COUNT (sources, ec->nodes),
- EC_COUNT (healed_sinks, ec->nodes));
+ if (EC_COUNT(healed_sinks, ec->nodes) == 0)
+ goto out;
- ret = ec_rebuild_data (frame, ec, fd, size[source], sources,
- healed_sinks);
- if (ret < 0)
- goto out;
+ gf_msg_debug(ec->xl->name, 0,
+ "%s: sources: %d, sinks: "
+ "%d",
+ uuid_utoa(fd->inode->gfid), EC_COUNT(sources, ec->nodes),
+ EC_COUNT(healed_sinks, ec->nodes));
+
+ ret = ec_rebuild_data(frame, ec, fd, size[source], sources, healed_sinks);
+ if (ret < 0)
+ goto out;
- ret = ec_restore_time_and_adjust_versions (frame, ec, fd, sources,
- healed_sinks, versions,
- dirty, size);
+ ret = ec_restore_time_and_adjust_versions(
+ frame, ec, fd, sources, healed_sinks, versions, dirty, size);
out:
- cluster_replies_wipe (replies, ec->nodes);
- return ret;
+ cluster_replies_wipe(replies, ec->nodes);
+ return ret;
}
int
-ec_heal_data (call_frame_t *frame, ec_t *ec, gf_boolean_t block, inode_t *inode,
- unsigned char *sources, unsigned char *healed_sinks)
-{
- unsigned char *locked_on = NULL;
- unsigned char *up_subvols = NULL;
- unsigned char *output = NULL;
- default_args_cbk_t *replies = NULL;
- fd_t *fd = NULL;
- loc_t loc = {0};
- char selfheal_domain[1024] = {0};
- int ret = 0;
-
- EC_REPLIES_ALLOC (replies, ec->nodes);
-
- locked_on = alloca0(ec->nodes);
- output = alloca0(ec->nodes);
- up_subvols = alloca0(ec->nodes);
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
-
- fd = fd_create (inode, 0);
- if (!fd) {
- ret = -ENOMEM;
- goto out;
- }
+ec_heal_data(call_frame_t *frame, ec_t *ec, gf_boolean_t block, inode_t *inode,
+ unsigned char *sources, unsigned char *healed_sinks)
+{
+ unsigned char *locked_on = NULL;
+ unsigned char *up_subvols = NULL;
+ unsigned char *output = NULL;
+ default_args_cbk_t *replies = NULL;
+ fd_t *fd = NULL;
+ loc_t loc = {0};
+ char selfheal_domain[1024] = {0};
+ int ret = 0;
+
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+
+ locked_on = alloca0(ec->nodes);
+ output = alloca0(ec->nodes);
+ up_subvols = alloca0(ec->nodes);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+
+ fd = fd_create(inode, 0);
+ if (!fd) {
+ ret = -ENOMEM;
+ goto out;
+ }
- ec_mask_to_char_array (ec->xl_up, up_subvols, ec->nodes);
+ ec_mask_to_char_array(ec->xl_up, up_subvols, ec->nodes);
- ret = cluster_open (ec->xl_list, up_subvols, ec->nodes, replies, output,
- frame, ec->xl, &loc, O_RDWR|O_LARGEFILE, fd, NULL);
- if (ret <= ec->fragments) {
- ret = -ENOTCONN;
- goto out;
- }
+ ret = cluster_open(ec->xl_list, up_subvols, ec->nodes, replies, output,
+ frame, ec->xl, &loc, O_RDWR | O_LARGEFILE, fd, NULL);
+ if (ret <= ec->fragments) {
+ ret = -ENOTCONN;
+ goto out;
+ }
- fd_bind (fd);
- sprintf (selfheal_domain, "%s:self-heal", ec->xl->name);
- /*If other processes are already doing the heal, don't block*/
- if (block) {
- ret = cluster_inodelk (ec->xl_list, output, ec->nodes, replies,
- locked_on, frame, ec->xl,
- selfheal_domain, inode, 0, 0);
- } else {
- ret = cluster_tiebreaker_inodelk (ec->xl_list, output,
- ec->nodes, replies, locked_on,
- frame, ec->xl,
- selfheal_domain, inode, 0, 0);
- }
- {
- if (ret <= ec->fragments) {
- gf_msg_debug (ec->xl->name, 0, "%s: Skipping heal "
- "as only %d number of subvolumes could "
- "be locked", uuid_utoa (inode->gfid), ret);
- ret = -ENOTCONN;
- goto unlock;
- }
- ret = __ec_heal_data (frame, ec, fd, locked_on, sources,
- healed_sinks);
- }
+ fd_bind(fd);
+ sprintf(selfheal_domain, "%s:self-heal", ec->xl->name);
+ /*If other processes are already doing the heal, don't block*/
+ if (block) {
+ ret = cluster_inodelk(ec->xl_list, output, ec->nodes, replies,
+ locked_on, frame, ec->xl, selfheal_domain, inode,
+ 0, 0);
+ } else {
+ ret = cluster_tiebreaker_inodelk(ec->xl_list, output, ec->nodes,
+ replies, locked_on, frame, ec->xl,
+ selfheal_domain, inode, 0, 0);
+ }
+ {
+ if (ret <= ec->fragments) {
+ gf_msg_debug(ec->xl->name, 0,
+ "%s: Skipping heal "
+ "as only %d number of subvolumes could "
+ "be locked",
+ uuid_utoa(inode->gfid), ret);
+ ret = -ENOTCONN;
+ goto unlock;
+ }
+ ret = __ec_heal_data(frame, ec, fd, locked_on, sources, healed_sinks);
+ }
unlock:
- cluster_uninodelk (ec->xl_list, locked_on, ec->nodes, replies, output,
- frame, ec->xl, selfheal_domain, inode, 0, 0);
+ cluster_uninodelk(ec->xl_list, locked_on, ec->nodes, replies, output, frame,
+ ec->xl, selfheal_domain, inode, 0, 0);
out:
- if (fd)
- fd_unref (fd);
- loc_wipe (&loc);
- cluster_replies_wipe (replies, ec->nodes);
- return ret;
+ if (fd)
+ fd_unref(fd);
+ loc_wipe(&loc);
+ cluster_replies_wipe(replies, ec->nodes);
+ return ret;
+}
+
+int
+ec_heal_purge_stale_index(call_frame_t *frame, ec_t *ec, inode_t *inode)
+{
+ int i = 0;
+ int ret = 0;
+ dict_t **xattr = NULL;
+ loc_t loc = {0};
+ uint64_t dirty_xattr[EC_VERSION_SIZE] = {0};
+ unsigned char *on = NULL;
+ default_args_cbk_t *replies = NULL;
+ dict_t *dict = NULL;
+
+ /* Allocate the required memory */
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+ on = alloca0(ec->nodes);
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ xattr = GF_CALLOC(ec->nodes, sizeof(*xattr), gf_common_mt_pointer);
+ if (!xattr) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ dict = dict_new();
+ if (!dict) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ for (i = 0; i < ec->nodes; i++) {
+ xattr[i] = dict;
+ on[i] = 1;
+ }
+ ret = dict_set_static_bin(dict, EC_XATTR_DIRTY, dirty_xattr,
+ (sizeof(*dirty_xattr) * EC_VERSION_SIZE));
+ if (ret < 0) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ PARALLEL_FOP_ONLIST(ec->xl_list, on, ec->nodes, replies, frame,
+ ec_wind_xattrop_parallel, &loc, GF_XATTROP_ADD_ARRAY64,
+ xattr, NULL);
+out:
+ if (dict) {
+ dict_unref(dict);
+ }
+ if (xattr) {
+ GF_FREE(xattr);
+ }
+ cluster_replies_wipe(replies, ec->nodes);
+ loc_wipe(&loc);
+ return ret;
}
void
-ec_heal_do (xlator_t *this, void *data, loc_t *loc, int32_t partial)
-{
- call_frame_t *frame = NULL;
- unsigned char *participants = NULL;
- unsigned char *msources = NULL;
- unsigned char *mhealed_sinks = NULL;
- unsigned char *sources = NULL;
- unsigned char *healed_sinks = NULL;
- ec_t *ec = NULL;
- int ret = 0;
- int op_ret = 0;
- int op_errno = 0;
- intptr_t mgood = 0;
- intptr_t mbad = 0;
- intptr_t good = 0;
- intptr_t bad = 0;
- ec_fop_data_t *fop = data;
- gf_boolean_t blocking = _gf_false;
- ec_heal_need_t need_heal = EC_HEAL_NONEED;
- unsigned char *up_subvols = NULL;
- char up_bricks[32];
-
- ec = this->private;
-
- /* If it is heal request from getxattr, complete the heal and then
- * unwind, if it is ec_heal with NULL as frame then no need to block
- * the heal as the caller doesn't care about its completion. In case
- * of heald whichever gets tiebreaking inodelk will take care of the
- * heal, so no need to block*/
- if (fop->req_frame && !ec->shd.iamshd)
- blocking = _gf_true;
-
- frame = create_frame (this, this->ctx->pool);
- if (!frame)
- goto out;
+ec_heal_do(xlator_t *this, void *data, loc_t *loc, int32_t partial)
+{
+ call_frame_t *frame = NULL;
+ unsigned char *participants = NULL;
+ unsigned char *msources = NULL;
+ unsigned char *mhealed_sinks = NULL;
+ unsigned char *sources = NULL;
+ unsigned char *healed_sinks = NULL;
+ ec_t *ec = NULL;
+ int ret = 0;
+ int op_ret = 0;
+ int op_errno = 0;
+ intptr_t mgood = 0;
+ intptr_t mbad = 0;
+ intptr_t good = 0;
+ intptr_t bad = 0;
+ uint32_t pending = 0;
+ ec_fop_data_t *fop = data;
+ gf_boolean_t blocking = _gf_false;
+ ec_heal_need_t need_heal = EC_HEAL_NONEED;
+ unsigned char *up_subvols = NULL;
+ char up_bricks[32];
+
+ ec = this->private;
+
+ /* If it is heal request from getxattr, complete the heal and then
+ * unwind, if it is ec_heal with NULL as frame then no need to block
+ * the heal as the caller doesn't care about its completion. In case
+ * of heald whichever gets tiebreaking inodelk will take care of the
+ * heal, so no need to block*/
+ if (fop->req_frame && !ec->shd.iamshd)
+ blocking = _gf_true;
+
+ frame = create_frame(this, this->ctx->pool);
+ if (!frame)
+ goto out;
- ec_owner_set(frame, frame->root);
- /*Do heal as root*/
- frame->root->uid = 0;
- frame->root->gid = 0;
- /*Mark the fops as internal*/
- frame->root->pid = GF_CLIENT_PID_SELF_HEALD;
- participants = alloca0(ec->nodes);
- ec_mask_to_char_array (ec->xl_up, participants, ec->nodes);
-
- up_subvols = alloca0(ec->nodes);
- ec_mask_to_char_array (ec->xl_up, up_subvols, ec->nodes);
-
- if (loc->name && strlen (loc->name)) {
- ret = ec_heal_name (frame, ec, loc->parent, (char *)loc->name,
- participants);
- if (ret == 0) {
- gf_msg_debug (this->name, 0, "%s: name heal "
- "successful on %lX", loc->path,
- ec_char_array_to_mask (participants,
- ec->nodes));
- } else {
- gf_msg_debug (this->name, 0, "%s: name heal "
- "failed. ret = %d, subvolumes up = %s",
- loc->path, ret,
- ec_bin(up_bricks, sizeof(up_bricks), ec->xl_up,
- ec->nodes));
- }
+ ec_owner_set(frame, frame->root);
+ /*Do heal as root*/
+ frame->root->uid = 0;
+ frame->root->gid = 0;
+ /*Mark the fops as internal*/
+ frame->root->pid = GF_CLIENT_PID_SELF_HEALD;
+ participants = alloca0(ec->nodes);
+ ec_mask_to_char_array(ec->xl_up, participants, ec->nodes);
+
+ up_subvols = alloca0(ec->nodes);
+ ec_mask_to_char_array(ec->xl_up, up_subvols, ec->nodes);
+
+ if (loc->name && strlen(loc->name)) {
+ ret = ec_heal_name(frame, ec, loc->parent, (char *)loc->name,
+ participants);
+ if (ret >= 0) {
+ gf_msg_debug(this->name, 0,
+ "%s: name heal "
+ "successful on %" PRIXPTR,
+ loc->path,
+ ec_char_array_to_mask(participants, ec->nodes));
+ } else {
+ gf_msg_debug(
+ this->name, 0,
+ "%s: name heal "
+ "failed. ret = %d, subvolumes up = %s",
+ loc->path, ret,
+ ec_bin(up_bricks, sizeof(up_bricks), ec->xl_up, ec->nodes));
}
+ }
- /* Mount triggers heal only when it detects that it must need heal, shd
- * triggers heals periodically which need not be thorough*/
- ec_heal_inspect (frame, ec, loc->inode, up_subvols, _gf_false,
- !ec->shd.iamshd, &need_heal);
-
- if (need_heal == EC_HEAL_NONEED) {
- gf_msg (ec->xl->name, GF_LOG_DEBUG, 0,
- EC_MSG_HEAL_FAIL, "Heal is not required for : %s ",
- uuid_utoa(loc->gfid));
- goto out;
+ /* Mount triggers heal only when it detects that it must need heal, shd
+ * triggers heals periodically which need not be thorough*/
+ if (ec->shd.iamshd && (ret <= 0)) {
+ ec_heal_inspect(frame, ec, loc->inode, up_subvols, _gf_false, _gf_false,
+ &need_heal);
+
+ if (need_heal == EC_HEAL_PURGE_INDEX) {
+ gf_msg(ec->xl->name, GF_LOG_INFO, 0, EC_MSG_HEAL_FAIL,
+ "Index entry needs to be purged for: %s ",
+ uuid_utoa(loc->gfid));
+ /* We need to send zero-xattrop so that stale index entry could be
+ * removed. We need not take lock on this entry to do so as
+ * xattrop on a brick is atomic. */
+ ec_heal_purge_stale_index(frame, ec, loc->inode);
+ goto out;
+ } else if (need_heal == EC_HEAL_NONEED) {
+ gf_msg(ec->xl->name, GF_LOG_DEBUG, 0, EC_MSG_HEAL_FAIL,
+ "Heal is not required for : %s ", uuid_utoa(loc->gfid));
+ goto out;
}
+ }
- msources = alloca0(ec->nodes);
- mhealed_sinks = alloca0(ec->nodes);
- ret = ec_heal_metadata (frame, ec, loc->inode, msources, mhealed_sinks);
- if (ret == 0) {
- mgood = ec_char_array_to_mask (msources, ec->nodes);
- mbad = ec_char_array_to_mask (mhealed_sinks, ec->nodes);
- } else {
- op_ret = -1;
- op_errno = -ret;
- }
- sources = alloca0(ec->nodes);
- healed_sinks = alloca0(ec->nodes);
- if (IA_ISREG (loc->inode->ia_type)) {
- ret = ec_heal_data (frame, ec, blocking, loc->inode, sources,
- healed_sinks);
- } else if (IA_ISDIR (loc->inode->ia_type) && !partial) {
- ret = ec_heal_entry (frame, ec, loc->inode, sources,
- healed_sinks);
- } else {
- ret = 0;
- memcpy (sources, participants, ec->nodes);
- memcpy (healed_sinks, participants, ec->nodes);
- }
+ sources = alloca0(ec->nodes);
+ healed_sinks = alloca0(ec->nodes);
+ if (IA_ISREG(loc->inode->ia_type)) {
+ ret = ec_heal_data(frame, ec, blocking, loc->inode, sources,
+ healed_sinks);
+ } else if (IA_ISDIR(loc->inode->ia_type) && !partial) {
+ ret = ec_heal_entry(frame, ec, loc->inode, sources, healed_sinks,
+ &pending);
+ } else {
+ ret = 0;
+ memcpy(sources, participants, ec->nodes);
+ memcpy(healed_sinks, participants, ec->nodes);
+ }
- if (ret == 0) {
- good = ec_char_array_to_mask (sources, ec->nodes);
- bad = ec_char_array_to_mask (healed_sinks, ec->nodes);
- } else {
- op_ret = -1;
- op_errno = -ret;
- }
+ if (ret == 0) {
+ good = ec_char_array_to_mask(sources, ec->nodes);
+ bad = ec_char_array_to_mask(healed_sinks, ec->nodes);
+ } else {
+ op_ret = -1;
+ op_errno = -ret;
+ }
+ msources = alloca0(ec->nodes);
+ mhealed_sinks = alloca0(ec->nodes);
+ ret = ec_heal_metadata(frame, ec, loc->inode, msources, mhealed_sinks);
+ if (ret == 0) {
+ mgood = ec_char_array_to_mask(msources, ec->nodes);
+ mbad = ec_char_array_to_mask(mhealed_sinks, ec->nodes);
+ } else {
+ op_ret = -1;
+ op_errno = -ret;
+ }
out:
- if (fop->cbks.heal) {
- fop->cbks.heal (fop->req_frame, fop, fop->xl, op_ret,
- op_errno, ec_char_array_to_mask (participants,
- ec->nodes),
- mgood & good, mbad & bad, NULL);
- }
- if (frame)
- STACK_DESTROY (frame->root);
- return;
+ ec_reset_entry_healing(fop);
+ if (fop->cbks.heal) {
+ fop->cbks.heal(fop->req_frame, fop->data, fop->xl, op_ret, op_errno,
+ ec_char_array_to_mask(participants, ec->nodes),
+ mgood & good, mbad & bad, pending, NULL);
+ }
+ if (frame)
+ STACK_DESTROY(frame->root);
+
+ return;
}
int
-ec_synctask_heal_wrap (void *opaque)
+ec_synctask_heal_wrap(void *opaque)
{
- ec_fop_data_t *fop = opaque;
- ec_heal_do (fop->xl, fop, &fop->loc[0], fop->int32);
- return 0;
+ ec_fop_data_t *fop = opaque;
+ ec_heal_do(fop->xl, fop, &fop->loc[0], fop->int32);
+ return 0;
}
int
-ec_heal_done (int ret, call_frame_t *heal, void *opaque)
+ec_heal_done(int ret, call_frame_t *heal, void *opaque)
{
- if (opaque)
- ec_fop_data_release (opaque);
- return 0;
+ if (opaque)
+ ec_fop_data_release(opaque);
+ return 0;
}
-ec_fop_data_t*
-__ec_dequeue_heals (ec_t *ec)
+ec_fop_data_t *
+__ec_dequeue_heals(ec_t *ec)
{
- ec_fop_data_t *fop = NULL;
+ ec_fop_data_t *fop = NULL;
- if (list_empty (&ec->heal_waiting))
- goto none;
+ if (list_empty(&ec->heal_waiting))
+ goto none;
- if ((ec->background_heals > 0) && (ec->healers >= ec->background_heals))
- goto none;
+ if ((ec->background_heals > 0) && (ec->healers >= ec->background_heals))
+ goto none;
- fop = list_entry(ec->heal_waiting.next, ec_fop_data_t, healer);
- ec->heal_waiters--;
- list_del_init(&fop->healer);
- list_add(&fop->healer, &ec->healing);
- ec->healers++;
- return fop;
+ fop = list_entry(ec->heal_waiting.next, ec_fop_data_t, healer);
+ ec->heal_waiters--;
+ list_del_init(&fop->healer);
+ list_add(&fop->healer, &ec->healing);
+ ec->healers++;
+ return fop;
none:
- gf_msg_debug (ec->xl->name, 0, "Num healers: %d, Num Waiters: %d",
- ec->healers, ec->heal_waiters);
- return NULL;
+ gf_msg_debug(ec->xl->name, 0, "Num healers: %d, Num Waiters: %d",
+ ec->healers, ec->heal_waiters);
+ return NULL;
}
void
-ec_heal_fail (ec_t *ec, ec_fop_data_t *fop)
+ec_heal_fail(ec_t *ec, ec_fop_data_t *fop)
{
- if (fop->cbks.heal) {
- fop->cbks.heal (fop->req_frame, NULL, ec->xl, -1, fop->error, 0, 0,
- 0, NULL);
- }
- ec_fop_data_release (fop);
+ if (fop->cbks.heal) {
+ fop->cbks.heal(fop->req_frame, fop->data, ec->xl, -1, fop->error, 0, 0,
+ 0, 0, NULL);
+ }
+ ec_fop_data_release(fop);
}
void
-ec_launch_heal (ec_t *ec, ec_fop_data_t *fop)
+ec_launch_heal(ec_t *ec, ec_fop_data_t *fop)
{
- int ret = 0;
+ int ret = 0;
+ call_frame_t *frame = NULL;
- ret = synctask_new (ec->xl->ctx->env, ec_synctask_heal_wrap,
- ec_heal_done, NULL, fop);
- if (ret < 0) {
- ec_fop_set_error(fop, ENOMEM);
- ec_heal_fail (ec, fop);
- }
+ frame = create_frame(ec->xl, ec->xl->ctx->pool);
+ if (!frame) {
+ ret = -1;
+ goto out;
+ }
+
+ ec_owner_set(frame, frame->root);
+ /*Do heal as root*/
+ frame->root->uid = 0;
+ frame->root->gid = 0;
+ /*Mark the fops as internal*/
+ frame->root->pid = GF_CLIENT_PID_SELF_HEALD;
+
+ ret = synctask_new(ec->xl->ctx->env, ec_synctask_heal_wrap, ec_heal_done,
+ frame, fop);
+out:
+ if (ret < 0) {
+ ec_fop_set_error(fop, ENOMEM);
+ ec_heal_fail(ec, fop);
+ }
+
+ if (frame)
+ STACK_DESTROY(frame->root);
}
void
-ec_handle_healers_done (ec_fop_data_t *fop)
+ec_handle_healers_done(ec_fop_data_t *fop)
{
- ec_t *ec = fop->xl->private;
- ec_fop_data_t *heal_fop = NULL;
+ ec_t *ec = fop->xl->private;
+ ec_fop_data_t *heal_fop = NULL;
- if (list_empty (&fop->healer))
- return;
+ if (list_empty(&fop->healer))
+ return;
- LOCK (&ec->lock);
+ LOCK(&ec->lock);
- list_del_init (&fop->healer);
+ list_del_init(&fop->healer);
- do {
- ec->healers--;
- heal_fop = __ec_dequeue_heals (ec);
+ do {
+ ec->healers--;
+ heal_fop = __ec_dequeue_heals(ec);
- if ((heal_fop != NULL) && ec->shutdown) {
- /* This will prevent ec_handle_healers_done() to be
- * called recursively. That would be problematic if
- * the queue is too big. */
- list_del_init(&heal_fop->healer);
+ if ((heal_fop != NULL) && ec->shutdown) {
+ /* This will prevent ec_handle_healers_done() to be
+ * called recursively. That would be problematic if
+ * the queue is too big. */
+ list_del_init(&heal_fop->healer);
- UNLOCK(&ec->lock);
+ UNLOCK(&ec->lock);
- ec_fop_set_error(fop, ENOTCONN);
- ec_heal_fail(ec, heal_fop);
+ ec_fop_set_error(fop, ENOTCONN);
+ ec_heal_fail(ec, heal_fop);
- LOCK(&ec->lock);
- }
- } while ((heal_fop != NULL) && ec->shutdown);
+ LOCK(&ec->lock);
+ }
+ } while ((heal_fop != NULL) && ec->shutdown);
- UNLOCK (&ec->lock);
+ UNLOCK(&ec->lock);
- if (heal_fop)
- ec_launch_heal (ec, heal_fop);
+ if (heal_fop)
+ ec_launch_heal(ec, heal_fop);
+}
+
+gf_boolean_t
+ec_is_entry_healing(ec_fop_data_t *fop)
+{
+ ec_inode_t *ctx = NULL;
+ int32_t heal_count = 0;
+ loc_t *loc = NULL;
+
+ loc = &fop->loc[0];
+
+ LOCK(&loc->inode->lock);
+ {
+ ctx = __ec_inode_get(loc->inode, fop->xl);
+ if (ctx) {
+ heal_count = ctx->heal_count;
+ }
+ }
+ UNLOCK(&loc->inode->lock);
+ GF_ASSERT(heal_count >= 0);
+ return heal_count;
}
void
-ec_heal_throttle (xlator_t *this, ec_fop_data_t *fop)
-{
- gf_boolean_t can_heal = _gf_true;
- ec_t *ec = this->private;
-
- if (fop->req_frame == NULL) {
-
- LOCK (&ec->lock);
- {
- if ((ec->background_heals > 0) &&
- (ec->heal_wait_qlen + ec->background_heals) >
- (ec->heal_waiters + ec->healers)) {
- list_add_tail(&fop->healer, &ec->heal_waiting);
- ec->heal_waiters++;
- fop = __ec_dequeue_heals (ec);
- } else {
- can_heal = _gf_false;
- }
+ec_heal_throttle(xlator_t *this, ec_fop_data_t *fop)
+{
+ gf_boolean_t can_heal = _gf_true;
+ ec_t *ec = this->private;
+ ec_fop_data_t *fop_rel = NULL;
+
+ if (fop->req_frame == NULL) {
+ LOCK(&ec->lock);
+ {
+ if ((ec->background_heals > 0) &&
+ (ec->heal_wait_qlen + ec->background_heals) >
+ (ec->heal_waiters + ec->healers)) {
+ if (!ec_is_entry_healing(fop)) {
+ list_add_tail(&fop->healer, &ec->heal_waiting);
+ ec->heal_waiters++;
+ ec_set_entry_healing(fop);
+ } else {
+ fop_rel = fop;
}
- UNLOCK (&ec->lock);
+ fop = __ec_dequeue_heals(ec);
+ } else {
+ can_heal = _gf_false;
+ }
}
+ UNLOCK(&ec->lock);
+ }
- if (can_heal) {
- if (fop)
- ec_launch_heal (ec, fop);
- } else {
- gf_msg_debug (this->name, 0, "Max number of heals are "
- "pending, background self-heal rejected");
- ec_fop_set_error(fop, EBUSY);
- ec_heal_fail (ec, fop);
+ if (can_heal) {
+ if (fop) {
+ if (fop->req_frame != NULL) {
+ ec_set_entry_healing(fop);
+ }
+ ec_launch_heal(ec, fop);
}
+ } else {
+ gf_msg_debug(this->name, 0,
+ "Max number of heals are "
+ "pending, background self-heal rejected");
+ ec_fop_set_error(fop, EBUSY);
+ ec_heal_fail(ec, fop);
+ }
+ if (fop_rel) {
+ ec_heal_done(0, NULL, fop_rel);
+ }
}
void
-ec_heal (call_frame_t *frame, xlator_t *this, uintptr_t target,
- int32_t minimum, fop_heal_cbk_t func, void *data, loc_t *loc,
- int32_t partial, dict_t *xdata)
+ec_heal(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_heal_cbk_t func, void *data, loc_t *loc,
+ int32_t partial, dict_t *xdata)
{
- ec_cbk_t callback = { .heal = func };
+ ec_cbk_t callback = {.heal = func};
ec_fop_data_t *fop = NULL;
int32_t err = EINVAL;
- gf_msg_trace ("ec", 0, "EC(HEAL) %p", frame);
+ gf_msg_trace("ec", 0, "EC(HEAL) %p", frame);
VALIDATE_OR_GOTO(this, fail);
GF_VALIDATE_OR_GOTO(this->name, this->private, fail);
- if (!loc || !loc->inode || gf_uuid_is_null (loc->inode->gfid))
- goto fail;
+ if (!loc || !loc->inode || gf_uuid_is_null(loc->inode->gfid))
+ goto fail;
if (frame && frame->local)
- goto fail;
- fop = ec_fop_data_allocate (frame, this, EC_FOP_HEAL, 0, target, minimum,
- NULL, NULL, callback, data);
+ goto fail;
+ fop = ec_fop_data_allocate(frame, this, EC_FOP_HEAL, 0, target, fop_flags,
+ NULL, NULL, callback, data);
err = ENOMEM;
@@ -2735,454 +2908,460 @@ ec_heal (call_frame_t *frame, xlator_t *this, uintptr_t target,
if (xdata)
fop->xdata = dict_ref(xdata);
- ec_heal_throttle (this, fop);
+ ec_heal_throttle(this, fop);
return;
fail:
if (fop)
- ec_fop_data_release (fop);
+ ec_fop_data_release(fop);
if (func)
- func (frame, NULL, this, -1, err, 0, 0, 0, NULL);
+ func(frame, data, this, -1, err, 0, 0, 0, 0, NULL);
}
int
-ec_replace_heal_done (int ret, call_frame_t *heal, void *opaque)
+ec_replace_heal_done(int ret, call_frame_t *heal, void *opaque)
{
- ec_t *ec = opaque;
+ ec_t *ec = opaque;
+ gf_boolean_t last_fop = _gf_false;
- gf_msg_debug (ec->xl->name, 0,
- "getxattr on bricks is done ret %d", ret);
- return 0;
+ if (GF_ATOMIC_DEC(ec->async_fop_count) == 0) {
+ LOCK(&ec->lock);
+ {
+ last_fop = __ec_is_last_fop(ec);
+ }
+ UNLOCK(&ec->lock);
+ }
+ gf_msg_debug(ec->xl->name, 0, "getxattr on bricks is done ret %d", ret);
+
+ if (last_fop)
+ ec_pending_fops_completed(ec);
+
+ return 0;
}
int32_t
-ec_replace_heal (ec_t *ec, inode_t *inode)
+ec_replace_heal(ec_t *ec, inode_t *inode)
{
- loc_t loc = {0};
- int ret = 0;
+ loc_t loc = {0};
+ int ret = 0;
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
- ret = syncop_getxattr (ec->xl, &loc, NULL, EC_XATTR_HEAL,
- NULL, NULL);
- if (ret < 0)
- gf_msg_debug (ec->xl->name, 0,
- "Heal failed for replace brick ret = %d", ret);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+ ret = syncop_getxattr(ec->xl, &loc, NULL, EC_XATTR_HEAL, NULL, NULL);
+ if (ret < 0)
+ gf_msg_debug(ec->xl->name, 0, "Heal failed for replace brick ret = %d",
+ ret);
- /* Once the root inode has been checked, it might have triggered a
- * self-heal on it after a replace brick command or for some other
- * reason. It can also happen that the volume already had damaged
- * files in the index, even if the heal on the root directory failed.
- * In both cases we need to wake all index healers to continue
- * healing remaining entries that are marked as dirty. */
- ec_shd_index_healer_wake(ec);
+ /* Once the root inode has been checked, it might have triggered a
+ * self-heal on it after a replace brick command or for some other
+ * reason. It can also happen that the volume already had damaged
+ * files in the index, even if the heal on the root directory failed.
+ * In both cases we need to wake all index healers to continue
+ * healing remaining entries that are marked as dirty. */
+ ec_shd_index_healer_wake(ec);
- loc_wipe (&loc);
- return ret;
+ loc_wipe(&loc);
+ return ret;
}
int32_t
-ec_replace_brick_heal_wrap (void *opaque)
+ec_replace_brick_heal_wrap(void *opaque)
{
- ec_t *ec = opaque;
- inode_table_t *itable = NULL;
- int32_t ret = -1;
+ ec_t *ec = opaque;
+ inode_table_t *itable = NULL;
+ int32_t ret = -1;
- if (ec->xl->itable)
- itable = ec->xl->itable;
- else
- goto out;
- ret = ec_replace_heal (ec, itable->root);
+ if (ec->xl->itable)
+ itable = ec->xl->itable;
+ else
+ goto out;
+
+ if (xlator_is_cleanup_starting(ec->xl))
+ goto out;
+
+ ret = ec_replace_heal(ec, itable->root);
out:
- return ret;
+ return ret;
}
int32_t
-ec_launch_replace_heal (ec_t *ec)
+ec_launch_replace_heal(ec_t *ec)
{
- int ret = -1;
+ int ret = -1;
- if (!ec)
- return ret;
- ret = synctask_new (ec->xl->ctx->env, ec_replace_brick_heal_wrap,
- ec_replace_heal_done, NULL, ec);
- if (ret < 0) {
- gf_msg_debug (ec->xl->name, 0,
- "Heal failed for replace brick ret = %d", ret);
- }
- return ret;
+ ret = synctask_new(ec->xl->ctx->env, ec_replace_brick_heal_wrap,
+ ec_replace_heal_done, NULL, ec);
+
+ if (ret < 0) {
+ gf_msg_debug(ec->xl->name, 0, "Heal failed for replace brick ret = %d",
+ ret);
+ ec_replace_heal_done(-1, NULL, ec);
+ }
+
+ return ret;
}
int32_t
ec_set_heal_info(dict_t **dict_rsp, char *status)
{
- dict_t *dict = NULL;
- int ret = 0;
+ dict_t *dict = NULL;
+ int ret = 0;
- dict = dict_new ();
- if (!dict) {
- ret = -ENOMEM;
- goto out;
- }
- ret = dict_set_str (dict, "heal-info", status);
- if (ret) {
- gf_msg (THIS->name, GF_LOG_WARNING, -ret,
- EC_MSG_HEAL_FAIL,
- "Failed to set heal-info key to "
- "%s", status);
- dict_unref(dict);
- dict = NULL;
- }
- *dict_rsp = dict;
+ dict = dict_new();
+ if (!dict) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ ret = dict_set_str(dict, "heal-info", status);
+ if (ret) {
+ gf_msg(THIS->name, GF_LOG_WARNING, -ret, EC_MSG_HEAL_FAIL,
+ "Failed to set heal-info key to "
+ "%s",
+ status);
+ dict_unref(dict);
+ dict = NULL;
+ }
+ *dict_rsp = dict;
out:
- return ret;
+ return ret;
}
static int32_t
-_need_heal_calculate (ec_t *ec, uint64_t *dirty, unsigned char *sources,
- gf_boolean_t self_locked, int32_t lock_count,
- ec_heal_need_t *need_heal)
-{
- int i = 0;
- int source_count = 0;
-
- source_count = EC_COUNT (sources, ec->nodes);
- if (source_count == ec->nodes) {
- *need_heal = EC_HEAL_NONEED;
- if (self_locked || lock_count == 0) {
- for (i = 0; i < ec->nodes; i++) {
- if (dirty[i]) {
- *need_heal = EC_HEAL_MUST;
- goto out;
- }
- }
- } else {
- for (i = 0; i < ec->nodes; i++) {
- /* Since each lock can only increment the dirty
- * count once, if dirty is > 1 it means that
- * another operation has left the dirty count
- * set and this indicates a problem in the
- * inode.*/
- if (dirty[i] > 1) {
- *need_heal = EC_HEAL_MUST;
- goto out;
- }
- }
+_need_heal_calculate(ec_t *ec, uint64_t *dirty, unsigned char *sources,
+ gf_boolean_t self_locked, int32_t lock_count,
+ ec_heal_need_t *need_heal, uint64_t *versions)
+{
+ int i = 0;
+ int source_count = 0;
+
+ source_count = EC_COUNT(sources, ec->nodes);
+ if (source_count == ec->nodes) {
+ *need_heal = EC_HEAL_NONEED;
+ if (self_locked || lock_count == 0) {
+ for (i = 0; i < ec->nodes; i++) {
+ if (dirty[i] || (versions[i] != versions[0])) {
+ *need_heal = EC_HEAL_MUST;
+ goto out;
}
+ }
+ /* If lock count is 0, all dirty flags are 0 and all the
+ * versions are macthing then why are we here. It looks
+ * like something went wrong while removing the index entries
+ * after completing a successful heal or fop. In this case
+ * we need to remove this index entry to avoid triggering heal
+ * in a loop and causing lookups again and again*/
+ *need_heal = EC_HEAL_PURGE_INDEX;
} else {
- *need_heal = EC_HEAL_MUST;
+ for (i = 0; i < ec->nodes; i++) {
+ /* Since each lock can only increment the dirty
+ * count once, if dirty is > 1 it means that
+ * another operation has left the dirty count
+ * set and this indicates a problem in the
+ * inode.*/
+ if (dirty[i] > 1) {
+ *need_heal = EC_HEAL_MUST;
+ goto out;
+ }
+ if (dirty[i] != dirty[0] || (versions[i] != versions[0])) {
+ *need_heal = EC_HEAL_MAYBE;
+ }
+ }
}
+ } else {
+ *need_heal = EC_HEAL_MUST;
+ }
out:
- return source_count;
+ return source_count;
}
static int32_t
-ec_need_metadata_heal (ec_t *ec, inode_t *inode, default_args_cbk_t *replies,
- int32_t lock_count, gf_boolean_t self_locked,
- gf_boolean_t thorough, ec_heal_need_t *need_heal)
-{
- uint64_t *dirty = NULL;
- unsigned char *sources = NULL;
- unsigned char *healed_sinks = NULL;
- uint64_t *meta_versions = NULL;
- int ret = 0;
- int i = 0;
-
- sources = alloca0(ec->nodes);
- healed_sinks = alloca0(ec->nodes);
- dirty = alloca0 (ec->nodes * sizeof (*dirty));
- meta_versions = alloca0 (ec->nodes * sizeof (*meta_versions));
- ret = ec_heal_metadata_find_direction (ec, replies, meta_versions,
- dirty, sources, healed_sinks);
- if (ret < 0 && ret != -EIO) {
- goto out;
- }
+ec_need_metadata_heal(ec_t *ec, inode_t *inode, default_args_cbk_t *replies,
+ int32_t lock_count, gf_boolean_t self_locked,
+ gf_boolean_t thorough, ec_heal_need_t *need_heal)
+{
+ uint64_t *dirty = NULL;
+ unsigned char *sources = NULL;
+ unsigned char *healed_sinks = NULL;
+ uint64_t *meta_versions = NULL;
+ int ret = 0;
+
+ sources = alloca0(ec->nodes);
+ healed_sinks = alloca0(ec->nodes);
+ dirty = alloca0(ec->nodes * sizeof(*dirty));
+ meta_versions = alloca0(ec->nodes * sizeof(*meta_versions));
+ ret = ec_heal_metadata_find_direction(ec, replies, meta_versions, dirty,
+ sources, healed_sinks);
+ if (ret < 0 && ret != -EIO) {
+ goto out;
+ }
- ret = _need_heal_calculate (ec, dirty, sources, self_locked, lock_count,
- need_heal);
- if (ret == ec->nodes && *need_heal == EC_HEAL_NONEED) {
- for (i = 1; i < ec->nodes; i++) {
- if (meta_versions[i] != meta_versions[0]) {
- *need_heal = EC_HEAL_MUST;
- goto out;
- }
- }
- }
+ ret = _need_heal_calculate(ec, dirty, sources, self_locked, lock_count,
+ need_heal, meta_versions);
out:
- return ret;
+ return ret;
}
static int32_t
-ec_need_data_heal (ec_t *ec, inode_t *inode, default_args_cbk_t *replies,
- int32_t lock_count, gf_boolean_t self_locked,
- gf_boolean_t thorough, ec_heal_need_t *need_heal)
-{
- uint64_t *dirty = NULL;
- unsigned char *sources = NULL;
- unsigned char *healed_sinks = NULL;
- uint64_t *data_versions = NULL;
- uint64_t *size = NULL;
- int ret = 0;
-
- sources = alloca0(ec->nodes);
- healed_sinks = alloca0(ec->nodes);
- dirty = alloca0 (ec->nodes * sizeof (*dirty));
- data_versions = alloca0 (ec->nodes * sizeof (*data_versions));
- size = alloca0 (ec->nodes * sizeof (*size));
-
- /* When dd is going on and heal info is called there is a very good
- * chance for on disk sizes to mismatch eventhough nothing is wrong
- * we don't need ondisk size check there. But if the file is either
- * self-locked or the caller wants a thorough check then make sure to
- * perform on disk check also. */
- ret = ec_heal_data_find_direction (ec, replies, data_versions,
- dirty, size, sources, healed_sinks,
- self_locked || thorough,
- EC_COMBINE_XDATA);
- if (ret < 0 && ret != -EIO) {
- goto out;
- }
+ec_need_data_heal(ec_t *ec, inode_t *inode, default_args_cbk_t *replies,
+ int32_t lock_count, gf_boolean_t self_locked,
+ gf_boolean_t thorough, ec_heal_need_t *need_heal)
+{
+ uint64_t *dirty = NULL;
+ unsigned char *sources = NULL;
+ unsigned char *healed_sinks = NULL;
+ uint64_t *data_versions = NULL;
+ uint64_t *size = NULL;
+ int ret = 0;
+
+ sources = alloca0(ec->nodes);
+ healed_sinks = alloca0(ec->nodes);
+ dirty = alloca0(ec->nodes * sizeof(*dirty));
+ data_versions = alloca0(ec->nodes * sizeof(*data_versions));
+ size = alloca0(ec->nodes * sizeof(*size));
+
+ /* When dd is going on and heal info is called there is a very good
+ * chance for on disk sizes to mismatch even though nothing is wrong
+ * we don't need ondisk size check there. But if the file is either
+ * self-locked or the caller wants a thorough check then make sure to
+ * perform on disk check also. */
+ ret = ec_heal_data_find_direction(
+ ec, replies, data_versions, dirty, size, sources, healed_sinks,
+ self_locked || thorough, EC_COMBINE_XDATA);
+ if (ret < 0 && ret != -EIO) {
+ goto out;
+ }
- ret = _need_heal_calculate (ec, dirty, sources, self_locked, lock_count,
- need_heal);
+ ret = _need_heal_calculate(ec, dirty, sources, self_locked, lock_count,
+ need_heal, data_versions);
out:
- return ret;
+ return ret;
}
static int32_t
-ec_need_entry_heal (ec_t *ec, inode_t *inode, default_args_cbk_t *replies,
- int32_t lock_count, gf_boolean_t self_locked,
- gf_boolean_t thorough, ec_heal_need_t *need_heal)
-{
- uint64_t *dirty = NULL;
- unsigned char *sources = NULL;
- unsigned char *healed_sinks = NULL;
- uint64_t *data_versions = NULL;
- int ret = 0;
-
- sources = alloca0(ec->nodes);
- healed_sinks = alloca0(ec->nodes);
- dirty = alloca0 (ec->nodes * sizeof (*dirty));
- data_versions = alloca0 (ec->nodes * sizeof (*data_versions));
-
- ret = ec_heal_entry_find_direction (ec, replies, data_versions,
- dirty, sources, healed_sinks);
- if (ret < 0 && ret != -EIO) {
- goto out;
- }
+ec_need_entry_heal(ec_t *ec, inode_t *inode, default_args_cbk_t *replies,
+ int32_t lock_count, gf_boolean_t self_locked,
+ gf_boolean_t thorough, ec_heal_need_t *need_heal)
+{
+ uint64_t *dirty = NULL;
+ unsigned char *sources = NULL;
+ unsigned char *healed_sinks = NULL;
+ uint64_t *data_versions = NULL;
+ int ret = 0;
+
+ sources = alloca0(ec->nodes);
+ healed_sinks = alloca0(ec->nodes);
+ dirty = alloca0(ec->nodes * sizeof(*dirty));
+ data_versions = alloca0(ec->nodes * sizeof(*data_versions));
+
+ ret = ec_heal_entry_find_direction(ec, replies, data_versions, dirty,
+ sources, healed_sinks);
+ if (ret < 0 && ret != -EIO) {
+ goto out;
+ }
- ret = _need_heal_calculate (ec, dirty, sources, self_locked, lock_count,
- need_heal);
+ ret = _need_heal_calculate(ec, dirty, sources, self_locked, lock_count,
+ need_heal, data_versions);
out:
- return ret;
+ return ret;
}
static int32_t
-ec_need_heal (ec_t *ec, inode_t *inode, default_args_cbk_t *replies,
- int32_t lock_count, gf_boolean_t self_locked,
- gf_boolean_t thorough, ec_heal_need_t *need_heal)
+ec_need_heal(ec_t *ec, inode_t *inode, default_args_cbk_t *replies,
+ int32_t lock_count, gf_boolean_t self_locked,
+ gf_boolean_t thorough, ec_heal_need_t *need_heal)
{
- int ret = 0;
+ int ret = 0;
+ ret = ec_need_metadata_heal(ec, inode, replies, lock_count, self_locked,
+ thorough, need_heal);
+ if (ret < 0)
+ goto out;
- ret = ec_need_metadata_heal (ec, inode, replies, lock_count,
- self_locked, thorough, need_heal);
- if (ret < 0)
- goto out;
-
- if (*need_heal == EC_HEAL_MUST)
- goto out;
+ if (*need_heal == EC_HEAL_MUST)
+ goto out;
- if (inode->ia_type == IA_IFREG) {
- ret = ec_need_data_heal (ec, inode, replies, lock_count,
- self_locked, thorough, need_heal);
- } else if (inode->ia_type == IA_IFDIR) {
- ret = ec_need_entry_heal (ec, inode, replies, lock_count,
- self_locked, thorough, need_heal);
- }
+ if (inode->ia_type == IA_IFREG) {
+ ret = ec_need_data_heal(ec, inode, replies, lock_count, self_locked,
+ thorough, need_heal);
+ } else if (inode->ia_type == IA_IFDIR) {
+ ret = ec_need_entry_heal(ec, inode, replies, lock_count, self_locked,
+ thorough, need_heal);
+ }
out:
- return ret;
+ return ret;
}
int32_t
-ec_heal_inspect (call_frame_t *frame, ec_t *ec,
- inode_t *inode, unsigned char *locked_on,
- gf_boolean_t self_locked, gf_boolean_t thorough,
- ec_heal_need_t *need_heal)
-{
- loc_t loc = {0};
- int i = 0;
- int ret = 0;
- dict_t *xdata = NULL;
- uint64_t zero_array[2] = {0};
- uint64_t zero_value = 0;
- unsigned char *output = NULL;
- default_args_cbk_t *replies = NULL;
- int32_t lock_count = 0;
-
- EC_REPLIES_ALLOC (replies, ec->nodes);
- output = alloca0 (ec->nodes);
-
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
-
- xdata = dict_new ();
- if (!xdata ||
- dict_set_static_bin (xdata, EC_XATTR_VERSION, zero_array,
- sizeof (zero_array)) ||
- dict_set_static_bin (xdata, EC_XATTR_DIRTY, zero_array,
- sizeof (zero_array)) ||
- dict_set_static_bin (xdata, EC_XATTR_SIZE, &zero_value,
- sizeof (zero_value))) {
- ret = -ENOMEM;
- goto out;
- }
+ec_heal_inspect(call_frame_t *frame, ec_t *ec, inode_t *inode,
+ unsigned char *locked_on, gf_boolean_t self_locked,
+ gf_boolean_t thorough, ec_heal_need_t *need_heal)
+{
+ loc_t loc = {0};
+ int i = 0;
+ int ret = 0;
+ dict_t *xdata = NULL;
+ uint64_t zero_array[2] = {0};
+ uint64_t zero_value = 0;
+ unsigned char *output = NULL;
+ default_args_cbk_t *replies = NULL;
+ int32_t lock_count = 0;
+
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ output = alloca0(ec->nodes);
+
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+
+ xdata = dict_new();
+ if (!xdata ||
+ dict_set_static_bin(xdata, EC_XATTR_VERSION, zero_array,
+ sizeof(zero_array)) ||
+ dict_set_static_bin(xdata, EC_XATTR_DIRTY, zero_array,
+ sizeof(zero_array)) ||
+ dict_set_static_bin(xdata, EC_XATTR_SIZE, &zero_value,
+ sizeof(zero_value))) {
+ ret = -ENOMEM;
+ goto out;
+ }
- if (!self_locked) {
- ret = dict_set_str(xdata, GLUSTERFS_INODELK_DOM_COUNT,
- ec->xl->name);
- if (ret) {
- ret = -ENOMEM;
- goto out;
- }
+ if (!self_locked) {
+ ret = dict_set_str(xdata, GLUSTERFS_INODELK_DOM_COUNT, ec->xl->name);
+ if (ret) {
+ ret = -ENOMEM;
+ goto out;
}
+ }
- ret = cluster_lookup (ec->xl_list, locked_on, ec->nodes, replies,
- output, frame, ec->xl, &loc, xdata);
+ ret = cluster_lookup(ec->xl_list, locked_on, ec->nodes, replies, output,
+ frame, ec->xl, &loc, xdata);
- if (ret != ec->nodes) {
- ret = ec->nodes;
- *need_heal = EC_HEAL_MUST;
- goto out;
- }
+ if (ret != ec->nodes) {
+ ret = ec->nodes;
+ *need_heal = EC_HEAL_MUST;
+ goto out;
+ }
- if (self_locked)
- goto need_heal;
+ if (self_locked)
+ goto need_heal;
- for (i = 0; i < ec->nodes; i++) {
- if (!output[i] || !replies[i].xdata) {
- continue;
- }
- if ((dict_get_int32 (replies[i].xdata, GLUSTERFS_INODELK_COUNT,
- &lock_count) == 0) && lock_count > 0) {
- break;
- }
+ for (i = 0; i < ec->nodes; i++) {
+ if (!output[i] || !replies[i].xdata) {
+ continue;
}
-need_heal:
- ret = ec_need_heal (ec, inode, replies, lock_count,
- self_locked, thorough, need_heal);
-
- if (!self_locked && *need_heal == EC_HEAL_MUST) {
- *need_heal = EC_HEAL_MAYBE;
+ if ((dict_get_int32(replies[i].xdata, GLUSTERFS_INODELK_COUNT,
+ &lock_count) == 0) &&
+ lock_count > 0) {
+ break;
}
+ }
+need_heal:
+ ret = ec_need_heal(ec, inode, replies, lock_count, self_locked, thorough,
+ need_heal);
out:
- cluster_replies_wipe (replies, ec->nodes);
- loc_wipe (&loc);
- if (xdata) {
- dict_unref(xdata);
- }
- return ret;
+ cluster_replies_wipe(replies, ec->nodes);
+ loc_wipe(&loc);
+ if (xdata) {
+ dict_unref(xdata);
+ }
+ return ret;
}
int32_t
-ec_heal_locked_inspect (call_frame_t *frame, ec_t *ec, inode_t *inode,
- ec_heal_need_t *need_heal)
-{
- unsigned char *locked_on = NULL;
- unsigned char *up_subvols = NULL;
- unsigned char *output = NULL;
- default_args_cbk_t *replies = NULL;
- int ret = 0;
-
- EC_REPLIES_ALLOC (replies, ec->nodes);
- locked_on = alloca0(ec->nodes);
- output = alloca0(ec->nodes);
- up_subvols = alloca0(ec->nodes);
- ec_mask_to_char_array (ec->xl_up, up_subvols, ec->nodes);
-
- ret = cluster_inodelk (ec->xl_list, up_subvols, ec->nodes,
- replies, locked_on, frame, ec->xl,
- ec->xl->name, inode, 0, 0);
- if (ret != ec->nodes) {
- *need_heal = EC_HEAL_MUST;
- goto unlock;
- }
- ret = ec_heal_inspect (frame, ec, inode, locked_on, _gf_true, _gf_true,
- need_heal);
+ec_heal_locked_inspect(call_frame_t *frame, ec_t *ec, inode_t *inode,
+ ec_heal_need_t *need_heal)
+{
+ unsigned char *locked_on = NULL;
+ unsigned char *up_subvols = NULL;
+ unsigned char *output = NULL;
+ default_args_cbk_t *replies = NULL;
+ int ret = 0;
+
+ EC_REPLIES_ALLOC(replies, ec->nodes);
+ locked_on = alloca0(ec->nodes);
+ output = alloca0(ec->nodes);
+ up_subvols = alloca0(ec->nodes);
+ ec_mask_to_char_array(ec->xl_up, up_subvols, ec->nodes);
+
+ ret = cluster_inodelk(ec->xl_list, up_subvols, ec->nodes, replies,
+ locked_on, frame, ec->xl, ec->xl->name, inode, 0, 0);
+ if (ret != ec->nodes) {
+ *need_heal = EC_HEAL_MUST;
+ goto unlock;
+ }
+ ret = ec_heal_inspect(frame, ec, inode, locked_on, _gf_true, _gf_true,
+ need_heal);
unlock:
- cluster_uninodelk (ec->xl_list, locked_on, ec->nodes,
- replies, output, frame, ec->xl,
- ec->xl->name, inode, 0, 0);
- cluster_replies_wipe (replies, ec->nodes);
- return ret;
+ cluster_uninodelk(ec->xl_list, locked_on, ec->nodes, replies, output, frame,
+ ec->xl, ec->xl->name, inode, 0, 0);
+ cluster_replies_wipe(replies, ec->nodes);
+ return ret;
}
int32_t
-ec_get_heal_info (xlator_t *this, loc_t *entry_loc, dict_t **dict_rsp)
+ec_get_heal_info(xlator_t *this, loc_t *entry_loc, dict_t **dict_rsp)
{
- int ret = -ENOMEM;
- ec_heal_need_t need_heal = EC_HEAL_NONEED;
- call_frame_t *frame = NULL;
- ec_t *ec = NULL;
- unsigned char *up_subvols = NULL;
- loc_t loc = {0, };
-
- VALIDATE_OR_GOTO(this, out);
- GF_VALIDATE_OR_GOTO(this->name, entry_loc, out);
+ int ret = -ENOMEM;
+ ec_heal_need_t need_heal = EC_HEAL_NONEED;
+ call_frame_t *frame = NULL;
+ ec_t *ec = NULL;
+ unsigned char *up_subvols = NULL;
+ loc_t loc = {
+ 0,
+ };
- ec = this->private;
- up_subvols = alloca0(ec->nodes);
- ec_mask_to_char_array (ec->xl_up, up_subvols, ec->nodes);
+ VALIDATE_OR_GOTO(this, out);
+ GF_VALIDATE_OR_GOTO(this->name, entry_loc, out);
- if (EC_COUNT (up_subvols, ec->nodes) != ec->nodes) {
- need_heal = EC_HEAL_MUST;
- goto set_heal;
- }
- frame = create_frame (this, this->ctx->pool);
- if (!frame) {
- goto out;
- }
- ec_owner_set(frame, frame->root);
- frame->root->uid = 0;
- frame->root->gid = 0;
- frame->root->pid = GF_CLIENT_PID_SELF_HEALD;
+ ec = this->private;
+ up_subvols = alloca0(ec->nodes);
+ ec_mask_to_char_array(ec->xl_up, up_subvols, ec->nodes);
- if (loc_copy(&loc, entry_loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR,
- ENOMEM, EC_MSG_LOC_COPY_FAIL,
- "Failed to copy a location.");
- goto out;
- }
- if (!loc.inode) {
- ret = syncop_inode_find (this, this, loc.gfid,
- &loc.inode, NULL, NULL);
- if (ret < 0)
- goto out;
- }
-
- ret = ec_heal_inspect (frame, ec, loc.inode, up_subvols, _gf_false,
- _gf_false, &need_heal);
- if (ret == ec->nodes && need_heal == EC_HEAL_NONEED) {
- goto set_heal;
- }
- need_heal = EC_HEAL_NONEED;
- ret = ec_heal_locked_inspect (frame, ec, loc.inode,
- &need_heal);
+ if (EC_COUNT(up_subvols, ec->nodes) != ec->nodes) {
+ need_heal = EC_HEAL_MUST;
+ goto set_heal;
+ }
+ frame = create_frame(this, this->ctx->pool);
+ if (!frame) {
+ goto out;
+ }
+ ec_owner_set(frame, frame->root);
+ frame->root->uid = 0;
+ frame->root->gid = 0;
+ frame->root->pid = GF_CLIENT_PID_SELF_HEALD;
+
+ if (loc_copy(&loc, entry_loc) != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
+ goto out;
+ }
+ if (!loc.inode) {
+ ret = syncop_inode_find(this, this, loc.gfid, &loc.inode, NULL, NULL);
if (ret < 0)
- goto out;
+ goto out;
+ }
+
+ ret = ec_heal_inspect(frame, ec, loc.inode, up_subvols, _gf_false,
+ _gf_false, &need_heal);
+ if (ret == ec->nodes && need_heal != EC_HEAL_MAYBE) {
+ goto set_heal;
+ }
+ need_heal = EC_HEAL_NONEED;
+ ret = ec_heal_locked_inspect(frame, ec, loc.inode, &need_heal);
+ if (ret < 0)
+ goto out;
set_heal:
- if (need_heal == EC_HEAL_MUST) {
- ret = ec_set_heal_info (dict_rsp, "heal");
- } else {
- ret = ec_set_heal_info (dict_rsp, "no-heal");
- }
+ if (need_heal == EC_HEAL_MUST) {
+ ret = ec_set_heal_info(dict_rsp, "heal");
+ } else {
+ ret = ec_set_heal_info(dict_rsp, "no-heal");
+ }
out:
- if (frame) {
- STACK_DESTROY (frame->root);
- }
- loc_wipe (&loc);
- return ret;
+ if (frame) {
+ STACK_DESTROY(frame->root);
+ }
+ loc_wipe(&loc);
+ return ret;
}
diff --git a/xlators/cluster/ec/src/ec-heald.c b/xlators/cluster/ec/src/ec-heald.c
index 65d8b5c4284..5c1586bc9c5 100644
--- a/xlators/cluster/ec/src/ec-heald.c
+++ b/xlators/cluster/ec/src/ec-heald.c
@@ -8,611 +8,674 @@
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "defaults.h"
-#include "compat-errno.h"
+#include <glusterfs/defaults.h>
+#include <glusterfs/compat-errno.h>
#include "ec.h"
#include "ec-messages.h"
#include "ec-heald.h"
#include "ec-mem-types.h"
-#include "syncop.h"
-#include "syncop-utils.h"
+#include <glusterfs/syncop.h>
+#include <glusterfs/syncop-utils.h>
#include "protocol-common.h"
-#define ASSERT_LOCAL(this, healer) \
- do { \
- if (!ec_shd_is_subvol_local (this, healer->subvol)) { \
- healer->local = _gf_false; \
- if (safe_break (healer)) { \
- break; \
- } else { \
- continue; \
- } \
- } else { \
- healer->local = _gf_true; \
- } \
- } while (0);
-
-
-#define NTH_INDEX_HEALER(this, n) (&((((ec_t *)this->private))->shd.index_healers[n]))
-#define NTH_FULL_HEALER(this, n) (&((((ec_t *)this->private))->shd.full_healers[n]))
+#define NTH_INDEX_HEALER(this, n) \
+ (&((((ec_t *)this->private))->shd.index_healers[n]))
+#define NTH_FULL_HEALER(this, n) \
+ (&((((ec_t *)this->private))->shd.full_healers[n]))
gf_boolean_t
-ec_shd_is_subvol_local (xlator_t *this, int subvol)
+ec_shd_is_subvol_local(xlator_t *this, int subvol)
{
- ec_t *ec = NULL;
- gf_boolean_t is_local = _gf_false;
- loc_t loc = {0, };
-
- ec = this->private;
- loc.inode = this->itable->root;
- syncop_is_subvol_local (ec->xl_list[subvol], &loc, &is_local);
- return is_local;
+ ec_t *ec = NULL;
+ gf_boolean_t is_local = _gf_false;
+ loc_t loc = {
+ 0,
+ };
+
+ ec = this->private;
+ loc.inode = this->itable->root;
+ syncop_is_subvol_local(ec->xl_list[subvol], &loc, &is_local);
+ return is_local;
}
char *
-ec_subvol_name (xlator_t *this, int subvol)
+ec_subvol_name(xlator_t *this, int subvol)
{
- ec_t *ec = NULL;
+ ec_t *ec = NULL;
- ec = this->private;
- if (subvol < 0 || subvol > ec->nodes)
- return NULL;
+ ec = this->private;
+ if (subvol < 0 || subvol > ec->nodes)
+ return NULL;
- return ec->xl_list[subvol]->name;
+ return ec->xl_list[subvol]->name;
}
int
-__ec_shd_healer_wait (struct subvol_healer *healer)
+__ec_shd_healer_wait(struct subvol_healer *healer)
{
- ec_t *ec = NULL;
- struct timespec wait_till = {0, };
- int ret = 0;
+ ec_t *ec = NULL;
+ struct timespec wait_till = {
+ 0,
+ };
+ int ret = 0;
- ec = healer->this->private;
+ ec = healer->this->private;
disabled_loop:
- wait_till.tv_sec = time (NULL) + 60;
-
- while (!healer->rerun) {
- ret = pthread_cond_timedwait (&healer->cond,
- &healer->mutex,
- &wait_till);
- if (ret == ETIMEDOUT)
- break;
- }
-
- ret = healer->rerun;
- healer->rerun = 0;
-
- if (!ec->shd.enabled || !ec->up)
- goto disabled_loop;
+ wait_till.tv_sec = gf_time() + ec->shd.timeout;
- return ret;
-}
+ while (!healer->rerun) {
+ ret = pthread_cond_timedwait(&healer->cond, &healer->mutex, &wait_till);
+ if (ret == ETIMEDOUT)
+ break;
+ }
+ if (ec->shutdown) {
+ healer->running = _gf_false;
+ return -1;
+ }
-int
-ec_shd_healer_wait (struct subvol_healer *healer)
-{
- int ret = 0;
+ ret = healer->rerun;
+ healer->rerun = 0;
- pthread_mutex_lock (&healer->mutex);
- {
- ret = __ec_shd_healer_wait (healer);
- }
- pthread_mutex_unlock (&healer->mutex);
+ if (!ec->shd.enabled || !ec->up)
+ goto disabled_loop;
- return ret;
+ return ret;
}
-
-gf_boolean_t
-safe_break (struct subvol_healer *healer)
+int
+ec_shd_healer_wait(struct subvol_healer *healer)
{
- gf_boolean_t ret = _gf_false;
-
- pthread_mutex_lock (&healer->mutex);
- {
- if (healer->rerun)
- goto unlock;
+ int ret = 0;
- healer->running = _gf_false;
- ret = _gf_true;
- }
-unlock:
- pthread_mutex_unlock (&healer->mutex);
+ pthread_mutex_lock(&healer->mutex);
+ {
+ ret = __ec_shd_healer_wait(healer);
+ }
+ pthread_mutex_unlock(&healer->mutex);
- return ret;
+ return ret;
}
int
-ec_shd_index_inode (xlator_t *this, xlator_t *subvol, inode_t **inode)
+ec_shd_index_inode(xlator_t *this, xlator_t *subvol, inode_t **inode)
{
- loc_t rootloc = {0, };
- int ret = 0;
- dict_t *xattr = NULL;
- void *index_gfid = NULL;
-
- *inode = NULL;
- rootloc.inode = inode_ref (this->itable->root);
- gf_uuid_copy (rootloc.gfid, rootloc.inode->gfid);
-
- ret = syncop_getxattr (subvol, &rootloc, &xattr,
- GF_XATTROP_INDEX_GFID, NULL, NULL);
- if (ret < 0)
- goto out;
- if (!xattr) {
- ret = -EINVAL;
- goto out;
- }
-
- ret = dict_get_ptr (xattr, GF_XATTROP_INDEX_GFID, &index_gfid);
- if (ret)
- goto out;
-
- gf_msg_debug (this->name, 0, "index-dir gfid for %s: %s",
- subvol->name, uuid_utoa (index_gfid));
-
- ret = syncop_inode_find (this, subvol, index_gfid,
- inode, NULL, NULL);
+ loc_t rootloc = {
+ 0,
+ };
+ int ret = 0;
+ dict_t *xattr = NULL;
+ void *index_gfid = NULL;
+
+ *inode = NULL;
+ rootloc.inode = inode_ref(this->itable->root);
+ gf_uuid_copy(rootloc.gfid, rootloc.inode->gfid);
+
+ ret = syncop_getxattr(subvol, &rootloc, &xattr, GF_XATTROP_INDEX_GFID, NULL,
+ NULL);
+ if (ret < 0)
+ goto out;
+ if (!xattr) {
+ ret = -EINVAL;
+ goto out;
+ }
+
+ ret = dict_get_ptr(xattr, GF_XATTROP_INDEX_GFID, &index_gfid);
+ if (ret)
+ goto out;
+
+ gf_msg_debug(this->name, 0, "index-dir gfid for %s: %s", subvol->name,
+ uuid_utoa(index_gfid));
+
+ ret = syncop_inode_find(this, subvol, index_gfid, inode, NULL, NULL);
out:
- loc_wipe (&rootloc);
+ loc_wipe(&rootloc);
- if (xattr)
- dict_unref (xattr);
+ if (xattr)
+ dict_unref(xattr);
- return ret;
+ return ret;
}
int
-ec_shd_index_purge (xlator_t *subvol, inode_t *inode, char *name)
+ec_shd_index_purge(xlator_t *subvol, inode_t *inode, char *name)
{
- loc_t loc = {0, };
- int ret = 0;
+ loc_t loc = {
+ 0,
+ };
+ int ret = 0;
- loc.parent = inode_ref (inode);
- loc.name = name;
+ loc.parent = inode_ref(inode);
+ loc.name = name;
- ret = syncop_unlink (subvol, &loc, NULL, NULL);
+ ret = syncop_unlink(subvol, &loc, NULL, NULL);
- loc_wipe (&loc);
- return ret;
+ loc_wipe(&loc);
+ return ret;
}
-int
-ec_shd_selfheal (struct subvol_healer *healer, int child, loc_t *loc)
+static gf_boolean_t
+ec_is_heal_completed(char *status)
{
- int32_t ret;
-
- ret = syncop_getxattr (healer->this, loc, NULL, EC_XATTR_HEAL, NULL,
- NULL);
- if ((ret >= 0) && (loc->inode->ia_type == IA_IFDIR)) {
- /* If we have just healed a directory, it's possible that
- * other index entries have appeared to be healed. We put a
- * mark so that we can check it later and restart a scan
- * without delay. */
- healer->rerun = _gf_true;
- }
-
- return ret;
+ char *bad_pos = NULL;
+ char *zero_pos = NULL;
+
+ if (!status) {
+ return _gf_false;
+ }
+
+ /*Logic:
+ * Status will be of the form Good: <binary>, Bad: <binary>
+ * If heal completes, if we do strchr for '0' it should be present after
+ * 'Bad:' i.e. strRchr for ':'
+ * */
+
+ zero_pos = strchr(status, '0');
+ bad_pos = strrchr(status, ':');
+ if (!zero_pos || !bad_pos) {
+ /*malformed status*/
+ return _gf_false;
+ }
+
+ if (zero_pos > bad_pos) {
+ return _gf_true;
+ }
+
+ return _gf_false;
}
-
int
-ec_shd_index_heal (xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
- void *data)
+ec_shd_selfheal(struct subvol_healer *healer, int child, loc_t *loc,
+ gf_boolean_t full)
{
- struct subvol_healer *healer = data;
- ec_t *ec = NULL;
- loc_t loc = {0};
- int ret = 0;
-
- ec = healer->this->private;
- if (ec->xl_up_count <= ec->fragments) {
- return -ENOTCONN;
+ dict_t *xdata = NULL;
+ dict_t *dict = NULL;
+ uint32_t count;
+ int32_t ret;
+ char *heal_status = NULL;
+ ec_t *ec = healer->this->private;
+
+ GF_ATOMIC_INC(ec->stats.shd.attempted);
+ ret = syncop_getxattr(healer->this, loc, &dict, EC_XATTR_HEAL, NULL,
+ &xdata);
+ if (ret == 0) {
+ if (dict && (dict_get_str(dict, EC_XATTR_HEAL, &heal_status) == 0)) {
+ if (ec_is_heal_completed(heal_status)) {
+ GF_ATOMIC_INC(ec->stats.shd.completed);
+ }
}
- if (!ec->shd.enabled)
- return -EBUSY;
+ }
+
+ if (!full && (loc->inode->ia_type == IA_IFDIR)) {
+ /* If we have just healed a directory, it's possible that
+ * other index entries have appeared to be healed. */
+ if ((xdata != NULL) &&
+ (dict_get_uint32(xdata, EC_XATTR_HEAL_NEW, &count) == 0) &&
+ (count > 0)) {
+ /* Force a rerun of the index healer. */
+ gf_msg_debug(healer->this->name, 0, "%d more entries to heal",
+ count);
+
+ healer->rerun = _gf_true;
+ }
+ }
- gf_msg_debug (healer->this->name, 0, "got entry: %s",
- entry->d_name);
+ if (xdata != NULL) {
+ dict_unref(xdata);
+ }
- ret = gf_uuid_parse (entry->d_name, loc.gfid);
- if (ret)
- return 0;
+ if (dict) {
+ dict_unref(dict);
+ }
- /* If this fails with ENOENT/ESTALE index is stale */
- ret = syncop_gfid_to_path (healer->this->itable, subvol, loc.gfid,
- (char **)&loc.path);
- if (ret < 0)
- goto out;
+ return ret;
+}
- ret = syncop_inode_find (healer->this, healer->this, loc.gfid,
- &loc.inode, NULL, NULL);
- if (ret < 0)
- goto out;
+int
+ec_shd_index_heal(xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
+ void *data)
+{
+ struct subvol_healer *healer = data;
+ ec_t *ec = NULL;
+ loc_t loc = {0};
+ int ret = 0;
+
+ ec = healer->this->private;
+ if (ec->xl_up_count <= ec->fragments) {
+ return -ENOTCONN;
+ }
+ if (!ec->shd.enabled)
+ return -EBUSY;
+
+ gf_msg_debug(healer->this->name, 0, "got entry: %s", entry->d_name);
+
+ ret = gf_uuid_parse(entry->d_name, loc.gfid);
+ if (ret)
+ return 0;
- ec_shd_selfheal (healer, healer->subvol, &loc);
-out:
- if (ret == -ENOENT || ret == -ESTALE) {
- gf_msg (healer->this->name, GF_LOG_DEBUG, 0,
- EC_MSG_HEAL_FAIL, "Purging index for gfid %s:",
- uuid_utoa(loc.gfid));
- ec_shd_index_purge (subvol, parent->inode, entry->d_name);
- }
- loc_wipe (&loc);
+ /* If this fails with ENOENT/ESTALE index is stale */
+ ret = syncop_gfid_to_path(healer->this->itable, subvol, loc.gfid,
+ (char **)&loc.path);
+ if (ret < 0)
+ goto out;
- return 0;
+ ret = syncop_inode_find(healer->this, healer->this, loc.gfid, &loc.inode,
+ NULL, NULL);
+ if (ret < 0)
+ goto out;
+
+ ec_shd_selfheal(healer, healer->subvol, &loc, _gf_false);
+out:
+ if (ret == -ENOENT || ret == -ESTALE) {
+ gf_msg(healer->this->name, GF_LOG_DEBUG, 0, EC_MSG_HEAL_FAIL,
+ "Purging index for gfid %s:", uuid_utoa(loc.gfid));
+ ec_shd_index_purge(subvol, parent->inode, entry->d_name);
+ }
+ loc_wipe(&loc);
+
+ return 0;
}
int
-ec_shd_index_sweep (struct subvol_healer *healer)
+ec_shd_index_sweep(struct subvol_healer *healer)
{
- loc_t loc = {0};
- ec_t *ec = NULL;
- int ret = 0;
- xlator_t *subvol = NULL;
- dict_t *xdata = NULL;
-
- ec = healer->this->private;
- subvol = ec->xl_list[healer->subvol];
-
- ret = ec_shd_index_inode (healer->this, subvol, &loc.inode);
- if (ret < 0) {
- gf_msg (healer->this->name, GF_LOG_WARNING, errno,
- EC_MSG_INDEX_DIR_GET_FAIL,
- "unable to get index-dir on %s", subvol->name);
- goto out;
- }
-
- xdata = dict_new ();
- if (!xdata || dict_set_int32 (xdata, "get-gfid-type", 1)) {
- ret = -ENOMEM;
- goto out;
- }
-
- ret = syncop_mt_dir_scan (NULL, subvol, &loc, GF_CLIENT_PID_SELF_HEALD,
- healer, ec_shd_index_heal, xdata,
- ec->shd.max_threads, ec->shd.wait_qlength);
+ loc_t loc = {0};
+ ec_t *ec = NULL;
+ int ret = 0;
+ xlator_t *subvol = NULL;
+ dict_t *xdata = NULL;
+
+ ec = healer->this->private;
+ subvol = ec->xl_list[healer->subvol];
+
+ ret = ec_shd_index_inode(healer->this, subvol, &loc.inode);
+ if (ret < 0) {
+ gf_msg(healer->this->name, GF_LOG_WARNING, errno,
+ EC_MSG_INDEX_DIR_GET_FAIL, "unable to get index-dir on %s",
+ subvol->name);
+ goto out;
+ }
+
+ xdata = dict_new();
+ if (!xdata || dict_set_int32(xdata, "get-gfid-type", 1)) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ _mask_cancellation();
+ ret = syncop_mt_dir_scan(NULL, subvol, &loc, GF_CLIENT_PID_SELF_HEALD,
+ healer, ec_shd_index_heal, xdata,
+ ec->shd.max_threads, ec->shd.wait_qlength);
+ _unmask_cancellation();
out:
- if (xdata)
- dict_unref (xdata);
- loc_wipe (&loc);
+ if (xdata)
+ dict_unref(xdata);
+ loc_wipe(&loc);
- return ret;
+ return ret;
}
int
-ec_shd_full_heal (xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
- void *data)
+ec_shd_full_heal(xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
+ void *data)
{
- struct subvol_healer *healer = data;
- xlator_t *this = healer->this;
- ec_t *ec = NULL;
- loc_t loc = {0};
- int ret = 0;
-
- ec = this->private;
- if (ec->xl_up_count <= ec->fragments) {
- return -ENOTCONN;
- }
- if (!ec->shd.enabled)
- return -EBUSY;
-
- if (gf_uuid_is_null(entry->d_stat.ia_gfid)) {
- /* It's possible that an entry has been removed just after
- * being seen in a directory but before getting its stat info.
- * In this case we'll receive a NULL gfid here. Since the file
- * doesn't exist anymore, we can safely ignore it. */
- return 0;
- }
+ struct subvol_healer *healer = data;
+ xlator_t *this = healer->this;
+ ec_t *ec = NULL;
+ loc_t loc = {0};
+ int ret = 0;
+
+ ec = this->private;
+
+ if (this->cleanup_starting) {
+ return -ENOTCONN;
+ }
+
+ if (ec->xl_up_count <= ec->fragments) {
+ return -ENOTCONN;
+ }
+ if (!ec->shd.enabled)
+ return -EBUSY;
+
+ if (gf_uuid_is_null(entry->d_stat.ia_gfid)) {
+ /* It's possible that an entry has been removed just after
+ * being seen in a directory but before getting its stat info.
+ * In this case we'll receive a NULL gfid here. Since the file
+ * doesn't exist anymore, we can safely ignore it. */
+ return 0;
+ }
- loc.parent = inode_ref (parent->inode);
- loc.name = entry->d_name;
- gf_uuid_copy (loc.gfid, entry->d_stat.ia_gfid);
+ loc.parent = inode_ref(parent->inode);
+ loc.name = entry->d_name;
+ gf_uuid_copy(loc.gfid, entry->d_stat.ia_gfid);
- /* If this fails with ENOENT/ESTALE index is stale */
- ret = syncop_gfid_to_path (this->itable, subvol, loc.gfid,
- (char **)&loc.path);
- if (ret < 0)
- goto out;
+ /* If this fails with ENOENT/ESTALE index is stale */
+ ret = syncop_gfid_to_path(this->itable, subvol, loc.gfid,
+ (char **)&loc.path);
+ if (ret < 0)
+ goto out;
- ret = syncop_inode_find (this, this, loc.gfid,
- &loc.inode, NULL, NULL);
- if (ret < 0)
- goto out;
+ ret = syncop_inode_find(this, this, loc.gfid, &loc.inode, NULL, NULL);
+ if (ret < 0)
+ goto out;
- ec_shd_selfheal (healer, healer->subvol, &loc);
+ ec_shd_selfheal(healer, healer->subvol, &loc, _gf_true);
- ret = 0;
+ ret = 0;
out:
- loc_wipe (&loc);
- return ret;
+ loc_wipe(&loc);
+ return ret;
}
int
-ec_shd_full_sweep (struct subvol_healer *healer, inode_t *inode)
+ec_shd_full_sweep(struct subvol_healer *healer, inode_t *inode)
{
- ec_t *ec = NULL;
- loc_t loc = {0};
-
- ec = healer->this->private;
- loc.inode = inode;
- return syncop_ftw (ec->xl_list[healer->subvol], &loc,
- GF_CLIENT_PID_SELF_HEALD, healer,
- ec_shd_full_heal);
+ ec_t *ec = NULL;
+ loc_t loc = {0};
+ int ret = -1;
+
+ ec = healer->this->private;
+ loc.inode = inode;
+ _mask_cancellation();
+ ret = syncop_ftw(ec->xl_list[healer->subvol], &loc,
+ GF_CLIENT_PID_SELF_HEALD, healer, ec_shd_full_heal);
+ _unmask_cancellation();
+ return ret;
}
-
void *
-ec_shd_index_healer (void *data)
+ec_shd_index_healer(void *data)
{
- struct subvol_healer *healer = NULL;
- xlator_t *this = NULL;
-
- healer = data;
- THIS = this = healer->this;
- ec_t *ec = this->private;
-
- for (;;) {
- ec_shd_healer_wait (healer);
-
- ASSERT_LOCAL(this, healer);
-
-
- if (ec->xl_up_count > ec->fragments) {
- gf_msg_debug (this->name, 0,
- "starting index sweep on subvol %s",
- ec_subvol_name (this, healer->subvol));
- ec_shd_index_sweep (healer);
- }
- gf_msg_debug (this->name, 0,
- "finished index sweep on subvol %s",
- ec_subvol_name (this, healer->subvol));
+ struct subvol_healer *healer = NULL;
+ xlator_t *this = NULL;
+ int run = 0;
+
+ healer = data;
+ THIS = this = healer->this;
+ ec_t *ec = this->private;
+
+ for (;;) {
+ run = ec_shd_healer_wait(healer);
+ if (run == -1)
+ break;
+
+ if (ec->xl_up_count > ec->fragments) {
+ gf_msg_debug(this->name, 0, "starting index sweep on subvol %s",
+ ec_subvol_name(this, healer->subvol));
+ ec_shd_index_sweep(healer);
}
+ gf_msg_debug(this->name, 0, "finished index sweep on subvol %s",
+ ec_subvol_name(this, healer->subvol));
+ }
- return NULL;
+ return NULL;
}
-
void *
-ec_shd_full_healer (void *data)
+ec_shd_full_healer(void *data)
{
- struct subvol_healer *healer = NULL;
- xlator_t *this = NULL;
- loc_t rootloc = {0};
-
- int run = 0;
-
- healer = data;
- THIS = this = healer->this;
- ec_t *ec = this->private;
-
- rootloc.inode = this->itable->root;
- for (;;) {
- pthread_mutex_lock (&healer->mutex);
- {
- run = __ec_shd_healer_wait (healer);
- if (!run)
- healer->running = _gf_false;
- }
- pthread_mutex_unlock (&healer->mutex);
-
- if (!run)
- break;
-
- ASSERT_LOCAL(this, healer);
-
-
- if (ec->xl_up_count > ec->fragments) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- EC_MSG_FULL_SWEEP_START,
- "starting full sweep on subvol %s",
- ec_subvol_name (this, healer->subvol));
-
- ec_shd_selfheal (healer, healer->subvol, &rootloc);
- ec_shd_full_sweep (healer, this->itable->root);
- }
-
- gf_msg (this->name, GF_LOG_INFO, 0,
- EC_MSG_FULL_SWEEP_STOP,
- "finished full sweep on subvol %s",
- ec_subvol_name (this, healer->subvol));
+ struct subvol_healer *healer = NULL;
+ xlator_t *this = NULL;
+ loc_t rootloc = {0};
+
+ int run = 0;
+
+ healer = data;
+ THIS = this = healer->this;
+ ec_t *ec = this->private;
+
+ rootloc.inode = this->itable->root;
+ for (;;) {
+ run = ec_shd_healer_wait(healer);
+ if (run < 0) {
+ break;
+ } else if (run == 0) {
+ continue;
}
- return NULL;
-}
+ if (ec->xl_up_count > ec->fragments) {
+ gf_msg(this->name, GF_LOG_INFO, 0, EC_MSG_FULL_SWEEP_START,
+ "starting full sweep on subvol %s",
+ ec_subvol_name(this, healer->subvol));
+ ec_shd_selfheal(healer, healer->subvol, &rootloc, _gf_true);
+ ec_shd_full_sweep(healer, this->itable->root);
+ }
+
+ gf_msg(this->name, GF_LOG_INFO, 0, EC_MSG_FULL_SWEEP_STOP,
+ "finished full sweep on subvol %s",
+ ec_subvol_name(this, healer->subvol));
+ }
+
+ return NULL;
+}
int
-ec_shd_healer_init (xlator_t *this, struct subvol_healer *healer)
+ec_shd_healer_init(xlator_t *this, struct subvol_healer *healer)
{
- int ret = 0;
+ int ret = 0;
- ret = pthread_mutex_init (&healer->mutex, NULL);
- if (ret)
- goto out;
+ ret = pthread_mutex_init(&healer->mutex, NULL);
+ if (ret)
+ goto out;
- ret = pthread_cond_init (&healer->cond, NULL);
- if (ret)
- goto out;
+ ret = pthread_cond_init(&healer->cond, NULL);
+ if (ret)
+ goto out;
- healer->this = this;
- healer->running = _gf_false;
- healer->rerun = _gf_false;
- healer->local = _gf_false;
+ healer->this = this;
+ healer->running = _gf_false;
+ healer->rerun = _gf_false;
out:
- return ret;
+ return ret;
}
-
int
-ec_shd_healer_spawn (xlator_t *this, struct subvol_healer *healer,
- void *(threadfn)(void *))
+ec_shd_healer_spawn(xlator_t *this, struct subvol_healer *healer,
+ void *(threadfn)(void *))
{
- int ret = 0;
-
- pthread_mutex_lock (&healer->mutex);
- {
- if (healer->running) {
- pthread_cond_signal (&healer->cond);
- } else {
- ret = gf_thread_create (&healer->thread, NULL,
- threadfn, healer, "ecshd");
- if (ret)
- goto unlock;
- healer->running = 1;
- }
-
- healer->rerun = 1;
+ int ret = 0;
+
+ pthread_mutex_lock(&healer->mutex);
+ {
+ if (healer->running) {
+ pthread_cond_signal(&healer->cond);
+ } else {
+ ret = gf_thread_create(&healer->thread, NULL, threadfn, healer,
+ "ecshd");
+ if (ret)
+ goto unlock;
+ healer->running = 1;
}
+
+ healer->rerun = 1;
+ }
unlock:
- pthread_mutex_unlock (&healer->mutex);
+ pthread_mutex_unlock(&healer->mutex);
- return ret;
+ return ret;
}
int
-ec_shd_full_healer_spawn (xlator_t *this, int subvol)
+ec_shd_full_healer_spawn(xlator_t *this, int subvol)
{
- return ec_shd_healer_spawn (this, NTH_FULL_HEALER (this, subvol),
- ec_shd_full_healer);
-}
+ if (xlator_is_cleanup_starting(this))
+ return -1;
+ return ec_shd_healer_spawn(this, NTH_FULL_HEALER(this, subvol),
+ ec_shd_full_healer);
+}
int
-ec_shd_index_healer_spawn (xlator_t *this, int subvol)
+ec_shd_index_healer_spawn(xlator_t *this, int subvol)
{
- return ec_shd_healer_spawn (this, NTH_INDEX_HEALER (this, subvol),
- ec_shd_index_healer);
+ if (xlator_is_cleanup_starting(this))
+ return -1;
+
+ return ec_shd_healer_spawn(this, NTH_INDEX_HEALER(this, subvol),
+ ec_shd_index_healer);
}
void
ec_shd_index_healer_wake(ec_t *ec)
{
- int32_t i;
+ int32_t i;
- for (i = 0; i < ec->nodes; i++) {
- if (((ec->xl_up >> i) & 1) != 0) {
- ec_shd_index_healer_spawn(ec->xl, i);
- }
+ for (i = 0; i < ec->nodes; i++) {
+ if (((ec->xl_up >> i) & 1) != 0) {
+ ec_shd_index_healer_spawn(ec->xl, i);
}
+ }
}
int
-ec_selfheal_daemon_init (xlator_t *this)
+ec_selfheal_daemon_init(xlator_t *this)
{
- ec_t *ec = NULL;
- ec_self_heald_t *shd = NULL;
- int ret = -1;
- int i = 0;
-
- ec = this->private;
- shd = &ec->shd;
-
- shd->index_healers = GF_CALLOC (sizeof(*shd->index_healers),
- ec->nodes,
- ec_mt_subvol_healer_t);
- if (!shd->index_healers)
- goto out;
-
- for (i = 0; i < ec->nodes; i++) {
- shd->index_healers[i].subvol = i;
- ret = ec_shd_healer_init (this, &shd->index_healers[i]);
- if (ret)
- goto out;
- }
+ ec_t *ec = NULL;
+ ec_self_heald_t *shd = NULL;
+ int ret = -1;
+ int i = 0;
+
+ ec = this->private;
+ shd = &ec->shd;
+
+ shd->index_healers = GF_CALLOC(sizeof(*shd->index_healers), ec->nodes,
+ ec_mt_subvol_healer_t);
+ if (!shd->index_healers)
+ goto out;
+
+ for (i = 0; i < ec->nodes; i++) {
+ shd->index_healers[i].subvol = i;
+ ret = ec_shd_healer_init(this, &shd->index_healers[i]);
+ if (ret)
+ goto out;
+ }
- shd->full_healers = GF_CALLOC (sizeof(*shd->full_healers),
- ec->nodes,
- ec_mt_subvol_healer_t);
- if (!shd->full_healers)
- goto out;
-
- for (i = 0; i < ec->nodes; i++) {
- shd->full_healers[i].subvol = i;
- ret = ec_shd_healer_init (this, &shd->full_healers[i]);
- if (ret)
- goto out;
- }
+ shd->full_healers = GF_CALLOC(sizeof(*shd->full_healers), ec->nodes,
+ ec_mt_subvol_healer_t);
+ if (!shd->full_healers)
+ goto out;
+
+ for (i = 0; i < ec->nodes; i++) {
+ shd->full_healers[i].subvol = i;
+ ret = ec_shd_healer_init(this, &shd->full_healers[i]);
+ if (ret)
+ goto out;
+ }
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
-
int
-ec_heal_op (xlator_t *this, dict_t *output, gf_xl_afr_op_t op, int xl_id)
+ec_heal_op(xlator_t *this, dict_t *output, gf_xl_afr_op_t op, int xl_id)
{
- char key[64] = {0};
- int op_ret = 0;
- ec_t *ec = NULL;
- int i = 0;
- GF_UNUSED int ret = 0;
-
- ec = this->private;
-
- op_ret = -1;
- for (i = 0; i < ec->nodes; i++) {
- snprintf (key, sizeof (key), "%d-%d-status", xl_id, i);
-
- if (((ec->xl_up >> i) & 1) == 0) {
- ret = dict_set_str (output, key, "Brick is not connected");
- } else if (!ec->up) {
- ret = dict_set_str (output, key,
- "Disperse subvolume is not up");
- } else if (!ec_shd_is_subvol_local (this, i)) {
- ret = dict_set_str (output, key, "Brick is remote");
- } else {
- ret = dict_set_str (output, key, "Started self-heal");
- if (op == GF_SHD_OP_HEAL_FULL) {
- ec_shd_full_healer_spawn (this, i);
- } else if (op == GF_SHD_OP_HEAL_INDEX) {
- ec_shd_index_healer_spawn (this, i);
- }
- op_ret = 0;
- }
+ char key[64] = {0};
+ int op_ret = 0;
+ ec_t *ec = NULL;
+ int i = 0;
+ GF_UNUSED int ret = 0;
+
+ ec = this->private;
+
+ op_ret = -1;
+ for (i = 0; i < ec->nodes; i++) {
+ snprintf(key, sizeof(key), "%d-%d-status", xl_id, i);
+
+ if (((ec->xl_up >> i) & 1) == 0) {
+ ret = dict_set_str(output, key, "Brick is not connected");
+ } else if (!ec->up) {
+ ret = dict_set_str(output, key, "Disperse subvolume is not up");
+ } else if (!ec_shd_is_subvol_local(this, i)) {
+ ret = dict_set_str(output, key, "Brick is remote");
+ } else {
+ ret = dict_set_str(output, key, "Started self-heal");
+ if (op == GF_SHD_OP_HEAL_FULL) {
+ ec_shd_full_healer_spawn(this, i);
+ } else if (op == GF_SHD_OP_HEAL_INDEX) {
+ ec_shd_index_healer_spawn(this, i);
+ }
+ op_ret = 0;
}
- return op_ret;
+ }
+ return op_ret;
}
int
-ec_xl_op (xlator_t *this, dict_t *input, dict_t *output)
+ec_xl_op(xlator_t *this, dict_t *input, dict_t *output)
{
- gf_xl_afr_op_t op = GF_SHD_OP_INVALID;
- int ret = 0;
- int xl_id = 0;
+ gf_xl_afr_op_t op = GF_SHD_OP_INVALID;
+ int ret = 0;
+ int xl_id = 0;
- ret = dict_get_int32 (input, "xl-op", (int32_t *)&op);
- if (ret)
- goto out;
+ ret = dict_get_int32(input, "xl-op", (int32_t *)&op);
+ if (ret)
+ goto out;
- ret = dict_get_int32 (input, this->name, &xl_id);
- if (ret)
- goto out;
+ ret = dict_get_int32(input, this->name, &xl_id);
+ if (ret)
+ goto out;
- ret = dict_set_int32 (output, this->name, xl_id);
- if (ret)
- goto out;
+ ret = dict_set_int32(output, this->name, xl_id);
+ if (ret)
+ goto out;
- switch (op) {
+ switch (op) {
case GF_SHD_OP_HEAL_FULL:
- ret = ec_heal_op (this, output, op, xl_id);
- break;
+ ret = ec_heal_op(this, output, op, xl_id);
+ break;
case GF_SHD_OP_HEAL_INDEX:
- ret = ec_heal_op (this, output, op, xl_id);
- break;
+ ret = ec_heal_op(this, output, op, xl_id);
+ break;
default:
- ret = -1;
- break;
- }
+ ret = -1;
+ break;
+ }
out:
- dict_del (output, this->name);
- return ret;
+ dict_del(output, this->name);
+ return ret;
+}
+
+void
+ec_destroy_healer_object(xlator_t *this, struct subvol_healer *healer)
+{
+ if (!healer)
+ return;
+
+ pthread_cond_destroy(&healer->cond);
+ pthread_mutex_destroy(&healer->mutex);
+}
+
+void
+ec_selfheal_daemon_fini(xlator_t *this)
+{
+ struct subvol_healer *healer = NULL;
+ ec_self_heald_t *shd = NULL;
+ ec_t *priv = NULL;
+ int i = 0;
+
+ priv = this->private;
+ if (!priv)
+ return;
+
+ shd = &priv->shd;
+ if (!shd->iamshd)
+ return;
+
+ for (i = 0; i < priv->nodes; i++) {
+ healer = &shd->index_healers[i];
+ ec_destroy_healer_object(this, healer);
+
+ healer = &shd->full_healers[i];
+ ec_destroy_healer_object(this, healer);
+ }
+
+ GF_FREE(shd->index_healers);
+ GF_FREE(shd->full_healers);
}
diff --git a/xlators/cluster/ec/src/ec-heald.h b/xlators/cluster/ec/src/ec-heald.h
index 2a8488124c4..6c7da4edc10 100644
--- a/xlators/cluster/ec/src/ec-heald.h
+++ b/xlators/cluster/ec/src/ec-heald.h
@@ -11,17 +11,20 @@
#ifndef __EC_HEALD_H__
#define __EC_HEALD_H__
-#include "xlator.h"
-
-#include "ec-types.h"
+#include "ec-types.h" // for ec_t
+#include "glusterfs/dict.h" // for dict_t
+#include "glusterfs/globals.h" // for xlator_t
int
-ec_xl_op (xlator_t *this, dict_t *input, dict_t *output);
+ec_xl_op(xlator_t *this, dict_t *input, dict_t *output);
int
-ec_selfheal_daemon_init (xlator_t *this);
+ec_selfheal_daemon_init(xlator_t *this);
void
ec_shd_index_healer_wake(ec_t *ec);
+void
+ec_selfheal_daemon_fini(xlator_t *this);
+
#endif /* __EC_HEALD_H__ */
diff --git a/xlators/cluster/ec/src/ec-helpers.c b/xlators/cluster/ec/src/ec-helpers.c
index 83f96ba6cb2..48f54475e01 100644
--- a/xlators/cluster/ec/src/ec-helpers.c
+++ b/xlators/cluster/ec/src/ec-helpers.c
@@ -10,7 +10,7 @@
#include <libgen.h>
-#include "byte-order.h"
+#include <glusterfs/byte-order.h>
#include "ec.h"
#include "ec-mem-types.h"
@@ -19,25 +19,20 @@
#include "ec-method.h"
#include "ec-helpers.h"
-static const char * ec_fop_list[] =
-{
- [-EC_FOP_HEAL] = "HEAL"
-};
+static const char *ec_fop_list[] = {[-EC_FOP_HEAL] = "HEAL"};
-const char * ec_bin(char * str, size_t size, uint64_t value, int32_t digits)
+const char *
+ec_bin(char *str, size_t size, uint64_t value, int32_t digits)
{
str += size;
- if (size-- < 1)
- {
+ if (size-- < 1) {
goto failed;
}
*--str = 0;
- while ((value != 0) || (digits > 0))
- {
- if (size-- < 1)
- {
+ while ((value != 0) || (digits > 0)) {
+ if (size-- < 1) {
goto failed;
}
*--str = '0' + (value & 1);
@@ -51,21 +46,22 @@ failed:
return "<buffer too small>";
}
-const char * ec_fop_name(int32_t id)
+const char *
+ec_fop_name(int32_t id)
{
- if (id >= 0)
- {
+ if (id >= 0) {
return gf_fop_list[id];
}
return ec_fop_list[-id];
}
-void ec_trace(const char * event, ec_fop_data_t * fop, const char * fmt, ...)
+void
+ec_trace(const char *event, ec_fop_data_t *fop, const char *fmt, ...)
{
char str1[32], str2[32], str3[32];
- char * msg;
- ec_t * ec = fop->xl->private;
+ char *msg;
+ ec_t *ec = fop->xl->private;
va_list args;
int32_t ret;
@@ -73,28 +69,28 @@ void ec_trace(const char * event, ec_fop_data_t * fop, const char * fmt, ...)
ret = vasprintf(&msg, fmt, args);
va_end(args);
- if (ret < 0)
- {
+ if (ret < 0) {
msg = "<memory allocation error>";
}
- gf_msg_trace ("ec", 0, "%s(%s) %p(%p) [refs=%d, winds=%d, jobs=%d] "
- "frame=%p/%p, min/exp=%d/%d, err=%d state=%d "
- "{%s:%s:%s} %s",
- event, ec_fop_name(fop->id), fop, fop->parent, fop->refs,
- fop->winds, fop->jobs, fop->req_frame, fop->frame, fop->minimum,
- fop->expected, fop->error, fop->state,
- ec_bin(str1, sizeof(str1), fop->mask, ec->nodes),
- ec_bin(str2, sizeof(str2), fop->remaining, ec->nodes),
- ec_bin(str3, sizeof(str3), fop->good, ec->nodes), msg);
+ gf_msg_trace("ec", 0,
+ "%s(%s) %p(%p) [refs=%d, winds=%d, jobs=%d] "
+ "frame=%p/%p, min/exp=%d/%d, err=%d state=%d "
+ "{%s:%s:%s} %s",
+ event, ec_fop_name(fop->id), fop, fop->parent, fop->refs,
+ fop->winds, fop->jobs, fop->req_frame, fop->frame,
+ fop->minimum, fop->expected, fop->error, fop->state,
+ ec_bin(str1, sizeof(str1), fop->mask, ec->nodes),
+ ec_bin(str2, sizeof(str2), fop->remaining, ec->nodes),
+ ec_bin(str3, sizeof(str3), fop->good, ec->nodes), msg);
- if (ret >= 0)
- {
+ if (ret >= 0) {
free(msg);
}
}
-int32_t ec_bits_consume(uint64_t * n)
+int32_t
+ec_bits_consume(uint64_t *n)
{
uint64_t tmp;
@@ -105,21 +101,18 @@ int32_t ec_bits_consume(uint64_t * n)
return gf_bits_index(tmp);
}
-size_t ec_iov_copy_to(void * dst, struct iovec * vector, int32_t count,
- off_t offset, size_t size)
+size_t
+ec_iov_copy_to(void *dst, struct iovec *vector, int32_t count, off_t offset,
+ size_t size)
{
int32_t i = 0;
size_t total = 0, len = 0;
- while (i < count)
- {
- if (offset < vector[i].iov_len)
- {
- while ((i < count) && (size > 0))
- {
+ while (i < count) {
+ if (offset < vector[i].iov_len) {
+ while ((i < count) && (size > 0)) {
len = size;
- if (len > vector[i].iov_len - offset)
- {
+ if (len > vector[i].iov_len - offset) {
len = vector[i].iov_len - offset;
}
memcpy(dst, vector[i++].iov_base + offset, len);
@@ -139,15 +132,15 @@ size_t ec_iov_copy_to(void * dst, struct iovec * vector, int32_t count,
return total;
}
-int32_t ec_buffer_alloc(xlator_t *xl, size_t size, struct iobref **piobref,
- void **ptr)
+int32_t
+ec_buffer_alloc(xlator_t *xl, size_t size, struct iobref **piobref, void **ptr)
{
struct iobref *iobref = NULL;
struct iobuf *iobuf = NULL;
int32_t ret = -ENOMEM;
- iobuf = iobuf_get_page_aligned (xl->ctx->iobuf_pool, size,
- EC_METHOD_WORD_SIZE);
+ iobuf = iobuf_get_page_aligned(xl->ctx->iobuf_pool, size,
+ EC_METHOD_WORD_SIZE);
if (iobuf == NULL) {
goto out;
}
@@ -186,12 +179,12 @@ out:
return ret;
}
-int32_t ec_dict_set_array(dict_t *dict, char *key, uint64_t value[],
- int32_t size)
+int32_t
+ec_dict_set_array(dict_t *dict, char *key, uint64_t value[], int32_t size)
{
- int ret = -1;
- uint64_t *ptr = NULL;
- int32_t vindex;
+ int ret = -1;
+ uint64_t *ptr = NULL;
+ int32_t vindex;
if (value == NULL) {
return -EINVAL;
@@ -202,19 +195,18 @@ int32_t ec_dict_set_array(dict_t *dict, char *key, uint64_t value[],
return -ENOMEM;
}
for (vindex = 0; vindex < size; vindex++) {
- ptr[vindex] = hton64(value[vindex]);
+ ptr[vindex] = hton64(value[vindex]);
}
ret = dict_set_bin(dict, key, ptr, sizeof(uint64_t) * size);
if (ret)
- GF_FREE (ptr);
+ GF_FREE(ptr);
return ret;
}
-
int32_t
-ec_dict_get_array (dict_t *dict, char *key, uint64_t value[], int32_t size)
+ec_dict_get_array(dict_t *dict, char *key, uint64_t value[], int32_t size)
{
- void *ptr;
+ void *ptr;
int32_t len;
int32_t vindex;
int32_t old_size = 0;
@@ -228,44 +220,43 @@ ec_dict_get_array (dict_t *dict, char *key, uint64_t value[], int32_t size)
return err;
}
- if (len > (size * sizeof(uint64_t)) || (len % sizeof (uint64_t))) {
+ if (len > (size * sizeof(uint64_t)) || (len % sizeof(uint64_t))) {
return -EINVAL;
}
- memset (value, 0, size * sizeof(uint64_t));
/* 3.6 version ec would have stored version in 64 bit. In that case treat
* metadata versions same as data*/
- old_size = min (size, len/sizeof(uint64_t));
+ old_size = min(size, len / sizeof(uint64_t));
for (vindex = 0; vindex < old_size; vindex++) {
- value[vindex] = ntoh64(*((uint64_t *)ptr + vindex));
+ value[vindex] = ntoh64(*((uint64_t *)ptr + vindex));
}
if (old_size < size) {
- for (vindex = old_size; vindex < size; vindex++) {
- value[vindex] = value[old_size-1];
- }
+ for (vindex = old_size; vindex < size; vindex++) {
+ value[vindex] = value[old_size - 1];
+ }
}
return 0;
}
int32_t
-ec_dict_del_array (dict_t *dict, char *key, uint64_t value[], int32_t size)
+ec_dict_del_array(dict_t *dict, char *key, uint64_t value[], int32_t size)
{
int ret = 0;
- ret = ec_dict_get_array (dict, key, value, size);
+ ret = ec_dict_get_array(dict, key, value, size);
if (ret == 0)
- dict_del(dict, key);
+ dict_del(dict, key);
return ret;
}
-
-int32_t ec_dict_set_number(dict_t * dict, char * key, uint64_t value)
+int32_t
+ec_dict_set_number(dict_t *dict, char *key, uint64_t value)
{
- int ret = -1;
- uint64_t * ptr;
+ int ret = -1;
+ uint64_t *ptr;
ptr = GF_MALLOC(sizeof(value), gf_common_mt_char);
if (ptr == NULL) {
@@ -276,14 +267,15 @@ int32_t ec_dict_set_number(dict_t * dict, char * key, uint64_t value)
ret = dict_set_bin(dict, key, ptr, sizeof(value));
if (ret)
- GF_FREE (ptr);
+ GF_FREE(ptr);
return ret;
}
-int32_t ec_dict_del_number(dict_t * dict, char * key, uint64_t * value)
+int32_t
+ec_dict_del_number(dict_t *dict, char *key, uint64_t *value)
{
- void * ptr;
+ void *ptr;
int32_t len, err;
if (dict == NULL) {
@@ -304,24 +296,23 @@ int32_t ec_dict_del_number(dict_t * dict, char * key, uint64_t * value)
return 0;
}
-int32_t ec_dict_set_config(dict_t * dict, char * key, ec_config_t * config)
+int32_t
+ec_dict_set_config(dict_t *dict, char *key, ec_config_t *config)
{
int ret = -1;
- uint64_t * ptr, data;
+ uint64_t *ptr, data;
- if (config->version > EC_CONFIG_VERSION)
- {
- gf_msg ("ec", GF_LOG_ERROR, EINVAL,
- EC_MSG_UNSUPPORTED_VERSION,
- "Trying to store an unsupported config "
- "version (%u)", config->version);
+ if (config->version > EC_CONFIG_VERSION) {
+ gf_msg("ec", GF_LOG_ERROR, EINVAL, EC_MSG_UNSUPPORTED_VERSION,
+ "Trying to store an unsupported config "
+ "version (%u)",
+ config->version);
return -EINVAL;
}
ptr = GF_MALLOC(sizeof(uint64_t), gf_common_mt_char);
- if (ptr == NULL)
- {
+ if (ptr == NULL) {
return -ENOMEM;
}
@@ -336,14 +327,15 @@ int32_t ec_dict_set_config(dict_t * dict, char * key, ec_config_t * config)
ret = dict_set_bin(dict, key, ptr, sizeof(uint64_t));
if (ret)
- GF_FREE (ptr);
+ GF_FREE(ptr);
return ret;
}
-int32_t ec_dict_del_config(dict_t * dict, char * key, ec_config_t * config)
+int32_t
+ec_dict_del_config(dict_t *dict, char *key, ec_config_t *config)
{
- void * ptr;
+ void *ptr;
uint64_t data;
int32_t len, err;
@@ -373,12 +365,9 @@ int32_t ec_dict_del_config(dict_t * dict, char * key, ec_config_t * config)
}
config->version = (data >> 56) & 0xff;
- if (config->version > EC_CONFIG_VERSION)
- {
- gf_msg ("ec", GF_LOG_ERROR, EINVAL,
- EC_MSG_UNSUPPORTED_VERSION,
- "Found an unsupported config version (%u)",
- config->version);
+ if (config->version > EC_CONFIG_VERSION) {
+ gf_msg("ec", GF_LOG_ERROR, EINVAL, EC_MSG_UNSUPPORTED_VERSION,
+ "Found an unsupported config version (%u)", config->version);
return -EINVAL;
}
@@ -394,7 +383,8 @@ int32_t ec_dict_del_config(dict_t * dict, char * key, ec_config_t * config)
return 0;
}
-gf_boolean_t ec_loc_gfid_check(xlator_t *xl, uuid_t dst, uuid_t src)
+gf_boolean_t
+ec_loc_gfid_check(xlator_t *xl, uuid_t dst, uuid_t src)
{
if (gf_uuid_is_null(src)) {
return _gf_true;
@@ -407,9 +397,8 @@ gf_boolean_t ec_loc_gfid_check(xlator_t *xl, uuid_t dst, uuid_t src)
}
if (gf_uuid_compare(dst, src) != 0) {
- gf_msg (xl->name, GF_LOG_WARNING, 0,
- EC_MSG_GFID_MISMATCH,
- "Mismatching GFID's in loc");
+ gf_msg(xl->name, GF_LOG_WARNING, 0, EC_MSG_GFID_MISMATCH,
+ "Mismatching GFID's in loc");
return _gf_false;
}
@@ -417,7 +406,8 @@ gf_boolean_t ec_loc_gfid_check(xlator_t *xl, uuid_t dst, uuid_t src)
return _gf_true;
}
-int32_t ec_loc_setup_inode(xlator_t *xl, inode_table_t *table, loc_t *loc)
+int32_t
+ec_loc_setup_inode(xlator_t *xl, inode_table_t *table, loc_t *loc)
{
int32_t ret = -EINVAL;
@@ -428,7 +418,7 @@ int32_t ec_loc_setup_inode(xlator_t *xl, inode_table_t *table, loc_t *loc)
} else if (table != NULL) {
if (!gf_uuid_is_null(loc->gfid)) {
loc->inode = inode_find(table, loc->gfid);
- } else if (loc->path && strchr (loc->path, '/')) {
+ } else if (loc->path && strchr(loc->path, '/')) {
loc->inode = inode_resolve(table, (char *)loc->path);
}
}
@@ -439,7 +429,8 @@ out:
return ret;
}
-int32_t ec_loc_setup_parent(xlator_t *xl, inode_table_t *table, loc_t *loc)
+int32_t
+ec_loc_setup_parent(xlator_t *xl, inode_table_t *table, loc_t *loc)
{
char *path, *parent;
int32_t ret = -EINVAL;
@@ -451,13 +442,11 @@ int32_t ec_loc_setup_parent(xlator_t *xl, inode_table_t *table, loc_t *loc)
} else if (table != NULL) {
if (!gf_uuid_is_null(loc->pargfid)) {
loc->parent = inode_find(table, loc->pargfid);
- } else if (loc->path && strchr (loc->path, '/')) {
+ } else if (loc->path && strchr(loc->path, '/')) {
path = gf_strdup(loc->path);
if (path == NULL) {
- gf_msg (xl->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY,
- "Unable to duplicate path '%s'",
- loc->path);
+ gf_msg(xl->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Unable to duplicate path '%s'", loc->path);
ret = -ENOMEM;
@@ -484,9 +473,10 @@ out:
return ret;
}
-int32_t ec_loc_setup_path(xlator_t *xl, loc_t *loc)
+int32_t
+ec_loc_setup_path(xlator_t *xl, loc_t *loc)
{
- uuid_t root = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1};
+ static uuid_t root = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1};
char *name;
int32_t ret = -EINVAL;
@@ -514,10 +504,8 @@ int32_t ec_loc_setup_path(xlator_t *xl, loc_t *loc)
if (loc->name != NULL) {
if (strcmp(loc->name, name) != 0) {
- gf_msg (xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_INVALID_LOC_NAME,
- "Invalid name '%s' in loc",
- loc->name);
+ gf_msg(xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_INVALID_LOC_NAME,
+ "Invalid name '%s' in loc", loc->name);
goto out;
}
@@ -532,7 +520,8 @@ out:
return ret;
}
-int32_t ec_loc_parent(xlator_t *xl, loc_t *loc, loc_t *parent)
+int32_t
+ec_loc_parent(xlator_t *xl, loc_t *loc, loc_t *parent)
{
inode_table_t *table = NULL;
char *str = NULL;
@@ -549,24 +538,20 @@ int32_t ec_loc_parent(xlator_t *xl, loc_t *loc, loc_t *parent)
if (!gf_uuid_is_null(loc->pargfid)) {
gf_uuid_copy(parent->gfid, loc->pargfid);
}
- if (loc->path && strchr (loc->path, '/')) {
+ if (loc->path && strchr(loc->path, '/')) {
str = gf_strdup(loc->path);
if (str == NULL) {
- gf_msg (xl->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY,
- "Unable to duplicate path '%s'",
- loc->path);
+ gf_msg(xl->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Unable to duplicate path '%s'", loc->path);
- goto out;
+ goto out;
}
parent->path = gf_strdup(dirname(str));
if (parent->path == NULL) {
- gf_msg (xl->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY,
- "Unable to duplicate path '%s'",
- dirname(str));
+ gf_msg(xl->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Unable to duplicate path '%s'", dirname(str));
- goto out;
+ goto out;
}
}
@@ -583,9 +568,8 @@ int32_t ec_loc_parent(xlator_t *xl, loc_t *loc, loc_t *parent)
if ((parent->inode == NULL) && (parent->path == NULL) &&
gf_uuid_is_null(parent->gfid)) {
- gf_msg (xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_LOC_PARENT_INODE_MISSING,
- "Parent inode missing for loc_t");
+ gf_msg(xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_LOC_PARENT_INODE_MISSING,
+ "Parent inode missing for loc_t");
ret = -EINVAL;
@@ -604,8 +588,8 @@ out:
return ret;
}
-int32_t ec_loc_update(xlator_t *xl, loc_t *loc, inode_t *inode,
- struct iatt *iatt)
+int32_t
+ec_loc_update(xlator_t *xl, loc_t *loc, inode_t *inode, struct iatt *iatt)
{
inode_table_t *table = NULL;
int32_t ret = -EINVAL;
@@ -646,9 +630,10 @@ out:
return ret;
}
-int32_t ec_loc_from_fd(xlator_t * xl, loc_t * loc, fd_t * fd)
+int32_t
+ec_loc_from_fd(xlator_t *xl, loc_t *loc, fd_t *fd)
{
- ec_fd_t * ctx;
+ ec_fd_t *ctx;
int32_t ret = -ENOMEM;
memset(loc, 0, sizeof(*loc));
@@ -673,7 +658,8 @@ out:
return ret;
}
-int32_t ec_loc_from_loc(xlator_t * xl, loc_t * dst, loc_t * src)
+int32_t
+ec_loc_from_loc(xlator_t *xl, loc_t *dst, loc_t *src)
{
int32_t ret = -ENOMEM;
@@ -696,62 +682,62 @@ out:
return ret;
}
-void ec_owner_set(call_frame_t * frame, void * owner)
+void
+ec_owner_set(call_frame_t *frame, void *owner)
{
set_lk_owner_from_ptr(&frame->root->lk_owner, owner);
}
-void ec_owner_copy(call_frame_t *frame, gf_lkowner_t *owner)
+void
+ec_owner_copy(call_frame_t *frame, gf_lkowner_t *owner)
{
- lk_owner_copy (&frame->root->lk_owner, owner);
+ lk_owner_copy(&frame->root->lk_owner, owner);
}
static void
-ec_stripe_cache_init (ec_t *ec, ec_inode_t *ctx)
+ec_stripe_cache_init(ec_t *ec, ec_inode_t *ctx)
{
- ec_stripe_list_t *stripe_cache = NULL;
+ ec_stripe_list_t *stripe_cache = NULL;
- stripe_cache = &(ctx->stripe_cache);
- if (stripe_cache->max == 0) {
- stripe_cache->max = ec->stripe_cache;
- }
+ stripe_cache = &(ctx->stripe_cache);
+ if (stripe_cache->max == 0) {
+ stripe_cache->max = ec->stripe_cache;
+ }
}
-ec_inode_t * __ec_inode_get(inode_t * inode, xlator_t * xl)
+ec_inode_t *
+__ec_inode_get(inode_t *inode, xlator_t *xl)
{
- ec_inode_t * ctx = NULL;
+ ec_inode_t *ctx = NULL;
uint64_t value = 0;
- if ((__inode_ctx_get(inode, xl, &value) != 0) || (value == 0))
- {
+ if ((__inode_ctx_get(inode, xl, &value) != 0) || (value == 0)) {
ctx = GF_MALLOC(sizeof(*ctx), ec_mt_ec_inode_t);
- if (ctx != NULL)
- {
+ if (ctx != NULL) {
memset(ctx, 0, sizeof(*ctx));
INIT_LIST_HEAD(&ctx->heal);
INIT_LIST_HEAD(&ctx->stripe_cache.lru);
+ ctx->heal_count = 0;
value = (uint64_t)(uintptr_t)ctx;
- if (__inode_ctx_set(inode, xl, &value) != 0)
- {
+ if (__inode_ctx_set(inode, xl, &value) != 0) {
GF_FREE(ctx);
return NULL;
}
}
- }
- else
- {
+ } else {
ctx = (ec_inode_t *)(uintptr_t)value;
}
if (ctx)
- ec_stripe_cache_init (xl->private, ctx);
+ ec_stripe_cache_init(xl->private, ctx);
return ctx;
}
-ec_inode_t * ec_inode_get(inode_t * inode, xlator_t * xl)
+ec_inode_t *
+ec_inode_get(inode_t *inode, xlator_t *xl)
{
- ec_inode_t * ctx = NULL;
+ ec_inode_t *ctx = NULL;
LOCK(&inode->lock);
@@ -762,32 +748,40 @@ ec_inode_t * ec_inode_get(inode_t * inode, xlator_t * xl)
return ctx;
}
-ec_fd_t * __ec_fd_get(fd_t * fd, xlator_t * xl)
+ec_fd_t *
+__ec_fd_get(fd_t *fd, xlator_t *xl)
{
int i = 0;
- ec_fd_t * ctx = NULL;
+ ec_fd_t *ctx = NULL;
+ ec_inode_t *ictx = NULL;
uint64_t value = 0;
ec_t *ec = xl->private;
if ((__fd_ctx_get(fd, xl, &value) != 0) || (value == 0)) {
- ctx = GF_MALLOC(sizeof(*ctx) + (sizeof (ec_fd_status_t) * ec->nodes),
+ ctx = GF_MALLOC(sizeof(*ctx) + (sizeof(ec_fd_status_t) * ec->nodes),
ec_mt_ec_fd_t);
if (ctx != NULL) {
memset(ctx, 0, sizeof(*ctx));
for (i = 0; i < ec->nodes; i++) {
- if (fd_is_anonymous (fd)) {
- ctx->fd_status[i] = EC_FD_OPENED;
+ if (fd_is_anonymous(fd)) {
+ ctx->fd_status[i] = EC_FD_OPENED;
} else {
- ctx->fd_status[i] = EC_FD_NOT_OPENED;
+ ctx->fd_status[i] = EC_FD_NOT_OPENED;
}
}
value = (uint64_t)(uintptr_t)ctx;
if (__fd_ctx_set(fd, xl, value) != 0) {
- GF_FREE (ctx);
+ GF_FREE(ctx);
return NULL;
}
+ /* Only refering bad-version so no need for lock
+ * */
+ ictx = __ec_inode_get(fd->inode, xl);
+ if (ictx) {
+ ctx->bad_version = ictx->bad_version;
+ }
}
} else {
ctx = (ec_fd_t *)(uintptr_t)value;
@@ -804,9 +798,10 @@ ec_fd_t * __ec_fd_get(fd_t * fd, xlator_t * xl)
return ctx;
}
-ec_fd_t * ec_fd_get(fd_t * fd, xlator_t * xl)
+ec_fd_t *
+ec_fd_get(fd_t *fd, xlator_t *xl)
{
- ec_fd_t * ctx = NULL;
+ ec_fd_t *ctx = NULL;
LOCK(&fd->lock);
@@ -818,37 +813,36 @@ ec_fd_t * ec_fd_get(fd_t * fd, xlator_t * xl)
}
gf_boolean_t
-ec_is_internal_xattr (dict_t *dict, char *key, data_t *value, void *data)
+ec_is_internal_xattr(dict_t *dict, char *key, data_t *value, void *data)
{
- if (key &&
- (strncmp (key, EC_XATTR_PREFIX, strlen (EC_XATTR_PREFIX)) == 0))
- return _gf_true;
+ if (key && (strncmp(key, EC_XATTR_PREFIX, SLEN(EC_XATTR_PREFIX)) == 0))
+ return _gf_true;
- return _gf_false;
+ return _gf_false;
}
void
-ec_filter_internal_xattrs (dict_t *xattr)
+ec_filter_internal_xattrs(dict_t *xattr)
{
- dict_foreach_match (xattr, ec_is_internal_xattr, NULL,
- dict_remove_foreach_fn, NULL);
+ dict_foreach_match(xattr, ec_is_internal_xattr, NULL,
+ dict_remove_foreach_fn, NULL);
}
gf_boolean_t
-ec_is_data_fop (glusterfs_fop_t fop)
+ec_is_data_fop(glusterfs_fop_t fop)
{
- switch (fop) {
+ switch (fop) {
case GF_FOP_WRITE:
case GF_FOP_TRUNCATE:
case GF_FOP_FTRUNCATE:
case GF_FOP_FALLOCATE:
case GF_FOP_DISCARD:
case GF_FOP_ZEROFILL:
- return _gf_true;
+ return _gf_true;
default:
- return _gf_false;
- }
- return _gf_false;
+ return _gf_false;
+ }
+ return _gf_false;
}
/*
gf_boolean_t
diff --git a/xlators/cluster/ec/src/ec-helpers.h b/xlators/cluster/ec/src/ec-helpers.h
index 71b8978abf8..015db514e05 100644
--- a/xlators/cluster/ec/src/ec-helpers.h
+++ b/xlators/cluster/ec/src/ec-helpers.h
@@ -17,60 +17,77 @@
#define EC_IS_ERR(_x) (((uintptr_t)(_x) & ~0xfffULL) == ~0xfffULL)
#define EC_GET_ERR(_x) ((int32_t)(intptr_t)(_x))
-#define EC_ALIGN_CHECK(_ptr, _align) \
- ((((uintptr_t)(_ptr)) & ((_align) - 1)) == 0)
-
-const char * ec_bin(char * str, size_t size, uint64_t value, int32_t digits);
-const char * ec_fop_name(int32_t id);
-void ec_trace(const char * event, ec_fop_data_t * fop, const char * fmt, ...);
-int32_t ec_bits_consume(uint64_t * n);
-size_t ec_iov_copy_to(void * dst, struct iovec * vector, int32_t count,
- off_t offset, size_t size);
-int32_t ec_buffer_alloc(xlator_t *xl, size_t size, struct iobref **piobref,
- void **ptr);
-int32_t ec_dict_set_array(dict_t *dict, char *key,
- uint64_t *value, int32_t size);
-int32_t ec_dict_get_array (dict_t *dict, char *key, uint64_t value[],
- int32_t size);
-
-int32_t ec_dict_del_array(dict_t *dict, char *key,
- uint64_t *value, int32_t size);
-int32_t ec_dict_set_number(dict_t * dict, char * key, uint64_t value);
-int32_t ec_dict_del_number(dict_t * dict, char * key, uint64_t * value);
-int32_t ec_dict_set_config(dict_t * dict, char * key, ec_config_t * config);
-int32_t ec_dict_del_config(dict_t * dict, char * key, ec_config_t * config);
-
-int32_t ec_loc_parent(xlator_t *xl, loc_t *loc, loc_t *parent);
-int32_t ec_loc_update(xlator_t *xl, loc_t *loc, inode_t *inode,
- struct iatt *iatt);
-
-int32_t ec_loc_from_fd(xlator_t * xl, loc_t * loc, fd_t * fd);
-int32_t ec_loc_from_loc(xlator_t * xl, loc_t * dst, loc_t * src);
-
-void ec_owner_set(call_frame_t * frame, void * owner);
-void ec_owner_copy(call_frame_t * frame, gf_lkowner_t * owner);
-
-ec_inode_t * __ec_inode_get(inode_t * inode, xlator_t * xl);
-ec_inode_t * ec_inode_get(inode_t * inode, xlator_t * xl);
-ec_fd_t * __ec_fd_get(fd_t * fd, xlator_t * xl);
-ec_fd_t * ec_fd_get(fd_t * fd, xlator_t * xl);
+#define EC_ALIGN_CHECK(_ptr, _align) ((((uintptr_t)(_ptr)) & ((_align)-1)) == 0)
+
+const char *
+ec_bin(char *str, size_t size, uint64_t value, int32_t digits);
+const char *
+ec_fop_name(int32_t id);
+void
+ec_trace(const char *event, ec_fop_data_t *fop, const char *fmt, ...);
+int32_t
+ec_bits_consume(uint64_t *n);
+size_t
+ec_iov_copy_to(void *dst, struct iovec *vector, int32_t count, off_t offset,
+ size_t size);
+int32_t
+ec_buffer_alloc(xlator_t *xl, size_t size, struct iobref **piobref, void **ptr);
+int32_t
+ec_dict_set_array(dict_t *dict, char *key, uint64_t *value, int32_t size);
+int32_t
+ec_dict_get_array(dict_t *dict, char *key, uint64_t value[], int32_t size);
+
+int32_t
+ec_dict_del_array(dict_t *dict, char *key, uint64_t *value, int32_t size);
+int32_t
+ec_dict_set_number(dict_t *dict, char *key, uint64_t value);
+int32_t
+ec_dict_del_number(dict_t *dict, char *key, uint64_t *value);
+int32_t
+ec_dict_set_config(dict_t *dict, char *key, ec_config_t *config);
+int32_t
+ec_dict_del_config(dict_t *dict, char *key, ec_config_t *config);
+
+int32_t
+ec_loc_parent(xlator_t *xl, loc_t *loc, loc_t *parent);
+int32_t
+ec_loc_update(xlator_t *xl, loc_t *loc, inode_t *inode, struct iatt *iatt);
+
+int32_t
+ec_loc_from_fd(xlator_t *xl, loc_t *loc, fd_t *fd);
+int32_t
+ec_loc_from_loc(xlator_t *xl, loc_t *dst, loc_t *src);
+
+void
+ec_owner_set(call_frame_t *frame, void *owner);
+void
+ec_owner_copy(call_frame_t *frame, gf_lkowner_t *owner);
+
+ec_inode_t *
+__ec_inode_get(inode_t *inode, xlator_t *xl);
+ec_inode_t *
+ec_inode_get(inode_t *inode, xlator_t *xl);
+ec_fd_t *
+__ec_fd_get(fd_t *fd, xlator_t *xl);
+ec_fd_t *
+ec_fd_get(fd_t *fd, xlator_t *xl);
static inline uint32_t
ec_adjust_size_down(ec_t *ec, uint64_t *value, gf_boolean_t scale)
{
- uint64_t head, tmp;
+ uint64_t head, tmp;
- tmp = *value;
- head = tmp % ec->stripe_size;
- tmp -= head;
+ tmp = *value;
+ head = tmp % ec->stripe_size;
+ tmp -= head;
- if (scale) {
- tmp /= ec->fragments;
- }
+ if (scale) {
+ tmp /= ec->fragments;
+ }
- *value = tmp;
+ *value = tmp;
- return (uint32_t)head;
+ return (uint32_t)head;
}
/* This function can cause an overflow if the passed value is too near to the
@@ -79,35 +96,35 @@ ec_adjust_size_down(ec_t *ec, uint64_t *value, gf_boolean_t scale)
static inline int32_t
ec_adjust_size_up(ec_t *ec, uint64_t *value, gf_boolean_t scale)
{
- uint64_t tmp;
- int32_t tail;
-
- tmp = *value;
- /* We first adjust the value down. This never causes overflow. */
- tail = ec_adjust_size_down(ec, &tmp, scale);
-
- /* If the value was already aligned, tail will be 0 and nothing else
- * needs to be done. */
- if (tail != 0) {
- /* Otherwise, we need to compute the real tail and adjust the
- * returned value to the next stripe. */
- tail = ec->stripe_size - tail;
- if (scale) {
- tmp += ec->fragment_size;
- } else {
- tmp += ec->stripe_size;
- /* If no scaling is requested there's a posibility of
- * overflow. */
- if (tmp < ec->stripe_size) {
- tmp = UINT64_MAX;
- tail = -tail;
- }
- }
+ uint64_t tmp;
+ int32_t tail;
+
+ tmp = *value;
+ /* We first adjust the value down. This never causes overflow. */
+ tail = ec_adjust_size_down(ec, &tmp, scale);
+
+ /* If the value was already aligned, tail will be 0 and nothing else
+ * needs to be done. */
+ if (tail != 0) {
+ /* Otherwise, we need to compute the real tail and adjust the
+ * returned value to the next stripe. */
+ tail = ec->stripe_size - tail;
+ if (scale) {
+ tmp += ec->fragment_size;
+ } else {
+ tmp += ec->stripe_size;
+ /* If no scaling is requested there's a possibility of
+ * overflow. */
+ if (tmp < ec->stripe_size) {
+ tmp = UINT64_MAX;
+ tail = -tail;
+ }
}
+ }
- *value = tmp;
+ *value = tmp;
- return tail;
+ return tail;
}
/* This function is equivalent to ec_adjust_size_down() but with a potentially
@@ -115,19 +132,19 @@ ec_adjust_size_up(ec_t *ec, uint64_t *value, gf_boolean_t scale)
static inline uint32_t
ec_adjust_offset_down(ec_t *ec, off_t *value, gf_boolean_t scale)
{
- off_t head, tmp;
+ off_t head, tmp;
- tmp = *value;
- head = tmp % ec->stripe_size;
- tmp -= head;
+ tmp = *value;
+ head = tmp % ec->stripe_size;
+ tmp -= head;
- if (scale) {
- tmp /= ec->fragments;
- }
+ if (scale) {
+ tmp /= ec->fragments;
+ }
- *value = tmp;
+ *value = tmp;
- return (uint32_t)head;
+ return (uint32_t)head;
}
/* This function is equivalent to ec_adjust_size_up() but with a potentially
@@ -135,48 +152,49 @@ ec_adjust_offset_down(ec_t *ec, off_t *value, gf_boolean_t scale)
static inline int32_t
ec_adjust_offset_up(ec_t *ec, off_t *value, gf_boolean_t scale)
{
- uint64_t tail, tmp;
-
- /* An offset is a signed type that can only have positive values, so
- * we take advantage of this to avoid overflows. We simply convert it
- * to an unsigned integer and operate normally. This won't cause an
- * overflow. Overflow is only checked when converting back to an
- * off_t. */
- tmp = *value;
- tail = ec->stripe_size;
- tail -= (tmp + tail - 1) % tail + 1;
- tmp += tail;
- if (scale) {
- /* If we are scaling, we'll never get an overflow. */
- tmp /= ec->fragments;
- } else {
- /* Check if there has been an overflow. */
- if ((off_t)tmp < 0) {
- tmp = GF_OFF_MAX;
- tail = -tail;
- }
+ uint64_t tail, tmp;
+
+ /* An offset is a signed type that can only have positive values, so
+ * we take advantage of this to avoid overflows. We simply convert it
+ * to an unsigned integer and operate normally. This won't cause an
+ * overflow. Overflow is only checked when converting back to an
+ * off_t. */
+ tmp = *value;
+ tail = ec->stripe_size;
+ tail -= (tmp + tail - 1) % tail + 1;
+ tmp += tail;
+ if (scale) {
+ /* If we are scaling, we'll never get an overflow. */
+ tmp /= ec->fragments;
+ } else {
+ /* Check if there has been an overflow. */
+ if ((off_t)tmp < 0) {
+ tmp = GF_OFF_MAX;
+ tail = -tail;
}
+ }
- *value = (off_t)tmp;
+ *value = (off_t)tmp;
- return (int32_t)tail;
+ return (int32_t)tail;
}
-static inline int32_t ec_is_power_of_2(uint32_t value)
+static inline int32_t
+ec_is_power_of_2(uint32_t value)
{
return (value != 0) && ((value & (value - 1)) == 0);
}
gf_boolean_t
-ec_is_internal_xattr (dict_t *dict, char *key, data_t *value, void *data);
+ec_is_internal_xattr(dict_t *dict, char *key, data_t *value, void *data);
void
-ec_filter_internal_xattrs (dict_t *xattr);
+ec_filter_internal_xattrs(dict_t *xattr);
gf_boolean_t
-ec_is_data_fop (glusterfs_fop_t fop);
+ec_is_data_fop(glusterfs_fop_t fop);
int32_t
-ec_launch_replace_heal (ec_t *ec);
+ec_launch_replace_heal(ec_t *ec);
#endif /* __EC_HELPERS_H__ */
diff --git a/xlators/cluster/ec/src/ec-inode-read.c b/xlators/cluster/ec/src/ec-inode-read.c
index d58ed9e5795..dad5f4d7018 100644
--- a/xlators/cluster/ec/src/ec-inode-read.c
+++ b/xlators/cluster/ec/src/ec-inode-read.c
@@ -8,9 +8,6 @@
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "defaults.h"
-
#include "ec.h"
#include "ec-messages.h"
#include "ec-helpers.h"
@@ -21,8 +18,9 @@
/* FOP: access */
-int32_t ec_access_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, dict_t * xdata)
+int32_t
+ec_access_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
ec_fop_data_t *fop = NULL;
ec_cbk_data_t *cbk = NULL;
@@ -35,27 +33,27 @@ int32_t ec_access_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
- cbk = ec_cbk_data_allocate (frame, this, fop, GF_FOP_ACCESS,
- idx, op_ret, op_errno);
+ cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_ACCESS, idx, op_ret,
+ op_errno);
if (cbk) {
if (xdata)
- cbk->xdata = dict_ref (xdata);
- ec_combine (cbk, NULL);
+ cbk->xdata = dict_ref(xdata);
+ ec_combine(cbk, NULL);
}
out:
- if (fop != NULL)
- {
- ec_complete (fop);
+ if (fop != NULL) {
+ ec_complete(fop);
}
return 0;
}
-void ec_wind_access(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_access(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -67,19 +65,19 @@ void ec_wind_access(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
int32_t
ec_manager_access(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t *cbk = NULL;
+ ec_cbk_data_t *cbk = NULL;
- switch (state) {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_LOCK:
- ec_lock_prepare_inode (fop, &fop->loc[0], EC_QUERY_INFO, 0,
- LLONG_MAX);
- ec_lock (fop);
+ ec_lock_prepare_inode(fop, &fop->loc[0], EC_QUERY_INFO, 0,
+ EC_RANGE_FULL);
+ ec_lock(fop);
return EC_STATE_DISPATCH;
case EC_STATE_DISPATCH:
- ec_dispatch_one (fop);
+ ec_dispatch_one(fop);
return EC_STATE_PREPARE_ANSWER;
@@ -92,12 +90,11 @@ ec_manager_access(ec_fop_data_t *fop, int32_t state)
case EC_STATE_REPORT:
cbk = fop->answer;
- GF_ASSERT (cbk);
+ GF_ASSERT(cbk);
if (fop->cbks.access != NULL) {
if (cbk) {
- fop->cbks.access(fop->req_frame, fop, fop->xl,
- cbk->op_ret, cbk->op_errno,
- cbk->xdata);
+ fop->cbks.access(fop->req_frame, fop, fop->xl, cbk->op_ret,
+ cbk->op_errno, cbk->xdata);
}
}
return EC_STATE_LOCK_REUSE;
@@ -108,8 +105,8 @@ ec_manager_access(ec_fop_data_t *fop, int32_t state)
case -EC_STATE_PREPARE_ANSWER:
case -EC_STATE_REPORT:
if (fop->cbks.access != NULL) {
- fop->cbks.access(fop->req_frame, fop, fop->xl, -1,
- fop->error, NULL);
+ fop->cbks.access(fop->req_frame, fop, fop->xl, -1, fop->error,
+ NULL);
}
return EC_STATE_LOCK_REUSE;
@@ -126,31 +123,30 @@ ec_manager_access(ec_fop_data_t *fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE,
- "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
- }
+ }
}
-void ec_access(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_access_cbk_t func, void * data,
- loc_t * loc, int32_t mask, dict_t * xdata)
+void
+ec_access(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_access_cbk_t func, void *data, loc_t *loc,
+ int32_t mask, dict_t *xdata)
{
- ec_cbk_t callback = { .access = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.access = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(ACCESS) %p", frame);
+ gf_msg_trace("ec", 0, "EC(ACCESS) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_ACCESS, EC_FLAG_LOCK_SHARED,
- target, minimum, ec_wind_access,
+ target, fop_flags, ec_wind_access,
ec_manager_access, callback, data);
if (fop == NULL) {
goto out;
@@ -160,9 +156,8 @@ void ec_access(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL,
- "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
@@ -170,10 +165,9 @@ void ec_access(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -191,14 +185,13 @@ out:
/* FOP: getxattr */
-int32_t ec_combine_getxattr(ec_fop_data_t * fop, ec_cbk_data_t * dst,
- ec_cbk_data_t * src)
+int32_t
+ec_combine_getxattr(ec_fop_data_t *fop, ec_cbk_data_t *dst, ec_cbk_data_t *src)
{
- if (!ec_dict_compare(dst->dict, src->dict))
- {
- gf_msg (fop->xl->name, GF_LOG_NOTICE, 0,
- EC_MSG_DICT_MISMATCH, "Mismatching dictionary in "
- "answers of 'GF_FOP_GETXATTR'");
+ if (!ec_dict_compare(dst->dict, src->dict)) {
+ gf_msg(fop->xl->name, GF_LOG_NOTICE, 0, EC_MSG_DICT_MISMATCH,
+ "Mismatching dictionary in "
+ "answers of 'GF_FOP_GETXATTR'");
return 0;
}
@@ -206,12 +199,12 @@ int32_t ec_combine_getxattr(ec_fop_data_t * fop, ec_cbk_data_t * dst,
return 1;
}
-int32_t ec_getxattr_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, dict_t * dict,
- dict_t * xdata)
+int32_t
+ec_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -221,38 +214,30 @@ int32_t ec_getxattr_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_GETXATTR, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (op_ret >= 0)
- {
- if (dict != NULL)
- {
+ if (cbk != NULL) {
+ if (op_ret >= 0) {
+ if (dict != NULL) {
cbk->dict = dict_ref(dict);
- if (cbk->dict == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ if (cbk->dict == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
}
}
- if (xdata != NULL)
- {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -262,15 +247,15 @@ int32_t ec_getxattr_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_getxattr(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_getxattr(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -280,43 +265,43 @@ void ec_wind_getxattr(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
}
void
-ec_handle_special_xattrs (ec_fop_data_t *fop)
+ec_handle_special_xattrs(ec_fop_data_t *fop)
{
- ec_cbk_data_t *cbk = NULL;
- /* Stime may not be available on all the bricks, so even if some of the
- * subvols succeed the operation, treat it as answer.*/
- if (fop->str[0] &&
- fnmatch (GF_XATTR_STIME_PATTERN, fop->str[0], 0) == 0) {
- if (!fop->answer || (fop->answer->op_ret < 0)) {
- list_for_each_entry (cbk, &fop->cbk_list, list) {
- if (cbk->op_ret >= 0) {
- fop->answer = cbk;
- break;
- }
- }
+ ec_cbk_data_t *cbk = NULL;
+ /* Stime may not be available on all the bricks, so even if some of the
+ * subvols succeed the operation, treat it as answer.*/
+ if (fop->str[0] && fnmatch(GF_XATTR_STIME_PATTERN, fop->str[0], 0) == 0) {
+ if (!fop->answer || (fop->answer->op_ret < 0)) {
+ list_for_each_entry(cbk, &fop->cbk_list, list)
+ {
+ if (cbk->op_ret >= 0) {
+ fop->answer = cbk;
+ break;
}
+ }
}
+ }
}
-int32_t ec_manager_getxattr(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_getxattr(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_LOCK:
/* clear-locks commands must be done without any locks acquired
to avoid interferences. */
if ((fop->str[0] == NULL) ||
(strncmp(fop->str[0], GF_XATTR_CLRLK_CMD,
- strlen(GF_XATTR_CLRLK_CMD)) != 0)) {
+ SLEN(GF_XATTR_CLRLK_CMD)) != 0)) {
if (fop->fd == NULL) {
- ec_lock_prepare_inode(fop, &fop->loc[0], EC_QUERY_INFO,
- 0, LLONG_MAX);
+ ec_lock_prepare_inode(fop, &fop->loc[0], EC_QUERY_INFO, 0,
+ EC_RANGE_FULL);
} else {
ec_lock_prepare_fd(fop, fop->fd, EC_QUERY_INFO, 0,
- LLONG_MAX);
+ EC_RANGE_FULL);
}
ec_lock(fop);
}
@@ -333,11 +318,11 @@ int32_t ec_manager_getxattr(ec_fop_data_t * fop, int32_t state)
return EC_STATE_PREPARE_ANSWER;
case EC_STATE_PREPARE_ANSWER:
- ec_handle_special_xattrs (fop);
+ ec_handle_special_xattrs(fop);
if (fop->minimum == EC_MINIMUM_ALL) {
cbk = ec_fop_prepare_answer(fop, _gf_true);
} else {
- if (ec_dispatch_one_retry (fop, &cbk)) {
+ if (ec_dispatch_one_retry(fop, &cbk)) {
return EC_STATE_DISPATCH;
}
}
@@ -347,10 +332,10 @@ int32_t ec_manager_getxattr(ec_fop_data_t * fop, int32_t state)
err = ec_dict_combine(cbk, EC_COMBINE_DICT);
if (!ec_cbk_set_error(cbk, -err, _gf_true)) {
if (cbk->xdata != NULL)
- ec_filter_internal_xattrs (cbk->xdata);
+ ec_filter_internal_xattrs(cbk->xdata);
if (cbk->dict != NULL)
- ec_filter_internal_xattrs (cbk->dict);
+ ec_filter_internal_xattrs(cbk->dict);
}
}
@@ -361,8 +346,7 @@ int32_t ec_manager_getxattr(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.getxattr != NULL)
- {
+ if (fop->cbks.getxattr != NULL) {
fop->cbks.getxattr(fop->req_frame, fop, fop->xl, cbk->op_ret,
cbk->op_errno, cbk->dict, cbk->xdata);
}
@@ -376,10 +360,9 @@ int32_t ec_manager_getxattr(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.getxattr != NULL)
- {
- fop->cbks.getxattr(fop->req_frame, fop, fop->xl, -1,
- fop->error, NULL, NULL);
+ if (fop->cbks.getxattr != NULL) {
+ fop->cbks.getxattr(fop->req_frame, fop, fop->xl, -1, fop->error,
+ NULL, NULL);
}
return EC_STATE_LOCK_REUSE;
@@ -397,26 +380,44 @@ int32_t ec_manager_getxattr(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE,
- "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-int32_t ec_getxattr_heal_cbk(call_frame_t *frame, void *cookie, xlator_t *xl,
- int32_t op_ret, int32_t op_errno, uintptr_t mask,
- uintptr_t good, uintptr_t bad, dict_t *xdata)
+int32_t
+ec_getxattr_heal_cbk(call_frame_t *frame, void *cookie, xlator_t *xl,
+ int32_t op_ret, int32_t op_errno, uintptr_t mask,
+ uintptr_t good, uintptr_t bad, uint32_t pending,
+ dict_t *xdata)
{
- ec_fop_data_t *fop = cookie;
- fop_getxattr_cbk_t func = fop->data;
+ fop_getxattr_cbk_t func = cookie;
ec_t *ec = xl->private;
dict_t *dict = NULL;
char *str;
char bin1[65], bin2[65];
+ /* We try to return the 'pending' information in xdata, but if this cannot
+ * be set, we will ignore it silently. We prefer to report the success or
+ * failure of the heal itself. */
+ if (xdata == NULL) {
+ xdata = dict_new();
+ } else {
+ dict_ref(xdata);
+ }
+ if (xdata != NULL) {
+ if (dict_set_uint32(xdata, EC_XATTR_HEAL_NEW, pending) != 0) {
+ /* dict_set_uint32() is marked as 'warn_unused_result' and gcc
+ * enforces to check the result in this case. However we don't
+ * really care if it succeeded or not. We'll just do the same.
+ *
+ * This empty 'if' avoids the warning, and it will be removed by
+ * the optimizer. */
+ }
+ }
+
if (op_ret >= 0) {
dict = dict_new();
if (dict == NULL) {
@@ -450,25 +451,28 @@ int32_t ec_getxattr_heal_cbk(call_frame_t *frame, void *cookie, xlator_t *xl,
}
out:
- func(frame, NULL, xl, op_ret, op_errno, dict, NULL);
+ func(frame, NULL, xl, op_ret, op_errno, dict, xdata);
if (dict != NULL) {
dict_unref(dict);
}
+ if (xdata != NULL) {
+ dict_unref(xdata);
+ }
return 0;
}
void
-ec_getxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
- int32_t minimum, fop_getxattr_cbk_t func, void *data,
- loc_t *loc, const char *name, dict_t *xdata)
+ec_getxattr(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_getxattr_cbk_t func, void *data, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- ec_cbk_t callback = { .getxattr = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.getxattr = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(GETXATTR) %p", frame);
+ gf_msg_trace("ec", 0, "EC(GETXATTR) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
@@ -476,25 +480,23 @@ ec_getxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
/* Special handling of an explicit self-heal request */
if ((name != NULL) && (strcmp(name, EC_XATTR_HEAL) == 0)) {
- ec_heal(frame, this, target, EC_MINIMUM_ONE, ec_getxattr_heal_cbk,
- func, loc, 0, NULL);
+ ec_heal(frame, this, target, EC_MINIMUM_ONE, ec_getxattr_heal_cbk, func,
+ loc, 0, NULL);
return;
}
- fop = ec_fop_data_allocate(frame, this, GF_FOP_GETXATTR,
- EC_FLAG_LOCK_SHARED, target, minimum,
- ec_wind_getxattr, ec_manager_getxattr, callback,
- data);
+ fop = ec_fop_data_allocate(
+ frame, this, GF_FOP_GETXATTR, EC_FLAG_LOCK_SHARED, target, fop_flags,
+ ec_wind_getxattr, ec_manager_getxattr, callback, data);
if (fop == NULL) {
goto out;
}
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL,
- "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
@@ -503,15 +505,14 @@ ec_getxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
/* In case of list-node-uuids xattr, set flag to indicate
* the same and use node-uuid xattr for winding fop */
if (XATTR_IS_NODE_UUID_LIST(name)) {
- fop->int32 = 1;
- fop->str[0] = gf_strdup(GF_XATTR_NODE_UUID_KEY);
+ fop->int32 = 1;
+ fop->str[0] = gf_strdup(GF_XATTR_NODE_UUID_KEY);
} else {
- fop->str[0] = gf_strdup(name);
+ fop->str[0] = gf_strdup(name);
}
if (fop->str[0] == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY,
- "Failed to duplicate a string.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to duplicate a string.");
goto out;
}
@@ -519,10 +520,9 @@ ec_getxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -532,20 +532,20 @@ ec_getxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
out:
if (fop != NULL) {
- ec_manager (fop, error);
+ ec_manager(fop, error);
} else {
- func (frame, NULL, this, -1, error, NULL, NULL);
+ func(frame, NULL, this, -1, error, NULL, NULL);
}
}
/* FOP: fgetxattr */
-int32_t ec_fgetxattr_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, dict_t * dict,
- dict_t * xdata)
+int32_t
+ec_fgetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -555,38 +555,30 @@ int32_t ec_fgetxattr_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_FGETXATTR, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (op_ret >= 0)
- {
- if (dict != NULL)
- {
+ if (cbk != NULL) {
+ if (op_ret >= 0) {
+ if (dict != NULL) {
cbk->dict = dict_ref(dict);
- if (cbk->dict == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ if (cbk->dict == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
}
}
- if (xdata != NULL)
- {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -596,8 +588,7 @@ int32_t ec_fgetxattr_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
@@ -605,7 +596,7 @@ out:
}
void
-ec_wind_fgetxattr (ec_t *ec, ec_fop_data_t *fop, int32_t idx)
+ec_wind_fgetxattr(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -615,24 +606,23 @@ ec_wind_fgetxattr (ec_t *ec, ec_fop_data_t *fop, int32_t idx)
}
void
-ec_fgetxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
- int32_t minimum, fop_fgetxattr_cbk_t func, void *data,
- fd_t *fd, const char *name, dict_t *xdata)
+ec_fgetxattr(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fgetxattr_cbk_t func, void *data, fd_t *fd,
+ const char *name, dict_t *xdata)
{
- ec_cbk_t callback = { .fgetxattr = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.fgetxattr = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(FGETXATTR) %p", frame);
+ gf_msg_trace("ec", 0, "EC(FGETXATTR) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_FGETXATTR,
- EC_FLAG_LOCK_SHARED, target, minimum,
- ec_wind_fgetxattr, ec_manager_getxattr,
- callback, data);
+ fop = ec_fop_data_allocate(
+ frame, this, GF_FOP_FGETXATTR, EC_FLAG_LOCK_SHARED, target, fop_flags,
+ ec_wind_fgetxattr, ec_manager_getxattr, callback, data);
if (fop == NULL) {
goto out;
}
@@ -642,10 +632,9 @@ ec_fgetxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL,
- "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -653,8 +642,8 @@ ec_fgetxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
if (name != NULL) {
fop->str[0] = gf_strdup(name);
if (fop->str[0] == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY, "Failed to duplicate a string.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to duplicate a string.");
goto out;
}
@@ -662,9 +651,9 @@ ec_fgetxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -674,22 +663,21 @@ ec_fgetxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
out:
if (fop != NULL) {
- ec_manager (fop, error);
+ ec_manager(fop, error);
} else {
- func (frame, NULL, this, -1, error, NULL, NULL);
+ func(frame, NULL, this, -1, error, NULL, NULL);
}
}
/* FOP: open */
-int32_t ec_combine_open(ec_fop_data_t * fop, ec_cbk_data_t * dst,
- ec_cbk_data_t * src)
+int32_t
+ec_combine_open(ec_fop_data_t *fop, ec_cbk_data_t *dst, ec_cbk_data_t *src)
{
- if (dst->fd != src->fd)
- {
- gf_msg (fop->xl->name, GF_LOG_NOTICE, 0,
- EC_MSG_FD_MISMATCH, "Mismatching fd in answers "
- "of 'GF_FOP_OPEN': %p <-> %p",
+ if (dst->fd != src->fd) {
+ gf_msg(fop->xl->name, GF_LOG_NOTICE, 0, EC_MSG_FD_MISMATCH,
+ "Mismatching fd in answers "
+ "of 'GF_FOP_OPEN': %p <-> %p",
dst->fd, src->fd);
return 0;
@@ -698,12 +686,12 @@ int32_t ec_combine_open(ec_fop_data_t * fop, ec_cbk_data_t * dst,
return 1;
}
-int32_t ec_open_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, fd_t * fd,
- dict_t * xdata)
+int32_t
+ec_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -713,36 +701,31 @@ int32_t ec_open_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_OPEN, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (op_ret >= 0)
- {
- if (fd != NULL)
- {
+ if (cbk != NULL) {
+ if (op_ret >= 0) {
+ if (fd != NULL) {
cbk->fd = fd_ref(fd);
- if (cbk->fd == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL, "Failed to reference a "
- "file descriptor.");
+ if (cbk->fd == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
}
}
- if (xdata != NULL)
- {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -750,20 +733,19 @@ int32_t ec_open_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
ec_combine(cbk, ec_combine_open);
- ec_update_fd_status (fd, this, idx, op_ret);
-
+ ec_update_fd_status(fd, this, idx, op_ret);
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_open(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_open(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -772,10 +754,10 @@ void ec_wind_open(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
&fop->loc[0], fop->int32, fop->fd, fop->xdata);
}
-int32_t ec_open_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *prebuf, struct iatt *postbuf,
- dict_t *xdata)
+int32_t
+ec_open_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
ec_fop_data_t *fop = cookie;
int32_t error = 0;
@@ -792,14 +774,14 @@ int32_t ec_open_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
return 0;
}
-int32_t ec_manager_open(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_open(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
- ec_fd_t * ctx;
+ ec_cbk_data_t *cbk;
+ ec_fd_t *ctx;
int32_t err;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
LOCK(&fop->fd->lock);
@@ -811,13 +793,15 @@ int32_t ec_manager_open(ec_fop_data_t * fop, int32_t state)
return EC_STATE_REPORT;
}
- err = ec_loc_from_loc(fop->xl, &ctx->loc, &fop->loc[0]);
- if (err != 0) {
- UNLOCK(&fop->fd->lock);
+ if (!ctx->loc.inode) {
+ err = ec_loc_from_loc(fop->xl, &ctx->loc, &fop->loc[0]);
+ if (err != 0) {
+ UNLOCK(&fop->fd->lock);
- fop->error = -err;
+ fop->error = -err;
- return EC_STATE_REPORT;
+ return EC_STATE_REPORT;
+ }
}
ctx->flags = fop->int32;
@@ -833,7 +817,7 @@ int32_t ec_manager_open(ec_fop_data_t * fop, int32_t state)
fop->uint32 = fop->int32 & O_TRUNC;
fop->int32 &= ~(O_APPEND | O_TRUNC);
- /* Fall through */
+ /* Fall through */
case EC_STATE_DISPATCH:
ec_dispatch_all(fop);
@@ -864,8 +848,8 @@ int32_t ec_manager_open(ec_fop_data_t * fop, int32_t state)
if (fop->uint32 != 0) {
ec_sleep(fop);
ec_ftruncate(fop->req_frame, fop->xl, cbk->mask,
- fop->minimum, ec_open_truncate_cbk,
- fop, cbk->fd, 0, NULL);
+ fop->minimum, ec_open_truncate_cbk, fop,
+ cbk->fd, 0, NULL);
}
}
}
@@ -877,8 +861,7 @@ int32_t ec_manager_open(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.open != NULL)
- {
+ if (fop->cbks.open != NULL) {
fop->cbks.open(fop->req_frame, fop, fop->xl, cbk->op_ret,
cbk->op_errno, cbk->fd, cbk->xdata);
}
@@ -891,8 +874,7 @@ int32_t ec_manager_open(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.open != NULL)
- {
+ if (fop->cbks.open != NULL) {
fop->cbks.open(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL, NULL);
}
@@ -900,30 +882,30 @@ int32_t ec_manager_open(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_open(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_open_cbk_t func, void * data, loc_t * loc,
- int32_t flags, fd_t * fd, dict_t * xdata)
+void
+ec_open(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_open_cbk_t func, void *data, loc_t *loc,
+ int32_t flags, fd_t *fd, dict_t *xdata)
{
- ec_cbk_t callback = { .open = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.open = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(OPEN) %p", frame);
+ gf_msg_trace("ec", 0, "EC(OPEN) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_OPEN, EC_FLAG_LOCK_SHARED,
- target, minimum, ec_wind_open, ec_manager_open,
+ target, fop_flags, ec_wind_open, ec_manager_open,
callback, data);
if (fop == NULL) {
goto out;
@@ -933,8 +915,8 @@ void ec_open(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
@@ -942,9 +924,9 @@ void ec_open(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL, "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -952,9 +934,9 @@ void ec_open(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -972,13 +954,13 @@ out:
/* FOP: readlink */
-int32_t ec_combine_readlink(ec_fop_data_t * fop, ec_cbk_data_t * dst,
- ec_cbk_data_t * src)
+int32_t
+ec_combine_readlink(ec_fop_data_t *fop, ec_cbk_data_t *dst, ec_cbk_data_t *src)
{
if (!ec_iatt_combine(fop, dst->iatt, src->iatt, 1)) {
- gf_msg (fop->xl->name, GF_LOG_NOTICE, 0,
- EC_MSG_IATT_MISMATCH, "Mismatching iatt in "
- "answers of 'GF_FOP_READLINK'");
+ gf_msg(fop->xl->name, GF_LOG_NOTICE, 0, EC_MSG_IATT_MISMATCH,
+ "Mismatching iatt in "
+ "answers of 'GF_FOP_READLINK'");
return 0;
}
@@ -987,13 +969,13 @@ int32_t ec_combine_readlink(ec_fop_data_t * fop, ec_cbk_data_t * dst,
}
int32_t
-ec_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, const char *path,
- struct iatt *buf, dict_t *xdata)
+ec_readlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, const char *path,
+ struct iatt *buf, dict_t *xdata)
{
- ec_fop_data_t *fop = NULL;
- ec_cbk_data_t *cbk = NULL;
- int32_t idx = (int32_t)(uintptr_t)cookie;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
+ int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
@@ -1002,23 +984,23 @@ ec_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
- cbk = ec_cbk_data_allocate (frame, this, fop, fop->id,
- idx, op_ret, op_errno);
+ cbk = ec_cbk_data_allocate(frame, this, fop, fop->id, idx, op_ret,
+ op_errno);
if (cbk) {
- if (xdata)
- cbk->xdata = dict_ref (xdata);
-
- if (cbk->op_ret >= 0) {
- cbk->iatt[0] = *buf;
- cbk->str = gf_strdup (path);
- if (!cbk->str) {
- ec_cbk_set_error(cbk, ENOMEM, _gf_true);
- }
+ if (xdata)
+ cbk->xdata = dict_ref(xdata);
+
+ if (cbk->op_ret >= 0) {
+ cbk->iatt[0] = *buf;
+ cbk->str = gf_strdup(path);
+ if (!cbk->str) {
+ ec_cbk_set_error(cbk, ENOMEM, _gf_true);
}
- ec_combine (cbk, NULL);
+ }
+ ec_combine(cbk, NULL);
}
out:
@@ -1028,7 +1010,8 @@ out:
return 0;
}
-void ec_wind_readlink(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_readlink(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -1037,21 +1020,21 @@ void ec_wind_readlink(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
&fop->loc[0], fop->size, fop->xdata);
}
-int32_t ec_manager_readlink(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_readlink(ec_fop_data_t *fop, int32_t state)
{
ec_cbk_data_t *cbk = NULL;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_LOCK:
- ec_lock_prepare_inode (fop, &fop->loc[0], EC_QUERY_INFO, 0,
- LLONG_MAX);
- ec_lock (fop);
+ ec_lock_prepare_inode(fop, &fop->loc[0], EC_QUERY_INFO, 0,
+ EC_RANGE_FULL);
+ ec_lock(fop);
return EC_STATE_DISPATCH;
case EC_STATE_DISPATCH:
- ec_dispatch_one (fop);
+ ec_dispatch_one(fop);
return EC_STATE_PREPARE_ANSWER;
@@ -1068,11 +1051,11 @@ int32_t ec_manager_readlink(ec_fop_data_t * fop, int32_t state)
case EC_STATE_REPORT:
cbk = fop->answer;
- GF_ASSERT (cbk);
+ GF_ASSERT(cbk);
if (fop->cbks.readlink != NULL) {
- fop->cbks.readlink (fop->req_frame, fop, fop->xl, cbk->op_ret,
- cbk->op_errno, cbk->str, &cbk->iatt[0],
- cbk->xdata);
+ fop->cbks.readlink(fop->req_frame, fop, fop->xl, cbk->op_ret,
+ cbk->op_errno, cbk->str, &cbk->iatt[0],
+ cbk->xdata);
}
return EC_STATE_LOCK_REUSE;
@@ -1083,8 +1066,8 @@ int32_t ec_manager_readlink(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_PREPARE_ANSWER:
case -EC_STATE_REPORT:
if (fop->cbks.readlink != NULL) {
- fop->cbks.readlink(fop->req_frame, fop, fop->xl, -1,
- fop->error, NULL, NULL, NULL);
+ fop->cbks.readlink(fop->req_frame, fop, fop->xl, -1, fop->error,
+ NULL, NULL, NULL);
}
return EC_STATE_LOCK_REUSE;
@@ -1100,32 +1083,31 @@ int32_t ec_manager_readlink(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_readlink(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_readlink_cbk_t func, void * data,
- loc_t * loc, size_t size, dict_t * xdata)
+void
+ec_readlink(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_readlink_cbk_t func, void *data, loc_t *loc,
+ size_t size, dict_t *xdata)
{
- ec_cbk_t callback = { .readlink = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.readlink = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(READLINK) %p", frame);
+ gf_msg_trace("ec", 0, "EC(READLINK) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_READLINK,
- EC_FLAG_LOCK_SHARED, target, minimum,
- ec_wind_readlink, ec_manager_readlink, callback,
- data);
+ fop = ec_fop_data_allocate(
+ frame, this, GF_FOP_READLINK, EC_FLAG_LOCK_SHARED, target, fop_flags,
+ ec_wind_readlink, ec_manager_readlink, callback, data);
if (fop == NULL) {
goto out;
}
@@ -1134,8 +1116,8 @@ void ec_readlink(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
@@ -1143,9 +1125,9 @@ void ec_readlink(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -1163,13 +1145,14 @@ out:
/* FOP: readv */
-int32_t ec_readv_rebuild(ec_t * ec, ec_fop_data_t * fop, ec_cbk_data_t * cbk)
+int32_t
+ec_readv_rebuild(ec_t *ec, ec_fop_data_t *fop, ec_cbk_data_t *cbk)
{
struct iovec vector[1];
ec_cbk_data_t *ans = NULL;
struct iobref *iobref = NULL;
void *ptr;
- size_t fsize = 0, size = 0, max = 0;
+ uint64_t fsize = 0, size = 0, max = 0;
int32_t pos, err = -ENOMEM;
if (cbk->op_ret < 0) {
@@ -1256,22 +1239,21 @@ out:
return err;
}
-int32_t ec_combine_readv(ec_fop_data_t * fop, ec_cbk_data_t * dst,
- ec_cbk_data_t * src)
+int32_t
+ec_combine_readv(ec_fop_data_t *fop, ec_cbk_data_t *dst, ec_cbk_data_t *src)
{
- if (!ec_vector_compare(dst->vector, dst->int32, src->vector, src->int32))
- {
- gf_msg (fop->xl->name, GF_LOG_NOTICE, 0,
- EC_MSG_VECTOR_MISMATCH, "Mismatching vector in "
- "answers of 'GF_FOP_READ'");
+ if (!ec_vector_compare(dst->vector, dst->int32, src->vector, src->int32)) {
+ gf_msg(fop->xl->name, GF_LOG_NOTICE, 0, EC_MSG_VECTOR_MISMATCH,
+ "Mismatching vector in "
+ "answers of 'GF_FOP_READ'");
return 0;
}
if (!ec_iatt_combine(fop, dst->iatt, src->iatt, 1)) {
- gf_msg (fop->xl->name, GF_LOG_NOTICE, 0,
- EC_MSG_IATT_MISMATCH, "Mismatching iatt in "
- "answers of 'GF_FOP_READ'");
+ gf_msg(fop->xl->name, GF_LOG_NOTICE, 0, EC_MSG_IATT_MISMATCH,
+ "Mismatching iatt in "
+ "answers of 'GF_FOP_READ'");
return 0;
}
@@ -1279,14 +1261,14 @@ int32_t ec_combine_readv(ec_fop_data_t * fop, ec_cbk_data_t * dst,
return 1;
}
-int32_t ec_readv_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, struct iovec * vector,
- int32_t count, struct iatt * stbuf,
- struct iobref * iobref, dict_t * xdata)
+int32_t
+ec_readv_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iovec *vector, int32_t count,
+ struct iatt *stbuf, struct iobref *iobref, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
- ec_t * ec = this->private;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
+ ec_t *ec = this->private;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -1296,8 +1278,8 @@ int32_t ec_readv_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_READ, idx, op_ret,
op_errno);
@@ -1308,9 +1290,9 @@ int32_t ec_readv_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
if (count > 0) {
cbk->vector = iov_dup(vector, count);
if (cbk->vector == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY, "Failed to duplicate a "
- "vector list.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to duplicate a "
+ "vector list.");
goto out;
}
@@ -1322,9 +1304,9 @@ int32_t ec_readv_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
if (iobref != NULL) {
cbk->buffers = iobref_ref(iobref);
if (cbk->buffers == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_BUF_REF_FAIL, "Failed to reference a "
- "buffer.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_BUF_REF_FAIL,
+ "Failed to reference a "
+ "buffer.");
goto out;
}
@@ -1333,9 +1315,9 @@ int32_t ec_readv_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
if (cbk->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -1356,7 +1338,8 @@ out:
return 0;
}
-void ec_wind_readv(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_readv(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -1365,12 +1348,13 @@ void ec_wind_readv(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->size, fop->offset, fop->uint32, fop->xdata);
}
-int32_t ec_manager_readv(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_readv(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
+ ec_t *ec = fop->xl->private;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
fop->user_size = fop->size;
fop->head = ec_adjust_offset_down(fop->xl->private, &fop->offset,
@@ -1378,7 +1362,7 @@ int32_t ec_manager_readv(ec_fop_data_t * fop, int32_t state)
fop->size += fop->head;
ec_adjust_size_up(fop->xl->private, &fop->size, _gf_true);
- /* Fall through */
+ /* Fall through */
case EC_STATE_LOCK:
ec_lock_prepare_fd(fop, fop->fd, EC_QUERY_INFO, fop->offset,
@@ -1388,6 +1372,9 @@ int32_t ec_manager_readv(ec_fop_data_t * fop, int32_t state)
return EC_STATE_DISPATCH;
case EC_STATE_DISPATCH:
+ if (ec->read_mask) {
+ fop->mask &= ec->read_mask;
+ }
ec_dispatch_min(fop);
return EC_STATE_PREPARE_ANSWER;
@@ -1397,8 +1384,7 @@ int32_t ec_manager_readv(ec_fop_data_t * fop, int32_t state)
if (cbk != NULL) {
int32_t err;
- ec_iatt_rebuild(fop->xl->private, cbk->iatt, 1,
- cbk->count);
+ ec_iatt_rebuild(fop->xl->private, cbk->iatt, 1, cbk->count);
err = ec_readv_rebuild(fop->xl->private, fop, cbk);
if (err != 0) {
@@ -1413,8 +1399,7 @@ int32_t ec_manager_readv(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.readv != NULL)
- {
+ if (fop->cbks.readv != NULL) {
fop->cbks.readv(fop->req_frame, fop, fop->xl, cbk->op_ret,
cbk->op_errno, cbk->vector, cbk->int32,
&cbk->iatt[0], cbk->buffers, cbk->xdata);
@@ -1429,8 +1414,7 @@ int32_t ec_manager_readv(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.readv != NULL)
- {
+ if (fop->cbks.readv != NULL) {
fop->cbks.readv(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL, 0, NULL, NULL, NULL);
}
@@ -1450,30 +1434,30 @@ int32_t ec_manager_readv(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_readv(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_readv_cbk_t func, void * data, fd_t * fd,
- size_t size, off_t offset, uint32_t flags, dict_t * xdata)
+void
+ec_readv(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_readv_cbk_t func, void *data, fd_t *fd,
+ size_t size, off_t offset, uint32_t flags, dict_t *xdata)
{
- ec_cbk_t callback = { .readv = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.readv = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(READ) %p", frame);
+ gf_msg_trace("ec", 0, "EC(READ) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_READ, EC_FLAG_LOCK_SHARED,
- target, minimum, ec_wind_readv,
+ target, fop_flags, ec_wind_readv,
ec_manager_readv, callback, data);
if (fop == NULL) {
goto out;
@@ -1488,9 +1472,9 @@ void ec_readv(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL, "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -1498,9 +1482,9 @@ void ec_readv(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -1518,9 +1502,9 @@ out:
/* FOP: seek */
-int32_t ec_seek_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, off_t offset,
- dict_t *xdata)
+int32_t
+ec_seek_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, off_t offset, dict_t *xdata)
{
ec_fop_data_t *fop = NULL;
ec_cbk_data_t *cbk = NULL;
@@ -1534,8 +1518,8 @@ int32_t ec_seek_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_SEEK, idx, op_ret,
op_errno);
@@ -1563,7 +1547,8 @@ out:
return 0;
}
-void ec_wind_seek(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
+void
+ec_wind_seek(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -1572,126 +1557,128 @@ void ec_wind_seek(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
fop->offset, fop->seek, fop->xdata);
}
-int32_t ec_manager_seek(ec_fop_data_t *fop, int32_t state)
+int32_t
+ec_manager_seek(ec_fop_data_t *fop, int32_t state)
{
ec_cbk_data_t *cbk;
- size_t size;
+ uint64_t size;
switch (state) {
- case EC_STATE_INIT:
- fop->user_size = fop->offset;
- fop->head = ec_adjust_offset_down(fop->xl->private, &fop->offset,
- _gf_true);
+ case EC_STATE_INIT:
+ fop->user_size = fop->offset;
+ fop->head = ec_adjust_offset_down(fop->xl->private, &fop->offset,
+ _gf_true);
- /* Fall through */
+ /* Fall through */
- case EC_STATE_LOCK:
- ec_lock_prepare_fd(fop, fop->fd, EC_QUERY_INFO, fop->offset, LLONG_MAX);
- ec_lock(fop);
+ case EC_STATE_LOCK:
+ ec_lock_prepare_fd(fop, fop->fd, EC_QUERY_INFO, fop->offset,
+ EC_RANGE_FULL);
+ ec_lock(fop);
- return EC_STATE_DISPATCH;
+ return EC_STATE_DISPATCH;
- case EC_STATE_DISPATCH:
- /* This shouldn't fail because we have the inode locked. */
- GF_ASSERT(ec_get_inode_size(fop, fop->locks[0].lock->loc.inode,
- &size));
+ case EC_STATE_DISPATCH:
+ /* This shouldn't fail because we have the inode locked. */
+ GF_ASSERT(
+ ec_get_inode_size(fop, fop->locks[0].lock->loc.inode, &size));
- if (fop->user_size >= size) {
- ec_fop_set_error(fop, ENXIO);
+ if (fop->user_size >= size) {
+ ec_fop_set_error(fop, ENXIO);
- return EC_STATE_REPORT;
- }
+ return EC_STATE_REPORT;
+ }
- ec_dispatch_one(fop);
+ ec_dispatch_one(fop);
- return EC_STATE_PREPARE_ANSWER;
+ return EC_STATE_PREPARE_ANSWER;
- case EC_STATE_PREPARE_ANSWER:
- if (ec_dispatch_one_retry(fop, &cbk)) {
- return EC_STATE_DISPATCH;
- }
- if ((cbk != NULL) && (cbk->op_ret >= 0)) {
- ec_t *ec = fop->xl->private;
+ case EC_STATE_PREPARE_ANSWER:
+ if (ec_dispatch_one_retry(fop, &cbk)) {
+ return EC_STATE_DISPATCH;
+ }
+ if ((cbk != NULL) && (cbk->op_ret >= 0)) {
+ ec_t *ec = fop->xl->private;
- /* This shouldn't fail because we have the inode locked. */
- GF_ASSERT(ec_get_inode_size(fop, fop->locks[0].lock->loc.inode,
- &size));
+ /* This shouldn't fail because we have the inode locked. */
+ GF_ASSERT(ec_get_inode_size(fop, fop->locks[0].lock->loc.inode,
+ &size));
- cbk->offset *= ec->fragments;
- if (cbk->offset < fop->user_size) {
- cbk->offset = fop->user_size;
- }
- if (cbk->offset > size) {
- cbk->offset = size;
+ cbk->offset *= ec->fragments;
+ if (cbk->offset < fop->user_size) {
+ cbk->offset = fop->user_size;
+ }
+ if (cbk->offset > size) {
+ cbk->offset = size;
+ }
}
- }
- return EC_STATE_REPORT;
+ return EC_STATE_REPORT;
- case EC_STATE_REPORT:
- cbk = fop->answer;
+ case EC_STATE_REPORT:
+ cbk = fop->answer;
- GF_ASSERT(cbk != NULL);
+ GF_ASSERT(cbk != NULL);
- if (fop->cbks.seek != NULL) {
- fop->cbks.seek(fop->req_frame, fop, fop->xl, cbk->op_ret,
- cbk->op_errno, cbk->offset, cbk->xdata);
- }
+ if (fop->cbks.seek != NULL) {
+ fop->cbks.seek(fop->req_frame, fop, fop->xl, cbk->op_ret,
+ cbk->op_errno, cbk->offset, cbk->xdata);
+ }
- return EC_STATE_LOCK_REUSE;
+ return EC_STATE_LOCK_REUSE;
- case -EC_STATE_INIT:
- case -EC_STATE_LOCK:
- case -EC_STATE_DISPATCH:
- case -EC_STATE_PREPARE_ANSWER:
- case -EC_STATE_REPORT:
- GF_ASSERT(fop->error != 0);
+ case -EC_STATE_INIT:
+ case -EC_STATE_LOCK:
+ case -EC_STATE_DISPATCH:
+ case -EC_STATE_PREPARE_ANSWER:
+ case -EC_STATE_REPORT:
+ GF_ASSERT(fop->error != 0);
- if (fop->cbks.seek != NULL) {
- fop->cbks.seek(fop->req_frame, fop, fop->xl, -1, fop->error, 0,
- NULL);
- }
+ if (fop->cbks.seek != NULL) {
+ fop->cbks.seek(fop->req_frame, fop, fop->xl, -1, fop->error, 0,
+ NULL);
+ }
- return EC_STATE_LOCK_REUSE;
+ return EC_STATE_LOCK_REUSE;
- case -EC_STATE_LOCK_REUSE:
- case EC_STATE_LOCK_REUSE:
- ec_lock_reuse(fop);
+ case -EC_STATE_LOCK_REUSE:
+ case EC_STATE_LOCK_REUSE:
+ ec_lock_reuse(fop);
- return EC_STATE_UNLOCK;
+ return EC_STATE_UNLOCK;
- case -EC_STATE_UNLOCK:
- case EC_STATE_UNLOCK:
- ec_unlock(fop);
+ case -EC_STATE_UNLOCK:
+ case EC_STATE_UNLOCK:
+ ec_unlock(fop);
- return EC_STATE_END;
+ return EC_STATE_END;
- default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, 0,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s", state,
- ec_fop_name(fop->id));
+ default:
+ gf_msg(fop->xl->name, GF_LOG_ERROR, 0, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
- return EC_STATE_END;
+ return EC_STATE_END;
}
}
-void ec_seek(call_frame_t *frame, xlator_t *this, uintptr_t target,
- int32_t minimum, fop_seek_cbk_t func, void *data, fd_t *fd,
- off_t offset, gf_seek_what_t what, dict_t *xdata)
+void
+ec_seek(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_seek_cbk_t func, void *data, fd_t *fd,
+ off_t offset, gf_seek_what_t what, dict_t *xdata)
{
- ec_cbk_t callback = { .seek = func };
+ ec_cbk_t callback = {.seek = func};
ec_fop_data_t *fop = NULL;
int32_t error = EIO;
- gf_msg_trace ("ec", 0, "EC(SEEK) %p", frame);
+ gf_msg_trace("ec", 0, "EC(SEEK) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_SEEK, EC_FLAG_LOCK_SHARED,
- target, minimum, ec_wind_seek,
- ec_manager_seek, callback, data);
+ target, fop_flags, ec_wind_seek, ec_manager_seek,
+ callback, data);
if (fop == NULL) {
goto out;
}
@@ -1720,13 +1707,13 @@ out:
/* FOP: stat */
-int32_t ec_combine_stat(ec_fop_data_t * fop, ec_cbk_data_t * dst,
- ec_cbk_data_t * src)
+int32_t
+ec_combine_stat(ec_fop_data_t *fop, ec_cbk_data_t *dst, ec_cbk_data_t *src)
{
if (!ec_iatt_combine(fop, dst->iatt, src->iatt, 1)) {
- gf_msg (fop->xl->name, GF_LOG_NOTICE, 0,
- EC_MSG_IATT_MISMATCH, "Mismatching iatt in "
- "answers of 'GF_FOP_STAT'");
+ gf_msg(fop->xl->name, GF_LOG_NOTICE, 0, EC_MSG_IATT_MISMATCH,
+ "Mismatching iatt in "
+ "answers of 'GF_FOP_STAT'");
return 0;
}
@@ -1734,12 +1721,12 @@ int32_t ec_combine_stat(ec_fop_data_t * fop, ec_cbk_data_t * dst,
return 1;
}
-int32_t ec_stat_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, struct iatt * buf,
- dict_t * xdata)
+int32_t
+ec_stat_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *buf, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -1749,28 +1736,23 @@ int32_t ec_stat_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_STAT, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (op_ret >= 0)
- {
- if (buf != NULL)
- {
+ if (cbk != NULL) {
+ if (op_ret >= 0) {
+ if (buf != NULL) {
cbk->iatt[0] = *buf;
}
}
- if (xdata != NULL)
- {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -1780,15 +1762,15 @@ int32_t ec_stat_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_stat(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_stat(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -1797,19 +1779,20 @@ void ec_wind_stat(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
&fop->loc[0], fop->xdata);
}
-int32_t ec_manager_stat(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_stat(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_LOCK:
if (fop->fd == NULL) {
ec_lock_prepare_inode(fop, &fop->loc[0], EC_QUERY_INFO, 0,
- LLONG_MAX);
+ EC_RANGE_FULL);
} else {
- ec_lock_prepare_fd(fop, fop->fd, EC_QUERY_INFO, 0, LLONG_MAX);
+ ec_lock_prepare_fd(fop, fop->fd, EC_QUERY_INFO, 0,
+ EC_RANGE_FULL);
}
ec_lock(fop);
@@ -1825,8 +1808,7 @@ int32_t ec_manager_stat(ec_fop_data_t * fop, int32_t state)
if (cbk != NULL) {
if (cbk->iatt[0].ia_type == IA_IFREG) {
- ec_iatt_rebuild(fop->xl->private, cbk->iatt, 1,
- cbk->count);
+ ec_iatt_rebuild(fop->xl->private, cbk->iatt, 1, cbk->count);
/* This shouldn't fail because we have the inode locked. */
GF_ASSERT(ec_get_inode_size(fop,
@@ -1842,18 +1824,13 @@ int32_t ec_manager_stat(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->id == GF_FOP_STAT)
- {
- if (fop->cbks.stat != NULL)
- {
+ if (fop->id == GF_FOP_STAT) {
+ if (fop->cbks.stat != NULL) {
fop->cbks.stat(fop->req_frame, fop, fop->xl, cbk->op_ret,
cbk->op_errno, &cbk->iatt[0], cbk->xdata);
}
- }
- else
- {
- if (fop->cbks.fstat != NULL)
- {
+ } else {
+ if (fop->cbks.fstat != NULL) {
fop->cbks.fstat(fop->req_frame, fop, fop->xl, cbk->op_ret,
cbk->op_errno, &cbk->iatt[0], cbk->xdata);
}
@@ -1868,18 +1845,13 @@ int32_t ec_manager_stat(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->id == GF_FOP_STAT)
- {
- if (fop->cbks.stat != NULL)
- {
- fop->cbks.stat(fop->req_frame, fop, fop->xl, -1,
- fop->error, NULL, NULL);
+ if (fop->id == GF_FOP_STAT) {
+ if (fop->cbks.stat != NULL) {
+ fop->cbks.stat(fop->req_frame, fop, fop->xl, -1, fop->error,
+ NULL, NULL);
}
- }
- else
- {
- if (fop->cbks.fstat != NULL)
- {
+ } else {
+ if (fop->cbks.fstat != NULL) {
fop->cbks.fstat(fop->req_frame, fop, fop->xl, -1,
fop->error, NULL, NULL);
}
@@ -1900,30 +1872,30 @@ int32_t ec_manager_stat(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE, "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_stat(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_stat_cbk_t func, void * data, loc_t * loc,
- dict_t * xdata)
+void
+ec_stat(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_stat_cbk_t func, void *data, loc_t *loc,
+ dict_t *xdata)
{
- ec_cbk_t callback = { .stat = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.stat = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(STAT) %p", frame);
+ gf_msg_trace("ec", 0, "EC(STAT) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_STAT, EC_FLAG_LOCK_SHARED,
- target, minimum, ec_wind_stat, ec_manager_stat,
+ target, fop_flags, ec_wind_stat, ec_manager_stat,
callback, data);
if (fop == NULL) {
goto out;
@@ -1931,8 +1903,8 @@ void ec_stat(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL, "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
@@ -1940,9 +1912,9 @@ void ec_stat(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -1960,12 +1932,12 @@ out:
/* FOP: fstat */
-int32_t ec_fstat_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, struct iatt * buf,
- dict_t * xdata)
+int32_t
+ec_fstat_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *buf, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -1975,28 +1947,23 @@ int32_t ec_fstat_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_FSTAT, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (op_ret >= 0)
- {
- if (buf != NULL)
- {
+ if (cbk != NULL) {
+ if (op_ret >= 0) {
+ if (buf != NULL) {
cbk->iatt[0] = *buf;
}
}
- if (xdata != NULL)
- {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -2006,15 +1973,15 @@ int32_t ec_fstat_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_fstat(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_fstat(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -2023,23 +1990,24 @@ void ec_wind_fstat(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->xdata);
}
-void ec_fstat(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fstat_cbk_t func, void * data, fd_t * fd,
- dict_t * xdata)
+void
+ec_fstat(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fstat_cbk_t func, void *data, fd_t *fd,
+ dict_t *xdata)
{
- ec_cbk_t callback = { .fstat = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.fstat = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(FSTAT) %p", frame);
+ gf_msg_trace("ec", 0, "EC(FSTAT) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_FSTAT, EC_FLAG_LOCK_SHARED,
- target, minimum, ec_wind_fstat, ec_manager_stat,
- callback, data);
+ target, fop_flags, ec_wind_fstat,
+ ec_manager_stat, callback, data);
if (fop == NULL) {
goto out;
}
@@ -2049,9 +2017,9 @@ void ec_fstat(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL, "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -2059,9 +2027,9 @@ void ec_fstat(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL, "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
diff --git a/xlators/cluster/ec/src/ec-inode-write.c b/xlators/cluster/ec/src/ec-inode-write.c
index bf45a867971..9b5fe2a7fdc 100644
--- a/xlators/cluster/ec/src/ec-inode-write.c
+++ b/xlators/cluster/ec/src/ec-inode-write.c
@@ -8,10 +8,6 @@
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "defaults.h"
-
-#include "ec.h"
#include "ec-messages.h"
#include "ec-helpers.h"
#include "ec-common.h"
@@ -21,56 +17,55 @@
#include "ec-mem-types.h"
int32_t
-ec_update_writev_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct iatt *prebuf, struct iatt *postbuf,
- dict_t *xdata)
+ec_update_writev_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- ec_fop_data_t *fop = cookie;
- ec_cbk_data_t *cbk = NULL;
+ ec_fop_data_t *fop = cookie;
+ ec_cbk_data_t *cbk = NULL;
ec_fop_data_t *parent = fop->parent;
- int i = 0;
+ int i = 0;
ec_trace("UPDATE_WRITEV_CBK", cookie, "ret=%d, errno=%d, parent-fop=%s",
- op_ret, op_errno, ec_fop_name (parent->id));
+ op_ret, op_errno, ec_fop_name(parent->id));
if (op_ret < 0) {
- ec_fop_set_error (parent, op_errno);
- goto out;
+ ec_fop_set_error(parent, op_errno);
+ goto out;
}
- cbk = ec_cbk_data_allocate (parent->frame, this, parent,
- parent->id, 0, op_ret, op_errno);
+ cbk = ec_cbk_data_allocate(parent->frame, this, parent, parent->id, 0,
+ op_ret, op_errno);
if (!cbk) {
- ec_fop_set_error (parent, ENOMEM);
- goto out;
+ ec_fop_set_error(parent, ENOMEM);
+ goto out;
}
if (xdata)
- cbk->xdata = dict_ref (xdata);
+ cbk->xdata = dict_ref(xdata);
if (prebuf)
- cbk->iatt[i++] = *prebuf;
+ cbk->iatt[i++] = *prebuf;
if (postbuf)
- cbk->iatt[i++] = *postbuf;
+ cbk->iatt[i++] = *postbuf;
- LOCK (&parent->lock);
+ LOCK(&parent->lock);
{
- parent->good &= fop->good;
+ parent->good &= fop->good;
- if (gf_bits_count (parent->good) < parent->minimum) {
- __ec_fop_set_error (parent, EIO);
- } else if (fop->error == 0 && parent->answer == NULL) {
- parent->answer = cbk;
- }
+ if (gf_bits_count(parent->good) < parent->minimum) {
+ __ec_fop_set_error(parent, EIO);
+ } else if (fop->error == 0 && parent->answer == NULL) {
+ parent->answer = cbk;
+ }
}
- UNLOCK (&parent->lock);
+ UNLOCK(&parent->lock);
out:
return 0;
}
-int32_t ec_update_write(ec_fop_data_t *fop, uintptr_t mask, off_t offset,
- size_t size)
+static int32_t
+ec_update_write(ec_fop_data_t *fop, uintptr_t mask, off_t offset, uint64_t size)
{
struct iobref *iobref = NULL;
struct iobuf *iobuf = NULL;
@@ -90,13 +85,14 @@ int32_t ec_update_write(ec_fop_data_t *fop, uintptr_t mask, off_t offset,
goto out;
}
+ if (fop->locks[0].lock)
+ ec_lock_update_good(fop->locks[0].lock, fop);
vector.iov_base = iobuf->ptr;
vector.iov_len = size;
memset(vector.iov_base, 0, vector.iov_len);
- ec_writev(fop->frame, fop->xl, mask, fop->minimum,
- ec_update_writev_cbk, NULL, fop->fd, &vector, 1,
- offset, 0, iobref, NULL);
+ ec_writev(fop->frame, fop->xl, mask, fop->minimum, ec_update_writev_cbk,
+ NULL, fop->fd, &vector, 1, offset, 0, iobref, NULL);
err = 0;
@@ -112,62 +108,63 @@ out:
}
int
-ec_inode_write_cbk (call_frame_t *frame, xlator_t *this, void *cookie,
- int op_ret, int op_errno, struct iatt *prestat,
- struct iatt *poststat, dict_t *xdata)
+ec_inode_write_cbk(call_frame_t *frame, xlator_t *this, void *cookie,
+ int op_ret, int op_errno, struct iatt *prestat,
+ struct iatt *poststat, dict_t *xdata)
{
- ec_fop_data_t *fop = NULL;
- ec_cbk_data_t *cbk = NULL;
- int i = 0;
- int idx = 0;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
+ int i = 0;
+ int idx = 0;
- VALIDATE_OR_GOTO (this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, frame->local, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
+ VALIDATE_OR_GOTO(this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame->local, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = frame->local;
- idx = (int32_t)(uintptr_t) cookie;
+ fop = frame->local;
+ idx = (int32_t)(uintptr_t)cookie;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
- cbk = ec_cbk_data_allocate (frame, this, fop, fop->id, idx, op_ret,
- op_errno);
- if (!cbk)
- goto out;
+ cbk = ec_cbk_data_allocate(frame, this, fop, fop->id, idx, op_ret,
+ op_errno);
+ if (!cbk)
+ goto out;
- if (op_ret < 0)
- goto out;
+ if (op_ret < 0)
+ goto out;
- if (xdata)
- cbk->xdata = dict_ref (xdata);
+ if (xdata)
+ cbk->xdata = dict_ref(xdata);
- if (prestat)
- cbk->iatt[i++] = *prestat;
+ if (prestat)
+ cbk->iatt[i++] = *prestat;
- if (poststat)
- cbk->iatt[i++] = *poststat;
+ if (poststat)
+ cbk->iatt[i++] = *poststat;
out:
- if (cbk)
- ec_combine (cbk, ec_combine_write);
+ if (cbk)
+ ec_combine(cbk, ec_combine_write);
- if (fop)
- ec_complete (fop);
- return 0;
+ if (fop)
+ ec_complete(fop);
+ return 0;
}
/* FOP: removexattr */
-int32_t ec_removexattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xdata)
+int32_t
+ec_removexattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- return ec_inode_write_cbk (frame, this, cookie, op_ret, op_errno,
- NULL, NULL, xdata);
+ return ec_inode_write_cbk(frame, this, cookie, op_ret, op_errno, NULL, NULL,
+ xdata);
}
-void ec_wind_removexattr(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_removexattr(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -177,130 +174,126 @@ void ec_wind_removexattr(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
}
void
-ec_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
+ec_xattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
- ec_fop_data_t *fop = cookie;
- switch (fop->id) {
+ ec_fop_data_t *fop = cookie;
+ switch (fop->id) {
case GF_FOP_SETXATTR:
- if (fop->cbks.setxattr) {
- fop->cbks.setxattr (frame, cookie, this, op_ret,
- op_errno, xdata);
- }
- break;
+ if (fop->cbks.setxattr) {
+ QUORUM_CBK(fop->cbks.setxattr, fop, frame, cookie, this, op_ret,
+ op_errno, xdata);
+ }
+ break;
case GF_FOP_REMOVEXATTR:
- if (fop->cbks.removexattr) {
- fop->cbks.removexattr (frame, cookie, this, op_ret,
- op_errno, xdata);
- }
- break;
+ if (fop->cbks.removexattr) {
+ QUORUM_CBK(fop->cbks.removexattr, fop, frame, cookie, this,
+ op_ret, op_errno, xdata);
+ }
+ break;
case GF_FOP_FSETXATTR:
- if (fop->cbks.fsetxattr) {
- fop->cbks.fsetxattr (frame, cookie, this, op_ret,
- op_errno, xdata);
- }
- break;
+ if (fop->cbks.fsetxattr) {
+ QUORUM_CBK(fop->cbks.fsetxattr, fop, frame, cookie, this,
+ op_ret, op_errno, xdata);
+ }
+ break;
case GF_FOP_FREMOVEXATTR:
- if (fop->cbks.fremovexattr) {
- fop->cbks.fremovexattr (frame, cookie, this, op_ret,
- op_errno, xdata);
- }
- break;
- }
+ if (fop->cbks.fremovexattr) {
+ QUORUM_CBK(fop->cbks.fremovexattr, fop, frame, cookie, this,
+ op_ret, op_errno, xdata);
+ }
+ break;
+ }
}
int32_t
-ec_manager_xattr (ec_fop_data_t *fop, int32_t state)
+ec_manager_xattr(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state) {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_LOCK:
- if (fop->fd == NULL) {
- ec_lock_prepare_inode(fop, &fop->loc[0],
- EC_UPDATE_META | EC_QUERY_INFO,
- 0, LLONG_MAX);
- } else {
- ec_lock_prepare_fd(fop, fop->fd,
- EC_UPDATE_META | EC_QUERY_INFO,
- 0, LLONG_MAX);
- }
- ec_lock(fop);
+ if (fop->fd == NULL) {
+ ec_lock_prepare_inode(fop, &fop->loc[0],
+ EC_UPDATE_META | EC_QUERY_INFO, 0,
+ EC_RANGE_FULL);
+ } else {
+ ec_lock_prepare_fd(fop, fop->fd, EC_UPDATE_META | EC_QUERY_INFO,
+ 0, EC_RANGE_FULL);
+ }
+ ec_lock(fop);
- return EC_STATE_DISPATCH;
+ return EC_STATE_DISPATCH;
case EC_STATE_DISPATCH:
- ec_dispatch_all(fop);
+ ec_dispatch_all(fop);
- return EC_STATE_PREPARE_ANSWER;
+ return EC_STATE_PREPARE_ANSWER;
case EC_STATE_PREPARE_ANSWER:
- ec_fop_prepare_answer(fop, _gf_false);
+ ec_fop_prepare_answer(fop, _gf_false);
- return EC_STATE_REPORT;
+ return EC_STATE_REPORT;
case EC_STATE_REPORT:
- cbk = fop->answer;
+ cbk = fop->answer;
- GF_ASSERT(cbk != NULL);
+ GF_ASSERT(cbk != NULL);
- ec_xattr_cbk (fop->req_frame, fop, fop->xl, cbk->op_ret,
- cbk->op_errno, cbk->xdata);
+ ec_xattr_cbk(fop->req_frame, fop, fop->xl, cbk->op_ret,
+ cbk->op_errno, cbk->xdata);
- return EC_STATE_LOCK_REUSE;
+ return EC_STATE_LOCK_REUSE;
case -EC_STATE_INIT:
case -EC_STATE_LOCK:
case -EC_STATE_DISPATCH:
case -EC_STATE_PREPARE_ANSWER:
case -EC_STATE_REPORT:
- GF_ASSERT(fop->error != 0);
+ GF_ASSERT(fop->error != 0);
- ec_xattr_cbk (fop->req_frame, fop, fop->xl, -1, fop->error,
- NULL);
+ ec_xattr_cbk(fop->req_frame, fop, fop->xl, -1, fop->error, NULL);
- return EC_STATE_LOCK_REUSE;
+ return EC_STATE_LOCK_REUSE;
case -EC_STATE_LOCK_REUSE:
case EC_STATE_LOCK_REUSE:
- ec_lock_reuse(fop);
+ ec_lock_reuse(fop);
- return EC_STATE_UNLOCK;
+ return EC_STATE_UNLOCK;
case -EC_STATE_UNLOCK:
case EC_STATE_UNLOCK:
- ec_unlock(fop);
+ ec_unlock(fop);
- return EC_STATE_END;
+ return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE,
- "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
- return EC_STATE_END;
- }
+ return EC_STATE_END;
+ }
}
void
-ec_removexattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
- int32_t minimum, fop_removexattr_cbk_t func, void *data,
- loc_t *loc, const char *name, dict_t *xdata)
+ec_removexattr(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_removexattr_cbk_t func, void *data,
+ loc_t *loc, const char *name, dict_t *xdata)
{
- ec_cbk_t callback = { .removexattr = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.removexattr = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(REMOVEXATTR) %p", frame);
+ gf_msg_trace("ec", 0, "EC(REMOVEXATTR) %p", frame);
- VALIDATE_OR_GOTO (this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
+ VALIDATE_OR_GOTO(this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_REMOVEXATTR, 0, target,
- minimum, ec_wind_removexattr, ec_manager_xattr,
+ fop_flags, ec_wind_removexattr, ec_manager_xattr,
callback, data);
if (fop == NULL) {
goto out;
@@ -308,9 +301,8 @@ ec_removexattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL,
- "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
@@ -318,20 +310,18 @@ ec_removexattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
if (name != NULL) {
fop->str[0] = gf_strdup(name);
if (fop->str[0] == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY,
- "Failed to duplicate a string.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to duplicate a string.");
goto out;
}
}
if (xdata != NULL) {
- fop->xdata = dict_copy_with_ref (xdata, NULL);
+ fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -341,23 +331,24 @@ ec_removexattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
out:
if (fop != NULL) {
- ec_manager (fop, error);
+ ec_manager(fop, error);
} else {
- func (frame, NULL, this, -1, error, NULL);
+ func(frame, NULL, this, -1, error, NULL);
}
}
/* FOP: fremovexattr */
-int32_t ec_fremovexattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xdata)
+int32_t
+ec_fremovexattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- return ec_inode_write_cbk (frame, this, cookie, op_ret, op_errno,
- NULL, NULL, xdata);
+ return ec_inode_write_cbk(frame, this, cookie, op_ret, op_errno, NULL, NULL,
+ xdata);
}
-void ec_wind_fremovexattr(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_fremovexattr(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -367,23 +358,23 @@ void ec_wind_fremovexattr(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
}
void
-ec_fremovexattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
- int32_t minimum, fop_fremovexattr_cbk_t func, void *data,
- fd_t *fd, const char *name, dict_t *xdata)
+ec_fremovexattr(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fremovexattr_cbk_t func, void *data,
+ fd_t *fd, const char *name, dict_t *xdata)
{
- ec_cbk_t callback = { .fremovexattr = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.fremovexattr = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(FREMOVEXATTR) %p", frame);
+ gf_msg_trace("ec", 0, "EC(FREMOVEXATTR) %p", frame);
- VALIDATE_OR_GOTO (this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
+ VALIDATE_OR_GOTO(this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_FREMOVEXATTR, 0, target,
- minimum, ec_wind_fremovexattr, ec_manager_xattr,
- callback, data);
+ fop_flags, ec_wind_fremovexattr,
+ ec_manager_xattr, callback, data);
if (fop == NULL) {
goto out;
}
@@ -393,10 +384,9 @@ ec_fremovexattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL,
- "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -404,9 +394,8 @@ ec_fremovexattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
if (name != NULL) {
fop->str[0] = gf_strdup(name);
if (fop->str[0] == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY,
- "Failed to duplicate a string.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to duplicate a string.");
goto out;
}
@@ -414,10 +403,9 @@ ec_fremovexattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -427,24 +415,25 @@ ec_fremovexattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
out:
if (fop != NULL) {
- ec_manager (fop, error);
+ ec_manager(fop, error);
} else {
- func (frame, NULL, this, -1, error, NULL);
+ func(frame, NULL, this, -1, error, NULL);
}
}
/* FOP: setattr */
-int32_t ec_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *prestat, struct iatt *poststat,
- dict_t *xdata)
+int32_t
+ec_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prestat,
+ struct iatt *poststat, dict_t *xdata)
{
- return ec_inode_write_cbk (frame, this, cookie, op_ret, op_errno,
- prestat, poststat, xdata);
+ return ec_inode_write_cbk(frame, this, cookie, op_ret, op_errno, prestat,
+ poststat, xdata);
}
-void ec_wind_setattr(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_setattr(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -453,21 +442,21 @@ void ec_wind_setattr(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
&fop->loc[0], &fop->iatt, fop->int32, fop->xdata);
}
-int32_t ec_manager_setattr(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_setattr(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_LOCK:
if (fop->fd == NULL) {
ec_lock_prepare_inode(fop, &fop->loc[0],
- EC_UPDATE_META | EC_QUERY_INFO,
- 0, LLONG_MAX);
+ EC_UPDATE_META | EC_QUERY_INFO, 0,
+ EC_RANGE_FULL);
} else {
ec_lock_prepare_fd(fop, fop->fd, EC_UPDATE_META | EC_QUERY_INFO,
- 0, LLONG_MAX);
+ 0, EC_RANGE_FULL);
}
ec_lock(fop);
@@ -482,8 +471,7 @@ int32_t ec_manager_setattr(ec_fop_data_t * fop, int32_t state)
cbk = ec_fop_prepare_answer(fop, _gf_false);
if (cbk != NULL) {
if (cbk->iatt[0].ia_type == IA_IFREG) {
- ec_iatt_rebuild(fop->xl->private, cbk->iatt, 2,
- cbk->count);
+ ec_iatt_rebuild(fop->xl->private, cbk->iatt, 2, cbk->count);
/* This shouldn't fail because we have the inode locked. */
GF_ASSERT(ec_get_inode_size(fop,
@@ -500,24 +488,17 @@ int32_t ec_manager_setattr(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->id == GF_FOP_SETATTR)
- {
- if (fop->cbks.setattr != NULL)
- {
- fop->cbks.setattr(fop->req_frame, fop, fop->xl,
- cbk->op_ret, cbk->op_errno,
- &cbk->iatt[0], &cbk->iatt[1],
- cbk->xdata);
+ if (fop->id == GF_FOP_SETATTR) {
+ if (fop->cbks.setattr != NULL) {
+ QUORUM_CBK(fop->cbks.setattr, fop, fop->req_frame, fop,
+ fop->xl, cbk->op_ret, cbk->op_errno,
+ &cbk->iatt[0], &cbk->iatt[1], cbk->xdata);
}
- }
- else
- {
- if (fop->cbks.fsetattr != NULL)
- {
- fop->cbks.fsetattr(fop->req_frame, fop, fop->xl,
- cbk->op_ret, cbk->op_errno,
- &cbk->iatt[0], &cbk->iatt[1],
- cbk->xdata);
+ } else {
+ if (fop->cbks.fsetattr != NULL) {
+ QUORUM_CBK(fop->cbks.fsetattr, fop, fop->req_frame, fop,
+ fop->xl, cbk->op_ret, cbk->op_errno,
+ &cbk->iatt[0], &cbk->iatt[1], cbk->xdata);
}
}
@@ -530,18 +511,13 @@ int32_t ec_manager_setattr(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->id == GF_FOP_SETATTR)
- {
- if (fop->cbks.setattr != NULL)
- {
+ if (fop->id == GF_FOP_SETATTR) {
+ if (fop->cbks.setattr != NULL) {
fop->cbks.setattr(fop->req_frame, fop, fop->xl, -1,
fop->error, NULL, NULL, NULL);
}
- }
- else
- {
- if (fop->cbks.fsetattr != NULL)
- {
+ } else {
+ if (fop->cbks.fsetattr != NULL) {
fop->cbks.fsetattr(fop->req_frame, fop, fop->xl, -1,
fop->error, NULL, NULL, NULL);
}
@@ -562,33 +538,31 @@ int32_t ec_manager_setattr(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE,
- "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_setattr(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_setattr_cbk_t func, void * data,
- loc_t * loc, struct iatt * stbuf, int32_t valid,
- dict_t * xdata)
+void
+ec_setattr(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_setattr_cbk_t func, void *data, loc_t *loc,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- ec_cbk_t callback = { .setattr = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.setattr = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(SETATTR) %p", frame);
+ gf_msg_trace("ec", 0, "EC(SETATTR) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_SETATTR, 0, target, minimum,
- ec_wind_setattr, ec_manager_setattr, callback,
- data);
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_SETATTR, 0, target,
+ fop_flags, ec_wind_setattr, ec_manager_setattr,
+ callback, data);
if (fop == NULL) {
goto out;
}
@@ -597,9 +571,8 @@ void ec_setattr(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL,
- "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
@@ -610,10 +583,9 @@ void ec_setattr(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -631,16 +603,17 @@ out:
/* FOP: fsetattr */
-int32_t ec_fsetattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *prestat, struct iatt *poststat,
- dict_t *xdata)
+int32_t
+ec_fsetattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prestat,
+ struct iatt *poststat, dict_t *xdata)
{
- return ec_inode_write_cbk (frame, this, cookie, op_ret, op_errno,
- prestat, poststat, xdata);
+ return ec_inode_write_cbk(frame, this, cookie, op_ret, op_errno, prestat,
+ poststat, xdata);
}
-void ec_wind_fsetattr(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_fsetattr(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -649,22 +622,23 @@ void ec_wind_fsetattr(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->fd, &fop->iatt, fop->int32, fop->xdata);
}
-void ec_fsetattr(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fsetattr_cbk_t func, void * data,
- fd_t * fd, struct iatt * stbuf, int32_t valid, dict_t * xdata)
+void
+ec_fsetattr(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fsetattr_cbk_t func, void *data, fd_t *fd,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- ec_cbk_t callback = { .fsetattr = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.fsetattr = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(FSETATTR) %p", frame);
+ gf_msg_trace("ec", 0, "EC(FSETATTR) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_FSETATTR, 0, target,
- minimum, ec_wind_fsetattr, ec_manager_setattr,
+ fop_flags, ec_wind_fsetattr, ec_manager_setattr,
callback, data);
if (fop == NULL) {
goto out;
@@ -677,10 +651,9 @@ void ec_fsetattr(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL,
- "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -691,10 +664,9 @@ void ec_fsetattr(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -712,14 +684,16 @@ out:
/* FOP: setxattr */
-int32_t ec_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+int32_t
+ec_setxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- return ec_inode_write_cbk (frame, this, cookie, op_ret, op_errno,
- NULL, NULL, xdata);
+ return ec_inode_write_cbk(frame, this, cookie, op_ret, op_errno, NULL, NULL,
+ xdata);
}
-void ec_wind_setxattr(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_setxattr(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -729,22 +703,22 @@ void ec_wind_setxattr(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
}
void
-ec_setxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
- int32_t minimum, fop_setxattr_cbk_t func, void *data,
- loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata)
+ec_setxattr(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_setxattr_cbk_t func, void *data, loc_t *loc,
+ dict_t *dict, int32_t flags, dict_t *xdata)
{
- ec_cbk_t callback = { .setxattr = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.setxattr = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(SETXATTR) %p", frame);
+ gf_msg_trace("ec", 0, "EC(SETXATTR) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_SETXATTR, 0, target,
- minimum, ec_wind_setxattr, ec_manager_xattr,
+ fop_flags, ec_wind_setxattr, ec_manager_xattr,
callback, data);
if (fop == NULL) {
goto out;
@@ -754,9 +728,8 @@ ec_setxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL,
- "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
@@ -764,10 +737,9 @@ ec_setxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
if (dict != NULL) {
fop->dict = dict_copy_with_ref(dict, NULL);
if (fop->dict == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -775,10 +747,9 @@ ec_setxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -788,45 +759,41 @@ ec_setxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
out:
if (fop != NULL) {
- ec_manager (fop, error);
+ ec_manager(fop, error);
} else {
- func (frame, NULL, this, -1, error, NULL);
+ func(frame, NULL, this, -1, error, NULL);
}
}
/* FOP: fsetxattr */
int32_t
-ec_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+ec_fsetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
- VALIDATE_OR_GOTO (this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, frame->local, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
+ VALIDATE_OR_GOTO(this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame->local, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_FSETXATTR, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (xdata != NULL)
- {
+ if (cbk != NULL) {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -836,15 +803,15 @@ ec_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_fsetxattr(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_fsetxattr(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -854,22 +821,22 @@ void ec_wind_fsetxattr(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
}
void
-ec_fsetxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
- int32_t minimum, fop_fsetxattr_cbk_t func, void *data,
- fd_t *fd, dict_t *dict, int32_t flags, dict_t *xdata)
+ec_fsetxattr(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fsetxattr_cbk_t func, void *data, fd_t *fd,
+ dict_t *dict, int32_t flags, dict_t *xdata)
{
- ec_cbk_t callback = { .fsetxattr = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.fsetxattr = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(FSETXATTR) %p", frame);
+ gf_msg_trace("ec", 0, "EC(FSETXATTR) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_FSETXATTR, 0, target,
- minimum, ec_wind_fsetxattr, ec_manager_xattr,
+ fop_flags, ec_wind_fsetxattr, ec_manager_xattr,
callback, data);
if (fop == NULL) {
goto out;
@@ -882,10 +849,9 @@ ec_fsetxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL,
- "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -893,10 +859,9 @@ ec_fsetxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
if (dict != NULL) {
fop->dict = dict_copy_with_ref(dict, NULL);
if (fop->dict == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -904,10 +869,9 @@ ec_fsetxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -917,9 +881,9 @@ ec_fsetxattr (call_frame_t *frame, xlator_t *this, uintptr_t target,
out:
if (fop != NULL) {
- ec_manager (fop, error);
+ ec_manager(fop, error);
} else {
- func (frame, NULL, this, -1, error, NULL);
+ func(frame, NULL, this, -1, error, NULL);
}
}
@@ -929,162 +893,161 @@ out:
*
*********************************************************************/
-int32_t ec_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+int32_t
+ec_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- return ec_inode_write_cbk (frame, this, cookie, op_ret, op_errno,
- prebuf, postbuf, xdata);
+ return ec_inode_write_cbk(frame, this, cookie, op_ret, op_errno, prebuf,
+ postbuf, xdata);
}
-void ec_wind_fallocate(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
+void
+ec_wind_fallocate(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
STACK_WIND_COOKIE(fop->frame, ec_fallocate_cbk, (void *)(uintptr_t)idx,
ec->xl_list[idx], ec->xl_list[idx]->fops->fallocate,
- fop->fd, fop->int32, fop->offset,
- fop->size, fop->xdata);
+ fop->fd, fop->int32, fop->offset, fop->size, fop->xdata);
}
-int32_t ec_manager_fallocate(ec_fop_data_t *fop, int32_t state)
+int32_t
+ec_manager_fallocate(ec_fop_data_t *fop, int32_t state)
{
ec_cbk_data_t *cbk = NULL;
switch (state) {
- case EC_STATE_INIT:
- if (fop->size == 0) {
+ case EC_STATE_INIT:
+ if (fop->size == 0) {
ec_fop_set_error(fop, EINVAL);
return EC_STATE_REPORT;
- }
- if (fop->int32 & (FALLOC_FL_COLLAPSE_RANGE
- |FALLOC_FL_INSERT_RANGE
- |FALLOC_FL_ZERO_RANGE
- |FALLOC_FL_PUNCH_HOLE)) {
+ }
+ if (fop->int32 &
+ (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE |
+ FALLOC_FL_ZERO_RANGE | FALLOC_FL_PUNCH_HOLE)) {
ec_fop_set_error(fop, ENOTSUP);
return EC_STATE_REPORT;
- }
- fop->user_size = fop->offset + fop->size;
- fop->head = ec_adjust_offset_down (fop->xl->private, &fop->offset,
- _gf_true);
- fop->size += fop->head;
- ec_adjust_size_up (fop->xl->private, &fop->size, _gf_true);
+ }
+ fop->user_size = fop->offset + fop->size;
+ fop->head = ec_adjust_offset_down(fop->xl->private, &fop->offset,
+ _gf_true);
+ fop->size += fop->head;
+ ec_adjust_size_up(fop->xl->private, &fop->size, _gf_true);
- /* Fall through */
+ /* Fall through */
- case EC_STATE_LOCK:
- ec_lock_prepare_fd(fop, fop->fd,
- EC_UPDATE_DATA | EC_UPDATE_META | EC_QUERY_INFO,
- fop->offset, fop->size);
- ec_lock(fop);
+ case EC_STATE_LOCK:
+ ec_lock_prepare_fd(fop, fop->fd,
+ EC_UPDATE_DATA | EC_UPDATE_META | EC_QUERY_INFO,
+ fop->offset, fop->size);
+ ec_lock(fop);
- return EC_STATE_DISPATCH;
+ return EC_STATE_DISPATCH;
- case EC_STATE_DISPATCH:
+ case EC_STATE_DISPATCH:
- ec_dispatch_all(fop);
+ ec_dispatch_all(fop);
- return EC_STATE_PREPARE_ANSWER;
+ return EC_STATE_PREPARE_ANSWER;
- case EC_STATE_PREPARE_ANSWER:
- cbk = ec_fop_prepare_answer(fop, _gf_false);
- if (cbk != NULL) {
- ec_iatt_rebuild(fop->xl->private, cbk->iatt, 2,
- cbk->count);
+ case EC_STATE_PREPARE_ANSWER:
+ cbk = ec_fop_prepare_answer(fop, _gf_false);
+ if (cbk != NULL) {
+ ec_iatt_rebuild(fop->xl->private, cbk->iatt, 2, cbk->count);
/* This shouldn't fail because we have the inode locked. */
LOCK(&fop->locks[0].lock->loc.inode->lock);
{
- GF_ASSERT(__ec_get_inode_size(fop,
- fop->locks[0].lock->loc.inode,
- &cbk->iatt[0].ia_size));
+ GF_ASSERT(__ec_get_inode_size(fop,
+ fop->locks[0].lock->loc.inode,
+ &cbk->iatt[0].ia_size));
- /*If mode has FALLOC_FL_KEEP_SIZE keep the size */
- if (fop->int32 & FALLOC_FL_KEEP_SIZE) {
- cbk->iatt[1].ia_size = cbk->iatt[0].ia_size;
- } else if (fop->user_size > cbk->iatt[0].ia_size) {
- cbk->iatt[1].ia_size = fop->user_size;
-
- /* This shouldn't fail because we have the inode
- * locked. */
- GF_ASSERT(__ec_set_inode_size(fop,
- fop->locks[0].lock->loc.inode,
- cbk->iatt[1].ia_size));
- } else {
- cbk->iatt[1].ia_size = cbk->iatt[0].ia_size;
- }
+ /*If mode has FALLOC_FL_KEEP_SIZE keep the size */
+ if (fop->int32 & FALLOC_FL_KEEP_SIZE) {
+ cbk->iatt[1].ia_size = cbk->iatt[0].ia_size;
+ } else if (fop->user_size > cbk->iatt[0].ia_size) {
+ cbk->iatt[1].ia_size = fop->user_size;
+
+ /* This shouldn't fail because we have the inode
+ * locked. */
+ GF_ASSERT(__ec_set_inode_size(
+ fop, fop->locks[0].lock->loc.inode,
+ cbk->iatt[1].ia_size));
+ } else {
+ cbk->iatt[1].ia_size = cbk->iatt[0].ia_size;
+ }
}
UNLOCK(&fop->locks[0].lock->loc.inode->lock);
- }
+ }
- return EC_STATE_REPORT;
+ return EC_STATE_REPORT;
- case EC_STATE_REPORT:
- cbk = fop->answer;
+ case EC_STATE_REPORT:
+ cbk = fop->answer;
- GF_ASSERT(cbk != NULL);
+ GF_ASSERT(cbk != NULL);
- if (fop->cbks.fallocate != NULL) {
- fop->cbks.fallocate(fop->req_frame, fop, fop->xl, cbk->op_ret,
- cbk->op_errno, &cbk->iatt[0], &cbk->iatt[1],
- cbk->xdata);
- }
+ if (fop->cbks.fallocate != NULL) {
+ QUORUM_CBK(fop->cbks.fallocate, fop, fop->req_frame, fop,
+ fop->xl, cbk->op_ret, cbk->op_errno, &cbk->iatt[0],
+ &cbk->iatt[1], cbk->xdata);
+ }
- return EC_STATE_LOCK_REUSE;
+ return EC_STATE_LOCK_REUSE;
- case -EC_STATE_INIT:
- case -EC_STATE_LOCK:
- case -EC_STATE_DISPATCH:
- case -EC_STATE_PREPARE_ANSWER:
- case -EC_STATE_REPORT:
- GF_ASSERT(fop->error != 0);
+ case -EC_STATE_INIT:
+ case -EC_STATE_LOCK:
+ case -EC_STATE_DISPATCH:
+ case -EC_STATE_PREPARE_ANSWER:
+ case -EC_STATE_REPORT:
+ GF_ASSERT(fop->error != 0);
- if (fop->cbks.fallocate != NULL) {
+ if (fop->cbks.fallocate != NULL) {
fop->cbks.fallocate(fop->req_frame, fop, fop->xl, -1,
fop->error, NULL, NULL, NULL);
- }
+ }
- return EC_STATE_LOCK_REUSE;
+ return EC_STATE_LOCK_REUSE;
- case -EC_STATE_LOCK_REUSE:
- case EC_STATE_LOCK_REUSE:
- ec_lock_reuse(fop);
+ case -EC_STATE_LOCK_REUSE:
+ case EC_STATE_LOCK_REUSE:
+ ec_lock_reuse(fop);
- return EC_STATE_UNLOCK;
+ return EC_STATE_UNLOCK;
- case -EC_STATE_UNLOCK:
- case EC_STATE_UNLOCK:
- ec_unlock(fop);
+ case -EC_STATE_UNLOCK:
+ case EC_STATE_UNLOCK:
+ ec_unlock(fop);
- return EC_STATE_END;
+ return EC_STATE_END;
- default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE,
- "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ default:
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
- return EC_STATE_END;
+ return EC_STATE_END;
}
}
-void ec_fallocate(call_frame_t *frame, xlator_t *this, uintptr_t target,
- int32_t minimum, fop_fallocate_cbk_t func, void *data, fd_t *fd,
- int32_t mode, off_t offset, size_t len, dict_t *xdata)
+void
+ec_fallocate(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fallocate_cbk_t func, void *data, fd_t *fd,
+ int32_t mode, off_t offset, size_t len, dict_t *xdata)
{
- ec_cbk_t callback = { .fallocate = func };
+ ec_cbk_t callback = {.fallocate = func};
ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(FALLOCATE) %p", frame);
+ gf_msg_trace("ec", 0, "EC(FALLOCATE) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_FALLOCATE, 0, target,
- minimum, ec_wind_fallocate, ec_manager_fallocate,
- callback, data);
+ fop_flags, ec_wind_fallocate,
+ ec_manager_fallocate, callback, data);
if (fop == NULL) {
goto out;
}
@@ -1097,22 +1060,20 @@ void ec_fallocate(call_frame_t *frame, xlator_t *this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL,
- "Failed to reference a "
- "file descriptor.");
- goto out;
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
+ goto out;
}
}
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
- goto out;
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
+ goto out;
}
}
@@ -1131,66 +1092,70 @@ out:
* File Operation : Discard
*
*********************************************************************/
-void ec_update_discard_write(ec_fop_data_t *fop, uintptr_t mask)
+void
+ec_update_discard_write(ec_fop_data_t *fop, uintptr_t mask)
{
- ec_t *ec = fop->xl->private;
- off_t off_head = 0;
- off_t off_tail = 0;
- size_t size_head = 0;
- size_t size_tail = 0;
- int error = 0;
+ ec_t *ec = fop->xl->private;
+ off_t off_head = 0;
+ off_t off_tail = 0;
+ uint64_t size_head = 0;
+ uint64_t size_tail = 0;
+ int error = 0;
off_head = fop->offset * ec->fragments - fop->int32;
if (fop->size == 0) {
- error = ec_update_write (fop, mask, off_head, fop->user_size);
+ error = ec_update_write(fop, mask, off_head, fop->user_size);
} else {
- size_head = fop->int32;
- size_tail = (off_head + fop->user_size) % ec->stripe_size;
- off_tail = off_head + fop->user_size - size_tail;
- if (size_head) {
- error = ec_update_write (fop, mask, off_head, size_head);
- if (error) {
- goto out;
- }
- }
- if (size_tail) {
- error = ec_update_write (fop, mask, off_tail, size_tail);
+ size_head = fop->int32;
+ size_tail = (off_head + fop->user_size) % ec->stripe_size;
+ off_tail = off_head + fop->user_size - size_tail;
+ if (size_head) {
+ error = ec_update_write(fop, mask, off_head, size_head);
+ if (error) {
+ goto out;
}
+ }
+ if (size_tail) {
+ error = ec_update_write(fop, mask, off_tail, size_tail);
+ }
}
out:
if (error)
- ec_fop_set_error (fop, -error);
+ ec_fop_set_error(fop, -error);
}
-void ec_discard_adjust_offset_size(ec_fop_data_t *fop)
+void
+ec_discard_adjust_offset_size(ec_fop_data_t *fop)
{
- ec_t *ec = fop->xl->private;
+ ec_t *ec = fop->xl->private;
- fop->user_size = fop->size;
- /* If discard length covers atleast a fragment on brick, we will
- * perform discard operation(when fop->size is non-zero) else we just
- * write zeros.
- */
- fop->int32 = ec_adjust_offset_up(ec, &fop->offset, _gf_true);
- fop->frag_range.first = fop->offset;
- if (fop->size < fop->int32) {
- fop->size = 0;
- } else {
- fop->size -= fop->int32;
- ec_adjust_size_down(ec, &fop->size, _gf_true);
- }
- fop->frag_range.last = fop->offset + fop->size;
+ fop->user_size = fop->size;
+ /* If discard length covers at least a fragment on brick, we will
+ * perform discard operation(when fop->size is non-zero) else we just
+ * write zeros.
+ */
+ fop->int32 = ec_adjust_offset_up(ec, &fop->offset, _gf_true);
+ fop->frag_range.first = fop->offset;
+ if (fop->size < fop->int32) {
+ fop->size = 0;
+ } else {
+ fop->size -= fop->int32;
+ ec_adjust_size_down(ec, &fop->size, _gf_true);
+ }
+ fop->frag_range.last = fop->offset + fop->size;
}
-int32_t ec_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+int32_t
+ec_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- return ec_inode_write_cbk (frame, this, cookie, op_ret, op_errno,
- prebuf, postbuf, xdata);
+ return ec_inode_write_cbk(frame, this, cookie, op_ret, op_errno, prebuf,
+ postbuf, xdata);
}
-void ec_wind_discard(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
+void
+ec_wind_discard(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -1199,145 +1164,143 @@ void ec_wind_discard(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
fop->fd, fop->offset, fop->size, fop->xdata);
}
-int32_t ec_manager_discard(ec_fop_data_t *fop, int32_t state)
+int32_t
+ec_manager_discard(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t *cbk = NULL;
- off_t fl_start = 0;
- size_t fl_size = 0;
-
+ ec_cbk_data_t *cbk = NULL;
+ off_t fl_start = 0;
+ uint64_t fl_size = 0;
switch (state) {
- case EC_STATE_INIT:
- if ((fop->size <= 0) || (fop->offset < 0)) {
+ case EC_STATE_INIT:
+ if ((fop->size <= 0) || (fop->offset < 0)) {
ec_fop_set_error(fop, EINVAL);
return EC_STATE_REPORT;
- }
- /* Because of the head/tail writes, "discard" happens on the remaining
- * regions, but we need to compute region including head/tail writes
- * so compute them separately*/
- fl_start = fop->offset;
- fl_size = fop->size;
- fl_size += ec_adjust_offset_down (fop->xl->private, &fl_start,
- _gf_true);
- ec_adjust_size_up (fop->xl->private, &fl_size, _gf_true);
+ }
+ /* Because of the head/tail writes, "discard" happens on the
+ * remaining regions, but we need to compute region including
+ * head/tail writes so compute them separately*/
+ fl_start = fop->offset;
+ fl_size = fop->size;
+ fl_size += ec_adjust_offset_down(fop->xl->private, &fl_start,
+ _gf_true);
+ ec_adjust_size_up(fop->xl->private, &fl_size, _gf_true);
- ec_discard_adjust_offset_size(fop);
+ ec_discard_adjust_offset_size(fop);
- /* Fall through */
+ /* Fall through */
- case EC_STATE_LOCK:
- ec_lock_prepare_fd(fop, fop->fd,
- EC_UPDATE_DATA | EC_UPDATE_META |
- EC_QUERY_INFO, fl_start, fl_size);
- ec_lock(fop);
+ case EC_STATE_LOCK:
+ ec_lock_prepare_fd(fop, fop->fd,
+ EC_UPDATE_DATA | EC_UPDATE_META | EC_QUERY_INFO,
+ fl_start, fl_size);
+ ec_lock(fop);
- return EC_STATE_DISPATCH;
+ return EC_STATE_DISPATCH;
- case EC_STATE_DISPATCH:
+ case EC_STATE_DISPATCH:
- /* Dispatch discard fop only if we have whole fragment
- * to deallocate */
- if (fop->size) {
+ /* Dispatch discard fop only if we have whole fragment
+ * to deallocate */
+ if (fop->size) {
ec_dispatch_all(fop);
return EC_STATE_DELAYED_START;
- } else {
- /*Assume discard to have succeeded on mask*/
- fop->good = fop->mask;
- }
+ } else {
+ /* Assume discard to have succeeded on all bricks */
+ ec_succeed_all(fop);
+ }
- /* Fall through */
+ /* Fall through */
- case EC_STATE_DELAYED_START:
+ case EC_STATE_DELAYED_START:
- if (fop->size) {
+ if (fop->size) {
if (fop->answer && fop->answer->op_ret == 0)
- ec_update_discard_write (fop, fop->answer->mask);
- } else {
- ec_update_discard_write (fop, fop->mask);
- }
+ ec_update_discard_write(fop, fop->answer->mask);
+ } else {
+ ec_update_discard_write(fop, fop->mask);
+ }
- return EC_STATE_PREPARE_ANSWER;
+ return EC_STATE_PREPARE_ANSWER;
- case EC_STATE_PREPARE_ANSWER:
- cbk = ec_fop_prepare_answer(fop, _gf_false);
- if (cbk != NULL) {
- ec_iatt_rebuild(fop->xl->private, cbk->iatt, 2,
- cbk->count);
+ case EC_STATE_PREPARE_ANSWER:
+ cbk = ec_fop_prepare_answer(fop, _gf_false);
+ if (cbk != NULL) {
+ ec_iatt_rebuild(fop->xl->private, cbk->iatt, 2, cbk->count);
/* This shouldn't fail because we have the inode locked. */
GF_ASSERT(ec_get_inode_size(fop, fop->locks[0].lock->loc.inode,
&cbk->iatt[0].ia_size));
cbk->iatt[1].ia_size = cbk->iatt[0].ia_size;
- }
- return EC_STATE_REPORT;
+ }
+ return EC_STATE_REPORT;
- case EC_STATE_REPORT:
- cbk = fop->answer;
+ case EC_STATE_REPORT:
+ cbk = fop->answer;
- GF_ASSERT(cbk != NULL);
+ GF_ASSERT(cbk != NULL);
- if (fop->cbks.discard != NULL) {
- fop->cbks.discard(fop->req_frame, fop, fop->xl, cbk->op_ret,
- cbk->op_errno, &cbk->iatt[0], &cbk->iatt[1],
- cbk->xdata);
- }
+ if (fop->cbks.discard != NULL) {
+ QUORUM_CBK(fop->cbks.discard, fop, fop->req_frame, fop, fop->xl,
+ cbk->op_ret, cbk->op_errno, &cbk->iatt[0],
+ &cbk->iatt[1], cbk->xdata);
+ }
- return EC_STATE_LOCK_REUSE;
+ return EC_STATE_LOCK_REUSE;
- case -EC_STATE_INIT:
- case -EC_STATE_LOCK:
- case -EC_STATE_DISPATCH:
- case -EC_STATE_DELAYED_START:
- case -EC_STATE_PREPARE_ANSWER:
- case -EC_STATE_REPORT:
- GF_ASSERT(fop->error != 0);
+ case -EC_STATE_INIT:
+ case -EC_STATE_LOCK:
+ case -EC_STATE_DISPATCH:
+ case -EC_STATE_DELAYED_START:
+ case -EC_STATE_PREPARE_ANSWER:
+ case -EC_STATE_REPORT:
+ GF_ASSERT(fop->error != 0);
- if (fop->cbks.discard != NULL) {
- fop->cbks.discard(fop->req_frame, fop, fop->xl, -1,
- fop->error, NULL, NULL, NULL);
- }
+ if (fop->cbks.discard != NULL) {
+ fop->cbks.discard(fop->req_frame, fop, fop->xl, -1, fop->error,
+ NULL, NULL, NULL);
+ }
- return EC_STATE_LOCK_REUSE;
+ return EC_STATE_LOCK_REUSE;
- case -EC_STATE_LOCK_REUSE:
- case EC_STATE_LOCK_REUSE:
- ec_lock_reuse(fop);
+ case -EC_STATE_LOCK_REUSE:
+ case EC_STATE_LOCK_REUSE:
+ ec_lock_reuse(fop);
- return EC_STATE_UNLOCK;
+ return EC_STATE_UNLOCK;
- case -EC_STATE_UNLOCK:
- case EC_STATE_UNLOCK:
- ec_unlock(fop);
+ case -EC_STATE_UNLOCK:
+ case EC_STATE_UNLOCK:
+ ec_unlock(fop);
- return EC_STATE_END;
+ return EC_STATE_END;
- default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE,
- "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ default:
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
- return EC_STATE_END;
+ return EC_STATE_END;
}
}
-void ec_discard(call_frame_t *frame, xlator_t *this, uintptr_t target,
- int32_t minimum, fop_discard_cbk_t func, void *data, fd_t *fd,
- off_t offset, size_t len, dict_t *xdata)
+void
+ec_discard(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_discard_cbk_t func, void *data, fd_t *fd,
+ off_t offset, size_t len, dict_t *xdata)
{
- ec_cbk_t callback = { .discard = func };
+ ec_cbk_t callback = {.discard = func};
ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(DISCARD) %p", frame);
+ gf_msg_trace("ec", 0, "EC(DISCARD) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_DISCARD, 0, target,
- minimum, ec_wind_discard, ec_manager_discard,
+ fop_flags, ec_wind_discard, ec_manager_discard,
callback, data);
if (fop == NULL) {
goto out;
@@ -1371,33 +1334,35 @@ out:
*
*********************************************************************/
-int32_t ec_update_truncate_write (ec_fop_data_t *fop, uintptr_t mask)
+int32_t
+ec_update_truncate_write(ec_fop_data_t *fop, uintptr_t mask)
{
- ec_t *ec = fop->xl->private;
- size_t size = fop->offset * ec->fragments - fop->user_size;
- return ec_update_write (fop, mask, fop->user_size, size);
+ ec_t *ec = fop->xl->private;
+ uint64_t size = fop->offset * ec->fragments - fop->user_size;
+ return ec_update_write(fop, mask, fop->user_size, size);
}
-int32_t ec_truncate_open_cbk(call_frame_t * frame, void * cookie,
- xlator_t * this, int32_t op_ret, int32_t op_errno,
- fd_t * fd, dict_t * xdata)
+int32_t
+ec_truncate_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- ec_fop_data_t * fop = cookie;
+ ec_fop_data_t *fop = cookie;
int32_t err;
fop->parent->good &= fop->good;
if (op_ret >= 0) {
- fd_bind (fd);
- err = ec_update_truncate_write (fop->parent, fop->answer->mask);
+ fd_bind(fd);
+ err = ec_update_truncate_write(fop->parent, fop->answer->mask);
if (err != 0) {
- ec_fop_set_error (fop->parent, -err);
+ ec_fop_set_error(fop->parent, -err);
}
}
return 0;
}
-int32_t ec_truncate_clean(ec_fop_data_t * fop)
+int32_t
+ec_truncate_clean(ec_fop_data_t *fop)
{
if (fop->fd == NULL) {
fop->fd = fd_create(fop->loc[0].inode, fop->frame->root->pid);
@@ -1406,24 +1371,25 @@ int32_t ec_truncate_clean(ec_fop_data_t * fop)
}
ec_open(fop->frame, fop->xl, fop->answer->mask, fop->minimum,
- ec_truncate_open_cbk, fop, &fop->loc[0], O_RDWR, fop->fd,
- NULL);
+ ec_truncate_open_cbk, fop, &fop->loc[0], O_RDWR, fop->fd, NULL);
return 0;
} else {
- return ec_update_truncate_write (fop, fop->answer->mask);
+ return ec_update_truncate_write(fop, fop->answer->mask);
}
}
-int32_t ec_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prestat,
- struct iatt *poststat, dict_t *xdata)
+int32_t
+ec_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prestat,
+ struct iatt *poststat, dict_t *xdata)
{
- return ec_inode_write_cbk (frame, this, cookie, op_ret, op_errno,
- prestat, poststat, xdata);
+ return ec_inode_write_cbk(frame, this, cookie, op_ret, op_errno, prestat,
+ poststat, xdata);
}
-void ec_wind_truncate(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_truncate(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -1432,29 +1398,35 @@ void ec_wind_truncate(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
&fop->loc[0], fop->offset, fop->xdata);
}
-int32_t ec_manager_truncate(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_truncate(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
+ off_t offset_down;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
fop->user_size = fop->offset;
ec_adjust_offset_up(fop->xl->private, &fop->offset, _gf_true);
fop->frag_range.first = fop->offset;
fop->frag_range.last = UINT64_MAX;
- /* Fall through */
+ /* Fall through */
case EC_STATE_LOCK:
+ offset_down = fop->user_size;
+ ec_adjust_offset_down(fop->xl->private, &offset_down, _gf_true);
+
if (fop->id == GF_FOP_TRUNCATE) {
- ec_lock_prepare_inode(fop, &fop->loc[0],
- EC_UPDATE_DATA | EC_UPDATE_META |
- EC_QUERY_INFO, fop->offset, LLONG_MAX);
+ ec_lock_prepare_inode(
+ fop, &fop->loc[0],
+ EC_UPDATE_DATA | EC_UPDATE_META | EC_QUERY_INFO,
+ offset_down, EC_RANGE_FULL);
} else {
- ec_lock_prepare_fd(fop, fop->fd,
- EC_UPDATE_DATA | EC_UPDATE_META |
- EC_QUERY_INFO, fop->offset, LLONG_MAX);
+ ec_lock_prepare_fd(
+ fop, fop->fd,
+ EC_UPDATE_DATA | EC_UPDATE_META | EC_QUERY_INFO,
+ offset_down, EC_RANGE_FULL);
}
ec_lock(fop);
@@ -1470,8 +1442,7 @@ int32_t ec_manager_truncate(ec_fop_data_t * fop, int32_t state)
if (cbk != NULL) {
int32_t err;
- ec_iatt_rebuild(fop->xl->private, cbk->iatt, 2,
- cbk->count);
+ ec_iatt_rebuild(fop->xl->private, cbk->iatt, 2, cbk->count);
/* This shouldn't fail because we have the inode locked. */
/* Inode size doesn't need to be updated under locks, because
@@ -1499,24 +1470,17 @@ int32_t ec_manager_truncate(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->id == GF_FOP_TRUNCATE)
- {
- if (fop->cbks.truncate != NULL)
- {
- fop->cbks.truncate(fop->req_frame, fop, fop->xl,
- cbk->op_ret, cbk->op_errno,
- &cbk->iatt[0], &cbk->iatt[1],
- cbk->xdata);
+ if (fop->id == GF_FOP_TRUNCATE) {
+ if (fop->cbks.truncate != NULL) {
+ QUORUM_CBK(fop->cbks.truncate, fop, fop->req_frame, fop,
+ fop->xl, cbk->op_ret, cbk->op_errno,
+ &cbk->iatt[0], &cbk->iatt[1], cbk->xdata);
}
- }
- else
- {
- if (fop->cbks.ftruncate != NULL)
- {
- fop->cbks.ftruncate(fop->req_frame, fop, fop->xl,
- cbk->op_ret, cbk->op_errno,
- &cbk->iatt[0], &cbk->iatt[1],
- cbk->xdata);
+ } else {
+ if (fop->cbks.ftruncate != NULL) {
+ QUORUM_CBK(fop->cbks.ftruncate, fop, fop->req_frame, fop,
+ fop->xl, cbk->op_ret, cbk->op_errno,
+ &cbk->iatt[0], &cbk->iatt[1], cbk->xdata);
}
}
@@ -1529,18 +1493,13 @@ int32_t ec_manager_truncate(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->id == GF_FOP_TRUNCATE)
- {
- if (fop->cbks.truncate != NULL)
- {
+ if (fop->id == GF_FOP_TRUNCATE) {
+ if (fop->cbks.truncate != NULL) {
fop->cbks.truncate(fop->req_frame, fop, fop->xl, -1,
fop->error, NULL, NULL, NULL);
}
- }
- else
- {
- if (fop->cbks.ftruncate != NULL)
- {
+ } else {
+ if (fop->cbks.ftruncate != NULL) {
fop->cbks.ftruncate(fop->req_frame, fop, fop->xl, -1,
fop->error, NULL, NULL, NULL);
}
@@ -1561,31 +1520,30 @@ int32_t ec_manager_truncate(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE,
- "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_truncate(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_truncate_cbk_t func, void * data,
- loc_t * loc, off_t offset, dict_t * xdata)
+void
+ec_truncate(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_truncate_cbk_t func, void *data, loc_t *loc,
+ off_t offset, dict_t *xdata)
{
- ec_cbk_t callback = { .truncate = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.truncate = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(TRUNCATE) %p", frame);
+ gf_msg_trace("ec", 0, "EC(TRUNCATE) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_TRUNCATE, 0, target,
- minimum, ec_wind_truncate, ec_manager_truncate,
+ fop_flags, ec_wind_truncate, ec_manager_truncate,
callback, data);
if (fop == NULL) {
goto out;
@@ -1595,9 +1553,8 @@ void ec_truncate(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL,
- "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
@@ -1605,10 +1562,9 @@ void ec_truncate(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -1626,16 +1582,17 @@ out:
/* FOP: ftruncate */
-int32_t ec_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *prestat, struct iatt *poststat,
- dict_t *xdata)
+int32_t
+ec_ftruncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prestat,
+ struct iatt *poststat, dict_t *xdata)
{
- return ec_inode_write_cbk (frame, this, cookie, op_ret, op_errno,
- prestat, poststat, xdata);
+ return ec_inode_write_cbk(frame, this, cookie, op_ret, op_errno, prestat,
+ poststat, xdata);
}
-void ec_wind_ftruncate(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_ftruncate(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -1644,23 +1601,24 @@ void ec_wind_ftruncate(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->fd, fop->offset, fop->xdata);
}
-void ec_ftruncate(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_ftruncate_cbk_t func, void * data,
- fd_t * fd, off_t offset, dict_t * xdata)
+void
+ec_ftruncate(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_ftruncate_cbk_t func, void *data, fd_t *fd,
+ off_t offset, dict_t *xdata)
{
- ec_cbk_t callback = { .ftruncate = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.ftruncate = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(FTRUNCATE) %p", frame);
+ gf_msg_trace("ec", 0, "EC(FTRUNCATE) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_FTRUNCATE, 0, target,
- minimum, ec_wind_ftruncate, ec_manager_truncate,
- callback, data);
+ fop_flags, ec_wind_ftruncate,
+ ec_manager_truncate, callback, data);
if (fop == NULL) {
goto out;
}
@@ -1672,10 +1630,9 @@ void ec_ftruncate(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL,
- "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -1683,10 +1640,9 @@ void ec_ftruncate(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -1704,95 +1660,91 @@ out:
/* FOP: writev */
static ec_stripe_t *
-ec_allocate_stripe (ec_t *ec, ec_stripe_list_t *stripe_cache)
+ec_allocate_stripe(ec_t *ec, ec_stripe_list_t *stripe_cache)
{
- ec_stripe_t *stripe = NULL;
+ ec_stripe_t *stripe = NULL;
- if (stripe_cache->count >= stripe_cache->max) {
- GF_ASSERT (!list_empty(&stripe_cache->lru));
- stripe = list_first_entry(&stripe_cache->lru, ec_stripe_t, lru);
- list_move_tail(&stripe->lru, &stripe_cache->lru);
- GF_ATOMIC_INC(ec->stats.stripe_cache.evicts);
+ if (stripe_cache->count >= stripe_cache->max) {
+ GF_ASSERT(!list_empty(&stripe_cache->lru));
+ stripe = list_first_entry(&stripe_cache->lru, ec_stripe_t, lru);
+ list_move_tail(&stripe->lru, &stripe_cache->lru);
+ GF_ATOMIC_INC(ec->stats.stripe_cache.evicts);
+ } else {
+ stripe = GF_MALLOC(sizeof(ec_stripe_t) + ec->stripe_size,
+ ec_mt_ec_stripe_t);
+ if (stripe != NULL) {
+ stripe_cache->count++;
+ list_add_tail(&stripe->lru, &stripe_cache->lru);
+ GF_ATOMIC_INC(ec->stats.stripe_cache.allocs);
} else {
- stripe = GF_MALLOC (sizeof (ec_stripe_t) + ec->stripe_size,
- ec_mt_ec_stripe_t);
- if (stripe != NULL) {
- stripe_cache->count++;
- list_add_tail (&stripe->lru, &stripe_cache->lru);
- GF_ATOMIC_INC(ec->stats.stripe_cache.allocs);
- } else {
- GF_ATOMIC_INC(ec->stats.stripe_cache.errors);
- }
+ GF_ATOMIC_INC(ec->stats.stripe_cache.errors);
}
+ }
- return stripe;
+ return stripe;
}
static void
-ec_write_stripe_data (ec_t *ec, ec_fop_data_t *fop,
- ec_stripe_t *stripe)
+ec_write_stripe_data(ec_t *ec, ec_fop_data_t *fop, ec_stripe_t *stripe)
{
- off_t base;
+ off_t base;
- base = fop->size - ec->stripe_size;
- memcpy(stripe->data, fop->vector[0].iov_base + base, ec->stripe_size);
- stripe->frag_offset = fop->frag_range.last - ec->fragment_size;
+ base = fop->size - ec->stripe_size;
+ memcpy(stripe->data, fop->vector[0].iov_base + base, ec->stripe_size);
+ stripe->frag_offset = fop->frag_range.last - ec->fragment_size;
}
static void
-ec_add_stripe_in_cache (ec_t *ec, ec_fop_data_t *fop)
+ec_add_stripe_in_cache(ec_t *ec, ec_fop_data_t *fop)
{
- ec_inode_t *ctx = NULL;
- ec_stripe_t *stripe = NULL;
- ec_stripe_list_t *stripe_cache = NULL;
- gf_boolean_t failed = _gf_true;
+ ec_inode_t *ctx = NULL;
+ ec_stripe_t *stripe = NULL;
+ ec_stripe_list_t *stripe_cache = NULL;
+ gf_boolean_t failed = _gf_true;
- LOCK(&fop->fd->inode->lock);
+ LOCK(&fop->fd->inode->lock);
- ctx = __ec_inode_get (fop->fd->inode, fop->xl);
- if (ctx == NULL) {
- goto out;
- }
-
- stripe_cache = &ctx->stripe_cache;
- if (stripe_cache->max > 0) {
- stripe = ec_allocate_stripe (ec, stripe_cache);
- if (stripe == NULL) {
- goto out;
- }
+ ctx = __ec_inode_get(fop->fd->inode, fop->xl);
+ if (ctx == NULL) {
+ goto out;
+ }
- ec_write_stripe_data (ec, fop, stripe);
+ stripe_cache = &ctx->stripe_cache;
+ if (stripe_cache->max > 0) {
+ stripe = ec_allocate_stripe(ec, stripe_cache);
+ if (stripe == NULL) {
+ goto out;
}
- failed = _gf_false;
+ ec_write_stripe_data(ec, fop, stripe);
+ }
+
+ failed = _gf_false;
out:
- UNLOCK(&fop->fd->inode->lock);
+ UNLOCK(&fop->fd->inode->lock);
- if (failed) {
- gf_msg (ec->xl->name, GF_LOG_DEBUG, ENOMEM,
- EC_MSG_FILE_DESC_REF_FAIL,
- "Failed to create and add stripe in cache");
- }
+ if (failed) {
+ gf_msg(ec->xl->name, GF_LOG_DEBUG, ENOMEM, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to create and add stripe in cache");
+ }
}
-int32_t ec_writev_merge_tail(call_frame_t * frame, void * cookie,
- xlator_t * this, int32_t op_ret, int32_t op_errno,
- struct iovec * vector, int32_t count,
- struct iatt * stbuf, struct iobref * iobref,
- dict_t * xdata)
+int32_t
+ec_writev_merge_tail(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iovec *vector,
+ int32_t count, struct iatt *stbuf, struct iobref *iobref,
+ dict_t *xdata)
{
- ec_t * ec = this->private;
- ec_fop_data_t * fop = frame->local;
- size_t size, base, tmp;
+ ec_t *ec = this->private;
+ ec_fop_data_t *fop = frame->local;
+ uint64_t size, base, tmp;
- if (op_ret >= 0)
- {
+ if (op_ret >= 0) {
tmp = 0;
size = fop->size - fop->user_size - fop->head;
base = ec->stripe_size - size;
- if (op_ret > base)
- {
+ if (op_ret > base) {
tmp = min(op_ret - base, size);
ec_iov_copy_to(fop->vector[0].iov_base + fop->size - size, vector,
count, base, tmp);
@@ -1800,49 +1752,44 @@ int32_t ec_writev_merge_tail(call_frame_t * frame, void * cookie,
size -= tmp;
}
- if (size > 0)
- {
+ if (size > 0) {
memset(fop->vector[0].iov_base + fop->size - size, 0, size);
}
if (ec->stripe_cache) {
- ec_add_stripe_in_cache (ec, fop);
+ ec_add_stripe_in_cache(ec, fop);
}
}
return 0;
}
-int32_t ec_writev_merge_head(call_frame_t * frame, void * cookie,
- xlator_t * this, int32_t op_ret, int32_t op_errno,
- struct iovec * vector, int32_t count,
- struct iatt * stbuf, struct iobref * iobref,
- dict_t * xdata)
+int32_t
+ec_writev_merge_head(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iovec *vector,
+ int32_t count, struct iatt *stbuf, struct iobref *iobref,
+ dict_t *xdata)
{
- ec_t * ec = this->private;
- ec_fop_data_t * fop = frame->local;
- size_t size, base;
+ ec_t *ec = this->private;
+ ec_fop_data_t *fop = frame->local;
+ uint64_t size, base;
- if (op_ret >= 0)
- {
+ if (op_ret >= 0) {
size = fop->head;
base = 0;
- if (op_ret > 0)
- {
+ if (op_ret > 0) {
base = min(op_ret, size);
ec_iov_copy_to(fop->vector[0].iov_base, vector, count, 0, base);
size -= base;
}
- if (size > 0)
- {
+ if (size > 0) {
memset(fop->vector[0].iov_base + base, 0, size);
}
size = fop->size - fop->user_size - fop->head;
- if ((size > 0) && (fop->size == ec->stripe_size))
- {
+ if ((size > 0) && (fop->size == ec->stripe_size)) {
ec_writev_merge_tail(frame, cookie, this, op_ret, op_errno, vector,
count, stbuf, iobref, xdata);
}
@@ -1852,7 +1799,7 @@ int32_t ec_writev_merge_head(call_frame_t * frame, void * cookie,
}
static int
-ec_make_internal_fop_xdata (dict_t **xdata)
+ec_make_internal_fop_xdata(dict_t **xdata)
{
dict_t *dict = NULL;
@@ -1861,16 +1808,16 @@ ec_make_internal_fop_xdata (dict_t **xdata)
dict = dict_new();
if (!dict)
- goto out;
+ goto out;
- if (dict_set_str (dict, GLUSTERFS_INTERNAL_FOP_KEY, "yes"))
- goto out;
+ if (dict_set_str(dict, GLUSTERFS_INTERNAL_FOP_KEY, "yes"))
+ goto out;
*xdata = dict;
return 0;
out:
if (dict)
- dict_unref (dict);
+ dict_unref(dict);
return -1;
}
@@ -1889,8 +1836,7 @@ ec_writev_prepare_buffers(ec_t *ec, ec_fop_data_t *fop)
ec_adjust_size_up(ec, &fop->size, _gf_false);
fop->frag_range.last = fop->frag_range.first + fop->size / ec->fragments;
- if ((fop->int32 != 1) || (fop->head != 0) ||
- (fop->size > fop->user_size) ||
+ if ((fop->int32 != 1) || (fop->head != 0) || (fop->size > fop->user_size) ||
!EC_ALIGN_CHECK(fop->vector[0].iov_base, EC_METHOD_WORD_SIZE)) {
err = ec_buffer_alloc(ec->xl, fop->size, &iobref, &ptr);
if (err != 0) {
@@ -1935,98 +1881,116 @@ out:
}
static void
-ec_merge_stripe_head_locked (ec_t *ec, ec_fop_data_t *fop, ec_stripe_t *stripe)
+ec_merge_stripe_head_locked(ec_t *ec, ec_fop_data_t *fop, ec_stripe_t *stripe)
{
- size_t head, size;
+ uint32_t head, size;
- head = fop->head;
- memcpy(fop->vector[0].iov_base, stripe->data, head);
+ head = fop->head;
+ memcpy(fop->vector[0].iov_base, stripe->data, head);
+ size = ec->stripe_size - head;
+ if (size > fop->user_size) {
+ head += fop->user_size;
size = ec->stripe_size - head;
- if (size > fop->user_size) {
- head += fop->user_size;
- size = ec->stripe_size - head;
- memcpy(fop->vector[0].iov_base + head, stripe->data + head,
- size);
- }
+ memcpy(fop->vector[0].iov_base + head, stripe->data + head, size);
+ }
}
static void
-ec_merge_stripe_tail_locked (ec_t *ec, ec_fop_data_t *fop, ec_stripe_t *stripe)
+ec_merge_stripe_tail_locked(ec_t *ec, ec_fop_data_t *fop, ec_stripe_t *stripe)
{
- size_t head, tail;
- off_t offset;
+ uint32_t head, tail;
+ off_t offset;
- offset = fop->user_size + fop->head;
- tail = fop->size - offset;
- head = ec->stripe_size - tail;
+ offset = fop->user_size + fop->head;
+ tail = fop->size - offset;
+ head = ec->stripe_size - tail;
- memcpy(fop->vector[0].iov_base + offset, stripe->data + head, tail);
+ memcpy(fop->vector[0].iov_base + offset, stripe->data + head, tail);
}
static ec_stripe_t *
-ec_get_stripe_from_cache_locked (ec_t *ec, ec_fop_data_t *fop,
- uint64_t frag_offset)
+ec_get_stripe_from_cache_locked(ec_t *ec, ec_fop_data_t *fop,
+ uint64_t frag_offset)
{
- ec_inode_t *ctx = NULL;
- ec_stripe_t *stripe = NULL;
- ec_stripe_list_t *stripe_cache = NULL;
-
- ctx = __ec_inode_get (fop->fd->inode, fop->xl);
- if (ctx == NULL) {
- GF_ATOMIC_INC(ec->stats.stripe_cache.errors);
- return NULL;
- }
+ ec_inode_t *ctx = NULL;
+ ec_stripe_t *stripe = NULL;
+ ec_stripe_list_t *stripe_cache = NULL;
- stripe_cache = &ctx->stripe_cache;
- list_for_each_entry (stripe, &stripe_cache->lru, lru) {
- if (stripe->frag_offset == frag_offset) {
- list_move_tail (&stripe->lru, &stripe_cache->lru);
- GF_ATOMIC_INC(ec->stats.stripe_cache.hits);
- return stripe;
- }
+ ctx = __ec_inode_get(fop->fd->inode, fop->xl);
+ if (ctx == NULL) {
+ GF_ATOMIC_INC(ec->stats.stripe_cache.errors);
+ return NULL;
+ }
+
+ stripe_cache = &ctx->stripe_cache;
+ list_for_each_entry(stripe, &stripe_cache->lru, lru)
+ {
+ if (stripe->frag_offset == frag_offset) {
+ list_move_tail(&stripe->lru, &stripe_cache->lru);
+ GF_ATOMIC_INC(ec->stats.stripe_cache.hits);
+ return stripe;
}
+ }
- GF_ATOMIC_INC(ec->stats.stripe_cache.misses);
+ GF_ATOMIC_INC(ec->stats.stripe_cache.misses);
- return NULL;
+ return NULL;
}
static gf_boolean_t
-ec_get_and_merge_stripe (ec_t *ec, ec_fop_data_t *fop, ec_stripe_part_t which)
+ec_get_and_merge_stripe(ec_t *ec, ec_fop_data_t *fop, ec_stripe_part_t which)
{
- uint64_t frag_offset;
- ec_stripe_t *stripe = NULL;
- gf_boolean_t found = _gf_false;
+ uint64_t frag_offset;
+ ec_stripe_t *stripe = NULL;
+ gf_boolean_t found = _gf_false;
- if (!ec->stripe_cache) {
- return found;
- }
+ if (!ec->stripe_cache) {
+ return found;
+ }
- LOCK(&fop->fd->inode->lock);
- if (which == EC_STRIPE_HEAD) {
- frag_offset = fop->frag_range.first;
- stripe = ec_get_stripe_from_cache_locked(ec, fop, frag_offset);
- if (stripe) {
- ec_merge_stripe_head_locked (ec, fop, stripe);
- found = _gf_true;
- }
+ LOCK(&fop->fd->inode->lock);
+ if (which == EC_STRIPE_HEAD) {
+ frag_offset = fop->frag_range.first;
+ stripe = ec_get_stripe_from_cache_locked(ec, fop, frag_offset);
+ if (stripe) {
+ ec_merge_stripe_head_locked(ec, fop, stripe);
+ found = _gf_true;
}
+ }
- if (which == EC_STRIPE_TAIL) {
- frag_offset = fop->frag_range.last - ec->fragment_size;
- stripe = ec_get_stripe_from_cache_locked(ec, fop, frag_offset);
- if (stripe) {
- ec_merge_stripe_tail_locked (ec, fop, stripe);
- found = _gf_true;
- }
+ if (which == EC_STRIPE_TAIL) {
+ frag_offset = fop->frag_range.last - ec->fragment_size;
+ stripe = ec_get_stripe_from_cache_locked(ec, fop, frag_offset);
+ if (stripe) {
+ ec_merge_stripe_tail_locked(ec, fop, stripe);
+ found = _gf_true;
}
- UNLOCK(&fop->fd->inode->lock);
+ }
+ UNLOCK(&fop->fd->inode->lock);
- return found;
+ return found;
+}
+
+static uintptr_t
+ec_get_lock_good_mask(inode_t *inode, xlator_t *xl)
+{
+ ec_lock_t *lock = NULL;
+ ec_inode_t *ictx = NULL;
+ LOCK(&inode->lock);
+ {
+ ictx = __ec_inode_get(inode, xl);
+ if (ictx)
+ lock = ictx->inode_lock;
+ }
+ UNLOCK(&inode->lock);
+ if (lock)
+ return lock->good_mask;
+ return 0;
}
-void ec_writev_start(ec_fop_data_t *fop)
+void
+ec_writev_start(ec_fop_data_t *fop)
{
ec_t *ec = fop->xl->private;
ec_fd_t *ctx;
@@ -2061,41 +2025,51 @@ void ec_writev_start(ec_fop_data_t *fop)
if (err != 0) {
goto failed_fd;
}
+ tail = fop->size - fop->user_size - fop->head;
if (fop->head > 0) {
- found_stripe = ec_get_and_merge_stripe (ec, fop, EC_STRIPE_HEAD);
- if (!found_stripe) {
- if (ec_make_internal_fop_xdata (&xdata)) {
- err = -ENOMEM;
- goto failed_xdata;
+ if (current > fop->offset) {
+ found_stripe = ec_get_and_merge_stripe(ec, fop, EC_STRIPE_HEAD);
+ if (!found_stripe) {
+ if (ec_make_internal_fop_xdata(&xdata)) {
+ err = -ENOMEM;
+ goto failed_xdata;
}
- ec_readv(fop->frame, fop->xl, -1, EC_MINIMUM_MIN,
- ec_writev_merge_head,
- NULL, fd, ec->stripe_size, fop->offset, 0, xdata);
+ ec_readv(fop->frame, fop->xl,
+ ec_get_lock_good_mask(fop->fd->inode, fop->xl),
+ EC_MINIMUM_MIN, ec_writev_merge_head, NULL, fd,
+ ec->stripe_size, fop->offset, 0, xdata);
+ }
+ } else {
+ memset(fop->vector[0].iov_base, 0, fop->head);
+ memset(fop->vector[0].iov_base + fop->size - tail, 0, tail);
+ if (ec->stripe_cache && (fop->size <= ec->stripe_size)) {
+ ec_add_stripe_in_cache(ec, fop);
+ }
}
}
- tail = fop->size - fop->user_size - fop->head;
if ((tail > 0) && ((fop->head == 0) || (fop->size > ec->stripe_size))) {
- /* Current locking scheme will make sure the 'current' below will
- * never decrease while the fop is in progress, so the checks will
- * work as expected
- */
+ /* Current locking scheme will make sure the 'current' below will
+ * never decrease while the fop is in progress, so the checks will
+ * work as expected
+ */
if (current > fop->offset + fop->head + fop->user_size) {
- found_stripe = ec_get_and_merge_stripe (ec, fop, EC_STRIPE_TAIL);
+ found_stripe = ec_get_and_merge_stripe(ec, fop, EC_STRIPE_TAIL);
if (!found_stripe) {
- if (ec_make_internal_fop_xdata (&xdata)) {
- err = -ENOMEM;
- goto failed_xdata;
- }
- ec_readv(fop->frame, fop->xl, -1, EC_MINIMUM_MIN,
- ec_writev_merge_tail, NULL, fd, ec->stripe_size,
- fop->offset + fop->size - ec->stripe_size,
- 0, xdata);
+ if (ec_make_internal_fop_xdata(&xdata)) {
+ err = -ENOMEM;
+ goto failed_xdata;
+ }
+ ec_readv(fop->frame, fop->xl,
+ ec_get_lock_good_mask(fop->fd->inode, fop->xl),
+ EC_MINIMUM_MIN, ec_writev_merge_tail, NULL, fd,
+ ec->stripe_size,
+ fop->offset + fop->size - ec->stripe_size, 0, xdata);
}
} else {
memset(fop->vector[0].iov_base + fop->size - tail, 0, tail);
if (ec->stripe_cache) {
- ec_add_stripe_in_cache (ec, fop);
+ ec_add_stripe_in_cache(ec, fop);
}
}
}
@@ -2112,23 +2086,25 @@ failed:
ec_fop_set_error(fop, -err);
}
-int32_t ec_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prestat,
- struct iatt *poststat, dict_t *xdata)
+int32_t
+ec_writev_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *prestat, struct iatt *poststat,
+ dict_t *xdata)
{
- ec_t *ec = NULL;
- if (this && this->private) {
- ec = this->private;
- if ((op_ret > 0) && ((op_ret % ec->fragment_size) != 0)) {
- op_ret = -1;
- op_errno = EIO;
- }
+ ec_t *ec = NULL;
+ if (this && this->private) {
+ ec = this->private;
+ if ((op_ret > 0) && ((op_ret % ec->fragment_size) != 0)) {
+ op_ret = -1;
+ op_errno = EIO;
}
- return ec_inode_write_cbk (frame, this, cookie, op_ret, op_errno,
- prestat, poststat, xdata);
+ }
+ return ec_inode_write_cbk(frame, this, cookie, op_ret, op_errno, prestat,
+ poststat, xdata);
}
-void ec_wind_writev(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_writev(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -2141,9 +2117,9 @@ void ec_wind_writev(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
vector[0].iov_len = size;
STACK_WIND_COOKIE(fop->frame, ec_writev_cbk, (void *)(uintptr_t)idx,
- ec->xl_list[idx], ec->xl_list[idx]->fops->writev,
- fop->fd, vector, 1, fop->offset / ec->fragments,
- fop->uint32, fop->buffers, fop->xdata);
+ ec->xl_list[idx], ec->xl_list[idx]->fops->writev, fop->fd,
+ vector, 1, fop->offset / ec->fragments, fop->uint32,
+ fop->buffers, fop->xdata);
}
static void
@@ -2161,35 +2137,34 @@ ec_writev_encode(ec_fop_data_t *fop)
fop->vector[0].iov_base, blocks);
}
-int32_t ec_manager_writev(ec_fop_data_t *fop, int32_t state)
+int32_t
+ec_manager_writev(ec_fop_data_t *fop, int32_t state)
{
ec_cbk_data_t *cbk;
ec_fd_t *ctx = NULL;
- ec_t *ec = fop->xl->private;
+ ec_t *ec = fop->xl->private;
off_t fl_start = 0;
- size_t fl_size = LLONG_MAX;
+ uint64_t fl_size = LONG_MAX;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
case EC_STATE_LOCK:
- ctx = ec_fd_get(fop->fd, fop->xl);
- if (ctx != NULL) {
- if ((ctx->flags & O_APPEND) == 0) {
- off_t user_size = 0;
- off_t head = 0;
-
- fl_start = fop->offset;
- user_size = iov_length(fop->vector, fop->int32);
- head = ec_adjust_offset_down(ec, &fl_start,
- _gf_true);
- fl_size = user_size + head;
- ec_adjust_size_up(ec, &fl_size, _gf_true);
- }
+ ctx = ec_fd_get(fop->fd, fop->xl);
+ if (ctx != NULL) {
+ if ((ctx->flags & O_APPEND) == 0) {
+ off_t user_size = 0;
+ off_t head = 0;
+
+ fl_start = fop->offset;
+ user_size = iov_length(fop->vector, fop->int32);
+ head = ec_adjust_offset_down(ec, &fl_start, _gf_true);
+ fl_size = user_size + head;
+ ec_adjust_size_up(ec, &fl_size, _gf_true);
}
+ }
ec_lock_prepare_fd(fop, fop->fd,
- EC_UPDATE_DATA | EC_UPDATE_META |
- EC_QUERY_INFO, fl_start, fl_size);
+ EC_UPDATE_DATA | EC_UPDATE_META | EC_QUERY_INFO,
+ fl_start, fl_size);
ec_lock(fop);
return EC_STATE_DISPATCH;
@@ -2215,31 +2190,30 @@ int32_t ec_manager_writev(ec_fop_data_t *fop, int32_t state)
cbk = ec_fop_prepare_answer(fop, _gf_false);
if (cbk != NULL) {
ec_t *ec = fop->xl->private;
- size_t size;
+ uint64_t size;
- ec_iatt_rebuild(fop->xl->private, cbk->iatt, 2,
- cbk->count);
+ ec_iatt_rebuild(fop->xl->private, cbk->iatt, 2, cbk->count);
/* This shouldn't fail because we have the inode locked. */
LOCK(&fop->fd->inode->lock);
{
- GF_ASSERT(__ec_get_inode_size(fop, fop->fd->inode,
- &cbk->iatt[0].ia_size));
- cbk->iatt[1].ia_size = cbk->iatt[0].ia_size;
- size = fop->offset + fop->head + fop->user_size;
- if (size > cbk->iatt[0].ia_size) {
- /* Only update inode size if this is a top level fop.
- * Otherwise this is an internal write and the top
- * level fop should take care of the real inode size.
- */
- if (fop->parent == NULL) {
- /* This shouldn't fail because we have the inode
- * locked. */
- GF_ASSERT(__ec_set_inode_size(fop,
- fop->fd->inode, size));
- }
- cbk->iatt[1].ia_size = size;
+ GF_ASSERT(__ec_get_inode_size(fop, fop->fd->inode,
+ &cbk->iatt[0].ia_size));
+ cbk->iatt[1].ia_size = cbk->iatt[0].ia_size;
+ size = fop->offset + fop->head + fop->user_size;
+ if (size > cbk->iatt[0].ia_size) {
+ /* Only update inode size if this is a top level fop.
+ * Otherwise this is an internal write and the top
+ * level fop should take care of the real inode size.
+ */
+ if (fop->parent == NULL) {
+ /* This shouldn't fail because we have the inode
+ * locked. */
+ GF_ASSERT(
+ __ec_set_inode_size(fop, fop->fd->inode, size));
}
+ cbk->iatt[1].ia_size = size;
+ }
}
UNLOCK(&fop->fd->inode->lock);
@@ -2263,11 +2237,10 @@ int32_t ec_manager_writev(ec_fop_data_t *fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.writev != NULL)
- {
- fop->cbks.writev(fop->req_frame, fop, fop->xl, cbk->op_ret,
- cbk->op_errno, &cbk->iatt[0], &cbk->iatt[1],
- cbk->xdata);
+ if (fop->cbks.writev != NULL) {
+ QUORUM_CBK(fop->cbks.writev, fop, fop->req_frame, fop, fop->xl,
+ cbk->op_ret, cbk->op_errno, &cbk->iatt[0],
+ &cbk->iatt[1], cbk->xdata);
}
return EC_STATE_LOCK_REUSE;
@@ -2278,7 +2251,7 @@ int32_t ec_manager_writev(ec_fop_data_t *fop, int32_t state)
fop->frame->root->uid = fop->uid;
fop->frame->root->gid = fop->gid;
- /* Fall through */
+ /* Fall through */
case -EC_STATE_INIT:
case -EC_STATE_LOCK:
@@ -2287,8 +2260,7 @@ int32_t ec_manager_writev(ec_fop_data_t *fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.writev != NULL)
- {
+ if (fop->cbks.writev != NULL) {
fop->cbks.writev(fop->req_frame, fop, fop->xl, -1, fop->error,
NULL, NULL, NULL);
}
@@ -2308,31 +2280,30 @@ int32_t ec_manager_writev(ec_fop_data_t *fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE,
- "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_writev(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_writev_cbk_t func, void * data, fd_t * fd,
- struct iovec * vector, int32_t count, off_t offset,
- uint32_t flags, struct iobref * iobref, dict_t * xdata)
+void
+ec_writev(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_writev_cbk_t func, void *data, fd_t *fd,
+ struct iovec *vector, int32_t count, off_t offset, uint32_t flags,
+ struct iobref *iobref, dict_t *xdata)
{
- ec_cbk_t callback = { .writev = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.writev = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(WRITE) %p", frame);
+ gf_msg_trace("ec", 0, "EC(WRITE) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_WRITE, 0, target, minimum,
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_WRITE, 0, target, fop_flags,
ec_wind_writev, ec_manager_writev, callback,
data);
if (fop == NULL) {
@@ -2348,10 +2319,9 @@ void ec_writev(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL,
- "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -2359,10 +2329,9 @@ void ec_writev(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (count > 0) {
fop->vector = iov_dup(vector, count);
if (fop->vector == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY,
- "Failed to duplicate a "
- "vector list.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to duplicate a "
+ "vector list.");
goto out;
}
@@ -2371,10 +2340,9 @@ void ec_writev(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (iobref != NULL) {
fop->buffers = iobref_ref(iobref);
if (fop->buffers == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_BUF_REF_FAIL,
- "Failed to reference a "
- "buffer.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_BUF_REF_FAIL,
+ "Failed to reference a "
+ "buffer.");
goto out;
}
@@ -2382,10 +2350,9 @@ void ec_writev(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_copy_with_ref(xdata, NULL);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
diff --git a/xlators/cluster/ec/src/ec-locks.c b/xlators/cluster/ec/src/ec-locks.c
index 996035de90b..601960d6154 100644
--- a/xlators/cluster/ec/src/ec-locks.c
+++ b/xlators/cluster/ec/src/ec-locks.c
@@ -8,52 +8,76 @@
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "defaults.h"
-
#include "ec-helpers.h"
#include "ec-common.h"
#include "ec-combine.h"
-#include "ec-method.h"
#include "ec-fops.h"
#include "ec-messages.h"
#define EC_LOCK_MODE_NONE 0
-#define EC_LOCK_MODE_INC 1
-#define EC_LOCK_MODE_ALL 2
+#define EC_LOCK_MODE_INC 1
+#define EC_LOCK_MODE_ALL 2
-int32_t ec_lock_check(ec_fop_data_t *fop, uintptr_t *mask)
+int32_t
+ec_lock_check(ec_fop_data_t *fop, uintptr_t *mask)
{
ec_t *ec = fop->xl->private;
ec_cbk_data_t *ans = NULL;
ec_cbk_data_t *cbk = NULL;
- uintptr_t locked = 0, notlocked = 0;
+ uintptr_t locked = 0;
+ int32_t good = 0;
+ int32_t eagain = 0;
+ int32_t estale = 0;
int32_t error = -1;
- list_for_each_entry(ans, &fop->cbk_list, list) {
+ /* There are some errors that we'll handle in an special way while trying
+ * to acquire a lock.
+ *
+ * EAGAIN: If it's found during a parallel non-blocking lock request, we
+ * consider that there's contention on the inode, so we consider
+ * the acquisition a failure and try again with a sequential
+ * blocking lock request. This will ensure that we get a lock on
+ * as many bricks as possible (ignoring EAGAIN here would cause
+ * unnecessary triggers of self-healing).
+ *
+ * If it's found during a sequential blocking lock request, it's
+ * considered an error. Lock will only succeed if there are
+ * enough other bricks locked.
+ *
+ * ESTALE: This can appear during parallel or sequential lock request if
+ * the inode has just been unlinked. We consider this error is
+ * not recoverable, but we also don't consider it as fatal. So,
+ * if it happens during parallel lock, we won't attempt a
+ * sequential one unless there are EAGAIN errors on other
+ * bricks (and are enough to form a quorum), but if we reach
+ * quorum counting the ESTALE bricks, we consider the whole
+ * result of the operation is ESTALE instead of EIO.
+ */
+
+ list_for_each_entry(ans, &fop->cbk_list, list)
+ {
if (ans->op_ret >= 0) {
if (locked != 0) {
error = EIO;
}
locked |= ans->mask;
+ good = ans->count;
cbk = ans;
- } else {
- if (ans->op_errno == EAGAIN) {
- switch (fop->uint32) {
- case EC_LOCK_MODE_NONE:
- case EC_LOCK_MODE_ALL:
- /* Goal is to treat non-blocking lock as failure
- * even if there is a signle EAGAIN*/
- notlocked |= ans->mask;
- break;
- }
- }
+ } else if (ans->op_errno == ESTALE) {
+ estale += ans->count;
+ } else if ((ans->op_errno == EAGAIN) &&
+ (fop->uint32 != EC_LOCK_MODE_INC)) {
+ eagain += ans->count;
}
}
if (error == -1) {
- if (gf_bits_count(locked | notlocked) >= ec->fragments) {
- if (notlocked == 0) {
+ /* If we have enough quorum with succeeded and EAGAIN answers, we
+ * ignore for now any ESTALE answer. If there are EAGAIN answers,
+ * we retry with a sequential blocking lock request if needed.
+ * Otherwise we succeed. */
+ if ((good + eagain) >= ec->fragments) {
+ if (eagain == 0) {
if (fop->answer == NULL) {
fop->answer = cbk;
}
@@ -63,24 +87,31 @@ int32_t ec_lock_check(ec_fop_data_t *fop, uintptr_t *mask)
error = 0;
} else {
switch (fop->uint32) {
- case EC_LOCK_MODE_NONE:
- error = EAGAIN;
- break;
-
- case EC_LOCK_MODE_ALL:
- fop->uint32 = EC_LOCK_MODE_INC;
- break;
-
- default:
- error = EIO;
- break;
+ case EC_LOCK_MODE_NONE:
+ error = EAGAIN;
+ break;
+ case EC_LOCK_MODE_ALL:
+ fop->uint32 = EC_LOCK_MODE_INC;
+ break;
+ default:
+ /* This shouldn't happen because eagain cannot be > 0
+ * when fop->uint32 is EC_LOCK_MODE_INC. */
+ error = EIO;
+ break;
}
}
} else {
- if (fop->answer && fop->answer->op_ret < 0)
- error = fop->answer->op_errno;
- else
- error = EIO;
+ /* We have been unable to find enough candidates that will be able
+ * to take the lock. If we have quorum on some answer, we return
+ * it. Otherwise we check if ESTALE answers allow us to reach
+ * quorum. If so, we return ESTALE. */
+ if (fop->answer && fop->answer->op_ret < 0) {
+ error = fop->answer->op_errno;
+ } else if ((good + eagain + estale) >= ec->fragments) {
+ error = ESTALE;
+ } else {
+ error = EIO;
+ }
}
}
@@ -89,28 +120,25 @@ int32_t ec_lock_check(ec_fop_data_t *fop, uintptr_t *mask)
return error;
}
-int32_t ec_lock_unlocked(call_frame_t * frame, void * cookie,
- xlator_t * this, int32_t op_ret, int32_t op_errno,
- dict_t * xdata)
+int32_t
+ec_lock_unlocked(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- if (op_ret < 0)
- {
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- EC_MSG_UNLOCK_FAILED,
- "Failed to unlock an entry/inode");
+ if (op_ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, EC_MSG_UNLOCK_FAILED,
+ "Failed to unlock an entry/inode");
}
return 0;
}
-int32_t ec_lock_lk_unlocked(call_frame_t * frame, void * cookie,
- xlator_t * this, int32_t op_ret, int32_t op_errno,
- struct gf_flock * flock, dict_t * xdata)
+int32_t
+ec_lock_lk_unlocked(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct gf_flock *flock,
+ dict_t *xdata)
{
- if (op_ret < 0)
- {
- gf_msg(this->name, GF_LOG_WARNING, op_errno,
- EC_MSG_LK_UNLOCK_FAILED,
+ if (op_ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, EC_MSG_LK_UNLOCK_FAILED,
"Failed to unlock an lk");
}
@@ -119,11 +147,12 @@ int32_t ec_lock_lk_unlocked(call_frame_t * frame, void * cookie,
/* FOP: entrylk */
-int32_t ec_entrylk_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, dict_t * xdata)
+int32_t
+ec_entrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -133,20 +162,16 @@ int32_t ec_entrylk_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_ENTRYLK, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (xdata != NULL)
- {
+ if (cbk != NULL) {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg(this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
"Failed to reference a "
"dictionary.");
@@ -158,15 +183,15 @@ int32_t ec_entrylk_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_entrylk(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_entrylk(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -176,20 +201,19 @@ void ec_wind_entrylk(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->entrylk_type, fop->xdata);
}
-int32_t ec_manager_entrylk(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_entrylk(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
- if (fop->entrylk_cmd == ENTRYLK_LOCK)
- {
+ if (fop->entrylk_cmd == ENTRYLK_LOCK) {
fop->uint32 = EC_LOCK_MODE_ALL;
fop->entrylk_cmd = ENTRYLK_LOCK_NB;
}
- /* Fall through */
+ /* Fall through */
case EC_STATE_DISPATCH:
ec_dispatch_all(fop);
@@ -201,15 +225,14 @@ int32_t ec_manager_entrylk(ec_fop_data_t * fop, int32_t state)
if (fop->entrylk_cmd != ENTRYLK_UNLOCK) {
uintptr_t mask;
- ec_fop_set_error (fop, ec_lock_check(fop, &mask));
+ ec_fop_set_error(fop, ec_lock_check(fop, &mask));
if (fop->error != 0) {
if (mask != 0) {
if (fop->id == GF_FOP_ENTRYLK) {
- ec_entrylk(fop->frame, fop->xl, mask, 1,
- ec_lock_unlocked, NULL, fop->str[0],
- &fop->loc[0], fop->str[1],
- ENTRYLK_UNLOCK, fop->entrylk_type,
- fop->xdata);
+ ec_entrylk(
+ fop->frame, fop->xl, mask, 1, ec_lock_unlocked,
+ NULL, fop->str[0], &fop->loc[0], fop->str[1],
+ ENTRYLK_UNLOCK, fop->entrylk_type, fop->xdata);
} else {
ec_fentrylk(fop->frame, fop->xl, mask, 1,
ec_lock_unlocked, NULL, fop->str[0],
@@ -238,18 +261,13 @@ int32_t ec_manager_entrylk(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->id == GF_FOP_ENTRYLK)
- {
- if (fop->cbks.entrylk != NULL)
- {
- fop->cbks.entrylk(fop->req_frame, fop, fop->xl,
- cbk->op_ret, cbk->op_errno, cbk->xdata);
+ if (fop->id == GF_FOP_ENTRYLK) {
+ if (fop->cbks.entrylk != NULL) {
+ fop->cbks.entrylk(fop->req_frame, fop, fop->xl, cbk->op_ret,
+ cbk->op_errno, cbk->xdata);
}
- }
- else
- {
- if (fop->cbks.fentrylk != NULL)
- {
+ } else {
+ if (fop->cbks.fentrylk != NULL) {
fop->cbks.fentrylk(fop->req_frame, fop, fop->xl,
cbk->op_ret, cbk->op_errno, cbk->xdata);
}
@@ -262,18 +280,13 @@ int32_t ec_manager_entrylk(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->id == GF_FOP_ENTRYLK)
- {
- if (fop->cbks.entrylk != NULL)
- {
+ if (fop->id == GF_FOP_ENTRYLK) {
+ if (fop->cbks.entrylk != NULL) {
fop->cbks.entrylk(fop->req_frame, fop, fop->xl, -1,
fop->error, NULL);
}
- }
- else
- {
- if (fop->cbks.fentrylk != NULL)
- {
+ } else {
+ if (fop->cbks.fentrylk != NULL) {
fop->cbks.fentrylk(fop->req_frame, fop, fop->xl, -1,
fop->error, NULL);
}
@@ -282,33 +295,31 @@ int32_t ec_manager_entrylk(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE,
- "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_entrylk(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_entrylk_cbk_t func, void * data,
- const char * volume, loc_t * loc, const char * basename,
- entrylk_cmd cmd, entrylk_type type, dict_t * xdata)
+void
+ec_entrylk(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_entrylk_cbk_t func, void *data,
+ const char *volume, loc_t *loc, const char *basename,
+ entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
{
- ec_cbk_t callback = { .entrylk = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.entrylk = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(ENTRYLK) %p", frame);
+ gf_msg_trace("ec", 0, "EC(ENTRYLK) %p", frame);
- VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_ENTRYLK, 0, target, minimum,
- ec_wind_entrylk, ec_manager_entrylk, callback,
- data);
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_ENTRYLK, 0, target,
+ fop_flags, ec_wind_entrylk, ec_manager_entrylk,
+ callback, data);
if (fop == NULL) {
goto out;
}
@@ -319,18 +330,16 @@ void ec_entrylk(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (volume != NULL) {
fop->str[0] = gf_strdup(volume);
if (fop->str[0] == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY,
- "Failed to duplicate a string.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to duplicate a string.");
goto out;
}
}
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL,
- "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
@@ -338,9 +347,8 @@ void ec_entrylk(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (basename != NULL) {
fop->str[1] = gf_strdup(basename);
if (fop->str[1] == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY,
- "Failed to duplicate a string.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to duplicate a string.");
goto out;
}
@@ -348,10 +356,9 @@ void ec_entrylk(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -369,11 +376,12 @@ out:
/* FOP: fentrylk */
-int32_t ec_fentrylk_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, dict_t * xdata)
+int32_t
+ec_fentrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -383,22 +391,18 @@ int32_t ec_fentrylk_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_FENTRYLK, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (xdata != NULL)
- {
+ if (cbk != NULL) {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -408,15 +412,15 @@ int32_t ec_fentrylk_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_fentrylk(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_fentrylk(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -426,23 +430,23 @@ void ec_wind_fentrylk(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->entrylk_type, fop->xdata);
}
-void ec_fentrylk(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_fentrylk_cbk_t func, void * data,
- const char * volume, fd_t * fd, const char * basename,
- entrylk_cmd cmd, entrylk_type type, dict_t * xdata)
+void
+ec_fentrylk(call_frame_t *frame, xlator_t *this, uintptr_t target,
+ uint32_t fop_flags, fop_fentrylk_cbk_t func, void *data,
+ const char *volume, fd_t *fd, const char *basename, entrylk_cmd cmd,
+ entrylk_type type, dict_t *xdata)
{
- ec_cbk_t callback = { .fentrylk = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.fentrylk = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(FENTRYLK) %p", frame);
+ gf_msg_trace("ec", 0, "EC(FENTRYLK) %p", frame);
- VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_FENTRYLK, 0, target,
- minimum, ec_wind_fentrylk, ec_manager_entrylk,
+ fop_flags, ec_wind_fentrylk, ec_manager_entrylk,
callback, data);
if (fop == NULL) {
goto out;
@@ -456,9 +460,8 @@ void ec_fentrylk(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (volume != NULL) {
fop->str[0] = gf_strdup(volume);
if (fop->str[0] == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY,
- "Failed to duplicate a string.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to duplicate a string.");
goto out;
}
@@ -466,10 +469,9 @@ void ec_fentrylk(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL,
- "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -477,9 +479,8 @@ void ec_fentrylk(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (basename != NULL) {
fop->str[1] = gf_strdup(basename);
if (fop->str[1] == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY,
- "Failed to duplicate a string.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to duplicate a string.");
goto out;
}
@@ -487,10 +488,9 @@ void ec_fentrylk(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -508,11 +508,12 @@ out:
/* FOP: inodelk */
-int32_t ec_inodelk_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, dict_t * xdata)
+int32_t
+ec_inodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -522,22 +523,18 @@ int32_t ec_inodelk_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_INODELK, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (xdata != NULL)
- {
+ if (cbk != NULL) {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -547,15 +544,15 @@ int32_t ec_inodelk_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_inodelk(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_inodelk(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -565,24 +562,22 @@ void ec_wind_inodelk(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->xdata);
}
-int32_t ec_manager_inodelk(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_inodelk(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
- fop->flock.l_len += ec_adjust_offset_down(fop->xl->private,
- &fop->flock.l_start,
- _gf_true);
+ fop->flock.l_len += ec_adjust_offset_down(
+ fop->xl->private, &fop->flock.l_start, _gf_true);
ec_adjust_offset_up(fop->xl->private, &fop->flock.l_len, _gf_true);
- if ((fop->int32 == F_SETLKW) && (fop->flock.l_type != F_UNLCK))
- {
+ if ((fop->int32 == F_SETLKW) && (fop->flock.l_type != F_UNLCK)) {
fop->uint32 = EC_LOCK_MODE_ALL;
fop->int32 = F_SETLK;
}
- /* Fall through */
+ /* Fall through */
case EC_STATE_DISPATCH:
ec_dispatch_all(fop);
@@ -594,7 +589,7 @@ int32_t ec_manager_inodelk(ec_fop_data_t * fop, int32_t state)
if (fop->flock.l_type != F_UNLCK) {
uintptr_t mask;
- ec_fop_set_error (fop, ec_lock_check(fop, &mask));
+ ec_fop_set_error(fop, ec_lock_check(fop, &mask));
if (fop->error != 0) {
if (mask != 0) {
ec_t *ec = fop->xl->private;
@@ -641,18 +636,13 @@ int32_t ec_manager_inodelk(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->id == GF_FOP_INODELK)
- {
- if (fop->cbks.inodelk != NULL)
- {
- fop->cbks.inodelk(fop->req_frame, fop, fop->xl,
- cbk->op_ret, cbk->op_errno, cbk->xdata);
+ if (fop->id == GF_FOP_INODELK) {
+ if (fop->cbks.inodelk != NULL) {
+ fop->cbks.inodelk(fop->req_frame, fop, fop->xl, cbk->op_ret,
+ cbk->op_errno, cbk->xdata);
}
- }
- else
- {
- if (fop->cbks.finodelk != NULL)
- {
+ } else {
+ if (fop->cbks.finodelk != NULL) {
fop->cbks.finodelk(fop->req_frame, fop, fop->xl,
cbk->op_ret, cbk->op_errno, cbk->xdata);
}
@@ -665,18 +655,13 @@ int32_t ec_manager_inodelk(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->id == GF_FOP_INODELK)
- {
- if (fop->cbks.inodelk != NULL)
- {
+ if (fop->id == GF_FOP_INODELK) {
+ if (fop->cbks.inodelk != NULL) {
fop->cbks.inodelk(fop->req_frame, fop, fop->xl, -1,
fop->error, NULL);
}
- }
- else
- {
- if (fop->cbks.finodelk != NULL)
- {
+ } else {
+ if (fop->cbks.finodelk != NULL) {
fop->cbks.finodelk(fop->req_frame, fop, fop->xl, -1,
fop->error, NULL);
}
@@ -685,55 +670,52 @@ int32_t ec_manager_inodelk(ec_fop_data_t * fop, int32_t state)
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE,
- "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_inodelk (call_frame_t *frame, xlator_t *this, gf_lkowner_t *owner,
- uintptr_t target, int32_t minimum, fop_inodelk_cbk_t func,
- void *data, const char *volume, loc_t *loc, int32_t cmd,
- struct gf_flock *flock, dict_t *xdata)
+void
+ec_inodelk(call_frame_t *frame, xlator_t *this, gf_lkowner_t *owner,
+ uintptr_t target, uint32_t fop_flags, fop_inodelk_cbk_t func,
+ void *data, const char *volume, loc_t *loc, int32_t cmd,
+ struct gf_flock *flock, dict_t *xdata)
{
- ec_cbk_t callback = { .inodelk = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.inodelk = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(INODELK) %p", frame);
+ gf_msg_trace("ec", 0, "EC(INODELK) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_INODELK, 0, target, minimum,
- ec_wind_inodelk, ec_manager_inodelk, callback,
- data);
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_INODELK, 0, target,
+ fop_flags, ec_wind_inodelk, ec_manager_inodelk,
+ callback, data);
if (fop == NULL) {
goto out;
}
fop->int32 = cmd;
- ec_owner_copy (fop->frame, owner);
+ ec_owner_copy(fop->frame, owner);
if (volume != NULL) {
fop->str[0] = gf_strdup(volume);
if (fop->str[0] == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY,
- "Failed to duplicate a string.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to duplicate a string.");
goto out;
}
}
if (loc != NULL) {
if (loc_copy(&fop->loc[0], loc) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_LOC_COPY_FAIL,
- "Failed to copy a location.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_LOC_COPY_FAIL,
+ "Failed to copy a location.");
goto out;
}
@@ -753,10 +735,9 @@ void ec_inodelk (call_frame_t *frame, xlator_t *this, gf_lkowner_t *owner,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -774,11 +755,12 @@ out:
/* FOP: finodelk */
-int32_t ec_finodelk_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, dict_t * xdata)
+int32_t
+ec_finodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -788,22 +770,18 @@ int32_t ec_finodelk_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_FINODELK, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (xdata != NULL)
- {
+ if (cbk != NULL) {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -813,15 +791,15 @@ int32_t ec_finodelk_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_finodelk(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_finodelk(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -831,23 +809,24 @@ void ec_wind_finodelk(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->xdata);
}
-void ec_finodelk(call_frame_t *frame, xlator_t *this, gf_lkowner_t *owner,
- uintptr_t target, int32_t minimum, fop_finodelk_cbk_t func,
- void *data, const char *volume, fd_t *fd, int32_t cmd,
- struct gf_flock *flock, dict_t *xdata)
+void
+ec_finodelk(call_frame_t *frame, xlator_t *this, gf_lkowner_t *owner,
+ uintptr_t target, uint32_t fop_flags, fop_finodelk_cbk_t func,
+ void *data, const char *volume, fd_t *fd, int32_t cmd,
+ struct gf_flock *flock, dict_t *xdata)
{
- ec_cbk_t callback = { .finodelk = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.finodelk = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(FINODELK) %p", frame);
+ gf_msg_trace("ec", 0, "EC(FINODELK) %p", frame);
VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
fop = ec_fop_data_allocate(frame, this, GF_FOP_FINODELK, 0, target,
- minimum, ec_wind_finodelk, ec_manager_inodelk,
+ fop_flags, ec_wind_finodelk, ec_manager_inodelk,
callback, data);
if (fop == NULL) {
goto out;
@@ -856,14 +835,13 @@ void ec_finodelk(call_frame_t *frame, xlator_t *this, gf_lkowner_t *owner,
fop->use_fd = 1;
fop->int32 = cmd;
- ec_owner_copy (fop->frame, owner);
+ ec_owner_copy(fop->frame, owner);
if (volume != NULL) {
fop->str[0] = gf_strdup(volume);
if (fop->str[0] == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY,
- "Failed to duplicate a string.");
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to duplicate a string.");
goto out;
}
@@ -871,10 +849,9 @@ void ec_finodelk(call_frame_t *frame, xlator_t *this, gf_lkowner_t *owner,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -894,10 +871,9 @@ void ec_finodelk(call_frame_t *frame, xlator_t *this, gf_lkowner_t *owner,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -915,15 +891,13 @@ out:
/* FOP: lk */
-int32_t ec_combine_lk(ec_fop_data_t * fop, ec_cbk_data_t * dst,
- ec_cbk_data_t * src)
+int32_t
+ec_combine_lk(ec_fop_data_t *fop, ec_cbk_data_t *dst, ec_cbk_data_t *src)
{
- if (!ec_flock_compare(&dst->flock, &src->flock))
- {
- gf_msg (fop->xl->name, GF_LOG_NOTICE, 0,
- EC_MSG_LOCK_MISMATCH,
- "Mismatching lock in "
- "answers of 'GF_FOP_LK'");
+ if (!ec_flock_compare(&dst->flock, &src->flock)) {
+ gf_msg(fop->xl->name, GF_LOG_NOTICE, 0, EC_MSG_LOCK_MISMATCH,
+ "Mismatching lock in "
+ "answers of 'GF_FOP_LK'");
return 0;
}
@@ -931,12 +905,12 @@ int32_t ec_combine_lk(ec_fop_data_t * fop, ec_cbk_data_t * dst,
return 1;
}
-int32_t ec_lk_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
- int32_t op_ret, int32_t op_errno, struct gf_flock * flock,
- dict_t * xdata)
+int32_t
+ec_lk_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct gf_flock *flock, dict_t *xdata)
{
- ec_fop_data_t * fop = NULL;
- ec_cbk_data_t * cbk = NULL;
+ ec_fop_data_t *fop = NULL;
+ ec_cbk_data_t *cbk = NULL;
int32_t idx = (int32_t)(uintptr_t)cookie;
VALIDATE_OR_GOTO(this, out);
@@ -946,39 +920,32 @@ int32_t ec_lk_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
fop = frame->local;
- ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx,
- frame, op_ret, op_errno);
+ ec_trace("CBK", fop, "idx=%d, frame=%p, op_ret=%d, op_errno=%d", idx, frame,
+ op_ret, op_errno);
cbk = ec_cbk_data_allocate(frame, this, fop, GF_FOP_LK, idx, op_ret,
op_errno);
- if (cbk != NULL)
- {
- if (op_ret >= 0)
- {
- if (flock != NULL)
- {
+ if (cbk != NULL) {
+ if (op_ret >= 0) {
+ if (flock != NULL) {
cbk->flock.l_type = flock->l_type;
cbk->flock.l_whence = flock->l_whence;
cbk->flock.l_start = flock->l_start;
cbk->flock.l_len = flock->l_len;
cbk->flock.l_pid = flock->l_pid;
cbk->flock.l_owner.len = flock->l_owner.len;
- if (flock->l_owner.len > 0)
- {
+ if (flock->l_owner.len > 0) {
memcpy(cbk->flock.l_owner.data, flock->l_owner.data,
flock->l_owner.len);
}
}
}
- if (xdata != NULL)
- {
+ if (xdata != NULL) {
cbk->xdata = dict_ref(xdata);
- if (cbk->xdata == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ if (cbk->xdata == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
@@ -988,15 +955,15 @@ int32_t ec_lk_cbk(call_frame_t * frame, void * cookie, xlator_t * this,
}
out:
- if (fop != NULL)
- {
+ if (fop != NULL) {
ec_complete(fop);
}
return 0;
}
-void ec_wind_lk(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
+void
+ec_wind_lk(ec_t *ec, ec_fop_data_t *fop, int32_t idx)
{
ec_trace("WIND", fop, "idx=%d", idx);
@@ -1005,20 +972,19 @@ void ec_wind_lk(ec_t * ec, ec_fop_data_t * fop, int32_t idx)
fop->int32, &fop->flock, fop->xdata);
}
-int32_t ec_manager_lk(ec_fop_data_t * fop, int32_t state)
+int32_t
+ec_manager_lk(ec_fop_data_t *fop, int32_t state)
{
- ec_cbk_data_t * cbk;
+ ec_cbk_data_t *cbk;
- switch (state)
- {
+ switch (state) {
case EC_STATE_INIT:
- if ((fop->int32 == F_SETLKW) && (fop->flock.l_type != F_UNLCK))
- {
+ if ((fop->int32 == F_SETLKW) && (fop->flock.l_type != F_UNLCK)) {
fop->uint32 = EC_LOCK_MODE_ALL;
fop->int32 = F_SETLK;
}
- /* Fall through */
+ /* Fall through */
case EC_STATE_DISPATCH:
ec_dispatch_all(fop);
@@ -1030,7 +996,7 @@ int32_t ec_manager_lk(ec_fop_data_t * fop, int32_t state)
if (fop->flock.l_type != F_UNLCK) {
uintptr_t mask;
- ec_fop_set_error (fop, ec_lock_check(fop, &mask));
+ ec_fop_set_error(fop, ec_lock_check(fop, &mask));
if (fop->error != 0) {
if (mask != 0) {
struct gf_flock flock = {0};
@@ -1040,11 +1006,10 @@ int32_t ec_manager_lk(ec_fop_data_t * fop, int32_t state)
flock.l_start = fop->flock.l_start;
flock.l_len = fop->flock.l_len;
flock.l_pid = fop->flock.l_pid;
- lk_owner_copy (&flock.l_owner, &fop->flock.l_owner);
+ lk_owner_copy(&flock.l_owner, &fop->flock.l_owner);
- ec_lk(fop->frame, fop->xl, mask, 1,
- ec_lock_lk_unlocked, NULL, fop->fd, F_SETLK,
- &flock, fop->xdata);
+ ec_lk(fop->frame, fop->xl, mask, 1, ec_lock_lk_unlocked,
+ NULL, fop->fd, F_SETLK, &flock, fop->xdata);
}
if (fop->error < 0) {
@@ -1068,8 +1033,7 @@ int32_t ec_manager_lk(ec_fop_data_t * fop, int32_t state)
GF_ASSERT(cbk != NULL);
- if (fop->cbks.lk != NULL)
- {
+ if (fop->cbks.lk != NULL) {
fop->cbks.lk(fop->req_frame, fop, fop->xl, cbk->op_ret,
cbk->op_errno, &cbk->flock, cbk->xdata);
}
@@ -1081,40 +1045,36 @@ int32_t ec_manager_lk(ec_fop_data_t * fop, int32_t state)
case -EC_STATE_REPORT:
GF_ASSERT(fop->error != 0);
- if (fop->cbks.lk != NULL)
- {
- fop->cbks.lk(fop->req_frame, fop, fop->xl, -1, fop->error,
- NULL, NULL);
+ if (fop->cbks.lk != NULL) {
+ fop->cbks.lk(fop->req_frame, fop, fop->xl, -1, fop->error, NULL,
+ NULL);
}
-
return EC_STATE_END;
default:
- gf_msg (fop->xl->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_UNHANDLED_STATE,
- "Unhandled state %d for %s",
- state, ec_fop_name(fop->id));
+ gf_msg(fop->xl->name, GF_LOG_ERROR, EINVAL, EC_MSG_UNHANDLED_STATE,
+ "Unhandled state %d for %s", state, ec_fop_name(fop->id));
return EC_STATE_END;
}
}
-void ec_lk(call_frame_t * frame, xlator_t * this, uintptr_t target,
- int32_t minimum, fop_lk_cbk_t func, void * data, fd_t * fd,
- int32_t cmd, struct gf_flock * flock, dict_t * xdata)
+void
+ec_lk(call_frame_t *frame, xlator_t *this, uintptr_t target, uint32_t fop_flags,
+ fop_lk_cbk_t func, void *data, fd_t *fd, int32_t cmd,
+ struct gf_flock *flock, dict_t *xdata)
{
- ec_cbk_t callback = { .lk = func };
- ec_fop_data_t * fop = NULL;
+ ec_cbk_t callback = {.lk = func};
+ ec_fop_data_t *fop = NULL;
int32_t error = ENOMEM;
- gf_msg_trace ("ec", 0, "EC(LK) %p", frame);
+ gf_msg_trace("ec", 0, "EC(LK) %p", frame);
- VALIDATE_OR_GOTO(this, out);
GF_VALIDATE_OR_GOTO(this->name, frame, out);
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- fop = ec_fop_data_allocate(frame, this, GF_FOP_LK, 0, target, minimum,
+ fop = ec_fop_data_allocate(frame, this, GF_FOP_LK, 0, target, fop_flags,
ec_wind_lk, ec_manager_lk, callback, data);
if (fop == NULL) {
goto out;
@@ -1127,10 +1087,9 @@ void ec_lk(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (fd != NULL) {
fop->fd = fd_ref(fd);
if (fop->fd == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_FILE_DESC_REF_FAIL,
- "Failed to reference a "
- "file descriptor.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_FILE_DESC_REF_FAIL,
+ "Failed to reference a "
+ "file descriptor.");
goto out;
}
@@ -1150,10 +1109,9 @@ void ec_lk(call_frame_t * frame, xlator_t * this, uintptr_t target,
if (xdata != NULL) {
fop->xdata = dict_ref(xdata);
if (fop->xdata == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_DICT_REF_FAIL,
- "Failed to reference a "
- "dictionary.");
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_DICT_REF_FAIL,
+ "Failed to reference a "
+ "dictionary.");
goto out;
}
diff --git a/xlators/cluster/ec/src/ec-mem-types.h b/xlators/cluster/ec/src/ec-mem-types.h
index 8109a422d9d..3252c4c1c58 100644
--- a/xlators/cluster/ec/src/ec-mem-types.h
+++ b/xlators/cluster/ec/src/ec-mem-types.h
@@ -11,15 +11,13 @@
#ifndef __EC_MEM_TYPES_H__
#define __EC_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
-enum gf_ec_mem_types_
-{
+enum gf_ec_mem_types_ {
ec_mt_ec_t = gf_common_mt_end + 1,
ec_mt_xlator_t,
ec_mt_ec_inode_t,
ec_mt_ec_fd_t,
- ec_mt_ec_heal_t,
ec_mt_subvol_healer_t,
ec_mt_ec_gf_t,
ec_mt_ec_code_t,
diff --git a/xlators/cluster/ec/src/ec-messages.h b/xlators/cluster/ec/src/ec-messages.h
index f5138ad8e57..72e98f11286 100644
--- a/xlators/cluster/ec/src/ec-messages.h
+++ b/xlators/cluster/ec/src/ec-messages.h
@@ -11,7 +11,7 @@
#ifndef _EC_MESSAGES_H_
#define _EC_MESSAGES_H_
-#include "glfs-message-id.h"
+#include <glusterfs/glfs-message-id.h>
/* To add new message IDs, append new identifiers at the end of the list.
*
@@ -23,83 +23,39 @@
* glfs-message-id.h.
*/
-GLFS_MSGID(EC,
- EC_MSG_INVALID_CONFIG,
- EC_MSG_HEAL_FAIL,
- EC_MSG_DICT_COMBINE_FAIL,
- EC_MSG_STIME_COMBINE_FAIL,
- EC_MSG_INVALID_DICT_NUMS,
- EC_MSG_IATT_COMBINE_FAIL,
- EC_MSG_INVALID_FORMAT,
- EC_MSG_DICT_GET_FAILED,
- EC_MSG_UNHANDLED_STATE,
- EC_MSG_FILE_DESC_REF_FAIL,
- EC_MSG_LOC_COPY_FAIL,
- EC_MSG_BUF_REF_FAIL,
- EC_MSG_DICT_REF_FAIL,
- EC_MSG_LK_UNLOCK_FAILED,
- EC_MSG_UNLOCK_FAILED,
- EC_MSG_LOC_PARENT_INODE_MISSING,
- EC_MSG_INVALID_LOC_NAME,
- EC_MSG_NO_MEMORY,
- EC_MSG_GFID_MISMATCH,
- EC_MSG_UNSUPPORTED_VERSION,
- EC_MSG_FD_CREATE_FAIL,
- EC_MSG_READDIRP_REQ_PREP_FAIL,
- EC_MSG_LOOKUP_REQ_PREP_FAIL,
- EC_MSG_INODE_REF_FAIL,
- EC_MSG_LOOKUP_READAHEAD_FAIL,
- EC_MSG_FRAME_MISMATCH,
- EC_MSG_XLATOR_MISMATCH,
- EC_MSG_VECTOR_MISMATCH,
- EC_MSG_IATT_MISMATCH,
- EC_MSG_FD_MISMATCH,
- EC_MSG_DICT_MISMATCH,
- EC_MSG_INDEX_DIR_GET_FAIL,
- EC_MSG_PREOP_LOCK_FAILED,
- EC_MSG_CHILDS_INSUFFICIENT,
- EC_MSG_OP_EXEC_UNAVAIL,
- EC_MSG_UNLOCK_DELAY_FAILED,
- EC_MSG_SIZE_VERS_UPDATE_FAIL,
- EC_MSG_INVALID_REQUEST,
- EC_MSG_INVALID_LOCK_TYPE,
- EC_MSG_SIZE_VERS_GET_FAIL,
- EC_MSG_FILE_SIZE_GET_FAIL,
- EC_MSG_FOP_MISMATCH,
- EC_MSG_SUBVOL_ID_DICT_SET_FAIL,
- EC_MSG_SUBVOL_BUILD_FAIL,
- EC_MSG_XLATOR_INIT_FAIL,
- EC_MSG_NO_PARENTS,
- EC_MSG_TIMER_CREATE_FAIL,
- EC_MSG_TOO_MANY_SUBVOLS,
- EC_MSG_DATA_UNAVAILABLE,
- EC_MSG_INODE_REMOVE_FAIL,
- EC_MSG_INVALID_REDUNDANCY,
- EC_MSG_XLATOR_PARSE_OPT_FAIL,
- EC_MSG_OP_FAIL_ON_SUBVOLS,
- EC_MSG_INVALID_INODE,
- EC_MSG_LOCK_MISMATCH,
- EC_MSG_XDATA_MISMATCH,
- EC_MSG_HEALING_INFO,
- EC_MSG_HEAL_SUCCESS,
- EC_MSG_FULL_SWEEP_START,
- EC_MSG_FULL_SWEEP_STOP,
- EC_MSG_INVALID_FOP,
- EC_MSG_EC_UP,
- EC_MSG_EC_DOWN,
- EC_MSG_SIZE_XATTR_GET_FAIL,
- EC_MSG_VER_XATTR_GET_FAIL,
- EC_MSG_CONFIG_XATTR_GET_FAIL,
- EC_MSG_CONFIG_XATTR_INVALID,
- EC_MSG_EXTENSION,
- EC_MSG_EXTENSION_NONE,
- EC_MSG_EXTENSION_UNKNOWN,
- EC_MSG_EXTENSION_UNSUPPORTED,
- EC_MSG_EXTENSION_FAILED,
- EC_MSG_NO_GF,
- EC_MSG_MATRIX_FAILED,
- EC_MSG_DYN_CREATE_FAILED,
- EC_MSG_DYN_CODEGEN_FAILED
-);
+GLFS_MSGID(EC, EC_MSG_INVALID_CONFIG, EC_MSG_HEAL_FAIL,
+ EC_MSG_DICT_COMBINE_FAIL, EC_MSG_STIME_COMBINE_FAIL,
+ EC_MSG_INVALID_DICT_NUMS, EC_MSG_IATT_COMBINE_FAIL,
+ EC_MSG_INVALID_FORMAT, EC_MSG_DICT_GET_FAILED,
+ EC_MSG_UNHANDLED_STATE, EC_MSG_FILE_DESC_REF_FAIL,
+ EC_MSG_LOC_COPY_FAIL, EC_MSG_BUF_REF_FAIL, EC_MSG_DICT_REF_FAIL,
+ EC_MSG_LK_UNLOCK_FAILED, EC_MSG_UNLOCK_FAILED,
+ EC_MSG_LOC_PARENT_INODE_MISSING, EC_MSG_INVALID_LOC_NAME,
+ EC_MSG_NO_MEMORY, EC_MSG_GFID_MISMATCH, EC_MSG_UNSUPPORTED_VERSION,
+ EC_MSG_FD_CREATE_FAIL, EC_MSG_READDIRP_REQ_PREP_FAIL,
+ EC_MSG_LOOKUP_REQ_PREP_FAIL, EC_MSG_INODE_REF_FAIL,
+ EC_MSG_LOOKUP_READAHEAD_FAIL, EC_MSG_FRAME_MISMATCH,
+ EC_MSG_XLATOR_MISMATCH, EC_MSG_VECTOR_MISMATCH, EC_MSG_IATT_MISMATCH,
+ EC_MSG_FD_MISMATCH, EC_MSG_DICT_MISMATCH, EC_MSG_INDEX_DIR_GET_FAIL,
+ EC_MSG_PREOP_LOCK_FAILED, EC_MSG_CHILDS_INSUFFICIENT,
+ EC_MSG_OP_EXEC_UNAVAIL, EC_MSG_UNLOCK_DELAY_FAILED,
+ EC_MSG_SIZE_VERS_UPDATE_FAIL, EC_MSG_INVALID_REQUEST,
+ EC_MSG_INVALID_LOCK_TYPE, EC_MSG_SIZE_VERS_GET_FAIL,
+ EC_MSG_FILE_SIZE_GET_FAIL, EC_MSG_FOP_MISMATCH,
+ EC_MSG_SUBVOL_ID_DICT_SET_FAIL, EC_MSG_SUBVOL_BUILD_FAIL,
+ EC_MSG_XLATOR_INIT_FAIL, EC_MSG_NO_PARENTS, EC_MSG_TIMER_CREATE_FAIL,
+ EC_MSG_TOO_MANY_SUBVOLS, EC_MSG_DATA_UNAVAILABLE,
+ EC_MSG_INODE_REMOVE_FAIL, EC_MSG_INVALID_REDUNDANCY,
+ EC_MSG_XLATOR_PARSE_OPT_FAIL, EC_MSG_OP_FAIL_ON_SUBVOLS,
+ EC_MSG_INVALID_INODE, EC_MSG_LOCK_MISMATCH, EC_MSG_XDATA_MISMATCH,
+ EC_MSG_HEALING_INFO, EC_MSG_HEAL_SUCCESS, EC_MSG_FULL_SWEEP_START,
+ EC_MSG_FULL_SWEEP_STOP, EC_MSG_INVALID_FOP, EC_MSG_EC_UP,
+ EC_MSG_EC_DOWN, EC_MSG_SIZE_XATTR_GET_FAIL,
+ EC_MSG_VER_XATTR_GET_FAIL, EC_MSG_CONFIG_XATTR_GET_FAIL,
+ EC_MSG_CONFIG_XATTR_INVALID, EC_MSG_EXTENSION, EC_MSG_EXTENSION_NONE,
+ EC_MSG_EXTENSION_UNKNOWN, EC_MSG_EXTENSION_UNSUPPORTED,
+ EC_MSG_EXTENSION_FAILED, EC_MSG_NO_GF, EC_MSG_MATRIX_FAILED,
+ EC_MSG_DYN_CREATE_FAILED, EC_MSG_DYN_CODEGEN_FAILED,
+ EC_MSG_THREAD_CLEANUP_FAILED, EC_MSG_FD_BAD);
#endif /* !_EC_MESSAGES_H_ */
diff --git a/xlators/cluster/ec/src/ec-method.c b/xlators/cluster/ec/src/ec-method.c
index fdc89391bfb..55faed0b193 100644
--- a/xlators/cluster/ec/src/ec-method.c
+++ b/xlators/cluster/ec/src/ec-method.c
@@ -88,11 +88,9 @@ ec_method_matrix_init(ec_matrix_list_t *list, ec_matrix_t *matrix,
matrix->rows);
for (i = 0; i < matrix->rows; i++) {
matrix->row_data[i].values = matrix->values + i * matrix->columns;
- matrix->row_data[i].func.interleaved =
- ec_code_build_interleaved(matrix->code,
- EC_METHOD_WORD_SIZE,
- matrix->row_data[i].values,
- matrix->columns);
+ matrix->row_data[i].func.interleaved = ec_code_build_interleaved(
+ matrix->code, EC_METHOD_WORD_SIZE, matrix->row_data[i].values,
+ matrix->columns);
}
} else {
matrix->rows = list->rows;
@@ -100,10 +98,9 @@ ec_method_matrix_init(ec_matrix_list_t *list, ec_matrix_t *matrix,
matrix->columns, rows, matrix->rows);
for (i = 0; i < matrix->rows; i++) {
matrix->row_data[i].values = matrix->values + i * matrix->columns;
- matrix->row_data[i].func.linear =
- ec_code_build_linear(matrix->code, EC_METHOD_WORD_SIZE,
- matrix->row_data[i].values,
- matrix->columns);
+ matrix->row_data[i].func.linear = ec_code_build_linear(
+ matrix->code, EC_METHOD_WORD_SIZE, matrix->row_data[i].values,
+ matrix->columns);
}
}
}
@@ -266,8 +263,8 @@ ec_method_setup(xlator_t *xl, ec_matrix_list_t *list, const char *gen)
int32_t err;
matrix = GF_MALLOC(sizeof(ec_matrix_t) +
- sizeof(ec_matrix_row_t) * list->rows +
- sizeof(uint32_t) * list->columns * list->rows,
+ sizeof(ec_matrix_row_t) * list->rows +
+ sizeof(uint32_t) * list->columns * list->rows,
ec_mt_ec_matrix_t);
if (matrix == NULL) {
err = -ENOMEM;
@@ -310,9 +307,10 @@ ec_method_init(xlator_t *xl, ec_matrix_list_t *list, uint32_t columns,
INIT_LIST_HEAD(&list->lru);
int32_t err;
- list->pool = mem_pool_new_fn(xl->ctx, sizeof(ec_matrix_t) +
- sizeof(ec_matrix_row_t) * columns +
- sizeof(uint32_t) * columns * columns,
+ list->pool = mem_pool_new_fn(xl->ctx,
+ sizeof(ec_matrix_t) +
+ sizeof(ec_matrix_row_t) * columns +
+ sizeof(uint32_t) * columns * columns,
128, "ec_matrix_t");
if (list->pool == NULL) {
err = -ENOMEM;
@@ -370,8 +368,8 @@ ec_method_fini(ec_matrix_list_t *list)
GF_ASSERT(list->count == 0);
- if (list->pool)/*Init was successful*/
- LOCK_DESTROY(&list->lock);
+ if (list->pool) /*Init was successful*/
+ LOCK_DESTROY(&list->lock);
ec_method_matrix_release(list->encode);
GF_FREE(list->encode);
@@ -393,29 +391,28 @@ ec_method_update(xlator_t *xl, ec_matrix_list_t *list, const char *gen)
}
void
-ec_method_encode(ec_matrix_list_t *list, size_t size, void *in, void **out)
+ec_method_encode(ec_matrix_list_t *list, uint64_t size, void *in, void **out)
{
ec_matrix_t *matrix;
- size_t pos;
+ uint64_t pos;
uint32_t i;
matrix = list->encode;
for (pos = 0; pos < size; pos += list->stripe) {
for (i = 0; i < matrix->rows; i++) {
- matrix->row_data[i].func.linear(out[i], in, pos,
- matrix->row_data[i].values,
- list->columns);
+ matrix->row_data[i].func.linear(
+ out[i], in, pos, matrix->row_data[i].values, list->columns);
out[i] += EC_METHOD_CHUNK_SIZE;
}
}
}
int32_t
-ec_method_decode(ec_matrix_list_t *list, size_t size, uintptr_t mask,
+ec_method_decode(ec_matrix_list_t *list, uint64_t size, uintptr_t mask,
uint32_t *rows, void **in, void *out)
{
ec_matrix_t *matrix;
- size_t pos;
+ uint64_t pos;
uint32_t i;
matrix = ec_method_matrix_get(list, mask, rows);
@@ -424,9 +421,8 @@ ec_method_decode(ec_matrix_list_t *list, size_t size, uintptr_t mask,
}
for (pos = 0; pos < size; pos += EC_METHOD_CHUNK_SIZE) {
for (i = 0; i < matrix->rows; i++) {
- matrix->row_data[i].func.interleaved(out, in, pos,
- matrix->row_data[i].values,
- list->columns);
+ matrix->row_data[i].func.interleaved(
+ out, in, pos, matrix->row_data[i].values, list->columns);
out += EC_METHOD_CHUNK_SIZE;
}
}
diff --git a/xlators/cluster/ec/src/ec-method.h b/xlators/cluster/ec/src/ec-method.h
index 9ba5069ff0e..f91233b2f88 100644
--- a/xlators/cluster/ec/src/ec-method.h
+++ b/xlators/cluster/ec/src/ec-method.h
@@ -11,8 +11,6 @@
#ifndef __EC_METHOD_H__
#define __EC_METHOD_H__
-#include "xlator.h"
-
#include "ec-types.h"
#include "ec-galois.h"
@@ -24,7 +22,7 @@
/* Determines the maximum size of the matrix used to encode/decode data */
#define EC_METHOD_MAX_FRAGMENTS 16
/* Determines the maximum number of usable elements in the Galois Field */
-#define EC_METHOD_MAX_NODES (EC_GF_SIZE - 1)
+#define EC_METHOD_MAX_NODES (EC_GF_SIZE - 1)
#define EC_METHOD_WORD_SIZE 64
@@ -34,16 +32,17 @@ int32_t
ec_method_init(xlator_t *xl, ec_matrix_list_t *list, uint32_t columns,
uint32_t rows, uint32_t max, const char *gen);
-void ec_method_fini(ec_matrix_list_t *list);
+void
+ec_method_fini(ec_matrix_list_t *list);
int32_t
ec_method_update(xlator_t *xl, ec_matrix_list_t *list, const char *gen);
void
-ec_method_encode(ec_matrix_list_t *list, size_t size, void *in, void **out);
+ec_method_encode(ec_matrix_list_t *list, uint64_t size, void *in, void **out);
int32_t
-ec_method_decode(ec_matrix_list_t *list, size_t size, uintptr_t mask,
+ec_method_decode(ec_matrix_list_t *list, uint64_t size, uintptr_t mask,
uint32_t *rows, void **in, void *out);
#endif /* __EC_METHOD_H__ */
diff --git a/xlators/cluster/ec/src/ec-types.h b/xlators/cluster/ec/src/ec-types.h
index 15b4c77abfe..de9b89bb2c9 100644
--- a/xlators/cluster/ec/src/ec-types.h
+++ b/xlators/cluster/ec/src/ec-types.h
@@ -11,10 +11,10 @@
#ifndef __EC_TYPES_H__
#define __EC_TYPES_H__
-#include "xlator.h"
-#include "timer.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/timer.h>
#include "libxlator.h"
-#include "atomic.h"
+#include <glusterfs/atomic.h>
#define EC_GF_MAX_REGS 16
@@ -128,182 +128,175 @@ typedef void (*ec_wind_f)(ec_t *, ec_fop_data_t *, int32_t);
typedef int32_t (*ec_handler_f)(ec_fop_data_t *, int32_t);
typedef void (*ec_resume_f)(ec_fop_data_t *, int32_t);
-enum _ec_read_policy {
- EC_ROUND_ROBIN,
- EC_GFID_HASH,
- EC_READ_POLICY_MAX
-};
+enum _ec_read_policy { EC_ROUND_ROBIN, EC_GFID_HASH, EC_READ_POLICY_MAX };
enum _ec_heal_need {
- EC_HEAL_NONEED,
- EC_HEAL_MAYBE,
- EC_HEAL_MUST
+ EC_HEAL_NONEED,
+ EC_HEAL_MAYBE,
+ EC_HEAL_MUST,
+ EC_HEAL_PURGE_INDEX
};
-enum _ec_stripe_part {
- EC_STRIPE_HEAD,
- EC_STRIPE_TAIL
-};
+enum _ec_stripe_part { EC_STRIPE_HEAD, EC_STRIPE_TAIL };
/* Enumartions to indicate FD status. */
-typedef enum {
- EC_FD_NOT_OPENED,
- EC_FD_OPENED,
- EC_FD_OPENING
-} ec_fd_status_t;
+typedef enum { EC_FD_NOT_OPENED, EC_FD_OPENED, EC_FD_OPENING } ec_fd_status_t;
struct _ec_config {
uint32_t version;
- uint8_t algorithm;
- uint8_t gf_word_size;
- uint8_t bricks;
- uint8_t redundancy;
+ uint8_t algorithm;
+ uint8_t gf_word_size;
+ uint8_t bricks;
+ uint8_t redundancy;
uint32_t chunk_size;
};
struct _ec_fd {
- loc_t loc;
+ loc_t loc;
uintptr_t open;
- int32_t flags;
+ int32_t flags;
+ uint64_t bad_version;
ec_fd_status_t fd_status[0];
};
struct _ec_stripe {
- struct list_head lru; /* LRU list member */
- uint64_t frag_offset; /* Fragment offset of this stripe */
- char data[]; /* Contents of the stripe */
+ struct list_head lru; /* LRU list member */
+ uint64_t frag_offset; /* Fragment offset of this stripe */
+ char data[]; /* Contents of the stripe */
};
struct _ec_stripe_list {
- struct list_head lru;
- uint32_t count;
- uint32_t max;
+ struct list_head lru;
+ uint32_t count;
+ uint32_t max;
};
struct _ec_inode {
- ec_lock_t *inode_lock;
- gf_boolean_t have_info;
- gf_boolean_t have_config;
- gf_boolean_t have_version;
- gf_boolean_t have_size;
- ec_config_t config;
- uint64_t pre_version[2];
- uint64_t post_version[2];
- uint64_t pre_size;
- uint64_t post_size;
- uint64_t dirty[2];
- struct list_head heal;
- ec_stripe_list_t stripe_cache;
+ ec_lock_t *inode_lock;
+ gf_boolean_t have_info;
+ gf_boolean_t have_config;
+ gf_boolean_t have_version;
+ gf_boolean_t have_size;
+ int32_t heal_count;
+ ec_config_t config;
+ uint64_t pre_version[2];
+ uint64_t post_version[2];
+ uint64_t pre_size;
+ uint64_t post_size;
+ uint64_t dirty[2];
+ struct list_head heal;
+ ec_stripe_list_t stripe_cache;
+ uint64_t bad_version;
};
-
typedef int32_t (*fop_heal_cbk_t)(call_frame_t *, void *, xlator_t *, int32_t,
int32_t, uintptr_t, uintptr_t, uintptr_t,
- dict_t *);
+ uint32_t, dict_t *);
typedef int32_t (*fop_fheal_cbk_t)(call_frame_t *, void *, xlator_t *, int32_t,
int32_t, uintptr_t, uintptr_t, uintptr_t,
- dict_t *);
+ uint32_t, dict_t *);
union _ec_cbk {
- fop_access_cbk_t access;
- fop_create_cbk_t create;
- fop_discard_cbk_t discard;
- fop_entrylk_cbk_t entrylk;
- fop_fentrylk_cbk_t fentrylk;
- fop_fallocate_cbk_t fallocate;
- fop_flush_cbk_t flush;
- fop_fsync_cbk_t fsync;
- fop_fsyncdir_cbk_t fsyncdir;
- fop_getxattr_cbk_t getxattr;
- fop_fgetxattr_cbk_t fgetxattr;
- fop_heal_cbk_t heal;
- fop_fheal_cbk_t fheal;
- fop_inodelk_cbk_t inodelk;
- fop_finodelk_cbk_t finodelk;
- fop_link_cbk_t link;
- fop_lk_cbk_t lk;
- fop_lookup_cbk_t lookup;
- fop_mkdir_cbk_t mkdir;
- fop_mknod_cbk_t mknod;
- fop_open_cbk_t open;
- fop_opendir_cbk_t opendir;
- fop_readdir_cbk_t readdir;
- fop_readdirp_cbk_t readdirp;
- fop_readlink_cbk_t readlink;
- fop_readv_cbk_t readv;
- fop_removexattr_cbk_t removexattr;
+ fop_access_cbk_t access;
+ fop_create_cbk_t create;
+ fop_discard_cbk_t discard;
+ fop_entrylk_cbk_t entrylk;
+ fop_fentrylk_cbk_t fentrylk;
+ fop_fallocate_cbk_t fallocate;
+ fop_flush_cbk_t flush;
+ fop_fsync_cbk_t fsync;
+ fop_fsyncdir_cbk_t fsyncdir;
+ fop_getxattr_cbk_t getxattr;
+ fop_fgetxattr_cbk_t fgetxattr;
+ fop_heal_cbk_t heal;
+ fop_fheal_cbk_t fheal;
+ fop_inodelk_cbk_t inodelk;
+ fop_finodelk_cbk_t finodelk;
+ fop_link_cbk_t link;
+ fop_lk_cbk_t lk;
+ fop_lookup_cbk_t lookup;
+ fop_mkdir_cbk_t mkdir;
+ fop_mknod_cbk_t mknod;
+ fop_open_cbk_t open;
+ fop_opendir_cbk_t opendir;
+ fop_readdir_cbk_t readdir;
+ fop_readdirp_cbk_t readdirp;
+ fop_readlink_cbk_t readlink;
+ fop_readv_cbk_t readv;
+ fop_removexattr_cbk_t removexattr;
fop_fremovexattr_cbk_t fremovexattr;
- fop_rename_cbk_t rename;
- fop_rmdir_cbk_t rmdir;
- fop_setattr_cbk_t setattr;
- fop_fsetattr_cbk_t fsetattr;
- fop_setxattr_cbk_t setxattr;
- fop_fsetxattr_cbk_t fsetxattr;
- fop_stat_cbk_t stat;
- fop_fstat_cbk_t fstat;
- fop_statfs_cbk_t statfs;
- fop_symlink_cbk_t symlink;
- fop_truncate_cbk_t truncate;
- fop_ftruncate_cbk_t ftruncate;
- fop_unlink_cbk_t unlink;
- fop_writev_cbk_t writev;
- fop_xattrop_cbk_t xattrop;
- fop_fxattrop_cbk_t fxattrop;
- fop_zerofill_cbk_t zerofill;
- fop_seek_cbk_t seek;
- fop_ipc_cbk_t ipc;
+ fop_rename_cbk_t rename;
+ fop_rmdir_cbk_t rmdir;
+ fop_setattr_cbk_t setattr;
+ fop_fsetattr_cbk_t fsetattr;
+ fop_setxattr_cbk_t setxattr;
+ fop_fsetxattr_cbk_t fsetxattr;
+ fop_stat_cbk_t stat;
+ fop_fstat_cbk_t fstat;
+ fop_statfs_cbk_t statfs;
+ fop_symlink_cbk_t symlink;
+ fop_truncate_cbk_t truncate;
+ fop_ftruncate_cbk_t ftruncate;
+ fop_unlink_cbk_t unlink;
+ fop_writev_cbk_t writev;
+ fop_xattrop_cbk_t xattrop;
+ fop_fxattrop_cbk_t fxattrop;
+ fop_zerofill_cbk_t zerofill;
+ fop_seek_cbk_t seek;
+ fop_ipc_cbk_t ipc;
};
struct _ec_lock {
- ec_inode_t *ctx;
- gf_timer_t *timer;
+ ec_inode_t *ctx;
+ gf_timer_t *timer;
/* List of owners of this lock. All fops added to this list are running
* concurrently. */
- struct list_head owners;
+ struct list_head owners;
/* List of fops waiting to be an owner of the lock. Fops are added to this
* list when the current owner has an incompatible access (conflicting lock)
* or the lock is not acquired yet. */
- struct list_head waiting;
+ struct list_head waiting;
/* List of fops that will wait until the next unlock/lock cycle. This
* happens when the currently acquired lock is decided to be released as
* soon as possible. In this case, all frozen fops will be continued only
* after the lock is reacquired. */
- struct list_head frozen;
-
- uintptr_t mask;
- uintptr_t good_mask;
- uintptr_t healing;
- uint32_t refs_owners; /* Refs for fops owning the lock */
- uint32_t refs_pending; /* Refs assigned to fops being prepared */
- uint32_t waiting_flags; /*Track xattrop/dirty marking*/
- gf_boolean_t acquired;
- gf_boolean_t unlock_now;
- gf_boolean_t release;
- gf_boolean_t query;
- fd_t *fd;
- loc_t loc;
+ struct list_head frozen;
+
+ uintptr_t mask;
+ uintptr_t good_mask;
+ uintptr_t healing;
+ uint32_t refs_owners; /* Refs for fops owning the lock */
+ uint32_t refs_pending; /* Refs assigned to fops being prepared */
+ uint32_t waiting_flags; /*Track xattrop/dirty marking*/
+ gf_boolean_t acquired;
+ gf_boolean_t contention;
+ gf_boolean_t unlock_now;
+ gf_boolean_t release;
+ gf_boolean_t query;
+ fd_t *fd;
+ loc_t loc;
union {
- entrylk_type type;
- struct gf_flock flock;
+ entrylk_type type;
+ struct gf_flock flock;
};
};
struct _ec_lock_link {
- ec_lock_t *lock;
- ec_fop_data_t *fop;
- struct list_head owner_list;
- struct list_head wait_list;
- gf_boolean_t update[2];
- gf_boolean_t dirty[2];
- gf_boolean_t optimistic_changelog;
- loc_t *base;
- uint64_t size;
- uint32_t waiting_flags;
- off_t fl_start;
- off_t fl_end;
+ ec_lock_t *lock;
+ ec_fop_data_t *fop;
+ struct list_head owner_list;
+ struct list_head wait_list;
+ gf_boolean_t update[2];
+ gf_boolean_t dirty[2];
+ gf_boolean_t optimistic_changelog;
+ loc_t *base;
+ uint64_t size;
+ uint32_t waiting_flags;
+ off_t fl_start;
+ off_t fl_end;
};
/* This structure keeps a range of fragment offsets affected by a fop. Since
@@ -320,111 +313,114 @@ struct _ec_fragment_range {
/* EC xlator data structure to collect all the data required to perform
* the file operation.*/
struct _ec_fop_data {
- int32_t id; /* ID of the file operation */
- int32_t refs;
- int32_t state;
- int32_t minimum; /* Mininum number of successful
- operation required to conclude a
- fop as successful */
- int32_t expected;
- int32_t winds;
- int32_t jobs;
- int32_t error;
- ec_fop_data_t *parent;
- xlator_t *xl; /* points to EC xlator */
- call_frame_t *req_frame; /* frame of the calling xlator */
- call_frame_t *frame; /* frame used by this fop */
- struct list_head cbk_list; /* sorted list of groups of answers */
- struct list_head answer_list; /* list of answers */
- struct list_head pending_list; /* member of ec_t.pending_fops */
- ec_cbk_data_t *answer; /* accepted answer */
- int32_t lock_count;
- int32_t locked;
- ec_lock_link_t locks[2];
- int32_t first_lock;
- gf_lock_t lock;
-
- uint32_t flags;
- uint32_t first;
- uintptr_t mask;
- uintptr_t healing; /*Dispatch is done but call is successful only
- if fop->minimum number of subvolumes succeed
- which are not healing*/
- uintptr_t remaining;
- uintptr_t received; /* Mask of responses */
- uintptr_t good;
-
- uid_t uid;
- gid_t gid;
-
- ec_wind_f wind; /* Function to wind to */
- ec_handler_f handler; /* FOP manager function */
- ec_resume_f resume;
- ec_cbk_t cbks; /* Callback function for this FOP */
- void *data;
- ec_heal_t *heal;
- struct list_head healer;
-
- uint64_t user_size;
- uint32_t head;
-
- int32_t use_fd; /* Indicates whether this FOP uses FD or
- not */
-
- dict_t *xdata;
- dict_t *dict;
- int32_t int32;
- uint32_t uint32;
- uint64_t size;
- off_t offset;
- mode_t mode[2];
- entrylk_cmd entrylk_cmd;
- entrylk_type entrylk_type;
+ int32_t id; /* ID of the file operation */
+ int32_t refs;
+ int32_t state;
+ uint32_t minimum; /* Minimum number of successful
+ operation required to conclude a
+ fop as successful */
+ int32_t expected;
+ int32_t winds;
+ int32_t jobs;
+ int32_t error;
+ ec_fop_data_t *parent;
+ xlator_t *xl; /* points to EC xlator */
+ call_frame_t *req_frame; /* frame of the calling xlator */
+ call_frame_t *frame; /* frame used by this fop */
+ struct list_head cbk_list; /* sorted list of groups of answers */
+ struct list_head answer_list; /* list of answers */
+ struct list_head pending_list; /* member of ec_t.pending_fops */
+ ec_cbk_data_t *answer; /* accepted answer */
+ int32_t lock_count;
+ int32_t locked;
+ gf_lock_t lock;
+ ec_lock_link_t locks[2];
+ int32_t first_lock;
+
+ uint32_t fop_flags; /* Flags passed by the caller. */
+ uint32_t flags; /* Internal flags. */
+ uint32_t first;
+ uintptr_t mask;
+ uintptr_t healing; /*Dispatch is done but call is successful only
+ if fop->minimum number of subvolumes succeed
+ which are not healing*/
+ uintptr_t remaining;
+ uintptr_t received; /* Mask of responses */
+ uintptr_t good;
+
+ uid_t uid;
+ gid_t gid;
+
+ ec_wind_f wind; /* Function to wind to */
+ ec_handler_f handler; /* FOP manager function */
+ ec_resume_f resume;
+ ec_cbk_t cbks; /* Callback function for this FOP */
+ void *data;
+ ec_heal_t *heal;
+ struct list_head healer;
+
+ uint64_t user_size;
+ uint32_t head;
+
+ int32_t use_fd; /* Indicates whether this FOP uses FD or
+ not */
+
+ dict_t *xdata;
+ dict_t *dict;
+ int32_t int32;
+ uint32_t uint32;
+ uint64_t size;
+ off_t offset;
+ mode_t mode[2];
+ entrylk_cmd entrylk_cmd;
+ entrylk_type entrylk_type;
gf_xattrop_flags_t xattrop_flags;
- dev_t dev;
- inode_t *inode;
- fd_t *fd; /* FD of the file on which FOP is
- being carried upon */
- struct iatt iatt;
- char *str[2];
- loc_t loc[2]; /* Holds the location details for
- the file */
- struct gf_flock flock;
- struct iovec *vector;
- struct iobref *buffers;
- gf_seek_what_t seek;
- ec_fragment_range_t frag_range; /* This will hold the range of stripes
- affected by the fop. */
+ dev_t dev;
+ inode_t *inode;
+ fd_t *fd; /* FD of the file on which FOP is
+ being carried upon */
+ struct iatt iatt;
+ char *str[2];
+ loc_t loc[2]; /* Holds the location details for
+ the file */
+ struct gf_flock flock;
+ struct iovec *vector;
+ struct iobref *buffers;
+ gf_seek_what_t seek;
+ ec_fragment_range_t frag_range; /* This will hold the range of stripes
+ affected by the fop. */
+ char *errstr; /*String of fop name, path and gfid
+ to be used in gf_msg. */
};
struct _ec_cbk_data {
- struct list_head list; /* item in the sorted list of groups */
- struct list_head answer_list; /* item in the list of answers */
- ec_fop_data_t *fop;
- ec_cbk_data_t *next; /* next answer in the same group */
- uint32_t idx;
- int32_t op_ret;
- int32_t op_errno;
- int32_t count;
- uintptr_t mask;
-
- dict_t *xdata;
- dict_t *dict;
- int32_t int32;
- uintptr_t uintptr[3];
- uint64_t size;
- uint64_t version[2];
- inode_t *inode;
- fd_t *fd;
- struct statvfs statvfs;
- struct iatt iatt[5];
- struct gf_flock flock;
- struct iovec *vector;
- struct iobref *buffers;
- char *str;
- gf_dirent_t entries;
- off_t offset;
- gf_seek_what_t what;
+ struct list_head list; /* item in the sorted list of groups */
+ struct list_head answer_list; /* item in the list of answers */
+ ec_fop_data_t *fop;
+ ec_cbk_data_t *next; /* next answer in the same group */
+ uint32_t idx;
+ int32_t op_ret;
+ int32_t op_errno;
+ int32_t count;
+ uintptr_t mask;
+
+ dict_t *xdata;
+ dict_t *dict;
+ int32_t int32;
+ uintptr_t uintptr[3];
+ uint64_t size;
+ uint64_t version[2];
+ inode_t *inode;
+ fd_t *fd;
+ struct statvfs statvfs;
+ struct iatt iatt[5];
+ struct gf_flock flock;
+ struct iovec *vector;
+ struct iobref *buffers;
+ char *str;
+ gf_dirent_t entries;
+ off_t offset;
+ gf_seek_what_t what;
};
enum _ec_gf_opcode {
@@ -439,26 +435,26 @@ enum _ec_gf_opcode {
struct _ec_gf_op {
ec_gf_opcode_t op;
- uint32_t arg1;
- uint32_t arg2;
- uint32_t arg3;
+ uint32_t arg1;
+ uint32_t arg2;
+ uint32_t arg3;
};
struct _ec_gf_mul {
- uint32_t regs;
- uint32_t map[EC_GF_MAX_REGS];
+ uint32_t regs;
+ uint32_t map[EC_GF_MAX_REGS];
ec_gf_op_t *ops;
};
struct _ec_gf {
- uint32_t bits;
- uint32_t size;
- uint32_t mod;
- uint32_t min_ops;
- uint32_t max_ops;
- uint32_t avg_ops;
- uint32_t *log;
- uint32_t *pow;
+ uint32_t bits;
+ uint32_t size;
+ uint32_t mod;
+ uint32_t min_ops;
+ uint32_t max_ops;
+ uint32_t avg_ops;
+ uint32_t *log;
+ uint32_t *pow;
ec_gf_mul_t **table;
};
@@ -481,208 +477,214 @@ struct _ec_code_gen {
};
struct _ec_code {
- gf_lock_t lock;
- struct list_head spaces;
- ec_gf_t *gf;
- ec_code_gen_t *gen;
+ gf_lock_t lock;
+ struct list_head spaces;
+ ec_gf_t *gf;
+ ec_code_gen_t *gen;
};
struct _ec_code_arg {
- uint32_t value;
+ uint32_t value;
};
struct _ec_code_op {
- ec_gf_opcode_t op;
- ec_code_arg_t arg1;
- ec_code_arg_t arg2;
- ec_code_arg_t arg3;
+ ec_gf_opcode_t op;
+ ec_code_arg_t arg1;
+ ec_code_arg_t arg2;
+ ec_code_arg_t arg3;
};
struct _ec_code_builder {
- ec_code_t *code;
- uint64_t address;
- uint8_t *data;
- uint32_t size;
- int32_t error;
- uint32_t regs;
- uint32_t bits;
- uint32_t width;
- uint32_t count;
- uint32_t base;
- uint32_t map[EC_GF_MAX_REGS];
- gf_boolean_t linear;
- uint64_t loop;
- ec_code_op_t ops[0];
+ ec_code_t *code;
+ uint64_t address;
+ uint8_t *data;
+ uint32_t size;
+ int32_t error;
+ uint32_t regs;
+ uint32_t bits;
+ uint32_t width;
+ uint32_t count;
+ uint32_t base;
+ uint32_t map[EC_GF_MAX_REGS];
+ gf_boolean_t linear;
+ uint64_t loop;
+ ec_code_op_t ops[0];
};
struct _ec_code_chunk {
- struct list_head list;
- size_t size;
- ec_code_space_t *space;
+ struct list_head list;
+ size_t size;
+ ec_code_space_t *space;
};
struct _ec_code_space {
- struct list_head list;
- struct list_head chunks;
- ec_code_t *code;
- void *exec;
- size_t size;
+ struct list_head list;
+ struct list_head chunks;
+ ec_code_t *code;
+ void *exec;
+ size_t size;
};
-
union _ec_code_func {
- ec_code_func_linear_t linear;
+ ec_code_func_linear_t linear;
ec_code_func_interleaved_t interleaved;
};
struct _ec_matrix_row {
- ec_code_func_t func;
- uint32_t *values;
+ ec_code_func_t func;
+ uint32_t *values;
};
struct _ec_matrix {
struct list_head lru;
- uint32_t refs;
- uint32_t columns;
- uint32_t rows;
- uintptr_t mask;
- ec_code_t *code;
- uint32_t *values;
- ec_matrix_row_t row_data[0];
+ uint32_t refs;
+ uint32_t columns;
+ uint32_t rows;
+ uintptr_t mask;
+ ec_code_t *code;
+ uint32_t *values;
+ ec_matrix_row_t row_data[0];
};
struct _ec_matrix_list {
- struct list_head lru;
- gf_lock_t lock;
- uint32_t columns;
- uint32_t rows;
- uint32_t max;
- uint32_t count;
- uint32_t stripe;
- struct mem_pool *pool;
- ec_gf_t *gf;
- ec_code_t *code;
- ec_matrix_t *encode;
- ec_matrix_t **objects;
+ struct list_head lru;
+ gf_lock_t lock;
+ uint32_t columns;
+ uint32_t rows;
+ uint32_t max;
+ uint32_t count;
+ uint32_t stripe;
+ struct mem_pool *pool;
+ ec_gf_t *gf;
+ ec_code_t *code;
+ ec_matrix_t *encode;
+ ec_matrix_t **objects;
};
struct _ec_heal {
- struct list_head list;
- gf_lock_t lock;
- xlator_t *xl;
- ec_fop_data_t *fop;
- void *data;
- ec_fop_data_t *lookup;
- loc_t loc;
- struct iatt iatt;
- char *symlink;
- fd_t *fd;
- int32_t partial;
- int32_t done;
- int32_t error;
- gf_boolean_t nameheal;
- uintptr_t available;
- uintptr_t good;
- uintptr_t bad;
- uintptr_t open;
- uintptr_t fixed;
- uint64_t offset;
- uint64_t size;
- uint64_t total_size;
- uint64_t version[2];
- uint64_t raw_size;
+ struct list_head list;
+ gf_lock_t lock;
+ xlator_t *xl;
+ ec_fop_data_t *fop;
+ void *data;
+ ec_fop_data_t *lookup;
+ loc_t loc;
+ struct iatt iatt;
+ char *symlink;
+ fd_t *fd;
+ int32_t partial;
+ int32_t done;
+ int32_t error;
+ gf_boolean_t nameheal;
+ uintptr_t available;
+ uintptr_t good;
+ uintptr_t bad;
+ uintptr_t open;
+ uintptr_t fixed;
+ uint64_t offset;
+ uint64_t size;
+ uint64_t total_size;
+ uint64_t version[2];
+ uint64_t raw_size;
};
struct subvol_healer {
- xlator_t *this;
- int subvol;
- gf_boolean_t local;
- gf_boolean_t running;
- gf_boolean_t rerun;
- pthread_mutex_t mutex;
- pthread_cond_t cond;
- pthread_t thread;
+ xlator_t *this;
+ int subvol;
+ gf_boolean_t running;
+ gf_boolean_t rerun;
+ pthread_mutex_t mutex;
+ pthread_cond_t cond;
+ pthread_t thread;
};
struct _ec_self_heald {
- gf_boolean_t iamshd;
- gf_boolean_t enabled;
- int timeout;
- uint32_t max_threads;
- uint32_t wait_qlength;
- struct subvol_healer *index_healers;
- struct subvol_healer *full_healers;
+ gf_boolean_t iamshd;
+ gf_boolean_t enabled;
+ int timeout;
+ uint32_t max_threads;
+ uint32_t wait_qlength;
+ struct subvol_healer *index_healers;
+ struct subvol_healer *full_healers;
};
struct _ec_statistics {
- struct {
- gf_atomic_t hits; /* Cache hits. */
- gf_atomic_t misses; /* Cache misses. */
- gf_atomic_t updates; /* Number of times an existing stripe has
- been updated with new content. */
- gf_atomic_t invals; /* Number of times an existing stripe has
- been invalidated because of truncates
- or discards. */
- gf_atomic_t evicts; /* Number of times that an existing entry
- has been evicted to make room for newer
- entries. */
- gf_atomic_t allocs; /* Number of memory allocations made to
- store stripes. */
- gf_atomic_t errors; /* Number of errors that have caused extra
- requests. (Basically memory allocation
- errors). */
- } stripe_cache;
+ struct {
+ gf_atomic_t hits; /* Cache hits. */
+ gf_atomic_t misses; /* Cache misses. */
+ gf_atomic_t updates; /* Number of times an existing stripe has
+ been updated with new content. */
+ gf_atomic_t invals; /* Number of times an existing stripe has
+ been invalidated because of truncates
+ or discards. */
+ gf_atomic_t evicts; /* Number of times that an existing entry
+ has been evicted to make room for newer
+ entries. */
+ gf_atomic_t allocs; /* Number of memory allocations made to
+ store stripes. */
+ gf_atomic_t errors; /* Number of errors that have caused extra
+ requests. (Basically memory allocation
+ errors). */
+ } stripe_cache;
+ struct {
+ gf_atomic_t attempted; /*Number of heals attempted on
+ files/directories*/
+ gf_atomic_t completed; /*Number of heals complted on files/directories*/
+ } shd;
};
struct _ec {
- xlator_t *xl;
- int32_t healers;
- int32_t heal_waiters;
- int32_t nodes; /* Total number of bricks(n) */
- int32_t bits_for_nodes;
- int32_t fragments; /* Data bricks(k) */
- int32_t redundancy; /* Redundant bricks(m) */
- uint32_t fragment_size; /* Size of fragment/chunk on a
- brick. */
- uint32_t stripe_size; /* (fragment_size * fragments)
- maximum size of user data
- stored in one stripe. */
- int32_t up; /* Represents whether EC volume is
- up or not. */
- uint32_t idx;
- uint32_t xl_up_count; /* Number of UP bricks. */
- uintptr_t xl_up; /* Bit flag representing UP
- bricks */
- uint32_t xl_notify_count; /* Number of notifications. */
- uintptr_t xl_notify; /* Bit flag representing
- notification for bricks. */
- uintptr_t node_mask;
- xlator_t **xl_list;
- gf_lock_t lock;
- gf_timer_t *timer;
- gf_boolean_t shutdown;
- gf_boolean_t eager_lock;
- gf_boolean_t other_eager_lock;
- gf_boolean_t optimistic_changelog;
- gf_boolean_t parallel_writes;
- uint32_t stripe_cache;
- uint32_t background_heals;
- uint32_t heal_wait_qlen;
- uint32_t self_heal_window_size; /* max size of read/writes */
- uint32_t eager_lock_timeout;
- uint32_t other_eager_lock_timeout;
- struct list_head pending_fops;
- struct list_head heal_waiting;
- struct list_head healing;
- struct mem_pool *fop_pool;
- struct mem_pool *cbk_pool;
- struct mem_pool *lock_pool;
- ec_self_heald_t shd;
- char vol_uuid[UUID_SIZE + 1];
- dict_t *leaf_to_subvolid;
- ec_read_policy_t read_policy;
- ec_matrix_list_t matrix;
- ec_statistics_t stats;
+ xlator_t *xl;
+ int32_t healers;
+ int32_t heal_waiters;
+ int32_t nodes; /* Total number of bricks(n) */
+ int32_t bits_for_nodes;
+ int32_t fragments; /* Data bricks(k) */
+ int32_t redundancy; /* Redundant bricks(m) */
+ uint32_t fragment_size; /* Size of fragment/chunk on a
+ brick. */
+ uint32_t stripe_size; /* (fragment_size * fragments)
+ maximum size of user data
+ stored in one stripe. */
+ int32_t up; /* Represents whether EC volume is
+ up or not. */
+ uint32_t idx;
+ uint32_t xl_up_count; /* Number of UP bricks. */
+ uintptr_t xl_up; /* Bit flag representing UP
+ bricks */
+ uint32_t xl_notify_count; /* Number of notifications. */
+ uintptr_t xl_notify; /* Bit flag representing
+ notification for bricks. */
+ uintptr_t node_mask;
+ uintptr_t read_mask; /*Stores user defined read-mask*/
+ gf_atomic_t async_fop_count; /* Number of on going asynchronous fops. */
+ xlator_t **xl_list;
+ gf_lock_t lock;
+ gf_timer_t *timer;
+ gf_boolean_t shutdown;
+ gf_boolean_t eager_lock;
+ gf_boolean_t other_eager_lock;
+ gf_boolean_t optimistic_changelog;
+ gf_boolean_t parallel_writes;
+ uint32_t stripe_cache;
+ uint32_t quorum_count;
+ uint32_t background_heals;
+ uint32_t heal_wait_qlen;
+ uint32_t self_heal_window_size; /* max size of read/writes */
+ uint32_t eager_lock_timeout;
+ uint32_t other_eager_lock_timeout;
+ struct list_head pending_fops;
+ struct list_head heal_waiting;
+ struct list_head healing;
+ struct mem_pool *fop_pool;
+ struct mem_pool *cbk_pool;
+ struct mem_pool *lock_pool;
+ ec_self_heald_t shd;
+ char vol_uuid[UUID_SIZE + 1];
+ dict_t *leaf_to_subvolid;
+ ec_read_policy_t read_policy;
+ ec_matrix_list_t matrix;
+ ec_statistics_t stats;
};
#endif /* __EC_TYPES_H__ */
diff --git a/xlators/cluster/ec/src/ec.c b/xlators/cluster/ec/src/ec.c
index 31e9c0ec584..7344be4968d 100644
--- a/xlators/cluster/ec/src/ec.c
+++ b/xlators/cluster/ec/src/ec.c
@@ -8,10 +8,10 @@
cases as published by the Free Software Foundation.
*/
-#include "defaults.h"
-#include "statedump.h"
-#include "compat-errno.h"
-#include "upcall-utils.h"
+#include <glusterfs/defaults.h>
+#include <glusterfs/statedump.h>
+#include <glusterfs/compat-errno.h>
+#include <glusterfs/upcall-utils.h>
#include "ec.h"
#include "ec-messages.h"
@@ -23,50 +23,48 @@
#include "ec-method.h"
#include "ec-code.h"
#include "ec-heald.h"
-#include "events.h"
+#include <glusterfs/events.h>
static char *ec_read_policies[EC_READ_POLICY_MAX + 1] = {
- [EC_ROUND_ROBIN] = "round-robin",
- [EC_GFID_HASH] = "gfid-hash",
- [EC_READ_POLICY_MAX] = NULL
-};
+ [EC_ROUND_ROBIN] = "round-robin",
+ [EC_GFID_HASH] = "gfid-hash",
+ [EC_READ_POLICY_MAX] = NULL};
#define EC_INTERNAL_XATTR_OR_GOTO(name, xattr, op_errno, label) \
- do { \
- if (ec_is_internal_xattr (NULL, (char *)name, NULL, NULL)) { \
- op_errno = EPERM; \
- goto label; \
- } \
- if (name && (strlen (name) == 0) && xattr) { \
- /* Bulk [f]removexattr/[f]setxattr */ \
- GF_IF_INTERNAL_XATTR_GOTO (EC_XATTR_PREFIX"*", xattr, \
- op_errno, label); \
- } \
- } while (0)
-
-int32_t ec_parse_options(xlator_t * this)
-{
- ec_t * ec = this->private;
+ do { \
+ if (ec_is_internal_xattr(NULL, (char *)name, NULL, NULL)) { \
+ op_errno = EPERM; \
+ goto label; \
+ } \
+ if (name && (strlen(name) == 0) && xattr) { \
+ /* Bulk [f]removexattr/[f]setxattr */ \
+ GF_IF_INTERNAL_XATTR_GOTO(EC_XATTR_PREFIX "*", xattr, op_errno, \
+ label); \
+ } \
+ } while (0)
+
+int32_t
+ec_parse_options(xlator_t *this)
+{
+ ec_t *ec = this->private;
int32_t error = EINVAL;
uintptr_t mask;
GF_OPTION_INIT("redundancy", ec->redundancy, int32, out);
ec->fragments = ec->nodes - ec->redundancy;
if ((ec->redundancy < 1) || (ec->redundancy >= ec->fragments) ||
- (ec->fragments > EC_MAX_FRAGMENTS))
- {
- gf_msg (this->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_INVALID_REDUNDANCY,
- "Invalid redundancy (must be between "
- "1 and %d)", (ec->nodes - 1) / 2);
+ (ec->fragments > EC_MAX_FRAGMENTS)) {
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL, EC_MSG_INVALID_REDUNDANCY,
+ "Invalid redundancy (must be between "
+ "1 and %d)",
+ (ec->nodes - 1) / 2);
goto out;
}
ec->bits_for_nodes = 1;
mask = 2;
- while (ec->nodes > mask)
- {
+ while (ec->nodes > mask) {
ec->bits_for_nodes++;
mask <<= 1;
}
@@ -74,9 +72,10 @@ int32_t ec_parse_options(xlator_t * this)
ec->fragment_size = EC_METHOD_CHUNK_SIZE;
ec->stripe_size = ec->fragment_size * ec->fragments;
- gf_msg_debug ("ec", 0, "Initialized with: nodes=%u, fragments=%u, "
- "stripe_size=%u, node_mask=%lX",
- ec->nodes, ec->fragments, ec->stripe_size, ec->node_mask);
+ gf_msg_debug("ec", 0,
+ "Initialized with: nodes=%u, fragments=%u, "
+ "stripe_size=%u, node_mask=%" PRIxFAST32,
+ ec->nodes, ec->fragments, ec->stripe_size, ec->node_mask);
error = 0;
@@ -84,30 +83,28 @@ out:
return error;
}
-int32_t ec_prepare_childs(xlator_t * this)
+int32_t
+ec_prepare_childs(xlator_t *this)
{
- ec_t * ec = this->private;
- xlator_list_t * child = NULL;
+ ec_t *ec = this->private;
+ xlator_list_t *child = NULL;
int32_t count = 0;
- for (child = this->children; child != NULL; child = child->next)
- {
+ for (child = this->children; child != NULL; child = child->next) {
count++;
}
- if (count > EC_MAX_NODES)
- {
- gf_msg (this->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_TOO_MANY_SUBVOLS, "Too many subvolumes");
+ if (count > EC_MAX_NODES) {
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL, EC_MSG_TOO_MANY_SUBVOLS,
+ "Too many subvolumes");
return EINVAL;
}
ec->nodes = count;
ec->xl_list = GF_CALLOC(count, sizeof(ec->xl_list[0]), ec_mt_xlator_t);
- if (ec->xl_list == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY, "Allocation of xlator list failed");
+ if (ec->xl_list == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Allocation of xlator list failed");
return ENOMEM;
}
@@ -115,8 +112,7 @@ int32_t ec_prepare_childs(xlator_t * this)
ec->xl_up_count = 0;
count = 0;
- for (child = this->children; child != NULL; child = child->next)
- {
+ for (child = this->children; child != NULL; child = child->next) {
ec->xl_list[count++] = child->xlator;
}
@@ -125,43 +121,42 @@ int32_t ec_prepare_childs(xlator_t * this)
/* This function transforms the subvol to subvol-id*/
static int
-_subvol_to_subvolid (dict_t *this, char *key, data_t *value, void *data)
-{
- ec_t *ec = data;
- xlator_t *subvol = NULL;
- int i = 0;
- int ret = -1;
-
- subvol = data_to_ptr (value);
- for (i = 0; i < ec->nodes; i++) {
- if (ec->xl_list[i] == subvol) {
- ret = dict_set_int32 (this, key, i);
- /* -1 stops dict_foreach and returns -1*/
- if (ret < 0)
- ret = -1;
- goto out;
- }
+_subvol_to_subvolid(dict_t *this, char *key, data_t *value, void *data)
+{
+ ec_t *ec = data;
+ xlator_t *subvol = NULL;
+ int i = 0;
+ int ret = -1;
+
+ subvol = data_to_ptr(value);
+ for (i = 0; i < ec->nodes; i++) {
+ if (ec->xl_list[i] == subvol) {
+ ret = dict_set_int32(this, key, i);
+ /* -1 stops dict_foreach and returns -1*/
+ if (ret < 0)
+ ret = -1;
+ goto out;
}
+ }
out:
- return ret;
+ return ret;
}
int
-ec_subvol_to_subvol_id_transform (ec_t *ec, dict_t *leaf_to_subvolid)
+ec_subvol_to_subvol_id_transform(ec_t *ec, dict_t *leaf_to_subvolid)
{
- return dict_foreach (leaf_to_subvolid, _subvol_to_subvolid, ec);
+ return dict_foreach(leaf_to_subvolid, _subvol_to_subvolid, ec);
}
-void __ec_destroy_private(xlator_t * this)
+void
+__ec_destroy_private(xlator_t *this)
{
- ec_t * ec = this->private;
+ ec_t *ec = this->private;
- if (ec != NULL)
- {
+ if (ec != NULL) {
LOCK(&ec->lock);
- if (ec->timer != NULL)
- {
+ if (ec->timer != NULL) {
gf_timer_call_cancel(this->ctx, ec->timer);
ec->timer = NULL;
}
@@ -179,31 +174,27 @@ void __ec_destroy_private(xlator_t * this)
sleep(2);
this->private = NULL;
- if (ec->xl_list != NULL)
- {
+ if (ec->xl_list != NULL) {
GF_FREE(ec->xl_list);
ec->xl_list = NULL;
}
- if (ec->fop_pool != NULL)
- {
+ if (ec->fop_pool != NULL) {
mem_pool_destroy(ec->fop_pool);
}
- if (ec->cbk_pool != NULL)
- {
+ if (ec->cbk_pool != NULL) {
mem_pool_destroy(ec->cbk_pool);
}
- if (ec->lock_pool != NULL)
- {
+ if (ec->lock_pool != NULL) {
mem_pool_destroy(ec->lock_pool);
}
LOCK_DESTROY(&ec->lock);
if (ec->leaf_to_subvolid)
- dict_unref (ec->leaf_to_subvolid);
+ dict_unref(ec->leaf_to_subvolid);
ec_method_fini(&ec->matrix);
@@ -211,13 +202,13 @@ void __ec_destroy_private(xlator_t * this)
}
}
-int32_t mem_acct_init(xlator_t * this)
+int32_t
+mem_acct_init(xlator_t *this)
{
- if (xlator_mem_acct_init(this, ec_mt_end + 1) != 0)
- {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY, "Memory accounting initialization "
- "failed.");
+ if (xlator_mem_acct_init(this, ec_mt_end + 1) != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Memory accounting initialization "
+ "failed.");
return -1;
}
@@ -226,442 +217,582 @@ int32_t mem_acct_init(xlator_t * this)
}
void
-ec_configure_background_heal_opts (ec_t *ec, int background_heals,
- int heal_wait_qlen)
+ec_configure_background_heal_opts(ec_t *ec, int background_heals,
+ int heal_wait_qlen)
{
- if (background_heals == 0) {
- ec->heal_wait_qlen = 0;
- } else {
- ec->heal_wait_qlen = heal_wait_qlen;
- }
- ec->background_heals = background_heals;
+ if (background_heals == 0) {
+ ec->heal_wait_qlen = 0;
+ } else {
+ ec->heal_wait_qlen = heal_wait_qlen;
+ }
+ ec->background_heals = background_heals;
}
int
-ec_assign_read_policy (ec_t *ec, char *read_policy)
+ec_assign_read_policy(ec_t *ec, char *read_policy)
{
- int read_policy_idx = -1;
+ int read_policy_idx = -1;
- read_policy_idx = gf_get_index_by_elem (ec_read_policies, read_policy);
- if (read_policy_idx < 0 || read_policy_idx >= EC_READ_POLICY_MAX)
- return -1;
+ read_policy_idx = gf_get_index_by_elem(ec_read_policies, read_policy);
+ if (read_policy_idx < 0 || read_policy_idx >= EC_READ_POLICY_MAX)
+ return -1;
- ec->read_policy = read_policy_idx;
- return 0;
+ ec->read_policy = read_policy_idx;
+ return 0;
}
int32_t
-reconfigure (xlator_t *this, dict_t *options)
-{
- ec_t *ec = this->private;
- char *read_policy = NULL;
- char *extensions = NULL;
- uint32_t heal_wait_qlen = 0;
- uint32_t background_heals = 0;
- int32_t ret = -1;
- int32_t err;
-
- GF_OPTION_RECONF ("cpu-extensions", extensions, options, str, failed);
-
- GF_OPTION_RECONF ("self-heal-daemon", ec->shd.enabled, options, bool,
- failed);
- GF_OPTION_RECONF ("iam-self-heal-daemon", ec->shd.iamshd, options,
- bool, failed);
- GF_OPTION_RECONF ("eager-lock", ec->eager_lock, options,
- bool, failed);
- GF_OPTION_RECONF ("other-eager-lock", ec->other_eager_lock, options,
- bool, failed);
- GF_OPTION_RECONF ("eager-lock-timeout", ec->eager_lock_timeout,
- options, uint32, failed);
- GF_OPTION_RECONF ("other-eager-lock-timeout",
- ec->other_eager_lock_timeout, options, uint32,
- failed);
- GF_OPTION_RECONF ("background-heals", background_heals, options,
- uint32, failed);
- GF_OPTION_RECONF ("heal-wait-qlength", heal_wait_qlen, options,
- uint32, failed);
- GF_OPTION_RECONF ("self-heal-window-size", ec->self_heal_window_size,
- options, uint32, failed);
- GF_OPTION_RECONF ("heal-timeout", ec->shd.timeout, options,
- int32, failed);
- ec_configure_background_heal_opts (ec, background_heals,
- heal_wait_qlen);
- GF_OPTION_RECONF ("shd-max-threads", ec->shd.max_threads,
- options, uint32, failed);
- GF_OPTION_RECONF ("shd-wait-qlength", ec->shd.wait_qlength,
- options, uint32, failed);
-
- GF_OPTION_RECONF ("read-policy", read_policy, options, str, failed);
-
- GF_OPTION_RECONF ("optimistic-change-log", ec->optimistic_changelog,
- options, bool, failed);
- GF_OPTION_RECONF ("parallel-writes", ec->parallel_writes,
- options, bool, failed);
- GF_OPTION_RECONF ("stripe-cache", ec->stripe_cache, options, uint32,
- failed);
- ret = 0;
- if (ec_assign_read_policy (ec, read_policy)) {
- ret = -1;
- }
+reconfigure(xlator_t *this, dict_t *options)
+{
+ ec_t *ec = this->private;
+ char *read_policy = NULL;
+ char *extensions = NULL;
+ uint32_t heal_wait_qlen = 0;
+ uint32_t background_heals = 0;
+ int32_t ret = -1;
+ int32_t err;
- err = ec_method_update(this, &ec->matrix, extensions);
- if (err != 0) {
- ret = -1;
- }
+ GF_OPTION_RECONF("cpu-extensions", extensions, options, str, failed);
+
+ GF_OPTION_RECONF("self-heal-daemon", ec->shd.enabled, options, bool,
+ failed);
+ GF_OPTION_RECONF("iam-self-heal-daemon", ec->shd.iamshd, options, bool,
+ failed);
+ GF_OPTION_RECONF("eager-lock", ec->eager_lock, options, bool, failed);
+ GF_OPTION_RECONF("other-eager-lock", ec->other_eager_lock, options, bool,
+ failed);
+ GF_OPTION_RECONF("eager-lock-timeout", ec->eager_lock_timeout, options,
+ uint32, failed);
+ GF_OPTION_RECONF("other-eager-lock-timeout", ec->other_eager_lock_timeout,
+ options, uint32, failed);
+ GF_OPTION_RECONF("background-heals", background_heals, options, uint32,
+ failed);
+ GF_OPTION_RECONF("heal-wait-qlength", heal_wait_qlen, options, uint32,
+ failed);
+ GF_OPTION_RECONF("self-heal-window-size", ec->self_heal_window_size,
+ options, uint32, failed);
+ GF_OPTION_RECONF("heal-timeout", ec->shd.timeout, options, int32, failed);
+ ec_configure_background_heal_opts(ec, background_heals, heal_wait_qlen);
+ GF_OPTION_RECONF("shd-max-threads", ec->shd.max_threads, options, uint32,
+ failed);
+ GF_OPTION_RECONF("shd-wait-qlength", ec->shd.wait_qlength, options, uint32,
+ failed);
+
+ GF_OPTION_RECONF("read-policy", read_policy, options, str, failed);
+
+ GF_OPTION_RECONF("optimistic-change-log", ec->optimistic_changelog, options,
+ bool, failed);
+ GF_OPTION_RECONF("parallel-writes", ec->parallel_writes, options, bool,
+ failed);
+ GF_OPTION_RECONF("stripe-cache", ec->stripe_cache, options, uint32, failed);
+ GF_OPTION_RECONF("quorum-count", ec->quorum_count, options, uint32, failed);
+ ret = 0;
+ if (ec_assign_read_policy(ec, read_policy)) {
+ ret = -1;
+ }
+
+ err = ec_method_update(this, &ec->matrix, extensions);
+ if (err != 0) {
+ ret = -1;
+ }
failed:
- return ret;
+ return ret;
}
glusterfs_event_t
-ec_get_event_from_state (ec_t *ec)
-{
- int down_count = 0;
-
- if (ec->xl_up_count >= ec->fragments) {
- /* If ec is up but some subvolumes are yet to notify, give
- * grace time for other subvols to notify to prevent start of
- * I/O which may result in self-heals */
- if (ec->xl_notify_count < ec->nodes)
- return GF_EVENT_MAXVAL;
-
- return GF_EVENT_CHILD_UP;
- } else {
- down_count = ec->xl_notify_count - ec->xl_up_count;
- if (down_count > ec->redundancy)
- return GF_EVENT_CHILD_DOWN;
- }
+ec_get_event_from_state(ec_t *ec)
+{
+ int down_count = 0;
+
+ if (ec->xl_up_count >= ec->fragments) {
+ /* If ec is up but some subvolumes are yet to notify, give
+ * grace time for other subvols to notify to prevent start of
+ * I/O which may result in self-heals */
+ if (ec->xl_notify_count < ec->nodes)
+ return GF_EVENT_MAXVAL;
+
+ return GF_EVENT_CHILD_UP;
+ } else {
+ down_count = ec->xl_notify_count - ec->xl_up_count;
+ if (down_count > ec->redundancy)
+ return GF_EVENT_CHILD_DOWN;
+ }
- return GF_EVENT_MAXVAL;
+ return GF_EVENT_MAXVAL;
}
void
-ec_up (xlator_t *this, ec_t *ec)
+ec_up(xlator_t *this, ec_t *ec)
{
- if (ec->timer != NULL) {
- gf_timer_call_cancel (this->ctx, ec->timer);
- ec->timer = NULL;
- }
+ char str1[32], str2[32];
- ec->up = 1;
- gf_msg (this->name, GF_LOG_INFO, 0, EC_MSG_EC_UP, "Going UP");
+ if (ec->timer != NULL) {
+ gf_timer_call_cancel(this->ctx, ec->timer);
+ ec->timer = NULL;
+ }
+
+ ec->up = 1;
+ gf_msg(this->name, GF_LOG_INFO, 0, EC_MSG_EC_UP,
+ "Going UP : Child UP = %s Child Notify = %s",
+ ec_bin(str1, sizeof(str1), ec->xl_up, ec->nodes),
+ ec_bin(str2, sizeof(str2), ec->xl_notify, ec->nodes));
- gf_event (EVENT_EC_MIN_BRICKS_UP, "subvol=%s", this->name);
+ gf_event(EVENT_EC_MIN_BRICKS_UP, "subvol=%s", this->name);
}
void
-ec_down (xlator_t *this, ec_t *ec)
+ec_down(xlator_t *this, ec_t *ec)
{
- if (ec->timer != NULL) {
- gf_timer_call_cancel(this->ctx, ec->timer);
- ec->timer = NULL;
- }
+ char str1[32], str2[32];
- ec->up = 0;
- gf_msg (this->name, GF_LOG_INFO, 0, EC_MSG_EC_DOWN, "Going DOWN");
+ if (ec->timer != NULL) {
+ gf_timer_call_cancel(this->ctx, ec->timer);
+ ec->timer = NULL;
+ }
+
+ ec->up = 0;
+ gf_msg(this->name, GF_LOG_INFO, 0, EC_MSG_EC_DOWN,
+ "Going DOWN : Child UP = %s Child Notify = %s",
+ ec_bin(str1, sizeof(str1), ec->xl_up, ec->nodes),
+ ec_bin(str2, sizeof(str2), ec->xl_notify, ec->nodes));
- gf_event (EVENT_EC_MIN_BRICKS_NOT_UP, "subvol=%s", this->name);
+ gf_event(EVENT_EC_MIN_BRICKS_NOT_UP, "subvol=%s", this->name);
}
void
-ec_notify_cbk (void *data)
+ec_notify_cbk(void *data)
{
- ec_t *ec = data;
- glusterfs_event_t event = GF_EVENT_MAXVAL;
- gf_boolean_t propagate = _gf_false;
+ ec_t *ec = data;
+ glusterfs_event_t event = GF_EVENT_MAXVAL;
+ gf_boolean_t propagate = _gf_false;
+ gf_boolean_t launch_heal = _gf_false;
- LOCK(&ec->lock);
- {
- if (!ec->timer) {
- /*
- * Either child_up/child_down is already sent to parent
- * This is a spurious wake up.
- */
- goto unlock;
- }
-
- gf_timer_call_cancel (ec->xl->ctx, ec->timer);
- ec->timer = NULL;
-
- /* The timeout has expired, so any subvolume that has not
- * already reported its state, will be considered to be down.
- * We mark as if all bricks had reported. */
- ec->xl_notify = (1ULL << ec->nodes) - 1ULL;
- ec->xl_notify_count = ec->nodes;
-
- /* Since we have marked all subvolumes as notified, it's
- * guaranteed that ec_get_event_from_state() will return
- * CHILD_UP or CHILD_DOWN, but not MAXVAL. */
- event = ec_get_event_from_state (ec);
- if (event == GF_EVENT_CHILD_UP) {
- /* We are ready to bring the volume up. If there are
- * still bricks DOWN, they will be healed when they
- * come up. */
- ec_up (ec->xl, ec);
- }
-
- propagate = _gf_true;
+ LOCK(&ec->lock);
+ {
+ if (!ec->timer) {
+ /*
+ * Either child_up/child_down is already sent to parent
+ * This is a spurious wake up.
+ */
+ goto unlock;
}
-unlock:
- UNLOCK(&ec->lock);
- if (propagate) {
- if ((event == GF_EVENT_CHILD_UP) && ec->shd.iamshd) {
- /* We have just brought the volume UP, so we trigger
- * a self-heal check on the root directory. */
- ec_launch_replace_heal (ec);
- }
-
- default_notify (ec->xl, event, NULL);
+ gf_timer_call_cancel(ec->xl->ctx, ec->timer);
+ ec->timer = NULL;
+
+ /* The timeout has expired, so any subvolume that has not
+ * already reported its state, will be considered to be down.
+ * We mark as if all bricks had reported. */
+ ec->xl_notify = (1ULL << ec->nodes) - 1ULL;
+ ec->xl_notify_count = ec->nodes;
+
+ /* Since we have marked all subvolumes as notified, it's
+ * guaranteed that ec_get_event_from_state() will return
+ * CHILD_UP or CHILD_DOWN, but not MAXVAL. */
+ event = ec_get_event_from_state(ec);
+ if (event == GF_EVENT_CHILD_UP) {
+ /* We are ready to bring the volume up. If there are
+ * still bricks DOWN, they will be healed when they
+ * come up. */
+ ec_up(ec->xl, ec);
+
+ if (ec->shd.iamshd && !ec->shutdown) {
+ launch_heal = _gf_true;
+ GF_ATOMIC_INC(ec->async_fop_count);
+ }
}
+ propagate = _gf_true;
+ }
+unlock:
+ UNLOCK(&ec->lock);
+
+ if (launch_heal) {
+ /* We have just brought the volume UP, so we trigger
+ * a self-heal check on the root directory. */
+ ec_launch_replace_heal(ec);
+ }
+ if (propagate) {
+ default_notify(ec->xl, event, NULL);
+ }
}
void
-ec_launch_notify_timer (xlator_t *this, ec_t *ec)
-{
- struct timespec delay = {0, };
-
- gf_msg_debug (this->name, 0, "Initiating child-down timer");
- delay.tv_sec = 10;
- delay.tv_nsec = 0;
- ec->timer = gf_timer_call_after (this->ctx, delay, ec_notify_cbk, ec);
- if (ec->timer == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_TIMER_CREATE_FAIL, "Cannot create timer "
- "for delayed initialization");
- }
+ec_launch_notify_timer(xlator_t *this, ec_t *ec)
+{
+ struct timespec delay = {
+ 0,
+ };
+
+ gf_msg_debug(this->name, 0, "Initiating child-down timer");
+ delay.tv_sec = 10;
+ delay.tv_nsec = 0;
+ ec->timer = gf_timer_call_after(this->ctx, delay, ec_notify_cbk, ec);
+ if (ec->timer == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_TIMER_CREATE_FAIL,
+ "Cannot create timer "
+ "for delayed initialization");
+ }
}
gf_boolean_t
ec_disable_delays(ec_t *ec)
{
- ec->shutdown = _gf_true;
+ ec->shutdown = _gf_true;
- return list_empty (&ec->pending_fops);
+ return __ec_is_last_fop(ec);
}
void
-ec_pending_fops_completed(ec_t *ec)
+ec_cleanup_healer_object(ec_t *ec)
{
- if (ec->shutdown) {
- default_notify (ec->xl, GF_EVENT_PARENT_DOWN, NULL);
+ struct subvol_healer *healer = NULL;
+ ec_self_heald_t *shd = NULL;
+ void *res = NULL;
+ int i = 0;
+ gf_boolean_t is_join = _gf_false;
+
+ shd = &ec->shd;
+ if (!shd->iamshd)
+ return;
+
+ for (i = 0; i < ec->nodes; i++) {
+ healer = &shd->index_healers[i];
+ pthread_mutex_lock(&healer->mutex);
+ {
+ healer->rerun = 1;
+ if (healer->running) {
+ pthread_cond_signal(&healer->cond);
+ is_join = _gf_true;
+ }
+ }
+ pthread_mutex_unlock(&healer->mutex);
+ if (is_join) {
+ pthread_join(healer->thread, &res);
+ is_join = _gf_false;
}
+
+ healer = &shd->full_healers[i];
+ pthread_mutex_lock(&healer->mutex);
+ {
+ healer->rerun = 1;
+ if (healer->running) {
+ pthread_cond_signal(&healer->cond);
+ is_join = _gf_true;
+ }
+ }
+ pthread_mutex_unlock(&healer->mutex);
+ if (is_join) {
+ pthread_join(healer->thread, &res);
+ is_join = _gf_false;
+ }
+ }
+}
+void
+ec_pending_fops_completed(ec_t *ec)
+{
+ if (ec->shutdown) {
+ default_notify(ec->xl, GF_EVENT_PARENT_DOWN, NULL);
+ }
}
static gf_boolean_t
ec_set_up_state(ec_t *ec, uintptr_t index_mask, uintptr_t new_state)
{
- uintptr_t current_state = 0;
+ uintptr_t current_state = 0;
- if ((ec->xl_notify & index_mask) == 0) {
- ec->xl_notify |= index_mask;
- ec->xl_notify_count++;
- }
- current_state = ec->xl_up & index_mask;
- if (current_state != new_state) {
- ec->xl_up ^= index_mask;
- ec->xl_up_count += (current_state ? -1 : 1);
+ if (xlator_is_cleanup_starting(ec->xl))
+ return _gf_false;
- return _gf_true;
- }
+ if ((ec->xl_notify & index_mask) == 0) {
+ ec->xl_notify |= index_mask;
+ ec->xl_notify_count++;
+ }
+ current_state = ec->xl_up & index_mask;
+ if (current_state != new_state) {
+ ec->xl_up ^= index_mask;
+ ec->xl_up_count += (current_state ? -1 : 1);
- return _gf_false;
+ return _gf_true;
+ }
+
+ return _gf_false;
}
static gf_boolean_t
ec_upcall(ec_t *ec, struct gf_upcall *upcall)
{
- struct gf_upcall_cache_invalidation *ci = NULL;
- struct gf_upcall_inodelk_contention *lc = NULL;
- inode_t *inode;
+ struct gf_upcall_cache_invalidation *ci = NULL;
+ struct gf_upcall_inodelk_contention *lc = NULL;
+ inode_t *inode;
+ inode_table_t *table;
- switch (upcall->event_type) {
+ switch (upcall->event_type) {
case GF_UPCALL_CACHE_INVALIDATION:
- ci = upcall->data;
- ci->flags |= UP_INVAL_ATTR;
- return _gf_true;
+ ci = upcall->data;
+ ci->flags |= UP_INVAL_ATTR;
+ return _gf_true;
case GF_UPCALL_INODELK_CONTENTION:
- lc = upcall->data;
- if (strcmp(lc->domain, ec->xl->name) != 0) {
- /* The lock is not owned by EC, ignore it. */
- return _gf_true;
- }
- inode = inode_find(((xlator_t *)ec->xl->graph->top)->itable,
- upcall->gfid);
- /* If inode is not found, it means that it's already released,
- * so we can ignore it. Probably it has been released and
- * destroyed while the contention notification was being sent.
- */
- if (inode != NULL) {
- ec_lock_release(ec, inode);
- inode_unref(inode);
- }
-
+ lc = upcall->data;
+ if (strcmp(lc->domain, ec->xl->name) != 0) {
+ /* The lock is not owned by EC, ignore it. */
+ return _gf_true;
+ }
+ table = ((xlator_t *)ec->xl->graph->top)->itable;
+ if (table == NULL) {
+ /* Self-heal daemon doesn't have an inode table on the top
+ * xlator because it doesn't need it. In this case we should
+ * use the inode table managed by EC itself where all inodes
+ * being healed should be present. However self-heal doesn't
+ * use eager-locking and inodelk's are already released as
+ * soon as possible. In this case we can safely ignore these
+ * notifications. */
return _gf_false;
+ }
+ inode = inode_find(table, upcall->gfid);
+ /* If inode is not found, it means that it's already released,
+ * so we can ignore it. Probably it has been released and
+ * destroyed while the contention notification was being sent.
+ */
+ if (inode != NULL) {
+ ec_lock_release(ec, inode);
+ inode_unref(inode);
+ }
+
+ return _gf_false;
default:
- return _gf_true;
- }
+ return _gf_true;
+ }
}
int32_t
-ec_notify (xlator_t *this, int32_t event, void *data, void *data2)
-{
- ec_t *ec = this->private;
- int32_t idx = 0;
- int32_t error = 0;
- glusterfs_event_t old_event = GF_EVENT_MAXVAL;
- dict_t *input = NULL;
- dict_t *output = NULL;
- gf_boolean_t propagate = _gf_true;
- gf_boolean_t needs_shd_check = _gf_false;
- int32_t orig_event = event;
- uintptr_t mask = 0;
-
- gf_msg_trace (this->name, 0, "NOTIFY(%d): %p, %p",
- event, data, data2);
-
- if (event == GF_EVENT_UPCALL) {
- propagate = ec_upcall(ec, data);
- goto done;
+ec_notify(xlator_t *this, int32_t event, void *data, void *data2)
+{
+ ec_t *ec = this->private;
+ int32_t idx = 0;
+ int32_t error = 0;
+ glusterfs_event_t old_event = GF_EVENT_MAXVAL;
+ dict_t *input = NULL;
+ dict_t *output = NULL;
+ gf_boolean_t propagate = _gf_true;
+ gf_boolean_t needs_shd_check = _gf_false;
+ int32_t orig_event = event;
+ uintptr_t mask = 0;
+
+ gf_msg_trace(this->name, 0, "NOTIFY(%d): %p, %p", event, data, data2);
+
+ if (event == GF_EVENT_UPCALL) {
+ propagate = ec_upcall(ec, data);
+ goto done;
+ }
+
+ if (event == GF_EVENT_TRANSLATOR_OP) {
+ if (!ec->up) {
+ error = -1;
+ } else {
+ input = data;
+ output = data2;
+ error = ec_xl_op(this, input, output);
+ }
+ goto out;
+ }
+
+ for (idx = 0; idx < ec->nodes; idx++) {
+ if (ec->xl_list[idx] == data) {
+ break;
}
+ }
- if (event == GF_EVENT_TRANSLATOR_OP) {
- if (!ec->up) {
- error = -1;
- } else {
- input = data;
- output = data2;
- error = ec_xl_op (this, input, output);
- }
- goto out;
+ LOCK(&ec->lock);
+
+ if (event == GF_EVENT_PARENT_UP) {
+ /*
+ * Start a timer which sends appropriate event to parent
+ * xlator to prevent the 'mount' syscall from hanging.
+ */
+ ec_launch_notify_timer(this, ec);
+ goto unlock;
+ } else if (event == GF_EVENT_PARENT_DOWN) {
+ /* If there aren't pending fops running after we have waken up
+ * them, we immediately propagate the notification. */
+ propagate = ec_disable_delays(ec);
+ ec_cleanup_healer_object(ec);
+ goto unlock;
+ }
+
+ if (idx < ec->nodes) { /* CHILD_* events */
+ old_event = ec_get_event_from_state(ec);
+
+ mask = 1ULL << idx;
+ if (event == GF_EVENT_CHILD_UP) {
+ /* We need to trigger a selfheal if a brick changes
+ * to UP state. */
+ if (ec_set_up_state(ec, mask, mask) && ec->shd.iamshd &&
+ !ec->shutdown) {
+ needs_shd_check = _gf_true;
+ }
+ } else if (event == GF_EVENT_CHILD_DOWN) {
+ ec_set_up_state(ec, mask, 0);
}
- for (idx = 0; idx < ec->nodes; idx++) {
- if (ec->xl_list[idx] == data) {
- break;
- }
+ event = ec_get_event_from_state(ec);
+
+ if (event == GF_EVENT_CHILD_UP) {
+ if (!ec->up) {
+ ec_up(this, ec);
+ }
+ } else {
+ /* If the volume is not UP, it's irrelevant if one
+ * brick has come up. We cannot heal anything. */
+ needs_shd_check = _gf_false;
+
+ if ((event == GF_EVENT_CHILD_DOWN) && ec->up) {
+ ec_down(this, ec);
+ }
}
- LOCK (&ec->lock);
-
- if (event == GF_EVENT_PARENT_UP) {
- /*
- * Start a timer which sends appropriate event to parent
- * xlator to prevent the 'mount' syscall from hanging.
- */
- ec_launch_notify_timer (this, ec);
- goto unlock;
- } else if (event == GF_EVENT_PARENT_DOWN) {
- /* If there aren't pending fops running after we have waken up
- * them, we immediately propagate the notification. */
- propagate = ec_disable_delays(ec);
- goto unlock;
+ if (event != GF_EVENT_MAXVAL) {
+ if (event == old_event) {
+ if (orig_event == GF_EVENT_CHILD_UP)
+ event = GF_EVENT_SOME_DESCENDENT_UP;
+ else /* orig_event has to be GF_EVENT_CHILD_DOWN */
+ event = GF_EVENT_SOME_DESCENDENT_DOWN;
+ }
+ } else {
+ propagate = _gf_false;
+ needs_shd_check = _gf_false;
}
- if (idx < ec->nodes) { /* CHILD_* events */
- old_event = ec_get_event_from_state (ec);
-
- mask = 1ULL << idx;
- if (event == GF_EVENT_CHILD_UP) {
- /* We need to trigger a selfheal if a brick changes
- * to UP state. */
- needs_shd_check = ec_set_up_state(ec, mask, mask);
- } else if (event == GF_EVENT_CHILD_DOWN) {
- ec_set_up_state(ec, mask, 0);
- }
-
- event = ec_get_event_from_state (ec);
-
- if (event == GF_EVENT_CHILD_UP) {
- if (!ec->up) {
- ec_up (this, ec);
- }
- } else {
- /* If the volume is not UP, it's irrelevant if one
- * brick has come up. We cannot heal anything. */
- needs_shd_check = _gf_false;
-
- if ((event == GF_EVENT_CHILD_DOWN) && ec->up) {
- ec_down (this, ec);
- }
- }
-
- if (event != GF_EVENT_MAXVAL) {
- if (event == old_event) {
- if (orig_event == GF_EVENT_CHILD_UP)
- event = GF_EVENT_SOME_DESCENDENT_UP;
- else /* orig_event has to be GF_EVENT_CHILD_DOWN */
- event = GF_EVENT_SOME_DESCENDENT_DOWN;
- }
- } else {
- propagate = _gf_false;
- }
+ if (needs_shd_check) {
+ GF_ATOMIC_INC(ec->async_fop_count);
}
+ }
unlock:
- UNLOCK (&ec->lock);
+ UNLOCK(&ec->lock);
done:
- if (propagate) {
- if (needs_shd_check && ec->shd.iamshd) {
- ec_launch_replace_heal (ec);
- }
-
- error = default_notify (this, event, data);
- }
+ if (needs_shd_check) {
+ ec_launch_replace_heal(ec);
+ }
+ if (propagate) {
+ error = default_notify(this, event, data);
+ }
out:
- return error;
+ return error;
}
int32_t
-notify (xlator_t *this, int32_t event, void *data, ...)
+notify(xlator_t *this, int32_t event, void *data, ...)
{
- int ret = -1;
- va_list ap;
- void *data2 = NULL;
+ int ret = -1;
+ va_list ap;
+ void *data2 = NULL;
- va_start (ap, data);
- data2 = va_arg (ap, dict_t*);
- va_end (ap);
- ret = ec_notify (this, event, data, data2);
+ va_start(ap, data);
+ data2 = va_arg(ap, dict_t *);
+ va_end(ap);
+ ret = ec_notify(this, event, data, data2);
- return ret;
+ return ret;
}
static void
ec_statistics_init(ec_t *ec)
{
- GF_ATOMIC_INIT(ec->stats.stripe_cache.hits, 0);
- GF_ATOMIC_INIT(ec->stats.stripe_cache.misses, 0);
- GF_ATOMIC_INIT(ec->stats.stripe_cache.updates, 0);
- GF_ATOMIC_INIT(ec->stats.stripe_cache.invals, 0);
- GF_ATOMIC_INIT(ec->stats.stripe_cache.evicts, 0);
- GF_ATOMIC_INIT(ec->stats.stripe_cache.allocs, 0);
- GF_ATOMIC_INIT(ec->stats.stripe_cache.errors, 0);
+ GF_ATOMIC_INIT(ec->stats.stripe_cache.hits, 0);
+ GF_ATOMIC_INIT(ec->stats.stripe_cache.misses, 0);
+ GF_ATOMIC_INIT(ec->stats.stripe_cache.updates, 0);
+ GF_ATOMIC_INIT(ec->stats.stripe_cache.invals, 0);
+ GF_ATOMIC_INIT(ec->stats.stripe_cache.evicts, 0);
+ GF_ATOMIC_INIT(ec->stats.stripe_cache.allocs, 0);
+ GF_ATOMIC_INIT(ec->stats.stripe_cache.errors, 0);
+ GF_ATOMIC_INIT(ec->stats.shd.attempted, 0);
+ GF_ATOMIC_INIT(ec->stats.shd.completed, 0);
+}
+
+static int
+ec_assign_read_mask(ec_t *ec, char *read_mask_str)
+{
+ char *mask = NULL;
+ char *maskptr = NULL;
+ char *saveptr = NULL;
+ char *id_str = NULL;
+ int id = 0;
+ int ret = 0;
+ uintptr_t read_mask = 0;
+
+ if (!read_mask_str) {
+ ec->read_mask = 0;
+ ret = 0;
+ goto out;
+ }
+
+ mask = gf_strdup(read_mask_str);
+ if (!mask) {
+ ret = -1;
+ goto out;
+ }
+ maskptr = mask;
+
+ for (;;) {
+ id_str = strtok_r(maskptr, ":", &saveptr);
+ if (id_str == NULL)
+ break;
+ if (gf_string2int(id_str, &id)) {
+ gf_msg(ec->xl->name, GF_LOG_ERROR, 0, EC_MSG_XLATOR_INIT_FAIL,
+ "In read-mask \"%s\" id %s is not a valid integer",
+ read_mask_str, id_str);
+ ret = -1;
+ goto out;
+ }
+
+ if ((id < 0) || (id >= ec->nodes)) {
+ gf_msg(ec->xl->name, GF_LOG_ERROR, 0, EC_MSG_XLATOR_INIT_FAIL,
+ "In read-mask \"%s\" id %d is not in range [0 - %d]",
+ read_mask_str, id, ec->nodes - 1);
+ ret = -1;
+ goto out;
+ }
+ read_mask |= (1UL << id);
+ maskptr = NULL;
+ }
+
+ if (gf_bits_count(read_mask) < ec->fragments) {
+ gf_msg(ec->xl->name, GF_LOG_ERROR, 0, EC_MSG_XLATOR_INIT_FAIL,
+ "read-mask \"%s\" should contain at least %d ids", read_mask_str,
+ ec->fragments);
+ ret = -1;
+ goto out;
+ }
+ ec->read_mask = read_mask;
+ ret = 0;
+out:
+ GF_FREE(mask);
+ return ret;
}
int32_t
-init (xlator_t *this)
+init(xlator_t *this)
{
- ec_t *ec = NULL;
+ ec_t *ec = NULL;
char *read_policy = NULL;
- char *extensions = NULL;
+ char *extensions = NULL;
int32_t err;
+ char *read_mask_str = NULL;
- if (this->parents == NULL)
- {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- EC_MSG_NO_PARENTS, "Volume does not have parents.");
+ if (this->parents == NULL) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, EC_MSG_NO_PARENTS,
+ "Volume does not have parents.");
}
ec = GF_MALLOC(sizeof(*ec), ec_mt_ec_t);
- if (ec == NULL)
- {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY, "Failed to allocate private memory.");
+ if (ec == NULL) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to allocate private memory.");
return -1;
}
@@ -672,6 +803,7 @@ init (xlator_t *this)
ec->xl = this;
LOCK_INIT(&ec->lock);
+ GF_ATOMIC_INIT(ec->async_fop_count, 0);
INIT_LIST_HEAD(&ec->pending_fops);
INIT_LIST_HEAD(&ec->heal_waiting);
INIT_LIST_HEAD(&ec->healing);
@@ -680,26 +812,23 @@ init (xlator_t *this)
ec->cbk_pool = mem_pool_new(ec_cbk_data_t, 4096);
ec->lock_pool = mem_pool_new(ec_lock_t, 1024);
if ((ec->fop_pool == NULL) || (ec->cbk_pool == NULL) ||
- (ec->lock_pool == NULL))
- {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- EC_MSG_NO_MEMORY, "Failed to create memory pools.");
+ (ec->lock_pool == NULL)) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, EC_MSG_NO_MEMORY,
+ "Failed to create memory pools.");
goto failed;
}
- if (ec_prepare_childs(this) != 0)
- {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_XLATOR_INIT_FAIL, "Failed to initialize xlator");
+ if (ec_prepare_childs(this) != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_XLATOR_INIT_FAIL,
+ "Failed to initialize xlator");
goto failed;
}
- if (ec_parse_options(this) != 0)
- {
- gf_msg (this->name, GF_LOG_ERROR, EINVAL,
- EC_MSG_XLATOR_PARSE_OPT_FAIL, "Failed to parse xlator options");
+ if (ec_parse_options(this) != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL, EC_MSG_XLATOR_PARSE_OPT_FAIL,
+ "Failed to parse xlator options");
goto failed;
}
@@ -709,58 +838,65 @@ init (xlator_t *this)
err = ec_method_init(this, &ec->matrix, ec->fragments, ec->nodes,
ec->nodes * 2, extensions);
if (err != 0) {
- gf_msg (this->name, GF_LOG_ERROR, -err, EC_MSG_MATRIX_FAILED,
- "Failed to initialize matrix management");
+ gf_msg(this->name, GF_LOG_ERROR, -err, EC_MSG_MATRIX_FAILED,
+ "Failed to initialize matrix management");
goto failed;
}
- GF_OPTION_INIT ("self-heal-daemon", ec->shd.enabled, bool, failed);
- GF_OPTION_INIT ("iam-self-heal-daemon", ec->shd.iamshd, bool, failed);
- GF_OPTION_INIT ("eager-lock", ec->eager_lock, bool, failed);
- GF_OPTION_INIT ("other-eager-lock", ec->other_eager_lock, bool, failed);
- GF_OPTION_INIT ("eager-lock-timeout", ec->eager_lock_timeout, uint32,
- failed);
- GF_OPTION_INIT ("other-eager-lock-timeout", ec->other_eager_lock_timeout,
- uint32, failed);
- GF_OPTION_INIT ("background-heals", ec->background_heals, uint32, failed);
- GF_OPTION_INIT ("heal-wait-qlength", ec->heal_wait_qlen, uint32, failed);
- GF_OPTION_INIT ("self-heal-window-size", ec->self_heal_window_size, uint32,
- failed);
- ec_configure_background_heal_opts (ec, ec->background_heals,
- ec->heal_wait_qlen);
- GF_OPTION_INIT ("read-policy", read_policy, str, failed);
- if (ec_assign_read_policy (ec, read_policy))
- goto failed;
-
- GF_OPTION_INIT ("shd-max-threads", ec->shd.max_threads, uint32, failed);
- GF_OPTION_INIT ("shd-wait-qlength", ec->shd.wait_qlength, uint32, failed);
- GF_OPTION_INIT ("optimistic-change-log", ec->optimistic_changelog, bool, failed);
- GF_OPTION_INIT ("parallel-writes", ec->parallel_writes, bool, failed);
- GF_OPTION_INIT ("stripe-cache", ec->stripe_cache, uint32, failed);
-
- this->itable = inode_table_new (EC_SHD_INODE_LRU_LIMIT, this);
+ GF_OPTION_INIT("self-heal-daemon", ec->shd.enabled, bool, failed);
+ GF_OPTION_INIT("iam-self-heal-daemon", ec->shd.iamshd, bool, failed);
+ GF_OPTION_INIT("eager-lock", ec->eager_lock, bool, failed);
+ GF_OPTION_INIT("other-eager-lock", ec->other_eager_lock, bool, failed);
+ GF_OPTION_INIT("eager-lock-timeout", ec->eager_lock_timeout, uint32,
+ failed);
+ GF_OPTION_INIT("other-eager-lock-timeout", ec->other_eager_lock_timeout,
+ uint32, failed);
+ GF_OPTION_INIT("background-heals", ec->background_heals, uint32, failed);
+ GF_OPTION_INIT("heal-wait-qlength", ec->heal_wait_qlen, uint32, failed);
+ GF_OPTION_INIT("self-heal-window-size", ec->self_heal_window_size, uint32,
+ failed);
+ ec_configure_background_heal_opts(ec, ec->background_heals,
+ ec->heal_wait_qlen);
+ GF_OPTION_INIT("read-policy", read_policy, str, failed);
+ if (ec_assign_read_policy(ec, read_policy))
+ goto failed;
+
+ GF_OPTION_INIT("heal-timeout", ec->shd.timeout, int32, failed);
+ GF_OPTION_INIT("shd-max-threads", ec->shd.max_threads, uint32, failed);
+ GF_OPTION_INIT("shd-wait-qlength", ec->shd.wait_qlength, uint32, failed);
+ GF_OPTION_INIT("optimistic-change-log", ec->optimistic_changelog, bool,
+ failed);
+ GF_OPTION_INIT("parallel-writes", ec->parallel_writes, bool, failed);
+ GF_OPTION_INIT("stripe-cache", ec->stripe_cache, uint32, failed);
+ GF_OPTION_INIT("quorum-count", ec->quorum_count, uint32, failed);
+ GF_OPTION_INIT("ec-read-mask", read_mask_str, str, failed);
+
+ if (ec_assign_read_mask(ec, read_mask_str))
+ goto failed;
+
+ this->itable = inode_table_new(EC_SHD_INODE_LRU_LIMIT, this);
if (!this->itable)
- goto failed;
+ goto failed;
if (ec->shd.iamshd)
- ec_selfheal_daemon_init (this);
- gf_msg_debug (this->name, 0, "Disperse translator initialized.");
+ ec_selfheal_daemon_init(this);
+ gf_msg_debug(this->name, 0, "Disperse translator initialized.");
- ec->leaf_to_subvolid = dict_new ();
+ ec->leaf_to_subvolid = dict_new();
if (!ec->leaf_to_subvolid)
- goto failed;
- if (glusterfs_reachable_leaves (this, ec->leaf_to_subvolid)) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_SUBVOL_BUILD_FAIL, "Failed to build subvol "
- "dictionary");
+ goto failed;
+ if (glusterfs_reachable_leaves(this, ec->leaf_to_subvolid)) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_SUBVOL_BUILD_FAIL,
+ "Failed to build subvol "
+ "dictionary");
goto failed;
}
- if (ec_subvol_to_subvol_id_transform (ec, ec->leaf_to_subvolid) < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- EC_MSG_SUBVOL_ID_DICT_SET_FAIL, "Failed to build subvol-id "
- "dictionary");
+ if (ec_subvol_to_subvol_id_transform(ec, ec->leaf_to_subvolid) < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, EC_MSG_SUBVOL_ID_DICT_SET_FAIL,
+ "Failed to build subvol-id "
+ "dictionary");
goto failed;
}
@@ -774,13 +910,16 @@ failed:
return -1;
}
-void fini(xlator_t * this)
+void
+fini(xlator_t *this)
{
+ ec_selfheal_daemon_fini(this);
__ec_destroy_private(this);
}
-int32_t ec_gf_access(call_frame_t * frame, xlator_t * this, loc_t * loc,
- int32_t mask, dict_t * xdata)
+int32_t
+ec_gf_access(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+ dict_t *xdata)
{
ec_access(frame, this, -1, EC_MINIMUM_ONE, default_access_cbk, NULL, loc,
mask, xdata);
@@ -788,9 +927,9 @@ int32_t ec_gf_access(call_frame_t * frame, xlator_t * this, loc_t * loc,
return 0;
}
-int32_t ec_gf_create(call_frame_t * frame, xlator_t * this, loc_t * loc,
- int32_t flags, mode_t mode, mode_t umask, fd_t * fd,
- dict_t * xdata)
+int32_t
+ec_gf_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
ec_create(frame, this, -1, EC_MINIMUM_MIN, default_create_cbk, NULL, loc,
flags, mode, umask, fd, xdata);
@@ -798,53 +937,58 @@ int32_t ec_gf_create(call_frame_t * frame, xlator_t * this, loc_t * loc,
return 0;
}
-int32_t ec_gf_discard(call_frame_t * frame, xlator_t * this, fd_t * fd,
- off_t offset, size_t len, dict_t * xdata)
+int32_t
+ec_gf_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
{
- ec_discard(frame, this, -1, EC_MINIMUM_MIN, default_discard_cbk,
- NULL, fd, offset, len, xdata);
+ ec_discard(frame, this, -1, EC_MINIMUM_MIN, default_discard_cbk, NULL, fd,
+ offset, len, xdata);
return 0;
}
-int32_t ec_gf_entrylk(call_frame_t * frame, xlator_t * this,
- const char * volume, loc_t * loc, const char * basename,
- entrylk_cmd cmd, entrylk_type type, dict_t * xdata)
+int32_t
+ec_gf_entrylk(call_frame_t *frame, xlator_t *this, const char *volume,
+ loc_t *loc, const char *basename, entrylk_cmd cmd,
+ entrylk_type type, dict_t *xdata)
{
- int32_t minimum = EC_MINIMUM_ALL;
+ uint32_t fop_flags = EC_MINIMUM_ALL;
+
if (cmd == ENTRYLK_UNLOCK)
- minimum = EC_MINIMUM_ONE;
- ec_entrylk(frame, this, -1, minimum, default_entrylk_cbk, NULL,
- volume, loc, basename, cmd, type, xdata);
+ fop_flags = EC_MINIMUM_ONE;
+ ec_entrylk(frame, this, -1, fop_flags, default_entrylk_cbk, NULL, volume,
+ loc, basename, cmd, type, xdata);
return 0;
}
-int32_t ec_gf_fentrylk(call_frame_t * frame, xlator_t * this,
- const char * volume, fd_t * fd, const char * basename,
- entrylk_cmd cmd, entrylk_type type, dict_t * xdata)
+int32_t
+ec_gf_fentrylk(call_frame_t *frame, xlator_t *this, const char *volume,
+ fd_t *fd, const char *basename, entrylk_cmd cmd,
+ entrylk_type type, dict_t *xdata)
{
- int32_t minimum = EC_MINIMUM_ALL;
+ uint32_t fop_flags = EC_MINIMUM_ALL;
+
if (cmd == ENTRYLK_UNLOCK)
- minimum = EC_MINIMUM_ONE;
- ec_fentrylk(frame, this, -1, minimum, default_fentrylk_cbk, NULL,
- volume, fd, basename, cmd, type, xdata);
+ fop_flags = EC_MINIMUM_ONE;
+ ec_fentrylk(frame, this, -1, fop_flags, default_fentrylk_cbk, NULL, volume,
+ fd, basename, cmd, type, xdata);
return 0;
}
-int32_t ec_gf_fallocate(call_frame_t * frame, xlator_t * this, fd_t * fd,
- int32_t mode, off_t offset, size_t len,
- dict_t * xdata)
+int32_t
+ec_gf_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+ off_t offset, size_t len, dict_t *xdata)
{
- ec_fallocate(frame, this, -1, EC_MINIMUM_MIN, default_fallocate_cbk,
- NULL, fd, mode, offset, len, xdata);
+ ec_fallocate(frame, this, -1, EC_MINIMUM_MIN, default_fallocate_cbk, NULL,
+ fd, mode, offset, len, xdata);
return 0;
}
-int32_t ec_gf_flush(call_frame_t * frame, xlator_t * this, fd_t * fd,
- dict_t * xdata)
+int32_t
+ec_gf_flush(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
ec_flush(frame, this, -1, EC_MINIMUM_MIN, default_flush_cbk, NULL, fd,
xdata);
@@ -852,8 +996,9 @@ int32_t ec_gf_flush(call_frame_t * frame, xlator_t * this, fd_t * fd,
return 0;
}
-int32_t ec_gf_fsync(call_frame_t * frame, xlator_t * this, fd_t * fd,
- int32_t datasync, dict_t * xdata)
+int32_t
+ec_gf_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+ dict_t *xdata)
{
ec_fsync(frame, this, -1, EC_MINIMUM_MIN, default_fsync_cbk, NULL, fd,
datasync, xdata);
@@ -861,136 +1006,139 @@ int32_t ec_gf_fsync(call_frame_t * frame, xlator_t * this, fd_t * fd,
return 0;
}
-int32_t ec_gf_fsyncdir(call_frame_t * frame, xlator_t * this, fd_t * fd,
- int32_t datasync, dict_t * xdata)
+int32_t
+ec_gf_fsyncdir(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+ dict_t *xdata)
{
- ec_fsyncdir(frame, this, -1, EC_MINIMUM_MIN, default_fsyncdir_cbk, NULL,
- fd, datasync, xdata);
+ ec_fsyncdir(frame, this, -1, EC_MINIMUM_MIN, default_fsyncdir_cbk, NULL, fd,
+ datasync, xdata);
return 0;
}
int
-ec_marker_populate_args (call_frame_t *frame, int type, int *gauge,
- xlator_t **subvols)
+ec_marker_populate_args(call_frame_t *frame, int type, int *gauge,
+ xlator_t **subvols)
{
- xlator_t *this = frame->this;
- ec_t *ec = this->private;
+ xlator_t *this = frame->this;
+ ec_t *ec = this->private;
- memcpy (subvols, ec->xl_list, sizeof (*subvols) * ec->nodes);
+ memcpy(subvols, ec->xl_list, sizeof(*subvols) * ec->nodes);
- if (type == MARKER_XTIME_TYPE) {
- /*Don't error out on ENOENT/ENOTCONN */
- gauge[MCNT_NOTFOUND] = 0;
- gauge[MCNT_ENOTCONN] = 0;
- }
+ if (type == MARKER_XTIME_TYPE) {
+ /*Don't error out on ENOENT/ENOTCONN */
+ gauge[MCNT_NOTFOUND] = 0;
+ gauge[MCNT_ENOTCONN] = 0;
+ }
- return ec->nodes;
+ return ec->nodes;
}
int32_t
-ec_handle_heal_commands (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+ec_handle_heal_commands(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- dict_t *dict_rsp = NULL;
- int op_ret = -1;
- int op_errno = ENOMEM;
+ dict_t *dict_rsp = NULL;
+ int op_ret = -1;
+ int op_errno = ENOMEM;
- if (!name || strcmp (name, GF_HEAL_INFO))
- return -1;
+ if (!name || strcmp(name, GF_HEAL_INFO))
+ return -1;
- op_errno = -ec_get_heal_info (this, loc, &dict_rsp);
- if (op_errno <= 0) {
- op_errno = op_ret = 0;
- }
+ op_errno = -ec_get_heal_info(this, loc, &dict_rsp);
+ if (op_errno <= 0) {
+ op_errno = op_ret = 0;
+ }
- STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict_rsp, NULL);
- if (dict_rsp)
- dict_unref (dict_rsp);
- return 0;
+ STACK_UNWIND_STRICT(getxattr, frame, op_ret, op_errno, dict_rsp, NULL);
+ if (dict_rsp)
+ dict_unref(dict_rsp);
+ return 0;
}
int32_t
-ec_gf_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+ec_gf_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- int error = 0;
- ec_t *ec = this->private;
- int32_t minimum = EC_MINIMUM_ONE;
+ int error = 0;
+ ec_t *ec = this->private;
+ int32_t fop_flags = EC_MINIMUM_ONE;
- if (name && strcmp (name, EC_XATTR_HEAL) != 0) {
- EC_INTERNAL_XATTR_OR_GOTO(name, NULL, error, out);
- }
+ if (name && strcmp(name, EC_XATTR_HEAL) != 0) {
+ EC_INTERNAL_XATTR_OR_GOTO(name, NULL, error, out);
+ }
- if (ec_handle_heal_commands (frame, this, loc, name, xdata) == 0)
- return 0;
+ if (ec_handle_heal_commands(frame, this, loc, name, xdata) == 0)
+ return 0;
- if (cluster_handle_marker_getxattr (frame, loc, name, ec->vol_uuid,
- NULL, ec_marker_populate_args) == 0)
- return 0;
+ if (cluster_handle_marker_getxattr(frame, loc, name, ec->vol_uuid, NULL,
+ ec_marker_populate_args) == 0)
+ return 0;
- if (name &&
- ((fnmatch (GF_XATTR_STIME_PATTERN, name, 0) == 0) ||
- XATTR_IS_NODE_UUID(name) || XATTR_IS_NODE_UUID_LIST(name))) {
- minimum = EC_MINIMUM_ALL;
- }
+ if (name && ((fnmatch(GF_XATTR_STIME_PATTERN, name, 0) == 0) ||
+ XATTR_IS_NODE_UUID(name) || XATTR_IS_NODE_UUID_LIST(name))) {
+ fop_flags = EC_MINIMUM_ALL;
+ }
- ec_getxattr (frame, this, -1, minimum, default_getxattr_cbk,
- NULL, loc, name, xdata);
+ ec_getxattr(frame, this, -1, fop_flags, default_getxattr_cbk, NULL, loc,
+ name, xdata);
- return 0;
+ return 0;
out:
- error = ENODATA;
- STACK_UNWIND_STRICT (getxattr, frame, -1, error, NULL, NULL);
- return 0;
+ error = ENODATA;
+ STACK_UNWIND_STRICT(getxattr, frame, -1, error, NULL, NULL);
+ return 0;
}
int32_t
-ec_gf_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
+ec_gf_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
+ dict_t *xdata)
{
- int error = 0;
+ int error = 0;
- EC_INTERNAL_XATTR_OR_GOTO(name, NULL, error, out);
+ EC_INTERNAL_XATTR_OR_GOTO(name, NULL, error, out);
- ec_fgetxattr (frame, this, -1, EC_MINIMUM_ONE, default_fgetxattr_cbk,
- NULL, fd, name, xdata);
- return 0;
+ ec_fgetxattr(frame, this, -1, EC_MINIMUM_ONE, default_fgetxattr_cbk, NULL,
+ fd, name, xdata);
+ return 0;
out:
- error = ENODATA;
- STACK_UNWIND_STRICT (fgetxattr, frame, -1, error, NULL, NULL);
- return 0;
+ error = ENODATA;
+ STACK_UNWIND_STRICT(fgetxattr, frame, -1, error, NULL, NULL);
+ return 0;
}
-int32_t ec_gf_inodelk(call_frame_t * frame, xlator_t * this,
- const char * volume, loc_t * loc, int32_t cmd,
- struct gf_flock * flock, dict_t * xdata)
+int32_t
+ec_gf_inodelk(call_frame_t *frame, xlator_t *this, const char *volume,
+ loc_t *loc, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
{
- int32_t minimum = EC_MINIMUM_ALL;
+ int32_t fop_flags = EC_MINIMUM_ALL;
+
if (flock->l_type == F_UNLCK)
- minimum = EC_MINIMUM_ONE;
+ fop_flags = EC_MINIMUM_ONE;
- ec_inodelk(frame, this, &frame->root->lk_owner, -1, minimum,
+ ec_inodelk(frame, this, &frame->root->lk_owner, -1, fop_flags,
default_inodelk_cbk, NULL, volume, loc, cmd, flock, xdata);
return 0;
}
-int32_t ec_gf_finodelk(call_frame_t * frame, xlator_t * this,
- const char * volume, fd_t * fd, int32_t cmd,
- struct gf_flock * flock, dict_t * xdata)
+int32_t
+ec_gf_finodelk(call_frame_t *frame, xlator_t *this, const char *volume,
+ fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
{
- int32_t minimum = EC_MINIMUM_ALL;
+ int32_t fop_flags = EC_MINIMUM_ALL;
+
if (flock->l_type == F_UNLCK)
- minimum = EC_MINIMUM_ONE;
- ec_finodelk(frame, this, &frame->root->lk_owner, -1, minimum,
+ fop_flags = EC_MINIMUM_ONE;
+ ec_finodelk(frame, this, &frame->root->lk_owner, -1, fop_flags,
default_finodelk_cbk, NULL, volume, fd, cmd, flock, xdata);
return 0;
}
-int32_t ec_gf_link(call_frame_t * frame, xlator_t * this, loc_t * oldloc,
- loc_t * newloc, dict_t * xdata)
+int32_t
+ec_gf_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
ec_link(frame, this, -1, EC_MINIMUM_MIN, default_link_cbk, NULL, oldloc,
newloc, xdata);
@@ -998,20 +1146,22 @@ int32_t ec_gf_link(call_frame_t * frame, xlator_t * this, loc_t * oldloc,
return 0;
}
-int32_t ec_gf_lk(call_frame_t * frame, xlator_t * this, fd_t * fd,
- int32_t cmd, struct gf_flock * flock, dict_t * xdata)
+int32_t
+ec_gf_lk(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
+ struct gf_flock *flock, dict_t *xdata)
{
- int32_t minimum = EC_MINIMUM_ALL;
+ int32_t fop_flags = EC_MINIMUM_ALL;
+
if (flock->l_type == F_UNLCK)
- minimum = EC_MINIMUM_ONE;
- ec_lk(frame, this, -1, minimum, default_lk_cbk, NULL, fd, cmd,
- flock, xdata);
+ fop_flags = EC_MINIMUM_ONE;
+ ec_lk(frame, this, -1, fop_flags, default_lk_cbk, NULL, fd, cmd, flock,
+ xdata);
return 0;
}
-int32_t ec_gf_lookup(call_frame_t * frame, xlator_t * this, loc_t * loc,
- dict_t * xdata)
+int32_t
+ec_gf_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
ec_lookup(frame, this, -1, EC_MINIMUM_MIN, default_lookup_cbk, NULL, loc,
xdata);
@@ -1019,8 +1169,9 @@ int32_t ec_gf_lookup(call_frame_t * frame, xlator_t * this, loc_t * loc,
return 0;
}
-int32_t ec_gf_mkdir(call_frame_t * frame, xlator_t * this, loc_t * loc,
- mode_t mode, mode_t umask, dict_t * xdata)
+int32_t
+ec_gf_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
ec_mkdir(frame, this, -1, EC_MINIMUM_MIN, default_mkdir_cbk, NULL, loc,
mode, umask, xdata);
@@ -1028,8 +1179,9 @@ int32_t ec_gf_mkdir(call_frame_t * frame, xlator_t * this, loc_t * loc,
return 0;
}
-int32_t ec_gf_mknod(call_frame_t * frame, xlator_t * this, loc_t * loc,
- mode_t mode, dev_t rdev, mode_t umask, dict_t * xdata)
+int32_t
+ec_gf_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *xdata)
{
ec_mknod(frame, this, -1, EC_MINIMUM_MIN, default_mknod_cbk, NULL, loc,
mode, rdev, umask, xdata);
@@ -1037,17 +1189,19 @@ int32_t ec_gf_mknod(call_frame_t * frame, xlator_t * this, loc_t * loc,
return 0;
}
-int32_t ec_gf_open(call_frame_t * frame, xlator_t * this, loc_t * loc,
- int32_t flags, fd_t * fd, dict_t * xdata)
+int32_t
+ec_gf_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
{
- ec_open(frame, this, -1, EC_MINIMUM_MIN, default_open_cbk, NULL, loc,
- flags, fd, xdata);
+ ec_open(frame, this, -1, EC_MINIMUM_MIN, default_open_cbk, NULL, loc, flags,
+ fd, xdata);
return 0;
}
-int32_t ec_gf_opendir(call_frame_t * frame, xlator_t * this, loc_t * loc,
- fd_t * fd, dict_t * xdata)
+int32_t
+ec_gf_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata)
{
ec_opendir(frame, this, -1, EC_MINIMUM_MIN, default_opendir_cbk, NULL, loc,
fd, xdata);
@@ -1055,8 +1209,9 @@ int32_t ec_gf_opendir(call_frame_t * frame, xlator_t * this, loc_t * loc,
return 0;
}
-int32_t ec_gf_readdir(call_frame_t * frame, xlator_t * this, fd_t * fd,
- size_t size, off_t offset, dict_t * xdata)
+int32_t
+ec_gf_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *xdata)
{
ec_readdir(frame, this, -1, EC_MINIMUM_ONE, default_readdir_cbk, NULL, fd,
size, offset, xdata);
@@ -1064,17 +1219,19 @@ int32_t ec_gf_readdir(call_frame_t * frame, xlator_t * this, fd_t * fd,
return 0;
}
-int32_t ec_gf_readdirp(call_frame_t * frame, xlator_t * this, fd_t * fd,
- size_t size, off_t offset, dict_t * xdata)
+int32_t
+ec_gf_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *xdata)
{
- ec_readdirp(frame, this, -1, EC_MINIMUM_ONE, default_readdirp_cbk, NULL,
- fd, size, offset, xdata);
+ ec_readdirp(frame, this, -1, EC_MINIMUM_ONE, default_readdirp_cbk, NULL, fd,
+ size, offset, xdata);
return 0;
}
-int32_t ec_gf_readlink(call_frame_t * frame, xlator_t * this, loc_t * loc,
- size_t size, dict_t * xdata)
+int32_t
+ec_gf_readlink(call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+ dict_t *xdata)
{
ec_readlink(frame, this, -1, EC_MINIMUM_ONE, default_readlink_cbk, NULL,
loc, size, xdata);
@@ -1082,60 +1239,63 @@ int32_t ec_gf_readlink(call_frame_t * frame, xlator_t * this, loc_t * loc,
return 0;
}
-int32_t ec_gf_readv(call_frame_t * frame, xlator_t * this, fd_t * fd,
- size_t size, off_t offset, uint32_t flags, dict_t * xdata)
+int32_t
+ec_gf_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- ec_readv(frame, this, -1, EC_MINIMUM_MIN, default_readv_cbk, NULL, fd,
- size, offset, flags, xdata);
+ ec_readv(frame, this, -1, EC_MINIMUM_MIN, default_readv_cbk, NULL, fd, size,
+ offset, flags, xdata);
return 0;
}
int32_t
-ec_gf_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+ec_gf_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- int error = 0;
+ int error = 0;
- EC_INTERNAL_XATTR_OR_GOTO (name, xdata, error, out);
+ EC_INTERNAL_XATTR_OR_GOTO(name, xdata, error, out);
- ec_removexattr (frame, this, -1, EC_MINIMUM_MIN,
- default_removexattr_cbk, NULL, loc, name, xdata);
+ ec_removexattr(frame, this, -1, EC_MINIMUM_MIN, default_removexattr_cbk,
+ NULL, loc, name, xdata);
- return 0;
+ return 0;
out:
- STACK_UNWIND_STRICT (removexattr, frame, -1, error, NULL);
- return 0;
+ STACK_UNWIND_STRICT(removexattr, frame, -1, error, NULL);
+ return 0;
}
int32_t
-ec_gf_fremovexattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
+ec_gf_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata)
{
- int error = 0;
+ int error = 0;
- EC_INTERNAL_XATTR_OR_GOTO (name, xdata, error, out);
+ EC_INTERNAL_XATTR_OR_GOTO(name, xdata, error, out);
- ec_fremovexattr (frame, this, -1, EC_MINIMUM_MIN,
- default_fremovexattr_cbk, NULL, fd, name, xdata);
+ ec_fremovexattr(frame, this, -1, EC_MINIMUM_MIN, default_fremovexattr_cbk,
+ NULL, fd, name, xdata);
- return 0;
+ return 0;
out:
- STACK_UNWIND_STRICT (fremovexattr, frame, -1, error, NULL);
- return 0;
+ STACK_UNWIND_STRICT(fremovexattr, frame, -1, error, NULL);
+ return 0;
}
-int32_t ec_gf_rename(call_frame_t * frame, xlator_t * this, loc_t * oldloc,
- loc_t * newloc, dict_t * xdata)
+int32_t
+ec_gf_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- ec_rename(frame, this, -1, EC_MINIMUM_MIN, default_rename_cbk, NULL,
- oldloc, newloc, xdata);
+ ec_rename(frame, this, -1, EC_MINIMUM_MIN, default_rename_cbk, NULL, oldloc,
+ newloc, xdata);
return 0;
}
-int32_t ec_gf_rmdir(call_frame_t * frame, xlator_t * this, loc_t * loc,
- int xflags, dict_t * xdata)
+int32_t
+ec_gf_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflags,
+ dict_t *xdata)
{
ec_rmdir(frame, this, -1, EC_MINIMUM_MIN, default_rmdir_cbk, NULL, loc,
xflags, xdata);
@@ -1143,8 +1303,9 @@ int32_t ec_gf_rmdir(call_frame_t * frame, xlator_t * this, loc_t * loc,
return 0;
}
-int32_t ec_gf_setattr(call_frame_t * frame, xlator_t * this, loc_t * loc,
- struct iatt * stbuf, int32_t valid, dict_t * xdata)
+int32_t
+ec_gf_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
ec_setattr(frame, this, -1, EC_MINIMUM_MIN, default_setattr_cbk, NULL, loc,
stbuf, valid, xdata);
@@ -1152,51 +1313,52 @@ int32_t ec_gf_setattr(call_frame_t * frame, xlator_t * this, loc_t * loc,
return 0;
}
-int32_t ec_gf_fsetattr(call_frame_t * frame, xlator_t * this, fd_t * fd,
- struct iatt * stbuf, int32_t valid, dict_t * xdata)
+int32_t
+ec_gf_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- ec_fsetattr(frame, this, -1, EC_MINIMUM_MIN, default_fsetattr_cbk, NULL,
- fd, stbuf, valid, xdata);
+ ec_fsetattr(frame, this, -1, EC_MINIMUM_MIN, default_fsetattr_cbk, NULL, fd,
+ stbuf, valid, xdata);
return 0;
}
int32_t
-ec_gf_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *dict, int32_t flags, dict_t *xdata)
+ec_gf_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- int error = 0;
+ int error = 0;
- EC_INTERNAL_XATTR_OR_GOTO ("", dict, error, out);
+ EC_INTERNAL_XATTR_OR_GOTO("", dict, error, out);
- ec_setxattr (frame, this, -1, EC_MINIMUM_MIN, default_setxattr_cbk,
- NULL, loc, dict, flags, xdata);
+ ec_setxattr(frame, this, -1, EC_MINIMUM_MIN, default_setxattr_cbk, NULL,
+ loc, dict, flags, xdata);
- return 0;
+ return 0;
out:
- STACK_UNWIND_STRICT (setxattr, frame, -1, error, NULL);
- return 0;
+ STACK_UNWIND_STRICT(setxattr, frame, -1, error, NULL);
+ return 0;
}
int32_t
-ec_gf_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- dict_t *dict, int32_t flags, dict_t *xdata)
+ec_gf_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- int error = 0;
+ int error = 0;
- EC_INTERNAL_XATTR_OR_GOTO ("", dict, error, out);
+ EC_INTERNAL_XATTR_OR_GOTO("", dict, error, out);
- ec_fsetxattr (frame, this, -1, EC_MINIMUM_MIN, default_fsetxattr_cbk,
- NULL, fd, dict, flags, xdata);
+ ec_fsetxattr(frame, this, -1, EC_MINIMUM_MIN, default_fsetxattr_cbk, NULL,
+ fd, dict, flags, xdata);
- return 0;
+ return 0;
out:
- STACK_UNWIND_STRICT (fsetxattr, frame, -1, error, NULL);
- return 0;
+ STACK_UNWIND_STRICT(fsetxattr, frame, -1, error, NULL);
+ return 0;
}
-int32_t ec_gf_stat(call_frame_t * frame, xlator_t * this, loc_t * loc,
- dict_t * xdata)
+int32_t
+ec_gf_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
ec_stat(frame, this, -1, EC_MINIMUM_MIN, default_stat_cbk, NULL, loc,
xdata);
@@ -1204,8 +1366,8 @@ int32_t ec_gf_stat(call_frame_t * frame, xlator_t * this, loc_t * loc,
return 0;
}
-int32_t ec_gf_fstat(call_frame_t * frame, xlator_t * this, fd_t * fd,
- dict_t * xdata)
+int32_t
+ec_gf_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
ec_fstat(frame, this, -1, EC_MINIMUM_MIN, default_fstat_cbk, NULL, fd,
xdata);
@@ -1213,8 +1375,8 @@ int32_t ec_gf_fstat(call_frame_t * frame, xlator_t * this, fd_t * fd,
return 0;
}
-int32_t ec_gf_statfs(call_frame_t * frame, xlator_t * this, loc_t * loc,
- dict_t * xdata)
+int32_t
+ec_gf_statfs(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
ec_statfs(frame, this, -1, EC_MINIMUM_MIN, default_statfs_cbk, NULL, loc,
xdata);
@@ -1222,9 +1384,9 @@ int32_t ec_gf_statfs(call_frame_t * frame, xlator_t * this, loc_t * loc,
return 0;
}
-int32_t ec_gf_symlink(call_frame_t * frame, xlator_t * this,
- const char * linkname, loc_t * loc, mode_t umask,
- dict_t * xdata)
+int32_t
+ec_gf_symlink(call_frame_t *frame, xlator_t *this, const char *linkname,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
ec_symlink(frame, this, -1, EC_MINIMUM_MIN, default_symlink_cbk, NULL,
linkname, loc, umask, xdata);
@@ -1232,8 +1394,9 @@ int32_t ec_gf_symlink(call_frame_t * frame, xlator_t * this,
return 0;
}
-int32_t ec_gf_truncate(call_frame_t * frame, xlator_t * this, loc_t * loc,
- off_t offset, dict_t * xdata)
+int32_t
+ec_gf_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
ec_truncate(frame, this, -1, EC_MINIMUM_MIN, default_truncate_cbk, NULL,
loc, offset, xdata);
@@ -1241,8 +1404,9 @@ int32_t ec_gf_truncate(call_frame_t * frame, xlator_t * this, loc_t * loc,
return 0;
}
-int32_t ec_gf_ftruncate(call_frame_t * frame, xlator_t * this, fd_t * fd,
- off_t offset, dict_t * xdata)
+int32_t
+ec_gf_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
ec_ftruncate(frame, this, -1, EC_MINIMUM_MIN, default_ftruncate_cbk, NULL,
fd, offset, xdata);
@@ -1250,8 +1414,9 @@ int32_t ec_gf_ftruncate(call_frame_t * frame, xlator_t * this, fd_t * fd,
return 0;
}
-int32_t ec_gf_unlink(call_frame_t * frame, xlator_t * this, loc_t * loc,
- int xflags, dict_t * xdata)
+int32_t
+ec_gf_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflags,
+ dict_t *xdata)
{
ec_unlink(frame, this, -1, EC_MINIMUM_MIN, default_unlink_cbk, NULL, loc,
xflags, xdata);
@@ -1259,9 +1424,10 @@ int32_t ec_gf_unlink(call_frame_t * frame, xlator_t * this, loc_t * loc,
return 0;
}
-int32_t ec_gf_writev(call_frame_t * frame, xlator_t * this, fd_t * fd,
- struct iovec * vector, int32_t count, off_t offset,
- uint32_t flags, struct iobref * iobref, dict_t * xdata)
+int32_t
+ec_gf_writev(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int32_t count, off_t offset, uint32_t flags,
+ struct iobref *iobref, dict_t *xdata)
{
ec_writev(frame, this, -1, EC_MINIMUM_MIN, default_writev_cbk, NULL, fd,
vector, count, offset, flags, iobref, xdata);
@@ -1269,9 +1435,9 @@ int32_t ec_gf_writev(call_frame_t * frame, xlator_t * this, fd_t * fd,
return 0;
}
-int32_t ec_gf_xattrop(call_frame_t * frame, xlator_t * this, loc_t * loc,
- gf_xattrop_flags_t optype, dict_t * xattr,
- dict_t * xdata)
+int32_t
+ec_gf_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
ec_xattrop(frame, this, -1, EC_MINIMUM_MIN, default_xattrop_cbk, NULL, loc,
optype, xattr, xdata);
@@ -1279,47 +1445,49 @@ int32_t ec_gf_xattrop(call_frame_t * frame, xlator_t * this, loc_t * loc,
return 0;
}
-int32_t ec_gf_fxattrop(call_frame_t * frame, xlator_t * this, fd_t * fd,
- gf_xattrop_flags_t optype, dict_t * xattr,
- dict_t * xdata)
+int32_t
+ec_gf_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
- ec_fxattrop(frame, this, -1, EC_MINIMUM_MIN, default_fxattrop_cbk, NULL,
- fd, optype, xattr, xdata);
+ ec_fxattrop(frame, this, -1, EC_MINIMUM_MIN, default_fxattrop_cbk, NULL, fd,
+ optype, xattr, xdata);
return 0;
}
-int32_t ec_gf_zerofill(call_frame_t * frame, xlator_t * this, fd_t * fd,
- off_t offset, off_t len, dict_t * xdata)
+int32_t
+ec_gf_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ off_t len, dict_t *xdata)
{
default_zerofill_failure_cbk(frame, ENOTSUP);
return 0;
}
-int32_t ec_gf_seek(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- gf_seek_what_t what, dict_t *xdata)
+int32_t
+ec_gf_seek(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ gf_seek_what_t what, dict_t *xdata)
{
- ec_seek(frame, this, -1, EC_MINIMUM_ONE, default_seek_cbk, NULL, fd,
- offset, what, xdata);
+ ec_seek(frame, this, -1, EC_MINIMUM_ONE, default_seek_cbk, NULL, fd, offset,
+ what, xdata);
return 0;
}
-int32_t ec_gf_ipc(call_frame_t *frame, xlator_t *this, int32_t op,
- dict_t *xdata)
+int32_t
+ec_gf_ipc(call_frame_t *frame, xlator_t *this, int32_t op, dict_t *xdata)
{
ec_ipc(frame, this, -1, EC_MINIMUM_MIN, default_ipc_cbk, NULL, op, xdata);
return 0;
}
-int32_t ec_gf_forget(xlator_t * this, inode_t * inode)
+int32_t
+ec_gf_forget(xlator_t *this, inode_t *inode)
{
uint64_t value = 0;
- ec_inode_t * ctx = NULL;
+ ec_inode_t *ctx = NULL;
- if ((inode_ctx_del(inode, this, &value) == 0) && (value != 0))
- {
+ if ((inode_ctx_del(inode, this, &value) == 0) && (value != 0)) {
ctx = (ec_inode_t *)(uintptr_t)value;
/* We can only forget an inode if it has been unlocked, so the stripe
* cache should also be empty. */
@@ -1330,38 +1498,41 @@ int32_t ec_gf_forget(xlator_t * this, inode_t * inode)
return 0;
}
-void ec_gf_release_fd(xlator_t * this, fd_t * fd)
+void
+ec_gf_release_fd(xlator_t *this, fd_t *fd)
{
uint64_t value = 0;
- ec_fd_t * ctx = NULL;
+ ec_fd_t *ctx = NULL;
- if ((fd_ctx_del(fd, this, &value) == 0) && (value != 0))
- {
+ if ((fd_ctx_del(fd, this, &value) == 0) && (value != 0)) {
ctx = (ec_fd_t *)(uintptr_t)value;
loc_wipe(&ctx->loc);
GF_FREE(ctx);
}
}
-int32_t ec_gf_release(xlator_t * this, fd_t * fd)
+int32_t
+ec_gf_release(xlator_t *this, fd_t *fd)
{
ec_gf_release_fd(this, fd);
return 0;
}
-int32_t ec_gf_releasedir(xlator_t * this, fd_t * fd)
+int32_t
+ec_gf_releasedir(xlator_t *this, fd_t *fd)
{
ec_gf_release_fd(this, fd);
return 0;
}
-int32_t ec_dump_private(xlator_t *this)
+int32_t
+ec_dump_private(xlator_t *this)
{
ec_t *ec = NULL;
- char key_prefix[GF_DUMP_MAX_BUF_LEN];
- char tmp[65];
+ char key_prefix[GF_DUMP_MAX_BUF_LEN];
+ char tmp[65];
GF_ASSERT(this);
@@ -1369,7 +1540,8 @@ int32_t ec_dump_private(xlator_t *this)
GF_ASSERT(ec);
snprintf(key_prefix, GF_DUMP_MAX_BUF_LEN, "%s.%s", this->type, this->name);
- gf_proc_dump_add_section(key_prefix);
+ gf_proc_dump_add_section("%s", key_prefix);
+ gf_proc_dump_write("up", "%u", ec->up);
gf_proc_dump_write("nodes", "%u", ec->nodes);
gf_proc_dump_write("redundancy", "%u", ec->redundancy);
gf_proc_dump_write("fragment_size", "%u", ec->fragment_size);
@@ -1377,108 +1549,106 @@ int32_t ec_dump_private(xlator_t *this)
gf_proc_dump_write("childs_up", "%u", ec->xl_up_count);
gf_proc_dump_write("childs_up_mask", "%s",
ec_bin(tmp, sizeof(tmp), ec->xl_up, ec->nodes));
+ if (ec->read_mask) {
+ gf_proc_dump_write("read-mask", "%s",
+ ec_bin(tmp, sizeof(tmp), ec->read_mask, ec->nodes));
+ }
gf_proc_dump_write("background-heals", "%d", ec->background_heals);
gf_proc_dump_write("heal-wait-qlength", "%d", ec->heal_wait_qlen);
- gf_proc_dump_write("self-heal-window-size", "%"PRIu32,
+ gf_proc_dump_write("self-heal-window-size", "%" PRIu32,
ec->self_heal_window_size);
gf_proc_dump_write("healers", "%d", ec->healers);
gf_proc_dump_write("heal-waiters", "%d", ec->heal_waiters);
gf_proc_dump_write("read-policy", "%s", ec_read_policies[ec->read_policy]);
gf_proc_dump_write("parallel-writes", "%d", ec->parallel_writes);
+ gf_proc_dump_write("quorum-count", "%u", ec->quorum_count);
snprintf(key_prefix, GF_DUMP_MAX_BUF_LEN, "%s.%s.stats.stripe_cache",
this->type, this->name);
- gf_proc_dump_add_section(key_prefix);
+ gf_proc_dump_add_section("%s", key_prefix);
- gf_proc_dump_write("hits", "%llu",
+ gf_proc_dump_write("hits", "%" GF_PRI_ATOMIC,
GF_ATOMIC_GET(ec->stats.stripe_cache.hits));
- gf_proc_dump_write("misses", "%llu",
+ gf_proc_dump_write("misses", "%" GF_PRI_ATOMIC,
GF_ATOMIC_GET(ec->stats.stripe_cache.misses));
- gf_proc_dump_write("updates", "%llu",
+ gf_proc_dump_write("updates", "%" GF_PRI_ATOMIC,
GF_ATOMIC_GET(ec->stats.stripe_cache.updates));
- gf_proc_dump_write("invalidations", "%llu",
+ gf_proc_dump_write("invalidations", "%" GF_PRI_ATOMIC,
GF_ATOMIC_GET(ec->stats.stripe_cache.invals));
- gf_proc_dump_write("evicts", "%llu",
+ gf_proc_dump_write("evicts", "%" GF_PRI_ATOMIC,
GF_ATOMIC_GET(ec->stats.stripe_cache.evicts));
- gf_proc_dump_write("allocations", "%llu",
+ gf_proc_dump_write("allocations", "%" GF_PRI_ATOMIC,
GF_ATOMIC_GET(ec->stats.stripe_cache.allocs));
- gf_proc_dump_write("errors", "%llu",
+ gf_proc_dump_write("errors", "%" GF_PRI_ATOMIC,
GF_ATOMIC_GET(ec->stats.stripe_cache.errors));
+ gf_proc_dump_write("heals-attempted", "%" GF_PRI_ATOMIC,
+ GF_ATOMIC_GET(ec->stats.shd.attempted));
+ gf_proc_dump_write("heals-completed", "%" GF_PRI_ATOMIC,
+ GF_ATOMIC_GET(ec->stats.shd.completed));
return 0;
}
-struct xlator_fops fops =
-{
- .lookup = ec_gf_lookup,
- .stat = ec_gf_stat,
- .fstat = ec_gf_fstat,
- .truncate = ec_gf_truncate,
- .ftruncate = ec_gf_ftruncate,
- .access = ec_gf_access,
- .readlink = ec_gf_readlink,
- .mknod = ec_gf_mknod,
- .mkdir = ec_gf_mkdir,
- .unlink = ec_gf_unlink,
- .rmdir = ec_gf_rmdir,
- .symlink = ec_gf_symlink,
- .rename = ec_gf_rename,
- .link = ec_gf_link,
- .create = ec_gf_create,
- .open = ec_gf_open,
- .readv = ec_gf_readv,
- .writev = ec_gf_writev,
- .flush = ec_gf_flush,
- .fsync = ec_gf_fsync,
- .opendir = ec_gf_opendir,
- .readdir = ec_gf_readdir,
- .readdirp = ec_gf_readdirp,
- .fsyncdir = ec_gf_fsyncdir,
- .statfs = ec_gf_statfs,
- .setxattr = ec_gf_setxattr,
- .getxattr = ec_gf_getxattr,
- .fsetxattr = ec_gf_fsetxattr,
- .fgetxattr = ec_gf_fgetxattr,
- .removexattr = ec_gf_removexattr,
- .fremovexattr = ec_gf_fremovexattr,
- .lk = ec_gf_lk,
- .inodelk = ec_gf_inodelk,
- .finodelk = ec_gf_finodelk,
- .entrylk = ec_gf_entrylk,
- .fentrylk = ec_gf_fentrylk,
- .xattrop = ec_gf_xattrop,
- .fxattrop = ec_gf_fxattrop,
- .setattr = ec_gf_setattr,
- .fsetattr = ec_gf_fsetattr,
- .fallocate = ec_gf_fallocate,
- .discard = ec_gf_discard,
- .zerofill = ec_gf_zerofill,
- .seek = ec_gf_seek,
- .ipc = ec_gf_ipc
-};
-
-struct xlator_cbks cbks =
-{
- .forget = ec_gf_forget,
- .release = ec_gf_release,
- .releasedir = ec_gf_releasedir
-};
-
-struct xlator_dumpops dumpops = {
- .priv = ec_dump_private
-};
-
-struct volume_options options[] =
-{
+struct xlator_fops fops = {.lookup = ec_gf_lookup,
+ .stat = ec_gf_stat,
+ .fstat = ec_gf_fstat,
+ .truncate = ec_gf_truncate,
+ .ftruncate = ec_gf_ftruncate,
+ .access = ec_gf_access,
+ .readlink = ec_gf_readlink,
+ .mknod = ec_gf_mknod,
+ .mkdir = ec_gf_mkdir,
+ .unlink = ec_gf_unlink,
+ .rmdir = ec_gf_rmdir,
+ .symlink = ec_gf_symlink,
+ .rename = ec_gf_rename,
+ .link = ec_gf_link,
+ .create = ec_gf_create,
+ .open = ec_gf_open,
+ .readv = ec_gf_readv,
+ .writev = ec_gf_writev,
+ .flush = ec_gf_flush,
+ .fsync = ec_gf_fsync,
+ .opendir = ec_gf_opendir,
+ .readdir = ec_gf_readdir,
+ .readdirp = ec_gf_readdirp,
+ .fsyncdir = ec_gf_fsyncdir,
+ .statfs = ec_gf_statfs,
+ .setxattr = ec_gf_setxattr,
+ .getxattr = ec_gf_getxattr,
+ .fsetxattr = ec_gf_fsetxattr,
+ .fgetxattr = ec_gf_fgetxattr,
+ .removexattr = ec_gf_removexattr,
+ .fremovexattr = ec_gf_fremovexattr,
+ .lk = ec_gf_lk,
+ .inodelk = ec_gf_inodelk,
+ .finodelk = ec_gf_finodelk,
+ .entrylk = ec_gf_entrylk,
+ .fentrylk = ec_gf_fentrylk,
+ .xattrop = ec_gf_xattrop,
+ .fxattrop = ec_gf_fxattrop,
+ .setattr = ec_gf_setattr,
+ .fsetattr = ec_gf_fsetattr,
+ .fallocate = ec_gf_fallocate,
+ .discard = ec_gf_discard,
+ .zerofill = ec_gf_zerofill,
+ .seek = ec_gf_seek,
+ .ipc = ec_gf_ipc};
+
+struct xlator_cbks cbks = {.forget = ec_gf_forget,
+ .release = ec_gf_release,
+ .releasedir = ec_gf_releasedir};
+
+struct xlator_dumpops dumpops = {.priv = ec_dump_private};
+
+struct volume_options options[] = {
+ {.key = {"redundancy"},
+ .type = GF_OPTION_TYPE_INT,
+ .default_value = "{{ volume.redundancy }}",
+ .description = "Maximum number of bricks that can fail "
+ "simultaneously without losing data."},
{
- .key = { "redundancy" },
- .type = GF_OPTION_TYPE_INT,
- .default_value = "{{ volume.redundancy }}",
- .description = "Maximum number of bricks that can fail "
- "simultaneously without losing data."
- },
- {
- .key = { "self-heal-daemon" },
+ .key = {"self-heal-daemon"},
.type = GF_OPTION_TYPE_BOOL,
.description = "self-heal daemon enable/disable",
.default_value = "enable",
@@ -1486,193 +1656,218 @@ struct volume_options options[] =
.flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
.tags = {"disperse"},
},
- { .key = {"iam-self-heal-daemon"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "This option differentiates if the disperse "
- "translator is running as part of self-heal-daemon "
- "or not."
- },
- { .key = {"eager-lock"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = {GD_OP_VERSION_3_7_10},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
- .tags = {"disperse"},
- .description = "Enable/Disable eager lock for regular files on a "
- "disperse volume. If a fop takes a lock and completes "
- "its operation, it waits for next 1 second before "
- "releasing the lock, to see if the lock can be reused "
- "for next fop from the same client. If ec finds any lock "
- "contention within 1 second it releases the lock "
- "immediately before time expires. This improves the "
- "performance of file operations. However, as it takes "
- "lock on first brick, for few operations like read, "
- "discovery of lock contention might take long time and "
- "can actually degrade the performance. If eager lock is "
- "disabled, lock will be released as soon as fop "
- "completes."
- },
- { .key = {"other-eager-lock"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = { GD_OP_VERSION_3_13_0 },
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
- .tags = { "disperse" },
- .description = "It's equivalent to the eager-lock option but for non "
- "regular files."
- },
- { .key = {"eager-lock-timeout"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 60,
- .default_value = "1",
- .op_version = { GD_OP_VERSION_4_0_0 },
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
- .tags = { "disperse", "locks", "timeout" },
- .description = "Maximum time (in seconds) that a lock on an inode is "
- "kept held if no new operations on the inode are "
- "received."
- },
- { .key = {"other-eager-lock-timeout"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 60,
- .default_value = "1",
- .op_version = { GD_OP_VERSION_4_0_0 },
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
- .tags = { "disperse", "locks", "timeout" },
- .description = "It's equivalent ot eager-lock-timeout option but for "
- "non regular files."
- },
- { .key = {"background-heals"},
- .type = GF_OPTION_TYPE_INT,
- .min = 0,/*Disabling background heals*/
- .max = 256,
- .default_value = "8",
- .op_version = {GD_OP_VERSION_3_7_3},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
- .tags = {"disperse"},
- .description = "This option can be used to control number of parallel"
- " heals",
- },
- { .key = {"heal-wait-qlength"},
- .type = GF_OPTION_TYPE_INT,
- .min = 0,
- .max = 65536, /*Around 100MB as of now with sizeof(ec_fop_data_t) at 1800*/
- .default_value = "128",
- .op_version = {GD_OP_VERSION_3_7_3},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
- .tags = {"disperse"},
- .description = "This option can be used to control number of heals"
- " that can wait",
- },
- { .key = {"heal-timeout"},
- .type = GF_OPTION_TYPE_INT,
- .min = 60,
- .max = INT_MAX,
- .default_value = "600",
- .op_version = {GD_OP_VERSION_3_7_3},
- .flags = OPT_FLAG_SETTABLE,
- .tags = {"disperse"},
- .description = "time interval for checking the need to self-heal "
- "in self-heal-daemon"
- },
- { .key = {"read-policy" },
- .type = GF_OPTION_TYPE_STR,
- .value = {"round-robin", "gfid-hash"},
- .default_value = "gfid-hash",
- .op_version = {GD_OP_VERSION_3_7_6},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
- .tags = {"disperse"},
- .description = "inode-read fops happen only on 'k' number of bricks in"
- " n=k+m disperse subvolume. 'round-robin' selects the read"
- " subvolume using round-robin algo. 'gfid-hash' selects read"
- " subvolume based on hash of the gfid of that file/directory.",
- },
- { .key = {"shd-max-threads"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 64,
- .default_value = "1",
- .op_version = {GD_OP_VERSION_3_9_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"disperse"},
- .description = "Maximum number of parallel heals SHD can do per local "
- "brick. This can substantially lower heal times, "
- "but can also crush your bricks if you don't have "
- "the storage hardware to support this."
+ {.key = {"iam-self-heal-daemon"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "This option differentiates if the disperse "
+ "translator is running as part of self-heal-daemon "
+ "or not."},
+ {.key = {"eager-lock"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .op_version = {GD_OP_VERSION_3_7_10},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
+ .tags = {"disperse"},
+ .description = "Enable/Disable eager lock for regular files on a "
+ "disperse volume. If a fop takes a lock and completes "
+ "its operation, it waits for next 1 second before "
+ "releasing the lock, to see if the lock can be reused "
+ "for next fop from the same client. If ec finds any lock "
+ "contention within 1 second it releases the lock "
+ "immediately before time expires. This improves the "
+ "performance of file operations. However, as it takes "
+ "lock on first brick, for few operations like read, "
+ "discovery of lock contention might take long time and "
+ "can actually degrade the performance. If eager lock is "
+ "disabled, lock will be released as soon as fop "
+ "completes."},
+ {.key = {"other-eager-lock"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .op_version = {GD_OP_VERSION_3_13_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
+ .tags = {"disperse"},
+ .description = "It's equivalent to the eager-lock option but for non "
+ "regular files."},
+ {.key = {"eager-lock-timeout"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 60,
+ .default_value = "1",
+ .op_version = {GD_OP_VERSION_4_0_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
+ .tags = {"disperse", "locks", "timeout"},
+ .description = "Maximum time (in seconds) that a lock on an inode is "
+ "kept held if no new operations on the inode are "
+ "received."},
+ {.key = {"other-eager-lock-timeout"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 60,
+ .default_value = "1",
+ .op_version = {GD_OP_VERSION_4_0_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
+ .tags = {"disperse", "locks", "timeout"},
+ .description = "It's equivalent to eager-lock-timeout option but for "
+ "non regular files."},
+ {
+ .key = {"background-heals"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0, /*Disabling background heals*/
+ .max = 256,
+ .default_value = "8",
+ .op_version = {GD_OP_VERSION_3_7_3},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
+ .tags = {"disperse"},
+ .description = "This option can be used to control number of parallel"
+ " heals",
},
- { .key = {"shd-wait-qlength"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 65536,
- .default_value = "1024",
- .op_version = {GD_OP_VERSION_3_9_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"disperse"},
- .description = "This option can be used to control number of heals"
- " that can wait in SHD per subvolume"
+ {
+ .key = {"heal-wait-qlength"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max =
+ 65536, /*Around 100MB as of now with sizeof(ec_fop_data_t) at 1800*/
+ .default_value = "128",
+ .op_version = {GD_OP_VERSION_3_7_3},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
+ .tags = {"disperse"},
+ .description = "This option can be used to control number of heals"
+ " that can wait",
},
+ {.key = {"heal-timeout"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 60,
+ .max = INT_MAX,
+ .default_value = "600",
+ .op_version = {GD_OP_VERSION_3_7_3},
+ .flags = OPT_FLAG_SETTABLE,
+ .tags = {"disperse"},
+ .description = "time interval for checking the need to self-heal "
+ "in self-heal-daemon"},
{
- .key = { "cpu-extensions" },
+ .key = {"read-policy"},
.type = GF_OPTION_TYPE_STR,
- .value = { "none", "auto", "x64", "sse", "avx" },
- .default_value = "auto",
- .op_version = {GD_OP_VERSION_3_9_0},
+ .value = {"round-robin", "gfid-hash"},
+ .default_value = "gfid-hash",
+ .op_version = {GD_OP_VERSION_3_7_6},
.flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
.tags = {"disperse"},
- .description = "force the cpu extensions to be used to accelerate the "
- "galois field computations."
- },
- { .key = {"self-heal-window-size"},
- .type = GF_OPTION_TYPE_INT,
- .min = 1,
- .max = 1024,
- .default_value = "1",
- .op_version = {GD_OP_VERSION_3_11_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
- .tags = {"disperse"},
- .description = "Maximum number blocks(128KB) per file for which "
- "self-heal process would be applied simultaneously."
+ .description =
+ "inode-read fops happen only on 'k' number of bricks in"
+ " n=k+m disperse subvolume. 'round-robin' selects the read"
+ " subvolume using round-robin algo. 'gfid-hash' selects read"
+ " subvolume based on hash of the gfid of that file/directory.",
},
- { .key = {"optimistic-change-log"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .op_version = {GD_OP_VERSION_3_10_1},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT,
- .tags = {"disperse"},
- .description = "Set/Unset dirty flag for every update fop at the start"
- "of the fop. If OFF, this option impacts performance of"
- "entry operations or metadata operations as it will"
- "set dirty flag at the start and unset it at the end of"
- "ALL update fop. If ON and all the bricks are good,"
- "dirty flag will be set at the start only for file fops"
- "For metadata and entry fops dirty flag will not be set"
- "at the start, if all the bricks are good. This does"
- "not impact performance for metadata operations and"
- "entry operation but has a very small window to miss"
- "marking entry as dirty in case it is required to be"
- "healed"
+ {.key = {"shd-max-threads"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 64,
+ .default_value = "1",
+ .op_version = {GD_OP_VERSION_3_9_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"disperse"},
+ .description = "Maximum number of parallel heals SHD can do per local "
+ "brick. This can substantially lower heal times, "
+ "but can also crush your bricks if you don't have "
+ "the storage hardware to support this."},
+ {.key = {"shd-wait-qlength"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 65536,
+ .default_value = "1024",
+ .op_version = {GD_OP_VERSION_3_9_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"disperse"},
+ .description = "This option can be used to control number of heals"
+ " that can wait in SHD per subvolume"},
+ {.key = {"cpu-extensions"},
+ .type = GF_OPTION_TYPE_STR,
+ .value = {"none", "auto", "x64", "sse", "avx"},
+ .default_value = "auto",
+ .op_version = {GD_OP_VERSION_3_9_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
+ .tags = {"disperse"},
+ .description = "force the cpu extensions to be used to accelerate the "
+ "galois field computations."},
+ {.key = {"self-heal-window-size"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 1,
+ .max = 1024,
+ .default_value = "1",
+ .op_version = {GD_OP_VERSION_3_11_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
+ .tags = {"disperse"},
+ .description = "Maximum number blocks(128KB) per file for which "
+ "self-heal process would be applied simultaneously."},
+ {.key = {"optimistic-change-log"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .op_version = {GD_OP_VERSION_3_10_1},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT,
+ .tags = {"disperse"},
+ .description = "Set/Unset dirty flag for every update fop at the start"
+ "of the fop. If OFF, this option impacts performance of"
+ "entry operations or metadata operations as it will"
+ "set dirty flag at the start and unset it at the end of"
+ "ALL update fop. If ON and all the bricks are good,"
+ "dirty flag will be set at the start only for file fops"
+ "For metadata and entry fops dirty flag will not be set"
+ "at the start, if all the bricks are good. This does"
+ "not impact performance for metadata operations and"
+ "entry operation but has a very small window to miss"
+ "marking entry as dirty in case it is required to be"
+ "healed"},
+ {.key = {"parallel-writes"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "This controls if writes can be wound in parallel as long"
+ "as it doesn't modify same stripes"},
+ {.key = {"stripe-cache"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0, /*Disabling stripe_cache*/
+ .max = EC_STRIPE_CACHE_MAX_SIZE,
+ .default_value = "4",
+ .description = "This option will keep the last stripe of write fop"
+ "in memory. If next write falls in this stripe, we need"
+ "not to read it again from backend and we can save READ"
+ "fop going over the network. This will improve performance,"
+ "specially for sequential writes. However, this will also"
+ "lead to extra memory consumption, maximum "
+ "(cache size * stripe size) Bytes per open file."},
+ {
+ .key = {"quorum-count"},
+ .type = GF_OPTION_TYPE_INT,
+ .default_value = "0",
+ .description =
+ "This option can be used to define how many successes on"
+ "the bricks constitute a success to the application. This"
+ " count should be in the range"
+ "[disperse-data-count, disperse-count] (inclusive)",
},
- { .key = {"parallel-writes"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .description = "This controls if writes can be wound in parallel as long"
- "as it doesn't modify same stripes"
+ {
+ .key = {"ec-read-mask"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = NULL,
+ .description = "This option can be used to choose which bricks can be"
+ " used for reading data/metadata of a file/directory",
},
- { .key = {"stripe-cache"},
- .type = GF_OPTION_TYPE_INT,
- .min = 0,/*Disabling stripe_cache*/
- .max = EC_STRIPE_CACHE_MAX_SIZE,
- .default_value = "4",
- .description = "This option will keep the last stripe of write fop"
- "in memory. If next write falls in this stripe, we need"
- "not to read it again from backend and we can save READ"
- "fop going over the network. This will improve performance,"
- "specially for sequential writes. However, this will also"
- "lead to extra memory consumption, maximum "
- "(cache size * stripe size) Bytes per open file."
+ {
+ .key = {NULL},
},
- { .key = {NULL} }
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .notify = notify,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1},
+ .dumpops = &dumpops,
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "disperse",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/cluster/ec/src/ec.h b/xlators/cluster/ec/src/ec.h
index e2ec9cf04f5..6f6de6d5981 100644
--- a/xlators/cluster/ec/src/ec.h
+++ b/xlators/cluster/ec/src/ec.h
@@ -13,15 +13,16 @@
#include "ec-method.h"
-#define EC_XATTR_PREFIX "trusted.ec."
-#define EC_XATTR_CONFIG EC_XATTR_PREFIX"config"
-#define EC_XATTR_SIZE EC_XATTR_PREFIX"size"
-#define EC_XATTR_VERSION EC_XATTR_PREFIX"version"
-#define EC_XATTR_HEAL EC_XATTR_PREFIX"heal"
-#define EC_XATTR_DIRTY EC_XATTR_PREFIX"dirty"
-#define EC_STRIPE_CACHE_MAX_SIZE 10
+#define EC_XATTR_PREFIX "trusted.ec."
+#define EC_XATTR_CONFIG EC_XATTR_PREFIX "config"
+#define EC_XATTR_SIZE EC_XATTR_PREFIX "size"
+#define EC_XATTR_VERSION EC_XATTR_PREFIX "version"
+#define EC_XATTR_HEAL EC_XATTR_PREFIX "heal"
+#define EC_XATTR_HEAL_NEW EC_XATTR_PREFIX "heal-new"
+#define EC_XATTR_DIRTY EC_XATTR_PREFIX "dirty"
+#define EC_STRIPE_CACHE_MAX_SIZE 10
#define EC_VERSION_SIZE 2
-#define EC_SHD_INODE_LRU_LIMIT 10
+#define EC_SHD_INODE_LRU_LIMIT 10
#define EC_MAX_FRAGMENTS EC_METHOD_MAX_FRAGMENTS
/* The maximum number of nodes is derived from the maximum allowed fragments
diff --git a/xlators/cluster/stripe/src/Makefile.am b/xlators/cluster/stripe/src/Makefile.am
deleted file mode 100644
index 2b594567db1..00000000000
--- a/xlators/cluster/stripe/src/Makefile.am
+++ /dev/null
@@ -1,22 +0,0 @@
-xlator_LTLIBRARIES = stripe.la
-xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/cluster
-
-stripe_la_LDFLAGS = -module $(GF_XLATOR_DEFAULT_LDFLAGS)
-
-
-stripe_la_SOURCES = stripe.c stripe-helpers.c \
- $(top_builddir)/xlators/lib/src/libxlator.c
-
-stripe_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-
-noinst_HEADERS = stripe.h stripe-mem-types.h \
- $(top_builddir)/xlators/lib/src/libxlator.h
-
-AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
- -I$(top_srcdir)/xlators/lib/src \
- -I$(top_srcdir)/rpc/xdr/src -I$(top_builddir)/rpc/xdr/src
-
-AM_CFLAGS = -Wall $(GF_CFLAGS)
-
-CLEANFILES =
-
diff --git a/xlators/cluster/stripe/src/stripe-helpers.c b/xlators/cluster/stripe/src/stripe-helpers.c
deleted file mode 100644
index 217f4d2b6e8..00000000000
--- a/xlators/cluster/stripe/src/stripe-helpers.c
+++ /dev/null
@@ -1,677 +0,0 @@
-/*
- Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#include <fnmatch.h>
-
-#include "stripe.h"
-#include "byte-order.h"
-#include "mem-types.h"
-#include "logging.h"
-
-void
-stripe_local_wipe (stripe_local_t *local)
-{
- if (!local)
- goto out;
-
- loc_wipe (&local->loc);
- loc_wipe (&local->loc2);
-
- if (local->fd)
- fd_unref (local->fd);
-
- if (local->inode)
- inode_unref (local->inode);
-
- if (local->xattr)
- dict_unref (local->xattr);
-
- if (local->xdata)
- dict_unref (local->xdata);
-
-out:
- return;
-}
-
-
-
-int
-stripe_aggregate (dict_t *this, char *key, data_t *value, void *data)
-{
- dict_t *dst = NULL;
- int64_t *ptr = 0, *size = NULL;
- int32_t ret = -1;
-
- dst = data;
-
- if (strcmp (key, QUOTA_SIZE_KEY) == 0) {
- ret = dict_get_bin (dst, key, (void **)&size);
- if (ret < 0) {
- size = GF_CALLOC (1, sizeof (int64_t),
- gf_common_mt_char);
- if (size == NULL) {
- gf_log ("stripe", GF_LOG_WARNING,
- "memory allocation failed");
- goto out;
- }
- ret = dict_set_bin (dst, key, size, sizeof (int64_t));
- if (ret < 0) {
- gf_log ("stripe", GF_LOG_WARNING,
- "stripe aggregate dict set failed");
- GF_FREE (size);
- goto out;
- }
- }
-
- ptr = data_to_bin (value);
- if (ptr == NULL) {
- gf_log ("stripe", GF_LOG_WARNING, "data to bin failed");
- goto out;
- }
-
- *size = hton64 (ntoh64 (*size) + ntoh64 (*ptr));
- } else if (strcmp (key, GF_CONTENT_KEY)) {
- /* No need to aggregate 'CONTENT' data */
- ret = dict_set (dst, key, value);
- if (ret)
- gf_log ("stripe", GF_LOG_WARNING, "xattr dict set failed");
- }
-
-out:
- return 0;
-}
-
-
-void
-stripe_aggregate_xattr (dict_t *dst, dict_t *src)
-{
- if ((dst == NULL) || (src == NULL)) {
- goto out;
- }
-
- dict_foreach (src, stripe_aggregate, dst);
-out:
- return;
-}
-
-
-int32_t
-stripe_xattr_aggregate (char *buffer, stripe_local_t *local, int32_t *total)
-{
- int32_t i = 0;
- int32_t ret = -1;
- int32_t len = 0;
- char *sbuf = NULL;
- stripe_xattr_sort_t *xattr = NULL;
-
- if (!buffer || !local || !local->xattr_list)
- goto out;
-
- sbuf = buffer;
-
- for (i = 0; i < local->nallocs; i++) {
- xattr = local->xattr_list + i;
- len = xattr->xattr_len - 1; /* length includes \0 */
-
- if (len && xattr && xattr->xattr_value) {
- memcpy (buffer, xattr->xattr_value, len);
- buffer += len;
- *buffer++ = ' ';
- }
- }
-
- *--buffer = '\0';
- if (total)
- *total = buffer - sbuf;
- ret = 0;
-
- out:
- return ret;
-}
-
-int32_t
-stripe_free_xattr_str (stripe_local_t *local)
-{
- int32_t i = 0;
- int32_t ret = -1;
- stripe_xattr_sort_t *xattr = NULL;
-
- if (!local || !local->xattr_list)
- goto out;
-
- for (i = 0; i < local->nallocs; i++) {
- xattr = local->xattr_list + i;
-
- if (xattr && xattr->xattr_value)
- GF_FREE (xattr->xattr_value);
- }
-
- ret = 0;
- out:
- return ret;
-}
-
-
-int32_t
-stripe_fill_lockinfo_xattr (xlator_t *this, stripe_local_t *local,
- void **xattr_serz)
-{
- int32_t ret = -1, i = 0, len = 0;
- dict_t *tmp1 = NULL, *tmp2 = NULL;
- char *buf = NULL;
- stripe_xattr_sort_t *xattr = NULL;
-
- if (xattr_serz == NULL) {
- goto out;
- }
-
- tmp2 = dict_new ();
-
- if (tmp2 == NULL) {
- goto out;
- }
-
- for (i = 0; i < local->nallocs; i++) {
- xattr = local->xattr_list + i;
- len = xattr->xattr_len;
-
- if (len && xattr && xattr->xattr_value) {
- ret = dict_reset (tmp2);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "dict_reset failed (%s)",
- strerror (-ret));
- }
-
- ret = dict_unserialize (xattr->xattr_value,
- xattr->xattr_len,
- &tmp2);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "dict_unserialize failed (%s)",
- strerror (-ret));
- ret = -1;
- goto out;
- }
-
- tmp1 = dict_copy (tmp2, tmp1);
- if (tmp1 == NULL) {
- gf_log (this->name, GF_LOG_WARNING,
- "dict_copy failed (%s)",
- strerror (-ret));
- ret = -1;
- goto out;
- }
- }
- }
-
- len = dict_serialized_length (tmp1);
- if (len > 0) {
- buf = GF_CALLOC (1, len, gf_common_mt_dict_t);
- if (buf == NULL) {
- ret = -1;
- goto out;
- }
-
- ret = dict_serialize (tmp1, buf);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "dict_serialize failed (%s)", strerror (-ret));
- GF_FREE(buf);
- ret = -1;
- goto out;
- }
-
- *xattr_serz = buf;
- }
-
- ret = 0;
-out:
- if (tmp1 != NULL) {
- dict_unref (tmp1);
- }
-
- if (tmp2 != NULL) {
- dict_unref (tmp2);
- }
-
- return ret;
-}
-
-
-int32_t
-stripe_fill_pathinfo_xattr (xlator_t *this, stripe_local_t *local,
- char **xattr_serz)
-{
- int ret = -1;
- int32_t padding = 0;
- int32_t tlen = 0;
- char stripe_size_str[20] = {0,};
- char *pathinfo_serz = NULL;
-
- if (!local) {
- gf_log (this->name, GF_LOG_ERROR, "Possible NULL deref");
- goto out;
- }
-
- (void) snprintf (stripe_size_str, 20, "%"PRId64,
- (long long) (local->fctx) ? local->fctx->stripe_size : 0);
-
- /* extra bytes for decorations (brackets and <>'s) */
- padding = strlen (this->name) + strlen (STRIPE_PATHINFO_HEADER)
- + strlen (stripe_size_str) + 7;
- local->xattr_total_len += (padding + 2);
-
- pathinfo_serz = GF_CALLOC (local->xattr_total_len, sizeof (char),
- gf_common_mt_char);
- if (!pathinfo_serz)
- goto out;
-
- /* xlator info */
- (void) sprintf (pathinfo_serz, "(<"STRIPE_PATHINFO_HEADER"%s:[%s]> ",
- this->name, stripe_size_str);
-
- ret = stripe_xattr_aggregate (pathinfo_serz + padding, local, &tlen);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "Cannot aggregate pathinfo list");
- GF_FREE(pathinfo_serz);
- goto out;
- }
-
- *(pathinfo_serz + padding + tlen) = ')';
- *(pathinfo_serz + padding + tlen + 1) = '\0';
-
- *xattr_serz = pathinfo_serz;
-
- ret = 0;
- out:
- return ret;
-}
-
-/**
- * stripe_get_matching_bs - Get the matching block size for the given path.
- */
-int32_t
-stripe_get_matching_bs (const char *path, stripe_private_t *priv)
-{
- struct stripe_options *trav = NULL;
- uint64_t block_size = 0;
-
- GF_VALIDATE_OR_GOTO ("stripe", priv, out);
- GF_VALIDATE_OR_GOTO ("stripe", path, out);
-
- LOCK (&priv->lock);
- {
- block_size = priv->block_size;
- trav = priv->pattern;
- while (trav) {
- if (!fnmatch (trav->path_pattern, path, FNM_NOESCAPE)) {
- block_size = trav->block_size;
- break;
- }
- trav = trav->next;
- }
- }
- UNLOCK (&priv->lock);
-
-out:
- return block_size;
-}
-
-int32_t
-stripe_ctx_handle (xlator_t *this, call_frame_t *prev, stripe_local_t *local,
- dict_t *dict)
-{
- char key[256] = {0,};
- data_t *data = NULL;
- int32_t index = 0;
- stripe_private_t *priv = NULL;
-
- priv = this->private;
-
-
- if (!local->fctx) {
- local->fctx = GF_CALLOC (1, sizeof (stripe_fd_ctx_t),
- gf_stripe_mt_stripe_fd_ctx_t);
- if (!local->fctx) {
- local->op_errno = ENOMEM;
- local->op_ret = -1;
- goto out;
- }
-
- local->fctx->static_array = 0;
- }
- /* Stripe block size */
- sprintf (key, "trusted.%s.stripe-size", this->name);
- data = dict_get (dict, key);
- if (!data) {
- local->xattr_self_heal_needed = 1;
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to get stripe-size");
- goto out;
- } else {
- if (!local->fctx->stripe_size) {
- local->fctx->stripe_size =
- data_to_int64 (data);
- }
-
- if (local->fctx->stripe_size != data_to_int64 (data)) {
- gf_log (this->name, GF_LOG_WARNING,
- "stripe-size mismatch in blocks");
- local->xattr_self_heal_needed = 1;
- }
- }
-
- /* Stripe count */
- sprintf (key, "trusted.%s.stripe-count", this->name);
- data = dict_get (dict, key);
-
- if (!data) {
- local->xattr_self_heal_needed = 1;
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to get stripe-count");
- goto out;
- }
- if (!local->fctx->xl_array) {
- local->fctx->stripe_count = data_to_int32 (data);
- if (!local->fctx->stripe_count) {
- gf_log (this->name, GF_LOG_ERROR,
- "error with stripe-count xattr");
- local->op_ret = -1;
- local->op_errno = EIO;
- goto out;
- }
-
- local->fctx->xl_array = GF_CALLOC (local->fctx->stripe_count,
- sizeof (xlator_t *),
- gf_stripe_mt_xlator_t);
-
- if (!local->fctx->xl_array) {
- local->op_errno = ENOMEM;
- local->op_ret = -1;
- goto out;
- }
- }
- if (local->fctx->stripe_count != data_to_int32 (data)) {
- gf_log (this->name, GF_LOG_ERROR,
- "error with stripe-count xattr (%d != %d)",
- local->fctx->stripe_count, data_to_int32 (data));
- local->op_ret = -1;
- local->op_errno = EIO;
- goto out;
- }
-
- /* index */
- sprintf (key, "trusted.%s.stripe-index", this->name);
- data = dict_get (dict, key);
- if (!data) {
- local->xattr_self_heal_needed = 1;
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to get stripe-index");
- goto out;
- }
- index = data_to_int32 (data);
- if (index > priv->child_count) {
- gf_log (this->name, GF_LOG_ERROR,
- "error with stripe-index xattr (%d)", index);
- local->op_ret = -1;
- local->op_errno = EIO;
- goto out;
- }
- if (local->fctx->xl_array) {
- if (!local->fctx->xl_array[index])
- local->fctx->xl_array[index] = prev->this;
- }
-
- sprintf(key, "trusted.%s.stripe-coalesce", this->name);
- data = dict_get(dict, key);
- if (!data) {
- /*
- * The file was probably created prior to coalesce support.
- * Assume non-coalesce mode for this file to maintain backwards
- * compatibility.
- */
- gf_log(this->name, GF_LOG_DEBUG, "missing stripe-coalesce "
- "attr, assume non-coalesce mode");
- local->fctx->stripe_coalesce = 0;
- } else {
- local->fctx->stripe_coalesce = data_to_int32(data);
- }
-
-
-out:
- return 0;
-}
-
-int32_t
-stripe_xattr_request_build (xlator_t *this, dict_t *dict, uint64_t stripe_size,
- uint32_t stripe_count, uint32_t stripe_index,
- uint32_t stripe_coalesce)
-{
- char key[256] = {0,};
- int32_t ret = -1;
-
- sprintf (key, "trusted.%s.stripe-size", this->name);
- ret = dict_set_int64 (dict, key, stripe_size);
- if (ret) {
- gf_log (this->name, GF_LOG_WARNING,
- "failed to set %s in xattr_req dict", key);
- goto out;
- }
-
- sprintf (key, "trusted.%s.stripe-count", this->name);
- ret = dict_set_int32 (dict, key, stripe_count);
- if (ret) {
- gf_log (this->name, GF_LOG_WARNING,
- "failed to set %s in xattr_req dict", key);
- goto out;
- }
-
- sprintf (key, "trusted.%s.stripe-index", this->name);
- ret = dict_set_int32 (dict, key, stripe_index);
- if (ret) {
- gf_log (this->name, GF_LOG_WARNING,
- "failed to set %s in xattr_req dict", key);
- goto out;
- }
-
- sprintf(key, "trusted.%s.stripe-coalesce", this->name);
- ret = dict_set_int32(dict, key, stripe_coalesce);
- if (ret) {
- gf_log(this->name, GF_LOG_WARNING,
- "failed to set %s in xattr_req_dict", key);
- goto out;
- }
-out:
- return ret;
-}
-
-
-static int
-set_default_block_size (stripe_private_t *priv, char *num)
-{
-
- int ret = -1;
- GF_VALIDATE_OR_GOTO ("stripe", THIS, out);
- GF_VALIDATE_OR_GOTO (THIS->name, priv, out);
- GF_VALIDATE_OR_GOTO (THIS->name, num, out);
-
-
- if (gf_string2bytesize_uint64 (num, &priv->block_size) != 0) {
- gf_log (THIS->name, GF_LOG_ERROR,
- "invalid number format \"%s\"", num);
- goto out;
- }
-
- ret = 0;
-
- out:
- return ret;
-
-}
-
-
-int
-set_stripe_block_size (xlator_t *this, stripe_private_t *priv, char *data)
-{
- int ret = -1;
- char *tmp_str = NULL;
- char *tmp_str1 = NULL;
- char *dup_str = NULL;
- char *stripe_str = NULL;
- char *pattern = NULL;
- char *num = NULL;
- struct stripe_options *temp_stripeopt = NULL;
- struct stripe_options *stripe_opt = NULL;
-
- if (!this || !priv || !data)
- goto out;
-
- /* Get the pattern for striping.
- "option block-size *avi:10MB" etc */
- stripe_str = strtok_r (data, ",", &tmp_str);
- while (stripe_str) {
- dup_str = gf_strdup (stripe_str);
- stripe_opt = GF_CALLOC (1, sizeof (struct stripe_options),
- gf_stripe_mt_stripe_options);
- if (!stripe_opt) {
- goto out;
- }
-
- pattern = strtok_r (dup_str, ":", &tmp_str1);
- num = strtok_r (NULL, ":", &tmp_str1);
- if (!num) {
- num = pattern;
- pattern = "*";
- ret = set_default_block_size (priv, num);
- if (ret)
- goto out;
- }
- if (gf_string2bytesize_uint64 (num, &stripe_opt->block_size) != 0) {
- gf_log (this->name, GF_LOG_ERROR,
- "invalid number format \"%s\"", num);
- goto out;
- }
-
- if (stripe_opt->block_size < STRIPE_MIN_BLOCK_SIZE) {
- gf_log (this->name, GF_LOG_ERROR, "Invalid Block-size: "
- "%s. Should be atleast %llu bytes", num,
- STRIPE_MIN_BLOCK_SIZE);
- goto out;
- }
- if (stripe_opt->block_size % 512) {
- gf_log (this->name, GF_LOG_ERROR, "Block-size: %s should"
- " be a multiple of 512 bytes", num);
- goto out;
- }
-
- memcpy (stripe_opt->path_pattern, pattern, strlen (pattern));
-
- gf_log (this->name, GF_LOG_DEBUG,
- "block-size : pattern %s : size %"PRId64,
- stripe_opt->path_pattern, stripe_opt->block_size);
-
- if (priv->pattern)
- temp_stripeopt = NULL;
- else
- temp_stripeopt = priv->pattern;
-
- stripe_opt->next = temp_stripeopt;
-
- priv->pattern = stripe_opt;
- stripe_opt = NULL;
-
- GF_FREE (dup_str);
- dup_str = NULL;
-
- stripe_str = strtok_r (NULL, ",", &tmp_str);
- }
-
- ret = 0;
-out:
-
- GF_FREE (dup_str);
-
- GF_FREE (stripe_opt);
-
- return ret;
-}
-
-int32_t
-stripe_iatt_merge (struct iatt *from, struct iatt *to)
-{
- if (to->ia_size < from->ia_size)
- to->ia_size = from->ia_size;
- if (to->ia_mtime < from->ia_mtime)
- to->ia_mtime = from->ia_mtime;
- if (to->ia_ctime < from->ia_ctime)
- to->ia_ctime = from->ia_ctime;
- if (to->ia_atime < from->ia_atime)
- to->ia_atime = from->ia_atime;
- return 0;
-}
-
-off_t
-coalesced_offset(off_t offset, uint64_t stripe_size, int stripe_count)
-{
- size_t line_size = 0;
- uint64_t stripe_num = 0;
- off_t coalesced_offset = 0;
-
- line_size = stripe_size * stripe_count;
- stripe_num = offset / line_size;
-
- coalesced_offset = (stripe_num * stripe_size) +
- (offset % stripe_size);
-
- return coalesced_offset;
-}
-
-off_t
-uncoalesced_size(off_t size, uint64_t stripe_size, int stripe_count,
- int stripe_index)
-{
- uint64_t nr_full_stripe_chunks = 0, mod = 0;
-
- if (!size)
- return size;
-
- /*
- * Estimate the number of fully written stripes from the
- * local file size. Each stripe_size chunk corresponds to
- * a stripe.
- */
- nr_full_stripe_chunks = (size / stripe_size) * stripe_count;
- mod = size % stripe_size;
-
- if (!mod) {
- /*
- * There is no remainder, thus we could have overestimated
- * the size of the file in terms of chunks. Trim the number
- * of chunks by the following stripe members and leave it
- * up to those nodes to respond with a larger size (if
- * necessary).
- */
- nr_full_stripe_chunks -= stripe_count -
- (stripe_index + 1);
- size = nr_full_stripe_chunks * stripe_size;
- } else {
- /*
- * There is a remainder and thus we own the last chunk of the
- * file. Add the preceding stripe members of the final stripe
- * along with the remainder to calculate the exact size.
- */
- nr_full_stripe_chunks += stripe_index;
- size = nr_full_stripe_chunks * stripe_size + mod;
- }
-
- return size;
-}
diff --git a/xlators/cluster/stripe/src/stripe-mem-types.h b/xlators/cluster/stripe/src/stripe-mem-types.h
deleted file mode 100644
index e9ac9cf4648..00000000000
--- a/xlators/cluster/stripe/src/stripe-mem-types.h
+++ /dev/null
@@ -1,31 +0,0 @@
-/*
- Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-
-#ifndef __STRIPE_MEM_TYPES_H__
-#define __STRIPE_MEM_TYPES_H__
-
-#include "mem-types.h"
-
-enum gf_stripe_mem_types_ {
- gf_stripe_mt_iovec = gf_common_mt_end + 1,
- gf_stripe_mt_stripe_replies,
- gf_stripe_mt_stripe_fd_ctx_t,
- gf_stripe_mt_char,
- gf_stripe_mt_int8_t,
- gf_stripe_mt_int32_t,
- gf_stripe_mt_xlator_t,
- gf_stripe_mt_stripe_private_t,
- gf_stripe_mt_stripe_options,
- gf_stripe_mt_xattr_sort_t,
- gf_stripe_mt_end
-};
-#endif
-
diff --git a/xlators/cluster/stripe/src/stripe.c b/xlators/cluster/stripe/src/stripe.c
deleted file mode 100644
index fae80c3e200..00000000000
--- a/xlators/cluster/stripe/src/stripe.c
+++ /dev/null
@@ -1,5782 +0,0 @@
-/*
- Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-/**
- * xlators/cluster/stripe:
- * Stripe translator, stripes the data across its child nodes,
- * as per the options given in the volfile. The striping works
- * fairly simple. It writes files at different offset as per
- * calculation. So, 'ls -l' output at the real posix level will
- * show file size bigger than the actual size. But when one does
- * 'df' or 'du <file>', real size of the file on the server is shown.
- *
- * WARNING:
- * Stripe translator can't regenerate data if a child node gets disconnected.
- * So, no 'self-heal' for stripe. Hence the advice, use stripe only when its
- * very much necessary, or else, use it in combination with AFR, to have a
- * backup copy.
- */
-#include <fnmatch.h>
-
-#include "stripe.h"
-#include "libxlator.h"
-#include "byte-order.h"
-#include "statedump.h"
-
-struct volume_options options[];
-
-int32_t
-stripe_sh_chown_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop, dict_t *xdata)
-{
- int callcnt = -1;
- stripe_local_t *local = NULL;
-
- if (!this || !frame || !frame->local) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- STRIPE_STACK_DESTROY (frame);
- }
-out:
- return 0;
-}
-
-int32_t
-stripe_sh_make_entry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
-
- if (!frame || !frame->local || !cookie || !this) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- STACK_WIND (frame, stripe_sh_chown_cbk, prev->this,
- prev->this->fops->setattr, &local->loc,
- &local->stbuf, (GF_SET_ATTR_UID | GF_SET_ATTR_GID), NULL);
-
-out:
- return 0;
-}
-
-int32_t
-stripe_entry_self_heal (call_frame_t *frame, xlator_t *this,
- stripe_local_t *local)
-{
- xlator_list_t *trav = NULL;
- call_frame_t *rframe = NULL;
- stripe_local_t *rlocal = NULL;
- stripe_private_t *priv = NULL;
- dict_t *xdata = NULL;
- int ret = 0;
-
- if (!local || !this || !frame) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- if (!(IA_ISREG (local->stbuf.ia_type) ||
- IA_ISDIR (local->stbuf.ia_type)))
- return 0;
-
- priv = this->private;
- trav = this->children;
- rframe = copy_frame (frame);
- if (!rframe) {
- goto out;
- }
- rlocal = mem_get0 (this->local_pool);
- if (!rlocal) {
- goto out;
- }
- rframe->local = rlocal;
- rlocal->call_count = priv->child_count;
- loc_copy (&rlocal->loc, &local->loc);
- memcpy (&rlocal->stbuf, &local->stbuf, sizeof (struct iatt));
-
- xdata = dict_new ();
- if (!xdata)
- goto out;
-
- ret = dict_set_gfuuid (xdata, "gfid-req", local->stbuf.ia_gfid, true);
- if (ret)
- gf_log (this->name, GF_LOG_WARNING,
- "%s: failed to set gfid-req", local->loc.path);
-
- while (trav) {
- if (IA_ISREG (local->stbuf.ia_type)) {
- STACK_WIND (rframe, stripe_sh_make_entry_cbk,
- trav->xlator, trav->xlator->fops->mknod,
- &local->loc,
- st_mode_from_ia (local->stbuf.ia_prot,
- local->stbuf.ia_type),
- 0, 0, xdata);
- }
- if (IA_ISDIR (local->stbuf.ia_type)) {
- STACK_WIND (rframe, stripe_sh_make_entry_cbk,
- trav->xlator, trav->xlator->fops->mkdir,
- &local->loc,
- st_mode_from_ia (local->stbuf.ia_prot,
- local->stbuf.ia_type),
- 0, xdata);
- }
- trav = trav->next;
- }
-
- if (xdata)
- dict_unref (xdata);
- return 0;
-
-out:
- if (rframe)
- STRIPE_STACK_DESTROY (rframe);
-
- return 0;
-}
-
-
-int32_t
-stripe_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata, struct iatt *postparent)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
- int ret = 0;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- if ((op_errno != ENOENT) && (op_errno != ESTALE))
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned error %s",
- prev->this->name,
- strerror (op_errno));
- if (local->op_errno != ESTALE)
- local->op_errno = op_errno;
- if (((op_errno != ENOENT) && (op_errno != ENOTCONN)
- && (op_errno != ESTALE)) ||
- (prev->this == FIRST_CHILD (this)))
- local->failed = 1;
- if (op_errno == ENOENT)
- local->entry_self_heal_needed = 1;
- }
-
- if (op_ret >= 0) {
- local->op_ret = 0;
- if (IA_ISREG (buf->ia_type)) {
- ret = stripe_ctx_handle (this, prev, local,
- xdata);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR,
- "Error getting fctx info from"
- " dict");
- }
-
- if (FIRST_CHILD(this) == prev->this) {
- local->stbuf = *buf;
- local->postparent = *postparent;
- local->inode = inode_ref (inode);
- if (xdata)
- local->xdata = dict_ref (xdata);
- if (local->xattr) {
- stripe_aggregate_xattr (local->xdata,
- local->xattr);
- dict_unref (local->xattr);
- local->xattr = NULL;
- }
- }
-
- if (!local->xdata && !local->xattr) {
- local->xattr = dict_ref (xdata);
- } else if (local->xdata) {
- stripe_aggregate_xattr (local->xdata, xdata);
- } else if (local->xattr) {
- stripe_aggregate_xattr (local->xattr, xdata);
- }
-
- local->stbuf_blocks += buf->ia_blocks;
- local->postparent_blocks += postparent->ia_blocks;
-
- correct_file_size(buf, local->fctx, prev);
-
- if (local->stbuf_size < buf->ia_size)
- local->stbuf_size = buf->ia_size;
- if (local->postparent_size < postparent->ia_size)
- local->postparent_size = postparent->ia_size;
-
- if (gf_uuid_is_null (local->ia_gfid))
- gf_uuid_copy (local->ia_gfid, buf->ia_gfid);
-
- /* Make sure the gfid on all the nodes are same */
- if (gf_uuid_compare (local->ia_gfid, buf->ia_gfid)) {
- gf_log (this->name, GF_LOG_WARNING,
- "%s: gfid different on subvolume %s",
- local->loc.path, prev->this->name);
- }
- }
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (local->op_ret == 0 && local->entry_self_heal_needed &&
- !gf_uuid_is_null (local->loc.inode->gfid))
- stripe_entry_self_heal (frame, this, local);
-
- if (local->failed)
- local->op_ret = -1;
-
- if (local->op_ret != -1) {
- local->stbuf.ia_blocks = local->stbuf_blocks;
- local->stbuf.ia_size = local->stbuf_size;
- local->postparent.ia_blocks = local->postparent_blocks;
- local->postparent.ia_size = local->postparent_size;
- inode_ctx_put (local->inode, this,
- (uint64_t) (long)local->fctx);
- }
-
- STRIPE_STACK_UNWIND (lookup, frame, local->op_ret,
- local->op_errno, local->inode,
- &local->stbuf, local->xdata,
- &local->postparent);
- }
-out:
- return 0;
-}
-
-int32_t
-stripe_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- xlator_list_t *trav = NULL;
- stripe_private_t *priv = NULL;
- int32_t op_errno = EINVAL;
- int64_t filesize = 0;
- int ret = 0;
- uint64_t tmpctx = 0;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- priv = this->private;
- trav = this->children;
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- frame->local = local;
- loc_copy (&local->loc, loc);
-
- inode_ctx_get (local->inode, this, &tmpctx);
- if (tmpctx)
- local->fctx = (stripe_fd_ctx_t*) (long)tmpctx;
-
- /* quick-read friendly changes */
- if (xdata && dict_get (xdata, GF_CONTENT_KEY)) {
- ret = dict_get_int64 (xdata, GF_CONTENT_KEY, &filesize);
- if (!ret && (filesize > priv->block_size))
- dict_del (xdata, GF_CONTENT_KEY);
- }
-
- /* get stripe-size xattr on lookup. This would be required for
- * open/read/write/pathinfo calls. Hence we send down the request
- * even when type == IA_INVAL */
-
- /*
- * We aren't guaranteed to have xdata here. We need the format info for
- * the file, so allocate xdata if necessary.
- */
- if (!xdata)
- xdata = dict_new();
- else
- xdata = dict_ref(xdata);
-
- if (xdata && (IA_ISREG (loc->inode->ia_type) ||
- (loc->inode->ia_type == IA_INVAL))) {
- ret = stripe_xattr_request_build (this, xdata, 8, 4, 4, 0);
- if (ret)
- gf_log (this->name , GF_LOG_ERROR, "Failed to build"
- " xattr request for %s", loc->path);
-
- }
-
- /* Every time in stripe lookup, all child nodes
- should be looked up */
- local->call_count = priv->child_count;
- while (trav) {
- STACK_WIND (frame, stripe_lookup_cbk, trav->xlator,
- trav->xlator->fops->lookup, loc, xdata);
- trav = trav->next;
- }
-
- dict_unref(xdata);
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
- prev = cookie;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned error %s",
- prev->this->name, strerror (op_errno));
- local->op_errno = op_errno;
- if ((op_errno != ENOENT) ||
- (prev->this == FIRST_CHILD (this)))
- local->failed = 1;
- }
-
- if (op_ret == 0) {
- local->op_ret = 0;
-
- if (FIRST_CHILD(this) == prev->this) {
- local->stbuf = *buf;
- }
-
- local->stbuf_blocks += buf->ia_blocks;
-
- correct_file_size(buf, local->fctx, prev);
-
- if (local->stbuf_size < buf->ia_size)
- local->stbuf_size = buf->ia_size;
- }
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (local->failed)
- local->op_ret = -1;
-
- if (local->op_ret != -1) {
- local->stbuf.ia_size = local->stbuf_size;
- local->stbuf.ia_blocks = local->stbuf_blocks;
- }
-
- STRIPE_STACK_UNWIND (stat, frame, local->op_ret,
- local->op_errno, &local->stbuf, NULL);
- }
-out:
- return 0;
-}
-
-int32_t
-stripe_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
-{
- xlator_list_t *trav = NULL;
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- stripe_fd_ctx_t *fctx = NULL;
- int32_t op_errno = EINVAL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->path, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- priv = this->private;
- trav = this->children;
-
- if (priv->first_child_down) {
- op_errno = ENOTCONN;
- goto err;
- }
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- frame->local = local;
- local->call_count = priv->child_count;
-
- if (IA_ISREG(loc->inode->ia_type)) {
- inode_ctx_get(loc->inode, this, (uint64_t *) &fctx);
- if (!fctx)
- goto err;
- local->fctx = fctx;
- }
-
- while (trav) {
- STACK_WIND (frame, stripe_stat_cbk, trav->xlator,
- trav->xlator->fops->stat, loc, NULL);
- trav = trav->next;
- }
-
- return 0;
-
-err:
- STRIPE_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct statvfs *stbuf, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- int32_t callcnt = 0;
-
- if (!this || !frame || !frame->local) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
- local = frame->local;
-
- LOCK(&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret && (op_errno != ENOTCONN)) {
- local->op_errno = op_errno;
- }
- if (op_ret == 0) {
- struct statvfs *dict_buf = &local->statvfs_buf;
- dict_buf->f_bsize = stbuf->f_bsize;
- dict_buf->f_frsize = stbuf->f_frsize;
- dict_buf->f_blocks += stbuf->f_blocks;
- dict_buf->f_bfree += stbuf->f_bfree;
- dict_buf->f_bavail += stbuf->f_bavail;
- dict_buf->f_files += stbuf->f_files;
- dict_buf->f_ffree += stbuf->f_ffree;
- dict_buf->f_favail += stbuf->f_favail;
- dict_buf->f_fsid = stbuf->f_fsid;
- dict_buf->f_flag = stbuf->f_flag;
- dict_buf->f_namemax = stbuf->f_namemax;
- local->op_ret = 0;
- }
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- STRIPE_STACK_UNWIND (statfs, frame, local->op_ret,
- local->op_errno, &local->statvfs_buf, NULL);
- }
-out:
- return 0;
-}
-
-int32_t
-stripe_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- xlator_list_t *trav = NULL;
- stripe_private_t *priv = NULL;
- int32_t op_errno = EINVAL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
-
- trav = this->children;
- priv = this->private;
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- local->op_errno = ENOTCONN;
- frame->local = local;
-
- local->call_count = priv->child_count;
- while (trav) {
- STACK_WIND (frame, stripe_statfs_cbk, trav->xlator,
- trav->xlator->fops->statfs, loc, NULL);
- trav = trav->next;
- }
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
- return 0;
-}
-
-
-
-int32_t
-stripe_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned error %s",
- prev->this->name, strerror (op_errno));
- local->op_errno = op_errno;
- if ((op_errno != ENOENT) ||
- (prev->this == FIRST_CHILD (this)))
- local->failed = 1;
- }
-
- if (op_ret == 0) {
- local->op_ret = 0;
- if (FIRST_CHILD(this) == prev->this) {
- local->pre_buf = *prebuf;
- local->post_buf = *postbuf;
- }
-
- local->prebuf_blocks += prebuf->ia_blocks;
- local->postbuf_blocks += postbuf->ia_blocks;
-
- correct_file_size(prebuf, local->fctx, prev);
- correct_file_size(postbuf, local->fctx, prev);
-
- if (local->prebuf_size < prebuf->ia_size)
- local->prebuf_size = prebuf->ia_size;
-
- if (local->postbuf_size < postbuf->ia_size)
- local->postbuf_size = postbuf->ia_size;
- }
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (local->failed)
- local->op_ret = -1;
-
- if (local->op_ret != -1) {
- local->pre_buf.ia_blocks = local->prebuf_blocks;
- local->pre_buf.ia_size = local->prebuf_size;
- local->post_buf.ia_blocks = local->postbuf_blocks;
- local->post_buf.ia_size = local->postbuf_size;
- }
-
- STRIPE_STACK_UNWIND (truncate, frame, local->op_ret,
- local->op_errno, &local->pre_buf,
- &local->post_buf, NULL);
- }
-out:
- return 0;
-}
-
-int32_t
-stripe_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- stripe_fd_ctx_t *fctx = NULL;
- int32_t op_errno = EINVAL;
- int i, eof_idx;
- off_t dest_offset, tmp_offset;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->path, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- priv = this->private;
-
- if (priv->first_child_down) {
- op_errno = ENOTCONN;
- goto err;
- }
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- frame->local = local;
- local->call_count = priv->child_count;
-
- inode_ctx_get(loc->inode, this, (uint64_t *) &fctx);
- if (!fctx) {
- gf_log(this->name, GF_LOG_ERROR, "no stripe context");
- op_errno = EINVAL;
- goto err;
- }
-
- local->fctx = fctx;
- eof_idx = (offset / fctx->stripe_size) % fctx->stripe_count;
-
- for (i = 0; i < fctx->stripe_count; i++) {
- if (!fctx->xl_array[i]) {
- gf_log(this->name, GF_LOG_ERROR,
- "no xlator at index %d", i);
- op_errno = EINVAL;
- goto err;
- }
-
- if (fctx->stripe_coalesce) {
- /*
- * The node that owns EOF is truncated to the exact
- * coalesced offset. Nodes prior to this index should
- * be rounded up to the size of the complete stripe,
- * while nodes after this index should be rounded down
- * to the size of the previous stripe.
- */
- if (i < eof_idx)
- tmp_offset = roof(offset, fctx->stripe_size *
- fctx->stripe_count);
- else if (i > eof_idx)
- tmp_offset = floor(offset, fctx->stripe_size *
- fctx->stripe_count);
- else
- tmp_offset = offset;
-
- dest_offset = coalesced_offset(tmp_offset,
- fctx->stripe_size, fctx->stripe_count);
- } else {
- dest_offset = offset;
- }
-
- STACK_WIND(frame, stripe_truncate_cbk, fctx->xl_array[i],
- fctx->xl_array[i]->fops->truncate, loc, dest_offset,
- NULL);
- }
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned error %s",
- prev->this->name, strerror (op_errno));
- local->op_errno = op_errno;
- if ((op_errno != ENOENT) ||
- (prev->this == FIRST_CHILD (this)))
- local->failed = 1;
- }
-
- if (op_ret == 0) {
- local->op_ret = 0;
-
- if (FIRST_CHILD(this) == prev->this) {
- local->pre_buf = *preop;
- local->post_buf = *postop;
- }
-
- local->prebuf_blocks += preop->ia_blocks;
- local->postbuf_blocks += postop->ia_blocks;
-
- correct_file_size(preop, local->fctx, prev);
- correct_file_size(postop, local->fctx, prev);
-
- if (local->prebuf_size < preop->ia_size)
- local->prebuf_size = preop->ia_size;
- if (local->postbuf_size < postop->ia_size)
- local->postbuf_size = postop->ia_size;
- }
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (local->failed)
- local->op_ret = -1;
-
- if (local->op_ret != -1) {
- local->pre_buf.ia_blocks = local->prebuf_blocks;
- local->pre_buf.ia_size = local->prebuf_size;
- local->post_buf.ia_blocks = local->postbuf_blocks;
- local->post_buf.ia_size = local->postbuf_size;
- }
-
- STRIPE_STACK_UNWIND (setattr, frame, local->op_ret,
- local->op_errno, &local->pre_buf,
- &local->post_buf, NULL);
- }
-out:
- return 0;
-}
-
-
-int32_t
-stripe_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
-{
- xlator_list_t *trav = NULL;
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- stripe_fd_ctx_t *fctx = NULL;
- int32_t op_errno = EINVAL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->path, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- priv = this->private;
- trav = this->children;
-
- if (priv->first_child_down) {
- op_errno = ENOTCONN;
- goto err;
- }
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- frame->local = local;
- if (!IA_ISDIR (loc->inode->ia_type) &&
- !IA_ISREG (loc->inode->ia_type)) {
- local->call_count = 1;
- STACK_WIND (frame, stripe_setattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setattr,
- loc, stbuf, valid, NULL);
- return 0;
- }
-
- if (IA_ISREG(loc->inode->ia_type)) {
- inode_ctx_get(loc->inode, this, (uint64_t *) &fctx);
- if (!fctx)
- goto err;
- local->fctx = fctx;
- }
-
- local->call_count = priv->child_count;
- while (trav) {
- STACK_WIND (frame, stripe_setattr_cbk,
- trav->xlator, trav->xlator->fops->setattr,
- loc, stbuf, valid, NULL);
- trav = trav->next;
- }
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- xlator_list_t *trav = NULL;
- int32_t op_errno = EINVAL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (fd->inode, err);
-
- priv = this->private;
- trav = this->children;
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- frame->local = local;
- local->call_count = priv->child_count;
-
- while (trav) {
- STACK_WIND (frame, stripe_setattr_cbk, trav->xlator,
- trav->xlator->fops->fsetattr, fd, stbuf, valid, NULL);
- trav = trav->next;
- }
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
-}
-
-int32_t
-stripe_stack_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent,
- dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned error %s",
- prev->this->name, strerror (op_errno));
- local->op_errno = op_errno;
- if ((op_errno != ENOENT) ||
- (prev->this == FIRST_CHILD (this)))
- local->failed = 1;
- }
-
- if (op_ret == 0) {
- local->op_ret = 0;
-
- local->stbuf.ia_blocks += buf->ia_blocks;
- local->preparent.ia_blocks += preoldparent->ia_blocks;
- local->postparent.ia_blocks += postoldparent->ia_blocks;
- local->pre_buf.ia_blocks += prenewparent->ia_blocks;
- local->post_buf.ia_blocks += postnewparent->ia_blocks;
-
- correct_file_size(buf, local->fctx, prev);
-
- if (local->stbuf.ia_size < buf->ia_size)
- local->stbuf.ia_size = buf->ia_size;
-
- if (local->preparent.ia_size < preoldparent->ia_size)
- local->preparent.ia_size = preoldparent->ia_size;
-
- if (local->postparent.ia_size < postoldparent->ia_size)
- local->postparent.ia_size = postoldparent->ia_size;
-
- if (local->pre_buf.ia_size < prenewparent->ia_size)
- local->pre_buf.ia_size = prenewparent->ia_size;
-
- if (local->post_buf.ia_size < postnewparent->ia_size)
- local->post_buf.ia_size = postnewparent->ia_size;
- }
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (local->failed)
- local->op_ret = -1;
-
- STRIPE_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
- &local->stbuf, &local->preparent,
- &local->postparent, &local->pre_buf,
- &local->post_buf, NULL);
- }
-out:
- return 0;
-}
-
-int32_t
-stripe_first_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent,
- dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- xlator_list_t *trav = NULL;
-
- if (!this || !frame || !frame->local) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- op_errno = EINVAL;
- goto unwind;
- }
-
- if (op_ret == -1) {
- goto unwind;
- }
-
- local = frame->local;
- trav = this->children;
-
- local->stbuf = *buf;
- local->preparent = *preoldparent;
- local->postparent = *postoldparent;
- local->pre_buf = *prenewparent;
- local->post_buf = *postnewparent;
-
- local->op_ret = 0;
- local->call_count--;
-
- trav = trav->next; /* Skip first child */
- while (trav) {
- STACK_WIND (frame, stripe_stack_rename_cbk,
- trav->xlator, trav->xlator->fops->rename,
- &local->loc, &local->loc2, NULL);
- trav = trav->next;
- }
- return 0;
-
-unwind:
- STRIPE_STACK_UNWIND (rename, frame, -1, op_errno, buf, preoldparent,
- postoldparent, prenewparent, postnewparent, NULL);
- return 0;
-}
-
-int32_t
-stripe_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
- loc_t *newloc, dict_t *xdata)
-{
- stripe_private_t *priv = NULL;
- stripe_local_t *local = NULL;
- xlator_list_t *trav = NULL;
- stripe_fd_ctx_t *fctx = NULL;
- int32_t op_errno = EINVAL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (oldloc, err);
- VALIDATE_OR_GOTO (oldloc->path, err);
- VALIDATE_OR_GOTO (oldloc->inode, err);
- VALIDATE_OR_GOTO (newloc, err);
-
- priv = this->private;
- trav = this->children;
-
- /* If any one node is down, don't allow rename */
- if (priv->nodes_down) {
- op_errno = ENOTCONN;
- goto err;
- }
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- frame->local = local;
-
- local->op_ret = -1;
- loc_copy (&local->loc, oldloc);
- loc_copy (&local->loc2, newloc);
-
- local->call_count = priv->child_count;
-
- if (IA_ISREG(oldloc->inode->ia_type)) {
- inode_ctx_get(oldloc->inode, this, (uint64_t *) &fctx);
- if (!fctx)
- goto err;
- local->fctx = fctx;
- }
-
- STACK_WIND (frame, stripe_first_rename_cbk, trav->xlator,
- trav->xlator->fops->rename, oldloc, newloc, NULL);
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL,
- NULL, NULL, NULL);
- return 0;
-}
-int32_t
-stripe_first_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG, "%s returned %s",
- prev->this->name, strerror (op_errno));
- goto out;
- }
- local->op_ret = 0;
- local->preparent = *preparent;
- local->postparent = *postparent;
- local->preparent_blocks += preparent->ia_blocks;
- local->postparent_blocks += postparent->ia_blocks;
-
- STRIPE_STACK_UNWIND(unlink, frame, local->op_ret, local->op_errno,
- &local->preparent, &local->postparent, xdata);
- return 0;
-out:
- STRIPE_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
-
- return 0;
-}
-
-
-
-
-int32_t
-stripe_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG, "%s returned %s",
- prev->this->name, strerror (op_errno));
- local->op_errno = op_errno;
- if (op_errno != ENOENT) {
- local->failed = 1;
- local->op_ret = op_ret;
- }
- }
- }
- UNLOCK (&frame->lock);
-
- if (callcnt == 1) {
- if (local->failed) {
- op_errno = local->op_errno;
- goto out;
- }
- STACK_WIND(frame, stripe_first_unlink_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->unlink, &local->loc,
- local->xflag, local->xdata);
- }
- return 0;
-out:
- STRIPE_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
-
- return 0;
-}
-
-int32_t
-stripe_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int xflag, dict_t *xdata)
-{
- xlator_list_t *trav = NULL;
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- int32_t op_errno = EINVAL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->path, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- priv = this->private;
- trav = this->children;
-
- if (priv->first_child_down) {
- op_errno = ENOTCONN;
- goto err;
- }
-
- /* Don't unlink a file if a node is down */
- if (priv->nodes_down) {
- op_errno = ENOTCONN;
- goto err;
- }
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- loc_copy (&local->loc, loc);
- local->xflag = xflag;
-
- if (xdata)
- local->xdata = dict_ref (xdata);
-
- frame->local = local;
- local->call_count = priv->child_count;
- trav = trav->next; /* Skip the first child */
-
- while (trav) {
- STACK_WIND (frame, stripe_unlink_cbk,
- trav->xlator, trav->xlator->fops->unlink,
- loc, xflag, xdata);
- trav = trav->next;
- }
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_first_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
-
- if (!this || !frame || !frame->local) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- op_errno = EINVAL;
- goto err;
- }
-
- if (op_ret == -1) {
- goto err;
- }
-
- local = frame->local;
- local->op_ret = 0;
-
- local->call_count--; /* First child successful */
-
- local->preparent = *preparent;
- local->postparent = *postparent;
- local->preparent_size = preparent->ia_size;
- local->postparent_size = postparent->ia_size;
- local->preparent_blocks += preparent->ia_blocks;
- local->postparent_blocks += postparent->ia_blocks;
-
- STRIPE_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
- &local->preparent, &local->postparent, xdata);
- return 0;
-err:
- STRIPE_STACK_UNWIND (rmdir, frame, op_ret, op_errno, NULL, NULL, NULL);
- return 0;
-
-}
-
-int32_t
-stripe_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG, "%s returned %s",
- prev->this->name, strerror (op_errno));
- if (op_errno != ENOENT)
- local->failed = 1;
- }
- }
- UNLOCK (&frame->lock);
-
- if (callcnt == 1) {
- if (local->failed)
- goto out;
- STACK_WIND (frame, stripe_first_rmdir_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->rmdir, &local->loc,
- local->flags, NULL);
- }
- return 0;
-out:
- STRIPE_STACK_UNWIND (rmdir, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
-}
-
-int32_t
-stripe_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags, dict_t *xdata)
-{
- xlator_list_t *trav = NULL;
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- int32_t op_errno = EINVAL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->path, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- priv = this->private;
- trav = this->children;
-
- /* don't delete a directory if any of the subvolume is down */
- if (priv->nodes_down) {
- op_errno = ENOTCONN;
- goto err;
- }
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- frame->local = local;
- loc_copy (&local->loc, loc);
- local->flags = flags;
- local->call_count = priv->child_count;
- trav = trav->next; /* skip the first child */
-
- while (trav) {
- STACK_WIND (frame, stripe_rmdir_cbk, trav->xlator,
- trav->xlator->fops->rmdir, loc, flags, NULL);
- trav = trav->next;
- }
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (rmdir, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_mknod_ifreg_fail_unlink_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
-
- if (!this || !frame || !frame->local) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- STRIPE_STACK_UNWIND (mknod, frame, local->op_ret, local->op_errno,
- local->inode, &local->stbuf,
- &local->preparent, &local->postparent, NULL);
- }
-out:
- return 0;
-}
-
-
-/**
- */
-int32_t
-stripe_mknod_ifreg_setxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- xlator_list_t *trav = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- priv = this->private;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned error %s",
- prev->this->name, strerror (op_errno));
- local->op_ret = -1;
- local->op_errno = op_errno;
- }
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (local->op_ret == -1) {
- local->call_count = priv->child_count;
- while (trav) {
- STACK_WIND (frame,
- stripe_mknod_ifreg_fail_unlink_cbk,
- trav->xlator,
- trav->xlator->fops->unlink,
- &local->loc, 0, NULL);
- trav = trav->next;
- }
- return 0;
- }
-
- STRIPE_STACK_UNWIND (mknod, frame, local->op_ret, local->op_errno,
- local->inode, &local->stbuf,
- &local->preparent, &local->postparent, NULL);
- }
-out:
- return 0;
-}
-
-int32_t
-stripe_mknod_ifreg_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- call_frame_t *prev = NULL;
- xlator_list_t *trav = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- priv = this->private;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned error %s",
- prev->this->name, strerror (op_errno));
- if ((op_errno != ENOENT) ||
- (prev->this == FIRST_CHILD (this)))
- local->failed = 1;
- local->op_errno = op_errno;
- }
- if (op_ret >= 0) {
- local->op_ret = op_ret;
-
- /* Can be used as a mechanism to understand if mknod
- was successful in at least one place */
- if (gf_uuid_is_null (local->ia_gfid))
- gf_uuid_copy (local->ia_gfid, buf->ia_gfid);
-
- if (stripe_ctx_handle(this, prev, local, xdata))
- gf_log(this->name, GF_LOG_ERROR,
- "Error getting fctx info from dict");
-
- local->stbuf_blocks += buf->ia_blocks;
- local->preparent_blocks += preparent->ia_blocks;
- local->postparent_blocks += postparent->ia_blocks;
-
- correct_file_size(buf, local->fctx, prev);
-
- if (local->stbuf_size < buf->ia_size)
- local->stbuf_size = buf->ia_size;
- if (local->preparent_size < preparent->ia_size)
- local->preparent_size = preparent->ia_size;
- if (local->postparent_size < postparent->ia_size)
- local->postparent_size = postparent->ia_size;
- }
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (local->failed)
- local->op_ret = -1;
-
- if ((local->op_ret == -1) && !gf_uuid_is_null (local->ia_gfid)) {
- /* ia_gfid set means, at least on one node 'mknod'
- is successful */
- local->call_count = priv->child_count;
- trav = this->children;
- while (trav) {
- STACK_WIND (frame,
- stripe_mknod_ifreg_fail_unlink_cbk,
- trav->xlator,
- trav->xlator->fops->unlink,
- &local->loc, 0, NULL);
- trav = trav->next;
- }
- return 0;
- }
-
-
- if (local->op_ret != -1) {
- local->preparent.ia_blocks = local->preparent_blocks;
- local->preparent.ia_size = local->preparent_size;
- local->postparent.ia_blocks = local->postparent_blocks;
- local->postparent.ia_size = local->postparent_size;
- local->stbuf.ia_size = local->stbuf_size;
- local->stbuf.ia_blocks = local->stbuf_blocks;
- inode_ctx_put (local->inode, this,
- (uint64_t)(long) local->fctx);
-
- }
- STRIPE_STACK_UNWIND (mknod, frame, local->op_ret, local->op_errno,
- local->inode, &local->stbuf,
- &local->preparent, &local->postparent, NULL);
- }
-out:
- return 0;
-}
-
-
-int32_t
-stripe_mknod_first_ifreg_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- call_frame_t *prev = NULL;
- xlator_list_t *trav = NULL;
- int i = 1;
- dict_t *dict = NULL;
- int ret = 0;
- int need_unref = 0;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- priv = this->private;
- local = frame->local;
- trav = this->children;
-
- local->call_count--;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG, "%s returned error %s",
- prev->this->name, strerror (op_errno));
- local->failed = 1;
- local->op_errno = op_errno;
- goto out;
- }
-
- local->op_ret = op_ret;
-
- local->stbuf = *buf;
- local->preparent = *preparent;
- local->postparent = *postparent;
-
- if (gf_uuid_is_null (local->ia_gfid))
- gf_uuid_copy (local->ia_gfid, buf->ia_gfid);
- local->preparent.ia_blocks = local->preparent_blocks;
- local->preparent.ia_size = local->preparent_size;
- local->postparent.ia_blocks = local->postparent_blocks;
- local->postparent.ia_size = local->postparent_size;
- local->stbuf.ia_size = local->stbuf_size;
- local->stbuf.ia_blocks = local->stbuf_blocks;
-
- trav = trav->next;
- while (trav) {
- if (priv->xattr_supported) {
- dict = dict_new ();
- if (!dict) {
- gf_log (this->name, GF_LOG_ERROR,
- "failed to allocate dict %s", local->loc.path);
- }
- need_unref = 1;
-
- dict_copy (local->xattr, dict);
-
- ret = stripe_xattr_request_build (this, dict,
- local->stripe_size,
- priv->child_count, i,
- priv->coalesce);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to build xattr request");
-
- } else {
- dict = local->xattr;
- }
-
- STACK_WIND (frame, stripe_mknod_ifreg_cbk,
- trav->xlator, trav->xlator->fops->mknod,
- &local->loc, local->mode, local->rdev, 0, dict);
- trav = trav->next;
- i++;
-
- if (dict && need_unref)
- dict_unref (dict);
- }
-
- return 0;
-
-out:
-
- STRIPE_STACK_UNWIND (mknod, frame, op_ret, op_errno, NULL, NULL, NULL, NULL, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_single_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- STRIPE_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
- return 0;
-}
-
-
-int
-stripe_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- dev_t rdev, mode_t umask, dict_t *xdata)
-{
- stripe_private_t *priv = NULL;
- stripe_local_t *local = NULL;
- int32_t op_errno = EINVAL;
- int32_t i = 0;
- dict_t *dict = NULL;
- int ret = 0;
- int need_unref = 0;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->path, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- priv = this->private;
-
- if (priv->first_child_down) {
- op_errno = ENOTCONN;
- goto err;
- }
-
- if (S_ISREG(mode)) {
- /* NOTE: on older kernels (older than 2.6.9),
- creat() fops is sent as mknod() + open(). Hence handling
- S_IFREG files is necessary */
- if (priv->nodes_down) {
- gf_log (this->name, GF_LOG_WARNING,
- "Some node down, returning EIO");
- op_errno = EIO;
- goto err;
- }
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- local->op_errno = ENOTCONN;
- local->stripe_size = stripe_get_matching_bs (loc->path, priv);
- frame->local = local;
- local->inode = inode_ref (loc->inode);
- loc_copy (&local->loc, loc);
- local->xattr = dict_copy_with_ref (xdata, NULL);
- local->mode = mode;
- local->umask = umask;
- local->rdev = rdev;
-
- /* Every time in stripe lookup, all child nodes should
- be looked up */
- local->call_count = priv->child_count;
-
- if (priv->xattr_supported) {
- dict = dict_new ();
- if (!dict) {
- gf_log (this->name, GF_LOG_ERROR,
- "failed to allocate dict %s", loc->path);
- }
- need_unref = 1;
-
- dict_copy (xdata, dict);
-
- ret = stripe_xattr_request_build (this, dict,
- local->stripe_size,
- priv->child_count,
- i, priv->coalesce);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR,
- "failed to build xattr request");
- } else {
- dict = xdata;
- }
-
- STACK_WIND (frame, stripe_mknod_first_ifreg_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->mknod,
- loc, mode, rdev, umask, dict);
-
- if (dict && need_unref)
- dict_unref (dict);
- return 0;
- }
-
- STACK_WIND (frame, stripe_single_mknod_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->mknod,
- loc, mode, rdev, umask, xdata);
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned error %s",
- prev->this->name, strerror (op_errno));
- local->op_errno = op_errno;
- if ((op_errno != ENOENT) ||
- (prev->this == FIRST_CHILD (this)))
- local->failed = 1;
- }
-
- if (op_ret >= 0) {
- local->op_ret = 0;
-
- local->stbuf_blocks += buf->ia_blocks;
- local->preparent_blocks += preparent->ia_blocks;
- local->postparent_blocks += postparent->ia_blocks;
-
- if (local->stbuf_size < buf->ia_size)
- local->stbuf_size = buf->ia_size;
- if (local->preparent_size < preparent->ia_size)
- local->preparent_size = preparent->ia_size;
- if (local->postparent_size < postparent->ia_size)
- local->postparent_size = postparent->ia_size;
- }
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (local->failed != -1) {
- local->preparent.ia_blocks = local->preparent_blocks;
- local->preparent.ia_size = local->preparent_size;
- local->postparent.ia_blocks = local->postparent_blocks;
- local->postparent.ia_size = local->postparent_size;
- local->stbuf.ia_size = local->stbuf_size;
- local->stbuf.ia_blocks = local->stbuf_blocks;
- }
- STRIPE_STACK_UNWIND (mkdir, frame, local->op_ret,
- local->op_errno, local->inode,
- &local->stbuf, &local->preparent,
- &local->postparent, NULL);
- }
-out:
- return 0;
-}
-
-
-int32_t
-stripe_first_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
- xlator_list_t *trav = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
- trav = this->children;
-
- local->call_count--; /* first child is successful */
- trav = trav->next; /* skip first child */
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG, "%s returned error %s",
- prev->this->name, strerror (op_errno));
- local->op_errno = op_errno;
- goto out;
- }
-
- local->op_ret = 0;
-
- local->inode = inode_ref (inode);
- local->stbuf = *buf;
- local->postparent = *postparent;
- local->preparent = *preparent;
-
- local->stbuf_blocks += buf->ia_blocks;
- local->preparent_blocks += preparent->ia_blocks;
- local->postparent_blocks += postparent->ia_blocks;
-
- local->stbuf_size = buf->ia_size;
- local->preparent_size = preparent->ia_size;
- local->postparent_size = postparent->ia_size;
-
- while (trav) {
- STACK_WIND (frame, stripe_mkdir_cbk, trav->xlator,
- trav->xlator->fops->mkdir, &local->loc, local->mode,
- local->umask, local->xdata);
- trav = trav->next;
- }
- return 0;
-out:
- STRIPE_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL,
- NULL, NULL);
-
- return 0;
-
-}
-
-
-int
-stripe_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- mode_t umask, dict_t *xdata)
-{
- stripe_private_t *priv = NULL;
- stripe_local_t *local = NULL;
- xlator_list_t *trav = NULL;
- int32_t op_errno = 1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->path, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- priv = this->private;
- trav = this->children;
-
- if (priv->first_child_down) {
- op_errno = ENOTCONN;
- goto err;
- }
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- local->call_count = priv->child_count;
- if (xdata)
- local->xdata = dict_ref (xdata);
- local->mode = mode;
- local->umask = umask;
- loc_copy (&local->loc, loc);
- frame->local = local;
-
- /* Every time in stripe lookup, all child nodes should be looked up */
- STACK_WIND (frame, stripe_first_mkdir_cbk, trav->xlator,
- trav->xlator->fops->mkdir, loc, mode, umask, xdata);
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
- stripe_fd_ctx_t *fctx = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned error %s",
- prev->this->name, strerror (op_errno));
- local->op_errno = op_errno;
- if ((op_errno != ENOENT) ||
- (prev->this == FIRST_CHILD (this)))
- local->failed = 1;
- }
-
- if (op_ret >= 0) {
- local->op_ret = 0;
-
- if (IA_ISREG(inode->ia_type)) {
- inode_ctx_get(inode, this, (uint64_t *) &fctx);
- if (!fctx) {
- gf_log(this->name, GF_LOG_ERROR,
- "failed to get stripe context");
- op_ret = -1;
- op_errno = EINVAL;
- }
- }
-
- if (FIRST_CHILD(this) == prev->this) {
- local->inode = inode_ref (inode);
- local->stbuf = *buf;
- local->postparent = *postparent;
- local->preparent = *preparent;
- }
- local->stbuf_blocks += buf->ia_blocks;
- local->preparent_blocks += preparent->ia_blocks;
- local->postparent_blocks += postparent->ia_blocks;
-
- correct_file_size(buf, fctx, prev);
-
- if (local->stbuf_size < buf->ia_size)
- local->stbuf_size = buf->ia_size;
- if (local->preparent_size < preparent->ia_size)
- local->preparent_size = preparent->ia_size;
- if (local->postparent_size < postparent->ia_size)
- local->postparent_size = postparent->ia_size;
- }
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (local->failed)
- local->op_ret = -1;
-
- if (local->op_ret != -1) {
- local->preparent.ia_blocks = local->preparent_blocks;
- local->preparent.ia_size = local->preparent_size;
- local->postparent.ia_blocks = local->postparent_blocks;
- local->postparent.ia_size = local->postparent_size;
- local->stbuf.ia_size = local->stbuf_size;
- local->stbuf.ia_blocks = local->stbuf_blocks;
- }
- STRIPE_STACK_UNWIND (link, frame, local->op_ret,
- local->op_errno, local->inode,
- &local->stbuf, &local->preparent,
- &local->postparent, NULL);
- }
-out:
- return 0;
-}
-
-int32_t
-stripe_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc, dict_t *xdata)
-{
- xlator_list_t *trav = NULL;
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- int32_t op_errno = 1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (oldloc, err);
- VALIDATE_OR_GOTO (oldloc->path, err);
- VALIDATE_OR_GOTO (oldloc->inode, err);
-
- priv = this->private;
- trav = this->children;
-
- /* If any one node is down, don't allow link operation */
- if (priv->nodes_down) {
- op_errno = ENOTCONN;
- goto err;
- }
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- frame->local = local;
- local->call_count = priv->child_count;
-
- /* Every time in stripe lookup, all child
- nodes should be looked up */
- while (trav) {
- STACK_WIND (frame, stripe_link_cbk,
- trav->xlator, trav->xlator->fops->link,
- oldloc, newloc, NULL);
- trav = trav->next;
- }
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
- return 0;
-}
-
-int32_t
-stripe_create_fail_unlink_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
-
- if (!this || !frame || !frame->local) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- STRIPE_STACK_UNWIND (create, frame, local->op_ret, local->op_errno,
- local->fd, local->inode, &local->stbuf,
- &local->preparent, &local->postparent, NULL);
- }
-out:
- return 0;
-}
-
-
-int32_t
-stripe_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd,
- inode_t *inode, struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- call_frame_t *prev = NULL;
- xlator_list_t *trav = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- priv = this->private;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned error %s",
- prev->this->name, strerror (op_errno));
- local->failed = 1;
- local->op_errno = op_errno;
- }
-
- if (op_ret >= 0) {
- if (IA_ISREG(buf->ia_type)) {
- if (stripe_ctx_handle(this, prev, local, xdata))
- gf_log(this->name, GF_LOG_ERROR,
- "Error getting fctx info from "
- "dict");
- }
-
- local->op_ret = op_ret;
-
- local->stbuf_blocks += buf->ia_blocks;
- local->preparent_blocks += preparent->ia_blocks;
- local->postparent_blocks += postparent->ia_blocks;
-
- correct_file_size(buf, local->fctx, prev);
-
- if (local->stbuf_size < buf->ia_size)
- local->stbuf_size = buf->ia_size;
- if (local->preparent_size < preparent->ia_size)
- local->preparent_size = preparent->ia_size;
- if (local->postparent_size < postparent->ia_size)
- local->postparent_size = postparent->ia_size;
- }
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (local->failed)
- local->op_ret = -1;
-
- if (local->op_ret == -1) {
- local->call_count = priv->child_count;
- trav = this->children;
- while (trav) {
- STACK_WIND (frame,
- stripe_create_fail_unlink_cbk,
- trav->xlator,
- trav->xlator->fops->unlink,
- &local->loc, 0, NULL);
- trav = trav->next;
- }
-
- return 0;
- }
-
- if (local->op_ret >= 0) {
- local->preparent.ia_blocks = local->preparent_blocks;
- local->preparent.ia_size = local->preparent_size;
- local->postparent.ia_blocks = local->postparent_blocks;
- local->postparent.ia_size = local->postparent_size;
- local->stbuf.ia_size = local->stbuf_size;
- local->stbuf.ia_blocks = local->stbuf_blocks;
-
- stripe_copy_xl_array(local->fctx->xl_array,
- priv->xl_array,
- local->fctx->stripe_count);
- inode_ctx_put(local->inode, this,
- (uint64_t) local->fctx);
- }
-
- /* Create itself has failed.. so return
- without setxattring */
- STRIPE_STACK_UNWIND (create, frame, local->op_ret,
- local->op_errno, local->fd,
- local->inode, &local->stbuf,
- &local->preparent, &local->postparent, NULL);
- }
-
-out:
- return 0;
-}
-
-
-
-int32_t
-stripe_first_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd,
- inode_t *inode, struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- call_frame_t *prev = NULL;
- xlator_list_t *trav = NULL;
- int i = 1;
- dict_t *dict = NULL;
- loc_t *loc = NULL;
- int32_t need_unref = 0;
- int32_t ret = -1;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- priv = this->private;
- local = frame->local;
- trav = this->children;
- loc = &local->loc;
-
- --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG, "%s returned error %s",
- prev->this->name, strerror (op_errno));
- local->failed = 1;
- local->op_errno = op_errno;
- }
-
- local->op_ret = 0;
- /* Get the mapping in inode private */
- /* Get the stat buf right */
- local->stbuf = *buf;
- local->preparent = *preparent;
- local->postparent = *postparent;
-
- local->stbuf_blocks += buf->ia_blocks;
- local->preparent_blocks += preparent->ia_blocks;
- local->postparent_blocks += postparent->ia_blocks;
-
- if (local->stbuf_size < buf->ia_size)
- local->stbuf_size = buf->ia_size;
- if (local->preparent_size < preparent->ia_size)
- local->preparent_size = preparent->ia_size;
- if (local->postparent_size < postparent->ia_size)
- local->postparent_size = postparent->ia_size;
-
- if (local->failed)
- local->op_ret = -1;
-
- if (local->op_ret == -1) {
- local->call_count = 1;
- STACK_WIND (frame, stripe_create_fail_unlink_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->unlink,
- &local->loc, 0, NULL);
- return 0;
- }
-
- if (local->op_ret >= 0) {
- local->preparent.ia_blocks = local->preparent_blocks;
- local->preparent.ia_size = local->preparent_size;
- local->postparent.ia_blocks = local->postparent_blocks;
- local->postparent.ia_size = local->postparent_size;
- local->stbuf.ia_size = local->stbuf_size;
- local->stbuf.ia_blocks = local->stbuf_blocks;
- }
-
- /* Send a setxattr request to nodes where the
- files are created */
- trav = trav->next;
- while (trav) {
- if (priv->xattr_supported) {
- dict = dict_new ();
- if (!dict) {
- gf_log (this->name, GF_LOG_ERROR,
- "failed to allocate dict %s", loc->path);
- }
- need_unref = 1;
-
- dict_copy (local->xattr, dict);
-
- ret = stripe_xattr_request_build (this, dict,
- local->stripe_size,
- priv->child_count,
- i, priv->coalesce);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR,
- "failed to build xattr request");
- } else {
- dict = local->xattr;
- }
-
- STACK_WIND (frame, stripe_create_cbk, trav->xlator,
- trav->xlator->fops->create, &local->loc,
- local->flags, local->mode, local->umask, local->fd,
- dict);
- trav = trav->next;
- if (need_unref && dict)
- dict_unref (dict);
- i++;
- }
-
-out:
- return 0;
-}
-
-
-
-/**
- * stripe_create - If a block-size is specified for the 'name', create the
- * file in all the child nodes. If not, create it in only first child.
- *
- * @name- complete path of the file to be created.
- */
-int32_t
-stripe_create (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int32_t flags, mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
-{
- stripe_private_t *priv = NULL;
- stripe_local_t *local = NULL;
- int32_t op_errno = EINVAL;
- int ret = 0;
- int need_unref = 0;
- int i = 0;
- dict_t *dict = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->path, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- priv = this->private;
-
- /* files created in O_APPEND mode does not allow lseek() on fd */
- flags &= ~O_APPEND;
-
- if (priv->first_child_down || priv->nodes_down) {
- gf_log (this->name, GF_LOG_DEBUG,
- "First node down, returning EIO");
- op_errno = EIO;
- goto err;
- }
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- local->op_errno = ENOTCONN;
- local->stripe_size = stripe_get_matching_bs (loc->path, priv);
- frame->local = local;
- local->inode = inode_ref (loc->inode);
- loc_copy (&local->loc, loc);
- local->fd = fd_ref (fd);
- local->flags = flags;
- local->mode = mode;
- local->umask = umask;
- if (xdata)
- local->xattr = dict_ref (xdata);
-
- local->call_count = priv->child_count;
- /* Send a setxattr request to nodes where the
- files are created */
-
- if (priv->xattr_supported) {
- dict = dict_new ();
- if (!dict) {
- gf_log (this->name, GF_LOG_ERROR,
- "failed to allocate dict %s", loc->path);
- }
- need_unref = 1;
-
- dict_copy (xdata, dict);
-
- ret = stripe_xattr_request_build (this, dict,
- local->stripe_size,
- priv->child_count,
- i, priv->coalesce);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR,
- "failed to build xattr request");
- } else {
- dict = xdata;
- }
-
-
- STACK_WIND (frame, stripe_first_create_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->create, loc, flags, mode,
- umask, fd, dict);
-
- if (need_unref && dict)
- dict_unref (dict);
-
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL,
- NULL, NULL, xdata);
- return 0;
-}
-
-int32_t
-stripe_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
-
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned error %s",
- prev->this->name, strerror (op_errno));
- if ((op_errno != ENOENT) ||
- (prev->this == FIRST_CHILD (this)))
- local->failed = 1;
- local->op_errno = op_errno;
- }
-
- if (op_ret >= 0)
- local->op_ret = op_ret;
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (local->failed)
- local->op_ret = -1;
-
- STRIPE_STACK_UNWIND (open, frame, local->op_ret,
- local->op_errno, local->fd, xdata);
- }
-out:
- return 0;
-}
-
-
-/**
- * stripe_open -
- */
-int32_t
-stripe_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int32_t flags, fd_t *fd, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- xlator_list_t *trav = NULL;
- int32_t op_errno = 1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->path, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- priv = this->private;
- trav = this->children;
-
- if (priv->first_child_down) {
- op_errno = ENOTCONN;
- goto err;
- }
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- /* files opened in O_APPEND mode does not allow lseek() on fd */
- flags &= ~O_APPEND;
-
- local->fd = fd_ref (fd);
- frame->local = local;
- loc_copy (&local->loc, loc);
-
- /* Striped files */
- local->flags = flags;
- local->call_count = priv->child_count;
- local->stripe_size = stripe_get_matching_bs (loc->path, priv);
-
- while (trav) {
- STACK_WIND (frame, stripe_open_cbk, trav->xlator,
- trav->xlator->fops->open,
- &local->loc, local->flags, local->fd,
- xdata);
- trav = trav->next;
- }
- return 0;
-err:
- STRIPE_STACK_UNWIND (open, frame, -1, op_errno, NULL, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned error %s",
- prev->this->name, strerror (op_errno));
- local->op_ret = -1;
- local->op_errno = op_errno;
- }
-
- if (op_ret >= 0)
- local->op_ret = op_ret;
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- STRIPE_STACK_UNWIND (opendir, frame, local->op_ret,
- local->op_errno, local->fd, NULL);
- }
-out:
- return 0;
-}
-
-
-int32_t
-stripe_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd, dict_t *xdata)
-{
- xlator_list_t *trav = NULL;
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- int32_t op_errno = EINVAL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->path, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- priv = this->private;
- trav = this->children;
-
- if (priv->first_child_down) {
- op_errno = ENOTCONN;
- goto err;
- }
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- frame->local = local;
- local->call_count = priv->child_count;
- local->fd = fd_ref (fd);
-
- while (trav) {
- STACK_WIND (frame, stripe_opendir_cbk, trav->xlator,
- trav->xlator->fops->opendir, loc, fd, NULL);
- trav = trav->next;
- }
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (opendir, frame, -1, op_errno, NULL, NULL);
- return 0;
-}
-
-int32_t
-stripe_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned error %s",
- prev->this->name, strerror (op_errno));
- local->op_errno = op_errno;
- if ((op_errno != ENOENT) ||
- (prev->this == FIRST_CHILD (this)))
- local->failed = 1;
- }
- if (op_ret >= 0) {
- if (FIRST_CHILD(this) == prev->this) {
- /* First successful call, copy the *lock */
- local->op_ret = op_ret;
- local->lock = *lock;
- }
- }
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (local->failed)
- local->op_ret = -1;
- STRIPE_STACK_UNWIND (lk, frame, local->op_ret,
- local->op_errno, &local->lock, NULL);
- }
-out:
- return 0;
-}
-
-int32_t
-stripe_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
- struct gf_flock *lock, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- xlator_list_t *trav = NULL;
- stripe_private_t *priv = NULL;
- int32_t op_errno = EINVAL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (fd->inode, err);
-
- trav = this->children;
- priv = this->private;
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- frame->local = local;
- local->call_count = priv->child_count;
-
- while (trav) {
- STACK_WIND (frame, stripe_lk_cbk, trav->xlator,
- trav->xlator->fops->lk, fd, cmd, lock, NULL);
- trav = trav->next;
- }
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (lk, frame, -1, op_errno, NULL, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned %s",
- prev->this->name, strerror (op_errno));
- local->op_errno = op_errno;
- if ((op_errno != ENOENT) ||
- (prev->this == FIRST_CHILD (this)))
- local->failed = 1;
- }
- if (op_ret >= 0)
- local->op_ret = op_ret;
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (local->failed)
- local->op_ret = -1;
-
- STRIPE_STACK_UNWIND (flush, frame, local->op_ret,
- local->op_errno, NULL);
- }
-out:
- return 0;
-}
-
-int32_t
-stripe_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- xlator_list_t *trav = NULL;
- int32_t op_errno = 1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (fd->inode, err);
-
- priv = this->private;
- trav = this->children;
-
- if (priv->first_child_down) {
- op_errno = ENOTCONN;
- goto err;
- }
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- frame->local = local;
- local->call_count = priv->child_count;
-
- while (trav) {
- STACK_WIND (frame, stripe_flush_cbk, trav->xlator,
- trav->xlator->fops->flush, fd, NULL);
- trav = trav->next;
- }
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (flush, frame, -1, op_errno, NULL);
- return 0;
-}
-
-
-
-int32_t
-stripe_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned %s",
- prev->this->name, strerror (op_errno));
- local->op_errno = op_errno;
- if ((op_errno != ENOENT) ||
- (prev->this == FIRST_CHILD (this)))
- local->failed = 1;
- }
- if (op_ret >= 0) {
- local->op_ret = op_ret;
- if (FIRST_CHILD(this) == prev->this) {
- local->pre_buf = *prebuf;
- local->post_buf = *postbuf;
- }
- local->prebuf_blocks += prebuf->ia_blocks;
- local->postbuf_blocks += postbuf->ia_blocks;
-
- correct_file_size(prebuf, local->fctx, prev);
- correct_file_size(postbuf, local->fctx, prev);
-
- if (local->prebuf_size < prebuf->ia_size)
- local->prebuf_size = prebuf->ia_size;
-
- if (local->postbuf_size < postbuf->ia_size)
- local->postbuf_size = postbuf->ia_size;
- }
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (local->failed)
- local->op_ret = -1;
-
- if (local->op_ret != -1) {
- local->pre_buf.ia_blocks = local->prebuf_blocks;
- local->pre_buf.ia_size = local->prebuf_size;
- local->post_buf.ia_blocks = local->postbuf_blocks;
- local->post_buf.ia_size = local->postbuf_size;
- }
-
- STRIPE_STACK_UNWIND (fsync, frame, local->op_ret,
- local->op_errno, &local->pre_buf,
- &local->post_buf, NULL);
- }
-out:
- return 0;
-}
-
-int32_t
-stripe_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- xlator_list_t *trav = NULL;
- stripe_fd_ctx_t *fctx = NULL;
- int32_t op_errno = 1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (fd->inode, err);
-
- priv = this->private;
- trav = this->children;
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- frame->local = local;
-
- inode_ctx_get(fd->inode, this, (uint64_t *) &fctx);
- if (!fctx) {
- op_errno = EINVAL;
- goto err;
- }
- local->fctx = fctx;
- local->op_ret = -1;
- local->call_count = priv->child_count;
-
- while (trav) {
- STACK_WIND (frame, stripe_fsync_cbk, trav->xlator,
- trav->xlator->fops->fsync, fd, flags, NULL);
- trav = trav->next;
- }
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
-}
-
-int32_t
-stripe_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned error %s",
- prev->this->name, strerror (op_errno));
- local->op_errno = op_errno;
- if ((op_errno != ENOENT) ||
- (prev->this == FIRST_CHILD (this)))
- local->failed = 1;
- }
-
- if (op_ret == 0) {
- local->op_ret = 0;
-
- if (FIRST_CHILD(this) == prev->this)
- local->stbuf = *buf;
-
- local->stbuf_blocks += buf->ia_blocks;
-
- correct_file_size(buf, local->fctx, prev);
-
- if (local->stbuf_size < buf->ia_size)
- local->stbuf_size = buf->ia_size;
- }
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (local->failed)
- local->op_ret = -1;
-
- if (local->op_ret != -1) {
- local->stbuf.ia_size = local->stbuf_size;
- local->stbuf.ia_blocks = local->stbuf_blocks;
- }
-
- STRIPE_STACK_UNWIND (fstat, frame, local->op_ret,
- local->op_errno, &local->stbuf, NULL);
- }
-
-out:
- return 0;
-}
-
-int32_t
-stripe_fstat (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- xlator_list_t *trav = NULL;
- stripe_fd_ctx_t *fctx = NULL;
- int32_t op_errno = 1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (fd->inode, err);
-
- priv = this->private;
- trav = this->children;
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- frame->local = local;
- local->call_count = priv->child_count;
-
- if (IA_ISREG(fd->inode->ia_type)) {
- inode_ctx_get(fd->inode, this, (uint64_t *) &fctx);
- if (!fctx)
- goto err;
- local->fctx = fctx;
- }
-
- while (trav) {
- STACK_WIND (frame, stripe_fstat_cbk, trav->xlator,
- trav->xlator->fops->fstat, fd, NULL);
- trav = trav->next;
- }
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (fstat, frame, -1, op_errno, NULL, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- stripe_fd_ctx_t *fctx = NULL;
- int i, eof_idx;
- off_t dest_offset, tmp_offset;
- int32_t op_errno = 1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (fd->inode, err);
-
- priv = this->private;
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- frame->local = local;
- local->call_count = priv->child_count;
-
- inode_ctx_get(fd->inode, this, (uint64_t *) &fctx);
- if (!fctx) {
- gf_log(this->name, GF_LOG_ERROR, "no stripe context");
- op_errno = EINVAL;
- goto err;
- }
- if (!fctx->stripe_count) {
- gf_log(this->name, GF_LOG_ERROR, "no stripe count");
- op_errno = EINVAL;
- goto err;
- }
-
- local->fctx = fctx;
- eof_idx = (offset / fctx->stripe_size) % fctx->stripe_count;
-
- for (i = 0; i < fctx->stripe_count; i++) {
- if (!fctx->xl_array[i]) {
- gf_log(this->name, GF_LOG_ERROR, "no xlator at index "
- "%d", i);
- op_errno = EINVAL;
- goto err;
- }
-
- if (fctx->stripe_coalesce) {
- if (i < eof_idx)
- tmp_offset = roof(offset, fctx->stripe_size *
- fctx->stripe_count);
- else if (i > eof_idx)
- tmp_offset = floor(offset, fctx->stripe_size *
- fctx->stripe_count);
- else
- tmp_offset = offset;
-
- dest_offset = coalesced_offset(tmp_offset,
- fctx->stripe_size, fctx->stripe_count);
- } else {
- dest_offset = offset;
- }
-
- STACK_WIND(frame, stripe_truncate_cbk, fctx->xl_array[i],
- fctx->xl_array[i]->fops->ftruncate, fd, dest_offset,
- NULL);
- }
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s returned %s",
- prev->this->name, strerror (op_errno));
- local->op_errno = op_errno;
- if ((op_errno != ENOENT) ||
- (prev->this == FIRST_CHILD (this)))
- local->failed = 1;
- }
- if (op_ret >= 0)
- local->op_ret = op_ret;
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (local->failed)
- local->op_ret = -1;
-
- STRIPE_STACK_UNWIND (fsyncdir, frame, local->op_ret,
- local->op_errno, NULL);
- }
-out:
- return 0;
-}
-
-int32_t
-stripe_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- xlator_list_t *trav = NULL;
- int32_t op_errno = 1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (fd->inode, err);
-
- priv = this->private;
- trav = this->children;
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- frame->local = local;
- local->call_count = priv->child_count;
-
- while (trav) {
- STACK_WIND (frame, stripe_fsyncdir_cbk, trav->xlator,
- trav->xlator->fops->fsyncdir, fd, flags, NULL);
- trav = trav->next;
- }
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (fsyncdir, frame, -1, op_errno, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_readv_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
-{
- int32_t i = 0;
- int32_t callcnt = 0;
- int32_t count = 0;
- stripe_local_t *local = NULL;
- struct iovec *vec = NULL;
- struct iatt tmp_stbuf = {0,};
- struct iobref *tmp_iobref = NULL;
- struct iobuf *iobuf = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- local = frame->local;
- prev = cookie;
-
- LOCK (&frame->lock);
- {
- callcnt = --local->call_count;
- if (op_ret != -1) {
- correct_file_size(buf, local->fctx, prev);
- if (local->stbuf_size < buf->ia_size)
- local->stbuf_size = buf->ia_size;
- }
- }
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- op_ret = 0;
-
- /* Keep extra space for filling in '\0's */
- vec = GF_CALLOC ((local->count * 2), sizeof (struct iovec),
- gf_stripe_mt_iovec);
- if (!vec) {
- op_ret = -1;
- goto done;
- }
-
- for (i = 0; i < local->wind_count; i++) {
- if (local->replies[i].op_ret) {
- memcpy ((vec + count), local->replies[i].vector,
- (local->replies[i].count * sizeof (struct iovec)));
- count += local->replies[i].count;
- op_ret += local->replies[i].op_ret;
- }
- if ((local->replies[i].op_ret <
- local->replies[i].requested_size) &&
- (local->stbuf_size > (local->offset + op_ret))) {
- /* Fill in 0s here */
- vec[count].iov_len =
- (local->replies[i].requested_size -
- local->replies[i].op_ret);
- iobuf = iobuf_get2 (this->ctx->iobuf_pool,
- vec[count].iov_len);
- if (!iobuf) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
- op_ret = -1;
- op_errno = ENOMEM;
- goto done;
- }
- memset (iobuf->ptr, 0, vec[count].iov_len);
- vec[count].iov_base = iobuf->ptr;
-
- iobref_add (local->iobref, iobuf);
- iobuf_unref(iobuf);
-
- op_ret += vec[count].iov_len;
- count++;
- }
- GF_FREE (local->replies[i].vector);
- }
-
- /* ENOENT signals EOF to the NFS-server */
- if (op_ret != -1 && op_ret < local->readv_size &&
- (local->offset + op_ret == buf->ia_size))
- op_errno = ENOENT;
-
- /* FIXME: notice that st_ino, and st_dev (gen) will be
- * different than what inode will have. Make sure this doesn't
- * cause any bugs at higher levels */
- memcpy (&tmp_stbuf, &local->replies[0].stbuf,
- sizeof (struct iatt));
- tmp_stbuf.ia_size = local->stbuf_size;
-
- done:
- GF_FREE (local->replies);
- tmp_iobref = local->iobref;
- STRIPE_STACK_UNWIND (readv, frame, op_ret, op_errno, vec,
- count, &tmp_stbuf, tmp_iobref, NULL);
-
- iobref_unref (tmp_iobref);
- GF_FREE (vec);
- }
-out:
- return 0;
-}
-
-/**
- * stripe_readv_cbk - get all the striped reads, and order it properly, send it
- * to above layer after putting it in a single vector.
- */
-int32_t
-stripe_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iovec *vector,
- int32_t count, struct iatt *stbuf, struct iobref *iobref, dict_t *xdata)
-{
- int32_t index = 0;
- int32_t callcnt = 0;
- int32_t final_count = 0;
- int32_t need_to_check_proper_size = 0;
- call_frame_t *mframe = NULL;
- stripe_local_t *mlocal = NULL;
- stripe_local_t *local = NULL;
- struct iovec *final_vec = NULL;
- struct iatt tmp_stbuf = {0,};
- struct iatt *tmp_stbuf_p = NULL; //need it for a warning
- struct iobref *tmp_iobref = NULL;
- stripe_fd_ctx_t *fctx = NULL;
- call_frame_t *prev = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto end;
- }
-
- local = frame->local;
- index = local->node_index;
- prev = cookie;
- mframe = local->orig_frame;
- if (!mframe)
- goto out;
-
- mlocal = mframe->local;
- if (!mlocal)
- goto out;
-
- fctx = mlocal->fctx;
-
- LOCK (&mframe->lock);
- {
- mlocal->replies[index].op_ret = op_ret;
- mlocal->replies[index].op_errno = op_errno;
- mlocal->replies[index].requested_size = local->readv_size;
- if (op_ret >= 0) {
- mlocal->replies[index].stbuf = *stbuf;
- mlocal->replies[index].count = count;
- mlocal->replies[index].vector = iov_dup (vector, count);
-
- correct_file_size(stbuf, fctx, prev);
-
- if (local->stbuf_size < stbuf->ia_size)
- local->stbuf_size = stbuf->ia_size;
- local->stbuf_blocks += stbuf->ia_blocks;
-
- if (!mlocal->iobref)
- mlocal->iobref = iobref_new ();
- iobref_merge (mlocal->iobref, iobref);
- }
- callcnt = ++mlocal->call_count;
- }
- UNLOCK(&mframe->lock);
-
- if (callcnt == mlocal->wind_count) {
- op_ret = 0;
-
- for (index=0; index < mlocal->wind_count; index++) {
- /* check whether each stripe returned
- * 'expected' number of bytes */
- if (mlocal->replies[index].op_ret == -1) {
- op_ret = -1;
- op_errno = mlocal->replies[index].op_errno;
- break;
- }
- /* TODO: handle the 'holes' within the read range
- properly */
- if (mlocal->replies[index].op_ret <
- mlocal->replies[index].requested_size) {
- need_to_check_proper_size = 1;
- }
-
- op_ret += mlocal->replies[index].op_ret;
- mlocal->count += mlocal->replies[index].count;
- }
- if (op_ret == -1)
- goto done;
- if (need_to_check_proper_size)
- goto check_size;
-
- final_vec = GF_CALLOC (mlocal->count, sizeof (struct iovec),
- gf_stripe_mt_iovec);
-
- if (!final_vec) {
- op_ret = -1;
- goto done;
- }
-
- for (index = 0; index < mlocal->wind_count; index++) {
- memcpy ((final_vec + final_count),
- mlocal->replies[index].vector,
- (mlocal->replies[index].count *
- sizeof (struct iovec)));
- final_count += mlocal->replies[index].count;
- GF_FREE (mlocal->replies[index].vector);
- }
-
- /* FIXME: notice that st_ino, and st_dev (gen) will be
- * different than what inode will have. Make sure this doesn't
- * cause any bugs at higher levels */
- memcpy (&tmp_stbuf, &mlocal->replies[0].stbuf,
- sizeof (struct iatt));
- tmp_stbuf.ia_size = local->stbuf_size;
- tmp_stbuf.ia_blocks = local->stbuf_blocks;
-
- done:
- /* */
- GF_FREE (mlocal->replies);
- tmp_iobref = mlocal->iobref;
- /* work around for nfs truncated read. Bug 3774 */
- tmp_stbuf_p = &tmp_stbuf;
- WIPE (tmp_stbuf_p);
- STRIPE_STACK_UNWIND (readv, mframe, op_ret, op_errno, final_vec,
- final_count, &tmp_stbuf, tmp_iobref, NULL);
-
- iobref_unref (tmp_iobref);
- GF_FREE (final_vec);
- }
-
- goto out;
-
-check_size:
- mlocal->call_count = fctx->stripe_count;
-
- for (index = 0; index < fctx->stripe_count; index++) {
- STACK_WIND (mframe, stripe_readv_fstat_cbk,
- (fctx->xl_array[index]),
- (fctx->xl_array[index])->fops->fstat,
- mlocal->fd, NULL);
- }
-
-out:
- STRIPE_STACK_DESTROY (frame);
-end:
- return 0;
-}
-
-
-int32_t
-stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
- size_t size, off_t offset, uint32_t flags, dict_t *xdata)
-{
- int32_t op_errno = EINVAL;
- int32_t idx = 0;
- int32_t index = 0;
- int32_t num_stripe = 0;
- int32_t off_index = 0;
- size_t frame_size = 0;
- off_t rounded_end = 0;
- uint64_t tmp_fctx = 0;
- uint64_t stripe_size = 0;
- off_t rounded_start = 0;
- off_t frame_offset = offset;
- off_t dest_offset = 0;
- stripe_local_t *local = NULL;
- call_frame_t *rframe = NULL;
- stripe_local_t *rlocal = NULL;
- stripe_fd_ctx_t *fctx = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (fd->inode, err);
-
- inode_ctx_get (fd->inode, this, &tmp_fctx);
- if (!tmp_fctx) {
- op_errno = EBADFD;
- goto err;
- }
- fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
- stripe_size = fctx->stripe_size;
-
- STRIPE_VALIDATE_FCTX (fctx, err);
-
- if (!stripe_size) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Wrong stripe size for the file");
- goto err;
- }
- /* The file is stripe across the child nodes. Send the read request
- * to the child nodes appropriately after checking which region of
- * the file is in which child node. Always '0-<stripe_size>' part of
- * the file resides in the first child.
- */
- rounded_start = floor (offset, stripe_size);
- rounded_end = roof (offset+size, stripe_size);
- num_stripe = (rounded_end- rounded_start)/stripe_size;
-
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- frame->local = local;
-
- /* This is where all the vectors should be copied. */
- local->replies = GF_CALLOC (num_stripe, sizeof (struct stripe_replies),
- gf_stripe_mt_stripe_replies);
- if (!local->replies) {
- op_errno = ENOMEM;
- goto err;
- }
-
- off_index = (offset / stripe_size) % fctx->stripe_count;
- local->wind_count = num_stripe;
- local->readv_size = size;
- local->offset = offset;
- local->fd = fd_ref (fd);
- local->fctx = fctx;
-
- for (index = off_index; index < (num_stripe + off_index); index++) {
- rframe = copy_frame (frame);
- rlocal = mem_get0 (this->local_pool);
- if (!rlocal) {
- op_errno = ENOMEM;
- goto err;
- }
-
- frame_size = min (roof (frame_offset+1, stripe_size),
- (offset + size)) - frame_offset;
-
- rlocal->node_index = index - off_index;
- rlocal->orig_frame = frame;
- rlocal->readv_size = frame_size;
- rframe->local = rlocal;
- idx = (index % fctx->stripe_count);
-
- if (fctx->stripe_coalesce)
- dest_offset = coalesced_offset(frame_offset,
- stripe_size, fctx->stripe_count);
- else
- dest_offset = frame_offset;
-
- STACK_WIND (rframe, stripe_readv_cbk, fctx->xl_array[idx],
- fctx->xl_array[idx]->fops->readv,
- fd, frame_size, dest_offset, flags, xdata);
-
- frame_offset += frame_size;
- }
-
- return 0;
-err:
- if (rframe)
- STRIPE_STACK_DESTROY (rframe);
-
- STRIPE_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, NULL, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- stripe_local_t *mlocal = NULL;
- call_frame_t *prev = NULL;
- call_frame_t *mframe = NULL;
- struct stripe_replies *reply = NULL;
- int32_t i = 0;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
- mframe = local->orig_frame;
- mlocal = mframe->local;
-
- LOCK(&frame->lock);
- {
- callcnt = ++mlocal->call_count;
-
- mlocal->replies[local->node_index].op_ret = op_ret;
- mlocal->replies[local->node_index].op_errno = op_errno;
-
- if (op_ret >= 0) {
- mlocal->post_buf = *postbuf;
- mlocal->pre_buf = *prebuf;
-
- mlocal->prebuf_blocks += prebuf->ia_blocks;
- mlocal->postbuf_blocks += postbuf->ia_blocks;
-
- correct_file_size(prebuf, mlocal->fctx, prev);
- correct_file_size(postbuf, mlocal->fctx, prev);
-
- if (mlocal->prebuf_size < prebuf->ia_size)
- mlocal->prebuf_size = prebuf->ia_size;
- if (mlocal->postbuf_size < postbuf->ia_size)
- mlocal->postbuf_size = postbuf->ia_size;
- }
- }
- UNLOCK (&frame->lock);
-
- if ((callcnt == mlocal->wind_count) && mlocal->unwind) {
- mlocal->pre_buf.ia_size = mlocal->prebuf_size;
- mlocal->pre_buf.ia_blocks = mlocal->prebuf_blocks;
- mlocal->post_buf.ia_size = mlocal->postbuf_size;
- mlocal->post_buf.ia_blocks = mlocal->postbuf_blocks;
-
- /*
- * Only return the number of consecutively written bytes up until
- * the first error. Only return an error if it occurs first.
- *
- * When a short write occurs, the application should retry at the
- * appropriate offset, at which point we'll potentially pass back
- * the error.
- */
- for (i = 0, reply = mlocal->replies; i < mlocal->wind_count;
- i++, reply++) {
- if (reply->op_ret == -1) {
- gf_log(this->name, GF_LOG_DEBUG, "reply %d "
- "returned error %s", i,
- strerror(reply->op_errno));
- if (!mlocal->op_ret) {
- mlocal->op_ret = -1;
- mlocal->op_errno = reply->op_errno;
- }
- break;
- }
-
- mlocal->op_ret += reply->op_ret;
-
- if (reply->op_ret < reply->requested_size)
- break;
- }
-
- GF_FREE(mlocal->replies);
-
- STRIPE_STACK_UNWIND (writev, mframe, mlocal->op_ret,
- mlocal->op_errno, &mlocal->pre_buf,
- &mlocal->post_buf, NULL);
- }
-out:
- STRIPE_STACK_DESTROY(frame);
- return 0;
-}
-
-int32_t
-stripe_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count, off_t offset,
- uint32_t flags, struct iobref *iobref, dict_t *xdata)
-{
- struct iovec *tmp_vec = NULL;
- stripe_local_t *local = NULL;
- stripe_fd_ctx_t *fctx = NULL;
- int32_t op_errno = 1;
- int32_t idx = 0;
- int32_t total_size = 0;
- int32_t offset_offset = 0;
- int32_t remaining_size = 0;
- int32_t tmp_count = count;
- off_t fill_size = 0;
- uint64_t stripe_size = 0;
- uint64_t tmp_fctx = 0;
- off_t dest_offset = 0;
- off_t rounded_start = 0;
- off_t rounded_end = 0;
- int32_t total_chunks = 0;
- call_frame_t *wframe = NULL;
- stripe_local_t *wlocal = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (fd->inode, err);
-
- inode_ctx_get (fd->inode, this, &tmp_fctx);
- if (!tmp_fctx) {
- op_errno = EINVAL;
- goto err;
- }
- fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
- stripe_size = fctx->stripe_size;
-
- STRIPE_VALIDATE_FCTX (fctx, err);
-
- /* File has to be stripped across the child nodes */
- total_size = iov_length (vector, count);
- remaining_size = total_size;
-
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- frame->local = local;
- local->stripe_size = stripe_size;
- local->fctx = fctx;
-
- if (!stripe_size) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Wrong stripe size for the file");
- op_errno = EINVAL;
- goto err;
- }
-
- rounded_start = floor(offset, stripe_size);
- rounded_end = roof(offset + total_size, stripe_size);
- total_chunks = (rounded_end - rounded_start) / stripe_size;
- local->replies = GF_CALLOC(total_chunks, sizeof(struct stripe_replies),
- gf_stripe_mt_stripe_replies);
- if (!local->replies) {
- op_errno = ENOMEM;
- goto err;
- }
-
- total_chunks = 0;
- while (1) {
- wframe = copy_frame(frame);
- wlocal = mem_get0(this->local_pool);
- if (!wlocal) {
- op_errno = ENOMEM;
- goto err;
- }
- wlocal->orig_frame = frame;
- wframe->local = wlocal;
-
- /* Send striped chunk of the vector to child
- nodes appropriately. */
- idx = (((offset + offset_offset) /
- local->stripe_size) % fctx->stripe_count);
-
- fill_size = (local->stripe_size -
- ((offset + offset_offset) % local->stripe_size));
- if (fill_size > remaining_size)
- fill_size = remaining_size;
-
- remaining_size -= fill_size;
-
- tmp_count = iov_subset (vector, count, offset_offset,
- offset_offset + fill_size, NULL);
- tmp_vec = GF_CALLOC (tmp_count, sizeof (struct iovec),
- gf_stripe_mt_iovec);
- if (!tmp_vec) {
- op_errno = ENOMEM;
- goto err;
- }
- tmp_count = iov_subset (vector, count, offset_offset,
- offset_offset + fill_size, tmp_vec);
-
- local->wind_count++;
- if (remaining_size == 0)
- local->unwind = 1;
-
- /*
- * Store off the request index (with respect to the chunk of the
- * initial offset) and the size of the request. This is required
- * in the callback to calculate an appropriate return value in
- * the event of a write failure in one or more requests.
- */
- wlocal->node_index = total_chunks;
- local->replies[total_chunks].requested_size = fill_size;
-
- dest_offset = offset + offset_offset;
- if (fctx->stripe_coalesce)
- dest_offset = coalesced_offset(dest_offset,
- local->stripe_size, fctx->stripe_count);
-
- STACK_WIND (wframe, stripe_writev_cbk, fctx->xl_array[idx],
- fctx->xl_array[idx]->fops->writev, fd, tmp_vec,
- tmp_count, dest_offset, flags, iobref,
- xdata);
-
- GF_FREE (tmp_vec);
- offset_offset += fill_size;
- total_chunks++;
- if (remaining_size == 0)
- break;
- }
-
- return 0;
-err:
- if (wframe)
- STRIPE_STACK_DESTROY(wframe);
-
- STRIPE_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- stripe_local_t *mlocal = NULL;
- call_frame_t *prev = NULL;
- call_frame_t *mframe = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
- mframe = local->orig_frame;
- mlocal = mframe->local;
-
- LOCK(&frame->lock);
- {
- callcnt = ++mlocal->call_count;
-
- if (op_ret == 0) {
- mlocal->post_buf = *postbuf;
- mlocal->pre_buf = *prebuf;
-
- mlocal->prebuf_blocks += prebuf->ia_blocks;
- mlocal->postbuf_blocks += postbuf->ia_blocks;
-
- correct_file_size(prebuf, mlocal->fctx, prev);
- correct_file_size(postbuf, mlocal->fctx, prev);
-
- if (mlocal->prebuf_size < prebuf->ia_size)
- mlocal->prebuf_size = prebuf->ia_size;
- if (mlocal->postbuf_size < postbuf->ia_size)
- mlocal->postbuf_size = postbuf->ia_size;
- }
-
- /* return the first failure */
- if (mlocal->op_ret == 0) {
- mlocal->op_ret = op_ret;
- mlocal->op_errno = op_errno;
- }
- }
- UNLOCK (&frame->lock);
-
- if ((callcnt == mlocal->wind_count) && mlocal->unwind) {
- mlocal->pre_buf.ia_size = mlocal->prebuf_size;
- mlocal->pre_buf.ia_blocks = mlocal->prebuf_blocks;
- mlocal->post_buf.ia_size = mlocal->postbuf_size;
- mlocal->post_buf.ia_blocks = mlocal->postbuf_blocks;
-
- STRIPE_STACK_UNWIND (fallocate, mframe, mlocal->op_ret,
- mlocal->op_errno, &mlocal->pre_buf,
- &mlocal->post_buf, NULL);
- }
-out:
- STRIPE_STACK_DESTROY(frame);
- return 0;
-}
-
-int32_t
-stripe_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
- off_t offset, size_t len, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- stripe_fd_ctx_t *fctx = NULL;
- int32_t op_errno = 1;
- int32_t idx = 0;
- int32_t offset_offset = 0;
- int32_t remaining_size = 0;
- off_t fill_size = 0;
- uint64_t stripe_size = 0;
- uint64_t tmp_fctx = 0;
- off_t dest_offset = 0;
- call_frame_t *fframe = NULL;
- stripe_local_t *flocal = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (fd->inode, err);
-
- inode_ctx_get (fd->inode, this, &tmp_fctx);
- if (!tmp_fctx) {
- op_errno = EINVAL;
- goto err;
- }
- fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
- stripe_size = fctx->stripe_size;
-
- STRIPE_VALIDATE_FCTX (fctx, err);
-
- remaining_size = len;
-
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- frame->local = local;
- local->stripe_size = stripe_size;
- local->fctx = fctx;
-
- if (!stripe_size) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Wrong stripe size for the file");
- op_errno = EINVAL;
- goto err;
- }
-
- while (1) {
- fframe = copy_frame(frame);
- flocal = mem_get0(this->local_pool);
- if (!flocal) {
- op_errno = ENOMEM;
- goto err;
- }
- flocal->orig_frame = frame;
- fframe->local = flocal;
-
- /* send fallocate request to the associated child node */
- idx = (((offset + offset_offset) /
- local->stripe_size) % fctx->stripe_count);
-
- fill_size = (local->stripe_size -
- ((offset + offset_offset) % local->stripe_size));
- if (fill_size > remaining_size)
- fill_size = remaining_size;
-
- remaining_size -= fill_size;
-
- local->wind_count++;
- if (remaining_size == 0)
- local->unwind = 1;
-
- dest_offset = offset + offset_offset;
- if (fctx->stripe_coalesce)
- dest_offset = coalesced_offset(dest_offset,
- local->stripe_size, fctx->stripe_count);
-
- /*
- * TODO: Create a separate handler for coalesce mode that sends a
- * single fallocate per-child (since the ranges are linear).
- */
- STACK_WIND(fframe, stripe_fallocate_cbk, fctx->xl_array[idx],
- fctx->xl_array[idx]->fops->fallocate, fd, mode,
- dest_offset, fill_size, xdata);
-
- offset_offset += fill_size;
- if (remaining_size == 0)
- break;
- }
-
- return 0;
-err:
- if (fframe)
- STRIPE_STACK_DESTROY(fframe);
-
- STRIPE_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
-}
-
-
-int32_t
-stripe_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- stripe_local_t *mlocal = NULL;
- call_frame_t *prev = NULL;
- call_frame_t *mframe = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
- mframe = local->orig_frame;
- mlocal = mframe->local;
-
- LOCK(&frame->lock);
- {
- callcnt = ++mlocal->call_count;
-
- if (op_ret == 0) {
- mlocal->post_buf = *postbuf;
- mlocal->pre_buf = *prebuf;
-
- mlocal->prebuf_blocks += prebuf->ia_blocks;
- mlocal->postbuf_blocks += postbuf->ia_blocks;
-
- correct_file_size(prebuf, mlocal->fctx, prev);
- correct_file_size(postbuf, mlocal->fctx, prev);
-
- if (mlocal->prebuf_size < prebuf->ia_size)
- mlocal->prebuf_size = prebuf->ia_size;
- if (mlocal->postbuf_size < postbuf->ia_size)
- mlocal->postbuf_size = postbuf->ia_size;
- }
-
- /* return the first failure */
- if (mlocal->op_ret == 0) {
- mlocal->op_ret = op_ret;
- mlocal->op_errno = op_errno;
- }
- }
- UNLOCK (&frame->lock);
-
- if ((callcnt == mlocal->wind_count) && mlocal->unwind) {
- mlocal->pre_buf.ia_size = mlocal->prebuf_size;
- mlocal->pre_buf.ia_blocks = mlocal->prebuf_blocks;
- mlocal->post_buf.ia_size = mlocal->postbuf_size;
- mlocal->post_buf.ia_blocks = mlocal->postbuf_blocks;
-
- STRIPE_STACK_UNWIND (discard, mframe, mlocal->op_ret,
- mlocal->op_errno, &mlocal->pre_buf,
- &mlocal->post_buf, NULL);
- }
-out:
- if (frame)
- STRIPE_STACK_DESTROY(frame);
-
- return 0;
-}
-
-int32_t
-stripe_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- size_t len, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- stripe_fd_ctx_t *fctx = NULL;
- int32_t op_errno = 1;
- int32_t idx = 0;
- int32_t offset_offset = 0;
- int32_t remaining_size = 0;
- off_t fill_size = 0;
- uint64_t stripe_size = 0;
- uint64_t tmp_fctx = 0;
- off_t dest_offset = 0;
- call_frame_t *fframe = NULL;
- stripe_local_t *flocal = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (fd->inode, err);
-
- inode_ctx_get (fd->inode, this, &tmp_fctx);
- if (!tmp_fctx) {
- op_errno = EINVAL;
- goto err;
- }
- fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
- stripe_size = fctx->stripe_size;
-
- STRIPE_VALIDATE_FCTX (fctx, err);
-
- remaining_size = len;
-
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- frame->local = local;
- local->stripe_size = stripe_size;
- local->fctx = fctx;
-
- if (!stripe_size) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Wrong stripe size for the file");
- op_errno = EINVAL;
- goto err;
- }
-
- while (1) {
- fframe = copy_frame(frame);
- flocal = mem_get0(this->local_pool);
- if (!flocal) {
- op_errno = ENOMEM;
- goto err;
- }
- flocal->orig_frame = frame;
- fframe->local = flocal;
-
- /* send discard request to the associated child node */
- idx = (((offset + offset_offset) /
- local->stripe_size) % fctx->stripe_count);
-
- fill_size = (local->stripe_size -
- ((offset + offset_offset) % local->stripe_size));
- if (fill_size > remaining_size)
- fill_size = remaining_size;
-
- remaining_size -= fill_size;
-
- local->wind_count++;
- if (remaining_size == 0)
- local->unwind = 1;
-
- dest_offset = offset + offset_offset;
- if (fctx->stripe_coalesce)
- dest_offset = coalesced_offset(dest_offset,
- local->stripe_size, fctx->stripe_count);
-
- /*
- * TODO: Create a separate handler for coalesce mode that sends a
- * single discard per-child (since the ranges are linear).
- */
- STACK_WIND(fframe, stripe_discard_cbk, fctx->xl_array[idx],
- fctx->xl_array[idx]->fops->discard, fd, dest_offset,
- fill_size, xdata);
-
- offset_offset += fill_size;
- if (remaining_size == 0)
- break;
- }
-
- return 0;
-err:
- if (fframe)
- STRIPE_STACK_DESTROY(fframe);
-
- STRIPE_STACK_UNWIND (discard, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
-}
-
-int32_t
-stripe_zerofill_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
-{
- int32_t callcnt = 0;
- stripe_local_t *local = NULL;
- stripe_local_t *mlocal = NULL;
- call_frame_t *prev = NULL;
- call_frame_t *mframe = NULL;
-
- GF_ASSERT (frame);
-
- if (!this || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
-
- prev = cookie;
- local = frame->local;
- mframe = local->orig_frame;
- mlocal = mframe->local;
-
- LOCK(&frame->lock);
- {
- callcnt = ++mlocal->call_count;
-
- if (op_ret == 0) {
- mlocal->post_buf = *postbuf;
- mlocal->pre_buf = *prebuf;
-
- mlocal->prebuf_blocks += prebuf->ia_blocks;
- mlocal->postbuf_blocks += postbuf->ia_blocks;
-
- correct_file_size(prebuf, mlocal->fctx, prev);
- correct_file_size(postbuf, mlocal->fctx, prev);
-
- if (mlocal->prebuf_size < prebuf->ia_size)
- mlocal->prebuf_size = prebuf->ia_size;
- if (mlocal->postbuf_size < postbuf->ia_size)
- mlocal->postbuf_size = postbuf->ia_size;
- }
-
- /* return the first failure */
- if (mlocal->op_ret == 0) {
- mlocal->op_ret = op_ret;
- mlocal->op_errno = op_errno;
- }
- }
- UNLOCK (&frame->lock);
-
- if ((callcnt == mlocal->wind_count) && mlocal->unwind) {
- mlocal->pre_buf.ia_size = mlocal->prebuf_size;
- mlocal->pre_buf.ia_blocks = mlocal->prebuf_blocks;
- mlocal->post_buf.ia_size = mlocal->postbuf_size;
- mlocal->post_buf.ia_blocks = mlocal->postbuf_blocks;
-
- STRIPE_STACK_UNWIND (zerofill, mframe, mlocal->op_ret,
- mlocal->op_errno, &mlocal->pre_buf,
- &mlocal->post_buf, NULL);
- }
-out:
- STRIPE_STACK_DESTROY(frame);
- return 0;
-}
-
-int32_t
-stripe_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- off_t len, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- stripe_fd_ctx_t *fctx = NULL;
- int32_t op_errno = 1;
- int32_t idx = 0;
- int32_t offset_offset = 0;
- int32_t remaining_size = 0;
- off_t fill_size = 0;
- uint64_t stripe_size = 0;
- uint64_t tmp_fctx = 0;
- off_t dest_offset = 0;
- call_frame_t *fframe = NULL;
- stripe_local_t *flocal = NULL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
- VALIDATE_OR_GOTO (fd->inode, err);
-
- inode_ctx_get (fd->inode, this, &tmp_fctx);
- if (!tmp_fctx) {
- op_errno = EINVAL;
- goto err;
- }
- fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
- stripe_size = fctx->stripe_size;
-
- STRIPE_VALIDATE_FCTX (fctx, err);
-
- remaining_size = len;
-
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- frame->local = local;
- local->stripe_size = stripe_size;
- local->fctx = fctx;
-
- if (!stripe_size) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Wrong stripe size for the file");
- op_errno = EINVAL;
- goto err;
- }
-
- while (1) {
- fframe = copy_frame(frame);
- flocal = mem_get0(this->local_pool);
- if (!flocal) {
- op_errno = ENOMEM;
- goto err;
- }
- flocal->orig_frame = frame;
- fframe->local = flocal;
-
- idx = (((offset + offset_offset) /
- local->stripe_size) % fctx->stripe_count);
-
- fill_size = (local->stripe_size -
- ((offset + offset_offset) % local->stripe_size));
- if (fill_size > remaining_size)
- fill_size = remaining_size;
-
- remaining_size -= fill_size;
-
- local->wind_count++;
- if (remaining_size == 0)
- local->unwind = 1;
-
- dest_offset = offset + offset_offset;
- if (fctx->stripe_coalesce)
- dest_offset = coalesced_offset(dest_offset,
- local->stripe_size,
- fctx->stripe_count);
-
- STACK_WIND(fframe, stripe_zerofill_cbk, fctx->xl_array[idx],
- fctx->xl_array[idx]->fops->zerofill, fd,
- dest_offset, fill_size, xdata);
- offset_offset += fill_size;
- if (remaining_size == 0)
- break;
- }
-
- return 0;
-err:
- if (fframe)
- STRIPE_STACK_DESTROY(fframe);
-
- STRIPE_STACK_UNWIND (zerofill, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
-}
-
-int32_t
-stripe_seek (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- gf_seek_what_t what, dict_t *xdata)
-{
- /* TBD */
- gf_log (this->name, GF_LOG_INFO, "seek called on %s.",
- uuid_utoa (fd->inode->gfid));
- STRIPE_STACK_UNWIND (seek, frame, -1, ENOTSUP, 0, NULL);
- return 0;
-}
-
-int32_t
-stripe_release (xlator_t *this, fd_t *fd)
-{
- return 0;
-}
-
-int
-stripe_forget (xlator_t *this, inode_t *inode)
-{
- uint64_t tmp_fctx = 0;
- stripe_fd_ctx_t *fctx = NULL;
-
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (inode, err);
-
- (void) inode_ctx_del (inode, this, &tmp_fctx);
- if (!tmp_fctx) {
- goto err;
- }
-
- fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
-
- if (!fctx->static_array)
- GF_FREE (fctx->xl_array);
-
- GF_FREE (fctx);
-err:
- return 0;
-}
-
-int32_t
-notify (xlator_t *this, int32_t event, void *data, ...)
-{
- stripe_private_t *priv = NULL;
- int down_client = 0;
- int i = 0;
- gf_boolean_t heard_from_all_children = _gf_false;
-
- if (!this)
- return 0;
-
- priv = this->private;
- if (!priv)
- return 0;
-
- switch (event)
- {
- case GF_EVENT_CHILD_UP:
- {
- /* get an index number to set */
- for (i = 0; i < priv->child_count; i++) {
- if (data == priv->xl_array[i])
- break;
- }
-
- if (priv->child_count == i) {
- gf_log (this->name, GF_LOG_ERROR,
- "got GF_EVENT_CHILD_UP bad subvolume %s",
- data? ((xlator_t *)data)->name: NULL);
- break;
- }
-
- LOCK (&priv->lock);
- {
- if (data == FIRST_CHILD (this))
- priv->first_child_down = 0;
- priv->last_event[i] = event;
- }
- UNLOCK (&priv->lock);
- }
- break;
- case GF_EVENT_CHILD_CONNECTING:
- {
- // 'CONNECTING' doesn't ensure its CHILD_UP, so do nothing
- goto out;
- }
- case GF_EVENT_CHILD_DOWN:
- {
- /* get an index number to set */
- for (i = 0; i < priv->child_count; i++) {
- if (data == priv->xl_array[i])
- break;
- }
-
- if (priv->child_count == i) {
- gf_log (this->name, GF_LOG_ERROR,
- "got GF_EVENT_CHILD_DOWN bad subvolume %s",
- data? ((xlator_t *)data)->name: NULL);
- break;
- }
-
- LOCK (&priv->lock);
- {
- if (data == FIRST_CHILD (this))
- priv->first_child_down = 1;
- priv->last_event[i] = event;
- }
- UNLOCK (&priv->lock);
- }
- break;
-
- default:
- {
- /* */
- default_notify (this, event, data);
- goto out;
- }
- break;
- }
-
- // Consider child as down if it's last_event is not CHILD_UP
- for (i = 0, down_client = 0; i < priv->child_count; i++)
- if (priv->last_event[i] != GF_EVENT_CHILD_UP)
- down_client++;
-
- LOCK (&priv->lock);
- {
- priv->nodes_down = down_client;
- }
- UNLOCK (&priv->lock);
-
- heard_from_all_children = _gf_true;
- for (i = 0; i < priv->child_count; i++)
- if (!priv->last_event[i])
- heard_from_all_children = _gf_false;
-
- if (heard_from_all_children)
- default_notify (this, event, data);
-out:
- return 0;
-}
-
-int
-stripe_setxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno, dict_t *xdata)
-{
- int ret = -1;
- int call_cnt = 0;
- stripe_local_t *local = NULL;
-
- if (!frame || !frame->local || !this) {
- gf_log ("", GF_LOG_ERROR, "Possible NULL deref");
- return ret;
- }
-
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- call_cnt = --local->wind_count;
-
- /**
- * We overwrite ->op_* values here for subsequent faliure
- * conditions, hence we propagate the last errno down the
- * stack.
- */
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto unlock;
- }
- }
-
- unlock:
- UNLOCK (&frame->lock);
-
- if (!call_cnt) {
- STRIPE_STACK_UNWIND (setxattr, frame, local->op_ret,
- local->op_errno, xdata);
- }
-
- return 0;
-}
-
-#ifdef HAVE_BD_XLATOR
-int
-stripe_is_bd (dict_t *this, char *key, data_t *value, void *data)
-{
- gf_boolean_t *is_bd = data;
-
- if (data == NULL)
- return 0;
-
- if (XATTR_IS_BD (key))
- *is_bd = _gf_true;
-
- return 0;
-}
-
-static gf_boolean_t
-stripe_setxattr_is_bd (dict_t *dict)
-{
- gf_boolean_t is_bd = _gf_false;
-
- if (dict == NULL)
- goto out;
-
- dict_foreach (dict, stripe_is_bd, &is_bd);
-out:
- return is_bd;
-}
-#else
-#define stripe_setxattr_is_bd(dict) _gf_false
-#endif
-
-int
-stripe_setxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *dict, int flags, dict_t *xdata)
-{
- int32_t op_errno = EINVAL;
- xlator_list_t *trav = NULL;
- stripe_private_t *priv = NULL;
- stripe_local_t *local = NULL;
- int i = 0;
- gf_boolean_t is_bd = _gf_false;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- GF_IF_INTERNAL_XATTR_GOTO ("trusted.*stripe*", dict,
- op_errno, err);
-
- priv = this->private;
- trav = this->children;
-
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- frame->local = local;
- local->wind_count = priv->child_count;
- local->op_ret = local->op_errno = 0;
-
- is_bd = stripe_setxattr_is_bd (dict);
-
- /**
- * Set xattrs for directories on all subvolumes. Additionally
- * this power is only given to a special client. Bd xlator
- * also needs xattrs for regular files (ie LVs)
- */
- if (((frame->root->pid == GF_CLIENT_PID_GSYNCD) &&
- IA_ISDIR (loc->inode->ia_type)) || is_bd) {
- for (i = 0; i < priv->child_count; i++, trav = trav->next) {
- STACK_WIND (frame, stripe_setxattr_cbk,
- trav->xlator, trav->xlator->fops->setxattr,
- loc, dict, flags, xdata);
- }
- } else {
- local->wind_count = 1;
- STACK_WIND (frame, stripe_setxattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setxattr,
- loc, dict, flags, xdata);
- }
-
- return 0;
-err:
- STRIPE_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
- return 0;
-}
-
-
-int
-stripe_fsetxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno, dict_t *xdata)
-{
- STRIPE_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, xdata);
- return 0;
-}
-
-
-int
-stripe_is_special_key (dict_t *this,
- char *key,
- data_t *value,
- void *data)
-{
- gf_boolean_t *is_special = NULL;
-
- if (data == NULL) {
- goto out;
- }
-
- is_special = data;
-
- if (XATTR_IS_LOCKINFO (key) || XATTR_IS_BD (key))
- *is_special = _gf_true;
-
-out:
- return 0;
-}
-
-int32_t
-stripe_fsetxattr_everyone_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *xdata)
-{
- int call_count = 0;
- stripe_local_t *local = NULL;
-
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- call_count = --local->wind_count;
-
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- }
- }
- UNLOCK (&frame->lock);
-
- if (call_count == 0) {
- STRIPE_STACK_UNWIND (fsetxattr, frame, local->op_ret,
- local->op_errno, NULL);
- }
- return 0;
-}
-
-int
-stripe_fsetxattr_to_everyone (call_frame_t *frame, xlator_t *this, fd_t *fd,
- dict_t *dict, int flags, dict_t *xdata)
-{
- xlator_list_t *trav = NULL;
- stripe_private_t *priv = NULL;
- int ret = -1;
- stripe_local_t *local = NULL;
-
- priv = this->private;
-
- local = mem_get0 (this->local_pool);
- if (local == NULL) {
- goto out;
- }
-
- frame->local = local;
-
- local->wind_count = priv->child_count;
-
- trav = this->children;
-
- while (trav) {
- STACK_WIND (frame, stripe_fsetxattr_everyone_cbk,
- trav->xlator, trav->xlator->fops->fsetxattr,
- fd, dict, flags, xdata);
- trav = trav->next;
- }
-
- ret = 0;
-out:
- return ret;
-}
-
-static gf_boolean_t
-stripe_fsetxattr_is_special (dict_t *dict)
-{
- gf_boolean_t is_spl = _gf_false;
-
- if (dict == NULL) {
- goto out;
- }
-
- dict_foreach (dict, stripe_is_special_key, &is_spl);
-
-out:
- return is_spl;
-}
-
-int
-stripe_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- dict_t *dict, int flags, dict_t *xdata)
-{
- int32_t op_ret = -1, ret = -1, op_errno = EINVAL;
- gf_boolean_t is_spl = _gf_false;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- GF_IF_INTERNAL_XATTR_GOTO ("trusted.*stripe*", dict,
- op_errno, err);
-
- is_spl = stripe_fsetxattr_is_special (dict);
- if (is_spl) {
- ret = stripe_fsetxattr_to_everyone (frame, this, fd, dict,
- flags, xdata);
- if (ret < 0) {
- op_errno = ENOMEM;
- goto err;
- }
-
- goto out;
- }
-
- STACK_WIND (frame, stripe_fsetxattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetxattr,
- fd, dict, flags, xdata);
-out:
- return 0;
-err:
- STRIPE_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, NULL);
- return 0;
-}
-
-int
-stripe_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- STRIPE_STACK_UNWIND (removexattr, frame, op_ret, op_errno, xdata);
- return 0;
-}
-
-int
-stripe_removexattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
-{
- int32_t op_errno = EINVAL;
-
- VALIDATE_OR_GOTO (this, err);
-
- GF_IF_NATIVE_XATTR_GOTO ("trusted.*stripe*",
- name, op_errno, err);
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (loc, err);
-
- STACK_WIND (frame, stripe_removexattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->removexattr,
- loc, name, xdata);
- return 0;
-err:
- STRIPE_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
- return 0;
-}
-
-
-int
-stripe_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- STRIPE_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, xdata);
- return 0;
-}
-
-int
-stripe_fremovexattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, const char *name, dict_t *xdata)
-{
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- GF_IF_NATIVE_XATTR_GOTO ("trusted.*stripe*",
- name, op_errno, err);
-
- STACK_WIND (frame, stripe_fremovexattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fremovexattr,
- fd, name, xdata);
- return 0;
- err:
- STRIPE_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, xdata);
- return 0;
-}
-
-int32_t
-stripe_readdirp_lookup_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf,
- dict_t *xattr, struct iatt *parent)
-{
- stripe_local_t *local = NULL;
- call_frame_t *main_frame = NULL;
- stripe_local_t *main_local = NULL;
- gf_dirent_t *entry = NULL;
- call_frame_t *prev = NULL;
- int done = 0;
-
- local = frame->local;
- prev = cookie;
-
- entry = local->dirent;
-
- main_frame = local->orig_frame;
- main_local = main_frame->local;
- LOCK (&frame->lock);
- {
-
- local->call_count--;
- if (!local->call_count)
- done = 1;
- if (op_ret == -1) {
- local->op_errno = op_errno;
- local->op_ret = op_ret;
- goto unlock;
- }
-
- if (stripe_ctx_handle(this, prev, local, xattr))
- gf_log(this->name, GF_LOG_ERROR,
- "Error getting fctx info from dict.");
-
- correct_file_size(stbuf, local->fctx, prev);
-
- stripe_iatt_merge (stbuf, &entry->d_stat);
- local->stbuf_blocks += stbuf->ia_blocks;
- }
-unlock:
- UNLOCK(&frame->lock);
-
- if (done) {
- inode_ctx_put (entry->inode, this,
- (uint64_t) (long)local->fctx);
-
- done = 0;
- LOCK (&main_frame->lock);
- {
- main_local->wind_count--;
- if (!main_local->wind_count)
- done = 1;
- if (local->op_ret == -1) {
- main_local->op_errno = local->op_errno;
- main_local->op_ret = local->op_ret;
- }
- entry->d_stat.ia_blocks = local->stbuf_blocks;
- }
- UNLOCK (&main_frame->lock);
- if (done) {
- main_frame->local = NULL;
- STRIPE_STACK_UNWIND (readdir, main_frame,
- main_local->op_ret,
- main_local->op_errno,
- &main_local->entries, NULL);
- gf_dirent_free (&main_local->entries);
- stripe_local_wipe (main_local);
- mem_put (main_local);
- }
- frame->local = NULL;
- stripe_local_wipe (local);
- mem_put (local);
- STRIPE_STACK_DESTROY (frame);
- }
-
- return 0;
-}
-
-int32_t
-stripe_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- gf_dirent_t *orig_entries, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- call_frame_t *prev = NULL;
- gf_dirent_t *local_entry = NULL;
- gf_dirent_t *tmp_entry = NULL;
- xlator_list_t *trav = NULL;
- loc_t loc = {0, };
- int32_t count = 0;
- stripe_private_t *priv = NULL;
- int32_t subvols = 0;
- dict_t *xattrs = NULL;
- call_frame_t *local_frame = NULL;
- stripe_local_t *local_ent = NULL;
-
- if (!this || !frame || !frame->local || !cookie) {
- gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
- goto out;
- }
- prev = cookie;
- local = frame->local;
- trav = this->children;
- priv = this->private;
-
- subvols = priv->child_count;
-
- LOCK (&frame->lock);
- {
- local->op_errno = op_errno;
- local->op_ret = op_ret;
-
- if (op_ret != -1) {
- list_splice_init (&orig_entries->list,
- &local->entries.list);
- local->wind_count = op_ret;
- }
-
- }
- UNLOCK (&frame->lock);
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_WARNING, "%s returned error %s",
- prev->this->name, strerror (op_errno));
- goto out;
- }
-
- xattrs = dict_new ();
- if (xattrs)
- (void) stripe_xattr_request_build (this, xattrs, 0, 0, 0, 0);
- count = op_ret;
- list_for_each_entry_safe (local_entry, tmp_entry,
- (&local->entries.list), list) {
-
- if (!local_entry)
- break;
- if (!IA_ISREG (local_entry->d_stat.ia_type) || !local_entry->inode) {
- LOCK (&frame->lock);
- {
- local->wind_count--;
- count = local->wind_count;
- }
- UNLOCK (&frame->lock);
- continue;
- }
-
- local_frame = copy_frame (frame);
-
- if (!local_frame) {
- op_errno = ENOMEM;
- op_ret = -1;
- goto out;
- }
-
- local_ent = mem_get0 (this->local_pool);
- if (!local_ent) {
- op_errno = ENOMEM;
- op_ret = -1;
- goto out;
- }
-
- loc.inode = inode_ref (local_entry->inode);
-
- gf_uuid_copy (loc.gfid, local_entry->d_stat.ia_gfid);
-
- local_ent->orig_frame = frame;
-
- local_ent->call_count = subvols;
-
- local_ent->dirent = local_entry;
-
- local_frame->local = local_ent;
-
- trav = this->children;
- while (trav) {
- STACK_WIND (local_frame, stripe_readdirp_lookup_cbk,
- trav->xlator, trav->xlator->fops->lookup,
- &loc, xattrs);
- trav = trav->next;
- }
- loc_wipe (&loc);
- }
-out:
- if (!count) {
- /* all entries are directories */
- frame->local = NULL;
- STRIPE_STACK_UNWIND (readdir, frame,
- (local ? local->op_ret : -1),
- (local ? local->op_errno : EINVAL),
- (local ? &local->entries : NULL),
- NULL);
- gf_dirent_free (&local->entries);
- stripe_local_wipe (local);
- mem_put (local);
- }
- if (xattrs)
- dict_unref (xattrs);
- return 0;
-
-}
-int32_t
-stripe_readdirp (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t off, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- xlator_list_t *trav = NULL;
- int op_errno = -1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- priv = this->private;
- trav = this->children;
-
- if (priv->first_child_down) {
- op_errno = ENOTCONN;
- goto err;
- }
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- frame->local = local;
-
- local->fd = fd_ref (fd);
-
- local->wind_count = 0;
-
- local->count = 0;
- local->op_ret = -1;
- INIT_LIST_HEAD(&local->entries);
-
- if (!trav)
- goto err;
-
- STACK_WIND (frame, stripe_readdirp_cbk, trav->xlator,
- trav->xlator->fops->readdirp, fd, size, off, xdata);
- return 0;
-err:
- op_errno = (op_errno == -1) ? errno : op_errno;
- STRIPE_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
-
- return 0;
-
-}
-
-int32_t
-mem_acct_init (xlator_t *this)
-{
- int ret = -1;
-
- if (!this)
- goto out;
-
- ret = xlator_mem_acct_init (this, gf_stripe_mt_end + 1);
-
- if (ret != 0) {
- gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
- "failed");
- goto out;
- }
-
-out:
- return ret;
-}
-
-static int
-clear_pattern_list (stripe_private_t *priv)
-{
- struct stripe_options *prev = NULL;
- struct stripe_options *trav = NULL;
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO ("stripe", priv, out);
-
- trav = priv->pattern;
- priv->pattern = NULL;
- while (trav) {
- prev = trav;
- trav = trav->next;
- GF_FREE (prev);
- }
-
- ret = 0;
- out:
- return ret;
-
-
-}
-
-
-int
-reconfigure (xlator_t *this, dict_t *options)
-{
-
- stripe_private_t *priv = NULL;
- data_t *data = NULL;
- int ret = -1;
- volume_option_t *opt = NULL;
-
- GF_ASSERT (this);
- GF_ASSERT (this->private);
-
- priv = this->private;
-
-
- ret = 0;
- LOCK (&priv->lock);
- {
- ret = clear_pattern_list (priv);
- if (ret)
- goto unlock;
-
- data = dict_get (options, "block-size");
- if (data) {
- ret = set_stripe_block_size (this, priv, data->data);
- if (ret)
- goto unlock;
- } else {
- opt = xlator_volume_option_get (this, "block-size");
- if (!opt) {
- gf_log (this->name, GF_LOG_WARNING,
- "option 'block-size' not found");
- ret = -1;
- goto unlock;
- }
-
- if (gf_string2bytesize_uint64 (opt->default_value, &priv->block_size)){
- gf_log (this->name, GF_LOG_ERROR,
- "Unable to set default block-size ");
- ret = -1;
- goto unlock;
- }
- }
-
- GF_OPTION_RECONF("coalesce", priv->coalesce, options, bool,
- unlock);
- }
- unlock:
- UNLOCK (&priv->lock);
- if (ret)
- goto out;
-
- ret = 0;
- out:
- return ret;
-
-}
-
-/**
- * init - This function is called when xlator-graph gets initialized.
- * The option given in volfiles are parsed here.
- * @this -
- */
-int32_t
-init (xlator_t *this)
-{
- stripe_private_t *priv = NULL;
- volume_option_t *opt = NULL;
- xlator_list_t *trav = NULL;
- data_t *data = NULL;
- int32_t count = 0;
- int ret = -1;
-
- if (!this)
- goto out;
-
- trav = this->children;
- while (trav) {
- count++;
- trav = trav->next;
- }
-
- if (!count) {
- gf_log (this->name, GF_LOG_ERROR,
- "stripe configured without \"subvolumes\" option. "
- "exiting");
- goto out;
- }
-
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "dangling volume. check volfile ");
- }
-
- if (count == 1) {
- gf_log (this->name, GF_LOG_ERROR,
- "stripe configured with only one \"subvolumes\" option."
- " please check the volume. exiting");
- goto out;
- }
-
- priv = GF_CALLOC (1, sizeof (stripe_private_t),
- gf_stripe_mt_stripe_private_t);
-
- if (!priv)
- goto out;
- priv->xl_array = GF_CALLOC (count, sizeof (xlator_t *),
- gf_stripe_mt_xlator_t);
- if (!priv->xl_array)
- goto out;
-
- priv->last_event = GF_CALLOC (count, sizeof (int),
- gf_stripe_mt_int32_t);
- if (!priv->last_event)
- goto out;
-
- priv->child_count = count;
- LOCK_INIT (&priv->lock);
-
- trav = this->children;
- count = 0;
- while (trav) {
- priv->xl_array[count++] = trav->xlator;
- trav = trav->next;
- }
-
- if (count > 256) {
- gf_log (this->name, GF_LOG_ERROR,
- "maximum number of stripe subvolumes supported "
- "is 256");
- goto out;
- }
-
- ret = 0;
- LOCK (&priv->lock);
- {
- opt = xlator_volume_option_get (this, "block-size");
- if (!opt) {
- gf_log (this->name, GF_LOG_WARNING,
- "option 'block-size' not found");
- ret = -1;
- goto unlock;
- }
- if (gf_string2bytesize_uint64 (opt->default_value, &priv->block_size)){
- gf_log (this->name, GF_LOG_ERROR,
- "Unable to set default block-size ");
- ret = -1;
- goto unlock;
- }
- /* option stripe-pattern *avi:1GB,*pdf:16K */
- data = dict_get (this->options, "block-size");
- if (data) {
- ret = set_stripe_block_size (this, priv, data->data);
- if (ret)
- goto unlock;
- }
- }
- unlock:
- UNLOCK (&priv->lock);
- if (ret)
- goto out;
-
- GF_OPTION_INIT ("use-xattr", priv->xattr_supported, bool, out);
- /* notify related */
- priv->nodes_down = priv->child_count;
-
- GF_OPTION_INIT("coalesce", priv->coalesce, bool, out);
-
- this->local_pool = mem_pool_new (stripe_local_t, 128);
- if (!this->local_pool) {
- ret = -1;
- gf_log (this->name, GF_LOG_ERROR,
- "failed to create local_t's memory pool");
- goto out;
- }
-
- this->private = priv;
-
- ret = 0;
-out:
- if (ret) {
- if (priv) {
- GF_FREE (priv->xl_array);
- GF_FREE (priv);
- }
- }
- return ret;
-}
-
-/**
- * fini - Free all the private variables
- * @this -
- */
-void
-fini (xlator_t *this)
-{
- stripe_private_t *priv = NULL;
- struct stripe_options *prev = NULL;
- struct stripe_options *trav = NULL;
-
- if (!this)
- goto out;
-
- priv = this->private;
- if (priv) {
- this->private = NULL;
- GF_FREE (priv->xl_array);
-
- trav = priv->pattern;
- while (trav) {
- prev = trav;
- trav = trav->next;
- GF_FREE (prev);
- }
- GF_FREE (priv->last_event);
- LOCK_DESTROY (&priv->lock);
- GF_FREE (priv);
- }
-
-out:
- return;
-}
-
-int32_t
-stripe_getxattr_unwind (call_frame_t *frame,
- int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
-
-{
- STRIPE_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
- return 0;
-}
-
-int
-stripe_internal_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr,
- dict_t *xdata)
-{
-
- char size_key[256] = {0,};
- char index_key[256] = {0,};
- char count_key[256] = {0,};
- char coalesce_key[256] = {0,};
-
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (frame->local, out);
-
- if (!xattr || (op_ret == -1))
- goto out;
-
- sprintf (size_key, "trusted.%s.stripe-size", this->name);
- sprintf (count_key, "trusted.%s.stripe-count", this->name);
- sprintf (index_key, "trusted.%s.stripe-index", this->name);
- sprintf (coalesce_key, "trusted.%s.stripe-coalesce", this->name);
-
- dict_del (xattr, size_key);
- dict_del (xattr, count_key);
- dict_del (xattr, index_key);
- dict_del (xattr, coalesce_key);
-
-out:
- STRIPE_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr, xdata);
-
- return 0;
-
-}
-
-int
-stripe_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
-{
- int call_cnt = 0;
- stripe_local_t *local = NULL;
-
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (frame->local, out);
-
- local = frame->local;
-
- LOCK (&frame->lock);
- {
- call_cnt = --local->wind_count;
- }
- UNLOCK (&frame->lock);
-
- if (!xattr || (op_ret < 0))
- goto out;
-
- local->op_ret = 0;
-
- if (!local->xattr) {
- local->xattr = dict_ref (xattr);
- } else {
- stripe_aggregate_xattr (local->xattr, xattr);
- }
-
-out:
- if (!call_cnt) {
- STRIPE_STACK_UNWIND (getxattr, frame, local->op_ret, op_errno,
- local->xattr, xdata);
- }
-
- return 0;
-}
-
-int32_t
-stripe_vgetxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- int32_t callcnt = 0;
- int32_t ret = -1;
- long cky = 0;
- void *xattr_val = NULL;
- void *xattr_serz = NULL;
- stripe_xattr_sort_t *xattr = NULL;
- dict_t *stripe_xattr = NULL;
-
- if (!frame || !frame->local || !this) {
- gf_log ("", GF_LOG_ERROR, "Possible NULL deref");
- return ret;
- }
-
- local = frame->local;
- cky = (long) cookie;
-
- if (local->xsel[0] == '\0') {
- gf_log (this->name, GF_LOG_ERROR, "Empty xattr in cbk");
- return ret;
- }
-
- LOCK (&frame->lock);
- {
- callcnt = --local->wind_count;
-
- if (!dict || (op_ret < 0))
- goto out;
-
- if (!local->xattr_list)
- local->xattr_list = (stripe_xattr_sort_t *)
- GF_CALLOC (local->nallocs,
- sizeof (stripe_xattr_sort_t),
- gf_stripe_mt_xattr_sort_t);
-
- if (local->xattr_list) {
- xattr = local->xattr_list + (int32_t) cky;
-
- ret = dict_get_ptr_and_len (dict, local->xsel,
- &xattr_val,
- &xattr->xattr_len);
- if (xattr->xattr_len == 0)
- goto out;
-
- xattr->pos = cky;
- xattr->xattr_value = gf_memdup (xattr_val,
- xattr->xattr_len);
-
- if (xattr->xattr_value != NULL)
- local->xattr_total_len += xattr->xattr_len + 1;
- }
- }
- out:
- UNLOCK (&frame->lock);
-
- if (!callcnt) {
- if (!local->xattr_total_len)
- goto unwind;
-
- stripe_xattr = dict_new ();
- if (!stripe_xattr)
- goto unwind;
-
- /* select filler based on ->xsel */
- if (XATTR_IS_PATHINFO (local->xsel))
- ret = stripe_fill_pathinfo_xattr (this, local,
- (char **)&xattr_serz);
- else if (XATTR_IS_LOCKINFO (local->xsel)) {
- ret = stripe_fill_lockinfo_xattr (this, local,
- &xattr_serz);
- } else {
- gf_log (this->name, GF_LOG_WARNING,
- "Unknown xattr in xattr request");
- goto unwind;
- }
-
- if (!ret) {
- ret = dict_set_dynptr (stripe_xattr, local->xsel,
- xattr_serz,
- local->xattr_total_len);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR,
- "Can't set %s key in dict",
- local->xsel);
- }
-
- unwind:
- /*
- * Among other things, STRIPE_STACK_UNWIND will free "local"
- * for us. That means we can't dereference it afterward.
- * Fortunately, the actual result is in stripe_xattr now, so we
- * can simply clean up before unwinding.
- */
- ret = stripe_free_xattr_str (local);
- GF_FREE (local->xattr_list);
- local->xattr_list = NULL;
-
- STRIPE_STACK_UNWIND (getxattr, frame, op_ret, op_errno,
- stripe_xattr, NULL);
-
- if (stripe_xattr)
- dict_unref (stripe_xattr);
- }
-
- return ret;
-}
-
-int
-stripe_marker_populate_args (call_frame_t *frame, int type, int *gauge,
- xlator_t **subvols)
-{
- xlator_t *this = frame->this;
- stripe_private_t *priv = this->private;
- stripe_local_t *local = frame->local;
- int count = 0;
-
- count = priv->child_count;
- if (MARKER_XTIME_TYPE == type) {
- if (!IA_FILE_OR_DIR (local->loc.inode->ia_type))
- count = 1;
- }
- memcpy (subvols, priv->xl_array, sizeof (*subvols) * count);
-
- return count;
-}
-
-int32_t
-stripe_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- xlator_list_t *trav = NULL;
- stripe_private_t *priv = NULL;
- int32_t op_errno = EINVAL;
- int i = 0;
- int ret = 0;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
- VALIDATE_OR_GOTO (loc->path, err);
- VALIDATE_OR_GOTO (loc->inode, err);
-
- priv = this->private;
- trav = this->children;
-
- /* Initialization */
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- local->op_ret = -1;
- frame->local = local;
- loc_copy (&local->loc, loc);
-
-
- if (name && strncmp (name, QUOTA_SIZE_KEY,
- strlen (QUOTA_SIZE_KEY)) == 0) {
- local->wind_count = priv->child_count;
-
- for (i = 0, trav=this->children; i < priv->child_count; i++,
- trav = trav->next) {
- STACK_WIND (frame, stripe_getxattr_cbk,
- trav->xlator, trav->xlator->fops->getxattr,
- loc, name, xdata);
- }
-
- return 0;
- }
-
- if (name && (XATTR_IS_PATHINFO (name))) {
- if (IA_ISREG (loc->inode->ia_type)) {
- ret = inode_ctx_get (loc->inode, this,
- (uint64_t *) &local->fctx);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR,
- "stripe size unavailable from fctx"
- " relying on pathinfo could lead to"
- " wrong results");
- }
-
- local->nallocs = local->wind_count = priv->child_count;
- (void) strncpy (local->xsel, name, strlen (name));
-
- /**
- * for xattrs that need info from all childs, fill ->xsel
- * as above and call the filler function in cbk based on
- * it
- */
- for (i = 0, trav = this->children; i < priv->child_count; i++,
- trav = trav->next) {
- STACK_WIND_COOKIE (frame, stripe_vgetxattr_cbk,
- (void *) (long) i, trav->xlator,
- trav->xlator->fops->getxattr,
- loc, name, xdata);
- }
-
- return 0;
- }
-
- if (cluster_handle_marker_getxattr (frame, loc, name, priv->vol_uuid,
- stripe_getxattr_unwind,
- stripe_marker_populate_args) == 0)
- return 0;
-
- STACK_WIND (frame, stripe_internal_getxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
-
- return 0;
-
-err:
- STRIPE_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
- return 0;
-}
-
-static gf_boolean_t
-stripe_is_special_xattr (const char *name)
-{
- gf_boolean_t is_spl = _gf_false;
-
- if (!name) {
- goto out;
- }
-
- if (!strncmp (name, GF_XATTR_LOCKINFO_KEY,
- strlen (GF_XATTR_LOCKINFO_KEY))
- || XATTR_IS_PATHINFO (name))
- is_spl = _gf_true;
-out:
- return is_spl;
-}
-
-int32_t
-stripe_fgetxattr_from_everyone (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
-{
- stripe_local_t *local = NULL;
- stripe_private_t *priv = NULL;
- int32_t ret = -1, op_errno = 0;
- int i = 0;
- xlator_list_t *trav = NULL;
-
- priv = this->private;
-
- local = mem_get0 (this->local_pool);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local->op_ret = -1;
- frame->local = local;
-
- strncpy (local->xsel, name, strlen (name));
- local->nallocs = local->wind_count = priv->child_count;
-
- for (i = 0, trav = this->children; i < priv->child_count; i++,
- trav = trav->next) {
- STACK_WIND_COOKIE (frame, stripe_vgetxattr_cbk,
- (void *) (long) i, trav->xlator,
- trav->xlator->fops->fgetxattr,
- fd, name, xdata);
- }
-
- return 0;
-
-err:
- STACK_UNWIND_STRICT (fgetxattr, frame, -1, op_errno, NULL, NULL);
- return ret;
-}
-
-int32_t
-stripe_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
-{
- if (stripe_is_special_xattr (name)) {
- stripe_fgetxattr_from_everyone (frame, this, fd, name, xdata);
- goto out;
- }
-
- STACK_WIND (frame, stripe_internal_getxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
-
-out:
- return 0;
-}
-
-
-
-int32_t
-stripe_priv_dump (xlator_t *this)
-{
- char key[GF_DUMP_MAX_BUF_LEN];
- int i = 0;
- stripe_private_t *priv = NULL;
- int ret = -1;
- struct stripe_options *options = NULL;
-
- GF_VALIDATE_OR_GOTO ("stripe", this, out);
-
- priv = this->private;
- if (!priv)
- goto out;
-
- ret = TRY_LOCK (&priv->lock);
- if (ret != 0)
- goto out;
-
- gf_proc_dump_add_section("xlator.cluster.stripe.%s.priv", this->name);
- gf_proc_dump_write("child_count","%d", priv->child_count);
-
- for (i = 0; i < priv->child_count; i++) {
- sprintf (key, "subvolumes[%d]", i);
- gf_proc_dump_write (key, "%s.%s", priv->xl_array[i]->type,
- priv->xl_array[i]->name);
- }
-
- options = priv->pattern;
- while (options != NULL) {
- gf_proc_dump_write ("path_pattern", "%s", priv->pattern->path_pattern);
- gf_proc_dump_write ("options_block_size", "%ul", options->block_size);
-
- options = options->next;
- }
-
- gf_proc_dump_write ("block_size", "%ul", priv->block_size);
- gf_proc_dump_write ("nodes-down", "%d", priv->nodes_down);
- gf_proc_dump_write ("first-child_down", "%d", priv->first_child_down);
- gf_proc_dump_write ("xattr_supported", "%d", priv->xattr_supported);
-
- UNLOCK (&priv->lock);
-
-out:
- return ret;
-}
-
-struct xlator_fops fops = {
- .stat = stripe_stat,
- .unlink = stripe_unlink,
- .rename = stripe_rename,
- .link = stripe_link,
- .truncate = stripe_truncate,
- .create = stripe_create,
- .open = stripe_open,
- .readv = stripe_readv,
- .writev = stripe_writev,
- .statfs = stripe_statfs,
- .flush = stripe_flush,
- .fsync = stripe_fsync,
- .ftruncate = stripe_ftruncate,
- .fstat = stripe_fstat,
- .mkdir = stripe_mkdir,
- .rmdir = stripe_rmdir,
- .lk = stripe_lk,
- .opendir = stripe_opendir,
- .fsyncdir = stripe_fsyncdir,
- .setattr = stripe_setattr,
- .fsetattr = stripe_fsetattr,
- .lookup = stripe_lookup,
- .mknod = stripe_mknod,
- .setxattr = stripe_setxattr,
- .fsetxattr = stripe_fsetxattr,
- .getxattr = stripe_getxattr,
- .fgetxattr = stripe_fgetxattr,
- .removexattr = stripe_removexattr,
- .fremovexattr = stripe_fremovexattr,
- .readdirp = stripe_readdirp,
- .fallocate = stripe_fallocate,
- .discard = stripe_discard,
- .zerofill = stripe_zerofill,
- .seek = stripe_seek,
-};
-
-struct xlator_cbks cbks = {
- .release = stripe_release,
- .forget = stripe_forget,
-};
-
-struct xlator_dumpops dumpops = {
- .priv = stripe_priv_dump,
-};
-
-struct volume_options options[] = {
- { .key = {"block-size"},
- .type = GF_OPTION_TYPE_SIZE_LIST,
- .default_value = "128KB",
- .min = STRIPE_MIN_BLOCK_SIZE,
- .description = "Size of the stripe unit that would be read "
- "from or written to the striped servers.",
- .op_version = { 1 },
- .tags = { "stripe" },
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
- },
- { .key = {"use-xattr"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "true",
- .description = "handle the stripe without the xattr",
- .tags = { "stripe", "dev-only" },
- .flags = OPT_FLAG_CLIENT_OPT,
- },
- { .key = {"coalesce"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "true",
- .description = "Enable/Disable coalesce mode to flatten striped "
- "files as stored on the server (i.e., eliminate holes "
- "caused by the traditional format).",
- .op_version = { 1 },
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE,
- .tags = { "stripe" },
- },
- { .key = {NULL} },
-};
diff --git a/xlators/cluster/stripe/src/stripe.h b/xlators/cluster/stripe/src/stripe.h
deleted file mode 100644
index 1e2fcb4e659..00000000000
--- a/xlators/cluster/stripe/src/stripe.h
+++ /dev/null
@@ -1,281 +0,0 @@
-/*
- Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-
-#ifndef _STRIPE_H_
-#define _STRIPE_H_
-
-#include "xlator.h"
-#include "logging.h"
-#include "defaults.h"
-#include "common-utils.h"
-#include "compat.h"
-#include "compat-errno.h"
-#include "stripe-mem-types.h"
-#include "libxlator.h"
-#include <fnmatch.h>
-#include <signal.h>
-
-#define STRIPE_PATHINFO_HEADER "STRIPE:"
-#define STRIPE_MIN_BLOCK_SIZE (16*GF_UNIT_KB)
-
-#define STRIPE_STACK_UNWIND(fop, frame, params ...) do { \
- stripe_local_t *__local = NULL; \
- if (frame) { \
- __local = frame->local; \
- frame->local = NULL; \
- } \
- STACK_UNWIND_STRICT (fop, frame, params); \
- if (__local) { \
- stripe_local_wipe(__local); \
- mem_put (__local); \
- } \
- } while (0)
-
-#define STRIPE_STACK_DESTROY(frame) do { \
- stripe_local_t *__local = NULL; \
- __local = frame->local; \
- frame->local = NULL; \
- STACK_DESTROY (frame->root); \
- if (__local) { \
- stripe_local_wipe (__local); \
- mem_put (__local); \
- } \
- } while (0)
-
-#define STRIPE_VALIDATE_FCTX(fctx, label) do { \
- int idx = 0; \
- if (!fctx) { \
- op_errno = EINVAL; \
- goto label; \
- } \
- for (idx = 0; idx < fctx->stripe_count; idx++) { \
- if (!fctx->xl_array[idx]) { \
- gf_log (this->name, GF_LOG_ERROR, \
- "fctx->xl_array[%d] is NULL", \
- idx); \
- op_errno = ESTALE; \
- goto label; \
- } \
- } \
- } while (0)
-
-typedef struct stripe_xattr_sort {
- int pos;
- int xattr_len;
- char *xattr_value;
-} stripe_xattr_sort_t;
-
-/**
- * struct stripe_options : This keeps the pattern and the block-size
- * information, which is used for striping on a file.
- */
-struct stripe_options {
- struct stripe_options *next;
- char path_pattern[256];
- uint64_t block_size;
-};
-
-/**
- * Private structure for stripe translator
- */
-struct stripe_private {
- struct stripe_options *pattern;
- xlator_t **xl_array;
- uint64_t block_size;
- gf_lock_t lock;
- uint8_t nodes_down;
- int8_t first_child_down;
- int *last_event;
- int8_t child_count;
- gf_boolean_t xattr_supported; /* default yes */
- gf_boolean_t coalesce;
- char vol_uuid[UUID_SIZE + 1];
-};
-
-/**
- * Used to keep info about the replies received from readv/writev calls
- */
-struct stripe_replies {
- struct iovec *vector;
- int32_t count; //count of vector
- int32_t op_ret; //op_ret of readv
- int32_t op_errno;
- int32_t requested_size;
- struct iatt stbuf; /* 'stbuf' is also a part of reply */
-};
-
-typedef struct _stripe_fd_ctx {
- off_t stripe_size;
- int stripe_count;
- int stripe_coalesce;
- int static_array;
- xlator_t **xl_array;
-} stripe_fd_ctx_t;
-
-
-/**
- * Local structure to be passed with all the frames in case of STACK_WIND
- */
-struct stripe_local; /* this itself is used inside the structure; */
-
-struct stripe_local {
- struct stripe_local *next;
- call_frame_t *orig_frame;
-
- stripe_fd_ctx_t *fctx;
-
- /* Used by _cbk functions */
- struct iatt stbuf;
- struct iatt pre_buf;
- struct iatt post_buf;
- struct iatt preparent;
- struct iatt postparent;
-
- off_t stbuf_size;
- off_t prebuf_size;
- off_t postbuf_size;
- off_t preparent_size;
- off_t postparent_size;
-
- blkcnt_t stbuf_blocks;
- blkcnt_t prebuf_blocks;
- blkcnt_t postbuf_blocks;
- blkcnt_t preparent_blocks;
- blkcnt_t postparent_blocks;
-
- struct stripe_replies *replies;
- struct statvfs statvfs_buf;
- dir_entry_t *entry;
-
- int8_t revalidate;
- int8_t failed;
- int8_t unwind;
-
- size_t readv_size;
- int32_t entry_count;
- int32_t node_index;
- int32_t call_count;
- int32_t wind_count; /* used instead of child_cound
- in case of read and write */
- int32_t op_ret;
- int32_t op_errno;
- int32_t count;
- int32_t flags;
- char *name;
- inode_t *inode;
-
- loc_t loc;
- loc_t loc2;
-
- mode_t mode;
- dev_t rdev;
- /* For File I/O fops */
- dict_t *xdata;
-
- stripe_xattr_sort_t *xattr_list;
- int32_t xattr_total_len;
- int32_t nallocs;
- char xsel[256];
-
- /* General usage */
- off_t offset;
- off_t stripe_size;
-
- int xattr_self_heal_needed;
- int entry_self_heal_needed;
-
- int8_t *list;
- struct gf_flock lock;
- fd_t *fd;
- void *value;
- struct iobref *iobref;
- gf_dirent_t entries;
- gf_dirent_t *dirent;
- dict_t *xattr;
- uuid_t ia_gfid;
-
- int xflag;
- mode_t umask;
-};
-
-typedef struct stripe_local stripe_local_t;
-typedef struct stripe_private stripe_private_t;
-
-/*
- * Determine the stripe index of a particular frame based on the translator.
- */
-static inline int32_t stripe_get_frame_index(stripe_fd_ctx_t *fctx,
- call_frame_t *prev)
-{
- int32_t i, idx = -1;
-
- for (i = 0; i < fctx->stripe_count; i++) {
- if (fctx->xl_array[i] == prev->this) {
- idx = i;
- break;
- }
- }
-
- return idx;
-}
-
-static inline void stripe_copy_xl_array(xlator_t **dst, xlator_t **src,
- int count)
-{
- int i;
-
- for (i = 0; i < count; i++)
- dst[i] = src[i];
-}
-
-void stripe_local_wipe (stripe_local_t *local);
-int32_t stripe_ctx_handle (xlator_t *this, call_frame_t *prev,
- stripe_local_t *local, dict_t *dict);
-void stripe_aggregate_xattr (dict_t *dst, dict_t *src);
-int32_t stripe_xattr_request_build (xlator_t *this, dict_t *dict,
- uint64_t stripe_size, uint32_t stripe_count,
- uint32_t stripe_index,
- uint32_t stripe_coalesce);
-int32_t stripe_get_matching_bs (const char *path, stripe_private_t *priv);
-int set_stripe_block_size (xlator_t *this, stripe_private_t *priv, char *data);
-int32_t stripe_iatt_merge (struct iatt *from, struct iatt *to);
-int32_t stripe_fill_pathinfo_xattr (xlator_t *this, stripe_local_t *local,
- char **xattr_serz);
-int32_t stripe_free_xattr_str (stripe_local_t *local);
-int32_t stripe_xattr_aggregate (char *buffer, stripe_local_t *local,
- int32_t *total);
-off_t coalesced_offset(off_t offset, uint64_t stripe_size, int stripe_count);
-off_t uncoalesced_size(off_t size, uint64_t stripe_size, int stripe_count,
- int stripe_index);
-int32_t
-stripe_fill_lockinfo_xattr (xlator_t *this, stripe_local_t *local,
- void **xattr_serz);
-
-/*
- * Adjust the size attribute for files if coalesce is enabled.
- */
-static inline void correct_file_size(struct iatt *buf, stripe_fd_ctx_t *fctx,
- call_frame_t *prev)
-{
- int index;
-
- if (!IA_ISREG(buf->ia_type))
- return;
-
- if (!fctx || !fctx->stripe_coalesce)
- return;
-
- index = stripe_get_frame_index(fctx, prev);
- buf->ia_size = uncoalesced_size(buf->ia_size, fctx->stripe_size,
- fctx->stripe_count, index);
-}
-
-#endif /* _STRIPE_H_ */
diff --git a/xlators/debug/delay-gen/src/delay-gen-mem-types.h b/xlators/debug/delay-gen/src/delay-gen-mem-types.h
index 8c54f5f922e..c89a9217193 100644
--- a/xlators/debug/delay-gen/src/delay-gen-mem-types.h
+++ b/xlators/debug/delay-gen/src/delay-gen-mem-types.h
@@ -8,15 +8,14 @@
* cases as published by the Free Software Foundation.
*/
-
#ifndef __DELAY_GEN_MEM_TYPES_H__
#define __DELAY_GEN_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_delay_gen_mem_types_ {
- gf_delay_gen_mt_dg_t = gf_common_mt_end + 1,
- gf_delay_gen_mt_end
+ gf_delay_gen_mt_dg_t = gf_common_mt_end + 1,
+ gf_delay_gen_mt_end
};
#endif /* __DELAY_GEN_MEM_TYPES_H__ */
diff --git a/xlators/debug/delay-gen/src/delay-gen-messages.h b/xlators/debug/delay-gen/src/delay-gen-messages.h
index a9046ca14bf..bc98cec2885 100644
--- a/xlators/debug/delay-gen/src/delay-gen-messages.h
+++ b/xlators/debug/delay-gen/src/delay-gen-messages.h
@@ -11,7 +11,7 @@
#ifndef __DELAY_GEN_MESSAGES_H__
#define __DELAY_GEN_MESSAGES_H__
-#include "glfs-message-id.h"
+#include <glusterfs/glfs-message-id.h>
/* To add new message IDs, append new identifiers at the end of the list.
*
diff --git a/xlators/debug/delay-gen/src/delay-gen.c b/xlators/debug/delay-gen/src/delay-gen.c
index 849ebeaabcd..4698f1fd785 100644
--- a/xlators/debug/delay-gen/src/delay-gen.c
+++ b/xlators/debug/delay-gen/src/delay-gen.c
@@ -8,741 +8,690 @@
* cases as published by the Free Software Foundation.
*/
-
#include "delay-gen.h"
-#define DELAY_GRANULARITY (1 << 20)
+#define DELAY_GRANULARITY (1 << 20)
#define DG_FOP(fop, name, frame, this, args...) \
- do { \
- delay_gen (this, fop); \
- default_##name (frame, this, args); \
- } while (0)
+ do { \
+ delay_gen(this, fop); \
+ default_##name(frame, this, args); \
+ } while (0)
int
-delay_gen (xlator_t *this, int fop)
+delay_gen(xlator_t *this, int fop)
{
- dg_t *dg = this->private;
+ dg_t *dg = this->private;
- if (!dg->enable[fop] || !dg->delay_ppm)
- return 0;
+ if (!dg->enable[fop] || !dg->delay_ppm)
+ return 0;
- if ((rand () % DELAY_GRANULARITY) < dg->delay_ppm)
- usleep (dg->delay_duration);
+ if ((rand() % DELAY_GRANULARITY) < dg->delay_ppm)
+ gf_nanosleep(dg->delay_duration * GF_US_IN_NS);
- return 0;
+ return 0;
}
int32_t
-dg_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
+dg_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_RENAME, rename, frame, this, oldloc, newloc, xdata);
- return 0;
+ DG_FOP(GF_FOP_RENAME, rename, frame, this, oldloc, newloc, xdata);
+ return 0;
}
-
int32_t
-dg_ipc (call_frame_t *frame, xlator_t *this, int32_t op, dict_t *xdata)
+dg_ipc(call_frame_t *frame, xlator_t *this, int32_t op, dict_t *xdata)
{
- DG_FOP (GF_FOP_IPC, ipc, frame, this, op, xdata);
- return 0;
+ DG_FOP(GF_FOP_IPC, ipc, frame, this, op, xdata);
+ return 0;
}
int32_t
-dg_setactivelk (call_frame_t *frame, xlator_t *this, loc_t *loc,
- lock_migration_info_t *locklist, dict_t *xdata)
+dg_setactivelk(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ lock_migration_info_t *locklist, dict_t *xdata)
{
- DG_FOP (GF_FOP_SETACTIVELK, setactivelk, frame, this, loc,
- locklist, xdata);
- return 0;
+ DG_FOP(GF_FOP_SETACTIVELK, setactivelk, frame, this, loc, locklist, xdata);
+ return 0;
}
int32_t
-dg_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+dg_flush(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- DG_FOP (GF_FOP_FLUSH, flush, frame, this, fd, xdata);
- return 0;
+ DG_FOP(GF_FOP_FLUSH, flush, frame, this, fd, xdata);
+ return 0;
}
int32_t
-dg_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t off, dict_t *xdata)
+dg_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *xdata)
{
- DG_FOP (GF_FOP_READDIR, readdir, frame, this, fd, size, off, xdata);
- return 0;
+ DG_FOP(GF_FOP_READDIR, readdir, frame, this, fd, size, off, xdata);
+ return 0;
}
int32_t
-dg_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
- int32_t flags, dict_t *xdata)
+dg_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- DG_FOP (GF_FOP_SETXATTR, setxattr, frame, this, loc, dict, flags,
- xdata);
- return 0;
+ DG_FOP(GF_FOP_SETXATTR, setxattr, frame, this, loc, dict, flags, xdata);
+ return 0;
}
int32_t
-dg_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- dev_t rdev, mode_t umask, dict_t *xdata)
+dg_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *xdata)
{
- DG_FOP (GF_FOP_MKNOD, mknod, frame, this, loc, mode, rdev, umask,
- xdata);
- return 0;
+ DG_FOP(GF_FOP_MKNOD, mknod, frame, this, loc, mode, rdev, umask, xdata);
+ return 0;
}
int32_t
-dg_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
- int32_t flags, dict_t *xdata)
+dg_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- DG_FOP (GF_FOP_FSETXATTR, fsetxattr, frame, this, fd, dict, flags,
- xdata);
- return 0;
+ DG_FOP(GF_FOP_FSETXATTR, fsetxattr, frame, this, fd, dict, flags, xdata);
+ return 0;
}
int32_t
-dg_readv (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, uint32_t flags, dict_t *xdata)
+dg_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- DG_FOP (GF_FOP_READ, readv, frame, this, fd, size, offset, flags,
- xdata);
- return 0;
+ DG_FOP(GF_FOP_READ, readv, frame, this, fd, size, offset, flags, xdata);
+ return 0;
}
int32_t
-dg_inodelk (call_frame_t *frame, xlator_t *this, const char *volume,
- loc_t *loc, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
+dg_inodelk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ int32_t cmd, struct gf_flock *lock, dict_t *xdata)
{
- DG_FOP (GF_FOP_INODELK, inodelk, frame, this, volume, loc, cmd, lock,
- xdata);
- return 0;
+ DG_FOP(GF_FOP_INODELK, inodelk, frame, this, volume, loc, cmd, lock, xdata);
+ return 0;
}
int32_t
-dg_fremovexattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
+dg_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_FREMOVEXATTR, fremovexattr, frame, this, fd, name,
- xdata);
- return 0;
+ DG_FOP(GF_FOP_FREMOVEXATTR, fremovexattr, frame, this, fd, name, xdata);
+ return 0;
}
int32_t
-dg_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- fd_t *fd, dict_t *xdata)
+dg_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
{
- DG_FOP (GF_FOP_OPEN, open, frame, this, loc, flags, fd, xdata);
- return 0;
+ DG_FOP(GF_FOP_OPEN, open, frame, this, loc, flags, fd, xdata);
+ return 0;
}
int32_t
-dg_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+dg_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- DG_FOP (GF_FOP_XATTROP, xattrop, frame, this, loc, flags, dict, xdata);
- return 0;
+ DG_FOP(GF_FOP_XATTROP, xattrop, frame, this, loc, flags, dict, xdata);
+ return 0;
}
int32_t
-dg_entrylk (call_frame_t *frame, xlator_t *this, const char *volume,
- loc_t *loc, const char *basename, entrylk_cmd cmd,
- entrylk_type type, dict_t *xdata)
+dg_entrylk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_ENTRYLK, entrylk, frame, this, volume,
- loc, basename, cmd, type, xdata);
- return 0;
+ DG_FOP(GF_FOP_ENTRYLK, entrylk, frame, this, volume, loc, basename, cmd,
+ type, xdata);
+ return 0;
}
int32_t
-dg_getactivelk (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+dg_getactivelk(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- DG_FOP (GF_FOP_GETACTIVELK, getactivelk, frame, this, loc, xdata);
- return 0;
+ DG_FOP(GF_FOP_GETACTIVELK, getactivelk, frame, this, loc, xdata);
+ return 0;
}
int32_t
-dg_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
- fd_t *fd, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
+dg_finodelk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ int32_t cmd, struct gf_flock *lock, dict_t *xdata)
{
- DG_FOP (GF_FOP_FINODELK, finodelk, frame, this, volume, fd, cmd, lock,
- xdata);
- return 0;
+ DG_FOP(GF_FOP_FINODELK, finodelk, frame, this, volume, fd, cmd, lock,
+ xdata);
+ return 0;
}
int32_t
-dg_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+dg_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- DG_FOP (GF_FOP_CREATE, create, frame, this, loc, flags, mode, umask, fd,
- xdata);
- return 0;
+ DG_FOP(GF_FOP_CREATE, create, frame, this, loc, flags, mode, umask, fd,
+ xdata);
+ return 0;
}
int32_t
-dg_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- size_t len, dict_t *xdata)
+dg_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
{
- DG_FOP (GF_FOP_DISCARD, discard, frame, this, fd, offset, len, xdata);
- return 0;
+ DG_FOP(GF_FOP_DISCARD, discard, frame, this, fd, offset, len, xdata);
+ return 0;
}
int32_t
-dg_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- mode_t umask, dict_t *xdata)
+dg_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- DG_FOP (GF_FOP_MKDIR, mkdir, frame, this, loc, mode, umask, xdata);
- return 0;
+ DG_FOP(GF_FOP_MKDIR, mkdir, frame, this, loc, mode, umask, xdata);
+ return 0;
}
int32_t
-dg_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
- struct gf_flock *lock, dict_t *xdata)
+dg_lk(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
+ struct gf_flock *lock, dict_t *xdata)
{
- DG_FOP (GF_FOP_LK, lk, frame, this, fd, cmd, lock, xdata);
- return 0;
+ DG_FOP(GF_FOP_LK, lk, frame, this, fd, cmd, lock, xdata);
+ return 0;
}
int32_t
-dg_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count, off_t off, uint32_t flags,
- struct iobref *iobref, dict_t *xdata)
+dg_writev(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
+ int32_t count, off_t off, uint32_t flags, struct iobref *iobref,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_WRITE, writev, frame, this, fd,
- vector, count, off, flags, iobref, xdata);
- return 0;
+ DG_FOP(GF_FOP_WRITE, writev, frame, this, fd, vector, count, off, flags,
+ iobref, xdata);
+ return 0;
}
int32_t
-dg_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
- dict_t *xdata)
+dg_access(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_ACCESS, access, frame, this, loc, mask, xdata);
- return 0;
+ DG_FOP(GF_FOP_ACCESS, access, frame, this, loc, mask, xdata);
+ return 0;
}
int32_t
-dg_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+dg_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- DG_FOP (GF_FOP_LOOKUP, lookup, frame, this, loc, xdata);
- return 0;
+ DG_FOP(GF_FOP_LOOKUP, lookup, frame, this, loc, xdata);
+ return 0;
}
-
-
int32_t
-dg_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int32_t flags, dict_t *xdata)
+dg_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_RMDIR, rmdir, frame, this, loc, flags, xdata);
- return 0;
+ DG_FOP(GF_FOP_RMDIR, rmdir, frame, this, loc, flags, xdata);
+ return 0;
}
-
-
int32_t
-dg_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t keep_size,
- off_t offset, size_t len, dict_t *xdata)
+dg_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t keep_size,
+ off_t offset, size_t len, dict_t *xdata)
{
- DG_FOP (GF_FOP_FALLOCATE, fallocate, frame, this, fd, keep_size, offset,
- len, xdata);
- return 0;
+ DG_FOP(GF_FOP_FALLOCATE, fallocate, frame, this, fd, keep_size, offset, len,
+ xdata);
+ return 0;
}
-
-
int32_t
-dg_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+dg_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- DG_FOP (GF_FOP_FSTAT, fstat, frame, this, fd, xdata);
- return 0;
+ DG_FOP(GF_FOP_FSTAT, fstat, frame, this, fd, xdata);
+ return 0;
}
-
-
int32_t
-dg_lease (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct gf_lease *lease, dict_t *xdata)
+dg_lease(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct gf_lease *lease, dict_t *xdata)
{
- DG_FOP (GF_FOP_LEASE, lease, frame, this, loc, lease, xdata);
- return 0;
+ DG_FOP(GF_FOP_LEASE, lease, frame, this, loc, lease, xdata);
+ return 0;
}
-
-
int32_t
-dg_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+dg_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- DG_FOP (GF_FOP_STAT, stat, frame, this, loc, xdata);
- return 0;
+ DG_FOP(GF_FOP_STAT, stat, frame, this, loc, xdata);
+ return 0;
}
-
-
int32_t
-dg_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
- dict_t *xdata)
+dg_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_TRUNCATE, truncate, frame, this, loc, offset, xdata);
- return 0;
+ DG_FOP(GF_FOP_TRUNCATE, truncate, frame, this, loc, offset, xdata);
+ return 0;
}
-
-
int32_t
-dg_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+dg_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, const char *name,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_GETXATTR, getxattr, frame, this, loc, name, xdata);
- return 0;
+ DG_FOP(GF_FOP_GETXATTR, getxattr, frame, this, loc, name, xdata);
+ return 0;
}
-
-
int32_t
-dg_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
- loc_t *loc, mode_t umask, dict_t *xdata)
+dg_symlink(call_frame_t *frame, xlator_t *this, const char *linkpath,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
- DG_FOP (GF_FOP_SYMLINK, symlink, frame, this, linkpath, loc, umask,
- xdata);
- return 0;
+ DG_FOP(GF_FOP_SYMLINK, symlink, frame, this, linkpath, loc, umask, xdata);
+ return 0;
}
-
-
int32_t
-dg_zerofill (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- off_t len, dict_t *xdata)
+dg_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ off_t len, dict_t *xdata)
{
- DG_FOP (GF_FOP_ZEROFILL, zerofill, frame, this, fd, offset, len, xdata);
- return 0;
+ DG_FOP(GF_FOP_ZEROFILL, zerofill, frame, this, fd, offset, len, xdata);
+ return 0;
}
-
-
int32_t
-dg_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
- dict_t *xdata)
+dg_fsyncdir(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_FSYNCDIR, fsyncdir, frame, this, fd, flags, xdata);
- return 0;
+ DG_FOP(GF_FOP_FSYNCDIR, fsyncdir, frame, this, fd, flags, xdata);
+ return 0;
}
-
-
int32_t
-dg_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
- dict_t *xdata)
+dg_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_FGETXATTR, fgetxattr, frame, this, fd, name, xdata);
- return 0;
+ DG_FOP(GF_FOP_FGETXATTR, fgetxattr, frame, this, fd, name, xdata);
+ return 0;
}
-
-
int32_t
-dg_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t off, dict_t *xdata)
+dg_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *xdata)
{
- DG_FOP (GF_FOP_READDIRP, readdirp, frame, this, fd, size, off, xdata);
- return 0;
+ DG_FOP(GF_FOP_READDIRP, readdirp, frame, this, fd, size, off, xdata);
+ return 0;
}
-
-
int32_t
-dg_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
+dg_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_LINK, link, frame, this, oldloc, newloc, xdata);
- return 0;
+ DG_FOP(GF_FOP_LINK, link, frame, this, oldloc, newloc, xdata);
+ return 0;
}
-
-
int32_t
-dg_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
- gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+dg_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- DG_FOP (GF_FOP_FXATTROP, fxattrop, frame, this, fd, flags, dict, xdata);
- return 0;
+ DG_FOP(GF_FOP_FXATTROP, fxattrop, frame, this, fd, flags, dict, xdata);
+ return 0;
}
-
-
int32_t
-dg_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- dict_t *xdata)
+dg_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_FTRUNCATE, ftruncate, frame, this, fd, offset, xdata);
- return 0;
+ DG_FOP(GF_FOP_FTRUNCATE, ftruncate, frame, this, fd, offset, xdata);
+ return 0;
}
-
-
int32_t
-dg_rchecksum (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- int32_t len, dict_t *xdata)
+dg_rchecksum(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ int32_t len, dict_t *xdata)
{
- DG_FOP (GF_FOP_RCHECKSUM, rchecksum, frame, this, fd, offset, len,
- xdata);
- return 0;
+ DG_FOP(GF_FOP_RCHECKSUM, rchecksum, frame, this, fd, offset, len, xdata);
+ return 0;
}
-
-
int32_t
-dg_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- dict_t *xdata)
+dg_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_UNLINK, unlink, frame, this, loc, flags, xdata);
- return 0;
+ DG_FOP(GF_FOP_UNLINK, unlink, frame, this, loc, flags, xdata);
+ return 0;
}
-
-
int32_t
-dg_fentrylk (call_frame_t *frame, xlator_t *this, const char *volume,
- fd_t *fd, const char *basename, entrylk_cmd cmd,
- entrylk_type type, dict_t *xdata)
+dg_fentrylk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_FENTRYLK, fentrylk, frame, this, volume, fd, basename,
- cmd, type, xdata);
- return 0;
+ DG_FOP(GF_FOP_FENTRYLK, fentrylk, frame, this, volume, fd, basename, cmd,
+ type, xdata);
+ return 0;
}
-
-
int32_t
-dg_getspec (call_frame_t *frame, xlator_t *this, const char *key,
- int32_t flags)
+dg_getspec(call_frame_t *frame, xlator_t *this, const char *key, int32_t flags)
{
- DG_FOP (GF_FOP_GETSPEC, getspec, frame, this, key, flags);
- return 0;
+ DG_FOP(GF_FOP_GETSPEC, getspec, frame, this, key, flags);
+ return 0;
}
-
-
int32_t
-dg_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+dg_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *stbuf,
+ int32_t valid, dict_t *xdata)
{
- DG_FOP (GF_FOP_SETATTR, setattr, frame, this, loc, stbuf, valid, xdata);
- return 0;
+ DG_FOP(GF_FOP_SETATTR, setattr, frame, this, loc, stbuf, valid, xdata);
+ return 0;
}
-
-
int32_t
-dg_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
- dict_t *xdata)
+dg_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_FSYNC, fsync, frame, this, fd, flags, xdata);
- return 0;
+ DG_FOP(GF_FOP_FSYNC, fsync, frame, this, fd, flags, xdata);
+ return 0;
}
-
-
int32_t
-dg_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+dg_statfs(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- DG_FOP (GF_FOP_STATFS, statfs, frame, this, loc, xdata);
- return 0;
+ DG_FOP(GF_FOP_STATFS, statfs, frame, this, loc, xdata);
+ return 0;
}
-
-
int32_t
-dg_seek (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- gf_seek_what_t what, dict_t *xdata)
+dg_seek(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ gf_seek_what_t what, dict_t *xdata)
{
- DG_FOP (GF_FOP_SEEK, seek, frame, this, fd, offset, what, xdata);
- return 0;
+ DG_FOP(GF_FOP_SEEK, seek, frame, this, fd, offset, what, xdata);
+ return 0;
}
-
-
int32_t
-dg_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+dg_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
+ int32_t valid, dict_t *xdata)
{
- DG_FOP (GF_FOP_FSETATTR, fsetattr, frame, this, fd,
- stbuf, valid, xdata);
- return 0;
+ DG_FOP(GF_FOP_FSETATTR, fsetattr, frame, this, fd, stbuf, valid, xdata);
+ return 0;
}
-
-
int32_t
-dg_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
- dict_t *xdata)
+dg_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_OPENDIR, opendir, frame, this, loc, fd, xdata);
- return 0;
+ DG_FOP(GF_FOP_OPENDIR, opendir, frame, this, loc, fd, xdata);
+ return 0;
}
-
-
int32_t
-dg_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
- dict_t *xdata)
+dg_readlink(call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+ dict_t *xdata)
{
- DG_FOP (GF_FOP_READLINK, readlink, frame, this, loc, size, xdata);
- return 0;
+ DG_FOP(GF_FOP_READLINK, readlink, frame, this, loc, size, xdata);
+ return 0;
}
-
-
int32_t
-dg_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+dg_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- DG_FOP (GF_FOP_REMOVEXATTR, removexattr, frame, this, loc, name, xdata);
- return 0;
+ DG_FOP(GF_FOP_REMOVEXATTR, removexattr, frame, this, loc, name, xdata);
+ return 0;
}
int32_t
-dg_forget (xlator_t *this, inode_t *inode)
+dg_forget(xlator_t *this, inode_t *inode)
{
- return 0;
+ return 0;
}
int32_t
-dg_release (xlator_t *this, fd_t *fd)
+dg_release(xlator_t *this, fd_t *fd)
{
- return 0;
+ return 0;
}
int32_t
-dg_releasedir (xlator_t *this, fd_t *fd)
+dg_releasedir(xlator_t *this, fd_t *fd)
{
- return 0;
+ return 0;
}
static int
-delay_gen_parse_fill_fops (dg_t *dg, char *enable_fops)
-{
- char *op_no_str = NULL;
- int op_no = -1;
- int i = 0;
- int ret = 0;
- xlator_t *this = THIS;
- char *saveptr = NULL;
- char *dup_enable_fops = NULL;
-
- if (strlen (enable_fops) == 0) {
- for (i = GF_FOP_NULL + 1; i < GF_FOP_MAXVALUE; i++)
- dg->enable[i] = 1;
- } else {
- dup_enable_fops = gf_strdup (enable_fops);
- if (!dup_enable_fops) {
- ret = -1;
- goto out;
- }
- op_no_str = strtok_r (dup_enable_fops, ",", &saveptr);
- while (op_no_str) {
- op_no = gf_fop_int (op_no_str);
- if (op_no == -1) {
- gf_log (this->name, GF_LOG_WARNING,
- "Wrong option value %s", op_no_str);
- ret = -1;
- goto out;
- } else {
- dg->enable[op_no] = 1;
- }
-
- op_no_str = strtok_r (NULL, ",", &saveptr);
- }
+delay_gen_parse_fill_fops(dg_t *dg, char *enable_fops)
+{
+ char *op_no_str = NULL;
+ int op_no = -1;
+ int i = 0;
+ int ret = 0;
+ xlator_t *this = THIS;
+ char *saveptr = NULL;
+ char *dup_enable_fops = NULL;
+
+ if (strlen(enable_fops) == 0) {
+ for (i = GF_FOP_NULL + 1; i < GF_FOP_MAXVALUE; i++)
+ dg->enable[i] = 1;
+ } else {
+ dup_enable_fops = gf_strdup(enable_fops);
+ if (!dup_enable_fops) {
+ ret = -1;
+ goto out;
}
+ op_no_str = strtok_r(dup_enable_fops, ",", &saveptr);
+ while (op_no_str) {
+ op_no = gf_fop_int(op_no_str);
+ if (op_no == -1) {
+ gf_log(this->name, GF_LOG_WARNING, "Wrong option value %s",
+ op_no_str);
+ ret = -1;
+ goto out;
+ } else {
+ dg->enable[op_no] = 1;
+ }
+
+ op_no_str = strtok_r(NULL, ",", &saveptr);
+ }
+ }
out:
- GF_FREE (dup_enable_fops);
- return ret;
+ GF_FREE(dup_enable_fops);
+ return ret;
}
void
-delay_gen_set_delay_ppm (dg_t *dg, double percent)
+delay_gen_set_delay_ppm(dg_t *dg, double percent)
{
- double ppm;
+ double ppm;
- ppm = (percent / 100.0) * (double) DELAY_GRANULARITY;
- dg->delay_ppm = ppm;
+ ppm = (percent / 100.0) * (double)DELAY_GRANULARITY;
+ dg->delay_ppm = ppm;
}
int32_t
-init (xlator_t *this)
+init(xlator_t *this)
{
- dg_t *dg = NULL;
- int32_t ret = 0;
- double delay_percent = 0;
- char *delay_enable_fops = NULL;
+ dg_t *dg = NULL;
+ int32_t ret = 0;
+ double delay_percent = 0;
+ char *delay_enable_fops = NULL;
- if (!this->children || this->children->next) {
- gf_log (this->name, GF_LOG_ERROR,
- "delay-gen not configured with one subvolume");
- ret = -1;
- goto out;
- }
-
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "dangling volume. check volfile ");
- }
+ if (!this->children || this->children->next) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "delay-gen not configured with one subvolume");
+ ret = -1;
+ goto out;
+ }
- dg = GF_CALLOC (1, sizeof (*dg), gf_delay_gen_mt_dg_t);
+ if (!this->parents) {
+ gf_log(this->name, GF_LOG_WARNING, "dangling volume. check volfile ");
+ }
- if (!dg) {
- ret = -1;
- goto out;
- }
+ dg = GF_CALLOC(1, sizeof(*dg), gf_delay_gen_mt_dg_t);
+ if (!dg) {
ret = -1;
+ goto out;
+ }
- GF_OPTION_INIT ("delay-percentage", delay_percent, percent, out);
- GF_OPTION_INIT ("enable", delay_enable_fops, str, out);
- GF_OPTION_INIT ("delay-duration", dg->delay_duration, int32, out);
+ ret = -1;
- delay_gen_set_delay_ppm (dg, delay_percent);
+ GF_OPTION_INIT("delay-percentage", delay_percent, percent, out);
+ GF_OPTION_INIT("enable", delay_enable_fops, str, out);
+ GF_OPTION_INIT("delay-duration", dg->delay_duration, int32, out);
- ret = delay_gen_parse_fill_fops (dg, delay_enable_fops);
- if (ret)
- goto out;
+ delay_gen_set_delay_ppm(dg, delay_percent);
- this->private = dg;
+ ret = delay_gen_parse_fill_fops(dg, delay_enable_fops);
+ if (ret)
+ goto out;
- ret = 0;
+ this->private = dg;
+
+ ret = 0;
out:
- if (ret)
- GF_FREE (dg);
- return ret;
+ if (ret)
+ GF_FREE(dg);
+ return ret;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- GF_FREE (this->private);
+ GF_FREE(this->private);
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
-
- if (!this)
- return ret;
+ int ret = -1;
- ret = xlator_mem_acct_init (this, gf_delay_gen_mt_end + 1);
+ if (!this)
+ return ret;
- if (ret != 0) {
- gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
- " failed");
- return ret;
- }
+ ret = xlator_mem_acct_init(this, gf_delay_gen_mt_end + 1);
+ if (ret != 0) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Memory accounting init"
+ " failed");
return ret;
+ }
+
+ return ret;
}
int32_t
-reconfigure (xlator_t *this, dict_t *dict)
+reconfigure(xlator_t *this, dict_t *dict)
{
- /*At the moment I don't see any need to implement this. In future
- *if this is needed we can add code here.
- */
- return 0;
+ /*At the moment I don't see any need to implement this. In future
+ *if this is needed we can add code here.
+ */
+ return 0;
}
int
-notify (xlator_t *this, int event, void *data, ...)
+notify(xlator_t *this, int event, void *data, ...)
{
- return default_notify (this, event, data);
+ return default_notify(this, event, data);
}
-
struct xlator_fops fops = {
- .rename = dg_rename,
- .ipc = dg_ipc,
- .setactivelk = dg_setactivelk,
- .flush = dg_flush,
- .readdir = dg_readdir,
- .setxattr = dg_setxattr,
- .mknod = dg_mknod,
- .fsetxattr = dg_fsetxattr,
- .readv = dg_readv,
- .inodelk = dg_inodelk,
- .fremovexattr = dg_fremovexattr,
- .open = dg_open,
- .xattrop = dg_xattrop,
- .entrylk = dg_entrylk,
- .getactivelk = dg_getactivelk,
- .finodelk = dg_finodelk,
- .create = dg_create,
- .discard = dg_discard,
- .mkdir = dg_mkdir,
- .lk = dg_lk,
- .writev = dg_writev,
- .access = dg_access,
- .lookup = dg_lookup,
- .rmdir = dg_rmdir,
- .fallocate = dg_fallocate,
- .fstat = dg_fstat,
- .lease = dg_lease,
- .stat = dg_stat,
- .truncate = dg_truncate,
- .getxattr = dg_getxattr,
- .symlink = dg_symlink,
- .zerofill = dg_zerofill,
- .fsyncdir = dg_fsyncdir,
- .fgetxattr = dg_fgetxattr,
- .readdirp = dg_readdirp,
- .link = dg_link,
- .fxattrop = dg_fxattrop,
- .ftruncate = dg_ftruncate,
- .rchecksum = dg_rchecksum,
- .unlink = dg_unlink,
- .fentrylk = dg_fentrylk,
- .getspec = dg_getspec,
- .setattr = dg_setattr,
- .fsync = dg_fsync,
- .statfs = dg_statfs,
- .seek = dg_seek,
- .fsetattr = dg_fsetattr,
- .opendir = dg_opendir,
- .readlink = dg_readlink,
- .removexattr = dg_removexattr,
+ .rename = dg_rename,
+ .ipc = dg_ipc,
+ .setactivelk = dg_setactivelk,
+ .flush = dg_flush,
+ .readdir = dg_readdir,
+ .setxattr = dg_setxattr,
+ .mknod = dg_mknod,
+ .fsetxattr = dg_fsetxattr,
+ .readv = dg_readv,
+ .inodelk = dg_inodelk,
+ .fremovexattr = dg_fremovexattr,
+ .open = dg_open,
+ .xattrop = dg_xattrop,
+ .entrylk = dg_entrylk,
+ .getactivelk = dg_getactivelk,
+ .finodelk = dg_finodelk,
+ .create = dg_create,
+ .discard = dg_discard,
+ .mkdir = dg_mkdir,
+ .lk = dg_lk,
+ .writev = dg_writev,
+ .access = dg_access,
+ .lookup = dg_lookup,
+ .rmdir = dg_rmdir,
+ .fallocate = dg_fallocate,
+ .fstat = dg_fstat,
+ .lease = dg_lease,
+ .stat = dg_stat,
+ .truncate = dg_truncate,
+ .getxattr = dg_getxattr,
+ .symlink = dg_symlink,
+ .zerofill = dg_zerofill,
+ .fsyncdir = dg_fsyncdir,
+ .fgetxattr = dg_fgetxattr,
+ .readdirp = dg_readdirp,
+ .link = dg_link,
+ .fxattrop = dg_fxattrop,
+ .ftruncate = dg_ftruncate,
+ .rchecksum = dg_rchecksum,
+ .unlink = dg_unlink,
+ .fentrylk = dg_fentrylk,
+ .getspec = dg_getspec,
+ .setattr = dg_setattr,
+ .fsync = dg_fsync,
+ .statfs = dg_statfs,
+ .seek = dg_seek,
+ .fsetattr = dg_fsetattr,
+ .opendir = dg_opendir,
+ .readlink = dg_readlink,
+ .removexattr = dg_removexattr,
};
struct xlator_cbks cbks = {
- .forget = dg_forget,
- .release = dg_release,
- .releasedir = dg_releasedir,
+ .forget = dg_forget,
+ .release = dg_release,
+ .releasedir = dg_releasedir,
};
struct volume_options options[] = {
- { .key = {"delay-percentage"},
- .type = GF_OPTION_TYPE_PERCENT,
- .default_value = "10%",
- .description = "Percentage delay of operations when enabled.",
- .op_version = {GD_OP_VERSION_3_13_0},
- .flags = OPT_FLAG_SETTABLE,
- .tags = {"delay-gen"},
- },
-
- { .key = {"delay-duration"},
- .type = GF_OPTION_TYPE_INT,
- .description = "Delay duration in micro seconds",
- .default_value = "100000",
- .op_version = {GD_OP_VERSION_3_13_0},
- .flags = OPT_FLAG_SETTABLE,
- .tags = {"delay-gen"},
- },
-
- { .key = {"enable"},
- .type = GF_OPTION_TYPE_STR,
- .description = "Accepts a string which takes ',' separated fop "
- "strings to denote which fops are enabled for delay",
- .op_version = {GD_OP_VERSION_3_13_0},
- .flags = OPT_FLAG_SETTABLE,
- .tags = {"delay-gen"},
- .default_value = "",
- },
-
- { .key = {NULL} }
+ {
+ .key = {"delay-percentage"},
+ .type = GF_OPTION_TYPE_PERCENT,
+ .default_value = "10%",
+ .description = "Percentage delay of operations when enabled.",
+ .op_version = {GD_OP_VERSION_3_13_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .tags = {"delay-gen"},
+ },
+
+ {
+ .key = {"delay-duration"},
+ .type = GF_OPTION_TYPE_INT,
+ .description = "Delay duration in micro seconds",
+ .default_value = "100000",
+ .op_version = {GD_OP_VERSION_3_13_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .tags = {"delay-gen"},
+ },
+
+ {
+ .key = {"enable"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Accepts a string which takes ',' separated fop "
+ "strings to denote which fops are enabled for delay",
+ .op_version = {GD_OP_VERSION_3_13_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .tags = {"delay-gen"},
+ .default_value = "",
+ },
+
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .notify = notify,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {GD_OP_VERSION_3_12_0},
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "delay-gen",
+ .category = GF_TECH_PREVIEW,
};
diff --git a/xlators/debug/delay-gen/src/delay-gen.h b/xlators/debug/delay-gen/src/delay-gen.h
index e1111c782e2..afa95e5eb2d 100644
--- a/xlators/debug/delay-gen/src/delay-gen.h
+++ b/xlators/debug/delay-gen/src/delay-gen.h
@@ -8,21 +8,20 @@
* cases as published by the Free Software Foundation.
*/
-
#ifndef __DELAY_GEN_H__
#define __DELAY_GEN_H__
#include "delay-gen-mem-types.h"
#include "delay-gen-messages.h"
-#include "glusterfs.h"
-#include "xlator.h"
-#include "defaults.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
typedef struct {
- int enable[GF_FOP_MAXVALUE];
- int op_count;
- int delay_ppm;
- int delay_duration;
+ int enable[GF_FOP_MAXVALUE];
+ int op_count;
+ int delay_ppm;
+ int delay_duration;
} dg_t;
#endif /* __DELAY_GEN_H__ */
diff --git a/xlators/debug/error-gen/src/error-gen-mem-types.h b/xlators/debug/error-gen/src/error-gen-mem-types.h
index f02280535df..b9b713af8fc 100644
--- a/xlators/debug/error-gen/src/error-gen-mem-types.h
+++ b/xlators/debug/error-gen/src/error-gen-mem-types.h
@@ -11,10 +11,10 @@
#ifndef __ERROR_GEN_MEM_TYPES_H__
#define __ERROR_GEN_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_error_gen_mem_types_ {
- gf_error_gen_mt_eg_t = gf_common_mt_end + 1,
- gf_error_gen_mt_end
+ gf_error_gen_mt_eg_t = gf_common_mt_end + 1,
+ gf_error_gen_mt_end
};
#endif
diff --git a/xlators/debug/error-gen/src/error-gen.c b/xlators/debug/error-gen/src/error-gen.c
index 6724fcf4c22..d45655ef4c3 100644
--- a/xlators/debug/error-gen/src/error-gen.c
+++ b/xlators/debug/error-gen/src/error-gen.c
@@ -7,10 +7,10 @@
later), or the GNU General Public License, version 2 (GPLv2), in all
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
+#include <glusterfs/xlator.h>
#include "error-gen.h"
-#include "statedump.h"
-#include "defaults.h"
+#include <glusterfs/statedump.h>
+#include <glusterfs/defaults.h>
/*
* The user can specify an error probability as a float percentage, but we
@@ -26,1688 +26,1638 @@
* something completely different and this number is irrelevant. See error_gen
* for the legacy code.)
*/
-#define FAILURE_GRANULARITY (1 << 20)
+#define FAILURE_GRANULARITY (1 << 20)
sys_error_t error_no_list[] = {
- [GF_FOP_LOOKUP] = { .error_no_count = 4,
- .error_no = {ENOENT,ENOTDIR,
- ENAMETOOLONG,EAGAIN}},
- [GF_FOP_STAT] = { .error_no_count = 7,
- .error_no = {EACCES,EBADF,EFAULT,
- ENAMETOOLONG,ENOENT,
- ENOMEM,ENOTDIR}},
- [GF_FOP_READLINK] = { .error_no_count = 8,
- .error_no = {EACCES,EFAULT,EINVAL,EIO,
- ENAMETOOLONG,ENOENT,ENOMEM,
- ENOTDIR}},
- [GF_FOP_MKNOD] = { .error_no_count = 11,
- .error_no = {EACCES,EEXIST,EFAULT,
- EINVAL,ENAMETOOLONG,
- ENOENT,ENOMEM,ENOSPC,
- ENOTDIR,EPERM,EROFS}},
- [GF_FOP_MKDIR] = { .error_no_count = 10,
- .error_no = {EACCES,EEXIST,EFAULT,
- ENAMETOOLONG,ENOENT,
- ENOMEM,ENOSPC,ENOTDIR,
- EPERM,EROFS}},
- [GF_FOP_UNLINK] = { .error_no_count = 10,
- .error_no = {EACCES,EBUSY,EFAULT,EIO,
- EISDIR,ENAMETOOLONG,
- ENOENT,ENOMEM,ENOTDIR,
- EPERM,EROFS}},
- [GF_FOP_RMDIR] = { .error_no_count = 8,
- .error_no = {EACCES,EBUSY,EFAULT,
- ENOMEM,ENOTDIR,ENOTEMPTY,
- EPERM,EROFS}},
- [GF_FOP_SYMLINK] = { .error_no_count = 11,
- .error_no = {EACCES,EEXIST,EFAULT,EIO,
- ENAMETOOLONG,ENOENT,ENOMEM,
- ENOSPC,ENOTDIR,EPERM,
- EROFS}},
- [GF_FOP_RENAME] = { .error_no_count = 13,
- .error_no = {EACCES,EBUSY,EFAULT,
- EINVAL,EISDIR,EMLINK,
- ENAMETOOLONG,ENOENT,ENOMEM,
- ENOSPC,ENOTDIR,EEXIST,
- EXDEV}},
- [GF_FOP_LINK] = { .error_no_count = 13,
- .error_no = {EACCES,EFAULT,EEXIST,EIO,
- EMLINK,ENAMETOOLONG,
- ENOENT,ENOMEM,ENOSPC,
- ENOTDIR,EPERM,EROFS,
- EXDEV}},
- [GF_FOP_TRUNCATE] = { .error_no_count = 10,
- .error_no = {EACCES,EFAULT,EFBIG,
- EINTR,EINVAL,EIO,EISDIR,
- ENAMETOOLONG,ENOENT,
- EISDIR}},
- [GF_FOP_CREATE] = {.error_no_count = 10,
- .error_no = {EACCES,EEXIST,EFAULT,
- EISDIR,EMFILE,ENAMETOOLONG,
- ENFILE,ENODEV,ENOENT,
- ENODEV}},
- [GF_FOP_OPEN] = { .error_no_count = 10,
- .error_no = {EACCES,EEXIST,EFAULT,
- EISDIR,EMFILE,
- ENAMETOOLONG,ENFILE,
- ENODEV,ENOENT,ENOMEM}},
- [GF_FOP_READ] = { .error_no_count = 5,
- .error_no = {EINVAL,EBADF,EFAULT,EISDIR,
- ENAMETOOLONG}},
- [GF_FOP_WRITE] = { .error_no_count = 7,
- .error_no = {EINVAL,EBADF,EFAULT,EISDIR,
- ENAMETOOLONG,ENOSPC,
- GF_ERROR_SHORT_WRITE}},
- [GF_FOP_STATFS] = {.error_no_count = 10,
- .error_no = {EACCES,EBADF,EFAULT,EINTR,
- EIO,ENAMETOOLONG,ENOENT,
- ENOMEM,ENOSYS,ENOTDIR}},
- [GF_FOP_FLUSH] = { .error_no_count = 5,
- .error_no = {EACCES,EFAULT,
- ENAMETOOLONG,ENOSYS,
- ENOENT}},
- [GF_FOP_FSYNC] = { .error_no_count = 4,
- .error_no = {EBADF,EIO,EROFS,EINVAL}},
- [GF_FOP_SETXATTR] = { .error_no_count = 4,
- .error_no = {EACCES,EBADF,EINTR,
- ENAMETOOLONG}},
- [GF_FOP_GETXATTR] = { .error_no_count = 4,
- .error_no = {EACCES,EBADF,ENAMETOOLONG,
- EINTR}},
- [GF_FOP_REMOVEXATTR] = { .error_no_count = 4,
- .error_no = {EACCES,EBADF,ENAMETOOLONG,
- EINTR}},
- [GF_FOP_FSETXATTR] = { .error_no_count = 4,
- .error_no = {EACCES,EBADF,EINTR,
- ENAMETOOLONG}},
- [GF_FOP_FGETXATTR] = { .error_no_count = 4,
- .error_no = {EACCES,EBADF,ENAMETOOLONG,
- EINTR}},
- [GF_FOP_FREMOVEXATTR] = { .error_no_count = 4,
- .error_no = {EACCES,EBADF,ENAMETOOLONG,
- EINTR}},
- [GF_FOP_OPENDIR] = { .error_no_count = 8,
- .error_no = {EACCES,EEXIST,EFAULT,
- EISDIR,EMFILE,
- ENAMETOOLONG,ENFILE,
- ENODEV}},
- [GF_FOP_READDIR] = { .error_no_count = 5,
- .error_no = {EINVAL,EACCES,EBADF,
- EMFILE,ENOENT}},
- [GF_FOP_READDIRP] = { .error_no_count = 5,
- .error_no = {EINVAL,EACCES,EBADF,
- EMFILE,ENOENT}},
- [GF_FOP_FSYNCDIR] = { .error_no_count = 4,
- .error_no = {EBADF,EIO,EROFS,EINVAL}},
- [GF_FOP_ACCESS] = { .error_no_count = 8,
- .error_no = {EACCES,ENAMETOOLONG,
- ENOENT,ENOTDIR,EROFS,
- EFAULT,EINVAL,EIO}},
- [GF_FOP_FTRUNCATE] = { .error_no_count = 9,
- .error_no = {EACCES,EFAULT,EFBIG,
- EINTR,EINVAL,EIO,EISDIR,
- ENAMETOOLONG,ENOENT}},
- [GF_FOP_FSTAT] = { .error_no_count = 7,
- .error_no = {EACCES,EBADF,EFAULT,
- ENAMETOOLONG,ENOENT,
- ENOMEM,ENOTDIR}},
- [GF_FOP_LK] = { .error_no_count = 4,
- .error_no = {EACCES,EFAULT,ENOENT,
- EINTR}},
- [GF_FOP_XATTROP] = { .error_no_count = 5,
- .error_no = {EACCES,EFAULT,
- ENAMETOOLONG,ENOSYS,
- ENOENT}},
- [GF_FOP_FXATTROP] = { .error_no_count = 4,
- .error_no = {EBADF,EIO,EROFS,EINVAL}},
- [GF_FOP_INODELK] = { .error_no_count = 4,
- .error_no = {EACCES,EBADF,EINTR,
- ENAMETOOLONG}},
- [GF_FOP_FINODELK] = { .error_no_count = 4,
- .error_no = {EACCES,EBADF,EINTR,
- ENAMETOOLONG}},
- [GF_FOP_ENTRYLK] = { .error_no_count = 4,
- .error_no = {EACCES,EBADF,
- ENAMETOOLONG,EINTR}},
- [GF_FOP_FENTRYLK] = { .error_no_count = 10,
- .error_no = {EACCES,EEXIST,EFAULT,
- EISDIR,EMFILE,
- ENAMETOOLONG,ENFILE,
- ENODEV,ENOENT,ENOMEM}},
- [GF_FOP_SETATTR] = {.error_no_count = 11,
- .error_no = {EACCES,EFAULT,EIO,
- ENAMETOOLONG,ENOENT,
- ENOMEM,ENOTDIR,EPERM,
- EROFS,EBADF,EIO}},
- [GF_FOP_FSETATTR] = { .error_no_count = 11,
- .error_no = {EACCES,EFAULT,EIO,
- ENAMETOOLONG,ENOENT,
- ENOMEM,ENOTDIR,EPERM,
- EROFS,EBADF,EIO}},
- [GF_FOP_GETSPEC] = { .error_no_count = 4,
- .error_no = {EACCES,EBADF,ENAMETOOLONG,
- EINTR}}
-};
+ [GF_FOP_LOOKUP] = {.error_no_count = 4,
+ .error_no = {ENOENT, ENOTDIR, ENAMETOOLONG, EAGAIN}},
+ [GF_FOP_STAT] = {.error_no_count = 6,
+ .error_no = {EACCES, EFAULT, ENAMETOOLONG, ENOENT, ENOMEM,
+ ENOTDIR}},
+ [GF_FOP_READLINK] = {.error_no_count = 8,
+ .error_no = {EACCES, EFAULT, EINVAL, EIO, ENAMETOOLONG,
+ ENOENT, ENOMEM, ENOTDIR}},
+ [GF_FOP_MKNOD] = {.error_no_count = 11,
+ .error_no = {EACCES, EEXIST, EFAULT, EINVAL, ENAMETOOLONG,
+ ENOENT, ENOMEM, ENOSPC, ENOTDIR, EPERM,
+ EROFS}},
+ [GF_FOP_MKDIR] = {.error_no_count = 10,
+ .error_no = {EACCES, EEXIST, EFAULT, ENAMETOOLONG, ENOENT,
+ ENOMEM, ENOSPC, ENOTDIR, EPERM, EROFS}},
+ [GF_FOP_UNLINK] = {.error_no_count = 10,
+ .error_no = {EACCES, EBUSY, EFAULT, EIO, EISDIR,
+ ENAMETOOLONG, ENOENT, ENOMEM, ENOTDIR,
+ EPERM, EROFS}},
+ [GF_FOP_RMDIR] = {.error_no_count = 8,
+ .error_no = {EACCES, EBUSY, EFAULT, ENOMEM, ENOTDIR,
+ ENOTEMPTY, EPERM, EROFS}},
+ [GF_FOP_SYMLINK] = {.error_no_count = 11,
+ .error_no = {EACCES, EEXIST, EFAULT, EIO, ENAMETOOLONG,
+ ENOENT, ENOMEM, ENOSPC, ENOTDIR, EPERM,
+ EROFS}},
+ [GF_FOP_RENAME] = {.error_no_count = 13,
+ .error_no = {EACCES, EBUSY, EFAULT, EINVAL, EISDIR,
+ EMLINK, ENAMETOOLONG, ENOENT, ENOMEM,
+ ENOSPC, ENOTDIR, EEXIST, EXDEV}},
+ [GF_FOP_LINK] = {.error_no_count = 13,
+ .error_no = {EACCES, EFAULT, EEXIST, EIO, EMLINK,
+ ENAMETOOLONG, ENOENT, ENOMEM, ENOSPC, ENOTDIR,
+ EPERM, EROFS, EXDEV}},
+ [GF_FOP_TRUNCATE] = {.error_no_count = 10,
+ .error_no = {EACCES, EFAULT, EFBIG, EINTR, EINVAL, EIO,
+ EISDIR, ENAMETOOLONG, ENOENT, EISDIR}},
+ [GF_FOP_CREATE] = {.error_no_count = 10,
+ .error_no = {EACCES, EEXIST, EFAULT, EISDIR, EMFILE,
+ ENAMETOOLONG, ENFILE, ENODEV, ENOENT,
+ ENODEV}},
+ [GF_FOP_OPEN] = {.error_no_count = 10,
+ .error_no = {EACCES, EEXIST, EFAULT, EISDIR, EMFILE,
+ ENAMETOOLONG, ENFILE, ENODEV, ENOENT,
+ ENOMEM}},
+ [GF_FOP_READ] = {.error_no_count = 5,
+ .error_no = {EINVAL, EBADF, EFAULT, EISDIR, ENAMETOOLONG}},
+ [GF_FOP_WRITE] = {.error_no_count = 7,
+ .error_no = {EINVAL, EBADF, EFAULT, EISDIR, ENAMETOOLONG,
+ ENOSPC, GF_ERROR_SHORT_WRITE}},
+ [GF_FOP_STATFS] = {.error_no_count = 9,
+ .error_no = {EACCES, EFAULT, EINTR, EIO, ENAMETOOLONG,
+ ENOENT, ENOMEM, ENOSYS, ENOTDIR}},
+ [GF_FOP_FLUSH] = {.error_no_count = 5,
+ .error_no = {EACCES, EFAULT, ENAMETOOLONG, ENOSYS,
+ ENOENT}},
+ [GF_FOP_FSYNC] = {.error_no_count = 4,
+ .error_no = {EBADF, EIO, EROFS, EINVAL}},
+ [GF_FOP_SETXATTR] = {.error_no_count = 3,
+ .error_no = {EACCES, EINTR, ENAMETOOLONG}},
+ [GF_FOP_GETXATTR] = {.error_no_count = 3,
+ .error_no = {EACCES, ENAMETOOLONG, EINTR}},
+ [GF_FOP_REMOVEXATTR] = {.error_no_count = 3,
+ .error_no = {EACCES, ENAMETOOLONG, EINTR}},
+ [GF_FOP_FSETXATTR] = {.error_no_count = 4,
+ .error_no = {EACCES, EBADF, EINTR, ENAMETOOLONG}},
+ [GF_FOP_FGETXATTR] = {.error_no_count = 4,
+ .error_no = {EACCES, EBADF, ENAMETOOLONG, EINTR}},
+ [GF_FOP_FREMOVEXATTR] = {.error_no_count = 4,
+ .error_no = {EACCES, EBADF, ENAMETOOLONG, EINTR}},
+ [GF_FOP_OPENDIR] = {.error_no_count = 8,
+ .error_no = {EACCES, EEXIST, EFAULT, EISDIR, EMFILE,
+ ENAMETOOLONG, ENFILE, ENODEV}},
+ [GF_FOP_READDIR] = {.error_no_count = 5,
+ .error_no = {EINVAL, EACCES, EBADF, EMFILE, ENOENT}},
+ [GF_FOP_READDIRP] = {.error_no_count = 5,
+ .error_no = {EINVAL, EACCES, EBADF, EMFILE, ENOENT}},
+ [GF_FOP_FSYNCDIR] = {.error_no_count = 4,
+ .error_no = {EBADF, EIO, EROFS, EINVAL}},
+ [GF_FOP_ACCESS] = {.error_no_count = 8,
+ .error_no = {EACCES, ENAMETOOLONG, ENOENT, ENOTDIR,
+ EROFS, EFAULT, EINVAL, EIO}},
+ [GF_FOP_FTRUNCATE] = {.error_no_count = 9,
+ .error_no = {EACCES, EFAULT, EFBIG, EINTR, EINVAL,
+ EIO, EISDIR, ENAMETOOLONG, ENOENT}},
+ [GF_FOP_FSTAT] = {.error_no_count = 7,
+ .error_no = {EACCES, EBADF, EFAULT, ENAMETOOLONG, ENOENT,
+ ENOMEM, ENOTDIR}},
+ [GF_FOP_LK] = {.error_no_count = 4,
+ .error_no = {EACCES, EFAULT, ENOENT, EINTR}},
+ [GF_FOP_XATTROP] = {.error_no_count = 5,
+ .error_no = {EACCES, EFAULT, ENAMETOOLONG, ENOSYS,
+ ENOENT}},
+ [GF_FOP_FXATTROP] = {.error_no_count = 4,
+ .error_no = {EBADF, EIO, EROFS, EINVAL}},
+ [GF_FOP_INODELK] = {.error_no_count = 3,
+ .error_no = {EACCES, EINTR, ENAMETOOLONG}},
+ [GF_FOP_FINODELK] = {.error_no_count = 4,
+ .error_no = {EACCES, EBADF, EINTR, ENAMETOOLONG}},
+ [GF_FOP_ENTRYLK] = {.error_no_count = 3,
+ .error_no = {EACCES, ENAMETOOLONG, EINTR}},
+ [GF_FOP_FENTRYLK] = {.error_no_count = 10,
+ .error_no = {EACCES, EEXIST, EFAULT, EISDIR, EMFILE,
+ ENAMETOOLONG, ENFILE, ENODEV, ENOENT,
+ ENOMEM}},
+ [GF_FOP_SETATTR] = {.error_no_count = 10,
+ .error_no = {EACCES, EFAULT, EIO, ENAMETOOLONG, ENOENT,
+ ENOMEM, ENOTDIR, EPERM, EROFS, EIO}},
+ [GF_FOP_FSETATTR] = {.error_no_count = 11,
+ .error_no = {EACCES, EFAULT, EIO, ENAMETOOLONG, ENOENT,
+ ENOMEM, ENOTDIR, EPERM, EROFS, EBADF,
+ EIO}},
+ [GF_FOP_GETSPEC] = {.error_no_count = 3,
+ .error_no = {EACCES, ENAMETOOLONG, EINTR}}};
int
-generate_rand_no (int op_no)
+generate_rand_no(int op_no)
{
- int rand_no = 0;
+ int rand_no = 0;
+ int error_no_list_size = 0;
+
+ error_no_list_size = sizeof(error_no_list) / sizeof(error_no_list[0]);
- if (op_no < GF_FOP_MAXVALUE)
- /* coverty[DC.WEAK_CRYPTO] */
- rand_no = rand () % error_no_list[op_no].error_no_count;
- return rand_no;
+ if (op_no < error_no_list_size)
+ /* coverity[DC.WEAK_CRYPTO] */
+ rand_no = rand() % error_no_list[op_no].error_no_count;
+ return rand_no;
}
int
-conv_errno_to_int (char **error_no)
+conv_errno_to_int(char **error_no)
{
- if (!strcmp ((*error_no), "ENOENT"))
- return ENOENT;
- else if (!strcmp ((*error_no), "ENOTDIR"))
- return ENOTDIR;
- else if (!strcmp ((*error_no), "ENAMETOOLONG"))
- return ENAMETOOLONG;
- else if (!strcmp ((*error_no), "EACCES"))
- return EACCES;
- else if (!strcmp ((*error_no), "EBADF"))
- return EBADF;
- else if (!strcmp ((*error_no), "EFAULT"))
- return EFAULT;
- else if (!strcmp ((*error_no), "ENOMEM"))
- return ENOMEM;
- else if (!strcmp ((*error_no), "EINVAL"))
- return EINVAL;
- else if (!strcmp ((*error_no), "EIO"))
- return EIO;
- else if (!strcmp ((*error_no), "EEXIST"))
- return EEXIST;
- else if (!strcmp ((*error_no), "ENOSPC"))
- return ENOSPC;
- else if (!strcmp ((*error_no), "EPERM"))
- return EPERM;
- else if (!strcmp ((*error_no), "EROFS"))
- return EROFS;
- else if (!strcmp ((*error_no), "EBUSY"))
- return EBUSY;
- else if (!strcmp ((*error_no), "EISDIR"))
- return EISDIR;
- else if (!strcmp ((*error_no), "ENOTEMPTY"))
- return ENOTEMPTY;
- else if (!strcmp ((*error_no), "EMLINK"))
- return EMLINK;
- else if (!strcmp ((*error_no), "ENODEV"))
- return ENODEV;
- else if (!strcmp ((*error_no), "EXDEV"))
- return EXDEV;
- else if (!strcmp ((*error_no), "EMFILE"))
- return EMFILE;
- else if (!strcmp ((*error_no), "ENFILE"))
- return ENFILE;
- else if (!strcmp ((*error_no), "ENOSYS"))
- return ENOSYS;
- else if (!strcmp ((*error_no), "EINTR"))
- return EINTR;
- else if (!strcmp ((*error_no), "EFBIG"))
- return EFBIG;
- else if (!strcmp((*error_no), "GF_ERROR_SHORT_WRITE"))
- return GF_ERROR_SHORT_WRITE;
- else
- return EAGAIN;
+ if (!strcmp((*error_no), "ENOENT"))
+ return ENOENT;
+ else if (!strcmp((*error_no), "ENOTDIR"))
+ return ENOTDIR;
+ else if (!strcmp((*error_no), "ENAMETOOLONG"))
+ return ENAMETOOLONG;
+ else if (!strcmp((*error_no), "EACCES"))
+ return EACCES;
+ else if (!strcmp((*error_no), "EBADF"))
+ return EBADF;
+ else if (!strcmp((*error_no), "EFAULT"))
+ return EFAULT;
+ else if (!strcmp((*error_no), "ENOMEM"))
+ return ENOMEM;
+ else if (!strcmp((*error_no), "EINVAL"))
+ return EINVAL;
+ else if (!strcmp((*error_no), "EIO"))
+ return EIO;
+ else if (!strcmp((*error_no), "EEXIST"))
+ return EEXIST;
+ else if (!strcmp((*error_no), "ENOSPC"))
+ return ENOSPC;
+ else if (!strcmp((*error_no), "EPERM"))
+ return EPERM;
+ else if (!strcmp((*error_no), "EROFS"))
+ return EROFS;
+ else if (!strcmp((*error_no), "EBUSY"))
+ return EBUSY;
+ else if (!strcmp((*error_no), "EISDIR"))
+ return EISDIR;
+ else if (!strcmp((*error_no), "ENOTEMPTY"))
+ return ENOTEMPTY;
+ else if (!strcmp((*error_no), "EMLINK"))
+ return EMLINK;
+ else if (!strcmp((*error_no), "ENODEV"))
+ return ENODEV;
+ else if (!strcmp((*error_no), "EXDEV"))
+ return EXDEV;
+ else if (!strcmp((*error_no), "EMFILE"))
+ return EMFILE;
+ else if (!strcmp((*error_no), "ENFILE"))
+ return ENFILE;
+ else if (!strcmp((*error_no), "ENOSYS"))
+ return ENOSYS;
+ else if (!strcmp((*error_no), "EINTR"))
+ return EINTR;
+ else if (!strcmp((*error_no), "EFBIG"))
+ return EFBIG;
+ else if (!strcmp((*error_no), "GF_ERROR_SHORT_WRITE"))
+ return GF_ERROR_SHORT_WRITE;
+ else
+ return EAGAIN;
}
int
-error_gen (xlator_t *this, int op_no)
+error_gen(xlator_t *this, int op_no)
{
- eg_t *egp = NULL;
- int count = 0;
- int error_no_int = 0;
- int rand_no = 0;
- int ret = 0;
- gf_boolean_t should_err = _gf_false;
-
- egp = this->private;
-
- if (egp->random_failure) {
- /*
- * I honestly don't know why anyone would use this "feature"
- * but I'll try to preserve its functionality anyway. Without
- * locking twice to update failure_iter_no and egp->op_count
- * separately, then not locking at all to update
- * egp->failure_iter_no. That's not needed for compatibility,
- * and it's abhorrently wrong. I have *some* standards.
- */
- LOCK (&egp->lock);
- {
- count = ++(egp->op_count);
- error_no_int = egp->error_no_int;
- if ((count % egp->failure_iter_no) == 0) {
- egp->op_count = 0;
- /* coverty[DC.WEAK_CRYPTO] */
- egp->failure_iter_no = 3
- + (rand () % GF_UNIVERSAL_ANSWER);
- should_err = _gf_true;
- }
- }
- UNLOCK (&egp->lock);
- } else {
- /*
- * It turns out that rand() is almost universally implemented
- * as a linear congruential PRNG, which is about as cheap as
- * it gets. This gets us real random behavior, including
- * phenomena like streaks and dry spells, with controllable
- * long-term probability, cheaply.
- */
- if ((rand () % FAILURE_GRANULARITY) < egp->failure_iter_no) {
- should_err = _gf_true;
- }
+ eg_t *egp = NULL;
+ int count = 0;
+ int error_no_int = 0;
+ int rand_no = 0;
+ int ret = 0;
+ gf_boolean_t should_err = _gf_false;
+ int error_no_list_size = 0;
+
+ egp = this->private;
+
+ if (egp->random_failure) {
+ /*
+ * I honestly don't know why anyone would use this "feature"
+ * but I'll try to preserve its functionality anyway. Without
+ * locking twice to update failure_iter_no and egp->op_count
+ * separately, then not locking at all to update
+ * egp->failure_iter_no. That's not needed for compatibility,
+ * and it's abhorrently wrong. I have *some* standards.
+ */
+ LOCK(&egp->lock);
+ {
+ count = ++(egp->op_count);
+ error_no_int = egp->error_no_int;
+ if ((count % egp->failure_iter_no) == 0) {
+ egp->op_count = 0;
+ /* coverity[DC.WEAK_CRYPTO] */
+ egp->failure_iter_no = 3 + (rand() % GF_UNIVERSAL_ANSWER);
+ should_err = _gf_true;
+ }
}
-
- if (should_err) {
- if (error_no_int)
- ret = error_no_int;
- else {
- rand_no = generate_rand_no (op_no);
- if (op_no >= GF_FOP_MAXVALUE)
- op_no = 0;
- if (rand_no >= error_no_list[op_no].error_no_count)
- rand_no = 0;
- ret = error_no_list[op_no].error_no[rand_no];
- }
+ UNLOCK(&egp->lock);
+ } else {
+ /*
+ * It turns out that rand() is almost universally implemented
+ * as a linear congruential PRNG, which is about as cheap as
+ * it gets. This gets us real random behavior, including
+ * phenomena like streaks and dry spells, with controllable
+ * long-term probability, cheaply.
+ */
+ if ((rand() % FAILURE_GRANULARITY) < egp->failure_iter_no) {
+ should_err = _gf_true;
}
+ }
+
+ error_no_list_size = sizeof(error_no_list) / sizeof(error_no_list[0]);
+ if (should_err) {
+ if (error_no_int)
+ ret = error_no_int;
+ else {
+ rand_no = generate_rand_no(op_no);
+ if (op_no >= error_no_list_size)
+ op_no = 0;
+ if (rand_no >= error_no_list[op_no].error_no_count)
+ rand_no = 0;
+ ret = error_no_list[op_no].error_no[rand_no];
+ }
+ }
- return ret;
+ return ret;
}
int
-error_gen_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *xdata)
+error_gen_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_LOOKUP];
-
- if (enable)
- op_errno = error_gen (this, GF_FOP_LOOKUP);
-
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (lookup, frame, -1, op_errno, NULL, NULL, NULL,
- NULL);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup,
- loc, xdata);
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
+
+ egp = this->private;
+ enable = egp->enable[GF_FOP_LOOKUP];
+
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_LOOKUP);
+
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(lookup, frame, -1, op_errno, NULL, NULL, NULL,
+ NULL);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->lookup,
+ loc, xdata);
+ return 0;
}
int
-error_gen_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+error_gen_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- egp = this->private;
- enable = egp->enable[GF_FOP_STAT];
+ egp = this->private;
+ enable = egp->enable[GF_FOP_STAT];
- if (enable)
- op_errno = error_gen (this, GF_FOP_STAT);
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_STAT);
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (stat, frame, -1, op_errno, NULL, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->stat, loc, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(stat, frame, -1, op_errno, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->stat,
+ loc, xdata);
+ return 0;
}
int
-error_gen_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+error_gen_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- egp = this->private;
- enable = egp->enable[GF_FOP_SETATTR];
+ egp = this->private;
+ enable = egp->enable[GF_FOP_SETATTR];
- if (enable)
- op_errno = error_gen (this, GF_FOP_SETATTR);
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_SETATTR);
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (setattr, frame, -1, op_errno, NULL, NULL, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setattr,
- loc, stbuf, valid, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(setattr, frame, -1, op_errno, NULL, NULL, xdata);
return 0;
-}
+ }
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->setattr,
+ loc, stbuf, valid, xdata);
+ return 0;
+}
int
-error_gen_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+error_gen_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- egp = this->private;
- enable = egp->enable[GF_FOP_FSETATTR];
+ egp = this->private;
+ enable = egp->enable[GF_FOP_FSETATTR];
- if (enable)
- op_errno = error_gen (this, GF_FOP_FSETATTR);
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_FSETATTR);
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (fsetattr, frame, -1, op_errno, NULL, NULL, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetattr,
- fd, stbuf, valid, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(fsetattr, frame, -1, op_errno, NULL, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->fsetattr,
+ fd, stbuf, valid, xdata);
+ return 0;
}
int
-error_gen_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc,
- off_t offset, dict_t *xdata)
+error_gen_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ off_t offset, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_TRUNCATE];
-
- if (enable)
- op_errno = error_gen (this, GF_FOP_TRUNCATE);
-
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (truncate, frame, -1, op_errno,
- NULL, NULL, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->truncate,
- loc, offset, xdata);
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
+
+ egp = this->private;
+ enable = egp->enable[GF_FOP_TRUNCATE];
+
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_TRUNCATE);
+
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(truncate, frame, -1, op_errno, NULL, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->truncate,
+ loc, offset, xdata);
+ return 0;
}
int
-error_gen_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd,
- off_t offset, dict_t *xdata)
+error_gen_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp =NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_FTRUNCATE];
-
- if (enable)
- op_errno = error_gen (this, GF_FOP_FTRUNCATE);
-
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (ftruncate, frame, -1, op_errno,
- NULL, NULL, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->ftruncate,
- fd, offset, xdata);
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
+
+ egp = this->private;
+ enable = egp->enable[GF_FOP_FTRUNCATE];
+
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_FTRUNCATE);
+
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(ftruncate, frame, -1, op_errno, NULL, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+ return 0;
}
int
-error_gen_access (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int32_t mask, dict_t *xdata)
+error_gen_access(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+ dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_ACCESS];
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- if (enable)
- op_errno = error_gen (this, GF_FOP_ACCESS);
+ egp = this->private;
+ enable = egp->enable[GF_FOP_ACCESS];
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (access, frame, -1, op_errno, xdata);
- return 0;
- }
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_ACCESS);
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->access,
- loc, mask, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(access, frame, -1, op_errno, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->access,
+ loc, mask, xdata);
+ return 0;
}
int
-error_gen_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc,
- size_t size, dict_t *xdata)
+error_gen_readlink(call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+ dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_READLINK];
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- if (enable)
- op_errno = error_gen (this, GF_FOP_READLINK);
+ egp = this->private;
+ enable = egp->enable[GF_FOP_READLINK];
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (readlink, frame, -1, op_errno, NULL, NULL, xdata);
- return 0;
- }
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_READLINK);
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readlink,
- loc, size, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(readlink, frame, -1, op_errno, NULL, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->readlink,
+ loc, size, xdata);
+ return 0;
}
int
-error_gen_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
- mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata)
+error_gen_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_MKNOD];
-
- if (enable)
- op_errno = error_gen (this, GF_FOP_MKNOD);
-
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (mknod, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mknod,
- loc, mode, rdev, umask, xdata);
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
+
+ egp = this->private;
+ enable = egp->enable[GF_FOP_MKNOD];
+
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_MKNOD);
+
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->mknod,
+ loc, mode, rdev, umask, xdata);
+ return 0;
}
int
-error_gen_mkdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata)
+error_gen_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_MKDIR];
-
- if (enable)
- op_errno = error_gen (this, GF_FOP_MKDIR);
-
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (mkdir, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mkdir,
- loc, mode, umask, xdata);
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
+
+ egp = this->private;
+ enable = egp->enable[GF_FOP_MKDIR];
+
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_MKDIR);
+
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->mkdir,
+ loc, mode, umask, xdata);
+ return 0;
}
int
-error_gen_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
- dict_t *xdata)
+error_gen_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_UNLINK];
-
- if (enable)
- op_errno = error_gen (this, GF_FOP_UNLINK);
-
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (unlink, frame, -1, op_errno, NULL, NULL,
- xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->unlink,
- loc, xflag, xdata);
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
+
+ egp = this->private;
+ enable = egp->enable[GF_FOP_UNLINK];
+
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_UNLINK);
+
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(unlink, frame, -1, op_errno, NULL, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->unlink,
+ loc, xflag, xdata);
+ return 0;
}
int
-error_gen_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
- dict_t *xdata)
+error_gen_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_RMDIR];
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- if (enable)
- op_errno = error_gen (this, GF_FOP_RMDIR);
+ egp = this->private;
+ enable = egp->enable[GF_FOP_RMDIR];
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (rmdir, frame, -1, op_errno, NULL, NULL, xdata);
- return 0;
- }
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_RMDIR);
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rmdir,
- loc, flags, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(rmdir, frame, -1, op_errno, NULL, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->rmdir,
+ loc, flags, xdata);
+ return 0;
}
int
-error_gen_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
- loc_t *loc, mode_t umask, dict_t *xdata)
+error_gen_symlink(call_frame_t *frame, xlator_t *this, const char *linkpath,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_SYMLINK];
-
- if (enable)
- op_errno = error_gen (this, GF_FOP_SYMLINK);
-
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (symlink, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL); /* pre & post parent attr */
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->symlink,
- linkpath, loc, umask, xdata);
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
+
+ egp = this->private;
+ enable = egp->enable[GF_FOP_SYMLINK];
+
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_SYMLINK);
+
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(symlink, frame, -1, op_errno, NULL, NULL, NULL,
+ NULL, NULL); /* pre & post parent attr */
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->symlink,
+ linkpath, loc, umask, xdata);
+ return 0;
}
int
-error_gen_rename (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+error_gen_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_RENAME];
-
- if (enable)
- op_errno = error_gen (this, GF_FOP_RENAME);
-
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (rename, frame, -1, op_errno, NULL,
- NULL, NULL, NULL, NULL, NULL);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rename,
- oldloc, newloc, xdata);
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
+
+ egp = this->private;
+ enable = egp->enable[GF_FOP_RENAME];
+
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_RENAME);
+
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL, NULL);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->rename,
+ oldloc, newloc, xdata);
+ return 0;
}
int
-error_gen_link (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+error_gen_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_LINK];
-
- if (enable)
- op_errno = error_gen (this, GF_FOP_LINK);
-
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (link, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->link,
- oldloc, newloc, xdata);
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
+
+ egp = this->private;
+ enable = egp->enable[GF_FOP_LINK];
+
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_LINK);
+
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(link, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->link,
+ oldloc, newloc, xdata);
+ return 0;
}
int
-error_gen_create (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int32_t flags, mode_t mode, mode_t umask, fd_t *fd,
- dict_t *xdata)
+error_gen_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_CREATE];
-
- if (enable)
- op_errno = error_gen (this, GF_FOP_CREATE);
-
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (create, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL, NULL);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->create,
- loc, flags, mode, umask, fd, xdata);
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
+
+ egp = this->private;
+ enable = egp->enable[GF_FOP_CREATE];
+
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_CREATE);
+
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(create, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL, NULL);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->create,
+ loc, flags, mode, umask, fd, xdata);
+ return 0;
}
int
-error_gen_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int32_t flags, fd_t *fd, dict_t *xdata)
+error_gen_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- egp = this->private;
- enable = egp->enable[GF_FOP_OPEN];
+ egp = this->private;
+ enable = egp->enable[GF_FOP_OPEN];
- if (enable)
- op_errno = error_gen (this, GF_FOP_OPEN);
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_OPEN);
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (open, frame, -1, op_errno, NULL, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->open,
- loc, flags, fd, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(open, frame, -1, op_errno, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->open,
+ loc, flags, fd, xdata);
+ return 0;
}
int
-error_gen_readv (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata)
+error_gen_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_READ];
-
- if (enable)
- op_errno = error_gen (this, GF_FOP_READ);
-
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (readv, frame, -1, op_errno, NULL, 0,
- NULL, NULL, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readv,
- fd, size, offset, flags, xdata);
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
+
+ egp = this->private;
+ enable = egp->enable[GF_FOP_READ];
+
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_READ);
+
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(readv, frame, -1, op_errno, NULL, 0, NULL, NULL,
+ xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->readv,
+ fd, size, offset, flags, xdata);
+ return 0;
}
int
-error_gen_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count,
- off_t off, uint32_t flags, struct iobref *iobref, dict_t *xdata)
+error_gen_writev(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int32_t count, off_t off, uint32_t flags,
+ struct iobref *iobref, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
- struct iovec *shortvec = NULL;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_WRITE];
-
- if (enable)
- op_errno = error_gen (this, GF_FOP_WRITE);
-
- if (op_errno == GF_ERROR_SHORT_WRITE) {
-
- /*
- * A short write error returns some value less than what was
- * requested from a write. To simulate this, replace the vector
- * with one half the size;
- */
- shortvec = iov_dup(vector, 1);
- shortvec->iov_len /= 2;
- goto wind;
- } else if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (writev, frame, -1, op_errno, NULL, NULL, xdata);
- return 0;
- }
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
+ struct iovec *shortvec = NULL;
+
+ egp = this->private;
+ enable = egp->enable[GF_FOP_WRITE];
+
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_WRITE);
+
+ if (op_errno == GF_ERROR_SHORT_WRITE) {
+ /*
+ * A short write error returns some value less than what was
+ * requested from a write. To simulate this, replace the vector
+ * with one half the size;
+ */
+ shortvec = iov_dup(vector, 1);
+ shortvec->iov_len /= 2;
+ count = 1;
+ goto wind;
+ } else if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(writev, frame, -1, op_errno, NULL, NULL, xdata);
+ return 0;
+ }
wind:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->writev,
- fd, shortvec?shortvec:vector,
- count, off, flags, iobref, xdata);
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->writev,
+ fd, shortvec ? shortvec : vector, count, off, flags, iobref,
+ xdata);
- if (shortvec)
- GF_FREE (shortvec);
- return 0;
+ if (shortvec)
+ GF_FREE(shortvec);
+ return 0;
}
int
-error_gen_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+error_gen_flush(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_FLUSH];
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- if (enable)
- op_errno = error_gen (this, GF_FOP_FLUSH);
+ egp = this->private;
+ enable = egp->enable[GF_FOP_FLUSH];
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (flush, frame, -1, op_errno, xdata);
- return 0;
- }
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_FLUSH);
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->flush,
- fd, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(flush, frame, -1, op_errno, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->flush,
+ fd, xdata);
+ return 0;
}
int
-error_gen_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags, dict_t *xdata)
+error_gen_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+ dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_FSYNC];
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- if (enable)
- op_errno = error_gen (this, GF_FOP_FSYNC);
+ egp = this->private;
+ enable = egp->enable[GF_FOP_FSYNC];
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (fsync, frame, -1, op_errno, NULL, NULL, xdata);
- return 0;
- }
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_FSYNC);
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsync,
- fd, flags, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(fsync, frame, -1, op_errno, NULL, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->fsync,
+ fd, flags, xdata);
+ return 0;
}
int
-error_gen_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+error_gen_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- egp = this->private;
- enable = egp->enable[GF_FOP_FSTAT];
+ egp = this->private;
+ enable = egp->enable[GF_FOP_FSTAT];
- if (enable)
- op_errno = error_gen (this, GF_FOP_FSTAT);
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_FSTAT);
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (fstat, frame, -1, op_errno, NULL, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fstat,
- fd, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(fstat, frame, -1, op_errno, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->fstat,
+ fd, xdata);
+ return 0;
}
int
-error_gen_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd, dict_t *xdata)
+error_gen_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_OPENDIR];
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- if (enable)
- op_errno = error_gen (this, GF_FOP_OPENDIR);
+ egp = this->private;
+ enable = egp->enable[GF_FOP_OPENDIR];
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (opendir, frame, -1, op_errno, NULL, xdata);
- return 0;
- }
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_OPENDIR);
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->opendir,
- loc, fd, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(opendir, frame, -1, op_errno, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->opendir,
+ loc, fd, xdata);
+ return 0;
}
int
-error_gen_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int32_t flags, dict_t *xdata)
+error_gen_fsyncdir(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+ dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_FSYNCDIR];
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- if (enable)
- op_errno = error_gen (this, GF_FOP_FSYNCDIR);
+ egp = this->private;
+ enable = egp->enable[GF_FOP_FSYNCDIR];
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (fsyncdir, frame, -1, op_errno, xdata);
- return 0;
- }
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_FSYNCDIR);
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsyncdir,
- fd, flags, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(fsyncdir, frame, -1, op_errno, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->fsyncdir,
+ fd, flags, xdata);
+ return 0;
}
int
-error_gen_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+error_gen_statfs(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_STATFS];
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- if (enable)
- op_errno = error_gen (this, GF_FOP_STATFS);
+ egp = this->private;
+ enable = egp->enable[GF_FOP_STATFS];
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (statfs, frame, -1, op_errno, NULL, xdata);
- return 0;
- }
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_STATFS);
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->statfs,
- loc, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(statfs, frame, -1, op_errno, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->statfs,
+ loc, xdata);
+ return 0;
}
int
-error_gen_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *dict, int32_t flags, dict_t *xdata)
+error_gen_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ dict_t *dict, int32_t flags, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- egp = this->private;
- enable = egp->enable[GF_FOP_SETXATTR];
+ egp = this->private;
+ enable = egp->enable[GF_FOP_SETXATTR];
- if (enable)
- op_errno = error_gen (this, GF_FOP_SETXATTR);
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_SETXATTR);
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (setxattr, frame, -1, op_errno, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setxattr,
- loc, dict, flags, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(setxattr, frame, -1, op_errno, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->setxattr,
+ loc, dict, flags, xdata);
+ return 0;
}
int
-error_gen_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+error_gen_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- egp = this->private;
- enable = egp->enable[GF_FOP_GETXATTR];
+ egp = this->private;
+ enable = egp->enable[GF_FOP_GETXATTR];
- if (enable)
- op_errno = error_gen (this, GF_FOP_GETXATTR);
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_GETXATTR);
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (getxattr, frame, -1, op_errno, NULL, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr,
- loc, name, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(getxattr, frame, -1, op_errno, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->getxattr,
+ loc, name, xdata);
+ return 0;
}
int
-error_gen_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- dict_t *dict, int32_t flags, dict_t *xdata)
+error_gen_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_FSETXATTR];
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- if (enable)
- op_errno = error_gen (this, GF_FOP_FSETXATTR);
+ egp = this->private;
+ enable = egp->enable[GF_FOP_FSETXATTR];
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (fsetxattr, frame, -1, op_errno, xdata);
- return 0;
- }
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_FSETXATTR);
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetxattr,
- fd, dict, flags, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(fsetxattr, frame, -1, op_errno, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+ return 0;
}
int
-error_gen_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
+error_gen_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_FGETXATTR];
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- if (enable)
- op_errno = error_gen (this, GF_FOP_FGETXATTR);
+ egp = this->private;
+ enable = egp->enable[GF_FOP_FGETXATTR];
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (fgetxattr, frame, -1, op_errno, NULL, xdata);
- return 0;
- }
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_FGETXATTR);
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fgetxattr,
- fd, name, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(fgetxattr, frame, -1, op_errno, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
+ return 0;
}
int
-error_gen_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+error_gen_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- egp = this->private;
- enable = egp->enable[GF_FOP_XATTROP];
+ egp = this->private;
+ enable = egp->enable[GF_FOP_XATTROP];
- if (enable)
- op_errno = error_gen (this, GF_FOP_XATTROP);
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_XATTROP);
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (xattrop, frame, -1, op_errno, NULL, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->xattrop,
- loc, flags, dict, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(xattrop, frame, -1, op_errno, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->xattrop,
+ loc, flags, dict, xdata);
+ return 0;
}
int
-error_gen_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
- gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+error_gen_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- egp = this->private;
- enable = egp->enable[GF_FOP_FXATTROP];
+ egp = this->private;
+ enable = egp->enable[GF_FOP_FXATTROP];
- if (enable)
- op_errno = error_gen (this, GF_FOP_FXATTROP);
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_FXATTROP);
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (fxattrop, frame, -1, op_errno, NULL, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fxattrop,
- fd, flags, dict, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(fxattrop, frame, -1, op_errno, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->fxattrop,
+ fd, flags, dict, xdata);
+ return 0;
}
int
-error_gen_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+error_gen_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- egp = this->private;
- enable = egp->enable[GF_FOP_REMOVEXATTR];
+ egp = this->private;
+ enable = egp->enable[GF_FOP_REMOVEXATTR];
- if (enable)
- op_errno = error_gen (this, GF_FOP_REMOVEXATTR);
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_REMOVEXATTR);
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (removexattr, frame, -1, op_errno, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->removexattr,
- loc, name, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(removexattr, frame, -1, op_errno, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
+ return 0;
}
int
-error_gen_fremovexattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
+error_gen_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- egp = this->private;
- enable = egp->enable[GF_FOP_FREMOVEXATTR];
+ egp = this->private;
+ enable = egp->enable[GF_FOP_FREMOVEXATTR];
- if (enable)
- op_errno = error_gen (this, GF_FOP_FREMOVEXATTR);
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_FREMOVEXATTR);
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (fremovexattr, frame, -1, op_errno, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fremovexattr,
- fd, name, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(fremovexattr, frame, -1, op_errno, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fremovexattr, fd, name, xdata);
+ return 0;
}
int
-error_gen_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
- struct gf_flock *lock, dict_t *xdata)
+error_gen_lk(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
+ struct gf_flock *lock, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_LK];
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- if (enable)
- op_errno = error_gen (this, GF_FOP_LK);
+ egp = this->private;
+ enable = egp->enable[GF_FOP_LK];
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (lk, frame, -1, op_errno, NULL, xdata);
- return 0;
- }
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_LK);
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lk,
- fd, cmd, lock, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(lk, frame, -1, op_errno, NULL, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->lk, fd,
+ cmd, lock, xdata);
+ return 0;
}
int
-error_gen_inodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, int32_t cmd,
- struct gf_flock *lock, dict_t *xdata)
+error_gen_inodelk(call_frame_t *frame, xlator_t *this, const char *volume,
+ loc_t *loc, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_INODELK];
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- if (enable)
- op_errno = error_gen (this, GF_FOP_INODELK);
+ egp = this->private;
+ enable = egp->enable[GF_FOP_INODELK];
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (inodelk, frame, -1, op_errno, xdata);
- return 0;
- }
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_INODELK);
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->inodelk,
- volume, loc, cmd, lock, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(inodelk, frame, -1, op_errno, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->inodelk,
+ volume, loc, cmd, lock, xdata);
+ return 0;
}
int
-error_gen_finodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, fd_t *fd, int32_t cmd,
- struct gf_flock *lock, dict_t *xdata)
+error_gen_finodelk(call_frame_t *frame, xlator_t *this, const char *volume,
+ fd_t *fd, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_FINODELK];
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- if (enable)
- op_errno = error_gen (this, GF_FOP_FINODELK);
+ egp = this->private;
+ enable = egp->enable[GF_FOP_FINODELK];
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (finodelk, frame, -1, op_errno, xdata);
- return 0;
- }
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_FINODELK);
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->finodelk,
- volume, fd, cmd, lock, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(finodelk, frame, -1, op_errno, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->finodelk,
+ volume, fd, cmd, lock, xdata);
+ return 0;
}
int
-error_gen_entrylk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, const char *basename,
- entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
+error_gen_entrylk(call_frame_t *frame, xlator_t *this, const char *volume,
+ loc_t *loc, const char *basename, entrylk_cmd cmd,
+ entrylk_type type, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_ENTRYLK];
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- if (enable)
- op_errno = error_gen (this, GF_FOP_ENTRYLK);
+ egp = this->private;
+ enable = egp->enable[GF_FOP_ENTRYLK];
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (entrylk, frame, -1, op_errno, xdata);
- return 0;
- }
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_ENTRYLK);
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->entrylk,
- volume, loc, basename, cmd, type, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(entrylk, frame, -1, op_errno, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->entrylk,
+ volume, loc, basename, cmd, type, xdata);
+ return 0;
}
int
-error_gen_fentrylk (call_frame_t *frame, xlator_t *this,
- const char *volume, fd_t *fd, const char *basename,
- entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
+error_gen_fentrylk(call_frame_t *frame, xlator_t *this, const char *volume,
+ fd_t *fd, const char *basename, entrylk_cmd cmd,
+ entrylk_type type, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_FENTRYLK];
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- if (enable)
- op_errno = error_gen (this, GF_FOP_FENTRYLK);
+ egp = this->private;
+ enable = egp->enable[GF_FOP_FENTRYLK];
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (fentrylk, frame, -1, op_errno, xdata);
- return 0;
- }
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_FENTRYLK);
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fentrylk,
- volume, fd, basename, cmd, type, xdata);
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(fentrylk, frame, -1, op_errno, xdata);
return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->fentrylk,
+ volume, fd, basename, cmd, type, xdata);
+ return 0;
}
int
-error_gen_getspec (call_frame_t *frame, xlator_t *this, const char *key,
- int32_t flags)
+error_gen_getspec(call_frame_t *frame, xlator_t *this, const char *key,
+ int32_t flags)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_GETSPEC];
-
- if (enable)
- op_errno = error_gen (this, GF_FOP_GETSPEC);
-
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (getspec, frame, -1, op_errno, NULL);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getspec,
- key, flags);
- return 0;
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
+
+ egp = this->private;
+ enable = egp->enable[GF_FOP_GETSPEC];
+
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_GETSPEC);
+
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(getspec, frame, -1, op_errno, NULL);
+ return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->getspec,
+ key, flags);
+ return 0;
}
int
-error_gen_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
- size_t size, off_t off, dict_t *xdata)
+error_gen_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *xdata)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
-
- egp = this->private;
- enable = egp->enable[GF_FOP_READDIR];
-
- if (enable)
- op_errno = error_gen (this, GF_FOP_READDIR);
-
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (readdir, frame, -1, op_errno, NULL, xdata);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdir,
- fd, size, off, xdata);
- return 0;
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
+
+ egp = this->private;
+ enable = egp->enable[GF_FOP_READDIR];
+
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_READDIR);
+
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(readdir, frame, -1, op_errno, NULL, xdata);
+ return 0;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->readdir,
+ fd, size, off, xdata);
+ return 0;
}
int
-error_gen_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t off, dict_t *dict)
+error_gen_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *dict)
{
- int op_errno = 0;
- eg_t *egp = NULL;
- int enable = 1;
+ int op_errno = 0;
+ eg_t *egp = NULL;
+ int enable = 1;
- egp = this->private;
- enable = egp->enable[GF_FOP_READDIRP];
+ egp = this->private;
+ enable = egp->enable[GF_FOP_READDIRP];
- if (enable)
- op_errno = error_gen (this, GF_FOP_READDIRP);
+ if (enable)
+ op_errno = error_gen(this, GF_FOP_READDIRP);
- if (op_errno) {
- GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
- STACK_UNWIND_STRICT (readdirp, frame, -1, op_errno, NULL, NULL);
- return 0;
- }
+ if (op_errno) {
+ GF_ERROR(this, "unwind(-1, %s)", strerror(op_errno));
+ STACK_UNWIND_STRICT(readdirp, frame, -1, op_errno, NULL, NULL);
+ return 0;
+ }
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdirp,
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->readdirp,
fd, size, off, dict);
- return 0;
+ return 0;
}
static void
-error_gen_set_failure (eg_t *pvt, double percent)
+error_gen_set_failure(eg_t *pvt, double percent)
{
- double ppm;
+ double ppm;
- GF_ASSERT (pvt);
+ GF_ASSERT(pvt);
- ppm = (percent / 100.0) * (double)FAILURE_GRANULARITY;
- pvt->failure_iter_no = (int)ppm;
+ ppm = (percent / 100.0) * (double)FAILURE_GRANULARITY;
+ pvt->failure_iter_no = (int)ppm;
}
static void
-error_gen_parse_fill_fops (eg_t *pvt, char *enable_fops)
+error_gen_parse_fill_fops(eg_t *pvt, char *enable_fops)
{
- char *op_no_str = NULL;
- int op_no = -1;
- int i = 0;
- xlator_t *this = THIS;
- char *saveptr = NULL;
+ char *op_no_str = NULL;
+ int op_no = -1;
+ int i = 0;
+ xlator_t *this = THIS;
+ char *saveptr = NULL;
+
+ GF_ASSERT(pvt);
+ GF_ASSERT(this);
- GF_ASSERT (pvt);
- GF_ASSERT (this);
+ for (i = 0; i < GF_FOP_MAXVALUE; i++)
+ pvt->enable[i] = 0;
+ if (!enable_fops) {
+ gf_log(this->name, GF_LOG_WARNING, "All fops are enabled.");
for (i = 0; i < GF_FOP_MAXVALUE; i++)
- pvt->enable[i] = 0;
-
- if (!enable_fops) {
- gf_log (this->name, GF_LOG_WARNING,
- "All fops are enabled.");
- for (i = 0; i < GF_FOP_MAXVALUE; i++)
- pvt->enable[i] = 1;
- } else {
- op_no_str = strtok_r (enable_fops, ",", &saveptr);
- while (op_no_str) {
- op_no = gf_fop_int (op_no_str);
- if (op_no == -1) {
- gf_log (this->name, GF_LOG_WARNING,
- "Wrong option value %s", op_no_str);
- } else
- pvt->enable[op_no] = 1;
-
- op_no_str = strtok_r (NULL, ",", &saveptr);
- }
+ pvt->enable[i] = 1;
+ } else {
+ op_no_str = strtok_r(enable_fops, ",", &saveptr);
+ while (op_no_str) {
+ op_no = gf_fop_int(op_no_str);
+ if (op_no == -1) {
+ gf_log(this->name, GF_LOG_WARNING, "Wrong option value %s",
+ op_no_str);
+ } else
+ pvt->enable[op_no] = 1;
+
+ op_no_str = strtok_r(NULL, ",", &saveptr);
}
+ }
}
int32_t
-error_gen_priv_dump (xlator_t *this)
+error_gen_priv_dump(xlator_t *this)
{
- char key_prefix[GF_DUMP_MAX_BUF_LEN];
- int ret = -1;
- eg_t *conf = NULL;
+ char key_prefix[GF_DUMP_MAX_BUF_LEN];
+ int ret = -1;
+ eg_t *conf = NULL;
- if (!this)
- goto out;
+ if (!this)
+ goto out;
- conf = this->private;
- if (!conf)
- goto out;
+ conf = this->private;
+ if (!conf)
+ goto out;
- ret = TRY_LOCK(&conf->lock);
- if (ret != 0) {
- return ret;
- }
+ ret = TRY_LOCK(&conf->lock);
+ if (ret != 0) {
+ return ret;
+ }
- gf_proc_dump_add_section("xlator.debug.error-gen.%s.priv", this->name);
- gf_proc_dump_build_key(key_prefix,"xlator.debug.error-gen","%s.priv",
- this->name);
+ gf_proc_dump_add_section("xlator.debug.error-gen.%s.priv", this->name);
+ gf_proc_dump_build_key(key_prefix, "xlator.debug.error-gen", "%s.priv",
+ this->name);
- gf_proc_dump_write("op_count", "%d", conf->op_count);
- gf_proc_dump_write("failure_iter_no", "%d", conf->failure_iter_no);
- gf_proc_dump_write("error_no", "%s", conf->error_no);
- gf_proc_dump_write("random_failure", "%d", conf->random_failure);
+ gf_proc_dump_write("op_count", "%d", conf->op_count);
+ gf_proc_dump_write("failure_iter_no", "%d", conf->failure_iter_no);
+ gf_proc_dump_write("error_no", "%d", conf->error_no_int);
+ gf_proc_dump_write("random_failure", "%d", conf->random_failure);
- UNLOCK(&conf->lock);
+ UNLOCK(&conf->lock);
out:
- return ret;
+ return ret;
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
-
- if (!this)
- return ret;
+ int ret = -1;
- ret = xlator_mem_acct_init (this, gf_error_gen_mt_end + 1);
+ if (!this)
+ return ret;
- if (ret != 0) {
- gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
- " failed");
- return ret;
- }
+ ret = xlator_mem_acct_init(this, gf_error_gen_mt_end + 1);
+ if (ret != 0) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Memory accounting init"
+ " failed");
return ret;
+ }
+
+ return ret;
}
int
-reconfigure (xlator_t *this, dict_t *options)
+reconfigure(xlator_t *this, dict_t *options)
{
- eg_t *pvt = NULL;
- int32_t ret = 0;
- char *error_enable_fops = NULL;
- double failure_percent_dbl = 0.0;
+ eg_t *pvt = NULL;
+ int32_t ret = 0;
+ char *error_enable_fops = NULL;
+ char *error_no = NULL;
+ double failure_percent_dbl = 0.0;
- if (!this || !this->private)
- goto out;
+ if (!this || !this->private)
+ goto out;
- pvt = this->private;
+ pvt = this->private;
- ret = -1;
+ ret = -1;
- GF_OPTION_RECONF ("error-no", pvt->error_no, options, str, out);
+ GF_OPTION_RECONF("error-no", error_no, options, str, out);
- if (pvt->error_no)
- pvt->error_no_int = conv_errno_to_int (&pvt->error_no);
+ if (error_no)
+ pvt->error_no_int = conv_errno_to_int(&error_no);
- GF_OPTION_RECONF ("failure", failure_percent_dbl, options, percent,
- out);
+ GF_OPTION_RECONF("failure", failure_percent_dbl, options, percent, out);
- GF_OPTION_RECONF ("enable", error_enable_fops, options, str, out);
+ GF_OPTION_RECONF("enable", error_enable_fops, options, str, out);
- GF_OPTION_RECONF ("random-failure", pvt->random_failure, options,
- bool, out);
+ GF_OPTION_RECONF("random-failure", pvt->random_failure, options, bool, out);
- error_gen_parse_fill_fops (pvt, error_enable_fops);
- error_gen_set_failure (pvt, failure_percent_dbl);
+ error_gen_parse_fill_fops(pvt, error_enable_fops);
+ error_gen_set_failure(pvt, failure_percent_dbl);
- ret = 0;
+ ret = 0;
out:
- gf_log (this->name, GF_LOG_DEBUG, "reconfigure returning %d", ret);
- return ret;
+ gf_log(this ? this->name : "error-gen", GF_LOG_DEBUG,
+ "reconfigure returning %d", ret);
+ return ret;
}
int
-init (xlator_t *this)
+init(xlator_t *this)
{
- eg_t *pvt = NULL;
- int32_t ret = 0;
- char *error_enable_fops = NULL;
- double failure_percent_dbl = 0.0;
-
- if (!this->children || this->children->next) {
- gf_log (this->name, GF_LOG_ERROR,
- "error-gen not configured with one subvolume");
- ret = -1;
- goto out;
- }
-
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "dangling volume. check volfile ");
- }
-
- pvt = GF_CALLOC (1, sizeof (eg_t), gf_error_gen_mt_eg_t);
+ eg_t *pvt = NULL;
+ int32_t ret = 0;
+ char *error_enable_fops = NULL;
+ char *error_no = NULL;
+ double failure_percent_dbl = 0.0;
+
+ if (!this->children || this->children->next) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "error-gen not configured with one subvolume");
+ ret = -1;
+ goto out;
+ }
- if (!pvt) {
- ret = -1;
- goto out;
- }
+ if (!this->parents) {
+ gf_log(this->name, GF_LOG_WARNING, "dangling volume. check volfile ");
+ }
- LOCK_INIT (&pvt->lock);
+ pvt = GF_CALLOC(1, sizeof(eg_t), gf_error_gen_mt_eg_t);
+ if (!pvt) {
ret = -1;
+ goto out;
+ }
+
+ LOCK_INIT(&pvt->lock);
- GF_OPTION_INIT ("error-no", pvt->error_no, str, out);
+ ret = -1;
- if (pvt->error_no)
- pvt->error_no_int = conv_errno_to_int (&pvt->error_no);
+ GF_OPTION_INIT("error-no", error_no, str, out);
- GF_OPTION_INIT ("failure", failure_percent_dbl, percent, out);
+ if (error_no)
+ pvt->error_no_int = conv_errno_to_int(&error_no);
- GF_OPTION_INIT ("enable", error_enable_fops, str, out);
+ GF_OPTION_INIT("failure", failure_percent_dbl, percent, out);
- GF_OPTION_INIT ("random-failure", pvt->random_failure, bool, out);
+ GF_OPTION_INIT("enable", error_enable_fops, str, out);
+ GF_OPTION_INIT("random-failure", pvt->random_failure, bool, out);
- error_gen_parse_fill_fops (pvt, error_enable_fops);
- error_gen_set_failure (pvt, failure_percent_dbl);
+ error_gen_parse_fill_fops(pvt, error_enable_fops);
+ error_gen_set_failure(pvt, failure_percent_dbl);
- this->private = pvt;
+ this->private = pvt;
- /* Give some seed value here */
- srand (time(NULL));
+ /* Give some seed value here. */
+ srand(gf_time());
- ret = 0;
+ ret = 0;
out:
- if (ret)
- GF_FREE (pvt);
- return ret;
+ if (ret)
+ GF_FREE(pvt);
+ return ret;
}
-
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- eg_t *pvt = NULL;
+ eg_t *pvt = NULL;
- if (!this)
- return;
- pvt = this->private;
-
- if (pvt) {
- LOCK_DESTROY (&pvt->lock);
- GF_FREE (pvt);
- gf_log (this->name, GF_LOG_DEBUG, "fini called");
- }
+ if (!this)
return;
+ pvt = this->private;
+
+ if (pvt) {
+ LOCK_DESTROY(&pvt->lock);
+ GF_FREE(pvt);
+ gf_log(this->name, GF_LOG_DEBUG, "fini called");
+ }
+ return;
}
struct xlator_dumpops dumpops = {
- .priv = error_gen_priv_dump,
+ .priv = error_gen_priv_dump,
};
struct xlator_cbks cbks;
struct xlator_fops fops = {
- .lookup = error_gen_lookup,
- .stat = error_gen_stat,
- .readlink = error_gen_readlink,
- .mknod = error_gen_mknod,
- .mkdir = error_gen_mkdir,
- .unlink = error_gen_unlink,
- .rmdir = error_gen_rmdir,
- .symlink = error_gen_symlink,
- .rename = error_gen_rename,
- .link = error_gen_link,
- .truncate = error_gen_truncate,
- .create = error_gen_create,
- .open = error_gen_open,
- .readv = error_gen_readv,
- .writev = error_gen_writev,
- .statfs = error_gen_statfs,
- .flush = error_gen_flush,
- .fsync = error_gen_fsync,
- .setxattr = error_gen_setxattr,
- .getxattr = error_gen_getxattr,
- .removexattr = error_gen_removexattr,
- .fsetxattr = error_gen_fsetxattr,
- .fgetxattr = error_gen_fgetxattr,
- .fremovexattr = error_gen_fremovexattr,
- .opendir = error_gen_opendir,
- .readdir = error_gen_readdir,
- .readdirp = error_gen_readdirp,
- .fsyncdir = error_gen_fsyncdir,
- .access = error_gen_access,
- .ftruncate = error_gen_ftruncate,
- .fstat = error_gen_fstat,
- .lk = error_gen_lk,
- .xattrop = error_gen_xattrop,
- .fxattrop = error_gen_fxattrop,
- .inodelk = error_gen_inodelk,
- .finodelk = error_gen_finodelk,
- .entrylk = error_gen_entrylk,
- .fentrylk = error_gen_fentrylk,
- .setattr = error_gen_setattr,
- .fsetattr = error_gen_fsetattr,
- .getspec = error_gen_getspec,
+ .lookup = error_gen_lookup,
+ .stat = error_gen_stat,
+ .readlink = error_gen_readlink,
+ .mknod = error_gen_mknod,
+ .mkdir = error_gen_mkdir,
+ .unlink = error_gen_unlink,
+ .rmdir = error_gen_rmdir,
+ .symlink = error_gen_symlink,
+ .rename = error_gen_rename,
+ .link = error_gen_link,
+ .truncate = error_gen_truncate,
+ .create = error_gen_create,
+ .open = error_gen_open,
+ .readv = error_gen_readv,
+ .writev = error_gen_writev,
+ .statfs = error_gen_statfs,
+ .flush = error_gen_flush,
+ .fsync = error_gen_fsync,
+ .setxattr = error_gen_setxattr,
+ .getxattr = error_gen_getxattr,
+ .removexattr = error_gen_removexattr,
+ .fsetxattr = error_gen_fsetxattr,
+ .fgetxattr = error_gen_fgetxattr,
+ .fremovexattr = error_gen_fremovexattr,
+ .opendir = error_gen_opendir,
+ .readdir = error_gen_readdir,
+ .readdirp = error_gen_readdirp,
+ .fsyncdir = error_gen_fsyncdir,
+ .access = error_gen_access,
+ .ftruncate = error_gen_ftruncate,
+ .fstat = error_gen_fstat,
+ .lk = error_gen_lk,
+ .xattrop = error_gen_xattrop,
+ .fxattrop = error_gen_fxattrop,
+ .inodelk = error_gen_inodelk,
+ .finodelk = error_gen_finodelk,
+ .entrylk = error_gen_entrylk,
+ .fentrylk = error_gen_fentrylk,
+ .setattr = error_gen_setattr,
+ .fsetattr = error_gen_fsetattr,
+ .getspec = error_gen_getspec,
};
struct volume_options options[] = {
- { .key = {"failure"},
- .type = GF_OPTION_TYPE_PERCENT,
- .description = "Percentage failure of operations when enabled.",
- },
-
- { .key = {"error-no"},
- .value = {"ENOENT","ENOTDIR","ENAMETOOLONG","EACCES","EBADF",
- "EFAULT","ENOMEM","EINVAL","EIO","EEXIST","ENOSPC",
- "EPERM","EROFS","EBUSY","EISDIR","ENOTEMPTY","EMLINK"
- "ENODEV","EXDEV","EMFILE","ENFILE","ENOSYS","EINTR",
- "EFBIG","EAGAIN","GF_ERROR_SHORT_WRITE"},
- .type = GF_OPTION_TYPE_STR,
- .op_version = {3},
- .tags = {"error-gen"},
- .flags = OPT_FLAG_SETTABLE,
-
- },
-
- { .key = {"random-failure"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .op_version = {3},
- .tags = {"error-gen"},
- .flags = OPT_FLAG_SETTABLE,
- },
-
- { .key = {"enable", "error-fops"},
- .type = GF_OPTION_TYPE_STR,
- .description = "Accepts a string which takes ',' separated fop "
- "strings to denote which fops are enabled for error",
- .op_version = {3},
- .tags = {"error-gen"},
- .flags = OPT_FLAG_SETTABLE,
- },
-
- { .key = {NULL} }
+ {
+ .key = {"failure"},
+ .type = GF_OPTION_TYPE_PERCENT,
+ .description = "Percentage failure of operations when enabled.",
+ },
+
+ {
+ .key = {"error-no"},
+ .value = {"ENOENT",
+ "ENOTDIR",
+ "ENAMETOOLONG",
+ "EACCES",
+ "EBADF",
+ "EFAULT",
+ "ENOMEM",
+ "EINVAL",
+ "EIO",
+ "EEXIST",
+ "ENOSPC",
+ "EPERM",
+ "EROFS",
+ "EBUSY",
+ "EISDIR",
+ "ENOTEMPTY",
+ "EMLINK"
+ "ENODEV",
+ "EXDEV",
+ "EMFILE",
+ "ENFILE",
+ "ENOSYS",
+ "EINTR",
+ "EFBIG",
+ "EAGAIN",
+ "GF_ERROR_SHORT_WRITE"},
+ .type = GF_OPTION_TYPE_STR,
+ .op_version = {3},
+ .tags = {"error-gen"},
+ .flags = OPT_FLAG_SETTABLE,
+
+ },
+
+ {
+ .key = {"random-failure"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .op_version = {3},
+ .tags = {"error-gen"},
+ .flags = OPT_FLAG_SETTABLE,
+ },
+
+ {
+ .key = {"enable", "error-fops"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Accepts a string which takes ',' separated fop "
+ "strings to denote which fops are enabled for error",
+ .op_version = {3},
+ .tags = {"error-gen"},
+ .flags = OPT_FLAG_SETTABLE,
+ },
+
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1},
+ .dumpops = &dumpops,
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "error-gen",
+ .category = GF_TECH_PREVIEW,
};
diff --git a/xlators/debug/error-gen/src/error-gen.h b/xlators/debug/error-gen/src/error-gen.h
index 2cd95de335a..2478cd5b21c 100644
--- a/xlators/debug/error-gen/src/error-gen.h
+++ b/xlators/debug/error-gen/src/error-gen.h
@@ -22,29 +22,28 @@
* range.
*/
enum GF_PSEUDO_ERRORS {
- GF_ERROR_SHORT_WRITE = 1000, /* short writev return value */
- GF_ERROR_MAX
+ GF_ERROR_SHORT_WRITE = 1000, /* short writev return value */
+ GF_ERROR_MAX
};
typedef struct {
- int enable[GF_FOP_MAXVALUE];
- int op_count;
- /*
- * This is only an iteration number in the random-failure case. For
- * the normal controlled-probability case, it's actually a numerator
- * for the failure probability (see FAILURE_GRANULARITY declaration).
- * It's just not worth blowing up the diff by changing it.
- */
- int failure_iter_no;
- char *error_no;
- int error_no_int;
- gf_boolean_t random_failure;
- gf_lock_t lock;
+ int enable[GF_FOP_MAXVALUE];
+ int op_count;
+ /*
+ * This is only an iteration number in the random-failure case. For
+ * the normal controlled-probability case, it's actually a numerator
+ * for the failure probability (see FAILURE_GRANULARITY declaration).
+ * It's just not worth blowing up the diff by changing it.
+ */
+ int failure_iter_no;
+ int error_no_int;
+ gf_boolean_t random_failure;
+ gf_lock_t lock;
} eg_t;
typedef struct {
- int error_no_count;
- int error_no[20];
+ int error_no_count;
+ int error_no[20];
} sys_error_t;
#endif
diff --git a/xlators/debug/io-stats/src/io-stats-mem-types.h b/xlators/debug/io-stats/src/io-stats-mem-types.h
index 9dde9373264..51d38d8b97c 100644
--- a/xlators/debug/io-stats/src/io-stats-mem-types.h
+++ b/xlators/debug/io-stats/src/io-stats-mem-types.h
@@ -11,18 +11,17 @@
#ifndef __IO_STATS_MEM_TYPES_H__
#define __IO_STATS_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
extern const char *__progname;
enum gf_io_stats_mem_types_ {
- gf_io_stats_mt_ios_conf = gf_common_mt_end + 1,
- gf_io_stats_mt_ios_fd,
- gf_io_stats_mt_ios_stat,
- gf_io_stats_mt_ios_stat_list,
- gf_io_stats_mt_ios_sample_buf,
- gf_io_stats_mt_ios_sample,
- gf_io_stats_mt_end
+ gf_io_stats_mt_ios_conf = gf_common_mt_end + 1,
+ gf_io_stats_mt_ios_fd,
+ gf_io_stats_mt_ios_stat,
+ gf_io_stats_mt_ios_stat_list,
+ gf_io_stats_mt_ios_sample_buf,
+ gf_io_stats_mt_ios_sample,
+ gf_io_stats_mt_end
};
#endif
-
diff --git a/xlators/debug/io-stats/src/io-stats.c b/xlators/debug/io-stats/src/io-stats.c
index 6ddad96d8aa..aa00c446e5a 100644
--- a/xlators/debug/io-stats/src/io-stats.c
+++ b/xlators/debug/io-stats/src/io-stats.c
@@ -7,8 +7,8 @@
later), or the GNU General Public License, version 2 (GPLv2), in all
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "syscall.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/syscall.h>
/**
* xlators/debug/io_stats :
@@ -17,9 +17,9 @@
*
* a) total read data - since process start, last interval and per fd
* b) total write data - since process start, last interval and per fd
- * c) counts of read IO block size - since process start, last interval and per fd
- * d) counts of write IO block size - since process start, last interval and per fd
- * e) counts of all FOP types passing through it
+ * c) counts of read IO block size - since process start, last interval and per
+ * fd d) counts of write IO block size - since process start, last interval and
+ * per fd e) counts of all FOP types passing through it
*
* Usage: setfattr -n trusted.io-stats-dump /tmp/filename /mnt/gluster
* output is written to /tmp/filename.<iostats xlator instance name>
@@ -28,187 +28,199 @@
#include <fnmatch.h>
#include <errno.h>
-#include "glusterfs.h"
-#include "xlator.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/xlator.h>
#include "io-stats-mem-types.h"
#include <stdarg.h>
-#include "defaults.h"
-#include "logging.h"
-#include "cli1-xdr.h"
-#include "statedump.h"
-#include "syncop.h"
+#include <glusterfs/defaults.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/statedump.h>
+#include <glusterfs/syncop.h>
#include <pwd.h>
#include <grp.h>
-#include "upcall-utils.h"
+#include <glusterfs/upcall-utils.h>
+#include <glusterfs/async.h>
#define MAX_LIST_MEMBERS 100
#define DEFAULT_PWD_BUF_SZ 16384
#define DEFAULT_GRP_BUF_SZ 16384
#define IOS_BLOCK_COUNT_SIZE 32
+#define IOS_STATS_DUMP_DIR DEFAULT_VAR_RUN_DIRECTORY
+
typedef enum {
- IOS_STATS_TYPE_NONE,
- IOS_STATS_TYPE_OPEN,
- IOS_STATS_TYPE_READ,
- IOS_STATS_TYPE_WRITE,
- IOS_STATS_TYPE_OPENDIR,
- IOS_STATS_TYPE_READDIRP,
- IOS_STATS_TYPE_READ_THROUGHPUT,
- IOS_STATS_TYPE_WRITE_THROUGHPUT,
- IOS_STATS_TYPE_MAX
-}ios_stats_type_t;
+ IOS_STATS_TYPE_NONE,
+ IOS_STATS_TYPE_OPEN,
+ IOS_STATS_TYPE_READ,
+ IOS_STATS_TYPE_WRITE,
+ IOS_STATS_TYPE_OPENDIR,
+ IOS_STATS_TYPE_READDIRP,
+ IOS_STATS_TYPE_READ_THROUGHPUT,
+ IOS_STATS_TYPE_WRITE_THROUGHPUT,
+ IOS_STATS_TYPE_MAX
+} ios_stats_type_t;
typedef enum {
- IOS_STATS_THRU_READ,
- IOS_STATS_THRU_WRITE,
- IOS_STATS_THRU_MAX,
-}ios_stats_thru_t;
+ IOS_STATS_THRU_READ,
+ IOS_STATS_THRU_WRITE,
+ IOS_STATS_THRU_MAX,
+} ios_stats_thru_t;
+
+/* This is same as gf1_cli_info_op */
+/* had to be defined here again, so we have modularity between
+ xdr, xlator, and library functions */
+typedef enum ios_info_op {
+ GF_IOS_INFO_NONE = 0,
+ GF_IOS_INFO_ALL = 1,
+ GF_IOS_INFO_INCREMENTAL = 2,
+ GF_IOS_INFO_CUMULATIVE = 3,
+ GF_IOS_INFO_CLEAR = 4,
+} ios_info_op_t;
struct ios_stat_lat {
- struct timeval time;
- double throughput;
+ struct timeval time;
+ double throughput;
};
struct ios_stat {
- gf_lock_t lock;
- uuid_t gfid;
- char *filename;
- gf_atomic_t counters[IOS_STATS_TYPE_MAX];
- struct ios_stat_lat thru_counters [IOS_STATS_THRU_MAX];
- gf_atomic_t refcnt;
+ gf_lock_t lock;
+ uuid_t gfid;
+ char *filename;
+ gf_atomic_t counters[IOS_STATS_TYPE_MAX];
+ struct ios_stat_lat thru_counters[IOS_STATS_THRU_MAX];
+ gf_atomic_t refcnt;
};
struct ios_stat_list {
- struct list_head list;
- struct ios_stat *iosstat;
- double value;
+ struct list_head list;
+ struct ios_stat *iosstat;
+ double value;
};
struct ios_stat_head {
- gf_lock_t lock;
- double min_cnt;
- uint64_t members;
- struct ios_stat_list *iosstats;
+ gf_lock_t lock;
+ double min_cnt;
+ uint64_t members;
+ struct ios_stat_list *iosstats;
};
typedef struct _ios_sample_t {
- uid_t uid;
- gid_t gid;
- char identifier[UNIX_PATH_MAX];
- glusterfs_fop_t fop_type;
- struct timeval timestamp;
- double elapsed;
+ uid_t uid;
+ gid_t gid;
+ char identifier[UNIX_PATH_MAX];
+ glusterfs_fop_t fop_type;
+ struct timeval timestamp;
+ double elapsed;
} ios_sample_t;
-
typedef struct _ios_sample_buf_t {
- uint64_t pos; /* Position in write buffer */
- uint64_t size; /* Size of ring buffer */
- uint64_t collected; /* Number of samples we've collected */
- uint64_t observed; /* Number of FOPs we've observed */
- ios_sample_t *ios_samples; /* Our list of samples */
+ uint64_t pos; /* Position in write buffer */
+ uint64_t size; /* Size of ring buffer */
+ uint64_t collected; /* Number of samples we've collected */
+ uint64_t observed; /* Number of FOPs we've observed */
+ ios_sample_t *ios_samples; /* Our list of samples */
} ios_sample_buf_t;
-
struct ios_lat {
- double min;
- double max;
- double avg;
- uint64_t total;
+ double min;
+ double max;
+ double avg;
+ uint64_t total;
};
struct ios_global_stats {
- gf_atomic_t data_written;
- gf_atomic_t data_read;
- gf_atomic_t block_count_write[IOS_BLOCK_COUNT_SIZE];
- gf_atomic_t block_count_read[IOS_BLOCK_COUNT_SIZE];
- gf_atomic_t fop_hits[GF_FOP_MAXVALUE];
- gf_atomic_t upcall_hits[GF_UPCALL_FLAGS_MAXVALUE];
- struct timeval started_at;
- struct ios_lat latency[GF_FOP_MAXVALUE];
- uint64_t nr_opens;
- uint64_t max_nr_opens;
- struct timeval max_openfd_time;
+ gf_atomic_t data_written;
+ gf_atomic_t data_read;
+ gf_atomic_t block_count_write[IOS_BLOCK_COUNT_SIZE];
+ gf_atomic_t block_count_read[IOS_BLOCK_COUNT_SIZE];
+ gf_atomic_t fop_hits[GF_FOP_MAXVALUE];
+ gf_atomic_t upcall_hits[GF_UPCALL_FLAGS_MAXVALUE];
+ time_t started_at;
+ struct ios_lat latency[GF_FOP_MAXVALUE];
+ uint64_t nr_opens;
+ uint64_t max_nr_opens;
+ struct timeval max_openfd_time;
};
typedef enum {
- IOS_DUMP_TYPE_NONE = 0,
- IOS_DUMP_TYPE_FILE = 1,
- IOS_DUMP_TYPE_DICT = 2,
- IOS_DUMP_TYPE_JSON_FILE = 3,
- IOS_DUMP_TYPE_SAMPLES = 4,
- IOS_DUMP_TYPE_MAX = 5
+ IOS_DUMP_TYPE_NONE = 0,
+ IOS_DUMP_TYPE_FILE = 1,
+ IOS_DUMP_TYPE_DICT = 2,
+ IOS_DUMP_TYPE_JSON_FILE = 3,
+ IOS_DUMP_TYPE_SAMPLES = 4,
+ IOS_DUMP_TYPE_MAX = 5
} ios_dump_type_t;
struct ios_conf {
- gf_lock_t lock;
- struct ios_global_stats cumulative;
- uint64_t increment;
- struct ios_global_stats incremental;
- gf_boolean_t dump_fd_stats;
- gf_boolean_t count_fop_hits;
- gf_boolean_t measure_latency;
- struct ios_stat_head list[IOS_STATS_TYPE_MAX];
- struct ios_stat_head thru_list[IOS_STATS_THRU_MAX];
- int32_t ios_dump_interval;
- pthread_t dump_thread;
- gf_boolean_t dump_thread_should_die;
- gf_lock_t ios_sampling_lock;
- int32_t ios_sample_interval;
- int32_t ios_sample_buf_size;
- ios_sample_buf_t *ios_sample_buf;
- struct dnscache *dnscache;
- int32_t ios_dnscache_ttl_sec;
- /*
- * What we really need here is just a unique value to keep files
- * created by this instance distinct from those created by any other.
- * On the client side this isn't a problem, so we just use the
- * translator name. On the server side conflicts can occur, so the
- * volfile-generation code automatically sets this (via an option)
- * to be the brick path.
- *
- * NB While the *field* name has changed, it didn't seem worth changing
- * all of the cases where "xlator_name" is used as a *variable* name.
- */
- char *unique_id;
- ios_dump_type_t dump_format;
- char *dump_format_str;
+ gf_lock_t lock;
+ struct ios_global_stats cumulative;
+ uint64_t increment;
+ struct ios_global_stats incremental;
+ gf_boolean_t dump_fd_stats;
+ gf_boolean_t count_fop_hits;
+ gf_boolean_t measure_latency;
+ struct ios_stat_head list[IOS_STATS_TYPE_MAX];
+ struct ios_stat_head thru_list[IOS_STATS_THRU_MAX];
+ int32_t ios_dump_interval;
+ pthread_t dump_thread;
+ gf_boolean_t dump_thread_should_die;
+ gf_boolean_t dump_thread_running;
+ gf_lock_t ios_sampling_lock;
+ int32_t ios_sample_interval;
+ int32_t ios_sample_buf_size;
+ ios_sample_buf_t *ios_sample_buf;
+ struct dnscache *dnscache;
+ int32_t ios_dnscache_ttl_sec;
+ /*
+ * What we really need here is just a unique value to keep files
+ * created by this instance distinct from those created by any other.
+ * On the client side this isn't a problem, so we just use the
+ * translator name. On the server side conflicts can occur, so the
+ * volfile-generation code automatically sets this (via an option)
+ * to be the brick path.
+ *
+ * NB While the *field* name has changed, it didn't seem worth changing
+ * all of the cases where "xlator_name" is used as a *variable* name.
+ */
+ char *unique_id;
+ ios_dump_type_t dump_format;
};
-
struct ios_fd {
- char *filename;
- gf_atomic_t data_written;
- gf_atomic_t data_read;
- gf_atomic_t block_count_write[IOS_BLOCK_COUNT_SIZE];
- gf_atomic_t block_count_read[IOS_BLOCK_COUNT_SIZE];
- struct timeval opened_at;
+ char *filename;
+ gf_atomic_t data_written;
+ gf_atomic_t data_read;
+ gf_atomic_t block_count_write[IOS_BLOCK_COUNT_SIZE];
+ gf_atomic_t block_count_read[IOS_BLOCK_COUNT_SIZE];
+ struct timeval opened_at;
};
struct ios_dump_args {
- ios_dump_type_t type;
- union {
- FILE *logfp;
- dict_t *dict;
- } u;
+ ios_dump_type_t type;
+ union {
+ FILE *logfp;
+ dict_t *dict;
+ } u;
};
-typedef int (*block_dump_func) (xlator_t *, struct ios_dump_args*,
- int , int , uint64_t ) ;
+typedef int (*block_dump_func)(xlator_t *, struct ios_dump_args *, int, int,
+ uint64_t);
struct ios_local {
- struct timeval wind_at;
- struct timeval unwind_at;
+ struct timeval wind_at;
+ struct timeval unwind_at;
};
struct volume_options options[];
static int
-is_fop_latency_started (call_frame_t *frame)
+is_fop_latency_started(call_frame_t *frame)
{
- GF_ASSERT (frame);
- struct timeval epoch = {0,};
- return memcmp (&frame->begin, &epoch, sizeof (epoch));
+ GF_ASSERT(frame);
+ struct timeval epoch = {
+ 0,
+ };
+ return memcmp(&frame->begin, &epoch, sizeof(epoch));
}
#define _IOS_SAMP_DIR DEFAULT_LOG_FILE_DIRECTORY "/samples"
@@ -218,355 +230,337 @@ is_fop_latency_started (call_frame_t *frame)
#define _IOS_DUMP_DIR DATADIR "/db/glusterd/stats"
#endif
-#define END_FOP_LATENCY(frame, op) \
- do { \
- struct ios_conf *conf = NULL; \
- \
- conf = this->private; \
- if (conf && conf->measure_latency) { \
- timespec_now (&frame->end); \
- update_ios_latency (conf, frame, GF_FOP_##op); \
- } \
- } while (0)
-
-#define START_FOP_LATENCY(frame) \
- do { \
- struct ios_conf *conf = NULL; \
- \
- conf = this->private; \
- if (conf && conf->measure_latency) { \
- timespec_now (&frame->begin); \
- } else { \
- memset (&frame->begin, 0, sizeof (frame->begin));\
- } \
- } while (0)
-
-
-#define BUMP_FOP(op) \
- do { \
- struct ios_conf *conf = NULL; \
- \
- conf = this->private; \
- if (!conf) \
- break; \
- GF_ATOMIC_INC (conf->cumulative.fop_hits[GF_FOP_##op]); \
- GF_ATOMIC_INC (conf->incremental.fop_hits[GF_FOP_##op]);\
- } while (0)
-
-#define UPDATE_PROFILE_STATS(frame, op) \
- do { \
- struct ios_conf *conf = NULL; \
- \
- if (!is_fop_latency_started (frame)) \
- break; \
- conf = this->private; \
- if (conf && conf->measure_latency && \
- conf->count_fop_hits) { \
- BUMP_FOP (op); \
- timespec_now (&frame->end); \
- update_ios_latency (conf, frame, GF_FOP_##op); \
- } \
- } while (0)
+#define END_FOP_LATENCY(frame, op) \
+ do { \
+ struct ios_conf *conf = NULL; \
+ \
+ conf = this->private; \
+ if (conf && conf->measure_latency) { \
+ timespec_now(&frame->end); \
+ update_ios_latency(conf, frame, GF_FOP_##op); \
+ } \
+ } while (0)
+
+#define START_FOP_LATENCY(frame) \
+ do { \
+ struct ios_conf *conf = NULL; \
+ \
+ conf = this->private; \
+ if (conf && conf->measure_latency) { \
+ timespec_now(&frame->begin); \
+ } else { \
+ memset(&frame->begin, 0, sizeof(frame->begin)); \
+ } \
+ } while (0)
+
+#define BUMP_FOP(op) \
+ do { \
+ struct ios_conf *conf = NULL; \
+ \
+ conf = this->private; \
+ if (!conf) \
+ break; \
+ GF_ATOMIC_INC(conf->cumulative.fop_hits[GF_FOP_##op]); \
+ GF_ATOMIC_INC(conf->incremental.fop_hits[GF_FOP_##op]); \
+ } while (0)
+
+#define UPDATE_PROFILE_STATS(frame, op) \
+ do { \
+ struct ios_conf *conf = NULL; \
+ \
+ if (!is_fop_latency_started(frame)) \
+ break; \
+ conf = this->private; \
+ if (conf && conf->measure_latency && conf->count_fop_hits) { \
+ BUMP_FOP(op); \
+ timespec_now(&frame->end); \
+ update_ios_latency(conf, frame, GF_FOP_##op); \
+ } \
+ } while (0)
#define BUMP_THROUGHPUT(iosstat, type) \
- do { \
- struct ios_conf *conf = NULL; \
- double elapsed; \
- struct timespec *begin, *end; \
- double throughput; \
- int flag = 0; \
+ do { \
+ struct ios_conf *conf = NULL; \
+ double elapsed; \
+ struct timespec *begin, *end; \
+ double throughput; \
+ int flag = 0; \
+ struct timeval tv = { \
+ 0, \
+ }; \
\
- begin = &frame->begin; \
- end = &frame->end; \
+ begin = &frame->begin; \
+ end = &frame->end; \
\
- elapsed = ((end->tv_sec - begin->tv_sec) * 1e9 \
- + (end->tv_nsec - begin->tv_nsec)) / 1000; \
- throughput = op_ret / elapsed; \
+ elapsed = gf_tsdiff(begin, end) / 1000.0; \
+ throughput = op_ret / elapsed; \
\
- conf = this->private; \
- LOCK (&iosstat->lock); \
- { \
- if (iosstat->thru_counters[type].throughput \
- <= throughput) { \
- iosstat->thru_counters[type].throughput = \
- throughput; \
- gettimeofday (&iosstat-> \
- thru_counters[type].time, NULL); \
- flag = 1; \
- } \
- } \
- UNLOCK (&iosstat->lock); \
- if (flag) \
- ios_stat_add_to_list (&conf->thru_list[type], \
- throughput, iosstat); \
- } while (0)
-
+ conf = this->private; \
+ gettimeofday(&tv, NULL); \
+ LOCK(&iosstat->lock); \
+ { \
+ if (iosstat->thru_counters[type].throughput <= throughput) { \
+ iosstat->thru_counters[type].throughput = throughput; \
+ memcpy(&iosstat->thru_counters[type].time, &tv, \
+ sizeof(struct timeval)); \
+ flag = 1; \
+ } \
+ } \
+ UNLOCK(&iosstat->lock); \
+ if (flag) \
+ ios_stat_add_to_list(&conf->thru_list[type], throughput, iosstat); \
+ } while (0)
static int
-ios_fd_ctx_get (fd_t *fd, xlator_t *this, struct ios_fd **iosfd)
+ios_fd_ctx_get(fd_t *fd, xlator_t *this, struct ios_fd **iosfd)
{
- uint64_t iosfd64 = 0;
- unsigned long iosfdlong = 0;
- int ret = 0;
+ uint64_t iosfd64 = 0;
+ unsigned long iosfdlong = 0;
+ int ret = 0;
- ret = fd_ctx_get (fd, this, &iosfd64);
- iosfdlong = iosfd64;
- if (ret != -1)
- *iosfd = (void *) iosfdlong;
+ ret = fd_ctx_get(fd, this, &iosfd64);
+ iosfdlong = iosfd64;
+ if (ret != -1)
+ *iosfd = (void *)iosfdlong;
- return ret;
+ return ret;
}
-
static int
-ios_fd_ctx_set (fd_t *fd, xlator_t *this, struct ios_fd *iosfd)
+ios_fd_ctx_set(fd_t *fd, xlator_t *this, struct ios_fd *iosfd)
{
- uint64_t iosfd64 = 0;
- int ret = 0;
+ uint64_t iosfd64 = 0;
+ int ret = 0;
- iosfd64 = (unsigned long) iosfd;
- ret = fd_ctx_set (fd, this, iosfd64);
+ iosfd64 = (unsigned long)iosfd;
+ ret = fd_ctx_set(fd, this, iosfd64);
- return ret;
+ return ret;
}
-
static int
-ios_stat_ref (struct ios_stat *iosstat)
+ios_stat_ref(struct ios_stat *iosstat)
{
- uint64_t refcnt = 0;
- refcnt = GF_ATOMIC_INC (iosstat->refcnt);
+ uint64_t refcnt = 0;
+ refcnt = GF_ATOMIC_INC(iosstat->refcnt);
- return refcnt;
+ return refcnt;
}
-
static int
-ios_stat_unref (struct ios_stat *iosstat)
+ios_stat_unref(struct ios_stat *iosstat)
{
- int cleanup = 0;
- uint64_t refcnt = 0;
+ int cleanup = 0;
+ uint64_t refcnt = 0;
- refcnt = GF_ATOMIC_DEC (iosstat->refcnt);
- if (refcnt == 0) {
- if (iosstat->filename) {
- GF_FREE (iosstat->filename);
- iosstat->filename = NULL;
- }
- cleanup = 1;
+ refcnt = GF_ATOMIC_DEC(iosstat->refcnt);
+ if (refcnt == 0) {
+ if (iosstat->filename) {
+ GF_FREE(iosstat->filename);
+ iosstat->filename = NULL;
}
+ cleanup = 1;
+ }
- if (cleanup) {
- LOCK_DESTROY (&iosstat->lock);
- GF_FREE (iosstat);
- iosstat = NULL;
- }
+ if (cleanup) {
+ LOCK_DESTROY(&iosstat->lock);
+ GF_FREE(iosstat);
+ iosstat = NULL;
+ }
- return 0;
+ return 0;
}
-
static int
-ios_stat_add_to_list (struct ios_stat_head *list_head, uint64_t value,
- struct ios_stat *iosstat)
-{
- struct ios_stat_list *new = NULL;
- struct ios_stat_list *entry = NULL;
- struct ios_stat_list *t = NULL;
- struct ios_stat_list *list_entry = NULL;
- struct ios_stat_list *tmp = NULL;
- struct ios_stat_list *last = NULL;
- struct ios_stat *stat = NULL;
- int cnt = 0;
- int found = 0;
- int reposition = 0;
- double min_count = 0;
-
- LOCK (&list_head->lock);
+ios_stat_add_to_list(struct ios_stat_head *list_head, uint64_t value,
+ struct ios_stat *iosstat)
+{
+ struct ios_stat_list *new = NULL;
+ struct ios_stat_list *entry = NULL;
+ struct ios_stat_list *t = NULL;
+ struct ios_stat_list *list_entry = NULL;
+ struct ios_stat_list *tmp = NULL;
+ struct ios_stat_list *last = NULL;
+ struct ios_stat *stat = NULL;
+ int cnt = 0;
+ int found = 0;
+ int reposition = 0;
+ double min_count = 0;
+
+ LOCK(&list_head->lock);
+ {
+ if (list_head->min_cnt == 0)
+ list_head->min_cnt = value;
+ if ((list_head->members == MAX_LIST_MEMBERS) &&
+ (list_head->min_cnt > value))
+ goto out;
+
+ list_for_each_entry_safe(entry, t, &list_head->iosstats->list, list)
{
-
- if (list_head->min_cnt == 0)
+ cnt++;
+ if (cnt == list_head->members)
+ last = entry;
+
+ if (!gf_uuid_compare(iosstat->gfid, entry->iosstat->gfid)) {
+ list_entry = entry;
+ found = cnt;
+ entry->value = value;
+ if (!reposition) {
+ if (cnt == list_head->members)
list_head->min_cnt = value;
- if ((list_head->members == MAX_LIST_MEMBERS) &&
- (list_head->min_cnt > value))
- goto out;
-
- list_for_each_entry_safe (entry, t,
- &list_head->iosstats->list, list) {
- cnt++;
- if (cnt == list_head->members)
- last = entry;
-
- if (!gf_uuid_compare (iosstat->gfid,
- entry->iosstat->gfid)) {
- list_entry = entry;
- found = cnt;
- entry->value = value;
- if (!reposition) {
- if (cnt == list_head->members)
- list_head->min_cnt = value;
- goto out;
- }
- break;
- } else if (entry->value <= value && !reposition) {
- reposition = cnt;
- tmp = entry;
- if (cnt == list_head->members - 1)
- min_count = entry->value;
- }
- }
- if (found) {
- list_del (&list_entry->list);
- list_add_tail (&list_entry->list, &tmp->list);
- if (min_count)
- list_head->min_cnt = min_count;
- goto out;
- } else if (list_head->members == MAX_LIST_MEMBERS && reposition) {
- new = GF_CALLOC (1, sizeof (*new),
- gf_io_stats_mt_ios_stat_list);
- new->iosstat = iosstat;
- new->value = value;
- ios_stat_ref (iosstat);
- list_add_tail (&new->list, &tmp->list);
- if (last) {
- stat = last->iosstat;
- last->iosstat = NULL;
- ios_stat_unref (stat);
- list_del (&last->list);
- GF_FREE (last);
- }
- if (reposition == MAX_LIST_MEMBERS)
- list_head->min_cnt = value;
- else if (min_count) {
- list_head->min_cnt = min_count;
- }
- } else if (list_head->members < MAX_LIST_MEMBERS) {
- new = GF_CALLOC (1, sizeof (*new),
- gf_io_stats_mt_ios_stat_list);
- new->iosstat = iosstat;
- new->value = value;
- ios_stat_ref (iosstat);
- if (reposition) {
- list_add_tail (&new->list, &tmp->list);
- } else {
- list_add_tail (&new->list, &entry->list);
- }
- list_head->members++;
- if (list_head->min_cnt > value)
- list_head->min_cnt = value;
+ goto out;
}
+ break;
+ } else if (entry->value <= value && !reposition) {
+ reposition = cnt;
+ tmp = entry;
+ if (cnt == list_head->members - 1)
+ min_count = entry->value;
+ }
+ }
+ if (found) {
+ list_del(&list_entry->list);
+ list_add_tail(&list_entry->list, &tmp->list);
+ if (min_count)
+ list_head->min_cnt = min_count;
+ goto out;
+ } else if (list_head->members == MAX_LIST_MEMBERS && reposition) {
+ new = GF_CALLOC(1, sizeof(*new), gf_io_stats_mt_ios_stat_list);
+ new->iosstat = iosstat;
+ new->value = value;
+ ios_stat_ref(iosstat);
+ list_add_tail(&new->list, &tmp->list);
+ if (last) {
+ stat = last->iosstat;
+ last->iosstat = NULL;
+ ios_stat_unref(stat);
+ list_del(&last->list);
+ GF_FREE(last);
+ }
+ if (reposition == MAX_LIST_MEMBERS)
+ list_head->min_cnt = value;
+ else if (min_count) {
+ list_head->min_cnt = min_count;
+ }
+ } else if (list_head->members < MAX_LIST_MEMBERS) {
+ new = GF_CALLOC(1, sizeof(*new), gf_io_stats_mt_ios_stat_list);
+ new->iosstat = iosstat;
+ new->value = value;
+ ios_stat_ref(iosstat);
+ if (reposition) {
+ list_add_tail(&new->list, &tmp->list);
+ } else {
+ list_add_tail(&new->list, &entry->list);
+ }
+ list_head->members++;
+ if (list_head->min_cnt > value)
+ list_head->min_cnt = value;
}
+ }
out:
- UNLOCK (&list_head->lock);
- return 0;
+ UNLOCK(&list_head->lock);
+ return 0;
}
-
static void
-ios_bump_read (xlator_t *this, fd_t *fd, size_t len)
-{
- struct ios_conf *conf = NULL;
- struct ios_fd *iosfd = NULL;
- int lb2 = 0;
-
- conf = this->private;
- lb2 = log_base2 (len);
- ios_fd_ctx_get (fd, this, &iosfd);
- if (!conf)
- return;
-
- GF_ATOMIC_ADD (conf->cumulative.data_read, len);
- GF_ATOMIC_ADD (conf->incremental.data_read, len);
- GF_ATOMIC_INC (conf->cumulative.block_count_read[lb2]);
- GF_ATOMIC_INC (conf->incremental.block_count_read[lb2]);
-
- if (iosfd) {
- GF_ATOMIC_ADD (iosfd->data_read, len);
- GF_ATOMIC_INC (iosfd->block_count_read[lb2]);
- }
-}
+ios_bump_read(xlator_t *this, fd_t *fd, size_t len)
+{
+ struct ios_conf *conf = NULL;
+ struct ios_fd *iosfd = NULL;
+ int lb2 = 0;
+ conf = this->private;
+ lb2 = log_base2(len);
+ ios_fd_ctx_get(fd, this, &iosfd);
+ if (!conf)
+ return;
-static void
-ios_bump_write (xlator_t *this, fd_t *fd, size_t len)
-{
- struct ios_conf *conf = NULL;
- struct ios_fd *iosfd = NULL;
- int lb2 = 0;
-
- conf = this->private;
- lb2 = log_base2 (len);
- ios_fd_ctx_get (fd, this, &iosfd);
- if (!conf)
- return;
-
- GF_ATOMIC_ADD (conf->cumulative.data_written, len);
- GF_ATOMIC_ADD (conf->incremental.data_written, len);
- GF_ATOMIC_INC (conf->cumulative.block_count_write[lb2]);
- GF_ATOMIC_INC (conf->incremental.block_count_write[lb2]);
-
- if (iosfd) {
- GF_ATOMIC_ADD (iosfd->data_written, len);
- GF_ATOMIC_INC (iosfd->block_count_write[lb2]);
- }
-}
+ GF_ATOMIC_ADD(conf->cumulative.data_read, len);
+ GF_ATOMIC_ADD(conf->incremental.data_read, len);
+ GF_ATOMIC_INC(conf->cumulative.block_count_read[lb2]);
+ GF_ATOMIC_INC(conf->incremental.block_count_read[lb2]);
+ if (iosfd) {
+ GF_ATOMIC_ADD(iosfd->data_read, len);
+ GF_ATOMIC_INC(iosfd->block_count_read[lb2]);
+ }
+}
static void
-ios_bump_upcall (xlator_t *this, gf_upcall_flags_t event)
+ios_bump_write(xlator_t *this, fd_t *fd, size_t len)
{
- struct ios_conf *conf = NULL;
+ struct ios_conf *conf = NULL;
+ struct ios_fd *iosfd = NULL;
+ int lb2 = 0;
- conf = this->private;
- if (!conf)
- return;
- if (conf->count_fop_hits) {
- GF_ATOMIC_INC (conf->cumulative.upcall_hits[event]);
- GF_ATOMIC_INC (conf->incremental.upcall_hits[event]);
- }
+ conf = this->private;
+ lb2 = log_base2(len);
+ ios_fd_ctx_get(fd, this, &iosfd);
+ if (!conf)
+ return;
+
+ GF_ATOMIC_ADD(conf->cumulative.data_written, len);
+ GF_ATOMIC_ADD(conf->incremental.data_written, len);
+ GF_ATOMIC_INC(conf->cumulative.block_count_write[lb2]);
+ GF_ATOMIC_INC(conf->incremental.block_count_write[lb2]);
+
+ if (iosfd) {
+ GF_ATOMIC_ADD(iosfd->data_written, len);
+ GF_ATOMIC_INC(iosfd->block_count_write[lb2]);
+ }
}
+static void
+ios_bump_upcall(xlator_t *this, gf_upcall_flags_t event)
+{
+ struct ios_conf *conf = NULL;
+
+ conf = this->private;
+ if (!conf)
+ return;
+ if (conf->count_fop_hits) {
+ GF_ATOMIC_INC(conf->cumulative.upcall_hits[event]);
+ GF_ATOMIC_INC(conf->incremental.upcall_hits[event]);
+ }
+}
static void
-ios_bump_stats (xlator_t *this, struct ios_stat *iosstat,
- ios_stats_type_t type)
+ios_bump_stats(xlator_t *this, struct ios_stat *iosstat, ios_stats_type_t type)
{
- struct ios_conf *conf = NULL;
- uint64_t value = 0;
+ struct ios_conf *conf = NULL;
+ uint64_t value = 0;
- conf = this->private;
+ conf = this->private;
- value = GF_ATOMIC_INC (iosstat->counters[type]);
- ios_stat_add_to_list (&conf->list[type], value, iosstat);
+ value = GF_ATOMIC_INC(iosstat->counters[type]);
+ ios_stat_add_to_list(&conf->list[type], value, iosstat);
}
-
int
-ios_inode_ctx_set (inode_t *inode, xlator_t *this, struct ios_stat *iosstat)
+ios_inode_ctx_set(inode_t *inode, xlator_t *this, struct ios_stat *iosstat)
{
- uint64_t iosstat64 = 0;
- int ret = 0;
+ uint64_t iosstat64 = 0;
+ int ret = 0;
- ios_stat_ref (iosstat);
- iosstat64 = (unsigned long) iosstat;
- ret = inode_ctx_put (inode, this, iosstat64);
- return ret;
+ ios_stat_ref(iosstat);
+ iosstat64 = (unsigned long)iosstat;
+ ret = inode_ctx_put(inode, this, iosstat64);
+ return ret;
}
int
-ios_inode_ctx_get (inode_t *inode, xlator_t *this, struct ios_stat **iosstat)
+ios_inode_ctx_get(inode_t *inode, xlator_t *this, struct ios_stat **iosstat)
{
- uint64_t iosstat64 = 0;
- unsigned long iosstatlong = 0;
- int ret = 0;
+ uint64_t iosstat64 = 0;
+ unsigned long iosstatlong = 0;
+ int ret = 0;
- ret = inode_ctx_get (inode, this, &iosstat64);
- iosstatlong = iosstat64;
- if (ret != -1)
- *iosstat = (void *) iosstatlong;
-
- return ret;
+ ret = inode_ctx_get(inode, this, &iosstat64);
+ iosstatlong = iosstat64;
+ if (ret != -1)
+ *iosstat = (void *)iosstatlong;
+ return ret;
}
/*
@@ -581,598 +575,561 @@ ios_inode_ctx_get (inode_t *inode, xlator_t *this, struct ios_stat **iosstat)
*
*/
ios_sample_buf_t *
-ios_create_sample_buf (size_t buf_size)
+ios_create_sample_buf(size_t buf_size)
{
- ios_sample_buf_t *ios_sample_buf = NULL;
- ios_sample_t *ios_samples = NULL;
+ ios_sample_buf_t *ios_sample_buf = NULL;
+ ios_sample_t *ios_samples = NULL;
- ios_sample_buf = GF_CALLOC (1,
- sizeof (*ios_sample_buf),
- gf_io_stats_mt_ios_sample_buf);
- if (!ios_sample_buf)
- goto err;
+ ios_sample_buf = GF_CALLOC(1, sizeof(*ios_sample_buf),
+ gf_io_stats_mt_ios_sample_buf);
+ if (!ios_sample_buf)
+ goto err;
- ios_samples = GF_CALLOC (buf_size,
- sizeof (*ios_samples),
- gf_io_stats_mt_ios_sample);
+ ios_samples = GF_CALLOC(buf_size, sizeof(*ios_samples),
+ gf_io_stats_mt_ios_sample);
- if (!ios_samples)
- goto err;
+ if (!ios_samples)
+ goto err;
- ios_sample_buf->ios_samples = ios_samples;
- ios_sample_buf->size = buf_size;
- ios_sample_buf->pos = 0;
- ios_sample_buf->observed = 0;
- ios_sample_buf->collected = 0;
+ ios_sample_buf->ios_samples = ios_samples;
+ ios_sample_buf->size = buf_size;
+ ios_sample_buf->pos = 0;
+ ios_sample_buf->observed = 0;
+ ios_sample_buf->collected = 0;
- return ios_sample_buf;
+ return ios_sample_buf;
err:
- GF_FREE (ios_sample_buf);
- return NULL;
+ GF_FREE(ios_sample_buf);
+ return NULL;
}
void
-ios_destroy_sample_buf (ios_sample_buf_t *ios_sample_buf)
+ios_destroy_sample_buf(ios_sample_buf_t *ios_sample_buf)
{
- GF_FREE (ios_sample_buf->ios_samples);
- GF_FREE (ios_sample_buf);
+ GF_FREE(ios_sample_buf->ios_samples);
+ GF_FREE(ios_sample_buf);
}
static int
-ios_init_sample_buf (struct ios_conf *conf)
+ios_init_sample_buf(struct ios_conf *conf)
{
- int32_t ret = -1;
+ int32_t ret = -1;
- GF_ASSERT (conf);
- LOCK (&conf->lock);
- conf->ios_sample_buf = ios_create_sample_buf (
- conf->ios_sample_buf_size);
- if (!conf->ios_sample_buf)
- goto out;
- ret = 0;
+ GF_ASSERT(conf);
+ LOCK(&conf->lock);
+ conf->ios_sample_buf = ios_create_sample_buf(conf->ios_sample_buf_size);
+ if (!conf->ios_sample_buf)
+ goto out;
+ ret = 0;
out:
- UNLOCK (&conf->lock);
- return ret;
+ UNLOCK(&conf->lock);
+ return ret;
}
static int
-ios_stats_cleanup (xlator_t *this, inode_t *inode)
+ios_stats_cleanup(xlator_t *this, inode_t *inode)
{
+ struct ios_stat *iosstat = NULL;
+ uint64_t iosstat64 = 0;
- struct ios_stat *iosstat = NULL;
- uint64_t iosstat64 = 0;
-
- inode_ctx_del (inode, this, &iosstat64);
- if (!iosstat64) {
- gf_log (this->name, GF_LOG_WARNING,
- "could not get inode ctx");
- return 0;
- }
- iosstat = (void *) (long)iosstat64;
- if (iosstat) {
- ios_stat_unref (iosstat);
- }
+ inode_ctx_del(inode, this, &iosstat64);
+ if (!iosstat64) {
+ gf_log(this->name, GF_LOG_WARNING, "could not get inode ctx");
return 0;
-}
-
-#define ios_log(this, logfp, fmt ...) \
- do { \
- if (logfp) { \
- fprintf (logfp, fmt); \
- fprintf (logfp, "\n"); \
- } \
- gf_log (this->name, GF_LOG_DEBUG, fmt); \
- } while (0)
+ }
+ iosstat = (void *)(long)iosstat64;
+ if (iosstat) {
+ ios_stat_unref(iosstat);
+ }
+ return 0;
+}
+
+#define ios_log(this, logfp, fmt...) \
+ do { \
+ if (logfp) { \
+ fprintf(logfp, fmt); \
+ fprintf(logfp, "\n"); \
+ } \
+ gf_log(this->name, GF_LOG_DEBUG, fmt); \
+ } while (0)
int
-ios_dump_file_stats (struct ios_stat_head *list_head, xlator_t *this,
- FILE *logfp)
+ios_dump_file_stats(struct ios_stat_head *list_head, xlator_t *this,
+ FILE *logfp)
{
- struct ios_stat_list *entry = NULL;
+ struct ios_stat_list *entry = NULL;
- LOCK (&list_head->lock);
+ LOCK(&list_head->lock);
+ {
+ list_for_each_entry(entry, &list_head->iosstats->list, list)
{
- list_for_each_entry (entry, &list_head->iosstats->list, list) {
- ios_log (this, logfp, "%-12.0f %s",
- entry->value, entry->iosstat->filename);
- }
+ ios_log(this, logfp, "%-12.0f %s", entry->value,
+ entry->iosstat->filename);
}
- UNLOCK (&list_head->lock);
- return 0;
+ }
+ UNLOCK(&list_head->lock);
+ return 0;
}
int
-ios_dump_throughput_stats (struct ios_stat_head *list_head, xlator_t *this,
- FILE *logfp, ios_stats_thru_t type)
+ios_dump_throughput_stats(struct ios_stat_head *list_head, xlator_t *this,
+ FILE *logfp, ios_stats_thru_t type)
{
- struct ios_stat_list *entry = NULL;
- struct timeval time = {0, };
- char timestr[256] = {0, };
+ struct ios_stat_list *entry = NULL;
+ char timestr[GF_TIMESTR_SIZE] = {
+ 0,
+ };
- LOCK (&list_head->lock);
+ LOCK(&list_head->lock);
+ {
+ list_for_each_entry(entry, &list_head->iosstats->list, list)
{
- list_for_each_entry (entry, &list_head->iosstats->list, list) {
- gf_time_fmt (timestr, sizeof timestr,
- entry->iosstat->thru_counters[type].time.tv_sec,
- gf_timefmt_FT);
- snprintf (timestr + strlen (timestr), sizeof timestr - strlen (timestr),
- ".%"GF_PRI_SUSECONDS, time.tv_usec);
-
- ios_log (this, logfp, "%s \t %-10.2f \t %s",
- timestr, entry->value, entry->iosstat->filename);
- }
+ gf_time_fmt_tv(timestr, sizeof timestr,
+ &entry->iosstat->thru_counters[type].time,
+ gf_timefmt_FT);
+
+ ios_log(this, logfp, "%s \t %-10.2f \t %s", timestr, entry->value,
+ entry->iosstat->filename);
}
- UNLOCK (&list_head->lock);
- return 0;
+ }
+ UNLOCK(&list_head->lock);
+ return 0;
}
int
-_io_stats_get_key_prefix (xlator_t *this, char **key_prefix) {
- char *key_root = "gluster";
- char *xlator_name = NULL;
- char *instance_name = NULL;
- size_t key_len = 0;
- int bytes_written = 0;
- int i = 0;
- int ret = 0;
- struct ios_conf *conf = this->private;
-
- xlator_name = strdupa (conf->unique_id);
- for (i = 0; i < strlen (xlator_name); i++) {
- if (xlator_name[i] == '/')
- xlator_name[i] = '_';
+_io_stats_get_key_prefix(xlator_t *this, char **key_prefix)
+{
+ char *key_root = "gluster";
+ char *xlator_name = NULL;
+ char *instance_name = NULL;
+ size_t key_len = 0;
+ int bytes_written = 0;
+ int i = 0;
+ int ret = 0;
+ struct ios_conf *conf = this->private;
+
+ xlator_name = strdupa(conf->unique_id);
+ for (i = 0; i < strlen(xlator_name); i++) {
+ if (xlator_name[i] == '/')
+ xlator_name[i] = '_';
+ }
+
+ instance_name = this->instance_name;
+ if (this->name && strcmp(this->name, "glustershd") == 0) {
+ xlator_name = "shd";
+ } else if (this->prev && strcmp(this->prev->name, "nfs-server") == 0) {
+ xlator_name = "nfsd";
+ if (this->prev->instance_name)
+ instance_name = strdupa(this->prev->instance_name);
+ }
+
+ if (strcmp(__progname, "glusterfsd") == 0)
+ key_root = "gluster.brick";
+
+ if (instance_name) {
+ /* +3 for 2 x "." + NULL */
+ key_len = strlen(key_root) + strlen(xlator_name) +
+ strlen(instance_name) + 3;
+ *key_prefix = GF_CALLOC(key_len, sizeof(char), gf_common_mt_char);
+ if (!*key_prefix) {
+ ret = -ENOMEM;
+ goto err;
}
-
- instance_name = this->instance_name;
- if (this->name && strcmp (this->name, "glustershd") == 0) {
- xlator_name = "shd";
- } else if (this->prev &&
- strcmp (this->prev->name, "nfs-server") == 0) {
- xlator_name = "nfsd";
- if (this->prev->instance_name)
- instance_name = strdupa (this->prev->instance_name);
+ bytes_written = snprintf(*key_prefix, key_len, "%s.%s.%s", key_root,
+ xlator_name, instance_name);
+ if (bytes_written != key_len - 1) {
+ ret = -EINVAL;
+ goto err;
}
-
- if (strcmp (__progname, "glusterfsd") == 0)
- key_root = "gluster.brick";
-
- if (instance_name) {
- /* +3 for 2 x "." + NULL */
- key_len = strlen (key_root) + strlen (xlator_name) +
- strlen (instance_name) + 3;
- *key_prefix = GF_CALLOC (key_len, sizeof (char),
- gf_common_mt_char);
- if (!*key_prefix) {
- ret = -ENOMEM;
- goto err;
- }
- bytes_written = snprintf (*key_prefix, key_len, "%s.%s.%s",
- key_root, xlator_name, instance_name);
- if (bytes_written != key_len - 1) {
- ret = -EINVAL;
- goto err;
- }
- } else {
- /* +2 for 1 x "." + NULL */
- key_len = strlen (key_root) + strlen (xlator_name) + 2;
- *key_prefix = GF_CALLOC (key_len, sizeof (char),
- gf_common_mt_char);
- if (!*key_prefix) {
- ret = -ENOMEM;
- goto err;
- }
- bytes_written = snprintf (*key_prefix, key_len, "%s.%s",
- key_root, xlator_name);
- if (bytes_written != key_len - 1) {
- ret = -EINVAL;
- goto err;
- }
+ } else {
+ /* +2 for 1 x "." + NULL */
+ key_len = strlen(key_root) + strlen(xlator_name) + 2;
+ *key_prefix = GF_CALLOC(key_len, sizeof(char), gf_common_mt_char);
+ if (!*key_prefix) {
+ ret = -ENOMEM;
+ goto err;
}
- return 0;
+ bytes_written = snprintf(*key_prefix, key_len, "%s.%s", key_root,
+ xlator_name);
+ if (bytes_written != key_len - 1) {
+ ret = -EINVAL;
+ goto err;
+ }
+ }
+ return 0;
err:
- GF_FREE (*key_prefix);
- *key_prefix = NULL;
- return ret;
+ GF_FREE(*key_prefix);
+ *key_prefix = NULL;
+ return ret;
}
int
-io_stats_dump_global_to_json_logfp (xlator_t *this,
- struct ios_global_stats *stats, struct timeval *now, int interval,
- FILE *logfp)
-{
- int i = 0;
- int j = 0;
- struct ios_conf *conf = NULL;
- char *key_prefix = NULL;
- char *str_prefix = NULL;
- char *lc_fop_name = NULL;
- int ret = 1; /* Default to error */
- int rw_size;
- char *rw_unit = NULL;
- uint64_t fop_hits;
- float fop_lat_ave;
- float fop_lat_min;
- float fop_lat_max;
- double interval_sec;
- double fop_ave_usec = 0.0;
- double fop_ave_usec_sum = 0.0;
- double weighted_fop_ave_usec = 0.0;
- double weighted_fop_ave_usec_sum = 0.0;
- long total_fop_hits = 0;
- loc_t unused_loc = {0, };
- dict_t *xattr = NULL;
-
- interval_sec = ((now->tv_sec * 1000000.0 + now->tv_usec) -
- (stats->started_at.tv_sec * 1000000.0 +
- stats->started_at.tv_usec)) / 1000000.0;
-
- conf = this->private;
-
- ret = _io_stats_get_key_prefix (this, &key_prefix);
- if (ret) {
- goto out;
+io_stats_dump_global_to_json_logfp(xlator_t *this,
+ struct ios_global_stats *stats, time_t now,
+ int interval, FILE *logfp)
+{
+ int i = 0;
+ int j = 0;
+ struct ios_conf *conf = NULL;
+ char *key_prefix = NULL;
+ char *str_prefix = NULL;
+ char *lc_fop_name = NULL;
+ int ret = 1; /* Default to error */
+ int rw_size;
+ char *rw_unit = NULL;
+ uint64_t fop_hits;
+ float fop_lat_ave;
+ float fop_lat_min;
+ float fop_lat_max;
+ double interval_sec;
+ double fop_ave_usec = 0.0;
+ double fop_ave_usec_sum = 0.0;
+ double weighted_fop_ave_usec = 0.0;
+ double weighted_fop_ave_usec_sum = 0.0;
+ long total_fop_hits = 0;
+ loc_t unused_loc = {
+ 0,
+ };
+ dict_t *xattr = NULL;
+
+ interval_sec = (double)(now - stats->started_at);
+
+ conf = this->private;
+
+ ret = _io_stats_get_key_prefix(this, &key_prefix);
+ if (ret) {
+ goto out;
+ }
+
+ if (interval == -1) {
+ str_prefix = "aggr";
+
+ } else {
+ str_prefix = "inter";
+ }
+ ios_log(this, logfp, "{");
+
+ for (i = 0; i < 31; i++) {
+ rw_size = (1 << i);
+ if (rw_size >= 1024 * 1024) {
+ rw_size = rw_size / (1024 * 1024);
+ rw_unit = "mb";
+ } else if (rw_size >= 1024) {
+ rw_size = rw_size / 1024;
+ rw_unit = "kb";
+ } else {
+ rw_unit = "b";
}
if (interval == -1) {
- str_prefix = "aggr";
-
+ ios_log(this, logfp, "\"%s.%s.read_%d%s\": %" GF_PRI_ATOMIC ",",
+ key_prefix, str_prefix, rw_size, rw_unit,
+ GF_ATOMIC_GET(stats->block_count_read[i]));
+ ios_log(this, logfp, "\"%s.%s.write_%d%s\": %" GF_PRI_ATOMIC ",",
+ key_prefix, str_prefix, rw_size, rw_unit,
+ GF_ATOMIC_GET(stats->block_count_write[i]));
} else {
- str_prefix = "inter";
+ ios_log(this, logfp, "\"%s.%s.read_%d%s_per_sec\": %0.2lf,",
+ key_prefix, str_prefix, rw_size, rw_unit,
+ (double)(GF_ATOMIC_GET(stats->block_count_read[i]) /
+ interval_sec));
+ ios_log(this, logfp, "\"%s.%s.write_%d%s_per_sec\": %0.2lf,",
+ key_prefix, str_prefix, rw_size, rw_unit,
+ (double)(GF_ATOMIC_GET(stats->block_count_write[i]) /
+ interval_sec));
}
- ios_log (this, logfp, "{");
-
- for (i = 0; i < 31; i++) {
- rw_size = (1 << i);
- if (rw_size >= 1024 * 1024) {
- rw_size = rw_size / (1024 * 1024);
- rw_unit = "mb";
- } else if (rw_size >= 1024) {
- rw_size = rw_size / 1024;
- rw_unit = "kb";
- } else {
- rw_unit = "b";
- }
-
- if (interval == -1) {
- ios_log (this, logfp,
- "\"%s.%s.read_%d%s\": \"%"GF_PRI_ATOMIC"\",",
- key_prefix, str_prefix, rw_size, rw_unit,
- GF_ATOMIC_GET (stats->block_count_read[i]));
- ios_log (this, logfp,
- "\"%s.%s.write_%d%s\": \"%"GF_PRI_ATOMIC"\",",
- key_prefix, str_prefix, rw_size, rw_unit,
- GF_ATOMIC_GET (stats->block_count_write[i]));
- } else {
- ios_log (this, logfp,
- "\"%s.%s.read_%d%s_per_sec\": \"%0.2lf\",",
- key_prefix, str_prefix, rw_size, rw_unit,
- (double)
- (GF_ATOMIC_GET (stats->block_count_read[i]) /
- interval_sec));
- ios_log (this, logfp,
- "\"%s.%s.write_%d%s_per_sec\": \"%0.2lf\",",
- key_prefix, str_prefix, rw_size, rw_unit,
- (double)
- (GF_ATOMIC_GET (stats->block_count_write[i]) /
- interval_sec));
- }
+ }
+
+ if (interval == -1) {
+ ios_log(this, logfp, "\"%s.%s.fds.open_count\": %" PRId64 ",",
+ key_prefix, str_prefix, conf->cumulative.nr_opens);
+ ios_log(this, logfp, "\"%s.%s.fds.max_open_count\": %" PRId64 ",",
+ key_prefix, str_prefix, conf->cumulative.max_nr_opens);
+ }
+
+ for (i = 0; i < GF_FOP_MAXVALUE; i++) {
+ lc_fop_name = strdupa(gf_fop_list[i]);
+ for (j = 0; lc_fop_name[j]; j++) {
+ lc_fop_name[j] = tolower(lc_fop_name[j]);
}
- if (interval == -1) {
- ios_log (this, logfp, "\"%s.%s.fds.open_count\": \"%"PRId64
- "\",", key_prefix, str_prefix,
- conf->cumulative.nr_opens);
- ios_log (this, logfp,
- "\"%s.%s.fds.max_open_count\": \"%"PRId64"\",",
- key_prefix, str_prefix, conf->cumulative.max_nr_opens);
+ fop_hits = GF_ATOMIC_GET(stats->fop_hits[i]);
+ fop_lat_ave = 0.0;
+ fop_lat_min = 0.0;
+ fop_lat_max = 0.0;
+ if (fop_hits) {
+ if (stats->latency[i].avg) {
+ fop_lat_ave = stats->latency[i].avg;
+ fop_lat_min = stats->latency[i].min;
+ fop_lat_max = stats->latency[i].max;
+ }
}
-
- for (i = 0; i < GF_FOP_MAXVALUE; i++) {
- lc_fop_name = strdupa (gf_fop_list[i]);
- for (j = 0; lc_fop_name[j]; j++) {
- lc_fop_name[j] = tolower (lc_fop_name[j]);
- }
-
- fop_hits = GF_ATOMIC_GET (stats->fop_hits[i]);
- fop_lat_ave = 0.0;
- fop_lat_min = 0.0;
- fop_lat_max = 0.0;
- if (fop_hits) {
- if (stats->latency[i].avg) {
- fop_lat_ave = stats->latency[i].avg;
- fop_lat_min = stats->latency[i].min;
- fop_lat_max = stats->latency[i].max;
- }
- }
- if (interval == -1) {
- ios_log (this, logfp,
- "\"%s.%s.fop.%s.count\": \"%"GF_PRI_ATOMIC"\",",
- key_prefix, str_prefix, lc_fop_name,
- fop_hits);
- } else {
- ios_log (this, logfp,
- "\"%s.%s.fop.%s.per_sec\": \"%0.2lf\",",
- key_prefix, str_prefix, lc_fop_name,
- (double)(fop_hits / interval_sec));
- }
-
- ios_log (this, logfp,
- "\"%s.%s.fop.%s.latency_ave_usec\": \"%0.2lf\",",
- key_prefix, str_prefix, lc_fop_name, fop_lat_ave);
- ios_log (this, logfp,
- "\"%s.%s.fop.%s.latency_min_usec\": \"%0.2lf\",",
- key_prefix, str_prefix, lc_fop_name, fop_lat_min);
- ios_log (this, logfp,
- "\"%s.%s.fop.%s.latency_max_usec\": \"%0.2lf\",",
- key_prefix, str_prefix, lc_fop_name, fop_lat_max);
-
- fop_ave_usec_sum += fop_lat_ave;
- weighted_fop_ave_usec_sum += fop_hits * fop_lat_ave;
- total_fop_hits += fop_hits;
+ if (interval == -1) {
+ ios_log(this, logfp, "\"%s.%s.fop.%s.count\": %" GF_PRI_ATOMIC ",",
+ key_prefix, str_prefix, lc_fop_name, fop_hits);
+ } else {
+ ios_log(this, logfp, "\"%s.%s.fop.%s.per_sec\": %0.2lf,",
+ key_prefix, str_prefix, lc_fop_name,
+ (double)(fop_hits / interval_sec));
}
- if (total_fop_hits) {
- weighted_fop_ave_usec =
- weighted_fop_ave_usec_sum/total_fop_hits;
- /* Extra key that does not print out an entry w/ 0.00 for
- * intervals with no data
- */
- ios_log (this, logfp,
- "\"%s.%s.fop.weighted_latency_ave_usec_nozerofill\": "
- "\"%0.4lf\",",
- key_prefix, str_prefix, weighted_fop_ave_usec);
- }
- ios_log (this, logfp,
- "\"%s.%s.fop.weighted_latency_ave_usec\": \"%0.4lf\",",
- key_prefix, str_prefix, weighted_fop_ave_usec);
- ios_log (this, logfp,
- "\"%s.%s.fop.weighted_fop_count\": \"%ld\",",
- key_prefix, str_prefix, total_fop_hits);
-
- fop_ave_usec = fop_ave_usec_sum/GF_FOP_MAXVALUE;
- ios_log (this, logfp,
- "\"%s.%s.fop.unweighted_latency_ave_usec\":\"%0.4lf\",",
- key_prefix, str_prefix, fop_ave_usec);
-
- for (i = 0; i < GF_UPCALL_FLAGS_MAXVALUE; i++) {
- lc_fop_name = strdupa (gf_upcall_list[i]);
- for (j = 0; lc_fop_name[j]; j++) {
- lc_fop_name[j] = tolower (lc_fop_name[j]);
- }
- fop_hits = GF_ATOMIC_GET (stats->upcall_hits[i]);
- if (interval == -1) {
- ios_log (this, logfp,
- "\"%s.%s.fop.%s.count\": \"%"GF_PRI_ATOMIC"\",",
- key_prefix, str_prefix, lc_fop_name,
- fop_hits);
- } else {
- ios_log (this, logfp,
- "\"%s.%s.fop.%s.per_sec\": \"%0.2lf\",",
- key_prefix, str_prefix, lc_fop_name,
- (double)(fop_hits / interval_sec));
- }
+ ios_log(this, logfp, "\"%s.%s.fop.%s.latency_ave_usec\": %0.2lf,",
+ key_prefix, str_prefix, lc_fop_name, fop_lat_ave);
+ ios_log(this, logfp, "\"%s.%s.fop.%s.latency_min_usec\": %0.2lf,",
+ key_prefix, str_prefix, lc_fop_name, fop_lat_min);
+ ios_log(this, logfp, "\"%s.%s.fop.%s.latency_max_usec\": %0.2lf,",
+ key_prefix, str_prefix, lc_fop_name, fop_lat_max);
+
+ fop_ave_usec_sum += fop_lat_ave;
+ weighted_fop_ave_usec_sum += fop_hits * fop_lat_ave;
+ total_fop_hits += fop_hits;
+ }
+
+ if (total_fop_hits) {
+ weighted_fop_ave_usec = weighted_fop_ave_usec_sum / total_fop_hits;
+ /* Extra key that does not print out an entry w/ 0.00 for
+ * intervals with no data
+ */
+ ios_log(this, logfp,
+ "\"%s.%s.fop.weighted_latency_ave_usec_nozerofill\": "
+ "%0.4lf,",
+ key_prefix, str_prefix, weighted_fop_ave_usec);
+ }
+ ios_log(this, logfp, "\"%s.%s.fop.weighted_latency_ave_usec\": %0.4lf,",
+ key_prefix, str_prefix, weighted_fop_ave_usec);
+ ios_log(this, logfp, "\"%s.%s.fop.weighted_fop_count\": %ld,", key_prefix,
+ str_prefix, total_fop_hits);
+
+ fop_ave_usec = fop_ave_usec_sum / GF_FOP_MAXVALUE;
+ ios_log(this, logfp, "\"%s.%s.fop.unweighted_latency_ave_usec\":%0.4lf,",
+ key_prefix, str_prefix, fop_ave_usec);
+
+ for (i = 0; i < GF_UPCALL_FLAGS_MAXVALUE; i++) {
+ lc_fop_name = strdupa(gf_upcall_list[i]);
+ for (j = 0; lc_fop_name[j]; j++) {
+ lc_fop_name[j] = tolower(lc_fop_name[j]);
}
-
- ret = syncop_getxattr (this, &unused_loc, &xattr,
- IO_THREADS_QUEUE_SIZE_KEY, NULL, NULL);
- if (xattr) {
- /*
- * Iterate over the dictionary returned to us by io-threads and
- * dump the results to the stats file.
- */
- data_pair_t *curr = NULL;
-
- dict_foreach_inline (xattr, curr) {
- ios_log (this, logfp,
- "\"%s.%s.%s.queue_size\": \"%d\",",
- key_prefix, str_prefix, curr->key,
- data_to_int32 (curr->value));
- }
-
- /* Free the dictionary */
- dict_unref (xattr);
+ fop_hits = GF_ATOMIC_GET(stats->upcall_hits[i]);
+ if (interval == -1) {
+ ios_log(this, logfp, "\"%s.%s.fop.%s.count\": %" GF_PRI_ATOMIC ",",
+ key_prefix, str_prefix, lc_fop_name, fop_hits);
} else {
- gf_log (this->name, GF_LOG_WARNING,
- "Unable to get queue size counts from "
- "the io-threads translator!");
+ ios_log(this, logfp, "\"%s.%s.fop.%s.per_sec\": %0.2lf,",
+ key_prefix, str_prefix, lc_fop_name,
+ (double)(fop_hits / interval_sec));
}
+ }
- if (interval == -1) {
- ios_log (this, logfp, "\"%s.%s.uptime\": \"%"PRId64"\",",
- key_prefix, str_prefix,
- (uint64_t) (now->tv_sec - stats->started_at.tv_sec));
- ios_log (this, logfp, "\"%s.%s.bytes_read\": \""
- "%"GF_PRI_ATOMIC"\",", key_prefix, str_prefix,
- GF_ATOMIC_GET (stats->data_read));
- ios_log (this, logfp, "\"%s.%s.bytes_written\": \""
- "%"GF_PRI_ATOMIC"\"", key_prefix, str_prefix,
- GF_ATOMIC_GET (stats->data_written));
- } else {
- ios_log (this, logfp,
- "\"%s.%s.sample_interval_sec\": \"%0.2lf\",",
- key_prefix, str_prefix,
- interval_sec);
- ios_log (this, logfp,
- "\"%s.%s.bytes_read_per_sec\": \"%0.2lf\",",
- key_prefix, str_prefix, (double)
- (GF_ATOMIC_GET (stats->data_read) / interval_sec));
- ios_log (this, logfp,
- "\"%s.%s.bytes_written_per_sec\": \"%0.2lf\"",
- key_prefix, str_prefix, (double)
- (GF_ATOMIC_GET (stats->data_written) / interval_sec));
+ ret = syncop_getxattr(this, &unused_loc, &xattr, IO_THREADS_QUEUE_SIZE_KEY,
+ NULL, NULL);
+ if (xattr) {
+ /*
+ * Iterate over the dictionary returned to us by io-threads and
+ * dump the results to the stats file.
+ */
+ data_pair_t *curr = NULL;
+
+ dict_foreach_inline(xattr, curr)
+ {
+ ios_log(this, logfp, "\"%s.%s.%s.queue_size\": %d,", key_prefix,
+ str_prefix, curr->key, data_to_int32(curr->value));
}
- ios_log (this, logfp, "}");
- ret = 0;
+ /* Free the dictionary */
+ dict_unref(xattr);
+ } else {
+ gf_log(this->name, GF_LOG_WARNING,
+ "Unable to get queue size counts from "
+ "the io-threads translator!");
+ }
+
+ if (interval == -1) {
+ ios_log(this, logfp, "\"%s.%s.uptime\": %" PRIu64 ",", key_prefix,
+ str_prefix, (uint64_t)(now - stats->started_at));
+ ios_log(this, logfp,
+ "\"%s.%s.bytes_read\": "
+ "%" GF_PRI_ATOMIC ",",
+ key_prefix, str_prefix, GF_ATOMIC_GET(stats->data_read));
+ ios_log(this, logfp,
+ "\"%s.%s.bytes_written\": "
+ "%" GF_PRI_ATOMIC "",
+ key_prefix, str_prefix, GF_ATOMIC_GET(stats->data_written));
+ } else {
+ ios_log(this, logfp, "\"%s.%s.sample_interval_sec\": %0.2lf,",
+ key_prefix, str_prefix, interval_sec);
+ ios_log(this, logfp, "\"%s.%s.bytes_read_per_sec\": %0.2lf,",
+ key_prefix, str_prefix,
+ (double)(GF_ATOMIC_GET(stats->data_read) / interval_sec));
+ ios_log(this, logfp, "\"%s.%s.bytes_written_per_sec\": %0.2lf",
+ key_prefix, str_prefix,
+ (double)(GF_ATOMIC_GET(stats->data_written) / interval_sec));
+ }
+
+ ios_log(this, logfp, "}");
+ ret = 0;
out:
- GF_FREE (key_prefix);
- return ret;
+ GF_FREE(key_prefix);
+ return ret;
}
char *
-_resolve_username (xlator_t *this, uid_t uid)
+_resolve_username(xlator_t *this, uid_t uid)
{
- struct passwd pwd;
- struct passwd *pwd_result = NULL;
- size_t pwd_buf_len;
- char *pwd_buf = NULL;
- char *ret = NULL;
+ struct passwd pwd;
+ struct passwd *pwd_result = NULL;
+ size_t pwd_buf_len;
+ char *pwd_buf = NULL;
+ char *ret = NULL;
- /* Prepare our buffer for the uid->username translation */
+ /* Prepare our buffer for the uid->username translation */
#ifdef _SC_GETGR_R_SIZE_MAX
- pwd_buf_len = sysconf (_SC_GETGR_R_SIZE_MAX);
+ pwd_buf_len = sysconf(_SC_GETGR_R_SIZE_MAX);
#else
- pwd_buf_len = -1;
+ pwd_buf_len = -1;
#endif
- if (pwd_buf_len == -1) {
- pwd_buf_len = DEFAULT_PWD_BUF_SZ; /* per the man page */
- }
+ if (pwd_buf_len == -1) {
+ pwd_buf_len = DEFAULT_PWD_BUF_SZ; /* per the man page */
+ }
- pwd_buf = alloca (pwd_buf_len);
- if (!pwd_buf)
- goto err;
+ pwd_buf = alloca(pwd_buf_len);
+ if (!pwd_buf)
+ goto err;
- getpwuid_r (uid, &pwd, pwd_buf, pwd_buf_len,
- &pwd_result);
- if (!pwd_result)
- goto err;
+ getpwuid_r(uid, &pwd, pwd_buf, pwd_buf_len, &pwd_result);
+ if (!pwd_result)
+ goto err;
- ret = gf_strdup (pwd.pw_name);
- if (ret)
- return ret;
- else
- gf_log (this->name, GF_LOG_ERROR,
- "gf_strdup failed, failing username "
- "resolution.");
-err:
+ ret = gf_strdup(pwd.pw_name);
+ if (ret)
return ret;
+ else
+ gf_log(this->name, GF_LOG_ERROR,
+ "gf_strdup failed, failing username "
+ "resolution.");
+err:
+ return ret;
}
char *
-_resolve_group_name (xlator_t *this, gid_t gid)
+_resolve_group_name(xlator_t *this, gid_t gid)
{
- struct group grp;
- struct group *grp_result = NULL;
- size_t grp_buf_len;
- char *grp_buf = NULL;
- char *ret = NULL;
+ struct group grp;
+ struct group *grp_result = NULL;
+ size_t grp_buf_len;
+ char *grp_buf = NULL;
+ char *ret = NULL;
- /* Prepare our buffer for the gid->group name translation */
+ /* Prepare our buffer for the gid->group name translation */
#ifdef _SC_GETGR_R_SIZE_MAX
- grp_buf_len = sysconf (_SC_GETGR_R_SIZE_MAX);
+ grp_buf_len = sysconf(_SC_GETGR_R_SIZE_MAX);
#else
- grp_buf_len = -1;
+ grp_buf_len = -1;
#endif
- if (grp_buf_len == -1) {
- grp_buf_len = DEFAULT_GRP_BUF_SZ; /* per the man page */
- }
+ if (grp_buf_len == -1) {
+ grp_buf_len = DEFAULT_GRP_BUF_SZ; /* per the man page */
+ }
- grp_buf = alloca (grp_buf_len);
- if (!grp_buf) {
- goto err;
- }
+ grp_buf = alloca(grp_buf_len);
+ if (!grp_buf) {
+ goto err;
+ }
- if (getgrgid_r (gid, &grp, grp_buf, grp_buf_len, &grp_result) != 0)
- goto err;
+ if (getgrgid_r(gid, &grp, grp_buf, grp_buf_len, &grp_result) != 0)
+ goto err;
- if (!grp_result)
- goto err;
+ if (!grp_result)
+ goto err;
- ret = gf_strdup (grp.gr_name);
- if (ret)
- return ret;
- else
- gf_log (this->name, GF_LOG_ERROR,
- "gf_strdup failed, failing username "
- "resolution.");
-err:
+ ret = gf_strdup(grp.gr_name);
+ if (ret)
return ret;
+ else
+ gf_log(this->name, GF_LOG_ERROR,
+ "gf_strdup failed, failing username "
+ "resolution.");
+err:
+ return ret;
}
-
/*
* This function writes out a latency sample to a given file descriptor
* and beautifies the output in the process.
*/
void
-_io_stats_write_latency_sample (xlator_t *this, ios_sample_t *sample,
- FILE *logfp)
-{
- double epoch_time = 0.00;
- char *xlator_name = NULL;
- char *instance_name = NULL;
- char *hostname = NULL;
- char *identifier = NULL;
- char *port = NULL;
- char *port_pos = NULL;
- char *group_name = NULL;
- char *username = NULL;
- struct ios_conf *conf = NULL;
-
- conf = this->private;
-
- epoch_time = (sample->timestamp).tv_sec +
- ((sample->timestamp).tv_usec / 1000000.0);
-
- if (strlen (sample->identifier) == 0) {
- hostname = "Unknown";
- port = "Unknown";
- } else {
- identifier = strdupa (sample->identifier);
- port_pos = strrchr (identifier, ':');
- if (!port_pos || strlen(port_pos) < 2)
- goto err;
- port = strdupa (port_pos + 1);
- if (!port)
- goto err;
- *port_pos = '\0';
- hostname = gf_rev_dns_lookup_cached (identifier,
- conf->dnscache);
- if (!hostname)
- hostname = "Unknown";
- }
-
- xlator_name = conf->unique_id;
- if (!xlator_name || strlen (xlator_name) == 0)
- xlator_name = "Unknown";
-
- instance_name = this->instance_name;
- if (!instance_name || strlen (instance_name) == 0)
- instance_name = "N/A";
-
- /* Resolve the UID to a string username */
- username = _resolve_username (this, sample->uid);
+_io_stats_write_latency_sample(xlator_t *this, ios_sample_t *sample,
+ FILE *logfp)
+{
+ double epoch_time = 0.00;
+ char *xlator_name = NULL;
+ char *instance_name = NULL;
+ char *hostname = NULL;
+ char *identifier = NULL;
+ char *port = NULL;
+ char *port_pos = NULL;
+ char *group_name = NULL;
+ char *username = NULL;
+ struct ios_conf *conf = NULL;
+
+ conf = this->private;
+
+ epoch_time = (sample->timestamp).tv_sec +
+ ((sample->timestamp).tv_usec / 1000000.0);
+
+ if (strlen(sample->identifier) == 0) {
+ hostname = "Unknown";
+ port = "Unknown";
+ } else {
+ identifier = strdupa(sample->identifier);
+ port_pos = strrchr(identifier, ':');
+ if (!port_pos || strlen(port_pos) < 2)
+ goto err;
+ port = strdupa(port_pos + 1);
+ if (!port)
+ goto err;
+ *port_pos = '\0';
+ hostname = gf_rev_dns_lookup_cached(identifier, conf->dnscache);
+ if (!hostname)
+ hostname = "Unknown";
+ }
+
+ xlator_name = conf->unique_id;
+ if (!xlator_name || strlen(xlator_name) == 0)
+ xlator_name = "Unknown";
+
+ instance_name = this->instance_name;
+ if (!instance_name || strlen(instance_name) == 0)
+ instance_name = "N/A";
+
+ /* Resolve the UID to a string username */
+ username = _resolve_username(this, sample->uid);
+ if (!username) {
+ username = GF_MALLOC(30, gf_common_mt_char);
if (!username) {
- username = GF_MALLOC (30, gf_common_mt_char);
- if (!username) {
- goto out;
- }
- sprintf (username, "%d", (int32_t)sample->uid);
+ goto out;
}
+ sprintf(username, "%d", (int32_t)sample->uid);
+ }
- /* Resolve the GID to a string group name */
- group_name = _resolve_group_name (this, sample->gid);
+ /* Resolve the GID to a string group name */
+ group_name = _resolve_group_name(this, sample->gid);
+ if (!group_name) {
+ group_name = GF_MALLOC(30, gf_common_mt_char);
if (!group_name) {
- group_name = GF_MALLOC (30, gf_common_mt_char);
- if (!group_name) {
- goto out;
- }
- sprintf (group_name, "%d", (int32_t)sample->gid);
+ goto out;
}
-
- ios_log (this, logfp,
- "%0.6lf,%s,%s,%0.4lf,%s,%s,%s,%s,%s,%s",
- epoch_time, fop_enum_to_pri_string (sample->fop_type),
- gf_fop_string (sample->fop_type),
- sample->elapsed, xlator_name, instance_name, username,
- group_name, hostname, port);
- goto out;
+ sprintf(group_name, "%d", (int32_t)sample->gid);
+ }
+
+ ios_log(this, logfp, "%0.6lf,%s,%s,%0.4lf,%s,%s,%s,%s,%s,%s", epoch_time,
+ fop_enum_to_pri_string(sample->fop_type),
+ gf_fop_string(sample->fop_type), sample->elapsed, xlator_name,
+ instance_name, username, group_name, hostname, port);
+ goto out;
err:
- gf_log (this->name, GF_LOG_ERROR,
- "Error parsing socket identifier");
+ gf_log(this->name, GF_LOG_ERROR, "Error parsing socket identifier");
out:
- GF_FREE (group_name);
- GF_FREE (username);
+ GF_FREE(group_name);
+ GF_FREE(username);
}
/*
@@ -1181,3409 +1138,3343 @@ out:
* contents of the saved reference.
*/
int
-io_stats_dump_latency_samples_logfp (xlator_t *this, FILE *logfp)
-{
- uint64_t i = 0;
- struct ios_conf *conf = NULL;
- ios_sample_buf_t *sample_buf = NULL;
- int ret = 1; /* Default to error */
-
- conf = this->private;
-
- /* Save pointer to old buffer; the CS equivalent of
- * Indiana Jones: https://www.youtube.com/watch?v=Pr-8AP0To4k,
- * though ours will end better I hope!
- */
- sample_buf = conf->ios_sample_buf;
- if (!sample_buf) {
- gf_log (this->name, GF_LOG_WARNING,
- "Sampling buffer is null, bailing!");
- goto out;
- }
-
- /* Empty case, nothing to do, exit. */
- if (sample_buf->collected == 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "No samples, dump not required.");
- ret = 0;
- goto out;
- }
-
- /* Init a new buffer, so we are free to work on the one we saved a
- * reference to above.
- */
- if (ios_init_sample_buf (conf) != 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "Failed to init new sampling buffer, out of memory?");
- goto out;
- }
+io_stats_dump_latency_samples_logfp(xlator_t *this, FILE *logfp)
+{
+ uint64_t i = 0;
+ struct ios_conf *conf = NULL;
+ ios_sample_buf_t *sample_buf = NULL;
+ int ret = 1; /* Default to error */
+
+ conf = this->private;
+
+ /* Save pointer to old buffer; the CS equivalent of
+ * Indiana Jones: https://www.youtube.com/watch?v=Pr-8AP0To4k,
+ * though ours will end better I hope!
+ */
+ sample_buf = conf->ios_sample_buf;
+ if (!sample_buf) {
+ gf_log(this->name, GF_LOG_WARNING, "Sampling buffer is null, bailing!");
+ goto out;
+ }
- /* Wrap-around case, dump from pos to sample_buf->size -1
- * and then from 0 to sample_buf->pos (covered off by
- * "simple case")
- */
- if (sample_buf->collected > sample_buf->pos + 1) {
- for (i = sample_buf->pos; i < sample_buf->size; i++) {
- _io_stats_write_latency_sample (this,
- &(sample_buf->ios_samples[i]), logfp);
- }
+ /* Empty case, nothing to do, exit. */
+ if (sample_buf->collected == 0) {
+ gf_log(this->name, GF_LOG_DEBUG, "No samples, dump not required.");
+ ret = 0;
+ goto out;
+ }
+
+ /* Init a new buffer, so we are free to work on the one we saved a
+ * reference to above.
+ */
+ if (ios_init_sample_buf(conf) != 0) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "Failed to init new sampling buffer, out of memory?");
+ goto out;
+ }
+
+ /* Wrap-around case, dump from pos to sample_buf->size -1
+ * and then from 0 to sample_buf->pos (covered off by
+ * "simple case")
+ */
+ if (sample_buf->collected > sample_buf->pos + 1) {
+ for (i = sample_buf->pos; i < sample_buf->size; i++) {
+ _io_stats_write_latency_sample(this, &(sample_buf->ios_samples[i]),
+ logfp);
}
+ }
- /* Simple case: Dump from 0 to sample_buf->pos */
- for (i = 0; i < sample_buf->pos; i++) {
- _io_stats_write_latency_sample (this,
- &(sample_buf->ios_samples[i]), logfp);
- }
- ios_destroy_sample_buf (sample_buf);
+ /* Simple case: Dump from 0 to sample_buf->pos */
+ for (i = 0; i < sample_buf->pos; i++) {
+ _io_stats_write_latency_sample(this, &(sample_buf->ios_samples[i]),
+ logfp);
+ }
+ ios_destroy_sample_buf(sample_buf);
out:
- return ret;
+ return ret;
}
int
-io_stats_dump_global_to_logfp (xlator_t *this, struct ios_global_stats *stats,
- struct timeval *now, int interval, FILE *logfp)
-{
- int i = 0;
- int per_line = 0;
- int index = 0;
- struct ios_stat_head *list_head = NULL;
- struct ios_conf *conf = NULL;
- char timestr[256] = {0, };
- char str_header[128] = {0};
- char str_read[128] = {0};
- char str_write[128] = {0};
- uint64_t fop_hits = 0;
- uint64_t block_count_read = 0;
- uint64_t block_count_write = 0;
-
- conf = this->private;
-
- if (interval == -1)
- ios_log (this, logfp, "\n=== Cumulative stats ===");
+io_stats_dump_global_to_logfp(xlator_t *this, struct ios_global_stats *stats,
+ time_t now, int interval, FILE *logfp)
+{
+ int i = 0;
+ int per_line = 0;
+ int index = 0;
+ struct ios_stat_head *list_head = NULL;
+ struct ios_conf *conf = NULL;
+ char timestr[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ char str_header[128] = {0};
+ char str_read[128] = {0};
+ char str_write[128] = {0};
+ uint64_t fop_hits = 0;
+ uint64_t block_count_read = 0;
+ uint64_t block_count_write = 0;
+
+ conf = this->private;
+
+ if (interval == -1)
+ ios_log(this, logfp, "\n=== Cumulative stats ===");
+ else
+ ios_log(this, logfp, "\n=== Interval %d stats ===", interval);
+ ios_log(this, logfp, " Duration : %" PRIu64 " secs",
+ (uint64_t)(now - stats->started_at));
+ ios_log(this, logfp, " BytesRead : %" GF_PRI_ATOMIC,
+ GF_ATOMIC_GET(stats->data_read));
+ ios_log(this, logfp, " BytesWritten : %" GF_PRI_ATOMIC "\n",
+ GF_ATOMIC_GET(stats->data_written));
+
+ snprintf(str_header, sizeof(str_header), "%-12s %c", "Block Size", ':');
+ snprintf(str_read, sizeof(str_read), "%-12s %c", "Read Count", ':');
+ snprintf(str_write, sizeof(str_write), "%-12s %c", "Write Count", ':');
+ index = 14;
+ for (i = 0; i < IOS_BLOCK_COUNT_SIZE; i++) {
+ block_count_read = GF_ATOMIC_GET(stats->block_count_read[i]);
+ block_count_write = GF_ATOMIC_GET(stats->block_count_write[i]);
+ if ((block_count_read == 0) && (block_count_write == 0))
+ continue;
+ per_line++;
+
+ snprintf(str_header + index, sizeof(str_header) - index, "%16dB+",
+ (1 << i));
+ if (block_count_read)
+ snprintf(str_read + index, sizeof(str_read) - index, "%18" PRId64,
+ block_count_read);
else
- ios_log (this, logfp, "\n=== Interval %d stats ===",
- interval);
- ios_log (this, logfp, " Duration : %"PRId64" secs",
- (uint64_t) (now->tv_sec - stats->started_at.tv_sec));
- ios_log (this, logfp, " BytesRead : %"GF_PRI_ATOMIC,
- GF_ATOMIC_GET (stats->data_read));
- ios_log (this, logfp, " BytesWritten : %"GF_PRI_ATOMIC"\n",
- GF_ATOMIC_GET (stats->data_written));
-
- snprintf (str_header, sizeof (str_header), "%-12s %c", "Block Size", ':');
- snprintf (str_read, sizeof (str_read), "%-12s %c", "Read Count", ':');
- snprintf (str_write, sizeof (str_write), "%-12s %c", "Write Count", ':');
- index = 14;
- for (i = 0; i < IOS_BLOCK_COUNT_SIZE; i++) {
- block_count_read = GF_ATOMIC_GET (stats->block_count_read[i]);
- block_count_write = GF_ATOMIC_GET (stats->block_count_write[i]);
- if ((block_count_read == 0) && (block_count_write == 0))
- continue;
- per_line++;
-
- snprintf (str_header+index, sizeof (str_header)-index,
- "%16dB+", (1<<i));
- if (block_count_read)
- snprintf (str_read+index, sizeof (str_read)-index,
- "%18"PRId64, block_count_read);
- else snprintf (str_read+index, sizeof (str_read)-index,
- "%18s", "0");
- if (block_count_write)
- snprintf (str_write+index, sizeof (str_write)-index,
- "%18"GF_PRI_ATOMIC, block_count_write);
- else snprintf (str_write+index, sizeof (str_write)-index,
- "%18s", "0");
-
- index += 18;
- if (per_line == 3) {
- ios_log (this, logfp, "%s", str_header);
- ios_log (this, logfp, "%s", str_read);
- ios_log (this, logfp, "%s\n", str_write);
-
- memset (str_header, 0, sizeof (str_header));
- memset (str_read, 0, sizeof (str_read));
- memset (str_write, 0, sizeof (str_write));
-
- snprintf (str_header, sizeof (str_header), "%-12s %c",
- "Block Size", ':');
- snprintf (str_read, sizeof (str_read), "%-12s %c",
- "Read Count", ':');
- snprintf (str_write, sizeof (str_write), "%-12s %c",
- "Write Count", ':');
-
- index = 14;
- per_line = 0;
- }
- }
-
- if (per_line != 0) {
- ios_log (this, logfp, "%s", str_header);
- ios_log (this, logfp, "%s", str_read);
- ios_log (this, logfp, "%s\n", str_write);
- }
-
- ios_log (this, logfp, "%-13s %10s %14s %14s %14s", "Fop",
- "Call Count", "Avg-Latency", "Min-Latency",
- "Max-Latency");
- ios_log (this, logfp, "%-13s %10s %14s %14s %14s", "---", "----------",
- "-----------", "-----------", "-----------");
-
- for (i = 0; i < GF_FOP_MAXVALUE; i++) {
- fop_hits = GF_ATOMIC_GET (stats->fop_hits[i]);
- if (fop_hits && !stats->latency[i].avg)
- ios_log (this, logfp, "%-13s %10"GF_PRI_ATOMIC" %11s "
- "us %11s us %11s us", gf_fop_list[i],
- fop_hits, "0", "0", "0");
- else if (fop_hits && stats->latency[i].avg)
- ios_log (this, logfp, "%-13s %10"GF_PRI_ATOMIC" "
- "%11.2lf us %11.2lf us %11.2lf us",
- gf_fop_list[i], fop_hits,
- stats->latency[i].avg, stats->latency[i].min,
- stats->latency[i].max);
- }
-
- for (i = 0; i < GF_UPCALL_FLAGS_MAXVALUE; i++) {
- fop_hits = GF_ATOMIC_GET (stats->upcall_hits[i]);
- if (fop_hits)
- ios_log (this, logfp, "%-13s %10"PRId64" %11s "
- "us %11s us %11s us", gf_upcall_list[i],
- fop_hits, "0", "0", "0");
+ snprintf(str_read + index, sizeof(str_read) - index, "%18s", "0");
+ if (block_count_write)
+ snprintf(str_write + index, sizeof(str_write) - index,
+ "%18" GF_PRI_ATOMIC, block_count_write);
+ else
+ snprintf(str_write + index, sizeof(str_write) - index, "%18s", "0");
+
+ index += 18;
+ if (per_line == 3) {
+ ios_log(this, logfp, "%s", str_header);
+ ios_log(this, logfp, "%s", str_read);
+ ios_log(this, logfp, "%s\n", str_write);
+
+ snprintf(str_header, sizeof(str_header), "%-12s %c", "Block Size",
+ ':');
+ snprintf(str_read, sizeof(str_read), "%-12s %c", "Read Count", ':');
+ snprintf(str_write, sizeof(str_write), "%-12s %c", "Write Count",
+ ':');
+
+ index = 14;
+ per_line = 0;
}
-
- ios_log (this, logfp, "------ ----- ----- ----- ----- ----- ----- ----- "
- " ----- ----- ----- -----\n");
-
- if (interval == -1) {
- LOCK (&conf->lock);
- {
- gf_time_fmt (timestr, sizeof timestr,
- conf->cumulative.max_openfd_time.tv_sec,
- gf_timefmt_FT);
- snprintf (timestr + strlen (timestr), sizeof timestr - strlen (timestr),
- ".%"GF_PRI_SUSECONDS,
- conf->cumulative.max_openfd_time.tv_usec);
- ios_log (this, logfp, "Current open fd's: %"PRId64
- " Max open fd's: %"PRId64" time %s",
- conf->cumulative.nr_opens,
- conf->cumulative.max_nr_opens, timestr);
- }
- UNLOCK (&conf->lock);
- ios_log (this, logfp, "\n==========Open File Stats========");
- ios_log (this, logfp, "\nCOUNT: \t FILE NAME");
- list_head = &conf->list[IOS_STATS_TYPE_OPEN];
- ios_dump_file_stats (list_head, this, logfp);
-
-
- ios_log (this, logfp, "\n==========Read File Stats========");
- ios_log (this, logfp, "\nCOUNT: \t FILE NAME");
- list_head = &conf->list[IOS_STATS_TYPE_READ];
- ios_dump_file_stats (list_head, this, logfp);
-
- ios_log (this, logfp, "\n==========Write File Stats========");
- ios_log (this, logfp, "\nCOUNT: \t FILE NAME");
- list_head = &conf->list[IOS_STATS_TYPE_WRITE];
- ios_dump_file_stats (list_head, this, logfp);
-
- ios_log (this, logfp, "\n==========Directory open stats========");
- ios_log (this, logfp, "\nCOUNT: \t DIRECTORY NAME");
- list_head = &conf->list[IOS_STATS_TYPE_OPENDIR];
- ios_dump_file_stats (list_head, this, logfp);
-
- ios_log (this, logfp, "\n========Directory readdirp Stats=======");
- ios_log (this, logfp, "\nCOUNT: \t DIRECTORY NAME");
- list_head = &conf->list[IOS_STATS_TYPE_READDIRP];
- ios_dump_file_stats (list_head, this, logfp);
-
- ios_log (this, logfp, "\n========Read Throughput File Stats=====");
- ios_log (this, logfp, "\nTIMESTAMP \t\t\t THROUGHPUT(KBPS)"
- "\tFILE NAME");
- list_head = &conf->thru_list[IOS_STATS_THRU_READ];
- ios_dump_throughput_stats(list_head, this, logfp,
- IOS_STATS_THRU_READ);
-
- ios_log (this, logfp, "\n======Write Throughput File Stats======");
- ios_log (this, logfp, "\nTIMESTAMP \t\t\t THROUGHPUT(KBPS)"
- "\tFILE NAME");
- list_head = &conf->thru_list[IOS_STATS_THRU_WRITE];
- ios_dump_throughput_stats (list_head, this, logfp,
- IOS_STATS_THRU_WRITE);
+ }
+
+ if (per_line != 0) {
+ ios_log(this, logfp, "%s", str_header);
+ ios_log(this, logfp, "%s", str_read);
+ ios_log(this, logfp, "%s\n", str_write);
+ }
+
+ ios_log(this, logfp, "%-13s %10s %14s %14s %14s", "Fop", "Call Count",
+ "Avg-Latency", "Min-Latency", "Max-Latency");
+ ios_log(this, logfp, "%-13s %10s %14s %14s %14s", "---", "----------",
+ "-----------", "-----------", "-----------");
+
+ for (i = 0; i < GF_FOP_MAXVALUE; i++) {
+ fop_hits = GF_ATOMIC_GET(stats->fop_hits[i]);
+ if (fop_hits && !stats->latency[i].avg)
+ ios_log(this, logfp,
+ "%-13s %10" GF_PRI_ATOMIC
+ " %11s "
+ "us %11s us %11s us",
+ gf_fop_list[i], fop_hits, "0", "0", "0");
+ else if (fop_hits && stats->latency[i].avg)
+ ios_log(this, logfp,
+ "%-13s %10" GF_PRI_ATOMIC
+ " "
+ "%11.2lf us %11.2lf us %11.2lf us",
+ gf_fop_list[i], fop_hits, stats->latency[i].avg,
+ stats->latency[i].min, stats->latency[i].max);
+ }
+
+ for (i = 0; i < GF_UPCALL_FLAGS_MAXVALUE; i++) {
+ fop_hits = GF_ATOMIC_GET(stats->upcall_hits[i]);
+ if (fop_hits)
+ ios_log(this, logfp,
+ "%-13s %10" PRId64
+ " %11s "
+ "us %11s us %11s us",
+ gf_upcall_list[i], fop_hits, "0", "0", "0");
+ }
+
+ ios_log(this, logfp,
+ "------ ----- ----- ----- ----- ----- ----- ----- "
+ " ----- ----- ----- -----\n");
+
+ if (interval == -1) {
+ LOCK(&conf->lock);
+ {
+ gf_time_fmt_tv(timestr, sizeof timestr,
+ &conf->cumulative.max_openfd_time, gf_timefmt_FT);
+ ios_log(this, logfp,
+ "Current open fd's: %" PRId64 " Max open fd's: %" PRId64
+ " time %s",
+ conf->cumulative.nr_opens, conf->cumulative.max_nr_opens,
+ timestr);
}
- return 0;
+ UNLOCK(&conf->lock);
+ ios_log(this, logfp, "\n==========Open File Stats========");
+ ios_log(this, logfp, "\nCOUNT: \t FILE NAME");
+ list_head = &conf->list[IOS_STATS_TYPE_OPEN];
+ ios_dump_file_stats(list_head, this, logfp);
+
+ ios_log(this, logfp, "\n==========Read File Stats========");
+ ios_log(this, logfp, "\nCOUNT: \t FILE NAME");
+ list_head = &conf->list[IOS_STATS_TYPE_READ];
+ ios_dump_file_stats(list_head, this, logfp);
+
+ ios_log(this, logfp, "\n==========Write File Stats========");
+ ios_log(this, logfp, "\nCOUNT: \t FILE NAME");
+ list_head = &conf->list[IOS_STATS_TYPE_WRITE];
+ ios_dump_file_stats(list_head, this, logfp);
+
+ ios_log(this, logfp, "\n==========Directory open stats========");
+ ios_log(this, logfp, "\nCOUNT: \t DIRECTORY NAME");
+ list_head = &conf->list[IOS_STATS_TYPE_OPENDIR];
+ ios_dump_file_stats(list_head, this, logfp);
+
+ ios_log(this, logfp, "\n========Directory readdirp Stats=======");
+ ios_log(this, logfp, "\nCOUNT: \t DIRECTORY NAME");
+ list_head = &conf->list[IOS_STATS_TYPE_READDIRP];
+ ios_dump_file_stats(list_head, this, logfp);
+
+ ios_log(this, logfp, "\n========Read Throughput File Stats=====");
+ ios_log(this, logfp,
+ "\nTIMESTAMP \t\t\t THROUGHPUT(KBPS)"
+ "\tFILE NAME");
+ list_head = &conf->thru_list[IOS_STATS_THRU_READ];
+ ios_dump_throughput_stats(list_head, this, logfp, IOS_STATS_THRU_READ);
+
+ ios_log(this, logfp, "\n======Write Throughput File Stats======");
+ ios_log(this, logfp,
+ "\nTIMESTAMP \t\t\t THROUGHPUT(KBPS)"
+ "\tFILE NAME");
+ list_head = &conf->thru_list[IOS_STATS_THRU_WRITE];
+ ios_dump_throughput_stats(list_head, this, logfp, IOS_STATS_THRU_WRITE);
+ }
+ return 0;
}
int
-io_stats_dump_global_to_dict (xlator_t *this, struct ios_global_stats *stats,
- struct timeval *now, int interval, dict_t *dict)
-{
- int ret = 0;
- char key[256] = {0};
- uint64_t sec = 0;
- int i = 0;
- uint64_t count = 0;
- uint64_t fop_hits = 0;
-
- GF_ASSERT (stats);
- GF_ASSERT (now);
- GF_ASSERT (dict);
- GF_ASSERT (this);
-
- if (interval == -1)
- snprintf (key, sizeof (key), "cumulative");
- else
- snprintf (key, sizeof (key), "interval");
- ret = dict_set_int32 (dict, key, interval);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR, "failed to set "
- "interval %d", interval);
-
- memset (key, 0, sizeof (key));
- snprintf (key, sizeof (key), "%d-duration", interval);
- sec = (uint64_t) (now->tv_sec - stats->started_at.tv_sec);
- ret = dict_set_uint64 (dict, key, sec);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "failed to set "
- "duration(%d) - %"PRId64, interval, sec);
+io_stats_dump_global_to_dict(xlator_t *this, struct ios_global_stats *stats,
+ time_t now, int interval, dict_t *dict)
+{
+ int ret = 0;
+ char key[64] = {0};
+ uint64_t sec = 0;
+ int i = 0;
+ uint64_t count = 0;
+ uint64_t fop_hits = 0;
+
+ GF_ASSERT(stats);
+ GF_ASSERT(now);
+ GF_ASSERT(dict);
+ GF_ASSERT(this);
+
+ if (interval == -1)
+ snprintf(key, sizeof(key), "cumulative");
+ else
+ snprintf(key, sizeof(key), "interval");
+ ret = dict_set_int32(dict, key, interval);
+ if (ret)
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to set "
+ "interval %d",
+ interval);
+
+ snprintf(key, sizeof(key), "%d-duration", interval);
+ sec = now - stats->started_at;
+ ret = dict_set_uint64(dict, key, sec);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to set "
+ "duration(%d) - %" PRId64,
+ interval, sec);
+ goto out;
+ }
+
+ snprintf(key, sizeof(key), "%d-total-read", interval);
+ ret = dict_set_uint64(dict, key, GF_ATOMIC_GET(stats->data_read));
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to set total "
+ "read(%d) - %" GF_PRI_ATOMIC,
+ interval, GF_ATOMIC_GET(stats->data_read));
+ goto out;
+ }
+
+ snprintf(key, sizeof(key), "%d-total-write", interval);
+ ret = dict_set_uint64(dict, key, GF_ATOMIC_GET(stats->data_written));
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to set total "
+ "write(%d) - %" GF_PRI_ATOMIC,
+ interval, GF_ATOMIC_GET(stats->data_written));
+ goto out;
+ }
+ for (i = 0; i < 32; i++) {
+ count = GF_ATOMIC_GET(stats->block_count_read[i]);
+ if (count) {
+ snprintf(key, sizeof(key), "%d-read-%d", interval, (1 << i));
+ ret = dict_set_uint64(dict, key, count);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to "
+ "set read-%db+, with: %" PRId64,
+ (1 << i), count);
goto out;
+ }
}
-
- memset (key, 0, sizeof (key));
- snprintf (key, sizeof (key), "%d-total-read", interval);
- ret = dict_set_uint64 (dict, key, GF_ATOMIC_GET (stats->data_read));
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "failed to set total "
- "read(%d) - %"GF_PRI_ATOMIC, interval,
- GF_ATOMIC_GET (stats->data_read));
+ }
+
+ for (i = 0; i < IOS_BLOCK_COUNT_SIZE; i++) {
+ count = GF_ATOMIC_GET(stats->block_count_write[i]);
+ if (count) {
+ snprintf(key, sizeof(key), "%d-write-%d", interval, (1 << i));
+ ret = dict_set_uint64(dict, key, count);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to "
+ "set write-%db+, with: %" PRId64,
+ (1 << i), count);
goto out;
+ }
}
-
- memset (key, 0, sizeof (key));
- snprintf (key, sizeof (key), "%d-total-write", interval);
- ret = dict_set_uint64 (dict, key, GF_ATOMIC_GET (stats->data_written));
+ }
+
+ for (i = 0; i < GF_FOP_MAXVALUE; i++) {
+ fop_hits = GF_ATOMIC_GET(stats->fop_hits[i]);
+ if (fop_hits == 0)
+ continue;
+ snprintf(key, sizeof(key), "%d-%d-hits", interval, i);
+ ret = dict_set_uint64(dict, key, fop_hits);
if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "failed to set total "
- "write(%d) - %"GF_PRI_ATOMIC, interval,
- GF_ATOMIC_GET (stats->data_written));
- goto out;
- }
- for (i = 0; i < 32; i++) {
- count = GF_ATOMIC_GET (stats->block_count_read[i]);
- if (count) {
- memset (key, 0, sizeof (key));
- snprintf (key, sizeof (key), "%d-read-%d", interval,
- (1 << i));
- ret = dict_set_uint64 (dict, key, count);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "failed to "
- "set read-%db+, with: %"PRId64,
- (1<<i), count);
- goto out;
- }
- }
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to set "
+ "%s-fop-hits: %" GF_PRI_ATOMIC,
+ gf_fop_list[i], fop_hits);
+ goto out;
}
- for (i = 0; i < IOS_BLOCK_COUNT_SIZE; i++) {
- count = GF_ATOMIC_GET (stats->block_count_write[i]);
- if (count) {
- snprintf (key, sizeof (key), "%d-write-%d", interval,
- (1<<i));
- ret = dict_set_uint64 (dict, key, count);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "failed to "
- "set write-%db+, with: %"PRId64,
- (1<<i), count);
- goto out;
- }
- }
+ if (stats->latency[i].avg == 0)
+ continue;
+ snprintf(key, sizeof(key), "%d-%d-avglatency", interval, i);
+ ret = dict_set_double(dict, key, stats->latency[i].avg);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to set %s "
+ "avglatency(%d) with %f",
+ gf_fop_list[i], interval, stats->latency[i].avg);
+ goto out;
}
-
- for (i = 0; i < GF_FOP_MAXVALUE; i++) {
- fop_hits = GF_ATOMIC_GET (stats->fop_hits[i]);
- if (fop_hits == 0)
- continue;
- snprintf (key, sizeof (key), "%d-%d-hits", interval, i);
- ret = dict_set_uint64 (dict, key, fop_hits);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "failed to set "
- "%s-fop-hits: %"GF_PRI_ATOMIC, gf_fop_list[i],
- fop_hits);
- goto out;
- }
-
- if (stats->latency[i].avg == 0)
- continue;
- snprintf (key, sizeof (key), "%d-%d-avglatency", interval, i);
- ret = dict_set_double (dict, key, stats->latency[i].avg);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "failed to set %s "
- "avglatency(%d) with %f", gf_fop_list[i],
- interval, stats->latency[i].avg);
- goto out;
- }
- snprintf (key, sizeof (key), "%d-%d-minlatency", interval, i);
- ret = dict_set_double (dict, key, stats->latency[i].min);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "failed to set %s "
- "minlatency(%d) with %f", gf_fop_list[i],
- interval, stats->latency[i].min);
- goto out;
- }
- snprintf (key, sizeof (key), "%d-%d-maxlatency", interval, i);
- ret = dict_set_double (dict, key, stats->latency[i].max);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "failed to set %s "
- "maxlatency(%d) with %f", gf_fop_list[i],
- interval, stats->latency[i].max);
- goto out;
- }
+ snprintf(key, sizeof(key), "%d-%d-minlatency", interval, i);
+ ret = dict_set_double(dict, key, stats->latency[i].min);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to set %s "
+ "minlatency(%d) with %f",
+ gf_fop_list[i], interval, stats->latency[i].min);
+ goto out;
}
- for (i = 0; i < GF_UPCALL_FLAGS_MAXVALUE; i++) {
- fop_hits = GF_ATOMIC_GET (stats->upcall_hits[i]);
- if (fop_hits == 0)
- continue;
- snprintf (key, sizeof (key), "%d-%d-upcall-hits", interval, i);
- ret = dict_set_uint64 (dict, key, fop_hits);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "failed to "
- "set %s-upcall-hits: %"PRIu64,
- gf_upcall_list[i], fop_hits);
- goto out;
- }
+ snprintf(key, sizeof(key), "%d-%d-maxlatency", interval, i);
+ ret = dict_set_double(dict, key, stats->latency[i].max);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to set %s "
+ "maxlatency(%d) with %f",
+ gf_fop_list[i], interval, stats->latency[i].max);
+ goto out;
}
+ }
+ for (i = 0; i < GF_UPCALL_FLAGS_MAXVALUE; i++) {
+ fop_hits = GF_ATOMIC_GET(stats->upcall_hits[i]);
+ if (fop_hits == 0)
+ continue;
+ snprintf(key, sizeof(key), "%d-%d-upcall-hits", interval, i);
+ ret = dict_set_uint64(dict, key, fop_hits);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to "
+ "set %s-upcall-hits: %" PRIu64,
+ gf_upcall_list[i], fop_hits);
+ goto out;
+ }
+ }
out:
- gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret);
- return ret;
+ gf_log(this->name, GF_LOG_DEBUG, "returning %d", ret);
+ return ret;
}
int
-io_stats_dump_global (xlator_t *this, struct ios_global_stats *stats,
- struct timeval *now, int interval,
- struct ios_dump_args *args)
+io_stats_dump_global(xlator_t *this, struct ios_global_stats *stats, time_t now,
+ int interval, struct ios_dump_args *args)
{
- int ret = -1;
-
- GF_ASSERT (args);
- GF_ASSERT (now);
- GF_ASSERT (stats);
- GF_ASSERT (this);
-
+ int ret = -1;
+ GF_ASSERT(args);
+ GF_ASSERT(now);
+ GF_ASSERT(stats);
+ GF_ASSERT(this);
- switch (args->type) {
+ switch (args->type) {
case IOS_DUMP_TYPE_JSON_FILE:
- ret = io_stats_dump_global_to_json_logfp (
- this, stats, now, interval, args->u.logfp);
- break;
+ ret = io_stats_dump_global_to_json_logfp(this, stats, now, interval,
+ args->u.logfp);
+ break;
case IOS_DUMP_TYPE_FILE:
- ret = io_stats_dump_global_to_logfp (this, stats, now,
- interval, args->u.logfp);
- break;
+ ret = io_stats_dump_global_to_logfp(this, stats, now, interval,
+ args->u.logfp);
+ break;
case IOS_DUMP_TYPE_DICT:
- ret = io_stats_dump_global_to_dict (this, stats, now,
- interval, args->u.dict);
- break;
+ ret = io_stats_dump_global_to_dict(this, stats, now, interval,
+ args->u.dict);
+ break;
default:
- GF_ASSERT (0);
- ret = -1;
- break;
- }
- return ret;
+ GF_ASSERT(0);
+ ret = -1;
+ break;
+ }
+ return ret;
}
int
-ios_dump_args_init (struct ios_dump_args *args, ios_dump_type_t type,
- void *output)
+ios_dump_args_init(struct ios_dump_args *args, ios_dump_type_t type,
+ void *output)
{
- int ret = 0;
+ int ret = 0;
- GF_ASSERT (args);
- GF_ASSERT (type > IOS_DUMP_TYPE_NONE && type < IOS_DUMP_TYPE_MAX);
- GF_ASSERT (output);
+ GF_ASSERT(args);
+ GF_ASSERT(type > IOS_DUMP_TYPE_NONE && type < IOS_DUMP_TYPE_MAX);
+ GF_ASSERT(output);
- args->type = type;
- switch (args->type) {
+ args->type = type;
+ switch (args->type) {
case IOS_DUMP_TYPE_JSON_FILE:
case IOS_DUMP_TYPE_FILE:
- args->u.logfp = output;
- break;
+ args->u.logfp = output;
+ break;
case IOS_DUMP_TYPE_DICT:
- args->u.dict = output;
- break;
+ args->u.dict = output;
+ break;
default:
- GF_ASSERT (0);
- ret = -1;
- }
+ GF_ASSERT(0);
+ ret = -1;
+ }
- return ret;
+ return ret;
}
static void
-ios_global_stats_clear (struct ios_global_stats *stats, struct timeval *now)
+ios_global_stats_clear(struct ios_global_stats *stats, time_t now)
{
- GF_ASSERT (stats);
- GF_ASSERT (now);
+ GF_ASSERT(stats);
+ GF_ASSERT(now);
- memset (stats, 0, sizeof (*stats));
- stats->started_at = *now;
+ memset(stats, 0, sizeof(*stats));
+ stats->started_at = now;
}
int
-io_stats_dump (xlator_t *this, struct ios_dump_args *args,
- gf1_cli_info_op op, gf_boolean_t is_peek)
+io_stats_dump(xlator_t *this, struct ios_dump_args *args, ios_info_op_t op,
+ gf_boolean_t is_peek)
{
- struct ios_conf *conf = NULL;
- struct ios_global_stats cumulative = { };
- struct ios_global_stats incremental = { };
- int increment = 0;
- struct timeval now;
+ struct ios_conf *conf = NULL;
+ struct ios_global_stats cumulative = {};
+ struct ios_global_stats incremental = {};
+ int increment = 0;
+ time_t now = 0;
- GF_ASSERT (this);
- GF_ASSERT (args);
- GF_ASSERT (args->type > IOS_DUMP_TYPE_NONE);
- GF_ASSERT (args->type < IOS_DUMP_TYPE_MAX);
+ GF_ASSERT(this);
+ GF_ASSERT(args);
+ GF_ASSERT(args->type > IOS_DUMP_TYPE_NONE);
+ GF_ASSERT(args->type < IOS_DUMP_TYPE_MAX);
- conf = this->private;
+ conf = this->private;
+ now = gf_time();
- gettimeofday (&now, NULL);
- LOCK (&conf->lock);
- {
- if (op == GF_CLI_INFO_ALL ||
- op == GF_CLI_INFO_CUMULATIVE)
- cumulative = conf->cumulative;
+ LOCK(&conf->lock);
+ {
+ if (op == GF_IOS_INFO_ALL || op == GF_IOS_INFO_CUMULATIVE)
+ cumulative = conf->cumulative;
- if (op == GF_CLI_INFO_ALL ||
- op == GF_CLI_INFO_INCREMENTAL) {
- incremental = conf->incremental;
- increment = conf->increment;
+ if (op == GF_IOS_INFO_ALL || op == GF_IOS_INFO_INCREMENTAL) {
+ incremental = conf->incremental;
+ increment = conf->increment;
- if (!is_peek) {
- increment = conf->increment++;
+ if (!is_peek) {
+ increment = conf->increment++;
- ios_global_stats_clear (&conf->incremental,
- &now);
- }
- }
+ ios_global_stats_clear(&conf->incremental, now);
+ }
}
- UNLOCK (&conf->lock);
+ }
+ UNLOCK(&conf->lock);
- if (op == GF_CLI_INFO_ALL ||
- op == GF_CLI_INFO_CUMULATIVE)
- io_stats_dump_global (this, &cumulative, &now, -1, args);
+ if (op == GF_IOS_INFO_ALL || op == GF_IOS_INFO_CUMULATIVE)
+ io_stats_dump_global(this, &cumulative, now, -1, args);
- if (op == GF_CLI_INFO_ALL ||
- op == GF_CLI_INFO_INCREMENTAL)
- io_stats_dump_global (this, &incremental, &now, increment, args);
+ if (op == GF_IOS_INFO_ALL || op == GF_IOS_INFO_INCREMENTAL)
+ io_stats_dump_global(this, &incremental, now, increment, args);
- return 0;
+ return 0;
}
-
int
-io_stats_dump_fd (xlator_t *this, struct ios_fd *iosfd)
+io_stats_dump_fd(xlator_t *this, struct ios_fd *iosfd)
{
- struct ios_conf *conf = NULL;
- struct timeval now;
- uint64_t sec = 0;
- uint64_t usec = 0;
- int i = 0;
- uint64_t data_read = 0;
- uint64_t data_written = 0;
- uint64_t block_count_read = 0;
- uint64_t block_count_write = 0;
+ struct ios_conf *conf = NULL;
+ struct timeval now;
+ int i = 0;
+ double usecs = 0;
+ uint64_t data_read = 0;
+ uint64_t data_written = 0;
+ uint64_t block_count_read = 0;
+ uint64_t block_count_write = 0;
- conf = this->private;
+ conf = this->private;
- if (!conf->dump_fd_stats)
- return 0;
+ if (!conf->dump_fd_stats)
+ return 0;
- if (!iosfd)
- return 0;
+ if (!iosfd)
+ return 0;
- gettimeofday (&now, NULL);
+ gettimeofday(&now, NULL);
+ usecs = gf_tvdiff(&iosfd->opened_at, &now);
- if (iosfd->opened_at.tv_usec > now.tv_usec) {
- now.tv_usec += 1000000;
- now.tv_usec--;
- }
+ gf_log(this->name, GF_LOG_INFO, "--- fd stats ---");
- sec = now.tv_sec - iosfd->opened_at.tv_sec;
- usec = now.tv_usec - iosfd->opened_at.tv_usec;
-
- gf_log (this->name, GF_LOG_INFO,
- "--- fd stats ---");
-
- if (iosfd->filename)
- gf_log (this->name, GF_LOG_INFO,
- " Filename : %s",
- iosfd->filename);
-
- if (sec)
- gf_log (this->name, GF_LOG_INFO,
- " Lifetime : %"PRId64"secs, %"PRId64"usecs",
- sec, usec);
-
- data_read = GF_ATOMIC_GET (iosfd->data_read);
- if (data_read)
- gf_log (this->name, GF_LOG_INFO,
- " BytesRead : %"PRId64" bytes", data_read);
-
- data_written = GF_ATOMIC_GET (iosfd->data_written);
- if (data_written)
- gf_log (this->name, GF_LOG_INFO,
- " BytesWritten : %"PRId64" bytes",
- data_written);
-
- for (i = 0; i < 32; i++) {
- block_count_read = GF_ATOMIC_GET (iosfd->block_count_read[i]);
- if (block_count_read)
- gf_log (this->name, GF_LOG_INFO, " Read %06db+ :"
- "%"PRId64, (1 << i), block_count_read);
- }
- for (i = 0; i < IOS_BLOCK_COUNT_SIZE; i++) {
- block_count_write = GF_ATOMIC_GET (iosfd->block_count_write[i]);
- if (block_count_write)
- gf_log (this->name, GF_LOG_INFO,
- "Write %06db+ : %"PRId64,
- (1 << i), block_count_write);
- }
- return 0;
-}
+ if (iosfd->filename)
+ gf_log(this->name, GF_LOG_INFO, " Filename : %s", iosfd->filename);
-void collect_ios_latency_sample (struct ios_conf *conf,
- glusterfs_fop_t fop_type, double elapsed,
- call_frame_t *frame)
-{
- ios_sample_buf_t *ios_sample_buf = NULL;
- ios_sample_t *ios_sample = NULL;
- struct timespec *timestamp = NULL;
- call_stack_t *root = NULL;
+ if (usecs)
+ gf_log(this->name, GF_LOG_INFO, " Lifetime : %lf secs", usecs);
+ data_read = GF_ATOMIC_GET(iosfd->data_read);
+ if (data_read)
+ gf_log(this->name, GF_LOG_INFO, " BytesRead : %" PRId64 " bytes",
+ data_read);
- ios_sample_buf = conf->ios_sample_buf;
- LOCK (&conf->ios_sampling_lock);
- if (conf->ios_sample_interval == 0 ||
- ios_sample_buf->observed % conf->ios_sample_interval != 0)
- goto out;
+ data_written = GF_ATOMIC_GET(iosfd->data_written);
+ if (data_written)
+ gf_log(this->name, GF_LOG_INFO, " BytesWritten : %" PRId64 " bytes",
+ data_written);
- timestamp = &frame->begin;
- root = frame->root;
-
- ios_sample = &(ios_sample_buf->ios_samples[ios_sample_buf->pos]);
- ios_sample->elapsed = elapsed;
- ios_sample->fop_type = fop_type;
- ios_sample->uid = root->uid;
- ios_sample->gid = root->gid;
- (ios_sample->timestamp).tv_sec = timestamp->tv_sec;
- (ios_sample->timestamp).tv_usec = timestamp->tv_nsec / 1000;
- memcpy (&ios_sample->identifier, &root->identifier,
- sizeof (root->identifier));
-
- /* We've reached the end of the circular buffer, start from the
- * beginning. */
- if (ios_sample_buf->pos == (ios_sample_buf->size - 1))
- ios_sample_buf->pos = 0;
- else
- ios_sample_buf->pos++;
- ios_sample_buf->collected++;
+ for (i = 0; i < 32; i++) {
+ block_count_read = GF_ATOMIC_GET(iosfd->block_count_read[i]);
+ if (block_count_read)
+ gf_log(this->name, GF_LOG_INFO,
+ " Read %06db+ :"
+ "%" PRId64,
+ (1 << i), block_count_read);
+ }
+ for (i = 0; i < IOS_BLOCK_COUNT_SIZE; i++) {
+ block_count_write = GF_ATOMIC_GET(iosfd->block_count_write[i]);
+ if (block_count_write)
+ gf_log(this->name, GF_LOG_INFO, "Write %06db+ : %" PRId64, (1 << i),
+ block_count_write);
+ }
+ return 0;
+}
+
+void
+collect_ios_latency_sample(struct ios_conf *conf, glusterfs_fop_t fop_type,
+ double elapsed, call_frame_t *frame)
+{
+ ios_sample_buf_t *ios_sample_buf = NULL;
+ ios_sample_t *ios_sample = NULL;
+ struct timespec *timestamp = NULL;
+ call_stack_t *root = NULL;
+
+ ios_sample_buf = conf->ios_sample_buf;
+ LOCK(&conf->ios_sampling_lock);
+ if (conf->ios_sample_interval == 0 ||
+ ios_sample_buf->observed % conf->ios_sample_interval != 0)
+ goto out;
+
+ timestamp = &frame->begin;
+ root = frame->root;
+
+ ios_sample = &(ios_sample_buf->ios_samples[ios_sample_buf->pos]);
+ ios_sample->elapsed = elapsed;
+ ios_sample->fop_type = fop_type;
+ ios_sample->uid = root->uid;
+ ios_sample->gid = root->gid;
+ (ios_sample->timestamp).tv_sec = timestamp->tv_sec;
+ (ios_sample->timestamp).tv_usec = timestamp->tv_nsec / 1000;
+ memcpy(&ios_sample->identifier, &root->identifier,
+ sizeof(root->identifier));
+
+ /* We've reached the end of the circular buffer, start from the
+ * beginning. */
+ if (ios_sample_buf->pos == (ios_sample_buf->size - 1))
+ ios_sample_buf->pos = 0;
+ else
+ ios_sample_buf->pos++;
+ ios_sample_buf->collected++;
out:
- ios_sample_buf->observed++;
- UNLOCK (&conf->ios_sampling_lock);
- return;
+ ios_sample_buf->observed++;
+ UNLOCK(&conf->ios_sampling_lock);
+ return;
}
static void
-update_ios_latency_stats (struct ios_global_stats *stats, double elapsed,
- glusterfs_fop_t op)
+update_ios_latency_stats(struct ios_global_stats *stats, double elapsed,
+ glusterfs_fop_t op)
{
- double avg;
+ double avg;
- GF_ASSERT (stats);
+ GF_ASSERT(stats);
- stats->latency[op].total += elapsed;
+ stats->latency[op].total += elapsed;
- if (!stats->latency[op].min)
- stats->latency[op].min = elapsed;
- if (stats->latency[op].min > elapsed)
- stats->latency[op].min = elapsed;
- if (stats->latency[op].max < elapsed)
- stats->latency[op].max = elapsed;
+ if (!stats->latency[op].min)
+ stats->latency[op].min = elapsed;
+ if (stats->latency[op].min > elapsed)
+ stats->latency[op].min = elapsed;
+ if (stats->latency[op].max < elapsed)
+ stats->latency[op].max = elapsed;
- avg = stats->latency[op].avg;
+ avg = stats->latency[op].avg;
- stats->latency[op].avg = avg + (elapsed - avg) /
- GF_ATOMIC_GET (stats->fop_hits[op]);
+ stats->latency[op].avg = avg + (elapsed - avg) /
+ GF_ATOMIC_GET(stats->fop_hits[op]);
}
int
-update_ios_latency (struct ios_conf *conf, call_frame_t *frame,
- glusterfs_fop_t op)
+update_ios_latency(struct ios_conf *conf, call_frame_t *frame,
+ glusterfs_fop_t op)
{
- double elapsed;
- struct timespec *begin, *end;
+ double elapsed;
+ struct timespec *begin, *end;
- begin = &frame->begin;
- end = &frame->end;
+ begin = &frame->begin;
+ end = &frame->end;
- elapsed = ((end->tv_sec - begin->tv_sec) * 1e9
- + (end->tv_nsec - begin->tv_nsec)) / 1000;
+ elapsed = gf_tsdiff(begin, end) / 1000.0;
- update_ios_latency_stats (&conf->cumulative, elapsed, op);
- update_ios_latency_stats (&conf->incremental, elapsed, op);
- collect_ios_latency_sample (conf, op, elapsed, frame);
+ update_ios_latency_stats(&conf->cumulative, elapsed, op);
+ update_ios_latency_stats(&conf->incremental, elapsed, op);
+ collect_ios_latency_sample(conf, op, elapsed, frame);
- return 0;
+ return 0;
}
int32_t
-io_stats_dump_stats_to_dict (xlator_t *this, dict_t *resp,
- ios_stats_type_t flags, int32_t list_cnt)
-{
- struct ios_conf *conf = NULL;
- int cnt = 0;
- char key[256];
- struct ios_stat_head *list_head = NULL;
- struct ios_stat_list *entry = NULL;
- int ret = -1;
- ios_stats_thru_t index = IOS_STATS_THRU_MAX;
- char timestr[256] = {0, };
- char *dict_timestr = NULL;
-
- conf = this->private;
-
- switch (flags) {
- case IOS_STATS_TYPE_OPEN:
- list_head = &conf->list[IOS_STATS_TYPE_OPEN];
- LOCK (&conf->lock);
- {
- ret = dict_set_uint64 (resp, "current-open",
- conf->cumulative.nr_opens);
- if (ret)
- goto unlock;
- ret = dict_set_uint64 (resp, "max-open",
- conf->cumulative.max_nr_opens);
-
- gf_time_fmt (timestr, sizeof timestr,
- conf->cumulative.max_openfd_time.tv_sec,
- gf_timefmt_FT);
- if (conf->cumulative.max_openfd_time.tv_sec)
- snprintf (timestr + strlen (timestr), sizeof timestr - strlen (timestr),
- ".%"GF_PRI_SUSECONDS,
- conf->cumulative.max_openfd_time.tv_usec);
-
- dict_timestr = gf_strdup (timestr);
- if (!dict_timestr)
- goto unlock;
- ret = dict_set_dynstr (resp, "max-openfd-time",
- dict_timestr);
- if (ret)
- goto unlock;
- }
+io_stats_dump_stats_to_dict(xlator_t *this, dict_t *resp,
+ ios_stats_type_t flags, int32_t list_cnt)
+{
+ struct ios_conf *conf = NULL;
+ int cnt = 0;
+ char key[32];
+ int keylen;
+ struct ios_stat_head *list_head = NULL;
+ struct ios_stat_list *entry = NULL;
+ int ret = -1;
+ ios_stats_thru_t index = IOS_STATS_THRU_MAX;
+ char timestr[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ char *dict_timestr = NULL;
+
+ conf = this->private;
+
+ switch (flags) {
+ case IOS_STATS_TYPE_OPEN:
+ list_head = &conf->list[IOS_STATS_TYPE_OPEN];
+ LOCK(&conf->lock);
+ {
+ ret = dict_set_uint64(resp, "current-open",
+ conf->cumulative.nr_opens);
+ if (ret)
+ goto unlock;
+ ret = dict_set_uint64(resp, "max-open",
+ conf->cumulative.max_nr_opens);
+
+ gf_time_fmt_tv(timestr, sizeof timestr,
+ &conf->cumulative.max_openfd_time,
+ gf_timefmt_FT);
+
+ dict_timestr = gf_strdup(timestr);
+ if (!dict_timestr)
+ goto unlock;
+ ret = dict_set_dynstr(resp, "max-openfd-time", dict_timestr);
+ if (ret)
+ goto unlock;
+ }
unlock:
- UNLOCK (&conf->lock);
- /* Do not proceed if we came here because of some error
- * during the dict operation */
- if (ret)
- goto out;
- break;
- case IOS_STATS_TYPE_READ:
- list_head = &conf->list[IOS_STATS_TYPE_READ];
- break;
- case IOS_STATS_TYPE_WRITE:
- list_head = &conf->list[IOS_STATS_TYPE_WRITE];
- break;
- case IOS_STATS_TYPE_OPENDIR:
- list_head = &conf->list[IOS_STATS_TYPE_OPENDIR];
- break;
- case IOS_STATS_TYPE_READDIRP:
- list_head = &conf->list[IOS_STATS_TYPE_READDIRP];
- break;
- case IOS_STATS_TYPE_READ_THROUGHPUT:
- list_head = &conf->thru_list[IOS_STATS_THRU_READ];
- index = IOS_STATS_THRU_READ;
- break;
- case IOS_STATS_TYPE_WRITE_THROUGHPUT:
- list_head = &conf->thru_list[IOS_STATS_THRU_WRITE];
- index = IOS_STATS_THRU_WRITE;
- break;
-
- default:
- goto out;
- }
- ret = dict_set_int32 (resp, "top-op", flags);
- if (!list_cnt)
+ UNLOCK(&conf->lock);
+ /* Do not proceed if we came here because of some error
+ * during the dict operation */
+ if (ret)
goto out;
- LOCK (&list_head->lock);
- {
- list_for_each_entry (entry, &list_head->iosstats->list, list) {
+ break;
+ case IOS_STATS_TYPE_READ:
+ list_head = &conf->list[IOS_STATS_TYPE_READ];
+ break;
+ case IOS_STATS_TYPE_WRITE:
+ list_head = &conf->list[IOS_STATS_TYPE_WRITE];
+ break;
+ case IOS_STATS_TYPE_OPENDIR:
+ list_head = &conf->list[IOS_STATS_TYPE_OPENDIR];
+ break;
+ case IOS_STATS_TYPE_READDIRP:
+ list_head = &conf->list[IOS_STATS_TYPE_READDIRP];
+ break;
+ case IOS_STATS_TYPE_READ_THROUGHPUT:
+ list_head = &conf->thru_list[IOS_STATS_THRU_READ];
+ index = IOS_STATS_THRU_READ;
+ break;
+ case IOS_STATS_TYPE_WRITE_THROUGHPUT:
+ list_head = &conf->thru_list[IOS_STATS_THRU_WRITE];
+ index = IOS_STATS_THRU_WRITE;
+ break;
- cnt++;
- snprintf (key, 256, "%s-%d", "filename", cnt);
- ret = dict_set_str (resp, key, entry->iosstat->filename);
- if (ret)
- goto unlock_list_head;
- snprintf (key, 256, "%s-%d", "value",cnt);
- ret = dict_set_uint64 (resp, key, entry->value);
- if (ret)
- goto unlock_list_head;
- if (index != IOS_STATS_THRU_MAX) {
- snprintf (key, 256, "%s-%d", "time-sec", cnt);
- ret = dict_set_int32 (resp, key,
- entry->iosstat->thru_counters[index].time.tv_sec);
- if (ret)
- goto unlock_list_head;
- snprintf (key, 256, "%s-%d", "time-usec", cnt);
- ret = dict_set_int32 (resp, key,
- entry->iosstat->thru_counters[index].time.tv_usec);
- if (ret)
- goto unlock_list_head;
- }
- if (cnt == list_cnt)
- break;
-
- }
+ default:
+ goto out;
+ }
+ ret = dict_set_int32_sizen(resp, "top-op", flags);
+ if (!list_cnt)
+ goto out;
+ LOCK(&list_head->lock);
+ {
+ list_for_each_entry(entry, &list_head->iosstats->list, list)
+ {
+ cnt++;
+ keylen = snprintf(key, sizeof(key), "filename-%d", cnt);
+ ret = dict_set_strn(resp, key, keylen, entry->iosstat->filename);
+ if (ret)
+ goto unlock_list_head;
+ snprintf(key, sizeof(key), "value-%d", cnt);
+ ret = dict_set_uint64(resp, key, entry->value);
+ if (ret)
+ goto unlock_list_head;
+ if (index != IOS_STATS_THRU_MAX) {
+ keylen = snprintf(key, sizeof(key), "time-sec-%d", cnt);
+ ret = dict_set_int32n(
+ resp, key, keylen,
+ entry->iosstat->thru_counters[index].time.tv_sec);
+ if (ret)
+ goto unlock_list_head;
+ keylen = snprintf(key, sizeof(key), "time-usec-%d", cnt);
+ ret = dict_set_int32n(
+ resp, key, keylen,
+ entry->iosstat->thru_counters[index].time.tv_usec);
+ if (ret)
+ goto unlock_list_head;
+ }
+ if (cnt == list_cnt)
+ break;
}
+ }
unlock_list_head:
- UNLOCK (&list_head->lock);
- /* ret is !=0 if some dict operation in the above critical region
- * failed. */
- if (ret)
- goto out;
- ret = dict_set_int32 (resp, "members", cnt);
- out:
- return ret;
+ UNLOCK(&list_head->lock);
+ /* ret is !=0 if some dict operation in the above critical region
+ * failed. */
+ if (ret)
+ goto out;
+ ret = dict_set_int32_sizen(resp, "members", cnt);
+out:
+ return ret;
}
-static struct ios_stat*
-ios_init_iosstat (xlator_t *this, char *path, uuid_t gfid, inode_t *inode)
+static struct ios_stat *
+ios_init_iosstat(xlator_t *this, char *path, uuid_t gfid, inode_t *inode)
{
- struct ios_stat *iosstat = NULL;
- int i = 0;
+ struct ios_stat *iosstat = NULL;
+ int i = 0;
- iosstat = GF_CALLOC (1, sizeof (*iosstat), gf_io_stats_mt_ios_stat);
- if (!iosstat)
- goto out;
+ iosstat = GF_CALLOC(1, sizeof(*iosstat), gf_io_stats_mt_ios_stat);
+ if (!iosstat)
+ goto out;
- iosstat->filename = gf_strdup (path);
- gf_uuid_copy (iosstat->gfid, gfid);
- LOCK_INIT (&iosstat->lock);
+ iosstat->filename = gf_strdup(path);
+ gf_uuid_copy(iosstat->gfid, gfid);
+ LOCK_INIT(&iosstat->lock);
- for (i = 0; i < IOS_STATS_TYPE_MAX; i++)
- GF_ATOMIC_INIT (iosstat->counters[i], 0);
+ for (i = 0; i < IOS_STATS_TYPE_MAX; i++)
+ GF_ATOMIC_INIT(iosstat->counters[i], 0);
- ios_inode_ctx_set (inode, this, iosstat);
+ ios_inode_ctx_set(inode, this, iosstat);
out:
- return iosstat;
+ return iosstat;
}
int
-io_stats_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd,
- inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
-{
- struct ios_fd *iosfd = NULL;
- char *path = NULL;
- struct ios_stat *iosstat = NULL;
- struct ios_conf *conf = NULL;
-
- conf = this->private;
-
- path = frame->local;
- frame->local = NULL;
-
- if (!path)
- goto unwind;
-
- if (op_ret < 0) {
- GF_FREE (path);
- goto unwind;
- }
-
- iosfd = GF_CALLOC (1, sizeof (*iosfd), gf_io_stats_mt_ios_fd);
- if (!iosfd) {
- GF_FREE (path);
- goto unwind;
- }
-
- iosfd->filename = path;
- gettimeofday (&iosfd->opened_at, NULL);
-
- ios_fd_ctx_set (fd, this, iosfd);
- LOCK (&conf->lock);
- {
- conf->cumulative.nr_opens++;
- if (conf->cumulative.nr_opens > conf->cumulative.max_nr_opens) {
- conf->cumulative.max_nr_opens = conf->cumulative.nr_opens;
- conf->cumulative.max_openfd_time = iosfd->opened_at;
- }
+io_stats_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ struct ios_fd *iosfd = NULL;
+ char *path = NULL;
+ struct ios_stat *iosstat = NULL;
+ struct ios_conf *conf = NULL;
+
+ conf = this->private;
+
+ path = frame->local;
+ frame->local = NULL;
+
+ if (!path)
+ goto unwind;
+
+ if (op_ret < 0) {
+ GF_FREE(path);
+ goto unwind;
+ }
+
+ iosfd = GF_CALLOC(1, sizeof(*iosfd), gf_io_stats_mt_ios_fd);
+ if (!iosfd) {
+ GF_FREE(path);
+ goto unwind;
+ }
+
+ iosfd->filename = path;
+ gettimeofday(&iosfd->opened_at, NULL);
+
+ ios_fd_ctx_set(fd, this, iosfd);
+ LOCK(&conf->lock);
+ {
+ conf->cumulative.nr_opens++;
+ if (conf->cumulative.nr_opens > conf->cumulative.max_nr_opens) {
+ conf->cumulative.max_nr_opens = conf->cumulative.nr_opens;
+ conf->cumulative.max_openfd_time = iosfd->opened_at;
}
- UNLOCK (&conf->lock);
+ }
+ UNLOCK(&conf->lock);
- iosstat = ios_init_iosstat (this, path, buf->ia_gfid, inode);
- if (!iosstat)
- GF_FREE (path);
+ iosstat = ios_init_iosstat(this, path, buf->ia_gfid, inode);
+ if (!iosstat)
+ GF_FREE(path);
unwind:
- UPDATE_PROFILE_STATS (frame, CREATE);
- STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
- preparent, postparent, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, CREATE);
+ STACK_UNWIND_STRICT(create, frame, op_ret, op_errno, fd, inode, buf,
+ preparent, postparent, xdata);
+ return 0;
}
-
int
-io_stats_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
-{
- struct ios_fd *iosfd = NULL;
- char *path = NULL;
- struct ios_stat *iosstat = NULL;
- struct ios_conf *conf = NULL;
- int i = 0;
-
- conf = this->private;
- path = frame->local;
- frame->local = NULL;
-
- if (!path)
- goto unwind;
-
- if (op_ret < 0) {
- GF_FREE (path);
- goto unwind;
- }
-
- iosfd = GF_CALLOC (1, sizeof (*iosfd), gf_io_stats_mt_ios_fd);
- if (!iosfd) {
- GF_FREE (path);
- goto unwind;
- }
-
- iosfd->filename = path;
- GF_ATOMIC_INIT (iosfd->data_read, 0);
- GF_ATOMIC_INIT (iosfd->data_written, 0);
- for (i = 0; i < IOS_BLOCK_COUNT_SIZE; i++) {
- GF_ATOMIC_INIT (iosfd->block_count_write[i], 0);
- GF_ATOMIC_INIT (iosfd->block_count_read[i], 0);
- }
- gettimeofday (&iosfd->opened_at, NULL);
-
- ios_fd_ctx_set (fd, this, iosfd);
-
- ios_inode_ctx_get (fd->inode, this, &iosstat);
- if (!iosstat) {
- iosstat = ios_init_iosstat (this, path, fd->inode->gfid,
- fd->inode);
- }
-
- LOCK (&conf->lock);
- {
- conf->cumulative.nr_opens++;
- if (conf->cumulative.nr_opens > conf->cumulative.max_nr_opens) {
- conf->cumulative.max_nr_opens = conf->cumulative.nr_opens;
- conf->cumulative.max_openfd_time = iosfd->opened_at;
- }
- }
- UNLOCK (&conf->lock);
- if (iosstat) {
- ios_bump_stats (this, iosstat, IOS_STATS_TYPE_OPEN);
- iosstat = NULL;
+io_stats_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+ struct ios_fd *iosfd = NULL;
+ char *path = NULL;
+ struct ios_stat *iosstat = NULL;
+ struct ios_conf *conf = NULL;
+ int i = 0;
+
+ conf = this->private;
+ path = frame->local;
+ frame->local = NULL;
+
+ if (!path)
+ goto unwind;
+
+ if (op_ret < 0) {
+ GF_FREE(path);
+ goto unwind;
+ }
+
+ iosfd = GF_CALLOC(1, sizeof(*iosfd), gf_io_stats_mt_ios_fd);
+ if (!iosfd) {
+ GF_FREE(path);
+ goto unwind;
+ }
+
+ iosfd->filename = path;
+ GF_ATOMIC_INIT(iosfd->data_read, 0);
+ GF_ATOMIC_INIT(iosfd->data_written, 0);
+ for (i = 0; i < IOS_BLOCK_COUNT_SIZE; i++) {
+ GF_ATOMIC_INIT(iosfd->block_count_write[i], 0);
+ GF_ATOMIC_INIT(iosfd->block_count_read[i], 0);
+ }
+ gettimeofday(&iosfd->opened_at, NULL);
+
+ ios_fd_ctx_set(fd, this, iosfd);
+
+ ios_inode_ctx_get(fd->inode, this, &iosstat);
+ if (!iosstat) {
+ iosstat = ios_init_iosstat(this, path, fd->inode->gfid, fd->inode);
+ }
+
+ LOCK(&conf->lock);
+ {
+ conf->cumulative.nr_opens++;
+ if (conf->cumulative.nr_opens > conf->cumulative.max_nr_opens) {
+ conf->cumulative.max_nr_opens = conf->cumulative.nr_opens;
+ conf->cumulative.max_openfd_time = iosfd->opened_at;
}
+ }
+ UNLOCK(&conf->lock);
+ if (iosstat) {
+ ios_bump_stats(this, iosstat, IOS_STATS_TYPE_OPEN);
+ iosstat = NULL;
+ }
unwind:
- UPDATE_PROFILE_STATS (frame, OPEN);
-
- STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, OPEN);
+ STACK_UNWIND_STRICT(open, frame, op_ret, op_errno, fd, xdata);
+ return 0;
}
-
int
-io_stats_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
+io_stats_stat_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, STAT);
- STACK_UNWIND_STRICT (stat, frame, op_ret, op_errno, buf, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, STAT);
+ STACK_UNWIND_STRICT(stat, frame, op_ret, op_errno, buf, xdata);
+ return 0;
}
-
int
-io_stats_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iovec *vector, int32_t count,
- struct iatt *buf, struct iobref *iobref, dict_t *xdata)
+io_stats_readv_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iovec *vector,
+ int32_t count, struct iatt *buf, struct iobref *iobref,
+ dict_t *xdata)
{
- int len = 0;
- fd_t *fd = NULL;
- struct ios_stat *iosstat = NULL;
-
- fd = frame->local;
- frame->local = NULL;
+ int len = 0;
+ fd_t *fd = NULL;
+ struct ios_stat *iosstat = NULL;
- if (op_ret > 0) {
- len = iov_length (vector, count);
- ios_bump_read (this, fd, len);
- }
+ fd = frame->local;
+ frame->local = NULL;
- UPDATE_PROFILE_STATS (frame, READ);
- ios_inode_ctx_get (fd->inode, this, &iosstat);
+ if (op_ret > 0) {
+ len = iov_length(vector, count);
+ ios_bump_read(this, fd, len);
+ }
- if (iosstat) {
- ios_bump_stats (this, iosstat, IOS_STATS_TYPE_READ);
- BUMP_THROUGHPUT (iosstat, IOS_STATS_THRU_READ);
- iosstat = NULL;
- }
+ UPDATE_PROFILE_STATS(frame, READ);
+ ios_inode_ctx_get(fd->inode, this, &iosstat);
- STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno,
- vector, count, buf, iobref, xdata);
- return 0;
+ if (iosstat) {
+ ios_bump_stats(this, iosstat, IOS_STATS_TYPE_READ);
+ BUMP_THROUGHPUT(iosstat, IOS_STATS_THRU_READ);
+ iosstat = NULL;
+ }
+ STACK_UNWIND_STRICT(readv, frame, op_ret, op_errno, vector, count, buf,
+ iobref, xdata);
+ return 0;
}
-
int
-io_stats_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
-{
- struct ios_stat *iosstat = NULL;
- inode_t *inode = NULL;
-
- UPDATE_PROFILE_STATS (frame, WRITE);
- if (frame->local){
- inode = frame->local;
- frame->local = NULL;
- ios_inode_ctx_get (inode, this, &iosstat);
- if (iosstat) {
- ios_bump_stats (this, iosstat, IOS_STATS_TYPE_WRITE);
- BUMP_THROUGHPUT (iosstat, IOS_STATS_THRU_WRITE);
- inode = NULL;
- iosstat = NULL;
- }
- }
+io_stats_writev_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ struct ios_stat *iosstat = NULL;
+ inode_t *inode = NULL;
- STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, WRITE);
+ if (frame->local) {
+ inode = frame->local;
+ frame->local = NULL;
+ ios_inode_ctx_get(inode, this, &iosstat);
+ if (iosstat) {
+ ios_bump_stats(this, iosstat, IOS_STATS_TYPE_WRITE);
+ BUMP_THROUGHPUT(iosstat, IOS_STATS_THRU_WRITE);
+ inode = NULL;
+ iosstat = NULL;
+ }
+ }
+ STACK_UNWIND_STRICT(writev, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
+int
+io_stats_copy_file_range_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *stbuf, struct iatt *prebuf_dst,
+ struct iatt *postbuf_dst, dict_t *xdata)
+{
+ UPDATE_PROFILE_STATS(frame, COPY_FILE_RANGE);
-
+ STACK_UNWIND_STRICT(copy_file_range, frame, op_ret, op_errno, stbuf,
+ prebuf_dst, postbuf_dst, xdata);
+ return 0;
+}
int
-io_stats_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *buf, dict_t *xdata)
+io_stats_readdirp_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *buf,
+ dict_t *xdata)
{
- struct ios_stat *iosstat = NULL;
- inode_t *inode = frame->local;
+ struct ios_stat *iosstat = NULL;
+ inode_t *inode = frame->local;
- frame->local = NULL;
+ frame->local = NULL;
- UPDATE_PROFILE_STATS (frame, READDIRP);
+ UPDATE_PROFILE_STATS(frame, READDIRP);
- ios_inode_ctx_get (inode, this, &iosstat);
+ ios_inode_ctx_get(inode, this, &iosstat);
- if (iosstat) {
- ios_bump_stats (this, iosstat, IOS_STATS_TYPE_READDIRP);
- iosstat = NULL;
- }
+ if (iosstat) {
+ ios_bump_stats(this, iosstat, IOS_STATS_TYPE_READDIRP);
+ iosstat = NULL;
+ }
- STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, buf, xdata);
- return 0;
+ STACK_UNWIND_STRICT(readdirp, frame, op_ret, op_errno, buf, xdata);
+ return 0;
}
-
int
-io_stats_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *buf, dict_t *xdata)
+io_stats_readdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *buf,
+ dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, READDIR);
- STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, buf, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, READDIR);
+ STACK_UNWIND_STRICT(readdir, frame, op_ret, op_errno, buf, xdata);
+ return 0;
}
-
int
-io_stats_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+io_stats_fsync_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, FSYNC);
- STACK_UNWIND_STRICT (fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, FSYNC);
+ STACK_UNWIND_STRICT(fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+ return 0;
}
-
int
-io_stats_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preop, struct iatt *postop, dict_t *xdata)
+io_stats_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preop,
+ struct iatt *postop, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, SETATTR);
- STACK_UNWIND_STRICT (setattr, frame, op_ret, op_errno, preop, postop, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, SETATTR);
+ STACK_UNWIND_STRICT(setattr, frame, op_ret, op_errno, preop, postop, xdata);
+ return 0;
}
-
int
-io_stats_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
+io_stats_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, UNLINK);
- STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno,
- preparent, postparent, xdata);
- return 0;
-
+ UPDATE_PROFILE_STATS(frame, UNLINK);
+ STACK_UNWIND_STRICT(unlink, frame, op_ret, op_errno, preparent, postparent,
+ xdata);
+ return 0;
}
-
int
-io_stats_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent, dict_t *xdata)
-{
- UPDATE_PROFILE_STATS (frame, RENAME);
- STACK_UNWIND_STRICT (rename, frame, op_ret, op_errno, buf,
- preoldparent, postoldparent,
- prenewparent, postnewparent, xdata);
- return 0;
+io_stats_rename_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ struct iatt *preoldparent, struct iatt *postoldparent,
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
+{
+ UPDATE_PROFILE_STATS(frame, RENAME);
+ STACK_UNWIND_STRICT(rename, frame, op_ret, op_errno, buf, preoldparent,
+ postoldparent, prenewparent, postnewparent, xdata);
+ return 0;
}
-
int
-io_stats_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, const char *buf,
- struct iatt *sbuf, dict_t *xdata)
+io_stats_readlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, const char *buf,
+ struct iatt *sbuf, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, READLINK);
- STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, buf, sbuf, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, READLINK);
+ STACK_UNWIND_STRICT(readlink, frame, op_ret, op_errno, buf, sbuf, xdata);
+ return 0;
}
-
int
-io_stats_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf,
- dict_t *xdata, struct iatt *postparent)
-{
- UPDATE_PROFILE_STATS (frame, LOOKUP);
- STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf, xdata,
- postparent);
- return 0;
+io_stats_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+{
+ UPDATE_PROFILE_STATS(frame, LOOKUP);
+ STACK_UNWIND_STRICT(lookup, frame, op_ret, op_errno, inode, buf, xdata,
+ postparent);
+ return 0;
}
-
int
-io_stats_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
-{
- UPDATE_PROFILE_STATS (frame, SYMLINK);
- STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
- return 0;
+io_stats_symlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ UPDATE_PROFILE_STATS(frame, SYMLINK);
+ STACK_UNWIND_STRICT(symlink, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
+ return 0;
}
-
int
-io_stats_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
-{
- UPDATE_PROFILE_STATS (frame, MKNOD);
- STACK_UNWIND_STRICT (mknod, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
- return 0;
+io_stats_mknod_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ UPDATE_PROFILE_STATS(frame, MKNOD);
+ STACK_UNWIND_STRICT(mknod, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
+ return 0;
}
-
int
-io_stats_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+io_stats_mkdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- char *path = frame->local;
+ char *path = frame->local;
- if (!path)
- goto unwind;
+ if (!path)
+ goto unwind;
- UPDATE_PROFILE_STATS (frame, MKDIR);
- if (op_ret < 0)
- goto unwind;
+ UPDATE_PROFILE_STATS(frame, MKDIR);
+ if (op_ret < 0)
+ goto unwind;
- /* allocate a struct ios_stat and set the inode ctx */
- ios_init_iosstat (this, path, buf->ia_gfid, inode);
+ /* allocate a struct ios_stat and set the inode ctx */
+ ios_init_iosstat(this, path, buf->ia_gfid, inode);
unwind:
- /* local is assigned with path */
- GF_FREE (frame->local);
- frame->local = NULL;
- STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
- return 0;
+ /* local is assigned with path */
+ GF_FREE(frame->local);
+ frame->local = NULL;
+ STACK_UNWIND_STRICT(mkdir, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
+ return 0;
}
-
int
-io_stats_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
-{
- UPDATE_PROFILE_STATS (frame, LINK);
- STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
- return 0;
+io_stats_link_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ UPDATE_PROFILE_STATS(frame, LINK);
+ STACK_UNWIND_STRICT(link, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
+ return 0;
}
-
int
-io_stats_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+io_stats_flush_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, FLUSH);
- STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, FLUSH);
+ STACK_UNWIND_STRICT(flush, frame, op_ret, op_errno, xdata);
+ return 0;
}
-
int
-io_stats_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+io_stats_opendir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- struct ios_stat *iosstat = NULL;
- int ret = -1;
+ struct ios_stat *iosstat = NULL;
+ int ret = -1;
- UPDATE_PROFILE_STATS (frame, OPENDIR);
- if (op_ret < 0)
- goto unwind;
+ UPDATE_PROFILE_STATS(frame, OPENDIR);
+ if (op_ret < 0)
+ goto unwind;
- ios_fd_ctx_set (fd, this, 0);
+ ios_fd_ctx_set(fd, this, 0);
- ret = ios_inode_ctx_get (fd->inode, this, &iosstat);
- if (!ret)
- ios_bump_stats (this, iosstat, IOS_STATS_TYPE_OPENDIR);
+ ret = ios_inode_ctx_get(fd->inode, this, &iosstat);
+ if (!ret)
+ ios_bump_stats(this, iosstat, IOS_STATS_TYPE_OPENDIR);
unwind:
- STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd, xdata);
- return 0;
+ STACK_UNWIND_STRICT(opendir, frame, op_ret, op_errno, fd, xdata);
+ return 0;
}
-
int
-io_stats_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
+io_stats_rmdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
+ UPDATE_PROFILE_STATS(frame, RMDIR);
- UPDATE_PROFILE_STATS (frame, RMDIR);
-
- STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno,
- preparent, postparent, xdata);
- return 0;
+ STACK_UNWIND_STRICT(rmdir, frame, op_ret, op_errno, preparent, postparent,
+ xdata);
+ return 0;
}
-
int
-io_stats_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+io_stats_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, TRUNCATE);
- STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno,
- prebuf, postbuf, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, TRUNCATE);
+ STACK_UNWIND_STRICT(truncate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
-
int
-io_stats_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct statvfs *buf, dict_t *xdata)
+io_stats_statfs_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct statvfs *buf,
+ dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, STATFS);
- STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, buf, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, STATFS);
+ STACK_UNWIND_STRICT(statfs, frame, op_ret, op_errno, buf, xdata);
+ return 0;
}
-
int
-io_stats_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+io_stats_setxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, SETXATTR);
- STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, SETXATTR);
+ STACK_UNWIND_STRICT(setxattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
-
int
-io_stats_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+io_stats_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, GETXATTR);
- STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, GETXATTR);
+ STACK_UNWIND_STRICT(getxattr, frame, op_ret, op_errno, dict, xdata);
+ return 0;
}
-
int
-io_stats_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+io_stats_removexattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, REMOVEXATTR);
- STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, REMOVEXATTR);
+ STACK_UNWIND_STRICT(removexattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-io_stats_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+io_stats_fsetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, FSETXATTR);
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, FSETXATTR);
+ STACK_UNWIND_STRICT(fsetxattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
-
int
-io_stats_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+io_stats_fgetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, FGETXATTR);
- STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, FGETXATTR);
+ STACK_UNWIND_STRICT(fgetxattr, frame, op_ret, op_errno, dict, xdata);
+ return 0;
}
-
int
-io_stats_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+io_stats_fremovexattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, FREMOVEXATTR);
- STACK_UNWIND_STRICT (fremovexattr, frame, op_ret, op_errno, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, FREMOVEXATTR);
+ STACK_UNWIND_STRICT(fremovexattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
-
int
-io_stats_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+io_stats_fsyncdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, FSYNCDIR);
- STACK_UNWIND_STRICT (fsyncdir, frame, op_ret, op_errno, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, FSYNCDIR);
+ STACK_UNWIND_STRICT(fsyncdir, frame, op_ret, op_errno, xdata);
+ return 0;
}
-
int
-io_stats_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+io_stats_access_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, ACCESS);
- STACK_UNWIND_STRICT (access, frame, op_ret, op_errno, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, ACCESS);
+ STACK_UNWIND_STRICT(access, frame, op_ret, op_errno, xdata);
+ return 0;
}
-
int
-io_stats_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+io_stats_ftruncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, FTRUNCATE);
- STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno,
- prebuf, postbuf, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, FTRUNCATE);
+ STACK_UNWIND_STRICT(ftruncate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
-
int
-io_stats_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
+io_stats_fstat_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, FSTAT);
- STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, buf, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, FSTAT);
+ STACK_UNWIND_STRICT(fstat, frame, op_ret, op_errno, buf, xdata);
+ return 0;
}
-
int
io_stats_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- UPDATE_PROFILE_STATS(frame, FALLOCATE);
- STACK_UNWIND_STRICT(fallocate, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, FALLOCATE);
+ STACK_UNWIND_STRICT(fallocate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
-
int
io_stats_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- UPDATE_PROFILE_STATS(frame, DISCARD);
- STACK_UNWIND_STRICT(discard, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, DISCARD);
+ STACK_UNWIND_STRICT(discard, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
int
io_stats_zerofill_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- UPDATE_PROFILE_STATS(frame, ZEROFILL);
- STACK_UNWIND_STRICT(zerofill, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, ZEROFILL);
+ STACK_UNWIND_STRICT(zerofill, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
int32_t
-io_stats_ipc_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+io_stats_ipc_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- UPDATE_PROFILE_STATS(frame, IPC);
- STACK_UNWIND_STRICT (ipc, frame, op_ret, op_errno,
- xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, IPC);
+ STACK_UNWIND_STRICT(ipc, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-io_stats_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
+io_stats_lk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+ dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, LK);
- STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, lock, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, LK);
+ STACK_UNWIND_STRICT(lk, frame, op_ret, op_errno, lock, xdata);
+ return 0;
}
-
int
-io_stats_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+io_stats_entrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, ENTRYLK);
- STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, ENTRYLK);
+ STACK_UNWIND_STRICT(entrylk, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-io_stats_fentrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+io_stats_fentrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, FENTRYLK);
- STACK_UNWIND_STRICT (fentrylk, frame, op_ret, op_errno, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, FENTRYLK);
+ STACK_UNWIND_STRICT(fentrylk, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-io_stats_rchecksum_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+io_stats_rchecksum_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, uint32_t weak_checksum,
uint8_t *strong_checksum, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, RCHECKSUM);
- STACK_UNWIND_STRICT (rchecksum, frame, op_ret, op_errno, weak_checksum,
- strong_checksum, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, RCHECKSUM);
+ STACK_UNWIND_STRICT(rchecksum, frame, op_ret, op_errno, weak_checksum,
+ strong_checksum, xdata);
+ return 0;
}
int
-io_stats_seek_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, off_t offset,
- dict_t *xdata)
+io_stats_seek_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, off_t offset, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, SEEK);
- STACK_UNWIND_STRICT (seek, frame, op_ret, op_errno, offset, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, SEEK);
+ STACK_UNWIND_STRICT(seek, frame, op_ret, op_errno, offset, xdata);
+ return 0;
}
int
-io_stats_lease_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct gf_lease *lease, dict_t *xdata)
+io_stats_lease_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct gf_lease *lease,
+ dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, LEASE);
- STACK_UNWIND_STRICT (lease, frame, op_ret, op_errno, lease, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, LEASE);
+ STACK_UNWIND_STRICT(lease, frame, op_ret, op_errno, lease, xdata);
+ return 0;
}
int
-io_stats_getactivelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- lock_migration_info_t *locklist,
- dict_t *xdata)
-{
- UPDATE_PROFILE_STATS (frame, GETACTIVELK);
- STACK_UNWIND_STRICT (getactivelk, frame, op_ret, op_errno,
- locklist, xdata);
- return 0;
+io_stats_getactivelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ lock_migration_info_t *locklist, dict_t *xdata)
+{
+ UPDATE_PROFILE_STATS(frame, GETACTIVELK);
+ STACK_UNWIND_STRICT(getactivelk, frame, op_ret, op_errno, locklist, xdata);
+ return 0;
}
int
-io_stats_setactivelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+io_stats_setactivelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, SETACTIVELK);
- STACK_UNWIND_STRICT (setactivelk, frame, op_ret, op_errno, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, SETACTIVELK);
+ STACK_UNWIND_STRICT(setactivelk, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-io_stats_compound_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, void *data,
- dict_t *xdata)
+io_stats_compound_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, void *data,
+ dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, COMPOUND);
- STACK_UNWIND_STRICT (compound, frame, op_ret, op_errno, data, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, COMPOUND);
+ STACK_UNWIND_STRICT(compound, frame, op_ret, op_errno, data, xdata);
+ return 0;
}
int
-io_stats_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+io_stats_xattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, XATTROP);
- STACK_UNWIND_STRICT (xattrop, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, XATTROP);
+ STACK_UNWIND_STRICT(xattrop, frame, op_ret, op_errno, dict, xdata);
+ return 0;
}
-
int
-io_stats_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+io_stats_fxattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, FXATTROP);
- STACK_UNWIND_STRICT (fxattrop, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, FXATTROP);
+ STACK_UNWIND_STRICT(fxattrop, frame, op_ret, op_errno, dict, xdata);
+ return 0;
}
-
int
-io_stats_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+io_stats_inodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- UPDATE_PROFILE_STATS (frame, INODELK);
- STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, INODELK);
+ STACK_UNWIND_STRICT(inodelk, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-io_stats_entrylk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, const char *basename,
- entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
+io_stats_entrylk(call_frame_t *frame, xlator_t *this, const char *volume,
+ loc_t *loc, const char *basename, entrylk_cmd cmd,
+ entrylk_type type, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_entrylk_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->entrylk,
- volume, loc, basename, cmd, type, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, volume, loc, basename, cmd,
+ type, xdata);
+ return 0;
}
int
-io_stats_fentrylk (call_frame_t *frame, xlator_t *this,
- const char *volume, fd_t *fd, const char *basename,
- entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
+io_stats_fentrylk(call_frame_t *frame, xlator_t *this, const char *volume,
+ fd_t *fd, const char *basename, entrylk_cmd cmd,
+ entrylk_type type, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_fentrylk_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fentrylk,
- volume, fd, basename, cmd, type, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_fentrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fentrylk, volume, fd, basename, cmd,
+ type, xdata);
+ return 0;
}
-
int
-io_stats_inodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
+io_stats_inodelk(call_frame_t *frame, xlator_t *this, const char *volume,
+ loc_t *loc, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
{
+ START_FOP_LATENCY(frame);
- START_FOP_LATENCY (frame);
-
- STACK_WIND (frame, io_stats_inodelk_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->inodelk,
- volume, loc, cmd, flock, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_inodelk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->inodelk, volume, loc, cmd, flock,
+ xdata);
+ return 0;
}
-
int
-io_stats_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+io_stats_finodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
-
- UPDATE_PROFILE_STATS (frame, FINODELK);
- STACK_UNWIND_STRICT (finodelk, frame, op_ret, op_errno, xdata);
- return 0;
+ UPDATE_PROFILE_STATS(frame, FINODELK);
+ STACK_UNWIND_STRICT(finodelk, frame, op_ret, op_errno, xdata);
+ return 0;
}
-
int
-io_stats_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
- fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
+io_stats_finodelk(call_frame_t *frame, xlator_t *this, const char *volume,
+ fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_finodelk_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->finodelk,
- volume, fd, cmd, flock, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_finodelk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->finodelk, volume, fd, cmd, flock,
+ xdata);
+ return 0;
}
-
int
-io_stats_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+io_stats_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_xattrop_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->xattrop,
- loc, flags, dict, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_xattrop_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->xattrop, loc, flags, dict, xdata);
+ return 0;
}
-
int
-io_stats_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
- gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+io_stats_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_fxattrop_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fxattrop,
- fd, flags, dict, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_fxattrop_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fxattrop, fd, flags, dict, xdata);
+ return 0;
}
-
int
-io_stats_lookup (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xdata)
+io_stats_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_lookup_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup,
- loc, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, xdata);
+ return 0;
}
-
int
-io_stats_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+io_stats_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_stat_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->stat,
- loc, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_stat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->stat, loc, xdata);
+ return 0;
}
-
int
-io_stats_readlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc, size_t size, dict_t *xdata)
+io_stats_readlink(call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+ dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_readlink_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readlink,
- loc, size, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_readlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readlink, loc, size, xdata);
+ return 0;
}
-
int
-io_stats_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
- mode_t mode, dev_t dev, mode_t umask, dict_t *xdata)
+io_stats_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t dev, mode_t umask, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_mknod_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mknod,
- loc, mode, dev, umask, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_mknod_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, loc, mode, dev, umask, xdata);
+ return 0;
}
-
int
-io_stats_mkdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata)
+io_stats_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- if (loc->path)
- frame->local = gf_strdup (loc->path);
+ if (loc->path)
+ frame->local = gf_strdup(loc->path);
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_mkdir_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mkdir,
- loc, mode, umask, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_mkdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
+ return 0;
}
-
int
-io_stats_unlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int xflag, dict_t *xdata)
+io_stats_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_unlink_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->unlink,
- loc, xflag, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_unlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
+ return 0;
}
-
int
-io_stats_rmdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int flags, dict_t *xdata)
+io_stats_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_rmdir_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rmdir,
- loc, flags, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_rmdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rmdir, loc, flags, xdata);
+ return 0;
}
-
int
-io_stats_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
- loc_t *loc, mode_t umask, dict_t *xdata)
+io_stats_symlink(call_frame_t *frame, xlator_t *this, const char *linkpath,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_symlink_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->symlink,
- linkpath, loc, umask, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_symlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->symlink, linkpath, loc, umask, xdata);
+ return 0;
}
-
int
-io_stats_rename (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+io_stats_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_rename_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rename,
- oldloc, newloc, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_rename_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
+ return 0;
}
-
int
-io_stats_link (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+io_stats_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_link_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->link,
- oldloc, newloc, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_link_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
+ return 0;
}
-
int
-io_stats_setattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, struct iatt *stbuf, int32_t valid, dict_t *xdata)
+io_stats_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_setattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setattr,
- loc, stbuf, valid, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_setattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid, xdata);
+ return 0;
}
-
int
-io_stats_truncate (call_frame_t *frame, xlator_t *this,
- loc_t *loc, off_t offset, dict_t *xdata)
+io_stats_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_truncate_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->truncate,
- loc, offset, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_truncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
+ return 0;
}
-
int
-io_stats_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int32_t flags, fd_t *fd, dict_t *xdata)
+io_stats_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
{
- if (loc->path)
- frame->local = gf_strdup (loc->path);
+ if (loc->path)
+ frame->local = gf_strdup(loc->path);
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_open_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->open,
- loc, flags, fd, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_open_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
+ return 0;
}
-
int
-io_stats_create (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, mode_t mode,
- mode_t umask, fd_t *fd, dict_t *xdata)
+io_stats_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- if (loc->path)
- frame->local = gf_strdup (loc->path);
+ if (loc->path)
+ frame->local = gf_strdup(loc->path);
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_create_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->create,
- loc, flags, mode, umask, fd, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_create_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, flags, mode, umask, fd,
+ xdata);
+ return 0;
}
-
int
-io_stats_readv (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata)
+io_stats_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- frame->local = fd;
+ frame->local = fd;
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_readv_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readv,
- fd, size, offset, flags, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_readv_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readv, fd, size, offset, flags, xdata);
+ return 0;
}
-
int
-io_stats_writev (call_frame_t *frame, xlator_t *this,
- fd_t *fd, struct iovec *vector,
- int32_t count, off_t offset,
- uint32_t flags, struct iobref *iobref, dict_t *xdata)
+io_stats_writev(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int32_t count, off_t offset,
+ uint32_t flags, struct iobref *iobref, dict_t *xdata)
{
- int len = 0;
+ int len = 0;
- if (fd->inode)
- frame->local = fd->inode;
- len = iov_length (vector, count);
+ if (fd->inode)
+ frame->local = fd->inode;
+ len = iov_length(vector, count);
- ios_bump_write (this, fd, len);
- START_FOP_LATENCY (frame);
-
- STACK_WIND (frame, io_stats_writev_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->writev,
- fd, vector, count, offset, flags, iobref, xdata);
- return 0;
+ ios_bump_write(this, fd, len);
+ START_FOP_LATENCY(frame);
+ STACK_WIND(frame, io_stats_writev_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector, count, offset,
+ flags, iobref, xdata);
+ return 0;
}
-
int
-io_stats_statfs (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xdata)
+io_stats_copy_file_range(call_frame_t *frame, xlator_t *this, fd_t *fd_in,
+ off_t off_in, fd_t *fd_out, off_t off_out, size_t len,
+ uint32_t flags, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_statfs_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->statfs,
- loc, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_copy_file_range_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->copy_file_range, fd_in, off_in, fd_out,
+ off_out, len, flags, xdata);
+ return 0;
}
-
int
-io_stats_flush (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *xdata)
+io_stats_statfs(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_flush_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->flush,
- fd, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_statfs_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->statfs, loc, xdata);
+ return 0;
}
-
int
-io_stats_fsync (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int32_t flags, dict_t *xdata)
+io_stats_flush(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_fsync_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsync,
- fd, flags, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_flush_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->flush, fd, xdata);
+ return 0;
}
-
int
-conditional_dump (dict_t *dict, char *key, data_t *value, void *data)
-{
- struct {
- xlator_t *this;
- inode_t *inode;
- const char *path;
- } *stub;
- xlator_t *this = NULL;
- char *filename = NULL;
- FILE *logfp = NULL;
- struct ios_dump_args args = {0};
- int pid, namelen;
- char dump_key[100];
- char *slash_ptr = NULL;
-
- stub = data;
- this = stub->this;
-
- /* Create a file name that is appended with the io-stats instance
- name as well. This helps when there is more than a single io-stats
- instance in the graph, or the client and server processes are running
- on the same node */
- /* hmmm... no check for this */
- /* name format: <passed in path/filename>.<xlator name slashes to -> */
- namelen = value->len + strlen (this->name) + 2; /* '.' and '\0' */
- filename = alloca0 (namelen);
- memcpy (filename, data_to_str (value), value->len);
- memcpy (filename + value->len, ".", 1);
- memcpy (filename + value->len + 1, this->name, strlen(this->name));
-
- /* convert any slashes to '-' so that fopen works correctly */
- slash_ptr = strchr (filename + value->len + 1, '/');
- while (slash_ptr) {
- *slash_ptr = '-';
- slash_ptr = strchr (slash_ptr, '/');
- }
+io_stats_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+ dict_t *xdata)
+{
+ START_FOP_LATENCY(frame);
- pid = getpid ();
+ STACK_WIND(frame, io_stats_fsync_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsync, fd, flags, xdata);
+ return 0;
+}
- if (!strncmp (filename, "", 1)) {
- gf_log (this->name, GF_LOG_ERROR, "No filename given");
- return -1;
- }
- logfp = fopen (filename, "w+");
- if (!logfp) {
- gf_log (this->name, GF_LOG_ERROR, "failed to open %s "
- "for writing", filename);
- return -1;
- }
- sprintf (dump_key, "*io*stat*%d_json_dump", pid);
- if (fnmatch (dump_key, key, 0) == 0) {
- (void) ios_dump_args_init (
- &args, IOS_DUMP_TYPE_JSON_FILE,
- logfp);
- } else {
- (void) ios_dump_args_init (&args, IOS_DUMP_TYPE_FILE,
- logfp);
- }
- io_stats_dump (this, &args, GF_CLI_INFO_ALL, _gf_false);
- fclose (logfp);
- return 0;
+int
+conditional_dump(dict_t *dict, char *key, data_t *value, void *data)
+{
+ struct {
+ xlator_t *this;
+ inode_t *inode;
+ const char *path;
+ } * stub;
+ xlator_t *this = NULL;
+ char *filename = NULL;
+ FILE *logfp = NULL;
+ struct ios_dump_args args = {0};
+ int pid, namelen, dirlen;
+ char dump_key[100];
+ char *slash_ptr = NULL;
+ char *path_in_value = NULL;
+ char *identifier = NULL;
+ struct ios_conf *conf = NULL;
+
+ stub = data;
+ this = stub->this;
+ conf = this->private;
+
+ /* Don't do this on 'brick-side', only do this on client side */
+ /* Addresses CVE-2018-14659 */
+ if (this->ctx->process_mode != GF_CLIENT_PROCESS) {
+ gf_log(this->name, GF_LOG_DEBUG,
+ "taking io-stats dump using setxattr not permitted on brick."
+ " Use 'gluster profile' instead");
+ return -1;
+ }
+
+ /* Create a file name that is appended with the io-stats instance
+ name as well. This helps when there is more than a single io-stats
+ instance in the graph, or the client and server processes are running
+ on the same node */
+ /* For the sanity of where the file should be located, we should make
+ sure file is written only inside RUNDIR (ie, /var/run/gluster) */
+ /* TODO: provide an option to dump it to different directory of
+ choice, based on options */
+ /* name format: /var/run/gluster/<passed in path/filename>.<xlator name
+ * slashes to -> */
+
+ path_in_value = alloca0(value->len + 1);
+
+ /* We need a memcpy here because of the way dict_unserialize works */
+
+ memcpy(path_in_value, data_to_str(value), value->len);
+ path_in_value[value->len] = '\0';
+
+ if (strstr(path_in_value, "../")) {
+ gf_log(this->name, GF_LOG_ERROR, "%s: no \"../\" allowed in path",
+ path_in_value);
+ return -1;
+ }
+
+ if (path_in_value[0] == '/') {
+ path_in_value = path_in_value + 1;
+ }
+
+ dirlen = strlen(IOS_STATS_DUMP_DIR);
+ if (conf->unique_id) {
+ /* this->name will be the same for all bricks of the volume */
+ identifier = conf->unique_id;
+ } else {
+ identifier = this->name;
+ }
+
+ namelen = (dirlen + value->len + strlen(identifier) + 3);
+ /* +3 for '/', '.' and '\0' added in snprintf below*/
+
+ filename = alloca0(namelen);
+ snprintf(filename, namelen, "%s/%s.%s", IOS_STATS_DUMP_DIR, path_in_value,
+ identifier);
+
+ /* convert any slashes to '-' so that fopen works correctly */
+ slash_ptr = strchr(filename + dirlen + 1, '/');
+ while (slash_ptr) {
+ *slash_ptr = '-';
+ slash_ptr = strchr(slash_ptr, '/');
+ }
+
+ pid = getpid();
+
+ if (!strncmp(filename, "", 1)) {
+ gf_log(this->name, GF_LOG_ERROR, "No filename given");
+ return -1;
+ }
+ logfp = fopen(filename, "w+");
+ if (!logfp) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to open %s "
+ "for writing",
+ filename);
+ return -1;
+ }
+ sprintf(dump_key, "*io*stat*%d_json_dump", pid);
+ if (fnmatch(dump_key, key, 0) == 0) {
+ (void)ios_dump_args_init(&args, IOS_DUMP_TYPE_JSON_FILE, logfp);
+ } else {
+ (void)ios_dump_args_init(&args, IOS_DUMP_TYPE_FILE, logfp);
+ }
+ io_stats_dump(this, &args, GF_IOS_INFO_ALL, _gf_false);
+ fclose(logfp);
+ return 0;
}
int
-_ios_destroy_dump_thread (struct ios_conf *conf) {
- conf->dump_thread_should_die = _gf_true;
- if (conf->ios_dump_interval > 0) {
- (void) pthread_cancel (conf->dump_thread);
- (void) pthread_join (conf->dump_thread, NULL);
- }
- return 0;
+_ios_destroy_dump_thread(struct ios_conf *conf)
+{
+ conf->dump_thread_should_die = _gf_true;
+ if (conf->dump_thread_running) {
+ (void)pthread_cancel(conf->dump_thread);
+ (void)pthread_join(conf->dump_thread, NULL);
+ }
+ return 0;
}
void *
-_ios_dump_thread (xlator_t *this) {
- struct ios_conf *conf = NULL;
- FILE *stats_logfp = NULL;
- FILE *samples_logfp = NULL;
- struct ios_dump_args args = {0};
- int i;
- int stats_bytes_written = 0;
- int samples_bytes_written = 0;
- char stats_filename[PATH_MAX];
- char samples_filename[PATH_MAX];
- char *xlator_name;
- char *instance_name;
- gf_boolean_t log_stats_fopen_failure = _gf_true;
- gf_boolean_t log_samples_fopen_failure = _gf_true;
- int old_cancel_type;
-
- conf = this->private;
- gf_log (this->name, GF_LOG_INFO, "IO stats dump thread started, "
- "polling IO stats every %d seconds", conf->ios_dump_interval);
- xlator_name = strdupa (conf->unique_id);
- for (i = 0; i < strlen (xlator_name); i++) {
- if (xlator_name[i] == '/')
- xlator_name[i] = '_';
+_ios_dump_thread(xlator_t *this)
+{
+ struct ios_conf *conf = NULL;
+ FILE *stats_logfp = NULL;
+ FILE *samples_logfp = NULL;
+ struct ios_dump_args args = {0};
+ int i;
+ int stats_bytes_written = 0;
+ int samples_bytes_written = 0;
+ char stats_filename[PATH_MAX];
+ char samples_filename[PATH_MAX];
+ char *xlator_name;
+ char *instance_name;
+ gf_boolean_t log_stats_fopen_failure = _gf_true;
+ gf_boolean_t log_samples_fopen_failure = _gf_true;
+ int old_cancel_type;
+
+ conf = this->private;
+ gf_log(this->name, GF_LOG_INFO,
+ "IO stats dump thread started, "
+ "polling IO stats every %d seconds",
+ conf->ios_dump_interval);
+ xlator_name = strdupa(conf->unique_id);
+ for (i = 0; i < strlen(xlator_name); i++) {
+ if (xlator_name[i] == '/')
+ xlator_name[i] = '_';
+ }
+ instance_name = this->instance_name;
+ if (this->name && strcmp(this->name, "glustershd") == 0) {
+ xlator_name = "shd";
+ } else if (this->prev && strcmp(this->prev->name, "nfs-server") == 0) {
+ xlator_name = "nfsd";
+ instance_name = this->prev->instance_name;
+ }
+ if (sys_mkdir(_IOS_DUMP_DIR, S_IRWXU | S_IRWXO | S_IRWXG) == (-1)) {
+ if (errno != EEXIST) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "could not create stats-dump directory %s", _IOS_DUMP_DIR);
+ goto out;
}
- instance_name = this->instance_name;
- if (this->name && strcmp (this->name, "glustershd") == 0) {
- xlator_name = "shd";
- } else if (this->prev &&
- strcmp (this->prev->name, "nfs-server") == 0) {
- xlator_name = "nfsd";
- instance_name = this->prev->instance_name;
+ }
+ if (sys_mkdir(_IOS_SAMP_DIR, S_IRWXU | S_IRWXO | S_IRWXG) == (-1)) {
+ if (errno != EEXIST) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "could not create stats-sample directory %s", _IOS_SAMP_DIR);
+ goto out;
}
- if (sys_mkdir (_IOS_DUMP_DIR, S_IRWXU | S_IRWXO | S_IRWXG) == (-1)) {
- if (errno != EEXIST) {
- gf_log (this->name, GF_LOG_ERROR,
- "could not create stats-dump directory %s",
- _IOS_DUMP_DIR);
- goto out;
- }
- }
- if (sys_mkdir (_IOS_SAMP_DIR, S_IRWXU | S_IRWXO | S_IRWXG) == (-1)) {
- if (errno != EEXIST) {
- gf_log (this->name, GF_LOG_ERROR,
- "could not create stats-sample directory %s",
- _IOS_SAMP_DIR);
- goto out;
- }
- }
- if (instance_name) {
- stats_bytes_written = snprintf (stats_filename, PATH_MAX,
- "%s/%s_%s_%s.dump", _IOS_DUMP_DIR,
- __progname, xlator_name, instance_name);
- samples_bytes_written = snprintf (samples_filename, PATH_MAX,
- "%s/%s_%s_%s.samp", _IOS_SAMP_DIR,
- __progname, xlator_name, instance_name);
- } else {
- stats_bytes_written = snprintf (stats_filename, PATH_MAX,
- "%s/%s_%s.dump", _IOS_DUMP_DIR, __progname,
- xlator_name);
- samples_bytes_written = snprintf (samples_filename, PATH_MAX,
- "%s/%s_%s.samp", _IOS_SAMP_DIR, __progname,
- xlator_name);
- }
- if ((stats_bytes_written >= PATH_MAX) ||
- (samples_bytes_written >= PATH_MAX)) {
- gf_log (this->name, GF_LOG_ERROR,
- "Invalid path for stats dump (%s) and/or latency "
- "samples (%s)", stats_filename, samples_filename);
- goto out;
+ }
+ if (instance_name) {
+ stats_bytes_written = snprintf(stats_filename, PATH_MAX,
+ "%s/%s_%s_%s.dump", _IOS_DUMP_DIR,
+ __progname, xlator_name, instance_name);
+ samples_bytes_written = snprintf(
+ samples_filename, PATH_MAX, "%s/%s_%s_%s.samp", _IOS_SAMP_DIR,
+ __progname, xlator_name, instance_name);
+ } else {
+ stats_bytes_written = snprintf(stats_filename, PATH_MAX,
+ "%s/%s_%s.dump", _IOS_DUMP_DIR,
+ __progname, xlator_name);
+ samples_bytes_written = snprintf(samples_filename, PATH_MAX,
+ "%s/%s_%s.samp", _IOS_SAMP_DIR,
+ __progname, xlator_name);
+ }
+ if ((stats_bytes_written >= PATH_MAX) ||
+ (samples_bytes_written >= PATH_MAX)) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Invalid path for stats dump (%s) and/or latency "
+ "samples (%s)",
+ stats_filename, samples_filename);
+ goto out;
+ }
+ while (1) {
+ if (conf->dump_thread_should_die)
+ break;
+ (void)pthread_setcanceltype(PTHREAD_CANCEL_ASYNCHRONOUS,
+ &old_cancel_type);
+ sleep(conf->ios_dump_interval);
+ (void)pthread_setcanceltype(PTHREAD_CANCEL_DEFERRED, &old_cancel_type);
+ /*
+ * It's not clear whether we should reopen this each time, or
+ * just hold it open and rewind/truncate on each iteration.
+ * Leaving it alone for now.
+ */
+ stats_logfp = fopen(stats_filename, "w+");
+ if (stats_logfp) {
+ (void)ios_dump_args_init(&args, conf->dump_format, stats_logfp);
+ io_stats_dump(this, &args, GF_IOS_INFO_ALL, _gf_false);
+ fclose(stats_logfp);
+ log_stats_fopen_failure = _gf_true;
+ } else if (log_stats_fopen_failure) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "could not open stats-dump file %s (%s)", stats_filename,
+ strerror(errno));
+ log_stats_fopen_failure = _gf_false;
}
- while (1) {
- if (conf->dump_thread_should_die)
- break;
- (void) pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS,
- &old_cancel_type);
- sleep (conf->ios_dump_interval);
- (void) pthread_setcanceltype (PTHREAD_CANCEL_DEFERRED,
- &old_cancel_type);
- /*
- * It's not clear whether we should reopen this each time, or
- * just hold it open and rewind/truncate on each iteration.
- * Leaving it alone for now.
- */
- stats_logfp = fopen (stats_filename, "w+");
- if (stats_logfp) {
- (void) ios_dump_args_init (&args,
- conf->dump_format,
- stats_logfp);
- io_stats_dump (this, &args, GF_CLI_INFO_ALL, _gf_false);
- fclose (stats_logfp);
- log_stats_fopen_failure = _gf_true;
- } else if (log_stats_fopen_failure) {
- gf_log (this->name, GF_LOG_ERROR,
- "could not open stats-dump file %s (%s)",
- stats_filename, strerror(errno));
- log_stats_fopen_failure = _gf_false;
- }
- samples_logfp = fopen (samples_filename, "w+");
- if (samples_logfp) {
- io_stats_dump_latency_samples_logfp (this,
- samples_logfp);
- fclose (samples_logfp);
- log_samples_fopen_failure = _gf_true;
- } else if (log_samples_fopen_failure) {
- gf_log (this->name, GF_LOG_ERROR,
- "could not open samples-dump file %s (%s)",
- samples_filename, strerror(errno));
- log_samples_fopen_failure = _gf_false;
- }
+ samples_logfp = fopen(samples_filename, "w+");
+ if (samples_logfp) {
+ io_stats_dump_latency_samples_logfp(this, samples_logfp);
+ fclose(samples_logfp);
+ log_samples_fopen_failure = _gf_true;
+ } else if (log_samples_fopen_failure) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "could not open samples-dump file %s (%s)", samples_filename,
+ strerror(errno));
+ log_samples_fopen_failure = _gf_false;
}
+ }
out:
- gf_log (this->name, GF_LOG_INFO, "IO stats dump thread terminated");
- return NULL;
+ conf->dump_thread_running = _gf_false;
+ gf_log(this->name, GF_LOG_INFO, "IO stats dump thread terminated");
+ return NULL;
}
static gf_boolean_t
-match_special_xattr (dict_t *d, char *k, data_t *val, void *mdata)
+match_special_xattr(dict_t *d, char *k, data_t *val, void *mdata)
{
- gf_boolean_t ret = _gf_false;
- if (fnmatch ("*io*stat*dump", k, 0) == 0) {
- ret = _gf_true;
- }
+ gf_boolean_t ret = _gf_false;
+ if (fnmatch("*io*stat*dump", k, 0) == 0) {
+ ret = _gf_true;
+ }
- return ret;
+ return ret;
}
int
-io_stats_setxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *dict,
- int32_t flags, dict_t *xdata)
+io_stats_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- struct {
- xlator_t *this;
- inode_t *inode;
- const char *path;
- } stub;
+ struct {
+ xlator_t *this;
+ inode_t *inode;
+ const char *path;
+ } stub;
- stub.this = this;
- stub.inode = loc->inode;
- stub.path = loc->path;
+ stub.this = this;
+ stub.inode = loc->inode;
+ stub.path = loc->path;
- (void) dict_foreach_match (dict, match_special_xattr, NULL,
- conditional_dump, &stub);
+ (void)dict_foreach_match(dict, match_special_xattr, NULL, conditional_dump,
+ &stub);
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_setxattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setxattr,
- loc, dict, flags, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_setxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, loc, dict, flags, xdata);
+ return 0;
}
-
int
-io_stats_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
+io_stats_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_getxattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr,
- loc, name, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_getxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+ return 0;
}
-
int
-io_stats_removexattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
+io_stats_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_removexattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->removexattr,
- loc, name, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_removexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
+ return 0;
}
-
int
-io_stats_fsetxattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *dict,
- int32_t flags, dict_t *xdata)
+io_stats_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_fsetxattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetxattr,
- fd, dict, flags, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_fsetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+ return 0;
}
-
int
-io_stats_fgetxattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, const char *name, dict_t *xdata)
+io_stats_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_fgetxattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fgetxattr,
- fd, name, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_fgetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
+ return 0;
}
-
int
-io_stats_fremovexattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, const char *name, dict_t *xdata)
+io_stats_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_fremovexattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fremovexattr,
- fd, name, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_fremovexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fremovexattr, fd, name, xdata);
+ return 0;
}
-
int
-io_stats_opendir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, fd_t *fd, dict_t *xdata)
+io_stats_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata)
{
+ START_FOP_LATENCY(frame);
- START_FOP_LATENCY (frame);
-
- STACK_WIND (frame, io_stats_opendir_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->opendir,
- loc, fd, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_opendir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->opendir, loc, fd, xdata);
+ return 0;
}
int
-io_stats_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, dict_t *dict)
+io_stats_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *dict)
{
- frame->local = fd->inode;
- START_FOP_LATENCY (frame);
+ frame->local = fd->inode;
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_readdirp_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdirp,
- fd, size, offset, dict);
- return 0;
+ STACK_WIND(frame, io_stats_readdirp_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdirp, fd, size, offset, dict);
+ return 0;
}
-
int
-io_stats_readdir (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset, dict_t *xdata)
+io_stats_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_readdir_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdir,
- fd, size, offset, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_readdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdir, fd, size, offset, xdata);
+ return 0;
}
-
int
-io_stats_fsyncdir (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int32_t datasync, dict_t *xdata)
+io_stats_fsyncdir(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ int32_t datasync, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_fsyncdir_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsyncdir,
- fd, datasync, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_fsyncdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsyncdir, fd, datasync, xdata);
+ return 0;
}
-
int
-io_stats_access (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t mask, dict_t *xdata)
+io_stats_access(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+ dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_access_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->access,
- loc, mask, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_access_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->access, loc, mask, xdata);
+ return 0;
}
-
int
-io_stats_ftruncate (call_frame_t *frame, xlator_t *this,
- fd_t *fd, off_t offset, dict_t *xdata)
+io_stats_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_ftruncate_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->ftruncate,
- fd, offset, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_ftruncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+ return 0;
}
-
int
-io_stats_fsetattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, struct iatt *stbuf, int32_t valid, dict_t *xdata)
+io_stats_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_setattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetattr,
- fd, stbuf, valid, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_setattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetattr, fd, stbuf, valid, xdata);
+ return 0;
}
-
int
-io_stats_fstat (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *xdata)
+io_stats_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_fstat_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fstat,
- fd, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_fstat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fstat, fd, xdata);
+ return 0;
}
-
int
io_stats_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
- off_t offset, size_t len, dict_t *xdata)
+ off_t offset, size_t len, dict_t *xdata)
{
- START_FOP_LATENCY(frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND(frame, io_stats_fallocate_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fallocate, fd, mode, offset, len,
- xdata);
+ STACK_WIND(frame, io_stats_fallocate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fallocate, fd, mode, offset, len,
+ xdata);
- return 0;
+ return 0;
}
-
int
io_stats_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- size_t len, dict_t *xdata)
+ size_t len, dict_t *xdata)
{
- START_FOP_LATENCY(frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND(frame, io_stats_discard_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->discard, fd, offset, len, xdata);
+ STACK_WIND(frame, io_stats_discard_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->discard, fd, offset, len, xdata);
- return 0;
+ return 0;
}
int
io_stats_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- off_t len, dict_t *xdata)
+ off_t len, dict_t *xdata)
{
- START_FOP_LATENCY(frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND(frame, io_stats_zerofill_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->zerofill, fd, offset, len, xdata);
+ STACK_WIND(frame, io_stats_zerofill_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->zerofill, fd, offset, len, xdata);
- return 0;
+ return 0;
}
int32_t
-io_stats_ipc (call_frame_t *frame, xlator_t *this, int32_t op, dict_t *xdata)
+io_stats_ipc(call_frame_t *frame, xlator_t *this, int32_t op, dict_t *xdata)
{
- START_FOP_LATENCY(frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_ipc_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->ipc, op, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_ipc_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ipc, op, xdata);
+ return 0;
}
int
-io_stats_lk (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
+io_stats_lk(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
+ struct gf_flock *lock, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_lk_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lk,
- fd, cmd, lock, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_lk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lk, fd, cmd, lock, xdata);
+ return 0;
}
int
-io_stats_rchecksum (call_frame_t *frame, xlator_t *this,
- fd_t *fd, off_t offset, int32_t len, dict_t *xdata)
+io_stats_rchecksum(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ int32_t len, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_rchecksum_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rchecksum,
- fd, offset, len, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_rchecksum_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rchecksum, fd, offset, len, xdata);
+ return 0;
}
int
-io_stats_seek (call_frame_t *frame, xlator_t *this, fd_t *fd,
- off_t offset, gf_seek_what_t what, dict_t *xdata)
+io_stats_seek(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ gf_seek_what_t what, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_seek_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->seek,
- fd, offset, what, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_seek_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->seek, fd, offset, what, xdata);
+ return 0;
}
int
-io_stats_lease (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct gf_lease *lease, dict_t *xdata)
+io_stats_lease(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct gf_lease *lease, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_lease_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lease,
- loc, lease, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_lease_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lease, loc, lease, xdata);
+ return 0;
}
int
-io_stats_getactivelk (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *xdata)
+io_stats_getactivelk(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_getactivelk_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getactivelk,
- loc, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_getactivelk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->getactivelk, loc, xdata);
+ return 0;
}
int
-io_stats_setactivelk (call_frame_t *frame, xlator_t *this, loc_t *loc,
- lock_migration_info_t *locklist, dict_t *xdata)
+io_stats_setactivelk(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ lock_migration_info_t *locklist, dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_setactivelk_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setactivelk,
- loc, locklist, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_setactivelk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setactivelk, loc, locklist, xdata);
+ return 0;
}
int
-io_stats_compound (call_frame_t *frame, xlator_t *this,
- void *args, dict_t *xdata)
+io_stats_compound(call_frame_t *frame, xlator_t *this, void *args,
+ dict_t *xdata)
{
- START_FOP_LATENCY (frame);
+ START_FOP_LATENCY(frame);
- STACK_WIND (frame, io_stats_compound_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->compound,
- args, xdata);
- return 0;
+ STACK_WIND(frame, io_stats_compound_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->compound, args, xdata);
+ return 0;
}
int
-io_stats_release (xlator_t *this, fd_t *fd)
+io_stats_release(xlator_t *this, fd_t *fd)
{
- struct ios_fd *iosfd = NULL;
- struct ios_conf *conf = NULL;
-
- BUMP_FOP (RELEASE);
+ struct ios_fd *iosfd = NULL;
+ struct ios_conf *conf = NULL;
- conf = this->private;
+ BUMP_FOP(RELEASE);
- LOCK (&conf->lock);
+ conf = this->private;
+ if (conf) {
+ LOCK(&conf->lock);
{
- conf->cumulative.nr_opens--;
+ conf->cumulative.nr_opens--;
}
- UNLOCK (&conf->lock);
+ UNLOCK(&conf->lock);
+ }
- ios_fd_ctx_get (fd, this, &iosfd);
- if (iosfd) {
- io_stats_dump_fd (this, iosfd);
+ ios_fd_ctx_get(fd, this, &iosfd);
+ if (iosfd) {
+ io_stats_dump_fd(this, iosfd);
- GF_FREE (iosfd->filename);
- GF_FREE (iosfd);
- }
+ GF_FREE(iosfd->filename);
+ GF_FREE(iosfd);
+ }
- return 0;
+ return 0;
}
-
int
-io_stats_releasedir (xlator_t *this, fd_t *fd)
+io_stats_releasedir(xlator_t *this, fd_t *fd)
{
- BUMP_FOP (RELEASEDIR);
+ BUMP_FOP(RELEASEDIR);
- return 0;
+ return 0;
}
-
int
-io_stats_forget (xlator_t *this, inode_t *inode)
+io_stats_forget(xlator_t *this, inode_t *inode)
{
- BUMP_FOP (FORGET);
- ios_stats_cleanup (this, inode);
- return 0;
+ BUMP_FOP(FORGET);
+ ios_stats_cleanup(this, inode);
+ return 0;
}
static int
-ios_init_top_stats (struct ios_conf *conf)
+ios_init_top_stats(struct ios_conf *conf)
{
- int i = 0;
+ int i = 0;
- GF_ASSERT (conf);
+ GF_ASSERT(conf);
- for (i = 0; i <IOS_STATS_TYPE_MAX; i++) {
- conf->list[i].iosstats = GF_CALLOC (1,
- sizeof(*conf->list[i].iosstats),
- gf_io_stats_mt_ios_stat);
+ for (i = 0; i < IOS_STATS_TYPE_MAX; i++) {
+ conf->list[i].iosstats = GF_CALLOC(1, sizeof(*conf->list[i].iosstats),
+ gf_io_stats_mt_ios_stat);
- if (!conf->list[i].iosstats)
- return -1;
+ if (!conf->list[i].iosstats)
+ return -1;
- INIT_LIST_HEAD(&conf->list[i].iosstats->list);
- LOCK_INIT (&conf->list[i].lock);
- }
+ INIT_LIST_HEAD(&conf->list[i].iosstats->list);
+ LOCK_INIT(&conf->list[i].lock);
+ }
- for (i = 0; i < IOS_STATS_THRU_MAX; i ++) {
- conf->thru_list[i].iosstats = GF_CALLOC (1,
- sizeof (*conf->thru_list[i].iosstats),
- gf_io_stats_mt_ios_stat);
+ for (i = 0; i < IOS_STATS_THRU_MAX; i++) {
+ conf->thru_list[i].iosstats = GF_CALLOC(
+ 1, sizeof(*conf->thru_list[i].iosstats), gf_io_stats_mt_ios_stat);
- if (!conf->thru_list[i].iosstats)
- return -1;
+ if (!conf->thru_list[i].iosstats)
+ return -1;
- INIT_LIST_HEAD(&conf->thru_list[i].iosstats->list);
- LOCK_INIT (&conf->thru_list[i].lock);
- }
+ INIT_LIST_HEAD(&conf->thru_list[i].iosstats->list);
+ LOCK_INIT(&conf->thru_list[i].lock);
+ }
- return 0;
+ return 0;
}
static void
-ios_destroy_top_stats (struct ios_conf *conf)
-{
- int i = 0;
- struct ios_stat_head *list_head = NULL;
- struct ios_stat_list *entry = NULL;
- struct ios_stat_list *tmp = NULL;
- struct ios_stat_list *list = NULL;
- struct ios_stat *stat = NULL;
-
- GF_ASSERT (conf);
-
- LOCK (&conf->lock);
-
- conf->cumulative.nr_opens = 0;
- conf->cumulative.max_nr_opens = 0;
- conf->cumulative.max_openfd_time.tv_sec = 0;
- conf->cumulative.max_openfd_time.tv_usec = 0;
-
- for (i = 0; i < IOS_STATS_TYPE_MAX; i++) {
- list_head = &conf->list[i];
- if (!list_head)
- continue;
- list_for_each_entry_safe (entry, tmp,
- &list_head->iosstats->list, list) {
- list = entry;
- stat = list->iosstat;
- ios_stat_unref (stat);
- list_del (&list->list);
- GF_FREE (list);
- list_head->members--;
- }
- GF_FREE (list_head->iosstats);
- }
+ios_destroy_top_stats(struct ios_conf *conf)
+{
+ int i = 0;
+ struct ios_stat_head *list_head = NULL;
+ struct ios_stat_list *entry = NULL;
+ struct ios_stat_list *tmp = NULL;
+ struct ios_stat_list *list = NULL;
+ struct ios_stat *stat = NULL;
- for (i = 0; i < IOS_STATS_THRU_MAX; i++) {
- list_head = &conf->thru_list[i];
- if (!list_head)
- continue;
- list_for_each_entry_safe (entry, tmp,
- &list_head->iosstats->list, list) {
- list = entry;
- stat = list->iosstat;
- ios_stat_unref (stat);
- list_del (&list->list);
- GF_FREE (list);
- list_head->members--;
- }
- GF_FREE (list_head->iosstats);
+ GF_ASSERT(conf);
+
+ LOCK(&conf->lock);
+
+ conf->cumulative.nr_opens = 0;
+ conf->cumulative.max_nr_opens = 0;
+ conf->cumulative.max_openfd_time.tv_sec = 0;
+ conf->cumulative.max_openfd_time.tv_usec = 0;
+
+ for (i = 0; i < IOS_STATS_TYPE_MAX; i++) {
+ list_head = &conf->list[i];
+ if (!list_head)
+ continue;
+ list_for_each_entry_safe(entry, tmp, &list_head->iosstats->list, list)
+ {
+ list = entry;
+ stat = list->iosstat;
+ ios_stat_unref(stat);
+ list_del(&list->list);
+ GF_FREE(list);
+ list_head->members--;
+ }
+ GF_FREE(list_head->iosstats);
+ }
+
+ for (i = 0; i < IOS_STATS_THRU_MAX; i++) {
+ list_head = &conf->thru_list[i];
+ if (!list_head)
+ continue;
+ list_for_each_entry_safe(entry, tmp, &list_head->iosstats->list, list)
+ {
+ list = entry;
+ stat = list->iosstat;
+ ios_stat_unref(stat);
+ list_del(&list->list);
+ GF_FREE(list);
+ list_head->members--;
}
+ GF_FREE(list_head->iosstats);
+ }
- UNLOCK (&conf->lock);
+ UNLOCK(&conf->lock);
- return;
+ return;
}
-static int
-io_stats_clear (struct ios_conf *conf)
+static void
+io_stats_clear(struct ios_conf *conf)
{
- struct timeval now;
- int ret = -1;
+ time_t now = 0;
- GF_ASSERT (conf);
+ GF_ASSERT(conf);
+ now = gf_time();
- if (!gettimeofday (&now, NULL))
- {
- LOCK (&conf->lock);
- {
- ios_global_stats_clear (&conf->cumulative, &now);
- ios_global_stats_clear (&conf->incremental, &now);
- conf->increment = 0;
- }
- UNLOCK (&conf->lock);
- ret = 0;
- }
-
- return ret;
+ LOCK(&conf->lock);
+ {
+ ios_global_stats_clear(&conf->cumulative, now);
+ ios_global_stats_clear(&conf->incremental, now);
+ conf->increment = 0;
+ }
+ UNLOCK(&conf->lock);
}
int32_t
-io_priv (xlator_t *this)
-{
- int i;
- char key[GF_DUMP_MAX_BUF_LEN];
- char key_prefix_cumulative[GF_DUMP_MAX_BUF_LEN];
- char key_prefix_incremental[GF_DUMP_MAX_BUF_LEN];
- double min, max, avg;
- uint64_t count, total;
- struct ios_conf *conf = NULL;
-
- conf = this->private;
- if (!conf)
- return -1;
-
- if(!conf->count_fop_hits || !conf->measure_latency)
- return -1;
-
- gf_proc_dump_write ("cumulative.data_read", "%"GF_PRI_ATOMIC,
- GF_ATOMIC_GET (conf->cumulative.data_read));
- gf_proc_dump_write ("cumulative.data_written", "%"GF_PRI_ATOMIC,
- GF_ATOMIC_GET (conf->cumulative.data_written));
-
- gf_proc_dump_write ("incremental.data_read", "%"GF_PRI_ATOMIC,
- GF_ATOMIC_GET (conf->incremental.data_read));
- gf_proc_dump_write ("incremental.data_written", "%"GF_PRI_ATOMIC,
- GF_ATOMIC_GET (conf->incremental.data_written));
-
- snprintf (key_prefix_cumulative, GF_DUMP_MAX_BUF_LEN, "%s.cumulative",
- this->name);
- snprintf (key_prefix_incremental, GF_DUMP_MAX_BUF_LEN, "%s.incremental",
- this->name);
-
- for (i = 0; i < GF_FOP_MAXVALUE; i++) {
- count = GF_ATOMIC_GET (conf->cumulative.fop_hits[i]);
- total = conf->cumulative.latency[i].total;
- min = conf->cumulative.latency[i].min;
- max = conf->cumulative.latency[i].max;
- avg = conf->cumulative.latency[i].avg;
-
- gf_proc_dump_build_key (key, key_prefix_cumulative, "%s",
- (char *)gf_fop_list[i]);
-
- gf_proc_dump_write (key,"%"PRId64",%"PRId64",%.03f,%.03f,%.03f",
- count, total, min, max, avg);
-
- count = GF_ATOMIC_GET (conf->incremental.fop_hits[i]);
- total = conf->incremental.latency[i].total;
- min = conf->incremental.latency[i].min;
- max = conf->incremental.latency[i].max;
- avg = conf->incremental.latency[i].avg;
-
- gf_proc_dump_build_key (key, key_prefix_incremental, "%s",
- (char *)gf_fop_list[i]);
-
- gf_proc_dump_write (key,"%"PRId64",%"PRId64",%.03f,%.03f,%.03f",
- count, total, min, max, avg);
+io_priv(xlator_t *this)
+{
+ int i;
+ char key[GF_DUMP_MAX_BUF_LEN];
+ char key_prefix_cumulative[GF_DUMP_MAX_BUF_LEN];
+ char key_prefix_incremental[GF_DUMP_MAX_BUF_LEN];
+ double min, max, avg;
+ uint64_t count, total;
+ struct ios_conf *conf = NULL;
- }
+ conf = this->private;
+ if (!conf)
+ return -1;
- return 0;
+ if (!conf->count_fop_hits || !conf->measure_latency)
+ return -1;
+
+ gf_proc_dump_write("cumulative.data_read", "%" GF_PRI_ATOMIC,
+ GF_ATOMIC_GET(conf->cumulative.data_read));
+ gf_proc_dump_write("cumulative.data_written", "%" GF_PRI_ATOMIC,
+ GF_ATOMIC_GET(conf->cumulative.data_written));
+
+ gf_proc_dump_write("incremental.data_read", "%" GF_PRI_ATOMIC,
+ GF_ATOMIC_GET(conf->incremental.data_read));
+ gf_proc_dump_write("incremental.data_written", "%" GF_PRI_ATOMIC,
+ GF_ATOMIC_GET(conf->incremental.data_written));
+
+ snprintf(key_prefix_cumulative, GF_DUMP_MAX_BUF_LEN, "%s.cumulative",
+ this->name);
+ snprintf(key_prefix_incremental, GF_DUMP_MAX_BUF_LEN, "%s.incremental",
+ this->name);
+
+ for (i = 0; i < GF_FOP_MAXVALUE; i++) {
+ count = GF_ATOMIC_GET(conf->cumulative.fop_hits[i]);
+ total = conf->cumulative.latency[i].total;
+ min = conf->cumulative.latency[i].min;
+ max = conf->cumulative.latency[i].max;
+ avg = conf->cumulative.latency[i].avg;
+
+ gf_proc_dump_build_key(key, key_prefix_cumulative, "%s",
+ (char *)gf_fop_list[i]);
+
+ gf_proc_dump_write(key, "%" PRId64 ",%" PRId64 ",%.03f,%.03f,%.03f",
+ count, total, min, max, avg);
+
+ count = GF_ATOMIC_GET(conf->incremental.fop_hits[i]);
+ total = conf->incremental.latency[i].total;
+ min = conf->incremental.latency[i].min;
+ max = conf->incremental.latency[i].max;
+ avg = conf->incremental.latency[i].avg;
+
+ gf_proc_dump_build_key(key, key_prefix_incremental, "%s",
+ (char *)gf_fop_list[i]);
+
+ gf_proc_dump_write(key, "%" PRId64 ",%" PRId64 ",%.03f,%.03f,%.03f",
+ count, total, min, max, avg);
+ }
+
+ return 0;
}
static void
-ios_set_log_format_code (struct ios_conf *conf)
+ios_set_log_format_code(struct ios_conf *conf, char *dump_format_str)
{
- if (strcmp (conf->dump_format_str, "json") == 0)
- conf->dump_format = IOS_DUMP_TYPE_JSON_FILE;
- else if (strcmp (conf->dump_format_str, "text") == 0)
- conf->dump_format = IOS_DUMP_TYPE_FILE;
- else if (strcmp (conf->dump_format_str, "dict") == 0)
- conf->dump_format = IOS_DUMP_TYPE_DICT;
- else if (strcmp (conf->dump_format_str, "samples") == 0)
- conf->dump_format = IOS_DUMP_TYPE_SAMPLES;
+ if (strcmp(dump_format_str, "json") == 0)
+ conf->dump_format = IOS_DUMP_TYPE_JSON_FILE;
+ else if (strcmp(dump_format_str, "text") == 0)
+ conf->dump_format = IOS_DUMP_TYPE_FILE;
+ else if (strcmp(dump_format_str, "dict") == 0)
+ conf->dump_format = IOS_DUMP_TYPE_DICT;
+ else if (strcmp(dump_format_str, "samples") == 0)
+ conf->dump_format = IOS_DUMP_TYPE_SAMPLES;
}
-int
-reconfigure (xlator_t *this, dict_t *options)
-{
- struct ios_conf *conf = NULL;
- int ret = -1;
- char *sys_log_str = NULL;
- char *log_format_str = NULL;
- char *logger_str = NULL;
- int sys_log_level = -1;
- char *log_str = NULL;
- int log_level = -1;
- int log_format = -1;
- int logger = -1;
- uint32_t log_buf_size = 0;
- uint32_t log_flush_timeout = 0;
- int32_t old_dump_interval;
-
- if (!this || !this->private)
- goto out;
-
- conf = this->private;
+void
+xlator_set_loglevel(xlator_t *this, int log_level)
+{
+ glusterfs_ctx_t *ctx = NULL;
+ glusterfs_graph_t *active = NULL;
+ xlator_t *top = NULL;
+ xlator_t *trav = this;
- GF_OPTION_RECONF ("dump-fd-stats", conf->dump_fd_stats, options, bool,
- out);
+ ctx = this->ctx;
+ GF_ASSERT(ctx);
+ active = ctx->active;
+ top = active->first;
- GF_OPTION_RECONF ("count-fop-hits", conf->count_fop_hits, options, bool,
- out);
+ if (log_level == -1)
+ return;
- GF_OPTION_RECONF ("latency-measurement", conf->measure_latency,
- options, bool, out);
+ if (ctx->cmd_args.brick_mux) {
+ /* Set log-level for all brick xlators */
+ top->loglevel = log_level;
- old_dump_interval = conf->ios_dump_interval;
- GF_OPTION_RECONF ("ios-dump-interval", conf->ios_dump_interval, options,
- int32, out);
- if ((old_dump_interval <= 0) && (conf->ios_dump_interval > 0)) {
- ret = gf_thread_create (&conf->dump_thread, NULL,
- (void *) &_ios_dump_thread, this, "iosdump");
- if (ret) {
- gf_log (this ? this->name : "io-stats",
- GF_LOG_ERROR, "Failed to start thread"
- "while reconfigure. Returning %d", ret);
- goto out;
- }
- }
+ /* Set log-level for parent xlator */
+ if (this->parents)
+ this->parents->xlator->loglevel = log_level;
- GF_OPTION_RECONF ("ios-sample-interval", conf->ios_sample_interval,
- options, int32, out);
- GF_OPTION_RECONF ("ios-dump-format", conf->dump_format_str, options,
- str, out);
- ios_set_log_format_code (conf);
- GF_OPTION_RECONF ("ios-sample-buf-size", conf->ios_sample_buf_size,
- options, int32, out);
- GF_OPTION_RECONF ("sys-log-level", sys_log_str, options, str, out);
- if (sys_log_str) {
- sys_log_level = glusterd_check_log_level (sys_log_str);
- set_sys_log_level (sys_log_level);
+ while (trav) {
+ trav->loglevel = log_level;
+ trav = trav->next;
}
+ } else {
+ gf_log_set_loglevel(this->ctx, log_level);
+ }
+}
- GF_OPTION_RECONF ("log-level", log_str, options, str, out);
- if (log_str) {
- log_level = glusterd_check_log_level (log_str);
- gf_log_set_loglevel (this->ctx, log_level);
- }
+int
+reconfigure(xlator_t *this, dict_t *options)
+{
+ struct ios_conf *conf = NULL;
+ int ret = -1;
+ char *sys_log_str = NULL;
+ char *log_format_str = NULL;
+ char *logger_str = NULL;
+ char *dump_format_str = NULL;
+ int sys_log_level = -1;
+ char *log_str = NULL;
+ int log_level = -1;
+ int log_format = -1;
+ int logger = -1;
+ uint32_t log_buf_size = 0;
+ uint32_t log_flush_timeout = 0;
+ int32_t old_dump_interval;
+ int32_t threads;
+
+ if (!this || !this->private)
+ goto out;
- GF_OPTION_RECONF ("logger", logger_str, options, str, out);
- if (logger_str) {
- logger = gf_check_logger (logger_str);
- gf_log_set_logger (logger);
- }
+ conf = this->private;
- GF_OPTION_RECONF ("log-format", log_format_str, options, str, out);
- if (log_format_str) {
- log_format = gf_check_log_format (log_format_str);
- gf_log_set_logformat (log_format);
- }
+ GF_OPTION_RECONF("dump-fd-stats", conf->dump_fd_stats, options, bool, out);
- GF_OPTION_RECONF ("log-buf-size", log_buf_size, options, uint32, out);
- gf_log_set_log_buf_size (log_buf_size);
+ GF_OPTION_RECONF("count-fop-hits", conf->count_fop_hits, options, bool,
+ out);
- GF_OPTION_RECONF ("log-flush-timeout", log_flush_timeout, options,
- time, out);
- gf_log_set_log_flush_timeout (log_flush_timeout);
+ GF_OPTION_RECONF("latency-measurement", conf->measure_latency, options,
+ bool, out);
- ret = 0;
+ old_dump_interval = conf->ios_dump_interval;
+ GF_OPTION_RECONF("ios-dump-interval", conf->ios_dump_interval, options,
+ int32, out);
+ if ((old_dump_interval <= 0) && (conf->ios_dump_interval > 0)) {
+ conf->dump_thread_running = _gf_true;
+ conf->dump_thread_should_die = _gf_false;
+ ret = gf_thread_create(&conf->dump_thread, NULL,
+ (void *)&_ios_dump_thread, this, "iosdump");
+ if (ret) {
+ conf->dump_thread_running = _gf_false;
+ gf_log(this ? this->name : "io-stats", GF_LOG_ERROR,
+ "Failed to start thread"
+ "while reconfigure. Returning %d",
+ ret);
+ goto out;
+ }
+ } else if ((old_dump_interval > 0) && (conf->ios_dump_interval == 0)) {
+ _ios_destroy_dump_thread(conf);
+ }
+
+ GF_OPTION_RECONF("ios-sample-interval", conf->ios_sample_interval, options,
+ int32, out);
+ GF_OPTION_RECONF("ios-dump-format", dump_format_str, options, str, out);
+ ios_set_log_format_code(conf, dump_format_str);
+ GF_OPTION_RECONF("ios-sample-buf-size", conf->ios_sample_buf_size, options,
+ int32, out);
+ GF_OPTION_RECONF("sys-log-level", sys_log_str, options, str, out);
+ if (sys_log_str) {
+ sys_log_level = glusterd_check_log_level(sys_log_str);
+ set_sys_log_level(sys_log_level);
+ }
+
+ GF_OPTION_RECONF("log-level", log_str, options, str, out);
+ if (log_str) {
+ log_level = glusterd_check_log_level(log_str);
+ /* Set loglevel for all children and server xlators */
+ xlator_set_loglevel(this, log_level);
+ }
+
+ GF_OPTION_RECONF("logger", logger_str, options, str, out);
+ if (logger_str) {
+ logger = gf_check_logger(logger_str);
+ gf_log_set_logger(logger);
+ }
+
+ GF_OPTION_RECONF("log-format", log_format_str, options, str, out);
+ if (log_format_str) {
+ log_format = gf_check_log_format(log_format_str);
+ gf_log_set_logformat(log_format);
+ }
+
+ GF_OPTION_RECONF("log-buf-size", log_buf_size, options, uint32, out);
+ gf_log_set_log_buf_size(log_buf_size);
+
+ GF_OPTION_RECONF("log-flush-timeout", log_flush_timeout, options, time,
+ out);
+ gf_log_set_log_flush_timeout(log_flush_timeout);
+
+ GF_OPTION_RECONF("threads", threads, options, int32, out);
+ gf_async_adjust_threads(threads);
+
+ ret = 0;
out:
- gf_log (this ? this->name : "io-stats",
- GF_LOG_DEBUG, "reconfigure returning %d", ret);
- return ret;
+ gf_log(this ? this->name : "io-stats", GF_LOG_DEBUG,
+ "reconfigure returning %d", ret);
+ return ret;
}
-
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- if (!this)
- return ret;
-
- ret = xlator_mem_acct_init (this, gf_io_stats_mt_end + 1);
+ if (!this)
+ return ret;
- if (ret != 0) {
- gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
- " failed");
- return ret;
- }
+ ret = xlator_mem_acct_init(this, gf_io_stats_mt_end + 1);
+ if (ret != 0) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Memory accounting init"
+ " failed");
return ret;
+ }
+
+ return ret;
}
void
-ios_conf_destroy (struct ios_conf *conf)
+ios_conf_destroy(struct ios_conf *conf)
{
- if (!conf)
- return;
+ if (!conf)
+ return;
- ios_destroy_top_stats (conf);
- _ios_destroy_dump_thread (conf);
- ios_destroy_sample_buf (conf->ios_sample_buf);
- LOCK_DESTROY (&conf->lock);
- GF_FREE(conf->dnscache);
- GF_FREE(conf);
+ ios_destroy_top_stats(conf);
+ _ios_destroy_dump_thread(conf);
+ ios_destroy_sample_buf(conf->ios_sample_buf);
+ LOCK_DESTROY(&conf->lock);
+ gf_dnscache_deinit(conf->dnscache);
+ GF_FREE(conf);
}
static void
-ios_init_stats (struct ios_global_stats *stats)
+ios_init_stats(struct ios_global_stats *stats)
{
- int i = 0;
+ int i = 0;
- GF_ATOMIC_INIT (stats->data_read, 0);
- GF_ATOMIC_INIT (stats->data_written, 0);
+ GF_ATOMIC_INIT(stats->data_read, 0);
+ GF_ATOMIC_INIT(stats->data_written, 0);
- for (i = 0; i < IOS_BLOCK_COUNT_SIZE; i++) {
- GF_ATOMIC_INIT (stats->block_count_write[i], 0);
- GF_ATOMIC_INIT (stats->block_count_read[i], 0);
- }
+ for (i = 0; i < IOS_BLOCK_COUNT_SIZE; i++) {
+ GF_ATOMIC_INIT(stats->block_count_write[i], 0);
+ GF_ATOMIC_INIT(stats->block_count_read[i], 0);
+ }
- for (i = 0; i < GF_FOP_MAXVALUE; i++)
- GF_ATOMIC_INIT (stats->fop_hits[i], 0);
+ for (i = 0; i < GF_FOP_MAXVALUE; i++)
+ GF_ATOMIC_INIT(stats->fop_hits[i], 0);
- for (i = 0; i < GF_UPCALL_FLAGS_MAXVALUE; i++)
- GF_ATOMIC_INIT (stats->upcall_hits[i], 0);
+ for (i = 0; i < GF_UPCALL_FLAGS_MAXVALUE; i++)
+ GF_ATOMIC_INIT(stats->upcall_hits[i], 0);
- gettimeofday (&stats->started_at, NULL);
+ stats->started_at = gf_time();
}
int
-init (xlator_t *this)
-{
- struct ios_conf *conf = NULL;
- char *sys_log_str = NULL;
- char *logger_str = NULL;
- char *log_format_str = NULL;
- int logger = -1;
- int log_format = -1;
- int sys_log_level = -1;
- char *log_str = NULL;
- int log_level = -1;
- int ret = -1;
- uint32_t log_buf_size = 0;
- uint32_t log_flush_timeout = 0;
-
- if (!this)
- return -1;
-
- if (!this->children) {
- gf_log (this->name, GF_LOG_ERROR,
- "io_stats translator requires atleast one subvolume");
- return -1;
- }
-
- if (!this->parents) {
- /* This is very much valid as io-stats currently is loaded
- * on top of volumes on both client and server, hence this is
- * not an warning message */
- gf_log (this->name, GF_LOG_DEBUG,
- "dangling volume. check volfile ");
- }
-
- conf = GF_CALLOC (1, sizeof(*conf), gf_io_stats_mt_ios_conf);
-
- if (!conf)
- goto out;
-
- if (dict_get_str (this->options, "unique-id", &conf->unique_id) != 0) {
- /* This is always set on servers, so we must be a client. */
- conf->unique_id = this->name;
- }
-
- /*
- * Init it just after calloc, so that we are sure the lock is inited
- * in case of error paths.
- */
- LOCK_INIT (&conf->lock);
- LOCK_INIT (&conf->ios_sampling_lock);
+init(xlator_t *this)
+{
+ struct ios_conf *conf = NULL;
+ char *volume_id = NULL;
+ char *sys_log_str = NULL;
+ char *logger_str = NULL;
+ char *log_format_str = NULL;
+ char *dump_format_str = NULL;
+ int logger = -1;
+ int log_format = -1;
+ int sys_log_level = -1;
+ char *log_str = NULL;
+ int log_level = -1;
+ int ret = -1;
+ uint32_t log_buf_size = 0;
+ uint32_t log_flush_timeout = 0;
+ int32_t threads;
+
+ if (!this)
+ return -1;
+
+ if (!this->children) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "io_stats translator requires at least one subvolume");
+ return -1;
+ }
+
+ if (!this->parents) {
+ /* This is very much valid as io-stats currently is loaded
+ * on top of volumes on both client and server, hence this is
+ * not an warning message */
+ gf_log(this->name, GF_LOG_DEBUG, "dangling volume. check volfile ");
+ }
+
+ conf = GF_CALLOC(1, sizeof(*conf), gf_io_stats_mt_ios_conf);
+
+ if (!conf)
+ goto out;
- ios_init_stats (&conf->cumulative);
- ios_init_stats (&conf->incremental);
+ if (dict_get_str(this->options, "unique-id", &conf->unique_id) != 0) {
+ /* This is always set on servers, so we must be a client. */
+ conf->unique_id = this->name;
+ }
+
+ ret = dict_get_strn(this->options, "volume-id", SLEN("volume-id"),
+ &volume_id);
+ if (!ret) {
+ strncpy(this->graph->volume_id, volume_id, GF_UUID_BUF_SIZE);
+ }
+ /*
+ * Init it just after calloc, so that we are sure the lock is inited
+ * in case of error paths.
+ */
+ LOCK_INIT(&conf->lock);
+ LOCK_INIT(&conf->ios_sampling_lock);
+
+ ios_init_stats(&conf->cumulative);
+ ios_init_stats(&conf->incremental);
+
+ ret = ios_init_top_stats(conf);
+ if (ret)
+ goto out;
- ret = ios_init_top_stats (conf);
- if (ret)
- goto out;
+ GF_OPTION_INIT("dump-fd-stats", conf->dump_fd_stats, bool, out);
- GF_OPTION_INIT ("dump-fd-stats", conf->dump_fd_stats, bool, out);
+ GF_OPTION_INIT("count-fop-hits", conf->count_fop_hits, bool, out);
- GF_OPTION_INIT ("count-fop-hits", conf->count_fop_hits, bool, out);
+ GF_OPTION_INIT("latency-measurement", conf->measure_latency, bool, out);
- GF_OPTION_INIT ("latency-measurement", conf->measure_latency,
- bool, out);
+ GF_OPTION_INIT("ios-dump-interval", conf->ios_dump_interval, int32, out);
- GF_OPTION_INIT ("ios-dump-interval", conf->ios_dump_interval,
- int32, out);
+ GF_OPTION_INIT("ios-sample-interval", conf->ios_sample_interval, int32,
+ out);
- GF_OPTION_INIT ("ios-sample-interval", conf->ios_sample_interval,
- int32, out);
+ GF_OPTION_INIT("ios-dump-format", dump_format_str, str, out);
+ ios_set_log_format_code(conf, dump_format_str);
- GF_OPTION_INIT ("ios-dump-format", conf->dump_format_str, str, out);
- ios_set_log_format_code (conf);
+ GF_OPTION_INIT("ios-sample-buf-size", conf->ios_sample_buf_size, int32,
+ out);
- GF_OPTION_INIT ("ios-sample-buf-size", conf->ios_sample_buf_size,
- int32, out);
+ ret = ios_init_sample_buf(conf);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR, "Out of memory.");
+ goto out;
+ }
- ret = ios_init_sample_buf (conf);
+ GF_OPTION_INIT("ios-dnscache-ttl-sec", conf->ios_dnscache_ttl_sec, int32,
+ out);
+ conf->dnscache = gf_dnscache_init(conf->ios_dnscache_ttl_sec);
+ if (!conf->dnscache) {
+ ret = -1;
+ goto out;
+ }
+
+ GF_OPTION_INIT("sys-log-level", sys_log_str, str, out);
+ if (sys_log_str) {
+ sys_log_level = glusterd_check_log_level(sys_log_str);
+ set_sys_log_level(sys_log_level);
+ }
+
+ GF_OPTION_INIT("log-level", log_str, str, out);
+ if (log_str) {
+ log_level = glusterd_check_log_level(log_str);
+ if (DEFAULT_LOG_LEVEL != log_level)
+ gf_log_set_loglevel(this->ctx, log_level);
+ }
+
+ GF_OPTION_INIT("logger", logger_str, str, out);
+ if (logger_str) {
+ logger = gf_check_logger(logger_str);
+ gf_log_set_logger(logger);
+ }
+
+ GF_OPTION_INIT("log-format", log_format_str, str, out);
+ if (log_format_str) {
+ log_format = gf_check_log_format(log_format_str);
+ gf_log_set_logformat(log_format);
+ }
+
+ GF_OPTION_INIT("log-buf-size", log_buf_size, uint32, out);
+ gf_log_set_log_buf_size(log_buf_size);
+
+ GF_OPTION_INIT("log-flush-timeout", log_flush_timeout, time, out);
+ gf_log_set_log_flush_timeout(log_flush_timeout);
+
+ GF_OPTION_INIT("threads", threads, int32, out);
+ gf_async_adjust_threads(threads);
+
+ this->private = conf;
+ if (conf->ios_dump_interval > 0) {
+ conf->dump_thread_running = _gf_true;
+ conf->dump_thread_should_die = _gf_false;
+ ret = gf_thread_create(&conf->dump_thread, NULL,
+ (void *)&_ios_dump_thread, this, "iosdump");
if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "Out of memory.");
- goto out;
- }
-
- GF_OPTION_INIT ("ios-dnscache-ttl-sec", conf->ios_dnscache_ttl_sec,
- int32, out);
- conf->dnscache = gf_dnscache_init (conf->ios_dnscache_ttl_sec);
-
- GF_OPTION_INIT ("sys-log-level", sys_log_str, str, out);
- if (sys_log_str) {
- sys_log_level = glusterd_check_log_level (sys_log_str);
- set_sys_log_level (sys_log_level);
- }
-
- GF_OPTION_INIT ("log-level", log_str, str, out);
- if (log_str) {
- log_level = glusterd_check_log_level (log_str);
- if (DEFAULT_LOG_LEVEL != log_level)
- gf_log_set_loglevel (this->ctx, log_level);
- }
-
- GF_OPTION_INIT ("logger", logger_str, str, out);
- if (logger_str) {
- logger = gf_check_logger (logger_str);
- gf_log_set_logger (logger);
- }
-
- GF_OPTION_INIT ("log-format", log_format_str, str, out);
- if (log_format_str) {
- log_format = gf_check_log_format (log_format_str);
- gf_log_set_logformat (log_format);
+ conf->dump_thread_running = _gf_false;
+ gf_log(this ? this->name : "io-stats", GF_LOG_ERROR,
+ "Failed to start thread"
+ "in init. Returning %d",
+ ret);
+ goto out;
}
-
- GF_OPTION_INIT ("log-buf-size", log_buf_size, uint32, out);
- gf_log_set_log_buf_size (log_buf_size);
-
- GF_OPTION_INIT ("log-flush-timeout", log_flush_timeout, time, out);
- gf_log_set_log_flush_timeout (log_flush_timeout);
-
- this->private = conf;
- if (conf->ios_dump_interval > 0) {
- ret = gf_thread_create (&conf->dump_thread, NULL,
- (void *) &_ios_dump_thread,
- this, "iosdump");
- if (ret) {
- gf_log (this ? this->name : "io-stats",
- GF_LOG_ERROR, "Failed to start thread"
- "in init. Returning %d", ret);
- goto out;
- }
- }
- return 0;
+ }
+ return 0;
out:
- ios_conf_destroy (conf);
- return ret;
+ ios_conf_destroy(conf);
+ return ret;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- struct ios_conf *conf = NULL;
+ struct ios_conf *conf = NULL;
- if (!this)
- return;
+ if (!this)
+ return;
- conf = this->private;
+ conf = this->private;
- ios_conf_destroy (conf);
- this->private = NULL;
- gf_log (this->name, GF_LOG_INFO,
- "io-stats translator unloaded");
- return;
+ ios_conf_destroy(conf);
+ this->private = NULL;
+ gf_log(this->name, GF_LOG_INFO, "io-stats translator unloaded");
+ return;
}
int
-notify (xlator_t *this, int32_t event, void *data, ...)
-{
- int ret = 0;
- struct ios_dump_args args = {0};
- dict_t *output = NULL;
- dict_t *dict = NULL;
- int32_t op = 0;
- int32_t list_cnt = 0;
- double throughput = 0;
- double time = 0;
- gf_boolean_t is_peek = _gf_false;
- va_list ap;
- struct gf_upcall *up_data = NULL;
- struct gf_upcall_cache_invalidation *up_ci = NULL;
-
- dict = data;
- va_start (ap, data);
- output = va_arg (ap, dict_t*);
- va_end (ap);
- switch (event) {
+notify(xlator_t *this, int32_t event, void *data, ...)
+{
+ int ret = 0;
+ struct ios_dump_args args = {0};
+ dict_t *output = NULL;
+ dict_t *dict = NULL;
+ int32_t op = 0;
+ int32_t list_cnt = 0;
+ double throughput = 0;
+ double time = 0;
+ gf_boolean_t is_peek = _gf_false;
+ va_list ap;
+ struct gf_upcall *up_data = NULL;
+ struct gf_upcall_cache_invalidation *up_ci = NULL;
+
+ dict = data;
+ va_start(ap, data);
+ output = va_arg(ap, dict_t *);
+ va_end(ap);
+ switch (event) {
case GF_EVENT_TRANSLATOR_INFO:
- ret = dict_get_str_boolean (dict, "clear-stats", _gf_false);
+ ret = dict_get_str_boolean(dict, "clear-stats", _gf_false);
+ if (ret) {
+ ret = dict_set_int32(output, "top-op", op);
if (ret) {
- ret = dict_set_int32 (output, "top-op", op);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to set top-op in dict");
- goto out;
- }
- ios_destroy_top_stats (this->private);
- ret = ios_init_top_stats (this->private);
+ gf_log(this->name, GF_LOG_ERROR,
+ "Failed to set top-op in dict");
+ goto out;
+ }
+ ios_destroy_top_stats(this->private);
+ ret = ios_init_top_stats(this->private);
+ if (ret)
+ gf_log(this->name, GF_LOG_ERROR,
+ "Failed to reset top stats");
+ ret = dict_set_int32(output, "stats-cleared", ret ? 0 : 1);
+ if (ret)
+ gf_log(this->name, GF_LOG_ERROR,
+ "Failed to set stats-cleared"
+ " in dict");
+ goto out;
+ }
+
+ ret = dict_get_int32(dict, "top-op", &op);
+ if (!ret) {
+ ret = dict_get_int32(dict, "list-cnt", &list_cnt);
+ if (op > IOS_STATS_TYPE_NONE && op < IOS_STATS_TYPE_MAX)
+ ret = io_stats_dump_stats_to_dict(this, output, op,
+ list_cnt);
+ if (op == IOS_STATS_TYPE_READ_THROUGHPUT ||
+ op == IOS_STATS_TYPE_WRITE_THROUGHPUT) {
+ ret = dict_get_double(dict, "throughput", &throughput);
+ if (!ret) {
+ ret = dict_get_double(dict, "time", &time);
+ if (ret)
+ goto out;
+ ret = dict_set_double(output, "throughput", throughput);
if (ret)
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to reset top stats");
- ret = dict_set_int32 (output, "stats-cleared",
- ret ? 0 : 1);
+ goto out;
+ ret = dict_set_double(output, "time", time);
if (ret)
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to set stats-cleared"
- " in dict");
- goto out;
+ goto out;
+ }
+ ret = 0;
}
+ } else {
+ ret = dict_get_int32(dict, "info-op", &op);
+ if (ret || op < GF_IOS_INFO_ALL || GF_IOS_INFO_CLEAR < op)
+ op = GF_IOS_INFO_ALL;
- ret = dict_get_int32 (dict, "top-op", &op);
- if (!ret) {
- ret = dict_get_int32 (dict, "list-cnt", &list_cnt);
- if (op > IOS_STATS_TYPE_NONE &&
- op < IOS_STATS_TYPE_MAX)
- ret = io_stats_dump_stats_to_dict (this, output,
- op, list_cnt);
- if (op == IOS_STATS_TYPE_READ_THROUGHPUT ||
- op == IOS_STATS_TYPE_WRITE_THROUGHPUT) {
- ret = dict_get_double (dict, "throughput",
- &throughput);
- if (!ret) {
- ret = dict_get_double (dict, "time",
- &time);
- if (ret)
- goto out;
- ret = dict_set_double (output,
- "throughput", throughput);
- if (ret)
- goto out;
- ret = dict_set_double (output, "time",
- time);
- if (ret)
- goto out;
- }
- ret = 0;
-
- }
+ ret = dict_set_int32(output, "info-op", op);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Failed to set info-op in dict");
+ goto out;
+ }
+
+ if (GF_IOS_INFO_CLEAR == op) {
+ io_stats_clear(this->private);
+
+ ret = dict_set_int32(output, "stats-cleared", 1);
+ if (ret)
+ gf_log(this->name, GF_LOG_ERROR,
+ "Failed to set stats-cleared"
+ " in dict");
} else {
- ret = dict_get_int32 (dict, "info-op", &op);
- if (ret || op < GF_CLI_INFO_ALL ||
- GF_CLI_INFO_CLEAR < op)
- op = GF_CLI_INFO_ALL;
-
- ret = dict_set_int32 (output, "info-op", op);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to set info-op in dict");
- goto out;
- }
-
- if (GF_CLI_INFO_CLEAR == op) {
- ret = io_stats_clear (this->private);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to clear info stats");
-
- ret = dict_set_int32 (output, "stats-cleared",
- ret ? 0 : 1);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to set stats-cleared"
- " in dict");
- }
- else {
- ret = dict_get_str_boolean (dict, "peek",
- _gf_false);
- if (-1 != ret)
- is_peek = ret;
-
- (void) ios_dump_args_init (&args,
- IOS_DUMP_TYPE_DICT, output);
- ret = io_stats_dump (this, &args, op, is_peek);
- }
+ ret = dict_get_str_boolean(dict, "peek", _gf_false);
+ if (-1 != ret)
+ is_peek = ret;
+
+ (void)ios_dump_args_init(&args, IOS_DUMP_TYPE_DICT, output);
+ ret = io_stats_dump(this, &args, op, is_peek);
}
- break;
+ }
+ break;
case GF_EVENT_UPCALL:
- up_data = (struct gf_upcall *)data;
- ios_bump_upcall (this, GF_UPCALL);
+ up_data = (struct gf_upcall *)data;
+ ios_bump_upcall(this, GF_UPCALL);
- switch (up_data->event_type) {
+ switch (up_data->event_type) {
case GF_UPCALL_RECALL_LEASE:
- ios_bump_upcall (this, GF_UPCALL_LEASE_RECALL);
- break;
+ ios_bump_upcall(this, GF_UPCALL_LEASE_RECALL);
+ break;
case GF_UPCALL_CACHE_INVALIDATION:
- up_ci = (struct gf_upcall_cache_invalidation *)up_data->data;
- if (up_ci->flags & (UP_XATTR | UP_XATTR_RM))
- ios_bump_upcall (this, GF_UPCALL_CI_XATTR);
- if (up_ci->flags & IATT_UPDATE_FLAGS)
- ios_bump_upcall (this, GF_UPCALL_CI_STAT);
- if (up_ci->flags & UP_RENAME_FLAGS)
- ios_bump_upcall (this, GF_UPCALL_CI_RENAME);
- if (up_ci->flags & UP_FORGET)
- ios_bump_upcall (this, GF_UPCALL_CI_FORGET);
- if (up_ci->flags & UP_NLINK)
- ios_bump_upcall (this, GF_UPCALL_CI_NLINK);
- break;
+ up_ci = (struct gf_upcall_cache_invalidation *)
+ up_data->data;
+ if (up_ci->flags & (UP_XATTR | UP_XATTR_RM))
+ ios_bump_upcall(this, GF_UPCALL_CI_XATTR);
+ if (up_ci->flags & IATT_UPDATE_FLAGS)
+ ios_bump_upcall(this, GF_UPCALL_CI_STAT);
+ if (up_ci->flags & UP_RENAME_FLAGS)
+ ios_bump_upcall(this, GF_UPCALL_CI_RENAME);
+ if (up_ci->flags & UP_FORGET)
+ ios_bump_upcall(this, GF_UPCALL_CI_FORGET);
+ if (up_ci->flags & UP_NLINK)
+ ios_bump_upcall(this, GF_UPCALL_CI_NLINK);
+ break;
default:
- gf_msg_debug (this->name, 0, "Unknown upcall event "
- "type :%d", up_data->event_type);
- break;
- }
+ gf_msg_debug(this->name, 0,
+ "Unknown upcall event "
+ "type :%d",
+ up_data->event_type);
+ break;
+ }
- default_notify (this, event, data);
- break;
+ default_notify(this, event, data);
+ break;
default:
- default_notify (this, event, data);
- break;
-
- }
+ default_notify(this, event, data);
+ break;
+ }
out:
- return ret;
+ return ret;
}
-struct xlator_dumpops dumpops = {
- .priv = io_priv
-};
+struct xlator_dumpops dumpops = {.priv = io_priv};
struct xlator_fops fops = {
- .stat = io_stats_stat,
- .readlink = io_stats_readlink,
- .mknod = io_stats_mknod,
- .mkdir = io_stats_mkdir,
- .unlink = io_stats_unlink,
- .rmdir = io_stats_rmdir,
- .symlink = io_stats_symlink,
- .rename = io_stats_rename,
- .link = io_stats_link,
- .truncate = io_stats_truncate,
- .open = io_stats_open,
- .readv = io_stats_readv,
- .writev = io_stats_writev,
- .statfs = io_stats_statfs,
- .flush = io_stats_flush,
- .fsync = io_stats_fsync,
- .setxattr = io_stats_setxattr,
- .getxattr = io_stats_getxattr,
- .removexattr = io_stats_removexattr,
- .fsetxattr = io_stats_fsetxattr,
- .fgetxattr = io_stats_fgetxattr,
- .fremovexattr = io_stats_fremovexattr,
- .opendir = io_stats_opendir,
- .readdir = io_stats_readdir,
- .readdirp = io_stats_readdirp,
- .fsyncdir = io_stats_fsyncdir,
- .access = io_stats_access,
- .ftruncate = io_stats_ftruncate,
- .fstat = io_stats_fstat,
- .create = io_stats_create,
- .lk = io_stats_lk,
- .inodelk = io_stats_inodelk,
- .finodelk = io_stats_finodelk,
- .entrylk = io_stats_entrylk,
- .fentrylk = io_stats_fentrylk,
- .lookup = io_stats_lookup,
- .xattrop = io_stats_xattrop,
- .fxattrop = io_stats_fxattrop,
- .setattr = io_stats_setattr,
- .fsetattr = io_stats_fsetattr,
- .fallocate = io_stats_fallocate,
- .discard = io_stats_discard,
- .zerofill = io_stats_zerofill,
- .ipc = io_stats_ipc,
- .rchecksum = io_stats_rchecksum,
- .seek = io_stats_seek,
- .lease = io_stats_lease,
- .getactivelk = io_stats_getactivelk,
- .setactivelk = io_stats_setactivelk,
- .compound = io_stats_compound,
+ .stat = io_stats_stat,
+ .readlink = io_stats_readlink,
+ .mknod = io_stats_mknod,
+ .mkdir = io_stats_mkdir,
+ .unlink = io_stats_unlink,
+ .rmdir = io_stats_rmdir,
+ .symlink = io_stats_symlink,
+ .rename = io_stats_rename,
+ .link = io_stats_link,
+ .truncate = io_stats_truncate,
+ .open = io_stats_open,
+ .readv = io_stats_readv,
+ .writev = io_stats_writev,
+ .statfs = io_stats_statfs,
+ .flush = io_stats_flush,
+ .fsync = io_stats_fsync,
+ .setxattr = io_stats_setxattr,
+ .getxattr = io_stats_getxattr,
+ .removexattr = io_stats_removexattr,
+ .fsetxattr = io_stats_fsetxattr,
+ .fgetxattr = io_stats_fgetxattr,
+ .fremovexattr = io_stats_fremovexattr,
+ .opendir = io_stats_opendir,
+ .readdir = io_stats_readdir,
+ .readdirp = io_stats_readdirp,
+ .fsyncdir = io_stats_fsyncdir,
+ .access = io_stats_access,
+ .ftruncate = io_stats_ftruncate,
+ .fstat = io_stats_fstat,
+ .create = io_stats_create,
+ .lk = io_stats_lk,
+ .inodelk = io_stats_inodelk,
+ .finodelk = io_stats_finodelk,
+ .entrylk = io_stats_entrylk,
+ .fentrylk = io_stats_fentrylk,
+ .lookup = io_stats_lookup,
+ .xattrop = io_stats_xattrop,
+ .fxattrop = io_stats_fxattrop,
+ .setattr = io_stats_setattr,
+ .fsetattr = io_stats_fsetattr,
+ .fallocate = io_stats_fallocate,
+ .discard = io_stats_discard,
+ .zerofill = io_stats_zerofill,
+ .ipc = io_stats_ipc,
+ .rchecksum = io_stats_rchecksum,
+ .seek = io_stats_seek,
+ .lease = io_stats_lease,
+ .getactivelk = io_stats_getactivelk,
+ .setactivelk = io_stats_setactivelk,
+ .compound = io_stats_compound,
+ .copy_file_range = io_stats_copy_file_range,
};
struct xlator_cbks cbks = {
- .release = io_stats_release,
- .releasedir = io_stats_releasedir,
- .forget = io_stats_forget,
+ .release = io_stats_release,
+ .releasedir = io_stats_releasedir,
+ .forget = io_stats_forget,
};
struct volume_options options[] = {
- { .key = {"dump-fd-stats"},
- .op_version = {1},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "If on stats related to file-operations would be "
- "tracked inside GlusterFS data-structures."
- },
- { .key = { "ios-dump-interval" },
- .type = GF_OPTION_TYPE_INT,
- .op_version = {1},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .min = 0,
- .max = 3600,
- .default_value = "0",
- .description = "Interval (in seconds) at which to auto-dump "
- "statistics. Zero disables automatic dumping."
- },
- { .key = { "ios-sample-interval" },
- .type = GF_OPTION_TYPE_INT,
- .op_version = {1},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .min = 0,
- .max = 65535,
- .default_value = "0",
- .description = "Interval in which we want to collect FOP latency "
- "samples. 2 means collect a sample every 2nd FOP."
- },
- { .key = {"ios-dump-format"},
- .type = GF_OPTION_TYPE_STR,
- .op_version = {GD_OP_VERSION_3_12_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .default_value = "json",
- .description = " The dump-format option specifies the format in which"
- " to dump the statistics. Select between \"text\", "
- "\"json\", \"dict\" and \"samples\". Default is "
- "\"json\".",
- .value = { "text", "json", "dict", "samples"}
- },
- { .key = { "ios-sample-buf-size" },
- .type = GF_OPTION_TYPE_INT,
- .op_version = {1},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .min = 1024,
- .max = 1024*1024,
- .default_value = "65535",
- .description = "The maximum size of our FOP sampling ring buffer."
- },
- { .key = { "ios-dnscache-ttl-sec" },
- .type = GF_OPTION_TYPE_INT,
- .op_version = {1},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .min = 1,
- .max = 3600 * 72,
- .default_value = "86400",
- .description = "The interval after wish a cached DNS entry will be "
- "re-validated. Default: 24 hrs"
- },
- { .key = { "latency-measurement" },
- .type = GF_OPTION_TYPE_BOOL,
- .op_version = {1},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .default_value = "off",
- .description = "If on stats related to the latency of each operation "
- "would be tracked inside GlusterFS data-structures. "
- },
- { .key = {"count-fop-hits"},
- .type = GF_OPTION_TYPE_BOOL,
- .op_version = {1},
- .flags = OPT_FLAG_SETTABLE,
- .tags = {"io-stats"},
- },
- { .key = {"log-level"},
- .type = GF_OPTION_TYPE_STR,
- .value = { "DEBUG", "WARNING", "ERROR", "INFO",
- "CRITICAL", "NONE", "TRACE"}
- },
-
- /* These are synthetic entries to assist validation of CLI's *
- * volume set command */
- { .key = {"client-log-level"},
- .type = GF_OPTION_TYPE_STR,
- .op_version = {1},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .default_value = "INFO",
- .description = "Changes the log-level of the clients",
- .value = { "DEBUG", "WARNING", "ERROR", "INFO",
- "CRITICAL", "NONE", "TRACE"}
- },
- { .key = {"sys-log-level"},
- .type = GF_OPTION_TYPE_STR,
- .default_value = "CRITICAL",
- .description = "Gluster's syslog log-level",
- .value = { "WARNING", "ERROR", "INFO", "CRITICAL"}
- },
- { .key = {"brick-log-level"},
- .type = GF_OPTION_TYPE_STR,
- .op_version = {1},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .default_value = "INFO",
- .description = "Changes the log-level of the bricks",
- .value = { "DEBUG", "WARNING", "ERROR", "INFO",
- "CRITICAL", "NONE", "TRACE"}
- },
- { .key = {"logger"},
- .type = GF_OPTION_TYPE_STR,
- .value = { GF_LOGGER_GLUSTER_LOG, GF_LOGGER_SYSLOG}
- },
- { .key = {"client-logger"},
- .type = GF_OPTION_TYPE_STR,
- .op_version = {GD_OP_VERSION_3_6_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .default_value = GF_LOGGER_GLUSTER_LOG,
- .description = "Changes the logging sub-system to log to, for the "
- "clients",
- .value = { GF_LOGGER_GLUSTER_LOG, GF_LOGGER_SYSLOG}
- },
- { .key = {"brick-logger"},
- .type = GF_OPTION_TYPE_STR,
- .op_version = {GD_OP_VERSION_3_6_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .default_value = GF_LOGGER_GLUSTER_LOG,
- .description = "Changes the logging sub-system to log to, for the "
- "bricks",
- .value = { GF_LOGGER_GLUSTER_LOG, GF_LOGGER_SYSLOG}
- },
- { .key = {"log-format"},
- .type = GF_OPTION_TYPE_STR,
- .value = { GF_LOG_FORMAT_NO_MSG_ID, GF_LOG_FORMAT_WITH_MSG_ID}
- },
- { .key = {"client-log-format"},
- .type = GF_OPTION_TYPE_STR,
- .op_version = {GD_OP_VERSION_3_6_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .default_value = GF_LOG_FORMAT_WITH_MSG_ID,
- .description = "Changes log format for the clients",
- .value = { GF_LOG_FORMAT_NO_MSG_ID, GF_LOG_FORMAT_WITH_MSG_ID}
- },
- { .key = {"brick-log-format"},
- .type = GF_OPTION_TYPE_STR,
- .op_version = {GD_OP_VERSION_3_6_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .default_value = GF_LOG_FORMAT_WITH_MSG_ID,
- .description = "Changes the log format for the bricks",
- .value = { GF_LOG_FORMAT_NO_MSG_ID, GF_LOG_FORMAT_WITH_MSG_ID}
- },
- { .key = {"log-buf-size"},
- .type = GF_OPTION_TYPE_INT,
- .min = GF_LOG_LRU_BUFSIZE_MIN,
- .max = GF_LOG_LRU_BUFSIZE_MAX,
- .default_value = "5",
- },
- { .key = {"client-log-buf-size"},
- .type = GF_OPTION_TYPE_INT,
- .op_version = {GD_OP_VERSION_3_6_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .min = GF_LOG_LRU_BUFSIZE_MIN,
- .max = GF_LOG_LRU_BUFSIZE_MAX,
- .default_value = "5",
- .description = "This option determines the maximum number of unique "
- "log messages that can be buffered for a time equal to"
- " the value of the option client-log-flush-timeout."
- },
- { .key = {"brick-log-buf-size"},
- .type = GF_OPTION_TYPE_INT,
- .op_version = {GD_OP_VERSION_3_6_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .min = GF_LOG_LRU_BUFSIZE_MIN,
- .max = GF_LOG_LRU_BUFSIZE_MAX,
- .default_value = "5",
- .description = "This option determines the maximum number of unique "
- "log messages that can be buffered for a time equal to"
- " the value of the option brick-log-flush-timeout."
- },
- { .key = {"log-flush-timeout"},
- .type = GF_OPTION_TYPE_TIME,
- .min = GF_LOG_FLUSH_TIMEOUT_MIN,
- .max = GF_LOG_FLUSH_TIMEOUT_MAX,
- .default_value = "120",
- },
- { .key = {"client-log-flush-timeout"},
- .type = GF_OPTION_TYPE_TIME,
- .op_version = {GD_OP_VERSION_3_6_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .min = GF_LOG_FLUSH_TIMEOUT_MIN,
- .max = GF_LOG_FLUSH_TIMEOUT_MAX,
- .default_value = "120",
- .description = "This option determines the maximum number of unique "
- "log messages that can be buffered for a time equal to"
- " the value of the option client-log-flush-timeout."
- },
- { .key = {"brick-log-flush-timeout"},
- .type = GF_OPTION_TYPE_TIME,
- .op_version = {GD_OP_VERSION_3_6_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"io-stats"},
- .min = GF_LOG_FLUSH_TIMEOUT_MIN,
- .max = GF_LOG_FLUSH_TIMEOUT_MAX,
- .default_value = "120",
- .description = "This option determines the maximum number of unique "
- "log messages that can be buffered for a time equal to"
- " the value of the option brick-log-flush-timeout."
- },
- { .key = {"unique-id"},
- .type = GF_OPTION_TYPE_STR,
- .default_value = "/no/such/path",
- .description = "Unique ID for our files."
- },
- { .key = {NULL} },
+ {.key = {"dump-fd-stats"},
+ .op_version = {1},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "If on stats related to file-operations would be "
+ "tracked inside GlusterFS data-structures."},
+ {.key = {"ios-dump-interval"},
+ .type = GF_OPTION_TYPE_INT,
+ .op_version = {1},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .min = 0,
+ .max = 3600,
+ .default_value = "0",
+ .description = "Interval (in seconds) at which to auto-dump "
+ "statistics. Zero disables automatic dumping."},
+ {.key = {"ios-sample-interval"},
+ .type = GF_OPTION_TYPE_INT,
+ .op_version = {1},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .min = 0,
+ .max = 65535,
+ .default_value = "0",
+ .description = "Interval in which we want to collect FOP latency "
+ "samples. 2 means collect a sample every 2nd FOP."},
+ {.key = {"ios-dump-format"},
+ .type = GF_OPTION_TYPE_STR,
+ .op_version = {GD_OP_VERSION_3_12_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .default_value = "json",
+ .description = " The dump-format option specifies the format in which"
+ " to dump the statistics. Select between \"text\", "
+ "\"json\", \"dict\" and \"samples\". Default is "
+ "\"json\".",
+ .value = {"text", "json", "dict", "samples"}},
+ {.key = {"ios-sample-buf-size"},
+ .type = GF_OPTION_TYPE_INT,
+ .op_version = {1},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .min = 1024,
+ .max = 1024 * 1024,
+ .default_value = "65535",
+ .description = "The maximum size of our FOP sampling ring buffer."},
+ {.key = {"ios-dnscache-ttl-sec"},
+ .type = GF_OPTION_TYPE_INT,
+ .op_version = {1},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .min = 1,
+ .max = 3600 * 72,
+ .default_value = "86400",
+ .description = "The interval after wish a cached DNS entry will be "
+ "re-validated. Default: 24 hrs"},
+ {.key = {"latency-measurement"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .op_version = {1},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .default_value = "off",
+ .description = "If on stats related to the latency of each operation "
+ "would be tracked inside GlusterFS data-structures. "},
+ {
+ .key = {"count-fop-hits"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .op_version = {1},
+ .flags = OPT_FLAG_SETTABLE,
+ .tags = {"io-stats"},
+ },
+ {.key = {"log-level"},
+ .type = GF_OPTION_TYPE_STR,
+ .value = {"DEBUG", "WARNING", "ERROR", "INFO", "CRITICAL", "NONE",
+ "TRACE"}},
+
+ /* These are synthetic entries to assist validation of CLI's *
+ * volume set command */
+ {.key = {"client-log-level"},
+ .type = GF_OPTION_TYPE_STR,
+ .op_version = {1},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .default_value = "INFO",
+ .description = "Changes the log-level of the clients",
+ .value = {"DEBUG", "WARNING", "ERROR", "INFO", "CRITICAL", "NONE",
+ "TRACE"}},
+ {.key = {"sys-log-level"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = "CRITICAL",
+ .description = "Gluster's syslog log-level",
+ .value = {"WARNING", "ERROR", "INFO", "CRITICAL"}},
+ {.key = {"brick-log-level"},
+ .type = GF_OPTION_TYPE_STR,
+ .op_version = {1},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .default_value = "INFO",
+ .description = "Changes the log-level of the bricks",
+ .value = {"DEBUG", "WARNING", "ERROR", "INFO", "CRITICAL", "NONE",
+ "TRACE"}},
+ {.key = {"logger"},
+ .type = GF_OPTION_TYPE_STR,
+ .value = {GF_LOGGER_GLUSTER_LOG, GF_LOGGER_SYSLOG}},
+ {.key = {"client-logger"},
+ .type = GF_OPTION_TYPE_STR,
+ .op_version = {GD_OP_VERSION_3_6_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .default_value = GF_LOGGER_GLUSTER_LOG,
+ .description = "Changes the logging sub-system to log to, for the "
+ "clients",
+ .value = {GF_LOGGER_GLUSTER_LOG, GF_LOGGER_SYSLOG}},
+ {.key = {"brick-logger"},
+ .type = GF_OPTION_TYPE_STR,
+ .op_version = {GD_OP_VERSION_3_6_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .default_value = GF_LOGGER_GLUSTER_LOG,
+ .description = "Changes the logging sub-system to log to, for the "
+ "bricks",
+ .value = {GF_LOGGER_GLUSTER_LOG, GF_LOGGER_SYSLOG}},
+ {.key = {"log-format"},
+ .type = GF_OPTION_TYPE_STR,
+ .value = {GF_LOG_FORMAT_NO_MSG_ID, GF_LOG_FORMAT_WITH_MSG_ID}},
+ {.key = {"client-log-format"},
+ .type = GF_OPTION_TYPE_STR,
+ .op_version = {GD_OP_VERSION_3_6_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .default_value = GF_LOG_FORMAT_WITH_MSG_ID,
+ .description = "Changes log format for the clients",
+ .value = {GF_LOG_FORMAT_NO_MSG_ID, GF_LOG_FORMAT_WITH_MSG_ID}},
+ {.key = {"brick-log-format"},
+ .type = GF_OPTION_TYPE_STR,
+ .op_version = {GD_OP_VERSION_3_6_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .default_value = GF_LOG_FORMAT_WITH_MSG_ID,
+ .description = "Changes the log format for the bricks",
+ .value = {GF_LOG_FORMAT_NO_MSG_ID, GF_LOG_FORMAT_WITH_MSG_ID}},
+ {
+ .key = {"log-buf-size"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = GF_LOG_LRU_BUFSIZE_MIN,
+ .max = GF_LOG_LRU_BUFSIZE_MAX,
+ .default_value = "5",
+ },
+ {.key = {"client-log-buf-size"},
+ .type = GF_OPTION_TYPE_INT,
+ .op_version = {GD_OP_VERSION_3_6_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .min = GF_LOG_LRU_BUFSIZE_MIN,
+ .max = GF_LOG_LRU_BUFSIZE_MAX,
+ .default_value = "5",
+ .description = "This option determines the maximum number of unique "
+ "log messages that can be buffered for a time equal to"
+ " the value of the option client-log-flush-timeout."},
+ {.key = {"brick-log-buf-size"},
+ .type = GF_OPTION_TYPE_INT,
+ .op_version = {GD_OP_VERSION_3_6_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .min = GF_LOG_LRU_BUFSIZE_MIN,
+ .max = GF_LOG_LRU_BUFSIZE_MAX,
+ .default_value = "5",
+ .description = "This option determines the maximum number of unique "
+ "log messages that can be buffered for a time equal to"
+ " the value of the option brick-log-flush-timeout."},
+ {
+ .key = {"log-flush-timeout"},
+ .type = GF_OPTION_TYPE_TIME,
+ .min = GF_LOG_FLUSH_TIMEOUT_MIN,
+ .max = GF_LOG_FLUSH_TIMEOUT_MAX,
+ .default_value = "120",
+ },
+ {.key = {"client-log-flush-timeout"},
+ .type = GF_OPTION_TYPE_TIME,
+ .op_version = {GD_OP_VERSION_3_6_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .min = GF_LOG_FLUSH_TIMEOUT_MIN,
+ .max = GF_LOG_FLUSH_TIMEOUT_MAX,
+ .default_value = "120",
+ .description = "This option determines the maximum number of unique "
+ "log messages that can be buffered for a time equal to"
+ " the value of the option client-log-flush-timeout."},
+ {.key = {"brick-log-flush-timeout"},
+ .type = GF_OPTION_TYPE_TIME,
+ .op_version = {GD_OP_VERSION_3_6_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"io-stats"},
+ .min = GF_LOG_FLUSH_TIMEOUT_MIN,
+ .max = GF_LOG_FLUSH_TIMEOUT_MAX,
+ .default_value = "120",
+ .description = "This option determines the maximum number of unique "
+ "log messages that can be buffered for a time equal to"
+ " the value of the option brick-log-flush-timeout."},
+ {.key = {"unique-id"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = "/no/such/path",
+ .description = "Unique ID for our files."},
+ {.key = {"global-threading"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .op_version = {GD_OP_VERSION_6_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .tags = {"io-stats", "threading"},
+ .description = "This option enables the global threading support for "
+ "bricks. If enabled, it's recommended to also enable "
+ "'performance.iot-pass-through'"},
+ {.key = {"threads"}, .type = GF_OPTION_TYPE_INT},
+ {.key = {"brick-threads"},
+ .type = GF_OPTION_TYPE_INT,
+ .default_value = "16",
+ .min = 0,
+ .max = GF_ASYNC_MAX_THREADS,
+ .op_version = {GD_OP_VERSION_6_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"io-stats", "threading"},
+ .description = "When global threading is used, this value determines the "
+ "maximum amount of threads that can be created on bricks"},
+ {.key = {"client-threads"},
+ .type = GF_OPTION_TYPE_INT,
+ .default_value = "16",
+ .min = 0,
+ .max = GF_ASYNC_MAX_THREADS,
+ .op_version = {GD_OP_VERSION_6_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC | OPT_FLAG_CLIENT_OPT,
+ .tags = {"io-stats", "threading"},
+ .description = "When global threading is used, this value determines the "
+ "maximum amount of threads that can be created on clients"},
+ {.key = {"volume-id"},
+ .type = GF_OPTION_TYPE_STR,
+ .op_version = {GD_OP_VERSION_7_1},
+ .tags = {"global", "volume-id"},
+ .description =
+ "This option points to the 'unique' UUID particular to this "
+ "volume, which would be set in 'graph->volume_id'"},
+ {.key = {NULL}},
+};
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .notify = notify,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .dumpops = &dumpops,
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "io-stats",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/debug/sink/src/sink.c b/xlators/debug/sink/src/sink.c
index dfd5685a969..9822bbb732e 100644
--- a/xlators/debug/sink/src/sink.c
+++ b/xlators/debug/sink/src/sink.c
@@ -8,41 +8,41 @@
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "defaults.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
int32_t
-init (xlator_t *this)
+init(xlator_t *this)
{
- return 0;
+ return 0;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- return;
+ return;
}
/*
* notify - when parent sends PARENT_UP, send CHILD_UP event from here
*/
int32_t
-notify (xlator_t *this, int32_t event, void *data, ...)
+notify(xlator_t *this, int32_t event, void *data, ...)
{
- switch (event) {
+ switch (event) {
case GF_EVENT_PARENT_UP:
- /* Tell the parent that this xlator is up */
- default_notify (this, GF_EVENT_CHILD_UP, data);
- break;
+ /* Tell the parent that this xlator is up */
+ default_notify(this, GF_EVENT_CHILD_UP, data);
+ break;
case GF_EVENT_PARENT_DOWN:
- /* Tell the parent that this xlator is down */
- default_notify (this, GF_EVENT_CHILD_DOWN, data);
- break;
+ /* Tell the parent that this xlator is down */
+ default_notify(this, GF_EVENT_CHILD_DOWN, data);
+ break;
default:
- break;
- }
+ break;
+ }
- return 0;
+ return 0;
}
/*
@@ -53,27 +53,42 @@ notify (xlator_t *this, int32_t event, void *data, ...)
* extended to support different LOOKUPs too.
*/
static int32_t
-sink_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+sink_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- struct iatt stbuf = { 0, };
- struct iatt postparent = { 0, };
+ struct iatt stbuf = {
+ 0,
+ };
+ struct iatt postparent = {
+ 0,
+ };
- /* the root of the volume always need to be a directory */
- stbuf.ia_type = IA_IFDIR;
+ /* the root of the volume always need to be a directory */
+ stbuf.ia_type = IA_IFDIR;
- STACK_UNWIND_STRICT (lookup, frame, 0, 0, loc ? loc->inode : NULL,
- &stbuf, xdata, &postparent);
+ STACK_UNWIND_STRICT(lookup, frame, 0, 0, loc ? loc->inode : NULL, &stbuf,
+ xdata, &postparent);
- return 0;
+ return 0;
}
struct xlator_fops fops = {
- .lookup = sink_lookup,
+ .lookup = sink_lookup,
};
-struct xlator_cbks cbks = {
-};
+struct xlator_cbks cbks = {};
struct volume_options options[] = {
- { .key = {NULL} },
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .notify = notify,
+ .op_version = {GD_OP_VERSION_3_12_0},
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "sink",
+ .category = GF_TECH_PREVIEW,
};
diff --git a/xlators/debug/trace/src/trace-mem-types.h b/xlators/debug/trace/src/trace-mem-types.h
index 9fa7d97c2ca..18a7e0414a6 100644
--- a/xlators/debug/trace/src/trace-mem-types.h
+++ b/xlators/debug/trace/src/trace-mem-types.h
@@ -8,14 +8,13 @@
cases as published by the Free Software Foundation.
*/
-
#ifndef __TRACE_MEM_TYPES_H__
#define __TRACE_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_trace_mem_types_ {
- gf_trace_mt_trace_conf_t = gf_common_mt_end + 1,
- gf_trace_mt_end
+ gf_trace_mt_trace_conf_t = gf_common_mt_end + 1,
+ gf_trace_mt_end
};
#endif
diff --git a/xlators/debug/trace/src/trace.c b/xlators/debug/trace/src/trace.c
index be2f84d936f..6ed0ca00342 100644
--- a/xlators/debug/trace/src/trace.c
+++ b/xlators/debug/trace/src/trace.c
@@ -17,3351 +17,3518 @@
* their _cbk functions, which later passes the call to next layer.
* Very helpful translator for debugging.
*/
-#define TRACE_STAT_TO_STR(buf, str) trace_stat_to_str (buf, str, sizeof (str))
+#define TRACE_STAT_TO_STR(buf, str) trace_stat_to_str(buf, str, sizeof(str))
static void
trace_stat_to_str(struct iatt *buf, char *str, size_t len)
{
- char atime_buf[256] = {0,};
- char mtime_buf[256] = {0,};
- char ctime_buf[256] = {0,};
-
- if (!buf)
- return;
+ char atime_buf[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ char mtime_buf[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ char ctime_buf[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+
+ if (!buf)
+ return;
- gf_time_fmt (atime_buf, sizeof atime_buf, buf->ia_atime,
- gf_timefmt_dirent);
+ gf_time_fmt(atime_buf, sizeof atime_buf, buf->ia_atime, gf_timefmt_dirent);
- gf_time_fmt (mtime_buf, sizeof mtime_buf, buf->ia_mtime,
- gf_timefmt_dirent);
+ gf_time_fmt(mtime_buf, sizeof mtime_buf, buf->ia_mtime, gf_timefmt_dirent);
- gf_time_fmt (ctime_buf, sizeof ctime_buf, buf->ia_ctime,
- gf_timefmt_dirent);
+ gf_time_fmt(ctime_buf, sizeof ctime_buf, buf->ia_ctime, gf_timefmt_dirent);
- snprintf (str, len, "gfid=%s ino=%"PRIu64", mode=%o, "
- "nlink=%"GF_PRI_NLINK", uid=%u, gid=%u, size=%"PRIu64", "
- "blocks=%"PRIu64", atime=%s mtime=%s ctime=%s "
- "atime_sec=%"PRId64", atime_nsec=%"PRIu32","
- " mtime_sec=%"PRId64", mtime_nsec=%"PRIu32", "
- "ctime_sec=%"PRId64", ctime_nsec=%"PRIu32"",
- uuid_utoa (buf->ia_gfid), buf->ia_ino,
- st_mode_from_ia (buf->ia_prot, buf->ia_type), buf->ia_nlink,
- buf->ia_uid, buf->ia_gid, buf->ia_size, buf->ia_blocks,
- atime_buf, mtime_buf, ctime_buf,
- buf->ia_atime, buf->ia_atime_nsec,
- buf->ia_mtime, buf->ia_mtime_nsec,
- buf->ia_ctime, buf->ia_ctime_nsec);
+ snprintf(str, len,
+ "gfid=%s ino=%" PRIu64
+ ", mode=%o, "
+ "nlink=%" GF_PRI_NLINK ", uid=%u, gid=%u, size=%" PRIu64
+ ", "
+ "blocks=%" PRIu64
+ ", atime=%s mtime=%s ctime=%s "
+ "atime_sec=%" PRId64 ", atime_nsec=%" PRIu32
+ ","
+ " mtime_sec=%" PRId64 ", mtime_nsec=%" PRIu32
+ ", "
+ "ctime_sec=%" PRId64 ", ctime_nsec=%" PRIu32 "",
+ uuid_utoa(buf->ia_gfid), buf->ia_ino,
+ st_mode_from_ia(buf->ia_prot, buf->ia_type), buf->ia_nlink,
+ buf->ia_uid, buf->ia_gid, buf->ia_size, buf->ia_blocks, atime_buf,
+ mtime_buf, ctime_buf, buf->ia_atime, buf->ia_atime_nsec,
+ buf->ia_mtime, buf->ia_mtime_nsec, buf->ia_ctime,
+ buf->ia_ctime_nsec);
}
-
int
-dump_history_trace (circular_buffer_t *cb, void *data)
+dump_history_trace(circular_buffer_t *cb, void *data)
{
- char timestr[256] = {0,};
+ char timestr[GF_TIMESTR_SIZE] = {
+ 0,
+ };
- /* Since we are continuing with adding entries to the buffer even when
- gettimeofday () fails, it's safe to check tm and then dump the time
- at which the entry was added to the buffer */
+ /* Since we are continuing with adding entries to the buffer even when
+ gettimeofday () fails, it's safe to check tm and then dump the time
+ at which the entry was added to the buffer */
- gf_time_fmt (timestr, sizeof timestr, cb->tv.tv_sec, gf_timefmt_Ymd_T);
- snprintf (timestr + strlen (timestr), 256 - strlen (timestr),
- ".%"GF_PRI_SUSECONDS, cb->tv.tv_usec);
- gf_proc_dump_write ("TIME", "%s", timestr);
+ gf_time_fmt_tv(timestr, sizeof timestr, &cb->tv, gf_timefmt_Ymd_T);
+ gf_proc_dump_write("TIME", "%s", timestr);
- gf_proc_dump_write ("FOP", "%s\n", cb->data);
+ gf_proc_dump_write("FOP", "%s\n", (char *)cb->data);
- return 0;
+ return 0;
}
int
-trace_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd,
- inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+trace_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- char statstr[4096] = {0, };
- char preparentstr[4096] = {0, };
- char postparentstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_CREATE].enabled) {
- char string[4096] = {0,};
- if (op_ret >= 0) {
- TRACE_STAT_TO_STR (buf, statstr);
- TRACE_STAT_TO_STR (preparent, preparentstr);
- TRACE_STAT_TO_STR (postparent, postparentstr);
-
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s (op_ret=%d, fd=%p"
- "*stbuf {%s}, *preparent {%s}, "
- "*postparent = {%s})",
- frame->root->unique,
- uuid_utoa (inode->gfid), op_ret, fd,
- statstr, preparentstr, postparentstr);
-
- /* for 'release' log */
- fd_ctx_set (fd, this, 0);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": (op_ret=%d, op_errno=%d)",
- frame->root->unique, op_ret,
- op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (create, frame, op_ret, op_errno, fd, inode, buf,
- preparent, postparent, xdata);
- return 0;
-}
+ char statstr[1024] = {
+ 0,
+ };
+ char preparentstr[1024] = {
+ 0,
+ };
+ char postparentstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
-int
-trace_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
-{
- trace_conf_t *conf = NULL;
+ conf = this->private;
- conf = this->private;
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_CREATE].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret >= 0) {
+ TRACE_STAT_TO_STR(buf, statstr);
+ TRACE_STAT_TO_STR(preparent, preparentstr);
+ TRACE_STAT_TO_STR(postparent, postparentstr);
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_OPEN].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, op_errno=%d, "
- "*fd=%p", frame->root->unique,
- uuid_utoa (frame->local), op_ret, op_errno,
- fd);
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s (op_ret=%d, fd=%p"
+ "*stbuf {%s}, *preparent {%s}, "
+ "*postparent = {%s})",
+ frame->root->unique, uuid_utoa(inode->gfid), op_ret, fd,
+ statstr, preparentstr, postparentstr);
- LOG_ELEMENT (conf, string);
+ /* for 'release' log */
+ fd_ctx_set(fd, this, 0);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": (op_ret=%d, op_errno=%d)",
+ frame->root->unique, op_ret, op_errno);
}
-
+ LOG_ELEMENT(conf, string);
+ }
out:
- /* for 'release' log */
- if (op_ret >= 0)
- fd_ctx_set (fd, this, 0);
-
- TRACE_STACK_UNWIND (open, frame, op_ret, op_errno, fd, xdata);
- return 0;
+ TRACE_STACK_UNWIND(create, frame, op_ret, op_errno, fd, inode, buf,
+ preparent, postparent, xdata);
+ return 0;
}
int
-trace_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- dict_t *xdata)
+trace_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- char statstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_STAT].enabled) {
- char string[4096] = {0,};
- if (op_ret == 0) {
- TRACE_STAT_TO_STR (buf, statstr);
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d buf=%s",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret,
- statstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "op_errno=%d)",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret,
- op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (stat, frame, op_ret, op_errno, buf, xdata);
- return 0;
-}
+ trace_conf_t *conf = NULL;
-int
-trace_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iovec *vector,
- int32_t count, struct iatt *buf, struct iobref *iobref,
- dict_t *xdata)
-{
- char statstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_READ].enabled) {
- char string[4096] = {0,};
- if (op_ret >= 0) {
- TRACE_STAT_TO_STR (buf, statstr);
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d buf=%s",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret,
- statstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "op_errno=%d)",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret,
- op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (readv, frame, op_ret, op_errno, vector, count,
- buf, iobref, xdata);
- return 0;
-}
-
-int
-trace_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
-{
- char preopstr[4096] = {0, };
- char postopstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_WRITE].enabled) {
- char string[4096] = {0,};
- if (op_ret >= 0) {
- TRACE_STAT_TO_STR (prebuf, preopstr);
- TRACE_STAT_TO_STR (postbuf, postopstr);
-
- snprintf (string, sizeof (string),
- "%"PRId64": (op_ret=%d, "
- "*prebuf = {%s}, *postbuf = {%s})",
- frame->root->unique, op_ret,
- preopstr, postopstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "op_errno=%d", frame->root->unique,
- uuid_utoa (frame->local), op_ret,
- op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (writev, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
- return 0;
-}
-
-int
-trace_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *buf,
- dict_t *xdata)
-{
- trace_conf_t *conf = NULL;
+ conf = this->private;
- conf = this->private;
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_OPEN].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, op_errno=%d, "
+ "*fd=%p",
+ frame->root->unique, uuid_utoa(frame->local), op_ret, op_errno,
+ fd);
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_READDIR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64" : gfid=%s op_ret=%d, op_errno=%d",
- frame->root->unique, uuid_utoa (frame->local),
- op_ret, op_errno);
+ LOG_ELEMENT(conf, string);
+ }
- LOG_ELEMENT (conf, string);
- }
out:
- TRACE_STACK_UNWIND (readdir, frame, op_ret, op_errno, buf, xdata);
+ /* for 'release' log */
+ if (op_ret >= 0)
+ fd_ctx_set(fd, this, 0);
- return 0;
+ TRACE_STACK_UNWIND(open, frame, op_ret, op_errno, fd, xdata);
+ return 0;
}
int
-trace_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *buf,
- dict_t *xdata)
+trace_stat_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
{
- int count = 0;
- char statstr[4096] = {0,};
- char string[4096] = {0,};
- trace_conf_t *conf = NULL;
- gf_dirent_t *entry = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_READDIRP].enabled) {
- snprintf (string, sizeof (string),
- "%"PRId64" : gfid=%s op_ret=%d, op_errno=%d",
- frame->root->unique, uuid_utoa (frame->local),
- op_ret, op_errno);
-
- LOG_ELEMENT (conf, string);
- }
- if (op_ret < 0)
- goto out;
-
- list_for_each_entry (entry, &buf->list, list) {
- count++;
- TRACE_STAT_TO_STR (&entry->d_stat, statstr);
- snprintf (string, sizeof (string), "entry no. %d, pargfid=%s, "
- "bname=%s *buf {%s}", count, uuid_utoa (frame->local),
- entry->d_name, statstr);
- LOG_ELEMENT (conf, string);
- }
-
-out:
- TRACE_STACK_UNWIND (readdirp, frame, op_ret, op_errno, buf, xdata);
- return 0;
-}
-
-int
-trace_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+ char statstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_STAT].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret == 0) {
+ TRACE_STAT_TO_STR(buf, statstr);
+ (void)snprintf(
+ string, sizeof(string), "%" PRId64 ": gfid=%s op_ret=%d buf=%s",
+ frame->root->unique, uuid_utoa(frame->local), op_ret, statstr);
+ } else {
+ (void)snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "op_errno=%d)",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(stat, frame, op_ret, op_errno, buf, xdata);
+ return 0;
+}
+
+int
+trace_readv_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iovec *vector,
+ int32_t count, struct iatt *buf, struct iobref *iobref,
+ dict_t *xdata)
{
- char preopstr[4096] = {0, };
- char postopstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FSYNC].enabled) {
- char string[4096] = {0,};
- if (op_ret == 0) {
- TRACE_STAT_TO_STR (prebuf, preopstr);
- TRACE_STAT_TO_STR (postbuf, postopstr);
-
- snprintf (string, sizeof (string),
- "%"PRId64": (op_ret=%d, "
- "*prebuf = {%s}, *postbuf = {%s}",
- frame->root->unique, op_ret,
- preopstr, postopstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "op_errno=%d", frame->root->unique,
- uuid_utoa (frame->local), op_ret,
- op_errno);
-
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (fsync, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
-
- return 0;
-}
-
-int
-trace_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *statpre, struct iatt *statpost, dict_t *xdata)
+ char statstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_READ].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret >= 0) {
+ TRACE_STAT_TO_STR(buf, statstr);
+ snprintf(
+ string, sizeof(string), "%" PRId64 ": gfid=%s op_ret=%d buf=%s",
+ frame->root->unique, uuid_utoa(frame->local), op_ret, statstr);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "op_errno=%d)",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(readv, frame, op_ret, op_errno, vector, count, buf,
+ iobref, xdata);
+ return 0;
+}
+
+int
+trace_writev_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ char preopstr[1024] = {
+ 0,
+ };
+ char postopstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_WRITE].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret >= 0) {
+ TRACE_STAT_TO_STR(prebuf, preopstr);
+ TRACE_STAT_TO_STR(postbuf, postopstr);
+
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": (op_ret=%d, "
+ "*prebuf = {%s}, *postbuf = {%s})",
+ frame->root->unique, op_ret, preopstr, postopstr);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(writev, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+ return 0;
+}
+
+int
+trace_readdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *buf,
+ dict_t *xdata)
{
- char preopstr[4096] = {0, };
- char postopstr[4096] = {0, };
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_SETATTR].enabled) {
- char string[4096] = {0,};
- if (op_ret == 0) {
- TRACE_STAT_TO_STR (statpre, preopstr);
- TRACE_STAT_TO_STR (statpost, postopstr);
-
- snprintf (string, sizeof (string),
- "%"PRId64": (op_ret=%d, "
- "*prebuf = {%s}, *postbuf = {%s})",
- frame->root->unique, op_ret,
- preopstr, postopstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "op_errno=%d)", frame->root->unique,
- uuid_utoa (frame->local), op_ret,
- op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (setattr, frame, op_ret, op_errno, statpre,
- statpost, xdata);
- return 0;
-}
-
-int
-trace_fsetattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *statpre, struct iatt *statpost, dict_t *xdata)
-{
- char preopstr[4096] = {0, };
- char postopstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FSETATTR].enabled) {
- char string[4096] = {0,};
- if (op_ret == 0) {
- TRACE_STAT_TO_STR (statpre, preopstr);
- TRACE_STAT_TO_STR (statpost, postopstr);
-
- snprintf (string, sizeof (string),
- "%"PRId64": (op_ret=%d, "
- "*prebuf = {%s}, *postbuf = {%s})",
- frame->root->unique, op_ret,
- preopstr, postopstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, op_errno=%d)",
- frame->root->unique, uuid_utoa (frame->local),
- op_ret, op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (fsetattr, frame, op_ret, op_errno,
- statpre, statpost, xdata);
- return 0;
-}
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_READDIR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 " : gfid=%s op_ret=%d, op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
-int
-trace_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
-{
- char preparentstr[4096] = {0, };
- char postparentstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_UNLINK].enabled) {
- char string[4096] = {0,};
- if (op_ret == 0) {
- TRACE_STAT_TO_STR (preparent, preparentstr);
- TRACE_STAT_TO_STR (postparent, postparentstr);
-
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- " *preparent = {%s}, "
- "*postparent = {%s})",
- frame->root->unique,
- uuid_utoa (frame->local),
- op_ret, preparentstr,
- postparentstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "op_errno=%d)",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret,
- op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- TRACE_STACK_UNWIND (unlink, frame, op_ret, op_errno,
- preparent, postparent, xdata);
- return 0;
-}
+ TRACE_STACK_UNWIND(readdir, frame, op_ret, op_errno, buf, xdata);
-int
-trace_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent,
- dict_t *xdata)
-{
- char statstr[4096] = {0, };
- char preoldparentstr[4096] = {0, };
- char postoldparentstr[4096] = {0, };
- char prenewparentstr[4096] = {0, };
- char postnewparentstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_RENAME].enabled) {
- char string[4096] = {0,};
- if (op_ret == 0) {
- TRACE_STAT_TO_STR (buf, statstr);
- TRACE_STAT_TO_STR (preoldparent, preoldparentstr);
- TRACE_STAT_TO_STR (postoldparent, postoldparentstr);
- TRACE_STAT_TO_STR (prenewparent, prenewparentstr);
- TRACE_STAT_TO_STR (postnewparent, postnewparentstr);
-
- snprintf (string, sizeof (string),
- "%"PRId64": (op_ret=%d, "
- "*stbuf = {%s}, *preoldparent = {%s},"
- " *postoldparent = {%s}"
- " *prenewparent = {%s}, "
- "*postnewparent = {%s})",
- frame->root->unique, op_ret, statstr,
- preoldparentstr, postoldparentstr,
- prenewparentstr, postnewparentstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "op_errno=%d", frame->root->unique,
- uuid_utoa (frame->local),
- op_ret, op_errno);
-
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (rename, frame, op_ret, op_errno, buf,
- preoldparent, postoldparent,
- prenewparent, postnewparent, xdata);
- return 0;
-}
-
-int
-trace_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- const char *buf, struct iatt *stbuf, dict_t *xdata)
-{
- char statstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_READLINK].enabled) {
- char string[4096] = {0,};
- if (op_ret == 0) {
- TRACE_STAT_TO_STR (stbuf, statstr);
- snprintf (string, sizeof (string),
- "%"PRId64": (op_ret=%d, op_errno=%d,"
- "buf=%s, stbuf = { %s })",
- frame->root->unique, op_ret, op_errno,
- buf, statstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "op_errno=%d",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret,
- op_errno);
- }
-
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (readlink, frame, op_ret, op_errno, buf, stbuf,
- xdata);
- return 0;
-}
-
-int
-trace_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf,
- dict_t *xdata, struct iatt *postparent)
-{
- char statstr[4096] = {0, };
- char postparentstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_LOOKUP].enabled) {
- char string[4096] = {0,};
- if (op_ret == 0) {
- TRACE_STAT_TO_STR (buf, statstr);
- TRACE_STAT_TO_STR (postparent, postparentstr);
- /* print buf->ia_gfid instead of inode->gfid,
- * since if the inode is not yet linked to the
- * inode table (fresh lookup) then null gfid
- * will be printed.
- */
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s (op_ret=%d "
- "*buf {%s}, *postparent {%s}",
- frame->root->unique,
- uuid_utoa (buf->ia_gfid),
- op_ret, statstr, postparentstr);
-
- /* For 'forget' */
- inode_ctx_put (inode, this, 0);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "op_errno=%d)",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret,
- op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, buf,
- xdata, postparent);
- return 0;
+ return 0;
}
int
-trace_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent,
+trace_readdirp_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *buf,
dict_t *xdata)
{
- char statstr[4096] = {0, };
- char preparentstr[4096] = {0, };
- char postparentstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_SYMLINK].enabled) {
- char string[4096] = {0,};
- if (op_ret == 0) {
- TRACE_STAT_TO_STR (buf, statstr);
- TRACE_STAT_TO_STR (preparent, preparentstr);
- TRACE_STAT_TO_STR (postparent, postparentstr);
-
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s (op_ret=%d "
- "*stbuf = {%s}, *preparent = {%s}, "
- "*postparent = {%s})",
- frame->root->unique,
- uuid_utoa (inode->gfid),
- op_ret, statstr, preparentstr,
- postparentstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": op_ret=%d, op_errno=%d",
- frame->root->unique, op_ret,
- op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (symlink, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
- return 0;
-}
-
-int
-trace_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
-{
- char statstr[4096] = {0, };
- char preparentstr[4096] = {0, };
- char postparentstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- char string[4096] = {0,};
- if (trace_fop_names[GF_FOP_MKNOD].enabled) {
- if (op_ret == 0) {
- TRACE_STAT_TO_STR (buf, statstr);
- TRACE_STAT_TO_STR (preparent, preparentstr);
- TRACE_STAT_TO_STR (postparent, postparentstr);
-
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s (op_ret=%d "
- "*stbuf = {%s}, *preparent = {%s}, "
- "*postparent = {%s})",
- frame->root->unique,
- uuid_utoa (inode->gfid),
- op_ret, statstr, preparentstr,
- postparentstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": (op_ret=%d, op_errno=%d)",
- frame->root->unique, op_ret,
- op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
- return 0;
-}
-
-int
-trace_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
-{
- char statstr[4096] = {0, };
- char preparentstr[4096] = {0, };
- char postparentstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_MKDIR].enabled) {
- char string[4096] = {0,};
- if (op_ret == 0) {
- TRACE_STAT_TO_STR (buf, statstr);
- TRACE_STAT_TO_STR (preparent, preparentstr);
- TRACE_STAT_TO_STR (postparent, postparentstr);
-
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s (op_ret=%d "
- ", *stbuf = {%s}, *prebuf = {%s}, "
- "*postbuf = {%s} )",
- frame->root->unique,
- uuid_utoa (inode->gfid),
- op_ret, statstr, preparentstr,
- postparentstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": (op_ret=%d, op_errno=%d)",
- frame->root->unique, op_ret,
- op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (mkdir, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
- return 0;
-}
-
-int
-trace_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
-{
- char statstr[4096] = {0, };
- char preparentstr[4096] = {0, };
- char postparentstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- char string[4096] = {0,};
- if (trace_fop_names[GF_FOP_LINK].enabled) {
- if (op_ret == 0) {
- TRACE_STAT_TO_STR (buf, statstr);
- TRACE_STAT_TO_STR (preparent, preparentstr);
- TRACE_STAT_TO_STR (postparent, postparentstr);
-
- snprintf (string, sizeof (string),
- "%"PRId64": (op_ret=%d, "
- "*stbuf = {%s}, *prebuf = {%s},"
- " *postbuf = {%s})",
- frame->root->unique, op_ret,
- statstr, preparentstr, postparentstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "op_errno=%d",
- frame->root->unique,
- uuid_utoa (frame->local),
- op_ret, op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (link, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
- return 0;
-}
-
-int
-trace_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+ int count = 0;
+ char statstr[1024] = {
+ 0,
+ };
+ char string[4096] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+ gf_dirent_t *entry = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_READDIRP].enabled) {
+ snprintf(string, sizeof(string),
+ "%" PRId64 " : gfid=%s op_ret=%d, op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+
+ LOG_ELEMENT(conf, string);
+ }
+ if (op_ret < 0)
+ goto out;
+
+ list_for_each_entry(entry, &buf->list, list)
+ {
+ count++;
+ TRACE_STAT_TO_STR(&entry->d_stat, statstr);
+ snprintf(string, sizeof(string),
+ "entry no. %d, pargfid=%s, "
+ "bname=%s *buf {%s}",
+ count, uuid_utoa(frame->local), entry->d_name, statstr);
+ LOG_ELEMENT(conf, string);
+ }
+
+out:
+ TRACE_STACK_UNWIND(readdirp, frame, op_ret, op_errno, buf, xdata);
+ return 0;
+}
+
+int
+trace_fsync_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ char preopstr[1024] = {
+ 0,
+ };
+ char postopstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FSYNC].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret == 0) {
+ TRACE_STAT_TO_STR(prebuf, preopstr);
+ TRACE_STAT_TO_STR(postbuf, postopstr);
+
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": (op_ret=%d, "
+ "*prebuf = {%s}, *postbuf = {%s}",
+ frame->root->unique, op_ret, preopstr, postopstr);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+
+ return 0;
+}
+
+int
+trace_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
+{
+ char preopstr[1024] = {
+ 0,
+ };
+ char postopstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_SETATTR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret == 0) {
+ TRACE_STAT_TO_STR(statpre, preopstr);
+ TRACE_STAT_TO_STR(statpost, postopstr);
+
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": (op_ret=%d, "
+ "*prebuf = {%s}, *postbuf = {%s})",
+ frame->root->unique, op_ret, preopstr, postopstr);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "op_errno=%d)",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(setattr, frame, op_ret, op_errno, statpre, statpost,
+ xdata);
+ return 0;
+}
+
+int
+trace_fsetattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
+{
+ char preopstr[1024] = {
+ 0,
+ };
+ char postopstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FSETATTR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret == 0) {
+ TRACE_STAT_TO_STR(statpre, preopstr);
+ TRACE_STAT_TO_STR(statpost, postopstr);
+
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": (op_ret=%d, "
+ "*prebuf = {%s}, *postbuf = {%s})",
+ frame->root->unique, op_ret, preopstr, postopstr);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s op_ret=%d, op_errno=%d)",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(fsetattr, frame, op_ret, op_errno, statpre, statpost,
+ xdata);
+ return 0;
+}
+
+int
+trace_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ char preparentstr[1024] = {
+ 0,
+ };
+ char postparentstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_UNLINK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret == 0) {
+ TRACE_STAT_TO_STR(preparent, preparentstr);
+ TRACE_STAT_TO_STR(postparent, postparentstr);
+
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ " *preparent = {%s}, "
+ "*postparent = {%s})",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ preparentstr, postparentstr);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "op_errno=%d)",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(unlink, frame, op_ret, op_errno, preparent, postparent,
+ xdata);
+ return 0;
+}
+
+int
+trace_rename_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ struct iatt *preoldparent, struct iatt *postoldparent,
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
+{
+ char statstr[1024] = {
+ 0,
+ };
+ char preoldparentstr[1024] = {
+ 0,
+ };
+ char postoldparentstr[1024] = {
+ 0,
+ };
+ char prenewparentstr[1024] = {
+ 0,
+ };
+ char postnewparentstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_RENAME].enabled) {
+ char string[6044] = {
+ 0,
+ };
+ if (op_ret == 0) {
+ TRACE_STAT_TO_STR(buf, statstr);
+ TRACE_STAT_TO_STR(preoldparent, preoldparentstr);
+ TRACE_STAT_TO_STR(postoldparent, postoldparentstr);
+ TRACE_STAT_TO_STR(prenewparent, prenewparentstr);
+ TRACE_STAT_TO_STR(postnewparent, postnewparentstr);
+
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": (op_ret=%d, "
+ "*stbuf = {%s}, *preoldparent = {%s},"
+ " *postoldparent = {%s}"
+ " *prenewparent = {%s}, "
+ "*postnewparent = {%s})",
+ frame->root->unique, op_ret, statstr, preoldparentstr,
+ postoldparentstr, prenewparentstr, postnewparentstr);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(rename, frame, op_ret, op_errno, buf, preoldparent,
+ postoldparent, prenewparent, postnewparent, xdata);
+ return 0;
+}
+
+int
+trace_readlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, const char *buf,
+ struct iatt *stbuf, dict_t *xdata)
+{
+ char statstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_READLINK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret == 0) {
+ TRACE_STAT_TO_STR(stbuf, statstr);
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": (op_ret=%d, op_errno=%d,"
+ "buf=%s, stbuf = { %s })",
+ frame->root->unique, op_ret, op_errno, buf, statstr);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+ }
+
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(readlink, frame, op_ret, op_errno, buf, stbuf, xdata);
+ return 0;
+}
+
+int
+trace_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+{
+ char statstr[1024] = {
+ 0,
+ };
+ char postparentstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_LOOKUP].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret == 0) {
+ TRACE_STAT_TO_STR(buf, statstr);
+ TRACE_STAT_TO_STR(postparent, postparentstr);
+ /* print buf->ia_gfid instead of inode->gfid,
+ * since if the inode is not yet linked to the
+ * inode table (fresh lookup) then null gfid
+ * will be printed.
+ */
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s (op_ret=%d "
+ "*buf {%s}, *postparent {%s}",
+ frame->root->unique, uuid_utoa(buf->ia_gfid), op_ret,
+ statstr, postparentstr);
+
+ /* For 'forget' */
+ inode_ctx_put(inode, this, 0);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "op_errno=%d)",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(lookup, frame, op_ret, op_errno, inode, buf, xdata,
+ postparent);
+ return 0;
+}
+
+int
+trace_symlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ char statstr[1024] = {
+ 0,
+ };
+ char preparentstr[1024] = {
+ 0,
+ };
+ char postparentstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_SYMLINK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret == 0) {
+ TRACE_STAT_TO_STR(buf, statstr);
+ TRACE_STAT_TO_STR(preparent, preparentstr);
+ TRACE_STAT_TO_STR(postparent, postparentstr);
+
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s (op_ret=%d "
+ "*stbuf = {%s}, *preparent = {%s}, "
+ "*postparent = {%s})",
+ frame->root->unique, uuid_utoa(inode->gfid), op_ret,
+ statstr, preparentstr, postparentstr);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": op_ret=%d, op_errno=%d", frame->root->unique,
+ op_ret, op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(symlink, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
+ return 0;
+}
+
+int
+trace_mknod_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ char statstr[1024] = {
+ 0,
+ };
+ char preparentstr[1024] = {
+ 0,
+ };
+ char postparentstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ char string[4096] = {
+ 0,
+ };
+ if (trace_fop_names[GF_FOP_MKNOD].enabled) {
+ if (op_ret == 0) {
+ TRACE_STAT_TO_STR(buf, statstr);
+ TRACE_STAT_TO_STR(preparent, preparentstr);
+ TRACE_STAT_TO_STR(postparent, postparentstr);
+
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s (op_ret=%d "
+ "*stbuf = {%s}, *preparent = {%s}, "
+ "*postparent = {%s})",
+ frame->root->unique, uuid_utoa(inode->gfid), op_ret,
+ statstr, preparentstr, postparentstr);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": (op_ret=%d, op_errno=%d)",
+ frame->root->unique, op_ret, op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(mknod, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
+ return 0;
+}
+
+int
+trace_mkdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ char statstr[1024] = {
+ 0,
+ };
+ char preparentstr[1024] = {
+ 0,
+ };
+ char postparentstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_MKDIR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret == 0) {
+ TRACE_STAT_TO_STR(buf, statstr);
+ TRACE_STAT_TO_STR(preparent, preparentstr);
+ TRACE_STAT_TO_STR(postparent, postparentstr);
+
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s (op_ret=%d "
+ ", *stbuf = {%s}, *prebuf = {%s}, "
+ "*postbuf = {%s} )",
+ frame->root->unique, uuid_utoa(inode->gfid), op_ret,
+ statstr, preparentstr, postparentstr);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": (op_ret=%d, op_errno=%d)",
+ frame->root->unique, op_ret, op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(mkdir, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
+ return 0;
+}
+
+int
+trace_link_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ char statstr[1024] = {
+ 0,
+ };
+ char preparentstr[1024] = {
+ 0,
+ };
+ char postparentstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ char string[4096] = {
+ 0,
+ };
+ if (trace_fop_names[GF_FOP_LINK].enabled) {
+ if (op_ret == 0) {
+ TRACE_STAT_TO_STR(buf, statstr);
+ TRACE_STAT_TO_STR(preparent, preparentstr);
+ TRACE_STAT_TO_STR(postparent, postparentstr);
+
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": (op_ret=%d, "
+ "*stbuf = {%s}, *prebuf = {%s},"
+ " *postbuf = {%s})",
+ frame->root->unique, op_ret, statstr, preparentstr,
+ postparentstr);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(link, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
+ return 0;
+}
+
+int
+trace_flush_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ char string[4096] = {
+ 0,
+ };
+ if (trace_fop_names[GF_FOP_FLUSH].enabled) {
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s op_ret=%d, op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(flush, frame, op_ret, op_errno, xdata);
+ return 0;
+}
+
+int
+trace_opendir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ char string[4096] = {
+ 0,
+ };
+ if (trace_fop_names[GF_FOP_OPENDIR].enabled) {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, op_errno=%d,"
+ " fd=%p",
+ frame->root->unique, uuid_utoa(frame->local), op_ret, op_errno,
+ fd);
+
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ /* for 'releasedir' log */
+ if (op_ret >= 0)
+ fd_ctx_set(fd, this, 0);
+
+ TRACE_STACK_UNWIND(opendir, frame, op_ret, op_errno, fd, xdata);
+ return 0;
+}
+
+int
+trace_rmdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ char preparentstr[1024] = {
+ 0,
+ };
+ char postparentstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_RMDIR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret == 0) {
+ TRACE_STAT_TO_STR(preparent, preparentstr);
+ TRACE_STAT_TO_STR(postparent, postparentstr);
+
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "*prebuf={%s}, *postbuf={%s}",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ preparentstr, postparentstr);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(rmdir, frame, op_ret, op_errno, preparent, postparent,
+ xdata);
+ return 0;
+}
+
+int
+trace_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ char preopstr[1024] = {
+ 0,
+ };
+ char postopstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_TRUNCATE].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret == 0) {
+ TRACE_STAT_TO_STR(prebuf, preopstr);
+ TRACE_STAT_TO_STR(postbuf, postopstr);
+
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": (op_ret=%d, "
+ "*prebuf = {%s}, *postbuf = {%s} )",
+ frame->root->unique, op_ret, preopstr, postopstr);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(truncate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
+}
+
+int
+trace_statfs_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct statvfs *buf,
+ dict_t *xdata)
+{
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_STATFS].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret == 0) {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": ({f_bsize=%lu, "
+ "f_frsize=%lu, "
+ "f_blocks=%" GF_PRI_FSBLK ", f_bfree=%" GF_PRI_FSBLK
+ ", "
+ "f_bavail=%" GF_PRI_FSBLK
+ ", "
+ "f_files=%" GF_PRI_FSBLK
+ ", "
+ "f_ffree=%" GF_PRI_FSBLK
+ ", "
+ "f_favail=%" GF_PRI_FSBLK
+ ", "
+ "f_fsid=%lu, f_flag=%lu, "
+ "f_namemax=%lu}) => ret=%d",
+ frame->root->unique, buf->f_bsize, buf->f_frsize,
+ buf->f_blocks, buf->f_bfree, buf->f_bavail, buf->f_files,
+ buf->f_ffree, buf->f_favail, buf->f_fsid, buf->f_flag,
+ buf->f_namemax, op_ret);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": (op_ret=%d, "
+ "op_errno=%d)",
+ frame->root->unique, op_ret, op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(statfs, frame, op_ret, op_errno, buf, xdata);
+ return 0;
+}
+
+int
+trace_setxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- char string[4096] = {0,};
- if (trace_fop_names[GF_FOP_FLUSH].enabled) {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
- frame->root->unique, uuid_utoa (frame->local),
- op_ret, op_errno);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_SETXATTR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s op_ret=%d, op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- TRACE_STACK_UNWIND (flush, frame, op_ret, op_errno, xdata);
- return 0;
+ TRACE_STACK_UNWIND(setxattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-trace_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+trace_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- char string[4096] = {0,};
- if (trace_fop_names[GF_FOP_OPENDIR].enabled) {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, op_errno=%d,"
- " fd=%p",
- frame->root->unique, uuid_utoa (frame->local),
- op_ret, op_errno, fd);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_GETXATTR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, op_errno=%d,"
+ " dict=%p",
+ frame->root->unique, uuid_utoa(frame->local), op_ret, op_errno,
+ dict);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- /* for 'releasedir' log */
- if (op_ret >= 0)
- fd_ctx_set (fd, this, 0);
-
- TRACE_STACK_UNWIND (opendir, frame, op_ret, op_errno, fd, xdata);
- return 0;
-}
-
-int
-trace_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
-{
- char preparentstr[4096] = {0, };
- char postparentstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_RMDIR].enabled) {
- char string[4096] = {0,};
- if (op_ret == 0) {
- TRACE_STAT_TO_STR (preparent, preparentstr);
- TRACE_STAT_TO_STR (postparent, postparentstr);
-
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "*prebuf={%s}, *postbuf={%s}",
- frame->root->unique,
- uuid_utoa (frame->local),
- op_ret, preparentstr, postparentstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "op_errno=%d", frame->root->unique,
- uuid_utoa (frame->local),
- op_ret, op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (rmdir, frame, op_ret, op_errno,
- preparent, postparent, xdata);
- return 0;
-}
-
-int
-trace_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
-{
- char preopstr[4096] = {0, };
- char postopstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_TRUNCATE].enabled) {
- char string[4096] = {0,};
- if (op_ret == 0) {
- TRACE_STAT_TO_STR (prebuf, preopstr);
- TRACE_STAT_TO_STR (postbuf, postopstr);
-
- snprintf (string, sizeof (string),
- "%"PRId64": (op_ret=%d, "
- "*prebuf = {%s}, *postbuf = {%s} )",
- frame->root->unique, op_ret,
- preopstr, postopstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "op_errno=%d", frame->root->unique,
- uuid_utoa (frame->local), op_ret,
- op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (truncate, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
- return 0;
-}
+ TRACE_STACK_UNWIND(getxattr, frame, op_ret, op_errno, dict, xdata);
-int
-trace_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct statvfs *buf,
- dict_t *xdata)
-{
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_STATFS].enabled) {
- char string[4096] = {0,};
- if (op_ret == 0) {
- snprintf (string, sizeof (string),
- "%"PRId64": ({f_bsize=%lu, "
- "f_frsize=%lu, "
- "f_blocks=%"GF_PRI_FSBLK
- ", f_bfree=%"GF_PRI_FSBLK", "
- "f_bavail=%"GF_PRI_FSBLK", "
- "f_files=%"GF_PRI_FSBLK", "
- "f_ffree=%"GF_PRI_FSBLK", "
- "f_favail=%"GF_PRI_FSBLK", "
- "f_fsid=%lu, f_flag=%lu, "
- "f_namemax=%lu}) => ret=%d",
- frame->root->unique, buf->f_bsize,
- buf->f_frsize, buf->f_blocks,
- buf->f_bfree, buf->f_bavail,
- buf->f_files, buf->f_ffree,
- buf->f_favail, buf->f_fsid,
- buf->f_flag, buf->f_namemax, op_ret);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": (op_ret=%d, "
- "op_errno=%d)",
- frame->root->unique, op_ret,
- op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (statfs, frame, op_ret, op_errno, buf, xdata);
- return 0;
+ return 0;
}
int
-trace_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+trace_fsetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_SETXATTR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret,
- op_errno);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FSETXATTR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s op_ret=%d, op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- TRACE_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
- return 0;
+ TRACE_STACK_UNWIND(fsetxattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-trace_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+trace_fgetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, dict_t *dict,
dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_GETXATTR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, op_errno=%d,"
- " dict=%p", frame->root->unique,
- uuid_utoa (frame->local), op_ret, op_errno,
- dict);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FGETXATTR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, op_errno=%d,"
+ " dict=%p",
+ frame->root->unique, uuid_utoa(frame->local), op_ret, op_errno,
+ dict);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- TRACE_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
+ TRACE_STACK_UNWIND(fgetxattr, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ return 0;
}
int
-trace_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+trace_removexattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FSETXATTR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret, op_errno);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_REMOVEXATTR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s op_ret=%d, op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- TRACE_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, xdata);
- return 0;
-}
+ TRACE_STACK_UNWIND(removexattr, frame, op_ret, op_errno, xdata);
-int
-trace_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
-{
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FGETXATTR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, op_errno=%d,"
- " dict=%p", frame->root->unique,
- uuid_utoa (frame->local), op_ret, op_errno,
- dict);
-
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, dict, xdata);
-
- return 0;
+ return 0;
}
int
-trace_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+trace_fsyncdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_REMOVEXATTR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret, op_errno);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FSYNCDIR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s op_ret=%d, op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- TRACE_STACK_UNWIND (removexattr, frame, op_ret, op_errno, xdata);
-
- return 0;
+ TRACE_STACK_UNWIND(fsyncdir, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-trace_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+trace_access_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FSYNCDIR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret, op_errno);
-
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (fsyncdir, frame, op_ret, op_errno, xdata);
- return 0;
-}
-
-int
-trace_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_ACCESS].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "op_errno=%d)",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(access, frame, op_ret, op_errno, xdata);
+ return 0;
+}
+
+int
+trace_ftruncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ char prebufstr[1024] = {
+ 0,
+ };
+ char postbufstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FTRUNCATE].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret == 0) {
+ TRACE_STAT_TO_STR(prebuf, prebufstr);
+ TRACE_STAT_TO_STR(postbuf, postbufstr);
+
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": op_ret=%d, "
+ "*prebuf = {%s}, *postbuf = {%s} )",
+ frame->root->unique, op_ret, prebufstr, postbufstr);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(ftruncate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
+}
+
+int
+trace_fstat_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
+{
+ char statstr[1024] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FSTAT].enabled) {
+ char string[4096] = {0.};
+ if (op_ret == 0) {
+ TRACE_STAT_TO_STR(buf, statstr);
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d "
+ "buf=%s",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ statstr);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+ }
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(fstat, frame, op_ret, op_errno, buf, xdata);
+ return 0;
+}
+
+int
+trace_lk_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
+{
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_LK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (op_ret == 0) {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "{l_type=%d, l_whence=%d, "
+ "l_start=%" PRId64
+ ", "
+ "l_len=%" PRId64 ", l_pid=%u})",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ lock->l_type, lock->l_whence, lock->l_start, lock->l_len,
+ lock->l_pid);
+ } else {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d, "
+ "op_errno=%d)",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
+ }
+
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(lk, frame, op_ret, op_errno, lock, xdata);
+ return 0;
+}
+
+int
+trace_entrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
-
- conf = this->private;
+ trace_conf_t *conf = NULL;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_ACCESS].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "op_errno=%d)", frame->root->unique,
- uuid_utoa (frame->local), op_ret, op_errno);
+ conf = this->private;
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (access, frame, op_ret, op_errno, xdata);
- return 0;
-}
-
-int
-trace_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
-{
- char prebufstr[4096] = {0, };
- char postbufstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FTRUNCATE].enabled) {
- char string[4096] = {0,};
- if (op_ret == 0) {
- TRACE_STAT_TO_STR (prebuf, prebufstr);
- TRACE_STAT_TO_STR (postbuf, postbufstr);
-
- snprintf (string, sizeof (string),
- "%"PRId64": op_ret=%d, "
- "*prebuf = {%s}, *postbuf = {%s} )",
- frame->root->unique, op_ret,
- prebufstr, postbufstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "op_errno=%d", frame->root->unique,
- uuid_utoa (frame->local), op_ret,
- op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (ftruncate, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
- return 0;
-}
-
-int
-trace_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
-{
- char statstr[4096] = {0, };
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FSTAT].enabled) {
- char string[4096] = {0.};
- if (op_ret == 0) {
- TRACE_STAT_TO_STR (buf, statstr);
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d "
- "buf=%s", frame->root->unique,
- uuid_utoa (frame->local), op_ret,
- statstr);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "op_errno=%d", frame->root->unique,
- uuid_utoa (frame->local), op_ret,
- op_errno);
- }
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf, xdata);
- return 0;
-}
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_ENTRYLK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s op_ret=%d, op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
-int
-trace_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
- dict_t *xdata)
-{
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_LK].enabled) {
- char string[4096] = {0,};
- if (op_ret == 0) {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "{l_type=%d, l_whence=%d, "
- "l_start=%"PRId64", "
- "l_len=%"PRId64", l_pid=%u})",
- frame->root->unique,
- uuid_utoa (frame->local),
- op_ret, lock->l_type, lock->l_whence,
- lock->l_start, lock->l_len,
- lock->l_pid);
- } else {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, "
- "op_errno=%d)", frame->root->unique,
- uuid_utoa (frame->local), op_ret,
- op_errno);
- }
-
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- TRACE_STACK_UNWIND (lk, frame, op_ret, op_errno, lock, xdata);
- return 0;
+ TRACE_STACK_UNWIND(entrylk, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-trace_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+trace_fentrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_ENTRYLK].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret, op_errno);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FENTRYLK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s op_ret=%d, op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- TRACE_STACK_UNWIND (entrylk, frame, op_ret, op_errno, xdata);
- return 0;
+ TRACE_STACK_UNWIND(fentrylk, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-trace_fentrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+trace_xattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FENTRYLK].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret, op_errno);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_XATTROP].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s op_ret=%d, op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- TRACE_STACK_UNWIND (fentrylk, frame, op_ret, op_errno, xdata);
- return 0;
+ TRACE_STACK_UNWIND(xattrop, frame, op_ret, op_errno, dict, xdata);
+ return 0;
}
int
-trace_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+trace_fxattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, dict_t *dict,
dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_XATTROP].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret, op_errno);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FXATTROP].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s op_ret=%d, op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- TRACE_STACK_UNWIND (xattrop, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ TRACE_STACK_UNWIND(fxattrop, frame, op_ret, op_errno, dict, xdata);
+ return 0;
}
int
-trace_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
+trace_inodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FXATTROP].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret, op_errno);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_INODELK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s op_ret=%d, op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- TRACE_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ TRACE_STACK_UNWIND(inodelk, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-trace_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+trace_finodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_INODELK].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
- frame->root->unique,
- uuid_utoa (frame->local),op_ret, op_errno);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FINODELK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s op_ret=%d, op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (inodelk, frame, op_ret, op_errno, xdata);
- return 0;
-}
-
-int
-trace_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FINODELK].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret, op_errno);
-
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- TRACE_STACK_UNWIND (finodelk, frame, op_ret, op_errno, xdata);
- return 0;
+ TRACE_STACK_UNWIND(finodelk, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-trace_rchecksum_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- uint32_t weak_checksum, uint8_t *strong_checksum,
- dict_t *xdata)
+trace_rchecksum_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, uint32_t weak_checksum,
+ uint8_t *strong_checksum, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_RCHECKSUM].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d op_errno=%d",
- frame->root->unique,
- uuid_utoa (frame->local), op_ret, op_errno);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_RCHECKSUM].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s op_ret=%d op_errno=%d",
+ frame->root->unique, uuid_utoa(frame->local), op_ret,
+ op_errno);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- TRACE_STACK_UNWIND (rchecksum, frame, op_ret, op_errno, weak_checksum,
- strong_checksum, xdata);
+ TRACE_STACK_UNWIND(rchecksum, frame, op_ret, op_errno, weak_checksum,
+ strong_checksum, xdata);
- return 0;
+ return 0;
}
/* *_cbk section over <----------> fop section start */
int
-trace_entrylk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, const char *basename,
- entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
+trace_entrylk(call_frame_t *frame, xlator_t *this, const char *volume,
+ loc_t *loc, const char *basename, entrylk_cmd cmd,
+ entrylk_type type, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_ENTRYLK].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s volume=%s, (path=%s "
- "basename=%s, cmd=%s, type=%s)",
- frame->root->unique,
- uuid_utoa (loc->inode->gfid),
- volume, loc->path, basename,
- ((cmd == ENTRYLK_LOCK) ? "ENTRYLK_LOCK" :
- "ENTRYLK_UNLOCK"),
- ((type == ENTRYLK_RDLCK) ? "ENTRYLK_RDLCK" :
- "ENTRYLK_WRLCK"));
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_ENTRYLK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s volume=%s, (path=%s "
+ "basename=%s, cmd=%s, type=%s)",
+ frame->root->unique, uuid_utoa(loc->inode->gfid), volume,
+ loc->path, basename,
+ ((cmd == ENTRYLK_LOCK) ? "ENTRYLK_LOCK" : "ENTRYLK_UNLOCK"),
+ ((type == ENTRYLK_RDLCK) ? "ENTRYLK_RDLCK" : "ENTRYLK_WRLCK"));
- frame->local = loc->inode->gfid;
+ frame->local = loc->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_entrylk_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->entrylk,
- volume, loc, basename, cmd, type, xdata);
- return 0;
+ STACK_WIND(frame, trace_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, volume, loc, basename, cmd,
+ type, xdata);
+ return 0;
}
int
-trace_inodelk (call_frame_t *frame, xlator_t *this, const char *volume,
- loc_t *loc, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
+trace_inodelk(call_frame_t *frame, xlator_t *this, const char *volume,
+ loc_t *loc, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
{
- char *cmd_str = NULL;
- char *type_str = NULL;
- trace_conf_t *conf = NULL;
+ char *cmd_str = NULL;
+ char *type_str = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_INODELK].enabled) {
- char string[4096] = {0,};
- switch (cmd) {
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_INODELK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ switch (cmd) {
#if F_GETLK != F_GETLK64
- case F_GETLK64:
+ case F_GETLK64:
#endif
- case F_GETLK:
- cmd_str = "GETLK";
- break;
+ case F_GETLK:
+ cmd_str = "GETLK";
+ break;
#if F_SETLK != F_SETLK64
- case F_SETLK64:
+ case F_SETLK64:
#endif
- case F_SETLK:
- cmd_str = "SETLK";
- break;
+ case F_SETLK:
+ cmd_str = "SETLK";
+ break;
#if F_SETLKW != F_SETLKW64
- case F_SETLKW64:
+ case F_SETLKW64:
#endif
- case F_SETLKW:
- cmd_str = "SETLKW";
- break;
-
- default:
- cmd_str = "UNKNOWN";
- break;
- }
-
- switch (flock->l_type) {
- case F_RDLCK:
- type_str = "READ";
- break;
- case F_WRLCK:
- type_str = "WRITE";
- break;
- case F_UNLCK:
- type_str = "UNLOCK";
- break;
- default:
- type_str = "UNKNOWN";
- break;
- }
-
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s volume=%s, (path=%s "
- "cmd=%s, type=%s, start=%llu, len=%llu, "
- "pid=%llu)", frame->root->unique,
- uuid_utoa (loc->inode->gfid), volume,
- loc->path, cmd_str, type_str,
- (unsigned long long)flock->l_start,
- (unsigned long long) flock->l_len,
- (unsigned long long) flock->l_pid);
-
- frame->local = loc->inode->gfid;
-
- LOG_ELEMENT (conf, string);
+ case F_SETLKW:
+ cmd_str = "SETLKW";
+ break;
+
+ default:
+ cmd_str = "UNKNOWN";
+ break;
+ }
+
+ switch (flock->l_type) {
+ case F_RDLCK:
+ type_str = "READ";
+ break;
+ case F_WRLCK:
+ type_str = "WRITE";
+ break;
+ case F_UNLCK:
+ type_str = "UNLOCK";
+ break;
+ default:
+ type_str = "UNKNOWN";
+ break;
}
+ snprintf(
+ string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s volume=%s, (path=%s "
+ "cmd=%s, type=%s, start=%llu, len=%llu, "
+ "pid=%llu)",
+ frame->root->unique, uuid_utoa(loc->inode->gfid), volume, loc->path,
+ cmd_str, type_str, (unsigned long long)flock->l_start,
+ (unsigned long long)flock->l_len, (unsigned long long)flock->l_pid);
+
+ frame->local = loc->inode->gfid;
+
+ LOG_ELEMENT(conf, string);
+ }
+
out:
- STACK_WIND (frame, trace_inodelk_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->inodelk,
- volume, loc, cmd, flock, xdata);
- return 0;
+ STACK_WIND(frame, trace_inodelk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->inodelk, volume, loc, cmd, flock,
+ xdata);
+ return 0;
}
int
-trace_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
- fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
+trace_finodelk(call_frame_t *frame, xlator_t *this, const char *volume,
+ fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
{
- char *cmd_str = NULL;
- char *type_str = NULL;
- trace_conf_t *conf = NULL;
+ char *cmd_str = NULL;
+ char *type_str = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FINODELK].enabled) {
- char string[4096] = {0,};
- switch (cmd) {
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FINODELK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ switch (cmd) {
#if F_GETLK != F_GETLK64
- case F_GETLK64:
+ case F_GETLK64:
#endif
- case F_GETLK:
- cmd_str = "GETLK";
- break;
+ case F_GETLK:
+ cmd_str = "GETLK";
+ break;
#if F_SETLK != F_SETLK64
- case F_SETLK64:
+ case F_SETLK64:
#endif
- case F_SETLK:
- cmd_str = "SETLK";
- break;
+ case F_SETLK:
+ cmd_str = "SETLK";
+ break;
#if F_SETLKW != F_SETLKW64
- case F_SETLKW64:
+ case F_SETLKW64:
#endif
- case F_SETLKW:
- cmd_str = "SETLKW";
- break;
-
- default:
- cmd_str = "UNKNOWN";
- break;
- }
-
- switch (flock->l_type) {
- case F_RDLCK:
- type_str = "READ";
- break;
- case F_WRLCK:
- type_str = "WRITE";
- break;
- case F_UNLCK:
- type_str = "UNLOCK";
- break;
- default:
- type_str = "UNKNOWN";
- break;
- }
-
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s volume=%s, (fd =%p "
- "cmd=%s, type=%s, start=%llu, len=%llu, "
- "pid=%llu)", frame->root->unique,
- uuid_utoa (fd->inode->gfid), volume, fd,
- cmd_str, type_str,
- (unsigned long long) flock->l_start,
- (unsigned long long) flock->l_len,
- (unsigned long long) flock->l_pid);
-
- frame->local = fd->inode->gfid;
-
- LOG_ELEMENT (conf, string);
+ case F_SETLKW:
+ cmd_str = "SETLKW";
+ break;
+
+ default:
+ cmd_str = "UNKNOWN";
+ break;
}
+
+ switch (flock->l_type) {
+ case F_RDLCK:
+ type_str = "READ";
+ break;
+ case F_WRLCK:
+ type_str = "WRITE";
+ break;
+ case F_UNLCK:
+ type_str = "UNLOCK";
+ break;
+ default:
+ type_str = "UNKNOWN";
+ break;
+ }
+
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s volume=%s, (fd =%p "
+ "cmd=%s, type=%s, start=%llu, len=%llu, "
+ "pid=%llu)",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), volume, fd,
+ cmd_str, type_str, (unsigned long long)flock->l_start,
+ (unsigned long long)flock->l_len,
+ (unsigned long long)flock->l_pid);
+
+ frame->local = fd->inode->gfid;
+
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_finodelk_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->finodelk,
- volume, fd, cmd, flock, xdata);
- return 0;
+ STACK_WIND(frame, trace_finodelk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->finodelk, volume, fd, cmd, flock,
+ xdata);
+ return 0;
}
int
-trace_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+trace_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_XATTROP].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s (path=%s flags=%d)",
- frame->root->unique,
- uuid_utoa (loc->inode->gfid), loc->path,
- flags);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_XATTROP].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s (path=%s flags=%d)", frame->root->unique,
+ uuid_utoa(loc->inode->gfid), loc->path, flags);
- frame->local = loc->inode->gfid;
+ frame->local = loc->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_xattrop_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->xattrop,
- loc, flags, dict, xdata);
+ STACK_WIND(frame, trace_xattrop_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->xattrop, loc, flags, dict, xdata);
- return 0;
+ return 0;
}
int
-trace_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
- gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+trace_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FXATTROP].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s fd=%p, flags=%d",
- frame->root->unique,
- uuid_utoa (fd->inode->gfid), fd, flags);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FXATTROP].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string), "%" PRId64 ": gfid=%s fd=%p, flags=%d",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), fd, flags);
- frame->local = fd->inode->gfid;
+ frame->local = fd->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_fxattrop_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fxattrop,
- fd, flags, dict, xdata);
+ STACK_WIND(frame, trace_fxattrop_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fxattrop, fd, flags, dict, xdata);
- return 0;
+ return 0;
}
int
-trace_lookup (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xdata)
+trace_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_LOOKUP].enabled) {
- char string[4096] = {0,};
- /* TODO: print all the keys mentioned in xattr_req */
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s",
- frame->root->unique,
- uuid_utoa (loc->inode->gfid), loc->path);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_LOOKUP].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ /* TODO: print all the keys mentioned in xattr_req */
+ snprintf(string, sizeof(string), "%" PRId64 ": gfid=%s path=%s",
+ frame->root->unique, uuid_utoa(loc->inode->gfid), loc->path);
- frame->local = loc->inode->gfid;
+ frame->local = loc->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_lookup_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup,
- loc, xdata);
+ STACK_WIND(frame, trace_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, xdata);
- return 0;
+ return 0;
}
int
-trace_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+trace_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_STAT].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s",
- frame->root->unique,
- uuid_utoa (loc->inode->gfid), loc->path);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_STAT].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string), "%" PRId64 ": gfid=%s path=%s",
+ frame->root->unique, uuid_utoa(loc->inode->gfid), loc->path);
- frame->local = loc->inode->gfid;
+ frame->local = loc->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_stat_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->stat,
- loc, xdata);
+ STACK_WIND(frame, trace_stat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->stat, loc, xdata);
- return 0;
+ return 0;
}
int
-trace_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
- dict_t *xdata)
+trace_readlink(call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+ dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_READLINK].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s, "
- "size=%"GF_PRI_SIZET")", frame->root->unique,
- uuid_utoa (loc->inode->gfid), loc->path,
- size);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_READLINK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s path=%s, "
+ "size=%" GF_PRI_SIZET ")",
+ frame->root->unique, uuid_utoa(loc->inode->gfid), loc->path,
+ size);
- frame->local = loc->inode->gfid;
+ frame->local = loc->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_readlink_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readlink,
- loc, size, xdata);
+ STACK_WIND(frame, trace_readlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readlink, loc, size, xdata);
- return 0;
+ return 0;
}
int
-trace_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
- mode_t mode, dev_t dev, mode_t umask, dict_t *xdata)
+trace_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t dev, mode_t umask, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_MKNOD].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s mode=%d "
- "umask=0%o, dev=%"GF_PRI_DEV")",
- frame->root->unique,
- uuid_utoa (loc->inode->gfid), loc->path,
- mode, umask, dev);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_MKNOD].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s path=%s mode=%d "
+ "umask=0%o, dev=%" GF_PRI_DEV ")",
+ frame->root->unique, uuid_utoa(loc->inode->gfid), loc->path,
+ mode, umask, dev);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_mknod_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mknod,
- loc, mode, dev, umask, xdata);
+ STACK_WIND(frame, trace_mknod_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, loc, mode, dev, umask, xdata);
- return 0;
+ return 0;
}
int
-trace_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- mode_t umask, dict_t *xdata)
+trace_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_MKDIR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s mode=%d"
- " umask=0%o", frame->root->unique,
- uuid_utoa (loc->inode->gfid), loc->path,
- mode, umask);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_MKDIR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s path=%s mode=%d"
+ " umask=0%o",
+ frame->root->unique, uuid_utoa(loc->inode->gfid), loc->path,
+ mode, umask);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_mkdir_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mkdir,
- loc, mode, umask, xdata);
- return 0;
+ STACK_WIND(frame, trace_mkdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
+ return 0;
}
int
-trace_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
- dict_t *xdata)
+trace_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_UNLINK].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s flag=%d",
- frame->root->unique,
- uuid_utoa (loc->inode->gfid), loc->path,
- xflag);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_UNLINK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string), "%" PRId64 ": gfid=%s path=%s flag=%d",
+ frame->root->unique, uuid_utoa(loc->inode->gfid), loc->path,
+ xflag);
- frame->local = loc->inode->gfid;
+ frame->local = loc->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_unlink_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->unlink,
- loc, xflag, xdata);
- return 0;
+ STACK_WIND(frame, trace_unlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
+ return 0;
}
int
-trace_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
- dict_t *xdata)
+trace_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_RMDIR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s flags=%d",
- frame->root->unique,
- uuid_utoa (loc->inode->gfid), loc->path,
- flags);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_RMDIR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s path=%s flags=%d", frame->root->unique,
+ uuid_utoa(loc->inode->gfid), loc->path, flags);
- frame->local = loc->inode->gfid;
+ frame->local = loc->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_rmdir_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rmdir,
- loc, flags, xdata);
+ STACK_WIND(frame, trace_rmdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rmdir, loc, flags, xdata);
- return 0;
+ return 0;
}
int
-trace_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
- loc_t *loc, mode_t umask, dict_t *xdata)
+trace_symlink(call_frame_t *frame, xlator_t *this, const char *linkpath,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_SYMLINK].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s linkpath=%s, path=%s"
- " umask=0%o", frame->root->unique,
- uuid_utoa (loc->inode->gfid), linkpath,
- loc->path, umask);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_SYMLINK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s linkpath=%s, path=%s"
+ " umask=0%o",
+ frame->root->unique, uuid_utoa(loc->inode->gfid), linkpath,
+ loc->path, umask);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_symlink_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->symlink,
- linkpath, loc, umask, xdata);
+ STACK_WIND(frame, trace_symlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->symlink, linkpath, loc, umask, xdata);
- return 0;
+ return 0;
}
int
-trace_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
+trace_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- char oldgfid[50] = {0,};
- char newgfid[50] = {0,};
- trace_conf_t *conf = NULL;
+ char oldgfid[50] = {
+ 0,
+ };
+ char newgfid[50] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_RENAME].enabled) {
- char string[4096] = {0,};
- if (newloc->inode)
- uuid_utoa_r (newloc->inode->gfid, newgfid);
- else
- strcpy (newgfid, "0");
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_RENAME].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (newloc->inode)
+ uuid_utoa_r(newloc->inode->gfid, newgfid);
+ else
+ strcpy(newgfid, "0");
- uuid_utoa_r (oldloc->inode->gfid, oldgfid);
+ uuid_utoa_r(oldloc->inode->gfid, oldgfid);
- snprintf (string, sizeof (string),
- "%"PRId64": oldgfid=%s oldpath=%s --> "
- "newgfid=%s newpath=%s",
- frame->root->unique, oldgfid,
- oldloc->path, newgfid, newloc->path);
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": oldgfid=%s oldpath=%s --> "
+ "newgfid=%s newpath=%s",
+ frame->root->unique, oldgfid, oldloc->path, newgfid,
+ newloc->path);
- frame->local = oldloc->inode->gfid;
+ frame->local = oldloc->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_rename_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rename,
- oldloc, newloc, xdata);
+ STACK_WIND(frame, trace_rename_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
- return 0;
+ return 0;
}
int
-trace_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
+trace_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- char oldgfid[50] = {0,};
- char newgfid[50] = {0,};
- trace_conf_t *conf = NULL;
+ char oldgfid[50] = {
+ 0,
+ };
+ char newgfid[50] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_LINK].enabled) {
- char string[4096] = {0,};
- if (newloc->inode)
- uuid_utoa_r (newloc->inode->gfid, newgfid);
- else
- strcpy (newgfid, "0");
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_LINK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (newloc->inode)
+ uuid_utoa_r(newloc->inode->gfid, newgfid);
+ else
+ strcpy(newgfid, "0");
- uuid_utoa_r (oldloc->inode->gfid, oldgfid);
+ uuid_utoa_r(oldloc->inode->gfid, oldgfid);
- snprintf (string, sizeof (string),
- "%"PRId64": oldgfid=%s oldpath=%s --> "
- "newgfid=%s newpath=%s", frame->root->unique,
- oldgfid, oldloc->path, newgfid,
- newloc->path);
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": oldgfid=%s oldpath=%s --> "
+ "newgfid=%s newpath=%s",
+ frame->root->unique, oldgfid, oldloc->path, newgfid,
+ newloc->path);
- frame->local = oldloc->inode->gfid;
+ frame->local = oldloc->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_link_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->link,
- oldloc, newloc, xdata);
- return 0;
+ STACK_WIND(frame, trace_link_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
+ return 0;
}
int
-trace_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+trace_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- char actime_str[256] = {0,};
- char modtime_str[256] = {0,};
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_SETATTR].enabled) {
- char string[4096] = {0,};
- if (valid & GF_SET_ATTR_MODE) {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s mode=%o)",
- frame->root->unique,
- uuid_utoa (loc->inode->gfid),
- loc->path,
- st_mode_from_ia (stbuf->ia_prot,
- stbuf->ia_type));
-
- LOG_ELEMENT (conf, string);
- memset (string, 0 , sizeof (string));
- }
-
- if (valid & (GF_SET_ATTR_UID | GF_SET_ATTR_GID)) {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s uid=%o,"
- " gid=%o", frame->root->unique,
- uuid_utoa (loc->inode->gfid),
- loc->path, stbuf->ia_uid,
- stbuf->ia_gid);
-
- LOG_ELEMENT (conf, string);
- memset (string, 0 , sizeof (string));
- }
-
- if (valid & (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME)) {
- gf_time_fmt (actime_str, sizeof actime_str,
- stbuf->ia_atime, gf_timefmt_bdT);
-
- gf_time_fmt (modtime_str, sizeof modtime_str,
- stbuf->ia_mtime, gf_timefmt_bdT);
-
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s "
- "ia_atime=%s, ia_mtime=%s",
- frame->root->unique,
- uuid_utoa (loc->inode->gfid),
- loc->path, actime_str, modtime_str);
-
- LOG_ELEMENT (conf, string);
- memset (string, 0 , sizeof (string));
- }
- frame->local = loc->inode->gfid;
+ char actime_str[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ char modtime_str[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_SETATTR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (valid & GF_SET_ATTR_MODE) {
+ snprintf(
+ string, sizeof(string), "%" PRId64 ": gfid=%s path=%s mode=%o)",
+ frame->root->unique, uuid_utoa(loc->inode->gfid), loc->path,
+ st_mode_from_ia(stbuf->ia_prot, stbuf->ia_type));
+
+ LOG_ELEMENT(conf, string);
+ memset(string, 0, sizeof(string));
+ }
+
+ if (valid & (GF_SET_ATTR_UID | GF_SET_ATTR_GID)) {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s path=%s uid=%o,"
+ " gid=%o",
+ frame->root->unique, uuid_utoa(loc->inode->gfid),
+ loc->path, stbuf->ia_uid, stbuf->ia_gid);
+
+ LOG_ELEMENT(conf, string);
+ memset(string, 0, sizeof(string));
+ }
+
+ if (valid & (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME)) {
+ gf_time_fmt(actime_str, sizeof actime_str, stbuf->ia_atime,
+ gf_timefmt_bdT);
+
+ gf_time_fmt(modtime_str, sizeof modtime_str, stbuf->ia_mtime,
+ gf_timefmt_bdT);
+
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s path=%s "
+ "ia_atime=%s, ia_mtime=%s",
+ frame->root->unique, uuid_utoa(loc->inode->gfid),
+ loc->path, actime_str, modtime_str);
+
+ LOG_ELEMENT(conf, string);
+ memset(string, 0, sizeof(string));
}
+ frame->local = loc->inode->gfid;
+ }
out:
- STACK_WIND (frame, trace_setattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setattr,
- loc, stbuf, valid, xdata);
+ STACK_WIND(frame, trace_setattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid, xdata);
- return 0;
+ return 0;
}
int
-trace_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+trace_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- char actime_str[256] = {0,};
- char modtime_str[256] = {0,};
- trace_conf_t *conf = NULL;
+ char actime_str[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ char modtime_str[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FSETATTR].enabled) {
- char string[4096] = {0,};
- if (valid & GF_SET_ATTR_MODE) {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s fd=%p, mode=%o",
- frame->root->unique,
- uuid_utoa (fd->inode->gfid), fd,
- st_mode_from_ia (stbuf->ia_prot,
- stbuf->ia_type));
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FSETATTR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ if (valid & GF_SET_ATTR_MODE) {
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s fd=%p, mode=%o", frame->root->unique,
+ uuid_utoa(fd->inode->gfid), fd,
+ st_mode_from_ia(stbuf->ia_prot, stbuf->ia_type));
- LOG_ELEMENT (conf, string);
- memset (string, 0, sizeof (string));
- }
+ LOG_ELEMENT(conf, string);
+ memset(string, 0, sizeof(string));
+ }
- if (valid & (GF_SET_ATTR_UID | GF_SET_ATTR_GID)) {
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s fd=%p, uid=%o, "
- "gid=%o", frame->root->unique,
- uuid_utoa (fd->inode->gfid),
- fd, stbuf->ia_uid, stbuf->ia_gid);
+ if (valid & (GF_SET_ATTR_UID | GF_SET_ATTR_GID)) {
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s fd=%p, uid=%o, "
+ "gid=%o",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), fd,
+ stbuf->ia_uid, stbuf->ia_gid);
- LOG_ELEMENT (conf, string);
- memset (string, 0, sizeof (string));
- }
+ LOG_ELEMENT(conf, string);
+ memset(string, 0, sizeof(string));
+ }
- if (valid & (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME)) {
- gf_time_fmt (actime_str, sizeof actime_str,
- stbuf->ia_atime, gf_timefmt_bdT);
+ if (valid & (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME)) {
+ gf_time_fmt(actime_str, sizeof actime_str, stbuf->ia_atime,
+ gf_timefmt_bdT);
- gf_time_fmt (modtime_str, sizeof modtime_str,
- stbuf->ia_mtime, gf_timefmt_bdT);
+ gf_time_fmt(modtime_str, sizeof modtime_str, stbuf->ia_mtime,
+ gf_timefmt_bdT);
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s fd=%p "
- "ia_atime=%s, ia_mtime=%s",
- frame->root->unique,
- uuid_utoa (fd->inode->gfid),
- fd, actime_str, modtime_str);
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s fd=%p "
+ "ia_atime=%s, ia_mtime=%s",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), fd,
+ actime_str, modtime_str);
- LOG_ELEMENT (conf, string);
- memset (string, 0, sizeof (string));
- }
- frame->local = fd->inode->gfid;
+ LOG_ELEMENT(conf, string);
+ memset(string, 0, sizeof(string));
}
+ frame->local = fd->inode->gfid;
+ }
out:
- STACK_WIND (frame, trace_fsetattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetattr,
- fd, stbuf, valid, xdata);
+ STACK_WIND(frame, trace_fsetattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetattr, fd, stbuf, valid, xdata);
- return 0;
+ return 0;
}
static int
-trace_seek_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, off_t offset, dict_t *xdata)
-{
- trace_conf_t *conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_SEEK].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s op_ret=%d op_errno=%d, "
- "offset=%"PRId64"", frame->root->unique,
- uuid_utoa (frame->local), op_ret, op_errno, offset);
- LOG_ELEMENT (conf, string);
- }
-out:
- TRACE_STACK_UNWIND (seek, frame, op_ret, op_errno, offset, xdata);
- return 0;
+trace_seek_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, off_t offset, dict_t *xdata)
+{
+ trace_conf_t *conf = this->private;
+
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_SEEK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s op_ret=%d op_errno=%d, "
+ "offset=%" PRId64 "",
+ frame->root->unique, uuid_utoa(frame->local), op_ret, op_errno,
+ offset);
+ LOG_ELEMENT(conf, string);
+ }
+out:
+ TRACE_STACK_UNWIND(seek, frame, op_ret, op_errno, offset, xdata);
+ return 0;
}
static int
-trace_seek (call_frame_t *frame, xlator_t *this, fd_t *fd,
- off_t offset, gf_seek_what_t what, dict_t *xdata)
+trace_seek(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ gf_seek_what_t what, dict_t *xdata)
{
- trace_conf_t *conf = this->private;
+ trace_conf_t *conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_SEEK].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string), "%"PRId64": gfid=%s fd=%p "
- "offset=%"PRId64" what=%d", frame->root->unique,
- uuid_utoa (fd->inode->gfid), fd, offset, what);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_SEEK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s fd=%p "
+ "offset=%" PRId64 " what=%d",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), fd, offset,
+ what);
- frame->local = fd->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ frame->local = fd->inode->gfid;
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_seek_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->seek, fd, offset, what, xdata);
+ STACK_WIND(frame, trace_seek_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->seek, fd, offset, what, xdata);
- return 0;
+ return 0;
}
int
-trace_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc,
- off_t offset, dict_t *xdata)
+trace_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_TRUNCATE].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s, "
- "offset=%"PRId64"", frame->root->unique,
- uuid_utoa (loc->inode->gfid), loc->path,
- offset);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_TRUNCATE].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s path=%s, "
+ "offset=%" PRId64 "",
+ frame->root->unique, uuid_utoa(loc->inode->gfid), loc->path,
+ offset);
- frame->local = loc->inode->gfid;
+ frame->local = loc->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_truncate_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->truncate,
- loc, offset, xdata);
+ STACK_WIND(frame, trace_truncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
- return 0;
+ return 0;
}
int
-trace_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int32_t flags, fd_t *fd, dict_t *xdata)
+trace_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_OPEN].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s flags=%d fd=%p",
- frame->root->unique,
- uuid_utoa (loc->inode->gfid), loc->path,
- flags, fd);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_OPEN].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s path=%s flags=%d fd=%p",
+ frame->root->unique, uuid_utoa(loc->inode->gfid), loc->path,
+ flags, fd);
- frame->local = loc->inode->gfid;
+ frame->local = loc->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_open_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->open,
- loc, flags, fd, xdata);
- return 0;
+ STACK_WIND(frame, trace_open_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
+ return 0;
}
int
-trace_create (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int32_t flags, mode_t mode, mode_t umask, fd_t *fd,
- dict_t *xdata)
+trace_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_CREATE].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s, fd=%p, "
- "flags=0%o mode=0%o umask=0%o",
- frame->root->unique,
- uuid_utoa (loc->inode->gfid), loc->path,
- fd, flags, mode, umask);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_CREATE].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s path=%s, fd=%p, "
+ "flags=0%o mode=0%o umask=0%o",
+ frame->root->unique, uuid_utoa(loc->inode->gfid), loc->path,
+ fd, flags, mode, umask);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_create_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->create,
- loc, flags, mode, umask, fd, xdata);
- return 0;
+ STACK_WIND(frame, trace_create_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, flags, mode, umask, fd,
+ xdata);
+ return 0;
}
int
-trace_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
- size_t size, off_t offset, uint32_t flags, dict_t *xdata)
+trace_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_READ].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s fd=%p, size=%"
- GF_PRI_SIZET"offset=%"PRId64" flags=0%x)",
- frame->root->unique,
- uuid_utoa (fd->inode->gfid), fd, size,
- offset, flags);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_READ].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s fd=%p, size=%" GF_PRI_SIZET
+ "offset=%" PRId64 " flags=0%x)",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), fd, size,
+ offset, flags);
- frame->local = fd->inode->gfid;
+ frame->local = fd->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_readv_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readv,
- fd, size, offset, flags, xdata);
- return 0;
+ STACK_WIND(frame, trace_readv_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readv, fd, size, offset, flags, xdata);
+ return 0;
}
int
-trace_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count,
- off_t offset, uint32_t flags, struct iobref *iobref, dict_t *xdata)
+trace_writev(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int32_t count, off_t offset, uint32_t flags,
+ struct iobref *iobref, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
- int i = 0;
- size_t total_size = 0;
+ trace_conf_t *conf = NULL;
+ int i = 0;
+ size_t total_size = 0;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_WRITE].enabled) {
- char string[4096] = {0,};
- for (i = 0; i < count; i++)
- total_size += vector[i].iov_len;
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_WRITE].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ for (i = 0; i < count; i++)
+ total_size += vector[i].iov_len;
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s fd=%p, count=%d, "
- " offset=%"PRId64" flags=0%x write_size=%zu",
- frame->root->unique,
- uuid_utoa (fd->inode->gfid), fd, count,
- offset, flags, total_size);
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s fd=%p, count=%d, "
+ " offset=%" PRId64 " flags=0%x write_size=%zu",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), fd, count,
+ offset, flags, total_size);
- frame->local = fd->inode->gfid;
+ frame->local = fd->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_writev_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->writev,
- fd, vector, count, offset, flags, iobref, xdata);
- return 0;
+ STACK_WIND(frame, trace_writev_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector, count, offset,
+ flags, iobref, xdata);
+ return 0;
}
int
-trace_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+trace_statfs(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_STATFS].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s",
- frame->root->unique, (loc->inode)?
- uuid_utoa (loc->inode->gfid):"0", loc->path);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_STATFS].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string), "%" PRId64 ": gfid=%s path=%s",
+ frame->root->unique,
+ (loc->inode) ? uuid_utoa(loc->inode->gfid) : "0", loc->path);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_statfs_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->statfs,
- loc, xdata);
- return 0;
+ STACK_WIND(frame, trace_statfs_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->statfs, loc, xdata);
+ return 0;
}
int
-trace_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+trace_flush(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FLUSH].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s fd=%p",
- frame->root->unique,
- uuid_utoa (fd->inode->gfid), fd);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FLUSH].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string), "%" PRId64 ": gfid=%s fd=%p",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), fd);
- frame->local = fd->inode->gfid;
+ frame->local = fd->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_flush_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->flush,
- fd, xdata);
- return 0;
+ STACK_WIND(frame, trace_flush_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->flush, fd, xdata);
+ return 0;
}
int
-trace_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
- dict_t *xdata)
+trace_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+ dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FSYNC].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s flags=%d fd=%p",
- frame->root->unique,
- uuid_utoa (fd->inode->gfid), flags, fd);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FSYNC].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string), "%" PRId64 ": gfid=%s flags=%d fd=%p",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), flags, fd);
- frame->local = fd->inode->gfid;
+ frame->local = fd->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_fsync_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsync,
- fd, flags, xdata);
- return 0;
+ STACK_WIND(frame, trace_fsync_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsync, fd, flags, xdata);
+ return 0;
}
int
-trace_setxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata)
+trace_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_SETXATTR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s flags=%d",
- frame->root->unique,
- uuid_utoa (loc->inode->gfid), loc->path,
- flags);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_SETXATTR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s path=%s flags=%d", frame->root->unique,
+ uuid_utoa(loc->inode->gfid), loc->path, flags);
- frame->local = loc->inode->gfid;
+ frame->local = loc->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_setxattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setxattr,
- loc, dict, flags, xdata);
- return 0;
+ STACK_WIND(frame, trace_setxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, loc, dict, flags, xdata);
+ return 0;
}
int
-trace_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
+trace_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_GETXATTR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s name=%s",
- frame->root->unique,
- uuid_utoa (loc->inode->gfid), loc->path,
- name);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_GETXATTR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string), "%" PRId64 ": gfid=%s path=%s name=%s",
+ frame->root->unique, uuid_utoa(loc->inode->gfid), loc->path,
+ name);
- frame->local = loc->inode->gfid;
+ frame->local = loc->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_getxattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr,
- loc, name, xdata);
- return 0;
+ STACK_WIND(frame, trace_getxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+ return 0;
}
int
-trace_removexattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
+trace_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_REMOVEXATTR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s name=%s",
- frame->root->unique,
- uuid_utoa (loc->inode->gfid), loc->path,
- name);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_REMOVEXATTR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string), "%" PRId64 ": gfid=%s path=%s name=%s",
+ frame->root->unique, uuid_utoa(loc->inode->gfid), loc->path,
+ name);
- frame->local = loc->inode->gfid;
+ frame->local = loc->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_removexattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->removexattr,
- loc, name, xdata);
+ STACK_WIND(frame, trace_removexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
- return 0;
+ return 0;
}
int
-trace_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
- dict_t *xdata)
+trace_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_OPENDIR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s fd=%p",
- frame->root->unique,
- uuid_utoa (loc->inode->gfid), loc->path, fd);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_OPENDIR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string), "%" PRId64 ": gfid=%s path=%s fd=%p",
+ frame->root->unique, uuid_utoa(loc->inode->gfid), loc->path,
+ fd);
- frame->local = loc->inode->gfid;
+ frame->local = loc->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_opendir_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->opendir,
- loc, fd, xdata);
- return 0;
+ STACK_WIND(frame, trace_opendir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->opendir, loc, fd, xdata);
+ return 0;
}
int
-trace_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, dict_t *dict)
+trace_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *dict)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_READDIRP].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s fd=%p, size=%"GF_PRI_SIZET
- ", offset=%"PRId64" dict=%p",
- frame->root->unique,
- uuid_utoa (fd->inode->gfid), fd, size,
- offset, dict);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_READDIRP].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s fd=%p, size=%" GF_PRI_SIZET
+ ", offset=%" PRId64 " dict=%p",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), fd, size,
+ offset, dict);
- frame->local = fd->inode->gfid;
+ frame->local = fd->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_readdirp_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdirp,
- fd, size, offset, dict);
+ STACK_WIND(frame, trace_readdirp_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdirp, fd, size, offset, dict);
- return 0;
+ return 0;
}
int
-trace_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
- size_t size, off_t offset, dict_t *xdata)
+trace_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_READDIR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s fd=%p, size=%"GF_PRI_SIZET
- ", offset=%"PRId64,
- frame->root->unique,
- uuid_utoa (fd->inode->gfid), fd, size,
- offset);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_READDIR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s fd=%p, size=%" GF_PRI_SIZET
+ ", offset=%" PRId64,
+ frame->root->unique, uuid_utoa(fd->inode->gfid), fd, size,
+ offset);
- frame->local = fd->inode->gfid;
+ frame->local = fd->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_readdir_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdir,
- fd, size, offset, xdata);
+ STACK_WIND(frame, trace_readdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdir, fd, size, offset, xdata);
- return 0;
+ return 0;
}
int
-trace_fsyncdir (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int32_t datasync, dict_t *xdata)
+trace_fsyncdir(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+ dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FSYNCDIR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s datasync=%d fd=%p",
- frame->root->unique,
- uuid_utoa (fd->inode->gfid), datasync, fd);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FSYNCDIR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s datasync=%d fd=%p", frame->root->unique,
+ uuid_utoa(fd->inode->gfid), datasync, fd);
- frame->local = fd->inode->gfid;
+ frame->local = fd->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_fsyncdir_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsyncdir,
- fd, datasync, xdata);
- return 0;
+ STACK_WIND(frame, trace_fsyncdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsyncdir, fd, datasync, xdata);
+ return 0;
}
int
-trace_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
- dict_t *xdata)
+trace_access(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+ dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_ACCESS].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s path=%s mask=0%o",
- frame->root->unique,
- uuid_utoa (loc->inode->gfid),
- loc->path, mask);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_ACCESS].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s path=%s mask=0%o", frame->root->unique,
+ uuid_utoa(loc->inode->gfid), loc->path, mask);
- frame->local = loc->inode->gfid;
+ frame->local = loc->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_access_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->access,
- loc, mask, xdata);
- return 0;
+ STACK_WIND(frame, trace_access_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->access, loc, mask, xdata);
+ return 0;
}
int32_t
-trace_rchecksum (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- int32_t len, dict_t *xdata)
+trace_rchecksum(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ int32_t len, dict_t *xdata)
{
+ trace_conf_t *conf = NULL;
- trace_conf_t *conf = NULL;
-
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_RCHECKSUM].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s offset=%"PRId64
- "len=%u fd=%p", frame->root->unique,
- uuid_utoa (fd->inode->gfid), offset, len, fd);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_RCHECKSUM].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s offset=%" PRId64 "len=%u fd=%p",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), offset, len,
+ fd);
- frame->local = fd->inode->gfid;
+ frame->local = fd->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_rchecksum_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rchecksum,
- fd, offset, len, xdata);
-
- return 0;
+ STACK_WIND(frame, trace_rchecksum_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rchecksum, fd, offset, len, xdata);
+ return 0;
}
int32_t
-trace_fentrylk (call_frame_t *frame, xlator_t *this, const char *volume,
- fd_t *fd, const char *basename, entrylk_cmd cmd,
- entrylk_type type, dict_t *xdata)
-{
- trace_conf_t *conf = NULL;
-
- conf = this->private;
-
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FENTRYLK].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s volume=%s, (fd=%p "
- "basename=%s, cmd=%s, type=%s)",
- frame->root->unique,
- uuid_utoa (fd->inode->gfid), volume, fd,
- basename,
- ((cmd == ENTRYLK_LOCK) ? "ENTRYLK_LOCK" :
- "ENTRYLK_UNLOCK"),
- ((type == ENTRYLK_RDLCK) ? "ENTRYLK_RDLCK" :
- "ENTRYLK_WRLCK"));
-
- frame->local = fd->inode->gfid;
-
- LOG_ELEMENT (conf, string);
- }
+trace_fentrylk(call_frame_t *frame, xlator_t *this, const char *volume,
+ fd_t *fd, const char *basename, entrylk_cmd cmd,
+ entrylk_type type, dict_t *xdata)
+{
+ trace_conf_t *conf = NULL;
-out:
- STACK_WIND (frame, trace_fentrylk_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fentrylk,
- volume, fd, basename, cmd, type, xdata);
- return 0;
+ conf = this->private;
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FENTRYLK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s volume=%s, (fd=%p "
+ "basename=%s, cmd=%s, type=%s)",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), volume, fd,
+ basename,
+ ((cmd == ENTRYLK_LOCK) ? "ENTRYLK_LOCK" : "ENTRYLK_UNLOCK"),
+ ((type == ENTRYLK_RDLCK) ? "ENTRYLK_RDLCK" : "ENTRYLK_WRLCK"));
+
+ frame->local = fd->inode->gfid;
+
+ LOG_ELEMENT(conf, string);
+ }
+
+out:
+ STACK_WIND(frame, trace_fentrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fentrylk, volume, fd, basename, cmd,
+ type, xdata);
+ return 0;
}
int32_t
-trace_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
+trace_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
+ dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FGETXATTR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s fd=%p name=%s",
- frame->root->unique,
- uuid_utoa (fd->inode->gfid), fd, name);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FGETXATTR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string), "%" PRId64 ": gfid=%s fd=%p name=%s",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), fd, name);
- frame->local = fd->inode->gfid;
+ frame->local = fd->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_fgetxattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fgetxattr,
- fd, name, xdata);
- return 0;
+ STACK_WIND(frame, trace_fgetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
+ return 0;
}
int32_t
-trace_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- dict_t *dict, int32_t flags, dict_t *xdata)
+trace_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FSETXATTR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s fd=%p flags=%d",
- frame->root->unique,
- uuid_utoa (fd->inode->gfid), fd, flags);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FSETXATTR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string), "%" PRId64 ": gfid=%s fd=%p flags=%d",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), fd, flags);
- frame->local = fd->inode->gfid;
+ frame->local = fd->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_fsetxattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetxattr,
- fd, dict, flags, xdata);
- return 0;
+ STACK_WIND(frame, trace_fsetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+ return 0;
}
int
-trace_ftruncate (call_frame_t *frame, xlator_t *this,
- fd_t *fd, off_t offset, dict_t *xdata)
+trace_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FTRUNCATE].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s offset=%"PRId64" fd=%p",
- frame->root->unique,
- uuid_utoa (fd->inode->gfid), offset, fd);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FTRUNCATE].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64 ": gfid=%s offset=%" PRId64 " fd=%p",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), offset, fd);
- frame->local = fd->inode->gfid;
+ frame->local = fd->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_ftruncate_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->ftruncate,
- fd, offset, xdata);
+ STACK_WIND(frame, trace_ftruncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
- return 0;
+ return 0;
}
int
-trace_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+trace_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_FSTAT].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s fd=%p",
- frame->root->unique,
- uuid_utoa (fd->inode->gfid), fd);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_FSTAT].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string), "%" PRId64 ": gfid=%s fd=%p",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), fd);
- frame->local = fd->inode->gfid;
+ frame->local = fd->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_fstat_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fstat,
- fd, xdata);
- return 0;
+ STACK_WIND(frame, trace_fstat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fstat, fd, xdata);
+ return 0;
}
int
-trace_lk (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int32_t cmd, struct gf_flock *lock, dict_t *xdata)
+trace_lk(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
+ struct gf_flock *lock, dict_t *xdata)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_LK].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "%"PRId64": gfid=%s fd=%p, cmd=%d, "
- "lock {l_type=%d, "
- "l_whence=%d, l_start=%"PRId64", "
- "l_len=%"PRId64", l_pid=%u})",
- frame->root->unique,
- uuid_utoa (fd->inode->gfid), fd, cmd,
- lock->l_type, lock->l_whence,
- lock->l_start, lock->l_len, lock->l_pid);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_LK].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string),
+ "%" PRId64
+ ": gfid=%s fd=%p, cmd=%d, "
+ "lock {l_type=%d, "
+ "l_whence=%d, l_start=%" PRId64
+ ", "
+ "l_len=%" PRId64 ", l_pid=%u})",
+ frame->root->unique, uuid_utoa(fd->inode->gfid), fd, cmd,
+ lock->l_type, lock->l_whence, lock->l_start, lock->l_len,
+ lock->l_pid);
- frame->local = fd->inode->gfid;
+ frame->local = fd->inode->gfid;
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- STACK_WIND (frame, trace_lk_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lk,
- fd, cmd, lock, xdata);
- return 0;
+ STACK_WIND(frame, trace_lk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lk, fd, cmd, lock, xdata);
+ return 0;
}
int32_t
-trace_forget (xlator_t *this, inode_t *inode)
+trace_forget(xlator_t *this, inode_t *inode)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
- /* If user want to understand when a lookup happens,
- he should know about 'forget' too */
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_LOOKUP].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "gfid=%s", uuid_utoa (inode->gfid));
+ conf = this->private;
+ /* If user want to understand when a lookup happens,
+ he should know about 'forget' too */
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_LOOKUP].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string), "gfid=%s", uuid_utoa(inode->gfid));
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- return 0;
+ return 0;
}
int32_t
-trace_releasedir (xlator_t *this, fd_t *fd)
+trace_releasedir(xlator_t *this, fd_t *fd)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_OPENDIR].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "gfid=%s fd=%p",
- uuid_utoa (fd->inode->gfid), fd);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_OPENDIR].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string), "gfid=%s fd=%p",
+ uuid_utoa(fd->inode->gfid), fd);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- return 0;
+ return 0;
}
int32_t
-trace_release (xlator_t *this, fd_t *fd)
+trace_release(xlator_t *this, fd_t *fd)
{
- trace_conf_t *conf = NULL;
+ trace_conf_t *conf = NULL;
- conf = this->private;
+ conf = this->private;
- if (!conf->log_file && !conf->log_history)
- goto out;
- if (trace_fop_names[GF_FOP_OPEN].enabled ||
- trace_fop_names[GF_FOP_CREATE].enabled) {
- char string[4096] = {0,};
- snprintf (string, sizeof (string),
- "gfid=%s fd=%p",
- uuid_utoa (fd->inode->gfid), fd);
+ if (!conf->log_file && !conf->log_history)
+ goto out;
+ if (trace_fop_names[GF_FOP_OPEN].enabled ||
+ trace_fop_names[GF_FOP_CREATE].enabled) {
+ char string[4096] = {
+ 0,
+ };
+ snprintf(string, sizeof(string), "gfid=%s fd=%p",
+ uuid_utoa(fd->inode->gfid), fd);
- LOG_ELEMENT (conf, string);
- }
+ LOG_ELEMENT(conf, string);
+ }
out:
- return 0;
+ return 0;
}
-
void
-enable_all_calls (int enabled)
+enable_all_calls(int enabled)
{
- int i;
+ int i;
- for (i = 0; i < GF_FOP_MAXVALUE; i++)
- trace_fop_names[i].enabled = enabled;
+ for (i = 0; i < GF_FOP_MAXVALUE; i++)
+ trace_fop_names[i].enabled = enabled;
}
void
-enable_call (const char *name, int enabled)
+enable_call(const char *name, int enabled)
{
- int i;
- for (i = 0; i < GF_FOP_MAXVALUE; i++)
- if (!strcasecmp(trace_fop_names[i].name, name))
- trace_fop_names[i].enabled = enabled;
+ int i;
+ for (i = 0; i < GF_FOP_MAXVALUE; i++)
+ if (!strcasecmp(trace_fop_names[i].name, name))
+ trace_fop_names[i].enabled = enabled;
}
-
/*
include = 1 for "include-ops"
= 0 for "exclude-ops"
*/
void
-process_call_list (const char *list, int include)
+process_call_list(const char *list, int include)
{
- enable_all_calls (include ? 0 : 1);
+ enable_all_calls(include ? 0 : 1);
- char *call = strsep ((char **)&list, ",");
+ char *call = strsep((char **)&list, ",");
- while (call) {
- enable_call (call, include);
- call = strsep ((char **)&list, ",");
- }
+ while (call) {
+ enable_call(call, include);
+ call = strsep((char **)&list, ",");
+ }
}
int32_t
-trace_dump_history (xlator_t *this)
-{
- int ret = -1;
- char key_prefix[GF_DUMP_MAX_BUF_LEN] = {0,};
- trace_conf_t *conf = NULL;
-
- GF_VALIDATE_OR_GOTO ("trace", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->history, out);
-
- conf = this->private;
- // Is it ok to return silently if log-history option his off?
- if (conf && conf->log_history == _gf_true) {
- gf_proc_dump_build_key (key_prefix, "xlator.debug.trace",
- "history");
- gf_proc_dump_add_section (key_prefix);
- eh_dump (this->history, NULL, dump_history_trace);
- }
- ret = 0;
+trace_dump_history(xlator_t *this)
+{
+ int ret = -1;
+ char key_prefix[GF_DUMP_MAX_BUF_LEN] = {
+ 0,
+ };
+ trace_conf_t *conf = NULL;
+
+ GF_VALIDATE_OR_GOTO("trace", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->history, out);
+
+ conf = this->private;
+ // Is it ok to return silently if log-history option his off?
+ if (conf && conf->log_history == _gf_true) {
+ gf_proc_dump_build_key(key_prefix, "xlator.debug.trace", "history");
+ gf_proc_dump_add_section("%s", key_prefix);
+ eh_dump(this->history, NULL, dump_history_trace);
+ }
+ ret = 0;
out:
- return ret;
+ return ret;
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- if (!this)
- return ret;
-
- ret = xlator_mem_acct_init (this, gf_trace_mt_end + 1);
+ if (!this)
+ return ret;
- if (ret != 0) {
- gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
- " failed");
- return ret;
- }
+ ret = xlator_mem_acct_init(this, gf_trace_mt_end + 1);
+ if (ret != 0) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Memory accounting init"
+ " failed");
return ret;
+ }
+
+ return ret;
}
int
-reconfigure (xlator_t *this, dict_t *options)
+reconfigure(xlator_t *this, dict_t *options)
{
- int32_t ret = -1;
- trace_conf_t *conf = NULL;
- char *includes = NULL, *excludes = NULL;
+ int32_t ret = -1;
+ trace_conf_t *conf = NULL;
+ char *includes = NULL, *excludes = NULL;
- GF_VALIDATE_OR_GOTO ("quick-read", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
- GF_VALIDATE_OR_GOTO (this->name, options, out);
+ GF_VALIDATE_OR_GOTO("quick-read", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
+ GF_VALIDATE_OR_GOTO(this->name, options, out);
- conf = this->private;
+ conf = this->private;
- includes = data_to_str (dict_get (options, "include-ops"));
- excludes = data_to_str (dict_get (options, "exclude-ops"));
+ includes = data_to_str(dict_get(options, "include-ops"));
+ excludes = data_to_str(dict_get(options, "exclude-ops"));
- {
- int i;
- for (i = 0; i < GF_FOP_MAXVALUE; i++) {
- if (gf_fop_list[i])
- strncpy (trace_fop_names[i].name,
- gf_fop_list[i],
- sizeof (trace_fop_names[i].name));
- else
- strncpy (trace_fop_names[i].name, ":0",
- sizeof (trace_fop_names[i].name));
- trace_fop_names[i].enabled = 1;
- trace_fop_names[i].name[sizeof (
- trace_fop_names[i].name) - 1] = 0;
- }
+ {
+ int i;
+ for (i = 0; i < GF_FOP_MAXVALUE; i++) {
+ if (gf_fop_list[i])
+ strncpy(trace_fop_names[i].name, gf_fop_list[i],
+ sizeof(trace_fop_names[i].name));
+ else
+ strncpy(trace_fop_names[i].name, ":0",
+ sizeof(trace_fop_names[i].name));
+ trace_fop_names[i].enabled = 1;
+ trace_fop_names[i].name[sizeof(trace_fop_names[i].name) - 1] = 0;
}
+ }
- if (includes && excludes) {
- gf_log (this->name,
- GF_LOG_ERROR,
- "must specify only one of 'include-ops' and "
- "'exclude-ops'");
- goto out;
- }
+ if (includes && excludes) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "must specify only one of 'include-ops' and "
+ "'exclude-ops'");
+ goto out;
+ }
- if (includes)
- process_call_list (includes, 1);
- if (excludes)
- process_call_list (excludes, 0);
+ if (includes)
+ process_call_list(includes, 1);
+ if (excludes)
+ process_call_list(excludes, 0);
- /* Should resizing of the event-history be allowed in reconfigure?
- * for which a new event_history might have to be allocated and the
- * older history has to be freed.
- */
- GF_OPTION_RECONF ("log-file", conf->log_file, options, bool, out);
+ /* Should resizing of the event-history be allowed in reconfigure?
+ * for which a new event_history might have to be allocated and the
+ * older history has to be freed.
+ */
+ GF_OPTION_RECONF("log-file", conf->log_file, options, bool, out);
- GF_OPTION_RECONF ("log-history", conf->log_history, options, bool, out);
+ GF_OPTION_RECONF("log-history", conf->log_history, options, bool, out);
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int32_t
-init (xlator_t *this)
-{
- dict_t *options = NULL;
- char *includes = NULL, *excludes = NULL;
- char *forced_loglevel = NULL;
- eh_t *history = NULL;
- int ret = -1;
- size_t history_size = TRACE_DEFAULT_HISTORY_SIZE;
- trace_conf_t *conf = NULL;
-
- if (!this)
- return -1;
-
- if (!this->children || this->children->next) {
- gf_log (this->name, GF_LOG_ERROR,
- "trace translator requires one subvolume");
- return -1;
- }
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "dangling volume. check volfile ");
- }
-
- conf = GF_CALLOC (1, sizeof (trace_conf_t), gf_trace_mt_trace_conf_t);
- if (!conf) {
- gf_log (this->name, GF_LOG_ERROR, "cannot allocate "
- "xl->private");
- return -1;
- }
-
- options = this->options;
- includes = data_to_str (dict_get (options, "include-ops"));
- excludes = data_to_str (dict_get (options, "exclude-ops"));
-
- {
- int i;
- for (i = 0; i < GF_FOP_MAXVALUE; i++) {
- if (gf_fop_list[i])
- strncpy (trace_fop_names[i].name,
- gf_fop_list[i],
- sizeof (trace_fop_names[i].name));
- else
- strncpy (trace_fop_names[i].name, ":O",
- sizeof (trace_fop_names[i].name));
- trace_fop_names[i].enabled = 1;
- trace_fop_names[i].name[sizeof (
- trace_fop_names[i].name) - 1] = 0;
- }
- }
-
- if (includes && excludes) {
- gf_log (this->name,
- GF_LOG_ERROR,
- "must specify only one of 'include-ops' and "
- "'exclude-ops'");
- return -1;
- }
-
- if (includes)
- process_call_list (includes, 1);
- if (excludes)
- process_call_list (excludes, 0);
-
-
- GF_OPTION_INIT ("history-size", conf->history_size, size, out);
-
- gf_log (this->name, GF_LOG_INFO, "history size %"GF_PRI_SIZET,
- history_size);
-
- GF_OPTION_INIT ("log-file", conf->log_file, bool, out);
-
- gf_log (this->name, GF_LOG_INFO, "logging to file %s",
- (conf->log_file == _gf_true)?"enabled":"disabled");
-
- GF_OPTION_INIT ("log-history", conf->log_history, bool, out);
-
- gf_log (this->name, GF_LOG_DEBUG, "logging to history %s",
- (conf->log_history == _gf_true)?"enabled":"disabled");
-
- history = eh_new (history_size, _gf_false, NULL);
- if (!history) {
- gf_log (this->name, GF_LOG_ERROR, "event history cannot be "
- "initialized");
- return -1;
- }
-
- this->history = history;
-
- conf->trace_log_level = GF_LOG_INFO;
-
- if (dict_get (options, "force-log-level")) {
- forced_loglevel = data_to_str (dict_get (options,
- "force-log-level"));
- if (!forced_loglevel)
- goto setloglevel;
-
- if (strcmp (forced_loglevel, "INFO") == 0)
- conf->trace_log_level = GF_LOG_INFO;
- else if (strcmp (forced_loglevel, "TRACE") == 0)
- conf->trace_log_level = GF_LOG_TRACE;
- else if (strcmp (forced_loglevel, "ERROR") == 0)
- conf->trace_log_level = GF_LOG_ERROR;
- else if (strcmp (forced_loglevel, "DEBUG") == 0)
- conf->trace_log_level = GF_LOG_DEBUG;
- else if (strcmp (forced_loglevel, "WARNING") == 0)
- conf->trace_log_level = GF_LOG_WARNING;
- else if (strcmp (forced_loglevel, "CRITICAL") == 0)
- conf->trace_log_level = GF_LOG_CRITICAL;
- else if (strcmp (forced_loglevel, "NONE") == 0)
- conf->trace_log_level = GF_LOG_NONE;
- }
+init(xlator_t *this)
+{
+ dict_t *options = NULL;
+ char *includes = NULL, *excludes = NULL;
+ char *forced_loglevel = NULL;
+ eh_t *history = NULL;
+ int ret = -1;
+ uint64_t history_size = TRACE_DEFAULT_HISTORY_SIZE;
+ trace_conf_t *conf = NULL;
+
+ if (!this)
+ return -1;
+
+ if (!this->children || this->children->next) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "trace translator requires one subvolume");
+ return -1;
+ }
+ if (!this->parents) {
+ gf_log(this->name, GF_LOG_WARNING, "dangling volume. check volfile ");
+ }
+
+ conf = GF_CALLOC(1, sizeof(trace_conf_t), gf_trace_mt_trace_conf_t);
+ if (!conf) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "cannot allocate "
+ "xl->private");
+ return -1;
+ }
+
+ options = this->options;
+ includes = data_to_str(dict_get(options, "include-ops"));
+ excludes = data_to_str(dict_get(options, "exclude-ops"));
+
+ {
+ int i;
+ for (i = 0; i < GF_FOP_MAXVALUE; i++) {
+ if (gf_fop_list[i])
+ strncpy(trace_fop_names[i].name, gf_fop_list[i],
+ sizeof(trace_fop_names[i].name));
+ else
+ strncpy(trace_fop_names[i].name, ":O",
+ sizeof(trace_fop_names[i].name));
+ trace_fop_names[i].enabled = 1;
+ trace_fop_names[i].name[sizeof(trace_fop_names[i].name) - 1] = 0;
+ }
+ }
+
+ if (includes && excludes) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "must specify only one of 'include-ops' and "
+ "'exclude-ops'");
+ return -1;
+ }
+
+ if (includes)
+ process_call_list(includes, 1);
+ if (excludes)
+ process_call_list(excludes, 0);
+
+ GF_OPTION_INIT("history-size", history_size, size, out);
+ conf->history_size = history_size;
+
+ gf_log(this->name, GF_LOG_INFO, "history size %" PRIu64, history_size);
+
+ GF_OPTION_INIT("log-file", conf->log_file, bool, out);
+
+ gf_log(this->name, GF_LOG_INFO, "logging to file %s",
+ (conf->log_file == _gf_true) ? "enabled" : "disabled");
+
+ GF_OPTION_INIT("log-history", conf->log_history, bool, out);
+
+ gf_log(this->name, GF_LOG_DEBUG, "logging to history %s",
+ (conf->log_history == _gf_true) ? "enabled" : "disabled");
+
+ history = eh_new(history_size, _gf_false, NULL);
+ if (!history) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "event history cannot be "
+ "initialized");
+ return -1;
+ }
+
+ this->history = history;
+
+ conf->trace_log_level = GF_LOG_INFO;
+
+ if (dict_get(options, "force-log-level")) {
+ forced_loglevel = data_to_str(dict_get(options, "force-log-level"));
+ if (!forced_loglevel)
+ goto setloglevel;
+
+ if (strcmp(forced_loglevel, "INFO") == 0)
+ conf->trace_log_level = GF_LOG_INFO;
+ else if (strcmp(forced_loglevel, "TRACE") == 0)
+ conf->trace_log_level = GF_LOG_TRACE;
+ else if (strcmp(forced_loglevel, "ERROR") == 0)
+ conf->trace_log_level = GF_LOG_ERROR;
+ else if (strcmp(forced_loglevel, "DEBUG") == 0)
+ conf->trace_log_level = GF_LOG_DEBUG;
+ else if (strcmp(forced_loglevel, "WARNING") == 0)
+ conf->trace_log_level = GF_LOG_WARNING;
+ else if (strcmp(forced_loglevel, "CRITICAL") == 0)
+ conf->trace_log_level = GF_LOG_CRITICAL;
+ else if (strcmp(forced_loglevel, "NONE") == 0)
+ conf->trace_log_level = GF_LOG_NONE;
+ }
setloglevel:
- gf_log_set_loglevel (this->ctx, conf->trace_log_level);
- this->private = conf;
- ret = 0;
-out:
- if (ret == -1) {
- if (history)
- GF_FREE (history);
- if (conf)
- GF_FREE (conf);
- }
+ gf_log_set_loglevel(this->ctx, conf->trace_log_level);
+ this->private = conf;
+ ret = 0;
+out:
+ if (ret == -1) {
+ if (history)
+ GF_FREE(history);
+ if (conf)
+ GF_FREE(conf);
+ }
- return ret;
+ return ret;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- if (!this)
- return;
+ if (!this)
+ return;
- if (this->history)
- eh_destroy (this->history);
+ if (this->history)
+ eh_destroy(this->history);
- gf_log (this->name, GF_LOG_INFO,
- "trace translator unloaded");
- return;
+ gf_log(this->name, GF_LOG_INFO, "trace translator unloaded");
+ return;
}
struct xlator_fops fops = {
- .stat = trace_stat,
- .readlink = trace_readlink,
- .mknod = trace_mknod,
- .mkdir = trace_mkdir,
- .unlink = trace_unlink,
- .rmdir = trace_rmdir,
- .symlink = trace_symlink,
- .rename = trace_rename,
- .link = trace_link,
- .truncate = trace_truncate,
- .open = trace_open,
- .readv = trace_readv,
- .writev = trace_writev,
- .statfs = trace_statfs,
- .flush = trace_flush,
- .fsync = trace_fsync,
- .setxattr = trace_setxattr,
- .getxattr = trace_getxattr,
- .fsetxattr = trace_fsetxattr,
- .fgetxattr = trace_fgetxattr,
- .removexattr = trace_removexattr,
- .opendir = trace_opendir,
- .readdir = trace_readdir,
- .readdirp = trace_readdirp,
- .fsyncdir = trace_fsyncdir,
- .access = trace_access,
- .ftruncate = trace_ftruncate,
- .fstat = trace_fstat,
- .create = trace_create,
- .lk = trace_lk,
- .inodelk = trace_inodelk,
- .finodelk = trace_finodelk,
- .entrylk = trace_entrylk,
- .fentrylk = trace_fentrylk,
- .lookup = trace_lookup,
- .rchecksum = trace_rchecksum,
- .xattrop = trace_xattrop,
- .fxattrop = trace_fxattrop,
- .setattr = trace_setattr,
- .fsetattr = trace_fsetattr,
- .seek = trace_seek,
+ .stat = trace_stat,
+ .readlink = trace_readlink,
+ .mknod = trace_mknod,
+ .mkdir = trace_mkdir,
+ .unlink = trace_unlink,
+ .rmdir = trace_rmdir,
+ .symlink = trace_symlink,
+ .rename = trace_rename,
+ .link = trace_link,
+ .truncate = trace_truncate,
+ .open = trace_open,
+ .readv = trace_readv,
+ .writev = trace_writev,
+ .statfs = trace_statfs,
+ .flush = trace_flush,
+ .fsync = trace_fsync,
+ .setxattr = trace_setxattr,
+ .getxattr = trace_getxattr,
+ .fsetxattr = trace_fsetxattr,
+ .fgetxattr = trace_fgetxattr,
+ .removexattr = trace_removexattr,
+ .opendir = trace_opendir,
+ .readdir = trace_readdir,
+ .readdirp = trace_readdirp,
+ .fsyncdir = trace_fsyncdir,
+ .access = trace_access,
+ .ftruncate = trace_ftruncate,
+ .fstat = trace_fstat,
+ .create = trace_create,
+ .lk = trace_lk,
+ .inodelk = trace_inodelk,
+ .finodelk = trace_finodelk,
+ .entrylk = trace_entrylk,
+ .fentrylk = trace_fentrylk,
+ .lookup = trace_lookup,
+ .rchecksum = trace_rchecksum,
+ .xattrop = trace_xattrop,
+ .fxattrop = trace_fxattrop,
+ .setattr = trace_setattr,
+ .fsetattr = trace_fsetattr,
+ .seek = trace_seek,
};
struct xlator_cbks cbks = {
- .release = trace_release,
- .releasedir = trace_releasedir,
- .forget = trace_forget,
+ .release = trace_release,
+ .releasedir = trace_releasedir,
+ .forget = trace_forget,
};
struct volume_options options[] = {
- { .key = {"include-ops", "include"},
- .type = GF_OPTION_TYPE_STR,
- /*.value = { ""} */
- },
- { .key = {"exclude-ops", "exclude"},
- .type = GF_OPTION_TYPE_STR
- /*.value = { ""} */
- },
- { .key = {"history-size"},
- .type = GF_OPTION_TYPE_SIZET,
- .default_value = "1024",
- },
- { .key = {"log-file"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "no",
- },
- { .key = {"log-history"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "no",
- },
- { .key = {NULL} },
+ {
+ .key = {"include-ops", "include"},
+ .type = GF_OPTION_TYPE_STR,
+ /*.value = { ""} */
+ },
+ {
+ .key = {"exclude-ops", "exclude"},
+ .type = GF_OPTION_TYPE_STR
+ /*.value = { ""} */
+ },
+ {
+ .key = {"history-size"},
+ .type = GF_OPTION_TYPE_SIZET,
+ .default_value = "1024",
+ },
+ {
+ .key = {"log-file"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "no",
+ },
+ {
+ .key = {"log-history"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "no",
+ },
+ {.key = {NULL}},
};
-struct xlator_dumpops dumpops = {
- .history = trace_dump_history
+struct xlator_dumpops dumpops = {.history = trace_dump_history};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1},
+ .dumpops = &dumpops,
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "trace",
+ .category = GF_TECH_PREVIEW,
};
diff --git a/xlators/debug/trace/src/trace.h b/xlators/debug/trace/src/trace.h
index 3b5f7891d00..b16304799da 100644
--- a/xlators/debug/trace/src/trace.h
+++ b/xlators/debug/trace/src/trace.h
@@ -10,47 +10,46 @@
#include <time.h>
#include <errno.h>
-#include "glusterfs.h"
-#include "xlator.h"
-#include "common-utils.h"
-#include "event-history.h"
-#include "logging.h"
-#include "circ-buff.h"
-#include "statedump.h"
-#include "options.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/common-utils.h>
+#include <glusterfs/event-history.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/circ-buff.h>
+#include <glusterfs/statedump.h>
+#include <glusterfs/options.h>
#define TRACE_DEFAULT_HISTORY_SIZE 1024
typedef struct {
- /* Since the longest fop name is fremovexattr i.e 12 characters, array size
- * is kept 24, i.e double of the maximum.
- */
- char name[24];
- int enabled;
+ /* Since the longest fop name is fremovexattr i.e 12 characters, array size
+ * is kept 24, i.e double of the maximum.
+ */
+ char name[24];
+ int enabled;
} trace_fop_name_t;
trace_fop_name_t trace_fop_names[GF_FOP_MAXVALUE];
typedef struct {
- gf_boolean_t log_file;
- gf_boolean_t log_history;
- size_t history_size;
- int trace_log_level;
+ gf_boolean_t log_file;
+ gf_boolean_t log_history;
+ uint64_t history_size;
+ int trace_log_level;
} trace_conf_t;
-#define TRACE_STACK_UNWIND(op, frame, params ...) \
- do { \
- frame->local = NULL; \
- STACK_UNWIND_STRICT (op, frame, params); \
- } while (0);
-
-#define LOG_ELEMENT(_conf, _string) \
- do { \
- if (_conf) { \
- if ((_conf->log_history) == _gf_true) \
- gf_log_eh ("%s", _string); \
- if ((_conf->log_file) == _gf_true) \
- gf_log (THIS->name, _conf->trace_log_level, \
- "%s", _string); \
- } \
- } while (0);
+#define TRACE_STACK_UNWIND(op, frame, params...) \
+ do { \
+ frame->local = NULL; \
+ STACK_UNWIND_STRICT(op, frame, params); \
+ } while (0);
+
+#define LOG_ELEMENT(_conf, _string) \
+ do { \
+ if (_conf) { \
+ if ((_conf->log_history) == _gf_true) \
+ gf_log_eh("%s", _string); \
+ if ((_conf->log_file) == _gf_true) \
+ gf_log(THIS->name, _conf->trace_log_level, "%s", _string); \
+ } \
+ } while (0);
diff --git a/xlators/encryption/Makefile.am b/xlators/encryption/Makefile.am
deleted file mode 100644
index 36efc6698bd..00000000000
--- a/xlators/encryption/Makefile.am
+++ /dev/null
@@ -1,3 +0,0 @@
-SUBDIRS = rot-13 crypt
-
-CLEANFILES =
diff --git a/xlators/encryption/crypt/src/Makefile.am b/xlators/encryption/crypt/src/Makefile.am
deleted file mode 100644
index 05fd3d5096b..00000000000
--- a/xlators/encryption/crypt/src/Makefile.am
+++ /dev/null
@@ -1,26 +0,0 @@
-if ENABLE_CRYPT_XLATOR
-
-xlator_LTLIBRARIES = crypt.la
-xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/encryption
-
-crypt_la_LDFLAGS = -module $(GF_XLATOR_DEFAULT_LDFLAGS)
-
-crypt_la_SOURCES = keys.c data.c metadata.c atom.c crypt.c
-crypt_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la \
- -lssl -lcrypto
-
-noinst_HEADERS = crypt-common.h crypt-mem-types.h crypt.h metadata.h
-
-AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
- -I$(top_srcdir)/rpc/xdr/src -I$(top_builddir)/rpc/xdr/src
-
-AM_CFLAGS = -Wall $(GF_CFLAGS)
-
-CLEANFILES =
-
-else
-
-noinst_DIST = keys.c data.c metadata.c atom.c crypt.c
-noinst_HEADERS = crypt-common.h crypt-mem-types.h crypt.h metadata.h
-
-endif
diff --git a/xlators/encryption/crypt/src/atom.c b/xlators/encryption/crypt/src/atom.c
deleted file mode 100644
index 6d7b351b370..00000000000
--- a/xlators/encryption/crypt/src/atom.c
+++ /dev/null
@@ -1,957 +0,0 @@
-/*
- Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#include "defaults.h"
-#include "crypt-common.h"
-#include "crypt.h"
-
-/*
- * Glossary
- *
- *
- * cblock (or cipher block). A logical unit in a file.
- * cblock size is defined as the number of bits
- * in an input (or output) block of the block
- * cipher (*). Cipher block size is a property of
- * cipher algorithm. E.g. cblock size is 64 bits
- * for DES, 128 bits for AES, etc.
- *
- * atomic cipher A cipher algorithm, which requires some chunks of
- * algorithm text to be padded at left and(or) right sides before
- * cipher transaform.
- *
- *
- * block (atom) Minimal chunk of file's data, which doesn't require
- * padding. We'll consider logical units in a file of
- * block size (atom size).
- *
- * cipher algorithm Atomic cipher algorithm, which requires the last
- * with EOF issue incomplete cblock in a file to be padded with some
- * data (usually zeros).
- *
- *
- * operation, which reading/writing from offset, which is not aligned to
- * forms a gap at to atom size
- * the beginning
- *
- *
- * operation, which reading/writing count bytes starting from offset off,
- * forms a gap at so that off+count is not aligned to atom_size
- * the end
- *
- * head block the first atom affected by an operation, which forms
- * a gap at the beginning, or(and) at the end.
- * Сomment. Head block has at least one gap (either at
- * the beginning, or at the end)
- *
- *
- * tail block the last atom different from head, affected by an
- * operation, which forms a gap at the end.
- * Сomment: Tail block has exactly one gap (at the end).
- *
- *
- * partial block head or tail block
- *
- *
- * full block block without gaps.
- *
- *
- * (*) Recommendation for Block Cipher Modes of Operation
- * Methods and Techniques
- * NIST Special Publication 800-38A Edition 2001
- */
-
-/*
- * atom->offset_at()
- */
-static off_t offset_at_head(struct avec_config *conf)
-{
- return conf->aligned_offset;
-}
-
-static off_t offset_at_hole_head(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return offset_at_head(get_hole_conf(frame));
-}
-
-static off_t offset_at_data_head(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return offset_at_head(get_data_conf(frame));
-}
-
-
-static off_t offset_at_tail(struct avec_config *conf,
- struct object_cipher_info *object)
-{
- return conf->aligned_offset +
- (conf->off_in_head ? get_atom_size(object) : 0) +
- (conf->nr_full_blocks << get_atom_bits(object));
-}
-
-static off_t offset_at_hole_tail(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return offset_at_tail(get_hole_conf(frame), object);
-}
-
-
-static off_t offset_at_data_tail(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return offset_at_tail(get_data_conf(frame), object);
-}
-
-static off_t offset_at_full(struct avec_config *conf,
- struct object_cipher_info *object)
-{
- return conf->aligned_offset +
- (conf->off_in_head ? get_atom_size(object) : 0);
-}
-
-static off_t offset_at_data_full(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return offset_at_full(get_data_conf(frame), object);
-}
-
-static off_t offset_at_hole_full(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return offset_at_full(get_hole_conf(frame), object);
-}
-
-/*
- * atom->io_size_nopad()
- */
-
-static uint32_t io_size_nopad_head(struct avec_config *conf,
- struct object_cipher_info *object)
-{
- uint32_t gap_at_beg;
- uint32_t gap_at_end;
-
- check_head_block(conf);
-
- gap_at_beg = conf->off_in_head;
-
- if (has_tail_block(conf) || has_full_blocks(conf) || conf->off_in_tail == 0 )
- gap_at_end = 0;
- else
- gap_at_end = get_atom_size(object) - conf->off_in_tail;
-
- return get_atom_size(object) - (gap_at_beg + gap_at_end);
-}
-
-static uint32_t io_size_nopad_tail(struct avec_config *conf,
- struct object_cipher_info *object)
-{
- check_tail_block(conf);
- return conf->off_in_tail;
-}
-
-static uint32_t io_size_nopad_full(struct avec_config *conf,
- struct object_cipher_info *object)
-{
- check_full_block(conf);
- return get_atom_size(object);
-}
-
-static uint32_t io_size_nopad_data_head(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return io_size_nopad_head(get_data_conf(frame), object);
-}
-
-static uint32_t io_size_nopad_hole_head(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return io_size_nopad_head(get_hole_conf(frame), object);
-}
-
-static uint32_t io_size_nopad_data_tail(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return io_size_nopad_tail(get_data_conf(frame), object);
-}
-
-static uint32_t io_size_nopad_hole_tail(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return io_size_nopad_tail(get_hole_conf(frame), object);
-}
-
-static uint32_t io_size_nopad_data_full(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return io_size_nopad_full(get_data_conf(frame), object);
-}
-
-static uint32_t io_size_nopad_hole_full(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return io_size_nopad_full(get_hole_conf(frame), object);
-}
-
-static uint32_t offset_in_head(struct avec_config *conf)
-{
- check_cursor_head(conf);
-
- return conf->off_in_head;
-}
-
-static uint32_t offset_in_tail(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return 0;
-}
-
-static uint32_t offset_in_full(struct avec_config *conf,
- struct object_cipher_info *object)
-{
- check_cursor_full(conf);
-
- if (has_head_block(conf))
- return (conf->cursor - 1) << get_atom_bits(object);
- else
- return conf->cursor << get_atom_bits(object);
-}
-
-static uint32_t offset_in_data_head(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return offset_in_head(get_data_conf(frame));
-}
-
-static uint32_t offset_in_hole_head(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return offset_in_head(get_hole_conf(frame));
-}
-
-static uint32_t offset_in_data_full(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return offset_in_full(get_data_conf(frame), object);
-}
-
-static uint32_t offset_in_hole_full(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return offset_in_full(get_hole_conf(frame), object);
-}
-
-/*
- * atom->rmw()
- */
-/*
- * Pre-conditions:
- * @vec contains plain text of the latest
- * version.
- *
- * Uptodate gaps of the @partial block with
- * this plain text, encrypt the whole block
- * and write the result to disk.
- */
-static int32_t rmw_partial_block(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iovec *vec,
- int32_t count,
- struct iatt *stbuf,
- struct iobref *iobref,
- struct rmw_atom *atom)
-{
- size_t was_read = 0;
- uint64_t file_size;
- crypt_local_t *local = frame->local;
- struct object_cipher_info *object = &local->info->cinfo;
-
- struct iovec *partial = atom->get_iovec(frame, 0);
- struct avec_config *conf = atom->get_config(frame);
- end_writeback_handler_t end_writeback_partial_block;
-#if DEBUG_CRYPT
- gf_boolean_t check_last_cblock = _gf_false;
-#endif
- local->op_ret = op_ret;
- local->op_errno = op_errno;
-
- if (op_ret < 0)
- goto exit;
-
- file_size = local->cur_file_size;
- was_read = op_ret;
-
- if (atom->locality == HEAD_ATOM && conf->off_in_head) {
- /*
- * head atom with a non-uptodate gap
- * at the beginning
- *
- * fill the gap with plain text of the
- * latest version. Convert a part of hole
- * (if any) to zeros.
- */
- int32_t i;
- int32_t copied = 0;
- int32_t to_gap; /* amount of data needed to uptodate
- the gap at the beginning */
-#if 0
- int32_t hole = 0; /* The part of the hole which
- * got in the head block */
-#endif /* 0 */
- to_gap = conf->off_in_head;
-
- if (was_read < to_gap) {
- if (file_size >
- offset_at_head(conf) + was_read) {
- /*
- * It is impossible to uptodate
- * head block: too few bytes have
- * been read from disk, so that
- * partial write is impossible.
- *
- * It could happen because of many
- * reasons: IO errors, (meta)data
- * corruption in the local file system,
- * etc.
- */
- gf_log(this->name, GF_LOG_WARNING,
- "Can not uptodate a gap at the beginning");
- local->op_ret = -1;
- local->op_errno = EIO;
- goto exit;
- }
-#if 0
- hole = to_gap - was_read;
-#endif /* 0 */
- to_gap = was_read;
- }
- /*
- * uptodate the gap at the beginning
- */
- for (i = 0; i < count && copied < to_gap; i++) {
- int32_t to_copy;
-
- to_copy = vec[i].iov_len;
- if (to_copy > to_gap - copied)
- to_copy = to_gap - copied;
-
- memcpy(partial->iov_base, vec[i].iov_base, to_copy);
- copied += to_copy;
- }
-#if 0
- /*
- * If possible, convert part of the
- * hole, which got in the head block
- */
- ret = TRY_LOCK(&local->hole_lock);
- if (!ret) {
- if (local->hole_handled)
- /*
- * already converted by
- * crypt_writev_cbk()
- */
- UNLOCK(&local->hole_lock);
- else {
- /*
- * convert the part of the hole
- * which got in the head block
- * to zeros.
- *
- * Update the orig_offset to make
- * sure writev_cbk() won't care
- * about this part of the hole.
- *
- */
- memset(partial->iov_base + to_gap, 0, hole);
-
- conf->orig_offset -= hole;
- conf->orig_size += hole;
- UNLOCK(&local->hole_lock);
- }
- }
- else /*
- * conversion is being performed
- * by crypt_writev_cbk()
- */
- ;
-#endif /* 0 */
- }
- if (atom->locality == TAIL_ATOM ||
- (!has_tail_block(conf) && conf->off_in_tail)) {
- /*
- * tail atom, or head atom with a non-uptodate
- * gap at the end.
- *
- * fill the gap at the end of the block
- * with plain text of the latest version.
- * Pad the result, (if needed)
- */
- int32_t i;
- int32_t to_gap;
- int copied;
- off_t off_in_tail;
- int32_t to_copy;
-
- off_in_tail = conf->off_in_tail;
- to_gap = conf->gap_in_tail;
-
- if (to_gap && was_read < off_in_tail + to_gap) {
- /*
- * It is impossible to uptodate
- * the gap at the end: too few bytes
- * have been read from disk, so that
- * partial write is impossible.
- *
- * It could happen because of many
- * reasons: IO errors, (meta)data
- * corruption in the local file system,
- * etc.
- */
- gf_log(this->name, GF_LOG_WARNING,
- "Can not uptodate a gap at the end");
- local->op_ret = -1;
- local->op_errno = EIO;
- goto exit;
- }
- /*
- * uptodate the gap at the end
- */
- copied = 0;
- to_copy = to_gap;
- for(i = count - 1; i >= 0 && to_copy > 0; i--) {
- uint32_t from_vec, off_in_vec;
-
- off_in_vec = 0;
- from_vec = vec[i].iov_len;
- if (from_vec > to_copy) {
- off_in_vec = from_vec - to_copy;
- from_vec = to_copy;
- }
- memcpy(partial->iov_base +
- off_in_tail + to_gap - copied - from_vec,
- vec[i].iov_base + off_in_vec,
- from_vec);
-
- gf_log(this->name, GF_LOG_DEBUG,
- "uptodate %d bytes at tail. Offset at target(source): %d(%d)",
- (int)from_vec,
- (int)off_in_tail + to_gap - copied - from_vec,
- (int)off_in_vec);
-
- copied += from_vec;
- to_copy -= from_vec;
- }
- partial->iov_len = off_in_tail + to_gap;
-
- if (object_alg_should_pad(object)) {
- int32_t resid = 0;
- resid = partial->iov_len & (object_alg_blksize(object) - 1);
- if (resid) {
- /*
- * append a new EOF padding
- */
- local->eof_padding_size =
- object_alg_blksize(object) - resid;
-
- gf_log(this->name, GF_LOG_DEBUG,
- "set padding size %d",
- local->eof_padding_size);
-
- memset(partial->iov_base + partial->iov_len,
- 1,
- local->eof_padding_size);
- partial->iov_len += local->eof_padding_size;
-#if DEBUG_CRYPT
- gf_log(this->name, GF_LOG_DEBUG,
- "pad cblock with %d zeros:",
- local->eof_padding_size);
- dump_cblock(this,
- (unsigned char *)partial->iov_base +
- partial->iov_len - object_alg_blksize(object));
- check_last_cblock = _gf_true;
-#endif
- }
- }
- }
- /*
- * encrypt the whole block
- */
- encrypt_aligned_iov(object,
- partial,
- 1,
- atom->offset_at(frame, object));
-#if DEBUG_CRYPT
- if (check_last_cblock == _gf_true) {
- gf_log(this->name, GF_LOG_DEBUG,
- "encrypt last cblock with offset %llu",
- (unsigned long long)atom->offset_at(frame, object));
- dump_cblock(this, (unsigned char *)partial->iov_base +
- partial->iov_len - object_alg_blksize(object));
- }
-#endif
- set_local_io_params_writev(frame, object, atom,
- atom->offset_at(frame, object),
- iov_length(partial, 1));
- /*
- * write the whole block to disk
- */
- end_writeback_partial_block = dispatch_end_writeback(local->fop);
- conf->cursor ++;
- STACK_WIND(frame,
- end_writeback_partial_block,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->writev,
- local->fd,
- partial,
- 1,
- atom->offset_at(frame, object),
- local->flags,
- local->iobref_data,
- local->xdata);
-
- gf_log("crypt", GF_LOG_DEBUG,
- "submit partial block: %d bytes from %d offset",
- (int)iov_length(partial, 1),
- (int)atom->offset_at(frame, object));
- exit:
- return 0;
-}
-
-/*
- * Perform a (read-)modify-write sequence.
- * This should be performed only after approval
- * of upper server-side manager, i.e. the caller
- * needs to make sure this is his turn to rmw.
- */
-void submit_partial(call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- atom_locality_type ltype)
-{
- int32_t ret;
- dict_t *dict;
- struct rmw_atom *atom;
- crypt_local_t *local = frame->local;
- struct object_cipher_info *object = &local->info->cinfo;
-
- atom = atom_by_types(local->active_setup, ltype);
- /*
- * To perform the "read" component of the read-modify-write
- * sequence the crypt translator does stack_wind to itself.
- *
- * Pass current file size to crypt_readv()
- */
- dict = dict_new();
- if (!dict) {
- /*
- * FIXME: Handle the error
- */
- gf_log("crypt", GF_LOG_WARNING, "Can not alloc dict");
- return;
- }
- ret = dict_set(dict,
- FSIZE_XATTR_PREFIX,
- data_from_uint64(local->cur_file_size));
- if (ret) {
- /*
- * FIXME: Handle the error
- */
- dict_unref(dict);
- gf_log("crypt", GF_LOG_WARNING, "Can not set dict");
- goto exit;
- }
- STACK_WIND(frame,
- atom->rmw,
- this,
- this->fops->readv, /* crypt_readv */
- fd,
- atom->count_to_uptodate(frame, object), /* count */
- atom->offset_at(frame, object), /* offset to read from */
- 0,
- dict);
- exit:
- dict_unref(dict);
-}
-
-/*
- * submit blocks of FULL_ATOM type
- */
-void submit_full(call_frame_t *frame, xlator_t *this)
-{
- crypt_local_t *local = frame->local;
- struct object_cipher_info *object = &local->info->cinfo;
- struct rmw_atom *atom = atom_by_types(local->active_setup, FULL_ATOM);
- uint32_t count; /* total number of full blocks to submit */
- uint32_t granularity; /* number of blocks to submit in one iteration */
-
- uint64_t off_in_file; /* start offset in the file, bytes */
- uint32_t off_in_atom; /* start offset in the atom, blocks */
- uint32_t blocks_written = 0; /* blocks written for this submit */
-
- struct avec_config *conf = atom->get_config(frame);
- end_writeback_handler_t end_writeback_full_block;
- /*
- * Write full blocks by groups of granularity size.
- */
- end_writeback_full_block = dispatch_end_writeback(local->fop);
-
- if (is_ordered_mode(frame)) {
- uint32_t skip = has_head_block(conf) ? 1 : 0;
- count = 1;
- granularity = 1;
- /*
- * calculate start offset using cursor value;
- * here we should take into accout head block,
- * which corresponds to cursor value 0.
- */
- off_in_file = atom->offset_at(frame, object) +
- ((conf->cursor - skip) << get_atom_bits(object));
- off_in_atom = conf->cursor - skip;
- }
- else {
- /*
- * in parallel mode
- */
- count = conf->nr_full_blocks;
- granularity = MAX_IOVEC;
- off_in_file = atom->offset_at(frame, object);
- off_in_atom = 0;
- }
- while (count) {
- uint32_t blocks_to_write = count;
-
- if (blocks_to_write > granularity)
- blocks_to_write = granularity;
- if (conf->type == HOLE_ATOM)
- /*
- * reset iovec before encryption
- */
- memset(atom->get_iovec(frame, 0)->iov_base,
- 0,
- get_atom_size(object));
- /*
- * encrypt the group
- */
- encrypt_aligned_iov(object,
- atom->get_iovec(frame,
- off_in_atom +
- blocks_written),
- blocks_to_write,
- off_in_file + (blocks_written <<
- get_atom_bits(object)));
-
- set_local_io_params_writev(frame, object, atom,
- off_in_file + (blocks_written << get_atom_bits(object)),
- blocks_to_write << get_atom_bits(object));
-
- conf->cursor += blocks_to_write;
-
- STACK_WIND(frame,
- end_writeback_full_block,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->writev,
- local->fd,
- atom->get_iovec(frame, off_in_atom + blocks_written),
- blocks_to_write,
- off_in_file + (blocks_written << get_atom_bits(object)),
- local->flags,
- local->iobref_data ? local->iobref_data : local->iobref,
- local->xdata);
-
- gf_log("crypt", GF_LOG_DEBUG, "submit %d full blocks from %d offset",
- blocks_to_write,
- (int)(off_in_file + (blocks_written << get_atom_bits(object))));
-
- count -= blocks_to_write;
- blocks_written += blocks_to_write;
- }
- return;
-}
-
-static int32_t rmw_data_head(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iovec *vec,
- int32_t count,
- struct iatt *stbuf,
- struct iobref *iobref,
- dict_t *xdata)
-{
- return rmw_partial_block(frame,
- cookie,
- this,
- op_ret,
- op_errno,
- vec,
- count,
- stbuf,
- iobref,
- atom_by_types(DATA_ATOM, HEAD_ATOM));
-}
-
-static int32_t rmw_data_tail(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iovec *vec,
- int32_t count,
- struct iatt *stbuf,
- struct iobref *iobref,
- dict_t *xdata)
-{
- return rmw_partial_block(frame,
- cookie,
- this,
- op_ret,
- op_errno,
- vec,
- count,
- stbuf,
- iobref,
- atom_by_types(DATA_ATOM, TAIL_ATOM));
-}
-
-static int32_t rmw_hole_head(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iovec *vec,
- int32_t count,
- struct iatt *stbuf,
- struct iobref *iobref,
- dict_t *xdata)
-{
- return rmw_partial_block(frame,
- cookie,
- this,
- op_ret,
- op_errno,
- vec,
- count,
- stbuf,
- iobref,
- atom_by_types(HOLE_ATOM, HEAD_ATOM));
-}
-
-static int32_t rmw_hole_tail(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iovec *vec,
- int32_t count,
- struct iatt *stbuf,
- struct iobref *iobref,
- dict_t *xdata)
-{
- return rmw_partial_block(frame,
- cookie,
- this,
- op_ret,
- op_errno,
- vec,
- count,
- stbuf,
- iobref,
- atom_by_types(HOLE_ATOM, TAIL_ATOM));
-}
-
-/*
- * atom->count_to_uptodate()
- */
-static uint32_t count_to_uptodate_head(struct avec_config *conf,
- struct object_cipher_info *object)
-{
- if (conf->acount == 1 && conf->off_in_tail)
- return get_atom_size(object);
- else
- /* there is no need to read the whole head block */
- return conf->off_in_head;
-}
-
-static uint32_t count_to_uptodate_tail(struct avec_config *conf,
- struct object_cipher_info *object)
-{
- /* we need to read the whole tail block */
- return get_atom_size(object);
-}
-
-static uint32_t count_to_uptodate_data_head(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return count_to_uptodate_head(get_data_conf(frame), object);
-}
-
-static uint32_t count_to_uptodate_data_tail(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return count_to_uptodate_tail(get_data_conf(frame), object);
-}
-
-static uint32_t count_to_uptodate_hole_head(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return count_to_uptodate_head(get_hole_conf(frame), object);
-}
-
-static uint32_t count_to_uptodate_hole_tail(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- return count_to_uptodate_tail(get_hole_conf(frame), object);
-}
-
-/* atom->get_config() */
-
-static struct avec_config *get_config_data(call_frame_t *frame)
-{
- return &((crypt_local_t *)frame->local)->data_conf;
-}
-
-static struct avec_config *get_config_hole(call_frame_t *frame)
-{
- return &((crypt_local_t *)frame->local)->hole_conf;
-}
-
-/*
- * atom->get_iovec()
- */
-static struct iovec *get_iovec_hole_head(call_frame_t *frame,
- uint32_t count)
-{
- struct avec_config *conf = get_hole_conf(frame);
-
- return conf->avec;
-}
-
-static struct iovec *get_iovec_hole_full(call_frame_t *frame,
- uint32_t count)
-{
- struct avec_config *conf = get_hole_conf(frame);
-
- return conf->avec + (conf->off_in_head ? 1 : 0);
-}
-
-static struct iovec *get_iovec_hole_tail(call_frame_t *frame,
- uint32_t count)
-{
- struct avec_config *conf = get_hole_conf(frame);
-
- return conf->avec + (conf->blocks_in_pool - 1);
-}
-
-static struct iovec *get_iovec_data_head(call_frame_t *frame,
- uint32_t count)
-{
- struct avec_config *conf = get_data_conf(frame);
-
- return conf->avec;
-}
-
-static struct iovec *get_iovec_data_full(call_frame_t *frame,
- uint32_t count)
-{
- struct avec_config *conf = get_data_conf(frame);
-
- return conf->avec + (conf->off_in_head ? 1 : 0) + count;
-}
-
-static struct iovec *get_iovec_data_tail(call_frame_t *frame,
- uint32_t count)
-{
- struct avec_config *conf = get_data_conf(frame);
-
- return conf->avec +
- (conf->off_in_head ? 1 : 0) +
- conf->nr_full_blocks;
-}
-
-static struct rmw_atom atoms[LAST_DATA_TYPE][LAST_LOCALITY_TYPE] = {
- [DATA_ATOM][HEAD_ATOM] =
- { .locality = HEAD_ATOM,
- .rmw = rmw_data_head,
- .offset_at = offset_at_data_head,
- .offset_in = offset_in_data_head,
- .get_iovec = get_iovec_data_head,
- .io_size_nopad = io_size_nopad_data_head,
- .count_to_uptodate = count_to_uptodate_data_head,
- .get_config = get_config_data
- },
- [DATA_ATOM][TAIL_ATOM] =
- { .locality = TAIL_ATOM,
- .rmw = rmw_data_tail,
- .offset_at = offset_at_data_tail,
- .offset_in = offset_in_tail,
- .get_iovec = get_iovec_data_tail,
- .io_size_nopad = io_size_nopad_data_tail,
- .count_to_uptodate = count_to_uptodate_data_tail,
- .get_config = get_config_data
- },
- [DATA_ATOM][FULL_ATOM] =
- { .locality = FULL_ATOM,
- .offset_at = offset_at_data_full,
- .offset_in = offset_in_data_full,
- .get_iovec = get_iovec_data_full,
- .io_size_nopad = io_size_nopad_data_full,
- .get_config = get_config_data
- },
- [HOLE_ATOM][HEAD_ATOM] =
- { .locality = HEAD_ATOM,
- .rmw = rmw_hole_head,
- .offset_at = offset_at_hole_head,
- .offset_in = offset_in_hole_head,
- .get_iovec = get_iovec_hole_head,
- .io_size_nopad = io_size_nopad_hole_head,
- .count_to_uptodate = count_to_uptodate_hole_head,
- .get_config = get_config_hole
- },
- [HOLE_ATOM][TAIL_ATOM] =
- { .locality = TAIL_ATOM,
- .rmw = rmw_hole_tail,
- .offset_at = offset_at_hole_tail,
- .offset_in = offset_in_tail,
- .get_iovec = get_iovec_hole_tail,
- .io_size_nopad = io_size_nopad_hole_tail,
- .count_to_uptodate = count_to_uptodate_hole_tail,
- .get_config = get_config_hole
- },
- [HOLE_ATOM][FULL_ATOM] =
- { .locality = FULL_ATOM,
- .offset_at = offset_at_hole_full,
- .offset_in = offset_in_hole_full,
- .get_iovec = get_iovec_hole_full,
- .io_size_nopad = io_size_nopad_hole_full,
- .get_config = get_config_hole
- }
-};
-
-struct rmw_atom *atom_by_types(atom_data_type data,
- atom_locality_type locality)
-{
- return &atoms[data][locality];
-}
-
-/*
- Local variables:
- c-indentation-style: "K&R"
- mode-name: "LC"
- c-basic-offset: 8
- tab-width: 8
- fill-column: 80
- scroll-step: 1
- End:
-*/
diff --git a/xlators/encryption/crypt/src/crypt-common.h b/xlators/encryption/crypt/src/crypt-common.h
deleted file mode 100644
index 7c212ad5d25..00000000000
--- a/xlators/encryption/crypt/src/crypt-common.h
+++ /dev/null
@@ -1,141 +0,0 @@
-/*
- Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#ifndef __CRYPT_COMMON_H__
-#define __CRYPT_COMMON_H__
-
-#define INVAL_SUBVERSION_NUMBER (0xff)
-#define CRYPT_INVAL_OP (GF_FOP_NULL)
-
-#define CRYPTO_FORMAT_PREFIX "trusted.glusterfs.crypt.att.cfmt"
-#define FSIZE_XATTR_PREFIX "trusted.glusterfs.crypt.att.size"
-#define SUBREQ_PREFIX "trusted.glusterfs.crypt.msg.sreq"
-#define FSIZE_MSG_PREFIX "trusted.glusterfs.crypt.msg.size"
-#define DE_MSG_PREFIX "trusted.glusterfs.crypt.msg.dent"
-#define REQUEST_ID_PREFIX "trusted.glusterfs.crypt.msg.rqid"
-#define MSGFLAGS_PREFIX "trusted.glusterfs.crypt.msg.xfgs"
-
-
-/* messages for crypt_open() */
-#define MSGFLAGS_REQUEST_MTD_RLOCK 1 /* take read lock and don't unlock */
-#define MSGFLAGS_REQUEST_MTD_WLOCK 2 /* take write lock and don't unlock */
-
-#define AES_BLOCK_BITS (4) /* AES_BLOCK_SIZE == 1 << AES_BLOCK_BITS */
-
-#define noop do {; } while (0)
-#define cassert(cond) ({ switch (-1) { case (cond): case 0: break; } })
-#define __round_mask(x, y) ((__typeof__(x))((y)-1))
-#define round_up(x, y) ((((x)-1) | __round_mask(x, y))+1)
-
-/*
- * Format of file's metadata
- */
-struct crypt_format {
- uint8_t loader_id; /* version of metadata loader */
- uint8_t versioned[0]; /* file's metadata of specific version */
-} __attribute__((packed));
-
-typedef enum {
- AES_CIPHER_ALG,
- LAST_CIPHER_ALG
-} cipher_alg_t;
-
-typedef enum {
- XTS_CIPHER_MODE,
- LAST_CIPHER_MODE
-} cipher_mode_t;
-
-typedef enum {
- MTD_LOADER_V1,
- LAST_MTD_LOADER
-} mtd_loader_id;
-
-static inline void msgflags_set_mtd_rlock(uint32_t *flags)
-{
- *flags |= MSGFLAGS_REQUEST_MTD_RLOCK;
-}
-
-static inline void msgflags_set_mtd_wlock(uint32_t *flags)
-{
- *flags |= MSGFLAGS_REQUEST_MTD_WLOCK;
-}
-
-static inline gf_boolean_t msgflags_check_mtd_rlock(uint32_t *flags)
-{
- return *flags & MSGFLAGS_REQUEST_MTD_RLOCK;
-}
-
-static inline gf_boolean_t msgflags_check_mtd_wlock(uint32_t *flags)
-{
- return *flags & MSGFLAGS_REQUEST_MTD_WLOCK;
-}
-
-static inline gf_boolean_t msgflags_check_mtd_lock(uint32_t *flags)
-{
- return msgflags_check_mtd_rlock(flags) ||
- msgflags_check_mtd_wlock(flags);
-}
-
-/*
- * returns number of logical blocks occupied
- * (maybe partially) by @count bytes
- * at offset @start.
- */
-static inline off_t logical_blocks_occupied(uint64_t start, off_t count,
- int blkbits)
-{
- return ((start + count - 1) >> blkbits) - (start >> blkbits) + 1;
-}
-
-/*
- * are two bytes (represented by offsets @off1
- * and @off2 respectively) in the same logical
- * block.
- */
-static inline int in_same_lblock(uint64_t off1, uint64_t off2,
- int blkbits)
-{
- return off1 >> blkbits == off2 >> blkbits;
-}
-
-static inline void dump_cblock(xlator_t *this, unsigned char *buf)
-{
- gf_log(this->name, GF_LOG_DEBUG,
- "dump cblock: %x %x %x %x %x %x %x %x %x %x %x %x %x %x %x %x",
- (buf)[0],
- (buf)[1],
- (buf)[2],
- (buf)[3],
- (buf)[4],
- (buf)[5],
- (buf)[6],
- (buf)[7],
- (buf)[8],
- (buf)[9],
- (buf)[10],
- (buf)[11],
- (buf)[12],
- (buf)[13],
- (buf)[14],
- (buf)[15]);
-}
-
-#endif /* __CRYPT_COMMON_H__ */
-
-/*
- Local variables:
- c-indentation-style: "K&R"
- mode-name: "LC"
- c-basic-offset: 8
- tab-width: 8
- fill-column: 80
- scroll-step: 1
- End:
-*/
diff --git a/xlators/encryption/crypt/src/crypt-mem-types.h b/xlators/encryption/crypt/src/crypt-mem-types.h
deleted file mode 100644
index 1954c579423..00000000000
--- a/xlators/encryption/crypt/src/crypt-mem-types.h
+++ /dev/null
@@ -1,45 +0,0 @@
-/*
- Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-
-#ifndef __CRYPT_MEM_TYPES_H__
-#define __CRYPT_MEM_TYPES_H__
-
-#include "mem-types.h"
-
-enum gf_crypt_mem_types_ {
- gf_crypt_mt_priv = gf_common_mt_end + 1,
- gf_crypt_mt_inode,
- gf_crypt_mt_data,
- gf_crypt_mt_mtd,
- gf_crypt_mt_loc,
- gf_crypt_mt_iatt,
- gf_crypt_mt_key,
- gf_crypt_mt_iovec,
- gf_crypt_mt_char,
- gf_crypt_mt_local,
- gf_crypt_mt_end,
-};
-
-#endif /* __CRYPT_MEM_TYPES_H__ */
-
-/*
- Local variables:
- c-indentation-style: "K&R"
- mode-name: "LC"
- c-basic-offset: 8
- tab-width: 8
- fill-column: 80
- scroll-step: 1
- End:
-*/
-
-
-
diff --git a/xlators/encryption/crypt/src/crypt.c b/xlators/encryption/crypt/src/crypt.c
deleted file mode 100644
index 9e5bcf79c30..00000000000
--- a/xlators/encryption/crypt/src/crypt.c
+++ /dev/null
@@ -1,4525 +0,0 @@
-/*
- Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-#include <ctype.h>
-#include <sys/uio.h>
-
-#include "glusterfs.h"
-#include "xlator.h"
-#include "logging.h"
-#include "defaults.h"
-
-#include "crypt-common.h"
-#include "crypt.h"
-
-static void init_inode_info_head(struct crypt_inode_info *info, fd_t *fd);
-static int32_t init_inode_info_tail(struct crypt_inode_info *info,
- struct master_cipher_info *master);
-static int32_t prepare_for_submit_hole(call_frame_t *frame, xlator_t *this,
- uint64_t from, off_t size);
-static int32_t load_file_size(call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- dict_t *dict, dict_t *xdata);
-static void do_ordered_submit(call_frame_t *frame, xlator_t *this,
- atom_data_type dtype);
-static void do_parallel_submit(call_frame_t *frame, xlator_t *this,
- atom_data_type dtype);
-static void put_one_call_open(call_frame_t *frame);
-static void put_one_call_readv(call_frame_t *frame, xlator_t *this);
-static void put_one_call_writev(call_frame_t *frame, xlator_t *this);
-static void put_one_call_ftruncate(call_frame_t *frame, xlator_t *this);
-static void free_avec(struct iovec *avec, char **pool, int blocks_in_pool);
-static void free_avec_data(crypt_local_t *local);
-static void free_avec_hole(crypt_local_t *local);
-
-static crypt_local_t *crypt_alloc_local(call_frame_t *frame, xlator_t *this,
- glusterfs_fop_t fop)
-{
- crypt_local_t *local = NULL;
-
- local = GF_CALLOC (1, sizeof (*local), gf_crypt_mt_local);
- if (!local) {
- gf_log(this->name, GF_LOG_ERROR, "out of memory");
- return NULL;
- }
- local->fop = fop;
- LOCK_INIT(&local->hole_lock);
- LOCK_INIT(&local->call_lock);
- LOCK_INIT(&local->rw_count_lock);
-
- frame->local = local;
- return local;
-}
-
-struct crypt_inode_info *get_crypt_inode_info(inode_t *inode, xlator_t *this)
-{
- int ret;
- uint64_t value = 0;
- struct crypt_inode_info *info;
-
- ret = inode_ctx_get(inode, this, &value);
- if (ret == -1) {
- gf_log (this->name, GF_LOG_WARNING,
- "Can not get inode info");
- return NULL;
- }
- info = (struct crypt_inode_info *)(long)value;
- if (info == NULL) {
- gf_log (this->name, GF_LOG_WARNING,
- "Can not obtain inode info");
- return NULL;
- }
- return info;
-}
-
-static struct crypt_inode_info *local_get_inode_info(crypt_local_t *local,
- xlator_t *this)
-{
- if (local->info)
- return local->info;
- local->info = get_crypt_inode_info(local->fd->inode, this);
- return local->info;
-}
-
-static struct crypt_inode_info *alloc_inode_info(crypt_local_t *local,
- loc_t *loc)
-{
- struct crypt_inode_info *info;
-
- info = GF_CALLOC(1, sizeof(*info), gf_crypt_mt_inode);
- if (!info) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- gf_log ("crypt", GF_LOG_WARNING,
- "Can not allocate inode info");
- return NULL;
- }
- memset(info, 0, sizeof(*info));
-#if DEBUG_CRYPT
- info->loc = GF_CALLOC(1, sizeof(*loc), gf_crypt_mt_loc);
- if (!info->loc) {
- gf_log("crypt", GF_LOG_WARNING, "Can not allocate loc");
- GF_FREE(info);
- return NULL;
- }
- if (loc_copy(info->loc, loc)){
- GF_FREE(info->loc);
- GF_FREE(info);
- return NULL;
- }
-#endif /* DEBUG_CRYPT */
-
- local->info = info;
- return info;
-}
-
-static void free_inode_info(struct crypt_inode_info *info)
-{
-#if DEBUG_CRYPT
- loc_wipe(info->loc);
- GF_FREE(info->loc);
-#endif
- memset(info, 0, sizeof(*info));
- GF_FREE(info);
-}
-
-int crypt_forget (xlator_t *this, inode_t *inode)
-{
- uint64_t ctx_addr = 0;
- if (!inode_ctx_del (inode, this, &ctx_addr))
- free_inode_info((struct crypt_inode_info *)(long)ctx_addr);
- return 0;
-}
-
-#if DEBUG_CRYPT
-static void check_read(call_frame_t *frame, xlator_t *this, int32_t read,
- struct iovec *vec, int32_t count, struct iatt *stbuf)
-{
- crypt_local_t *local = frame->local;
- struct object_cipher_info *object = get_object_cinfo(local->info);
- struct avec_config *conf = &local->data_conf;
- uint32_t resid = stbuf->ia_size & (object_alg_blksize(object) - 1);
-
- if (read <= 0)
- return;
- if (read != iov_length(vec, count))
- gf_log ("crypt", GF_LOG_DEBUG,
- "op_ret differs from amount of read bytes");
-
- if (object_alg_should_pad(object) && (read & (object_alg_blksize(object) - 1)))
- gf_log ("crypt", GF_LOG_DEBUG,
- "bad amount of read bytes (!= 0 mod(cblock size))");
-
- if (conf->aligned_offset + read >
- stbuf->ia_size + (resid ? object_alg_blksize(object) - resid : 0))
- gf_log ("crypt", GF_LOG_DEBUG,
- "bad amount of read bytes (too large))");
-
-}
-
-#define PT_BYTES_TO_DUMP (32)
-static void dump_plain_text(crypt_local_t *local, struct iovec *avec)
-{
- int32_t to_dump;
- char str[PT_BYTES_TO_DUMP + 1];
-
- if (!avec)
- return;
- to_dump = avec->iov_len;
- if (to_dump > PT_BYTES_TO_DUMP)
- to_dump = PT_BYTES_TO_DUMP;
- memcpy(str, avec->iov_base, to_dump);
- memset(str + to_dump, '0', 1);
- gf_log("crypt", GF_LOG_DEBUG, "Read file: %s", str);
-}
-
-static int32_t data_conf_invariant(struct avec_config *conf)
-{
- return conf->acount ==
- !!has_head_block(conf) +
- !!has_tail_block(conf)+
- conf->nr_full_blocks;
-}
-
-static int32_t hole_conf_invariant(struct avec_config *conf)
-{
- return conf->blocks_in_pool ==
- !!has_head_block(conf) +
- !!has_tail_block(conf)+
- !!has_full_blocks(conf);
-}
-
-static void crypt_check_conf(struct avec_config *conf)
-{
- int32_t ret = 0;
- const char *msg;
-
- switch (conf->type) {
- case DATA_ATOM:
- msg = "data";
- ret = data_conf_invariant(conf);
- break;
- case HOLE_ATOM:
- msg = "hole";
- ret = hole_conf_invariant(conf);
- break;
- default:
- msg = "unknown";
- }
- if (!ret)
- gf_log("crypt", GF_LOG_DEBUG, "bad %s conf", msg);
-}
-
-static void check_buf(call_frame_t *frame, xlator_t *this, struct iatt *buf)
-{
- crypt_local_t *local = frame->local;
- struct object_cipher_info *object = &local->info->cinfo;
- uint64_t local_file_size;
-
- switch(local->fop) {
- case GF_FOP_FTRUNCATE:
- return;
- case GF_FOP_WRITE:
- local_file_size = local->new_file_size;
- break;
- case GF_FOP_READ:
- if (parent_is_crypt_xlator(frame, this))
- return;
- local_file_size = local->cur_file_size;
- break;
- default:
- gf_log("crypt", GF_LOG_DEBUG, "bad file operation");
- return;
- }
- if (buf->ia_size != round_up(local_file_size,
- object_alg_blksize(object)))
- gf_log("crypt", GF_LOG_DEBUG,
- "bad ia_size in buf (%llu), should be %llu",
- (unsigned long long)buf->ia_size,
- (unsigned long long)round_up(local_file_size,
- object_alg_blksize(object)));
-}
-
-#else
-#define check_read(frame, this, op_ret, vec, count, stbuf) noop
-#define dump_plain_text(local, avec) noop
-#define crypt_check_conf(conf) noop
-#define check_buf(frame, this, buf) noop
-#endif /* DEBUG_CRYPT */
-
-/*
- * Pre-conditions:
- * @vec represents a ciphertext of expanded size and
- * aligned offset.
- *
- * Compound a temporal vector @avec with block-aligned
- * components, decrypt and fix it up to represent a chunk
- * of data corresponding to the original size and offset.
- * Pass the result to the next translator.
- */
-int32_t crypt_readv_cbk(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iovec *vec,
- int32_t count,
- struct iatt *stbuf,
- struct iobref *iobref,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
- struct avec_config *conf = &local->data_conf;
- struct object_cipher_info *object = &local->info->cinfo;
-
- struct iovec *avec;
- uint32_t i;
- uint32_t to_vec;
- uint32_t to_user;
-
- check_buf(frame, this, stbuf);
- check_read(frame, this, op_ret, vec, count, stbuf);
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- local->iobref = iobref_ref(iobref);
-
- local->buf = *stbuf;
- local->buf.ia_size = local->cur_file_size;
-
- if (op_ret <= 0 || count == 0 || vec[0].iov_len == 0)
- goto put_one_call;
-
- if (conf->orig_offset >= local->cur_file_size) {
- local->op_ret = 0;
- goto put_one_call;
- }
- /*
- * correct config params with real file size
- * and actual amount of bytes read
- */
- set_config_offsets(frame, this,
- conf->orig_offset, op_ret, DATA_ATOM, 0);
-
- if (conf->orig_offset + conf->orig_size > local->cur_file_size)
- conf->orig_size = local->cur_file_size - conf->orig_offset;
- /*
- * calculate amount of data to be returned
- * to user.
- */
- to_user = op_ret;
- if (conf->aligned_offset + to_user <= conf->orig_offset) {
- gf_log(this->name, GF_LOG_WARNING, "Incomplete read");
- local->op_ret = -1;
- local->op_errno = EIO;
- goto put_one_call;
- }
- to_user -= (conf->aligned_offset - conf->orig_offset);
-
- if (to_user > conf->orig_size)
- to_user = conf->orig_size;
- local->rw_count = to_user;
-
- op_errno = set_config_avec_data(this, local,
- conf, object, vec, count);
- if (op_errno) {
- local->op_ret = -1;
- local->op_errno = op_errno;
- goto put_one_call;
- }
- avec = conf->avec;
-#if DEBUG_CRYPT
- if (conf->off_in_tail != 0 &&
- conf->off_in_tail < object_alg_blksize(object) &&
- object_alg_should_pad(object))
- gf_log(this->name, GF_LOG_DEBUG, "Bad offset in tail %d",
- conf->off_in_tail);
- if (iov_length(vec, count) != 0 &&
- in_same_lblock(conf->orig_offset + iov_length(vec, count) - 1,
- local->cur_file_size - 1,
- object_alg_blkbits(object))) {
- gf_log(this->name, GF_LOG_DEBUG, "Compound last cblock");
- dump_cblock(this,
- (unsigned char *)(avec[conf->acount - 1].iov_base) +
- avec[conf->acount - 1].iov_len - object_alg_blksize(object));
- dump_cblock(this,
- (unsigned char *)(vec[count - 1].iov_base) +
- vec[count - 1].iov_len - object_alg_blksize(object));
- }
-#endif
- decrypt_aligned_iov(object, avec,
- conf->acount, conf->aligned_offset);
- /*
- * pass proper plain data to user
- */
- avec[0].iov_base += (conf->aligned_offset - conf->orig_offset);
- avec[0].iov_len -= (conf->aligned_offset - conf->orig_offset);
-
- to_vec = to_user;
- for (i = 0; i < conf->acount; i++) {
- if (avec[i].iov_len > to_vec)
- avec[i].iov_len = to_vec;
- to_vec -= avec[i].iov_len;
- }
- put_one_call:
- put_one_call_readv(frame, this);
- return 0;
-}
-
-static int32_t do_readv(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *dict,
- dict_t *xdata)
-{
- data_t *data;
- crypt_local_t *local = frame->local;
-
- if (op_ret < 0)
- goto error;
- /*
- * extract regular file size
- */
- data = dict_get(dict, FSIZE_XATTR_PREFIX);
- if (!data) {
- gf_log("crypt", GF_LOG_WARNING, "Regular file size not found");
- op_errno = EIO;
- goto error;
- }
- local->cur_file_size = data_to_uint64(data);
-
- get_one_call(frame);
- STACK_WIND(frame,
- crypt_readv_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->readv,
- local->fd,
- /*
- * FIXME: read amount can be reduced
- */
- local->data_conf.expanded_size,
- local->data_conf.aligned_offset,
- local->flags,
- local->xdata);
- return 0;
- error:
- local->op_ret = -1;
- local->op_errno = op_errno;
-
- get_one_call(frame);
- put_one_call_readv(frame, this);
- return 0;
-}
-
-static int32_t crypt_readv_finodelk_cbk(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
-
- if (op_ret < 0)
- goto error;
- /*
- * An access has been granted,
- * retrieve file size
- */
- STACK_WIND(frame,
- do_readv,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fgetxattr,
- local->fd,
- FSIZE_XATTR_PREFIX,
- NULL);
- return 0;
- error:
- fd_unref(local->fd);
- if (local->xdata)
- dict_unref(local->xdata);
- STACK_UNWIND_STRICT(readv,
- frame,
- -1,
- op_errno,
- NULL,
- 0,
- NULL,
- NULL,
- NULL);
- return 0;
-}
-
-static int32_t readv_trivial_completion(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iatt *buf,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
-
- if (op_ret < 0) {
- gf_log(this->name, GF_LOG_WARNING,
- "stat failed (%d)", op_errno);
- goto error;
- }
- local->buf = *buf;
- STACK_WIND(frame,
- load_file_size,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr,
- local->loc,
- FSIZE_XATTR_PREFIX,
- NULL);
- return 0;
- error:
- STACK_UNWIND_STRICT(readv, frame, op_ret, op_errno,
- NULL, 0, NULL, NULL, NULL);
- return 0;
-}
-
-int32_t crypt_readv(call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- size_t size,
- off_t offset,
- uint32_t flags, dict_t *xdata)
-{
- int32_t ret;
- crypt_local_t *local;
- struct crypt_inode_info *info;
- struct gf_flock lock = {0, };
-
-#if DEBUG_CRYPT
- gf_log("crypt", GF_LOG_DEBUG, "reading %d bytes from offset %llu",
- (int)size, (long long)offset);
- if (parent_is_crypt_xlator(frame, this))
- gf_log("crypt", GF_LOG_DEBUG, "parent is crypt");
-#endif
- local = crypt_alloc_local(frame, this, GF_FOP_READ);
- if (!local) {
- ret = ENOMEM;
- goto error;
- }
- if (size == 0)
- goto trivial;
-
- local->fd = fd_ref(fd);
- local->flags = flags;
-
- info = local_get_inode_info(local, this);
- if (info == NULL) {
- ret = EINVAL;
- fd_unref(fd);
- goto error;
- }
- if (!object_alg_atomic(&info->cinfo)) {
- ret = EINVAL;
- fd_unref(fd);
- goto error;
- }
- set_config_offsets(frame, this, offset, size,
- DATA_ATOM, 0);
- if (parent_is_crypt_xlator(frame, this)) {
- data_t *data;
- /*
- * We are called by crypt_writev (or cypt_ftruncate)
- * to perform the "read" component of the read-modify-write
- * (or read-prune-write) sequence for some atom;
- *
- * don't ask for access:
- * it has already been acquired
- *
- * Retrieve current file size
- */
- if (!xdata) {
- gf_log("crypt", GF_LOG_WARNING,
- "Regular file size hasn't been passed");
- ret = EIO;
- goto error;
- }
- data = dict_get(xdata, FSIZE_XATTR_PREFIX);
- if (!data) {
- gf_log("crypt", GF_LOG_WARNING,
- "Regular file size not found");
- ret = EIO;
- goto error;
- }
- local->old_file_size =
- local->cur_file_size = data_to_uint64(data);
-
- get_one_call(frame);
- STACK_WIND(frame,
- crypt_readv_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readv,
- local->fd,
- /*
- * FIXME: read amount can be reduced
- */
- local->data_conf.expanded_size,
- local->data_conf.aligned_offset,
- flags,
- NULL);
- return 0;
- }
- if (xdata)
- local->xdata = dict_ref(xdata);
-
- lock.l_len = 0;
- lock.l_start = 0;
- lock.l_type = F_RDLCK;
- lock.l_whence = SEEK_SET;
-
- STACK_WIND(frame,
- crypt_readv_finodelk_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->finodelk,
- this->name,
- fd,
- F_SETLKW,
- &lock,
- NULL);
- return 0;
- trivial:
- STACK_WIND(frame,
- readv_trivial_completion,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fstat,
- fd,
- NULL);
- return 0;
- error:
- STACK_UNWIND_STRICT(readv,
- frame,
- -1,
- ret,
- NULL,
- 0,
- NULL,
- NULL,
- NULL);
- return 0;
-}
-
-void set_local_io_params_writev(call_frame_t *frame,
- struct object_cipher_info *object,
- struct rmw_atom *atom,
- off_t io_offset,
- uint32_t io_size)
-{
- crypt_local_t *local = frame->local;
-
- local->io_offset = io_offset;
- local->io_size = io_size;
-
- local->io_offset_nopad =
- atom->offset_at(frame, object) + atom->offset_in(frame, object);
-
- gf_log("crypt", GF_LOG_DEBUG,
- "set nopad offset to %llu",
- (unsigned long long)local->io_offset_nopad);
-
- local->io_size_nopad = atom->io_size_nopad(frame, object);
-
- gf_log("crypt", GF_LOG_DEBUG,
- "set nopad size to %llu",
- (unsigned long long)local->io_size_nopad);
-
- local->update_disk_file_size = 0;
- /*
- * NOTE: eof_padding_size is 0 for all full atoms;
- * For head and tail atoms it will be set up at rmw_partial block()
- */
- local->new_file_size = local->cur_file_size;
-
- if (local->io_offset_nopad + local->io_size_nopad > local->cur_file_size) {
-
- local->new_file_size = local->io_offset_nopad + local->io_size_nopad;
-
- gf_log("crypt", GF_LOG_DEBUG,
- "set new file size to %llu",
- (unsigned long long)local->new_file_size);
-
- local->update_disk_file_size = 1;
- }
-}
-
-void set_local_io_params_ftruncate(call_frame_t *frame,
- struct object_cipher_info *object)
-{
- uint32_t resid;
- crypt_local_t *local = frame->local;
- struct avec_config *conf = &local->data_conf;
-
- resid = conf->orig_offset & (object_alg_blksize(object) - 1);
- if (resid) {
- local->eof_padding_size =
- object_alg_blksize(object) - resid;
- local->new_file_size = conf->aligned_offset;
- local->update_disk_file_size = 0;
- /*
- * file size will be updated
- * in the ->writev() stack,
- * when submitting file tail
- */
- } else {
- local->eof_padding_size = 0;
- local->new_file_size = conf->orig_offset;
- local->update_disk_file_size = 1;
- /*
- * file size will be updated
- * in this ->ftruncate stack
- */
- }
-}
-
-static void submit_head(call_frame_t *frame, xlator_t *this)
-{
- crypt_local_t *local = frame->local;
- submit_partial(frame, this, local->fd, HEAD_ATOM);
-}
-
-static void submit_tail(call_frame_t *frame, xlator_t *this)
-{
- crypt_local_t *local = frame->local;
- submit_partial(frame, this, local->fd, TAIL_ATOM);
-}
-
-static void submit_hole(call_frame_t *frame, xlator_t *this)
-{
- /*
- * hole conversion always means
- * appended write and goes in ordered fashion
- */
- do_ordered_submit(frame, this, HOLE_ATOM);
-}
-
-static void submit_data(call_frame_t *frame, xlator_t *this)
-{
- if (is_ordered_mode(frame)) {
- do_ordered_submit(frame, this, DATA_ATOM);
- return;
- }
- gf_log("crypt", GF_LOG_WARNING, "Bad submit mode");
- get_nr_calls(frame, nr_calls_data(frame));
- do_parallel_submit(frame, this, DATA_ATOM);
- return;
-}
-
-/*
- * heplers called by writev_cbk, fruncate_cbk in ordered mode
- */
-
-static int32_t should_submit_hole(crypt_local_t *local)
-{
- struct avec_config *conf = &local->hole_conf;
-
- return conf->avec != NULL;
-}
-
-static int32_t should_resume_submit_hole(crypt_local_t *local)
-{
- struct avec_config *conf = &local->hole_conf;
-
- if (local->fop == GF_FOP_WRITE && has_tail_block(conf))
- /*
- * Don't submit a part of hole, which
- * fits into a data block:
- * this part of hole will be converted
- * as a gap filled by zeros in data head
- * block.
- */
- return conf->cursor < conf->acount - 1;
- else
- return conf->cursor < conf->acount;
-}
-
-static int32_t should_resume_submit_data(call_frame_t *frame)
-{
- crypt_local_t *local = frame->local;
- struct avec_config *conf = &local->data_conf;
-
- if (is_ordered_mode(frame))
- return conf->cursor < conf->acount;
- /*
- * parallel writes
- */
- return 0;
-}
-
-static int32_t should_submit_data_after_hole(crypt_local_t *local)
-{
- return local->data_conf.avec != NULL;
-}
-
-static void update_local_file_params(call_frame_t *frame,
- xlator_t *this,
- struct iatt *prebuf,
- struct iatt *postbuf)
-{
- crypt_local_t *local = frame->local;
-
- check_buf(frame, this, postbuf);
-
- local->prebuf = *prebuf;
- local->postbuf = *postbuf;
-
- local->prebuf.ia_size = local->cur_file_size;
- local->postbuf.ia_size = local->new_file_size;
-
- local->cur_file_size = local->new_file_size;
-}
-
-static int32_t end_writeback_writev(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iatt *prebuf,
- struct iatt *postbuf,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
-
- if (op_ret <= 0) {
- gf_log(this->name, GF_LOG_WARNING,
- "writev iteration failed");
- goto put_one_call;
- }
- /*
- * op_ret includes paddings (atom's head, atom's tail and EOF)
- */
- if (op_ret < local->io_size) {
- gf_log(this->name, GF_LOG_WARNING,
- "Incomplete writev iteration");
- goto put_one_call;
- }
- op_ret -= local->eof_padding_size;
- local->op_ret = op_ret;
-
- update_local_file_params(frame, this, prebuf, postbuf);
-
- if (data_write_in_progress(local)) {
-
- LOCK(&local->rw_count_lock);
- local->rw_count += op_ret;
- UNLOCK(&local->rw_count_lock);
-
- if (should_resume_submit_data(frame))
- submit_data(frame, this);
- }
- else {
- /*
- * hole conversion is going on;
- * don't take into account written zeros
- */
- if (should_resume_submit_hole(local))
- submit_hole(frame, this);
-
- else if (should_submit_data_after_hole(local))
- submit_data(frame, this);
- }
- put_one_call:
- put_one_call_writev(frame, this);
- return 0;
-}
-
-#define crypt_writev_cbk end_writeback_writev
-
-#define HOLE_WRITE_CHUNK_BITS 12
-#define HOLE_WRITE_CHUNK_SIZE (1 << HOLE_WRITE_CHUNK_BITS)
-
-/*
- * Convert hole of size @size at offset @off to
- * zeros and prepare respective iovecs for submit.
- * The hole lock should be held.
- *
- * Pre-conditions:
- * @local->file_size is set and valid.
- */
-int32_t prepare_for_submit_hole(call_frame_t *frame, xlator_t *this,
- uint64_t off, off_t size)
-{
- int32_t ret;
- crypt_local_t *local = frame->local;
- struct object_cipher_info *object = &local->info->cinfo;
-
- set_config_offsets(frame, this, off, size, HOLE_ATOM, 1);
-
- ret = set_config_avec_hole(this, local,
- &local->hole_conf, object, local->fop);
- crypt_check_conf(&local->hole_conf);
-
- return ret;
-}
-
-/*
- * prepare for submit @count bytes at offset @from
- */
-int32_t prepare_for_submit_data(call_frame_t *frame, xlator_t *this,
- off_t from, int32_t size, struct iovec *vec,
- int32_t vec_count, int32_t setup_gap)
-{
- uint32_t ret;
- crypt_local_t *local = frame->local;
- struct object_cipher_info *object = &local->info->cinfo;
-
- set_config_offsets(frame, this, from, size,
- DATA_ATOM, setup_gap);
-
- ret = set_config_avec_data(this, local,
- &local->data_conf, object, vec, vec_count);
- crypt_check_conf(&local->data_conf);
-
- return ret;
-}
-
-static void free_avec(struct iovec *avec,
- char **pool, int blocks_in_pool)
-{
- if (!avec)
- return;
- GF_FREE(pool);
- GF_FREE(avec);
-}
-
-static void free_avec_data(crypt_local_t *local)
-{
- return free_avec(local->data_conf.avec,
- local->data_conf.pool,
- local->data_conf.blocks_in_pool);
-}
-
-static void free_avec_hole(crypt_local_t *local)
-{
- return free_avec(local->hole_conf.avec,
- local->hole_conf.pool,
- local->hole_conf.blocks_in_pool);
-}
-
-
-static void do_parallel_submit(call_frame_t *frame, xlator_t *this,
- atom_data_type dtype)
-{
- crypt_local_t *local = frame->local;
- struct avec_config *conf;
-
- local->active_setup = dtype;
- conf = conf_by_type(frame, dtype);
-
- if (has_head_block(conf))
- submit_head(frame, this);
-
- if (has_full_blocks(conf))
- submit_full(frame, this);
-
- if (has_tail_block(conf))
- submit_tail(frame, this);
- return;
-}
-
-static void do_ordered_submit(call_frame_t *frame, xlator_t *this,
- atom_data_type dtype)
-{
- crypt_local_t *local = frame->local;
- struct avec_config *conf;
-
- local->active_setup = dtype;
- conf = conf_by_type(frame, dtype);
-
- if (should_submit_head_block(conf)) {
- get_one_call_nolock(frame);
- submit_head(frame, this);
- }
- else if (should_submit_full_block(conf)) {
- get_one_call_nolock(frame);
- submit_full(frame, this);
- }
- else if (should_submit_tail_block(conf)) {
- get_one_call_nolock(frame);
- submit_tail(frame, this);
- }
- else
- gf_log("crypt", GF_LOG_DEBUG,
- "nothing has been submitted in ordered mode");
- return;
-}
-
-static int32_t do_writev(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *dict,
- dict_t *xdata)
-{
- data_t *data;
- crypt_local_t *local = frame->local;
- struct object_cipher_info *object = &local->info->cinfo;
- /*
- * extract regular file size
- */
- data = dict_get(dict, FSIZE_XATTR_PREFIX);
- if (!data) {
- gf_log("crypt", GF_LOG_WARNING, "Regular file size not found");
- op_ret = -1;
- op_errno = EIO;
- goto error;
- }
- local->old_file_size = local->cur_file_size = data_to_uint64(data);
-
- set_gap_at_end(frame, object, &local->data_conf, DATA_ATOM);
-
- if (local->cur_file_size < local->data_conf.orig_offset) {
- /*
- * Set up hole config
- */
- op_errno = prepare_for_submit_hole(frame,
- this,
- local->cur_file_size,
- local->data_conf.orig_offset - local->cur_file_size);
- if (op_errno) {
- local->op_ret = -1;
- local->op_errno = op_errno;
- goto error;
- }
- }
- if (should_submit_hole(local))
- submit_hole(frame, this);
- else
- submit_data(frame, this);
- return 0;
- error:
- get_one_call_nolock(frame);
- put_one_call_writev(frame, this);
- return 0;
-}
-
-static int32_t crypt_writev_finodelk_cbk(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
-
- if (op_ret < 0)
- goto error;
- /*
- * An access has been granted,
- * retrieve file size first
- */
- STACK_WIND(frame,
- do_writev,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fgetxattr,
- local->fd,
- FSIZE_XATTR_PREFIX,
- NULL);
- return 0;
- error:
- get_one_call_nolock(frame);
- put_one_call_writev(frame, this);
- return 0;
-}
-
-static int32_t writev_trivial_completion(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iatt *buf,
- dict_t *dict)
-{
- crypt_local_t *local = frame->local;
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- local->prebuf = *buf;
- local->postbuf = *buf;
-
- local->prebuf.ia_size = local->cur_file_size;
- local->postbuf.ia_size = local->cur_file_size;
-
- get_one_call(frame);
- put_one_call_writev(frame, this);
- return 0;
-}
-
-int crypt_writev(call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- struct iovec *vec,
- int32_t count,
- off_t offset,
- uint32_t flags,
- struct iobref *iobref,
- dict_t *xdata)
-{
- int32_t ret;
- crypt_local_t *local;
- struct crypt_inode_info *info;
- struct gf_flock lock = {0, };
-#if DEBUG_CRYPT
- gf_log ("crypt", GF_LOG_DEBUG, "writing %d bytes from offset %llu",
- (int)iov_length(vec, count), (long long)offset);
-#endif
- local = crypt_alloc_local(frame, this, GF_FOP_WRITE);
- if (!local) {
- ret = ENOMEM;
- goto error;
- }
- local->fd = fd_ref(fd);
-
- if (iobref)
- local->iobref = iobref_ref(iobref);
- /*
- * to update real file size on the server
- */
- local->xattr = dict_new();
- if (!local->xattr) {
- ret = ENOMEM;
- goto error;
- }
- local->flags = flags;
-
- info = local_get_inode_info(local, this);
- if (info == NULL) {
- ret = EINVAL;
- goto error;
- }
- if (!object_alg_atomic(&info->cinfo)) {
- ret = EINVAL;
- goto error;
- }
- if (iov_length(vec, count) == 0)
- goto trivial;
-
- ret = prepare_for_submit_data(frame, this, offset,
- iov_length(vec, count),
- vec, count, 0 /* don't setup gup
- in tail: we don't
- know file size yet */);
- if (ret)
- goto error;
-
- if (parent_is_crypt_xlator(frame, this)) {
- data_t *data;
- /*
- * we are called by shinking crypt_ftruncate(),
- * which doesn't perform hole conversion;
- *
- * don't ask for access:
- * it has already been acquired
- */
-
- /*
- * extract file size
- */
- if (!xdata) {
- gf_log("crypt", GF_LOG_WARNING,
- "Regular file size hasn't been passed");
- ret = EIO;
- goto error;
- }
- data = dict_get(xdata, FSIZE_XATTR_PREFIX);
- if (!data) {
- gf_log("crypt", GF_LOG_WARNING,
- "Regular file size not found");
- ret = EIO;
- goto error;
- }
- local->old_file_size =
- local->cur_file_size = data_to_uint64(data);
-
- submit_data(frame, this);
- return 0;
- }
- if (xdata)
- local->xdata = dict_ref(xdata);
- /*
- * lock the file and retrieve its size
- */
- lock.l_len = 0;
- lock.l_start = 0;
- lock.l_type = F_WRLCK;
- lock.l_whence = SEEK_SET;
-
- STACK_WIND(frame,
- crypt_writev_finodelk_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->finodelk,
- this->name,
- fd,
- F_SETLKW,
- &lock,
- NULL);
- return 0;
- trivial:
- STACK_WIND(frame,
- writev_trivial_completion,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fstat,
- fd,
- NULL);
- return 0;
- error:
- if (local && local->fd)
- fd_unref(fd);
- if (local && local->iobref)
- iobref_unref(iobref);
- if (local && local->xdata)
- dict_unref(xdata);
- if (local && local->xattr)
- dict_unref(local->xattr);
- if (local && local->info)
- free_inode_info(local->info);
-
- STACK_UNWIND_STRICT(writev, frame, -1, ret, NULL, NULL, NULL);
- return 0;
-}
-
-int32_t prepare_for_prune(call_frame_t *frame, xlator_t *this, uint64_t offset)
-{
- set_config_offsets(frame, this,
- offset,
- 0, /* count */
- DATA_ATOM,
- 0 /* since we prune, there is no
- gap in tail to uptodate */);
- return 0;
-}
-
-/*
- * Finish the read-prune-modify sequence
- *
- * Can be invoked as
- * 1) ->ftruncate_cbk() for cblock-aligned, or trivial prune
- * 2) ->writev_cbk() for non-cblock-aligned prune
- */
-
-static int32_t prune_complete(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iatt *prebuf,
- struct iatt *postbuf,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
-
- update_local_file_params(frame, this, prebuf, postbuf);
-
- put_one_call_ftruncate(frame, this);
- return 0;
-}
-
-/*
- * This is called as ->ftruncate_cbk()
- *
- * Perform the "write" component of the
- * read-prune-write sequence.
- *
- * submuit the rest of the file
- */
-static int32_t prune_submit_file_tail(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iatt *prebuf,
- struct iatt *postbuf,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
- struct avec_config *conf = &local->data_conf;
- dict_t *dict;
-
- if (op_ret < 0)
- goto put_one_call;
-
- if (local->xdata) {
- dict_unref(local->xdata);
- local->xdata = NULL;
- }
- if (xdata)
- local->xdata = dict_ref(xdata);
-
- dict = dict_new();
- if (!dict) {
- op_errno = ENOMEM;
- goto error;
- }
-
- update_local_file_params(frame, this, prebuf, postbuf);
- local->new_file_size = conf->orig_offset;
-
- /*
- * The rest of the file is a partial block and, hence,
- * should be written via RMW sequence, so the crypt xlator
- * does STACK_WIND to itself.
- *
- * Pass current file size to crypt_writev()
- */
- op_errno = dict_set(dict,
- FSIZE_XATTR_PREFIX,
- data_from_uint64(local->cur_file_size));
- if (op_errno) {
- gf_log("crypt", GF_LOG_WARNING,
- "can not set key to update file size");
- dict_unref(dict);
- goto error;
- }
- gf_log("crypt", GF_LOG_DEBUG,
- "passing current file size (%llu) to crypt_writev",
- (unsigned long long)local->cur_file_size);
- /*
- * Padding will be filled with
- * zeros by rmw_partial_block()
- */
- STACK_WIND(frame,
- prune_complete,
- this,
- this->fops->writev, /* crypt_writev */
- local->fd,
- &local->vec,
- 1,
- conf->aligned_offset, /* offset to write from */
- 0,
- local->iobref,
- dict);
-
- dict_unref(dict);
- return 0;
- error:
- local->op_ret = -1;
- local->op_errno = op_errno;
- put_one_call:
- put_one_call_ftruncate(frame, this);
- return 0;
-}
-
-/*
- * This is called as a callback of ->writev() invoked in behalf
- * of ftruncate(): it can be
- * 1) ordered writes issued by hole conversion in the case of
- * expanded truncate, or
- * 2) an rmw partial data block issued by non-cblock-aligned
- * prune.
- */
-int32_t end_writeback_ftruncate(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iatt *prebuf,
- struct iatt *postbuf,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
- /*
- * if nothing has been written,
- * then it must be an error
- */
- local->op_ret = op_ret;
- local->op_errno = op_errno;
-
- if (op_ret < 0)
- goto put_one_call;
-
- update_local_file_params(frame, this, prebuf, postbuf);
-
- if (data_write_in_progress(local))
- /* case (2) */
- goto put_one_call;
- /* case (1) */
- if (should_resume_submit_hole(local))
- submit_hole(frame, this);
- /*
- * case of hole, when we should't resume
- */
- put_one_call:
- put_one_call_ftruncate(frame, this);
- return 0;
-}
-
-/*
- * Perform prune and write components of the
- * read-prune-write sequence.
- *
- * Called as ->readv_cbk()
- *
- * Pre-conditions:
- * @vec contains the latest atom of the file
- * (plain text)
- */
-static int32_t prune_write(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iovec *vec,
- int32_t count,
- struct iatt *stbuf,
- struct iobref *iobref,
- dict_t *xdata)
-{
- int32_t i;
- size_t to_copy;
- size_t copied = 0;
- crypt_local_t *local = frame->local;
- struct avec_config *conf = &local->data_conf;
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- if (op_ret == -1)
- goto put_one_call;
-
- /*
- * At first, uptodate head block
- */
- if (iov_length(vec, count) < conf->off_in_head) {
- gf_log(this->name, GF_LOG_WARNING,
- "Failed to uptodate head block for prune");
- local->op_ret = -1;
- local->op_errno = EIO;
- goto put_one_call;
- }
- local->vec.iov_len = conf->off_in_head;
- local->vec.iov_base = GF_CALLOC(1, local->vec.iov_len,
- gf_crypt_mt_data);
-
- if (local->vec.iov_base == NULL) {
- gf_log(this->name, GF_LOG_WARNING,
- "Failed to calloc head block for prune");
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto put_one_call;
- }
- for (i = 0; i < count; i++) {
- to_copy = vec[i].iov_len;
- if (to_copy > local->vec.iov_len - copied)
- to_copy = local->vec.iov_len - copied;
-
- memcpy((char *)local->vec.iov_base + copied,
- vec[i].iov_base,
- to_copy);
- copied += to_copy;
- if (copied == local->vec.iov_len)
- break;
- }
- /*
- * perform prune with aligned offset
- * (i.e. at this step we prune a bit
- * more then it is needed
- */
- STACK_WIND(frame,
- prune_submit_file_tail,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->ftruncate,
- local->fd,
- conf->aligned_offset,
- local->xdata);
- return 0;
- put_one_call:
- put_one_call_ftruncate(frame, this);
- return 0;
-}
-
-/*
- * Perform a read-prune-write sequence
- */
-int32_t read_prune_write(call_frame_t *frame, xlator_t *this)
-{
- int32_t ret = 0;
- dict_t *dict = NULL;
- crypt_local_t *local = frame->local;
- struct avec_config *conf = &local->data_conf;
- struct object_cipher_info *object = &local->info->cinfo;
-
- set_local_io_params_ftruncate(frame, object);
- get_one_call_nolock(frame);
-
- if ((conf->orig_offset & (object_alg_blksize(object) - 1)) == 0) {
- /*
- * cblock-aligned prune:
- * we don't need read and write components,
- * just cut file body
- */
- gf_log("crypt", GF_LOG_DEBUG,
- "prune without RMW (at offset %llu",
- (unsigned long long)conf->orig_offset);
-
- STACK_WIND(frame,
- prune_complete,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->ftruncate,
- local->fd,
- conf->orig_offset,
- local->xdata);
- return 0;
- }
- gf_log("crypt", GF_LOG_DEBUG,
- "prune with RMW (at offset %llu",
- (unsigned long long)conf->orig_offset);
- /*
- * We are about to perform the "read" component of the
- * read-prune-write sequence. It means that we need to
- * read encrypted data from disk and decrypt it.
- * So, the crypt translator does STACK_WIND to itself.
- *
- * Pass current file size to crypt_readv()
-
- */
- dict = dict_new();
- if (!dict) {
- gf_log("crypt", GF_LOG_WARNING, "Can not alloc dict");
- ret = ENOMEM;
- goto exit;
- }
- ret = dict_set(dict,
- FSIZE_XATTR_PREFIX,
- data_from_uint64(local->cur_file_size));
- if (ret) {
- gf_log("crypt", GF_LOG_WARNING, "Can not set dict");
- goto exit;
- }
- STACK_WIND(frame,
- prune_write,
- this,
- this->fops->readv, /* crypt_readv */
- local->fd,
- get_atom_size(object), /* bytes to read */
- conf->aligned_offset, /* offset to read from */
- 0,
- dict);
- exit:
- if (dict)
- dict_unref(dict);
- return ret;
-}
-
-/*
- * File prune is more complicated than expand.
- * First we need to read the latest atom to not lose info
- * needed for proper update. Also we need to make sure that
- * every component of read-prune-write sequence leaves data
- * consistent
- *
- * Non-cblock aligned prune is performed as read-prune-write
- * sequence:
- *
- * 1) read the latest atom;
- * 2) perform cblock-aligned prune
- * 3) issue a write request for the end-of-file
- */
-int32_t prune_file(call_frame_t *frame, xlator_t *this, uint64_t offset)
-{
- int32_t ret;
-
- ret = prepare_for_prune(frame, this, offset);
- if (ret)
- return ret;
- return read_prune_write(frame, this);
-}
-
-int32_t expand_file(call_frame_t *frame, xlator_t *this,
- uint64_t offset)
-{
- int32_t ret;
- crypt_local_t *local = frame->local;
-
- ret = prepare_for_submit_hole(frame, this,
- local->old_file_size,
- offset - local->old_file_size);
- if (ret)
- return ret;
- submit_hole(frame, this);
- return 0;
-}
-
-static int32_t ftruncate_trivial_completion(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iatt *buf,
- dict_t *dict)
-{
- crypt_local_t *local = frame->local;
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- local->prebuf = *buf;
- local->postbuf = *buf;
-
- local->prebuf.ia_size = local->cur_file_size;
- local->postbuf.ia_size = local->cur_file_size;
-
- get_one_call(frame);
- put_one_call_ftruncate(frame, this);
- return 0;
-}
-
-static int32_t do_ftruncate(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *dict,
- dict_t *xdata)
-{
- data_t *data;
- crypt_local_t *local = frame->local;
-
- if (op_ret)
- goto error;
- /*
- * extract regular file size
- */
- data = dict_get(dict, FSIZE_XATTR_PREFIX);
- if (!data) {
- gf_log("crypt", GF_LOG_WARNING, "Regular file size not found");
- op_errno = EIO;
- goto error;
- }
- local->old_file_size = local->cur_file_size = data_to_uint64(data);
-
- if (local->data_conf.orig_offset == local->cur_file_size) {
-#if DEBUG_CRYPT
- gf_log("crypt", GF_LOG_DEBUG,
- "trivial ftruncate (current file size %llu)",
- (unsigned long long)local->cur_file_size);
-#endif
- goto trivial;
- }
- else if (local->data_conf.orig_offset < local->cur_file_size) {
-#if DEBUG_CRYPT
- gf_log("crypt", GF_LOG_DEBUG, "prune from %llu to %llu",
- (unsigned long long)local->cur_file_size,
- (unsigned long long)local->data_conf.orig_offset);
-#endif
- op_errno = prune_file(frame,
- this,
- local->data_conf.orig_offset);
- }
- else {
-#if DEBUG_CRYPT
- gf_log("crypt", GF_LOG_DEBUG, "expand from %llu to %llu",
- (unsigned long long)local->cur_file_size,
- (unsigned long long)local->data_conf.orig_offset);
-#endif
- op_errno = expand_file(frame,
- this,
- local->data_conf.orig_offset);
- }
- if (op_errno)
- goto error;
- return 0;
- trivial:
- STACK_WIND(frame,
- ftruncate_trivial_completion,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fstat,
- local->fd,
- NULL);
- return 0;
- error:
- /*
- * finish with ftruncate
- */
- local->op_ret = -1;
- local->op_errno = op_errno;
-
- get_one_call_nolock(frame);
- put_one_call_ftruncate(frame, this);
- return 0;
-}
-
-static int32_t crypt_ftruncate_finodelk_cbk(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
-
- if (op_ret < 0)
- goto error;
- /*
- * An access has been granted,
- * retrieve file size first
- */
- STACK_WIND(frame,
- do_ftruncate,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fgetxattr,
- local->fd,
- FSIZE_XATTR_PREFIX,
- NULL);
- return 0;
- error:
- get_one_call_nolock(frame);
- put_one_call_ftruncate(frame, this);
- return 0;
-}
-
-/*
- * ftruncate is performed in 2 steps:
- * . receive file size;
- * . expand or prune file.
- */
-static int32_t crypt_ftruncate(call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- off_t offset,
- dict_t *xdata)
-{
- int32_t ret;
- crypt_local_t *local;
- struct crypt_inode_info *info;
- struct gf_flock lock = {0, };
-
- local = crypt_alloc_local(frame, this, GF_FOP_FTRUNCATE);
- if (!local) {
- ret = ENOMEM;
- goto error;
- }
- local->xattr = dict_new();
- if (!local->xattr) {
- ret = ENOMEM;
- goto error;
- }
- local->fd = fd_ref(fd);
- info = local_get_inode_info(local, this);
- if (info == NULL) {
- ret = EINVAL;
- goto error;
- }
- if (!object_alg_atomic(&info->cinfo)) {
- ret = EINVAL;
- goto error;
- }
- local->data_conf.orig_offset = offset;
- if (xdata)
- local->xdata = dict_ref(xdata);
-
- lock.l_len = 0;
- lock.l_start = 0;
- lock.l_type = F_WRLCK;
- lock.l_whence = SEEK_SET;
-
- STACK_WIND(frame,
- crypt_ftruncate_finodelk_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->finodelk,
- this->name,
- fd,
- F_SETLKW,
- &lock,
- NULL);
- return 0;
- error:
- if (local && local->fd)
- fd_unref(fd);
- if (local && local->xdata)
- dict_unref(xdata);
- if (local && local->xattr)
- dict_unref(local->xattr);
- if (local && local->info)
- free_inode_info(local->info);
-
- STACK_UNWIND_STRICT(ftruncate, frame, -1, ret, NULL, NULL, NULL);
- return 0;
-}
-
-/* ->flush_cbk() */
-int32_t truncate_end(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
-
- STACK_UNWIND_STRICT(truncate,
- frame,
- op_ret,
- op_errno,
- &local->prebuf,
- &local->postbuf,
- local->xdata);
- return 0;
-}
-
-/* ftruncate_cbk() */
-int32_t truncate_flush(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iatt *prebuf,
- struct iatt *postbuf,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
- fd_t *fd = local->fd;
- local->prebuf = *prebuf;
- local->postbuf = *postbuf;
-
- STACK_WIND(frame,
- truncate_end,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->flush,
- fd,
- NULL);
- fd_unref(fd);
- return 0;
-}
-
-/*
- * is called as ->open_cbk()
- */
-static int32_t truncate_begin(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- fd_t *fd,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
-
- if (op_ret < 0) {
- fd_unref(fd);
- STACK_UNWIND_STRICT(truncate,
- frame,
- op_ret,
- op_errno, NULL, NULL, NULL);
- return 0;
- } else {
- fd_bind (fd);
- }
- /*
- * crypt_truncate() is implemented via crypt_ftruncate(),
- * so the crypt xlator does STACK_WIND to itself here
- */
- STACK_WIND(frame,
- truncate_flush,
- this,
- this->fops->ftruncate, /* crypt_ftruncate */
- fd,
- local->offset,
- NULL);
- return 0;
-}
-
-/*
- * crypt_truncate() is implemented via crypt_ftruncate() as a
- * sequence crypt_open() - crypt_ftruncate() - truncate_flush()
- */
-int32_t crypt_truncate(call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- off_t offset,
- dict_t *xdata)
-{
- fd_t *fd;
- crypt_local_t *local;
-
-#if DEBUG_CRYPT
- gf_log(this->name, GF_LOG_DEBUG,
- "truncate file %s at offset %llu",
- loc->path, (unsigned long long)offset);
-#endif
- local = crypt_alloc_local(frame, this, GF_FOP_TRUNCATE);
- if (!local)
- goto error;
-
- fd = fd_create(loc->inode, frame->root->pid);
- if (!fd) {
- gf_log(this->name, GF_LOG_ERROR, "Can not create fd");
- goto error;
- }
- local->fd = fd;
- local->offset = offset;
- local->xdata = xdata;
- STACK_WIND(frame,
- truncate_begin,
- this,
- this->fops->open, /* crypt_open() */
- loc,
- O_RDWR,
- fd,
- NULL);
- return 0;
- error:
- STACK_UNWIND_STRICT(truncate, frame, -1, EINVAL, NULL, NULL, NULL);
- return 0;
-}
-
-end_writeback_handler_t dispatch_end_writeback(glusterfs_fop_t fop)
-{
- switch (fop) {
- case GF_FOP_WRITE:
- return end_writeback_writev;
- case GF_FOP_FTRUNCATE:
- return end_writeback_ftruncate;
- default:
- gf_log("crypt", GF_LOG_WARNING, "Bad wb operation %d", fop);
- return NULL;
- }
-}
-
-/*
- * true, if the caller needs metadata string
- */
-static int32_t is_custom_mtd(dict_t *xdata)
-{
- data_t *data;
- uint32_t flags;
-
- if (!xdata)
- return 0;
-
- data = dict_get(xdata, MSGFLAGS_PREFIX);
- if (!data)
- return 0;
- if (data->len != sizeof(uint32_t)) {
- gf_log("crypt", GF_LOG_WARNING,
- "Bad msgflags size (%d)", data->len);
- return -1;
- }
- flags = *((uint32_t *)data->data);
- return msgflags_check_mtd_lock(&flags);
-}
-
-static int32_t crypt_open_done(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- if (op_ret < 0)
- gf_log(this->name, GF_LOG_WARNING, "mtd unlock failed (%d)",
- op_errno);
- put_one_call_open(frame);
- return 0;
-}
-
-static void crypt_open_tail(call_frame_t *frame, xlator_t *this)
-{
- struct gf_flock lock = {0, };
- crypt_local_t *local = frame->local;
-
- lock.l_type = F_UNLCK;
- lock.l_whence = SEEK_SET;
- lock.l_start = 0;
- lock.l_len = 0;
- lock.l_pid = 0;
-
- STACK_WIND(frame,
- crypt_open_done,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->finodelk,
- this->name,
- local->fd,
- F_SETLKW,
- &lock,
- NULL);
-}
-
-/*
- * load private inode info at open time
- * called as ->fgetxattr_cbk()
- */
-static int load_mtd_open(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *dict,
- dict_t *xdata)
-{
- int32_t ret;
- gf_boolean_t upload_info;
- data_t *mtd;
- uint64_t value = 0;
- struct crypt_inode_info *info;
- crypt_local_t *local = frame->local;
- crypt_private_t *priv = this->private;
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
-
- if (local->fd->inode->ia_type == IA_IFLNK)
- goto exit;
- if (op_ret < 0)
- goto exit;
- /*
- * first, check for cached info
- */
- ret = inode_ctx_get(local->fd->inode, this, &value);
- if (ret != -1) {
- info = (struct crypt_inode_info *)(long)value;
- if (info == NULL) {
- gf_log(this->name, GF_LOG_WARNING,
- "Inode info expected, but not found");
- local->op_ret = -1;
- local->op_errno = EIO;
- goto exit;
- }
- /*
- * info has been found in the cache
- */
- upload_info = _gf_false;
- }
- else {
- /*
- * info hasn't been found in the cache.
- */
- info = alloc_inode_info(local, local->loc);
- if (!info) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto exit;
- }
- init_inode_info_head(info, local->fd);
- upload_info = _gf_true;
- }
- /*
- * extract metadata
- */
- mtd = dict_get(dict, CRYPTO_FORMAT_PREFIX);
- if (!mtd) {
- local->op_ret = -1;
- local->op_errno = ENOENT;
- gf_log (this->name, GF_LOG_WARNING,
- "Format string wasn't found");
- goto exit;
- }
- /*
- * authenticate metadata against the path
- */
- ret = open_format((unsigned char *)mtd->data,
- mtd->len,
- local->loc,
- info,
- get_master_cinfo(priv),
- local,
- upload_info);
- if (ret) {
- local->op_ret = -1;
- local->op_errno = ret;
- goto exit;
- }
- if (upload_info) {
- ret = init_inode_info_tail(info, get_master_cinfo(priv));
- if (ret) {
- local->op_ret = -1;
- local->op_errno = ret;
- goto exit;
- }
- ret = inode_ctx_put(local->fd->inode,
- this, (uint64_t)(long)info);
- if (ret == -1) {
- local->op_ret = -1;
- local->op_errno = EIO;
- goto exit;
- }
- }
- if (local->custom_mtd) {
- /*
- * pass the metadata string to the customer
- */
- ret = dict_set_static_bin(local->xdata,
- CRYPTO_FORMAT_PREFIX,
- mtd->data,
- mtd->len);
- if (ret) {
- local->op_ret = -1;
- local->op_errno = ret;
- goto exit;
- }
- }
- exit:
- if (!local->custom_mtd)
- crypt_open_tail(frame, this);
- else
- put_one_call_open(frame);
- return 0;
-}
-
-static int32_t crypt_open_finodelk_cbk(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
-
- if (op_ret < 0) {
- gf_log(this->name, GF_LOG_WARNING, "finodelk (LOCK) failed");
- goto exit;
- }
- STACK_WIND(frame,
- load_mtd_open,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fgetxattr,
- local->fd,
- CRYPTO_FORMAT_PREFIX,
- NULL);
- return 0;
- exit:
- put_one_call_open(frame);
- return 0;
-}
-
-/*
- * verify metadata against the specified pathname
- */
-static int32_t crypt_open_cbk(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- fd_t *fd,
- dict_t *xdata)
-{
- struct gf_flock lock = {0, };
- crypt_local_t *local = frame->local;
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
-
- if (local->fd->inode->ia_type == IA_IFLNK)
- goto exit;
- if (op_ret < 0)
- goto exit;
- if (xdata)
- local->xdata = dict_ref(xdata);
- else if (local->custom_mtd){
- local->xdata = dict_new();
- if (!local->xdata) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- gf_log ("crypt", GF_LOG_ERROR,
- "Can not get new dict for mtd string");
- goto exit;
- }
- }
- lock.l_len = 0;
- lock.l_start = 0;
- lock.l_type = local->custom_mtd ? F_WRLCK : F_RDLCK;
- lock.l_whence = SEEK_SET;
-
- STACK_WIND(frame,
- crypt_open_finodelk_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->finodelk,
- this->name,
- fd,
- F_SETLKW,
- &lock,
- NULL);
- return 0;
- exit:
- put_one_call_open(frame);
- return 0;
-}
-
-static int32_t crypt_open(call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- int32_t flags,
- fd_t *fd,
- dict_t *xdata)
-{
- int32_t ret = ENOMEM;
- crypt_local_t *local;
-
- local = crypt_alloc_local(frame, this, GF_FOP_OPEN);
- if (!local)
- goto error;
- local->loc = GF_CALLOC(1, sizeof(*loc), gf_crypt_mt_loc);
- if (!local->loc) {
- ret = ENOMEM;
- goto error;
- }
- memset(local->loc, 0, sizeof(*local->loc));
- ret = loc_copy(local->loc, loc);
- if (ret) {
- GF_FREE(local->loc);
- goto error;
- }
- local->fd = fd_ref(fd);
-
- ret = is_custom_mtd(xdata);
- if (ret < 0) {
- loc_wipe(local->loc);
- GF_FREE(local->loc);
- ret = EINVAL;
- goto error;
- }
- local->custom_mtd = ret;
-
- if ((flags & O_ACCMODE) == O_WRONLY)
- /*
- * we can't open O_WRONLY, because
- * we need to do read-modify-write
- */
- flags = (flags & ~O_ACCMODE) | O_RDWR;
- /*
- * Make sure that out translated offsets
- * and counts won't be ignored
- */
- flags &= ~O_APPEND;
- get_one_call_nolock(frame);
- STACK_WIND(frame,
- crypt_open_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->open,
- loc,
- flags,
- fd,
- xdata);
- return 0;
- error:
- STACK_UNWIND_STRICT(open,
- frame,
- -1,
- ret,
- NULL,
- NULL);
- return 0;
-}
-
-static int32_t init_inode_info_tail(struct crypt_inode_info *info,
- struct master_cipher_info *master)
-{
- int32_t ret;
- struct object_cipher_info *object = &info->cinfo;
-
-#if DEBUG_CRYPT
- gf_log("crypt", GF_LOG_DEBUG, "Init inode info for object %s",
- uuid_utoa(info->oid));
-#endif
- ret = data_cipher_algs[object->o_alg][object->o_mode].set_private(info,
- master);
- if (ret) {
- gf_log("crypt", GF_LOG_ERROR, "Set private info failed");
- return ret;
- }
- return 0;
-}
-
-/*
- * Init inode info at ->create() time
- */
-static void init_inode_info_create(struct crypt_inode_info *info,
- struct master_cipher_info *master,
- data_t *data)
-{
- struct object_cipher_info *object;
-
- info->nr_minor = CRYPT_XLATOR_ID;
- memcpy(info->oid, data->data, data->len);
-
- object = &info->cinfo;
-
- object->o_alg = master->m_alg;
- object->o_mode = master->m_mode;
- object->o_block_bits = master->m_block_bits;
- object->o_dkey_size = master->m_dkey_size;
-}
-
-static void init_inode_info_head(struct crypt_inode_info *info, fd_t *fd)
-{
- memcpy(info->oid, fd->inode->gfid, sizeof(uuid_t));
-}
-
-static int32_t crypt_create_done(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
-{
- crypt_private_t *priv = this->private;
- crypt_local_t *local = frame->local;
- struct crypt_inode_info *info = local->info;
- fd_t *local_fd = local->fd;
- dict_t *local_xdata = local->xdata;
- inode_t *local_inode = local->inode;
-
- if (op_ret < 0) {
- free_inode_info(info);
- goto unwind;
- }
- op_errno = init_inode_info_tail(info, get_master_cinfo(priv));
- if (op_errno) {
- op_ret = -1;
- free_inode_info(info);
- goto unwind;
- }
- /*
- * FIXME: drop major subversion number
- */
- op_ret = inode_ctx_put(local->fd->inode, this, (uint64_t)(long)info);
- if (op_ret == -1) {
- op_errno = EIO;
- free_inode_info(info);
- goto unwind;
- }
- unwind:
- free_format(local);
- STACK_UNWIND_STRICT(create,
- frame,
- op_ret,
- op_errno,
- local_fd,
- local_inode,
- &local->buf,
- &local->prebuf,
- &local->postbuf,
- local_xdata);
- fd_unref(local_fd);
- inode_unref(local_inode);
- if (local_xdata)
- dict_unref(local_xdata);
- return 0;
-}
-
-static int crypt_create_tail(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *xdata)
-{
- struct gf_flock lock = {0, };
- crypt_local_t *local = frame->local;
- fd_t *local_fd = local->fd;
- dict_t *local_xdata = local->xdata;
- inode_t *local_inode = local->inode;
-
- dict_unref(local->xattr);
-
- if (op_ret < 0)
- goto error;
-
- lock.l_type = F_UNLCK;
- lock.l_whence = SEEK_SET;
- lock.l_start = 0;
- lock.l_len = 0;
- lock.l_pid = 0;
-
- STACK_WIND(frame,
- crypt_create_done,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->finodelk,
- this->name,
- local->fd,
- F_SETLKW,
- &lock,
- NULL);
- return 0;
- error:
- free_inode_info(local->info);
- free_format(local);
-
- STACK_UNWIND_STRICT(create,
- frame,
- op_ret,
- op_errno,
- local_fd,
- local_inode,
- &local->buf,
- &local->prebuf,
- &local->postbuf,
- local_xdata);
-
- fd_unref(local_fd);
- inode_unref(local_inode);
- if (local_xdata)
- dict_unref(local_xdata);
- return 0;
-}
-
-static int32_t crypt_create_finodelk_cbk(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
- struct crypt_inode_info *info = local->info;
-
- if (op_ret < 0)
- goto error;
-
- STACK_WIND(frame,
- crypt_create_tail,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetxattr,
- local->fd,
- local->xattr, /* CRYPTO_FORMAT_PREFIX */
- 0,
- NULL);
- return 0;
- error:
- free_inode_info(info);
- free_format(local);
- fd_unref(local->fd);
- dict_unref(local->xattr);
- if (local->xdata)
- dict_unref(local->xdata);
-
- STACK_UNWIND_STRICT(create,
- frame,
- op_ret,
- op_errno,
- NULL,
- NULL,
- NULL,
- NULL,
- NULL,
- NULL);
- return 0;
-}
-
-/*
- * Create and store crypt-specific format on disk;
- * Populate cache with private inode info
- */
-static int32_t crypt_create_cbk(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- fd_t *fd,
- inode_t *inode,
- struct iatt *buf,
- struct iatt *preparent,
- struct iatt *postparent,
- dict_t *xdata)
-{
- struct gf_flock lock = {0, };
- crypt_local_t *local = frame->local;
- struct crypt_inode_info *info = local->info;
-
- if (op_ret < 0)
- goto error;
- if (xdata)
- local->xdata = dict_ref(xdata);
- local->inode = inode_ref(inode);
- local->buf = *buf;
- local->prebuf = *preparent;
- local->postbuf = *postparent;
-
- lock.l_len = 0;
- lock.l_start = 0;
- lock.l_type = F_WRLCK;
- lock.l_whence = SEEK_SET;
-
- STACK_WIND(frame,
- crypt_create_finodelk_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->finodelk,
- this->name,
- local->fd,
- F_SETLKW,
- &lock,
- NULL);
- return 0;
- error:
- free_inode_info(info);
- free_format(local);
- fd_unref(local->fd);
- dict_unref(local->xattr);
-
- STACK_UNWIND_STRICT(create,
- frame,
- op_ret,
- op_errno,
- NULL, NULL, NULL,
- NULL, NULL, NULL);
- return 0;
-}
-
-static int32_t crypt_create(call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- int32_t flags,
- mode_t mode,
- mode_t umask,
- fd_t *fd,
- dict_t *xdata)
-{
- int ret;
- data_t *data;
- crypt_local_t *local;
- crypt_private_t *priv;
- struct master_cipher_info *master;
- struct crypt_inode_info *info;
-
- priv = this->private;
- master = get_master_cinfo(priv);
-
- if (master_alg_atomic(master)) {
- /*
- * We can't open O_WRONLY, because we
- * need to do read-modify-write.
- */
- if ((flags & O_ACCMODE) == O_WRONLY)
- flags = (flags & ~O_ACCMODE) | O_RDWR;
- /*
- * Make sure that out translated offsets
- * and counts won't be ignored
- */
- flags &= ~O_APPEND;
- }
- local = crypt_alloc_local(frame, this, GF_FOP_CREATE);
- if (!local) {
- ret = ENOMEM;
- goto error;
- }
- data = dict_get(xdata, "gfid-req");
- if (!data) {
- ret = EINVAL;
- gf_log("crypt", GF_LOG_WARNING, "gfid not found");
- goto error;
- }
- if (data->len != sizeof(uuid_t)) {
- ret = EINVAL;
- gf_log("crypt", GF_LOG_WARNING,
- "bad gfid size (%d), should be %d",
- (int)data->len, (int)sizeof(uuid_t));
- goto error;
- }
- info = alloc_inode_info(local, loc);
- if (!info){
- ret = ENOMEM;
- goto error;
- }
- /*
- * NOTE:
- * format has to be created BEFORE
- * proceeding to the untrusted server
- */
- ret = alloc_format_create(local);
- if (ret) {
- free_inode_info(info);
- goto error;
- }
- init_inode_info_create(info, master, data);
-
- ret = create_format(local->format,
- loc,
- info,
- master);
- if (ret) {
- free_inode_info(info);
- goto error;
- }
- local->xattr = dict_new();
- if (!local->xattr) {
- free_inode_info(info);
- free_format(local);
- goto error;
- }
- ret = dict_set_static_bin(local->xattr,
- CRYPTO_FORMAT_PREFIX,
- local->format,
- new_format_size());
- if (ret) {
- dict_unref(local->xattr);
- free_inode_info(info);
- free_format(local);
- goto error;
- }
- ret = dict_set(local->xattr, FSIZE_XATTR_PREFIX, data_from_uint64(0));
- if (ret) {
- dict_unref(local->xattr);
- free_inode_info(info);
- free_format(local);
- goto error;
- }
- local->fd = fd_ref(fd);
-
- STACK_WIND(frame,
- crypt_create_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->create,
- loc,
- flags,
- mode,
- umask,
- fd,
- xdata);
- return 0;
- error:
- gf_log("crypt", GF_LOG_WARNING, "can not create file");
- STACK_UNWIND_STRICT(create,
- frame,
- -1,
- ret,
- NULL, NULL, NULL,
- NULL, NULL, NULL);
- return 0;
-}
-
-/*
- * FIXME: this should depends on the version of format string
- */
-static int32_t filter_crypt_xattr(dict_t *dict,
- char *key, data_t *value, void *data)
-{
- dict_del(dict, key);
- return 0;
-}
-
-static int32_t crypt_fsetxattr(call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- dict_t *dict,
- int32_t flags, dict_t *xdata)
-{
- dict_foreach_fnmatch(dict, "trusted.glusterfs.crypt*",
- filter_crypt_xattr, NULL);
- STACK_WIND(frame,
- default_fsetxattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetxattr,
- fd,
- dict,
- flags,
- xdata);
- return 0;
-}
-
-/*
- * TBD: verify file metadata before wind
- */
-static int32_t crypt_setxattr(call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- dict_t *dict,
- int32_t flags, dict_t *xdata)
-{
- dict_foreach_fnmatch(dict, "trusted.glusterfs.crypt*",
- filter_crypt_xattr, NULL);
- STACK_WIND(frame,
- default_setxattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setxattr,
- loc,
- dict,
- flags,
- xdata);
- return 0;
-}
-
-/*
- * called as flush_cbk()
- */
-static int32_t linkop_end(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
- linkop_unwind_handler_t unwind_fn;
- unwind_fn = linkop_unwind_dispatch(local->fop);
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
-
- if (op_ret < 0 &&
- op_errno == ENOENT &&
- local->loc->inode->ia_type == IA_IFLNK) {
- local->op_ret = 0;
- local->op_errno = 0;
- }
- unwind_fn(frame);
- return 0;
-}
-
-/*
- * unpin inode on the server
- */
-static int32_t link_flush(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- inode_t *inode,
- struct iatt *buf,
- struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
-
- if (op_ret < 0)
- goto error;
- if (local->xdata) {
- dict_unref(local->xdata);
- local->xdata = NULL;
- }
- if (xdata)
- local->xdata = dict_ref(xdata);
- local->inode = inode_ref(inode);
- local->buf = *buf;
- local->prebuf = *preparent;
- local->postbuf = *postparent;
-
- STACK_WIND(frame,
- linkop_end,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->flush,
- local->fd,
- NULL);
- return 0;
- error:
- local->op_ret = -1;
- local->op_errno = op_errno;
- link_unwind(frame);
- return 0;
-}
-
-void link_unwind(call_frame_t *frame)
-{
- crypt_local_t *local = frame->local;
- dict_t *xdata;
- dict_t *xattr;
- inode_t *inode;
-
- if (!local) {
- STACK_UNWIND_STRICT(link,
- frame,
- -1,
- ENOMEM,
- NULL,
- NULL,
- NULL,
- NULL,
- NULL);
- return;
- }
- xdata = local->xdata;
- xattr = local->xattr;
- inode = local->inode;
-
- if (local->loc){
- loc_wipe(local->loc);
- GF_FREE(local->loc);
- }
- if (local->newloc) {
- loc_wipe(local->newloc);
- GF_FREE(local->newloc);
- }
- if (local->fd)
- fd_unref(local->fd);
- if (local->format)
- GF_FREE(local->format);
-
- STACK_UNWIND_STRICT(link,
- frame,
- local->op_ret,
- local->op_errno,
- inode,
- &local->buf,
- &local->prebuf,
- &local->postbuf,
- xdata);
- if (xdata)
- dict_unref(xdata);
- if (xattr)
- dict_unref(xattr);
- if (inode)
- inode_unref(inode);
-}
-
-void link_wind(call_frame_t *frame, xlator_t *this)
-{
- crypt_local_t *local = frame->local;
-
- STACK_WIND(frame,
- link_flush,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->link,
- local->loc,
- local->newloc,
- local->xdata);
-}
-
-/*
- * unlink()
- */
-static int32_t unlink_flush(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
-
- if (op_ret < 0)
- goto error;
- local->prebuf = *preparent;
- local->postbuf = *postparent;
- if (local->xdata) {
- dict_unref(local->xdata);
- local->xdata = NULL;
- }
- if (xdata)
- local->xdata = dict_ref(xdata);
-
- STACK_WIND(frame,
- linkop_end,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->flush,
- local->fd,
- NULL);
- return 0;
- error:
- local->op_ret = -1;
- local->op_errno = op_errno;
- unlink_unwind(frame);
- return 0;
-}
-
-void unlink_unwind(call_frame_t *frame)
-{
- crypt_local_t *local = frame->local;
- dict_t *xdata;
- dict_t *xattr;
-
- if (!local) {
- STACK_UNWIND_STRICT(unlink,
- frame,
- -1,
- ENOMEM,
- NULL,
- NULL,
- NULL);
- return;
- }
- xdata = local->xdata;
- xattr = local->xattr;
- if (local->loc){
- loc_wipe(local->loc);
- GF_FREE(local->loc);
- }
- if (local->fd)
- fd_unref(local->fd);
- if (local->format)
- GF_FREE(local->format);
-
- STACK_UNWIND_STRICT(unlink,
- frame,
- local->op_ret,
- local->op_errno,
- &local->prebuf,
- &local->postbuf,
- xdata);
- if (xdata)
- dict_unref(xdata);
- if (xattr)
- dict_unref(xattr);
-}
-
-void unlink_wind(call_frame_t *frame, xlator_t *this)
-{
- crypt_local_t *local = frame->local;
-
- STACK_WIND(frame,
- unlink_flush,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->unlink,
- local->loc,
- local->flags,
- local->xdata);
-}
-
-void rename_unwind(call_frame_t *frame)
-{
- crypt_local_t *local = frame->local;
- dict_t *xdata;
- dict_t *xattr;
- struct iatt *prenewparent;
- struct iatt *postnewparent;
-
- if (!local) {
- STACK_UNWIND_STRICT(rename,
- frame,
- -1,
- ENOMEM,
- NULL,
- NULL,
- NULL,
- NULL,
- NULL,
- NULL);
- return;
- }
- xdata = local->xdata;
- xattr = local->xattr;
- prenewparent = local->prenewparent;
- postnewparent = local->postnewparent;
-
- if (local->loc){
- loc_wipe(local->loc);
- GF_FREE(local->loc);
- }
- if (local->newloc){
- loc_wipe(local->newloc);
- GF_FREE(local->newloc);
- }
- if (local->fd)
- fd_unref(local->fd);
- if (local->format)
- GF_FREE(local->format);
-
- STACK_UNWIND_STRICT(rename,
- frame,
- local->op_ret,
- local->op_errno,
- &local->buf,
- &local->prebuf,
- &local->postbuf,
- prenewparent,
- postnewparent,
- xdata);
- if (xdata)
- dict_unref(xdata);
- if (xattr)
- dict_unref(xattr);
- if (prenewparent)
- GF_FREE(prenewparent);
- if (postnewparent)
- GF_FREE(postnewparent);
-}
-
-/*
- * called as flush_cbk()
- */
-static int32_t rename_end(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
-
- rename_unwind(frame);
- return 0;
-}
-
-static int32_t rename_flush(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iatt *buf,
- struct iatt *preoldparent,
- struct iatt *postoldparent,
- struct iatt *prenewparent,
- struct iatt *postnewparent,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
-
- if (op_ret < 0)
- goto error;
- dict_unref(local->xdata);
- local->xdata = NULL;
- if (xdata)
- local->xdata = dict_ref(xdata);
-
- local->buf = *buf;
- local->prebuf = *preoldparent;
- local->postbuf = *postoldparent;
- if (prenewparent) {
- local->prenewparent = GF_CALLOC(1, sizeof(*prenewparent),
- gf_crypt_mt_iatt);
- if (!local->prenewparent) {
- op_errno = ENOMEM;
- goto error;
- }
- *local->prenewparent = *prenewparent;
- }
- if (postnewparent) {
- local->postnewparent = GF_CALLOC(1, sizeof(*postnewparent),
- gf_crypt_mt_iatt);
- if (!local->postnewparent) {
- op_errno = ENOMEM;
- goto error;
- }
- *local->postnewparent = *postnewparent;
- }
- STACK_WIND(frame,
- rename_end,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->flush,
- local->fd,
- NULL);
- return 0;
- error:
- local->op_ret = -1;
- local->op_errno = op_errno;
- rename_unwind(frame);
- return 0;
-}
-
-void rename_wind(call_frame_t *frame, xlator_t *this)
-{
- crypt_local_t *local = frame->local;
-
- STACK_WIND(frame,
- rename_flush,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rename,
- local->loc,
- local->newloc,
- local->xdata);
-}
-
-static int32_t __do_linkop(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
- linkop_wind_handler_t wind_fn;
- linkop_unwind_handler_t unwind_fn;
-
- wind_fn = linkop_wind_dispatch(local->fop);
- unwind_fn = linkop_unwind_dispatch(local->fop);
-
- local->op_ret = op_ret;
- local->op_errno = op_errno;
-
- if (op_ret >= 0)
- wind_fn(frame, this);
- else {
- gf_log(this->name, GF_LOG_WARNING, "mtd unlock failed (%d)",
- op_errno);
- unwind_fn(frame);
- }
- return 0;
-}
-
-static int32_t do_linkop(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *xdata)
-{
- struct gf_flock lock = {0, };
- crypt_local_t *local = frame->local;
- linkop_unwind_handler_t unwind_fn;
-
- unwind_fn = linkop_unwind_dispatch(local->fop);
- local->op_ret = op_ret;
- local->op_errno = op_errno;
-
- if(op_ret < 0)
- goto error;
-
- lock.l_type = F_UNLCK;
- lock.l_whence = SEEK_SET;
- lock.l_start = 0;
- lock.l_len = 0;
- lock.l_pid = 0;
-
- STACK_WIND(frame,
- __do_linkop,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->finodelk,
- this->name,
- local->fd,
- F_SETLKW,
- &lock,
- NULL);
- return 0;
- error:
- unwind_fn(frame);
- return 0;
-}
-
-/*
- * Update the metadata string (against the new pathname);
- * submit the result
- */
-static int32_t linkop_begin(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- fd_t *fd,
- dict_t *xdata)
-{
- gf_boolean_t upload_info;
- crypt_local_t *local = frame->local;
- crypt_private_t *priv = this->private;
- struct crypt_inode_info *info;
- data_t *old_mtd;
- uint32_t new_mtd_size;
- uint64_t value = 0;
- void (*unwind_fn)(call_frame_t *frame);
- mtd_op_t mop;
-
- unwind_fn = linkop_unwind_dispatch(local->fop);
- mop = linkop_mtdop_dispatch(local->fop);
-
- if (op_ret < 0) {
- /*
- * verification failed
- */
- goto error;
- } else {
- fd_bind (fd);
- }
-
- old_mtd = dict_get(xdata, CRYPTO_FORMAT_PREFIX);
- if (!old_mtd) {
- op_errno = EIO;
- gf_log (this->name, GF_LOG_DEBUG,
- "Metadata string wasn't found");
- goto error;
- }
- new_mtd_size = format_size(mop, old_mtd->len);
- op_errno = alloc_format(local, new_mtd_size);
- if (op_errno)
- goto error;
- /*
- * check for cached info
- */
- op_ret = inode_ctx_get(fd->inode, this, &value);
- if (op_ret != -1) {
- info = (struct crypt_inode_info *)(long)value;
- if (info == NULL) {
- gf_log (this->name, GF_LOG_WARNING,
- "Inode info was not found");
- op_errno = EINVAL;
- goto error;
- }
- /*
- * info was found in the cache
- */
- local->info = info;
- upload_info = _gf_false;
- }
- else {
- /*
- * info wasn't found in the cache;
- */
- info = alloc_inode_info(local, local->loc);
- if (!info)
- goto error;
- init_inode_info_head(info, fd);
- local->info = info;
- upload_info = _gf_true;
- }
- op_errno = open_format((unsigned char *)old_mtd->data,
- old_mtd->len,
- local->loc,
- info,
- get_master_cinfo(priv),
- local,
- upload_info);
- if (op_errno)
- goto error;
- if (upload_info == _gf_true) {
- op_errno = init_inode_info_tail(info,
- get_master_cinfo(priv));
- if (op_errno)
- goto error;
- op_errno = inode_ctx_put(fd->inode, this,
- (uint64_t)(long)(info));
- if (op_errno == -1) {
- op_errno = EIO;
- goto error;
- }
- }
- /*
- * update the format string (append/update/cup a MAC)
- */
- op_errno = update_format(local->format,
- (unsigned char *)old_mtd->data,
- old_mtd->len,
- local->mac_idx,
- mop,
- local->newloc,
- info,
- get_master_cinfo(priv),
- local);
- if (op_errno)
- goto error;
- /*
- * store the new format string on the server
- */
- if (new_mtd_size) {
- op_errno = dict_set_static_bin(local->xattr,
- CRYPTO_FORMAT_PREFIX,
- local->format,
- new_mtd_size);
- if (op_errno)
- goto error;
- }
- STACK_WIND(frame,
- do_linkop,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setxattr,
- local->loc,
- local->xattr,
- 0,
- NULL);
- return 0;
- error:
- local->op_ret = -1;
- local->op_errno = op_errno;
- unwind_fn(frame);
- return 0;
-}
-
-static int32_t linkop_grab_local(call_frame_t *frame,
- xlator_t *this,
- loc_t *oldloc,
- loc_t *newloc,
- int flags, dict_t *xdata,
- glusterfs_fop_t op)
-{
- int32_t ret = ENOMEM;
- fd_t *fd;
- crypt_local_t *local;
-
- local = crypt_alloc_local(frame, this, op);
- if (!local)
- goto error;
- if (xdata)
- local->xdata = dict_ref(xdata);
-
- fd = fd_create(oldloc->inode, frame->root->pid);
- if (!fd) {
- gf_log(this->name, GF_LOG_ERROR, "Can not create fd");
- goto error;
- }
- local->fd = fd;
- local->flags = flags;
- local->loc = GF_CALLOC(1, sizeof(*oldloc), gf_crypt_mt_loc);
- if (!local->loc)
- goto error;
- memset(local->loc, 0, sizeof(*local->loc));
- ret = loc_copy(local->loc, oldloc);
- if (ret) {
- GF_FREE(local->loc);
- local->loc = NULL;
- goto error;
- }
- if (newloc) {
- local->newloc = GF_CALLOC(1, sizeof(*newloc), gf_crypt_mt_loc);
- if (!local->newloc) {
- loc_wipe(local->loc);
- GF_FREE(local->loc);
- goto error;
- }
- memset(local->newloc, 0, sizeof(*local->newloc));
- ret = loc_copy(local->newloc, newloc);
- if (ret) {
- loc_wipe(local->loc);
- GF_FREE(local->loc);
- GF_FREE(local->newloc);
- goto error;
- }
- }
- local->xattr = dict_new();
- if (!local->xattr) {
- gf_log(this->name, GF_LOG_ERROR, "Can not create dict");
- ret = ENOMEM;
- goto error;
- }
- return 0;
-
-error:
- if (local) {
- if (local->xdata)
- dict_unref(local->xdata);
- if (local->fd)
- fd_unref(local->fd);
- local->fd = 0;
- local->loc = NULL;
- local->newloc = NULL;
- local->op_ret = -1;
- local->op_errno = ret;
- }
-
- return ret;
-}
-
-/*
- * read and verify locked metadata against the old pathname (via open);
- * update the metadata string in accordance with the new pathname;
- * submit modified metadata;
- * wind;
- */
-static int32_t linkop(call_frame_t *frame,
- xlator_t *this,
- loc_t *oldloc,
- loc_t *newloc,
- int flags,
- dict_t *xdata,
- glusterfs_fop_t op)
-{
- int32_t ret;
- dict_t *dict;
- crypt_local_t *local;
- void (*unwind_fn)(call_frame_t *frame);
- void (*wind_fn)(call_frame_t *frame, xlator_t *this);
-
- wind_fn = linkop_wind_dispatch(op);
- unwind_fn = linkop_unwind_dispatch(op);
-
- ret = linkop_grab_local(frame, this, oldloc, newloc, flags, xdata, op);
- local = frame->local;
- if (ret)
- goto error;
-
- if (local->fd->inode->ia_type == IA_IFLNK)
- goto wind;
-
- dict = dict_new();
- if (!dict) {
- gf_log(this->name, GF_LOG_ERROR, "Can not create dict");
- ret = ENOMEM;
- goto error;
- }
- /*
- * Set a message to crypt_open() that we need
- * locked metadata string.
- * All link operations (link, unlink, rename)
- * need write lock
- */
- msgflags_set_mtd_wlock(&local->msgflags);
- ret = dict_set_static_bin(dict,
- MSGFLAGS_PREFIX,
- &local->msgflags,
- sizeof(local->msgflags));
- if (ret) {
- gf_log(this->name, GF_LOG_ERROR, "Can not set dict");
- dict_unref(dict);
- goto error;
- }
- /*
- * verify metadata against the old pathname
- * and retrieve locked metadata string
- */
- STACK_WIND(frame,
- linkop_begin,
- this,
- this->fops->open, /* crypt_open() */
- oldloc,
- O_RDWR,
- local->fd,
- dict);
- dict_unref(dict);
- return 0;
-
-wind:
- wind_fn(frame, this);
- return 0;
-
-error:
- local->op_ret = -1;
- local->op_errno = ret;
- unwind_fn(frame);
- return 0;
-}
-
-static int32_t crypt_link(call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
-{
- return linkop(frame, this, oldloc, newloc, 0, xdata, GF_FOP_LINK);
-}
-
-static int32_t crypt_unlink(call_frame_t *frame, xlator_t *this,
- loc_t *loc, int flags, dict_t *xdata)
-{
- return linkop(frame, this, loc, NULL, flags, xdata, GF_FOP_UNLINK);
-}
-
-static int32_t crypt_rename(call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
-{
- return linkop(frame, this, oldloc, newloc, 0, xdata, GF_FOP_RENAME);
-}
-
-static void put_one_call_open(call_frame_t *frame)
-{
- crypt_local_t *local = frame->local;
- if (put_one_call(local)) {
- fd_t *fd = local->fd;
- loc_t *loc = local->loc;
- dict_t *xdata = local->xdata;
-
- STACK_UNWIND_STRICT(open,
- frame,
- local->op_ret,
- local->op_errno,
- fd,
- xdata);
- fd_unref(fd);
- if (xdata)
- dict_unref(xdata);
- loc_wipe(loc);
- GF_FREE(loc);
- }
-}
-
-static int32_t __crypt_readv_done(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
- fd_t *local_fd = local->fd;
- dict_t *local_xdata = local->xdata;
- /* read deals with data configs only */
- struct iovec *avec = local->data_conf.avec;
- char **pool = local->data_conf.pool;
- int blocks_in_pool = local->data_conf.blocks_in_pool;
- struct iobref *iobref = local->iobref;
- struct iobref *iobref_data = local->iobref_data;
-
- if (op_ret < 0) {
- gf_log(this->name, GF_LOG_WARNING,
- "readv unlock failed (%d)", op_errno);
- if (local->op_ret >= 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- }
- }
- dump_plain_text(local, avec);
-
- gf_log("crypt", GF_LOG_DEBUG,
- "readv: ret_to_user: %d, iovec len: %d, ia_size: %llu",
- (int)(local->rw_count > 0 ? local->rw_count : local->op_ret),
- (int)(local->rw_count > 0 ? iov_length(avec, local->data_conf.acount) : 0),
- (unsigned long long)local->buf.ia_size);
-
- STACK_UNWIND_STRICT(readv,
- frame,
- local->rw_count > 0 ? local->rw_count : local->op_ret,
- local->op_errno,
- avec,
- avec ? local->data_conf.acount : 0,
- &local->buf,
- local->iobref,
- local_xdata);
-
- free_avec(avec, pool, blocks_in_pool);
- fd_unref(local_fd);
- if (local_xdata)
- dict_unref(local_xdata);
- if (iobref)
- iobref_unref(iobref);
- if (iobref_data)
- iobref_unref(iobref_data);
- return 0;
-}
-
-static void crypt_readv_done(call_frame_t *frame, xlator_t *this)
-{
- if (parent_is_crypt_xlator(frame, this))
- /*
- * don't unlock (it will be done by the parent)
- */
- __crypt_readv_done(frame, NULL, this, 0, 0, NULL);
- else {
- crypt_local_t *local = frame->local;
- struct gf_flock lock = {0, };
-
- lock.l_type = F_UNLCK;
- lock.l_whence = SEEK_SET;
- lock.l_start = 0;
- lock.l_len = 0;
- lock.l_pid = 0;
-
- STACK_WIND(frame,
- __crypt_readv_done,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->finodelk,
- this->name,
- local->fd,
- F_SETLKW,
- &lock,
- NULL);
- }
-}
-
-static void put_one_call_readv(call_frame_t *frame, xlator_t *this)
-{
- crypt_local_t *local = frame->local;
- if (put_one_call(local))
- crypt_readv_done(frame, this);
-}
-
-static int32_t __crypt_writev_done(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
- fd_t *local_fd = local->fd;
- dict_t *local_xdata = local->xdata;
- int32_t ret_to_user;
-
- if (local->xattr)
- dict_unref(local->xattr);
- /*
- * Calculate amout of butes to be returned
- * to user. We need to subtract paddings that
- * have been written as a part of atom.
- */
- /*
- * subtract head padding
- */
- if (local->rw_count == 0)
- /*
- * Nothing has been written, it must be an error
- */
- ret_to_user = local->op_ret;
- else if (local->rw_count <= local->data_conf.off_in_head) {
- gf_log("crypt", GF_LOG_WARNING, "Incomplete write");
- ret_to_user = 0;
- }
- else
- ret_to_user = local->rw_count -
- local->data_conf.off_in_head;
- /*
- * subtract tail padding
- */
- if (ret_to_user > local->data_conf.orig_size)
- ret_to_user = local->data_conf.orig_size;
-
- if (local->iobref)
- iobref_unref(local->iobref);
- if (local->iobref_data)
- iobref_unref(local->iobref_data);
- free_avec_data(local);
- free_avec_hole(local);
-
- gf_log("crypt", GF_LOG_DEBUG,
- "writev: ret_to_user: %d", ret_to_user);
-
- STACK_UNWIND_STRICT(writev,
- frame,
- ret_to_user,
- local->op_errno,
- &local->prebuf,
- &local->postbuf,
- local_xdata);
- fd_unref(local_fd);
- if (local_xdata)
- dict_unref(local_xdata);
- return 0;
-}
-
-static int32_t crypt_writev_done(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
-
- if (op_ret < 0)
- gf_log("crypt", GF_LOG_WARNING, "can not update file size");
-
- if (parent_is_crypt_xlator(frame, this))
- /*
- * don't unlock (it will be done by the parent)
- */
- __crypt_writev_done(frame, NULL, this, 0, 0, NULL);
- else {
- struct gf_flock lock = {0, };
-
- lock.l_type = F_UNLCK;
- lock.l_whence = SEEK_SET;
- lock.l_start = 0;
- lock.l_len = 0;
- lock.l_pid = 0;
-
- STACK_WIND(frame,
- __crypt_writev_done,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->finodelk,
- this->name,
- local->fd,
- F_SETLKW,
- &lock,
- NULL);
- }
- return 0;
-}
-
-static void put_one_call_writev(call_frame_t *frame, xlator_t *this)
-{
- crypt_local_t *local = frame->local;
- if (put_one_call(local)) {
- if (local->update_disk_file_size) {
- int32_t ret;
- /*
- * update file size, unlock the file and unwind
- */
- ret = dict_set(local->xattr,
- FSIZE_XATTR_PREFIX,
- data_from_uint64(local->cur_file_size));
- if (ret) {
- gf_log("crypt", GF_LOG_WARNING,
- "can not set key to update file size");
- crypt_writev_done(frame, NULL,
- this, 0, 0, NULL);
- return;
- }
- gf_log("crypt", GF_LOG_DEBUG,
- "Updating disk file size to %llu",
- (unsigned long long)local->cur_file_size);
- STACK_WIND(frame,
- crypt_writev_done,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetxattr,
- local->fd,
- local->xattr, /* CRYPTO_FORMAT_PREFIX */
- 0,
- NULL);
- }
- else
- crypt_writev_done(frame, NULL, this, 0, 0, NULL);
- }
-}
-
-static int32_t __crypt_ftruncate_done(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
- fd_t *local_fd = local->fd;
- dict_t *local_xdata = local->xdata;
- char *iobase = local->vec.iov_base;
-
- if (op_ret < 0) {
- gf_log(this->name, GF_LOG_WARNING,
- "ftruncate unlock failed (%d)", op_errno);
- if (local->op_ret >= 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- }
- }
- if (local->iobref_data)
- iobref_unref(local->iobref_data);
- free_avec_data(local);
- free_avec_hole(local);
-
- gf_log("crypt", GF_LOG_DEBUG,
- "ftruncate, return to user: presize=%llu, postsize=%llu",
- (unsigned long long)local->prebuf.ia_size,
- (unsigned long long)local->postbuf.ia_size);
-
- STACK_UNWIND_STRICT(ftruncate,
- frame,
- ((local->op_ret < 0) ? -1 : 0),
- local->op_errno,
- &local->prebuf,
- &local->postbuf,
- local_xdata);
- fd_unref(local_fd);
- if (local_xdata)
- dict_unref(local_xdata);
- if (iobase)
- GF_FREE(iobase);
- return 0;
-}
-
-static int32_t crypt_ftruncate_done(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
- struct gf_flock lock = {0, };
-
- dict_unref(local->xattr);
- if (op_ret < 0)
- gf_log("crypt", GF_LOG_WARNING, "can not update file size");
-
- lock.l_type = F_UNLCK;
- lock.l_whence = SEEK_SET;
- lock.l_start = 0;
- lock.l_len = 0;
- lock.l_pid = 0;
-
- STACK_WIND(frame,
- __crypt_ftruncate_done,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->finodelk,
- this->name,
- local->fd,
- F_SETLKW,
- &lock,
- NULL);
- return 0;
-}
-
-static void put_one_call_ftruncate(call_frame_t *frame, xlator_t *this)
-{
- crypt_local_t *local = frame->local;
- if (put_one_call(local)) {
- if (local->update_disk_file_size) {
- int32_t ret;
- /*
- * update file size, unlock the file and unwind
- */
- ret = dict_set(local->xattr,
- FSIZE_XATTR_PREFIX,
- data_from_uint64(local->cur_file_size));
- if (ret) {
- gf_log("crypt", GF_LOG_WARNING,
- "can not set key to update file size");
- crypt_ftruncate_done(frame, NULL,
- this, 0, 0, NULL);
- return;
- }
- gf_log("crypt", GF_LOG_DEBUG,
- "Updating disk file size to %llu",
- (unsigned long long)local->cur_file_size);
- STACK_WIND(frame,
- crypt_ftruncate_done,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetxattr,
- local->fd,
- local->xattr, /* CRYPTO_FORMAT_PREFIX */
- 0,
- NULL);
- }
- else
- crypt_ftruncate_done(frame, NULL, this, 0, 0, NULL);
- }
-}
-
-/*
- * load regular file size for some FOPs
- */
-static int32_t load_file_size(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- dict_t *dict,
- dict_t *xdata)
-{
- data_t *data;
- crypt_local_t *local = frame->local;
-
- dict_t *local_xdata = local->xdata;
- inode_t *local_inode = local->inode;
-
- if (op_ret < 0)
- goto unwind;
- /*
- * load regular file size
- */
- data = dict_get(dict, FSIZE_XATTR_PREFIX);
- if (!data) {
- if (local->xdata)
- dict_unref(local->xdata);
- gf_log("crypt", GF_LOG_WARNING, "Regular file size not found");
- op_ret = -1;
- op_errno = EIO;
- goto unwind;
- }
- local->buf.ia_size = data_to_uint64(data);
-
- gf_log(this->name, GF_LOG_DEBUG,
- "FOP %d: Translate regular file to %llu",
- local->fop,
- (unsigned long long)local->buf.ia_size);
- unwind:
- if (local->fd)
- fd_unref(local->fd);
- if (local->loc) {
- loc_wipe(local->loc);
- GF_FREE(local->loc);
- }
- switch (local->fop) {
- case GF_FOP_FSTAT:
- STACK_UNWIND_STRICT(fstat,
- frame,
- op_ret,
- op_errno,
- op_ret >= 0 ? &local->buf : NULL,
- local->xdata);
- break;
- case GF_FOP_STAT:
- STACK_UNWIND_STRICT(stat,
- frame,
- op_ret,
- op_errno,
- op_ret >= 0 ? &local->buf : NULL,
- local->xdata);
- break;
- case GF_FOP_LOOKUP:
- STACK_UNWIND_STRICT(lookup,
- frame,
- op_ret,
- op_errno,
- op_ret >= 0 ? local->inode : NULL,
- op_ret >= 0 ? &local->buf : NULL,
- local->xdata,
- op_ret >= 0 ? &local->postbuf : NULL);
- break;
- case GF_FOP_READ:
- STACK_UNWIND_STRICT(readv,
- frame,
- op_ret,
- op_errno,
- NULL,
- 0,
- op_ret >= 0 ? &local->buf : NULL,
- NULL,
- NULL);
- break;
- default:
- gf_log(this->name, GF_LOG_WARNING,
- "Improper file operation %d", local->fop);
- }
- if (local_xdata)
- dict_unref(local_xdata);
- if (local_inode)
- inode_unref(local_inode);
- return 0;
-}
-
-static int32_t crypt_stat_common_cbk(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iatt *buf, dict_t *xdata)
-{
- crypt_local_t *local = frame->local;
-
- if (op_ret < 0)
- goto unwind;
- if (!IA_ISREG(buf->ia_type))
- goto unwind;
-
- local->buf = *buf;
- if (xdata)
- local->xdata = dict_ref(xdata);
-
- switch (local->fop) {
- case GF_FOP_FSTAT:
- STACK_WIND(frame,
- load_file_size,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fgetxattr,
- local->fd,
- FSIZE_XATTR_PREFIX,
- NULL);
- break;
- case GF_FOP_STAT:
- STACK_WIND(frame,
- load_file_size,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr,
- local->loc,
- FSIZE_XATTR_PREFIX,
- NULL);
- break;
- default:
- gf_log (this->name, GF_LOG_WARNING,
- "Improper file operation %d", local->fop);
- }
- return 0;
- unwind:
- if (local->fd)
- fd_unref(local->fd);
- if (local->loc) {
- loc_wipe(local->loc);
- GF_FREE(local->loc);
- }
- switch (local->fop) {
- case GF_FOP_FSTAT:
- STACK_UNWIND_STRICT(fstat,
- frame,
- op_ret,
- op_errno,
- op_ret >= 0 ? buf : NULL,
- op_ret >= 0 ? xdata : NULL);
- break;
- case GF_FOP_STAT:
- STACK_UNWIND_STRICT(stat,
- frame,
- op_ret,
- op_errno,
- op_ret >= 0 ? buf : NULL,
- op_ret >= 0 ? xdata : NULL);
- break;
- default:
- gf_log (this->name, GF_LOG_WARNING,
- "Improper file operation %d", local->fop);
- }
- return 0;
-}
-
-static int32_t crypt_fstat(call_frame_t *frame,
- xlator_t *this,
- fd_t *fd, dict_t *xdata)
-{
- crypt_local_t *local;
-
- local = crypt_alloc_local(frame, this, GF_FOP_FSTAT);
- if (!local)
- goto error;
- local->fd = fd_ref(fd);
- STACK_WIND(frame,
- crypt_stat_common_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fstat,
- fd,
- xdata);
- return 0;
- error:
- STACK_UNWIND_STRICT(fstat,
- frame,
- -1,
- ENOMEM,
- NULL,
- NULL);
- return 0;
-}
-
-static int32_t crypt_stat(call_frame_t *frame,
- xlator_t *this,
- loc_t *loc, dict_t *xdata)
-{
- int32_t ret;
- crypt_local_t *local;
-
- local = crypt_alloc_local(frame, this, GF_FOP_STAT);
- if (!local)
- goto error;
- local->loc = GF_CALLOC(1, sizeof(*loc), gf_crypt_mt_loc);
- if (!local->loc)
- goto error;
- memset(local->loc, 0, sizeof(*local->loc));
- ret = loc_copy(local->loc, loc);
- if (ret) {
- GF_FREE(local->loc);
- goto error;
- }
- STACK_WIND(frame,
- crypt_stat_common_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->stat,
- loc,
- xdata);
- return 0;
- error:
- STACK_UNWIND_STRICT(stat,
- frame,
- -1,
- ENOMEM,
- NULL,
- NULL);
- return 0;
-}
-
-static int32_t crypt_lookup_cbk(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- inode_t *inode,
- struct iatt *buf, dict_t *xdata,
- struct iatt *postparent)
-{
- crypt_local_t *local = frame->local;
-
- if (op_ret < 0)
- goto unwind;
- if (!IA_ISREG(buf->ia_type))
- goto unwind;
-
- local->inode = inode_ref(inode);
- local->buf = *buf;
- local->postbuf = *postparent;
- if (xdata)
- local->xdata = dict_ref(xdata);
- gf_uuid_copy(local->loc->gfid, buf->ia_gfid);
-
- STACK_WIND(frame,
- load_file_size,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr,
- local->loc,
- FSIZE_XATTR_PREFIX,
- NULL);
- return 0;
- unwind:
- loc_wipe(local->loc);
- GF_FREE(local->loc);
- STACK_UNWIND_STRICT(lookup,
- frame,
- op_ret,
- op_errno,
- inode,
- buf,
- xdata,
- postparent);
- return 0;
-}
-
-static int32_t crypt_lookup(call_frame_t *frame,
- xlator_t *this,
- loc_t *loc, dict_t *xdata)
-{
- int32_t ret;
- crypt_local_t *local;
-
- local = crypt_alloc_local(frame, this, GF_FOP_LOOKUP);
- if (!local)
- goto error;
- local->loc = GF_CALLOC(1, sizeof(*loc), gf_crypt_mt_loc);
- if (!local->loc)
- goto error;
- memset(local->loc, 0, sizeof(*local->loc));
- ret = loc_copy(local->loc, loc);
- if (ret) {
- GF_FREE(local->loc);
- goto error;
- }
- gf_log(this->name, GF_LOG_DEBUG, "Lookup %s", loc->path);
- STACK_WIND(frame,
- crypt_lookup_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup,
- loc,
- xdata);
- return 0;
- error:
- STACK_UNWIND_STRICT(lookup,
- frame,
- -1,
- ENOMEM,
- NULL,
- NULL,
- NULL,
- NULL);
- return 0;
-}
-
-/*
- * for every regular directory entry find its real file size
- * and update stat's buf properly
- */
-static int32_t crypt_readdirp_cbk(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- gf_dirent_t *entries, dict_t *xdata)
-{
- gf_dirent_t *entry = NULL;
-
- if (op_ret < 0)
- goto unwind;
-
- list_for_each_entry (entry, (&entries->list), list) {
- data_t *data;
-
- if (!IA_ISREG(entry->d_stat.ia_type))
- continue;
- data = dict_get(entry->dict, FSIZE_XATTR_PREFIX);
- if (!data){
- gf_log("crypt", GF_LOG_WARNING,
- "Regular file size of direntry not found");
- op_errno = EIO;
- op_ret = -1;
- break;
- }
- entry->d_stat.ia_size = data_to_uint64(data);
- }
- unwind:
- STACK_UNWIND_STRICT(readdirp, frame, op_ret, op_errno, entries, xdata);
- return 0;
-}
-
-/*
- * ->readdirp() fills in-core inodes, so we need to set proper
- * file sizes for all directory entries of the parent @fd.
- * Actual updates take place in ->crypt_readdirp_cbk()
- */
-static int32_t crypt_readdirp(call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset,
- dict_t *xdata)
-{
- int32_t ret = ENOMEM;
-
- if (!xdata) {
- xdata = dict_new();
- if (!xdata)
- goto error;
- }
- else
- dict_ref(xdata);
- /*
- * make sure that we'll have real file sizes at ->readdirp_cbk()
- */
- ret = dict_set(xdata, FSIZE_XATTR_PREFIX, data_from_uint64(0));
- if (ret) {
- dict_unref(xdata);
- goto error;
- }
- STACK_WIND(frame,
- crypt_readdirp_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdirp,
- fd,
- size,
- offset,
- xdata);
- dict_unref(xdata);
- return 0;
- error:
- STACK_UNWIND_STRICT(readdirp, frame, -1, ret, NULL, NULL);
- return 0;
-}
-
-static int32_t crypt_access(call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- int32_t mask, dict_t *xdata)
-{
- gf_log(this->name, GF_LOG_WARNING,
- "NFS mounts of encrypted volumes are unsupported");
- STACK_UNWIND_STRICT(access, frame, -1, EPERM, NULL);
- return 0;
-}
-
-int32_t master_set_block_size (xlator_t *this, crypt_private_t *priv,
- dict_t *options)
-{
- uint64_t block_size = 0;
- struct master_cipher_info *master = get_master_cinfo(priv);
-
- if (options != NULL)
- GF_OPTION_RECONF("block-size", block_size, options,
- size_uint64, error);
- else
- GF_OPTION_INIT("block-size", block_size, size_uint64, error);
-
- switch (block_size) {
- case 512:
- master->m_block_bits = 9;
- break;
- case 1024:
- master->m_block_bits = 10;
- break;
- case 2048:
- master->m_block_bits = 11;
- break;
- case 4096:
- master->m_block_bits = 12;
- break;
- default:
- gf_log("crypt", GF_LOG_ERROR,
- "FATAL: unsupported block size %llu",
- (unsigned long long)block_size);
- goto error;
- }
- return 0;
- error:
- return -1;
-}
-
-int32_t master_set_alg(xlator_t *this, crypt_private_t *priv)
-{
- struct master_cipher_info *master = get_master_cinfo(priv);
- master->m_alg = AES_CIPHER_ALG;
- return 0;
-}
-
-int32_t master_set_mode(xlator_t *this, crypt_private_t *priv)
-{
- struct master_cipher_info *master = get_master_cinfo(priv);
- master->m_mode = XTS_CIPHER_MODE;
- return 0;
-}
-
-/*
- * set key size in bits to the master info
- * Pre-conditions: cipher mode in the master info is uptodate.
- */
-static int master_set_data_key_size (xlator_t *this, crypt_private_t *priv,
- dict_t *options)
-{
- int32_t ret;
- uint64_t key_size = 0;
- struct master_cipher_info *master = get_master_cinfo(priv);
-
- if (options != NULL)
- GF_OPTION_RECONF("data-key-size", key_size, options,
- uint64, error);
- else
- GF_OPTION_INIT("data-key-size", key_size, uint64, error);
-
- ret = data_cipher_algs[master->m_alg][master->m_mode].check_key(key_size);
- if (ret) {
- gf_log("crypt", GF_LOG_ERROR,
- "FATAL: wrong bin key size %llu for alg %d mode %d",
- (unsigned long long)key_size,
- (int)master->m_alg,
- (int)master->m_mode);
- goto error;
- }
- master->m_dkey_size = key_size;
- return 0;
- error:
- return -1;
-}
-
-static int is_hex(char *s) {
- return ('0' <= *s && *s <= '9') || ('a' <= *s && *s <= 'f');
-}
-
-static int parse_hex_buf(xlator_t *this, char *src, unsigned char *dst,
- int hex_size)
-{
- int i;
- int hex_byte = 0;
-
- for (i = 0; i < (hex_size / 2); i++) {
- if (!is_hex(src + i*2) || !is_hex(src + i*2 + 1)) {
- gf_log("crypt", GF_LOG_ERROR,
- "FATAL: not hex symbol in key");
- return -1;
- }
- if (sscanf(src + i*2, "%2x", &hex_byte) != 1) {
- gf_log("crypt", GF_LOG_ERROR,
- "FATAL: can not parse hex key");
- return -1;
- }
- dst[i] = hex_byte & 0xff;
- }
- return 0;
-}
-
-/*
- * Parse options;
- * install master volume key
- */
-int32_t master_set_master_vol_key(xlator_t *this, crypt_private_t *priv)
-{
- int32_t ret;
- FILE *file = NULL;
-
- int32_t key_size;
- char *opt_key_file_pathname = NULL;
-
- unsigned char bin_buf[MASTER_VOL_KEY_SIZE];
- char hex_buf[2 * MASTER_VOL_KEY_SIZE];
-
- struct master_cipher_info *master = get_master_cinfo(priv);
- /*
- * extract master key passed via option
- */
- GF_OPTION_INIT("master-key", opt_key_file_pathname, path, bad_key);
-
- if (!opt_key_file_pathname) {
- gf_log(this->name, GF_LOG_ERROR, "FATAL: missing master key");
- return -1;
- }
- gf_log(this->name, GF_LOG_DEBUG, "handling file key %s",
- opt_key_file_pathname);
-
- file = fopen(opt_key_file_pathname, "r");
- if (file == NULL) {
- gf_log(this->name, GF_LOG_ERROR,
- "FATAL: can not open file with master key");
- return -1;
- }
- /*
- * extract hex key
- */
- key_size = fread(hex_buf, 1, sizeof(hex_buf), file);
- if (key_size < sizeof(hex_buf)) {
- gf_log(this->name, GF_LOG_ERROR,
- "FATAL: master key is too short");
- goto bad_key;
- }
- ret = parse_hex_buf(this, hex_buf, bin_buf, key_size);
- if (ret)
- goto bad_key;
- memcpy(master->m_key, bin_buf, MASTER_VOL_KEY_SIZE);
- memset(hex_buf, 0, sizeof(hex_buf));
- fclose(file);
-
- memset(bin_buf, 0, sizeof(bin_buf));
- return 0;
- bad_key:
- gf_log(this->name, GF_LOG_ERROR, "FATAL: bad master key");
- if (file)
- fclose(file);
- memset(bin_buf, 0, sizeof(bin_buf));
- return -1;
-}
-
-/*
- * Derive volume key for object-id authentication
- */
-int32_t master_set_nmtd_vol_key(xlator_t *this, crypt_private_t *priv)
-{
- return get_nmtd_vol_key(get_master_cinfo(priv));
-}
-
-int32_t crypt_init_xlator(xlator_t *this)
-{
- int32_t ret;
- crypt_private_t *priv = this->private;
-
- ret = master_set_alg(this, priv);
- if (ret)
- return ret;
- ret = master_set_mode(this, priv);
- if (ret)
- return ret;
- ret = master_set_block_size(this, priv, NULL);
- if (ret)
- return ret;
- ret = master_set_data_key_size(this, priv, NULL);
- if (ret)
- return ret;
- ret = master_set_master_vol_key(this, priv);
- if (ret)
- return ret;
- return master_set_nmtd_vol_key(this, priv);
-}
-
-static int32_t crypt_alloc_private(xlator_t *this)
-{
- this->private = GF_CALLOC(1, sizeof(crypt_private_t), gf_crypt_mt_priv);
- if (!this->private) {
- gf_log("crypt", GF_LOG_ERROR,
- "Can not allocate memory for private data");
- return ENOMEM;
- }
- return 0;
-}
-
-static void crypt_free_private(xlator_t *this)
-{
- crypt_private_t *priv = this->private;
- if (priv) {
- memset(priv, 0, sizeof(*priv));
- GF_FREE(priv);
- }
-}
-
-int32_t
-mem_acct_init (xlator_t *this)
-{
- int ret = -1;
-
- if (!this)
- return ret;
-
- ret = xlator_mem_acct_init (this, gf_crypt_mt_end);
-
- if (ret != 0) {
- gf_log(this->name, GF_LOG_ERROR, "Memory accounting init"
- "failed");
- return ret;
- }
-
- return ret;
-}
-
-int32_t reconfigure (xlator_t *this, dict_t *options)
-{
- int32_t ret = -1;
- crypt_private_t *priv = NULL;
-
- GF_VALIDATE_OR_GOTO ("crypt", this, error);
- GF_VALIDATE_OR_GOTO (this->name, this->private, error);
- GF_VALIDATE_OR_GOTO (this->name, options, error);
-
- priv = this->private;
-
- ret = master_set_block_size(this, priv, options);
- if (ret) {
- gf_log("this->name", GF_LOG_ERROR,
- "Failed to reconfure block size");
- goto error;
- }
- ret = master_set_data_key_size(this, priv, options);
- if (ret) {
- gf_log("this->name", GF_LOG_ERROR,
- "Failed to reconfure data key size");
- goto error;
- }
- return 0;
- error:
- return ret;
-}
-
-int32_t init(xlator_t *this)
-{
- int32_t ret;
-
- if (!this->children || this->children->next) {
- gf_log ("crypt", GF_LOG_ERROR,
- "FATAL: crypt should have exactly one child");
- return EINVAL;
- }
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "dangling volume. check volfile ");
- }
- ret = crypt_alloc_private(this);
- if (ret)
- return ret;
- ret = crypt_init_xlator(this);
- if (ret)
- goto error;
- this->local_pool = mem_pool_new(crypt_local_t, 64);
- if (!this->local_pool) {
- gf_log(this->name, GF_LOG_ERROR,
- "failed to create local_t's memory pool");
- ret = ENOMEM;
- goto error;
- }
- gf_log ("crypt", GF_LOG_INFO, "crypt xlator loaded");
- return 0;
- error:
- crypt_free_private(this);
- return ret;
-}
-
-void fini (xlator_t *this)
-{
- crypt_free_private(this);
-}
-
-struct xlator_fops fops = {
- .readv = crypt_readv,
- .writev = crypt_writev,
- .truncate = crypt_truncate,
- .ftruncate = crypt_ftruncate,
- .setxattr = crypt_setxattr,
- .fsetxattr = crypt_fsetxattr,
- .link = crypt_link,
- .unlink = crypt_unlink,
- .rename = crypt_rename,
- .open = crypt_open,
- .create = crypt_create,
- .stat = crypt_stat,
- .fstat = crypt_fstat,
- .lookup = crypt_lookup,
- .readdirp = crypt_readdirp,
- .access = crypt_access
-};
-
-struct xlator_cbks cbks = {
- .forget = crypt_forget
-};
-
-struct volume_options options[] = {
- { .key = {"master-key"},
- .type = GF_OPTION_TYPE_PATH,
- .description = "Pathname of regular file which contains master volume key"
- },
- { .key = {"data-key-size"},
- .type = GF_OPTION_TYPE_SIZET,
- .description = "Data key size (bits)",
- .min = 256,
- .max = 512,
- .default_value = "256",
- },
- { .key = {"block-size"},
- .type = GF_OPTION_TYPE_SIZET,
- .description = "Atom size (bits)",
- .min = 512,
- .max = 4096,
- .default_value = "4096"
- },
- { .key = {NULL} },
-};
-
-/*
- Local variables:
- c-indentation-style: "K&R"
- mode-name: "LC"
- c-basic-offset: 8
- tab-width: 8
- fill-column: 80
- scroll-step: 1
- End:
-*/
diff --git a/xlators/encryption/crypt/src/crypt.h b/xlators/encryption/crypt/src/crypt.h
deleted file mode 100644
index 6c014535f0a..00000000000
--- a/xlators/encryption/crypt/src/crypt.h
+++ /dev/null
@@ -1,890 +0,0 @@
-/*
- Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#ifndef __CRYPT_H__
-#define __CRYPT_H__
-
-#include <openssl/aes.h>
-#include <openssl/evp.h>
-#include <openssl/sha.h>
-#include <openssl/hmac.h>
-#include <openssl/cmac.h>
-#include <openssl/modes.h>
-#include "crypt-mem-types.h"
-#include "compat.h"
-
-#define CRYPT_XLATOR_ID (0)
-
-#define MAX_IOVEC_BITS (3)
-#define MAX_IOVEC (1 << MAX_IOVEC_BITS)
-#define KEY_FACTOR_BITS (6)
-
-#define DEBUG_CRYPT (0)
-#define TRIVIAL_TFM (0)
-
-#define CRYPT_MIN_BLOCK_BITS (9)
-#define CRYPT_MAX_BLOCK_BITS (12)
-
-#define MASTER_VOL_KEY_SIZE (32)
-#define NMTD_VOL_KEY_SIZE (16)
-
-#if !defined(GF_LINUX_HOST_OS)
-typedef off_t loff_t;
-#endif
-
-struct crypt_key {
- uint32_t len;
- const char *label;
-};
-
-/*
- * Add new key types to the end of this
- * enumeration but before LAST_KEY_TYPE
- */
-typedef enum {
- MASTER_VOL_KEY,
- NMTD_VOL_KEY,
- NMTD_LINK_KEY,
- EMTD_FILE_KEY,
- DATA_FILE_KEY_256,
- DATA_FILE_KEY_512,
- LAST_KEY_TYPE
-}crypt_key_type;
-
-struct kderive_context {
- const unsigned char *pkey;/* parent key */
- uint32_t pkey_len; /* parent key size, bits */
- uint32_t ckey_len; /* child key size, bits */
- unsigned char *fid; /* fixed input data, NIST 800-108, 5.1 */
- uint32_t fid_len; /* fid len, bytes */
- unsigned char *out; /* contains child keying material */
- uint32_t out_len; /* out len, bytes */
-};
-
-typedef enum {
- DATA_ATOM,
- HOLE_ATOM,
- LAST_DATA_TYPE
-}atom_data_type;
-
-typedef enum {
- HEAD_ATOM,
- TAIL_ATOM,
- FULL_ATOM,
- LAST_LOCALITY_TYPE
-}atom_locality_type;
-
-typedef enum {
- MTD_CREATE,
- MTD_APPEND,
- MTD_OVERWRITE,
- MTD_CUT,
- MTD_LAST_OP
-} mtd_op_t;
-
-struct xts128_context {
- void *key1, *key2;
- block128_f block1,block2;
-};
-
-struct object_cipher_info {
- cipher_alg_t o_alg;
- cipher_mode_t o_mode;
- uint32_t o_block_bits;
- uint32_t o_dkey_size; /* raw data key size in bits */
- union {
- struct {
- unsigned char ivec[16];
- AES_KEY dkey[2];
- AES_KEY tkey; /* key used for tweaking */
- XTS128_CONTEXT xts;
- } aes_xts;
- } u;
-};
-
-struct master_cipher_info {
- /*
- * attributes inherited by newly created regular files
- */
- cipher_alg_t m_alg;
- cipher_mode_t m_mode;
- uint32_t m_block_bits;
- uint32_t m_dkey_size; /* raw key size in bits */
- /*
- * master key
- */
- unsigned char m_key[MASTER_VOL_KEY_SIZE];
- /*
- * volume key for oid authentication
- */
- unsigned char m_nmtd_key[NMTD_VOL_KEY_SIZE];
-};
-
-/*
-* This info is not changed during file's life
- */
-struct crypt_inode_info {
-#if DEBUG_CRYPT
- loc_t *loc; /* pathname that the file has been
- opened, or created with */
-#endif
- uint16_t nr_minor;
- uuid_t oid;
- struct object_cipher_info cinfo;
-};
-
-/*
- * this should locate in secure memory
- */
-typedef struct {
- struct master_cipher_info master;
-} crypt_private_t;
-
-static inline struct master_cipher_info *get_master_cinfo(crypt_private_t *priv)
-{
- return &priv->master;
-}
-
-static inline struct object_cipher_info *get_object_cinfo(struct crypt_inode_info
- *info)
-{
- return &info->cinfo;
-}
-
-/*
- * this describes layouts and properties
- * of atoms in an aligned vector
- */
-struct avec_config {
- uint32_t atom_size;
- atom_data_type type;
- size_t orig_size;
- off_t orig_offset;
- size_t expanded_size;
- off_t aligned_offset;
-
- uint32_t off_in_head;
- uint32_t off_in_tail;
- uint32_t gap_in_tail;
- uint32_t nr_full_blocks;
-
- struct iovec *avec; /* aligned vector */
- uint32_t acount; /* number of avec components. The same
- * as number of occupied logical blocks */
- char **pool;
- uint32_t blocks_in_pool;
- uint32_t cursor; /* makes sense only for ordered writes,
- * so there is no races on this counter.
- *
- * Cursor is per-config object, we don't
- * reset cursor for atoms of different
- * localities (head, tail, full)
- */
-};
-
-
-typedef struct {
- glusterfs_fop_t fop; /* code of FOP this local info built for */
- fd_t *fd;
- inode_t *inode;
- loc_t *loc;
- int32_t mac_idx;
- loc_t *newloc;
- int32_t flags;
- int32_t wbflags;
- struct crypt_inode_info *info;
- struct iobref *iobref;
- struct iobref *iobref_data;
- off_t offset;
-
- uint64_t old_file_size; /* per FOP, retrieved under lock held */
- uint64_t cur_file_size; /* per iteration, before issuing IOs */
- uint64_t new_file_size; /* per iteration, after issuing IOs */
-
- uint64_t io_offset; /* offset of IOs issued per iteration */
- uint64_t io_offset_nopad; /* offset of user's data in the atom */
- uint32_t io_size; /* size of IOs issued per iteration */
- uint32_t io_size_nopad; /* size of user's data in the IOs */
- uint32_t eof_padding_size; /* size od EOF padding in the IOs */
-
- gf_lock_t call_lock; /* protect nr_calls from many cbks */
- int32_t nr_calls;
-
- atom_data_type active_setup; /* which setup (hole or date)
- is currently active */
- /* data setup */
- struct avec_config data_conf;
-
- /* hole setup */
- int hole_conv_in_proggress;
- gf_lock_t hole_lock; /* protect hole config from many cbks */
- int hole_handled;
- struct avec_config hole_conf;
- struct iatt buf;
- struct iatt prebuf;
- struct iatt postbuf;
- struct iatt *prenewparent;
- struct iatt *postnewparent;
- int32_t op_ret;
- int32_t op_errno;
- int32_t rw_count; /* total read or written */
- gf_lock_t rw_count_lock; /* protect the counter above */
- unsigned char *format; /* for create, update format string */
- uint32_t format_size;
- uint32_t msgflags; /* messages for crypt_open() */
- dict_t *xdata;
- dict_t *xattr;
- struct iovec vec; /* contains last file's atom for
- read-prune-write sequence */
- gf_boolean_t custom_mtd;
- /*
- * the next 3 fields are used by readdir and friends
- */
- gf_dirent_t *de; /* directory entry */
- char *de_path; /* pathname of directory entry */
- uint32_t de_prefix_len; /* length of the parent's pathname */
- gf_dirent_t *entries;
-
- uint32_t update_disk_file_size:1;
-} crypt_local_t;
-
-/* This represents a (read)modify-write atom */
-struct rmw_atom {
- atom_locality_type locality;
- /*
- * read-modify-write sequence of the atom
- */
- int32_t (*rmw)(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iovec *vec,
- int32_t count,
- struct iatt *stbuf,
- struct iobref *iobref,
- dict_t *xdata);
- /*
- * offset of the logical block in a file
- */
- loff_t (*offset_at)(call_frame_t *frame,
- struct object_cipher_info *object);
- /*
- * IO offset in an atom
- */
- uint32_t (*offset_in)(call_frame_t *frame,
- struct object_cipher_info *object);
- /*
- * number of bytes of plain text of this atom that user
- * wants to read/write.
- * It can be smaller than atom_size in the case of head
- * or tail atoms.
- */
- uint32_t (*io_size_nopad)(call_frame_t *frame,
- struct object_cipher_info *object);
- /*
- * which iovec represents the atom
- */
- struct iovec *(*get_iovec)(call_frame_t *frame, uint32_t count);
- /*
- * how many bytes of partial block should be uptodated by
- * reading from disk.
- * This is used to perform a read component of RMW (read-modify-write).
- */
- uint32_t (*count_to_uptodate)(call_frame_t *frame, struct object_cipher_info *object);
- struct avec_config *(*get_config)(call_frame_t *frame);
-};
-
-struct data_cipher_alg {
- gf_boolean_t atomic; /* true means that algorithm requires
- to pad data before cipher transform */
- gf_boolean_t should_pad; /* true means that algorithm requires
- to pad the end of file with extra-data */
- uint32_t blkbits; /* blksize = 1 << blkbits */
- /*
- * any preliminary sanity checks goes here
- */
- int32_t (*init)(void);
- /*
- * set alg-mode specific inode info
- */
- int32_t (*set_private)(struct crypt_inode_info *info,
- struct master_cipher_info *master);
- /*
- * check alg-mode specific data key
- */
- int32_t (*check_key)(uint32_t key_size);
- void (*set_iv)(off_t offset, struct object_cipher_info *object);
- int32_t (*encrypt)(const unsigned char *from, unsigned char *to,
- size_t length, off_t offset, const int enc,
- struct object_cipher_info *object);
-};
-
-/*
- * version-dependent metadata loader
- */
-struct crypt_mtd_loader {
- /*
- * return core format size
- */
- size_t (*format_size)(mtd_op_t op, size_t old_size);
- /*
- * pack version-specific metadata of an object
- * at ->create()
- */
- int32_t (*create_format)(unsigned char *wire,
- loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master);
- /*
- * extract version-specific metadata of an object
- * at ->open() time
- */
- int32_t (*open_format)(unsigned char *wire,
- int32_t len,
- loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master,
- crypt_local_t *local,
- gf_boolean_t load_info);
- int32_t (*update_format)(unsigned char *new,
- unsigned char *old,
- size_t old_len,
- int32_t mac_idx,
- mtd_op_t op,
- loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master,
- crypt_local_t *local);
-};
-
-typedef int32_t (*end_writeback_handler_t)(call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iatt *prebuf,
- struct iatt *postbuf,
- dict_t *xdata);
-typedef void (*linkop_wind_handler_t)(call_frame_t *frame, xlator_t *this);
-typedef void (*linkop_unwind_handler_t)(call_frame_t *frame);
-
-
-/* Declarations */
-
-/* keys.c */
-extern struct crypt_key crypt_keys[LAST_KEY_TYPE];
-int32_t get_nmtd_vol_key(struct master_cipher_info *master);
-int32_t get_nmtd_link_key(loc_t *loc,
- struct master_cipher_info *master,
- unsigned char *result);
-int32_t get_emtd_file_key(struct crypt_inode_info *info,
- struct master_cipher_info *master,
- unsigned char *result);
-int32_t get_data_file_key(struct crypt_inode_info *info,
- struct master_cipher_info *master,
- uint32_t keysize,
- unsigned char *key);
-/* data.c */
-extern struct data_cipher_alg data_cipher_algs[LAST_CIPHER_ALG][LAST_CIPHER_MODE];
-void encrypt_aligned_iov(struct object_cipher_info *object,
- struct iovec *vec,
- int count,
- off_t off);
-void decrypt_aligned_iov(struct object_cipher_info *object,
- struct iovec *vec,
- int count,
- off_t off);
-int32_t align_iov_by_atoms(xlator_t *this,
- crypt_local_t *local,
- struct object_cipher_info *object,
- struct iovec *vec /* input vector */,
- int32_t count /* number of vec components */,
- struct iovec *avec /* aligned vector */,
- char **blocks /* pool of blocks */,
- uint32_t *blocks_allocated,
- struct avec_config *conf);
-int32_t set_config_avec_data(xlator_t *this,
- crypt_local_t *local,
- struct avec_config *conf,
- struct object_cipher_info *object,
- struct iovec *vec,
- int32_t vec_count);
-int32_t set_config_avec_hole(xlator_t *this,
- crypt_local_t *local,
- struct avec_config *conf,
- struct object_cipher_info *object,
- glusterfs_fop_t fop);
-void set_gap_at_end(call_frame_t *frame, struct object_cipher_info *object,
- struct avec_config *conf, atom_data_type dtype);
-void set_config_offsets(call_frame_t *frame,
- xlator_t *this,
- uint64_t offset,
- uint64_t count,
- atom_data_type dtype,
- int32_t setup_gap_in_tail);
-
-/* metadata.c */
-extern struct crypt_mtd_loader mtd_loaders [LAST_MTD_LOADER];
-
-int32_t alloc_format(crypt_local_t *local, size_t size);
-int32_t alloc_format_create(crypt_local_t *local);
-void free_format(crypt_local_t *local);
-size_t format_size(mtd_op_t op, size_t old_size);
-size_t new_format_size(void);
-int32_t open_format(unsigned char *str, int32_t len, loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master, crypt_local_t *local,
- gf_boolean_t load_info);
-int32_t update_format(unsigned char *new, unsigned char *old,
- size_t old_len, int32_t mac_idx, mtd_op_t op, loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master,
- crypt_local_t *local);
-int32_t create_format(unsigned char *wire,
- loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master);
-
-/* atom.c */
-struct rmw_atom *atom_by_types(atom_data_type data,
- atom_locality_type locality);
-void submit_partial(call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- atom_locality_type ltype);
-void submit_full(call_frame_t *frame, xlator_t *this);
-
-/* crypt.c */
-
-end_writeback_handler_t dispatch_end_writeback(glusterfs_fop_t fop);
-void set_local_io_params_writev(call_frame_t *frame,
- struct object_cipher_info *object,
- struct rmw_atom *atom, off_t io_offset,
- uint32_t io_size);
-void link_wind(call_frame_t *frame, xlator_t *this);
-void unlink_wind(call_frame_t *frame, xlator_t *this);
-void link_unwind(call_frame_t *frame);
-void unlink_unwind(call_frame_t *frame);
-void rename_wind(call_frame_t *frame, xlator_t *this);
-void rename_unwind(call_frame_t *frame);
-
-/* Inline functions */
-
-static inline int32_t crypt_xlator_id(void)
-{
- return CRYPT_XLATOR_ID;
-}
-
-static inline mtd_loader_id current_mtd_loader(void)
-{
- return MTD_LOADER_V1;
-}
-
-static inline uint32_t master_key_size (void)
-{
- return crypt_keys[MASTER_VOL_KEY].len >> 3;
-}
-
-static inline uint32_t nmtd_vol_key_size (void)
-{
- return crypt_keys[NMTD_VOL_KEY].len >> 3;
-}
-
-static inline uint32_t alg_mode_blkbits(cipher_alg_t alg,
- cipher_mode_t mode)
-{
- return data_cipher_algs[alg][mode].blkbits;
-}
-
-static inline uint32_t alg_mode_blksize(cipher_alg_t alg,
- cipher_mode_t mode)
-{
- return 1 << alg_mode_blkbits(alg, mode);
-}
-
-static inline gf_boolean_t alg_mode_atomic(cipher_alg_t alg,
- cipher_mode_t mode)
-{
- return data_cipher_algs[alg][mode].atomic;
-}
-
-static inline gf_boolean_t alg_mode_should_pad(cipher_alg_t alg,
- cipher_mode_t mode)
-{
- return data_cipher_algs[alg][mode].should_pad;
-}
-
-static inline uint32_t master_alg_blksize(struct master_cipher_info *mr)
-{
- return alg_mode_blksize(mr->m_alg, mr->m_mode);
-}
-
-static inline uint32_t master_alg_blkbits(struct master_cipher_info *mr)
-{
- return alg_mode_blkbits(mr->m_alg, mr->m_mode);
-}
-
-static inline gf_boolean_t master_alg_atomic(struct master_cipher_info *mr)
-{
- return alg_mode_atomic(mr->m_alg, mr->m_mode);
-}
-
-static inline gf_boolean_t master_alg_should_pad(struct master_cipher_info *mr)
-{
- return alg_mode_should_pad(mr->m_alg, mr->m_mode);
-}
-
-static inline uint32_t object_alg_blksize(struct object_cipher_info *ob)
-{
- return alg_mode_blksize(ob->o_alg, ob->o_mode);
-}
-
-static inline uint32_t object_alg_blkbits(struct object_cipher_info *ob)
-{
- return alg_mode_blkbits(ob->o_alg, ob->o_mode);
-}
-
-static inline gf_boolean_t object_alg_atomic(struct object_cipher_info *ob)
-{
- return alg_mode_atomic(ob->o_alg, ob->o_mode);
-}
-
-static inline gf_boolean_t object_alg_should_pad(struct object_cipher_info *ob)
-{
- return alg_mode_should_pad(ob->o_alg, ob->o_mode);
-}
-
-static inline uint32_t aes_raw_key_size(struct master_cipher_info *master)
-{
- return master->m_dkey_size >> 3;
-}
-
-static inline struct avec_config *get_hole_conf(call_frame_t *frame)
-{
- return &(((crypt_local_t *)frame->local)->hole_conf);
-}
-
-static inline struct avec_config *get_data_conf(call_frame_t *frame)
-{
- return &(((crypt_local_t *)frame->local)->data_conf);
-}
-
-static inline int32_t get_atom_bits (struct object_cipher_info *object)
-{
- return object->o_block_bits;
-}
-
-static inline int32_t get_atom_size (struct object_cipher_info *object)
-{
- return 1 << get_atom_bits(object);
-}
-
-static inline int32_t has_head_block(struct avec_config *conf)
-{
- return conf->off_in_head ||
- (conf->acount == 1 && conf->off_in_tail);
-}
-
-static inline int32_t has_tail_block(struct avec_config *conf)
-{
- return conf->off_in_tail && conf->acount > 1;
-}
-
-static inline int32_t has_full_blocks(struct avec_config *conf)
-{
- return conf->nr_full_blocks;
-}
-
-static inline int32_t should_submit_head_block(struct avec_config *conf)
-{
- return has_head_block(conf) && (conf->cursor == 0);
-}
-
-static inline int32_t should_submit_tail_block(struct avec_config *conf)
-{
- return has_tail_block(conf) && (conf->cursor == conf->acount - 1);
-}
-
-static inline int32_t should_submit_full_block(struct avec_config *conf)
-{
- uint32_t start = has_head_block(conf) ? 1 : 0;
-
- return has_full_blocks(conf) &&
- conf->cursor >= start &&
- conf->cursor < start + conf->nr_full_blocks;
-}
-
-#if DEBUG_CRYPT
-static inline void crypt_check_input_len(size_t len,
- struct object_cipher_info *object)
-{
- if (object_alg_should_pad(object) && (len & (object_alg_blksize(object) - 1)))
- gf_log ("crypt", GF_LOG_DEBUG, "bad input len: %d", (int)len);
-}
-
-static inline void check_head_block(struct avec_config *conf)
-{
- if (!has_head_block(conf))
- gf_log("crypt", GF_LOG_DEBUG, "not a head atom");
-}
-
-static inline void check_tail_block(struct avec_config *conf)
-{
- if (!has_tail_block(conf))
- gf_log("crypt", GF_LOG_DEBUG, "not a tail atom");
-}
-
-static inline void check_full_block(struct avec_config *conf)
-{
- if (!has_full_blocks(conf))
- gf_log("crypt", GF_LOG_DEBUG, "not a full atom");
-}
-
-static inline void check_cursor_head(struct avec_config *conf)
-{
- if (!has_head_block(conf))
- gf_log("crypt",
- GF_LOG_DEBUG, "Illegal call of head atom method");
- else if (conf->cursor != 0)
- gf_log("crypt",
- GF_LOG_DEBUG, "Cursor (%d) is not at head atom",
- conf->cursor);
-}
-
-static inline void check_cursor_full(struct avec_config *conf)
-{
- if (!has_full_blocks(conf))
- gf_log("crypt",
- GF_LOG_DEBUG, "Illegal call of full atom method");
- if (has_head_block(conf) && (conf->cursor == 0))
- gf_log("crypt",
- GF_LOG_DEBUG, "Cursor is not at full atom");
-}
-
-/*
- * FIXME: use avec->iov_len to check setup
- */
-static inline int data_local_invariant(crypt_local_t *local)
-{
- return 0;
-}
-
-#else
-#define crypt_check_input_len(len, object) noop
-#define check_head_block(conf) noop
-#define check_tail_block(conf) noop
-#define check_full_block(conf) noop
-#define check_cursor_head(conf) noop
-#define check_cursor_full(conf) noop
-
-#endif /* DEBUG_CRYPT */
-
-static inline struct avec_config *conf_by_type(call_frame_t *frame,
- atom_data_type dtype)
-{
- struct avec_config *conf = NULL;
-
- switch (dtype) {
- case HOLE_ATOM:
- conf = get_hole_conf(frame);
- break;
- case DATA_ATOM:
- conf = get_data_conf(frame);
- break;
- default:
- gf_log("crypt", GF_LOG_DEBUG, "bad atom type");
- }
- return conf;
-}
-
-static inline uint32_t nr_calls_head(struct avec_config *conf)
-{
- return has_head_block(conf) ? 1 : 0;
-}
-
-static inline uint32_t nr_calls_tail(struct avec_config *conf)
-{
- return has_tail_block(conf) ? 1 : 0;
-}
-
-static inline uint32_t nr_calls_full(struct avec_config *conf)
-{
- switch(conf->type) {
- case HOLE_ATOM:
- return has_full_blocks(conf);
- case DATA_ATOM:
- return has_full_blocks(conf) ?
- logical_blocks_occupied(0,
- conf->nr_full_blocks,
- MAX_IOVEC_BITS) : 0;
- default:
- gf_log("crypt", GF_LOG_DEBUG, "bad atom data type");
- return 0;
- }
-}
-
-static inline uint32_t nr_calls(struct avec_config *conf)
-{
- return nr_calls_head(conf) + nr_calls_tail(conf) + nr_calls_full(conf);
-}
-
-static inline uint32_t nr_calls_data(call_frame_t *frame)
-{
- return nr_calls(get_data_conf(frame));
-}
-
-static inline uint32_t nr_calls_hole(call_frame_t *frame)
-{
- return nr_calls(get_hole_conf(frame));
-}
-
-static inline void get_one_call_nolock(call_frame_t *frame)
-{
- crypt_local_t *local = frame->local;
-
- ++local->nr_calls;
-
- //gf_log("crypt", GF_LOG_DEBUG, "get %d calls", 1);
-}
-
-static inline void get_one_call(call_frame_t *frame)
-{
- crypt_local_t *local = frame->local;
-
- LOCK(&local->call_lock);
- get_one_call_nolock(frame);
- UNLOCK(&local->call_lock);
-}
-
-static inline void get_nr_calls_nolock(call_frame_t *frame, int32_t nr)
-{
- crypt_local_t *local = frame->local;
-
- local->nr_calls += nr;
-
- //gf_log("crypt", GF_LOG_DEBUG, "get %d calls", nr);
-}
-
-static inline void get_nr_calls(call_frame_t *frame, int32_t nr)
-{
- crypt_local_t *local = frame->local;
-
- LOCK(&local->call_lock);
- get_nr_calls_nolock(frame, nr);
- UNLOCK(&local->call_lock);
-}
-
-static inline int put_one_call(crypt_local_t *local)
-{
- uint32_t last = 0;
-
- LOCK(&local->call_lock);
- if (--local->nr_calls == 0)
- last = 1;
-
- //gf_log("crypt", GF_LOG_DEBUG, "put %d calls", 1);
-
- UNLOCK(&local->call_lock);
- return last;
-}
-
-static inline int is_appended_write(call_frame_t *frame)
-{
- crypt_local_t *local = frame->local;
- struct avec_config *conf = get_data_conf(frame);
-
- return conf->orig_offset + conf->orig_size > local->old_file_size;
-}
-
-static inline int is_ordered_mode(call_frame_t *frame)
-{
-#if 0
- crypt_local_t *local = frame->local;
- return local->fop == GF_FOP_FTRUNCATE ||
- (local->fop == GF_FOP_WRITE && is_appended_write(frame));
-#endif
- return 1;
-}
-
-static inline int32_t hole_conv_completed(crypt_local_t *local)
-{
- struct avec_config *conf = &local->hole_conf;
- return conf->cursor == conf->acount;
-}
-
-static inline int32_t data_write_in_progress(crypt_local_t *local)
-{
- return local->active_setup == DATA_ATOM;
-}
-
-static inline int32_t parent_is_crypt_xlator(call_frame_t *frame,
- xlator_t *this)
-{
- return frame->parent->this == this;
-}
-
-static inline linkop_wind_handler_t linkop_wind_dispatch(glusterfs_fop_t fop)
-{
- switch(fop){
- case GF_FOP_LINK:
- return link_wind;
- case GF_FOP_UNLINK:
- return unlink_wind;
- case GF_FOP_RENAME:
- return rename_wind;
- default:
- gf_log("crypt", GF_LOG_ERROR, "Bad link operation %d", fop);
- return NULL;
- }
-}
-
-static inline linkop_unwind_handler_t linkop_unwind_dispatch(glusterfs_fop_t fop)
-{
- switch(fop){
- case GF_FOP_LINK:
- return link_unwind;
- case GF_FOP_UNLINK:
- return unlink_unwind;
- case GF_FOP_RENAME:
- return rename_unwind;
- default:
- gf_log("crypt", GF_LOG_ERROR, "Bad link operation %d", fop);
- return NULL;
- }
-}
-
-static inline mtd_op_t linkop_mtdop_dispatch(glusterfs_fop_t fop)
-{
- switch (fop) {
- case GF_FOP_LINK:
- return MTD_APPEND;
- case GF_FOP_UNLINK:
- return MTD_CUT;
- case GF_FOP_RENAME:
- return MTD_OVERWRITE;
- default:
- gf_log("crypt", GF_LOG_WARNING, "Bad link operation %d", fop);
- return MTD_LAST_OP;
- }
-}
-
-#endif /* __CRYPT_H__ */
-
-/*
- Local variables:
- c-indentation-style: "K&R"
- mode-name: "LC"
- c-basic-offset: 8
- tab-width: 8
- fill-column: 80
- scroll-step: 1
- End:
-*/
diff --git a/xlators/encryption/crypt/src/data.c b/xlators/encryption/crypt/src/data.c
deleted file mode 100644
index 8b467bf2c60..00000000000
--- a/xlators/encryption/crypt/src/data.c
+++ /dev/null
@@ -1,764 +0,0 @@
-/*
- Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#include "defaults.h"
-#include "crypt-common.h"
-#include "crypt.h"
-
-static void set_iv_aes_xts(off_t offset, struct object_cipher_info *object)
-{
- unsigned char *ivec;
-
- ivec = object->u.aes_xts.ivec;
-
- /* convert the tweak into a little-endian byte
- * array (IEEE P1619/D16, May 2007, section 5.1)
- */
-
- *((uint64_t *)ivec) = htole64(offset);
-
- /* ivec is padded with zeroes */
-}
-
-static int32_t aes_set_keys_common(unsigned char *raw_key, uint32_t key_size,
- AES_KEY *keys)
-{
- int32_t ret;
-
- ret = AES_set_encrypt_key(raw_key,
- key_size,
- &keys[AES_ENCRYPT]);
- if (ret) {
- gf_log("crypt", GF_LOG_ERROR, "Set encrypt key failed");
- return ret;
- }
- ret = AES_set_decrypt_key(raw_key,
- key_size,
- &keys[AES_DECRYPT]);
- if (ret) {
- gf_log("crypt", GF_LOG_ERROR, "Set decrypt key failed");
- return ret;
- }
- return 0;
-}
-
-/*
- * set private cipher info for xts mode
- */
-static int32_t set_private_aes_xts(struct crypt_inode_info *info,
- struct master_cipher_info *master)
-{
- int ret;
- struct object_cipher_info *object = get_object_cinfo(info);
- unsigned char *data_key;
- uint32_t subkey_size;
-
- /* init tweak value */
- memset(object->u.aes_xts.ivec, 0, 16);
-
- data_key = GF_CALLOC(1, object->o_dkey_size, gf_crypt_mt_key);
- if (!data_key)
- return ENOMEM;
-
- /*
- * retrieve data keying meterial
- */
- ret = get_data_file_key(info, master, object->o_dkey_size, data_key);
- if (ret) {
- gf_log("crypt", GF_LOG_ERROR, "Failed to retrieve data key");
- GF_FREE(data_key);
- return ret;
- }
- /*
- * parse compound xts key
- */
- subkey_size = object->o_dkey_size >> 4; /* (xts-key-size-in-bytes / 2) */
- /*
- * install key for data encryption
- */
- ret = aes_set_keys_common(data_key,
- subkey_size << 3, object->u.aes_xts.dkey);
- if (ret) {
- GF_FREE(data_key);
- return ret;
- }
- /*
- * set up key used to encrypt tweaks
- */
- ret = AES_set_encrypt_key(data_key + subkey_size,
- object->o_dkey_size / 2,
- &object->u.aes_xts.tkey);
- if (ret < 0)
- gf_log("crypt", GF_LOG_ERROR, "Set tweak key failed");
-
- GF_FREE(data_key);
- return ret;
-}
-
-static int32_t aes_xts_init(void)
-{
- cassert(AES_BLOCK_SIZE == (1 << AES_BLOCK_BITS));
- return 0;
-}
-
-static int32_t check_key_aes_xts(uint32_t keysize)
-{
- switch(keysize) {
- case 256:
- case 512:
- return 0;
- default:
- break;
- }
- return -1;
-}
-
-static int32_t encrypt_aes_xts(const unsigned char *from,
- unsigned char *to, size_t length,
- off_t offset, const int enc,
- struct object_cipher_info *object)
-{
- XTS128_CONTEXT ctx;
- if (enc) {
- ctx.key1 = &object->u.aes_xts.dkey[AES_ENCRYPT];
- ctx.block1 = (block128_f)AES_encrypt;
- }
- else {
- ctx.key1 = &object->u.aes_xts.dkey[AES_DECRYPT];
- ctx.block1 = (block128_f)AES_decrypt;
- }
- ctx.key2 = &object->u.aes_xts.tkey;
- ctx.block2 = (block128_f)AES_encrypt;
-
- return CRYPTO_xts128_encrypt(&ctx,
- object->u.aes_xts.ivec,
- from,
- to,
- length, enc);
-}
-
-/*
- * Cipher input chunk @from of length @len;
- * @to: result of cipher transform;
- * @off: offset in a file (must be cblock-aligned);
- */
-static void cipher_data(struct object_cipher_info *object,
- char *from,
- char *to,
- off_t off,
- size_t len,
- const int enc)
-{
- crypt_check_input_len(len, object);
-
-#if TRIVIAL_TFM && DEBUG_CRYPT
- return;
-#endif
- data_cipher_algs[object->o_alg][object->o_mode].set_iv(off, object);
- data_cipher_algs[object->o_alg][object->o_mode].encrypt
- ((const unsigned char *)from,
- (unsigned char *)to,
- len,
- off,
- enc,
- object);
-}
-
-#define MAX_CIPHER_CHUNK (1 << 30)
-
-/*
- * Do cipher (encryption/decryption) transform of a
- * continuous region of memory.
- *
- * @len: a number of bytes to transform;
- * @buf: data to transform;
- * @off: offset in a file, should be block-aligned
- * for atomic cipher modes and ksize-aligned
- * for other modes).
- * @dir: direction of transform (encrypt/decrypt).
- */
-static void cipher_region(struct object_cipher_info *object,
- char *from,
- char *to,
- off_t off,
- size_t len,
- int dir)
-{
- while (len > 0) {
- size_t to_cipher;
-
- to_cipher = len;
- if (to_cipher > MAX_CIPHER_CHUNK)
- to_cipher = MAX_CIPHER_CHUNK;
-
- /* this will reset IV */
- cipher_data(object,
- from,
- to,
- off,
- to_cipher,
- dir);
- from += to_cipher;
- to += to_cipher;
- off += to_cipher;
- len -= to_cipher;
- }
-}
-
-/*
- * Do cipher transform (encryption/decryption) of
- * plaintext/ciphertext represented by @vec.
- *
- * Pre-conditions: @vec represents a continuous piece
- * of data in a file at offset @off to be ciphered
- * (encrypted/decrypted).
- * @count is the number of vec's components. All the
- * components must be block-aligned, the caller is
- * responsible for this. @dir is "direction" of
- * transform (encrypt/decrypt).
- */
-static void cipher_aligned_iov(struct object_cipher_info *object,
- struct iovec *vec,
- int count,
- off_t off,
- int32_t dir)
-{
- int i;
- int len = 0;
-
- for (i = 0; i < count; i++) {
- cipher_region(object,
- vec[i].iov_base,
- vec[i].iov_base,
- off + len,
- vec[i].iov_len,
- dir);
- len += vec[i].iov_len;
- }
-}
-
-void encrypt_aligned_iov(struct object_cipher_info *object,
- struct iovec *vec,
- int count,
- off_t off)
-{
- cipher_aligned_iov(object, vec, count, off, 1);
-}
-
-void decrypt_aligned_iov(struct object_cipher_info *object,
- struct iovec *vec,
- int count,
- off_t off)
-{
- cipher_aligned_iov(object, vec, count, off, 0);
-}
-
-#if DEBUG_CRYPT
-static void compound_stream(struct iovec *vec, int count, char *buf, off_t skip)
-{
- int i;
- int off = 0;
- for (i = 0; i < count; i++) {
- memcpy(buf + off,
- vec[i].iov_base + skip,
- vec[i].iov_len - skip);
-
- off += (vec[i].iov_len - skip);
- skip = 0;
- }
-}
-
-static void check_iovecs(struct iovec *vec, int cnt,
- struct iovec *avec, int acnt, uint32_t off_in_head)
-{
- char *s1, *s2;
- uint32_t size, asize;
-
- size = iov_length(vec, cnt);
- asize = iov_length(avec, acnt) - off_in_head;
- if (size != asize) {
- gf_log("crypt", GF_LOG_DEBUG, "size %d is not eq asize %d",
- size, asize);
- return;
- }
- s1 = GF_CALLOC(1, size, gf_crypt_mt_data);
- if (!s1) {
- gf_log("crypt", GF_LOG_DEBUG, "Can not allocate stream ");
- return;
- }
- s2 = GF_CALLOC(1, asize, gf_crypt_mt_data);
- if (!s2) {
- GF_FREE(s1);
- gf_log("crypt", GF_LOG_DEBUG, "Can not allocate stream ");
- return;
- }
- compound_stream(vec, cnt, s1, 0);
- compound_stream(avec, acnt, s2, off_in_head);
- if (memcmp(s1, s2, size))
- gf_log("crypt", GF_LOG_DEBUG, "chunks of different data");
- GF_FREE(s1);
- GF_FREE(s2);
-}
-
-#else
-#define check_iovecs(vec, count, avec, avecn, off) noop
-#endif /* DEBUG_CRYPT */
-
-static char *data_alloc_block(xlator_t *this, crypt_local_t *local,
- int32_t block_size)
-{
- struct iobuf *iobuf = NULL;
-
- iobuf = iobuf_get2(this->ctx->iobuf_pool, block_size);
- if (!iobuf) {
- gf_log("crypt", GF_LOG_ERROR,
- "Failed to get iobuf");
- return NULL;
- }
- if (!local->iobref_data) {
- local->iobref_data = iobref_new();
- if (!local->iobref_data) {
- gf_log("crypt", GF_LOG_ERROR,
- "Failed to get iobref");
- iobuf_unref(iobuf);
- return NULL;
- }
- }
- iobref_add(local->iobref_data, iobuf);
- return iobuf->ptr;
-}
-
-/*
- * Compound @avec, which represent the same data
- * chunk as @vec, but has aligned components of
- * specified block size. Alloc blocks, if needed.
- * In particular, incomplete head and tail blocks
- * must be allocated.
- * Put number of allocated blocks to @num_blocks.
- *
- * Example:
- *
- * input: data chunk represented by 4 components
- * [AB],[BC],[CD],[DE];
- * output: 5 logical blocks (0, 1, 2, 3, 4).
- *
- * A B C D E
- * *-----*+------*-+---*----+--------+-*
- * | || | | | | | |
- * *-+-----+*------+-*---+----*--------*-+------*
- * 0 1 2 3 4
- *
- * 0 - incomplete compound (head);
- * 1, 2 - full compound;
- * 3 - full non-compound (the case of reuse);
- * 4 - incomplete non-compound (tail).
- */
-int32_t align_iov_by_atoms(xlator_t *this,
- crypt_local_t *local,
- struct object_cipher_info *object,
- struct iovec *vec /* input vector */,
- int32_t count /* number of vec components */,
- struct iovec *avec /* aligned vector */,
- char **blocks /* pool of blocks */,
- uint32_t *blocks_allocated,
- struct avec_config *conf)
-{
- int vecn = 0; /* number of the current component in vec */
- int avecn = 0; /* number of the current component in avec */
- off_t vec_off = 0; /* offset in the current vec component,
- * i.e. the number of bytes have already
- * been copied */
- int32_t block_size = get_atom_size(object);
- size_t to_process; /* number of vec's bytes to copy and(or) re-use */
- int32_t off_in_head = conf->off_in_head;
-
- to_process = iov_length(vec, count);
-
- while (to_process > 0) {
- if (off_in_head ||
- vec[vecn].iov_len - vec_off < block_size) {
- /*
- * less than block_size:
- * the case of incomplete (head or tail),
- * or compound block
- */
- size_t copied = 0;
- /*
- * populate the pool with a new block
- */
- blocks[*blocks_allocated] = data_alloc_block(this,
- local,
- block_size);
- if (!blocks[*blocks_allocated])
- return -ENOMEM;
- memset(blocks[*blocks_allocated], 0, off_in_head);
- /*
- * fill the block with vec components
- */
- do {
- size_t to_copy;
-
- to_copy = vec[vecn].iov_len - vec_off;
- if (to_copy > block_size - off_in_head)
- to_copy = block_size - off_in_head;
-
- memcpy(blocks[*blocks_allocated] + off_in_head + copied,
- vec[vecn].iov_base + vec_off,
- to_copy);
-
- copied += to_copy;
- to_process -= to_copy;
-
- vec_off += to_copy;
- if (vec_off == vec[vecn].iov_len) {
- /* finished with this vecn */
- vec_off = 0;
- vecn++;
- }
- } while (copied < (block_size - off_in_head) && to_process > 0);
- /*
- * update avec
- */
- avec[avecn].iov_len = off_in_head + copied;
- avec[avecn].iov_base = blocks[*blocks_allocated];
-
- (*blocks_allocated)++;
- off_in_head = 0;
- } else {
- /*
- * the rest of the current vec component
- * is not less than block_size, so reuse
- * the memory buffer of the component.
- */
- size_t to_reuse;
- to_reuse = (to_process > block_size ?
- block_size :
- to_process);
- avec[avecn].iov_len = to_reuse;
- avec[avecn].iov_base = vec[vecn].iov_base + vec_off;
-
- vec_off += to_reuse;
- if (vec_off == vec[vecn].iov_len) {
- /* finished with this vecn */
- vec_off = 0;
- vecn++;
- }
- to_process -= to_reuse;
- }
- avecn++;
- }
- check_iovecs(vec, count, avec, avecn, conf->off_in_head);
- return 0;
-}
-
-/*
- * allocate and setup aligned vector for data submission
- * Pre-condition: @conf is set.
- */
-int32_t set_config_avec_data(xlator_t *this,
- crypt_local_t *local,
- struct avec_config *conf,
- struct object_cipher_info *object,
- struct iovec *vec,
- int32_t vec_count)
-{
- int32_t ret = ENOMEM;
- struct iovec *avec;
- char **pool;
- uint32_t blocks_in_pool = 0;
-
- conf->type = DATA_ATOM;
-
- avec = GF_CALLOC(conf->acount, sizeof(*avec), gf_crypt_mt_iovec);
- if (!avec)
- return ret;
- pool = GF_CALLOC(conf->acount, sizeof(pool), gf_crypt_mt_char);
- if (!pool) {
- GF_FREE(avec);
- return ret;
- }
- if (!vec) {
- /*
- * degenerated case: no data
- */
- pool[0] = data_alloc_block(this, local, get_atom_size(object));
- if (!pool[0])
- goto free;
- blocks_in_pool = 1;
- avec->iov_base = pool[0];
- avec->iov_len = conf->off_in_tail;
- }
- else {
- ret = align_iov_by_atoms(this, local, object, vec, vec_count,
- avec, pool, &blocks_in_pool, conf);
- if (ret)
- goto free;
- }
- conf->avec = avec;
- conf->pool = pool;
- conf->blocks_in_pool = blocks_in_pool;
- return 0;
- free:
- GF_FREE(avec);
- GF_FREE(pool);
- return ret;
-}
-
-/*
- * allocate and setup aligned vector for hole submission
- */
-int32_t set_config_avec_hole(xlator_t *this,
- crypt_local_t *local,
- struct avec_config *conf,
- struct object_cipher_info *object,
- glusterfs_fop_t fop)
-{
- uint32_t i, idx;
- struct iovec *avec;
- char **pool;
- uint32_t num_blocks;
- uint32_t blocks_in_pool = 0;
-
- conf->type = HOLE_ATOM;
-
- num_blocks = conf->acount -
- (conf->nr_full_blocks ? conf->nr_full_blocks - 1 : 0);
-
- switch (fop) {
- case GF_FOP_WRITE:
- /*
- * hole goes before data
- */
- if (num_blocks == 1 && conf->off_in_tail != 0)
- /*
- * we won't submit a hole which fits into
- * a data atom: this part of hole will be
- * submitted with data write
- */
- return 0;
- break;
- case GF_FOP_FTRUNCATE:
- /*
- * expanding truncate, hole goes after data,
- * and will be submited in any case.
- */
- break;
- default:
- gf_log("crypt", GF_LOG_WARNING,
- "bad file operation %d", fop);
- return 0;
- }
- avec = GF_CALLOC(num_blocks, sizeof(*avec), gf_crypt_mt_iovec);
- if (!avec)
- return ENOMEM;
- pool = GF_CALLOC(num_blocks, sizeof(pool), gf_crypt_mt_char);
- if (!pool) {
- GF_FREE(avec);
- return ENOMEM;
- }
- for (i = 0; i < num_blocks; i++) {
- pool[i] = data_alloc_block(this, local, get_atom_size(object));
- if (pool[i] == NULL)
- goto free;
- blocks_in_pool++;
- }
- if (has_head_block(conf)) {
- /* set head block */
- idx = 0;
- avec[idx].iov_base = pool[idx];
- avec[idx].iov_len = get_atom_size(object);
- memset(avec[idx].iov_base + conf->off_in_head,
- 0,
- get_atom_size(object) - conf->off_in_head);
- }
- if (has_tail_block(conf)) {
- /* set tail block */
- idx = num_blocks - 1;
- avec[idx].iov_base = pool[idx];
- avec[idx].iov_len = get_atom_size(object);
- memset(avec[idx].iov_base, 0, conf->off_in_tail);
- }
- if (has_full_blocks(conf)) {
- /* set full block */
- idx = conf->off_in_head ? 1 : 0;
- avec[idx].iov_base = pool[idx];
- avec[idx].iov_len = get_atom_size(object);
- /*
- * since we re-use the buffer,
- * zeroes will be set every time
- * before encryption, see submit_full()
- */
- }
- conf->avec = avec;
- conf->pool = pool;
- conf->blocks_in_pool = blocks_in_pool;
- return 0;
- free:
- GF_FREE(avec);
- GF_FREE(pool);
- return ENOMEM;
-}
-
-/* A helper for setting up config of partial atoms (which
- * participate in read-modify-write sequence).
- *
- * Calculate and setup precise amount of "extra-bytes"
- * that should be uptodated at the end of partial (not
- * necessarily tail!) block.
- *
- * Pre-condition: local->old_file_size is valid!
- * @conf contains setup, which is enough for correct calculation
- * of has_tail_block(), ->get_offset().
- */
-void set_gap_at_end(call_frame_t *frame, struct object_cipher_info *object,
- struct avec_config *conf, atom_data_type dtype)
-{
- uint32_t to_block;
- crypt_local_t *local = frame->local;
- uint64_t old_file_size = local->old_file_size;
- struct rmw_atom *partial = atom_by_types(dtype,
- has_tail_block(conf) ?
- TAIL_ATOM : HEAD_ATOM);
-
- if (old_file_size <= partial->offset_at(frame, object))
- to_block = 0;
- else {
- to_block = old_file_size - partial->offset_at(frame, object);
- if (to_block > get_atom_size(object))
- to_block = get_atom_size(object);
- }
- if (to_block > conf->off_in_tail)
- conf->gap_in_tail = to_block - conf->off_in_tail;
- else
- /*
- * nothing to uptodate
- */
- conf->gap_in_tail = 0;
-}
-
-/*
- * fill struct avec_config with offsets layouts
- */
-void set_config_offsets(call_frame_t *frame,
- xlator_t *this,
- uint64_t offset,
- uint64_t count,
- atom_data_type dtype,
- int32_t set_gap)
-{
- crypt_local_t *local;
- struct object_cipher_info *object;
- struct avec_config *conf;
- uint32_t resid;
-
- uint32_t atom_size;
- uint32_t atom_bits;
-
- size_t orig_size;
- off_t orig_offset;
- size_t expanded_size;
- off_t aligned_offset;
-
- uint32_t off_in_head = 0;
- uint32_t off_in_tail = 0;
- uint32_t nr_full_blocks;
- int32_t size_full_blocks;
-
- uint32_t acount; /* number of alifned components to write.
- * The same as number of occupied logical
- * blocks (atoms)
- */
- local = frame->local;
- object = &local->info->cinfo;
- conf = (dtype == DATA_ATOM ?
- get_data_conf(frame) : get_hole_conf(frame));
-
- orig_offset = offset;
- orig_size = count;
-
- atom_size = get_atom_size(object);
- atom_bits = get_atom_bits(object);
-
- /*
- * Round-down the start,
- * round-up the end.
- */
- resid = offset & (uint64_t)(atom_size - 1);
-
- if (resid)
- off_in_head = resid;
- aligned_offset = offset - off_in_head;
- expanded_size = orig_size + off_in_head;
-
- /* calculate tail,
- expand size forward */
- resid = (offset + orig_size) & (uint64_t)(atom_size - 1);
-
- if (resid) {
- off_in_tail = resid;
- expanded_size += (atom_size - off_in_tail);
- }
- /*
- * calculate number of occupied blocks
- */
- acount = expanded_size >> atom_bits;
- /*
- * calculate number of full blocks
- */
- size_full_blocks = expanded_size;
- if (off_in_head)
- size_full_blocks -= atom_size;
- if (off_in_tail && size_full_blocks > 0)
- size_full_blocks -= atom_size;
- nr_full_blocks = size_full_blocks >> atom_bits;
-
- conf->atom_size = atom_size;
- conf->orig_size = orig_size;
- conf->orig_offset = orig_offset;
- conf->expanded_size = expanded_size;
- conf->aligned_offset = aligned_offset;
-
- conf->off_in_head = off_in_head;
- conf->off_in_tail = off_in_tail;
- conf->nr_full_blocks = nr_full_blocks;
- conf->acount = acount;
- /*
- * Finally, calculate precise amount of
- * "extra-bytes" that should be uptodated
- * at the end.
- * Only if RMW is expected.
- */
- if (off_in_tail && set_gap)
- set_gap_at_end(frame, object, conf, dtype);
-}
-
-struct data_cipher_alg data_cipher_algs[LAST_CIPHER_ALG][LAST_CIPHER_MODE] = {
- [AES_CIPHER_ALG][XTS_CIPHER_MODE] =
- { .atomic = _gf_true,
- .should_pad = _gf_true,
- .blkbits = AES_BLOCK_BITS,
- .init = aes_xts_init,
- .set_private = set_private_aes_xts,
- .check_key = check_key_aes_xts,
- .set_iv = set_iv_aes_xts,
- .encrypt = encrypt_aes_xts
- }
-};
-
-/*
- Local variables:
- c-indentation-style: "K&R"
- mode-name: "LC"
- c-basic-offset: 8
- tab-width: 8
- fill-column: 80
- scroll-step: 1
- End:
-*/
diff --git a/xlators/encryption/crypt/src/keys.c b/xlators/encryption/crypt/src/keys.c
deleted file mode 100644
index e9da55960c8..00000000000
--- a/xlators/encryption/crypt/src/keys.c
+++ /dev/null
@@ -1,310 +0,0 @@
-/*
- Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#include "defaults.h"
-#include "crypt-common.h"
-#include "crypt.h"
-
-/* Key hierarchy
-
- +----------------+
- | MASTER_VOL_KEY |
- +-------+--------+
- |
- |
- +----------------+----------------+
- | | |
- | | |
- +-------+------+ +-------+-------+ +------+--------+
- | NMTD_VOL_KEY | | EMTD_FILE_KEY | | DATA_FILE_KEY |
- +-------+------+ +---------------+ +---------------+
- |
- |
- +-------+-------+
- | NMTD_LINK_KEY |
- +---------------+
-
- */
-
-#if DEBUG_CRYPT
-static void check_prf_iters(uint32_t num_iters)
-{
- if (num_iters == 0)
- gf_log ("crypt", GF_LOG_DEBUG,
- "bad number of prf iterations : %d", num_iters);
-}
-#else
-#define check_prf_iters(num_iters) noop
-#endif /* DEBUG_CRYPT */
-
-unsigned char crypt_fake_oid[16] =
- {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
-
-/*
- * derive key in the counter mode using
- * sha256-based HMAC as PRF, see
- * NIST Special Publication 800-108, 5.1)
- */
-
-#define PRF_OUTPUT_SIZE SHA256_DIGEST_LENGTH
-
-static int32_t kderive_init(struct kderive_context *ctx,
- const unsigned char *pkey, /* parent key */
- uint32_t pkey_size, /* parent key size */
- const unsigned char *idctx, /* id-context */
- uint32_t idctx_size,
- crypt_key_type type /* type of child key */)
-{
- unsigned char *pos;
- uint32_t llen = strlen(crypt_keys[type].label);
- /*
- * Compoud the fixed input data for KDF:
- * [i]_2 || Label || 0x00 || Id-Context || [L]_2),
- * NIST SP 800-108, 5.1
- */
- ctx->fid_len =
- sizeof(uint32_t) +
- llen +
- 1 +
- idctx_size +
- sizeof(uint32_t);
-
- ctx->fid = GF_CALLOC(ctx->fid_len, 1, gf_crypt_mt_key);
- if (!ctx->fid)
- return ENOMEM;
- ctx->out_len = round_up(crypt_keys[type].len >> 3,
- PRF_OUTPUT_SIZE);
- ctx->out = GF_CALLOC(ctx->out_len, 1, gf_crypt_mt_key);
- if (!ctx->out) {
- GF_FREE(ctx->fid);
- return ENOMEM;
- }
- ctx->pkey = pkey;
- ctx->pkey_len = pkey_size;
- ctx->ckey_len = crypt_keys[type].len;
-
- pos = ctx->fid;
-
- /* counter will be set up in kderive_rfn() */
- pos += sizeof(uint32_t);
-
- memcpy(pos, crypt_keys[type].label, llen);
- pos += llen;
-
- /* set up zero octet */
- *pos = 0;
- pos += 1;
-
- memcpy(pos, idctx, idctx_size);
- pos += idctx_size;
-
- *((uint32_t *)pos) = htobe32(ctx->ckey_len);
-
- return 0;
-}
-
-static void kderive_update(struct kderive_context *ctx)
-{
- uint32_t i;
-#if (OPENSSL_VERSION_NUMBER < 0x1010002f)
- HMAC_CTX hctx;
-#endif
- HMAC_CTX *phctx = NULL;
- unsigned char *pos = ctx->out;
- uint32_t *p_iter = (uint32_t *)ctx->fid;
- uint32_t num_iters = ctx->out_len / PRF_OUTPUT_SIZE;
-
- check_prf_iters(num_iters);
-
-#if (OPENSSL_VERSION_NUMBER < 0x1010002f)
- HMAC_CTX_init(&hctx);
- phctx = &hctx;
-#else
- phctx = HMAC_CTX_new();
- /* I guess we presume it was successful? */
-#endif
- for (i = 0; i < num_iters; i++) {
- /*
- * update the iteration number in the fid
- */
- *p_iter = htobe32(i);
- HMAC_Init_ex(phctx,
- ctx->pkey, ctx->pkey_len >> 3,
- EVP_sha256(),
- NULL);
- HMAC_Update(phctx, ctx->fid, ctx->fid_len);
- HMAC_Final(phctx, pos, NULL);
-
- pos += PRF_OUTPUT_SIZE;
- }
-#if (OPENSSL_VERSION_NUMBER < 0x1010002f)
- HMAC_CTX_cleanup(phctx);
-#else
- HMAC_CTX_free(phctx);
-#endif
-}
-
-static void kderive_final(struct kderive_context *ctx, unsigned char *child)
-{
- memcpy(child, ctx->out, ctx->ckey_len >> 3);
- GF_FREE(ctx->fid);
- GF_FREE(ctx->out);
- memset(ctx, 0, sizeof(*ctx));
-}
-
-/*
- * derive per-volume key for object ids aithentication
- */
-int32_t get_nmtd_vol_key(struct master_cipher_info *master)
-{
- int32_t ret;
- struct kderive_context ctx;
-
- ret = kderive_init(&ctx,
- master->m_key,
- master_key_size(),
- crypt_fake_oid, sizeof(uuid_t), NMTD_VOL_KEY);
- if (ret)
- return ret;
- kderive_update(&ctx);
- kderive_final(&ctx, master->m_nmtd_key);
- return 0;
-}
-
-/*
- * derive per-link key for aithentication of non-encrypted
- * meta-data (nmtd)
- */
-int32_t get_nmtd_link_key(loc_t *loc,
- struct master_cipher_info *master,
- unsigned char *result)
-{
- int32_t ret;
- struct kderive_context ctx;
-
- ret = kderive_init(&ctx,
- master->m_nmtd_key,
- nmtd_vol_key_size(),
- (const unsigned char *)loc->path,
- strlen(loc->path), NMTD_LINK_KEY);
- if (ret)
- return ret;
- kderive_update(&ctx);
- kderive_final(&ctx, result);
- return 0;
-}
-
-/*
- * derive per-file key for encryption and authentication
- * of encrypted part of metadata (emtd)
- */
-int32_t get_emtd_file_key(struct crypt_inode_info *info,
- struct master_cipher_info *master,
- unsigned char *result)
-{
- int32_t ret;
- struct kderive_context ctx;
-
- ret = kderive_init(&ctx,
- master->m_key,
- master_key_size(),
- info->oid, sizeof(uuid_t), EMTD_FILE_KEY);
- if (ret)
- return ret;
- kderive_update(&ctx);
- kderive_final(&ctx, result);
- return 0;
-}
-
-static int32_t data_key_type_by_size(uint32_t keysize, crypt_key_type *type)
-{
- int32_t ret = 0;
- switch (keysize) {
- case 256:
- *type = DATA_FILE_KEY_256;
- break;
- case 512:
- *type = DATA_FILE_KEY_512;
- break;
- default:
- gf_log("crypt", GF_LOG_ERROR, "Unsupported data key size %d",
- keysize);
- ret = ENOTSUP;
- break;
- }
- return ret;
-}
-
-/*
- * derive per-file key for data encryption
- */
-int32_t get_data_file_key(struct crypt_inode_info *info,
- struct master_cipher_info *master,
- uint32_t keysize,
- unsigned char *key)
-{
- int32_t ret;
- struct kderive_context ctx;
- crypt_key_type type;
-
- ret = data_key_type_by_size(keysize, &type);
- if (ret)
- return ret;
- ret = kderive_init(&ctx,
- master->m_key,
- master_key_size(),
- info->oid, sizeof(uuid_t), type);
- if (ret)
- return ret;
- kderive_update(&ctx);
- kderive_final(&ctx, key);
- return 0;
-}
-
-/*
- * NOTE: Don't change existing keys: it will break compatibility;
- */
-struct crypt_key crypt_keys[LAST_KEY_TYPE] = {
- [MASTER_VOL_KEY] =
- { .len = MASTER_VOL_KEY_SIZE << 3,
- .label = "volume-master",
- },
- [NMTD_VOL_KEY] =
- { .len = NMTD_VOL_KEY_SIZE << 3,
- .label = "volume-nmtd-key-generation"
- },
- [NMTD_LINK_KEY] =
- { .len = 128,
- .label = "link-nmtd-authentication"
- },
- [EMTD_FILE_KEY] =
- { .len = 128,
- .label = "file-emtd-encryption-and-auth"
- },
- [DATA_FILE_KEY_256] =
- { .len = 256,
- .label = "file-data-encryption-256"
- },
- [DATA_FILE_KEY_512] =
- { .len = 512,
- .label = "file-data-encryption-512"
- }
-};
-
-/*
- Local variables:
- c-indentation-style: "K&R"
- mode-name: "LC"
- c-basic-offset: 8
- tab-width: 8
- fill-column: 80
- scroll-step: 1
- End:
-*/
diff --git a/xlators/encryption/crypt/src/metadata.c b/xlators/encryption/crypt/src/metadata.c
deleted file mode 100644
index 1364f825a98..00000000000
--- a/xlators/encryption/crypt/src/metadata.c
+++ /dev/null
@@ -1,614 +0,0 @@
-/*
- Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#include "defaults.h"
-#include "crypt-common.h"
-#include "crypt.h"
-#include "metadata.h"
-
-int32_t alloc_format(crypt_local_t *local, size_t size)
-{
- if (size > 0) {
- local->format = GF_CALLOC(1, size, gf_crypt_mt_mtd);
- if (!local->format)
- return ENOMEM;
- }
- local->format_size = size;
- return 0;
-}
-
-int32_t alloc_format_create(crypt_local_t *local)
-{
- return alloc_format(local, new_format_size());
-}
-
-void free_format(crypt_local_t *local)
-{
- GF_FREE(local->format);
-}
-
-/*
- * Check compatibility with extracted metadata
- */
-static int32_t check_file_metadata(struct crypt_inode_info *info)
-{
- struct object_cipher_info *object = &info->cinfo;
-
- if (info->nr_minor != CRYPT_XLATOR_ID) {
- gf_log("crypt", GF_LOG_WARNING,
- "unsupported minor subversion %d", info->nr_minor);
- return EINVAL;
- }
- if (object->o_alg > LAST_CIPHER_ALG) {
- gf_log("crypt", GF_LOG_WARNING,
- "unsupported cipher algorithm %d",
- object->o_alg);
- return EINVAL;
- }
- if (object->o_mode > LAST_CIPHER_MODE) {
- gf_log("crypt", GF_LOG_WARNING,
- "unsupported cipher mode %d",
- object->o_mode);
- return EINVAL;
- }
- if (object->o_block_bits < CRYPT_MIN_BLOCK_BITS ||
- object->o_block_bits > CRYPT_MAX_BLOCK_BITS) {
- gf_log("crypt", GF_LOG_WARNING, "unsupported block bits %d",
- object->o_block_bits);
- return EINVAL;
- }
- /* TBD: check data key size */
- return 0;
-}
-
-static size_t format_size_v1(mtd_op_t op, size_t old_size)
-{
-
- switch (op) {
- case MTD_CREATE:
- return sizeof(struct mtd_format_v1);
- case MTD_OVERWRITE:
- return old_size;
- case MTD_APPEND:
- return old_size + NMTD_8_MAC_SIZE;
- case MTD_CUT:
- if (old_size > sizeof(struct mtd_format_v1))
- return old_size - NMTD_8_MAC_SIZE;
- else
- return 0;
- default:
- gf_log("crypt", GF_LOG_WARNING, "Bad mtd operation");
- return 0;
- }
-}
-
-/*
- * Calculate size of the updated format string.
- * Returned zero means that we don't need to update the format string.
- */
-size_t format_size(mtd_op_t op, size_t old_size)
-{
- size_t versioned;
-
- versioned = mtd_loaders[current_mtd_loader()].format_size(op,
- old_size - sizeof(struct crypt_format));
- if (versioned != 0)
- return versioned + sizeof(struct crypt_format);
- return 0;
-}
-
-/*
- * size of the format string of newly created file (nr_links = 1)
- */
-size_t new_format_size(void)
-{
- return format_size(MTD_CREATE, 0);
-}
-
-/*
- * Calculate per-link MAC by pathname
- */
-static int32_t calc_link_mac_v1(struct mtd_format_v1 *fmt,
- loc_t *loc,
- unsigned char *result,
- struct crypt_inode_info *info,
- struct master_cipher_info *master)
-{
- int32_t ret;
- unsigned char nmtd_link_key[16];
- CMAC_CTX *cctx;
- size_t len;
-
- ret = get_nmtd_link_key(loc, master, nmtd_link_key);
- if (ret) {
- gf_log("crypt", GF_LOG_ERROR, "Can not get nmtd link key");
- return -1;
- }
- cctx = CMAC_CTX_new();
- if (!cctx) {
- gf_log("crypt", GF_LOG_ERROR, "CMAC_CTX_new failed");
- return -1;
- }
- ret = CMAC_Init(cctx, nmtd_link_key, sizeof(nmtd_link_key),
- EVP_aes_128_cbc(), 0);
- if (!ret) {
- gf_log("crypt", GF_LOG_ERROR, "CMAC_Init failed");
- CMAC_CTX_free(cctx);
- return -1;
- }
- ret = CMAC_Update(cctx, get_NMTD_V1(info), SIZE_OF_NMTD_V1);
- if (!ret) {
- gf_log("crypt", GF_LOG_ERROR, "CMAC_Update failed");
- CMAC_CTX_free(cctx);
- return -1;
- }
- ret = CMAC_Final(cctx, result, &len);
- CMAC_CTX_free(cctx);
- if (!ret) {
- gf_log("crypt", GF_LOG_ERROR, "CMAC_Final failed");
- return -1;
- }
- return 0;
-}
-
-/*
- * Create per-link MAC of index @idx by pathname
- */
-static int32_t create_link_mac_v1(struct mtd_format_v1 *fmt,
- uint32_t idx,
- loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master)
-{
- int32_t ret;
- unsigned char *mac;
- unsigned char cmac[16];
-
- mac = get_NMTD_V1_MAC(fmt) + idx * SIZE_OF_NMTD_V1_MAC;
-
- ret = calc_link_mac_v1(fmt, loc, cmac, info, master);
- if (ret)
- return -1;
- memcpy(mac, cmac, SIZE_OF_NMTD_V1_MAC);
- return 0;
-}
-
-static int32_t create_format_v1(unsigned char *wire,
- loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master)
-{
- int32_t ret;
- struct mtd_format_v1 *fmt;
- unsigned char mtd_key[16];
- AES_KEY EMTD_KEY;
- unsigned char nmtd_link_key[16];
- uint32_t ad;
- GCM128_CONTEXT *gctx;
-
- fmt = (struct mtd_format_v1 *)wire;
-
- fmt->minor_id = info->nr_minor;
- fmt->alg_id = AES_CIPHER_ALG;
- fmt->dkey_factor = master->m_dkey_size >> KEY_FACTOR_BITS;
- fmt->block_bits = master->m_block_bits;
- fmt->mode_id = master->m_mode;
- /*
- * retrieve keys for the parts of metadata
- */
- ret = get_emtd_file_key(info, master, mtd_key);
- if (ret)
- return ret;
- ret = get_nmtd_link_key(loc, master, nmtd_link_key);
- if (ret)
- return ret;
-
- AES_set_encrypt_key(mtd_key, sizeof(mtd_key)*8, &EMTD_KEY);
-
- gctx = CRYPTO_gcm128_new(&EMTD_KEY, (block128_f)AES_encrypt);
-
- /* TBD: Check return values */
-
- CRYPTO_gcm128_setiv(gctx, info->oid, sizeof(uuid_t));
-
- ad = htole32(MTD_LOADER_V1);
- ret = CRYPTO_gcm128_aad(gctx, (const unsigned char *)&ad, sizeof(ad));
- if (ret) {
- gf_log("crypt", GF_LOG_ERROR, " CRYPTO_gcm128_aad failed");
- CRYPTO_gcm128_release(gctx);
- return ret;
- }
- ret = CRYPTO_gcm128_encrypt(gctx,
- get_EMTD_V1(fmt),
- get_EMTD_V1(fmt),
- SIZE_OF_EMTD_V1);
- if (ret) {
- gf_log("crypt", GF_LOG_ERROR, " CRYPTO_gcm128_encrypt failed");
- CRYPTO_gcm128_release(gctx);
- return ret;
- }
- /*
- * set MAC of encrypted part of metadata
- */
- CRYPTO_gcm128_tag(gctx, get_EMTD_V1_MAC(fmt), SIZE_OF_EMTD_V1_MAC);
- CRYPTO_gcm128_release(gctx);
- /*
- * set the first MAC of non-encrypted part of metadata
- */
- return create_link_mac_v1(fmt, 0, loc, info, master);
-}
-
-/*
- * Called by fops:
- * ->create();
- * ->link();
- *
- * Pack common and version-specific parts of file's metadata
- * Pre-conditions: @info contains valid object-id.
- */
-int32_t create_format(unsigned char *wire,
- loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master)
-{
- struct crypt_format *fmt = (struct crypt_format *)wire;
-
- fmt->loader_id = current_mtd_loader();
-
- wire += sizeof(struct crypt_format);
- return mtd_loaders[current_mtd_loader()].create_format(wire, loc,
- info, master);
-}
-
-/*
- * Append or overwrite per-link mac of @mac_idx index
- * in accordance with the new pathname
- */
-int32_t appov_link_mac_v1(unsigned char *new,
- unsigned char *old,
- uint32_t old_size,
- int32_t mac_idx,
- loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master,
- crypt_local_t *local)
-{
- memcpy(new, old, old_size);
- return create_link_mac_v1((struct mtd_format_v1 *)new, mac_idx,
- loc, info, master);
-}
-
-/*
- * Cut per-link mac of @mac_idx index
- */
-static int32_t cut_link_mac_v1(unsigned char *new,
- unsigned char *old,
- uint32_t old_size,
- int32_t mac_idx,
- loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master,
- crypt_local_t *local)
-{
- memcpy(new,
- old,
- sizeof(struct mtd_format_v1) + NMTD_8_MAC_SIZE * (mac_idx - 1));
-
- memcpy(new + sizeof(struct mtd_format_v1) + NMTD_8_MAC_SIZE * (mac_idx - 1),
- old + sizeof(struct mtd_format_v1) + NMTD_8_MAC_SIZE * mac_idx,
- old_size - (sizeof(struct mtd_format_v1) + NMTD_8_MAC_SIZE * mac_idx));
- return 0;
-}
-
-int32_t update_format_v1(unsigned char *new,
- unsigned char *old,
- size_t old_len,
- int32_t mac_idx, /* of old name */
- mtd_op_t op,
- loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master,
- crypt_local_t *local)
-{
- switch (op) {
- case MTD_APPEND:
- mac_idx = 1 + (old_len - sizeof(struct mtd_format_v1))/8;
- case MTD_OVERWRITE:
- return appov_link_mac_v1(new, old, old_len, mac_idx,
- loc, info, master, local);
- case MTD_CUT:
- return cut_link_mac_v1(new, old, old_len, mac_idx,
- loc, info, master, local);
- default:
- gf_log("crypt", GF_LOG_ERROR, "Bad mtd operation %d", op);
- return -1;
- }
-}
-
-/*
- * Called by fops:
- *
- * ->link()
- * ->unlink()
- * ->rename()
- *
- */
-int32_t update_format(unsigned char *new,
- unsigned char *old,
- size_t old_len,
- int32_t mac_idx,
- mtd_op_t op,
- loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master,
- crypt_local_t *local)
-{
- if (!new)
- return 0;
- memcpy(new, old, sizeof(struct crypt_format));
-
- old += sizeof(struct crypt_format);
- new += sizeof(struct crypt_format);
- old_len -= sizeof(struct crypt_format);
-
- return mtd_loaders[current_mtd_loader()].update_format(new, old,
- old_len,
- mac_idx, op,
- loc, info,
- master, local);
-}
-
-/*
- * Perform preliminary checks of found metadata
- * Return < 0 on errors;
- * Return number of object-id MACs (>= 1) on success
- */
-int32_t check_format_v1(uint32_t len, unsigned char *wire)
-{
- uint32_t nr_links;
-
- if (len < sizeof(struct mtd_format_v1)) {
- gf_log("crypt", GF_LOG_ERROR,
- "v1-loader: bad metadata size %d", len);
- goto error;
- }
- len -= sizeof(struct mtd_format_v1);
- if (len % sizeof(nmtd_8_mac_t)) {
- gf_log("crypt", GF_LOG_ERROR,
- "v1-loader: bad metadata format");
- goto error;
- }
- nr_links = 1 + len / sizeof(nmtd_8_mac_t);
- if (nr_links > _POSIX_LINK_MAX)
- goto error;
- return nr_links;
- error:
- return EIO;
-}
-
-/*
- * Verify per-link MAC specified by index @idx
- *
- * return:
- * -1 on errors;
- * 0 on failed verification;
- * 1 on successful verification
- */
-static int32_t verify_link_mac_v1(struct mtd_format_v1 *fmt,
- uint32_t idx /* index of the mac to verify */,
- loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master)
-{
- int32_t ret;
- unsigned char *mac;
- unsigned char cmac[16];
-
- mac = get_NMTD_V1_MAC(fmt) + idx * SIZE_OF_NMTD_V1_MAC;
-
- ret = calc_link_mac_v1(fmt, loc, cmac, info, master);
- if (ret)
- return -1;
- if (memcmp(cmac, mac, SIZE_OF_NMTD_V1_MAC))
- return 0;
- return 1;
-}
-
-/*
- * Lookup per-link MAC by pathname.
- *
- * return index of the MAC, if it was found;
- * return < 0 on errors, or if the MAC wasn't found
- */
-static int32_t lookup_link_mac_v1(struct mtd_format_v1 *fmt,
- uint32_t nr_macs,
- loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master)
-{
- int32_t ret;
- uint32_t idx;
-
- for (idx = 0; idx < nr_macs; idx++) {
- ret = verify_link_mac_v1(fmt, idx, loc, info, master);
- if (ret < 0)
- return ret;
- if (ret > 0)
- return idx;
- }
- return -ENOENT;
-}
-
-/*
- * Extract version-specific part of metadata
- */
-static int32_t open_format_v1(unsigned char *wire,
- int32_t len,
- loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master,
- crypt_local_t *local,
- gf_boolean_t load_info)
-{
- int32_t ret;
- int32_t num_nmtd_macs;
- struct mtd_format_v1 *fmt;
- unsigned char mtd_key[16];
- AES_KEY EMTD_KEY;
- GCM128_CONTEXT *gctx;
- uint32_t ad;
- emtd_8_mac_t gmac;
- struct object_cipher_info *object;
-
- num_nmtd_macs = check_format_v1(len, wire);
- if (num_nmtd_macs <= 0)
- return EIO;
-
- ret = lookup_link_mac_v1((struct mtd_format_v1 *)wire,
- num_nmtd_macs, loc, info, master);
- if (ret < 0) {
- gf_log("crypt", GF_LOG_ERROR, "NMTD verification failed");
- return EINVAL;
- }
-
- local->mac_idx = ret;
- if (load_info == _gf_false)
- /* the case of partial open */
- return 0;
-
- fmt = GF_CALLOC(1, len, gf_crypt_mt_mtd);
- if (!fmt)
- return ENOMEM;
- memcpy(fmt, wire, len);
-
- object = &info->cinfo;
-
- ret = get_emtd_file_key(info, master, mtd_key);
- if (ret) {
- gf_log("crypt", GF_LOG_ERROR, "Can not retrieve metadata key");
- goto out;
- }
- /*
- * decrypt encrypted meta-data
- */
- ret = AES_set_encrypt_key(mtd_key, sizeof(mtd_key)*8, &EMTD_KEY);
- if (ret < 0) {
- gf_log("crypt", GF_LOG_ERROR, "Can not set encrypt key");
- ret = EIO;
- goto out;
- }
- gctx = CRYPTO_gcm128_new(&EMTD_KEY, (block128_f)AES_encrypt);
- if (!gctx) {
- gf_log("crypt", GF_LOG_ERROR, "Can not alloc gcm context");
- ret = ENOMEM;
- goto out;
- }
- CRYPTO_gcm128_setiv(gctx, info->oid, sizeof(uuid_t));
-
- ad = htole32(MTD_LOADER_V1);
- ret = CRYPTO_gcm128_aad(gctx, (const unsigned char *)&ad, sizeof(ad));
- if (ret) {
- gf_log("crypt", GF_LOG_ERROR, " CRYPTO_gcm128_aad failed");
- CRYPTO_gcm128_release(gctx);
- ret = EIO;
- goto out;
- }
- ret = CRYPTO_gcm128_decrypt(gctx,
- get_EMTD_V1(fmt),
- get_EMTD_V1(fmt),
- SIZE_OF_EMTD_V1);
- if (ret) {
- gf_log("crypt", GF_LOG_ERROR, " CRYPTO_gcm128_decrypt failed");
- CRYPTO_gcm128_release(gctx);
- ret = EIO;
- goto out;
- }
- /*
- * verify metadata
- */
- CRYPTO_gcm128_tag(gctx, gmac, sizeof(gmac));
- CRYPTO_gcm128_release(gctx);
- if (memcmp(gmac, get_EMTD_V1_MAC(fmt), SIZE_OF_EMTD_V1_MAC)) {
- gf_log("crypt", GF_LOG_ERROR, "EMTD verification failed");
- ret = EINVAL;
- goto out;
- }
- /*
- * load verified metadata to the private part of inode
- */
- info->nr_minor = fmt->minor_id;
-
- object->o_alg = fmt->alg_id;
- object->o_dkey_size = fmt->dkey_factor << KEY_FACTOR_BITS;
- object->o_block_bits = fmt->block_bits;
- object->o_mode = fmt->mode_id;
-
- ret = check_file_metadata(info);
- out:
- GF_FREE(fmt);
- return ret;
-}
-
-/*
- * perform metadata authentication against @loc->path;
- * extract crypt-specific attribtes and populate @info
- * with them (optional)
- */
-int32_t open_format(unsigned char *str,
- int32_t len,
- loc_t *loc,
- struct crypt_inode_info *info,
- struct master_cipher_info *master,
- crypt_local_t *local,
- gf_boolean_t load_info)
-{
- struct crypt_format *fmt;
- if (len < sizeof(*fmt)) {
- gf_log("crypt", GF_LOG_ERROR, "Bad core format");
- return EIO;
- }
- fmt = (struct crypt_format *)str;
-
- if (fmt->loader_id >= LAST_MTD_LOADER) {
- gf_log("crypt", GF_LOG_ERROR,
- "Unsupported loader id %d", fmt->loader_id);
- return EINVAL;
- }
- str += sizeof(*fmt);
- len -= sizeof(*fmt);
-
- return mtd_loaders[fmt->loader_id].open_format(str,
- len,
- loc,
- info,
- master,
- local,
- load_info);
-}
-
-struct crypt_mtd_loader mtd_loaders [LAST_MTD_LOADER] = {
- [MTD_LOADER_V1] =
- {.format_size = format_size_v1,
- .create_format = create_format_v1,
- .open_format = open_format_v1,
- .update_format = update_format_v1
- }
-};
-
-/*
- Local variables:
- c-indentation-style: "K&R"
- mode-name: "LC"
- c-basic-offset: 8
- tab-width: 8
- fill-column: 80
- scroll-step: 1
- End:
-*/
diff --git a/xlators/encryption/crypt/src/metadata.h b/xlators/encryption/crypt/src/metadata.h
deleted file mode 100644
index b67ae25b58c..00000000000
--- a/xlators/encryption/crypt/src/metadata.h
+++ /dev/null
@@ -1,74 +0,0 @@
-/*
- Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#ifndef __METADATA_H__
-#define __METADATA_H__
-
-#define NMTD_8_MAC_SIZE (8)
-#define EMTD_8_MAC_SIZE (8)
-
-typedef uint8_t nmtd_8_mac_t[NMTD_8_MAC_SIZE];
-typedef uint8_t emtd_8_mac_t[EMTD_8_MAC_SIZE] ;
-
-/*
- * Version "v1" of file's metadata.
- * Metadata of this version has 4 components:
- *
- * 1) EMTD (Encrypted part of MeTaData);
- * 2) NMTD (Non-encrypted part of MeTaData);
- * 3) EMTD_MAC; (EMTD Message Authentication Code);
- * 4) Array of per-link NMTD MACs (for every (hard)link it includes
- * exactly one MAC)
- */
-struct mtd_format_v1 {
- /* EMTD, encrypted part of meta-data */
- uint8_t alg_id; /* cipher algorithm id (only AES for now) */
- uint8_t mode_id; /* cipher mode id; (only XTS for now) */
- uint8_t block_bits; /* encoded block size */
- uint8_t minor_id; /* client translator id */
- uint8_t dkey_factor; /* encoded size of the data key */
- /* MACs */
- emtd_8_mac_t gmac; /* MAC of the encrypted meta-data, 8 bytes */
- nmtd_8_mac_t omac; /* per-link MACs of the non-encrypted
- * meta-data: at least one such MAC is always
- * present */
-} __attribute__((packed));
-
-/*
- * NMTD, the non-encrypted part of metadata of version "v1"
- * is file's gfid, which is generated on trusted machines.
- */
-#define SIZE_OF_NMTD_V1 (sizeof(uuid_t))
-#define SIZE_OF_EMTD_V1 (offsetof(struct mtd_format_v1, gmac) - \
- offsetof(struct mtd_format_v1, alg_id))
-#define SIZE_OF_NMTD_V1_MAC (NMTD_8_MAC_SIZE)
-#define SIZE_OF_EMTD_V1_MAC (EMTD_8_MAC_SIZE)
-
-static inline unsigned char *get_EMTD_V1(struct mtd_format_v1 *format)
-{
- return &format->alg_id;
-}
-
-static inline unsigned char *get_NMTD_V1(struct crypt_inode_info *info)
-{
- return info->oid;
-}
-
-static inline unsigned char *get_EMTD_V1_MAC(struct mtd_format_v1 *format)
-{
- return format->gmac;
-}
-
-static inline unsigned char *get_NMTD_V1_MAC(struct mtd_format_v1 *format)
-{
- return format->omac;
-}
-
-#endif /* __METADATA_H__ */
diff --git a/xlators/encryption/rot-13/src/rot-13.c b/xlators/encryption/rot-13/src/rot-13.c
deleted file mode 100644
index 6ec1b47c87b..00000000000
--- a/xlators/encryption/rot-13/src/rot-13.c
+++ /dev/null
@@ -1,196 +0,0 @@
-/*
- Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-#include <ctype.h>
-#include <sys/uio.h>
-
-#include "glusterfs.h"
-#include "xlator.h"
-#include "logging.h"
-
-#include "rot-13.h"
-
-/*
- * This is a rot13 ``encryption'' xlator. It rot13's data when
- * writing to disk and rot13's it back when reading it.
- * This xlator is meant as an example, NOT FOR PRODUCTION
- * USE ;) (hence no error-checking)
- */
-
-void
-rot13 (char *buf, int len)
-{
- int i;
- for (i = 0; i < len; i++) {
- if (buf[i] >= 'a' && buf[i] <= 'z')
- buf[i] = 'a' + ((buf[i] - 'a' + 13) % 26);
- else if (buf[i] >= 'A' && buf[i] <= 'Z')
- buf[i] = 'A' + ((buf[i] - 'A' + 13) % 26);
- }
-}
-
-void
-rot13_iovec (struct iovec *vector, int count)
-{
- int i;
- for (i = 0; i < count; i++) {
- rot13 (vector[i].iov_base, vector[i].iov_len);
- }
-}
-
-int32_t
-rot13_readv_cbk (call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iovec *vector,
- int32_t count,
- struct iatt *stbuf,
- struct iobref *iobref, dict_t *xdata)
-{
- rot_13_private_t *priv = (rot_13_private_t *)this->private;
-
- if (priv->decrypt_read)
- rot13_iovec (vector, count);
-
- STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector, count,
- stbuf, iobref, xdata);
- return 0;
-}
-
-int32_t
-rot13_readv (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- size_t size,
- off_t offset, uint32_t flags, dict_t *xdata)
-{
- STACK_WIND (frame,
- rot13_readv_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->readv,
- fd, size, offset, flags, xdata);
- return 0;
-}
-
-int32_t
-rot13_writev_cbk (call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
-{
- STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
- return 0;
-}
-
-int32_t
-rot13_writev (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- struct iovec *vector,
- int32_t count,
- off_t offset, uint32_t flags,
- struct iobref *iobref, dict_t *xdata)
-{
- rot_13_private_t *priv = (rot_13_private_t *)this->private;
- if (priv->encrypt_write)
- rot13_iovec (vector, count);
-
- STACK_WIND (frame,
- rot13_writev_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->writev,
- fd, vector, count, offset, flags,
- iobref, xdata);
- return 0;
-}
-
-int32_t
-init (xlator_t *this)
-{
- data_t *data = NULL;
- rot_13_private_t *priv = NULL;
-
- if (!this->children || this->children->next) {
- gf_log ("rot13", GF_LOG_ERROR,
- "FATAL: rot13 should have exactly one child");
- return -1;
- }
-
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "dangling volume. check volfile ");
- }
-
- priv = GF_CALLOC (sizeof (rot_13_private_t), 1, 0);
- if (!priv)
- return -1;
-
- priv->decrypt_read = 1;
- priv->encrypt_write = 1;
-
- data = dict_get (this->options, "encrypt-write");
- if (data) {
- if (gf_string2boolean (data->data, &priv->encrypt_write) == -1) {
- gf_log (this->name, GF_LOG_ERROR,
- "encrypt-write takes only boolean options");
- GF_FREE (priv);
- return -1;
- }
- }
-
- data = dict_get (this->options, "decrypt-read");
- if (data) {
- if (gf_string2boolean (data->data, &priv->decrypt_read) == -1) {
- gf_log (this->name, GF_LOG_ERROR,
- "decrypt-read takes only boolean options");
- GF_FREE (priv);
- return -1;
- }
- }
-
- this->private = priv;
- gf_log ("rot13", GF_LOG_DEBUG, "rot13 xlator loaded");
- return 0;
-}
-
-void
-fini (xlator_t *this)
-{
- rot_13_private_t *priv = this->private;
-
- if (!priv)
- return;
- this->private = NULL;
- GF_FREE (priv);
-
- return;
-}
-
-struct xlator_fops fops = {
- .readv = rot13_readv,
- .writev = rot13_writev
-};
-
-struct xlator_cbks cbks;
-
-struct volume_options options[] = {
- { .key = {"encrypt-write"},
- .type = GF_OPTION_TYPE_BOOL
- },
- { .key = {"decrypt-read"},
- .type = GF_OPTION_TYPE_BOOL
- },
- { .key = {NULL} },
-};
diff --git a/xlators/features/Makefile.am b/xlators/features/Makefile.am
index 81fbe856e78..c57897f11ea 100644
--- a/xlators/features/Makefile.am
+++ b/xlators/features/Makefile.am
@@ -1,6 +1,14 @@
-SUBDIRS = locks quota read-only quiesce marker index barrier \
- arbiter compress changelog changetimerecorder \
- gfid-access $(GLUPY_SUBDIR) upcall snapview-client snapview-server \
- trash shard bit-rot leases selinux sdfs namespace cloudsync thin-arbiter \
- utime
+if BUILD_CLOUDSYNC
+ CLOUDSYNC_DIR = cloudsync
+endif
+
+if BUILD_METADISP
+ METADISP_DIR = metadisp
+endif
+
+SUBDIRS = locks quota read-only quiesce marker index barrier arbiter upcall \
+ compress changelog gfid-access snapview-client snapview-server trash \
+ shard bit-rot leases selinux sdfs namespace $(CLOUDSYNC_DIR) thin-arbiter \
+ utime $(METADISP_DIR)
+
CLEANFILES =
diff --git a/xlators/features/arbiter/src/arbiter-mem-types.h b/xlators/features/arbiter/src/arbiter-mem-types.h
index ccf864cdef0..05d18374c46 100644
--- a/xlators/features/arbiter/src/arbiter-mem-types.h
+++ b/xlators/features/arbiter/src/arbiter-mem-types.h
@@ -9,10 +9,10 @@
#ifndef __ARBITER_MEM_TYPES_H__
#define __ARBITER_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
typedef enum gf_arbiter_mem_types_ {
- gf_arbiter_mt_inode_ctx_t = gf_common_mt_end + 1,
- gf_arbiter_mt_end
+ gf_arbiter_mt_inode_ctx_t = gf_common_mt_end + 1,
+ gf_arbiter_mt_end
} gf_arbiter_mem_types_t;
#endif
diff --git a/xlators/features/arbiter/src/arbiter.c b/xlators/features/arbiter/src/arbiter.c
index cf0841653b7..83a97e3354b 100644
--- a/xlators/features/arbiter/src/arbiter.c
+++ b/xlators/features/arbiter/src/arbiter.c
@@ -10,362 +10,371 @@
#include "arbiter.h"
#include "arbiter-mem-types.h"
-#include "glusterfs.h"
-#include "xlator.h"
-#include "logging.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/logging.h>
static arbiter_inode_ctx_t *
-__arbiter_inode_ctx_get (inode_t *inode, xlator_t *this)
+__arbiter_inode_ctx_get(inode_t *inode, xlator_t *this)
{
-
- arbiter_inode_ctx_t *ctx = NULL;
- int ret = 0;
- uint64_t ctx_addr = 0;
-
- ret = __inode_ctx_get (inode, this, &ctx_addr);
- if (ret == 0) {
- ctx = (arbiter_inode_ctx_t *) (long) ctx_addr;
- goto out;
- }
-
- ctx = GF_CALLOC (1, sizeof (*ctx), gf_arbiter_mt_inode_ctx_t);
- if (!ctx)
- goto out;
-
- ret = __inode_ctx_put (inode, this, (uint64_t)ctx);
- if (ret) {
- GF_FREE (ctx);
- ctx = NULL;
- gf_log_callingfn (this->name, GF_LOG_ERROR, "failed to "
- "set the inode ctx (%s)",
- uuid_utoa (inode->gfid));
- }
+ arbiter_inode_ctx_t *ctx = NULL;
+ int ret = 0;
+ uint64_t ctx_addr = 0;
+
+ ret = __inode_ctx_get(inode, this, &ctx_addr);
+ if (ret == 0) {
+ ctx = (arbiter_inode_ctx_t *)(long)ctx_addr;
+ goto out;
+ }
+
+ ctx = GF_CALLOC(1, sizeof(*ctx), gf_arbiter_mt_inode_ctx_t);
+ if (!ctx)
+ goto out;
+
+ ret = __inode_ctx_put(inode, this, (uint64_t)(uintptr_t)ctx);
+ if (ret) {
+ GF_FREE(ctx);
+ ctx = NULL;
+ gf_log_callingfn(this->name, GF_LOG_ERROR,
+ "failed to "
+ "set the inode ctx (%s)",
+ uuid_utoa(inode->gfid));
+ }
out:
- return ctx;
+ return ctx;
}
static arbiter_inode_ctx_t *
-arbiter_inode_ctx_get (inode_t *inode, xlator_t *this)
+arbiter_inode_ctx_get(inode_t *inode, xlator_t *this)
{
- arbiter_inode_ctx_t *ctx = NULL;
-
- LOCK(&inode->lock);
- {
- ctx = __arbiter_inode_ctx_get (inode, this);
- }
- UNLOCK(&inode->lock);
- return ctx;
+ arbiter_inode_ctx_t *ctx = NULL;
+
+ LOCK(&inode->lock);
+ {
+ ctx = __arbiter_inode_ctx_get(inode, this);
+ }
+ UNLOCK(&inode->lock);
+ return ctx;
}
int32_t
-arbiter_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+arbiter_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata, struct iatt *postparent)
{
- arbiter_inode_ctx_t *ctx = NULL;
-
- if (op_ret != 0)
- goto unwind;
- ctx = arbiter_inode_ctx_get (inode, this);
- if (!ctx) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
- memcpy (&ctx->iattbuf, buf, sizeof (ctx->iattbuf));
+ arbiter_inode_ctx_t *ctx = NULL;
+
+ if (op_ret != 0)
+ goto unwind;
+ ctx = arbiter_inode_ctx_get(inode, this);
+ if (!ctx) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ memcpy(&ctx->iattbuf, buf, sizeof(ctx->iattbuf));
unwind:
- STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
- xdata, postparent);
- return 0;
+ STACK_UNWIND_STRICT(lookup, frame, op_ret, op_errno, inode, buf, xdata,
+ postparent);
+ return 0;
}
int32_t
-arbiter_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+arbiter_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- STACK_WIND (frame, arbiter_lookup_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, loc, xdata);
- return 0;
+ STACK_WIND(frame, arbiter_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, xdata);
+ return 0;
}
int32_t
-arbiter_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
- dict_t *xdata)
+arbiter_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- arbiter_inode_ctx_t *ctx = NULL;
- struct iatt *buf = NULL;
- int32_t op_ret = 0;
- int32_t op_errno = 0;
-
- ctx = arbiter_inode_ctx_get (loc->inode, this);
- if (!ctx) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
- buf = &ctx->iattbuf;
+ arbiter_inode_ctx_t *ctx = NULL;
+ struct iatt *buf = NULL;
+ int32_t op_ret = 0;
+ int32_t op_errno = 0;
+
+ ctx = arbiter_inode_ctx_get(loc->inode, this);
+ if (!ctx) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ buf = &ctx->iattbuf;
unwind:
- STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, buf, buf, NULL);
- return 0;
+ STACK_UNWIND_STRICT(truncate, frame, op_ret, op_errno, buf, buf, NULL);
+ return 0;
}
int32_t
-arbiter_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- dict_t *xdata)
+arbiter_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- arbiter_inode_ctx_t *ctx = NULL;
- struct iatt *buf = NULL;
- int32_t op_ret = 0;
- int32_t op_errno = 0;
-
- ctx = arbiter_inode_ctx_get (fd->inode, this);
- if (!ctx) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
- buf = &ctx->iattbuf;
+ arbiter_inode_ctx_t *ctx = NULL;
+ struct iatt *buf = NULL;
+ int32_t op_ret = 0;
+ int32_t op_errno = 0;
+
+ ctx = arbiter_inode_ctx_get(fd->inode, this);
+ if (!ctx) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ buf = &ctx->iattbuf;
unwind:
- STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno, buf, buf,
- NULL);
- return 0;
+ STACK_UNWIND_STRICT(ftruncate, frame, op_ret, op_errno, buf, buf, NULL);
+ return 0;
}
-dict_t*
-arbiter_fill_writev_xdata (fd_t *fd, dict_t *xdata, xlator_t *this)
+dict_t *
+arbiter_fill_writev_xdata(fd_t *fd, dict_t *xdata, xlator_t *this)
{
- dict_t *rsp_xdata = NULL;
- int32_t ret = 0;
- int is_append = 1;
-
- if (!fd || !fd->inode || gf_uuid_is_null (fd->inode->gfid)) {
- goto out;
- }
-
- if (!xdata)
- goto out;
-
- rsp_xdata = dict_new();
- if (!rsp_xdata)
- goto out;
-
- if (dict_get (xdata, GLUSTERFS_OPEN_FD_COUNT)) {
- ret = dict_set_uint32 (rsp_xdata, GLUSTERFS_OPEN_FD_COUNT,
- fd->inode->fd_count);
- if (ret < 0) {
- gf_msg_debug (this->name, 0, "Failed to set dict value"
- " for GLUSTERFS_OPEN_FD_COUNT");
- }
+ dict_t *rsp_xdata = NULL;
+ int32_t ret = 0;
+ int is_append = 1;
+
+ if (!fd || !fd->inode || gf_uuid_is_null(fd->inode->gfid)) {
+ goto out;
+ }
+
+ if (!xdata)
+ goto out;
+
+ rsp_xdata = dict_new();
+ if (!rsp_xdata)
+ goto out;
+
+ if (dict_get(xdata, GLUSTERFS_OPEN_FD_COUNT)) {
+ ret = dict_set_uint32(rsp_xdata, GLUSTERFS_OPEN_FD_COUNT,
+ fd->inode->fd_count);
+ if (ret < 0) {
+ gf_msg_debug(this->name, 0,
+ "Failed to set dict value"
+ " for GLUSTERFS_OPEN_FD_COUNT");
}
- if (dict_get (xdata, GLUSTERFS_WRITE_IS_APPEND)) {
- ret = dict_set_uint32 (rsp_xdata, GLUSTERFS_WRITE_IS_APPEND,
- is_append);
- if (ret < 0) {
- gf_msg_debug (this->name, 0, "Failed to set dict value"
- " for GLUSTERFS_WRITE_IS_APPEND");
- }
+ }
+ if (dict_get(xdata, GLUSTERFS_WRITE_IS_APPEND)) {
+ ret = dict_set_uint32(rsp_xdata, GLUSTERFS_WRITE_IS_APPEND, is_append);
+ if (ret < 0) {
+ gf_msg_debug(this->name, 0,
+ "Failed to set dict value"
+ " for GLUSTERFS_WRITE_IS_APPEND");
}
+ }
out:
- return rsp_xdata;
+ return rsp_xdata;
}
int32_t
-arbiter_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count, off_t off, uint32_t flags,
- struct iobref *iobref, dict_t *xdata)
+arbiter_writev(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int32_t count, off_t off, uint32_t flags,
+ struct iobref *iobref, dict_t *xdata)
{
- arbiter_inode_ctx_t *ctx = NULL;
- struct iatt *buf = NULL;
- dict_t *rsp_xdata = NULL;
- int op_ret = 0;
- int op_errno = 0;
-
- ctx = arbiter_inode_ctx_get (fd->inode, this);
- if (!ctx) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
- buf = &ctx->iattbuf;
- op_ret = iov_length (vector, count);
- rsp_xdata = arbiter_fill_writev_xdata (fd, xdata, this);
+ arbiter_inode_ctx_t *ctx = NULL;
+ struct iatt *buf = NULL;
+ dict_t *rsp_xdata = NULL;
+ int op_ret = 0;
+ int op_errno = 0;
+
+ ctx = arbiter_inode_ctx_get(fd->inode, this);
+ if (!ctx) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ buf = &ctx->iattbuf;
+ op_ret = iov_length(vector, count);
+ rsp_xdata = arbiter_fill_writev_xdata(fd, xdata, this);
unwind:
- STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, buf, buf,
- rsp_xdata);
- if (rsp_xdata)
- dict_unref (rsp_xdata);
- return 0;
+ STACK_UNWIND_STRICT(writev, frame, op_ret, op_errno, buf, buf, rsp_xdata);
+ if (rsp_xdata)
+ dict_unref(rsp_xdata);
+ return 0;
}
int32_t
-arbiter_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd,
+arbiter_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd,
int32_t keep_size, off_t offset, size_t len, dict_t *xdata)
{
- arbiter_inode_ctx_t *ctx = NULL;
- struct iatt *buf = NULL;
- int op_ret = 0;
- int op_errno = 0;
-
- ctx = arbiter_inode_ctx_get (fd->inode, this);
- if (!ctx) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
- buf = &ctx->iattbuf;
+ arbiter_inode_ctx_t *ctx = NULL;
+ struct iatt *buf = NULL;
+ int op_ret = 0;
+ int op_errno = 0;
+
+ ctx = arbiter_inode_ctx_get(fd->inode, this);
+ if (!ctx) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ buf = &ctx->iattbuf;
unwind:
- STACK_UNWIND_STRICT(fallocate, frame, op_ret, op_errno, buf, buf, NULL);
- return 0;
+ STACK_UNWIND_STRICT(fallocate, frame, op_ret, op_errno, buf, buf, NULL);
+ return 0;
}
int32_t
-arbiter_discard (call_frame_t *frame, xlator_t *this, fd_t *fd,
- off_t offset, size_t len, dict_t *xdata)
+arbiter_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
{
- arbiter_inode_ctx_t *ctx = NULL;
- struct iatt *buf = NULL;
- int op_ret = 0;
- int op_errno = 0;
-
- ctx = arbiter_inode_ctx_get (fd->inode, this);
- if (!ctx) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
- buf = &ctx->iattbuf;
+ arbiter_inode_ctx_t *ctx = NULL;
+ struct iatt *buf = NULL;
+ int op_ret = 0;
+ int op_errno = 0;
+
+ ctx = arbiter_inode_ctx_get(fd->inode, this);
+ if (!ctx) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ buf = &ctx->iattbuf;
unwind:
- STACK_UNWIND_STRICT(discard, frame, op_ret, op_errno, buf, buf, NULL);
- return 0;
+ STACK_UNWIND_STRICT(discard, frame, op_ret, op_errno, buf, buf, NULL);
+ return 0;
}
int32_t
-arbiter_zerofill (call_frame_t *frame, xlator_t *this, fd_t *fd,
- off_t offset, off_t len, dict_t *xdata)
+arbiter_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ off_t len, dict_t *xdata)
{
- arbiter_inode_ctx_t *ctx = NULL;
- struct iatt *buf = NULL;
- int op_ret = 0;
- int op_errno = 0;
-
- ctx = arbiter_inode_ctx_get (fd->inode, this);
- if (!ctx) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
- buf = &ctx->iattbuf;
+ arbiter_inode_ctx_t *ctx = NULL;
+ struct iatt *buf = NULL;
+ int op_ret = 0;
+ int op_errno = 0;
+
+ ctx = arbiter_inode_ctx_get(fd->inode, this);
+ if (!ctx) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+ buf = &ctx->iattbuf;
unwind:
- STACK_UNWIND_STRICT(zerofill, frame, op_ret, op_errno, buf, buf, NULL);
- return 0;
+ STACK_UNWIND_STRICT(zerofill, frame, op_ret, op_errno, buf, buf, NULL);
+ return 0;
}
static int32_t
-arbiter_readv (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, uint32_t flags, dict_t *xdata)
+arbiter_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- STACK_UNWIND_STRICT (readv, frame, -1, ENOSYS, NULL, 0, NULL, NULL,
- NULL);
- return 0;
+ STACK_UNWIND_STRICT(readv, frame, -1, ENOSYS, NULL, 0, NULL, NULL, NULL);
+ return 0;
}
static int32_t
-arbiter_seek (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- gf_seek_what_t what, dict_t *xdata)
+arbiter_seek(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ gf_seek_what_t what, dict_t *xdata)
{
- STACK_UNWIND_STRICT (seek, frame, -1, ENOSYS, 0, xdata);
- return 0;
+ STACK_UNWIND_STRICT(seek, frame, -1, ENOSYS, 0, xdata);
+ return 0;
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
-
- ret = xlator_mem_acct_init (this, gf_arbiter_mt_end + 1);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR, "Memory accounting "
- "initialization failed.");
- return ret;
+ int ret = -1;
+
+ ret = xlator_mem_acct_init(this, gf_arbiter_mt_end + 1);
+ if (ret)
+ gf_log(this->name, GF_LOG_ERROR,
+ "Memory accounting "
+ "initialization failed.");
+ return ret;
}
int
-reconfigure (xlator_t *this, dict_t *options)
+reconfigure(xlator_t *this, dict_t *options)
{
-
- return 0;
+ return 0;
}
int
-arbiter_forget (xlator_t *this, inode_t *inode)
+arbiter_forget(xlator_t *this, inode_t *inode)
{
- arbiter_inode_ctx_t *ctx = NULL;
- uint64_t ctx_addr = 0;
-
- inode_ctx_del (inode, this, &ctx_addr);
- if (!ctx_addr)
- return 0;
- ctx = (arbiter_inode_ctx_t *) (long) ctx_addr;
- GF_FREE (ctx);
+ arbiter_inode_ctx_t *ctx = NULL;
+ uint64_t ctx_addr = 0;
+
+ inode_ctx_del(inode, this, &ctx_addr);
+ if (!ctx_addr)
return 0;
+ ctx = (arbiter_inode_ctx_t *)(long)ctx_addr;
+ GF_FREE(ctx);
+ return 0;
}
int32_t
-init (xlator_t *this)
+init(xlator_t *this)
{
+ if (!this->children || this->children->next) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "'arbiter' not configured with exactly one child");
+ return -1;
+ }
- if (!this->children || this->children->next) {
- gf_log (this->name, GF_LOG_ERROR,
- "'arbiter' not configured with exactly one child");
- return -1;
- }
-
- if (!this->parents)
- gf_log (this->name, GF_LOG_ERROR,
- "dangling volume. check volfile ");
+ if (!this->parents)
+ gf_log(this->name, GF_LOG_ERROR, "dangling volume. check volfile ");
- return 0;
+ return 0;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- return;
+ return;
}
struct xlator_fops fops = {
- .lookup = arbiter_lookup,
-
- /* Return success for these inode write FOPS without winding it down to
- * posix; this is needed for AFR write transaction logic to work.*/
- .truncate = arbiter_truncate,
- .writev = arbiter_writev,
- .ftruncate = arbiter_ftruncate,
- .fallocate = arbiter_fallocate,
- .discard = arbiter_discard,
- .zerofill = arbiter_zerofill,
-
- /* AFR is not expected to wind these inode read FOPS initiated by the
- * application to the arbiter brick. But in case a bug causes them
- * to be called, we return ENOSYS. */
- .readv = arbiter_readv,
- .seek = arbiter_seek,
-
- /* The following inode read FOPS initiated by the application are not
- * wound by AFR either but internal logic like shd, glfsheal and
- * client side healing in AFR will send them for selfheal/ inode refresh
- * operations etc.,so we need to wind them down to posix:
- *
- * (f)stat, readdir(p), readlink, (f)getxattr.*/
-
- /* All other FOPs not listed here are safe to be wound down to posix.*/
+ .lookup = arbiter_lookup,
+
+ /* Return success for these inode write FOPS without winding it down to
+ * posix; this is needed for AFR write transaction logic to work.*/
+ .truncate = arbiter_truncate,
+ .writev = arbiter_writev,
+ .ftruncate = arbiter_ftruncate,
+ .fallocate = arbiter_fallocate,
+ .discard = arbiter_discard,
+ .zerofill = arbiter_zerofill,
+
+ /* AFR is not expected to wind these inode read FOPS initiated by the
+ * application to the arbiter brick. But in case a bug causes them
+ * to be called, we return ENOSYS. */
+ .readv = arbiter_readv,
+ .seek = arbiter_seek,
+
+ /* The following inode read FOPS initiated by the application are not
+ * wound by AFR either but internal logic like shd, glfsheal and
+ * client side healing in AFR will send them for selfheal/ inode refresh
+ * operations etc.,so we need to wind them down to posix:
+ *
+ * (f)stat, readdir(p), readlink, (f)getxattr.*/
+
+ /* All other FOPs not listed here are safe to be wound down to posix.*/
};
struct xlator_cbks cbks = {
- .forget = arbiter_forget,
+ .forget = arbiter_forget,
};
struct volume_options options[] = {
- { .key = {NULL} },
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "arbiter",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/features/arbiter/src/arbiter.h b/xlators/features/arbiter/src/arbiter.h
index f52b45fad20..546db7b751a 100644
--- a/xlators/features/arbiter/src/arbiter.h
+++ b/xlators/features/arbiter/src/arbiter.h
@@ -11,11 +11,11 @@
#ifndef _ARBITER_H
#define _ARBITER_H
-#include "locking.h"
-#include "common-utils.h"
+#include <glusterfs/locking.h>
+#include <glusterfs/common-utils.h>
typedef struct arbiter_inode_ctx_ {
- struct iatt iattbuf;
+ struct iatt iattbuf;
} arbiter_inode_ctx_t;
#endif /* _ARBITER_H */
diff --git a/xlators/features/barrier/src/barrier-mem-types.h b/xlators/features/barrier/src/barrier-mem-types.h
index 36647a66966..71ed7898d9c 100644
--- a/xlators/features/barrier/src/barrier-mem-types.h
+++ b/xlators/features/barrier/src/barrier-mem-types.h
@@ -11,10 +11,10 @@
#ifndef __BARRIER_MEM_TYPES_H__
#define __BARRIER_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_barrier_mem_types_ {
- gf_barrier_mt_priv_t = gf_common_mt_end + 1,
- gf_barrier_mt_end
+ gf_barrier_mt_priv_t = gf_common_mt_end + 1,
+ gf_barrier_mt_end
};
#endif
diff --git a/xlators/features/barrier/src/barrier.c b/xlators/features/barrier/src/barrier.c
index 4401e3459a9..852bbacb99d 100644
--- a/xlators/features/barrier/src/barrier.c
+++ b/xlators/features/barrier/src/barrier.c
@@ -9,795 +9,801 @@
*/
#include "barrier.h"
-#include "defaults.h"
-#include "call-stub.h"
+#include <glusterfs/defaults.h>
+#include <glusterfs/call-stub.h>
-#include "statedump.h"
+#include <glusterfs/statedump.h>
void
-barrier_local_set_gfid (call_frame_t *frame, uuid_t gfid, xlator_t *this)
-{
- if (gfid) {
- uuid_t *id = GF_MALLOC (sizeof (uuid_t), gf_common_mt_uuid_t);
- if (!id) {
- gf_log (this->name, GF_LOG_WARNING, "Could not set gfid"
- ". gfid will not be dumped in statedump file.");
- return;
- }
- gf_uuid_copy (*id, gfid);
- frame->local = id;
+barrier_local_set_gfid(call_frame_t *frame, uuid_t gfid, xlator_t *this)
+{
+ if (gfid) {
+ uuid_t *id = GF_MALLOC(sizeof(uuid_t), gf_common_mt_uuid_t);
+ if (!id) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "Could not set gfid"
+ ". gfid will not be dumped in statedump file.");
+ return;
}
+ gf_uuid_copy(*id, gfid);
+ frame->local = id;
+ }
}
void
-barrier_local_free_gfid (call_frame_t *frame)
+barrier_local_free_gfid(call_frame_t *frame)
{
- if (frame->local) {
- GF_FREE (frame->local);
- frame->local = NULL;
- }
+ if (frame->local) {
+ GF_FREE(frame->local);
+ frame->local = NULL;
+ }
}
int32_t
-barrier_truncate_cbk_resume (call_frame_t *frame, void *cookie, xlator_t *this,
+barrier_truncate_cbk_resume(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata)
+{
+ barrier_local_free_gfid(frame);
+ STACK_UNWIND_STRICT(truncate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
+}
+
+int32_t
+barrier_ftruncate_cbk_resume(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno,
struct iatt *prebuf, struct iatt *postbuf,
dict_t *xdata)
{
- barrier_local_free_gfid (frame);
- STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
- return 0;
+ barrier_local_free_gfid(frame);
+ STACK_UNWIND_STRICT(ftruncate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
int32_t
-barrier_ftruncate_cbk_resume (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *prebuf, struct iatt *postbuf,
- dict_t *xdata)
-{
- barrier_local_free_gfid (frame);
- STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
- return 0;
+barrier_unlink_cbk_resume(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
+{
+ barrier_local_free_gfid(frame);
+ STACK_UNWIND_STRICT(unlink, frame, op_ret, op_errno, preparent, postparent,
+ xdata);
+ return 0;
}
int32_t
-barrier_unlink_cbk_resume (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
-{
- barrier_local_free_gfid (frame);
- STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno, preparent,
- postparent, xdata);
- return 0;
+barrier_rmdir_cbk_resume(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
+{
+ barrier_local_free_gfid(frame);
+ STACK_UNWIND_STRICT(rmdir, frame, op_ret, op_errno, preparent, postparent,
+ xdata);
+ return 0;
}
int32_t
-barrier_rmdir_cbk_resume (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent,
+barrier_rename_cbk_resume(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ struct iatt *preoldparent, struct iatt *postoldparent,
+ struct iatt *prenewparent, struct iatt *postnewparent,
dict_t *xdata)
{
- barrier_local_free_gfid (frame);
- STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno, preparent,
- postparent, xdata);
- return 0;
+ barrier_local_free_gfid(frame);
+ STACK_UNWIND_STRICT(rename, frame, op_ret, op_errno, buf, preoldparent,
+ postoldparent, prenewparent, postnewparent, xdata);
+ return 0;
}
int32_t
-barrier_rename_cbk_resume (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent,
- dict_t *xdata)
-{
- barrier_local_free_gfid (frame);
- STACK_UNWIND_STRICT (rename, frame, op_ret, op_errno, buf, preoldparent,
- postoldparent, prenewparent, postnewparent, xdata);
- return 0;
+barrier_writev_cbk_resume(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ barrier_local_free_gfid(frame);
+ STACK_UNWIND_STRICT(writev, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
int32_t
-barrier_writev_cbk_resume (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *prebuf, struct iatt *postbuf,
- dict_t *xdata)
-{
- barrier_local_free_gfid (frame);
- STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
- return 0;
+barrier_fsync_cbk_resume(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ barrier_local_free_gfid(frame);
+ STACK_UNWIND_STRICT(fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+ return 0;
}
int32_t
-barrier_fsync_cbk_resume (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+barrier_removexattr_cbk_resume(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ dict_t *xdata)
{
- barrier_local_free_gfid (frame);
- STACK_UNWIND_STRICT (fsync, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
- return 0;
+ barrier_local_free_gfid(frame);
+ STACK_UNWIND_STRICT(removexattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
int32_t
-barrier_removexattr_cbk_resume (call_frame_t *frame, void *cookie,
+barrier_fremovexattr_cbk_resume(call_frame_t *frame, void *cookie,
xlator_t *this, int32_t op_ret,
int32_t op_errno, dict_t *xdata)
{
- barrier_local_free_gfid (frame);
- STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
- return 0;
+ barrier_local_free_gfid(frame);
+ STACK_UNWIND_STRICT(fremovexattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
int32_t
-barrier_fremovexattr_cbk_resume (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
+barrier_writev_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- barrier_local_free_gfid (frame);
- STACK_UNWIND_STRICT (fremovexattr, frame, op_ret, op_errno, xdata);
- return 0;
+ BARRIER_FOP_CBK(writev, out, frame, this, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+out:
+ return 0;
}
int32_t
-barrier_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf,
- dict_t *xdata)
+barrier_fremovexattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- BARRIER_FOP_CBK (writev, out, frame, this, op_ret, op_errno,
- prebuf, postbuf, xdata);
+ BARRIER_FOP_CBK(fremovexattr, out, frame, this, op_ret, op_errno, xdata);
out:
- return 0;
+ return 0;
}
int32_t
-barrier_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+barrier_removexattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- BARRIER_FOP_CBK (fremovexattr, out, frame, this, op_ret, op_errno,
- xdata);
+ BARRIER_FOP_CBK(removexattr, out, frame, this, op_ret, op_errno, xdata);
out:
- return 0;
+ return 0;
}
int32_t
-barrier_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+barrier_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- BARRIER_FOP_CBK (removexattr, out, frame, this, op_ret, op_errno,
- xdata);
+ BARRIER_FOP_CBK(truncate, out, frame, this, op_ret, op_errno, prebuf,
+ postbuf, xdata);
out:
- return 0;
+ return 0;
}
int32_t
-barrier_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+barrier_ftruncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
struct iatt *postbuf, dict_t *xdata)
{
- BARRIER_FOP_CBK (truncate, out, frame, this, op_ret, op_errno, prebuf,
- postbuf, xdata);
+ BARRIER_FOP_CBK(ftruncate, out, frame, this, op_ret, op_errno, prebuf,
+ postbuf, xdata);
out:
- return 0;
+ return 0;
}
int32_t
-barrier_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
-{
- BARRIER_FOP_CBK (ftruncate, out, frame, this, op_ret, op_errno, prebuf,
- postbuf, xdata);
+barrier_rename_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ struct iatt *preoldparent, struct iatt *postoldparent,
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
+{
+ BARRIER_FOP_CBK(rename, out, frame, this, op_ret, op_errno, buf,
+ preoldparent, postoldparent, prenewparent, postnewparent,
+ xdata);
out:
- return 0;
+ return 0;
}
int32_t
-barrier_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent,
- dict_t *xdata)
-{
- BARRIER_FOP_CBK (rename, out, frame, this, op_ret, op_errno, buf,
- preoldparent, postoldparent, prenewparent,
- postnewparent, xdata);
+barrier_rmdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ BARRIER_FOP_CBK(rmdir, out, frame, this, op_ret, op_errno, preparent,
+ postparent, xdata);
out:
- return 0;
+ return 0;
}
int32_t
-barrier_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+barrier_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *preparent,
struct iatt *postparent, dict_t *xdata)
{
- BARRIER_FOP_CBK (rmdir, out, frame, this, op_ret, op_errno, preparent,
- postparent, xdata);
+ BARRIER_FOP_CBK(unlink, out, frame, this, op_ret, op_errno, preparent,
+ postparent, xdata);
out:
- return 0;
+ return 0;
}
int32_t
-barrier_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+barrier_fsync_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- BARRIER_FOP_CBK (unlink, out, frame, this, op_ret, op_errno, preparent,
- postparent, xdata);
+ BARRIER_FOP_CBK(fsync, out, frame, this, op_ret, op_errno, prebuf, postbuf,
+ xdata);
out:
- return 0;
+ return 0;
}
int32_t
-barrier_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+barrier_writev(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int32_t count, off_t off, uint32_t flags,
+ struct iobref *iobref, dict_t *xdata)
{
- BARRIER_FOP_CBK (fsync, out, frame, this, op_ret, op_errno,
- prebuf, postbuf, xdata);
-out:
- return 0;
-}
-
-int32_t
-barrier_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count, off_t off, uint32_t flags,
- struct iobref *iobref, dict_t *xdata)
-{
- if (!((flags | fd->flags) & (O_SYNC | O_DSYNC))) {
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->writev,
- fd, vector, count, off, flags, iobref, xdata);
-
- return 0;
- }
+ if (!((flags | fd->flags) & (O_SYNC | O_DSYNC))) {
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector, count, off,
+ flags, iobref, xdata);
- barrier_local_set_gfid (frame, fd->inode->gfid, this);
- STACK_WIND (frame, barrier_writev_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->writev, fd, vector, count,
- off, flags, iobref, xdata);
return 0;
+ }
+
+ barrier_local_set_gfid(frame, fd->inode->gfid, this);
+ STACK_WIND(frame, barrier_writev_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector, count, off, flags,
+ iobref, xdata);
+ return 0;
}
int32_t
-barrier_fremovexattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
+barrier_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata)
{
- barrier_local_set_gfid (frame, fd->inode->gfid, this);
- STACK_WIND (frame, barrier_fremovexattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fremovexattr,
- fd, name, xdata);
- return 0;
+ barrier_local_set_gfid(frame, fd->inode->gfid, this);
+ STACK_WIND(frame, barrier_fremovexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fremovexattr, fd, name, xdata);
+ return 0;
}
int32_t
-barrier_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+barrier_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- barrier_local_set_gfid (frame, loc->inode->gfid, this);
- STACK_WIND (frame, barrier_removexattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->removexattr,
- loc, name, xdata);
- return 0;
+ barrier_local_set_gfid(frame, loc->inode->gfid, this);
+ STACK_WIND(frame, barrier_removexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
+ return 0;
}
int32_t
-barrier_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc,
- off_t offset, dict_t *xdata)
+barrier_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- barrier_local_set_gfid (frame, loc->inode->gfid, this);
- STACK_WIND (frame, barrier_truncate_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->truncate,
- loc, offset, xdata);
- return 0;
+ barrier_local_set_gfid(frame, loc->inode->gfid, this);
+ STACK_WIND(frame, barrier_truncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
+ return 0;
}
-
int32_t
-barrier_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
- loc_t *newloc, dict_t *xdata)
+barrier_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata)
{
- barrier_local_set_gfid (frame, oldloc->inode->gfid, this);
- STACK_WIND (frame, barrier_rename_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->rename,
- oldloc, newloc, xdata);
- return 0;
+ barrier_local_set_gfid(frame, oldloc->inode->gfid, this);
+ STACK_WIND(frame, barrier_rename_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
+ return 0;
}
int
-barrier_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
- dict_t *xdata)
+barrier_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
{
- barrier_local_set_gfid (frame, loc->inode->gfid, this);
- STACK_WIND (frame, barrier_rmdir_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->rmdir,
- loc, flags, xdata);
- return 0;
+ barrier_local_set_gfid(frame, loc->inode->gfid, this);
+ STACK_WIND(frame, barrier_rmdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rmdir, loc, flags, xdata);
+ return 0;
}
int32_t
-barrier_unlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int xflag, dict_t *xdata)
+barrier_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
{
- barrier_local_set_gfid (frame, loc->inode->gfid, this);
- STACK_WIND (frame, barrier_unlink_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->unlink,
- loc, xflag, xdata);
- return 0;
+ barrier_local_set_gfid(frame, loc->inode->gfid, this);
+ STACK_WIND(frame, barrier_unlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
+ return 0;
}
int32_t
-barrier_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd,
- off_t offset, dict_t *xdata)
+barrier_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- barrier_local_set_gfid (frame, fd->inode->gfid, this);
- STACK_WIND (frame, barrier_ftruncate_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->ftruncate,
- fd, offset, xdata);
- return 0;
+ barrier_local_set_gfid(frame, fd->inode->gfid, this);
+ STACK_WIND(frame, barrier_ftruncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+ return 0;
}
int32_t
-barrier_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int32_t flags, dict_t *xdata)
+barrier_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+ dict_t *xdata)
{
- barrier_local_set_gfid (frame, fd->inode->gfid, this);
- STACK_WIND (frame, barrier_fsync_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fsync,
- fd, flags, xdata);
- return 0;
+ barrier_local_set_gfid(frame, fd->inode->gfid, this);
+ STACK_WIND(frame, barrier_fsync_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsync, fd, flags, xdata);
+ return 0;
}
call_stub_t *
-__barrier_dequeue (xlator_t *this, struct list_head *queue)
+__barrier_dequeue(xlator_t *this, struct list_head *queue)
{
- call_stub_t *stub = NULL;
- barrier_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ barrier_priv_t *priv = NULL;
- priv = this->private;
- GF_ASSERT (priv);
+ priv = this->private;
+ GF_ASSERT(priv);
- if (list_empty (queue))
- goto out;
+ if (list_empty(queue))
+ goto out;
- stub = list_entry (queue->next, call_stub_t, list);
- list_del_init (&stub->list);
+ stub = list_entry(queue->next, call_stub_t, list);
+ list_del_init(&stub->list);
out:
- return stub;
+ return stub;
}
void
-barrier_dequeue_all (xlator_t *this, struct list_head *queue)
+barrier_dequeue_all(xlator_t *this, struct list_head *queue)
{
- call_stub_t *stub = NULL;
+ call_stub_t *stub = NULL;
- gf_log (this->name, GF_LOG_INFO, "Dequeuing all the barriered fops");
+ gf_log(this->name, GF_LOG_INFO, "Dequeuing all the barriered fops");
- /* TODO: Start the below task in a new thread */
- while ((stub = __barrier_dequeue (this, queue)))
- call_resume (stub);
+ /* TODO: Start the below task in a new thread */
+ while ((stub = __barrier_dequeue(this, queue)))
+ call_resume(stub);
- gf_log (this->name, GF_LOG_INFO, "Dequeuing the barriered fops is "
- "finished");
- return;
+ gf_log(this->name, GF_LOG_INFO,
+ "Dequeuing the barriered fops is "
+ "finished");
+ return;
}
void
-barrier_timeout (void *data)
+barrier_timeout(void *data)
{
- xlator_t *this = NULL;
- barrier_priv_t *priv = NULL;
- struct list_head queue = {0,};
+ xlator_t *this = NULL;
+ barrier_priv_t *priv = NULL;
+ struct list_head queue = {
+ 0,
+ };
- this = data;
- THIS = this;
- priv = this->private;
+ this = data;
+ THIS = this;
+ priv = this->private;
- INIT_LIST_HEAD (&queue);
+ INIT_LIST_HEAD(&queue);
- gf_log (this->name, GF_LOG_CRITICAL, "Disabling barrier because of "
- "the barrier timeout.");
+ gf_log(this->name, GF_LOG_CRITICAL,
+ "Disabling barrier because of "
+ "the barrier timeout.");
- LOCK (&priv->lock);
- {
- __barrier_disable (this, &queue);
- }
- UNLOCK (&priv->lock);
+ LOCK(&priv->lock);
+ {
+ __barrier_disable(this, &queue);
+ }
+ UNLOCK(&priv->lock);
- barrier_dequeue_all (this, &queue);
+ barrier_dequeue_all(this, &queue);
- return;
+ return;
}
void
-__barrier_enqueue (xlator_t *this, call_stub_t *stub)
+__barrier_enqueue(xlator_t *this, call_stub_t *stub)
{
- barrier_priv_t *priv = NULL;
+ barrier_priv_t *priv = NULL;
- priv = this->private;
- GF_ASSERT (priv);
+ priv = this->private;
+ GF_ASSERT(priv);
- list_add_tail (&stub->list, &priv->queue);
- priv->queue_size++;
+ list_add_tail(&stub->list, &priv->queue);
+ priv->queue_size++;
- return;
+ return;
}
void
-__barrier_disable (xlator_t *this, struct list_head *queue)
+__barrier_disable(xlator_t *this, struct list_head *queue)
{
- GF_UNUSED int ret = 0;
- barrier_priv_t *priv = NULL;
+ GF_UNUSED int ret = 0;
+ barrier_priv_t *priv = NULL;
- priv = this->private;
- GF_ASSERT (priv);
+ priv = this->private;
+ GF_ASSERT(priv);
- if (priv->timer) {
- ret = gf_timer_call_cancel (this->ctx, priv->timer);
- priv->timer = NULL;
- }
+ if (priv->timer) {
+ ret = gf_timer_call_cancel(this->ctx, priv->timer);
+ priv->timer = NULL;
+ }
- list_splice_init (&priv->queue, queue);
- priv->queue_size = 0;
- priv->barrier_enabled = _gf_false;
+ list_splice_init(&priv->queue, queue);
+ priv->queue_size = 0;
+ priv->barrier_enabled = _gf_false;
}
int
-__barrier_enable (xlator_t *this, barrier_priv_t *priv)
+__barrier_enable(xlator_t *this, barrier_priv_t *priv)
{
- int ret = -1;
+ int ret = -1;
- priv->timer = gf_timer_call_after (this->ctx, priv->timeout,
- barrier_timeout, (void *) this);
- if (!priv->timer) {
- gf_log (this->name, GF_LOG_CRITICAL, "Couldn't add barrier "
- "timeout event.");
- goto out;
- }
+ priv->timer = gf_timer_call_after(this->ctx, priv->timeout, barrier_timeout,
+ (void *)this);
+ if (!priv->timer) {
+ gf_log(this->name, GF_LOG_CRITICAL,
+ "Couldn't add barrier "
+ "timeout event.");
+ goto out;
+ }
- priv->barrier_enabled = _gf_true;
- ret = 0;
+ priv->barrier_enabled = _gf_true;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-notify (xlator_t *this, int event, void *data, ...)
-{
- barrier_priv_t *priv = NULL;
- dict_t *dict = NULL;
- gf_boolean_t past = _gf_false;
- int ret = -1;
- int barrier_enabled = _gf_false;
- struct list_head queue = {0,};
-
- priv = this->private;
- GF_ASSERT (priv);
- INIT_LIST_HEAD (&queue);
-
- switch (event) {
- case GF_EVENT_TRANSLATOR_OP:
- {
- dict = data;
- barrier_enabled = dict_get_str_boolean (dict, "barrier", -1);
-
- if (barrier_enabled == -1) {
- gf_log (this->name, GF_LOG_ERROR, "Could not fetch "
- " barrier key from the dictionary.");
- goto out;
- }
-
- LOCK (&priv->lock);
- {
- past = priv->barrier_enabled;
-
- switch (past) {
- case _gf_false:
- if (barrier_enabled) {
- ret = __barrier_enable (this,priv);
- if (ret)
- goto unlock;
- } else {
- gf_log (this->name, GF_LOG_ERROR,
- "Already disabled.");
- goto unlock;
- }
- break;
-
- case _gf_true:
- if (!barrier_enabled) {
- __barrier_disable(this, &queue);
- } else {
- gf_log (this->name, GF_LOG_ERROR,
- "Already enabled");
- goto unlock;
- }
- break;
- }
+notify(xlator_t *this, int event, void *data, ...)
+{
+ barrier_priv_t *priv = this->private;
+ dict_t *dict = NULL;
+ int ret = -1;
+ int barrier_enabled = _gf_false;
+ struct list_head queue = {
+ 0,
+ };
+
+ GF_ASSERT(priv);
+ INIT_LIST_HEAD(&queue);
+
+ switch (event) {
+ case GF_EVENT_TRANSLATOR_OP: {
+ dict = data;
+ barrier_enabled = dict_get_str_boolean(dict, "barrier", -1);
+
+ if (barrier_enabled == -1) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Could not fetch "
+ " barrier key from the dictionary.");
+ goto out;
+ }
+
+ LOCK(&priv->lock);
+ {
+ if (!priv->barrier_enabled) {
+ if (barrier_enabled) {
+ ret = __barrier_enable(this, priv);
+ } else {
+ UNLOCK(&priv->lock);
+ gf_log(this->name, GF_LOG_ERROR, "Already disabled.");
+ goto post_unlock;
+ }
+ } else {
+ if (!barrier_enabled) {
+ __barrier_disable(this, &queue);
ret = 0;
+ } else {
+ UNLOCK(&priv->lock);
+ gf_log(this->name, GF_LOG_ERROR, "Already enabled");
+ goto post_unlock;
+ }
}
-unlock:
- UNLOCK (&priv->lock);
+ }
+ UNLOCK(&priv->lock);
+ post_unlock:
+ if (!list_empty(&queue))
+ barrier_dequeue_all(this, &queue);
- if (!list_empty (&queue))
- barrier_dequeue_all (this, &queue);
-
- break;
- }
- default:
- {
- default_notify (this, event, data);
- ret = 0;
- goto out;
+ break;
}
+ default: {
+ default_notify(this, event, data);
+ ret = 0;
+ goto out;
}
+ }
out:
- return ret;
+ return ret;
}
int
-reconfigure (xlator_t *this, dict_t *options)
-{
- barrier_priv_t *priv = NULL;
- gf_boolean_t past = _gf_false;
- int ret = -1;
- gf_boolean_t barrier_enabled = _gf_false;
- uint32_t timeout = {0,};
- struct list_head queue = {0,};
-
- priv = this->private;
- GF_ASSERT (priv);
-
- GF_OPTION_RECONF ("barrier", barrier_enabled, options, bool, out);
- GF_OPTION_RECONF ("barrier-timeout", timeout, options, time, out);
-
- INIT_LIST_HEAD (&queue);
-
- LOCK (&priv->lock);
- {
- past = priv->barrier_enabled;
-
- switch (past) {
- case _gf_false:
- if (barrier_enabled) {
- ret = __barrier_enable (this, priv);
- if (ret) {
- goto unlock;
- }
- }
- break;
-
- case _gf_true:
- if (!barrier_enabled) {
- __barrier_disable (this, &queue);
-
- }
- break;
+reconfigure(xlator_t *this, dict_t *options)
+{
+ barrier_priv_t *priv = NULL;
+ int ret = -1;
+ gf_boolean_t barrier_enabled = _gf_false;
+ uint32_t timeout = {
+ 0,
+ };
+ struct list_head queue = {
+ 0,
+ };
+
+ priv = this->private;
+ GF_ASSERT(priv);
+
+ GF_OPTION_RECONF("barrier", barrier_enabled, options, bool, out);
+ GF_OPTION_RECONF("barrier-timeout", timeout, options, time, out);
+
+ INIT_LIST_HEAD(&queue);
+
+ LOCK(&priv->lock);
+ {
+ if (!priv->barrier_enabled) {
+ if (barrier_enabled) {
+ ret = __barrier_enable(this, priv);
+ if (ret) {
+ goto unlock;
}
- priv->timeout.tv_sec = timeout;
- ret = 0;
+ }
+ } else {
+ if (!barrier_enabled) {
+ __barrier_disable(this, &queue);
+ }
}
+ priv->timeout.tv_sec = timeout;
+ ret = 0;
+ }
unlock:
- UNLOCK (&priv->lock);
+ UNLOCK(&priv->lock);
- if (!list_empty (&queue))
- barrier_dequeue_all (this, &queue);
+ if (!list_empty(&queue))
+ barrier_dequeue_all(this, &queue);
out:
- return ret;
+ return ret;
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- ret = xlator_mem_acct_init (this, gf_barrier_mt_end + 1);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR, "Memory accounting "
- "initialization failed.");
+ ret = xlator_mem_acct_init(this, gf_barrier_mt_end + 1);
+ if (ret)
+ gf_log(this->name, GF_LOG_ERROR,
+ "Memory accounting "
+ "initialization failed.");
- return ret;
+ return ret;
}
int
-init (xlator_t *this)
+init(xlator_t *this)
{
- int ret = -1;
- barrier_priv_t *priv = NULL;
- uint32_t timeout = {0,};
+ int ret = -1;
+ barrier_priv_t *priv = NULL;
+ uint32_t timeout = {
+ 0,
+ };
- if (!this->children || this->children->next) {
- gf_log (this->name, GF_LOG_ERROR,
- "'barrier' not configured with exactly one child");
- goto out;
- }
+ if (!this->children || this->children->next) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "'barrier' not configured with exactly one child");
+ goto out;
+ }
- if (!this->parents)
- gf_log (this->name, GF_LOG_WARNING,
- "dangling volume. check volfile ");
+ if (!this->parents)
+ gf_log(this->name, GF_LOG_WARNING, "dangling volume. check volfile ");
- priv = GF_CALLOC (1, sizeof (*priv), gf_barrier_mt_priv_t);
- if (!priv)
- goto out;
+ priv = GF_CALLOC(1, sizeof(*priv), gf_barrier_mt_priv_t);
+ if (!priv)
+ goto out;
- LOCK_INIT (&priv->lock);
+ LOCK_INIT(&priv->lock);
- GF_OPTION_INIT ("barrier", priv->barrier_enabled, bool, out);
- GF_OPTION_INIT ("barrier-timeout", timeout, time, out);
- priv->timeout.tv_sec = timeout;
+ GF_OPTION_INIT("barrier", priv->barrier_enabled, bool, out);
+ GF_OPTION_INIT("barrier-timeout", timeout, time, out);
+ priv->timeout.tv_sec = timeout;
- INIT_LIST_HEAD (&priv->queue);
+ INIT_LIST_HEAD(&priv->queue);
- if (priv->barrier_enabled) {
- ret = __barrier_enable (this, priv);
- if (ret == -1)
- goto out;
- }
+ if (priv->barrier_enabled) {
+ ret = __barrier_enable(this, priv);
+ if (ret == -1)
+ goto out;
+ }
- this->private = priv;
- ret = 0;
+ this->private = priv;
+ ret = 0;
out:
- return ret;
+ if (ret && priv)
+ GF_FREE(priv);
+
+ return ret;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- barrier_priv_t *priv = NULL;
- struct list_head queue = {0,};
+ barrier_priv_t *priv = NULL;
+ struct list_head queue = {
+ 0,
+ };
- priv = this->private;
- if (!priv)
- goto out;
+ priv = this->private;
+ if (!priv)
+ goto out;
- INIT_LIST_HEAD (&queue);
+ INIT_LIST_HEAD(&queue);
- gf_log (this->name, GF_LOG_INFO, "Disabling barriering and dequeuing "
- "all the queued fops");
- LOCK (&priv->lock);
- {
- __barrier_disable (this, &queue);
- }
- UNLOCK (&priv->lock);
+ gf_log(this->name, GF_LOG_INFO,
+ "Disabling barriering and dequeuing "
+ "all the queued fops");
+ LOCK(&priv->lock);
+ {
+ __barrier_disable(this, &queue);
+ }
+ UNLOCK(&priv->lock);
- if (!list_empty (&queue))
- barrier_dequeue_all (this, &queue);
+ if (!list_empty(&queue))
+ barrier_dequeue_all(this, &queue);
- this->private = NULL;
+ this->private = NULL;
- LOCK_DESTROY (&priv->lock);
- GF_FREE (priv);
+ LOCK_DESTROY(&priv->lock);
+ GF_FREE(priv);
out:
- return;
+ return;
}
static void
-barrier_dump_stub (call_stub_t *stub, char *prefix)
+barrier_dump_stub(call_stub_t *stub, char *prefix)
{
- char key[GF_DUMP_MAX_BUF_LEN] = {0,};
+ char key[GF_DUMP_MAX_BUF_LEN] = {
+ 0,
+ };
- gf_proc_dump_build_key (key, prefix, "fop");
- gf_proc_dump_write (key, "%s", gf_fop_list[stub->fop]);
+ gf_proc_dump_build_key(key, prefix, "fop");
+ gf_proc_dump_write(key, "%s", gf_fop_list[stub->fop]);
- if (stub->frame->local) {
- gf_proc_dump_build_key (key, prefix, "gfid");
- gf_proc_dump_write (key, "%s",
- uuid_utoa (*(uuid_t*)(stub->frame->local)));
- }
- if (stub->args.loc.path) {
- gf_proc_dump_build_key (key, prefix, "path");
- gf_proc_dump_write (key, "%s", stub->args.loc.path);
- }
- if (stub->args.loc.name) {
- gf_proc_dump_build_key (key, prefix, "name");
- gf_proc_dump_write (key, "%s", stub->args.loc.name);
- }
+ if (stub->frame->local) {
+ gf_proc_dump_build_key(key, prefix, "gfid");
+ gf_proc_dump_write(key, "%s",
+ uuid_utoa(*(uuid_t *)(stub->frame->local)));
+ }
+ if (stub->args.loc.path) {
+ gf_proc_dump_build_key(key, prefix, "path");
+ gf_proc_dump_write(key, "%s", stub->args.loc.path);
+ }
+ if (stub->args.loc.name) {
+ gf_proc_dump_build_key(key, prefix, "name");
+ gf_proc_dump_write(key, "%s", stub->args.loc.name);
+ }
- return;
+ return;
}
static void
-__barrier_dump_queue (barrier_priv_t *priv)
+__barrier_dump_queue(barrier_priv_t *priv)
{
- call_stub_t *stub = NULL;
- char key[GF_DUMP_MAX_BUF_LEN] = {0,};
- int i = 0;
+ call_stub_t *stub = NULL;
+ char key[GF_DUMP_MAX_BUF_LEN] = {
+ 0,
+ };
+ int i = 0;
- GF_VALIDATE_OR_GOTO ("barrier", priv, out);
+ GF_VALIDATE_OR_GOTO("barrier", priv, out);
- list_for_each_entry (stub, &priv->queue, list) {
- snprintf (key, sizeof (key), "stub.%d", i++);
- gf_proc_dump_add_section (key);
- barrier_dump_stub(stub, key);
- }
+ list_for_each_entry(stub, &priv->queue, list)
+ {
+ snprintf(key, sizeof(key), "stub.%d", i++);
+ gf_proc_dump_add_section("%s", key);
+ barrier_dump_stub(stub, key);
+ }
out:
- return;
+ return;
}
int
-barrier_dump_priv (xlator_t *this)
-{
- int ret = -1;
- char key[GF_DUMP_MAX_BUF_LEN] = {0,};
- barrier_priv_t *priv = NULL;
-
- GF_VALIDATE_OR_GOTO ("barrier", this, out);
-
- priv = this->private;
- if (!priv)
- return 0;
-
- gf_proc_dump_build_key (key, "xlator.features.barrier", "priv");
- gf_proc_dump_add_section (key);
-
- LOCK (&priv->lock);
- {
- gf_proc_dump_build_key (key, "barrier", "enabled");
- gf_proc_dump_write (key, "%d", priv->barrier_enabled);
- gf_proc_dump_build_key (key, "barrier", "timeout");
- gf_proc_dump_write (key, "%"PRId64, priv->timeout.tv_sec);
- if (priv->barrier_enabled) {
- gf_proc_dump_build_key (key, "barrier", "queue_size");
- gf_proc_dump_write (key, "%d", priv->queue_size);
- __barrier_dump_queue (priv);
- }
+barrier_dump_priv(xlator_t *this)
+{
+ int ret = -1;
+ char key[GF_DUMP_MAX_BUF_LEN] = {
+ 0,
+ };
+ barrier_priv_t *priv = NULL;
+
+ GF_VALIDATE_OR_GOTO("barrier", this, out);
+
+ priv = this->private;
+ if (!priv)
+ return 0;
+
+ gf_proc_dump_build_key(key, "xlator.features.barrier", "priv");
+ gf_proc_dump_add_section("%s", key);
+ gf_proc_dump_build_key(key, "barrier", "enabled");
+
+ LOCK(&priv->lock);
+ {
+ gf_proc_dump_write(key, "%d", priv->barrier_enabled);
+ gf_proc_dump_build_key(key, "barrier", "timeout");
+ gf_proc_dump_write(key, "%ld", priv->timeout.tv_sec);
+ if (priv->barrier_enabled) {
+ gf_proc_dump_build_key(key, "barrier", "queue_size");
+ gf_proc_dump_write(key, "%d", priv->queue_size);
+ __barrier_dump_queue(priv);
}
- UNLOCK (&priv->lock);
+ }
+ UNLOCK(&priv->lock);
out:
- return ret;
+ return ret;
}
struct xlator_fops fops = {
- /* Barrier Class fops */
- .rmdir = barrier_rmdir,
- .unlink = barrier_unlink,
- .rename = barrier_rename,
- .removexattr = barrier_removexattr,
- .fremovexattr = barrier_fremovexattr,
- .truncate = barrier_truncate,
- .ftruncate = barrier_ftruncate,
- .fsync = barrier_fsync,
-
- /* Writes with only O_SYNC flag */
- .writev = barrier_writev,
+ /* Barrier Class fops */
+ .rmdir = barrier_rmdir,
+ .unlink = barrier_unlink,
+ .rename = barrier_rename,
+ .removexattr = barrier_removexattr,
+ .fremovexattr = barrier_fremovexattr,
+ .truncate = barrier_truncate,
+ .ftruncate = barrier_ftruncate,
+ .fsync = barrier_fsync,
+
+ /* Writes with only O_SYNC flag */
+ .writev = barrier_writev,
};
struct xlator_dumpops dumpops = {
- .priv = barrier_dump_priv,
+ .priv = barrier_dump_priv,
};
struct xlator_cbks cbks;
struct volume_options options[] = {
- { .key = {"barrier"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "disable",
- .op_version = {GD_OP_VERSION_3_6_0},
- .flags = OPT_FLAG_SETTABLE,
- .description = "When \"enabled\", blocks acknowledgements to application "
- "for file operations such as rmdir, rename, unlink, "
- "removexattr, fremovexattr, truncate, ftruncate, "
- "write (with O_SYNC), fsync. It is turned \"off\" by "
- "default."
- },
- { .key = {"barrier-timeout"},
- .type = GF_OPTION_TYPE_TIME,
- .default_value = BARRIER_TIMEOUT,
- .op_version = {GD_OP_VERSION_3_6_0},
- .flags = OPT_FLAG_SETTABLE,
- .description = "After 'timeout' seconds since the time 'barrier' "
- "option was set to \"on\", acknowledgements to file "
- "operations are no longer blocked and previously "
- "blocked acknowledgements are sent to the application"
- },
- { .key = {NULL} },
+ {.key = {"barrier"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "disable",
+ .op_version = {GD_OP_VERSION_3_6_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .description = "When \"enabled\", blocks acknowledgements to application "
+ "for file operations such as rmdir, rename, unlink, "
+ "removexattr, fremovexattr, truncate, ftruncate, "
+ "write (with O_SYNC), fsync. It is turned \"off\" by "
+ "default."},
+ {.key = {"barrier-timeout"},
+ .type = GF_OPTION_TYPE_TIME,
+ .default_value = BARRIER_TIMEOUT,
+ .op_version = {GD_OP_VERSION_3_6_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .description = "After 'timeout' seconds since the time 'barrier' "
+ "option was set to \"on\", acknowledgements to file "
+ "operations are no longer blocked and previously "
+ "blocked acknowledgements are sent to the application"},
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .notify = notify,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .dumpops = &dumpops,
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "barrier",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/features/barrier/src/barrier.h b/xlators/features/barrier/src/barrier.h
index 0d646f90474..1337f311f7d 100644
--- a/xlators/features/barrier/src/barrier.h
+++ b/xlators/features/barrier/src/barrier.h
@@ -12,71 +12,78 @@
#define __BARRIER_H__
#include "barrier-mem-types.h"
-#include "xlator.h"
-#include "timer.h"
-#include "call-stub.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/timer.h>
+#include <glusterfs/call-stub.h>
-#define BARRIER_FOP_CBK(fop_name, label, frame, this, params ...) \
- do { \
- barrier_priv_t *_priv = NULL; \
- call_stub_t *_stub = NULL; \
- gf_boolean_t _barrier_enabled= _gf_false; \
- struct list_head queue = {0, }; \
- \
- INIT_LIST_HEAD (&queue); \
- \
- _priv = this->private; \
- GF_ASSERT (_priv); \
- \
- LOCK (&_priv->lock); \
- { \
- if (_priv->barrier_enabled) { \
- _barrier_enabled = _priv->barrier_enabled;\
- \
- _stub = fop_##fop_name##_cbk_stub \
- (frame, \
- barrier_##fop_name##_cbk_resume,\
- params); \
- if (!_stub) { \
- __barrier_disable (this, &queue);\
- goto unlock; \
- } \
- \
- __barrier_enqueue (this, _stub); \
- } \
- } \
-unlock: \
- UNLOCK (&_priv->lock); \
- \
- if (_stub) \
- goto label; \
- \
- if (_barrier_enabled && !_stub) { \
- gf_log (this->name, GF_LOG_CRITICAL, \
- "Failed to barrier FOPs, disabling " \
- "barrier. FOP: %s, ERROR: %s", \
- #fop_name, strerror (ENOMEM)); \
- barrier_dequeue_all (this, &queue); \
- } \
- barrier_local_free_gfid (frame); \
- STACK_UNWIND_STRICT (fop_name, frame, params); \
- goto label; \
- } while (0)
+#define BARRIER_FOP_CBK(fop_name, label, frame, this, params...) \
+ do { \
+ barrier_priv_t *_priv = NULL; \
+ call_stub_t *_stub = NULL; \
+ gf_boolean_t _barrier_enabled = _gf_false; \
+ struct list_head queue = { \
+ 0, \
+ }; \
+ \
+ INIT_LIST_HEAD(&queue); \
+ \
+ _priv = this->private; \
+ GF_ASSERT(_priv); \
+ \
+ LOCK(&_priv->lock); \
+ { \
+ if (_priv->barrier_enabled) { \
+ _barrier_enabled = _priv->barrier_enabled; \
+ \
+ _stub = fop_##fop_name##_cbk_stub( \
+ frame, barrier_##fop_name##_cbk_resume, params); \
+ if (!_stub) { \
+ __barrier_disable(this, &queue); \
+ goto unlock; \
+ } \
+ \
+ __barrier_enqueue(this, _stub); \
+ } \
+ } \
+ unlock: \
+ UNLOCK(&_priv->lock); \
+ \
+ if (_stub) \
+ goto label; \
+ \
+ if (_barrier_enabled && !_stub) { \
+ gf_log(this->name, GF_LOG_CRITICAL, \
+ "Failed to barrier FOPs, disabling " \
+ "barrier. FOP: %s, ERROR: %s", \
+ #fop_name, strerror(ENOMEM)); \
+ barrier_dequeue_all(this, &queue); \
+ } \
+ barrier_local_free_gfid(frame); \
+ STACK_UNWIND_STRICT(fop_name, frame, params); \
+ goto label; \
+ } while (0)
typedef struct {
- gf_timer_t *timer;
- gf_boolean_t barrier_enabled;
- gf_lock_t lock;
- struct list_head queue;
- struct timespec timeout;
- uint32_t queue_size;
+ gf_timer_t *timer;
+ gf_lock_t lock;
+ struct list_head queue;
+ struct timespec timeout;
+ uint32_t queue_size;
+ gf_boolean_t barrier_enabled;
+ char _pad[3]; /* manual padding */
} barrier_priv_t;
-int __barrier_enable (xlator_t *this, barrier_priv_t *priv);
-void __barrier_enqueue (xlator_t *this, call_stub_t *stub);
-void __barrier_disable (xlator_t *this, struct list_head *queue);
-void barrier_timeout (void *data);
-void barrier_dequeue_all (xlator_t *this, struct list_head *queue);
-call_stub_t *__barrier_dequeue (xlator_t *this, struct list_head *queue);
+int
+__barrier_enable(xlator_t *this, barrier_priv_t *priv);
+void
+__barrier_enqueue(xlator_t *this, call_stub_t *stub);
+void
+__barrier_disable(xlator_t *this, struct list_head *queue);
+void
+barrier_timeout(void *data);
+void
+barrier_dequeue_all(xlator_t *this, struct list_head *queue);
+call_stub_t *
+__barrier_dequeue(xlator_t *this, struct list_head *queue);
#endif
diff --git a/xlators/features/bit-rot/src/bitd/bit-rot-bitd-messages.h b/xlators/features/bit-rot/src/bitd/bit-rot-bitd-messages.h
index 89a67cc10b1..5bc5103a27c 100644
--- a/xlators/features/bit-rot/src/bitd/bit-rot-bitd-messages.h
+++ b/xlators/features/bit-rot/src/bitd/bit-rot-bitd-messages.h
@@ -11,7 +11,7 @@
#ifndef _BITROT_BITD_MESSAGES_H_
#define _BITROT_BITD_MESSAGES_H_
-#include "glfs-message-id.h"
+#include <glusterfs/glfs-message-id.h>
/* To add new message IDs, append new identifiers at the end of the list.
*
@@ -23,62 +23,79 @@
* glfs-message-id.h.
*/
-GLFS_MSGID(BITROT_BITD,
- BRB_MSG_FD_CREATE_FAILED,
- BRB_MSG_READV_FAILED,
- BRB_MSG_BLOCK_READ_FAILED,
- BRB_MSG_CALC_CHECKSUM_FAILED,
- BRB_MSG_NO_MEMORY,
- BRB_MSG_GET_SIGN_FAILED,
- BRB_MSG_SET_SIGN_FAILED,
- BRB_MSG_OP_FAILED,
- BRB_MSG_READ_AND_SIGN_FAILED,
- BRB_MSG_SIGN_FAILED,
- BRB_MSG_GET_SUBVOL_FAILED,
- BRB_MSG_SET_TIMER_FAILED,
- BRB_MSG_GET_INFO_FAILED,
- BRB_MSG_PATH_FAILED,
- BRB_MSG_MARK_BAD_FILE,
- BRB_MSG_TRIGGER_SIGN,
- BRB_MSG_REGISTER_FAILED,
- BRB_MSG_CRAWLING_START,
- BRB_MSG_SPAWN_FAILED,
- BRB_MSG_INVALID_SUBVOL_CHILD,
- BRB_MSG_SKIP_OBJECT,
- BRB_MSG_NO_CHILD,
- BRB_MSG_CHECKSUM_MISMATCH,
- BRB_MSG_MARK_CORRUPTED,
- BRB_MSG_CRAWLING_FINISH,
- BRB_MSG_CALC_ERROR,
- BRB_MSG_LOOKUP_FAILED,
- BRB_MSG_PARTIAL_VERSION_PRESENCE,
- BRB_MSG_MEM_ACNT_FAILED,
- BRB_MSG_TIMER_WHEEL_UNAVAILABLE,
- BRB_MSG_BITROT_LOADED,
- BRB_MSG_SCALE_DOWN_FAILED,
- BRB_MSG_SCALE_UP_FAILED,
- BRB_MSG_SCALE_DOWN_SCRUBBER,
- BRB_MSG_SCALING_UP_SCRUBBER,
- BRB_MSG_UNKNOWN_THROTTLE,
- BRB_MSG_RATE_LIMIT_INFO,
- BRB_MSG_SCRUB_INFO,
- BRB_MSG_CONNECTED_TO_BRICK,
- BRB_MSG_BRICK_INFO,
- BRB_MSG_SUBVOL_CONNECT_FAILED,
- BRB_MSG_INVALID_SUBVOL,
- BRB_MSG_RESCHEDULE_SCRUBBER_FAILED,
- BRB_MSG_SCRUB_START,
- BRB_MSG_SCRUB_FINISH,
- BRB_MSG_SCRUB_RUNNING,
- BRB_MSG_SCRUB_RESCHEDULED,
- BRB_MSG_SCRUB_TUNABLE,
- BRB_MSG_SCRUB_THREAD_CLEANUP,
- BRB_MSG_SCRUBBER_CLEANED,
- BRB_MSG_GENERIC_SSM_INFO,
- BRB_MSG_ZERO_TIMEOUT_BUG,
- BRB_MSG_BAD_OBJ_READDIR_FAIL,
- BRB_MSG_SSM_FAILED,
- BRB_MSG_SCRUB_WAIT_FAILED
-);
+GLFS_MSGID(BITROT_BITD, BRB_MSG_FD_CREATE_FAILED, BRB_MSG_READV_FAILED,
+ BRB_MSG_BLOCK_READ_FAILED, BRB_MSG_CALC_CHECKSUM_FAILED,
+ BRB_MSG_NO_MEMORY, BRB_MSG_GET_SIGN_FAILED, BRB_MSG_SET_SIGN_FAILED,
+ BRB_MSG_OP_FAILED, BRB_MSG_READ_AND_SIGN_FAILED, BRB_MSG_SIGN_FAILED,
+ BRB_MSG_GET_SUBVOL_FAILED, BRB_MSG_SET_TIMER_FAILED,
+ BRB_MSG_GET_INFO_FAILED, BRB_MSG_PATH_FAILED, BRB_MSG_MARK_BAD_FILE,
+ BRB_MSG_TRIGGER_SIGN, BRB_MSG_REGISTER_FAILED,
+ BRB_MSG_CRAWLING_START, BRB_MSG_SPAWN_FAILED,
+ BRB_MSG_INVALID_SUBVOL_CHILD, BRB_MSG_SKIP_OBJECT, BRB_MSG_NO_CHILD,
+ BRB_MSG_CHECKSUM_MISMATCH, BRB_MSG_MARK_CORRUPTED,
+ BRB_MSG_CRAWLING_FINISH, BRB_MSG_CALC_ERROR, BRB_MSG_LOOKUP_FAILED,
+ BRB_MSG_PARTIAL_VERSION_PRESENCE, BRB_MSG_MEM_ACNT_FAILED,
+ BRB_MSG_TIMER_WHEEL_UNAVAILABLE, BRB_MSG_BITROT_LOADED,
+ BRB_MSG_SCALE_DOWN_FAILED, BRB_MSG_SCALE_UP_FAILED,
+ BRB_MSG_SCALE_DOWN_SCRUBBER, BRB_MSG_SCALING_UP_SCRUBBER,
+ BRB_MSG_UNKNOWN_THROTTLE, BRB_MSG_RATE_LIMIT_INFO,
+ BRB_MSG_SCRUB_INFO, BRB_MSG_CONNECTED_TO_BRICK, BRB_MSG_BRICK_INFO,
+ BRB_MSG_SUBVOL_CONNECT_FAILED, BRB_MSG_INVALID_SUBVOL,
+ BRB_MSG_RESCHEDULE_SCRUBBER_FAILED, BRB_MSG_SCRUB_START,
+ BRB_MSG_SCRUB_FINISH, BRB_MSG_SCRUB_RUNNING,
+ BRB_MSG_SCRUB_RESCHEDULED, BRB_MSG_SCRUB_TUNABLE,
+ BRB_MSG_SCRUB_THREAD_CLEANUP, BRB_MSG_SCRUBBER_CLEANED,
+ BRB_MSG_GENERIC_SSM_INFO, BRB_MSG_ZERO_TIMEOUT_BUG,
+ BRB_MSG_BAD_OBJ_READDIR_FAIL, BRB_MSG_SSM_FAILED,
+ BRB_MSG_SCRUB_WAIT_FAILED, BRB_MSG_TRIGGER_SIGN_FAILED,
+ BRB_MSG_EVENT_UNHANDLED, BRB_MSG_COULD_NOT_SCHEDULE_SCRUB,
+ BRB_MSG_THREAD_CREATION_FAILED, BRB_MSG_MEM_POOL_ALLOC,
+ BRB_MSG_SAVING_HASH_FAILED);
+#define BRB_MSG_FD_CREATE_FAILED_STR "failed to create fd for the inode"
+#define BRB_MSG_READV_FAILED_STR "readv failed"
+#define BRB_MSG_BLOCK_READ_FAILED_STR "reading block failed"
+#define BRB_MSG_NO_MEMORY_STR "failed to allocate memory"
+#define BRB_MSG_CALC_CHECKSUM_FAILED_STR "calculating checksum failed"
+#define BRB_MSG_GET_SIGN_FAILED_STR "failed to get the signature"
+#define BRB_MSG_SET_SIGN_FAILED_STR "signing failed"
+#define BRB_MSG_OP_FAILED_STR "failed on object"
+#define BRB_MSG_TRIGGER_SIGN_FAILED_STR "Could not trigger signing"
+#define BRB_MSG_READ_AND_SIGN_FAILED_STR "reading and signing of object failed"
+#define BRB_MSG_SET_TIMER_FAILED_STR "Failed to allocate object expiry timer"
+#define BRB_MSG_GET_SUBVOL_FAILED_STR \
+ "failed to get the subvolume for the brick"
+#define BRB_MSG_PATH_FAILED_STR "path failed"
+#define BRB_MSG_SKIP_OBJECT_STR "Entry is marked corrupted. skipping"
+#define BRB_MSG_PARTIAL_VERSION_PRESENCE_STR \
+ "PArtial version xattr presence detected, ignoring"
+#define BRB_MSG_TRIGGER_SIGN_STR "Triggering signing"
+#define BRB_MSG_CRAWLING_START_STR \
+ "Crawling brick, scanning for unsigned objects"
+#define BRB_MSG_CRAWLING_FINISH_STR "Completed crawling brick"
+#define BRB_MSG_REGISTER_FAILED_STR "Register to changelog failed"
+#define BRB_MSG_SPAWN_FAILED_STR "failed to spawn"
+#define BRB_MSG_CONNECTED_TO_BRICK_STR "Connected to brick"
+#define BRB_MSG_LOOKUP_FAILED_STR "lookup on root failed"
+#define BRB_MSG_GET_INFO_FAILED_STR "failed to get stub info"
+#define BRB_MSG_SCRUB_THREAD_CLEANUP_STR "Error cleaning up scanner thread"
+#define BRB_MSG_SCRUBBER_CLEANED_STR "clened up scrubber for brick"
+#define BRB_MSG_SUBVOL_CONNECT_FAILED_STR \
+ "callback handler for subvolume failed"
+#define BRB_MSG_MEM_ACNT_FAILED_STR "Memory accounting init failed"
+#define BRB_MSG_EVENT_UNHANDLED_STR "Event unhandled for child"
+#define BRB_MSG_INVALID_SUBVOL_STR "Got event from invalid subvolume"
+#define BRB_MSG_RESCHEDULE_SCRUBBER_FAILED_STR \
+ "on demand scrub schedule failed. Scrubber is not in pending state."
+#define BRB_MSG_COULD_NOT_SCHEDULE_SCRUB_STR \
+ "Could not schedule ondemand scrubbing. Scrubbing will continue " \
+ "according to old frequency."
+#define BRB_MSG_THREAD_CREATION_FAILED_STR "thread creation failed"
+#define BRB_MSG_RATE_LIMIT_INFO_STR "Rate Limit Info"
+#define BRB_MSG_MEM_POOL_ALLOC_STR "failed to allocate mem-pool for timer"
+#define BRB_MSG_NO_CHILD_STR "FATAL: no children"
+#define BRB_MSG_TIMER_WHEEL_UNAVAILABLE_STR "global timer wheel unavailable"
+#define BRB_MSG_BITROT_LOADED_STR "bit-rot xlator loaded"
+#define BRB_MSG_SAVING_HASH_FAILED_STR \
+ "failed to allocate memory for saving hash of the object"
#endif /* !_BITROT_BITD_MESSAGES_H_ */
diff --git a/xlators/features/bit-rot/src/bitd/bit-rot-scrub-status.c b/xlators/features/bit-rot/src/bitd/bit-rot-scrub-status.c
index 2f9aaf31a52..5cef2ffa5e5 100644
--- a/xlators/features/bit-rot/src/bitd/bit-rot-scrub-status.c
+++ b/xlators/features/bit-rot/src/bitd/bit-rot-scrub-status.c
@@ -9,69 +9,70 @@
*/
#include <string.h>
+#include <stdio.h>
#include "bit-rot-scrub-status.h"
void
-br_inc_unsigned_file_count (br_scrub_stats_t *scrub_stat)
+br_inc_unsigned_file_count(br_scrub_stats_t *scrub_stat)
{
- if (!scrub_stat)
- return;
+ if (!scrub_stat)
+ return;
- pthread_mutex_lock (&scrub_stat->lock);
- {
- scrub_stat->unsigned_files++;
- }
- pthread_mutex_unlock (&scrub_stat->lock);
+ pthread_mutex_lock(&scrub_stat->lock);
+ {
+ scrub_stat->unsigned_files++;
+ }
+ pthread_mutex_unlock(&scrub_stat->lock);
}
void
-br_inc_scrubbed_file (br_scrub_stats_t *scrub_stat)
+br_inc_scrubbed_file(br_scrub_stats_t *scrub_stat)
{
- if (!scrub_stat)
- return;
+ if (!scrub_stat)
+ return;
- pthread_mutex_lock (&scrub_stat->lock);
- {
- scrub_stat->scrubbed_files++;
- }
- pthread_mutex_unlock (&scrub_stat->lock);
+ pthread_mutex_lock(&scrub_stat->lock);
+ {
+ scrub_stat->scrubbed_files++;
+ }
+ pthread_mutex_unlock(&scrub_stat->lock);
}
void
-br_update_scrub_start_time (br_scrub_stats_t *scrub_stat, struct timeval *tv)
+br_update_scrub_start_time(br_scrub_stats_t *scrub_stat, time_t time)
{
- if (!scrub_stat)
- return;
+ if (!scrub_stat)
+ return;
- pthread_mutex_lock (&scrub_stat->lock);
- {
- scrub_stat->scrub_start_tv.tv_sec = tv->tv_sec;
- }
- pthread_mutex_unlock (&scrub_stat->lock);
+ pthread_mutex_lock(&scrub_stat->lock);
+ {
+ scrub_stat->scrub_start_time = time;
+ }
+ pthread_mutex_unlock(&scrub_stat->lock);
}
void
-br_update_scrub_finish_time (br_scrub_stats_t *scrub_stat, char *timestr,
- struct timeval *tv)
+br_update_scrub_finish_time(br_scrub_stats_t *scrub_stat, char *timestr,
+ time_t time)
{
- int lst_size = 0;
+ int lst_size = 0;
- if (!scrub_stat)
- return;
+ if (!scrub_stat)
+ return;
- lst_size = sizeof (scrub_stat->last_scrub_time);
- pthread_mutex_lock (&scrub_stat->lock);
- {
- scrub_stat->scrub_end_tv.tv_sec = tv->tv_sec;
+ lst_size = sizeof(scrub_stat->last_scrub_time);
+ if (strlen(timestr) >= lst_size)
+ return;
- scrub_stat->scrub_duration =
- scrub_stat->scrub_end_tv.tv_sec -
- scrub_stat->scrub_start_tv.tv_sec;
+ pthread_mutex_lock(&scrub_stat->lock);
+ {
+ scrub_stat->scrub_end_time = time;
- strncpy (scrub_stat->last_scrub_time, timestr,
- lst_size-1);
- scrub_stat->last_scrub_time[lst_size-1] = '\0';
- }
- pthread_mutex_unlock (&scrub_stat->lock);
+ scrub_stat->scrub_duration = scrub_stat->scrub_end_time -
+ scrub_stat->scrub_start_time;
+
+ snprintf(scrub_stat->last_scrub_time, lst_size, "%s", timestr);
+ }
+ pthread_mutex_unlock(&scrub_stat->lock);
}
diff --git a/xlators/features/bit-rot/src/bitd/bit-rot-scrub-status.h b/xlators/features/bit-rot/src/bitd/bit-rot-scrub-status.h
index 02bd0fab04e..f022aa831eb 100644
--- a/xlators/features/bit-rot/src/bitd/bit-rot-scrub-status.h
+++ b/xlators/features/bit-rot/src/bitd/bit-rot-scrub-status.h
@@ -15,34 +15,36 @@
#include <sys/time.h>
#include <pthread.h>
+#include <glusterfs/common-utils.h>
+
struct br_scrub_stats {
- uint64_t scrubbed_files; /* Total number of scrubbed file */
+ uint64_t scrubbed_files; /* Total number of scrubbed files. */
- uint64_t unsigned_files; /* Total number of unsigned file */
+ uint64_t unsigned_files; /* Total number of unsigned files. */
- uint64_t scrub_duration; /* Duration of last scrub */
+ uint64_t scrub_duration; /* Duration of last scrub. */
- char last_scrub_time[1024]; /*last scrub completion time */
+ char last_scrub_time[GF_TIMESTR_SIZE]; /* Last scrub completion time. */
- struct timeval scrub_start_tv; /* Scrubbing starting time*/
+ time_t scrub_start_time; /* Scrubbing starting time. */
- struct timeval scrub_end_tv; /* Scrubbing finishing time */
+ time_t scrub_end_time; /* Scrubbing finishing time. */
- int8_t scrub_running; /* Scrub running or not */
+ int8_t scrub_running; /* Whether scrub running or not. */
- pthread_mutex_t lock;
+ pthread_mutex_t lock;
};
typedef struct br_scrub_stats br_scrub_stats_t;
void
-br_inc_unsigned_file_count (br_scrub_stats_t *scrub_stat);
+br_inc_unsigned_file_count(br_scrub_stats_t *scrub_stat);
void
-br_inc_scrubbed_file (br_scrub_stats_t *scrub_stat);
+br_inc_scrubbed_file(br_scrub_stats_t *scrub_stat);
void
-br_update_scrub_start_time (br_scrub_stats_t *scrub_stat, struct timeval *tv);
+br_update_scrub_start_time(br_scrub_stats_t *scrub_stat, time_t time);
void
-br_update_scrub_finish_time (br_scrub_stats_t *scrub_stat, char *timestr,
- struct timeval *tv);
+br_update_scrub_finish_time(br_scrub_stats_t *scrub_stat, char *timestr,
+ time_t time);
#endif /* __BIT_ROT_SCRUB_STATUS_H__ */
diff --git a/xlators/features/bit-rot/src/bitd/bit-rot-scrub.c b/xlators/features/bit-rot/src/bitd/bit-rot-scrub.c
index ee996525f76..289dd53f610 100644
--- a/xlators/features/bit-rot/src/bitd/bit-rot-scrub.c
+++ b/xlators/features/bit-rot/src/bitd/bit-rot-scrub.c
@@ -12,32 +12,32 @@
#include <ctype.h>
#include <sys/uio.h>
-#include "glusterfs.h"
-#include "logging.h"
-#include "common-utils.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/common-utils.h>
#include "bit-rot-scrub.h"
#include <pthread.h>
#include "bit-rot-bitd-messages.h"
#include "bit-rot-scrub-status.h"
-#include "events.h"
+#include <glusterfs/events.h>
struct br_scrubbers {
- pthread_t scrubthread;
+ pthread_t scrubthread;
- struct list_head list;
+ struct list_head list;
};
struct br_fsscan_entry {
- void *data;
+ void *data;
- loc_t parent;
+ loc_t parent;
- gf_dirent_t *entry;
+ gf_dirent_t *entry;
- struct br_scanfs *fsscan; /* backpointer to subvolume scanner */
+ struct br_scanfs *fsscan; /* backpointer to subvolume scanner */
- struct list_head list;
+ struct list_head list;
};
/**
@@ -46,34 +46,32 @@ struct br_fsscan_entry {
* to the dictionary value.
*/
static int32_t
-bitd_fetch_signature (xlator_t *this, br_child_t *child,
- fd_t *fd, dict_t **xattr, br_isignature_out_t **sign)
+bitd_fetch_signature(xlator_t *this, br_child_t *child, fd_t *fd,
+ dict_t **xattr, br_isignature_out_t **sign)
{
- int32_t ret = -1;
-
- ret = syncop_fgetxattr (child->xl, fd, xattr,
- GLUSTERFS_GET_OBJECT_SIGNATURE, NULL, NULL);
- if (ret < 0) {
- br_log_object (this, "fgetxattr", fd->inode->gfid, -ret);
- goto out;
- }
-
- ret = dict_get_ptr
- (*xattr, GLUSTERFS_GET_OBJECT_SIGNATURE, (void **) sign);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_SIGN_FAILED,
- "failed to extract signature info [GFID: %s]",
- uuid_utoa (fd->inode->gfid));
- goto unref_dict;
- }
-
- return 0;
-
- unref_dict:
- dict_unref (*xattr);
- out:
- return -1;
-
+ int32_t ret = -1;
+
+ ret = syncop_fgetxattr(child->xl, fd, xattr, GLUSTERFS_GET_OBJECT_SIGNATURE,
+ NULL, NULL);
+ if (ret < 0) {
+ br_log_object(this, "fgetxattr", fd->inode->gfid, -ret);
+ goto out;
+ }
+
+ ret = dict_get_ptr(*xattr, GLUSTERFS_GET_OBJECT_SIGNATURE, (void **)sign);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_SIGN_FAILED,
+ "failed to extract signature info [GFID: %s]",
+ uuid_utoa(fd->inode->gfid));
+ goto unref_dict;
+ }
+
+ return 0;
+
+unref_dict:
+ dict_unref(*xattr);
+out:
+ return -1;
}
/**
@@ -86,88 +84,87 @@ bitd_fetch_signature (xlator_t *this, br_child_t *child,
*/
int32_t
-bitd_scrub_post_compute_check (xlator_t *this,
- br_child_t *child,
- fd_t *fd, unsigned long version,
- br_isignature_out_t **signature,
- br_scrub_stats_t *scrub_stat,
- gf_boolean_t skip_stat)
+bitd_scrub_post_compute_check(xlator_t *this, br_child_t *child, fd_t *fd,
+ unsigned long version,
+ br_isignature_out_t **signature,
+ br_scrub_stats_t *scrub_stat,
+ gf_boolean_t skip_stat)
{
- int32_t ret = 0;
- size_t signlen = 0;
- dict_t *xattr = NULL;
- br_isignature_out_t *signptr = NULL;
-
- ret = bitd_fetch_signature (this, child, fd, &xattr, &signptr);
- if (ret < 0) {
- if (!skip_stat)
- br_inc_unsigned_file_count (scrub_stat);
- goto out;
- }
-
- /**
- * Either the object got dirtied during the time the signature was
- * calculated OR the version we saved during pre-compute check does
- * not match now, implying that the object got dirtied and signed in
- * between scrubs pre & post compute checks (checksum window).
- *
- * The log entry looks pretty ugly, but helps in debugging..
- */
- if (signptr->stale || (signptr->version != version)) {
- if (!skip_stat)
- br_inc_unsigned_file_count (scrub_stat);
- gf_msg_debug (this->name, 0, "<STAGE: POST> Object [GFID: %s] "
- "either has a stale signature OR underwent "
- "signing during checksumming {Stale: %d | "
- "Version: %lu,%lu}", uuid_utoa (fd->inode->gfid),
- (signptr->stale) ? 1 : 0, version,
- signptr->version);
- ret = -1;
- goto unref_dict;
- }
-
- signlen = signptr->signaturelen;
- *signature = GF_CALLOC (1, sizeof (br_isignature_out_t) + signlen,
- gf_common_mt_char);
-
- (void) memcpy (*signature, signptr,
- sizeof (br_isignature_out_t) + signlen);
-
- unref_dict:
- dict_unref (xattr);
- out:
- return ret;
+ int32_t ret = 0;
+ size_t signlen = 0;
+ dict_t *xattr = NULL;
+ br_isignature_out_t *signptr = NULL;
+ ret = bitd_fetch_signature(this, child, fd, &xattr, &signptr);
+ if (ret < 0) {
+ if (!skip_stat)
+ br_inc_unsigned_file_count(scrub_stat);
+ goto out;
+ }
+
+ /**
+ * Either the object got dirtied during the time the signature was
+ * calculated OR the version we saved during pre-compute check does
+ * not match now, implying that the object got dirtied and signed in
+ * between scrubs pre & post compute checks (checksum window).
+ *
+ * The log entry looks pretty ugly, but helps in debugging..
+ */
+ if (signptr->stale || (signptr->version != version)) {
+ if (!skip_stat)
+ br_inc_unsigned_file_count(scrub_stat);
+ gf_msg_debug(this->name, 0,
+ "<STAGE: POST> Object [GFID: %s] "
+ "either has a stale signature OR underwent "
+ "signing during checksumming {Stale: %d | "
+ "Version: %lu,%lu}",
+ uuid_utoa(fd->inode->gfid), (signptr->stale) ? 1 : 0,
+ version, signptr->version);
+ ret = -1;
+ goto unref_dict;
+ }
+
+ signlen = signptr->signaturelen;
+ *signature = GF_MALLOC(sizeof(br_isignature_out_t) + signlen,
+ gf_common_mt_char);
+
+ (void)memcpy(*signature, signptr, sizeof(br_isignature_out_t) + signlen);
+
+ (*signature)->signaturelen = signlen;
+
+unref_dict:
+ dict_unref(xattr);
+out:
+ return ret;
}
static int32_t
-bitd_signature_staleness (xlator_t *this,
- br_child_t *child, fd_t *fd,
- int *stale, unsigned long *version,
- br_scrub_stats_t *scrub_stat, gf_boolean_t skip_stat)
+bitd_signature_staleness(xlator_t *this, br_child_t *child, fd_t *fd,
+ int *stale, unsigned long *version,
+ br_scrub_stats_t *scrub_stat, gf_boolean_t skip_stat)
{
- int32_t ret = -1;
- dict_t *xattr = NULL;
- br_isignature_out_t *signptr = NULL;
+ int32_t ret = -1;
+ dict_t *xattr = NULL;
+ br_isignature_out_t *signptr = NULL;
- ret = bitd_fetch_signature (this, child, fd, &xattr, &signptr);
- if (ret < 0) {
- if (!skip_stat)
- br_inc_unsigned_file_count (scrub_stat);
- goto out;
- }
+ ret = bitd_fetch_signature(this, child, fd, &xattr, &signptr);
+ if (ret < 0) {
+ if (!skip_stat)
+ br_inc_unsigned_file_count(scrub_stat);
+ goto out;
+ }
- /**
- * save verison for validation in post compute stage
- * c.f. bitd_scrub_post_compute_check()
- */
- *stale = signptr->stale ? 1 : 0;
- *version = signptr->version;
+ /**
+ * save version for validation in post compute stage
+ * c.f. bitd_scrub_post_compute_check()
+ */
+ *stale = signptr->stale ? 1 : 0;
+ *version = signptr->version;
- dict_unref (xattr);
+ dict_unref(xattr);
- out:
- return ret;
+out:
+ return ret;
}
/**
@@ -179,101 +176,102 @@ bitd_signature_staleness (xlator_t *this,
* - has stale signature
*/
int32_t
-bitd_scrub_pre_compute_check (xlator_t *this, br_child_t *child,
- fd_t *fd, unsigned long *version,
- br_scrub_stats_t *scrub_stat,
- gf_boolean_t skip_stat)
+bitd_scrub_pre_compute_check(xlator_t *this, br_child_t *child, fd_t *fd,
+ unsigned long *version,
+ br_scrub_stats_t *scrub_stat,
+ gf_boolean_t skip_stat)
{
- int stale = 0;
- int32_t ret = -1;
-
- if (bitd_is_bad_file (this, child, NULL, fd)) {
- gf_msg (this->name, GF_LOG_WARNING, 0, BRB_MSG_SKIP_OBJECT,
- "Object [GFID: %s] is marked corrupted, skipping..",
- uuid_utoa (fd->inode->gfid));
- goto out;
- }
-
- ret = bitd_signature_staleness (this, child, fd, &stale, version,
- scrub_stat, skip_stat);
- if (!ret && stale) {
- if (!skip_stat)
- br_inc_unsigned_file_count (scrub_stat);
- gf_msg_debug (this->name, 0, "<STAGE: PRE> Object [GFID: %s] "
- "has stale signature",
- uuid_utoa (fd->inode->gfid));
- ret = -1;
- }
+ int stale = 0;
+ int32_t ret = -1;
+
+ if (bitd_is_bad_file(this, child, NULL, fd)) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, BRB_MSG_SKIP_OBJECT,
+ "Object [GFID: %s] is marked corrupted, skipping..",
+ uuid_utoa(fd->inode->gfid));
+ goto out;
+ }
+
+ ret = bitd_signature_staleness(this, child, fd, &stale, version, scrub_stat,
+ skip_stat);
+ if (!ret && stale) {
+ if (!skip_stat)
+ br_inc_unsigned_file_count(scrub_stat);
+ gf_msg_debug(this->name, 0,
+ "<STAGE: PRE> Object [GFID: %s] "
+ "has stale signature",
+ uuid_utoa(fd->inode->gfid));
+ ret = -1;
+ }
- out:
- return ret;
+out:
+ return ret;
}
/* static int */
int
-bitd_compare_ckum (xlator_t *this,
- br_isignature_out_t *sign,
- unsigned char *md, inode_t *linked_inode,
- gf_dirent_t *entry, fd_t *fd, br_child_t *child, loc_t *loc)
+bitd_compare_ckum(xlator_t *this, br_isignature_out_t *sign, unsigned char *md,
+ inode_t *linked_inode, gf_dirent_t *entry, fd_t *fd,
+ br_child_t *child, loc_t *loc)
{
- int ret = -1;
- dict_t *xattr = NULL;
-
- GF_VALIDATE_OR_GOTO ("bit-rot", this, out);
- GF_VALIDATE_OR_GOTO (this->name, sign, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
- GF_VALIDATE_OR_GOTO (this->name, child, out);
- GF_VALIDATE_OR_GOTO (this->name, linked_inode, out);
- GF_VALIDATE_OR_GOTO (this->name, md, out);
- GF_VALIDATE_OR_GOTO (this->name, entry, out);
-
- if (strncmp
- (sign->signature, (char *) md, strlen (sign->signature)) == 0) {
- gf_msg_debug (this->name, 0, "%s [GFID: %s | Brick: %s] "
- "matches calculated checksum", loc->path,
- uuid_utoa (linked_inode->gfid),
- child->brick_path);
- return 0;
- }
-
- gf_msg (this->name, GF_LOG_DEBUG, 0, BRB_MSG_CHECKSUM_MISMATCH,
- "Object checksum mismatch: %s [GFID: %s | Brick: %s]",
- loc->path, uuid_utoa (linked_inode->gfid), child->brick_path);
- gf_msg (this->name, GF_LOG_ALERT, 0, BRB_MSG_CHECKSUM_MISMATCH,
- "CORRUPTION DETECTED: Object %s {Brick: %s | GFID: %s}",
- loc->path, child->brick_path, uuid_utoa (linked_inode->gfid));
-
- /* Perform bad-file marking */
- xattr = dict_new ();
- if (!xattr) {
- ret = -1;
- goto out;
- }
-
- ret = dict_set_int32 (xattr, BITROT_OBJECT_BAD_KEY, _gf_true);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_MARK_BAD_FILE,
- "Error setting bad-file marker for %s [GFID: %s | "
- "Brick: %s]", loc->path, uuid_utoa (linked_inode->gfid),
- child->brick_path);
- goto dictfree;
- }
-
- gf_msg (this->name, GF_LOG_ALERT, 0, BRB_MSG_MARK_CORRUPTED, "Marking"
- " %s [GFID: %s | Brick: %s] as corrupted..", loc->path,
- uuid_utoa (linked_inode->gfid), child->brick_path);
- gf_event (EVENT_BITROT_BAD_FILE, "gfid=%s;path=%s;brick=%s",
- uuid_utoa (linked_inode->gfid), loc->path, child->brick_path);
- ret = syncop_fsetxattr (child->xl, fd, xattr, 0, NULL, NULL);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_MARK_BAD_FILE,
- "Error marking object %s [GFID: %s] as corrupted",
- loc->path, uuid_utoa (linked_inode->gfid));
-
- dictfree:
- dict_unref (xattr);
- out:
- return ret;
+ int ret = -1;
+ dict_t *xattr = NULL;
+
+ GF_VALIDATE_OR_GOTO("bit-rot", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, sign, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
+ GF_VALIDATE_OR_GOTO(this->name, child, out);
+ GF_VALIDATE_OR_GOTO(this->name, linked_inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, md, out);
+ GF_VALIDATE_OR_GOTO(this->name, entry, out);
+
+ if (strncmp(sign->signature, (char *)md, sign->signaturelen) == 0) {
+ gf_msg_debug(this->name, 0,
+ "%s [GFID: %s | Brick: %s] "
+ "matches calculated checksum",
+ loc->path, uuid_utoa(linked_inode->gfid),
+ child->brick_path);
+ return 0;
+ }
+
+ gf_msg(this->name, GF_LOG_DEBUG, 0, BRB_MSG_CHECKSUM_MISMATCH,
+ "Object checksum mismatch: %s [GFID: %s | Brick: %s]", loc->path,
+ uuid_utoa(linked_inode->gfid), child->brick_path);
+ gf_msg(this->name, GF_LOG_ALERT, 0, BRB_MSG_CHECKSUM_MISMATCH,
+ "CORRUPTION DETECTED: Object %s {Brick: %s | GFID: %s}", loc->path,
+ child->brick_path, uuid_utoa(linked_inode->gfid));
+
+ /* Perform bad-file marking */
+ xattr = dict_new();
+ if (!xattr) {
+ ret = -1;
+ goto out;
+ }
+
+ ret = dict_set_int32(xattr, BITROT_OBJECT_BAD_KEY, _gf_true);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_MARK_BAD_FILE,
+ "Error setting bad-file marker for %s [GFID: %s | "
+ "Brick: %s]",
+ loc->path, uuid_utoa(linked_inode->gfid), child->brick_path);
+ goto dictfree;
+ }
+
+ gf_msg(this->name, GF_LOG_ALERT, 0, BRB_MSG_MARK_CORRUPTED,
+ "Marking"
+ " %s [GFID: %s | Brick: %s] as corrupted..",
+ loc->path, uuid_utoa(linked_inode->gfid), child->brick_path);
+ gf_event(EVENT_BITROT_BAD_FILE, "gfid=%s;path=%s;brick=%s",
+ uuid_utoa(linked_inode->gfid), loc->path, child->brick_path);
+ ret = syncop_fsetxattr(child->xl, fd, xattr, 0, NULL, NULL);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_MARK_BAD_FILE,
+ "Error marking object %s [GFID: %s] as corrupted", loc->path,
+ uuid_utoa(linked_inode->gfid));
+
+dictfree:
+ dict_unref(xattr);
+out:
+ return ret;
}
/**
@@ -284,540 +282,550 @@ bitd_compare_ckum (xlator_t *this,
* signs with SHA256).
*/
int
-br_scrubber_scrub_begin (xlator_t *this, struct br_fsscan_entry *fsentry)
+br_scrubber_scrub_begin(xlator_t *this, struct br_fsscan_entry *fsentry)
{
- int32_t ret = -1;
- fd_t *fd = NULL;
- loc_t loc = {0, };
- struct iatt iatt = {0, };
- struct iatt parent_buf = {0, };
- pid_t pid = 0;
- br_child_t *child = NULL;
- unsigned char *md = NULL;
- inode_t *linked_inode = NULL;
- br_isignature_out_t *sign = NULL;
- unsigned long signedversion = 0;
- gf_dirent_t *entry = NULL;
- br_private_t *priv = NULL;
- loc_t *parent = NULL;
- gf_boolean_t skip_stat = _gf_false;
- uuid_t shard_root_gfid = {0,};
-
-
- GF_VALIDATE_OR_GOTO ("bit-rot", fsentry, out);
-
- entry = fsentry->entry;
- parent = &fsentry->parent;
- child = fsentry->data;
-
- priv = this->private;
-
- GF_VALIDATE_OR_GOTO ("bit-rot", entry, out);
- GF_VALIDATE_OR_GOTO ("bit-rot", parent, out);
- GF_VALIDATE_OR_GOTO ("bit-rot", child, out);
- GF_VALIDATE_OR_GOTO ("bit-rot", priv, out);
-
- pid = GF_CLIENT_PID_SCRUB;
-
- ret = br_prepare_loc (this, child, parent, entry, &loc);
- if (!ret)
- goto out;
-
- syncopctx_setfspid (&pid);
-
- ret = syncop_lookup (child->xl, &loc, &iatt, &parent_buf, NULL, NULL);
- if (ret) {
- br_log_object_path (this, "lookup", loc.path, -ret);
- goto out;
- }
-
- linked_inode = inode_link (loc.inode, parent->inode, loc.name, &iatt);
- if (linked_inode)
- inode_lookup (linked_inode);
-
- gf_msg_debug (this->name, 0, "Scrubbing object %s [GFID: %s]",
- entry->d_name, uuid_utoa (linked_inode->gfid));
-
- if (iatt.ia_type != IA_IFREG) {
- gf_msg_debug (this->name, 0, "%s is not a regular file",
- entry->d_name);
- ret = 0;
- goto unref_inode;
- }
-
- if (IS_DHT_LINKFILE_MODE ((&iatt))) {
- gf_msg_debug (this->name, 0, "%s is a dht sticky bit file",
- entry->d_name);
- ret = 0;
- goto unref_inode;
- }
-
- /* skip updating scrub statistics for shard entries */
- gf_uuid_parse (SHARD_ROOT_GFID, shard_root_gfid);
- if (gf_uuid_compare (loc.pargfid, shard_root_gfid) == 0)
- skip_stat = _gf_true;
-
- /**
- * open() an fd for subsequent opertaions
- */
- fd = fd_create (linked_inode, 0);
- if (!fd) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_FD_CREATE_FAILED,
- "failed to create fd for inode %s",
- uuid_utoa (linked_inode->gfid));
- goto unref_inode;
- }
-
- ret = syncop_open (child->xl, &loc, O_RDWR, fd, NULL, NULL);
- if (ret) {
- br_log_object (this, "open", linked_inode->gfid, -ret);
- ret = -1;
- goto unrefd;
- }
-
- fd_bind (fd);
-
- /**
- * perform pre compute checks before initiating checksum
- * computation
- * - presence of bad object
- * - signature staleness
- */
- ret = bitd_scrub_pre_compute_check (this, child, fd, &signedversion,
- &priv->scrub_stat, skip_stat);
- if (ret)
- goto unrefd; /* skip this object */
-
- /* if all's good, proceed to calculate the hash */
- md = GF_CALLOC (SHA256_DIGEST_LENGTH, sizeof (*md),
- gf_common_mt_char);
- if (!md)
- goto unrefd;
-
- ret = br_calculate_obj_checksum (md, child, fd, &iatt);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_CALC_ERROR,
- "error calculating hash for object [GFID: %s]",
- uuid_utoa (fd->inode->gfid));
- ret = -1;
- goto free_md;
- }
-
- /**
- * perform post compute checks as an object's signature may have
- * become stale while scrubber calculated checksum.
- */
- ret = bitd_scrub_post_compute_check (this, child, fd, signedversion,
- &sign, &priv->scrub_stat,
- skip_stat);
- if (ret)
- goto free_md;
-
- ret = bitd_compare_ckum (this, sign, md,
- linked_inode, entry, fd, child, &loc);
-
- if (!skip_stat)
- br_inc_scrubbed_file (&priv->scrub_stat);
-
- GF_FREE (sign); /* alloced on post-compute */
-
- /** fd_unref() takes care of closing fd.. like syncop_close() */
+ int32_t ret = -1;
+ fd_t *fd = NULL;
+ loc_t loc = {
+ 0,
+ };
+ struct iatt iatt = {
+ 0,
+ };
+ struct iatt parent_buf = {
+ 0,
+ };
+ pid_t pid = 0;
+ br_child_t *child = NULL;
+ unsigned char *md = NULL;
+ inode_t *linked_inode = NULL;
+ br_isignature_out_t *sign = NULL;
+ unsigned long signedversion = 0;
+ gf_dirent_t *entry = NULL;
+ br_private_t *priv = NULL;
+ loc_t *parent = NULL;
+ gf_boolean_t skip_stat = _gf_false;
+ uuid_t shard_root_gfid = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("bit-rot", fsentry, out);
+
+ entry = fsentry->entry;
+ parent = &fsentry->parent;
+ child = fsentry->data;
+
+ priv = this->private;
+
+ GF_VALIDATE_OR_GOTO("bit-rot", entry, out);
+ GF_VALIDATE_OR_GOTO("bit-rot", parent, out);
+ GF_VALIDATE_OR_GOTO("bit-rot", child, out);
+ GF_VALIDATE_OR_GOTO("bit-rot", priv, out);
+
+ pid = GF_CLIENT_PID_SCRUB;
+
+ ret = br_prepare_loc(this, child, parent, entry, &loc);
+ if (!ret)
+ goto out;
+
+ syncopctx_setfspid(&pid);
+
+ ret = syncop_lookup(child->xl, &loc, &iatt, &parent_buf, NULL, NULL);
+ if (ret) {
+ br_log_object_path(this, "lookup", loc.path, -ret);
+ goto out;
+ }
+
+ linked_inode = inode_link(loc.inode, parent->inode, loc.name, &iatt);
+ if (linked_inode)
+ inode_lookup(linked_inode);
+
+ gf_msg_debug(this->name, 0, "Scrubbing object %s [GFID: %s]", entry->d_name,
+ uuid_utoa(linked_inode->gfid));
+
+ if (iatt.ia_type != IA_IFREG) {
+ gf_msg_debug(this->name, 0, "%s is not a regular file", entry->d_name);
+ ret = 0;
+ goto unref_inode;
+ }
- free_md:
- GF_FREE (md);
- unrefd:
- fd_unref (fd);
- unref_inode:
- inode_unref (linked_inode);
- out:
- loc_wipe (&loc);
- return ret;
+ if (IS_DHT_LINKFILE_MODE((&iatt))) {
+ gf_msg_debug(this->name, 0, "%s is a dht sticky bit file",
+ entry->d_name);
+ ret = 0;
+ goto unref_inode;
+ }
+
+ /* skip updating scrub statistics for shard entries */
+ gf_uuid_parse(SHARD_ROOT_GFID, shard_root_gfid);
+ if (gf_uuid_compare(loc.pargfid, shard_root_gfid) == 0)
+ skip_stat = _gf_true;
+
+ /**
+ * open() an fd for subsequent operations
+ */
+ fd = fd_create(linked_inode, 0);
+ if (!fd) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_FD_CREATE_FAILED,
+ "failed to create fd for inode %s",
+ uuid_utoa(linked_inode->gfid));
+ goto unref_inode;
+ }
+
+ ret = syncop_open(child->xl, &loc, O_RDWR, fd, NULL, NULL);
+ if (ret) {
+ br_log_object(this, "open", linked_inode->gfid, -ret);
+ ret = -1;
+ goto unrefd;
+ }
+
+ fd_bind(fd);
+
+ /**
+ * perform pre compute checks before initiating checksum
+ * computation
+ * - presence of bad object
+ * - signature staleness
+ */
+ ret = bitd_scrub_pre_compute_check(this, child, fd, &signedversion,
+ &priv->scrub_stat, skip_stat);
+ if (ret)
+ goto unrefd; /* skip this object */
+
+ /* if all's good, proceed to calculate the hash */
+ md = GF_MALLOC(SHA256_DIGEST_LENGTH, gf_common_mt_char);
+ if (!md)
+ goto unrefd;
+
+ ret = br_calculate_obj_checksum(md, child, fd, &iatt);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_CALC_ERROR,
+ "error calculating hash for object [GFID: %s]",
+ uuid_utoa(fd->inode->gfid));
+ ret = -1;
+ goto free_md;
+ }
+
+ /**
+ * perform post compute checks as an object's signature may have
+ * become stale while scrubber calculated checksum.
+ */
+ ret = bitd_scrub_post_compute_check(this, child, fd, signedversion, &sign,
+ &priv->scrub_stat, skip_stat);
+ if (ret)
+ goto free_md;
+
+ ret = bitd_compare_ckum(this, sign, md, linked_inode, entry, fd, child,
+ &loc);
+
+ if (!skip_stat)
+ br_inc_scrubbed_file(&priv->scrub_stat);
+
+ GF_FREE(sign); /* allocated on post-compute */
+
+ /** fd_unref() takes care of closing fd.. like syncop_close() */
+
+free_md:
+ GF_FREE(md);
+unrefd:
+ fd_unref(fd);
+unref_inode:
+ inode_unref(linked_inode);
+out:
+ loc_wipe(&loc);
+ return ret;
}
static void
-_br_lock_cleaner (void *arg)
+_br_lock_cleaner(void *arg)
{
- pthread_mutex_t *mutex = arg;
+ pthread_mutex_t *mutex = arg;
- pthread_mutex_unlock (mutex);
+ pthread_mutex_unlock(mutex);
}
static void
-wait_for_scrubbing (xlator_t *this, struct br_scanfs *fsscan)
+wait_for_scrubbing(xlator_t *this, struct br_scanfs *fsscan)
{
- br_private_t *priv = NULL;
- struct br_scrubber *fsscrub = NULL;
+ br_private_t *priv = NULL;
+ struct br_scrubber *fsscrub = NULL;
- priv = this->private;
- fsscrub = &priv->fsscrub;
+ priv = this->private;
+ fsscrub = &priv->fsscrub;
- pthread_cleanup_push (_br_lock_cleaner, &fsscan->waitlock);
- pthread_mutex_lock (&fsscan->waitlock);
+ pthread_cleanup_push(_br_lock_cleaner, &fsscan->waitlock);
+ pthread_mutex_lock(&fsscan->waitlock);
+ {
+ pthread_cleanup_push(_br_lock_cleaner, &fsscrub->mutex);
+ pthread_mutex_lock(&fsscrub->mutex);
{
- pthread_cleanup_push (_br_lock_cleaner, &fsscrub->mutex);
- pthread_mutex_lock (&fsscrub->mutex);
- {
- list_replace_init (&fsscan->queued, &fsscan->ready);
-
- /* wake up scrubbers */
- pthread_cond_broadcast (&fsscrub->cond);
- }
- pthread_mutex_unlock (&fsscrub->mutex);
- pthread_cleanup_pop (0);
-
- while (fsscan->entries != 0)
- pthread_cond_wait
- (&fsscan->waitcond, &fsscan->waitlock);
+ list_replace_init(&fsscan->queued, &fsscan->ready);
+
+ /* wake up scrubbers */
+ pthread_cond_broadcast(&fsscrub->cond);
}
- pthread_mutex_unlock (&fsscan->waitlock);
- pthread_cleanup_pop (0);
+ pthread_mutex_unlock(&fsscrub->mutex);
+ pthread_cleanup_pop(0);
+
+ while (fsscan->entries != 0)
+ pthread_cond_wait(&fsscan->waitcond, &fsscan->waitlock);
+ }
+ pthread_mutex_unlock(&fsscan->waitlock);
+ pthread_cleanup_pop(0);
}
static void
-_br_fsscan_inc_entry_count (struct br_scanfs *fsscan)
+_br_fsscan_inc_entry_count(struct br_scanfs *fsscan)
{
- fsscan->entries++;
+ fsscan->entries++;
}
static void
-_br_fsscan_dec_entry_count (struct br_scanfs *fsscan)
+_br_fsscan_dec_entry_count(struct br_scanfs *fsscan)
{
- if (--fsscan->entries == 0) {
- pthread_mutex_lock (&fsscan->waitlock);
- {
- pthread_cond_signal (&fsscan->waitcond);
- }
- pthread_mutex_unlock (&fsscan->waitlock);
+ if (--fsscan->entries == 0) {
+ pthread_mutex_lock(&fsscan->waitlock);
+ {
+ pthread_cond_signal(&fsscan->waitcond);
}
+ pthread_mutex_unlock(&fsscan->waitlock);
+ }
}
static void
-_br_fsscan_collect_entry (struct br_scanfs *fsscan,
- struct br_fsscan_entry *fsentry)
+_br_fsscan_collect_entry(struct br_scanfs *fsscan,
+ struct br_fsscan_entry *fsentry)
{
- list_add_tail (&fsentry->list, &fsscan->queued);
- _br_fsscan_inc_entry_count (fsscan);
+ list_add_tail(&fsentry->list, &fsscan->queued);
+ _br_fsscan_inc_entry_count(fsscan);
}
-#define NR_ENTRIES (1<<7) /* ..bulk scrubbing */
+#define NR_ENTRIES (1 << 7) /* ..bulk scrubbing */
int
-br_fsscanner_handle_entry (xlator_t *subvol,
- gf_dirent_t *entry, loc_t *parent, void *data)
+br_fsscanner_handle_entry(xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
+ void *data)
{
- int32_t ret = -1;
- int scrub = 0;
- br_child_t *child = NULL;
- xlator_t *this = NULL;
- struct br_scanfs *fsscan = NULL;
- struct br_fsscan_entry *fsentry = NULL;
+ int32_t ret = -1;
+ int scrub = 0;
+ br_child_t *child = NULL;
+ xlator_t *this = NULL;
+ struct br_scanfs *fsscan = NULL;
+ struct br_fsscan_entry *fsentry = NULL;
- GF_VALIDATE_OR_GOTO ("bit-rot", subvol, error_return);
- GF_VALIDATE_OR_GOTO ("bit-rot", data, error_return);
+ GF_VALIDATE_OR_GOTO("bit-rot", subvol, error_return);
+ GF_VALIDATE_OR_GOTO("bit-rot", data, error_return);
- child = data;
- this = child->this;
- fsscan = &child->fsscan;
+ child = data;
+ this = child->this;
+ fsscan = &child->fsscan;
- _mask_cancellation ();
+ _mask_cancellation();
- fsentry = GF_CALLOC (1, sizeof (*fsentry), gf_br_mt_br_fsscan_entry_t);
- if (!fsentry)
- goto error_return;
+ fsentry = GF_CALLOC(1, sizeof(*fsentry), gf_br_mt_br_fsscan_entry_t);
+ if (!fsentry)
+ goto error_return;
- {
- fsentry->data = data;
- fsentry->fsscan = &child->fsscan;
+ {
+ fsentry->data = data;
+ fsentry->fsscan = &child->fsscan;
+
+ /* copy parent loc */
+ ret = loc_copy(&fsentry->parent, parent);
+ if (ret)
+ goto dealloc;
- /* copy parent loc */
- ret = loc_copy (&fsentry->parent, parent);
- if (ret)
- goto dealloc;
+ /* copy child entry */
+ fsentry->entry = entry_copy(entry);
+ if (!fsentry->entry)
+ goto locwipe;
- /* copy child entry */
- fsentry->entry = entry_copy (entry);
- if (!fsentry->entry)
- goto locwipe;
+ INIT_LIST_HEAD(&fsentry->list);
+ }
- INIT_LIST_HEAD (&fsentry->list);
- }
+ LOCK(&fsscan->entrylock);
+ {
+ _br_fsscan_collect_entry(fsscan, fsentry);
- LOCK (&fsscan->entrylock);
- {
- _br_fsscan_collect_entry (fsscan, fsentry);
-
- /**
- * need not be a equality check as entries may be pushed
- * back onto the scanned queue when thread(s) are cleaned.
- */
- if (fsscan->entries >= NR_ENTRIES)
- scrub = 1;
- }
- UNLOCK (&fsscan->entrylock);
+ /**
+ * need not be a equality check as entries may be pushed
+ * back onto the scanned queue when thread(s) are cleaned.
+ */
+ if (fsscan->entries >= NR_ENTRIES)
+ scrub = 1;
+ }
+ UNLOCK(&fsscan->entrylock);
- _unmask_cancellation ();
+ _unmask_cancellation();
- if (scrub)
- wait_for_scrubbing (this, fsscan);
+ if (scrub)
+ wait_for_scrubbing(this, fsscan);
- return 0;
+ return 0;
- locwipe:
- loc_wipe (&fsentry->parent);
- dealloc:
- GF_FREE (fsentry);
- error_return:
- return -1;
+locwipe:
+ loc_wipe(&fsentry->parent);
+dealloc:
+ GF_FREE(fsentry);
+error_return:
+ return -1;
}
int32_t
-br_fsscan_deactivate (xlator_t *this)
+br_fsscan_deactivate(xlator_t *this)
{
- int ret = 0;
- br_private_t *priv = NULL;
- br_scrub_state_t nstate = 0;
- struct br_monitor *scrub_monitor = NULL;
-
- priv = this->private;
- scrub_monitor = &priv->scrub_monitor;
-
- ret = gf_tw_del_timer (priv->timer_wheel, scrub_monitor->timer);
- if (ret == 0) {
- nstate = BR_SCRUB_STATE_STALLED;
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
- "Volume is under active scrubbing. Pausing scrub..");
- } else {
- nstate = BR_SCRUB_STATE_PAUSED;
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
- "Scrubber paused");
- }
-
- _br_monitor_set_scrub_state (scrub_monitor, nstate);
-
- return 0;
+ int ret = 0;
+ br_private_t *priv = NULL;
+ br_scrub_state_t nstate = 0;
+ struct br_monitor *scrub_monitor = NULL;
+
+ priv = this->private;
+ scrub_monitor = &priv->scrub_monitor;
+
+ ret = gf_tw_del_timer(priv->timer_wheel, scrub_monitor->timer);
+ if (ret == 0) {
+ nstate = BR_SCRUB_STATE_STALLED;
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
+ "Volume is under active scrubbing. Pausing scrub..");
+ } else {
+ nstate = BR_SCRUB_STATE_PAUSED;
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
+ "Scrubber paused");
+ }
+
+ _br_monitor_set_scrub_state(scrub_monitor, nstate);
+
+ return 0;
}
static void
-br_scrubber_log_time (xlator_t *this, const char *sfx)
+br_scrubber_log_time(xlator_t *this, const char *sfx)
{
- char timestr[1024] = {0,};
- struct timeval tv = {0,};
- br_private_t *priv = NULL;
-
- priv = this->private;
-
- gettimeofday (&tv, NULL);
- gf_time_fmt (timestr, sizeof (timestr), tv.tv_sec, gf_timefmt_FT);
-
- if (strcasecmp (sfx, "started") == 0) {
- br_update_scrub_start_time (&priv->scrub_stat, &tv);
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_START,
- "Scrubbing %s at %s", sfx, timestr);
- } else {
- br_update_scrub_finish_time (&priv->scrub_stat, timestr, &tv);
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_FINISH,
- "Scrubbing %s at %s", sfx, timestr);
- }
+ char timestr[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ br_private_t *priv = NULL;
+ time_t now = 0;
+
+ now = gf_time();
+ priv = this->private;
+
+ gf_time_fmt(timestr, sizeof(timestr), now, gf_timefmt_FT);
+
+ if (strcasecmp(sfx, "started") == 0) {
+ br_update_scrub_start_time(&priv->scrub_stat, now);
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_START,
+ "Scrubbing %s at %s", sfx, timestr);
+ } else {
+ br_update_scrub_finish_time(&priv->scrub_stat, timestr, now);
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_FINISH,
+ "Scrubbing %s at %s", sfx, timestr);
+ }
}
static void
-br_fsscanner_log_time (xlator_t *this, br_child_t *child, const char *sfx)
+br_fsscanner_log_time(xlator_t *this, br_child_t *child, const char *sfx)
{
- char timestr[1024] = {0,};
- struct timeval tv = {0,};
-
- gettimeofday (&tv, NULL);
- gf_time_fmt (timestr, sizeof (timestr), tv.tv_sec, gf_timefmt_FT);
-
- if (strcasecmp (sfx, "started") == 0) {
- gf_msg_debug (this->name, 0, "Scrubbing \"%s\" %s at %s",
- child->brick_path, sfx, timestr);
- } else {
- gf_msg_debug (this->name, 0, "Scrubbing \"%s\" %s at %s",
- child->brick_path, sfx, timestr);
- }
+ char timestr[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ time_t now = 0;
+
+ now = gf_time();
+ gf_time_fmt(timestr, sizeof(timestr), now, gf_timefmt_FT);
+
+ if (strcasecmp(sfx, "started") == 0) {
+ gf_msg_debug(this->name, 0, "Scrubbing \"%s\" %s at %s",
+ child->brick_path, sfx, timestr);
+ } else {
+ gf_msg_debug(this->name, 0, "Scrubbing \"%s\" %s at %s",
+ child->brick_path, sfx, timestr);
+ }
}
void
-br_child_set_scrub_state (br_child_t *child, gf_boolean_t state)
+br_child_set_scrub_state(br_child_t *child, gf_boolean_t state)
{
- child->active_scrubbing = state;
+ child->active_scrubbing = state;
}
static void
-br_fsscanner_wait_until_kicked (xlator_t *this, br_child_t *child)
+br_fsscanner_wait_until_kicked(xlator_t *this, br_child_t *child)
{
- br_private_t *priv = NULL;
- struct br_monitor *scrub_monitor = NULL;
-
- priv = this->private;
- scrub_monitor = &priv->scrub_monitor;
-
- pthread_cleanup_push (_br_lock_cleaner, &scrub_monitor->wakelock);
- pthread_mutex_lock (&scrub_monitor->wakelock);
+ br_private_t *priv = NULL;
+ struct br_monitor *scrub_monitor = NULL;
+
+ priv = this->private;
+ scrub_monitor = &priv->scrub_monitor;
+
+ pthread_cleanup_push(_br_lock_cleaner, &scrub_monitor->wakelock);
+ pthread_mutex_lock(&scrub_monitor->wakelock);
+ {
+ while (!scrub_monitor->kick)
+ pthread_cond_wait(&scrub_monitor->wakecond,
+ &scrub_monitor->wakelock);
+
+ /* Child lock is to synchronize with disconnect events */
+ pthread_cleanup_push(_br_lock_cleaner, &child->lock);
+ pthread_mutex_lock(&child->lock);
{
- while (!scrub_monitor->kick)
- pthread_cond_wait (&scrub_monitor->wakecond,
- &scrub_monitor->wakelock);
-
- /* Child lock is to synchronize with disconnect events */
- pthread_cleanup_push (_br_lock_cleaner, &child->lock);
- pthread_mutex_lock (&child->lock);
- {
- scrub_monitor->active_child_count++;
- br_child_set_scrub_state (child, _gf_true);
- }
- pthread_mutex_unlock (&child->lock);
- pthread_cleanup_pop (0);
- }
- pthread_mutex_unlock (&scrub_monitor->wakelock);
- pthread_cleanup_pop (0);
+ scrub_monitor->active_child_count++;
+ br_child_set_scrub_state(child, _gf_true);
+ }
+ pthread_mutex_unlock(&child->lock);
+ pthread_cleanup_pop(0);
+ }
+ pthread_mutex_unlock(&scrub_monitor->wakelock);
+ pthread_cleanup_pop(0);
}
static void
-br_scrubber_entry_control (xlator_t *this)
+br_scrubber_entry_control(xlator_t *this)
{
- br_private_t *priv = NULL;
- struct br_monitor *scrub_monitor = NULL;
-
- priv = this->private;
- scrub_monitor = &priv->scrub_monitor;
-
- LOCK (&scrub_monitor->lock);
- {
- /* Move the state to BR_SCRUB_STATE_ACTIVE */
- if (scrub_monitor->state == BR_SCRUB_STATE_PENDING)
- scrub_monitor->state = BR_SCRUB_STATE_ACTIVE;
- br_scrubber_log_time (this, "started");
- priv->scrub_stat.scrub_running = 1;
- }
- UNLOCK (&scrub_monitor->lock);
+ br_private_t *priv = NULL;
+ struct br_monitor *scrub_monitor = NULL;
+
+ priv = this->private;
+ scrub_monitor = &priv->scrub_monitor;
+
+ LOCK(&scrub_monitor->lock);
+ {
+ /* Move the state to BR_SCRUB_STATE_ACTIVE */
+ if (scrub_monitor->state == BR_SCRUB_STATE_PENDING)
+ scrub_monitor->state = BR_SCRUB_STATE_ACTIVE;
+ br_scrubber_log_time(this, "started");
+ priv->scrub_stat.scrub_running = 1;
+ }
+ UNLOCK(&scrub_monitor->lock);
}
static void
-br_scrubber_exit_control (xlator_t *this)
+br_scrubber_exit_control(xlator_t *this)
{
- br_private_t *priv = NULL;
- struct br_monitor *scrub_monitor = NULL;
+ br_private_t *priv = NULL;
+ struct br_monitor *scrub_monitor = NULL;
- priv = this->private;
- scrub_monitor = &priv->scrub_monitor;
+ priv = this->private;
+ scrub_monitor = &priv->scrub_monitor;
- LOCK (&scrub_monitor->lock);
- {
- br_scrubber_log_time (this, "finished");
- priv->scrub_stat.scrub_running = 0;
-
- if (scrub_monitor->state == BR_SCRUB_STATE_ACTIVE) {
- (void) br_fsscan_activate (this);
- } else {
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
- "Volume waiting to get rescheduled..");
- }
+ LOCK(&scrub_monitor->lock);
+ {
+ br_scrubber_log_time(this, "finished");
+ priv->scrub_stat.scrub_running = 0;
+
+ if (scrub_monitor->state == BR_SCRUB_STATE_ACTIVE) {
+ (void)br_fsscan_activate(this);
+ } else {
+ UNLOCK(&scrub_monitor->lock);
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
+ "Volume waiting to get rescheduled..");
+ return;
}
- UNLOCK (&scrub_monitor->lock);
+ }
+ UNLOCK(&scrub_monitor->lock);
}
static void
-br_fsscanner_entry_control (xlator_t *this, br_child_t *child)
+br_fsscanner_entry_control(xlator_t *this, br_child_t *child)
{
- br_fsscanner_log_time (this, child, "started");
+ br_fsscanner_log_time(this, child, "started");
}
static void
-br_fsscanner_exit_control (xlator_t *this, br_child_t *child)
+br_fsscanner_exit_control(xlator_t *this, br_child_t *child)
{
- br_private_t *priv = NULL;
- struct br_monitor *scrub_monitor = NULL;
-
- priv = this->private;
- scrub_monitor = &priv->scrub_monitor;
-
- if (!_br_is_child_connected (child)) {
- gf_msg (this->name, GF_LOG_WARNING, 0, BRB_MSG_SCRUB_INFO,
- "Brick [%s] disconnected while scrubbing. Scrubbing "
- "might be incomplete", child->brick_path);
- }
-
- br_fsscanner_log_time (this, child, "finished");
-
- pthread_cleanup_push (_br_lock_cleaner, &scrub_monitor->wakelock);
- pthread_mutex_lock (&scrub_monitor->wakelock);
+ br_private_t *priv = NULL;
+ struct br_monitor *scrub_monitor = NULL;
+
+ priv = this->private;
+ scrub_monitor = &priv->scrub_monitor;
+
+ if (!_br_is_child_connected(child)) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, BRB_MSG_SCRUB_INFO,
+ "Brick [%s] disconnected while scrubbing. Scrubbing "
+ "might be incomplete",
+ child->brick_path);
+ }
+
+ br_fsscanner_log_time(this, child, "finished");
+
+ pthread_cleanup_push(_br_lock_cleaner, &scrub_monitor->wakelock);
+ pthread_mutex_lock(&scrub_monitor->wakelock);
+ {
+ scrub_monitor->active_child_count--;
+ pthread_cleanup_push(_br_lock_cleaner, &child->lock);
+ pthread_mutex_lock(&child->lock);
{
- scrub_monitor->active_child_count--;
- pthread_cleanup_push (_br_lock_cleaner, &child->lock);
- pthread_mutex_lock (&child->lock);
- {
- br_child_set_scrub_state (child, _gf_false);
- }
- pthread_mutex_unlock (&child->lock);
- pthread_cleanup_pop (0);
-
- if (scrub_monitor->active_child_count == 0) {
- /* The last child has finished scrubbing.
- * Set the kick to false and wake up other
- * children who are waiting for the last
- * child to complete scrubbing.
- */
- scrub_monitor->kick = _gf_false;
- pthread_cond_broadcast (&scrub_monitor->wakecond);
-
- /* Signal monitor thread waiting for the all
- * the children to finish scrubbing.
- */
- pthread_cleanup_push (_br_lock_cleaner,
- &scrub_monitor->donelock);
- pthread_mutex_lock (&scrub_monitor->donelock);
- {
- scrub_monitor->done = _gf_true;
- pthread_cond_signal (&scrub_monitor->donecond);
- }
- pthread_mutex_unlock (&scrub_monitor->donelock);
- pthread_cleanup_pop (0);
- } else {
- while (scrub_monitor->active_child_count)
- pthread_cond_wait (&scrub_monitor->wakecond,
- &scrub_monitor->wakelock);
- }
+ br_child_set_scrub_state(child, _gf_false);
+ }
+ pthread_mutex_unlock(&child->lock);
+ pthread_cleanup_pop(0);
+
+ if (scrub_monitor->active_child_count == 0) {
+ /* The last child has finished scrubbing.
+ * Set the kick to false and wake up other
+ * children who are waiting for the last
+ * child to complete scrubbing.
+ */
+ scrub_monitor->kick = _gf_false;
+ pthread_cond_broadcast(&scrub_monitor->wakecond);
+
+ /* Signal monitor thread waiting for the all
+ * the children to finish scrubbing.
+ */
+ pthread_cleanup_push(_br_lock_cleaner, &scrub_monitor->donelock);
+ pthread_mutex_lock(&scrub_monitor->donelock);
+ {
+ scrub_monitor->done = _gf_true;
+ pthread_cond_signal(&scrub_monitor->donecond);
+ }
+ pthread_mutex_unlock(&scrub_monitor->donelock);
+ pthread_cleanup_pop(0);
+ } else {
+ while (scrub_monitor->active_child_count)
+ pthread_cond_wait(&scrub_monitor->wakecond,
+ &scrub_monitor->wakelock);
}
- pthread_mutex_unlock (&scrub_monitor->wakelock);
- pthread_cleanup_pop (0);
+ }
+ pthread_mutex_unlock(&scrub_monitor->wakelock);
+ pthread_cleanup_pop(0);
}
void *
-br_fsscanner (void *arg)
+br_fsscanner(void *arg)
{
- loc_t loc = {0,};
- br_child_t *child = NULL;
- xlator_t *this = NULL;
- struct br_scanfs *fsscan = NULL;
-
- child = arg;
- this = child->this;
- fsscan = &child->fsscan;
-
- THIS = this;
- loc.inode = child->table->root;
-
- while (1) {
- br_fsscanner_wait_until_kicked (this, child);
- {
- /* precursor for scrub */
- br_fsscanner_entry_control (this, child);
-
- /* scrub */
- (void) syncop_ftw (child->xl,
- &loc, GF_CLIENT_PID_SCRUB,
- child, br_fsscanner_handle_entry);
- if (!list_empty (&fsscan->queued))
- wait_for_scrubbing (this, fsscan);
-
- /* scrub exit criteria */
- br_fsscanner_exit_control (this, child);
- }
+ loc_t loc = {
+ 0,
+ };
+ br_child_t *child = NULL;
+ xlator_t *this = NULL;
+ struct br_scanfs *fsscan = NULL;
+
+ child = arg;
+ this = child->this;
+ fsscan = &child->fsscan;
+
+ THIS = this;
+ loc.inode = child->table->root;
+
+ while (1) {
+ br_fsscanner_wait_until_kicked(this, child);
+ {
+ /* precursor for scrub */
+ br_fsscanner_entry_control(this, child);
+
+ /* scrub */
+ (void)syncop_ftw(child->xl, &loc, GF_CLIENT_PID_SCRUB, child,
+ br_fsscanner_handle_entry);
+ if (!list_empty(&fsscan->queued))
+ wait_for_scrubbing(this, fsscan);
+
+ /* scrub exit criteria */
+ br_fsscanner_exit_control(this, child);
}
+ }
- return NULL;
+ return NULL;
}
/**
@@ -828,263 +836,268 @@ br_fsscanner (void *arg)
* non-pending timer.
*/
void
-br_kickstart_scanner (struct gf_tw_timer_list *timer,
- void *data, unsigned long calltime)
+br_kickstart_scanner(struct gf_tw_timer_list *timer, void *data,
+ unsigned long calltime)
{
- xlator_t *this = NULL;
- struct br_monitor *scrub_monitor = data;
- br_private_t *priv = NULL;
-
- THIS = this = scrub_monitor->this;
- priv = this->private;
-
- /* Reset scrub statistics */
- priv->scrub_stat.scrubbed_files = 0;
- priv->scrub_stat.unsigned_files = 0;
-
- /* Moves state from PENDING to ACTIVE */
- (void) br_scrubber_entry_control (this);
-
- /* kickstart scanning.. */
- pthread_mutex_lock (&scrub_monitor->wakelock);
- {
- scrub_monitor->kick = _gf_true;
- GF_ASSERT (scrub_monitor->active_child_count == 0);
- pthread_cond_broadcast (&scrub_monitor->wakecond);
- }
- pthread_mutex_unlock (&scrub_monitor->wakelock);
-
- return;
+ xlator_t *this = NULL;
+ struct br_monitor *scrub_monitor = data;
+ br_private_t *priv = NULL;
+
+ THIS = this = scrub_monitor->this;
+ priv = this->private;
+
+ /* Reset scrub statistics */
+ priv->scrub_stat.scrubbed_files = 0;
+ priv->scrub_stat.unsigned_files = 0;
+
+ /* Moves state from PENDING to ACTIVE */
+ (void)br_scrubber_entry_control(this);
+
+ /* kickstart scanning.. */
+ pthread_mutex_lock(&scrub_monitor->wakelock);
+ {
+ scrub_monitor->kick = _gf_true;
+ GF_ASSERT(scrub_monitor->active_child_count == 0);
+ pthread_cond_broadcast(&scrub_monitor->wakecond);
+ }
+ pthread_mutex_unlock(&scrub_monitor->wakelock);
+
+ return;
}
static uint32_t
-br_fsscan_calculate_delta (uint32_t times)
+br_fsscan_calculate_delta(uint32_t times)
{
- return times;
+ return times;
}
-#define BR_SCRUB_ONDEMAND (1)
-#define BR_SCRUB_MINUTE (60)
-#define BR_SCRUB_HOURLY (60 * 60)
-#define BR_SCRUB_DAILY (1 * 24 * 60 * 60)
-#define BR_SCRUB_WEEKLY (7 * 24 * 60 * 60)
-#define BR_SCRUB_BIWEEKLY (14 * 24 * 60 * 60)
-#define BR_SCRUB_MONTHLY (30 * 24 * 60 * 60)
+#define BR_SCRUB_ONDEMAND (1)
+#define BR_SCRUB_MINUTE (60)
+#define BR_SCRUB_HOURLY (60 * 60)
+#define BR_SCRUB_DAILY (1 * 24 * 60 * 60)
+#define BR_SCRUB_WEEKLY (7 * 24 * 60 * 60)
+#define BR_SCRUB_BIWEEKLY (14 * 24 * 60 * 60)
+#define BR_SCRUB_MONTHLY (30 * 24 * 60 * 60)
static unsigned int
-br_fsscan_calculate_timeout (scrub_freq_t freq)
+br_fsscan_calculate_timeout(scrub_freq_t freq)
{
- uint32_t timo = 0;
+ uint32_t timo = 0;
- switch (freq) {
+ switch (freq) {
case BR_FSSCRUB_FREQ_MINUTE:
- timo = br_fsscan_calculate_delta (BR_SCRUB_MINUTE);
- break;
+ timo = br_fsscan_calculate_delta(BR_SCRUB_MINUTE);
+ break;
case BR_FSSCRUB_FREQ_HOURLY:
- timo = br_fsscan_calculate_delta (BR_SCRUB_HOURLY);
- break;
+ timo = br_fsscan_calculate_delta(BR_SCRUB_HOURLY);
+ break;
case BR_FSSCRUB_FREQ_DAILY:
- timo = br_fsscan_calculate_delta (BR_SCRUB_DAILY);
- break;
+ timo = br_fsscan_calculate_delta(BR_SCRUB_DAILY);
+ break;
case BR_FSSCRUB_FREQ_WEEKLY:
- timo = br_fsscan_calculate_delta (BR_SCRUB_WEEKLY);
- break;
+ timo = br_fsscan_calculate_delta(BR_SCRUB_WEEKLY);
+ break;
case BR_FSSCRUB_FREQ_BIWEEKLY:
- timo = br_fsscan_calculate_delta (BR_SCRUB_BIWEEKLY);
- break;
+ timo = br_fsscan_calculate_delta(BR_SCRUB_BIWEEKLY);
+ break;
case BR_FSSCRUB_FREQ_MONTHLY:
- timo = br_fsscan_calculate_delta (BR_SCRUB_MONTHLY);
- break;
+ timo = br_fsscan_calculate_delta(BR_SCRUB_MONTHLY);
+ break;
default:
- timo = 0;
- }
+ timo = 0;
+ }
- return timo;
+ return timo;
}
int32_t
-br_fsscan_schedule (xlator_t *this)
+br_fsscan_schedule(xlator_t *this)
{
- uint32_t timo = 0;
- br_private_t *priv = NULL;
- struct timeval tv = {0,};
- char timestr[1024] = {0,};
- struct br_scrubber *fsscrub = NULL;
- struct gf_tw_timer_list *timer = NULL;
- struct br_monitor *scrub_monitor = NULL;
-
- priv = this->private;
- fsscrub = &priv->fsscrub;
- scrub_monitor = &priv->scrub_monitor;
-
- (void) gettimeofday (&tv, NULL);
- scrub_monitor->boot = tv.tv_sec;
-
- timo = br_fsscan_calculate_timeout (fsscrub->frequency);
- if (timo == 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_ZERO_TIMEOUT_BUG,
- "BUG: Zero schedule timeout");
- goto error_return;
- }
-
- scrub_monitor->timer = GF_CALLOC (1, sizeof (*scrub_monitor->timer),
- gf_br_stub_mt_br_scanner_freq_t);
- if (!scrub_monitor->timer)
- goto error_return;
-
- timer = scrub_monitor->timer;
- INIT_LIST_HEAD (&timer->entry);
-
- timer->data = scrub_monitor;
- timer->expires = timo;
- timer->function = br_kickstart_scanner;
-
- gf_tw_add_timer (priv->timer_wheel, timer);
- _br_monitor_set_scrub_state (scrub_monitor, BR_SCRUB_STATE_PENDING);
-
- gf_time_fmt (timestr, sizeof (timestr),
- (scrub_monitor->boot + timo), gf_timefmt_FT);
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, "Scrubbing is "
- "scheduled to run at %s", timestr);
-
- return 0;
-
- error_return:
- return -1;
+ uint32_t timo = 0;
+ br_private_t *priv = NULL;
+ char timestr[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ struct br_scrubber *fsscrub = NULL;
+ struct gf_tw_timer_list *timer = NULL;
+ struct br_monitor *scrub_monitor = NULL;
+
+ priv = this->private;
+ fsscrub = &priv->fsscrub;
+ scrub_monitor = &priv->scrub_monitor;
+
+ scrub_monitor->boot = gf_time();
+
+ timo = br_fsscan_calculate_timeout(fsscrub->frequency);
+ if (timo == 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_ZERO_TIMEOUT_BUG,
+ "BUG: Zero schedule timeout");
+ goto error_return;
+ }
+
+ scrub_monitor->timer = GF_CALLOC(1, sizeof(*scrub_monitor->timer),
+ gf_br_stub_mt_br_scanner_freq_t);
+ if (!scrub_monitor->timer)
+ goto error_return;
+
+ timer = scrub_monitor->timer;
+ INIT_LIST_HEAD(&timer->entry);
+
+ timer->data = scrub_monitor;
+ timer->expires = timo;
+ timer->function = br_kickstart_scanner;
+
+ gf_tw_add_timer(priv->timer_wheel, timer);
+ _br_monitor_set_scrub_state(scrub_monitor, BR_SCRUB_STATE_PENDING);
+
+ gf_time_fmt(timestr, sizeof(timestr), (scrub_monitor->boot + timo),
+ gf_timefmt_FT);
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
+ "Scrubbing is "
+ "scheduled to run at %s",
+ timestr);
+
+ return 0;
+
+error_return:
+ return -1;
}
int32_t
-br_fsscan_activate (xlator_t *this)
+br_fsscan_activate(xlator_t *this)
{
- uint32_t timo = 0;
- char timestr[1024] = {0,};
- struct timeval now = {0,};
- br_private_t *priv = NULL;
- struct br_scrubber *fsscrub = NULL;
- struct br_monitor *scrub_monitor = NULL;
-
- priv = this->private;
- fsscrub = &priv->fsscrub;
- scrub_monitor = &priv->scrub_monitor;
-
- (void) gettimeofday (&now, NULL);
- timo = br_fsscan_calculate_timeout (fsscrub->frequency);
- if (timo == 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_ZERO_TIMEOUT_BUG,
- "BUG: Zero schedule timeout");
- return -1;
- }
+ uint32_t timo = 0;
+ char timestr[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ time_t now = 0;
+ br_private_t *priv = NULL;
+ struct br_scrubber *fsscrub = NULL;
+ struct br_monitor *scrub_monitor = NULL;
+
+ priv = this->private;
+ fsscrub = &priv->fsscrub;
+ scrub_monitor = &priv->scrub_monitor;
+
+ now = gf_time();
+ timo = br_fsscan_calculate_timeout(fsscrub->frequency);
+ if (timo == 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_ZERO_TIMEOUT_BUG,
+ "BUG: Zero schedule timeout");
+ return -1;
+ }
- pthread_mutex_lock (&scrub_monitor->donelock);
- {
- scrub_monitor->done = _gf_false;
- }
- pthread_mutex_unlock (&scrub_monitor->donelock);
+ pthread_mutex_lock(&scrub_monitor->donelock);
+ {
+ scrub_monitor->done = _gf_false;
+ }
+ pthread_mutex_unlock(&scrub_monitor->donelock);
- gf_time_fmt (timestr, sizeof (timestr),
- (now.tv_sec + timo), gf_timefmt_FT);
- (void) gf_tw_mod_timer (priv->timer_wheel, scrub_monitor->timer, timo);
+ gf_time_fmt(timestr, sizeof(timestr), now + timo, gf_timefmt_FT);
+ (void)gf_tw_mod_timer(priv->timer_wheel, scrub_monitor->timer, timo);
- _br_monitor_set_scrub_state (scrub_monitor, BR_SCRUB_STATE_PENDING);
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO, "Scrubbing is "
- "rescheduled to run at %s", timestr);
+ _br_monitor_set_scrub_state(scrub_monitor, BR_SCRUB_STATE_PENDING);
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
+ "Scrubbing is "
+ "rescheduled to run at %s",
+ timestr);
- return 0;
+ return 0;
}
int32_t
-br_fsscan_reschedule (xlator_t *this)
+br_fsscan_reschedule(xlator_t *this)
{
- int32_t ret = 0;
- uint32_t timo = 0;
- char timestr[1024] = {0,};
- struct timeval now = {0,};
- br_private_t *priv = NULL;
- struct br_scrubber *fsscrub = NULL;
- struct br_monitor *scrub_monitor = NULL;
-
- priv = this->private;
- fsscrub = &priv->fsscrub;
- scrub_monitor = &priv->scrub_monitor;
-
- if (!fsscrub->frequency_reconf)
- return 0;
-
- (void) gettimeofday (&now, NULL);
- timo = br_fsscan_calculate_timeout (fsscrub->frequency);
- if (timo == 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_ZERO_TIMEOUT_BUG,
- "BUG: Zero schedule timeout");
- return -1;
- }
+ int32_t ret = 0;
+ uint32_t timo = 0;
+ char timestr[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ time_t now = 0;
+ br_private_t *priv = NULL;
+ struct br_scrubber *fsscrub = NULL;
+ struct br_monitor *scrub_monitor = NULL;
+
+ priv = this->private;
+ fsscrub = &priv->fsscrub;
+ scrub_monitor = &priv->scrub_monitor;
+
+ if (!fsscrub->frequency_reconf)
+ return 0;
- gf_time_fmt (timestr, sizeof (timestr),
- (now.tv_sec + timo), gf_timefmt_FT);
+ now = gf_time();
+ timo = br_fsscan_calculate_timeout(fsscrub->frequency);
+ if (timo == 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_ZERO_TIMEOUT_BUG,
+ "BUG: Zero schedule timeout");
+ return -1;
+ }
- pthread_mutex_lock (&scrub_monitor->donelock);
- {
- scrub_monitor->done = _gf_false;
- }
- pthread_mutex_unlock (&scrub_monitor->donelock);
-
- ret = gf_tw_mod_timer_pending (priv->timer_wheel, scrub_monitor->timer, timo);
- if (ret == 0)
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
- "Scrubber is currently running and would be "
- "rescheduled after completion");
- else {
- _br_monitor_set_scrub_state (scrub_monitor, BR_SCRUB_STATE_PENDING);
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
- "Scrubbing rescheduled to run at %s", timestr);
- }
+ gf_time_fmt(timestr, sizeof(timestr), now + timo, gf_timefmt_FT);
- return 0;
+ pthread_mutex_lock(&scrub_monitor->donelock);
+ {
+ scrub_monitor->done = _gf_false;
+ }
+ pthread_mutex_unlock(&scrub_monitor->donelock);
+
+ ret = gf_tw_mod_timer_pending(priv->timer_wheel, scrub_monitor->timer,
+ timo);
+ if (ret == 0)
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
+ "Scrubber is currently running and would be "
+ "rescheduled after completion");
+ else {
+ _br_monitor_set_scrub_state(scrub_monitor, BR_SCRUB_STATE_PENDING);
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
+ "Scrubbing rescheduled to run at %s", timestr);
+ }
+
+ return 0;
}
int32_t
-br_fsscan_ondemand (xlator_t *this)
+br_fsscan_ondemand(xlator_t *this)
{
- int32_t ret = 0;
- uint32_t timo = 0;
- char timestr[1024] = {0,};
- struct timeval now = {0,};
- br_private_t *priv = NULL;
- struct br_monitor *scrub_monitor = NULL;
-
- priv = this->private;
- scrub_monitor = &priv->scrub_monitor;
-
- (void) gettimeofday (&now, NULL);
-
- timo = BR_SCRUB_ONDEMAND;
-
- gf_time_fmt (timestr, sizeof (timestr),
- (now.tv_sec + timo), gf_timefmt_FT);
-
- pthread_mutex_lock (&scrub_monitor->donelock);
- {
- scrub_monitor->done = _gf_false;
- }
- pthread_mutex_unlock (&scrub_monitor->donelock);
-
- ret = gf_tw_mod_timer_pending (priv->timer_wheel, scrub_monitor->timer,
- timo);
- if (ret == 0)
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
- "Scrubber is currently running and would be "
- "rescheduled after completion");
- else {
- _br_monitor_set_scrub_state (scrub_monitor,
- BR_SCRUB_STATE_PENDING);
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
- "Ondemand Scrubbing scheduled to run at %s", timestr);
- }
-
- return 0;
+ int32_t ret = 0;
+ uint32_t timo = 0;
+ char timestr[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ time_t now = 0;
+ br_private_t *priv = NULL;
+ struct br_monitor *scrub_monitor = NULL;
+
+ priv = this->private;
+ scrub_monitor = &priv->scrub_monitor;
+
+ now = gf_time();
+ timo = BR_SCRUB_ONDEMAND;
+ gf_time_fmt(timestr, sizeof(timestr), now + timo, gf_timefmt_FT);
+
+ pthread_mutex_lock(&scrub_monitor->donelock);
+ {
+ scrub_monitor->done = _gf_false;
+ }
+ pthread_mutex_unlock(&scrub_monitor->donelock);
+
+ ret = gf_tw_mod_timer_pending(priv->timer_wheel, scrub_monitor->timer,
+ timo);
+ if (ret == 0)
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
+ "Scrubber is currently running and would be "
+ "rescheduled after completion");
+ else {
+ _br_monitor_set_scrub_state(scrub_monitor, BR_SCRUB_STATE_PENDING);
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
+ "Ondemand Scrubbing scheduled to run at %s", timestr);
+ }
+
+ return 0;
}
-#define BR_SCRUB_THREAD_SCALE_LAZY 0
-#define BR_SCRUB_THREAD_SCALE_NORMAL 0.4
+#define BR_SCRUB_THREAD_SCALE_LAZY 0
+#define BR_SCRUB_THREAD_SCALE_NORMAL 0.4
#define BR_SCRUB_THREAD_SCALE_AGGRESSIVE 1.0
#ifndef M_E
@@ -1097,111 +1110,105 @@ br_fsscan_ondemand (xlator_t *this)
* the scale based on the number of processor cores too.
*/
static unsigned int
-br_scrubber_calc_scale (xlator_t *this,
- br_private_t *priv, scrub_throttle_t throttle)
+br_scrubber_calc_scale(xlator_t *this, br_private_t *priv,
+ scrub_throttle_t throttle)
{
- unsigned int scale = 0;
+ unsigned int scale = 0;
- switch (throttle) {
+ switch (throttle) {
case BR_SCRUB_THROTTLE_VOID:
case BR_SCRUB_THROTTLE_STALLED:
- scale = 0;
- break;
+ scale = 0;
+ break;
case BR_SCRUB_THROTTLE_LAZY:
- scale = priv->child_count *
- pow (M_E, BR_SCRUB_THREAD_SCALE_LAZY);
- break;
+ scale = priv->child_count * pow(M_E, BR_SCRUB_THREAD_SCALE_LAZY);
+ break;
case BR_SCRUB_THROTTLE_NORMAL:
- scale = priv->child_count *
- pow (M_E, BR_SCRUB_THREAD_SCALE_NORMAL);
- break;
+ scale = priv->child_count * pow(M_E, BR_SCRUB_THREAD_SCALE_NORMAL);
+ break;
case BR_SCRUB_THROTTLE_AGGRESSIVE:
- scale = priv->child_count *
- pow (M_E, BR_SCRUB_THREAD_SCALE_AGGRESSIVE);
- break;
+ scale = priv->child_count *
+ pow(M_E, BR_SCRUB_THREAD_SCALE_AGGRESSIVE);
+ break;
default:
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_UNKNOWN_THROTTLE,
- "Unknown throttle %d", throttle);
- }
-
- return scale;
+ gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_UNKNOWN_THROTTLE,
+ "Unknown throttle %d", throttle);
+ }
+ return scale;
}
static br_child_t *
-_br_scrubber_get_next_child (struct br_scrubber *fsscrub)
+_br_scrubber_get_next_child(struct br_scrubber *fsscrub)
{
- br_child_t *child = NULL;
+ br_child_t *child = NULL;
- child = list_first_entry (&fsscrub->scrublist, br_child_t, list);
- list_rotate_left (&fsscrub->scrublist);
+ child = list_first_entry(&fsscrub->scrublist, br_child_t, list);
+ list_rotate_left(&fsscrub->scrublist);
- return child;
+ return child;
}
static void
-_br_scrubber_get_entry (br_child_t *child, struct br_fsscan_entry **fsentry)
+_br_scrubber_get_entry(br_child_t *child, struct br_fsscan_entry **fsentry)
{
- struct br_scanfs *fsscan = &child->fsscan;
+ struct br_scanfs *fsscan = &child->fsscan;
- if (list_empty (&fsscan->ready))
- return;
- *fsentry = list_first_entry
- (&fsscan->ready, struct br_fsscan_entry, list);
- list_del_init (&(*fsentry)->list);
+ if (list_empty(&fsscan->ready))
+ return;
+ *fsentry = list_first_entry(&fsscan->ready, struct br_fsscan_entry, list);
+ list_del_init(&(*fsentry)->list);
}
static void
-_br_scrubber_find_scrubbable_entry (struct br_scrubber *fsscrub,
- struct br_fsscan_entry **fsentry)
+_br_scrubber_find_scrubbable_entry(struct br_scrubber *fsscrub,
+ struct br_fsscan_entry **fsentry)
{
- br_child_t *child = NULL;
- br_child_t *firstchild = NULL;
+ br_child_t *child = NULL;
+ br_child_t *firstchild = NULL;
- while (1) {
- while (list_empty (&fsscrub->scrublist))
- pthread_cond_wait (&fsscrub->cond, &fsscrub->mutex);
+ while (1) {
+ while (list_empty(&fsscrub->scrublist))
+ pthread_cond_wait(&fsscrub->cond, &fsscrub->mutex);
- firstchild = NULL;
- for (child = _br_scrubber_get_next_child (fsscrub);
- child != firstchild;
- child = _br_scrubber_get_next_child (fsscrub)) {
+ firstchild = NULL;
+ for (child = _br_scrubber_get_next_child(fsscrub); child != firstchild;
+ child = _br_scrubber_get_next_child(fsscrub)) {
+ if (!firstchild)
+ firstchild = child;
- if (!firstchild)
- firstchild = child;
-
- _br_scrubber_get_entry (child, fsentry);
- if (*fsentry)
- break;
- }
+ _br_scrubber_get_entry(child, fsentry);
+ if (*fsentry)
+ break;
+ }
- if (*fsentry)
- break;
+ if (*fsentry)
+ break;
- /* nothing to work on.. wait till available */
- pthread_cond_wait (&fsscrub->cond, &fsscrub->mutex);
- }
+ /* nothing to work on.. wait till available */
+ pthread_cond_wait(&fsscrub->cond, &fsscrub->mutex);
+ }
}
static void
-br_scrubber_pick_entry (struct br_scrubber *fsscrub,
- struct br_fsscan_entry **fsentry)
+br_scrubber_pick_entry(struct br_scrubber *fsscrub,
+ struct br_fsscan_entry **fsentry)
{
- pthread_cleanup_push (_br_lock_cleaner, &fsscrub->mutex);
+ pthread_cleanup_push(_br_lock_cleaner, &fsscrub->mutex);
- pthread_mutex_lock (&fsscrub->mutex);
- {
- *fsentry = NULL;
- _br_scrubber_find_scrubbable_entry (fsscrub, fsentry);
- }
- pthread_mutex_unlock (&fsscrub->mutex);
+ pthread_mutex_lock(&fsscrub->mutex);
+ {
+ *fsentry = NULL;
+ _br_scrubber_find_scrubbable_entry(fsscrub, fsentry);
+ }
+ pthread_mutex_unlock(&fsscrub->mutex);
- pthread_cleanup_pop (0);
+ pthread_cleanup_pop(0);
}
struct br_scrub_entry {
- gf_boolean_t scrubbed;
- struct br_fsscan_entry *fsentry;
+ gf_boolean_t scrubbed;
+ struct br_fsscan_entry *fsentry;
};
/**
@@ -1211,682 +1218,702 @@ struct br_scrub_entry {
* in the ->pending queue or when an object is undergoing scrubbing.
*/
static void
-br_scrubber_entry_handle (void *arg)
+br_scrubber_entry_handle(void *arg)
{
- struct br_scanfs *fsscan = NULL;
- struct br_scrub_entry *sentry = NULL;
- struct br_fsscan_entry *fsentry = NULL;
+ struct br_scanfs *fsscan = NULL;
+ struct br_scrub_entry *sentry = NULL;
+ struct br_fsscan_entry *fsentry = NULL;
- sentry = arg;
+ sentry = arg;
- fsentry = sentry->fsentry;
- fsscan = fsentry->fsscan;
+ fsentry = sentry->fsentry;
+ fsscan = fsentry->fsscan;
- LOCK (&fsscan->entrylock);
- {
- if (sentry->scrubbed) {
- _br_fsscan_dec_entry_count (fsscan);
-
- /* cleanup ->entry */
- fsentry->data = NULL;
- fsentry->fsscan = NULL;
- loc_wipe (&fsentry->parent);
- gf_dirent_entry_free (fsentry->entry);
-
- GF_FREE (sentry->fsentry);
- } else {
- /* (re)queue the entry again for scrub */
- _br_fsscan_collect_entry (fsscan, sentry->fsentry);
- }
+ LOCK(&fsscan->entrylock);
+ {
+ if (sentry->scrubbed) {
+ _br_fsscan_dec_entry_count(fsscan);
+
+ /* cleanup ->entry */
+ fsentry->data = NULL;
+ fsentry->fsscan = NULL;
+ loc_wipe(&fsentry->parent);
+ gf_dirent_entry_free(fsentry->entry);
+
+ GF_FREE(sentry->fsentry);
+ } else {
+ /* (re)queue the entry again for scrub */
+ _br_fsscan_collect_entry(fsscan, sentry->fsentry);
}
- UNLOCK (&fsscan->entrylock);
+ }
+ UNLOCK(&fsscan->entrylock);
}
static void
-br_scrubber_scrub_entry (xlator_t *this, struct br_fsscan_entry *fsentry)
+br_scrubber_scrub_entry(xlator_t *this, struct br_fsscan_entry *fsentry)
{
- struct br_scrub_entry sentry = {0, };
-
- sentry.scrubbed = 0;
- sentry.fsentry = fsentry;
-
- pthread_cleanup_push (br_scrubber_entry_handle, &sentry);
- {
- (void) br_scrubber_scrub_begin (this, fsentry);
- sentry.scrubbed = 1;
- }
- pthread_cleanup_pop (1);
+ struct br_scrub_entry sentry = {
+ 0,
+ };
+
+ sentry.scrubbed = 0;
+ sentry.fsentry = fsentry;
+
+ pthread_cleanup_push(br_scrubber_entry_handle, &sentry);
+ {
+ (void)br_scrubber_scrub_begin(this, fsentry);
+ sentry.scrubbed = 1;
+ }
+ pthread_cleanup_pop(1);
}
-void *br_scrubber_proc (void *arg)
+void *
+br_scrubber_proc(void *arg)
{
- xlator_t *this = NULL;
- struct br_scrubber *fsscrub = NULL;
- struct br_fsscan_entry *fsentry = NULL;
+ xlator_t *this = NULL;
+ struct br_scrubber *fsscrub = NULL;
+ struct br_fsscan_entry *fsentry = NULL;
- fsscrub = arg;
- THIS = this = fsscrub->this;
+ fsscrub = arg;
+ THIS = this = fsscrub->this;
- while (1) {
- br_scrubber_pick_entry (fsscrub, &fsentry);
- br_scrubber_scrub_entry (this, fsentry);
- sleep (1);
- }
+ while (1) {
+ br_scrubber_pick_entry(fsscrub, &fsentry);
+ br_scrubber_scrub_entry(this, fsentry);
+ sleep(1);
+ }
- return NULL;
+ return NULL;
}
static int32_t
-br_scrubber_scale_up (xlator_t *this,
- struct br_scrubber *fsscrub,
- unsigned int v1, unsigned int v2)
+br_scrubber_scale_up(xlator_t *this, struct br_scrubber *fsscrub,
+ unsigned int v1, unsigned int v2)
{
- int i = 0;
- int32_t ret = -1;
- int diff = 0;
- struct br_scrubbers *scrub = NULL;
-
- diff = (int)(v2 - v1);
-
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCALING_UP_SCRUBBER,
- "Scaling up scrubbers [%d => %d]", v1, v2);
-
- for (i = 0; i < diff; i++) {
- scrub = GF_CALLOC (diff, sizeof (*scrub),
- gf_br_mt_br_scrubber_t);
- if (!scrub)
- break;
-
- INIT_LIST_HEAD (&scrub->list);
- ret = gf_thread_create (&scrub->scrubthread,
- NULL, br_scrubber_proc, fsscrub,
- "brsproc");
- if (ret)
- break;
-
- fsscrub->nr_scrubbers++;
- list_add_tail (&scrub->list, &fsscrub->scrubbers);
- }
+ int i = 0;
+ int32_t ret = -1;
+ int diff = 0;
+ struct br_scrubbers *scrub = NULL;
- if ((i != diff) && !scrub)
- goto error_return;
+ diff = (int)(v2 - v1);
- if (i != diff) /* degraded scaling.. */
- gf_msg (this->name, GF_LOG_WARNING, 0, BRB_MSG_SCALE_UP_FAILED,
- "Could not fully scale up to %d scrubber(s). Spawned "
- "%d/%d [total scrubber(s): %d]", v2, i, diff, (v1 + i));
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCALING_UP_SCRUBBER,
+ "Scaling up scrubbers [%d => %d]", v1, v2);
- return 0;
+ for (i = 0; i < diff; i++) {
+ scrub = GF_CALLOC(diff, sizeof(*scrub), gf_br_mt_br_scrubber_t);
+ if (!scrub)
+ break;
- error_return:
- return -1;
+ INIT_LIST_HEAD(&scrub->list);
+ ret = gf_thread_create(&scrub->scrubthread, NULL, br_scrubber_proc,
+ fsscrub, "brsproc");
+ if (ret)
+ break;
+
+ fsscrub->nr_scrubbers++;
+ list_add_tail(&scrub->list, &fsscrub->scrubbers);
+ }
+
+ if ((i != diff) && !scrub)
+ goto error_return;
+
+ if (i != diff) /* degraded scaling.. */
+ gf_msg(this->name, GF_LOG_WARNING, 0, BRB_MSG_SCALE_UP_FAILED,
+ "Could not fully scale up to %d scrubber(s). Spawned "
+ "%d/%d [total scrubber(s): %d]",
+ v2, i, diff, (v1 + i));
+
+ return 0;
+
+error_return:
+ return -1;
}
static int32_t
-br_scrubber_scale_down (xlator_t *this,
- struct br_scrubber *fsscrub,
- unsigned int v1, unsigned int v2)
+br_scrubber_scale_down(xlator_t *this, struct br_scrubber *fsscrub,
+ unsigned int v1, unsigned int v2)
{
- int i = 0;
- int diff = 0;
- int32_t ret = -1;
- struct br_scrubbers *scrub = NULL;
-
- diff = (int)(v1 - v2);
-
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCALE_DOWN_SCRUBBER,
- "Scaling down scrubbers [%d => %d]", v1, v2);
+ int i = 0;
+ int diff = 0;
+ int32_t ret = -1;
+ struct br_scrubbers *scrub = NULL;
- for (i = 0 ; i < diff; i++) {
- scrub = list_first_entry
- (&fsscrub->scrubbers, struct br_scrubbers, list);
+ diff = (int)(v1 - v2);
- list_del_init (&scrub->list);
- ret = gf_thread_cleanup_xint (scrub->scrubthread);
- if (ret)
- break;
- GF_FREE (scrub);
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCALE_DOWN_SCRUBBER,
+ "Scaling down scrubbers [%d => %d]", v1, v2);
- fsscrub->nr_scrubbers--;
- }
+ for (i = 0; i < diff; i++) {
+ scrub = list_first_entry(&fsscrub->scrubbers, struct br_scrubbers,
+ list);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- BRB_MSG_SCALE_DOWN_FAILED, "Could not fully scale down "
- "to %d scrubber(s). Terminated %d/%d [total "
- "scrubber(s): %d]", v1, i, diff, (v2 - i));
- ret = 0;
- }
+ list_del_init(&scrub->list);
+ ret = gf_thread_cleanup_xint(scrub->scrubthread);
+ if (ret)
+ break;
+ GF_FREE(scrub);
+
+ fsscrub->nr_scrubbers--;
+ }
+
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, BRB_MSG_SCALE_DOWN_FAILED,
+ "Could not fully scale down "
+ "to %d scrubber(s). Terminated %d/%d [total "
+ "scrubber(s): %d]",
+ v1, i, diff, (v2 - i));
+ ret = 0;
+ }
- return ret;
+ return ret;
}
static int32_t
-br_scrubber_configure (xlator_t *this, br_private_t *priv,
- struct br_scrubber *fsscrub, scrub_throttle_t nthrottle)
+br_scrubber_configure(xlator_t *this, br_private_t *priv,
+ struct br_scrubber *fsscrub, scrub_throttle_t nthrottle)
{
- int32_t ret = 0;
- unsigned int v1 = 0;
- unsigned int v2 = 0;
+ int32_t ret = 0;
+ unsigned int v1 = 0;
+ unsigned int v2 = 0;
- v1 = fsscrub->nr_scrubbers;
- v2 = br_scrubber_calc_scale (this, priv, nthrottle);
+ v1 = fsscrub->nr_scrubbers;
+ v2 = br_scrubber_calc_scale(this, priv, nthrottle);
- if (v1 == v2)
- return 0;
+ if (v1 == v2)
+ return 0;
- if (v1 > v2)
- ret = br_scrubber_scale_down (this, fsscrub, v1, v2);
- else
- ret = br_scrubber_scale_up (this, fsscrub, v1, v2);
+ if (v1 > v2)
+ ret = br_scrubber_scale_down(this, fsscrub, v1, v2);
+ else
+ ret = br_scrubber_scale_up(this, fsscrub, v1, v2);
- return ret;
+ return ret;
}
static int32_t
-br_scrubber_fetch_option (xlator_t *this,
- char *opt, dict_t *options, char **value)
+br_scrubber_fetch_option(xlator_t *this, char *opt, dict_t *options,
+ char **value)
{
- if (options)
- GF_OPTION_RECONF (opt, *value, options, str, error_return);
- else
- GF_OPTION_INIT (opt, *value, str, error_return);
+ if (options)
+ GF_OPTION_RECONF(opt, *value, options, str, error_return);
+ else
+ GF_OPTION_INIT(opt, *value, str, error_return);
- return 0;
+ return 0;
- error_return:
- return -1;
+error_return:
+ return -1;
}
/* internal "throttle" override */
-#define BR_SCRUB_STALLED "STALLED"
+#define BR_SCRUB_STALLED "STALLED"
/* TODO: token buket spec */
static int32_t
-br_scrubber_handle_throttle (xlator_t *this, br_private_t *priv,
- dict_t *options, gf_boolean_t scrubstall)
+br_scrubber_handle_throttle(xlator_t *this, br_private_t *priv, dict_t *options,
+ gf_boolean_t scrubstall)
{
- int32_t ret = 0;
- char *tmp = NULL;
- struct br_scrubber *fsscrub = NULL;
- scrub_throttle_t nthrottle = BR_SCRUB_THROTTLE_VOID;
-
- fsscrub = &priv->fsscrub;
- fsscrub->throttle_reconf = _gf_false;
-
- ret = br_scrubber_fetch_option (this, "scrub-throttle", options, &tmp);
- if (ret)
- goto error_return;
-
- if (scrubstall)
- tmp = BR_SCRUB_STALLED;
-
- if (strcasecmp (tmp, "lazy") == 0)
- nthrottle = BR_SCRUB_THROTTLE_LAZY;
- else if (strcasecmp (tmp, "normal") == 0)
- nthrottle = BR_SCRUB_THROTTLE_NORMAL;
- else if (strcasecmp (tmp, "aggressive") == 0)
- nthrottle = BR_SCRUB_THROTTLE_AGGRESSIVE;
- else if (strcasecmp (tmp, BR_SCRUB_STALLED) == 0)
- nthrottle = BR_SCRUB_THROTTLE_STALLED;
- else
- goto error_return;
-
- /* on failure old throttling value is preserved */
- ret = br_scrubber_configure (this, priv, fsscrub, nthrottle);
- if (ret)
- goto error_return;
-
- if (fsscrub->throttle != nthrottle)
- fsscrub->throttle_reconf = _gf_true;
-
- fsscrub->throttle = nthrottle;
- return 0;
-
- error_return:
- return -1;
+ int32_t ret = 0;
+ char *tmp = NULL;
+ struct br_scrubber *fsscrub = NULL;
+ scrub_throttle_t nthrottle = BR_SCRUB_THROTTLE_VOID;
+
+ fsscrub = &priv->fsscrub;
+ fsscrub->throttle_reconf = _gf_false;
+
+ ret = br_scrubber_fetch_option(this, "scrub-throttle", options, &tmp);
+ if (ret)
+ goto error_return;
+
+ if (scrubstall)
+ tmp = BR_SCRUB_STALLED;
+
+ if (strcasecmp(tmp, "lazy") == 0)
+ nthrottle = BR_SCRUB_THROTTLE_LAZY;
+ else if (strcasecmp(tmp, "normal") == 0)
+ nthrottle = BR_SCRUB_THROTTLE_NORMAL;
+ else if (strcasecmp(tmp, "aggressive") == 0)
+ nthrottle = BR_SCRUB_THROTTLE_AGGRESSIVE;
+ else if (strcasecmp(tmp, BR_SCRUB_STALLED) == 0)
+ nthrottle = BR_SCRUB_THROTTLE_STALLED;
+ else
+ goto error_return;
+
+ /* on failure old throttling value is preserved */
+ ret = br_scrubber_configure(this, priv, fsscrub, nthrottle);
+ if (ret)
+ goto error_return;
+
+ if (fsscrub->throttle != nthrottle)
+ fsscrub->throttle_reconf = _gf_true;
+
+ fsscrub->throttle = nthrottle;
+ return 0;
+
+error_return:
+ return -1;
}
static int32_t
-br_scrubber_handle_stall (xlator_t *this, br_private_t *priv,
- dict_t *options, gf_boolean_t *scrubstall)
+br_scrubber_handle_stall(xlator_t *this, br_private_t *priv, dict_t *options,
+ gf_boolean_t *scrubstall)
{
- int32_t ret = 0;
- char *tmp = NULL;
+ int32_t ret = 0;
+ char *tmp = NULL;
- ret = br_scrubber_fetch_option (this, "scrub-state", options, &tmp);
- if (ret)
- goto error_return;
+ ret = br_scrubber_fetch_option(this, "scrub-state", options, &tmp);
+ if (ret)
+ goto error_return;
- if (strcasecmp (tmp, "pause") == 0) /* anything else is active */
- *scrubstall = _gf_true;
+ if (strcasecmp(tmp, "pause") == 0) /* anything else is active */
+ *scrubstall = _gf_true;
- return 0;
+ return 0;
- error_return:
- return -1;
+error_return:
+ return -1;
}
static int32_t
-br_scrubber_handle_freq (xlator_t *this, br_private_t *priv,
- dict_t *options, gf_boolean_t scrubstall)
+br_scrubber_handle_freq(xlator_t *this, br_private_t *priv, dict_t *options,
+ gf_boolean_t scrubstall)
{
- int32_t ret = -1;
- char *tmp = NULL;
- scrub_freq_t frequency = BR_FSSCRUB_FREQ_HOURLY;
- struct br_scrubber *fsscrub = NULL;
-
- fsscrub = &priv->fsscrub;
- fsscrub->frequency_reconf = _gf_true;
-
- ret = br_scrubber_fetch_option (this, "scrub-freq", options, &tmp);
- if (ret)
- goto error_return;
-
- if (scrubstall)
- tmp = BR_SCRUB_STALLED;
-
- if (strcasecmp (tmp, "hourly") == 0) {
- frequency = BR_FSSCRUB_FREQ_HOURLY;
- } else if (strcasecmp (tmp, "daily") == 0) {
- frequency = BR_FSSCRUB_FREQ_DAILY;
- } else if (strcasecmp (tmp, "weekly") == 0) {
- frequency = BR_FSSCRUB_FREQ_WEEKLY;
- } else if (strcasecmp (tmp, "biweekly") == 0) {
- frequency = BR_FSSCRUB_FREQ_BIWEEKLY;
- } else if (strcasecmp (tmp, "monthly") == 0) {
- frequency = BR_FSSCRUB_FREQ_MONTHLY;
- } else if (strcasecmp (tmp, "minute") == 0) {
- frequency = BR_FSSCRUB_FREQ_MINUTE;
- } else if (strcasecmp (tmp, BR_SCRUB_STALLED) == 0) {
- frequency = BR_FSSCRUB_FREQ_STALLED;
- } else
- goto error_return;
-
- if (fsscrub->frequency == frequency)
- fsscrub->frequency_reconf = _gf_false;
- else
- fsscrub->frequency = frequency;
-
- return 0;
-
- error_return:
- return -1;
+ int32_t ret = -1;
+ char *tmp = NULL;
+ scrub_freq_t frequency = BR_FSSCRUB_FREQ_HOURLY;
+ struct br_scrubber *fsscrub = NULL;
+
+ fsscrub = &priv->fsscrub;
+ fsscrub->frequency_reconf = _gf_true;
+
+ ret = br_scrubber_fetch_option(this, "scrub-freq", options, &tmp);
+ if (ret)
+ goto error_return;
+
+ if (scrubstall)
+ tmp = BR_SCRUB_STALLED;
+
+ if (strcasecmp(tmp, "hourly") == 0) {
+ frequency = BR_FSSCRUB_FREQ_HOURLY;
+ } else if (strcasecmp(tmp, "daily") == 0) {
+ frequency = BR_FSSCRUB_FREQ_DAILY;
+ } else if (strcasecmp(tmp, "weekly") == 0) {
+ frequency = BR_FSSCRUB_FREQ_WEEKLY;
+ } else if (strcasecmp(tmp, "biweekly") == 0) {
+ frequency = BR_FSSCRUB_FREQ_BIWEEKLY;
+ } else if (strcasecmp(tmp, "monthly") == 0) {
+ frequency = BR_FSSCRUB_FREQ_MONTHLY;
+ } else if (strcasecmp(tmp, "minute") == 0) {
+ frequency = BR_FSSCRUB_FREQ_MINUTE;
+ } else if (strcasecmp(tmp, BR_SCRUB_STALLED) == 0) {
+ frequency = BR_FSSCRUB_FREQ_STALLED;
+ } else
+ goto error_return;
+
+ if (fsscrub->frequency == frequency)
+ fsscrub->frequency_reconf = _gf_false;
+ else
+ fsscrub->frequency = frequency;
+
+ return 0;
+
+error_return:
+ return -1;
}
-static void br_scrubber_log_option (xlator_t *this,
- br_private_t *priv, gf_boolean_t scrubstall)
+static void
+br_scrubber_log_option(xlator_t *this, br_private_t *priv,
+ gf_boolean_t scrubstall)
{
- struct br_scrubber *fsscrub = &priv->fsscrub;
- char *scrub_throttle_str[] = {
- [BR_SCRUB_THROTTLE_LAZY] = "lazy",
- [BR_SCRUB_THROTTLE_NORMAL] = "normal",
- [BR_SCRUB_THROTTLE_AGGRESSIVE] = "aggressive",
- };
-
- char *scrub_freq_str[] = {
- [BR_FSSCRUB_FREQ_HOURLY] = "hourly",
- [BR_FSSCRUB_FREQ_DAILY] = "daily",
- [BR_FSSCRUB_FREQ_WEEKLY] = "weekly",
- [BR_FSSCRUB_FREQ_BIWEEKLY] = "biweekly",
- [BR_FSSCRUB_FREQ_MONTHLY] = "monthly (30 days)",
- [BR_FSSCRUB_FREQ_MINUTE] = "every minute",
- };
-
- if (scrubstall)
- return; /* logged as pause */
-
- if (fsscrub->frequency_reconf || fsscrub->throttle_reconf) {
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_TUNABLE,
- "SCRUB TUNABLES:: [Frequency: %s, Throttle: %s]",
- scrub_freq_str[fsscrub->frequency],
- scrub_throttle_str[fsscrub->throttle]);
- }
+ struct br_scrubber *fsscrub = &priv->fsscrub;
+ char *scrub_throttle_str[] = {
+ [BR_SCRUB_THROTTLE_LAZY] = "lazy",
+ [BR_SCRUB_THROTTLE_NORMAL] = "normal",
+ [BR_SCRUB_THROTTLE_AGGRESSIVE] = "aggressive",
+ [BR_SCRUB_THROTTLE_STALLED] = "stalled",
+ };
+
+ char *scrub_freq_str[] = {
+ [0] = "",
+ [BR_FSSCRUB_FREQ_HOURLY] = "hourly",
+ [BR_FSSCRUB_FREQ_DAILY] = "daily",
+ [BR_FSSCRUB_FREQ_WEEKLY] = "weekly",
+ [BR_FSSCRUB_FREQ_BIWEEKLY] = "biweekly",
+ [BR_FSSCRUB_FREQ_MONTHLY] = "monthly (30 days)",
+ [BR_FSSCRUB_FREQ_MINUTE] = "every minute",
+ };
+
+ if (scrubstall)
+ return; /* logged as pause */
+
+ if (fsscrub->frequency_reconf || fsscrub->throttle_reconf) {
+ if (fsscrub->throttle == BR_SCRUB_THROTTLE_VOID)
+ return;
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_TUNABLE,
+ "SCRUB TUNABLES:: [Frequency: %s, Throttle: %s]",
+ scrub_freq_str[fsscrub->frequency],
+ scrub_throttle_str[fsscrub->throttle]);
+ }
}
int32_t
-br_scrubber_handle_options (xlator_t *this, br_private_t *priv, dict_t *options)
+br_scrubber_handle_options(xlator_t *this, br_private_t *priv, dict_t *options)
{
- int32_t ret = 0;
- gf_boolean_t scrubstall = _gf_false; /* not as dangerous as it sounds */
+ int32_t ret = 0;
+ gf_boolean_t scrubstall = _gf_false; /* not as dangerous as it sounds */
- ret = br_scrubber_handle_stall (this, priv, options, &scrubstall);
- if (ret)
- goto error_return;
+ ret = br_scrubber_handle_stall(this, priv, options, &scrubstall);
+ if (ret)
+ goto error_return;
- ret = br_scrubber_handle_throttle (this, priv, options, scrubstall);
- if (ret)
- goto error_return;
+ ret = br_scrubber_handle_throttle(this, priv, options, scrubstall);
+ if (ret)
+ goto error_return;
- ret = br_scrubber_handle_freq (this, priv, options, scrubstall);
- if (ret)
- goto error_return;
+ ret = br_scrubber_handle_freq(this, priv, options, scrubstall);
+ if (ret)
+ goto error_return;
- br_scrubber_log_option (this, priv, scrubstall);
+ br_scrubber_log_option(this, priv, scrubstall);
- return 0;
+ return 0;
- error_return:
- return -1;
+error_return:
+ return -1;
}
inode_t *
-br_lookup_bad_obj_dir (xlator_t *this, br_child_t *child, uuid_t gfid)
+br_lookup_bad_obj_dir(xlator_t *this, br_child_t *child, uuid_t gfid)
{
- struct iatt statbuf = {0, };
- inode_table_t *table = NULL;
- int32_t ret = -1;
- loc_t loc = {0, };
- inode_t *linked_inode = NULL;
- int32_t op_errno = 0;
-
- GF_VALIDATE_OR_GOTO ("bit-rot-scrubber", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
- GF_VALIDATE_OR_GOTO (this->name, child, out);
-
- table = child->table;
-
- loc.inode = inode_new (table);
- if (!loc.inode) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- BRB_MSG_NO_MEMORY, "failed to allocate a new inode for"
- "bad object directory");
- goto out;
- }
-
- gf_uuid_copy (loc.gfid, gfid);
-
- ret = syncop_lookup (child->xl, &loc, &statbuf, NULL, NULL, NULL);
- if (ret < 0) {
- op_errno = -ret;
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRB_MSG_LOOKUP_FAILED, "failed to lookup the bad "
- "objects directory (gfid: %s (%s))", uuid_utoa (gfid),
- strerror (op_errno));
- goto out;
- }
-
- linked_inode = inode_link (loc.inode, NULL, NULL, &statbuf);
- if (linked_inode)
- inode_lookup (linked_inode);
+ struct iatt statbuf = {
+ 0,
+ };
+ inode_table_t *table = NULL;
+ int32_t ret = -1;
+ loc_t loc = {
+ 0,
+ };
+ inode_t *linked_inode = NULL;
+ int32_t op_errno = 0;
+
+ GF_VALIDATE_OR_GOTO("bit-rot-scrubber", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
+ GF_VALIDATE_OR_GOTO(this->name, child, out);
+
+ table = child->table;
+
+ loc.inode = inode_new(table);
+ if (!loc.inode) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY,
+ "failed to allocate a new inode for"
+ "bad object directory");
+ goto out;
+ }
+
+ gf_uuid_copy(loc.gfid, gfid);
+
+ ret = syncop_lookup(child->xl, &loc, &statbuf, NULL, NULL, NULL);
+ if (ret < 0) {
+ op_errno = -ret;
+ gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_LOOKUP_FAILED,
+ "failed to lookup the bad "
+ "objects directory (gfid: %s (%s))",
+ uuid_utoa(gfid), strerror(op_errno));
+ goto out;
+ }
+
+ linked_inode = inode_link(loc.inode, NULL, NULL, &statbuf);
+ if (linked_inode)
+ inode_lookup(linked_inode);
out:
- loc_wipe (&loc);
- return linked_inode;
+ loc_wipe(&loc);
+ return linked_inode;
}
int32_t
-br_read_bad_object_dir (xlator_t *this, br_child_t *child, fd_t *fd,
- dict_t *dict)
+br_read_bad_object_dir(xlator_t *this, br_child_t *child, fd_t *fd,
+ dict_t *dict)
{
- gf_dirent_t entries;
- gf_dirent_t *entry = NULL;
- int32_t ret = -1;
- off_t offset = 0;
- int32_t count = 0;
- char key[PATH_MAX] = {0, };
- dict_t *out_dict = NULL;
-
- INIT_LIST_HEAD (&entries.list);
-
- while ((ret = syncop_readdir (child->xl, fd, 131072, offset, &entries,
- NULL, &out_dict))) {
- if (ret < 0)
- goto out;
-
- list_for_each_entry (entry, &entries.list, list) {
- offset = entry->d_off;
-
- snprintf (key, sizeof (key), "quarantine-%d", count);
-
- /*
- * ignore the dict_set errors for now. The intention is
- * to get as many bad objects as possible instead of
- * erroring out at the first failure.
- */
- ret = dict_set_dynstr_with_alloc (dict, key,
- entry->d_name);
- if (!ret)
- count++;
-
- if (out_dict) {
- dict_copy (out_dict, dict);
- dict_unref (out_dict);
- out_dict = NULL;
- }
- }
-
- gf_dirent_free (&entries);
- }
-
- ret = count;
- ret = dict_set_int32 (dict, "count", count);
+ gf_dirent_t entries;
+ gf_dirent_t *entry = NULL;
+ int32_t ret = -1;
+ off_t offset = 0;
+ int32_t count = 0;
+ char key[32] = {
+ 0,
+ };
+ dict_t *out_dict = NULL;
+
+ INIT_LIST_HEAD(&entries.list);
+
+ while ((ret = syncop_readdir(child->xl, fd, 131072, offset, &entries, NULL,
+ &out_dict))) {
+ if (ret < 0)
+ goto out;
+
+ list_for_each_entry(entry, &entries.list, list)
+ {
+ offset = entry->d_off;
-out:
- return ret;
-}
+ snprintf(key, sizeof(key), "quarantine-%d", count);
-int32_t
-br_get_bad_objects_from_child (xlator_t *this, dict_t *dict, br_child_t *child)
-{
- inode_t *inode = NULL;
- inode_table_t *table = NULL;
- fd_t *fd = NULL;
- int32_t ret = -1;
- loc_t loc = {0, };
- int32_t op_errno = 0;
-
- GF_VALIDATE_OR_GOTO ("bit-rot-scrubber", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
- GF_VALIDATE_OR_GOTO (this->name, child, out);
- GF_VALIDATE_OR_GOTO (this->name, dict, out);
-
- table = child->table;
-
- inode = inode_find (table, BR_BAD_OBJ_CONTAINER);
- if (!inode) {
- inode = br_lookup_bad_obj_dir (this, child,
- BR_BAD_OBJ_CONTAINER);
- if (!inode)
- goto out;
- }
+ /*
+ * ignore the dict_set errors for now. The intention is
+ * to get as many bad objects as possible instead of
+ * erroring out at the first failure.
+ */
+ ret = dict_set_dynstr_with_alloc(dict, key, entry->d_name);
+ if (!ret)
+ count++;
- fd = fd_create (inode, 0);
- if (!fd) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- BRB_MSG_FD_CREATE_FAILED, "fd creation for the bad "
- "objects directory failed (gfid: %s)",
- uuid_utoa (BR_BAD_OBJ_CONTAINER));
- goto out;
+ if (out_dict) {
+ dict_copy(out_dict, dict);
+ dict_unref(out_dict);
+ out_dict = NULL;
+ }
}
- loc.inode = inode;
- gf_uuid_copy (loc.gfid, inode->gfid);
-
- ret = syncop_opendir (child->xl, &loc, fd, NULL, NULL);
- if (ret < 0) {
- op_errno = -ret;
- fd_unref (fd);
- fd = NULL;
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- BRB_MSG_FD_CREATE_FAILED, "failed to open the bad "
- "objects directory %s",
- uuid_utoa (BR_BAD_OBJ_CONTAINER));
- goto out;
- }
+ gf_dirent_free(&entries);
+ }
- fd_bind (fd);
+ ret = count;
+ ret = dict_set_int32_sizen(dict, "count", count);
- ret = br_read_bad_object_dir (this, child, fd, dict);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRB_MSG_BAD_OBJ_READDIR_FAIL, "readdir of the bad "
- "objects directory (%s) failed ",
- uuid_utoa (BR_BAD_OBJ_CONTAINER));
- goto out;
- }
+out:
+ return ret;
+}
- ret = 0;
+int32_t
+br_get_bad_objects_from_child(xlator_t *this, dict_t *dict, br_child_t *child)
+{
+ inode_t *inode = NULL;
+ inode_table_t *table = NULL;
+ fd_t *fd = NULL;
+ int32_t ret = -1;
+ loc_t loc = {
+ 0,
+ };
+ int32_t op_errno = 0;
+
+ GF_VALIDATE_OR_GOTO("bit-rot-scrubber", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
+ GF_VALIDATE_OR_GOTO(this->name, child, out);
+ GF_VALIDATE_OR_GOTO(this->name, dict, out);
+
+ table = child->table;
+
+ inode = inode_find(table, BR_BAD_OBJ_CONTAINER);
+ if (!inode) {
+ inode = br_lookup_bad_obj_dir(this, child, BR_BAD_OBJ_CONTAINER);
+ if (!inode)
+ goto out;
+ }
+
+ fd = fd_create(inode, 0);
+ if (!fd) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_FD_CREATE_FAILED,
+ "fd creation for the bad "
+ "objects directory failed (gfid: %s)",
+ uuid_utoa(BR_BAD_OBJ_CONTAINER));
+ goto out;
+ }
+
+ loc.inode = inode;
+ gf_uuid_copy(loc.gfid, inode->gfid);
+
+ ret = syncop_opendir(child->xl, &loc, fd, NULL, NULL);
+ if (ret < 0) {
+ op_errno = -ret;
+ fd_unref(fd);
+ fd = NULL;
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, BRB_MSG_FD_CREATE_FAILED,
+ "failed to open the bad "
+ "objects directory %s",
+ uuid_utoa(BR_BAD_OBJ_CONTAINER));
+ goto out;
+ }
+
+ fd_bind(fd);
+
+ ret = br_read_bad_object_dir(this, child, fd, dict);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, BRB_MSG_BAD_OBJ_READDIR_FAIL,
+ "readdir of the bad "
+ "objects directory (%s) failed ",
+ uuid_utoa(BR_BAD_OBJ_CONTAINER));
+ goto out;
+ }
+
+ ret = 0;
out:
- loc_wipe (&loc);
- if (fd)
- fd_unref (fd);
- return ret;
+ loc_wipe(&loc);
+ if (fd)
+ fd_unref(fd);
+ return ret;
}
int32_t
-br_collect_bad_objects_of_child (xlator_t *this, br_child_t *child,
- dict_t *dict, dict_t *child_dict,
- int32_t total_count)
+br_collect_bad_objects_of_child(xlator_t *this, br_child_t *child, dict_t *dict,
+ dict_t *child_dict, int32_t total_count)
{
-
- int32_t ret = -1;
- int32_t count = 0;
- char key[PATH_MAX] = {0, };
- char main_key[PATH_MAX] = {0, };
- int32_t j = 0;
- int32_t tmp_count = 0;
- char *entry = NULL;
- char tmp[PATH_MAX] = {0, };
- char *path = NULL;
-
- ret = dict_get_int32 (child_dict, "count", &count);
+ int32_t ret = -1;
+ int32_t count = 0;
+ char key[32] = {
+ 0,
+ };
+ char main_key[32] = {
+ 0,
+ };
+ int32_t j = 0;
+ int32_t tmp_count = 0;
+ char *entry = NULL;
+ char tmp[PATH_MAX] = {
+ 0,
+ };
+ char *path = NULL;
+ int32_t len = 0;
+
+ ret = dict_get_int32_sizen(child_dict, "count", &count);
+ if (ret)
+ goto out;
+
+ tmp_count = total_count;
+
+ for (j = 0; j < count; j++) {
+ len = snprintf(key, sizeof(key), "quarantine-%d", j);
+ ret = dict_get_strn(child_dict, key, len, &entry);
if (ret)
- goto out;
-
- tmp_count = total_count;
-
- for (j = 0; j < count; j++) {
- snprintf (key, PATH_MAX, "quarantine-%d", j);
- ret = dict_get_str (child_dict, key, &entry);
- if (ret)
- continue;
-
- ret = dict_get_str (child_dict, entry, &path);
- snprintf (tmp, PATH_MAX, "%s ==> BRICK: %s\n path: %s",
- entry, child->brick_path, path);
- snprintf (main_key, PATH_MAX, "quarantine-%d",
- tmp_count);
-
- ret = dict_set_dynstr_with_alloc (dict, main_key, tmp);
- if (!ret)
- tmp_count++;
- path = NULL;
+ continue;
+
+ ret = dict_get_str(child_dict, entry, &path);
+ len = snprintf(tmp, PATH_MAX, "%s ==> BRICK: %s\n path: %s", entry,
+ child->brick_path, path);
+ if ((len < 0) || (len >= PATH_MAX)) {
+ continue;
}
+ snprintf(main_key, sizeof(main_key), "quarantine-%d", tmp_count);
- ret = tmp_count;
+ ret = dict_set_dynstr_with_alloc(dict, main_key, tmp);
+ if (!ret)
+ tmp_count++;
+ path = NULL;
+ }
+
+ ret = tmp_count;
out:
- return ret;
+ return ret;
}
int32_t
-br_collect_bad_objects_from_children (xlator_t *this, dict_t *dict)
+br_collect_bad_objects_from_children(xlator_t *this, dict_t *dict)
{
- int32_t ret = -1;
- dict_t *child_dict = NULL;
- int32_t i = 0;
- int32_t total_count = 0;
- br_child_t *child = NULL;
- br_private_t *priv = NULL;
- dict_t *tmp_dict = NULL;
-
- priv = this->private;
- tmp_dict = dict;
-
- for (i = 0; i < priv->child_count; i++) {
- child = &priv->children[i];
- GF_ASSERT (child);
- if (!_br_is_child_connected (child))
- continue;
-
- child_dict = dict_new ();
- if (!child_dict) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- BRB_MSG_NO_MEMORY, "failed to allocate dict");
- continue;
- }
- ret = br_get_bad_objects_from_child (this, child_dict, child);
- /*
- * Continue asking the remaining children for the list of
- * bad objects even though getting the list from one of them
- * fails.
- */
- if (ret) {
- dict_unref (child_dict);
- continue;
- }
-
- ret = br_collect_bad_objects_of_child (this, child, tmp_dict,
- child_dict, total_count);
- if (ret < 0) {
- dict_unref (child_dict);
- continue;
- }
-
- total_count = ret;
- dict_unref (child_dict);
- child_dict = NULL;
+ int32_t ret = -1;
+ dict_t *child_dict = NULL;
+ int32_t i = 0;
+ int32_t total_count = 0;
+ br_child_t *child = NULL;
+ br_private_t *priv = NULL;
+ dict_t *tmp_dict = NULL;
+
+ priv = this->private;
+ tmp_dict = dict;
+
+ for (i = 0; i < priv->child_count; i++) {
+ child = &priv->children[i];
+ GF_ASSERT(child);
+ if (!_br_is_child_connected(child))
+ continue;
+
+ child_dict = dict_new();
+ if (!child_dict) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY,
+ "failed to allocate dict");
+ continue;
+ }
+ ret = br_get_bad_objects_from_child(this, child_dict, child);
+ /*
+ * Continue asking the remaining children for the list of
+ * bad objects even though getting the list from one of them
+ * fails.
+ */
+ if (ret) {
+ dict_unref(child_dict);
+ continue;
}
- ret = dict_set_int32 (tmp_dict, "total-count", total_count);
+ ret = br_collect_bad_objects_of_child(this, child, tmp_dict, child_dict,
+ total_count);
+ if (ret < 0) {
+ dict_unref(child_dict);
+ continue;
+ }
+
+ total_count = ret;
+ dict_unref(child_dict);
+ child_dict = NULL;
+ }
- return ret;
+ ret = dict_set_int32(tmp_dict, "total-count", total_count);
+
+ return ret;
}
int32_t
-br_get_bad_objects_list (xlator_t *this, dict_t **dict)
+br_get_bad_objects_list(xlator_t *this, dict_t **dict)
{
- int32_t ret = -1;
- dict_t *tmp_dict = NULL;
+ int32_t ret = -1;
+ dict_t *tmp_dict = NULL;
- GF_VALIDATE_OR_GOTO ("bir-rot-scrubber", this, out);
- GF_VALIDATE_OR_GOTO (this->name, dict, out);
+ GF_VALIDATE_OR_GOTO("bir-rot-scrubber", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, dict, out);
- tmp_dict = *dict;
+ tmp_dict = *dict;
+ if (!tmp_dict) {
+ tmp_dict = dict_new();
if (!tmp_dict) {
- tmp_dict = dict_new ();
- if (!tmp_dict) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- BRB_MSG_NO_MEMORY, "failed to allocate dict");
- goto out;
- }
- *dict = tmp_dict;
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY,
+ "failed to allocate dict");
+ goto out;
}
+ *dict = tmp_dict;
+ }
- ret = br_collect_bad_objects_from_children (this, tmp_dict);
+ ret = br_collect_bad_objects_from_children(this, tmp_dict);
out:
- return ret;
+ return ret;
}
static int
-wait_for_scrub_to_finish (xlator_t *this)
+wait_for_scrub_to_finish(xlator_t *this)
{
- int ret = -1;
- br_private_t *priv = NULL;
- struct br_monitor *scrub_monitor = NULL;
-
- priv = this->private;
- scrub_monitor = &priv->scrub_monitor;
-
- GF_VALIDATE_OR_GOTO ("bit-rot", scrub_monitor, out);
- GF_VALIDATE_OR_GOTO ("bit-rot", this, out);
-
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
- "Waiting for all children to start and finish scrub");
-
- pthread_mutex_lock (&scrub_monitor->donelock);
- {
- while (!scrub_monitor->done)
- pthread_cond_wait (&scrub_monitor->donecond,
- &scrub_monitor->donelock);
- }
- pthread_mutex_unlock (&scrub_monitor->donelock);
- ret = 0;
+ int ret = -1;
+ br_private_t *priv = NULL;
+ struct br_monitor *scrub_monitor = NULL;
+
+ priv = this->private;
+ scrub_monitor = &priv->scrub_monitor;
+
+ GF_VALIDATE_OR_GOTO("bit-rot", scrub_monitor, out);
+ GF_VALIDATE_OR_GOTO("bit-rot", this, out);
+
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_INFO,
+ "Waiting for all children to start and finish scrub");
+
+ pthread_mutex_lock(&scrub_monitor->donelock);
+ {
+ while (!scrub_monitor->done)
+ pthread_cond_wait(&scrub_monitor->donecond,
+ &scrub_monitor->donelock);
+ }
+ pthread_mutex_unlock(&scrub_monitor->donelock);
+ ret = 0;
out:
- return ret;
+ return ret;
}
/**
@@ -1894,153 +1921,150 @@ out:
* thread that takes care of state machine.
*/
void *
-br_monitor_thread (void *arg)
+br_monitor_thread(void *arg)
{
- int32_t ret = 0;
- xlator_t *this = NULL;
- br_private_t *priv = NULL;
- struct br_monitor *scrub_monitor = NULL;
-
- this = arg;
- priv = this->private;
-
- /*
- * Since, this is the topmost xlator, THIS has to be set by bit-rot
- * xlator itself (STACK_WIND wont help in this case). Also it has
- * to be done for each thread that gets spawned. Otherwise, a new
- * thread will get global_xlator's pointer when it does "THIS".
- */
- THIS = this;
-
- scrub_monitor = &priv->scrub_monitor;
-
- pthread_mutex_lock (&scrub_monitor->mutex);
- {
- while (!scrub_monitor->inited)
- pthread_cond_wait (&scrub_monitor->cond,
- &scrub_monitor->mutex);
- }
- pthread_mutex_unlock (&scrub_monitor->mutex);
-
- /* this needs to be serialized with reconfigure() */
- pthread_mutex_lock (&priv->lock);
- {
- ret = br_scrub_state_machine (this, _gf_false);
- }
- pthread_mutex_unlock (&priv->lock);
+ int32_t ret = 0;
+ xlator_t *this = NULL;
+ br_private_t *priv = NULL;
+ struct br_monitor *scrub_monitor = NULL;
+
+ this = arg;
+ priv = this->private;
+
+ /*
+ * Since, this is the topmost xlator, THIS has to be set by bit-rot
+ * xlator itself (STACK_WIND won't help in this case). Also it has
+ * to be done for each thread that gets spawned. Otherwise, a new
+ * thread will get global_xlator's pointer when it does "THIS".
+ */
+ THIS = this;
+
+ scrub_monitor = &priv->scrub_monitor;
+
+ pthread_mutex_lock(&scrub_monitor->mutex);
+ {
+ while (!scrub_monitor->inited)
+ pthread_cond_wait(&scrub_monitor->cond, &scrub_monitor->mutex);
+ }
+ pthread_mutex_unlock(&scrub_monitor->mutex);
+
+ /* this needs to be serialized with reconfigure() */
+ pthread_mutex_lock(&priv->lock);
+ {
+ ret = br_scrub_state_machine(this, _gf_false);
+ }
+ pthread_mutex_unlock(&priv->lock);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, BRB_MSG_SSM_FAILED,
+ "Scrub state machine failed");
+ goto out;
+ }
+
+ while (1) {
+ /* Wait for all children to finish scrubbing */
+ ret = wait_for_scrub_to_finish(this);
if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- BRB_MSG_SSM_FAILED,
- "Scrub state machine failed");
- goto out;
+ gf_msg(this->name, GF_LOG_ERROR, -ret, BRB_MSG_SCRUB_WAIT_FAILED,
+ "Scrub wait failed");
+ goto out;
}
- while (1) {
- /* Wait for all children to finish scrubbing */
- ret = wait_for_scrub_to_finish (this);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- BRB_MSG_SCRUB_WAIT_FAILED,
- "Scrub wait failed");
- goto out;
- }
-
- /* scrub exit criteria: Move the state to PENDING */
- br_scrubber_exit_control (this);
- }
+ /* scrub exit criteria: Move the state to PENDING */
+ br_scrubber_exit_control(this);
+ }
out:
- return NULL;
+ return NULL;
}
static void
-br_set_scrub_state (struct br_monitor *scrub_monitor, br_scrub_state_t state)
+br_set_scrub_state(struct br_monitor *scrub_monitor, br_scrub_state_t state)
{
- LOCK (&scrub_monitor->lock);
- {
- _br_monitor_set_scrub_state (scrub_monitor, state);
- }
- UNLOCK (&scrub_monitor->lock);
+ LOCK(&scrub_monitor->lock);
+ {
+ _br_monitor_set_scrub_state(scrub_monitor, state);
+ }
+ UNLOCK(&scrub_monitor->lock);
}
int32_t
-br_scrubber_monitor_init (xlator_t *this, br_private_t *priv)
+br_scrubber_monitor_init(xlator_t *this, br_private_t *priv)
{
- struct br_monitor *scrub_monitor = NULL;
- int ret = 0;
+ struct br_monitor *scrub_monitor = NULL;
+ int ret = 0;
- scrub_monitor = &priv->scrub_monitor;
+ scrub_monitor = &priv->scrub_monitor;
- LOCK_INIT (&scrub_monitor->lock);
- scrub_monitor->this = this;
+ LOCK_INIT(&scrub_monitor->lock);
+ scrub_monitor->this = this;
- scrub_monitor->inited = _gf_false;
- pthread_mutex_init (&scrub_monitor->mutex, NULL);
- pthread_cond_init (&scrub_monitor->cond, NULL);
+ scrub_monitor->inited = _gf_false;
+ pthread_mutex_init(&scrub_monitor->mutex, NULL);
+ pthread_cond_init(&scrub_monitor->cond, NULL);
- scrub_monitor->kick = _gf_false;
- scrub_monitor->active_child_count = 0;
- pthread_mutex_init (&scrub_monitor->wakelock, NULL);
- pthread_cond_init (&scrub_monitor->wakecond, NULL);
+ scrub_monitor->kick = _gf_false;
+ scrub_monitor->active_child_count = 0;
+ pthread_mutex_init(&scrub_monitor->wakelock, NULL);
+ pthread_cond_init(&scrub_monitor->wakecond, NULL);
- scrub_monitor->done = _gf_false;
- pthread_mutex_init (&scrub_monitor->donelock, NULL);
- pthread_cond_init (&scrub_monitor->donecond, NULL);
-
- /* Set the state to INACTIVE */
- br_set_scrub_state (&priv->scrub_monitor, BR_SCRUB_STATE_INACTIVE);
-
- /* Start the monitor thread */
- ret = gf_thread_create (&scrub_monitor->thread, NULL,
- br_monitor_thread, this, "brmon");
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- BRB_MSG_SPAWN_FAILED, "monitor thread creation failed");
- ret = -1;
- goto err;
- }
+ scrub_monitor->done = _gf_false;
+ pthread_mutex_init(&scrub_monitor->donelock, NULL);
+ pthread_cond_init(&scrub_monitor->donecond, NULL);
- return 0;
+ /* Set the state to INACTIVE */
+ br_set_scrub_state(&priv->scrub_monitor, BR_SCRUB_STATE_INACTIVE);
+
+ /* Start the monitor thread */
+ ret = gf_thread_create(&scrub_monitor->thread, NULL, br_monitor_thread,
+ this, "brmon");
+ if (ret != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret, BRB_MSG_SPAWN_FAILED,
+ "monitor thread creation failed");
+ ret = -1;
+ goto err;
+ }
+
+ return 0;
err:
- pthread_mutex_destroy (&scrub_monitor->mutex);
- pthread_cond_destroy (&scrub_monitor->cond);
+ pthread_mutex_destroy(&scrub_monitor->mutex);
+ pthread_cond_destroy(&scrub_monitor->cond);
- pthread_mutex_destroy (&scrub_monitor->wakelock);
- pthread_cond_destroy (&scrub_monitor->wakecond);
+ pthread_mutex_destroy(&scrub_monitor->wakelock);
+ pthread_cond_destroy(&scrub_monitor->wakecond);
- pthread_mutex_destroy (&scrub_monitor->donelock);
- pthread_cond_destroy (&scrub_monitor->donecond);
+ pthread_mutex_destroy(&scrub_monitor->donelock);
+ pthread_cond_destroy(&scrub_monitor->donecond);
- LOCK_DESTROY (&scrub_monitor->lock);
+ LOCK_DESTROY(&scrub_monitor->lock);
- return ret;
+ return ret;
}
int32_t
-br_scrubber_init (xlator_t *this, br_private_t *priv)
+br_scrubber_init(xlator_t *this, br_private_t *priv)
{
- struct br_scrubber *fsscrub = NULL;
- int ret = 0;
+ struct br_scrubber *fsscrub = NULL;
+ int ret = 0;
- priv->tbf = tbf_init (NULL, 0);
- if (!priv->tbf)
- return -1;
+ priv->tbf = tbf_init(NULL, 0);
+ if (!priv->tbf)
+ return -1;
- ret = br_scrubber_monitor_init (this, priv);
- if (ret)
- return -1;
+ ret = br_scrubber_monitor_init(this, priv);
+ if (ret)
+ return -1;
- fsscrub = &priv->fsscrub;
+ fsscrub = &priv->fsscrub;
- fsscrub->this = this;
- fsscrub->throttle = BR_SCRUB_THROTTLE_VOID;
+ fsscrub->this = this;
+ fsscrub->throttle = BR_SCRUB_THROTTLE_VOID;
- pthread_mutex_init (&fsscrub->mutex, NULL);
- pthread_cond_init (&fsscrub->cond, NULL);
+ pthread_mutex_init(&fsscrub->mutex, NULL);
+ pthread_cond_init(&fsscrub->cond, NULL);
- fsscrub->nr_scrubbers = 0;
- INIT_LIST_HEAD (&fsscrub->scrubbers);
- INIT_LIST_HEAD (&fsscrub->scrublist);
+ fsscrub->nr_scrubbers = 0;
+ INIT_LIST_HEAD(&fsscrub->scrubbers);
+ INIT_LIST_HEAD(&fsscrub->scrublist);
- return 0;
+ return 0;
}
diff --git a/xlators/features/bit-rot/src/bitd/bit-rot-scrub.h b/xlators/features/bit-rot/src/bitd/bit-rot-scrub.h
index 8cc88ec153e..4e5f67bc021 100644
--- a/xlators/features/bit-rot/src/bitd/bit-rot-scrub.h
+++ b/xlators/features/bit-rot/src/bitd/bit-rot-scrub.h
@@ -11,27 +11,36 @@
#ifndef __BIT_ROT_SCRUB_H__
#define __BIT_ROT_SCRUB_H__
-#include "xlator.h"
+#include <glusterfs/xlator.h>
#include "bit-rot.h"
-void *br_fsscanner (void *);
+void *
+br_fsscanner(void *);
-int32_t br_fsscan_schedule (xlator_t *);
-int32_t br_fsscan_reschedule (xlator_t *);
-int32_t br_fsscan_activate (xlator_t *);
-int32_t br_fsscan_deactivate (xlator_t *);
-int32_t br_fsscan_ondemand (xlator_t *);
+int32_t
+br_fsscan_schedule(xlator_t *);
+int32_t
+br_fsscan_reschedule(xlator_t *);
+int32_t
+br_fsscan_activate(xlator_t *);
+int32_t
+br_fsscan_deactivate(xlator_t *);
+int32_t
+br_fsscan_ondemand(xlator_t *);
-int32_t br_scrubber_handle_options (xlator_t *, br_private_t *, dict_t *);
+int32_t
+br_scrubber_handle_options(xlator_t *, br_private_t *, dict_t *);
int32_t
-br_scrubber_monitor_init (xlator_t *, br_private_t *);
+br_scrubber_monitor_init(xlator_t *, br_private_t *);
-int32_t br_scrubber_init (xlator_t *, br_private_t *);
+int32_t
+br_scrubber_init(xlator_t *, br_private_t *);
-int32_t br_collect_bad_objects_from_children (xlator_t *this, dict_t *dict);
+int32_t
+br_collect_bad_objects_from_children(xlator_t *this, dict_t *dict);
void
-br_child_set_scrub_state (br_child_t *, gf_boolean_t);
+br_child_set_scrub_state(br_child_t *, gf_boolean_t);
#endif /* __BIT_ROT_SCRUB_H__ */
diff --git a/xlators/features/bit-rot/src/bitd/bit-rot-ssm.c b/xlators/features/bit-rot/src/bitd/bit-rot-ssm.c
index af887a1ff8b..753e31a3b23 100644
--- a/xlators/features/bit-rot/src/bitd/bit-rot-ssm.c
+++ b/xlators/features/bit-rot/src/bitd/bit-rot-ssm.c
@@ -12,78 +12,79 @@
#include "bit-rot-scrub.h"
#include "bit-rot-bitd-messages.h"
-int br_scrub_ssm_noop (xlator_t *this)
+int
+br_scrub_ssm_noop(xlator_t *this)
{
- return 0;
+ return 0;
}
int
-br_scrub_ssm_state_pause (xlator_t *this)
+br_scrub_ssm_state_pause(xlator_t *this)
{
- br_private_t *priv = NULL;
- struct br_monitor *scrub_monitor = NULL;
+ br_private_t *priv = NULL;
+ struct br_monitor *scrub_monitor = NULL;
- priv = this->private;
- scrub_monitor = &priv->scrub_monitor;
+ priv = this->private;
+ scrub_monitor = &priv->scrub_monitor;
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_GENERIC_SSM_INFO,
- "Scrubber paused");
- _br_monitor_set_scrub_state (scrub_monitor, BR_SCRUB_STATE_PAUSED);
- return 0;
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_GENERIC_SSM_INFO,
+ "Scrubber paused");
+ _br_monitor_set_scrub_state(scrub_monitor, BR_SCRUB_STATE_PAUSED);
+ return 0;
}
int
-br_scrub_ssm_state_ipause (xlator_t *this)
+br_scrub_ssm_state_ipause(xlator_t *this)
{
- br_private_t *priv = NULL;
- struct br_monitor *scrub_monitor = NULL;
+ br_private_t *priv = NULL;
+ struct br_monitor *scrub_monitor = NULL;
- priv = this->private;
- scrub_monitor = &priv->scrub_monitor;
+ priv = this->private;
+ scrub_monitor = &priv->scrub_monitor;
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_GENERIC_SSM_INFO,
- "Scrubber paused");
- _br_monitor_set_scrub_state (scrub_monitor, BR_SCRUB_STATE_IPAUSED);
- return 0;
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_GENERIC_SSM_INFO,
+ "Scrubber paused");
+ _br_monitor_set_scrub_state(scrub_monitor, BR_SCRUB_STATE_IPAUSED);
+ return 0;
}
int
-br_scrub_ssm_state_active (xlator_t *this)
+br_scrub_ssm_state_active(xlator_t *this)
{
- br_private_t *priv = NULL;
- struct br_monitor *scrub_monitor = NULL;
+ br_private_t *priv = NULL;
+ struct br_monitor *scrub_monitor = NULL;
- priv = this->private;
- scrub_monitor = &priv->scrub_monitor;
+ priv = this->private;
+ scrub_monitor = &priv->scrub_monitor;
- if (scrub_monitor->done) {
- (void) br_fsscan_activate (this);
- } else {
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_GENERIC_SSM_INFO,
- "Scrubbing resumed");
- _br_monitor_set_scrub_state (scrub_monitor, BR_SCRUB_STATE_ACTIVE);
- }
+ if (scrub_monitor->done) {
+ (void)br_fsscan_activate(this);
+ } else {
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_GENERIC_SSM_INFO,
+ "Scrubbing resumed");
+ _br_monitor_set_scrub_state(scrub_monitor, BR_SCRUB_STATE_ACTIVE);
+ }
- return 0;
+ return 0;
}
int
-br_scrub_ssm_state_stall (xlator_t *this)
+br_scrub_ssm_state_stall(xlator_t *this)
{
- br_private_t *priv = NULL;
- struct br_monitor *scrub_monitor = NULL;
+ br_private_t *priv = NULL;
+ struct br_monitor *scrub_monitor = NULL;
- priv = this->private;
- scrub_monitor = &priv->scrub_monitor;
+ priv = this->private;
+ scrub_monitor = &priv->scrub_monitor;
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_GENERIC_SSM_INFO,
- "Volume is under active scrubbing. Pausing scrub..");
- _br_monitor_set_scrub_state (scrub_monitor, BR_SCRUB_STATE_STALLED);
- return 0;
+ gf_msg(this->name, GF_LOG_INFO, 0, BRB_MSG_GENERIC_SSM_INFO,
+ "Volume is under active scrubbing. Pausing scrub..");
+ _br_monitor_set_scrub_state(scrub_monitor, BR_SCRUB_STATE_STALLED);
+ return 0;
}
-static br_scrub_ssm_call *
-br_scrub_ssm[BR_SCRUB_MAXSTATES][BR_SCRUB_MAXEVENTS] = {
+static br_scrub_ssm_call *br_scrub_ssm[BR_SCRUB_MAXSTATES][BR_SCRUB_MAXEVENTS] =
+ {
/* INACTIVE */
{br_fsscan_schedule, br_scrub_ssm_state_ipause, br_scrub_ssm_noop},
/* PENDING */
@@ -99,25 +100,25 @@ br_scrub_ssm[BR_SCRUB_MAXSTATES][BR_SCRUB_MAXEVENTS] = {
};
int32_t
-br_scrub_state_machine (xlator_t *this, gf_boolean_t scrub_ondemand)
+br_scrub_state_machine(xlator_t *this, gf_boolean_t scrub_ondemand)
{
- br_private_t *priv = NULL;
- br_scrub_ssm_call *call = NULL;
- struct br_scrubber *fsscrub = NULL;
- br_scrub_state_t currstate = 0;
- br_scrub_event_t event = 0;
- struct br_monitor *scrub_monitor = NULL;
-
- priv = this->private;
- fsscrub = &priv->fsscrub;
- scrub_monitor = &priv->scrub_monitor;
-
- currstate = scrub_monitor->state;
- if (scrub_ondemand)
- event = BR_SCRUB_EVENT_ONDEMAND;
- else
- event = _br_child_get_scrub_event (fsscrub);
-
- call = br_scrub_ssm[currstate][event];
- return call (this);
+ br_private_t *priv = NULL;
+ br_scrub_ssm_call *call = NULL;
+ struct br_scrubber *fsscrub = NULL;
+ br_scrub_state_t currstate = 0;
+ br_scrub_event_t event = 0;
+ struct br_monitor *scrub_monitor = NULL;
+
+ priv = this->private;
+ fsscrub = &priv->fsscrub;
+ scrub_monitor = &priv->scrub_monitor;
+
+ currstate = scrub_monitor->state;
+ if (scrub_ondemand)
+ event = BR_SCRUB_EVENT_ONDEMAND;
+ else
+ event = _br_child_get_scrub_event(fsscrub);
+
+ call = br_scrub_ssm[currstate][event];
+ return call(this);
}
diff --git a/xlators/features/bit-rot/src/bitd/bit-rot-ssm.h b/xlators/features/bit-rot/src/bitd/bit-rot-ssm.h
index 8609477180b..37b45a42eac 100644
--- a/xlators/features/bit-rot/src/bitd/bit-rot-ssm.h
+++ b/xlators/features/bit-rot/src/bitd/bit-rot-ssm.h
@@ -11,27 +11,28 @@
#ifndef __BIT_ROT_SSM_H__
#define __BIT_ROT_SSM_H__
-#include "xlator.h"
+#include <glusterfs/xlator.h>
typedef enum br_scrub_state {
- BR_SCRUB_STATE_INACTIVE = 0,
- BR_SCRUB_STATE_PENDING,
- BR_SCRUB_STATE_ACTIVE,
- BR_SCRUB_STATE_PAUSED,
- BR_SCRUB_STATE_IPAUSED,
- BR_SCRUB_STATE_STALLED,
- BR_SCRUB_MAXSTATES,
+ BR_SCRUB_STATE_INACTIVE = 0,
+ BR_SCRUB_STATE_PENDING,
+ BR_SCRUB_STATE_ACTIVE,
+ BR_SCRUB_STATE_PAUSED,
+ BR_SCRUB_STATE_IPAUSED,
+ BR_SCRUB_STATE_STALLED,
+ BR_SCRUB_MAXSTATES,
} br_scrub_state_t;
typedef enum br_scrub_event {
- BR_SCRUB_EVENT_SCHEDULE = 0,
- BR_SCRUB_EVENT_PAUSE,
- BR_SCRUB_EVENT_ONDEMAND,
- BR_SCRUB_MAXEVENTS,
+ BR_SCRUB_EVENT_SCHEDULE = 0,
+ BR_SCRUB_EVENT_PAUSE,
+ BR_SCRUB_EVENT_ONDEMAND,
+ BR_SCRUB_MAXEVENTS,
} br_scrub_event_t;
struct br_monitor;
-int32_t br_scrub_state_machine (xlator_t *, gf_boolean_t);
+int32_t
+br_scrub_state_machine(xlator_t *, gf_boolean_t);
#endif /* __BIT_ROT_SSM_H__ */
diff --git a/xlators/features/bit-rot/src/bitd/bit-rot.c b/xlators/features/bit-rot/src/bitd/bit-rot.c
index db97ce54735..a2f1c343a1d 100644
--- a/xlators/features/bit-rot/src/bitd/bit-rot.c
+++ b/xlators/features/bit-rot/src/bitd/bit-rot.c
@@ -9,84 +9,81 @@
*/
#include <ctype.h>
-#include <sys/uio.h>
-#include "glusterfs.h"
-#include "xlator.h"
-#include "logging.h"
-#include "compat-errno.h"
+#include <glusterfs/logging.h>
+#include <glusterfs/compat-errno.h>
#include "bit-rot.h"
#include "bit-rot-scrub.h"
#include <pthread.h>
#include "bit-rot-bitd-messages.h"
-#define BR_HASH_CALC_READ_SIZE (128 * 1024)
+#define BR_HASH_CALC_READ_SIZE (128 * 1024)
-typedef int32_t (br_child_handler)(xlator_t *, br_child_t *);
+typedef int32_t(br_child_handler)(xlator_t *, br_child_t *);
struct br_child_event {
- xlator_t *this;
+ xlator_t *this;
- br_child_t *child;
+ br_child_t *child;
- br_child_handler *call;
+ br_child_handler *call;
- struct list_head list;
+ struct list_head list;
};
static int
-br_find_child_index (xlator_t *this, xlator_t *child)
+br_find_child_index(xlator_t *this, xlator_t *child)
{
- br_private_t *priv = NULL;
- int i = -1;
- int index = -1;
+ br_private_t *priv = NULL;
+ int i = -1;
+ int index = -1;
- GF_VALIDATE_OR_GOTO ("bit-rot", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
- GF_VALIDATE_OR_GOTO (this->name, child, out);
+ GF_VALIDATE_OR_GOTO("bit-rot", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
+ GF_VALIDATE_OR_GOTO(this->name, child, out);
- priv = this->private;
+ priv = this->private;
- for (i = 0; i < priv->child_count; i++) {
- if (child == priv->children[i].xl) {
- index = i;
- break;
- }
+ for (i = 0; i < priv->child_count; i++) {
+ if (child == priv->children[i].xl) {
+ index = i;
+ break;
}
+ }
out:
- return index;
+ return index;
}
br_child_t *
-br_get_child_from_brick_path (xlator_t *this, char *brick_path)
+br_get_child_from_brick_path(xlator_t *this, char *brick_path)
{
- br_private_t *priv = NULL;
- br_child_t *child = NULL;
- br_child_t *tmp = NULL;
- int i = 0;
+ br_private_t *priv = NULL;
+ br_child_t *child = NULL;
+ br_child_t *tmp = NULL;
+ int i = 0;
- GF_VALIDATE_OR_GOTO ("bit-rot", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
- GF_VALIDATE_OR_GOTO (this->name, brick_path, out);
+ GF_VALIDATE_OR_GOTO("bit-rot", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
+ GF_VALIDATE_OR_GOTO(this->name, brick_path, out);
- priv = this->private;
+ priv = this->private;
- pthread_mutex_lock (&priv->lock);
- {
- for (i = 0; i < priv->child_count; i++) {
- tmp = &priv->children[i];
- if (!strcmp (tmp->brick_path, brick_path)) {
- child = tmp;
- break;
- }
- }
+ pthread_mutex_lock(&priv->lock);
+ {
+ for (i = 0; i < priv->child_count; i++) {
+ tmp = &priv->children[i];
+ if (!strcmp(tmp->brick_path, brick_path)) {
+ child = tmp;
+ break;
+ }
}
- pthread_mutex_unlock (&priv->lock);
+ }
+ pthread_mutex_unlock(&priv->lock);
out:
- return child;
+ return child;
}
/**
@@ -94,18 +91,18 @@ out:
* needed -- later.
*/
void *
-br_brick_init (void *xl, struct gf_brick_spec *brick)
+br_brick_init(void *xl, struct gf_brick_spec *brick)
{
- return brick;
+ return brick;
}
/**
* and cleanup things here when allocated br_brick_init().
*/
void
-br_brick_fini (void *xl, char *brick, void *data)
+br_brick_fini(void *xl, char *brick, void *data)
{
- return;
+ return;
}
/**
@@ -117,109 +114,109 @@ br_brick_fini (void *xl, char *brick, void *data)
* change stub to handle this change.
*/
static br_isignature_t *
-br_prepare_signature (const unsigned char *sign,
- unsigned long hashlen,
- int8_t hashtype, br_object_t *object)
+br_prepare_signature(const unsigned char *sign, unsigned long hashlen,
+ int8_t hashtype, br_object_t *object)
{
- br_isignature_t *signature = NULL;
+ br_isignature_t *signature = NULL;
- /* TODO: use mem-pool */
- signature = GF_CALLOC (1, signature_size (hashlen + 1),
- gf_br_stub_mt_signature_t);
- if (!signature)
- return NULL;
+ /* TODO: use mem-pool */
+ signature = GF_CALLOC(1, signature_size(hashlen + 1),
+ gf_br_stub_mt_signature_t);
+ if (!signature)
+ return NULL;
- /* object version */
- signature->signedversion = object->signedversion;
+ /* object version */
+ signature->signedversion = object->signedversion;
- /* signature length & type */
- signature->signaturelen = hashlen;
- signature->signaturetype = hashtype;
+ /* signature length & type */
+ signature->signaturelen = hashlen;
+ signature->signaturetype = hashtype;
- /* signature itself */
- memcpy (signature->signature, (char *)sign, hashlen);
- signature->signature[hashlen+1] = '\0';
+ /* signature itself */
+ memcpy(signature->signature, (char *)sign, hashlen);
+ signature->signature[hashlen + 1] = '\0';
- return signature;
+ return signature;
}
gf_boolean_t
-bitd_is_bad_file (xlator_t *this, br_child_t *child, loc_t *loc, fd_t *fd)
+bitd_is_bad_file(xlator_t *this, br_child_t *child, loc_t *loc, fd_t *fd)
{
- int32_t ret = -1;
- dict_t *xattr = NULL;
- inode_t *inode = NULL;
- gf_boolean_t bad_file = _gf_false;
+ int32_t ret = -1;
+ dict_t *xattr = NULL;
+ inode_t *inode = NULL;
+ gf_boolean_t bad_file = _gf_false;
- GF_VALIDATE_OR_GOTO ("bit-rot", this, out);
+ GF_VALIDATE_OR_GOTO("bit-rot", this, out);
- inode = (loc) ? loc->inode : fd->inode;
+ inode = (loc) ? loc->inode : fd->inode;
- if (fd)
- ret = syncop_fgetxattr (child->xl, fd, &xattr,
- BITROT_OBJECT_BAD_KEY, NULL, NULL);
- else if (loc)
- ret = syncop_getxattr (child->xl, loc,
- &xattr, BITROT_OBJECT_BAD_KEY, NULL,
- NULL);
+ if (fd)
+ ret = syncop_fgetxattr(child->xl, fd, &xattr, BITROT_OBJECT_BAD_KEY,
+ NULL, NULL);
+ else if (loc)
+ ret = syncop_getxattr(child->xl, loc, &xattr, BITROT_OBJECT_BAD_KEY,
+ NULL, NULL);
- if (!ret) {
- gf_msg_debug (this->name, 0, "[GFID: %s] is marked corrupted",
- uuid_utoa (inode->gfid));
- bad_file = _gf_true;
- }
+ if (!ret) {
+ gf_msg_debug(this->name, 0, "[GFID: %s] is marked corrupted",
+ uuid_utoa(inode->gfid));
+ bad_file = _gf_true;
+ }
- if (xattr)
- dict_unref (xattr);
+ if (xattr)
+ dict_unref(xattr);
out:
- return bad_file;
+ return bad_file;
}
/**
* Do a lookup on the gfid present within the object.
*/
static int32_t
-br_object_lookup (xlator_t *this, br_object_t *object,
- struct iatt *iatt, inode_t **linked_inode)
+br_object_lookup(xlator_t *this, br_object_t *object, struct iatt *iatt,
+ inode_t **linked_inode)
{
- int ret = -EINVAL;
- loc_t loc = {0, };
- inode_t *inode = NULL;
-
- GF_VALIDATE_OR_GOTO ("bit-rot", this, out);
- GF_VALIDATE_OR_GOTO (this->name, object, out);
-
- inode = inode_find (object->child->table, object->gfid);
-
- if (inode)
- loc.inode = inode;
- else
- loc.inode = inode_new (object->child->table);
-
- if (!loc.inode) {
- ret = -ENOMEM;
- goto out;
- }
-
- gf_uuid_copy (loc.gfid, object->gfid);
-
- ret = syncop_lookup (object->child->xl, &loc, iatt, NULL, NULL, NULL);
- if (ret < 0)
- goto out;
-
- /*
- * The file might have been deleted by the application
- * after getting the event, but before doing a lookup.
- * So use linked_inode after inode_link is done.
- */
- *linked_inode = inode_link (loc.inode, NULL, NULL, iatt);
- if (*linked_inode)
- inode_lookup (*linked_inode);
+ int ret = -EINVAL;
+ loc_t loc = {
+ 0,
+ };
+ inode_t *inode = NULL;
+
+ GF_VALIDATE_OR_GOTO("bit-rot", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, object, out);
+
+ inode = inode_find(object->child->table, object->gfid);
+
+ if (inode)
+ loc.inode = inode;
+ else
+ loc.inode = inode_new(object->child->table);
+
+ if (!loc.inode) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ gf_uuid_copy(loc.gfid, object->gfid);
+
+ ret = syncop_lookup(object->child->xl, &loc, iatt, NULL, NULL, NULL);
+ if (ret < 0)
+ goto out;
+
+ /*
+ * The file might have been deleted by the application
+ * after getting the event, but before doing a lookup.
+ * So use linked_inode after inode_link is done.
+ */
+ *linked_inode = inode_link(loc.inode, NULL, NULL, iatt);
+ if (*linked_inode)
+ inode_lookup(*linked_inode);
out:
- loc_wipe (&loc);
- return ret;
+ loc_wipe(&loc);
+ return ret;
}
/**
@@ -228,43 +225,44 @@ out:
* passing xdata -- may be use frame->root->pid itself.
*/
static int32_t
-br_object_open (xlator_t *this,
- br_object_t *object, inode_t *inode, fd_t **openfd)
+br_object_open(xlator_t *this, br_object_t *object, inode_t *inode,
+ fd_t **openfd)
{
- int32_t ret = -1;
- fd_t *fd = NULL;
- loc_t loc = {0, };
-
- GF_VALIDATE_OR_GOTO ("bit-rot", this, out);
- GF_VALIDATE_OR_GOTO (this->name, object, out);
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
-
- ret = -EINVAL;
- fd = fd_create (inode, 0);
- if (!fd) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_FD_CREATE_FAILED,
- "failed to create fd for the inode %s",
- uuid_utoa (inode->gfid));
- goto out;
- }
-
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
-
- ret = syncop_open (object->child->xl, &loc, O_RDONLY, fd, NULL, NULL);
- if (ret) {
- br_log_object (this, "open", inode->gfid, -ret);
- fd_unref (fd);
- fd = NULL;
- } else {
- fd_bind (fd);
- *openfd = fd;
- }
-
- loc_wipe (&loc);
+ int32_t ret = -1;
+ fd_t *fd = NULL;
+ loc_t loc = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("bit-rot", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, object, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
+
+ ret = -EINVAL;
+ fd = fd_create(inode, 0);
+ if (!fd) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_FD_CREATE_FAILED,
+ "gfid=%s", uuid_utoa(inode->gfid), NULL);
+ goto out;
+ }
+
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
+
+ ret = syncop_open(object->child->xl, &loc, O_RDONLY, fd, NULL, NULL);
+ if (ret) {
+ br_log_object(this, "open", inode->gfid, -ret);
+ fd_unref(fd);
+ fd = NULL;
+ } else {
+ fd_bind(fd);
+ *openfd = fd;
+ }
+
+ loc_wipe(&loc);
out:
- return ret;
+ return ret;
}
/**
@@ -272,287 +270,282 @@ out:
* and return the buffer.
*/
static int32_t
-br_object_read_block_and_sign (xlator_t *this, fd_t *fd, br_child_t *child,
- off_t offset, size_t size, SHA256_CTX *sha256)
+br_object_read_block_and_sign(xlator_t *this, fd_t *fd, br_child_t *child,
+ off_t offset, size_t size, SHA256_CTX *sha256)
{
- int32_t ret = -1;
- tbf_t *tbf = NULL;
- struct iovec *iovec = NULL;
- struct iobref *iobref = NULL;
- br_private_t *priv = NULL;
- int count = 0;
- int i = 0;
-
- GF_VALIDATE_OR_GOTO ("bit-rot", this, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
- GF_VALIDATE_OR_GOTO (this->name, fd->inode, out);
- GF_VALIDATE_OR_GOTO (this->name, child, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
-
- priv = this->private;
-
- GF_VALIDATE_OR_GOTO (this->name, priv->tbf, out);
- tbf = priv->tbf;
-
- ret = syncop_readv (child->xl, fd,
- size, offset, 0, &iovec, &count, &iobref, NULL,
- NULL);
+ int32_t ret = -1;
+ tbf_t *tbf = NULL;
+ struct iovec *iovec = NULL;
+ struct iobref *iobref = NULL;
+ br_private_t *priv = NULL;
+ int count = 0;
+ int i = 0;
+
+ GF_VALIDATE_OR_GOTO("bit-rot", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd->inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, child, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
+
+ priv = this->private;
+
+ GF_VALIDATE_OR_GOTO(this->name, priv->tbf, out);
+ tbf = priv->tbf;
+
+ ret = syncop_readv(child->xl, fd, size, offset, 0, &iovec, &count, &iobref,
+ NULL, NULL, NULL);
+
+ if (ret < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, BRB_MSG_READV_FAILED,
+ "gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ ret = -1;
+ goto out;
+ }
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, errno, BRB_MSG_READV_FAILED,
- "readv on %s failed", uuid_utoa (fd->inode->gfid));
- ret = -1;
- goto out;
- }
+ if (ret == 0)
+ goto out;
- if (ret == 0)
- goto out;
-
- for (i = 0; i < count; i++) {
- TBF_THROTTLE_BEGIN (tbf, TBF_OP_HASH, iovec[i].iov_len);
- {
- SHA256_Update (sha256, (const unsigned char *)
- (iovec[i].iov_base), iovec[i].iov_len);
- }
- TBF_THROTTLE_BEGIN (tbf, TBF_OP_HASH, iovec[i].iov_len);
+ for (i = 0; i < count; i++) {
+ TBF_THROTTLE_BEGIN(tbf, TBF_OP_HASH, iovec[i].iov_len);
+ {
+ SHA256_Update(sha256, (const unsigned char *)(iovec[i].iov_base),
+ iovec[i].iov_len);
}
+ TBF_THROTTLE_BEGIN(tbf, TBF_OP_HASH, iovec[i].iov_len);
+ }
- out:
- if (iovec)
- GF_FREE (iovec);
+out:
+ if (iovec)
+ GF_FREE(iovec);
- if (iobref)
- iobref_unref (iobref);
+ if (iobref)
+ iobref_unref(iobref);
- return ret;
+ return ret;
}
int32_t
-br_calculate_obj_checksum (unsigned char *md,
- br_child_t *child, fd_t *fd, struct iatt *iatt)
+br_calculate_obj_checksum(unsigned char *md, br_child_t *child, fd_t *fd,
+ struct iatt *iatt)
{
- int32_t ret = -1;
- off_t offset = 0;
- size_t block = BR_HASH_CALC_READ_SIZE;
- xlator_t *this = NULL;
-
- SHA256_CTX sha256;
-
- GF_VALIDATE_OR_GOTO ("bit-rot", child, out);
- GF_VALIDATE_OR_GOTO ("bit-rot", iatt, out);
- GF_VALIDATE_OR_GOTO ("bit-rot", fd, out);
+ int32_t ret = -1;
+ off_t offset = 0;
+ size_t block = BR_HASH_CALC_READ_SIZE;
+ xlator_t *this = NULL;
- this = child->this;
+ SHA256_CTX sha256;
- SHA256_Init (&sha256);
+ GF_VALIDATE_OR_GOTO("bit-rot", child, out);
+ GF_VALIDATE_OR_GOTO("bit-rot", iatt, out);
+ GF_VALIDATE_OR_GOTO("bit-rot", fd, out);
- while (1) {
- ret = br_object_read_block_and_sign (this, fd, child,
- offset, block, &sha256);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRB_MSG_BLOCK_READ_FAILED, "reading block with "
- "offset %lu of object %s failed", offset,
- uuid_utoa (fd->inode->gfid));
- break;
- }
+ this = child->this;
- if (ret == 0)
- break;
+ SHA256_Init(&sha256);
- offset += ret;
+ while (1) {
+ ret = br_object_read_block_and_sign(this, fd, child, offset, block,
+ &sha256);
+ if (ret < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_BLOCK_READ_FAILED,
+ "offset=%" PRIu64, offset, "object-gfid=%s",
+ uuid_utoa(fd->inode->gfid), NULL);
+ break;
}
if (ret == 0)
- SHA256_Final (md, &sha256);
+ break;
- out:
- return ret;
+ offset += ret;
+ }
+
+ if (ret == 0)
+ SHA256_Final(md, &sha256);
+
+out:
+ return ret;
}
static int32_t
-br_object_checksum (unsigned char *md,
- br_object_t *object, fd_t *fd, struct iatt *iatt)
+br_object_checksum(unsigned char *md, br_object_t *object, fd_t *fd,
+ struct iatt *iatt)
{
- return br_calculate_obj_checksum (md, object->child, fd, iatt);
+ return br_calculate_obj_checksum(md, object->child, fd, iatt);
}
static int32_t
-br_object_read_sign (inode_t *linked_inode, fd_t *fd, br_object_t *object,
- struct iatt *iatt)
+br_object_read_sign(inode_t *linked_inode, fd_t *fd, br_object_t *object,
+ struct iatt *iatt)
{
- int32_t ret = -1;
- xlator_t *this = NULL;
- dict_t *xattr = NULL;
- unsigned char *md = NULL;
- br_isignature_t *sign = NULL;
-
- GF_VALIDATE_OR_GOTO ("bit-rot", object, out);
- GF_VALIDATE_OR_GOTO ("bit-rot", linked_inode, out);
- GF_VALIDATE_OR_GOTO ("bit-rot", fd, out);
-
- this = object->this;
-
- md = GF_CALLOC (SHA256_DIGEST_LENGTH, sizeof (*md), gf_common_mt_char);
- if (!md) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY,
- "failed to allocate memory for saving hash of the "
- "object %s", uuid_utoa (fd->inode->gfid));
- goto out;
- }
-
- ret = br_object_checksum (md, object, fd, iatt);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRB_MSG_CALC_CHECKSUM_FAILED, "calculating checksum "
- "for the object %s failed",
- uuid_utoa (linked_inode->gfid));
- goto free_signature;
- }
-
- sign = br_prepare_signature (md, SHA256_DIGEST_LENGTH,
- BR_SIGNATURE_TYPE_SHA256, object);
- if (!sign) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_SIGN_FAILED,
- "failed to get the signature for the object %s",
- uuid_utoa (fd->inode->gfid));
- goto free_signature;
- }
-
- xattr = dict_for_key_value
- (GLUSTERFS_SET_OBJECT_SIGNATURE,
- (void *)sign, signature_size (SHA256_DIGEST_LENGTH), _gf_true);
-
- if (!xattr) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_SET_SIGN_FAILED,
- "dict allocation for signing failed for the object %s",
- uuid_utoa (fd->inode->gfid));
- goto free_isign;
- }
-
- ret = syncop_fsetxattr (object->child->xl, fd, xattr, 0, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_SET_SIGN_FAILED,
- "fsetxattr of signature to the object %s failed",
- uuid_utoa (fd->inode->gfid));
- goto unref_dict;
- }
-
- ret = 0;
-
- unref_dict:
- dict_unref (xattr);
- free_isign:
- GF_FREE (sign);
- free_signature:
- GF_FREE (md);
- out:
- return ret;
+ int32_t ret = -1;
+ xlator_t *this = NULL;
+ dict_t *xattr = NULL;
+ unsigned char *md = NULL;
+ br_isignature_t *sign = NULL;
+
+ GF_VALIDATE_OR_GOTO("bit-rot", object, out);
+ GF_VALIDATE_OR_GOTO("bit-rot", linked_inode, out);
+ GF_VALIDATE_OR_GOTO("bit-rot", fd, out);
+
+ this = object->this;
+
+ md = GF_MALLOC(SHA256_DIGEST_LENGTH, gf_common_mt_char);
+ if (!md) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_SAVING_HASH_FAILED,
+ "object-gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto out;
+ }
+
+ ret = br_object_checksum(md, object, fd, iatt);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_CALC_CHECKSUM_FAILED,
+ "object-gfid=%s", uuid_utoa(linked_inode->gfid), NULL);
+ goto free_signature;
+ }
+
+ sign = br_prepare_signature(md, SHA256_DIGEST_LENGTH,
+ BR_SIGNATURE_TYPE_SHA256, object);
+ if (!sign) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_SIGN_FAILED,
+ "object-gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto free_signature;
+ }
+
+ xattr = dict_for_key_value(GLUSTERFS_SET_OBJECT_SIGNATURE, (void *)sign,
+ signature_size(SHA256_DIGEST_LENGTH), _gf_true);
+
+ if (!xattr) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_SET_SIGN_FAILED,
+ "dict-allocation object-gfid=%s", uuid_utoa(fd->inode->gfid),
+ NULL);
+ goto free_isign;
+ }
+
+ ret = syncop_fsetxattr(object->child->xl, fd, xattr, 0, NULL, NULL);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_SET_SIGN_FAILED,
+ "fsetxattr object-gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto unref_dict;
+ }
+
+ ret = 0;
+
+unref_dict:
+ dict_unref(xattr);
+free_isign:
+ GF_FREE(sign);
+free_signature:
+ GF_FREE(md);
+out:
+ return ret;
}
-static int br_object_sign_softerror (int32_t op_errno)
+static int
+br_object_sign_softerror(int32_t op_errno)
{
- return ((op_errno == ENOENT) || (op_errno == ESTALE)
- || (op_errno == ENODATA));
+ return ((op_errno == ENOENT) || (op_errno == ESTALE) ||
+ (op_errno == ENODATA));
}
void
-br_log_object (xlator_t *this, char *op, uuid_t gfid, int32_t op_errno)
+br_log_object(xlator_t *this, char *op, uuid_t gfid, int32_t op_errno)
{
- int softerror = br_object_sign_softerror (op_errno);
- if (softerror) {
- gf_msg_debug (this->name, 0, "%s() failed on object %s "
- "[reason: %s]", op, uuid_utoa (gfid),
- strerror (op_errno));
- } else {
- gf_msg (this->name, GF_LOG_ERROR, op_errno, BRB_MSG_OP_FAILED,
- "%s() failed on object %s", op, uuid_utoa (gfid));
- }
+ int softerror = br_object_sign_softerror(op_errno);
+ if (softerror) {
+ gf_msg_debug(this->name, 0,
+ "%s() failed on object %s "
+ "[reason: %s]",
+ op, uuid_utoa(gfid), strerror(op_errno));
+ } else {
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno, BRB_MSG_OP_FAILED, "op=%s",
+ op, "gfid=%s", uuid_utoa(gfid), NULL);
+ }
}
void
-br_log_object_path (xlator_t *this, char *op,
- const char *path, int32_t op_errno)
+br_log_object_path(xlator_t *this, char *op, const char *path, int32_t op_errno)
{
- int softerror = br_object_sign_softerror (op_errno);
- if (softerror) {
- gf_msg_debug (this->name, 0, "%s() failed on object %s "
- "[reason: %s]", op, path, strerror (op_errno));
- } else {
- gf_msg (this->name, GF_LOG_ERROR, op_errno, BRB_MSG_OP_FAILED,
- "%s() failed on object %s", op, path);
- }
+ int softerror = br_object_sign_softerror(op_errno);
+ if (softerror) {
+ gf_msg_debug(this->name, 0,
+ "%s() failed on object %s "
+ "[reason: %s]",
+ op, path, strerror(op_errno));
+ } else {
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno, BRB_MSG_OP_FAILED, "op=%s",
+ op, "path=%s", path, NULL);
+ }
}
static void
-br_trigger_sign (xlator_t *this, br_child_t *child,
- inode_t *linked_inode, loc_t *loc, gf_boolean_t need_reopen)
+br_trigger_sign(xlator_t *this, br_child_t *child, inode_t *linked_inode,
+ loc_t *loc, gf_boolean_t need_reopen)
{
- fd_t *fd = NULL;
- int32_t ret = -1;
- uint32_t val = 0;
- dict_t *dict = NULL;
- pid_t pid = GF_CLIENT_PID_BITD;
-
- syncopctx_setfspid (&pid);
-
- val = (need_reopen == _gf_true) ? BR_OBJECT_REOPEN : BR_OBJECT_RESIGN;
-
- dict = dict_new ();
- if (!dict)
- goto out;
-
- ret = dict_set_uint32 (dict, BR_REOPEN_SIGN_HINT_KEY, val);
- if (ret)
- goto cleanup_dict;
-
- ret = -1;
- fd = fd_create (linked_inode, 0);
- if (!fd) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_FD_CREATE_FAILED,
- "Failed to create fd [GFID %s]",
- uuid_utoa (linked_inode->gfid));
- goto cleanup_dict;
- }
-
- ret = syncop_open (child->xl, loc, O_RDWR, fd, NULL, NULL);
- if (ret) {
- br_log_object (this, "open", linked_inode->gfid, -ret);
- goto unref_fd;
- }
-
- fd_bind (fd);
-
- ret = syncop_fsetxattr (child->xl, fd, dict, 0, NULL, NULL);
- if (ret)
- br_log_object (this, "fsetxattr", linked_inode->gfid, -ret);
-
- /* passthough: fd_unref() */
-
- unref_fd:
- fd_unref (fd);
- cleanup_dict:
- dict_unref (dict);
- out:
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, 0, BRB_MSG_TRIGGER_SIGN,
- "Could not trigger signingd for %s (reopen hint: %d)",
- uuid_utoa (linked_inode->gfid), val);
- }
+ fd_t *fd = NULL;
+ int32_t ret = -1;
+ uint32_t val = 0;
+ dict_t *dict = NULL;
+ pid_t pid = GF_CLIENT_PID_BITD;
+
+ syncopctx_setfspid(&pid);
+
+ val = (need_reopen == _gf_true) ? BR_OBJECT_REOPEN : BR_OBJECT_RESIGN;
+
+ dict = dict_new();
+ if (!dict)
+ goto out;
+
+ ret = dict_set_uint32(dict, BR_REOPEN_SIGN_HINT_KEY, val);
+ if (ret)
+ goto cleanup_dict;
+
+ ret = -1;
+ fd = fd_create(linked_inode, 0);
+ if (!fd) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_FD_CREATE_FAILED,
+ "gfid=%s", uuid_utoa(linked_inode->gfid), NULL);
+ goto cleanup_dict;
+ }
+
+ ret = syncop_open(child->xl, loc, O_RDWR, fd, NULL, NULL);
+ if (ret) {
+ br_log_object(this, "open", linked_inode->gfid, -ret);
+ goto unref_fd;
+ }
+
+ fd_bind(fd);
+
+ ret = syncop_fsetxattr(child->xl, fd, dict, 0, NULL, NULL);
+ if (ret)
+ br_log_object(this, "fsetxattr", linked_inode->gfid, -ret);
+
+ /* passthough: fd_unref() */
+
+unref_fd:
+ fd_unref(fd);
+cleanup_dict:
+ dict_unref(dict);
+out:
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRB_MSG_TRIGGER_SIGN_FAILED,
+ "gfid=%s", uuid_utoa(linked_inode->gfid), "reopen-hint-val=%d",
+ val, NULL);
+ }
}
static void
-br_object_resign (xlator_t *this,
- br_object_t *object, inode_t *linked_inode)
+br_object_resign(xlator_t *this, br_object_t *object, inode_t *linked_inode)
{
- loc_t loc = {0, };
+ loc_t loc = {
+ 0,
+ };
- loc.inode = inode_ref (linked_inode);
- gf_uuid_copy (loc.gfid, linked_inode->gfid);
+ loc.inode = inode_ref(linked_inode);
+ gf_uuid_copy(loc.gfid, linked_inode->gfid);
- br_trigger_sign (this, object->child, linked_inode, &loc, _gf_false);
+ br_trigger_sign(this, object->child, linked_inode, &loc, _gf_false);
- loc_wipe (&loc);
+ loc_wipe(&loc);
}
/**
@@ -560,122 +553,123 @@ br_object_resign (xlator_t *this,
* some form of priority scheduling and/or read burstness to avoid starving
* (or kicking) client I/O's.
*/
-static int32_t br_sign_object (br_object_t *object)
+static int32_t
+br_sign_object(br_object_t *object)
{
- int32_t ret = -1;
- inode_t *linked_inode = NULL;
- xlator_t *this = NULL;
- fd_t *fd = NULL;
- struct iatt iatt = {0, };
- pid_t pid = GF_CLIENT_PID_BITD;
- br_sign_state_t sign_info = BR_SIGN_NORMAL;
-
- GF_VALIDATE_OR_GOTO ("bit-rot", object, out);
-
- this = object->this;
-
- /**
- * FIXME: This is required as signing an object is restricted to
- * clients with special frame->root->pid. Change the way client
- * pid is set.
- */
- syncopctx_setfspid (&pid);
-
- ret = br_object_lookup (this, object, &iatt, &linked_inode);
- if (ret) {
- br_log_object (this, "lookup", object->gfid, -ret);
- goto out;
- }
-
- /**
- * For fd's that have notified for reopening, we send an explicit
- * open() followed by a dummy write() call. This triggers the
- * actual signing of the object.
- */
- sign_info = ntohl (object->sign_info);
- if (sign_info == BR_SIGN_REOPEN_WAIT) {
- br_object_resign (this, object, linked_inode);
- goto unref_inode;
- }
-
- ret = br_object_open (this, object, linked_inode, &fd);
- if (!fd) {
- br_log_object (this, "open", object->gfid, -ret);
- goto unref_inode;
- }
-
- /**
- * we have an open file descriptor on the object. from here on,
- * do not be generous to file operation errors.
- */
- gf_msg_debug (this->name, 0, "Signing object [%s]",
- uuid_utoa (linked_inode->gfid));
-
- ret = br_object_read_sign (linked_inode, fd, object, &iatt);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRB_MSG_READ_AND_SIGN_FAILED, "reading and signing of "
- "the object %s failed", uuid_utoa (linked_inode->gfid));
- goto unref_fd;
- }
-
- ret = 0;
-
- unref_fd:
- fd_unref (fd);
- unref_inode:
- inode_unref (linked_inode);
- out:
- return ret;
+ int32_t ret = -1;
+ inode_t *linked_inode = NULL;
+ xlator_t *this = NULL;
+ fd_t *fd = NULL;
+ struct iatt iatt = {
+ 0,
+ };
+ pid_t pid = GF_CLIENT_PID_BITD;
+ br_sign_state_t sign_info = BR_SIGN_NORMAL;
+
+ GF_VALIDATE_OR_GOTO("bit-rot", object, out);
+
+ this = object->this;
+
+ /**
+ * FIXME: This is required as signing an object is restricted to
+ * clients with special frame->root->pid. Change the way client
+ * pid is set.
+ */
+ syncopctx_setfspid(&pid);
+
+ ret = br_object_lookup(this, object, &iatt, &linked_inode);
+ if (ret) {
+ br_log_object(this, "lookup", object->gfid, -ret);
+ goto out;
+ }
+
+ /**
+ * For fd's that have notified for reopening, we send an explicit
+ * open() followed by a dummy write() call. This triggers the
+ * actual signing of the object.
+ */
+ sign_info = ntohl(object->sign_info);
+ if (sign_info == BR_SIGN_REOPEN_WAIT) {
+ br_object_resign(this, object, linked_inode);
+ goto unref_inode;
+ }
+
+ ret = br_object_open(this, object, linked_inode, &fd);
+ if (!fd) {
+ br_log_object(this, "open", object->gfid, -ret);
+ goto unref_inode;
+ }
+
+ /**
+ * we have an open file descriptor on the object. from here on,
+ * do not be generous to file operation errors.
+ */
+ gf_msg_debug(this->name, 0, "Signing object [%s]",
+ uuid_utoa(linked_inode->gfid));
+
+ ret = br_object_read_sign(linked_inode, fd, object, &iatt);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_READ_AND_SIGN_FAILED,
+ "gfid=%s", uuid_utoa(linked_inode->gfid), NULL);
+ goto unref_fd;
+ }
+
+ ret = 0;
+
+unref_fd:
+ fd_unref(fd);
+unref_inode:
+ inode_unref(linked_inode);
+out:
+ return ret;
}
-static br_object_t *__br_pick_object (br_private_t *priv)
+static br_object_t *
+__br_pick_object(br_private_t *priv)
{
- br_object_t *object = NULL;
+ br_object_t *object = NULL;
- while (list_empty (&priv->obj_queue->objects)) {
- pthread_cond_wait (&priv->object_cond, &priv->lock);
- }
+ while (list_empty(&priv->obj_queue->objects)) {
+ pthread_cond_wait(&priv->object_cond, &priv->lock);
+ }
- object = list_first_entry
- (&priv->obj_queue->objects, br_object_t, list);
- list_del_init (&object->list);
+ object = list_first_entry(&priv->obj_queue->objects, br_object_t, list);
+ list_del_init(&object->list);
- return object;
+ return object;
}
/**
* This is the place where the signing of the objects is triggered.
*/
void *
-br_process_object (void *arg)
+br_process_object(void *arg)
{
- xlator_t *this = NULL;
- br_object_t *object = NULL;
- br_private_t *priv = NULL;
- int32_t ret = -1;
-
- this = arg;
- priv = this->private;
-
- THIS = this;
-
- for (;;) {
- pthread_mutex_lock (&priv->lock);
- {
- object = __br_pick_object (priv);
- }
- pthread_mutex_unlock (&priv->lock);
-
- ret = br_sign_object (object);
- if (ret && !br_object_sign_softerror (-ret))
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRB_MSG_SIGN_FAILED, "SIGNING FAILURE [%s]",
- uuid_utoa (object->gfid));
- GF_FREE (object);
+ xlator_t *this = NULL;
+ br_object_t *object = NULL;
+ br_private_t *priv = NULL;
+ int32_t ret = -1;
+
+ this = arg;
+ priv = this->private;
+
+ THIS = this;
+
+ for (;;) {
+ pthread_mutex_lock(&priv->lock);
+ {
+ object = __br_pick_object(priv);
}
+ pthread_mutex_unlock(&priv->lock);
- return NULL;
+ ret = br_sign_object(object);
+ if (ret && !br_object_sign_softerror(-ret))
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_SET_SIGN_FAILED,
+ "gfid=%s", uuid_utoa(object->gfid), NULL);
+ GF_FREE(object);
+ }
+
+ return NULL;
}
/**
@@ -692,98 +686,97 @@ br_process_object (void *arg)
* NOTE: use call_time to instrument signing time in br_sign_object().
*/
void
-br_add_object_to_queue (struct gf_tw_timer_list *timer,
- void *data, unsigned long call_time)
+br_add_object_to_queue(struct gf_tw_timer_list *timer, void *data,
+ unsigned long call_time)
{
- br_object_t *object = NULL;
- xlator_t *this = NULL;
- br_private_t *priv = NULL;
-
- object = data;
- this = object->this;
- priv = this->private;
-
- THIS = this;
-
- pthread_mutex_lock (&priv->lock);
- {
- list_add_tail (&object->list, &priv->obj_queue->objects);
- pthread_cond_broadcast (&priv->object_cond);
- }
- pthread_mutex_unlock (&priv->lock);
-
- if (timer)
- mem_put (timer);
- return;
+ br_object_t *object = NULL;
+ xlator_t *this = NULL;
+ br_private_t *priv = NULL;
+
+ object = data;
+ this = object->this;
+ priv = this->private;
+
+ THIS = this;
+
+ pthread_mutex_lock(&priv->lock);
+ {
+ list_add_tail(&object->list, &priv->obj_queue->objects);
+ pthread_cond_broadcast(&priv->object_cond);
+ }
+ pthread_mutex_unlock(&priv->lock);
+
+ if (timer)
+ mem_put(timer);
+ return;
}
static br_object_t *
-br_initialize_object (xlator_t *this, br_child_t *child, changelog_event_t *ev)
+br_initialize_object(xlator_t *this, br_child_t *child, changelog_event_t *ev)
{
- br_object_t *object = NULL;
+ br_object_t *object = NULL;
- object = GF_CALLOC (1, sizeof (*object), gf_br_mt_br_object_t);
- if (!object)
- goto out;
- INIT_LIST_HEAD (&object->list);
+ object = GF_CALLOC(1, sizeof(*object), gf_br_mt_br_object_t);
+ if (!object)
+ goto out;
+ INIT_LIST_HEAD(&object->list);
- object->this = this;
- object->child = child;
- gf_uuid_copy (object->gfid, ev->u.releasebr.gfid);
+ object->this = this;
+ object->child = child;
+ gf_uuid_copy(object->gfid, ev->u.releasebr.gfid);
- /* NOTE: it's BE, but no worry */
- object->signedversion = ev->u.releasebr.version;
- object->sign_info = ev->u.releasebr.sign_info;
+ /* NOTE: it's BE, but no worry */
+ object->signedversion = ev->u.releasebr.version;
+ object->sign_info = ev->u.releasebr.sign_info;
out:
- return object;
+ return object;
}
static struct gf_tw_timer_list *
-br_initialize_timer (xlator_t *this, br_object_t *object, br_child_t *child,
- changelog_event_t *ev)
+br_initialize_timer(xlator_t *this, br_object_t *object, br_child_t *child,
+ changelog_event_t *ev)
{
- br_private_t *priv = NULL;
- struct gf_tw_timer_list *timer = NULL;
+ br_private_t *priv = NULL;
+ struct gf_tw_timer_list *timer = NULL;
- priv = this->private;
+ priv = this->private;
- timer = mem_get0 (child->timer_pool);
- if (!timer)
- goto out;
- INIT_LIST_HEAD (&timer->entry);
+ timer = mem_get0(child->timer_pool);
+ if (!timer)
+ goto out;
+ INIT_LIST_HEAD(&timer->entry);
- timer->expires = priv->expiry_time;
- if (!timer->expires)
- timer->expires = 1;
+ timer->expires = priv->expiry_time;
+ if (!timer->expires)
+ timer->expires = 1;
- timer->data = object;
- timer->function = br_add_object_to_queue;
- gf_tw_add_timer (priv->timer_wheel, timer);
+ timer->data = object;
+ timer->function = br_add_object_to_queue;
+ gf_tw_add_timer(priv->timer_wheel, timer);
out:
- return timer;
+ return timer;
}
static int32_t
-br_schedule_object_reopen (xlator_t *this, br_object_t *object,
- br_child_t *child, changelog_event_t *ev)
+br_schedule_object_reopen(xlator_t *this, br_object_t *object,
+ br_child_t *child, changelog_event_t *ev)
{
- struct gf_tw_timer_list *timer = NULL;
-
- timer = br_initialize_timer (this, object, child, ev);
- if (!timer)
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_SET_TIMER_FAILED,
- "Failed to allocate object expiry timer [GFID: %s]",
- uuid_utoa (object->gfid));
- return timer ? 0 : -1;
+ struct gf_tw_timer_list *timer = NULL;
+
+ timer = br_initialize_timer(this, object, child, ev);
+ if (!timer)
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_SET_TIMER_FAILED,
+ "gfid=%s", uuid_utoa(object->gfid), NULL);
+ return timer ? 0 : -1;
}
static int32_t
-br_object_quicksign (xlator_t *this, br_object_t *object)
+br_object_quicksign(xlator_t *this, br_object_t *object)
{
- br_add_object_to_queue (NULL, object, 0ULL);
- return 0;
+ br_add_object_to_queue(NULL, object, 0ULL);
+ return 0;
}
/**
@@ -796,148 +789,146 @@ br_object_quicksign (xlator_t *this, br_object_t *object)
* object as a single alloc and bifurcate their respective pointers.
*/
void
-br_brick_callback (void *xl, char *brick,
- void *data, changelog_event_t *ev)
+br_brick_callback(void *xl, char *brick, void *data, changelog_event_t *ev)
{
- int32_t ret = 0;
- uuid_t gfid = {0,};
- xlator_t *this = NULL;
- br_object_t *object = NULL;
- br_child_t *child = NULL;
- br_sign_state_t sign_info = BR_SIGN_INVALID;
-
- this = xl;
-
- GF_VALIDATE_OR_GOTO (this->name, ev, out);
- GF_VALIDATE_OR_GOTO ("bit-rot", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
-
- GF_ASSERT (ev->ev_type == CHANGELOG_OP_TYPE_BR_RELEASE);
- GF_ASSERT (!gf_uuid_is_null (ev->u.releasebr.gfid));
-
- gf_uuid_copy (gfid, ev->u.releasebr.gfid);
-
- gf_msg_debug (this->name, 0, "RELEASE EVENT [GFID %s]",
- uuid_utoa (gfid));
-
- child = br_get_child_from_brick_path (this, brick);
- if (!child) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_SUBVOL_FAILED,
- "failed to get the subvolume for the brick %s", brick);
- goto out;
- }
-
- object = br_initialize_object (this, child, ev);
- if (!object) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY,
- "failed to allocate object memory [GFID: %s]",
- uuid_utoa (gfid));
- goto out;
- }
-
- /* sanity check */
- sign_info = ntohl (object->sign_info);
- GF_ASSERT (sign_info != BR_SIGN_NORMAL);
-
- if (sign_info == BR_SIGN_REOPEN_WAIT)
- ret = br_schedule_object_reopen (this, object, child, ev);
- else
- ret = br_object_quicksign (this, object);
-
- if (ret)
- goto free_object;
-
- gf_msg_debug (this->name, 0, "->callback: brick [%s], type [%d]\n",
- brick, ev->ev_type);
- return;
-
- free_object:
- GF_FREE (object);
- out:
- return;
+ int32_t ret = 0;
+ uuid_t gfid = {
+ 0,
+ };
+ xlator_t *this = NULL;
+ br_object_t *object = NULL;
+ br_child_t *child = NULL;
+ br_sign_state_t sign_info = BR_SIGN_INVALID;
+
+ this = xl;
+
+ GF_VALIDATE_OR_GOTO(this->name, ev, out);
+ GF_VALIDATE_OR_GOTO("bit-rot", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
+
+ GF_ASSERT(ev->ev_type == CHANGELOG_OP_TYPE_BR_RELEASE);
+ GF_ASSERT(!gf_uuid_is_null(ev->u.releasebr.gfid));
+
+ gf_uuid_copy(gfid, ev->u.releasebr.gfid);
+
+ gf_msg_debug(this->name, 0, "RELEASE EVENT [GFID %s]", uuid_utoa(gfid));
+
+ child = br_get_child_from_brick_path(this, brick);
+ if (!child) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_SUBVOL_FAILED,
+ "brick=%s", brick, NULL);
+ goto out;
+ }
+
+ object = br_initialize_object(this, child, ev);
+ if (!object) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY,
+ "object-gfid=%s", uuid_utoa(gfid), NULL);
+ goto out;
+ }
+
+ /* sanity check */
+ sign_info = ntohl(object->sign_info);
+ GF_ASSERT(sign_info != BR_SIGN_NORMAL);
+
+ if (sign_info == BR_SIGN_REOPEN_WAIT)
+ ret = br_schedule_object_reopen(this, object, child, ev);
+ else
+ ret = br_object_quicksign(this, object);
+
+ if (ret)
+ goto free_object;
+
+ gf_msg_debug(this->name, 0, "->callback: brick [%s], type [%d]\n", brick,
+ ev->ev_type);
+ return;
+
+free_object:
+ GF_FREE(object);
+out:
+ return;
}
void
-br_fill_brick_spec (struct gf_brick_spec *brick, char *path)
+br_fill_brick_spec(struct gf_brick_spec *brick, char *path)
{
- brick->brick_path = gf_strdup (path);
- brick->filter = CHANGELOG_OP_TYPE_BR_RELEASE;
-
- brick->init = br_brick_init;
- brick->fini = br_brick_fini;
- brick->callback = br_brick_callback;
- brick->connected = NULL;
- brick->disconnected = NULL;
+ brick->brick_path = gf_strdup(path);
+ brick->filter = CHANGELOG_OP_TYPE_BR_RELEASE;
+
+ brick->init = br_brick_init;
+ brick->fini = br_brick_fini;
+ brick->callback = br_brick_callback;
+ brick->connected = NULL;
+ brick->disconnected = NULL;
}
static gf_boolean_t
-br_check_object_need_sign (xlator_t *this, dict_t *xattr, br_child_t *child)
+br_check_object_need_sign(xlator_t *this, dict_t *xattr, br_child_t *child)
{
- int32_t ret = -1;
- gf_boolean_t need_sign = _gf_false;
- br_isignature_out_t *sign = NULL;
-
- GF_VALIDATE_OR_GOTO ("bit-rot", this, out);
- GF_VALIDATE_OR_GOTO (this->name, xattr, out);
- GF_VALIDATE_OR_GOTO (this->name, child, out);
-
- ret = dict_get_ptr (xattr, GLUSTERFS_GET_OBJECT_SIGNATURE,
- (void **)&sign);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_SIGN_FAILED,
- "failed to get object signature info");
- goto out;
- }
+ int32_t ret = -1;
+ gf_boolean_t need_sign = _gf_false;
+ br_isignature_out_t *sign = NULL;
- /* Object has been opened and hence dirty. Do not sign it */
- if (sign->stale)
- need_sign = _gf_true;
+ GF_VALIDATE_OR_GOTO("bit-rot", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, xattr, out);
+ GF_VALIDATE_OR_GOTO(this->name, child, out);
-out:
- return need_sign;
-}
+ ret = dict_get_ptr(xattr, GLUSTERFS_GET_OBJECT_SIGNATURE, (void **)&sign);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_SIGN_FAILED,
+ "object-info", NULL);
+ goto out;
+ }
+ /* Object has been opened and hence dirty. Do not sign it */
+ if (sign->stale)
+ need_sign = _gf_true;
+out:
+ return need_sign;
+}
int32_t
-br_prepare_loc (xlator_t *this, br_child_t *child, loc_t *parent,
- gf_dirent_t *entry, loc_t *loc)
+br_prepare_loc(xlator_t *this, br_child_t *child, loc_t *parent,
+ gf_dirent_t *entry, loc_t *loc)
{
- int32_t ret = -1;
- inode_t *inode = NULL;
-
- inode = inode_grep (child->table, parent->inode, entry->d_name);
- if (!inode)
- loc->inode = inode_new (child->table);
- else {
- loc->inode = inode;
- if (loc->inode->ia_type != IA_IFREG) {
- gf_msg_debug (this->name, 0, "%s is not a regular "
- "file", entry->d_name);
- ret = 0;
- goto out;
- }
- }
-
- loc->parent = inode_ref (parent->inode);
- gf_uuid_copy (loc->pargfid, parent->inode->gfid);
-
- ret = inode_path (parent->inode, entry->d_name, (char **)&loc->path);
- if (ret < 0 || !loc->path) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_PATH_FAILED,
- "inode_path on %s (parent: %s) failed", entry->d_name,
- uuid_utoa (parent->inode->gfid));
- goto out;
- }
-
- loc->name = strrchr (loc->path, '/');
- if (loc->name)
- loc->name++;
-
- ret = 1;
+ int32_t ret = -1;
+ inode_t *inode = NULL;
+
+ inode = inode_grep(child->table, parent->inode, entry->d_name);
+ if (!inode)
+ loc->inode = inode_new(child->table);
+ else {
+ loc->inode = inode;
+ if (loc->inode->ia_type != IA_IFREG) {
+ gf_msg_debug(this->name, 0,
+ "%s is not a regular "
+ "file",
+ entry->d_name);
+ ret = 0;
+ goto out;
+ }
+ }
+
+ loc->parent = inode_ref(parent->inode);
+ gf_uuid_copy(loc->pargfid, parent->inode->gfid);
+
+ ret = inode_path(parent->inode, entry->d_name, (char **)&loc->path);
+ if (ret < 0 || !loc->path) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_PATH_FAILED,
+ "inode_path=%s", entry->d_name, "parent-gfid=%s",
+ uuid_utoa(parent->inode->gfid), NULL);
+ goto out;
+ }
+
+ loc->name = strrchr(loc->path, '/');
+ if (loc->name)
+ loc->name++;
+
+ ret = 1;
out:
- return ret;
+ return ret;
}
/**
@@ -947,154 +938,174 @@ out:
* last run for whatever reason (node crashes, reboots, etc..) become
* candidates for signing. This allows the signature to "catch up" with
* the current state of the object. Triggering signing is easy: perform
- * an open() followed by a close() therby resulting in call boomerang.
+ * an open() followed by a close() thereby resulting in call boomerang.
* (though not back to itself :))
*/
int
-bitd_oneshot_crawl (xlator_t *subvol,
- gf_dirent_t *entry, loc_t *parent, void *data)
+bitd_oneshot_crawl(xlator_t *subvol, gf_dirent_t *entry, loc_t *parent,
+ void *data)
{
- int op_errno = 0;
- br_child_t *child = NULL;
- xlator_t *this = NULL;
- loc_t loc = {0, };
- struct iatt iatt = {0, };
- struct iatt parent_buf = {0, };
- dict_t *xattr = NULL;
- int32_t ret = -1;
- inode_t *linked_inode = NULL;
- gf_boolean_t need_signing = _gf_false;
-
- GF_VALIDATE_OR_GOTO ("bit-rot", subvol, out);
- GF_VALIDATE_OR_GOTO ("bit-rot", data, out);
-
- child = data;
- this = child->this;
-
- ret = br_prepare_loc (this, child, parent, entry, &loc);
- if (!ret)
- goto out;
-
- ret = syncop_lookup (child->xl, &loc, &iatt, &parent_buf, NULL, NULL);
- if (ret) {
- br_log_object_path (this, "lookup", loc.path, -ret);
- goto out;
- }
-
- linked_inode = inode_link (loc.inode, parent->inode, loc.name, &iatt);
- if (linked_inode)
- inode_lookup (linked_inode);
-
- if (iatt.ia_type != IA_IFREG) {
- gf_msg_debug (this->name, 0, "%s is not a regular file, "
- "skipping..", entry->d_name);
- ret = 0;
- goto unref_inode;
- }
+ int op_errno = 0;
+ br_child_t *child = NULL;
+ xlator_t *this = NULL;
+ loc_t loc = {
+ 0,
+ };
+ struct iatt iatt = {
+ 0,
+ };
+ struct iatt parent_buf = {
+ 0,
+ };
+ dict_t *xattr = NULL;
+ int32_t ret = -1;
+ inode_t *linked_inode = NULL;
+ gf_boolean_t need_signing = _gf_false;
+ gf_boolean_t need_reopen = _gf_true;
+
+ GF_VALIDATE_OR_GOTO("bit-rot", subvol, out);
+ GF_VALIDATE_OR_GOTO("bit-rot", data, out);
+
+ child = data;
+ this = child->this;
+
+ ret = br_prepare_loc(this, child, parent, entry, &loc);
+ if (!ret)
+ goto out;
+
+ ret = syncop_lookup(child->xl, &loc, &iatt, &parent_buf, NULL, NULL);
+ if (ret) {
+ br_log_object_path(this, "lookup", loc.path, -ret);
+ goto out;
+ }
+
+ linked_inode = inode_link(loc.inode, parent->inode, loc.name, &iatt);
+ if (linked_inode)
+ inode_lookup(linked_inode);
+
+ if (iatt.ia_type != IA_IFREG) {
+ gf_msg_debug(this->name, 0,
+ "%s is not a regular file, "
+ "skipping..",
+ entry->d_name);
+ ret = 0;
+ goto unref_inode;
+ }
+
+ /**
+ * As of now, 2 cases are possible and handled.
+ * 1) GlusterFS is upgraded from a previous version which does not
+ * have any idea about bit-rot and have data in the filesystem.
+ * In this case syncop_getxattr fails with ENODATA and the object
+ * is signed. (In real, when crawler sends lookup, bit-rot-stub
+ * creates the xattrs before returning lookup reply)
+ * 2) Bit-rot was not enabled or BitD was does for some reasons, during
+ * which some files were created, but since BitD was down, were not
+ * signed.
+ * If the file was just created and was being written some data when
+ * the down BitD came up, then bit-rot stub should be intelligent to
+ * identify this case (by comparing the ongoing version or by checking
+ * if there are any fds present for that inode) and handle properly.
+ */
+
+ if (bitd_is_bad_file(this, child, &loc, NULL)) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRB_MSG_SKIP_OBJECT, "path=%s",
+ loc.path, NULL);
+ goto unref_inode;
+ }
+
+ ret = syncop_getxattr(child->xl, &loc, &xattr,
+ GLUSTERFS_GET_OBJECT_SIGNATURE, NULL, NULL);
+ if (ret < 0) {
+ op_errno = -ret;
+ br_log_object(this, "getxattr", linked_inode->gfid, op_errno);
/**
- * As of now, 2 cases are possible and handled.
- * 1) GlusterFS is upgraded from a previous version which does not
- * have any idea about bit-rot and have data in the filesystem.
- * In this case syncop_getxattr fails with ENODATA and the object
- * is signed. (In real, when crawler sends lookup, bit-rot-stub
- * creates the xattrs before returning lookup reply)
- * 2) Bit-rot was not enabled or BitD was dows for some reasons, during
- * which some files were created, but since BitD was down, were not
- * signed.
- * If the file was just created and was being written some data when
- * the down BitD came up, then bit-rot stub should be intelligent to
- * identify this case (by comparing the ongoing version or by checking
- * if there are any fds present for that inode) and handle properly.
+ * No need to sign the zero byte objects as the signing
+ * happens upon first modification of the object.
*/
+ if (op_errno == ENODATA && (iatt.ia_size != 0))
+ need_signing = _gf_true;
+ if (op_errno == EINVAL)
+ gf_smsg(this->name, GF_LOG_WARNING, 0,
+ BRB_MSG_PARTIAL_VERSION_PRESENCE, "gfid=%s",
+ uuid_utoa(linked_inode->gfid), NULL);
+ } else {
+ need_signing = br_check_object_need_sign(this, xattr, child);
- if (bitd_is_bad_file (this, child, &loc, NULL)) {
- gf_msg (this->name, GF_LOG_WARNING, 0, BRB_MSG_SKIP_OBJECT,
- "Entry [%s] is marked corrupted.. skipping.", loc.path);
- goto unref_inode;
- }
-
- ret = syncop_getxattr (child->xl, &loc, &xattr,
- GLUSTERFS_GET_OBJECT_SIGNATURE, NULL, NULL);
- if (ret < 0) {
- op_errno = -ret;
- br_log_object (this, "getxattr", linked_inode->gfid, op_errno);
-
- /**
- * No need to sign the zero byte objects as the signing
- * happens upon first modification of the object.
- */
- if (op_errno == ENODATA && (iatt.ia_size != 0))
- need_signing = _gf_true;
- if (op_errno == EINVAL)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- BRB_MSG_PARTIAL_VERSION_PRESENCE, "Partial "
- "version xattr presence detected, ignoring "
- "[GFID: %s]", uuid_utoa (linked_inode->gfid));
- } else {
- need_signing = br_check_object_need_sign (this, xattr, child);
+ /*
+ * If we are here means, bitrot daemon has started. Is it just
+ * a simple restart of the daemon or is it started because the
+ * feature is enabled is something hard to determine. Hence,
+ * if need_signing is false (because bit-rot version and signature
+ * are present), then still go ahead and sign it.
+ */
+ if (!need_signing) {
+ need_signing = _gf_true;
+ need_reopen = _gf_true;
}
+ }
- if (!need_signing)
- goto unref_dict;
+ if (!need_signing)
+ goto unref_dict;
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_TRIGGER_SIGN,
- "Triggering signing for %s [GFID: %s | Brick: %s]",
- loc.path, uuid_utoa (linked_inode->gfid), child->brick_path);
- br_trigger_sign (this, child, linked_inode, &loc, _gf_true);
+ gf_smsg(this->name, GF_LOG_INFO, 0, BRB_MSG_TRIGGER_SIGN, "path=%s",
+ loc.path, "gfid=%s", uuid_utoa(linked_inode->gfid), "Brick-path=%s",
+ child->brick_path, NULL);
+ br_trigger_sign(this, child, linked_inode, &loc, need_reopen);
- ret = 0;
+ ret = 0;
- unref_dict:
- if (xattr)
- dict_unref (xattr);
- unref_inode:
- inode_unref (linked_inode);
- out:
- loc_wipe (&loc);
+unref_dict:
+ if (xattr)
+ dict_unref(xattr);
+unref_inode:
+ inode_unref(linked_inode);
+out:
+ loc_wipe(&loc);
- return ret;
+ return ret;
}
#define BR_CRAWL_THROTTLE_COUNT 50
-#define BR_CRAWL_THROTTLE_ZZZ 5
+#define BR_CRAWL_THROTTLE_ZZZ 5
void *
-br_oneshot_signer (void *arg)
+br_oneshot_signer(void *arg)
{
- loc_t loc = {0,};
- xlator_t *this = NULL;
- br_child_t *child = NULL;
+ loc_t loc = {
+ 0,
+ };
+ xlator_t *this = NULL;
+ br_child_t *child = NULL;
- child = arg;
- this = child->this;
+ child = arg;
+ this = child->this;
- THIS = this;
+ THIS = this;
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_CRAWLING_START,
- "Crawling brick [%s], scanning for unsigned objects",
- child->brick_path);
+ gf_smsg(this->name, GF_LOG_INFO, 0, BRB_MSG_CRAWLING_START, "brick-path=%s",
+ child->brick_path, NULL);
- loc.inode = child->table->root;
- (void) syncop_ftw_throttle
- (child->xl, &loc,
- GF_CLIENT_PID_BITD, child, bitd_oneshot_crawl,
- BR_CRAWL_THROTTLE_COUNT, BR_CRAWL_THROTTLE_ZZZ);
+ loc.inode = child->table->root;
+ (void)syncop_ftw_throttle(child->xl, &loc, GF_CLIENT_PID_BITD, child,
+ bitd_oneshot_crawl, BR_CRAWL_THROTTLE_COUNT,
+ BR_CRAWL_THROTTLE_ZZZ);
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_CRAWLING_FINISH,
- "Completed crawling brick [%s]", child->brick_path);
+ gf_smsg(this->name, GF_LOG_INFO, 0, BRB_MSG_CRAWLING_FINISH,
+ "brick-path=%s", child->brick_path, NULL);
- return NULL;
+ return NULL;
}
static void
-br_set_child_state (br_child_t *child, br_child_state_t state)
+br_set_child_state(br_child_t *child, br_child_state_t state)
{
- pthread_mutex_lock (&child->lock);
- {
- _br_set_child_state (child, state);
- }
- pthread_mutex_unlock (&child->lock);
+ pthread_mutex_lock(&child->lock);
+ {
+ _br_set_child_state(child, state);
+ }
+ pthread_mutex_unlock(&child->lock);
}
/**
@@ -1109,160 +1120,157 @@ br_set_child_state (br_child_t *child, br_child_state_t state)
* notifications.
*/
static int32_t
-br_enact_signer (xlator_t *this, br_child_t *child, br_stub_init_t *stub)
+br_enact_signer(xlator_t *this, br_child_t *child, br_stub_init_t *stub)
{
- int32_t ret = 0;
- br_private_t *priv = NULL;
- struct gf_brick_spec *brick = NULL;
-
- priv = this->private;
-
- brick = GF_CALLOC (1, sizeof (struct gf_brick_spec),
- gf_common_mt_gf_brick_spec_t);
- if (!brick)
- goto error_return;
-
- br_fill_brick_spec (brick, stub->export);
- ret = gf_changelog_register_generic
- (brick, 1, 1, this->ctx->cmd_args.log_file, -1, this);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- BRB_MSG_REGISTER_FAILED, "Register to changelog "
- "failed");
- goto dealloc;
- }
-
- child->threadrunning = 0;
- ret = gf_thread_create (&child->thread, NULL, br_oneshot_signer, child,
- "brosign");
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0, BRB_MSG_SPAWN_FAILED,
- "failed to spawn FS crawler thread");
- else
- child->threadrunning = 1;
-
- /* it's OK to continue, "old" objects would be signed when modified */
- list_add_tail (&child->list, &priv->signing);
- return 0;
-
- dealloc:
- GF_FREE (brick);
- error_return:
- return -1;
+ int32_t ret = 0;
+ br_private_t *priv = NULL;
+ struct gf_brick_spec *brick = NULL;
+
+ priv = this->private;
+
+ brick = GF_CALLOC(1, sizeof(struct gf_brick_spec),
+ gf_common_mt_gf_brick_spec_t);
+ if (!brick)
+ goto error_return;
+
+ br_fill_brick_spec(brick, stub->export);
+ ret = gf_changelog_register_generic(brick, 1, 1,
+ this->ctx->cmd_args.log_file, -1, this);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, BRB_MSG_REGISTER_FAILED, NULL);
+ goto dealloc;
+ }
+
+ child->threadrunning = 0;
+ ret = gf_thread_create(&child->thread, NULL, br_oneshot_signer, child,
+ "brosign");
+ if (ret)
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRB_MSG_SPAWN_FAILED,
+ "FS-crawler-thread", NULL);
+ else
+ child->threadrunning = 1;
+
+ /* it's OK to continue, "old" objects would be signed when modified */
+ list_add_tail(&child->list, &priv->signing);
+ return 0;
+
+dealloc:
+ GF_FREE(brick);
+error_return:
+ return -1;
}
static int32_t
-br_launch_scrubber (xlator_t *this, br_child_t *child,
- struct br_scanfs *fsscan, struct br_scrubber *fsscrub)
+br_launch_scrubber(xlator_t *this, br_child_t *child, struct br_scanfs *fsscan,
+ struct br_scrubber *fsscrub)
{
- int32_t ret = -1;
- br_private_t *priv = NULL;
- struct br_monitor *scrub_monitor = NULL;
-
- priv = this->private;
-
- scrub_monitor = &priv->scrub_monitor;
- ret = gf_thread_create (&child->thread, NULL, br_fsscanner, child,
- "brfsscan");
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ALERT, 0, BRB_MSG_SPAWN_FAILED,
- "failed to spawn bitrot scrubber daemon [Brick: %s]",
- child->brick_path);
- goto error_return;
- }
-
- /* Signal monitor to kick off state machine*/
- pthread_mutex_lock (&scrub_monitor->mutex);
- {
- if (!scrub_monitor->inited)
- pthread_cond_signal (&scrub_monitor->cond);
- scrub_monitor->inited = _gf_true;
- }
- pthread_mutex_unlock (&scrub_monitor->mutex);
+ int32_t ret = -1;
+ br_private_t *priv = NULL;
+ struct br_monitor *scrub_monitor = NULL;
+
+ priv = this->private;
+
+ scrub_monitor = &priv->scrub_monitor;
+ ret = gf_thread_create(&child->thread, NULL, br_fsscanner, child,
+ "brfsscan");
+ if (ret != 0) {
+ gf_smsg(this->name, GF_LOG_ALERT, 0, BRB_MSG_SPAWN_FAILED,
+ "bitrot-scrubber-daemon Brick-path=%s", child->brick_path,
+ NULL);
+ goto error_return;
+ }
+
+ /* Signal monitor to kick off state machine*/
+ pthread_mutex_lock(&scrub_monitor->mutex);
+ {
+ if (!scrub_monitor->inited)
+ pthread_cond_signal(&scrub_monitor->cond);
+ scrub_monitor->inited = _gf_true;
+ }
+ pthread_mutex_unlock(&scrub_monitor->mutex);
+
+ /**
+ * Everything has been setup.. add this subvolume to scrubbers
+ * list.
+ */
+ pthread_mutex_lock(&fsscrub->mutex);
+ {
+ list_add_tail(&child->list, &fsscrub->scrublist);
+ pthread_cond_broadcast(&fsscrub->cond);
+ }
+ pthread_mutex_unlock(&fsscrub->mutex);
+
+ return 0;
- /**
- * Everything has been setup.. add this subvolume to scrubbers
- * list.
- */
- pthread_mutex_lock (&fsscrub->mutex);
- {
- list_add_tail (&child->list, &fsscrub->scrublist);
- pthread_cond_broadcast (&fsscrub->cond);
- }
- pthread_mutex_unlock (&fsscrub->mutex);
-
- return 0;
-
- error_return:
- return -1;
+error_return:
+ return -1;
}
static int32_t
-br_enact_scrubber (xlator_t *this, br_child_t *child)
+br_enact_scrubber(xlator_t *this, br_child_t *child)
{
- int32_t ret = 0;
- br_private_t *priv = NULL;
- struct br_scanfs *fsscan = NULL;
- struct br_scrubber *fsscrub = NULL;
+ int32_t ret = 0;
+ br_private_t *priv = NULL;
+ struct br_scanfs *fsscan = NULL;
+ struct br_scrubber *fsscrub = NULL;
- priv = this->private;
+ priv = this->private;
- fsscan = &child->fsscan;
- fsscrub = &priv->fsscrub;
+ fsscan = &child->fsscan;
+ fsscrub = &priv->fsscrub;
- /**
- * if this child already witnesses a successful connection earlier
- * there's no need to initialize mutexes, condvars, etc..
- */
- if (_br_child_witnessed_connection (child))
- return br_launch_scrubber (this, child, fsscan, fsscrub);
+ /**
+ * if this child already witnesses a successful connection earlier
+ * there's no need to initialize mutexes, condvars, etc..
+ */
+ if (_br_child_witnessed_connection(child))
+ return br_launch_scrubber(this, child, fsscan, fsscrub);
- LOCK_INIT (&fsscan->entrylock);
- pthread_mutex_init (&fsscan->waitlock, NULL);
- pthread_cond_init (&fsscan->waitcond, NULL);
+ LOCK_INIT(&fsscan->entrylock);
+ pthread_mutex_init(&fsscan->waitlock, NULL);
+ pthread_cond_init(&fsscan->waitcond, NULL);
- fsscan->entries = 0;
- INIT_LIST_HEAD (&fsscan->queued);
- INIT_LIST_HEAD (&fsscan->ready);
+ fsscan->entries = 0;
+ INIT_LIST_HEAD(&fsscan->queued);
+ INIT_LIST_HEAD(&fsscan->ready);
- ret = br_launch_scrubber (this, child, fsscan, fsscrub);
- if (ret)
- goto error_return;
+ ret = br_launch_scrubber(this, child, fsscan, fsscrub);
+ if (ret)
+ goto error_return;
- return 0;
+ return 0;
- error_return:
- LOCK_DESTROY (&fsscan->entrylock);
- pthread_mutex_destroy (&fsscan->waitlock);
- pthread_cond_destroy (&fsscan->waitcond);
+error_return:
+ LOCK_DESTROY(&fsscan->entrylock);
+ pthread_mutex_destroy(&fsscan->waitlock);
+ pthread_cond_destroy(&fsscan->waitcond);
- return -1;
+ return -1;
}
static int32_t
-br_child_enaction (xlator_t *this, br_child_t *child, br_stub_init_t *stub)
+br_child_enaction(xlator_t *this, br_child_t *child, br_stub_init_t *stub)
{
- int32_t ret = -1;
- br_private_t *priv = this->private;
+ int32_t ret = -1;
+ br_private_t *priv = this->private;
- pthread_mutex_lock (&child->lock);
- {
- if (priv->iamscrubber)
- ret = br_enact_scrubber (this, child);
- else
- ret = br_enact_signer (this, child, stub);
-
- if (!ret) {
- child->witnessed = 1;
- _br_set_child_state (child, BR_CHILD_STATE_CONNECTED);
- gf_msg (this->name, GF_LOG_INFO,
- 0, BRB_MSG_CONNECTED_TO_BRICK,
- "Connected to brick %s..", child->brick_path);
- }
+ pthread_mutex_lock(&child->lock);
+ {
+ if (priv->iamscrubber)
+ ret = br_enact_scrubber(this, child);
+ else
+ ret = br_enact_signer(this, child, stub);
+
+ if (!ret) {
+ child->witnessed = 1;
+ _br_set_child_state(child, BR_CHILD_STATE_CONNECTED);
+ gf_smsg(this->name, GF_LOG_INFO, 0, BRB_MSG_CONNECTED_TO_BRICK,
+ "brick-path=%s", child->brick_path, NULL);
}
- pthread_mutex_unlock (&child->lock);
+ }
+ pthread_mutex_unlock(&child->lock);
- return ret;
+ return ret;
}
/**
@@ -1273,129 +1281,130 @@ br_child_enaction (xlator_t *this, br_child_t *child, br_stub_init_t *stub)
* process either acts as a signer or a scrubber.
*/
int32_t
-br_brick_connect (xlator_t *this, br_child_t *child)
+br_brick_connect(xlator_t *this, br_child_t *child)
{
- int32_t ret = -1;
- loc_t loc = {0, };
- struct iatt buf = {0, };
- struct iatt parent = {0, };
- br_stub_init_t *stub = NULL;
- dict_t *xattr = NULL;
- int op_errno = 0;
-
- GF_VALIDATE_OR_GOTO ("bit-rot", this, out);
- GF_VALIDATE_OR_GOTO (this->name, child, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
-
- br_child_set_scrub_state (child, _gf_false);
- br_set_child_state (child, BR_CHILD_STATE_INITIALIZING);
-
- loc.inode = inode_ref (child->table->root);
- gf_uuid_copy (loc.gfid, loc.inode->gfid);
- loc.path = gf_strdup ("/");
-
- ret = syncop_lookup (child->xl, &loc, &buf, &parent, NULL, NULL);
- if (ret) {
- op_errno = -ret;
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- BRB_MSG_LOOKUP_FAILED, "lookup on root failed");
- goto wipeloc;
- }
-
- ret = syncop_getxattr (child->xl, &loc, &xattr,
- GLUSTERFS_GET_BR_STUB_INIT_TIME, NULL, NULL);
- if (ret) {
- op_errno = -ret;
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- BRB_MSG_GET_INFO_FAILED, "failed to get stub info");
- goto wipeloc;
- }
-
- ret = dict_get_ptr (xattr, GLUSTERFS_GET_BR_STUB_INIT_TIME,
- (void **)&stub);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_INFO_FAILED,
- "failed to extract stub information");
- goto free_dict;
- }
-
- memcpy (child->brick_path, stub->export, strlen (stub->export) + 1);
- child->tv.tv_sec = ntohl (stub->timebuf[0]);
- child->tv.tv_usec = ntohl (stub->timebuf[1]);
-
- ret = br_child_enaction (this, child, stub);
-
- free_dict:
- dict_unref (xattr);
- wipeloc:
- loc_wipe (&loc);
- out:
- if (ret)
- br_set_child_state (child, BR_CHILD_STATE_CONNFAILED);
- return ret;
+ int32_t ret = -1;
+ loc_t loc = {
+ 0,
+ };
+ struct iatt buf = {
+ 0,
+ };
+ struct iatt parent = {
+ 0,
+ };
+ br_stub_init_t *stub = NULL;
+ dict_t *xattr = NULL;
+ int op_errno = 0;
+
+ GF_VALIDATE_OR_GOTO("bit-rot", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, child, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
+
+ br_child_set_scrub_state(child, _gf_false);
+ br_set_child_state(child, BR_CHILD_STATE_INITIALIZING);
+
+ loc.inode = inode_ref(child->table->root);
+ gf_uuid_copy(loc.gfid, loc.inode->gfid);
+ loc.path = gf_strdup("/");
+
+ ret = syncop_lookup(child->xl, &loc, &buf, &parent, NULL, NULL);
+ if (ret) {
+ op_errno = -ret;
+ ret = -1;
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno, BRB_MSG_LOOKUP_FAILED,
+ NULL);
+ goto wipeloc;
+ }
+
+ ret = syncop_getxattr(child->xl, &loc, &xattr,
+ GLUSTERFS_GET_BR_STUB_INIT_TIME, NULL, NULL);
+ if (ret) {
+ op_errno = -ret;
+ ret = -1;
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno, BRB_MSG_GET_INFO_FAILED,
+ NULL);
+ goto wipeloc;
+ }
+
+ ret = dict_get_ptr(xattr, GLUSTERFS_GET_BR_STUB_INIT_TIME, (void **)&stub);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_GET_INFO_FAILED, NULL);
+ goto free_dict;
+ }
+
+ memcpy(child->brick_path, stub->export, strlen(stub->export) + 1);
+ child->tv.tv_sec = ntohl(stub->timebuf[0]);
+ child->tv.tv_usec = ntohl(stub->timebuf[1]);
+
+ ret = br_child_enaction(this, child, stub);
+
+free_dict:
+ dict_unref(xattr);
+wipeloc:
+ loc_wipe(&loc);
+out:
+ if (ret)
+ br_set_child_state(child, BR_CHILD_STATE_CONNFAILED);
+ return ret;
}
/* TODO: cleanup signer */
static int32_t
-br_cleanup_signer (xlator_t *this, br_child_t *child)
+br_cleanup_signer(xlator_t *this, br_child_t *child)
{
- return 0;
+ return 0;
}
static int32_t
-br_cleanup_scrubber (xlator_t *this, br_child_t *child)
+br_cleanup_scrubber(xlator_t *this, br_child_t *child)
{
- int32_t ret = 0;
- br_private_t *priv = NULL;
- struct br_scrubber *fsscrub = NULL;
- struct br_monitor *scrub_monitor = NULL;
-
- priv = this->private;
- fsscrub = &priv->fsscrub;
- scrub_monitor = &priv->scrub_monitor;
-
- if (_br_is_child_scrub_active (child)) {
- scrub_monitor->active_child_count--;
- br_child_set_scrub_state (child, _gf_false);
- }
-
- /**
- * 0x0: child (brick) goes out of rotation
- *
- * This is fully safe w.r.t. entries for this child being actively
- * scrubbed. Each of the scrubber thread(s) would finish scrubbing
- * the entry (probably failing due to disconnection) and either
- * putting the entry back into the queue or continuing further.
- * Either way, pending entries for this child's queue need not be
- * drained; entries just sit there in the queued/ready list to be
- * consumed later upon re-connection.
- */
- pthread_mutex_lock (&fsscrub->mutex);
- {
- list_del_init (&child->list);
- }
- pthread_mutex_unlock (&fsscrub->mutex);
-
- /**
- * 0x1: cleanup scanner thread
- *
- * The pending timer needs to be removed _after_ cleaning up the
- * filesystem scanner (scheduling the next scrub time is not a
- * cancellation point).
- */
- ret = gf_thread_cleanup_xint (child->thread);
- if (ret)
- gf_msg (this->name, GF_LOG_INFO,
- 0, BRB_MSG_SCRUB_THREAD_CLEANUP,
- "Error cleaning up scanner thread");
-
- gf_msg (this->name, GF_LOG_INFO,
- 0, BRB_MSG_SCRUBBER_CLEANED,
- "Cleaned up scrubber for brick [%s]", child->brick_path);
-
- return 0;
+ int32_t ret = 0;
+ br_private_t *priv = NULL;
+ struct br_scrubber *fsscrub = NULL;
+ struct br_monitor *scrub_monitor = NULL;
+
+ priv = this->private;
+ fsscrub = &priv->fsscrub;
+ scrub_monitor = &priv->scrub_monitor;
+
+ if (_br_is_child_scrub_active(child)) {
+ scrub_monitor->active_child_count--;
+ br_child_set_scrub_state(child, _gf_false);
+ }
+
+ /**
+ * 0x0: child (brick) goes out of rotation
+ *
+ * This is fully safe w.r.t. entries for this child being actively
+ * scrubbed. Each of the scrubber thread(s) would finish scrubbing
+ * the entry (probably failing due to disconnection) and either
+ * putting the entry back into the queue or continuing further.
+ * Either way, pending entries for this child's queue need not be
+ * drained; entries just sit there in the queued/ready list to be
+ * consumed later upon re-connection.
+ */
+ pthread_mutex_lock(&fsscrub->mutex);
+ {
+ list_del_init(&child->list);
+ }
+ pthread_mutex_unlock(&fsscrub->mutex);
+
+ /**
+ * 0x1: cleanup scanner thread
+ *
+ * The pending timer needs to be removed _after_ cleaning up the
+ * filesystem scanner (scheduling the next scrub time is not a
+ * cancellation point).
+ */
+ ret = gf_thread_cleanup_xint(child->thread);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUB_THREAD_CLEANUP, NULL);
+
+ gf_smsg(this->name, GF_LOG_INFO, 0, BRB_MSG_SCRUBBER_CLEANED,
+ "brick-path=%s", child->brick_path, NULL);
+
+ return 0;
}
/**
@@ -1404,38 +1413,38 @@ br_cleanup_scrubber (xlator_t *this, br_child_t *child)
* the inode table, it's just reused taking care of stale inodes)
*/
int32_t
-br_brick_disconnect (xlator_t *this, br_child_t *child)
+br_brick_disconnect(xlator_t *this, br_child_t *child)
{
- int32_t ret = 0;
- struct br_monitor *scrub_monitor = NULL;
- br_private_t *priv = this->private;
+ int32_t ret = 0;
+ struct br_monitor *scrub_monitor = NULL;
+ br_private_t *priv = this->private;
+
+ scrub_monitor = &priv->scrub_monitor;
+
+ /* Lock order should be wakelock and then child lock to
+ * dead locks.
+ */
+ pthread_mutex_lock(&scrub_monitor->wakelock);
+ {
+ pthread_mutex_lock(&child->lock);
+ {
+ if (!_br_is_child_connected(child))
+ goto unblock;
- scrub_monitor = &priv->scrub_monitor;
+ /* child is on death row.. */
+ _br_set_child_state(child, BR_CHILD_STATE_DISCONNECTED);
- /* Lock order should be wakelock and then child lock to
- * dead locks.
- */
- pthread_mutex_lock (&scrub_monitor->wakelock);
- {
- pthread_mutex_lock (&child->lock);
- {
- if (!_br_is_child_connected (child))
- goto unblock;
-
- /* child is on death row.. */
- _br_set_child_state (child, BR_CHILD_STATE_DISCONNECTED);
-
- if (priv->iamscrubber)
- ret = br_cleanup_scrubber (this, child);
- else
- ret = br_cleanup_signer (this, child);
- }
- unblock:
- pthread_mutex_unlock (&child->lock);
+ if (priv->iamscrubber)
+ ret = br_cleanup_scrubber(this, child);
+ else
+ ret = br_cleanup_signer(this, child);
}
- pthread_mutex_unlock (&scrub_monitor->wakelock);
+ unblock:
+ pthread_mutex_unlock(&child->lock);
+ }
+ pthread_mutex_unlock(&scrub_monitor->wakelock);
- return ret;
+ return ret;
}
/**
@@ -1444,338 +1453,341 @@ br_brick_disconnect (xlator_t *this, br_child_t *child)
* information regarding that brick (such as brick path).
*/
void *
-br_handle_events (void *arg)
+br_handle_events(void *arg)
{
- int32_t ret = 0;
- xlator_t *this = NULL;
- br_private_t *priv = NULL;
- br_child_t *child = NULL;
- struct br_child_event *childev = NULL;
-
- this = arg;
- priv = this->private;
+ int32_t ret = 0;
+ xlator_t *this = NULL;
+ br_private_t *priv = NULL;
+ br_child_t *child = NULL;
+ struct br_child_event *childev = NULL;
+
+ this = arg;
+ priv = this->private;
+
+ /*
+ * Since, this is the topmost xlator, THIS has to be set by bit-rot
+ * xlator itself (STACK_WIND won't help in this case). Also it has
+ * to be done for each thread that gets spawned. Otherwise, a new
+ * thread will get global_xlator's pointer when it does "THIS".
+ */
+ THIS = this;
+
+ while (1) {
+ pthread_mutex_lock(&priv->lock);
+ {
+ while (list_empty(&priv->bricks))
+ pthread_cond_wait(&priv->cond, &priv->lock);
- /*
- * Since, this is the topmost xlator, THIS has to be set by bit-rot
- * xlator itself (STACK_WIND wont help in this case). Also it has
- * to be done for each thread that gets spawned. Otherwise, a new
- * thread will get global_xlator's pointer when it does "THIS".
- */
- THIS = this;
-
- while (1) {
- pthread_mutex_lock (&priv->lock);
- {
- while (list_empty (&priv->bricks))
- pthread_cond_wait (&priv->cond, &priv->lock);
-
- childev = list_first_entry
- (&priv->bricks, struct br_child_event, list);
- list_del_init (&childev->list);
- }
- pthread_mutex_unlock (&priv->lock);
-
- child = childev->child;
- ret = childev->call (this, child);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRB_MSG_SUBVOL_CONNECT_FAILED,
- "callback handler for subvolume [%s] failed",
- child->xl->name);
- GF_FREE (childev);
+ childev = list_first_entry(&priv->bricks, struct br_child_event,
+ list);
+ list_del_init(&childev->list);
}
+ pthread_mutex_unlock(&priv->lock);
- return NULL;
+ child = childev->child;
+ ret = childev->call(this, child);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_SUBVOL_CONNECT_FAILED,
+ "name=%s", child->xl->name, NULL);
+ GF_FREE(childev);
+ }
+
+ return NULL;
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int32_t ret = -1;
-
- if (!this)
- return ret;
+ int32_t ret = -1;
- ret = xlator_mem_acct_init (this, gf_br_stub_mt_end + 1);
+ if (!this)
+ return ret;
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_WARNING, 0, BRB_MSG_MEM_ACNT_FAILED,
- "Memory accounting init failed");
- return ret;
- }
+ ret = xlator_mem_acct_init(this, gf_br_stub_mt_end + 1);
+ if (ret != 0) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRB_MSG_MEM_ACNT_FAILED, NULL);
return ret;
+ }
+
+ return ret;
}
static void
-_br_qchild_event (xlator_t *this, br_child_t *child, br_child_handler *call)
+_br_qchild_event(xlator_t *this, br_child_t *child, br_child_handler *call)
{
- br_private_t *priv = NULL;
- struct br_child_event *childev = NULL;
+ br_private_t *priv = NULL;
+ struct br_child_event *childev = NULL;
- priv = this->private;
+ priv = this->private;
- childev = GF_CALLOC (1, sizeof (*childev), gf_br_mt_br_child_event_t);
- if (!childev) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY,
- "Event unhandled for child.. [Brick: %s]",
- child->xl->name);
- return;
- }
+ childev = GF_CALLOC(1, sizeof(*childev), gf_br_mt_br_child_event_t);
+ if (!childev) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_EVENT_UNHANDLED,
+ "Brick-name=%s", child->xl->name, NULL);
+ return;
+ }
- INIT_LIST_HEAD (&childev->list);
- childev->this = this;
- childev->child = child;
- childev->call = call;
+ INIT_LIST_HEAD(&childev->list);
+ childev->this = this;
+ childev->child = child;
+ childev->call = call;
- list_add_tail (&childev->list, &priv->bricks);
+ list_add_tail(&childev->list, &priv->bricks);
}
int
-br_scrubber_status_get (xlator_t *this, dict_t **dict)
+br_scrubber_status_get(xlator_t *this, dict_t **dict)
{
- int ret = -1;
- br_private_t *priv = NULL;
- struct br_scrub_stats *scrub_stats = NULL;
-
- priv = this->private;
-
- GF_VALIDATE_OR_GOTO ("bit-rot", priv, out);
-
- scrub_stats = &priv->scrub_stat;
-
- ret = br_get_bad_objects_list (this, dict);
- if (ret) {
- gf_msg_debug (this->name, 0, "Failed to collect corrupt "
- "files");
- }
-
- ret = dict_set_int8 (*dict, "scrub-running",
- scrub_stats->scrub_running);
- if (ret) {
- gf_msg_debug (this->name, 0, "Failed setting scrub_running "
- "entry to the dictionary");
- }
-
- ret = dict_set_uint64 (*dict, "scrubbed-files",
- scrub_stats->scrubbed_files);
- if (ret) {
- gf_msg_debug (this->name, 0, "Failed to setting scrubbed file "
- "entry to the dictionary");
- }
-
- ret = dict_set_uint64 (*dict, "unsigned-files",
- scrub_stats->unsigned_files);
- if (ret) {
- gf_msg_debug (this->name, 0, "Failed to set unsigned file count"
- " entry to the dictionary");
- }
-
- ret = dict_set_uint64 (*dict, "scrub-duration",
- scrub_stats->scrub_duration);
- if (ret) {
- gf_msg_debug (this->name, 0, "Failed to set scrub duration"
- " entry to the dictionary");
- }
-
- ret = dict_set_dynstr_with_alloc (*dict, "last-scrub-time",
- scrub_stats->last_scrub_time);
- if (ret) {
- gf_msg_debug (this->name, 0, "Failed to set "
- "last scrub time value");
- }
+ int ret = -1;
+ br_private_t *priv = NULL;
+ struct br_scrub_stats *scrub_stats = NULL;
+
+ priv = this->private;
+
+ GF_VALIDATE_OR_GOTO("bit-rot", priv, out);
+
+ scrub_stats = &priv->scrub_stat;
+
+ ret = br_get_bad_objects_list(this, dict);
+ if (ret) {
+ gf_msg_debug(this->name, 0,
+ "Failed to collect corrupt "
+ "files");
+ }
+
+ ret = dict_set_int8(*dict, "scrub-running", scrub_stats->scrub_running);
+ if (ret) {
+ gf_msg_debug(this->name, 0,
+ "Failed setting scrub_running "
+ "entry to the dictionary");
+ }
+
+ ret = dict_set_uint64(*dict, "scrubbed-files", scrub_stats->scrubbed_files);
+ if (ret) {
+ gf_msg_debug(this->name, 0,
+ "Failed to setting scrubbed file "
+ "entry to the dictionary");
+ }
+
+ ret = dict_set_uint64(*dict, "unsigned-files", scrub_stats->unsigned_files);
+ if (ret) {
+ gf_msg_debug(this->name, 0,
+ "Failed to set unsigned file count"
+ " entry to the dictionary");
+ }
+
+ ret = dict_set_uint64(*dict, "scrub-duration", scrub_stats->scrub_duration);
+ if (ret) {
+ gf_msg_debug(this->name, 0,
+ "Failed to set scrub duration"
+ " entry to the dictionary");
+ }
+
+ ret = dict_set_dynstr_with_alloc(*dict, "last-scrub-time",
+ scrub_stats->last_scrub_time);
+ if (ret) {
+ gf_msg_debug(this->name, 0,
+ "Failed to set "
+ "last scrub time value");
+ }
out:
- return ret;
+ return ret;
}
int
-notify (xlator_t *this, int32_t event, void *data, ...)
+notify(xlator_t *this, int32_t event, void *data, ...)
{
- int idx = -1;
- int ret = -1;
- xlator_t *subvol = NULL;
- br_child_t *child = NULL;
- br_private_t *priv = NULL;
- dict_t *output = NULL;
- va_list ap;
- struct br_monitor *scrub_monitor = NULL;
+ int idx = -1;
+ int ret = -1;
+ xlator_t *subvol = NULL;
+ br_child_t *child = NULL;
+ br_private_t *priv = NULL;
+ dict_t *output = NULL;
+ va_list ap;
+ struct br_monitor *scrub_monitor = NULL;
- subvol = (xlator_t *)data;
- priv = this->private;
- scrub_monitor = &priv->scrub_monitor;
+ subvol = (xlator_t *)data;
+ priv = this->private;
+ scrub_monitor = &priv->scrub_monitor;
- gf_msg_trace (this->name, 0, "Notification received: %d", event);
+ gf_msg_trace(this->name, 0, "Notification received: %d", event);
- idx = br_find_child_index (this, subvol);
+ idx = br_find_child_index(this, subvol);
- switch (event) {
+ switch (event) {
case GF_EVENT_CHILD_UP:
- if (idx < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRB_MSG_INVALID_SUBVOL, "Got event %d from "
- "invalid subvolume", event);
- goto out;
- }
-
- pthread_mutex_lock (&priv->lock);
- {
- child = &priv->children[idx];
- if (child->child_up == 1)
- goto unblock_0;
- priv->up_children++;
-
- child->child_up = 1;
- child->xl = subvol;
- if (!child->table)
- child->table = inode_table_new (4096, subvol);
-
- _br_qchild_event (this, child, br_brick_connect);
- pthread_cond_signal (&priv->cond);
- }
+ if (idx < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_INVALID_SUBVOL,
+ "event=%d", event, NULL);
+ goto out;
+ }
+
+ pthread_mutex_lock(&priv->lock);
+ {
+ child = &priv->children[idx];
+ if (child->child_up == 1)
+ goto unblock_0;
+ priv->up_children++;
+
+ child->child_up = 1;
+ child->xl = subvol;
+ if (!child->table)
+ child->table = inode_table_new(4096, subvol);
+
+ _br_qchild_event(this, child, br_brick_connect);
+ pthread_cond_signal(&priv->cond);
+ }
unblock_0:
- pthread_mutex_unlock (&priv->lock);
+ pthread_mutex_unlock(&priv->lock);
- if (priv->up_children == priv->child_count)
- default_notify (this, event, data);
- break;
+ if (priv->up_children == priv->child_count)
+ default_notify(this, event, data);
+ break;
case GF_EVENT_CHILD_DOWN:
- if (idx < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRB_MSG_INVALID_SUBVOL_CHILD,
- "Got event %d from invalid subvolume", event);
- goto out;
- }
-
- pthread_mutex_lock (&priv->lock);
- {
- child = &priv->children[idx];
- if (child->child_up == 0)
- goto unblock_1;
-
- child->child_up = 0;
- priv->up_children--;
-
- _br_qchild_event (this, child, br_brick_disconnect);
- pthread_cond_signal (&priv->cond);
- }
+ if (idx < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_INVALID_SUBVOL,
+ "event=%d", event, NULL);
+ goto out;
+ }
+
+ pthread_mutex_lock(&priv->lock);
+ {
+ child = &priv->children[idx];
+ if (child->child_up == 0)
+ goto unblock_1;
+
+ child->child_up = 0;
+ priv->up_children--;
+
+ _br_qchild_event(this, child, br_brick_disconnect);
+ pthread_cond_signal(&priv->cond);
+ }
unblock_1:
- pthread_mutex_unlock (&priv->lock);
+ pthread_mutex_unlock(&priv->lock);
- if (priv->up_children == 0)
- default_notify (this, event, data);
- break;
+ if (priv->up_children == 0)
+ default_notify(this, event, data);
+ break;
case GF_EVENT_SCRUB_STATUS:
- gf_msg_debug (this->name, GF_LOG_INFO, "BitRot scrub status "
- "called");
- va_start (ap, data);
- output = va_arg (ap, dict_t *);
- va_end (ap);
-
- ret = br_scrubber_status_get (this, &output);
- gf_msg_debug (this->name, 0, "returning %d", ret);
- break;
+ gf_msg_debug(this->name, GF_LOG_INFO,
+ "BitRot scrub status "
+ "called");
+ va_start(ap, data);
+ output = va_arg(ap, dict_t *);
+ va_end(ap);
+
+ ret = br_scrubber_status_get(this, &output);
+ gf_msg_debug(this->name, 0, "returning %d", ret);
+ break;
case GF_EVENT_SCRUB_ONDEMAND:
- gf_log (this->name, GF_LOG_INFO, "BitRot scrub ondemand "
- "called");
-
- if (scrub_monitor->state != BR_SCRUB_STATE_PENDING) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRB_MSG_RESCHEDULE_SCRUBBER_FAILED,
- "on demand scrub schedule failed. Scrubber is "
- "not in pending state. Current state is %d",
- scrub_monitor->state);
- return -2;
- }
-
- /* Needs synchronization with reconfigure thread */
- pthread_mutex_lock (&priv->lock);
- {
- ret = br_scrub_state_machine (this, _gf_true);
- }
- pthread_mutex_unlock (&priv->lock);
-
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRB_MSG_RESCHEDULE_SCRUBBER_FAILED,
- "Could not schedule ondemand scrubbing. "
- "Scrubbing will continue according to "
- "old frequency.");
- }
- gf_msg_debug (this->name, 0, "returning %d", ret);
- break;
+ gf_log(this->name, GF_LOG_INFO,
+ "BitRot scrub ondemand "
+ "called");
+
+ if (scrub_monitor->state != BR_SCRUB_STATE_PENDING) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ BRB_MSG_RESCHEDULE_SCRUBBER_FAILED, "Current-state=%d",
+ scrub_monitor->state, NULL);
+ return -2;
+ }
+
+ /* Needs synchronization with reconfigure thread */
+ pthread_mutex_lock(&priv->lock);
+ {
+ ret = br_scrub_state_machine(this, _gf_true);
+ }
+ pthread_mutex_unlock(&priv->lock);
+
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ BRB_MSG_COULD_NOT_SCHEDULE_SCRUB, NULL);
+ }
+ gf_msg_debug(this->name, 0, "returning %d", ret);
+ break;
default:
- default_notify (this, event, data);
- }
+ default_notify(this, event, data);
+ }
- out:
- return 0;
+out:
+ return 0;
}
-/**
- * Initialize signer specific structures, spawn worker threads.
- */
-
static void
-br_fini_signer (xlator_t *this, br_private_t *priv)
+br_fini_signer(xlator_t *this, br_private_t *priv)
{
- int i = 0;
+ int i = 0;
- for (; i < BR_WORKERS; i++) {
- (void) gf_thread_cleanup_xint (priv->obj_queue->workers[i]);
- }
+ if (priv == NULL)
+ return;
+
+ for (; i < priv->signer_th_count; i++) {
+ (void)gf_thread_cleanup_xint(priv->obj_queue->workers[i]);
+ }
+ GF_FREE(priv->obj_queue->workers);
- pthread_cond_destroy (&priv->object_cond);
+ pthread_cond_destroy(&priv->object_cond);
}
+/**
+ * Initialize signer specific structures, spawn worker threads.
+ */
+
static int32_t
-br_init_signer (xlator_t *this, br_private_t *priv)
+br_init_signer(xlator_t *this, br_private_t *priv)
{
- int i = 0;
- int32_t ret = -1;
-
- /* initialize gfchangelog xlator context */
- ret = gf_changelog_init (this);
- if (ret)
- goto out;
-
- pthread_cond_init (&priv->object_cond, NULL);
-
- priv->obj_queue = GF_CALLOC (1, sizeof (*priv->obj_queue),
- gf_br_mt_br_ob_n_wk_t);
- if (!priv->obj_queue)
- goto cleanup_cond;
- INIT_LIST_HEAD (&priv->obj_queue->objects);
-
- for (i = 0; i < BR_WORKERS; i++) {
- ret = gf_thread_create (&priv->obj_queue->workers[i], NULL,
- br_process_object, this, "brpobj");
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- BRB_MSG_SPAWN_FAILED, "thread creation"
- " failed");
- ret = -1;
- goto cleanup_threads;
- }
+ int i = 0;
+ int32_t ret = -1;
+
+ /* initialize gfchangelog xlator context */
+ ret = gf_changelog_init(this);
+ if (ret)
+ goto out;
+
+ pthread_cond_init(&priv->object_cond, NULL);
+
+ priv->obj_queue = GF_CALLOC(1, sizeof(*priv->obj_queue),
+ gf_br_mt_br_ob_n_wk_t);
+ if (!priv->obj_queue)
+ goto cleanup_cond;
+ INIT_LIST_HEAD(&priv->obj_queue->objects);
+
+ priv->obj_queue->workers = GF_CALLOC(
+ priv->signer_th_count, sizeof(pthread_t), gf_br_mt_br_worker_t);
+ if (!priv->obj_queue->workers)
+ goto cleanup_obj_queue;
+
+ for (i = 0; i < priv->signer_th_count; i++) {
+ ret = gf_thread_create(&priv->obj_queue->workers[i], NULL,
+ br_process_object, this, "brpobj");
+ if (ret != 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret,
+ BRB_MSG_THREAD_CREATION_FAILED, NULL);
+ ret = -1;
+ goto cleanup_threads;
}
+ }
- return 0;
+ return 0;
- cleanup_threads:
- for (i--; i >= 0; i--) {
- (void) gf_thread_cleanup_xint (priv->obj_queue->workers[i]);
- }
+cleanup_threads:
+ for (i--; i >= 0; i--) {
+ (void)gf_thread_cleanup_xint(priv->obj_queue->workers[i]);
+ }
+ GF_FREE(priv->obj_queue->workers);
- GF_FREE (priv->obj_queue);
+cleanup_obj_queue:
+ GF_FREE(priv->obj_queue);
- cleanup_cond:
- /* that's explicit */
- pthread_cond_destroy (&priv->object_cond);
- out:
- return -1;
+cleanup_cond:
+ /* that's explicit */
+ pthread_cond_destroy(&priv->object_cond);
+out:
+ return -1;
}
/**
@@ -1784,360 +1796,358 @@ br_init_signer (xlator_t *this, br_private_t *priv)
* throttle.
*/
static int32_t
-br_rate_limit_signer (xlator_t *this, int child_count, int numbricks)
+br_rate_limit_signer(xlator_t *this, int child_count, int numbricks)
{
- br_private_t *priv = NULL;
- tbf_opspec_t spec = {0,};
-
- priv = this->private;
-
- spec.op = TBF_OP_HASH;
- spec.rate = 0;
- spec.maxlimit = 0;
-
-/**
- * OK. Most implementations of TBF I've come across generate tokens
- * every second (UML, etc..) and some chose sub-second granularity
- * (blk-iothrottle cgroups). TBF algorithm itself does not enforce
- * any logic for choosing generation interval and it seems pretty
- * logical as one could jack up token count per interval w.r.t.
- * generation rate.
- *
- * Value used here is chosen based on a series of test(s) performed
- * to balance object signing time and not maxing out on all available
- * CPU cores. It's obvious to have seconds granularity and jack up
- * token count per interval, thereby achieving close to similar
- * results. Let's stick to this as it seems to be working fine for
- * the set of ops that are throttled.
- **/
- spec.token_gen_interval = 600000; /* In usec */
-
+ br_private_t *priv = NULL;
+ tbf_opspec_t spec = {
+ 0,
+ };
+
+ priv = this->private;
+
+ spec.op = TBF_OP_HASH;
+ spec.rate = 0;
+ spec.maxlimit = 0;
+
+ /**
+ * OK. Most implementations of TBF I've come across generate tokens
+ * every second (UML, etc..) and some chose sub-second granularity
+ * (blk-iothrottle cgroups). TBF algorithm itself does not enforce
+ * any logic for choosing generation interval and it seems pretty
+ * logical as one could jack up token count per interval w.r.t.
+ * generation rate.
+ *
+ * Value used here is chosen based on a series of test(s) performed
+ * to balance object signing time and not maxing out on all available
+ * CPU cores. It's obvious to have seconds granularity and jack up
+ * token count per interval, thereby achieving close to similar
+ * results. Let's stick to this as it seems to be working fine for
+ * the set of ops that are throttled.
+ **/
+ spec.token_gen_interval = 600000; /* In usec */
#ifdef BR_RATE_LIMIT_SIGNER
- double contribution = 0;
- contribution = ((double)1 - ((double)child_count / (double)numbricks));
- if (contribution == 0)
- contribution = 1;
- spec.rate = BR_HASH_CALC_READ_SIZE * contribution;
- spec.maxlimit = BR_WORKERS * BR_HASH_CALC_READ_SIZE;
+ double contribution = 0;
+ contribution = ((double)1 - ((double)child_count / (double)numbricks));
+ if (contribution == 0)
+ contribution = 1;
+ spec.rate = BR_HASH_CALC_READ_SIZE * contribution;
+ spec.maxlimit = priv->signer_th_count * BR_HASH_CALC_READ_SIZE;
#endif
- if (!spec.rate)
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_RATE_LIMIT_INFO,
- "[Rate Limit Info] \"FULL THROTTLE\"");
- else
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_RATE_LIMIT_INFO,
- "[Rate Limit Info] \"tokens/sec (rate): %lu, "
- "maxlimit: %lu\"", spec.rate, spec.maxlimit);
+ if (!spec.rate)
+ gf_smsg(this->name, GF_LOG_INFO, 0, BRB_MSG_RATE_LIMIT_INFO,
+ "FULL THROTTLE", NULL);
+ else
+ gf_smsg(this->name, GF_LOG_INFO, 0, BRB_MSG_RATE_LIMIT_INFO,
+ "tokens/sec-rate=%lu", spec.rate, "maxlimit=%lu", spec.maxlimit,
+ NULL);
- priv->tbf = tbf_init (&spec, 1);
- return priv->tbf ? 0 : -1;
+ priv->tbf = tbf_init(&spec, 1);
+ return priv->tbf ? 0 : -1;
}
static int32_t
-br_signer_handle_options (xlator_t *this, br_private_t *priv, dict_t *options)
+br_signer_handle_options(xlator_t *this, br_private_t *priv, dict_t *options)
{
- if (options)
- GF_OPTION_RECONF ("expiry-time", priv->expiry_time,
- options, uint32, error_return);
- else
- GF_OPTION_INIT ("expiry-time", priv->expiry_time,
- uint32, error_return);
-
- return 0;
+ if (options) {
+ GF_OPTION_RECONF("expiry-time", priv->expiry_time, options, uint32,
+ error_return);
+ GF_OPTION_RECONF("signer-threads", priv->signer_th_count, options,
+ uint32, error_return);
+ } else {
+ GF_OPTION_INIT("expiry-time", priv->expiry_time, uint32, error_return);
+ GF_OPTION_INIT("signer-threads", priv->signer_th_count, uint32,
+ error_return);
+ }
+
+ return 0;
error_return:
- return -1;
+ return -1;
}
static int32_t
-br_signer_init (xlator_t *this, br_private_t *priv)
+br_signer_init(xlator_t *this, br_private_t *priv)
{
- int32_t ret = 0;
- int numbricks = 0;
+ int32_t ret = 0;
+ int numbricks = 0;
- GF_OPTION_INIT ("expiry-time", priv->expiry_time, uint32, error_return);
- GF_OPTION_INIT ("brick-count", numbricks, int32, error_return);
+ GF_OPTION_INIT("expiry-time", priv->expiry_time, uint32, error_return);
+ GF_OPTION_INIT("brick-count", numbricks, int32, error_return);
+ GF_OPTION_INIT("signer-threads", priv->signer_th_count, uint32,
+ error_return);
- ret = br_rate_limit_signer (this, priv->child_count, numbricks);
- if (ret)
- goto error_return;
-
- ret = br_init_signer (this, priv);
- if (ret)
- goto cleanup_tbf;
+ ret = br_rate_limit_signer(this, priv->child_count, numbricks);
+ if (ret)
+ goto error_return;
- return 0;
+ ret = br_init_signer(this, priv);
+ if (ret)
+ goto cleanup_tbf;
- cleanup_tbf:
- /* cleanup TBF */
- error_return:
- return -1;
+ return 0;
+cleanup_tbf:
+ /* cleanup TBF */
+error_return:
+ return -1;
}
static void
-br_free_scrubber_monitor (xlator_t *this, br_private_t *priv)
+br_free_scrubber_monitor(xlator_t *this, br_private_t *priv)
{
- struct br_monitor *scrub_monitor = &priv->scrub_monitor;
+ struct br_monitor *scrub_monitor = &priv->scrub_monitor;
- if (scrub_monitor->timer) {
- (void) gf_tw_del_timer (priv->timer_wheel, scrub_monitor->timer);
+ if (scrub_monitor->timer) {
+ (void)gf_tw_del_timer(priv->timer_wheel, scrub_monitor->timer);
- GF_FREE (scrub_monitor->timer);
- scrub_monitor->timer = NULL;
- }
+ GF_FREE(scrub_monitor->timer);
+ scrub_monitor->timer = NULL;
+ }
- (void) gf_thread_cleanup_xint (scrub_monitor->thread);
+ (void)gf_thread_cleanup_xint(scrub_monitor->thread);
- /* Clean up cond and mutex variables */
- pthread_mutex_destroy (&scrub_monitor->mutex);
- pthread_cond_destroy (&scrub_monitor->cond);
+ /* Clean up cond and mutex variables */
+ pthread_mutex_destroy(&scrub_monitor->mutex);
+ pthread_cond_destroy(&scrub_monitor->cond);
- pthread_mutex_destroy (&scrub_monitor->wakelock);
- pthread_cond_destroy (&scrub_monitor->wakecond);
+ pthread_mutex_destroy(&scrub_monitor->wakelock);
+ pthread_cond_destroy(&scrub_monitor->wakecond);
- pthread_mutex_destroy (&scrub_monitor->donelock);
- pthread_cond_destroy (&scrub_monitor->donecond);
+ pthread_mutex_destroy(&scrub_monitor->donelock);
+ pthread_cond_destroy(&scrub_monitor->donecond);
- LOCK_DESTROY (&scrub_monitor->lock);
+ LOCK_DESTROY(&scrub_monitor->lock);
}
static void
-br_free_children (xlator_t *this, br_private_t *priv, int count)
+br_free_children(xlator_t *this, br_private_t *priv, int count)
{
- br_child_t *child = NULL;
+ br_child_t *child = NULL;
- for (--count; count >= 0; count--) {
- child = &priv->children[count];
- mem_pool_destroy (child->timer_pool);
- pthread_mutex_destroy (&child->lock);
- }
+ for (--count; count >= 0; count--) {
+ child = &priv->children[count];
+ mem_pool_destroy(child->timer_pool);
+ pthread_mutex_destroy(&child->lock);
+ }
- GF_FREE (priv->children);
- priv->children = NULL;
+ GF_FREE(priv->children);
+ priv->children = NULL;
}
static int
-br_init_children (xlator_t *this, br_private_t *priv)
+br_init_children(xlator_t *this, br_private_t *priv)
{
- int i = 0;
- br_child_t *child = NULL;
- xlator_list_t *trav = NULL;
-
- priv->child_count = xlator_subvolume_count (this);
- priv->children = GF_CALLOC (priv->child_count, sizeof (*priv->children),
- gf_br_mt_br_child_t);
- if (!priv->children)
- goto err;
-
- trav = this->children;
- while (trav) {
- child = &priv->children[i];
-
- pthread_mutex_init (&child->lock, NULL);
- child->witnessed = 0;
-
- br_set_child_state (child, BR_CHILD_STATE_DISCONNECTED);
-
- child->this = this;
- child->xl = trav->xlator;
-
- child->timer_pool = mem_pool_new
- (struct gf_tw_timer_list, 4096);
- if (!child->timer_pool) {
- gf_msg (this->name, GF_LOG_ERROR,
- ENOMEM, BRB_MSG_NO_MEMORY,
- "failed to allocate mem-pool for timer");
- errno = ENOMEM;
- goto freechild;
- }
-
- INIT_LIST_HEAD (&child->list);
-
- i++;
- trav = trav->next;
+ int i = 0;
+ br_child_t *child = NULL;
+ xlator_list_t *trav = NULL;
+
+ priv->child_count = xlator_subvolume_count(this);
+ priv->children = GF_CALLOC(priv->child_count, sizeof(*priv->children),
+ gf_br_mt_br_child_t);
+ if (!priv->children)
+ goto err;
+
+ trav = this->children;
+ while (trav) {
+ child = &priv->children[i];
+
+ pthread_mutex_init(&child->lock, NULL);
+ child->witnessed = 0;
+
+ br_set_child_state(child, BR_CHILD_STATE_DISCONNECTED);
+
+ child->this = this;
+ child->xl = trav->xlator;
+
+ child->timer_pool = mem_pool_new(struct gf_tw_timer_list, 4096);
+ if (!child->timer_pool) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_MEM_POOL_ALLOC,
+ NULL);
+ errno = ENOMEM;
+ goto freechild;
}
- return 0;
+ INIT_LIST_HEAD(&child->list);
+
+ i++;
+ trav = trav->next;
+ }
- freechild:
- br_free_children (this, priv, i);
- err:
- return -1;
+ return 0;
+
+freechild:
+ br_free_children(this, priv, i);
+err:
+ return -1;
}
int32_t
-init (xlator_t *this)
+init(xlator_t *this)
{
- int32_t ret = -1;
- br_private_t *priv = NULL;
-
- if (!this->children) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRB_MSG_NO_CHILD,
- "FATAL: no children");
- goto out;
- }
-
- priv = GF_CALLOC (1, sizeof (*priv), gf_br_mt_br_private_t);
- if (!priv) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY,
- "failed to allocate memory (->priv)");
- goto out;
- }
+ int32_t ret = -1;
+ br_private_t *priv = NULL;
- GF_OPTION_INIT ("scrubber", priv->iamscrubber, bool, free_priv);
+ if (!this->children) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_NO_CHILD, NULL);
+ goto out;
+ }
- ret = br_init_children (this, priv);
- if (ret)
- goto free_priv;
+ priv = GF_CALLOC(1, sizeof(*priv), gf_br_mt_br_private_t);
+ if (!priv) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, BRB_MSG_NO_MEMORY, NULL);
+ goto out;
+ }
- pthread_mutex_init (&priv->lock, NULL);
- pthread_cond_init (&priv->cond, NULL);
+ GF_OPTION_INIT("scrubber", priv->iamscrubber, bool, free_priv);
- INIT_LIST_HEAD (&priv->bricks);
- INIT_LIST_HEAD (&priv->signing);
+ ret = br_init_children(this, priv);
+ if (ret)
+ goto free_priv;
- priv->timer_wheel = glusterfs_ctx_tw_get (this->ctx);
- if (!priv->timer_wheel) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRB_MSG_TIMER_WHEEL_UNAVAILABLE,
- "global timer wheel unavailable");
- goto cleanup;
- }
+ pthread_mutex_init(&priv->lock, NULL);
+ pthread_cond_init(&priv->cond, NULL);
- this->private = priv;
+ INIT_LIST_HEAD(&priv->bricks);
+ INIT_LIST_HEAD(&priv->signing);
- if (!priv->iamscrubber) {
- ret = br_signer_init (this, priv);
- if (!ret)
- ret = br_signer_handle_options (this, priv, NULL);
- } else {
- ret = br_scrubber_init (this, priv);
- if (!ret)
- ret = br_scrubber_handle_options (this, priv, NULL);
- }
+ priv->timer_wheel = glusterfs_ctx_tw_get(this->ctx);
+ if (!priv->timer_wheel) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_TIMER_WHEEL_UNAVAILABLE,
+ NULL);
+ goto cleanup;
+ }
- if (ret)
- goto cleanup;
+ this->private = priv;
- ret = gf_thread_create (&priv->thread, NULL, br_handle_events, this,
- "brhevent");
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- BRB_MSG_SPAWN_FAILED, "thread creation failed");
- ret = -1;
- }
+ if (!priv->iamscrubber) {
+ ret = br_signer_init(this, priv);
+ if (!ret)
+ ret = br_signer_handle_options(this, priv, NULL);
+ } else {
+ ret = br_scrubber_init(this, priv);
+ if (!ret)
+ ret = br_scrubber_handle_options(this, priv, NULL);
+ }
- if (!ret) {
- gf_msg (this->name, GF_LOG_INFO, 0, BRB_MSG_BITROT_LOADED,
- "bit-rot xlator loaded in \"%s\" mode",
- (priv->iamscrubber) ? "SCRUBBER" : "SIGNER");
- return 0;
- }
+ if (ret)
+ goto cleanup;
+
+ ret = gf_thread_create(&priv->thread, NULL, br_handle_events, this,
+ "brhevent");
+ if (ret != 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, -ret, BRB_MSG_THREAD_CREATION_FAILED,
+ NULL);
+ ret = -1;
+ }
+
+ if (!ret) {
+ gf_smsg(this->name, GF_LOG_INFO, 0, BRB_MSG_BITROT_LOADED, "mode=%s",
+ (priv->iamscrubber) ? "SCRUBBER" : "SIGNER", NULL);
+ return 0;
+ }
- cleanup:
- (void) pthread_cond_destroy (&priv->cond);
- (void) pthread_mutex_destroy (&priv->lock);
+cleanup:
+ (void)pthread_cond_destroy(&priv->cond);
+ (void)pthread_mutex_destroy(&priv->lock);
- br_free_children (this, priv, priv->child_count);
+ br_free_children(this, priv, priv->child_count);
- free_priv:
- GF_FREE (priv);
- out:
- this->private = NULL;
- return -1;
+free_priv:
+ GF_FREE(priv);
+out:
+ this->private = NULL;
+ return -1;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- br_private_t *priv = this->private;
+ br_private_t *priv = this->private;
- if (!priv)
- return;
+ if (!priv)
+ return;
- if (!priv->iamscrubber)
- br_fini_signer (this, priv);
- else
- (void) br_free_scrubber_monitor (this, priv);
+ if (!priv->iamscrubber)
+ br_fini_signer(this, priv);
+ else
+ (void)br_free_scrubber_monitor(this, priv);
- br_free_children (this, priv, priv->child_count);
+ br_free_children(this, priv, priv->child_count);
- this->private = NULL;
- GF_FREE (priv);
+ this->private = NULL;
+ GF_FREE(priv);
- glusterfs_ctx_tw_put (this->ctx);
+ glusterfs_ctx_tw_put(this->ctx);
- return;
+ return;
}
static void
-br_reconfigure_monitor (xlator_t *this)
+br_reconfigure_monitor(xlator_t *this)
{
- int32_t ret = 0;
-
- ret = br_scrub_state_machine (this, _gf_false);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRB_MSG_RESCHEDULE_SCRUBBER_FAILED,
- "Could not reschedule scrubber for the volume. Scrubbing "
- "will continue according to old frequency.");
- }
+ int32_t ret = 0;
+
+ ret = br_scrub_state_machine(this, _gf_false);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRB_MSG_COULD_NOT_SCHEDULE_SCRUB,
+ NULL);
+ }
}
static int
-br_reconfigure_scrubber (xlator_t *this, dict_t *options)
+br_reconfigure_scrubber(xlator_t *this, dict_t *options)
{
- int32_t ret = -1;
- br_private_t *priv = NULL;
+ int32_t ret = -1;
+ br_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- pthread_mutex_lock (&priv->lock);
- {
- ret = br_scrubber_handle_options (this, priv, options);
- }
- pthread_mutex_unlock (&priv->lock);
+ pthread_mutex_lock(&priv->lock);
+ {
+ ret = br_scrubber_handle_options(this, priv, options);
+ }
+ pthread_mutex_unlock(&priv->lock);
- if (ret)
- goto err;
+ if (ret)
+ goto err;
- /* change state for all _up_ subvolume(s) */
- pthread_mutex_lock (&priv->lock);
- {
- br_reconfigure_monitor (this);
- }
- pthread_mutex_unlock (&priv->lock);
+ /* change state for all _up_ subvolume(s) */
+ pthread_mutex_lock(&priv->lock);
+ {
+ br_reconfigure_monitor(this);
+ }
+ pthread_mutex_unlock(&priv->lock);
- err:
- return ret;
+err:
+ return ret;
}
static int
-br_reconfigure_signer (xlator_t *this, dict_t *options)
+br_reconfigure_signer(xlator_t *this, dict_t *options)
{
- br_private_t *priv = this->private;
+ br_private_t *priv = this->private;
- return br_signer_handle_options (this, priv, options);
+ return br_signer_handle_options(this, priv, options);
}
int
-reconfigure (xlator_t *this, dict_t *options)
+reconfigure(xlator_t *this, dict_t *options)
{
- int ret = 0;
- br_private_t *priv = NULL;
+ int ret = 0;
+ br_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->iamscrubber)
- ret = br_reconfigure_scrubber (this, options);
- else
- ret = br_reconfigure_signer (this, options);
+ if (priv->iamscrubber)
+ ret = br_reconfigure_scrubber(this, options);
+ else
+ ret = br_reconfigure_signer(this, options);
- return ret;
+ return ret;
}
struct xlator_fops fops;
@@ -2145,49 +2155,78 @@ struct xlator_fops fops;
struct xlator_cbks cbks;
struct volume_options options[] = {
- { .key = {"expiry-time"},
- .type = GF_OPTION_TYPE_INT,
- .default_value = SIGNING_TIMEOUT,
- .op_version = {GD_OP_VERSION_3_7_0},
- .flags = OPT_FLAG_SETTABLE,
- .description = "Waiting time for an object on which it waits "
- "before it is signed",
- },
- { .key = {"brick-count"},
- .type = GF_OPTION_TYPE_STR,
- .description = "Total number of bricks for the current node for "
- "all volumes in the trusted storage pool.",
- },
- { .key = {"scrubber", "scrub"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "false",
- .op_version = {GD_OP_VERSION_3_7_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_FORCE,
- .description = "option to run as a scrubber",
- },
- { .key = {"scrub-throttle"},
- .type = GF_OPTION_TYPE_STR,
- .default_value = "lazy",
- .op_version = {GD_OP_VERSION_3_7_0},
- .flags = OPT_FLAG_SETTABLE,
- .description = "Scrub-throttle value is a measure of how fast "
- "or slow the scrubber scrubs the filesystem for "
- "volume <VOLNAME>",
- },
- { .key = {"scrub-freq"},
- .type = GF_OPTION_TYPE_STR,
- .default_value = "biweekly",
- .op_version = {GD_OP_VERSION_3_7_0},
- .flags = OPT_FLAG_SETTABLE,
- .description = "Scrub frequency for volume <VOLNAME>",
- },
- { .key = {"scrub-state"},
- .type = GF_OPTION_TYPE_STR,
- .default_value = "active",
- .op_version = {GD_OP_VERSION_4_0_0},
- .flags = OPT_FLAG_SETTABLE,
- .description = "Pause/Resume scrub. Upon resume, scrubber "
- "continues from where it left off.",
- },
- { .key = {NULL} },
+ {
+ .key = {"expiry-time"},
+ .type = GF_OPTION_TYPE_INT,
+ .default_value = SIGNING_TIMEOUT,
+ .op_version = {GD_OP_VERSION_3_7_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .description = "Waiting time for an object on which it waits "
+ "before it is signed",
+ },
+ {
+ .key = {"brick-count"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Total number of bricks for the current node for "
+ "all volumes in the trusted storage pool.",
+ },
+ {
+ .key = {"scrubber", "scrub"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "false",
+ .op_version = {GD_OP_VERSION_3_7_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_FORCE,
+ .description = "option to run as a scrubber",
+ },
+ {
+ .key = {"scrub-throttle"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = "lazy",
+ .op_version = {GD_OP_VERSION_3_7_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .description = "Scrub-throttle value is a measure of how fast "
+ "or slow the scrubber scrubs the filesystem for "
+ "volume <VOLNAME>",
+ },
+ {
+ .key = {"scrub-freq"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = "biweekly",
+ .op_version = {GD_OP_VERSION_3_7_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .description = "Scrub frequency for volume <VOLNAME>",
+ },
+ {
+ .key = {"scrub-state"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = "active",
+ .op_version = {GD_OP_VERSION_4_0_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .description = "Pause/Resume scrub. Upon resume, scrubber "
+ "continues from where it left off.",
+ },
+ {
+ .key = {"signer-threads"},
+ .type = GF_OPTION_TYPE_INT,
+ .default_value = BR_WORKERS,
+ .op_version = {GD_OP_VERSION_8_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .description = "Number of signing process threads. As a best "
+ "practice, set this to the number of processor cores",
+ },
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .notify = notify,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "bit-rot",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/features/bit-rot/src/bitd/bit-rot.h b/xlators/features/bit-rot/src/bitd/bit-rot.h
index c2c0c225792..8ac7dcdac3d 100644
--- a/xlators/features/bit-rot/src/bitd/bit-rot.h
+++ b/xlators/features/bit-rot/src/bitd/bit-rot.h
@@ -11,17 +11,17 @@
#ifndef __BIT_ROT_H__
#define __BIT_ROT_H__
-#include "glusterfs.h"
-#include "logging.h"
-#include "dict.h"
-#include "xlator.h"
-#include "defaults.h"
-#include "syncop.h"
-#include "syncop-utils.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/dict.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
+#include <glusterfs/syncop.h>
+#include <glusterfs/syncop-utils.h>
#include "changelog.h"
#include "timer-wheel.h"
-#include "throttle-tbf.h"
+#include <glusterfs/throttle-tbf.h>
#include "bit-rot-ssm.h"
#include "bit-rot-common.h"
@@ -30,278 +30,273 @@
#include <openssl/sha.h>
-/**
- * TODO: make this configurable. As a best practice, set this to the
- * number of processor cores.
- */
-#define BR_WORKERS 4
-
typedef enum scrub_throttle {
- BR_SCRUB_THROTTLE_VOID = -1,
- BR_SCRUB_THROTTLE_LAZY = 0,
- BR_SCRUB_THROTTLE_NORMAL = 1,
- BR_SCRUB_THROTTLE_AGGRESSIVE = 2,
- BR_SCRUB_THROTTLE_STALLED = 3,
+ BR_SCRUB_THROTTLE_VOID = -1,
+ BR_SCRUB_THROTTLE_LAZY = 0,
+ BR_SCRUB_THROTTLE_NORMAL = 1,
+ BR_SCRUB_THROTTLE_AGGRESSIVE = 2,
+ BR_SCRUB_THROTTLE_STALLED = 3,
} scrub_throttle_t;
typedef enum scrub_freq {
- BR_FSSCRUB_FREQ_HOURLY = 1,
- BR_FSSCRUB_FREQ_DAILY,
- BR_FSSCRUB_FREQ_WEEKLY,
- BR_FSSCRUB_FREQ_BIWEEKLY,
- BR_FSSCRUB_FREQ_MONTHLY,
- BR_FSSCRUB_FREQ_MINUTE,
- BR_FSSCRUB_FREQ_STALLED,
+ BR_FSSCRUB_FREQ_HOURLY = 1,
+ BR_FSSCRUB_FREQ_DAILY,
+ BR_FSSCRUB_FREQ_WEEKLY,
+ BR_FSSCRUB_FREQ_BIWEEKLY,
+ BR_FSSCRUB_FREQ_MONTHLY,
+ BR_FSSCRUB_FREQ_MINUTE,
+ BR_FSSCRUB_FREQ_STALLED,
} scrub_freq_t;
-#define signature_size(hl) (sizeof (br_isignature_t) + hl + 1)
+#define signature_size(hl) (sizeof(br_isignature_t) + hl + 1)
struct br_scanfs {
- gf_lock_t entrylock;
+ gf_lock_t entrylock;
- pthread_mutex_t waitlock;
- pthread_cond_t waitcond;
+ pthread_mutex_t waitlock;
+ pthread_cond_t waitcond;
- unsigned int entries;
- struct list_head queued;
- struct list_head ready;
+ unsigned int entries;
+ struct list_head queued;
+ struct list_head ready;
};
/* just need three states to track child status */
typedef enum br_child_state {
- BR_CHILD_STATE_CONNECTED = 1,
- BR_CHILD_STATE_INITIALIZING,
- BR_CHILD_STATE_CONNFAILED,
- BR_CHILD_STATE_DISCONNECTED,
+ BR_CHILD_STATE_CONNECTED = 1,
+ BR_CHILD_STATE_INITIALIZING,
+ BR_CHILD_STATE_CONNFAILED,
+ BR_CHILD_STATE_DISCONNECTED,
} br_child_state_t;
struct br_child {
- pthread_mutex_t lock; /* protects child state */
- char witnessed; /* witnessed at least one succesfull
- connection */
- br_child_state_t c_state; /* current state of this child */
+ pthread_mutex_t lock; /* protects child state */
+ char witnessed; /* witnessed at least one successful
+ connection */
+ br_child_state_t c_state; /* current state of this child */
- char child_up; /* Indicates whether this child is
- up or not */
- xlator_t *xl; /* client xlator corresponding to
- this child */
- inode_table_t *table; /* inode table for this child */
- char brick_path[PATH_MAX]; /* brick export directory of this
- child */
- struct list_head list; /* hook to attach to the list of
- UP children */
- xlator_t *this; /* Bit rot xlator */
+ char child_up; /* Indicates whether this child is
+ up or not */
+ xlator_t *xl; /* client xlator corresponding to
+ this child */
+ inode_table_t *table; /* inode table for this child */
+ char brick_path[PATH_MAX]; /* brick export directory of this
+ child */
+ struct list_head list; /* hook to attach to the list of
+ UP children */
+ xlator_t *this; /* Bit rot xlator */
- pthread_t thread; /* initial crawler for unsigned
- object(s) or scrub crawler */
- int threadrunning; /* active thread */
+ pthread_t thread; /* initial crawler for unsigned
+ object(s) or scrub crawler */
+ int threadrunning; /* active thread */
- struct mem_pool *timer_pool; /* timer-wheel's timer mem-pool */
+ struct mem_pool *timer_pool; /* timer-wheel's timer mem-pool */
- struct timeval tv;
+ struct timeval tv;
- struct br_scanfs fsscan; /* per subvolume FS scanner */
+ struct br_scanfs fsscan; /* per subvolume FS scanner */
- gf_boolean_t active_scrubbing; /* Actively scrubbing or not */
+ gf_boolean_t active_scrubbing; /* Actively scrubbing or not */
};
typedef struct br_child br_child_t;
struct br_obj_n_workers {
- struct list_head objects; /* queue of objects expired from the
- timer wheel and ready to be picked
- up for signing */
- pthread_t workers[BR_WORKERS]; /* Threads which pick up the objects
- from the above queue and start
- signing each object */
+ struct list_head objects; /* queue of objects expired from the
+ timer wheel and ready to be picked
+ up for signing */
+ pthread_t *workers; /* Threads which pick up the objects
+ from the above queue and start
+ signing each object */
};
struct br_scrubber {
- xlator_t *this;
+ xlator_t *this;
- scrub_throttle_t throttle;
+ scrub_throttle_t throttle;
- /**
- * frequency of scanning for this subvolume. this should
- * normally be per-child, but since all childs follow the
- * same frequency for a volume, this option ends up here
- * instead of br_child_t.
- */
- scrub_freq_t frequency;
+ /**
+ * frequency of scanning for this subvolume. this should
+ * normally be per-child, but since all children follow the
+ * same frequency for a volume, this option ends up here
+ * instead of br_child_t.
+ */
+ scrub_freq_t frequency;
- gf_boolean_t frequency_reconf;
- gf_boolean_t throttle_reconf;
+ gf_boolean_t frequency_reconf;
+ gf_boolean_t throttle_reconf;
- pthread_mutex_t mutex;
- pthread_cond_t cond;
+ pthread_mutex_t mutex;
+ pthread_cond_t cond;
- unsigned int nr_scrubbers;
- struct list_head scrubbers;
+ unsigned int nr_scrubbers;
+ struct list_head scrubbers;
- /**
- * list of "rotatable" subvolume(s) undergoing scrubbing
- */
- struct list_head scrublist;
+ /**
+ * list of "rotatable" subvolume(s) undergoing scrubbing
+ */
+ struct list_head scrublist;
};
struct br_monitor {
- gf_lock_t lock;
- pthread_t thread; /* Monitor thread */
-
- gf_boolean_t inited;
- pthread_mutex_t mutex;
- pthread_cond_t cond; /* Thread starts and will be waiting on cond.
- First child which is up wakes this up */
-
- xlator_t *this;
- /* scheduler */
- uint32_t boot;
-
- int32_t active_child_count; /* Number of children currently scrubbing */
- gf_boolean_t kick; /* This variable tracks the scrubber is
- * kicked or not. Both 'kick' and
- * 'active_child_count' uses the same pair
- * of mutex-cond variable, i.e, wakelock and
- * wakecond. */
-
- pthread_mutex_t wakelock;
- pthread_cond_t wakecond;
-
- gf_boolean_t done;
- pthread_mutex_t donelock;
- pthread_cond_t donecond;
-
- struct gf_tw_timer_list *timer;
- br_scrub_state_t state; /* current scrub state */
+ gf_lock_t lock;
+ pthread_t thread; /* Monitor thread */
+
+ gf_boolean_t inited;
+ pthread_mutex_t mutex;
+ pthread_cond_t cond; /* Thread starts and will be waiting on cond.
+ First child which is up wakes this up */
+
+ xlator_t *this;
+ /* scheduler */
+ uint32_t boot;
+
+ int32_t active_child_count; /* Number of children currently scrubbing */
+ gf_boolean_t kick; /* This variable tracks the scrubber is
+ * kicked or not. Both 'kick' and
+ * 'active_child_count' uses the same pair
+ * of mutex-cond variable, i.e, wakelock and
+ * wakecond. */
+
+ pthread_mutex_t wakelock;
+ pthread_cond_t wakecond;
+
+ gf_boolean_t done;
+ pthread_mutex_t donelock;
+ pthread_cond_t donecond;
+
+ struct gf_tw_timer_list *timer;
+ br_scrub_state_t state; /* current scrub state */
};
typedef struct br_obj_n_workers br_obj_n_workers_t;
typedef struct br_private br_private_t;
-typedef void (*br_scrubbed_file_update) (br_private_t *priv);
+typedef void (*br_scrubbed_file_update)(br_private_t *priv);
struct br_private {
- pthread_mutex_t lock;
+ pthread_mutex_t lock;
+
+ struct list_head bricks; /* list of bricks from which enents
+ have been received */
- struct list_head bricks; /* list of bricks from which enents
- have been received */
+ struct list_head signing;
- struct list_head signing;
+ pthread_cond_t object_cond; /* handling signing of objects */
+ int child_count;
+ br_child_t *children; /* list of subvolumes */
+ int up_children;
- pthread_cond_t object_cond; /* handling signing of objects */
- int child_count;
- br_child_t *children; /* list of subvolumes */
- int up_children;
+ pthread_cond_t cond; /* handling CHILD_UP notifications */
+ pthread_t thread; /* thread for connecting each UP
+ child with changelog */
- pthread_cond_t cond; /* handling CHILD_UP notifications */
- pthread_t thread; /* thread for connecting each UP
- child with changelog */
+ struct tvec_base *timer_wheel; /* timer wheel where the objects which
+ changelog has sent sits and waits
+ for expiry */
+ br_obj_n_workers_t *obj_queue; /* place holder for all the objects
+ that are expired from timer wheel
+ and ready to be picked up for
+ signing and the workers which sign
+ the objects */
- struct tvec_base *timer_wheel; /* timer wheel where the objects which
- changelog has sent sits and waits
- for expiry */
- br_obj_n_workers_t *obj_queue; /* place holder for all the objects
- that are expired from timer wheel
- and ready to be picked up for
- signing and the workers which sign
- the objects */
+ uint32_t expiry_time; /* objects "wait" time */
- uint32_t expiry_time; /* objects "wait" time */
+ uint32_t signer_th_count; /* Number of signing process threads */
- tbf_t *tbf; /* token bucket filter */
+ tbf_t *tbf; /* token bucket filter */
- gf_boolean_t iamscrubber; /* function as a fs scrubber */
+ gf_boolean_t iamscrubber; /* function as a fs scrubber */
- struct br_scrub_stats scrub_stat; /* statistics of scrub*/
+ struct br_scrub_stats scrub_stat; /* statistics of scrub*/
- struct br_scrubber fsscrub; /* scrubbers for this subvolume */
+ struct br_scrubber fsscrub; /* scrubbers for this subvolume */
- struct br_monitor scrub_monitor; /* scrubber monitor */
+ struct br_monitor scrub_monitor; /* scrubber monitor */
};
struct br_object {
- xlator_t *this;
+ xlator_t *this;
- uuid_t gfid;
+ uuid_t gfid;
- unsigned long signedversion; /* version aginst which this object will
- be signed */
- br_child_t *child; /* object's subvolume */
+ unsigned long signedversion; /* version against which this object will
+ be signed */
+ br_child_t *child; /* object's subvolume */
- int sign_info;
+ int sign_info;
- struct list_head list; /* hook to add to the queue once the
- object is expired from timer wheel */
- void *data;
+ struct list_head list; /* hook to add to the queue once the
+ object is expired from timer wheel */
+ void *data;
};
typedef struct br_object br_object_t;
-typedef int32_t (br_scrub_ssm_call) (xlator_t *);
+typedef int32_t(br_scrub_ssm_call)(xlator_t *);
void
-br_log_object (xlator_t *, char *, uuid_t, int32_t);
+br_log_object(xlator_t *, char *, uuid_t, int32_t);
void
-br_log_object_path (xlator_t *, char *, const char *, int32_t);
+br_log_object_path(xlator_t *, char *, const char *, int32_t);
int32_t
-br_calculate_obj_checksum (unsigned char *,
- br_child_t *, fd_t *, struct iatt *);
+br_calculate_obj_checksum(unsigned char *, br_child_t *, fd_t *, struct iatt *);
int32_t
-br_prepare_loc (xlator_t *, br_child_t *, loc_t *, gf_dirent_t *, loc_t *);
+br_prepare_loc(xlator_t *, br_child_t *, loc_t *, gf_dirent_t *, loc_t *);
gf_boolean_t
-bitd_is_bad_file (xlator_t *, br_child_t *, loc_t *, fd_t *);
+bitd_is_bad_file(xlator_t *, br_child_t *, loc_t *, fd_t *);
static inline void
-_br_set_child_state (br_child_t *child, br_child_state_t state)
+_br_set_child_state(br_child_t *child, br_child_state_t state)
{
- child->c_state = state;
+ child->c_state = state;
}
static inline int
-_br_is_child_connected (br_child_t *child)
+_br_is_child_connected(br_child_t *child)
{
- return (child->c_state == BR_CHILD_STATE_CONNECTED);
+ return (child->c_state == BR_CHILD_STATE_CONNECTED);
}
static inline int
-_br_is_child_scrub_active (br_child_t *child)
+_br_is_child_scrub_active(br_child_t *child)
{
- return child->active_scrubbing;
+ return child->active_scrubbing;
}
static inline int
-_br_child_failed_conn (br_child_t *child)
+_br_child_failed_conn(br_child_t *child)
{
- return (child->c_state == BR_CHILD_STATE_CONNFAILED);
+ return (child->c_state == BR_CHILD_STATE_CONNFAILED);
}
static inline int
-_br_child_witnessed_connection (br_child_t *child)
+_br_child_witnessed_connection(br_child_t *child)
{
- return (child->witnessed == 1);
+ return (child->witnessed == 1);
}
/* scrub state */
static inline void
-_br_monitor_set_scrub_state (struct br_monitor *scrub_monitor,
- br_scrub_state_t state)
+_br_monitor_set_scrub_state(struct br_monitor *scrub_monitor,
+ br_scrub_state_t state)
{
- scrub_monitor->state = state;
+ scrub_monitor->state = state;
}
static inline br_scrub_event_t
-_br_child_get_scrub_event (struct br_scrubber *fsscrub)
+_br_child_get_scrub_event(struct br_scrubber *fsscrub)
{
- return (fsscrub->frequency == BR_FSSCRUB_FREQ_STALLED)
- ? BR_SCRUB_EVENT_PAUSE : BR_SCRUB_EVENT_SCHEDULE;
+ return (fsscrub->frequency == BR_FSSCRUB_FREQ_STALLED)
+ ? BR_SCRUB_EVENT_PAUSE
+ : BR_SCRUB_EVENT_SCHEDULE;
}
int32_t
-br_get_bad_objects_list (xlator_t *this, dict_t **dict);
-
+br_get_bad_objects_list(xlator_t *this, dict_t **dict);
#endif /* __BIT_ROT_H__ */
diff --git a/xlators/features/bit-rot/src/stub/bit-rot-common.h b/xlators/features/bit-rot/src/stub/bit-rot-common.h
index 2afc9f47c29..20561aa7764 100644
--- a/xlators/features/bit-rot/src/stub/bit-rot-common.h
+++ b/xlators/features/bit-rot/src/stub/bit-rot-common.h
@@ -11,74 +11,74 @@
#ifndef __BIT_ROT_COMMON_H__
#define __BIT_ROT_COMMON_H__
-#include "glusterfs.h"
+#include <glusterfs/glusterfs.h>
#include "bit-rot-object-version.h"
-#define BR_VXATTR_VERSION (1 << 0)
+#define BR_VXATTR_VERSION (1 << 0)
#define BR_VXATTR_SIGNATURE (1 << 1)
#define BR_VXATTR_SIGN_MISSING (BR_VXATTR_SIGNATURE)
-#define BR_VXATTR_ALL_MISSING \
- (BR_VXATTR_VERSION | BR_VXATTR_SIGNATURE)
+#define BR_VXATTR_ALL_MISSING (BR_VXATTR_VERSION | BR_VXATTR_SIGNATURE)
-#define BR_BAD_OBJ_CONTAINER (uuid_t){0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 8}
+#define BR_BAD_OBJ_CONTAINER \
+ (uuid_t) { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 8 }
typedef enum br_vxattr_state {
- BR_VXATTR_STATUS_FULL = 0,
- BR_VXATTR_STATUS_MISSING = 1,
- BR_VXATTR_STATUS_UNSIGNED = 2,
- BR_VXATTR_STATUS_INVALID = 3,
+ BR_VXATTR_STATUS_FULL = 0,
+ BR_VXATTR_STATUS_MISSING = 1,
+ BR_VXATTR_STATUS_UNSIGNED = 2,
+ BR_VXATTR_STATUS_INVALID = 3,
} br_vxattr_status_t;
typedef enum br_sign_state {
- BR_SIGN_INVALID = -1,
- BR_SIGN_NORMAL = 0,
- BR_SIGN_REOPEN_WAIT = 1,
- BR_SIGN_QUICK = 2,
+ BR_SIGN_INVALID = -1,
+ BR_SIGN_NORMAL = 0,
+ BR_SIGN_REOPEN_WAIT = 1,
+ BR_SIGN_QUICK = 2,
} br_sign_state_t;
static inline br_vxattr_status_t
-br_version_xattr_state (dict_t *xattr, br_version_t **obuf,
- br_signature_t **sbuf, gf_boolean_t *objbad)
+br_version_xattr_state(dict_t *xattr, br_version_t **obuf,
+ br_signature_t **sbuf, gf_boolean_t *objbad)
{
- int32_t ret = 0;
- int32_t vxattr = 0;
- br_vxattr_status_t status;
- void *data = NULL;
-
- /**
- * The key being present in the dict indicates the xattr was set on
- * disk. The presence of xattr itself as of now is suffecient to say
- * the the object is bad.
- */
- *objbad = _gf_false;
- ret = dict_get_bin (xattr, BITROT_OBJECT_BAD_KEY, (void **)&data);
- if (!ret)
- *objbad = _gf_true;
-
- ret = dict_get_bin (xattr, BITROT_CURRENT_VERSION_KEY, (void **)obuf);
- if (ret)
- vxattr |= BR_VXATTR_VERSION;
-
- ret = dict_get_bin (xattr, BITROT_SIGNING_VERSION_KEY, (void **)sbuf);
- if (ret)
- vxattr |= BR_VXATTR_SIGNATURE;
-
- switch (vxattr) {
+ int32_t ret = 0;
+ int32_t vxattr = 0;
+ br_vxattr_status_t status;
+ void *data = NULL;
+
+ /**
+ * The key being present in the dict indicates the xattr was set on
+ * disk. The presence of xattr itself as of now is suffecient to say
+ * the the object is bad.
+ */
+ *objbad = _gf_false;
+ ret = dict_get_bin(xattr, BITROT_OBJECT_BAD_KEY, (void **)&data);
+ if (!ret)
+ *objbad = _gf_true;
+
+ ret = dict_get_bin(xattr, BITROT_CURRENT_VERSION_KEY, (void **)obuf);
+ if (ret)
+ vxattr |= BR_VXATTR_VERSION;
+
+ ret = dict_get_bin(xattr, BITROT_SIGNING_VERSION_KEY, (void **)sbuf);
+ if (ret)
+ vxattr |= BR_VXATTR_SIGNATURE;
+
+ switch (vxattr) {
case 0:
- status = BR_VXATTR_STATUS_FULL;
- break;
+ status = BR_VXATTR_STATUS_FULL;
+ break;
case BR_VXATTR_SIGN_MISSING:
- status = BR_VXATTR_STATUS_UNSIGNED;
- break;
+ status = BR_VXATTR_STATUS_UNSIGNED;
+ break;
case BR_VXATTR_ALL_MISSING:
- status = BR_VXATTR_STATUS_MISSING;
- break;
+ status = BR_VXATTR_STATUS_MISSING;
+ break;
default:
- status = BR_VXATTR_STATUS_INVALID;
- }
+ status = BR_VXATTR_STATUS_INVALID;
+ }
- return status;
+ return status;
}
/**
@@ -86,13 +86,13 @@ br_version_xattr_state (dict_t *xattr, br_version_t **obuf,
* signing.
*/
typedef struct br_isignature_in {
- int8_t signaturetype; /* signature type */
+ int8_t signaturetype; /* signature type */
- unsigned long signedversion; /* version against which the
- object was signed */
+ unsigned long signedversion; /* version against which the
+ object was signed */
- size_t signaturelen; /* signature length */
- char signature[0]; /* object signature */
+ size_t signaturelen; /* signature length */
+ char signature[0]; /* object signature */
} br_isignature_t;
/**
@@ -100,80 +100,79 @@ typedef struct br_isignature_in {
* verification.
*/
typedef struct br_isignature_out {
- char stale; /* stale signature? */
+ char stale; /* stale signature? */
- unsigned long version; /* current signed version */
+ unsigned long version; /* current signed version */
- uint32_t time[2]; /* time when the object
- got dirtied */
+ uint32_t time[2]; /* time when the object
+ got dirtied */
- int8_t signaturetype; /* hash type */
- size_t signaturelen; /* signature length */
- char signature[0]; /* signature (hash) */
+ int8_t signaturetype; /* hash type */
+ size_t signaturelen; /* signature length */
+ char signature[0]; /* signature (hash) */
} br_isignature_out_t;
typedef struct br_stub_init {
- uint32_t timebuf[2];
- char export[PATH_MAX];
+ uint32_t timebuf[2];
+ char export[PATH_MAX];
} br_stub_init_t;
typedef enum {
- BR_SIGNATURE_TYPE_VOID = -1, /* object is not signed */
- BR_SIGNATURE_TYPE_ZERO = 0, /* min boundary */
- BR_SIGNATURE_TYPE_SHA256 = 1, /* signed with SHA256 */
- BR_SIGNATURE_TYPE_MAX = 2, /* max boundary */
+ BR_SIGNATURE_TYPE_VOID = -1, /* object is not signed */
+ BR_SIGNATURE_TYPE_ZERO = 0, /* min boundary */
+ BR_SIGNATURE_TYPE_SHA256 = 1, /* signed with SHA256 */
+ BR_SIGNATURE_TYPE_MAX = 2, /* max boundary */
} br_signature_type;
/* BitRot stub start time (virtual xattr) */
-#define GLUSTERFS_GET_BR_STUB_INIT_TIME "trusted.glusterfs.bit-rot.stub-init"
+#define GLUSTERFS_GET_BR_STUB_INIT_TIME "trusted.glusterfs.bit-rot.stub-init"
/* signing/reopen hint */
#define BR_OBJECT_RESIGN 0
-#define BR_OBJECT_REOPEN 1
-#define BR_REOPEN_SIGN_HINT_KEY "trusted.glusterfs.bit-rot.reopen-hint"
+#define BR_OBJECT_REOPEN 1
+#define BR_REOPEN_SIGN_HINT_KEY "trusted.glusterfs.bit-rot.reopen-hint"
static inline int
-br_is_signature_type_valid (int8_t signaturetype)
+br_is_signature_type_valid(int8_t signaturetype)
{
- return ((signaturetype > BR_SIGNATURE_TYPE_ZERO)
- && (signaturetype < BR_SIGNATURE_TYPE_MAX));
+ return ((signaturetype > BR_SIGNATURE_TYPE_ZERO) &&
+ (signaturetype < BR_SIGNATURE_TYPE_MAX));
}
static inline void
-br_set_default_ongoingversion (br_version_t *buf, uint32_t *tv)
+br_set_default_ongoingversion(br_version_t *buf, uint32_t *tv)
{
- buf->ongoingversion = BITROT_DEFAULT_CURRENT_VERSION;
- buf->timebuf[0] = tv[0];
- buf->timebuf[1] = tv[1];
+ buf->ongoingversion = BITROT_DEFAULT_CURRENT_VERSION;
+ buf->timebuf[0] = tv[0];
+ buf->timebuf[1] = tv[1];
}
static inline void
-br_set_default_signature (br_signature_t *buf, size_t *size)
+br_set_default_signature(br_signature_t *buf, size_t *size)
{
- buf->signaturetype = (int8_t) BR_SIGNATURE_TYPE_VOID;
- buf->signedversion = BITROT_DEFAULT_SIGNING_VERSION;
+ buf->signaturetype = (int8_t)BR_SIGNATURE_TYPE_VOID;
+ buf->signedversion = BITROT_DEFAULT_SIGNING_VERSION;
- *size = sizeof (br_signature_t); /* no signature */
+ *size = sizeof(br_signature_t); /* no signature */
}
static inline void
-br_set_ongoingversion (br_version_t *buf,
- unsigned long version, uint32_t *tv)
+br_set_ongoingversion(br_version_t *buf, unsigned long version, uint32_t *tv)
{
- buf->ongoingversion = version;
- buf->timebuf[0] = tv[0];
- buf->timebuf[1] = tv[1];
+ buf->ongoingversion = version;
+ buf->timebuf[0] = tv[0];
+ buf->timebuf[1] = tv[1];
}
static inline void
-br_set_signature (br_signature_t *buf,
- br_isignature_t *sign, size_t signaturelen, size_t *size)
+br_set_signature(br_signature_t *buf, br_isignature_t *sign,
+ size_t signaturelen, size_t *size)
{
- buf->signaturetype = sign->signaturetype;
- buf->signedversion = ntohl (sign->signedversion);
+ buf->signaturetype = sign->signaturetype;
+ buf->signedversion = ntohl(sign->signedversion);
- memcpy (buf->signature, sign->signature, signaturelen);
- *size = sizeof (br_signature_t) + signaturelen;
+ memcpy(buf->signature, sign->signature, signaturelen);
+ *size = sizeof(br_signature_t) + signaturelen;
}
#endif /* __BIT_ROT_COMMON_H__ */
diff --git a/xlators/features/bit-rot/src/stub/bit-rot-object-version.h b/xlators/features/bit-rot/src/stub/bit-rot-object-version.h
index 1f2497aebe9..7ae6a5200df 100644
--- a/xlators/features/bit-rot/src/stub/bit-rot-object-version.h
+++ b/xlators/features/bit-rot/src/stub/bit-rot-object-version.h
@@ -15,16 +15,16 @@
* on-disk formats for ongoing version and object signature.
*/
typedef struct br_version {
- unsigned long ongoingversion;
- uint32_t timebuf[2];
+ unsigned long ongoingversion;
+ uint32_t timebuf[2];
} br_version_t;
-typedef struct __attribute__ ((__packed__)) br_signature {
- int8_t signaturetype;
+typedef struct __attribute__((__packed__)) br_signature {
+ int8_t signaturetype;
- unsigned long signedversion;
+ unsigned long signedversion;
- char signature[0];
+ char signature[0];
} br_signature_t;
#endif
diff --git a/xlators/features/bit-rot/src/stub/bit-rot-stub-helpers.c b/xlators/features/bit-rot/src/stub/bit-rot-stub-helpers.c
index fff7a7eedf2..8ac13a09941 100644
--- a/xlators/features/bit-rot/src/stub/bit-rot-stub-helpers.c
+++ b/xlators/features/bit-rot/src/stub/bit-rot-stub-helpers.c
@@ -11,89 +11,88 @@
#include "bit-rot-stub.h"
br_stub_fd_t *
-br_stub_fd_new (void)
+br_stub_fd_new(void)
{
- br_stub_fd_t *br_stub_fd = NULL;
+ br_stub_fd_t *br_stub_fd = NULL;
- br_stub_fd = GF_CALLOC (1, sizeof (*br_stub_fd),
- gf_br_stub_mt_br_stub_fd_t);
+ br_stub_fd = GF_CALLOC(1, sizeof(*br_stub_fd), gf_br_stub_mt_br_stub_fd_t);
- return br_stub_fd;
+ return br_stub_fd;
}
int
-__br_stub_fd_ctx_set (xlator_t *this, fd_t *fd, br_stub_fd_t *br_stub_fd)
+__br_stub_fd_ctx_set(xlator_t *this, fd_t *fd, br_stub_fd_t *br_stub_fd)
{
- uint64_t value = 0;
- int ret = -1;
+ uint64_t value = 0;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO ("bit-rot-stub", this, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
- GF_VALIDATE_OR_GOTO (this->name, br_stub_fd, out);
+ GF_VALIDATE_OR_GOTO("bit-rot-stub", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
+ GF_VALIDATE_OR_GOTO(this->name, br_stub_fd, out);
- value = (uint64_t)(long) br_stub_fd;
+ value = (uint64_t)(long)br_stub_fd;
- ret = __fd_ctx_set (fd, this, value);
+ ret = __fd_ctx_set(fd, this, value);
out:
- return ret;
+ return ret;
}
br_stub_fd_t *
-__br_stub_fd_ctx_get (xlator_t *this, fd_t *fd)
+__br_stub_fd_ctx_get(xlator_t *this, fd_t *fd)
{
- br_stub_fd_t *br_stub_fd = NULL;
- uint64_t value = 0;
- int ret = -1;
+ br_stub_fd_t *br_stub_fd = NULL;
+ uint64_t value = 0;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO ("bit-rot-stub", this, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
+ GF_VALIDATE_OR_GOTO("bit-rot-stub", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
- ret = __fd_ctx_get (fd, this, &value);
- if (ret)
- return NULL;
+ ret = __fd_ctx_get(fd, this, &value);
+ if (ret)
+ return NULL;
- br_stub_fd = (br_stub_fd_t *) ((long) value);
+ br_stub_fd = (br_stub_fd_t *)((long)value);
out:
- return br_stub_fd;
+ return br_stub_fd;
}
br_stub_fd_t *
-br_stub_fd_ctx_get (xlator_t *this, fd_t *fd)
+br_stub_fd_ctx_get(xlator_t *this, fd_t *fd)
{
- br_stub_fd_t *br_stub_fd = NULL;
+ br_stub_fd_t *br_stub_fd = NULL;
- GF_VALIDATE_OR_GOTO ("bit-rot-stub", this, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
+ GF_VALIDATE_OR_GOTO("bit-rot-stub", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
- LOCK (&fd->lock);
- {
- br_stub_fd = __br_stub_fd_ctx_get (this, fd);
- }
- UNLOCK (&fd->lock);
+ LOCK(&fd->lock);
+ {
+ br_stub_fd = __br_stub_fd_ctx_get(this, fd);
+ }
+ UNLOCK(&fd->lock);
out:
- return br_stub_fd;
+ return br_stub_fd;
}
int32_t
-br_stub_fd_ctx_set (xlator_t *this, fd_t *fd, br_stub_fd_t *br_stub_fd)
+br_stub_fd_ctx_set(xlator_t *this, fd_t *fd, br_stub_fd_t *br_stub_fd)
{
- int32_t ret = -1;
+ int32_t ret = -1;
- GF_VALIDATE_OR_GOTO ("bit-rot-stub", this, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
- GF_VALIDATE_OR_GOTO (this->name, br_stub_fd, out);
+ GF_VALIDATE_OR_GOTO("bit-rot-stub", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
+ GF_VALIDATE_OR_GOTO(this->name, br_stub_fd, out);
- LOCK (&fd->lock);
- {
- ret = __br_stub_fd_ctx_set (this, fd, br_stub_fd);
- }
- UNLOCK (&fd->lock);
+ LOCK(&fd->lock);
+ {
+ ret = __br_stub_fd_ctx_set(this, fd, br_stub_fd);
+ }
+ UNLOCK(&fd->lock);
out:
- return ret;
+ return ret;
}
/**
@@ -101,114 +100,112 @@ out:
* @gfid: gfid of the bad object being added to the bad objects directory
*/
int
-br_stub_add (xlator_t *this, uuid_t gfid)
+br_stub_add(xlator_t *this, uuid_t gfid)
{
- char gfid_path[BR_PATH_MAX_PLUS] = {0};
- char bad_gfid_path[BR_PATH_MAX_PLUS] = {0};
- int ret = 0;
- br_stub_private_t *priv = NULL;
- struct stat st = {0};
-
- priv = this->private;
- GF_ASSERT_AND_GOTO_WITH_ERROR (this->name, !gf_uuid_is_null (gfid),
- out, errno, EINVAL);
+ char gfid_path[BR_PATH_MAX_PLUS] = {0};
+ char bad_gfid_path[BR_PATH_MAX_PLUS] = {0};
+ int ret = 0;
+ br_stub_private_t *priv = NULL;
+ struct stat st = {0};
+
+ priv = this->private;
+ GF_ASSERT_AND_GOTO_WITH_ERROR(this->name, !gf_uuid_is_null(gfid), out,
+ errno, EINVAL);
+
+ snprintf(gfid_path, sizeof(gfid_path), "%s/%s", priv->stub_basepath,
+ uuid_utoa(gfid));
+
+ ret = sys_stat(gfid_path, &st);
+ if (!ret)
+ goto out;
+ snprintf(bad_gfid_path, sizeof(bad_gfid_path), "%s/stub-%s",
+ priv->stub_basepath, uuid_utoa(priv->bad_object_dir_gfid));
+
+ ret = sys_link(bad_gfid_path, gfid_path);
+ if (ret) {
+ if ((errno != ENOENT) && (errno != EMLINK) && (errno != EEXIST))
+ goto out;
- snprintf (gfid_path, sizeof (gfid_path), "%s/%s",
- priv->stub_basepath, uuid_utoa (gfid));
-
- ret = sys_stat (gfid_path, &st);
- if (!ret)
- goto out;
- snprintf (bad_gfid_path, sizeof (bad_gfid_path), "%s/stub-%s",
- priv->stub_basepath, uuid_utoa (priv->bad_object_dir_gfid));
-
- ret = sys_link (bad_gfid_path, gfid_path);
- if (ret) {
- if ((errno != ENOENT) && (errno != EMLINK) && (errno != EEXIST))
- goto out;
-
- /*
- * Continue with success. At least we'll have half of the
- * functionality, in the sense, object is marked bad and
- * would be inaccessible. It's only scrub status that would
- * show up less number of objects. That's fine as we'll have
- * the log files that will have the missing information.
- */
- gf_msg (this->name, GF_LOG_WARNING, errno, BRS_MSG_LINK_FAIL,
- "failed to record gfid [%s]", uuid_utoa (gfid));
- }
+ /*
+ * Continue with success. At least we'll have half of the
+ * functionality, in the sense, object is marked bad and
+ * would be inaccessible. It's only scrub status that would
+ * show up less number of objects. That's fine as we'll have
+ * the log files that will have the missing information.
+ */
+ gf_smsg(this->name, GF_LOG_WARNING, errno, BRS_MSG_LINK_FAIL, "gfid=%s",
+ uuid_utoa(gfid), NULL);
+ }
- return 0;
+ return 0;
out:
- return -1;
+ return -1;
}
int
-br_stub_del (xlator_t *this, uuid_t gfid)
+br_stub_del(xlator_t *this, uuid_t gfid)
{
- int32_t op_errno __attribute__((unused)) = 0;
- br_stub_private_t *priv = NULL;
- int ret = 0;
- char gfid_path[BR_PATH_MAX_PLUS] = {0};
-
- priv = this->private;
- GF_ASSERT_AND_GOTO_WITH_ERROR (this->name, !gf_uuid_is_null (gfid),
- out, op_errno, EINVAL);
- snprintf (gfid_path, sizeof (gfid_path), "%s/%s",
- priv->stub_basepath, uuid_utoa (gfid));
- ret = sys_unlink (gfid_path);
- if (ret && (errno != ENOENT)) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- BRS_MSG_BAD_OBJ_UNLINK_FAIL,
- "%s: failed to delete bad object link from quarantine "
- "directory", gfid_path);
- ret = -errno;
- goto out;
- }
-
- ret = 0;
+ int32_t op_errno __attribute__((unused)) = 0;
+ br_stub_private_t *priv = NULL;
+ int ret = 0;
+ char gfid_path[BR_PATH_MAX_PLUS] = {0};
+
+ priv = this->private;
+ GF_ASSERT_AND_GOTO_WITH_ERROR(this->name, !gf_uuid_is_null(gfid), out,
+ op_errno, EINVAL);
+ snprintf(gfid_path, sizeof(gfid_path), "%s/%s", priv->stub_basepath,
+ uuid_utoa(gfid));
+ ret = sys_unlink(gfid_path);
+ if (ret && (errno != ENOENT)) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, BRS_MSG_BAD_OBJ_UNLINK_FAIL,
+ "path=%s", gfid_path, NULL);
+ ret = -errno;
+ goto out;
+ }
+
+ ret = 0;
out:
- return ret;
+ return ret;
}
static int
-br_stub_check_stub_directory (xlator_t *this, char *fullpath)
+br_stub_check_stub_directory(xlator_t *this, char *fullpath)
{
- int ret = 0;
- struct stat st = {0,};
- char oldpath[BR_PATH_MAX_PLUS] = {0};
- br_stub_private_t *priv = NULL;
-
- priv = this->private;
-
- snprintf (oldpath, sizeof (oldpath), "%s/%s",
- priv->export, OLD_BR_STUB_QUARANTINE_DIR);
-
- ret = sys_stat (fullpath, &st);
- if (!ret && !S_ISDIR (st.st_mode))
- goto error_return;
- if (ret) {
- if (errno != ENOENT)
- goto error_return;
- ret = sys_stat (oldpath, &st);
- if (ret)
- ret = mkdir_p (fullpath, 0600, _gf_true);
- else
- ret = sys_rename (oldpath, fullpath);
- }
-
+ int ret = 0;
+ struct stat st = {
+ 0,
+ };
+ char oldpath[BR_PATH_MAX_PLUS] = {0};
+ br_stub_private_t *priv = NULL;
+
+ priv = this->private;
+
+ snprintf(oldpath, sizeof(oldpath), "%s/%s", priv->export,
+ OLD_BR_STUB_QUARANTINE_DIR);
+
+ ret = sys_stat(fullpath, &st);
+ if (!ret && !S_ISDIR(st.st_mode))
+ goto error_return;
+ if (ret) {
+ if (errno != ENOENT)
+ goto error_return;
+ ret = sys_stat(oldpath, &st);
if (ret)
- gf_msg (this->name, GF_LOG_ERROR, errno,
- BRS_MSG_BAD_OBJECT_DIR_FAIL,
- "failed to create stub directory [%s]", fullpath);
- return ret;
+ ret = mkdir_p(fullpath, 0600, _gf_true);
+ else
+ ret = sys_rename(oldpath, fullpath);
+ }
+
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, errno, BRS_MSG_BAD_OBJECT_DIR_FAIL,
+ "create-path=%s", fullpath, NULL);
+ return ret;
error_return:
- gf_msg (this->name, GF_LOG_ERROR, errno,
- BRS_MSG_BAD_OBJECT_DIR_FAIL,
- "Failed to verify stub directory [%s]", fullpath);
- return -1;
+ gf_smsg(this->name, GF_LOG_ERROR, errno, BRS_MSG_BAD_OBJECT_DIR_FAIL,
+ "verify-path=%s", fullpath, NULL);
+ return -1;
}
/**
@@ -216,393 +213,399 @@ error_return:
* directory.
*/
static int
-br_stub_check_stub_file (xlator_t *this, char *path)
+br_stub_check_stub_file(xlator_t *this, char *path)
{
- int ret = 0;
- int fd = -1;
- struct stat st = {0,};
-
- ret = sys_stat (path, &st);
- if (!ret && !S_ISREG (st.st_mode))
- goto error_return;
- if (ret) {
- if (errno != ENOENT)
- goto error_return;
- fd = sys_creat (path, 0);
- if (fd < 0)
- gf_msg (this->name, GF_LOG_ERROR, errno,
- BRS_MSG_BAD_OBJECT_DIR_FAIL,
- "Failed ot create stub file [%s]", path);
- }
-
- if (fd >= 0) {
- sys_close (fd);
- ret = 0;
- }
+ int ret = 0;
+ int fd = -1;
+ struct stat st = {
+ 0,
+ };
+
+ ret = sys_stat(path, &st);
+ if (!ret && !S_ISREG(st.st_mode))
+ goto error_return;
+ if (ret) {
+ if (errno != ENOENT)
+ goto error_return;
+ fd = sys_creat(path, 0);
+ if (fd < 0)
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ BRS_MSG_BAD_OBJECT_DIR_FAIL, "create-path=%s", path, NULL);
+ }
+
+ if (fd >= 0) {
+ sys_close(fd);
+ ret = 0;
+ }
- return ret;
+ return ret;
error_return:
- gf_msg (this->name, GF_LOG_ERROR, errno,
- BRS_MSG_BAD_OBJECT_DIR_FAIL, "Failed ot verify stub file [%s]", path);
- return -1;
+ gf_smsg(this->name, GF_LOG_ERROR, errno, BRS_MSG_BAD_OBJECT_DIR_FAIL,
+ "verify-path=%s", path, NULL);
+ return -1;
}
int
-br_stub_dir_create (xlator_t *this, br_stub_private_t *priv)
+br_stub_dir_create(xlator_t *this, br_stub_private_t *priv)
{
- int ret = -1;
- char fullpath[BR_PATH_MAX_PLUS] = {0,};
- char stub_gfid_path[BR_PATH_MAX_PLUS] = {0,};
-
- gf_uuid_copy (priv->bad_object_dir_gfid, BR_BAD_OBJ_CONTAINER);
-
- strncpy (fullpath, priv->stub_basepath, sizeof (fullpath));
-
- snprintf (stub_gfid_path, sizeof (stub_gfid_path), "%s/stub-%s",
- priv->stub_basepath, uuid_utoa (priv->bad_object_dir_gfid));
-
- ret = br_stub_check_stub_directory (this, fullpath);
- if (ret)
- goto out;
- ret = br_stub_check_stub_file (this, stub_gfid_path);
- if (ret)
- goto out;
-
- return 0;
+ int ret = -1;
+ char fullpath[BR_PATH_MAX_PLUS] = {
+ 0,
+ };
+ char stub_gfid_path[BR_PATH_MAX_PLUS] = {
+ 0,
+ };
+
+ gf_uuid_copy(priv->bad_object_dir_gfid, BR_BAD_OBJ_CONTAINER);
+
+ if (snprintf(fullpath, sizeof(fullpath), "%s", priv->stub_basepath) >=
+ sizeof(fullpath))
+ goto out;
+
+ if (snprintf(stub_gfid_path, sizeof(stub_gfid_path), "%s/stub-%s",
+ priv->stub_basepath, uuid_utoa(priv->bad_object_dir_gfid)) >=
+ sizeof(stub_gfid_path))
+ goto out;
+
+ ret = br_stub_check_stub_directory(this, fullpath);
+ if (ret)
+ goto out;
+ ret = br_stub_check_stub_file(this, stub_gfid_path);
+ if (ret)
+ goto out;
+
+ return 0;
out:
- return -1;
+ return -1;
}
call_stub_t *
-__br_stub_dequeue (struct list_head *callstubs)
+__br_stub_dequeue(struct list_head *callstubs)
{
- call_stub_t *stub = NULL;
+ call_stub_t *stub = NULL;
- if (!list_empty (callstubs)) {
- stub = list_entry (callstubs->next, call_stub_t, list);
- list_del_init (&stub->list);
- }
+ if (!list_empty(callstubs)) {
+ stub = list_entry(callstubs->next, call_stub_t, list);
+ list_del_init(&stub->list);
+ }
- return stub;
+ return stub;
}
void
-__br_stub_enqueue (struct list_head *callstubs, call_stub_t *stub)
+__br_stub_enqueue(struct list_head *callstubs, call_stub_t *stub)
{
- list_add_tail (&stub->list, callstubs);
+ list_add_tail(&stub->list, callstubs);
}
void
-br_stub_worker_enqueue (xlator_t *this, call_stub_t *stub)
+br_stub_worker_enqueue(xlator_t *this, call_stub_t *stub)
{
- br_stub_private_t *priv = NULL;
-
- priv = this->private;
- pthread_mutex_lock (&priv->container.bad_lock);
- {
- __br_stub_enqueue (&priv->container.bad_queue, stub);
- pthread_cond_signal (&priv->container.bad_cond);
- }
- pthread_mutex_unlock (&priv->container.bad_lock);
+ br_stub_private_t *priv = NULL;
+
+ priv = this->private;
+ pthread_mutex_lock(&priv->container.bad_lock);
+ {
+ __br_stub_enqueue(&priv->container.bad_queue, stub);
+ pthread_cond_signal(&priv->container.bad_cond);
+ }
+ pthread_mutex_unlock(&priv->container.bad_lock);
}
void *
-br_stub_worker (void *data)
+br_stub_worker(void *data)
{
- br_stub_private_t *priv = NULL;
- xlator_t *this = NULL;
- call_stub_t *stub = NULL;
-
+ br_stub_private_t *priv = NULL;
+ xlator_t *this = NULL;
+ call_stub_t *stub = NULL;
- THIS = data;
- this = data;
- priv = this->private;
+ THIS = data;
+ this = data;
+ priv = this->private;
- for (;;) {
- pthread_mutex_lock (&priv->container.bad_lock);
- {
- while (list_empty (&priv->container.bad_queue)) {
- (void) pthread_cond_wait (&priv->container.bad_cond,
- &priv->container.bad_lock);
- }
-
- stub = __br_stub_dequeue (&priv->container.bad_queue);
- }
- pthread_mutex_unlock (&priv->container.bad_lock);
+ for (;;) {
+ pthread_mutex_lock(&priv->container.bad_lock);
+ {
+ while (list_empty(&priv->container.bad_queue)) {
+ (void)pthread_cond_wait(&priv->container.bad_cond,
+ &priv->container.bad_lock);
+ }
- if (stub) /* guard against spurious wakeups */
- call_resume (stub);
+ stub = __br_stub_dequeue(&priv->container.bad_queue);
}
+ pthread_mutex_unlock(&priv->container.bad_lock);
- return NULL;
+ if (stub) /* guard against spurious wakeups */
+ call_resume(stub);
+ }
+
+ return NULL;
}
int32_t
-br_stub_lookup_wrapper (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xattr_req)
+br_stub_lookup_wrapper(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ dict_t *xattr_req)
{
- br_stub_private_t *priv = NULL;
- struct stat lstatbuf = {0};
- int ret = 0;
- int32_t op_errno = EINVAL;
- int32_t op_ret = -1;
- struct iatt stbuf = {0, };
- struct iatt postparent = {0,};
- dict_t *xattr = NULL;
- gf_boolean_t ver_enabled = _gf_false;
-
- BR_STUB_VER_ENABLED_IN_CALLPATH(frame, ver_enabled);
- priv = this->private;
- BR_STUB_VER_COND_GOTO (priv, (!ver_enabled), done);
-
- VALIDATE_OR_GOTO (loc, done);
- if (gf_uuid_compare (loc->gfid, priv->bad_object_dir_gfid))
- goto done;
-
- ret = sys_lstat (priv->stub_basepath, &lstatbuf);
- if (ret) {
- gf_msg_debug (this->name, errno, "Stat failed on stub bad "
- "object dir");
- op_errno = errno;
- goto done;
- } else if (!S_ISDIR (lstatbuf.st_mode)) {
- gf_msg_debug (this->name, errno, "bad object container is not "
- "a directory");
- op_errno = ENOTDIR;
- goto done;
- }
-
- iatt_from_stat (&stbuf, &lstatbuf);
- gf_uuid_copy (stbuf.ia_gfid, priv->bad_object_dir_gfid);
-
- op_ret = op_errno = 0;
- xattr = dict_new ();
- if (!xattr) {
- op_ret = -1;
- op_errno = ENOMEM;
- }
+ br_stub_private_t *priv = NULL;
+ struct stat lstatbuf = {0};
+ int ret = 0;
+ int32_t op_errno = EINVAL;
+ int32_t op_ret = -1;
+ struct iatt stbuf = {
+ 0,
+ };
+ struct iatt postparent = {
+ 0,
+ };
+ dict_t *xattr = NULL;
+ gf_boolean_t ver_enabled = _gf_false;
+
+ BR_STUB_VER_ENABLED_IN_CALLPATH(frame, ver_enabled);
+ priv = this->private;
+ BR_STUB_VER_COND_GOTO(priv, (!ver_enabled), done);
+
+ VALIDATE_OR_GOTO(loc, done);
+ if (gf_uuid_compare(loc->gfid, priv->bad_object_dir_gfid))
+ goto done;
+
+ ret = sys_lstat(priv->stub_basepath, &lstatbuf);
+ if (ret) {
+ gf_msg_debug(this->name, errno,
+ "Stat failed on stub bad "
+ "object dir");
+ op_errno = errno;
+ goto done;
+ } else if (!S_ISDIR(lstatbuf.st_mode)) {
+ gf_msg_debug(this->name, errno,
+ "bad object container is not "
+ "a directory");
+ op_errno = ENOTDIR;
+ goto done;
+ }
+
+ iatt_from_stat(&stbuf, &lstatbuf);
+ gf_uuid_copy(stbuf.ia_gfid, priv->bad_object_dir_gfid);
+
+ op_ret = op_errno = 0;
+ xattr = dict_new();
+ if (!xattr) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ }
done:
- STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno,
- loc->inode, &stbuf, xattr, &postparent);
- if (xattr)
- dict_unref (xattr);
- return 0;
+ STACK_UNWIND_STRICT(lookup, frame, op_ret, op_errno, loc->inode, &stbuf,
+ xattr, &postparent);
+ if (xattr)
+ dict_unref(xattr);
+ return 0;
}
static int
-is_bad_gfid_file_current (char *filename, uuid_t gfid)
+is_bad_gfid_file_current(char *filename, uuid_t gfid)
{
- char current_stub_gfid[GF_UUID_BUF_SIZE + 16] = {0, };
+ char current_stub_gfid[GF_UUID_BUF_SIZE + 16] = {
+ 0,
+ };
- snprintf (current_stub_gfid, sizeof current_stub_gfid,
- "stub-%s", uuid_utoa(gfid));
- return (!strcmp(filename, current_stub_gfid));
+ snprintf(current_stub_gfid, sizeof current_stub_gfid, "stub-%s",
+ uuid_utoa(gfid));
+ return (!strcmp(filename, current_stub_gfid));
}
static void
-check_delete_stale_bad_file (xlator_t *this, char *filename)
+check_delete_stale_bad_file(xlator_t *this, char *filename)
{
- int ret = 0;
- struct stat st = {0};
- char filepath[BR_PATH_MAX_PLUS] = {0};
- br_stub_private_t *priv = NULL;
+ int ret = 0;
+ struct stat st = {0};
+ char filepath[BR_PATH_MAX_PLUS] = {0};
+ br_stub_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (is_bad_gfid_file_current (filename, priv->bad_object_dir_gfid))
- return;
+ if (is_bad_gfid_file_current(filename, priv->bad_object_dir_gfid))
+ return;
- snprintf (filepath, sizeof (filepath), "%s/%s",
- priv->stub_basepath, filename);
+ snprintf(filepath, sizeof(filepath), "%s/%s", priv->stub_basepath,
+ filename);
- ret = sys_stat (filepath, &st);
- if (!ret && st.st_nlink == 1)
- sys_unlink (filepath);
+ ret = sys_stat(filepath, &st);
+ if (!ret && st.st_nlink == 1)
+ sys_unlink(filepath);
}
static int
-br_stub_fill_readdir (fd_t *fd, br_stub_fd_t *fctx, DIR *dir, off_t off,
- size_t size, gf_dirent_t *entries)
+br_stub_fill_readdir(fd_t *fd, br_stub_fd_t *fctx, DIR *dir, off_t off,
+ size_t size, gf_dirent_t *entries)
{
- off_t in_case = -1;
- off_t last_off = 0;
- size_t filled = 0;
- int count = 0;
- int32_t this_size = -1;
- gf_dirent_t *this_entry = NULL;
- xlator_t *this = NULL;
- struct dirent *entry = NULL;
- struct dirent scratch[2] = {{0,},};
-
- this = THIS;
- if (!off) {
- rewinddir (dir);
- } else {
- seekdir (dir, off);
+ off_t in_case = -1;
+ off_t last_off = 0;
+ size_t filled = 0;
+ int count = 0;
+ int32_t this_size = -1;
+ gf_dirent_t *this_entry = NULL;
+ xlator_t *this = NULL;
+ struct dirent *entry = NULL;
+ struct dirent scratch[2] = {
+ {
+ 0,
+ },
+ };
+
+ this = THIS;
+ if (!off) {
+ rewinddir(dir);
+ } else {
+ seekdir(dir, off);
#ifndef GF_LINUX_HOST_OS
- if ((u_long)telldir(dir) != off &&
- off != fctx->bad_object.dir_eof) {
- gf_msg (THIS->name, GF_LOG_ERROR, 0,
- BRS_MSG_BAD_OBJECT_DIR_SEEK_FAIL,
- "seekdir(0x%llx) failed on dir=%p: "
- "Invalid argument (offset reused from "
- "another DIR * structure?)", off, dir);
- errno = EINVAL;
- count = -1;
- goto out;
- }
-#endif /* GF_LINUX_HOST_OS */
+ if ((u_long)telldir(dir) != off && off != fctx->bad_object.dir_eof) {
+ gf_smsg(THIS->name, GF_LOG_ERROR, 0,
+ BRS_MSG_BAD_OBJECT_DIR_SEEK_FAIL, "off=(0x%llx)", off,
+ "dir=%p", dir, NULL);
+ errno = EINVAL;
+ count = -1;
+ goto out;
}
-
- while (filled <= size) {
- in_case = (u_long)telldir (dir);
-
- if (in_case == -1) {
- gf_msg (THIS->name, GF_LOG_ERROR, 0,
- BRS_MSG_BAD_OBJECT_DIR_TELL_FAIL,
- "telldir failed on dir=%p: %s",
- dir, strerror (errno));
- goto out;
- }
-
- errno = 0;
- entry = sys_readdir (dir, scratch);
- if (!entry || errno != 0) {
- if (errno == EBADF) {
- gf_msg (THIS->name, GF_LOG_WARNING, 0,
- BRS_MSG_BAD_OBJECT_DIR_READ_FAIL,
- "readdir failed on dir=%p: %s",
- dir, strerror (errno));
- goto out;
- }
- break;
- }
-
- if (!strcmp (entry->d_name, ".") ||
- !strcmp (entry->d_name, ".."))
- continue;
-
- if (!strncmp (entry->d_name, "stub-",
- strlen ("stub-"))) {
- check_delete_stale_bad_file (this, entry->d_name);
- continue;
- }
-
- this_size = max (sizeof (gf_dirent_t),
- sizeof (gfs3_dirplist))
- + strlen (entry->d_name) + 1;
-
- if (this_size + filled > size) {
- seekdir (dir, in_case);
-#ifndef GF_LINUX_HOST_OS
- if ((u_long)telldir(dir) != in_case &&
- in_case != fctx->bad_object.dir_eof) {
- gf_msg (THIS->name, GF_LOG_ERROR, 0,
- BRS_MSG_BAD_OBJECT_DIR_SEEK_FAIL,
- "seekdir(0x%llx) failed on dir=%p: "
- "Invalid argument (offset reused from "
- "another DIR * structure?)",
- in_case, dir);
- errno = EINVAL;
- count = -1;
- goto out;
- }
#endif /* GF_LINUX_HOST_OS */
- break;
- }
-
- this_entry = gf_dirent_for_name (entry->d_name);
-
- if (!this_entry) {
- gf_msg (THIS->name, GF_LOG_ERROR, 0,
- BRS_MSG_NO_MEMORY,
- "could not create gf_dirent for entry %s: (%s)",
- entry->d_name, strerror (errno));
- goto out;
- }
- /*
- * we store the offset of next entry here, which is
- * probably not intended, but code using syncop_readdir()
- * (glfs-heal.c, afr-self-heald.c, pump.c) rely on it
- * for directory read resumption.
- */
- last_off = (u_long)telldir(dir);
- this_entry->d_off = last_off;
- this_entry->d_ino = entry->d_ino;
-
- list_add_tail (&this_entry->list, &entries->list);
-
- filled += this_size;
- count++;
- }
+ }
- if ((!sys_readdir (dir, scratch) && (errno == 0))) {
- /* Indicate EOF */
- errno = ENOENT;
- /* Remember EOF offset for later detection */
- fctx->bad_object.dir_eof = last_off;
+ while (filled <= size) {
+ in_case = (u_long)telldir(dir);
+
+ if (in_case == -1) {
+ gf_smsg(THIS->name, GF_LOG_ERROR, 0,
+ BRS_MSG_BAD_OBJECT_DIR_TELL_FAIL, "dir=%p", dir, "err=%s",
+ strerror(errno), NULL);
+ goto out;
}
-out:
- return count;
-}
-int32_t
-br_stub_readdir_wrapper (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t off, dict_t *xdata)
-{
- br_stub_fd_t *fctx = NULL;
- DIR *dir = NULL;
- int ret = -1;
- int32_t op_ret = -1;
- int32_t op_errno = 0;
- int count = 0;
- gf_dirent_t entries;
- gf_boolean_t xdata_unref = _gf_false;
- dict_t *dict = NULL;
-
- INIT_LIST_HEAD (&entries.list);
-
- fctx = br_stub_fd_ctx_get (this, fd);
- if (!fctx) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- BRS_MSG_GET_FD_CONTEXT_FAILED,
- "pfd is NULL, fd=%p", fd);
- op_errno = -ret;
- goto done;
+ errno = 0;
+ entry = sys_readdir(dir, scratch);
+ if (!entry || errno != 0) {
+ if (errno == EBADF) {
+ gf_smsg(THIS->name, GF_LOG_WARNING, 0,
+ BRS_MSG_BAD_OBJECT_DIR_READ_FAIL, "dir=%p", dir,
+ "err=%s", strerror(errno), NULL);
+ goto out;
+ }
+ break;
}
- dir = fctx->bad_object.dir;
+ if (!strcmp(entry->d_name, ".") || !strcmp(entry->d_name, ".."))
+ continue;
- if (!dir) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- BRS_MSG_BAD_HANDLE_DIR_NULL,
- "dir is NULL for fd=%p", fd);
- op_errno = EINVAL;
- goto done;
+ if (!strncmp(entry->d_name, "stub-", strlen("stub-"))) {
+ check_delete_stale_bad_file(this, entry->d_name);
+ continue;
}
- count = br_stub_fill_readdir (fd, fctx, dir, off, size, &entries);
+ this_size = max(sizeof(gf_dirent_t), sizeof(gfs3_dirplist)) +
+ strlen(entry->d_name) + 1;
- /* pick ENOENT to indicate EOF */
- op_errno = errno;
- op_ret = count;
+ if (this_size + filled > size) {
+ seekdir(dir, in_case);
+#ifndef GF_LINUX_HOST_OS
+ if ((u_long)telldir(dir) != in_case &&
+ in_case != fctx->bad_object.dir_eof) {
+ gf_smsg(THIS->name, GF_LOG_ERROR, 0,
+ BRS_MSG_BAD_OBJECT_DIR_SEEK_FAIL, "in_case=(0x%llx)",
+ in_case, "dir=%p", dir, NULL);
+ errno = EINVAL;
+ count = -1;
+ goto out;
+ }
+#endif /* GF_LINUX_HOST_OS */
+ break;
+ }
+
+ this_entry = gf_dirent_for_name(entry->d_name);
- dict = xdata;
- (void) br_stub_bad_objects_path (this, fd, &entries, &dict);
- if (!xdata && dict) {
- xdata = dict;
- xdata_unref = _gf_true;
+ if (!this_entry) {
+ gf_smsg(THIS->name, GF_LOG_ERROR, 0,
+ BRS_MSG_CREATE_GF_DIRENT_FAILED, "entry-name=%s",
+ entry->d_name, "err=%s", strerror(errno), NULL);
+ goto out;
}
+ /*
+ * we store the offset of next entry here, which is
+ * probably not intended, but code using syncop_readdir()
+ * (glfs-heal.c, afr-self-heald.c, pump.c) rely on it
+ * for directory read resumption.
+ */
+ last_off = (u_long)telldir(dir);
+ this_entry->d_off = last_off;
+ this_entry->d_ino = entry->d_ino;
+
+ list_add_tail(&this_entry->list, &entries->list);
+
+ filled += this_size;
+ count++;
+ }
+
+ if ((!sys_readdir(dir, scratch) && (errno == 0))) {
+ /* Indicate EOF */
+ errno = ENOENT;
+ /* Remember EOF offset for later detection */
+ fctx->bad_object.dir_eof = last_off;
+ }
+out:
+ return count;
+}
+
+int32_t
+br_stub_readdir_wrapper(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ size_t size, off_t off, dict_t *xdata)
+{
+ br_stub_fd_t *fctx = NULL;
+ DIR *dir = NULL;
+ int ret = -1;
+ int32_t op_ret = -1;
+ int32_t op_errno = 0;
+ int count = 0;
+ gf_dirent_t entries;
+ gf_boolean_t xdata_unref = _gf_false;
+ dict_t *dict = NULL;
+
+ INIT_LIST_HEAD(&entries.list);
+
+ fctx = br_stub_fd_ctx_get(this, fd);
+ if (!fctx) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_GET_FD_CONTEXT_FAILED,
+ "fd=%p", fd, NULL);
+ op_errno = -ret;
+ goto done;
+ }
+
+ dir = fctx->bad_object.dir;
+
+ if (!dir) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_BAD_HANDLE_DIR_NULL,
+ "fd=%p", fd, NULL);
+ op_errno = EINVAL;
+ goto done;
+ }
+
+ count = br_stub_fill_readdir(fd, fctx, dir, off, size, &entries);
+
+ /* pick ENOENT to indicate EOF */
+ op_errno = errno;
+ op_ret = count;
+
+ dict = xdata;
+ (void)br_stub_bad_objects_path(this, fd, &entries, &dict);
+ if (!xdata && dict) {
+ xdata = dict;
+ xdata_unref = _gf_true;
+ }
done:
- STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, &entries, xdata);
- gf_dirent_free (&entries);
- if (xdata_unref)
- dict_unref (xdata);
- return 0;
+ STACK_UNWIND_STRICT(readdir, frame, op_ret, op_errno, &entries, xdata);
+ gf_dirent_free(&entries);
+ if (xdata_unref)
+ dict_unref(xdata);
+ return 0;
}
/**
@@ -638,168 +641,156 @@ done:
* path will not be printed in scrub and only the gfid will be there.
**/
int
-br_stub_bad_objects_path (xlator_t *this, fd_t *fd, gf_dirent_t *entries,
- dict_t **dict)
+br_stub_bad_objects_path(xlator_t *this, fd_t *fd, gf_dirent_t *entries,
+ dict_t **dict)
{
- gf_dirent_t *entry = NULL;
- inode_t *inode = NULL;
- char *hpath = NULL;
- uuid_t gfid = {0};
- int ret = -1;
- dict_t *tmp_dict = NULL;
- char str_gfid[64] = {0};
-
- if (list_empty(&entries->list))
- return 0;
+ gf_dirent_t *entry = NULL;
+ inode_t *inode = NULL;
+ char *hpath = NULL;
+ uuid_t gfid = {0};
+ int ret = -1;
+ dict_t *tmp_dict = NULL;
+ char str_gfid[64] = {0};
+
+ if (list_empty(&entries->list))
+ return 0;
- tmp_dict = *dict;
+ tmp_dict = *dict;
+ if (!tmp_dict) {
+ tmp_dict = dict_new();
+ /*
+ * If the allocation of dict fails then no need treat it
+ * it as a error. This path (or function) is executed when
+ * "gluster volume bitrot <volume name> scrub status" is
+ * executed, to get the list of the corrupted objects.
+ * And the motive of this function is to get the paths of
+ * the corrupted objects. If the dict allocation fails, then
+ * the scrub status will only show the gfids of those corrupted
+ * objects (which is the behavior as of the time of this patch
+ * being worked upon). So just return and only the gfids will
+ * be shown.
+ */
if (!tmp_dict) {
- tmp_dict = dict_new ();
- /*
- * If the allocation of dict fails then no need treat it
- * it as a error. This path (or function) is executed when
- * "gluster volume bitrot <volume name> scrub status" is
- * executed, to get the list of the corrupted objects.
- * And the motive of this function is to get the paths of
- * the corrupted objects. If the dict allocation fails, then
- * the scrub status will only show the gfids of those corrupted
- * objects (which is the behavior as of the time of this patch
- * being worked upon). So just return and only the gfids will
- * be shown.
- */
- if (!tmp_dict) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRS_MSG_NO_MEMORY,
- "failed to allocate new dict for saving the paths "
- "of the corrupted objects. Scrub status will only "
- "display the gfid");
- goto out;
- }
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_ALLOC_FAILED, NULL);
+ goto out;
}
+ }
- list_for_each_entry (entry, &entries->list, list) {
- gf_uuid_clear (gfid);
- gf_uuid_parse (entry->d_name, gfid);
-
- inode = inode_find (fd->inode->table, gfid);
-
- /* No need to check the return value here.
- * Because @hpath is examined.
- */
- (void) br_stub_get_path_of_gfid (this, fd->inode, inode,
- gfid, &hpath);
-
- if (hpath) {
- gf_msg_debug (this->name, 0, "path of the corrupted "
- "object (gfid: %s) is %s",
- uuid_utoa (gfid), hpath);
- br_stub_entry_xattr_fill (this, hpath, entry, tmp_dict);
- } else
- gf_msg (this->name, GF_LOG_WARNING, 0,
- BRS_MSG_PATH_GET_FAILED,
- "failed to get the path for the inode %s",
- uuid_utoa_r (gfid, str_gfid));
-
- inode = NULL;
- hpath = NULL;
- }
+ list_for_each_entry(entry, &entries->list, list)
+ {
+ gf_uuid_clear(gfid);
+ gf_uuid_parse(entry->d_name, gfid);
+
+ inode = inode_find(fd->inode->table, gfid);
+
+ /* No need to check the return value here.
+ * Because @hpath is examined.
+ */
+ (void)br_stub_get_path_of_gfid(this, fd->inode, inode, gfid, &hpath);
- ret = 0;
- *dict = tmp_dict;
+ if (hpath) {
+ gf_msg_debug(this->name, 0,
+ "path of the corrupted "
+ "object (gfid: %s) is %s",
+ uuid_utoa(gfid), hpath);
+ br_stub_entry_xattr_fill(this, hpath, entry, tmp_dict);
+ } else
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_PATH_GET_FAILED,
+ "gfid=%s", uuid_utoa_r(gfid, str_gfid), NULL);
+
+ inode = NULL;
+ hpath = NULL;
+ }
+
+ ret = 0;
+ *dict = tmp_dict;
out:
- return ret;
- }
+ return ret;
+}
int
-br_stub_get_path_of_gfid (xlator_t *this, inode_t *parent, inode_t *inode,
- uuid_t gfid, char **path)
+br_stub_get_path_of_gfid(xlator_t *this, inode_t *parent, inode_t *inode,
+ uuid_t gfid, char **path)
{
- int32_t ret = -1;
- char gfid_str[64] = {0};
-
- GF_VALIDATE_OR_GOTO ("bitrot-stub", this, out);
- GF_VALIDATE_OR_GOTO (this->name, parent, out);
- GF_VALIDATE_OR_GOTO (this->name, path, out);
-
- /* Above, No need to validate the @inode for hard resolution. Because
- * inode can be NULL and if it is NULL, then syncop_gfid_to_path_hard
- * will allocate a new inode and proceed. So no need to bother about
- * @inode. Because we need it only to send a syncop_getxattr call
- * from inside syncop_gfid_to_path_hard. And getxattr fetches the
- * path from the backend.
- */
-
- ret = syncop_gfid_to_path_hard (parent->table, FIRST_CHILD (this), gfid,
- inode, path, _gf_true);
+ int32_t ret = -1;
+ char gfid_str[64] = {0};
+
+ GF_VALIDATE_OR_GOTO("bitrot-stub", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, parent, out);
+ GF_VALIDATE_OR_GOTO(this->name, path, out);
+
+ /* Above, No need to validate the @inode for hard resolution. Because
+ * inode can be NULL and if it is NULL, then syncop_gfid_to_path_hard
+ * will allocate a new inode and proceed. So no need to bother about
+ * @inode. Because we need it only to send a syncop_getxattr call
+ * from inside syncop_gfid_to_path_hard. And getxattr fetches the
+ * path from the backend.
+ */
+
+ ret = syncop_gfid_to_path_hard(parent->table, FIRST_CHILD(this), gfid,
+ inode, path, _gf_true);
+ if (ret < 0)
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_PATH_GET_FAILED,
+ "gfid=%s", uuid_utoa_r(gfid, gfid_str), NULL);
+
+ /*
+ * Try with soft resolution of path if hard resolve fails. Because
+ * checking the xattr on disk to get the path of a inode (or gfid)
+ * is dependent on whether that option is enabled in the posix
+ * xlator or not. If it is not enabled, then hard resolution by
+ * checking the on disk xattr fails.
+ *
+ * Thus in such situations fall back to the soft resolution which
+ * mainly depends on the inode_path() function. And for using
+ * inode_path, @inode has to be linked i.e. a successful lookup should
+ * have happened on the gfid (or the path) to link the inode to the
+ * inode table. And if @inode is NULL, means, the inode has not been
+ * found in the inode table and better not to do inode_path() on the
+ * inode which has not been linked.
+ */
+ if (ret < 0 && inode) {
+ ret = syncop_gfid_to_path_hard(parent->table, FIRST_CHILD(this), gfid,
+ inode, path, _gf_false);
if (ret < 0)
- gf_msg (this->name, GF_LOG_WARNING, 0, BRS_MSG_PATH_GET_FAILED,
- "failed to get the path xattr from disk for the "
- " gfid %s. Trying to get path from the memory",
- uuid_utoa_r (gfid, gfid_str));
-
- /*
- * Try with soft resolution of path if hard resolve fails. Because
- * checking the xattr on disk to get the path of a inode (or gfid)
- * is dependent on whether that option is enabled in the posix
- * xlator or not. If it is not enabled, then hard resolution by
- * checking the on disk xattr fails.
- *
- * Thus in such situations fall back to the soft resolution which
- * mainly depends on the inode_path() function. And for using
- * inode_path, @inode has to be linked i.e. a successful lookup should
- * have happened on the gfid (or the path) to link the inode to the
- * inode table. And if @inode is NULL, means, the inode has not been
- * found in the inode table and better not to do inode_path() on the
- * inode which has not been linked.
- */
- if (ret < 0 && inode) {
- ret = syncop_gfid_to_path_hard (parent->table,
- FIRST_CHILD (this), gfid, inode,
- path, _gf_false);
- if (ret < 0)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- BRS_MSG_PATH_GET_FAILED,
- "failed to get the path from the memory for gfid %s",
- uuid_utoa_r (gfid, gfid_str));
- }
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_PATH_GET_FAILED,
+ "from-memory gfid=%s", uuid_utoa_r(gfid, gfid_str), NULL);
+ }
out:
- return ret;
+ return ret;
}
-
/**
-* NOTE: If the file has multiple hardlinks (in gluster volume
-* namespace), the path would be one of the hardlinks. Its upto
-* the user to find the remaining hardlinks (using find -samefile)
-* and remove them.
-**/
+ * NOTE: If the file has multiple hardlinks (in gluster volume
+ * namespace), the path would be one of the hardlinks. Its up to
+ * the user to find the remaining hardlinks (using find -samefile)
+ * and remove them.
+ **/
void
-br_stub_entry_xattr_fill (xlator_t *this, char *hpath, gf_dirent_t *entry,
- dict_t *dict)
+br_stub_entry_xattr_fill(xlator_t *this, char *hpath, gf_dirent_t *entry,
+ dict_t *dict)
{
- int32_t ret = -1;
-
- GF_VALIDATE_OR_GOTO ("bit-rot-stub", this, out);
- GF_VALIDATE_OR_GOTO (this->name, hpath, out);
-
- /*
- * Use the entry->d_name (which is nothing but the gfid of the
- * corrupted object) as the key. And the value will be the actual
- * path of that object (or file).
- *
- * ALso ignore the dict_set errors. scrubber will get the gfid of
- * the corrupted object for sure. So, for now lets just log the
- * dict_set_dynstr failure and move on.
- */
-
- ret = dict_set_dynstr (dict, entry->d_name, hpath);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0, BRS_MSG_DICT_SET_FAILED,
- "failed to set the actual path %s as the value in the "
- "dict for the corrupted object %s", hpath,
- entry->d_name);
+ int32_t ret = -1;
+
+ GF_VALIDATE_OR_GOTO("bit-rot-stub", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, hpath, out);
+
+ /*
+ * Use the entry->d_name (which is nothing but the gfid of the
+ * corrupted object) as the key. And the value will be the actual
+ * path of that object (or file).
+ *
+ * ALso ignore the dict_set errors. scrubber will get the gfid of
+ * the corrupted object for sure. So, for now lets just log the
+ * dict_set_dynstr failure and move on.
+ */
+
+ ret = dict_set_dynstr(dict, entry->d_name, hpath);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_DICT_SET_FAILED,
+ "path=%s", hpath, "object-name=%s", entry->d_name, NULL);
out:
- return;
+ return;
}
diff --git a/xlators/features/bit-rot/src/stub/bit-rot-stub-mem-types.h b/xlators/features/bit-rot/src/stub/bit-rot-stub-mem-types.h
index 1652a3ebf8d..9d93caf069f 100644
--- a/xlators/features/bit-rot/src/stub/bit-rot-stub-mem-types.h
+++ b/xlators/features/bit-rot/src/stub/bit-rot-stub-mem-types.h
@@ -11,25 +11,26 @@
#ifndef _BR_MEM_TYPES_H
#define _BR_MEM_TYPES_H
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum br_mem_types {
- gf_br_stub_mt_private_t = gf_common_mt_end + 1,
- gf_br_stub_mt_version_t,
- gf_br_stub_mt_inode_ctx_t,
- gf_br_stub_mt_signature_t,
- gf_br_mt_br_private_t,
- gf_br_mt_br_child_t,
- gf_br_mt_br_object_t,
- gf_br_mt_br_ob_n_wk_t,
- gf_br_mt_br_scrubber_t,
- gf_br_mt_br_fsscan_entry_t,
- gf_br_stub_mt_br_stub_fd_t,
- gf_br_stub_mt_br_scanner_freq_t,
- gf_br_stub_mt_sigstub_t,
- gf_br_mt_br_child_event_t,
- gf_br_stub_mt_misc,
- gf_br_stub_mt_end,
+ gf_br_stub_mt_private_t = gf_common_mt_end + 1,
+ gf_br_stub_mt_version_t,
+ gf_br_stub_mt_inode_ctx_t,
+ gf_br_stub_mt_signature_t,
+ gf_br_mt_br_private_t,
+ gf_br_mt_br_child_t,
+ gf_br_mt_br_object_t,
+ gf_br_mt_br_ob_n_wk_t,
+ gf_br_mt_br_scrubber_t,
+ gf_br_mt_br_fsscan_entry_t,
+ gf_br_stub_mt_br_stub_fd_t,
+ gf_br_stub_mt_br_scanner_freq_t,
+ gf_br_stub_mt_sigstub_t,
+ gf_br_mt_br_child_event_t,
+ gf_br_stub_mt_misc,
+ gf_br_mt_br_worker_t,
+ gf_br_stub_mt_end,
};
#endif
diff --git a/xlators/features/bit-rot/src/stub/bit-rot-stub-messages.h b/xlators/features/bit-rot/src/stub/bit-rot-stub-messages.h
index 2ade4f03997..6c15a166f18 100644
--- a/xlators/features/bit-rot/src/stub/bit-rot-stub-messages.h
+++ b/xlators/features/bit-rot/src/stub/bit-rot-stub-messages.h
@@ -11,7 +11,7 @@
#ifndef _BITROT_STUB_MESSAGES_H_
#define _BITROT_STUB_MESSAGES_H_
-#include "glfs-message-id.h"
+#include <glusterfs/glfs-message-id.h>
/* To add new message IDs, append new identifiers at the end of the list.
*
@@ -23,40 +23,95 @@
* glfs-message-id.h.
*/
-GLFS_MSGID(BITROT_STUB,
- BRS_MSG_NO_MEMORY,
- BRS_MSG_SET_EVENT_FAILED,
- BRS_MSG_MEM_ACNT_FAILED,
- BRS_MSG_CREATE_FRAME_FAILED,
- BRS_MSG_SET_CONTEXT_FAILED,
- BRS_MSG_CHANGE_VERSION_FAILED,
- BRS_MSG_ADD_FD_TO_LIST_FAILED,
- BRS_MSG_SET_FD_CONTEXT_FAILED,
- BRS_MSG_CREATE_ANONYMOUS_FD_FAILED,
- BRS_MSG_NO_CHILD,
- BRS_MSG_STUB_ALLOC_FAILED,
- BRS_MSG_GET_INODE_CONTEXT_FAILED,
- BRS_MSG_CANCEL_SIGN_THREAD_FAILED,
- BRS_MSG_ADD_FD_TO_INODE,
- BRS_MSG_SIGN_VERSION_ERROR,
- BRS_MSG_BAD_OBJ_MARK_FAIL,
- BRS_MSG_NON_SCRUB_BAD_OBJ_MARK,
- BRS_MSG_REMOVE_INTERNAL_XATTR,
- BRS_MSG_SET_INTERNAL_XATTR,
- BRS_MSG_BAD_OBJECT_ACCESS,
- BRS_MSG_BAD_CONTAINER_FAIL,
- BRS_MSG_BAD_OBJECT_DIR_FAIL,
- BRS_MSG_BAD_OBJECT_DIR_SEEK_FAIL,
- BRS_MSG_BAD_OBJECT_DIR_TELL_FAIL,
- BRS_MSG_BAD_OBJECT_DIR_READ_FAIL,
- BRS_MSG_GET_FD_CONTEXT_FAILED,
- BRS_MSG_BAD_HANDLE_DIR_NULL,
- BRS_MSG_BAD_OBJ_THREAD_FAIL,
- BRS_MSG_BAD_OBJ_DIR_CLOSE_FAIL,
- BRS_MSG_LINK_FAIL,
- BRS_MSG_BAD_OBJ_UNLINK_FAIL,
- BRS_MSG_DICT_SET_FAILED,
- BRS_MSG_PATH_GET_FAILED
-);
+GLFS_MSGID(BITROT_STUB, BRS_MSG_NO_MEMORY, BRS_MSG_SET_EVENT_FAILED,
+ BRS_MSG_MEM_ACNT_FAILED, BRS_MSG_CREATE_FRAME_FAILED,
+ BRS_MSG_SET_CONTEXT_FAILED, BRS_MSG_CHANGE_VERSION_FAILED,
+ BRS_MSG_ADD_FD_TO_LIST_FAILED, BRS_MSG_SET_FD_CONTEXT_FAILED,
+ BRS_MSG_CREATE_ANONYMOUS_FD_FAILED, BRS_MSG_NO_CHILD,
+ BRS_MSG_STUB_ALLOC_FAILED, BRS_MSG_GET_INODE_CONTEXT_FAILED,
+ BRS_MSG_CANCEL_SIGN_THREAD_FAILED, BRS_MSG_ADD_FD_TO_INODE,
+ BRS_MSG_SIGN_VERSION_ERROR, BRS_MSG_BAD_OBJ_MARK_FAIL,
+ BRS_MSG_NON_SCRUB_BAD_OBJ_MARK, BRS_MSG_REMOVE_INTERNAL_XATTR,
+ BRS_MSG_SET_INTERNAL_XATTR, BRS_MSG_BAD_OBJECT_ACCESS,
+ BRS_MSG_BAD_CONTAINER_FAIL, BRS_MSG_BAD_OBJECT_DIR_FAIL,
+ BRS_MSG_BAD_OBJECT_DIR_SEEK_FAIL, BRS_MSG_BAD_OBJECT_DIR_TELL_FAIL,
+ BRS_MSG_BAD_OBJECT_DIR_READ_FAIL, BRS_MSG_GET_FD_CONTEXT_FAILED,
+ BRS_MSG_BAD_HANDLE_DIR_NULL, BRS_MSG_BAD_OBJ_THREAD_FAIL,
+ BRS_MSG_BAD_OBJ_DIR_CLOSE_FAIL, BRS_MSG_LINK_FAIL,
+ BRS_MSG_BAD_OBJ_UNLINK_FAIL, BRS_MSG_DICT_SET_FAILED,
+ BRS_MSG_PATH_GET_FAILED, BRS_MSG_NULL_LOCAL,
+ BRS_MSG_SPAWN_SIGN_THRD_FAILED, BRS_MSG_KILL_SIGN_THREAD,
+ BRS_MSG_NON_BITD_PID, BRS_MSG_SIGN_PREPARE_FAIL,
+ BRS_MSG_USING_DEFAULT_THREAD_SIZE, BRS_MSG_ALLOC_MEM_FAILED,
+ BRS_MSG_DICT_ALLOC_FAILED, BRS_MSG_CREATE_GF_DIRENT_FAILED,
+ BRS_MSG_ALLOC_FAILED, BRS_MSG_PATH_XATTR_GET_FAILED,
+ BRS_MSG_VERSION_PREPARE_FAIL);
+#define BRS_MSG_MEM_ACNT_FAILED_STR "Memory accounting init failed"
+#define BRS_MSG_BAD_OBJ_THREAD_FAIL_STR "pthread_init failed"
+#define BRS_MSG_USING_DEFAULT_THREAD_SIZE_STR "Using default thread stack size"
+#define BRS_MSG_NO_CHILD_STR "FATAL: no children"
+#define BRS_MSG_SPAWN_SIGN_THRD_FAILED_STR \
+ "failed to create the new thread for signer"
+#define BRS_MSG_BAD_CONTAINER_FAIL_STR \
+ "failed to launch the thread for storing bad gfids"
+#define BRS_MSG_CANCEL_SIGN_THREAD_FAILED_STR \
+ "Could not cancel sign serializer thread"
+#define BRS_MSG_KILL_SIGN_THREAD_STR "killed the signer thread"
+#define BRS_MSG_GET_INODE_CONTEXT_FAILED_STR \
+ "failed to init the inode context for the inode"
+#define BRS_MSG_ADD_FD_TO_INODE_STR "failed to add fd to the inode"
+#define BRS_MSG_NO_MEMORY_STR "local allocation failed"
+#define BRS_MSG_BAD_OBJECT_ACCESS_STR "bad object accessed. Returning"
+#define BRS_MSG_SIGN_VERSION_ERROR_STR "Signing version exceeds current version"
+#define BRS_MSG_NON_BITD_PID_STR \
+ "PID from where signature request came, does not belong to bit-rot " \
+ "daemon. Unwinding the fop"
+#define BRS_MSG_SIGN_PREPARE_FAIL_STR \
+ "failed to prepare the signature. Unwinding the fop"
+#define BRS_MSG_VERSION_PREPARE_FAIL_STR \
+ "failed to prepare the version. Unwinding the fop"
+#define BRS_MSG_STUB_ALLOC_FAILED_STR "failed to allocate stub fop, Unwinding"
+#define BRS_MSG_BAD_OBJ_MARK_FAIL_STR "failed to mark object as bad"
+#define BRS_MSG_NON_SCRUB_BAD_OBJ_MARK_STR \
+ "bad object marking is not from the scrubber"
+#define BRS_MSG_ALLOC_MEM_FAILED_STR "failed to allocate memory"
+#define BRS_MSG_SET_INTERNAL_XATTR_STR "called on the internal xattr"
+#define BRS_MSG_REMOVE_INTERNAL_XATTR_STR "removexattr called on internal xattr"
+#define BRS_MSG_CREATE_ANONYMOUS_FD_FAILED_STR \
+ "failed to create anonymous fd for the inode"
+#define BRS_MSG_ADD_FD_TO_LIST_FAILED_STR "failed add fd to the list"
+#define BRS_MSG_SET_FD_CONTEXT_FAILED_STR \
+ "failed to set the fd context for the file"
+#define BRS_MSG_NULL_LOCAL_STR "local is NULL"
+#define BRS_MSG_DICT_ALLOC_FAILED_STR \
+ "dict allocation failed: cannot send IPC FOP to changelog"
+#define BRS_MSG_SET_EVENT_FAILED_STR "cannot set release event in dict"
+#define BRS_MSG_CREATE_FRAME_FAILED_STR "create_frame() failure"
+#define BRS_MSG_BAD_OBJ_DIR_CLOSE_FAIL_STR "closedir error"
+#define BRS_MSG_LINK_FAIL_STR "failed to record gfid"
+#define BRS_MSG_BAD_OBJ_UNLINK_FAIL_STR \
+ "failed to delete bad object link from quaratine directory"
+#define BRS_MSG_BAD_OBJECT_DIR_FAIL_STR "failed stub directory"
+#define BRS_MSG_BAD_OBJECT_DIR_SEEK_FAIL_STR \
+ "seekdir failed. Invalid argument (offset reused from another DIR * " \
+ "structure)"
+#define BRS_MSG_BAD_OBJECT_DIR_TELL_FAIL_STR "telldir failed on dir"
+#define BRS_MSG_BAD_OBJECT_DIR_READ_FAIL_STR "readdir failed on dir"
+#define BRS_MSG_CREATE_GF_DIRENT_FAILED_STR "could not create gf_dirent"
+#define BRS_MSG_GET_FD_CONTEXT_FAILED_STR "pfd is NULL"
+#define BRS_MSG_BAD_HANDLE_DIR_NULL_STR "dir if NULL"
+#define BRS_MSG_ALLOC_FAILED_STR \
+ "failed to allocate new dict for saving the paths of the corrupted " \
+ "objects. Scrub status will only display the gfid"
+#define BRS_MSG_PATH_GET_FAILED_STR "failed to get the path"
+#define BRS_MSG_PATH_XATTR_GET_FAILED_STR \
+ "failed to get the path xattr from disk for the gfid. Trying to get path " \
+ "from the memory"
+#define BRS_MSG_DICT_SET_FAILED_STR \
+ "failed to set the actual path as the value in the dict for the " \
+ "corrupted object"
+#define BRS_MSG_SET_CONTEXT_FAILED_STR \
+ "could not set fd context for release callback"
+#define BRS_MSG_CHANGE_VERSION_FAILED_STR "change version failed"
#endif /* !_BITROT_STUB_MESSAGES_H_ */
diff --git a/xlators/features/bit-rot/src/stub/bit-rot-stub.c b/xlators/features/bit-rot/src/stub/bit-rot-stub.c
index 76a4c3f6962..447dd47ff41 100644
--- a/xlators/features/bit-rot/src/stub/bit-rot-stub.c
+++ b/xlators/features/bit-rot/src/stub/bit-rot-stub.c
@@ -12,180 +12,193 @@
#include <sys/uio.h>
#include <signal.h>
-#include "glusterfs.h"
-#include "xlator.h"
-#include "logging.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/logging.h>
#include "changelog.h"
-#include "compat-errno.h"
-#include "call-stub.h"
+#include <glusterfs/compat-errno.h>
+#include <glusterfs/call-stub.h>
#include "bit-rot-stub.h"
#include "bit-rot-stub-mem-types.h"
#include "bit-rot-stub-messages.h"
#include "bit-rot-common.h"
-#define BR_STUB_REQUEST_COOKIE 0x1
+#define BR_STUB_REQUEST_COOKIE 0x1
-void *br_stub_signth (void *);
+void
+br_stub_lock_cleaner(void *arg)
+{
+ pthread_mutex_t *clean_mutex = arg;
+
+ pthread_mutex_unlock(clean_mutex);
+ return;
+}
+
+void *
+br_stub_signth(void *);
struct br_stub_signentry {
- unsigned long v;
+ unsigned long v;
- call_stub_t *stub;
+ call_stub_t *stub;
- struct list_head list;
+ struct list_head list;
};
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int32_t ret = -1;
-
- if (!this)
- return ret;
-
- ret = xlator_mem_acct_init (this, gf_br_stub_mt_end + 1);
-
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_WARNING, 0, BRS_MSG_MEM_ACNT_FAILED,
- "Memory accounting init failed");
- return ret;
- }
+ int32_t ret = -1;
+ if (!this)
return ret;
-}
-
-int32_t
-br_stub_bad_object_container_init (xlator_t *this, br_stub_private_t *priv)
-{
- pthread_attr_t w_attr;
- int32_t ret = -1;
-
- ret = pthread_cond_init(&priv->container.bad_cond, NULL);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_BAD_OBJ_THREAD_FAIL,
- "pthread_cond_init failed (%d)", ret);
- goto out;
- }
- ret = pthread_mutex_init(&priv->container.bad_lock, NULL);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_BAD_OBJ_THREAD_FAIL,
- "pthread_mutex_init failed (%d)", ret);
- goto cleanup_cond;
- }
+ ret = xlator_mem_acct_init(this, gf_br_stub_mt_end + 1);
- ret = pthread_attr_init (&w_attr);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_BAD_OBJ_THREAD_FAIL,
- "pthread_attr_init failed (%d)", ret);
- goto cleanup_lock;
- }
-
- ret = pthread_attr_setstacksize (&w_attr, BAD_OBJECT_THREAD_STACK_SIZE);
- if (ret == EINVAL) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- BRS_MSG_BAD_OBJ_THREAD_FAIL,
- "Using default thread stack size");
- }
-
- INIT_LIST_HEAD (&priv->container.bad_queue);
- ret = br_stub_dir_create (this, priv);
- if (ret < 0)
- goto cleanup_lock;
+ if (ret != 0) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_MEM_ACNT_FAILED, NULL);
+ return ret;
+ }
- ret = gf_thread_create (&priv->container.thread, &w_attr,
- br_stub_worker, this, "brswrker");
- if (ret)
- goto cleanup_attr;
+ return ret;
+}
- return 0;
+int
+br_stub_bad_object_container_init(xlator_t *this, br_stub_private_t *priv)
+{
+ pthread_attr_t w_attr;
+ int ret = -1;
+
+ ret = pthread_cond_init(&priv->container.bad_cond, NULL);
+ if (ret != 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_BAD_OBJ_THREAD_FAIL,
+ "cond_init ret=%d", ret, NULL);
+ goto out;
+ }
+
+ ret = pthread_mutex_init(&priv->container.bad_lock, NULL);
+ if (ret != 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_BAD_OBJ_THREAD_FAIL,
+ "mutex_init ret=%d", ret, NULL);
+ goto cleanup_cond;
+ }
+
+ ret = pthread_attr_init(&w_attr);
+ if (ret != 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_BAD_OBJ_THREAD_FAIL,
+ "attr_init ret=%d", ret, NULL);
+ goto cleanup_lock;
+ }
+
+ ret = pthread_attr_setstacksize(&w_attr, BAD_OBJECT_THREAD_STACK_SIZE);
+ if (ret == EINVAL) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0,
+ BRS_MSG_USING_DEFAULT_THREAD_SIZE, NULL);
+ }
+
+ INIT_LIST_HEAD(&priv->container.bad_queue);
+ ret = br_stub_dir_create(this, priv);
+ if (ret < 0)
+ goto cleanup_lock;
+
+ ret = gf_thread_create(&priv->container.thread, &w_attr, br_stub_worker,
+ this, "brswrker");
+ if (ret)
+ goto cleanup_attr;
+
+ return 0;
cleanup_attr:
- pthread_attr_destroy (&w_attr);
+ pthread_attr_destroy(&w_attr);
cleanup_lock:
- pthread_mutex_destroy (&priv->container.bad_lock);
+ pthread_mutex_destroy(&priv->container.bad_lock);
cleanup_cond:
- pthread_cond_destroy (&priv->container.bad_cond);
+ pthread_cond_destroy(&priv->container.bad_cond);
out:
- return -1;
+ return -1;
}
int32_t
-init (xlator_t *this)
+init(xlator_t *this)
{
- int32_t ret = 0;
- char *tmp = NULL;
- struct timeval tv = {0,};
- br_stub_private_t *priv = NULL;
+ int ret = 0;
+ char *tmp = NULL;
+ struct timeval tv = {
+ 0,
+ };
+ br_stub_private_t *priv = NULL;
- if (!this->children) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRS_MSG_NO_CHILD,
- "FATAL: no children");
- goto error_return;
- }
+ if (!this->children) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_NO_CHILD, NULL);
+ goto error_return;
+ }
- priv = GF_CALLOC (1, sizeof (*priv), gf_br_stub_mt_private_t);
- if (!priv)
- goto error_return;
+ priv = GF_CALLOC(1, sizeof(*priv), gf_br_stub_mt_private_t);
+ if (!priv)
+ goto error_return;
- priv->local_pool = mem_pool_new (br_stub_local_t, 512);
- if (!priv->local_pool)
- goto free_priv;
+ priv->local_pool = mem_pool_new(br_stub_local_t, 512);
+ if (!priv->local_pool)
+ goto free_priv;
- GF_OPTION_INIT ("bitrot", priv->do_versioning, bool, free_mempool);
+ GF_OPTION_INIT("bitrot", priv->do_versioning, bool, free_mempool);
- GF_OPTION_INIT ("export", tmp, str, free_mempool);
- strncpy (priv->export, tmp, PATH_MAX-1);
- priv->export[PATH_MAX-1] = '\0';
+ GF_OPTION_INIT("export", tmp, str, free_mempool);
- (void) snprintf (priv->stub_basepath, sizeof (priv->stub_basepath),
- "%s/%s", priv->export, BR_STUB_QUARANTINE_DIR);
+ if (snprintf(priv->export, PATH_MAX, "%s", tmp) >= PATH_MAX)
+ goto free_mempool;
- (void) gettimeofday (&tv, NULL);
+ if (snprintf(priv->stub_basepath, sizeof(priv->stub_basepath), "%s/%s",
+ priv->export,
+ BR_STUB_QUARANTINE_DIR) >= sizeof(priv->stub_basepath))
+ goto free_mempool;
- /* boot time is in network endian format */
- priv->boot[0] = htonl (tv.tv_sec);
- priv->boot[1] = htonl (tv.tv_usec);
+ (void)gettimeofday(&tv, NULL);
- pthread_mutex_init (&priv->lock, NULL);
- pthread_cond_init (&priv->cond, NULL);
- INIT_LIST_HEAD (&priv->squeue);
+ /* boot time is in network endian format */
+ priv->boot[0] = htonl(tv.tv_sec);
+ priv->boot[1] = htonl(tv.tv_usec);
- /* Thread creations need 'this' to be passed so that THIS can be
- * assigned inside the thread. So setting this->private here.
- */
- this->private = priv;
+ pthread_mutex_init(&priv->lock, NULL);
+ pthread_cond_init(&priv->cond, NULL);
+ INIT_LIST_HEAD(&priv->squeue);
- ret = gf_thread_create (&priv->signth, NULL, br_stub_signth, this,
- "brssign");
- if (ret != 0)
- goto cleanup_lock;
+ /* Thread creations need 'this' to be passed so that THIS can be
+ * assigned inside the thread. So setting this->private here.
+ */
+ this->private = priv;
+ if (!priv->do_versioning)
+ return 0;
- ret = br_stub_bad_object_container_init (this, priv);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRS_MSG_BAD_CONTAINER_FAIL,
- "failed to launch the thread for storing bad gfids");
- goto cleanup_lock;
- }
+ ret = gf_thread_create(&priv->signth, NULL, br_stub_signth, this,
+ "brssign");
+ if (ret != 0) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_SPAWN_SIGN_THRD_FAILED,
+ NULL);
+ goto cleanup_lock;
+ }
- gf_msg_debug (this->name, 0, "bit-rot stub loaded");
+ ret = br_stub_bad_object_container_init(this, priv);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_BAD_CONTAINER_FAIL, NULL);
+ goto cleanup_lock;
+ }
- return 0;
+ gf_msg_debug(this->name, 0, "bit-rot stub loaded");
- cleanup_lock:
- pthread_cond_destroy (&priv->cond);
- pthread_mutex_destroy (&priv->lock);
- free_mempool:
- mem_pool_destroy (priv->local_pool);
- free_priv:
- GF_FREE (priv);
- this->private = NULL;
- error_return:
- return -1;
+ return 0;
+
+cleanup_lock:
+ pthread_cond_destroy(&priv->cond);
+ pthread_mutex_destroy(&priv->lock);
+free_mempool:
+ mem_pool_destroy(priv->local_pool);
+ priv->local_pool = NULL;
+free_priv:
+ GF_FREE(priv);
+ this->private = NULL;
+error_return:
+ return -1;
}
/* TODO:
@@ -196,188 +209,234 @@ init (xlator_t *this)
* incremented upon modification.
* So object versioning is tied to bitrot daemon's signing. In future, object
* versioning might be necessary for other things as well apart from bit-rot
- * detection (well thats the objective of bringing in object-versioning :)).
+ * detection (well that's the objective of bringing in object-versioning :)).
* In that case, better to make versioning a new option and letting it to be
* enabled despite bit-rot detection is not needed.
* Ex: ICAP.
*/
int32_t
-reconfigure (xlator_t *this, dict_t *options)
+reconfigure(xlator_t *this, dict_t *options)
{
- int32_t ret = -1;
- br_stub_private_t *priv = NULL;
+ int32_t ret = -1;
+ br_stub_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- GF_OPTION_RECONF ("bitrot", priv->do_versioning, options, bool, out);
+ GF_OPTION_RECONF("bitrot", priv->do_versioning, options, bool, err);
+ if (priv->do_versioning && !priv->signth) {
+ ret = gf_thread_create(&priv->signth, NULL, br_stub_signth, this,
+ "brssign");
+ if (ret != 0) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0,
+ BRS_MSG_SPAWN_SIGN_THRD_FAILED, NULL);
+ goto err;
+ }
- ret = 0;
-out:
- return ret;
+ ret = br_stub_bad_object_container_init(this, priv);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_BAD_CONTAINER_FAIL,
+ NULL);
+ goto err;
+ }
+ } else {
+ if (priv->signth) {
+ if (gf_thread_cleanup_xint(priv->signth)) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ BRS_MSG_CANCEL_SIGN_THREAD_FAILED, NULL);
+ } else {
+ gf_smsg(this->name, GF_LOG_INFO, 0, BRS_MSG_KILL_SIGN_THREAD,
+ NULL);
+ priv->signth = 0;
+ }
+ }
+
+ if (priv->container.thread) {
+ if (gf_thread_cleanup_xint(priv->container.thread)) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ BRS_MSG_CANCEL_SIGN_THREAD_FAILED, NULL);
+ }
+ priv->container.thread = 0;
+ }
+ }
+
+ ret = 0;
+ return ret;
+err:
+ if (priv->signth) {
+ if (gf_thread_cleanup_xint(priv->signth)) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ BRS_MSG_CANCEL_SIGN_THREAD_FAILED, NULL);
+ }
+ priv->signth = 0;
+ }
+
+ if (priv->container.thread) {
+ if (gf_thread_cleanup_xint(priv->container.thread)) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ BRS_MSG_CANCEL_SIGN_THREAD_FAILED, NULL);
+ }
+ priv->container.thread = 0;
+ }
+ ret = -1;
+ return ret;
}
-
int
-notify (xlator_t *this, int event, void *data, ...)
-{
- br_stub_private_t *priv = NULL;
-
- if (!this)
- return 0;
-
- priv = this->private;
- if (!priv)
- return 0;
-
- switch (event) {
- case GF_EVENT_CLEANUP:
- if (priv->signth) {
- (void) gf_thread_cleanup_xint (priv->signth);
- priv->signth = 0;
- }
- if (priv->container.thread) {
- (void) gf_thread_cleanup_xint (priv->container.thread);
- priv->container.thread = 0;
- }
- break;
- }
- default_notify (this, event, data);
+notify(xlator_t *this, int event, void *data, ...)
+{
+ br_stub_private_t *priv = NULL;
+
+ if (!this)
return 0;
+
+ priv = this->private;
+ if (!priv)
+ return 0;
+
+ default_notify(this, event, data);
+ return 0;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- int32_t ret = 0;
- br_stub_private_t *priv = this->private;
- struct br_stub_signentry *sigstub = NULL;
- call_stub_t *stub = NULL;
+ int32_t ret = 0;
+ br_stub_private_t *priv = this->private;
+ struct br_stub_signentry *sigstub = NULL;
+ call_stub_t *stub = NULL;
- if (!priv)
- return;
+ if (!priv)
+ return;
- ret = gf_thread_cleanup_xint (priv->signth);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_CANCEL_SIGN_THREAD_FAILED,
- "Could not cancel sign serializer thread");
- goto out;
- }
+ if (!priv->do_versioning)
+ goto cleanup;
- while (!list_empty (&priv->squeue)) {
- sigstub = list_first_entry (&priv->squeue,
- struct br_stub_signentry, list);
- list_del_init (&sigstub->list);
+ ret = gf_thread_cleanup_xint(priv->signth);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_CANCEL_SIGN_THREAD_FAILED,
+ NULL);
+ goto out;
+ }
+ priv->signth = 0;
- call_stub_destroy (sigstub->stub);
- GF_FREE (sigstub);
- }
+ while (!list_empty(&priv->squeue)) {
+ sigstub = list_first_entry(&priv->squeue, struct br_stub_signentry,
+ list);
+ list_del_init(&sigstub->list);
- pthread_mutex_destroy (&priv->lock);
- pthread_cond_destroy (&priv->cond);
+ call_stub_destroy(sigstub->stub);
+ GF_FREE(sigstub);
+ }
- ret = gf_thread_cleanup_xint (priv->container.thread);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_CANCEL_SIGN_THREAD_FAILED,
- "Could not cancel sign serializer thread");
- goto out;
- }
+ ret = gf_thread_cleanup_xint(priv->container.thread);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_CANCEL_SIGN_THREAD_FAILED,
+ NULL);
+ goto out;
+ }
+
+ priv->container.thread = 0;
+
+ while (!list_empty(&priv->container.bad_queue)) {
+ stub = list_first_entry(&priv->container.bad_queue, call_stub_t, list);
+ list_del_init(&stub->list);
+ call_stub_destroy(stub);
+ }
- while (!list_empty (&priv->container.bad_queue)) {
- stub = list_first_entry (&priv->container.bad_queue, call_stub_t,
- list);
- list_del_init (&stub->list);
- call_stub_destroy (stub);
- };
+ pthread_mutex_destroy(&priv->container.bad_lock);
+ pthread_cond_destroy(&priv->container.bad_cond);
- pthread_mutex_destroy (&priv->container.bad_lock);
- pthread_cond_destroy (&priv->container.bad_cond);
+cleanup:
+ pthread_mutex_destroy(&priv->lock);
+ pthread_cond_destroy(&priv->cond);
- this->private = NULL;
- GF_FREE (priv);
+ if (priv->local_pool) {
+ mem_pool_destroy(priv->local_pool);
+ priv->local_pool = NULL;
+ }
- out:
- return;
+ this->private = NULL;
+ GF_FREE(priv);
+
+out:
+ return;
}
static int
-br_stub_alloc_versions (br_version_t **obuf,
- br_signature_t **sbuf, size_t signaturelen)
+br_stub_alloc_versions(br_version_t **obuf, br_signature_t **sbuf,
+ size_t signaturelen)
{
- void *mem = NULL;
- size_t size = 0;
+ void *mem = NULL;
+ size_t size = 0;
- if (obuf)
- size += sizeof (br_version_t);
- if (sbuf)
- size += sizeof (br_signature_t) + signaturelen;
+ if (obuf)
+ size += sizeof(br_version_t);
+ if (sbuf)
+ size += sizeof(br_signature_t) + signaturelen;
- mem = GF_CALLOC (1, size, gf_br_stub_mt_version_t);
- if (!mem)
- goto error_return;
+ mem = GF_CALLOC(1, size, gf_br_stub_mt_version_t);
+ if (!mem)
+ goto error_return;
- if (obuf) {
- *obuf = (br_version_t *)mem;
- mem = ((char *)mem + sizeof (br_version_t));
- }
- if (sbuf) {
- *sbuf = (br_signature_t *)mem;
- }
+ if (obuf) {
+ *obuf = (br_version_t *)mem;
+ mem = ((char *)mem + sizeof(br_version_t));
+ }
+ if (sbuf) {
+ *sbuf = (br_signature_t *)mem;
+ }
- return 0;
+ return 0;
- error_return:
- return -1;
+error_return:
+ return -1;
}
static void
-br_stub_dealloc_versions (void *mem)
+br_stub_dealloc_versions(void *mem)
{
- GF_FREE (mem);
+ GF_FREE(mem);
}
static br_stub_local_t *
-br_stub_alloc_local (xlator_t *this)
+br_stub_alloc_local(xlator_t *this)
{
- br_stub_private_t *priv = this->private;
+ br_stub_private_t *priv = this->private;
- return mem_get0 (priv->local_pool);
+ return mem_get0(priv->local_pool);
}
static void
-br_stub_dealloc_local (br_stub_local_t *ptr)
+br_stub_dealloc_local(br_stub_local_t *ptr)
{
- if (!ptr)
- return;
+ if (!ptr)
+ return;
- mem_put (ptr);
+ mem_put(ptr);
}
static int
-br_stub_prepare_version_request (xlator_t *this, dict_t *dict,
+br_stub_prepare_version_request(xlator_t *this, dict_t *dict,
br_version_t *obuf, unsigned long oversion)
{
- br_stub_private_t *priv = NULL;
+ br_stub_private_t *priv = NULL;
- priv = this->private;
- br_set_ongoingversion (obuf, oversion, priv->boot);
+ priv = this->private;
+ br_set_ongoingversion(obuf, oversion, priv->boot);
- return dict_set_static_bin (dict, BITROT_CURRENT_VERSION_KEY,
- (void *)obuf, sizeof (br_version_t));
+ return dict_set_bin(dict, BITROT_CURRENT_VERSION_KEY, (void *)obuf,
+ sizeof(br_version_t));
}
static int
-br_stub_prepare_signing_request (dict_t *dict,
- br_signature_t *sbuf,
- br_isignature_t *sign, size_t signaturelen)
+br_stub_prepare_signing_request(dict_t *dict, br_signature_t *sbuf,
+ br_isignature_t *sign, size_t signaturelen)
{
- size_t size = 0;
+ size_t size = 0;
- br_set_signature (sbuf, sign, signaturelen, &size);
+ br_set_signature(sbuf, sign, signaturelen, &size);
- return dict_set_static_bin (dict, BITROT_SIGNING_VERSION_KEY,
- (void *)sbuf, size);
+ return dict_set_bin(dict, BITROT_SIGNING_VERSION_KEY, (void *)sbuf, size);
}
/**
@@ -387,253 +446,245 @@ br_stub_prepare_signing_request (dict_t *dict,
* initializes the transient inode version.
*/
static int
-br_stub_init_inode_versions (xlator_t *this, fd_t *fd, inode_t *inode,
- unsigned long version, gf_boolean_t markdirty,
- gf_boolean_t bad_object, uint64_t *ctx_addr)
-{
- int32_t ret = 0;
- br_stub_inode_ctx_t *ctx = NULL;
-
- ctx = GF_CALLOC (1, sizeof (br_stub_inode_ctx_t),
- gf_br_stub_mt_inode_ctx_t);
- if (!ctx)
- goto error_return;
-
- INIT_LIST_HEAD (&ctx->fd_list);
- (markdirty) ? __br_stub_mark_inode_dirty (ctx)
- : __br_stub_mark_inode_synced (ctx);
- __br_stub_set_ongoing_version (ctx, version);
-
- if (bad_object)
- __br_stub_mark_object_bad (ctx);
-
- if (fd) {
- ret = br_stub_add_fd_to_inode (this, fd, ctx);
- if (ret)
- goto free_ctx;
- }
+br_stub_init_inode_versions(xlator_t *this, fd_t *fd, inode_t *inode,
+ unsigned long version, gf_boolean_t markdirty,
+ gf_boolean_t bad_object, uint64_t *ctx_addr)
+{
+ int32_t ret = 0;
+ br_stub_inode_ctx_t *ctx = NULL;
+
+ ctx = GF_CALLOC(1, sizeof(br_stub_inode_ctx_t), gf_br_stub_mt_inode_ctx_t);
+ if (!ctx)
+ goto error_return;
- ret = br_stub_set_inode_ctx (this, inode, ctx);
+ INIT_LIST_HEAD(&ctx->fd_list);
+ (markdirty) ? __br_stub_mark_inode_dirty(ctx)
+ : __br_stub_mark_inode_synced(ctx);
+ __br_stub_set_ongoing_version(ctx, version);
+
+ if (bad_object)
+ __br_stub_mark_object_bad(ctx);
+
+ if (fd) {
+ ret = br_stub_add_fd_to_inode(this, fd, ctx);
if (ret)
- goto free_ctx;
+ goto free_ctx;
+ }
- if (ctx_addr)
- *ctx_addr = (uint64_t) ctx;
- return 0;
+ ret = br_stub_set_inode_ctx(this, inode, ctx);
+ if (ret)
+ goto free_ctx;
+
+ if (ctx_addr)
+ *ctx_addr = (uint64_t)(uintptr_t)ctx;
+ return 0;
free_ctx:
- GF_FREE (ctx);
- error_return:
- return -1;
+ GF_FREE(ctx);
+error_return:
+ return -1;
}
/**
* modify the ongoing version of an inode.
*/
static int
-br_stub_mod_inode_versions (xlator_t *this,
- fd_t *fd, inode_t *inode, unsigned long version)
+br_stub_mod_inode_versions(xlator_t *this, fd_t *fd, inode_t *inode,
+ unsigned long version)
{
- int32_t ret = -1;
- br_stub_inode_ctx_t *ctx = 0;
+ int32_t ret = -1;
+ br_stub_inode_ctx_t *ctx = 0;
- LOCK (&inode->lock);
- {
- ctx = __br_stub_get_ongoing_version_ctx (this, inode, NULL);
- if (ctx == NULL)
- goto unblock;
- if (__br_stub_is_inode_dirty (ctx)) {
- __br_stub_set_ongoing_version (ctx, version);
- __br_stub_mark_inode_synced (ctx);
- }
-
- ret = 0;
+ LOCK(&inode->lock);
+ {
+ ctx = __br_stub_get_ongoing_version_ctx(this, inode, NULL);
+ if (ctx == NULL)
+ goto unblock;
+ if (__br_stub_is_inode_dirty(ctx)) {
+ __br_stub_set_ongoing_version(ctx, version);
+ __br_stub_mark_inode_synced(ctx);
}
+
+ ret = 0;
+ }
unblock:
- UNLOCK (&inode->lock);
+ UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
static void
-br_stub_fill_local (br_stub_local_t *local,
- call_stub_t *stub, fd_t *fd, inode_t *inode, uuid_t gfid,
- int versioningtype, unsigned long memversion)
+br_stub_fill_local(br_stub_local_t *local, call_stub_t *stub, fd_t *fd,
+ inode_t *inode, uuid_t gfid, int versioningtype,
+ unsigned long memversion)
{
- local->fopstub = stub;
- local->versioningtype = versioningtype;
- local->u.context.version = memversion;
- if (fd)
- local->u.context.fd = fd_ref (fd);
- if (inode)
- local->u.context.inode = inode_ref (inode);
- gf_uuid_copy (local->u.context.gfid, gfid);
+ local->fopstub = stub;
+ local->versioningtype = versioningtype;
+ local->u.context.version = memversion;
+ if (fd)
+ local->u.context.fd = fd_ref(fd);
+ if (inode)
+ local->u.context.inode = inode_ref(inode);
+ gf_uuid_copy(local->u.context.gfid, gfid);
}
static void
-br_stub_cleanup_local (br_stub_local_t *local)
+br_stub_cleanup_local(br_stub_local_t *local)
{
- if (!local)
- return;
+ if (!local)
+ return;
- local->fopstub = NULL;
- local->versioningtype = 0;
- local->u.context.version = 0;
- if (local->u.context.fd) {
- fd_unref (local->u.context.fd);
- local->u.context.fd = NULL;
- }
- if (local->u.context.inode) {
- inode_unref (local->u.context.inode);
- local->u.context.inode = NULL;
- }
- memset (local->u.context.gfid, '\0', sizeof (uuid_t));
+ local->fopstub = NULL;
+ local->versioningtype = 0;
+ local->u.context.version = 0;
+ if (local->u.context.fd) {
+ fd_unref(local->u.context.fd);
+ local->u.context.fd = NULL;
+ }
+ if (local->u.context.inode) {
+ inode_unref(local->u.context.inode);
+ local->u.context.inode = NULL;
+ }
+ memset(local->u.context.gfid, '\0', sizeof(uuid_t));
}
static int
-br_stub_need_versioning (xlator_t *this,
- fd_t *fd, gf_boolean_t *versioning,
- gf_boolean_t *modified, br_stub_inode_ctx_t **ctx)
-{
- int32_t ret = -1;
- uint64_t ctx_addr = 0;
- br_stub_inode_ctx_t *c = NULL;
- unsigned long version = BITROT_DEFAULT_CURRENT_VERSION;
-
- *versioning = _gf_false;
- *modified = _gf_false;
-
- /* Bitrot stub inode context was initialized only in lookup, create
- * and mknod cbk path. Object versioning was enabled by default
- * irrespective of bitrot enbaled or not. But it's made optional now.
- * As a consequence there could be cases where getting inode ctx would
- * fail because it's not set yet.
- * e.g., If versioning (with bitrot enable) is enabled while I/O is
- * happening, it could directly get other fops like writev without
- * lookup, where getting inode ctx would fail. Hence initialize the
- * inode ctx on failure to get ctx. This is done in all places where
- * applicable.
- */
- ret = br_stub_get_inode_ctx (this, fd->inode, &ctx_addr);
- if (ret < 0) {
- ret = br_stub_init_inode_versions (this, fd, fd->inode, version,
- _gf_true, _gf_false,
- &ctx_addr);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_GET_INODE_CONTEXT_FAILED, "failed to "
- " init the inode context for the inode %s",
- uuid_utoa (fd->inode->gfid));
- goto error_return;
- }
+br_stub_need_versioning(xlator_t *this, fd_t *fd, gf_boolean_t *versioning,
+ gf_boolean_t *modified, br_stub_inode_ctx_t **ctx)
+{
+ int32_t ret = -1;
+ uint64_t ctx_addr = 0;
+ br_stub_inode_ctx_t *c = NULL;
+ unsigned long version = BITROT_DEFAULT_CURRENT_VERSION;
+
+ *versioning = _gf_false;
+ *modified = _gf_false;
+
+ /* Bitrot stub inode context was initialized only in lookup, create
+ * and mknod cbk path. Object versioning was enabled by default
+ * irrespective of bitrot enabled or not. But it's made optional now.
+ * As a consequence there could be cases where getting inode ctx would
+ * fail because it's not set yet.
+ * e.g., If versioning (with bitrot enable) is enabled while I/O is
+ * happening, it could directly get other fops like writev without
+ * lookup, where getting inode ctx would fail. Hence initialize the
+ * inode ctx on failure to get ctx. This is done in all places where
+ * applicable.
+ */
+ ret = br_stub_get_inode_ctx(this, fd->inode, &ctx_addr);
+ if (ret < 0) {
+ ret = br_stub_init_inode_versions(this, fd, fd->inode, version,
+ _gf_true, _gf_false, &ctx_addr);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ BRS_MSG_GET_INODE_CONTEXT_FAILED, "gfid=%s",
+ uuid_utoa(fd->inode->gfid), NULL);
+ goto error_return;
}
+ }
- c = (br_stub_inode_ctx_t *) (long) ctx_addr;
+ c = (br_stub_inode_ctx_t *)(long)ctx_addr;
- LOCK (&fd->inode->lock);
- {
- if (__br_stub_is_inode_dirty (c))
- *versioning = _gf_true;
- if (__br_stub_is_inode_modified (c))
- *modified = _gf_true;
- }
- UNLOCK (&fd->inode->lock);
+ LOCK(&fd->inode->lock);
+ {
+ if (__br_stub_is_inode_dirty(c))
+ *versioning = _gf_true;
+ if (__br_stub_is_inode_modified(c))
+ *modified = _gf_true;
+ }
+ UNLOCK(&fd->inode->lock);
- if (ctx)
- *ctx = c;
- return 0;
+ if (ctx)
+ *ctx = c;
+ return 0;
- error_return:
- return -1;
+error_return:
+ return -1;
}
static int32_t
-br_stub_anon_fd_ctx (xlator_t *this, fd_t *fd, br_stub_inode_ctx_t *ctx)
-{
- int32_t ret = -1;
- br_stub_fd_t *br_stub_fd = NULL;
-
- br_stub_fd = br_stub_fd_ctx_get (this, fd);
- if (!br_stub_fd) {
- ret = br_stub_add_fd_to_inode (this, fd, ctx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_ADD_FD_TO_INODE, "failed to add fd to "
- "the inode (gfid: %s)",
- uuid_utoa (fd->inode->gfid));
- goto out;
- }
+br_stub_anon_fd_ctx(xlator_t *this, fd_t *fd, br_stub_inode_ctx_t *ctx)
+{
+ int32_t ret = -1;
+ br_stub_fd_t *br_stub_fd = NULL;
+
+ br_stub_fd = br_stub_fd_ctx_get(this, fd);
+ if (!br_stub_fd) {
+ ret = br_stub_add_fd_to_inode(this, fd, ctx);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_ADD_FD_TO_INODE,
+ "gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto out;
}
+ }
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
static int
-br_stub_versioning_prep (call_frame_t *frame,
- xlator_t *this, fd_t *fd, br_stub_inode_ctx_t *ctx)
+br_stub_versioning_prep(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ br_stub_inode_ctx_t *ctx)
{
- int32_t ret = -1;
- br_stub_local_t *local = NULL;
+ int32_t ret = -1;
+ br_stub_local_t *local = NULL;
- local = br_stub_alloc_local (this);
- if (!local) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM, BRS_MSG_NO_MEMORY,
- "local allocation failed (gfid: %s)",
- uuid_utoa (fd->inode->gfid));
- goto error_return;
- }
+ local = br_stub_alloc_local(this);
+ if (!local) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, BRS_MSG_NO_MEMORY, "gfid=%s",
+ uuid_utoa(fd->inode->gfid), NULL);
+ goto error_return;
+ }
- if (fd_is_anonymous (fd)) {
- ret = br_stub_anon_fd_ctx (this, fd, ctx);
- if (ret)
- goto free_local;
- }
+ if (fd_is_anonymous(fd)) {
+ ret = br_stub_anon_fd_ctx(this, fd, ctx);
+ if (ret)
+ goto free_local;
+ }
- frame->local = local;
+ frame->local = local;
- return 0;
+ return 0;
- free_local:
- br_stub_dealloc_local (local);
- error_return:
- return -1;
+free_local:
+ br_stub_dealloc_local(local);
+error_return:
+ return -1;
}
static int
-br_stub_mark_inode_modified (xlator_t *this, br_stub_local_t *local)
-{
- fd_t *fd = NULL;
- int32_t ret = 0;
- uint64_t ctx_addr = 0;
- br_stub_inode_ctx_t *ctx = NULL;
- unsigned long version = BITROT_DEFAULT_CURRENT_VERSION;
-
- fd = local->u.context.fd;
-
- ret = br_stub_get_inode_ctx (this, fd->inode, &ctx_addr);
- if (ret < 0) {
- ret = br_stub_init_inode_versions (this, fd, fd->inode, version,
- _gf_true, _gf_false,
- &ctx_addr);
- if (ret)
- goto error_return;
- }
+br_stub_mark_inode_modified(xlator_t *this, br_stub_local_t *local)
+{
+ fd_t *fd = NULL;
+ int32_t ret = 0;
+ uint64_t ctx_addr = 0;
+ br_stub_inode_ctx_t *ctx = NULL;
+ unsigned long version = BITROT_DEFAULT_CURRENT_VERSION;
- ctx = (br_stub_inode_ctx_t *) (long) ctx_addr;
+ fd = local->u.context.fd;
- LOCK (&fd->inode->lock);
- {
- __br_stub_set_inode_modified (ctx);
- }
- UNLOCK (&fd->inode->lock);
+ ret = br_stub_get_inode_ctx(this, fd->inode, &ctx_addr);
+ if (ret < 0) {
+ ret = br_stub_init_inode_versions(this, fd, fd->inode, version,
+ _gf_true, _gf_false, &ctx_addr);
+ if (ret)
+ goto error_return;
+ }
- return 0;
+ ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
- error_return:
- return -1;
+ LOCK(&fd->inode->lock);
+ {
+ __br_stub_set_inode_modified(ctx);
+ }
+ UNLOCK(&fd->inode->lock);
+
+ return 0;
+
+error_return:
+ return -1;
}
/**
@@ -645,71 +696,68 @@ br_stub_mark_inode_modified (xlator_t *this, br_stub_local_t *local)
* and error is returned upwards.
*/
static int
-br_stub_check_bad_object (xlator_t *this, inode_t *inode, int32_t *op_ret,
- int32_t *op_errno)
-{
- int ret = -1;
- unsigned long version = BITROT_DEFAULT_CURRENT_VERSION;
-
- ret = br_stub_is_bad_object (this, inode);
- if (ret == -2) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRS_MSG_BAD_OBJECT_ACCESS,
- "%s is a bad object. Returning",
- uuid_utoa (inode->gfid));
- *op_ret = -1;
- *op_errno = EIO;
- }
-
- if (ret == -1) {
- ret = br_stub_init_inode_versions (this, NULL, inode, version,
- _gf_true, _gf_false, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_GET_INODE_CONTEXT_FAILED,
- "failed to init inode context for %s",
- uuid_utoa (inode->gfid));
- *op_ret = -1;
- *op_errno = EINVAL;
- }
+br_stub_check_bad_object(xlator_t *this, inode_t *inode, int32_t *op_ret,
+ int32_t *op_errno)
+{
+ int ret = -1;
+ unsigned long version = BITROT_DEFAULT_CURRENT_VERSION;
+
+ ret = br_stub_is_bad_object(this, inode);
+ if (ret == -2) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_BAD_OBJECT_ACCESS,
+ "gfid=%s", uuid_utoa(inode->gfid), NULL);
+ *op_ret = -1;
+ *op_errno = EIO;
+ }
+
+ if (ret == -1) {
+ ret = br_stub_init_inode_versions(this, NULL, inode, version, _gf_true,
+ _gf_false, NULL);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ BRS_MSG_GET_INODE_CONTEXT_FAILED, "gfid=%s",
+ uuid_utoa(inode->gfid), NULL);
+ *op_ret = -1;
+ *op_errno = EINVAL;
}
+ }
- return ret;
+ return ret;
}
/**
* callback for inode/fd versioning
*/
int
-br_stub_fd_incversioning_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
-{
- fd_t *fd = NULL;
- inode_t *inode = NULL;
- unsigned long version = 0;
- br_stub_local_t *local = NULL;
-
- local = (br_stub_local_t *)frame->local;
- if (op_ret < 0)
- goto done;
- fd = local->u.context.fd;
- inode = local->u.context.inode;
- version = local->u.context.version;
-
- op_ret = br_stub_mod_inode_versions (this, fd, inode, version);
- if (op_ret < 0)
- op_errno = EINVAL;
-
- done:
- if (op_ret < 0) {
- frame->local = NULL;
- call_unwind_error (local->fopstub, -1, op_errno);
- br_stub_cleanup_local (local);
- br_stub_dealloc_local (local);
- } else {
- call_resume (local->fopstub);
- }
- return 0;
+br_stub_fd_incversioning_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
+{
+ fd_t *fd = NULL;
+ inode_t *inode = NULL;
+ unsigned long version = 0;
+ br_stub_local_t *local = NULL;
+
+ local = (br_stub_local_t *)frame->local;
+ if (op_ret < 0)
+ goto done;
+ fd = local->u.context.fd;
+ inode = local->u.context.inode;
+ version = local->u.context.version;
+
+ op_ret = br_stub_mod_inode_versions(this, fd, inode, version);
+ if (op_ret < 0)
+ op_errno = EINVAL;
+
+done:
+ if (op_ret < 0) {
+ frame->local = NULL;
+ call_unwind_error(local->fopstub, -1, op_errno);
+ br_stub_cleanup_local(local);
+ br_stub_dealloc_local(local);
+ } else {
+ call_resume(local->fopstub);
+ }
+ return 0;
}
/**
@@ -742,102 +790,101 @@ br_stub_fd_incversioning_cbk (call_frame_t *frame,
/**
* perform full or incremental versioning on an inode pointd by an
* fd. incremental versioning is done when an inode is dirty and a
- * writeback is trigerred.
+ * writeback is triggered.
*/
int
-br_stub_fd_versioning (xlator_t *this, call_frame_t *frame,
- call_stub_t *stub, dict_t *dict, fd_t *fd,
- br_stub_version_cbk *callback, unsigned long memversion,
- int versioningtype, int durable)
+br_stub_fd_versioning(xlator_t *this, call_frame_t *frame, call_stub_t *stub,
+ dict_t *dict, fd_t *fd, br_stub_version_cbk *callback,
+ unsigned long memversion, int versioningtype, int durable)
{
- int32_t ret = -1;
- int flags = 0;
- dict_t *xdata = NULL;
- br_stub_local_t *local = NULL;
+ int32_t ret = -1;
+ int flags = 0;
+ dict_t *xdata = NULL;
+ br_stub_local_t *local = NULL;
- xdata = dict_new ();
- if (!xdata)
- goto done;
+ xdata = dict_new();
+ if (!xdata)
+ goto done;
- ret = dict_set_int32 (xdata, GLUSTERFS_INTERNAL_FOP_KEY, 1);
- if (ret)
- goto dealloc_xdata;
+ ret = dict_set_int32(xdata, GLUSTERFS_INTERNAL_FOP_KEY, 1);
+ if (ret)
+ goto dealloc_xdata;
- if (durable) {
- ret = dict_set_int32 (xdata, GLUSTERFS_DURABLE_OP, 0);
- if (ret)
- goto dealloc_xdata;
- }
+ if (durable) {
+ ret = dict_set_int32(xdata, GLUSTERFS_DURABLE_OP, 0);
+ if (ret)
+ goto dealloc_xdata;
+ }
- local = frame->local;
+ local = frame->local;
- br_stub_fill_local (local, stub, fd,
- fd->inode, fd->inode->gfid,
- versioningtype, memversion);
+ br_stub_fill_local(local, stub, fd, fd->inode, fd->inode->gfid,
+ versioningtype, memversion);
- STACK_WIND (frame, callback,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->fsetxattr,
- fd, dict, flags, xdata);
+ STACK_WIND(frame, callback, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
- ret = 0;
+ ret = 0;
- dealloc_xdata:
- dict_unref (xdata);
- done:
- return ret;
+dealloc_xdata:
+ dict_unref(xdata);
+done:
+ return ret;
}
static int
-br_stub_perform_incversioning (xlator_t *this,
- call_frame_t *frame, call_stub_t *stub,
- fd_t *fd, br_stub_inode_ctx_t *ctx)
-{
- int32_t ret = -1;
- dict_t *dict = NULL;
- br_version_t *obuf = NULL;
- unsigned long writeback_version = 0;
- int op_errno = 0;
- br_stub_local_t *local = NULL;
-
- op_errno = EINVAL;
- local = frame->local;
-
- writeback_version = __br_stub_writeback_version (ctx);
-
- op_errno = ENOMEM;
- dict = dict_new ();
- if (!dict)
- goto done;
- ret = br_stub_alloc_versions (&obuf, NULL, 0);
- if (ret)
- goto dealloc_dict;
- ret = br_stub_prepare_version_request (this, dict,
- obuf, writeback_version);
- if (ret)
- goto dealloc_versions;
-
- ret = br_stub_fd_versioning
- (this, frame, stub, dict,
- fd, br_stub_fd_incversioning_cbk, writeback_version,
- BR_STUB_INCREMENTAL_VERSIONING, !WRITEBACK_DURABLE);
-
- dealloc_versions:
- br_stub_dealloc_versions (obuf);
- dealloc_dict:
- dict_unref (dict);
- done:
- if (ret) {
- if (local)
- frame->local = NULL;
- call_unwind_error (stub, -1, op_errno);
- if (local) {
- br_stub_cleanup_local (local);
- br_stub_dealloc_local (local);
- }
+br_stub_perform_incversioning(xlator_t *this, call_frame_t *frame,
+ call_stub_t *stub, fd_t *fd,
+ br_stub_inode_ctx_t *ctx)
+{
+ int32_t ret = -1;
+ dict_t *dict = NULL;
+ br_version_t *obuf = NULL;
+ unsigned long writeback_version = 0;
+ int op_errno = 0;
+ br_stub_local_t *local = NULL;
+
+ op_errno = EINVAL;
+ local = frame->local;
+
+ writeback_version = __br_stub_writeback_version(ctx);
+
+ op_errno = ENOMEM;
+ dict = dict_new();
+ if (!dict)
+ goto out;
+ ret = br_stub_alloc_versions(&obuf, NULL, 0);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_ALLOC_MEM_FAILED,
+ "gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto out;
+ }
+ ret = br_stub_prepare_version_request(this, dict, obuf, writeback_version);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_VERSION_PREPARE_FAIL,
+ "gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ br_stub_dealloc_versions(obuf);
+ goto out;
+ }
+
+ ret = br_stub_fd_versioning(
+ this, frame, stub, dict, fd, br_stub_fd_incversioning_cbk,
+ writeback_version, BR_STUB_INCREMENTAL_VERSIONING, !WRITEBACK_DURABLE);
+out:
+ if (dict)
+ dict_unref(dict);
+ if (ret) {
+ if (local)
+ frame->local = NULL;
+ call_unwind_error(stub, -1, op_errno);
+ if (local) {
+ br_stub_cleanup_local(local);
+ br_stub_dealloc_local(local);
}
+ }
- return ret;
+ return ret;
}
/** {{{ */
@@ -845,247 +892,271 @@ br_stub_perform_incversioning (xlator_t *this,
/* fsetxattr() */
int32_t
-br_stub_perform_objsign (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *dict, int flags, dict_t *xdata)
+br_stub_perform_objsign(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ dict_t *dict, int flags, dict_t *xdata)
{
- STACK_WIND (frame, default_fsetxattr_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->fsetxattr, fd,
- dict, flags, xdata);
+ STACK_WIND(frame, default_fsetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
- dict_unref (xdata);
- return 0;
+ dict_unref(xdata);
+ return 0;
}
void *
-br_stub_signth (void *arg)
+br_stub_signth(void *arg)
{
- xlator_t *this = arg;
- br_stub_private_t *priv = this->private;
- struct br_stub_signentry *sigstub = NULL;
+ xlator_t *this = arg;
+ br_stub_private_t *priv = this->private;
+ struct br_stub_signentry *sigstub = NULL;
- THIS = this;
- while (1) {
-
- pthread_mutex_lock (&priv->lock);
- {
- while (list_empty (&priv->squeue))
- pthread_cond_wait (&priv->cond, &priv->lock);
+ THIS = this;
+ while (1) {
+ /*
+ * Disabling bit-rot feature leads to this particular thread
+ * getting cleaned up by reconfigure via a call to the function
+ * gf_thread_cleanup_xint (which in turn calls pthread_cancel
+ * and pthread_join). But, if this thread had held the mutex
+ * &priv->lock at the time of cancellation, then it leads to
+ * deadlock in future when bit-rot feature is enabled (which
+ * again spawns this thread which cant hold the lock as the
+ * mutex is still held by the previous instance of the thread
+ * which got killed). Also, the br_stub_handle_object_signature
+ * function which is called whenever file has to be signed
+ * also gets blocked as it too attempts to acquire &priv->lock.
+ *
+ * So, arrange for the lock to be unlocked as part of the
+ * cleanup of this thread using pthread_cleanup_push and
+ * pthread_cleanup_pop.
+ */
+ pthread_cleanup_push(br_stub_lock_cleaner, &priv->lock);
+ pthread_mutex_lock(&priv->lock);
+ {
+ while (list_empty(&priv->squeue))
+ pthread_cond_wait(&priv->cond, &priv->lock);
- sigstub = list_first_entry
- (&priv->squeue, struct br_stub_signentry, list);
- list_del_init (&sigstub->list);
- }
- pthread_mutex_unlock (&priv->lock);
+ sigstub = list_first_entry(&priv->squeue, struct br_stub_signentry,
+ list);
+ list_del_init(&sigstub->list);
+ }
+ pthread_mutex_unlock(&priv->lock);
+ pthread_cleanup_pop(0);
- call_resume (sigstub->stub);
+ call_resume(sigstub->stub);
- GF_FREE (sigstub);
- }
+ GF_FREE(sigstub);
+ }
- return NULL;
+ return NULL;
}
static gf_boolean_t
-br_stub_internal_xattr (dict_t *dict) {
-
- if (dict_get (dict, GLUSTERFS_SET_OBJECT_SIGNATURE) ||
- dict_get (dict, GLUSTERFS_GET_OBJECT_SIGNATURE) ||
- dict_get (dict, BR_REOPEN_SIGN_HINT_KEY) ||
- dict_get (dict, BITROT_OBJECT_BAD_KEY) ||
- dict_get (dict, BITROT_SIGNING_VERSION_KEY) ||
- dict_get (dict, BITROT_CURRENT_VERSION_KEY))
- return _gf_true;
+br_stub_internal_xattr(dict_t *dict)
+{
+ if (dict_get(dict, GLUSTERFS_SET_OBJECT_SIGNATURE) ||
+ dict_get(dict, GLUSTERFS_GET_OBJECT_SIGNATURE) ||
+ dict_get(dict, BR_REOPEN_SIGN_HINT_KEY) ||
+ dict_get(dict, BITROT_OBJECT_BAD_KEY) ||
+ dict_get(dict, BITROT_SIGNING_VERSION_KEY) ||
+ dict_get(dict, BITROT_CURRENT_VERSION_KEY))
+ return _gf_true;
- return _gf_false;
+ return _gf_false;
}
int
-orderq (struct list_head *elem1, struct list_head *elem2)
+orderq(struct list_head *elem1, struct list_head *elem2)
{
- struct br_stub_signentry *s1 = NULL;
- struct br_stub_signentry *s2 = NULL;
+ struct br_stub_signentry *s1 = NULL;
+ struct br_stub_signentry *s2 = NULL;
- s1 = list_entry (elem1, struct br_stub_signentry, list);
- s2 = list_entry (elem2, struct br_stub_signentry, list);
+ s1 = list_entry(elem1, struct br_stub_signentry, list);
+ s2 = list_entry(elem2, struct br_stub_signentry, list);
- return (s1->v > s2->v);
+ return (s1->v > s2->v);
}
static int
-br_stub_compare_sign_version (xlator_t *this,
- inode_t *inode,
- br_signature_t *sbuf,
- dict_t *dict, int *fakesuccess)
-{
- int32_t ret = -1;
- uint64_t tmp_ctx = 0;
- gf_boolean_t invalid = _gf_false;
- br_stub_inode_ctx_t *ctx = NULL;
-
- GF_VALIDATE_OR_GOTO ("bit-rot-stub", this, out);
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
- GF_VALIDATE_OR_GOTO (this->name, sbuf, out);
- GF_VALIDATE_OR_GOTO (this->name, dict, out);
-
- ret = br_stub_get_inode_ctx (this, inode, &tmp_ctx);
- if (ret) {
- dict_del (dict, BITROT_SIGNING_VERSION_KEY);
- goto out;
- }
-
- ctx = (br_stub_inode_ctx_t *)(long)tmp_ctx;
-
- LOCK (&inode->lock);
- {
- if (ctx->currentversion < sbuf->signedversion) {
- invalid = _gf_true;
- } else if (ctx->currentversion > sbuf->signedversion) {
- gf_msg_debug (this->name, 0, "\"Signing version\" "
- "(%lu) lower than \"Current version \" "
- "(%lu)", ctx->currentversion,
- sbuf->signedversion);
- *fakesuccess = 1;
- }
- }
- UNLOCK (&inode->lock);
-
- if (invalid) {
- ret = -1;
- gf_msg (this->name, GF_LOG_WARNING, 0,
- BRS_MSG_SIGN_VERSION_ERROR, "Signing version exceeds "
- "current version [%lu > %lu]", sbuf->signedversion,
- ctx->currentversion);
- }
+br_stub_compare_sign_version(xlator_t *this, inode_t *inode,
+ br_signature_t *sbuf, dict_t *dict,
+ int *fakesuccess)
+{
+ int32_t ret = -1;
+ uint64_t tmp_ctx = 0;
+ gf_boolean_t invalid = _gf_false;
+ br_stub_inode_ctx_t *ctx = NULL;
+
+ GF_VALIDATE_OR_GOTO("bit-rot-stub", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, sbuf, out);
+ GF_VALIDATE_OR_GOTO(this->name, dict, out);
+
+ ret = br_stub_get_inode_ctx(this, inode, &tmp_ctx);
+ if (ret) {
+ dict_del(dict, BITROT_SIGNING_VERSION_KEY);
+ goto out;
+ }
+
+ ctx = (br_stub_inode_ctx_t *)(long)tmp_ctx;
+
+ LOCK(&inode->lock);
+ {
+ if (ctx->currentversion < sbuf->signedversion) {
+ invalid = _gf_true;
+ } else if (ctx->currentversion > sbuf->signedversion) {
+ gf_msg_debug(this->name, 0,
+ "\"Signing version\" "
+ "(%lu) lower than \"Current version \" "
+ "(%lu)",
+ ctx->currentversion, sbuf->signedversion);
+ *fakesuccess = 1;
+ }
+ }
+ UNLOCK(&inode->lock);
+
+ if (invalid) {
+ ret = -1;
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_SIGN_VERSION_ERROR,
+ "Signing-ver=%lu", sbuf->signedversion, "current-ver=%lu",
+ ctx->currentversion, NULL);
+ }
- out:
- return ret;
+out:
+ return ret;
}
static int
-br_stub_prepare_signature (xlator_t *this,
- dict_t *dict, inode_t *inode,
- br_isignature_t *sign, int *fakesuccess)
+br_stub_prepare_signature(xlator_t *this, dict_t *dict, inode_t *inode,
+ br_isignature_t *sign, int *fakesuccess)
{
- int32_t ret = 0;
- size_t signaturelen = 0;
- br_signature_t *sbuf = NULL;
-
- if (!br_is_signature_type_valid (sign->signaturetype))
- goto error_return;
+ int32_t ret = -1;
+ size_t signaturelen = 0;
+ br_signature_t *sbuf = NULL;
- signaturelen = sign->signaturelen;
- ret = br_stub_alloc_versions (NULL, &sbuf, signaturelen);
- if (ret)
- goto error_return;
- ret = br_stub_prepare_signing_request (dict, sbuf, sign, signaturelen);
- if (ret)
- goto dealloc_versions;
+ if (!br_is_signature_type_valid(sign->signaturetype))
+ goto out;
- ret = br_stub_compare_sign_version (this, inode,
- sbuf, dict, fakesuccess);
- if (ret)
- goto dealloc_versions;
-
- return 0;
-
- dealloc_versions:
- br_stub_dealloc_versions (sbuf);
- error_return:
- return -1;
+ signaturelen = sign->signaturelen;
+ ret = br_stub_alloc_versions(NULL, &sbuf, signaturelen);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_ALLOC_MEM_FAILED,
+ "gfid=%s", uuid_utoa(inode->gfid), NULL);
+ ret = -1;
+ goto out;
+ }
+ ret = br_stub_prepare_signing_request(dict, sbuf, sign, signaturelen);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_SIGN_PREPARE_FAIL,
+ "gfid=%s", uuid_utoa(inode->gfid), NULL);
+ ret = -1;
+ br_stub_dealloc_versions(sbuf);
+ goto out;
+ }
+
+ /* At this point sbuf has been added to dict, so the memory will be freed
+ * when the data from the dict is destroyed
+ */
+ ret = br_stub_compare_sign_version(this, inode, sbuf, dict, fakesuccess);
+out:
+ return ret;
}
static void
-br_stub_handle_object_signature (call_frame_t *frame,
- xlator_t *this, fd_t *fd, dict_t *dict,
- br_isignature_t *sign, dict_t *xdata)
-{
- int32_t ret = -1;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- int fakesuccess = 0;
- br_stub_private_t *priv = NULL;
- struct br_stub_signentry *sigstub = NULL;
-
- priv = this->private;
-
- if (frame->root->pid != GF_CLIENT_PID_BITD)
- goto dofop;
-
- ret = br_stub_prepare_signature (this, dict,
- fd->inode, sign, &fakesuccess);
- if (ret)
- goto dofop;
- if (fakesuccess) {
- op_ret = op_errno = 0;
- goto dofop;
- }
-
- dict_del (dict, GLUSTERFS_SET_OBJECT_SIGNATURE);
-
- ret = -1;
- if (!xdata) {
- xdata = dict_new ();
- if (!xdata)
- goto dofop;
- } else {
- dict_ref (xdata);
- }
+br_stub_handle_object_signature(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ dict_t *dict, br_isignature_t *sign,
+ dict_t *xdata)
+{
+ int32_t ret = -1;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ int fakesuccess = 0;
+ br_stub_private_t *priv = NULL;
+ struct br_stub_signentry *sigstub = NULL;
+
+ priv = this->private;
+
+ if (frame->root->pid != GF_CLIENT_PID_BITD) {
+ gf_smsg(this->name, GF_LOG_WARNING, op_errno, BRS_MSG_NON_BITD_PID,
+ "PID=%d", frame->root->pid, NULL);
+ goto dofop;
+ }
+
+ ret = br_stub_prepare_signature(this, dict, fd->inode, sign, &fakesuccess);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_SIGN_PREPARE_FAIL,
+ "gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto dofop;
+ }
+ if (fakesuccess) {
+ op_ret = op_errno = 0;
+ goto dofop;
+ }
+
+ dict_del(dict, GLUSTERFS_SET_OBJECT_SIGNATURE);
+
+ ret = -1;
+ if (!xdata) {
+ xdata = dict_new();
+ if (!xdata)
+ goto dofop;
+ } else {
+ dict_ref(xdata);
+ }
- ret = dict_set_int32 (xdata, GLUSTERFS_DURABLE_OP, 0);
- if (ret)
- goto unref_dict;
+ ret = dict_set_int32(xdata, GLUSTERFS_DURABLE_OP, 0);
+ if (ret)
+ goto unref_dict;
- /* prepare dispatch stub to order object signing */
- sigstub = GF_CALLOC (1, sizeof (*sigstub), gf_br_stub_mt_sigstub_t);
- if (!sigstub)
- goto unref_dict;
+ /* prepare dispatch stub to order object signing */
+ sigstub = GF_CALLOC(1, sizeof(*sigstub), gf_br_stub_mt_sigstub_t);
+ if (!sigstub)
+ goto unref_dict;
- INIT_LIST_HEAD (&sigstub->list);
- sigstub->v = ntohl (sign->signedversion);
- sigstub->stub = fop_fsetxattr_stub (frame, br_stub_perform_objsign,
- fd, dict, 0, xdata);
- if (!sigstub->stub)
- goto cleanup_stub;
+ INIT_LIST_HEAD(&sigstub->list);
+ sigstub->v = ntohl(sign->signedversion);
+ sigstub->stub = fop_fsetxattr_stub(frame, br_stub_perform_objsign, fd, dict,
+ 0, xdata);
+ if (!sigstub->stub)
+ goto cleanup_stub;
- pthread_mutex_lock (&priv->lock);
- {
- list_add_order (&sigstub->list, &priv->squeue, orderq);
- pthread_cond_signal (&priv->cond);
- }
- pthread_mutex_unlock (&priv->lock);
+ pthread_mutex_lock(&priv->lock);
+ {
+ list_add_order(&sigstub->list, &priv->squeue, orderq);
+ pthread_cond_signal(&priv->cond);
+ }
+ pthread_mutex_unlock(&priv->lock);
- return;
+ return;
- cleanup_stub:
- GF_FREE (sigstub);
- unref_dict:
- dict_unref (xdata);
- dofop:
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, NULL);
+cleanup_stub:
+ GF_FREE(sigstub);
+unref_dict:
+ dict_unref(xdata);
+dofop:
+ STACK_UNWIND_STRICT(fsetxattr, frame, op_ret, op_errno, NULL);
}
int32_t
-br_stub_fsetxattr_resume (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+br_stub_fsetxattr_resume(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- int32_t ret = -1;
- br_stub_local_t *local = NULL;
+ int32_t ret = -1;
+ br_stub_local_t *local = NULL;
- local = frame->local;
- frame->local = NULL;
+ local = frame->local;
+ frame->local = NULL;
- ret = br_stub_mark_inode_modified (this, local);
- if (ret) {
- op_ret = -1;
- op_errno = EINVAL;
- }
+ ret = br_stub_mark_inode_modified(this, local);
+ if (ret) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ }
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
+ STACK_UNWIND_STRICT(fsetxattr, frame, op_ret, op_errno, xdata);
- br_stub_cleanup_local (local);
- br_stub_dealloc_local (local);
+ br_stub_cleanup_local(local);
+ br_stub_dealloc_local(local);
- return 0;
+ return 0;
}
/**
@@ -1123,65 +1194,62 @@ br_stub_fsetxattr_resume (call_frame_t *frame, void *cookie, xlator_t *this,
* }
*/
static void
-br_stub_handle_object_reopen (call_frame_t *frame,
- xlator_t *this, fd_t *fd, uint32_t val)
-{
- int32_t ret = -1;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- call_stub_t *stub = NULL;
- gf_boolean_t inc_version = _gf_false;
- gf_boolean_t modified = _gf_false;
- br_stub_inode_ctx_t *ctx = NULL;
- br_stub_local_t *local = NULL;
- gf_boolean_t goback = _gf_true;
-
- ret = br_stub_need_versioning (this, fd, &inc_version, &modified, &ctx);
- if (ret)
- goto unwind;
-
- LOCK (&fd->inode->lock);
- {
- if ((val == BR_OBJECT_REOPEN) && inc_version)
- goback = _gf_false;
- if (val == BR_OBJECT_RESIGN &&
- ctx->info_sign == BR_SIGN_NORMAL) {
- __br_stub_mark_inode_synced (ctx);
- __br_stub_set_inode_modified (ctx);
- }
- (void) __br_stub_inode_sign_state (ctx, GF_FOP_FSETXATTR, fd);
- }
- UNLOCK (&fd->inode->lock);
-
- if (goback) {
- op_ret = op_errno = 0;
- goto unwind;
- }
-
- ret = br_stub_versioning_prep (frame, this, fd, ctx);
- if (ret)
- goto unwind;
- local = frame->local;
-
- stub = fop_fsetxattr_cbk_stub (frame, br_stub_fsetxattr_resume,
- 0, 0, NULL);
- if (!stub) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRS_MSG_STUB_ALLOC_FAILED,
- "failed to allocate stub for fsetxattr fop (gfid: %s),"
- " unwinding", uuid_utoa (fd->inode->gfid));
- goto cleanup_local;
- }
-
- (void) br_stub_perform_incversioning (this, frame, stub, fd, ctx);
- return;
+br_stub_handle_object_reopen(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ uint32_t val)
+{
+ int32_t ret = -1;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ call_stub_t *stub = NULL;
+ gf_boolean_t inc_version = _gf_false;
+ gf_boolean_t modified = _gf_false;
+ br_stub_inode_ctx_t *ctx = NULL;
+ br_stub_local_t *local = NULL;
+ gf_boolean_t goback = _gf_true;
+
+ ret = br_stub_need_versioning(this, fd, &inc_version, &modified, &ctx);
+ if (ret)
+ goto unwind;
+
+ LOCK(&fd->inode->lock);
+ {
+ if ((val == BR_OBJECT_REOPEN) && inc_version)
+ goback = _gf_false;
+ if (val == BR_OBJECT_RESIGN && ctx->info_sign == BR_SIGN_NORMAL) {
+ __br_stub_mark_inode_synced(ctx);
+ __br_stub_set_inode_modified(ctx);
+ }
+ (void)__br_stub_inode_sign_state(ctx, GF_FOP_FSETXATTR, fd);
+ }
+ UNLOCK(&fd->inode->lock);
+
+ if (goback) {
+ op_ret = op_errno = 0;
+ goto unwind;
+ }
+
+ ret = br_stub_versioning_prep(frame, this, fd, ctx);
+ if (ret)
+ goto unwind;
+ local = frame->local;
+
+ stub = fop_fsetxattr_cbk_stub(frame, br_stub_fsetxattr_resume, 0, 0, NULL);
+ if (!stub) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_STUB_ALLOC_FAILED,
+ "fsetxattr gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto cleanup_local;
+ }
+
+ (void)br_stub_perform_incversioning(this, frame, stub, fd, ctx);
+ return;
+
+cleanup_local:
+ br_stub_cleanup_local(local);
+ br_stub_dealloc_local(local);
- cleanup_local:
- br_stub_cleanup_local (local);
- br_stub_dealloc_local (local);
-
- unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, NULL);
+unwind:
+ frame->local = NULL;
+ STACK_UNWIND_STRICT(fsetxattr, frame, op_ret, op_errno, NULL);
}
/**
@@ -1192,89 +1260,83 @@ br_stub_handle_object_reopen (call_frame_t *frame,
* to mark the object as bad.
*/
int
-br_stub_fsetxattr_bad_object_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
-{
- br_stub_local_t *local = NULL;
- int32_t ret = -1;
-
- local = frame->local;
- frame->local = NULL;
-
- if (op_ret < 0)
- goto unwind;
-
- /*
- * What to do if marking the object as bad fails? (i.e. in memory
- * marking within the inode context. If we are here means fsetxattr
- * fop has succeeded on disk and the bad object xattr has been set).
- * We can return failure to scruber, but there is nothing the scrubber
- * can do with it (it might assume that the on disk setxattr itself has
- * failed). The main purpose of this operation is to help identify the
- * bad object by checking the inode context itself (thus avoiding the
- * necessity of doing a getxattr fop on the disk).
- *
- * So as of now, success itself is being returned even though inode
- * context set operation fails.
- * In future if there is any change in the policy which can handle this,
- * then appropriate response should be sent (i.e. success or error).
- */
- ret = br_stub_mark_object_bad (this, local->u.context.inode);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0, BRS_MSG_BAD_OBJ_MARK_FAIL,
- "failed to mark object %s as bad",
- uuid_utoa (local->u.context.inode->gfid));
-
- ret = br_stub_add (this, local->u.context.inode->gfid);
+br_stub_fsetxattr_bad_object_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
+{
+ br_stub_local_t *local = NULL;
+ int32_t ret = -1;
+
+ local = frame->local;
+ frame->local = NULL;
+
+ if (op_ret < 0)
+ goto unwind;
+
+ /*
+ * What to do if marking the object as bad fails? (i.e. in memory
+ * marking within the inode context. If we are here means fsetxattr
+ * fop has succeeded on disk and the bad object xattr has been set).
+ * We can return failure to scruber, but there is nothing the scrubber
+ * can do with it (it might assume that the on disk setxattr itself has
+ * failed). The main purpose of this operation is to help identify the
+ * bad object by checking the inode context itself (thus avoiding the
+ * necessity of doing a getxattr fop on the disk).
+ *
+ * So as of now, success itself is being returned even though inode
+ * context set operation fails.
+ * In future if there is any change in the policy which can handle this,
+ * then appropriate response should be sent (i.e. success or error).
+ */
+ ret = br_stub_mark_object_bad(this, local->u.context.inode);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_BAD_OBJ_MARK_FAIL,
+ "gfid=%s", uuid_utoa(local->u.context.inode->gfid), NULL);
+
+ ret = br_stub_add(this, local->u.context.inode->gfid);
unwind:
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
- br_stub_cleanup_local (local);
- br_stub_dealloc_local (local);
- return 0;
+ STACK_UNWIND_STRICT(fsetxattr, frame, op_ret, op_errno, xdata);
+ br_stub_cleanup_local(local);
+ br_stub_dealloc_local(local);
+ return 0;
}
static int32_t
-br_stub_handle_bad_object_key (call_frame_t *frame, xlator_t *this, fd_t *fd,
- dict_t *dict, int flags, dict_t *xdata)
-{
- br_stub_local_t *local = NULL;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
-
- if (frame->root->pid != GF_CLIENT_PID_SCRUB) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_NON_SCRUB_BAD_OBJ_MARK, "bad object marking "
- "on %s is not from the scrubber",
- uuid_utoa (fd->inode->gfid));
- goto unwind;
- }
-
- local = br_stub_alloc_local (this);
- if (!local) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRS_MSG_NO_MEMORY,
- "failed to allocate memory for fsetxattr on %s",
- uuid_utoa (fd->inode->gfid));
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
+br_stub_handle_bad_object_key(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ dict_t *dict, int flags, dict_t *xdata)
+{
+ br_stub_local_t *local = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+
+ if (frame->root->pid != GF_CLIENT_PID_SCRUB) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_NON_SCRUB_BAD_OBJ_MARK,
+ "gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto unwind;
+ }
+
+ local = br_stub_alloc_local(this);
+ if (!local) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_ALLOC_MEM_FAILED,
+ "fsetxattr gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
- br_stub_fill_local (local, NULL, fd, fd->inode,
- fd->inode->gfid, BR_STUB_NO_VERSIONING, 0);
- frame->local = local;
+ br_stub_fill_local(local, NULL, fd, fd->inode, fd->inode->gfid,
+ BR_STUB_NO_VERSIONING, 0);
+ frame->local = local;
- STACK_WIND (frame, br_stub_fsetxattr_bad_object_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fsetxattr, fd, dict, flags,
- xdata);
- return 0;
+ STACK_WIND(frame, br_stub_fsetxattr_bad_object_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+ return 0;
unwind:
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, NULL);
- return 0;
+ STACK_UNWIND_STRICT(fsetxattr, frame, op_ret, op_errno, NULL);
+ return 0;
}
-
/**
* As of now, versioning is done by the stub (though as a setxattr
* operation) as part of inode modification operations such as writev,
@@ -1290,126 +1352,121 @@ unwind:
*
*/
static int32_t
-br_stub_handle_internal_xattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- char *key)
+br_stub_handle_internal_xattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ char *key)
{
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_SET_INTERNAL_XATTR, "setxattr called"
- " on the internal xattr %s for inode %s", key,
- uuid_utoa (fd->inode->gfid));
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_SET_INTERNAL_XATTR,
+ "setxattr key=%s", key, "inode-gfid=%s", uuid_utoa(fd->inode->gfid),
+ NULL);
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, NULL);
- return 0;
+ STACK_UNWIND_STRICT(fsetxattr, frame, op_ret, op_errno, NULL);
+ return 0;
}
static void
-br_stub_dump_xattr (xlator_t *this, dict_t *dict, int *op_errno)
-{
- char *format = "(%s:%s)";
- char *dump = NULL;
-
- dump = GF_CALLOC (1, BR_STUB_DUMP_STR_SIZE, gf_br_stub_mt_misc);
- if (!dump) {
- *op_errno = ENOMEM;
- goto out;
- }
- dict_dump_to_str (dict, dump, BR_STUB_DUMP_STR_SIZE, format);
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_SET_INTERNAL_XATTR, "fsetxattr called on "
- "internal xattr %s", dump);
- out:
- if (dump) {
- GF_FREE (dump);
- }
- return;
+br_stub_dump_xattr(xlator_t *this, dict_t *dict, int *op_errno)
+{
+ char *format = "(%s:%s)";
+ char *dump = NULL;
+
+ dump = GF_CALLOC(1, BR_STUB_DUMP_STR_SIZE, gf_br_stub_mt_misc);
+ if (!dump) {
+ *op_errno = ENOMEM;
+ goto out;
+ }
+ dict_dump_to_str(dict, dump, BR_STUB_DUMP_STR_SIZE, format);
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_SET_INTERNAL_XATTR,
+ "fsetxattr dump=%s", dump, NULL);
+out:
+ if (dump) {
+ GF_FREE(dump);
+ }
+ return;
}
int
-br_stub_fsetxattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *dict, int flags, dict_t *xdata)
-{
- int32_t ret = 0;
- uint32_t val = 0;
- br_isignature_t *sign = NULL;
- br_stub_private_t *priv = NULL;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
-
- priv = this->private;
-
- if ((frame->root->pid != GF_CLIENT_PID_BITD &&
- frame->root->pid != GF_CLIENT_PID_SCRUB) &&
- br_stub_internal_xattr (dict)) {
- br_stub_dump_xattr (this, dict, &op_errno);
- goto unwind;
- }
-
- if (!priv->do_versioning)
- goto wind;
-
- if (!IA_ISREG (fd->inode->ia_type))
- goto wind;
-
- /* object signature request */
- ret = dict_get_bin (dict, GLUSTERFS_SET_OBJECT_SIGNATURE,
- (void **) &sign);
- if (!ret) {
- br_stub_handle_object_signature (frame, this,
- fd, dict, sign, xdata);
- goto done;
- }
-
- /* signing xattr */
- if (dict_get(dict, BITROT_SIGNING_VERSION_KEY)) {
- br_stub_handle_internal_xattr (frame, this, fd,
- BITROT_SIGNING_VERSION_KEY);
- goto done;
- }
-
- /* version xattr */
- if (dict_get(dict, BITROT_CURRENT_VERSION_KEY)) {
- br_stub_handle_internal_xattr (frame, this, fd,
- BITROT_CURRENT_VERSION_KEY);
- goto done;
- }
-
- if (dict_get (dict, GLUSTERFS_GET_OBJECT_SIGNATURE)) {
- br_stub_handle_internal_xattr (frame, this, fd,
- GLUSTERFS_GET_OBJECT_SIGNATURE);
- goto done;
- }
-
- /* object reopen request */
- ret = dict_get_uint32 (dict, BR_REOPEN_SIGN_HINT_KEY, &val);
- if (!ret) {
- br_stub_handle_object_reopen (frame, this, fd, val);
- goto done;
- }
-
- /* handle bad object */
- if (dict_get (dict, BITROT_OBJECT_BAD_KEY)) {
- br_stub_handle_bad_object_key (frame, this, fd,
- dict, flags, xdata);
- goto done;
- }
+br_stub_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int flags, dict_t *xdata)
+{
+ int32_t ret = 0;
+ uint32_t val = 0;
+ br_isignature_t *sign = NULL;
+ br_stub_private_t *priv = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+
+ priv = this->private;
+
+ if ((frame->root->pid != GF_CLIENT_PID_BITD &&
+ frame->root->pid != GF_CLIENT_PID_SCRUB) &&
+ br_stub_internal_xattr(dict)) {
+ br_stub_dump_xattr(this, dict, &op_errno);
+ goto unwind;
+ }
+
+ if (!priv->do_versioning)
+ goto wind;
+
+ if (!IA_ISREG(fd->inode->ia_type))
+ goto wind;
+
+ /* object signature request */
+ ret = dict_get_bin(dict, GLUSTERFS_SET_OBJECT_SIGNATURE, (void **)&sign);
+ if (!ret) {
+ gf_msg_debug(this->name, 0, "got SIGNATURE request on %s",
+ uuid_utoa(fd->inode->gfid));
+ br_stub_handle_object_signature(frame, this, fd, dict, sign, xdata);
+ goto done;
+ }
+
+ /* signing xattr */
+ if (dict_get(dict, BITROT_SIGNING_VERSION_KEY)) {
+ br_stub_handle_internal_xattr(frame, this, fd,
+ BITROT_SIGNING_VERSION_KEY);
+ goto done;
+ }
+
+ /* version xattr */
+ if (dict_get(dict, BITROT_CURRENT_VERSION_KEY)) {
+ br_stub_handle_internal_xattr(frame, this, fd,
+ BITROT_CURRENT_VERSION_KEY);
+ goto done;
+ }
+
+ if (dict_get(dict, GLUSTERFS_GET_OBJECT_SIGNATURE)) {
+ br_stub_handle_internal_xattr(frame, this, fd,
+ GLUSTERFS_GET_OBJECT_SIGNATURE);
+ goto done;
+ }
+
+ /* object reopen request */
+ ret = dict_get_uint32(dict, BR_REOPEN_SIGN_HINT_KEY, &val);
+ if (!ret) {
+ br_stub_handle_object_reopen(frame, this, fd, val);
+ goto done;
+ }
+
+ /* handle bad object */
+ if (dict_get(dict, BITROT_OBJECT_BAD_KEY)) {
+ br_stub_handle_bad_object_key(frame, this, fd, dict, flags, xdata);
+ goto done;
+ }
wind:
- STACK_WIND (frame, default_fsetxattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fsetxattr, fd, dict, flags,
- xdata);
- return 0;
+ STACK_WIND(frame, default_fsetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+ return 0;
unwind:
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, NULL);
+ STACK_UNWIND_STRICT(fsetxattr, frame, op_ret, op_errno, NULL);
done:
- return 0;
+ return 0;
}
-
/**
* Currently BitD and scrubber are doing fsetxattr to either sign the object
* or to mark it as bad. Hence setxattr on any of those keys is denied directly
@@ -1418,82 +1475,76 @@ done:
* check has to be added below.
*/
int
-br_stub_setxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *dict, int flags, dict_t *xdata)
+br_stub_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int flags, dict_t *xdata)
{
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
- if (br_stub_internal_xattr (dict)) {
- br_stub_dump_xattr (this, dict, &op_errno);
- goto unwind;
- }
+ if (br_stub_internal_xattr(dict)) {
+ br_stub_dump_xattr(this, dict, &op_errno);
+ goto unwind;
+ }
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setxattr, loc, dict, flags,
- xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->setxattr,
+ loc, dict, flags, xdata);
+ return 0;
unwind:
- STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, NULL);
- return 0;
+ STACK_UNWIND_STRICT(setxattr, frame, op_ret, op_errno, NULL);
+ return 0;
}
/** }}} */
-
/** {{{ */
/* {f}removexattr() */
int32_t
-br_stub_removexattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
-{
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
-
- if (!strcmp (BITROT_OBJECT_BAD_KEY, name) ||
- !strcmp (BITROT_SIGNING_VERSION_KEY, name) ||
- !strcmp (BITROT_CURRENT_VERSION_KEY, name)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- BRS_MSG_REMOVE_INTERNAL_XATTR, "removexattr called"
- " on internal xattr %s for file %s", name, loc->path);
- goto unwind;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->removexattr,
- loc, name, xdata);
- return 0;
+br_stub_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
+{
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+
+ if (!strcmp(BITROT_OBJECT_BAD_KEY, name) ||
+ !strcmp(BITROT_SIGNING_VERSION_KEY, name) ||
+ !strcmp(BITROT_CURRENT_VERSION_KEY, name)) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_REMOVE_INTERNAL_XATTR,
+ "name=%s", name, "file-path=%s", loc->path, NULL);
+ goto unwind;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
+ return 0;
unwind:
- STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, NULL);
- return 0;
+ STACK_UNWIND_STRICT(removexattr, frame, op_ret, op_errno, NULL);
+ return 0;
}
int32_t
-br_stub_fremovexattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, const char *name, dict_t *xdata)
-{
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
-
- if (!strcmp (BITROT_OBJECT_BAD_KEY, name) ||
- !strcmp (BITROT_SIGNING_VERSION_KEY, name) ||
- !strcmp (BITROT_CURRENT_VERSION_KEY, name)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- BRS_MSG_REMOVE_INTERNAL_XATTR, "removexattr called"
- " on internal xattr %s for inode %s", name,
- uuid_utoa (fd->inode->gfid));
- goto unwind;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fremovexattr,
- fd, name, xdata);
- return 0;
+br_stub_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata)
+{
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+
+ if (!strcmp(BITROT_OBJECT_BAD_KEY, name) ||
+ !strcmp(BITROT_SIGNING_VERSION_KEY, name) ||
+ !strcmp(BITROT_CURRENT_VERSION_KEY, name)) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_REMOVE_INTERNAL_XATTR,
+ "name=%s", name, "inode-gfid=%s", uuid_utoa(fd->inode->gfid),
+ NULL);
+ goto unwind;
+ }
+
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fremovexattr, fd, name, xdata);
+ return 0;
unwind:
- STACK_UNWIND_STRICT (fremovexattr, frame, op_ret, op_errno, NULL);
- return 0;
+ STACK_UNWIND_STRICT(fremovexattr, frame, op_ret, op_errno, NULL);
+ return 0;
}
/** }}} */
@@ -1503,17 +1554,17 @@ unwind:
/* {f}getxattr() */
int
-br_stub_listxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
+br_stub_listxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
{
- if (op_ret < 0)
- goto unwind;
+ if (op_ret < 0)
+ goto unwind;
- br_stub_remove_vxattrs (xattr);
+ br_stub_remove_vxattrs(xattr, _gf_true);
- unwind:
- STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, xattr, xdata);
- return 0;
+unwind:
+ STACK_UNWIND_STRICT(getxattr, frame, op_ret, op_errno, xattr, xdata);
+ return 0;
}
/**
@@ -1552,383 +1603,380 @@ br_stub_listxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
* then treat the object as stale.
*/
char
-br_stub_is_object_stale (xlator_t *this, call_frame_t *frame, inode_t *inode,
- br_version_t *obuf, br_signature_t *sbuf)
-{
- uint64_t ctx_addr = 0;
- br_stub_inode_ctx_t *ctx = NULL;
- int32_t ret = -1;
- char stale = 0;
-
- if (obuf->ongoingversion == sbuf->signedversion)
- goto out;
-
- if (frame->root->pid == GF_CLIENT_PID_SCRUB) {
- stale = 1;
- goto out;
- }
-
- ret = br_stub_get_inode_ctx (this, inode, &ctx_addr);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_GET_INODE_CONTEXT_FAILED, "failed to get the "
- "inode context for %s", uuid_utoa (inode->gfid));
- goto out;
- }
-
- ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
-
- LOCK (&inode->lock);
- {
- if ((!__br_stub_is_inode_dirty (ctx) &&
- ctx->info_sign != BR_SIGN_NORMAL) ||
- __br_stub_is_inode_dirty (ctx))
- stale = 1;
- }
- UNLOCK (&inode->lock);
+br_stub_is_object_stale(xlator_t *this, call_frame_t *frame, inode_t *inode,
+ br_version_t *obuf, br_signature_t *sbuf)
+{
+ uint64_t ctx_addr = 0;
+ br_stub_inode_ctx_t *ctx = NULL;
+ int32_t ret = -1;
+ char stale = 0;
+
+ if (obuf->ongoingversion == sbuf->signedversion)
+ goto out;
+
+ if (frame->root->pid == GF_CLIENT_PID_SCRUB) {
+ stale = 1;
+ goto out;
+ }
+
+ ret = br_stub_get_inode_ctx(this, inode, &ctx_addr);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_GET_INODE_CONTEXT_FAILED,
+ "gfid=%s", uuid_utoa(inode->gfid), NULL);
+ goto out;
+ }
+
+ ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
+
+ LOCK(&inode->lock);
+ {
+ if ((!__br_stub_is_inode_dirty(ctx) &&
+ ctx->info_sign != BR_SIGN_NORMAL) ||
+ __br_stub_is_inode_dirty(ctx))
+ stale = 1;
+ }
+ UNLOCK(&inode->lock);
out:
- return stale;
+ return stale;
}
int
-br_stub_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
-{
- int32_t ret = 0;
- size_t totallen = 0;
- size_t signaturelen = 0;
- br_stub_private_t *priv = NULL;
- br_version_t *obuf = NULL;
- br_signature_t *sbuf = NULL;
- br_isignature_out_t *sign = NULL;
- br_vxattr_status_t status;
- br_stub_local_t *local = NULL;
- inode_t *inode = NULL;
- gf_boolean_t bad_object = _gf_false;
- gf_boolean_t ver_enabled = _gf_false;
-
- BR_STUB_VER_ENABLED_IN_CALLPATH(frame, ver_enabled);
- priv = this->private;
-
- if (op_ret < 0)
- goto unwind;
- BR_STUB_VER_COND_GOTO (priv, (!ver_enabled), delkeys);
-
- if (cookie != (void *) BR_STUB_REQUEST_COOKIE)
- goto unwind;
-
- local = frame->local;
- frame->local = NULL;
- inode = local->u.context.inode;
-
- op_ret = -1;
- status = br_version_xattr_state (xattr, &obuf, &sbuf, &bad_object);
-
- op_errno = EIO;
- if (bad_object)
- goto delkeys;
-
- op_errno = EINVAL;
- if (status == BR_VXATTR_STATUS_INVALID)
- goto delkeys;
-
- op_errno = ENODATA;
- if ((status == BR_VXATTR_STATUS_MISSING)
- || (status == BR_VXATTR_STATUS_UNSIGNED))
- goto delkeys;
-
- /**
- * okay.. we have enough information to satisfy the request,
- * namely: version and signing extended attribute. what's
- * pending is the signature length -- that's figured out
- * indirectly via the size of the _whole_ xattr and the
- * on-disk signing xattr header size.
- */
- op_errno = EINVAL;
- ret = dict_get_uint32 (xattr, BITROT_SIGNING_XATTR_SIZE_KEY,
- (uint32_t *)&signaturelen);
- if (ret)
- goto delkeys;
-
- signaturelen -= sizeof (br_signature_t);
- totallen = sizeof (br_isignature_out_t) + signaturelen;
-
- op_errno = ENOMEM;
- sign = GF_CALLOC (1, totallen, gf_br_stub_mt_signature_t);
- if (!sign)
- goto delkeys;
-
- sign->time[0] = obuf->timebuf[0];
- sign->time[1] = obuf->timebuf[1];
-
- /* Object's dirty state & current signed version */
- sign->version = sbuf->signedversion;
- sign->stale = br_stub_is_object_stale (this, frame, inode, obuf, sbuf);
-
- /* Object's signature */
- sign->signaturelen = signaturelen;
- sign->signaturetype = sbuf->signaturetype;
- (void) memcpy (sign->signature, sbuf->signature, signaturelen);
-
+br_stub_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
+{
+ int32_t ret = 0;
+ size_t totallen = 0;
+ size_t signaturelen = 0;
+ br_stub_private_t *priv = NULL;
+ br_version_t *obuf = NULL;
+ br_signature_t *sbuf = NULL;
+ br_isignature_out_t *sign = NULL;
+ br_vxattr_status_t status;
+ br_stub_local_t *local = NULL;
+ inode_t *inode = NULL;
+ gf_boolean_t bad_object = _gf_false;
+ gf_boolean_t ver_enabled = _gf_false;
+
+ BR_STUB_VER_ENABLED_IN_CALLPATH(frame, ver_enabled);
+ priv = this->private;
+
+ if (op_ret < 0)
+ goto unwind;
+ BR_STUB_VER_COND_GOTO(priv, (!ver_enabled), delkeys);
+
+ if (cookie != (void *)BR_STUB_REQUEST_COOKIE)
+ goto unwind;
+
+ local = frame->local;
+ frame->local = NULL;
+ if (!local) {
+ op_ret = -1;
op_errno = EINVAL;
- ret = dict_set_bin (xattr, GLUSTERFS_GET_OBJECT_SIGNATURE,
- (void *)sign, totallen);
- if (ret < 0) {
- GF_FREE (sign);
- goto delkeys;
- }
- op_errno = 0;
- op_ret = totallen;
+ goto unwind;
+ }
+ inode = local->u.context.inode;
+
+ op_ret = -1;
+ status = br_version_xattr_state(xattr, &obuf, &sbuf, &bad_object);
+
+ op_errno = EIO;
+ if (bad_object)
+ goto delkeys;
+
+ op_errno = EINVAL;
+ if (status == BR_VXATTR_STATUS_INVALID)
+ goto delkeys;
+
+ op_errno = ENODATA;
+ if ((status == BR_VXATTR_STATUS_MISSING) ||
+ (status == BR_VXATTR_STATUS_UNSIGNED))
+ goto delkeys;
+
+ /**
+ * okay.. we have enough information to satisfy the request,
+ * namely: version and signing extended attribute. what's
+ * pending is the signature length -- that's figured out
+ * indirectly via the size of the _whole_ xattr and the
+ * on-disk signing xattr header size.
+ */
+ op_errno = EINVAL;
+ ret = dict_get_uint32(xattr, BITROT_SIGNING_XATTR_SIZE_KEY,
+ (uint32_t *)&signaturelen);
+ if (ret)
+ goto delkeys;
+
+ signaturelen -= sizeof(br_signature_t);
+ totallen = sizeof(br_isignature_out_t) + signaturelen;
+
+ op_errno = ENOMEM;
+ sign = GF_CALLOC(1, totallen, gf_br_stub_mt_signature_t);
+ if (!sign)
+ goto delkeys;
+
+ sign->time[0] = obuf->timebuf[0];
+ sign->time[1] = obuf->timebuf[1];
+
+ /* Object's dirty state & current signed version */
+ sign->version = sbuf->signedversion;
+ sign->stale = br_stub_is_object_stale(this, frame, inode, obuf, sbuf);
+
+ /* Object's signature */
+ sign->signaturelen = signaturelen;
+ sign->signaturetype = sbuf->signaturetype;
+ (void)memcpy(sign->signature, sbuf->signature, signaturelen);
+
+ op_errno = EINVAL;
+ ret = dict_set_bin(xattr, GLUSTERFS_GET_OBJECT_SIGNATURE, (void *)sign,
+ totallen);
+ if (ret < 0) {
+ GF_FREE(sign);
+ goto delkeys;
+ }
+ op_errno = 0;
+ op_ret = totallen;
+
+delkeys:
+ br_stub_remove_vxattrs(xattr, _gf_true);
- delkeys:
- br_stub_remove_vxattrs (xattr);
-
- unwind:
- STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, xattr, xdata);
- if (local) {
- br_stub_cleanup_local (local);
- br_stub_dealloc_local (local);
- }
- return 0;
+unwind:
+ STACK_UNWIND_STRICT(getxattr, frame, op_ret, op_errno, xattr, xdata);
+ br_stub_cleanup_local(local);
+ br_stub_dealloc_local(local);
+ return 0;
}
static void
-br_stub_send_stub_init_time (call_frame_t *frame, xlator_t *this)
+br_stub_send_stub_init_time(call_frame_t *frame, xlator_t *this)
{
- int op_ret = 0;
- int op_errno = 0;
- dict_t *xattr = NULL;
- br_stub_init_t stub = {{0,},};
- br_stub_private_t *priv = NULL;
+ int op_ret = 0;
+ int op_errno = 0;
+ dict_t *xattr = NULL;
+ br_stub_init_t stub = {
+ {
+ 0,
+ },
+ };
+ br_stub_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- xattr = dict_new ();
- if (!xattr) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
+ xattr = dict_new();
+ if (!xattr) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
- stub.timebuf[0] = priv->boot[0];
- stub.timebuf[1] = priv->boot[1];
- memcpy (stub.export, priv->export, strlen (priv->export) + 1);
+ stub.timebuf[0] = priv->boot[0];
+ stub.timebuf[1] = priv->boot[1];
+ memcpy(stub.export, priv->export, strlen(priv->export) + 1);
- op_ret = dict_set_static_bin (xattr, GLUSTERFS_GET_BR_STUB_INIT_TIME,
- (void *) &stub, sizeof (br_stub_init_t));
- if (op_ret < 0) {
- op_errno = EINVAL;
- goto unwind;
- }
+ op_ret = dict_set_static_bin(xattr, GLUSTERFS_GET_BR_STUB_INIT_TIME,
+ (void *)&stub, sizeof(br_stub_init_t));
+ if (op_ret < 0) {
+ op_errno = EINVAL;
+ goto unwind;
+ }
- op_ret = sizeof (br_stub_init_t);
+ op_ret = sizeof(br_stub_init_t);
- unwind:
- STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, xattr, NULL);
+unwind:
+ STACK_UNWIND_STRICT(getxattr, frame, op_ret, op_errno, xattr, NULL);
- if (xattr)
- dict_unref (xattr);
+ if (xattr)
+ dict_unref(xattr);
}
int
-br_stub_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
-{
- void *cookie = NULL;
- uuid_t rootgfid = {0, };
- fop_getxattr_cbk_t cbk = br_stub_getxattr_cbk;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- br_stub_local_t *local = NULL;
- br_stub_private_t *priv = NULL;
-
-
- GF_VALIDATE_OR_GOTO ("bit-rot-stub", this, unwind);
- GF_VALIDATE_OR_GOTO (this->name, loc, unwind);
- GF_VALIDATE_OR_GOTO (this->name, this->private, unwind);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, unwind);
-
- rootgfid[15] = 1;
-
- if (!name) {
- cbk = br_stub_listxattr_cbk;
- goto wind;
- }
-
- if (br_stub_is_internal_xattr (name))
- goto unwind;
-
- priv = this->private;
- BR_STUB_VER_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
-
- /**
- * If xattr is node-uuid and the inode is marked bad, return EIO.
- * Returning EIO would result in AFR to choose correct node-uuid
- * coresponding to the subvolume * where the good copy of the
- * file resides.
- */
- if (IA_ISREG (loc->inode->ia_type) && XATTR_IS_NODE_UUID (name) &&
- br_stub_check_bad_object (this, loc->inode, &op_ret, &op_errno)) {
- goto unwind;
- }
-
- /**
- * this special extended attribute is allowed only on root
- */
- if (name
- && (strncmp (name, GLUSTERFS_GET_BR_STUB_INIT_TIME,
- strlen (GLUSTERFS_GET_BR_STUB_INIT_TIME)) == 0)
- && ((gf_uuid_compare (loc->gfid, rootgfid) == 0)
- || (gf_uuid_compare (loc->inode->gfid, rootgfid) == 0))) {
- BR_STUB_RESET_LOCAL_NULL (frame);
- br_stub_send_stub_init_time (frame, this);
- return 0;
+br_stub_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
+{
+ void *cookie = NULL;
+ static uuid_t rootgfid = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1};
+ fop_getxattr_cbk_t cbk = br_stub_getxattr_cbk;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ br_stub_local_t *local = NULL;
+ br_stub_private_t *priv = NULL;
+
+ GF_VALIDATE_OR_GOTO("bit-rot-stub", this, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, loc, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, unwind);
+
+ if (!name) {
+ cbk = br_stub_listxattr_cbk;
+ goto wind;
+ }
+
+ if (br_stub_is_internal_xattr(name))
+ goto unwind;
+
+ priv = this->private;
+ BR_STUB_VER_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
+
+ /**
+ * If xattr is node-uuid and the inode is marked bad, return EIO.
+ * Returning EIO would result in AFR to choose correct node-uuid
+ * corresponding to the subvolume * where the good copy of the
+ * file resides.
+ */
+ if (IA_ISREG(loc->inode->ia_type) && XATTR_IS_NODE_UUID(name) &&
+ br_stub_check_bad_object(this, loc->inode, &op_ret, &op_errno)) {
+ goto unwind;
+ }
+
+ /**
+ * this special extended attribute is allowed only on root
+ */
+ if (name &&
+ (strncmp(name, GLUSTERFS_GET_BR_STUB_INIT_TIME,
+ sizeof(GLUSTERFS_GET_BR_STUB_INIT_TIME) - 1) == 0) &&
+ ((gf_uuid_compare(loc->gfid, rootgfid) == 0) ||
+ (gf_uuid_compare(loc->inode->gfid, rootgfid) == 0))) {
+ BR_STUB_RESET_LOCAL_NULL(frame);
+ br_stub_send_stub_init_time(frame, this);
+ return 0;
+ }
+
+ if (!IA_ISREG(loc->inode->ia_type))
+ goto wind;
+
+ if (name && (strncmp(name, GLUSTERFS_GET_OBJECT_SIGNATURE,
+ sizeof(GLUSTERFS_GET_OBJECT_SIGNATURE) - 1) == 0)) {
+ cookie = (void *)BR_STUB_REQUEST_COOKIE;
+
+ local = br_stub_alloc_local(this);
+ if (!local) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
}
- if (!IA_ISREG (loc->inode->ia_type))
- goto wind;
-
- if (name && (strncmp (name, GLUSTERFS_GET_OBJECT_SIGNATURE,
- strlen (GLUSTERFS_GET_OBJECT_SIGNATURE)) == 0)) {
- cookie = (void *) BR_STUB_REQUEST_COOKIE;
-
- local = br_stub_alloc_local (this);
- if (!local) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
-
- br_stub_fill_local (local, NULL, NULL, loc->inode,
- loc->inode->gfid,
- BR_STUB_NO_VERSIONING, 0);
- frame->local = local;
- }
+ br_stub_fill_local(local, NULL, NULL, loc->inode, loc->inode->gfid,
+ BR_STUB_NO_VERSIONING, 0);
+ frame->local = local;
+ }
- wind:
- STACK_WIND_COOKIE
- (frame, cbk, cookie, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->getxattr, loc, name, xdata);
- return 0;
+wind:
+ STACK_WIND_COOKIE(frame, cbk, cookie, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+ return 0;
unwind:
- BR_STUB_RESET_LOCAL_NULL (frame);
- STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, NULL, NULL);
- return 0;
+ BR_STUB_RESET_LOCAL_NULL(frame);
+ STACK_UNWIND_STRICT(getxattr, frame, op_ret, op_errno, NULL, NULL);
+ return 0;
}
int
-br_stub_fgetxattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, const char *name, dict_t *xdata)
-{
- void *cookie = NULL;
- uuid_t rootgfid = {0, };
- fop_fgetxattr_cbk_t cbk = br_stub_getxattr_cbk;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- br_stub_local_t *local = NULL;
- br_stub_private_t *priv = NULL;
-
- rootgfid[15] = 1;
- priv = this->private;
-
- if (!name) {
- cbk = br_stub_listxattr_cbk;
- goto wind;
- }
-
- if (br_stub_is_internal_xattr (name))
- goto unwind;
-
- BR_STUB_VER_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
-
- /**
- * If xattr is node-uuid and the inode is marked bad, return EIO.
- * Returning EIO would result in AFR to choose correct node-uuid
- * coresponding to the subvolume * where the good copy of the
- * file resides.
- */
- if (IA_ISREG (fd->inode->ia_type) && XATTR_IS_NODE_UUID (name) &&
- br_stub_check_bad_object (this, fd->inode, &op_ret, &op_errno)) {
- goto unwind;
- }
-
- /**
- * this special extended attribute is allowed only on root
- */
- if (name
- && (strncmp (name, GLUSTERFS_GET_BR_STUB_INIT_TIME,
- strlen (GLUSTERFS_GET_BR_STUB_INIT_TIME)) == 0)
- && (gf_uuid_compare (fd->inode->gfid, rootgfid) == 0)) {
- BR_STUB_RESET_LOCAL_NULL (frame);
- br_stub_send_stub_init_time (frame, this);
- return 0;
+br_stub_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata)
+{
+ void *cookie = NULL;
+ static uuid_t rootgfid = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1};
+ fop_fgetxattr_cbk_t cbk = br_stub_getxattr_cbk;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ br_stub_local_t *local = NULL;
+ br_stub_private_t *priv = NULL;
+
+ priv = this->private;
+
+ if (!name) {
+ cbk = br_stub_listxattr_cbk;
+ goto wind;
+ }
+
+ if (br_stub_is_internal_xattr(name))
+ goto unwind;
+
+ BR_STUB_VER_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
+
+ /**
+ * If xattr is node-uuid and the inode is marked bad, return EIO.
+ * Returning EIO would result in AFR to choose correct node-uuid
+ * corresponding to the subvolume * where the good copy of the
+ * file resides.
+ */
+ if (IA_ISREG(fd->inode->ia_type) && XATTR_IS_NODE_UUID(name) &&
+ br_stub_check_bad_object(this, fd->inode, &op_ret, &op_errno)) {
+ goto unwind;
+ }
+
+ /**
+ * this special extended attribute is allowed only on root
+ */
+ if (name &&
+ (strncmp(name, GLUSTERFS_GET_BR_STUB_INIT_TIME,
+ sizeof(GLUSTERFS_GET_BR_STUB_INIT_TIME) - 1) == 0) &&
+ (gf_uuid_compare(fd->inode->gfid, rootgfid) == 0)) {
+ BR_STUB_RESET_LOCAL_NULL(frame);
+ br_stub_send_stub_init_time(frame, this);
+ return 0;
+ }
+
+ if (!IA_ISREG(fd->inode->ia_type))
+ goto wind;
+
+ if (name && (strncmp(name, GLUSTERFS_GET_OBJECT_SIGNATURE,
+ sizeof(GLUSTERFS_GET_OBJECT_SIGNATURE) - 1) == 0)) {
+ cookie = (void *)BR_STUB_REQUEST_COOKIE;
+
+ local = br_stub_alloc_local(this);
+ if (!local) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
}
- if (!IA_ISREG (fd->inode->ia_type))
- goto wind;
-
- if (name && (strncmp (name, GLUSTERFS_GET_OBJECT_SIGNATURE,
- strlen (GLUSTERFS_GET_OBJECT_SIGNATURE)) == 0)) {
- cookie = (void *) BR_STUB_REQUEST_COOKIE;
-
- local = br_stub_alloc_local (this);
- if (!local) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
-
- br_stub_fill_local (local, NULL, fd, fd->inode,
- fd->inode->gfid,
- BR_STUB_NO_VERSIONING, 0);
- frame->local = local;
- }
+ br_stub_fill_local(local, NULL, fd, fd->inode, fd->inode->gfid,
+ BR_STUB_NO_VERSIONING, 0);
+ frame->local = local;
+ }
- wind:
- STACK_WIND_COOKIE
- (frame, cbk, cookie, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fgetxattr, fd, name, xdata);
- return 0;
+wind:
+ STACK_WIND_COOKIE(frame, cbk, cookie, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
+ return 0;
unwind:
- BR_STUB_RESET_LOCAL_NULL (frame);
- STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, NULL, NULL);
- return 0;
+ BR_STUB_RESET_LOCAL_NULL(frame);
+ STACK_UNWIND_STRICT(fgetxattr, frame, op_ret, op_errno, NULL, NULL);
+ return 0;
}
int32_t
-br_stub_readv (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata)
+br_stub_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- int32_t ret = -1;
- br_stub_private_t *priv = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ int32_t ret = -1;
+ br_stub_private_t *priv = NULL;
- GF_VALIDATE_OR_GOTO ("bit-rot-stub", this, unwind);
- GF_VALIDATE_OR_GOTO (this->name, frame, unwind);
- GF_VALIDATE_OR_GOTO (this->name, this->private, unwind);
- GF_VALIDATE_OR_GOTO (this->name, fd, unwind);
- GF_VALIDATE_OR_GOTO (this->name, fd->inode, unwind);
+ GF_VALIDATE_OR_GOTO("bit-rot-stub", this, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, frame, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, fd, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, fd->inode, unwind);
- priv = this->private;
- if (!priv->do_versioning)
- goto wind;
+ priv = this->private;
+ if (!priv->do_versioning)
+ goto wind;
- ret = br_stub_check_bad_object (this, fd->inode, &op_ret, &op_errno);
- if (ret)
- goto unwind;
+ ret = br_stub_check_bad_object(this, fd->inode, &op_ret, &op_errno);
+ if (ret)
+ goto unwind;
wind:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readv, fd, size, offset,
- flags, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->readv,
+ fd, size, offset, flags, xdata);
+ return 0;
unwind:
- STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, NULL, 0, NULL,
- NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(readv, frame, op_ret, op_errno, NULL, 0, NULL, NULL,
+ NULL);
+ return 0;
}
/**
@@ -1939,299 +1987,294 @@ unwind:
* fds.
*/
int32_t
-br_stub_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+br_stub_writev_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- int32_t ret = 0;
- br_stub_local_t *local = NULL;
+ int32_t ret = 0;
+ br_stub_local_t *local = NULL;
- local = frame->local;
- frame->local = NULL;
+ local = frame->local;
+ frame->local = NULL;
- if (op_ret < 0)
- goto unwind;
+ if (op_ret < 0)
+ goto unwind;
- ret = br_stub_mark_inode_modified (this, local);
- if (ret) {
- op_ret = -1;
- op_errno = EINVAL;
- }
+ ret = br_stub_mark_inode_modified(this, local);
+ if (ret) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ }
unwind:
- STACK_UNWIND_STRICT (writev, frame,
- op_ret, op_errno, prebuf, postbuf, xdata);
+ STACK_UNWIND_STRICT(writev, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
- br_stub_cleanup_local (local);
- br_stub_dealloc_local (local);
+ br_stub_cleanup_local(local);
+ br_stub_dealloc_local(local);
- return 0;
+ return 0;
}
int32_t
-br_stub_writev_resume (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count, off_t offset,
- uint32_t flags, struct iobref *iobref, dict_t *xdata)
+br_stub_writev_resume(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int32_t count, off_t offset,
+ uint32_t flags, struct iobref *iobref, dict_t *xdata)
{
- STACK_WIND (frame, br_stub_writev_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->writev, fd, vector, count,
- offset, flags, iobref, xdata);
- return 0;
+ STACK_WIND(frame, br_stub_writev_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector, count, offset,
+ flags, iobref, xdata);
+ return 0;
}
/**
* This is probably the most crucial part about the whole versioning thing.
* There's absolutely no differentiation as such between an anonymous fd
* and a regular fd except the fd context initialization. Object versioning
- * is perfomed when the inode is dirty. Parallel write operations are no
+ * is performed when the inode is dirty. Parallel write operations are no
* special with each write performing object versioning followed by marking
* the inode as non-dirty (synced). This is followed by the actual operation
* (writev() in this case) which on a success marks the inode as modified.
* This prevents signing of objects that have not been modified.
*/
int32_t
-br_stub_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count, off_t offset,
- uint32_t flags, struct iobref *iobref, dict_t *xdata)
-{
- call_stub_t *stub = NULL;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- gf_boolean_t inc_version = _gf_false;
- gf_boolean_t modified = _gf_false;
- br_stub_inode_ctx_t *ctx = NULL;
- int32_t ret = -1;
- fop_writev_cbk_t cbk = default_writev_cbk;
- br_stub_local_t *local = NULL;
- br_stub_private_t *priv = NULL;
-
- GF_VALIDATE_OR_GOTO ("bit-rot-stub", this, unwind);
- GF_VALIDATE_OR_GOTO (this->name, this->private, unwind);
- GF_VALIDATE_OR_GOTO (this->name, frame, unwind);
- GF_VALIDATE_OR_GOTO (this->name, fd, unwind);
-
- priv = this->private;
- if (!priv->do_versioning)
- goto wind;
-
- ret = br_stub_need_versioning (this, fd, &inc_version, &modified, &ctx);
- if (ret)
- goto unwind;
-
- ret = br_stub_check_bad_object (this, fd->inode, &op_ret, &op_errno);
- if (ret)
- goto unwind;
-
- /**
- * The inode is not dirty and also witnessed atleast one successful
- * modification operation. Therefore, subsequent operations need not
- * perform any special tracking.
- */
- if (!inc_version && modified)
- goto wind;
-
- /**
- * okay.. so, either the inode needs versioning or the modification
- * needs to be tracked. ->cbk is set to the appropriate callback
- * routine for this.
- * NOTE: ->local needs to be deallocated on failures from here on.
- */
- ret = br_stub_versioning_prep (frame, this, fd, ctx);
- if (ret)
- goto unwind;
-
- local = frame->local;
- if (!inc_version) {
- br_stub_fill_local (local, NULL, fd, fd->inode,
- fd->inode->gfid, BR_STUB_NO_VERSIONING, 0);
- cbk = br_stub_writev_cbk;
- goto wind;
- }
+br_stub_writev(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int32_t count, off_t offset,
+ uint32_t flags, struct iobref *iobref, dict_t *xdata)
+{
+ call_stub_t *stub = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ gf_boolean_t inc_version = _gf_false;
+ gf_boolean_t modified = _gf_false;
+ br_stub_inode_ctx_t *ctx = NULL;
+ int32_t ret = -1;
+ fop_writev_cbk_t cbk = default_writev_cbk;
+ br_stub_local_t *local = NULL;
+ br_stub_private_t *priv = NULL;
+
+ GF_VALIDATE_OR_GOTO("bit-rot-stub", this, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, frame, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, fd, unwind);
+
+ priv = this->private;
+ if (!priv->do_versioning)
+ goto wind;
+
+ ret = br_stub_need_versioning(this, fd, &inc_version, &modified, &ctx);
+ if (ret)
+ goto unwind;
+
+ ret = br_stub_check_bad_object(this, fd->inode, &op_ret, &op_errno);
+ if (ret)
+ goto unwind;
+
+ /**
+ * The inode is not dirty and also witnessed at least one successful
+ * modification operation. Therefore, subsequent operations need not
+ * perform any special tracking.
+ */
+ if (!inc_version && modified)
+ goto wind;
+
+ /**
+ * okay.. so, either the inode needs versioning or the modification
+ * needs to be tracked. ->cbk is set to the appropriate callback
+ * routine for this.
+ * NOTE: ->local needs to be deallocated on failures from here on.
+ */
+ ret = br_stub_versioning_prep(frame, this, fd, ctx);
+ if (ret)
+ goto unwind;
+
+ local = frame->local;
+ if (!inc_version) {
+ br_stub_fill_local(local, NULL, fd, fd->inode, fd->inode->gfid,
+ BR_STUB_NO_VERSIONING, 0);
+ cbk = br_stub_writev_cbk;
+ goto wind;
+ }
+
+ stub = fop_writev_stub(frame, br_stub_writev_resume, fd, vector, count,
+ offset, flags, iobref, xdata);
+
+ if (!stub) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_STUB_ALLOC_FAILED,
+ "write gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto cleanup_local;
+ }
+
+ /* Perform Versioning */
+ return br_stub_perform_incversioning(this, frame, stub, fd, ctx);
- stub = fop_writev_stub (frame, br_stub_writev_resume, fd, vector, count,
- offset, flags, iobref, xdata);
-
- if (!stub) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRS_MSG_STUB_ALLOC_FAILED,
- "failed to allocate stub for write fop (gfid: %s), "
- "unwinding", uuid_utoa (fd->inode->gfid));
- goto cleanup_local;
- }
-
- /* Perform Versioning */
- return br_stub_perform_incversioning (this, frame, stub, fd, ctx);
-
- wind:
- STACK_WIND (frame, cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->writev,
- fd, vector, count, offset, flags, iobref, xdata);
- return 0;
+wind:
+ STACK_WIND(frame, cbk, FIRST_CHILD(this), FIRST_CHILD(this)->fops->writev,
+ fd, vector, count, offset, flags, iobref, xdata);
+ return 0;
- cleanup_local:
- br_stub_cleanup_local (local);
- br_stub_dealloc_local (local);
+cleanup_local:
+ br_stub_cleanup_local(local);
+ br_stub_dealloc_local(local);
- unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, NULL, NULL,
- NULL);
+unwind:
+ frame->local = NULL;
+ STACK_UNWIND_STRICT(writev, frame, op_ret, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
int32_t
-br_stub_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+br_stub_ftruncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- int32_t ret = -1;
- br_stub_local_t *local = NULL;
+ int32_t ret = -1;
+ br_stub_local_t *local = NULL;
- local = frame->local;
- frame->local = NULL;
+ local = frame->local;
+ frame->local = NULL;
- if (op_ret < 0)
- goto unwind;
+ if (op_ret < 0)
+ goto unwind;
- ret = br_stub_mark_inode_modified (this, local);
- if (ret) {
- op_ret = -1;
- op_errno = EINVAL;
- }
+ ret = br_stub_mark_inode_modified(this, local);
+ if (ret) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ }
unwind:
- STACK_UNWIND_STRICT (ftruncate, frame,
- op_ret, op_errno, prebuf, postbuf, xdata);
+ STACK_UNWIND_STRICT(ftruncate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
- br_stub_cleanup_local (local);
- br_stub_dealloc_local (local);
+ br_stub_cleanup_local(local);
+ br_stub_dealloc_local(local);
- return 0;
+ return 0;
}
int32_t
-br_stub_ftruncate_resume (call_frame_t *frame, xlator_t *this, fd_t *fd,
- off_t offset, dict_t *xdata)
+br_stub_ftruncate_resume(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ off_t offset, dict_t *xdata)
{
- STACK_WIND (frame, br_stub_ftruncate_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
- return 0;
+ STACK_WIND(frame, br_stub_ftruncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+ return 0;
}
/* c.f. br_stub_writev() for explanation */
int32_t
-br_stub_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd,
- off_t offset, dict_t *xdata)
-{
- br_stub_local_t *local = NULL;
- call_stub_t *stub = NULL;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- gf_boolean_t inc_version = _gf_false;
- gf_boolean_t modified = _gf_false;
- br_stub_inode_ctx_t *ctx = NULL;
- int32_t ret = -1;
- fop_ftruncate_cbk_t cbk = default_ftruncate_cbk;
- br_stub_private_t *priv = NULL;
-
- GF_VALIDATE_OR_GOTO ("bit-rot-stub", this, unwind);
- GF_VALIDATE_OR_GOTO (this->name, this->private, unwind);
- GF_VALIDATE_OR_GOTO (this->name, frame, unwind);
- GF_VALIDATE_OR_GOTO (this->name, fd, unwind);
-
- priv = this->private;
- if (!priv->do_versioning)
- goto wind;
-
- ret = br_stub_need_versioning (this, fd, &inc_version, &modified, &ctx);
- if (ret)
- goto unwind;
-
- ret = br_stub_check_bad_object (this, fd->inode, &op_ret, &op_errno);
- if (ret)
- goto unwind;
-
- if (!inc_version && modified)
- goto wind;
-
- ret = br_stub_versioning_prep (frame, this, fd, ctx);
- if (ret)
- goto unwind;
-
- local = frame->local;
- if (!inc_version) {
- br_stub_fill_local (local, NULL, fd, fd->inode,
- fd->inode->gfid, BR_STUB_NO_VERSIONING, 0);
- cbk = br_stub_ftruncate_cbk;
- goto wind;
- }
+br_stub_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
+{
+ br_stub_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ gf_boolean_t inc_version = _gf_false;
+ gf_boolean_t modified = _gf_false;
+ br_stub_inode_ctx_t *ctx = NULL;
+ int32_t ret = -1;
+ fop_ftruncate_cbk_t cbk = default_ftruncate_cbk;
+ br_stub_private_t *priv = NULL;
+
+ GF_VALIDATE_OR_GOTO("bit-rot-stub", this, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, frame, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, fd, unwind);
+
+ priv = this->private;
+ if (!priv->do_versioning)
+ goto wind;
+
+ ret = br_stub_need_versioning(this, fd, &inc_version, &modified, &ctx);
+ if (ret)
+ goto unwind;
+
+ ret = br_stub_check_bad_object(this, fd->inode, &op_ret, &op_errno);
+ if (ret)
+ goto unwind;
+
+ if (!inc_version && modified)
+ goto wind;
+
+ ret = br_stub_versioning_prep(frame, this, fd, ctx);
+ if (ret)
+ goto unwind;
+
+ local = frame->local;
+ if (!inc_version) {
+ br_stub_fill_local(local, NULL, fd, fd->inode, fd->inode->gfid,
+ BR_STUB_NO_VERSIONING, 0);
+ cbk = br_stub_ftruncate_cbk;
+ goto wind;
+ }
+
+ stub = fop_ftruncate_stub(frame, br_stub_ftruncate_resume, fd, offset,
+ xdata);
+ if (!stub) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_STUB_ALLOC_FAILED,
+ "ftruncate gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto cleanup_local;
+ }
+
+ return br_stub_perform_incversioning(this, frame, stub, fd, ctx);
- stub = fop_ftruncate_stub (frame, br_stub_ftruncate_resume, fd, offset,
- xdata);
- if (!stub) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRS_MSG_STUB_ALLOC_FAILED,
- "failed to allocate stub for ftruncate fop (gfid: %s),"
- " unwinding", uuid_utoa (fd->inode->gfid));
- goto cleanup_local;
- }
-
- return br_stub_perform_incversioning (this, frame, stub, fd, ctx);
-
- wind:
- STACK_WIND (frame, cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
- return 0;
+wind:
+ STACK_WIND(frame, cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+ return 0;
- cleanup_local:
- br_stub_cleanup_local (local);
- br_stub_dealloc_local (local);
+cleanup_local:
+ br_stub_cleanup_local(local);
+ br_stub_dealloc_local(local);
- unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno, NULL, NULL,
- NULL);
+unwind:
+ frame->local = NULL;
+ STACK_UNWIND_STRICT(ftruncate, frame, op_ret, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
int32_t
-br_stub_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+br_stub_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- int32_t ret = 0;
- br_stub_local_t *local = NULL;
+ int32_t ret = 0;
+ br_stub_local_t *local = NULL;
- local = frame->local;
- frame->local = NULL;
+ local = frame->local;
+ frame->local = NULL;
- if (op_ret < 0)
- goto unwind;
+ if (op_ret < 0)
+ goto unwind;
- ret = br_stub_mark_inode_modified (this, local);
- if (ret) {
- op_ret = -1;
- op_errno = EINVAL;
- }
+ ret = br_stub_mark_inode_modified(this, local);
+ if (ret) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ }
unwind:
- STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
- br_stub_cleanup_local (local);
- br_stub_dealloc_local (local);
- return 0;
+ STACK_UNWIND_STRICT(truncate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ br_stub_cleanup_local(local);
+ br_stub_dealloc_local(local);
+ return 0;
}
int32_t
-br_stub_truncate_resume (call_frame_t *frame, xlator_t *this, loc_t *loc,
- off_t offset, dict_t *xdata)
+br_stub_truncate_resume(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ off_t offset, dict_t *xdata)
{
- br_stub_local_t *local = frame->local;
+ br_stub_local_t *local = frame->local;
- fd_unref (local->u.context.fd);
- STACK_WIND (frame, br_stub_ftruncate_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
- return 0;
+ fd_unref(local->u.context.fd);
+ STACK_WIND(frame, br_stub_ftruncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
+ return 0;
}
/**
@@ -2249,97 +2292,92 @@ br_stub_truncate_resume (call_frame_t *frame, xlator_t *this, loc_t *loc,
* c.f. br_writev_cbk() for explanation
*/
int32_t
-br_stub_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc,
- off_t offset, dict_t *xdata)
-{
- br_stub_local_t *local = NULL;
- call_stub_t *stub = NULL;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- gf_boolean_t inc_version = _gf_false;
- gf_boolean_t modified = _gf_false;
- br_stub_inode_ctx_t *ctx = NULL;
- int32_t ret = -1;
- fd_t *fd = NULL;
- fop_truncate_cbk_t cbk = default_truncate_cbk;
- br_stub_private_t *priv = NULL;
-
- GF_VALIDATE_OR_GOTO ("bit-rot-stub", this, unwind);
- GF_VALIDATE_OR_GOTO (this->name, this->private, unwind);
- GF_VALIDATE_OR_GOTO (this->name, frame, unwind);
- GF_VALIDATE_OR_GOTO (this->name, loc, unwind);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, unwind);
-
- priv = this->private;
- if (!priv->do_versioning)
- goto wind;
-
- fd = fd_anonymous (loc->inode);
- if (!fd) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_CREATE_ANONYMOUS_FD_FAILED, "failed to create "
- "anonymous fd for the inode %s",
- uuid_utoa (loc->inode->gfid));
- goto unwind;
- }
-
- ret = br_stub_need_versioning (this, fd, &inc_version, &modified, &ctx);
- if (ret)
- goto cleanup_fd;
-
- ret = br_stub_check_bad_object (this, fd->inode, &op_ret, &op_errno);
- if (ret)
- goto unwind;
-
- if (!inc_version && modified)
- goto wind;
-
- ret = br_stub_versioning_prep (frame, this, fd, ctx);
- if (ret)
- goto cleanup_fd;
-
- local = frame->local;
- if (!inc_version) {
- br_stub_fill_local (local, NULL, fd, fd->inode,
- fd->inode->gfid, BR_STUB_NO_VERSIONING, 0);
- cbk = br_stub_truncate_cbk;
- goto wind;
- }
-
- stub = fop_truncate_stub (frame, br_stub_truncate_resume, loc, offset,
- xdata);
- if (!stub) {
- gf_msg (this->name, GF_LOG_ERROR, 0, BRS_MSG_STUB_ALLOC_FAILED,
- "failed to allocate stub for truncate fop (gfid: %s), "
- "unwinding", uuid_utoa (fd->inode->gfid));
- goto cleanup_local;
- }
-
- return br_stub_perform_incversioning (this, frame, stub, fd, ctx);
+br_stub_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
+{
+ br_stub_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ gf_boolean_t inc_version = _gf_false;
+ gf_boolean_t modified = _gf_false;
+ br_stub_inode_ctx_t *ctx = NULL;
+ int32_t ret = -1;
+ fd_t *fd = NULL;
+ fop_truncate_cbk_t cbk = default_truncate_cbk;
+ br_stub_private_t *priv = NULL;
+
+ GF_VALIDATE_OR_GOTO("bit-rot-stub", this, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, frame, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, loc, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, unwind);
+
+ priv = this->private;
+ if (!priv->do_versioning)
+ goto wind;
+
+ fd = fd_anonymous(loc->inode);
+ if (!fd) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_CREATE_ANONYMOUS_FD_FAILED,
+ "inode-gfid=%s", uuid_utoa(loc->inode->gfid), NULL);
+ goto unwind;
+ }
+
+ ret = br_stub_need_versioning(this, fd, &inc_version, &modified, &ctx);
+ if (ret)
+ goto cleanup_fd;
+
+ ret = br_stub_check_bad_object(this, fd->inode, &op_ret, &op_errno);
+ if (ret)
+ goto unwind;
+
+ if (!inc_version && modified)
+ goto wind;
+
+ ret = br_stub_versioning_prep(frame, this, fd, ctx);
+ if (ret)
+ goto cleanup_fd;
+
+ local = frame->local;
+ if (!inc_version) {
+ br_stub_fill_local(local, NULL, fd, fd->inode, fd->inode->gfid,
+ BR_STUB_NO_VERSIONING, 0);
+ cbk = br_stub_truncate_cbk;
+ goto wind;
+ }
+
+ stub = fop_truncate_stub(frame, br_stub_truncate_resume, loc, offset,
+ xdata);
+ if (!stub) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_STUB_ALLOC_FAILED,
+ "truncate gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto cleanup_local;
+ }
- wind:
- STACK_WIND (frame, cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
- if (fd)
- fd_unref (fd);
- return 0;
+ return br_stub_perform_incversioning(this, frame, stub, fd, ctx);
- cleanup_local:
- br_stub_cleanup_local (local);
- br_stub_dealloc_local (local);
- cleanup_fd:
- fd_unref (fd);
- unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, NULL, NULL,
- NULL);
+wind:
+ STACK_WIND(frame, cbk, FIRST_CHILD(this), FIRST_CHILD(this)->fops->truncate,
+ loc, offset, xdata);
+ if (fd)
+ fd_unref(fd);
+ return 0;
+
+cleanup_local:
+ br_stub_cleanup_local(local);
+ br_stub_dealloc_local(local);
+cleanup_fd:
+ fd_unref(fd);
+unwind:
+ frame->local = NULL;
+ STACK_UNWIND_STRICT(truncate, frame, op_ret, op_errno, NULL, NULL, NULL);
- return 0;
+ return 0;
}
/** }}} */
-
/** {{{ */
/* open() */
@@ -2359,76 +2397,70 @@ br_stub_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc,
*/
int
-br_stub_open (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, fd_t *fd, dict_t *xdata)
-{
- int32_t ret = -1;
- br_stub_inode_ctx_t *ctx = NULL;
- uint64_t ctx_addr = 0;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- br_stub_private_t *priv = NULL;
- unsigned long version = BITROT_DEFAULT_CURRENT_VERSION;
-
- GF_VALIDATE_OR_GOTO ("bit-rot-stub", this, unwind);
- GF_VALIDATE_OR_GOTO (this->name, this->private, unwind);
- GF_VALIDATE_OR_GOTO (this->name, loc, unwind);
- GF_VALIDATE_OR_GOTO (this->name, fd, unwind);
- GF_VALIDATE_OR_GOTO (this->name, fd->inode, unwind);
-
- priv = this->private;
-
- if (!priv->do_versioning)
- goto wind;
-
- ret = br_stub_get_inode_ctx (this, fd->inode, &ctx_addr);
+br_stub_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
+{
+ int32_t ret = -1;
+ br_stub_inode_ctx_t *ctx = NULL;
+ uint64_t ctx_addr = 0;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ br_stub_private_t *priv = NULL;
+ unsigned long version = BITROT_DEFAULT_CURRENT_VERSION;
+
+ GF_VALIDATE_OR_GOTO("bit-rot-stub", this, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, loc, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, fd, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, fd->inode, unwind);
+
+ priv = this->private;
+
+ if (!priv->do_versioning)
+ goto wind;
+
+ ret = br_stub_get_inode_ctx(this, fd->inode, &ctx_addr);
+ if (ret) {
+ ret = br_stub_init_inode_versions(this, fd, fd->inode, version,
+ _gf_true, _gf_false, &ctx_addr);
if (ret) {
- ret = br_stub_init_inode_versions (this, fd, fd->inode, version,
- _gf_true, _gf_false,
- &ctx_addr);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_GET_INODE_CONTEXT_FAILED,
- "failed to init the inode context for "
- "the file %s (gfid: %s)", loc->path,
- uuid_utoa (fd->inode->gfid));
- goto unwind;
- }
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ BRS_MSG_GET_INODE_CONTEXT_FAILED, "path=%s", loc->path,
+ "gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto unwind;
}
+ }
- ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
+ ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
- ret = br_stub_check_bad_object (this, fd->inode, &op_ret, &op_errno);
- if (ret)
- goto unwind;
+ ret = br_stub_check_bad_object(this, fd->inode, &op_ret, &op_errno);
+ if (ret)
+ goto unwind;
- if (frame->root->pid == GF_CLIENT_PID_SCRUB)
- goto wind;
+ if (frame->root->pid == GF_CLIENT_PID_SCRUB)
+ goto wind;
- if (flags == O_RDONLY)
- goto wind;
+ if (flags == O_RDONLY)
+ goto wind;
- ret = br_stub_add_fd_to_inode (this, fd, ctx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_ADD_FD_TO_LIST_FAILED,
- "failed add fd to the list (gfid: %s)",
- uuid_utoa (fd->inode->gfid));
- goto unwind;
- }
+ ret = br_stub_add_fd_to_inode(this, fd, ctx);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_ADD_FD_TO_LIST_FAILED,
+ "gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto unwind;
+ }
wind:
- STACK_WIND (frame, default_open_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->open, loc, flags, fd, xdata);
- return 0;
+ STACK_WIND(frame, default_open_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
+ return 0;
unwind:
- STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(open, frame, op_ret, op_errno, NULL, NULL);
+ return 0;
}
/** }}} */
-
/** {{{ */
/* creat() */
@@ -2438,143 +2470,137 @@ unwind:
* fd to the inode context fd tracking list.
*/
int32_t
-br_stub_add_fd_to_inode (xlator_t *this, fd_t *fd, br_stub_inode_ctx_t *ctx)
+br_stub_add_fd_to_inode(xlator_t *this, fd_t *fd, br_stub_inode_ctx_t *ctx)
{
- int32_t ret = -1;
- br_stub_fd_t *br_stub_fd = NULL;
+ int32_t ret = -1;
+ br_stub_fd_t *br_stub_fd = NULL;
- ret = br_stub_require_release_call (this, fd, &br_stub_fd);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_SET_FD_CONTEXT_FAILED, "failed to set the fd "
- "context for the file (gfid: %s)",
- uuid_utoa (fd->inode->gfid));
- goto out;
- }
+ ret = br_stub_require_release_call(this, fd, &br_stub_fd);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_SET_FD_CONTEXT_FAILED,
+ "gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto out;
+ }
- LOCK (&fd->inode->lock);
- {
- list_add_tail (&ctx->fd_list, &br_stub_fd->list);
- }
- UNLOCK (&fd->inode->lock);
+ LOCK(&fd->inode->lock);
+ {
+ list_add_tail(&ctx->fd_list, &br_stub_fd->list);
+ }
+ UNLOCK(&fd->inode->lock);
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-br_stub_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, fd_t *fd, inode_t *inode,
- struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- int32_t ret = 0;
- uint64_t ctx_addr = 0;
- br_stub_inode_ctx_t *ctx = NULL;
- unsigned long version = BITROT_DEFAULT_CURRENT_VERSION;
- br_stub_private_t *priv = NULL;
-
- priv = this->private;
-
- if (op_ret < 0)
- goto unwind;
-
- if (!priv->do_versioning)
- goto unwind;
-
- ret = br_stub_get_inode_ctx (this, fd->inode, &ctx_addr);
- if (ret < 0) {
- ret = br_stub_init_inode_versions (this, fd, inode, version,
- _gf_true, _gf_false,
- &ctx_addr);
- if (ret) {
- op_ret = -1;
- op_errno = EINVAL;
- }
- } else {
- ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
- ret = br_stub_add_fd_to_inode (this, fd, ctx);
+br_stub_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, fd_t *fd, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ int32_t ret = 0;
+ uint64_t ctx_addr = 0;
+ br_stub_inode_ctx_t *ctx = NULL;
+ unsigned long version = BITROT_DEFAULT_CURRENT_VERSION;
+ br_stub_private_t *priv = NULL;
+
+ priv = this->private;
+
+ if (op_ret < 0)
+ goto unwind;
+
+ if (!priv->do_versioning)
+ goto unwind;
+
+ ret = br_stub_get_inode_ctx(this, fd->inode, &ctx_addr);
+ if (ret < 0) {
+ ret = br_stub_init_inode_versions(this, fd, inode, version, _gf_true,
+ _gf_false, &ctx_addr);
+ if (ret) {
+ op_ret = -1;
+ op_errno = EINVAL;
}
+ } else {
+ ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
+ ret = br_stub_add_fd_to_inode(this, fd, ctx);
+ }
unwind:
- STACK_UNWIND_STRICT (create, frame, op_ret, op_errno,
- fd, inode, stbuf, preparent, postparent, xdata);
- return 0;
+ STACK_UNWIND_STRICT(create, frame, op_ret, op_errno, fd, inode, stbuf,
+ preparent, postparent, xdata);
+ return 0;
}
int
-br_stub_create (call_frame_t *frame,
- xlator_t *this, loc_t *loc, int32_t flags,
- mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
-{
- GF_VALIDATE_OR_GOTO ("bit-rot-stub", this, unwind);
- GF_VALIDATE_OR_GOTO (this->name, loc, unwind);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, unwind);
- GF_VALIDATE_OR_GOTO (this->name, fd, unwind);
- GF_VALIDATE_OR_GOTO (this->name, fd->inode, unwind);
-
- STACK_WIND (frame, br_stub_create_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->create,
- loc, flags, mode, umask, fd, xdata);
- return 0;
+br_stub_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+{
+ GF_VALIDATE_OR_GOTO("bit-rot-stub", this, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, loc, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, fd, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, fd->inode, unwind);
+
+ STACK_WIND(frame, br_stub_create_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, flags, mode, umask, fd,
+ xdata);
+ return 0;
unwind:
- STACK_UNWIND_STRICT (create, frame, -1, EINVAL, NULL, NULL, NULL, NULL,
- NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(create, frame, -1, EINVAL, NULL, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
int
-br_stub_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode,
- struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- int32_t ret = -1;
- unsigned long version = BITROT_DEFAULT_CURRENT_VERSION;
- br_stub_private_t *priv = NULL;
-
- priv = this->private;
-
- if (op_ret < 0)
- goto unwind;
-
- if (!priv->do_versioning)
- goto unwind;
-
- ret = br_stub_init_inode_versions (this, NULL, inode, version,
- _gf_true, _gf_false, NULL);
- /**
- * Like lookup, if init_inode_versions fail, return EINVAL
- */
- if (ret) {
- op_ret = -1;
- op_errno = EINVAL;
- }
+br_stub_mknod_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, inode_t *inode, struct iatt *stbuf,
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
+{
+ int32_t ret = -1;
+ unsigned long version = BITROT_DEFAULT_CURRENT_VERSION;
+ br_stub_private_t *priv = NULL;
+
+ priv = this->private;
+
+ if (op_ret < 0)
+ goto unwind;
+
+ if (!priv->do_versioning)
+ goto unwind;
+
+ ret = br_stub_init_inode_versions(this, NULL, inode, version, _gf_true,
+ _gf_false, NULL);
+ /**
+ * Like lookup, if init_inode_versions fail, return EINVAL
+ */
+ if (ret) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ }
unwind:
- STACK_UNWIND_STRICT (mknod, frame, op_ret, op_errno,
- inode, stbuf, preparent, postparent, xdata);
- return 0;
+ STACK_UNWIND_STRICT(mknod, frame, op_ret, op_errno, inode, stbuf, preparent,
+ postparent, xdata);
+ return 0;
}
int
-br_stub_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t dev, mode_t umask, dict_t *xdata)
+br_stub_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t dev, mode_t umask, dict_t *xdata)
{
- GF_VALIDATE_OR_GOTO ("bit-rot-stub", this, unwind);
- GF_VALIDATE_OR_GOTO (this->name, loc, unwind);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, unwind);
+ GF_VALIDATE_OR_GOTO("bit-rot-stub", this, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, loc, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, unwind);
- STACK_WIND (frame, br_stub_mknod_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->mknod,
- loc, mode, dev, umask, xdata);
- return 0;
+ STACK_WIND(frame, br_stub_mknod_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, loc, mode, dev, umask, xdata);
+ return 0;
unwind:
- STACK_UNWIND_STRICT (mknod, frame, -1, EINVAL, NULL, NULL, NULL,
- NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(mknod, frame, -1, EINVAL, NULL, NULL, NULL, NULL, NULL);
+ return 0;
}
/** }}} */
@@ -2604,226 +2630,245 @@ unwind:
* creattion failure as the lookup failure.
*/
static int32_t
-br_stub_lookup_version (xlator_t *this,
- uuid_t gfid, inode_t *inode, dict_t *xattr)
-{
- unsigned long version = 0;
- br_version_t *obuf = NULL;
- br_signature_t *sbuf = NULL;
- br_vxattr_status_t status;
- gf_boolean_t bad_object = _gf_false;
-
- /**
- * versioning xattrs were requested from POSIX. if available, figure
- * out the correct version to use in the inode context (start with
- * the default version if unavailable). As of now versions are not
- * persisted on-disk. The inode is marked dirty, so that the first
- * operation (such as write(), etc..) triggers synchronization to
- * disk.
- */
- status = br_version_xattr_state (xattr, &obuf, &sbuf, &bad_object);
- version = ((status == BR_VXATTR_STATUS_FULL)
- || (status == BR_VXATTR_STATUS_UNSIGNED))
- ? obuf->ongoingversion : BITROT_DEFAULT_CURRENT_VERSION;
-
- /**
- * If signature is there, but version is not therem then that status is
- * is treated as INVALID. So in that case, we should not initialize the
- * inode context with wrong version names etc.
- */
- if (status == BR_VXATTR_STATUS_INVALID)
- return -1;
+br_stub_lookup_version(xlator_t *this, uuid_t gfid, inode_t *inode,
+ dict_t *xattr)
+{
+ unsigned long version = 0;
+ br_version_t *obuf = NULL;
+ br_signature_t *sbuf = NULL;
+ br_vxattr_status_t status;
+ gf_boolean_t bad_object = _gf_false;
+
+ /**
+ * versioning xattrs were requested from POSIX. if available, figure
+ * out the correct version to use in the inode context (start with
+ * the default version if unavailable). As of now versions are not
+ * persisted on-disk. The inode is marked dirty, so that the first
+ * operation (such as write(), etc..) triggers synchronization to
+ * disk.
+ */
+ status = br_version_xattr_state(xattr, &obuf, &sbuf, &bad_object);
+ version = ((status == BR_VXATTR_STATUS_FULL) ||
+ (status == BR_VXATTR_STATUS_UNSIGNED))
+ ? obuf->ongoingversion
+ : BITROT_DEFAULT_CURRENT_VERSION;
+
+ /**
+ * If signature is there, but version is not there then that status is
+ * is treated as INVALID. So in that case, we should not initialize the
+ * inode context with wrong version names etc.
+ */
+ if (status == BR_VXATTR_STATUS_INVALID)
+ return -1;
- return br_stub_init_inode_versions (this, NULL, inode, version,
- _gf_true, bad_object, NULL);
+ return br_stub_init_inode_versions(this, NULL, inode, version, _gf_true,
+ bad_object, NULL);
}
-
/** {{{ */
int32_t
-br_stub_opendir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, fd_t *fd, dict_t *xdata)
-{
- br_stub_private_t *priv = NULL;
- br_stub_fd_t *fd_ctx = NULL;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
-
- priv = this->private;
- if (gf_uuid_compare (fd->inode->gfid, priv->bad_object_dir_gfid))
- goto normal;
-
- fd_ctx = br_stub_fd_new ();
- if (!fd_ctx) {
- op_errno = ENOMEM;
- goto unwind;
- }
+br_stub_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata)
+{
+ br_stub_private_t *priv = NULL;
+ br_stub_fd_t *fd_ctx = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
- fd_ctx->bad_object.dir_eof = -1;
- fd_ctx->bad_object.dir = sys_opendir (priv->stub_basepath);
- if (!fd_ctx->bad_object.dir) {
- op_errno = errno;
- goto err_freectx;
- }
+ priv = this->private;
+ if (gf_uuid_compare(fd->inode->gfid, priv->bad_object_dir_gfid))
+ goto normal;
+
+ fd_ctx = br_stub_fd_new();
+ if (!fd_ctx) {
+ op_errno = ENOMEM;
+ goto unwind;
+ }
- op_ret = br_stub_fd_ctx_set (this, fd, fd_ctx);
- if (!op_ret)
- goto unwind;
+ fd_ctx->bad_object.dir_eof = -1;
+ fd_ctx->bad_object.dir = sys_opendir(priv->stub_basepath);
+ if (!fd_ctx->bad_object.dir) {
+ op_errno = errno;
+ goto err_freectx;
+ }
- sys_closedir (fd_ctx->bad_object.dir);
+ op_ret = br_stub_fd_ctx_set(this, fd, fd_ctx);
+ if (!op_ret)
+ goto unwind;
+
+ sys_closedir(fd_ctx->bad_object.dir);
err_freectx:
- GF_FREE (fd_ctx);
+ GF_FREE(fd_ctx);
unwind:
- STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd, NULL);
- return 0;
+ STACK_UNWIND_STRICT(opendir, frame, op_ret, op_errno, fd, NULL);
+ return 0;
normal:
- STACK_WIND (frame, default_opendir_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->opendir, loc, fd, xdata);
- return 0;
+ STACK_WIND(frame, default_opendir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->opendir, loc, fd, xdata);
+ return 0;
}
int32_t
-br_stub_readdir (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t off, dict_t *xdata)
-{
- call_stub_t *stub = NULL;
- br_stub_private_t *priv = NULL;
-
- priv = this->private;
- if (!priv->do_versioning)
- goto out;
-
- if (gf_uuid_compare (fd->inode->gfid, priv->bad_object_dir_gfid))
- goto out;
- stub = fop_readdir_stub (frame, br_stub_readdir_wrapper, fd, size, off,
- xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (readdir, frame, -1, ENOMEM, NULL, NULL);
- return 0;
- }
- br_stub_worker_enqueue (this, stub);
- return 0;
+br_stub_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *xdata)
+{
+ call_stub_t *stub = NULL;
+ br_stub_private_t *priv = NULL;
+
+ priv = this->private;
+ if (!priv->do_versioning)
+ goto out;
+
+ if (gf_uuid_compare(fd->inode->gfid, priv->bad_object_dir_gfid))
+ goto out;
+ stub = fop_readdir_stub(frame, br_stub_readdir_wrapper, fd, size, off,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(readdir, frame, -1, ENOMEM, NULL, NULL);
+ return 0;
+ }
+ br_stub_worker_enqueue(this, stub);
+ return 0;
out:
- STACK_WIND (frame, default_readdir_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdir, fd, size, off, xdata);
- return 0;
+ STACK_WIND(frame, default_readdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdir, fd, size, off, xdata);
+ return 0;
}
int
-br_stub_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, gf_dirent_t *entries,
- dict_t *dict)
-{
- int32_t ret = 0;
- uint64_t ctxaddr = 0;
- gf_dirent_t *entry = NULL;
- br_stub_private_t *priv = NULL;
- gf_boolean_t ver_enabled = _gf_false;
-
- BR_STUB_VER_ENABLED_IN_CALLPATH (frame, ver_enabled);
- priv = this->private;
- BR_STUB_VER_COND_GOTO (priv, (!ver_enabled), unwind);
-
- if (op_ret < 0)
- goto unwind;
-
- list_for_each_entry (entry, &entries->list, list) {
- if ((strcmp (entry->d_name, ".") == 0)
- || (strcmp (entry->d_name, "..") == 0))
- continue;
-
- if (!IA_ISREG (entry->d_stat.ia_type))
- continue;
-
- ret = br_stub_get_inode_ctx (this, entry->inode, &ctxaddr);
- if (ret < 0)
- ctxaddr = 0;
- if (ctxaddr) { /* already has the context */
- br_stub_remove_vxattrs (entry->dict);
- continue;
- }
-
- ret = br_stub_lookup_version
- (this, entry->inode->gfid, entry->inode, entry->dict);
- br_stub_remove_vxattrs (entry->dict);
- if (ret) {
- /**
- * there's no per-file granularity support in case of
- * failure. let's fail the entire request for now..
- */
- break;
- }
+br_stub_readdirp_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, gf_dirent_t *entries,
+ dict_t *dict)
+{
+ int32_t ret = 0;
+ uint64_t ctxaddr = 0;
+ gf_dirent_t *entry = NULL;
+ br_stub_private_t *priv = NULL;
+ gf_boolean_t ver_enabled = _gf_false;
+
+ BR_STUB_VER_ENABLED_IN_CALLPATH(frame, ver_enabled);
+ priv = this->private;
+ BR_STUB_VER_COND_GOTO(priv, (!ver_enabled), unwind);
+
+ if (op_ret < 0)
+ goto unwind;
+
+ list_for_each_entry(entry, &entries->list, list)
+ {
+ if ((strcmp(entry->d_name, ".") == 0) ||
+ (strcmp(entry->d_name, "..") == 0))
+ continue;
+
+ if (!IA_ISREG(entry->d_stat.ia_type))
+ continue;
+
+ /*
+ * Readdirp for most part is a bulk lookup for all the entries
+ * present in the directory being read. Ideally, for each
+ * entry, the handling should be similar to that of a lookup
+ * callback. But for now, just keeping this as it has been
+ * until now (which means, this comment has been added much
+ * later as part of a change that wanted to send the flag
+ * of true/false to br_stub_remove_vxattrs to indicate whether
+ * the bad-object xattr should be removed from the entry->dict
+ * or not). Until this change, the function br_stub_remove_vxattrs
+ * was just removing all the xattrs associated with bit-rot-stub
+ * (like version, bad-object, signature etc). But, there are
+ * scenarios where we only want to send bad-object xattr and not
+ * others. So this comment is part of that change which also
+ * mentions about another possible change that might be needed
+ * in future.
+ * But for now, adding _gf_true means functionally its same as
+ * what this function was doing before. Just remove all the stub
+ * related xattrs.
+ */
+ ret = br_stub_get_inode_ctx(this, entry->inode, &ctxaddr);
+ if (ret < 0)
+ ctxaddr = 0;
+ if (ctxaddr) { /* already has the context */
+ br_stub_remove_vxattrs(entry->dict, _gf_true);
+ continue;
}
+ ret = br_stub_lookup_version(this, entry->inode->gfid, entry->inode,
+ entry->dict);
+ br_stub_remove_vxattrs(entry->dict, _gf_true);
if (ret) {
- op_ret = -1;
- op_errno = EINVAL;
+ /**
+ * there's no per-file granularity support in case of
+ * failure. let's fail the entire request for now..
+ */
+ break;
}
+ }
- unwind:
- STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries, dict);
+ if (ret) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ }
- return 0;
+unwind:
+ STACK_UNWIND_STRICT(readdirp, frame, op_ret, op_errno, entries, dict);
+
+ return 0;
}
int
-br_stub_readdirp (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset, dict_t *dict)
+br_stub_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *dict)
{
- int32_t ret = -1;
- int op_errno = 0;
- gf_boolean_t xref = _gf_false;
- br_stub_private_t *priv = NULL;
-
- priv = this->private;
- BR_STUB_VER_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
-
- op_errno = ENOMEM;
- if (!dict) {
- dict = dict_new ();
- if (!dict)
- goto unwind;
- } else {
- dict = dict_ref (dict);
- }
+ int32_t ret = -1;
+ int op_errno = 0;
+ gf_boolean_t xref = _gf_false;
+ br_stub_private_t *priv = NULL;
- xref = _gf_true;
+ priv = this->private;
+ BR_STUB_VER_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
- op_errno = EINVAL;
- ret = dict_set_uint32 (dict, BITROT_CURRENT_VERSION_KEY, 0);
- if (ret)
- goto unwind;
- ret = dict_set_uint32 (dict, BITROT_SIGNING_VERSION_KEY, 0);
- if (ret)
- goto unwind;
- ret = dict_set_uint32 (dict, BITROT_OBJECT_BAD_KEY, 0);
- if (ret)
- goto unwind;
+ op_errno = ENOMEM;
+ if (!dict) {
+ dict = dict_new();
+ if (!dict)
+ goto unwind;
+ } else {
+ dict = dict_ref(dict);
+ }
+
+ xref = _gf_true;
+
+ op_errno = EINVAL;
+ ret = dict_set_uint32(dict, BITROT_CURRENT_VERSION_KEY, 0);
+ if (ret)
+ goto unwind;
+ ret = dict_set_uint32(dict, BITROT_SIGNING_VERSION_KEY, 0);
+ if (ret)
+ goto unwind;
+ ret = dict_set_uint32(dict, BITROT_OBJECT_BAD_KEY, 0);
+ if (ret)
+ goto unwind;
wind:
- STACK_WIND (frame, br_stub_readdirp_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->readdirp, fd, size,
- offset, dict);
- goto unref_dict;
+ STACK_WIND(frame, br_stub_readdirp_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdirp, fd, size, offset, dict);
+ goto unref_dict;
unwind:
- if (frame->local == (void *)0x1)
- frame->local = NULL;
- STACK_UNWIND_STRICT (readdirp, frame, -1, op_errno, NULL, NULL);
- return 0;
+ if (frame->local == (void *)0x1)
+ frame->local = NULL;
+ STACK_UNWIND_STRICT(readdirp, frame, -1, op_errno, NULL, NULL);
+ return 0;
unref_dict:
- if (xref)
- dict_unref (dict);
- return 0;
+ if (xref)
+ dict_unref(dict);
+ return 0;
}
/** }}} */
-
/** {{{ */
/* lookup() */
@@ -2837,203 +2882,213 @@ unref_dict:
* either forget () or lookup () will take care of removing the link.
*/
void
-br_stub_handle_lookup_error (xlator_t *this, inode_t *inode, int32_t op_errno)
+br_stub_handle_lookup_error(xlator_t *this, inode_t *inode, int32_t op_errno)
{
- int32_t ret = -1;
- uint64_t ctx_addr = 0;
- br_stub_inode_ctx_t *ctx = NULL;
+ int32_t ret = -1;
+ uint64_t ctx_addr = 0;
+ br_stub_inode_ctx_t *ctx = NULL;
- if (op_errno != ENOENT)
- goto out;
+ if (op_errno != ENOENT)
+ goto out;
- if (!inode_is_linked (inode))
- goto out;
+ if (!inode_is_linked(inode))
+ goto out;
- ret = br_stub_get_inode_ctx (this, inode, &ctx_addr);
- if (ret)
- goto out;
+ ret = br_stub_get_inode_ctx(this, inode, &ctx_addr);
+ if (ret)
+ goto out;
- ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
+ ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
- LOCK (&inode->lock);
- {
- if (__br_stub_is_bad_object (ctx))
- (void) br_stub_del (this, inode->gfid);
- }
- UNLOCK (&inode->lock);
-
- if (__br_stub_is_bad_object (ctx)) {
- /* File is not present, might be deleted for recovery,
- * del the bitrot inode context
- */
- ctx_addr = 0;
- inode_ctx_del (inode, this, &ctx_addr);
- if (ctx_addr) {
- ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
- GF_FREE (ctx);
- }
+ LOCK(&inode->lock);
+ {
+ if (__br_stub_is_bad_object(ctx))
+ (void)br_stub_del(this, inode->gfid);
+ }
+ UNLOCK(&inode->lock);
+
+ if (__br_stub_is_bad_object(ctx)) {
+ /* File is not present, might be deleted for recovery,
+ * del the bitrot inode context
+ */
+ ctx_addr = 0;
+ inode_ctx_del(inode, this, &ctx_addr);
+ if (ctx_addr) {
+ ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
+ GF_FREE(ctx);
}
+ }
out:
- return;
+ return;
}
int
-br_stub_lookup_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno, inode_t *inode,
- struct iatt *stbuf, dict_t *xattr, struct iatt *postparent)
+br_stub_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, inode_t *inode, struct iatt *stbuf,
+ dict_t *xattr, struct iatt *postparent)
{
- int32_t ret = 0;
- br_stub_private_t *priv = NULL;
- gf_boolean_t ver_enabled = _gf_false;
-
- BR_STUB_VER_ENABLED_IN_CALLPATH(frame, ver_enabled);
- priv = this->private;
-
- if (op_ret < 0) {
- (void) br_stub_handle_lookup_error (this, inode, op_errno);
- goto unwind;
- }
+ int32_t ret = 0;
+ br_stub_private_t *priv = NULL;
+ gf_boolean_t ver_enabled = _gf_false;
+ gf_boolean_t remove_bad_file_marker = _gf_true;
- BR_STUB_VER_COND_GOTO (priv, (!ver_enabled), delkey);
+ BR_STUB_VER_ENABLED_IN_CALLPATH(frame, ver_enabled);
+ priv = this->private;
- if (!IA_ISREG (stbuf->ia_type))
- goto unwind;
+ if (op_ret < 0) {
+ (void)br_stub_handle_lookup_error(this, inode, op_errno);
- /**
- * If the object is bad, then "bad inode" marker has to be sent back
- * in resoinse, for revalidated lookups as well. Some xlators such as
- * quick-read might cache the data in revalidated lookup as fresh
- * lookup would anyway have sent "bad inode" marker.
- * In general send bad inode marker for every lookup operation on the
- * bad object.
+ /*
+ * If the lookup error is not ENOENT, then it is better
+ * to send the bad file marker to the higher layer (if
+ * it has been set)
*/
- if (cookie != (void *) BR_STUB_REQUEST_COOKIE) {
- ret = br_stub_mark_xdata_bad_object (this, inode, xattr);
- if (ret) {
- op_ret = -1;
- op_errno = EIO;
- goto unwind;
- }
- goto delkey;
- }
-
- ret = br_stub_lookup_version (this, stbuf->ia_gfid, inode, xattr);
- if (ret < 0) {
- op_ret = -1;
- op_errno = EINVAL;
- goto delkey;
- }
-
+ if (op_errno != ENOENT)
+ remove_bad_file_marker = _gf_false;
+ goto delkey;
+ }
+
+ BR_STUB_VER_COND_GOTO(priv, (!ver_enabled), delkey);
+
+ if (!IA_ISREG(stbuf->ia_type))
+ goto unwind;
+
+ /**
+ * If the object is bad, then "bad inode" marker has to be sent back
+ * in resoinse, for revalidated lookups as well. Some xlators such as
+ * quick-read might cache the data in revalidated lookup as fresh
+ * lookup would anyway have sent "bad inode" marker.
+ * In general send bad inode marker for every lookup operation on the
+ * bad object.
+ */
+ if (cookie != (void *)BR_STUB_REQUEST_COOKIE) {
+ ret = br_stub_mark_xdata_bad_object(this, inode, xattr);
+ if (ret) {
+ op_ret = -1;
+ op_errno = EIO;
+ /*
+ * This flag ensures that in the label @delkey below,
+ * bad file marker is not removed from the dictinary,
+ * but other virtual xattrs (such as version, signature)
+ * are removed.
+ */
+ remove_bad_file_marker = _gf_false;
+ }
+ goto delkey;
+ }
+
+ ret = br_stub_lookup_version(this, stbuf->ia_gfid, inode, xattr);
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto delkey;
+ }
+
+ /**
+ * If the object is bad, send "bad inode" marker back in response
+ * for xlator(s) to act accordingly (such as quick-read, etc..)
+ */
+ ret = br_stub_mark_xdata_bad_object(this, inode, xattr);
+ if (ret) {
/**
- * If the object is bad, send "bad inode" marker back in response
- * for xlator(s) to act accordingly (such as quick-read, etc..)
+ * aaha! bad object, but sorry we would not
+ * satisfy the request on allocation failures.
*/
- ret = br_stub_mark_xdata_bad_object (this, inode, xattr);
- if (ret) {
- /**
- * aaha! bad object, but sorry we would not
- * satisfy the request on allocation failures.
- */
- op_ret = -1;
- op_errno = EIO;
- goto unwind;
- }
+ op_ret = -1;
+ op_errno = EIO;
+ goto delkey;
+ }
delkey:
- br_stub_remove_vxattrs (xattr);
+ br_stub_remove_vxattrs(xattr, remove_bad_file_marker);
unwind:
- STACK_UNWIND_STRICT (lookup, frame,
- op_ret, op_errno, inode, stbuf, xattr, postparent);
+ STACK_UNWIND_STRICT(lookup, frame, op_ret, op_errno, inode, stbuf, xattr,
+ postparent);
- return 0;
+ return 0;
}
int
-br_stub_lookup (call_frame_t *frame,
- xlator_t *this, loc_t *loc, dict_t *xdata)
-{
- int32_t ret = 0;
- int op_errno = 0;
- void *cookie = NULL;
- uint64_t ctx_addr = 0;
- gf_boolean_t xref = _gf_false;
- br_stub_private_t *priv = NULL;
- call_stub_t *stub = NULL;
-
- GF_VALIDATE_OR_GOTO ("bit-rot-stub", this, unwind);
- GF_VALIDATE_OR_GOTO (this->name, loc, unwind);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, unwind);
-
- priv = this->private;
-
- BR_STUB_VER_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
-
- if (!gf_uuid_compare (loc->gfid, priv->bad_object_dir_gfid) ||
- !gf_uuid_compare (loc->pargfid, priv->bad_object_dir_gfid)) {
-
- stub = fop_lookup_stub (frame, br_stub_lookup_wrapper, loc,
- xdata);
- if (!stub) {
- op_errno = ENOMEM;
- goto unwind;
- }
- br_stub_worker_enqueue (this, stub);
- return 0;
- }
+br_stub_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+ int32_t ret = 0;
+ int op_errno = 0;
+ void *cookie = NULL;
+ uint64_t ctx_addr = 0;
+ gf_boolean_t xref = _gf_false;
+ br_stub_private_t *priv = NULL;
+ call_stub_t *stub = NULL;
- ret = br_stub_get_inode_ctx (this, loc->inode, &ctx_addr);
- if (ret < 0)
- ctx_addr = 0;
- if (ctx_addr != 0)
- goto wind;
+ GF_VALIDATE_OR_GOTO("bit-rot-stub", this, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, loc, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, unwind);
- /**
- * fresh lookup: request version keys from POSIX
- */
- op_errno = ENOMEM;
- if (!xdata) {
- xdata = dict_new ();
- if (!xdata)
- goto unwind;
- } else {
- xdata = dict_ref (xdata);
+ priv = this->private;
+
+ BR_STUB_VER_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
+
+ if (!gf_uuid_compare(loc->gfid, priv->bad_object_dir_gfid) ||
+ !gf_uuid_compare(loc->pargfid, priv->bad_object_dir_gfid)) {
+ stub = fop_lookup_stub(frame, br_stub_lookup_wrapper, loc, xdata);
+ if (!stub) {
+ op_errno = ENOMEM;
+ goto unwind;
}
+ br_stub_worker_enqueue(this, stub);
+ return 0;
+ }
- xref = _gf_true;
+ ret = br_stub_get_inode_ctx(this, loc->inode, &ctx_addr);
+ if (ret < 0)
+ ctx_addr = 0;
+ if (ctx_addr != 0)
+ goto wind;
- /**
- * Requesting both xattrs provides a way of sanity checking the
- * object. Anomaly checking is done in cbk by examining absence
- * of either or both xattrs.
- */
- op_errno = EINVAL;
- ret = dict_set_uint32 (xdata, BITROT_CURRENT_VERSION_KEY, 0);
- if (ret)
- goto unwind;
- ret = dict_set_uint32 (xdata, BITROT_SIGNING_VERSION_KEY, 0);
- if (ret)
- goto unwind;
- ret = dict_set_uint32 (xdata, BITROT_OBJECT_BAD_KEY, 0);
- if (ret)
- goto unwind;
- cookie = (void *) BR_STUB_REQUEST_COOKIE;
+ /**
+ * fresh lookup: request version keys from POSIX
+ */
+ op_errno = ENOMEM;
+ if (!xdata) {
+ xdata = dict_new();
+ if (!xdata)
+ goto unwind;
+ } else {
+ xdata = dict_ref(xdata);
+ }
+
+ xref = _gf_true;
+
+ /**
+ * Requesting both xattrs provides a way of sanity checking the
+ * object. Anomaly checking is done in cbk by examining absence
+ * of either or both xattrs.
+ */
+ op_errno = EINVAL;
+ ret = dict_set_uint32(xdata, BITROT_CURRENT_VERSION_KEY, 0);
+ if (ret)
+ goto unwind;
+ ret = dict_set_uint32(xdata, BITROT_SIGNING_VERSION_KEY, 0);
+ if (ret)
+ goto unwind;
+ ret = dict_set_uint32(xdata, BITROT_OBJECT_BAD_KEY, 0);
+ if (ret)
+ goto unwind;
+ cookie = (void *)BR_STUB_REQUEST_COOKIE;
- wind:
- STACK_WIND_COOKIE (frame, br_stub_lookup_cbk, cookie,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->lookup,
- loc, xdata);
- goto dealloc_dict;
+wind:
+ STACK_WIND_COOKIE(frame, br_stub_lookup_cbk, cookie, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, xdata);
+ goto dealloc_dict;
- unwind:
- if (frame->local == (void *) 0x1)
- frame->local = NULL;
- STACK_UNWIND_STRICT (lookup, frame,
- -1, op_errno, NULL, NULL, NULL, NULL);
- dealloc_dict:
- if (xref)
- dict_unref (xdata);
- return 0;
+unwind:
+ if (frame->local == (void *)0x1)
+ frame->local = NULL;
+ STACK_UNWIND_STRICT(lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+dealloc_dict:
+ if (xref)
+ dict_unref(xdata);
+ return 0;
}
/** }}} */
@@ -3042,64 +3097,64 @@ br_stub_lookup (call_frame_t *frame,
/* stat */
int
-br_stub_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+br_stub_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- int32_t ret = 0;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- br_stub_private_t *priv = NULL;
+ int32_t ret = 0;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ br_stub_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (!priv->do_versioning)
- goto wind;
+ if (!priv->do_versioning)
+ goto wind;
- if (!IA_ISREG (loc->inode->ia_type))
- goto wind;
+ if (!IA_ISREG(loc->inode->ia_type))
+ goto wind;
- ret = br_stub_check_bad_object (this, loc->inode, &op_ret, &op_errno);
- if (ret)
- goto unwind;
+ ret = br_stub_check_bad_object(this, loc->inode, &op_ret, &op_errno);
+ if (ret)
+ goto unwind;
- wind:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->stat, loc, xdata);
- return 0;
+wind:
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->stat,
+ loc, xdata);
+ return 0;
unwind:
- STACK_UNWIND_STRICT (stat, frame, op_ret, op_errno, NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(stat, frame, op_ret, op_errno, NULL, NULL);
+ return 0;
}
/* fstat */
int
-br_stub_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+br_stub_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- int32_t ret = 0;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- br_stub_private_t *priv = NULL;
+ int32_t ret = 0;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ br_stub_private_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (!priv->do_versioning)
- goto wind;
+ if (!priv->do_versioning)
+ goto wind;
- if (!IA_ISREG (fd->inode->ia_type))
- goto wind;
+ if (!IA_ISREG(fd->inode->ia_type))
+ goto wind;
- ret = br_stub_check_bad_object (this, fd->inode, &op_ret, &op_errno);
- if (ret)
- goto unwind;
+ ret = br_stub_check_bad_object(this, fd->inode, &op_ret, &op_errno);
+ if (ret)
+ goto unwind;
- wind:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fstat, fd, xdata);
- return 0;
+wind:
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->fstat,
+ fd, xdata);
+ return 0;
unwind:
- STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(fstat, frame, op_ret, op_errno, NULL, NULL);
+ return 0;
}
/** }}} */
@@ -3109,114 +3164,114 @@ unwind:
/* unlink() */
int
-br_stub_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+br_stub_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *preparent,
struct iatt *postparent, dict_t *xdata)
{
- br_stub_local_t *local = NULL;
- inode_t *inode = NULL;
- uint64_t ctx_addr = 0;
- br_stub_inode_ctx_t *ctx = NULL;
- int32_t ret = -1;
- br_stub_private_t *priv = NULL;
- gf_boolean_t ver_enabled = _gf_false;
-
- BR_STUB_VER_ENABLED_IN_CALLPATH (frame, ver_enabled);
- priv = this->private;
- BR_STUB_VER_COND_GOTO (priv, (!ver_enabled), unwind);
-
- local = frame->local;
- frame->local = NULL;
-
- if (op_ret < 0)
- goto unwind;
-
- inode = local->u.context.inode;
- if (!IA_ISREG (inode->ia_type))
- goto unwind;
-
- ret = br_stub_get_inode_ctx (this, inode, &ctx_addr);
- if (ret) {
- /**
- * If the inode is bad AND context is not there, then there
- * is a possibility of the gfid of the object being listed
- * in the quarantine directory and will be shown in the
- * bad objects list. So continuing with the fop with a
- * warning log. The entry from the quarantine directory
- * has to be removed manually. Its not a good idea to fail
- * the fop, as the object has already been deleted.
- */
- gf_msg (this->name, GF_LOG_WARNING, 0,
- BRS_MSG_GET_INODE_CONTEXT_FAILED,
- "failed to get the context for the inode %s",
- uuid_utoa (inode->gfid));
- goto unwind;
- }
+ br_stub_local_t *local = NULL;
+ inode_t *inode = NULL;
+ uint64_t ctx_addr = 0;
+ br_stub_inode_ctx_t *ctx = NULL;
+ int32_t ret = -1;
+ br_stub_private_t *priv = NULL;
+ gf_boolean_t ver_enabled = _gf_false;
+
+ BR_STUB_VER_ENABLED_IN_CALLPATH(frame, ver_enabled);
+ priv = this->private;
+ BR_STUB_VER_COND_GOTO(priv, (!ver_enabled), unwind);
+
+ local = frame->local;
+ frame->local = NULL;
+
+ if (op_ret < 0)
+ goto unwind;
+
+ if (!local) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_NULL_LOCAL, NULL);
+ goto unwind;
+ }
+ inode = local->u.context.inode;
+ if (!IA_ISREG(inode->ia_type))
+ goto unwind;
+
+ ret = br_stub_get_inode_ctx(this, inode, &ctx_addr);
+ if (ret) {
+ /**
+ * If the inode is bad AND context is not there, then there
+ * is a possibility of the gfid of the object being listed
+ * in the quarantine directory and will be shown in the
+ * bad objects list. So continuing with the fop with a
+ * warning log. The entry from the quarantine directory
+ * has to be removed manually. Its not a good idea to fail
+ * the fop, as the object has already been deleted.
+ */
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_GET_INODE_CONTEXT_FAILED,
+ "inode-gfid=%s", uuid_utoa(inode->gfid), NULL);
+ goto unwind;
+ }
- ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
+ ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
- LOCK (&inode->lock);
- {
- /**
- * Ignoring the return value of br_stub_del ().
- * There is not much that can be done if unlinking
- * of the entry in the quarantine directory fails.
- * The failure is logged.
- */
- if (__br_stub_is_bad_object (ctx))
- (void) br_stub_del (this, inode->gfid);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ /**
+ * Ignoring the return value of br_stub_del ().
+ * There is not much that can be done if unlinking
+ * of the entry in the quarantine directory fails.
+ * The failure is logged.
+ */
+ if (__br_stub_is_bad_object(ctx))
+ (void)br_stub_del(this, inode->gfid);
+ }
+ UNLOCK(&inode->lock);
unwind:
- STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno, preparent,
- postparent, xdata);
- br_stub_cleanup_local (local);
- br_stub_dealloc_local (local);
- return 0;
+ STACK_UNWIND_STRICT(unlink, frame, op_ret, op_errno, preparent, postparent,
+ xdata);
+ br_stub_cleanup_local(local);
+ br_stub_dealloc_local(local);
+ return 0;
}
int
-br_stub_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int flag,
- dict_t *xdata)
+br_stub_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int flag,
+ dict_t *xdata)
{
- br_stub_local_t *local = NULL;
- int32_t op_ret = -1;
- int32_t op_errno = 0;
- br_stub_private_t *priv = NULL;
+ br_stub_local_t *local = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = 0;
+ br_stub_private_t *priv = NULL;
- priv = this->private;
- BR_STUB_VER_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+ priv = this->private;
+ BR_STUB_VER_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
- local = br_stub_alloc_local (this);
- if (!local) {
- op_ret = -1;
- op_errno = ENOMEM;
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM, BRS_MSG_NO_MEMORY,
- "failed to allocate memory for local (path: %s, gfid: %s)",
- loc->path, uuid_utoa (loc->inode->gfid));
- goto unwind;
- }
+ local = br_stub_alloc_local(this);
+ if (!local) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, BRS_MSG_ALLOC_MEM_FAILED,
+ "local path=%s", loc->path, "gfid=%s",
+ uuid_utoa(loc->inode->gfid), NULL);
+ goto unwind;
+ }
- br_stub_fill_local (local, NULL, NULL, loc->inode,
- loc->inode->gfid,
- BR_STUB_NO_VERSIONING, 0);
+ br_stub_fill_local(local, NULL, NULL, loc->inode, loc->inode->gfid,
+ BR_STUB_NO_VERSIONING, 0);
- frame->local = local;
+ frame->local = local;
wind:
- STACK_WIND (frame, br_stub_unlink_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->unlink, loc, flag, xdata);
- return 0;
+ STACK_WIND(frame, br_stub_unlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, flag, xdata);
+ return 0;
unwind:
- if (frame->local == (void *)0x1)
- frame->local = NULL;
- STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno, NULL, NULL, NULL);
- return 0;
+ if (frame->local == (void *)0x1)
+ frame->local = NULL;
+ STACK_UNWIND_STRICT(unlink, frame, op_ret, op_errno, NULL, NULL, NULL);
+ return 0;
}
-
/** }}} */
/** {{{ */
@@ -3224,20 +3279,20 @@ unwind:
/* forget() */
int
-br_stub_forget (xlator_t *this, inode_t *inode)
+br_stub_forget(xlator_t *this, inode_t *inode)
{
- uint64_t ctx_addr = 0;
- br_stub_inode_ctx_t *ctx = NULL;
+ uint64_t ctx_addr = 0;
+ br_stub_inode_ctx_t *ctx = NULL;
- inode_ctx_del (inode, this, &ctx_addr);
- if (!ctx_addr)
- return 0;
+ inode_ctx_del(inode, this, &ctx_addr);
+ if (!ctx_addr)
+ return 0;
- ctx = (br_stub_inode_ctx_t *) (long) ctx_addr;
+ ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
- GF_FREE (ctx);
+ GF_FREE(ctx);
- return 0;
+ return 0;
}
/** }}} */
@@ -3245,60 +3300,58 @@ br_stub_forget (xlator_t *this, inode_t *inode)
/** {{{ */
int32_t
-br_stub_noop (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+br_stub_noop(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
- STACK_DESTROY (frame->root);
- return 0;
+ STACK_DESTROY(frame->root);
+ return 0;
}
static void
-br_stub_send_ipc_fop (xlator_t *this, fd_t *fd, unsigned long releaseversion,
- int sign_info)
-{
- int32_t op = 0;
- int32_t ret = 0;
- dict_t *xdata = NULL;
- call_frame_t *frame = NULL;
- changelog_event_t ev = {0,};
-
- ev.ev_type = CHANGELOG_OP_TYPE_BR_RELEASE;
- ev.u.releasebr.version = releaseversion;
- ev.u.releasebr.sign_info = sign_info;
- gf_uuid_copy (ev.u.releasebr.gfid, fd->inode->gfid);
-
- xdata = dict_new ();
- if (!xdata) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM, BRS_MSG_NO_MEMORY,
- "dict allocation failed: cannot send IPC FOP "
- "to changelog");
- goto out;
- }
-
- ret = dict_set_static_bin (xdata,
- "RELEASE-EVENT", &ev, CHANGELOG_EV_SIZE);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, 0, BRS_MSG_SET_EVENT_FAILED,
- "cannot set release event in dict");
- goto dealloc_dict;
- }
+br_stub_send_ipc_fop(xlator_t *this, fd_t *fd, unsigned long releaseversion,
+ int sign_info)
+{
+ int32_t op = 0;
+ int32_t ret = 0;
+ dict_t *xdata = NULL;
+ call_frame_t *frame = NULL;
+ changelog_event_t ev = {
+ 0,
+ };
+
+ ev.ev_type = CHANGELOG_OP_TYPE_BR_RELEASE;
+ ev.u.releasebr.version = releaseversion;
+ ev.u.releasebr.sign_info = sign_info;
+ gf_uuid_copy(ev.u.releasebr.gfid, fd->inode->gfid);
+
+ xdata = dict_new();
+ if (!xdata) {
+ gf_smsg(this->name, GF_LOG_WARNING, ENOMEM, BRS_MSG_DICT_ALLOC_FAILED,
+ NULL);
+ goto out;
+ }
+
+ ret = dict_set_static_bin(xdata, "RELEASE-EVENT", &ev, CHANGELOG_EV_SIZE);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_SET_EVENT_FAILED, NULL);
+ goto dealloc_dict;
+ }
- frame = create_frame (this, this->ctx->pool);
- if (!frame) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- BRS_MSG_CREATE_FRAME_FAILED,
- "create_frame() failure");
- goto dealloc_dict;
- }
+ frame = create_frame(this, this->ctx->pool);
+ if (!frame) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_CREATE_FRAME_FAILED,
+ NULL);
+ goto dealloc_dict;
+ }
- op = GF_IPC_TARGET_CHANGELOG;
- STACK_WIND (frame, br_stub_noop, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->ipc, op, xdata);
+ op = GF_IPC_TARGET_CHANGELOG;
+ STACK_WIND(frame, br_stub_noop, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ipc, op, xdata);
- dealloc_dict:
- dict_unref (xdata);
- out:
- return;
+dealloc_dict:
+ dict_unref(xdata);
+out:
+ return;
}
/**
@@ -3309,7 +3362,7 @@ br_stub_send_ipc_fop (xlator_t *this, fd_t *fd, unsigned long releaseversion,
* 3) BR_SIGN_QUICK => reopen has happened and this release should trigger sign
* 2 events:
* 1) GF_FOP_RELEASE
- * 2) GF_FOP_WRITE (actually a dummy write fro BitD)
+ * 2) GF_FOP_WRITE (actually a dummy write for BitD)
*
* This is how states are changed based on events:
* EVENT: GF_FOP_RELEASE:
@@ -3322,115 +3375,113 @@ br_stub_send_ipc_fop (xlator_t *this, fd_t *fd, unsigned long releaseversion,
* set state = BR_SIGN_QUICK;
*/
br_sign_state_t
-__br_stub_inode_sign_state (br_stub_inode_ctx_t *ctx,
- glusterfs_fop_t fop, fd_t *fd)
+__br_stub_inode_sign_state(br_stub_inode_ctx_t *ctx, glusterfs_fop_t fop,
+ fd_t *fd)
{
- br_sign_state_t sign_info = BR_SIGN_INVALID;
-
- switch (fop) {
+ br_sign_state_t sign_info = BR_SIGN_INVALID;
+ switch (fop) {
case GF_FOP_FSETXATTR:
- sign_info = ctx->info_sign = BR_SIGN_QUICK;
- break;
+ sign_info = ctx->info_sign = BR_SIGN_QUICK;
+ break;
case GF_FOP_RELEASE:
- GF_ASSERT (ctx->info_sign != BR_SIGN_REOPEN_WAIT);
+ GF_ASSERT(ctx->info_sign != BR_SIGN_REOPEN_WAIT);
- if (ctx->info_sign == BR_SIGN_NORMAL) {
- sign_info = ctx->info_sign = BR_SIGN_REOPEN_WAIT;
- } else {
- sign_info = ctx->info_sign;
- ctx->info_sign = BR_SIGN_NORMAL;
- }
+ if (ctx->info_sign == BR_SIGN_NORMAL) {
+ sign_info = ctx->info_sign = BR_SIGN_REOPEN_WAIT;
+ } else {
+ sign_info = ctx->info_sign;
+ ctx->info_sign = BR_SIGN_NORMAL;
+ }
- break;
+ break;
default:
- break;
- }
+ break;
+ }
- return sign_info;
+ return sign_info;
}
int32_t
-br_stub_release (xlator_t *this, fd_t *fd)
-{
- int32_t ret = 0;
- int32_t flags = 0;
- inode_t *inode = NULL;
- unsigned long releaseversion = 0;
- br_stub_inode_ctx_t *ctx = NULL;
- uint64_t tmp = 0;
- br_stub_fd_t *br_stub_fd = NULL;
- int32_t signinfo = 0;
-
- inode = fd->inode;
-
- LOCK (&inode->lock);
- {
- ctx = __br_stub_get_ongoing_version_ctx (this, inode, NULL);
- if (ctx == NULL)
- goto unblock;
- br_stub_fd = br_stub_fd_ctx_get (this, fd);
- if (br_stub_fd) {
- list_del_init (&br_stub_fd->list);
- }
-
- ret = __br_stub_can_trigger_release
- (inode, ctx, &releaseversion);
- if (!ret)
- goto unblock;
-
- signinfo = __br_stub_inode_sign_state (ctx, GF_FOP_RELEASE, fd);
- signinfo = htonl (signinfo);
-
- /* inode back to initital state: mark dirty */
- if (ctx->info_sign == BR_SIGN_NORMAL) {
- __br_stub_mark_inode_dirty (ctx);
- __br_stub_unset_inode_modified (ctx);
- }
- }
- unblock:
- UNLOCK (&inode->lock);
+br_stub_release(xlator_t *this, fd_t *fd)
+{
+ int32_t ret = 0;
+ int32_t flags = 0;
+ inode_t *inode = NULL;
+ unsigned long releaseversion = 0;
+ br_stub_inode_ctx_t *ctx = NULL;
+ uint64_t tmp = 0;
+ br_stub_fd_t *br_stub_fd = NULL;
+ int32_t signinfo = 0;
+
+ inode = fd->inode;
+
+ LOCK(&inode->lock);
+ {
+ ctx = __br_stub_get_ongoing_version_ctx(this, inode, NULL);
+ if (ctx == NULL)
+ goto unblock;
+ br_stub_fd = br_stub_fd_ctx_get(this, fd);
+ if (br_stub_fd) {
+ list_del_init(&br_stub_fd->list);
+ }
+
+ ret = __br_stub_can_trigger_release(inode, ctx, &releaseversion);
+ if (!ret)
+ goto unblock;
+
+ signinfo = __br_stub_inode_sign_state(ctx, GF_FOP_RELEASE, fd);
+ signinfo = htonl(signinfo);
+
+ /* inode back to initital state: mark dirty */
+ if (ctx->info_sign == BR_SIGN_NORMAL) {
+ __br_stub_mark_inode_dirty(ctx);
+ __br_stub_unset_inode_modified(ctx);
+ }
+ }
+unblock:
+ UNLOCK(&inode->lock);
- if (ret) {
- gf_msg_debug (this->name, 0, "releaseversion: %lu | flags: %d "
- "| signinfo: %d",
- (unsigned long) ntohl (releaseversion), flags,
- ntohl(signinfo));
- br_stub_send_ipc_fop (this, fd, releaseversion, signinfo);
- }
+ if (ret) {
+ gf_msg_debug(this->name, 0,
+ "releaseversion: %lu | flags: %d "
+ "| signinfo: %d",
+ (unsigned long)ntohl(releaseversion), flags,
+ ntohl(signinfo));
+ br_stub_send_ipc_fop(this, fd, releaseversion, signinfo);
+ }
- ret = fd_ctx_del (fd, this, &tmp);
- br_stub_fd = (br_stub_fd_t *)(long)tmp;
+ ret = fd_ctx_del(fd, this, &tmp);
+ br_stub_fd = (br_stub_fd_t *)(long)tmp;
- GF_FREE (br_stub_fd);
+ GF_FREE(br_stub_fd);
- return 0;
+ return 0;
}
int32_t
-br_stub_releasedir (xlator_t *this, fd_t *fd)
+br_stub_releasedir(xlator_t *this, fd_t *fd)
{
- br_stub_fd_t *fctx = NULL;
- uint64_t ctx = 0;
- int ret = 0;
+ br_stub_fd_t *fctx = NULL;
+ uint64_t ctx = 0;
+ int ret = 0;
- ret = fd_ctx_del (fd, this, &ctx);
- if (ret < 0)
- goto out;
-
- fctx = (br_stub_fd_t *) (long) ctx;
- if (fctx->bad_object.dir) {
- ret = sys_closedir (fctx->bad_object.dir);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_BAD_OBJ_DIR_CLOSE_FAIL,
- "closedir error: %s", strerror (errno));
- }
+ ret = fd_ctx_del(fd, this, &ctx);
+ if (ret < 0)
+ goto out;
- GF_FREE (fctx);
+ fctx = (br_stub_fd_t *)(long)ctx;
+ if (fctx->bad_object.dir) {
+ ret = sys_closedir(fctx->bad_object.dir);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_BAD_OBJ_DIR_CLOSE_FAIL,
+ "error=%s", strerror(errno), NULL);
+ }
+
+ GF_FREE(fctx);
out:
- return 0;
+ return 0;
}
/** }}} */
@@ -3440,90 +3491,100 @@ out:
/* ictxmerge */
void
-br_stub_ictxmerge (xlator_t *this, fd_t *fd,
- inode_t *inode, inode_t *linked_inode)
-{
- int32_t ret = 0;
- uint64_t ctxaddr = 0;
- uint64_t lctxaddr = 0;
- br_stub_inode_ctx_t *ctx = NULL;
- br_stub_inode_ctx_t *lctx = NULL;
- br_stub_fd_t *br_stub_fd = NULL;
-
- ret = br_stub_get_inode_ctx (this, inode, &ctxaddr);
+br_stub_ictxmerge(xlator_t *this, fd_t *fd, inode_t *inode,
+ inode_t *linked_inode)
+{
+ int32_t ret = 0;
+ uint64_t ctxaddr = 0;
+ uint64_t lctxaddr = 0;
+ br_stub_inode_ctx_t *ctx = NULL;
+ br_stub_inode_ctx_t *lctx = NULL;
+ br_stub_fd_t *br_stub_fd = NULL;
+
+ ret = br_stub_get_inode_ctx(this, inode, &ctxaddr);
+ if (ret < 0)
+ goto done;
+ ctx = (br_stub_inode_ctx_t *)(uintptr_t)ctxaddr;
+
+ LOCK(&linked_inode->lock);
+ {
+ ret = __br_stub_get_inode_ctx(this, linked_inode, &lctxaddr);
if (ret < 0)
- goto done;
- ctx = (br_stub_inode_ctx_t *) ctxaddr;
+ goto unblock;
+ lctx = (br_stub_inode_ctx_t *)(uintptr_t)lctxaddr;
- LOCK (&linked_inode->lock);
- {
- ret = __br_stub_get_inode_ctx (this, linked_inode, &lctxaddr);
- if (ret < 0)
- goto unblock;
- lctx = (br_stub_inode_ctx_t *) lctxaddr;
-
- GF_ASSERT (list_is_singular (&ctx->fd_list));
- br_stub_fd = list_first_entry (&ctx->fd_list, br_stub_fd_t,
- list);
- if (br_stub_fd) {
- GF_ASSERT (br_stub_fd->fd == fd);
- list_move_tail (&br_stub_fd->list, &lctx->fd_list);
- }
+ GF_ASSERT(list_is_singular(&ctx->fd_list));
+ br_stub_fd = list_first_entry(&ctx->fd_list, br_stub_fd_t, list);
+ if (br_stub_fd) {
+ GF_ASSERT(br_stub_fd->fd == fd);
+ list_move_tail(&br_stub_fd->list, &lctx->fd_list);
}
+ }
unblock:
- UNLOCK (&linked_inode->lock);
+ UNLOCK(&linked_inode->lock);
- done:
- return;
+done:
+ return;
}
/** }}} */
-
struct xlator_fops fops = {
- .lookup = br_stub_lookup,
- .stat = br_stub_stat,
- .fstat = br_stub_fstat,
- .open = br_stub_open,
- .create = br_stub_create,
- .readdirp = br_stub_readdirp,
- .getxattr = br_stub_getxattr,
- .fgetxattr = br_stub_fgetxattr,
- .fsetxattr = br_stub_fsetxattr,
- .writev = br_stub_writev,
- .truncate = br_stub_truncate,
- .ftruncate = br_stub_ftruncate,
- .mknod = br_stub_mknod,
- .readv = br_stub_readv,
- .removexattr = br_stub_removexattr,
- .fremovexattr = br_stub_fremovexattr,
- .setxattr = br_stub_setxattr,
- .opendir = br_stub_opendir,
- .readdir = br_stub_readdir,
- .unlink = br_stub_unlink,
+ .lookup = br_stub_lookup,
+ .stat = br_stub_stat,
+ .fstat = br_stub_fstat,
+ .open = br_stub_open,
+ .create = br_stub_create,
+ .readdirp = br_stub_readdirp,
+ .getxattr = br_stub_getxattr,
+ .fgetxattr = br_stub_fgetxattr,
+ .fsetxattr = br_stub_fsetxattr,
+ .writev = br_stub_writev,
+ .truncate = br_stub_truncate,
+ .ftruncate = br_stub_ftruncate,
+ .mknod = br_stub_mknod,
+ .readv = br_stub_readv,
+ .removexattr = br_stub_removexattr,
+ .fremovexattr = br_stub_fremovexattr,
+ .setxattr = br_stub_setxattr,
+ .opendir = br_stub_opendir,
+ .readdir = br_stub_readdir,
+ .unlink = br_stub_unlink,
};
struct xlator_cbks cbks = {
- .forget = br_stub_forget,
- .release = br_stub_release,
- .ictxmerge = br_stub_ictxmerge,
+ .forget = br_stub_forget,
+ .release = br_stub_release,
+ .ictxmerge = br_stub_ictxmerge,
};
struct volume_options options[] = {
- { .key = {"bitrot"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .op_version = {GD_OP_VERSION_3_7_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_FORCE,
- .tags = {"bitrot"},
- .description = "enable/disable bitrot stub"
- },
- { .key = {"export"},
- .type = GF_OPTION_TYPE_PATH,
- .op_version = {GD_OP_VERSION_3_7_0},
- .tags = {"bitrot"},
- .description = "brick path for versioning",
- .default_value = "{{ brick.path }}"
- },
- { .key = {NULL} },
+ {.key = {"bitrot"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .op_version = {GD_OP_VERSION_3_7_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_FORCE,
+ .tags = {"bitrot"},
+ .description = "enable/disable bitrot stub"},
+ {.key = {"export"},
+ .type = GF_OPTION_TYPE_PATH,
+ .op_version = {GD_OP_VERSION_3_7_0},
+ .tags = {"bitrot"},
+ .description = "brick path for versioning",
+ .default_value = "{{ brick.path }}"},
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .notify = notify,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "bitrot-stub",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/features/bit-rot/src/stub/bit-rot-stub.h b/xlators/features/bit-rot/src/stub/bit-rot-stub.h
index ae4db0fd4f1..edd79a77e4f 100644
--- a/xlators/features/bit-rot/src/stub/bit-rot-stub.h
+++ b/xlators/features/bit-rot/src/stub/bit-rot-stub.h
@@ -1,329 +1,325 @@
- /*
- Copyright (c) 2015 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
+/*
+ Copyright (c) 2015 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#ifndef __BIT_ROT_STUB_H__
#define __BIT_ROT_STUB_H__
-#include "glusterfs.h"
-#include "logging.h"
-#include "dict.h"
-#include "xlator.h"
-#include "defaults.h"
-#include "call-stub.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/dict.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
+#include <glusterfs/call-stub.h>
#include "bit-rot-stub-mem-types.h"
-#include "syscall.h"
-#include "common-utils.h"
+#include <glusterfs/syscall.h>
+#include <glusterfs/common-utils.h>
#include "bit-rot-common.h"
#include "bit-rot-stub-messages.h"
#include "glusterfs3-xdr.h"
-#include "syncop.h"
-#include "syncop-utils.h"
+#include <glusterfs/syncop.h>
+#include <glusterfs/syncop-utils.h>
-#define BAD_OBJECT_THREAD_STACK_SIZE ((size_t)(1024*1024))
+#define BAD_OBJECT_THREAD_STACK_SIZE ((size_t)(1024 * 1024))
#define BR_STUB_DUMP_STR_SIZE 65536
-#define BR_PATH_MAX_EXTRA (PATH_MAX+1024)
-#define BR_PATH_MAX_PLUS (PATH_MAX+2048)
+#define BR_PATH_MAX_EXTRA (PATH_MAX + 1024)
+#define BR_PATH_MAX_PLUS (PATH_MAX + 2048)
/*
* Oops. Spelling mistake. Correcting it
*/
-#define OLD_BR_STUB_QUARANTINE_DIR GF_HIDDEN_PATH"/quanrantine"
-#define BR_STUB_QUARANTINE_DIR GF_HIDDEN_PATH"/quarantine"
+#define OLD_BR_STUB_QUARANTINE_DIR GF_HIDDEN_PATH "/quanrantine"
+#define BR_STUB_QUARANTINE_DIR GF_HIDDEN_PATH "/quarantine"
/* do not reference frame->local in cbk unless initialized.
* Assigned 0x1 marks verisoning flag between call path and
* cbk path.
*/
-#define BR_STUB_VER_NOT_ACTIVE_THEN_GOTO(frame, priv, label) do { \
- if (priv->do_versioning) \
- frame->local = (void *)0x1; \
- else \
- goto label; \
- } while (0)
-
-#define BR_STUB_VER_COND_GOTO(priv, cond, label) do { \
- if (!priv->do_versioning || cond) \
- goto label; \
- } while (0)
-
-#define BR_STUB_VER_ENABLED_IN_CALLPATH(frame, flag) do { \
- if (frame->local) \
- flag = _gf_true; \
- if (frame->local == (void *) 0x1) \
- frame->local = NULL; \
- } while (0)
-
-#define BR_STUB_RESET_LOCAL_NULL(frame) do { \
- if (frame->local == (void *)0x1) \
- frame->local = NULL; \
- } while (0)
-
-typedef int (br_stub_version_cbk) (call_frame_t *, void *,
- xlator_t *, int32_t, int32_t, dict_t *);
+#define BR_STUB_VER_NOT_ACTIVE_THEN_GOTO(frame, priv, label) \
+ do { \
+ if (priv->do_versioning) \
+ frame->local = (void *)0x1; \
+ else \
+ goto label; \
+ } while (0)
+
+#define BR_STUB_VER_COND_GOTO(priv, cond, label) \
+ do { \
+ if (!priv->do_versioning || cond) \
+ goto label; \
+ } while (0)
+
+#define BR_STUB_VER_ENABLED_IN_CALLPATH(frame, flag) \
+ do { \
+ if (frame->local) \
+ flag = _gf_true; \
+ if (frame->local == (void *)0x1) \
+ frame->local = NULL; \
+ } while (0)
+
+#define BR_STUB_RESET_LOCAL_NULL(frame) \
+ do { \
+ if (frame->local == (void *)0x1) \
+ frame->local = NULL; \
+ } while (0)
+
+typedef int(br_stub_version_cbk)(call_frame_t *, void *, xlator_t *, int32_t,
+ int32_t, dict_t *);
typedef struct br_stub_inode_ctx {
- int need_writeback; /* does the inode need
- a writeback to disk? */
- unsigned long currentversion; /* ongoing version */
-
- int info_sign;
- struct list_head fd_list; /* list of open fds or fds participating in
- write operations */
- gf_boolean_t bad_object;
+ int need_writeback; /* does the inode need
+ a writeback to disk? */
+ unsigned long currentversion; /* ongoing version */
+
+ int info_sign;
+ struct list_head fd_list; /* list of open fds or fds participating in
+ write operations */
+ gf_boolean_t bad_object;
} br_stub_inode_ctx_t;
typedef struct br_stub_fd {
- fd_t *fd;
- struct list_head list;
- struct bad_object_dir {
- DIR *dir;
- off_t dir_eof;
- } bad_object;
+ fd_t *fd;
+ struct list_head list;
+ struct bad_object_dir {
+ DIR *dir;
+ off_t dir_eof;
+ } bad_object;
} br_stub_fd_t;
-#define I_DIRTY (1<<0) /* inode needs writeback */
-#define I_MODIFIED (1<<1)
-#define WRITEBACK_DURABLE 1 /* writeback is durable */
+#define I_DIRTY (1 << 0) /* inode needs writeback */
+#define I_MODIFIED (1 << 1)
+#define WRITEBACK_DURABLE 1 /* writeback is durable */
/**
* This could just have been a plain struct without unions and all,
* but we may need additional things in the future.
*/
typedef struct br_stub_local {
- call_stub_t *fopstub; /* stub for original fop */
-
- int versioningtype; /* not much used atm */
-
- union {
- struct br_stub_ctx {
- fd_t *fd;
- uuid_t gfid;
- inode_t *inode;
- unsigned long version;
- } context;
- } u;
+ call_stub_t *fopstub; /* stub for original fop */
+
+ int versioningtype; /* not much used atm */
+
+ union {
+ struct br_stub_ctx {
+ fd_t *fd;
+ uuid_t gfid;
+ inode_t *inode;
+ unsigned long version;
+ } context;
+ } u;
} br_stub_local_t;
#define BR_STUB_NO_VERSIONING (1 << 0)
#define BR_STUB_INCREMENTAL_VERSIONING (1 << 1)
typedef struct br_stub_private {
- gf_boolean_t do_versioning;
+ gf_boolean_t do_versioning;
- uint32_t boot[2];
- char export[PATH_MAX];
+ uint32_t boot[2];
+ char export[PATH_MAX];
- pthread_mutex_t lock;
- pthread_cond_t cond;
+ pthread_mutex_t lock;
+ pthread_cond_t cond;
- struct list_head squeue; /* ordered signing queue */
- pthread_t signth;
- struct bad_objects_container {
- pthread_t thread;
- pthread_mutex_t bad_lock;
- pthread_cond_t bad_cond;
- struct list_head bad_queue;
- } container;
- struct mem_pool *local_pool;
+ struct list_head squeue; /* ordered signing queue */
+ pthread_t signth;
+ struct bad_objects_container {
+ pthread_t thread;
+ pthread_mutex_t bad_lock;
+ pthread_cond_t bad_cond;
+ struct list_head bad_queue;
+ } container;
+ struct mem_pool *local_pool;
- char stub_basepath[BR_PATH_MAX_EXTRA];
+ char stub_basepath[BR_PATH_MAX_EXTRA];
- uuid_t bad_object_dir_gfid;
+ uuid_t bad_object_dir_gfid;
} br_stub_private_t;
br_stub_fd_t *
-br_stub_fd_new (void);
-
+br_stub_fd_new(void);
int
-__br_stub_fd_ctx_set (xlator_t *this, fd_t *fd, br_stub_fd_t *br_stub_fd);
+__br_stub_fd_ctx_set(xlator_t *this, fd_t *fd, br_stub_fd_t *br_stub_fd);
br_stub_fd_t *
-__br_stub_fd_ctx_get (xlator_t *this, fd_t *fd);
+__br_stub_fd_ctx_get(xlator_t *this, fd_t *fd);
br_stub_fd_t *
-br_stub_fd_ctx_get (xlator_t *this, fd_t *fd);
+br_stub_fd_ctx_get(xlator_t *this, fd_t *fd);
int32_t
-br_stub_fd_ctx_set (xlator_t *this, fd_t *fd, br_stub_fd_t *br_stub_fd);
+br_stub_fd_ctx_set(xlator_t *this, fd_t *fd, br_stub_fd_t *br_stub_fd);
static inline gf_boolean_t
-__br_stub_is_bad_object (br_stub_inode_ctx_t *ctx)
+__br_stub_is_bad_object(br_stub_inode_ctx_t *ctx)
{
- return ctx->bad_object;
+ return ctx->bad_object;
}
static inline void
-__br_stub_mark_object_bad (br_stub_inode_ctx_t *ctx)
+__br_stub_mark_object_bad(br_stub_inode_ctx_t *ctx)
{
- ctx->bad_object = _gf_true;
+ ctx->bad_object = _gf_true;
}
/* inode writeback helpers */
static inline void
-__br_stub_mark_inode_dirty (br_stub_inode_ctx_t *ctx)
+__br_stub_mark_inode_dirty(br_stub_inode_ctx_t *ctx)
{
- ctx->need_writeback |= I_DIRTY;
+ ctx->need_writeback |= I_DIRTY;
}
static inline void
-__br_stub_mark_inode_synced (br_stub_inode_ctx_t *ctx)
+__br_stub_mark_inode_synced(br_stub_inode_ctx_t *ctx)
{
- ctx->need_writeback &= ~I_DIRTY;
+ ctx->need_writeback &= ~I_DIRTY;
}
static inline int
-__br_stub_is_inode_dirty (br_stub_inode_ctx_t *ctx)
+__br_stub_is_inode_dirty(br_stub_inode_ctx_t *ctx)
{
- return (ctx->need_writeback & I_DIRTY);
+ return (ctx->need_writeback & I_DIRTY);
}
/* inode mofification markers */
static inline void
-__br_stub_set_inode_modified (br_stub_inode_ctx_t *ctx)
+__br_stub_set_inode_modified(br_stub_inode_ctx_t *ctx)
{
- ctx->need_writeback |= I_MODIFIED;
+ ctx->need_writeback |= I_MODIFIED;
}
static inline void
-__br_stub_unset_inode_modified (br_stub_inode_ctx_t *ctx)
+__br_stub_unset_inode_modified(br_stub_inode_ctx_t *ctx)
{
- ctx->need_writeback &= ~I_MODIFIED;
+ ctx->need_writeback &= ~I_MODIFIED;
}
static inline int
-__br_stub_is_inode_modified (br_stub_inode_ctx_t *ctx)
+__br_stub_is_inode_modified(br_stub_inode_ctx_t *ctx)
{
- return (ctx->need_writeback & I_MODIFIED);
+ return (ctx->need_writeback & I_MODIFIED);
}
-
static inline int
-br_stub_require_release_call (xlator_t *this, fd_t *fd, br_stub_fd_t **fd_ctx)
+br_stub_require_release_call(xlator_t *this, fd_t *fd, br_stub_fd_t **fd_ctx)
{
- int32_t ret = 0;
- br_stub_fd_t *br_stub_fd = NULL;
+ int32_t ret = 0;
+ br_stub_fd_t *br_stub_fd = NULL;
- br_stub_fd = br_stub_fd_new ();
- if (!br_stub_fd)
- return -1;
+ br_stub_fd = br_stub_fd_new();
+ if (!br_stub_fd)
+ return -1;
- br_stub_fd->fd = fd;
- INIT_LIST_HEAD (&br_stub_fd->list);
+ br_stub_fd->fd = fd;
+ INIT_LIST_HEAD(&br_stub_fd->list);
- ret = br_stub_fd_ctx_set (this, fd, br_stub_fd);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- BRS_MSG_SET_CONTEXT_FAILED,
- "could not set fd context (for release callback");
- else
- *fd_ctx = br_stub_fd;
+ ret = br_stub_fd_ctx_set(this, fd, br_stub_fd);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_WARNING, 0, BRS_MSG_SET_CONTEXT_FAILED,
+ NULL);
+ else
+ *fd_ctx = br_stub_fd;
- return ret;
+ return ret;
}
/* get/set inode context helpers */
static inline int
-__br_stub_get_inode_ctx (xlator_t *this,
- inode_t *inode, uint64_t *ctx)
+__br_stub_get_inode_ctx(xlator_t *this, inode_t *inode, uint64_t *ctx)
{
- return __inode_ctx_get (inode, this, ctx);
+ return __inode_ctx_get(inode, this, ctx);
}
static inline int
-br_stub_get_inode_ctx (xlator_t *this,
- inode_t *inode, uint64_t *ctx)
+br_stub_get_inode_ctx(xlator_t *this, inode_t *inode, uint64_t *ctx)
{
- int ret = -1;
+ int ret = -1;
- LOCK (&inode->lock);
- {
- ret = __br_stub_get_inode_ctx (this, inode, ctx);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __br_stub_get_inode_ctx(this, inode, ctx);
+ }
+ UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
static inline int
-br_stub_set_inode_ctx (xlator_t *this,
- inode_t *inode, br_stub_inode_ctx_t *ctx)
+br_stub_set_inode_ctx(xlator_t *this, inode_t *inode, br_stub_inode_ctx_t *ctx)
{
- uint64_t ctx_addr = (uint64_t) ctx;
- return inode_ctx_set (inode, this, &ctx_addr);
+ uint64_t ctx_addr = (uint64_t)(uintptr_t)ctx;
+ return inode_ctx_set(inode, this, &ctx_addr);
}
/* version get/set helpers */
static inline unsigned long
-__br_stub_writeback_version (br_stub_inode_ctx_t *ctx)
+__br_stub_writeback_version(br_stub_inode_ctx_t *ctx)
{
- return (ctx->currentversion + 1);
+ return (ctx->currentversion + 1);
}
static inline void
-__br_stub_set_ongoing_version (br_stub_inode_ctx_t *ctx, unsigned long version)
+__br_stub_set_ongoing_version(br_stub_inode_ctx_t *ctx, unsigned long version)
{
- if (ctx->currentversion < version)
- ctx->currentversion = version;
- else
- gf_msg ("bit-rot-stub", GF_LOG_WARNING, 0,
- BRS_MSG_CHANGE_VERSION_FAILED, "current version: %lu"
- "new version: %lu", ctx->currentversion, version);
+ if (ctx->currentversion < version)
+ ctx->currentversion = version;
+ else
+ gf_smsg("bit-rot-stub", GF_LOG_WARNING, 0,
+ BRS_MSG_CHANGE_VERSION_FAILED, "current version=%lu",
+ ctx->currentversion, "new version=%lu", version, NULL);
}
static inline int
-__br_stub_can_trigger_release (inode_t *inode,
- br_stub_inode_ctx_t *ctx, unsigned long *version)
+__br_stub_can_trigger_release(inode_t *inode, br_stub_inode_ctx_t *ctx,
+ unsigned long *version)
{
- /**
- * If the inode is modified, then it has to be dirty. An inode is
- * marked dirty once version is increased. Its marked as modified
- * when the modification call (write/truncate) which triggered
- * the versioning is successful.
- */
- if (__br_stub_is_inode_modified (ctx)
- && list_empty (&ctx->fd_list)
- && (ctx->info_sign != BR_SIGN_REOPEN_WAIT)) {
-
- GF_ASSERT (__br_stub_is_inode_dirty (ctx) == 0);
+ /**
+ * If the inode is modified, then it has to be dirty. An inode is
+ * marked dirty once version is increased. Its marked as modified
+ * when the modification call (write/truncate) which triggered
+ * the versioning is successful.
+ */
+ if (__br_stub_is_inode_modified(ctx) && list_empty(&ctx->fd_list) &&
+ (ctx->info_sign != BR_SIGN_REOPEN_WAIT)) {
+ GF_ASSERT(__br_stub_is_inode_dirty(ctx) == 0);
- if (version)
- *version = htonl (ctx->currentversion);
- return 1;
- }
+ if (version)
+ *version = htonl(ctx->currentversion);
+ return 1;
+ }
- return 0;
+ return 0;
}
static inline int32_t
-br_stub_get_ongoing_version (xlator_t *this,
- inode_t *inode, unsigned long *version)
+br_stub_get_ongoing_version(xlator_t *this, inode_t *inode,
+ unsigned long *version)
{
- int32_t ret = 0;
- uint64_t ctx_addr = 0;
- br_stub_inode_ctx_t *ctx = NULL;
-
- LOCK (&inode->lock);
- {
- ret = __inode_ctx_get (inode, this, &ctx_addr);
- if (ret < 0)
- goto unblock;
- ctx = (br_stub_inode_ctx_t *) (long) ctx_addr;
- *version = ctx->currentversion;
- }
- unblock:
- UNLOCK (&inode->lock);
-
- return ret;
+ int32_t ret = 0;
+ uint64_t ctx_addr = 0;
+ br_stub_inode_ctx_t *ctx = NULL;
+
+ LOCK(&inode->lock);
+ {
+ ret = __inode_ctx_get(inode, this, &ctx_addr);
+ if (ret < 0)
+ goto unblock;
+ ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
+ *version = ctx->currentversion;
+ }
+unblock:
+ UNLOCK(&inode->lock);
+
+ return ret;
}
/**
@@ -332,45 +328,52 @@ br_stub_get_ongoing_version (xlator_t *this,
* *needs* to be valid in the caller.
*/
static inline br_stub_inode_ctx_t *
-__br_stub_get_ongoing_version_ctx (xlator_t *this,
- inode_t *inode, unsigned long *version)
+__br_stub_get_ongoing_version_ctx(xlator_t *this, inode_t *inode,
+ unsigned long *version)
{
- int32_t ret = 0;
- uint64_t ctx_addr = 0;
- br_stub_inode_ctx_t *ctx = NULL;
-
- ret = __inode_ctx_get (inode, this, &ctx_addr);
- if (ret < 0)
- return NULL;
- ctx = (br_stub_inode_ctx_t *) (long) ctx_addr;
- if (version)
- *version = ctx->currentversion;
-
- return ctx;
+ int32_t ret = 0;
+ uint64_t ctx_addr = 0;
+ br_stub_inode_ctx_t *ctx = NULL;
+
+ ret = __inode_ctx_get(inode, this, &ctx_addr);
+ if (ret < 0)
+ return NULL;
+ ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
+ if (version)
+ *version = ctx->currentversion;
+
+ return ctx;
}
/* filter for xattr fetch */
static inline int
-br_stub_is_internal_xattr (const char *name)
+br_stub_is_internal_xattr(const char *name)
{
- if (name
- && ((strncmp (name, BITROT_CURRENT_VERSION_KEY,
- strlen (BITROT_CURRENT_VERSION_KEY)) == 0)
- || (strncmp (name, BITROT_SIGNING_VERSION_KEY,
- strlen (BITROT_SIGNING_VERSION_KEY)) == 0)))
- return 1;
- return 0;
+ if (name && ((strncmp(name, BITROT_CURRENT_VERSION_KEY,
+ SLEN(BITROT_CURRENT_VERSION_KEY)) == 0) ||
+ (strncmp(name, BITROT_SIGNING_VERSION_KEY,
+ SLEN(BITROT_SIGNING_VERSION_KEY)) == 0)))
+ return 1;
+ return 0;
}
static inline void
-br_stub_remove_vxattrs (dict_t *xattr)
+br_stub_remove_vxattrs(dict_t *xattr, gf_boolean_t remove_bad_marker)
{
- if (xattr) {
- dict_del (xattr, BITROT_OBJECT_BAD_KEY);
- dict_del (xattr, BITROT_CURRENT_VERSION_KEY);
- dict_del (xattr, BITROT_SIGNING_VERSION_KEY);
- dict_del (xattr, BITROT_SIGNING_XATTR_SIZE_KEY);
- }
+ if (xattr) {
+ /*
+ * When a file is corrupted, bad-object should be
+ * set in the dict. But, other info such as version,
+ * signature etc should not be set. Hence the flag
+ * remove_bad_marker. The consumer should know whether
+ * to send the bad-object info in the dict or not.
+ */
+ if (remove_bad_marker)
+ dict_del(xattr, BITROT_OBJECT_BAD_KEY);
+ dict_del(xattr, BITROT_CURRENT_VERSION_KEY);
+ dict_del(xattr, BITROT_SIGNING_VERSION_KEY);
+ dict_del(xattr, BITROT_SIGNING_XATTR_SIZE_KEY);
+ }
}
/**
@@ -384,64 +387,60 @@ br_stub_remove_vxattrs (dict_t *xattr)
* errors can be made into enums.
*/
static inline int
-br_stub_is_bad_object (xlator_t *this, inode_t *inode)
+br_stub_is_bad_object(xlator_t *this, inode_t *inode)
{
- int bad_object = 0;
- gf_boolean_t tmp = _gf_false;
- uint64_t ctx_addr = 0;
- br_stub_inode_ctx_t *ctx = NULL;
- int32_t ret = -1;
-
- ret = br_stub_get_inode_ctx (this, inode, &ctx_addr);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_GET_INODE_CONTEXT_FAILED,
- "failed to get the inode context for the inode %s",
- uuid_utoa (inode->gfid));
- bad_object = -1;
- goto out;
- }
-
- ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
-
- LOCK (&inode->lock);
- {
- tmp = __br_stub_is_bad_object (ctx);
- if (tmp)
- bad_object = -2;
- }
- UNLOCK (&inode->lock);
+ int bad_object = 0;
+ gf_boolean_t tmp = _gf_false;
+ uint64_t ctx_addr = 0;
+ br_stub_inode_ctx_t *ctx = NULL;
+ int32_t ret = -1;
+
+ ret = br_stub_get_inode_ctx(this, inode, &ctx_addr);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_GET_INODE_CONTEXT_FAILED,
+ "inode-gfid=%s", uuid_utoa(inode->gfid), NULL);
+ bad_object = -1;
+ goto out;
+ }
+
+ ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
+
+ LOCK(&inode->lock);
+ {
+ tmp = __br_stub_is_bad_object(ctx);
+ if (tmp)
+ bad_object = -2;
+ }
+ UNLOCK(&inode->lock);
out:
- return bad_object;
+ return bad_object;
}
static inline int32_t
-br_stub_mark_object_bad (xlator_t *this, inode_t *inode)
+br_stub_mark_object_bad(xlator_t *this, inode_t *inode)
{
- int32_t ret = -1;
- uint64_t ctx_addr = 0;
- br_stub_inode_ctx_t *ctx = NULL;
-
- ret = br_stub_get_inode_ctx (this, inode, &ctx_addr);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- BRS_MSG_GET_INODE_CONTEXT_FAILED, "failed to get the "
- "inode context for the inode %s",
- uuid_utoa (inode->gfid));
- goto out;
- }
+ int32_t ret = -1;
+ uint64_t ctx_addr = 0;
+ br_stub_inode_ctx_t *ctx = NULL;
- ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
+ ret = br_stub_get_inode_ctx(this, inode, &ctx_addr);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, BRS_MSG_GET_INODE_CONTEXT_FAILED,
+ "inode-gfid=%s", uuid_utoa(inode->gfid), NULL);
+ goto out;
+ }
+
+ ctx = (br_stub_inode_ctx_t *)(long)ctx_addr;
- LOCK (&inode->lock);
- {
- __br_stub_mark_object_bad (ctx);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ __br_stub_mark_object_bad(ctx);
+ }
+ UNLOCK(&inode->lock);
out:
- return ret;
+ return ret;
}
/**
@@ -449,68 +448,68 @@ out:
* given to the caller and the caller has to decide what to do.
*/
static inline int32_t
-br_stub_mark_xdata_bad_object (xlator_t *this, inode_t *inode, dict_t *xdata)
+br_stub_mark_xdata_bad_object(xlator_t *this, inode_t *inode, dict_t *xdata)
{
- int32_t ret = 0;
+ int32_t ret = 0;
- if (br_stub_is_bad_object (this, inode) == -2)
- ret = dict_set_int32 (xdata, GLUSTERFS_BAD_INODE, 1);
+ if (br_stub_is_bad_object(this, inode) == -2)
+ ret = dict_set_int32(xdata, GLUSTERFS_BAD_INODE, 1);
- return ret;
+ return ret;
}
int32_t
-br_stub_add_fd_to_inode (xlator_t *this, fd_t *fd, br_stub_inode_ctx_t *ctx);
+br_stub_add_fd_to_inode(xlator_t *this, fd_t *fd, br_stub_inode_ctx_t *ctx);
br_sign_state_t
-__br_stub_inode_sign_state (br_stub_inode_ctx_t *ctx, glusterfs_fop_t fop,
- fd_t *fd);
+__br_stub_inode_sign_state(br_stub_inode_ctx_t *ctx, glusterfs_fop_t fop,
+ fd_t *fd);
int
-br_stub_dir_create (xlator_t *this, br_stub_private_t *priv);
+br_stub_dir_create(xlator_t *this, br_stub_private_t *priv);
int
-br_stub_add (xlator_t *this, uuid_t gfid);
+br_stub_add(xlator_t *this, uuid_t gfid);
int32_t
-br_stub_create_stub_gfid (xlator_t *this, char *stub_gfid_path, uuid_t gfid);
+br_stub_create_stub_gfid(xlator_t *this, char *stub_gfid_path, uuid_t gfid);
int
-br_stub_dir_create (xlator_t *this, br_stub_private_t *priv);
+br_stub_dir_create(xlator_t *this, br_stub_private_t *priv);
call_stub_t *
-__br_stub_dequeue (struct list_head *callstubs);
+__br_stub_dequeue(struct list_head *callstubs);
void
-__br_stub_enqueue (struct list_head *callstubs, call_stub_t *stub);
+__br_stub_enqueue(struct list_head *callstubs, call_stub_t *stub);
void
-br_stub_worker_enqueue (xlator_t *this, call_stub_t *stub);
+br_stub_worker_enqueue(xlator_t *this, call_stub_t *stub);
void *
-br_stub_worker (void *data);
+br_stub_worker(void *data);
int32_t
-br_stub_lookup_wrapper (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xattr_req);
+br_stub_lookup_wrapper(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ dict_t *xattr_req);
int32_t
-br_stub_readdir_wrapper (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t off, dict_t *xdata);
+br_stub_readdir_wrapper(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ size_t size, off_t off, dict_t *xdata);
int
-br_stub_del (xlator_t *this, uuid_t gfid);
+br_stub_del(xlator_t *this, uuid_t gfid);
int
-br_stub_bad_objects_path (xlator_t *this, fd_t *fd, gf_dirent_t *entries,
- dict_t **dict);
+br_stub_bad_objects_path(xlator_t *this, fd_t *fd, gf_dirent_t *entries,
+ dict_t **dict);
void
-br_stub_entry_xattr_fill (xlator_t *this, char *hpath, gf_dirent_t *entry,
- dict_t *dict);
+br_stub_entry_xattr_fill(xlator_t *this, char *hpath, gf_dirent_t *entry,
+ dict_t *dict);
int
-br_stub_get_path_of_gfid (xlator_t *this, inode_t *parent, inode_t *inode,
- uuid_t gfid, char **path);
+br_stub_get_path_of_gfid(xlator_t *this, inode_t *parent, inode_t *inode,
+ uuid_t gfid, char **path);
#endif /* __BIT_ROT_STUB_H__ */
diff --git a/xlators/features/changelog/lib/examples/c/get-changes-multi.c b/xlators/features/changelog/lib/examples/c/get-changes-multi.c
index 3741bdf6edc..5ea5bbb6630 100644
--- a/xlators/features/changelog/lib/examples/c/get-changes-multi.c
+++ b/xlators/features/changelog/lib/examples/c/get-changes-multi.c
@@ -25,64 +25,66 @@
#include "changelog.h"
-void *brick_init (void *xl, struct gf_brick_spec *brick)
+void *
+brick_init(void *xl, struct gf_brick_spec *brick)
{
- return brick;
+ return brick;
}
-void brick_fini (void *xl, char *brick, void *data)
+void
+brick_fini(void *xl, char *brick, void *data)
{
- return;
+ return;
}
-void brick_callback (void *xl, char *brick,
- void *data, changelog_event_t *ev)
+void
+brick_callback(void *xl, char *brick, void *data, changelog_event_t *ev)
{
- printf ("->callback: (brick,type) [%s:%d]\n", brick, ev->ev_type);
+ printf("->callback: (brick,type) [%s:%d]\n", brick, ev->ev_type);
}
-void fill_brick_spec (struct gf_brick_spec *brick, char *path)
+void
+fill_brick_spec(struct gf_brick_spec *brick, char *path)
{
- brick->brick_path = strdup (path);
- brick->filter = CHANGELOG_OP_TYPE_BR_RELEASE;
-
- brick->init = brick_init;
- brick->fini = brick_fini;
- brick->callback = brick_callback;
- brick->connected = NULL;
- brick->disconnected = NULL;
+ brick->brick_path = strdup(path);
+ brick->filter = CHANGELOG_OP_TYPE_BR_RELEASE;
+
+ brick->init = brick_init;
+ brick->fini = brick_fini;
+ brick->callback = brick_callback;
+ brick->connected = NULL;
+ brick->disconnected = NULL;
}
int
-main (int argc, char **argv)
+main(int argc, char **argv)
{
- int ret = 0;
- void *bricks = NULL;
- struct gf_brick_spec *brick = NULL;
+ int ret = 0;
+ void *bricks = NULL;
+ struct gf_brick_spec *brick = NULL;
- bricks = calloc (2, sizeof (struct gf_brick_spec));
- if (!bricks)
- goto error_return;
+ bricks = calloc(2, sizeof(struct gf_brick_spec));
+ if (!bricks)
+ goto error_return;
- brick = (struct gf_brick_spec *)bricks;
- fill_brick_spec (brick, "/export/z1/zwoop");
+ brick = (struct gf_brick_spec *)bricks;
+ fill_brick_spec(brick, "/export/z1/zwoop");
- brick++;
- fill_brick_spec (brick, "/export/z2/zwoop");
+ brick++;
+ fill_brick_spec(brick, "/export/z2/zwoop");
- ret = gf_changelog_init (NULL);
- if (ret)
- goto error_return;
+ ret = gf_changelog_init(NULL);
+ if (ret)
+ goto error_return;
- ret = gf_changelog_register_generic ((struct gf_brick_spec *)bricks, 2,
- 0, "/tmp/multi-changes.log", 9,
- NULL);
- if (ret)
- goto error_return;
+ ret = gf_changelog_register_generic((struct gf_brick_spec *)bricks, 2, 0,
+ "/tmp/multi-changes.log", 9, NULL);
+ if (ret)
+ goto error_return;
- /* let callbacks do the job */
- select (0, NULL, NULL, NULL, NULL);
+ /* let callbacks do the job */
+ select(0, NULL, NULL, NULL, NULL);
- error_return:
- return -1;
+error_return:
+ return -1;
}
diff --git a/xlators/features/changelog/lib/examples/c/get-changes.c b/xlators/features/changelog/lib/examples/c/get-changes.c
index ef766c566b6..8bc651c24a4 100644
--- a/xlators/features/changelog/lib/examples/c/get-changes.c
+++ b/xlators/features/changelog/lib/examples/c/get-changes.c
@@ -27,67 +27,67 @@
#include "changelog.h"
-#define handle_error(fn) \
- printf ("%s (reason: %s)\n", fn, strerror (errno))
+#define handle_error(fn) printf("%s (reason: %s)\n", fn, strerror(errno))
int
-main (int argc, char ** argv)
+main(int argc, char **argv)
{
- int i = 0;
- int ret = 0;
- ssize_t nr_changes = 0;
- ssize_t changes = 0;
- char fbuf[PATH_MAX] = {0,};
-
- ret = gf_changelog_init (NULL);
- if (ret) {
- handle_error ("Init failed");
- goto out;
+ int i = 0;
+ int ret = 0;
+ ssize_t nr_changes = 0;
+ ssize_t changes = 0;
+ char fbuf[PATH_MAX] = {
+ 0,
+ };
+
+ ret = gf_changelog_init(NULL);
+ if (ret) {
+ handle_error("Init failed");
+ goto out;
+ }
+
+ /* get changes for brick "/home/vshankar/export/yow/yow-1" */
+ ret = gf_changelog_register("/export/z1/zwoop", "/tmp/scratch",
+ "/tmp/change.log", 9, 5);
+ if (ret) {
+ handle_error("register failed");
+ goto out;
+ }
+
+ while (1) {
+ i = 0;
+ nr_changes = gf_changelog_scan();
+ if (nr_changes < 0) {
+ handle_error("scan(): ");
+ break;
}
- /* get changes for brick "/home/vshankar/export/yow/yow-1" */
- ret = gf_changelog_register ("/export/z1/zwoop",
- "/tmp/scratch", "/tmp/change.log", 9, 5);
- if (ret) {
- handle_error ("register failed");
- goto out;
- }
-
- while (1) {
- i = 0;
- nr_changes = gf_changelog_scan ();
- if (nr_changes < 0) {
- handle_error ("scan(): ");
- break;
- }
-
- if (nr_changes == 0)
- goto next;
+ if (nr_changes == 0)
+ goto next;
- printf ("Got %ld changelog files\n", nr_changes);
+ printf("Got %ld changelog files\n", nr_changes);
- while ( (changes =
- gf_changelog_next_change (fbuf, PATH_MAX)) > 0) {
- printf ("changelog file [%d]: %s\n", ++i, fbuf);
+ while ((changes = gf_changelog_next_change(fbuf, PATH_MAX)) > 0) {
+ printf("changelog file [%d]: %s\n", ++i, fbuf);
- /* process changelog */
- /* ... */
- /* ... */
- /* ... */
- /* done processing */
+ /* process changelog */
+ /* ... */
+ /* ... */
+ /* ... */
+ /* done processing */
- ret = gf_changelog_done (fbuf);
- if (ret)
- handle_error ("gf_changelog_done");
- }
+ ret = gf_changelog_done(fbuf);
+ if (ret)
+ handle_error("gf_changelog_done");
+ }
- if (changes == -1)
- handle_error ("gf_changelog_next_change");
+ if (changes == -1)
+ handle_error("gf_changelog_next_change");
- next:
- sleep (10);
- }
+ next:
+ sleep(10);
+ }
- out:
- return ret;
+out:
+ return ret;
}
diff --git a/xlators/features/changelog/lib/examples/c/get-history.c b/xlators/features/changelog/lib/examples/c/get-history.c
index ee3ec0ad100..3e888d75ca6 100644
--- a/xlators/features/changelog/lib/examples/c/get-history.c
+++ b/xlators/features/changelog/lib/examples/c/get-history.c
@@ -27,90 +27,90 @@
#include "changelog.h"
-#define handle_error(fn) \
- printf ("%s (reason: %s)\n", fn, strerror (errno))
+#define handle_error(fn) printf("%s (reason: %s)\n", fn, strerror(errno))
int
-main (int argc, char ** argv)
+main(int argc, char **argv)
{
- int i = 0;
- int ret = 0;
- ssize_t nr_changes = 0;
- ssize_t changes = 0;
- char fbuf[PATH_MAX] = {0,};
- unsigned long end_ts = 0;
-
- ret = gf_changelog_init (NULL);
- if (ret) {
- handle_error ("init failed");
- goto out;
+ int i = 0;
+ int ret = 0;
+ ssize_t nr_changes = 0;
+ ssize_t changes = 0;
+ char fbuf[PATH_MAX] = {
+ 0,
+ };
+ unsigned long end_ts = 0;
+
+ ret = gf_changelog_init(NULL);
+ if (ret) {
+ handle_error("init failed");
+ goto out;
+ }
+
+ ret = gf_changelog_register("/export/z1/zwoop", "/tmp/scratch_v1",
+ "/tmp/changes.log", 9, 5);
+ if (ret) {
+ handle_error("register failed");
+ goto out;
+ }
+
+ int a, b;
+ printf("give the two numbers start and end\t");
+ scanf("%d%d", &a, &b);
+ ret = gf_history_changelog("/export/z1/zwoop/.glusterfs/changelogs", a, b,
+ 3, &end_ts);
+ if (ret == -1) {
+ printf("history failed");
+ goto out;
+ }
+
+ printf("end time till when changelog available : %d , ret(%d) \t", end_ts,
+ ret);
+ fflush(stdout);
+
+ while (1) {
+ nr_changes = gf_history_changelog_scan();
+ printf("scanned, nr_changes : %d\n", nr_changes);
+ if (nr_changes < 0) {
+ handle_error("scan(): ");
+ break;
}
- ret = gf_changelog_register ("/export/z1/zwoop",
- "/tmp/scratch_v1", "/tmp/changes.log",
- 9, 5);
- if (ret) {
- handle_error ("register failed");
- goto out;
+ if (nr_changes == 0) {
+ printf("done scanning \n");
+ goto out;
}
- int a, b;
- printf ("give the two numbers start and end\t");
- scanf ("%d%d", &a, &b);
- ret = gf_history_changelog ("/export/z1/zwoop/.glusterfs/changelogs",
- a, b, 3, &end_ts);
- if (ret == -1) {
- printf ("history failed");
- goto out;
- }
+ printf("Got %ld changelog files\n", nr_changes);
+
+ while ((changes = gf_history_changelog_next_change(fbuf, PATH_MAX)) >
+ 0) {
+ printf("changelog file [%d]: %s\n", ++i, fbuf);
- printf ("end time till when changelog available : %d , ret(%d) \t", end_ts, ret);
- fflush(stdout);
-
- while (1) {
- nr_changes = gf_history_changelog_scan ();
- printf ("scanned, nr_changes : %d\n",nr_changes);
- if (nr_changes < 0) {
- handle_error ("scan(): ");
- break;
- }
-
- if (nr_changes == 0) {
- printf ("done scanning \n");
- goto out;
- }
-
- printf ("Got %ld changelog files\n", nr_changes);
-
- while ( (changes =
- gf_history_changelog_next_change (fbuf, PATH_MAX)) > 0) {
- printf ("changelog file [%d]: %s\n", ++i, fbuf);
-
- /* process changelog */
- /* ... */
- /* ... */
- /* ... */
- /* done processing */
-
- ret = gf_history_changelog_done (fbuf);
- if (ret)
- handle_error ("gf_changelog_done");
- }
- /*
- if (changes == -1)
- handle_error ("gf_changelog_next_change");
- if (nr_changes ==1){
- printf("continue scanning\n");
- }
-
- if(nr_changes == 0){
- printf("done scanning \n");
- goto out;
- }
- */
+ /* process changelog */
+ /* ... */
+ /* ... */
+ /* ... */
+ /* done processing */
+
+ ret = gf_history_changelog_done(fbuf);
+ if (ret)
+ handle_error("gf_changelog_done");
+ }
+ /*
+ if (changes == -1)
+ handle_error ("gf_changelog_next_change");
+ if (nr_changes ==1){
+ printf("continue scanning\n");
}
+ if(nr_changes == 0){
+ printf("done scanning \n");
+ goto out;
+ }
+ */
+ }
out:
- return ret;
+ return ret;
}
diff --git a/xlators/features/changelog/lib/examples/python/changes.py b/xlators/features/changelog/lib/examples/python/changes.py
index 077c1108cce..c410d3b000d 100644..100755
--- a/xlators/features/changelog/lib/examples/python/changes.py
+++ b/xlators/features/changelog/lib/examples/python/changes.py
@@ -1,4 +1,4 @@
-#!/usr/bin/python2
+#!/usr/bin/python3
from __future__ import print_function
import os
diff --git a/xlators/features/changelog/lib/examples/python/libgfchangelog.py b/xlators/features/changelog/lib/examples/python/libgfchangelog.py
index 2cdbf1152b9..2da9f2d2a8c 100644
--- a/xlators/features/changelog/lib/examples/python/libgfchangelog.py
+++ b/xlators/features/changelog/lib/examples/python/libgfchangelog.py
@@ -1,8 +1,10 @@
import os
from ctypes import *
+from ctypes.util import find_library
class Changes(object):
- libgfc = CDLL("libgfchangelog.so", mode=RTLD_GLOBAL, use_errno=True)
+ libgfc = CDLL(find_library("gfchangelog"), mode=RTLD_GLOBAL,
+ use_errno=True)
@classmethod
def geterrno(cls):
diff --git a/xlators/features/changelog/lib/src/Makefile.am b/xlators/features/changelog/lib/src/Makefile.am
index 92ee26827da..c933ec53ed2 100644
--- a/xlators/features/changelog/lib/src/Makefile.am
+++ b/xlators/features/changelog/lib/src/Makefile.am
@@ -1,7 +1,7 @@
libgfchangelog_la_CFLAGS = -Wall $(GF_CFLAGS) $(GF_DARWIN_LIBGLUSTERFS_CFLAGS) \
-DDATADIR=\"$(localstatedir)\"
-libgfchangelog_la_CPPFLAGS = $(GF_CPPFLAGS) -D__USE_FILE_OFFSET64 -fpic \
+libgfchangelog_la_CPPFLAGS = $(GF_CPPFLAGS) -D__USE_FILE_OFFSET64 -D__USE_LARGEFILE64 -fpic \
-I../../../src/ -I$(top_srcdir)/libglusterfs/src \
-I$(top_srcdir)/xlators/features/changelog/src \
-I$(top_srcdir)/rpc/xdr/src -I$(top_builddir)/rpc/xdr/src \
@@ -13,10 +13,10 @@ libgfchangelog_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la \
$(top_builddir)/rpc/xdr/src/libgfxdr.la \
$(top_builddir)/rpc/rpc-lib/src/libgfrpc.la
-libgfchangelog_la_LDFLAGS = $(GF_LDFLAGS) -version-info \
- $(LIBGFCHANGELOG_LT_VERSION)
+libgfchangelog_la_LDFLAGS = $(GF_LDFLAGS) \
+ -version-info $(LIBGFCHANGELOG_LT_VERSION) \
+ $(GF_NO_UNDEFINED)
-libgfchangelogdir = $(includedir)/glusterfs/gfchangelog
lib_LTLIBRARIES = libgfchangelog.la
CONTRIB_BUILDDIR = $(top_builddir)/contrib
diff --git a/xlators/features/changelog/lib/src/changelog-lib-messages.h b/xlators/features/changelog/lib/src/changelog-lib-messages.h
index 2061217b801..d7fe7274353 100644
--- a/xlators/features/changelog/lib/src/changelog-lib-messages.h
+++ b/xlators/features/changelog/lib/src/changelog-lib-messages.h
@@ -11,7 +11,7 @@
#ifndef _CHANGELOG_LIB_MESSAGES_H_
#define _CHANGELOG_LIB_MESSAGES_H_
-#include "glfs-message-id.h"
+#include <glusterfs/glfs-message-id.h>
/* To add new message IDs, append new identifiers at the end of the list.
*
@@ -23,39 +23,52 @@
* glfs-message-id.h.
*/
-GLFS_MSGID(CHANGELOG_LIB,
- CHANGELOG_LIB_MSG_OPEN_FAILED,
- CHANGELOG_LIB_MSG_FAILED_TO_RMDIR,
- CHANGELOG_LIB_MSG_SCRATCH_DIR_ENTRIES_CREATION_ERROR,
- CHANGELOG_LIB_MSG_THREAD_CREATION_FAILED,
- CHANGELOG_LIB_MSG_OPENDIR_ERROR,
- CHANGELOG_LIB_MSG_RENAME_FAILED,
- CHANGELOG_LIB_MSG_READ_ERROR,
- CHANGELOG_LIB_MSG_HTIME_ERROR,
- CHANGELOG_LIB_MSG_GET_TIME_ERROR,
- CHANGELOG_LIB_MSG_WRITE_FAILED,
- CHANGELOG_LIB_MSG_PTHREAD_ERROR,
- CHANGELOG_LIB_MSG_MMAP_FAILED,
- CHANGELOG_LIB_MSG_MUNMAP_FAILED,
- CHANGELOG_LIB_MSG_ASCII_ERROR,
- CHANGELOG_LIB_MSG_STAT_FAILED,
- CHANGELOG_LIB_MSG_GET_XATTR_FAILED,
- CHANGELOG_LIB_MSG_PUBLISH_ERROR,
- CHANGELOG_LIB_MSG_PARSE_ERROR,
- CHANGELOG_LIB_MSG_TOTAL_LOG_INFO,
- CHANGELOG_LIB_MSG_CLEANUP_ERROR,
- CHANGELOG_LIB_MSG_UNLINK_FAILED,
- CHANGELOG_LIB_MSG_NOTIFY_REGISTER_FAILED,
- CHANGELOG_LIB_MSG_INVOKE_RPC_FAILED,
- CHANGELOG_LIB_MSG_DRAINING_EVENT_INFO,
- CHANGELOG_LIB_MSG_CLEANING_BRICK_ENTRY_INFO,
- CHANGELOG_LIB_MSG_FREEING_ENTRY_INFO,
- CHANGELOG_LIB_MSG_XDR_DECODING_FAILED,
- CHANGELOG_LIB_MSG_NOTIFY_REGISTER_INFO,
- CHANGELOG_LIB_MSG_THREAD_CLEANUP_WARNING,
- CHANGELOG_LIB_MSG_COPY_FROM_BUFFER_FAILED,
- CHANGELOG_LIB_MSG_PTHREAD_JOIN_FAILED,
- CHANGELOG_LIB_MSG_HIST_FAILED
-);
+GLFS_MSGID(
+ CHANGELOG_LIB, CHANGELOG_LIB_MSG_OPEN_FAILED,
+ CHANGELOG_LIB_MSG_FAILED_TO_RMDIR,
+ CHANGELOG_LIB_MSG_SCRATCH_DIR_ENTRIES_CREATION_ERROR,
+ CHANGELOG_LIB_MSG_THREAD_CREATION_FAILED, CHANGELOG_LIB_MSG_OPENDIR_ERROR,
+ CHANGELOG_LIB_MSG_RENAME_FAILED, CHANGELOG_LIB_MSG_READ_ERROR,
+ CHANGELOG_LIB_MSG_HTIME_ERROR, CHANGELOG_LIB_MSG_GET_TIME_ERROR,
+ CHANGELOG_LIB_MSG_WRITE_FAILED, CHANGELOG_LIB_MSG_PTHREAD_ERROR,
+ CHANGELOG_LIB_MSG_MMAP_FAILED, CHANGELOG_LIB_MSG_MUNMAP_FAILED,
+ CHANGELOG_LIB_MSG_ASCII_ERROR, CHANGELOG_LIB_MSG_STAT_FAILED,
+ CHANGELOG_LIB_MSG_GET_XATTR_FAILED, CHANGELOG_LIB_MSG_PUBLISH_ERROR,
+ CHANGELOG_LIB_MSG_PARSE_ERROR, CHANGELOG_LIB_MSG_MIN_MAX_INFO,
+ CHANGELOG_LIB_MSG_CLEANUP_ERROR, CHANGELOG_LIB_MSG_UNLINK_FAILED,
+ CHANGELOG_LIB_MSG_NOTIFY_REGISTER_FAILED,
+ CHANGELOG_LIB_MSG_INVOKE_RPC_FAILED, CHANGELOG_LIB_MSG_DRAINING_EVENT_INFO,
+ CHANGELOG_LIB_MSG_CLEANING_BRICK_ENTRY_INFO,
+ CHANGELOG_LIB_MSG_FREEING_ENTRY_INFO, CHANGELOG_LIB_MSG_XDR_DECODING_FAILED,
+ CHANGELOG_LIB_MSG_NOTIFY_REGISTER_INFO,
+ CHANGELOG_LIB_MSG_THREAD_CLEANUP_WARNING,
+ CHANGELOG_LIB_MSG_COPY_FROM_BUFFER_FAILED,
+ CHANGELOG_LIB_MSG_PTHREAD_JOIN_FAILED, CHANGELOG_LIB_MSG_HIST_FAILED,
+ CHANGELOG_LIB_MSG_DRAINED_EVENT_INFO, CHANGELOG_LIB_MSG_PARSE_ERROR_CEASED,
+ CHANGELOG_LIB_MSG_REQUESTING_INFO, CHANGELOG_LIB_MSG_FINAL_INFO);
+
+#define CHANGELOG_LIB_MSG_NOTIFY_REGISTER_INFO_STR "Registering brick"
+#define CHANGELOG_LIB_MSG_RENAME_FAILED_STR "error moving changelog file"
+#define CHANGELOG_LIB_MSG_OPEN_FAILED_STR "cannot open changelog file"
+#define CHANGELOG_LIB_MSG_UNLINK_FAILED_STR "failed to unlink"
+#define CHANGELOG_LIB_MSG_FAILED_TO_RMDIR_STR "failed to rmdir"
+#define CHANGELOG_LIB_MSG_STAT_FAILED_STR "stat failed on changelog file"
+#define CHANGELOG_LIB_MSG_PARSE_ERROR_STR "could not parse changelog"
+#define CHANGELOG_LIB_MSG_PARSE_ERROR_CEASED_STR \
+ "parsing error, ceased publishing..."
+#define CHANGELOG_LIB_MSG_HTIME_ERROR_STR "fop failed on htime file"
+#define CHANGELOG_LIB_MSG_GET_XATTR_FAILED_STR \
+ "error extracting max timstamp from htime file"
+#define CHANGELOG_LIB_MSG_MIN_MAX_INFO_STR "changelogs min max"
+#define CHANGELOG_LIB_MSG_REQUESTING_INFO_STR "Requesting historical changelogs"
+#define CHANGELOG_LIB_MSG_FINAL_INFO_STR "FINAL"
+#define CHANGELOG_LIB_MSG_HIST_FAILED_STR \
+ "Requested changelog range is not available"
+#define CHANGELOG_LIB_MSG_GET_TIME_ERROR_STR "wrong result"
+#define CHANGELOG_LIB_MSG_CLEANING_BRICK_ENTRY_INFO_STR \
+ "Cleaning brick entry for brick"
+#define CHANGELOG_LIB_MSG_DRAINING_EVENT_INFO_STR "Draining event"
+#define CHANGELOG_LIB_MSG_DRAINED_EVENT_INFO_STR "Drained event"
+#define CHANGELOG_LIB_MSG_FREEING_ENTRY_INFO_STR "freeing entry"
#endif /* !_CHANGELOG_MESSAGES_H_ */
diff --git a/xlators/features/changelog/lib/src/gf-changelog-api.c b/xlators/features/changelog/lib/src/gf-changelog-api.c
index 372550c7acf..81a5cbfec10 100644
--- a/xlators/features/changelog/lib/src/gf-changelog-api.c
+++ b/xlators/features/changelog/lib/src/gf-changelog-api.c
@@ -8,10 +8,10 @@
cases as published by the Free Software Foundation.
*/
-#include "compat-uuid.h"
-#include "globals.h"
-#include "glusterfs.h"
-#include "syscall.h"
+#include <glusterfs/compat-uuid.h>
+#include <glusterfs/globals.h>
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/syscall.h>
#include "gf-changelog-helpers.h"
#include "gf-changelog-journal.h"
@@ -19,57 +19,54 @@
#include "changelog-lib-messages.h"
int
-gf_changelog_done (char *file)
+gf_changelog_done(char *file)
{
- int ret = -1;
- char *buffer = NULL;
- xlator_t *this = NULL;
- gf_changelog_journal_t *jnl = NULL;
- char to_path[PATH_MAX] = {0,};
-
- errno = EINVAL;
-
- this = THIS;
- if (!this)
- goto out;
-
- jnl = (gf_changelog_journal_t *) GF_CHANGELOG_GET_API_PTR (this);
- if (!jnl)
- goto out;
-
- if (!file || !strlen (file))
- goto out;
-
- /* make sure 'file' is inside ->jnl_working_dir */
- buffer = realpath (file, NULL);
- if (!buffer)
- goto out;
-
- if (strncmp (jnl->jnl_working_dir,
- buffer, strlen (jnl->jnl_working_dir)))
- goto out;
-
- (void) snprintf (to_path, PATH_MAX, "%s%s",
- jnl->jnl_processed_dir, basename (buffer));
- gf_msg_debug (this->name, 0,
- "moving %s to processed directory", file);
- ret = sys_rename (buffer, to_path);
- if (ret) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_RENAME_FAILED,
- "cannot move changelog file",
- "from=%s", file,
- "to=%s", to_path,
- NULL);
- goto out;
- }
-
- ret = 0;
+ int ret = -1;
+ char *buffer = NULL;
+ xlator_t *this = NULL;
+ gf_changelog_journal_t *jnl = NULL;
+ char to_path[PATH_MAX] = {
+ 0,
+ };
+
+ errno = EINVAL;
+
+ this = THIS;
+ if (!this)
+ goto out;
+
+ jnl = (gf_changelog_journal_t *)GF_CHANGELOG_GET_API_PTR(this);
+ if (!jnl)
+ goto out;
+
+ if (!file || !strlen(file))
+ goto out;
+
+ /* make sure 'file' is inside ->jnl_working_dir */
+ buffer = realpath(file, NULL);
+ if (!buffer)
+ goto out;
+
+ if (strncmp(jnl->jnl_working_dir, buffer, strlen(jnl->jnl_working_dir)))
+ goto out;
+
+ (void)snprintf(to_path, PATH_MAX, "%s%s", jnl->jnl_processed_dir,
+ basename(buffer));
+ gf_msg_debug(this->name, 0, "moving %s to processed directory", file);
+ ret = sys_rename(buffer, to_path);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_LIB_MSG_RENAME_FAILED, "from=%s", file, "to=%s",
+ to_path, NULL);
+ goto out;
+ }
+
+ ret = 0;
- out:
- if (buffer)
- free (buffer); /* allocated by realpath() */
- return ret;
+out:
+ if (buffer)
+ free(buffer); /* allocated by realpath() */
+ return ret;
}
/**
@@ -77,28 +74,28 @@ gf_changelog_done (char *file)
* for a set of changelogs, start from the beginning
*/
int
-gf_changelog_start_fresh ()
+gf_changelog_start_fresh()
{
- xlator_t *this = NULL;
- gf_changelog_journal_t *jnl = NULL;
+ xlator_t *this = NULL;
+ gf_changelog_journal_t *jnl = NULL;
- this = THIS;
- if (!this)
- goto out;
+ this = THIS;
+ if (!this)
+ goto out;
- errno = EINVAL;
+ errno = EINVAL;
- jnl = (gf_changelog_journal_t *) GF_CHANGELOG_GET_API_PTR (this);
- if (!jnl)
- goto out;
+ jnl = (gf_changelog_journal_t *)GF_CHANGELOG_GET_API_PTR(this);
+ if (!jnl)
+ goto out;
- if (gf_ftruncate (jnl->jnl_fd, 0))
- goto out;
+ if (gf_ftruncate(jnl->jnl_fd, 0))
+ goto out;
- return 0;
+ return 0;
- out:
- return -1;
+out:
+ return -1;
}
/**
@@ -107,40 +104,42 @@ gf_changelog_start_fresh ()
* consumed.
*/
ssize_t
-gf_changelog_next_change (char *bufptr, size_t maxlen)
+gf_changelog_next_change(char *bufptr, size_t maxlen)
{
- ssize_t size = -1;
- int tracker_fd = 0;
- xlator_t *this = NULL;
- gf_changelog_journal_t *jnl = NULL;
- char buffer[PATH_MAX] = {0,};
+ ssize_t size = -1;
+ int tracker_fd = 0;
+ xlator_t *this = NULL;
+ gf_changelog_journal_t *jnl = NULL;
+ char buffer[PATH_MAX] = {
+ 0,
+ };
- errno = EINVAL;
+ errno = EINVAL;
- this = THIS;
- if (!this)
- goto out;
+ this = THIS;
+ if (!this)
+ goto out;
- jnl = (gf_changelog_journal_t *) GF_CHANGELOG_GET_API_PTR (this);
- if (!jnl)
- goto out;
+ jnl = (gf_changelog_journal_t *)GF_CHANGELOG_GET_API_PTR(this);
+ if (!jnl)
+ goto out;
- tracker_fd = jnl->jnl_fd;
+ tracker_fd = jnl->jnl_fd;
- size = gf_readline (tracker_fd, buffer, maxlen);
- if (size < 0) {
- size = -1;
- goto out;
- }
+ size = gf_readline(tracker_fd, buffer, maxlen);
+ if (size < 0) {
+ size = -1;
+ goto out;
+ }
- if (size == 0)
- goto out;
+ if (size == 0)
+ goto out;
- memcpy (bufptr, buffer, size - 1);
- bufptr[size - 1] = '\0';
+ memcpy(bufptr, buffer, size - 1);
+ bufptr[size - 1] = '\0';
out:
- return size;
+ return size;
}
/**
@@ -152,70 +151,74 @@ out:
* This call also acts as a cancellation point for the consumer.
*/
ssize_t
-gf_changelog_scan ()
+gf_changelog_scan()
{
- int tracker_fd = 0;
- size_t off = 0;
- xlator_t *this = NULL;
- size_t nr_entries = 0;
- gf_changelog_journal_t *jnl = NULL;
- struct dirent *entry = NULL;
- struct dirent scratch[2] = {{0,},};
- char buffer[PATH_MAX] = {0,};
-
- this = THIS;
- if (!this)
- goto out;
-
- jnl = (gf_changelog_journal_t *) GF_CHANGELOG_GET_API_PTR (this);
- if (!jnl)
- goto out;
- if (JNL_IS_API_DISCONNECTED (jnl)) {
- errno = ENOTCONN;
- goto out;
+ int tracker_fd = 0;
+ size_t off = 0;
+ xlator_t *this = NULL;
+ size_t nr_entries = 0;
+ gf_changelog_journal_t *jnl = NULL;
+ struct dirent *entry = NULL;
+ struct dirent scratch[2] = {
+ {
+ 0,
+ },
+ };
+ char buffer[PATH_MAX] = {
+ 0,
+ };
+
+ this = THIS;
+ if (!this)
+ goto out;
+
+ jnl = (gf_changelog_journal_t *)GF_CHANGELOG_GET_API_PTR(this);
+ if (!jnl)
+ goto out;
+ if (JNL_IS_API_DISCONNECTED(jnl)) {
+ errno = ENOTCONN;
+ goto out;
+ }
+
+ errno = EINVAL;
+
+ tracker_fd = jnl->jnl_fd;
+ if (gf_ftruncate(tracker_fd, 0))
+ goto out;
+
+ rewinddir(jnl->jnl_dir);
+
+ for (;;) {
+ errno = 0;
+ entry = sys_readdir(jnl->jnl_dir, scratch);
+ if (!entry || errno != 0)
+ break;
+
+ if (!strcmp(basename(entry->d_name), ".") ||
+ !strcmp(basename(entry->d_name), ".."))
+ continue;
+
+ nr_entries++;
+
+ GF_CHANGELOG_FILL_BUFFER(jnl->jnl_processing_dir, buffer, off,
+ strlen(jnl->jnl_processing_dir));
+ GF_CHANGELOG_FILL_BUFFER(entry->d_name, buffer, off,
+ strlen(entry->d_name));
+ GF_CHANGELOG_FILL_BUFFER("\n", buffer, off, 1);
+
+ if (gf_changelog_write(tracker_fd, buffer, off) != off) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, CHANGELOG_LIB_MSG_WRITE_FAILED,
+ "error writing changelog filename"
+ " to tracker file");
+ break;
}
+ off = 0;
+ }
- errno = EINVAL;
-
- tracker_fd = jnl->jnl_fd;
- if (gf_ftruncate (tracker_fd, 0))
- goto out;
-
- rewinddir (jnl->jnl_dir);
-
- for (;;) {
- errno = 0;
- entry = sys_readdir (jnl->jnl_dir, scratch);
- if (!entry || errno != 0)
- break;
-
- if (!strcmp (basename (entry->d_name), ".")
- || !strcmp (basename (entry->d_name), ".."))
- continue;
-
- nr_entries++;
-
- GF_CHANGELOG_FILL_BUFFER (jnl->jnl_processing_dir,
- buffer, off,
- strlen (jnl->jnl_processing_dir));
- GF_CHANGELOG_FILL_BUFFER (entry->d_name, buffer,
- off, strlen (entry->d_name));
- GF_CHANGELOG_FILL_BUFFER ("\n", buffer, off, 1);
-
- if (gf_changelog_write (tracker_fd, buffer, off) != off) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_LIB_MSG_WRITE_FAILED,
- "error writing changelog filename"
- " to tracker file");
- break;
- }
- off = 0;
- }
-
- if (!entry) {
- if (gf_lseek (tracker_fd, 0, SEEK_SET) != -1)
- return nr_entries;
- }
- out:
- return -1;
+ if (!entry) {
+ if (gf_lseek(tracker_fd, 0, SEEK_SET) != -1)
+ return nr_entries;
+ }
+out:
+ return -1;
}
diff --git a/xlators/features/changelog/lib/src/gf-changelog-helpers.c b/xlators/features/changelog/lib/src/gf-changelog-helpers.c
index 45cf55778e4..75f8a6dfc08 100644
--- a/xlators/features/changelog/lib/src/gf-changelog-helpers.c
+++ b/xlators/features/changelog/lib/src/gf-changelog-helpers.c
@@ -11,40 +11,36 @@
#include "changelog-mem-types.h"
#include "gf-changelog-helpers.h"
#include "changelog-lib-messages.h"
-#include "syscall.h"
-
-ssize_t gf_changelog_read_path (int fd, char *buffer, size_t bufsize)
-{
- return sys_read (fd, buffer, bufsize);
-}
+#include <glusterfs/syscall.h>
size_t
-gf_changelog_write (int fd, char *buffer, size_t len)
+gf_changelog_write(int fd, char *buffer, size_t len)
{
- ssize_t size = 0;
- size_t written = 0;
+ ssize_t size = 0;
+ size_t written = 0;
- while (written < len) {
- size = sys_write (fd, buffer + written, len - written);
- if (size <= 0)
- break;
+ while (written < len) {
+ size = sys_write(fd, buffer + written, len - written);
+ if (size <= 0)
+ break;
- written += size;
- }
+ written += size;
+ }
- return written;
+ return written;
}
void
-gf_rfc3986_encode_space_newline (unsigned char *s, char *enc, char *estr)
+gf_rfc3986_encode_space_newline(unsigned char *s, char *enc, char *estr)
{
- for (; *s; s++) {
- if (estr[*s])
- sprintf(enc, "%c", estr[*s]);
- else
- sprintf(enc, "%%%02X", *s);
- while (*++enc);
- }
+ for (; *s; s++) {
+ if (estr[*s])
+ sprintf(enc, "%c", estr[*s]);
+ else
+ sprintf(enc, "%%%02X", *s);
+ while (*++enc)
+ ;
+ }
}
/**
@@ -57,163 +53,118 @@ gf_rfc3986_encode_space_newline (unsigned char *s, char *enc, char *estr)
* that can be done via @fflush(fp), @ftruncate(fd) and @fseek(fp),
* but this involves mixing POSIX file descriptors and stream FILE *).
*
- * NOTE: This implmentation still does work with more than one fd's
+ * NOTE: This implementation still does work with more than one fd's
* used to perform gf_readline(). For this very reason it's not
* made a part of libglusterfs.
*/
-static pthread_key_t rl_key;
-static pthread_once_t rl_once = PTHREAD_ONCE_INIT;
-
-static void
-readline_destructor (void *ptr)
-{
- GF_FREE (ptr);
-}
-
-static void
-readline_once (void)
-{
- pthread_key_create (&rl_key, readline_destructor);
-}
+static __thread read_line_t thread_tsd = {};
static ssize_t
-my_read (read_line_t *tsd, int fd, char *ptr)
-{
- if (tsd->rl_cnt <= 0) {
- tsd->rl_cnt = sys_read (fd, tsd->rl_buf, MAXLINE);
-
- if (tsd->rl_cnt < 0)
- return -1;
- else if (tsd->rl_cnt == 0)
- return 0;
- tsd->rl_bufptr = tsd->rl_buf;
- }
-
- tsd->rl_cnt--;
- *ptr = *tsd->rl_bufptr++;
- return 1;
-}
-
-static int
-gf_readline_init_once (read_line_t **tsd)
+my_read(read_line_t *tsd, int fd, char *ptr)
{
- if (pthread_once (&rl_once, readline_once) != 0)
- return -1;
+ if (tsd->rl_cnt <= 0) {
+ tsd->rl_cnt = sys_read(fd, tsd->rl_buf, MAXLINE);
- *tsd = pthread_getspecific (rl_key);
- if (*tsd)
- goto out;
-
- *tsd = GF_CALLOC (1, sizeof (**tsd),
- gf_changelog_mt_libgfchangelog_rl_t);
- if (!*tsd)
- return -1;
-
- if (pthread_setspecific (rl_key, *tsd) != 0)
- return -1;
+ if (tsd->rl_cnt < 0)
+ return -1;
+ else if (tsd->rl_cnt == 0)
+ return 0;
+ tsd->rl_bufptr = tsd->rl_buf;
+ }
- out:
- return 0;
+ tsd->rl_cnt--;
+ *ptr = *tsd->rl_bufptr++;
+ return 1;
}
ssize_t
-gf_readline (int fd, void *vptr, size_t maxlen)
+gf_readline(int fd, void *vptr, size_t maxlen)
{
- size_t n = 0;
- size_t rc = 0;
- char c = ' ';
- char *ptr = NULL;
- read_line_t *tsd = NULL;
-
- if (gf_readline_init_once (&tsd))
- return -1;
-
- ptr = vptr;
- for (n = 1; n < maxlen; n++) {
- if ( (rc = my_read (tsd, fd, &c)) == 1 ) {
- *ptr++ = c;
- if (c == '\n')
- break;
- } else if (rc == 0) {
- *ptr = '\0';
- return (n - 1);
- } else
- return -1;
- }
-
- *ptr = '\0';
- return n;
-
+ size_t n = 0;
+ size_t rc = 0;
+ char c = ' ';
+ char *ptr = NULL;
+ read_line_t *tsd = &thread_tsd;
+
+ ptr = vptr;
+ for (n = 1; n < maxlen; n++) {
+ if ((rc = my_read(tsd, fd, &c)) == 1) {
+ *ptr++ = c;
+ if (c == '\n')
+ break;
+ } else if (rc == 0) {
+ *ptr = '\0';
+ return (n - 1);
+ } else
+ return -1;
+ }
+
+ *ptr = '\0';
+ return n;
}
off_t
-gf_lseek (int fd, off_t offset, int whence)
+gf_lseek(int fd, off_t offset, int whence)
{
- off_t off = 0;
- read_line_t *tsd = NULL;
+ off_t off = 0;
+ read_line_t *tsd = &thread_tsd;
- if (gf_readline_init_once (&tsd))
- return -1;
+ off = sys_lseek(fd, offset, whence);
+ if (off == -1)
+ return -1;
- off = sys_lseek (fd, offset, whence);
- if (off == -1)
- return -1;
+ tsd->rl_cnt = 0;
+ tsd->rl_bufptr = tsd->rl_buf;
- tsd->rl_cnt = 0;
- tsd->rl_bufptr = tsd->rl_buf;
-
- return off;
+ return off;
}
int
-gf_ftruncate (int fd, off_t length)
+gf_ftruncate(int fd, off_t length)
{
- read_line_t *tsd = NULL;
+ read_line_t *tsd = &thread_tsd;
- if (gf_readline_init_once (&tsd))
- return -1;
+ if (sys_ftruncate(fd, 0))
+ return -1;
- if (sys_ftruncate (fd, 0))
- return -1;
+ tsd->rl_cnt = 0;
+ tsd->rl_bufptr = tsd->rl_buf;
- tsd->rl_cnt = 0;
- tsd->rl_bufptr = tsd->rl_buf;
-
- return 0;
+ return 0;
}
int
-gf_thread_cleanup (xlator_t *this, pthread_t thread)
+gf_thread_cleanup(xlator_t *this, pthread_t thread)
{
- int ret = 0;
- void *res = NULL;
-
- ret = pthread_cancel (thread);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- CHANGELOG_LIB_MSG_THREAD_CLEANUP_WARNING,
- "Failed to send cancellation to thread");
- goto error_return;
- }
-
- ret = pthread_join (thread, &res);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- CHANGELOG_LIB_MSG_THREAD_CLEANUP_WARNING,
- "failed to join thread");
- goto error_return;
- }
-
- if (res != PTHREAD_CANCELED) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- CHANGELOG_LIB_MSG_THREAD_CLEANUP_WARNING,
- "Thread could not be cleaned up");
- goto error_return;
- }
-
- return 0;
-
- error_return:
- return -1;
+ int ret = 0;
+ void *res = NULL;
+
+ ret = pthread_cancel(thread);
+ if (ret != 0) {
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ CHANGELOG_LIB_MSG_THREAD_CLEANUP_WARNING,
+ "Failed to send cancellation to thread");
+ goto error_return;
+ }
+
+ ret = pthread_join(thread, &res);
+ if (ret != 0) {
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ CHANGELOG_LIB_MSG_THREAD_CLEANUP_WARNING,
+ "failed to join thread");
+ goto error_return;
+ }
+
+ if (res != PTHREAD_CANCELED) {
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ CHANGELOG_LIB_MSG_THREAD_CLEANUP_WARNING,
+ "Thread could not be cleaned up");
+ goto error_return;
+ }
+
+ return 0;
+
+error_return:
+ return -1;
}
diff --git a/xlators/features/changelog/lib/src/gf-changelog-helpers.h b/xlators/features/changelog/lib/src/gf-changelog-helpers.h
index b05628ee70d..9c609d33172 100644
--- a/xlators/features/changelog/lib/src/gf-changelog-helpers.h
+++ b/xlators/features/changelog/lib/src/gf-changelog-helpers.h
@@ -14,36 +14,37 @@
#include <unistd.h>
#include <dirent.h>
#include <limits.h>
-#include "locking.h"
+#include <glusterfs/locking.h>
-#include <xlator.h>
+#include <glusterfs/xlator.h>
#include "changelog.h"
#include "changelog-rpc-common.h"
#include "gf-changelog-journal.h"
-#define GF_CHANGELOG_TRACKER "tracker"
+#define GF_CHANGELOG_TRACKER "tracker"
-#define GF_CHANGELOG_CURRENT_DIR ".current"
-#define GF_CHANGELOG_PROCESSED_DIR ".processed"
+#define GF_CHANGELOG_CURRENT_DIR ".current"
+#define GF_CHANGELOG_PROCESSED_DIR ".processed"
#define GF_CHANGELOG_PROCESSING_DIR ".processing"
-#define GF_CHANGELOG_HISTORY_DIR ".history"
+#define GF_CHANGELOG_HISTORY_DIR ".history"
#define TIMESTAMP_LENGTH 10
#ifndef MAXLINE
#define MAXLINE 4096
#endif
-#define GF_CHANGELOG_FILL_BUFFER(ptr, ascii, off, len) do { \
- memcpy (ascii + off, ptr, len); \
- off += len; \
- } while (0)
+#define GF_CHANGELOG_FILL_BUFFER(ptr, ascii, off, len) \
+ do { \
+ memcpy(ascii + off, ptr, len); \
+ off += len; \
+ } while (0)
typedef struct read_line {
- int rl_cnt;
- char *rl_bufptr;
- char rl_buf[MAXLINE];
+ int rl_cnt;
+ char *rl_bufptr;
+ char rl_buf[MAXLINE];
} read_line_t;
struct gf_changelog;
@@ -55,51 +56,50 @@ struct gf_event;
* ->next_seq holds the next _expected_ sequence number.
*/
struct gf_event_list {
- pthread_mutex_t lock; /* protects this structure */
- pthread_cond_t cond;
+ pthread_mutex_t lock; /* protects this structure */
+ pthread_cond_t cond;
- pthread_t invoker;
+ pthread_t invoker;
- unsigned long next_seq; /* next sequence number expected:
- zero during bootstrap */
+ unsigned long next_seq; /* next sequence number expected:
+ zero during bootstrap */
- struct gf_changelog *entry; /* backpointer to it's brick
- encapsulator (entry) */
- struct list_head events; /* list of events */
+ struct gf_changelog *entry; /* backpointer to it's brick
+ encapsulator (entry) */
+ struct list_head events; /* list of events */
};
/**
* include a refcount if it's of use by additional layers
*/
struct gf_event {
- int count;
+ int count;
- unsigned long seq;
+ unsigned long seq;
- struct list_head list;
+ struct list_head list;
- struct iovec iov[0];
+ struct iovec iov[0];
};
-#define GF_EVENT_CALLOC_SIZE(cnt, len) \
- (sizeof (struct gf_event) + (cnt * sizeof (struct iovec)) + len)
+#define GF_EVENT_CALLOC_SIZE(cnt, len) \
+ (sizeof(struct gf_event) + (cnt * sizeof(struct iovec)) + len)
/**
* assign the base address of the IO vector to the correct memory
o * area and set it's addressable length.
*/
-#define GF_EVENT_ASSIGN_IOVEC(vec, event, len, pos) \
- do { \
- vec->iov_base = ((char *)event) + \
- sizeof (struct gf_event) + \
- (event->count * sizeof (struct iovec)) + pos; \
- vec->iov_len = len; \
- pos += len; \
- } while (0)
+#define GF_EVENT_ASSIGN_IOVEC(vec, event, len, pos) \
+ do { \
+ vec->iov_base = ((char *)event) + sizeof(struct gf_event) + \
+ (event->count * sizeof(struct iovec)) + pos; \
+ vec->iov_len = len; \
+ pos += len; \
+ } while (0)
typedef enum gf_changelog_conn_state {
- GF_CHANGELOG_CONN_STATE_PENDING = 0,
- GF_CHANGELOG_CONN_STATE_ACCEPTED,
- GF_CHANGELOG_CONN_STATE_DISCONNECTED,
+ GF_CHANGELOG_CONN_STATE_PENDING = 0,
+ GF_CHANGELOG_CONN_STATE_ACCEPTED,
+ GF_CHANGELOG_CONN_STATE_DISCONNECTED,
} gf_changelog_conn_state_t;
/**
@@ -107,153 +107,149 @@ typedef enum gf_changelog_conn_state {
* notifications are streamed.
*/
typedef struct gf_changelog {
- gf_lock_t statelock;
- gf_changelog_conn_state_t connstate;
+ gf_lock_t statelock;
+ gf_changelog_conn_state_t connstate;
- xlator_t *this;
+ xlator_t *this;
- struct list_head list; /* list of instances */
+ struct list_head list; /* list of instances */
- char brick[PATH_MAX]; /* brick path for this end-point */
+ char brick[PATH_MAX]; /* brick path for this end-point */
- changelog_rpc_t grpc; /* rpc{-clnt,svc} for this brick */
-#define RPC_PROBER(ent) ent->grpc.rpc
-#define RPC_REBORP(ent) ent->grpc.svc
-#define RPC_SOCK(ent) ent->grpc.sock
+ changelog_rpc_t grpc; /* rpc{-clnt,svc} for this brick */
+#define RPC_PROBER(ent) ent->grpc.rpc
+#define RPC_REBORP(ent) ent->grpc.svc
+#define RPC_SOCK(ent) ent->grpc.sock
- unsigned int notify; /* notification flag(s) */
+ unsigned int notify; /* notification flag(s) */
- FINI *fini; /* destructor callback */
- CALLBACK *callback; /* event callback dispatcher */
- CONNECT *connected; /* connect callback */
- DISCONNECT *disconnected; /* disconnection callback */
+ FINI *fini; /* destructor callback */
+ CALLBACK *callback; /* event callback dispatcher */
+ CONNECT *connected; /* connect callback */
+ DISCONNECT *disconnected; /* disconnection callback */
- void *ptr; /* owner specific private data */
- xlator_t *invokerxl; /* consumers _this_, if valid,
- assigned to THIS before cbk is
- invoked */
+ void *ptr; /* owner specific private data */
+ xlator_t *invokerxl; /* consumers _this_, if valid,
+ assigned to THIS before cbk is
+ invoked */
- gf_boolean_t ordered;
+ gf_boolean_t ordered;
- void (*queueevent) (struct gf_event_list *, struct gf_event *);
- void (*pickevent) (struct gf_event_list *, struct gf_event **);
+ void (*queueevent)(struct gf_event_list *, struct gf_event *);
+ void (*pickevent)(struct gf_event_list *, struct gf_event **);
- struct gf_event_list event;
+ struct gf_event_list event;
} gf_changelog_t;
static inline int
-gf_changelog_filter_check (gf_changelog_t *entry, changelog_event_t *event)
+gf_changelog_filter_check(gf_changelog_t *entry, changelog_event_t *event)
{
- if (event->ev_type & entry->notify)
- return 1;
- return 0;
+ if (event->ev_type & entry->notify)
+ return 1;
+ return 0;
}
-#define GF_NEED_ORDERED_EVENTS(ent) (ent->ordered == _gf_true)
+#define GF_NEED_ORDERED_EVENTS(ent) (ent->ordered == _gf_true)
/** private structure */
typedef struct gf_private {
- pthread_mutex_t lock; /* protects ->connections, cleanups */
- pthread_cond_t cond;
+ pthread_mutex_t lock; /* protects ->connections, cleanups */
+ pthread_cond_t cond;
- void *api; /* pointer for API access */
+ void *api; /* pointer for API access */
- pthread_t poller; /* event poller thread */
- pthread_t connectionjanitor; /* connection cleaner */
+ pthread_t poller; /* event poller thread */
+ pthread_t connectionjanitor; /* connection cleaner */
- struct list_head connections; /* list of connections */
- struct list_head cleanups; /* list of connection to be
- cleaned up */
+ struct list_head connections; /* list of connections */
+ struct list_head cleanups; /* list of connection to be
+ cleaned up */
} gf_private_t;
-#define GF_CHANGELOG_GET_API_PTR(this) (((gf_private_t *) this->private)->api)
+#define GF_CHANGELOG_GET_API_PTR(this) (((gf_private_t *)this->private)->api)
/**
* upcall: invoke callback with _correct_ THIS
*/
-#define GF_CHANGELOG_INVOKE_CBK(this, cbk, brick, args ...) \
- do { \
- xlator_t *old_this = NULL; \
- xlator_t *invokerxl = NULL; \
- \
- invokerxl = entry->invokerxl; \
- old_this = this; \
- \
- if (invokerxl) { \
- THIS = invokerxl; \
- } \
- \
- cbk (invokerxl, brick, args); \
- THIS = old_this; \
- \
- } while (0)
-
-#define SAVE_THIS(xl) \
- do { \
- old_this = xl; \
- THIS = master; \
- } while (0)
-
-#define RESTORE_THIS() \
- do { \
- if (old_this) \
- THIS = old_this; \
- } while (0)
+#define GF_CHANGELOG_INVOKE_CBK(this, cbk, brick, args...) \
+ do { \
+ xlator_t *old_this = NULL; \
+ xlator_t *invokerxl = NULL; \
+ \
+ invokerxl = entry->invokerxl; \
+ old_this = this; \
+ \
+ if (invokerxl) { \
+ THIS = invokerxl; \
+ } \
+ \
+ cbk(invokerxl, brick, args); \
+ THIS = old_this; \
+ \
+ } while (0)
+
+#define SAVE_THIS(xl) \
+ do { \
+ old_this = xl; \
+ THIS = master; \
+ } while (0)
+
+#define RESTORE_THIS() \
+ do { \
+ if (old_this) \
+ THIS = old_this; \
+ } while (0)
/** APIs and the rest */
void *
-gf_changelog_process (void *data);
-
-ssize_t
-gf_changelog_read_path (int fd, char *buffer, size_t bufsize);
+gf_changelog_process(void *data);
void
-gf_rfc3986_encode_space_newline (unsigned char *s, char *enc, char *estr);
+gf_rfc3986_encode_space_newline(unsigned char *s, char *enc, char *estr);
size_t
-gf_changelog_write (int fd, char *buffer, size_t len);
+gf_changelog_write(int fd, char *buffer, size_t len);
ssize_t
-gf_readline (int fd, void *vptr, size_t maxlen);
+gf_readline(int fd, void *vptr, size_t maxlen);
int
-gf_ftruncate (int fd, off_t length);
+gf_ftruncate(int fd, off_t length);
off_t
-gf_lseek (int fd, off_t offset, int whence);
+gf_lseek(int fd, off_t offset, int whence);
int
-gf_changelog_consume (xlator_t *this,
- gf_changelog_journal_t *jnl,
- char *from_path, gf_boolean_t no_publish);
+gf_changelog_consume(xlator_t *this, gf_changelog_journal_t *jnl,
+ char *from_path, gf_boolean_t no_publish);
int
-gf_changelog_publish (xlator_t *this,
- gf_changelog_journal_t *jnl, char *from_path);
+gf_changelog_publish(xlator_t *this, gf_changelog_journal_t *jnl,
+ char *from_path);
int
-gf_thread_cleanup (xlator_t *this, pthread_t thread);
+gf_thread_cleanup(xlator_t *this, pthread_t thread);
void *
-gf_changelog_callback_invoker (void *arg);
+gf_changelog_callback_invoker(void *arg);
int
-gf_cleanup_event (xlator_t *, struct gf_event_list *);
+gf_cleanup_event(xlator_t *, struct gf_event_list *);
/* (un)ordered event queueing */
void
-queue_ordered_event (struct gf_event_list *, struct gf_event *);
+queue_ordered_event(struct gf_event_list *, struct gf_event *);
void
-queue_unordered_event (struct gf_event_list *, struct gf_event *);
+queue_unordered_event(struct gf_event_list *, struct gf_event *);
/* (un)ordered event picking */
void
-pick_event_ordered (struct gf_event_list *, struct gf_event **);
+pick_event_ordered(struct gf_event_list *, struct gf_event **);
void
-pick_event_unordered (struct gf_event_list *, struct gf_event **);
+pick_event_unordered(struct gf_event_list *, struct gf_event **);
/* connection janitor thread */
void *
-gf_changelog_connection_janitor (void *);
+gf_changelog_connection_janitor(void *);
#endif
diff --git a/xlators/features/changelog/lib/src/gf-changelog-journal-handler.c b/xlators/features/changelog/lib/src/gf-changelog-journal-handler.c
index 07de76e7cda..7f6e2329e71 100644
--- a/xlators/features/changelog/lib/src/gf-changelog-journal-handler.c
+++ b/xlators/features/changelog/lib/src/gf-changelog-journal-handler.c
@@ -8,11 +8,11 @@
cases as published by the Free Software Foundation.
*/
-#include "compat-uuid.h"
-#include "globals.h"
-#include "glusterfs.h"
-#include "syscall.h"
-#include "compat-errno.h"
+#include <glusterfs/compat-uuid.h>
+#include <glusterfs/globals.h>
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/syscall.h>
+#include <glusterfs/compat-errno.h>
#include "gf-changelog-helpers.h"
@@ -25,112 +25,107 @@
extern int byebye;
-enum changelog_versions {
- VERSION_1_1 = 0,
- VERSION_1_2 = 1
-};
+enum changelog_versions { VERSION_1_1 = 0, VERSION_1_2 = 1 };
/**
* number of gfid records after fop number
*/
-int nr_gfids[2][GF_FOP_MAXVALUE] = {
- {
- [GF_FOP_MKNOD] = 1,
- [GF_FOP_MKDIR] = 1,
- [GF_FOP_UNLINK] = 1,
- [GF_FOP_RMDIR] = 1,
- [GF_FOP_SYMLINK] = 1,
- [GF_FOP_RENAME] = 2,
- [GF_FOP_LINK] = 1,
- [GF_FOP_CREATE] = 1,
- },
- {
- [GF_FOP_MKNOD] = 1,
- [GF_FOP_MKDIR] = 1,
- [GF_FOP_UNLINK] = 2,
- [GF_FOP_RMDIR] = 2,
- [GF_FOP_SYMLINK] = 1,
- [GF_FOP_RENAME] = 2,
- [GF_FOP_LINK] = 1,
- [GF_FOP_CREATE] = 1,
- }
-};
-
-int nr_extra_recs[2][GF_FOP_MAXVALUE] = {
- {
- [GF_FOP_MKNOD] = 3,
- [GF_FOP_MKDIR] = 3,
- [GF_FOP_UNLINK] = 0,
- [GF_FOP_RMDIR] = 0,
- [GF_FOP_SYMLINK] = 0,
- [GF_FOP_RENAME] = 0,
- [GF_FOP_LINK] = 0,
- [GF_FOP_CREATE] = 3,
- },
- {
- [GF_FOP_MKNOD] = 3,
- [GF_FOP_MKDIR] = 3,
- [GF_FOP_UNLINK] = 0,
- [GF_FOP_RMDIR] = 0,
- [GF_FOP_SYMLINK] = 0,
- [GF_FOP_RENAME] = 0,
- [GF_FOP_LINK] = 0,
- [GF_FOP_CREATE] = 3,
- }
-};
+int nr_gfids[2][GF_FOP_MAXVALUE] = {{
+ [GF_FOP_MKNOD] = 1,
+ [GF_FOP_MKDIR] = 1,
+ [GF_FOP_UNLINK] = 1,
+ [GF_FOP_RMDIR] = 1,
+ [GF_FOP_SYMLINK] = 1,
+ [GF_FOP_RENAME] = 2,
+ [GF_FOP_LINK] = 1,
+ [GF_FOP_CREATE] = 1,
+ },
+ {
+ [GF_FOP_MKNOD] = 1,
+ [GF_FOP_MKDIR] = 1,
+ [GF_FOP_UNLINK] = 2,
+ [GF_FOP_RMDIR] = 2,
+ [GF_FOP_SYMLINK] = 1,
+ [GF_FOP_RENAME] = 2,
+ [GF_FOP_LINK] = 1,
+ [GF_FOP_CREATE] = 1,
+ }};
+
+int nr_extra_recs[2][GF_FOP_MAXVALUE] = {{
+ [GF_FOP_MKNOD] = 3,
+ [GF_FOP_MKDIR] = 3,
+ [GF_FOP_UNLINK] = 0,
+ [GF_FOP_RMDIR] = 0,
+ [GF_FOP_SYMLINK] = 0,
+ [GF_FOP_RENAME] = 0,
+ [GF_FOP_LINK] = 0,
+ [GF_FOP_CREATE] = 3,
+ },
+ {
+ [GF_FOP_MKNOD] = 3,
+ [GF_FOP_MKDIR] = 3,
+ [GF_FOP_UNLINK] = 0,
+ [GF_FOP_RMDIR] = 0,
+ [GF_FOP_SYMLINK] = 0,
+ [GF_FOP_RENAME] = 0,
+ [GF_FOP_LINK] = 0,
+ [GF_FOP_CREATE] = 3,
+ }};
static char *
-binary_to_ascii (uuid_t uuid)
+binary_to_ascii(uuid_t uuid)
{
- return uuid_utoa (uuid);
+ return uuid_utoa(uuid);
}
static char *
-conv_noop (char *ptr) { return ptr; }
-
-#define VERIFY_SEPARATOR(ptr, plen, perr) \
- { \
- if (*(ptr + plen) != '\0') { \
- perr = 1; \
- break; \
- } \
- }
+conv_noop(char *ptr)
+{
+ return ptr;
+}
-#define MOVER_MOVE(mover, nleft, bytes) \
- { \
- mover += bytes; \
- nleft -= bytes; \
- } \
-
-#define PARSE_GFID(mov, ptr, le, fn, perr) \
- { \
- VERIFY_SEPARATOR (mov, le, perr); \
- ptr = fn (mov); \
- if (!ptr) { \
- perr = 1; \
- break; \
- } \
- }
+#define VERIFY_SEPARATOR(ptr, plen, perr) \
+ { \
+ if (*(ptr + plen) != '\0') { \
+ perr = 1; \
+ break; \
+ } \
+ }
-#define FILL_AND_MOVE(pt, buf, of, mo, nl, le) \
- { \
- GF_CHANGELOG_FILL_BUFFER (pt, buf, of, strlen (pt)); \
- MOVER_MOVE (mo, nl, le); \
- }
+#define MOVER_MOVE(mover, nleft, bytes) \
+ { \
+ mover += bytes; \
+ nleft -= bytes; \
+ }
+
+#define PARSE_GFID(mov, ptr, le, fn, perr) \
+ { \
+ VERIFY_SEPARATOR(mov, le, perr); \
+ ptr = fn(mov); \
+ if (!ptr) { \
+ perr = 1; \
+ break; \
+ } \
+ }
+#define FILL_AND_MOVE(pt, buf, of, mo, nl, le) \
+ { \
+ GF_CHANGELOG_FILL_BUFFER(pt, buf, of, strlen(pt)); \
+ MOVER_MOVE(mo, nl, le); \
+ }
-#define PARSE_GFID_MOVE(ptr, uuid, mover, nleft, perr) \
- { \
- memcpy (uuid, mover, sizeof (uuid_t)); \
- ptr = binary_to_ascii (uuid); \
- if (!ptr) { \
- perr = 1; \
- break; \
- } \
- MOVER_MOVE (mover, nleft, sizeof (uuid_t)); \
- } \
+#define PARSE_GFID_MOVE(ptr, uuid, mover, nleft, perr) \
+ { \
+ memcpy(uuid, mover, sizeof(uuid_t)); \
+ ptr = binary_to_ascii(uuid); \
+ if (!ptr) { \
+ perr = 1; \
+ break; \
+ } \
+ MOVER_MOVE(mover, nleft, sizeof(uuid_t)); \
+ }
-#define LINE_BUFSIZE (3*PATH_MAX) /* enough buffer for extra chars too */
+#define LINE_BUFSIZE (3 * PATH_MAX) /* enough buffer for extra chars too */
/**
* using mmap() makes parsing easy. fgets() cannot be used here as
@@ -145,111 +140,107 @@ conv_noop (char *ptr) { return ptr; }
*/
static int
-gf_changelog_parse_binary (xlator_t *this,
- gf_changelog_journal_t *jnl,
- int from_fd, int to_fd,
- size_t start_offset, struct stat *stbuf,
- int version_idx)
+gf_changelog_parse_binary(xlator_t *this, gf_changelog_journal_t *jnl,
+ int from_fd, int to_fd, size_t start_offset,
+ struct stat *stbuf, int version_idx)
{
- int ret = -1;
- off_t off = 0;
- off_t nleft = 0;
- uuid_t uuid = {0,};
- char *ptr = NULL;
- char *bname_start = NULL;
- char *bname_end = NULL;
- char *mover = NULL;
- void *start = NULL;
- char current_mover = ' ';
- size_t blen = 0;
- int parse_err = 0;
- char *ascii = NULL;
-
- ascii = GF_CALLOC (LINE_BUFSIZE, sizeof(char), gf_common_mt_char);
-
- nleft = stbuf->st_size;
-
- start = mmap (NULL, nleft, PROT_READ, MAP_PRIVATE, from_fd, 0);
- if (start == MAP_FAILED) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_MMAP_FAILED,
- "mmap() error");
- goto out;
- }
-
- mover = start;
+ int ret = -1;
+ off_t off = 0;
+ off_t nleft = 0;
+ uuid_t uuid = {
+ 0,
+ };
+ char *ptr = NULL;
+ char *bname_start = NULL;
+ char *bname_end = NULL;
+ char *mover = NULL;
+ void *start = NULL;
+ char current_mover = ' ';
+ size_t blen = 0;
+ int parse_err = 0;
+ char *ascii = NULL;
+
+ ascii = GF_CALLOC(LINE_BUFSIZE, sizeof(char), gf_common_mt_char);
+
+ nleft = stbuf->st_size;
+
+ start = mmap(NULL, nleft, PROT_READ, MAP_PRIVATE, from_fd, 0);
+ if (start == MAP_FAILED) {
+ gf_msg(this->name, GF_LOG_ERROR, errno, CHANGELOG_LIB_MSG_MMAP_FAILED,
+ "mmap() error");
+ goto out;
+ }
- MOVER_MOVE (mover, nleft, start_offset);
+ mover = start;
- while (nleft > 0) {
+ MOVER_MOVE(mover, nleft, start_offset);
- off = blen = 0;
- ptr = bname_start = bname_end = NULL;
+ while (nleft > 0) {
+ off = blen = 0;
+ ptr = bname_start = bname_end = NULL;
- current_mover = *mover;
+ current_mover = *mover;
- switch (current_mover) {
- case 'D':
- case 'M':
- MOVER_MOVE (mover, nleft, 1);
- PARSE_GFID_MOVE (ptr, uuid, mover, nleft, parse_err);
+ switch (current_mover) {
+ case 'D':
+ case 'M':
+ MOVER_MOVE(mover, nleft, 1);
+ PARSE_GFID_MOVE(ptr, uuid, mover, nleft, parse_err);
- break;
+ break;
- case 'E':
- MOVER_MOVE (mover, nleft, 1);
- PARSE_GFID_MOVE (ptr, uuid, mover, nleft, parse_err);
+ case 'E':
+ MOVER_MOVE(mover, nleft, 1);
+ PARSE_GFID_MOVE(ptr, uuid, mover, nleft, parse_err);
- bname_start = mover;
- bname_end = strchr (mover, '\n');
- if (bname_end == NULL) {
- parse_err = 1;
- break;
- }
+ bname_start = mover;
+ bname_end = strchr(mover, '\n');
+ if (bname_end == NULL) {
+ parse_err = 1;
+ break;
+ }
- blen = bname_end - bname_start;
- MOVER_MOVE (mover, nleft, blen);
+ blen = bname_end - bname_start;
+ MOVER_MOVE(mover, nleft, blen);
- break;
+ break;
- default:
- parse_err = 1;
- }
+ default:
+ parse_err = 1;
+ }
- if (parse_err)
- break;
+ if (parse_err)
+ break;
+
+ GF_CHANGELOG_FILL_BUFFER(&current_mover, ascii, off, 1);
+ GF_CHANGELOG_FILL_BUFFER(" ", ascii, off, 1);
+ GF_CHANGELOG_FILL_BUFFER(ptr, ascii, off, strlen(ptr));
+ if (blen)
+ GF_CHANGELOG_FILL_BUFFER(bname_start, ascii, off, blen);
+ GF_CHANGELOG_FILL_BUFFER("\n", ascii, off, 1);
+
+ if (gf_changelog_write(to_fd, ascii, off) != off) {
+ gf_msg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_LIB_MSG_ASCII_ERROR,
+ "processing binary changelog failed due to "
+ " error in writing ascii change");
+ break;
+ }
- GF_CHANGELOG_FILL_BUFFER (&current_mover, ascii, off, 1);
- GF_CHANGELOG_FILL_BUFFER (" ", ascii, off, 1);
- GF_CHANGELOG_FILL_BUFFER (ptr, ascii, off, strlen (ptr));
- if (blen)
- GF_CHANGELOG_FILL_BUFFER (bname_start,
- ascii, off, blen);
- GF_CHANGELOG_FILL_BUFFER ("\n", ascii, off, 1);
-
- if (gf_changelog_write (to_fd, ascii, off) != off) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_ASCII_ERROR,
- "processing binary changelog failed due to "
- " error in writing ascii change");
- break;
- }
+ MOVER_MOVE(mover, nleft, 1);
+ }
- MOVER_MOVE (mover, nleft, 1);
- }
+ if ((nleft == 0) && (!parse_err))
+ ret = 0;
- if ((nleft == 0) && (!parse_err))
- ret = 0;
-
- if (munmap (start, stbuf->st_size))
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_MUNMAP_FAILED,
- "munmap() error");
- out:
- if (ascii)
- GF_FREE (ascii);
- return ret;
+ if (munmap(start, stbuf->st_size))
+ gf_msg(this->name, GF_LOG_ERROR, errno, CHANGELOG_LIB_MSG_MUNMAP_FAILED,
+ "munmap() error");
+out:
+ if (ascii)
+ GF_FREE(ascii);
+ return ret;
}
/**
@@ -258,798 +249,781 @@ gf_changelog_parse_binary (xlator_t *this,
* - use fop name rather than fop number
*/
static int
-gf_changelog_parse_ascii (xlator_t *this,
- gf_changelog_journal_t *jnl,
- int from_fd, int to_fd,
- size_t start_offset, struct stat *stbuf,
- int version_idx)
+gf_changelog_parse_ascii(xlator_t *this, gf_changelog_journal_t *jnl,
+ int from_fd, int to_fd, size_t start_offset,
+ struct stat *stbuf, int version_idx)
{
- int ng = 0;
- int ret = -1;
- int fop = 0;
- int len = 0;
- off_t off = 0;
- off_t nleft = 0;
- char *ptr = NULL;
- char *eptr = NULL;
- void *start = NULL;
- char *mover = NULL;
- int parse_err = 0;
- char current_mover = ' ';
- char *ascii = NULL;
- const char *fopname = NULL;
-
- ascii = GF_CALLOC (LINE_BUFSIZE, sizeof(char), gf_common_mt_char);
-
- nleft = stbuf->st_size;
-
- start = mmap (NULL, nleft, PROT_READ, MAP_PRIVATE, from_fd, 0);
- if (start == MAP_FAILED) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_MMAP_FAILED,
- "mmap() error");
- goto out;
- }
+ int ng = 0;
+ int ret = -1;
+ int fop = 0;
+ int len = 0;
+ off_t off = 0;
+ off_t nleft = 0;
+ char *ptr = NULL;
+ char *eptr = NULL;
+ void *start = NULL;
+ char *mover = NULL;
+ int parse_err = 0;
+ char current_mover = ' ';
+ char *ascii = NULL;
+ const char *fopname = NULL;
+
+ ascii = GF_CALLOC(LINE_BUFSIZE, sizeof(char), gf_common_mt_char);
+
+ nleft = stbuf->st_size;
+
+ start = mmap(NULL, nleft, PROT_READ, MAP_PRIVATE, from_fd, 0);
+ if (start == MAP_FAILED) {
+ gf_msg(this->name, GF_LOG_ERROR, errno, CHANGELOG_LIB_MSG_MMAP_FAILED,
+ "mmap() error");
+ goto out;
+ }
- mover = start;
+ mover = start;
- MOVER_MOVE (mover, nleft, start_offset);
+ MOVER_MOVE(mover, nleft, start_offset);
- while (nleft > 0) {
- off = 0;
- current_mover = *mover;
+ while (nleft > 0) {
+ off = 0;
+ current_mover = *mover;
- GF_CHANGELOG_FILL_BUFFER (&current_mover, ascii, off, 1);
- GF_CHANGELOG_FILL_BUFFER (" ", ascii, off, 1);
+ GF_CHANGELOG_FILL_BUFFER(&current_mover, ascii, off, 1);
+ GF_CHANGELOG_FILL_BUFFER(" ", ascii, off, 1);
- switch (current_mover) {
- case 'D':
- MOVER_MOVE (mover, nleft, 1);
+ switch (current_mover) {
+ case 'D':
+ MOVER_MOVE(mover, nleft, 1);
- /* target gfid */
- PARSE_GFID (mover, ptr, UUID_CANONICAL_FORM_LEN,
- conv_noop, parse_err);
- FILL_AND_MOVE(ptr, ascii, off,
- mover, nleft, UUID_CANONICAL_FORM_LEN);
- break;
- case 'M':
- MOVER_MOVE (mover, nleft, 1);
+ /* target gfid */
+ PARSE_GFID(mover, ptr, UUID_CANONICAL_FORM_LEN, conv_noop,
+ parse_err);
+ FILL_AND_MOVE(ptr, ascii, off, mover, nleft,
+ UUID_CANONICAL_FORM_LEN);
+ break;
+ case 'M':
+ MOVER_MOVE(mover, nleft, 1);
+
+ /* target gfid */
+ PARSE_GFID(mover, ptr, UUID_CANONICAL_FORM_LEN, conv_noop,
+ parse_err);
+ FILL_AND_MOVE(ptr, ascii, off, mover, nleft,
+ UUID_CANONICAL_FORM_LEN);
+ FILL_AND_MOVE(" ", ascii, off, mover, nleft, 1);
+
+ /* fop */
+ len = strlen(mover);
+ VERIFY_SEPARATOR(mover, len, parse_err);
+
+ fop = atoi(mover);
+ fopname = gf_fop_list[fop];
+ if (fopname == NULL) {
+ parse_err = 1;
+ break;
+ }
- /* target gfid */
- PARSE_GFID (mover, ptr, UUID_CANONICAL_FORM_LEN,
- conv_noop, parse_err);
- FILL_AND_MOVE (ptr, ascii, off,
- mover, nleft, UUID_CANONICAL_FORM_LEN);
- FILL_AND_MOVE (" ", ascii, off, mover, nleft, 1);
+ MOVER_MOVE(mover, nleft, len);
- /* fop */
- len = strlen (mover);
- VERIFY_SEPARATOR (mover, len, parse_err);
+ len = strlen(fopname);
+ GF_CHANGELOG_FILL_BUFFER(fopname, ascii, off, len);
- fop = atoi (mover);
- fopname = gf_fop_list[fop];
- if (fopname == NULL) {
- parse_err = 1;
- break;
- }
+ break;
- MOVER_MOVE (mover, nleft, len);
+ case 'E':
+ MOVER_MOVE(mover, nleft, 1);
+
+ /* target gfid */
+ PARSE_GFID(mover, ptr, UUID_CANONICAL_FORM_LEN, conv_noop,
+ parse_err);
+ FILL_AND_MOVE(ptr, ascii, off, mover, nleft,
+ UUID_CANONICAL_FORM_LEN);
+ FILL_AND_MOVE(" ", ascii, off, mover, nleft, 1);
+
+ /* fop */
+ len = strlen(mover);
+ VERIFY_SEPARATOR(mover, len, parse_err);
+
+ fop = atoi(mover);
+ fopname = gf_fop_list[fop];
+ if (fopname == NULL) {
+ parse_err = 1;
+ break;
+ }
- len = strlen (fopname);
- GF_CHANGELOG_FILL_BUFFER (fopname, ascii, off, len);
+ MOVER_MOVE(mover, nleft, len);
- break;
+ len = strlen(fopname);
+ GF_CHANGELOG_FILL_BUFFER(fopname, ascii, off, len);
- case 'E':
- MOVER_MOVE (mover, nleft, 1);
-
- /* target gfid */
- PARSE_GFID (mover, ptr, UUID_CANONICAL_FORM_LEN,
- conv_noop, parse_err);
- FILL_AND_MOVE (ptr, ascii, off,
- mover, nleft, UUID_CANONICAL_FORM_LEN);
- FILL_AND_MOVE (" ", ascii, off,
- mover, nleft, 1);
-
- /* fop */
- len = strlen (mover);
- VERIFY_SEPARATOR (mover, len, parse_err);
-
- fop = atoi (mover);
- fopname = gf_fop_list[fop];
- if (fopname == NULL) {
- parse_err = 1;
- break;
- }
-
- MOVER_MOVE (mover, nleft, len);
-
- len = strlen (fopname);
- GF_CHANGELOG_FILL_BUFFER (fopname, ascii, off, len);
-
- ng = nr_extra_recs[version_idx][fop];
- for (; ng > 0; ng--) {
- MOVER_MOVE (mover, nleft, 1);
- len = strlen (mover);
- VERIFY_SEPARATOR (mover, len, parse_err);
-
- GF_CHANGELOG_FILL_BUFFER (" ", ascii, off, 1);
- FILL_AND_MOVE (mover, ascii,
- off, mover, nleft, len);
- }
-
- /* pargfid + bname */
- ng = nr_gfids[version_idx][fop];
- while (ng-- > 0) {
- MOVER_MOVE (mover, nleft, 1);
- len = strlen (mover);
- if (!len) {
- MOVER_MOVE (mover, nleft, 1);
- continue;
- }
-
- GF_CHANGELOG_FILL_BUFFER (" ", ascii, off, 1);
-
- PARSE_GFID (mover, ptr, len,
- conv_noop, parse_err);
- eptr = calloc (3, strlen (ptr));
- if (!eptr) {
- parse_err = 1;
- break;
- }
-
- gf_rfc3986_encode_space_newline (
- (unsigned char *) ptr,
- eptr,
- jnl->rfc3986_space_newline);
- FILL_AND_MOVE (eptr, ascii, off,
- mover, nleft, len);
- free (eptr);
- }
+ ng = nr_extra_recs[version_idx][fop];
+ for (; ng > 0; ng--) {
+ MOVER_MOVE(mover, nleft, 1);
+ len = strlen(mover);
+ VERIFY_SEPARATOR(mover, len, parse_err);
- break;
- default:
- parse_err = 1;
+ GF_CHANGELOG_FILL_BUFFER(" ", ascii, off, 1);
+ FILL_AND_MOVE(mover, ascii, off, mover, nleft, len);
}
- if (parse_err)
+ /* pargfid + bname */
+ ng = nr_gfids[version_idx][fop];
+ while (ng-- > 0) {
+ MOVER_MOVE(mover, nleft, 1);
+ len = strlen(mover);
+ if (!len) {
+ MOVER_MOVE(mover, nleft, 1);
+ continue;
+ }
+
+ GF_CHANGELOG_FILL_BUFFER(" ", ascii, off, 1);
+
+ PARSE_GFID(mover, ptr, len, conv_noop, parse_err);
+ eptr = calloc(3, strlen(ptr));
+ if (!eptr) {
+ parse_err = 1;
break;
+ }
- GF_CHANGELOG_FILL_BUFFER ("\n", ascii, off, 1);
-
- if (gf_changelog_write (to_fd, ascii, off) != off) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_ASCII_ERROR,
- "processing ascii changelog failed due to "
- " error in writing change");
- break;
+ gf_rfc3986_encode_space_newline((unsigned char *)ptr, eptr,
+ jnl->rfc3986_space_newline);
+ FILL_AND_MOVE(eptr, ascii, off, mover, nleft, len);
+ free(eptr);
}
- MOVER_MOVE (mover, nleft, 1);
+ break;
+ default:
+ parse_err = 1;
+ }
+
+ if (parse_err)
+ break;
+
+ GF_CHANGELOG_FILL_BUFFER("\n", ascii, off, 1);
+ if (gf_changelog_write(to_fd, ascii, off) != off) {
+ gf_msg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_LIB_MSG_ASCII_ERROR,
+ "processing ascii changelog failed due to "
+ " error in writing change");
+ break;
}
- if ((nleft == 0) && (!parse_err))
- ret = 0;
+ MOVER_MOVE(mover, nleft, 1);
+ }
+
+ if ((nleft == 0) && (!parse_err))
+ ret = 0;
- if (munmap (start, stbuf->st_size))
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_MUNMAP_FAILED,
- "munmap() error");
+ if (munmap(start, stbuf->st_size))
+ gf_msg(this->name, GF_LOG_ERROR, errno, CHANGELOG_LIB_MSG_MUNMAP_FAILED,
+ "munmap() error");
- out:
- if (ascii)
- GF_FREE (ascii);
+out:
+ if (ascii)
+ GF_FREE(ascii);
- return ret;
+ return ret;
}
static int
-gf_changelog_decode (xlator_t *this, gf_changelog_journal_t *jnl,
- int from_fd, int to_fd, struct stat *stbuf, int *zerob)
+gf_changelog_decode(xlator_t *this, gf_changelog_journal_t *jnl, int from_fd,
+ int to_fd, struct stat *stbuf, int *zerob)
{
- int ret = -1;
- int encoding = -1;
- int major_version = -1;
- int minor_version = -1;
- int version_idx = -1;
- size_t elen = 0;
- char buffer[1024] = {0,};
-
- CHANGELOG_GET_HEADER_INFO (from_fd, buffer, 1024, encoding,
- major_version, minor_version, elen);
- if (encoding == -1) /* unknown encoding */
- goto out;
-
- if (major_version == -1) /* unknown major version */
- goto out;
-
- if (minor_version == -1) /* unknown minor version */
- goto out;
-
- if (!CHANGELOG_VALID_ENCODING (encoding))
- goto out;
-
- if (elen == stbuf->st_size) {
- *zerob = 1;
- goto out;
- }
+ int ret = -1;
+ int encoding = -1;
+ int major_version = -1;
+ int minor_version = -1;
+ int version_idx = -1;
+ size_t elen = 0;
+ char buffer[1024] = {
+ 0,
+ };
+
+ CHANGELOG_GET_HEADER_INFO(from_fd, buffer, sizeof(buffer), encoding,
+ major_version, minor_version, elen);
+ if (encoding == -1) /* unknown encoding */
+ goto out;
+
+ if (major_version == -1) /* unknown major version */
+ goto out;
+
+ if (minor_version == -1) /* unknown minor version */
+ goto out;
+
+ if (!CHANGELOG_VALID_ENCODING(encoding))
+ goto out;
+
+ if (elen == stbuf->st_size) {
+ *zerob = 1;
+ goto out;
+ }
- if (major_version == 1 && minor_version == 1) {
- version_idx = VERSION_1_1;
- } else if (major_version == 1 && minor_version == 2) {
- version_idx = VERSION_1_2;
- }
+ if (major_version == 1 && minor_version == 1) {
+ version_idx = VERSION_1_1;
+ } else if (major_version == 1 && minor_version == 2) {
+ version_idx = VERSION_1_2;
+ }
- if (version_idx == -1) /* unknown version number */
- goto out;
+ if (version_idx == -1) /* unknown version number */
+ goto out;
- /**
- * start processing after the header
- */
- if (sys_lseek (from_fd, elen, SEEK_SET) < 0) {
- goto out;
- }
- switch (encoding) {
+ /**
+ * start processing after the header
+ */
+ if (sys_lseek(from_fd, elen, SEEK_SET) < 0) {
+ goto out;
+ }
+ switch (encoding) {
case CHANGELOG_ENCODE_BINARY:
- /**
- * this ideally should have been a part of changelog-encoders.c
- * (ie. part of the changelog translator).
- */
- ret = gf_changelog_parse_binary (this, jnl, from_fd,
- to_fd, elen, stbuf,
- version_idx);
- break;
+ /**
+ * this ideally should have been a part of changelog-encoders.c
+ * (ie. part of the changelog translator).
+ */
+ ret = gf_changelog_parse_binary(this, jnl, from_fd, to_fd, elen,
+ stbuf, version_idx);
+ break;
case CHANGELOG_ENCODE_ASCII:
- ret = gf_changelog_parse_ascii (this, jnl, from_fd,
- to_fd, elen, stbuf,
- version_idx);
- break;
- }
+ ret = gf_changelog_parse_ascii(this, jnl, from_fd, to_fd, elen,
+ stbuf, version_idx);
+ break;
+ }
out:
- return ret;
+ return ret;
}
int
-gf_changelog_publish (xlator_t *this,
- gf_changelog_journal_t *jnl, char *from_path)
+gf_changelog_publish(xlator_t *this, gf_changelog_journal_t *jnl,
+ char *from_path)
{
- int ret = 0;
- char dest[PATH_MAX] = {0,};
- char to_path[PATH_MAX] = {0,};
- struct stat stbuf = {0,};
-
- (void) snprintf (to_path, PATH_MAX, "%s%s",
- jnl->jnl_current_dir, basename (from_path));
-
- /* handle zerob file that wont exist in current */
- ret = sys_stat (to_path, &stbuf);
- if (ret) {
- if (errno == ENOENT)
- ret = 0;
- goto out;
- }
+ int ret = 0;
+ char dest[PATH_MAX] = {
+ 0,
+ };
+ char to_path[PATH_MAX] = {
+ 0,
+ };
+ struct stat stbuf = {
+ 0,
+ };
+
+ if (snprintf(to_path, PATH_MAX, "%s%s", jnl->jnl_current_dir,
+ basename(from_path)) >= PATH_MAX)
+ return -1;
- (void) snprintf (dest, PATH_MAX, "%s%s",
- jnl->jnl_processing_dir, basename (from_path));
+ /* handle zerob file that won't exist in current */
+ ret = sys_stat(to_path, &stbuf);
+ if (ret) {
+ if (errno == ENOENT)
+ ret = 0;
+ goto out;
+ }
- ret = sys_rename (to_path, dest);
- if (ret) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_RENAME_FAILED,
- "error moving changelog to processing dir",
- "path=%s", to_path,
- NULL);
- }
+ if (snprintf(dest, PATH_MAX, "%s%s", jnl->jnl_processing_dir,
+ basename(from_path)) >= PATH_MAX)
+ return -1;
+
+ ret = sys_rename(to_path, dest);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_LIB_MSG_RENAME_FAILED, "from=%s", to_path, "to=%s",
+ dest, NULL);
+ }
out:
- return ret;
+ return ret;
}
int
-gf_changelog_consume (xlator_t *this,
- gf_changelog_journal_t *jnl,
- char *from_path, gf_boolean_t no_publish)
+gf_changelog_consume(xlator_t *this, gf_changelog_journal_t *jnl,
+ char *from_path, gf_boolean_t no_publish)
{
- int ret = -1;
- int fd1 = 0;
- int fd2 = 0;
- int zerob = 0;
- struct stat stbuf = {0,};
- char dest[PATH_MAX] = {0,};
- char to_path[PATH_MAX] = {0,};
-
- ret = sys_stat (from_path, &stbuf);
- if (ret || !S_ISREG(stbuf.st_mode)) {
- ret = -1;
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_STAT_FAILED,
- "stat failed on changelog file",
- "path=%s", from_path,
- NULL);
- goto out;
- }
+ int ret = -1;
+ int fd1 = 0;
+ int fd2 = 0;
+ int zerob = 0;
+ struct stat stbuf = {
+ 0,
+ };
+ char dest[PATH_MAX] = {
+ 0,
+ };
+ char to_path[PATH_MAX] = {
+ 0,
+ };
+
+ if (snprintf(to_path, PATH_MAX, "%s%s", jnl->jnl_current_dir,
+ basename(from_path)) >= PATH_MAX)
+ goto out;
+ if (snprintf(dest, PATH_MAX, "%s%s", jnl->jnl_processing_dir,
+ basename(from_path)) >= PATH_MAX)
+ goto out;
+
+ ret = sys_stat(from_path, &stbuf);
+ if (ret || !S_ISREG(stbuf.st_mode)) {
+ ret = -1;
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_LIB_MSG_STAT_FAILED,
+ "path=%s", from_path, NULL);
+ goto out;
+ }
- fd1 = open (from_path, O_RDONLY);
- if (fd1 < 0) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_OPEN_FAILED,
- "cannot open changelog file",
- "path=%s", from_path,
- NULL);
- goto out;
- }
+ fd1 = open(from_path, O_RDONLY);
+ if (fd1 < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_LIB_MSG_OPEN_FAILED,
+ "path=%s", from_path, NULL);
+ goto out;
+ }
- (void) snprintf (to_path, PATH_MAX, "%s%s",
- jnl->jnl_current_dir, basename (from_path));
- (void) snprintf (dest, PATH_MAX, "%s%s",
- jnl->jnl_processing_dir, basename (from_path));
-
- fd2 = open (to_path, O_CREAT | O_TRUNC | O_RDWR,
- S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
- if (fd2 < 0) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_OPEN_FAILED,
- "cannot create ascii changelog file",
- "path=%s", to_path,
- NULL);
+ fd2 = open(to_path, O_CREAT | O_TRUNC | O_RDWR,
+ S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
+ if (fd2 < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_LIB_MSG_OPEN_FAILED,
+ "path=%s", to_path, NULL);
+ goto close_fd;
+ } else {
+ ret = gf_changelog_decode(this, jnl, fd1, fd2, &stbuf, &zerob);
+
+ sys_close(fd2);
+
+ if (!ret) {
+ /* move it to processing on a successful
+ decode */
+ if (no_publish == _gf_true)
goto close_fd;
- } else {
- ret = gf_changelog_decode (this, jnl, fd1,
- fd2, &stbuf, &zerob);
-
- sys_close (fd2);
-
- if (!ret) {
- /* move it to processing on a successful
- decode */
- if (no_publish == _gf_true)
- goto close_fd;
- ret = sys_rename (to_path, dest);
- if (ret)
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_RENAME_FAILED,
- "error moving changelog to processing dir",
- "path=%s", to_path,
- NULL);
- }
+ ret = sys_rename(to_path, dest);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_LIB_MSG_RENAME_FAILED, "from=%s", to_path,
+ "to=%s", dest, NULL);
+ }
- /* remove it from .current if it's an empty file */
- if (zerob) {
- /* zerob changelogs must be unlinked */
- ret = sys_unlink (to_path);
- if (ret)
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_UNLINK_FAILED,
- "could not unlink empty changelog",
- "path=%s", to_path,
- NULL);
- }
+ /* remove it from .current if it's an empty file */
+ if (zerob) {
+ /* zerob changelogs must be unlinked */
+ ret = sys_unlink(to_path);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_LIB_MSG_UNLINK_FAILED, "name=empty changelog",
+ "path=%s", to_path, NULL);
}
+ }
- close_fd:
- sys_close (fd1);
+close_fd:
+ sys_close(fd1);
- out:
- return ret;
+out:
+ return ret;
}
void *
-gf_changelog_process (void *data)
+gf_changelog_process(void *data)
{
- xlator_t *this = NULL;
- gf_changelog_journal_t *jnl = NULL;
- gf_changelog_entry_t *entry = NULL;
- gf_changelog_processor_t *jnl_proc = NULL;
-
- jnl = data;
- jnl_proc = jnl->jnl_proc;
- THIS = jnl->this;
- this = jnl->this;
-
- while (1) {
- pthread_mutex_lock (&jnl_proc->lock);
- {
- while (list_empty (&jnl_proc->entries)) {
- jnl_proc->waiting = _gf_true;
- pthread_cond_wait
- (&jnl_proc->cond, &jnl_proc->lock);
- }
-
- entry = list_first_entry (&jnl_proc->entries,
- gf_changelog_entry_t, list);
- if (entry)
- list_del (&entry->list);
-
- jnl_proc->waiting = _gf_false;
- }
- pthread_mutex_unlock (&jnl_proc->lock);
+ xlator_t *this = NULL;
+ gf_changelog_journal_t *jnl = NULL;
+ gf_changelog_entry_t *entry = NULL;
+ gf_changelog_processor_t *jnl_proc = NULL;
+
+ jnl = data;
+ jnl_proc = jnl->jnl_proc;
+ THIS = jnl->this;
+ this = jnl->this;
+
+ while (1) {
+ pthread_mutex_lock(&jnl_proc->lock);
+ {
+ while (list_empty(&jnl_proc->entries)) {
+ jnl_proc->waiting = _gf_true;
+ pthread_cond_wait(&jnl_proc->cond, &jnl_proc->lock);
+ }
- if (entry) {
- (void) gf_changelog_consume (this, jnl,
- entry->path, _gf_false);
- GF_FREE (entry);
- }
+ entry = list_first_entry(&jnl_proc->entries, gf_changelog_entry_t,
+ list);
+ if (entry)
+ list_del(&entry->list);
+
+ jnl_proc->waiting = _gf_false;
}
+ pthread_mutex_unlock(&jnl_proc->lock);
- return NULL;
+ if (entry) {
+ (void)gf_changelog_consume(this, jnl, entry->path, _gf_false);
+ GF_FREE(entry);
+ }
+ }
+
+ return NULL;
}
void
-gf_changelog_queue_journal (gf_changelog_processor_t *jnl_proc,
- changelog_event_t *event)
+gf_changelog_queue_journal(gf_changelog_processor_t *jnl_proc,
+ changelog_event_t *event)
{
- size_t len = 0;
- gf_changelog_entry_t *entry = NULL;
+ size_t len = 0;
+ gf_changelog_entry_t *entry = NULL;
- entry = GF_CALLOC (1, sizeof (gf_changelog_entry_t),
- gf_changelog_mt_libgfchangelog_entry_t);
- if (!entry)
- return;
- INIT_LIST_HEAD (&entry->list);
+ entry = GF_CALLOC(1, sizeof(gf_changelog_entry_t),
+ gf_changelog_mt_libgfchangelog_entry_t);
+ if (!entry)
+ return;
+ INIT_LIST_HEAD(&entry->list);
- len = strlen (event->u.journal.path);
- (void)memcpy (entry->path, event->u.journal.path, len+1);
- entry->path[len] = '\0';
+ len = strlen(event->u.journal.path);
+ (void)memcpy(entry->path, event->u.journal.path, len + 1);
+ entry->path[len] = '\0';
- pthread_mutex_lock (&jnl_proc->lock);
- {
- list_add_tail (&entry->list, &jnl_proc->entries);
- if (jnl_proc->waiting)
- pthread_cond_signal (&jnl_proc->cond);
- }
- pthread_mutex_unlock (&jnl_proc->lock);
+ pthread_mutex_lock(&jnl_proc->lock);
+ {
+ list_add_tail(&entry->list, &jnl_proc->entries);
+ if (jnl_proc->waiting)
+ pthread_cond_signal(&jnl_proc->cond);
+ }
+ pthread_mutex_unlock(&jnl_proc->lock);
- return;
+ return;
}
void
-gf_changelog_handle_journal (void *xl, char *brick,
- void *cbkdata, changelog_event_t *event)
+gf_changelog_handle_journal(void *xl, char *brick, void *cbkdata,
+ changelog_event_t *event)
{
- gf_changelog_journal_t *jnl = NULL;
- gf_changelog_processor_t *jnl_proc = NULL;
+ gf_changelog_journal_t *jnl = NULL;
+ gf_changelog_processor_t *jnl_proc = NULL;
- jnl = cbkdata;
- jnl_proc = jnl->jnl_proc;
+ jnl = cbkdata;
+ jnl_proc = jnl->jnl_proc;
- gf_changelog_queue_journal (jnl_proc, event);
+ gf_changelog_queue_journal(jnl_proc, event);
}
void
-gf_changelog_journal_disconnect (void *xl, char *brick, void *data)
+gf_changelog_journal_disconnect(void *xl, char *brick, void *data)
{
- gf_changelog_journal_t *jnl = NULL;
+ gf_changelog_journal_t *jnl = NULL;
- jnl = data;
+ jnl = data;
- pthread_spin_lock (&jnl->lock);
- {
- JNL_SET_API_STATE (jnl, JNL_API_DISCONNECTED);
- };
- pthread_spin_unlock (&jnl->lock);
+ pthread_spin_lock(&jnl->lock);
+ {
+ JNL_SET_API_STATE(jnl, JNL_API_DISCONNECTED);
+ };
+ pthread_spin_unlock(&jnl->lock);
}
void
-gf_changelog_journal_connect (void *xl, char *brick, void *data)
+gf_changelog_journal_connect(void *xl, char *brick, void *data)
{
- gf_changelog_journal_t *jnl = NULL;
+ gf_changelog_journal_t *jnl = NULL;
- jnl = data;
+ jnl = data;
- pthread_spin_lock (&jnl->lock);
- {
- JNL_SET_API_STATE (jnl, JNL_API_CONNECTED);
- };
- pthread_spin_unlock (&jnl->lock);
+ pthread_spin_lock(&jnl->lock);
+ {
+ JNL_SET_API_STATE(jnl, JNL_API_CONNECTED);
+ };
+ pthread_spin_unlock(&jnl->lock);
- return;
+ return;
}
void
-gf_changelog_cleanup_processor (gf_changelog_journal_t *jnl)
+gf_changelog_cleanup_processor(gf_changelog_journal_t *jnl)
{
- int ret = 0;
- xlator_t *this = NULL;
- gf_changelog_processor_t *jnl_proc = NULL;
-
- this = THIS;
- if (!this || !jnl || !jnl->jnl_proc)
- goto error_return;
-
- jnl_proc = jnl->jnl_proc;
-
- ret = gf_thread_cleanup (this, jnl_proc->processor);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_LIB_MSG_CLEANUP_ERROR,
- "failed to cleanup processor thread");
- goto error_return;
- }
+ int ret = 0;
+ xlator_t *this = NULL;
+ gf_changelog_processor_t *jnl_proc = NULL;
- (void)pthread_mutex_destroy (&jnl_proc->lock);
- (void)pthread_cond_destroy (&jnl_proc->cond);
+ this = THIS;
+ if (!this || !jnl || !jnl->jnl_proc)
+ goto error_return;
- GF_FREE (jnl_proc);
+ jnl_proc = jnl->jnl_proc;
- error_return:
- return;
+ ret = gf_thread_cleanup(this, jnl_proc->processor);
+ if (ret != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, CHANGELOG_LIB_MSG_CLEANUP_ERROR,
+ "failed to cleanup processor thread");
+ goto error_return;
+ }
+
+ (void)pthread_mutex_destroy(&jnl_proc->lock);
+ (void)pthread_cond_destroy(&jnl_proc->cond);
+
+ GF_FREE(jnl_proc);
+
+error_return:
+ return;
}
int
-gf_changelog_init_processor (gf_changelog_journal_t *jnl)
+gf_changelog_init_processor(gf_changelog_journal_t *jnl)
{
- int ret = -1;
- gf_changelog_processor_t *jnl_proc = NULL;
+ int ret = -1;
+ gf_changelog_processor_t *jnl_proc = NULL;
- jnl_proc = GF_CALLOC (1, sizeof (gf_changelog_processor_t),
- gf_changelog_mt_libgfchangelog_t);
- if (!jnl_proc)
- goto error_return;
-
- ret = pthread_mutex_init (&jnl_proc->lock, NULL);
- if (ret != 0)
- goto free_jnl_proc;
- ret = pthread_cond_init (&jnl_proc->cond, NULL);
- if (ret != 0)
- goto cleanup_mutex;
-
- INIT_LIST_HEAD (&jnl_proc->entries);
- jnl_proc->waiting = _gf_false;
- jnl->jnl_proc = jnl_proc;
-
- ret = gf_thread_create (&jnl_proc->processor,
- NULL, gf_changelog_process, jnl, "clogproc");
- if (ret != 0) {
- jnl->jnl_proc = NULL;
- goto cleanup_cond;
- }
+ jnl_proc = GF_CALLOC(1, sizeof(gf_changelog_processor_t),
+ gf_changelog_mt_libgfchangelog_t);
+ if (!jnl_proc)
+ goto error_return;
+
+ ret = pthread_mutex_init(&jnl_proc->lock, NULL);
+ if (ret != 0)
+ goto free_jnl_proc;
+ ret = pthread_cond_init(&jnl_proc->cond, NULL);
+ if (ret != 0)
+ goto cleanup_mutex;
+
+ INIT_LIST_HEAD(&jnl_proc->entries);
+ jnl_proc->waiting = _gf_false;
+ jnl->jnl_proc = jnl_proc;
+
+ ret = gf_thread_create(&jnl_proc->processor, NULL, gf_changelog_process,
+ jnl, "clogproc");
+ if (ret != 0) {
+ jnl->jnl_proc = NULL;
+ goto cleanup_cond;
+ }
- return 0;
+ return 0;
- cleanup_cond:
- (void) pthread_cond_destroy (&jnl_proc->cond);
- cleanup_mutex:
- (void) pthread_mutex_destroy (&jnl_proc->lock);
- free_jnl_proc:
- GF_FREE (jnl_proc);
- error_return:
- return -1;
+cleanup_cond:
+ (void)pthread_cond_destroy(&jnl_proc->cond);
+cleanup_mutex:
+ (void)pthread_mutex_destroy(&jnl_proc->lock);
+free_jnl_proc:
+ GF_FREE(jnl_proc);
+error_return:
+ return -1;
}
static void
-gf_changelog_cleanup_fds (gf_changelog_journal_t *jnl)
+gf_changelog_cleanup_fds(gf_changelog_journal_t *jnl)
{
- /* tracker fd */
- if (jnl->jnl_fd != -1)
- sys_close (jnl->jnl_fd);
- /* processing dir */
- if (jnl->jnl_dir)
- sys_closedir (jnl->jnl_dir);
-
- if (jnl->jnl_working_dir)
- free (jnl->jnl_working_dir); /* allocated by realpath */
+ /* tracker fd */
+ if (jnl->jnl_fd != -1)
+ sys_close(jnl->jnl_fd);
+ /* processing dir */
+ if (jnl->jnl_dir)
+ sys_closedir(jnl->jnl_dir);
+
+ if (jnl->jnl_working_dir)
+ free(jnl->jnl_working_dir); /* allocated by realpath */
}
static int
-gf_changelog_open_dirs (xlator_t *this, gf_changelog_journal_t *jnl)
+gf_changelog_open_dirs(xlator_t *this, gf_changelog_journal_t *jnl)
{
- int ret = -1;
- DIR *dir = NULL;
- int tracker_fd = 0;
- char tracker_path[PATH_MAX] = {0,};
-
- /* .current */
- (void) snprintf (jnl->jnl_current_dir, PATH_MAX,
- "%s/"GF_CHANGELOG_CURRENT_DIR"/",
- jnl->jnl_working_dir);
- ret = recursive_rmdir (jnl->jnl_current_dir);
- if (ret) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_FAILED_TO_RMDIR,
- "Failed to rmdir",
- "path=%s", jnl->jnl_current_dir,
- NULL);
- goto out;
- }
- ret = mkdir_p (jnl->jnl_current_dir, 0600, _gf_false);
- if (ret)
- goto out;
+ int ret = -1;
+ DIR *dir = NULL;
+ int tracker_fd = 0;
+ char tracker_path[PATH_MAX] = {
+ 0,
+ };
+
+ /* .current */
+ (void)snprintf(jnl->jnl_current_dir, PATH_MAX,
+ "%s/" GF_CHANGELOG_CURRENT_DIR "/", jnl->jnl_working_dir);
+ ret = recursive_rmdir(jnl->jnl_current_dir);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_LIB_MSG_FAILED_TO_RMDIR, "path=%s",
+ jnl->jnl_current_dir, NULL);
+ goto out;
+ }
+ ret = mkdir_p(jnl->jnl_current_dir, 0600, _gf_false);
+ if (ret)
+ goto out;
+
+ /* .processed */
+ (void)snprintf(jnl->jnl_processed_dir, PATH_MAX,
+ "%s/" GF_CHANGELOG_PROCESSED_DIR "/", jnl->jnl_working_dir);
+ ret = mkdir_p(jnl->jnl_processed_dir, 0600, _gf_false);
+ if (ret)
+ goto out;
+
+ /* .processing */
+ (void)snprintf(jnl->jnl_processing_dir, PATH_MAX,
+ "%s/" GF_CHANGELOG_PROCESSING_DIR "/", jnl->jnl_working_dir);
+ ret = recursive_rmdir(jnl->jnl_processing_dir);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_LIB_MSG_FAILED_TO_RMDIR, "path=%s",
+ jnl->jnl_processing_dir, NULL);
+ goto out;
+ }
- /* .processed */
- (void) snprintf (jnl->jnl_processed_dir, PATH_MAX,
- "%s/"GF_CHANGELOG_PROCESSED_DIR"/",
- jnl->jnl_working_dir);
- ret = mkdir_p (jnl->jnl_processed_dir, 0600, _gf_false);
- if (ret)
- goto out;
-
- /* .processing */
- (void) snprintf (jnl->jnl_processing_dir, PATH_MAX,
- "%s/"GF_CHANGELOG_PROCESSING_DIR"/",
- jnl->jnl_working_dir);
- ret = recursive_rmdir (jnl->jnl_processing_dir);
- if (ret) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_FAILED_TO_RMDIR,
- "Failed to rmdir",
- "path=%s", jnl->jnl_processing_dir,
- NULL);
- goto out;
- }
+ ret = mkdir_p(jnl->jnl_processing_dir, 0600, _gf_false);
+ if (ret)
+ goto out;
- ret = mkdir_p (jnl->jnl_processing_dir, 0600, _gf_false);
- if (ret)
- goto out;
-
- dir = sys_opendir (jnl->jnl_processing_dir);
- if (!dir) {
- gf_msg ("", GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_OPENDIR_ERROR,
- "opendir() error");
- goto out;
- }
+ dir = sys_opendir(jnl->jnl_processing_dir);
+ if (!dir) {
+ gf_msg("", GF_LOG_ERROR, errno, CHANGELOG_LIB_MSG_OPENDIR_ERROR,
+ "opendir() error");
+ goto out;
+ }
- jnl->jnl_dir = dir;
+ jnl->jnl_dir = dir;
- (void) snprintf (tracker_path, PATH_MAX,
- "%s/"GF_CHANGELOG_TRACKER, jnl->jnl_working_dir);
+ (void)snprintf(tracker_path, PATH_MAX, "%s/" GF_CHANGELOG_TRACKER,
+ jnl->jnl_working_dir);
- tracker_fd = open (tracker_path, O_CREAT | O_APPEND | O_RDWR,
- S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
- if (tracker_fd < 0) {
- sys_closedir (jnl->jnl_dir);
- ret = -1;
- goto out;
- }
+ tracker_fd = open(tracker_path, O_CREAT | O_APPEND | O_RDWR,
+ S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
+ if (tracker_fd < 0) {
+ sys_closedir(jnl->jnl_dir);
+ ret = -1;
+ goto out;
+ }
- jnl->jnl_fd = tracker_fd;
- ret = 0;
- out:
- return ret;
+ jnl->jnl_fd = tracker_fd;
+ ret = 0;
+out:
+ return ret;
}
int
-gf_changelog_init_history (xlator_t *this,
- gf_changelog_journal_t *jnl,
- char *brick_path)
+gf_changelog_init_history(xlator_t *this, gf_changelog_journal_t *jnl,
+ char *brick_path)
{
- int i = 0;
- int ret = 0;
- char hist_scratch_dir[PATH_MAX] = {0,};
+ int i = 0;
+ int ret = 0;
+ char hist_scratch_dir[PATH_MAX] = {
+ 0,
+ };
- jnl->hist_jnl = GF_CALLOC (1, sizeof (*jnl),
- gf_changelog_mt_libgfchangelog_t);
- if (!jnl->hist_jnl)
- goto error_return;
+ jnl->hist_jnl = GF_CALLOC(1, sizeof(*jnl),
+ gf_changelog_mt_libgfchangelog_t);
+ if (!jnl->hist_jnl)
+ goto error_return;
- jnl->hist_jnl->jnl_dir = NULL;
- jnl->hist_jnl->jnl_fd = -1;
+ jnl->hist_jnl->jnl_dir = NULL;
+ jnl->hist_jnl->jnl_fd = -1;
- (void) snprintf (hist_scratch_dir, PATH_MAX,
- "%s/"GF_CHANGELOG_HISTORY_DIR"/",
- jnl->jnl_working_dir);
+ (void)snprintf(hist_scratch_dir, PATH_MAX,
+ "%s/" GF_CHANGELOG_HISTORY_DIR "/", jnl->jnl_working_dir);
- ret = mkdir_p (hist_scratch_dir, 0600, _gf_false);
- if (ret)
- goto dealloc_hist;
-
- jnl->hist_jnl->jnl_working_dir = realpath (hist_scratch_dir, NULL);
- if (!jnl->hist_jnl->jnl_working_dir)
- goto dealloc_hist;
-
- ret = gf_changelog_open_dirs (this, jnl->hist_jnl);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_LIB_MSG_OPENDIR_ERROR,
- "could not create entries in history scratch dir");
- goto dealloc_hist;
- }
+ ret = mkdir_p(hist_scratch_dir, 0600, _gf_false);
+ if (ret)
+ goto dealloc_hist;
- (void) strncpy (jnl->hist_jnl->jnl_brickpath, brick_path, PATH_MAX-1);
- jnl->hist_jnl->jnl_brickpath[PATH_MAX-1] = 0;
+ jnl->hist_jnl->jnl_working_dir = realpath(hist_scratch_dir, NULL);
+ if (!jnl->hist_jnl->jnl_working_dir)
+ goto dealloc_hist;
- for (i = 0; i < 256; i++) {
- jnl->hist_jnl->rfc3986_space_newline[i] =
- (i == ' ' || i == '\n' || i == '%') ? 0 : i;
- }
+ ret = gf_changelog_open_dirs(this, jnl->hist_jnl);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, CHANGELOG_LIB_MSG_OPENDIR_ERROR,
+ "could not create entries in history scratch dir");
+ goto dealloc_hist;
+ }
- return 0;
+ if (snprintf(jnl->hist_jnl->jnl_brickpath, PATH_MAX, "%s", brick_path) >=
+ PATH_MAX)
+ goto dealloc_hist;
- dealloc_hist:
- GF_FREE (jnl->hist_jnl);
- jnl->hist_jnl = NULL;
- error_return:
- return -1;
+ for (i = 0; i < 256; i++) {
+ jnl->hist_jnl->rfc3986_space_newline[i] = (i == ' ' || i == '\n' ||
+ i == '%')
+ ? 0
+ : i;
+ }
+
+ return 0;
+
+dealloc_hist:
+ GF_FREE(jnl->hist_jnl);
+ jnl->hist_jnl = NULL;
+error_return:
+ return -1;
}
void
-gf_changelog_journal_fini (void *xl, char *brick, void *data)
+gf_changelog_journal_fini(void *xl, char *brick, void *data)
{
- gf_changelog_journal_t *jnl = NULL;
+ gf_changelog_journal_t *jnl = NULL;
- jnl = data;
+ jnl = data;
- gf_changelog_cleanup_processor (jnl);
+ gf_changelog_cleanup_processor(jnl);
- gf_changelog_cleanup_fds (jnl);
- if (jnl->hist_jnl)
- gf_changelog_cleanup_fds (jnl->hist_jnl);
+ gf_changelog_cleanup_fds(jnl);
+ if (jnl->hist_jnl)
+ gf_changelog_cleanup_fds(jnl->hist_jnl);
- GF_FREE (jnl);
+ GF_FREE(jnl);
}
void *
-gf_changelog_journal_init (void *xl, struct gf_brick_spec *brick)
+gf_changelog_journal_init(void *xl, struct gf_brick_spec *brick)
{
- int i = 0;
- int ret = 0;
- xlator_t *this = NULL;
- struct stat buf = {0,};
- char *scratch_dir = NULL;
- gf_changelog_journal_t *jnl = NULL;
-
- this = xl;
- scratch_dir = (char *) brick->ptr;
-
- jnl = GF_CALLOC (1, sizeof (gf_changelog_journal_t),
- gf_changelog_mt_libgfchangelog_t);
- if (!jnl)
- goto error_return;
-
- if (sys_stat (scratch_dir, &buf) && errno == ENOENT) {
- ret = mkdir_p (scratch_dir, 0600, _gf_true);
- if (ret)
- goto dealloc_private;
- }
-
- jnl->jnl_working_dir = realpath (scratch_dir, NULL);
- if (!jnl->jnl_working_dir)
- goto dealloc_private;
-
- ret = gf_changelog_open_dirs (this, jnl);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_LIB_MSG_OPENDIR_ERROR,
- "could not create entries in scratch dir");
- goto dealloc_private;
- }
+ int i = 0;
+ int ret = 0;
+ xlator_t *this = NULL;
+ struct stat buf = {
+ 0,
+ };
+ char *scratch_dir = NULL;
+ gf_changelog_journal_t *jnl = NULL;
+
+ this = xl;
+ scratch_dir = (char *)brick->ptr;
+
+ jnl = GF_CALLOC(1, sizeof(gf_changelog_journal_t),
+ gf_changelog_mt_libgfchangelog_t);
+ if (!jnl)
+ goto error_return;
+
+ if (snprintf(jnl->jnl_brickpath, PATH_MAX, "%s", brick->brick_path) >=
+ PATH_MAX)
+ goto dealloc_private;
+
+ if (sys_stat(scratch_dir, &buf) && errno == ENOENT) {
+ ret = mkdir_p(scratch_dir, 0600, _gf_true);
+ if (ret)
+ goto dealloc_private;
+ }
- (void) strncpy (jnl->jnl_brickpath, brick->brick_path, PATH_MAX-1);
- jnl->jnl_brickpath[PATH_MAX-1] = 0;
+ jnl->jnl_working_dir = realpath(scratch_dir, NULL);
+ if (!jnl->jnl_working_dir)
+ goto dealloc_private;
- /* RFC 3986 {de,en}coding */
- for (i = 0; i < 256; i++) {
- jnl->rfc3986_space_newline[i] =
- (i == ' ' || i == '\n' || i == '%') ? 0 : i;
- }
+ ret = gf_changelog_open_dirs(this, jnl);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, CHANGELOG_LIB_MSG_OPENDIR_ERROR,
+ "could not create entries in scratch dir");
+ goto dealloc_private;
+ }
- ret = gf_changelog_init_history (this, jnl, brick->brick_path);
- if (ret)
- goto cleanup_fds;
+ /* RFC 3986 {de,en}coding */
+ for (i = 0; i < 256; i++) {
+ jnl->rfc3986_space_newline[i] = (i == ' ' || i == '\n' || i == '%') ? 0
+ : i;
+ }
- /* initialize journal processor */
- jnl->this = this;
- ret = gf_changelog_init_processor (jnl);
- if (ret)
- goto cleanup_fds;
-
- JNL_SET_API_STATE (jnl, JNL_API_CONN_INPROGESS);
- ret = pthread_spin_init (&jnl->lock, 0);
- if (ret != 0)
- goto cleanup_processor;
- return jnl;
-
- cleanup_processor:
- gf_changelog_cleanup_processor (jnl);
- cleanup_fds:
- gf_changelog_cleanup_fds (jnl);
- if (jnl->hist_jnl)
- gf_changelog_cleanup_fds (jnl->hist_jnl);
- dealloc_private:
- GF_FREE (jnl);
- error_return:
- return NULL;
+ ret = gf_changelog_init_history(this, jnl, brick->brick_path);
+ if (ret)
+ goto cleanup_fds;
+
+ /* initialize journal processor */
+ jnl->this = this;
+ ret = gf_changelog_init_processor(jnl);
+ if (ret)
+ goto cleanup_fds;
+
+ JNL_SET_API_STATE(jnl, JNL_API_CONN_INPROGESS);
+ ret = pthread_spin_init(&jnl->lock, 0);
+ if (ret != 0)
+ goto cleanup_processor;
+ return jnl;
+
+cleanup_processor:
+ gf_changelog_cleanup_processor(jnl);
+cleanup_fds:
+ gf_changelog_cleanup_fds(jnl);
+ if (jnl->hist_jnl)
+ gf_changelog_cleanup_fds(jnl->hist_jnl);
+dealloc_private:
+ GF_FREE(jnl);
+error_return:
+ return NULL;
}
diff --git a/xlators/features/changelog/lib/src/gf-changelog-journal.h b/xlators/features/changelog/lib/src/gf-changelog-journal.h
index 46d50f159d9..ba5b9bf827e 100644
--- a/xlators/features/changelog/lib/src/gf-changelog-journal.h
+++ b/xlators/features/changelog/lib/src/gf-changelog-journal.h
@@ -17,91 +17,91 @@
#include "changelog.h"
enum api_conn {
- JNL_API_CONNECTED,
- JNL_API_CONN_INPROGESS,
- JNL_API_DISCONNECTED,
+ JNL_API_CONNECTED,
+ JNL_API_CONN_INPROGESS,
+ JNL_API_DISCONNECTED,
};
typedef struct gf_changelog_entry {
- char path[PATH_MAX];
+ char path[PATH_MAX];
- struct list_head list;
+ struct list_head list;
} gf_changelog_entry_t;
typedef struct gf_changelog_processor {
- pthread_mutex_t lock; /* protects ->entries */
- pthread_cond_t cond; /* waiter during empty list */
- gf_boolean_t waiting;
+ pthread_mutex_t lock; /* protects ->entries */
+ pthread_cond_t cond; /* waiter during empty list */
+ gf_boolean_t waiting;
- pthread_t processor; /* thread-id of journal processing thread */
+ pthread_t processor; /* thread-id of journal processing thread */
- struct list_head entries;
+ struct list_head entries;
} gf_changelog_processor_t;
typedef struct gf_changelog_journal {
- DIR *jnl_dir; /* 'processing' directory stream */
+ DIR *jnl_dir; /* 'processing' directory stream */
- int jnl_fd; /* fd to the tracker file */
+ int jnl_fd; /* fd to the tracker file */
- char jnl_brickpath[PATH_MAX]; /* brick path for this end-point */
+ char jnl_brickpath[PATH_MAX]; /* brick path for this end-point */
- gf_changelog_processor_t *jnl_proc;
+ gf_changelog_processor_t *jnl_proc;
- char *jnl_working_dir; /* scratch directory */
+ char *jnl_working_dir; /* scratch directory */
- char jnl_current_dir[PATH_MAX];
- char jnl_processed_dir[PATH_MAX];
- char jnl_processing_dir[PATH_MAX];
+ char jnl_current_dir[PATH_MAX];
+ char jnl_processed_dir[PATH_MAX];
+ char jnl_processing_dir[PATH_MAX];
- char rfc3986_space_newline[256]; /* RFC 3986 string encoding */
+ char rfc3986_space_newline[256]; /* RFC 3986 string encoding */
- struct gf_changelog_journal *hist_jnl;
- int hist_done; /* holds 0 done scanning,
- 1 keep scanning and -1 error */
+ struct gf_changelog_journal *hist_jnl;
+ int hist_done; /* holds 0 done scanning,
+ 1 keep scanning and -1 error */
- pthread_spinlock_t lock;
- int connected;
- xlator_t *this;
+ pthread_spinlock_t lock;
+ int connected;
+ xlator_t *this;
} gf_changelog_journal_t;
-#define JNL_SET_API_STATE(jnl, state) (jnl->connected = state)
-#define JNL_IS_API_DISCONNECTED(jnl) (jnl->connected == JNL_API_DISCONNECTED)
+#define JNL_SET_API_STATE(jnl, state) (jnl->connected = state)
+#define JNL_IS_API_DISCONNECTED(jnl) (jnl->connected == JNL_API_DISCONNECTED)
/* History API */
typedef struct gf_changelog_history_data {
- int len;
+ int len;
- int htime_fd;
+ int htime_fd;
- /* parallelism count */
- int n_parallel;
+ /* parallelism count */
+ int n_parallel;
- /* history from, to indexes */
- unsigned long from;
- unsigned long to;
- xlator_t *this;
+ /* history from, to indexes */
+ unsigned long from;
+ unsigned long to;
+ xlator_t *this;
} gf_changelog_history_data_t;
typedef struct gf_changelog_consume_data {
- /** set of inputs */
+ /** set of inputs */
- /* fd to read from */
- int fd;
+ /* fd to read from */
+ int fd;
- /* from @offset */
- off_t offset;
+ /* from @offset */
+ off_t offset;
- xlator_t *this;
+ xlator_t *this;
- gf_changelog_journal_t *jnl;
+ gf_changelog_journal_t *jnl;
- /** set of outputs */
+ /** set of outputs */
- /* return value */
- int retval;
+ /* return value */
+ int retval;
- /* journal processed */
- char changelog[PATH_MAX];
+ /* journal processed */
+ char changelog[PATH_MAX];
} gf_changelog_consume_data_t;
/* event handler */
diff --git a/xlators/features/changelog/lib/src/gf-changelog-reborp.c b/xlators/features/changelog/lib/src/gf-changelog-reborp.c
index d1b3059a60c..56b11cbb705 100644
--- a/xlators/features/changelog/lib/src/gf-changelog-reborp.c
+++ b/xlators/features/changelog/lib/src/gf-changelog-reborp.c
@@ -15,147 +15,130 @@
#include "changelog-rpc-common.h"
#include "changelog-lib-messages.h"
-#include "syscall.h"
+#include <glusterfs/syscall.h>
/**
* Reverse socket: actual data transfer handler. Connection
* initiator is PROBER, data transfer is REBORP.
*/
-struct rpcsvc_program *gf_changelog_reborp_programs[];
+static struct rpcsvc_program *gf_changelog_reborp_programs[];
void *
-gf_changelog_connection_janitor (void *arg)
+gf_changelog_connection_janitor(void *arg)
{
- int32_t ret = 0;
- xlator_t *this = NULL;
- gf_private_t *priv = NULL;
- gf_changelog_t *entry = NULL;
- struct gf_event *event = NULL;
- struct gf_event_list *ev = NULL;
- unsigned long drained = 0;
-
- this = arg;
- THIS = this;
-
- priv = this->private;
-
- while (1) {
- pthread_mutex_lock (&priv->lock);
- {
- while (list_empty (&priv->cleanups))
- pthread_cond_wait (&priv->cond, &priv->lock);
-
- entry = list_first_entry (&priv->cleanups,
- gf_changelog_t, list);
- list_del_init (&entry->list);
- }
- pthread_mutex_unlock (&priv->lock);
-
- drained = 0;
- ev = &entry->event;
-
- gf_smsg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_LIB_MSG_CLEANING_BRICK_ENTRY_INFO,
- "Cleaning brick entry for brick",
- "brick=%s", entry->brick,
- NULL);
-
- /* 0x0: disbale rpc-clnt */
- rpc_clnt_disable (RPC_PROBER (entry));
-
- /* 0x1: cleanup callback invoker thread */
- ret = gf_cleanup_event (this, ev);
- if (ret)
- continue;
-
- /* 0x2: drain pending events */
- while (!list_empty (&ev->events)) {
- event = list_first_entry (&ev->events,
- struct gf_event, list);
- gf_smsg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_LIB_MSG_DRAINING_EVENT_INFO,
- "Draining event",
- "seq=%lu", event->seq,
- "payload=%d", event->count,
- NULL);
-
- GF_FREE (event);
- drained++;
- }
-
- gf_smsg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_LIB_MSG_DRAINING_EVENT_INFO,
- "Drained events",
- "num=%lu", drained,
- NULL);
-
- /* 0x3: freeup brick entry */
- gf_smsg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_LIB_MSG_FREEING_ENTRY_INFO,
- "freeing entry",
- "entry=%p", entry,
- NULL);
- LOCK_DESTROY (&entry->statelock);
- GF_FREE (entry);
+ int32_t ret = 0;
+ xlator_t *this = NULL;
+ gf_private_t *priv = NULL;
+ gf_changelog_t *entry = NULL;
+ struct gf_event *event = NULL;
+ struct gf_event_list *ev = NULL;
+ unsigned long drained = 0;
+
+ this = arg;
+ THIS = this;
+
+ priv = this->private;
+
+ while (1) {
+ pthread_mutex_lock(&priv->lock);
+ {
+ while (list_empty(&priv->cleanups))
+ pthread_cond_wait(&priv->cond, &priv->lock);
+
+ entry = list_first_entry(&priv->cleanups, gf_changelog_t, list);
+ list_del_init(&entry->list);
+ }
+ pthread_mutex_unlock(&priv->lock);
+
+ drained = 0;
+ ev = &entry->event;
+
+ gf_smsg(this->name, GF_LOG_INFO, 0,
+ CHANGELOG_LIB_MSG_CLEANING_BRICK_ENTRY_INFO, "brick=%s",
+ entry->brick, NULL);
+
+ /* 0x0: disable rpc-clnt */
+ rpc_clnt_disable(RPC_PROBER(entry));
+
+ /* 0x1: cleanup callback invoker thread */
+ ret = gf_cleanup_event(this, ev);
+ if (ret)
+ continue;
+
+ /* 0x2: drain pending events */
+ while (!list_empty(&ev->events)) {
+ event = list_first_entry(&ev->events, struct gf_event, list);
+ gf_smsg(this->name, GF_LOG_INFO, 0,
+ CHANGELOG_LIB_MSG_DRAINING_EVENT_INFO, "seq=%lu",
+ event->seq, "payload=%d", event->count, NULL);
+
+ GF_FREE(event);
+ drained++;
}
- return NULL;
+ gf_smsg(this->name, GF_LOG_INFO, 0,
+ CHANGELOG_LIB_MSG_DRAINED_EVENT_INFO, "num=%lu", drained, NULL);
+
+ /* 0x3: freeup brick entry */
+ gf_smsg(this->name, GF_LOG_INFO, 0,
+ CHANGELOG_LIB_MSG_FREEING_ENTRY_INFO, "entry=%p", entry, NULL);
+ LOCK_DESTROY(&entry->statelock);
+ GF_FREE(entry);
+ }
+
+ return NULL;
}
int
-gf_changelog_reborp_rpcsvc_notify (rpcsvc_t *rpc, void *mydata,
- rpcsvc_event_t event, void *data)
+gf_changelog_reborp_rpcsvc_notify(rpcsvc_t *rpc, void *mydata,
+ rpcsvc_event_t event, void *data)
{
- int ret = 0;
- xlator_t *this = NULL;
- gf_changelog_t *entry = NULL;
+ int ret = 0;
+ xlator_t *this = NULL;
+ gf_changelog_t *entry = NULL;
- if (!(event == RPCSVC_EVENT_ACCEPT ||
- event == RPCSVC_EVENT_DISCONNECT))
- return 0;
+ if (!(event == RPCSVC_EVENT_ACCEPT || event == RPCSVC_EVENT_DISCONNECT))
+ return 0;
- entry = mydata;
- this = entry->this;
+ entry = mydata;
+ this = entry->this;
- switch (event) {
+ switch (event) {
case RPCSVC_EVENT_ACCEPT:
- ret = sys_unlink (RPC_SOCK(entry));
- if (ret != 0)
- gf_smsg (this->name, GF_LOG_WARNING, errno,
- CHANGELOG_LIB_MSG_UNLINK_FAILED,
- "failed to unlink "
- "reverse socket",
- "path=%s", RPC_SOCK (entry),
- NULL);
- if (entry->connected)
- GF_CHANGELOG_INVOKE_CBK (this, entry->connected,
- entry->brick, entry->ptr);
- break;
+ ret = sys_unlink(RPC_SOCK(entry));
+ if (ret != 0)
+ gf_smsg(this->name, GF_LOG_WARNING, errno,
+ CHANGELOG_LIB_MSG_UNLINK_FAILED, "name=reverse socket",
+ "path=%s", RPC_SOCK(entry), NULL);
+ if (entry->connected)
+ GF_CHANGELOG_INVOKE_CBK(this, entry->connected, entry->brick,
+ entry->ptr);
+ break;
case RPCSVC_EVENT_DISCONNECT:
- if (entry->disconnected)
- GF_CHANGELOG_INVOKE_CBK (this, entry->disconnected,
- entry->brick, entry->ptr);
- /* passthrough */
+ if (entry->disconnected)
+ GF_CHANGELOG_INVOKE_CBK(this, entry->disconnected, entry->brick,
+ entry->ptr);
+ /* passthrough */
default:
- break;
- }
+ break;
+ }
- return 0;
+ return 0;
}
rpcsvc_t *
-gf_changelog_reborp_init_rpc_listner (xlator_t *this,
- char *path, char *sock, void *cbkdata)
+gf_changelog_reborp_init_rpc_listner(xlator_t *this, char *path, char *sock,
+ void *cbkdata)
{
- CHANGELOG_MAKE_TMP_SOCKET_PATH (path, sock, UNIX_PATH_MAX);
- return changelog_rpc_server_init (this, sock, cbkdata,
- gf_changelog_reborp_rpcsvc_notify,
- gf_changelog_reborp_programs);
+ CHANGELOG_MAKE_TMP_SOCKET_PATH(path, sock, UNIX_PATH_MAX);
+ return changelog_rpc_server_init(this, sock, cbkdata,
+ gf_changelog_reborp_rpcsvc_notify,
+ gf_changelog_reborp_programs);
}
/**
- * This is dirty and painful as of now untill there is event filtering in the
+ * This is dirty and painful as of now until there is event filtering in the
* server. The entire event buffer is scanned and interested events are picked,
* whereas we _should_ be notified with the events we were interested in
* (selected at the time of probe). As of now this is complete BS and needs
@@ -164,29 +147,27 @@ gf_changelog_reborp_init_rpc_listner (xlator_t *this,
* @FIXME: cleanup this bugger once server filters events.
*/
void
-gf_changelog_invoke_callback (gf_changelog_t *entry,
- struct iovec **vec, int payloadcnt)
+gf_changelog_invoke_callback(gf_changelog_t *entry, struct iovec **vec,
+ int payloadcnt)
{
- int i = 0;
- int evsize = 0;
- xlator_t *this = NULL;
- changelog_event_t *event = NULL;
-
- this = entry->this;
-
- for (; i < payloadcnt; i++) {
- event = (changelog_event_t *)vec[i]->iov_base;
- evsize = vec[i]->iov_len / CHANGELOG_EV_SIZE;
-
- for (; evsize > 0; evsize--, event++) {
- if (gf_changelog_filter_check (entry, event)) {
- GF_CHANGELOG_INVOKE_CBK (this,
- entry->callback,
- entry->brick,
- entry->ptr, event);
- }
- }
+ int i = 0;
+ int evsize = 0;
+ xlator_t *this = NULL;
+ changelog_event_t *event = NULL;
+
+ this = entry->this;
+
+ for (; i < payloadcnt; i++) {
+ event = (changelog_event_t *)vec[i]->iov_base;
+ evsize = vec[i]->iov_len / CHANGELOG_EV_SIZE;
+
+ for (; evsize > 0; evsize--, event++) {
+ if (gf_changelog_filter_check(entry, event)) {
+ GF_CHANGELOG_INVOKE_CBK(this, entry->callback, entry->brick,
+ entry->ptr, event);
+ }
}
+ }
}
/**
@@ -197,218 +178,218 @@ gf_changelog_invoke_callback (gf_changelog_t *entry,
*/
int
-__is_expected_sequence (struct gf_event_list *ev, struct gf_event *event)
+__is_expected_sequence(struct gf_event_list *ev, struct gf_event *event)
{
- return (ev->next_seq == event->seq);
+ return (ev->next_seq == event->seq);
}
int
-__can_process_event (struct gf_event_list *ev, struct gf_event **event)
+__can_process_event(struct gf_event_list *ev, struct gf_event **event)
{
- *event = list_first_entry (&ev->events, struct gf_event, list);
+ *event = list_first_entry(&ev->events, struct gf_event, list);
- if (__is_expected_sequence (ev, *event)) {
- list_del (&(*event)->list);
- ev->next_seq++;
- return 1;
- }
+ if (__is_expected_sequence(ev, *event)) {
+ list_del(&(*event)->list);
+ ev->next_seq++;
+ return 1;
+ }
- return 0;
+ return 0;
}
void
-pick_event_ordered (struct gf_event_list *ev, struct gf_event **event)
+pick_event_ordered(struct gf_event_list *ev, struct gf_event **event)
{
- pthread_mutex_lock (&ev->lock);
- {
- while (list_empty (&ev->events)
- || !__can_process_event (ev, event))
- pthread_cond_wait (&ev->cond, &ev->lock);
- }
- pthread_mutex_unlock (&ev->lock);
+ pthread_mutex_lock(&ev->lock);
+ {
+ while (list_empty(&ev->events) || !__can_process_event(ev, event))
+ pthread_cond_wait(&ev->cond, &ev->lock);
+ }
+ pthread_mutex_unlock(&ev->lock);
}
void
-pick_event_unordered (struct gf_event_list *ev, struct gf_event **event)
+pick_event_unordered(struct gf_event_list *ev, struct gf_event **event)
{
- pthread_mutex_lock (&ev->lock);
- {
- while (list_empty (&ev->events))
- pthread_cond_wait (&ev->cond, &ev->lock);
- *event = list_first_entry (&ev->events, struct gf_event, list);
- list_del (&(*event)->list);
- }
- pthread_mutex_unlock (&ev->lock);
+ pthread_mutex_lock(&ev->lock);
+ {
+ while (list_empty(&ev->events))
+ pthread_cond_wait(&ev->cond, &ev->lock);
+ *event = list_first_entry(&ev->events, struct gf_event, list);
+ list_del(&(*event)->list);
+ }
+ pthread_mutex_unlock(&ev->lock);
}
void *
-gf_changelog_callback_invoker (void *arg)
+gf_changelog_callback_invoker(void *arg)
{
- xlator_t *this = NULL;
- gf_changelog_t *entry = NULL;
- struct iovec *vec = NULL;
- struct gf_event *event = NULL;
- struct gf_event_list *ev = NULL;
+ xlator_t *this = NULL;
+ gf_changelog_t *entry = NULL;
+ struct iovec *vec = NULL;
+ struct gf_event *event = NULL;
+ struct gf_event_list *ev = NULL;
- ev = arg;
- entry = ev->entry;
- THIS = this = entry->this;
+ ev = arg;
+ entry = ev->entry;
+ THIS = this = entry->this;
- while (1) {
- entry->pickevent (ev, &event);
+ while (1) {
+ entry->pickevent(ev, &event);
- vec = (struct iovec *) &event->iov;
- gf_changelog_invoke_callback (entry, &vec, event->count);
+ vec = (struct iovec *)&event->iov;
+ gf_changelog_invoke_callback(entry, &vec, event->count);
- GF_FREE (event);
- }
+ GF_FREE(event);
+ }
- return NULL;
+ return NULL;
}
static int
-orderfn (struct list_head *pos1, struct list_head *pos2)
+orderfn(struct list_head *pos1, struct list_head *pos2)
{
- struct gf_event *event1 = NULL;
- struct gf_event *event2 = NULL;
+ struct gf_event *event1 = NULL;
+ struct gf_event *event2 = NULL;
- event1 = list_entry (pos1, struct gf_event, list);
- event2 = list_entry (pos2, struct gf_event, list);
+ event1 = list_entry(pos1, struct gf_event, list);
+ event2 = list_entry(pos2, struct gf_event, list);
- if (event1->seq > event2->seq)
- return 1;
- return -1;
+ if (event1->seq > event2->seq)
+ return 1;
+ return -1;
}
void
-queue_ordered_event (struct gf_event_list *ev, struct gf_event *event)
+queue_ordered_event(struct gf_event_list *ev, struct gf_event *event)
{
- /* add event to the ordered event list and wake up listner(s) */
- pthread_mutex_lock (&ev->lock);
- {
- list_add_order (&event->list, &ev->events, orderfn);
- if (!ev->next_seq)
- ev->next_seq = event->seq;
- if (ev->next_seq == event->seq)
- pthread_cond_signal (&ev->cond);
- }
- pthread_mutex_unlock (&ev->lock);
+ /* add event to the ordered event list and wake up listener(s) */
+ pthread_mutex_lock(&ev->lock);
+ {
+ list_add_order(&event->list, &ev->events, orderfn);
+ if (!ev->next_seq)
+ ev->next_seq = event->seq;
+ if (ev->next_seq == event->seq)
+ pthread_cond_signal(&ev->cond);
+ }
+ pthread_mutex_unlock(&ev->lock);
}
void
-queue_unordered_event (struct gf_event_list *ev, struct gf_event *event)
+queue_unordered_event(struct gf_event_list *ev, struct gf_event *event)
{
- /* add event to the tail of the queue and wake up listener(s) */
- pthread_mutex_lock (&ev->lock);
- {
- list_add_tail (&event->list, &ev->events);
- pthread_cond_signal (&ev->cond);
- }
- pthread_mutex_unlock (&ev->lock);
+ /* add event to the tail of the queue and wake up listener(s) */
+ pthread_mutex_lock(&ev->lock);
+ {
+ list_add_tail(&event->list, &ev->events);
+ pthread_cond_signal(&ev->cond);
+ }
+ pthread_mutex_unlock(&ev->lock);
}
int
-gf_changelog_event_handler (rpcsvc_request_t *req,
- xlator_t *this, gf_changelog_t *entry)
+gf_changelog_event_handler(rpcsvc_request_t *req, xlator_t *this,
+ gf_changelog_t *entry)
{
- int i = 0;
- size_t payloadlen = 0;
- ssize_t len = 0;
- int payloadcnt = 0;
- changelog_event_req rpc_req = {0,};
- changelog_event_rsp rpc_rsp = {0,};
- struct iovec *vec = NULL;
- struct gf_event *event = NULL;
- struct gf_event_list *ev = NULL;
-
- ev = &entry->event;
-
- len = xdr_to_generic (req->msg[0],
- &rpc_req, (xdrproc_t)xdr_changelog_event_req);
- if (len < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_LIB_MSG_XDR_DECODING_FAILED,
- "xdr decoding failed");
- req->rpc_err = GARBAGE_ARGS;
- goto handle_xdr_error;
- }
-
- if (len < req->msg[0].iov_len) {
- payloadcnt = 1;
- payloadlen = (req->msg[0].iov_len - len);
- }
- for (i = 1; i < req->count; i++) {
- payloadcnt++;
- payloadlen += req->msg[i].iov_len;
- }
-
- event = GF_CALLOC (1, GF_EVENT_CALLOC_SIZE (payloadcnt, payloadlen),
- gf_changelog_mt_libgfchangelog_event_t);
- if (!event)
- goto handle_xdr_error;
- INIT_LIST_HEAD (&event->list);
-
- payloadlen = 0;
- event->seq = rpc_req.seq;
- event->count = payloadcnt;
-
- /* deep copy IO vectors */
- vec = &event->iov[0];
- GF_EVENT_ASSIGN_IOVEC (vec, event,
- (req->msg[0].iov_len - len), payloadlen);
- (void) memcpy (vec->iov_base,
- req->msg[0].iov_base + len, vec->iov_len);
-
- for (i = 1; i < req->count; i++) {
- vec = &event->iov[i];
- GF_EVENT_ASSIGN_IOVEC (vec, event,
- req->msg[i].iov_len, payloadlen);
- (void) memcpy (event->iov[i].iov_base,
- req->msg[i].iov_base, req->msg[i].iov_len);
- }
-
- gf_msg_debug (this->name, 0,
- "seq: %lu [%s] (time: %lu.%lu), (vec: %d, len: %zd)",
- rpc_req.seq, entry->brick, rpc_req.tv_sec,
- rpc_req.tv_usec, payloadcnt, payloadlen);
-
- /* dispatch event */
- entry->queueevent (ev, event);
-
- /* ack sequence number */
- rpc_rsp.op_ret = 0;
- rpc_rsp.seq = rpc_req.seq;
-
- goto submit_rpc;
-
- handle_xdr_error:
- rpc_rsp.op_ret = -1;
- rpc_rsp.seq = 0; /* invalid */
- submit_rpc:
- return changelog_rpc_sumbit_reply (req, &rpc_rsp, NULL, 0, NULL,
- (xdrproc_t)xdr_changelog_event_rsp);
+ int i = 0;
+ size_t payloadlen = 0;
+ ssize_t len = 0;
+ int payloadcnt = 0;
+ changelog_event_req rpc_req = {
+ 0,
+ };
+ changelog_event_rsp rpc_rsp = {
+ 0,
+ };
+ struct iovec *vec = NULL;
+ struct gf_event *event = NULL;
+ struct gf_event_list *ev = NULL;
+
+ ev = &entry->event;
+
+ len = xdr_to_generic(req->msg[0], &rpc_req,
+ (xdrproc_t)xdr_changelog_event_req);
+ if (len < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_LIB_MSG_XDR_DECODING_FAILED, "xdr decoding failed");
+ req->rpc_err = GARBAGE_ARGS;
+ goto handle_xdr_error;
+ }
+
+ if (len < req->msg[0].iov_len) {
+ payloadcnt = 1;
+ payloadlen = (req->msg[0].iov_len - len);
+ }
+ for (i = 1; i < req->count; i++) {
+ payloadcnt++;
+ payloadlen += req->msg[i].iov_len;
+ }
+
+ event = GF_CALLOC(1, GF_EVENT_CALLOC_SIZE(payloadcnt, payloadlen),
+ gf_changelog_mt_libgfchangelog_event_t);
+ if (!event)
+ goto handle_xdr_error;
+ INIT_LIST_HEAD(&event->list);
+
+ payloadlen = 0;
+ event->seq = rpc_req.seq;
+ event->count = payloadcnt;
+
+ /* deep copy IO vectors */
+ vec = &event->iov[0];
+ GF_EVENT_ASSIGN_IOVEC(vec, event, (req->msg[0].iov_len - len), payloadlen);
+ (void)memcpy(vec->iov_base, req->msg[0].iov_base + len, vec->iov_len);
+
+ for (i = 1; i < req->count; i++) {
+ vec = &event->iov[i];
+ GF_EVENT_ASSIGN_IOVEC(vec, event, req->msg[i].iov_len, payloadlen);
+ (void)memcpy(event->iov[i].iov_base, req->msg[i].iov_base,
+ req->msg[i].iov_len);
+ }
+
+ gf_msg_debug(this->name, 0,
+ "seq: %" PRIu64 " [%s] (time: %" PRIu64 ".%" PRIu64
+ "), "
+ "(vec: %d, len: %zd)",
+ rpc_req.seq, entry->brick, rpc_req.tv_sec, rpc_req.tv_usec,
+ payloadcnt, payloadlen);
+
+ /* dispatch event */
+ entry->queueevent(ev, event);
+
+ /* ack sequence number */
+ rpc_rsp.op_ret = 0;
+ rpc_rsp.seq = rpc_req.seq;
+
+ goto submit_rpc;
+
+handle_xdr_error:
+ rpc_rsp.op_ret = -1;
+ rpc_rsp.seq = 0; /* invalid */
+submit_rpc:
+ return changelog_rpc_sumbit_reply(req, &rpc_rsp, NULL, 0, NULL,
+ (xdrproc_t)xdr_changelog_event_rsp);
}
int
-gf_changelog_reborp_handle_event (rpcsvc_request_t *req)
+gf_changelog_reborp_handle_event(rpcsvc_request_t *req)
{
- xlator_t *this = NULL;
- rpcsvc_t *svc = NULL;
- gf_changelog_t *entry = NULL;
+ xlator_t *this = NULL;
+ rpcsvc_t *svc = NULL;
+ gf_changelog_t *entry = NULL;
- svc = rpcsvc_request_service (req);
- entry = svc->mydata;
+ svc = rpcsvc_request_service(req);
+ entry = svc->mydata;
- this = THIS = entry->this;
+ this = THIS = entry->this;
- return gf_changelog_event_handler (req, this, entry);
+ return gf_changelog_event_handler(req, this, entry);
}
-rpcsvc_actor_t gf_changelog_reborp_actors[CHANGELOG_REV_PROC_MAX] = {
- [CHANGELOG_REV_PROC_EVENT] = {
- "CHANGELOG EVENT HANDLER", CHANGELOG_REV_PROC_EVENT,
- gf_changelog_reborp_handle_event, NULL, 0, DRC_NA
- },
+static rpcsvc_actor_t gf_changelog_reborp_actors[CHANGELOG_REV_PROC_MAX] = {
+ [CHANGELOG_REV_PROC_EVENT] = {"CHANGELOG EVENT HANDLER",
+ gf_changelog_reborp_handle_event, NULL,
+ CHANGELOG_REV_PROC_EVENT, DRC_NA, 0},
};
/**
@@ -417,16 +398,16 @@ rpcsvc_actor_t gf_changelog_reborp_actors[CHANGELOG_REV_PROC_MAX] = {
* and that's required to invoke the callback with the appropriate
* brick path and it's private data.
*/
-struct rpcsvc_program gf_changelog_reborp_prog = {
- .progname = "LIBGFCHANGELOG REBORP",
- .prognum = CHANGELOG_REV_RPC_PROCNUM,
- .progver = CHANGELOG_REV_RPC_PROCVER,
- .numactors = CHANGELOG_REV_PROC_MAX,
- .actors = gf_changelog_reborp_actors,
- .synctask = _gf_false,
+static struct rpcsvc_program gf_changelog_reborp_prog = {
+ .progname = "LIBGFCHANGELOG REBORP",
+ .prognum = CHANGELOG_REV_RPC_PROCNUM,
+ .progver = CHANGELOG_REV_RPC_PROCVER,
+ .numactors = CHANGELOG_REV_PROC_MAX,
+ .actors = gf_changelog_reborp_actors,
+ .synctask = _gf_false,
};
-struct rpcsvc_program *gf_changelog_reborp_programs[] = {
- &gf_changelog_reborp_prog,
- NULL,
+static struct rpcsvc_program *gf_changelog_reborp_programs[] = {
+ &gf_changelog_reborp_prog,
+ NULL,
};
diff --git a/xlators/features/changelog/lib/src/gf-changelog-rpc.c b/xlators/features/changelog/lib/src/gf-changelog-rpc.c
index c1139423d6d..8ec6ffbcebc 100644
--- a/xlators/features/changelog/lib/src/gf-changelog-rpc.c
+++ b/xlators/features/changelog/lib/src/gf-changelog-rpc.c
@@ -16,32 +16,32 @@ struct rpc_clnt_program gf_changelog_clnt;
/* TODO: piggyback reconnect to called (upcall) */
int
-gf_changelog_rpc_notify (struct rpc_clnt *rpc,
- void *mydata, rpc_clnt_event_t event, void *data)
+gf_changelog_rpc_notify(struct rpc_clnt *rpc, void *mydata,
+ rpc_clnt_event_t event, void *data)
{
- switch (event) {
+ switch (event) {
case RPC_CLNT_CONNECT:
- rpc_clnt_set_connected (&rpc->conn);
- break;
+ break;
case RPC_CLNT_DISCONNECT:
case RPC_CLNT_MSG:
case RPC_CLNT_DESTROY:
case RPC_CLNT_PING:
- break;
- }
+ break;
+ }
- return 0;
+ return 0;
}
struct rpc_clnt *
-gf_changelog_rpc_init (xlator_t *this, gf_changelog_t *entry)
+gf_changelog_rpc_init(xlator_t *this, gf_changelog_t *entry)
{
- char sockfile[UNIX_PATH_MAX] = {0,};
+ char sockfile[UNIX_PATH_MAX] = {
+ 0,
+ };
- CHANGELOG_MAKE_SOCKET_PATH (entry->brick,
- sockfile, UNIX_PATH_MAX);
- return changelog_rpc_client_init (this, entry,
- sockfile, gf_changelog_rpc_notify);
+ CHANGELOG_MAKE_SOCKET_PATH(entry->brick, sockfile, UNIX_PATH_MAX);
+ return changelog_rpc_client_init(this, entry, sockfile,
+ gf_changelog_rpc_notify);
}
/**
@@ -49,51 +49,50 @@ gf_changelog_rpc_init (xlator_t *this, gf_changelog_t *entry)
*/
int
-gf_probe_changelog_cbk (struct rpc_req *req,
- struct iovec *iovec, int count, void *myframe)
+gf_probe_changelog_cbk(struct rpc_req *req, struct iovec *iovec, int count,
+ void *myframe)
{
- return 0;
+ return 0;
}
int
-gf_probe_changelog_filter (call_frame_t *frame, xlator_t *this, void *data)
+gf_probe_changelog_filter(call_frame_t *frame, xlator_t *this, void *data)
{
- char *sock = NULL;
- gf_changelog_t *entry = NULL;
- changelog_probe_req req = {0,};
-
- entry = data;
- sock = RPC_SOCK (entry);
-
- (void) memcpy (&req.sock, sock, strlen (sock));
- req.filter = entry->notify;
-
- /* invoke RPC */
- return changelog_rpc_sumbit_req (RPC_PROBER (entry), (void *) &req,
- frame, &gf_changelog_clnt,
- CHANGELOG_RPC_PROBE_FILTER, NULL, 0,
- NULL, this, gf_probe_changelog_cbk,
- (xdrproc_t) xdr_changelog_probe_req);
+ char *sock = NULL;
+ gf_changelog_t *entry = NULL;
+ changelog_probe_req req = {
+ 0,
+ };
+
+ entry = data;
+ sock = RPC_SOCK(entry);
+
+ (void)memcpy(&req.sock, sock, strlen(sock));
+ req.filter = entry->notify;
+
+ /* invoke RPC */
+ return changelog_rpc_sumbit_req(
+ RPC_PROBER(entry), (void *)&req, frame, &gf_changelog_clnt,
+ CHANGELOG_RPC_PROBE_FILTER, NULL, 0, NULL, this, gf_probe_changelog_cbk,
+ (xdrproc_t)xdr_changelog_probe_req);
}
int
-gf_changelog_invoke_rpc (xlator_t *this, gf_changelog_t *entry, int procidx)
+gf_changelog_invoke_rpc(xlator_t *this, gf_changelog_t *entry, int procidx)
{
- return changelog_invoke_rpc (this, RPC_PROBER (entry),
- &gf_changelog_clnt, procidx, entry);
+ return changelog_invoke_rpc(this, RPC_PROBER(entry), &gf_changelog_clnt,
+ procidx, entry);
}
struct rpc_clnt_procedure gf_changelog_procs[CHANGELOG_RPC_PROC_MAX] = {
- [CHANGELOG_RPC_PROC_NULL] = {"NULL", NULL},
- [CHANGELOG_RPC_PROBE_FILTER] = {
- "PROBE FILTER", gf_probe_changelog_filter
- },
+ [CHANGELOG_RPC_PROC_NULL] = {"NULL", NULL},
+ [CHANGELOG_RPC_PROBE_FILTER] = {"PROBE FILTER", gf_probe_changelog_filter},
};
struct rpc_clnt_program gf_changelog_clnt = {
- .progname = "LIBGFCHANGELOG",
- .prognum = CHANGELOG_RPC_PROGNUM,
- .progver = CHANGELOG_RPC_PROGVER,
- .numproc = CHANGELOG_RPC_PROC_MAX,
- .proctable = gf_changelog_procs,
+ .progname = "LIBGFCHANGELOG",
+ .prognum = CHANGELOG_RPC_PROGNUM,
+ .progver = CHANGELOG_RPC_PROGVER,
+ .numproc = CHANGELOG_RPC_PROC_MAX,
+ .proctable = gf_changelog_procs,
};
diff --git a/xlators/features/changelog/lib/src/gf-changelog-rpc.h b/xlators/features/changelog/lib/src/gf-changelog-rpc.h
index 1c982eef809..5c82d6f1c08 100644
--- a/xlators/features/changelog/lib/src/gf-changelog-rpc.h
+++ b/xlators/features/changelog/lib/src/gf-changelog-rpc.h
@@ -11,16 +11,18 @@
#ifndef __GF_CHANGELOG_RPC_H
#define __GF_CHANGELOG_RPC_H
-#include "xlator.h"
+#include <glusterfs/xlator.h>
#include "gf-changelog-helpers.h"
#include "changelog-rpc-common.h"
-struct rpc_clnt *gf_changelog_rpc_init (xlator_t *, gf_changelog_t *);
+struct rpc_clnt *
+gf_changelog_rpc_init(xlator_t *, gf_changelog_t *);
-int gf_changelog_invoke_rpc (xlator_t *, gf_changelog_t *, int);
+int
+gf_changelog_invoke_rpc(xlator_t *, gf_changelog_t *, int);
rpcsvc_t *
-gf_changelog_reborp_init_rpc_listner (xlator_t *, char *, char *, void *);
+gf_changelog_reborp_init_rpc_listner(xlator_t *, char *, char *, void *);
#endif
diff --git a/xlators/features/changelog/lib/src/gf-changelog.c b/xlators/features/changelog/lib/src/gf-changelog.c
index 6ad6c63740d..57c3d39ef76 100644
--- a/xlators/features/changelog/lib/src/gf-changelog.c
+++ b/xlators/features/changelog/lib/src/gf-changelog.c
@@ -22,11 +22,11 @@
#endif
#include <string.h>
-#include "globals.h"
-#include "glusterfs.h"
-#include "logging.h"
-#include "defaults.h"
-#include "syncop.h"
+#include <glusterfs/globals.h>
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/defaults.h>
+#include <glusterfs/syncop.h>
#include "gf-changelog-rpc.h"
#include "gf-changelog-helpers.h"
@@ -45,291 +45,315 @@
*/
xlator_t *master = NULL;
-static inline
-gf_private_t *gf_changelog_alloc_priv ()
+static inline gf_private_t *
+gf_changelog_alloc_priv()
{
- int ret = 0;
- gf_private_t *priv = NULL;
-
- priv = GF_CALLOC (1, sizeof (*priv), gf_changelog_mt_priv_t);
- if (!priv)
- goto error_return;
- INIT_LIST_HEAD (&priv->connections);
- INIT_LIST_HEAD (&priv->cleanups);
-
- ret = pthread_mutex_init (&priv->lock, NULL);
- if (ret != 0)
- goto free_priv;
- ret = pthread_cond_init (&priv->cond, NULL);
- if (ret != 0)
- goto cleanup_mutex;
-
- priv->api = NULL;
- return priv;
-
- cleanup_mutex:
- (void) pthread_mutex_destroy (&priv->lock);
- free_priv:
- GF_FREE (priv);
- error_return:
- return NULL;
+ int ret = 0;
+ gf_private_t *priv = NULL;
+
+ priv = GF_CALLOC(1, sizeof(*priv), gf_changelog_mt_priv_t);
+ if (!priv)
+ goto error_return;
+ INIT_LIST_HEAD(&priv->connections);
+ INIT_LIST_HEAD(&priv->cleanups);
+
+ ret = pthread_mutex_init(&priv->lock, NULL);
+ if (ret != 0)
+ goto free_priv;
+ ret = pthread_cond_init(&priv->cond, NULL);
+ if (ret != 0)
+ goto cleanup_mutex;
+
+ priv->api = NULL;
+ return priv;
+
+cleanup_mutex:
+ (void)pthread_mutex_destroy(&priv->lock);
+free_priv:
+ GF_FREE(priv);
+error_return:
+ return NULL;
}
-#define GF_CHANGELOG_EVENT_POOL_SIZE 16384
+#define GF_CHANGELOG_EVENT_POOL_SIZE 16384
#define GF_CHANGELOG_EVENT_THREAD_COUNT 4
static int
-gf_changelog_ctx_defaults_init (glusterfs_ctx_t *ctx)
+gf_changelog_ctx_defaults_init(glusterfs_ctx_t *ctx)
{
- cmd_args_t *cmd_args = NULL;
- struct rlimit lim = {0, };
- call_pool_t *pool = NULL;
- int ret = -1;
+ cmd_args_t *cmd_args = NULL;
+ struct rlimit lim = {
+ 0,
+ };
+ call_pool_t *pool = NULL;
+ int ret = -1;
+
+ ret = xlator_mem_acct_init(THIS, gf_changelog_mt_end);
+ if (ret != 0)
+ return -1;
- ret = xlator_mem_acct_init (THIS, gf_changelog_mt_end);
- if (ret != 0)
- return -1;
+ ctx->process_uuid = generate_glusterfs_ctx_id();
+ if (!ctx->process_uuid)
+ return -1;
- ctx->process_uuid = generate_glusterfs_ctx_id ();
- if (!ctx->process_uuid)
- return -1;
+ ctx->page_size = 128 * GF_UNIT_KB;
- ctx->page_size = 128 * GF_UNIT_KB;
+ ctx->iobuf_pool = iobuf_pool_new();
+ if (!ctx->iobuf_pool)
+ goto free_pool;
- ctx->iobuf_pool = iobuf_pool_new ();
- if (!ctx->iobuf_pool)
- return -1;
+ ctx->event_pool = gf_event_pool_new(GF_CHANGELOG_EVENT_POOL_SIZE,
+ GF_CHANGELOG_EVENT_THREAD_COUNT);
+ if (!ctx->event_pool)
+ goto free_pool;
- ctx->event_pool = event_pool_new (GF_CHANGELOG_EVENT_POOL_SIZE,
- GF_CHANGELOG_EVENT_THREAD_COUNT);
- if (!ctx->event_pool)
- return -1;
+ pool = GF_CALLOC(1, sizeof(call_pool_t),
+ gf_changelog_mt_libgfchangelog_call_pool_t);
+ if (!pool)
+ goto free_pool;
- pool = GF_CALLOC (1, sizeof (call_pool_t),
- gf_changelog_mt_libgfchangelog_call_pool_t);
- if (!pool)
- return -1;
+ /* frame_mem_pool size 112 * 64 */
+ pool->frame_mem_pool = mem_pool_new(call_frame_t, 32);
+ if (!pool->frame_mem_pool)
+ goto free_pool;
- /* frame_mem_pool size 112 * 64 */
- pool->frame_mem_pool = mem_pool_new (call_frame_t, 32);
- if (!pool->frame_mem_pool)
- return -1;
+ /* stack_mem_pool size 256 * 128 */
+ pool->stack_mem_pool = mem_pool_new(call_stack_t, 16);
- /* stack_mem_pool size 256 * 128 */
- pool->stack_mem_pool = mem_pool_new (call_stack_t, 16);
+ if (!pool->stack_mem_pool)
+ goto free_pool;
- if (!pool->stack_mem_pool)
- return -1;
+ ctx->stub_mem_pool = mem_pool_new(call_stub_t, 16);
+ if (!ctx->stub_mem_pool)
+ goto free_pool;
- ctx->stub_mem_pool = mem_pool_new (call_stub_t, 16);
- if (!ctx->stub_mem_pool)
- return -1;
+ ctx->dict_pool = mem_pool_new(dict_t, 32);
+ if (!ctx->dict_pool)
+ goto free_pool;
- ctx->dict_pool = mem_pool_new (dict_t, 32);
- if (!ctx->dict_pool)
- return -1;
+ ctx->dict_pair_pool = mem_pool_new(data_pair_t, 512);
+ if (!ctx->dict_pair_pool)
+ goto free_pool;
- ctx->dict_pair_pool = mem_pool_new (data_pair_t, 512);
- if (!ctx->dict_pair_pool)
- return -1;
+ ctx->dict_data_pool = mem_pool_new(data_t, 512);
+ if (!ctx->dict_data_pool)
+ goto free_pool;
- ctx->dict_data_pool = mem_pool_new (data_t, 512);
- if (!ctx->dict_data_pool)
- return -1;
+ ctx->logbuf_pool = mem_pool_new(log_buf_t, 256);
+ if (!ctx->logbuf_pool)
+ goto free_pool;
- ctx->logbuf_pool = mem_pool_new (log_buf_t, 256);
- if (!ctx->logbuf_pool)
- return -1;
+ INIT_LIST_HEAD(&pool->all_frames);
+ LOCK_INIT(&pool->lock);
+ ctx->pool = pool;
- INIT_LIST_HEAD (&pool->all_frames);
- LOCK_INIT (&pool->lock);
- ctx->pool = pool;
+ LOCK_INIT(&ctx->lock);
- LOCK_INIT (&ctx->lock);
+ cmd_args = &ctx->cmd_args;
- cmd_args = &ctx->cmd_args;
+ INIT_LIST_HEAD(&cmd_args->xlator_options);
- INIT_LIST_HEAD (&cmd_args->xlator_options);
+ lim.rlim_cur = RLIM_INFINITY;
+ lim.rlim_max = RLIM_INFINITY;
+ setrlimit(RLIMIT_CORE, &lim);
- lim.rlim_cur = RLIM_INFINITY;
- lim.rlim_max = RLIM_INFINITY;
- setrlimit (RLIMIT_CORE, &lim);
+ return 0;
- return 0;
+free_pool:
+ if (pool) {
+ GF_FREE(pool->frame_mem_pool);
+
+ GF_FREE(pool->stack_mem_pool);
+
+ GF_FREE(pool);
+ }
+
+ GF_FREE(ctx->stub_mem_pool);
+
+ GF_FREE(ctx->dict_pool);
+
+ GF_FREE(ctx->dict_pair_pool);
+
+ GF_FREE(ctx->dict_data_pool);
+
+ GF_FREE(ctx->logbuf_pool);
+
+ GF_FREE(ctx->iobuf_pool);
+
+ GF_FREE(ctx->event_pool);
+
+ return -1;
}
/* TODO: cleanup ctx defaults */
void
-gf_changelog_cleanup_this (xlator_t *this)
+gf_changelog_cleanup_this(xlator_t *this)
{
- glusterfs_ctx_t *ctx = NULL;
+ glusterfs_ctx_t *ctx = NULL;
- if (!this)
- return;
+ if (!this)
+ return;
- ctx = this->ctx;
- syncenv_destroy (ctx->env);
- free (ctx);
+ ctx = this->ctx;
+ syncenv_destroy(ctx->env);
+ free(ctx);
- this->private = NULL;
- this->ctx = NULL;
+ this->private = NULL;
+ this->ctx = NULL;
- mem_pools_fini ();
+ mem_pools_fini();
}
static int
-gf_changelog_init_context ()
+gf_changelog_init_context()
{
- glusterfs_ctx_t *ctx = NULL;
+ glusterfs_ctx_t *ctx = NULL;
- ctx = glusterfs_ctx_new ();
- if (!ctx)
- goto error_return;
+ ctx = glusterfs_ctx_new();
+ if (!ctx)
+ goto error_return;
- if (glusterfs_globals_init (ctx))
- goto free_ctx;
+ if (glusterfs_globals_init(ctx))
+ goto free_ctx;
- THIS->ctx = ctx;
- if (gf_changelog_ctx_defaults_init (ctx))
- goto free_ctx;
+ THIS->ctx = ctx;
+ if (gf_changelog_ctx_defaults_init(ctx))
+ goto free_ctx;
- ctx->env = syncenv_new (0, 0, 0);
- if (!ctx->env)
- goto free_ctx;
- return 0;
+ ctx->env = syncenv_new(0, 0, 0);
+ if (!ctx->env)
+ goto free_ctx;
+ return 0;
- free_ctx:
- free (ctx);
- THIS->ctx = NULL;
- error_return:
- return -1;
+free_ctx:
+ free(ctx);
+ THIS->ctx = NULL;
+error_return:
+ return -1;
}
static int
-gf_changelog_init_master ()
+gf_changelog_init_master()
{
- int ret = 0;
+ int ret = 0;
- mem_pools_init_early ();
- ret = gf_changelog_init_context ();
- mem_pools_init_late ();
+ ret = gf_changelog_init_context();
+ mem_pools_init();
- return ret;
+ return ret;
}
/* TODO: cleanup clnt/svc on failure */
int
-gf_changelog_setup_rpc (xlator_t *this,
- gf_changelog_t *entry, int proc)
+gf_changelog_setup_rpc(xlator_t *this, gf_changelog_t *entry, int proc)
{
- int ret = 0;
- rpcsvc_t *svc = NULL;
- struct rpc_clnt *rpc = NULL;
-
- /**
- * Initialize a connect back socket. A probe() RPC call to the server
- * triggers a reverse connect.
- */
- svc = gf_changelog_reborp_init_rpc_listner (this, entry->brick,
- RPC_SOCK (entry), entry);
- if (!svc)
- goto error_return;
- RPC_REBORP (entry) = svc;
-
- /* Initialize an RPC client */
- rpc = gf_changelog_rpc_init (this, entry);
- if (!rpc)
- goto error_return;
- RPC_PROBER (entry) = rpc;
-
- /**
- * @FIXME
- * till we have connection state machine, let's delay the RPC call
- * for now..
- */
- sleep (2);
-
- /**
- * Probe changelog translator for reverse connection. After a successful
- * call, there's less use of the client and can be disconnected, but
- * let's leave the connection active for any future RPC calls.
- */
- ret = gf_changelog_invoke_rpc (this, entry, proc);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_LIB_MSG_INVOKE_RPC_FAILED,
- "Could not initiate probe RPC, bailing out!!!");
- goto error_return;
- }
-
- return 0;
-
- error_return:
- return -1;
+ int ret = 0;
+ rpcsvc_t *svc = NULL;
+ struct rpc_clnt *rpc = NULL;
+
+ /**
+ * Initialize a connect back socket. A probe() RPC call to the server
+ * triggers a reverse connect.
+ */
+ svc = gf_changelog_reborp_init_rpc_listner(this, entry->brick,
+ RPC_SOCK(entry), entry);
+ if (!svc)
+ goto error_return;
+ RPC_REBORP(entry) = svc;
+
+ /* Initialize an RPC client */
+ rpc = gf_changelog_rpc_init(this, entry);
+ if (!rpc)
+ goto error_return;
+ RPC_PROBER(entry) = rpc;
+
+ /**
+ * @FIXME
+ * till we have connection state machine, let's delay the RPC call
+ * for now..
+ */
+ sleep(2);
+
+ /**
+ * Probe changelog translator for reverse connection. After a successful
+ * call, there's less use of the client and can be disconnected, but
+ * let's leave the connection active for any future RPC calls.
+ */
+ ret = gf_changelog_invoke_rpc(this, entry, proc);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, CHANGELOG_LIB_MSG_INVOKE_RPC_FAILED,
+ "Could not initiate probe RPC, bailing out!!!");
+ goto error_return;
+ }
+
+ return 0;
+
+error_return:
+ return -1;
}
int
-gf_cleanup_event (xlator_t *this, struct gf_event_list *ev)
+gf_cleanup_event(xlator_t *this, struct gf_event_list *ev)
{
- int ret = 0;
-
- ret = gf_thread_cleanup (this, ev->invoker);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, -ret,
- CHANGELOG_LIB_MSG_CLEANUP_ERROR,
- "cannot cleanup callback invoker thread."
- " Not freeing resources");
- return -1;
- }
+ int ret = 0;
+
+ ret = gf_thread_cleanup(this, ev->invoker);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, -ret,
+ CHANGELOG_LIB_MSG_CLEANUP_ERROR,
+ "cannot cleanup callback invoker thread."
+ " Not freeing resources");
+ return -1;
+ }
- ev->entry = NULL;
+ ev->entry = NULL;
- return 0;
+ return 0;
}
static int
-gf_init_event (gf_changelog_t *entry)
+gf_init_event(gf_changelog_t *entry)
{
- int ret = 0;
- struct gf_event_list *ev = NULL;
-
- ev = &entry->event;
- ev->entry = entry;
-
- ret = pthread_mutex_init (&ev->lock, NULL);
- if (ret != 0)
- goto error_return;
- ret = pthread_cond_init (&ev->cond, NULL);
- if (ret != 0)
- goto cleanup_mutex;
- INIT_LIST_HEAD (&ev->events);
-
- ev->next_seq = 0; /* bootstrap sequencing */
-
- if (GF_NEED_ORDERED_EVENTS (entry)) {
- entry->pickevent = pick_event_ordered;
- entry->queueevent = queue_ordered_event;
- } else {
- entry->pickevent = pick_event_unordered;
- entry->queueevent = queue_unordered_event;
- }
-
- ret = gf_thread_create (&ev->invoker, NULL,
- gf_changelog_callback_invoker, ev, "clogcbki");
- if (ret != 0) {
- entry->pickevent = NULL;
- entry->queueevent = NULL;
- goto cleanup_cond;
- }
-
- return 0;
-
- cleanup_cond:
- (void) pthread_cond_destroy (&ev->cond);
- cleanup_mutex:
- (void) pthread_mutex_destroy (&ev->lock);
- error_return:
- return -1;
+ int ret = 0;
+ struct gf_event_list *ev = NULL;
+
+ ev = &entry->event;
+ ev->entry = entry;
+
+ ret = pthread_mutex_init(&ev->lock, NULL);
+ if (ret != 0)
+ goto error_return;
+ ret = pthread_cond_init(&ev->cond, NULL);
+ if (ret != 0)
+ goto cleanup_mutex;
+ INIT_LIST_HEAD(&ev->events);
+
+ ev->next_seq = 0; /* bootstrap sequencing */
+
+ if (GF_NEED_ORDERED_EVENTS(entry)) {
+ entry->pickevent = pick_event_ordered;
+ entry->queueevent = queue_ordered_event;
+ } else {
+ entry->pickevent = pick_event_unordered;
+ entry->queueevent = queue_unordered_event;
+ }
+
+ ret = gf_thread_create(&ev->invoker, NULL, gf_changelog_callback_invoker,
+ ev, "clogcbki");
+ if (ret != 0) {
+ entry->pickevent = NULL;
+ entry->queueevent = NULL;
+ goto cleanup_cond;
+ }
+
+ return 0;
+
+cleanup_cond:
+ (void)pthread_cond_destroy(&ev->cond);
+cleanup_mutex:
+ (void)pthread_mutex_destroy(&ev->lock);
+error_return:
+ return -1;
}
/**
@@ -339,250 +363,241 @@ gf_init_event (gf_changelog_t *entry)
* - destroy rpc{-clnt, svc}
*/
int
-gf_cleanup_brick_connection (xlator_t *this, gf_changelog_t *entry)
+gf_cleanup_brick_connection(xlator_t *this, gf_changelog_t *entry)
{
- return 0;
+ return 0;
}
int
-gf_cleanup_connections (xlator_t *this)
+gf_cleanup_connections(xlator_t *this)
{
- return 0;
+ return 0;
}
static int
-gf_setup_brick_connection (xlator_t *this,
- struct gf_brick_spec *brick,
- gf_boolean_t ordered, void *xl)
+gf_setup_brick_connection(xlator_t *this, struct gf_brick_spec *brick,
+ gf_boolean_t ordered, void *xl)
{
- int ret = 0;
- gf_private_t *priv = NULL;
- gf_changelog_t *entry = NULL;
-
- priv = this->private;
-
- if (!brick->callback || !brick->init || !brick->fini)
- goto error_return;
-
- entry = GF_CALLOC (1, sizeof (*entry),
- gf_changelog_mt_libgfchangelog_t);
- if (!entry)
- goto error_return;
- INIT_LIST_HEAD (&entry->list);
-
- LOCK_INIT (&entry->statelock);
- entry->connstate = GF_CHANGELOG_CONN_STATE_PENDING;
-
- entry->notify = brick->filter;
- (void) strncpy (entry->brick, brick->brick_path, PATH_MAX-1);
- entry->brick[PATH_MAX-1] = 0;
-
- entry->this = this;
- entry->invokerxl = xl;
-
- entry->ordered = ordered;
- ret = gf_init_event (entry);
- if (ret)
- goto free_entry;
-
- entry->fini = brick->fini;
- entry->callback = brick->callback;
- entry->connected = brick->connected;
- entry->disconnected = brick->disconnected;
-
- entry->ptr = brick->init (this, brick);
- if (!entry->ptr)
- goto cleanup_event;
- priv->api = entry->ptr; /* pointer to API, if required */
-
- pthread_mutex_lock (&priv->lock);
- {
- list_add_tail (&entry->list, &priv->connections);
- }
- pthread_mutex_unlock (&priv->lock);
-
- ret = gf_changelog_setup_rpc (this, entry, CHANGELOG_RPC_PROBE_FILTER);
- if (ret)
- goto cleanup_event;
- return 0;
-
- cleanup_event:
- (void) gf_cleanup_event (this, &entry->event);
- free_entry:
- gf_msg_debug (this->name, 0, "freeing entry %p", entry);
- list_del (&entry->list); /* FIXME: kludge for now */
- GF_FREE (entry);
- error_return:
- return -1;
+ int ret = 0;
+ gf_private_t *priv = NULL;
+ gf_changelog_t *entry = NULL;
+
+ priv = this->private;
+
+ if (!brick->callback || !brick->init || !brick->fini)
+ goto error_return;
+
+ entry = GF_CALLOC(1, sizeof(*entry), gf_changelog_mt_libgfchangelog_t);
+ if (!entry)
+ goto error_return;
+ INIT_LIST_HEAD(&entry->list);
+
+ LOCK_INIT(&entry->statelock);
+ entry->connstate = GF_CHANGELOG_CONN_STATE_PENDING;
+
+ entry->notify = brick->filter;
+ if (snprintf(entry->brick, PATH_MAX, "%s", brick->brick_path) >= PATH_MAX)
+ goto free_entry;
+
+ entry->this = this;
+ entry->invokerxl = xl;
+
+ entry->ordered = ordered;
+ ret = gf_init_event(entry);
+ if (ret)
+ goto free_entry;
+
+ entry->fini = brick->fini;
+ entry->callback = brick->callback;
+ entry->connected = brick->connected;
+ entry->disconnected = brick->disconnected;
+
+ entry->ptr = brick->init(this, brick);
+ if (!entry->ptr)
+ goto cleanup_event;
+ priv->api = entry->ptr; /* pointer to API, if required */
+
+ pthread_mutex_lock(&priv->lock);
+ {
+ list_add_tail(&entry->list, &priv->connections);
+ }
+ pthread_mutex_unlock(&priv->lock);
+
+ ret = gf_changelog_setup_rpc(this, entry, CHANGELOG_RPC_PROBE_FILTER);
+ if (ret)
+ goto cleanup_event;
+ return 0;
+
+cleanup_event:
+ (void)gf_cleanup_event(this, &entry->event);
+free_entry:
+ gf_msg_debug(this->name, 0, "freeing entry %p", entry);
+ list_del(&entry->list); /* FIXME: kludge for now */
+ GF_FREE(entry);
+error_return:
+ return -1;
}
int
-gf_changelog_register_brick (xlator_t *this,
- struct gf_brick_spec *brick,
- gf_boolean_t ordered, void *xl)
+gf_changelog_register_brick(xlator_t *this, struct gf_brick_spec *brick,
+ gf_boolean_t ordered, void *xl)
{
- return gf_setup_brick_connection (this, brick, ordered, xl);
+ return gf_setup_brick_connection(this, brick, ordered, xl);
}
static int
-gf_changelog_setup_logging (xlator_t *this, char *logfile, int loglevel)
+gf_changelog_setup_logging(xlator_t *this, char *logfile, int loglevel)
{
- /* passing ident as NULL means to use default ident for syslog */
- if (gf_log_init (this->ctx, logfile, NULL))
- return -1;
+ /* passing ident as NULL means to use default ident for syslog */
+ if (gf_log_init(this->ctx, logfile, NULL))
+ return -1;
- gf_log_set_loglevel (this->ctx, (loglevel == -1) ? GF_LOG_INFO :
- loglevel);
- return 0;
+ gf_log_set_loglevel(this->ctx, (loglevel == -1) ? GF_LOG_INFO : loglevel);
+ return 0;
}
static int
-gf_changelog_set_master (xlator_t *master, void *xl)
+gf_changelog_set_master(xlator_t *master, void *xl)
{
- int32_t ret = 0;
- xlator_t *this = NULL;
- xlator_t *old_this = NULL;
- gf_private_t *priv = NULL;
-
- this = xl;
- if (!this || !this->ctx) {
- ret = gf_changelog_init_master ();
- if (ret)
- return -1;
- this = THIS;
- }
+ int32_t ret = 0;
+ xlator_t *this = NULL;
+ xlator_t *old_this = NULL;
+ gf_private_t *priv = NULL;
+
+ this = xl;
+ if (!this || !this->ctx) {
+ ret = gf_changelog_init_master();
+ if (ret)
+ return -1;
+ this = THIS;
+ }
- master->ctx = this->ctx;
+ master->ctx = this->ctx;
- INIT_LIST_HEAD (&master->volume_options);
- SAVE_THIS (THIS);
+ INIT_LIST_HEAD(&master->volume_options);
+ SAVE_THIS(THIS);
- ret = xlator_mem_acct_init (THIS, gf_changelog_mt_end);
- if (ret != 0)
- goto restore_this;
+ ret = xlator_mem_acct_init(THIS, gf_changelog_mt_end);
+ if (ret != 0)
+ goto restore_this;
- priv = gf_changelog_alloc_priv ();
- if (!priv) {
- ret = -1;
- goto restore_this;
- }
+ priv = gf_changelog_alloc_priv();
+ if (!priv) {
+ ret = -1;
+ goto restore_this;
+ }
- if (!xl) {
- /* poller thread */
- ret = gf_thread_create (&priv->poller,
- NULL, changelog_rpc_poller, THIS,
- "clogpoll");
- if (ret != 0) {
- GF_FREE (priv);
- gf_msg (master->name, GF_LOG_ERROR, 0,
- CHANGELOG_LIB_MSG_THREAD_CREATION_FAILED,
- "failed to spawn poller thread");
- goto restore_this;
- }
+ if (!xl) {
+ /* poller thread */
+ ret = gf_thread_create(&priv->poller, NULL, changelog_rpc_poller, THIS,
+ "clogpoll");
+ if (ret != 0) {
+ GF_FREE(priv);
+ gf_msg(master->name, GF_LOG_ERROR, 0,
+ CHANGELOG_LIB_MSG_THREAD_CREATION_FAILED,
+ "failed to spawn poller thread");
+ goto restore_this;
}
+ }
- master->private = priv;
+ master->private = priv;
- restore_this:
- RESTORE_THIS ();
+restore_this:
+ RESTORE_THIS();
- return ret;
+ return ret;
}
int
-gf_changelog_init (void *xl)
+gf_changelog_init(void *xl)
{
- int ret = 0;
- gf_private_t *priv = NULL;
-
- if (master)
- return 0;
-
- master = calloc (1, sizeof (*master));
- if (!master)
- goto error_return;
-
- master->name = strdup ("gfchangelog");
- if (!master->name)
- goto dealloc_master;
-
- ret = gf_changelog_set_master (master, xl);
- if (ret)
- goto dealloc_name;
-
- priv = master->private;
- ret = gf_thread_create (&priv->connectionjanitor, NULL,
- gf_changelog_connection_janitor, master,
- "clogjan");
- if (ret != 0) {
- /* TODO: cleanup priv, mutex (poller thread for !xl) */
- goto dealloc_name;
- }
+ int ret = 0;
+ gf_private_t *priv = NULL;
+ if (master)
return 0;
- dealloc_name:
- free (master->name);
- dealloc_master:
- free (master);
- master = NULL;
- error_return:
- return -1;
+ master = calloc(1, sizeof(*master));
+ if (!master)
+ goto error_return;
+
+ master->name = strdup("gfchangelog");
+ if (!master->name)
+ goto dealloc_master;
+
+ ret = gf_changelog_set_master(master, xl);
+ if (ret)
+ goto dealloc_name;
+
+ priv = master->private;
+ ret = gf_thread_create(&priv->connectionjanitor, NULL,
+ gf_changelog_connection_janitor, master, "clogjan");
+ if (ret != 0) {
+ /* TODO: cleanup priv, mutex (poller thread for !xl) */
+ goto dealloc_name;
+ }
+
+ return 0;
+
+dealloc_name:
+ free(master->name);
+dealloc_master:
+ free(master);
+ master = NULL;
+error_return:
+ return -1;
}
int
-gf_changelog_register_generic (struct gf_brick_spec *bricks, int count,
- int ordered, char *logfile, int lvl, void *xl)
+gf_changelog_register_generic(struct gf_brick_spec *bricks, int count,
+ int ordered, char *logfile, int lvl, void *xl)
{
- int ret = 0;
- xlator_t *this = NULL;
- xlator_t *old_this = NULL;
- struct gf_brick_spec *brick = NULL;
- gf_boolean_t need_order = _gf_false;
+ int ret = 0;
+ xlator_t *this = NULL;
+ xlator_t *old_this = NULL;
+ struct gf_brick_spec *brick = NULL;
+ gf_boolean_t need_order = _gf_false;
- SAVE_THIS (xl);
+ SAVE_THIS(xl);
- this = THIS;
- if (!this)
- goto error_return;
+ this = THIS;
+ if (!this)
+ goto error_return;
- ret = gf_changelog_setup_logging (this, logfile, lvl);
- if (ret)
- goto error_return;
-
- need_order = (ordered) ? _gf_true : _gf_false;
-
- brick = bricks;
- while (count--) {
- gf_smsg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_LIB_MSG_NOTIFY_REGISTER_INFO,
- "Registering brick",
- "brick=%s", brick->brick_path,
- "notify_filter=%d", brick->filter,
- NULL);
-
- ret = gf_changelog_register_brick (this, brick, need_order, xl);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_LIB_MSG_NOTIFY_REGISTER_FAILED,
- "Error registering with changelog xlator");
- break;
- }
-
- brick++;
+ ret = gf_changelog_setup_logging(this, logfile, lvl);
+ if (ret)
+ goto error_return;
+
+ need_order = (ordered) ? _gf_true : _gf_false;
+
+ brick = bricks;
+ while (count--) {
+ gf_smsg(this->name, GF_LOG_INFO, 0,
+ CHANGELOG_LIB_MSG_NOTIFY_REGISTER_INFO, "brick=%s",
+ brick->brick_path, "notify_filter=%d", brick->filter, NULL);
+
+ ret = gf_changelog_register_brick(this, brick, need_order, xl);
+ if (ret != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_LIB_MSG_NOTIFY_REGISTER_FAILED,
+ "Error registering with changelog xlator");
+ break;
}
- if (ret != 0)
- goto cleanup_inited_bricks;
+ brick++;
+ }
- RESTORE_THIS();
- return 0;
+ if (ret != 0)
+ goto cleanup_inited_bricks;
- cleanup_inited_bricks:
- gf_cleanup_connections (this);
- error_return:
- RESTORE_THIS();
- return -1;
+ RESTORE_THIS();
+ return 0;
+
+cleanup_inited_bricks:
+ gf_cleanup_connections(this);
+error_return:
+ RESTORE_THIS();
+ return -1;
}
/**
@@ -609,27 +624,29 @@ gf_changelog_register_generic (struct gf_brick_spec *bricks, int count,
* For generic API, refer gf_changelog_register_generic().
*/
int
-gf_changelog_register (char *brick_path, char *scratch_dir,
- char *log_file, int log_level, int max_reconnects)
+gf_changelog_register(char *brick_path, char *scratch_dir, char *log_file,
+ int log_level, int max_reconnects)
{
- struct gf_brick_spec brick = {0,};
+ struct gf_brick_spec brick = {
+ 0,
+ };
- if (master)
- THIS = master;
- else
- return -1;
+ if (master)
+ THIS = master;
+ else
+ return -1;
- brick.brick_path = brick_path;
- brick.filter = CHANGELOG_OP_TYPE_JOURNAL;
+ brick.brick_path = brick_path;
+ brick.filter = CHANGELOG_OP_TYPE_JOURNAL;
- brick.init = gf_changelog_journal_init;
- brick.fini = gf_changelog_journal_fini;
- brick.callback = gf_changelog_handle_journal;
- brick.connected = gf_changelog_journal_connect;
- brick.disconnected = gf_changelog_journal_disconnect;
+ brick.init = gf_changelog_journal_init;
+ brick.fini = gf_changelog_journal_fini;
+ brick.callback = gf_changelog_handle_journal;
+ brick.connected = gf_changelog_journal_connect;
+ brick.disconnected = gf_changelog_journal_disconnect;
- brick.ptr = scratch_dir;
+ brick.ptr = scratch_dir;
- return gf_changelog_register_generic (&brick, 1, 1,
- log_file, log_level, NULL);
+ return gf_changelog_register_generic(&brick, 1, 1, log_file, log_level,
+ NULL);
}
diff --git a/xlators/features/changelog/lib/src/gf-history-changelog.c b/xlators/features/changelog/lib/src/gf-history-changelog.c
index 2f245a725f6..a16219f3664 100644
--- a/xlators/features/changelog/lib/src/gf-history-changelog.c
+++ b/xlators/features/changelog/lib/src/gf-history-changelog.c
@@ -8,10 +8,10 @@
#endif
#include <string.h>
-#include "globals.h"
-#include "glusterfs.h"
-#include "logging.h"
-#include "syscall.h"
+#include <glusterfs/globals.h>
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/syscall.h>
#include "gf-changelog-helpers.h"
#include "gf-changelog-journal.h"
@@ -36,62 +36,60 @@
* -1: On error.
*/
int
-gf_history_changelog_done (char *file)
+gf_history_changelog_done(char *file)
{
- int ret = -1;
- char *buffer = NULL;
- xlator_t *this = NULL;
- gf_changelog_journal_t *jnl = NULL;
- gf_changelog_journal_t *hist_jnl = NULL;
- char to_path[PATH_MAX] = {0,};
+ int ret = -1;
+ char *buffer = NULL;
+ xlator_t *this = NULL;
+ gf_changelog_journal_t *jnl = NULL;
+ gf_changelog_journal_t *hist_jnl = NULL;
+ char to_path[PATH_MAX] = {
+ 0,
+ };
+
+ errno = EINVAL;
+
+ this = THIS;
+ if (!this)
+ goto out;
+
+ jnl = (gf_changelog_journal_t *)GF_CHANGELOG_GET_API_PTR(this);
+ if (!jnl)
+ goto out;
+
+ hist_jnl = jnl->hist_jnl;
+ if (!hist_jnl)
+ goto out;
+
+ if (!file || !strlen(file))
+ goto out;
+
+ /* make sure 'file' is inside ->jnl_working_dir */
+ buffer = realpath(file, NULL);
+ if (!buffer)
+ goto out;
+
+ if (strncmp(hist_jnl->jnl_working_dir, buffer,
+ strlen(hist_jnl->jnl_working_dir)))
+ goto out;
+
+ (void)snprintf(to_path, PATH_MAX, "%s%s", hist_jnl->jnl_processed_dir,
+ basename(buffer));
+ gf_msg_debug(this->name, 0, "moving %s to processed directory", file);
+ ret = sys_rename(buffer, to_path);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_LIB_MSG_RENAME_FAILED, "from=%s", file, "to=%s",
+ to_path, NULL);
+ goto out;
+ }
+
+ ret = 0;
- errno = EINVAL;
-
- this = THIS;
- if (!this)
- goto out;
-
- jnl = (gf_changelog_journal_t *) GF_CHANGELOG_GET_API_PTR (this);
- if (!jnl)
- goto out;
-
- hist_jnl = jnl->hist_jnl;
- if (!hist_jnl)
- goto out;
-
- if (!file || !strlen (file))
- goto out;
-
- /* make sure 'file' is inside ->jnl_working_dir */
- buffer = realpath (file, NULL);
- if (!buffer)
- goto out;
-
- if (strncmp (hist_jnl->jnl_working_dir,
- buffer, strlen (hist_jnl->jnl_working_dir)))
- goto out;
-
- (void) snprintf (to_path, PATH_MAX, "%s%s",
- hist_jnl->jnl_processed_dir, basename (buffer));
- gf_msg_debug (this->name, 0,
- "moving %s to processed directory", file);
- ret = sys_rename (buffer, to_path);
- if (ret) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_RENAME_FAILED,
- "cannot move changelog file",
- "from=%s", file,
- "to=%s", to_path,
- NULL);
- goto out;
- }
-
- ret = 0;
-
- out:
- if (buffer)
- free (buffer); /* allocated by realpath() */
- return ret;
+out:
+ if (buffer)
+ free(buffer); /* allocated by realpath() */
+ return ret;
}
/**
@@ -105,33 +103,33 @@ gf_history_changelog_done (char *file)
* -1: On error.
*/
int
-gf_history_changelog_start_fresh ()
+gf_history_changelog_start_fresh()
{
- xlator_t *this = NULL;
- gf_changelog_journal_t *jnl = NULL;
- gf_changelog_journal_t *hist_jnl = NULL;
+ xlator_t *this = NULL;
+ gf_changelog_journal_t *jnl = NULL;
+ gf_changelog_journal_t *hist_jnl = NULL;
- this = THIS;
- if (!this)
- goto out;
+ this = THIS;
+ if (!this)
+ goto out;
- errno = EINVAL;
+ errno = EINVAL;
- jnl = (gf_changelog_journal_t *) GF_CHANGELOG_GET_API_PTR (this);
- if (!jnl)
- goto out;
+ jnl = (gf_changelog_journal_t *)GF_CHANGELOG_GET_API_PTR(this);
+ if (!jnl)
+ goto out;
- hist_jnl = jnl->hist_jnl;
- if (!hist_jnl)
- goto out;
+ hist_jnl = jnl->hist_jnl;
+ if (!hist_jnl)
+ goto out;
- if (gf_ftruncate (hist_jnl->jnl_fd, 0))
- goto out;
+ if (gf_ftruncate(hist_jnl->jnl_fd, 0))
+ goto out;
- return 0;
+ return 0;
- out:
- return -1;
+out:
+ return -1;
}
/**
@@ -150,50 +148,52 @@ gf_history_changelog_start_fresh ()
* -1 : On error.
*/
ssize_t
-gf_history_changelog_next_change (char *bufptr, size_t maxlen)
+gf_history_changelog_next_change(char *bufptr, size_t maxlen)
{
- ssize_t size = -1;
- int tracker_fd = 0;
- xlator_t *this = NULL;
- gf_changelog_journal_t *jnl = NULL;
- gf_changelog_journal_t *hist_jnl = NULL;
- char buffer[PATH_MAX] = {0,};
-
- if (maxlen > PATH_MAX) {
- errno = ENAMETOOLONG;
- goto out;
- }
+ ssize_t size = -1;
+ int tracker_fd = 0;
+ xlator_t *this = NULL;
+ gf_changelog_journal_t *jnl = NULL;
+ gf_changelog_journal_t *hist_jnl = NULL;
+ char buffer[PATH_MAX] = {
+ 0,
+ };
- errno = EINVAL;
+ if (maxlen > PATH_MAX) {
+ errno = ENAMETOOLONG;
+ goto out;
+ }
- this = THIS;
- if (!this)
- goto out;
+ errno = EINVAL;
- jnl = (gf_changelog_journal_t *) GF_CHANGELOG_GET_API_PTR (this);
- if (!jnl)
- goto out;
+ this = THIS;
+ if (!this)
+ goto out;
- hist_jnl = jnl->hist_jnl;
- if (!hist_jnl)
- goto out;
+ jnl = (gf_changelog_journal_t *)GF_CHANGELOG_GET_API_PTR(this);
+ if (!jnl)
+ goto out;
- tracker_fd = hist_jnl->jnl_fd;
+ hist_jnl = jnl->hist_jnl;
+ if (!hist_jnl)
+ goto out;
- size = gf_readline (tracker_fd, buffer, maxlen);
- if (size < 0) {
- size = -1;
- goto out;
- }
+ tracker_fd = hist_jnl->jnl_fd;
- if (size == 0)
- goto out;
+ size = gf_readline(tracker_fd, buffer, maxlen);
+ if (size < 0) {
+ size = -1;
+ goto out;
+ }
- memcpy (bufptr, buffer, size - 1);
- bufptr[size - 1] = '\0';
+ if (size == 0)
+ goto out;
+
+ memcpy(bufptr, buffer, size - 1);
+ bufptr[size - 1] = '\0';
out:
- return size;
+ return size;
}
/**
@@ -214,97 +214,100 @@ out:
*
*/
ssize_t
-gf_history_changelog_scan ()
+gf_history_changelog_scan()
{
- int tracker_fd = 0;
- size_t off = 0;
- xlator_t *this = NULL;
- size_t nr_entries = 0;
- gf_changelog_journal_t *jnl = NULL;
- gf_changelog_journal_t *hist_jnl = NULL;
- struct dirent *entry = NULL;
- struct dirent scratch[2] = {{0,},};
- char buffer[PATH_MAX] = {0,};
- static int is_last_scan;
-
- this = THIS;
- if (!this)
- goto out;
+ int tracker_fd = 0;
+ size_t off = 0;
+ xlator_t *this = NULL;
+ size_t nr_entries = 0;
+ gf_changelog_journal_t *jnl = NULL;
+ gf_changelog_journal_t *hist_jnl = NULL;
+ struct dirent *entry = NULL;
+ struct dirent scratch[2] = {
+ {
+ 0,
+ },
+ };
+ char buffer[PATH_MAX] = {
+ 0,
+ };
+ static int is_last_scan;
+
+ this = THIS;
+ if (!this)
+ goto out;
+
+ jnl = (gf_changelog_journal_t *)GF_CHANGELOG_GET_API_PTR(this);
+ if (!jnl)
+ goto out;
+ if (JNL_IS_API_DISCONNECTED(jnl)) {
+ errno = ENOTCONN;
+ goto out;
+ }
+
+ hist_jnl = jnl->hist_jnl;
+ if (!hist_jnl)
+ goto out;
+
+retry:
+ if (is_last_scan == 1)
+ return 0;
+ if (hist_jnl->hist_done == 0)
+ is_last_scan = 1;
- jnl = (gf_changelog_journal_t *) GF_CHANGELOG_GET_API_PTR (this);
- if (!jnl)
- goto out;
- if (JNL_IS_API_DISCONNECTED (jnl)) {
- errno = ENOTCONN;
- goto out;
- }
+ errno = EINVAL;
+ if (hist_jnl->hist_done == -1)
+ goto out;
- hist_jnl = jnl->hist_jnl;
- if (!hist_jnl)
- goto out;
+ tracker_fd = hist_jnl->jnl_fd;
- retry:
- if (is_last_scan == 1)
- return 0;
- if (hist_jnl->hist_done == 0)
- is_last_scan = 1;
+ if (gf_ftruncate(tracker_fd, 0))
+ goto out;
- errno = EINVAL;
- if (hist_jnl->hist_done == -1)
- goto out;
+ rewinddir(hist_jnl->jnl_dir);
- tracker_fd = hist_jnl->jnl_fd;
+ for (;;) {
+ errno = 0;
+ entry = sys_readdir(hist_jnl->jnl_dir, scratch);
+ if (!entry || errno != 0)
+ break;
- if (gf_ftruncate (tracker_fd, 0))
- goto out;
+ if (strcmp(basename(entry->d_name), ".") == 0 ||
+ strcmp(basename(entry->d_name), "..") == 0)
+ continue;
+
+ nr_entries++;
- rewinddir (hist_jnl->jnl_dir);
-
- for (;;) {
- errno = 0;
- entry = sys_readdir (hist_jnl->jnl_dir, scratch);
- if (!entry || errno != 0)
- break;
-
- if (strcmp (basename (entry->d_name), ".") == 0 ||
- strcmp (basename (entry->d_name), "..") == 0)
- continue;
-
- nr_entries++;
-
- GF_CHANGELOG_FILL_BUFFER (hist_jnl->jnl_processing_dir,
- buffer, off,
- strlen (hist_jnl->jnl_processing_dir));
- GF_CHANGELOG_FILL_BUFFER (entry->d_name, buffer,
- off, strlen (entry->d_name));
- GF_CHANGELOG_FILL_BUFFER ("\n", buffer, off, 1);
-
- if (gf_changelog_write (tracker_fd, buffer, off) != off) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_LIB_MSG_WRITE_FAILED,
- "error writing changelog filename"
- " to tracker file");
- break;
- }
- off = 0;
+ GF_CHANGELOG_FILL_BUFFER(hist_jnl->jnl_processing_dir, buffer, off,
+ strlen(hist_jnl->jnl_processing_dir));
+ GF_CHANGELOG_FILL_BUFFER(entry->d_name, buffer, off,
+ strlen(entry->d_name));
+ GF_CHANGELOG_FILL_BUFFER("\n", buffer, off, 1);
+
+ if (gf_changelog_write(tracker_fd, buffer, off) != off) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, CHANGELOG_LIB_MSG_WRITE_FAILED,
+ "error writing changelog filename"
+ " to tracker file");
+ break;
}
+ off = 0;
+ }
+
+ gf_msg_debug(this->name, 0, "hist_done %d, is_last_scan: %d",
+ hist_jnl->hist_done, is_last_scan);
- gf_msg_debug (this->name, 0,
- "hist_done %d, is_last_scan: %d",
- hist_jnl->hist_done, is_last_scan);
-
- if (!entry) {
- if (gf_lseek (tracker_fd, 0, SEEK_SET) != -1) {
- if (nr_entries > 0)
- return nr_entries;
- else {
- sleep(1);
- goto retry;
- }
- }
+ if (!entry) {
+ if (gf_lseek(tracker_fd, 0, SEEK_SET) != -1) {
+ if (nr_entries > 0)
+ return nr_entries;
+ else {
+ sleep(1);
+ goto retry;
+ }
}
- out:
- return -1;
+ }
+out:
+ return -1;
}
/*
@@ -312,36 +315,36 @@ gf_history_changelog_scan ()
* Returns 0 on success(updates given time-stamp), -1 on failure.
*/
int
-gf_history_get_timestamp (int fd, int index, int len,
- unsigned long *ts)
+gf_history_get_timestamp(int fd, int index, int len, unsigned long *ts)
{
- xlator_t *this = NULL;
- int n_read = -1;
- char path_buf[PATH_MAX]= {0,};
- char *iter = path_buf;
- size_t offset = index * (len+1);
- unsigned long value = 0;
- int ret = 0;
-
- this = THIS;
- if (!this) {
- return -1;
- }
-
- n_read = sys_pread (fd, path_buf, len, offset);
- if (n_read < 0 ) {
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_READ_ERROR,
- "could not read from htime file");
- goto out;
- }
- iter+= len - TIMESTAMP_LENGTH;
- sscanf (iter, "%lu",&value);
+ xlator_t *this = NULL;
+ int n_read = -1;
+ char path_buf[PATH_MAX] = {
+ 0,
+ };
+ char *iter = path_buf;
+ size_t offset = index * (len + 1);
+ unsigned long value = 0;
+ int ret = 0;
+
+ this = THIS;
+ if (!this) {
+ return -1;
+ }
+
+ n_read = sys_pread(fd, path_buf, len, offset);
+ if (n_read < 0) {
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, errno, CHANGELOG_LIB_MSG_READ_ERROR,
+ "could not read from htime file");
+ goto out;
+ }
+ iter += len - TIMESTAMP_LENGTH;
+ sscanf(iter, "%lu", &value);
out:
- if(ret == 0)
- *ts = value;
- return ret;
+ if (ret == 0)
+ *ts = value;
+ return ret;
}
/*
@@ -349,38 +352,37 @@ out:
* Checks whether @value is there next to @target_index or not
*/
int
-gf_history_check ( int fd, int target_index, unsigned long value, int len)
+gf_history_check(int fd, int target_index, unsigned long value, int len)
{
- int ret = 0;
- unsigned long ts1 = 0;
- unsigned long ts2 = 0;
-
- if (target_index == 0) {
- ret = gf_history_get_timestamp (fd, target_index, len, &ts1);
- if (ret == -1)
- goto out;
- if (value <= ts1)
- goto out;
- else {
- ret = -1;
- goto out;
- }
- }
+ int ret = 0;
+ unsigned long ts1 = 0;
+ unsigned long ts2 = 0;
- ret = gf_history_get_timestamp (fd, target_index, len, &ts1);
- if (ret ==-1)
- goto out;
- ret = gf_history_get_timestamp (fd, target_index -1, len, &ts2);
- if (ret ==-1)
- goto out;
-
- if ( (value <= ts1) && (value > ts2) ) {
- goto out;
- }
- else
- ret = -1;
+ if (target_index == 0) {
+ ret = gf_history_get_timestamp(fd, target_index, len, &ts1);
+ if (ret == -1)
+ goto out;
+ if (value <= ts1)
+ goto out;
+ else {
+ ret = -1;
+ goto out;
+ }
+ }
+
+ ret = gf_history_get_timestamp(fd, target_index, len, &ts1);
+ if (ret == -1)
+ goto out;
+ ret = gf_history_get_timestamp(fd, target_index - 1, len, &ts2);
+ if (ret == -1)
+ goto out;
+
+ if ((value <= ts1) && (value > ts2)) {
+ goto out;
+ } else
+ ret = -1;
out:
- return ret;
+ return ret;
}
/*
@@ -400,83 +402,74 @@ out:
*/
int
-gf_history_b_search (int fd, unsigned long value,
- unsigned long from, unsigned long to, int len)
+gf_history_b_search(int fd, unsigned long value, unsigned long from,
+ unsigned long to, int len)
{
- int m_index = -1;
- unsigned long cur_value = 0;
- unsigned long ts1 = 0;
- int ret = 0;
-
- m_index = (from + to)/2;
-
- if ( (to - from) <=1 ) {
- /* either one or 2 changelogs left */
- if ( to != from ) {
- /* check if value is less or greater than to
- * return accordingly
- */
- ret = gf_history_get_timestamp (fd, from, len, &ts1);
- if (ret ==-1)
- goto out;
- if ( ts1 >= value) {
- /* actually compatision should be
- * exactly == but considering
- *
- * case of only 2 changelogs in htime file
- */
- return from;
- }
- else
- return to;
- }
- else
- return to;
- }
-
- ret = gf_history_get_timestamp (fd, m_index, len, &cur_value);
+ int m_index = -1;
+ unsigned long cur_value = 0;
+ unsigned long ts1 = 0;
+ int ret = 0;
+
+ m_index = (from + to) / 2;
+
+ if ((to - from) <= 1) {
+ /* either one or 2 changelogs left */
+ if (to != from) {
+ /* check if value is less or greater than to
+ * return accordingly
+ */
+ ret = gf_history_get_timestamp(fd, from, len, &ts1);
+ if (ret == -1)
+ goto out;
+ if (ts1 >= value) {
+ /* actually compatision should be
+ * exactly == but considering
+ *
+ * case of only 2 changelogs in htime file
+ */
+ return from;
+ } else
+ return to;
+ } else
+ return to;
+ }
+
+ ret = gf_history_get_timestamp(fd, m_index, len, &cur_value);
+ if (ret == -1)
+ goto out;
+ if (cur_value == value) {
+ return m_index;
+ } else if (value > cur_value) {
+ ret = gf_history_get_timestamp(fd, m_index + 1, len, &cur_value);
if (ret == -1)
- goto out;
- if (cur_value == value) {
+ goto out;
+ if (value < cur_value)
+ return m_index + 1;
+ else
+ return gf_history_b_search(fd, value, m_index + 1, to, len);
+ } else {
+ if (m_index == 0) {
+ /* we are sure that values exists
+ * in this htime file
+ */
+ return 0;
+ } else {
+ ret = gf_history_get_timestamp(fd, m_index - 1, len, &cur_value);
+ if (ret == -1)
+ goto out;
+ if (value > cur_value) {
return m_index;
+ } else
+ return gf_history_b_search(fd, value, from, m_index - 1, len);
}
- else if (value > cur_value) {
- ret = gf_history_get_timestamp (fd, m_index+1, len, &cur_value);
- if (ret == -1)
- goto out;
- if (value < cur_value)
- return m_index + 1;
- else
- return gf_history_b_search (fd, value,
- m_index+1, to, len);
- }
- else {
- if (m_index ==0) {
- /* we are sure that values exists
- * in this htime file
- */
- return 0;
- }
- else {
- ret = gf_history_get_timestamp (fd, m_index-1, len,
- &cur_value);
- if (ret == -1)
- goto out;
- if (value > cur_value) {
- return m_index;
- }
- else
- return gf_history_b_search (fd, value, from,
- m_index-1, len);
- }
- }
+ }
out:
- return -1;
+ return -1;
}
/*
* Description: Checks if the changelog path is usable or not,
- * which is differenciated by checking for "changelog"
+ * which is differentiated by checking for "changelog"
* in the path and not "CHANGELOG".
*
* Returns:
@@ -484,65 +477,59 @@ out:
* 0 : No, Not usable ( contains, "changelog")
*/
int
-gf_is_changelog_usable (char *cl_path)
+gf_is_changelog_usable(char *cl_path)
{
- int ret = -1;
- const char low_c[] = "changelog";
- char *str_ret = NULL;
- char *bname = NULL;
+ int ret = -1;
+ const char low_c[] = "changelog";
+ char *str_ret = NULL;
+ char *bname = NULL;
- bname = basename (cl_path);
+ bname = basename(cl_path);
- str_ret = strstr (bname, low_c);
+ str_ret = strstr(bname, low_c);
- if (str_ret != NULL)
- ret = 0;
- else
- ret = 1;
-
- return ret;
+ if (str_ret != NULL)
+ ret = 0;
+ else
+ ret = 1;
+ return ret;
}
void *
-gf_changelog_consume_wrap (void* data)
+gf_changelog_consume_wrap(void *data)
{
- int ret = -1;
- ssize_t nread = 0;
- xlator_t *this = NULL;
- gf_changelog_consume_data_t *ccd = NULL;
-
- ccd = (gf_changelog_consume_data_t *) data;
- this = ccd->this;
-
- ccd->retval = -1;
-
- nread = sys_pread (ccd->fd, ccd->changelog, PATH_MAX-1, ccd->offset);
- if (nread < 0) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_READ_ERROR,
- "cannot read from history metadata file");
- goto out;
- }
-
- /* TODO: handle short reads and EOF. */
- if (gf_is_changelog_usable (ccd->changelog) == 1) {
-
- ret = gf_changelog_consume (ccd->this,
- ccd->jnl, ccd->changelog, _gf_true);
- if (ret) {
- gf_smsg (this->name, GF_LOG_ERROR,
- 0, CHANGELOG_LIB_MSG_PARSE_ERROR,
- "could not parse changelog",
- "name=%s", ccd->changelog,
- NULL);
- goto out;
- }
+ int ret = -1;
+ ssize_t nread = 0;
+ xlator_t *this = NULL;
+ gf_changelog_consume_data_t *ccd = NULL;
+
+ ccd = (gf_changelog_consume_data_t *)data;
+ this = ccd->this;
+
+ ccd->retval = -1;
+
+ nread = sys_pread(ccd->fd, ccd->changelog, PATH_MAX - 1, ccd->offset);
+ if (nread < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, errno, CHANGELOG_LIB_MSG_READ_ERROR,
+ "cannot read from history metadata file");
+ goto out;
+ }
+
+ /* TODO: handle short reads and EOF. */
+ if (gf_is_changelog_usable(ccd->changelog) == 1) {
+ ret = gf_changelog_consume(ccd->this, ccd->jnl, ccd->changelog,
+ _gf_true);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_LIB_MSG_PARSE_ERROR,
+ "name=%s", ccd->changelog, NULL);
+ goto out;
}
- ccd->retval = 0;
+ }
+ ccd->retval = 0;
- out:
- return NULL;
+out:
+ return NULL;
}
/**
@@ -551,133 +538,132 @@ gf_changelog_consume_wrap (void* data)
* to index "to" in open htime file whose fd is "fd".
*/
-#define MAX_PARALLELS 10
+#define MAX_PARALLELS 10
void *
-gf_history_consume (void * data)
+gf_history_consume(void *data)
{
- xlator_t *this = NULL;
- gf_changelog_journal_t *jnl = NULL;
- gf_changelog_journal_t *hist_jnl = NULL;
- int ret = 0;
- int iter = 0;
- int fd = -1;
- int from = -1;
- int to = -1;
- int len = -1;
- int n_parallel = 0;
- int n_envoked = 0;
- gf_boolean_t publish = _gf_true;
- pthread_t th_id[MAX_PARALLELS] = {0,};
- gf_changelog_history_data_t *hist_data = NULL;
- gf_changelog_consume_data_t ccd[MAX_PARALLELS] = {{0},};
- gf_changelog_consume_data_t *curr = NULL;
- char thread_name[GF_THREAD_NAMEMAX] = {0,};
-
- hist_data = (gf_changelog_history_data_t *) data;
- if (hist_data == NULL) {
- ret = -1;
- goto out;
- }
-
- fd = hist_data->htime_fd;
- from = hist_data->from;
- to = hist_data->to;
- len = hist_data->len;
- n_parallel = hist_data->n_parallel;
-
- THIS = hist_data->this;
- this = hist_data->this;
- if (!this) {
- ret = -1;
- goto out;
- }
-
- jnl = (gf_changelog_journal_t *) GF_CHANGELOG_GET_API_PTR (this);
- if (!jnl) {
- ret = -1;
- goto out;
- }
-
- hist_jnl = jnl->hist_jnl;
- if (!hist_jnl) {
- ret = -1;
- goto out;
- }
-
- while (from <= to) {
- n_envoked = 0;
-
- for (iter = 0 ; (iter < n_parallel) && (from <= to); iter++) {
- curr = &ccd[iter];
-
- curr->this = this;
- curr->jnl = hist_jnl;
- curr->fd = fd;
- curr->offset = from * (len + 1);
-
- curr->retval = 0;
- memset (curr->changelog, '\0', PATH_MAX);
- snprintf (thread_name, sizeof(thread_name), "%s%d",
- "clogc", iter + 1);
-
- ret = gf_thread_create (&th_id[iter], NULL,
- gf_changelog_consume_wrap, curr,
- thread_name);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, ret,
- CHANGELOG_LIB_MSG_THREAD_CREATION_FAILED
- , "could not create consume-thread");
- goto sync;
- } else
- n_envoked++;
-
- from++;
- }
-
- sync:
- for (iter = 0; iter < n_envoked; iter++) {
- ret = pthread_join (th_id[iter], NULL);
- if (ret) {
- publish = _gf_false;
- gf_msg (this->name, GF_LOG_ERROR, ret,
- CHANGELOG_LIB_MSG_PTHREAD_JOIN_FAILED,
- "pthread_join() error");
- /* try to join the rest */
- continue;
- }
-
- if (publish == _gf_false)
- continue;
-
- curr = &ccd[iter];
- if (ccd->retval) {
- publish = _gf_false;
- gf_msg (this->name, GF_LOG_ERROR,
- 0, CHANGELOG_LIB_MSG_PARSE_ERROR,
- "parsing error, ceased publishing...");
- continue;
- }
-
- ret = gf_changelog_publish (curr->this,
- curr->jnl, curr->changelog);
- if (ret) {
- publish = _gf_false;
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_LIB_MSG_PUBLISH_ERROR,
- "publish error, ceased publishing...");
- }
- }
- }
-
- /* informing "parsing done". */
- hist_jnl->hist_done = (publish == _gf_true) ? 0 : -1;
+ xlator_t *this = NULL;
+ gf_changelog_journal_t *jnl = NULL;
+ gf_changelog_journal_t *hist_jnl = NULL;
+ int ret = 0;
+ int iter = 0;
+ int fd = -1;
+ int from = -1;
+ int to = -1;
+ int len = -1;
+ int n_parallel = 0;
+ int n_envoked = 0;
+ gf_boolean_t publish = _gf_true;
+ pthread_t th_id[MAX_PARALLELS] = {
+ 0,
+ };
+ gf_changelog_history_data_t *hist_data = NULL;
+ gf_changelog_consume_data_t ccd[MAX_PARALLELS] = {
+ {0},
+ };
+ gf_changelog_consume_data_t *curr = NULL;
+
+ hist_data = (gf_changelog_history_data_t *)data;
+ if (hist_data == NULL) {
+ ret = -1;
+ goto out;
+ }
+
+ fd = hist_data->htime_fd;
+ from = hist_data->from;
+ to = hist_data->to;
+ len = hist_data->len;
+ n_parallel = hist_data->n_parallel;
+
+ THIS = hist_data->this;
+ this = hist_data->this;
+ if (!this) {
+ ret = -1;
+ goto out;
+ }
+
+ jnl = (gf_changelog_journal_t *)GF_CHANGELOG_GET_API_PTR(this);
+ if (!jnl) {
+ ret = -1;
+ goto out;
+ }
+
+ hist_jnl = jnl->hist_jnl;
+ if (!hist_jnl) {
+ ret = -1;
+ goto out;
+ }
+
+ while (from <= to) {
+ n_envoked = 0;
+
+ for (iter = 0; (iter < n_parallel) && (from <= to); iter++) {
+ curr = &ccd[iter];
+
+ curr->this = this;
+ curr->jnl = hist_jnl;
+ curr->fd = fd;
+ curr->offset = from * (len + 1);
+
+ curr->retval = 0;
+ memset(curr->changelog, '\0', PATH_MAX);
+
+ ret = gf_thread_create(&th_id[iter], NULL,
+ gf_changelog_consume_wrap, curr,
+ "clogc%03hx", (iter + 1) & 0x3ff);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, ret,
+ CHANGELOG_LIB_MSG_THREAD_CREATION_FAILED,
+ "could not create consume-thread");
+ goto sync;
+ } else
+ n_envoked++;
+
+ from++;
+ }
+
+ sync:
+ for (iter = 0; iter < n_envoked; iter++) {
+ ret = pthread_join(th_id[iter], NULL);
+ if (ret) {
+ publish = _gf_false;
+ gf_msg(this->name, GF_LOG_ERROR, ret,
+ CHANGELOG_LIB_MSG_PTHREAD_JOIN_FAILED,
+ "pthread_join() error");
+ /* try to join the rest */
+ continue;
+ }
+
+ if (publish == _gf_false)
+ continue;
+
+ curr = &ccd[iter];
+ if (ccd->retval) {
+ publish = _gf_false;
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_LIB_MSG_PARSE_ERROR_CEASED, NULL);
+ continue;
+ }
+
+ ret = gf_changelog_publish(curr->this, curr->jnl, curr->changelog);
+ if (ret) {
+ publish = _gf_false;
+ gf_msg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_LIB_MSG_PUBLISH_ERROR,
+ "publish error, ceased publishing...");
+ }
+ }
+ }
+
+ /* informing "parsing done". */
+ hist_jnl->hist_done = (publish == _gf_true) ? 0 : -1;
out:
- if (fd != -1)
- (void) sys_close (fd);
- GF_FREE (hist_data);
- return NULL;
+ if (fd != -1)
+ (void)sys_close(fd);
+ GF_FREE(hist_data);
+ return NULL;
}
/**
@@ -707,317 +693,328 @@ out:
* -2 : Ignore this metadata file and process next
*/
int
-gf_changelog_extract_min_max (const char *dname, const char *htime_dir,
- int *fd, unsigned long *total,
- unsigned long *min_ts, unsigned long *max_ts)
+gf_changelog_extract_min_max(const char *dname, const char *htime_dir, int *fd,
+ unsigned long *total, unsigned long *min_ts,
+ unsigned long *max_ts)
{
- int ret = -1;
- xlator_t *this = NULL;
- char htime_file[PATH_MAX] = {0,};
- struct stat stbuf = {0,};
- char *iter = NULL;
- char x_value[30] = {0,};
-
- this = THIS;
-
- snprintf (htime_file, PATH_MAX, "%s/%s", htime_dir, dname);
+ int ret = -1;
+ xlator_t *this = NULL;
+ char htime_file[PATH_MAX] = {
+ 0,
+ };
+ struct stat stbuf = {
+ 0,
+ };
+ char *iter = NULL;
+ char x_value[30] = {
+ 0,
+ };
+
+ this = THIS;
+
+ snprintf(htime_file, PATH_MAX, "%s/%s", htime_dir, dname);
+
+ iter = (htime_file + strlen(htime_file) - TIMESTAMP_LENGTH);
+ sscanf(iter, "%lu", min_ts);
+
+ ret = sys_stat(htime_file, &stbuf);
+ if (ret) {
+ ret = -1;
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_LIB_MSG_HTIME_ERROR,
+ "op=stat", "path=%s", htime_file, NULL);
+ goto out;
+ }
+
+ /* ignore everything except regular files */
+ if (!S_ISREG(stbuf.st_mode)) {
+ ret = -2;
+ goto out;
+ }
+
+ *fd = open(htime_file, O_RDONLY);
+ if (*fd < 0) {
+ ret = -1;
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_LIB_MSG_HTIME_ERROR,
+ "op=open", "path=%s", htime_file, NULL);
+ goto out;
+ }
+
+ /* Looks good, extract max timestamp */
+ ret = sys_fgetxattr(*fd, HTIME_KEY, x_value, sizeof(x_value));
+ if (ret < 0) {
+ ret = -1;
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_LIB_MSG_GET_XATTR_FAILED, "path=%s", htime_file,
+ NULL);
+ goto out;
+ }
+
+ sscanf(x_value, "%lu:%lu", max_ts, total);
+ gf_smsg(this->name, GF_LOG_INFO, 0, CHANGELOG_LIB_MSG_MIN_MAX_INFO,
+ "min=%lu", *min_ts, "max=%lu", *max_ts, "total_changelogs=%lu",
+ *total, NULL);
+
+ ret = 0;
- iter = (htime_file + strlen (htime_file) - TIMESTAMP_LENGTH);
- sscanf (iter ,"%lu",min_ts);
+out:
+ return ret;
+}
- ret = sys_stat (htime_file, &stbuf);
+/* gf_history_changelog returns actual_end and spawns threads to
+ * parse historical changelogs. The return values are as follows.
+ * 0 : On success
+ * 1 : Successful, but partial historical changelogs available,
+ * end time falls into different htime file or future time
+ * -2 : Error, requested historical changelog not available, not
+ * even partial
+ * -1 : On any error
+ */
+int
+gf_history_changelog(char *changelog_dir, unsigned long start,
+ unsigned long end, int n_parallel,
+ unsigned long *actual_end)
+{
+ int ret = 0;
+ int len = -1;
+ int fd = -1;
+ int n_read = -1;
+ unsigned long min_ts = 0;
+ unsigned long max_ts = 0;
+ unsigned long end2 = 0;
+ unsigned long ts1 = 0;
+ unsigned long ts2 = 0;
+ unsigned long to = 0;
+ unsigned long from = 0;
+ unsigned long total_changelog = 0;
+ xlator_t *this = NULL;
+ gf_changelog_journal_t *jnl = NULL;
+ gf_changelog_journal_t *hist_jnl = NULL;
+ gf_changelog_history_data_t *hist_data = NULL;
+ DIR *dirp = NULL;
+ struct dirent *entry = NULL;
+ struct dirent scratch[2] = {
+ {
+ 0,
+ },
+ };
+ pthread_t consume_th = 0;
+ char htime_dir[PATH_MAX] = {
+ 0,
+ };
+ char buffer[PATH_MAX] = {
+ 0,
+ };
+ gf_boolean_t partial_history = _gf_false;
+
+ pthread_attr_t attr;
+
+ this = THIS;
+ if (!this) {
+ ret = -1;
+ goto out;
+ }
+
+ ret = pthread_attr_init(&attr);
+ if (ret != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, errno, CHANGELOG_LIB_MSG_PTHREAD_ERROR,
+ "Pthread init failed");
+ return -1;
+ }
+
+ jnl = (gf_changelog_journal_t *)GF_CHANGELOG_GET_API_PTR(this);
+ if (!jnl) {
+ ret = -1;
+ goto out;
+ }
+
+ hist_jnl = (gf_changelog_journal_t *)jnl->hist_jnl;
+ if (!hist_jnl) {
+ ret = -1;
+ goto out;
+ }
+
+ gf_smsg(this->name, GF_LOG_INFO, 0, CHANGELOG_LIB_MSG_REQUESTING_INFO,
+ "start=%lu", start, "end=%lu", end, NULL);
+
+ /* basic sanity check */
+ if (start > end || n_parallel <= 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_LIB_MSG_HIST_FAILED,
+ "start=%lu", start, "end=%lu", end, "thread_count=%d",
+ n_parallel, NULL);
+ ret = -1;
+ goto out;
+ }
+
+ /* cap parallelism count */
+ if (n_parallel > MAX_PARALLELS)
+ n_parallel = MAX_PARALLELS;
+
+ CHANGELOG_FILL_HTIME_DIR(changelog_dir, htime_dir);
+
+ dirp = sys_opendir(htime_dir);
+ if (dirp == NULL) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_LIB_MSG_HTIME_ERROR,
+ "op=opendir", "path=%s", htime_dir, NULL);
+ ret = -1;
+ goto out;
+ }
+
+ for (;;) {
+ errno = 0;
+
+ entry = sys_readdir(dirp, scratch);
+
+ if (!entry || errno != 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_LIB_MSG_HIST_FAILED, "start=%lu", start,
+ "end=%lu", end, NULL);
+ ret = -2;
+ break;
+ }
+
+ ret = gf_changelog_extract_min_max(entry->d_name, htime_dir, &fd,
+ &total_changelog, &min_ts, &max_ts);
if (ret) {
- ret = -1;
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_HTIME_ERROR,
- "stat() failed on htime file",
- "path=%s", htime_file,
- NULL);
- goto out;
- }
-
- /* ignore everything except regular files */
- if (!S_ISREG (stbuf.st_mode)) {
- ret = -2;
- goto out;
+ if (-2 == ret)
+ continue;
+ goto out;
}
- *fd = open (htime_file, O_RDONLY);
- if (*fd < 0) {
+ if (start >= min_ts && start < max_ts) {
+ /**
+ * TODO: handle short reads later...
+ */
+ n_read = sys_read(fd, buffer, PATH_MAX);
+ if (n_read < 0) {
ret = -1;
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_HTIME_ERROR,
- "open() failed for htime file",
- "path=%s", htime_file,
- NULL);
+ gf_msg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_LIB_MSG_READ_ERROR,
+ "unable to read htime file");
goto out;
- }
+ }
- /* Looks good, extract max timestamp */
- ret = sys_fgetxattr (*fd, HTIME_KEY, x_value, sizeof (x_value));
- if (ret < 0) {
- ret = -1;
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_GET_XATTR_FAILED,
- "error extracting max timstamp from htime file"
- "path=%s", htime_file,
- NULL);
- goto out;
- }
-
- sscanf (x_value, "%lu:%lu", max_ts, total);
- gf_smsg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_LIB_MSG_TOTAL_LOG_INFO,
- "changelogs min max",
- "min=%lu", *min_ts,
- "max=%lu", *max_ts,
- "total_changelogs=%lu", *total,
- NULL);
+ len = strlen(buffer);
- ret = 0;
+ /**
+ * search @start in the htime file returning it's index
+ * (@from)
+ */
+ from = gf_history_b_search(fd, start, 0, total_changelog - 1, len);
- out:
- return ret;
-}
-
-int
-gf_history_changelog (char* changelog_dir, unsigned long start,
- unsigned long end, int n_parallel,
- unsigned long *actual_end)
-{
- int ret = 0;
- int len = -1;
- int fd = -1;
- int n_read = -1;
- unsigned long min_ts = 0;
- unsigned long max_ts = 0;
- unsigned long end2 = 0;
- unsigned long ts1 = 0;
- unsigned long ts2 = 0;
- unsigned long to = 0;
- unsigned long from = 0;
- unsigned long total_changelog = 0;
- xlator_t *this = NULL;
- gf_changelog_journal_t *jnl = NULL;
- gf_changelog_journal_t *hist_jnl = NULL;
- gf_changelog_history_data_t *hist_data = NULL;
- DIR *dirp = NULL;
- struct dirent *entry = NULL;
- struct dirent scratch[2] = {{0,},};
- pthread_t consume_th = 0;
- char htime_dir[PATH_MAX] = {0,};
- char buffer[PATH_MAX] = {0,};
-
- pthread_attr_t attr;
-
- this = THIS;
- if (!this) {
+ /* ensuring correctness of gf_b_search */
+ if (gf_history_check(fd, from, start, len) != 0) {
+ ret = -1;
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_LIB_MSG_GET_TIME_ERROR, "for=start",
+ "start=%lu", start, "idx=%lu", from, NULL);
+ goto out;
+ }
+
+ end2 = (end <= max_ts) ? end : max_ts;
+
+ /* Check if end falls out of same HTIME file. The end
+ * falling to a different htime file or changelog
+ * disable-enable is detected only after 20 seconds.
+ * This is required because, applications generally
+ * asks historical changelogs till current time and
+ * it is possible changelog is not rolled over yet.
+ * So, buffer time of default rollover time plus 5
+ * seconds is subtracted. If the application requests
+ * the end time with in half a minute of changelog
+ * disable, it's not detected as changelog disable and
+ * it's application's responsibility to retry after
+ * 20 seconds before confirming it as partial history.
+ */
+ if ((end - 20) > max_ts) {
+ partial_history = _gf_true;
+ }
+
+ /**
+ * search @end2 in htime file returning it's index (@to)
+ */
+ to = gf_history_b_search(fd, end2, 0, total_changelog - 1, len);
+
+ if (gf_history_check(fd, to, end2, len) != 0) {
ret = -1;
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_LIB_MSG_GET_TIME_ERROR, "for=end",
+ "start=%lu", end2, "idx=%lu", to, NULL);
goto out;
- }
+ }
- ret = pthread_attr_init (&attr);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_PTHREAD_ERROR,
- "Pthread init failed");
- return -1;
- }
+ ret = gf_history_get_timestamp(fd, from, len, &ts1);
+ if (ret == -1)
+ goto out;
- jnl = (gf_changelog_journal_t *) GF_CHANGELOG_GET_API_PTR (this);
- if (!jnl) {
- ret = -1;
+ ret = gf_history_get_timestamp(fd, to, len, &ts2);
+ if (ret == -1)
goto out;
- }
- hist_jnl = (gf_changelog_journal_t *) jnl->hist_jnl;
- if (!hist_jnl) {
+ gf_smsg(this->name, GF_LOG_INFO, 0, CHANGELOG_LIB_MSG_FINAL_INFO,
+ "from=%lu", ts1, "to=%lu", ts2, "changes=%lu",
+ (to - from + 1), NULL);
+
+ hist_data = GF_CALLOC(1, sizeof(gf_changelog_history_data_t),
+ gf_changelog_mt_history_data_t);
+
+ hist_data->htime_fd = fd;
+ hist_data->from = from;
+ hist_data->to = to;
+ hist_data->len = len;
+ hist_data->n_parallel = n_parallel;
+ hist_data->this = this;
+
+ ret = pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED);
+ if (ret != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, ret,
+ CHANGELOG_LIB_MSG_PTHREAD_ERROR,
+ "unable to sets the detach"
+ " state attribute");
ret = -1;
goto out;
- }
+ }
- /* basic sanity check */
- if (start > end || n_parallel <= 0) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_HIST_FAILED, "Sanity check failed",
- "start=%lu", start,
- "end=%lu", end,
- "thread_count=%d", n_parallel,
- NULL);
+ /* spawn a thread for background parsing & publishing */
+ ret = gf_thread_create(&consume_th, &attr, gf_history_consume,
+ hist_data, "cloghcon");
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, ret,
+ CHANGELOG_LIB_MSG_THREAD_CREATION_FAILED,
+ "creation of consume parent-thread"
+ " failed.");
ret = -1;
goto out;
- }
+ }
- /* cap parallelism count */
- if (n_parallel > MAX_PARALLELS)
- n_parallel = MAX_PARALLELS;
+ goto out;
- CHANGELOG_FILL_HTIME_DIR (changelog_dir, htime_dir);
-
- dirp = sys_opendir (htime_dir);
- if (dirp == NULL) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_HTIME_ERROR,
- "open dir on htime failed",
- "path=%s", htime_dir,
- NULL);
- ret = -1;
- goto out;
+ } else { /* end of range check */
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_LIB_MSG_HIST_FAILED, "start=%lu", start,
+ "end=%lu", end, "chlog_min=%lu", min_ts, "chlog_max=%lu",
+ max_ts, NULL);
}
-
- for (;;) {
-
- errno = 0;
-
- entry = sys_readdir (dirp, scratch);
-
- if (!entry || errno != 0)
- break;
-
- ret = gf_changelog_extract_min_max (entry->d_name, htime_dir,
- &fd, &total_changelog,
- &min_ts, &max_ts);
- if (ret) {
- if (-2 == ret)
- continue;
- goto out;
- }
-
- if (start >= min_ts && start < max_ts) {
- /**
- * TODO: handle short reads later...
- */
- n_read = sys_read (fd, buffer, PATH_MAX);
- if (n_read < 0) {
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_READ_ERROR,
- "unable to read htime file");
- goto out;
- }
-
- len = strlen (buffer);
-
- /**
- * search @start in the htime file returning it's index
- * (@from)
- */
- from = gf_history_b_search (fd, start, 0,
- total_changelog - 1, len);
-
- /* ensuring correctness of gf_b_search */
- if (gf_history_check (fd, from, start, len) != 0) {
- ret = -1;
- gf_smsg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_LIB_MSG_GET_TIME_ERROR,
- "wrong result for start",
- "start=%lu", start,
- "idx=%lu", from,
- NULL);
- goto out;
- }
-
- end2 = (end <= max_ts) ? end : max_ts;
-
- /**
- * search @end2 in htime file returning it's index (@to)
- */
- to = gf_history_b_search (fd, end2,
- 0, total_changelog - 1, len);
-
- if (gf_history_check (fd, to, end2, len) != 0) {
- ret = -1;
- gf_smsg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_LIB_MSG_GET_TIME_ERROR,
- "wrong result for end",
- "start=%lu", end2,
- "idx=%lu", to,
- NULL);
- goto out;
- }
-
- ret = gf_history_get_timestamp (fd, from, len, &ts1);
- if (ret == -1)
- goto out;
-
- ret = gf_history_get_timestamp (fd, to, len, &ts2);
- if (ret == -1)
- goto out;
-
- gf_smsg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_LIB_MSG_TOTAL_LOG_INFO,
- "FINAL",
- "from=%lu", ts1,
- "to=%lu", ts2,
- "changes=%lu", (to - from + 1),
- NULL);
-
- hist_data = GF_CALLOC (1,
- sizeof (gf_changelog_history_data_t),
- gf_changelog_mt_history_data_t);
-
- hist_data->htime_fd = fd;
- hist_data->from = from;
- hist_data->to = to;
- hist_data->len = len;
- hist_data->n_parallel = n_parallel;
- hist_data->this = this;
-
- ret = pthread_attr_setdetachstate
- (&attr, PTHREAD_CREATE_DETACHED);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ret,
- CHANGELOG_LIB_MSG_PTHREAD_ERROR,
- "unable to sets the detach"
- " state attribute");
- ret = -1;
- goto out;
- }
-
- /* spawn a thread for background parsing & publishing */
- ret = gf_thread_create (&consume_th, &attr,
- gf_history_consume, hist_data,
- "cloghcon");
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, ret,
- CHANGELOG_LIB_MSG_THREAD_CREATION_FAILED
- , "creation of consume parent-thread"
- " failed.");
- ret = -1;
- goto out;
- }
-
- goto out;
-
- } else {/* end of range check */
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_LIB_MSG_HIST_FAILED,
- "Requested changelog "
- "range is not available.",
- "start=%lu", start,
- "chlog_min=%lu", min_ts,
- "chlog_max=%lu", max_ts,
- NULL);
- ret = -2;
- goto out;
- }
- } /* end of readdir() */
+ } /* end of readdir() */
out:
- if (dirp != NULL)
- (void) sys_closedir (dirp);
+ if (dirp != NULL)
+ (void)sys_closedir(dirp);
- if (ret < 0) {
- if (fd != -1)
- (void) sys_close (fd);
- GF_FREE (hist_data);
- (void) pthread_attr_destroy (&attr);
+ if (ret < 0) {
+ if (fd != -1)
+ (void)sys_close(fd);
+ GF_FREE(hist_data);
+ (void)pthread_attr_destroy(&attr);
- return ret;
- }
+ return ret;
+ }
- hist_jnl->hist_done = 1;
- *actual_end = ts2;
+ hist_jnl->hist_done = 1;
+ *actual_end = ts2;
- return ret;
+ if (partial_history) {
+ ret = 1;
+ }
+
+ return ret;
}
diff --git a/xlators/features/changelog/src/changelog-barrier.c b/xlators/features/changelog/src/changelog-barrier.c
index ac1eb0e4397..0fb89ddb127 100644
--- a/xlators/features/changelog/src/changelog-barrier.c
+++ b/xlators/features/changelog/src/changelog-barrier.c
@@ -10,125 +10,122 @@
#include "changelog-helpers.h"
#include "changelog-messages.h"
-#include "call-stub.h"
+#include <glusterfs/call-stub.h>
/* Enqueue a stub*/
void
-__chlog_barrier_enqueue (xlator_t *this, call_stub_t *stub)
+__chlog_barrier_enqueue(xlator_t *this, call_stub_t *stub)
{
- changelog_priv_t *priv = NULL;
+ changelog_priv_t *priv = NULL;
- priv = this->private;
- GF_ASSERT (priv);
+ priv = this->private;
+ GF_ASSERT(priv);
- list_add_tail (&stub->list, &priv->queue);
- priv->queue_size++;
+ list_add_tail(&stub->list, &priv->queue);
+ priv->queue_size++;
- return;
+ return;
}
/* Dequeue a stub */
call_stub_t *
-__chlog_barrier_dequeue (xlator_t *this, struct list_head *queue)
+__chlog_barrier_dequeue(xlator_t *this, struct list_head *queue)
{
- call_stub_t *stub = NULL;
- changelog_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ changelog_priv_t *priv = NULL;
- priv = this->private;
- GF_ASSERT (priv);
+ priv = this->private;
+ GF_ASSERT(priv);
- if (list_empty (queue))
- goto out;
+ if (list_empty(queue))
+ goto out;
- stub = list_entry (queue->next, call_stub_t, list);
- list_del_init (&stub->list);
+ stub = list_entry(queue->next, call_stub_t, list);
+ list_del_init(&stub->list);
out:
- return stub;
+ return stub;
}
/* Dequeue all the stubs and call corresponding resume functions */
void
-chlog_barrier_dequeue_all (xlator_t *this, struct list_head *queue)
+chlog_barrier_dequeue_all(xlator_t *this, struct list_head *queue)
{
- call_stub_t *stub = NULL;
+ call_stub_t *stub = NULL;
- gf_msg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_MSG_BARRIER_INFO,
- "Dequeuing all the changelog barriered fops");
+ gf_smsg(this->name, GF_LOG_INFO, 0, CHANGELOG_MSG_DEQUEUING_BARRIER_FOPS,
+ NULL);
- while ((stub = __chlog_barrier_dequeue (this, queue)))
- call_resume (stub);
+ while ((stub = __chlog_barrier_dequeue(this, queue)))
+ call_resume(stub);
- gf_msg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_MSG_BARRIER_INFO,
- "Dequeuing changelog barriered fops is finished");
- return;
+ gf_smsg(this->name, GF_LOG_INFO, 0,
+ CHANGELOG_MSG_DEQUEUING_BARRIER_FOPS_FINISHED, NULL);
+ return;
}
/* Function called on changelog barrier timeout */
void
-chlog_barrier_timeout (void *data)
+chlog_barrier_timeout(void *data)
{
- xlator_t *this = NULL;
- changelog_priv_t *priv = NULL;
- struct list_head queue = {0,};
+ xlator_t *this = NULL;
+ changelog_priv_t *priv = NULL;
+ struct list_head queue = {
+ 0,
+ };
- this = data;
- THIS = this;
- priv = this->private;
+ this = data;
+ THIS = this;
+ priv = this->private;
- INIT_LIST_HEAD (&queue);
+ INIT_LIST_HEAD(&queue);
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_BARRIER_ERROR,
- "Disabling changelog barrier because of the timeout.");
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_BARRIER_TIMEOUT, NULL);
- LOCK (&priv->lock);
- {
- __chlog_barrier_disable (this, &queue);
- }
- UNLOCK (&priv->lock);
+ LOCK(&priv->lock);
+ {
+ __chlog_barrier_disable(this, &queue);
+ }
+ UNLOCK(&priv->lock);
- chlog_barrier_dequeue_all (this, &queue);
+ chlog_barrier_dequeue_all(this, &queue);
- return;
+ return;
}
/* Disable changelog barrier enable flag */
void
-__chlog_barrier_disable (xlator_t *this, struct list_head *queue)
+__chlog_barrier_disable(xlator_t *this, struct list_head *queue)
{
- changelog_priv_t *priv = this->private;
- GF_ASSERT (priv);
+ changelog_priv_t *priv = this->private;
+ GF_ASSERT(priv);
- if (priv->timer) {
- gf_timer_call_cancel (this->ctx, priv->timer);
- priv->timer = NULL;
- }
+ if (priv->timer) {
+ gf_timer_call_cancel(this->ctx, priv->timer);
+ priv->timer = NULL;
+ }
- list_splice_init (&priv->queue, queue);
- priv->queue_size = 0;
- priv->barrier_enabled = _gf_false;
+ list_splice_init(&priv->queue, queue);
+ priv->queue_size = 0;
+ priv->barrier_enabled = _gf_false;
}
/* Enable chagelog barrier enable with timer */
int
-__chlog_barrier_enable (xlator_t *this, changelog_priv_t *priv)
+__chlog_barrier_enable(xlator_t *this, changelog_priv_t *priv)
{
- int ret = -1;
-
- priv->timer = gf_timer_call_after (this->ctx, priv->timeout,
- chlog_barrier_timeout, (void *)this);
- if (!priv->timer) {
- gf_msg (this->name, GF_LOG_CRITICAL, 0,
- CHANGELOG_MSG_BARRIER_ERROR,
- "Couldn't add changelog barrier timeout event.");
- goto out;
- }
-
- priv->barrier_enabled = _gf_true;
- ret = 0;
+ int ret = -1;
+
+ priv->timer = gf_timer_call_after(this->ctx, priv->timeout,
+ chlog_barrier_timeout, (void *)this);
+ if (!priv->timer) {
+ gf_smsg(this->name, GF_LOG_CRITICAL, 0,
+ CHANGELOG_MSG_TIMEOUT_ADD_FAILED, NULL);
+ goto out;
+ }
+
+ priv->barrier_enabled = _gf_true;
+ ret = 0;
out:
- return ret;
+ return ret;
}
diff --git a/xlators/features/changelog/src/changelog-encoders.c b/xlators/features/changelog/src/changelog-encoders.c
index 95030236636..63754516c2e 100644
--- a/xlators/features/changelog/src/changelog-encoders.c
+++ b/xlators/features/changelog/src/changelog-encoders.c
@@ -11,117 +11,117 @@
#include "changelog-encoders.h"
size_t
-entry_fn (void *data, char *buffer, gf_boolean_t encode)
+entry_fn(void *data, char *buffer, gf_boolean_t encode)
{
- char *tmpbuf = NULL;
- size_t bufsz = 0;
- struct changelog_entry_fields *ce = NULL;
-
- ce = (struct changelog_entry_fields *) data;
-
- if (encode) {
- tmpbuf = uuid_utoa (ce->cef_uuid);
- CHANGELOG_FILL_BUFFER (buffer, bufsz, tmpbuf, strlen (tmpbuf));
- } else {
- CHANGELOG_FILL_BUFFER (buffer, bufsz,
- ce->cef_uuid, sizeof (uuid_t));
- }
-
- CHANGELOG_FILL_BUFFER (buffer, bufsz, "/", 1);
- CHANGELOG_FILL_BUFFER (buffer, bufsz,
- ce->cef_bname, strlen (ce->cef_bname));
- return bufsz;
+ char *tmpbuf = NULL;
+ size_t bufsz = 0;
+ struct changelog_entry_fields *ce = NULL;
+
+ ce = (struct changelog_entry_fields *)data;
+
+ if (encode) {
+ tmpbuf = uuid_utoa(ce->cef_uuid);
+ CHANGELOG_FILL_BUFFER(buffer, bufsz, tmpbuf, strlen(tmpbuf));
+ } else {
+ CHANGELOG_FILL_BUFFER(buffer, bufsz, ce->cef_uuid, sizeof(uuid_t));
+ }
+
+ CHANGELOG_FILL_BUFFER(buffer, bufsz, "/", 1);
+ CHANGELOG_FILL_BUFFER(buffer, bufsz, ce->cef_bname, strlen(ce->cef_bname));
+ return bufsz;
}
size_t
-del_entry_fn (void *data, char *buffer, gf_boolean_t encode)
+del_entry_fn(void *data, char *buffer, gf_boolean_t encode)
{
- char *tmpbuf = NULL;
- size_t bufsz = 0;
- struct changelog_entry_fields *ce = NULL;
-
- ce = (struct changelog_entry_fields *) data;
-
- if (encode) {
- tmpbuf = uuid_utoa (ce->cef_uuid);
- CHANGELOG_FILL_BUFFER (buffer, bufsz, tmpbuf, strlen (tmpbuf));
- } else {
- CHANGELOG_FILL_BUFFER (buffer, bufsz,
- ce->cef_uuid, sizeof (uuid_t));
- }
-
- CHANGELOG_FILL_BUFFER (buffer, bufsz, "/", 1);
- CHANGELOG_FILL_BUFFER (buffer, bufsz,
- ce->cef_bname, strlen (ce->cef_bname));
- CHANGELOG_FILL_BUFFER (buffer, bufsz, "\0", 1);
-
- if (ce->cef_path[0] == '\0') {
- CHANGELOG_FILL_BUFFER (buffer, bufsz, "\0", 1);
- } else {
- CHANGELOG_FILL_BUFFER (buffer, bufsz,
- ce->cef_path, strlen (ce->cef_path));
- }
-
- return bufsz;
+ char *tmpbuf = NULL;
+ size_t bufsz = 0;
+ struct changelog_entry_fields *ce = NULL;
+
+ ce = (struct changelog_entry_fields *)data;
+
+ if (encode) {
+ tmpbuf = uuid_utoa(ce->cef_uuid);
+ CHANGELOG_FILL_BUFFER(buffer, bufsz, tmpbuf, strlen(tmpbuf));
+ } else {
+ CHANGELOG_FILL_BUFFER(buffer, bufsz, ce->cef_uuid, sizeof(uuid_t));
+ }
+
+ CHANGELOG_FILL_BUFFER(buffer, bufsz, "/", 1);
+ CHANGELOG_FILL_BUFFER(buffer, bufsz, ce->cef_bname, strlen(ce->cef_bname));
+ CHANGELOG_FILL_BUFFER(buffer, bufsz, "\0", 1);
+
+ if (ce->cef_path[0] == '\0') {
+ CHANGELOG_FILL_BUFFER(buffer, bufsz, "\0", 1);
+ } else {
+ CHANGELOG_FILL_BUFFER(buffer, bufsz, ce->cef_path,
+ strlen(ce->cef_path));
+ }
+
+ return bufsz;
}
size_t
-fop_fn (void *data, char *buffer, gf_boolean_t encode)
+fop_fn(void *data, char *buffer, gf_boolean_t encode)
{
- char buf[10] = {0,};
- size_t bufsz = 0;
- glusterfs_fop_t fop = 0;
+ char buf[10] = {
+ 0,
+ };
+ size_t bufsz = 0;
+ glusterfs_fop_t fop = 0;
- fop = *(glusterfs_fop_t *) data;
+ fop = *(glusterfs_fop_t *)data;
- if (encode) {
- (void) snprintf (buf, sizeof (buf), "%d", fop);
- CHANGELOG_FILL_BUFFER (buffer, bufsz, buf, strlen (buf));
- } else
- CHANGELOG_FILL_BUFFER (buffer, bufsz, &fop, sizeof (fop));
+ if (encode) {
+ (void)snprintf(buf, sizeof(buf), "%d", fop);
+ CHANGELOG_FILL_BUFFER(buffer, bufsz, buf, strlen(buf));
+ } else
+ CHANGELOG_FILL_BUFFER(buffer, bufsz, &fop, sizeof(fop));
- return bufsz;
+ return bufsz;
}
size_t
-number_fn (void *data, char *buffer, gf_boolean_t encode)
+number_fn(void *data, char *buffer, gf_boolean_t encode)
{
- size_t bufsz = 0;
- unsigned int nr = 0;
- char buf[20] = {0,};
+ size_t bufsz = 0;
+ unsigned int nr = 0;
+ char buf[20] = {
+ 0,
+ };
- nr = *(unsigned int *) data;
+ nr = *(unsigned int *)data;
- if (encode) {
- (void) snprintf (buf, sizeof (buf), "%u", nr);
- CHANGELOG_FILL_BUFFER (buffer, bufsz, buf, strlen (buf));
- } else
- CHANGELOG_FILL_BUFFER (buffer, bufsz, &nr, sizeof (unsigned int));
+ if (encode) {
+ (void)snprintf(buf, sizeof(buf), "%u", nr);
+ CHANGELOG_FILL_BUFFER(buffer, bufsz, buf, strlen(buf));
+ } else
+ CHANGELOG_FILL_BUFFER(buffer, bufsz, &nr, sizeof(unsigned int));
- return bufsz;
+ return bufsz;
}
void
-entry_free_fn (void *data)
+entry_free_fn(void *data)
{
- changelog_opt_t *co = data;
+ changelog_opt_t *co = data;
- if (!co)
- return;
+ if (!co)
+ return;
- GF_FREE (co->co_entry.cef_bname);
+ GF_FREE(co->co_entry.cef_bname);
}
void
-del_entry_free_fn (void *data)
+del_entry_free_fn(void *data)
{
- changelog_opt_t *co = data;
+ changelog_opt_t *co = data;
- if (!co)
- return;
+ if (!co)
+ return;
- GF_FREE (co->co_entry.cef_bname);
- GF_FREE (co->co_entry.cef_path);
+ GF_FREE(co->co_entry.cef_bname);
+ GF_FREE(co->co_entry.cef_path);
}
/**
@@ -129,108 +129,104 @@ del_entry_free_fn (void *data)
*/
static void
-changelog_encode_write_xtra (changelog_log_data_t *cld,
- char *buffer, size_t *off, gf_boolean_t encode)
+changelog_encode_write_xtra(changelog_log_data_t *cld, char *buffer,
+ size_t *off, gf_boolean_t encode)
{
- int i = 0;
- size_t offset = 0;
- void *data = NULL;
- changelog_opt_t *co = NULL;
-
- offset = *off;
-
- co = (changelog_opt_t *) cld->cld_ptr;
-
- for (; i < cld->cld_xtra_records; i++, co++) {
- CHANGELOG_FILL_BUFFER (buffer, offset, "\0", 1);
-
- switch (co->co_type) {
- case CHANGELOG_OPT_REC_FOP:
- data = &co->co_fop;
- break;
- case CHANGELOG_OPT_REC_ENTRY:
- data = &co->co_entry;
- break;
- case CHANGELOG_OPT_REC_UINT32:
- data = &co->co_uint32;
- break;
- }
-
- if (co->co_convert)
- offset += co->co_convert (data,
- buffer + offset, encode);
- else /* no coversion: write it out as it is */
- CHANGELOG_FILL_BUFFER (buffer, offset,
- data, co->co_len);
+ int i = 0;
+ size_t offset = 0;
+ void *data = NULL;
+ changelog_opt_t *co = NULL;
+
+ offset = *off;
+
+ co = (changelog_opt_t *)cld->cld_ptr;
+
+ for (; i < cld->cld_xtra_records; i++, co++) {
+ CHANGELOG_FILL_BUFFER(buffer, offset, "\0", 1);
+
+ switch (co->co_type) {
+ case CHANGELOG_OPT_REC_FOP:
+ data = &co->co_fop;
+ break;
+ case CHANGELOG_OPT_REC_ENTRY:
+ data = &co->co_entry;
+ break;
+ case CHANGELOG_OPT_REC_UINT32:
+ data = &co->co_uint32;
+ break;
}
- *off = offset;
+ if (co->co_convert)
+ offset += co->co_convert(data, buffer + offset, encode);
+ else /* no coversion: write it out as it is */
+ CHANGELOG_FILL_BUFFER(buffer, offset, data, co->co_len);
+ }
+
+ *off = offset;
}
int
-changelog_encode_ascii (xlator_t *this, changelog_log_data_t *cld)
+changelog_encode_ascii(xlator_t *this, changelog_log_data_t *cld)
{
- size_t off = 0;
- size_t gfid_len = 0;
- char *gfid_str = NULL;
- char *buffer = NULL;
- changelog_priv_t *priv = NULL;
+ size_t off = 0;
+ size_t gfid_len = 0;
+ char *gfid_str = NULL;
+ char *buffer = NULL;
+ changelog_priv_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- gfid_str = uuid_utoa (cld->cld_gfid);
- gfid_len = strlen (gfid_str);
+ gfid_str = uuid_utoa(cld->cld_gfid);
+ gfid_len = strlen(gfid_str);
- /* extra bytes for decorations */
- buffer = alloca (gfid_len + cld->cld_ptr_len + 10);
- CHANGELOG_STORE_ASCII (priv, buffer,
- off, gfid_str, gfid_len, cld);
+ /* extra bytes for decorations */
+ buffer = alloca(gfid_len + cld->cld_ptr_len + 10);
+ CHANGELOG_STORE_ASCII(priv, buffer, off, gfid_str, gfid_len, cld);
- if (cld->cld_xtra_records)
- changelog_encode_write_xtra (cld, buffer, &off, _gf_true);
+ if (cld->cld_xtra_records)
+ changelog_encode_write_xtra(cld, buffer, &off, _gf_true);
- CHANGELOG_FILL_BUFFER (buffer, off, "\0", 1);
+ CHANGELOG_FILL_BUFFER(buffer, off, "\0", 1);
- return changelog_write_change (priv, buffer, off);
+ return changelog_write_change(priv, buffer, off);
}
int
-changelog_encode_binary (xlator_t *this, changelog_log_data_t *cld)
+changelog_encode_binary(xlator_t *this, changelog_log_data_t *cld)
{
- size_t off = 0;
- char *buffer = NULL;
- changelog_priv_t *priv = NULL;
+ size_t off = 0;
+ char *buffer = NULL;
+ changelog_priv_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- /* extra bytes for decorations */
- buffer = alloca (sizeof (uuid_t) + cld->cld_ptr_len + 10);
- CHANGELOG_STORE_BINARY (priv, buffer, off, cld->cld_gfid, cld);
+ /* extra bytes for decorations */
+ buffer = alloca(sizeof(uuid_t) + cld->cld_ptr_len + 10);
+ CHANGELOG_STORE_BINARY(priv, buffer, off, cld->cld_gfid, cld);
- if (cld->cld_xtra_records)
- changelog_encode_write_xtra (cld, buffer, &off, _gf_false);
+ if (cld->cld_xtra_records)
+ changelog_encode_write_xtra(cld, buffer, &off, _gf_false);
- CHANGELOG_FILL_BUFFER (buffer, off, "\0", 1);
+ CHANGELOG_FILL_BUFFER(buffer, off, "\0", 1);
- return changelog_write_change (priv, buffer, off);
+ return changelog_write_change(priv, buffer, off);
}
-static struct changelog_encoder
-cb_encoder[] = {
- [CHANGELOG_ENCODE_BINARY] =
+static struct changelog_encoder cb_encoder[] = {
+ [CHANGELOG_ENCODE_BINARY] =
{
- .encoder = CHANGELOG_ENCODE_BINARY,
- .encode = changelog_encode_binary,
+ .encoder = CHANGELOG_ENCODE_BINARY,
+ .encode = changelog_encode_binary,
},
- [CHANGELOG_ENCODE_ASCII] =
+ [CHANGELOG_ENCODE_ASCII] =
{
- .encoder = CHANGELOG_ENCODE_ASCII,
- .encode = changelog_encode_ascii,
+ .encoder = CHANGELOG_ENCODE_ASCII,
+ .encode = changelog_encode_ascii,
},
};
void
changelog_encode_change(changelog_priv_t *priv)
{
- priv->ce = &cb_encoder[priv->encode_mode];
+ priv->ce = &cb_encoder[priv->encode_mode];
}
diff --git a/xlators/features/changelog/src/changelog-encoders.h b/xlators/features/changelog/src/changelog-encoders.h
index d6a50cc9ef7..26252696d56 100644
--- a/xlators/features/changelog/src/changelog-encoders.h
+++ b/xlators/features/changelog/src/changelog-encoders.h
@@ -11,41 +11,39 @@
#ifndef _CHANGELOG_ENCODERS_H
#define _CHANGELOG_ENCODERS_H
-#include "xlator.h"
-#include "defaults.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
#include "changelog-helpers.h"
-#define CHANGELOG_STORE_ASCII(priv, buf, off, gfid, gfid_len, cld) do { \
- CHANGELOG_FILL_BUFFER (buffer, off, \
- priv->maps[cld->cld_type], 1); \
- CHANGELOG_FILL_BUFFER (buffer, \
- off, gfid, gfid_len); \
- } while (0)
+#define CHANGELOG_STORE_ASCII(priv, buf, off, gfid, gfid_len, cld) \
+ do { \
+ CHANGELOG_FILL_BUFFER(buffer, off, priv->maps[cld->cld_type], 1); \
+ CHANGELOG_FILL_BUFFER(buffer, off, gfid, gfid_len); \
+ } while (0)
-#define CHANGELOG_STORE_BINARY(priv, buf, off, gfid, cld) do { \
- CHANGELOG_FILL_BUFFER (buffer, off, \
- priv->maps[cld->cld_type], 1); \
- CHANGELOG_FILL_BUFFER (buffer, \
- off, gfid, sizeof (uuid_t)); \
- } while (0)
+#define CHANGELOG_STORE_BINARY(priv, buf, off, gfid, cld) \
+ do { \
+ CHANGELOG_FILL_BUFFER(buffer, off, priv->maps[cld->cld_type], 1); \
+ CHANGELOG_FILL_BUFFER(buffer, off, gfid, sizeof(uuid_t)); \
+ } while (0)
size_t
-entry_fn (void *data, char *buffer, gf_boolean_t encode);
+entry_fn(void *data, char *buffer, gf_boolean_t encode);
size_t
-del_entry_fn (void *data, char *buffer, gf_boolean_t encode);
+del_entry_fn(void *data, char *buffer, gf_boolean_t encode);
size_t
-fop_fn (void *data, char *buffer, gf_boolean_t encode);
+fop_fn(void *data, char *buffer, gf_boolean_t encode);
size_t
-number_fn (void *data, char *buffer, gf_boolean_t encode);
+number_fn(void *data, char *buffer, gf_boolean_t encode);
void
-entry_free_fn (void *data);
+entry_free_fn(void *data);
void
-del_entry_free_fn (void *data);
+del_entry_free_fn(void *data);
int
-changelog_encode_binary (xlator_t *, changelog_log_data_t *);
+changelog_encode_binary(xlator_t *, changelog_log_data_t *);
int
-changelog_encode_ascii (xlator_t *, changelog_log_data_t *);
+changelog_encode_ascii(xlator_t *, changelog_log_data_t *);
void
changelog_encode_change(changelog_priv_t *);
diff --git a/xlators/features/changelog/src/changelog-ev-handle.c b/xlators/features/changelog/src/changelog-ev-handle.c
index a7f485b019f..aa94459de5a 100644
--- a/xlators/features/changelog/src/changelog-ev-handle.c
+++ b/xlators/features/changelog/src/changelog-ev-handle.c
@@ -14,19 +14,19 @@
struct rpc_clnt_program changelog_ev_program;
-#define NR_IOVEC (MAX_IOVEC - 3)
+#define NR_IOVEC (MAX_IOVEC - 3)
struct ev_rpc_vec {
- int count;
- struct iovec vector[NR_IOVEC];
+ int count;
+ struct iovec vector[NR_IOVEC];
- /* sequence number */
- unsigned long seq;
+ /* sequence number */
+ unsigned long seq;
};
struct ev_rpc {
- rbuf_list_t *rlist;
- struct rpc_clnt *rpc;
- struct ev_rpc_vec vec;
+ rbuf_list_t *rlist;
+ struct rpc_clnt *rpc;
+ struct ev_rpc_vec vec;
};
/**
@@ -35,215 +35,229 @@ struct ev_rpc {
* intelligence can be built into the server.
*/
int
-changelog_event_dispatch_cbk (struct rpc_req *req,
- struct iovec *iov, int count, void *myframe)
+changelog_event_dispatch_cbk(struct rpc_req *req, struct iovec *iov, int count,
+ void *myframe)
{
- return 0;
+ return 0;
}
/* dispatcher RPC */
int
-changelog_dispatch_vec (call_frame_t *frame, xlator_t *this,
- struct rpc_clnt *rpc, struct ev_rpc_vec *vec)
+changelog_dispatch_vec(call_frame_t *frame, xlator_t *this,
+ struct rpc_clnt *rpc, struct ev_rpc_vec *vec)
{
- struct timeval tv = {0,};
- changelog_event_req req = {0,};
-
- (void) gettimeofday (&tv, NULL);
-
- /**
- * Event dispatch RPC header contains a sequence number for each
- * dispatch. This allows the receiver to order the request before
- * processing.
- */
- req.seq = vec->seq;
- req.tv_sec = tv.tv_sec;
- req.tv_usec = tv.tv_usec;
-
- return changelog_rpc_sumbit_req (rpc, (void *)&req,
- frame, &changelog_ev_program,
- CHANGELOG_REV_PROC_EVENT,
- vec->vector, vec->count, NULL,
- this, changelog_event_dispatch_cbk,
- (xdrproc_t) xdr_changelog_event_req);
- }
-
- int
- changelog_event_dispatch_rpc (call_frame_t *frame, xlator_t *this, void *data)
- {
- int idx = 0;
- int count = 0;
- int ret = 0;
- unsigned long sequence = 0;
- rbuf_iovec_t *rvec = NULL;
- struct ev_rpc *erpc = NULL;
- struct rlist_iter riter = {{0,},};
-
- /* dispatch NR_IOVEC IO vectors at a time. */
-
- erpc = data;
- sequence = erpc->rlist->seq[0];
-
- rlist_iter_init (&riter, erpc->rlist);
-
- rvec_for_each_entry (rvec, &riter) {
- idx = count % NR_IOVEC;
- if (++count == NR_IOVEC) {
- erpc->vec.vector[idx] = rvec->iov;
- erpc->vec.seq = sequence++;
- erpc->vec.count = NR_IOVEC;
-
- ret = changelog_dispatch_vec (frame, this,
- erpc->rpc, &erpc->vec);
- if (ret)
- break;
- count = 0;
- continue;
- }
-
- erpc->vec.vector[idx] = rvec->iov;
- }
-
- if (ret)
- goto error_return;
-
- idx = count % NR_IOVEC;
- if (idx) {
- erpc->vec.seq = sequence;
- erpc->vec.count = idx;
-
- ret = changelog_dispatch_vec (frame, this,
- erpc->rpc, &erpc->vec);
- }
-
- error_return:
- return ret;
+ struct timeval tv = {
+ 0,
+ };
+ changelog_event_req req = {
+ 0,
+ };
+
+ (void)gettimeofday(&tv, NULL);
+
+ /**
+ * Event dispatch RPC header contains a sequence number for each
+ * dispatch. This allows the receiver to order the request before
+ * processing.
+ */
+ req.seq = vec->seq;
+ req.tv_sec = tv.tv_sec;
+ req.tv_usec = tv.tv_usec;
+
+ return changelog_rpc_sumbit_req(
+ rpc, (void *)&req, frame, &changelog_ev_program,
+ CHANGELOG_REV_PROC_EVENT, vec->vector, vec->count, NULL, this,
+ changelog_event_dispatch_cbk, (xdrproc_t)xdr_changelog_event_req);
}
int
-changelog_rpc_notify (struct rpc_clnt *rpc,
- void *mydata, rpc_clnt_event_t event, void *data)
+changelog_event_dispatch_rpc(call_frame_t *frame, xlator_t *this, void *data)
{
- xlator_t *this = NULL;
- changelog_rpc_clnt_t *crpc = NULL;
- changelog_clnt_t *c_clnt = NULL;
- changelog_priv_t *priv = NULL;
- changelog_ev_selector_t *selection = NULL;
+ int idx = 0;
+ int count = 0;
+ int ret = 0;
+ unsigned long sequence = 0;
+ rbuf_iovec_t *rvec = NULL;
+ struct ev_rpc *erpc = NULL;
+ struct rlist_iter riter = {
+ {
+ 0,
+ },
+ };
+
+ /* dispatch NR_IOVEC IO vectors at a time. */
+
+ erpc = data;
+ sequence = erpc->rlist->seq[0];
+
+ rlist_iter_init(&riter, erpc->rlist);
+
+ rvec_for_each_entry(rvec, &riter)
+ {
+ idx = count % NR_IOVEC;
+ if (++count == NR_IOVEC) {
+ erpc->vec.vector[idx] = rvec->iov;
+ erpc->vec.seq = sequence++;
+ erpc->vec.count = NR_IOVEC;
- crpc = mydata;
- this = crpc->this;
- c_clnt = crpc->c_clnt;
+ ret = changelog_dispatch_vec(frame, this, erpc->rpc, &erpc->vec);
+ if (ret)
+ break;
+ count = 0;
+ continue;
+ }
+
+ erpc->vec.vector[idx] = rvec->iov;
+ }
- priv = this->private;
+ if (ret)
+ goto error_return;
- switch (event) {
+ idx = count % NR_IOVEC;
+ if (idx) {
+ erpc->vec.seq = sequence;
+ erpc->vec.count = idx;
+
+ ret = changelog_dispatch_vec(frame, this, erpc->rpc, &erpc->vec);
+ }
+
+error_return:
+ return ret;
+}
+
+int
+changelog_rpc_notify(struct rpc_clnt *rpc, void *mydata, rpc_clnt_event_t event,
+ void *data)
+{
+ xlator_t *this = NULL;
+ changelog_rpc_clnt_t *crpc = NULL;
+ changelog_clnt_t *c_clnt = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_ev_selector_t *selection = NULL;
+ uint64_t clntcnt = 0;
+ uint64_t xprtcnt = 0;
+
+ crpc = mydata;
+ this = crpc->this;
+ c_clnt = crpc->c_clnt;
+
+ priv = this->private;
+
+ switch (event) {
case RPC_CLNT_CONNECT:
- rpc_clnt_set_connected (&rpc->conn);
- selection = &priv->ev_selection;
+ selection = &priv->ev_selection;
+ GF_ATOMIC_INC(priv->clntcnt);
- LOCK (&c_clnt->wait_lock);
+ LOCK(&c_clnt->wait_lock);
+ {
+ LOCK(&c_clnt->active_lock);
{
- LOCK (&c_clnt->active_lock);
- {
- changelog_select_event (this, selection,
- crpc->filter);
- list_move_tail (&crpc->list, &c_clnt->active);
- }
- UNLOCK (&c_clnt->active_lock);
+ changelog_select_event(this, selection, crpc->filter);
+ list_move_tail(&crpc->list, &c_clnt->active);
}
- UNLOCK (&c_clnt->wait_lock);
+ UNLOCK(&c_clnt->active_lock);
+ }
+ UNLOCK(&c_clnt->wait_lock);
- break;
+ break;
case RPC_CLNT_DISCONNECT:
- rpc_clnt_disable (crpc->rpc);
+ rpc_clnt_disable(crpc->rpc);
- /* rpc_clnt_disable doesn't unref the rpc. It just marks
- * the rpc as disabled and cancels reconnection timer.
- * Hence unref the rpc object to free it.
- */
- rpc_clnt_unref (crpc->rpc);
+ /* rpc_clnt_disable doesn't unref the rpc. It just marks
+ * the rpc as disabled and cancels reconnection timer.
+ * Hence unref the rpc object to free it.
+ */
+ rpc_clnt_unref(crpc->rpc);
+ if (priv)
selection = &priv->ev_selection;
- LOCK (&crpc->lock);
- {
- changelog_deselect_event (this, selection,
- crpc->filter);
- changelog_set_disconnect_flag (crpc, _gf_true);
- }
- UNLOCK (&crpc->lock);
-
- break;
+ LOCK(&crpc->lock);
+ {
+ if (selection)
+ changelog_deselect_event(this, selection, crpc->filter);
+ changelog_set_disconnect_flag(crpc, _gf_true);
+ }
+ UNLOCK(&crpc->lock);
+ LOCK(&c_clnt->active_lock);
+ {
+ list_del_init(&crpc->list);
+ }
+ UNLOCK(&c_clnt->active_lock);
+
+ break;
case RPC_CLNT_MSG:
case RPC_CLNT_DESTROY:
- /* Free up mydata */
- changelog_rpc_clnt_unref (crpc);
- break;
+ /* Free up mydata */
+ changelog_rpc_clnt_unref(crpc);
+ clntcnt = GF_ATOMIC_DEC(priv->clntcnt);
+ xprtcnt = GF_ATOMIC_GET(priv->xprtcnt);
+ if (this->cleanup_starting) {
+ if (!clntcnt && !xprtcnt)
+ changelog_process_cleanup_event(this);
+ }
+ break;
case RPC_CLNT_PING:
- break;
- }
+ break;
+ }
- return 0;
+ return 0;
}
void *
-changelog_ev_connector (void *data)
+changelog_ev_connector(void *data)
{
- xlator_t *this = NULL;
- changelog_clnt_t *c_clnt = NULL;
- changelog_rpc_clnt_t *crpc = NULL;
+ xlator_t *this = NULL;
+ changelog_clnt_t *c_clnt = NULL;
+ changelog_rpc_clnt_t *crpc = NULL;
- c_clnt = data;
- this = c_clnt->this;
+ c_clnt = data;
+ this = c_clnt->this;
- while (1) {
- pthread_mutex_lock (&c_clnt->pending_lock);
- {
- while (list_empty (&c_clnt->pending))
- pthread_cond_wait (&c_clnt->pending_cond,
- &c_clnt->pending_lock);
- crpc = list_first_entry (&c_clnt->pending,
- changelog_rpc_clnt_t, list);
- crpc->rpc =
- changelog_rpc_client_init (this, crpc,
- crpc->sock,
- changelog_rpc_notify);
- if (!crpc->rpc) {
- gf_smsg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_RPC_CONNECT_ERROR,
- "failed to connect back",
- "path=%s", crpc->sock,
- NULL);
- crpc->cleanup (crpc);
- goto mutex_unlock;
- }
-
- LOCK (&c_clnt->wait_lock);
- {
- list_move_tail (&crpc->list, &c_clnt->waitq);
- }
- UNLOCK (&c_clnt->wait_lock);
- }
- mutex_unlock:
- pthread_mutex_unlock (&c_clnt->pending_lock);
+ while (1) {
+ pthread_mutex_lock(&c_clnt->pending_lock);
+ {
+ while (list_empty(&c_clnt->pending))
+ pthread_cond_wait(&c_clnt->pending_cond, &c_clnt->pending_lock);
+ crpc = list_first_entry(&c_clnt->pending, changelog_rpc_clnt_t,
+ list);
+ crpc->rpc = changelog_rpc_client_init(this, crpc, crpc->sock,
+ changelog_rpc_notify);
+ if (!crpc->rpc) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_MSG_RPC_CONNECT_ERROR, "path=%s", crpc->sock,
+ NULL);
+ crpc->cleanup(crpc);
+ goto mutex_unlock;
+ }
+
+ LOCK(&c_clnt->wait_lock);
+ {
+ list_move_tail(&crpc->list, &c_clnt->waitq);
+ }
+ UNLOCK(&c_clnt->wait_lock);
}
+ mutex_unlock:
+ pthread_mutex_unlock(&c_clnt->pending_lock);
+ }
- return NULL;
+ return NULL;
}
void
-changelog_ev_cleanup_connections (xlator_t *this, changelog_clnt_t *c_clnt)
+changelog_ev_cleanup_connections(xlator_t *this, changelog_clnt_t *c_clnt)
{
- changelog_rpc_clnt_t *crpc = NULL;
+ changelog_rpc_clnt_t *crpc = NULL;
- /* cleanup active connections */
- LOCK (&c_clnt->active_lock);
+ /* cleanup active connections */
+ LOCK(&c_clnt->active_lock);
+ {
+ list_for_each_entry(crpc, &c_clnt->active, list)
{
- list_for_each_entry (crpc, &c_clnt->active, list) {
- rpc_clnt_disable (crpc->rpc);
- }
+ rpc_clnt_disable(crpc->rpc);
}
- UNLOCK (&c_clnt->active_lock);
+ }
+ UNLOCK(&c_clnt->active_lock);
}
/**
@@ -254,147 +268,145 @@ changelog_ev_cleanup_connections (xlator_t *this, changelog_clnt_t *c_clnt)
*/
static changelog_rpc_clnt_t *
-get_client (changelog_clnt_t *c_clnt, struct list_head **next)
+get_client(changelog_clnt_t *c_clnt, struct list_head **next)
{
- changelog_rpc_clnt_t *crpc = NULL;
-
- LOCK (&c_clnt->active_lock);
- {
- if (*next == &c_clnt->active)
- goto unblock;
- crpc = list_entry (*next, changelog_rpc_clnt_t, list);
- /* ref rpc as DISCONNECT might unref the rpc asynchronously */
- changelog_rpc_clnt_ref (crpc);
- rpc_clnt_ref (crpc->rpc);
- *next = (*next)->next;
- }
- unblock:
- UNLOCK (&c_clnt->active_lock);
-
- return crpc;
+ changelog_rpc_clnt_t *crpc = NULL;
+
+ LOCK(&c_clnt->active_lock);
+ {
+ if (*next == &c_clnt->active)
+ goto unblock;
+ crpc = list_entry(*next, changelog_rpc_clnt_t, list);
+ /* ref rpc as DISCONNECT might unref the rpc asynchronously */
+ changelog_rpc_clnt_ref(crpc);
+ rpc_clnt_ref(crpc->rpc);
+ *next = (*next)->next;
+ }
+unblock:
+ UNLOCK(&c_clnt->active_lock);
+
+ return crpc;
}
static void
-put_client (changelog_clnt_t *c_clnt, changelog_rpc_clnt_t *crpc)
+put_client(changelog_clnt_t *c_clnt, changelog_rpc_clnt_t *crpc)
{
- LOCK (&c_clnt->active_lock);
- {
- rpc_clnt_unref (crpc->rpc);
- changelog_rpc_clnt_unref (crpc);
- }
- UNLOCK (&c_clnt->active_lock);
+ LOCK(&c_clnt->active_lock);
+ {
+ rpc_clnt_unref(crpc->rpc);
+ changelog_rpc_clnt_unref(crpc);
+ }
+ UNLOCK(&c_clnt->active_lock);
}
void
-_dispatcher (rbuf_list_t *rlist, void *arg)
+_dispatcher(rbuf_list_t *rlist, void *arg)
{
- xlator_t *this = NULL;
- changelog_clnt_t *c_clnt = NULL;
- changelog_rpc_clnt_t *crpc = NULL;
- struct ev_rpc erpc = {0,};
- struct list_head *next = NULL;
-
- c_clnt = arg;
- this = c_clnt->this;
-
- erpc.rlist = rlist;
- next = c_clnt->active.next;
-
- while (1) {
- crpc = get_client (c_clnt, &next);
- if (!crpc)
- break;
- erpc.rpc = crpc->rpc;
- (void) changelog_invoke_rpc (this, crpc->rpc,
- &changelog_ev_program,
- CHANGELOG_REV_PROC_EVENT, &erpc);
- put_client (c_clnt, crpc);
- }
+ xlator_t *this = NULL;
+ changelog_clnt_t *c_clnt = NULL;
+ changelog_rpc_clnt_t *crpc = NULL;
+ struct ev_rpc erpc = {
+ 0,
+ };
+ struct list_head *next = NULL;
+
+ c_clnt = arg;
+ this = c_clnt->this;
+
+ erpc.rlist = rlist;
+ next = c_clnt->active.next;
+
+ while (1) {
+ crpc = get_client(c_clnt, &next);
+ if (!crpc)
+ break;
+ erpc.rpc = crpc->rpc;
+ (void)changelog_invoke_rpc(this, crpc->rpc, &changelog_ev_program,
+ CHANGELOG_REV_PROC_EVENT, &erpc);
+ put_client(c_clnt, crpc);
+ }
}
/** this is called under rotbuff's lock */
void
-sequencer (rbuf_list_t *rlist, void *mydata)
+sequencer(rbuf_list_t *rlist, void *mydata)
{
- unsigned long range = 0;
- changelog_clnt_t *c_clnt = 0;
+ unsigned long range = 0;
+ changelog_clnt_t *c_clnt = 0;
- c_clnt = mydata;
+ c_clnt = mydata;
- range = (RLIST_ENTRY_COUNT (rlist)) / NR_IOVEC;
- if ((RLIST_ENTRY_COUNT (rlist)) % NR_IOVEC)
- range++;
- RLIST_STORE_SEQ (rlist, c_clnt->sequence, range);
+ range = (RLIST_ENTRY_COUNT(rlist)) / NR_IOVEC;
+ if ((RLIST_ENTRY_COUNT(rlist)) % NR_IOVEC)
+ range++;
+ RLIST_STORE_SEQ(rlist, c_clnt->sequence, range);
- c_clnt->sequence += range;
+ c_clnt->sequence += range;
}
void *
-changelog_ev_dispatch (void *data)
+changelog_ev_dispatch(void *data)
{
- int ret = 0;
- void *opaque = NULL;
- xlator_t *this = NULL;
- changelog_clnt_t *c_clnt = NULL;
- struct timeval tv = {0,};
-
- c_clnt = data;
- this = c_clnt->this;
-
- while (1) {
- /* TODO: change this to be pthread cond based.. later */
-
- tv.tv_sec = 1;
- tv.tv_usec = 0;
- select (0, NULL, NULL, NULL, &tv);
-
- ret = rbuf_get_buffer (c_clnt->rbuf,
- &opaque, sequencer, c_clnt);
- if (ret != RBUF_CONSUMABLE) {
- if (ret != RBUF_EMPTY)
- gf_smsg (this->name, GF_LOG_WARNING, 0,
- CHANGELOG_MSG_BUFFER_STARVATION_ERROR,
- "Failed to get buffer for RPC dispatch",
- "rbuf_retval=%d", ret,
- NULL);
- continue;
- }
-
- ret = rbuf_wait_for_completion (c_clnt->rbuf,
- opaque, _dispatcher, c_clnt);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- CHANGELOG_MSG_PUT_BUFFER_FAILED,
- "failed to put buffer after consumption");
-
+ int ret = 0;
+ void *opaque = NULL;
+ xlator_t *this = NULL;
+ changelog_clnt_t *c_clnt = NULL;
+ struct timeval tv = {
+ 0,
+ };
+
+ c_clnt = data;
+ this = c_clnt->this;
+
+ while (1) {
+ /* TODO: change this to be pthread cond based.. later */
+
+ tv.tv_sec = 1;
+ tv.tv_usec = 0;
+ select(0, NULL, NULL, NULL, &tv);
+
+ ret = rbuf_get_buffer(c_clnt->rbuf, &opaque, sequencer, c_clnt);
+ if (ret != RBUF_CONSUMABLE) {
+ if (ret != RBUF_EMPTY)
+ gf_smsg(this->name, GF_LOG_WARNING, 0,
+ CHANGELOG_MSG_BUFFER_STARVATION_ERROR,
+ "Failed to get buffer for RPC dispatch",
+ "rbuf_retval=%d", ret, NULL);
+ continue;
}
- return NULL;
+ ret = rbuf_wait_for_completion(c_clnt->rbuf, opaque, _dispatcher,
+ c_clnt);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_WARNING, 0,
+ CHANGELOG_MSG_PUT_BUFFER_FAILED, NULL);
+ }
+
+ return NULL;
}
void
-changelog_ev_queue_connection (changelog_clnt_t *c_clnt,
- changelog_rpc_clnt_t *crpc)
+changelog_ev_queue_connection(changelog_clnt_t *c_clnt,
+ changelog_rpc_clnt_t *crpc)
{
- pthread_mutex_lock (&c_clnt->pending_lock);
- {
- list_add_tail (&crpc->list, &c_clnt->pending);
- pthread_cond_signal (&c_clnt->pending_cond);
- }
- pthread_mutex_unlock (&c_clnt->pending_lock);
+ pthread_mutex_lock(&c_clnt->pending_lock);
+ {
+ list_add_tail(&crpc->list, &c_clnt->pending);
+ pthread_cond_signal(&c_clnt->pending_cond);
+ }
+ pthread_mutex_unlock(&c_clnt->pending_lock);
}
struct rpc_clnt_procedure changelog_ev_procs[CHANGELOG_REV_PROC_MAX] = {
- [CHANGELOG_REV_PROC_NULL] = {"NULL", NULL},
- [CHANGELOG_REV_PROC_EVENT] = {
- "EVENT DISPATCH", changelog_event_dispatch_rpc
- },
+ [CHANGELOG_REV_PROC_NULL] = {"NULL", NULL},
+ [CHANGELOG_REV_PROC_EVENT] = {"EVENT DISPATCH",
+ changelog_event_dispatch_rpc},
};
struct rpc_clnt_program changelog_ev_program = {
- .progname = "CHANGELOG EVENT DISPATCHER",
- .prognum = CHANGELOG_REV_RPC_PROCNUM,
- .progver = CHANGELOG_REV_RPC_PROCVER,
- .numproc = CHANGELOG_REV_PROC_MAX,
- .proctable = changelog_ev_procs,
+ .progname = "CHANGELOG EVENT DISPATCHER",
+ .prognum = CHANGELOG_REV_RPC_PROCNUM,
+ .progver = CHANGELOG_REV_RPC_PROCVER,
+ .numproc = CHANGELOG_REV_PROC_MAX,
+ .proctable = changelog_ev_procs,
};
diff --git a/xlators/features/changelog/src/changelog-ev-handle.h b/xlators/features/changelog/src/changelog-ev-handle.h
index e89af8793a8..cc1af58a276 100644
--- a/xlators/features/changelog/src/changelog-ev-handle.h
+++ b/xlators/features/changelog/src/changelog-ev-handle.h
@@ -11,68 +11,67 @@
#ifndef __CHANGELOG_EV_HANDLE_H
#define __CHANGELOG_EV_HANDLE_H
-#include "list.h"
-#include "xlator.h"
+#include <glusterfs/list.h>
+#include <glusterfs/xlator.h>
#include "rpc-clnt.h"
-#include "rot-buffs.h"
+#include <glusterfs/rot-buffs.h>
struct changelog_clnt;
typedef struct changelog_rpc_clnt {
- xlator_t *this;
+ xlator_t *this;
- gf_lock_t lock;
+ gf_lock_t lock;
- gf_atomic_t ref;
- gf_boolean_t disconnected;
+ gf_atomic_t ref;
+ gf_boolean_t disconnected;
- unsigned int filter;
- char sock[UNIX_PATH_MAX];
+ unsigned int filter;
+ char sock[UNIX_PATH_MAX];
- struct changelog_clnt *c_clnt; /* back pointer to list holder */
+ struct changelog_clnt *c_clnt; /* back pointer to list holder */
- struct rpc_clnt *rpc; /* RPC client endpoint */
+ struct rpc_clnt *rpc; /* RPC client endpoint */
- struct list_head list; /* ->pending, ->waitq, ->active */
+ struct list_head list; /* ->pending, ->waitq, ->active */
- void (*cleanup)
- (struct changelog_rpc_clnt *); /* cleanup handler */
+ void (*cleanup)(struct changelog_rpc_clnt *); /* cleanup handler */
} changelog_rpc_clnt_t;
static inline void
-changelog_rpc_clnt_ref (changelog_rpc_clnt_t *crpc)
+changelog_rpc_clnt_ref(changelog_rpc_clnt_t *crpc)
{
- GF_ATOMIC_INC (crpc->ref);
+ GF_ATOMIC_INC(crpc->ref);
}
static inline void
-changelog_set_disconnect_flag (changelog_rpc_clnt_t *crpc, gf_boolean_t flag)
+changelog_set_disconnect_flag(changelog_rpc_clnt_t *crpc, gf_boolean_t flag)
{
- crpc->disconnected = flag;
+ crpc->disconnected = flag;
}
static inline int
-changelog_rpc_clnt_is_disconnected (changelog_rpc_clnt_t *crpc)
+changelog_rpc_clnt_is_disconnected(changelog_rpc_clnt_t *crpc)
{
- return (crpc->disconnected == _gf_true);
+ return (crpc->disconnected == _gf_true);
}
static inline void
-changelog_rpc_clnt_unref (changelog_rpc_clnt_t *crpc)
+changelog_rpc_clnt_unref(changelog_rpc_clnt_t *crpc)
{
- gf_boolean_t gone = _gf_false;
- uint64_t ref = 0;
+ gf_boolean_t gone = _gf_false;
+ uint64_t ref = 0;
- ref = GF_ATOMIC_DEC (crpc->ref);
+ ref = GF_ATOMIC_DEC(crpc->ref);
- if (!ref && changelog_rpc_clnt_is_disconnected (crpc)) {
- list_del (&crpc->list);
- gone = _gf_true;
- }
+ if (!ref && changelog_rpc_clnt_is_disconnected(crpc)) {
+ list_del(&crpc->list);
+ gone = _gf_true;
+ }
- if (gone)
- crpc->cleanup (crpc);
+ if (gone)
+ crpc->cleanup(crpc);
}
/**
@@ -100,35 +99,38 @@ changelog_rpc_clnt_unref (changelog_rpc_clnt_t *crpc)
*/
typedef struct changelog_clnt {
- xlator_t *this;
+ xlator_t *this;
- /* pending connections */
- pthread_mutex_t pending_lock;
- pthread_cond_t pending_cond;
- struct list_head pending;
+ /* pending connections */
+ pthread_mutex_t pending_lock;
+ pthread_cond_t pending_cond;
+ struct list_head pending;
- /* current active connections */
- gf_lock_t active_lock;
- struct list_head active;
+ /* current active connections */
+ gf_lock_t active_lock;
+ struct list_head active;
- gf_lock_t wait_lock;
- struct list_head waitq;
+ gf_lock_t wait_lock;
+ struct list_head waitq;
- /* consumer part of rot-buffs */
- rbuf_t *rbuf;
- unsigned long sequence;
+ /* consumer part of rot-buffs */
+ rbuf_t *rbuf;
+ unsigned long sequence;
} changelog_clnt_t;
-void *changelog_ev_connector (void *);
+void *
+changelog_ev_connector(void *);
-void *changelog_ev_dispatch (void *);
+void *
+changelog_ev_dispatch(void *);
/* APIs */
void
-changelog_ev_queue_connection (changelog_clnt_t *, changelog_rpc_clnt_t *);
+changelog_ev_queue_connection(changelog_clnt_t *, changelog_rpc_clnt_t *);
void
-changelog_ev_cleanup_connections (xlator_t *, changelog_clnt_t *);
+changelog_ev_cleanup_connections(xlator_t *, changelog_clnt_t *);
+void
+changelog_process_cleanup_event(xlator_t *);
#endif
-
diff --git a/xlators/features/changelog/src/changelog-helpers.c b/xlators/features/changelog/src/changelog-helpers.c
index 306287ecd11..e561997d858 100644
--- a/xlators/features/changelog/src/changelog-helpers.c
+++ b/xlators/features/changelog/src/changelog-helpers.c
@@ -8,11 +8,11 @@
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "defaults.h"
-#include "logging.h"
-#include "iobuf.h"
-#include "syscall.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/iobuf.h>
+#include <glusterfs/syscall.h>
#include "changelog-helpers.h"
#include "changelog-encoders.h"
@@ -22,313 +22,281 @@
#include "changelog-encoders.h"
#include "changelog-rpc-common.h"
#include <pthread.h>
+#include <time.h>
static void
-changelog_cleanup_free_mutex (void *arg_mutex)
+changelog_cleanup_free_mutex(void *arg_mutex)
{
- pthread_mutex_t *p_mutex = (pthread_mutex_t*) arg_mutex;
+ pthread_mutex_t *p_mutex = (pthread_mutex_t *)arg_mutex;
if (p_mutex)
- pthread_mutex_unlock(p_mutex);
+ pthread_mutex_unlock(p_mutex);
}
int
-changelog_thread_cleanup (xlator_t *this, pthread_t thr_id)
+changelog_thread_cleanup(xlator_t *this, pthread_t thr_id)
{
- int ret = 0;
- void *retval = NULL;
-
- /* send a cancel request to the thread */
- ret = pthread_cancel (thr_id);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_PTHREAD_CANCEL_FAILED,
- "could not cancel thread");
- goto out;
- }
-
- ret = pthread_join (thr_id, &retval);
- if ((ret != 0) || (retval != PTHREAD_CANCELED)) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_PTHREAD_CANCEL_FAILED,
- "cancel request not adhered as expected");
- }
+ int ret = 0;
+ void *retval = NULL;
+
+ /* send a cancel request to the thread */
+ ret = pthread_cancel(thr_id);
+ if (ret != 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_PTHREAD_CANCEL_FAILED, NULL);
+ goto out;
+ }
+
+ ret = pthread_join(thr_id, &retval);
+ if ((ret != 0) || (retval != PTHREAD_CANCELED)) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_PTHREAD_CANCEL_FAILED, NULL);
+ }
- out:
- return ret;
+out:
+ return ret;
}
void *
-changelog_get_usable_buffer (changelog_local_t *local)
+changelog_get_usable_buffer(changelog_local_t *local)
{
- changelog_log_data_t *cld = NULL;
+ changelog_log_data_t *cld = NULL;
- if (!local)
- return NULL;
+ if (!local)
+ return NULL;
- cld = &local->cld;
- if (!cld->cld_iobuf)
- return NULL;
+ cld = &local->cld;
+ if (!cld->cld_iobuf)
+ return NULL;
- return cld->cld_iobuf->ptr;
+ return cld->cld_iobuf->ptr;
}
static int
-changelog_selector_index (unsigned int selector)
+changelog_selector_index(unsigned int selector)
{
- return (ffs (selector) - 1);
+ return (ffs(selector) - 1);
}
int
-changelog_ev_selected (xlator_t *this,
- changelog_ev_selector_t *selection,
- unsigned int selector)
+changelog_ev_selected(xlator_t *this, changelog_ev_selector_t *selection,
+ unsigned int selector)
{
- int idx = 0;
-
- idx = changelog_selector_index (selector);
- gf_msg_debug (this->name, 0,
- "selector ref count for %d (idx: %d): %d",
- selector, idx, selection->ref[idx]);
- /* this can be lockless */
- return (idx < CHANGELOG_EV_SELECTION_RANGE
- && (selection->ref[idx] > 0));
+ int idx = 0;
+
+ idx = changelog_selector_index(selector);
+ gf_msg_debug(this->name, 0, "selector ref count for %d (idx: %d): %d",
+ selector, idx, selection->ref[idx]);
+ /* this can be lockless */
+ return (idx < CHANGELOG_EV_SELECTION_RANGE && (selection->ref[idx] > 0));
}
void
-changelog_select_event (xlator_t *this,
- changelog_ev_selector_t *selection,
- unsigned int selector)
+changelog_select_event(xlator_t *this, changelog_ev_selector_t *selection,
+ unsigned int selector)
{
- int idx = 0;
-
- LOCK (&selection->reflock);
- {
- while (selector) {
- idx = changelog_selector_index (selector);
- if (idx < CHANGELOG_EV_SELECTION_RANGE) {
- selection->ref[idx]++;
- gf_msg_debug (this->name, 0,
- "selecting event %d", idx);
- }
- selector &= ~(1 << idx);
- }
- }
- UNLOCK (&selection->reflock);
+ int idx = 0;
+
+ LOCK(&selection->reflock);
+ {
+ while (selector) {
+ idx = changelog_selector_index(selector);
+ if (idx < CHANGELOG_EV_SELECTION_RANGE) {
+ selection->ref[idx]++;
+ gf_msg_debug(this->name, 0, "selecting event %d", idx);
+ }
+ selector &= ~(1 << idx);
+ }
+ }
+ UNLOCK(&selection->reflock);
}
void
-changelog_deselect_event (xlator_t *this,
- changelog_ev_selector_t *selection,
- unsigned int selector)
+changelog_deselect_event(xlator_t *this, changelog_ev_selector_t *selection,
+ unsigned int selector)
{
- int idx = 0;
-
- LOCK (&selection->reflock);
- {
- while (selector) {
- idx = changelog_selector_index (selector);
- if (idx < CHANGELOG_EV_SELECTION_RANGE) {
- selection->ref[idx]--;
- gf_msg_debug (this->name, 0,
- "de-selecting event %d", idx);
- }
- selector &= ~(1 << idx);
- }
- }
- UNLOCK (&selection->reflock);
+ int idx = 0;
+
+ LOCK(&selection->reflock);
+ {
+ while (selector) {
+ idx = changelog_selector_index(selector);
+ if (idx < CHANGELOG_EV_SELECTION_RANGE) {
+ selection->ref[idx]--;
+ gf_msg_debug(this->name, 0, "de-selecting event %d", idx);
+ }
+ selector &= ~(1 << idx);
+ }
+ }
+ UNLOCK(&selection->reflock);
}
int
-changelog_init_event_selection (xlator_t *this,
- changelog_ev_selector_t *selection)
+changelog_init_event_selection(xlator_t *this,
+ changelog_ev_selector_t *selection)
{
- int ret = 0;
- int j = CHANGELOG_EV_SELECTION_RANGE;
+ int ret = 0;
+ int j = CHANGELOG_EV_SELECTION_RANGE;
- ret = LOCK_INIT (&selection->reflock);
- if (ret != 0)
- return -1;
-
- LOCK (&selection->reflock);
- {
- while (j--) {
- selection->ref[j] = 0;
- }
- }
- UNLOCK (&selection->reflock);
-
- return 0;
-}
-
-int
-changelog_cleanup_event_selection (xlator_t *this,
- changelog_ev_selector_t *selection)
-{
- int j = CHANGELOG_EV_SELECTION_RANGE;
+ ret = LOCK_INIT(&selection->reflock);
+ if (ret != 0)
+ return -1;
- LOCK (&selection->reflock);
- {
- while (j--) {
- if (selection->ref[j] > 0)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- CHANGELOG_MSG_CLEANUP_ON_ACTIVE_REF,
- "changelog event selection cleaning up "
- " on active references");
- }
+ LOCK(&selection->reflock);
+ {
+ while (j--) {
+ selection->ref[j] = 0;
}
- UNLOCK (&selection->reflock);
+ }
+ UNLOCK(&selection->reflock);
- return LOCK_DESTROY (&selection->reflock);
+ return 0;
}
static void
-changelog_perform_dispatch (xlator_t *this,
- changelog_priv_t *priv, void *mem, size_t size)
+changelog_perform_dispatch(xlator_t *this, changelog_priv_t *priv, void *mem,
+ size_t size)
{
- char *buf = NULL;
- void *opaque = NULL;
-
- buf = rbuf_reserve_write_area (priv->rbuf, size, &opaque);
- if (!buf) {
- gf_msg_callingfn (this->name,
- GF_LOG_WARNING, 0,
- CHANGELOG_MSG_DISPATCH_EVENT_FAILED,
- "failed to dispatch event");
- return;
- }
+ char *buf = NULL;
+ void *opaque = NULL;
+
+ buf = rbuf_reserve_write_area(priv->rbuf, size, &opaque);
+ if (!buf) {
+ gf_msg_callingfn(this->name, GF_LOG_WARNING, 0,
+ CHANGELOG_MSG_DISPATCH_EVENT_FAILED,
+ "failed to dispatch event");
+ return;
+ }
- memcpy (buf, mem, size);
- rbuf_write_complete (opaque);
+ memcpy(buf, mem, size);
+ rbuf_write_complete(opaque);
}
void
-changelog_dispatch_event (xlator_t *this,
- changelog_priv_t *priv, changelog_event_t *ev)
+changelog_dispatch_event(xlator_t *this, changelog_priv_t *priv,
+ changelog_event_t *ev)
{
- changelog_ev_selector_t *selection = NULL;
+ changelog_ev_selector_t *selection = NULL;
- selection = &priv->ev_selection;
- if (changelog_ev_selected (this, selection, ev->ev_type)) {
- changelog_perform_dispatch (this, priv, ev, CHANGELOG_EV_SIZE);
- }
+ selection = &priv->ev_selection;
+ if (changelog_ev_selected(this, selection, ev->ev_type)) {
+ changelog_perform_dispatch(this, priv, ev, CHANGELOG_EV_SIZE);
+ }
}
void
-changelog_set_usable_record_and_length (changelog_local_t *local,
- size_t len, int xr)
+changelog_set_usable_record_and_length(changelog_local_t *local, size_t len,
+ int xr)
{
- changelog_log_data_t *cld = NULL;
+ changelog_log_data_t *cld = NULL;
- cld = &local->cld;
+ cld = &local->cld;
- cld->cld_ptr_len = len;
- cld->cld_xtra_records = xr;
+ cld->cld_ptr_len = len;
+ cld->cld_xtra_records = xr;
}
void
-changelog_local_cleanup (xlator_t *xl, changelog_local_t *local)
+changelog_local_cleanup(xlator_t *xl, changelog_local_t *local)
{
- int i = 0;
- changelog_opt_t *co = NULL;
- changelog_log_data_t *cld = NULL;
+ int i = 0;
+ changelog_opt_t *co = NULL;
+ changelog_log_data_t *cld = NULL;
- if (!local)
- return;
+ if (!local)
+ return;
- cld = &local->cld;
+ cld = &local->cld;
- /* cleanup dynamic allocation for extra records */
- if (cld->cld_xtra_records) {
- co = (changelog_opt_t *) cld->cld_ptr;
- for (; i < cld->cld_xtra_records; i++, co++)
- if (co->co_free)
- co->co_free (co);
- }
+ /* cleanup dynamic allocation for extra records */
+ if (cld->cld_xtra_records) {
+ co = (changelog_opt_t *)cld->cld_ptr;
+ for (; i < cld->cld_xtra_records; i++, co++)
+ if (co->co_free)
+ co->co_free(co);
+ }
- CHANGELOG_IOBUF_UNREF (cld->cld_iobuf);
+ CHANGELOG_IOBUF_UNREF(cld->cld_iobuf);
- if (local->inode)
- inode_unref (local->inode);
+ if (local->inode)
+ inode_unref(local->inode);
- mem_put (local);
+ mem_put(local);
}
int
-changelog_write (int fd, char *buffer, size_t len)
+changelog_write(int fd, char *buffer, size_t len)
{
- ssize_t size = 0;
- size_t written = 0;
+ ssize_t size = 0;
+ size_t written = 0;
- while (written < len) {
- size = sys_write (fd,
- buffer + written, len - written);
- if (size <= 0)
- break;
+ while (written < len) {
+ size = sys_write(fd, buffer + written, len - written);
+ if (size <= 0)
+ break;
- written += size;
- }
+ written += size;
+ }
- return (written != len);
+ return (written != len);
}
int
-htime_update (xlator_t *this,
- changelog_priv_t *priv, unsigned long ts,
- char * buffer)
+htime_update(xlator_t *this, changelog_priv_t *priv, time_t ts, char *buffer)
{
- char changelog_path[PATH_MAX+1] = {0,};
- int len = -1;
- char x_value[25] = {0,};
- /* time stamp(10) + : (1) + rolltime (12 ) + buffer (2) */
- int ret = 0;
-
- if (priv->htime_fd ==-1) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_HTIME_ERROR,
- "Htime fd not available for updation");
- ret = -1;
- goto out;
- }
- strncpy (changelog_path, buffer, PATH_MAX);
- len = strlen (changelog_path);
- changelog_path[len] = '\0'; /* redundant */
-
- if (changelog_write (priv->htime_fd, (void*) changelog_path, len+1 ) < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_HTIME_ERROR,
- "Htime file content write failed");
- ret =-1;
- goto out;
- }
-
- snprintf (x_value, sizeof x_value, "%lu:%d",
- ts, priv->rollover_count);
-
- if (sys_fsetxattr (priv->htime_fd, HTIME_KEY, x_value,
- strlen (x_value), XATTR_REPLACE)) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_HTIME_ERROR,
- "Htime xattr updation failed with XATTR_REPLACE",
- "changelog=%s", changelog_path,
- NULL);
-
- if (sys_fsetxattr (priv->htime_fd, HTIME_KEY, x_value,
- strlen (x_value), 0)) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_HTIME_ERROR,
- "Htime xattr updation failed",
- "changelog=%s", changelog_path,
- NULL);
- ret = -1;
- goto out;
- }
- }
-
- priv->rollover_count +=1;
+ char changelog_path[PATH_MAX + 1] = {
+ 0,
+ };
+ int len = -1;
+ char x_value[25] = {
+ 0,
+ };
+ /* time stamp(10) + : (1) + rolltime (12 ) + buffer (2) */
+ int ret = 0;
+
+ if (priv->htime_fd == -1) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_HTIME_ERROR,
+ "reason=fd not available", NULL);
+ ret = -1;
+ goto out;
+ }
+ len = snprintf(changelog_path, PATH_MAX, "%s", buffer);
+ if (len >= PATH_MAX) {
+ ret = -1;
+ goto out;
+ }
+ if (changelog_write(priv->htime_fd, (void *)changelog_path, len + 1) < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_HTIME_ERROR,
+ "reason=write failed", NULL);
+ ret = -1;
+ goto out;
+ }
+
+ len = snprintf(x_value, sizeof(x_value), "%ld:%d", ts,
+ priv->rollover_count);
+ if (len >= sizeof(x_value)) {
+ ret = -1;
+ goto out;
+ }
+
+ if (sys_fsetxattr(priv->htime_fd, HTIME_KEY, x_value, len, XATTR_REPLACE)) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_HTIME_ERROR,
+ "reason=xattr updation failed", "XATTR_REPLACE=true",
+ "changelog=%s", changelog_path, NULL);
+
+ if (sys_fsetxattr(priv->htime_fd, HTIME_KEY, x_value, len, 0)) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_HTIME_ERROR,
+ "reason=xattr updation failed", "changelog=%s",
+ changelog_path, NULL);
+ ret = -1;
+ goto out;
+ }
+ }
+
+ priv->rollover_count += 1;
out:
- return ret;
+ return ret;
}
/*
@@ -340,43 +308,45 @@ out:
* 0 : If NOT empty, proceed usual.
*/
int
-cl_is_empty (xlator_t *this, int fd)
+cl_is_empty(xlator_t *this, int fd)
{
- int ret = -1;
- size_t elen = 0;
- int encoding = -1;
- char buffer[1024] = {0,};
- struct stat stbuf = {0,};
- int major_version = -1;
- int minor_version = -1;
-
- ret = sys_fstat (fd, &stbuf);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_FSTAT_OP_FAILED,
- "Could not stat (CHANGELOG)");
- goto out;
- }
-
- ret = sys_lseek (fd, 0, SEEK_SET);
- if (ret == -1) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_LSEEK_OP_FAILED,
- "Could not lseek (CHANGELOG)");
- goto out;
- }
-
- CHANGELOG_GET_HEADER_INFO (fd, buffer, 1024, encoding,
- major_version, minor_version, elen);
-
- if (elen == stbuf.st_size) {
- ret = 1;
- } else {
- ret = 0;
- }
+ int ret = -1;
+ size_t elen = 0;
+ int encoding = -1;
+ char buffer[1024] = {
+ 0,
+ };
+ struct stat stbuf = {
+ 0,
+ };
+ int major_version = -1;
+ int minor_version = -1;
+
+ ret = sys_fstat(fd, &stbuf);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_FSTAT_OP_FAILED,
+ NULL);
+ goto out;
+ }
+
+ ret = sys_lseek(fd, 0, SEEK_SET);
+ if (ret == -1) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_LSEEK_OP_FAILED,
+ NULL);
+ goto out;
+ }
+
+ CHANGELOG_GET_HEADER_INFO(fd, buffer, sizeof(buffer), encoding,
+ major_version, minor_version, elen);
+
+ if (elen == stbuf.st_size) {
+ ret = 1;
+ } else {
+ ret = 0;
+ }
out:
- return ret;
+ return ret;
}
/*
@@ -388,162 +358,172 @@ out:
* -1 : Error
*/
int
-update_path (xlator_t *this, char *cl_path)
+update_path(xlator_t *this, char *cl_path)
{
- char low_cl[] = "changelog";
- char up_cl[] = "CHANGELOG";
- char *found = NULL;
- int ret = -1;
-
- found = strstr(cl_path, up_cl);
-
- if (found == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_LSEEK_OP_FAILED,
- "Could not find CHANGELOG in changelog path");
- goto out;
- } else {
- strncpy(found, low_cl, strlen(low_cl));
- }
-
- ret = 0;
+ const char low_cl[] = "changelog";
+ const char up_cl[] = "CHANGELOG";
+ char *found = NULL;
+ int ret = -1;
+
+ found = strstr(cl_path, up_cl);
+
+ if (found == NULL) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_PATH_NOT_FOUND,
+ NULL);
+ goto out;
+ } else {
+ memcpy(found, low_cl, sizeof(low_cl) - 1);
+ }
+
+ ret = 0;
out:
- return ret;
+ return ret;
}
static int
-changelog_rollover_changelog (xlator_t *this,
- changelog_priv_t *priv, unsigned long ts)
+changelog_rollover_changelog(xlator_t *this, changelog_priv_t *priv, time_t ts)
{
- int ret = -1;
- int notify = 0;
- int cl_empty_flag = 0;
- char ofile[PATH_MAX] = {0,};
- char nfile[PATH_MAX] = {0,};
- changelog_event_t ev = {0,};
-
- if (priv->changelog_fd != -1) {
- ret = sys_fsync (priv->changelog_fd);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_FSYNC_OP_FAILED,
- "fsync failed");
- }
- ret = cl_is_empty (this, priv->changelog_fd);
- if (ret == 1) {
- cl_empty_flag = 1;
- } else if (ret == -1) {
- /* Log error but proceed as usual */
- gf_msg (this->name, GF_LOG_WARNING, 0,
- CHANGELOG_MSG_DETECT_EMPTY_CHANGELOG_FAILED,
- "Error detecting empty changelog");
- }
- sys_close (priv->changelog_fd);
- priv->changelog_fd = -1;
- }
-
- (void) snprintf (ofile, PATH_MAX,
- "%s/"CHANGELOG_FILE_NAME, priv->changelog_dir);
- (void) snprintf (nfile, PATH_MAX,
- "%s/"CHANGELOG_FILE_NAME".%lu",
- priv->changelog_dir, ts);
-
- if (cl_empty_flag == 1) {
- ret = sys_unlink (ofile);
- if (ret) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_UNLINK_OP_FAILED,
- "error unlinking empty changelog",
- "path=%s", ofile,
- NULL);
- ret = 0; /* Error in unlinking empty changelog should
- not break further changelog operation, so
- reset return value to 0*/
- }
- } else {
- ret = sys_rename (ofile, nfile);
+ int ret = -1;
+ int notify = 0;
+ int cl_empty_flag = 0;
+ struct tm *gmt;
+ char yyyymmdd[40];
+ char ofile[PATH_MAX] = {
+ 0,
+ };
+ char nfile[PATH_MAX] = {
+ 0,
+ };
+ char nfile_dir[PATH_MAX] = {
+ 0,
+ };
+ changelog_event_t ev = {
+ 0,
+ };
+
+ if (priv->changelog_fd != -1) {
+ ret = sys_fsync(priv->changelog_fd);
+ if (ret < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_FSYNC_OP_FAILED, NULL);
+ }
+ ret = cl_is_empty(this, priv->changelog_fd);
+ if (ret == 1) {
+ cl_empty_flag = 1;
+ } else if (ret == -1) {
+ /* Log error but proceed as usual */
+ gf_smsg(this->name, GF_LOG_WARNING, 0,
+ CHANGELOG_MSG_DETECT_EMPTY_CHANGELOG_FAILED, NULL);
+ }
+ sys_close(priv->changelog_fd);
+ priv->changelog_fd = -1;
+ }
+
+ /* Get GMT time. */
+ gmt = gmtime(&ts);
+
+ strftime(yyyymmdd, sizeof(yyyymmdd), "%Y/%m/%d", gmt);
+
+ (void)snprintf(ofile, PATH_MAX, "%s/" CHANGELOG_FILE_NAME,
+ priv->changelog_dir);
+ (void)snprintf(nfile, PATH_MAX, "%s/%s/" CHANGELOG_FILE_NAME ".%ld",
+ priv->changelog_dir, yyyymmdd, ts);
+ (void)snprintf(nfile_dir, PATH_MAX, "%s/%s", priv->changelog_dir, yyyymmdd);
+
+ if (cl_empty_flag == 1) {
+ ret = sys_unlink(ofile);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_UNLINK_OP_FAILED, "path=%s", ofile, NULL);
+ ret = 0; /* Error in unlinking empty changelog should
+ not break further changelog operation, so
+ reset return value to 0*/
+ }
+ } else {
+ ret = sys_rename(ofile, nfile);
+
+ /* Changelog file rename gets ENOENT when parent dir doesn't exist */
+ if (errno == ENOENT) {
+ ret = mkdir_p(nfile_dir, 0600, _gf_true);
+
+ if ((ret == -1) && (EEXIST != errno)) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_MKDIR_ERROR, "%s", nfile_dir, NULL);
+ goto out;
+ }
- if (ret && (errno == ENOENT)) {
- ret = 0;
- goto out;
- }
- if (ret) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_RENAME_ERROR,
- "error renaming",
- "from=%s", ofile,
- "to=%s", nfile,
- NULL);
- }
+ ret = sys_rename(ofile, nfile);
}
- if (!ret && (cl_empty_flag == 0)) {
- notify = 1;
+ if (ret && (errno == ENOENT)) {
+ ret = 0;
+ goto out;
}
-
- if (!ret) {
- if (cl_empty_flag) {
- update_path (this, nfile);
- }
- ret = htime_update (this, priv, ts, nfile);
- if (ret == -1) {
- gf_msg (this->name, GF_LOG_ERROR,
- 0, CHANGELOG_MSG_HTIME_ERROR,
- "could not update htime file");
- goto out;
- }
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_RENAME_ERROR,
+ "from=%s", ofile, "to=%s", nfile, NULL);
}
+ }
+
+ if (!ret && (cl_empty_flag == 0)) {
+ notify = 1;
+ }
- if (notify) {
- ev.ev_type = CHANGELOG_OP_TYPE_JOURNAL;
- memcpy (ev.u.journal.path, nfile, strlen (nfile) + 1);
- changelog_dispatch_event (this, priv, &ev);
+ if (!ret) {
+ if (cl_empty_flag) {
+ update_path(this, nfile);
}
- out:
- /* If this is explicit rollover initiated by snapshot,
- * wakeup reconfigure thread waiting for changelog to
- * rollover. This should happen even in failure cases as
- * well otherwise snapshot will timeout and fail. Hence
- * moved under out.
- */
- if (priv->explicit_rollover) {
- priv->explicit_rollover = _gf_false;
-
- pthread_mutex_lock (&priv->bn.bnotify_mutex);
- {
- if (ret) {
- priv->bn.bnotify_error = _gf_true;
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_EXPLICIT_ROLLOVER_FAILED,
- "Fail snapshot because of "
- "previous errors");
- } else {
- gf_smsg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_MSG_BNOTIFY_INFO, "Explicit "
- "rollover changelog signaling "
- "bnotify",
- "changelog=%s", nfile,
- NULL);
- }
- priv->bn.bnotify = _gf_false;
- pthread_cond_signal (&priv->bn.bnotify_cond);
- }
- pthread_mutex_unlock (&priv->bn.bnotify_mutex);
+ ret = htime_update(this, priv, ts, nfile);
+ if (ret == -1) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_HTIME_ERROR,
+ NULL);
+ goto out;
}
- return ret;
+ }
+
+ if (notify) {
+ ev.ev_type = CHANGELOG_OP_TYPE_JOURNAL;
+ memcpy(ev.u.journal.path, nfile, strlen(nfile) + 1);
+ changelog_dispatch_event(this, priv, &ev);
+ }
+out:
+ /* If this is explicit rollover initiated by snapshot,
+ * wakeup reconfigure thread waiting for changelog to
+ * rollover. This should happen even in failure cases as
+ * well otherwise snapshot will timeout and fail. Hence
+ * moved under out.
+ */
+ if (priv->explicit_rollover) {
+ priv->explicit_rollover = _gf_false;
+
+ pthread_mutex_lock(&priv->bn.bnotify_mutex);
+ {
+ if (ret) {
+ priv->bn.bnotify_error = _gf_true;
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_MSG_EXPLICIT_ROLLOVER_FAILED, NULL);
+ } else {
+ gf_smsg(this->name, GF_LOG_INFO, 0, CHANGELOG_MSG_BNOTIFY_INFO,
+ "changelog=%s", nfile, NULL);
+ }
+ priv->bn.bnotify = _gf_false;
+ pthread_cond_signal(&priv->bn.bnotify_cond);
+ }
+ pthread_mutex_unlock(&priv->bn.bnotify_mutex);
+ }
+ return ret;
}
int
-filter_cur_par_dirs (const struct dirent *entry)
+filter_cur_par_dirs(const struct dirent *entry)
{
- if (entry == NULL)
- return 0;
+ if (entry == NULL)
+ return 0;
- if ((strcmp(entry->d_name, ".") == 0) ||
- (strcmp(entry->d_name, "..") == 0))
- return 0;
- else
- return 1;
+ if ((strcmp(entry->d_name, ".") == 0) || (strcmp(entry->d_name, "..") == 0))
+ return 0;
+ else
+ return 1;
}
/*
@@ -556,295 +536,284 @@ filter_cur_par_dirs (const struct dirent *entry)
*/
int
-find_current_htime (int ht_dir_fd, const char *ht_dir_path, char *ht_file_bname)
+find_current_htime(int ht_dir_fd, const char *ht_dir_path, char *ht_file_bname)
{
- struct dirent **namelist = NULL;
- int ret = 0;
- int cnt = 0;
- int i = 0;
- xlator_t *this = NULL;
-
- this = THIS;
- GF_ASSERT (this);
- GF_ASSERT (ht_dir_path);
-
- cnt = scandir (ht_dir_path, &namelist, filter_cur_par_dirs, alphasort);
- if (cnt < 0) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_SCAN_DIR_FAILED,
- "scandir failed");
- } else if (cnt > 0) {
- strncpy (ht_file_bname, namelist[cnt - 1]->d_name, NAME_MAX);
- ht_file_bname[NAME_MAX - 1] = 0;
-
- if (sys_fsetxattr (ht_dir_fd, HTIME_CURRENT, ht_file_bname,
- strlen (ht_file_bname), 0)) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_FSETXATTR_FAILED,
- "fsetxattr failed: HTIME_CURRENT");
- ret = -1;
- goto out;
- }
-
- if (sys_fsync (ht_dir_fd) < 0) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_FSYNC_OP_FAILED,
- "fsync failed");
- ret = -1;
- goto out;
- }
- }
+ struct dirent **namelist = NULL;
+ int ret = 0;
+ int cnt = 0;
+ int i = 0;
+ xlator_t *this = NULL;
+
+ this = THIS;
+ GF_ASSERT(this);
+ GF_ASSERT(ht_dir_path);
+
+ cnt = scandir(ht_dir_path, &namelist, filter_cur_par_dirs, alphasort);
+ if (cnt < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_SCAN_DIR_FAILED,
+ NULL);
+ } else if (cnt > 0) {
+ if (snprintf(ht_file_bname, NAME_MAX, "%s",
+ namelist[cnt - 1]->d_name) >= NAME_MAX) {
+ ret = -1;
+ goto out;
+ }
+ if (sys_fsetxattr(ht_dir_fd, HTIME_CURRENT, ht_file_bname,
+ strlen(ht_file_bname), 0)) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_FSETXATTR_FAILED, "HTIME_CURRENT", NULL);
+ ret = -1;
+ goto out;
+ }
+
+ if (sys_fsync(ht_dir_fd) < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_FSYNC_OP_FAILED, NULL);
+ ret = -1;
+ goto out;
+ }
+ }
- out:
- for (i = 0; i < cnt; i++)
- free (namelist[i]);
- free (namelist);
+out:
+ for (i = 0; i < cnt; i++)
+ free(namelist[i]);
+ free(namelist);
- if (ret)
- cnt = ret;
+ if (ret)
+ cnt = ret;
- return cnt;
+ return cnt;
}
/* Returns 0 on successful open of htime file
* returns -1 on failure or error
*/
int
-htime_open (xlator_t *this,
- changelog_priv_t *priv, unsigned long ts)
+htime_open(xlator_t *this, changelog_priv_t *priv, time_t ts)
{
- int ht_file_fd = -1;
- int ht_dir_fd = -1;
- int ret = 0;
- int cnt = 0;
- char ht_dir_path[PATH_MAX] = {0,};
- char ht_file_path[PATH_MAX] = {0,};
- char ht_file_bname[NAME_MAX] = {0,};
- char x_value[NAME_MAX] = {0,};
- int flags = 0;
- unsigned long min_ts = 0;
- unsigned long max_ts = 0;
- unsigned long total = 0;
- unsigned long total1 = 0;
- ssize_t size = 0;
- struct stat stat_buf = {0,};
- unsigned long record_len = 0;
-
- CHANGELOG_FILL_HTIME_DIR(priv->changelog_dir, ht_dir_path);
-
- /* Open htime directory to get HTIME_CURRENT */
- ht_dir_fd = open (ht_dir_path, O_RDONLY);
- if (ht_dir_fd == -1) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_OPEN_FAILED, "open failed",
- "path=%s", ht_dir_path,
- NULL);
- ret = -1;
- goto out;
- }
-
- size = sys_fgetxattr (ht_dir_fd, HTIME_CURRENT, ht_file_bname,
- sizeof (ht_file_bname));
- if (size < 0) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_FGETXATTR_FAILED, "Error extracting"
- " HTIME_CURRENT.");
-
- /* If upgrade scenario, find the latest HTIME.TSTAMP file
- * and use the same. If error, create a new HTIME.TSTAMP
- * file.
- */
- cnt = find_current_htime (ht_dir_fd, ht_dir_path,
- ht_file_bname);
- if (cnt <= 0) {
- gf_msg (this->name, GF_LOG_INFO, errno,
- CHANGELOG_MSG_HTIME_INFO,
- "HTIME_CURRENT not found. Changelog enabled"
- " before init");
- return htime_create (this, priv, ts);
- }
-
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_HTIME_ERROR, "Error extracting"
- " HTIME_CURRENT.");
- }
-
- gf_smsg (this->name, GF_LOG_INFO, 0, CHANGELOG_MSG_HTIME_INFO,
- "HTIME_CURRENT",
- "path=%s", ht_file_bname, NULL);
- (void) snprintf (ht_file_path, PATH_MAX, "%s/%s",
- ht_dir_path, ht_file_bname);
-
- /* Open in append mode as existing htime file is used */
- flags |= (O_RDWR | O_SYNC | O_APPEND);
- ht_file_fd = open (ht_file_path, flags,
- S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
- if (ht_file_fd < 0) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_OPEN_FAILED,
- "unable to open htime file",
- "path=%s", ht_file_path,
- NULL);
- ret = -1;
- goto out;
- }
-
- /* save this htime_fd in priv->htime_fd */
- priv->htime_fd = ht_file_fd;
-
- ret = sys_fstat (ht_file_fd, &stat_buf);
- if (ret < 0) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_HTIME_ERROR,
- "unable to stat htime file",
- "path=%s", ht_file_path,
- NULL);
- ret = -1;
- goto out;
- }
-
- /* Initialize rollover-number in priv to current number */
- size = sys_fgetxattr (ht_file_fd, HTIME_KEY, x_value, sizeof (x_value));
- if (size < 0) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_FGETXATTR_FAILED, "error extracting max"
- " timstamp from htime file",
- "path=%s", ht_file_path,
- NULL);
- ret = -1;
- goto out;
- }
-
- sscanf (x_value, "%lu:%lu", &max_ts, &total);
-
- /* 22 = 1(/) + 20(CHANGELOG.TIMESTAMP) + 1(\x00) */
- record_len = strlen(priv->changelog_dir) + 22;
- total1 = stat_buf.st_size/record_len;
- if (total != total1) {
- gf_smsg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_MSG_TOTAL_LOG_INFO,
- "Mismatch of changelog count. "
- "INIT CASE",
- "xattr_total=%lu", total,
- "size_total=%lu", total1,
- NULL);
- }
-
- gf_smsg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_MSG_TOTAL_LOG_INFO,
- "INIT CASE",
- "min=%lu", min_ts,
- "max=%lu", max_ts,
- "total_changelogs=%lu", total,
- NULL);
-
- if (total < total1)
- priv->rollover_count = total1 + 1;
- else
- priv->rollover_count = total + 1;
+ int ht_file_fd = -1;
+ int ht_dir_fd = -1;
+ int ret = 0;
+ int cnt = 0;
+ char ht_dir_path[PATH_MAX] = {
+ 0,
+ };
+ char ht_file_path[PATH_MAX] = {
+ 0,
+ };
+ char ht_file_bname[NAME_MAX] = {
+ 0,
+ };
+ char x_value[NAME_MAX] = {
+ 0,
+ };
+ int flags = 0;
+ unsigned long min_ts = 0;
+ unsigned long max_ts = 0;
+ unsigned long total = 0;
+ unsigned long total1 = 0;
+ ssize_t size = 0;
+ struct stat stat_buf = {
+ 0,
+ };
+ unsigned long record_len = 0;
+ int32_t len = 0;
+
+ CHANGELOG_FILL_HTIME_DIR(priv->changelog_dir, ht_dir_path);
+
+ /* Open htime directory to get HTIME_CURRENT */
+ ht_dir_fd = open(ht_dir_path, O_RDONLY);
+ if (ht_dir_fd == -1) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_OPEN_FAILED,
+ "path=%s", ht_dir_path, NULL);
+ ret = -1;
+ goto out;
+ }
+
+ size = sys_fgetxattr(ht_dir_fd, HTIME_CURRENT, ht_file_bname,
+ sizeof(ht_file_bname));
+ if (size < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_FGETXATTR_FAILED,
+ "name=HTIME_CURRENT", NULL);
+
+ /* If upgrade scenario, find the latest HTIME.TSTAMP file
+ * and use the same. If error, create a new HTIME.TSTAMP
+ * file.
+ */
+ cnt = find_current_htime(ht_dir_fd, ht_dir_path, ht_file_bname);
+ if (cnt <= 0) {
+ gf_smsg(this->name, GF_LOG_INFO, errno,
+ CHANGELOG_MSG_NO_HTIME_CURRENT, NULL);
+ sys_close(ht_dir_fd);
+ return htime_create(this, priv, ts);
+ }
+
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_HTIME_CURRENT_ERROR, NULL);
+ }
+
+ gf_smsg(this->name, GF_LOG_INFO, 0, CHANGELOG_MSG_HTIME_CURRENT, "path=%s",
+ ht_file_bname, NULL);
+ len = snprintf(ht_file_path, PATH_MAX, "%s/%s", ht_dir_path, ht_file_bname);
+ if ((len < 0) || (len >= PATH_MAX)) {
+ ret = -1;
+ goto out;
+ }
+
+ /* Open in append mode as existing htime file is used */
+ flags |= (O_RDWR | O_SYNC | O_APPEND);
+ ht_file_fd = open(ht_file_path, flags,
+ S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
+ if (ht_file_fd < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_OPEN_FAILED,
+ "path=%s", ht_file_path, NULL);
+ ret = -1;
+ goto out;
+ }
+
+ /* save this htime_fd in priv->htime_fd */
+ priv->htime_fd = ht_file_fd;
+
+ ret = sys_fstat(ht_file_fd, &stat_buf);
+ if (ret < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_HTIME_STAT_ERROR,
+ "path=%s", ht_file_path, NULL);
+ ret = -1;
+ goto out;
+ }
+
+ /* Initialize rollover-number in priv to current number */
+ size = sys_fgetxattr(ht_file_fd, HTIME_KEY, x_value, sizeof(x_value));
+ if (size < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_FGETXATTR_FAILED,
+ "name=%s", HTIME_KEY, "path=%s", ht_file_path, NULL);
+ ret = -1;
+ goto out;
+ }
+
+ sscanf(x_value, "%lu:%lu", &max_ts, &total);
+
+ /* 22 = 1(/) + 20(CHANGELOG.TIMESTAMP) + 1(\x00) */
+ record_len = strlen(priv->changelog_dir) + 22;
+ total1 = stat_buf.st_size / record_len;
+ if (total != total1) {
+ gf_smsg(this->name, GF_LOG_INFO, 0, CHANGELOG_MSG_TOTAL_LOG_INFO,
+ "xattr_total=%lu", total, "size_total=%lu", total1, NULL);
+ }
+
+ gf_smsg(this->name, GF_LOG_INFO, 0, CHANGELOG_MSG_TOTAL_LOG_INFO, "min=%lu",
+ min_ts, "max=%lu", max_ts, "total_changelogs=%lu", total, NULL);
+
+ if (total < total1)
+ priv->rollover_count = total1 + 1;
+ else
+ priv->rollover_count = total + 1;
out:
- if (ht_dir_fd != -1)
- sys_close (ht_dir_fd);
- return ret;
+ if (ht_dir_fd != -1)
+ sys_close(ht_dir_fd);
+ return ret;
}
/* Returns 0 on successful creation of htime file
* returns -1 on failure or error
*/
int
-htime_create (xlator_t *this,
- changelog_priv_t *priv, unsigned long ts)
+htime_create(xlator_t *this, changelog_priv_t *priv, time_t ts)
{
- int ht_file_fd = -1;
- int ht_dir_fd = -1;
- int ret = 0;
- char ht_dir_path[PATH_MAX] = {0,};
- char ht_file_path[PATH_MAX] = {0,};
- char ht_file_bname[NAME_MAX + 1] = {0,};
- int flags = 0;
-
- gf_smsg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_MSG_HTIME_INFO, "Changelog enable: Creating new "
- "HTIME file",
- "name=%lu", ts,
- NULL);
-
- CHANGELOG_FILL_HTIME_DIR(priv->changelog_dir, ht_dir_path);
-
- /* get the htime file name in ht_file_path */
- (void) snprintf (ht_file_path,PATH_MAX,"%s/%s.%lu",ht_dir_path,
- HTIME_FILE_NAME, ts);
-
- flags |= (O_CREAT | O_RDWR | O_SYNC);
- ht_file_fd = open (ht_file_path, flags,
- S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
- if (ht_file_fd < 0) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_OPEN_FAILED,
- "unable to create htime file",
- "path=%s", ht_file_path,
- NULL);
- ret = -1;
- goto out;
- }
-
- if (sys_fsetxattr (ht_file_fd, HTIME_KEY, HTIME_INITIAL_VALUE,
- sizeof (HTIME_INITIAL_VALUE)-1, 0)) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_FSETXATTR_FAILED,
- "Htime xattr initialization failed");
- ret = -1;
- goto out;
- }
-
- ret = sys_fsync (ht_file_fd);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_FSYNC_OP_FAILED,
- "fsync failed");
- goto out;
- }
-
- /* Set xattr HTIME_CURRENT on htime directory to htime filename */
- ht_dir_fd = open (ht_dir_path, O_RDONLY);
- if (ht_dir_fd == -1) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_OPEN_FAILED, "open failed",
- "path=%s", ht_dir_path,
- NULL);
- ret = -1;
- goto out;
- }
-
- (void) snprintf (ht_file_bname, sizeof (ht_file_bname), "%s.%lu",
- HTIME_FILE_NAME, ts);
- if (sys_fsetxattr (ht_dir_fd, HTIME_CURRENT, ht_file_bname,
- strlen (ht_file_bname), 0)) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_FSETXATTR_FAILED, "fsetxattr failed:"
- " HTIME_CURRENT");
- ret = -1;
- goto out;
- }
-
- ret = sys_fsync (ht_dir_fd);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_FSYNC_OP_FAILED,
- "fsync failed");
- goto out;
- }
-
- /* save this htime_fd in priv->htime_fd */
- priv->htime_fd = ht_file_fd;
- /* initialize rollover-number in priv to 1 */
- priv->rollover_count = 1;
+ int ht_file_fd = -1;
+ int ht_dir_fd = -1;
+ int ret = 0;
+ char ht_dir_path[PATH_MAX] = {
+ 0,
+ };
+ char ht_file_path[PATH_MAX] = {
+ 0,
+ };
+ char ht_file_bname[NAME_MAX + 1] = {
+ 0,
+ };
+ int flags = 0;
+ int32_t len = 0;
+
+ gf_smsg(this->name, GF_LOG_INFO, 0, CHANGELOG_MSG_NEW_HTIME_FILE,
+ "name=%ld", ts, NULL);
+
+ CHANGELOG_FILL_HTIME_DIR(priv->changelog_dir, ht_dir_path);
+
+ /* get the htime file name in ht_file_path */
+ len = snprintf(ht_file_path, PATH_MAX, "%s/%s.%ld", ht_dir_path,
+ HTIME_FILE_NAME, ts);
+ if ((len < 0) || (len >= PATH_MAX)) {
+ ret = -1;
+ goto out;
+ }
+
+ flags |= (O_CREAT | O_RDWR | O_SYNC);
+ ht_file_fd = open(ht_file_path, flags,
+ S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
+ if (ht_file_fd < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_OPEN_FAILED,
+ "path=%s", ht_file_path, NULL);
+ ret = -1;
+ goto out;
+ }
+
+ if (sys_fsetxattr(ht_file_fd, HTIME_KEY, HTIME_INITIAL_VALUE,
+ sizeof(HTIME_INITIAL_VALUE) - 1, 0)) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_XATTR_INIT_FAILED, NULL);
+ ret = -1;
+ goto out;
+ }
+
+ ret = sys_fsync(ht_file_fd);
+ if (ret < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_FSYNC_OP_FAILED,
+ NULL);
+ goto out;
+ }
+
+ /* save this htime_fd in priv->htime_fd */
+ priv->htime_fd = ht_file_fd;
+
+ ht_file_fd = -1;
+
+ /* Set xattr HTIME_CURRENT on htime directory to htime filename */
+ ht_dir_fd = open(ht_dir_path, O_RDONLY);
+ if (ht_dir_fd == -1) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_OPEN_FAILED,
+ "path=%s", ht_dir_path, NULL);
+ ret = -1;
+ goto out;
+ }
+
+ (void)snprintf(ht_file_bname, sizeof(ht_file_bname), "%s.%ld",
+ HTIME_FILE_NAME, ts);
+ if (sys_fsetxattr(ht_dir_fd, HTIME_CURRENT, ht_file_bname,
+ strlen(ht_file_bname), 0)) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_FSETXATTR_FAILED,
+ " HTIME_CURRENT", NULL);
+ ret = -1;
+ goto out;
+ }
+
+ ret = sys_fsync(ht_dir_fd);
+ if (ret < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_FSYNC_OP_FAILED,
+ NULL);
+ goto out;
+ }
+
+ /* initialize rollover-number in priv to 1 */
+ priv->rollover_count = 1;
out:
- if (ht_dir_fd != -1)
- sys_close (ht_dir_fd);
- return ret;
+ if (ht_dir_fd != -1)
+ sys_close(ht_dir_fd);
+ if (ht_file_fd != -1)
+ sys_close(ht_file_fd);
+ return ret;
}
/* Description:
@@ -856,49 +825,53 @@ out:
* -1 : On failure.
*/
int
-changelog_snap_open (xlator_t *this,
- changelog_priv_t *priv)
+changelog_snap_open(xlator_t *this, changelog_priv_t *priv)
{
- int fd = -1;
- int ret = 0;
- int flags = 0;
- char buffer[1024] = {0,};
- char c_snap_path[PATH_MAX] = {0,};
- char csnap_dir_path[PATH_MAX] = {0,};
-
- CHANGELOG_FILL_CSNAP_DIR(priv->changelog_dir, csnap_dir_path);
-
- (void) snprintf (c_snap_path, PATH_MAX,
- "%s/"CSNAP_FILE_NAME,
- csnap_dir_path);
-
- flags |= (O_CREAT | O_RDWR | O_TRUNC);
-
- fd = open (c_snap_path, flags,
- S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
- if (fd < 0) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_OPEN_FAILED, "unable to open file",
- "path=%s", c_snap_path,
- NULL);
- ret = -1;
- goto out;
- }
- priv->c_snap_fd = fd;
-
- (void) snprintf (buffer, 1024, CHANGELOG_HEADER,
- CHANGELOG_VERSION_MAJOR,
- CHANGELOG_VERSION_MINOR,
- priv->ce->encoder);
- ret = changelog_snap_write_change (priv, buffer, strlen (buffer));
- if (ret < 0) {
- sys_close (priv->c_snap_fd);
- priv->c_snap_fd = -1;
- goto out;
- }
+ int fd = -1;
+ int ret = 0;
+ int flags = 0;
+ char buffer[1024] = {
+ 0,
+ };
+ char c_snap_path[PATH_MAX] = {
+ 0,
+ };
+ char csnap_dir_path[PATH_MAX] = {
+ 0,
+ };
+ int32_t len = 0;
+
+ CHANGELOG_FILL_CSNAP_DIR(priv->changelog_dir, csnap_dir_path);
+
+ len = snprintf(c_snap_path, PATH_MAX, "%s/" CSNAP_FILE_NAME,
+ csnap_dir_path);
+ if ((len < 0) || (len >= PATH_MAX)) {
+ ret = -1;
+ goto out;
+ }
+
+ flags |= (O_CREAT | O_RDWR | O_TRUNC);
+
+ fd = open(c_snap_path, flags, S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
+ if (fd < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_OPEN_FAILED,
+ "path=%s", c_snap_path, NULL);
+ ret = -1;
+ goto out;
+ }
+ priv->c_snap_fd = fd;
+
+ (void)snprintf(buffer, 1024, CHANGELOG_HEADER, CHANGELOG_VERSION_MAJOR,
+ CHANGELOG_VERSION_MINOR, priv->ce->encoder);
+ ret = changelog_snap_write_change(priv, buffer, strlen(buffer));
+ if (ret < 0) {
+ sys_close(priv->c_snap_fd);
+ priv->c_snap_fd = -1;
+ goto out;
+ }
out:
- return ret;
+ return ret;
}
/*
@@ -909,17 +882,15 @@ out:
* -1 : On Failure.
*/
int
-changelog_snap_logging_start (xlator_t *this,
- changelog_priv_t *priv)
+changelog_snap_logging_start(xlator_t *this, changelog_priv_t *priv)
{
- int ret = 0;
+ int ret = 0;
- ret = changelog_snap_open (this, priv);
- gf_msg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_MSG_SNAP_INFO,
- "Now starting to log in call path");
+ ret = changelog_snap_open(this, priv);
+ gf_smsg(this->name, GF_LOG_INFO, 0, CHANGELOG_MSG_SNAP_INFO, "starting",
+ NULL);
- return ret;
+ return ret;
}
/*
@@ -930,120 +901,104 @@ changelog_snap_logging_start (xlator_t *this,
* -1 : On Failure.
*/
int
-changelog_snap_logging_stop (xlator_t *this,
- changelog_priv_t *priv)
+changelog_snap_logging_stop(xlator_t *this, changelog_priv_t *priv)
{
- int ret = 0;
+ int ret = 0;
- sys_close (priv->c_snap_fd);
- priv->c_snap_fd = -1;
+ sys_close(priv->c_snap_fd);
+ priv->c_snap_fd = -1;
- gf_msg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_MSG_SNAP_INFO,
- "Stopped to log in call path");
+ gf_smsg(this->name, GF_LOG_INFO, 0, CHANGELOG_MSG_SNAP_INFO, "Stopped",
+ NULL);
- return ret;
+ return ret;
}
int
-changelog_open_journal (xlator_t *this,
- changelog_priv_t *priv)
+changelog_open_journal(xlator_t *this, changelog_priv_t *priv)
{
- int fd = 0;
- int ret = -1;
- int flags = 0;
- char buffer[1024] = {0,};
- char changelog_path[PATH_MAX] = {0,};
-
- (void) snprintf (changelog_path, PATH_MAX,
- "%s/"CHANGELOG_FILE_NAME,
- priv->changelog_dir);
-
- flags |= (O_CREAT | O_RDWR);
- if (priv->fsync_interval == 0)
- flags |= O_SYNC;
-
- fd = open (changelog_path, flags,
- S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
- if (fd < 0) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_OPEN_FAILED,
- "unable to open/create changelog file."
- " change-logging will be"
- " inactive",
- "path=%s", changelog_path,
- NULL);
- goto out;
- }
-
- priv->changelog_fd = fd;
+ int fd = 0;
+ int ret = -1;
+ int flags = 0;
+ char buffer[1024] = {
+ 0,
+ };
+ char changelog_path[PATH_MAX] = {
+ 0,
+ };
+
+ (void)snprintf(changelog_path, PATH_MAX, "%s/" CHANGELOG_FILE_NAME,
+ priv->changelog_dir);
+
+ flags |= (O_CREAT | O_RDWR);
+ if (priv->fsync_interval == 0)
+ flags |= O_SYNC;
+
+ fd = open(changelog_path, flags, S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
+ if (fd < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_OPEN_FAILED,
+ "path=%s", changelog_path, NULL);
+ goto out;
+ }
+
+ priv->changelog_fd = fd;
+
+ (void)snprintf(buffer, 1024, CHANGELOG_HEADER, CHANGELOG_VERSION_MAJOR,
+ CHANGELOG_VERSION_MINOR, priv->ce->encoder);
+ ret = changelog_write_change(priv, buffer, strlen(buffer));
+ if (ret) {
+ sys_close(priv->changelog_fd);
+ priv->changelog_fd = -1;
+ goto out;
+ }
+
+ ret = 0;
- (void) snprintf (buffer, 1024, CHANGELOG_HEADER,
- CHANGELOG_VERSION_MAJOR,
- CHANGELOG_VERSION_MINOR,
- priv->ce->encoder);
- ret = changelog_write_change (priv, buffer, strlen (buffer));
- if (ret) {
- sys_close (priv->changelog_fd);
- priv->changelog_fd = -1;
- goto out;
- }
-
- ret = 0;
-
- out:
- return ret;
+out:
+ return ret;
}
int
-changelog_start_next_change (xlator_t *this,
- changelog_priv_t *priv,
- unsigned long ts, gf_boolean_t finale)
+changelog_start_next_change(xlator_t *this, changelog_priv_t *priv, time_t ts,
+ gf_boolean_t finale)
{
- int ret = -1;
+ int ret = -1;
- ret = changelog_rollover_changelog (this, priv, ts);
+ ret = changelog_rollover_changelog(this, priv, ts);
- if (!ret && !finale)
- ret = changelog_open_journal (this, priv);
+ if (!ret && !finale)
+ ret = changelog_open_journal(this, priv);
- return ret;
+ return ret;
}
/**
* return the length of entry
*/
size_t
-changelog_entry_length ()
+changelog_entry_length()
{
- return sizeof (changelog_log_data_t);
+ return sizeof(changelog_log_data_t);
}
-int
-changelog_fill_rollover_data (changelog_log_data_t *cld, gf_boolean_t is_last)
+void
+changelog_fill_rollover_data(changelog_log_data_t *cld, gf_boolean_t is_last)
{
- struct timeval tv = {0,};
-
- cld->cld_type = CHANGELOG_TYPE_ROLLOVER;
-
- if (gettimeofday (&tv, NULL))
- return -1;
-
- cld->cld_roll_time = (unsigned long) tv.tv_sec;
- cld->cld_finale = is_last;
- return 0;
+ cld->cld_type = CHANGELOG_TYPE_ROLLOVER;
+ cld->cld_roll_time = gf_time();
+ cld->cld_finale = is_last;
}
int
-changelog_snap_write_change (changelog_priv_t *priv, char *buffer, size_t len)
+changelog_snap_write_change(changelog_priv_t *priv, char *buffer, size_t len)
{
- return changelog_write (priv->c_snap_fd, buffer, len);
+ return changelog_write(priv->c_snap_fd, buffer, len);
}
int
-changelog_write_change (changelog_priv_t *priv, char *buffer, size_t len)
+changelog_write_change(changelog_priv_t *priv, char *buffer, size_t len)
{
- return changelog_write (priv->changelog_fd, buffer, len);
+ return changelog_write(priv->changelog_fd, buffer, len);
}
/*
@@ -1056,255 +1011,230 @@ changelog_write_change (changelog_priv_t *priv, char *buffer, size_t len)
* -1 : On Failure.
*/
int
-changelog_snap_handle_ascii_change (xlator_t *this,
- changelog_log_data_t *cld)
+changelog_snap_handle_ascii_change(xlator_t *this, changelog_log_data_t *cld)
{
- size_t off = 0;
- size_t gfid_len = 0;
- char *gfid_str = NULL;
- char *buffer = NULL;
- changelog_priv_t *priv = NULL;
- int ret = 0;
-
- if (this == NULL) {
- ret = -1;
- goto out;
- }
-
- priv = this->private;
-
- if (priv == NULL) {
- ret = -1;
- goto out;
- }
-
- gfid_str = uuid_utoa (cld->cld_gfid);
- gfid_len = strlen (gfid_str);
-
- /* extra bytes for decorations */
- buffer = alloca (gfid_len + cld->cld_ptr_len + 10);
- CHANGELOG_STORE_ASCII (priv, buffer,
- off, gfid_str, gfid_len, cld);
-
- CHANGELOG_FILL_BUFFER (buffer, off, "\0", 1);
-
- ret = changelog_snap_write_change (priv, buffer, off);
-
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_WRITE_FAILED,
- "error writing csnap to disk");
- }
- gf_msg (this->name, GF_LOG_INFO, 0, CHANGELOG_MSG_SNAP_INFO,
- "Successfully wrote to csnap");
- ret = 0;
+ size_t off = 0;
+ size_t gfid_len = 0;
+ char *gfid_str = NULL;
+ char *buffer = NULL;
+ changelog_priv_t *priv = NULL;
+ int ret = 0;
+
+ if (this == NULL) {
+ ret = -1;
+ goto out;
+ }
+
+ priv = this->private;
+
+ if (priv == NULL) {
+ ret = -1;
+ goto out;
+ }
+
+ gfid_str = uuid_utoa(cld->cld_gfid);
+ gfid_len = strlen(gfid_str);
+
+ /* extra bytes for decorations */
+ buffer = alloca(gfid_len + cld->cld_ptr_len + 10);
+ CHANGELOG_STORE_ASCII(priv, buffer, off, gfid_str, gfid_len, cld);
+
+ CHANGELOG_FILL_BUFFER(buffer, off, "\0", 1);
+
+ ret = changelog_snap_write_change(priv, buffer, off);
+
+ if (ret < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_WRITE_FAILED,
+ "csnap", NULL);
+ }
+ gf_smsg(this->name, GF_LOG_INFO, 0, CHANGELOG_MSG_WROTE_TO_CSNAP, NULL);
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-changelog_handle_change (xlator_t *this,
- changelog_priv_t *priv, changelog_log_data_t *cld)
+changelog_handle_change(xlator_t *this, changelog_priv_t *priv,
+ changelog_log_data_t *cld)
{
- int ret = 0;
-
- if (CHANGELOG_TYPE_IS_ROLLOVER (cld->cld_type)) {
- changelog_encode_change (priv);
- ret = changelog_start_next_change (this, priv,
- cld->cld_roll_time,
- cld->cld_finale);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_GET_TIME_OP_FAILED,
- "Problem rolling over changelog(s)");
- goto out;
- }
+ int ret = 0;
- /**
- * case when there is reconfigure done (disabling changelog) and there
- * are still fops that have updates in prgress.
- */
- if (priv->changelog_fd == -1)
- return 0;
-
- if (CHANGELOG_TYPE_IS_FSYNC (cld->cld_type)) {
- ret = sys_fsync (priv->changelog_fd);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_FSYNC_OP_FAILED,
- "fsync failed");
- }
- goto out;
- }
+ if (CHANGELOG_TYPE_IS_ROLLOVER(cld->cld_type)) {
+ changelog_encode_change(priv);
+ ret = changelog_start_next_change(this, priv, cld->cld_roll_time,
+ cld->cld_finale);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_MSG_GET_TIME_OP_FAILED, NULL);
+ goto out;
+ }
+
+ /**
+ * case when there is reconfigure done (disabling changelog) and there
+ * are still fops that have updates in prgress.
+ */
+ if (priv->changelog_fd == -1)
+ return 0;
- ret = priv->ce->encode (this, cld);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_WRITE_FAILED,
- "error writing changelog to disk");
+ if (CHANGELOG_TYPE_IS_FSYNC(cld->cld_type)) {
+ ret = sys_fsync(priv->changelog_fd);
+ if (ret < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_FSYNC_OP_FAILED, NULL);
}
+ goto out;
+ }
- out:
- return ret;
+ ret = priv->ce->encode(this, cld);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_WRITE_FAILED,
+ "changelog", NULL);
+ }
+
+out:
+ return ret;
}
changelog_local_t *
-changelog_local_init (xlator_t *this, inode_t *inode,
- uuid_t gfid, int xtra_records,
- gf_boolean_t update_flag)
+changelog_local_init(xlator_t *this, inode_t *inode, uuid_t gfid,
+ int xtra_records, gf_boolean_t update_flag)
{
- changelog_local_t *local = NULL;
- struct iobuf *iobuf = NULL;
+ changelog_local_t *local = NULL;
+ struct iobuf *iobuf = NULL;
+
+ /**
+ * We relax the presence of inode if @update_flag is true.
+ * The caller (implementation of the fop) needs to be careful to
+ * not blindly use local->inode.
+ */
+ if (!update_flag && !inode) {
+ gf_msg_callingfn(this->name, GF_LOG_WARNING, 0,
+ CHANGELOG_MSG_INODE_NOT_FOUND,
+ "inode needed for version checking !!!");
- /**
- * We relax the presence of inode if @update_flag is true.
- * The caller (implmentation of the fop) needs to be careful to
- * not blindly use local->inode.
- */
- if (!update_flag && !inode) {
- gf_msg_callingfn (this->name, GF_LOG_WARNING, 0,
- CHANGELOG_MSG_INODE_NOT_FOUND,
- "inode needed for version checking !!!");
- goto out;
- }
+ goto out;
+ }
- if (xtra_records) {
- iobuf = iobuf_get2 (this->ctx->iobuf_pool,
- xtra_records * CHANGELOG_OPT_RECORD_LEN);
- if (!iobuf)
- goto out;
- }
+ if (xtra_records) {
+ iobuf = iobuf_get2(this->ctx->iobuf_pool,
+ xtra_records * CHANGELOG_OPT_RECORD_LEN);
+ if (!iobuf)
+ goto out;
+ }
- local = mem_get0 (this->local_pool);
- if (!local) {
- CHANGELOG_IOBUF_UNREF (iobuf);
- goto out;
- }
+ local = mem_get0(this->local_pool);
+ if (!local) {
+ CHANGELOG_IOBUF_UNREF(iobuf);
+ goto out;
+ }
- local->update_no_check = update_flag;
+ local->update_no_check = update_flag;
- gf_uuid_copy (local->cld.cld_gfid, gfid);
+ gf_uuid_copy(local->cld.cld_gfid, gfid);
- local->cld.cld_iobuf = iobuf;
- local->cld.cld_xtra_records = 0; /* set by the caller */
+ local->cld.cld_iobuf = iobuf;
+ local->cld.cld_xtra_records = 0; /* set by the caller */
- if (inode)
- local->inode = inode_ref (inode);
+ if (inode)
+ local->inode = inode_ref(inode);
- out:
- return local;
+out:
+ return local;
}
int
-changelog_forget (xlator_t *this, inode_t *inode)
+changelog_forget(xlator_t *this, inode_t *inode)
{
- uint64_t ctx_addr = 0;
- changelog_inode_ctx_t *ctx = NULL;
+ uint64_t ctx_addr = 0;
+ changelog_inode_ctx_t *ctx = NULL;
- inode_ctx_del (inode, this, &ctx_addr);
- if (!ctx_addr)
- return 0;
+ inode_ctx_del(inode, this, &ctx_addr);
+ if (!ctx_addr)
+ return 0;
- ctx = (changelog_inode_ctx_t *) (long) ctx_addr;
- GF_FREE (ctx);
+ ctx = (changelog_inode_ctx_t *)(long)ctx_addr;
+ GF_FREE(ctx);
- return 0;
+ return 0;
}
int
-changelog_inject_single_event (xlator_t *this,
- changelog_priv_t *priv,
- changelog_log_data_t *cld)
+changelog_inject_single_event(xlator_t *this, changelog_priv_t *priv,
+ changelog_log_data_t *cld)
{
- return priv->cd.dispatchfn (this, priv, priv->cd.cd_data, cld, NULL);
+ return priv->cd.dispatchfn(this, priv, priv->cd.cd_data, cld, NULL);
}
/* Wait till all the black fops are drained */
void
-changelog_drain_black_fops (xlator_t *this, changelog_priv_t *priv)
+changelog_drain_black_fops(xlator_t *this, changelog_priv_t *priv)
{
- int ret = 0;
-
- /* clean up framework of pthread_mutex is required here as
- * 'reconfigure' terminates the changelog_rollover thread
- * on graph change.
- */
- pthread_cleanup_push (changelog_cleanup_free_mutex,
- &priv->dm.drain_black_mutex);
- ret = pthread_mutex_lock (&priv->dm.drain_black_mutex);
- if (ret)
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_PTHREAD_ERROR, "pthread error",
- "error=%d", ret,
- NULL);
- while (priv->dm.black_fop_cnt > 0) {
- gf_msg_debug (this->name, 0,
- "Condtional wait on black fops: %ld",
- priv->dm.black_fop_cnt);
- priv->dm.drain_wait_black = _gf_true;
- ret = pthread_cond_wait (&priv->dm.drain_black_cond,
- &priv->dm.drain_black_mutex);
- if (ret)
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_PTHREAD_COND_WAIT_FAILED,
- "pthread cond wait failed",
- "error=%d", ret,
- NULL);
- }
- priv->dm.drain_wait_black = _gf_false;
- ret = pthread_mutex_unlock (&priv->dm.drain_black_mutex);
+ int ret = 0;
+
+ /* clean up framework of pthread_mutex is required here as
+ * 'reconfigure' terminates the changelog_rollover thread
+ * on graph change.
+ */
+ pthread_cleanup_push(changelog_cleanup_free_mutex,
+ &priv->dm.drain_black_mutex);
+ ret = pthread_mutex_lock(&priv->dm.drain_black_mutex);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_PTHREAD_ERROR,
+ "error=%d", ret, NULL);
+ while (priv->dm.black_fop_cnt > 0) {
+ gf_msg_debug(this->name, 0, "Conditional wait on black fops: %ld",
+ priv->dm.black_fop_cnt);
+ priv->dm.drain_wait_black = _gf_true;
+ ret = pthread_cond_wait(&priv->dm.drain_black_cond,
+ &priv->dm.drain_black_mutex);
if (ret)
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_PTHREAD_ERROR, "pthread error",
- "error=%d", ret,
- NULL);
- pthread_cleanup_pop (0);
- gf_msg_debug (this->name, 0,
- "Woke up: Conditional wait on black fops");
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_PTHREAD_COND_WAIT_FAILED, "error=%d", ret,
+ NULL);
+ }
+ priv->dm.drain_wait_black = _gf_false;
+ ret = pthread_mutex_unlock(&priv->dm.drain_black_mutex);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_PTHREAD_ERROR,
+ "error=%d", ret, NULL);
+ pthread_cleanup_pop(0);
+ gf_msg_debug(this->name, 0, "Woke up: Conditional wait on black fops");
}
/* Wait till all the white fops are drained */
void
-changelog_drain_white_fops (xlator_t *this, changelog_priv_t *priv)
+changelog_drain_white_fops(xlator_t *this, changelog_priv_t *priv)
{
- int ret = 0;
-
- /* clean up framework of pthread_mutex is required here as
- * 'reconfigure' terminates the changelog_rollover thread
- * on graph change.
- */
- pthread_cleanup_push (changelog_cleanup_free_mutex,
- &priv->dm.drain_white_mutex);
- ret = pthread_mutex_lock (&priv->dm.drain_white_mutex);
+ int ret = 0;
+
+ /* clean up framework of pthread_mutex is required here as
+ * 'reconfigure' terminates the changelog_rollover thread
+ * on graph change.
+ */
+ pthread_cleanup_push(changelog_cleanup_free_mutex,
+ &priv->dm.drain_white_mutex);
+ ret = pthread_mutex_lock(&priv->dm.drain_white_mutex);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_PTHREAD_ERROR,
+ "error=%d", ret, NULL);
+ while (priv->dm.white_fop_cnt > 0) {
+ gf_msg_debug(this->name, 0, "Conditional wait on white fops : %ld",
+ priv->dm.white_fop_cnt);
+ priv->dm.drain_wait_white = _gf_true;
+ ret = pthread_cond_wait(&priv->dm.drain_white_cond,
+ &priv->dm.drain_white_mutex);
if (ret)
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_PTHREAD_ERROR, "pthread error",
- "error=%d", ret,
- NULL);
- while (priv->dm.white_fop_cnt > 0) {
- gf_msg_debug (this->name, 0,
- "Condtional wait on white fops : %ld",
- priv->dm.white_fop_cnt);
- priv->dm.drain_wait_white = _gf_true;
- ret = pthread_cond_wait (&priv->dm.drain_white_cond,
- &priv->dm.drain_white_mutex);
- if (ret)
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_PTHREAD_COND_WAIT_FAILED,
- "pthread cond wait failed",
- "error=%d", ret,
- NULL);
- }
- priv->dm.drain_wait_white = _gf_false;
- ret = pthread_mutex_unlock (&priv->dm.drain_white_mutex);
- if (ret)
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_PTHREAD_ERROR, "pthread error",
- "error=%d", ret,
- NULL);
- pthread_cleanup_pop (0);
- gf_msg_debug (this->name, 0,
- "Woke up: Conditional wait on white fops");
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_PTHREAD_COND_WAIT_FAILED, "error=%d", ret,
+ NULL);
+ }
+ priv->dm.drain_wait_white = _gf_false;
+ ret = pthread_mutex_unlock(&priv->dm.drain_white_mutex);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_PTHREAD_ERROR,
+ "error=%d", ret, NULL);
+ pthread_cleanup_pop(0);
+ gf_msg_debug(this->name, 0, "Woke up: Conditional wait on white fops");
}
/**
@@ -1312,193 +1242,194 @@ changelog_drain_white_fops (xlator_t *this, changelog_priv_t *priv)
* a certain time etc..). move them into separate routine.
*/
void *
-changelog_rollover (void *data)
+changelog_rollover(void *data)
{
- int ret = 0;
- xlator_t *this = NULL;
- struct timespec tv = {0,};
- changelog_log_data_t cld = {0,};
- changelog_time_slice_t *slice = NULL;
- changelog_priv_t *priv = data;
-
- this = priv->cr.this;
- slice = &priv->slice;
-
- while (1) {
- (void) pthread_testcancel();
-
- tv.tv_sec = time (NULL) + priv->rollover_time;
- tv.tv_nsec = 0;
- ret = 0; /* Reset ret to zero */
-
- /* The race between actual rollover and explicit rollover is
- * handled. If actual rollover is being done and the
- * explicit rollover event comes, the event is not missed.
- * Since explicit rollover sets 'cr.notify' to true, this
- * thread doesn't wait on 'pthread_cond_timedwait'.
- */
- pthread_cleanup_push (changelog_cleanup_free_mutex,
- &priv->cr.lock);
- pthread_mutex_lock (&priv->cr.lock);
- {
- while (ret == 0 && !priv->cr.notify)
- ret = pthread_cond_timedwait (&priv->cr.cond,
- &priv->cr.lock,
- &tv);
- if (ret == 0)
- priv->cr.notify = _gf_false;
- }
- pthread_mutex_unlock (&priv->cr.lock);
- pthread_cleanup_pop (0);
-
- if (ret == 0) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_MSG_BARRIER_INFO,
- "Explicit wakeup on barrier notify");
- priv->explicit_rollover = _gf_true;
- } else if (ret && ret != ETIMEDOUT) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_SELECT_FAILED,
- "pthread_cond_timedwait failed");
- continue;
- } else if (ret && ret == ETIMEDOUT) {
- gf_msg_debug (this->name, 0, "Wokeup on timeout");
- }
-
- /* Reading curent_color without lock is fine here
- * as it is only modified here and is next to reading.
- */
- if (priv->current_color == FOP_COLOR_BLACK) {
- LOCK(&priv->lock);
- priv->current_color = FOP_COLOR_WHITE;
- UNLOCK(&priv->lock);
- gf_msg_debug (this->name, 0, "Black fops"
- " to be drained:%ld",
- priv->dm.black_fop_cnt);
- changelog_drain_black_fops (this, priv);
- } else {
- LOCK(&priv->lock);
- priv->current_color = FOP_COLOR_BLACK;
- UNLOCK(&priv->lock);
- gf_msg_debug (this->name, 0, "White fops"
- " to be drained:%ld",
- priv->dm.white_fop_cnt);
- changelog_drain_white_fops (this, priv);
- }
-
- /* Adding delay of 1 second only during explicit rollover:
- *
- * Changelog rollover can happen either due to actual
- * or the explict rollover during snapshot. Actual
- * rollover is controlled by tuneable called 'rollover-time'.
- * The minimum granularity for rollover-time is 1 second.
- * Explicit rollover is asynchronous in nature and happens
- * during snapshot.
- *
- * Basically, rollover renames the current CHANGELOG file
- * to CHANGELOG.TIMESTAMP. Let's assume, at time 't1',
- * actual and explicit rollover raced against each
- * other and actual rollover won the race renaming the
- * CHANGELOG file to CHANGELOG.t1 and opens a new
- * CHANGELOG file. There is high chance that, an immediate
- * explicit rollover at time 't1' can happen with in the same
- * second to rename CHANGELOG file to CHANGELOG.t1 resulting in
- * purging the earlier CHANGELOG.t1 file created by actual
- * rollover. So adding a delay of 1 second guarantees unique
- * CHANGELOG.TIMESTAMP during explicit rollover.
- */
- if (priv->explicit_rollover == _gf_true)
- sleep (1);
-
- ret = changelog_fill_rollover_data (&cld, _gf_false);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_GET_TIME_OP_FAILED,
- "failed to fill rollover data");
- continue;
- }
+ int ret = 0;
+ xlator_t *this = NULL;
+ struct timespec tv = {
+ 0,
+ };
+ changelog_log_data_t cld = {
+ 0,
+ };
+ changelog_time_slice_t *slice = NULL;
+ changelog_priv_t *priv = data;
+
+ this = priv->cr.this;
+ slice = &priv->slice;
+
+ while (1) {
+ (void)pthread_testcancel();
+
+ tv.tv_sec = gf_time() + priv->rollover_time;
+ tv.tv_nsec = 0;
+ ret = 0; /* Reset ret to zero */
+
+ /* The race between actual rollover and explicit rollover is
+ * handled. If actual rollover is being done and the
+ * explicit rollover event comes, the event is not missed.
+ * Since explicit rollover sets 'cr.notify' to true, this
+ * thread doesn't wait on 'pthread_cond_timedwait'.
+ */
+ pthread_cleanup_push(changelog_cleanup_free_mutex, &priv->cr.lock);
+ pthread_mutex_lock(&priv->cr.lock);
+ {
+ while (ret == 0 && !priv->cr.notify)
+ ret = pthread_cond_timedwait(&priv->cr.cond, &priv->cr.lock,
+ &tv);
+ if (ret == 0)
+ priv->cr.notify = _gf_false;
+ }
+ pthread_mutex_unlock(&priv->cr.lock);
+ pthread_cleanup_pop(0);
+
+ if (ret == 0) {
+ gf_smsg(this->name, GF_LOG_INFO, 0, CHANGELOG_MSG_BARRIER_INFO,
+ NULL);
+ priv->explicit_rollover = _gf_true;
+ } else if (ret && ret != ETIMEDOUT) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_SELECT_FAILED, NULL);
+ continue;
+ } else if (ret && ret == ETIMEDOUT) {
+ gf_msg_debug(this->name, 0, "Wokeup on timeout");
+ }
+
+ /* Reading curent_color without lock is fine here
+ * as it is only modified here and is next to reading.
+ */
+ if (priv->current_color == FOP_COLOR_BLACK) {
+ LOCK(&priv->lock);
+ priv->current_color = FOP_COLOR_WHITE;
+ UNLOCK(&priv->lock);
+ gf_msg_debug(this->name, 0,
+ "Black fops"
+ " to be drained:%ld",
+ priv->dm.black_fop_cnt);
+ changelog_drain_black_fops(this, priv);
+ } else {
+ LOCK(&priv->lock);
+ priv->current_color = FOP_COLOR_BLACK;
+ UNLOCK(&priv->lock);
+ gf_msg_debug(this->name, 0,
+ "White fops"
+ " to be drained:%ld",
+ priv->dm.white_fop_cnt);
+ changelog_drain_white_fops(this, priv);
+ }
+
+ /* Adding delay of 1 second only during explicit rollover:
+ *
+ * Changelog rollover can happen either due to actual
+ * or the explicit rollover during snapshot. Actual
+ * rollover is controlled by tuneable called 'rollover-time'.
+ * The minimum granularity for rollover-time is 1 second.
+ * Explicit rollover is asynchronous in nature and happens
+ * during snapshot.
+ *
+ * Basically, rollover renames the current CHANGELOG file
+ * to CHANGELOG.TIMESTAMP. Let's assume, at time 't1',
+ * actual and explicit rollover raced against each
+ * other and actual rollover won the race renaming the
+ * CHANGELOG file to CHANGELOG.t1 and opens a new
+ * CHANGELOG file. There is high chance that, an immediate
+ * explicit rollover at time 't1' can happen with in the same
+ * second to rename CHANGELOG file to CHANGELOG.t1 resulting in
+ * purging the earlier CHANGELOG.t1 file created by actual
+ * rollover. So adding a delay of 1 second guarantees unique
+ * CHANGELOG.TIMESTAMP during explicit rollover.
+ */
+ if (priv->explicit_rollover == _gf_true)
+ sleep(1);
- _mask_cancellation ();
+ changelog_fill_rollover_data(&cld, _gf_false);
- LOCK (&priv->lock);
- {
- ret = changelog_inject_single_event (this, priv, &cld);
- if (!ret)
- SLICE_VERSION_UPDATE (slice);
- }
- UNLOCK (&priv->lock);
+ _mask_cancellation();
- _unmask_cancellation ();
+ LOCK(&priv->lock);
+ {
+ ret = changelog_inject_single_event(this, priv, &cld);
+ if (!ret)
+ SLICE_VERSION_UPDATE(slice);
}
+ UNLOCK(&priv->lock);
- return NULL;
+ _unmask_cancellation();
+ }
+
+ return NULL;
}
void *
-changelog_fsync_thread (void *data)
+changelog_fsync_thread(void *data)
{
- int ret = 0;
- xlator_t *this = NULL;
- struct timeval tv = {0,};
- changelog_log_data_t cld = {0,};
- changelog_priv_t *priv = data;
-
- this = priv->cf.this;
- cld.cld_type = CHANGELOG_TYPE_FSYNC;
-
- while (1) {
- (void) pthread_testcancel();
-
- tv.tv_sec = priv->fsync_interval;
- tv.tv_usec = 0;
+ int ret = 0;
+ xlator_t *this = NULL;
+ struct timeval tv = {
+ 0,
+ };
+ changelog_log_data_t cld = {
+ 0,
+ };
+ changelog_priv_t *priv = data;
+
+ this = priv->cf.this;
+ cld.cld_type = CHANGELOG_TYPE_FSYNC;
+
+ while (1) {
+ (void)pthread_testcancel();
+
+ tv.tv_sec = priv->fsync_interval;
+ tv.tv_usec = 0;
+
+ ret = select(0, NULL, NULL, NULL, &tv);
+ if (ret)
+ continue;
- ret = select (0, NULL, NULL, NULL, &tv);
- if (ret)
- continue;
+ _mask_cancellation();
- _mask_cancellation ();
+ ret = changelog_inject_single_event(this, priv, &cld);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_MSG_INJECT_FSYNC_FAILED, NULL);
- ret = changelog_inject_single_event (this, priv, &cld);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_INJECT_FSYNC_FAILED,
- "failed to inject fsync event");
+ _unmask_cancellation();
+ }
- _unmask_cancellation ();
- }
-
- return NULL;
+ return NULL;
}
/* macros for inode/changelog version checks */
-#define INODE_VERSION_UPDATE(priv, inode, iver, slice, type) do { \
- LOCK (&inode->lock); \
- { \
- LOCK (&priv->lock); \
- { \
- *iver = slice->changelog_version[type]; \
- } \
- UNLOCK (&priv->lock); \
- } \
- UNLOCK (&inode->lock); \
- } while (0)
-
-#define INODE_VERSION_EQUALS_SLICE(priv, ver, slice, type, upd) do { \
- LOCK (&priv->lock); \
- { \
- upd = (ver == slice->changelog_version[type]) \
- ? _gf_false : _gf_true; \
- } \
- UNLOCK (&priv->lock); \
- } while (0)
+#define INODE_VERSION_UPDATE(priv, inode, iver, slice, type) \
+ do { \
+ LOCK(&inode->lock); \
+ { \
+ LOCK(&priv->lock); \
+ { \
+ *iver = slice->changelog_version[type]; \
+ } \
+ UNLOCK(&priv->lock); \
+ } \
+ UNLOCK(&inode->lock); \
+ } while (0)
+
+#define INODE_VERSION_EQUALS_SLICE(priv, ver, slice, type, upd) \
+ do { \
+ LOCK(&priv->lock); \
+ { \
+ upd = (ver == slice->changelog_version[type]) ? _gf_false \
+ : _gf_true; \
+ } \
+ UNLOCK(&priv->lock); \
+ } while (0)
static int
-__changelog_inode_ctx_set (xlator_t *this,
- inode_t *inode, changelog_inode_ctx_t *ctx)
+__changelog_inode_ctx_set(xlator_t *this, inode_t *inode,
+ changelog_inode_ctx_t *ctx)
{
- uint64_t ctx_addr = (uint64_t) ctx;
- return __inode_ctx_set (inode, this, &ctx_addr);
+ uint64_t ctx_addr = (uint64_t)(uintptr_t)ctx;
+ return __inode_ctx_set(inode, this, &ctx_addr);
}
/**
@@ -1506,56 +1437,53 @@ __changelog_inode_ctx_set (xlator_t *this,
* for a particular type.
*/
changelog_inode_ctx_t *
-__changelog_inode_ctx_get (xlator_t *this,
- inode_t *inode, unsigned long **iver,
- unsigned long *version, changelog_log_type type)
+__changelog_inode_ctx_get(xlator_t *this, inode_t *inode, unsigned long **iver,
+ unsigned long *version, changelog_log_type type)
{
- int ret = 0;
- uint64_t ctx_addr = 0;
- changelog_inode_ctx_t *ctx = NULL;
-
- ret = __inode_ctx_get (inode, this, &ctx_addr);
- if (ret < 0)
- ctx_addr = 0;
- if (ctx_addr != 0) {
- ctx = (changelog_inode_ctx_t *) (long)ctx_addr;
- goto out;
- }
-
- ctx = GF_CALLOC (1, sizeof (*ctx), gf_changelog_mt_inode_ctx_t);
- if (!ctx)
- goto out;
+ int ret = 0;
+ uint64_t ctx_addr = 0;
+ changelog_inode_ctx_t *ctx = NULL;
+
+ ret = __inode_ctx_get(inode, this, &ctx_addr);
+ if (ret < 0)
+ ctx_addr = 0;
+ if (ctx_addr != 0) {
+ ctx = (changelog_inode_ctx_t *)(long)ctx_addr;
+ goto out;
+ }
+
+ ctx = GF_CALLOC(1, sizeof(*ctx), gf_changelog_mt_inode_ctx_t);
+ if (!ctx)
+ goto out;
+
+ ret = __changelog_inode_ctx_set(this, inode, ctx);
+ if (ret) {
+ GF_FREE(ctx);
+ ctx = NULL;
+ }
- ret = __changelog_inode_ctx_set (this, inode, ctx);
- if (ret) {
- GF_FREE (ctx);
- ctx = NULL;
- }
-
- out:
- if (ctx && iver && version) {
- *iver = CHANGELOG_INODE_VERSION_TYPE (ctx, type);
- *version = **iver;
- }
+out:
+ if (ctx && iver && version) {
+ *iver = CHANGELOG_INODE_VERSION_TYPE(ctx, type);
+ *version = **iver;
+ }
- return ctx;
+ return ctx;
}
static changelog_inode_ctx_t *
-changelog_inode_ctx_get (xlator_t *this,
- inode_t *inode, unsigned long **iver,
- unsigned long *version, changelog_log_type type)
+changelog_inode_ctx_get(xlator_t *this, inode_t *inode, unsigned long **iver,
+ unsigned long *version, changelog_log_type type)
{
- changelog_inode_ctx_t *ctx = NULL;
+ changelog_inode_ctx_t *ctx = NULL;
- LOCK (&inode->lock);
- {
- ctx = __changelog_inode_ctx_get (this,
- inode, iver, version, type);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ctx = __changelog_inode_ctx_get(this, inode, iver, version, type);
+ }
+ UNLOCK(&inode->lock);
- return ctx;
+ return ctx;
}
/**
@@ -1659,59 +1587,57 @@ changelog_inode_ctx_get (xlator_t *this,
* signifies an update was recorded in the current time slice).
*/
void
-changelog_update (xlator_t *this, changelog_priv_t *priv,
- changelog_local_t *local, changelog_log_type type)
+changelog_update(xlator_t *this, changelog_priv_t *priv,
+ changelog_local_t *local, changelog_log_type type)
{
- int ret = 0;
- unsigned long *iver = NULL;
- unsigned long version = 0;
- inode_t *inode = NULL;
- changelog_time_slice_t *slice = NULL;
- changelog_inode_ctx_t *ctx = NULL;
- changelog_log_data_t *cld_0 = NULL;
- changelog_log_data_t *cld_1 = NULL;
- changelog_local_t *next_local = NULL;
- gf_boolean_t need_upd = _gf_true;
-
- slice = &priv->slice;
-
- /**
- * for fops that do not require inode version checking
- */
- if (local->update_no_check)
- goto update;
+ int ret = 0;
+ unsigned long *iver = NULL;
+ unsigned long version = 0;
+ inode_t *inode = NULL;
+ changelog_time_slice_t *slice = NULL;
+ changelog_inode_ctx_t *ctx = NULL;
+ changelog_log_data_t *cld_0 = NULL;
+ changelog_log_data_t *cld_1 = NULL;
+ changelog_local_t *next_local = NULL;
+ gf_boolean_t need_upd = _gf_true;
- inode = local->inode;
+ slice = &priv->slice;
- ctx = changelog_inode_ctx_get (this,
- inode, &iver, &version, type);
- if (!ctx)
- goto update;
+ /**
+ * for fops that do not require inode version checking
+ */
+ if (local->update_no_check)
+ goto update;
- INODE_VERSION_EQUALS_SLICE (priv, version, slice, type, need_upd);
+ inode = local->inode;
- update:
- if (need_upd) {
- cld_0 = &local->cld;
- cld_0->cld_type = type;
+ ctx = changelog_inode_ctx_get(this, inode, &iver, &version, type);
+ if (!ctx)
+ goto update;
- if ( (next_local = local->prev_entry) != NULL ) {
- cld_1 = &next_local->cld;
- cld_1->cld_type = type;
- }
+ INODE_VERSION_EQUALS_SLICE(priv, version, slice, type, need_upd);
- ret = priv->cd.dispatchfn (this, priv,
- priv->cd.cd_data, cld_0, cld_1);
+update:
+ if (need_upd) {
+ cld_0 = &local->cld;
+ cld_0->cld_type = type;
- /**
- * update after the dispatcher has successfully done
- * it's job.
- */
- if (!local->update_no_check && iver && !ret)
- INODE_VERSION_UPDATE (priv, inode, iver, slice, type);
+ if ((next_local = local->prev_entry) != NULL) {
+ cld_1 = &next_local->cld;
+ cld_1->cld_type = type;
}
- return;
+ ret = priv->cd.dispatchfn(this, priv, priv->cd.cd_data, cld_0, cld_1);
+
+ /**
+ * update after the dispatcher has successfully done
+ * it's job.
+ */
+ if (!local->update_no_check && iver && !ret)
+ INODE_VERSION_UPDATE(priv, inode, iver, slice, type);
+ }
+
+ return;
}
/* Begin: Geo-rep snapshot dependency changes */
@@ -1727,228 +1653,221 @@ changelog_update (xlator_t *this, changelog_priv_t *priv,
*/
void
-changelog_color_fop_and_inc_cnt (xlator_t *this, changelog_priv_t *priv,
- changelog_local_t *local)
+changelog_color_fop_and_inc_cnt(xlator_t *this, changelog_priv_t *priv,
+ changelog_local_t *local)
{
- if (!priv || !local)
- return;
+ if (!priv || !local)
+ return;
- LOCK (&priv->lock);
- {
- local->color = priv->current_color;
- changelog_inc_fop_cnt (this, priv, local);
- }
- UNLOCK (&priv->lock);
+ LOCK(&priv->lock);
+ {
+ local->color = priv->current_color;
+ changelog_inc_fop_cnt(this, priv, local);
+ }
+ UNLOCK(&priv->lock);
}
/* Increments the respective fop counter based on the fop color */
void
-changelog_inc_fop_cnt (xlator_t *this, changelog_priv_t *priv,
- changelog_local_t *local)
+changelog_inc_fop_cnt(xlator_t *this, changelog_priv_t *priv,
+ changelog_local_t *local)
{
- int ret = 0;
-
- if (local) {
- if (local->color == FOP_COLOR_BLACK) {
- ret = pthread_mutex_lock (&priv->dm.drain_black_mutex);
- CHANGELOG_PTHREAD_ERROR_HANDLE_0 (ret, out);
- {
- priv->dm.black_fop_cnt++;
- }
- ret = pthread_mutex_unlock(&priv->dm.drain_black_mutex);
- CHANGELOG_PTHREAD_ERROR_HANDLE_0 (ret, out);
- } else {
- ret = pthread_mutex_lock (&priv->dm.drain_white_mutex);
- CHANGELOG_PTHREAD_ERROR_HANDLE_0 (ret, out);
- {
- priv->dm.white_fop_cnt++;
- }
- ret = pthread_mutex_unlock(&priv->dm.drain_white_mutex);
- CHANGELOG_PTHREAD_ERROR_HANDLE_0 (ret, out);
- }
- }
- out:
- return;
+ int ret = 0;
+
+ if (local) {
+ if (local->color == FOP_COLOR_BLACK) {
+ ret = pthread_mutex_lock(&priv->dm.drain_black_mutex);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_0(ret, out);
+ {
+ priv->dm.black_fop_cnt++;
+ }
+ ret = pthread_mutex_unlock(&priv->dm.drain_black_mutex);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_0(ret, out);
+ } else {
+ ret = pthread_mutex_lock(&priv->dm.drain_white_mutex);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_0(ret, out);
+ {
+ priv->dm.white_fop_cnt++;
+ }
+ ret = pthread_mutex_unlock(&priv->dm.drain_white_mutex);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_0(ret, out);
+ }
+ }
+out:
+ return;
}
/* Decrements the respective fop counter based on the fop color */
void
-changelog_dec_fop_cnt (xlator_t *this, changelog_priv_t *priv,
- changelog_local_t *local)
+changelog_dec_fop_cnt(xlator_t *this, changelog_priv_t *priv,
+ changelog_local_t *local)
{
- int ret = 0;
-
- if (local) {
- if (local->color == FOP_COLOR_BLACK) {
- ret = pthread_mutex_lock (&priv->dm.drain_black_mutex);
- CHANGELOG_PTHREAD_ERROR_HANDLE_0 (ret, out);
- {
- priv->dm.black_fop_cnt--;
- if (priv->dm.black_fop_cnt == 0 &&
- priv->dm.drain_wait_black == _gf_true) {
- ret = pthread_cond_signal (
- &priv->dm.drain_black_cond);
- CHANGELOG_PTHREAD_ERROR_HANDLE_0 (ret,
- out);
- gf_msg_debug (this->name, 0,
- "Signalled "
- "draining of black");
- }
- }
- ret = pthread_mutex_unlock(&priv->dm.drain_black_mutex);
- CHANGELOG_PTHREAD_ERROR_HANDLE_0 (ret, out);
- } else {
- ret = pthread_mutex_lock (&priv->dm.drain_white_mutex);
- CHANGELOG_PTHREAD_ERROR_HANDLE_0 (ret, out);
- {
- priv->dm.white_fop_cnt--;
- if (priv->dm.white_fop_cnt == 0 &&
- priv->dm.drain_wait_white == _gf_true) {
- ret = pthread_cond_signal (
- &priv->dm.drain_white_cond);
- CHANGELOG_PTHREAD_ERROR_HANDLE_0 (ret,
- out);
- gf_msg_debug (this->name, 0,
- "Signalled "
- "draining of white");
- }
- }
- ret = pthread_mutex_unlock(&priv->dm.drain_white_mutex);
- CHANGELOG_PTHREAD_ERROR_HANDLE_0 (ret, out);
+ int ret = 0;
+
+ if (local) {
+ if (local->color == FOP_COLOR_BLACK) {
+ ret = pthread_mutex_lock(&priv->dm.drain_black_mutex);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_0(ret, out);
+ {
+ priv->dm.black_fop_cnt--;
+ if (priv->dm.black_fop_cnt == 0 &&
+ priv->dm.drain_wait_black == _gf_true) {
+ ret = pthread_cond_signal(&priv->dm.drain_black_cond);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_2(
+ ret, out, priv->dm.drain_black_mutex);
+ gf_msg_debug(this->name, 0,
+ "Signalled "
+ "draining of black");
}
+ }
+ ret = pthread_mutex_unlock(&priv->dm.drain_black_mutex);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_0(ret, out);
+ } else {
+ ret = pthread_mutex_lock(&priv->dm.drain_white_mutex);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_0(ret, out);
+ {
+ priv->dm.white_fop_cnt--;
+ if (priv->dm.white_fop_cnt == 0 &&
+ priv->dm.drain_wait_white == _gf_true) {
+ ret = pthread_cond_signal(&priv->dm.drain_white_cond);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_2(
+ ret, out, priv->dm.drain_white_mutex);
+ gf_msg_debug(this->name, 0,
+ "Signalled "
+ "draining of white");
+ }
+ }
+ ret = pthread_mutex_unlock(&priv->dm.drain_white_mutex);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_0(ret, out);
}
- out:
- return;
+ }
+out:
+ return;
}
/* Write to a pipe setup between changelog main thread and changelog
* rollover thread to initiate explicit rollover of changelog journal.
*/
int
-changelog_barrier_notify (changelog_priv_t *priv, char *buf)
+changelog_barrier_notify(changelog_priv_t *priv, char *buf)
{
- int ret = 0;
-
- pthread_mutex_lock (&priv->cr.lock);
- {
- ret = pthread_cond_signal (&priv->cr.cond);
- priv->cr.notify = _gf_true;
- }
- pthread_mutex_unlock (&priv->cr.lock);
- return ret;
+ int ret = 0;
+
+ pthread_mutex_lock(&priv->cr.lock);
+ {
+ ret = pthread_cond_signal(&priv->cr.cond);
+ priv->cr.notify = _gf_true;
+ }
+ pthread_mutex_unlock(&priv->cr.lock);
+ return ret;
}
/* Clean up flags set on barrier notification */
void
-changelog_barrier_cleanup (xlator_t *this, changelog_priv_t *priv,
- struct list_head *queue)
+changelog_barrier_cleanup(xlator_t *this, changelog_priv_t *priv,
+ struct list_head *queue)
{
- int ret = 0;
-
- LOCK (&priv->bflags.lock);
- priv->bflags.barrier_ext = _gf_false;
- UNLOCK (&priv->bflags.lock);
-
- ret = pthread_mutex_lock (&priv->bn.bnotify_mutex);
- CHANGELOG_PTHREAD_ERROR_HANDLE_0 (ret, out);
- {
- priv->bn.bnotify = _gf_false;
- }
- ret = pthread_mutex_unlock (&priv->bn.bnotify_mutex);
- CHANGELOG_PTHREAD_ERROR_HANDLE_0 (ret, out);
-
- /* Disable changelog barrier and dequeue fops */
- LOCK (&priv->lock);
- {
- if (priv->barrier_enabled == _gf_true)
- __chlog_barrier_disable (this, queue);
- else
- ret = -1;
- }
- UNLOCK (&priv->lock);
- if (ret == 0)
- chlog_barrier_dequeue_all(this, queue);
+ int ret = 0;
+
+ LOCK(&priv->bflags.lock);
+ priv->bflags.barrier_ext = _gf_false;
+ UNLOCK(&priv->bflags.lock);
+
+ ret = pthread_mutex_lock(&priv->bn.bnotify_mutex);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_0(ret, out);
+ {
+ priv->bn.bnotify = _gf_false;
+ }
+ ret = pthread_mutex_unlock(&priv->bn.bnotify_mutex);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_0(ret, out);
+
+ /* Disable changelog barrier and dequeue fops */
+ LOCK(&priv->lock);
+ {
+ if (priv->barrier_enabled == _gf_true)
+ __chlog_barrier_disable(this, queue);
+ else
+ ret = -1;
+ }
+ UNLOCK(&priv->lock);
+ if (ret == 0)
+ chlog_barrier_dequeue_all(this, queue);
- out:
- return;
+out:
+ return;
}
/* End: Geo-Rep snapshot dependency changes */
int32_t
-changelog_fill_entry_buf (call_frame_t *frame, xlator_t *this,
- loc_t *loc, changelog_local_t **local)
+changelog_fill_entry_buf(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ changelog_local_t **local)
{
- changelog_opt_t *co = NULL;
- size_t xtra_len = 0;
- char *dup_path = NULL;
- char *bname = NULL;
- inode_t *parent = NULL;
-
- GF_ASSERT (this);
-
- parent = inode_parent (loc->inode, 0, 0);
- if (!parent) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_INODE_NOT_FOUND,
- "Parent inode not found",
- "gfid=%s", uuid_utoa (loc->inode->gfid),
- NULL);
- goto err;
- }
-
- CHANGELOG_INIT_NOCHECK (this, *local, loc->inode, loc->inode->gfid, 5);
- if (!(*local)) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_LOCAL_INIT_FAILED, "changelog local"
- " initiatilization failed");
- goto err;
- }
-
- co = changelog_get_usable_buffer (*local);
- if (!co) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_NO_MEMORY,
- "Failed to get buffer");
- goto err;
- }
-
- if (loc->inode->ia_type == IA_IFDIR) {
- CHANGLOG_FILL_FOP_NUMBER (co, GF_FOP_MKDIR, fop_fn, xtra_len);
- co++;
- CHANGELOG_FILL_UINT32 (co, S_IFDIR|0755, number_fn, xtra_len);
- co++;
- } else {
- CHANGLOG_FILL_FOP_NUMBER (co, GF_FOP_CREATE, fop_fn, xtra_len);
- co++;
- CHANGELOG_FILL_UINT32 (co, S_IFREG|0644, number_fn, xtra_len);
- co++;
- }
-
- CHANGELOG_FILL_UINT32 (co, frame->root->uid, number_fn, xtra_len);
+ changelog_opt_t *co = NULL;
+ size_t xtra_len = 0;
+ char *dup_path = NULL;
+ char *bname = NULL;
+ inode_t *parent = NULL;
+
+ GF_ASSERT(this);
+
+ parent = inode_parent(loc->inode, 0, 0);
+ if (!parent) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno, CHANGELOG_MSG_INODE_NOT_FOUND,
+ "type=parent", "gfid=%s", uuid_utoa(loc->inode->gfid), NULL);
+ goto err;
+ }
+
+ CHANGELOG_INIT_NOCHECK(this, *local, loc->inode, loc->inode->gfid, 5);
+ if (!(*local)) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_LOCAL_INIT_FAILED,
+ NULL);
+ goto err;
+ }
+
+ co = changelog_get_usable_buffer(*local);
+ if (!co) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_GET_BUFFER_FAILED,
+ NULL);
+ goto err;
+ }
+
+ if (loc->inode->ia_type == IA_IFDIR) {
+ CHANGLOG_FILL_FOP_NUMBER(co, GF_FOP_MKDIR, fop_fn, xtra_len);
co++;
-
- CHANGELOG_FILL_UINT32 (co, frame->root->gid, number_fn, xtra_len);
+ CHANGELOG_FILL_UINT32(co, S_IFDIR | 0755, number_fn, xtra_len);
+ co++;
+ } else {
+ CHANGLOG_FILL_FOP_NUMBER(co, GF_FOP_CREATE, fop_fn, xtra_len);
+ co++;
+ CHANGELOG_FILL_UINT32(co, S_IFREG | 0644, number_fn, xtra_len);
co++;
+ }
- dup_path = gf_strdup (loc->path);
- bname = basename (dup_path);
+ CHANGELOG_FILL_UINT32(co, frame->root->uid, number_fn, xtra_len);
+ co++;
- CHANGELOG_FILL_ENTRY (co, parent->gfid, bname, entry_fn, entry_free_fn,
- xtra_len, err);
- changelog_set_usable_record_and_length (*local, xtra_len, 5);
+ CHANGELOG_FILL_UINT32(co, frame->root->gid, number_fn, xtra_len);
+ co++;
- if (dup_path)
- GF_FREE (dup_path);
- if (parent)
- inode_unref (parent);
- return 0;
+ dup_path = gf_strdup(loc->path);
+ bname = basename(dup_path);
+
+ CHANGELOG_FILL_ENTRY(co, parent->gfid, bname, entry_fn, entry_free_fn,
+ xtra_len, err);
+ changelog_set_usable_record_and_length(*local, xtra_len, 5);
+
+ if (dup_path)
+ GF_FREE(dup_path);
+ if (parent)
+ inode_unref(parent);
+ return 0;
err:
- if (dup_path)
- GF_FREE (dup_path);
- if (parent)
- inode_unref (parent);
- return -1;
+ if (dup_path)
+ GF_FREE(dup_path);
+ if (parent)
+ inode_unref(parent);
+ return -1;
}
/*
@@ -1961,80 +1880,98 @@ err:
*/
int
-resolve_pargfid_to_path (xlator_t *this, const uuid_t pgfid,
- char **path, char *bname)
+resolve_pargfid_to_path(xlator_t *this, const uuid_t pgfid, char **path,
+ char *bname)
{
- char *linkname = NULL;
- char *dir_handle = NULL;
- char *pgfidstr = NULL;
- char *saveptr = NULL;
- ssize_t len = 0;
- int ret = 0;
- uuid_t tmp_gfid = {0, };
- uuid_t pargfid = {0, };
- changelog_priv_t *priv = NULL;
- char gpath[PATH_MAX] = {0,};
- char result[PATH_MAX] = {0,};
- char *dir_name = NULL;
- char pre_dir_name[PATH_MAX] = {0,};
-
- GF_ASSERT (this);
- priv = this->private;
- GF_ASSERT (priv);
-
- gf_uuid_copy (pargfid, pgfid);
- if (!path || gf_uuid_is_null (pargfid)) {
- ret = -1;
- goto out;
- }
+ char *linkname = NULL;
+ char *dir_handle = NULL;
+ char *pgfidstr = NULL;
+ char *saveptr = NULL;
+ ssize_t len = 0;
+ int ret = 0;
+ uuid_t tmp_gfid = {
+ 0,
+ };
+ uuid_t pargfid = {
+ 0,
+ };
+ changelog_priv_t *priv = NULL;
+ char gpath[PATH_MAX] = {
+ 0,
+ };
+ char result[PATH_MAX] = {
+ 0,
+ };
+ char *dir_name = NULL;
+ char pre_dir_name[PATH_MAX] = {
+ 0,
+ };
+
+ GF_ASSERT(this);
+ priv = this->private;
+ GF_ASSERT(priv);
+
+ gf_uuid_copy(pargfid, pgfid);
+ if (!path || gf_uuid_is_null(pargfid)) {
+ ret = -1;
+ goto out;
+ }
+
+ if (__is_root_gfid(pargfid)) {
+ if (bname)
+ *path = gf_strdup(bname);
+ else
+ *path = gf_strdup(".");
+ return ret;
+ }
- if (__is_root_gfid (pargfid)) {
- if (bname)
- *path = gf_strdup (bname);
- else
- *path = gf_strdup (".");
- return ret;
- }
+ dir_handle = alloca(PATH_MAX);
+ linkname = alloca(PATH_MAX);
+ (void)snprintf(gpath, PATH_MAX, "%s/.glusterfs/", priv->changelog_brick);
- dir_handle = alloca (PATH_MAX);
- linkname = alloca (PATH_MAX);
- (void) snprintf (gpath, PATH_MAX, "%s/.glusterfs/",
- priv->changelog_brick);
-
- while (!(__is_root_gfid (pargfid))) {
- snprintf (dir_handle, PATH_MAX, "%s/%02x/%02x/%s", gpath,
- pargfid[0], pargfid[1], uuid_utoa (pargfid));
-
- len = sys_readlink (dir_handle, linkname, PATH_MAX);
- if (len < 0) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_READLINK_OP_FAILED,
- "could not read the "
- "link from the gfid handle",
- "handle=%s", dir_handle,
- NULL);
- ret = -1;
- goto out;
- }
+ while (!(__is_root_gfid(pargfid))) {
+ len = snprintf(dir_handle, PATH_MAX, "%s/%02x/%02x/%s", gpath,
+ pargfid[0], pargfid[1], uuid_utoa(pargfid));
+ if ((len < 0) || (len >= PATH_MAX)) {
+ ret = -1;
+ goto out;
+ }
- linkname[len] = '\0';
+ len = sys_readlink(dir_handle, linkname, PATH_MAX);
+ if (len < 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_READLINK_OP_FAILED,
+ "could not read the "
+ "link from the gfid handle",
+ "handle=%s", dir_handle, NULL);
+ ret = -1;
+ goto out;
+ }
- pgfidstr = strtok_r (linkname + strlen("../../00/00/"), "/",
- &saveptr);
- dir_name = strtok_r (NULL, "/", &saveptr);
+ linkname[len] = '\0';
- snprintf (result, PATH_MAX, "%s/%s", dir_name, pre_dir_name);
- strncpy (pre_dir_name, result, sizeof(pre_dir_name));
+ pgfidstr = strtok_r(linkname + strlen("../../00/00/"), "/", &saveptr);
+ dir_name = strtok_r(NULL, "/", &saveptr);
- gf_uuid_parse (pgfidstr, tmp_gfid);
- gf_uuid_copy (pargfid, tmp_gfid);
+ len = snprintf(result, PATH_MAX, "%s/%s", dir_name, pre_dir_name);
+ if ((len < 0) || (len >= PATH_MAX)) {
+ ret = -1;
+ goto out;
+ }
+ if (snprintf(pre_dir_name, len + 1, "%s", result) >= len + 1) {
+ ret = -1;
+ goto out;
}
- if (bname)
- strncat (result, bname, strlen(bname) + 1);
+ gf_uuid_parse(pgfidstr, tmp_gfid);
+ gf_uuid_copy(pargfid, tmp_gfid);
+ }
- *path = gf_strdup (result);
+ if (bname)
+ strncat(result, bname, strlen(bname) + 1);
+
+ *path = gf_strdup(result);
out:
- return ret;
+ return ret;
}
diff --git a/xlators/features/changelog/src/changelog-helpers.h b/xlators/features/changelog/src/changelog-helpers.h
index 39a792891dd..38fa7590c32 100644
--- a/xlators/features/changelog/src/changelog-helpers.h
+++ b/xlators/features/changelog/src/changelog-helpers.h
@@ -11,14 +11,14 @@
#ifndef _CHANGELOG_HELPERS_H
#define _CHANGELOG_HELPERS_H
-#include "locking.h"
-#include "timer.h"
+#include <glusterfs/locking.h>
+#include <glusterfs/timer.h>
#include "pthread.h"
-#include "iobuf.h"
-#include "rot-buffs.h"
+#include <glusterfs/iobuf.h>
+#include <glusterfs/rot-buffs.h>
#include "changelog-misc.h"
-#include "call-stub.h"
+#include <glusterfs/call-stub.h>
#include "rpcsvc.h"
#include "changelog-ev-handle.h"
@@ -30,44 +30,44 @@
* the changelog entry
*/
typedef struct changelog_log_data {
- /* rollover related */
- unsigned long cld_roll_time;
+ /* rollover related */
+ time_t cld_roll_time;
- /* reopen changelog? */
- gf_boolean_t cld_finale;
+ /* reopen changelog? */
+ gf_boolean_t cld_finale;
- changelog_log_type cld_type;
+ changelog_log_type cld_type;
- /**
- * sincd gfid is _always_ a necessity, it's not a part
- * of the iobuf. by doing this we do not add any overhead
- * for data and metadata related fops.
- */
- uuid_t cld_gfid;
+ /**
+ * sincd gfid is _always_ a necessity, it's not a part
+ * of the iobuf. by doing this we do not add any overhead
+ * for data and metadata related fops.
+ */
+ uuid_t cld_gfid;
- /**
- * iobufs are used for optionals records: pargfid, path,
- * write offsets etc.. It's the fop implementers job
- * to allocate (iobuf_get() in the fop) and get unref'ed
- * in the callback (CHANGELOG_STACK_UNWIND).
- */
- struct iobuf *cld_iobuf;
+ /**
+ * iobufs are used for optionals records: pargfid, path,
+ * write offsets etc.. It's the fop implementers job
+ * to allocate (iobuf_get() in the fop) and get unref'ed
+ * in the callback (CHANGELOG_STACK_UNWIND).
+ */
+ struct iobuf *cld_iobuf;
#define cld_ptr cld_iobuf->ptr
- /**
- * after allocation you can point this to the length of
- * usable data, but make sure it does not exceed the
- * the size of the requested iobuf.
- */
- size_t cld_iobuf_len;
+ /**
+ * after allocation you can point this to the length of
+ * usable data, but make sure it does not exceed the
+ * the size of the requested iobuf.
+ */
+ size_t cld_iobuf_len;
#define cld_ptr_len cld_iobuf_len
- /**
- * number of optional records
- */
- int cld_xtra_records;
+ /**
+ * number of optional records
+ */
+ int cld_xtra_records;
} changelog_log_data_t;
/**
@@ -77,55 +77,48 @@ typedef struct changelog_log_data {
typedef struct changelog_priv changelog_priv_t;
typedef struct changelog_dispatcher {
- void *cd_data;
- int (*dispatchfn) (xlator_t *, changelog_priv_t *, void *,
- changelog_log_data_t *, changelog_log_data_t *);
+ void *cd_data;
+ int (*dispatchfn)(xlator_t *, changelog_priv_t *, void *,
+ changelog_log_data_t *, changelog_log_data_t *);
} changelog_dispatcher_t;
struct changelog_bootstrap {
- changelog_mode_t mode;
- int (*ctor) (xlator_t *, changelog_dispatcher_t *);
- int (*dtor) (xlator_t *, changelog_dispatcher_t *);
+ changelog_mode_t mode;
+ int (*ctor)(xlator_t *, changelog_dispatcher_t *);
+ int (*dtor)(xlator_t *, changelog_dispatcher_t *);
};
struct changelog_encoder {
- changelog_encoder_t encoder;
- int (*encode) (xlator_t *, changelog_log_data_t *);
+ changelog_encoder_t encoder;
+ int (*encode)(xlator_t *, changelog_log_data_t *);
};
-
/* xlator private */
typedef struct changelog_time_slice {
- /**
- * just in case we need nanosecond granularity some day.
- * field is unused as of now (maybe we'd need it later).
- */
- struct timeval tv_start;
-
- /**
- * version of changelog file, incremented each time changes
- * rollover.
- */
- unsigned long changelog_version[CHANGELOG_MAX_TYPE];
+ /**
+ * version of changelog file, incremented each time changes
+ * rollover.
+ */
+ unsigned long changelog_version[CHANGELOG_MAX_TYPE];
} changelog_time_slice_t;
typedef struct changelog_rollover {
- /* rollover thread */
- pthread_t rollover_th;
+ /* rollover thread */
+ pthread_t rollover_th;
- xlator_t *this;
+ xlator_t *this;
- pthread_mutex_t lock;
- pthread_cond_t cond;
- gf_boolean_t notify;
+ pthread_mutex_t lock;
+ pthread_cond_t cond;
+ gf_boolean_t notify;
} changelog_rollover_t;
typedef struct changelog_fsync {
- /* fsync() thread */
- pthread_t fsync_th;
+ /* fsync() thread */
+ pthread_t fsync_th;
- xlator_t *this;
+ xlator_t *this;
} changelog_fsync_t;
/* Draining during changelog rollover (for geo-rep snapshot dependency):
@@ -145,198 +138,219 @@ typedef struct changelog_fsync {
*/
typedef enum chlog_fop_color {
- FOP_COLOR_BLACK,
- FOP_COLOR_WHITE
+ FOP_COLOR_BLACK,
+ FOP_COLOR_WHITE
} chlog_fop_color_t;
/* Barrier notify variable */
typedef struct barrier_notify {
- pthread_mutex_t bnotify_mutex;
- pthread_cond_t bnotify_cond;
- gf_boolean_t bnotify;
- gf_boolean_t bnotify_error;
+ pthread_mutex_t bnotify_mutex;
+ pthread_cond_t bnotify_cond;
+ gf_boolean_t bnotify;
+ gf_boolean_t bnotify_error;
} barrier_notify_t;
/* Two separate mutex and conditional variable set is used
* to drain white and black fops. */
typedef struct drain_mgmt {
- pthread_mutex_t drain_black_mutex;
- pthread_cond_t drain_black_cond;
- pthread_mutex_t drain_white_mutex;
- pthread_cond_t drain_white_cond;
- /* Represents black fops count in-transit */
- unsigned long black_fop_cnt;
- /* Represents white fops count in-transit */
- unsigned long white_fop_cnt;
- gf_boolean_t drain_wait_black;
- gf_boolean_t drain_wait_white;
+ pthread_mutex_t drain_black_mutex;
+ pthread_cond_t drain_black_cond;
+ pthread_mutex_t drain_white_mutex;
+ pthread_cond_t drain_white_cond;
+ /* Represents black fops count in-transit */
+ unsigned long black_fop_cnt;
+ /* Represents white fops count in-transit */
+ unsigned long white_fop_cnt;
+ gf_boolean_t drain_wait_black;
+ gf_boolean_t drain_wait_white;
} drain_mgmt_t;
/* External barrier as a result of snap on/off indicating flag*/
typedef struct barrier_flags {
- gf_lock_t lock;
- gf_boolean_t barrier_ext;
+ gf_lock_t lock;
+ gf_boolean_t barrier_ext;
} barrier_flags_t;
/* Event selection */
typedef struct changelog_ev_selector {
- gf_lock_t reflock;
+ gf_lock_t reflock;
- /**
- * Array of references for each selection bit.
- */
- unsigned int ref[CHANGELOG_EV_SELECTION_RANGE];
+ /**
+ * Array of references for each selection bit.
+ */
+ unsigned int ref[CHANGELOG_EV_SELECTION_RANGE];
} changelog_ev_selector_t;
-
/* changelog's private structure */
struct changelog_priv {
- gf_boolean_t active;
+ /* changelog journalling */
+ gf_boolean_t active;
+
+ /* changelog live notifications */
+ gf_boolean_t rpc_active;
+
+ /* to generate unique socket file per brick */
+ char *changelog_brick;
+
+ /* logging directory */
+ char *changelog_dir;
- /* to generate unique socket file per brick */
- char *changelog_brick;
+ /* htime directory */
+ char *htime_dir;
- /* logging directory */
- char *changelog_dir;
+ /* one file for all changelog types */
+ int changelog_fd;
- /* htime directory */
- char *htime_dir;
+ /* htime fd for current changelog session */
+ int htime_fd;
- /* one file for all changelog types */
- int changelog_fd;
+ /* c_snap_fd is fd for call-path changelog */
+ int c_snap_fd;
- /* htime fd for current changelog session */
- int htime_fd;
+ /* rollover_count used by htime */
+ int rollover_count;
- /* c_snap_fd is fd for call-path changelog */
- int c_snap_fd;
+ gf_lock_t lock;
- /* rollover_count used by htime */
- int rollover_count;
+ /* lock to synchronize CSNAP updation */
+ gf_lock_t c_snap_lock;
- gf_lock_t lock;
+ /* written end of the pipe */
+ int wfd;
- /* lock to synchronize CSNAP updation */
- gf_lock_t c_snap_lock;
+ /* rollover time */
+ int32_t rollover_time;
- /* written end of the pipe */
- int wfd;
+ /* fsync() interval */
+ int32_t fsync_interval;
- /* rollover time */
- int32_t rollover_time;
+ /* changelog type maps */
+ const char *maps[CHANGELOG_MAX_TYPE];
- /* fsync() interval */
- int32_t fsync_interval;
+ /* time slicer */
+ changelog_time_slice_t slice;
- /* changelog type maps */
- const char *maps[CHANGELOG_MAX_TYPE];
+ /* context of the updater */
+ changelog_dispatcher_t cd;
- /* time slicer */
- changelog_time_slice_t slice;
+ /* context of the rollover thread */
+ changelog_rollover_t cr;
- /* context of the updater */
- changelog_dispatcher_t cd;
+ /* context of fsync thread */
+ changelog_fsync_t cf;
- /* context of the rollover thread */
- changelog_rollover_t cr;
+ /* operation mode */
+ changelog_mode_t op_mode;
- /* context of fsync thread */
- changelog_fsync_t cf;
+ /* bootstrap routine for 'current' logger */
+ struct changelog_bootstrap *cb;
- /* operation mode */
- changelog_mode_t op_mode;
+ /* encoder mode */
+ changelog_encoder_t encode_mode;
- /* bootstrap routine for 'current' logger */
- struct changelog_bootstrap *cb;
+ /* encoder */
+ struct changelog_encoder *ce;
- /* encoder mode */
- changelog_encoder_t encode_mode;
+ /**
+ * snapshot dependency changes
+ */
- /* encoder */
- struct changelog_encoder *ce;
+ /* Draining of fops*/
+ drain_mgmt_t dm;
- /**
- * snapshot dependency changes
- */
+ /* Represents the active color. Initially by default black */
+ chlog_fop_color_t current_color;
- /* Draining of fops*/
- drain_mgmt_t dm;
+ /* flag to determine explicit rollover is triggered */
+ gf_boolean_t explicit_rollover;
- /* Represents the active color. Initially by default black */
- chlog_fop_color_t current_color;
+ /* barrier notification variable protected by mutex */
+ barrier_notify_t bn;
- /* flag to determine explicit rollover is triggered */
- gf_boolean_t explicit_rollover;
+ /* barrier on/off indicating flags */
+ barrier_flags_t bflags;
- /* barrier notification variable protected by mutex */
- barrier_notify_t bn;
+ /* changelog barrier on/off indicating flag */
+ gf_boolean_t barrier_enabled;
+ struct list_head queue;
+ uint32_t queue_size;
+ gf_timer_t *timer;
+ struct timespec timeout;
- /* barrier on/off indicating flags */
- barrier_flags_t bflags;
+ /**
+ * buffers, RPC, event selection, notifications and other
+ * beasts.
+ */
- /* changelog barrier on/off indicating flag */
- gf_boolean_t barrier_enabled;
- struct list_head queue;
- uint32_t queue_size;
- gf_timer_t *timer;
- struct timespec timeout;
+ /* epoll pthread */
+ pthread_t poller;
- /**
- * buffers, RPC, event selection, notifications and other
- * beasts.
- */
+ /* rotational buffer */
+ rbuf_t *rbuf;
- /* epoll pthread */
- pthread_t poller;
+ /* changelog RPC server */
+ rpcsvc_t *rpc;
- /* rotational buffer */
- rbuf_t *rbuf;
+ /* event selection */
+ changelog_ev_selector_t ev_selection;
- /* changelog RPC server */
- rpcsvc_t *rpc;
+ /* client handling (reverse connection) */
+ pthread_t connector;
- /* event selection */
- changelog_ev_selector_t ev_selection;
+ int nr_dispatchers;
+ pthread_t *ev_dispatcher;
- /* client handling (reverse connection) */
- pthread_t connector;
+ changelog_clnt_t connections;
- int nr_dispatchers;
- pthread_t *ev_dispatcher;
+ /* glusterfind dependency to capture paths on deleted entries*/
+ gf_boolean_t capture_del_path;
- changelog_clnt_t connections;
+ /* Save total no. of listners */
+ gf_atomic_t listnercnt;
- /* glusterfind dependency to capture paths on deleted entries*/
- gf_boolean_t capture_del_path;
+ /* Save total no. of xprt are associated with listner */
+ gf_atomic_t xprtcnt;
+
+ /* Save xprt list */
+ struct list_head xprt_list;
+
+ /* Save total no. of client connection */
+ gf_atomic_t clntcnt;
+
+ /* Save cleanup brick in victim */
+ xlator_t *victim;
+
+ /* Status to save cleanup notify status */
+ gf_boolean_t notify_down;
};
struct changelog_local {
- inode_t *inode;
- gf_boolean_t update_no_check;
+ inode_t *inode;
+ gf_boolean_t update_no_check;
- changelog_log_data_t cld;
+ changelog_log_data_t cld;
- /**
- * ->prev_entry is used in cases when there needs to be
- * additional changelog entry for the parent (eg. rename)
- * It's analogous to ->next in single linked list world,
- * but we call it as ->prev_entry... ha ha ha
- */
- struct changelog_local *prev_entry;
+ /**
+ * ->prev_entry is used in cases when there needs to be
+ * additional changelog entry for the parent (eg. rename)
+ * It's analogous to ->next in single linked list world,
+ * but we call it as ->prev_entry... ha ha ha
+ */
+ struct changelog_local *prev_entry;
- /* snap dependency changes */
- chlog_fop_color_t color;
+ /* snap dependency changes */
+ chlog_fop_color_t color;
};
typedef struct changelog_local changelog_local_t;
/* inode version is stored in inode ctx */
typedef struct changelog_inode_ctx {
- unsigned long iversion[CHANGELOG_MAX_TYPE];
+ unsigned long iversion[CHANGELOG_MAX_TYPE];
} changelog_inode_ctx_t;
-#define CHANGELOG_INODE_VERSION_TYPE(ctx, type) &(ctx->iversion[type])
+#define CHANGELOG_INODE_VERSION_TYPE(ctx, type) &(ctx->iversion[type])
/**
* Optional Records:
@@ -344,269 +358,276 @@ typedef struct changelog_inode_ctx {
* @changelog_opt_t struct. The array is allocated via @iobufs.
*/
typedef enum {
- CHANGELOG_OPT_REC_FOP,
- CHANGELOG_OPT_REC_ENTRY,
- CHANGELOG_OPT_REC_UINT32,
+ CHANGELOG_OPT_REC_FOP,
+ CHANGELOG_OPT_REC_ENTRY,
+ CHANGELOG_OPT_REC_UINT32,
} changelog_optional_rec_type_t;
struct changelog_entry_fields {
- uuid_t cef_uuid;
- char *cef_bname;
- char *cef_path;
+ uuid_t cef_uuid;
+ char *cef_bname;
+ char *cef_path;
};
typedef struct {
- /**
- * @co_covert can be used to do post-processing of the record before
- * it's persisted to the CHANGELOG. If this is NULL, then the record
- * is persisted as per it's in memory format.
- */
- size_t (*co_convert) (void *data, char *buffer, gf_boolean_t encode);
-
- /* release routines */
- void (*co_free) (void *data);
-
- /* type of the field */
- changelog_optional_rec_type_t co_type;
-
- /**
- * sizeof of the 'valid' field in the union. This field is not used if
- * @co_convert is specified.
- */
- size_t co_len;
-
- union {
- unsigned int co_uint32;
- glusterfs_fop_t co_fop;
- struct changelog_entry_fields co_entry;
- };
+ /**
+ * @co_covert can be used to do post-processing of the record before
+ * it's persisted to the CHANGELOG. If this is NULL, then the record
+ * is persisted as per it's in memory format.
+ */
+ size_t (*co_convert)(void *data, char *buffer, gf_boolean_t encode);
+
+ /* release routines */
+ void (*co_free)(void *data);
+
+ /* type of the field */
+ changelog_optional_rec_type_t co_type;
+
+ /**
+ * sizeof of the 'valid' field in the union. This field is not used if
+ * @co_convert is specified.
+ */
+ size_t co_len;
+
+ union {
+ unsigned int co_uint32;
+ glusterfs_fop_t co_fop;
+ struct changelog_entry_fields co_entry;
+ };
} changelog_opt_t;
-#define CHANGELOG_OPT_RECORD_LEN sizeof (changelog_opt_t)
+#define CHANGELOG_OPT_RECORD_LEN sizeof(changelog_opt_t)
/**
* helpers routines
*/
int
-changelog_thread_cleanup (xlator_t *this, pthread_t thr_id);
+changelog_thread_cleanup(xlator_t *this, pthread_t thr_id);
void *
-changelog_get_usable_buffer (changelog_local_t *local);
+changelog_get_usable_buffer(changelog_local_t *local);
void
-changelog_set_usable_record_and_length (changelog_local_t *local,
- size_t len, int xr);
+changelog_set_usable_record_and_length(changelog_local_t *local, size_t len,
+ int xr);
void
-changelog_local_cleanup (xlator_t *xl, changelog_local_t *local);
+changelog_local_cleanup(xlator_t *xl, changelog_local_t *local);
changelog_local_t *
-changelog_local_init (xlator_t *this, inode_t *inode, uuid_t gfid,
- int xtra_records, gf_boolean_t update_flag);
-int
-changelog_start_next_change (xlator_t *this,
- changelog_priv_t *priv,
- unsigned long ts, gf_boolean_t finale);
+changelog_local_init(xlator_t *this, inode_t *inode, uuid_t gfid,
+ int xtra_records, gf_boolean_t update_flag);
int
-changelog_open_journal (xlator_t *this, changelog_priv_t *priv);
+changelog_start_next_change(xlator_t *this, changelog_priv_t *priv, time_t ts,
+ gf_boolean_t finale);
int
-changelog_fill_rollover_data (changelog_log_data_t *cld, gf_boolean_t is_last);
+changelog_open_journal(xlator_t *this, changelog_priv_t *priv);
+void
+changelog_fill_rollover_data(changelog_log_data_t *cld, gf_boolean_t is_last);
int
-changelog_inject_single_event (xlator_t *this,
- changelog_priv_t *priv,
- changelog_log_data_t *cld);
+changelog_inject_single_event(xlator_t *this, changelog_priv_t *priv,
+ changelog_log_data_t *cld);
size_t
-changelog_entry_length ();
+changelog_entry_length();
int
-changelog_write (int fd, char *buffer, size_t len);
+changelog_write(int fd, char *buffer, size_t len);
int
-changelog_write_change (changelog_priv_t *priv, char *buffer, size_t len);
+changelog_write_change(changelog_priv_t *priv, char *buffer, size_t len);
int
-changelog_handle_change (xlator_t *this,
- changelog_priv_t *priv, changelog_log_data_t *cld);
+changelog_handle_change(xlator_t *this, changelog_priv_t *priv,
+ changelog_log_data_t *cld);
void
-changelog_update (xlator_t *this, changelog_priv_t *priv,
- changelog_local_t *local, changelog_log_type type);
+changelog_update(xlator_t *this, changelog_priv_t *priv,
+ changelog_local_t *local, changelog_log_type type);
void *
-changelog_rollover (void *data);
+changelog_rollover(void *data);
void *
-changelog_fsync_thread (void *data);
+changelog_fsync_thread(void *data);
int
-changelog_forget (xlator_t *this, inode_t *inode);
+changelog_forget(xlator_t *this, inode_t *inode);
int
-htime_update (xlator_t *this, changelog_priv_t *priv,
- unsigned long ts, char * buffer);
+htime_update(xlator_t *this, changelog_priv_t *priv, time_t ts, char *buffer);
int
-htime_open (xlator_t *this, changelog_priv_t *priv, unsigned long ts);
+htime_open(xlator_t *this, changelog_priv_t *priv, time_t ts);
int
-htime_create (xlator_t *this, changelog_priv_t *priv, unsigned long ts);
+htime_create(xlator_t *this, changelog_priv_t *priv, time_t ts);
/* Geo-Rep snapshot dependency changes */
void
-changelog_color_fop_and_inc_cnt (xlator_t *this, changelog_priv_t *priv,
- changelog_local_t *local);
+changelog_color_fop_and_inc_cnt(xlator_t *this, changelog_priv_t *priv,
+ changelog_local_t *local);
void
-changelog_inc_fop_cnt (xlator_t *this, changelog_priv_t *priv,
- changelog_local_t *local);
+changelog_inc_fop_cnt(xlator_t *this, changelog_priv_t *priv,
+ changelog_local_t *local);
void
-changelog_dec_fop_cnt (xlator_t *this, changelog_priv_t *priv,
- changelog_local_t *local);
+changelog_dec_fop_cnt(xlator_t *this, changelog_priv_t *priv,
+ changelog_local_t *local);
int
-changelog_barrier_notify (changelog_priv_t *priv, char* buf);
+changelog_barrier_notify(changelog_priv_t *priv, char *buf);
void
-changelog_barrier_cleanup (xlator_t *this, changelog_priv_t *priv,
- struct list_head *queue);
+changelog_barrier_cleanup(xlator_t *this, changelog_priv_t *priv,
+ struct list_head *queue);
void
-changelog_drain_white_fops (xlator_t *this, changelog_priv_t *priv);
+changelog_drain_white_fops(xlator_t *this, changelog_priv_t *priv);
void
-changelog_drain_black_fops (xlator_t *this, changelog_priv_t *priv);
+changelog_drain_black_fops(xlator_t *this, changelog_priv_t *priv);
/* Crash consistency of changelog wrt snapshot */
int
-changelog_snap_logging_stop ( xlator_t *this, changelog_priv_t *priv);
+changelog_snap_logging_stop(xlator_t *this, changelog_priv_t *priv);
int
-changelog_snap_logging_start ( xlator_t *this, changelog_priv_t *priv);
+changelog_snap_logging_start(xlator_t *this, changelog_priv_t *priv);
int
-changelog_snap_open ( xlator_t *this, changelog_priv_t *priv);
+changelog_snap_open(xlator_t *this, changelog_priv_t *priv);
int
-changelog_snap_handle_ascii_change (xlator_t *this,
- changelog_log_data_t *cld);
+changelog_snap_handle_ascii_change(xlator_t *this, changelog_log_data_t *cld);
int
-changelog_snap_write_change (changelog_priv_t *priv, char *buffer, size_t len);
+changelog_snap_write_change(changelog_priv_t *priv, char *buffer, size_t len);
/* Changelog barrier routines */
-void __chlog_barrier_enqueue (xlator_t *this, call_stub_t *stub);
-void __chlog_barrier_disable (xlator_t *this, struct list_head *queue);
-void chlog_barrier_dequeue_all (xlator_t *this, struct list_head *queue);
-call_stub_t *__chlog_barrier_dequeue (xlator_t *this, struct list_head *queue);
-int __chlog_barrier_enable (xlator_t *this, changelog_priv_t *priv);
+void
+__chlog_barrier_enqueue(xlator_t *this, call_stub_t *stub);
+void
+__chlog_barrier_disable(xlator_t *this, struct list_head *queue);
+void
+chlog_barrier_dequeue_all(xlator_t *this, struct list_head *queue);
+call_stub_t *
+__chlog_barrier_dequeue(xlator_t *this, struct list_head *queue);
+int
+__chlog_barrier_enable(xlator_t *this, changelog_priv_t *priv);
int32_t
-changelog_fill_entry_buf (call_frame_t *frame, xlator_t *this,
- loc_t *loc, changelog_local_t **local);
+changelog_fill_entry_buf(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ changelog_local_t **local);
/* event selection routines */
-void changelog_select_event (xlator_t *,
- changelog_ev_selector_t *, unsigned int);
-void changelog_deselect_event (xlator_t *,
- changelog_ev_selector_t *, unsigned int);
-int changelog_init_event_selection (xlator_t *,
- changelog_ev_selector_t *);
-int changelog_cleanup_event_selection (xlator_t *,
- changelog_ev_selector_t *);
-int changelog_ev_selected (xlator_t *,
- changelog_ev_selector_t *, unsigned int);
void
-changelog_dispatch_event (xlator_t *, changelog_priv_t *, changelog_event_t *);
+changelog_select_event(xlator_t *, changelog_ev_selector_t *, unsigned int);
+void
+changelog_deselect_event(xlator_t *, changelog_ev_selector_t *, unsigned int);
+int
+changelog_init_event_selection(xlator_t *, changelog_ev_selector_t *);
+int
+changelog_ev_selected(xlator_t *, changelog_ev_selector_t *, unsigned int);
+void
+changelog_dispatch_event(xlator_t *, changelog_priv_t *, changelog_event_t *);
changelog_inode_ctx_t *
-__changelog_inode_ctx_get (xlator_t *, inode_t *, unsigned long **,
- unsigned long *, changelog_log_type);
+__changelog_inode_ctx_get(xlator_t *, inode_t *, unsigned long **,
+ unsigned long *, changelog_log_type);
int
-resolve_pargfid_to_path (xlator_t *this, const uuid_t gfid, char **path,
- char *bname);
+resolve_pargfid_to_path(xlator_t *this, const uuid_t gfid, char **path,
+ char *bname);
/* macros */
-#define CHANGELOG_STACK_UNWIND(fop, frame, params ...) do { \
- changelog_local_t *__local = NULL; \
- xlator_t *__xl = NULL; \
- if (frame) { \
- __local = frame->local; \
- __xl = frame->this; \
- frame->local = NULL; \
- } \
- STACK_UNWIND_STRICT (fop, frame, params); \
- if (__local && __local->prev_entry) \
- changelog_local_cleanup (__xl, \
- __local->prev_entry); \
- changelog_local_cleanup (__xl, __local); \
- } while (0)
-
-#define CHANGELOG_IOBUF_REF(iobuf) do { \
- if (iobuf) \
- iobuf_ref (iobuf); \
- } while (0)
-
-#define CHANGELOG_IOBUF_UNREF(iobuf) do { \
- if (iobuf) \
- iobuf_unref (iobuf); \
- } while (0)
-
-#define CHANGELOG_FILL_BUFFER(buffer, off, val, len) do { \
- memcpy (buffer + off, val, len); \
- off += len; \
- } while (0)
-
-#define SLICE_VERSION_UPDATE(slice) do { \
- int i = 0; \
- for (; i < CHANGELOG_MAX_TYPE; i++) { \
- slice->changelog_version[i]++; \
- } \
- } while (0)
-
-#define CHANGELOG_FILL_UINT32(co, number, converter, xlen) do { \
- co->co_convert = converter; \
- co->co_free = NULL; \
- co->co_type = CHANGELOG_OPT_REC_UINT32; \
- co->co_uint32 = number; \
- xlen += sizeof (unsigned int); \
- } while (0)
-
-#define CHANGLOG_FILL_FOP_NUMBER(co, fop, converter, xlen) do { \
- co->co_convert = converter; \
- co->co_free = NULL; \
- co->co_type = CHANGELOG_OPT_REC_FOP; \
- co->co_fop = fop; \
- xlen += sizeof (fop); \
- } while (0)
-
-#define CHANGELOG_FILL_ENTRY(co, pargfid, bname, \
- converter, freefn, xlen, label) \
- do { \
- co->co_convert = converter; \
- co->co_free = freefn; \
- co->co_type = CHANGELOG_OPT_REC_ENTRY; \
- gf_uuid_copy (co->co_entry.cef_uuid, pargfid); \
- co->co_entry.cef_bname = gf_strdup(bname); \
- if (!co->co_entry.cef_bname) \
- goto label; \
- xlen += (UUID_CANONICAL_FORM_LEN + strlen (bname)); \
- } while (0)
-
-#define CHANGELOG_FILL_ENTRY_DIR_PATH(co, pargfid, bname, converter, \
- del_freefn, xlen, label, capture_del) \
- do { \
- co->co_convert = converter; \
- co->co_free = del_freefn; \
- co->co_type = CHANGELOG_OPT_REC_ENTRY; \
- gf_uuid_copy (co->co_entry.cef_uuid, pargfid); \
- co->co_entry.cef_bname = gf_strdup(bname); \
- if (!co->co_entry.cef_bname) \
- goto label; \
- xlen += (UUID_CANONICAL_FORM_LEN + strlen (bname)); \
- if (!capture_del || resolve_pargfid_to_path (this, pargfid, \
- &(co->co_entry.cef_path), co->co_entry.cef_bname)) { \
- co->co_entry.cef_path = gf_strdup ("\0"); \
- xlen += 1; \
- } else { \
- xlen += (strlen (co->co_entry.cef_path)); \
- } \
- } while (0)
-
-#define CHANGELOG_INIT(this, local, inode, gfid, xrec) \
- local = changelog_local_init (this, inode, gfid, xrec, _gf_false)
-
-#define CHANGELOG_INIT_NOCHECK(this, local, inode, gfid, xrec) \
- local = changelog_local_init (this, inode, gfid, xrec, _gf_true)
-
-#define CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, label) do { \
- if (!priv->active) \
- goto label; \
- /* ignore rebalance process's activity. */ \
- if ((frame->root->pid == GF_CLIENT_PID_DEFRAG) || \
- (frame->root->pid == GF_CLIENT_PID_TIER_DEFRAG)) \
- goto label; \
- } while (0)
+#define CHANGELOG_STACK_UNWIND(fop, frame, params...) \
+ do { \
+ changelog_local_t *__local = NULL; \
+ xlator_t *__xl = NULL; \
+ if (frame) { \
+ __local = frame->local; \
+ __xl = frame->this; \
+ frame->local = NULL; \
+ } \
+ STACK_UNWIND_STRICT(fop, frame, params); \
+ if (__local && __local->prev_entry) \
+ changelog_local_cleanup(__xl, __local->prev_entry); \
+ changelog_local_cleanup(__xl, __local); \
+ } while (0)
+
+#define CHANGELOG_IOBUF_REF(iobuf) \
+ do { \
+ if (iobuf) \
+ iobuf_ref(iobuf); \
+ } while (0)
+
+#define CHANGELOG_IOBUF_UNREF(iobuf) \
+ do { \
+ if (iobuf) \
+ iobuf_unref(iobuf); \
+ } while (0)
+
+#define CHANGELOG_FILL_BUFFER(buffer, off, val, len) \
+ do { \
+ memcpy(buffer + off, val, len); \
+ off += len; \
+ } while (0)
+
+#define SLICE_VERSION_UPDATE(slice) \
+ do { \
+ int i = 0; \
+ for (; i < CHANGELOG_MAX_TYPE; i++) { \
+ slice->changelog_version[i]++; \
+ } \
+ } while (0)
+
+#define CHANGELOG_FILL_UINT32(co, number, converter, xlen) \
+ do { \
+ co->co_convert = converter; \
+ co->co_free = NULL; \
+ co->co_type = CHANGELOG_OPT_REC_UINT32; \
+ co->co_uint32 = number; \
+ xlen += sizeof(unsigned int); \
+ } while (0)
+
+#define CHANGLOG_FILL_FOP_NUMBER(co, fop, converter, xlen) \
+ do { \
+ co->co_convert = converter; \
+ co->co_free = NULL; \
+ co->co_type = CHANGELOG_OPT_REC_FOP; \
+ co->co_fop = fop; \
+ xlen += sizeof(fop); \
+ } while (0)
+
+#define CHANGELOG_FILL_ENTRY(co, pargfid, bname, converter, freefn, xlen, \
+ label) \
+ do { \
+ co->co_convert = converter; \
+ co->co_free = freefn; \
+ co->co_type = CHANGELOG_OPT_REC_ENTRY; \
+ gf_uuid_copy(co->co_entry.cef_uuid, pargfid); \
+ co->co_entry.cef_bname = gf_strdup(bname); \
+ if (!co->co_entry.cef_bname) \
+ goto label; \
+ xlen += (UUID_CANONICAL_FORM_LEN + strlen(bname)); \
+ } while (0)
+
+#define CHANGELOG_FILL_ENTRY_DIR_PATH(co, pargfid, bname, converter, \
+ del_freefn, xlen, label, capture_del) \
+ do { \
+ co->co_convert = converter; \
+ co->co_free = del_freefn; \
+ co->co_type = CHANGELOG_OPT_REC_ENTRY; \
+ gf_uuid_copy(co->co_entry.cef_uuid, pargfid); \
+ co->co_entry.cef_bname = gf_strdup(bname); \
+ if (!co->co_entry.cef_bname) \
+ goto label; \
+ xlen += (UUID_CANONICAL_FORM_LEN + strlen(bname)); \
+ if (!capture_del || \
+ resolve_pargfid_to_path(this, pargfid, &(co->co_entry.cef_path), \
+ co->co_entry.cef_bname)) { \
+ co->co_entry.cef_path = gf_strdup("\0"); \
+ xlen += 1; \
+ } else { \
+ xlen += (strlen(co->co_entry.cef_path)); \
+ } \
+ } while (0)
+
+#define CHANGELOG_INIT(this, local, inode, gfid, xrec) \
+ local = changelog_local_init(this, inode, gfid, xrec, _gf_false)
+
+#define CHANGELOG_INIT_NOCHECK(this, local, inode, gfid, xrec) \
+ local = changelog_local_init(this, inode, gfid, xrec, _gf_true)
+
+#define CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, label) \
+ do { \
+ if (!priv->active) \
+ goto label; \
+ /* ignore rebalance process's activity. */ \
+ if ((frame->root->pid == GF_CLIENT_PID_DEFRAG) || \
+ (frame->root->pid == GF_CLIENT_PID_TIER_DEFRAG)) \
+ goto label; \
+ } while (0)
/* If it is a METADATA entry and fop num being GF_FOP_NULL, don't
* log in the changelog as it is of no use. And also if it is
@@ -615,66 +636,81 @@ resolve_pargfid_to_path (xlator_t *this, const uuid_t gfid, char **path,
* to same changelog will be missed. Hence check for boundary
* condition.
*/
-#define CHANGELOG_OP_BOUNDARY_CHECK(frame, label) do { \
- if (frame->root->op <= GF_FOP_NULL || \
- frame->root->op >= GF_FOP_MAXVALUE) \
- goto label; \
- } while (0)
+#define CHANGELOG_OP_BOUNDARY_CHECK(frame, label) \
+ do { \
+ if (frame->root->op <= GF_FOP_NULL || \
+ frame->root->op >= GF_FOP_MAXVALUE) \
+ goto label; \
+ } while (0)
/**
* ignore internal fops for all clients except AFR self-heal daemon
*/
-#define CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO(frame, dict, label) do { \
- if ((frame->root->pid != GF_CLIENT_PID_SELF_HEALD) \
- && dict \
- && dict_get (dict, GLUSTERFS_INTERNAL_FOP_KEY)) \
- goto label; \
- } while (0)
-
-#define CHANGELOG_COND_GOTO(priv, cond, label) do { \
- if (!priv->active || cond) \
- goto label; \
- } while (0)
+#define CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO(frame, dict, label) \
+ do { \
+ if ((frame->root->pid != GF_CLIENT_PID_SELF_HEALD) && dict && \
+ dict_get(dict, GLUSTERFS_INTERNAL_FOP_KEY)) \
+ goto label; \
+ } while (0)
+
+#define CHANGELOG_COND_GOTO(priv, cond, label) \
+ do { \
+ if (!priv->active || cond) \
+ goto label; \
+ } while (0)
/* Begin: Geo-Rep snapshot dependency changes */
-#define DICT_ERROR -1
-#define BARRIER_OFF 0
-#define BARRIER_ON 1
-#define DICT_DEFAULT 2
-
-#define CHANGELOG_NOT_ON_THEN_GOTO(priv, ret, label) do { \
- if (!priv->active) { \
- gf_msg (this->name, GF_LOG_WARNING, 0, \
- CHANGELOG_MSG_NOT_ACTIVE, \
- "Changelog is not active, return success"); \
- ret = 0; \
- goto label; \
- } \
- } while (0)
+#define DICT_ERROR -1
+#define BARRIER_OFF 0
+#define BARRIER_ON 1
+#define DICT_DEFAULT 2
+
+#define CHANGELOG_NOT_ON_THEN_GOTO(priv, ret, label) \
+ do { \
+ if (!priv->active) { \
+ gf_smsg(this->name, GF_LOG_WARNING, 0, \
+ CHANGELOG_MSG_CHANGELOG_NOT_ACTIVE, NULL); \
+ ret = 0; \
+ goto label; \
+ } \
+ } while (0)
/* Log pthread error and goto label */
-#define CHANGELOG_PTHREAD_ERROR_HANDLE_0(ret, label) do { \
- if (ret) { \
- gf_smsg (this->name, GF_LOG_ERROR, \
- 0, CHANGELOG_MSG_PTHREAD_ERROR, \
- "pthread error", "error=%d", ret, NULL); \
- ret = -1; \
- goto label; \
- } \
- } while (0);
+#define CHANGELOG_PTHREAD_ERROR_HANDLE_0(ret, label) \
+ do { \
+ if (ret) { \
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_PTHREAD_ERROR, \
+ "error=%d", ret, NULL); \
+ ret = -1; \
+ goto label; \
+ } \
+ } while (0);
/* Log pthread error, set flag and goto label */
-#define CHANGELOG_PTHREAD_ERROR_HANDLE_1(ret, label, flag) do { \
- if (ret) { \
- gf_smsg (this->name, GF_LOG_ERROR, 0, \
- CHANGELOG_MSG_PTHREAD_ERROR, \
- "pthread error", "error=%d", ret, NULL); \
- ret = -1; \
- flag = _gf_true; \
- goto label; \
- } \
- } while (0)
+#define CHANGELOG_PTHREAD_ERROR_HANDLE_1(ret, label, flag) \
+ do { \
+ if (ret) { \
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_PTHREAD_ERROR, \
+ "error=%d", ret, NULL); \
+ ret = -1; \
+ flag = _gf_true; \
+ goto label; \
+ } \
+ } while (0)
+
+/* Log pthread error, unlock mutex and goto label */
+#define CHANGELOG_PTHREAD_ERROR_HANDLE_2(ret, label, mutex) \
+ do { \
+ if (ret) { \
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_PTHREAD_ERROR, \
+ "error=%d", ret, NULL); \
+ ret = -1; \
+ pthread_mutex_unlock(&mutex); \
+ goto label; \
+ } \
+ } while (0)
+
/* End: Geo-Rep snapshot dependency changes */
#endif /* _CHANGELOG_HELPERS_H */
diff --git a/xlators/features/changelog/src/changelog-mem-types.h b/xlators/features/changelog/src/changelog-mem-types.h
index 33fea31b979..a2d8a9cbe93 100644
--- a/xlators/features/changelog/src/changelog-mem-types.h
+++ b/xlators/features/changelog/src/changelog-mem-types.h
@@ -11,24 +11,24 @@
#ifndef _CHANGELOG_MEM_TYPES_H
#define _CHANGELOG_MEM_TYPES_H
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_changelog_mem_types {
- gf_changelog_mt_priv_t = gf_common_mt_end + 1,
- gf_changelog_mt_str_t = gf_common_mt_end + 2,
- gf_changelog_mt_batch_t = gf_common_mt_end + 3,
- gf_changelog_mt_rt_t = gf_common_mt_end + 4,
- gf_changelog_mt_inode_ctx_t = gf_common_mt_end + 5,
- gf_changelog_mt_rpc_clnt_t = gf_common_mt_end + 6,
- gf_changelog_mt_libgfchangelog_t = gf_common_mt_end + 7,
- gf_changelog_mt_libgfchangelog_entry_t = gf_common_mt_end + 8,
- gf_changelog_mt_libgfchangelog_rl_t = gf_common_mt_end + 9,
- gf_changelog_mt_changelog_buffer_t = gf_common_mt_end + 10,
- gf_changelog_mt_history_data_t = gf_common_mt_end + 11,
- gf_changelog_mt_libgfchangelog_call_pool_t = gf_common_mt_end + 12,
- gf_changelog_mt_libgfchangelog_event_t = gf_common_mt_end + 13,
- gf_changelog_mt_ev_dispatcher_t = gf_common_mt_end + 14,
- gf_changelog_mt_end
+ gf_changelog_mt_priv_t = gf_common_mt_end + 1,
+ gf_changelog_mt_str_t = gf_common_mt_end + 2,
+ gf_changelog_mt_batch_t = gf_common_mt_end + 3,
+ gf_changelog_mt_rt_t = gf_common_mt_end + 4,
+ gf_changelog_mt_inode_ctx_t = gf_common_mt_end + 5,
+ gf_changelog_mt_rpc_clnt_t = gf_common_mt_end + 6,
+ gf_changelog_mt_libgfchangelog_t = gf_common_mt_end + 7,
+ gf_changelog_mt_libgfchangelog_entry_t = gf_common_mt_end + 8,
+ gf_changelog_mt_libgfchangelog_rl_t = gf_common_mt_end + 9,
+ gf_changelog_mt_changelog_buffer_t = gf_common_mt_end + 10,
+ gf_changelog_mt_history_data_t = gf_common_mt_end + 11,
+ gf_changelog_mt_libgfchangelog_call_pool_t = gf_common_mt_end + 12,
+ gf_changelog_mt_libgfchangelog_event_t = gf_common_mt_end + 13,
+ gf_changelog_mt_ev_dispatcher_t = gf_common_mt_end + 14,
+ gf_changelog_mt_end
};
#endif
diff --git a/xlators/features/changelog/src/changelog-messages.h b/xlators/features/changelog/src/changelog-messages.h
index bbef2f2bde5..cb0e16c85d8 100644
--- a/xlators/features/changelog/src/changelog-messages.h
+++ b/xlators/features/changelog/src/changelog-messages.h
@@ -11,7 +11,7 @@
#ifndef _CHANGELOG_MESSAGES_H_
#define _CHANGELOG_MESSAGES_H_
-#include "glfs-message-id.h"
+#include <glusterfs/glfs-message-id.h>
/* To add new message IDs, append new identifiers at the end of the list.
*
@@ -23,63 +23,150 @@
* glfs-message-id.h.
*/
-GLFS_MSGID(CHANGELOG,
- CHANGELOG_MSG_OPEN_FAILED,
- CHANGELOG_MSG_NO_MEMORY,
- CHANGELOG_MSG_VOL_MISCONFIGURED,
- CHANGELOG_MSG_RENAME_ERROR,
- CHANGELOG_MSG_READ_ERROR,
- CHANGELOG_MSG_HTIME_ERROR,
- CHANGELOG_MSG_PTHREAD_MUTEX_INIT_FAILED,
- CHANGELOG_MSG_PTHREAD_COND_INIT_FAILED,
- CHANGELOG_MSG_CHILD_MISCONFIGURED,
- CHANGELOG_MSG_DIR_OPTIONS_NOT_SET,
- CHANGELOG_MSG_CLOSE_ERROR,
- CHANGELOG_MSG_PIPE_CREATION_ERROR,
- CHANGELOG_MSG_DICT_GET_FAILED,
- CHANGELOG_MSG_BARRIER_INFO,
- CHANGELOG_MSG_BARRIER_ERROR,
- CHANGELOG_MSG_GET_TIME_OP_FAILED,
- CHANGELOG_MSG_WRITE_FAILED,
- CHANGELOG_MSG_PTHREAD_ERROR,
- CHANGELOG_MSG_INODE_NOT_FOUND,
- CHANGELOG_MSG_FSYNC_OP_FAILED,
- CHANGELOG_MSG_TOTAL_LOG_INFO,
- CHANGELOG_MSG_SNAP_INFO,
- CHANGELOG_MSG_SELECT_FAILED,
- CHANGELOG_MSG_FCNTL_FAILED,
- CHANGELOG_MSG_BNOTIFY_INFO,
- CHANGELOG_MSG_ENTRY_BUF_INFO,
- CHANGELOG_MSG_NOT_ACTIVE,
- CHANGELOG_MSG_LOCAL_INIT_FAILED,
- CHANGELOG_MSG_NOTIFY_REGISTER_FAILED,
- CHANGELOG_MSG_PROGRAM_NAME_REG_FAILED,
- CHANGELOG_MSG_HANDLE_PROBE_ERROR,
- CHANGELOG_MSG_SET_FD_CONTEXT,
- CHANGELOG_MSG_FREEUP_FAILED,
- CHANGELOG_MSG_HTIME_INFO,
- CHANGELOG_MSG_RPC_SUBMIT_REPLY_FAILED,
- CHANGELOG_MSG_RPC_BUILD_ERROR,
- CHANGELOG_MSG_RPC_CONNECT_ERROR,
- CHANGELOG_MSG_RPC_START_ERROR,
- CHANGELOG_MSG_BUFFER_STARVATION_ERROR,
- CHANGELOG_MSG_SCAN_DIR_FAILED,
- CHANGELOG_MSG_FSETXATTR_FAILED,
- CHANGELOG_MSG_FGETXATTR_FAILED,
- CHANGELOG_MSG_CLEANUP_ON_ACTIVE_REF,
- CHANGELOG_MSG_DISPATCH_EVENT_FAILED,
- CHANGELOG_MSG_PUT_BUFFER_FAILED,
- CHANGELOG_MSG_PTHREAD_COND_WAIT_FAILED,
- CHANGELOG_MSG_PTHREAD_CANCEL_FAILED,
- CHANGELOG_MSG_INJECT_FSYNC_FAILED,
- CHANGELOG_MSG_CREATE_FRAME_FAILED,
- CHANGELOG_MSG_FSTAT_OP_FAILED,
- CHANGELOG_MSG_LSEEK_OP_FAILED,
- CHANGELOG_MSG_STRSTR_OP_FAILED,
- CHANGELOG_MSG_UNLINK_OP_FAILED,
- CHANGELOG_MSG_DETECT_EMPTY_CHANGELOG_FAILED,
- CHANGELOG_MSG_READLINK_OP_FAILED,
- CHANGELOG_MSG_EXPLICIT_ROLLOVER_FAILED
-);
+GLFS_MSGID(
+ CHANGELOG, CHANGELOG_MSG_OPEN_FAILED, CHANGELOG_MSG_BARRIER_FOP_FAILED,
+ CHANGELOG_MSG_VOL_MISCONFIGURED, CHANGELOG_MSG_RENAME_ERROR,
+ CHANGELOG_MSG_READ_ERROR, CHANGELOG_MSG_HTIME_ERROR,
+ CHANGELOG_MSG_PTHREAD_MUTEX_INIT_FAILED,
+ CHANGELOG_MSG_PTHREAD_COND_INIT_FAILED, CHANGELOG_MSG_CHILD_MISCONFIGURED,
+ CHANGELOG_MSG_DIR_OPTIONS_NOT_SET, CHANGELOG_MSG_CLOSE_ERROR,
+ CHANGELOG_MSG_PIPE_CREATION_ERROR, CHANGELOG_MSG_DICT_GET_FAILED,
+ CHANGELOG_MSG_BARRIER_INFO, CHANGELOG_MSG_BARRIER_ERROR,
+ CHANGELOG_MSG_GET_TIME_OP_FAILED, CHANGELOG_MSG_WRITE_FAILED,
+ CHANGELOG_MSG_PTHREAD_ERROR, CHANGELOG_MSG_INODE_NOT_FOUND,
+ CHANGELOG_MSG_FSYNC_OP_FAILED, CHANGELOG_MSG_TOTAL_LOG_INFO,
+ CHANGELOG_MSG_SNAP_INFO, CHANGELOG_MSG_SELECT_FAILED,
+ CHANGELOG_MSG_FCNTL_FAILED, CHANGELOG_MSG_BNOTIFY_INFO,
+ CHANGELOG_MSG_ENTRY_BUF_INFO, CHANGELOG_MSG_CHANGELOG_NOT_ACTIVE,
+ CHANGELOG_MSG_LOCAL_INIT_FAILED, CHANGELOG_MSG_NOTIFY_REGISTER_FAILED,
+ CHANGELOG_MSG_PROGRAM_NAME_REG_FAILED, CHANGELOG_MSG_HANDLE_PROBE_ERROR,
+ CHANGELOG_MSG_SET_FD_CONTEXT, CHANGELOG_MSG_FREEUP_FAILED,
+ CHANGELOG_MSG_RECONFIGURE, CHANGELOG_MSG_RPC_SUBMIT_REPLY_FAILED,
+ CHANGELOG_MSG_RPC_BUILD_ERROR, CHANGELOG_MSG_RPC_CONNECT_ERROR,
+ CHANGELOG_MSG_RPC_START_ERROR, CHANGELOG_MSG_BUFFER_STARVATION_ERROR,
+ CHANGELOG_MSG_SCAN_DIR_FAILED, CHANGELOG_MSG_FSETXATTR_FAILED,
+ CHANGELOG_MSG_FGETXATTR_FAILED, CHANGELOG_MSG_CLEANUP_ON_ACTIVE_REF,
+ CHANGELOG_MSG_DISPATCH_EVENT_FAILED, CHANGELOG_MSG_PUT_BUFFER_FAILED,
+ CHANGELOG_MSG_PTHREAD_COND_WAIT_FAILED, CHANGELOG_MSG_PTHREAD_CANCEL_FAILED,
+ CHANGELOG_MSG_INJECT_FSYNC_FAILED, CHANGELOG_MSG_CREATE_FRAME_FAILED,
+ CHANGELOG_MSG_FSTAT_OP_FAILED, CHANGELOG_MSG_LSEEK_OP_FAILED,
+ CHANGELOG_MSG_STRSTR_OP_FAILED, CHANGELOG_MSG_UNLINK_OP_FAILED,
+ CHANGELOG_MSG_DETECT_EMPTY_CHANGELOG_FAILED,
+ CHANGELOG_MSG_READLINK_OP_FAILED, CHANGELOG_MSG_EXPLICIT_ROLLOVER_FAILED,
+ CHANGELOG_MSG_RPCSVC_NOTIFY_FAILED, CHANGELOG_MSG_MEMORY_INIT_FAILED,
+ CHANGELOG_MSG_NO_MEMORY, CHANGELOG_MSG_HTIME_STAT_ERROR,
+ CHANGELOG_MSG_HTIME_CURRENT_ERROR, CHANGELOG_MSG_BNOTIFY_COND_INFO,
+ CHANGELOG_MSG_NO_HTIME_CURRENT, CHANGELOG_MSG_HTIME_CURRENT,
+ CHANGELOG_MSG_NEW_HTIME_FILE, CHANGELOG_MSG_MKDIR_ERROR,
+ CHANGELOG_MSG_PATH_NOT_FOUND, CHANGELOG_MSG_XATTR_INIT_FAILED,
+ CHANGELOG_MSG_WROTE_TO_CSNAP, CHANGELOG_MSG_UNUSED_0,
+ CHANGELOG_MSG_GET_BUFFER_FAILED, CHANGELOG_MSG_BARRIER_STATE_NOTIFY,
+ CHANGELOG_MSG_BARRIER_DISABLED, CHANGELOG_MSG_BARRIER_ALREADY_DISABLED,
+ CHANGELOG_MSG_BARRIER_ON_ERROR, CHANGELOG_MSG_BARRIER_ENABLE,
+ CHANGELOG_MSG_BARRIER_KEY_NOT_FOUND, CHANGELOG_MSG_ERROR_IN_DICT_GET,
+ CHANGELOG_MSG_UNUSED_1, CHANGELOG_MSG_UNUSED_2,
+ CHANGELOG_MSG_DEQUEUING_BARRIER_FOPS,
+ CHANGELOG_MSG_DEQUEUING_BARRIER_FOPS_FINISHED,
+ CHANGELOG_MSG_BARRIER_TIMEOUT, CHANGELOG_MSG_TIMEOUT_ADD_FAILED,
+ CHANGELOG_MSG_CLEANUP_ALREADY_SET);
+#define CHANGELOG_MSG_BARRIER_FOP_FAILED_STR \
+ "failed to barrier FOPs, disabling changelog barrier"
+#define CHANGELOG_MSG_MEMORY_INIT_FAILED_STR "memory accounting init failed"
+#define CHANGELOG_MSG_NO_MEMORY_STR "failed to create local memory pool"
+#define CHANGELOG_MSG_ENTRY_BUF_INFO_STR \
+ "Entry cannot be captured for gfid, Capturing DATA entry."
+#define CHANGELOG_MSG_PTHREAD_ERROR_STR "pthread error"
+#define CHANGELOG_MSG_PTHREAD_MUTEX_INIT_FAILED_STR "pthread_mutex_init failed"
+#define CHANGELOG_MSG_PTHREAD_COND_INIT_FAILED_STR "pthread_cond_init failed"
+#define CHANGELOG_MSG_HTIME_ERROR_STR "failed to update HTIME file"
+#define CHANGELOG_MSG_HTIME_STAT_ERROR_STR "unable to stat htime file"
+#define CHANGELOG_MSG_HTIME_CURRENT_ERROR_STR "Error extracting HTIME_CURRENT."
+#define CHANGELOG_MSG_UNLINK_OP_FAILED_STR "error unlinking empty changelog"
+#define CHANGELOG_MSG_RENAME_ERROR_STR "error renaming"
+#define CHANGELOG_MSG_MKDIR_ERROR_STR "unable to create directory"
+#define CHANGELOG_MSG_BNOTIFY_INFO_STR \
+ "Explicit rollover changelog signaling bnotify"
+#define CHANGELOG_MSG_BNOTIFY_COND_INFO_STR "Woke up: bnotify conditional wait"
+#define CHANGELOG_MSG_RECONFIGURE_STR "Reconfigure: Changelog Enable"
+#define CHANGELOG_MSG_NO_HTIME_CURRENT_STR \
+ "HTIME_CURRENT not found. Changelog enabled before init"
+#define CHANGELOG_MSG_HTIME_CURRENT_STR "HTIME_CURRENT"
+#define CHANGELOG_MSG_NEW_HTIME_FILE_STR \
+ "Changelog enable: Creating new HTIME file"
+#define CHANGELOG_MSG_FGETXATTR_FAILED_STR "fgetxattr failed"
+#define CHANGELOG_MSG_TOTAL_LOG_INFO_STR "changelog info"
+#define CHANGELOG_MSG_PTHREAD_COND_WAIT_FAILED_STR "pthread cond wait failed"
+#define CHANGELOG_MSG_INODE_NOT_FOUND_STR "inode not found"
+#define CHANGELOG_MSG_READLINK_OP_FAILED_STR \
+ "could not read the link from the gfid handle"
+#define CHANGELOG_MSG_OPEN_FAILED_STR "unable to open file"
+#define CHANGELOG_MSG_RPC_CONNECT_ERROR_STR "failed to connect back"
+#define CHANGELOG_MSG_BUFFER_STARVATION_ERROR_STR \
+ "Failed to get buffer for RPC dispatch"
+#define CHANGELOG_MSG_PTHREAD_CANCEL_FAILED_STR "could not cancel thread"
+#define CHANGELOG_MSG_FSTAT_OP_FAILED_STR "Could not stat (CHANGELOG)"
+#define CHANGELOG_MSG_LSEEK_OP_FAILED_STR "Could not lseek (changelog)"
+#define CHANGELOG_MSG_PATH_NOT_FOUND_STR \
+ "Could not find CHANGELOG in changelog path"
+#define CHANGELOG_MSG_FSYNC_OP_FAILED_STR "fsync failed"
+#define CHANGELOG_MSG_DETECT_EMPTY_CHANGELOG_FAILED_STR \
+ "Error detecting empty changelog"
+#define CHANGELOG_MSG_EXPLICIT_ROLLOVER_FAILED_STR \
+ "Fail snapshot because of previous errors"
+#define CHANGELOG_MSG_SCAN_DIR_FAILED_STR "scandir failed"
+#define CHANGELOG_MSG_FSETXATTR_FAILED_STR "fsetxattr failed"
+#define CHANGELOG_MSG_XATTR_INIT_FAILED_STR "Htime xattr initialization failed"
+#define CHANGELOG_MSG_SNAP_INFO_STR "log in call path"
+#define CHANGELOG_MSG_WRITE_FAILED_STR "error writing to disk"
+#define CHANGELOG_MSG_WROTE_TO_CSNAP_STR "Successfully wrote to csnap"
+#define CHANGELOG_MSG_GET_TIME_OP_FAILED_STR "Problem rolling over changelog(s)"
+#define CHANGELOG_MSG_BARRIER_INFO_STR "Explicit wakeup on barrier notify"
+#define CHANGELOG_MSG_SELECT_FAILED_STR "pthread_cond_timedwait failed"
+#define CHANGELOG_MSG_INJECT_FSYNC_FAILED_STR "failed to inject fsync event"
+#define CHANGELOG_MSG_LOCAL_INIT_FAILED_STR \
+ "changelog local initialization failed"
+#define CHANGELOG_MSG_GET_BUFFER_FAILED_STR "Failed to get buffer"
+#define CHANGELOG_MSG_SET_FD_CONTEXT_STR \
+ "could not set fd context(for release cbk)"
+#define CHANGELOG_MSG_DICT_GET_FAILED_STR "Barrier failed"
+#define CHANGELOG_MSG_BARRIER_STATE_NOTIFY_STR "Barrier notification"
+#define CHANGELOG_MSG_BARRIER_ERROR_STR \
+ "Received another barrier off notification while already off"
+#define CHANGELOG_MSG_BARRIER_DISABLED_STR "disabled changelog barrier"
+#define CHANGELOG_MSG_BARRIER_ALREADY_DISABLED_STR \
+ "Changelog barrier already disabled"
+#define CHANGELOG_MSG_BARRIER_ON_ERROR_STR \
+ "Received another barrier on notification when last one is not served yet"
+#define CHANGELOG_MSG_BARRIER_ENABLE_STR "Enabled changelog barrier"
+#define CHANGELOG_MSG_BARRIER_KEY_NOT_FOUND_STR "barrier key not found"
+#define CHANGELOG_MSG_ERROR_IN_DICT_GET_STR \
+ "Something went wrong in dict_get_str_boolean"
+#define CHANGELOG_MSG_DIR_OPTIONS_NOT_SET_STR "changelog-dir option is not set"
+#define CHANGELOG_MSG_FREEUP_FAILED_STR "could not cleanup bootstrapper"
+#define CHANGELOG_MSG_CHILD_MISCONFIGURED_STR \
+ "translator needs a single subvolume"
+#define CHANGELOG_MSG_VOL_MISCONFIGURED_STR \
+ "dangling volume. please check volfile"
+#define CHANGELOG_MSG_DEQUEUING_BARRIER_FOPS_STR \
+ "Dequeuing all the changelog barriered fops"
+#define CHANGELOG_MSG_DEQUEUING_BARRIER_FOPS_FINISHED_STR \
+ "Dequeuing changelog barriered fops is finished"
+#define CHANGELOG_MSG_BARRIER_TIMEOUT_STR \
+ "Disabling changelog barrier because of the timeout"
+#define CHANGELOG_MSG_TIMEOUT_ADD_FAILED_STR \
+ "Couldn't add changelog barrier timeout event"
+#define CHANGELOG_MSG_RPC_BUILD_ERROR_STR "failed to build rpc options"
+#define CHANGELOG_MSG_NOTIFY_REGISTER_FAILED_STR "failed to register notify"
+#define CHANGELOG_MSG_RPC_START_ERROR_STR "failed to start rpc"
+#define CHANGELOG_MSG_CREATE_FRAME_FAILED_STR "failed to create frame"
+#define CHANGELOG_MSG_RPC_SUBMIT_REPLY_FAILED_STR "failed to serialize reply"
+#define CHANGELOG_MSG_PROGRAM_NAME_REG_FAILED_STR "cannot register program"
+#define CHANGELOG_MSG_CHANGELOG_NOT_ACTIVE_STR \
+ "Changelog is not active, return success"
+#define CHANGELOG_MSG_PUT_BUFFER_FAILED_STR \
+ "failed to put buffer after consumption"
+#define CHANGELOG_MSG_CLEANUP_ALREADY_SET_STR \
+ "cleanup_starting flag is already set for xl"
+#define CHANGELOG_MSG_HANDLE_PROBE_ERROR_STR "xdr decoding error"
#endif /* !_CHANGELOG_MESSAGES_H_ */
diff --git a/xlators/features/changelog/src/changelog-misc.h b/xlators/features/changelog/src/changelog-misc.h
index 93af201879e..e2addc09414 100644
--- a/xlators/features/changelog/src/changelog-misc.h
+++ b/xlators/features/changelog/src/changelog-misc.h
@@ -11,10 +11,10 @@
#ifndef _CHANGELOG_MISC_H
#define _CHANGELOG_MISC_H
-#include "glusterfs.h"
-#include "common-utils.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/common-utils.h>
-#define CHANGELOG_MAX_TYPE 4
+#define CHANGELOG_MAX_TYPE 4
#define CHANGELOG_FILE_NAME "CHANGELOG"
#define HTIME_FILE_NAME "HTIME"
#define CSNAP_FILE_NAME "CHANGELOG.SNAP"
@@ -22,111 +22,110 @@
#define HTIME_CURRENT "trusted.glusterfs.current_htime"
#define HTIME_INITIAL_VALUE "0:0"
-#define CHANGELOG_VERSION_MAJOR 1
-#define CHANGELOG_VERSION_MINOR 2
+#define CHANGELOG_VERSION_MAJOR 1
+#define CHANGELOG_VERSION_MINOR 2
-#define CHANGELOG_UNIX_SOCK DEFAULT_VAR_RUN_DIRECTORY"/changelog-%s.sock"
-#define CHANGELOG_TMP_UNIX_SOCK DEFAULT_VAR_RUN_DIRECTORY"/.%s%lu.sock"
+#define CHANGELOG_UNIX_SOCK DEFAULT_VAR_RUN_DIRECTORY "/changelog-%s.sock"
+#define CHANGELOG_TMP_UNIX_SOCK DEFAULT_VAR_RUN_DIRECTORY "/.%s%lu.sock"
/**
* header starts with the version and the format of the changelog.
* 'version' not much of a use now.
*/
-#define CHANGELOG_HEADER \
- "GlusterFS Changelog | version: v%d.%d | encoding : %d\n"
-
-#define CHANGELOG_MAKE_SOCKET_PATH(brick_path, sockpath, len) do { \
- char xxh64[GF_XXH64_DIGEST_LENGTH*2+1] = {0,}; \
- gf_xxh64_wrapper ((unsigned char *)brick_path, \
- strlen(brick_path), \
- GF_XXHSUM64_DEFAULT_SEED, xxh64); \
- (void) snprintf (sockpath, len, \
- CHANGELOG_UNIX_SOCK, xxh64); \
- } while (0)
-
-#define CHANGELOG_MAKE_TMP_SOCKET_PATH(brick_path, sockpath, len) do { \
- unsigned long pid = 0; \
- char xxh64[GF_XXH64_DIGEST_LENGTH*2+1] = {0,}; \
- pid = (unsigned long) getpid (); \
- gf_xxh64_wrapper ((unsigned char *)brick_path, \
- strlen(brick_path), \
- GF_XXHSUM64_DEFAULT_SEED, xxh64); \
- (void) snprintf (sockpath, \
- len, CHANGELOG_TMP_UNIX_SOCK, \
- xxh64, pid); \
- } while (0)
-
+#define CHANGELOG_HEADER \
+ "GlusterFS Changelog | version: v%d.%d | encoding : %d\n"
+
+#define CHANGELOG_MAKE_SOCKET_PATH(brick_path, sockpath, len) \
+ do { \
+ char xxh64[GF_XXH64_DIGEST_LENGTH * 2 + 1] = { \
+ 0, \
+ }; \
+ gf_xxh64_wrapper((unsigned char *)brick_path, strlen(brick_path), \
+ GF_XXHSUM64_DEFAULT_SEED, xxh64); \
+ (void)snprintf(sockpath, len, CHANGELOG_UNIX_SOCK, xxh64); \
+ } while (0)
+
+#define CHANGELOG_MAKE_TMP_SOCKET_PATH(brick_path, sockpath, len) \
+ do { \
+ unsigned long pid = 0; \
+ char xxh64[GF_XXH64_DIGEST_LENGTH * 2 + 1] = { \
+ 0, \
+ }; \
+ pid = (unsigned long)getpid(); \
+ gf_xxh64_wrapper((unsigned char *)brick_path, strlen(brick_path), \
+ GF_XXHSUM64_DEFAULT_SEED, xxh64); \
+ (void)snprintf(sockpath, len, CHANGELOG_TMP_UNIX_SOCK, xxh64, pid); \
+ } while (0)
/**
* ... used by libgfchangelog.
*/
-#define CHANGELOG_GET_HEADER_INFO(fd, buffer, len, enc, maj, min, elen) do { \
- FILE *fp; \
- int fd_dup; \
- \
- enc = -1; \
- maj = -1; \
- min = -1; \
- fd_dup = dup (fd); \
- \
- if (fd_dup != -1) { \
- fp = fdopen (fd_dup, "r"); \
- if (fp) { \
- if (fgets (buffer, len, fp)) { \
- elen = strlen (buffer); \
- sscanf (buffer, \
- CHANGELOG_HEADER, \
- &maj, &min, &enc); \
- } \
- fclose (fp); \
- } else { \
- sys_close (fd_dup); \
- } \
- } \
- } while (0)
-
-#define CHANGELOG_FILL_HTIME_DIR(changelog_dir, path) do { \
- strncpy (path, changelog_dir, sizeof (path) - 1); \
- strcat (path, "/htime"); \
- } while(0)
-
-#define CHANGELOG_FILL_CSNAP_DIR(changelog_dir, path) do { \
- strncpy (path, changelog_dir, sizeof (path) - 1); \
- strcat (path, "/csnap"); \
- } while(0)
+#define CHANGELOG_GET_HEADER_INFO(fd, buffer, len, enc, maj, min, elen) \
+ do { \
+ FILE *fp; \
+ int fd_dup; \
+ \
+ enc = -1; \
+ maj = -1; \
+ min = -1; \
+ fd_dup = dup(fd); \
+ \
+ if (fd_dup != -1) { \
+ fp = fdopen(fd_dup, "r"); \
+ if (fp) { \
+ if (fgets(buffer, len, fp)) { \
+ elen = strlen(buffer); \
+ sscanf(buffer, CHANGELOG_HEADER, &maj, &min, &enc); \
+ } \
+ fclose(fp); \
+ } else { \
+ sys_close(fd_dup); \
+ } \
+ } \
+ } while (0)
+
+#define CHANGELOG_FILL_HTIME_DIR(changelog_dir, path) \
+ do { \
+ snprintf(path, sizeof(path), "%s/htime", changelog_dir); \
+ } while (0)
+
+#define CHANGELOG_FILL_CSNAP_DIR(changelog_dir, path) \
+ do { \
+ snprintf(path, sizeof(path), "%s/csnap", changelog_dir); \
+ } while (0)
/**
* everything after 'CHANGELOG_TYPE_METADATA_XATTR' are internal types
* (ie. none of the fops trigger this type of event), hence
* CHANGELOG_MAX_TYPE = 4
*/
typedef enum {
- CHANGELOG_TYPE_DATA = 0,
- CHANGELOG_TYPE_METADATA,
- CHANGELOG_TYPE_ENTRY,
- CHANGELOG_TYPE_METADATA_XATTR,
- CHANGELOG_TYPE_ROLLOVER,
- CHANGELOG_TYPE_FSYNC,
+ CHANGELOG_TYPE_DATA = 0,
+ CHANGELOG_TYPE_METADATA,
+ CHANGELOG_TYPE_ENTRY,
+ CHANGELOG_TYPE_METADATA_XATTR,
+ CHANGELOG_TYPE_ROLLOVER,
+ CHANGELOG_TYPE_FSYNC,
} changelog_log_type;
/* operation modes - RT for now */
typedef enum {
- CHANGELOG_MODE_RT = 0,
+ CHANGELOG_MODE_RT = 0,
} changelog_mode_t;
/* encoder types */
typedef enum {
- CHANGELOG_ENCODE_MIN = 0,
- CHANGELOG_ENCODE_BINARY,
- CHANGELOG_ENCODE_ASCII,
- CHANGELOG_ENCODE_MAX,
+ CHANGELOG_ENCODE_MIN = 0,
+ CHANGELOG_ENCODE_BINARY,
+ CHANGELOG_ENCODE_ASCII,
+ CHANGELOG_ENCODE_MAX,
} changelog_encoder_t;
-#define CHANGELOG_VALID_ENCODING(enc) \
- (enc > CHANGELOG_ENCODE_MIN && enc < CHANGELOG_ENCODE_MAX)
+#define CHANGELOG_VALID_ENCODING(enc) \
+ (enc > CHANGELOG_ENCODE_MIN && enc < CHANGELOG_ENCODE_MAX)
-#define CHANGELOG_TYPE_IS_ENTRY(type) (type == CHANGELOG_TYPE_ENTRY)
-#define CHANGELOG_TYPE_IS_ROLLOVER(type) (type == CHANGELOG_TYPE_ROLLOVER)
-#define CHANGELOG_TYPE_IS_FSYNC(type) (type == CHANGELOG_TYPE_FSYNC)
+#define CHANGELOG_TYPE_IS_ENTRY(type) (type == CHANGELOG_TYPE_ENTRY)
+#define CHANGELOG_TYPE_IS_ROLLOVER(type) (type == CHANGELOG_TYPE_ROLLOVER)
+#define CHANGELOG_TYPE_IS_FSYNC(type) (type == CHANGELOG_TYPE_FSYNC)
#endif /* _CHANGELOG_MISC_H */
diff --git a/xlators/features/changelog/src/changelog-rpc-common.c b/xlators/features/changelog/src/changelog-rpc-common.c
index 7bd5b688fa5..125246a17e1 100644
--- a/xlators/features/changelog/src/changelog-rpc-common.c
+++ b/xlators/features/changelog/src/changelog-rpc-common.c
@@ -11,7 +11,7 @@
#include "changelog-rpc-common.h"
#include "changelog-messages.h"
-#include "syscall.h"
+#include <glusterfs/syscall.h>
/**
*****************************************************
Client Interface
@@ -24,65 +24,63 @@
*/
void *
-changelog_rpc_poller (void *arg)
+changelog_rpc_poller(void *arg)
{
- xlator_t *this = arg;
+ xlator_t *this = arg;
- (void) event_dispatch (this->ctx->event_pool);
- return NULL;
+ (void)gf_event_dispatch(this->ctx->event_pool);
+ return NULL;
}
struct rpc_clnt *
-changelog_rpc_client_init (xlator_t *this, void *cbkdata,
- char *sockfile, rpc_clnt_notify_t fn)
+changelog_rpc_client_init(xlator_t *this, void *cbkdata, char *sockfile,
+ rpc_clnt_notify_t fn)
{
- int ret = 0;
- struct rpc_clnt *rpc = NULL;
- dict_t *options = NULL;
-
- if (!cbkdata)
- cbkdata = this;
-
- options = dict_new ();
- if (!options)
- goto error_return;
-
- ret = rpc_transport_unix_options_build (&options, sockfile, 0);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_RPC_BUILD_ERROR,
- "failed to build rpc options");
- goto dealloc_dict;
- }
-
- rpc = rpc_clnt_new (options, this, this->name, 16);
- if (!rpc)
- goto dealloc_dict;
-
- ret = rpc_clnt_register_notify (rpc, fn, cbkdata);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_NOTIFY_REGISTER_FAILED,
- "failed to register notify");
- goto dealloc_rpc_clnt;
- }
-
- ret = rpc_clnt_start (rpc);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_RPC_START_ERROR,
- "failed to start rpc");
- goto dealloc_rpc_clnt;
- }
-
- return rpc;
-
- dealloc_rpc_clnt:
- rpc_clnt_unref (rpc);
- dealloc_dict:
- dict_unref (options);
- error_return:
- return NULL;
+ int ret = 0;
+ struct rpc_clnt *rpc = NULL;
+ dict_t *options = NULL;
+
+ if (!cbkdata)
+ cbkdata = this;
+
+ options = dict_new();
+ if (!options)
+ goto error_return;
+
+ ret = rpc_transport_unix_options_build(options, sockfile, 0);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_RPC_BUILD_ERROR,
+ NULL);
+ goto dealloc_dict;
+ }
+
+ rpc = rpc_clnt_new(options, this, this->name, 16);
+ if (!rpc)
+ goto dealloc_dict;
+
+ ret = rpc_clnt_register_notify(rpc, fn, cbkdata);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_MSG_NOTIFY_REGISTER_FAILED, NULL);
+ goto dealloc_rpc_clnt;
+ }
+
+ ret = rpc_clnt_start(rpc);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_RPC_START_ERROR,
+ NULL);
+ goto dealloc_rpc_clnt;
+ }
+
+ dict_unref(options);
+ return rpc;
+
+dealloc_rpc_clnt:
+ rpc_clnt_unref(rpc);
+dealloc_dict:
+ dict_unref(options);
+error_return:
+ return NULL;
}
/**
@@ -90,96 +88,96 @@ changelog_rpc_client_init (xlator_t *this, void *cbkdata,
* RPC server.
*/
int
-changelog_rpc_sumbit_req (struct rpc_clnt *rpc, void *req,
- call_frame_t *frame, rpc_clnt_prog_t *prog,
- int procnum, struct iovec *payload, int payloadcnt,
- struct iobref *iobref, xlator_t *this,
- fop_cbk_fn_t cbkfn, xdrproc_t xdrproc)
+changelog_rpc_sumbit_req(struct rpc_clnt *rpc, void *req, call_frame_t *frame,
+ rpc_clnt_prog_t *prog, int procnum,
+ struct iovec *payload, int payloadcnt,
+ struct iobref *iobref, xlator_t *this,
+ fop_cbk_fn_t cbkfn, xdrproc_t xdrproc)
{
- int ret = 0;
- int count = 0;
- struct iovec iov = {0, };
- struct iobuf *iobuf = NULL;
- char new_iobref = 0;
- ssize_t xdr_size = 0;
-
- GF_ASSERT (this);
+ int ret = 0;
+ int count = 0;
+ struct iovec iov = {
+ 0,
+ };
+ struct iobuf *iobuf = NULL;
+ char new_iobref = 0;
+ ssize_t xdr_size = 0;
- if (req) {
- xdr_size = xdr_sizeof (xdrproc, req);
+ GF_ASSERT(this);
- iobuf = iobuf_get2 (this->ctx->iobuf_pool, xdr_size);
- if (!iobuf) {
- goto out;
- };
+ if (req) {
+ xdr_size = xdr_sizeof(xdrproc, req);
- if (!iobref) {
- iobref = iobref_new ();
- if (!iobref) {
- goto out;
- }
+ iobuf = iobuf_get2(this->ctx->iobuf_pool, xdr_size);
+ if (!iobuf) {
+ goto out;
+ };
- new_iobref = 1;
- }
+ if (!iobref) {
+ iobref = iobref_new();
+ if (!iobref) {
+ goto out;
+ }
- iobref_add (iobref, iobuf);
+ new_iobref = 1;
+ }
- iov.iov_base = iobuf->ptr;
- iov.iov_len = iobuf_size (iobuf);
+ iobref_add(iobref, iobuf);
- /* Create the xdr payload */
- ret = xdr_serialize_generic (iov, req, xdrproc);
- if (ret == -1) {
- goto out;
- }
+ iov.iov_base = iobuf->ptr;
+ iov.iov_len = iobuf_size(iobuf);
- iov.iov_len = ret;
- count = 1;
+ /* Create the xdr payload */
+ ret = xdr_serialize_generic(iov, req, xdrproc);
+ if (ret == -1) {
+ goto out;
}
- ret = rpc_clnt_submit (rpc, prog, procnum, cbkfn, &iov, count,
- payload, payloadcnt, iobref, frame, NULL,
- 0, NULL, 0, NULL);
+ iov.iov_len = ret;
+ count = 1;
+ }
+
+ ret = rpc_clnt_submit(rpc, prog, procnum, cbkfn, &iov, count, payload,
+ payloadcnt, iobref, frame, NULL, 0, NULL, 0, NULL);
- out:
- if (new_iobref)
- iobref_unref (iobref);
- if (iobuf)
- iobuf_unref (iobuf);
- return ret;
+out:
+ if (new_iobref)
+ iobref_unref(iobref);
+ if (iobuf)
+ iobuf_unref(iobuf);
+ return ret;
}
/**
* Entry point to perform a remote procedure call
*/
int
-changelog_invoke_rpc (xlator_t *this, struct rpc_clnt *rpc,
- rpc_clnt_prog_t *prog, int procidx, void *arg)
+changelog_invoke_rpc(xlator_t *this, struct rpc_clnt *rpc,
+ rpc_clnt_prog_t *prog, int procidx, void *arg)
{
- int ret = 0;
- call_frame_t *frame = NULL;
- rpc_clnt_procedure_t *proc = NULL;
-
- if (!this || !prog)
- goto error_return;
-
- frame = create_frame (this, this->ctx->pool);
- if (!frame) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_CREATE_FRAME_FAILED,
- "failed to create frame");
- goto error_return;
- }
+ int ret = 0;
+ call_frame_t *frame = NULL;
+ rpc_clnt_procedure_t *proc = NULL;
+
+ if (!this || !prog)
+ goto error_return;
+
+ frame = create_frame(this, this->ctx->pool);
+ if (!frame) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_CREATE_FRAME_FAILED,
+ NULL);
+ goto error_return;
+ }
- proc = &prog->proctable[procidx];
- if (proc->fn)
- ret = proc->fn (frame, this, arg);
+ proc = &prog->proctable[procidx];
+ if (proc->fn)
+ ret = proc->fn(frame, this, arg);
- STACK_DESTROY (frame->root);
- return ret;
+ STACK_DESTROY(frame->root);
+ return ret;
- error_return:
- return -1;
+error_return:
+ return -1;
}
/**
@@ -189,161 +187,173 @@ changelog_invoke_rpc (xlator_t *this, struct rpc_clnt *rpc,
*/
struct iobuf *
-__changelog_rpc_serialize_reply (rpcsvc_request_t *req, void *arg,
- struct iovec *outmsg, xdrproc_t xdrproc)
+__changelog_rpc_serialize_reply(rpcsvc_request_t *req, void *arg,
+ struct iovec *outmsg, xdrproc_t xdrproc)
{
- struct iobuf *iob = NULL;
- ssize_t retlen = 0;
- ssize_t rsp_size = 0;
+ struct iobuf *iob = NULL;
+ ssize_t retlen = 0;
+ ssize_t rsp_size = 0;
- rsp_size = xdr_sizeof (xdrproc, arg);
- iob = iobuf_get2 (req->svc->ctx->iobuf_pool, rsp_size);
- if (!iob)
- goto error_return;
+ rsp_size = xdr_sizeof(xdrproc, arg);
+ iob = iobuf_get2(req->svc->ctx->iobuf_pool, rsp_size);
+ if (!iob)
+ goto error_return;
- iobuf_to_iovec (iob, outmsg);
+ iobuf_to_iovec(iob, outmsg);
- retlen = xdr_serialize_generic (*outmsg, arg, xdrproc);
- if (retlen == -1)
- goto unref_iob;
+ retlen = xdr_serialize_generic(*outmsg, arg, xdrproc);
+ if (retlen == -1)
+ goto unref_iob;
- outmsg->iov_len = retlen;
- return iob;
+ outmsg->iov_len = retlen;
+ return iob;
- unref_iob:
- iobuf_unref (iob);
- error_return:
- return NULL;
+unref_iob:
+ iobuf_unref(iob);
+error_return:
+ return NULL;
}
int
-changelog_rpc_sumbit_reply (rpcsvc_request_t *req,
- void *arg, struct iovec *payload, int payloadcount,
- struct iobref *iobref, xdrproc_t xdrproc)
+changelog_rpc_sumbit_reply(rpcsvc_request_t *req, void *arg,
+ struct iovec *payload, int payloadcount,
+ struct iobref *iobref, xdrproc_t xdrproc)
{
- int ret = -1;
- struct iobuf *iob = NULL;
- struct iovec iov = {0,};
- char new_iobref = 0;
-
- if (!req)
- goto return_ret;
-
- if (!iobref) {
- iobref = iobref_new ();
- if (!iobref)
- goto return_ret;
- new_iobref = 1;
- }
-
- iob = __changelog_rpc_serialize_reply (req, arg, &iov, xdrproc);
- if (!iob)
- gf_msg ("", GF_LOG_ERROR, 0,
- CHANGELOG_MSG_RPC_SUBMIT_REPLY_FAILED,
- "failed to serialize reply");
- else
- iobref_add (iobref, iob);
-
- ret = rpcsvc_submit_generic (req, &iov,
- 1, payload, payloadcount, iobref);
-
- if (new_iobref)
- iobref_unref (iobref);
- if (iob)
- iobuf_unref (iob);
- return_ret:
- return ret;
+ int ret = -1;
+ struct iobuf *iob = NULL;
+ struct iovec iov = {
+ 0,
+ };
+ char new_iobref = 0;
+
+ if (!req)
+ goto return_ret;
+
+ if (!iobref) {
+ iobref = iobref_new();
+ if (!iobref)
+ goto return_ret;
+ new_iobref = 1;
+ }
+
+ iob = __changelog_rpc_serialize_reply(req, arg, &iov, xdrproc);
+ if (!iob)
+ gf_smsg("", GF_LOG_ERROR, 0, CHANGELOG_MSG_RPC_SUBMIT_REPLY_FAILED,
+ NULL);
+ else
+ iobref_add(iobref, iob);
+
+ ret = rpcsvc_submit_generic(req, &iov, 1, payload, payloadcount, iobref);
+
+ if (new_iobref)
+ iobref_unref(iobref);
+ if (iob)
+ iobuf_unref(iob);
+return_ret:
+ return ret;
}
void
-changelog_rpc_server_destroy (xlator_t *this, rpcsvc_t *rpc, char *sockfile,
- rpcsvc_notify_t fn, struct rpcsvc_program **progs)
+changelog_rpc_server_destroy(xlator_t *this, rpcsvc_t *rpc, char *sockfile,
+ rpcsvc_notify_t fn, struct rpcsvc_program **progs)
{
- rpcsvc_listener_t *listener = NULL;
- rpcsvc_listener_t *next = NULL;
- struct rpcsvc_program *prog = NULL;
-
- while (*progs) {
- prog = *progs;
- (void) rpcsvc_program_unregister (rpc, prog);
- progs++;
+ rpcsvc_listener_t *listener = NULL;
+ rpcsvc_listener_t *next = NULL;
+ struct rpcsvc_program *prog = NULL;
+ rpc_transport_t *trans = NULL;
+
+ if (!rpc)
+ return;
+
+ while (*progs) {
+ prog = *progs;
+ (void)rpcsvc_program_unregister(rpc, prog);
+ progs++;
+ }
+
+ list_for_each_entry_safe(listener, next, &rpc->listeners, list)
+ {
+ if (listener->trans) {
+ trans = listener->trans;
+ rpc_transport_disconnect(trans, _gf_false);
}
-
- list_for_each_entry_safe (listener, next, &rpc->listeners, list) {
- rpcsvc_listener_destroy (listener);
+ }
+
+ (void)rpcsvc_unregister_notify(rpc, fn, this);
+
+ /* TODO Avoid freeing rpc object in case of brick multiplex
+ after freeing rpc object svc->rpclock corrupted and it takes
+ more time to detach a brick
+ */
+ if (!this->cleanup_starting) {
+ if (rpc->rxpool) {
+ mem_pool_destroy(rpc->rxpool);
+ rpc->rxpool = NULL;
}
-
- (void) rpcsvc_unregister_notify (rpc, fn, this);
- sys_unlink (sockfile);
-
- GF_FREE (rpc);
+ GF_FREE(rpc);
+ }
}
rpcsvc_t *
-changelog_rpc_server_init (xlator_t *this, char *sockfile, void *cbkdata,
- rpcsvc_notify_t fn, struct rpcsvc_program **progs)
+changelog_rpc_server_init(xlator_t *this, char *sockfile, void *cbkdata,
+ rpcsvc_notify_t fn, struct rpcsvc_program **progs)
{
- int ret = 0;
- rpcsvc_t *rpc = NULL;
- dict_t *options = NULL;
- struct rpcsvc_program *prog = NULL;
-
- if (!cbkdata)
- cbkdata = this;
-
- options = dict_new ();
- if (!options)
- goto error_return;
-
- ret = rpcsvc_transport_unix_options_build (&options, sockfile);
- if (ret)
- goto dealloc_dict;
-
- rpc = rpcsvc_init (this, this->ctx, options, 8);
- if (rpc == NULL) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_RPC_START_ERROR,
- "failed to init rpc");
- goto dealloc_dict;
- }
+ int ret = 0;
+ rpcsvc_t *rpc = NULL;
+ dict_t *options = NULL;
+ struct rpcsvc_program *prog = NULL;
- ret = rpcsvc_register_notify (rpc, fn, cbkdata);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_NOTIFY_REGISTER_FAILED,
- "failed to register notify function");
- goto dealloc_rpc;
- }
+ if (!cbkdata)
+ cbkdata = this;
- ret = rpcsvc_create_listeners (rpc, options, this->name);
- if (ret != 1) {
- gf_msg_debug (this->name,
- 0, "failed to create listeners");
- goto dealloc_rpc;
- }
+ options = dict_new();
+ if (!options)
+ return NULL;
- while (*progs) {
- prog = *progs;
- ret = rpcsvc_program_register (rpc, prog, _gf_false);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_PROGRAM_NAME_REG_FAILED,
- "cannot register program "
- "(name: %s, prognum: %d, pogver: %d)",
- prog->progname, prog->prognum, prog->progver);
- goto dealloc_rpc;
- }
-
- progs++;
+ ret = rpcsvc_transport_unix_options_build(options, sockfile);
+ if (ret)
+ goto dealloc_dict;
+
+ rpc = rpcsvc_init(this, this->ctx, options, 8);
+ if (rpc == NULL) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_RPC_START_ERROR,
+ NULL);
+ goto dealloc_dict;
+ }
+
+ ret = rpcsvc_register_notify(rpc, fn, cbkdata);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_MSG_NOTIFY_REGISTER_FAILED, NULL);
+ goto dealloc_rpc;
+ }
+
+ ret = rpcsvc_create_listeners(rpc, options, this->name);
+ if (ret != 1) {
+ gf_msg_debug(this->name, 0, "failed to create listeners");
+ goto dealloc_rpc;
+ }
+
+ while (*progs) {
+ prog = *progs;
+ ret = rpcsvc_program_register(rpc, prog, _gf_false);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_MSG_PROGRAM_NAME_REG_FAILED, "name%s",
+ prog->progname, "prognum=%d", prog->prognum, "pogver=%d",
+ prog->progver, NULL);
+ goto dealloc_rpc;
}
- dict_unref (options);
- return rpc;
+ progs++;
+ }
- dealloc_rpc:
- GF_FREE (rpc);
- dealloc_dict:
- dict_unref (options);
- error_return:
- return NULL;
+ dict_unref(options);
+ return rpc;
+
+dealloc_rpc:
+ GF_FREE(rpc);
+dealloc_dict:
+ dict_unref(options);
+ return NULL;
}
diff --git a/xlators/features/changelog/src/changelog-rpc-common.h b/xlators/features/changelog/src/changelog-rpc-common.h
index 95c850c9400..4d9aa2c694b 100644
--- a/xlators/features/changelog/src/changelog-rpc-common.h
+++ b/xlators/features/changelog/src/changelog-rpc-common.h
@@ -13,8 +13,8 @@
#include "rpcsvc.h"
#include "rpc-clnt.h"
-#include "event.h"
-#include "call-stub.h"
+#include <glusterfs/gf-event.h>
+#include <glusterfs/call-stub.h>
#include "changelog-xdr.h"
#include "xdr-generic.h"
@@ -24,61 +24,62 @@
/**
* Let's keep this non-configurable for now.
*/
-#define NR_ROTT_BUFFS 4
+#define NR_ROTT_BUFFS 4
#define NR_DISPATCHERS (NR_ROTT_BUFFS - 1)
enum changelog_rpc_procnum {
- CHANGELOG_RPC_PROC_NULL = 0,
- CHANGELOG_RPC_PROBE_FILTER = 1,
- CHANGELOG_RPC_PROC_MAX = 2,
+ CHANGELOG_RPC_PROC_NULL = 0,
+ CHANGELOG_RPC_PROBE_FILTER = 1,
+ CHANGELOG_RPC_PROC_MAX = 2,
};
-#define CHANGELOG_RPC_PROGNUM 1885957735
-#define CHANGELOG_RPC_PROGVER 1
+#define CHANGELOG_RPC_PROGNUM 1885957735
+#define CHANGELOG_RPC_PROGVER 1
/**
* reverse connection: data xfer path
*/
enum changelog_reverse_rpc_procnum {
- CHANGELOG_REV_PROC_NULL = 0,
- CHANGELOG_REV_PROC_EVENT = 1,
- CHANGELOG_REV_PROC_MAX = 2,
+ CHANGELOG_REV_PROC_NULL = 0,
+ CHANGELOG_REV_PROC_EVENT = 1,
+ CHANGELOG_REV_PROC_MAX = 2,
};
-#define CHANGELOG_REV_RPC_PROCNUM 1886350951
-#define CHANGELOG_REV_RPC_PROCVER 1
+#define CHANGELOG_REV_RPC_PROCNUM 1886350951
+#define CHANGELOG_REV_RPC_PROCVER 1
typedef struct changelog_rpc {
- rpcsvc_t *svc;
- struct rpc_clnt *rpc;
- char sock[UNIX_PATH_MAX]; /* tied to server */
+ rpcsvc_t *svc;
+ struct rpc_clnt *rpc;
+ char sock[UNIX_PATH_MAX]; /* tied to server */
} changelog_rpc_t;
/* event poller */
-void *changelog_rpc_poller (void *);
+void *
+changelog_rpc_poller(void *);
/* CLIENT API */
struct rpc_clnt *
-changelog_rpc_client_init (xlator_t *, void *, char *, rpc_clnt_notify_t);
+changelog_rpc_client_init(xlator_t *, void *, char *, rpc_clnt_notify_t);
int
-changelog_rpc_sumbit_req (struct rpc_clnt *, void *, call_frame_t *,
- rpc_clnt_prog_t *, int , struct iovec *, int,
- struct iobref *, xlator_t *, fop_cbk_fn_t, xdrproc_t);
+changelog_rpc_sumbit_req(struct rpc_clnt *, void *, call_frame_t *,
+ rpc_clnt_prog_t *, int, struct iovec *, int,
+ struct iobref *, xlator_t *, fop_cbk_fn_t, xdrproc_t);
int
-changelog_invoke_rpc (xlator_t *, struct rpc_clnt *,
- rpc_clnt_prog_t *, int , void *);
+changelog_invoke_rpc(xlator_t *, struct rpc_clnt *, rpc_clnt_prog_t *, int,
+ void *);
/* SERVER API */
int
-changelog_rpc_sumbit_reply (rpcsvc_request_t *, void *,
- struct iovec *, int, struct iobref *, xdrproc_t);
+changelog_rpc_sumbit_reply(rpcsvc_request_t *, void *, struct iovec *, int,
+ struct iobref *, xdrproc_t);
rpcsvc_t *
-changelog_rpc_server_init (xlator_t *, char *, void*,
- rpcsvc_notify_t, struct rpcsvc_program **);
+changelog_rpc_server_init(xlator_t *, char *, void *, rpcsvc_notify_t,
+ struct rpcsvc_program **);
void
-changelog_rpc_server_destroy (xlator_t *, rpcsvc_t *, char *,
- rpcsvc_notify_t, struct rpcsvc_program **);
+changelog_rpc_server_destroy(xlator_t *, rpcsvc_t *, char *, rpcsvc_notify_t,
+ struct rpcsvc_program **);
#endif
diff --git a/xlators/features/changelog/src/changelog-rpc.c b/xlators/features/changelog/src/changelog-rpc.c
index 4b2b38cad51..440b88091a6 100644
--- a/xlators/features/changelog/src/changelog-rpc.c
+++ b/xlators/features/changelog/src/changelog-rpc.c
@@ -8,231 +8,346 @@
cases as published by the Free Software Foundation.
*/
-#include "syscall.h"
+#include <glusterfs/syscall.h>
#include "changelog-rpc.h"
#include "changelog-mem-types.h"
#include "changelog-ev-handle.h"
-struct rpcsvc_program *changelog_programs[];
+static struct rpcsvc_program *changelog_programs[];
static void
-changelog_cleanup_dispatchers (xlator_t *this,
- changelog_priv_t *priv, int count)
+changelog_cleanup_dispatchers(xlator_t *this, changelog_priv_t *priv, int count)
{
- for (count--; count >= 0; count--) {
- (void) changelog_thread_cleanup
- (this, priv->ev_dispatcher[count]);
- priv->ev_dispatcher[count] = 0;
- }
+ for (count--; count >= 0; count--) {
+ (void)changelog_thread_cleanup(this, priv->ev_dispatcher[count]);
+ priv->ev_dispatcher[count] = 0;
+ }
}
int
-changelog_cleanup_rpc_threads (xlator_t *this, changelog_priv_t *priv)
+changelog_cleanup_rpc_threads(xlator_t *this, changelog_priv_t *priv)
{
- int ret = 0;
- changelog_clnt_t *conn = NULL;
-
- conn = &priv->connections;
- if (!conn)
- return 0;
-
- /** terminate RPC thread(s) */
- ret = changelog_thread_cleanup (this, priv->connector);
- if (ret != 0)
- goto error_return;
- priv->connector = 0;
-
- /** terminate dispatcher thread(s) */
- changelog_cleanup_dispatchers (this, priv, priv->nr_dispatchers);
-
- /* TODO: what about pending and waiting connections? */
- changelog_ev_cleanup_connections (this, conn);
-
- /* destroy locks */
- ret = pthread_mutex_destroy (&conn->pending_lock);
- if (ret != 0)
- goto error_return;
- ret = pthread_cond_destroy (&conn->pending_cond);
- if (ret != 0)
- goto error_return;
- ret = LOCK_DESTROY (&conn->active_lock);
- if (ret != 0)
- goto error_return;
- ret = LOCK_DESTROY (&conn->wait_lock);
- if (ret != 0)
- goto error_return;
+ int ret = 0;
+ changelog_clnt_t *conn = NULL;
+
+ conn = &priv->connections;
+ if (!conn)
return 0;
- error_return:
- return -1;
+ /** terminate RPC thread(s) */
+ ret = changelog_thread_cleanup(this, priv->connector);
+ if (ret != 0)
+ goto error_return;
+ priv->connector = 0;
+
+ /** terminate dispatcher thread(s) */
+ changelog_cleanup_dispatchers(this, priv, priv->nr_dispatchers);
+
+ /* destroy locks */
+ ret = pthread_mutex_destroy(&conn->pending_lock);
+ if (ret != 0)
+ goto error_return;
+ ret = pthread_cond_destroy(&conn->pending_cond);
+ if (ret != 0)
+ goto error_return;
+ ret = LOCK_DESTROY(&conn->active_lock);
+ if (ret != 0)
+ goto error_return;
+ ret = LOCK_DESTROY(&conn->wait_lock);
+ if (ret != 0)
+ goto error_return;
+ return 0;
+
+error_return:
+ return -1;
}
static int
-changelog_init_rpc_threads (xlator_t *this, changelog_priv_t *priv,
- rbuf_t *rbuf, int nr_dispatchers)
+changelog_init_rpc_threads(xlator_t *this, changelog_priv_t *priv, rbuf_t *rbuf,
+ int nr_dispatchers)
+{
+ int j = 0;
+ int ret = 0;
+ changelog_clnt_t *conn = NULL;
+
+ conn = &priv->connections;
+
+ conn->this = this;
+ conn->rbuf = rbuf;
+ conn->sequence = 1; /* start with sequence number one */
+
+ INIT_LIST_HEAD(&conn->pending);
+ INIT_LIST_HEAD(&conn->active);
+ INIT_LIST_HEAD(&conn->waitq);
+
+ ret = pthread_mutex_init(&conn->pending_lock, NULL);
+ if (ret)
+ goto error_return;
+ ret = pthread_cond_init(&conn->pending_cond, NULL);
+ if (ret)
+ goto cleanup_pending_lock;
+
+ ret = LOCK_INIT(&conn->active_lock);
+ if (ret)
+ goto cleanup_pending_cond;
+ ret = LOCK_INIT(&conn->wait_lock);
+ if (ret)
+ goto cleanup_active_lock;
+
+ /* spawn reverse connection thread */
+ ret = gf_thread_create(&priv->connector, NULL, changelog_ev_connector, conn,
+ "clogecon");
+ if (ret != 0)
+ goto cleanup_wait_lock;
+
+ /* spawn dispatcher thread(s) */
+ priv->ev_dispatcher = GF_CALLOC(nr_dispatchers, sizeof(pthread_t),
+ gf_changelog_mt_ev_dispatcher_t);
+ if (!priv->ev_dispatcher)
+ goto cleanup_connector;
+
+ /* spawn dispatcher threads */
+ for (; j < nr_dispatchers; j++) {
+ ret = gf_thread_create(&priv->ev_dispatcher[j], NULL,
+ changelog_ev_dispatch, conn, "clogd%03hx",
+ j & 0x3ff);
+ if (ret != 0) {
+ changelog_cleanup_dispatchers(this, priv, j);
+ break;
+ }
+ }
+
+ if (ret != 0)
+ goto cleanup_connector;
+
+ priv->nr_dispatchers = nr_dispatchers;
+ return 0;
+
+cleanup_connector:
+ (void)pthread_cancel(priv->connector);
+cleanup_wait_lock:
+ LOCK_DESTROY(&conn->wait_lock);
+cleanup_active_lock:
+ LOCK_DESTROY(&conn->active_lock);
+cleanup_pending_cond:
+ (void)pthread_cond_destroy(&conn->pending_cond);
+cleanup_pending_lock:
+ (void)pthread_mutex_destroy(&conn->pending_lock);
+error_return:
+ return -1;
+}
+
+int
+changelog_rpcsvc_notify(rpcsvc_t *rpc, void *xl, rpcsvc_event_t event,
+ void *data)
{
- int j = 0;
- int ret = 0;
- changelog_clnt_t *conn = NULL;
- char thread_name[GF_THREAD_NAMEMAX] = {0,};
-
-
- conn = &priv->connections;
-
- conn->this = this;
- conn->rbuf = rbuf;
- conn->sequence = 1; /* start with sequence number one */
-
- INIT_LIST_HEAD (&conn->pending);
- INIT_LIST_HEAD (&conn->active);
- INIT_LIST_HEAD (&conn->waitq);
-
- ret = pthread_mutex_init (&conn->pending_lock, NULL);
- if (ret)
- goto error_return;
- ret = pthread_cond_init (&conn->pending_cond, NULL);
- if (ret)
- goto cleanup_pending_lock;
-
- ret = LOCK_INIT (&conn->active_lock);
- if (ret)
- goto cleanup_pending_cond;
- ret = LOCK_INIT (&conn->wait_lock);
- if (ret)
- goto cleanup_active_lock;
-
- /* spawn reverse connection thread */
- ret = gf_thread_create (&priv->connector,
- NULL, changelog_ev_connector, conn, "clogecon");
- if (ret != 0)
- goto cleanup_wait_lock;
-
- /* spawn dispatcher thread(s) */
- priv->ev_dispatcher = GF_CALLOC (nr_dispatchers, sizeof(pthread_t),
- gf_changelog_mt_ev_dispatcher_t);
- if (!priv->ev_dispatcher)
- goto cleanup_connector;
-
- /* spawn dispatcher threads */
- for (; j < nr_dispatchers; j++) {
- snprintf (thread_name, sizeof(thread_name),
- "%s%d", "clogd", j);
- ret = gf_thread_create (&priv->ev_dispatcher[j],
- NULL, changelog_ev_dispatch, conn,
- thread_name);
- if (ret != 0) {
- changelog_cleanup_dispatchers (this, priv, j);
- break;
+ xlator_t *this = NULL;
+ rpc_transport_t *trans = NULL;
+ rpc_transport_t *xprt = NULL;
+ rpc_transport_t *xp_next = NULL;
+ changelog_priv_t *priv = NULL;
+ uint64_t listnercnt = 0;
+ uint64_t xprtcnt = 0;
+ uint64_t clntcnt = 0;
+ rpcsvc_listener_t *listener = NULL;
+ rpcsvc_listener_t *next = NULL;
+ gf_boolean_t listner_found = _gf_false;
+ socket_private_t *sockpriv = NULL;
+
+ if (!xl || !data || !rpc) {
+ gf_msg_callingfn("changelog", GF_LOG_WARNING, 0,
+ CHANGELOG_MSG_RPCSVC_NOTIFY_FAILED,
+ "Calling rpc_notify without initializing");
+ goto out;
+ }
+
+ this = xl;
+ trans = data;
+ priv = this->private;
+
+ if (!priv) {
+ gf_msg_callingfn("changelog", GF_LOG_WARNING, 0,
+ CHANGELOG_MSG_RPCSVC_NOTIFY_FAILED,
+ "Calling rpc_notify without priv initializing");
+ goto out;
+ }
+
+ if (event == RPCSVC_EVENT_ACCEPT) {
+ GF_ATOMIC_INC(priv->xprtcnt);
+ LOCK(&priv->lock);
+ {
+ list_add_tail(&trans->list, &priv->xprt_list);
+ }
+ UNLOCK(&priv->lock);
+ goto out;
+ }
+
+ if (event == RPCSVC_EVENT_DISCONNECT) {
+ list_for_each_entry_safe(listener, next, &rpc->listeners, list)
+ {
+ if (listener && listener->trans) {
+ if (listener->trans == trans) {
+ listnercnt = GF_ATOMIC_DEC(priv->listnercnt);
+ listner_found = _gf_true;
+ rpcsvc_listener_destroy(listener);
}
+ }
}
- if (ret != 0)
- goto cleanup_connector;
+ if (listnercnt > 0) {
+ goto out;
+ }
+ if (listner_found) {
+ LOCK(&priv->lock);
+ list_for_each_entry_safe(xprt, xp_next, &priv->xprt_list, list)
+ {
+ sockpriv = (socket_private_t *)(xprt->private);
+ gf_log("changelog", GF_LOG_INFO,
+ "Send disconnect"
+ " on socket %d",
+ sockpriv->sock);
+ rpc_transport_disconnect(xprt, _gf_false);
+ }
+ UNLOCK(&priv->lock);
+ goto out;
+ }
+ LOCK(&priv->lock);
+ {
+ list_del_init(&trans->list);
+ }
+ UNLOCK(&priv->lock);
- priv->nr_dispatchers = nr_dispatchers;
- return 0;
+ xprtcnt = GF_ATOMIC_DEC(priv->xprtcnt);
+ clntcnt = GF_ATOMIC_GET(priv->clntcnt);
+ if (!xprtcnt && !clntcnt) {
+ changelog_process_cleanup_event(this);
+ }
+ }
- cleanup_connector:
- (void) pthread_cancel (priv->connector);
- cleanup_wait_lock:
- LOCK_DESTROY (&conn->wait_lock);
- cleanup_active_lock:
- LOCK_DESTROY (&conn->active_lock);
- cleanup_pending_cond:
- (void) pthread_cond_destroy (&conn->pending_cond);
- cleanup_pending_lock:
- (void) pthread_mutex_destroy (&conn->pending_lock);
- error_return:
- return -1;
+out:
+ return 0;
}
-int
-changelog_rpcsvc_notify (rpcsvc_t *rpc,
- void *xl, rpcsvc_event_t event, void *data)
+void
+changelog_process_cleanup_event(xlator_t *this)
{
- return 0;
+ gf_boolean_t cleanup_notify = _gf_false;
+ changelog_priv_t *priv = NULL;
+ char sockfile[UNIX_PATH_MAX] = {
+ 0,
+ };
+
+ if (!this)
+ return;
+ priv = this->private;
+ if (!priv)
+ return;
+
+ LOCK(&priv->lock);
+ {
+ cleanup_notify = priv->notify_down;
+ priv->notify_down = _gf_true;
+ }
+ UNLOCK(&priv->lock);
+
+ if (priv->victim && !cleanup_notify) {
+ default_notify(this, GF_EVENT_PARENT_DOWN, priv->victim);
+
+ if (priv->rpc) {
+ /* sockfile path could have been saved to avoid this */
+ CHANGELOG_MAKE_SOCKET_PATH(priv->changelog_brick, sockfile,
+ UNIX_PATH_MAX);
+ sys_unlink(sockfile);
+ (void)rpcsvc_unregister_notify(priv->rpc, changelog_rpcsvc_notify,
+ this);
+ if (priv->rpc->rxpool) {
+ mem_pool_destroy(priv->rpc->rxpool);
+ priv->rpc->rxpool = NULL;
+ }
+ GF_FREE(priv->rpc);
+ priv->rpc = NULL;
+ }
+ }
}
void
-changelog_destroy_rpc_listner (xlator_t *this, changelog_priv_t *priv)
+changelog_destroy_rpc_listner(xlator_t *this, changelog_priv_t *priv)
{
- char sockfile[UNIX_PATH_MAX] = {0,};
-
- /* sockfile path could have been saved to avoid this */
- CHANGELOG_MAKE_SOCKET_PATH (priv->changelog_brick,
- sockfile, UNIX_PATH_MAX);
- changelog_rpc_server_destroy (this,
- priv->rpc, sockfile,
- changelog_rpcsvc_notify,
- changelog_programs);
- (void) changelog_cleanup_rpc_threads (this, priv);
+ char sockfile[UNIX_PATH_MAX] = {
+ 0,
+ };
+
+ /* sockfile path could have been saved to avoid this */
+ CHANGELOG_MAKE_SOCKET_PATH(priv->changelog_brick, sockfile, UNIX_PATH_MAX);
+ changelog_rpc_server_destroy(this, priv->rpc, sockfile,
+ changelog_rpcsvc_notify, changelog_programs);
}
rpcsvc_t *
-changelog_init_rpc_listener (xlator_t *this, changelog_priv_t *priv,
+changelog_init_rpc_listener(xlator_t *this, changelog_priv_t *priv,
rbuf_t *rbuf, int nr_dispatchers)
{
- int ret = 0;
- char sockfile[UNIX_PATH_MAX] = {0,};
- rpcsvc_t *svcp;
-
- ret = changelog_init_rpc_threads (this, priv, rbuf, nr_dispatchers);
- if (ret)
- return NULL;
-
- CHANGELOG_MAKE_SOCKET_PATH (priv->changelog_brick,
- sockfile, UNIX_PATH_MAX);
- (void) sys_unlink (sockfile);
- svcp = changelog_rpc_server_init (this, sockfile, NULL,
- changelog_rpcsvc_notify,
- changelog_programs);
- return svcp;
+ int ret = 0;
+ char sockfile[UNIX_PATH_MAX] = {
+ 0,
+ };
+ rpcsvc_t *svcp;
+
+ ret = changelog_init_rpc_threads(this, priv, rbuf, nr_dispatchers);
+ if (ret)
+ return NULL;
+
+ CHANGELOG_MAKE_SOCKET_PATH(priv->changelog_brick, sockfile, UNIX_PATH_MAX);
+ (void)sys_unlink(sockfile);
+ svcp = changelog_rpc_server_init(
+ this, sockfile, NULL, changelog_rpcsvc_notify, changelog_programs);
+ return svcp;
}
void
-changelog_rpc_clnt_cleanup (changelog_rpc_clnt_t *crpc)
+changelog_rpc_clnt_cleanup(changelog_rpc_clnt_t *crpc)
{
- if (!crpc)
- return;
- crpc->c_clnt = NULL;
- LOCK_DESTROY (&crpc->lock);
- GF_FREE (crpc);
+ if (!crpc)
+ return;
+ crpc->c_clnt = NULL;
+ LOCK_DESTROY(&crpc->lock);
+ GF_FREE(crpc);
}
static changelog_rpc_clnt_t *
-changelog_rpc_clnt_init (xlator_t *this,
- changelog_probe_req *rpc_req, changelog_clnt_t *c_clnt)
+changelog_rpc_clnt_init(xlator_t *this, changelog_probe_req *rpc_req,
+ changelog_clnt_t *c_clnt)
{
- int ret = 0;
- changelog_rpc_clnt_t *crpc = NULL;
-
- crpc = GF_CALLOC (1, sizeof (*crpc), gf_changelog_mt_rpc_clnt_t);
- if (!crpc)
- goto error_return;
- INIT_LIST_HEAD (&crpc->list);
-
- /* Take a ref, the last unref will be on RPC_CLNT_DESTROY
- * which comes as a result of last rpc_clnt_unref.
- */
- GF_ATOMIC_INIT (crpc->ref, 1);
- changelog_set_disconnect_flag (crpc, _gf_false);
-
- crpc->filter = rpc_req->filter;
- (void) memcpy (crpc->sock, rpc_req->sock, strlen (rpc_req->sock));
-
- crpc->this = this;
- crpc->c_clnt = c_clnt;
- crpc->cleanup = changelog_rpc_clnt_cleanup;
-
- ret = LOCK_INIT (&crpc->lock);
- if (ret != 0)
- goto dealloc_crpc;
- return crpc;
-
- dealloc_crpc:
- GF_FREE (crpc);
- error_return:
- return NULL;
+ int ret = 0;
+ changelog_rpc_clnt_t *crpc = NULL;
+
+ crpc = GF_CALLOC(1, sizeof(*crpc), gf_changelog_mt_rpc_clnt_t);
+ if (!crpc)
+ goto error_return;
+ INIT_LIST_HEAD(&crpc->list);
+
+ /* Take a ref, the last unref will be on RPC_CLNT_DESTROY
+ * which comes as a result of last rpc_clnt_unref.
+ */
+ GF_ATOMIC_INIT(crpc->ref, 1);
+ changelog_set_disconnect_flag(crpc, _gf_false);
+
+ crpc->filter = rpc_req->filter;
+ (void)memcpy(crpc->sock, rpc_req->sock, strlen(rpc_req->sock));
+
+ crpc->this = this;
+ crpc->c_clnt = c_clnt;
+ crpc->cleanup = changelog_rpc_clnt_cleanup;
+
+ ret = LOCK_INIT(&crpc->lock);
+ if (ret != 0)
+ goto dealloc_crpc;
+ return crpc;
+
+dealloc_crpc:
+ GF_FREE(crpc);
+error_return:
+ return NULL;
}
/**
@@ -246,72 +361,80 @@ changelog_rpc_clnt_init (xlator_t *this,
*/
int
-changelog_handle_probe (rpcsvc_request_t *req)
+changelog_handle_probe(rpcsvc_request_t *req)
{
- int ret = 0;
- xlator_t *this = NULL;
- rpcsvc_t *svc = NULL;
- changelog_priv_t *priv = NULL;
- changelog_clnt_t *c_clnt = NULL;
- changelog_rpc_clnt_t *crpc = NULL;
-
- changelog_probe_req rpc_req = {0,};
- changelog_probe_rsp rpc_rsp = {0,};
-
- ret = xdr_to_generic (req->msg[0],
- &rpc_req, (xdrproc_t)xdr_changelog_probe_req);
- if (ret < 0) {
- gf_msg ("", GF_LOG_ERROR, 0,
- CHANGELOG_MSG_HANDLE_PROBE_ERROR,
- "xdr decoding error");
- req->rpc_err = GARBAGE_ARGS;
- goto handle_xdr_error;
- }
-
- /* ->xl hidden in rpcsvc */
- svc = rpcsvc_request_service (req);
- this = svc->xl;
- priv = this->private;
- c_clnt = &priv->connections;
-
- crpc = changelog_rpc_clnt_init (this, &rpc_req, c_clnt);
- if (!crpc)
- goto handle_xdr_error;
-
- changelog_ev_queue_connection (c_clnt, crpc);
- rpc_rsp.op_ret = 0;
-
- goto submit_rpc;
-
- handle_xdr_error:
- rpc_rsp.op_ret = -1;
- submit_rpc:
- (void) changelog_rpc_sumbit_reply (req, &rpc_rsp, NULL, 0, NULL,
- (xdrproc_t)xdr_changelog_probe_rsp);
+ int ret = 0;
+ xlator_t *this = NULL;
+ rpcsvc_t *svc = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_clnt_t *c_clnt = NULL;
+ changelog_rpc_clnt_t *crpc = NULL;
+
+ changelog_probe_req rpc_req = {
+ 0,
+ };
+ changelog_probe_rsp rpc_rsp = {
+ 0,
+ };
+
+ this = req->trans->xl;
+ if (this->cleanup_starting) {
+ gf_smsg(this->name, GF_LOG_DEBUG, 0, CHANGELOG_MSG_CLEANUP_ALREADY_SET,
+ NULL);
return 0;
+ }
+
+ ret = xdr_to_generic(req->msg[0], &rpc_req,
+ (xdrproc_t)xdr_changelog_probe_req);
+ if (ret < 0) {
+ gf_smsg("", GF_LOG_ERROR, 0, CHANGELOG_MSG_HANDLE_PROBE_ERROR, NULL);
+ req->rpc_err = GARBAGE_ARGS;
+ goto handle_xdr_error;
+ }
+
+ /* ->xl hidden in rpcsvc */
+ svc = rpcsvc_request_service(req);
+ this = svc->xl;
+ priv = this->private;
+ c_clnt = &priv->connections;
+
+ crpc = changelog_rpc_clnt_init(this, &rpc_req, c_clnt);
+ if (!crpc)
+ goto handle_xdr_error;
+
+ changelog_ev_queue_connection(c_clnt, crpc);
+ rpc_rsp.op_ret = 0;
+
+ goto submit_rpc;
+
+handle_xdr_error:
+ rpc_rsp.op_ret = -1;
+submit_rpc:
+ (void)changelog_rpc_sumbit_reply(req, &rpc_rsp, NULL, 0, NULL,
+ (xdrproc_t)xdr_changelog_probe_rsp);
+ return 0;
}
/**
* RPC declarations
*/
-rpcsvc_actor_t changelog_svc_actors[CHANGELOG_RPC_PROC_MAX] = {
- [CHANGELOG_RPC_PROBE_FILTER] = {
- "CHANGELOG PROBE FILTER", CHANGELOG_RPC_PROBE_FILTER,
- changelog_handle_probe, NULL, 0, DRC_NA
- },
+static rpcsvc_actor_t changelog_svc_actors[CHANGELOG_RPC_PROC_MAX] = {
+ [CHANGELOG_RPC_PROBE_FILTER] = {"CHANGELOG PROBE FILTER",
+ changelog_handle_probe, NULL,
+ CHANGELOG_RPC_PROBE_FILTER, DRC_NA, 0},
};
-struct rpcsvc_program changelog_svc_prog = {
- .progname = CHANGELOG_RPC_PROGNAME,
- .prognum = CHANGELOG_RPC_PROGNUM,
- .progver = CHANGELOG_RPC_PROGVER,
- .numactors = CHANGELOG_RPC_PROC_MAX,
- .actors = changelog_svc_actors,
- .synctask = _gf_true,
+static struct rpcsvc_program changelog_svc_prog = {
+ .progname = CHANGELOG_RPC_PROGNAME,
+ .prognum = CHANGELOG_RPC_PROGNUM,
+ .progver = CHANGELOG_RPC_PROGVER,
+ .numactors = CHANGELOG_RPC_PROC_MAX,
+ .actors = changelog_svc_actors,
+ .synctask = _gf_true,
};
-struct rpcsvc_program *changelog_programs[] = {
- &changelog_svc_prog,
- NULL,
+static struct rpcsvc_program *changelog_programs[] = {
+ &changelog_svc_prog,
+ NULL,
};
diff --git a/xlators/features/changelog/src/changelog-rpc.h b/xlators/features/changelog/src/changelog-rpc.h
index 9f1e2d223a3..b1707565249 100644
--- a/xlators/features/changelog/src/changelog-rpc.h
+++ b/xlators/features/changelog/src/changelog-rpc.h
@@ -11,21 +11,21 @@
#ifndef __CHANGELOG_RPC_H
#define __CHANGELOG_RPC_H
-#include "xlator.h"
+#include <glusterfs/xlator.h>
#include "changelog-helpers.h"
/* one time */
#include "socket.h"
#include "changelog-rpc-common.h"
-#define CHANGELOG_RPC_PROGNAME "GlusterFS Changelog"
+#define CHANGELOG_RPC_PROGNAME "GlusterFS Changelog"
rpcsvc_t *
-changelog_init_rpc_listener (xlator_t *, changelog_priv_t *, rbuf_t *, int);
+changelog_init_rpc_listener(xlator_t *, changelog_priv_t *, rbuf_t *, int);
void
-changelog_destroy_rpc_listner (xlator_t *, changelog_priv_t *);
+changelog_destroy_rpc_listner(xlator_t *, changelog_priv_t *);
int
-changelog_cleanup_rpc_threads (xlator_t *this, changelog_priv_t *priv);
+changelog_cleanup_rpc_threads(xlator_t *this, changelog_priv_t *priv);
#endif
diff --git a/xlators/features/changelog/src/changelog-rt.c b/xlators/features/changelog/src/changelog-rt.c
index c262820c64c..841545ae359 100644
--- a/xlators/features/changelog/src/changelog-rt.c
+++ b/xlators/features/changelog/src/changelog-rt.c
@@ -8,60 +8,59 @@
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "defaults.h"
-#include "logging.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
+#include <glusterfs/logging.h>
#include "changelog-rt.h"
#include "changelog-mem-types.h"
int
-changelog_rt_init (xlator_t *this, changelog_dispatcher_t *cd)
+changelog_rt_init(xlator_t *this, changelog_dispatcher_t *cd)
{
- changelog_rt_t *crt = NULL;
+ changelog_rt_t *crt = NULL;
- crt = GF_CALLOC (1, sizeof (*crt),
- gf_changelog_mt_rt_t);
- if (!crt)
- return -1;
+ crt = GF_CALLOC(1, sizeof(*crt), gf_changelog_mt_rt_t);
+ if (!crt)
+ return -1;
- LOCK_INIT (&crt->lock);
+ LOCK_INIT(&crt->lock);
- cd->cd_data = crt;
- cd->dispatchfn = &changelog_rt_enqueue;
+ cd->cd_data = crt;
+ cd->dispatchfn = &changelog_rt_enqueue;
- return 0;
+ return 0;
}
int
-changelog_rt_fini (xlator_t *this, changelog_dispatcher_t *cd)
+changelog_rt_fini(xlator_t *this, changelog_dispatcher_t *cd)
{
- changelog_rt_t *crt = NULL;
+ changelog_rt_t *crt = NULL;
- crt = cd->cd_data;
+ crt = cd->cd_data;
- LOCK_DESTROY (&crt->lock);
- GF_FREE (crt);
+ LOCK_DESTROY(&crt->lock);
+ GF_FREE(crt);
- return 0;
+ return 0;
}
int
-changelog_rt_enqueue (xlator_t *this, changelog_priv_t *priv, void *cbatch,
- changelog_log_data_t *cld_0, changelog_log_data_t *cld_1)
+changelog_rt_enqueue(xlator_t *this, changelog_priv_t *priv, void *cbatch,
+ changelog_log_data_t *cld_0, changelog_log_data_t *cld_1)
{
- int ret = 0;
- changelog_rt_t *crt = NULL;
+ int ret = 0;
+ changelog_rt_t *crt = NULL;
- crt = (changelog_rt_t *) cbatch;
+ crt = (changelog_rt_t *)cbatch;
- LOCK (&crt->lock);
- {
- ret = changelog_handle_change (this, priv, cld_0);
- if (!ret && cld_1)
- ret = changelog_handle_change (this, priv, cld_1);
- }
- UNLOCK (&crt->lock);
+ LOCK(&crt->lock);
+ {
+ ret = changelog_handle_change(this, priv, cld_0);
+ if (!ret && cld_1)
+ ret = changelog_handle_change(this, priv, cld_1);
+ }
+ UNLOCK(&crt->lock);
- return ret;
+ return ret;
}
diff --git a/xlators/features/changelog/src/changelog-rt.h b/xlators/features/changelog/src/changelog-rt.h
index 1fc2bbc5bb9..28b9827d85b 100644
--- a/xlators/features/changelog/src/changelog-rt.h
+++ b/xlators/features/changelog/src/changelog-rt.h
@@ -11,23 +11,23 @@
#ifndef _CHANGELOG_RT_H
#define _CHANGELOG_RT_H
-#include "locking.h"
-#include "timer.h"
+#include <glusterfs/locking.h>
+#include <glusterfs/timer.h>
#include "pthread.h"
#include "changelog-helpers.h"
/* unused as of now - may be you would need it later */
typedef struct changelog_rt {
- gf_lock_t lock;
+ gf_lock_t lock;
} changelog_rt_t;
int
-changelog_rt_init (xlator_t *this, changelog_dispatcher_t *cd);
+changelog_rt_init(xlator_t *this, changelog_dispatcher_t *cd);
int
-changelog_rt_fini (xlator_t *this, changelog_dispatcher_t *cd);
+changelog_rt_fini(xlator_t *this, changelog_dispatcher_t *cd);
int
-changelog_rt_enqueue (xlator_t *this, changelog_priv_t *priv, void *cbatch,
- changelog_log_data_t *cld_0, changelog_log_data_t *cld_1);
+changelog_rt_enqueue(xlator_t *this, changelog_priv_t *priv, void *cbatch,
+ changelog_log_data_t *cld_0, changelog_log_data_t *cld_1);
#endif /* _CHANGELOG_RT_H */
diff --git a/xlators/features/changelog/src/changelog.c b/xlators/features/changelog/src/changelog.c
index 464a28dde6e..6a6e5af859e 100644
--- a/xlators/features/changelog/src/changelog.c
+++ b/xlators/features/changelog/src/changelog.c
@@ -8,11 +8,11 @@
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "defaults.h"
-#include "syscall.h"
-#include "logging.h"
-#include "iobuf.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
+#include <glusterfs/syscall.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/iobuf.h>
#include "changelog-rt.h"
@@ -26,15 +26,20 @@
#include "changelog-rpc.h"
#include "errno.h"
-static struct changelog_bootstrap
-cb_bootstrap[] = {
- {
- .mode = CHANGELOG_MODE_RT,
- .ctor = changelog_rt_init,
- .dtor = changelog_rt_fini,
- },
+static struct changelog_bootstrap cb_bootstrap[] = {
+ {
+ .mode = CHANGELOG_MODE_RT,
+ .ctor = changelog_rt_init,
+ .dtor = changelog_rt_fini,
+ },
};
+static int
+changelog_init_rpc(xlator_t *this, changelog_priv_t *priv);
+
+static int
+changelog_init(xlator_t *this, changelog_priv_t *priv);
+
/* Entry operations - TYPE III */
/**
@@ -46,1111 +51,1052 @@ cb_bootstrap[] = {
/* rmdir */
int32_t
-changelog_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+changelog_rmdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_ENTRY);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (rmdir, frame, op_ret, op_errno,
- preparent, postparent, xdata);
- return 0;
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(rmdir, frame, op_ret, op_errno, preparent,
+ postparent, xdata);
+ return 0;
}
int32_t
-changelog_rmdir_resume (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int xflags, dict_t *xdata)
+changelog_rmdir_resume(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ int xflags, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
+ changelog_priv_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- gf_msg_debug (this->name, 0, "Dequeue rmdir");
- changelog_color_fop_and_inc_cnt (this, priv,
- frame->local);
- STACK_WIND (frame, changelog_rmdir_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->rmdir,
- loc, xflags, xdata);
- return 0;
+ gf_msg_debug(this->name, 0, "Dequeue rmdir");
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_rmdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rmdir, loc, xflags, xdata);
+ return 0;
}
int32_t
-changelog_rmdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int xflags, dict_t *xdata)
-{
- size_t xtra_len = 0;
- changelog_priv_t *priv = NULL;
- changelog_opt_t *co = NULL;
- call_stub_t *stub = NULL;
- struct list_head queue = {0, };
- gf_boolean_t barrier_enabled = _gf_false;
-
- INIT_LIST_HEAD (&queue);
-
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
-
- CHANGELOG_INIT_NOCHECK (this, frame->local,
- NULL, loc->inode->gfid, 2);
-
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
-
- CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
-
- co++;
- if (priv->capture_del_path) {
- CHANGELOG_FILL_ENTRY_DIR_PATH (co, loc->pargfid, loc->name,
- del_entry_fn, del_entry_free_fn,
- xtra_len, wind, _gf_true);
+changelog_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflags,
+ dict_t *xdata)
+{
+ size_t xtra_len = 0;
+ changelog_priv_t *priv = NULL;
+ changelog_opt_t *co = NULL;
+ call_stub_t *stub = NULL;
+ struct list_head queue = {
+ 0,
+ };
+ gf_boolean_t barrier_enabled = _gf_false;
+
+ INIT_LIST_HEAD(&queue);
+
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
+
+ CHANGELOG_INIT_NOCHECK(this, frame->local, NULL, loc->inode->gfid, 2);
+
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
+
+ CHANGLOG_FILL_FOP_NUMBER(co, frame->root->op, fop_fn, xtra_len);
+
+ co++;
+ if (priv->capture_del_path) {
+ CHANGELOG_FILL_ENTRY_DIR_PATH(co, loc->pargfid, loc->name, del_entry_fn,
+ del_entry_free_fn, xtra_len, wind,
+ _gf_true);
+ } else {
+ CHANGELOG_FILL_ENTRY_DIR_PATH(co, loc->pargfid, loc->name, del_entry_fn,
+ del_entry_free_fn, xtra_len, wind,
+ _gf_false);
+ }
+
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 2);
+
+ /* changelog barrier */
+ /* Color assignment and increment of fop_cnt for rmdir/unlink/rename
+ * should be made with in priv lock if changelog barrier is not enabled.
+ * Because if counter is not incremented yet, draining wakes up and
+ * publishes the changelog but later these fops might hit the disk and
+ * present in snapped volume but where as the intention is these fops
+ * should not be present in snapped volume.
+ */
+ LOCK(&priv->lock);
+ {
+ if ((barrier_enabled = priv->barrier_enabled)) {
+ stub = fop_rmdir_stub(frame, changelog_rmdir_resume, loc, xflags,
+ xdata);
+ if (!stub)
+ __chlog_barrier_disable(this, &queue);
+ else
+ __chlog_barrier_enqueue(this, stub);
} else {
- CHANGELOG_FILL_ENTRY_DIR_PATH (co, loc->pargfid, loc->name,
- del_entry_fn, del_entry_free_fn,
- xtra_len, wind, _gf_false);
- }
-
- changelog_set_usable_record_and_length (frame->local, xtra_len, 2);
-
-/* changelog barrier */
- /* Color assignment and increment of fop_cnt for rmdir/unlink/rename
- * should be made with in priv lock if changelog barrier is not enabled.
- * Because if counter is not incremented yet, draining wakes up and
- * publishes the changelog but later these fops might hit the disk and
- * present in snapped volume but where as the intention is these fops
- * should not be present in snapped volume.
- */
- LOCK (&priv->lock);
- {
- if ((barrier_enabled = priv->barrier_enabled)) {
- stub = fop_rmdir_stub (frame, changelog_rmdir_resume,
- loc, xflags, xdata);
- if (!stub)
- __chlog_barrier_disable (this, &queue);
- else
- __chlog_barrier_enqueue (this, stub);
- } else {
- ((changelog_local_t *)frame->local)->color
- = priv->current_color;
- changelog_inc_fop_cnt (this, priv, frame->local);
- }
- }
- UNLOCK (&priv->lock);
-
- if (barrier_enabled && stub) {
- gf_msg_debug (this->name, 0, "Enqueue rmdir");
- goto out;
- }
- if (barrier_enabled && !stub) {
- gf_smsg (this->name, GF_LOG_ERROR, ENOMEM,
- CHANGELOG_MSG_NO_MEMORY,
- "Failed to barrier FOPs, disabling changelog barrier",
- "fop=rmdir",
- NULL);
- chlog_barrier_dequeue_all (this, &queue);
- }
-
-/* changelog barrier */
-
- wind:
- STACK_WIND (frame, changelog_rmdir_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->rmdir,
- loc, xflags, xdata);
- out:
- return 0;
+ ((changelog_local_t *)frame->local)->color = priv->current_color;
+ changelog_inc_fop_cnt(this, priv, frame->local);
+ }
+ }
+ UNLOCK(&priv->lock);
+
+ if (barrier_enabled && stub) {
+ gf_msg_debug(this->name, 0, "Enqueue rmdir");
+ goto out;
+ }
+ if (barrier_enabled && !stub) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM,
+ CHANGELOG_MSG_BARRIER_FOP_FAILED, "fop=rmdir", NULL);
+ chlog_barrier_dequeue_all(this, &queue);
+ }
+
+ /* changelog barrier */
+
+wind:
+ STACK_WIND(frame, changelog_rmdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rmdir, loc, xflags, xdata);
+out:
+ return 0;
}
/* unlink */
int32_t
-changelog_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+changelog_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_ENTRY);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (unlink, frame, op_ret, op_errno,
- preparent, postparent, xdata);
- return 0;
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(unlink, frame, op_ret, op_errno, preparent,
+ postparent, xdata);
+ return 0;
}
int32_t
-changelog_unlink_resume (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int xflags, dict_t *xdata)
+changelog_unlink_resume(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ int xflags, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
+ changelog_priv_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- gf_msg_debug (this->name, 0, "Dequeue unlink");
- changelog_color_fop_and_inc_cnt
- (this, priv, frame->local);
- STACK_WIND (frame, changelog_unlink_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->unlink,
- loc, xflags, xdata);
- return 0;
+ gf_msg_debug(this->name, 0, "Dequeue unlink");
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_unlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, xflags, xdata);
+ return 0;
}
int32_t
-changelog_unlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int xflags, dict_t *xdata)
-{
- size_t xtra_len = 0;
- changelog_priv_t *priv = NULL;
- changelog_opt_t *co = NULL;
- call_stub_t *stub = NULL;
- struct list_head queue = {0, };
- gf_boolean_t barrier_enabled = _gf_false;
- dht_changelog_rename_info_t *info = NULL;
- int ret = 0;
- char *old_name = NULL;
- char *new_name = NULL;
- char *nname = NULL;
-
- INIT_LIST_HEAD (&queue);
- priv = this->private;
-
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
-
- ret = dict_get_bin (xdata, DHT_CHANGELOG_RENAME_OP_KEY, (void **)&info);
- if (!ret) { /* special case: unlink considered as rename */
- /* 3 == fop + oldloc + newloc */
- old_name = alloca (info->oldname_len);
- new_name = alloca (info->newname_len);
- CHANGELOG_INIT_NOCHECK (this, frame->local,
- NULL, loc->inode->gfid, 3);
-
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
-
- CHANGLOG_FILL_FOP_NUMBER (co, GF_FOP_RENAME, fop_fn, xtra_len);
-
- co++;
- strncpy (old_name, info->buffer, info->oldname_len);
- CHANGELOG_FILL_ENTRY (co, info->old_pargfid, old_name,
- entry_fn, entry_free_fn, xtra_len, wind);
-
- co++;
- /* new name resides just after old name */
- nname = info->buffer + info->oldname_len;
- strncpy (new_name, nname, info->newname_len);
- CHANGELOG_FILL_ENTRY (co, info->new_pargfid, new_name,
- entry_fn, entry_free_fn, xtra_len, wind);
-
- changelog_set_usable_record_and_length (frame->local,
- xtra_len, 3);
- } else { /* default unlink */
- CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, wind);
- CHANGELOG_INIT_NOCHECK (this, frame->local, NULL,
- loc->inode->gfid, 2);
-
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
-
- CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op,
- fop_fn, xtra_len);
-
- co++;
- if (priv->capture_del_path) {
- CHANGELOG_FILL_ENTRY_DIR_PATH (co, loc->pargfid,
- loc->name, del_entry_fn, del_entry_free_fn,
- xtra_len, wind, _gf_true);
- } else {
- CHANGELOG_FILL_ENTRY_DIR_PATH (co, loc->pargfid,
- loc->name, del_entry_fn, del_entry_free_fn,
- xtra_len, wind, _gf_false);
- }
+changelog_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflags,
+ dict_t *xdata)
+{
+ size_t xtra_len = 0;
+ changelog_priv_t *priv = NULL;
+ changelog_opt_t *co = NULL;
+ call_stub_t *stub = NULL;
+ struct list_head queue = {
+ 0,
+ };
+ gf_boolean_t barrier_enabled = _gf_false;
+ dht_changelog_rename_info_t *info = NULL;
+ int ret = 0;
+ char *old_name = NULL;
+ char *new_name = NULL;
+ char *nname = NULL;
+
+ INIT_LIST_HEAD(&queue);
+ priv = this->private;
+
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
+
+ ret = dict_get_bin(xdata, DHT_CHANGELOG_RENAME_OP_KEY, (void **)&info);
+ if (!ret) { /* special case: unlink considered as rename */
+ /* 3 == fop + oldloc + newloc */
+ old_name = alloca(info->oldname_len);
+ new_name = alloca(info->newname_len);
+ CHANGELOG_INIT_NOCHECK(this, frame->local, NULL, loc->inode->gfid, 3);
- changelog_set_usable_record_and_length (frame->local,
- xtra_len, 2);
- }
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
-/* changelog barrier */
- LOCK (&priv->lock);
- {
- if ((barrier_enabled = priv->barrier_enabled)) {
- stub = fop_unlink_stub (frame, changelog_unlink_resume,
- loc, xflags, xdata);
- if (!stub)
- __chlog_barrier_disable (this, &queue);
- else
- __chlog_barrier_enqueue (this, stub);
- } else {
- ((changelog_local_t *)frame->local)->color
- = priv->current_color;
- changelog_inc_fop_cnt (this, priv, frame->local);
- }
- }
- UNLOCK (&priv->lock);
+ CHANGLOG_FILL_FOP_NUMBER(co, GF_FOP_RENAME, fop_fn, xtra_len);
- if (barrier_enabled && stub) {
- gf_msg_debug (this->name, 0, "Enqueue unlink");
- goto out;
- }
- if (barrier_enabled && !stub) {
- gf_smsg (this->name, GF_LOG_ERROR, ENOMEM,
- CHANGELOG_MSG_NO_MEMORY,
- "Failed to barrier FOPs, disabling changelog barrier",
- "fop=unlink",
- NULL);
- chlog_barrier_dequeue_all (this, &queue);
- }
+ co++;
+ strncpy(old_name, info->buffer, info->oldname_len);
+ CHANGELOG_FILL_ENTRY(co, info->old_pargfid, old_name, entry_fn,
+ entry_free_fn, xtra_len, wind);
-/* changelog barrier */
+ co++;
+ /* new name resides just after old name */
+ nname = info->buffer + info->oldname_len;
+ strncpy(new_name, nname, info->newname_len);
+ CHANGELOG_FILL_ENTRY(co, info->new_pargfid, new_name, entry_fn,
+ entry_free_fn, xtra_len, wind);
+
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 3);
+ } else { /* default unlink */
+ CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO(frame, xdata, wind);
+ CHANGELOG_INIT_NOCHECK(this, frame->local, NULL, loc->inode->gfid, 2);
+
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
- wind:
- STACK_WIND (frame, changelog_unlink_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->unlink,
- loc, xflags, xdata);
- out:
- return 0;
+ CHANGLOG_FILL_FOP_NUMBER(co, frame->root->op, fop_fn, xtra_len);
+
+ co++;
+ if (priv->capture_del_path) {
+ CHANGELOG_FILL_ENTRY_DIR_PATH(co, loc->pargfid, loc->name,
+ del_entry_fn, del_entry_free_fn,
+ xtra_len, wind, _gf_true);
+ } else {
+ CHANGELOG_FILL_ENTRY_DIR_PATH(co, loc->pargfid, loc->name,
+ del_entry_fn, del_entry_free_fn,
+ xtra_len, wind, _gf_false);
+ }
+
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 2);
+ }
+
+ /* changelog barrier */
+ LOCK(&priv->lock);
+ {
+ if ((barrier_enabled = priv->barrier_enabled)) {
+ stub = fop_unlink_stub(frame, changelog_unlink_resume, loc, xflags,
+ xdata);
+ if (!stub)
+ __chlog_barrier_disable(this, &queue);
+ else
+ __chlog_barrier_enqueue(this, stub);
+ } else {
+ ((changelog_local_t *)frame->local)->color = priv->current_color;
+ changelog_inc_fop_cnt(this, priv, frame->local);
+ }
+ }
+ UNLOCK(&priv->lock);
+
+ if (barrier_enabled && stub) {
+ gf_msg_debug(this->name, 0, "Enqueue unlink");
+ goto out;
+ }
+ if (barrier_enabled && !stub) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM,
+ CHANGELOG_MSG_BARRIER_FOP_FAILED, "fop=unlink", NULL);
+ chlog_barrier_dequeue_all(this, &queue);
+ }
+
+ /* changelog barrier */
+
+wind:
+ STACK_WIND(frame, changelog_unlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, xflags, xdata);
+out:
+ return 0;
}
/* rename */
int32_t
-changelog_rename_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *buf, struct iatt *preoldparent,
- struct iatt *postoldparent, struct iatt *prenewparent,
- struct iatt *postnewparent, dict_t *xdata)
-{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
-
- priv = this->private;
- local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (rename, frame, op_ret, op_errno,
- buf, preoldparent, postoldparent,
- prenewparent, postnewparent, xdata);
- return 0;
+changelog_rename_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ struct iatt *preoldparent, struct iatt *postoldparent,
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
+{
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
+
+ priv = this->private;
+ local = frame->local;
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_ENTRY);
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(rename, frame, op_ret, op_errno, buf, preoldparent,
+ postoldparent, prenewparent, postnewparent, xdata);
+ return 0;
}
int32_t
-changelog_rename_resume (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+changelog_rename_resume(call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
+ changelog_priv_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- gf_msg_debug (this->name, 0, "Dequeue rename");
- changelog_color_fop_and_inc_cnt
- (this, priv, frame->local);
- STACK_WIND (frame, changelog_rename_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->rename,
- oldloc, newloc, xdata);
- return 0;
+ gf_msg_debug(this->name, 0, "Dequeue rename");
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_rename_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
+ return 0;
}
int32_t
-changelog_rename (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
-{
- size_t xtra_len = 0;
- changelog_priv_t *priv = NULL;
- changelog_opt_t *co = NULL;
- call_stub_t *stub = NULL;
- struct list_head queue = {0, };
- gf_boolean_t barrier_enabled = _gf_false;
- dht_changelog_rename_info_t *info = NULL;
- int ret = 0;
-
- INIT_LIST_HEAD (&queue);
-
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
-
- ret = dict_get_bin (xdata, DHT_CHANGELOG_RENAME_OP_KEY, (void **)&info);
- if (ret && oldloc->inode->ia_type != IA_IFDIR) {
- /* xdata "NOT" set for a non-directory,
- * Special rename => avoid logging */
- goto wind;
- }
-
- /* 3 == fop + oldloc + newloc */
- CHANGELOG_INIT_NOCHECK (this, frame->local,
- NULL, oldloc->inode->gfid, 3);
-
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
-
- CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
-
- co++;
- CHANGELOG_FILL_ENTRY (co, oldloc->pargfid, oldloc->name,
- entry_fn, entry_free_fn, xtra_len, wind);
-
- co++;
- CHANGELOG_FILL_ENTRY (co, newloc->pargfid, newloc->name,
- entry_fn, entry_free_fn, xtra_len, wind);
-
- changelog_set_usable_record_and_length (frame->local, xtra_len, 3);
-/* changelog barrier */
- LOCK (&priv->lock);
- {
- if ((barrier_enabled = priv->barrier_enabled)) {
- stub = fop_rename_stub (frame, changelog_rename_resume,
- oldloc, newloc, xdata);
- if (!stub)
- __chlog_barrier_disable (this, &queue);
- else
- __chlog_barrier_enqueue (this, stub);
- } else {
- ((changelog_local_t *)frame->local)->color
- = priv->current_color;
- changelog_inc_fop_cnt (this, priv, frame->local);
- }
- }
- UNLOCK (&priv->lock);
-
- if (barrier_enabled && stub) {
- gf_msg_debug (this->name, 0, "Enqueue rename");
- goto out;
- }
- if (barrier_enabled && !stub) {
- gf_smsg (this->name, GF_LOG_ERROR, ENOMEM,
- CHANGELOG_MSG_NO_MEMORY,
- "Failed to barrier FOPs, disabling changelog barrier",
- "fop=rename",
- NULL);
- chlog_barrier_dequeue_all (this, &queue);
- }
-/* changelog barrier */
-
- wind:
- STACK_WIND (frame, changelog_rename_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->rename,
- oldloc, newloc, xdata);
- out:
- return 0;
+changelog_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata)
+{
+ size_t xtra_len = 0;
+ changelog_priv_t *priv = NULL;
+ changelog_opt_t *co = NULL;
+ call_stub_t *stub = NULL;
+ struct list_head queue = {
+ 0,
+ };
+ gf_boolean_t barrier_enabled = _gf_false;
+ dht_changelog_rename_info_t *info = NULL;
+ int ret = 0;
+
+ INIT_LIST_HEAD(&queue);
+
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
+
+ ret = dict_get_bin(xdata, DHT_CHANGELOG_RENAME_OP_KEY, (void **)&info);
+ if (ret && oldloc->inode->ia_type != IA_IFDIR) {
+ /* xdata "NOT" set for a non-directory,
+ * Special rename => avoid logging */
+ goto wind;
+ }
+
+ /* 3 == fop + oldloc + newloc */
+ CHANGELOG_INIT_NOCHECK(this, frame->local, NULL, oldloc->inode->gfid, 3);
+
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
+
+ CHANGLOG_FILL_FOP_NUMBER(co, frame->root->op, fop_fn, xtra_len);
+
+ co++;
+ CHANGELOG_FILL_ENTRY(co, oldloc->pargfid, oldloc->name, entry_fn,
+ entry_free_fn, xtra_len, wind);
+
+ co++;
+ CHANGELOG_FILL_ENTRY(co, newloc->pargfid, newloc->name, entry_fn,
+ entry_free_fn, xtra_len, wind);
+
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 3);
+ /* changelog barrier */
+ LOCK(&priv->lock);
+ {
+ if ((barrier_enabled = priv->barrier_enabled)) {
+ stub = fop_rename_stub(frame, changelog_rename_resume, oldloc,
+ newloc, xdata);
+ if (!stub)
+ __chlog_barrier_disable(this, &queue);
+ else
+ __chlog_barrier_enqueue(this, stub);
+ } else {
+ ((changelog_local_t *)frame->local)->color = priv->current_color;
+ changelog_inc_fop_cnt(this, priv, frame->local);
+ }
+ }
+ UNLOCK(&priv->lock);
+
+ if (barrier_enabled && stub) {
+ gf_msg_debug(this->name, 0, "Enqueue rename");
+ goto out;
+ }
+ if (barrier_enabled && !stub) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM,
+ CHANGELOG_MSG_BARRIER_FOP_FAILED, "fop=rename", NULL);
+ chlog_barrier_dequeue_all(this, &queue);
+ }
+ /* changelog barrier */
+
+wind:
+ STACK_WIND(frame, changelog_rename_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
+out:
+ return 0;
}
/* link */
int32_t
-changelog_link_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this, int32_t op_ret,
- int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+changelog_link_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_ENTRY);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (link, frame, op_ret, op_errno,
- inode, buf, preparent, postparent, xdata);
- return 0;
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(link, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
+ return 0;
}
int32_t
-changelog_link_resume (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+changelog_link_resume(call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
+ changelog_priv_t *priv = NULL;
- GF_VALIDATE_OR_GOTO ("changelog", this, out);
- GF_VALIDATE_OR_GOTO ("changelog", this->fops, out);
- GF_VALIDATE_OR_GOTO ("changelog", frame, out);
+ GF_VALIDATE_OR_GOTO("changelog", this, out);
+ GF_VALIDATE_OR_GOTO("changelog", this->fops, out);
+ GF_VALIDATE_OR_GOTO("changelog", frame, out);
- priv = this->private;
+ priv = this->private;
- gf_msg_debug (this->name, 0, "Dequeuing link");
- changelog_color_fop_and_inc_cnt
- (this, priv, frame->local);
- STACK_WIND (frame, changelog_link_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->link,
- oldloc, newloc, xdata);
- return 0;
+ gf_msg_debug(this->name, 0, "Dequeuing link");
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_link_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
+ return 0;
out:
- return -1;
+ return -1;
}
int32_t
-changelog_link (call_frame_t *frame,
- xlator_t *this, loc_t *oldloc,
- loc_t *newloc, dict_t *xdata)
-{
- size_t xtra_len = 0;
- changelog_priv_t *priv = NULL;
- changelog_opt_t *co = NULL;
- call_stub_t *stub = NULL;
- struct list_head queue = {0, };
- gf_boolean_t barrier_enabled = _gf_false;
-
- priv = this->private;
-
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
- CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, wind);
-
- CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, oldloc->gfid, 2);
-
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
-
- CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
-
- co++;
- CHANGELOG_FILL_ENTRY (co, newloc->pargfid, newloc->name,
- entry_fn, entry_free_fn, xtra_len, wind);
-
- changelog_set_usable_record_and_length (frame->local, xtra_len, 2);
-
- LOCK (&priv->lock);
- {
- if ((barrier_enabled = priv->barrier_enabled)) {
- stub = fop_link_stub (frame, changelog_link_resume,
- oldloc, newloc, xdata);
- if (!stub)
- __chlog_barrier_disable (this, &queue);
- else
- __chlog_barrier_enqueue (this, stub);
- } else {
- ((changelog_local_t *)frame->local)->color
- = priv->current_color;
- changelog_inc_fop_cnt (this, priv, frame->local);
- }
- }
- UNLOCK (&priv->lock);
-
- if (barrier_enabled && stub) {
- gf_msg_debug (this->name, 0, "Enqueued link");
- goto out;
- }
-
- if (barrier_enabled && !stub) {
- gf_smsg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_NO_MEMORY,
- "Failed to barrier FOPs, disabling changelog barrier",
- "fop=link",
- NULL);
- chlog_barrier_dequeue_all (this, &queue);
- }
- wind:
- STACK_WIND (frame, changelog_link_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->link,
- oldloc, newloc, xdata);
+changelog_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata)
+{
+ size_t xtra_len = 0;
+ changelog_priv_t *priv = NULL;
+ changelog_opt_t *co = NULL;
+ call_stub_t *stub = NULL;
+ struct list_head queue = {
+ 0,
+ };
+ gf_boolean_t barrier_enabled = _gf_false;
+
+ priv = this->private;
+
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
+ CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO(frame, xdata, wind);
+
+ CHANGELOG_INIT_NOCHECK(this, frame->local, NULL, oldloc->gfid, 2);
+
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
+
+ CHANGLOG_FILL_FOP_NUMBER(co, frame->root->op, fop_fn, xtra_len);
+
+ co++;
+ CHANGELOG_FILL_ENTRY(co, newloc->pargfid, newloc->name, entry_fn,
+ entry_free_fn, xtra_len, wind);
+
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 2);
+
+ LOCK(&priv->lock);
+ {
+ if ((barrier_enabled = priv->barrier_enabled)) {
+ stub = fop_link_stub(frame, changelog_link_resume, oldloc, newloc,
+ xdata);
+ if (!stub)
+ __chlog_barrier_disable(this, &queue);
+ else
+ __chlog_barrier_enqueue(this, stub);
+ } else {
+ ((changelog_local_t *)frame->local)->color = priv->current_color;
+ changelog_inc_fop_cnt(this, priv, frame->local);
+ }
+ }
+ UNLOCK(&priv->lock);
+
+ if (barrier_enabled && stub) {
+ gf_msg_debug(this->name, 0, "Enqueued link");
+ goto out;
+ }
+
+ if (barrier_enabled && !stub) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_BARRIER_FOP_FAILED,
+ "fop=link", NULL);
+ chlog_barrier_dequeue_all(this, &queue);
+ }
+wind:
+ STACK_WIND(frame, changelog_link_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
out:
- return 0;
+ return 0;
}
/* mkdir */
int32_t
-changelog_mkdir_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this, int32_t op_ret,
- int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+changelog_mkdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_ENTRY);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (mkdir, frame, op_ret, op_errno,
- inode, buf, preparent, postparent, xdata);
- return 0;
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(mkdir, frame, op_ret, op_errno, inode, buf,
+ preparent, postparent, xdata);
+ return 0;
}
int32_t
-changelog_mkdir_resume (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode,
- mode_t umask, dict_t *xdata)
+changelog_mkdir_resume(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ mode_t mode, mode_t umask, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
+ changelog_priv_t *priv = NULL;
- GF_VALIDATE_OR_GOTO ("changelog", this, out);
- GF_VALIDATE_OR_GOTO ("changelog", this->fops, out);
- GF_VALIDATE_OR_GOTO ("changelog", frame, out);
+ GF_VALIDATE_OR_GOTO("changelog", this, out);
+ GF_VALIDATE_OR_GOTO("changelog", this->fops, out);
+ GF_VALIDATE_OR_GOTO("changelog", frame, out);
- priv = this->private;
+ priv = this->private;
- gf_msg_debug (this->name, 0, "Dequeuing mkdir");
- changelog_color_fop_and_inc_cnt
- (this, priv, frame->local);
- STACK_WIND (frame, changelog_mkdir_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->mkdir,
- loc, mode, umask, xdata);
- return 0;
+ gf_msg_debug(this->name, 0, "Dequeuing mkdir");
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_mkdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
+ return 0;
out:
- return -1;
+ return -1;
}
int32_t
-changelog_mkdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata)
-{
- int ret = -1;
- uuid_t gfid = {0,};
- size_t xtra_len = 0;
- changelog_priv_t *priv = NULL;
- changelog_opt_t *co = NULL;
- call_stub_t *stub = NULL;
- struct list_head queue = {0, };
- gf_boolean_t barrier_enabled = _gf_false;
-
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
-
- ret = dict_get_gfuuid (xdata, "gfid-req", &gfid);
- if (ret) {
- gf_msg_debug (this->name, 0,
- "failed to get gfid from dict");
- goto wind;
- }
-
- CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, gfid, 5);
-
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
-
- CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
- co++;
-
- CHANGELOG_FILL_UINT32 (co, S_IFDIR | mode, number_fn, xtra_len);
- co++;
-
- CHANGELOG_FILL_UINT32 (co, frame->root->uid, number_fn, xtra_len);
- co++;
-
- CHANGELOG_FILL_UINT32 (co, frame->root->gid, number_fn, xtra_len);
- co++;
-
- CHANGELOG_FILL_ENTRY (co, loc->pargfid, loc->name,
- entry_fn, entry_free_fn, xtra_len, wind);
-
- changelog_set_usable_record_and_length (frame->local, xtra_len, 5);
-
- LOCK (&priv->lock);
- {
- if ((barrier_enabled = priv->barrier_enabled)) {
- stub = fop_mkdir_stub (frame, changelog_mkdir_resume,
- loc, mode, umask, xdata);
- if (!stub)
- __chlog_barrier_disable (this, &queue);
- else
- __chlog_barrier_enqueue (this, stub);
- } else {
- ((changelog_local_t *)frame->local)->color
- = priv->current_color;
- changelog_inc_fop_cnt (this, priv, frame->local);
- }
- }
- UNLOCK (&priv->lock);
-
- if (barrier_enabled && stub) {
- gf_msg_debug (this->name, 0, "Enqueued mkdir");
- goto out;
- }
-
- if (barrier_enabled && !stub) {
- gf_smsg (this->name, GF_LOG_ERROR, ENOMEM,
- CHANGELOG_MSG_NO_MEMORY,
- "Failed to barrier FOPs, disabling changelog barrier",
- "fop=mkdir",
- NULL);
- chlog_barrier_dequeue_all (this, &queue);
- }
-
- wind:
- STACK_WIND (frame, changelog_mkdir_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->mkdir,
- loc, mode, umask, xdata);
+changelog_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
+{
+ int ret = -1;
+ uuid_t gfid = {
+ 0,
+ };
+ size_t xtra_len = 0;
+ changelog_priv_t *priv = NULL;
+ changelog_opt_t *co = NULL;
+ call_stub_t *stub = NULL;
+ struct list_head queue = {
+ 0,
+ };
+ gf_boolean_t barrier_enabled = _gf_false;
+
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
+
+ ret = dict_get_gfuuid(xdata, "gfid-req", &gfid);
+ if (ret) {
+ gf_msg_debug(this->name, 0, "failed to get gfid from dict");
+ goto wind;
+ }
+
+ CHANGELOG_INIT_NOCHECK(this, frame->local, NULL, gfid, 5);
+
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
+
+ CHANGLOG_FILL_FOP_NUMBER(co, frame->root->op, fop_fn, xtra_len);
+ co++;
+
+ CHANGELOG_FILL_UINT32(co, S_IFDIR | mode, number_fn, xtra_len);
+ co++;
+
+ CHANGELOG_FILL_UINT32(co, frame->root->uid, number_fn, xtra_len);
+ co++;
+
+ CHANGELOG_FILL_UINT32(co, frame->root->gid, number_fn, xtra_len);
+ co++;
+
+ CHANGELOG_FILL_ENTRY(co, loc->pargfid, loc->name, entry_fn, entry_free_fn,
+ xtra_len, wind);
+
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 5);
+
+ LOCK(&priv->lock);
+ {
+ if ((barrier_enabled = priv->barrier_enabled)) {
+ stub = fop_mkdir_stub(frame, changelog_mkdir_resume, loc, mode,
+ umask, xdata);
+ if (!stub)
+ __chlog_barrier_disable(this, &queue);
+ else
+ __chlog_barrier_enqueue(this, stub);
+ } else {
+ ((changelog_local_t *)frame->local)->color = priv->current_color;
+ changelog_inc_fop_cnt(this, priv, frame->local);
+ }
+ }
+ UNLOCK(&priv->lock);
+
+ if (barrier_enabled && stub) {
+ gf_msg_debug(this->name, 0, "Enqueued mkdir");
+ goto out;
+ }
+
+ if (barrier_enabled && !stub) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM,
+ CHANGELOG_MSG_BARRIER_FOP_FAILED, "fop=mkdir", NULL);
+ chlog_barrier_dequeue_all(this, &queue);
+ }
+
+wind:
+ STACK_WIND(frame, changelog_mkdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
out:
- return 0;
+ return 0;
}
/* symlink */
int32_t
-changelog_symlink_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+changelog_symlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_ENTRY);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (symlink, frame, op_ret, op_errno,
- inode, buf, preparent, postparent, xdata);
- return 0;
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(symlink, frame, op_ret, op_errno, inode, buf,
+ preparent, postparent, xdata);
+ return 0;
}
-
int32_t
-changelog_symlink_resume (call_frame_t *frame, xlator_t *this,
- const char *linkname, loc_t *loc,
- mode_t umask, dict_t *xdata)
+changelog_symlink_resume(call_frame_t *frame, xlator_t *this,
+ const char *linkname, loc_t *loc, mode_t umask,
+ dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
+ changelog_priv_t *priv = NULL;
- GF_VALIDATE_OR_GOTO ("changelog", this, out);
- GF_VALIDATE_OR_GOTO ("changelog", this->fops, out);
- GF_VALIDATE_OR_GOTO ("changelog", frame, out);
+ GF_VALIDATE_OR_GOTO("changelog", this, out);
+ GF_VALIDATE_OR_GOTO("changelog", this->fops, out);
+ GF_VALIDATE_OR_GOTO("changelog", frame, out);
- priv = this->private;
+ priv = this->private;
- gf_msg_debug (this->name, 0, "Dequeuing symlink");
- changelog_color_fop_and_inc_cnt
- (this, priv, frame->local);
- STACK_WIND (frame, changelog_symlink_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->symlink,
- linkname, loc, umask, xdata);
- return 0;
+ gf_msg_debug(this->name, 0, "Dequeuing symlink");
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_symlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->symlink, linkname, loc, umask, xdata);
+ return 0;
out:
- return -1;
+ return -1;
}
int32_t
-changelog_symlink (call_frame_t *frame, xlator_t *this,
- const char *linkname, loc_t *loc,
- mode_t umask, dict_t *xdata)
-{
- int ret = -1;
- size_t xtra_len = 0;
- uuid_t gfid = {0,};
- changelog_priv_t *priv = NULL;
- changelog_opt_t *co = NULL;
- call_stub_t *stub = NULL;
- struct list_head queue = {0, };
- gf_boolean_t barrier_enabled = _gf_false;
-
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
-
- ret = dict_get_gfuuid (xdata, "gfid-req", &gfid);
- if (ret) {
- gf_msg_debug (this->name, 0,
- "failed to get gfid from dict");
- goto wind;
- }
-
- CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, gfid, 2);
-
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
-
- CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
- co++;
-
- CHANGELOG_FILL_ENTRY (co, loc->pargfid, loc->name,
- entry_fn, entry_free_fn, xtra_len, wind);
-
- changelog_set_usable_record_and_length (frame->local, xtra_len, 2);
-
- LOCK (&priv->lock);
- {
- if ((barrier_enabled = priv->barrier_enabled)) {
- stub = fop_symlink_stub (frame,
- changelog_symlink_resume,
- linkname, loc, umask, xdata);
- if (!stub)
- __chlog_barrier_disable (this, &queue);
- else
- __chlog_barrier_enqueue (this, stub);
- } else {
- ((changelog_local_t *)frame->local)->color
- = priv->current_color;
- changelog_inc_fop_cnt (this, priv, frame->local);
- }
- }
- UNLOCK (&priv->lock);
-
- if (barrier_enabled && stub) {
- gf_msg_debug (this->name, 0, "Enqueued symlink");
- goto out;
- }
-
- if (barrier_enabled && !stub) {
- gf_smsg (this->name, GF_LOG_ERROR, ENOMEM,
- CHANGELOG_MSG_NO_MEMORY,
- "Failed to barrier FOPs, disabling changelog barrier",
- "fop=symlink",
- NULL);
- chlog_barrier_dequeue_all (this, &queue);
- }
-
- wind:
- STACK_WIND (frame, changelog_symlink_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->symlink,
- linkname, loc, umask, xdata);
+changelog_symlink(call_frame_t *frame, xlator_t *this, const char *linkname,
+ loc_t *loc, mode_t umask, dict_t *xdata)
+{
+ int ret = -1;
+ size_t xtra_len = 0;
+ uuid_t gfid = {
+ 0,
+ };
+ changelog_priv_t *priv = NULL;
+ changelog_opt_t *co = NULL;
+ call_stub_t *stub = NULL;
+ struct list_head queue = {
+ 0,
+ };
+ gf_boolean_t barrier_enabled = _gf_false;
+
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
+
+ ret = dict_get_gfuuid(xdata, "gfid-req", &gfid);
+ if (ret) {
+ gf_msg_debug(this->name, 0, "failed to get gfid from dict");
+ goto wind;
+ }
+
+ CHANGELOG_INIT_NOCHECK(this, frame->local, NULL, gfid, 2);
+
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
+
+ CHANGLOG_FILL_FOP_NUMBER(co, frame->root->op, fop_fn, xtra_len);
+ co++;
+
+ CHANGELOG_FILL_ENTRY(co, loc->pargfid, loc->name, entry_fn, entry_free_fn,
+ xtra_len, wind);
+
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 2);
+
+ LOCK(&priv->lock);
+ {
+ if ((barrier_enabled = priv->barrier_enabled)) {
+ stub = fop_symlink_stub(frame, changelog_symlink_resume, linkname,
+ loc, umask, xdata);
+ if (!stub)
+ __chlog_barrier_disable(this, &queue);
+ else
+ __chlog_barrier_enqueue(this, stub);
+ } else {
+ ((changelog_local_t *)frame->local)->color = priv->current_color;
+ changelog_inc_fop_cnt(this, priv, frame->local);
+ }
+ }
+ UNLOCK(&priv->lock);
+
+ if (barrier_enabled && stub) {
+ gf_msg_debug(this->name, 0, "Enqueued symlink");
+ goto out;
+ }
+
+ if (barrier_enabled && !stub) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM,
+ CHANGELOG_MSG_BARRIER_FOP_FAILED, "fop=symlink", NULL);
+ chlog_barrier_dequeue_all(this, &queue);
+ }
+
+wind:
+ STACK_WIND(frame, changelog_symlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->symlink, linkname, loc, umask, xdata);
out:
- return 0;
+ return 0;
}
/* mknod */
int32_t
-changelog_mknod_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+changelog_mknod_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_ENTRY);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (mknod, frame, op_ret, op_errno,
- inode, buf, preparent, postparent, xdata);
- return 0;
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(mknod, frame, op_ret, op_errno, inode, buf,
+ preparent, postparent, xdata);
+ return 0;
}
int32_t
-changelog_mknod_resume (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t rdev,
- mode_t umask, dict_t *xdata)
+changelog_mknod_resume(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
+ changelog_priv_t *priv = NULL;
- GF_VALIDATE_OR_GOTO ("changelog", this, out);
- GF_VALIDATE_OR_GOTO ("changelog", this->fops, out);
- GF_VALIDATE_OR_GOTO ("changelog", frame, out);
+ GF_VALIDATE_OR_GOTO("changelog", this, out);
+ GF_VALIDATE_OR_GOTO("changelog", this->fops, out);
+ GF_VALIDATE_OR_GOTO("changelog", frame, out);
- priv = this->private;
+ priv = this->private;
- gf_msg_debug (this->name, 0, "Dequeuing mknod");
- changelog_color_fop_and_inc_cnt
- (this, priv, frame->local);
- STACK_WIND (frame, changelog_mknod_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->mknod,
- loc, mode, rdev, umask, xdata);
- return 0;
+ gf_msg_debug(this->name, 0, "Dequeuing mknod");
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_mknod_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, loc, mode, rdev, umask, xdata);
+ return 0;
out:
- return -1;
+ return -1;
}
int32_t
-changelog_mknod (call_frame_t *frame,
- xlator_t *this, loc_t *loc,
- mode_t mode, dev_t dev, mode_t umask, dict_t *xdata)
-{
- int ret = -1;
- uuid_t gfid = {0,};
- size_t xtra_len = 0;
- changelog_priv_t *priv = NULL;
- changelog_opt_t *co = NULL;
- call_stub_t *stub = NULL;
- struct list_head queue = {0, };
- gf_boolean_t barrier_enabled = _gf_false;
-
- priv = this->private;
-
- /* Check whether changelog active */
- if (!(priv->active))
- goto wind;
-
- /* Check whether rebalance activity */
- if (frame->root->pid == GF_CLIENT_PID_DEFRAG)
- goto wind;
-
- /* If tier-dht linkto is SET, ignore about verifiying :
- * 1. Whether internal fop AND
- * 2. Whether tier rebalance process activity (this will help in
- * recording mknod if tier rebalance process calls this mknod) */
- if (!(dict_get (xdata, "trusted.tier.tier-dht.linkto"))) {
- CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, wind);
- if (frame->root->pid == GF_CLIENT_PID_TIER_DEFRAG)
- goto wind;
- }
-
- ret = dict_get_gfuuid (xdata, "gfid-req", &gfid);
- if (ret) {
- gf_msg_debug (this->name, 0,
- "failed to get gfid from dict");
- goto wind;
- }
-
- CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, gfid, 5);
-
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
-
- CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
- co++;
-
- CHANGELOG_FILL_UINT32 (co, mode, number_fn, xtra_len);
- co++;
-
- CHANGELOG_FILL_UINT32 (co, frame->root->uid, number_fn, xtra_len);
- co++;
-
- CHANGELOG_FILL_UINT32 (co, frame->root->gid, number_fn, xtra_len);
- co++;
-
- CHANGELOG_FILL_ENTRY (co, loc->pargfid, loc->name,
- entry_fn, entry_free_fn, xtra_len, wind);
-
- changelog_set_usable_record_and_length (frame->local, xtra_len, 5);
-
- LOCK (&priv->lock);
- {
- if ((barrier_enabled = priv->barrier_enabled)) {
- stub = fop_mknod_stub (frame, changelog_mknod_resume,
- loc, mode, dev, umask, xdata);
- if (!stub)
- __chlog_barrier_disable (this, &queue);
- else
- __chlog_barrier_enqueue (this, stub);
- } else {
- ((changelog_local_t *)frame->local)->color
- = priv->current_color;
- changelog_inc_fop_cnt (this, priv, frame->local);
- }
- }
- UNLOCK (&priv->lock);
-
- if (barrier_enabled && stub) {
- gf_msg_debug (this->name, 0, "Enqueued mknod");
- goto out;
- }
-
- if (barrier_enabled && !stub) {
- gf_smsg (this->name, GF_LOG_ERROR, ENOMEM,
- CHANGELOG_MSG_NO_MEMORY,
- "Failed to barrier FOPs, disabling changelog barrier",
- "fop=mknod",
- NULL);
- chlog_barrier_dequeue_all (this, &queue);
- }
-
- wind:
- STACK_WIND (frame, changelog_mknod_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->mknod,
- loc, mode, dev, umask, xdata);
+changelog_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t dev, mode_t umask, dict_t *xdata)
+{
+ int ret = -1;
+ uuid_t gfid = {
+ 0,
+ };
+ size_t xtra_len = 0;
+ changelog_priv_t *priv = NULL;
+ changelog_opt_t *co = NULL;
+ call_stub_t *stub = NULL;
+ struct list_head queue = {
+ 0,
+ };
+ gf_boolean_t barrier_enabled = _gf_false;
+
+ priv = this->private;
+
+ /* Check whether changelog active */
+ if (!(priv->active))
+ goto wind;
+
+ /* Check whether rebalance activity */
+ if (frame->root->pid == GF_CLIENT_PID_DEFRAG)
+ goto wind;
+
+ /* If tier-dht linkto is SET, ignore about verifiying :
+ * 1. Whether internal fop AND
+ * 2. Whether tier rebalance process activity (this will help in
+ * recording mknod if tier rebalance process calls this mknod) */
+ if (!(dict_get(xdata, "trusted.tier.tier-dht.linkto"))) {
+ CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO(frame, xdata, wind);
+ if (frame->root->pid == GF_CLIENT_PID_TIER_DEFRAG)
+ goto wind;
+ }
+
+ ret = dict_get_gfuuid(xdata, "gfid-req", &gfid);
+ if (ret) {
+ gf_msg_debug(this->name, 0, "failed to get gfid from dict");
+ goto wind;
+ }
+
+ CHANGELOG_INIT_NOCHECK(this, frame->local, NULL, gfid, 5);
+
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
+
+ CHANGLOG_FILL_FOP_NUMBER(co, frame->root->op, fop_fn, xtra_len);
+ co++;
+
+ CHANGELOG_FILL_UINT32(co, mode, number_fn, xtra_len);
+ co++;
+
+ CHANGELOG_FILL_UINT32(co, frame->root->uid, number_fn, xtra_len);
+ co++;
+
+ CHANGELOG_FILL_UINT32(co, frame->root->gid, number_fn, xtra_len);
+ co++;
+
+ CHANGELOG_FILL_ENTRY(co, loc->pargfid, loc->name, entry_fn, entry_free_fn,
+ xtra_len, wind);
+
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 5);
+
+ LOCK(&priv->lock);
+ {
+ if ((barrier_enabled = priv->barrier_enabled)) {
+ stub = fop_mknod_stub(frame, changelog_mknod_resume, loc, mode, dev,
+ umask, xdata);
+ if (!stub)
+ __chlog_barrier_disable(this, &queue);
+ else
+ __chlog_barrier_enqueue(this, stub);
+ } else {
+ ((changelog_local_t *)frame->local)->color = priv->current_color;
+ changelog_inc_fop_cnt(this, priv, frame->local);
+ }
+ }
+ UNLOCK(&priv->lock);
+
+ if (barrier_enabled && stub) {
+ gf_msg_debug(this->name, 0, "Enqueued mknod");
+ goto out;
+ }
+
+ if (barrier_enabled && !stub) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM,
+ CHANGELOG_MSG_BARRIER_FOP_FAILED, "fop=mknod", NULL);
+ chlog_barrier_dequeue_all(this, &queue);
+ }
+
+wind:
+ STACK_WIND(frame, changelog_mknod_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, loc, mode, dev, umask, xdata);
out:
- return 0;
+ return 0;
}
-/* creat */
+/* create */
int32_t
-changelog_create_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- fd_t *fd, inode_t *inode, struct iatt *buf,
- struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+changelog_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- int32_t ret = 0;
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
- changelog_event_t ev = {0,};
-
- priv = this->private;
- local = frame->local;
-
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
-
- /* fill the event structure.. similar to open() */
- ev.ev_type = CHANGELOG_OP_TYPE_CREATE;
- gf_uuid_copy (ev.u.create.gfid, buf->ia_gfid);
- ev.u.create.flags = fd->flags;
- changelog_dispatch_event (this, priv, &ev);
-
- if (changelog_ev_selected
- (this, &priv->ev_selection, CHANGELOG_OP_TYPE_RELEASE)) {
- ret = fd_ctx_set (fd, this, (uint64_t)(long) 0x1);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- CHANGELOG_MSG_SET_FD_CONTEXT,
- "could not set fd context (for release cbk)");
- }
+ int32_t ret = 0;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
+ changelog_event_t ev = {
+ 0,
+ };
- changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
-
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (create, frame,
- op_ret, op_errno, fd, inode,
- buf, preparent, postparent, xdata);
- return 0;
-}
+ priv = this->private;
+ local = frame->local;
-int32_t
-changelog_create_resume (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, mode_t mode,
- mode_t umask, fd_t *fd, dict_t *xdata)
-{
- changelog_priv_t *priv = NULL;
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- GF_VALIDATE_OR_GOTO ("changelog", this, out);
- GF_VALIDATE_OR_GOTO ("changelog", this->fops, out);
- GF_VALIDATE_OR_GOTO ("changelog", frame, out);
+ /* fill the event structure.. similar to open() */
+ ev.ev_type = CHANGELOG_OP_TYPE_CREATE;
+ gf_uuid_copy(ev.u.create.gfid, buf->ia_gfid);
+ ev.u.create.flags = fd->flags;
+ changelog_dispatch_event(this, priv, &ev);
- priv = this->private;
+ if (changelog_ev_selected(this, &priv->ev_selection,
+ CHANGELOG_OP_TYPE_RELEASE)) {
+ ret = fd_ctx_set(fd, this, (uint64_t)(long)0x1);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_WARNING, 0, CHANGELOG_MSG_SET_FD_CONTEXT,
+ NULL);
+ }
- gf_msg_debug (this->name, 0, "Dequeuing create");
- changelog_color_fop_and_inc_cnt
- (this, priv, frame->local);
- STACK_WIND (frame, changelog_create_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->create,
- loc, flags, mode, umask, fd, xdata);
- return 0;
+ changelog_update(this, priv, local, CHANGELOG_TYPE_ENTRY);
-out:
- return -1;
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(create, frame, op_ret, op_errno, fd, inode, buf,
+ preparent, postparent, xdata);
+ return 0;
}
int32_t
-changelog_create (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, mode_t mode,
- mode_t umask, fd_t *fd, dict_t *xdata)
-{
- int ret = -1;
- uuid_t gfid = {0,};
- changelog_opt_t *co = NULL;
- changelog_priv_t *priv = NULL;
- size_t xtra_len = 0;
- call_stub_t *stub = NULL;
- struct list_head queue = {0, };
- gf_boolean_t barrier_enabled = _gf_false;
-
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
-
- ret = dict_get_gfuuid (xdata, "gfid-req", &gfid);
- if (ret) {
- gf_msg_debug (this->name, 0,
- "failed to get gfid from dict");
- goto wind;
- }
-
- /* init with two extra records */
- CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, gfid, 5);
- if (!frame->local)
- goto wind;
-
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
-
- CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
- co++;
-
- CHANGELOG_FILL_UINT32 (co, mode, number_fn, xtra_len);
- co++;
-
- CHANGELOG_FILL_UINT32 (co, frame->root->uid, number_fn, xtra_len);
- co++;
-
- CHANGELOG_FILL_UINT32 (co, frame->root->gid, number_fn, xtra_len);
- co++;
-
- CHANGELOG_FILL_ENTRY (co, loc->pargfid, loc->name,
- entry_fn, entry_free_fn, xtra_len, wind);
+changelog_create_resume(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ int32_t flags, mode_t mode, mode_t umask, fd_t *fd,
+ dict_t *xdata)
+{
+ changelog_priv_t *priv = NULL;
- changelog_set_usable_record_and_length (frame->local, xtra_len, 5);
+ GF_VALIDATE_OR_GOTO("changelog", this, out);
+ GF_VALIDATE_OR_GOTO("changelog", this->fops, out);
+ GF_VALIDATE_OR_GOTO("changelog", frame, out);
- LOCK (&priv->lock);
- {
- if ((barrier_enabled = priv->barrier_enabled)) {
- stub = fop_create_stub (frame, changelog_create_resume,
- loc, flags, mode, umask, fd,
- xdata);
- if (!stub)
- __chlog_barrier_disable (this, &queue);
- else
- __chlog_barrier_enqueue (this, stub);
- } else {
- ((changelog_local_t *)frame->local)->color
- = priv->current_color;
- changelog_inc_fop_cnt (this, priv, frame->local);
- }
- }
- UNLOCK (&priv->lock);
+ priv = this->private;
- if (barrier_enabled && stub) {
- gf_msg_debug (this->name, 0, "Enqueued create");
- goto out;
- }
+ gf_msg_debug(this->name, 0, "Dequeuing create");
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_create_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, flags, mode, umask, fd,
+ xdata);
+ return 0;
- if (barrier_enabled && !stub) {
- gf_smsg (this->name, GF_LOG_ERROR, ENOMEM,
- CHANGELOG_MSG_NO_MEMORY,
- "Failed to barrier FOPs, disabling changelog barrier",
- "fop=create",
- NULL);
- chlog_barrier_dequeue_all (this, &queue);
- }
+out:
+ return -1;
+}
- wind:
- STACK_WIND (frame, changelog_create_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->create,
- loc, flags, mode, umask, fd, xdata);
+int32_t
+changelog_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+{
+ int ret = -1;
+ uuid_t gfid = {
+ 0,
+ };
+ changelog_opt_t *co = NULL;
+ changelog_priv_t *priv = NULL;
+ size_t xtra_len = 0;
+ call_stub_t *stub = NULL;
+ struct list_head queue = {
+ 0,
+ };
+ gf_boolean_t barrier_enabled = _gf_false;
+
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
+
+ ret = dict_get_gfuuid(xdata, "gfid-req", &gfid);
+ if (ret) {
+ gf_msg_debug(this->name, 0, "failed to get gfid from dict");
+ goto wind;
+ }
+
+ /* init with two extra records */
+ CHANGELOG_INIT_NOCHECK(this, frame->local, NULL, gfid, 5);
+ if (!frame->local)
+ goto wind;
+
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
+
+ CHANGLOG_FILL_FOP_NUMBER(co, frame->root->op, fop_fn, xtra_len);
+ co++;
+
+ CHANGELOG_FILL_UINT32(co, mode, number_fn, xtra_len);
+ co++;
+
+ CHANGELOG_FILL_UINT32(co, frame->root->uid, number_fn, xtra_len);
+ co++;
+
+ CHANGELOG_FILL_UINT32(co, frame->root->gid, number_fn, xtra_len);
+ co++;
+
+ CHANGELOG_FILL_ENTRY(co, loc->pargfid, loc->name, entry_fn, entry_free_fn,
+ xtra_len, wind);
+
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 5);
+
+ LOCK(&priv->lock);
+ {
+ if ((barrier_enabled = priv->barrier_enabled)) {
+ stub = fop_create_stub(frame, changelog_create_resume, loc, flags,
+ mode, umask, fd, xdata);
+ if (!stub)
+ __chlog_barrier_disable(this, &queue);
+ else
+ __chlog_barrier_enqueue(this, stub);
+ } else {
+ ((changelog_local_t *)frame->local)->color = priv->current_color;
+ changelog_inc_fop_cnt(this, priv, frame->local);
+ }
+ }
+ UNLOCK(&priv->lock);
+
+ if (barrier_enabled && stub) {
+ gf_msg_debug(this->name, 0, "Enqueued create");
+ goto out;
+ }
+
+ if (barrier_enabled && !stub) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM,
+ CHANGELOG_MSG_BARRIER_FOP_FAILED, "fop=create", NULL);
+ chlog_barrier_dequeue_all(this, &queue);
+ }
+
+wind:
+ STACK_WIND(frame, changelog_create_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, flags, mode, umask, fd,
+ xdata);
out:
- return 0;
+ return 0;
}
/* }}} */
-
/* Metadata modification fops - TYPE II */
/* {{{ */
@@ -1158,268 +1104,253 @@ out:
/* {f}setattr */
int32_t
-changelog_fsetattr_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this, int32_t op_ret,
- int32_t op_errno, struct iatt *preop_stbuf,
- struct iatt *postop_stbuf, dict_t *xdata)
+changelog_fsetattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *preop_stbuf, struct iatt *postop_stbuf,
+ dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
-
- priv = this->private;
- local = frame->local;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+ priv = this->private;
+ local = frame->local;
- changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA);
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (fsetattr, frame, op_ret, op_errno,
- preop_stbuf, postop_stbuf, xdata);
-
- return 0;
+ changelog_update(this, priv, local, CHANGELOG_TYPE_METADATA);
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(fsetattr, frame, op_ret, op_errno, preop_stbuf,
+ postop_stbuf, xdata);
+ return 0;
}
int32_t
-changelog_fsetattr (call_frame_t *frame,
- xlator_t *this, fd_t *fd,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+changelog_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_opt_t *co = NULL;
- size_t xtra_len = 0;
-
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+ changelog_priv_t *priv = NULL;
+ changelog_opt_t *co = NULL;
+ size_t xtra_len = 0;
- CHANGELOG_OP_BOUNDARY_CHECK (frame, wind);
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
- CHANGELOG_INIT (this, frame->local,
- fd->inode, fd->inode->gfid, 1);
- if (!frame->local)
- goto wind;
+ CHANGELOG_OP_BOUNDARY_CHECK(frame, wind);
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
+ CHANGELOG_INIT(this, frame->local, fd->inode, fd->inode->gfid, 1);
+ if (!frame->local)
+ goto wind;
- CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
- changelog_set_usable_record_and_length (frame->local, xtra_len, 1);
-
- wind:
- changelog_color_fop_and_inc_cnt (this, priv, frame->local);
- STACK_WIND (frame, changelog_fsetattr_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->fsetattr,
- fd, stbuf, valid, xdata);
- return 0;
+ CHANGLOG_FILL_FOP_NUMBER(co, frame->root->op, fop_fn, xtra_len);
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 1);
+wind:
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_fsetattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetattr, fd, stbuf, valid, xdata);
+ return 0;
}
int32_t
-changelog_setattr_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this, int32_t op_ret,
- int32_t op_errno, struct iatt *preop_stbuf,
- struct iatt *postop_stbuf, dict_t *xdata)
+changelog_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *preop_stbuf, struct iatt *postop_stbuf,
+ dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_METADATA);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (setattr, frame, op_ret, op_errno,
- preop_stbuf, postop_stbuf, xdata);
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(setattr, frame, op_ret, op_errno, preop_stbuf,
+ postop_stbuf, xdata);
- return 0;
+ return 0;
}
int32_t
-changelog_setattr (call_frame_t *frame,
- xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+changelog_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_opt_t *co = NULL;
- size_t xtra_len = 0;
- uuid_t shard_root_gfid = {0,};
+ changelog_priv_t *priv = NULL;
+ changelog_opt_t *co = NULL;
+ size_t xtra_len = 0;
+ uuid_t shard_root_gfid = {
+ 0,
+ };
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
- CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, wind);
+ CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO(frame, xdata, wind);
- /* Do not record META on .shard */
- gf_uuid_parse (SHARD_ROOT_GFID, shard_root_gfid);
- if (gf_uuid_compare (loc->gfid, shard_root_gfid) == 0) {
- goto wind;
- }
+ /* Do not record META on .shard */
+ gf_uuid_parse(SHARD_ROOT_GFID, shard_root_gfid);
+ if (gf_uuid_compare(loc->gfid, shard_root_gfid) == 0) {
+ goto wind;
+ }
- CHANGELOG_OP_BOUNDARY_CHECK (frame, wind);
+ CHANGELOG_OP_BOUNDARY_CHECK(frame, wind);
- CHANGELOG_INIT (this, frame->local,
- loc->inode, loc->inode->gfid, 1);
- if (!frame->local)
- goto wind;
+ CHANGELOG_INIT(this, frame->local, loc->inode, loc->inode->gfid, 1);
+ if (!frame->local)
+ goto wind;
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
- CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+ CHANGLOG_FILL_FOP_NUMBER(co, frame->root->op, fop_fn, xtra_len);
- changelog_set_usable_record_and_length (frame->local, xtra_len, 1);
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 1);
- wind:
- changelog_color_fop_and_inc_cnt (this, priv, frame->local);
- STACK_WIND (frame, changelog_setattr_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->setattr,
- loc, stbuf, valid, xdata);
- return 0;
+wind:
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_setattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid, xdata);
+ return 0;
}
/* {f}removexattr */
int32_t
-changelog_fremovexattr_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+changelog_fremovexattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA_XATTR);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_METADATA_XATTR);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, xdata);
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(fremovexattr, frame, op_ret, op_errno, xdata);
- return 0;
+ return 0;
}
int32_t
-changelog_fremovexattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, const char *name, dict_t *xdata)
+changelog_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_opt_t *co = NULL;
- size_t xtra_len = 0;
+ changelog_priv_t *priv = NULL;
+ changelog_opt_t *co = NULL;
+ size_t xtra_len = 0;
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
- CHANGELOG_OP_BOUNDARY_CHECK (frame, wind);
+ CHANGELOG_OP_BOUNDARY_CHECK(frame, wind);
- CHANGELOG_INIT (this, frame->local,
- fd->inode, fd->inode->gfid, 1);
+ CHANGELOG_INIT(this, frame->local, fd->inode, fd->inode->gfid, 1);
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
- CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+ CHANGLOG_FILL_FOP_NUMBER(co, frame->root->op, fop_fn, xtra_len);
- changelog_set_usable_record_and_length (frame->local, xtra_len, 1);
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 1);
- wind:
- changelog_color_fop_and_inc_cnt (this, priv, frame->local);
- STACK_WIND (frame, changelog_fremovexattr_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->fremovexattr,
- fd, name, xdata);
- return 0;
+wind:
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_fremovexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fremovexattr, fd, name, xdata);
+ return 0;
}
int32_t
-changelog_removexattr_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+changelog_removexattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA_XATTR);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_METADATA_XATTR);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (removexattr, frame, op_ret, op_errno, xdata);
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(removexattr, frame, op_ret, op_errno, xdata);
- return 0;
+ return 0;
}
int32_t
-changelog_removexattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
+changelog_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_opt_t *co = NULL;
- size_t xtra_len = 0;
+ changelog_priv_t *priv = NULL;
+ changelog_opt_t *co = NULL;
+ size_t xtra_len = 0;
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
- CHANGELOG_OP_BOUNDARY_CHECK (frame, wind);
+ CHANGELOG_OP_BOUNDARY_CHECK(frame, wind);
- CHANGELOG_INIT (this, frame->local,
- loc->inode, loc->inode->gfid, 1);
+ CHANGELOG_INIT(this, frame->local, loc->inode, loc->inode->gfid, 1);
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
- CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+ CHANGLOG_FILL_FOP_NUMBER(co, frame->root->op, fop_fn, xtra_len);
- changelog_set_usable_record_and_length (frame->local, xtra_len, 1);
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 1);
- wind:
- changelog_color_fop_and_inc_cnt (this, priv, frame->local);
- STACK_WIND (frame, changelog_removexattr_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->removexattr,
- loc, name, xdata);
- return 0;
+wind:
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_removexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
+ return 0;
}
/* {f}setxattr */
int32_t
-changelog_setxattr_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+changelog_setxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA_XATTR);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_METADATA_XATTR);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(setxattr, frame, op_ret, op_errno, xdata);
- return 0;
+ return 0;
}
/* changelog_handle_virtual_xattr:
@@ -1432,276 +1363,255 @@ changelog_setxattr_cbk (call_frame_t *frame,
* any other value: ENOTSUP is returned.
*/
static void
-changelog_handle_virtual_xattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *dict)
+changelog_handle_virtual_xattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ dict_t *dict)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
- int32_t value = 0;
- int ret = 0;
- int dict_ret = 0;
- gf_boolean_t valid = _gf_false;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
+ int32_t value = 0;
+ int ret = 0;
+ int dict_ret = 0;
+ gf_boolean_t valid = _gf_false;
- priv = this->private;
- GF_ASSERT (priv);
+ priv = this->private;
+ GF_ASSERT(priv);
- dict_ret = dict_get_int32 (dict, GF_XATTR_TRIGGER_SYNC, &value);
+ dict_ret = dict_get_int32(dict, GF_XATTR_TRIGGER_SYNC, &value);
- if ((dict_ret == 0 && value == 1) && ((loc->inode->ia_type == IA_IFDIR)
- || (loc->inode->ia_type == IA_IFREG)))
- valid = _gf_true;
+ if ((dict_ret == 0 && value == 1) && ((loc->inode->ia_type == IA_IFDIR) ||
+ (loc->inode->ia_type == IA_IFREG)))
+ valid = _gf_true;
- if (valid) {
- ret = changelog_fill_entry_buf (frame, this, loc, &local);
- if (ret) {
- gf_smsg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_MSG_ENTRY_BUF_INFO,
- "Entry cannot be"
- " captured for gfid, Capturing DATA"
- " entry.",
- "gfid=%s", uuid_utoa (loc->inode->gfid),
- NULL);
- goto unwind;
- }
- changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
-
- unwind:
- /* Capture DATA only if it's a file. */
- if (loc->inode->ia_type != IA_IFDIR)
- changelog_update (this, priv, frame->local,
- CHANGELOG_TYPE_DATA);
- /* Assign local to prev_entry, so unwind will take
- * care of cleanup. */
- ((changelog_local_t *)(frame->local))->prev_entry = local;
- CHANGELOG_STACK_UNWIND (setxattr, frame, 0, 0, NULL);
- return;
- } else {
- CHANGELOG_STACK_UNWIND (setxattr, frame, -1, ENOTSUP, NULL);
- return;
- }
+ if (valid) {
+ ret = changelog_fill_entry_buf(frame, this, loc, &local);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_INFO, 0, CHANGELOG_MSG_ENTRY_BUF_INFO,
+ "gfid=%s", uuid_utoa(loc->inode->gfid), NULL);
+ goto unwind;
+ }
+ changelog_update(this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+ /* Capture DATA only if it's a file. */
+ if (loc->inode->ia_type != IA_IFDIR)
+ changelog_update(this, priv, frame->local, CHANGELOG_TYPE_DATA);
+ /* Assign local to prev_entry, so unwind will take
+ * care of cleanup. */
+ ((changelog_local_t *)(frame->local))->prev_entry = local;
+ CHANGELOG_STACK_UNWIND(setxattr, frame, 0, 0, NULL);
+ return;
+ } else {
+ CHANGELOG_STACK_UNWIND(setxattr, frame, -1, ENOTSUP, NULL);
+ return;
+ }
}
int32_t
-changelog_setxattr (call_frame_t *frame,
- xlator_t *this, loc_t *loc,
- dict_t *dict, int32_t flags, dict_t *xdata)
+changelog_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ dict_t *dict, int32_t flags, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_opt_t *co = NULL;
- size_t xtra_len = 0;
+ changelog_priv_t *priv = NULL;
+ changelog_opt_t *co = NULL;
+ size_t xtra_len = 0;
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
- CHANGELOG_OP_BOUNDARY_CHECK (frame, wind);
+ CHANGELOG_OP_BOUNDARY_CHECK(frame, wind);
- CHANGELOG_INIT (this, frame->local,
- loc->inode, loc->inode->gfid, 1);
+ CHANGELOG_INIT(this, frame->local, loc->inode, loc->inode->gfid, 1);
- /* On setting this virtual xattr on a file, an explicit data
- * sync is triggered from geo-rep as CREATE|DATA entry is
- * recorded in changelog based on xattr value.
- */
- if (dict_get (dict, GF_XATTR_TRIGGER_SYNC)) {
- changelog_handle_virtual_xattr (frame, this, loc, dict);
- return 0;
- }
+ /* On setting this virtual xattr on a file, an explicit data
+ * sync is triggered from geo-rep as CREATE|DATA entry is
+ * recorded in changelog based on xattr value.
+ */
+ if (dict_get(dict, GF_XATTR_TRIGGER_SYNC)) {
+ changelog_handle_virtual_xattr(frame, this, loc, dict);
+ return 0;
+ }
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
- CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+ CHANGLOG_FILL_FOP_NUMBER(co, frame->root->op, fop_fn, xtra_len);
- changelog_set_usable_record_and_length (frame->local, xtra_len, 1);
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 1);
- wind:
- changelog_color_fop_and_inc_cnt (this, priv, frame->local);
- STACK_WIND (frame, changelog_setxattr_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->setxattr,
- loc, dict, flags, xdata);
- return 0;
+wind:
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_setxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, loc, dict, flags, xdata);
+ return 0;
}
int32_t
-changelog_fsetxattr_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
+changelog_fsetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA_XATTR);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_METADATA_XATTR);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, xdata);
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(fsetxattr, frame, op_ret, op_errno, xdata);
- return 0;
+ return 0;
}
int32_t
-changelog_fsetxattr (call_frame_t *frame,
- xlator_t *this, fd_t *fd, dict_t *dict,
- int32_t flags, dict_t *xdata)
+changelog_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_opt_t *co = NULL;
- size_t xtra_len = 0;
+ changelog_priv_t *priv = NULL;
+ changelog_opt_t *co = NULL;
+ size_t xtra_len = 0;
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
- CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, wind);
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
+ CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO(frame, xdata, wind);
- CHANGELOG_OP_BOUNDARY_CHECK (frame, wind);
+ CHANGELOG_OP_BOUNDARY_CHECK(frame, wind);
- CHANGELOG_INIT (this, frame->local,
- fd->inode, fd->inode->gfid, 1);
+ CHANGELOG_INIT(this, frame->local, fd->inode, fd->inode->gfid, 1);
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
- CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+ CHANGLOG_FILL_FOP_NUMBER(co, frame->root->op, fop_fn, xtra_len);
- changelog_set_usable_record_and_length (frame->local, xtra_len, 1);
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 1);
- wind:
- changelog_color_fop_and_inc_cnt (this, priv, frame->local);
- STACK_WIND (frame, changelog_fsetxattr_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->fsetxattr,
- fd, dict, flags, xdata);
- return 0;
+wind:
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_fsetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+ return 0;
}
int32_t
-changelog_xattrop_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xattr, dict_t *xdata)
+changelog_xattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_METADATA);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (xattrop, frame, op_ret, op_errno, xattr, xdata);
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(xattrop, frame, op_ret, op_errno, xattr, xdata);
- return 0;
+ return 0;
}
int32_t
-changelog_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
+changelog_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_opt_t *co = NULL;
- size_t xtra_len = 0;
- int ret = 0;
- void *size_attr = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_opt_t *co = NULL;
+ size_t xtra_len = 0;
+ int ret = 0;
+ void *size_attr = NULL;
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
- ret = dict_get_ptr (xattr, GF_XATTR_SHARD_FILE_SIZE, &size_attr);
- if (ret)
- goto wind;
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
+ ret = dict_get_ptr(xattr, GF_XATTR_SHARD_FILE_SIZE, &size_attr);
+ if (ret)
+ goto wind;
- CHANGELOG_OP_BOUNDARY_CHECK (frame, wind);
+ CHANGELOG_OP_BOUNDARY_CHECK(frame, wind);
- CHANGELOG_INIT (this, frame->local,
- loc->inode, loc->inode->gfid, 1);
+ CHANGELOG_INIT(this, frame->local, loc->inode, loc->inode->gfid, 1);
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
- CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+ CHANGLOG_FILL_FOP_NUMBER(co, frame->root->op, fop_fn, xtra_len);
- changelog_set_usable_record_and_length (frame->local, xtra_len, 1);
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 1);
- wind:
- changelog_color_fop_and_inc_cnt (this, priv, frame->local);
- STACK_WIND (frame, changelog_xattrop_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->xattrop,
- loc, optype, xattr, xdata);
- return 0;
+wind:
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_xattrop_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->xattrop, loc, optype, xattr, xdata);
+ return 0;
}
int32_t
-changelog_fxattrop_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xattr, dict_t *xdata)
+changelog_fxattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA_XATTR);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_METADATA_XATTR);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (fxattrop, frame,
- op_ret, op_errno, xattr, xdata);
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(fxattrop, frame, op_ret, op_errno, xattr, xdata);
- return 0;
+ return 0;
}
int32_t
-changelog_fxattrop (call_frame_t *frame,
- xlator_t *this, fd_t *fd, gf_xattrop_flags_t optype,
- dict_t *xattr, dict_t *xdata)
-{
- changelog_priv_t *priv = NULL;
- changelog_opt_t *co = NULL;
- size_t xtra_len = 0;
- void *size_attr = NULL;
- int ret = 0;
-
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
- ret = dict_get_ptr (xattr, GF_XATTR_SHARD_FILE_SIZE, &size_attr);
- if (ret)
- goto wind;
+changelog_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
+{
+ changelog_priv_t *priv = NULL;
+ changelog_opt_t *co = NULL;
+ size_t xtra_len = 0;
+ void *size_attr = NULL;
+ int ret = 0;
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
+ ret = dict_get_ptr(xattr, GF_XATTR_SHARD_FILE_SIZE, &size_attr);
+ if (ret)
+ goto wind;
- CHANGELOG_OP_BOUNDARY_CHECK (frame, wind);
+ CHANGELOG_OP_BOUNDARY_CHECK(frame, wind);
- CHANGELOG_INIT (this, frame->local,
- fd->inode, fd->inode->gfid, 1);
+ CHANGELOG_INIT(this, frame->local, fd->inode, fd->inode->gfid, 1);
- co = changelog_get_usable_buffer (frame->local);
- if (!co)
- goto wind;
+ co = changelog_get_usable_buffer(frame->local);
+ if (!co)
+ goto wind;
- CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+ CHANGLOG_FILL_FOP_NUMBER(co, frame->root->op, fop_fn, xtra_len);
- changelog_set_usable_record_and_length (frame->local, xtra_len, 1);
+ changelog_set_usable_record_and_length(frame->local, xtra_len, 1);
- wind:
- changelog_color_fop_and_inc_cnt (this, priv, frame->local);
- STACK_WIND (frame, changelog_fxattrop_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->fxattrop,
- fd, optype, xattr, xdata);
- return 0;
+wind:
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_fxattrop_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fxattrop, fd, optype, xattr, xdata);
+ return 0;
}
/* }}} */
-
/* Data modification fops - TYPE I */
/* {{{ */
@@ -1709,164 +1619,151 @@ changelog_fxattrop (call_frame_t *frame,
/* {f}truncate() */
int32_t
-changelog_truncate_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this, int32_t op_ret,
- int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+changelog_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_DATA);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_DATA);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (truncate, frame,
- op_ret, op_errno, prebuf, postbuf, xdata);
- return 0;
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(truncate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
int32_t
-changelog_truncate (call_frame_t *frame,
- xlator_t *this, loc_t *loc, off_t offset, dict_t *xdata)
+changelog_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ off_t offset, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
+ changelog_priv_t *priv = NULL;
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
- CHANGELOG_INIT (this, frame->local,
- loc->inode, loc->inode->gfid, 0);
- LOCK(&priv->c_snap_lock);
- {
- if (priv->c_snap_fd != -1 &&
- priv->barrier_enabled == _gf_true) {
- changelog_snap_handle_ascii_change (this,
- &( ((changelog_local_t *)(frame->local))->cld));
- }
+ CHANGELOG_INIT(this, frame->local, loc->inode, loc->inode->gfid, 0);
+ LOCK(&priv->c_snap_lock);
+ {
+ if (priv->c_snap_fd != -1 && priv->barrier_enabled == _gf_true) {
+ changelog_snap_handle_ascii_change(
+ this, &(((changelog_local_t *)(frame->local))->cld));
}
- UNLOCK(&priv->c_snap_lock);
-
+ }
+ UNLOCK(&priv->c_snap_lock);
- wind:
- changelog_color_fop_and_inc_cnt (this, priv, frame->local);
- STACK_WIND (frame, changelog_truncate_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->truncate,
- loc, offset, xdata);
- return 0;
+wind:
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_truncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
+ return 0;
}
int32_t
-changelog_ftruncate_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this, int32_t op_ret,
- int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+changelog_ftruncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_DATA);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_DATA);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (ftruncate, frame,
- op_ret, op_errno, prebuf, postbuf, xdata);
- return 0;
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(ftruncate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
int32_t
-changelog_ftruncate (call_frame_t *frame,
- xlator_t *this, fd_t *fd, off_t offset, dict_t *xdata)
+changelog_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
+ changelog_priv_t *priv = NULL;
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
- CHANGELOG_INIT (this, frame->local,
- fd->inode, fd->inode->gfid, 0);
- LOCK(&priv->c_snap_lock);
- {
- if (priv->c_snap_fd != -1 &&
- priv->barrier_enabled == _gf_true) {
- changelog_snap_handle_ascii_change (this,
- &( ((changelog_local_t *)(frame->local))->cld));
- }
+ CHANGELOG_INIT(this, frame->local, fd->inode, fd->inode->gfid, 0);
+ LOCK(&priv->c_snap_lock);
+ {
+ if (priv->c_snap_fd != -1 && priv->barrier_enabled == _gf_true) {
+ changelog_snap_handle_ascii_change(
+ this, &(((changelog_local_t *)(frame->local))->cld));
}
- UNLOCK(&priv->c_snap_lock);
+ }
+ UNLOCK(&priv->c_snap_lock);
- wind:
- changelog_color_fop_and_inc_cnt (this, priv, frame->local);
- STACK_WIND (frame, changelog_ftruncate_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->ftruncate,
- fd, offset, xdata);
- return 0;
+wind:
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_ftruncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+ return 0;
}
/* writev() */
int32_t
-changelog_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf,
- dict_t *xdata)
+changelog_writev_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
- changelog_local_t *local = NULL;
+ changelog_priv_t *priv = NULL;
+ changelog_local_t *local = NULL;
- priv = this->private;
- local = frame->local;
+ priv = this->private;
+ local = frame->local;
- CHANGELOG_COND_GOTO (priv, ((op_ret <= 0) || !local), unwind);
+ CHANGELOG_COND_GOTO(priv, ((op_ret <= 0) || !local), unwind);
- changelog_update (this, priv, local, CHANGELOG_TYPE_DATA);
+ changelog_update(this, priv, local, CHANGELOG_TYPE_DATA);
- unwind:
- changelog_dec_fop_cnt (this, priv, local);
- CHANGELOG_STACK_UNWIND (writev, frame,
- op_ret, op_errno, prebuf, postbuf, xdata);
- return 0;
+unwind:
+ changelog_dec_fop_cnt(this, priv, local);
+ CHANGELOG_STACK_UNWIND(writev, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
int32_t
-changelog_writev (call_frame_t *frame,
- xlator_t *this, fd_t *fd, struct iovec *vector,
- int32_t count, off_t offset, uint32_t flags,
- struct iobref *iobref, dict_t *xdata)
+changelog_writev(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int32_t count, off_t offset,
+ uint32_t flags, struct iobref *iobref, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
+ changelog_priv_t *priv = NULL;
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
- CHANGELOG_INIT (this, frame->local,
- fd->inode, fd->inode->gfid, 0);
- LOCK(&priv->c_snap_lock);
- {
- if (priv->c_snap_fd != -1 &&
- priv->barrier_enabled == _gf_true) {
- changelog_snap_handle_ascii_change (this,
- &( ((changelog_local_t *)(frame->local))->cld));
- }
+ CHANGELOG_INIT(this, frame->local, fd->inode, fd->inode->gfid, 0);
+ LOCK(&priv->c_snap_lock);
+ {
+ if (priv->c_snap_fd != -1 && priv->barrier_enabled == _gf_true) {
+ changelog_snap_handle_ascii_change(
+ this, &(((changelog_local_t *)(frame->local))->cld));
}
- UNLOCK(&priv->c_snap_lock);
+ }
+ UNLOCK(&priv->c_snap_lock);
- wind:
- changelog_color_fop_and_inc_cnt (this, priv, frame->local);
- STACK_WIND (frame, changelog_writev_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->writev, fd, vector,
- count, offset, flags, iobref, xdata);
- return 0;
+wind:
+ changelog_color_fop_and_inc_cnt(this, priv, frame->local);
+ STACK_WIND(frame, changelog_writev_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector, count, offset,
+ flags, iobref, xdata);
+ return 0;
}
/* }}} */
@@ -1875,81 +1772,79 @@ changelog_writev (call_frame_t *frame,
/* {{{ */
-
-
int
-changelog_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
-{
- int ret = 0;
- changelog_priv_t *priv = NULL;
- changelog_event_t ev = {0,};
- gf_boolean_t logopen = _gf_false;
-
- priv = this->private;
- if (frame->local) {
- frame->local = NULL;
- logopen = _gf_true;
- }
-
- CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !logopen), unwind);
-
- /* fill the event structure */
- ev.ev_type = CHANGELOG_OP_TYPE_OPEN;
- gf_uuid_copy (ev.u.open.gfid, fd->inode->gfid);
- ev.u.open.flags = fd->flags;
- changelog_dispatch_event (this, priv, &ev);
-
- if (changelog_ev_selected
- (this, &priv->ev_selection, CHANGELOG_OP_TYPE_RELEASE)) {
- ret = fd_ctx_set (fd, this, (uint64_t)(long) 0x1);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- CHANGELOG_MSG_SET_FD_CONTEXT,
- "could not set fd context (for release cbk)");
- }
+changelog_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
+{
+ int ret = 0;
+ changelog_priv_t *priv = NULL;
+ changelog_event_t ev = {
+ 0,
+ };
+ gf_boolean_t logopen = _gf_false;
+
+ priv = this->private;
+ if (frame->local) {
+ frame->local = NULL;
+ logopen = _gf_true;
+ }
+
+ CHANGELOG_COND_GOTO(priv, ((op_ret < 0) || !logopen), unwind);
+
+ /* fill the event structure */
+ ev.ev_type = CHANGELOG_OP_TYPE_OPEN;
+ gf_uuid_copy(ev.u.open.gfid, fd->inode->gfid);
+ ev.u.open.flags = fd->flags;
+ changelog_dispatch_event(this, priv, &ev);
+
+ if (changelog_ev_selected(this, &priv->ev_selection,
+ CHANGELOG_OP_TYPE_RELEASE)) {
+ ret = fd_ctx_set(fd, this, (uint64_t)(long)0x1);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_WARNING, 0, CHANGELOG_MSG_SET_FD_CONTEXT,
+ NULL);
+ }
- unwind:
- CHANGELOG_STACK_UNWIND (open, frame, op_ret, op_errno, fd, xdata);
- return 0;
+unwind:
+ CHANGELOG_STACK_UNWIND(open, frame, op_ret, op_errno, fd, xdata);
+ return 0;
}
int
-changelog_open (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int flags, fd_t *fd, dict_t *xdata)
+changelog_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ fd_t *fd, dict_t *xdata)
{
- changelog_priv_t *priv = NULL;
+ changelog_priv_t *priv = NULL;
- priv = this->private;
- CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+ priv = this->private;
+ CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, wind);
- frame->local = (void *)0x1; /* do not dereference in ->cbk */
+ frame->local = (void *)0x1; /* do not dereference in ->cbk */
- wind:
- STACK_WIND (frame, changelog_open_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->open, loc, flags, fd, xdata);
- return 0;
+wind:
+ STACK_WIND(frame, changelog_open_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
+ return 0;
}
/* }}} */
/* {{{ */
-
/* }}} */
int32_t
-_changelog_generic_dispatcher (dict_t *dict,
- char *key, data_t *value, void *data)
+_changelog_generic_dispatcher(dict_t *dict, char *key, data_t *value,
+ void *data)
{
- xlator_t *this = NULL;
- changelog_priv_t *priv = NULL;
+ xlator_t *this = NULL;
+ changelog_priv_t *priv = NULL;
- this = data;
- priv = this->private;
+ this = data;
+ priv = this->private;
- changelog_dispatch_event (this, priv, (changelog_event_t *)value->data);
- return 0;
+ changelog_dispatch_event(this, priv, (changelog_event_t *)value->data);
+ return 0;
}
/**
@@ -1958,46 +1853,45 @@ _changelog_generic_dispatcher (dict_t *dict,
* traverses the dictionary).
*/
int32_t
-changelog_ipc (call_frame_t *frame, xlator_t *this, int32_t op, dict_t *xdata)
+changelog_ipc(call_frame_t *frame, xlator_t *this, int32_t op, dict_t *xdata)
{
- if (op != GF_IPC_TARGET_CHANGELOG)
- goto wind;
+ if (op != GF_IPC_TARGET_CHANGELOG)
+ goto wind;
- /* it's for us, do the job */
- if (xdata)
- (void) dict_foreach (xdata,
- _changelog_generic_dispatcher, this);
+ /* it's for us, do the job */
+ if (xdata)
+ (void)dict_foreach(xdata, _changelog_generic_dispatcher, this);
- STACK_UNWIND_STRICT (ipc, frame, 0, 0, NULL);
- return 0;
+ STACK_UNWIND_STRICT(ipc, frame, 0, 0, NULL);
+ return 0;
- wind:
- STACK_WIND (frame, default_ipc_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->ipc, op, xdata);
- return 0;
+wind:
+ STACK_WIND(frame, default_ipc_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ipc, op, xdata);
+ return 0;
}
-
/* {{{ */
int32_t
-changelog_release (xlator_t *this, fd_t *fd)
+changelog_release(xlator_t *this, fd_t *fd)
{
- changelog_event_t ev = {0,};
- changelog_priv_t *priv = NULL;
+ changelog_event_t ev = {
+ 0,
+ };
+ changelog_priv_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- ev.ev_type = CHANGELOG_OP_TYPE_RELEASE;
- gf_uuid_copy (ev.u.release.gfid, fd->inode->gfid);
- changelog_dispatch_event (this, priv, &ev);
+ ev.ev_type = CHANGELOG_OP_TYPE_RELEASE;
+ gf_uuid_copy(ev.u.release.gfid, fd->inode->gfid);
+ changelog_dispatch_event(this, priv, &ev);
- (void) fd_ctx_del (fd, this, NULL);
+ (void)fd_ctx_del(fd, this, NULL);
- return 0;
+ return 0;
}
-
/* }}} */
/**
@@ -2012,1034 +1906,1084 @@ changelog_release (xlator_t *this, fd_t *fd)
* needed if there are more operation modes in the future.
*/
static void
-changelog_assign_opmode (changelog_priv_t *priv, char *mode)
+changelog_assign_opmode(changelog_priv_t *priv, char *mode)
{
- if ( strncmp (mode, "realtime", 8) == 0 ) {
- priv->op_mode = CHANGELOG_MODE_RT;
- }
+ if (strncmp(mode, "realtime", 8) == 0) {
+ priv->op_mode = CHANGELOG_MODE_RT;
+ }
}
static void
-changelog_assign_encoding (changelog_priv_t *priv, char *enc)
+changelog_assign_encoding(changelog_priv_t *priv, char *enc)
{
- if ( strncmp (enc, "binary", 6) == 0 ) {
- priv->encode_mode = CHANGELOG_ENCODE_BINARY;
- } else if ( strncmp (enc, "ascii", 5) == 0 ) {
- priv->encode_mode = CHANGELOG_ENCODE_ASCII;
- }
+ if (strncmp(enc, "binary", 6) == 0) {
+ priv->encode_mode = CHANGELOG_ENCODE_BINARY;
+ } else if (strncmp(enc, "ascii", 5) == 0) {
+ priv->encode_mode = CHANGELOG_ENCODE_ASCII;
+ }
}
static void
changelog_assign_barrier_timeout(changelog_priv_t *priv, uint32_t timeout)
{
- LOCK (&priv->lock);
- {
- priv->timeout.tv_sec = timeout;
- }
- UNLOCK (&priv->lock);
+ LOCK(&priv->lock);
+ {
+ priv->timeout.tv_sec = timeout;
+ }
+ UNLOCK(&priv->lock);
}
/* cleanup any helper threads that are running */
static void
-changelog_cleanup_helper_threads (xlator_t *this, changelog_priv_t *priv)
+changelog_cleanup_helper_threads(xlator_t *this, changelog_priv_t *priv)
{
- if (priv->cr.rollover_th) {
- (void) changelog_thread_cleanup (this, priv->cr.rollover_th);
- priv->cr.rollover_th = 0;
- }
+ if (priv->cr.rollover_th) {
+ (void)changelog_thread_cleanup(this, priv->cr.rollover_th);
+ priv->cr.rollover_th = 0;
+ }
- if (priv->cf.fsync_th) {
- (void) changelog_thread_cleanup (this, priv->cf.fsync_th);
- priv->cf.fsync_th = 0;
- }
+ if (priv->cf.fsync_th) {
+ (void)changelog_thread_cleanup(this, priv->cf.fsync_th);
+ priv->cf.fsync_th = 0;
+ }
}
/* spawn helper thread; cleaning up in case of errors */
static int
-changelog_spawn_helper_threads (xlator_t *this, changelog_priv_t *priv)
-{
- int ret = 0;
-
- /* Geo-Rep snapshot dependency:
- *
- * To implement explicit rollover of changlog journal on barrier
- * notification, a pipe is created to communicate between
- * 'changelog_rollover' thread and changelog main thread. The select
- * call used to wait till roll-over time in changelog_rollover thread
- * is modified to wait on read end of the pipe. When barrier
- * notification comes (i.e, in 'reconfigure'), select in
- * changelog_rollover thread is woken up explicitly by writing into
- * the write end of the pipe in 'reconfigure'.
- */
-
- priv->cr.notify = _gf_false;
- priv->cr.this = this;
- ret = gf_thread_create (&priv->cr.rollover_th,
- NULL, changelog_rollover, priv, "clogro");
- if (ret)
- goto out;
-
- if (priv->fsync_interval) {
- priv->cf.this = this;
- ret = gf_thread_create (&priv->cf.fsync_th,
- NULL, changelog_fsync_thread, priv,
- "clogfsyn");
- }
-
- if (ret)
- changelog_cleanup_helper_threads (this, priv);
+changelog_spawn_helper_threads(xlator_t *this, changelog_priv_t *priv)
+{
+ int ret = 0;
+
+ /* Geo-Rep snapshot dependency:
+ *
+ * To implement explicit rollover of changlog journal on barrier
+ * notification, a pipe is created to communicate between
+ * 'changelog_rollover' thread and changelog main thread. The select
+ * call used to wait till roll-over time in changelog_rollover thread
+ * is modified to wait on read end of the pipe. When barrier
+ * notification comes (i.e, in 'reconfigure'), select in
+ * changelog_rollover thread is woken up explicitly by writing into
+ * the write end of the pipe in 'reconfigure'.
+ */
+
+ priv->cr.notify = _gf_false;
+ priv->cr.this = this;
+ ret = gf_thread_create(&priv->cr.rollover_th, NULL, changelog_rollover,
+ priv, "clogro");
+ if (ret)
+ goto out;
+
+ if (priv->fsync_interval) {
+ priv->cf.this = this;
+ ret = gf_thread_create(&priv->cf.fsync_th, NULL, changelog_fsync_thread,
+ priv, "clogfsyn");
+ }
+
+ if (ret)
+ changelog_cleanup_helper_threads(this, priv);
- out:
- return ret;
+out:
+ return ret;
}
int
-notify (xlator_t *this, int event, void *data, ...)
-{
- changelog_priv_t *priv = NULL;
- dict_t *dict = NULL;
- char buf[1] = {1};
- int barrier = DICT_DEFAULT;
- gf_boolean_t bclean_req = _gf_false;
- int ret = 0;
- int ret1 = 0;
- struct list_head queue = {0, };
+notify(xlator_t *this, int event, void *data, ...)
+{
+ changelog_priv_t *priv = NULL;
+ dict_t *dict = NULL;
+ char buf[1] = {1};
+ int barrier = DICT_DEFAULT;
+ gf_boolean_t bclean_req = _gf_false;
+ int ret = 0;
+ int ret1 = 0;
+ struct list_head queue = {
+ 0,
+ };
+ uint64_t xprtcnt = 0;
+ uint64_t clntcnt = 0;
+ changelog_clnt_t *conn = NULL;
+ gf_boolean_t cleanup_notify = _gf_false;
+ char sockfile[UNIX_PATH_MAX] = {
+ 0,
+ };
+ rpcsvc_listener_t *listener = NULL;
+ rpcsvc_listener_t *next = NULL;
+
+ INIT_LIST_HEAD(&queue);
+
+ priv = this->private;
+ if (!priv)
+ goto out;
+
+ if (event == GF_EVENT_PARENT_DOWN) {
+ priv->victim = data;
+ gf_log(this->name, GF_LOG_INFO,
+ "cleanup changelog rpc connection of brick %s",
+ priv->victim->name);
+
+ if (priv->rpc_active) {
+ this->cleanup_starting = 1;
+ changelog_destroy_rpc_listner(this, priv);
+ conn = &priv->connections;
+ if (conn)
+ changelog_ev_cleanup_connections(this, conn);
+ xprtcnt = GF_ATOMIC_GET(priv->xprtcnt);
+ clntcnt = GF_ATOMIC_GET(priv->clntcnt);
+ if (!xprtcnt && !clntcnt) {
+ LOCK(&priv->lock);
+ {
+ cleanup_notify = priv->notify_down;
+ priv->notify_down = _gf_true;
+ }
+ UNLOCK(&priv->lock);
+ if (priv->rpc) {
+ list_for_each_entry_safe(listener, next,
+ &priv->rpc->listeners, list)
+ {
+ if (listener->trans) {
+ rpc_transport_unref(listener->trans);
+ }
+ }
+ rpcsvc_destroy(priv->rpc);
+ priv->rpc = NULL;
+ }
+ CHANGELOG_MAKE_SOCKET_PATH(priv->changelog_brick, sockfile,
+ UNIX_PATH_MAX);
+ sys_unlink(sockfile);
+ if (!cleanup_notify)
+ default_notify(this, GF_EVENT_PARENT_DOWN, data);
+ }
+ } else {
+ default_notify(this, GF_EVENT_PARENT_DOWN, data);
+ }
+ goto out;
+ }
- INIT_LIST_HEAD (&queue);
+ if (event == GF_EVENT_TRANSLATOR_OP) {
+ dict = data;
- priv = this->private;
- if (!priv)
- goto out;
+ barrier = dict_get_str_boolean(dict, "barrier", DICT_DEFAULT);
- if (event == GF_EVENT_CLEANUP) {
- /* terminate helper threads */
- changelog_cleanup_helper_threads (this, priv);
+ switch (barrier) {
+ case DICT_ERROR:
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_MSG_DICT_GET_FAILED, "dict_get_str_boolean",
+ NULL);
+ ret = -1;
+ goto out;
- /* terminate RPC server/threads */
- changelog_cleanup_rpc_threads (this, priv);
- }
+ case BARRIER_OFF:
+ gf_smsg(this->name, GF_LOG_INFO, 0,
+ CHANGELOG_MSG_BARRIER_STATE_NOTIFY, "off", NULL);
- if (event == GF_EVENT_TRANSLATOR_OP) {
+ CHANGELOG_NOT_ON_THEN_GOTO(priv, ret, out);
+ LOCK(&priv->c_snap_lock);
+ {
+ changelog_snap_logging_stop(this, priv);
+ }
+ UNLOCK(&priv->c_snap_lock);
- dict = data;
+ LOCK(&priv->bflags.lock);
+ {
+ if (priv->bflags.barrier_ext == _gf_false)
+ ret = -1;
+ }
+ UNLOCK(&priv->bflags.lock);
- barrier = dict_get_str_boolean (dict, "barrier", DICT_DEFAULT);
+ if (ret == -1) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_MSG_BARRIER_ERROR, NULL);
+ goto out;
+ }
- switch (barrier) {
- case DICT_ERROR:
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_DICT_GET_FAILED,
- "Barrier dict_get_str_boolean failed");
+ /* Stop changelog barrier and dequeue all fops */
+ LOCK(&priv->lock);
+ {
+ if (priv->barrier_enabled == _gf_true)
+ __chlog_barrier_disable(this, &queue);
+ else
ret = -1;
- goto out;
-
- case BARRIER_OFF:
- gf_msg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_MSG_BARRIER_INFO,
- "Barrier off notification");
+ }
+ UNLOCK(&priv->lock);
+ /* If ret = -1, then changelog barrier is already
+ * disabled because of error or timeout.
+ */
+ if (ret == 0) {
+ chlog_barrier_dequeue_all(this, &queue);
+ gf_smsg(this->name, GF_LOG_INFO, 0,
+ CHANGELOG_MSG_BARRIER_DISABLED, NULL);
+ } else {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_MSG_BARRIER_ALREADY_DISABLED, NULL);
+ }
- CHANGELOG_NOT_ON_THEN_GOTO(priv, ret, out);
- LOCK(&priv->c_snap_lock);
- {
- changelog_snap_logging_stop (this, priv);
- }
- UNLOCK(&priv->c_snap_lock);
+ LOCK(&priv->bflags.lock);
+ {
+ priv->bflags.barrier_ext = _gf_false;
+ }
+ UNLOCK(&priv->bflags.lock);
- LOCK (&priv->bflags.lock);
- {
- if (priv->bflags.barrier_ext == _gf_false)
- ret = -1;
- }
- UNLOCK (&priv->bflags.lock);
-
- if (ret == -1 ) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_BARRIER_ERROR,
- "Received another barrier off"
- " notification while already off");
- goto out;
- }
+ goto out;
- /* Stop changelog barrier and dequeue all fops */
- LOCK (&priv->lock);
- {
- if (priv->barrier_enabled == _gf_true)
- __chlog_barrier_disable (this, &queue);
- else
- ret = -1;
- }
- UNLOCK (&priv->lock);
- /* If ret = -1, then changelog barrier is already
- * disabled because of error or timeout.
- */
- if (ret == 0) {
- chlog_barrier_dequeue_all(this, &queue);
- gf_msg(this->name, GF_LOG_INFO, 0,
- CHANGELOG_MSG_BARRIER_INFO,
- "Disabled changelog barrier");
- } else {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_BARRIER_ERROR,
- "Changelog barrier already disabled");
- }
+ case BARRIER_ON:
+ gf_smsg(this->name, GF_LOG_INFO, 0,
+ CHANGELOG_MSG_BARRIER_STATE_NOTIFY, "on", NULL);
- LOCK (&priv->bflags.lock);
- {
- priv->bflags.barrier_ext = _gf_false;
- }
- UNLOCK (&priv->bflags.lock);
+ CHANGELOG_NOT_ON_THEN_GOTO(priv, ret, out);
+ LOCK(&priv->c_snap_lock);
+ {
+ changelog_snap_logging_start(this, priv);
+ }
+ UNLOCK(&priv->c_snap_lock);
- goto out;
+ LOCK(&priv->bflags.lock);
+ {
+ if (priv->bflags.barrier_ext == _gf_true)
+ ret = -1;
+ else
+ priv->bflags.barrier_ext = _gf_true;
+ }
+ UNLOCK(&priv->bflags.lock);
- case BARRIER_ON:
- gf_msg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_MSG_BARRIER_INFO,
- "Barrier on notification");
+ if (ret == -1) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_MSG_BARRIER_ON_ERROR, NULL);
+ goto out;
+ }
- CHANGELOG_NOT_ON_THEN_GOTO(priv, ret, out);
- LOCK(&priv->c_snap_lock);
- {
- changelog_snap_logging_start (this, priv);
- }
- UNLOCK(&priv->c_snap_lock);
-
- LOCK (&priv->bflags.lock);
- {
- if (priv->bflags.barrier_ext == _gf_true)
- ret = -1;
- else
- priv->bflags.barrier_ext = _gf_true;
- }
- UNLOCK (&priv->bflags.lock);
-
- if (ret == -1 ) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_BARRIER_ERROR,
- "Received another barrier on"
- "notification when last one is"
- "not served yet");
- goto out;
- }
+ ret = pthread_mutex_lock(&priv->bn.bnotify_mutex);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_1(ret, out, bclean_req);
+ {
+ priv->bn.bnotify = _gf_true;
+ }
+ ret = pthread_mutex_unlock(&priv->bn.bnotify_mutex);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_1(ret, out, bclean_req);
- ret = pthread_mutex_lock (&priv->bn.bnotify_mutex);
- CHANGELOG_PTHREAD_ERROR_HANDLE_1 (ret, out,
- bclean_req);
- {
- priv->bn.bnotify = _gf_true;
- }
- ret = pthread_mutex_unlock (&priv->bn.bnotify_mutex);
- CHANGELOG_PTHREAD_ERROR_HANDLE_1 (ret, out,
- bclean_req);
-
- /* Start changelog barrier */
- LOCK (&priv->lock);
- {
- ret = __chlog_barrier_enable (this, priv);
- }
- UNLOCK (&priv->lock);
- if (ret == -1) {
- changelog_barrier_cleanup (this, priv, &queue);
- goto out;
- }
+ /* Start changelog barrier */
+ LOCK(&priv->lock);
+ {
+ ret = __chlog_barrier_enable(this, priv);
+ }
+ UNLOCK(&priv->lock);
+ if (ret == -1) {
+ changelog_barrier_cleanup(this, priv, &queue);
+ goto out;
+ }
- gf_msg(this->name, GF_LOG_INFO, 0,
- CHANGELOG_MSG_BARRIER_INFO,
- "Enabled changelog barrier");
-
- ret = changelog_barrier_notify(priv, buf);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_WRITE_FAILED,
- "Explicit roll over: write failed");
- changelog_barrier_cleanup (this, priv, &queue);
- ret = -1;
- goto out;
- }
+ gf_smsg(this->name, GF_LOG_INFO, 0,
+ CHANGELOG_MSG_BARRIER_ENABLE, NULL);
- ret = pthread_mutex_lock (&priv->bn.bnotify_mutex);
- CHANGELOG_PTHREAD_ERROR_HANDLE_1 (ret, out,
- bclean_req);
- {
- /* The while condition check is required here to
- * handle spurious wakeup of cond wait that can
- * happen with pthreads. See man page */
- while (priv->bn.bnotify == _gf_true) {
- ret = pthread_cond_wait (
- &priv->bn.bnotify_cond,
- &priv->bn.bnotify_mutex);
- CHANGELOG_PTHREAD_ERROR_HANDLE_1 (ret,
- out,
- bclean_req);
- }
- if (priv->bn.bnotify_error == _gf_true) {
- ret = -1;
- priv->bn.bnotify_error = _gf_false;
- }
- }
- ret1 = pthread_mutex_unlock (&priv->bn.bnotify_mutex);
- CHANGELOG_PTHREAD_ERROR_HANDLE_1 (ret1, out,
- bclean_req);
- gf_msg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_MSG_BNOTIFY_INFO,
- "Woke up: bnotify conditional wait");
-
- goto out;
-
- case DICT_DEFAULT:
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_DICT_GET_FAILED,
- "barrier key not found");
- ret = -1;
- goto out;
+ ret = changelog_barrier_notify(priv, buf);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_MSG_WRITE_FAILED, "Explicit roll over",
+ NULL);
+ changelog_barrier_cleanup(this, priv, &queue);
+ ret = -1;
+ goto out;
+ }
- default:
- gf_msg (this->name, GF_LOG_ERROR, EINVAL,
- CHANGELOG_MSG_DICT_GET_FAILED,
- "Something went bad in dict_get_str_boolean");
+ ret = pthread_mutex_lock(&priv->bn.bnotify_mutex);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_1(ret, out, bclean_req);
+ {
+ /* The while condition check is required here to
+ * handle spurious wakeup of cond wait that can
+ * happen with pthreads. See man page */
+ while (priv->bn.bnotify == _gf_true) {
+ ret = pthread_cond_wait(&priv->bn.bnotify_cond,
+ &priv->bn.bnotify_mutex);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_1(ret, out, bclean_req);
+ }
+ if (priv->bn.bnotify_error == _gf_true) {
ret = -1;
- goto out;
+ priv->bn.bnotify_error = _gf_false;
+ }
}
- } else {
- ret = default_notify (this, event, data);
+ ret1 = pthread_mutex_unlock(&priv->bn.bnotify_mutex);
+ CHANGELOG_PTHREAD_ERROR_HANDLE_1(ret1, out, bclean_req);
+ gf_smsg(this->name, GF_LOG_INFO, 0,
+ CHANGELOG_MSG_BNOTIFY_COND_INFO, NULL);
+
+ goto out;
+
+ case DICT_DEFAULT:
+ gf_smsg(this->name, GF_LOG_ERROR, 0,
+ CHANGELOG_MSG_BARRIER_KEY_NOT_FOUND, NULL);
+ ret = -1;
+ goto out;
+
+ default:
+ gf_smsg(this->name, GF_LOG_ERROR, EINVAL,
+ CHANGELOG_MSG_ERROR_IN_DICT_GET, NULL);
+ ret = -1;
+ goto out;
}
+ } else {
+ ret = default_notify(this, event, data);
+ }
- out:
- if (bclean_req)
- changelog_barrier_cleanup (this, priv, &queue);
+out:
+ if (bclean_req)
+ changelog_barrier_cleanup(this, priv, &queue);
- return ret;
+ return ret;
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
-
- if (!this)
- return ret;
+ int ret = -1;
- ret = xlator_mem_acct_init (this, gf_changelog_mt_end + 1);
+ if (!this)
+ return ret;
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- CHANGELOG_MSG_NO_MEMORY, "Memory accounting"
- " init failed");
- return ret;
- }
+ ret = xlator_mem_acct_init(this, gf_changelog_mt_end + 1);
+ if (ret != 0) {
+ gf_smsg(this->name, GF_LOG_WARNING, ENOMEM,
+ CHANGELOG_MSG_MEMORY_INIT_FAILED, NULL);
return ret;
+ }
+
+ return ret;
}
static int
-changelog_init (xlator_t *this, changelog_priv_t *priv)
+changelog_init(xlator_t *this, changelog_priv_t *priv)
{
- int i = 0;
- int ret = -1;
- struct timeval tv = {0,};
- changelog_log_data_t cld = {0,};
+ int i = 0;
+ int ret = 0;
+ changelog_log_data_t cld = {
+ 0,
+ };
- ret = gettimeofday (&tv, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_GET_TIME_OP_FAILED,
- "gettimeofday() failure");
- goto out;
- }
-
- priv->slice.tv_start = tv;
-
- priv->maps[CHANGELOG_TYPE_DATA] = "D ";
- priv->maps[CHANGELOG_TYPE_METADATA] = "M ";
- priv->maps[CHANGELOG_TYPE_METADATA_XATTR] = "M ";
- priv->maps[CHANGELOG_TYPE_ENTRY] = "E ";
+ priv->maps[CHANGELOG_TYPE_DATA] = "D ";
+ priv->maps[CHANGELOG_TYPE_METADATA] = "M ";
+ priv->maps[CHANGELOG_TYPE_METADATA_XATTR] = "M ";
+ priv->maps[CHANGELOG_TYPE_ENTRY] = "E ";
- for (; i < CHANGELOG_MAX_TYPE; i++) {
- /* start with version 1 */
- priv->slice.changelog_version[i] = 1;
- }
+ for (; i < CHANGELOG_MAX_TYPE; i++) {
+ /* start with version 1 */
+ priv->slice.changelog_version[i] = 1;
+ }
- if (!priv->active)
- return ret;
+ if (!priv->active)
+ return ret;
- /**
- * start with a fresh changelog file every time. this is done
- * in case there was an encoding change. so... things are kept
- * simple here.
- */
- ret = changelog_fill_rollover_data (&cld, _gf_false);
- if(ret)
- goto out;
+ /**
+ * start with a fresh changelog file every time. this is done
+ * in case there was an encoding change. so... things are kept
+ * simple here.
+ */
+ changelog_fill_rollover_data(&cld, _gf_false);
- ret = htime_open (this, priv, cld.cld_roll_time);
- /* call htime open with cld's rollover_time */
- if (ret)
- goto out;
+ ret = htime_open(this, priv, cld.cld_roll_time);
+ /* call htime open with cld's rollover_time */
+ if (ret)
+ goto out;
- LOCK (&priv->lock);
- {
- ret = changelog_inject_single_event (this, priv, &cld);
- }
- UNLOCK (&priv->lock);
+ LOCK(&priv->lock);
+ {
+ ret = changelog_inject_single_event(this, priv, &cld);
+ }
+ UNLOCK(&priv->lock);
- /* ... and finally spawn the helpers threads */
- ret = changelog_spawn_helper_threads (this, priv);
+ /* ... and finally spawn the helpers threads */
+ ret = changelog_spawn_helper_threads(this, priv);
- out:
- return ret;
+out:
+ return ret;
}
/**
* Init barrier related condition variables and locks
*/
static int
-changelog_barrier_pthread_init (xlator_t *this, changelog_priv_t *priv)
-{
- gf_boolean_t bn_mutex_init = _gf_false;
- gf_boolean_t bn_cond_init = _gf_false;
- gf_boolean_t dm_mutex_black_init = _gf_false;
- gf_boolean_t dm_cond_black_init = _gf_false;
- gf_boolean_t dm_mutex_white_init = _gf_false;
- gf_boolean_t dm_cond_white_init = _gf_false;
- gf_boolean_t cr_mutex_init = _gf_false;
- gf_boolean_t cr_cond_init = _gf_false;
- int ret = 0;
-
- if ((ret = pthread_mutex_init(&priv->bn.bnotify_mutex, NULL)) != 0) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_PTHREAD_MUTEX_INIT_FAILED,
- "bnotify pthread_mutex_init failed",
- "ret=%d", ret,
- NULL);
- ret = -1;
- goto out;
- }
- bn_mutex_init = _gf_true;
-
- if ((ret = pthread_cond_init(&priv->bn.bnotify_cond, NULL)) != 0) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_PTHREAD_COND_INIT_FAILED,
- "bnotify pthread_cond_init failed",
- "ret=%d", ret,
- NULL);
- ret = -1;
- goto out;
- }
- bn_cond_init = _gf_true;
-
- if ((ret = pthread_mutex_init(&priv->dm.drain_black_mutex, NULL)) != 0)
- {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_PTHREAD_MUTEX_INIT_FAILED,
- "drain_black pthread_mutex_init failed",
- "ret=%d", ret,
- NULL);
- ret = -1;
- goto out;
- }
- dm_mutex_black_init = _gf_true;
-
- if ((ret = pthread_cond_init(&priv->dm.drain_black_cond, NULL)) != 0) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_PTHREAD_COND_INIT_FAILED,
- "drain_black pthread_cond_init failed",
- "ret=%d", ret,
- NULL);
- ret = -1;
- goto out;
- }
- dm_cond_black_init = _gf_true;
-
- if ((ret = pthread_mutex_init(&priv->dm.drain_white_mutex, NULL)) != 0)
- {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_PTHREAD_MUTEX_INIT_FAILED,
- "drain_white pthread_mutex_init failed",
- "ret=%d", ret,
- NULL);
- ret = -1;
- goto out;
- }
- dm_mutex_white_init = _gf_true;
-
- if ((ret = pthread_cond_init(&priv->dm.drain_white_cond, NULL)) != 0) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_PTHREAD_COND_INIT_FAILED,
- "drain_white pthread_cond_init failed",
- "ret=%d", ret,
- NULL);
- ret = -1;
- goto out;
- }
- dm_cond_white_init = _gf_true;
-
- if ((pthread_mutex_init(&priv->cr.lock, NULL)) != 0) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_PTHREAD_MUTEX_INIT_FAILED,
- "changelog_rollover lock init failed",
- "ret=%d", ret,
- NULL);
- ret = -1;
- goto out;
- }
- cr_mutex_init = _gf_true;
-
- if ((pthread_cond_init(&priv->cr.cond, NULL)) != 0) {
- gf_smsg (this->name, GF_LOG_ERROR, errno,
- CHANGELOG_MSG_PTHREAD_COND_INIT_FAILED,
- "changelog_rollover cond init failed",
- "ret=%d", ret,
- NULL);
- ret = -1;
- goto out;
- }
- cr_cond_init = _gf_true;
- out:
- if (ret) {
- if (bn_mutex_init)
- pthread_mutex_destroy(&priv->bn.bnotify_mutex);
- if (bn_cond_init)
- pthread_cond_destroy (&priv->bn.bnotify_cond);
- if (dm_mutex_black_init)
- pthread_mutex_destroy(&priv->dm.drain_black_mutex);
- if (dm_cond_black_init)
- pthread_cond_destroy (&priv->dm.drain_black_cond);
- if (dm_mutex_white_init)
- pthread_mutex_destroy(&priv->dm.drain_white_mutex);
- if (dm_cond_white_init)
- pthread_cond_destroy (&priv->dm.drain_white_cond);
- if (cr_mutex_init)
- pthread_mutex_destroy(&priv->cr.lock);
- if (cr_cond_init)
- pthread_cond_destroy (&priv->cr.cond);
- }
- return ret;
+changelog_barrier_pthread_init(xlator_t *this, changelog_priv_t *priv)
+{
+ gf_boolean_t bn_mutex_init = _gf_false;
+ gf_boolean_t bn_cond_init = _gf_false;
+ gf_boolean_t dm_mutex_black_init = _gf_false;
+ gf_boolean_t dm_cond_black_init = _gf_false;
+ gf_boolean_t dm_mutex_white_init = _gf_false;
+ gf_boolean_t dm_cond_white_init = _gf_false;
+ gf_boolean_t cr_mutex_init = _gf_false;
+ gf_boolean_t cr_cond_init = _gf_false;
+ int ret = 0;
+
+ if ((ret = pthread_mutex_init(&priv->bn.bnotify_mutex, NULL)) != 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_PTHREAD_MUTEX_INIT_FAILED, "name=bnotify",
+ "ret=%d", ret, NULL);
+ ret = -1;
+ goto out;
+ }
+ bn_mutex_init = _gf_true;
+
+ if ((ret = pthread_cond_init(&priv->bn.bnotify_cond, NULL)) != 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_PTHREAD_COND_INIT_FAILED, "name=bnotify",
+ "ret=%d", ret, NULL);
+ ret = -1;
+ goto out;
+ }
+ bn_cond_init = _gf_true;
+
+ if ((ret = pthread_mutex_init(&priv->dm.drain_black_mutex, NULL)) != 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_PTHREAD_MUTEX_INIT_FAILED, "name=drain_black",
+ "ret=%d", ret, NULL);
+ ret = -1;
+ goto out;
+ }
+ dm_mutex_black_init = _gf_true;
+
+ if ((ret = pthread_cond_init(&priv->dm.drain_black_cond, NULL)) != 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_PTHREAD_COND_INIT_FAILED, "name=drain_black",
+ "ret=%d", ret, NULL);
+ ret = -1;
+ goto out;
+ }
+ dm_cond_black_init = _gf_true;
+
+ if ((ret = pthread_mutex_init(&priv->dm.drain_white_mutex, NULL)) != 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_PTHREAD_MUTEX_INIT_FAILED, "name=drain_white",
+ "ret=%d", ret, NULL);
+ ret = -1;
+ goto out;
+ }
+ dm_mutex_white_init = _gf_true;
+
+ if ((ret = pthread_cond_init(&priv->dm.drain_white_cond, NULL)) != 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_PTHREAD_COND_INIT_FAILED, "name=drain_white",
+ "ret=%d", ret, NULL);
+ ret = -1;
+ goto out;
+ }
+ dm_cond_white_init = _gf_true;
+
+ if ((pthread_mutex_init(&priv->cr.lock, NULL)) != 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_PTHREAD_MUTEX_INIT_FAILED,
+ "name=changelog_rollover", "ret=%d", ret, NULL);
+ ret = -1;
+ goto out;
+ }
+ cr_mutex_init = _gf_true;
+
+ if ((pthread_cond_init(&priv->cr.cond, NULL)) != 0) {
+ gf_smsg(this->name, GF_LOG_ERROR, errno,
+ CHANGELOG_MSG_PTHREAD_COND_INIT_FAILED,
+ "changelog_rollover cond init failed", "ret=%d", ret, NULL);
+ ret = -1;
+ goto out;
+ }
+ cr_cond_init = _gf_true;
+out:
+ if (ret) {
+ if (bn_mutex_init)
+ pthread_mutex_destroy(&priv->bn.bnotify_mutex);
+ if (bn_cond_init)
+ pthread_cond_destroy(&priv->bn.bnotify_cond);
+ if (dm_mutex_black_init)
+ pthread_mutex_destroy(&priv->dm.drain_black_mutex);
+ if (dm_cond_black_init)
+ pthread_cond_destroy(&priv->dm.drain_black_cond);
+ if (dm_mutex_white_init)
+ pthread_mutex_destroy(&priv->dm.drain_white_mutex);
+ if (dm_cond_white_init)
+ pthread_cond_destroy(&priv->dm.drain_white_cond);
+ if (cr_mutex_init)
+ pthread_mutex_destroy(&priv->cr.lock);
+ if (cr_cond_init)
+ pthread_cond_destroy(&priv->cr.cond);
+ }
+ return ret;
}
/* Destroy barrier related condition variables and locks */
static void
-changelog_barrier_pthread_destroy (changelog_priv_t *priv)
+changelog_barrier_pthread_destroy(changelog_priv_t *priv)
{
- pthread_mutex_destroy (&priv->bn.bnotify_mutex);
- pthread_cond_destroy (&priv->bn.bnotify_cond);
- pthread_mutex_destroy (&priv->dm.drain_black_mutex);
- pthread_cond_destroy (&priv->dm.drain_black_cond);
- pthread_mutex_destroy (&priv->dm.drain_white_mutex);
- pthread_cond_destroy (&priv->dm.drain_white_cond);
- pthread_mutex_destroy(&priv->cr.lock);
- pthread_cond_destroy (&priv->cr.cond);
- LOCK_DESTROY (&priv->bflags.lock);
+ pthread_mutex_destroy(&priv->bn.bnotify_mutex);
+ pthread_cond_destroy(&priv->bn.bnotify_cond);
+ pthread_mutex_destroy(&priv->dm.drain_black_mutex);
+ pthread_cond_destroy(&priv->dm.drain_black_cond);
+ pthread_mutex_destroy(&priv->dm.drain_white_mutex);
+ pthread_cond_destroy(&priv->dm.drain_white_cond);
+ pthread_mutex_destroy(&priv->cr.lock);
+ pthread_cond_destroy(&priv->cr.cond);
+ LOCK_DESTROY(&priv->bflags.lock);
}
-int
-reconfigure (xlator_t *this, dict_t *options)
-{
- int ret = 0;
- char *tmp = NULL;
- changelog_priv_t *priv = NULL;
- gf_boolean_t active_earlier = _gf_true;
- gf_boolean_t active_now = _gf_true;
- changelog_time_slice_t *slice = NULL;
- changelog_log_data_t cld = {0,};
- char htime_dir[PATH_MAX] = {0,};
- char csnap_dir[PATH_MAX] = {0,};
- struct timeval tv = {0,};
- uint32_t timeout = 0;
-
- priv = this->private;
- if (!priv)
- goto out;
-
- ret = -1;
- active_earlier = priv->active;
-
- /* first stop the rollover and the fsync thread */
- changelog_cleanup_helper_threads (this, priv);
-
- GF_OPTION_RECONF ("changelog-dir", tmp, options, str, out);
- if (!tmp) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_DIR_OPTIONS_NOT_SET,
- "\"changelog-dir\" option is not set");
- goto out;
- }
-
- GF_FREE (priv->changelog_dir);
- priv->changelog_dir = gf_strdup (tmp);
- if (!priv->changelog_dir)
- goto out;
+static void
+changelog_cleanup_rpc(xlator_t *this, changelog_priv_t *priv)
+{
+ /* terminate rpc server */
+ if (!this->cleanup_starting)
+ changelog_destroy_rpc_listner(this, priv);
- ret = mkdir_p (priv->changelog_dir, 0600, _gf_true);
+ (void)changelog_cleanup_rpc_threads(this, priv);
+ /* cleanup rot buffs */
+ rbuf_dtor(priv->rbuf);
- if (ret)
- goto out;
- CHANGELOG_FILL_HTIME_DIR(priv->changelog_dir, htime_dir);
- ret = mkdir_p (htime_dir, 0600, _gf_true);
+ /* cleanup poller thread */
+ if (priv->poller)
+ (void)changelog_thread_cleanup(this, priv->poller);
+}
+int
+reconfigure(xlator_t *this, dict_t *options)
+{
+ int ret = 0;
+ char *tmp = NULL;
+ changelog_priv_t *priv = NULL;
+ gf_boolean_t active_earlier = _gf_true;
+ gf_boolean_t active_now = _gf_true;
+ gf_boolean_t rpc_active_earlier = _gf_true;
+ gf_boolean_t rpc_active_now = _gf_true;
+ gf_boolean_t iniate_rpc = _gf_false;
+ changelog_time_slice_t *slice = NULL;
+ changelog_log_data_t cld = {
+ 0,
+ };
+ char htime_dir[PATH_MAX] = {
+ 0,
+ };
+ char csnap_dir[PATH_MAX] = {
+ 0,
+ };
+ uint32_t timeout = 0;
+
+ priv = this->private;
+ if (!priv)
+ goto out;
+
+ ret = -1;
+ active_earlier = priv->active;
+ rpc_active_earlier = priv->rpc_active;
+
+ /* first stop the rollover and the fsync thread */
+ changelog_cleanup_helper_threads(this, priv);
+
+ GF_OPTION_RECONF("changelog-dir", tmp, options, str, out);
+ if (!tmp) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_DIR_OPTIONS_NOT_SET,
+ NULL);
+ goto out;
+ }
+
+ GF_FREE(priv->changelog_dir);
+ priv->changelog_dir = gf_strdup(tmp);
+ if (!priv->changelog_dir)
+ goto out;
+
+ ret = mkdir_p(priv->changelog_dir, 0600, _gf_true);
+
+ if (ret)
+ goto out;
+ CHANGELOG_FILL_HTIME_DIR(priv->changelog_dir, htime_dir);
+ ret = mkdir_p(htime_dir, 0600, _gf_true);
+
+ if (ret)
+ goto out;
+
+ CHANGELOG_FILL_CSNAP_DIR(priv->changelog_dir, csnap_dir);
+ ret = mkdir_p(csnap_dir, 0600, _gf_true);
+
+ if (ret)
+ goto out;
+
+ GF_OPTION_RECONF("changelog", active_now, options, bool, out);
+ GF_OPTION_RECONF("changelog-notification", rpc_active_now, options, bool,
+ out);
+
+ /* If journalling is enabled, enable rpc notifications */
+ if (active_now && !active_earlier) {
+ if (!rpc_active_earlier)
+ iniate_rpc = _gf_true;
+ }
+
+ if (rpc_active_now && !rpc_active_earlier) {
+ iniate_rpc = _gf_true;
+ }
+
+ /* TODO: Disable of changelog-notifications is not supported for now
+ * as there is no clean way of cleaning up of rpc resources
+ */
+
+ if (iniate_rpc) {
+ ret = changelog_init_rpc(this, priv);
if (ret)
- goto out;
+ goto out;
+ priv->rpc_active = _gf_true;
+ }
- CHANGELOG_FILL_CSNAP_DIR(priv->changelog_dir, csnap_dir);
- ret = mkdir_p (csnap_dir, 0600, _gf_true);
-
- if (ret)
- goto out;
+ /**
+ * changelog_handle_change() handles changes that could possibly
+ * have been submit changes before changelog deactivation.
+ */
+ if (!active_now)
+ priv->active = _gf_false;
- GF_OPTION_RECONF ("changelog", active_now, options, bool, out);
+ GF_OPTION_RECONF("op-mode", tmp, options, str, out);
+ changelog_assign_opmode(priv, tmp);
- /**
- * changelog_handle_change() handles changes that could possibly
- * have been submit changes before changelog deactivation.
- */
- if (!active_now)
- priv->active = _gf_false;
+ tmp = NULL;
- GF_OPTION_RECONF ("op-mode", tmp, options, str, out);
- changelog_assign_opmode (priv, tmp);
+ GF_OPTION_RECONF("encoding", tmp, options, str, out);
+ changelog_assign_encoding(priv, tmp);
- tmp = NULL;
+ GF_OPTION_RECONF("rollover-time", priv->rollover_time, options, int32, out);
+ GF_OPTION_RECONF("fsync-interval", priv->fsync_interval, options, int32,
+ out);
+ GF_OPTION_RECONF("changelog-barrier-timeout", timeout, options, time, out);
+ changelog_assign_barrier_timeout(priv, timeout);
- GF_OPTION_RECONF ("encoding", tmp, options, str, out);
- changelog_assign_encoding (priv, tmp);
+ GF_OPTION_RECONF("capture-del-path", priv->capture_del_path, options, bool,
+ out);
- GF_OPTION_RECONF ("rollover-time",
- priv->rollover_time, options, int32, out);
- GF_OPTION_RECONF ("fsync-interval",
- priv->fsync_interval, options, int32, out);
- GF_OPTION_RECONF ("changelog-barrier-timeout",
- timeout, options, time, out);
- changelog_assign_barrier_timeout (priv, timeout);
+ if (active_now || active_earlier) {
+ changelog_fill_rollover_data(&cld, !active_now);
- GF_OPTION_RECONF ("capture-del-path", priv->capture_del_path, options,
- bool, out);
+ slice = &priv->slice;
- if (active_now || active_earlier) {
- ret = changelog_fill_rollover_data (&cld, !active_now);
- if (ret)
- goto out;
+ LOCK(&priv->lock);
+ {
+ ret = changelog_inject_single_event(this, priv, &cld);
+ if (!ret && active_now)
+ SLICE_VERSION_UPDATE(slice);
+ }
+ UNLOCK(&priv->lock);
- slice = &priv->slice;
+ if (ret)
+ goto out;
- LOCK (&priv->lock);
- {
- ret = changelog_inject_single_event (this, priv, &cld);
- if (!ret && active_now)
- SLICE_VERSION_UPDATE (slice);
- }
- UNLOCK (&priv->lock);
-
- if (ret)
- goto out;
-
- if (active_now) {
- if (!active_earlier) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- CHANGELOG_MSG_HTIME_INFO,
- "Reconfigure: Changelog Enable");
- if (gettimeofday(&tv, NULL) ) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_HTIME_ERROR,
- "unable to fetch htime");
- ret = -1;
- goto out;
- }
- htime_create (this, priv, tv.tv_sec);
- }
- ret = changelog_spawn_helper_threads (this, priv);
- }
+ if (active_now) {
+ if (!active_earlier) {
+ gf_smsg(this->name, GF_LOG_INFO, 0, CHANGELOG_MSG_RECONFIGURE,
+ NULL);
+ htime_create(this, priv, gf_time());
+ }
+ ret = changelog_spawn_helper_threads(this, priv);
}
+ }
- out:
- if (ret) {
- /* TODO */
- } else {
- gf_msg_debug (this->name, 0,
- "changelog reconfigured");
- if (active_now && priv)
- priv->active = _gf_true;
- }
+out:
+ if (ret) {
+ /* TODO */
+ } else {
+ gf_msg_debug(this->name, 0, "changelog reconfigured");
+ if (active_now && priv)
+ priv->active = _gf_true;
+ }
- return ret;
+ return ret;
}
static void
-changelog_freeup_options (xlator_t *this, changelog_priv_t *priv)
+changelog_freeup_options(xlator_t *this, changelog_priv_t *priv)
{
- int ret = 0;
+ int ret = 0;
- ret = priv->cb->dtor (this, &priv->cd);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_FREEUP_FAILED,
- "could not cleanup bootstrapper");
- GF_FREE (priv->changelog_brick);
- GF_FREE (priv->changelog_dir);
+ ret = priv->cb->dtor(this, &priv->cd);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_FREEUP_FAILED, NULL);
+ GF_FREE(priv->changelog_brick);
+ GF_FREE(priv->changelog_dir);
}
static int
-changelog_init_options (xlator_t *this, changelog_priv_t *priv)
+changelog_init_options(xlator_t *this, changelog_priv_t *priv)
{
- int ret = 0;
- char *tmp = NULL;
- uint32_t timeout = 0;
- char htime_dir[PATH_MAX] = {0,};
- char csnap_dir[PATH_MAX] = {0,};
+ int ret = 0;
+ char *tmp = NULL;
+ uint32_t timeout = 0;
+ char htime_dir[PATH_MAX] = {
+ 0,
+ };
+ char csnap_dir[PATH_MAX] = {
+ 0,
+ };
- GF_OPTION_INIT ("changelog-brick", tmp, str, error_return);
- priv->changelog_brick = gf_strdup (tmp);
- if (!priv->changelog_brick)
- goto error_return;
+ GF_OPTION_INIT("changelog-brick", tmp, str, error_return);
+ priv->changelog_brick = gf_strdup(tmp);
+ if (!priv->changelog_brick)
+ goto error_return;
- tmp = NULL;
+ tmp = NULL;
- GF_OPTION_INIT ("changelog-dir", tmp, str, dealloc_1);
- priv->changelog_dir = gf_strdup (tmp);
- if (!priv->changelog_dir)
- goto dealloc_1;
+ GF_OPTION_INIT("changelog-dir", tmp, str, dealloc_1);
+ priv->changelog_dir = gf_strdup(tmp);
+ if (!priv->changelog_dir)
+ goto dealloc_1;
- tmp = NULL;
+ tmp = NULL;
- /**
- * create the directory even if change-logging would be inactive
- * so that consumers can _look_ into it (finding nothing...)
- */
- ret = mkdir_p (priv->changelog_dir, 0600, _gf_true);
+ /**
+ * create the directory even if change-logging would be inactive
+ * so that consumers can _look_ into it (finding nothing...)
+ */
+ ret = mkdir_p(priv->changelog_dir, 0600, _gf_true);
- if (ret)
- goto dealloc_2;
+ if (ret)
+ goto dealloc_2;
- CHANGELOG_FILL_HTIME_DIR (priv->changelog_dir, htime_dir);
- ret = mkdir_p (htime_dir, 0600, _gf_true);
- if (ret)
- goto dealloc_2;
+ CHANGELOG_FILL_HTIME_DIR(priv->changelog_dir, htime_dir);
+ ret = mkdir_p(htime_dir, 0600, _gf_true);
+ if (ret)
+ goto dealloc_2;
- CHANGELOG_FILL_CSNAP_DIR (priv->changelog_dir, csnap_dir);
- ret = mkdir_p (csnap_dir, 0600, _gf_true);
- if (ret)
- goto dealloc_2;
+ CHANGELOG_FILL_CSNAP_DIR(priv->changelog_dir, csnap_dir);
+ ret = mkdir_p(csnap_dir, 0600, _gf_true);
+ if (ret)
+ goto dealloc_2;
- GF_OPTION_INIT ("changelog", priv->active, bool, dealloc_2);
- GF_OPTION_INIT ("capture-del-path", priv->capture_del_path,
- bool, dealloc_2);
+ GF_OPTION_INIT("changelog", priv->active, bool, dealloc_2);
+ GF_OPTION_INIT("changelog-notification", priv->rpc_active, bool, dealloc_2);
+ GF_OPTION_INIT("capture-del-path", priv->capture_del_path, bool, dealloc_2);
- GF_OPTION_INIT ("op-mode", tmp, str, dealloc_2);
- changelog_assign_opmode (priv, tmp);
+ GF_OPTION_INIT("op-mode", tmp, str, dealloc_2);
+ changelog_assign_opmode(priv, tmp);
- tmp = NULL;
+ tmp = NULL;
- GF_OPTION_INIT ("encoding", tmp, str, dealloc_2);
- changelog_assign_encoding (priv, tmp);
- changelog_encode_change (priv);
+ GF_OPTION_INIT("encoding", tmp, str, dealloc_2);
+ changelog_assign_encoding(priv, tmp);
+ changelog_encode_change(priv);
- GF_OPTION_INIT ("rollover-time",
- priv->rollover_time, int32, dealloc_2);
+ GF_OPTION_INIT("rollover-time", priv->rollover_time, int32, dealloc_2);
- GF_OPTION_INIT ("fsync-interval",
- priv->fsync_interval, int32, dealloc_2);
+ GF_OPTION_INIT("fsync-interval", priv->fsync_interval, int32, dealloc_2);
- GF_OPTION_INIT ("changelog-barrier-timeout",
- timeout, time, dealloc_2);
- changelog_assign_barrier_timeout (priv, timeout);
+ GF_OPTION_INIT("changelog-barrier-timeout", timeout, time, dealloc_2);
+ changelog_assign_barrier_timeout(priv, timeout);
- GF_ASSERT (cb_bootstrap[priv->op_mode].mode == priv->op_mode);
- priv->cb = &cb_bootstrap[priv->op_mode];
+ GF_ASSERT(cb_bootstrap[priv->op_mode].mode == priv->op_mode);
+ priv->cb = &cb_bootstrap[priv->op_mode];
- /* ... now bootstrap the logger */
- ret = priv->cb->ctor (this, &priv->cd);
- if (ret)
- goto dealloc_2;
+ /* ... now bootstrap the logger */
+ ret = priv->cb->ctor(this, &priv->cd);
+ if (ret)
+ goto dealloc_2;
- priv->changelog_fd = -1;
+ priv->changelog_fd = -1;
- return 0;
+ return 0;
- dealloc_2:
- GF_FREE (priv->changelog_dir);
- dealloc_1:
- GF_FREE (priv->changelog_brick);
- error_return:
- return -1;
-}
-
-static void
-changelog_cleanup_rpc (xlator_t *this, changelog_priv_t *priv)
-{
- /* terminate rpc server */
- changelog_destroy_rpc_listner (this, priv);
-
- /* cleanup rot buffs */
- rbuf_dtor (priv->rbuf);
-
- /* cleanup poller thread */
- if (priv->poller)
- (void) changelog_thread_cleanup (this, priv->poller);
+dealloc_2:
+ GF_FREE(priv->changelog_dir);
+dealloc_1:
+ GF_FREE(priv->changelog_brick);
+error_return:
+ return -1;
}
static int
-changelog_init_rpc (xlator_t *this, changelog_priv_t *priv)
+changelog_init_rpc(xlator_t *this, changelog_priv_t *priv)
{
- rpcsvc_t *rpc = NULL;
- changelog_ev_selector_t *selection = NULL;
+ rpcsvc_t *rpc = NULL;
+ changelog_ev_selector_t *selection = NULL;
- selection = &priv->ev_selection;
+ selection = &priv->ev_selection;
- /* initialize event selection */
- changelog_init_event_selection (this, selection);
+ /* initialize event selection */
+ changelog_init_event_selection(this, selection);
- priv->rbuf = rbuf_init (NR_ROTT_BUFFS);
- if (!priv->rbuf)
- goto cleanup_thread;
+ priv->rbuf = rbuf_init(NR_ROTT_BUFFS);
+ if (!priv->rbuf)
+ goto cleanup_thread;
- rpc = changelog_init_rpc_listener (this, priv,
- priv->rbuf, NR_DISPATCHERS);
- if (!rpc)
- goto cleanup_rbuf;
- priv->rpc = rpc;
+ rpc = changelog_init_rpc_listener(this, priv, priv->rbuf, NR_DISPATCHERS);
+ if (!rpc)
+ goto cleanup_rbuf;
+ priv->rpc = rpc;
- return 0;
+ return 0;
- cleanup_rbuf:
- rbuf_dtor (priv->rbuf);
- cleanup_thread:
- if (priv->poller)
- (void) changelog_thread_cleanup (this, priv->poller);
+cleanup_rbuf:
+ rbuf_dtor(priv->rbuf);
+cleanup_thread:
+ if (priv->poller)
+ (void)changelog_thread_cleanup(this, priv->poller);
- return -1;
+ return -1;
}
int32_t
-init (xlator_t *this)
-{
- int ret = -1;
- changelog_priv_t *priv = NULL;
-
- GF_VALIDATE_OR_GOTO ("changelog", this, error_return);
-
- if (!this->children || this->children->next) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_CHILD_MISCONFIGURED,
- "translator needs a single subvolume");
- goto error_return;
- }
-
- if (!this->parents) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CHANGELOG_MSG_VOL_MISCONFIGURED,
- "dangling volume. please check volfile");
- goto error_return;
- }
-
- priv = GF_CALLOC (1, sizeof (*priv), gf_changelog_mt_priv_t);
- if (!priv)
- goto error_return;
-
- this->local_pool = mem_pool_new (changelog_local_t, 64);
- if (!this->local_pool) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- CHANGELOG_MSG_NO_MEMORY,
- "failed to create local memory pool");
- goto cleanup_priv;
- }
-
- LOCK_INIT (&priv->lock);
- LOCK_INIT (&priv->c_snap_lock);
-
- ret = changelog_init_options (this, priv);
- if (ret)
- goto cleanup_mempool;
-
- /* snap dependency changes */
- priv->dm.black_fop_cnt = 0;
- priv->dm.white_fop_cnt = 0;
- priv->dm.drain_wait_black = _gf_false;
- priv->dm.drain_wait_white = _gf_false;
- priv->current_color = FOP_COLOR_BLACK;
- priv->explicit_rollover = _gf_false;
-
- priv->cr.notify = _gf_false;
- /* Mutex is not needed as threads are not spawned yet */
- priv->bn.bnotify = _gf_false;
- priv->bn.bnotify_error = _gf_false;
- ret = changelog_barrier_pthread_init (this, priv);
- if (ret)
- goto cleanup_options;
- LOCK_INIT (&priv->bflags.lock);
- priv->bflags.barrier_ext = _gf_false;
-
- /* Changelog barrier init */
- INIT_LIST_HEAD (&priv->queue);
- priv->barrier_enabled = _gf_false;
-
+init(xlator_t *this)
+{
+ int ret = -1;
+ changelog_priv_t *priv = NULL;
+
+ GF_VALIDATE_OR_GOTO("changelog", this, error_return);
+
+ if (!this->children || this->children->next) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_CHILD_MISCONFIGURED,
+ NULL);
+ goto error_return;
+ }
+
+ if (!this->parents) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, CHANGELOG_MSG_VOL_MISCONFIGURED,
+ NULL);
+ goto error_return;
+ }
+
+ priv = GF_CALLOC(1, sizeof(*priv), gf_changelog_mt_priv_t);
+ if (!priv)
+ goto error_return;
+
+ this->local_pool = mem_pool_new(changelog_local_t, 64);
+ if (!this->local_pool) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, CHANGELOG_MSG_NO_MEMORY,
+ NULL);
+ goto cleanup_priv;
+ }
+
+ LOCK_INIT(&priv->lock);
+ LOCK_INIT(&priv->c_snap_lock);
+ GF_ATOMIC_INIT(priv->listnercnt, 0);
+ GF_ATOMIC_INIT(priv->clntcnt, 0);
+ GF_ATOMIC_INIT(priv->xprtcnt, 0);
+ INIT_LIST_HEAD(&priv->xprt_list);
+ priv->htime_fd = -1;
+
+ ret = changelog_init_options(this, priv);
+ if (ret)
+ goto cleanup_mempool;
+
+ /* snap dependency changes */
+ priv->dm.black_fop_cnt = 0;
+ priv->dm.white_fop_cnt = 0;
+ priv->dm.drain_wait_black = _gf_false;
+ priv->dm.drain_wait_white = _gf_false;
+ priv->current_color = FOP_COLOR_BLACK;
+ priv->explicit_rollover = _gf_false;
+
+ priv->cr.notify = _gf_false;
+ /* Mutex is not needed as threads are not spawned yet */
+ priv->bn.bnotify = _gf_false;
+ priv->bn.bnotify_error = _gf_false;
+ ret = changelog_barrier_pthread_init(this, priv);
+ if (ret)
+ goto cleanup_options;
+ LOCK_INIT(&priv->bflags.lock);
+ priv->bflags.barrier_ext = _gf_false;
+
+ /* Changelog barrier init */
+ INIT_LIST_HEAD(&priv->queue);
+ priv->barrier_enabled = _gf_false;
+
+ if (priv->rpc_active || priv->active) {
/* RPC ball rolling.. */
- ret = changelog_init_rpc (this, priv);
- if (ret)
- goto cleanup_barrier;
-
- ret = changelog_init (this, priv);
+ ret = changelog_init_rpc(this, priv);
if (ret)
- goto cleanup_rpc;
-
- gf_msg_debug (this->name, 0, "changelog translator loaded");
-
- this->private = priv;
- return 0;
-
- cleanup_rpc:
- changelog_cleanup_rpc (this, priv);
- cleanup_barrier:
- changelog_barrier_pthread_destroy (priv);
- cleanup_options:
- changelog_freeup_options (this, priv);
- cleanup_mempool:
- mem_pool_destroy (this->local_pool);
- cleanup_priv:
- GF_FREE (priv);
- error_return:
- this->private = NULL;
- return -1;
+ goto cleanup_barrier;
+ priv->rpc_active = _gf_true;
+ }
+
+ ret = changelog_init(this, priv);
+ if (ret)
+ goto cleanup_rpc;
+
+ gf_msg_debug(this->name, 0, "changelog translator loaded");
+
+ this->private = priv;
+ return 0;
+
+cleanup_rpc:
+ if (priv->rpc_active) {
+ changelog_cleanup_rpc(this, priv);
+ }
+cleanup_barrier:
+ changelog_barrier_pthread_destroy(priv);
+cleanup_options:
+ changelog_freeup_options(this, priv);
+cleanup_mempool:
+ mem_pool_destroy(this->local_pool);
+ this->local_pool = NULL;
+cleanup_priv:
+ GF_FREE(priv);
+error_return:
+ this->private = NULL;
+ return -1;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- changelog_priv_t *priv = NULL;
+ changelog_priv_t *priv = NULL;
+ struct list_head queue = {
+ 0,
+ };
+
+ priv = this->private;
+
+ if (priv) {
+ if (priv->active || priv->rpc_active) {
+ /* terminate RPC server/threads */
+ changelog_cleanup_rpc(this, priv);
+ GF_FREE(priv->ev_dispatcher);
+ }
+ /* call barrier_disable to cancel timer */
+ if (priv->barrier_enabled)
+ __chlog_barrier_disable(this, &queue);
- priv = this->private;
+ /* cleanup barrier related objects */
+ changelog_barrier_pthread_destroy(priv);
- if (priv) {
- /* terminate RPC server/threads */
- changelog_cleanup_rpc (this, priv);
+ /* cleanup helper threads */
+ changelog_cleanup_helper_threads(this, priv);
- /* cleanup barrier related objects */
- changelog_barrier_pthread_destroy (priv);
+ /* cleanup allocated options */
+ changelog_freeup_options(this, priv);
- /* cleanup allocated options */
- changelog_freeup_options (this, priv);
+ /* deallocate mempool */
+ mem_pool_destroy(this->local_pool);
- /* deallocate mempool */
- mem_pool_destroy (this->local_pool);
- /* finally, dealloac private variable */
- GF_FREE (priv);
+ if (priv->htime_fd != -1) {
+ sys_close(priv->htime_fd);
}
- this->private = NULL;
+ /* finally, dealloac private variable */
+ GF_FREE(priv);
+ }
- return;
+ this->private = NULL;
+ this->local_pool = NULL;
+
+ return;
}
struct xlator_fops fops = {
- .open = changelog_open,
- .mknod = changelog_mknod,
- .mkdir = changelog_mkdir,
- .create = changelog_create,
- .symlink = changelog_symlink,
- .writev = changelog_writev,
- .truncate = changelog_truncate,
- .ftruncate = changelog_ftruncate,
- .link = changelog_link,
- .rename = changelog_rename,
- .unlink = changelog_unlink,
- .rmdir = changelog_rmdir,
- .setattr = changelog_setattr,
- .fsetattr = changelog_fsetattr,
- .setxattr = changelog_setxattr,
- .fsetxattr = changelog_fsetxattr,
- .removexattr = changelog_removexattr,
- .fremovexattr = changelog_fremovexattr,
- .ipc = changelog_ipc,
- .xattrop = changelog_xattrop,
- .fxattrop = changelog_fxattrop,
+ .open = changelog_open,
+ .mknod = changelog_mknod,
+ .mkdir = changelog_mkdir,
+ .create = changelog_create,
+ .symlink = changelog_symlink,
+ .writev = changelog_writev,
+ .truncate = changelog_truncate,
+ .ftruncate = changelog_ftruncate,
+ .link = changelog_link,
+ .rename = changelog_rename,
+ .unlink = changelog_unlink,
+ .rmdir = changelog_rmdir,
+ .setattr = changelog_setattr,
+ .fsetattr = changelog_fsetattr,
+ .setxattr = changelog_setxattr,
+ .fsetxattr = changelog_fsetxattr,
+ .removexattr = changelog_removexattr,
+ .fremovexattr = changelog_fremovexattr,
+ .ipc = changelog_ipc,
+ .xattrop = changelog_xattrop,
+ .fxattrop = changelog_fxattrop,
};
struct xlator_cbks cbks = {
- .forget = changelog_forget,
- .release = changelog_release,
+ .forget = changelog_forget,
+ .release = changelog_release,
};
struct volume_options options[] = {
- {.key = {"changelog"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "enable/disable change-logging",
- .op_version = {3},
- .flags = OPT_FLAG_SETTABLE,
- .level = OPT_STATUS_BASIC,
- .tags = {"journal", "georep", "glusterfind"}
- },
- {.key = {"changelog-brick"},
- .type = GF_OPTION_TYPE_PATH,
- .description = "brick path to generate unique socket file name."
- " should be the export directory of the volume strictly.",
- .default_value = "{{ brick.path }}",
- .op_version = {3},
- .tags = {"journal"}
- },
- {.key = {"changelog-dir"},
- .type = GF_OPTION_TYPE_PATH,
- .description = "directory for the changelog files",
- .default_value = "{{ brick.path }}/.glusterfs/changelogs",
- .op_version = {3},
- .flags = OPT_FLAG_SETTABLE,
- .level = OPT_STATUS_ADVANCED,
- .tags = {"journal", "georep", "glusterfind"}
- },
- {.key = {"op-mode"},
- .type = GF_OPTION_TYPE_STR,
- .default_value = "realtime",
- .value = {"realtime"},
- .description = "operation mode - futuristic operation modes",
- .op_version = {3},
- .tags = {"journal"}
- },
- {.key = {"encoding"},
- .type = GF_OPTION_TYPE_STR,
- .default_value = "ascii",
- .value = {"binary", "ascii"},
- .description = "encoding type for changelogs",
- .op_version = {3},
- .flags = OPT_FLAG_SETTABLE,
- .level = OPT_STATUS_ADVANCED,
- .tags = {"journal"}
- },
- {.key = {"rollover-time"},
- .default_value = "15",
- .type = GF_OPTION_TYPE_TIME,
- .description = "time to switch to a new changelog file (in seconds)",
- .op_version = {3},
- .flags = OPT_FLAG_SETTABLE,
- .level = OPT_STATUS_ADVANCED,
- .tags = {"journal", "georep", "glusterfind"}
- },
- {.key = {"fsync-interval"},
- .type = GF_OPTION_TYPE_TIME,
- .default_value = "5",
- .description = "do not open CHANGELOG file with O_SYNC mode."
- " instead perform fsync() at specified intervals",
- .op_version = {3},
- .flags = OPT_FLAG_SETTABLE,
- .level = OPT_STATUS_ADVANCED,
- .tags = {"journal"}
- },
- { .key = {"changelog-barrier-timeout"},
- .type = GF_OPTION_TYPE_TIME,
- .default_value = BARRIER_TIMEOUT,
- .description = "After 'timeout' seconds since the time 'barrier' "
- "option was set to \"on\", unlink/rmdir/rename "
- "operations are no longer blocked and previously "
- "blocked fops are allowed to go through",
- .op_version = {3},
- .flags = OPT_FLAG_SETTABLE,
- .level = OPT_STATUS_ADVANCED,
- .tags = {"journal"}
- },
- {.key = {"capture-del-path"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "enable/disable capturing paths of deleted entries",
- .op_version = {3},
- .flags = OPT_FLAG_SETTABLE,
- .level = OPT_STATUS_BASIC,
- .tags = {"journal", "glusterfind"}
- },
- {.key = {NULL}
- },
+ {.key = {"changelog"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "enable/disable change-logging",
+ .op_version = {3},
+ .flags = OPT_FLAG_SETTABLE,
+ .level = OPT_STATUS_BASIC,
+ .tags = {"journal", "georep", "glusterfind"}},
+ {.key = {"changelog-notification"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "enable/disable changelog live notification",
+ .op_version = {3},
+ .level = OPT_STATUS_BASIC,
+ .tags = {"bitrot", "georep"}},
+ {.key = {"changelog-brick"},
+ .type = GF_OPTION_TYPE_PATH,
+ .description = "brick path to generate unique socket file name."
+ " should be the export directory of the volume strictly.",
+ .default_value = "{{ brick.path }}",
+ .op_version = {3},
+ .tags = {"journal"}},
+ {.key = {"changelog-dir"},
+ .type = GF_OPTION_TYPE_PATH,
+ .description = "directory for the changelog files",
+ .default_value = "{{ brick.path }}/.glusterfs/changelogs",
+ .op_version = {3},
+ .flags = OPT_FLAG_SETTABLE,
+ .level = OPT_STATUS_ADVANCED,
+ .tags = {"journal", "georep", "glusterfind"}},
+ {.key = {"op-mode"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = "realtime",
+ .value = {"realtime"},
+ .description = "operation mode - futuristic operation modes",
+ .op_version = {3},
+ .tags = {"journal"}},
+ {.key = {"encoding"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = "ascii",
+ .value = {"binary", "ascii"},
+ .description = "encoding type for changelogs",
+ .op_version = {3},
+ .flags = OPT_FLAG_SETTABLE,
+ .level = OPT_STATUS_ADVANCED,
+ .tags = {"journal"}},
+ {.key = {"rollover-time"},
+ .default_value = "15",
+ .type = GF_OPTION_TYPE_TIME,
+ .description = "time to switch to a new changelog file (in seconds)",
+ .op_version = {3},
+ .flags = OPT_FLAG_SETTABLE,
+ .level = OPT_STATUS_ADVANCED,
+ .tags = {"journal", "georep", "glusterfind"}},
+ {.key = {"fsync-interval"},
+ .type = GF_OPTION_TYPE_TIME,
+ .default_value = "5",
+ .description = "do not open CHANGELOG file with O_SYNC mode."
+ " instead perform fsync() at specified intervals",
+ .op_version = {3},
+ .flags = OPT_FLAG_SETTABLE,
+ .level = OPT_STATUS_ADVANCED,
+ .tags = {"journal"}},
+ {.key = {"changelog-barrier-timeout"},
+ .type = GF_OPTION_TYPE_TIME,
+ .default_value = BARRIER_TIMEOUT,
+ .description = "After 'timeout' seconds since the time 'barrier' "
+ "option was set to \"on\", unlink/rmdir/rename "
+ "operations are no longer blocked and previously "
+ "blocked fops are allowed to go through",
+ .op_version = {3},
+ .flags = OPT_FLAG_SETTABLE,
+ .level = OPT_STATUS_ADVANCED,
+ .tags = {"journal"}},
+ {.key = {"capture-del-path"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "enable/disable capturing paths of deleted entries",
+ .op_version = {3},
+ .flags = OPT_FLAG_SETTABLE,
+ .level = OPT_STATUS_BASIC,
+ .tags = {"journal", "glusterfind"}},
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .notify = notify,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "changelog",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/features/changetimerecorder/src/Makefile.am b/xlators/features/changetimerecorder/src/Makefile.am
deleted file mode 100644
index 620017e3309..00000000000
--- a/xlators/features/changetimerecorder/src/Makefile.am
+++ /dev/null
@@ -1,26 +0,0 @@
-xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
-
-# changetimerecorder can only get build when libgfdb is enabled
-if BUILD_GFDB
- xlator_LTLIBRARIES = changetimerecorder.la
-endif
-
-changetimerecorder_la_LDFLAGS = -module $(GF_XLATOR_DEFAULT_LDFLAGS)
-
-changetimerecorder_la_SOURCES = changetimerecorder.c \
- ctr-helper.c ctr-xlator-ctx.c
-
-changetimerecorder_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la\
- $(top_builddir)/libglusterfs/src/gfdb/libgfdb.la
-
-noinst_HEADERS = ctr-messages.h changetimerecorder.h ctr_mem_types.h \
- ctr-helper.h ctr-xlator-ctx.h
-
-AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
- -I$(top_srcdir)/libglusterfs/src/gfdb \
- -I$(top_srcdir)/rpc/xdr/src -I$(top_builddir)/rpc/xdr/src \
- -DDATADIR=\"$(localstatedir)\"
-
-AM_CFLAGS = -Wall $(GF_CFLAGS) $(SQLITE_CFLAGS)
-
-CLEANFILES =
diff --git a/xlators/features/changetimerecorder/src/changetimerecorder.c b/xlators/features/changetimerecorder/src/changetimerecorder.c
deleted file mode 100644
index e97e3ad6cfa..00000000000
--- a/xlators/features/changetimerecorder/src/changetimerecorder.c
+++ /dev/null
@@ -1,2523 +0,0 @@
-/*
- Copyright (c) 2015 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-#include <ctype.h>
-#include <sys/uio.h>
-
-#include "gfdb_sqlite3.h"
-#include "ctr-helper.h"
-#include "ctr-messages.h"
-#include "syscall.h"
-
-#include "changetimerecorder.h"
-#include "tier-ctr-interface.h"
-
-/*******************************inode forget***********************************/
-
-int
-ctr_forget (xlator_t *this, inode_t *inode)
-{
- fini_ctr_xlator_ctx (this, inode);
- return 0;
-}
-
-/************************** Look up heal **************************************/
-/*
-Problem: The CTR xlator records file meta (heat/hardlinks)
-into the data. This works fine for files which are created
-after ctr xlator is switched ON. But for files which were
-created before CTR xlator is ON, CTR xlator is not able to
-record either of the meta i.e heat or hardlinks. Thus making
-those files immune to promotions/demotions.
-
-Solution: The solution that is implemented in this patch is
-do ctr-db heal of all those pre-existent files, using named lookup.
-For this purpose we use the inode-xlator context variable option
-in gluster.
-The inode-xlator context variable for ctr xlator will have the
-following,
- a. A Lock for the context variable
- b. A hardlink list: This list represents the successful looked
- up hardlinks.
-These are the scenarios when the hardlink list is updated:
-1) Named-Lookup: Whenever a named lookup happens on a file, in the
- wind path we copy all required hardlink and inode information to
- ctr_db_record structure, which resides in the frame->local variable.
- We dont update the database in wind. During the unwind, we read the
- information from the ctr_db_record and ,
- Check if the inode context variable is created, if not we create it.
- Check if the hard link is there in the hardlink list.
- If its not there we add it to the list and send a update to the
- database using libgfdb.
- Please note: The database transaction can fail(and we ignore) as there
- already might be a record in the db. This update to the db is to heal
- if its not there.
- If its there in the list we ignore it.
-2) Inode Forget: Whenever an inode forget hits we clear the hardlink list in
- the inode context variable and delete the inode context variable.
- Please note: An inode forget may happen for two reason,
- a. when the inode is delete.
- b. the in-memory inode is evicted from the inode table due to cache limits.
-3) create: whenever a create happens we create the inode context variable and
- add the hardlink. The database updation is done as usual by ctr.
-4) link: whenever a hardlink is created for the inode, we create the inode
- context variable, if not present, and add the hardlink to the list.
-5) unlink: whenever a unlink happens we delete the hardlink from the list.
-6) mknod: same as create.
-7) rename: whenever a rename happens we update the hardlink in list. if the
- hardlink was not present for updation, we add the hardlink to the list.
-
-What is pending:
-1) This solution will only work for named lookups.
-2) We dont track afr-self-heal/dht-rebalancer traffic for healing.
-
-*/
-
-
-/* This function doesnot write anything to the db,
- * just created the local variable
- * for the frame and sets values for the ctr_db_record */
-static int
-ctr_lookup_wind(call_frame_t *frame,
- xlator_t *this,
- gf_ctr_inode_context_t *ctr_inode_cx)
-{
- int ret = -1;
- gf_ctr_private_t *_priv = NULL;
- gf_ctr_local_t *ctr_local = NULL;
-
- GF_ASSERT(frame);
- GF_ASSERT(frame->root);
- GF_ASSERT(this);
- IS_CTR_INODE_CX_SANE(ctr_inode_cx);
-
- _priv = this->private;
- GF_ASSERT (_priv);
-
- if (_priv->ctr_record_wind && ctr_inode_cx->ia_type != IA_IFDIR) {
-
- frame->local = init_ctr_local_t (this);
- if (!frame->local) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_CREATE_CTR_LOCAL_ERROR_WIND,
- "WIND: Error while creating ctr local");
- goto out;
- };
- ctr_local = frame->local;
- /*Definately no internal fops will reach here*/
- ctr_local->is_internal_fop = _gf_false;
- /*Dont record counters*/
- CTR_DB_REC(ctr_local).do_record_counters = _gf_false;
- /*Don't record time at all*/
- CTR_DB_REC(ctr_local).do_record_times = _gf_false;
-
- /* Copy gfid into db record*/
- gf_uuid_copy (CTR_DB_REC(ctr_local).gfid,
- *(ctr_inode_cx->gfid));
-
- /* Set fop_path and fop_type, required by libgfdb to make
- * decision while inserting the record */
- CTR_DB_REC(ctr_local).gfdb_fop_path = ctr_inode_cx->fop_path;
- CTR_DB_REC(ctr_local).gfdb_fop_type = ctr_inode_cx->fop_type;
-
- /* Copy hard link info*/
- gf_uuid_copy (CTR_DB_REC(ctr_local).pargfid,
- *((NEW_LINK_CX(ctr_inode_cx))->pargfid));
- strncpy (CTR_DB_REC(ctr_local).file_name,
- NEW_LINK_CX(ctr_inode_cx)->basename,
- sizeof(CTR_DB_REC(ctr_local).file_name));
-
- /* Since we are in lookup we can ignore errors while
- * Inserting in the DB, because there may be many
- * to write to the DB attempts for healing.
- * We dont want to log all failed attempts and
- * bloat the log*/
- ctr_local->gfdb_db_record.ignore_errors = _gf_true;
- }
-
- ret = 0;
-
-out:
-
- if (ret) {
- free_ctr_local (ctr_local);
- frame->local = NULL;
- }
-
- return ret;
-}
-
-
-/* This function inserts the ctr_db_record populated by ctr_lookup_wind
- * in to the db. It also destroys the frame->local created by ctr_lookup_wind */
-static int
-ctr_lookup_unwind (call_frame_t *frame,
- xlator_t *this)
-{
- int ret = -1;
- gf_ctr_private_t *_priv = NULL;
- gf_ctr_local_t *ctr_local = NULL;
-
- GF_ASSERT(frame);
- GF_ASSERT(this);
-
- _priv = this->private;
- GF_ASSERT (_priv);
-
- GF_ASSERT(_priv->_db_conn);
-
- ctr_local = frame->local;
-
- if (ctr_local && (ctr_local->ia_inode_type != IA_IFDIR)) {
-
- ret = insert_record(_priv->_db_conn,
- &ctr_local->gfdb_db_record);
- if (ret == -1) {
- gf_msg (this->name,
- _gfdb_log_level (GF_LOG_ERROR,
- ctr_local->
- gfdb_db_record.ignore_errors),
- 0, CTR_MSG_FILL_CTR_LOCAL_ERROR_UNWIND,
- "UNWIND: Error filling ctr local");
- goto out;
- }
- }
- ret = 0;
-out:
- free_ctr_local (ctr_local);
- frame->local = NULL;
- return ret;
-}
-
-/******************************************************************************
- *
- * FOPS HANDLING BELOW
- *
- * ***************************************************************************/
-
-/****************************LOOKUP********************************************/
-
-
-int32_t
-ctr_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *dict, struct iatt *postparent)
-{
- int ret = -1;
- ctr_xlator_ctx_t *ctr_xlator_ctx = NULL;
- gf_ctr_local_t *ctr_local = NULL;
- ctr_heal_ret_val_t ret_val = CTR_CTX_ERROR;
- gf_boolean_t _is_heal_needed = _gf_false;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
-
- /* if the lookup failed lookup dont do anything*/
- if (op_ret == -1) {
- gf_msg_trace (this->name, 0, "lookup failed with %s",
- strerror (op_errno));
- goto out;
- }
-
- /* Ignore directory lookups */
- if (inode->ia_type == IA_IFDIR) {
- goto out;
- }
-
- /* if frame local was not set by the ctr_lookup()
- * so dont so anything*/
- if (!frame->local) {
- goto out;
- }
-
- /* if the lookup is for dht link donot record*/
- if (dht_is_linkfile (buf, dict)) {
- gf_msg_trace (this->name, 0, "Ignoring Lookup "
- "for dht link file");
- goto out;
- }
-
- ctr_local = frame->local;
- /*Assign the proper inode type*/
- ctr_local->ia_inode_type = inode->ia_type;
-
- /* Copy gfid directly from inode */
- gf_uuid_copy (CTR_DB_REC(ctr_local).gfid, inode->gfid);
-
- /* Checking if gfid and parent gfid is valid */
- if (gf_uuid_is_null(CTR_DB_REC(ctr_local).gfid) ||
- gf_uuid_is_null(CTR_DB_REC(ctr_local).pargfid)) {
- gf_msg_trace (this->name, 0,
- "Invalid GFID");
- goto out;
- }
-
- /* if its a first entry
- * then mark the ctr_record for create
- * A create will attempt a file and a hard link created in the db*/
- ctr_xlator_ctx = get_ctr_xlator_ctx (this, inode);
- if (!ctr_xlator_ctx) {
- /* This marks inode heal */
- CTR_DB_REC(ctr_local).gfdb_fop_type = GFDB_FOP_CREATE_WRITE;
- _is_heal_needed = _gf_true;
- }
-
- /* Copy the correct gfid from resolved inode */
- gf_uuid_copy (CTR_DB_REC(ctr_local).gfid, inode->gfid);
-
- /* Add hard link to the list */
- ret_val = add_hard_link_ctx (frame, this, inode);
- if (ret_val == CTR_CTX_ERROR) {
- gf_msg_trace (this->name, 0,
- "Failed adding hardlink to list");
- goto out;
- }
- /* If inode needs healing then heal the hardlink also */
- else if (ret_val & CTR_TRY_INODE_HEAL) {
- /* This marks inode heal */
- CTR_DB_REC(ctr_local).gfdb_fop_type = GFDB_FOP_CREATE_WRITE;
- _is_heal_needed = _gf_true;
- }
- /* If hardlink needs healing */
- else if (ret_val & CTR_TRY_HARDLINK_HEAL) {
- _is_heal_needed = _gf_true;
- }
-
- /* If lookup heal needed */
- if (!_is_heal_needed)
- goto out;
-
- /* FINALLY HEAL : Inserts the ctr_db_record populated by ctr_lookup_wind
- * in to the db. It also destroys the frame->local
- * created by ctr_lookup_wind */
- ret = ctr_lookup_unwind(frame, this);
- if (ret) {
- gf_msg_trace (this->name, 0,
- "Failed healing/inserting link");
- }
-
-
-out:
- free_ctr_local ((gf_ctr_local_t *)frame->local);
- frame->local = NULL;
-
- STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
- dict, postparent);
-
- return 0;
-}
-
-
-
-int32_t
-ctr_lookup (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xdata)
-{
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
- gf_ctr_link_context_t ctr_link_cx;
- gf_ctr_link_context_t *_link_cx = &ctr_link_cx;
- int ret = -1;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, out);
-
- GF_ASSERT(frame);
- GF_ASSERT(frame->root);
-
- /* Dont handle nameless lookups*/
- if (!loc->parent || !loc->name)
- goto out;
-
- /*fill ctr link context*/
- FILL_CTR_LINK_CX(_link_cx, loc->parent->gfid, loc->name, out);
-
- /* Fill ctr inode context*/
- /* IA_IFREG : We assume its a file in the wind
- * but in the unwind we are sure what the inode is a file
- * or directory
- * gfid: we are just filling loc->gfid which is not correct.
- * In unwind we fill the correct gfid for successful lookup*/
- FILL_CTR_INODE_CONTEXT(_inode_cx, IA_IFREG,
- loc->gfid, _link_cx, NULL,
- GFDB_FOP_DENTRY_WRITE, GFDB_FOP_WIND);
-
- /* Create the frame->local and populate ctr_db_record
- * No writing to the db yet */
- ret = ctr_lookup_wind(frame, this, _inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_LINK_WIND_FAILED,
- "Failed to insert link wind");
- }
-
-out:
- STACK_WIND (frame, ctr_lookup_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, loc, xdata);
- return 0;
-}
-
-
-
-
-/****************************WRITEV********************************************/
-int32_t
-ctr_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf,
- dict_t *xdata)
-{
- int ret = -1;
-
- CTR_IS_DISABLED_THEN_GOTO (this, out);
- CTR_IF_FOP_FAILED_THEN_GOTO (this, op_ret, op_errno, out);
-
- ret = ctr_insert_unwind (frame, this,
- GFDB_FOP_INODE_WRITE, GFDB_FOP_UNWIND);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_WRITEV_UNWIND_FAILED,
- "Failed to insert writev unwind");
- }
-
-
-out:
- ctr_free_frame_local (frame);
-
- STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
-
- return 0;
-}
-
-int32_t
-ctr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count, off_t off,
- uint32_t flags,
- struct iobref *iobref, dict_t *xdata)
-{
- int ret = -1;
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, out);
-
- /*Fill ctr inode context*/
- FILL_CTR_INODE_CONTEXT(_inode_cx, fd->inode->ia_type,
- fd->inode->gfid, NULL, NULL,
- GFDB_FOP_INODE_WRITE, GFDB_FOP_WIND);
-
- /*record into the database*/
- ret = ctr_insert_wind(frame, this, _inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_WRITEV_WIND_FAILED,
- "Failed to insert writev wind");
- }
-
-out:
- STACK_WIND (frame, ctr_writev_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->writev, fd, vector, count,
- off, flags, iobref, xdata);
-
- return 0;
-}
-
-/******************************setattr*****************************************/
-
-int32_t
-ctr_setattr_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this, int32_t op_ret,
- int32_t op_errno, struct iatt *preop_stbuf,
- struct iatt *postop_stbuf, dict_t *xdata)
-{
-
- int ret = -1;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_FOP_FAILED_THEN_GOTO (this, op_ret, op_errno, out);
-
- ret = ctr_insert_unwind(frame, this,
- GFDB_FOP_INODE_WRITE, GFDB_FOP_UNWIND);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_SETATTR_UNWIND_FAILED,
- "Failed to insert setattr unwind");
- }
-
-out:
- ctr_free_frame_local (frame);
-
- STACK_UNWIND_STRICT (setattr, frame, op_ret, op_errno, preop_stbuf,
- postop_stbuf, xdata);
-
- return 0;
-}
-
-int32_t
-ctr_setattr (call_frame_t *frame,
- xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
-{
-
- int ret = -1;
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, out);
- CTR_RECORD_METADATA_HEAT_IS_DISABLED_THEN_GOTO (this, out);
-
- /*Fill ctr inode context*/
- FILL_CTR_INODE_CONTEXT(_inode_cx, loc->inode->ia_type,
- loc->inode->gfid, NULL, NULL, GFDB_FOP_INODE_WRITE,
- GFDB_FOP_WIND);
-
- /*record into the database*/
- ret = ctr_insert_wind(frame, this, _inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_SETATTR_WIND_FAILED,
- "Failed to insert setattr wind");
- }
-out:
-
- STACK_WIND (frame, ctr_setattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setattr, loc, stbuf,
- valid, xdata);
-
- return 0;
-}
-
-/*************************** fsetattr ***************************************/
-int32_t
-ctr_fsetattr_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this, int32_t op_ret,
- int32_t op_errno, struct iatt *preop_stbuf,
- struct iatt *postop_stbuf, dict_t *xdata)
-{
- int ret = -1;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_FOP_FAILED_THEN_GOTO (this, op_ret, op_errno, out);
-
- ret = ctr_insert_unwind(frame, this,
- GFDB_FOP_INODE_WRITE, GFDB_FOP_UNWIND);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_SETATTR_UNWIND_FAILED,
- "Failed to insert fsetattr unwind");
- }
-
-out:
- ctr_free_frame_local (frame);
-
- STACK_UNWIND_STRICT (fsetattr, frame, op_ret, op_errno,
- preop_stbuf, postop_stbuf, xdata);
-
- return 0;
-}
-
-
-int32_t
-ctr_fsetattr (call_frame_t *frame,
- xlator_t *this, fd_t *fd,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
-{
- int ret = -1;
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, out);
- CTR_RECORD_METADATA_HEAT_IS_DISABLED_THEN_GOTO (this, out);
-
- /*Fill ctr inode context*/
- FILL_CTR_INODE_CONTEXT(_inode_cx, fd->inode->ia_type,
- fd->inode->gfid, NULL, NULL, GFDB_FOP_INODE_WRITE,
- GFDB_FOP_WIND);
-
- /*record into the database*/
- ret = ctr_insert_wind(frame, this, _inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_SETATTR_WIND_FAILED,
- "Failed to insert fsetattr wind");
- }
-out:
- STACK_WIND (frame, ctr_fsetattr_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->fsetattr,
- fd, stbuf, valid, xdata);
-
- return 0;
-}
-/****************************fremovexattr************************************/
-
-int32_t
-ctr_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- int ret = -1;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_FOP_FAILED_THEN_GOTO (this, op_ret, op_errno, out);
-
- ret = ctr_insert_unwind(frame, this,
- GFDB_FOP_INODE_WRITE, GFDB_FOP_UNWIND);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_FREMOVEXATTR_UNWIND_FAILED,
- "Failed to insert fremovexattr unwind");
- }
-
-out:
- ctr_free_frame_local (frame);
-
- STACK_UNWIND_STRICT (fremovexattr, frame, op_ret, op_errno, xdata);
-
- return 0;
-}
-
-int32_t
-ctr_fremovexattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
-{
- int ret = -1;
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, out);
- CTR_RECORD_METADATA_HEAT_IS_DISABLED_THEN_GOTO (this, out);
-
- /*Fill ctr inode context*/
- FILL_CTR_INODE_CONTEXT(_inode_cx, fd->inode->ia_type,
- fd->inode->gfid, NULL, NULL, GFDB_FOP_INODE_WRITE,
- GFDB_FOP_WIND);
-
- /*record into the database*/
- ret = ctr_insert_wind(frame, this, _inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_FREMOVEXATTR_WIND_FAILED,
- "Failed to insert fremovexattr wind");
- }
-
-out:
- STACK_WIND (frame, ctr_fremovexattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fremovexattr,
- fd, name, xdata);
- return 0;
-}
-
-/****************************removexattr*************************************/
-
-int32_t
-ctr_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- int ret = -1;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_FOP_FAILED_THEN_GOTO (this, op_ret, op_errno, out);
- CTR_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, out);
-
-
- ret = ctr_insert_unwind(frame, this,
- GFDB_FOP_INODE_WRITE, GFDB_FOP_UNWIND);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_REMOVEXATTR_UNWIND_FAILED,
- "Failed to insert removexattr unwind");
- }
-
-out:
- ctr_free_frame_local (frame);
-
- STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
-
- return 0;
-}
-
-int32_t
-ctr_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
-{
- int ret = -1;
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, out);
- CTR_RECORD_METADATA_HEAT_IS_DISABLED_THEN_GOTO (this, out);
-
- /*Fill ctr inode context*/
- FILL_CTR_INODE_CONTEXT(_inode_cx, loc->inode->ia_type,
- loc->inode->gfid, NULL, NULL, GFDB_FOP_INODE_WRITE,
- GFDB_FOP_WIND);
-
- /*record into the database*/
- ret = ctr_insert_wind(frame, this, _inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_REMOVEXATTR_WIND_FAILED,
- "Failed to insert removexattr wind");
- }
-
-out:
- STACK_WIND (frame, ctr_removexattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->removexattr,
- loc, name, xdata);
- return 0;
-}
-
-/****************************truncate****************************************/
-
-int32_t
-ctr_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
-{
- int ret = -1;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_FOP_FAILED_THEN_GOTO (this, op_ret, op_errno, out);
-
- ret = ctr_insert_unwind(frame, this,
- GFDB_FOP_INODE_WRITE, GFDB_FOP_UNWIND);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_TRUNCATE_UNWIND_FAILED,
- "Failed to insert truncate unwind");
- }
-
-
-out:
- ctr_free_frame_local (frame);
-
- STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
-
- return 0;
-}
-
-int32_t
-ctr_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc,
- off_t offset, dict_t *xdata)
-{
- int ret = -1;
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, out);
-
- /*Fill ctr inode context*/
- FILL_CTR_INODE_CONTEXT(_inode_cx, loc->inode->ia_type,
- loc->inode->gfid, NULL, NULL, GFDB_FOP_INODE_WRITE,
- GFDB_FOP_WIND);
-
- /*record into the database*/
- ret = ctr_insert_wind(frame, this, _inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_TRUNCATE_WIND_FAILED,
- "Failed to insert truncate wind");
- }
-out:
- STACK_WIND (frame, ctr_truncate_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->truncate,
- loc, offset, xdata);
- return 0;
-}
-
-/****************************ftruncate***************************************/
-
-int32_t
-ctr_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
-{
- int ret = -1;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_FOP_FAILED_THEN_GOTO (this, op_ret, op_errno, out);
-
- ret = ctr_insert_unwind(frame, this,
- GFDB_FOP_INODE_WRITE, GFDB_FOP_UNWIND);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_FTRUNCATE_UNWIND_FAILED,
- "Failed to insert ftruncate unwind");
- }
-
-out:
- ctr_free_frame_local (frame);
-
- STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
-
- return 0;
-}
-
-int32_t
-ctr_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd,
- off_t offset, dict_t *xdata)
-{
- int ret = -1;
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, out);
-
- /*Fill ctr inode context*/
- FILL_CTR_INODE_CONTEXT(_inode_cx, fd->inode->ia_type,
- fd->inode->gfid, NULL, NULL, GFDB_FOP_INODE_WRITE,
- GFDB_FOP_WIND);
-
- /*record into the database*/
- ret = ctr_insert_wind(frame, this, _inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_FTRUNCATE_WIND_FAILED,
- "Failed to insert ftruncate wind");
- }
-
-out:
- STACK_WIND (frame, ctr_ftruncate_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->ftruncate,
- fd, offset, xdata);
- return 0;
-}
-
-/****************************rename******************************************/
-int32_t
-ctr_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent,
- dict_t *xdata)
-{
- int ret = -1;
- uint32_t remaining_links = -1;
- gf_ctr_local_t *ctr_local = NULL;
- gfdb_fop_type_t fop_type = GFDB_FOP_INVALID_OP;
- gfdb_fop_path_t fop_path = GFDB_FOP_INVALID;
-
- GF_ASSERT(frame);
- GF_ASSERT(this);
-
- CTR_IS_DISABLED_THEN_GOTO (this, out);
- CTR_IF_FOP_FAILED_THEN_GOTO (this, op_ret, op_errno, out);
-
- ret = ctr_insert_unwind (frame, this,
- GFDB_FOP_DENTRY_WRITE, GFDB_FOP_UNWIND);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_RENAME_UNWIND_FAILED,
- "Failed to insert rename unwind");
- goto out;
- }
-
- if (!xdata)
- goto out;
- /*
- *
- * Extracting GF_RESPONSE_LINK_COUNT_XDATA from POSIX Xlator
- * This is only set when we are overwriting hardlinks.
- *
- * */
- ret = dict_get_uint32 (xdata , GF_RESPONSE_LINK_COUNT_XDATA,
- &remaining_links);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_GET_CTR_RESPONSE_LINK_COUNT_XDATA_FAILED,
- "Failed to getting GF_RESPONSE_LINK_COUNT_XDATA");
- remaining_links = -1;
- goto out;
- }
-
- ctr_local = frame->local;
- if (!ctr_local) {
- gf_msg (this->name, GF_LOG_ERROR, 0, CTR_MSG_NULL_LOCAL,
- "ctr_local is NULL.");
- goto out;
- }
-
- /* This is not the only link */
- if (remaining_links > 1) {
- fop_type = GFDB_FOP_DENTRY_WRITE;
- fop_path = GFDB_FOP_UNDEL;
- }
- /* Last link that was deleted */
- else if (remaining_links == 1) {
- fop_type = GFDB_FOP_DENTRY_WRITE;
- fop_path = GFDB_FOP_UNDEL_ALL;
- } else {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_RENAME_UNWIND_FAILED,
- "Invalid link count from posix");
- goto out;
- }
-
- ret = ctr_delete_hard_link_from_db (this,
- CTR_DB_REC(ctr_local).old_gfid,
- CTR_DB_REC(ctr_local).pargfid,
- CTR_DB_REC(ctr_local).file_name,
- fop_type, fop_path);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_UNLINK_UNWIND_FAILED,
- "Failed to delete records of %s",
- CTR_DB_REC(ctr_local).old_file_name);
- }
-
-out:
- ctr_free_frame_local (frame);
-
- STACK_UNWIND_STRICT (rename, frame, op_ret, op_errno, buf,
- preoldparent, postoldparent, prenewparent,
- postnewparent,
- xdata);
-
- return 0;
-}
-
-int32_t
-ctr_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
- loc_t *newloc, dict_t *xdata)
-{
- int ret = -1;
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
- gf_ctr_link_context_t new_link_cx, old_link_cx;
- gf_ctr_link_context_t *_nlink_cx = &new_link_cx;
- gf_ctr_link_context_t *_olink_cx = &old_link_cx;
- int is_dict_created = 0;
- ctr_xlator_ctx_t *ctr_xlator_ctx = NULL;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, out);
-
- /*Fill old link context*/
- FILL_CTR_LINK_CX(_olink_cx, oldloc->pargfid, oldloc->name, out);
-
- /*Fill new link context*/
- FILL_CTR_LINK_CX(_nlink_cx, newloc->pargfid, newloc->name, out);
-
- /*Fill ctr inode context*/
- FILL_CTR_INODE_CONTEXT(_inode_cx, oldloc->inode->ia_type,
- oldloc->inode->gfid, _nlink_cx, _olink_cx,
- GFDB_FOP_DENTRY_WRITE, GFDB_FOP_WIND);
-
-
- /* If the rename is a overwrite of hardlink
- * rename ("file1", "file2")
- * file1 is hardlink for gfid say 00000000-0000-0000-0000-00000000000A
- * file2 is hardlink for gfid say 00000000-0000-0000-0000-00000000000B
- * so we are saving file2 gfid in old_gfid so that we delete entries
- * from the db during rename callback if the fop is successful
- * */
- if (newloc->inode) {
- /* This is the GFID from where the newloc hardlink will be
- * unlinked */
- _inode_cx->old_gfid = &newloc->inode->gfid;
- }
-
- /* Is a metatdata fop */
- _inode_cx->is_metadata_fop = _gf_true;
-
- /*record into the database*/
- ret = ctr_insert_wind(frame, this, _inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_RENAME_WIND_FAILED,
- "Failed to insert rename wind");
- } else {
- /* We are doing updation of hard link in inode context in wind
- * As we dont get the "inode" in the call back for rename */
- ret = update_hard_link_ctx (frame, this, oldloc->inode);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_UPDATE_HARDLINK_FAILED, "Failed "
- "updating hard link in ctr inode context");
- goto out;
- }
-
- /* If the newloc has an inode. i.e aquiring hardlink of an
- * exisitng file i.e overwritting a file.
- * */
- if (newloc->inode) {
-
- /* Getting the ctr inode context variable for
- * inode whose hardlink will be aquired during
- * the rename
- * */
- ctr_xlator_ctx = get_ctr_xlator_ctx (this,
- newloc->inode);
- if (!ctr_xlator_ctx) {
- /* Since there is no ctr inode context
- * so nothing more to do */
- ret = 0;
- goto out;
- }
-
- /* Deleting hardlink from context variable */
- ret = ctr_delete_hard_link (this, ctr_xlator_ctx,
- newloc->pargfid, newloc->name);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_DELETE_HARDLINK_FAILED,
- "Failed to delete hard link");
- goto out;
- }
-
- /* Requesting for number of hardlinks on the newloc
- * inode from POSIX.
- * */
- is_dict_created = set_posix_link_request (this, &xdata);
- if (is_dict_created == -1) {
- ret = -1;
- goto out;
- }
- }
- }
-
-out:
- STACK_WIND (frame, ctr_rename_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->rename,
- oldloc, newloc, xdata);
-
- if (is_dict_created == 1) {
- dict_unref (xdata);
- }
-
- return 0;
-}
-
-/****************************unlink******************************************/
-int32_t
-ctr_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- int ret = -1;
- uint32_t remaining_links = -1;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_FOP_FAILED_THEN_GOTO (this, op_ret, op_errno, out);
-
- if (!xdata)
- goto out;
-
- /*
- *
- * Extracting GF_RESPONSE_LINK_COUNT_XDATA from POSIX Xlator
- *
- * */
- ret = dict_get_uint32 (xdata , GF_RESPONSE_LINK_COUNT_XDATA,
- &remaining_links);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_GET_CTR_RESPONSE_LINK_COUNT_XDATA_FAILED,
- "Failed to getting GF_RESPONSE_LINK_COUNT_XDATA");
- remaining_links = -1;
- }
-
- /*This is not the only link*/
- if (remaining_links != 1) {
-
- ret = ctr_insert_unwind(frame, this, GFDB_FOP_DENTRY_WRITE,
- GFDB_FOP_UNDEL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_UNLINK_UNWIND_FAILED,
- "Failed to insert unlink unwind");
- }
- }
- /*Last link that was deleted*/
- else if (remaining_links == 1) {
-
- ret = ctr_insert_unwind(frame, this, GFDB_FOP_DENTRY_WRITE,
- GFDB_FOP_UNDEL_ALL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_UNLINK_UNWIND_FAILED,
- "Failed to insert unlink unwind");
- }
- }
-
-out:
- ctr_free_frame_local (frame);
-
- STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno, preparent,
- postparent, xdata);
-
- return 0;
-}
-
-int32_t
-ctr_unlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int xflag, dict_t *xdata)
-{
- int ret = -1;
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
- gf_ctr_link_context_t ctr_link_cx;
- gf_ctr_link_context_t *_link_cx = &ctr_link_cx;
- gf_boolean_t is_xdata_created = _gf_false;
- struct iatt dummy_stat = {0};
-
- GF_ASSERT (frame);
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
-
- /*Fill link context*/
- FILL_CTR_LINK_CX(_link_cx, loc->pargfid, loc->name, out);
-
- /*Fill ctr inode context*/
- FILL_CTR_INODE_CONTEXT(_inode_cx, loc->inode->ia_type,
- loc->inode->gfid, _link_cx, NULL,
- GFDB_FOP_DENTRY_WRITE, GFDB_FOP_WDEL);
-
- /*Internal FOP*/
- _inode_cx->is_internal_fop = is_internal_fop (frame, xdata);
-
- /* Is a metadata FOP */
- _inode_cx->is_metadata_fop = _gf_true;
-
- /* If its a internal FOP and dht link file donot record*/
- if (_inode_cx->is_internal_fop &&
- dht_is_linkfile (&dummy_stat, xdata)) {
- goto out;
- }
-
- /*record into the database*/
- ret = ctr_insert_wind(frame, this, _inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_UNLINK_UNWIND_FAILED,
- "Failed to insert unlink wind");
- } else {
- /* We are doing delete of hard link in inode context in wind
- * As we dont get the "inode" in the call back for rename */
- ret = delete_hard_link_ctx (frame, this, loc->inode);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_DELETE_HARDLINK_FAILED, "Failed "
- "deleting hard link from ctr inode context");
- }
- }
-
- /*
- *
- * Sending GF_REQUEST_LINK_COUNT_XDATA
- * to POSIX Xlator to send link count in unwind path
- *
- * */
- /*create xdata if NULL*/
- if (!xdata) {
- xdata = dict_new();
- is_xdata_created = (xdata) ? _gf_true : _gf_false;
- }
- if (!xdata) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_XDATA_NULL, "xdata is NULL :Cannot send "
- "GF_REQUEST_LINK_COUNT_XDATA to posix");
- goto out;
- }
-
- ret = dict_set_int32 (xdata, GF_REQUEST_LINK_COUNT_XDATA, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_SET_CTR_RESPONSE_LINK_COUNT_XDATA_FAILED,
- "Failed setting GF_REQUEST_LINK_COUNT_XDATA");
- if (is_xdata_created) {
- dict_unref (xdata);
- }
- goto out;
- }
-
-out:
- STACK_WIND (frame, ctr_unlink_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->unlink,
- loc, xflag, xdata);
-
- if (is_xdata_created)
- dict_unref (xdata);
-
- return 0;
-}
-
-/****************************fsync******************************************/
-int32_t
-ctr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
-{
- int ret = -1;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_FOP_FAILED_THEN_GOTO (this, op_ret, op_errno, out);
-
- ret = ctr_insert_unwind(frame, this, GFDB_FOP_INODE_WRITE,
- GFDB_FOP_UNWIND);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_FSYNC_UNWIND_FAILED,
- "Failed to insert fsync unwind");
- }
-
-out:
- ctr_free_frame_local (frame);
-
- STACK_UNWIND_STRICT (fsync, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
-
- return 0;
-}
-
-int32_t
-ctr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int32_t flags, dict_t *xdata)
-{
- int ret = -1;
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, out);
-
- /*Fill ctr inode context*/
- FILL_CTR_INODE_CONTEXT(_inode_cx, fd->inode->ia_type,
- fd->inode->gfid, NULL, NULL,
- GFDB_FOP_INODE_WRITE, GFDB_FOP_WIND);
-
- /*record into the database*/
- ret = ctr_insert_wind(frame, this, _inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_FSYNC_WIND_FAILED,
- "Failed to insert fsync wind");
- }
-
-out:
- STACK_WIND (frame, ctr_fsync_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fsync,
- fd, flags, xdata);
- return 0;
-}
-
-/****************************setxattr****************************************/
-
-int
-ctr_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- int ret = -1;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
-
- ret = ctr_insert_unwind(frame, this, GFDB_FOP_INODE_WRITE,
- GFDB_FOP_UNWIND);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_FSYNC_UNWIND_FAILED,
- "Failed to insert setxattr unwind");
- }
-
-out:
- ctr_free_frame_local (frame);
-
- STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
-
- return 0;
-}
-
-int
-ctr_setxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xattr, int flags, dict_t *xdata)
-{
- int ret = -1;
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, out);
- CTR_RECORD_METADATA_HEAT_IS_DISABLED_THEN_GOTO (this, out);
-
- /*Fill ctr inode context*/
- FILL_CTR_INODE_CONTEXT(_inode_cx, loc->inode->ia_type,
- loc->inode->gfid, NULL, NULL,
- GFDB_FOP_INODE_WRITE, GFDB_FOP_WIND);
-
- /*record into the database*/
- ret = ctr_insert_wind(frame, this, _inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_SETATTR_WIND_FAILED,
- "Failed to insert setxattr wind");
- }
-
-out:
- STACK_WIND (frame, ctr_setxattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setxattr,
- loc, xattr, flags, xdata);
- return 0;
-}
-/**************************** fsetxattr *************************************/
-int32_t
-ctr_fsetxattr_cbk (call_frame_t *frame,
- void *cookie, xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
-{
- int ret = -1;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_FOP_FAILED_THEN_GOTO (this, op_ret, op_errno, out);
-
- ret = ctr_insert_unwind(frame, this, GFDB_FOP_INODE_WRITE,
- GFDB_FOP_UNWIND);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_FSYNC_UNWIND_FAILED,
- "Failed to insert fsetxattr unwind");
- }
-
-out:
- ctr_free_frame_local (frame);
-
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
-
- return 0;
-}
-
-int32_t
-ctr_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
- int32_t flags, dict_t *xdata)
-{
- int ret = -1;
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, out);
- CTR_RECORD_METADATA_HEAT_IS_DISABLED_THEN_GOTO (this, out);
-
- /*Fill ctr inode context*/
- FILL_CTR_INODE_CONTEXT(_inode_cx, fd->inode->ia_type,
- fd->inode->gfid, NULL, NULL,
- GFDB_FOP_INODE_WRITE, GFDB_FOP_WIND);
-
- /*record into the database*/
- ret = ctr_insert_wind(frame, this, _inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_SETATTR_WIND_FAILED,
- "Failed to insert fsetxattr wind");
- }
-
-out:
- STACK_WIND (frame, ctr_fsetxattr_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->fsetxattr,
- fd, dict, flags, xdata);
- return 0;
-}
-/****************************mknod*******************************************/
-
-
-int32_t
-ctr_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- int ret = -1;
- ctr_heal_ret_val_t ret_val = CTR_CTX_ERROR;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_FOP_FAILED_THEN_GOTO (this, op_ret, op_errno, out);
-
- /* Add hard link to the list */
- ret_val = add_hard_link_ctx (frame, this, inode);
- if (ret_val == CTR_CTX_ERROR) {
- gf_msg_trace (this->name, 0, "Failed adding hard link");
- }
-
- ret = ctr_insert_unwind(frame, this, GFDB_FOP_CREATE_WRITE,
- GFDB_FOP_UNWIND);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_MKNOD_UNWIND_FAILED,
- "Failed to insert mknod unwind");
- }
-
-out:
- ctr_free_frame_local (frame);
-
- STACK_UNWIND_STRICT (mknod, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
-
- return 0;
-}
-
-
-int
-ctr_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata)
-{
- int ret = -1;
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
- gf_ctr_link_context_t ctr_link_cx;
- gf_ctr_link_context_t *_link_cx = &ctr_link_cx;
- uuid_t gfid = {0,};
- uuid_t *ptr_gfid = &gfid;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, out);
-
- GF_ASSERT(frame);
- GF_ASSERT(frame->root);
-
- /*get gfid from xdata dict*/
- ret = dict_get_gfuuid (xdata, "gfid-req", &gfid);
- if (ret) {
- gf_msg_debug (this->name, 0, "failed to get gfid from dict");
- goto out;
- }
-
- /*fill ctr link context*/
- FILL_CTR_LINK_CX (_link_cx, loc->pargfid, loc->name, out);
-
- /*Fill ctr inode context*/
- FILL_CTR_INODE_CONTEXT (_inode_cx, loc->inode->ia_type,
- *ptr_gfid, _link_cx, NULL,
- GFDB_FOP_CREATE_WRITE, GFDB_FOP_WIND);
-
- /*record into the database*/
- ret = ctr_insert_wind(frame, this, _inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_MKNOD_WIND_FAILED,
- "Failed to insert mknod wind");
- }
-
-out:
- STACK_WIND (frame, ctr_mknod_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->mknod,
- loc, mode, rdev, umask, xdata);
- return 0;
-}
-
-/****************************create******************************************/
-int
-ctr_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- fd_t *fd, inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
-{
- int ret = -1;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_FOP_FAILED_THEN_GOTO (this, op_ret, op_errno, out);
-
- ret = add_hard_link_ctx (frame, this, inode);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_ADD_HARDLINK_FAILED,
- "Failed adding hard link");
- }
-
- ret = ctr_insert_unwind(frame, this, GFDB_FOP_CREATE_WRITE,
- GFDB_FOP_UNWIND);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_CREATE_UNWIND_FAILED,
- "Failed to insert create unwind");
- }
-
-out:
- ctr_free_frame_local (frame);
-
- STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode,
- stbuf,
- preparent, postparent, xdata);
-
- return 0;
-}
-
-int
-ctr_create (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, mode_t mode,
- mode_t umask, fd_t *fd, dict_t *xdata)
-{
- int ret = -1;
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
- gf_ctr_link_context_t ctr_link_cx;
- gf_ctr_link_context_t *_link_cx = &ctr_link_cx;
- uuid_t gfid = {0,};
- uuid_t *ptr_gfid = &gfid;
- struct iatt dummy_stat = {0};
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
-
- GF_ASSERT(frame);
- GF_ASSERT(frame->root);
-
- /*Get GFID from Xdata dict*/
- ret = dict_get_gfuuid (xdata, "gfid-req", &gfid);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_GET_GFID_FROM_DICT_FAILED,
- "failed to get gfid from dict");
- goto out;
- }
-
- /*fill ctr link context*/
- FILL_CTR_LINK_CX(_link_cx, loc->pargfid, loc->name, out);
-
- /*Fill ctr inode context*/
- FILL_CTR_INODE_CONTEXT(_inode_cx, loc->inode->ia_type,
- *ptr_gfid, _link_cx, NULL,
- GFDB_FOP_CREATE_WRITE, GFDB_FOP_WIND);
-
- /*Internal FOP*/
- _inode_cx->is_internal_fop = is_internal_fop (frame, xdata);
-
- /* If its a internal FOP and dht link file donot record*/
- if (_inode_cx->is_internal_fop &&
- dht_is_linkfile (&dummy_stat, xdata)) {
- goto out;
- }
-
- /*record into the database*/
- ret = ctr_insert_wind(frame, this, &ctr_inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_CREATE_WIND_FAILED,
- "Failed to insert create wind");
- }
-out:
- STACK_WIND (frame, ctr_create_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->create,
- loc, flags, mode, umask, fd, xdata);
- return 0;
-}
-
-/****************************link********************************************/
-
-int
-ctr_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- int ret = -1;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_FOP_FAILED_THEN_GOTO (this, op_ret, op_errno, out);
-
- /* Add hard link to the list */
- ret = add_hard_link_ctx (frame, this, inode);
- if (ret) {
- gf_msg_trace (this->name, 0, "Failed adding hard link");
- }
-
- ret = ctr_insert_unwind(frame, this, GFDB_FOP_DENTRY_WRITE,
- GFDB_FOP_UNWIND);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_CREATE_UNWIND_FAILED,
- "Failed to insert create unwind");
- }
-
-out:
- ctr_free_frame_local (frame);
-
- STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, stbuf,
- preparent, postparent, xdata);
- return 0;
-}
-
-int
-ctr_link (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
-{
- int ret = -1;
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
- gf_ctr_link_context_t ctr_link_cx;
- gf_ctr_link_context_t *_link_cx = &ctr_link_cx;
- struct iatt dummy_stat = {0};
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
-
- GF_ASSERT(frame);
- GF_ASSERT(frame->root);
-
- /*fill ctr link context*/
- FILL_CTR_LINK_CX(_link_cx, newloc->pargfid, newloc->name, out);
-
- /*Fill ctr inode context*/
- FILL_CTR_INODE_CONTEXT(_inode_cx, oldloc->inode->ia_type,
- oldloc->inode->gfid, _link_cx, NULL,
- GFDB_FOP_DENTRY_WRITE, GFDB_FOP_WIND);
-
- /*Internal FOP*/
- _inode_cx->is_internal_fop = is_internal_fop (frame, xdata);
-
- /* Is a metadata fop */
- _inode_cx->is_metadata_fop = _gf_true;
-
- /* If its a internal FOP and dht link file donot record*/
- if (_inode_cx->is_internal_fop &&
- dht_is_linkfile (&dummy_stat, xdata)) {
- goto out;
- }
-
-
- /*record into the database*/
- ret = ctr_insert_wind(frame, this, _inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_LINK_WIND_FAILED,
- "Failed to insert link wind");
- }
-
-out:
- STACK_WIND (frame, ctr_link_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->link,
- oldloc, newloc, xdata);
- return 0;
-}
-
-/******************************readv*****************************************/
-int ctr_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- struct iovec *vector, int count, struct iatt *stbuf,
- struct iobref *iobref, dict_t *xdata) {
-
- int ret = -1;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_FOP_FAILED_THEN_GOTO (this, op_ret, op_errno, out);
-
- ret = ctr_insert_unwind(frame, this, GFDB_FOP_INODE_READ,
- GFDB_FOP_UNWIND);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_CREATE_UNWIND_FAILED,
- "Failed to insert create unwind");
- }
-
-out:
- ctr_free_frame_local (frame);
-
- STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector, count,
- stbuf, iobref, xdata);
- return 0;
-}
-
-
-int
-ctr_readv (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t off, uint32_t flags, dict_t *xdata)
-{
- int ret = -1;
- gf_ctr_inode_context_t ctr_inode_cx;
- gf_ctr_inode_context_t *_inode_cx = &ctr_inode_cx;
-
- CTR_IS_DISABLED_THEN_GOTO(this, out);
- CTR_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, out);
-
- /*Fill ctr inode context*/
- FILL_CTR_INODE_CONTEXT(_inode_cx, fd->inode->ia_type,
- fd->inode->gfid, NULL, NULL,
- GFDB_FOP_INODE_READ, GFDB_FOP_WIND);
-
- /*record into the database*/
- ret = ctr_insert_wind(frame, this, _inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_READV_WIND_FAILED,
- "Failed to insert readv wind");
- }
-
-out:
- STACK_WIND (frame, ctr_readv_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->readv,
- fd, size, off, flags, xdata);
- return 0;
-}
-
-/*******************************ctr_ipc****************************************/
-
-/*This is the call back function per record/file from data base*/
-static int
-ctr_db_query_callback (gfdb_query_record_t *gfdb_query_record,
- void *args) {
- int ret = -1;
- ctr_query_cbk_args_t *query_cbk_args = args;
-
- GF_VALIDATE_OR_GOTO ("ctr", query_cbk_args, out);
-
- ret = gfdb_write_query_record (query_cbk_args->query_fd,
- gfdb_query_record);
- if (ret) {
- gf_msg ("ctr", GF_LOG_ERROR, 0,
- CTR_MSG_FATAL_ERROR,
- "Failed to write to query file");
- goto out;
- }
-
- query_cbk_args->count++;
-
- ret = 0;
-out:
- return ret;
-}
-
-/* This function does all the db queries related to tiering and
- * generates/populates new/existing query file
- * inputs:
- * xlator_t *this : CTR Translator
- * void *conn_node : Database connection
- * char *query_file: the query file that needs to be updated
- * gfdb_ipc_ctr_params_t *ipc_ctr_params: the query parameters
- * Return:
- * On success 0
- * On failure -1
- * */
-int
-ctr_db_query (xlator_t *this,
- void *conn_node,
- char *query_file,
- gfdb_ipc_ctr_params_t *ipc_ctr_params)
-{
- int ret = -1;
- ctr_query_cbk_args_t query_cbk_args = {0};
-
- GF_VALIDATE_OR_GOTO ("ctr", this, out);
- GF_VALIDATE_OR_GOTO (this->name, conn_node, out);
- GF_VALIDATE_OR_GOTO (this->name, query_file, out);
- GF_VALIDATE_OR_GOTO (this->name, ipc_ctr_params, out);
-
- /*Query for eligible files from db*/
- query_cbk_args.query_fd = open (query_file,
- O_WRONLY | O_CREAT | O_APPEND,
- S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
- if (query_cbk_args.query_fd < 0) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CTR_MSG_FATAL_ERROR,
- "Failed to open query file %s", query_file);
- goto out;
- }
- if (!ipc_ctr_params->is_promote) {
- if (ipc_ctr_params->emergency_demote) {
- /* emergency demotion mode */
- ret = find_all (conn_node,
- ctr_db_query_callback,
- (void *)&query_cbk_args,
- ipc_ctr_params->query_limit);
- } else {
- if (ipc_ctr_params->write_freq_threshold == 0 &&
- ipc_ctr_params->read_freq_threshold == 0) {
- ret = find_unchanged_for_time (
- conn_node,
- ctr_db_query_callback,
- (void *)&query_cbk_args,
- &ipc_ctr_params->time_stamp);
- } else {
- ret = find_unchanged_for_time_freq (
- conn_node,
- ctr_db_query_callback,
- (void *)&query_cbk_args,
- &ipc_ctr_params->time_stamp,
- ipc_ctr_params->write_freq_threshold,
- ipc_ctr_params->read_freq_threshold,
- _gf_false);
- }
- }
- } else {
- if (ipc_ctr_params->write_freq_threshold == 0 &&
- ipc_ctr_params->read_freq_threshold == 0) {
- ret = find_recently_changed_files (
- conn_node,
- ctr_db_query_callback,
- (void *)&query_cbk_args,
- &ipc_ctr_params->time_stamp);
- } else {
- ret = find_recently_changed_files_freq (
- conn_node,
- ctr_db_query_callback,
- (void *)&query_cbk_args,
- &ipc_ctr_params->time_stamp,
- ipc_ctr_params->write_freq_threshold,
- ipc_ctr_params->read_freq_threshold,
- _gf_false);
- }
- }
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_FATAL_ERROR,
- "FATAL: query from db failed");
- goto out;
- }
-
- ret = clear_files_heat (conn_node);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_FATAL_ERROR,
- "FATAL: Failed to clear db entries");
- goto out;
- }
-
- ret = 0;
-out:
-
- if (!ret)
- ret = query_cbk_args.count;
-
- if (query_cbk_args.query_fd >= 0) {
- sys_close (query_cbk_args.query_fd);
- query_cbk_args.query_fd = -1;
- }
-
- return ret;
-}
-
-void *
-ctr_compact_thread (void *args)
-{
- int ret = -1;
- void *db_conn = NULL;
-
- xlator_t *this = NULL;
- gf_ctr_private_t *priv = NULL;
- gf_boolean_t compact_active = _gf_false;
- gf_boolean_t compact_mode_switched = _gf_false;
-
- this = (xlator_t *)args;
-
- GF_VALIDATE_OR_GOTO("ctr", this, out);
-
- priv = this->private;
-
- db_conn = priv->_db_conn;
- compact_active = priv->compact_active;
- compact_mode_switched = priv->compact_mode_switched;
-
- gf_msg ("ctr-compact", GF_LOG_INFO, 0, CTR_MSG_SET,
- "Starting compaction");
-
- ret = compact_db(db_conn, compact_active,
- compact_mode_switched);
-
- if (ret) {
- gf_msg ("ctr-compact", GF_LOG_ERROR, 0, CTR_MSG_SET,
- "Failed to perform the compaction");
- }
-
- ret = pthread_mutex_lock (&priv->compact_lock);
-
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, CTR_MSG_SET,
- "Failed to acquire lock");
- goto out;
- }
-
- /* We are done compaction on this brick. Set all flags to false */
- priv->compact_active = _gf_false;
- priv->compact_mode_switched = _gf_false;
-
- ret = pthread_mutex_unlock (&priv->compact_lock);
-
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, CTR_MSG_SET,
- "Failed to release lock");
- goto out;
- }
-
-out:
- return NULL;
-}
-
-int
-ctr_ipc_helper (xlator_t *this, dict_t *in_dict,
- dict_t *out_dict)
-{
- int ret = -1;
- char *ctr_ipc_ops = NULL;
- gf_ctr_private_t *priv = NULL;
- char *db_version = NULL;
- char *db_param_key = NULL;
- char *db_param = NULL;
- char *query_file = NULL;
- gfdb_ipc_ctr_params_t *ipc_ctr_params = NULL;
- int result = 0;
- pthread_t compact_thread;
-
- GF_VALIDATE_OR_GOTO ("ctr", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
- priv = this->private;
- GF_VALIDATE_OR_GOTO (this->name, priv->_db_conn, out);
- GF_VALIDATE_OR_GOTO (this->name, in_dict, out);
- GF_VALIDATE_OR_GOTO (this->name, out_dict, out);
-
- GET_DB_PARAM_FROM_DICT(this->name, in_dict, GFDB_IPC_CTR_KEY,
- ctr_ipc_ops, out);
-
- /*if its a db clear operation */
- if (strncmp (ctr_ipc_ops, GFDB_IPC_CTR_CLEAR_OPS,
- strlen (GFDB_IPC_CTR_CLEAR_OPS)) == 0) {
-
- ret = clear_files_heat (priv->_db_conn);
- if (ret)
- goto out;
-
- } /* if its a query operation, in which case its query + clear db*/
- else if (strncmp (ctr_ipc_ops, GFDB_IPC_CTR_QUERY_OPS,
- strlen (GFDB_IPC_CTR_QUERY_OPS)) == 0) {
-
- ret = dict_get_str (in_dict, GFDB_IPC_CTR_GET_QFILE_PATH,
- &query_file);
- if (ret) {
- gf_msg(this->name, GF_LOG_ERROR, 0, CTR_MSG_SET,
- "Failed extracting query file path");
- goto out;
- }
-
- ret = dict_get_bin (in_dict, GFDB_IPC_CTR_GET_QUERY_PARAMS,
- (void *)&ipc_ctr_params);
- if (ret) {
- gf_msg(this->name, GF_LOG_ERROR, 0, CTR_MSG_SET,
- "Failed extracting query parameters");
- goto out;
- }
-
- ret = ctr_db_query (this, priv->_db_conn, query_file,
- ipc_ctr_params);
-
- ret = dict_set_int32 (out_dict,
- GFDB_IPC_CTR_RET_QUERY_COUNT, ret);
- if (ret) {
- gf_msg(this->name, GF_LOG_ERROR, 0, CTR_MSG_SET,
- "Failed setting query reply");
- goto out;
- }
-
- } /* if its a query for db version */
- else if (strncmp (ctr_ipc_ops, GFDB_IPC_CTR_GET_DB_VERSION_OPS,
- strlen (GFDB_IPC_CTR_GET_DB_VERSION_OPS)) == 0) {
-
- ret = get_db_version (priv->_db_conn, &db_version);
- if (ret == -1 || !db_version) {
- gf_msg(this->name, GF_LOG_ERROR, 0, CTR_MSG_SET,
- "Failed extracting db version ");
- goto out;
- }
-
- SET_DB_PARAM_TO_DICT(this->name, out_dict,
- GFDB_IPC_CTR_RET_DB_VERSION,
- db_version, ret, error);
-
- } /* if its a query for a db setting */
- else if (strncmp (ctr_ipc_ops, GFDB_IPC_CTR_GET_DB_PARAM_OPS,
- strlen (GFDB_IPC_CTR_GET_DB_PARAM_OPS)) == 0) {
-
- ret = dict_get_str (in_dict, GFDB_IPC_CTR_GET_DB_KEY,
- &db_param_key);
- if (ret) {
- gf_msg(this->name, GF_LOG_ERROR, 0, CTR_MSG_SET,
- "Failed extracting db param key");
- goto out;
- }
-
- ret = get_db_params (priv->_db_conn, db_param_key, &db_param);
- if (ret == -1 || !db_param) {
- goto out;
- }
-
- SET_DB_PARAM_TO_DICT(this->name, out_dict,
- db_param_key,
- db_param, ret, error);
- } /* if its an attempt to compact the database */
- else if (strncmp (ctr_ipc_ops, GFDB_IPC_CTR_SET_COMPACT_PRAGMA,
- strlen (GFDB_IPC_CTR_SET_COMPACT_PRAGMA)) == 0) {
-
- ret = pthread_mutex_lock (&priv->compact_lock);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, CTR_MSG_SET,
- "Failed to acquire lock for compaction");
- goto out;
- }
-
- if ((priv->compact_active || priv->compact_mode_switched)) {
- /* Compaction in progress. LEAVE */
- gf_msg (this->name, GF_LOG_ERROR, 0, CTR_MSG_SET,
- "Compaction already in progress.");
- pthread_mutex_unlock (&priv->compact_lock);
- goto out;
- }
- /* At this point, we should be the only one on the brick */
- /* compacting */
-
- /* Grab the arguments from the dictionary */
- ret = dict_get_int32 (in_dict, "compact_active", &result);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, CTR_MSG_SET,
- "Failed to get compaction type");
- goto out;
- }
-
- if (result) {
- priv->compact_active = _gf_true;
- }
-
- ret = dict_get_int32 (in_dict, "compact_mode_switched"
- , &result);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, CTR_MSG_SET,
- "Failed to see if compaction switched");
- goto out;
- }
-
- if (result) {
- priv->compact_mode_switched = _gf_true;
- gf_msg ("ctr-compact", GF_LOG_TRACE, 0, CTR_MSG_SET,
- "Pre-thread: Compact mode switch is true");
- } else {
- gf_msg ("ctr-compact", GF_LOG_TRACE, 0, CTR_MSG_SET,
- "Pre-thread: Compact mode switch is false");
- }
-
- ret = pthread_mutex_unlock (&priv->compact_lock);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, CTR_MSG_SET,
- "Failed to release lock for compaction");
- goto out;
- }
-
- ret = gf_thread_create (&compact_thread, NULL,
- ctr_compact_thread, (void *)this,
- "ctrcomp");
-
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, CTR_MSG_SET,
- "Failed to spawn compaction thread");
- goto out;
- }
-
- goto out;
- } /* default case */
- else {
- goto out;
- }
-
- ret = 0;
- goto out;
-error:
- GF_FREE (db_param_key);
- GF_FREE (db_param);
- GF_FREE (db_version);
-out:
- return ret;
-}
-
-
-/* IPC Call from tier migrator to clear the heat on the DB */
-int32_t
-ctr_ipc (call_frame_t *frame, xlator_t *this, int32_t op,
- dict_t *in_dict)
-{
- int ret = -1;
- gf_ctr_private_t *priv = NULL;
- dict_t *out_dict = NULL;
-
- GF_ASSERT(this);
- priv = this->private;
- GF_ASSERT (priv);
- GF_ASSERT(priv->_db_conn);
- GF_VALIDATE_OR_GOTO (this->name, in_dict, wind);
-
-
- if (op != GF_IPC_TARGET_CTR)
- goto wind;
-
- out_dict = dict_new();
- if (!out_dict) {
- goto out;
- }
-
- ret = ctr_ipc_helper (this, in_dict, out_dict);
- if (ret) {
- gf_msg(this->name, GF_LOG_ERROR, 0, CTR_MSG_SET,
- "Failed in ctr_ipc_helper");
- }
-out:
-
- STACK_UNWIND_STRICT (ipc, frame, ret, 0, out_dict);
-
- if (out_dict)
- dict_unref(out_dict);
-
- return 0;
-
- wind:
- STACK_WIND (frame, default_ipc_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->ipc, op, in_dict);
-
-
-
- return 0;
-}
-
-
-/******************************************************************************/
-int
-reconfigure (xlator_t *this, dict_t *options)
-{
- char *temp_str = NULL;
- int ret = 0;
- gf_ctr_private_t *priv = NULL;
-
- priv = this->private;
- if (dict_get_str(options, "changetimerecorder.frequency",
- &temp_str)) {
- gf_msg(this->name, GF_LOG_TRACE, 0, CTR_MSG_SET, "set");
- }
-
- GF_OPTION_RECONF ("ctr-enabled", priv->enabled, options,
- bool, out);
-
- GF_OPTION_RECONF ("record-counters", priv->ctr_record_counter, options,
- bool, out);
-
- GF_OPTION_RECONF ("ctr-record-metadata-heat",
- priv->ctr_record_metadata_heat, options,
- bool, out);
-
- GF_OPTION_RECONF ("ctr_link_consistency", priv->ctr_link_consistency,
- options, bool, out);
-
- GF_OPTION_RECONF ("ctr_lookupheal_inode_timeout",
- priv->ctr_lookupheal_inode_timeout,
- options, uint64, out);
-
- GF_OPTION_RECONF ("ctr_lookupheal_link_timeout",
- priv->ctr_lookupheal_link_timeout,
- options, uint64, out);
-
- GF_OPTION_RECONF ("record-exit", priv->ctr_record_unwind, options,
- bool, out);
-
- GF_OPTION_RECONF ("record-entry", priv->ctr_record_wind, options,
- bool, out);
-
-
-
-
- /* If database is sqlite */
- if (priv->gfdb_db_type == GFDB_SQLITE3) {
-
- /* AUTOCHECKPOINT */
- if (dict_get_str (options, GFDB_SQL_PARAM_WAL_AUTOCHECK,
- &temp_str) == 0) {
- ret = set_db_params (priv->_db_conn,
- "wal_autocheckpoint", temp_str);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_SET_VALUE_TO_SQL_PARAM_FAILED,
- "Failed to set %s",
- GFDB_SQL_PARAM_WAL_AUTOCHECK);
- }
- }
-
- /* CACHE_SIZE */
- if (dict_get_str (options, GFDB_SQL_PARAM_CACHE_SIZE, &temp_str)
- == 0) {
- ret = set_db_params (priv->_db_conn, "cache_size",
- temp_str);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_SET_VALUE_TO_SQL_PARAM_FAILED,
- "Failed to set %s",
- GFDB_SQL_PARAM_CACHE_SIZE);
- }
- }
- }
-
- ret = 0;
-
-out:
-
- return ret;
-}
-
-/****************************init********************************************/
-
-int32_t
-init (xlator_t *this)
-{
- gf_ctr_private_t *priv = NULL;
- int ret_db = -1;
- dict_t *params_dict = NULL;
-
- GF_VALIDATE_OR_GOTO ("ctr", this, error);
-
- if (!this->children || this->children->next) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_FATAL_ERROR,
- "FATAL: ctr should have exactly one child");
- goto error;
- }
-
- if (!this->parents) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- CTR_MSG_DANGLING_VOLUME,
- "dangling volume. check volfile ");
- }
-
- priv = GF_CALLOC (1, sizeof (*priv), gf_ctr_mt_private_t);
- if (!priv) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- CTR_MSG_CALLOC_FAILED,
- "Calloc did not work!!!");
- goto error;
- }
-
- /*Default values for the translator*/
- priv->ctr_record_wind = _gf_true;
- priv->ctr_record_unwind = _gf_false;
- priv->ctr_hot_brick = _gf_false;
- priv->gfdb_db_type = GFDB_SQLITE3;
- priv->gfdb_sync_type = GFDB_DB_SYNC;
- priv->enabled = _gf_true;
- priv->_db_conn = NULL;
- priv->ctr_lookupheal_link_timeout =
- CTR_DEFAULT_HARDLINK_EXP_PERIOD;
- priv->ctr_lookupheal_inode_timeout =
- CTR_DEFAULT_INODE_EXP_PERIOD;
-
- /* For compaction */
- priv->compact_active = _gf_false;
- priv->compact_mode_switched = _gf_false;
- ret_db = pthread_mutex_init (&priv->compact_lock, NULL);
-
- if (ret_db) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_FATAL_ERROR,
- "FATAL: Failed initializing compaction mutex");
- goto error;
- }
-
- /*Extract ctr xlator options*/
- ret_db = extract_ctr_options (this, priv);
- if (ret_db) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_EXTRACT_CTR_XLATOR_OPTIONS_FAILED,
- "Failed extracting ctr xlator options");
- goto error;
- }
-
- params_dict = dict_new ();
- if (!params_dict) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INIT_DB_PARAMS_FAILED,
- "DB Params cannot initialized!");
- goto error;
- }
-
- /*Extract db params options*/
- ret_db = extract_db_params(this, params_dict, priv->gfdb_db_type);
- if (ret_db) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_EXTRACT_DB_PARAM_OPTIONS_FAILED,
- "Failed extracting db params options");
- goto error;
- }
-
- /*Create a memory pool for ctr xlator*/
- this->local_pool = mem_pool_new (gf_ctr_local_t, 64);
- if (!this->local_pool) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_CREATE_LOCAL_MEMORY_POOL_FAILED,
- "failed to create local memory pool");
- goto error;
- }
-
- /*Initialize Database Connection*/
- priv->_db_conn = init_db(params_dict, priv->gfdb_db_type);
- if (!priv->_db_conn) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_FATAL_ERROR,
- "FATAL: Failed initializing data base");
- goto error;
- }
-
-
- ret_db = 0;
- goto out;
-
-/*Error handling */
-error:
-
- if (this)
- mem_pool_destroy (this->local_pool);
-
- if (priv) {
- GF_FREE (priv->ctr_db_path);
- }
- GF_FREE (priv);
-
- if (params_dict)
- dict_unref (params_dict);
-
- return -1;
-
-out:
-
- if (params_dict)
- dict_unref (params_dict);
-
- this->private = (void *)priv;
- return 0;
-}
-
-int
-notify (xlator_t *this, int event, void *data, ...)
-{
-
- gf_ctr_private_t *priv = NULL;
- int ret = 0;
-
- priv = this->private;
-
- if (!priv)
- goto out;
-
- if (event == GF_EVENT_CLEANUP) {
- if (fini_db (priv->_db_conn)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- CTR_MSG_CLOSE_DB_CONN_FAILED, "Failed closing "
- "db connection");
- }
- if (priv->_db_conn)
- priv->_db_conn = NULL;
- }
- ret = default_notify (this, event, data);
-
-out:
- return ret;
-
-}
-
-int32_t
-mem_acct_init (xlator_t *this)
-{
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO ("ctr", this, out);
-
- ret = xlator_mem_acct_init (this, gf_ctr_mt_end + 1);
-
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_MEM_ACC_INIT_FAILED, "Memory accounting init"
- "failed");
- return ret;
- }
-out:
- return ret;
-}
-
-
-void
-fini (xlator_t *this)
-{
- gf_ctr_private_t *priv = NULL;
-
- priv = this->private;
-
- if (priv) {
- if (fini_db (priv->_db_conn)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- CTR_MSG_CLOSE_DB_CONN_FAILED, "Failed closing "
- "db connection");
- }
- GF_FREE (priv->ctr_db_path);
- if (pthread_mutex_destroy (&priv->compact_lock)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- CTR_MSG_CLOSE_DB_CONN_FAILED, "Failed to "
- "destroy the compaction mutex");
- }
- }
- GF_FREE (priv);
- mem_pool_destroy (this->local_pool);
-
- return;
-}
-
-struct xlator_fops fops = {
- /*lookup*/
- .lookup = ctr_lookup,
- /*write fops */
- .mknod = ctr_mknod,
- .create = ctr_create,
- .truncate = ctr_truncate,
- .ftruncate = ctr_ftruncate,
- .setxattr = ctr_setxattr,
- .fsetxattr = ctr_fsetxattr,
- .removexattr = ctr_removexattr,
- .fremovexattr = ctr_fremovexattr,
- .unlink = ctr_unlink,
- .link = ctr_link,
- .rename = ctr_rename,
- .writev = ctr_writev,
- .setattr = ctr_setattr,
- .fsetattr = ctr_fsetattr,
- /*read fops*/
- .readv = ctr_readv,
- /* IPC call*/
- .ipc = ctr_ipc
-};
-
-struct xlator_cbks cbks = {
- .forget = ctr_forget
-};
-
-struct volume_options options[] = {
- { .key = {"ctr-enabled",},
- .type = GF_OPTION_TYPE_BOOL,
- .value = {"on", "off"},
- .default_value = "off",
- .description = "Enables the CTR",
- .flags = OPT_FLAG_SETTABLE
- },
- { .key = {"record-entry"},
- .type = GF_OPTION_TYPE_BOOL,
- .value = {"on", "off"},
- .default_value = "on"
- },
- { .key = {"record-exit"},
- .type = GF_OPTION_TYPE_BOOL,
- .value = {"on", "off"},
- .default_value = "off"
- },
- { .key = {"record-counters"},
- .type = GF_OPTION_TYPE_BOOL,
- .value = {"on", "off"},
- .default_value = "off",
- .op_version = {GD_OP_VERSION_3_7_0},
- .flags = OPT_FLAG_SETTABLE,
- .tags = {}
- },
- { .key = {"ctr-record-metadata-heat"},
- .type = GF_OPTION_TYPE_BOOL,
- .value = {"on", "off"},
- .default_value = "off",
- .flags = OPT_FLAG_SETTABLE,
- .op_version = {GD_OP_VERSION_3_7_0},
- .tags = {}
- },
- { .key = {"ctr_link_consistency"},
- .type = GF_OPTION_TYPE_BOOL,
- .value = {"on", "off"},
- .default_value = "off",
- .flags = OPT_FLAG_SETTABLE,
- .op_version = {GD_OP_VERSION_3_7_0},
- .tags = {}
- },
- { .key = {"ctr_lookupheal_link_timeout"},
- .type = GF_OPTION_TYPE_INT,
- .default_value = "300",
- .flags = OPT_FLAG_SETTABLE,
- .op_version = {GD_OP_VERSION_3_7_2},
- .tags = {}
- },
- { .key = {"ctr_lookupheal_inode_timeout"},
- .type = GF_OPTION_TYPE_INT,
- .default_value = "300",
- .flags = OPT_FLAG_SETTABLE,
- .op_version = {GD_OP_VERSION_3_7_2},
- .tags = {}
- },
- { .key = {"hot-brick"},
- .type = GF_OPTION_TYPE_BOOL,
- .value = {"on", "off"},
- .default_value = "off"
- },
- { .key = {"db-type"},
- .type = GF_OPTION_TYPE_STR,
- .value = {"hashfile", "rocksdb", "changelog", "sqlite3",
- "hyperdex"},
- .default_value = "sqlite3",
- .op_version = {GD_OP_VERSION_3_7_0},
- .flags = OPT_FLAG_SETTABLE,
- .tags = {}
- },
- { .key = {"db-sync"},
- .type = GF_OPTION_TYPE_STR,
- .value = {"sync", "async"},
- .default_value = "sync"
- },
- { .key = {"db-path"},
- .type = GF_OPTION_TYPE_PATH
- },
- { .key = {"db-name"},
- .type = GF_OPTION_TYPE_STR
- },
- { .key = {GFDB_SQL_PARAM_SYNC},
- .type = GF_OPTION_TYPE_STR,
- .value = {"off", "normal", "full"},
- .default_value = "normal"
- },
- { .key = {GFDB_SQL_PARAM_JOURNAL_MODE},
- .type = GF_OPTION_TYPE_STR,
- .value = {"delete", "truncate", "persist", "memory", "wal", "off"},
- .default_value = "wal",
- .flags = OPT_FLAG_SETTABLE,
- .op_version = {GD_OP_VERSION_3_7_0},
- .tags = {}
- },
- { .key = {GFDB_SQL_PARAM_AUTO_VACUUM},
- .type = GF_OPTION_TYPE_STR,
- .value = {"off", "full", "incr"},
- .default_value = "off",
- .flags = OPT_FLAG_SETTABLE,
- .op_version = {GD_OP_VERSION_3_7_0},
- .tags = {}
- },
- { .key = {GFDB_SQL_PARAM_WAL_AUTOCHECK},
- .type = GF_OPTION_TYPE_INT,
- .default_value = "25000",
- .flags = OPT_FLAG_SETTABLE,
- .op_version = {GD_OP_VERSION_3_7_0},
- .tags = {}
- },
- { .key = {GFDB_SQL_PARAM_CACHE_SIZE},
- .type = GF_OPTION_TYPE_INT,
- .default_value = "12500",
- .flags = OPT_FLAG_SETTABLE,
- .op_version = {GD_OP_VERSION_3_7_0},
- .tags = {}
- },
- { .key = {GFDB_SQL_PARAM_PAGE_SIZE},
- .type = GF_OPTION_TYPE_INT,
- .default_value = "4096",
- .flags = OPT_FLAG_SETTABLE,
- .op_version = {GD_OP_VERSION_3_7_0},
- .tags = {}
- },
- { .key = {NULL} },
-};
diff --git a/xlators/features/changetimerecorder/src/changetimerecorder.h b/xlators/features/changetimerecorder/src/changetimerecorder.h
deleted file mode 100644
index 2a8bbd18c5b..00000000000
--- a/xlators/features/changetimerecorder/src/changetimerecorder.h
+++ /dev/null
@@ -1,21 +0,0 @@
-/*
- Copyright (c) 2006-2015 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#ifndef __CTR_H
-#define __CTR_H
-
-#include "glusterfs.h"
-#include "xlator.h"
-#include "logging.h"
-#include "common-utils.h"
-#include "ctr_mem_types.h"
-#include "ctr-helper.h"
-
-#endif /* __CTR_H */
diff --git a/xlators/features/changetimerecorder/src/ctr-helper.c b/xlators/features/changetimerecorder/src/ctr-helper.c
deleted file mode 100644
index 822526ca833..00000000000
--- a/xlators/features/changetimerecorder/src/ctr-helper.c
+++ /dev/null
@@ -1,307 +0,0 @@
-/*
- Copyright (c) 2015 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#include "gfdb_sqlite3.h"
-#include "ctr-helper.h"
-#include "ctr-messages.h"
-
-/*******************************************************************************
- *
- * Fill unwind into db record
- *
- ******************************************************************************/
-int
-fill_db_record_for_unwind(xlator_t *this,
- gf_ctr_local_t *ctr_local,
- gfdb_fop_type_t fop_type,
- gfdb_fop_path_t fop_path)
-{
- int ret = -1;
- gfdb_time_t *ctr_uwtime = NULL;
- gf_ctr_private_t *_priv = NULL;
-
- GF_ASSERT (this);
- _priv = this->private;
- GF_ASSERT (_priv);
-
- GF_ASSERT(ctr_local);
-
- /*If not unwind path error*/
- if (!isunwindpath(fop_path)) {
- gf_msg (this->name, GF_LOG_ERROR, 0, CTR_MSG_WRONG_FOP_PATH,
- "Wrong fop_path. Should be unwind");
- goto out;
- }
-
- ctr_uwtime = &CTR_DB_REC(ctr_local).gfdb_unwind_change_time;
- CTR_DB_REC(ctr_local).gfdb_fop_path = fop_path;
- CTR_DB_REC(ctr_local).gfdb_fop_type = fop_type;
-
- ret = gettimeofday (ctr_uwtime, NULL);
- if (ret == -1) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CTR_MSG_FILL_UNWIND_TIME_REC_ERROR, "Error "
- "filling unwind time record %s",
- strerror(errno));
- goto out;
- }
-
- /* Special case i.e if its a tier rebalance
- * + cold tier brick
- * + its a create/mknod FOP
- * we record unwind time as zero */
- if (ctr_local->client_pid == GF_CLIENT_PID_TIER_DEFRAG
- && (!_priv->ctr_hot_brick)
- && isdentrycreatefop(fop_type)) {
- memset(ctr_uwtime, 0, sizeof(*ctr_uwtime));
- }
- ret = 0;
-out:
- return ret;
-}
-
-
-/*******************************************************************************
- *
- * Fill wind into db record
- *
- ******************************************************************************/
-int
-fill_db_record_for_wind (xlator_t *this,
- gf_ctr_local_t *ctr_local,
- gf_ctr_inode_context_t *ctr_inode_cx)
-{
- int ret = -1;
- gfdb_time_t *ctr_wtime = NULL;
- gf_ctr_private_t *_priv = NULL;
-
- GF_ASSERT (this);
- _priv = this->private;
- GF_ASSERT (_priv);
- GF_ASSERT (ctr_local);
- IS_CTR_INODE_CX_SANE (ctr_inode_cx);
-
- /*if not wind path error!*/
- if (!iswindpath(ctr_inode_cx->fop_path)) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_WRONG_FOP_PATH,
- "Wrong fop_path. Should be wind");
- goto out;
- }
-
- ctr_wtime = &CTR_DB_REC(ctr_local).gfdb_wind_change_time;
- CTR_DB_REC(ctr_local).gfdb_fop_path = ctr_inode_cx->fop_path;
- CTR_DB_REC(ctr_local).gfdb_fop_type = ctr_inode_cx->fop_type;
- CTR_DB_REC(ctr_local).link_consistency = _priv->ctr_link_consistency;
-
- ret = gettimeofday (ctr_wtime, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- CTR_MSG_FILL_UNWIND_TIME_REC_ERROR,
- "Error filling wind time record %s",
- strerror(errno));
- goto out;
- }
-
- /* Special case i.e if its a tier rebalance
- * + cold tier brick
- * + its a create/mknod FOP
- * we record wind time as zero */
- if (ctr_local->client_pid == GF_CLIENT_PID_TIER_DEFRAG
- && (!_priv->ctr_hot_brick)
- && isdentrycreatefop(ctr_inode_cx->fop_type)) {
- memset(ctr_wtime, 0, sizeof(*ctr_wtime));
- }
-
- /* Copy gfid into db record */
- gf_uuid_copy (CTR_DB_REC(ctr_local).gfid, *(ctr_inode_cx->gfid));
-
- /* Copy older gfid if any */
- if (ctr_inode_cx->old_gfid &&
- (!gf_uuid_is_null (*(ctr_inode_cx->old_gfid)))) {
- gf_uuid_copy (CTR_DB_REC(ctr_local).old_gfid,
- *(ctr_inode_cx->old_gfid));
- }
-
- /*Hard Links*/
- if (isdentryfop(ctr_inode_cx->fop_type)) {
- /*new link fop*/
- if (NEW_LINK_CX(ctr_inode_cx)) {
- gf_uuid_copy (CTR_DB_REC(ctr_local).pargfid,
- *((NEW_LINK_CX(ctr_inode_cx))->pargfid));
- strcpy (CTR_DB_REC(ctr_local).file_name,
- NEW_LINK_CX(ctr_inode_cx)->basename);
- }
- /*rename fop*/
- if (OLD_LINK_CX(ctr_inode_cx)) {
- gf_uuid_copy (CTR_DB_REC(ctr_local).old_pargfid,
- *((OLD_LINK_CX(ctr_inode_cx))->pargfid));
- strcpy (CTR_DB_REC(ctr_local).old_file_name,
- OLD_LINK_CX(ctr_inode_cx)->basename);
- }
- }
-
- ret = 0;
-out:
- /*On error roll back and clean the record*/
- if (ret == -1) {
- CLEAR_CTR_DB_RECORD (ctr_local);
- }
- return ret;
-}
-
-
-/******************************************************************************
- *
- * CTR xlator init related functions
- *
- *
- * ****************************************************************************/
-static int
-extract_sql_params(xlator_t *this, dict_t *params_dict)
-{
- int ret = -1;
- char *db_path = NULL;
- char *db_name = NULL;
- char *db_full_path = NULL;
-
- GF_ASSERT (this);
- GF_ASSERT (params_dict);
-
- /*Extract the path of the db*/
- db_path = NULL;
- GET_DB_PARAM_FROM_DICT_DEFAULT(this->name, this->options, "db-path",
- db_path, "/var/run/gluster/");
-
- /*Extract the name of the db*/
- db_name = NULL;
- GET_DB_PARAM_FROM_DICT_DEFAULT(this->name, this->options, "db-name",
- db_name, "gf_ctr_db.db");
-
- /*Construct full path of the db*/
- ret = gf_asprintf(&db_full_path, "%s/%s", db_path, db_name);
- if (ret < 0) {
- gf_msg (GFDB_DATA_STORE, GF_LOG_ERROR, 0,
- CTR_MSG_CONSTRUCT_DB_PATH_FAILED,
- "Construction of full db path failed!");
- goto out;
- }
-
- /*Setting the SQL DB Path*/
- SET_DB_PARAM_TO_DICT(this->name, params_dict, GFDB_SQL_PARAM_DBPATH,
- db_full_path, ret, out);
-
- /*Extact rest of the sql params*/
- ret = gfdb_set_sql_params(this->name, this->options, params_dict);
- if (ret) {
- gf_msg (GFDB_DATA_STORE, GF_LOG_ERROR, 0,
- CTR_MSG_SET_VALUE_TO_SQL_PARAM_FAILED,
- "Failed setting values to sql param dict!");
- }
-
- ret = 0;
-
-out:
- if (ret)
- GF_FREE (db_full_path);
- return ret;
-}
-
-
-
-int extract_db_params(xlator_t *this, dict_t *params_dict,
- gfdb_db_type_t db_type) {
-
- int ret = -1;
-
- GF_ASSERT (this);
- GF_ASSERT (params_dict);
-
- switch (db_type) {
- case GFDB_SQLITE3:
- ret = extract_sql_params(this, params_dict);
- if (ret)
- goto out;
- break;
- case GFDB_ROCKS_DB:
- case GFDB_HYPERDEX:
- case GFDB_HASH_FILE_STORE:
- case GFDB_INVALID_DB:
- case GFDB_DB_END:
- goto out;
- }
- ret = 0;
-out:
- return ret;
-}
-
-int extract_ctr_options (xlator_t *this, gf_ctr_private_t *_priv) {
- int ret = -1;
- char *_val_str = NULL;
-
- GF_ASSERT (this);
- GF_ASSERT (_priv);
-
- /*Checking if the CTR Translator is enabled. By default its disabled*/
- _priv->enabled = _gf_false;
- GF_OPTION_INIT ("ctr-enabled", _priv->enabled, bool, out);
- if (!_priv->enabled) {
- gf_msg (GFDB_DATA_STORE, GF_LOG_INFO, 0,
- CTR_MSG_XLATOR_DISABLED,
- "CTR Xlator is disabled.");
- ret = 0;
- goto out;
- }
-
- /*Extract db type*/
- GF_OPTION_INIT ("db-type", _val_str, str, out);
- _priv->gfdb_db_type = gf_string2gfdbdbtype(_val_str);
-
- /*Extract flag for record on wind*/
- GF_OPTION_INIT ("record-entry", _priv->ctr_record_wind, bool, out);
-
- /*Extract flag for record on unwind*/
- GF_OPTION_INIT ("record-exit", _priv->ctr_record_unwind, bool, out);
-
- /*Extract flag for record on counters*/
- GF_OPTION_INIT ("record-counters", _priv->ctr_record_counter, bool,
- out);
-
- /* Extract flag for record metadata heat */
- GF_OPTION_INIT ("ctr-record-metadata-heat",
- _priv->ctr_record_metadata_heat, bool,
- out);
-
- /*Extract flag for link consistency*/
- GF_OPTION_INIT ("ctr_link_consistency", _priv->ctr_link_consistency,
- bool, out);
-
- /*Extract ctr_lookupheal_inode_timeout */
- GF_OPTION_INIT ("ctr_lookupheal_inode_timeout",
- _priv->ctr_lookupheal_inode_timeout,
- uint64, out);
-
- /*Extract ctr_lookupheal_link_timeout*/
- GF_OPTION_INIT ("ctr_lookupheal_link_timeout",
- _priv->ctr_lookupheal_link_timeout,
- uint64, out);
-
- /*Extract flag for hot tier brick*/
- GF_OPTION_INIT ("hot-brick", _priv->ctr_hot_brick, bool, out);
-
- /*Extract flag for sync mode*/
- GF_OPTION_INIT ("db-sync", _val_str, str, out);
- _priv->gfdb_sync_type = gf_string2gfdbdbsync(_val_str);
-
- ret = 0;
-
-out:
- return ret;
-}
diff --git a/xlators/features/changetimerecorder/src/ctr-helper.h b/xlators/features/changetimerecorder/src/ctr-helper.h
deleted file mode 100644
index 4fd4f745f4d..00000000000
--- a/xlators/features/changetimerecorder/src/ctr-helper.h
+++ /dev/null
@@ -1,927 +0,0 @@
-/*
- Copyright (c) 2015 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#ifndef __CTR_HELPER_H
-#define __CTR_HELPER_H
-
-
-#include "xlator.h"
-#include "ctr_mem_types.h"
-#include "iatt.h"
-#include "glusterfs.h"
-#include "xlator.h"
-#include "defaults.h"
-#include "logging.h"
-#include "common-utils.h"
-#include <time.h>
-#include <sys/time.h>
-#include <pthread.h>
-
-#include "gfdb_data_store.h"
-#include "ctr-xlator-ctx.h"
-#include "ctr-messages.h"
-
-#define CTR_DEFAULT_HARDLINK_EXP_PERIOD 300 /* Five mins */
-#define CTR_DEFAULT_INODE_EXP_PERIOD 300 /* Five mins */
-
-
-typedef struct ctr_query_cbk_args {
- int query_fd;
- int count;
-} ctr_query_cbk_args_t;
-
-
-/*CTR Xlator Private structure*/
-typedef struct gf_ctr_private {
- gf_boolean_t enabled;
- char *ctr_db_path;
- gf_boolean_t ctr_hot_brick;
- gf_boolean_t ctr_record_wind;
- gf_boolean_t ctr_record_unwind;
- gf_boolean_t ctr_record_counter;
- gf_boolean_t ctr_record_metadata_heat;
- gf_boolean_t ctr_link_consistency;
- gfdb_db_type_t gfdb_db_type;
- gfdb_sync_type_t gfdb_sync_type;
- gfdb_conn_node_t *_db_conn;
- uint64_t ctr_lookupheal_link_timeout;
- uint64_t ctr_lookupheal_inode_timeout;
- gf_boolean_t compact_active;
- gf_boolean_t compact_mode_switched;
- pthread_mutex_t compact_lock;
-} gf_ctr_private_t;
-
-
-/*
- * gf_ctr_local_t is the ctr xlator local data structure that is stored in
- * the call_frame of each FOP.
- *
- * gfdb_db_record: The gf_ctr_local contains a gfdb_db_record object, which is
- * used by the insert_record() api from the libgfdb. The gfdb_db_record object
- * will contain all the inode and hardlink(only for dentry fops: create,
- * mknod,link, unlink, rename).The ctr_local is keep alive till the unwind
- * call and will be release during the unwind. The same gfdb_db_record will
- * used for the unwind insert_record() api, to record unwind in the database.
- *
- * ia_inode_type in gf_ctr_local will tell the type of the inode. This is
- * important for during the unwind path. As we will not have the inode during
- * the unwind path. We would have include this in the gfdb_db_record itself
- * but currently we record only file inode information.
- *
- * is_internal_fop in gf_ctr_local will tell us if this is a internal fop and
- * take special/no action. We dont record change/acces times or increement heat
- * counter for internal fops from rebalancer.
- * */
-typedef struct gf_ctr_local {
- gfdb_db_record_t gfdb_db_record;
- ia_type_t ia_inode_type;
- gf_boolean_t is_internal_fop;
- gf_special_pid_t client_pid;
-} gf_ctr_local_t;
-/*
- * Easy access of gfdb_db_record of ctr_local
- * */
-#define CTR_DB_REC(ctr_local)\
- (ctr_local->gfdb_db_record)
-
-/*Clear db record*/
-#define CLEAR_CTR_DB_RECORD(ctr_local)\
-do {\
- ctr_local->gfdb_db_record.gfdb_fop_path = GFDB_FOP_INVALID;\
- memset(&(ctr_local->gfdb_db_record.gfdb_wind_change_time),\
- 0, sizeof(gfdb_time_t));\
- memset(&(ctr_local->gfdb_db_record.gfdb_unwind_change_time),\
- 0, sizeof(gfdb_time_t));\
- gf_uuid_clear (ctr_local->gfdb_db_record.gfid);\
- gf_uuid_clear (ctr_local->gfdb_db_record.pargfid);\
- memset(ctr_local->gfdb_db_record.file_name, 0, GF_NAME_MAX + 1);\
- memset(ctr_local->gfdb_db_record.old_file_name, 0, GF_NAME_MAX + 1);\
- ctr_local->gfdb_db_record.gfdb_fop_type = GFDB_FOP_INVALID_OP;\
- ctr_local->ia_inode_type = IA_INVAL;\
-} while (0)
-
-
-static gf_ctr_local_t *
-init_ctr_local_t (xlator_t *this) {
-
- gf_ctr_local_t *ctr_local = NULL;
-
- GF_ASSERT(this);
-
- ctr_local = mem_get0 (this->local_pool);
- if (!ctr_local) {
- gf_msg (GFDB_DATA_STORE, GF_LOG_ERROR, 0,
- CTR_MSG_CREATE_CTR_LOCAL_ERROR_WIND,
- "Error while creating ctr local");
- goto out;
- }
-
- CLEAR_CTR_DB_RECORD (ctr_local);
-out:
- return ctr_local;
-}
-
-static void
-free_ctr_local (gf_ctr_local_t *ctr_local)
-{
- if (ctr_local)
- mem_put (ctr_local);
-}
-
-
-
-/******************************************************************************
- *
- *
- * Context Carrier Structures
- *
- *
- * ****************************************************************************/
-
-/*
- * Context Carrier structures are used to carry relavent information about
- * inodes and links from the fops calls to the ctr_insert_wind.
- * These structure just have pointers to the original data and donot
- * do a deep copy of any data. This info is deep copied to
- * ctr_local->gfdb_db_record and passed to insert_record() api of libgfdb. This
- * info remains persistent for the unwind in ctr_local->gfdb_db_record
- * and once used will be destroyed.
- *
- * gf_ctr_link_context_t : Context structure for hard links
- * gf_ctr_inode_context_t : Context structure for inodes
- *
- * */
-
- /*Context Carrier Structure for hard links*/
-typedef struct gf_ctr_link_context {
- uuid_t *pargfid;
- const char *basename;
-} gf_ctr_link_context_t;
-
- /*Context Carrier Structure for inodes*/
-typedef struct gf_ctr_inode_context {
- ia_type_t ia_type;
- uuid_t *gfid;
- uuid_t *old_gfid;
- gf_ctr_link_context_t *new_link_cx;
- gf_ctr_link_context_t *old_link_cx;
- gfdb_fop_type_t fop_type;
- gfdb_fop_path_t fop_path;
- gf_boolean_t is_internal_fop;
- /* Indicating metadata fops */
- gf_boolean_t is_metadata_fop;
-} gf_ctr_inode_context_t;
-
-
-/*******************Util Macros for Context Carrier Structures*****************/
-
-/*Checks if ctr_link_cx is sane!*/
-#define IS_CTR_LINK_CX_SANE(ctr_link_cx)\
-do {\
- if (ctr_link_cx) {\
- if (ctr_link_cx->pargfid)\
- GF_ASSERT (*(ctr_link_cx->pargfid));\
- GF_ASSERT (ctr_link_cx->basename);\
- };\
-} while (0)
-
-/*Clear and fill the ctr_link_context with values*/
-#define FILL_CTR_LINK_CX(ctr_link_cx, _pargfid, _basename, label)\
-do {\
- GF_VALIDATE_OR_GOTO ("ctr", ctr_link_cx, label);\
- GF_VALIDATE_OR_GOTO ("ctr", _pargfid, label);\
- GF_VALIDATE_OR_GOTO ("ctr", _basename, label);\
- memset (ctr_link_cx, 0, sizeof (*ctr_link_cx));\
- ctr_link_cx->pargfid = &_pargfid;\
- ctr_link_cx->basename = _basename;\
-} while (0)
-
-#define NEW_LINK_CX(ctr_inode_cx)\
- ctr_inode_cx->new_link_cx\
-
-#define OLD_LINK_CX(ctr_inode_cx)\
- ctr_inode_cx->old_link_cx\
-
-/*Checks if ctr_inode_cx is sane!*/
-#define IS_CTR_INODE_CX_SANE(ctr_inode_cx)\
-do {\
- GF_ASSERT (ctr_inode_cx);\
- GF_ASSERT (ctr_inode_cx->gfid);\
- GF_ASSERT (*(ctr_inode_cx->gfid));\
- GF_ASSERT (ctr_inode_cx->fop_type != GFDB_FOP_INVALID_OP);\
- GF_ASSERT (ctr_inode_cx->fop_path != GFDB_FOP_INVALID);\
- IS_CTR_LINK_CX_SANE (NEW_LINK_CX(ctr_inode_cx));\
- IS_CTR_LINK_CX_SANE (OLD_LINK_CX(ctr_inode_cx));\
-} while (0)
-
-/*Clear and fill the ctr_inode_context with values*/
-#define FILL_CTR_INODE_CONTEXT(ctr_inode_cx,\
- _ia_type,\
- _gfid,\
- _new_link_cx,\
- _old_link_cx,\
- _fop_type,\
- _fop_path)\
-do {\
- GF_ASSERT (ctr_inode_cx);\
- GF_ASSERT (_gfid);\
- GF_ASSERT (_fop_type != GFDB_FOP_INVALID_OP);\
- GF_ASSERT (_fop_path != GFDB_FOP_INVALID);\
- memset(ctr_inode_cx, 0, sizeof(*ctr_inode_cx));\
- ctr_inode_cx->ia_type = _ia_type;\
- ctr_inode_cx->gfid = &_gfid;\
- IS_CTR_LINK_CX_SANE(NEW_LINK_CX(ctr_inode_cx));\
- if (_new_link_cx)\
- NEW_LINK_CX(ctr_inode_cx) = _new_link_cx;\
- IS_CTR_LINK_CX_SANE(OLD_LINK_CX(ctr_inode_cx));\
- if (_old_link_cx)\
- OLD_LINK_CX(ctr_inode_cx) = _old_link_cx;\
- ctr_inode_cx->fop_type = _fop_type;\
- ctr_inode_cx->fop_path = _fop_path;\
-} while (0)
-
-
-/******************************************************************************
- *
- * Util functions or macros used by
- * insert wind and insert unwind
- *
- * ****************************************************************************/
-/* Free ctr frame local */
-static inline void
-ctr_free_frame_local (call_frame_t *frame) {
- if (frame) {
- free_ctr_local ((gf_ctr_local_t *) frame->local);
- frame->local = NULL;
- }
-}
-
-/* Setting GF_REQUEST_LINK_COUNT_XDATA in dict
- * that has to be sent to POSIX Xlator to send
- * link count in unwind path.
- * return 0 for success with not creation of dict
- * return 1 for success with creation of dict
- * return -1 for failure.
- * */
-static inline int
-set_posix_link_request (xlator_t *this,
- dict_t **xdata)
-{
- int ret = -1;
- gf_boolean_t is_created = _gf_false;
-
- GF_VALIDATE_OR_GOTO ("ctr", this, out);
- GF_VALIDATE_OR_GOTO (this->name, xdata, out);
-
- /*create xdata if NULL*/
- if (!*xdata) {
- *xdata = dict_new();
- is_created = _gf_true;
- ret = 1;
- } else {
- ret = 0;
- }
-
- if (!*xdata) {
- gf_msg (this->name, GF_LOG_ERROR, 0, CTR_MSG_XDATA_NULL,
- "xdata is NULL :Cannot send "
- "GF_REQUEST_LINK_COUNT_XDATA to posix");
- ret = -1;
- goto out;
- }
-
- ret = dict_set_int32 (*xdata, GF_REQUEST_LINK_COUNT_XDATA, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_SET_CTR_RESPONSE_LINK_COUNT_XDATA_FAILED,
- "Failed setting GF_REQUEST_LINK_COUNT_XDATA");
- ret = -1;
- goto out;
- }
- ret = 0;
-out:
- if (ret == -1) {
- if (*xdata && is_created) {
- dict_unref (*xdata);
- }
- }
- return ret;
-}
-
-
-/*
- * If a bitrot fop
- * */
-#define BITROT_FOP(frame)\
- (frame->root->pid == GF_CLIENT_PID_BITD ||\
- frame->root->pid == GF_CLIENT_PID_SCRUB)
-
-
-/*
- * If a rebalancer fop
- * */
-#define REBALANCE_FOP(frame)\
- (frame->root->pid == GF_CLIENT_PID_DEFRAG)
-
-/*
- * If its a tiering rebalancer fop
- * */
-#define TIER_REBALANCE_FOP(frame)\
- (frame->root->pid == GF_CLIENT_PID_TIER_DEFRAG)
-
-/*
- * If its a AFR SELF HEAL
- * */
- #define AFR_SELF_HEAL_FOP(frame)\
- (frame->root->pid == GF_CLIENT_PID_SELF_HEALD)
-
-/*
- * if a rebalancer fop goto
- * */
-#define CTR_IF_REBALANCE_FOP_THEN_GOTO(frame, label)\
-do {\
- if (REBALANCE_FOP (frame))\
- goto label;\
-} while (0)
-
-/*
- * Internal fop
- *
- * */
-static inline gf_boolean_t
-is_internal_fop (call_frame_t *frame,
- dict_t *xdata)
-{
- gf_boolean_t ret = _gf_false;
-
- GF_ASSERT(frame);
- GF_ASSERT(frame->root);
-
- if (AFR_SELF_HEAL_FOP (frame)) {
- ret = _gf_true;
- }
- if (BITROT_FOP (frame)) {
- ret = _gf_true;
- }
- if (REBALANCE_FOP (frame) || TIER_REBALANCE_FOP (frame)) {
- ret = _gf_true;
- if (xdata && dict_get (xdata, CTR_ATTACH_TIER_LOOKUP)) {
- ret = _gf_false;
- }
- }
- if (xdata && dict_get (xdata, GLUSTERFS_INTERNAL_FOP_KEY)) {
- ret = _gf_true;
- }
-
- return ret;
-}
-
-#define CTR_IF_INTERNAL_FOP_THEN_GOTO(frame, dict, label)\
-do {\
- if (is_internal_fop (frame, dict)) \
- goto label; \
-} while (0)
-
-/* if fop has failed exit */
-#define CTR_IF_FOP_FAILED_THEN_GOTO(this, op_ret, op_errno, label)\
-do {\
- if (op_ret == -1) {\
- gf_msg_trace (this->name, 0, "Failed fop with %s",\
- strerror (op_errno));\
- goto label;\
- };\
-} while (0)
-
-/*
- * IS CTR Xlator is disabled then goto to label
- * */
- #define CTR_IS_DISABLED_THEN_GOTO(this, label)\
- do {\
- gf_ctr_private_t *_priv = NULL;\
- GF_ASSERT (this);\
- GF_ASSERT (this->private);\
- _priv = this->private;\
- if (!_priv->enabled)\
- goto label;\
- } while (0)
-
-/*
- * IS CTR record metadata heat is disabled then goto to label
- * */
- #define CTR_RECORD_METADATA_HEAT_IS_DISABLED_THEN_GOTO(this, label)\
- do {\
- gf_ctr_private_t *_priv = NULL;\
- GF_ASSERT (this);\
- GF_ASSERT (this->private);\
- _priv = this->private;\
- if (!_priv->ctr_record_metadata_heat)\
- goto label;\
- } while (0)
-
-int
-fill_db_record_for_unwind (xlator_t *this,
- gf_ctr_local_t *ctr_local,
- gfdb_fop_type_t fop_type,
- gfdb_fop_path_t fop_path);
-
-int
-fill_db_record_for_wind (xlator_t *this,
- gf_ctr_local_t *ctr_local,
- gf_ctr_inode_context_t *ctr_inode_cx);
-
-/*******************************************************************************
- * CTR INSERT WIND
- * *****************************************************************************
- * Function used to insert/update record into the database during a wind fop
- * This function creates ctr_local structure into the frame of the fop
- * call.
- * ****************************************************************************/
-
-static inline int
-ctr_insert_wind (call_frame_t *frame,
- xlator_t *this,
- gf_ctr_inode_context_t *ctr_inode_cx)
-{
- int ret = -1;
- gf_ctr_private_t *_priv = NULL;
- gf_ctr_local_t *ctr_local = NULL;
-
- GF_ASSERT(frame);
- GF_ASSERT(frame->root);
- GF_ASSERT(this);
- IS_CTR_INODE_CX_SANE(ctr_inode_cx);
-
- _priv = this->private;
- GF_ASSERT (_priv);
-
- GF_ASSERT(_priv->_db_conn);
-
- /*If record_wind option of CTR is on record wind for
- * regular files only*/
- if (_priv->ctr_record_wind && ctr_inode_cx->ia_type != IA_IFDIR) {
- frame->local = init_ctr_local_t (this);
- if (!frame->local) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_CREATE_CTR_LOCAL_ERROR_WIND,
- "WIND: Error while creating ctr local");
- goto out;
- };
- ctr_local = frame->local;
- ctr_local->client_pid = frame->root->pid;
- ctr_local->is_internal_fop = ctr_inode_cx->is_internal_fop;
-
- /* Decide whether to record counters or not */
- CTR_DB_REC(ctr_local).do_record_counters = _gf_false;
- /* If record counter is enabled */
- if (_priv->ctr_record_counter) {
- /* If not a internal fop */
- if (!(ctr_local->is_internal_fop)) {
- /* If its a metadata fop AND
- * record metadata heat
- * OR
- * its NOT a metadata fop */
- if ((ctr_inode_cx->is_metadata_fop
- && _priv->ctr_record_metadata_heat)
- ||
- (!ctr_inode_cx->is_metadata_fop)) {
- CTR_DB_REC(ctr_local).do_record_counters
- = _gf_true;
- }
- }
- }
-
- /* Decide whether to record times or not
- * For non internal FOPS record times as usual*/
- CTR_DB_REC(ctr_local).do_record_times = _gf_false;
- if (!ctr_local->is_internal_fop) {
- /* If its a metadata fop AND
- * record metadata heat
- * OR
- * its NOT a metadata fop */
- if ((ctr_inode_cx->is_metadata_fop &&
- _priv->ctr_record_metadata_heat)
- ||
- (!ctr_inode_cx->is_metadata_fop)) {
- CTR_DB_REC(ctr_local).do_record_times =
- (_priv->ctr_record_wind
- || _priv->ctr_record_unwind);
- }
- }
- /* when its a internal FOPS*/
- else {
- /* Record times only for create
- * i.e when the inode is created */
- CTR_DB_REC(ctr_local).do_record_times =
- (isdentrycreatefop(ctr_inode_cx->fop_type)) ?
- _gf_true : _gf_false;
- }
-
- /*Fill the db record for insertion*/
- ret = fill_db_record_for_wind (this, ctr_local, ctr_inode_cx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_FILL_CTR_LOCAL_ERROR_WIND,
- "WIND: Error filling ctr local");
- goto out;
- }
-
- /*Insert the db record*/
- ret = insert_record (_priv->_db_conn,
- &ctr_local->gfdb_db_record);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_RECORD_WIND_FAILED,
- "WIND: Inserting of record failed!");
- goto out;
- }
- }
- ret = 0;
-out:
-
- if (ret) {
- free_ctr_local (ctr_local);
- frame->local = NULL;
- }
-
- return ret;
-}
-
-
-
-
-/*******************************************************************************
- * CTR INSERT UNWIND
- * *****************************************************************************
- * Function used to insert/update record into the database during a unwind fop
- * This function destroys ctr_local structure into the frame of the fop
- * call at the end.
- * ****************************************************************************/
-static inline int
-ctr_insert_unwind (call_frame_t *frame,
- xlator_t *this,
- gfdb_fop_type_t fop_type,
- gfdb_fop_path_t fop_path)
-{
- int ret = -1;
- gf_ctr_private_t *_priv = NULL;
- gf_ctr_local_t *ctr_local = NULL;
-
- GF_ASSERT(frame);
- GF_ASSERT(this);
-
- _priv = this->private;
- GF_ASSERT (_priv);
-
- GF_ASSERT(_priv->_db_conn);
-
- ctr_local = frame->local;
-
- if (ctr_local
- && (_priv->ctr_record_unwind || isdentryfop(fop_type))
- && (ctr_local->ia_inode_type != IA_IFDIR)) {
-
- CTR_DB_REC(ctr_local).do_record_uwind_time =
- _priv->ctr_record_unwind;
-
- ret = fill_db_record_for_unwind(this, ctr_local, fop_type,
- fop_path);
- if (ret == -1) {
- gf_msg(this->name, GF_LOG_ERROR, 0,
- CTR_MSG_FILL_CTR_LOCAL_ERROR_UNWIND,
- "UNWIND: Error filling ctr local");
- goto out;
- }
-
- ret = insert_record(_priv->_db_conn,
- &ctr_local->gfdb_db_record);
- if (ret == -1) {
- gf_msg(this->name, GF_LOG_ERROR, 0,
- CTR_MSG_FILL_CTR_LOCAL_ERROR_UNWIND,
- "UNWIND: Error filling ctr local");
- goto out;
- }
- }
- ret = 0;
-out:
- return ret;
-}
-
-/******************************************************************************
- * Delete file/flink record/s from db
- * ****************************************************************************/
-static inline int
-ctr_delete_hard_link_from_db (xlator_t *this,
- uuid_t gfid,
- uuid_t pargfid,
- char *basename,
- gfdb_fop_type_t fop_type,
- gfdb_fop_path_t fop_path)
-{
- int ret = -1;
- gfdb_db_record_t gfdb_db_record;
- gf_ctr_private_t *_priv = NULL;
-
- _priv = this->private;
- GF_VALIDATE_OR_GOTO (this->name, _priv, out);
- GF_VALIDATE_OR_GOTO (this->name, (!gf_uuid_is_null (gfid)), out);
- GF_VALIDATE_OR_GOTO (this->name, (!gf_uuid_is_null (pargfid)), out);
- GF_VALIDATE_OR_GOTO (this->name, (fop_type == GFDB_FOP_DENTRY_WRITE),
- out);
- GF_VALIDATE_OR_GOTO (this->name,
- (fop_path == GFDB_FOP_UNDEL || GFDB_FOP_UNDEL_ALL),
- out);
-
- /* Set gfdb_db_record to 0 */
- memset (&gfdb_db_record, 0, sizeof(gfdb_db_record));
-
- /* Copy gfid into db record */
- gf_uuid_copy (gfdb_db_record.gfid, gfid);
-
- /* Copy pargid into db record */
- gf_uuid_copy (gfdb_db_record.pargfid, pargfid);
-
- /* Copy basename */
- strncpy (gfdb_db_record.file_name, basename, GF_NAME_MAX - 1);
-
- gfdb_db_record.gfdb_fop_path = fop_path;
- gfdb_db_record.gfdb_fop_type = fop_type;
-
- /*send delete request to db*/
- ret = insert_record (_priv->_db_conn, &gfdb_db_record);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_INSERT_RECORD_WIND_FAILED,
- "Failed to delete record. %s", basename);
- goto out;
- }
-
- ret = 0;
-out:
- return ret;
-}
-
-/******************************* Hard link function ***************************/
-
-static inline gf_boolean_t
-__is_inode_expired (ctr_xlator_ctx_t *ctr_xlator_ctx,
- gf_ctr_private_t *_priv,
- gfdb_time_t *current_time)
-{
- gf_boolean_t ret = _gf_false;
- uint64_t time_diff = 0;
-
- GF_ASSERT (ctr_xlator_ctx);
- GF_ASSERT (_priv);
- GF_ASSERT (current_time);
-
- time_diff = current_time->tv_sec -
- ctr_xlator_ctx->inode_heal_period;
-
- ret = (time_diff >= _priv->ctr_lookupheal_inode_timeout) ?
- _gf_true : _gf_false;
- return ret;
-}
-
-static inline gf_boolean_t
-__is_hardlink_expired (ctr_hard_link_t *ctr_hard_link,
- gf_ctr_private_t *_priv,
- gfdb_time_t *current_time)
-{
- gf_boolean_t ret = _gf_false;
- uint64_t time_diff = 0;
-
- GF_ASSERT (ctr_hard_link);
- GF_ASSERT (_priv);
- GF_ASSERT (current_time);
-
- time_diff = current_time->tv_sec -
- ctr_hard_link->hardlink_heal_period;
-
- ret = ret || (time_diff >= _priv->ctr_lookupheal_link_timeout) ?
- _gf_true : _gf_false;
-
- return ret;
-}
-
-
-/* Return values of heal*/
-typedef enum ctr_heal_ret_val {
- CTR_CTX_ERROR = -1,
- /* No healing required */
- CTR_TRY_NO_HEAL = 0,
- /* Try healing hard link */
- CTR_TRY_HARDLINK_HEAL = 1,
- /* Try healing inode */
- CTR_TRY_INODE_HEAL = 2,
-} ctr_heal_ret_val_t;
-
-
-
-/**
- * @brief Function to add hard link to the inode context variable.
- * The inode context maintainences a in-memory list. This is used
- * smart healing of database.
- * @param frame of the FOP
- * @param this is the Xlator instant
- * @param inode
- * @return Return ctr_heal_ret_val_t
- */
-
-static inline ctr_heal_ret_val_t
-add_hard_link_ctx (call_frame_t *frame,
- xlator_t *this,
- inode_t *inode)
-{
- ctr_heal_ret_val_t ret_val = CTR_TRY_NO_HEAL;
- int ret = -1;
- gf_ctr_local_t *ctr_local = NULL;
- ctr_xlator_ctx_t *ctr_xlator_ctx = NULL;
- ctr_hard_link_t *ctr_hard_link = NULL;
- gf_ctr_private_t *_priv = NULL;
- gfdb_time_t current_time = {0};
-
-
- GF_ASSERT (frame);
- GF_ASSERT (this);
- GF_ASSERT (inode);
- GF_ASSERT (this->private);
-
- _priv = this->private;
-
- ctr_local = frame->local;
- if (!ctr_local) {
- goto out;
- }
-
- ctr_xlator_ctx = init_ctr_xlator_ctx (this, inode);
- if (!ctr_xlator_ctx) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_ACCESS_CTR_INODE_CONTEXT_FAILED,
- "Failed accessing ctr inode context");
- goto out;
- }
-
- LOCK (&ctr_xlator_ctx->lock);
-
- /* Check if the hard link already exists
- * in the ctr inode context*/
- ctr_hard_link = ctr_search_hard_link_ctx (this,
- ctr_xlator_ctx,
- CTR_DB_REC(ctr_local).pargfid,
- CTR_DB_REC(ctr_local).file_name);
- /* if there then ignore */
- if (ctr_hard_link) {
-
- ret = gettimeofday (&current_time, NULL);
- if (ret == -1) {
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to get current time");
- ret_val = CTR_CTX_ERROR;
- goto unlock;
- }
-
- if (__is_hardlink_expired (ctr_hard_link,
- _priv, &current_time)) {
- ctr_hard_link->hardlink_heal_period =
- current_time.tv_sec;
- ret_val = ret_val | CTR_TRY_HARDLINK_HEAL;
- }
-
- if (__is_inode_expired (ctr_xlator_ctx,
- _priv, &current_time)) {
- ctr_xlator_ctx->inode_heal_period =
- current_time.tv_sec;
- ret_val = ret_val | CTR_TRY_INODE_HEAL;
- }
-
- goto unlock;
- }
-
- /* Add the hard link to the list*/
- ret = ctr_add_hard_link (this, ctr_xlator_ctx,
- CTR_DB_REC(ctr_local).pargfid,
- CTR_DB_REC(ctr_local).file_name);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_ADD_HARDLINK_TO_CTR_INODE_CONTEXT_FAILED,
- "Failed to add hardlink to the ctr inode context");
- ret_val = CTR_CTX_ERROR;
- goto unlock;
- }
-
- ret_val = CTR_TRY_NO_HEAL;
-unlock:
- UNLOCK (&ctr_xlator_ctx->lock);
-out:
- return ret_val;
-}
-
-static inline int
-delete_hard_link_ctx (call_frame_t *frame,
- xlator_t *this,
- inode_t *inode)
-{
- int ret = -1;
- ctr_xlator_ctx_t *ctr_xlator_ctx = NULL;
- gf_ctr_local_t *ctr_local = NULL;
-
- GF_ASSERT (frame);
- GF_ASSERT (this);
- GF_ASSERT (inode);
-
- ctr_local = frame->local;
- if (!ctr_local) {
- goto out;
- }
-
- ctr_xlator_ctx = get_ctr_xlator_ctx (this, inode);
- if (!ctr_xlator_ctx) {
- /* Since there is no ctr inode context so nothing more to do */
- ret = 0;
- goto out;
- }
-
- ret = ctr_delete_hard_link (this, ctr_xlator_ctx,
- CTR_DB_REC(ctr_local).pargfid,
- CTR_DB_REC(ctr_local).file_name);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_DELETE_HARDLINK_FAILED,
- "Failed to delete hard link");
- goto out;
- }
-
- ret = 0;
-
-out:
- return ret;
-}
-
-static inline int
-update_hard_link_ctx (call_frame_t *frame,
- xlator_t *this,
- inode_t *inode)
-{
- int ret = -1;
- ctr_xlator_ctx_t *ctr_xlator_ctx = NULL;
- gf_ctr_local_t *ctr_local = NULL;
-
- GF_ASSERT (frame);
- GF_ASSERT (this);
- GF_ASSERT (inode);
-
- ctr_local = frame->local;
- if (!ctr_local) {
- goto out;
- }
-
- ctr_xlator_ctx = init_ctr_xlator_ctx (this, inode);
- if (!ctr_xlator_ctx) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_ACCESS_CTR_INODE_CONTEXT_FAILED,
- "Failed accessing ctr inode context");
- goto out;
- }
-
- ret = ctr_update_hard_link (this, ctr_xlator_ctx,
- CTR_DB_REC(ctr_local).pargfid,
- CTR_DB_REC(ctr_local).file_name,
- CTR_DB_REC(ctr_local).old_pargfid,
- CTR_DB_REC(ctr_local).old_file_name);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_DELETE_HARDLINK_FAILED,
- "Failed to delete hard link");
- goto out;
- }
-
- ret = 0;
-
-out:
- return ret;
-}
-
-
-/******************************************************************************
- *
- * CTR xlator init related functions
- *
- *
- * ****************************************************************************/
-int
-extract_db_params (xlator_t *this,
- dict_t *params_dict,
- gfdb_db_type_t db_type);
-
-int
-extract_ctr_options (xlator_t *this,
- gf_ctr_private_t *_priv);
-
-#endif
diff --git a/xlators/features/changetimerecorder/src/ctr-messages.h b/xlators/features/changetimerecorder/src/ctr-messages.h
deleted file mode 100644
index bc9a9e0f3ab..00000000000
--- a/xlators/features/changetimerecorder/src/ctr-messages.h
+++ /dev/null
@@ -1,86 +0,0 @@
-/*
- Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
- */
-
-#ifndef _CTR_MESSAGES_H_
-#define _CTR_MESSAGES_H_
-
-#include "glfs-message-id.h"
-
-/* To add new message IDs, append new identifiers at the end of the list.
- *
- * Never remove a message ID. If it's not used anymore, you can rename it or
- * leave it as it is, but not delete it. This is to prevent reutilization of
- * IDs by other messages.
- *
- * The component name must match one of the entries defined in
- * glfs-message-id.h.
- */
-
-GLFS_MSGID(CTR,
- CTR_MSG_CREATE_CTR_LOCAL_ERROR_WIND,
- CTR_MSG_FILL_CTR_LOCAL_ERROR_UNWIND,
- CTR_MSG_FILL_CTR_LOCAL_ERROR_WIND,
- CTR_MSG_INSERT_LINK_WIND_FAILED,
- CTR_MSG_INSERT_WRITEV_WIND_FAILED,
- CTR_MSG_INSERT_WRITEV_UNWIND_FAILED,
- CTR_MSG_INSERT_SETATTR_WIND_FAILED,
- CTR_MSG_INSERT_SETATTR_UNWIND_FAILED,
- CTR_MSG_INSERT_FREMOVEXATTR_UNWIND_FAILED,
- CTR_MSG_INSERT_FREMOVEXATTR_WIND_FAILED,
- CTR_MSG_INSERT_REMOVEXATTR_WIND_FAILED,
- CTR_MSG_INSERT_REMOVEXATTR_UNWIND_FAILED,
- CTR_MSG_INSERT_TRUNCATE_WIND_FAILED,
- CTR_MSG_INSERT_TRUNCATE_UNWIND_FAILED,
- CTR_MSG_INSERT_FTRUNCATE_UNWIND_FAILED,
- CTR_MSG_INSERT_FTRUNCATE_WIND_FAILED,
- CTR_MSG_INSERT_RENAME_WIND_FAILED,
- CTR_MSG_INSERT_RENAME_UNWIND_FAILED,
- CTR_MSG_ACCESS_CTR_INODE_CONTEXT_FAILED,
- CTR_MSG_ADD_HARDLINK_FAILED,
- CTR_MSG_DELETE_HARDLINK_FAILED,
- CTR_MSG_UPDATE_HARDLINK_FAILED,
- CTR_MSG_GET_CTR_RESPONSE_LINK_COUNT_XDATA_FAILED,
- CTR_MSG_SET_CTR_RESPONSE_LINK_COUNT_XDATA_FAILED,
- CTR_MSG_INSERT_UNLINK_UNWIND_FAILED,
- CTR_MSG_INSERT_UNLINK_WIND_FAILED,
- CTR_MSG_XDATA_NULL,
- CTR_MSG_INSERT_FSYNC_WIND_FAILED,
- CTR_MSG_INSERT_FSYNC_UNWIND_FAILED,
- CTR_MSG_INSERT_MKNOD_UNWIND_FAILED,
- CTR_MSG_INSERT_MKNOD_WIND_FAILED,
- CTR_MSG_INSERT_CREATE_WIND_FAILED,
- CTR_MSG_INSERT_CREATE_UNWIND_FAILED,
- CTR_MSG_INSERT_RECORD_WIND_FAILED,
- CTR_MSG_INSERT_READV_WIND_FAILED,
- CTR_MSG_GET_GFID_FROM_DICT_FAILED,
- CTR_MSG_SET,
- CTR_MSG_FATAL_ERROR,
- CTR_MSG_DANGLING_VOLUME,
- CTR_MSG_CALLOC_FAILED,
- CTR_MSG_EXTRACT_CTR_XLATOR_OPTIONS_FAILED,
- CTR_MSG_INIT_DB_PARAMS_FAILED,
- CTR_MSG_CREATE_LOCAL_MEMORY_POOL_FAILED,
- CTR_MSG_MEM_ACC_INIT_FAILED,
- CTR_MSG_CLOSE_DB_CONN_FAILED,
- CTR_MSG_FILL_UNWIND_TIME_REC_ERROR,
- CTR_MSG_WRONG_FOP_PATH,
- CTR_MSG_CONSTRUCT_DB_PATH_FAILED,
- CTR_MSG_SET_VALUE_TO_SQL_PARAM_FAILED,
- CTR_MSG_XLATOR_DISABLED,
- CTR_MSG_HARDLINK_MISSING_IN_LIST,
- CTR_MSG_ADD_HARDLINK_TO_LIST_FAILED,
- CTR_MSG_INIT_LOCK_FAILED,
- CTR_MSG_COPY_FAILED,
- CTR_MSG_EXTRACT_DB_PARAM_OPTIONS_FAILED,
- CTR_MSG_ADD_HARDLINK_TO_CTR_INODE_CONTEXT_FAILED,
- CTR_MSG_NULL_LOCAL
-);
-
-#endif /* !_CTR_MESSAGES_H_ */
diff --git a/xlators/features/changetimerecorder/src/ctr-xlator-ctx.c b/xlators/features/changetimerecorder/src/ctr-xlator-ctx.c
deleted file mode 100644
index 7700ad40ba6..00000000000
--- a/xlators/features/changetimerecorder/src/ctr-xlator-ctx.c
+++ /dev/null
@@ -1,409 +0,0 @@
-/*
- Copyright (c) 2015 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#include "ctr-xlator-ctx.h"
-#include "ctr-messages.h"
-#include <time.h>
-#include <sys/time.h>
-
-#define IS_THE_ONLY_HARDLINK(ctr_hard_link)\
- (ctr_hard_link->list.next == ctr_hard_link->list.prev)
-
-
-static void
-fini_ctr_hard_link (ctr_hard_link_t **ctr_hard_link) {
-
- GF_ASSERT (ctr_hard_link);
-
- if (*ctr_hard_link)
- return;
- GF_FREE ((*ctr_hard_link)->base_name);
- GF_FREE (*ctr_hard_link);
- *ctr_hard_link = NULL;
-}
-
-
-/* Please lock the ctr_xlator_ctx before using this function */
-ctr_hard_link_t *
-ctr_search_hard_link_ctx (xlator_t *this,
- ctr_xlator_ctx_t *ctr_xlator_ctx,
- uuid_t pgfid,
- const char *base_name)
-{
- ctr_hard_link_t *_hard_link = NULL;
- ctr_hard_link_t *searched_hardlink = NULL;
-
- GF_ASSERT (this);
- GF_ASSERT (ctr_xlator_ctx);
-
- if (pgfid == NULL || base_name == NULL)
- goto out;
-
- /*linear search*/
- list_for_each_entry (_hard_link, &ctr_xlator_ctx->hardlink_list, list) {
- if (gf_uuid_compare (_hard_link->pgfid, pgfid) == 0
- && _hard_link->base_name
- && strcmp(_hard_link->base_name, base_name) == 0) {
- searched_hardlink = _hard_link;
- break;
- }
- }
-
-out:
- return searched_hardlink;
-}
-
-
-
-
-/* Please lock the ctr_xlator_ctx before using this function */
-int
-ctr_add_hard_link (xlator_t *this,
- ctr_xlator_ctx_t *ctr_xlator_ctx,
- uuid_t pgfid,
- const char *base_name)
-{
- int ret = -1;
- ctr_hard_link_t *ctr_hard_link = NULL;
- struct timeval current_time = {0};
-
- GF_ASSERT (this);
- GF_ASSERT (ctr_xlator_ctx);
-
- if (pgfid == NULL || base_name == NULL)
- goto out;
-
- ctr_hard_link = GF_CALLOC (1, sizeof (*ctr_hard_link),
- gf_ctr_mt_hard_link_t);
- if (!ctr_hard_link) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- CTR_MSG_CALLOC_FAILED, "Failed allocating "
- "ctr_hard_link");
- goto out;
- }
-
- /*Initialize the ctr_hard_link object and
- * Assign the values : parent GFID and basename*/
- INIT_LIST_HEAD (&ctr_hard_link->list);
- gf_uuid_copy (ctr_hard_link->pgfid, pgfid);
- ret = gf_asprintf(&ctr_hard_link->base_name, "%s", base_name);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_COPY_FAILED, "Failed copying basename"
- "to ctr_hard_link");
- goto error;
- }
-
- ret = gettimeofday (&current_time, NULL);
- if (ret == -1) {
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to get current time");
- goto error;
- }
-
- /*Add the hard link to the list*/
- list_add_tail (&ctr_hard_link->list,
- &ctr_xlator_ctx->hardlink_list);
-
- ctr_hard_link->hardlink_heal_period = current_time.tv_sec;
-
- /*aal izz well!*/
- ret = 0;
- goto out;
-error:
- GF_FREE (ctr_hard_link);
-out:
- return ret;
-}
-
-static void
-__delete_hard_link_from_list (ctr_hard_link_t **ctr_hard_link)
-{
- GF_ASSERT (ctr_hard_link);
- GF_ASSERT (*ctr_hard_link);
-
- /*Remove hard link from list*/
- list_del(&(*ctr_hard_link)->list);
- fini_ctr_hard_link (ctr_hard_link);
-}
-
-
-int
-ctr_delete_hard_link (xlator_t *this,
- ctr_xlator_ctx_t *ctr_xlator_ctx,
- uuid_t pgfid,
- const char *base_name)
-{
- int ret = -1;
- ctr_hard_link_t *ctr_hard_link = NULL;
-
- GF_ASSERT (this);
- GF_ASSERT (ctr_xlator_ctx);
-
-
- LOCK (&ctr_xlator_ctx->lock);
-
- /*Check if the hard link is present */
- ctr_hard_link = ctr_search_hard_link_ctx (this, ctr_xlator_ctx,
- pgfid, base_name);
- if (!ctr_hard_link) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_HARDLINK_MISSING_IN_LIST,
- "Hard link doesn't exist in the list");
- goto out;
- }
-
- __delete_hard_link_from_list (&ctr_hard_link);
- ctr_hard_link = NULL;
-
- ret = 0;
-out:
- UNLOCK (&ctr_xlator_ctx->lock);
-
- return ret;
-}
-
-
-
-
-int
-ctr_update_hard_link (xlator_t *this,
- ctr_xlator_ctx_t *ctr_xlator_ctx,
- uuid_t pgfid,
- const char *base_name,
- uuid_t old_pgfid,
- const char *old_base_name)
-{
- int ret = -1;
- ctr_hard_link_t *ctr_hard_link = NULL;
- struct timeval current_time = {0};
-
- GF_ASSERT (this);
- GF_ASSERT (ctr_xlator_ctx);
-
-
- LOCK (&ctr_xlator_ctx->lock);
-
- /*Check if the hard link is present */
- ctr_hard_link = ctr_search_hard_link_ctx (this, ctr_xlator_ctx,
- old_pgfid, old_base_name);
- if (!ctr_hard_link) {
- gf_msg_trace (this->name, 0, "Hard link doesn't exist"
- " in the list");
- /* Since the hard link is not present in the list
- * we add it to the list */
- ret = ctr_add_hard_link (this, ctr_xlator_ctx,
- pgfid, base_name);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_ADD_HARDLINK_TO_LIST_FAILED,
- "Failed adding hard link to the list");
- goto out;
- }
- ret = 0;
- goto out;
- }
-
- /* update the hard link */
- gf_uuid_copy (ctr_hard_link->pgfid, pgfid);
- GF_FREE (ctr_hard_link->base_name);
- ret = gf_asprintf(&ctr_hard_link->base_name, "%s", base_name);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- CTR_MSG_COPY_FAILED, "Failed copying basename"
- "to ctr_hard_link");
- /* delete the corrupted entry */
- __delete_hard_link_from_list (&ctr_hard_link);
- ctr_hard_link = NULL;
- goto out;
- }
-
- ret = gettimeofday (&current_time, NULL);
- if (ret == -1) {
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to get current time");
- ctr_hard_link->hardlink_heal_period = 0;
- } else {
- ctr_hard_link->hardlink_heal_period = current_time.tv_sec;
- }
-
- ret = 0;
-
-out:
- UNLOCK (&ctr_xlator_ctx->lock);
-
- return ret;
-}
-
-
-
-
-/* Delete all hardlinks */
-static int
-ctr_delete_all_hard_link (xlator_t *this,
- ctr_xlator_ctx_t *ctr_xlator_ctx)
-{
- int ret = -1;
- ctr_hard_link_t *ctr_hard_link = NULL;
- ctr_hard_link_t *tmp = NULL;
-
- GF_ASSERT (ctr_xlator_ctx);
-
- LOCK (&ctr_xlator_ctx->lock);
-
- list_for_each_entry_safe(ctr_hard_link, tmp,
- &ctr_xlator_ctx->hardlink_list, list)
- {
- /*Remove hard link from list*/
- __delete_hard_link_from_list (&ctr_hard_link);
- ctr_hard_link = NULL;
-
- }
-
-
- UNLOCK (&ctr_xlator_ctx->lock);
-
- ret = 0;
-
- return ret;
-}
-
-
-/* Please lock the inode before using this function */
-static ctr_xlator_ctx_t *
-__get_ctr_xlator_ctx (xlator_t *this,
- inode_t *inode)
-{
- int ret = 0;
- uint64_t _addr = 0;
- ctr_xlator_ctx_t *ctr_xlator_ctx = NULL;
-
- GF_ASSERT (this);
- GF_ASSERT (inode);
-
- ret = __inode_ctx_get (inode, this, &_addr);
- if (ret < 0)
- _addr = 0;
- if (_addr != 0) {
- ctr_xlator_ctx = (ctr_xlator_ctx_t *) (long)_addr;
- }
-
- return ctr_xlator_ctx;
-}
-
-
-ctr_xlator_ctx_t *
-init_ctr_xlator_ctx (xlator_t *this,
- inode_t *inode)
-{
- int ret = -1;
- uint64_t _addr = 0;
- ctr_xlator_ctx_t *ctr_xlator_ctx = NULL;
- struct timeval current_time = {0};
-
- GF_ASSERT (this);
- GF_ASSERT (inode);
-
- LOCK (&inode->lock);
- {
- ctr_xlator_ctx = __get_ctr_xlator_ctx (this, inode);
- if (ctr_xlator_ctx) {
- ret = 0;
- goto out;
- }
- ctr_xlator_ctx = GF_CALLOC (1, sizeof (*ctr_xlator_ctx),
- gf_ctr_mt_xlator_ctx);
- if (!ctr_xlator_ctx)
- goto out;
-
- ret = LOCK_INIT (&ctr_xlator_ctx->lock);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, ret,
- CTR_MSG_INIT_LOCK_FAILED,
- "Failed init lock %s", strerror(ret));
- goto out;
- }
- _addr = (uint64_t) ctr_xlator_ctx;
-
- ret = __inode_ctx_set (inode, this, &_addr);
- if (ret) {
- goto out;
- }
-
- INIT_LIST_HEAD (&ctr_xlator_ctx->hardlink_list);
-
- ret = gettimeofday (&current_time, NULL);
- if (ret == -1) {
- gf_log (this->name, GF_LOG_ERROR,
- "Failed to get current time");
- goto out;
- }
-
- ctr_xlator_ctx->inode_heal_period = current_time.tv_sec;
- }
- ret = 0;
-out:
- if (ret) {
- GF_FREE (ctr_xlator_ctx);
- ctr_xlator_ctx = NULL;
- }
-
- UNLOCK (&inode->lock);
-
- return ctr_xlator_ctx;
-}
-
-
-
-
-void
-fini_ctr_xlator_ctx (xlator_t *this,
- inode_t *inode)
-{
- int ret = 0;
- uint64_t _addr = 0;
- ctr_xlator_ctx_t *ctr_xlator_ctx = NULL;
-
-
- inode_ctx_del (inode, this, &_addr);
- if (!_addr)
- return;
-
- ctr_xlator_ctx = (ctr_xlator_ctx_t *) (long) _addr;
-
- ret = ctr_delete_all_hard_link (this, ctr_xlator_ctx);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING , 0,
- CTR_MSG_DELETE_HARDLINK_FAILED, "Failed deleting all "
- "hard links from inode context");
- }
-
- LOCK_DESTROY (&ctr_xlator_ctx->lock);
-
- GF_FREE (ctr_xlator_ctx);
-
-}
-
-
-
-
-ctr_xlator_ctx_t *
-get_ctr_xlator_ctx (xlator_t *this,
- inode_t *inode)
-{
- ctr_xlator_ctx_t *ctr_xlator_ctx = NULL;
-
- LOCK (&inode->lock);
- ctr_xlator_ctx = __get_ctr_xlator_ctx (this, inode);
- UNLOCK (&inode->lock);
-
- return ctr_xlator_ctx;
-}
-
diff --git a/xlators/features/changetimerecorder/src/ctr-xlator-ctx.h b/xlators/features/changetimerecorder/src/ctr-xlator-ctx.h
deleted file mode 100644
index 7f1c6cb1712..00000000000
--- a/xlators/features/changetimerecorder/src/ctr-xlator-ctx.h
+++ /dev/null
@@ -1,90 +0,0 @@
-/*
- Copyright (c) 2015 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#ifndef __CTR_XLATOR_CTX_H
-#define __CTR_XLATOR_CTX_H
-
-#include "xlator.h"
-#include "ctr_mem_types.h"
-#include "iatt.h"
-#include "glusterfs.h"
-#include "xlator.h"
-#include "logging.h"
-#include "locking.h"
-#include "common-utils.h"
-#include <time.h>
-#include <sys/time.h>
-
-typedef struct ctr_hard_link {
- uuid_t pgfid;
- char *base_name;
- /* Hardlink expiry : Defines the expiry period after which a
- * database heal is attempted. */
- uint64_t hardlink_heal_period;
- struct list_head list;
-} ctr_hard_link_t;
-
-typedef struct ctr_xlator_ctx {
- /* This represents the looked up hardlinks
- * NOTE: This doesn't represent all physical hardlinks of the inode*/
- struct list_head hardlink_list;
- uint64_t inode_heal_period;
- gf_lock_t lock;
-} ctr_xlator_ctx_t;
-
-
-ctr_hard_link_t *
-ctr_search_hard_link_ctx (xlator_t *this,
- ctr_xlator_ctx_t *ctr_xlator_ctx,
- uuid_t pgfid,
- const char *base_name);
-
-
-int
-ctr_add_hard_link (xlator_t *this,
- ctr_xlator_ctx_t *ctr_xlator_ctx,
- uuid_t pgfid,
- const char *base_name);
-
-
-
-int
-ctr_delete_hard_link (xlator_t *this,
- ctr_xlator_ctx_t *ctr_xlator_ctx,
- uuid_t pgfid,
- const char *base_name);
-
-
-int
-ctr_update_hard_link (xlator_t *this,
- ctr_xlator_ctx_t *ctr_xlator_ctx,
- uuid_t pgfid,
- const char *base_name,
- uuid_t old_pgfid,
- const char *old_base_name);
-
-
-ctr_xlator_ctx_t *
-get_ctr_xlator_ctx (xlator_t *this,
- inode_t *inode);
-
-
-
-
-ctr_xlator_ctx_t *
-init_ctr_xlator_ctx (xlator_t *this,
- inode_t *inode);
-
-
-void
-fini_ctr_xlator_ctx (xlator_t *this,
- inode_t *inode);
-
-#endif
diff --git a/xlators/features/changetimerecorder/src/ctr_mem_types.h b/xlators/features/changetimerecorder/src/ctr_mem_types.h
deleted file mode 100644
index f408c028e24..00000000000
--- a/xlators/features/changetimerecorder/src/ctr_mem_types.h
+++ /dev/null
@@ -1,24 +0,0 @@
-/*
- Copyright (c) 2008-2015 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-
-#ifndef __CTR_MEM_TYPES_H__
-#define __CTR_MEM_TYPES_H__
-
-#include "gfdb_mem-types.h"
-
-enum gf_ctr_mem_types_ {
- gf_ctr_mt_private_t = gfdb_mt_end + 1,
- gf_ctr_mt_xlator_ctx,
- gf_ctr_mt_hard_link_t,
- gf_ctr_mt_end
-};
-#endif
-
diff --git a/xlators/features/cloudsync/src/Makefile.am b/xlators/features/cloudsync/src/Makefile.am
index da660d7401b..e2a277e372b 100644
--- a/xlators/features/cloudsync/src/Makefile.am
+++ b/xlators/features/cloudsync/src/Makefile.am
@@ -1,9 +1,9 @@
+SUBDIRS = cloudsync-plugins
+
xlator_LTLIBRARIES = cloudsync.la
xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
-CLOUDSYNC_SRC = $(top_srcdir)/xlators/features/cloudsync/src
-
cloudsync_sources = cloudsync.c
CLOUDSYNC_SRC = $(top_srcdir)/xlators/features/cloudsync/src
@@ -12,9 +12,9 @@ CLOUDSYNC_BLD = $(top_builddir)/xlators/features/cloudsync/src
cloudsynccommon_sources = $(CLOUDSYNC_SRC)/cloudsync-common.c
noinst_HEADERS = $(CLOUDSYNC_BLD)/cloudsync.h \
- $(CLOUDSYNC_BLD)/cloudsync-mem-types.h \
- $(CLOUDSYNC_BLD)/cloudsync-messages.h \
- $(CLOUDSYNC_BLD)/cloudsync-common.h
+ $(CLOUDSYNC_BLD)/cloudsync-mem-types.h \
+ $(CLOUDSYNC_BLD)/cloudsync-messages.h \
+ $(CLOUDSYNC_BLD)/cloudsync-common.h
cloudsync_la_SOURCES = $(cloudsync_sources) $(cloudsynccommon_sources)
@@ -23,10 +23,10 @@ BUILT_SOURCES = cloudsync-autogen-fops.h
cloudsync_la_LDFLAGS = -module $(GF_XLATOR_DEFAULT_LDFLAGS)
-cloudsync_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-
-AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src -I$(top_srcdir)/rpc/xdr/src -I$(top_builddir)/rpc/xdr/src
+cloudsync_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la $(LIB_DL)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src -I$(top_srcdir)/rpc/xdr/src -I$(top_builddir)/rpc/xdr/src \
+ -DCS_PLUGINDIR=\"$(libdir)/glusterfs/$(PACKAGE_VERSION)/cloudsync-plugins\"
AM_CFLAGS = -Wall -fno-strict-aliasing $(GF_CFLAGS)
noinst_PYTHON = cloudsync-fops-c.py cloudsync-fops-h.py
@@ -44,4 +44,3 @@ CLEANFILES = $(nodist_cloudsync_la_SOURCES)
uninstall-local:
rm -f $(DESTDIR)$(xlatordir)/cloudsync.so
-
diff --git a/xlators/features/cloudsync/src/cloudsync-autogen-fops-tmpl.c b/xlators/features/cloudsync/src/cloudsync-autogen-fops-tmpl.c
index 6bb68cd170c..ee63f983980 100644
--- a/xlators/features/cloudsync/src/cloudsync-autogen-fops-tmpl.c
+++ b/xlators/features/cloudsync/src/cloudsync-autogen-fops-tmpl.c
@@ -20,11 +20,11 @@
#include <dlfcn.h>
-#include "glusterfs.h"
-#include "xlator.h"
-#include "defaults.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
#include "cloudsync.h"
#include "cloudsync-common.h"
-#include "call-stub.h"
+#include <glusterfs/call-stub.h>
#pragma generate
diff --git a/xlators/features/cloudsync/src/cloudsync-autogen-fops-tmpl.h b/xlators/features/cloudsync/src/cloudsync-autogen-fops-tmpl.h
index 2db2a9c88c7..d922c77d8aa 100644
--- a/xlators/features/cloudsync/src/cloudsync-autogen-fops-tmpl.h
+++ b/xlators/features/cloudsync/src/cloudsync-autogen-fops-tmpl.h
@@ -15,7 +15,7 @@
#ifndef _CLOUDSYNC_AUTOGEN_FOPS_H
#define _CLOUDSYNC_AUTOGEN_FOPS_H
-#include "xlator.h"
+#include <glusterfs/xlator.h>
#include "cloudsync.h"
#include "cloudsync-common.h"
diff --git a/xlators/features/cloudsync/src/cloudsync-common.c b/xlators/features/cloudsync/src/cloudsync-common.c
index d0d00decca3..445a31b90e7 100644
--- a/xlators/features/cloudsync/src/cloudsync-common.c
+++ b/xlators/features/cloudsync/src/cloudsync-common.c
@@ -11,34 +11,50 @@
#include "cloudsync-common.h"
void
-cs_local_wipe (xlator_t *this, cs_local_t *local)
+cs_xattrinfo_wipe(cs_local_t *local)
{
- if (!local)
- return;
+ if (local->xattrinfo.lxattr) {
+ if (local->xattrinfo.lxattr->file_path)
+ GF_FREE(local->xattrinfo.lxattr->file_path);
- loc_wipe (&local->loc);
+ if (local->xattrinfo.lxattr->volname)
+ GF_FREE(local->xattrinfo.lxattr->volname);
- if (local->fd) {
- fd_unref (local->fd);
- local->fd = NULL;
- }
+ GF_FREE(local->xattrinfo.lxattr);
+ }
+}
+
+void
+cs_local_wipe(xlator_t *this, cs_local_t *local)
+{
+ if (!local)
+ return;
+
+ loc_wipe(&local->loc);
+
+ if (local->fd) {
+ fd_unref(local->fd);
+ local->fd = NULL;
+ }
+
+ if (local->stub) {
+ call_stub_destroy(local->stub);
+ local->stub = NULL;
+ }
- if (local->stub) {
- call_stub_destroy (local->stub);
- local->stub = NULL;
- }
+ if (local->xattr_req)
+ dict_unref(local->xattr_req);
- if (local->xattr_req)
- dict_unref (local->xattr_req);
+ if (local->xattr_rsp)
+ dict_unref(local->xattr_rsp);
- if (local->xattr_rsp)
- dict_unref (local->xattr_rsp);
+ if (local->dlfd)
+ fd_unref(local->dlfd);
- if (local->dlfd)
- fd_unref (local->dlfd);
+ if (local->remotepath)
+ GF_FREE(local->remotepath);
- if (local->remotepath)
- GF_FREE (local->remotepath);
+ cs_xattrinfo_wipe(local);
- mem_put (local);
+ mem_put(local);
}
diff --git a/xlators/features/cloudsync/src/cloudsync-common.h b/xlators/features/cloudsync/src/cloudsync-common.h
index 3298ab0a6f2..11d233460a4 100644
--- a/xlators/features/cloudsync/src/cloudsync-common.h
+++ b/xlators/features/cloudsync/src/cloudsync-common.h
@@ -10,88 +10,125 @@
#ifndef _CLOUDSYNC_COMMON_H
#define _CLOUDSYNC_COMMON_H
-#include "glusterfs.h"
-#include "call-stub.h"
-#include "xlator.h"
-#include "syncop.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/call-stub.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/syncop.h>
+#include <glusterfs/compat-errno.h>
#include "cloudsync-mem-types.h"
#include "cloudsync-messages.h"
+typedef struct cs_loc_xattr {
+ char *file_path;
+ uuid_t uuid;
+ uuid_t gfid;
+ char *volname;
+} cs_loc_xattr_t;
+
+typedef struct cs_size_xattr {
+ uint64_t size;
+ uint64_t blksize;
+ uint64_t blocks;
+} cs_size_xattr_t;
+
typedef struct cs_local {
- loc_t loc;
- fd_t *fd;
- call_stub_t *stub;
- call_frame_t *main_frame;
- int op_errno;
- int op_ret;
- fd_t *dlfd;
- off_t dloffset;
- struct iatt stbuf;
- dict_t *xattr_rsp;
- dict_t *xattr_req;
- glusterfs_fop_t fop;
- gf_boolean_t locked;
- int call_cnt;
- inode_t *inode;
- char *remotepath;
+ loc_t loc;
+ fd_t *fd;
+ call_stub_t *stub;
+ call_frame_t *main_frame;
+ int op_errno;
+ int op_ret;
+ fd_t *dlfd;
+ off_t dloffset;
+ struct iatt stbuf;
+ dict_t *xattr_rsp;
+ dict_t *xattr_req;
+ glusterfs_fop_t fop;
+ gf_boolean_t locked;
+ int call_cnt;
+ inode_t *inode;
+ char *remotepath;
+
+ struct {
+ /* offset, flags and size are the information needed
+ * by read fop for remote read operation. These will be
+ * populated in cloudsync read fop, before being passed
+ * on to the plugin performing remote read.
+ */
+ off_t offset;
+ uint32_t flags;
+ size_t size;
+ cs_loc_xattr_t *lxattr;
+ } xattrinfo;
+
} cs_local_t;
-typedef int (*fop_download_t) (call_frame_t *frame, void *config);
+typedef int (*fop_download_t)(call_frame_t *frame, void *config);
+
+typedef int (*fop_remote_read_t)(call_frame_t *, void *);
-typedef void *(*store_init) (xlator_t *this);
+typedef void *(*store_init)(xlator_t *this);
-typedef int (*store_reconfigure) (xlator_t *this, dict_t *options);
+typedef int (*store_reconfigure)(xlator_t *this, dict_t *options);
-typedef void (*store_fini) (void *config);
+typedef void (*store_fini)(void *config);
struct cs_remote_stores {
- char *name; /* store name */
- void *config; /* store related information */
- fop_download_t dlfop; /* store specific download function */
- store_init init; /* store init to initialize store config */
- store_reconfigure reconfigure; /* reconfigure store config */
- store_fini fini;
- void *handle; /* shared library handle*/
+ char *name; /* store name */
+ void *config; /* store related information */
+ fop_download_t dlfop; /* store specific download function */
+ fop_remote_read_t rdfop; /* store specific read function */
+ store_init init; /* store init to initialize store config */
+ store_reconfigure reconfigure; /* reconfigure store config */
+ store_fini fini;
+ void *handle; /* shared library handle*/
};
typedef struct cs_private {
- xlator_t *this;
- struct cs_remote_stores *stores;
- gf_boolean_t abortdl;
- pthread_spinlock_t lock;
+ xlator_t *this;
+ struct cs_remote_stores *stores;
+ gf_boolean_t abortdl;
+ pthread_spinlock_t lock;
+ gf_boolean_t remote_read;
} cs_private_t;
void
-cs_local_wipe (xlator_t *this, cs_local_t *local);
-
-#define CS_STACK_UNWIND(fop, frame, params ...) do { \
- cs_local_t *__local = NULL; \
- xlator_t *__xl = NULL; \
- if (frame) { \
- __xl = frame->this; \
- __local = frame->local; \
- frame->local = NULL; \
- } \
- STACK_UNWIND_STRICT (fop, frame, params); \
- cs_local_wipe (__xl, __local); \
-} while (0)
-
-#define CS_STACK_DESTROY(frame) do { \
- cs_local_t *__local = NULL; \
- xlator_t *__xl = NULL; \
- __xl = frame->this; \
- __local = frame->local; \
- frame->local = NULL; \
- STACK_DESTROY (frame->root); \
- cs_local_wipe (__xl, __local); \
-} while (0)
+cs_local_wipe(xlator_t *this, cs_local_t *local);
+
+void
+cs_xattrinfo_wipe(cs_local_t *local);
+
+#define CS_STACK_UNWIND(fop, frame, params...) \
+ do { \
+ cs_local_t *__local = NULL; \
+ xlator_t *__xl = NULL; \
+ if (frame) { \
+ __xl = frame->this; \
+ __local = frame->local; \
+ frame->local = NULL; \
+ } \
+ STACK_UNWIND_STRICT(fop, frame, params); \
+ cs_local_wipe(__xl, __local); \
+ } while (0)
+
+#define CS_STACK_DESTROY(frame) \
+ do { \
+ cs_local_t *__local = NULL; \
+ xlator_t *__xl = NULL; \
+ __xl = frame->this; \
+ __local = frame->local; \
+ frame->local = NULL; \
+ STACK_DESTROY(frame->root); \
+ cs_local_wipe(__xl, __local); \
+ } while (0)
typedef struct store_methods {
- int (*fop_download) (call_frame_t *frame, void *config);
- /* return type should be the store config */
- void *(*fop_init) (xlator_t *this);
- int (*fop_reconfigure) (xlator_t *this, dict_t *options);
- void (*fop_fini) (void *config);
+ int (*fop_download)(call_frame_t *frame, void *config);
+ int (*fop_remote_read)(call_frame_t *, void *);
+ /* return type should be the store config */
+ void *(*fop_init)(xlator_t *this);
+ int (*fop_reconfigure)(xlator_t *this, dict_t *options);
+ void (*fop_fini)(void *config);
} store_methods_t;
#endif /* _CLOUDSYNC_COMMON_H */
diff --git a/xlators/features/cloudsync/src/cloudsync-fops-c.py b/xlators/features/cloudsync/src/cloudsync-fops-c.py
index fda56752871..c27df97ae58 100644..100755
--- a/xlators/features/cloudsync/src/cloudsync-fops-c.py
+++ b/xlators/features/cloudsync/src/cloudsync-fops-c.py
@@ -1,4 +1,4 @@
-#!/usr/bin/python
+#!/usr/bin/python3
from __future__ import print_function
import os
@@ -14,7 +14,7 @@ int32_t
cs_@NAME@ (call_frame_t *frame, xlator_t *this,
@LONG_ARGS@)
{
- int op_errno = -1;
+ int op_errno = EINVAL ;
cs_local_t *local = NULL;
int ret = 0;
cs_inode_ctx_t *ctx = NULL;
@@ -35,11 +35,19 @@ cs_@NAME@ (call_frame_t *frame, xlator_t *this,
__cs_inode_ctx_get (this, fd->inode, &ctx);
if (ctx)
- state = __cs_get_file_state (this, fd->inode, ctx);
+ state = __cs_get_file_state (fd->inode, ctx);
else
state = GF_CS_LOCAL;
- local->xattr_req = xdata ? dict_ref (xdata) : (xdata = dict_new ());
+ xdata = xdata ? dict_ref (xdata) : dict_new ();
+
+ if (!xdata) {
+ gf_msg (this->name, GF_LOG_ERROR, 0, 0, "insufficient memory");
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->xattr_req = xdata;
ret = dict_set_uint32 (local->xattr_req, GF_CS_OBJECT_STATUS, 1);
if (ret) {
@@ -137,15 +145,15 @@ cs_@NAME@_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
} else {
__cs_inode_ctx_update (this, fd->inode, val);
gf_msg (this->name, GF_LOG_INFO, 0, 0,
- " state = %ld", val);
+ " state = %" PRIu64, val);
if (local->call_cnt == 1 &&
(val == GF_CS_REMOTE ||
val == GF_CS_DOWNLOADING)) {
gf_msg (this->name, GF_LOG_INFO, 0,
0, " will repair and download "
- "the file, current state : %ld",
- val);
+ "the file, current state : %"
+ PRIu64, val);
goto repair;
} else {
gf_msg (this->name, GF_LOG_ERROR, 0, 0,
@@ -187,19 +195,29 @@ int32_t
cs_@NAME@ (call_frame_t *frame, xlator_t *this,
@LONG_ARGS@)
{
+ int op_errno = EINVAL;
cs_local_t *local = NULL;
int ret = 0;
local = cs_local_init (this, frame, loc, NULL, GF_FOP_@UPNAME@);
if (!local) {
gf_msg (this->name, GF_LOG_ERROR, 0, 0, "local is NULL");
+ op_errno = ENOMEM;
goto err;
}
if (loc->inode->ia_type == IA_IFDIR)
goto wind;
- local->xattr_req = xdata ? dict_ref (xdata) : dict_new ();
+ xdata = xdata ? dict_ref (xdata) : dict_new ();
+
+ if (!xdata) {
+ gf_msg (this->name, GF_LOG_ERROR, 0, 0, "insufficient memory");
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->xattr_req = xdata;
ret = dict_set_uint32 (local->xattr_req, GF_CS_OBJECT_STATUS, 1);
if (ret) {
@@ -215,7 +233,7 @@ wind:
return 0;
err:
- CS_STACK_UNWIND (@NAME@, frame, -1, errno, @CBK_ERROR_ARGS@);
+ CS_STACK_UNWIND (@NAME@, frame, -1, op_errno, @CBK_ERROR_ARGS@);
return 0;
}
@@ -274,7 +292,7 @@ fd_ops = ['readv', 'writev', 'flush', 'fsync', 'fsyncdir', 'ftruncate',
# These are the current actual lists used to generate the code
# The following list contains fops which are fd based that modifies data
-fd_data_modify_op_fop_template = ['readv', 'writev', 'flush', 'fsync',
+fd_data_modify_op_fop_template = ['writev', 'flush', 'fsync',
'ftruncate', 'rchecksum', 'fallocate',
'discard', 'zerofill', 'seek']
@@ -284,8 +302,8 @@ loc_stat_op_fop_template = ['lookup', 'stat', 'discover', 'access', 'setattr',
'getattr']
# These fops need a separate implementation
-special_fops = ['readdirp', 'statfs', 'setxattr', 'unlink', 'getxattr',
- 'truncate', 'fstat']
+special_fops = ['statfs', 'setxattr', 'unlink', 'getxattr',
+ 'truncate', 'fstat', 'readv', 'readdirp']
def gen_defaults():
for name in ops:
diff --git a/xlators/features/cloudsync/src/cloudsync-fops-h.py b/xlators/features/cloudsync/src/cloudsync-fops-h.py
index 37dfa5b10be..faa2de651a7 100644..100755
--- a/xlators/features/cloudsync/src/cloudsync-fops-h.py
+++ b/xlators/features/cloudsync/src/cloudsync-fops-h.py
@@ -1,4 +1,4 @@
-#!/usr/bin/python
+#!/usr/bin/python3
from __future__ import print_function
import os
@@ -16,7 +16,7 @@ cs_@NAME@ (call_frame_t *frame, xlator_t *this,
"""
def gen_defaults():
- for name, value in ops.iteritems():
+ for name, value in ops.items():
if name == 'getspec':
continue
print(generate(OP_FOP_TEMPLATE, name, fop_subs))
diff --git a/xlators/features/cloudsync/src/cloudsync-mem-types.h b/xlators/features/cloudsync/src/cloudsync-mem-types.h
index 6ebcb16552b..220346405d0 100644
--- a/xlators/features/cloudsync/src/cloudsync-mem-types.h
+++ b/xlators/features/cloudsync/src/cloudsync-mem-types.h
@@ -8,16 +8,15 @@
* cases as published by the Free Software Foundation.
*/
-
#ifndef __CLOUDSYNC_MEM_TYPES_H__
#define __CLOUDSYNC_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum cs_mem_types_ {
- gf_cs_mt_cs_private_t = gf_common_mt_end + 1,
- gf_cs_mt_cs_remote_stores_t,
- gf_cs_mt_cs_inode_ctx_t,
- gf_cs_mt_end
+ gf_cs_mt_cs_private_t = gf_common_mt_end + 1,
+ gf_cs_mt_cs_remote_stores_t,
+ gf_cs_mt_cs_inode_ctx_t,
+ gf_cs_mt_cs_lxattr_t,
+ gf_cs_mt_end
};
#endif /* __CLOUDSYNC_MEM_TYPES_H__ */
-
diff --git a/xlators/features/cloudsync/src/cloudsync-messages.h b/xlators/features/cloudsync/src/cloudsync-messages.h
index ad4b7d2e0b8..fb08f72de7f 100644
--- a/xlators/features/cloudsync/src/cloudsync-messages.h
+++ b/xlators/features/cloudsync/src/cloudsync-messages.h
@@ -8,12 +8,9 @@
* cases as published by the Free Software Foundation.
*/
-
#ifndef __CLOUDSYNC_MESSAGES_H__
#define __CLOUDSYNC_MESSAGES_H__
/*TODO: define relevant message ids */
-
#endif /* __CLOUDSYNC_MESSAGES_H__ */
-
diff --git a/xlators/features/changetimerecorder/Makefile.am b/xlators/features/cloudsync/src/cloudsync-plugins/Makefile.am
index a985f42a877..a985f42a877 100644
--- a/xlators/features/changetimerecorder/Makefile.am
+++ b/xlators/features/cloudsync/src/cloudsync-plugins/Makefile.am
diff --git a/xlators/features/cloudsync/src/cloudsync-plugins/src/Makefile.am b/xlators/features/cloudsync/src/cloudsync-plugins/src/Makefile.am
new file mode 100644
index 00000000000..fb6b0580c6d
--- /dev/null
+++ b/xlators/features/cloudsync/src/cloudsync-plugins/src/Makefile.am
@@ -0,0 +1,11 @@
+if BUILD_AMAZONS3_PLUGIN
+ AMAZONS3_DIR = cloudsyncs3
+endif
+
+if BUILD_CVLT_PLUGIN
+ CVLT_DIR = cvlt
+endif
+
+SUBDIRS = ${AMAZONS3_DIR} ${CVLT_DIR}
+
+CLEANFILES =
diff --git a/xlators/storage/bd/Makefile.am b/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/Makefile.am
index a985f42a877..a985f42a877 100644
--- a/xlators/storage/bd/Makefile.am
+++ b/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/Makefile.am
diff --git a/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/Makefile.am b/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/Makefile.am
new file mode 100644
index 00000000000..6509426ef87
--- /dev/null
+++ b/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/Makefile.am
@@ -0,0 +1,12 @@
+csp_LTLIBRARIES = cloudsyncs3.la
+cspdir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/cloudsync-plugins
+
+cloudsyncs3_la_SOURCES = libcloudsyncs3.c $(top_srcdir)/xlators/features/cloudsync/src/cloudsync-common.c
+cloudsyncs3_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+cloudsyncs3_la_LDFLAGS = -module -export-symbols $(top_srcdir)/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3.sym $(GF_XLATOR_LDFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src -I$(top_srcdir)/rpc/xdr/src -I$(top_builddir)/rpc/xdr/src -lcurlpp -lcryptopp
+noinst_HEADERS = libcloudsyncs3.h libcloudsyncs3-mem-types.h
+AM_CFLAGS = -Wall -fno-strict-aliasing $(GF_CFLAGS) -lcurl -lcrypto -I$(top_srcdir)/xlators/features/cloudsync/src
+CLEANFILES =
+
+EXTRA_DIST = libcloudsyncs3.sym
diff --git a/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3-mem-types.h b/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3-mem-types.h
new file mode 100644
index 00000000000..7ccfcc9f4b6
--- /dev/null
+++ b/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3-mem-types.h
@@ -0,0 +1,19 @@
+/*
+ * Copyright (c) 2018 Red Hat, Inc. <http://www.redhat.com>
+ * This file is part of GlusterFS.
+ *
+ * This file is licensed to you under your choice of the GNU Lesser
+ * General Public License, version 3 or any later version (LGPLv3 or
+ * later), or the GNU General Public License, version 2 (GPLv2), in all
+ * cases as published by the Free Software Foundation.
+ */
+
+#ifndef __LIBAWS_MEM_TYPES_H__
+#define __LIBAWS_MEM_TYPES_H__
+
+#include <glusterfs/mem-types.h>
+enum libaws_mem_types_ {
+ gf_libaws_mt_aws_private_t = gf_common_mt_end + 1,
+ gf_libaws_mt_end
+};
+#endif /* __CLOUDSYNC_MEM_TYPES_H__ */
diff --git a/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3.c b/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3.c
new file mode 100644
index 00000000000..23c3599825a
--- /dev/null
+++ b/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3.c
@@ -0,0 +1,584 @@
+/*
+ Copyright (c) 2018 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+*/
+
+#include <stdlib.h>
+#include <openssl/hmac.h>
+#include <openssl/evp.h>
+#include <openssl/bio.h>
+#include <openssl/buffer.h>
+#include <openssl/crypto.h>
+#include <curl/curl.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/glusterfs.h>
+#include "libcloudsyncs3.h"
+#include "cloudsync-common.h"
+
+#define RESOURCE_SIZE 4096
+
+store_methods_t store_ops = {
+ .fop_download = aws_download_s3,
+ .fop_init = aws_init,
+ .fop_reconfigure = aws_reconfigure,
+ .fop_fini = aws_fini,
+};
+
+typedef struct aws_private {
+ char *hostname;
+ char *bucketid;
+ char *awssekey;
+ char *awskeyid;
+ gf_boolean_t abortdl;
+ pthread_spinlock_t lock;
+} aws_private_t;
+
+void *
+aws_init(xlator_t *this)
+{
+ aws_private_t *priv = NULL;
+ char *temp_str = NULL;
+ int ret = 0;
+
+ priv = GF_CALLOC(1, sizeof(aws_private_t), gf_libaws_mt_aws_private_t);
+ if (!priv) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "insufficient memory");
+ return NULL;
+ }
+
+ priv->abortdl = _gf_false;
+
+ pthread_spin_init(&priv->lock, PTHREAD_PROCESS_PRIVATE);
+
+ pthread_spin_lock(&(priv->lock));
+ {
+ if (dict_get_str(this->options, "s3plugin-seckey", &temp_str) == 0) {
+ priv->awssekey = gf_strdup(temp_str);
+ if (!priv->awssekey) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, 0,
+ "initializing aws secret key failed");
+ ret = -1;
+ goto unlock;
+ }
+ }
+
+ if (dict_get_str(this->options, "s3plugin-keyid", &temp_str) == 0) {
+ priv->awskeyid = gf_strdup(temp_str);
+ if (!priv->awskeyid) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, 0,
+ "initializing aws key ID failed");
+ ret = -1;
+ goto unlock;
+ }
+ }
+
+ if (dict_get_str(this->options, "s3plugin-bucketid", &temp_str) == 0) {
+ priv->bucketid = gf_strdup(temp_str);
+ if (!priv->bucketid) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, 0,
+ "initializing aws bucketid failed");
+
+ ret = -1;
+ goto unlock;
+ }
+ }
+
+ if (dict_get_str(this->options, "s3plugin-hostname", &temp_str) == 0) {
+ priv->hostname = gf_strdup(temp_str);
+ if (!priv->hostname) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, 0,
+ "initializing aws hostname failed");
+
+ ret = -1;
+ goto unlock;
+ }
+ }
+
+ gf_msg_debug(this->name, 0,
+ "stored key: %s id: %s "
+ "bucketid %s hostname: %s",
+ priv->awssekey, priv->awskeyid, priv->bucketid,
+ priv->hostname);
+ }
+unlock:
+ pthread_spin_unlock(&(priv->lock));
+
+ if (ret == -1) {
+ GF_FREE(priv->awskeyid);
+ GF_FREE(priv->awssekey);
+ GF_FREE(priv->bucketid);
+ GF_FREE(priv->hostname);
+ GF_FREE(priv);
+ priv = NULL;
+ }
+
+ return (void *)priv;
+}
+
+int
+aws_reconfigure(xlator_t *this, dict_t *options)
+{
+ aws_private_t *priv = NULL;
+ char *temp_str = NULL;
+ int ret = 0;
+ cs_private_t *cspriv = NULL;
+
+ cspriv = this->private;
+
+ priv = cspriv->stores->config;
+
+ if (!priv) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "null priv");
+ return -1;
+ }
+
+ pthread_spin_lock(&(priv->lock));
+ {
+ if (dict_get_str(options, "s3plugin-seckey", &temp_str) == 0) {
+ priv->awssekey = gf_strdup(temp_str);
+ if (!priv->awssekey) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, 0,
+ "initializing aws secret key failed");
+ ret = -1;
+ goto out;
+ }
+ }
+
+ if (dict_get_str(options, "s3plugin-keyid", &temp_str) == 0) {
+ priv->awskeyid = gf_strdup(temp_str);
+ if (!priv->awskeyid) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, 0,
+ "initializing aws key ID failed");
+ ret = -1;
+ goto out;
+ }
+ }
+
+ if (dict_get_str(options, "s3plugin-bucketid", &temp_str) == 0) {
+ priv->bucketid = gf_strdup(temp_str);
+ if (!priv->bucketid) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, 0,
+ "initializing aws bucketid failed");
+ ret = -1;
+ goto out;
+ }
+ }
+
+ if (dict_get_str(options, "s3plugin-hostname", &temp_str) == 0) {
+ priv->hostname = gf_strdup(temp_str);
+ if (!priv->hostname) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, 0,
+ "initializing aws hostname failed");
+ ret = -1;
+ goto out;
+ }
+ }
+ }
+out:
+ pthread_spin_unlock(&(priv->lock));
+
+ gf_msg_debug(this->name, 0,
+ "stored key: %s id: %s "
+ "bucketid %s hostname: %s",
+ priv->awssekey, priv->awskeyid, priv->bucketid,
+ priv->hostname);
+
+ return ret;
+}
+
+void
+aws_fini(void *config)
+{
+ aws_private_t *priv = NULL;
+
+ priv = (aws_private_t *)priv;
+
+ if (priv) {
+ GF_FREE(priv->hostname);
+ GF_FREE(priv->bucketid);
+ GF_FREE(priv->awssekey);
+ GF_FREE(priv->awskeyid);
+
+ pthread_spin_destroy(&priv->lock);
+ GF_FREE(priv);
+ }
+}
+
+int32_t
+mem_acct_init(xlator_t *this)
+{
+ int ret = -1;
+
+ GF_VALIDATE_OR_GOTO("dht", this, out);
+
+ ret = xlator_mem_acct_init(this, gf_libaws_mt_end + 1);
+
+ if (ret != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "Memory accounting init failed");
+ return ret;
+ }
+out:
+ return ret;
+}
+char *
+aws_form_request(char *resource, char **date, char *reqtype, char *bucketid,
+ char *filepath)
+{
+ char httpdate[256];
+ time_t ctime;
+ struct tm *gtime = NULL;
+ char *sign_req = NULL;
+ int signreq_len = -1;
+ int date_len = -1;
+ int res_len = -1;
+
+ ctime = gf_time();
+ gtime = gmtime(&ctime);
+
+ date_len = strftime(httpdate, sizeof(httpdate),
+ "%a, %d %b %Y %H:%M:%S +0000", gtime);
+
+ *date = gf_strndup(httpdate, date_len);
+ if (*date == NULL) {
+ gf_msg("CS", GF_LOG_ERROR, ENOMEM, 0,
+ "memory allocation "
+ "failure for date");
+ goto out;
+ }
+
+ res_len = snprintf(resource, RESOURCE_SIZE, "%s/%s", bucketid, filepath);
+
+ gf_msg_debug("CS", 0, "resource %s", resource);
+
+ /* 6 accounts for the 4 new line chars, one forward slash and
+ * one null char */
+ signreq_len = res_len + date_len + strlen(reqtype) + 6;
+
+ sign_req = GF_MALLOC(signreq_len, gf_common_mt_char);
+ if (sign_req == NULL) {
+ gf_msg("CS", GF_LOG_ERROR, ENOMEM, 0,
+ "memory allocation "
+ "failure for sign_req");
+ goto out;
+ }
+
+ snprintf(sign_req, signreq_len, "%s\n\n%s\n%s\n/%s", reqtype, "", *date,
+ resource);
+
+out:
+ return sign_req;
+}
+
+char *
+aws_b64_encode(const unsigned char *input, int length)
+{
+ BIO *bio, *b64;
+ BUF_MEM *bptr;
+ char *buff = NULL;
+
+ b64 = BIO_new(BIO_f_base64());
+ bio = BIO_new(BIO_s_mem());
+ b64 = BIO_push(b64, bio);
+ BIO_write(b64, input, length);
+ BIO_flush(b64);
+ BIO_get_mem_ptr(b64, &bptr);
+
+ buff = GF_MALLOC(bptr->length, gf_common_mt_char);
+ memcpy(buff, bptr->data, bptr->length - 1);
+ buff[bptr->length - 1] = 0;
+
+ BIO_free_all(b64);
+
+ return buff;
+}
+
+char *
+aws_sign_request(char *const str, char *awssekey)
+{
+#if (OPENSSL_VERSION_NUMBER < 0x1010002f)
+ HMAC_CTX ctx;
+#endif
+ HMAC_CTX *pctx = NULL;
+ ;
+
+ unsigned char md[256];
+ unsigned len;
+ char *base64 = NULL;
+
+#if (OPENSSL_VERSION_NUMBER < 0x1010002f)
+ HMAC_CTX_init(&ctx);
+ pctx = &ctx;
+#else
+ pctx = HMAC_CTX_new();
+#endif
+ HMAC_Init_ex(pctx, awssekey, strlen(awssekey), EVP_sha1(), NULL);
+ HMAC_Update(pctx, (unsigned char *)str, strlen(str));
+ HMAC_Final(pctx, (unsigned char *)md, &len);
+
+#if (OPENSSL_VERSION_NUMBER < 0x1010002f)
+ HMAC_CTX_cleanup(pctx);
+#else
+ HMAC_CTX_free(pctx);
+#endif
+ base64 = aws_b64_encode(md, len);
+
+ return base64;
+}
+
+int
+aws_dlwritev_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata)
+{
+ aws_private_t *priv = NULL;
+
+ if (op_ret == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, op_errno,
+ "write failed "
+ ". Aborting Download");
+
+ priv = this->private;
+ pthread_spin_lock(&(priv->lock));
+ {
+ priv->abortdl = _gf_true;
+ }
+ pthread_spin_unlock(&(priv->lock));
+ }
+
+ CS_STACK_DESTROY(frame);
+
+ return op_ret;
+}
+
+size_t
+aws_write_callback(void *dlbuf, size_t size, size_t nitems, void *mainframe)
+{
+ call_frame_t *frame = NULL;
+ fd_t *dlfd = NULL;
+ int ret = 0;
+ cs_local_t *local = NULL;
+ struct iovec iov = {
+ 0,
+ };
+ struct iobref *iobref = NULL;
+ struct iobuf *iobuf = NULL;
+ struct iovec dliov = {
+ 0,
+ };
+ size_t tsize = 0;
+ xlator_t *this = NULL;
+ cs_private_t *xl_priv = NULL;
+ aws_private_t *priv = NULL;
+ call_frame_t *dlframe = NULL;
+
+ frame = (call_frame_t *)mainframe;
+ this = frame->this;
+ xl_priv = this->private;
+ priv = xl_priv->stores->config;
+
+ pthread_spin_lock(&(priv->lock));
+ {
+ /* returning size other than the size passed from curl will
+ * abort further download*/
+ if (priv->abortdl) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "aborting download");
+ pthread_spin_unlock(&(priv->lock));
+ return 0;
+ }
+ }
+ pthread_spin_unlock(&(priv->lock));
+
+ local = frame->local;
+ dlfd = local->dlfd;
+ tsize = size * nitems;
+
+ dliov.iov_base = (void *)dlbuf;
+ dliov.iov_len = tsize;
+
+ ret = iobuf_copy(this->ctx->iobuf_pool, &dliov, 1, &iobref, &iobuf, &iov);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "iobuf_copy failed");
+ goto out;
+ }
+
+ /* copy frame */
+ dlframe = copy_frame(frame);
+ if (!dlframe) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "copy_frame failed");
+ tsize = 0;
+ goto out;
+ }
+
+ STACK_WIND(dlframe, aws_dlwritev_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, dlfd, &iov, 1, local->dloffset,
+ 0, iobref, NULL);
+
+ local->dloffset += tsize;
+
+out:
+ if (iobuf)
+ iobuf_unref(iobuf);
+ if (iobref)
+ iobref_unref(iobref);
+
+ return tsize;
+}
+
+int
+aws_download_s3(call_frame_t *frame, void *config)
+{
+ char *buf;
+ int bufsize = -1;
+ CURL *handle = NULL;
+ struct curl_slist *slist = NULL;
+ struct curl_slist *tmp = NULL;
+ xlator_t *this = NULL;
+ int ret = 0;
+ int debug = 1;
+ CURLcode res;
+ char errbuf[CURL_ERROR_SIZE];
+ size_t len = 0;
+ long responsecode;
+ char *sign_req = NULL;
+ char *date = NULL;
+ char *const reqtype = "GET";
+ char *signature = NULL;
+ cs_local_t *local = NULL;
+ char resource[RESOURCE_SIZE] = {
+ 0,
+ };
+ aws_private_t *priv = NULL;
+
+ this = frame->this;
+
+ local = frame->local;
+
+ priv = (aws_private_t *)config;
+
+ if (!priv->bucketid || !priv->hostname || !priv->awssekey ||
+ !priv->awskeyid) {
+ ret = -1;
+ goto out;
+ }
+
+ sign_req = aws_form_request(resource, &date, reqtype, priv->bucketid,
+ local->remotepath);
+ if (!sign_req) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "null sign_req, "
+ "aborting download");
+ ret = -1;
+ goto out;
+ }
+
+ gf_msg_debug("CS", 0, "sign_req %s date %s", sign_req, date);
+
+ signature = aws_sign_request(sign_req, priv->awssekey);
+ if (!signature) {
+ gf_msg("CS", GF_LOG_ERROR, 0, 0,
+ "null signature, "
+ "aborting download");
+ ret = -1;
+ goto out;
+ }
+
+ handle = curl_easy_init();
+ this = frame->this;
+
+ /* special numbers 6, 20, 10 accounts for static characters in the
+ * below snprintf string format arguments*/
+ bufsize = strlen(date) + 6 + strlen(priv->awskeyid) + strlen(signature) +
+ 20 + strlen(priv->hostname) + 10;
+
+ buf = (char *)alloca(bufsize);
+ if (!buf) {
+ gf_msg("CS", GF_LOG_ERROR, ENOMEM, 0,
+ "mem allocation "
+ "failed for buf");
+ ret = -1;
+ goto out;
+ }
+
+ snprintf(buf, bufsize, "Date: %s", date);
+ slist = curl_slist_append(slist, buf);
+ snprintf(buf, bufsize, "Authorization: AWS %s:%s", priv->awskeyid,
+ signature);
+ slist = curl_slist_append(slist, buf);
+ snprintf(buf, bufsize, "https://%s/%s", priv->hostname, resource);
+
+ if (gf_log_get_loglevel() >= GF_LOG_DEBUG) {
+ tmp = slist;
+ while (tmp) {
+ gf_msg_debug(this->name, 0, "slist for curl - %s", tmp->data);
+ tmp = tmp->next;
+ }
+ }
+
+ curl_easy_setopt(handle, CURLOPT_HTTPHEADER, slist);
+ curl_easy_setopt(handle, CURLOPT_URL, buf);
+ curl_easy_setopt(handle, CURLOPT_WRITEFUNCTION, aws_write_callback);
+ curl_easy_setopt(handle, CURLOPT_WRITEDATA, frame);
+ curl_easy_setopt(handle, CURLOPT_VERBOSE, debug);
+ curl_easy_setopt(handle, CURLOPT_ERRORBUFFER, errbuf);
+
+ res = curl_easy_perform(handle);
+ if (res != CURLE_OK) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "download failed. err: %s\n",
+ curl_easy_strerror(res));
+ ret = -1;
+ len = strlen(errbuf);
+ if (len) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "curl failure %s", errbuf);
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "curl error "
+ "%s\n",
+ curl_easy_strerror(res));
+ }
+ }
+
+ if (res == CURLE_OK) {
+ curl_easy_getinfo(handle, CURLINFO_RESPONSE_CODE, &responsecode);
+ gf_msg_debug(this->name, 0, "response code %ld", responsecode);
+ if (responsecode != 200) {
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "curl download failed");
+ }
+ }
+
+ curl_slist_free_all(slist);
+ curl_easy_cleanup(handle);
+
+out:
+ if (sign_req)
+ GF_FREE(sign_req);
+ if (date)
+ GF_FREE(date);
+ if (signature)
+ GF_FREE(signature);
+
+ return ret;
+}
+
+struct volume_options cs_options[] = {
+ {.key = {"s3plugin-seckey"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "aws secret key"},
+ {.key = {"s3plugin-keyid"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "aws key ID"
+
+ },
+ {.key = {"s3plugin-bucketid"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "aws bucketid"},
+ {.key = {"s3plugin-hostname"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "aws hostname e.g. s3.amazonaws.com"},
+ {.key = {NULL}},
+};
diff --git a/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3.h b/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3.h
new file mode 100644
index 00000000000..85ae669486b
--- /dev/null
+++ b/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3.h
@@ -0,0 +1,50 @@
+/*
+ Copyright (c) 2018 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+*/
+#ifndef _LIBAWS_H
+#define _LIBAWS_H
+
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/call-stub.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/syncop.h>
+#include <curl/curl.h>
+#include "cloudsync-common.h"
+#include "libcloudsyncs3-mem-types.h"
+
+char *
+aws_b64_encode(const unsigned char *input, int length);
+
+size_t
+aws_write_callback(void *dlbuf, size_t size, size_t nitems, void *mainframe);
+
+int
+aws_download_s3(call_frame_t *frame, void *config);
+
+int
+aws_dlwritev_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata);
+
+void *
+aws_init(xlator_t *this);
+
+int
+aws_reconfigure(xlator_t *this, dict_t *options);
+
+char *
+aws_form_request(char *resource, char **date, char *reqtype, char *bucketid,
+ char *filepath);
+char *
+aws_sign_request(char *const str, char *awssekey);
+
+void
+aws_fini(void *config);
+
+#endif
diff --git a/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3.sym b/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3.sym
new file mode 100644
index 00000000000..0bc273670d5
--- /dev/null
+++ b/xlators/features/cloudsync/src/cloudsync-plugins/src/cloudsyncs3/src/libcloudsyncs3.sym
@@ -0,0 +1 @@
+store_ops
diff --git a/xlators/encryption/rot-13/Makefile.am b/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/Makefile.am
index d471a3f9243..a985f42a877 100644
--- a/xlators/encryption/rot-13/Makefile.am
+++ b/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/Makefile.am
@@ -1,3 +1,3 @@
SUBDIRS = src
-CLEANFILES =
+CLEANFILES =
diff --git a/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/Makefile.am b/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/Makefile.am
new file mode 100644
index 00000000000..b512464f157
--- /dev/null
+++ b/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/Makefile.am
@@ -0,0 +1,12 @@
+csp_LTLIBRARIES = cloudsynccvlt.la
+cspdir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/cloudsync-plugins
+
+cloudsynccvlt_la_SOURCES = libcvlt.c $(top_srcdir)/xlators/features/cloudsync/src/cloudsync-common.c
+cloudsynccvlt_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+cloudsynccvlt_la_LDFLAGS = -module -avoid-version -export-symbols $(top_srcdir)/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcloudsynccvlt.sym
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src -I$(top_srcdir)/rpc/xdr/src -I$(top_builddir)/rpc/xdr/src
+noinst_HEADERS = archivestore.h libcvlt.h libcvlt-mem-types.h cvlt-messages.h
+AM_CFLAGS = -Wall -fno-strict-aliasing $(GF_CFLAGS) -I$(top_srcdir)/xlators/features/cloudsync/src
+CLEANFILES =
+
+EXTRA_DIST = libcloudsynccvlt.sym
diff --git a/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/archivestore.h b/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/archivestore.h
new file mode 100644
index 00000000000..7230ef77337
--- /dev/null
+++ b/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/archivestore.h
@@ -0,0 +1,203 @@
+/*
+ Copyright (c) 2018 Commvault Systems, Inc. <http://www.commvault.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+*/
+
+#ifndef __ARCHIVESTORE_H__
+#define __ARCHIVESTORE_H__
+
+#include <stdlib.h>
+#include <stddef.h>
+#include <stdint.h>
+#include <dlfcn.h>
+#include <uuid/uuid.h>
+
+#define CS_XATTR_ARCHIVE_UUID "trusted.cloudsync.uuid"
+#define CS_XATTR_PRODUCT_ID "trusted.cloudsync.product-id"
+#define CS_XATTR_STORE_ID "trusted.cloudsync.store-id"
+
+struct _archstore_methods;
+typedef struct _archstore_methods archstore_methods_t;
+
+struct _archstore_desc {
+ void *priv; /* Private field for store mgmt. */
+ /* To be used only by archive store*/
+};
+typedef struct _archstore_desc archstore_desc_t;
+
+struct _archstore_info {
+ char *id; /* Identifier for the archivestore */
+ uint32_t idlen; /* Length of identifier string */
+ char *prod; /* Name of the data mgmt. product */
+ uint32_t prodlen; /* Length of the product string */
+};
+typedef struct _archstore_info archstore_info_t;
+
+struct _archstore_fileinfo {
+ uuid_t uuid; /* uuid of the file */
+ char *path; /* file path */
+ uint32_t pathlength; /* length of file path */
+};
+typedef struct _archstore_fileinfo archstore_fileinfo_t;
+
+struct _app_callback_info {
+ archstore_info_t *src_archstore;
+ archstore_fileinfo_t *src_archfile;
+ archstore_info_t *dest_archstore;
+ archstore_fileinfo_t *dest_archfile;
+};
+typedef struct _app_callback_info app_callback_info_t;
+
+typedef void (*app_callback_t)(archstore_desc_t *, app_callback_info_t *,
+ void *, int64_t, int32_t);
+
+enum _archstore_scan_type { FULL = 1, INCREMENTAL = 2 };
+typedef enum _archstore_scan_type archstore_scan_type_t;
+
+typedef int32_t archstore_errno_t;
+
+/*
+ * Initialize archive store.
+ * arg1 pointer to structure containing archive store information
+ * arg2 error number if any generated during the initialization
+ * arg3 name of the log file
+ */
+typedef int32_t (*init_archstore_t)(archstore_desc_t *, archstore_errno_t *,
+ const char *);
+
+/*
+ * Clean up archive store.
+ * arg1 pointer to structure containing archive store information
+ * arg2 error number if any generated during the cleanup
+ */
+typedef int32_t (*term_archstore_t)(archstore_desc_t *, archstore_errno_t *);
+
+/*
+ * Read the contents of the file from archive store
+ * arg1 pointer to structure containing archive store description
+ * arg2 pointer to structure containing archive store information
+ * arg3 pointer to structure containing information about file to be read
+ * arg4 offset in the file from which data should be read
+ * arg5 buffer where the data should be read
+ * arg6 number of bytes of data to be read
+ * arg7 error number if any generated during the read from file
+ * arg8 callback handler to be invoked after the data is read
+ * arg9 cookie to be passed when callback is invoked
+ */
+typedef int32_t (*read_archstore_t)(archstore_desc_t *, archstore_info_t *,
+ archstore_fileinfo_t *, off_t, char *,
+ size_t, archstore_errno_t *, app_callback_t,
+ void *);
+
+/*
+ * Restore the contents of the file from archive store
+ * This is basically in-place restore
+ * arg1 pointer to structure containing archive store description
+ * arg2 pointer to structure containing archive store information
+ * arg3 pointer to structure containing information about file to be restored
+ * arg4 error number if any generated during the file restore
+ * arg5 callback to be invoked after the file is restored
+ * arg6 cookie to be passed when callback is invoked
+ */
+typedef int32_t (*recall_archstore_t)(archstore_desc_t *, archstore_info_t *,
+ archstore_fileinfo_t *,
+ archstore_errno_t *, app_callback_t,
+ void *);
+
+/*
+ * Restore the contents of the file from archive store to a different store
+ * This is basically out-of-place restore
+ * arg1 pointer to structure containing archive store description
+ * arg2 pointer to structure containing source archive store information
+ * arg3 pointer to structure containing information about file to be restored
+ * arg4 pointer to structure containing destination archive store information
+ * arg5 pointer to structure containing information about the location to
+ which the file will be restored
+ * arg6 error number if any generated during the file restore
+ * arg7 callback to be invoked after the file is restored
+ * arg8 cookie to be passed when callback is invoked
+ */
+typedef int32_t (*restore_archstore_t)(archstore_desc_t *, archstore_info_t *,
+ archstore_fileinfo_t *,
+ archstore_info_t *,
+ archstore_fileinfo_t *,
+ archstore_errno_t *, app_callback_t,
+ void *);
+
+/*
+ * Archive the contents of the file to archive store
+ * arg1 pointer to structure containing archive store description
+ * arg2 pointer to structure containing source archive store information
+ * arg3 pointer to structure containing information about files to be archived
+ * arg4 pointer to structure containing destination archive store information
+ * arg5 pointer to structure containing information about files that failed
+ * to be archived
+ * arg6 error number if any generated during the file archival
+ * arg7 callback to be invoked after the file is archived
+ * arg8 cookie to be passed when callback is invoked
+ */
+typedef int32_t (*archive_archstore_t)(archstore_desc_t *, archstore_info_t *,
+ archstore_fileinfo_t *,
+ archstore_info_t *,
+ archstore_fileinfo_t *,
+ archstore_errno_t *, app_callback_t,
+ void *);
+
+/*
+ * Backup list of files provided in the input file
+ * arg1 pointer to structure containing archive store description
+ * arg2 pointer to structure containing source archive store information
+ * arg3 pointer to structure containing information about files to be backed up
+ * arg4 pointer to structure containing destination archive store information
+ * arg5 pointer to structure containing information about files that failed
+ * to be backed up
+ * arg6 error number if any generated during the file archival
+ * arg7 callback to be invoked after the file is archived
+ * arg8 cookie to be passed when callback is invoked
+ */
+typedef int32_t (*backup_archstore_t)(archstore_desc_t *, archstore_info_t *,
+ archstore_fileinfo_t *,
+ archstore_info_t *,
+ archstore_fileinfo_t *,
+ archstore_errno_t *, app_callback_t,
+ void *);
+
+/*
+ * Scan the contents of a store and determine the files which need to be
+ * backed up.
+ * arg1 pointer to structure containing archive store description
+ * arg2 pointer to structure containing archive store information
+ * arg3 type of scan whether full or incremental
+ * arg4 path to file that contains list of files to be backed up
+ * arg5 error number if any generated during scan operation
+ */
+typedef int32_t (*scan_archstore_t)(archstore_desc_t *, archstore_info_t *,
+ archstore_scan_type_t, char *,
+ archstore_errno_t *);
+
+struct _archstore_methods {
+ init_archstore_t init;
+ term_archstore_t fini;
+ backup_archstore_t backup;
+ archive_archstore_t archive;
+ scan_archstore_t scan;
+ restore_archstore_t restore;
+ recall_archstore_t recall;
+ read_archstore_t read;
+};
+
+typedef int (*get_archstore_methods_t)(archstore_methods_t *);
+
+/*
+ * Single function that will be invoked by applications for extracting
+ * the function pointers to all data management functions.
+ */
+int32_t
+get_archstore_methods(archstore_methods_t *);
+
+#endif /* End of __ARCHIVESTORE_H__ */
diff --git a/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/cvlt-messages.h b/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/cvlt-messages.h
new file mode 100644
index 00000000000..57c9aa77da0
--- /dev/null
+++ b/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/cvlt-messages.h
@@ -0,0 +1,30 @@
+/*
+ Copyright (c) 2015 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+ */
+
+#ifndef _CVLT_MESSAGES_H_
+#define _CVLT_MESSAGES_H_
+
+#include <glusterfs/glfs-message-id.h>
+
+/* To add new message IDs, append new identifiers at the end of the list.
+ *
+ * Never remove a message ID. If it's not used anymore, you can rename it or
+ * leave it as it is, but not delete it. This is to prevent reutilization of
+ * IDs by other messages.
+ *
+ * The component name must match one of the entries defined in
+ * glfs-message-id.h.
+ */
+
+GLFS_MSGID(CVLT, CVLT_EXTRACTION_FAILED, CVLT_FREE,
+ CVLT_RESOURCE_ALLOCATION_FAILED, CVLT_RESTORE_FAILED,
+ CVLT_READ_FAILED, CVLT_NO_MEMORY, CVLT_DLOPEN_FAILED);
+
+#endif /* !_CVLT_MESSAGES_H_ */
diff --git a/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcloudsynccvlt.sym b/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcloudsynccvlt.sym
new file mode 100644
index 00000000000..0bc273670d5
--- /dev/null
+++ b/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcloudsynccvlt.sym
@@ -0,0 +1 @@
+store_ops
diff --git a/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcvlt-mem-types.h b/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcvlt-mem-types.h
new file mode 100644
index 00000000000..c24fab8bfe7
--- /dev/null
+++ b/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcvlt-mem-types.h
@@ -0,0 +1,19 @@
+/*
+ * Copyright (c) 2018 Commvault Systems, Inc. <http://www.commvault.com>
+ * This file is part of GlusterFS.
+ *
+ * This file is licensed to you under your choice of the GNU Lesser
+ * General Public License, version 3 or any later version (LGPLv3 or
+ * later), or the GNU General Public License, version 2 (GPLv2), in all
+ * cases as published by the Free Software Foundation.
+ */
+
+#ifndef __LIBCVLT_MEM_TYPES_H__
+#define __LIBCVLT_MEM_TYPES_H__
+
+#include <glusterfs/mem-types.h>
+enum libcvlt_mem_types_ {
+ gf_libcvlt_mt_cvlt_private_t = gf_common_mt_end + 1,
+ gf_libcvlt_mt_end
+};
+#endif /* __LIBCVLT_MEM_TYPES_H__ */
diff --git a/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcvlt.c b/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcvlt.c
new file mode 100644
index 00000000000..5b7272bb448
--- /dev/null
+++ b/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcvlt.c
@@ -0,0 +1,842 @@
+#include <stdlib.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/glusterfs.h>
+#include "libcvlt.h"
+#include "cloudsync-common.h"
+#include "cvlt-messages.h"
+
+#define LIBARCHIVE_SO "libopenarchive.so"
+#define ALIGN_SIZE 4096
+#define CVLT_TRAILER "cvltv1"
+
+store_methods_t store_ops = {
+ .fop_download = cvlt_download,
+ .fop_init = cvlt_init,
+ .fop_reconfigure = cvlt_reconfigure,
+ .fop_fini = cvlt_fini,
+ .fop_remote_read = cvlt_read,
+};
+
+static const int32_t num_req = 32;
+static const int32_t num_iatt = 32;
+static char *plugin = "cvlt_cloudSync";
+
+int32_t
+mem_acct_init(xlator_t *this)
+{
+ int ret = -1;
+
+ if (!this)
+ return ret;
+
+ ret = xlator_mem_acct_init(this, gf_libcvlt_mt_end + 1);
+
+ if (ret != 0) {
+ return ret;
+ }
+
+ return ret;
+}
+
+static void
+cvlt_free_resources(archive_t *arch)
+{
+ /*
+ * We will release all the resources that were allocated by the xlator.
+ * Check whether there are any buffers which have not been released
+ * back to a mempool.
+ */
+
+ if (arch->handle) {
+ dlclose(arch->handle);
+ }
+
+ if (arch->iobuf_pool) {
+ iobuf_pool_destroy(arch->iobuf_pool);
+ }
+
+ if (arch->req_pool) {
+ mem_pool_destroy(arch->req_pool);
+ arch->req_pool = NULL;
+ }
+
+ return;
+}
+
+static int32_t
+cvlt_extract_store_fops(xlator_t *this, archive_t *arch)
+{
+ int32_t op_ret = -1;
+ get_archstore_methods_t get_archstore_methods;
+
+ /*
+ * libopenarchive.so defines methods for performing data management
+ * operations. We will extract the methods from library and these
+ * methods will be invoked for moving data between glusterfs volume
+ * and the data management product.
+ */
+
+ VALIDATE_OR_GOTO(arch, err);
+
+ arch->handle = dlopen(LIBARCHIVE_SO, RTLD_NOW);
+ if (!arch->handle) {
+ gf_msg(plugin, GF_LOG_ERROR, 0, CVLT_DLOPEN_FAILED,
+ " failed to open %s ", LIBARCHIVE_SO);
+ return op_ret;
+ }
+
+ dlerror(); /* Clear any existing error */
+
+ get_archstore_methods = dlsym(arch->handle, "get_archstore_methods");
+ if (!get_archstore_methods) {
+ gf_msg(plugin, GF_LOG_ERROR, 0, CVLT_EXTRACTION_FAILED,
+ " Error extracting get_archstore_methods()");
+ dlclose(arch->handle);
+ arch->handle = NULL;
+ return op_ret;
+ }
+
+ op_ret = get_archstore_methods(&(arch->fops));
+ if (op_ret) {
+ gf_msg(plugin, GF_LOG_ERROR, 0, CVLT_EXTRACTION_FAILED,
+ " Failed to extract methods in get_archstore_methods");
+ dlclose(arch->handle);
+ arch->handle = NULL;
+ return op_ret;
+ }
+
+err:
+ return op_ret;
+}
+
+static int32_t
+cvlt_alloc_resources(xlator_t *this, archive_t *arch, int num_req, int num_iatt)
+{
+ /*
+ * Initialize information about all the memory pools that will be
+ * used by this xlator.
+ */
+ arch->nreqs = 0;
+
+ arch->req_pool = NULL;
+
+ arch->handle = NULL;
+ arch->xl = this;
+
+ arch->req_pool = mem_pool_new(cvlt_request_t, num_req);
+ if (!arch->req_pool) {
+ goto err;
+ }
+
+ arch->iobuf_pool = iobuf_pool_new();
+ if (!arch->iobuf_pool) {
+ goto err;
+ }
+
+ if (cvlt_extract_store_fops(this, arch)) {
+ goto err;
+ }
+
+ return 0;
+
+err:
+
+ return -1;
+}
+
+static void
+cvlt_req_init(cvlt_request_t *req)
+{
+ sem_init(&(req->sem), 0, 0);
+
+ return;
+}
+
+static void
+cvlt_req_destroy(cvlt_request_t *req)
+{
+ if (req->iobuf) {
+ iobuf_unref(req->iobuf);
+ }
+
+ if (req->iobref) {
+ iobref_unref(req->iobref);
+ }
+
+ sem_destroy(&(req->sem));
+
+ return;
+}
+
+static cvlt_request_t *
+cvlt_alloc_req(archive_t *arch)
+{
+ cvlt_request_t *reqptr = NULL;
+
+ if (!arch) {
+ goto err;
+ }
+
+ if (arch->req_pool) {
+ reqptr = mem_get0(arch->req_pool);
+ if (reqptr) {
+ cvlt_req_init(reqptr);
+ }
+ }
+
+ if (reqptr) {
+ LOCK(&(arch->lock));
+ arch->nreqs++;
+ UNLOCK(&(arch->lock));
+ }
+
+err:
+ return reqptr;
+}
+
+static int32_t
+cvlt_free_req(archive_t *arch, cvlt_request_t *reqptr)
+{
+ if (!reqptr) {
+ goto err;
+ }
+
+ if (!arch) {
+ goto err;
+ }
+
+ if (arch->req_pool) {
+ /*
+ * Free the request resources if they exist.
+ */
+
+ cvlt_req_destroy(reqptr);
+ mem_put(reqptr);
+
+ LOCK(&(arch->lock));
+ arch->nreqs--;
+ UNLOCK(&(arch->lock));
+ }
+
+ return 0;
+
+err:
+ return -1;
+}
+
+static int32_t
+cvlt_init_xlator(xlator_t *this, archive_t *arch, int num_req, int num_iatt)
+{
+ int32_t ret = -1;
+ int32_t errnum = -1;
+ int32_t locked = 0;
+
+ /*
+ * Perform all the initializations needed for brining up the xlator.
+ */
+ if (!arch) {
+ goto err;
+ }
+
+ LOCK_INIT(&(arch->lock));
+ LOCK(&(arch->lock));
+
+ locked = 1;
+
+ ret = cvlt_alloc_resources(this, arch, num_req, num_iatt);
+
+ if (ret) {
+ goto err;
+ }
+
+ /*
+ * Now that the fops have been extracted initialize the store
+ */
+ ret = arch->fops.init(&(arch->descinfo), &errnum, plugin);
+ if (ret) {
+ goto err;
+ }
+
+ UNLOCK(&(arch->lock));
+ locked = 0;
+ ret = 0;
+
+ return ret;
+
+err:
+ if (arch) {
+ cvlt_free_resources(arch);
+
+ if (locked) {
+ UNLOCK(&(arch->lock));
+ }
+ }
+
+ return ret;
+}
+
+static int32_t
+cvlt_term_xlator(archive_t *arch)
+{
+ int32_t errnum = -1;
+
+ if (!arch) {
+ goto err;
+ }
+
+ LOCK(&(arch->lock));
+
+ /*
+ * Release the resources that have been allocated inside store
+ */
+ arch->fops.fini(&(arch->descinfo), &errnum);
+
+ cvlt_free_resources(arch);
+
+ UNLOCK(&(arch->lock));
+
+ GF_FREE(arch);
+
+ return 0;
+
+err:
+ return -1;
+}
+
+static int32_t
+cvlt_init_store_info(archive_t *priv, archstore_info_t *store_info)
+{
+ if (!store_info) {
+ return -1;
+ }
+
+ store_info->prod = priv->product_id;
+ store_info->prodlen = strlen(priv->product_id);
+
+ store_info->id = priv->store_id;
+ store_info->idlen = strlen(priv->store_id);
+
+ return 0;
+}
+
+static int32_t
+cvlt_init_file_info(cs_loc_xattr_t *xattr, archstore_fileinfo_t *file_info)
+{
+ if (!xattr || !file_info) {
+ return -1;
+ }
+
+ gf_uuid_copy(file_info->uuid, xattr->uuid);
+ file_info->path = xattr->file_path;
+ file_info->pathlength = strlen(xattr->file_path);
+
+ return 0;
+}
+
+static int32_t
+cvlt_init_gluster_store_info(cs_loc_xattr_t *xattr,
+ archstore_info_t *store_info)
+{
+ static char *product = "glusterfs";
+
+ if (!xattr || !store_info) {
+ return -1;
+ }
+
+ store_info->prod = product;
+ store_info->prodlen = strlen(product);
+
+ store_info->id = xattr->volname;
+ store_info->idlen = strlen(xattr->volname);
+
+ return 0;
+}
+
+static int32_t
+cvlt_init_gluster_file_info(cs_loc_xattr_t *xattr,
+ archstore_fileinfo_t *file_info)
+{
+ if (!xattr || !file_info) {
+ return -1;
+ }
+
+ gf_uuid_copy(file_info->uuid, xattr->gfid);
+ file_info->path = xattr->file_path;
+ file_info->pathlength = strlen(xattr->file_path);
+
+ return 0;
+}
+
+static void
+cvlt_copy_stat_info(struct iatt *buf, cs_size_xattr_t *xattrs)
+{
+ /*
+ * If the file was archived then the reported size will not be a
+ * correct one. We need to fix this.
+ */
+ if (buf && xattrs) {
+ buf->ia_size = xattrs->size;
+ buf->ia_blksize = xattrs->blksize;
+ buf->ia_blocks = xattrs->blocks;
+ }
+
+ return;
+}
+
+static void
+cvlt_readv_complete(archstore_desc_t *desc, app_callback_info_t *cbkinfo,
+ void *cookie, int64_t op_ret, int32_t op_errno)
+{
+ struct iovec iov;
+ xlator_t *this = NULL;
+ struct iatt postbuf = {
+ 0,
+ };
+ call_frame_t *frame = NULL;
+ cvlt_request_t *req = (cvlt_request_t *)cookie;
+ cs_local_t *local = NULL;
+ cs_private_t *cspriv = NULL;
+ archive_t *priv = NULL;
+
+ frame = req->frame;
+ this = frame->this;
+ local = frame->local;
+
+ cspriv = this->private;
+ priv = (archive_t *)cspriv->stores->config;
+
+ if (strcmp(priv->trailer, CVLT_TRAILER)) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ gf_msg_debug(plugin, 0,
+ " Read callback invoked offset:%" PRIu64 "bytes: %" PRIu64
+ " op : %d ret : %" PRId64 " errno : %d",
+ req->offset, req->bytes, req->op_type, op_ret, op_errno);
+
+ if (op_ret < 0) {
+ goto out;
+ }
+
+ req->iobref = iobref_new();
+ if (!req->iobref) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ iobref_add(req->iobref, req->iobuf);
+ iov.iov_base = iobuf_ptr(req->iobuf);
+ iov.iov_len = op_ret;
+
+ cvlt_copy_stat_info(&postbuf, &(req->szxattr));
+
+ /*
+ * Hack to notify higher layers of EOF.
+ */
+ if (!postbuf.ia_size || (req->offset + iov.iov_len >= postbuf.ia_size)) {
+ gf_msg_debug(plugin, 0, " signalling end-of-file for uuid=%s",
+ uuid_utoa(req->file_info.uuid));
+ op_errno = ENOENT;
+ }
+
+out:
+
+ STACK_UNWIND_STRICT(readv, frame, op_ret, op_errno, &iov, 1, &postbuf,
+ req->iobref, local->xattr_rsp);
+
+ cvlt_free_req(priv, req);
+
+ return;
+}
+
+static void
+cvlt_download_complete(archstore_desc_t *store, app_callback_info_t *cbk_info,
+ void *cookie, int64_t ret, int errcode)
+{
+ cvlt_request_t *req = (cvlt_request_t *)cookie;
+
+ gf_msg_debug(plugin, 0,
+ " Download callback invoked ret : %" PRId64 " errno : %d",
+ ret, errcode);
+
+ req->op_ret = ret;
+ req->op_errno = errcode;
+ sem_post(&(req->sem));
+
+ return;
+}
+
+void *
+cvlt_init(xlator_t *this)
+{
+ int ret = 0;
+ archive_t *priv = NULL;
+
+ if (!this->children || this->children->next) {
+ gf_msg(plugin, GF_LOG_ERROR, ENOMEM, 0,
+ "should have exactly one child");
+ ret = -1;
+ goto out;
+ }
+
+ if (!this->parents) {
+ gf_msg(plugin, GF_LOG_ERROR, ENOMEM, 0,
+ "dangling volume. check volfile");
+ ret = -1;
+ goto out;
+ }
+
+ priv = GF_CALLOC(1, sizeof(archive_t), gf_libcvlt_mt_cvlt_private_t);
+ if (!priv) {
+ ret = -1;
+ goto out;
+ }
+
+ priv->trailer = CVLT_TRAILER;
+ if (cvlt_init_xlator(this, priv, num_req, num_iatt)) {
+ gf_msg(plugin, GF_LOG_ERROR, ENOMEM, 0, "xlator init failed");
+ ret = -1;
+ goto out;
+ }
+
+ GF_OPTION_INIT("cloudsync-store-id", priv->store_id, str, out);
+ GF_OPTION_INIT("cloudsync-product-id", priv->product_id, str, out);
+
+ gf_msg(plugin, GF_LOG_INFO, 0, 0,
+ "store id is : %s "
+ "product id is : %s.",
+ priv->store_id, priv->product_id);
+out:
+ if (ret == -1) {
+ cvlt_term_xlator(priv);
+ return (NULL);
+ }
+ return priv;
+}
+
+int
+cvlt_reconfigure(xlator_t *this, dict_t *options)
+{
+ cs_private_t *cspriv = NULL;
+ archive_t *priv = NULL;
+
+ cspriv = this->private;
+ priv = (archive_t *)cspriv->stores->config;
+
+ if (strcmp(priv->trailer, CVLT_TRAILER))
+ goto out;
+
+ GF_OPTION_RECONF("cloudsync-store-id", priv->store_id, options, str, out);
+
+ GF_OPTION_RECONF("cloudsync-product-id", priv->product_id, options, str,
+ out);
+ gf_msg_debug(plugin, 0,
+ "store id is : %s "
+ "product id is : %s.",
+ priv->store_id, priv->product_id);
+ return 0;
+out:
+ return -1;
+}
+
+void
+cvlt_fini(void *config)
+{
+ archive_t *priv = NULL;
+
+ priv = (archive_t *)config;
+
+ if (strcmp(priv->trailer, CVLT_TRAILER))
+ return;
+
+ cvlt_term_xlator(priv);
+ gf_msg(plugin, GF_LOG_INFO, 0, CVLT_FREE, " released xlator resources");
+ return;
+}
+
+int
+cvlt_download(call_frame_t *frame, void *config)
+{
+ archive_t *parch = NULL;
+ cs_local_t *local = frame->local;
+ cs_loc_xattr_t *locxattr = local->xattrinfo.lxattr;
+ cvlt_request_t *req = NULL;
+ archstore_info_t dest_storeinfo;
+ archstore_fileinfo_t dest_fileinfo;
+ int32_t op_ret, op_errno;
+
+ parch = (archive_t *)config;
+
+ if (strcmp(parch->trailer, CVLT_TRAILER)) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ gf_msg_debug(plugin, 0, " download invoked for uuid = %s gfid=%s ",
+ locxattr->uuid, uuid_utoa(locxattr->gfid));
+
+ if (!(parch->fops.restore)) {
+ op_errno = ELIBBAD;
+ goto err;
+ }
+
+ /*
+ * Download needs to be processed. Allocate a request.
+ */
+ req = cvlt_alloc_req(parch);
+
+ if (!req) {
+ gf_msg(plugin, GF_LOG_ERROR, ENOMEM, CVLT_RESOURCE_ALLOCATION_FAILED,
+ " failed to allocated request for gfid=%s",
+ uuid_utoa(locxattr->gfid));
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ /*
+ * Initialize the request object.
+ */
+ req->op_type = CVLT_RESTORE_OP;
+ req->frame = frame;
+
+ /*
+ * The file is currently residing inside a data management store.
+ * To restore the file contents we need to provide the information
+ * about data management store.
+ */
+ op_ret = cvlt_init_store_info(parch, &(req->store_info));
+ if (op_ret < 0) {
+ gf_msg(plugin, GF_LOG_ERROR, 0, CVLT_EXTRACTION_FAILED,
+ " failed to extract store info for gfid=%s",
+ uuid_utoa(locxattr->gfid));
+ goto err;
+ }
+
+ op_ret = cvlt_init_file_info(locxattr, &(req->file_info));
+ if (op_ret < 0) {
+ gf_msg(plugin, GF_LOG_ERROR, 0, CVLT_EXTRACTION_FAILED,
+ " failed to extract file info for gfid=%s",
+ uuid_utoa(locxattr->gfid));
+ goto err;
+ }
+
+ /*
+ * We need to perform in-place restore of the file from data management
+ * store to gusterfs volume.
+ */
+ op_ret = cvlt_init_gluster_store_info(locxattr, &dest_storeinfo);
+ if (op_ret < 0) {
+ gf_msg(plugin, GF_LOG_ERROR, 0, CVLT_EXTRACTION_FAILED,
+ " failed to extract destination store info for gfid=%s",
+ uuid_utoa(locxattr->gfid));
+ goto err;
+ }
+
+ op_ret = cvlt_init_gluster_file_info(locxattr, &dest_fileinfo);
+ if (op_ret < 0) {
+ gf_msg(plugin, GF_LOG_ERROR, 0, CVLT_EXTRACTION_FAILED,
+ " failed to extract file info for gfid=%s",
+ uuid_utoa(locxattr->gfid));
+ goto err;
+ }
+
+ /*
+ * Submit the restore request.
+ */
+ op_ret = parch->fops.restore(&(parch->descinfo), &(req->store_info),
+ &(req->file_info), &dest_storeinfo,
+ &dest_fileinfo, &op_errno,
+ cvlt_download_complete, req);
+ if (op_ret < 0) {
+ gf_msg(plugin, GF_LOG_ERROR, 0, CVLT_RESTORE_FAILED,
+ " failed to restore file gfid=%s from data management store",
+ uuid_utoa(locxattr->gfid));
+ goto err;
+ }
+
+ /*
+ * Wait for the restore to complete.
+ */
+ sem_wait(&(req->sem));
+
+ if (req->op_ret < 0) {
+ gf_msg(plugin, GF_LOG_ERROR, 0, CVLT_RESTORE_FAILED,
+ " restored failed for gfid=%s", uuid_utoa(locxattr->gfid));
+ goto err;
+ }
+
+ if (req) {
+ cvlt_free_req(parch, req);
+ }
+
+ return 0;
+
+err:
+
+ if (req) {
+ cvlt_free_req(parch, req);
+ }
+
+ return -1;
+}
+
+int
+cvlt_read(call_frame_t *frame, void *config)
+{
+ int32_t op_ret = -1;
+ int32_t op_errno = 0;
+ archive_t *parch = NULL;
+ cvlt_request_t *req = NULL;
+ struct iovec iov = {
+ 0,
+ };
+ struct iobref *iobref;
+ size_t size = 0;
+ off_t off = 0;
+
+ cs_local_t *local = frame->local;
+ cs_loc_xattr_t *locxattr = local->xattrinfo.lxattr;
+
+ size = local->xattrinfo.size;
+ off = local->xattrinfo.offset;
+
+ parch = (archive_t *)config;
+
+ if (strcmp(parch->trailer, CVLT_TRAILER)) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ gf_msg_debug(plugin, 0,
+ " read invoked for gfid = %s offset = %" PRIu64
+ " file_size = %" PRIu64,
+ uuid_utoa(locxattr->gfid), off, local->stbuf.ia_size);
+
+ if (off >= local->stbuf.ia_size) {
+ /*
+ * Hack to notify higher layers of EOF.
+ */
+
+ op_errno = ENOENT;
+ op_ret = 0;
+
+ gf_msg(plugin, GF_LOG_ERROR, 0, CVLT_READ_FAILED,
+ " reporting end-of-file for gfid=%s", uuid_utoa(locxattr->gfid));
+
+ goto err;
+ }
+
+ if (!size) {
+ op_errno = EINVAL;
+
+ gf_msg(plugin, GF_LOG_ERROR, 0, CVLT_READ_FAILED,
+ " zero size read attempted on gfid=%s",
+ uuid_utoa(locxattr->gfid));
+ goto err;
+ }
+
+ if (!(parch->fops.read)) {
+ op_errno = ELIBBAD;
+ goto err;
+ }
+
+ /*
+ * The read request need to be processed. Allocate a request.
+ */
+ req = cvlt_alloc_req(parch);
+
+ if (!req) {
+ gf_msg(plugin, GF_LOG_ERROR, ENOMEM, CVLT_NO_MEMORY,
+ " failed to allocated request for gfid=%s",
+ uuid_utoa(locxattr->gfid));
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ req->iobuf = iobuf_get_page_aligned(parch->iobuf_pool, size, ALIGN_SIZE);
+ if (!req->iobuf) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ /*
+ * Initialize the request object.
+ */
+ req->op_type = CVLT_READ_OP;
+ req->offset = off;
+ req->bytes = size;
+ req->frame = frame;
+ req->szxattr.size = local->stbuf.ia_size;
+ req->szxattr.blocks = local->stbuf.ia_blocks;
+ req->szxattr.blksize = local->stbuf.ia_blksize;
+
+ /*
+ * The file is currently residing inside a data management store.
+ * To read the file contents we need to provide the information
+ * about data management store.
+ */
+ op_ret = cvlt_init_store_info(parch, &(req->store_info));
+ if (op_ret < 0) {
+ gf_msg(plugin, GF_LOG_ERROR, 0, CVLT_EXTRACTION_FAILED,
+ " failed to extract store info for gfid=%s"
+ " offset=%" PRIu64 " size=%" GF_PRI_SIZET
+ ", "
+ " buf=%p",
+ uuid_utoa(locxattr->gfid), off, size, req->iobuf->ptr);
+ goto err;
+ }
+
+ op_ret = cvlt_init_file_info(locxattr, &(req->file_info));
+ if (op_ret < 0) {
+ gf_msg(plugin, GF_LOG_ERROR, 0, CVLT_EXTRACTION_FAILED,
+ " failed to extract file info for gfid=%s"
+ " offset=%" PRIu64 " size=%" GF_PRI_SIZET
+ ", "
+ " buf=%p",
+ uuid_utoa(locxattr->gfid), off, size, req->iobuf->ptr);
+ goto err;
+ }
+
+ /*
+ * Submit the read request.
+ */
+ op_ret = parch->fops.read(&(parch->descinfo), &(req->store_info),
+ &(req->file_info), off, req->iobuf->ptr, size,
+ &op_errno, cvlt_readv_complete, req);
+
+ if (op_ret < 0) {
+ gf_msg(plugin, GF_LOG_ERROR, 0, CVLT_EXTRACTION_FAILED,
+ " read failed on gfid=%s"
+ " offset=%" PRIu64 " size=%" GF_PRI_SIZET
+ ", "
+ " buf=%p",
+ uuid_utoa(locxattr->gfid), off, size, req->iobuf->ptr);
+ goto err;
+ }
+
+ return 0;
+
+err:
+
+ iobref = iobref_new();
+ gf_msg_debug(plugin, 0, " read unwinding stack op_ret = %d, op_errno = %d",
+ op_ret, op_errno);
+
+ STACK_UNWIND_STRICT(readv, frame, op_ret, op_errno, &iov, 1,
+ &(local->stbuf), iobref, local->xattr_rsp);
+
+ if (iobref) {
+ iobref_unref(iobref);
+ }
+
+ if (req) {
+ cvlt_free_req(parch, req);
+ }
+
+ return 0;
+}
diff --git a/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcvlt.h b/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcvlt.h
new file mode 100644
index 00000000000..c45ac948f6c
--- /dev/null
+++ b/xlators/features/cloudsync/src/cloudsync-plugins/src/cvlt/src/libcvlt.h
@@ -0,0 +1,84 @@
+/*
+ Copyright (c) 2018 Commvault Systems, Inc. <http://www.commvault.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+*/
+#ifndef _LIBCVLT_H
+#define _LIBCVLT_H
+
+#include <semaphore.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/call-stub.h>
+#include <glusterfs/syncop.h>
+#include <glusterfs/compat-errno.h>
+#include "cloudsync-common.h"
+#include "libcvlt-mem-types.h"
+#include "archivestore.h"
+
+enum _cvlt_op {
+ CVLT_READ_OP = 1,
+ CVLT_WRITE_OP = 2,
+ CVLT_RESTORE_OP = 3,
+ CVLT_ARCHIVE_OP = 4,
+ CVLT_LOOKUP_OP = 5,
+ CVLT_XATTR_OP = 6,
+ CVLT_STAT_OP = 7,
+ CVLT_FSTAT_op = 8,
+ CVLT_UNDEF_OP = 127
+};
+typedef enum _cvlt_op cvlt_op_t;
+
+struct _archive;
+struct _cvlt_request {
+ uint64_t offset;
+ uint64_t bytes;
+ struct iobuf *iobuf;
+ struct iobref *iobref;
+ call_frame_t *frame;
+ cvlt_op_t op_type;
+ int32_t op_ret;
+ int32_t op_errno;
+ xlator_t *this;
+ sem_t sem;
+ archstore_info_t store_info;
+ archstore_fileinfo_t file_info;
+ cs_size_xattr_t szxattr;
+};
+typedef struct _cvlt_request cvlt_request_t;
+
+struct _archive {
+ gf_lock_t lock; /* lock for controlling access */
+ xlator_t *xl; /* xlator */
+ void *handle; /* handle returned from dlopen */
+ int32_t nreqs; /* num requests active */
+ struct mem_pool *req_pool; /* pool for requests */
+ struct iobuf_pool *iobuf_pool; /* iobuff pool */
+ archstore_desc_t descinfo; /* Archive store descriptor info */
+ archstore_methods_t fops; /* function pointers */
+ char *product_id;
+ char *store_id;
+ char *trailer;
+};
+typedef struct _archive archive_t;
+
+void *
+cvlt_init(xlator_t *);
+
+int
+cvlt_reconfigure(xlator_t *, dict_t *);
+
+void
+cvlt_fini(void *);
+
+int
+cvlt_download(call_frame_t *, void *);
+
+int
+cvlt_read(call_frame_t *, void *);
+
+#endif
diff --git a/xlators/features/cloudsync/src/cloudsync.c b/xlators/features/cloudsync/src/cloudsync.c
index ce3e41bff63..7f0b9e563b8 100644
--- a/xlators/features/cloudsync/src/cloudsync.c
+++ b/xlators/features/cloudsync/src/cloudsync.c
@@ -8,1666 +8,2069 @@
* cases as published by the Free Software Foundation.
*/
-#include "glusterfs.h"
-#include "xlator.h"
-#include "defaults.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
#include "cloudsync.h"
#include "cloudsync-common.h"
-#include "call-stub.h"
+#include <glusterfs/call-stub.h>
#include "cloudsync-autogen-fops.h"
+#include <string.h>
#include <dlfcn.h>
-void
-cs_cleanup_private (cs_private_t *priv)
+static void
+cs_cleanup_private(cs_private_t *priv)
{
- if (priv) {
- if (priv->stores) {
- priv->stores->fini (priv->stores->config);
- GF_FREE (priv->stores);
- }
-
- pthread_spin_destroy (&priv->lock);
- GF_FREE (priv);
+ if (priv) {
+ if (priv->stores) {
+ priv->stores->fini(priv->stores->config);
+ GF_FREE(priv->stores);
}
- return;
+ pthread_spin_destroy(&priv->lock);
+ GF_FREE(priv);
+ }
+
+ return;
}
+static struct cs_plugin plugins[] = {
+ {.name = "cloudsyncs3",
+ .library = "cloudsyncs3.so",
+ .description = "cloudsync s3 store."},
+#if defined(__linux__)
+ {.name = "cvlt",
+ .library = "cloudsynccvlt.so",
+ .description = "Commvault content store."},
+#endif
+ {.name = NULL},
+};
+
int
-cs_init (xlator_t *this)
-{
- cs_private_t *priv = NULL;
- gf_boolean_t per_vol = _gf_false;
- int ret = 0;
- char *libpath = ("libaws.so");
- store_methods_t *store_methods = NULL;
- void *handle = NULL;
-
- priv = GF_CALLOC (1, sizeof (*priv), gf_cs_mt_cs_private_t);
- if (!priv) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "insufficient memory");
- goto out;
+cs_init(xlator_t *this)
+{
+ cs_private_t *priv = NULL;
+ gf_boolean_t per_vol = _gf_false;
+ int ret = 0;
+ char *libpath = NULL;
+ store_methods_t *store_methods = NULL;
+ void *handle = NULL;
+ char *temp_str = NULL;
+ int index = 0;
+ char *libname = NULL;
+
+ priv = GF_CALLOC(1, sizeof(*priv), gf_cs_mt_cs_private_t);
+ if (!priv) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "insufficient memory");
+ goto out;
+ }
+
+ priv->this = this;
+
+ this->local_pool = mem_pool_new(cs_local_t, 512);
+ if (!this->local_pool) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, ENOMEM, "initialisation failed.");
+ ret = -1;
+ goto out;
+ }
+
+ this->private = priv;
+
+ GF_OPTION_INIT("cloudsync-remote-read", priv->remote_read, bool, out);
+
+ /* temp workaround. Should be configurable through glusterd*/
+ per_vol = _gf_true;
+
+ if (per_vol) {
+ if (dict_get_str_sizen(this->options, "cloudsync-storetype",
+ &temp_str) == 0) {
+ for (index = 0; plugins[index].name; index++) {
+ if (!strcmp(temp_str, plugins[index].name)) {
+ libname = plugins[index].library;
+ break;
+ }
+ }
+ } else {
+ ret = 0;
}
- priv->this = this;
+ if (!libname) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, 0, "no plugin enabled");
+ ret = 0;
+ goto out;
+ }
- this->local_pool = mem_pool_new (cs_local_t, 512);
- if (!this->local_pool) {
- gf_msg (this->name, GF_LOG_ERROR, 0, ENOMEM,
- "initialisation failed.");
- ret = -1;
- goto out;
+ ret = gf_asprintf(&libpath, "%s/%s", CS_PLUGINDIR, libname);
+ if (ret == -1) {
+ goto out;
}
- this->private = priv;
+ handle = dlopen(libpath, RTLD_NOW);
+ if (!handle) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, 0,
+ "could not "
+ "load the required library. %s",
+ dlerror());
+ ret = 0;
+ goto out;
+ } else {
+ gf_msg(this->name, GF_LOG_INFO, 0, 0,
+ "loading library:%s successful", libname);
+ }
- /* temp workaround. Should be configurable through glusterd*/
- per_vol = _gf_true;
+ priv->stores = GF_CALLOC(1, sizeof(struct cs_remote_stores),
+ gf_cs_mt_cs_remote_stores_t);
+ if (!priv->stores) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "Could not "
+ "allocate memory for priv->stores");
+ ret = -1;
+ goto out;
+ }
- if (per_vol) {
- /*TODO:Need to make it configurable. This is a temp workaround*/
- handle = dlopen (libpath, RTLD_NOW);
- if (!handle) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "could not load"
- " the required library. %s", dlerror ());
- goto out;
- } else {
- gf_msg (this->name, GF_LOG_INFO, 0, 0,
- "loading library:%s successful", libpath);
- }
+ (void)dlerror(); /* clear out previous error string */
+ /* load library methods */
+ store_methods = (store_methods_t *)dlsym(handle, "store_ops");
+ if (!store_methods) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "null store_methods %s",
+ dlerror());
+ ret = -1;
+ goto out;
+ }
- priv->stores = GF_CALLOC (1, sizeof (struct cs_remote_stores),
- gf_cs_mt_cs_remote_stores_t);
- if (!priv->stores) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "Could not "
- "allocate memory for priv->stores");
- ret = -1;
- goto out;
- }
+ (void)dlerror();
- (void) dlerror (); /* clear out previous error string */
+ if (priv->remote_read) {
+ priv->stores->rdfop = store_methods->fop_remote_read;
+ if (!priv->stores->rdfop) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "failed to get"
+ " read fop %s",
+ dlerror());
+ ret = -1;
+ goto out;
+ }
+ }
- /* load library methods */
- store_methods = (store_methods_t *) dlsym (handle, "store_ops");
- if (!store_methods) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0,
- "null store_methods %s", dlerror ());
- ret = -1;
- goto out;
- }
+ priv->stores->dlfop = store_methods->fop_download;
+ if (!priv->stores->dlfop) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "failed to get"
+ " download fop %s",
+ dlerror());
+ ret = -1;
+ goto out;
+ }
- (void) dlerror ();
+ (void)dlerror();
+ priv->stores->init = store_methods->fop_init;
+ if (!priv->stores->init) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "failed to get"
+ " init fop %s",
+ dlerror());
+ ret = -1;
+ goto out;
+ }
- priv->stores->dlfop = store_methods->fop_download;
- if (!priv->stores->dlfop) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "failed to get"
- " download fop %s", dlerror ());
- ret = -1;
- goto out;
- }
+ (void)dlerror();
+ priv->stores->reconfigure = store_methods->fop_reconfigure;
+ if (!priv->stores->reconfigure) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "failed to get"
+ " reconfigure fop %s",
+ dlerror());
+ ret = -1;
+ goto out;
+ }
- (void) dlerror ();
- priv->stores->init = store_methods->fop_init;
- if (!priv->stores->init) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "failed to get"
- " init fop %s", dlerror ());
- ret = -1;
- goto out;
- }
+ priv->stores->handle = handle;
- (void) dlerror ();
- priv->stores->reconfigure = store_methods->fop_reconfigure;
- if (!priv->stores->reconfigure) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "failed to get"
- " reconfigure fop %s", dlerror ());
- ret = -1;
- goto out;
- }
+ priv->stores->config = (void *)((priv->stores->init)(this));
+ if (!priv->stores->config) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "null config");
+ ret = -1;
+ goto out;
+ }
+ }
- priv->stores->handle = handle;
+ ret = 0;
- priv->stores->config = (void *) ((priv->stores->init) (this));
- if (!priv->stores->config) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "null config");
- ret = -1;
- goto out;
- }
+out:
+ if (ret == -1) {
+ if (this->local_pool) {
+ mem_pool_destroy(this->local_pool);
+ this->local_pool = NULL;
}
-out:
- if (ret == -1) {
- if (this->local_pool)
- mem_pool_destroy (this->local_pool);
+ cs_cleanup_private(priv);
- cs_cleanup_private (priv);
+ if (handle) {
+ dlclose(handle);
}
+ }
- return ret;
-}
+ GF_FREE(libpath);
-void
-cs_fini (xlator_t *this)
-{
- cs_private_t *priv = NULL;
- priv = this->private;
-
- cs_cleanup_private (priv);
+ return ret;
}
int
-cs_reconfigure (xlator_t *this, dict_t *options)
+cs_forget(xlator_t *this, inode_t *inode)
{
- cs_private_t *priv = NULL;
- int ret = 0;
+ uint64_t ctx_int = 0;
+ cs_inode_ctx_t *ctx = NULL;
- priv = this->private;
- if (!priv) {
- ret = -1;
- goto out;
- }
+ inode_ctx_del(inode, this, &ctx_int);
+ if (!ctx_int)
+ return 0;
- /* needed only for per volume configuration*/
- ret = priv->stores->reconfigure (this, options);
+ ctx = (cs_inode_ctx_t *)(uintptr_t)ctx_int;
-out:
- return ret;
+ GF_FREE(ctx);
+ return 0;
}
-int32_t
-cs_mem_acct_init (xlator_t *this)
+void
+cs_fini(xlator_t *this)
{
- int ret = -1;
+ cs_private_t *priv = NULL;
+ priv = this->private;
- GF_VALIDATE_OR_GOTO ("cloudsync", this, out);
+ cs_cleanup_private(priv);
+}
- ret = xlator_mem_acct_init (this, gf_cs_mt_end + 1);
+int
+cs_reconfigure(xlator_t *this, dict_t *options)
+{
+ cs_private_t *priv = NULL;
+ int ret = 0;
+
+ priv = this->private;
+ if (!priv) {
+ ret = -1;
+ goto out;
+ }
+
+ GF_OPTION_RECONF("cloudsync-remote-read", priv->remote_read, options, bool,
+ out);
+
+ /* needed only for per volume configuration*/
+ ret = priv->stores->reconfigure(this, options);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0,
- "Memory accounting init failed");
- return ret;
- }
out:
- return ret;
+ return ret;
}
int32_t
-cs_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
- dict_t *xdata)
+cs_mem_acct_init(xlator_t *this)
{
- gf_dirent_t *tmp = NULL;
- char *sxattr = NULL;
- uint64_t ia_size = 0;
- int ret = 0;
+ int ret = -1;
- list_for_each_entry (tmp, &entries->list, list) {
- ret = dict_get_str (tmp->dict, GF_CS_OBJECT_SIZE, &sxattr);
- if (ret) {
- gf_msg_trace (this->name, 0, "size xattr found");
- continue;
- }
+ GF_VALIDATE_OR_GOTO("cloudsync", this, out);
- ia_size = atoll (sxattr);
- tmp->d_stat.ia_size = ia_size;
- }
+ ret = xlator_mem_acct_init(this, gf_cs_mt_end + 1);
- STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno,
- entries, xdata);
- return 0;
+ if (ret != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "Memory accounting init failed");
+ return ret;
+ }
+out:
+ return ret;
}
-
int32_t
-cs_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t off, dict_t *xdata)
+cs_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *xdata)
{
- int ret = 0;
- int op_errno = ENOMEM;
+ int ret = 0;
+ int op_errno = ENOMEM;
+ if (!xdata) {
+ xdata = dict_new();
if (!xdata) {
- xdata = dict_new ();
- if (!xdata) {
- goto err;
- }
+ gf_msg(this->name, GF_LOG_ERROR, 0, ENOMEM,
+ "failed to create "
+ "dict");
+ goto err;
}
-
- ret = dict_set_int32 (xdata, GF_CS_OBJECT_SIZE, 1);
- if (ret) {
- goto err;
- }
-
- STACK_WIND (frame, cs_readdirp_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->readdirp,
- fd, size, off, xdata);
- return 0;
+ }
+
+ ret = dict_set_uint32(xdata, GF_CS_OBJECT_STATUS, 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "dict_set failed key:"
+ " %s",
+ GF_CS_OBJECT_STATUS);
+ goto err;
+ }
+
+ STACK_WIND(frame, default_readdirp_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdirp, fd, size, off, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (readdirp, frame, -1, op_errno,
- NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(readdirp, frame, -1, op_errno, NULL, NULL);
+ return 0;
}
-
int32_t
-cs_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+cs_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- cs_local_t *local = NULL;
- int ret = 0;
- uint64_t val = 0;
-
- local = frame->local;
-
- /* Do we need lock here? */
- local->call_cnt++;
-
- if (op_ret == -1) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "truncate failed");
- ret = dict_get_uint64 (xdata, GF_CS_OBJECT_STATUS, &val);
- if (ret == 0) {
- if (val == GF_CS_ERROR) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0,
- "could not get file state, unwinding");
- op_ret = -1;
- op_errno = EIO;
- goto unwind;
- } else {
- __cs_inode_ctx_update (this, local->loc.inode,
- val);
- gf_msg (this->name, GF_LOG_INFO, 0, 0,
- " state = %ld", val);
-
- if (local->call_cnt == 1 &&
- (val == GF_CS_REMOTE ||
- val == GF_CS_DOWNLOADING)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- 0, "will repair and download "
- "the file, current state : %ld",
- val);
- goto repair;
- } else {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0,
- "second truncate, Unwinding");
- goto unwind;
- }
- }
+ cs_local_t *local = NULL;
+ int ret = 0;
+ uint64_t val = 0;
+
+ local = frame->local;
+
+ local->call_cnt++;
+
+ if (op_ret == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "truncate failed");
+ ret = dict_get_uint64(xdata, GF_CS_OBJECT_STATUS, &val);
+ if (ret == 0) {
+ if (val == GF_CS_ERROR) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "could not get file state, unwinding");
+ op_ret = -1;
+ op_errno = EIO;
+ goto unwind;
+ } else {
+ __cs_inode_ctx_update(this, local->loc.inode, val);
+ gf_msg(this->name, GF_LOG_INFO, 0, 0, " state = %" PRIu64, val);
+
+ if (local->call_cnt == 1 &&
+ (val == GF_CS_REMOTE || val == GF_CS_DOWNLOADING)) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, 0,
+ "will repair and download "
+ "the file, current state : %" PRIu64,
+ val);
+ goto repair;
} else {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "file state "
- "could not be figured, unwinding");
- goto unwind;
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "second truncate, Unwinding");
+ goto unwind;
}
+ }
} else {
- /* successful write => file is local */
- __cs_inode_ctx_update (this, local->loc.inode, GF_CS_LOCAL);
- gf_msg (this->name, GF_LOG_INFO, 0, 0, "state : GF_CS_LOCAL"
- ", truncate successful");
-
- goto unwind;
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "file state "
+ "could not be figured, unwinding");
+ goto unwind;
}
+ } else {
+ /* successful write => file is local */
+ __cs_inode_ctx_update(this, local->loc.inode, GF_CS_LOCAL);
+ gf_msg(this->name, GF_LOG_INFO, 0, 0,
+ "state : GF_CS_LOCAL"
+ ", truncate successful");
+
+ goto unwind;
+ }
repair:
- ret = locate_and_execute (frame);
- if (ret) {
- goto unwind;
- }
+ ret = locate_and_execute(frame);
+ if (ret) {
+ goto unwind;
+ }
- return 0;
+ return 0;
unwind:
- CS_STACK_UNWIND (truncate, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
- return 0;
+ CS_STACK_UNWIND(truncate, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+ return 0;
}
-
int32_t
-cs_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
- dict_t *xdata)
+cs_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- int op_errno = -1;
- cs_local_t *local = NULL;
- int ret = 0;
- cs_inode_ctx_t *ctx = NULL;
- gf_cs_obj_state state = -1;
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
-
- local = cs_local_init (this, frame, loc, NULL, GF_FOP_TRUNCATE);
- if (!local) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "local init failed");
- op_errno = ENOMEM;
- goto err;
- }
-
- __cs_inode_ctx_get (this, loc->inode, &ctx);
-
- if (ctx)
- state = __cs_get_file_state (this, loc->inode, ctx);
- else
- state = GF_CS_LOCAL;
-
- local->xattr_req = xdata ? dict_ref (xdata) : (xdata = dict_new ());
-
- ret = dict_set_uint32 (local->xattr_req, GF_CS_OBJECT_STATUS, 1);
+ cs_local_t *local = NULL;
+ int ret = 0;
+ cs_inode_ctx_t *ctx = NULL;
+ gf_cs_obj_state state = -1;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+
+ local = cs_local_init(this, frame, loc, NULL, GF_FOP_TRUNCATE);
+ if (!local) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "local init failed");
+ goto err;
+ }
+
+ __cs_inode_ctx_get(this, loc->inode, &ctx);
+
+ if (ctx)
+ state = __cs_get_file_state(loc->inode, ctx);
+ else
+ state = GF_CS_LOCAL;
+
+ local->xattr_req = xdata ? dict_ref(xdata) : (xdata = dict_new());
+
+ ret = dict_set_uint32(local->xattr_req, GF_CS_OBJECT_STATUS, 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "dict_set failed key:"
+ " %s",
+ GF_CS_OBJECT_STATUS);
+ goto err;
+ }
+
+ local->stub = fop_truncate_stub(frame, cs_resume_truncate, loc, offset,
+ xdata);
+ if (!local->stub) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "insufficient memory");
+ goto err;
+ }
+
+ if (state == GF_CS_LOCAL) {
+ STACK_WIND(frame, cs_truncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
+
+ } else {
+ local->call_cnt++;
+ ret = locate_and_execute(frame);
if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "dict_set failed key:"
- " %s", GF_CS_OBJECT_STATUS);
- goto err;
- }
-
- local->stub = fop_truncate_stub (frame, cs_resume_truncate, loc, offset,
- xdata);
- if (!local->stub) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "insufficient memory");
- op_errno = ENOMEM;
- goto err;
- }
-
- if (state == GF_CS_LOCAL) {
- STACK_WIND (frame, cs_truncate_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->truncate,
- loc, offset, xdata);
-
- } else {
- local->call_cnt++;
- ret = locate_and_execute (frame);
- if (ret) {
- op_errno = ENOMEM;
- goto err;
- }
+ goto err;
}
+ }
- return 0;
+ return 0;
err:
- CS_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ CS_STACK_UNWIND(truncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+ return 0;
}
int32_t
-cs_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct statvfs *buf,
- dict_t *xdata)
+cs_statfs_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct statvfs *buf, dict_t *xdata)
{
- STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno,
- buf, xdata);
- return 0;
+ STACK_UNWIND_STRICT(statfs, frame, op_ret, op_errno, buf, xdata);
+ return 0;
}
-
int32_t
-cs_statfs (call_frame_t *frame, xlator_t *this,
- loc_t *loc,
- dict_t *xdata)
+cs_statfs(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- STACK_WIND (frame, cs_statfs_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->statfs,
- loc, xdata);
- return 0;
+ STACK_WIND(frame, cs_statfs_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->statfs, loc, xdata);
+ return 0;
}
-
int32_t
-cs_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
+cs_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno,
- dict, xdata);
- return 0;
+ STACK_UNWIND_STRICT(getxattr, frame, op_ret, op_errno, dict, xdata);
+ return 0;
}
-
int32_t
-cs_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc,
- const char *name,
- dict_t *xattr_req)
-{
- STACK_WIND (frame, cs_getxattr_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->getxattr,
- loc, name, xattr_req);
- return 0;
+cs_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, const char *name,
+ dict_t *xattr_req)
+{
+ STACK_WIND(frame, cs_getxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->getxattr, loc, name, xattr_req);
+ return 0;
}
int32_t
-cs_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+cs_setxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- cs_local_t *local = NULL;
+ cs_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->locked)
- cs_inodelk_unlock (frame);
+ if (local->locked)
+ cs_inodelk_unlock(frame);
- CS_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
+ CS_STACK_UNWIND(setxattr, frame, op_ret, op_errno, xdata);
- return 0;
+ return 0;
}
-
int32_t
-cs_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
- int32_t flags, dict_t *xdata)
+cs_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- data_t *tmp = NULL;
- cs_local_t *local = NULL;
- int ret = 0;
+ data_t *tmp = NULL;
+ cs_local_t *local = NULL;
+ int ret = 0;
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
- local = cs_local_init (this, frame, loc, NULL, GF_FOP_SETXATTR);
- if (!local) {
- ret = -1;
- goto err;
- }
-
- local->xattr_req = xdata ? dict_ref (xdata) : (xdata = dict_new ());
+ local = cs_local_init(this, frame, loc, NULL, GF_FOP_SETXATTR);
+ if (!local) {
+ ret = -1;
+ goto err;
+ }
- tmp = dict_get (dict, GF_CS_OBJECT_UPLOAD_COMPLETE);
- if (tmp) {
- /* Value of key should be the atime */
- local->stub = fop_setxattr_stub (frame, cs_resume_setxattr,
- loc, dict, flags, xdata);
+ local->xattr_req = xdata ? dict_ref(xdata) : (xdata = dict_new());
- if (!local->stub)
- goto err;
+ tmp = dict_get_sizen(dict, GF_CS_OBJECT_UPLOAD_COMPLETE);
+ if (tmp) {
+ /* Value of key should be the atime */
+ local->stub = fop_setxattr_stub(frame, cs_resume_setxattr, loc, dict,
+ flags, xdata);
- ret = locate_and_execute (frame);
- if (ret) {
- goto err;
- }
+ if (!local->stub)
+ goto err;
- return 0;
+ ret = locate_and_execute(frame);
+ if (ret) {
+ goto err;
}
- STACK_WIND (frame, cs_setxattr_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->setxattr,
- loc, dict, flags, xdata);
return 0;
+ }
+
+ STACK_WIND(frame, cs_setxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, loc, dict, flags, xdata);
+ return 0;
err:
- CS_STACK_UNWIND (setxattr, frame, -1, errno, NULL);
- return 0;
+ CS_STACK_UNWIND(setxattr, frame, -1, errno, NULL);
+ return 0;
}
-
int32_t
-cs_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
+cs_fgetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno,
- dict, xdata);
- return 0;
+ STACK_UNWIND_STRICT(fgetxattr, frame, op_ret, op_errno, dict, xdata);
+ return 0;
}
-
int32_t
-cs_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
- dict_t *xdata)
+cs_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
+ dict_t *xdata)
{
- STACK_WIND (frame, cs_fgetxattr_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->fgetxattr,
- fd, name, xdata);
- return 0;
+ STACK_WIND(frame, cs_fgetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
+ return 0;
}
-
-
int32_t
-cs_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+cs_fsetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno,
- xdata);
- return 0;
+ STACK_UNWIND_STRICT(fsetxattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
-
int32_t
-cs_fsetxattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd,
- dict_t *dict,
- int32_t flags,
- dict_t *xdata)
-{
- STACK_WIND (frame, cs_fsetxattr_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->fsetxattr,
- fd, dict, flags, xdata);
- return 0;
+cs_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata)
+{
+ STACK_WIND(frame, cs_fsetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+ return 0;
}
int32_t
-cs_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent,
- dict_t *xdata)
+cs_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno,
- preparent, postparent, xdata);
- return 0;
+ STACK_UNWIND_STRICT(unlink, frame, op_ret, op_errno, preparent, postparent,
+ xdata);
+ return 0;
}
-
int32_t
-cs_unlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc,
- int32_t flags,
- dict_t *xattr_req)
+cs_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ dict_t *xattr_req)
{
- cs_local_t *local = NULL;
- int ret = 0;
-
- local = cs_local_init (this, frame, loc, NULL, GF_FOP_UNLINK);
- if (!local)
- goto err;
-
- local->xattr_req = xattr_req ? dict_ref (xattr_req) : dict_new ();
-
- ret = dict_set_uint32 (local->xattr_req, GF_CS_OBJECT_STATUS, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "dict_set failed key:"
- " %s", GF_CS_OBJECT_STATUS);
- goto err;
- }
- STACK_WIND (frame, cs_unlink_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->unlink,
- loc, flags, local->xattr_req);
- return 0;
+ cs_local_t *local = NULL;
+ int ret = 0;
+
+ local = cs_local_init(this, frame, loc, NULL, GF_FOP_UNLINK);
+ if (!local)
+ goto err;
+
+ local->xattr_req = xattr_req ? dict_ref(xattr_req) : dict_new();
+
+ ret = dict_set_uint32(local->xattr_req, GF_CS_OBJECT_STATUS, 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "dict_set failed key:"
+ " %s",
+ GF_CS_OBJECT_STATUS);
+ goto err;
+ }
+ STACK_WIND(frame, cs_unlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, flags, local->xattr_req);
+ return 0;
err:
- CS_STACK_UNWIND (unlink, frame, -1, errno, NULL, NULL, NULL);
- return 0;
+ CS_STACK_UNWIND(unlink, frame, -1, errno, NULL, NULL, NULL);
+ return 0;
}
-
int32_t
-cs_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd,
- dict_t *xdata)
-{
- int ret = 0;
- uint64_t val = 0;
-
- if (op_ret == 0) {
- ret = dict_get_uint64 (xdata, GF_CS_OBJECT_STATUS, &val);
- if (!ret) {
- ret = __cs_inode_ctx_update (this, fd->inode, val);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0,
- "ctx update failed");
- }
- }
- } else {
- cs_inode_ctx_reset (this, fd->inode);
+cs_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+ int ret = 0;
+ uint64_t val = 0;
+
+ if (op_ret == 0) {
+ ret = dict_get_uint64(xdata, GF_CS_OBJECT_STATUS, &val);
+ if (!ret) {
+ ret = __cs_inode_ctx_update(this, fd->inode, val);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "ctx update failed");
+ }
}
+ } else {
+ cs_inode_ctx_reset(this, fd->inode);
+ }
- CS_STACK_UNWIND (open, frame, op_ret, op_errno, fd, xdata);
- return 0;
+ CS_STACK_UNWIND(open, frame, op_ret, op_errno, fd, xdata);
+ return 0;
}
-
int32_t
-cs_open (call_frame_t *frame, xlator_t *this,
- loc_t *loc,
- int32_t flags,
- fd_t *fd,
- dict_t *xattr_req)
+cs_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xattr_req)
{
- cs_local_t *local = NULL;
- int ret = 0;
-
- local = cs_local_init (this, frame, NULL, fd, GF_FOP_OPEN);
- if (!local)
- goto err;
-
- local->xattr_req = xattr_req ? dict_ref (xattr_req) : dict_new ();
-
- ret = dict_set_uint32 (local->xattr_req, GF_CS_OBJECT_STATUS, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "dict_set failed key:"
- " %s", GF_CS_OBJECT_STATUS);
- goto err;
- }
-
- STACK_WIND (frame, cs_open_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->open,
- loc, flags, fd, local->xattr_req);
- return 0;
+ cs_local_t *local = NULL;
+ int ret = 0;
+
+ local = cs_local_init(this, frame, NULL, fd, GF_FOP_OPEN);
+ if (!local)
+ goto err;
+
+ local->xattr_req = xattr_req ? dict_ref(xattr_req) : dict_new();
+
+ ret = dict_set_uint32(local->xattr_req, GF_CS_OBJECT_STATUS, 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "dict_set failed key:"
+ " %s",
+ GF_CS_OBJECT_STATUS);
+ goto err;
+ }
+
+ STACK_WIND(frame, cs_open_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->open, loc, flags, fd, local->xattr_req);
+ return 0;
err:
- CS_STACK_UNWIND (open, frame, -1, errno, NULL, NULL);
- return 0;
+ CS_STACK_UNWIND(open, frame, -1, errno, NULL, NULL);
+ return 0;
}
-
int32_t
-cs_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- dict_t *xdata)
-{
- int ret = 0;
- uint64_t val = 0;
- fd_t *fd = NULL;
- cs_local_t *local = NULL;
-
- local = frame->local;
-
- fd = local->fd;
-
- if (op_ret == 0) {
- ret = dict_get_uint64 (xdata, GF_CS_OBJECT_STATUS, &val);
- if (!ret) {
- gf_msg_debug (this->name, 0, "state %ld", val);
- ret = __cs_inode_ctx_update (this, fd->inode, val);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0,
- "ctx update failed");
- }
- }
- } else {
- cs_inode_ctx_reset (this, fd->inode);
+cs_fstat_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *buf, dict_t *xdata)
+{
+ int ret = 0;
+ uint64_t val = 0;
+ fd_t *fd = NULL;
+ cs_local_t *local = NULL;
+
+ local = frame->local;
+
+ fd = local->fd;
+
+ if (op_ret == 0) {
+ ret = dict_get_uint64(xdata, GF_CS_OBJECT_STATUS, &val);
+ if (!ret) {
+ gf_msg_debug(this->name, 0, "state %" PRIu64, val);
+ ret = __cs_inode_ctx_update(this, fd->inode, val);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "ctx update failed");
+ }
}
+ } else {
+ cs_inode_ctx_reset(this, fd->inode);
+ }
- CS_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf, xdata);
+ CS_STACK_UNWIND(fstat, frame, op_ret, op_errno, buf, xdata);
- return 0;
+ return 0;
}
-
int32_t
-cs_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xattr_req)
+cs_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xattr_req)
{
- cs_local_t *local = NULL;
- int ret = 0;
+ cs_local_t *local = NULL;
+ int ret = 0;
- local = cs_local_init (this, frame, NULL, fd, GF_FOP_FSTAT);
- if (!local)
- goto err;
+ local = cs_local_init(this, frame, NULL, fd, GF_FOP_FSTAT);
+ if (!local)
+ goto err;
- if (fd->inode->ia_type == IA_IFDIR)
- goto wind;
+ if (fd->inode->ia_type == IA_IFDIR)
+ goto wind;
- local->xattr_req = xattr_req ? dict_ref (xattr_req) : dict_new ();
+ local->xattr_req = xattr_req ? dict_ref(xattr_req) : dict_new();
- ret = dict_set_uint32 (local->xattr_req, GF_CS_OBJECT_STATUS, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "dict_set failed key:"
- " %s", GF_CS_OBJECT_STATUS);
- goto err;
- }
+ ret = dict_set_uint32(local->xattr_req, GF_CS_OBJECT_STATUS, 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "dict_set failed key:"
+ " %s",
+ GF_CS_OBJECT_STATUS);
+ goto err;
+ }
wind:
- STACK_WIND (frame, cs_fstat_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->fstat,
- fd, local->xattr_req);
- return 0;
+ STACK_WIND(frame, cs_fstat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fstat, fd, local->xattr_req);
+ return 0;
err:
- CS_STACK_UNWIND (fstat, frame, -1, errno, NULL, NULL);
- return 0;
+ CS_STACK_UNWIND(fstat, frame, -1, errno, NULL, NULL);
+ return 0;
}
cs_local_t *
-cs_local_init (xlator_t *this, call_frame_t *frame, loc_t *loc, fd_t *fd,
- glusterfs_fop_t fop)
+cs_local_init(xlator_t *this, call_frame_t *frame, loc_t *loc, fd_t *fd,
+ glusterfs_fop_t fop)
{
- cs_local_t *local = NULL;
- int ret = 0;
-
- local = mem_get0 (this->local_pool);
- if (!local)
- goto out;
+ cs_local_t *local = NULL;
+ int ret = 0;
- if (loc) {
- ret = loc_copy (&local->loc, loc);
- if (ret)
- goto out;
- }
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto out;
- if (fd) {
- local->fd = fd_ref (fd);
- }
-
- local->op_ret = -1;
- local->op_errno = EUCLEAN;
- local->fop = fop;
- local->dloffset = 0;
- frame->local = local;
- local->locked = _gf_false;
- local->call_cnt = 0;
+ if (loc) {
+ ret = loc_copy(&local->loc, loc);
+ if (ret)
+ goto out;
+ }
+
+ if (fd) {
+ local->fd = fd_ref(fd);
+ }
+
+ local->op_ret = -1;
+ local->op_errno = EUCLEAN;
+ local->fop = fop;
+ local->dloffset = 0;
+ frame->local = local;
+ local->locked = _gf_false;
+ local->call_cnt = 0;
out:
- if (ret) {
- if (local)
- mem_put (local);
- local = NULL;
- }
+ if (ret) {
+ if (local)
+ mem_put(local);
+ local = NULL;
+ }
- return local;
+ return local;
}
call_frame_t *
-cs_lock_frame (call_frame_t *parent_frame)
+cs_lock_frame(call_frame_t *parent_frame)
{
- call_frame_t *lock_frame = NULL;
+ call_frame_t *lock_frame = NULL;
- lock_frame = copy_frame (parent_frame);
+ lock_frame = copy_frame(parent_frame);
- if (lock_frame == NULL)
- goto out;
+ if (lock_frame == NULL)
+ goto out;
- set_lk_owner_from_ptr (&lock_frame->root->lk_owner, parent_frame->root);
+ set_lk_owner_from_ptr(&lock_frame->root->lk_owner, parent_frame->root);
out:
- return lock_frame;
-
+ return lock_frame;
}
void
-cs_lock_wipe (call_frame_t *lock_frame)
+cs_lock_wipe(call_frame_t *lock_frame)
{
- CS_STACK_DESTROY (lock_frame);
+ CS_STACK_DESTROY(lock_frame);
}
-
int32_t
-cs_inodelk_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+cs_inodelk_unlock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- cs_lock_wipe (frame);
+ cs_lock_wipe(frame);
- return 0;
+ return 0;
}
int
-cs_inodelk_unlock (call_frame_t *main_frame)
+cs_inodelk_unlock(call_frame_t *main_frame)
{
- xlator_t *this = NULL;
- struct gf_flock flock = {0,};
- call_frame_t *lock_frame = NULL;
- cs_local_t *lock_local = NULL;
- cs_local_t *main_local = NULL;
- int ret = 0;
+ xlator_t *this = NULL;
+ struct gf_flock flock = {
+ 0,
+ };
+ call_frame_t *lock_frame = NULL;
+ cs_local_t *lock_local = NULL;
+ cs_local_t *main_local = NULL;
+ int ret = 0;
- this = main_frame->this;
- main_local = main_frame->local;
+ this = main_frame->this;
+ main_local = main_frame->local;
- lock_frame = cs_lock_frame (main_frame);
- if (!lock_frame)
- goto out;
+ lock_frame = cs_lock_frame(main_frame);
+ if (!lock_frame)
+ goto out;
- lock_local = cs_local_init (this, lock_frame, NULL, NULL, 0);
- if (!lock_local)
- goto out;
+ lock_local = cs_local_init(this, lock_frame, NULL, NULL, 0);
+ if (!lock_local)
+ goto out;
- ret = cs_build_loc (&lock_local->loc, main_frame);
- if (ret) {
- goto out;
- }
+ ret = cs_build_loc(&lock_local->loc, main_frame);
+ if (ret) {
+ goto out;
+ }
- flock.l_type = F_UNLCK;
+ flock.l_type = F_UNLCK;
- main_local->locked = _gf_false;
+ main_local->locked = _gf_false;
- STACK_WIND (lock_frame, cs_inodelk_unlock_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->inodelk, CS_LOCK_DOMAIN,
- &lock_local->loc, F_SETLKW, &flock, NULL);
+ STACK_WIND(lock_frame, cs_inodelk_unlock_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->inodelk, CS_LOCK_DOMAIN,
+ &lock_local->loc, F_SETLKW, &flock, NULL);
- return 0;
+ return 0;
out:
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "Stale lock would be found on"
- " server");
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "Stale lock would be found on"
+ " server");
- if (lock_frame)
- cs_lock_wipe (lock_frame);
+ if (lock_frame)
+ cs_lock_wipe(lock_frame);
- return 0;
+ return 0;
}
-void *
-cs_download_task (void *arg)
+int
+cs_download_task(void *arg)
{
- call_frame_t *frame = NULL;
- xlator_t *this = NULL;
- cs_private_t *priv = NULL;
- int ret = -1;
- char *sign_req = NULL;
- fd_t *fd = NULL;
- cs_local_t *local = NULL;
- dict_t *dict = NULL;
- int *retval = NULL;
+ call_frame_t *frame = NULL;
+ xlator_t *this = NULL;
+ cs_private_t *priv = NULL;
+ int ret = -1;
+ char *sign_req = NULL;
+ fd_t *fd = NULL;
+ cs_local_t *local = NULL;
+ dict_t *dict = NULL;
+
+ frame = (call_frame_t *)arg;
+
+ this = frame->this;
+
+ priv = this->private;
+
+ if (!priv->stores) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "No remote store "
+ "plugins found");
+ ret = -1;
+ goto out;
+ }
+
+ local = frame->local;
+
+ if (local->fd)
+ fd = fd_anonymous(local->fd->inode);
+ else
+ fd = fd_anonymous(local->loc.inode);
+
+ if (!fd) {
+ gf_msg("CS", GF_LOG_ERROR, 0, 0, "fd creation failed");
+ ret = -1;
+ goto out;
+ }
+
+ local->dlfd = fd;
+ local->dloffset = 0;
+
+ dict = dict_new();
+ if (!dict) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, ENOMEM,
+ "failed to create "
+ "dict");
+ ret = -1;
+ goto out;
+ }
+
+ ret = dict_set_uint32(dict, GF_CS_OBJECT_DOWNLOADING, 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "dict_set failed");
+ ret = -1;
+ goto out;
+ }
+
+ ret = syncop_fsetxattr(this, local->fd, dict, 0, NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "fsetxattr failed "
+ "key %s",
+ GF_CS_OBJECT_DOWNLOADING);
+ ret = -1;
+ goto out;
+ }
+ /*this calling method is for per volume setting */
+ ret = priv->stores->dlfop(frame, priv->stores->config);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "download failed"
+ ", remotepath: %s",
+ local->remotepath);
+
+ /*using dlfd as it is anonymous and have RDWR flag*/
+ ret = syncop_ftruncate(FIRST_CHILD(this), local->dlfd, 0, NULL, NULL,
+ NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, -ret, "ftruncate failed");
+ } else {
+ gf_msg_debug(this->name, 0, "ftruncate succeed");
+ }
- frame = (call_frame_t *)arg;
+ ret = -1;
+ goto out;
+ } else {
+ gf_msg(this->name, GF_LOG_INFO, 0, 0,
+ "download success, path"
+ " : %s",
+ local->remotepath);
- this = frame->this;
+ ret = syncop_fremovexattr(this, local->fd, GF_CS_OBJECT_REMOTE, NULL,
+ NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, -ret,
+ "removexattr failed, remotexattr");
+ ret = -1;
+ goto out;
+ } else {
+ gf_msg_debug(this->name, 0,
+ "fremovexattr success, "
+ "path : %s",
+ local->remotepath);
+ }
- priv = this->private;
+ ret = syncop_fremovexattr(this, local->fd, GF_CS_OBJECT_DOWNLOADING,
+ NULL, NULL);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, -ret,
+ "removexattr failed, downloading xattr, path %s",
+ local->remotepath);
+ ret = -1;
+ goto out;
+ } else {
+ gf_msg_debug(this->name, 0,
+ "fremovexattr success"
+ " path %s",
+ local->remotepath);
+ }
+ }
- local = frame->local;
+out:
+ GF_FREE(sign_req);
- retval = GF_CALLOC (1, sizeof(int), gf_common_mt_int);
- if (!retval) {
- gf_log (this->name, GF_LOG_ERROR, "insufficient memory");
- ret = -1;
- goto out;
- }
+ if (dict)
+ dict_unref(dict);
- if (local->fd)
- fd = fd_anonymous (local->fd->inode);
- else
- fd = fd_anonymous (local->loc.inode);
+ if (fd) {
+ fd_unref(fd);
+ local->dlfd = NULL;
+ }
- if (!fd) {
- gf_msg ("CS", GF_LOG_ERROR, 0, 0, "fd creation failed");
- ret = -1;
- goto out;
- }
+ return ret;
+}
- local->dlfd = fd;
- local->dloffset = 0;
+int
+cs_download(call_frame_t *frame)
+{
+ int ret = 0;
+ cs_local_t *local = NULL;
+ xlator_t *this = NULL;
+
+ local = frame->local;
+ this = frame->this;
+
+ if (!local->remotepath) {
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "remote path not"
+ " available. Check posix logs to resolve");
+ goto out;
+ }
+
+ ret = cs_download_task((void *)frame);
+out:
+ return ret;
+}
- dict = dict_new ();
- if (!dict) {
- gf_msg (this->name, GF_LOG_ERROR, 0, ENOMEM, "failed to create "
- "dict");
- ret = -1;
- goto out;
- }
+int
+cs_set_xattr_req(call_frame_t *frame)
+{
+ cs_local_t *local = NULL;
+ GF_UNUSED int ret = 0;
+
+ local = frame->local;
+
+ /* When remote reads are performed (i.e. reads on remote store),
+ * there needs to be a way to associate a file on gluster volume
+ * with its correspnding file on the remote store. In order to do
+ * that, a unique key can be maintained as an xattr
+ * (GF_CS_XATTR_ARCHIVE_UUID)on the stub file on gluster bricks.
+ * This xattr should be provided to the plugin to
+ * perform the read fop on the correct file. This assumes that the file
+ * hierarchy and name need not be the same on remote store as that of
+ * the gluster volume.
+ */
+ ret = dict_set_sizen_str_sizen(local->xattr_req, GF_CS_XATTR_ARCHIVE_UUID,
+ "1");
+
+ return 0;
+}
- ret = dict_set_uint32 (dict, GF_CS_OBJECT_DOWNLOADING, 1);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "dict_set failed");
- ret = -1;
- goto out;
+int
+cs_update_xattrs(call_frame_t *frame, dict_t *xdata)
+{
+ cs_local_t *local = NULL;
+ xlator_t *this = NULL;
+ int size = -1;
+ GF_UNUSED int ret = 0;
+
+ local = frame->local;
+ this = frame->this;
+
+ local->xattrinfo.lxattr = GF_CALLOC(1, sizeof(cs_loc_xattr_t),
+ gf_cs_mt_cs_lxattr_t);
+ if (!local->xattrinfo.lxattr) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto err;
+ }
+
+ gf_uuid_copy(local->xattrinfo.lxattr->gfid, local->loc.gfid);
+
+ if (local->remotepath) {
+ local->xattrinfo.lxattr->file_path = gf_strdup(local->remotepath);
+ if (!local->xattrinfo.lxattr->file_path) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto err;
}
+ }
+
+ ret = dict_get_gfuuid(xdata, GF_CS_XATTR_ARCHIVE_UUID,
+ &(local->xattrinfo.lxattr->uuid));
+
+ if (ret) {
+ gf_uuid_clear(local->xattrinfo.lxattr->uuid);
+ }
+ size = strlen(this->name) - strlen("-cloudsync") + 1;
+ local->xattrinfo.lxattr->volname = GF_CALLOC(1, size, gf_common_mt_char);
+ if (!local->xattrinfo.lxattr->volname) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto err;
+ }
+ strncpy(local->xattrinfo.lxattr->volname, this->name, size - 1);
+ local->xattrinfo.lxattr->volname[size - 1] = '\0';
+
+ return 0;
+err:
+ cs_xattrinfo_wipe(local);
+ return -1;
+}
- ret = syncop_fsetxattr (this, local->fd, dict, 0, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "fsetxattr failed "
- "key %s", GF_CS_OBJECT_DOWNLOADING);
- ret = -1;
- goto out;
- }
- /*this calling method is for per volume setting */
- ret = priv->stores->dlfop (frame, priv->stores->config);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "download failed"
- ", remotepath: %s", local->remotepath);
-
- /*using dlfd as it is anonymous and have RDWR flag*/
- ret = syncop_ftruncate (FIRST_CHILD (this), local->dlfd, 0,
- NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, -ret,
- "ftruncate failed");
- } else {
- gf_msg_debug (this->name, 0, "ftruncate succeed");
- }
-
- ret = -1;
- goto out;
- } else {
- gf_msg (this->name, GF_LOG_INFO, 0, 0, "download success, path"
- " : %s", local->remotepath);
+int
+cs_serve_readv(call_frame_t *frame, off_t offset, size_t size, uint32_t flags)
+{
+ xlator_t *this = NULL;
+ cs_private_t *priv = NULL;
+ int ret = -1;
+ fd_t *fd = NULL;
+ cs_local_t *local = NULL;
+
+ local = frame->local;
+ this = frame->this;
+ priv = this->private;
+
+ if (!local->remotepath) {
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "remote path not"
+ " available. Check posix logs to resolve");
+ goto out;
+ }
+
+ if (!priv->stores) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "No remote store "
+ "plugins found");
+ ret = -1;
+ goto out;
+ }
+
+ if (local->fd) {
+ fd = fd_anonymous(local->fd->inode);
+ } else {
+ fd = fd_anonymous(local->loc.inode);
+ }
+
+ local->xattrinfo.size = size;
+ local->xattrinfo.offset = offset;
+ local->xattrinfo.flags = flags;
+
+ if (!fd) {
+ gf_msg("CS", GF_LOG_ERROR, 0, 0, "fd creation failed");
+ ret = -1;
+ goto out;
+ }
+
+ local->dlfd = fd;
+ local->dloffset = offset;
+
+ /*this calling method is for per volume setting */
+ ret = priv->stores->rdfop(frame, priv->stores->config);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "read failed"
+ ", remotepath: %s",
+ local->remotepath);
+ ret = -1;
+ goto out;
+ } else {
+ gf_msg(this->name, GF_LOG_INFO, 0, 0,
+ "read success, path"
+ " : %s",
+ local->remotepath);
+ }
- ret = syncop_fremovexattr (this, local->fd,
- GF_CS_OBJECT_REMOTE, NULL, NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, -ret,
- "removexattr failed, remotexattr");
- ret = -1;
- goto out;
- } else {
- gf_msg_debug (this->name, 0, "fremovexattr success, "
- "path : %s", local->remotepath);
- }
+out:
+ if (fd) {
+ fd_unref(fd);
+ local->dlfd = NULL;
+ }
+ return ret;
+}
- ret = syncop_fremovexattr (this, local->fd,
- GF_CS_OBJECT_DOWNLOADING, NULL,
- NULL);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, -ret,
- "removexattr failed, downloading xattr, path %s"
- , local->remotepath);
- ret = -1;
- goto out;
+int32_t
+cs_readv_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iovec *vector, int32_t count,
+ struct iatt *stbuf, struct iobref *iobref, dict_t *xdata)
+{
+ cs_local_t *local = NULL;
+ int ret = 0;
+ uint64_t val = 0;
+ fd_t *fd = NULL;
+
+ local = frame->local;
+ fd = local->fd;
+
+ local->call_cnt++;
+
+ if (op_ret == -1) {
+ ret = dict_get_uint64(xdata, GF_CS_OBJECT_STATUS, &val);
+ if (ret == 0) {
+ if (val == GF_CS_ERROR) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "could not get file state, unwinding");
+ op_ret = -1;
+ op_errno = EIO;
+ goto unwind;
+ } else {
+ __cs_inode_ctx_update(this, fd->inode, val);
+ gf_msg(this->name, GF_LOG_INFO, 0, 0, " state = %" PRIu64, val);
+
+ if (local->call_cnt == 1 &&
+ (val == GF_CS_REMOTE || val == GF_CS_DOWNLOADING)) {
+ gf_msg(this->name, GF_LOG_INFO, 0, 0,
+ " will read from remote : %" PRIu64, val);
+ goto repair;
} else {
- gf_msg_debug (this->name, 0, "fremovexattr success"
- " path %s", local->remotepath);
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "second readv, Unwinding");
+ goto unwind;
}
+ }
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "file state "
+ "could not be figured, unwinding");
+ goto unwind;
}
+ } else {
+ /* successful readv => file is local */
+ __cs_inode_ctx_update(this, fd->inode, GF_CS_LOCAL);
+ gf_msg(this->name, GF_LOG_INFO, 0, 0,
+ "state : GF_CS_LOCAL"
+ ", readv successful");
-out:
- GF_FREE (sign_req);
+ goto unwind;
+ }
- if (dict)
- dict_unref (dict);
+repair:
+ ret = locate_and_execute(frame);
+ if (ret) {
+ goto unwind;
+ }
- if (fd) {
- fd_unref (fd);
- local->dlfd = NULL;
- }
+ return 0;
- if (retval) {
- *retval = ret;
- pthread_exit (retval);
- } else {
- pthread_exit (&ret);
- }
+unwind:
+ CS_STACK_UNWIND(readv, frame, op_ret, op_errno, vector, count, stbuf,
+ iobref, xdata);
+
+ return 0;
}
-int
-cs_download (call_frame_t *frame)
+int32_t
+cs_resume_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- int *retval = NULL;
- int ret = 0;
- pthread_t dthread;
- cs_local_t *local = NULL;
- xlator_t *this = NULL;
+ int ret = 0;
- local = frame->local;
- this = frame->this;
+ ret = cs_resume_postprocess(this, frame, fd->inode);
+ if (ret) {
+ goto unwind;
+ }
- if (!local->remotepath) {
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "remote path not"
- " available. Check posix logs to resolve");
- goto out;
- }
+ cs_inodelk_unlock(frame);
- ret = gf_thread_create (&dthread, NULL, &cs_download_task,
- (void *)frame, "downloadthread");
+ STACK_WIND(frame, cs_readv_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readv, fd, size, offset, flags, xdata);
- pthread_join (dthread, (void **)&retval);
+ return 0;
- ret = *retval;
+unwind:
+ cs_inodelk_unlock(frame);
-out:
- if (retval)
- GF_FREE (retval);
+ cs_common_cbk(frame);
- return ret;
+ return 0;
}
-int
-cs_stat_check_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *stbuf, dict_t *xdata)
-{
- cs_local_t *local = NULL;
- call_stub_t *stub = NULL;
- char *filepath = NULL;
- int ret = 0;
- inode_t *inode = NULL;
- uint64_t val = 0;
-
- local = frame->local;
-
- if (op_ret == -1) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- gf_msg (this->name, GF_LOG_ERROR, 0, op_errno,
- "stat check failed");
- goto err;
- } else {
- if (local->fd)
- inode = local->fd->inode;
- else
- inode = local->loc.inode;
-
- if (!inode) {
- local->op_ret = -1;
- local->op_errno = EINVAL;
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "null inode "
- "returned");
- goto err;
- }
+int32_t
+cs_resume_remote_readv(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ size_t size, off_t offset, uint32_t flags, dict_t *xdata)
+{
+ int ret = 0;
+ cs_local_t *local = NULL;
+ gf_cs_obj_state state = -1;
+ cs_inode_ctx_t *ctx = NULL;
+
+ cs_inodelk_unlock(frame);
+
+ local = frame->local;
+ if (!local) {
+ ret = -1;
+ goto unwind;
+ }
+
+ __cs_inode_ctx_get(this, fd->inode, &ctx);
+
+ state = __cs_get_file_state(fd->inode, ctx);
+ if (state == GF_CS_ERROR) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "status is GF_CS_ERROR."
+ " Aborting readv");
+ local->op_ret = -1;
+ local->op_errno = EREMOTE;
+ ret = -1;
+ goto unwind;
+ }
+
+ /* Serve readv from remote store only if it is remote. */
+ gf_msg_debug(this->name, 0, "status of file %s is %d",
+ local->remotepath ? local->remotepath : "", state);
+
+ /* We will reach this condition if local inode ctx had REMOTE
+ * state when the control was in cs_readv but after stat
+ * we got an updated state saying that the file is LOCAL.
+ */
+ if (state == GF_CS_LOCAL) {
+ STACK_WIND(frame, cs_readv_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readv, fd, size, offset, flags,
+ xdata);
+ } else if (state == GF_CS_REMOTE) {
+ ret = cs_resume_remote_readv_postprocess(this, frame, fd->inode, offset,
+ size, flags);
+ /* Failed to submit the remote readv fop to plugin */
+ if (ret) {
+ local->op_ret = -1;
+ local->op_errno = EREMOTE;
+ goto unwind;
+ }
+ /* When the file is in any other intermediate state,
+ * we should not perform remote reads.
+ */
+ } else {
+ local->op_ret = -1;
+ local->op_errno = EINVAL;
+ goto unwind;
+ }
- ret = dict_get_uint64 (xdata, GF_CS_OBJECT_STATUS, &val);
- if (ret == 0) {
- if (val == GF_CS_ERROR) {
- cs_inode_ctx_reset (this, inode);
- local->op_ret = -1;
- local->op_errno = EIO;
- gf_msg (this->name, GF_LOG_ERROR, 0, 0,
- "status = GF_CS_ERROR. failed to get "
- " file state");
- goto err;
- } else {
- ret = __cs_inode_ctx_update (this, inode, val);
- gf_msg_debug (this->name, 0, "status : %lu",
- val);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0,
- "ctx update failed");
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto err;
- }
- }
- } else {
- gf_msg_debug (this->name, 0,
- "status not found in dict");
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto err;
- }
+ return 0;
- ret = dict_get_str (xdata, GF_CS_OBJECT_REMOTE, &filepath);
- if (filepath) {
- gf_msg_debug (this->name, 0, "filepath returned %s",
- filepath);
- local->remotepath = gf_strdup (filepath);
- if (!local->remotepath) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto err;
- }
- } else {
- gf_msg_debug (this->name, 0, "NULL filepath");
- }
+unwind:
+ cs_common_cbk(frame);
+
+ return 0;
+}
- local->op_ret = 0;
- local->xattr_rsp = dict_ref (xdata);
- memcpy (&local->stbuf, stbuf, sizeof (struct iatt));
+int32_t
+cs_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
+{
+ int op_errno = ENOMEM;
+ cs_local_t *local = NULL;
+ int ret = 0;
+ cs_inode_ctx_t *ctx = NULL;
+ gf_cs_obj_state state = -1;
+ cs_private_t *priv = NULL;
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+
+ priv = this->private;
+
+ local = cs_local_init(this, frame, NULL, fd, GF_FOP_READ);
+ if (!local) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "local init failed");
+ goto err;
+ }
+
+ __cs_inode_ctx_get(this, fd->inode, &ctx);
+
+ if (ctx)
+ state = __cs_get_file_state(fd->inode, ctx);
+ else
+ state = GF_CS_LOCAL;
+
+ local->xattr_req = xdata ? dict_ref(xdata) : (xdata = dict_new());
+
+ ret = dict_set_uint32(local->xattr_req, GF_CS_OBJECT_STATUS, 1);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "dict_set failed key:"
+ " %s",
+ GF_CS_OBJECT_STATUS);
+ goto err;
+ }
+
+ if (priv->remote_read) {
+ local->stub = fop_readv_stub(frame, cs_resume_remote_readv, fd, size,
+ offset, flags, xdata);
+ } else {
+ local->stub = fop_readv_stub(frame, cs_resume_readv, fd, size, offset,
+ flags, xdata);
+ }
+ if (!local->stub) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "insufficient memory");
+ goto err;
+ }
+
+ if (state == GF_CS_LOCAL) {
+ STACK_WIND(frame, cs_readv_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readv, fd, size, offset, flags,
+ xdata);
+ } else {
+ local->call_cnt++;
+ ret = locate_and_execute(frame);
+ if (ret) {
+ goto err;
}
+ }
- stub = local->stub;
- local->stub = NULL;
- call_resume (stub);
+ return 0;
- return 0;
err:
- cs_inodelk_unlock (frame);
-
- cs_common_cbk (frame);
+ CS_STACK_UNWIND(readv, frame, -1, op_errno, NULL, -1, NULL, NULL, NULL);
- return 0;
+ return 0;
}
int
-cs_do_stat_check (call_frame_t *main_frame)
+cs_resume_remote_readv_postprocess(xlator_t *this, call_frame_t *frame,
+ inode_t *inode, off_t offset, size_t size,
+ uint32_t flags)
{
- cs_local_t *local = NULL;
- xlator_t *this = NULL;
- int ret = 0;
+ int ret = 0;
- local = main_frame->local;
- this = main_frame->this;
+ ret = cs_serve_readv(frame, offset, size, flags);
- ret = dict_set_uint32 (local->xattr_req, GF_CS_OBJECT_REPAIR, 256);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "dict_set failed");
+ return ret;
+}
+
+int
+cs_stat_check_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *stbuf, dict_t *xdata)
+{
+ cs_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ char *filepath = NULL;
+ int ret = 0;
+ inode_t *inode = NULL;
+ uint64_t val = 0;
+
+ local = frame->local;
+
+ if (op_ret == -1) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ gf_msg(this->name, GF_LOG_ERROR, 0, op_errno, "stat check failed");
+ goto err;
+ } else {
+ if (local->fd)
+ inode = local->fd->inode;
+ else
+ inode = local->loc.inode;
+
+ if (!inode) {
+ local->op_ret = -1;
+ local->op_errno = EINVAL;
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "null inode "
+ "returned");
+ goto err;
+ }
+
+ ret = dict_get_uint64(xdata, GF_CS_OBJECT_STATUS, &val);
+ if (ret == 0) {
+ if (val == GF_CS_ERROR) {
+ cs_inode_ctx_reset(this, inode);
+ local->op_ret = -1;
+ local->op_errno = EIO;
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "status = GF_CS_ERROR. failed to get "
+ " file state");
goto err;
+ } else {
+ ret = __cs_inode_ctx_update(this, inode, val);
+ gf_msg_debug(this->name, 0, "status : %" PRIu64, val);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "ctx update failed");
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto err;
+ }
+ }
+ } else {
+ gf_msg_debug(this->name, 0, "status not found in dict");
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto err;
}
- if (local->fd) {
- STACK_WIND (main_frame, cs_stat_check_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fstat, local->fd,
- local->xattr_req);
+ ret = dict_get_str_sizen(xdata, GF_CS_OBJECT_REMOTE, &filepath);
+ if (filepath) {
+ gf_msg_debug(this->name, 0, "filepath returned %s", filepath);
+ local->remotepath = gf_strdup(filepath);
+ if (!local->remotepath) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto err;
+ }
} else {
- STACK_WIND (main_frame, cs_stat_check_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->stat, &local->loc,
- local->xattr_req);
+ gf_msg_debug(this->name, 0, "NULL filepath");
}
- return 0;
+ ret = cs_update_xattrs(frame, xdata);
+ if (ret)
+ goto err;
+ local->op_ret = 0;
+ local->xattr_rsp = dict_ref(xdata);
+ memcpy(&local->stbuf, stbuf, sizeof(struct iatt));
+ }
+
+ stub = local->stub;
+ local->stub = NULL;
+ call_resume(stub);
+
+ return 0;
err:
- cs_inodelk_unlock (main_frame);
+ cs_inodelk_unlock(frame);
- cs_common_cbk (main_frame);
+ cs_common_cbk(frame);
- return 0;
+ return 0;
+}
+
+int
+cs_do_stat_check(call_frame_t *main_frame)
+{
+ cs_local_t *local = NULL;
+ xlator_t *this = NULL;
+ int ret = 0;
+
+ local = main_frame->local;
+ this = main_frame->this;
+
+ ret = dict_set_uint32(local->xattr_req, GF_CS_OBJECT_REPAIR, 256);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "dict_set failed");
+ goto err;
+ }
+
+ cs_set_xattr_req(main_frame);
+
+ if (local->fd) {
+ STACK_WIND(main_frame, cs_stat_check_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fstat, local->fd, local->xattr_req);
+ } else {
+ STACK_WIND(main_frame, cs_stat_check_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->stat, &local->loc,
+ local->xattr_req);
+ }
+
+ return 0;
+
+err:
+ cs_inodelk_unlock(main_frame);
+
+ cs_common_cbk(main_frame);
+
+ return 0;
}
void
-cs_common_cbk (call_frame_t *frame)
+cs_common_cbk(call_frame_t *frame)
{
- glusterfs_fop_t fop = -1;
- cs_local_t *local = NULL;
+ glusterfs_fop_t fop = -1;
+ cs_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- fop = local->fop;
+ fop = local->fop;
- /*Note: Only the failure case needs to be handled here. Since for
- * successful stat check the fop will resume anyway. The unwind can
- * happen from the fop_cbk and each cbk can unlock the inodelk in case
- * a lock was taken before. The lock status can be stored in frame */
+ /*Note: Only the failure case needs to be handled here. Since for
+ * successful stat check the fop will resume anyway. The unwind can
+ * happen from the fop_cbk and each cbk can unlock the inodelk in case
+ * a lock was taken before. The lock status can be stored in frame */
- /* for failure case */
+ /* for failure case */
- /*TODO: add other fops*/
- switch (fop) {
+ /*TODO: add other fops*/
+ switch (fop) {
case GF_FOP_WRITE:
- CS_STACK_UNWIND (writev, frame, local->op_ret,
- local->op_errno, NULL, NULL, NULL);
- break;
+ CS_STACK_UNWIND(writev, frame, local->op_ret, local->op_errno, NULL,
+ NULL, NULL);
+ break;
case GF_FOP_SETXATTR:
- CS_STACK_UNWIND (setxattr, frame, local->op_ret,
- local->op_errno, NULL);
- break;
+ CS_STACK_UNWIND(setxattr, frame, local->op_ret, local->op_errno,
+ NULL);
+ break;
case GF_FOP_READ:
- CS_STACK_UNWIND (readv, frame, local->op_ret,
- local->op_errno, NULL, 0, NULL, NULL,
- NULL);
- break;
+ CS_STACK_UNWIND(readv, frame, local->op_ret, local->op_errno, NULL,
+ 0, NULL, NULL, NULL);
+ break;
case GF_FOP_FTRUNCATE:
- CS_STACK_UNWIND (ftruncate, frame, local->op_ret,
- local->op_errno, NULL, NULL, NULL);
- break;
-
-
+ CS_STACK_UNWIND(ftruncate, frame, local->op_ret, local->op_errno,
+ NULL, NULL, NULL);
+ break;
+
+ case GF_FOP_TRUNCATE:
+ CS_STACK_UNWIND(truncate, frame, local->op_ret, local->op_errno,
+ NULL, NULL, NULL);
+ break;
default:
- break;
- }
+ break;
+ }
- return;
+ return;
}
int
-cs_blocking_inodelk_cbk (call_frame_t *lock_frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+cs_blocking_inodelk_cbk(call_frame_t *lock_frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- cs_local_t *main_local = NULL;
- call_frame_t *main_frame = NULL;
- cs_local_t *lock_local = NULL;
+ cs_local_t *main_local = NULL;
+ call_frame_t *main_frame = NULL;
+ cs_local_t *lock_local = NULL;
- lock_local = lock_frame->local;
+ lock_local = lock_frame->local;
- main_frame = lock_local->main_frame;
- main_local = main_frame->local;
+ main_frame = lock_local->main_frame;
+ main_local = main_frame->local;
- if (op_ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "inodelk failed");
- main_local->op_errno = op_errno;
- main_local->op_ret = op_ret;
- goto err;
- }
+ if (op_ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "inodelk failed");
+ main_local->op_errno = op_errno;
+ main_local->op_ret = op_ret;
+ goto err;
+ }
- main_local->locked = _gf_true;
+ main_local->locked = _gf_true;
- cs_lock_wipe (lock_frame);
+ cs_lock_wipe(lock_frame);
- cs_do_stat_check (main_frame);
+ cs_do_stat_check(main_frame);
- return 0;
+ return 0;
err:
- cs_common_cbk (main_frame);
+ cs_common_cbk(main_frame);
- cs_lock_wipe (lock_frame);
+ cs_lock_wipe(lock_frame);
- return 0;
+ return 0;
}
int
-cs_build_loc (loc_t *loc, call_frame_t *frame)
+cs_build_loc(loc_t *loc, call_frame_t *frame)
{
- cs_local_t *local = NULL;
- int ret = -1;
+ cs_local_t *local = NULL;
+ int ret = -1;
- local = frame->local;
+ local = frame->local;
- if (local->fd) {
- loc->inode = inode_ref (local->fd->inode);
- if (loc->inode) {
- gf_uuid_copy (loc->gfid, loc->inode->gfid);
- ret = 0;
- goto out;
- } else {
- ret = -1;
- goto out;
- }
+ if (local->fd) {
+ loc->inode = inode_ref(local->fd->inode);
+ if (loc->inode) {
+ gf_uuid_copy(loc->gfid, loc->inode->gfid);
+ ret = 0;
+ goto out;
} else {
- loc->inode = inode_ref (local->loc.inode);
- if (loc->inode) {
- gf_uuid_copy (loc->gfid, loc->inode->gfid);
- ret = 0;
- goto out;
- } else {
- ret = -1;
- goto out;
- }
+ ret = -1;
+ goto out;
}
+ } else {
+ loc->inode = inode_ref(local->loc.inode);
+ if (loc->inode) {
+ gf_uuid_copy(loc->gfid, loc->inode->gfid);
+ ret = 0;
+ goto out;
+ } else {
+ ret = -1;
+ goto out;
+ }
+ }
out:
- return ret;
+ return ret;
}
int
-cs_blocking_inodelk (call_frame_t *parent_frame)
+cs_blocking_inodelk(call_frame_t *parent_frame)
{
- call_frame_t *lock_frame = NULL;
- cs_local_t *lock_local = NULL;
- xlator_t *this = NULL;
- struct gf_flock flock = {0,};
- int ret = 0;
-
- lock_frame = cs_lock_frame (parent_frame);
- if (!lock_frame) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "insuffcient memory");
- goto err;
- }
-
- this = parent_frame->this;
-
- lock_local = cs_local_init (this, lock_frame, NULL, NULL, 0);
- if (!lock_local) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "local init failed");
- goto err;
- }
-
- lock_local->main_frame = parent_frame;
-
- flock.l_type = F_WRLCK;
-
- ret = cs_build_loc (&lock_local->loc, parent_frame);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "build_loc failed");
- goto err;
- }
-
- STACK_WIND (lock_frame, cs_blocking_inodelk_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->inodelk, CS_LOCK_DOMAIN,
- &lock_local->loc, F_SETLKW, &flock, NULL);
-
- return 0;
+ call_frame_t *lock_frame = NULL;
+ cs_local_t *lock_local = NULL;
+ xlator_t *this = NULL;
+ struct gf_flock flock = {
+ 0,
+ };
+ int ret = 0;
+
+ this = parent_frame->this;
+
+ lock_frame = cs_lock_frame(parent_frame);
+ if (!lock_frame) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "insuffcient memory");
+ goto err;
+ }
+
+ lock_local = cs_local_init(this, lock_frame, NULL, NULL, 0);
+ if (!lock_local) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "local init failed");
+ goto err;
+ }
+
+ lock_local->main_frame = parent_frame;
+
+ flock.l_type = F_WRLCK;
+
+ ret = cs_build_loc(&lock_local->loc, parent_frame);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "build_loc failed");
+ goto err;
+ }
+
+ STACK_WIND(lock_frame, cs_blocking_inodelk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->inodelk, CS_LOCK_DOMAIN,
+ &lock_local->loc, F_SETLKW, &flock, NULL);
+
+ return 0;
err:
- if (lock_frame)
- cs_lock_wipe (lock_frame);
+ if (lock_frame)
+ cs_lock_wipe(lock_frame);
- return -1;
+ return -1;
}
int
-locate_and_execute (call_frame_t *frame)
+locate_and_execute(call_frame_t *frame)
{
- int ret = 0;
+ int ret = 0;
- ret = cs_blocking_inodelk (frame);
+ ret = cs_blocking_inodelk(frame);
- if (ret)
- return -1;
- else
- return 0;
+ if (ret)
+ return -1;
+ else
+ return 0;
}
int32_t
-cs_resume_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc,
- off_t offset, dict_t *xattr_req)
+cs_resume_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ off_t offset, dict_t *xattr_req)
{
- cs_local_t *local = NULL;
- int ret = 0;
+ cs_local_t *local = NULL;
+ int ret = 0;
- local = frame->local;
+ local = frame->local;
- ret = cs_resume_postprocess (this, frame, loc->inode);
- if (ret) {
- goto unwind;
- }
+ ret = cs_resume_postprocess(this, frame, loc->inode);
+ if (ret) {
+ goto unwind;
+ }
- cs_inodelk_unlock (frame);
+ cs_inodelk_unlock(frame);
- STACK_WIND (frame, cs_truncate_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->truncate,
- loc, offset, local->xattr_req);
+ STACK_WIND(frame, cs_truncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, loc, offset,
+ local->xattr_req);
- return 0;
+ return 0;
unwind:
- cs_inodelk_unlock (frame);
+ cs_inodelk_unlock(frame);
- cs_common_cbk (frame);
+ cs_common_cbk(frame);
- return 0;
+ return 0;
}
-
int32_t
-cs_resume_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *dict, int32_t flags, dict_t *xdata)
+cs_resume_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ dict_t *dict, int32_t flags, dict_t *xdata)
{
- cs_local_t *local = NULL;
- cs_inode_ctx_t *ctx = NULL;
- gf_cs_obj_state state = GF_CS_ERROR;
-
- local = frame->local;
-
- __cs_inode_ctx_get (this, loc->inode, &ctx);
-
- state = __cs_get_file_state (this, loc->inode, ctx);
-
- if (state == GF_CS_ERROR) {
- /* file is already remote */
- local->op_ret = -1;
- local->op_errno = EINVAL;
- gf_msg (this->name, GF_LOG_WARNING, 0, 0,
- "file %s , could not figure file state", loc->path);
- goto unwind;
- }
-
-
- if (state == GF_CS_REMOTE) {
- /* file is already remote */
- local->op_ret = -1;
- local->op_errno = EINVAL;
- gf_msg (this->name, GF_LOG_WARNING, 0, EINVAL,
- "file %s is already remote", loc->path);
- goto unwind;
- }
-
- if (state == GF_CS_DOWNLOADING) {
- gf_msg (this->name, GF_LOG_WARNING, 0, 0,
- " file is in downloading state.");
- local->op_ret = -1;
- local->op_errno = EINVAL;
- goto unwind;
- }
-
- STACK_WIND (frame, cs_setxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setxattr, loc, dict, flags,
- local->xattr_req);
-
- return 0;
+ cs_local_t *local = NULL;
+ cs_inode_ctx_t *ctx = NULL;
+ gf_cs_obj_state state = GF_CS_ERROR;
+
+ local = frame->local;
+
+ __cs_inode_ctx_get(this, loc->inode, &ctx);
+
+ state = __cs_get_file_state(loc->inode, ctx);
+
+ if (state == GF_CS_ERROR) {
+ /* file is already remote */
+ local->op_ret = -1;
+ local->op_errno = EINVAL;
+ gf_msg(this->name, GF_LOG_WARNING, 0, 0,
+ "file %s , could not figure file state", loc->path);
+ goto unwind;
+ }
+
+ if (state == GF_CS_REMOTE) {
+ /* file is already remote */
+ local->op_ret = -1;
+ local->op_errno = EINVAL;
+ gf_msg(this->name, GF_LOG_WARNING, 0, EINVAL,
+ "file %s is already remote", loc->path);
+ goto unwind;
+ }
+
+ if (state == GF_CS_DOWNLOADING) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, 0,
+ " file is in downloading state.");
+ local->op_ret = -1;
+ local->op_errno = EINVAL;
+ goto unwind;
+ }
+
+ STACK_WIND(frame, cs_setxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, loc, dict, flags,
+ local->xattr_req);
+
+ return 0;
unwind:
- cs_inodelk_unlock (frame);
+ cs_inodelk_unlock(frame);
- cs_common_cbk (frame);
+ cs_common_cbk(frame);
- return 0;
+ return 0;
}
-
-
-
gf_cs_obj_state
-__cs_get_file_state (xlator_t *this, inode_t *inode, cs_inode_ctx_t *ctx)
+__cs_get_file_state(inode_t *inode, cs_inode_ctx_t *ctx)
{
- gf_cs_obj_state state = -1;
+ gf_cs_obj_state state = -1;
- if (!ctx)
- return GF_CS_ERROR;
+ if (!ctx)
+ return GF_CS_ERROR;
- LOCK (&inode->lock);
- {
- state = ctx->state;
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ state = ctx->state;
+ }
+ UNLOCK(&inode->lock);
- return state;
+ return state;
}
void
-__cs_inode_ctx_get (xlator_t *this, inode_t *inode, cs_inode_ctx_t **ctx)
+__cs_inode_ctx_get(xlator_t *this, inode_t *inode, cs_inode_ctx_t **ctx)
{
- uint64_t ctxint = 0;
- int ret = 0;
+ uint64_t ctxint = 0;
+ int ret = 0;
- LOCK (&inode->lock);
- {
- ret = __inode_ctx_get (inode, this, &ctxint);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __inode_ctx_get(inode, this, &ctxint);
+ }
+ UNLOCK(&inode->lock);
- if (ret)
- *ctx = NULL;
- else
- *ctx = (cs_inode_ctx_t *)ctxint;
+ if (ret)
+ *ctx = NULL;
+ else
+ *ctx = (cs_inode_ctx_t *)(uintptr_t)ctxint;
- return;
+ return;
}
int
-__cs_inode_ctx_update (xlator_t *this, inode_t *inode, uint64_t val)
+__cs_inode_ctx_update(xlator_t *this, inode_t *inode, uint64_t val)
{
- cs_inode_ctx_t *ctx = NULL;
- uint64_t ctxint = 0;
- int ret = 0;
+ cs_inode_ctx_t *ctx = NULL;
+ uint64_t ctxint = 0;
+ int ret = 0;
- LOCK (&inode->lock);
- {
- ret = __inode_ctx_get (inode, this, &ctxint);
+ LOCK(&inode->lock);
+ {
+ ret = __inode_ctx_get(inode, this, &ctxint);
if (ret) {
- ctx = GF_CALLOC (1, sizeof (*ctx), gf_cs_mt_cs_inode_ctx_t);
- if (!ctx) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0,
- "ctx allocation failed");
- ret = -1;
- goto out;
- }
+ ctx = GF_CALLOC(1, sizeof(*ctx), gf_cs_mt_cs_inode_ctx_t);
+ if (!ctx) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "ctx allocation failed");
+ ret = -1;
+ goto out;
+ }
- ctx->state = val;
+ ctx->state = val;
- ctxint = (uint64_t) ctx;
+ ctxint = (uint64_t)(uintptr_t)ctx;
- ret = __inode_ctx_set (inode, this, &ctxint);
- if (ret) {
- GF_FREE (ctx);
- goto out;
- }
+ ret = __inode_ctx_set(inode, this, &ctxint);
+ if (ret) {
+ GF_FREE(ctx);
+ goto out;
+ }
} else {
- ctx = (cs_inode_ctx_t *) ctxint;
-
- ctx->state = val;
- }
+ ctx = (cs_inode_ctx_t *)(uintptr_t)ctxint;
+ ctx->state = val;
}
+ }
out:
- UNLOCK (&inode->lock);
+ UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
int
-cs_inode_ctx_reset (xlator_t *this, inode_t *inode)
+cs_inode_ctx_reset(xlator_t *this, inode_t *inode)
{
- cs_inode_ctx_t *ctx = NULL;
- uint64_t ctxint = 0;
+ cs_inode_ctx_t *ctx = NULL;
+ uint64_t ctxint = 0;
- inode_ctx_del (inode, this, &ctxint);
- if (!ctxint) {
- return 0;
- }
+ inode_ctx_del(inode, this, &ctxint);
+ if (!ctxint) {
+ return 0;
+ }
- ctx = (cs_inode_ctx_t *)ctxint;
+ ctx = (cs_inode_ctx_t *)(uintptr_t)ctxint;
- GF_FREE (ctx);
- return 0;
+ GF_FREE(ctx);
+ return 0;
}
int
-cs_resume_postprocess (xlator_t *this, call_frame_t *frame, inode_t *inode)
+cs_resume_postprocess(xlator_t *this, call_frame_t *frame, inode_t *inode)
{
- cs_local_t *local = NULL;
- gf_cs_obj_state state = -1;
- cs_inode_ctx_t *ctx = NULL;
- int ret = 0;
-
- local = frame->local;
- if (!local) {
- ret = -1;
- goto out;
- }
-
- __cs_inode_ctx_get (this, inode, &ctx);
-
- state = __cs_get_file_state (this, inode, ctx);
- if (state == GF_CS_ERROR) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "status is GF_CS_ERROR."
- " Aborting write");
- local->op_ret = -1;
- local->op_errno = EREMOTE;
- ret = -1;
- goto out;
- }
-
- if (state == GF_CS_REMOTE || state == GF_CS_DOWNLOADING) {
- gf_msg_debug (this->name, 0, "status is %d", state);
- ret = cs_download (frame);
- if (ret == 0) {
- gf_msg_debug (this->name, 0, "Winding for Final Write");
- } else {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0,
- " download failed, unwinding writev");
- local->op_ret = -1;
- local->op_errno = EREMOTE;
- ret = -1;
- }
+ cs_local_t *local = NULL;
+ gf_cs_obj_state state = -1;
+ cs_inode_ctx_t *ctx = NULL;
+ int ret = 0;
+
+ local = frame->local;
+ if (!local) {
+ ret = -1;
+ goto out;
+ }
+
+ __cs_inode_ctx_get(this, inode, &ctx);
+
+ state = __cs_get_file_state(inode, ctx);
+ if (state == GF_CS_ERROR) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "status is GF_CS_ERROR."
+ " Aborting write");
+ local->op_ret = -1;
+ local->op_errno = EREMOTE;
+ ret = -1;
+ goto out;
+ }
+
+ if (state == GF_CS_REMOTE || state == GF_CS_DOWNLOADING) {
+ gf_msg_debug(this->name, 0, "status is %d", state);
+ ret = cs_download(frame);
+ if (ret == 0) {
+ gf_msg_debug(this->name, 0, "Winding for Final Write");
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ " download failed, unwinding writev");
+ local->op_ret = -1;
+ local->op_errno = EREMOTE;
+ ret = -1;
}
+ }
out:
- return ret;
+ return ret;
}
+
int32_t
-cs_fdctx_to_dict (xlator_t *this,
- fd_t *fd,
- dict_t *dict)
+cs_fdctx_to_dict(xlator_t *this, fd_t *fd, dict_t *dict)
{
- return 0;
+ return 0;
}
-
int32_t
-cs_inode (xlator_t *this)
+cs_inode(xlator_t *this)
{
- return 0;
+ return 0;
}
-
int32_t
-cs_inode_to_dict (xlator_t *this,
- dict_t *dict)
+cs_inode_to_dict(xlator_t *this, dict_t *dict)
{
- return 0;
+ return 0;
}
-
int32_t
-cs_history (xlator_t *this)
+cs_history(xlator_t *this)
{
- return 0;
+ return 0;
}
-
int32_t
-cs_fd (xlator_t *this)
+cs_fd(xlator_t *this)
{
- return 0;
+ return 0;
}
-
int32_t
-cs_fd_to_dict (xlator_t *this,
- dict_t *dict)
+cs_fd_to_dict(xlator_t *this, dict_t *dict)
{
- return 0;
+ return 0;
}
-
int32_t
-cs_fdctx (xlator_t *this,
- fd_t *fd)
+cs_fdctx(xlator_t *this, fd_t *fd)
{
- return 0;
+ return 0;
}
-
int32_t
-cs_inodectx (xlator_t *this,
- inode_t *ino)
+cs_inodectx(xlator_t *this, inode_t *ino)
{
- return 0;
+ return 0;
}
-
int32_t
-cs_inodectx_to_dict (xlator_t *this,
- inode_t *ino,
- dict_t *dict)
+cs_inodectx_to_dict(xlator_t *this, inode_t *ino, dict_t *dict)
{
- return 0;
+ return 0;
}
-
int32_t
-cs_priv_to_dict (xlator_t *this,
- dict_t *dict, char *brickname)
+cs_priv_to_dict(xlator_t *this, dict_t *dict, char *brickname)
{
- return 0;
+ return 0;
}
-
int32_t
-cs_priv (xlator_t *this)
+cs_priv(xlator_t *this)
{
- return 0;
+ return 0;
}
int
-cs_notify (xlator_t *this, int event, void *data, ...)
+cs_notify(xlator_t *this, int event, void *data, ...)
{
- return default_notify (this, event, data);
+ return default_notify(this, event, data);
}
-
struct xlator_fops cs_fops = {
- .stat = cs_stat,
- .readdirp = cs_readdirp,
- .truncate = cs_truncate,
- .seek = cs_seek,
- .statfs = cs_statfs,
- .fallocate = cs_fallocate,
- .discard = cs_discard,
- .getxattr = cs_getxattr,
- .writev = cs_writev,
- .setxattr = cs_setxattr,
- .fgetxattr = cs_fgetxattr,
- .lookup = cs_lookup,
- .fsetxattr = cs_fsetxattr,
- .readv = cs_readv,
- .ftruncate = cs_ftruncate,
- .rchecksum = cs_rchecksum,
- .unlink = cs_unlink,
- .open = cs_open,
- .fstat = cs_fstat,
- .zerofill = cs_zerofill,
+ .stat = cs_stat,
+ .readdirp = cs_readdirp,
+ .truncate = cs_truncate,
+ .seek = cs_seek,
+ .statfs = cs_statfs,
+ .fallocate = cs_fallocate,
+ .discard = cs_discard,
+ .getxattr = cs_getxattr,
+ .writev = cs_writev,
+ .setxattr = cs_setxattr,
+ .fgetxattr = cs_fgetxattr,
+ .lookup = cs_lookup,
+ .fsetxattr = cs_fsetxattr,
+ .readv = cs_readv,
+ .ftruncate = cs_ftruncate,
+ .rchecksum = cs_rchecksum,
+ .unlink = cs_unlink,
+ .open = cs_open,
+ .fstat = cs_fstat,
+ .zerofill = cs_zerofill,
};
struct xlator_cbks cs_cbks = {
+ .forget = cs_forget,
};
struct xlator_dumpops cs_dumpops = {
- .fdctx_to_dict = cs_fdctx_to_dict,
- .inode = cs_inode,
- .inode_to_dict = cs_inode_to_dict,
- .history = cs_history,
- .fd = cs_fd,
- .fd_to_dict = cs_fd_to_dict,
- .fdctx = cs_fdctx,
- .inodectx = cs_inodectx,
- .inodectx_to_dict = cs_inodectx_to_dict,
- .priv_to_dict = cs_priv_to_dict,
- .priv = cs_priv,
+ .fdctx_to_dict = cs_fdctx_to_dict,
+ .inode = cs_inode,
+ .inode_to_dict = cs_inode_to_dict,
+ .history = cs_history,
+ .fd = cs_fd,
+ .fd_to_dict = cs_fd_to_dict,
+ .fdctx = cs_fdctx,
+ .inodectx = cs_inodectx,
+ .inodectx_to_dict = cs_inodectx_to_dict,
+ .priv_to_dict = cs_priv_to_dict,
+ .priv = cs_priv,
};
struct volume_options cs_options[] = {
- { .key = {NULL} },
+ {.key = {"cloudsync-storetype"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Defines which remote store is enabled"},
+ {.key = {"cloudsync-remote-read"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .description = "Defines a remote read fop when on"},
+ {.key = {"cloudsync-store-id"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Defines a volume wide store id"},
+ {.key = {"cloudsync-product-id"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Defines a volume wide product id"},
+ {.key = {NULL}},
};
xlator_api_t xlator_api = {
- .init = cs_init,
- .fini = cs_fini,
- .notify = cs_notify,
- .reconfigure = cs_reconfigure,
- .mem_acct_init = cs_mem_acct_init,
- .dumpops = &cs_dumpops,
- .fops = &cs_fops,
- .cbks = &cs_cbks,
- .options = cs_options,
- .identifier = "cloudsync",
+ .init = cs_init,
+ .fini = cs_fini,
+ .notify = cs_notify,
+ .reconfigure = cs_reconfigure,
+ .mem_acct_init = cs_mem_acct_init,
+ .dumpops = &cs_dumpops,
+ .fops = &cs_fops,
+ .cbks = &cs_cbks,
+ .options = cs_options,
+ .identifier = "cloudsync",
+ .category = GF_TECH_PREVIEW,
};
diff --git a/xlators/features/cloudsync/src/cloudsync.h b/xlators/features/cloudsync/src/cloudsync.h
index bd54d760864..d24141978d6 100644
--- a/xlators/features/cloudsync/src/cloudsync.h
+++ b/xlators/features/cloudsync/src/cloudsync.h
@@ -8,93 +8,116 @@
* cases as published by the Free Software Foundation.
*/
-
#ifndef __CLOUDSYNC_H__
#define __CLOUDSYNC_H__
-#include "glusterfs.h"
-#include "xlator.h"
-#include "defaults.h"
-#include "syncop.h"
-#include "call-stub.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
+#include <glusterfs/syncop.h>
+#include <glusterfs/call-stub.h>
#include "cloudsync-common.h"
#include "cloudsync-autogen-fops.h"
-
+#define ALIGN_SIZE 4096
#define CS_LOCK_DOMAIN "cs.protect.file.stat"
typedef struct cs_dlstore {
- off_t off;
- struct iovec *vector;
- int32_t count;
- struct iobref *iobref;
- uint32_t flags;
+ off_t off;
+ struct iovec *vector;
+ int32_t count;
+ struct iobref *iobref;
+ uint32_t flags;
} cs_dlstore;
typedef struct cs_inode_ctx {
- gf_cs_obj_state state;
+ cs_loc_xattr_t locxattr;
+ gf_cs_obj_state state;
} cs_inode_ctx_t;
+struct cs_plugin {
+ char *name; /* store name */
+ char *library; /* library to load for the given store */
+ char *description; /* description about the store */
+};
+
cs_local_t *
-cs_local_init (xlator_t *this, call_frame_t *frame, loc_t *loc, fd_t *fd,
- glusterfs_fop_t fop);
+cs_local_init(xlator_t *this, call_frame_t *frame, loc_t *loc, fd_t *fd,
+ glusterfs_fop_t fop);
int
-locate_and_execute (call_frame_t *frame);
-
+locate_and_execute(call_frame_t *frame);
int32_t
-cs_resume_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *dict, int32_t flags, dict_t *xdata);
+cs_resume_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ dict_t *dict, int32_t flags, dict_t *xdata);
int32_t
-cs_inodelk_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xdata);
+cs_inodelk_unlock_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata);
size_t
-cs_write_callback (void *lcurlbuf, size_t size, size_t nitems, void *frame);
+cs_write_callback(void *lcurlbuf, size_t size, size_t nitems, void *frame);
void
-cs_common_cbk (call_frame_t *frame);
+cs_common_cbk(call_frame_t *frame);
gf_boolean_t
-cs_is_file_remote (struct iatt *stbuf, dict_t *xattr);
+cs_is_file_remote(struct iatt *stbuf, dict_t *xattr);
int32_t
-cs_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata);
+cs_setxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata);
int
-cs_build_loc (loc_t *loc, call_frame_t *frame);
+cs_build_loc(loc_t *loc, call_frame_t *frame);
int
-cs_blocking_inodelk_cbk (call_frame_t *lock_frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata);
+cs_blocking_inodelk_cbk(call_frame_t *lock_frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata);
-int cs_read_authinfo(xlator_t *this);
+int
+cs_read_authinfo(xlator_t *this);
int
-__cs_inode_ctx_update (xlator_t *this, inode_t *inode, uint64_t val);
+__cs_inode_ctx_update(xlator_t *this, inode_t *inode, uint64_t val);
int
-cs_inode_ctx_reset (xlator_t *this, inode_t *inode);
+cs_inode_ctx_reset(xlator_t *this, inode_t *inode);
void
-__cs_inode_ctx_get (xlator_t *this, inode_t *inode, cs_inode_ctx_t **ctx);
+__cs_inode_ctx_get(xlator_t *this, inode_t *inode, cs_inode_ctx_t **ctx);
gf_cs_obj_state
-__cs_get_file_state (xlator_t *this, inode_t *inode, cs_inode_ctx_t *ctx);
+__cs_get_file_state(inode_t *inode, cs_inode_ctx_t *ctx);
int
-cs_inodelk_unlock (call_frame_t *main_frame);
+cs_inodelk_unlock(call_frame_t *main_frame);
int
-cs_resume_postprocess (xlator_t *this, call_frame_t *frame, inode_t *inode);
+cs_resume_postprocess(xlator_t *this, call_frame_t *frame, inode_t *inode);
int32_t
-cs_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata);
+cs_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata);
int32_t
-cs_resume_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc,
- off_t offset, dict_t *xattr_req);
-#endif /* __CLOUDSYNC_H__ */
+cs_resume_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ off_t offset, dict_t *xattr_req);
+int32_t
+cs_readv_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iovec *vector, int32_t count,
+ struct iatt *stbuf, struct iobref *iobref, dict_t *xdata);
+int32_t
+cs_resume_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata);
+int32_t
+cs_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata);
+
+int
+cs_resume_remote_readv_postprocess(xlator_t *this, call_frame_t *frame,
+ inode_t *inode, off_t offset, size_t size,
+ uint32_t flags);
+int
+cs_serve_readv(call_frame_t *frame, off_t offset, size_t size, uint32_t flags);
+#endif /* __CLOUDSYNC_H__ */
diff --git a/xlators/features/compress/src/cdc-helper.c b/xlators/features/compress/src/cdc-helper.c
index 0a9a0e3d29c..f973ff56cf5 100644
--- a/xlators/features/compress/src/cdc-helper.c
+++ b/xlators/features/compress/src/cdc-helper.c
@@ -8,9 +8,9 @@
cases as published by the Free Software Foundation.
*/
-#include "glusterfs.h"
-#include "logging.h"
-#include "syscall.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/syscall.h>
#include "cdc.h"
#include "cdc-mem-types.h"
@@ -34,118 +34,110 @@
* gzip_header is added only during debugging.
* Refer to the function cdc_dump_iovec_to_disk
*/
-static const char gzip_header[10] =
- {
- '\037', '\213', Z_DEFLATED, 0,
- 0, 0, 0, 0,
- 0, GF_CDC_OS_ID
- };
+static const char gzip_header[10] = {'\037', '\213', Z_DEFLATED, 0, 0, 0, 0,
+ 0, 0, GF_CDC_OS_ID};
static int32_t
-cdc_next_iovec (xlator_t *this, cdc_info_t *ci)
+cdc_next_iovec(xlator_t *this, cdc_info_t *ci)
{
- int ret = -1;
-
- ci->ncount++;
- /* check for iovec overflow -- should not happen */
- if (ci->ncount == MAX_IOVEC) {
- gf_log (this->name, GF_LOG_ERROR,
- "Zlib output buffer overflow"
- " ->ncount (%d) | ->MAX_IOVEC (%d)",
- ci->ncount, MAX_IOVEC);
- goto out;
- }
-
- ret = 0;
-
- out:
- return ret;
+ int ret = -1;
+
+ ci->ncount++;
+ /* check for iovec overflow -- should not happen */
+ if (ci->ncount == MAX_IOVEC) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Zlib output buffer overflow"
+ " ->ncount (%d) | ->MAX_IOVEC (%d)",
+ ci->ncount, MAX_IOVEC);
+ goto out;
+ }
+
+ ret = 0;
+
+out:
+ return ret;
}
static void
-cdc_put_long (unsigned char *string, unsigned long x)
+cdc_put_long(unsigned char *string, unsigned long x)
{
- string[0] = (unsigned char) (x & 0xff);
- string[1] = (unsigned char) ((x & 0xff00) >> 8);
- string[2] = (unsigned char) ((x & 0xff0000) >> 16);
- string[3] = (unsigned char) ((x & 0xff000000) >> 24);
+ string[0] = (unsigned char)(x & 0xff);
+ string[1] = (unsigned char)((x & 0xff00) >> 8);
+ string[2] = (unsigned char)((x & 0xff0000) >> 16);
+ string[3] = (unsigned char)((x & 0xff000000) >> 24);
}
static unsigned long
-cdc_get_long (unsigned char *buf)
+cdc_get_long(unsigned char *buf)
{
- return ((unsigned long) buf[0])
- | (((unsigned long) buf[1]) << 8)
- | (((unsigned long) buf[2]) << 16)
- | (((unsigned long) buf[3]) << 24);
+ return ((unsigned long)buf[0]) | (((unsigned long)buf[1]) << 8) |
+ (((unsigned long)buf[2]) << 16) | (((unsigned long)buf[3]) << 24);
}
static int32_t
-cdc_init_gzip_trailer (xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci)
+cdc_init_gzip_trailer(xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci)
{
- int ret = -1;
- char *buf = NULL;
+ int ret = -1;
+ char *buf = NULL;
- ret = cdc_next_iovec (this, ci);
- if (ret)
- goto out;
+ ret = cdc_next_iovec(this, ci);
+ if (ret)
+ goto out;
- buf = CURR_VEC(ci).iov_base =
- (char *) GF_CALLOC (1, GF_CDC_VALIDATION_SIZE,
- gf_cdc_mt_gzip_trailer_t);
+ buf = CURR_VEC(ci).iov_base = (char *)GF_CALLOC(1, GF_CDC_VALIDATION_SIZE,
+ gf_cdc_mt_gzip_trailer_t);
- if (!CURR_VEC(ci).iov_base)
- goto out;
+ if (!CURR_VEC(ci).iov_base)
+ goto out;
- CURR_VEC(ci).iov_len = GF_CDC_VALIDATION_SIZE;
+ CURR_VEC(ci).iov_len = GF_CDC_VALIDATION_SIZE;
- cdc_put_long ((unsigned char *)&buf[0], ci->crc);
- cdc_put_long ((unsigned char *)&buf[4], ci->stream.total_in);
+ cdc_put_long((unsigned char *)&buf[0], ci->crc);
+ cdc_put_long((unsigned char *)&buf[4], ci->stream.total_in);
- ret = 0;
+ ret = 0;
- out:
- return ret;
+out:
+ return ret;
}
static int32_t
-cdc_alloc_iobuf_and_init_vec (xlator_t *this,
- cdc_priv_t *priv, cdc_info_t *ci,
- int size)
+cdc_alloc_iobuf_and_init_vec(xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci,
+ int size)
{
- int ret = -1;
- int alloc_len = 0;
- struct iobuf *iobuf = NULL;
+ int ret = -1;
+ int alloc_len = 0;
+ struct iobuf *iobuf = NULL;
- ret = cdc_next_iovec (this, ci);
- if (ret)
- goto out;
+ ret = cdc_next_iovec(this, ci);
+ if (ret)
+ goto out;
- alloc_len = size ? size : ci->buffer_size;
+ alloc_len = size ? size : ci->buffer_size;
- iobuf = iobuf_get2 (this->ctx->iobuf_pool, alloc_len);
- if (!iobuf)
- goto out;
+ iobuf = iobuf_get2(this->ctx->iobuf_pool, alloc_len);
+ if (!iobuf)
+ goto out;
- ret = iobref_add (ci->iobref, iobuf);
- if (ret)
- goto out;
+ ret = iobref_add(ci->iobref, iobuf);
+ if (ret)
+ goto out;
- /* Initialize this iovec */
- CURR_VEC(ci).iov_base = iobuf->ptr;
- CURR_VEC(ci).iov_len = alloc_len;
+ /* Initialize this iovec */
+ CURR_VEC(ci).iov_base = iobuf->ptr;
+ CURR_VEC(ci).iov_len = alloc_len;
- ret = 0;
+ ret = 0;
- out:
- return ret;
+out:
+ return ret;
}
static void
-cdc_init_zlib_output_stream (cdc_priv_t *priv, cdc_info_t *ci, int size)
+cdc_init_zlib_output_stream(cdc_priv_t *priv, cdc_info_t *ci, int size)
{
- ci->stream.next_out = (unsigned char *) CURR_VEC(ci).iov_base;
- ci->stream.avail_out = size ? size : ci->buffer_size;
+ ci->stream.next_out = (unsigned char *)CURR_VEC(ci).iov_base;
+ ci->stream.avail_out = size ? size : ci->buffer_size;
}
/* This routine is for testing and debugging only.
@@ -153,391 +145,383 @@ cdc_init_zlib_output_stream (cdc_priv_t *priv, cdc_info_t *ci, int size)
* So each gzip dump file is at least 18 bytes in size.
*/
void
-cdc_dump_iovec_to_disk (xlator_t *this, cdc_info_t *ci, const char *file)
+cdc_dump_iovec_to_disk(xlator_t *this, cdc_info_t *ci, const char *file)
{
- int i = 0;
- int fd = 0;
- size_t written = 0;
- size_t total_written = 0;
-
- fd = open (file, O_WRONLY|O_CREAT|O_TRUNC, 0777 );
- if (fd < 0) {
- gf_log (this->name, GF_LOG_ERROR,
- "Cannot open file: %s", file);
- return;
- }
-
- written = sys_write (fd, (char *) gzip_header, 10);
+ int i = 0;
+ int fd = 0;
+ size_t written = 0;
+ size_t total_written = 0;
+
+ fd = open(file, O_WRONLY | O_CREAT | O_TRUNC, 0777);
+ if (fd < 0) {
+ gf_log(this->name, GF_LOG_ERROR, "Cannot open file: %s", file);
+ return;
+ }
+
+ written = sys_write(fd, (char *)gzip_header, 10);
+ total_written += written;
+ for (i = 0; i < ci->ncount; i++) {
+ written = sys_write(fd, (char *)ci->vec[i].iov_base,
+ ci->vec[i].iov_len);
total_written += written;
- for (i = 0; i < ci->ncount; i++) {
- written = sys_write (fd, (char *) ci->vec[i].iov_base, ci->vec[i].iov_len);
- total_written += written;
- }
+ }
- gf_log (this->name, GF_LOG_DEBUG,
- "dump'd %zu bytes to %s", total_written, GF_CDC_DEBUG_DUMP_FILE );
+ gf_log(this->name, GF_LOG_DEBUG, "dump'd %zu bytes to %s", total_written,
+ GF_CDC_DEBUG_DUMP_FILE);
- sys_close (fd);
+ sys_close(fd);
}
static int32_t
-cdc_flush_libz_buffer (cdc_priv_t *priv, xlator_t *this, cdc_info_t *ci,
- int (*libz_func)(z_streamp, int),
- int flush)
+cdc_flush_libz_buffer(cdc_priv_t *priv, xlator_t *this, cdc_info_t *ci,
+ int (*libz_func)(z_streamp, int), int flush)
{
- int32_t ret = Z_OK;
- int done = 0;
- unsigned int deflate_len = 0;
+ int32_t ret = Z_OK;
+ int done = 0;
+ unsigned int deflate_len = 0;
- for (;;) {
- deflate_len = ci->buffer_size - ci->stream.avail_out;
+ for (;;) {
+ deflate_len = ci->buffer_size - ci->stream.avail_out;
- if (deflate_len != 0) {
- CURR_VEC(ci).iov_len = deflate_len;
+ if (deflate_len != 0) {
+ CURR_VEC(ci).iov_len = deflate_len;
- ret = cdc_alloc_iobuf_and_init_vec (this, priv, ci, 0);
- if (ret) {
- ret = Z_MEM_ERROR;
- break;
- }
+ ret = cdc_alloc_iobuf_and_init_vec(this, priv, ci, 0);
+ if (ret) {
+ ret = Z_MEM_ERROR;
+ break;
+ }
- /* Re-position Zlib output buffer */
- cdc_init_zlib_output_stream (priv, ci, 0);
- }
+ /* Re-position Zlib output buffer */
+ cdc_init_zlib_output_stream(priv, ci, 0);
+ }
- if (done) {
- ci->ncount--;
- break;
- }
+ if (done) {
+ ci->ncount--;
+ break;
+ }
- ret = libz_func (&ci->stream, flush);
+ ret = libz_func(&ci->stream, flush);
- if (ret == Z_BUF_ERROR) {
- ret = Z_OK;
- ci->ncount--;
- break;
- }
+ if (ret == Z_BUF_ERROR) {
+ ret = Z_OK;
+ ci->ncount--;
+ break;
+ }
- done = (ci->stream.avail_out != 0 || ret == Z_STREAM_END);
+ done = (ci->stream.avail_out != 0 || ret == Z_STREAM_END);
- if (ret != Z_OK && ret != Z_STREAM_END)
- break;
- }
+ if (ret != Z_OK && ret != Z_STREAM_END)
+ break;
+ }
- return ret;
+ return ret;
}
static int32_t
-do_cdc_compress (struct iovec *vec, xlator_t *this, cdc_priv_t *priv,
- cdc_info_t *ci)
+do_cdc_compress(struct iovec *vec, xlator_t *this, cdc_priv_t *priv,
+ cdc_info_t *ci)
{
- int ret = -1;
+ int ret = -1;
- /* Initialize defalte */
- ret = deflateInit2 (&ci->stream, priv->cdc_level, Z_DEFLATED,
- priv->window_size, priv->mem_level,
- Z_DEFAULT_STRATEGY);
+ /* Initialize defalte */
+ ret = deflateInit2(&ci->stream, priv->cdc_level, Z_DEFLATED,
+ priv->window_size, priv->mem_level, Z_DEFAULT_STRATEGY);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "unable to init Zlib (retval: %d)", ret);
- goto out;
- }
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR, "unable to init Zlib (retval: %d)",
+ ret);
+ goto out;
+ }
- ret = cdc_alloc_iobuf_and_init_vec (this, priv, ci, 0);
- if (ret)
- goto out;
+ ret = cdc_alloc_iobuf_and_init_vec(this, priv, ci, 0);
+ if (ret)
+ goto out;
- /* setup output buffer */
- cdc_init_zlib_output_stream (priv, ci, 0);
+ /* setup output buffer */
+ cdc_init_zlib_output_stream(priv, ci, 0);
- /* setup input buffer */
- ci->stream.next_in = (unsigned char *) vec->iov_base;
- ci->stream.avail_in = vec->iov_len;
+ /* setup input buffer */
+ ci->stream.next_in = (unsigned char *)vec->iov_base;
+ ci->stream.avail_in = vec->iov_len;
- ci->crc = crc32 (ci->crc, (const Bytef *) vec->iov_base, vec->iov_len);
+ ci->crc = crc32(ci->crc, (const Bytef *)vec->iov_base, vec->iov_len);
- gf_log (this->name, GF_LOG_DEBUG, "crc=%lu len=%d buffer_size=%d",
- ci->crc, ci->stream.avail_in, ci->buffer_size);
+ gf_log(this->name, GF_LOG_DEBUG, "crc=%lu len=%d buffer_size=%d", ci->crc,
+ ci->stream.avail_in, ci->buffer_size);
- /* compress !! */
- while (ci->stream.avail_in != 0) {
- if (ci->stream.avail_out == 0) {
+ /* compress !! */
+ while (ci->stream.avail_in != 0) {
+ if (ci->stream.avail_out == 0) {
+ CURR_VEC(ci).iov_len = ci->buffer_size;
- CURR_VEC(ci).iov_len = ci->buffer_size;
+ ret = cdc_alloc_iobuf_and_init_vec(this, priv, ci, 0);
+ if (ret)
+ break;
- ret = cdc_alloc_iobuf_and_init_vec (this, priv, ci, 0);
- if (ret)
- break;
-
- /* Re-position Zlib output buffer */
- cdc_init_zlib_output_stream (priv, ci, 0);
- }
-
- ret = deflate (&ci->stream, Z_NO_FLUSH);
- if (ret != Z_OK)
- break;
+ /* Re-position Zlib output buffer */
+ cdc_init_zlib_output_stream(priv, ci, 0);
}
- out:
- return ret;
+ ret = deflate(&ci->stream, Z_NO_FLUSH);
+ if (ret != Z_OK)
+ break;
+ }
+
+out:
+ return ret;
}
int32_t
-cdc_compress (xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci,
- dict_t **xdata)
+cdc_compress(xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci, dict_t **xdata)
{
- int ret = -1;
- int i = 0;
+ int ret = -1;
+ int i = 0;
- ci->iobref = iobref_new ();
- if (!ci->iobref)
- goto out;
+ ci->iobref = iobref_new();
+ if (!ci->iobref)
+ goto out;
+ if (!*xdata) {
+ *xdata = dict_new();
if (!*xdata) {
- *xdata = dict_new ();
- if (!*xdata) {
- gf_log (this->name, GF_LOG_ERROR, "Cannot allocate xdata"
- " dict");
- goto out;
- }
- }
-
- /* data */
- for (i = 0; i < ci->count; i++) {
- ret = do_cdc_compress (&ci->vector[i], this, priv, ci);
- if (ret != Z_OK)
- goto deflate_cleanup_out;
- }
-
- /* flush zlib buffer */
- ret = cdc_flush_libz_buffer (priv, this, ci, deflate, Z_FINISH);
- if (!(ret == Z_OK || ret == Z_STREAM_END)) {
- gf_log (this->name, GF_LOG_ERROR,
- "Compression Error: ret (%d)", ret);
- ret = -1;
- goto deflate_cleanup_out;
- }
-
- /* trailer */
- ret = cdc_init_gzip_trailer (this, priv, ci);
- if (ret)
- goto deflate_cleanup_out;
-
- gf_log (this->name, GF_LOG_DEBUG,
- "Compressed %ld to %ld bytes",
- ci->stream.total_in, ci->stream.total_out);
-
- ci->nbytes = ci->stream.total_out + GF_CDC_VALIDATION_SIZE;
-
- /* set deflated canary value for identification */
- ret = dict_set_int32 (*xdata, GF_CDC_DEFLATE_CANARY_VAL, 1);
- if (ret) {
- /* Send uncompressed data if we can't _tell_ the client
- * that deflated data is on it's way. So, we just log
- * the faliure and continue as usual.
- */
- gf_log (this->name, GF_LOG_ERROR,
- "Data deflated, but could not set canary"
- " value in dict for identification");
+ gf_log(this->name, GF_LOG_ERROR,
+ "Cannot allocate xdata"
+ " dict");
+ goto out;
}
+ }
+
+ /* data */
+ for (i = 0; i < ci->count; i++) {
+ ret = do_cdc_compress(&ci->vector[i], this, priv, ci);
+ if (ret != Z_OK)
+ goto deflate_cleanup_out;
+ }
+
+ /* flush zlib buffer */
+ ret = cdc_flush_libz_buffer(priv, this, ci, deflate, Z_FINISH);
+ if (!(ret == Z_OK || ret == Z_STREAM_END)) {
+ gf_log(this->name, GF_LOG_ERROR, "Compression Error: ret (%d)", ret);
+ ret = -1;
+ goto deflate_cleanup_out;
+ }
+
+ /* trailer */
+ ret = cdc_init_gzip_trailer(this, priv, ci);
+ if (ret)
+ goto deflate_cleanup_out;
+
+ gf_log(this->name, GF_LOG_DEBUG, "Compressed %ld to %ld bytes",
+ ci->stream.total_in, ci->stream.total_out);
+
+ ci->nbytes = ci->stream.total_out + GF_CDC_VALIDATION_SIZE;
+
+ /* set deflated canary value for identification */
+ ret = dict_set_int32(*xdata, GF_CDC_DEFLATE_CANARY_VAL, 1);
+ if (ret) {
+ /* Send uncompressed data if we can't _tell_ the client
+ * that deflated data is on it's way. So, we just log
+ * the failure and continue as usual.
+ */
+ gf_log(this->name, GF_LOG_ERROR,
+ "Data deflated, but could not set canary"
+ " value in dict for identification");
+ }
- /* This is to be used in testing */
- if ( priv->debug ) {
- cdc_dump_iovec_to_disk (this, ci, GF_CDC_DEBUG_DUMP_FILE );
- }
+ /* This is to be used in testing */
+ if (priv->debug) {
+ cdc_dump_iovec_to_disk(this, ci, GF_CDC_DEBUG_DUMP_FILE);
+ }
- deflate_cleanup_out:
- (void) deflateEnd(&ci->stream);
+deflate_cleanup_out:
+ (void)deflateEnd(&ci->stream);
- out:
- return ret;
+out:
+ return ret;
}
-
/* deflate content is checked by the presence of a canary
* value in the dict as the key
*/
static int32_t
-cdc_check_content_for_deflate (dict_t *xdata)
+cdc_check_content_for_deflate(dict_t *xdata)
{
- return dict_get (xdata, GF_CDC_DEFLATE_CANARY_VAL) ? -1 : 0;
+ return dict_get(xdata, GF_CDC_DEFLATE_CANARY_VAL) ? -1 : 0;
}
static unsigned long
-cdc_extract_crc (char *trailer)
+cdc_extract_crc(char *trailer)
{
- return cdc_get_long ((unsigned char *) &trailer[0]);
+ return cdc_get_long((unsigned char *)&trailer[0]);
}
static unsigned long
-cdc_extract_size (char *trailer)
+cdc_extract_size(char *trailer)
{
- return cdc_get_long ((unsigned char *) &trailer[4]);
+ return cdc_get_long((unsigned char *)&trailer[4]);
}
static int32_t
-cdc_validate_inflate (cdc_info_t *ci, unsigned long crc,
- unsigned long len)
+cdc_validate_inflate(cdc_info_t *ci, unsigned long crc, unsigned long len)
{
- return !((crc == ci->crc)
- /* inflated length is hidden inside
- * Zlib stream struct */
- && (len == ci->stream.total_out));
+ return !((crc == ci->crc)
+ /* inflated length is hidden inside
+ * Zlib stream struct */
+ && (len == ci->stream.total_out));
}
static int32_t
-do_cdc_decompress (xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci)
+do_cdc_decompress(xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci)
{
- int ret = -1;
- int i = 0;
- int len = 0;
- char *inflte = NULL;
- char *trailer = NULL;
- struct iovec vec = {0,};
- unsigned long computed_crc = 0;
- unsigned long computed_len = 0;
-
- ret = inflateInit2 (&ci->stream, priv->window_size);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "Zlib: Unable to initialize inflate");
- goto out;
+ int ret = -1;
+ int i = 0;
+ int len = 0;
+ char *inflte = NULL;
+ char *trailer = NULL;
+ struct iovec vec = {
+ 0,
+ };
+ unsigned long computed_crc = 0;
+ unsigned long computed_len = 0;
+
+ ret = inflateInit2(&ci->stream, priv->window_size);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR, "Zlib: Unable to initialize inflate");
+ goto out;
+ }
+
+ vec = THIS_VEC(ci, 0);
+
+ trailer = (char *)(((char *)vec.iov_base) + vec.iov_len -
+ GF_CDC_VALIDATION_SIZE);
+
+ /* CRC of uncompressed data */
+ computed_crc = cdc_extract_crc(trailer);
+
+ /* size of uncomrpessed data */
+ computed_len = cdc_extract_size(trailer);
+
+ gf_log(this->name, GF_LOG_DEBUG, "crc=%lu len=%lu buffer_size=%d",
+ computed_crc, computed_len, ci->buffer_size);
+
+ inflte = vec.iov_base;
+ len = vec.iov_len - GF_CDC_VALIDATION_SIZE;
+
+ /* allocate buffer of the original length of the data */
+ ret = cdc_alloc_iobuf_and_init_vec(this, priv, ci, 0);
+ if (ret)
+ goto out;
+
+ /* setup output buffer */
+ cdc_init_zlib_output_stream(priv, ci, 0);
+
+ /* setup input buffer */
+ ci->stream.next_in = (unsigned char *)inflte;
+ ci->stream.avail_in = len;
+
+ while (ci->stream.avail_in != 0) {
+ if (ci->stream.avail_out == 0) {
+ CURR_VEC(ci).iov_len = ci->buffer_size;
+
+ ret = cdc_alloc_iobuf_and_init_vec(this, priv, ci, 0);
+ if (ret)
+ break;
+
+ /* Re-position Zlib output buffer */
+ cdc_init_zlib_output_stream(priv, ci, 0);
}
- vec = THIS_VEC(ci, 0);
-
- trailer = (char *) (((char *) vec.iov_base) + vec.iov_len
- - GF_CDC_VALIDATION_SIZE);
-
- /* CRC of uncompressed data */
- computed_crc = cdc_extract_crc (trailer);
-
- /* size of uncomrpessed data */
- computed_len = cdc_extract_size (trailer);
-
- gf_log (this->name, GF_LOG_DEBUG, "crc=%lu len=%lu buffer_size=%d",
- computed_crc, computed_len, ci->buffer_size);
-
- inflte = vec.iov_base ;
- len = vec.iov_len - GF_CDC_VALIDATION_SIZE;
-
- /* allocate buffer of the original length of the data */
- ret = cdc_alloc_iobuf_and_init_vec (this, priv, ci, 0);
- if (ret)
- goto out;
-
- /* setup output buffer */
- cdc_init_zlib_output_stream (priv, ci, 0);
-
- /* setup input buffer */
- ci->stream.next_in = (unsigned char *) inflte;
- ci->stream.avail_in = len;
-
- while (ci->stream.avail_in != 0) {
- if (ci->stream.avail_out == 0) {
- CURR_VEC(ci).iov_len = ci->buffer_size;
-
- ret = cdc_alloc_iobuf_and_init_vec (this, priv, ci, 0);
- if (ret)
- break;
-
- /* Re-position Zlib output buffer */
- cdc_init_zlib_output_stream (priv, ci, 0);
- }
-
- ret = inflate (&ci->stream, Z_NO_FLUSH);
- if (ret == Z_STREAM_ERROR)
- break;
- }
-
- /* flush zlib buffer */
- ret = cdc_flush_libz_buffer (priv, this, ci, inflate, Z_SYNC_FLUSH);
- if (!(ret == Z_OK || ret == Z_STREAM_END)) {
- gf_log (this->name, GF_LOG_ERROR,
- "Decompression Error: ret (%d)", ret);
- ret = -1;
- goto out;
- }
-
- /* compute CRC of the uncompresses data to check for
- * correctness */
-
- for (i = 0; i < ci->ncount; i++) {
- ci->crc = crc32 (ci->crc,
- (const Bytef *) ci->vec[i].iov_base,
- ci->vec[i].iov_len);
- }
-
- /* validate inflated data */
- ret = cdc_validate_inflate (ci, computed_crc, computed_len);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "Checksum or length mismatched in inflated data");
- }
-
- out:
- return ret;
+ ret = inflate(&ci->stream, Z_NO_FLUSH);
+ if (ret == Z_STREAM_ERROR)
+ break;
+ }
+
+ /* flush zlib buffer */
+ ret = cdc_flush_libz_buffer(priv, this, ci, inflate, Z_SYNC_FLUSH);
+ if (!(ret == Z_OK || ret == Z_STREAM_END)) {
+ gf_log(this->name, GF_LOG_ERROR, "Decompression Error: ret (%d)", ret);
+ ret = -1;
+ goto out;
+ }
+
+ /* compute CRC of the uncompresses data to check for
+ * correctness */
+
+ for (i = 0; i < ci->ncount; i++) {
+ ci->crc = crc32(ci->crc, (const Bytef *)ci->vec[i].iov_base,
+ ci->vec[i].iov_len);
+ }
+
+ /* validate inflated data */
+ ret = cdc_validate_inflate(ci, computed_crc, computed_len);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Checksum or length mismatched in inflated data");
+ }
+
+out:
+ return ret;
}
int32_t
-cdc_decompress (xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci,
- dict_t *xdata)
+cdc_decompress(xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci, dict_t *xdata)
{
- int32_t ret = -1;
-
- /* check for deflate content */
- if (!cdc_check_content_for_deflate (xdata)) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Content not deflated, passing through ...");
- goto passthrough_out;
- }
-
- ci->iobref = iobref_new ();
- if (!ci->iobref)
- goto passthrough_out;
-
- /* do we need to do this? can we assume that one iovec
- * will hold per request data every time?
- *
- * server/client protocol seems to deal with a single
- * iovec even if op_ret > 1M. So, it looks ok to
- * assume that a single iovec will contain all the
- * data (This saves us a lot from finding the trailer
- * and the data since it could have been split-up onto
- * two adjacent iovec's.
- *
- * But, in case this translator is loaded above quick-read
- * for some reason, then it's entirely possible that we get
- * multiple iovec's...
- *
- * This case (handled below) is not tested. (by loading the
- * xlator below quick-read)
- */
-
- /* @@ I_HOPE_THIS_IS_NEVER_HIT */
- if (ci->count > 1) {
- gf_log (this->name, GF_LOG_WARNING, "unable to handle"
- " multiple iovecs (%d in number)", ci->count);
- goto inflate_cleanup_out;
- /* TODO: coallate all iovecs in one */
- }
-
- ret = do_cdc_decompress (this, priv, ci);
- if (ret)
- goto inflate_cleanup_out;
-
- ci->nbytes = ci->stream.total_out;
-
- gf_log (this->name, GF_LOG_DEBUG,
- "Inflated %ld to %ld bytes",
- ci->stream.total_in, ci->stream.total_out);
-
- inflate_cleanup_out:
- (void) inflateEnd (&ci->stream);
-
- passthrough_out:
- return ret;
+ int32_t ret = -1;
+
+ /* check for deflate content */
+ if (!cdc_check_content_for_deflate(xdata)) {
+ gf_log(this->name, GF_LOG_DEBUG,
+ "Content not deflated, passing through ...");
+ goto passthrough_out;
+ }
+
+ ci->iobref = iobref_new();
+ if (!ci->iobref)
+ goto passthrough_out;
+
+ /* do we need to do this? can we assume that one iovec
+ * will hold per request data every time?
+ *
+ * server/client protocol seems to deal with a single
+ * iovec even if op_ret > 1M. So, it looks ok to
+ * assume that a single iovec will contain all the
+ * data (This saves us a lot from finding the trailer
+ * and the data since it could have been split-up onto
+ * two adjacent iovec's.
+ *
+ * But, in case this translator is loaded above quick-read
+ * for some reason, then it's entirely possible that we get
+ * multiple iovec's...
+ *
+ * This case (handled below) is not tested. (by loading the
+ * xlator below quick-read)
+ */
+
+ /* @@ I_HOPE_THIS_IS_NEVER_HIT */
+ if (ci->count > 1) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "unable to handle"
+ " multiple iovecs (%d in number)",
+ ci->count);
+ goto inflate_cleanup_out;
+ /* TODO: coallate all iovecs in one */
+ }
+
+ ret = do_cdc_decompress(this, priv, ci);
+ if (ret)
+ goto inflate_cleanup_out;
+
+ ci->nbytes = ci->stream.total_out;
+
+ gf_log(this->name, GF_LOG_DEBUG, "Inflated %ld to %ld bytes",
+ ci->stream.total_in, ci->stream.total_out);
+
+inflate_cleanup_out:
+ (void)inflateEnd(&ci->stream);
+
+passthrough_out:
+ return ret;
}
#endif
diff --git a/xlators/features/compress/src/cdc-mem-types.h b/xlators/features/compress/src/cdc-mem-types.h
index ead2c70ba6e..928afdd2efe 100644
--- a/xlators/features/compress/src/cdc-mem-types.h
+++ b/xlators/features/compress/src/cdc-mem-types.h
@@ -11,13 +11,13 @@
#ifndef __CDC_MEM_TYPES_H
#define __CDC_MEM_TYPES_H
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_cdc_mem_types {
- gf_cdc_mt_priv_t = gf_common_mt_end + 1,
- gf_cdc_mt_vec_t = gf_common_mt_end + 2,
- gf_cdc_mt_gzip_trailer_t = gf_common_mt_end + 3,
- gf_cdc_mt_end = gf_common_mt_end + 4,
+ gf_cdc_mt_priv_t = gf_common_mt_end + 1,
+ gf_cdc_mt_vec_t = gf_common_mt_end + 2,
+ gf_cdc_mt_gzip_trailer_t = gf_common_mt_end + 3,
+ gf_cdc_mt_end = gf_common_mt_end + 4,
};
#endif
diff --git a/xlators/features/compress/src/cdc.c b/xlators/features/compress/src/cdc.c
index e33d4efc1a1..b0b51e914ed 100644
--- a/xlators/features/compress/src/cdc.c
+++ b/xlators/features/compress/src/cdc.c
@@ -10,347 +10,339 @@
#include <sys/uio.h>
-#include "xlator.h"
-#include "defaults.h"
-#include "logging.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
+#include <glusterfs/logging.h>
#include "cdc.h"
#include "cdc-mem-types.h"
static void
-cdc_cleanup_iobref (cdc_info_t *ci)
+cdc_cleanup_iobref(cdc_info_t *ci)
{
- assert(ci->iobref != NULL);
- iobref_clear (ci->iobref);
+ assert(ci->iobref != NULL);
+ iobref_clear(ci->iobref);
}
int32_t
-cdc_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iovec *vector, int32_t count,
- struct iatt *stbuf, struct iobref *iobref,
- dict_t *xdata)
+cdc_readv_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iovec *vector, int32_t count,
+ struct iatt *stbuf, struct iobref *iobref, dict_t *xdata)
{
- int ret = -1;
- cdc_priv_t *priv = NULL;
- cdc_info_t ci = {0,};
-
- GF_VALIDATE_OR_GOTO ("cdc", this, default_out);
- GF_VALIDATE_OR_GOTO (this->name, frame, default_out);
-
- priv = this->private;
-
- if (op_ret <= 0)
- goto default_out;
-
- if ( (priv->min_file_size != 0)
- && (op_ret < priv->min_file_size) )
- goto default_out;
-
- ci.count = count;
- ci.ibytes = op_ret;
- ci.vector = vector;
- ci.buf = NULL;
- ci.iobref = NULL;
- ci.ncount = 0;
- ci.crc = 0;
- ci.buffer_size = GF_CDC_DEF_BUFFERSIZE;
-
-/* A readv compresses on the server side and decompresses on the client side
- */
- if (priv->op_mode == GF_CDC_MODE_SERVER) {
- ret = cdc_compress (this, priv, &ci, &xdata);
- } else if (priv->op_mode == GF_CDC_MODE_CLIENT) {
- ret = cdc_decompress (this, priv, &ci, xdata);
- } else {
- gf_log (this->name, GF_LOG_ERROR,
- "Invalid operation mode (%d)", priv->op_mode);
- }
-
- if (ret)
- goto default_out;
-
- STACK_UNWIND_STRICT (readv, frame, ci.nbytes, op_errno,
- ci.vec, ci.ncount, stbuf, iobref,
- xdata);
- cdc_cleanup_iobref (&ci);
- return 0;
-
- default_out:
- STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno,
- vector, count, stbuf, iobref, xdata);
- return 0;
+ int ret = -1;
+ cdc_priv_t *priv = NULL;
+ cdc_info_t ci = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("cdc", this, default_out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, default_out);
+
+ priv = this->private;
+
+ if (op_ret <= 0)
+ goto default_out;
+
+ if ((priv->min_file_size != 0) && (op_ret < priv->min_file_size))
+ goto default_out;
+
+ ci.count = count;
+ ci.ibytes = op_ret;
+ ci.vector = vector;
+ ci.buf = NULL;
+ ci.iobref = NULL;
+ ci.ncount = 0;
+ ci.crc = 0;
+ ci.buffer_size = GF_CDC_DEF_BUFFERSIZE;
+
+ /* A readv compresses on the server side and decompresses on the client side
+ */
+ if (priv->op_mode == GF_CDC_MODE_SERVER) {
+ ret = cdc_compress(this, priv, &ci, &xdata);
+ } else if (priv->op_mode == GF_CDC_MODE_CLIENT) {
+ ret = cdc_decompress(this, priv, &ci, xdata);
+ } else {
+ gf_log(this->name, GF_LOG_ERROR, "Invalid operation mode (%d)",
+ priv->op_mode);
+ }
+
+ if (ret)
+ goto default_out;
+
+ STACK_UNWIND_STRICT(readv, frame, ci.nbytes, op_errno, ci.vec, ci.ncount,
+ stbuf, iobref, xdata);
+ cdc_cleanup_iobref(&ci);
+ return 0;
+
+default_out:
+ STACK_UNWIND_STRICT(readv, frame, op_ret, op_errno, vector, count, stbuf,
+ iobref, xdata);
+ return 0;
}
int32_t
-cdc_readv (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset, uint32_t flags,
- dict_t *xdata)
+cdc_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- fop_readv_cbk_t cbk = NULL;
+ fop_readv_cbk_t cbk = NULL;
#ifdef HAVE_LIB_Z
- cbk = cdc_readv_cbk;
+ cbk = cdc_readv_cbk;
#else
- cbk = default_readv_cbk;
+ cbk = default_readv_cbk;
#endif
- STACK_WIND (frame, cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readv,
- fd, size, offset, flags, xdata);
- return 0;
+ STACK_WIND(frame, cbk, FIRST_CHILD(this), FIRST_CHILD(this)->fops->readv,
+ fd, size, offset, flags, xdata);
+ return 0;
}
int32_t
-cdc_writev_cbk (call_frame_t *frame,
- void *cookie,
- xlator_t *this,
- int32_t op_ret,
- int32_t op_errno,
- struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+cdc_writev_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
-
- STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf, xdata);
- return 0;
+ STACK_UNWIND_STRICT(writev, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
int32_t
-cdc_writev (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- struct iovec *vector,
- int32_t count,
- off_t offset,
- uint32_t flags,
- struct iobref *iobref, dict_t *xdata)
+cdc_writev(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
+ int32_t count, off_t offset, uint32_t flags, struct iobref *iobref,
+ dict_t *xdata)
{
- int ret = -1;
- cdc_priv_t *priv = NULL;
- cdc_info_t ci = {0,};
- size_t isize = 0;
-
- GF_VALIDATE_OR_GOTO ("cdc", this, default_out);
- GF_VALIDATE_OR_GOTO (this->name, frame, default_out);
-
- priv = this->private;
-
- isize = iov_length(vector, count);
-
- if (isize <= 0)
- goto default_out;
-
- if ( (priv->min_file_size != 0) && (isize < priv->min_file_size) )
- goto default_out;
-
- ci.count = count;
- ci.ibytes = isize;
- ci.vector = vector;
- ci.buf = NULL;
- ci.iobref = NULL;
- ci.ncount = 0;
- ci.crc = 0;
- ci.buffer_size = GF_CDC_DEF_BUFFERSIZE;
-
-/* A writev compresses on the client side and decompresses on the server side
- */
- if (priv->op_mode == GF_CDC_MODE_CLIENT) {
- ret = cdc_compress (this, priv, &ci, &xdata);
- } else if (priv->op_mode == GF_CDC_MODE_SERVER) {
- ret = cdc_decompress (this, priv, &ci, xdata);
- } else {
- gf_log (this->name, GF_LOG_ERROR, "Invalid operation mode (%d) ", priv->op_mode);
- }
-
- if (ret)
- goto default_out;
-
- STACK_WIND (frame,
- cdc_writev_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->writev,
- fd, ci.vec, ci.ncount, offset, flags,
- iobref, xdata);
-
- cdc_cleanup_iobref (&ci);
- return 0;
-
- default_out:
- STACK_WIND (frame,
- cdc_writev_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->writev,
- fd, vector, count, offset, flags,
- iobref, xdata);
- return 0;
+ int ret = -1;
+ cdc_priv_t *priv = NULL;
+ cdc_info_t ci = {
+ 0,
+ };
+ size_t isize = 0;
+
+ GF_VALIDATE_OR_GOTO("cdc", this, err);
+ GF_VALIDATE_OR_GOTO(this->name, frame, err);
+
+ priv = this->private;
+
+ isize = iov_length(vector, count);
+
+ if (isize <= 0)
+ goto default_out;
+
+ if ((priv->min_file_size != 0) && (isize < priv->min_file_size))
+ goto default_out;
+
+ ci.count = count;
+ ci.ibytes = isize;
+ ci.vector = vector;
+ ci.buf = NULL;
+ ci.iobref = NULL;
+ ci.ncount = 0;
+ ci.crc = 0;
+ ci.buffer_size = GF_CDC_DEF_BUFFERSIZE;
+
+ /* A writev compresses on the client side and decompresses on the server
+ * side
+ */
+ if (priv->op_mode == GF_CDC_MODE_CLIENT) {
+ ret = cdc_compress(this, priv, &ci, &xdata);
+ } else if (priv->op_mode == GF_CDC_MODE_SERVER) {
+ ret = cdc_decompress(this, priv, &ci, xdata);
+ } else {
+ gf_log(this->name, GF_LOG_ERROR, "Invalid operation mode (%d) ",
+ priv->op_mode);
+ }
+
+ if (ret)
+ goto default_out;
+
+ STACK_WIND(frame, cdc_writev_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, ci.vec, ci.ncount, offset,
+ flags, iobref, xdata);
+
+ cdc_cleanup_iobref(&ci);
+ return 0;
+
+default_out:
+ STACK_WIND(frame, cdc_writev_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector, count, offset,
+ flags, iobref, xdata);
+ return 0;
+err:
+ STACK_UNWIND_STRICT(writev, frame, -1, EINVAL, NULL, NULL, NULL);
+ return 0;
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- if (!this)
- return ret;
-
- ret = xlator_mem_acct_init (this, gf_cdc_mt_end);
+ if (!this)
+ return ret;
- if (ret != 0) {
- gf_log(this->name, GF_LOG_ERROR, "Memory accounting init"
- "failed");
- return ret;
- }
+ ret = xlator_mem_acct_init(this, gf_cdc_mt_end);
+ if (ret != 0) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Memory accounting init"
+ "failed");
return ret;
+ }
+
+ return ret;
}
int32_t
-init (xlator_t *this)
+init(xlator_t *this)
{
- int ret = -1;
- char *temp_str = NULL;
- cdc_priv_t *priv = NULL;
-
- GF_VALIDATE_OR_GOTO ("cdc", this, err);
-
- if (!this->children || this->children->next) {
- gf_log (this->name, GF_LOG_ERROR,
- "Need subvolume == 1");
- goto err;
- }
-
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "Dangling volume. Check volfile");
- }
-
- priv = GF_CALLOC (1, sizeof (*priv), gf_cdc_mt_priv_t);
- if (!priv) {
- goto err;
- }
-
- /* Check if debug mode is turned on */
- GF_OPTION_INIT ("debug", priv->debug, bool, err);
- if( priv->debug ) {
- gf_log (this->name, GF_LOG_DEBUG, "CDC debug option turned on");
- }
-
- /* Set Gzip Window Size */
- GF_OPTION_INIT ("window-size", priv->window_size, int32, err);
- if ( (priv->window_size > GF_CDC_MAX_WINDOWSIZE)
- || (priv->window_size < GF_CDC_DEF_WINDOWSIZE) ) {
- gf_log (this->name, GF_LOG_WARNING,
- "Invalid gzip window size (%d), using default",
- priv->window_size);
- priv->window_size = GF_CDC_DEF_WINDOWSIZE;
- }
-
- /* Set Gzip (De)Compression Level */
- GF_OPTION_INIT ("compression-level", priv->cdc_level, int32, err);
- if ( ((priv->cdc_level < 1) || (priv->cdc_level > 9))
- && (priv->cdc_level != GF_CDC_DEF_COMPRESSION) ) {
- gf_log (this->name, GF_LOG_WARNING,
- "Invalid gzip (de)compression level (%d),"
- " using default", priv->cdc_level);
- priv->cdc_level = GF_CDC_DEF_COMPRESSION;
- }
-
- /* Set Gzip Memory Level */
- GF_OPTION_INIT ("mem-level", priv->mem_level, int32, err);
- if ( (priv->mem_level < 1) || (priv->mem_level > 9) ) {
- gf_log (this->name, GF_LOG_WARNING,
- "Invalid gzip memory level, using the default");
- priv->mem_level = GF_CDC_DEF_MEMLEVEL;
- }
-
- /* Set min file size to enable compression */
- GF_OPTION_INIT ("min-size", priv->min_file_size, int32, err);
-
- /* Mode of operation - Server/Client */
- ret = dict_get_str (this->options, "mode", &temp_str);
- if (ret) {
- gf_log (this->name, GF_LOG_CRITICAL,
- "Operation mode not specified !!");
- goto err;
- }
-
- if (GF_CDC_MODE_IS_CLIENT (temp_str)) {
- priv->op_mode = GF_CDC_MODE_CLIENT;
- } else if (GF_CDC_MODE_IS_SERVER (temp_str)) {
- priv->op_mode = GF_CDC_MODE_SERVER;
- } else {
- gf_log (this->name, GF_LOG_CRITICAL,
- "Bogus operation mode (%s) specified", temp_str);
- goto err;
- }
-
- this->private = priv;
- gf_log (this->name, GF_LOG_DEBUG, "CDC xlator loaded in (%s) mode",temp_str);
- return 0;
-
- err:
- if (priv)
- GF_FREE (priv);
-
- return -1;
+ int ret = -1;
+ char *temp_str = NULL;
+ cdc_priv_t *priv = NULL;
+
+ GF_VALIDATE_OR_GOTO("cdc", this, err);
+
+ if (!this->children || this->children->next) {
+ gf_log(this->name, GF_LOG_ERROR, "Need subvolume == 1");
+ goto err;
+ }
+
+ if (!this->parents) {
+ gf_log(this->name, GF_LOG_WARNING, "Dangling volume. Check volfile");
+ }
+
+ priv = GF_CALLOC(1, sizeof(*priv), gf_cdc_mt_priv_t);
+ if (!priv) {
+ goto err;
+ }
+
+ /* Check if debug mode is turned on */
+ GF_OPTION_INIT("debug", priv->debug, bool, err);
+ if (priv->debug) {
+ gf_log(this->name, GF_LOG_DEBUG, "CDC debug option turned on");
+ }
+
+ /* Set Gzip Window Size */
+ GF_OPTION_INIT("window-size", priv->window_size, int32, err);
+ if ((priv->window_size > GF_CDC_MAX_WINDOWSIZE) ||
+ (priv->window_size < GF_CDC_DEF_WINDOWSIZE)) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "Invalid gzip window size (%d), using default",
+ priv->window_size);
+ priv->window_size = GF_CDC_DEF_WINDOWSIZE;
+ }
+
+ /* Set Gzip (De)Compression Level */
+ GF_OPTION_INIT("compression-level", priv->cdc_level, int32, err);
+ if (((priv->cdc_level < 1) || (priv->cdc_level > 9)) &&
+ (priv->cdc_level != GF_CDC_DEF_COMPRESSION)) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "Invalid gzip (de)compression level (%d),"
+ " using default",
+ priv->cdc_level);
+ priv->cdc_level = GF_CDC_DEF_COMPRESSION;
+ }
+
+ /* Set Gzip Memory Level */
+ GF_OPTION_INIT("mem-level", priv->mem_level, int32, err);
+ if ((priv->mem_level < 1) || (priv->mem_level > 9)) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "Invalid gzip memory level, using the default");
+ priv->mem_level = GF_CDC_DEF_MEMLEVEL;
+ }
+
+ /* Set min file size to enable compression */
+ GF_OPTION_INIT("min-size", priv->min_file_size, int32, err);
+
+ /* Mode of operation - Server/Client */
+ ret = dict_get_str(this->options, "mode", &temp_str);
+ if (ret) {
+ gf_log(this->name, GF_LOG_CRITICAL, "Operation mode not specified !!");
+ goto err;
+ }
+
+ if (GF_CDC_MODE_IS_CLIENT(temp_str)) {
+ priv->op_mode = GF_CDC_MODE_CLIENT;
+ } else if (GF_CDC_MODE_IS_SERVER(temp_str)) {
+ priv->op_mode = GF_CDC_MODE_SERVER;
+ } else {
+ gf_log(this->name, GF_LOG_CRITICAL,
+ "Bogus operation mode (%s) specified", temp_str);
+ goto err;
+ }
+
+ this->private = priv;
+ gf_log(this->name, GF_LOG_DEBUG, "CDC xlator loaded in (%s) mode",
+ temp_str);
+ return 0;
+
+err:
+ if (priv)
+ GF_FREE(priv);
+
+ return -1;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- cdc_priv_t *priv = this->private;
+ cdc_priv_t *priv = this->private;
- if (priv)
- GF_FREE (priv);
- this->private = NULL;
- return;
+ if (priv)
+ GF_FREE(priv);
+ this->private = NULL;
+ return;
}
struct xlator_fops fops = {
- .readv = cdc_readv,
- .writev = cdc_writev,
+ .readv = cdc_readv,
+ .writev = cdc_writev,
};
-struct xlator_cbks cbks = {
-};
+struct xlator_cbks cbks = {};
struct volume_options options[] = {
- { .key = {"window-size"},
- .default_value = "-15",
- .type = GF_OPTION_TYPE_INT,
- .description = "Size of the zlib history buffer."
- },
- { .key = {"mem-level"},
- .default_value = "8",
- .type = GF_OPTION_TYPE_INT,
- .description = "Memory allocated for internal compression state. "
- "1 uses minimum memory but is slow and reduces "
- "compression ratio; memLevel=9 uses maximum memory "
- "for optimal speed. The default value is 8."
- },
- { .key = {"compression-level"},
- .default_value = "-1",
- .type = GF_OPTION_TYPE_INT,
- .description = "Compression levels \n"
- "0 : no compression, 1 : best speed, \n"
- "9 : best compression, -1 : default compression "
- },
- { .key = {"min-size"},
- .default_value = "0",
- .type = GF_OPTION_TYPE_INT,
- .description = "Data is compressed only when its size exceeds this."
- },
- { .key = {"mode"},
- .value = {"server", "client"},
- .type = GF_OPTION_TYPE_STR,
- .description = "Set on the basis of where the xlator is loaded. "
- "This option should NOT be configured by user."
- },
- { .key = {"debug"},
- .default_value = "false",
- .type = GF_OPTION_TYPE_BOOL,
- .description = "This is used in testing. Will dump compressed data "
- "to disk as a gzip file."
- },
- { .key = {NULL}
- },
+ {.key = {"window-size"},
+ .default_value = "-15",
+ .type = GF_OPTION_TYPE_INT,
+ .description = "Size of the zlib history buffer."},
+ {.key = {"mem-level"},
+ .default_value = "8",
+ .type = GF_OPTION_TYPE_INT,
+ .description = "Memory allocated for internal compression state. "
+ "1 uses minimum memory but is slow and reduces "
+ "compression ratio; memLevel=9 uses maximum memory "
+ "for optimal speed. The default value is 8."},
+ {.key = {"compression-level"},
+ .default_value = "-1",
+ .type = GF_OPTION_TYPE_INT,
+ .description = "Compression levels \n"
+ "0 : no compression, 1 : best speed, \n"
+ "9 : best compression, -1 : default compression "},
+ {.key = {"min-size"},
+ .default_value = "0",
+ .type = GF_OPTION_TYPE_INT,
+ .description = "Data is compressed only when its size exceeds this."},
+ {.key = {"mode"},
+ .value = {"server", "client"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Set on the basis of where the xlator is loaded. "
+ "This option should NOT be configured by user."},
+ {.key = {"debug"},
+ .default_value = "false",
+ .type = GF_OPTION_TYPE_BOOL,
+ .description = "This is used in testing. Will dump compressed data "
+ "to disk as a gzip file."},
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {GD_OP_VERSION_3_9_0},
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "cdc",
+ .category = GF_TECH_PREVIEW,
};
diff --git a/xlators/features/compress/src/cdc.h b/xlators/features/compress/src/cdc.h
index 71f4d2317bb..cb87b06a989 100644
--- a/xlators/features/compress/src/cdc.h
+++ b/xlators/features/compress/src/cdc.h
@@ -15,41 +15,41 @@
#include "zlib.h"
#endif
-#include "xlator.h"
+#include <glusterfs/xlator.h>
#ifndef MAX_IOVEC
#define MAX_IOVEC 16
#endif
typedef struct cdc_priv {
- int window_size;
- int mem_level;
- int cdc_level;
- int min_file_size;
- int op_mode;
- gf_boolean_t debug;
- gf_lock_t lock;
+ int window_size;
+ int mem_level;
+ int cdc_level;
+ int min_file_size;
+ int op_mode;
+ gf_boolean_t debug;
+ gf_lock_t lock;
} cdc_priv_t;
typedef struct cdc_info {
- /* input bits */
- int count;
- int32_t ibytes;
- struct iovec *vector;
- struct iatt *buf;
-
- /* output bits */
- int ncount;
- int nbytes;
- int buffer_size;
- struct iovec vec[MAX_IOVEC];
- struct iobref *iobref;
-
- /* zlib bits */
+ /* input bits */
+ int count;
+ int32_t ibytes;
+ struct iovec *vector;
+ struct iatt *buf;
+
+ /* output bits */
+ int ncount;
+ int nbytes;
+ int buffer_size;
+ struct iovec vec[MAX_IOVEC];
+ struct iobref *iobref;
+
+ /* zlib bits */
#ifdef HAVE_LIB_Z
- z_stream stream;
+ z_stream stream;
#endif
- unsigned long crc;
+ unsigned long crc;
} cdc_info_t;
#define NVEC(ci) (ci->ncount - 1)
@@ -57,8 +57,8 @@ typedef struct cdc_info {
#define THIS_VEC(ci, i) ci->vector[i]
/* Gzip defaults */
-#define GF_CDC_DEF_WINDOWSIZE -15 /* default value */
-#define GF_CDC_MAX_WINDOWSIZE -8 /* max value */
+#define GF_CDC_DEF_WINDOWSIZE -15 /* default value */
+#define GF_CDC_MAX_WINDOWSIZE -8 /* max value */
#ifdef HAVE_LIB_Z
#define GF_CDC_DEF_COMPRESSION Z_DEFAULT_COMPRESSION
@@ -66,15 +66,15 @@ typedef struct cdc_info {
#define GF_CDC_DEF_COMPRESSION -1
#endif
-#define GF_CDC_DEF_MEMLEVEL 8
-#define GF_CDC_DEF_BUFFERSIZE 262144 // 256K - default compression buffer size
+#define GF_CDC_DEF_MEMLEVEL 8
+#define GF_CDC_DEF_BUFFERSIZE 262144 // 256K - default compression buffer size
/* Operation mode
* If xlator is loaded on client, readv decompresses and writev compresses
* If xlator is loaded on server, readv compresses and writev decompresses
*/
-#define GF_CDC_MODE_CLIENT 0
-#define GF_CDC_MODE_SERVER 1
+#define GF_CDC_MODE_CLIENT 0
+#define GF_CDC_MODE_SERVER 1
/* min size of data to do cmpression
* 0 == compress even 1byte
@@ -87,21 +87,13 @@ typedef struct cdc_info {
#define GF_CDC_DEFLATE_CANARY_VAL "deflate"
#define GF_CDC_DEBUG_DUMP_FILE "/tmp/cdcdump.gz"
-#define GF_CDC_MODE_IS_CLIENT(m) \
- (strcmp (m, "client") == 0)
+#define GF_CDC_MODE_IS_CLIENT(m) (strcmp(m, "client") == 0)
-#define GF_CDC_MODE_IS_SERVER(m) \
- (strcmp (m, "server") == 0)
+#define GF_CDC_MODE_IS_SERVER(m) (strcmp(m, "server") == 0)
int32_t
-cdc_compress (xlator_t *this,
- cdc_priv_t *priv,
- cdc_info_t *ci,
- dict_t **xdata);
+cdc_compress(xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci, dict_t **xdata);
int32_t
-cdc_decompress (xlator_t *this,
- cdc_priv_t *priv,
- cdc_info_t *ci,
- dict_t *xdata);
+cdc_decompress(xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci, dict_t *xdata);
#endif
diff --git a/xlators/features/gfid-access/src/gfid-access-mem-types.h b/xlators/features/gfid-access/src/gfid-access-mem-types.h
index 168d67b431f..1c4d0b93de2 100644
--- a/xlators/features/gfid-access/src/gfid-access-mem-types.h
+++ b/xlators/features/gfid-access/src/gfid-access-mem-types.h
@@ -11,13 +11,12 @@
#ifndef _GFID_ACCESS_MEM_TYPES_H
#define _GFID_ACCESS_MEM_TYPES_H
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_changelog_mem_types {
- gf_gfid_access_mt_priv_t = gf_common_mt_end + 1,
- gf_gfid_access_mt_gfid_t,
- gf_gfid_access_mt_end
+ gf_gfid_access_mt_priv_t = gf_common_mt_end + 1,
+ gf_gfid_access_mt_gfid_t,
+ gf_gfid_access_mt_end
};
#endif
-
diff --git a/xlators/features/gfid-access/src/gfid-access.c b/xlators/features/gfid-access/src/gfid-access.c
index 49d781c59ba..3fea5672a21 100644
--- a/xlators/features/gfid-access/src/gfid-access.c
+++ b/xlators/features/gfid-access/src/gfid-access.c
@@ -8,1421 +8,1413 @@
cases as published by the Free Software Foundation.
*/
#include "gfid-access.h"
-#include "inode.h"
-#include "byte-order.h"
-#include "statedump.h"
-
+#include <glusterfs/inode.h>
+#include <glusterfs/byte-order.h>
+#include <glusterfs/statedump.h>
int
-ga_valid_inode_loc_copy (loc_t *dst, loc_t *src, xlator_t *this)
+ga_valid_inode_loc_copy(loc_t *dst, loc_t *src, xlator_t *this)
{
- int ret = 0;
- uint64_t value = 0;
-
- /* if its an entry operation, on the virtual */
- /* directory inode as parent, we need to handle */
- /* it properly */
- ret = loc_copy (dst, src);
- if (ret < 0)
- goto out;
-
- /*
- * Change ALL virtual inodes with real-inodes in loc
- */
- if (dst->parent) {
- ret = inode_ctx_get (dst->parent, this, &value);
- if (ret < 0) {
- ret = 0; //real-inode
- goto out;
- }
- inode_unref (dst->parent);
- dst->parent = inode_ref ((inode_t*)value);
- gf_uuid_copy (dst->pargfid, dst->parent->gfid);
+ int ret = 0;
+ uint64_t value = 0;
+
+ /* if its an entry operation, on the virtual */
+ /* directory inode as parent, we need to handle */
+ /* it properly */
+ ret = loc_copy(dst, src);
+ if (ret < 0)
+ goto out;
+
+ /*
+ * Change ALL virtual inodes with real-inodes in loc
+ */
+ if (dst->parent) {
+ ret = inode_ctx_get(dst->parent, this, &value);
+ if (ret < 0) {
+ ret = 0; // real-inode
+ goto out;
}
+ inode_unref(dst->parent);
+ dst->parent = inode_ref((inode_t *)(uintptr_t)value);
+ gf_uuid_copy(dst->pargfid, dst->parent->gfid);
+ }
- if (dst->inode) {
- ret = inode_ctx_get (dst->inode, this, &value);
- if (ret < 0) {
- ret = 0; //real-inode
- goto out;
- }
- inode_unref (dst->inode);
- dst->inode = inode_ref ((inode_t*)value);
- gf_uuid_copy (dst->gfid, dst->inode->gfid);
+ if (dst->inode) {
+ ret = inode_ctx_get(dst->inode, this, &value);
+ if (ret < 0) {
+ ret = 0; // real-inode
+ goto out;
}
+ inode_unref(dst->inode);
+ dst->inode = inode_ref((inode_t *)(uintptr_t)value);
+ gf_uuid_copy(dst->gfid, dst->inode->gfid);
+ }
out:
- return ret;
+ return ret;
}
void
-ga_newfile_args_free (ga_newfile_args_t *args)
+ga_newfile_args_free(ga_newfile_args_t *args)
{
- if (!args)
- goto out;
+ if (!args)
+ goto out;
- GF_FREE (args->bname);
+ GF_FREE(args->bname);
- if (S_ISLNK (args->st_mode) && args->args.symlink.linkpath) {
- GF_FREE (args->args.symlink.linkpath);
- args->args.symlink.linkpath = NULL;
- }
+ if (S_ISLNK(args->st_mode) && args->args.symlink.linkpath) {
+ GF_FREE(args->args.symlink.linkpath);
+ args->args.symlink.linkpath = NULL;
+ }
- mem_put (args);
+ mem_put(args);
out:
- return;
+ return;
}
-
void
-ga_heal_args_free (ga_heal_args_t *args)
+ga_heal_args_free(ga_heal_args_t *args)
{
- if (!args)
- goto out;
+ if (!args)
+ goto out;
- GF_FREE (args->bname);
+ GF_FREE(args->bname);
- mem_put (args);
+ mem_put(args);
out:
- return;
+ return;
}
-
ga_newfile_args_t *
-ga_newfile_parse_args (xlator_t *this, data_t *data)
+ga_newfile_parse_args(xlator_t *this, data_t *data)
{
- ga_newfile_args_t *args = NULL;
- ga_private_t *priv = NULL;
- int len = 0;
- int blob_len = 0;
- int min_len = 0;
- void *blob = NULL;
-
- priv = this->private;
-
- blob = data->data;
- blob_len = data->len;
-
- min_len = sizeof (args->uid) + sizeof (args->gid) + sizeof (args->gfid)
- + sizeof (args->st_mode) + 2 + 2;
- if (blob_len < min_len) {
- gf_log (this->name, GF_LOG_ERROR,
- "Invalid length: Total length is less "
- "than minimum length.");
- goto err;
+ ga_newfile_args_t *args = NULL;
+ ga_private_t *priv = NULL;
+ int len = 0;
+ int blob_len = 0;
+ int min_len = 0;
+ void *blob = NULL;
+
+ priv = this->private;
+
+ blob = data->data;
+ blob_len = data->len;
+
+ min_len = sizeof(args->uid) + sizeof(args->gid) + sizeof(args->gfid) +
+ sizeof(args->st_mode) + 2 + 2;
+ if (blob_len < min_len) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Invalid length: Total length is less "
+ "than minimum length.");
+ goto err;
+ }
+
+ args = mem_get0(priv->newfile_args_pool);
+ if (args == NULL)
+ goto err;
+
+ args->uid = ntoh32(*(uint32_t *)blob);
+ blob += sizeof(uint32_t);
+ blob_len -= sizeof(uint32_t);
+
+ args->gid = ntoh32(*(uint32_t *)blob);
+ blob += sizeof(uint32_t);
+ blob_len -= sizeof(uint32_t);
+
+ memcpy(args->gfid, blob, sizeof(args->gfid));
+ blob += sizeof(args->gfid);
+ blob_len -= sizeof(args->gfid);
+
+ args->st_mode = ntoh32(*(uint32_t *)blob);
+ blob += sizeof(uint32_t);
+ blob_len -= sizeof(uint32_t);
+
+ len = strnlen(blob, blob_len);
+ if (len == blob_len) {
+ gf_log(this->name, GF_LOG_ERROR, "gfid: %s. No null byte present.",
+ args->gfid);
+ goto err;
+ }
+
+ args->bname = GF_MALLOC(len + 1, gf_common_mt_char);
+ if (args->bname == NULL)
+ goto err;
+
+ memcpy(args->bname, blob, (len + 1));
+ blob += (len + 1);
+ blob_len -= (len + 1);
+
+ if (S_ISDIR(args->st_mode)) {
+ if (blob_len < sizeof(uint32_t)) {
+ gf_log(this->name, GF_LOG_ERROR, "gfid: %s. Invalid length",
+ args->gfid);
+ goto err;
}
-
- args = mem_get0 (priv->newfile_args_pool);
- if (args == NULL)
- goto err;
-
- args->uid = ntoh32 (*(uint32_t *)blob);
- blob += sizeof (uint32_t);
- blob_len -= sizeof (uint32_t);
-
- args->gid = ntoh32 (*(uint32_t *)blob);
- blob += sizeof (uint32_t);
- blob_len -= sizeof (uint32_t);
-
- memcpy (args->gfid, blob, sizeof (args->gfid));
- blob += sizeof (args->gfid);
- blob_len -= sizeof (args->gfid);
-
- args->st_mode = ntoh32 (*(uint32_t *)blob);
- blob += sizeof (uint32_t);
- blob_len -= sizeof (uint32_t);
-
- len = strnlen (blob, blob_len);
+ args->args.mkdir.mode = ntoh32(*(uint32_t *)blob);
+ blob += sizeof(uint32_t);
+ blob_len -= sizeof(uint32_t);
+
+ if (blob_len < sizeof(uint32_t)) {
+ gf_log(this->name, GF_LOG_ERROR, "gfid: %s. Invalid length",
+ args->gfid);
+ goto err;
+ }
+ args->args.mkdir.umask = ntoh32(*(uint32_t *)blob);
+ blob_len -= sizeof(uint32_t);
+ if (blob_len < 0) {
+ gf_log(this->name, GF_LOG_ERROR, "gfid: %s. Invalid length",
+ args->gfid);
+ goto err;
+ }
+ } else if (S_ISLNK(args->st_mode)) {
+ len = strnlen(blob, blob_len);
if (len == blob_len) {
- gf_log (this->name, GF_LOG_ERROR,
- "gfid: %s. No null byte present.",
- args->gfid);
- goto err;
+ gf_log(this->name, GF_LOG_ERROR, "gfid: %s. Invalid length",
+ args->gfid);
+ goto err;
}
+ args->args.symlink.linkpath = GF_MALLOC(len + 1, gf_common_mt_char);
+ if (args->args.symlink.linkpath == NULL)
+ goto err;
- args->bname = GF_CALLOC (1, (len + 1), gf_common_mt_char);
- if (args->bname == NULL)
- goto err;
-
- memcpy (args->bname, blob, (len + 1));
- blob += (len + 1);
+ memcpy(args->args.symlink.linkpath, blob, (len + 1));
blob_len -= (len + 1);
-
- if (S_ISDIR (args->st_mode)) {
- if (blob_len < sizeof (uint32_t)) {
- gf_log (this->name, GF_LOG_ERROR,
- "gfid: %s. Invalid length",
- args->gfid);
- goto err;
- }
- args->args.mkdir.mode = ntoh32 (*(uint32_t *)blob);
- blob += sizeof (uint32_t);
- blob_len -= sizeof (uint32_t);
-
- if (blob_len < sizeof (uint32_t)) {
- gf_log (this->name, GF_LOG_ERROR,
- "gfid: %s. Invalid length",
- args->gfid);
- goto err;
- }
- args->args.mkdir.umask = ntoh32 (*(uint32_t *)blob);
- blob_len -= sizeof (uint32_t);
- if (blob_len < 0) {
- gf_log (this->name, GF_LOG_ERROR,
- "gfid: %s. Invalid length",
- args->gfid);
- goto err;
- }
- } else if (S_ISLNK (args->st_mode)) {
- len = strnlen (blob, blob_len);
- if (len == blob_len) {
- gf_log (this->name, GF_LOG_ERROR,
- "gfid: %s. Invalid length",
- args->gfid);
- goto err;
- }
- args->args.symlink.linkpath = GF_CALLOC (1, len + 1,
- gf_common_mt_char);
- if (args->args.symlink.linkpath == NULL)
- goto err;
-
- memcpy (args->args.symlink.linkpath, blob, (len + 1));
- blob_len -= (len + 1);
- } else {
- if (blob_len < sizeof (uint32_t)) {
- gf_log (this->name, GF_LOG_ERROR,
- "gfid: %s. Invalid length",
- args->gfid);
- goto err;
- }
- args->args.mknod.mode = ntoh32 (*(uint32_t *)blob);
- blob += sizeof (uint32_t);
- blob_len -= sizeof (uint32_t);
-
- if (blob_len < sizeof (uint32_t)) {
- gf_log (this->name, GF_LOG_ERROR,
- "gfid: %s. Invalid length",
- args->gfid);
- goto err;
- }
- args->args.mknod.rdev = ntoh32 (*(uint32_t *)blob);
- blob += sizeof (uint32_t);
- blob_len -= sizeof (uint32_t);
-
- if (blob_len < sizeof (uint32_t)) {
- gf_log (this->name, GF_LOG_ERROR,
- "gfid: %s. Invalid length",
- args->gfid);
- goto err;
- }
- args->args.mknod.umask = ntoh32 (*(uint32_t *)blob);
- blob_len -= sizeof (uint32_t);
+ } else {
+ if (blob_len < sizeof(uint32_t)) {
+ gf_log(this->name, GF_LOG_ERROR, "gfid: %s. Invalid length",
+ args->gfid);
+ goto err;
}
-
- if (blob_len) {
- gf_log (this->name, GF_LOG_ERROR,
- "gfid: %s. Invalid length",
- args->gfid);
- goto err;
+ args->args.mknod.mode = ntoh32(*(uint32_t *)blob);
+ blob += sizeof(uint32_t);
+ blob_len -= sizeof(uint32_t);
+
+ if (blob_len < sizeof(uint32_t)) {
+ gf_log(this->name, GF_LOG_ERROR, "gfid: %s. Invalid length",
+ args->gfid);
+ goto err;
}
+ args->args.mknod.rdev = ntoh32(*(uint32_t *)blob);
+ blob += sizeof(uint32_t);
+ blob_len -= sizeof(uint32_t);
+
+ if (blob_len < sizeof(uint32_t)) {
+ gf_log(this->name, GF_LOG_ERROR, "gfid: %s. Invalid length",
+ args->gfid);
+ goto err;
+ }
+ args->args.mknod.umask = ntoh32(*(uint32_t *)blob);
+ blob_len -= sizeof(uint32_t);
+ }
+
+ if (blob_len) {
+ gf_log(this->name, GF_LOG_ERROR, "gfid: %s. Invalid length",
+ args->gfid);
+ goto err;
+ }
- return args;
+ return args;
err:
- if (args)
- ga_newfile_args_free (args);
+ if (args)
+ ga_newfile_args_free(args);
- return NULL;
+ return NULL;
}
ga_heal_args_t *
-ga_heal_parse_args (xlator_t *this, data_t *data)
+ga_heal_parse_args(xlator_t *this, data_t *data)
{
- ga_heal_args_t *args = NULL;
- ga_private_t *priv = NULL;
- void *blob = NULL;
- int len = 0;
- int blob_len = 0;
+ ga_heal_args_t *args = NULL;
+ ga_private_t *priv = NULL;
+ void *blob = NULL;
+ int len = 0;
+ int blob_len = 0;
- blob = data->data;
- blob_len = data->len;
+ blob = data->data;
+ blob_len = data->len;
- priv = this->private;
+ priv = this->private;
- /* bname should at least contain a character */
- if (blob_len < (sizeof (args->gfid) + 2))
- goto err;
+ /* bname should at least contain a character */
+ if (blob_len < (sizeof(args->gfid) + 2))
+ goto err;
- args = mem_get0 (priv->heal_args_pool);
- if (!args)
- goto err;
+ args = mem_get0(priv->heal_args_pool);
+ if (!args)
+ goto err;
- memcpy (args->gfid, blob, sizeof (args->gfid));
- blob += sizeof (args->gfid);
- blob_len -= sizeof (args->gfid);
+ memcpy(args->gfid, blob, sizeof(args->gfid));
+ blob += sizeof(args->gfid);
+ blob_len -= sizeof(args->gfid);
- len = strnlen (blob, blob_len);
- if (len == blob_len)
- goto err;
+ len = strnlen(blob, blob_len);
+ if (len == blob_len)
+ goto err;
- args->bname = GF_CALLOC (1, len + 1, gf_common_mt_char);
- if (!args->bname)
- goto err;
+ args->bname = GF_MALLOC(len + 1, gf_common_mt_char);
+ if (!args->bname)
+ goto err;
- memcpy (args->bname, blob, len);
- blob_len -= (len + 1);
+ memcpy(args->bname, blob, len);
+ args->bname[len] = '\0';
+ blob_len -= (len + 1);
- if (blob_len)
- goto err;
+ if (blob_len)
+ goto err;
- return args;
+ return args;
err:
- if (args)
- ga_heal_args_free (args);
+ if (args)
+ ga_heal_args_free(args);
- return NULL;
+ return NULL;
}
static int32_t
-ga_fill_tmp_loc (loc_t *loc, xlator_t *this, uuid_t gfid,
- char *bname, dict_t *xdata, loc_t *new_loc)
+ga_fill_tmp_loc(loc_t *loc, xlator_t *this, uuid_t gfid, char *bname,
+ dict_t *xdata, loc_t *new_loc)
{
- int ret = -1;
- uint64_t value = 0;
- inode_t *parent = NULL;
- unsigned char *gfid_ptr = NULL;
-
- parent = loc->inode;
- ret = inode_ctx_get (loc->inode, this, &value);
- if (!ret) {
- parent = (void *)value;
- if (gf_uuid_is_null (parent->gfid))
- parent = loc->inode;
- }
-
- /* parent itself should be looked up */
- gf_uuid_copy (new_loc->pargfid, parent->gfid);
- new_loc->parent = inode_ref (parent);
-
- new_loc->inode = inode_grep (parent->table, parent, bname);
- if (!new_loc->inode) {
- new_loc->inode = inode_new (parent->table);
- gf_uuid_copy (new_loc->inode->gfid, gfid);
- }
-
- loc_path (new_loc, bname);
- if (new_loc->path) {
- new_loc->name = strrchr (new_loc->path, '/');
- if (new_loc->name)
- new_loc->name++;
- }
-
- gfid_ptr = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t);
- if (!gfid_ptr) {
- ret = -1;
- goto out;
- }
- gf_uuid_copy (gfid_ptr, gfid);
- ret = dict_set_gfuuid (xdata, "gfid-req", gfid_ptr, false);
- if (ret < 0)
- goto out;
-
- ret = 0;
+ int ret = -1;
+ uint64_t value = 0;
+ inode_t *parent = NULL;
+ unsigned char *gfid_ptr = NULL;
+
+ parent = loc->inode;
+ ret = inode_ctx_get(loc->inode, this, &value);
+ if (!ret) {
+ parent = (void *)(uintptr_t)value;
+ if (gf_uuid_is_null(parent->gfid))
+ parent = loc->inode;
+ }
+
+ /* parent itself should be looked up */
+ gf_uuid_copy(new_loc->pargfid, parent->gfid);
+ new_loc->parent = inode_ref(parent);
+
+ new_loc->inode = inode_grep(parent->table, parent, bname);
+ if (!new_loc->inode) {
+ new_loc->inode = inode_new(parent->table);
+ gf_uuid_copy(new_loc->inode->gfid, gfid);
+ }
+
+ loc_path(new_loc, bname);
+ if (new_loc->path) {
+ new_loc->name = strrchr(new_loc->path, '/');
+ if (new_loc->name)
+ new_loc->name++;
+ }
+
+ gfid_ptr = GF_MALLOC(sizeof(uuid_t), gf_common_mt_uuid_t);
+ if (!gfid_ptr) {
+ ret = -1;
+ goto out;
+ }
+ gf_uuid_copy(gfid_ptr, gfid);
+ ret = dict_set_gfuuid(xdata, "gfid-req", gfid_ptr, false);
+ if (ret < 0)
+ goto out;
+
+ ret = 0;
out:
- if (ret && gfid_ptr)
- GF_FREE (gfid_ptr);
- return ret;
+ if (ret && gfid_ptr)
+ GF_FREE(gfid_ptr);
+ return ret;
}
-
-
static gf_boolean_t
-__is_gfid_access_dir (uuid_t gfid)
+__is_gfid_access_dir(uuid_t gfid)
{
- uuid_t aux_gfid;
-
- memset (aux_gfid, 0, 16);
- aux_gfid[15] = GF_AUX_GFID;
+ static uuid_t aux_gfid = {0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, GF_AUX_GFID};
- if (gf_uuid_compare (gfid, aux_gfid) == 0)
- return _gf_true;
+ if (gf_uuid_compare(gfid, aux_gfid) == 0)
+ return _gf_true;
- return _gf_false;
+ return _gf_false;
}
int32_t
-ga_forget (xlator_t *this, inode_t *inode)
+ga_forget(xlator_t *this, inode_t *inode)
{
- int ret = -1;
- uint64_t value = 0;
- inode_t *tmp_inode = NULL;
+ int ret = -1;
+ uint64_t value = 0;
+ inode_t *tmp_inode = NULL;
- ret = inode_ctx_del (inode, this, &value);
- if (ret)
- goto out;
+ ret = inode_ctx_del(inode, this, &value);
+ if (ret)
+ goto out;
- tmp_inode = (void *)value;
- inode_unref (tmp_inode);
+ tmp_inode = (void *)(uintptr_t)value;
+ inode_unref(tmp_inode);
out:
- return 0;
+ return 0;
}
-
static int
-ga_heal_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *stat, dict_t *dict,
- struct iatt *postparent)
+ga_heal_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode, struct iatt *stat, dict_t *dict,
+ struct iatt *postparent)
{
- call_frame_t *orig_frame = NULL;
+ call_frame_t *orig_frame = NULL;
- orig_frame = frame->local;
- frame->local = NULL;
+ orig_frame = frame->local;
+ frame->local = NULL;
- /* don't worry about inode linking and other stuff. They'll happen on
- * the next lookup.
- */
- STACK_DESTROY (frame->root);
+ /* don't worry about inode linking and other stuff. They'll happen on
+ * the next lookup.
+ */
+ STACK_DESTROY(frame->root);
- STACK_UNWIND_STRICT (setxattr, orig_frame, op_ret, op_errno, dict);
+ STACK_UNWIND_STRICT(setxattr, orig_frame, op_ret, op_errno, dict);
- return 0;
+ return 0;
}
static int
-ga_newentry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+ga_newentry_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- ga_local_t *local = NULL;
+ ga_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- /* don't worry about inode linking and other stuff. They'll happen on
- * the next lookup.
- */
- frame->local = NULL;
- STACK_DESTROY (frame->root);
+ /* don't worry about inode linking and other stuff. They'll happen on
+ * the next lookup.
+ */
+ frame->local = NULL;
+ STACK_DESTROY(frame->root);
- STACK_UNWIND_STRICT (setxattr, local->orig_frame, op_ret,
- op_errno, xdata);
+ STACK_UNWIND_STRICT(setxattr, local->orig_frame, op_ret, op_errno, xdata);
- if (local->xdata)
- dict_unref (local->xdata);
- loc_wipe (&local->loc);
- mem_put (local);
+ if (local->xdata)
+ dict_unref(local->xdata);
+ loc_wipe(&local->loc);
+ mem_put(local);
- return 0;
+ return 0;
}
static int
-ga_newentry_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *stat, dict_t *xdata,
- struct iatt *postparent)
+ga_newentry_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *stat, dict_t *xdata,
+ struct iatt *postparent)
{
- ga_local_t *local = NULL;
+ ga_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if ((op_ret < 0) && ((op_errno != ENOENT) && (op_errno != ESTALE)))
- goto err;
+ if ((op_ret < 0) && ((op_errno != ENOENT) && (op_errno != ESTALE)))
+ goto err;
- STACK_WIND (frame, ga_newentry_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mknod, &local->loc, local->mode,
- local->rdev, local->umask, local->xdata);
- return 0;
+ STACK_WIND(frame, ga_newentry_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, &local->loc, local->mode,
+ local->rdev, local->umask, local->xdata);
+ return 0;
err:
- frame->local = NULL;
- STACK_DESTROY (frame->root);
- STACK_UNWIND_STRICT (setxattr, local->orig_frame, op_ret, op_errno,
- xdata);
- if (local->xdata)
- dict_unref (local->xdata);
- loc_wipe (&local->loc);
- mem_put (local);
-
- return 0;
+ frame->local = NULL;
+ STACK_DESTROY(frame->root);
+ STACK_UNWIND_STRICT(setxattr, local->orig_frame, op_ret, op_errno, xdata);
+ if (local->xdata)
+ dict_unref(local->xdata);
+ loc_wipe(&local->loc);
+ mem_put(local);
+
+ return 0;
}
int32_t
-ga_new_entry (call_frame_t *frame, xlator_t *this, loc_t *loc, data_t *data,
- dict_t *xdata)
+ga_new_entry(call_frame_t *frame, xlator_t *this, loc_t *loc, data_t *data,
+ dict_t *xdata)
{
- int ret = -1;
- ga_newfile_args_t *args = NULL;
- loc_t tmp_loc = {0,};
- call_frame_t *new_frame = NULL;
- ga_local_t *local = NULL;
- uuid_t gfid = {0,};
-
- args = ga_newfile_parse_args (this, data);
- if (!args)
- goto out;
-
- ret = gf_uuid_parse (args->gfid, gfid);
- if (ret)
- goto out;
-
- if (!xdata) {
- xdata = dict_new ();
- } else {
- xdata = dict_ref (xdata);
- }
-
- if (!xdata) {
- ret = -1;
- goto out;
- }
-
- ret = ga_fill_tmp_loc (loc, this, gfid,
- args->bname, xdata, &tmp_loc);
- if (ret)
- goto out;
-
- new_frame = copy_frame (frame);
- if (!new_frame)
- goto out;
-
- local = mem_get0 (this->local_pool);
- local->orig_frame = frame;
-
- loc_copy (&local->loc, &tmp_loc);
-
- new_frame->local = local;
- new_frame->root->uid = args->uid;
- new_frame->root->gid = args->gid;
-
- if (S_ISDIR (args->st_mode)) {
- STACK_WIND (new_frame, ga_newentry_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->mkdir,
- &tmp_loc, args->args.mkdir.mode,
- args->args.mkdir.umask, xdata);
- } else if (S_ISLNK (args->st_mode)) {
- STACK_WIND (new_frame, ga_newentry_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->symlink,
- args->args.symlink.linkpath,
- &tmp_loc, 0, xdata);
- } else {
- /* use 07777 (4 7s) for considering the Sticky bits etc) */
- ((ga_local_t *)new_frame->local)->mode =
- (S_IFMT & args->st_mode) | (07777 & args->args.mknod.mode);
-
- ((ga_local_t *)new_frame->local)->umask =
- args->args.mknod.umask;
- ((ga_local_t *)new_frame->local)->rdev = args->args.mknod.rdev;
- ((ga_local_t *)new_frame->local)->xdata = dict_ref (xdata);
-
- /* send a named lookup, so that dht can cleanup up stale linkto
- * files etc.
- */
- STACK_WIND (new_frame, ga_newentry_lookup_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->lookup,
- &tmp_loc, NULL);
- }
+ int ret = -1;
+ ga_newfile_args_t *args = NULL;
+ loc_t tmp_loc = {
+ 0,
+ };
+ call_frame_t *new_frame = NULL;
+ ga_local_t *local = NULL;
+ uuid_t gfid = {
+ 0,
+ };
+
+ if (!xdata) {
+ xdata = dict_new();
+ } else {
+ xdata = dict_ref(xdata);
+ }
+
+ if (!xdata) {
+ ret = -1;
+ goto out;
+ }
+
+ args = ga_newfile_parse_args(this, data);
+ if (!args)
+ goto out;
+
+ ret = gf_uuid_parse(args->gfid, gfid);
+ if (ret)
+ goto out;
+
+ ret = ga_fill_tmp_loc(loc, this, gfid, args->bname, xdata, &tmp_loc);
+ if (ret)
+ goto out;
+
+ new_frame = copy_frame(frame);
+ if (!new_frame)
+ goto out;
+
+ local = mem_get0(this->local_pool);
+ local->orig_frame = frame;
+
+ loc_copy(&local->loc, &tmp_loc);
+
+ new_frame->local = local;
+ new_frame->root->uid = args->uid;
+ new_frame->root->gid = args->gid;
+
+ if (S_ISDIR(args->st_mode)) {
+ STACK_WIND(new_frame, ga_newentry_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mkdir, &tmp_loc,
+ args->args.mkdir.mode, args->args.mkdir.umask, xdata);
+ } else if (S_ISLNK(args->st_mode)) {
+ STACK_WIND(new_frame, ga_newentry_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->symlink,
+ args->args.symlink.linkpath, &tmp_loc, 0, xdata);
+ } else {
+ /* use 07777 (4 7s) for considering the Sticky bits etc) */
+ ((ga_local_t *)new_frame->local)->mode = (S_IFMT & args->st_mode) |
+ (07777 &
+ args->args.mknod.mode);
+
+ ((ga_local_t *)new_frame->local)->umask = args->args.mknod.umask;
+ ((ga_local_t *)new_frame->local)->rdev = args->args.mknod.rdev;
+ ((ga_local_t *)new_frame->local)->xdata = dict_ref(xdata);
+
+ /* send a named lookup, so that dht can cleanup up stale linkto
+ * files etc.
+ */
+ STACK_WIND(new_frame, ga_newentry_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, &tmp_loc, NULL);
+ }
- ret = 0;
+ ret = 0;
out:
- ga_newfile_args_free (args);
+ ga_newfile_args_free(args);
- if (xdata)
- dict_unref (xdata);
+ if (xdata)
+ dict_unref(xdata);
- loc_wipe (&tmp_loc);
+ loc_wipe(&tmp_loc);
- return ret;
+ return ret;
}
int32_t
-ga_heal_entry (call_frame_t *frame, xlator_t *this, loc_t *loc, data_t *data,
- dict_t *xdata)
+ga_heal_entry(call_frame_t *frame, xlator_t *this, loc_t *loc, data_t *data,
+ dict_t *xdata)
{
- int ret = -1;
- ga_heal_args_t *args = NULL;
- loc_t tmp_loc = {0,};
- call_frame_t *new_frame = NULL;
- uuid_t gfid = {0,};
-
- args = ga_heal_parse_args (this, data);
- if (!args)
- goto out;
-
- ret = gf_uuid_parse (args->gfid, gfid);
- if (ret)
- goto out;
-
- if (!xdata)
- xdata = dict_new ();
- else
- xdata = dict_ref (xdata);
-
- if (!xdata) {
- ret = -1;
- goto out;
- }
-
- ret = ga_fill_tmp_loc (loc, this, gfid, args->bname,
- xdata, &tmp_loc);
- if (ret)
- goto out;
-
- new_frame = copy_frame (frame);
- if (!new_frame)
- goto out;
-
- new_frame->local = (void *)frame;
-
- STACK_WIND (new_frame, ga_heal_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->lookup,
- &tmp_loc, xdata);
-
- ret = 0;
+ int ret = -1;
+ ga_heal_args_t *args = NULL;
+ loc_t tmp_loc = {
+ 0,
+ };
+ call_frame_t *new_frame = NULL;
+ uuid_t gfid = {
+ 0,
+ };
+
+ args = ga_heal_parse_args(this, data);
+ if (!args)
+ goto out;
+
+ ret = gf_uuid_parse(args->gfid, gfid);
+ if (ret)
+ goto out;
+
+ if (!xdata)
+ xdata = dict_new();
+ else
+ xdata = dict_ref(xdata);
+
+ if (!xdata) {
+ ret = -1;
+ goto out;
+ }
+
+ ret = ga_fill_tmp_loc(loc, this, gfid, args->bname, xdata, &tmp_loc);
+ if (ret)
+ goto out;
+
+ new_frame = copy_frame(frame);
+ if (!new_frame)
+ goto out;
+
+ new_frame->local = (void *)frame;
+
+ STACK_WIND(new_frame, ga_heal_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, &tmp_loc, xdata);
+
+ ret = 0;
out:
- if (args)
- ga_heal_args_free (args);
+ if (args)
+ ga_heal_args_free(args);
- loc_wipe (&tmp_loc);
+ loc_wipe(&tmp_loc);
- if (xdata)
- dict_unref (xdata);
+ if (xdata)
+ dict_unref(xdata);
- return ret;
+ return ret;
}
int32_t
-ga_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- dict_t *xdata)
+ga_setxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
- return 0;
+ STACK_UNWIND_STRICT(setxattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
int32_t
-ga_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
- int32_t flags, dict_t *xdata)
+ga_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- data_t *data = NULL;
- int op_errno = ENOMEM;
- int ret = 0;
- loc_t ga_loc = {0, };
-
- GFID_ACCESS_INODE_OP_CHECK (loc, op_errno, err);
-
- data = dict_get (dict, GF_FUSE_AUX_GFID_NEWFILE);
- if (data) {
- ret = ga_new_entry (frame, this, loc, data, xdata);
- if (ret)
- goto err;
- return 0;
- }
+ data_t *data = NULL;
+ int op_errno = ENOMEM;
+ int ret = 0;
+ loc_t ga_loc = {
+ 0,
+ };
+
+ GFID_ACCESS_INODE_OP_CHECK(loc, op_errno, err);
+
+ data = dict_get(dict, GF_FUSE_AUX_GFID_NEWFILE);
+ if (data) {
+ ret = ga_new_entry(frame, this, loc, data, xdata);
+ if (ret)
+ goto err;
+ return 0;
+ }
- data = dict_get (dict, GF_FUSE_AUX_GFID_HEAL);
- if (data) {
- ret = ga_heal_entry (frame, this, loc, data, xdata);
- if (ret)
- goto err;
- return 0;
- }
+ data = dict_get(dict, GF_FUSE_AUX_GFID_HEAL);
+ if (data) {
+ ret = ga_heal_entry(frame, this, loc, data, xdata);
+ if (ret)
+ goto err;
+ return 0;
+ }
- //If the inode is a virtual inode change the inode otherwise perform
- //the operation on same inode
- ret = ga_valid_inode_loc_copy (&ga_loc, loc, this);
- if (ret < 0)
- goto err;
+ // If the inode is a virtual inode change the inode otherwise perform
+ // the operation on same inode
+ ret = ga_valid_inode_loc_copy(&ga_loc, loc, this);
+ if (ret < 0)
+ goto err;
- STACK_WIND (frame, ga_setxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setxattr, &ga_loc, dict, flags,
- xdata);
+ STACK_WIND(frame, ga_setxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, &ga_loc, dict, flags, xdata);
- loc_wipe (&ga_loc);
- return 0;
+ loc_wipe(&ga_loc);
+ return 0;
err:
- STACK_UNWIND_STRICT (setxattr, frame, -1, op_errno, xdata);
- return 0;
+ STACK_UNWIND_STRICT(setxattr, frame, -1, op_errno, xdata);
+ return 0;
}
-
int32_t
-ga_virtual_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+ga_virtual_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata, struct iatt *postparent)
{
- int ret = 0;
- inode_t *cbk_inode = NULL;
- inode_t *true_inode = NULL;
- uuid_t random_gfid = {0,};
- inode_t *linked_inode = NULL;
-
- if (frame->local)
- cbk_inode = frame->local;
- else
- cbk_inode = inode_ref (inode);
-
- frame->local = NULL;
- if (op_ret)
- goto unwind;
-
- if (!IA_ISDIR (buf->ia_type))
+ int ret = 0;
+ inode_t *cbk_inode = NULL;
+ inode_t *true_inode = NULL;
+ uuid_t random_gfid = {
+ 0,
+ };
+ inode_t *linked_inode = NULL;
+
+ if (frame->local)
+ cbk_inode = frame->local;
+ else
+ cbk_inode = inode_ref(inode);
+
+ frame->local = NULL;
+ if (op_ret)
+ goto unwind;
+
+ if (!IA_ISDIR(buf->ia_type))
+ goto unwind;
+
+ /* need to send back a different inode for linking in itable */
+ if (cbk_inode == inode) {
+ /* check if the inode is in the 'itable' or
+ if its just previously discover()'d inode */
+ true_inode = inode_find(inode->table, buf->ia_gfid);
+ if (!true_inode) {
+ /* This unref is for 'inode_ref()' done in beginning.
+ This is needed as cbk_inode is allocated new inode
+ whose unref is taken at the end*/
+ inode_unref(cbk_inode);
+ cbk_inode = inode_new(inode->table);
+
+ if (!cbk_inode) {
+ op_ret = -1;
+ op_errno = ENOMEM;
goto unwind;
+ }
+ /* the inode is not present in itable, ie, the actual
+ path is not yet looked up. Use the current inode
+ itself for now */
- /* need to send back a different inode for linking in itable */
- if (cbk_inode == inode) {
- /* check if the inode is in the 'itable' or
- if its just previously discover()'d inode */
- true_inode = inode_find (inode->table, buf->ia_gfid);
- if (!true_inode) {
- /* This unref is for 'inode_ref()' done in beginning.
- This is needed as cbk_inode is allocated new inode
- whose unref is taken at the end*/
- inode_unref (cbk_inode);
- cbk_inode = inode_new (inode->table);
-
- if (!cbk_inode) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
- /* the inode is not present in itable, ie, the actual
- path is not yet looked up. Use the current inode
- itself for now */
-
- linked_inode = inode_link (inode, NULL, NULL, buf);
- inode = linked_inode;
- } else {
- /* 'inode_ref()' has been done in inode_find() */
- inode = true_inode;
- }
-
- ret = inode_ctx_put (cbk_inode, this, (uint64_t)inode);
- if (ret) {
- gf_log (this->name, GF_LOG_WARNING,
- "failed to set the inode ctx with"
- "the actual inode");
- if (inode)
- inode_unref (inode);
- }
- inode = NULL;
+ linked_inode = inode_link(inode, NULL, NULL, buf);
+ inode = linked_inode;
+ } else {
+ /* 'inode_ref()' has been done in inode_find() */
+ inode = true_inode;
}
- if (!gf_uuid_is_null (cbk_inode->gfid)) {
- /* if the previous linked inode is used, use the
- same gfid */
- gf_uuid_copy (random_gfid, cbk_inode->gfid);
- } else {
- /* replace the buf->ia_gfid to a random gfid
- for directory, for files, what we received is fine */
- gf_uuid_generate (random_gfid);
+ ret = inode_ctx_put(cbk_inode, this, (uint64_t)(uintptr_t)inode);
+ if (ret) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "failed to set the inode ctx with"
+ "the actual inode");
+ if (inode)
+ inode_unref(inode);
}
+ inode = NULL;
+ }
+
+ if (!gf_uuid_is_null(cbk_inode->gfid)) {
+ /* if the previous linked inode is used, use the
+ same gfid */
+ gf_uuid_copy(random_gfid, cbk_inode->gfid);
+ } else {
+ /* replace the buf->ia_gfid to a random gfid
+ for directory, for files, what we received is fine */
+ gf_uuid_generate(random_gfid);
+ }
- gf_uuid_copy (buf->ia_gfid, random_gfid);
+ gf_uuid_copy(buf->ia_gfid, random_gfid);
- buf->ia_ino = gfid_to_ino (buf->ia_gfid);
+ buf->ia_ino = gfid_to_ino(buf->ia_gfid);
unwind:
- /* Lookup on non-existing gfid returns ESTALE.
- Convert into ENOENT for virtual lookup*/
- if (op_errno == ESTALE)
- op_errno = ENOENT;
+ /* Lookup on non-existing gfid returns ESTALE.
+ Convert into ENOENT for virtual lookup*/
+ if (op_errno == ESTALE)
+ op_errno = ENOENT;
- STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, cbk_inode, buf,
- xdata, postparent);
+ STACK_UNWIND_STRICT(lookup, frame, op_ret, op_errno, cbk_inode, buf, xdata,
+ postparent);
- /* Also handles inode_unref of frame->local if done in ga_lookup */
- if (cbk_inode)
- inode_unref (cbk_inode);
+ /* Also handles inode_unref of frame->local if done in ga_lookup */
+ if (cbk_inode)
+ inode_unref(cbk_inode);
- return 0;
+ return 0;
}
int32_t
-ga_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+ga_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode, struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
{
- ga_private_t *priv = NULL;
+ ga_private_t *priv = NULL;
- /* if the entry in question is not 'root',
- then follow the normal path */
- if (op_ret || !__is_root_gfid(buf->ia_gfid))
- goto unwind;
+ /* if the entry in question is not 'root',
+ then follow the normal path */
+ if (op_ret || !__is_root_gfid(buf->ia_gfid))
+ goto unwind;
- priv = this->private;
+ priv = this->private;
- /* do we need to copy root stbuf everytime? */
- /* mostly yes, as we want to have the 'stat' info show latest
- in every _cbk() */
+ /* do we need to copy root stbuf every time? */
+ /* mostly yes, as we want to have the 'stat' info show latest
+ in every _cbk() */
- /* keep the reference for root stat buf */
- priv->root_stbuf = *buf;
- priv->gfiddir_stbuf = priv->root_stbuf;
- priv->gfiddir_stbuf.ia_gfid[15] = GF_AUX_GFID;
- priv->gfiddir_stbuf.ia_ino = GF_AUX_GFID;
+ /* keep the reference for root stat buf */
+ priv->root_stbuf = *buf;
+ priv->gfiddir_stbuf = priv->root_stbuf;
+ priv->gfiddir_stbuf.ia_gfid[15] = GF_AUX_GFID;
+ priv->gfiddir_stbuf.ia_ino = GF_AUX_GFID;
unwind:
- STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
- xdata, postparent);
- return 0;
+ STACK_UNWIND_STRICT(lookup, frame, op_ret, op_errno, inode, buf, xdata,
+ postparent);
+ return 0;
}
int32_t
-ga_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+ga_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- ga_private_t *priv = NULL;
- int ret = -1;
- uuid_t tmp_gfid = {0,};
- loc_t tmp_loc = {0,};
- uint64_t value = 0;
- inode_t *inode = NULL;
- inode_t *true_inode = NULL;
- int32_t op_errno = ENOENT;
-
- priv = this->private;
-
- /* Handle nameless lookup on ".gfid" */
- if (!loc->parent && __is_gfid_access_dir(loc->gfid)) {
- STACK_UNWIND_STRICT (lookup, frame, 0, 0, loc->inode,
- &priv->gfiddir_stbuf, xdata,
- &priv->root_stbuf);
- return 0;
- }
-
- /* if its discover(), no need for any action here */
- if (!loc->name)
- goto wind;
-
- /* if its revalidate, and inode is not of type directory,
- proceed with 'wind' */
- if (loc->inode && loc->inode->ia_type &&
- !IA_ISDIR (loc->inode->ia_type)) {
-
- /* a revalidate on ".gfid/<dentry>" is possible, check for it */
- if (((loc->parent &&
- __is_gfid_access_dir (loc->parent->gfid)) ||
- __is_gfid_access_dir (loc->pargfid))) {
-
- /* here, just send 'loc->gfid' and 'loc->inode' */
- tmp_loc.inode = inode_ref (loc->inode);
- gf_uuid_copy (tmp_loc.gfid, loc->inode->gfid);
+ ga_private_t *priv = NULL;
+ int ret = -1;
+ uuid_t tmp_gfid = {
+ 0,
+ };
+ loc_t tmp_loc = {
+ 0,
+ };
+ uint64_t value = 0;
+ inode_t *inode = NULL;
+ inode_t *true_inode = NULL;
+ int32_t op_errno = ENOENT;
+
+ priv = this->private;
+
+ /* Handle nameless lookup on ".gfid" */
+ if (!loc->parent && __is_gfid_access_dir(loc->gfid)) {
+ STACK_UNWIND_STRICT(lookup, frame, 0, 0, loc->inode,
+ &priv->gfiddir_stbuf, xdata, &priv->root_stbuf);
+ return 0;
+ }
- STACK_WIND (frame, default_lookup_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup,
- &tmp_loc, xdata);
+ /* if its discover(), no need for any action here */
+ if (!loc->name)
+ goto wind;
- inode_unref (tmp_loc.inode);
+ /* if its revalidate, and inode is not of type directory,
+ proceed with 'wind' */
+ if (loc->inode && loc->inode->ia_type && !IA_ISDIR(loc->inode->ia_type)) {
+ /* a revalidate on ".gfid/<dentry>" is possible, check for it */
+ if (((loc->parent && __is_gfid_access_dir(loc->parent->gfid)) ||
+ __is_gfid_access_dir(loc->pargfid))) {
+ /* here, just send 'loc->gfid' and 'loc->inode' */
+ tmp_loc.inode = inode_ref(loc->inode);
+ gf_uuid_copy(tmp_loc.gfid, loc->inode->gfid);
- return 0;
- }
+ STACK_WIND(frame, default_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, &tmp_loc, xdata);
- /* not something to bother, continue the flow */
- goto wind;
- }
+ inode_unref(tmp_loc.inode);
- /* need to check if the lookup is on virtual dir */
- if ((loc->name && !strcmp (GF_GFID_DIR, loc->name)) &&
- ((loc->parent && __is_root_gfid (loc->parent->gfid)) ||
- __is_root_gfid (loc->pargfid))) {
- /* this means, the query is on '/.gfid', return the fake stat,
- and say success */
-
- STACK_UNWIND_STRICT (lookup, frame, 0, 0, loc->inode,
- &priv->gfiddir_stbuf, xdata,
- &priv->root_stbuf);
- return 0;
+ return 0;
}
- /* now, check if the lookup() is on an existing entry,
- but on gfid-path */
- if (!((loc->parent && __is_gfid_access_dir (loc->parent->gfid)) ||
- __is_gfid_access_dir (loc->pargfid))) {
- if (!loc->parent)
- goto wind;
-
- ret = inode_ctx_get (loc->parent, this, &value);
- if (ret)
- goto wind;
+ /* not something to bother, continue the flow */
+ goto wind;
+ }
- inode = (inode_t *) value;
+ /* need to check if the lookup is on virtual dir */
+ if ((loc->name && !strcmp(GF_GFID_DIR, loc->name)) &&
+ ((loc->parent && __is_root_gfid(loc->parent->gfid)) ||
+ __is_root_gfid(loc->pargfid))) {
+ /* this means, the query is on '/.gfid', return the fake stat,
+ and say success */
- ret = loc_copy_overload_parent (&tmp_loc, loc, inode);
- if (ret)
- goto err;
+ STACK_UNWIND_STRICT(lookup, frame, 0, 0, loc->inode,
+ &priv->gfiddir_stbuf, xdata, &priv->root_stbuf);
+ return 0;
+ }
- STACK_WIND (frame, ga_lookup_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->lookup, &tmp_loc, xdata);
+ /* now, check if the lookup() is on an existing entry,
+ but on gfid-path */
+ if (!((loc->parent && __is_gfid_access_dir(loc->parent->gfid)) ||
+ __is_gfid_access_dir(loc->pargfid))) {
+ if (!loc->parent)
+ goto wind;
- loc_wipe (&tmp_loc);
- return 0;
- }
-
- /* make sure the 'basename' is actually a 'canonical-gfid',
- otherwise, return error */
- ret = gf_uuid_parse (loc->name, tmp_gfid);
+ ret = inode_ctx_get(loc->parent, this, &value);
if (ret)
- goto err;
+ goto wind;
- /* if its fresh lookup, go ahead and send it down, if not,
- for directory, we need indirection to actual dir inode */
- if (!(loc->inode && loc->inode->ia_type))
- goto discover;
+ inode = (inode_t *)(uintptr_t)value;
- /* revalidate on directory */
- ret = inode_ctx_get (loc->inode, this, &value);
+ ret = loc_copy_overload_parent(&tmp_loc, loc, inode);
if (ret)
- goto err;
-
- inode = (void *)value;
+ goto err;
- /* valid inode, already looked up, work on that */
- if (inode->ia_type)
- goto discover;
+ STACK_WIND(frame, ga_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, &tmp_loc, xdata);
- /* check if the inode is in the 'itable' or
- if its just previously discover()'d inode */
- true_inode = inode_find (loc->inode->table, tmp_gfid);
- if (true_inode) {
- /* time do another lookup and update the context
- with proper inode */
- op_errno = ESTALE;
- /* 'inode_ref()' done in inode_find */
- inode_unref (true_inode);
- goto err;
- }
+ loc_wipe(&tmp_loc);
+ return 0;
+ }
+
+ /* make sure the 'basename' is actually a 'canonical-gfid',
+ otherwise, return error */
+ ret = gf_uuid_parse(loc->name, tmp_gfid);
+ if (ret)
+ goto err;
+
+ /* if its fresh lookup, go ahead and send it down, if not,
+ for directory, we need indirection to actual dir inode */
+ if (!(loc->inode && loc->inode->ia_type))
+ goto discover;
+
+ /* revalidate on directory */
+ ret = inode_ctx_get(loc->inode, this, &value);
+ if (ret)
+ goto err;
+
+ inode = (void *)(uintptr_t)value;
+
+ /* valid inode, already looked up, work on that */
+ if (inode->ia_type)
+ goto discover;
+
+ /* check if the inode is in the 'itable' or
+ if its just previously discover()'d inode */
+ true_inode = inode_find(loc->inode->table, tmp_gfid);
+ if (true_inode) {
+ /* time do another lookup and update the context
+ with proper inode */
+ op_errno = ESTALE;
+ /* 'inode_ref()' done in inode_find */
+ inode_unref(true_inode);
+ goto err;
+ }
discover:
- /* for the virtual entries, we don't need to send 'gfid-req' key, as
- for these entries, we don't want to 'set' a new gfid */
- if (xdata)
- dict_del (xdata, "gfid-req");
+ /* for the virtual entries, we don't need to send 'gfid-req' key, as
+ for these entries, we don't want to 'set' a new gfid */
+ if (xdata)
+ dict_del(xdata, "gfid-req");
- gf_uuid_copy (tmp_loc.gfid, tmp_gfid);
+ gf_uuid_copy(tmp_loc.gfid, tmp_gfid);
- /* if revalidate, then we need to have the proper reference */
- if (inode) {
- tmp_loc.inode = inode_ref (inode);
- frame->local = inode_ref (loc->inode);
- } else {
- tmp_loc.inode = inode_ref (loc->inode);
- }
+ /* if revalidate, then we need to have the proper reference */
+ if (inode) {
+ tmp_loc.inode = inode_ref(inode);
+ frame->local = inode_ref(loc->inode);
+ } else {
+ tmp_loc.inode = inode_ref(loc->inode);
+ }
- STACK_WIND (frame, ga_virtual_lookup_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, &tmp_loc, xdata);
+ STACK_WIND(frame, ga_virtual_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, &tmp_loc, xdata);
- inode_unref (tmp_loc.inode);
+ inode_unref(tmp_loc.inode);
- return 0;
+ return 0;
wind:
- /* used for all the normal lookup path */
- STACK_WIND (frame, ga_lookup_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, loc, xdata);
+ /* used for all the normal lookup path */
+ STACK_WIND(frame, ga_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, xdata);
- return 0;
+ return 0;
err:
- STACK_UNWIND_STRICT (lookup, frame, -1, op_errno, loc->inode,
- &priv->gfiddir_stbuf, xdata,
- &priv->root_stbuf);
- return 0;
+ STACK_UNWIND_STRICT(lookup, frame, -1, op_errno, loc->inode,
+ &priv->gfiddir_stbuf, xdata, &priv->root_stbuf);
+ return 0;
}
int
-ga_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- mode_t umask, dict_t *xdata)
+ga_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- int op_errno = ENOMEM;
+ int op_errno = ENOMEM;
- GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+ GFID_ACCESS_ENTRY_OP_CHECK(loc, op_errno, err);
- STACK_WIND (frame, default_mkdir_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mkdir, loc, mode, umask,
- xdata);
+ STACK_WIND(frame, default_mkdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
- return 0;
+ return 0;
err:
- STACK_UNWIND_STRICT (mkdir, frame, -1, op_errno, loc->inode,
- NULL, NULL, NULL, xdata);
- return 0;
+ STACK_UNWIND_STRICT(mkdir, frame, -1, op_errno, loc->inode, NULL, NULL,
+ NULL, xdata);
+ return 0;
}
-
int
-ga_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
- mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+ga_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- int op_errno = ENOMEM;
+ int op_errno = ENOMEM;
- GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+ GFID_ACCESS_ENTRY_OP_CHECK(loc, op_errno, err);
- STACK_WIND (frame, default_create_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->create,
- loc, flags, mode, umask, fd, xdata);
- return 0;
+ STACK_WIND(frame, default_create_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, flags, mode, umask, fd,
+ xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (create, frame, -1, op_errno, NULL,
- NULL, NULL, NULL, NULL, xdata);
-
- return 0;
+ STACK_UNWIND_STRICT(create, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL, xdata);
+ return 0;
}
int
-ga_symlink (call_frame_t *frame, xlator_t *this, const char *linkname,
- loc_t *loc, mode_t umask, dict_t *xdata)
+ga_symlink(call_frame_t *frame, xlator_t *this, const char *linkname,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
- int op_errno = ENOMEM;
+ int op_errno = ENOMEM;
- GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+ GFID_ACCESS_ENTRY_OP_CHECK(loc, op_errno, err);
- STACK_WIND (frame, default_symlink_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->symlink,
- linkname, loc, umask, xdata);
- return 0;
+ STACK_WIND(frame, default_symlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->symlink, linkname, loc, umask, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (symlink, frame, -1, op_errno, NULL,
- NULL, NULL, NULL, xdata);
+ STACK_UNWIND_STRICT(symlink, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ xdata);
- return 0;
+ return 0;
}
int
-ga_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- dev_t rdev, mode_t umask, dict_t *xdata)
+ga_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *xdata)
{
- int op_errno = ENOMEM;
+ int op_errno = ENOMEM;
- GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+ GFID_ACCESS_ENTRY_OP_CHECK(loc, op_errno, err);
- STACK_WIND (frame, default_mknod_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mknod, loc, mode, rdev,
- umask, xdata);
+ STACK_WIND(frame, default_mknod_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, loc, mode, rdev, umask, xdata);
- return 0;
+ return 0;
err:
- STACK_UNWIND_STRICT (mknod, frame, -1, op_errno, NULL,
- NULL, NULL, NULL, xdata);
+ STACK_UNWIND_STRICT(mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ xdata);
- return 0;
+ return 0;
}
int
-ga_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flag,
- dict_t *xdata)
+ga_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int flag,
+ dict_t *xdata)
{
- int op_errno = ENOMEM;
- int ret = -1;
- loc_t ga_loc = {0, };
+ int op_errno = ENOMEM;
+ int ret = -1;
+ loc_t ga_loc = {
+ 0,
+ };
- GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+ GFID_ACCESS_ENTRY_OP_CHECK(loc, op_errno, err);
- ret = ga_valid_inode_loc_copy (&ga_loc, loc, this);
- if (ret < 0)
- goto err;
+ ret = ga_valid_inode_loc_copy(&ga_loc, loc, this);
+ if (ret < 0)
+ goto err;
- STACK_WIND (frame, default_rmdir_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->rmdir,
- &ga_loc, flag, xdata);
+ STACK_WIND(frame, default_rmdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rmdir, &ga_loc, flag, xdata);
- loc_wipe (&ga_loc);
- return 0;
+ loc_wipe(&ga_loc);
+ return 0;
err:
- STACK_UNWIND_STRICT (rmdir, frame, -1, op_errno, NULL,
- NULL, xdata);
+ STACK_UNWIND_STRICT(rmdir, frame, -1, op_errno, NULL, NULL, xdata);
- return 0;
+ return 0;
}
int
-ga_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t xflag,
- dict_t *xdata)
+ga_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t xflag,
+ dict_t *xdata)
{
- int op_errno = ENOMEM;
- int ret = -1;
- loc_t ga_loc = {0, };
+ int op_errno = ENOMEM;
+ int ret = -1;
+ loc_t ga_loc = {
+ 0,
+ };
- GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+ GFID_ACCESS_ENTRY_OP_CHECK(loc, op_errno, err);
- ret = ga_valid_inode_loc_copy (&ga_loc, loc, this);
- if (ret < 0)
- goto err;
+ ret = ga_valid_inode_loc_copy(&ga_loc, loc, this);
+ if (ret < 0)
+ goto err;
- STACK_WIND (frame, default_unlink_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->unlink,
- &ga_loc, xflag, xdata);
+ STACK_WIND(frame, default_unlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, &ga_loc, xflag, xdata);
- loc_wipe (&ga_loc);
- return 0;
+ loc_wipe(&ga_loc);
+ return 0;
err:
- STACK_UNWIND_STRICT (unlink, frame, -1, op_errno, NULL,
- NULL, xdata);
+ STACK_UNWIND_STRICT(unlink, frame, -1, op_errno, NULL, NULL, xdata);
- return 0;
+ return 0;
}
int
-ga_rename (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+ga_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- int op_errno = ENOMEM;
- int ret = 0;
- loc_t ga_oldloc = {0, };
- loc_t ga_newloc = {0, };
-
- GFID_ACCESS_ENTRY_OP_CHECK (oldloc, op_errno, err);
- GFID_ACCESS_ENTRY_OP_CHECK (newloc, op_errno, err);
-
- ret = ga_valid_inode_loc_copy (&ga_oldloc, oldloc, this);
- if (ret < 0)
- goto err;
-
- ret = ga_valid_inode_loc_copy (&ga_newloc, newloc, this);
- if (ret < 0) {
- loc_wipe (&ga_oldloc);
- goto err;
- }
-
- STACK_WIND (frame, default_rename_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->rename,
- &ga_oldloc, &ga_newloc, xdata);
-
- loc_wipe (&ga_newloc);
- loc_wipe (&ga_oldloc);
- return 0;
+ int op_errno = ENOMEM;
+ int ret = 0;
+ loc_t ga_oldloc = {
+ 0,
+ };
+ loc_t ga_newloc = {
+ 0,
+ };
+
+ GFID_ACCESS_ENTRY_OP_CHECK(oldloc, op_errno, err);
+ GFID_ACCESS_ENTRY_OP_CHECK(newloc, op_errno, err);
+
+ ret = ga_valid_inode_loc_copy(&ga_oldloc, oldloc, this);
+ if (ret < 0)
+ goto err;
+
+ ret = ga_valid_inode_loc_copy(&ga_newloc, newloc, this);
+ if (ret < 0) {
+ loc_wipe(&ga_oldloc);
+ goto err;
+ }
+
+ STACK_WIND(frame, default_rename_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, &ga_oldloc, &ga_newloc, xdata);
+
+ loc_wipe(&ga_newloc);
+ loc_wipe(&ga_oldloc);
+ return 0;
err:
- STACK_UNWIND_STRICT (rename, frame, -1, op_errno, NULL,
- NULL, NULL, NULL, NULL, xdata);
+ STACK_UNWIND_STRICT(rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL, xdata);
- return 0;
+ return 0;
}
-
int
-ga_link (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+ga_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- int op_errno = ENOMEM;
- int ret = 0;
- loc_t ga_oldloc = {0, };
- loc_t ga_newloc = {0, };
-
- GFID_ACCESS_ENTRY_OP_CHECK (oldloc, op_errno, err);
- GFID_ACCESS_ENTRY_OP_CHECK (newloc, op_errno, err);
-
- ret = ga_valid_inode_loc_copy (&ga_oldloc, oldloc, this);
- if (ret < 0)
- goto err;
-
- ret = ga_valid_inode_loc_copy (&ga_newloc, newloc, this);
- if (ret < 0) {
- loc_wipe (&ga_oldloc);
- goto err;
- }
-
- STACK_WIND (frame, default_link_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->link,
- &ga_oldloc, &ga_newloc, xdata);
-
- loc_wipe (&ga_newloc);
- loc_wipe (&ga_oldloc);
- return 0;
+ int op_errno = ENOMEM;
+ int ret = 0;
+ loc_t ga_oldloc = {
+ 0,
+ };
+ loc_t ga_newloc = {
+ 0,
+ };
+
+ GFID_ACCESS_ENTRY_OP_CHECK(oldloc, op_errno, err);
+ GFID_ACCESS_ENTRY_OP_CHECK(newloc, op_errno, err);
+
+ ret = ga_valid_inode_loc_copy(&ga_oldloc, oldloc, this);
+ if (ret < 0)
+ goto err;
+
+ ret = ga_valid_inode_loc_copy(&ga_newloc, newloc, this);
+ if (ret < 0) {
+ loc_wipe(&ga_oldloc);
+ goto err;
+ }
+
+ STACK_WIND(frame, default_link_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->link, &ga_oldloc, &ga_newloc, xdata);
+
+ loc_wipe(&ga_newloc);
+ loc_wipe(&ga_oldloc);
+ return 0;
err:
- STACK_UNWIND_STRICT (link, frame, -1, op_errno, NULL,
- NULL, NULL, NULL, xdata);
+ STACK_UNWIND_STRICT(link, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ xdata);
- return 0;
+ return 0;
}
int32_t
-ga_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc,
- fd_t *fd, dict_t *xdata)
+ga_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata)
{
- int op_errno = ENOMEM;
+ int op_errno = ENOMEM;
- GFID_ACCESS_INODE_OP_CHECK (loc, op_errno, err);
+ GFID_ACCESS_INODE_OP_CHECK(loc, op_errno, err);
- /* also check if the loc->inode itself is virtual
- inode, if yes, return with failure, mainly because we
- can't handle all the readdirp and other things on it. */
- if (inode_ctx_get (loc->inode, this, NULL) == 0) {
- op_errno = ENOTSUP;
- goto err;
- }
+ /* also check if the loc->inode itself is virtual
+ inode, if yes, return with failure, mainly because we
+ can't handle all the readdirp and other things on it. */
+ if (inode_ctx_get(loc->inode, this, NULL) == 0) {
+ op_errno = ENOTSUP;
+ goto err;
+ }
- STACK_WIND (frame, default_opendir_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->opendir,
- loc, fd, xdata);
- return 0;
+ STACK_WIND(frame, default_opendir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->opendir, loc, fd, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (opendir, frame, -1, op_errno, NULL, xdata);
+ STACK_UNWIND_STRICT(opendir, frame, -1, op_errno, NULL, xdata);
- return 0;
+ return 0;
}
int32_t
-ga_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+ga_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, const char *name,
+ dict_t *xdata)
{
- int op_errno = ENOMEM;
- int ret = -1;
- loc_t ga_loc = {0, };
+ int op_errno = ENOMEM;
+ int ret = -1;
+ loc_t ga_loc = {
+ 0,
+ };
- GFID_ACCESS_INODE_OP_CHECK (loc, op_errno, err);
- ret = ga_valid_inode_loc_copy (&ga_loc, loc, this);
- if (ret < 0)
- goto err;
+ GFID_ACCESS_INODE_OP_CHECK(loc, op_errno, err);
+ ret = ga_valid_inode_loc_copy(&ga_loc, loc, this);
+ if (ret < 0)
+ goto err;
- STACK_WIND (frame, default_getxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr, &ga_loc, name, xdata);
+ STACK_WIND(frame, default_getxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->getxattr, &ga_loc, name, xdata);
- loc_wipe (&ga_loc);
+ loc_wipe(&ga_loc);
- return 0;
+ return 0;
err:
- STACK_UNWIND_STRICT (getxattr, frame, -1, op_errno, NULL, xdata);
+ STACK_UNWIND_STRICT(getxattr, frame, -1, op_errno, NULL, xdata);
- return 0;
+ return 0;
}
int32_t
-ga_stat (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *xdata)
+ga_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- int op_errno = ENOMEM;
- int ret = -1;
- loc_t ga_loc = {0, };
- ga_private_t *priv = NULL;
-
- priv = this->private;
- /* If stat is on ".gfid" itself, do not wind further,
- * return fake stat and return success.
- */
- if (__is_gfid_access_dir(loc->gfid))
- goto out;
-
- ret = ga_valid_inode_loc_copy (&ga_loc, loc, this);
- if (ret < 0)
- goto err;
-
- STACK_WIND (frame, default_stat_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->stat, &ga_loc, xdata);
-
- loc_wipe (&ga_loc);
- return 0;
+ int op_errno = ENOMEM;
+ int ret = -1;
+ loc_t ga_loc = {
+ 0,
+ };
+ ga_private_t *priv = NULL;
+
+ priv = this->private;
+ /* If stat is on ".gfid" itself, do not wind further,
+ * return fake stat and return success.
+ */
+ if (__is_gfid_access_dir(loc->gfid))
+ goto out;
+
+ ret = ga_valid_inode_loc_copy(&ga_loc, loc, this);
+ if (ret < 0)
+ goto err;
+
+ STACK_WIND(frame, default_stat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->stat, &ga_loc, xdata);
+
+ loc_wipe(&ga_loc);
+ return 0;
err:
- STACK_UNWIND_STRICT (stat, frame, -1, op_errno, NULL, xdata);
+ STACK_UNWIND_STRICT(stat, frame, -1, op_errno, NULL, xdata);
- return 0;
+ return 0;
out:
- STACK_UNWIND_STRICT (stat, frame, 0, 0, &priv->gfiddir_stbuf, xdata);
- return 0;
+ STACK_UNWIND_STRICT(stat, frame, 0, 0, &priv->gfiddir_stbuf, xdata);
+ return 0;
}
int32_t
-ga_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid,
- dict_t *xdata)
+ga_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *stbuf,
+ int32_t valid, dict_t *xdata)
{
- int op_errno = ENOMEM;
- int ret = -1;
- loc_t ga_loc = {0, };
-
- GFID_ACCESS_INODE_OP_CHECK (loc, op_errno, err);
- ret = ga_valid_inode_loc_copy (&ga_loc, loc, this);
- if (ret < 0)
- goto err;
-
- STACK_WIND (frame, default_setattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setattr, &ga_loc, stbuf, valid,
- xdata);
-
- loc_wipe (&ga_loc);
- return 0;
+ int op_errno = ENOMEM;
+ int ret = -1;
+ loc_t ga_loc = {
+ 0,
+ };
+
+ GFID_ACCESS_INODE_OP_CHECK(loc, op_errno, err);
+ ret = ga_valid_inode_loc_copy(&ga_loc, loc, this);
+ if (ret < 0)
+ goto err;
+
+ STACK_WIND(frame, default_setattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setattr, &ga_loc, stbuf, valid, xdata);
+
+ loc_wipe(&ga_loc);
+ return 0;
err:
- STACK_UNWIND_STRICT (setattr, frame, -1, op_errno, NULL, NULL, xdata);
+ STACK_UNWIND_STRICT(setattr, frame, -1, op_errno, NULL, NULL, xdata);
- return 0;
+ return 0;
}
int32_t
-ga_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+ga_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- int op_errno = ENOMEM;
- int ret = -1;
- loc_t ga_loc = {0, };
+ int op_errno = ENOMEM;
+ int ret = -1;
+ loc_t ga_loc = {
+ 0,
+ };
- GFID_ACCESS_INODE_OP_CHECK (loc, op_errno, err);
- ret = ga_valid_inode_loc_copy (&ga_loc, loc, this);
- if (ret < 0)
- goto err;
+ GFID_ACCESS_INODE_OP_CHECK(loc, op_errno, err);
+ ret = ga_valid_inode_loc_copy(&ga_loc, loc, this);
+ if (ret < 0)
+ goto err;
- STACK_WIND (frame, default_removexattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->removexattr, &ga_loc, name,
- xdata);
+ STACK_WIND(frame, default_removexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, &ga_loc, name, xdata);
- loc_wipe (&ga_loc);
- return 0;
+ loc_wipe(&ga_loc);
+ return 0;
err:
- STACK_UNWIND_STRICT (removexattr, frame, -1, op_errno, xdata);
+ STACK_UNWIND_STRICT(removexattr, frame, -1, op_errno, xdata);
- return 0;
+ return 0;
}
-
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
-
- if (!this)
- return ret;
+ int ret = -1;
- ret = xlator_mem_acct_init (this, gf_gfid_access_mt_end + 1);
+ if (!this)
+ return ret;
- if (ret != 0) {
- gf_log (this->name, GF_LOG_WARNING, "Memory accounting"
- " init failed");
- return ret;
- }
+ ret = xlator_mem_acct_init(this, gf_gfid_access_mt_end + 1);
+ if (ret != 0) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "Memory accounting"
+ " init failed");
return ret;
+ }
+
+ return ret;
}
int32_t
-init (xlator_t *this)
+init(xlator_t *this)
{
- ga_private_t *priv = NULL;
- int ret = -1;
-
- if (!this->children || this->children->next) {
- gf_log (this->name, GF_LOG_ERROR,
- "not configured with exactly one child. exiting");
- goto out;
- }
-
- /* This can be the top of graph in certain cases */
- if (!this->parents) {
- gf_log (this->name, GF_LOG_DEBUG,
- "dangling volume. check volfile ");
- }
-
- /* TODO: define a mem-type structure */
- priv = GF_CALLOC (1, sizeof (*priv), gf_gfid_access_mt_priv_t);
- if (!priv)
- goto out;
-
- priv->newfile_args_pool = mem_pool_new (ga_newfile_args_t, 512);
- if (!priv->newfile_args_pool)
- goto out;
-
- priv->heal_args_pool = mem_pool_new (ga_heal_args_t, 512);
- if (!priv->heal_args_pool)
- goto out;
-
- this->local_pool = mem_pool_new (ga_local_t, 16);
- if (!this->local_pool) {
- gf_log (this->name, GF_LOG_ERROR,
- "failed to create local_t's memory pool");
- goto out;
- }
-
- this->private = priv;
-
- ret = 0;
+ ga_private_t *priv = NULL;
+ int ret = -1;
+
+ if (!this->children || this->children->next) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "not configured with exactly one child. exiting");
+ goto out;
+ }
+
+ /* This can be the top of graph in certain cases */
+ if (!this->parents) {
+ gf_log(this->name, GF_LOG_DEBUG, "dangling volume. check volfile ");
+ }
+
+ /* TODO: define a mem-type structure */
+ priv = GF_CALLOC(1, sizeof(*priv), gf_gfid_access_mt_priv_t);
+ if (!priv)
+ goto out;
+
+ priv->newfile_args_pool = mem_pool_new(ga_newfile_args_t, 512);
+ if (!priv->newfile_args_pool)
+ goto out;
+
+ priv->heal_args_pool = mem_pool_new(ga_heal_args_t, 512);
+ if (!priv->heal_args_pool)
+ goto out;
+
+ this->local_pool = mem_pool_new(ga_local_t, 16);
+ if (!this->local_pool) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to create local_t's memory pool");
+ goto out;
+ }
+
+ this->private = priv;
+
+ ret = 0;
out:
- if (ret && priv) {
- if (priv->newfile_args_pool)
- mem_pool_destroy (priv->newfile_args_pool);
- GF_FREE (priv);
- }
+ if (ret && priv) {
+ if (priv->newfile_args_pool)
+ mem_pool_destroy(priv->newfile_args_pool);
+ GF_FREE(priv);
+ }
- return ret;
+ return ret;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- ga_private_t *priv = NULL;
- priv = this->private;
- this->private = NULL;
-
- if (priv) {
- if (priv->newfile_args_pool)
- mem_pool_destroy (priv->newfile_args_pool);
- if (priv->heal_args_pool)
- mem_pool_destroy (priv->heal_args_pool);
- GF_FREE (priv);
- }
-
- return;
+ ga_private_t *priv = NULL;
+ priv = this->private;
+ this->private = NULL;
+
+ if (priv) {
+ if (priv->newfile_args_pool)
+ mem_pool_destroy(priv->newfile_args_pool);
+ if (priv->heal_args_pool)
+ mem_pool_destroy(priv->heal_args_pool);
+ GF_FREE(priv);
+ }
+
+ return;
}
int32_t
-ga_dump_inodectx (xlator_t *this, inode_t *inode)
+ga_dump_inodectx(xlator_t *this, inode_t *inode)
{
- int ret = -1;
- uint64_t value = 0;
- inode_t *tmp_inode = NULL;
- char key_prefix[GF_DUMP_MAX_BUF_LEN] = {0, };
-
- ret = inode_ctx_get (inode, this, &value);
- if (ret == 0) {
- tmp_inode = (void*) value;
- gf_proc_dump_build_key (key_prefix, this->name, "inode");
- gf_proc_dump_add_section (key_prefix);
- gf_proc_dump_write ("real-gfid", "%s",
- uuid_utoa (tmp_inode->gfid));
- }
-
- return 0;
+ int ret = -1;
+ uint64_t value = 0;
+ inode_t *tmp_inode = NULL;
+ char key_prefix[GF_DUMP_MAX_BUF_LEN] = {
+ 0,
+ };
+
+ ret = inode_ctx_get(inode, this, &value);
+ if (ret == 0) {
+ tmp_inode = (void *)(uintptr_t)value;
+ gf_proc_dump_build_key(key_prefix, this->name, "inode");
+ gf_proc_dump_add_section("%s", key_prefix);
+ gf_proc_dump_write("real-gfid", "%s", uuid_utoa(tmp_inode->gfid));
+ }
+
+ return 0;
}
struct xlator_fops fops = {
- .lookup = ga_lookup,
-
- /* entry fops */
- .mkdir = ga_mkdir,
- .mknod = ga_mknod,
- .create = ga_create,
- .symlink = ga_symlink,
- .link = ga_link,
- .unlink = ga_unlink,
- .rmdir = ga_rmdir,
- .rename = ga_rename,
-
- /* handle any other directory operations here */
- .opendir = ga_opendir,
- .stat = ga_stat,
- .setattr = ga_setattr,
- .getxattr = ga_getxattr,
- .removexattr = ga_removexattr,
-
- /* special fop to handle more entry creations */
- .setxattr = ga_setxattr,
+ .lookup = ga_lookup,
+
+ /* entry fops */
+ .mkdir = ga_mkdir,
+ .mknod = ga_mknod,
+ .create = ga_create,
+ .symlink = ga_symlink,
+ .link = ga_link,
+ .unlink = ga_unlink,
+ .rmdir = ga_rmdir,
+ .rename = ga_rename,
+
+ /* handle any other directory operations here */
+ .opendir = ga_opendir,
+ .stat = ga_stat,
+ .setattr = ga_setattr,
+ .getxattr = ga_getxattr,
+ .removexattr = ga_removexattr,
+
+ /* special fop to handle more entry creations */
+ .setxattr = ga_setxattr,
};
struct xlator_cbks cbks = {
- .forget = ga_forget,
+ .forget = ga_forget,
};
struct xlator_dumpops dumpops = {
- .inodectx = ga_dump_inodectx,
+ .inodectx = ga_dump_inodectx,
};
struct volume_options options[] = {
- /* This translator doesn't take any options, or provide any options */
- { .key = {NULL} },
+ /* This translator doesn't take any options, or provide any options */
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1},
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "gfid-access",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/features/gfid-access/src/gfid-access.h b/xlators/features/gfid-access/src/gfid-access.h
index 2b5e4fd4184..b1e255e56c0 100644
--- a/xlators/features/gfid-access/src/gfid-access.h
+++ b/xlators/features/gfid-access/src/gfid-access.h
@@ -10,97 +10,97 @@
#ifndef __GFID_ACCESS_H__
#define __GFID_ACCESS_H__
-#include "glusterfs.h"
-#include "logging.h"
-#include "dict.h"
-#include "xlator.h"
-#include "defaults.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/dict.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
#include "gfid-access-mem-types.h"
#define UUID_CANONICAL_FORM_LEN 36
#define GF_FUSE_AUX_GFID_NEWFILE "glusterfs.gfid.newfile"
-#define GF_FUSE_AUX_GFID_HEAL "glusterfs.gfid.heal"
+#define GF_FUSE_AUX_GFID_HEAL "glusterfs.gfid.heal"
#define GF_GFID_KEY "GLUSTERFS_GFID"
#define GF_GFID_DIR ".gfid"
#define GF_AUX_GFID 0xd
-#define GFID_ACCESS_ENTRY_OP_CHECK(loc,err,lbl) do { \
- /* need to check if the lookup is on virtual dir */ \
- if ((loc->name && !strcmp (GF_GFID_DIR, loc->name)) && \
- ((loc->parent && \
- __is_root_gfid (loc->parent->gfid)) || \
- __is_root_gfid (loc->pargfid))) { \
- err = ENOTSUP; \
- goto lbl; \
- } \
- \
- /* now, check if the lookup() is on an existing */ \
- /* entry, but on gfid-path */ \
- if ((loc->parent && \
- __is_gfid_access_dir (loc->parent->gfid)) || \
- __is_gfid_access_dir (loc->pargfid)) { \
- err = EPERM; \
- goto lbl; \
- } \
- } while (0)
+#define GFID_ACCESS_ENTRY_OP_CHECK(loc, err, lbl) \
+ do { \
+ /* need to check if the lookup is on virtual dir */ \
+ if ((loc->name && !strcmp(GF_GFID_DIR, loc->name)) && \
+ ((loc->parent && __is_root_gfid(loc->parent->gfid)) || \
+ __is_root_gfid(loc->pargfid))) { \
+ err = ENOTSUP; \
+ goto lbl; \
+ } \
+ \
+ /* now, check if the lookup() is on an existing */ \
+ /* entry, but on gfid-path */ \
+ if ((loc->parent && __is_gfid_access_dir(loc->parent->gfid)) || \
+ __is_gfid_access_dir(loc->pargfid)) { \
+ err = EPERM; \
+ goto lbl; \
+ } \
+ } while (0)
-#define GFID_ACCESS_INODE_OP_CHECK(loc,err,lbl) do { \
- /*Check if it is on .gfid*/ \
- if (__is_gfid_access_dir(loc->gfid)) { \
- err = ENOTSUP; \
- goto lbl; \
- } \
- } while (0)
+#define GFID_ACCESS_INODE_OP_CHECK(loc, err, lbl) \
+ do { \
+ /*Check if it is on .gfid*/ \
+ if (__is_gfid_access_dir(loc->gfid)) { \
+ err = ENOTSUP; \
+ goto lbl; \
+ } \
+ } while (0)
typedef struct {
- unsigned int uid;
- unsigned int gid;
- char gfid[UUID_CANONICAL_FORM_LEN + 1];
- unsigned int st_mode;
- char *bname;
+ unsigned int uid;
+ unsigned int gid;
+ char gfid[UUID_CANONICAL_FORM_LEN + 1];
+ unsigned int st_mode;
+ char *bname;
- union {
- struct _symlink_in {
- char *linkpath;
- } __attribute__ ((__packed__)) symlink;
+ union {
+ struct _symlink_in {
+ char *linkpath;
+ } __attribute__((__packed__)) symlink;
- struct _mknod_in {
- unsigned int mode;
- unsigned int rdev;
- unsigned int umask;
- } __attribute__ ((__packed__)) mknod;
+ struct _mknod_in {
+ unsigned int mode;
+ unsigned int rdev;
+ unsigned int umask;
+ } __attribute__((__packed__)) mknod;
- struct _mkdir_in {
- unsigned int mode;
- unsigned int umask;
- } __attribute__ ((__packed__)) mkdir;
- } __attribute__ ((__packed__)) args;
+ struct _mkdir_in {
+ unsigned int mode;
+ unsigned int umask;
+ } __attribute__((__packed__)) mkdir;
+ } __attribute__((__packed__)) args;
} __attribute__((__packed__)) ga_newfile_args_t;
typedef struct {
- char gfid[UUID_CANONICAL_FORM_LEN + 1];
- char *bname; /* a null terminated basename */
+ char gfid[UUID_CANONICAL_FORM_LEN + 1];
+ char *bname; /* a null terminated basename */
} __attribute__((__packed__)) ga_heal_args_t;
struct ga_private {
- /* root inode's stbuf */
- struct iatt root_stbuf;
- struct iatt gfiddir_stbuf;
- struct mem_pool *newfile_args_pool;
- struct mem_pool *heal_args_pool;
+ /* root inode's stbuf */
+ struct iatt root_stbuf;
+ struct iatt gfiddir_stbuf;
+ struct mem_pool *newfile_args_pool;
+ struct mem_pool *heal_args_pool;
};
typedef struct ga_private ga_private_t;
struct __ga_local {
- call_frame_t *orig_frame;
- unsigned int uid;
- unsigned int gid;
- loc_t loc;
- mode_t mode;
- dev_t rdev;
- mode_t umask;
- dict_t *xdata;
+ call_frame_t *orig_frame;
+ unsigned int uid;
+ unsigned int gid;
+ loc_t loc;
+ mode_t mode;
+ dev_t rdev;
+ mode_t umask;
+ dict_t *xdata;
};
typedef struct __ga_local ga_local_t;
diff --git a/xlators/features/glupy/Makefile.am b/xlators/features/glupy/Makefile.am
deleted file mode 100644
index 060429ecf0f..00000000000
--- a/xlators/features/glupy/Makefile.am
+++ /dev/null
@@ -1,3 +0,0 @@
-SUBDIRS = src examples
-
-CLEANFILES =
diff --git a/xlators/features/glupy/doc/README.md b/xlators/features/glupy/doc/README.md
deleted file mode 100644
index 4b8b863ef39..00000000000
--- a/xlators/features/glupy/doc/README.md
+++ /dev/null
@@ -1,44 +0,0 @@
-This is just the very start for a GlusterFS[1] meta-translator that will
-allow translator code to be written in Python. It's based on the standard
-Python embedding (not extending) techniques, plus a dash of the ctypes module.
-The interface is a pretty minimal adaptation of the dispatches and callbacks
-from the C API[2] to Python, as follows:
-
-* Dispatch functions and callbacks must be defined on an "xlator" class
- derived from gluster.Translator so that they'll be auto-registered with
- the C translator during initialization.
-
-* For each dispatch or callback function you want to intercept, you define a
- Python function using the xxx\_fop\_t or xxx\_cbk\_t decorator.
-
-* The arguments for each operation are different, so you'll need to refer to
- the C API. GlusterFS-specific types are used (though only loc\_t is fully
- defined so far) and type correctness is enforced by ctypes.
-
-* If you do intercept a dispatch function, it is your responsibility to call
- xxx\_wind (like STACK\_WIND in the C API but operation-specific) to pass
- the request to the next translator. If you do not intercept a function, it
- will default the same way as for C (pass through to the same operation with
- the same arguments on the first child translator).
-
-* If you intercept a callback function, it is your responsibility to call
- xxx\_unwind (like STACK\_UNWIND\_STRICT in the C API) to pass the request back
- to the caller.
-
-So far only the lookup and create operations are handled this way, to support
-the "negative lookup" example. Now that the basic infrastructure is in place,
-adding more functions should be very quick, though with that much boilerplate I
-might pause to write a code generator. I also plan to add structure
-definitions and interfaces for some of the utility functions in libglusterfs
-(especially those having to do with inode and fd context) in the fairly near
-future. Note that you can also use ctypes to get at anything not explicitly
-exposed to Python already.
-
-_If you're coming here because of the Linux Journal article, please note that
-the code has evolved since that was written. The version that matches the
-article is here:_
-
-https://github.com/jdarcy/glupy/tree/4bbae91ba459ea46ef32f2966562492e4ca9187a
-
-[1] http://www.gluster.org
-[2] http://pl.atyp.us/hekafs.org/dist/xlator_api_2.html
diff --git a/xlators/features/glupy/doc/TESTING b/xlators/features/glupy/doc/TESTING
deleted file mode 100644
index e05f17f498f..00000000000
--- a/xlators/features/glupy/doc/TESTING
+++ /dev/null
@@ -1,9 +0,0 @@
-Loading a translator written in Python using the glupy meta translator
--------------------------------------------------------------------------------
-'test.vol' is a simple volfile with the debug-trace Python translator on top
-of a brick. The volfile can be mounted using the following command.
-
-$ glusterfs --debug -f test.vol /path/to/mntpt
-
-If then file operations are performed on the newly mounted file system, log
-output would be printed by the Python translator on the standard output.
diff --git a/xlators/features/glupy/doc/test.vol b/xlators/features/glupy/doc/test.vol
deleted file mode 100644
index 0751a488c1f..00000000000
--- a/xlators/features/glupy/doc/test.vol
+++ /dev/null
@@ -1,10 +0,0 @@
-volume vol-posix
- type storage/posix
- option directory /path/to/brick
-end-volume
-
-volume vol-glupy
- type features/glupy
- option module-name debug-trace
- subvolumes vol-posix
-end-volume
diff --git a/xlators/features/glupy/examples/Makefile.am b/xlators/features/glupy/examples/Makefile.am
deleted file mode 100644
index c26abeaafb6..00000000000
--- a/xlators/features/glupy/examples/Makefile.am
+++ /dev/null
@@ -1,5 +0,0 @@
-xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
-
-glupyexamplesdir = $(xlatordir)/glupy
-
-glupyexamples_PYTHON = negative.py helloworld.py debug-trace.py
diff --git a/xlators/features/glupy/examples/debug-trace.py b/xlators/features/glupy/examples/debug-trace.py
deleted file mode 100644
index ca726ed21cb..00000000000
--- a/xlators/features/glupy/examples/debug-trace.py
+++ /dev/null
@@ -1,777 +0,0 @@
-
-from __future__ import print_function
-import sys
-import stat
-from uuid import UUID
-from time import strftime, localtime
-from gluster.glupy import *
-
-# This translator was written primarily to test the fop entry point definitions
-# and structure definitions in 'glupy.py'.
-
-# It is similar to the C language debug-trace translator, which logs the
-# arguments passed to the fops and their corresponding cbk functions.
-
-dl.get_id.restype = c_long
-dl.get_id.argtypes = [ POINTER(call_frame_t) ]
-
-dl.get_rootunique.restype = c_uint64
-dl.get_rootunique.argtypes = [ POINTER(call_frame_t) ]
-
-def uuid2str (gfid):
- return str(UUID(''.join(map("{0:02x}".format, gfid))))
-
-
-def st_mode_from_ia (prot, filetype):
- st_mode = 0
- type_bit = 0
- prot_bit = 0
-
- if filetype == IA_IFREG:
- type_bit = stat.S_IFREG
- elif filetype == IA_IFDIR:
- type_bit = stat.S_IFDIR
- elif filetype == IA_IFLNK:
- type_bit = stat.S_IFLNK
- elif filetype == IA_IFBLK:
- type_bit = stat.S_IFBLK
- elif filetype == IA_IFCHR:
- type_bit = stat.S_IFCHR
- elif filetype == IA_IFIFO:
- type_bit = stat.S_IFIFO
- elif filetype == IA_IFSOCK:
- type_bit = stat.S_IFSOCK
- elif filetype == IA_INVAL:
- pass
-
-
- if prot.suid:
- prot_bit |= stat.S_ISUID
- if prot.sgid:
- prot_bit |= stat.S_ISGID
- if prot.sticky:
- prot_bit |= stat.S_ISVTX
-
- if prot.owner.read:
- prot_bit |= stat.S_IRUSR
- if prot.owner.write:
- prot_bit |= stat.S_IWUSR
- if prot.owner.execn:
- prot_bit |= stat.S_IXUSR
-
- if prot.group.read:
- prot_bit |= stat.S_IRGRP
- if prot.group.write:
- prot_bit |= stat.S_IWGRP
- if prot.group.execn:
- prot_bit |= stat.S_IXGRP
-
- if prot.other.read:
- prot_bit |= stat.S_IROTH
- if prot.other.write:
- prot_bit |= stat.S_IWOTH
- if prot.other.execn:
- prot_bit |= stat.S_IXOTH
-
- st_mode = (type_bit | prot_bit)
-
- return st_mode
-
-
-def trace_stat2str (buf):
- gfid = uuid2str(buf.contents.ia_gfid)
- mode = st_mode_from_ia(buf.contents.ia_prot, buf.contents.ia_type)
- atime_buf = strftime("[%b %d %H:%M:%S]",
- localtime(buf.contents.ia_atime))
- mtime_buf = strftime("[%b %d %H:%M:%S]",
- localtime(buf.contents.ia_mtime))
- ctime_buf = strftime("[%b %d %H:%M:%S]",
- localtime(buf.contents.ia_ctime))
- return ("(gfid={0:s}, ino={1:d}, mode={2:o}, nlink={3:d}, uid ={4:d}, "+
- "gid ={5:d}, size={6:d}, blocks={7:d}, atime={8:s}, mtime={9:s}, "+
- "ctime={10:s})").format(gfid, buf.contents.ia_no, mode,
- buf.contents.ia_nlink,
- buf.contents.ia_uid,
- buf.contents.ia_gid,
- buf.contents.ia_size,
- buf.contents.ia_blocks,
- atime_buf, mtime_buf,
- ctime_buf)
-
-class xlator(Translator):
-
- def __init__(self, c_this):
- Translator.__init__(self, c_this)
- self.gfids = {}
-
- def lookup_fop(self, frame, this, loc, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(loc.contents.gfid)
- print(("GLUPY TRACE LOOKUP FOP- {0:d}: gfid={1:s}; " +
- "path={2:s}").format(unique, gfid, loc.contents.path))
- self.gfids[key] = gfid
- dl.wind_lookup(frame, POINTER(xlator_t)(), loc, xdata)
- return 0
-
- def lookup_cbk(self, frame, cookie, this, op_ret, op_errno,
- inode, buf, xdata, postparent):
- unique =dl.get_rootunique(frame)
- key =dl.get_id(frame)
- if op_ret == 0:
- gfid = uuid2str(buf.contents.ia_gfid)
- statstr = trace_stat2str(buf)
- postparentstr = trace_stat2str(postparent)
- print(("GLUPY TRACE LOOKUP CBK- {0:d}: gfid={1:s}; "+
- "op_ret={2:d}; *buf={3:s}; " +
- "*postparent={4:s}").format(unique, gfid,
- op_ret, statstr,
- postparentstr))
- else:
- gfid = self.gfids[key]
- print(("GLUPY TRACE LOOKUP CBK - {0:d}: gfid={1:s};" +
- " op_ret={2:d}; op_errno={3:d}").format(unique,
- gfid,
- op_ret,
- op_errno))
- del self.gfids[key]
- dl.unwind_lookup(frame, cookie, this, op_ret, op_errno,
- inode, buf, xdata, postparent)
- return 0
-
- def create_fop(self, frame, this, loc, flags, mode, umask, fd,
- xdata):
- unique = dl.get_rootunique(frame)
- gfid = uuid2str(loc.contents.gfid)
- print(("GLUPY TRACE CREATE FOP- {0:d}: gfid={1:s}; path={2:s}; " +
- "fd={3:s}; flags=0{4:o}; mode=0{5:o}; " +
- "umask=0{6:o}").format(unique, gfid, loc.contents.path,
- fd, flags, mode, umask))
- dl.wind_create(frame, POINTER(xlator_t)(), loc, flags,mode,
- umask, fd, xdata)
- return 0
-
- def create_cbk(self, frame, cookie, this, op_ret, op_errno, fd,
- inode, buf, preparent, postparent, xdata):
- unique = dl.get_rootunique(frame)
- if op_ret >= 0:
- gfid = uuid2str(inode.contents.gfid)
- statstr = trace_stat2str(buf)
- preparentstr = trace_stat2str(preparent)
- postparentstr = trace_stat2str(postparent)
- print(("GLUPY TRACE CREATE CBK- {0:d}: gfid={1:s};" +
- " op_ret={2:d}; fd={3:s}; *stbuf={4:s}; " +
- "*preparent={5:s};" +
- " *postparent={6:s}").format(unique, gfid, op_ret,
- fd, statstr,
- preparentstr,
- postparentstr))
- else:
- print(("GLUPY TRACE CREATE CBK- {0:d}: op_ret={1:d}; " +
- "op_errno={2:d}").format(unique, op_ret, op_errno))
- dl.unwind_create(frame, cookie, this, op_ret, op_errno, fd,
- inode, buf, preparent, postparent, xdata)
- return 0
-
- def open_fop(self, frame, this, loc, flags, fd, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(loc.contents.inode.contents.gfid)
- print(("GLUPY TRACE OPEN FOP- {0:d}: gfid={1:s}; path={2:s}; "+
- "flags={3:d}; fd={4:s}").format(unique, gfid,
- loc.contents.path, flags,
- fd))
- self.gfids[key] = gfid
- dl.wind_open(frame, POINTER(xlator_t)(), loc, flags, fd, xdata)
- return 0
-
- def open_cbk(self, frame, cookie, this, op_ret, op_errno, fd, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- print(("GLUPY TRACE OPEN CBK- {0:d}: gfid={1:s}; op_ret={2:d}; "
- "op_errno={3:d}; *fd={4:s}").format(unique, gfid,
- op_ret, op_errno, fd))
- del self.gfids[key]
- dl.unwind_open(frame, cookie, this, op_ret, op_errno, fd,
- xdata)
- return 0
-
- def readv_fop(self, frame, this, fd, size, offset, flags, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(fd.contents.inode.contents.gfid)
- print(("GLUPY TRACE READV FOP- {0:d}: gfid={1:s}; "+
- "fd={2:s}; size ={3:d}; offset={4:d}; " +
- "flags=0{5:x}").format(unique, gfid, fd, size, offset,
- flags))
- self.gfids[key] = gfid
- dl.wind_readv (frame, POINTER(xlator_t)(), fd, size, offset,
- flags, xdata)
- return 0
-
- def readv_cbk(self, frame, cookie, this, op_ret, op_errno, vector,
- count, buf, iobref, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- if op_ret >= 0:
- statstr = trace_stat2str(buf)
- print(("GLUPY TRACE READV CBK- {0:d}: gfid={1:s}, "+
- "op_ret={2:d}; *buf={3:s};").format(unique, gfid,
- op_ret,
- statstr))
-
- else:
- print(("GLUPY TRACE READV CBK- {0:d}: gfid={1:s}, "+
- "op_ret={2:d}; op_errno={3:d}").format(unique,
- gfid,
- op_ret,
- op_errno))
- del self.gfids[key]
- dl.unwind_readv (frame, cookie, this, op_ret, op_errno,
- vector, count, buf, iobref, xdata)
- return 0
-
- def writev_fop(self, frame, this, fd, vector, count, offset, flags,
- iobref, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(fd.contents.inode.contents.gfid)
- print(("GLUPY TRACE WRITEV FOP- {0:d}: gfid={1:s}; " +
- "fd={2:s}; count={3:d}; offset={4:d}; " +
- "flags=0{5:x}").format(unique, gfid, fd, count, offset,
- flags))
- self.gfids[key] = gfid
- dl.wind_writev(frame, POINTER(xlator_t)(), fd, vector, count,
- offset, flags, iobref, xdata)
- return 0
-
- def writev_cbk(self, frame, cookie, this, op_ret, op_errno, prebuf,
- postbuf, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- if op_ret >= 0:
- preopstr = trace_stat2str(prebuf)
- postopstr = trace_stat2str(postbuf)
- print(("GLUPY TRACE WRITEV CBK- {0:d}: op_ret={1:d}; " +
- "*prebuf={2:s}; " +
- "*postbuf={3:s}").format(unique, op_ret, preopstr,
- postopstr))
- else:
- gfid = self.gfids[key]
- print(("GLUPY TRACE WRITEV CBK- {0:d}: gfid={1:s}; "+
- "op_ret={2:d}; op_errno={3:d}").format(unique,
- gfid,
- op_ret,
- op_errno))
- del self.gfids[key]
- dl.unwind_writev (frame, cookie, this, op_ret, op_errno,
- prebuf, postbuf, xdata)
- return 0
-
- def opendir_fop(self, frame, this, loc, fd, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(loc.contents.inode.contents.gfid)
- print(("GLUPY TRACE OPENDIR FOP- {0:d}: gfid={1:s}; path={2:s}; "+
- "fd={3:s}").format(unique, gfid, loc.contents.path, fd))
- self.gfids[key] = gfid
- dl.wind_opendir(frame, POINTER(xlator_t)(), loc, fd, xdata)
- return 0
-
- def opendir_cbk(self, frame, cookie, this, op_ret, op_errno, fd,
- xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- print(("GLUPY TRACE OPENDIR CBK- {0:d}: gfid={1:s}; op_ret={2:d};"+
- " op_errno={3:d}; fd={4:s}").format(unique, gfid, op_ret,
- op_errno, fd))
- del self.gfids[key]
- dl.unwind_opendir(frame, cookie, this, op_ret, op_errno,
- fd, xdata)
- return 0
-
- def readdir_fop(self, frame, this, fd, size, offset, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(fd.contents.inode.contents.gfid)
- print(("GLUPY TRACE READDIR FOP- {0:d}: gfid={1:s}; fd={2:s}; " +
- "size={3:d}; offset={4:d}").format(unique, gfid, fd, size,
- offset))
- self.gfids[key] = gfid
- dl.wind_readdir(frame, POINTER(xlator_t)(), fd, size, offset,
- xdata)
- return 0
-
- def readdir_cbk(self, frame, cookie, this, op_ret, op_errno, buf,
- xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- print(("GLUPY TRACE READDIR CBK- {0:d}: gfid={1:s}; op_ret={2:d};"+
- " op_errno={3:d}").format(unique, gfid, op_ret, op_errno))
- del self.gfids[key]
- dl.unwind_readdir(frame, cookie, this, op_ret, op_errno, buf,
- xdata)
- return 0
-
- def readdirp_fop(self, frame, this, fd, size, offset, dictionary):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(fd.contents.inode.contents.gfid)
- print(("GLUPY TRACE READDIRP FOP- {0:d}: gfid={1:s}; fd={2:s}; "+
- " size={3:d}; offset={4:d}").format(unique, gfid, fd, size,
- offset))
- self.gfids[key] = gfid
- dl.wind_readdirp(frame, POINTER(xlator_t)(), fd, size, offset,
- dictionary)
- return 0
-
- def readdirp_cbk(self, frame, cookie, this, op_ret, op_errno, buf,
- xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- print(("GLUPY TRACE READDIRP CBK- {0:d}: gfid={1:s}; "+
- "op_ret={2:d}; op_errno={3:d}").format(unique, gfid,
- op_ret, op_errno))
- del self.gfids[key]
- dl.unwind_readdirp(frame, cookie, this, op_ret, op_errno, buf,
- xdata)
- return 0
-
- def mkdir_fop(self, frame, this, loc, mode, umask, xdata):
- unique = dl.get_rootunique(frame)
- gfid = uuid2str(loc.contents.inode.contents.gfid)
- print(("GLUPY TRACE MKDIR FOP- {0:d}: gfid={1:s}; path={2:s}; " +
- "mode={3:d}; umask=0{4:o}").format(unique, gfid,
- loc.contents.path, mode,
- umask))
- dl.wind_mkdir(frame, POINTER(xlator_t)(), loc, mode, umask,
- xdata)
- return 0
-
- def mkdir_cbk(self, frame, cookie, this, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata):
- unique = dl.get_rootunique(frame)
- if op_ret == 0:
- gfid = uuid2str(inode.contents.gfid)
- statstr = trace_stat2str(buf)
- preparentstr = trace_stat2str(preparent)
- postparentstr = trace_stat2str(postparent)
- print(("GLUPY TRACE MKDIR CBK- {0:d}: gfid={1:s}; "+
- "op_ret={2:d}; *stbuf={3:s}; *prebuf={4:s}; "+
- "*postbuf={5:s} ").format(unique, gfid, op_ret,
- statstr,
- preparentstr,
- postparentstr))
- else:
- print(("GLUPY TRACE MKDIR CBK- {0:d}: op_ret={1:d}; "+
- "op_errno={2:d}").format(unique, op_ret, op_errno))
- dl.unwind_mkdir(frame, cookie, this, op_ret, op_errno, inode,
- buf, preparent, postparent, xdata)
- return 0
-
- def rmdir_fop(self, frame, this, loc, flags, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(loc.contents.inode.contents.gfid)
- print(("GLUPY TRACE RMDIR FOP- {0:d}: gfid={1:s}; path={2:s}; "+
- "flags={3:d}").format(unique, gfid, loc.contents.path,
- flags))
- self.gfids[key] = gfid
- dl.wind_rmdir(frame, POINTER(xlator_t)(), loc, flags, xdata)
- return 0
-
- def rmdir_cbk(self, frame, cookie, this, op_ret, op_errno, preparent,
- postparent, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- if op_ret == 0:
- preparentstr = trace_stat2str(preparent)
- postparentstr = trace_stat2str(postparent)
- print(("GLUPY TRACE RMDIR CBK- {0:d}: gfid={1:s}; "+
- "op_ret={2:d}; *prebuf={3:s}; "+
- "*postbuf={4:s}").format(unique, gfid, op_ret,
- preparentstr,
- postparentstr))
- else:
- print(("GLUPY TRACE RMDIR CBK- {0:d}: gfid={1:s}; "+
- "op_ret={2:d}; op_errno={3:d}").format(unique,
- gfid,
- op_ret,
- op_errno))
- del self.gfids[key]
- dl.unwind_rmdir(frame, cookie, this, op_ret, op_errno,
- preparent, postparent, xdata)
- return 0
-
- def stat_fop(self, frame, this, loc, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(loc.contents.inode.contents.gfid)
- print(("GLUPY TRACE STAT FOP- {0:d}: gfid={1:s}; " +
- " path={2:s}").format(unique, gfid, loc.contents.path))
- self.gfids[key] = gfid
- dl.wind_stat(frame, POINTER(xlator_t)(), loc, xdata)
- return 0
-
- def stat_cbk(self, frame, cookie, this, op_ret, op_errno, buf,
- xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- if op_ret == 0:
- statstr = trace_stat2str(buf)
- print(("GLUPY TRACE STAT CBK- {0:d}: gfid={1:s}; "+
- "op_ret={2:d}; *buf={3:s};").format(unique,
- gfid,
- op_ret,
- statstr))
- else:
- print(("GLUPY TRACE STAT CBK- {0:d}: gfid={1:s}; "+
- "op_ret={2:d}; op_errno={3:d}").format(unique,
- gfid,
- op_ret,
- op_errno))
- del self.gfids[key]
- dl.unwind_stat(frame, cookie, this, op_ret, op_errno,
- buf, xdata)
- return 0
-
- def fstat_fop(self, frame, this, fd, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(fd.contents.inode.contents.gfid)
- print(("GLUPY TRACE FSTAT FOP- {0:d}: gfid={1:s}; " +
- "fd={2:s}").format(unique, gfid, fd))
- self.gfids[key] = gfid
- dl.wind_fstat(frame, POINTER(xlator_t)(), fd, xdata)
- return 0
-
- def fstat_cbk(self, frame, cookie, this, op_ret, op_errno, buf,
- xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- if op_ret == 0:
- statstr = trace_stat2str(buf)
- print(("GLUPY TRACE FSTAT CBK- {0:d}: gfid={1:s} "+
- " op_ret={2:d}; *buf={3:s}").format(unique,
- gfid,
- op_ret,
- statstr))
- else:
- print(("GLUPY TRACE FSTAT CBK- {0:d}: gfid={1:s} "+
- "op_ret={2:d}; op_errno={3:d}").format(unique.
- gfid,
- op_ret,
- op_errno))
- del self.gfids[key]
- dl.unwind_fstat(frame, cookie, this, op_ret, op_errno,
- buf, xdata)
- return 0
-
- def statfs_fop(self, frame, this, loc, xdata):
- unique = dl.get_rootunique(frame)
- if loc.contents.inode:
- gfid = uuid2str(loc.contents.inode.contents.gfid)
- else:
- gfid = "0"
- print(("GLUPY TRACE STATFS FOP- {0:d}: gfid={1:s}; "+
- "path={2:s}").format(unique, gfid, loc.contents.path))
- dl.wind_statfs(frame, POINTER(xlator_t)(), loc, xdata)
- return 0
-
- def statfs_cbk(self, frame, cookie, this, op_ret, op_errno, buf,
- xdata):
- unique = dl.get_rootunique(frame)
- if op_ret == 0:
- #TBD: print buf (pointer to an iovec type object)
- print(("GLUPY TRACE STATFS CBK {0:d}: "+
- "op_ret={1:d}").format(unique, op_ret))
- else:
- print(("GLUPY TRACE STATFS CBK- {0:d}"+
- "op_ret={1:d}; op_errno={2:d}").format(unique,
- op_ret,
- op_errno))
- dl.unwind_statfs(frame, cookie, this, op_ret, op_errno,
- buf, xdata)
- return 0
-
- def getxattr_fop(self, frame, this, loc, name, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(loc.contents.inode.contents.gfid)
- print(("GLUPY TRACE GETXATTR FOP- {0:d}: gfid={1:s}; path={2:s};"+
- " name={3:s}").format(unique, gfid, loc.contents.path,
- name))
- self.gfids[key]=gfid
- dl.wind_getxattr(frame, POINTER(xlator_t)(), loc, name, xdata)
- return 0
-
- def getxattr_cbk(self, frame, cookie, this, op_ret, op_errno,
- dictionary, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- print(("GLUPY TRACE GETXATTR CBK- {0:d}: gfid={1:s}; "+
- "op_ret={2:d}; op_errno={3:d}; "+
- " dictionary={4:s}").format(unique, gfid, op_ret, op_errno,
- dictionary))
- del self.gfids[key]
- dl.unwind_getxattr(frame, cookie, this, op_ret, op_errno,
- dictionary, xdata)
- return 0
-
- def fgetxattr_fop(self, frame, this, fd, name, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(fd.contents.inode.contents.gfid)
- print(("GLUPY TRACE FGETXATTR FOP- {0:d}: gfid={1:s}; fd={2:s}; "+
- "name={3:s}").format(unique, gfid, fd, name))
- self.gfids[key] = gfid
- dl.wind_fgetxattr(frame, POINTER(xlator_t)(), fd, name, xdata)
- return 0
-
- def fgetxattr_cbk(self, frame, cookie, this, op_ret, op_errno,
- dictionary, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- print(("GLUPY TRACE FGETXATTR CBK- {0:d}: gfid={1:s}; "+
- "op_ret={2:d}; op_errno={3:d};"+
- " dictionary={4:s}").format(unique, gfid, op_ret,
- op_errno, dictionary))
- del self.gfids[key]
- dl.unwind_fgetxattr(frame, cookie, this, op_ret, op_errno,
- dictionary, xdata)
- return 0
-
- def setxattr_fop(self, frame, this, loc, dictionary, flags, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(loc.contents.inode.contents.gfid)
- print(("GLUPY TRACE SETXATTR FOP- {0:d}: gfid={1:s}; path={2:s};"+
- " flags={3:d}").format(unique, gfid, loc.contents.path,
- flags))
- self.gfids[key] = gfid
- dl.wind_setxattr(frame, POINTER(xlator_t)(), loc, dictionary,
- flags, xdata)
- return 0
-
- def setxattr_cbk(self, frame, cookie, this, op_ret, op_errno, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- print(("GLUPY TRACE SETXATTR CBK- {0:d}: gfid={1:s}; "+
- "op_ret={2:d}; op_errno={3:d}").format(unique, gfid,
- op_ret, op_errno))
- del self.gfids[key]
- dl.unwind_setxattr(frame, cookie, this, op_ret, op_errno,
- xdata)
- return 0
-
- def fsetxattr_fop(self, frame, this, fd, dictionary, flags, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(fd.contents.inode.contents.gfid)
- print(("GLUPY TRACE FSETXATTR FOP- {0:d}: gfid={1:s}; fd={2:p}; "+
- "flags={3:d}").format(unique, gfid, fd, flags))
- self.gfids[key] = gfid
- dl.wind_fsetxattr(frame, POINTER(xlator_t)(), fd, dictionary,
- flags, xdata)
- return 0
-
- def fsetxattr_cbk(self, frame, cookie, this, op_ret, op_errno, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- print(("GLUPY TRACE FSETXATTR CBK- {0:d}: gfid={1:s}; "+
- "op_ret={2:d}; op_errno={3:d}").format(unique, gfid,
- op_ret, op_errno))
- del self.gfids[key]
- dl.unwind_fsetxattr(frame, cookie, this, op_ret, op_errno,
- xdata)
- return 0
-
- def removexattr_fop(self, frame, this, loc, name, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(loc.contents.inode.contents.gfid)
- print(("GLUPY TRACE REMOVEXATTR FOP- {0:d}: gfid={1:s}; "+
- "path={2:s}; name={3:s}").format(unique, gfid,
- loc.contents.path,
- name))
- self.gfids[key] = gfid
- dl.wind_removexattr(frame, POINTER(xlator_t)(), loc, name,
- xdata)
- return 0
-
- def removexattr_cbk(self, frame, cookie, this, op_ret, op_errno,
- xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- print(("GLUPY TRACE REMOVEXATTR CBK- {0:d}: gfid={1:s} "+
- " op_ret={2:d}; op_errno={3:d}").format(unique, gfid,
- op_ret, op_errno))
- del self.gfids[key]
- dl.unwind_removexattr(frame, cookie, this, op_ret, op_errno,
- xdata)
- return 0
-
- def link_fop(self, frame, this, oldloc, newloc, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- if (newloc.contents.inode):
- newgfid = uuid2str(newloc.contents.inode.contents.gfid)
- else:
- newgfid = "0"
- oldgfid = uuid2str(oldloc.contents.inode.contents.gfid)
- print(("GLUPY TRACE LINK FOP-{0:d}: oldgfid={1:s}; oldpath={2:s};"+
- "newgfid={3:s};"+
- "newpath={4:s}").format(unique, oldgfid,
- oldloc.contents.path,
- newgfid,
- newloc.contents.path))
- self.gfids[key] = oldgfid
- dl.wind_link(frame, POINTER(xlator_t)(), oldloc, newloc,
- xdata)
- return 0
-
- def link_cbk(self, frame, cookie, this, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- if op_ret == 0:
- statstr = trace_stat2str(buf)
- preparentstr = trace_stat2str(preparent)
- postparentstr = trace_stat2str(postparent)
- print(("GLUPY TRACE LINK CBK- {0:d}: op_ret={1:d} "+
- "*stbuf={2:s}; *prebuf={3:s}; "+
- "*postbuf={4:s} ").format(unique, op_ret, statstr,
- preparentstr,
- postparentstr))
- else:
- print(("GLUPY TRACE LINK CBK- {0:d}: gfid={1:s}; "+
- "op_ret={2:d}; "+
- "op_errno={3:d}").format(unique, gfid,
- op_ret, op_errno))
- del self.gfids[key]
- dl.unwind_link(frame, cookie, this, op_ret, op_errno, inode,
- buf, preparent, postparent, xdata)
- return 0
-
- def unlink_fop(self, frame, this, loc, xflag, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(loc.contents.inode.contents.gfid)
- print(("GLUPY TRACE UNLINK FOP- {0:d}; gfid={1:s}; path={2:s}; "+
- "flag={3:d}").format(unique, gfid, loc.contents.path,
- xflag))
- self.gfids[key] = gfid
- dl.wind_unlink(frame, POINTER(xlator_t)(), loc, xflag,
- xdata)
- return 0
-
- def unlink_cbk(self, frame, cookie, this, op_ret, op_errno,
- preparent, postparent, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- if op_ret == 0:
- preparentstr = trace_stat2str(preparent)
- postparentstr = trace_stat2str(postparent)
- print(("GLUPY TRACE UNLINK CBK- {0:d}: gfid ={1:s}; "+
- "op_ret={2:d}; *prebuf={3:s}; "+
- "*postbuf={4:s} ").format(unique, gfid, op_ret,
- preparentstr,
- postparentstr))
- else:
- print(("GLUPY TRACE UNLINK CBK: {0:d}: gfid ={1:s}; "+
- "op_ret={2:d}; "+
- "op_errno={3:d}").format(unique, gfid, op_ret,
- op_errno))
- del self.gfids[key]
- dl.unwind_unlink(frame, cookie, this, op_ret, op_errno,
- preparent, postparent, xdata)
- return 0
-
- def readlink_fop(self, frame, this, loc, size, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(loc.contents.inode.contents.gfid)
- print(("GLUPY TRACE READLINK FOP- {0:d}: gfid={1:s}; path={2:s};"+
- " size={3:d}").format(unique, gfid, loc.contents.path,
- size))
- self.gfids[key] = gfid
- dl.wind_readlink(frame, POINTER(xlator_t)(), loc, size,
- xdata)
- return 0
-
- def readlink_cbk(self, frame, cookie, this, op_ret, op_errno,
- buf, stbuf, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- if op_ret == 0:
- statstr = trace_stat2str(stbuf)
- print(("GLUPY TRACE READLINK CBK- {0:d}: gfid={1:s} "+
- " op_ret={2:d}; op_errno={3:d}; *prebuf={4:s}; "+
- "*postbuf={5:s} ").format(unique, gfid,
- op_ret, op_errno,
- buf, statstr))
- else:
- print(("GLUPY TRACE READLINK CBK- {0:d}: gfid={1:s} "+
- " op_ret={2:d}; op_errno={3:d}").format(unique,
- gfid,
- op_ret,
- op_errno))
- del self.gfids[key]
- dl.unwind_readlink(frame, cookie, this, op_ret, op_errno, buf,
- stbuf, xdata)
- return 0
-
- def symlink_fop(self, frame, this, linkpath, loc, umask, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = uuid2str(loc.contents.inode.contents.gfid)
- print(("GLUPY TRACE SYMLINK FOP- {0:d}: gfid={1:s}; "+
- "linkpath={2:s}; path={3:s};"+
- "umask=0{4:o}").format(unique, gfid, linkpath,
- loc.contents.path, umask))
- self.gfids[key] = gfid
- dl.wind_symlink(frame, POINTER(xlator_t)(), linkpath, loc,
- umask, xdata)
- return 0
-
- def symlink_cbk(self, frame, cookie, this, op_ret, op_errno,
- inode, buf, preparent, postparent, xdata):
- unique = dl.get_rootunique(frame)
- key = dl.get_id(frame)
- gfid = self.gfids[key]
- if op_ret == 0:
- statstr = trace_stat2str(buf)
- preparentstr = trace_stat2str(preparent)
- postparentstr = trace_stat2str(postparent)
- print(("GLUPY TRACE SYMLINK CBK- {0:d}: gfid={1:s}; "+
- "op_ret={2:d}; *stbuf={3:s}; *preparent={4:s}; "+
- "*postparent={5:s}").format(unique, gfid,
- op_ret, statstr,
- preparentstr,
- postparentstr))
- else:
- print(("GLUPY TRACE SYMLINK CBK- {0:d}: gfid={1:s}; "+
- "op_ret={2:d}; op_errno={3:d}").format(unique,
- gfid,
- op_ret,
- op_errno))
- del self.gfids[key]
- dl.unwind_symlink(frame, cookie, this, op_ret, op_errno,
- inode, buf, preparent, postparent, xdata)
- return 0
diff --git a/xlators/features/glupy/examples/helloworld.py b/xlators/features/glupy/examples/helloworld.py
deleted file mode 100644
index 282f9207949..00000000000
--- a/xlators/features/glupy/examples/helloworld.py
+++ /dev/null
@@ -1,21 +0,0 @@
-
-from __future__ import print_function
-import sys
-from gluster.glupy import *
-
-class xlator (Translator):
-
- def __init__(self, c_this):
- Translator.__init__(self, c_this)
-
- def lookup_fop(self, frame, this, loc, xdata):
- print("Python xlator: Hello!")
- dl.wind_lookup(frame, POINTER(xlator_t)(), loc, xdata)
- return 0
-
- def lookup_cbk(self, frame, cookie, this, op_ret, op_errno, inode, buf,
- xdata, postparent):
- print("Python xlator: Hello again!")
- dl.unwind_lookup(frame, cookie, this, op_ret, op_errno, inode, buf,
- xdata, postparent)
- return 0
diff --git a/xlators/features/glupy/examples/negative.py b/xlators/features/glupy/examples/negative.py
deleted file mode 100644
index e44ff4deed2..00000000000
--- a/xlators/features/glupy/examples/negative.py
+++ /dev/null
@@ -1,93 +0,0 @@
-
-from __future__ import print_function
-import sys
-from uuid import UUID
-from gluster.glupy import *
-
-# Negative-lookup-caching example. If a file wasn't there the last time we
-# looked, it's probably still not there. This translator keeps track of
-# those failed lookups for us, and returns ENOENT without needing to pass the
-# call any further for repeated requests.
-
-# If we were doing this for real, we'd need separate caches for each xlator
-# instance. The easiest way to do this would be to have xlator.__init__
-# "register" each instance in a module-global dict, with the key as the C
-# translator address and the value as the xlator object itself. For testing
-# and teaching, it's sufficient just to have one cache. The keys are parent
-# GFIDs, and the entries are lists of names within that parent that we know
-# don't exist.
-cache = {}
-
-# TBD: we need a better way of handling per-request data (frame->local in C).
-dl.get_id.restype = c_long
-dl.get_id.argtypes = [ POINTER(call_frame_t) ]
-
-def uuid2str (gfid):
- return str(UUID(''.join(map("{0:02x}".format, gfid))))
-
-class xlator (Translator):
-
- def __init__ (self, c_this):
- self.requests = {}
- Translator.__init__(self,c_this)
-
- def lookup_fop (self, frame, this, loc, xdata):
- pargfid = uuid2str(loc.contents.pargfid)
- print("lookup FOP: %s:%s" % (pargfid, loc.contents.name))
- # Check the cache.
- if cache.has_key(pargfid):
- if loc.contents.name in cache[pargfid]:
- print("short-circuiting for %s:%s" % (pargfid,
- loc.contents.name))
- dl.unwind_lookup(frame,0,this,-1,2,None,None,None,None)
- return 0
- key = dl.get_id(frame)
- self.requests[key] = (pargfid, loc.contents.name[:])
- # TBD: get real child xl from init, pass it here
- dl.wind_lookup(frame,POINTER(xlator_t)(),loc,xdata)
- return 0
-
- def lookup_cbk (self, frame, cookie, this, op_ret, op_errno, inode, buf,
- xdata, postparent):
- print("lookup CBK: %d (%d)" % (op_ret, op_errno))
- key = dl.get_id(frame)
- pargfid, name = self.requests[key]
- # Update the cache.
- if op_ret == 0:
- print("found %s, removing from cache" % name)
- if cache.has_key(pargfid):
- cache[pargfid].discard(name)
- elif op_errno == 2: # ENOENT
- print("failed to find %s, adding to cache" % name)
- if cache.has_key(pargfid):
- cache[pargfid].add(name)
- else:
- cache[pargfid] = set([name])
- del self.requests[key]
- dl.unwind_lookup(frame,cookie,this,op_ret,op_errno,
- inode,buf,xdata,postparent)
- return 0
-
- def create_fop (self, frame, this, loc, flags, mode, umask, fd, xdata):
- pargfid = uuid2str(loc.contents.pargfid)
- print("create FOP: %s:%s" % (pargfid, loc.contents.name))
- key = dl.get_id(frame)
- self.requests[key] = (pargfid, loc.contents.name[:])
- # TBD: get real child xl from init, pass it here
- dl.wind_create(frame,POINTER(xlator_t)(),loc,flags,mode,umask,fd,xdata)
- return 0
-
- def create_cbk (self, frame, cookie, this, op_ret, op_errno, fd, inode,
- buf, preparent, postparent, xdata):
- print("create CBK: %d (%d)" % (op_ret, op_errno))
- key = dl.get_id(frame)
- pargfid, name = self.requests[key]
- # Update the cache.
- if op_ret == 0:
- print("created %s, removing from cache" % name)
- if cache.has_key(pargfid):
- cache[pargfid].discard(name)
- del self.requests[key]
- dl.unwind_create(frame,cookie,this,op_ret,op_errno,fd,inode,buf,
- preparent,postparent,xdata)
- return 0
diff --git a/xlators/features/glupy/src/Makefile.am b/xlators/features/glupy/src/Makefile.am
deleted file mode 100644
index f1802c3545e..00000000000
--- a/xlators/features/glupy/src/Makefile.am
+++ /dev/null
@@ -1,37 +0,0 @@
-xlator_LTLIBRARIES = glupy.la
-
-# Ensure GLUSTER_PYTHON_PATH is passed to glupy.so
-xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
-glupydir = $(xlatordir)/glupy
-
-AM_CPPFLAGS = $(PYTHONDEV_CPPFLAGS) $(GF_CPPFLAGS) \
- -I$(top_srcdir)/libglusterfs/src \
- -I$(top_srcdir)/rpc/xdr/src -I$(top_builddir)/rpc/xdr/src \
- -isystem $(BUILD_PYTHON_INC)
-
-AM_CFLAGS = $(PYTHONDEV_CPPFLAGS) -Wall -fno-strict-aliasing \
- -DGLUSTER_PYTHON_PATH=\"$(glupydir)\" \
- -DPATH_GLUSTERFS_GLUPY_MODULE=\"${xlatordir}/glupy${shrext_cmds}\" \
- $(GF_CFLAGS)
-
-# Flags to build glupy.so with
-glupy_la_LDFLAGS = $(PYTHONDEV_LDFLAGS) -module -avoid-version \
- -nostartfiles -export-symbols \
- $(top_srcdir)/xlators/features/glupy/src/glupy.sym
-
-glupy_la_SOURCES = glupy.c
-glupy_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la \
- -lpthread -l$(BUILD_PYTHON_LIB)
-
-noinst_HEADERS = glupy.h
-
-# Install __init__.py into the Python site-packages area
-pyglupydir = @BUILD_PYTHON_SITE_PACKAGES@/gluster
-pyglupy_PYTHON = __init__.py
-
-# Install glupy/__init_-.py into the Python site-packages area
-SUBDIRS = glupy
-
-CLEANFILES =
-
-EXTRA_DIST = glupy.sym
diff --git a/xlators/features/glupy/src/__init__.py.in b/xlators/features/glupy/src/__init__.py.in
deleted file mode 100644
index 2a66079de12..00000000000
--- a/xlators/features/glupy/src/__init__.py.in
+++ /dev/null
@@ -1,4 +0,0 @@
-#!/usr/bin/python2
-
-from pkgutil import extend_path
-__path__ = extend_path(__path__, __name__)
diff --git a/xlators/features/glupy/src/glupy.c b/xlators/features/glupy/src/glupy.c
deleted file mode 100644
index 46d1af5533a..00000000000
--- a/xlators/features/glupy/src/glupy.c
+++ /dev/null
@@ -1,2515 +0,0 @@
-/*
- Copyright (c) 2006-2014 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#include <ctype.h>
-#include <dlfcn.h>
-#include <sys/uio.h>
-#include <Python.h>
-
-#include "glusterfs.h"
-#include "xlator.h"
-#include "logging.h"
-#include "defaults.h"
-
-#include "glupy.h"
-
-/* UTILITY FUNCTIONS FOR FOP-SPECIFIC CODE */
-
-pthread_key_t gil_init_key;
-
-PyGILState_STATE
-glupy_enter (void)
-{
- if (!pthread_getspecific(gil_init_key)) {
- PyEval_ReleaseLock();
- (void)pthread_setspecific(gil_init_key,(void *)1);
- }
-
- return PyGILState_Ensure();
-}
-
-void
-glupy_leave (PyGILState_STATE gstate)
-{
- PyGILState_Release(gstate);
-}
-
-/* FOP: LOOKUP */
-
-int32_t
-glupy_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata, struct iatt *postparent)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_LOOKUP]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_lookup_cbk_t)(priv->cbks[GLUPY_LOOKUP]))(
- frame, cookie, this, op_ret, op_errno,
- inode, buf, xdata, postparent);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
- xdata, postparent);
- return 0;
-}
-
-int32_t
-glupy_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_LOOKUP]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_lookup_t)(priv->fops[GLUPY_LOOKUP]))(
- frame, this, loc, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_lookup_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, loc, xdata);
- return 0;
-}
-
-void
-wind_lookup (call_frame_t *frame, xlator_t *xl, loc_t *loc, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND(frame,glupy_lookup_cbk,xl,xl->fops->lookup,loc,xdata);
-}
-
-void
-unwind_lookup (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata, struct iatt *postparent)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT(lookup,frame,op_ret,op_errno,
- inode,buf,xdata,postparent);
-}
-
-void
-set_lookup_fop (long py_this, fop_lookup_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_LOOKUP] = (long)fop;
-}
-
-void
-set_lookup_cbk (long py_this, fop_lookup_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_LOOKUP] = (long)cbk;
-}
-
-/* FOP: CREATE */
-
-int32_t
-glupy_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_CREATE]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_create_cbk_t)(priv->cbks[GLUPY_CREATE]))(
- frame, cookie, this, op_ret, op_errno,
- fd, inode, buf, preparent, postparent, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
- preparent, postparent, xdata);
- return 0;
-}
-
-int32_t
-glupy_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_CREATE]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_create_t)(priv->fops[GLUPY_CREATE]))(
- frame, this, loc, flags, mode, umask, fd, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_create_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->create, loc, flags, mode, umask,
- fd, xdata);
- return 0;
-}
-
-void
-wind_create (call_frame_t *frame, xlator_t *xl, loc_t *loc, int32_t flags,
- mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND (frame, glupy_create_cbk,xl, xl->fops->create,
- loc, flags, mode, umask, fd, xdata);
-}
-
-void
-unwind_create (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
- preparent, postparent, xdata);
-}
-
-void
-set_create_fop (long py_this, fop_create_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_CREATE] = (long)fop;
-}
-
-void
-set_create_cbk (long py_this, fop_create_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_CREATE] = (long)cbk;
-}
-
-/* FOP: OPEN */
-
-int32_t
-glupy_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_OPEN]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_open_cbk_t)(priv->cbks[GLUPY_OPEN]))(
- frame, cookie, this, op_ret, op_errno,
- fd, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
- return 0;
-}
-
-int32_t
-glupy_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int32_t flags, fd_t *fd, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_OPEN]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_open_t)(priv->fops[GLUPY_OPEN]))(
- frame, this, loc, flags, fd, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_open_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
- return 0;
-}
-
-void
-wind_open (call_frame_t *frame, xlator_t *xl, loc_t *loc, int32_t flags,
- fd_t *fd, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND (frame, glupy_open_cbk, xl, xl->fops->open, loc, flags,
- fd, xdata);
-}
-
-void
-unwind_open (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
-}
-
-void
-set_open_fop (long py_this, fop_open_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
- priv->fops[GLUPY_OPEN] = (long)fop;
-}
-
-void
-set_open_cbk (long py_this, fop_open_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
- priv->cbks[GLUPY_OPEN] = (long)cbk;
-}
-
-/* FOP: READV */
-
-int32_t
-glupy_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iovec *vector,
- int32_t count, struct iatt *stbuf, struct iobref *iobref,
- dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_READV]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_readv_cbk_t)(priv->cbks[GLUPY_READV]))(
- frame, cookie, this, op_ret, op_errno,
- vector, count, stbuf, iobref, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector,
- count, stbuf, iobref, xdata);
- return 0;
-}
-
-int32_t
-glupy_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
- size_t size, off_t offset, uint32_t flags, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_READV]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_readv_t)(priv->fops[GLUPY_READV]))(
- frame, this, fd, size, offset, flags, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_readv_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readv, fd, size, offset,
- flags, xdata);
- return 0;
-}
-
-void
-wind_readv (call_frame_t *frame, xlator_t *xl, fd_t *fd, size_t size,
- off_t offset, uint32_t flags, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND (frame, glupy_readv_cbk, xl, xl->fops->readv, fd, size,
- offset, flags, xdata);
-}
-
-void
-unwind_readv (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iovec *vector,
- int32_t count, struct iatt *stbuf, struct iobref *iobref,
- dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector,
- count, stbuf, iobref, xdata);
-}
-
-void
-set_readv_fop (long py_this, fop_readv_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
- priv->fops[GLUPY_READV] = (long)fop;
-}
-
-void
-set_readv_cbk (long py_this, fop_readv_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
- priv->cbks[GLUPY_READV] = (long)cbk;
-}
-
-/* FOP: WRITEV */
-
-int32_t
-glupy_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_WRITEV]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_writev_cbk_t)(priv->cbks[GLUPY_WRITEV]))(
- frame, cookie, this, op_ret, op_errno,
- prebuf, postbuf, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
- return 0;
-}
-
-int32_t
-glupy_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count, off_t offset,
- uint32_t flags, struct iobref *iobref, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_WRITEV]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_writev_t)(priv->fops[GLUPY_WRITEV]))(
- frame, this, fd, vector, count, offset, flags,
- iobref, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_writev_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->writev, fd, vector, count,
- offset, flags, iobref, xdata);
- return 0;
-}
-
-void
-wind_writev (call_frame_t *frame, xlator_t *xl, fd_t *fd, struct iovec *vector,
- int32_t count, off_t offset, uint32_t flags, struct iobref *iobref,
- dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND (frame, glupy_writev_cbk, xl, xl->fops->writev, fd, vector,
- count, offset, flags, iobref, xdata);
-}
-
-void
-unwind_writev (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
-}
-
-void
-set_writev_fop (long py_this, fop_writev_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
- priv->fops[GLUPY_WRITEV] = (long)fop;
-}
-
-void
-set_writev_cbk (long py_this, fop_writev_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
- priv->cbks[GLUPY_WRITEV] = (long)cbk;
-}
-
-
-/* FOP: OPENDIR */
-
-int32_t
-glupy_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd,
- dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_OPENDIR]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_opendir_cbk_t)(priv->cbks[GLUPY_OPENDIR]))(
- frame, cookie, this, op_ret, op_errno,
- fd, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd, xdata);
- return 0;
-}
-
-int32_t
-glupy_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc,
- fd_t *fd, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_OPENDIR]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_opendir_t)(priv->fops[GLUPY_OPENDIR]))(
- frame, this, loc, fd, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_opendir_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->opendir, loc, fd, xdata);
- return 0;
-}
-
-void
-wind_opendir (call_frame_t *frame, xlator_t *xl, loc_t *loc, fd_t *fd, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND(frame,glupy_opendir_cbk,xl,xl->fops->opendir,loc,fd,xdata);
-}
-
-void
-unwind_opendir (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT(opendir,frame,op_ret,op_errno,
- fd,xdata);
-}
-
-void
-set_opendir_fop (long py_this, fop_opendir_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_OPENDIR] = (long)fop;
-}
-
-void
-set_opendir_cbk (long py_this, fop_opendir_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_OPENDIR] = (long)cbk;
-}
-
-/* FOP: READDIR */
-
-int32_t
-glupy_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
- dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_READDIR]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_readdir_cbk_t)(priv->cbks[GLUPY_READDIR]))(
- frame, cookie, this, op_ret, op_errno,
- entries, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, entries,
- xdata);
- return 0;
-}
-
-int32_t
-glupy_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
- size_t size, off_t offset, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_READDIR]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_readdir_t)(priv->fops[GLUPY_READDIR]))(
- frame, this, fd, size, offset, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_readdir_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdir,fd, size, offset, xdata);
- return 0;
-}
-
-void
-wind_readdir(call_frame_t *frame, xlator_t *xl, fd_t *fd, size_t size,
- off_t offset, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND(frame,glupy_readdir_cbk,xl,xl->fops->readdir,fd,size,offset,xdata);
-}
-
-void
-unwind_readdir (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
- dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT(readdir,frame,op_ret,op_errno,
- entries, xdata);
-}
-
-void
-set_readdir_fop (long py_this, fop_readdir_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_READDIR] = (long)fop;
-}
-
-void
-set_readdir_cbk (long py_this, fop_readdir_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_READDIR] = (long)cbk;
-}
-
-
-/* FOP: READDIRP */
-
-int32_t
-glupy_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
- dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_READDIRP]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_readdirp_cbk_t)(priv->cbks[GLUPY_READDIRP]))(
- frame, cookie, this, op_ret, op_errno,
- entries, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries,
- xdata);
- return 0;
-}
-
-int32_t
-glupy_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd,
- size_t size, off_t offset, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_READDIRP]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_readdirp_t)(priv->fops[GLUPY_READDIRP]))(
- frame, this, fd, size, offset, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_readdirp_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdirp,fd, size, offset, xdata);
- return 0;
-}
-
-void
-wind_readdirp (call_frame_t *frame, xlator_t *xl, fd_t *fd, size_t size,
- off_t offset, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND(frame,glupy_readdirp_cbk,xl,xl->fops->readdirp,fd,size,offset,xdata);
-}
-
-void
-unwind_readdirp (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
- dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT(readdirp,frame,op_ret,op_errno,
- entries, xdata);
-}
-
-void
-set_readdirp_fop (long py_this, fop_readdirp_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_READDIRP] = (long)fop;
-}
-
-void
-set_readdirp_cbk (long py_this, fop_readdirp_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_READDIRP] = (long)cbk;
-}
-
-
-/* FOP:STAT */
-
-int32_t
-glupy_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_STAT]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_stat_cbk_t)(priv->cbks[GLUPY_STAT]))(
- frame, cookie, this, op_ret, op_errno,
- buf, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (stat, frame, op_ret, op_errno, buf, xdata);
- return 0;
-}
-
-int32_t
-glupy_stat (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_STAT]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_stat_t)(priv->fops[GLUPY_STAT]))(
- frame, this, loc, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_stat_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->stat, loc, xdata);
- return 0;
-}
-
-void
-wind_stat (call_frame_t *frame, xlator_t *xl, loc_t *loc, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND(frame,glupy_stat_cbk,xl,xl->fops->stat,loc,xdata);
-}
-
-void
-unwind_stat (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT(stat,frame,op_ret,op_errno,
- buf,xdata);
-}
-
-void
-set_stat_fop (long py_this, fop_stat_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_STAT] = (long)fop;
-}
-
-void
-set_stat_cbk (long py_this, fop_stat_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_STAT] = (long)cbk;
-}
-
-
-/* FOP: FSTAT */
-
-int32_t
-glupy_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_FSTAT]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_fstat_cbk_t)(priv->cbks[GLUPY_FSTAT]))(
- frame, cookie, this, op_ret, op_errno,
- buf, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, buf, xdata);
- return 0;
-}
-
-int32_t
-glupy_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd,
- dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_FSTAT]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_fstat_t)(priv->fops[GLUPY_FSTAT]))(
- frame, this, fd, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_fstat_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fstat, fd, xdata);
- return 0;
-}
-
-void
-wind_fstat (call_frame_t *frame, xlator_t *xl, fd_t *fd, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND(frame,glupy_fstat_cbk,xl,xl->fops->fstat,fd,xdata);
-}
-
-void
-unwind_fstat (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT(fstat,frame,op_ret,op_errno,
- buf,xdata);
-}
-
-void
-set_fstat_fop (long py_this, fop_fstat_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_FSTAT] = (long)fop;
-}
-
-void
-set_fstat_cbk (long py_this, fop_fstat_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_FSTAT] = (long)cbk;
-}
-
-/* FOP:STATFS */
-
-int32_t
-glupy_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct statvfs *buf, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_STATFS]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_statfs_cbk_t)(priv->cbks[GLUPY_STATFS]))(
- frame, cookie, this, op_ret, op_errno,
- buf, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, buf, xdata);
- return 0;
-}
-
-int32_t
-glupy_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_STATFS]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_statfs_t)(priv->fops[GLUPY_STATFS]))(
- frame, this, loc, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_statfs_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->statfs, loc, xdata);
- return 0;
-}
-
-void
-wind_statfs (call_frame_t *frame, xlator_t *xl, loc_t *loc, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND(frame,glupy_statfs_cbk,xl,xl->fops->statfs,loc,xdata);
-}
-
-void
-unwind_statfs (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct statvfs *buf,
- dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT(statfs,frame,op_ret,op_errno,
- buf,xdata);
-}
-
-void
-set_statfs_fop (long py_this, fop_statfs_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_STATFS] = (long)fop;
-}
-
-void
-set_statfs_cbk (long py_this, fop_statfs_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_STATFS] = (long)cbk;
-}
-
-
-/* FOP: SETXATTR */
-
-int32_t
-glupy_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_SETXATTR]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_setxattr_cbk_t)(priv->cbks[GLUPY_SETXATTR]))(
- frame, cookie, this, op_ret, op_errno,
- xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
- return 0;
-}
-
-int32_t
-glupy_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *dict, int32_t flags, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_SETXATTR]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_setxattr_t)(priv->fops[GLUPY_SETXATTR]))(
- frame, this, loc, dict, flags, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_setxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setxattr, loc, dict,
- flags, xdata);
- return 0;
-}
-
-void
-wind_setxattr (call_frame_t *frame, xlator_t *xl, loc_t *loc,
- dict_t *dict, int32_t flags, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND (frame, glupy_setxattr_cbk, xl, xl->fops->setxattr,
- loc, dict, flags, xdata);
-}
-
-
-void
-unwind_setxattr (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
-
-}
-
-void
-set_setxattr_fop (long py_this, fop_setxattr_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_SETXATTR] = (long)fop;
-}
-
-void
-set_setxattr_cbk (long py_this, fop_setxattr_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_SETXATTR] = (long)cbk;
-}
-
-/* FOP: GETXATTR */
-
-int32_t
-glupy_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_GETXATTR]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_getxattr_cbk_t)(priv->cbks[GLUPY_GETXATTR]))(
- frame, cookie, this, op_ret, op_errno, dict,
- xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict,
- xdata);
- return 0;
-}
-
-int32_t
-glupy_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_GETXATTR]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_getxattr_t)(priv->fops[GLUPY_GETXATTR]))(
- frame, this, loc, name, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_getxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr, loc, name,
- xdata);
- return 0;
-}
-
-void
-wind_getxattr (call_frame_t *frame, xlator_t *xl, loc_t *loc,
- const char *name, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND (frame, glupy_getxattr_cbk, xl, xl->fops->getxattr,
- loc, name, xdata);
-}
-
-
-void
-unwind_getxattr (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict,
- xdata);
-
-}
-
-
-void
-set_getxattr_fop (long py_this, fop_getxattr_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_GETXATTR] = (long)fop;
-}
-
-
-void
-set_getxattr_cbk (long py_this, fop_getxattr_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_GETXATTR] = (long)cbk;
-}
-
-/* FOP: FSETXATTR */
-
-int32_t
-glupy_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_FSETXATTR]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_fsetxattr_cbk_t)(priv->cbks[GLUPY_FSETXATTR]))(
- frame, cookie, this, op_ret, op_errno,
- xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
- return 0;
-}
-
-int32_t
-glupy_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- dict_t *dict, int32_t flags, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_FSETXATTR]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_fsetxattr_t)(priv->fops[GLUPY_FSETXATTR]))(
- frame, this, fd, dict, flags, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_fsetxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetxattr, fd, dict,
- flags, xdata);
- return 0;
-}
-
-void
-wind_fsetxattr (call_frame_t *frame, xlator_t *xl, fd_t *fd,
- dict_t *dict, int32_t flags, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND (frame, glupy_fsetxattr_cbk, xl, xl->fops->fsetxattr,
- fd, dict, flags, xdata);
-}
-
-
-void
-unwind_fsetxattr (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
-
-}
-
-void
-set_fsetxattr_fop (long py_this, fop_fsetxattr_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_FSETXATTR] = (long)fop;
-}
-
-void
-set_fsetxattr_cbk (long py_this, fop_fsetxattr_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_FSETXATTR] = (long)cbk;
-}
-
-/* FOP: FGETXATTR */
-
-int32_t
-glupy_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_FGETXATTR]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_fgetxattr_cbk_t)(priv->cbks[GLUPY_FGETXATTR]))(
- frame, cookie, this, op_ret, op_errno, dict,
- xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict,
- xdata);
- return 0;
-}
-
-int32_t
-glupy_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_FGETXATTR]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_fgetxattr_t)(priv->fops[GLUPY_FGETXATTR]))(
- frame, this, fd, name, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_fgetxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fgetxattr, fd, name,
- xdata);
- return 0;
-}
-
-void
-wind_fgetxattr (call_frame_t *frame, xlator_t *xl, fd_t *fd,
- const char *name, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND (frame, glupy_fgetxattr_cbk, xl, xl->fops->fgetxattr,
- fd, name, xdata);
-}
-
-
-void
-unwind_fgetxattr (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict,
- xdata);
-
-}
-
-
-void
-set_fgetxattr_fop (long py_this, fop_fgetxattr_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_FGETXATTR] = (long)fop;
-}
-
-
-void
-set_fgetxattr_cbk (long py_this, fop_fgetxattr_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_FGETXATTR] = (long)cbk;
-}
-
-/* FOP:REMOVEXATTR */
-
-int32_t
-glupy_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_REMOVEXATTR]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_removexattr_cbk_t)(priv->cbks[GLUPY_REMOVEXATTR]))(
- frame, cookie, this, op_ret, op_errno, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
- return 0;
-}
-
-int32_t
-glupy_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_REMOVEXATTR]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_removexattr_t)(priv->fops[GLUPY_REMOVEXATTR]))(
- frame, this, loc, name, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_removexattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->removexattr, loc, name,
- xdata);
- return 0;
-}
-
-void
-wind_removexattr (call_frame_t *frame, xlator_t *xl, loc_t *loc,
- const char *name, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND (frame, glupy_removexattr_cbk, xl, xl->fops->removexattr,
- loc, name, xdata);
-}
-
-
-void
-unwind_removexattr (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
-
-}
-
-void
-set_removexattr_fop (long py_this, fop_removexattr_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_REMOVEXATTR] = (long)fop;
-}
-
-void
-set_removexattr_cbk (long py_this, fop_removexattr_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_REMOVEXATTR] = (long)cbk;
-}
-
-
-/* FOP:FREMOVEXATTR */
-
-int32_t
-glupy_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_FREMOVEXATTR]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_fremovexattr_cbk_t)(priv->cbks[GLUPY_FREMOVEXATTR]))(
- frame, cookie, this, op_ret, op_errno, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (fremovexattr, frame, op_ret, op_errno, xdata);
- return 0;
-}
-
-int32_t
-glupy_fremovexattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_FREMOVEXATTR]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_fremovexattr_t)(priv->fops[GLUPY_FREMOVEXATTR]))(
- frame, this, fd, name, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_fremovexattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fremovexattr, fd, name,
- xdata);
- return 0;
-}
-
-void
-wind_fremovexattr (call_frame_t *frame, xlator_t *xl, fd_t *fd,
- const char *name, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND (frame, glupy_fremovexattr_cbk, xl, xl->fops->fremovexattr,
- fd, name, xdata);
-}
-
-
-void
-unwind_fremovexattr (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT (fremovexattr, frame, op_ret, op_errno, xdata);
-
-}
-
-void
-set_fremovexattr_fop (long py_this, fop_fremovexattr_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_FREMOVEXATTR] = (long)fop;
-}
-
-void
-set_fremovexattr_cbk (long py_this, fop_fremovexattr_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_FREMOVEXATTR] = (long)cbk;
-}
-
-
-/* FOP: LINK*/
-int32_t
-glupy_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_LINK]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_link_cbk_t)(priv->cbks[GLUPY_LINK]))(
- frame, cookie, this, op_ret, op_errno,
- inode, buf, preparent, postparent, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
- return 0;
-}
-
-int32_t
-glupy_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_LINK]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_link_t)(priv->fops[GLUPY_LINK]))(
- frame, this, oldloc, newloc, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_link_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->link, oldloc, newloc,
- xdata);
- return 0;
-}
-
-void
-wind_link (call_frame_t *frame, xlator_t *xl, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND (frame, glupy_link_cbk, xl, xl->fops->link,
- oldloc, newloc, xdata);
-}
-
-void
-unwind_link (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
-}
-
-void
-set_link_fop (long py_this, fop_link_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_LINK] = (long)fop;
-}
-
-void
-set_link_cbk (long py_this, fop_link_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_LINK] = (long)cbk;
-}
-
-/* FOP: SYMLINK*/
-int32_t
-glupy_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_SYMLINK]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_symlink_cbk_t)(priv->cbks[GLUPY_SYMLINK]))(
- frame, cookie, this, op_ret, op_errno,
- inode, buf, preparent, postparent, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
- return 0;
-}
-
-int32_t
-glupy_symlink(call_frame_t *frame, xlator_t *this, const char *linkname,
- loc_t *loc, mode_t umask, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_SYMLINK]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_symlink_t)(priv->fops[GLUPY_SYMLINK]))(
- frame, this, linkname, loc, umask, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_symlink_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->symlink, linkname, loc,
- umask, xdata);
- return 0;
-}
-
-void
-wind_symlink (call_frame_t *frame, xlator_t *xl, const char *linkname,
- loc_t *loc, mode_t umask, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND (frame, glupy_symlink_cbk, xl, xl->fops->symlink,
- linkname, loc, umask, xdata);
-}
-
-void
-unwind_symlink (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
-}
-
-void
-set_symlink_fop (long py_this, fop_symlink_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_SYMLINK] = (long)fop;
-}
-
-void
-set_symlink_cbk (long py_this, fop_symlink_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_SYMLINK] = (long)cbk;
-}
-
-
-/* FOP: READLINK */
-int32_t
-glupy_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, const char *path,
- struct iatt *buf, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_READLINK]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_readlink_cbk_t)(priv->cbks[GLUPY_READLINK]))(
- frame, cookie, this, op_ret, op_errno,
- path, buf, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, path,
- buf, xdata);
- return 0;
-}
-
-int32_t
-glupy_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc,
- size_t size, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_READLINK]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_readlink_t)(priv->fops[GLUPY_READLINK]))(
- frame, this, loc, size, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_readlink_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readlink, loc,
- size, xdata);
- return 0;
-}
-
-void
-wind_readlink (call_frame_t *frame, xlator_t *xl, loc_t *loc,
- size_t size, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND (frame, glupy_readlink_cbk, xl, xl->fops->readlink,
- loc, size, xdata);
-}
-
-void
-unwind_readlink (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, const char *path,
- struct iatt *buf, dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, path, buf,
- xdata);
-}
-
-void
-set_readlink_fop (long py_this, fop_readlink_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_READLINK] = (long)fop;
-}
-
-void
-set_readlink_cbk (long py_this, fop_readlink_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_READLINK] = (long)cbk;
-}
-
-
-/* FOP: UNLINK */
-
-int32_t
-glupy_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_UNLINK]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_unlink_cbk_t)(priv->cbks[GLUPY_UNLINK]))(
- frame, cookie, this, op_ret, op_errno,
- preparent, postparent, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno, preparent,
- postparent, xdata);
- return 0;
-}
-
-int32_t
-glupy_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int xflags, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_UNLINK]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_unlink_t)(priv->fops[GLUPY_UNLINK]))(
- frame, this, loc, xflags, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_unlink_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->unlink, loc,
- xflags, xdata);
- return 0;
-}
-
-void
-wind_unlink (call_frame_t *frame, xlator_t *xl, loc_t *loc,
- int xflags, dict_t *xdata)
-{
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND (frame, glupy_unlink_cbk, xl, xl->fops->unlink,
- loc, xflags, xdata);
-}
-
-void
-unwind_unlink (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno,
- preparent, postparent, xdata);
-}
-
-void
-set_unlink_fop (long py_this, fop_unlink_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_UNLINK] = (long)fop;
-}
-
-void
-set_unlink_cbk (long py_this, fop_unlink_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_UNLINK] = (long)cbk;
-}
-
-
-/* FOP: MKDIR */
-
-int32_t
-glupy_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_MKDIR]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_mkdir_cbk_t)(priv->cbks[GLUPY_MKDIR]))(
- frame, cookie, this, op_ret, op_errno,
- inode, buf, preparent, postparent, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
- return 0;
-}
-
-int32_t
-glupy_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- mode_t umask, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_MKDIR]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_mkdir_t)(priv->fops[GLUPY_MKDIR]))(
- frame, this, loc, mode, umask, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_mkdir_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mkdir, loc, mode, umask,
- xdata);
- return 0;
-}
-
-void
-wind_mkdir (call_frame_t *frame, xlator_t *xl, loc_t *loc, mode_t mode,
- mode_t umask, dict_t *xdata)
-{
-
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND (frame, glupy_mkdir_cbk, xl, xl->fops->mkdir,
- loc, mode, umask, xdata);
-}
-
-void
-unwind_mkdir (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
-}
-
-void
-set_mkdir_fop (long py_this, fop_mkdir_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_MKDIR] = (long)fop;
-}
-
-void
-set_mkdir_cbk (long py_this, fop_mkdir_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_MKDIR] = (long)cbk;
-}
-
-
-/* FOP: RMDIR */
-
-int32_t
-glupy_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
-
- if (!priv->cbks[GLUPY_RMDIR]) {
- goto unwind;
- }
-
- gstate = glupy_enter();
- ret = ((fop_rmdir_cbk_t)(priv->cbks[GLUPY_RMDIR]))(
- frame, cookie, this, op_ret, op_errno,
- preparent, postparent, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-unwind:
- frame->local = NULL;
- STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno, preparent,
- postparent, xdata);
- return 0;
-}
-
-int32_t
-glupy_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int xflags, dict_t *xdata)
-{
- glupy_private_t *priv = this->private;
- PyGILState_STATE gstate;
- int32_t ret;
- static long next_id = 0;
-
- if (!priv->fops[GLUPY_RMDIR]) {
- goto wind;
- }
-
- gstate = glupy_enter();
- frame->local = (void *)++next_id;
- ret = ((fop_rmdir_t)(priv->fops[GLUPY_RMDIR]))(
- frame, this, loc, xflags, xdata);
- glupy_leave(gstate);
-
- return ret;
-
-wind:
- STACK_WIND (frame, glupy_rmdir_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rmdir, loc,
- xflags, xdata);
- return 0;
-}
-
-void
-wind_rmdir (call_frame_t *frame, xlator_t *xl, loc_t *loc,
- int xflags, dict_t *xdata)
-{
-
- xlator_t *this = THIS;
-
- if (!xl || (xl == this)) {
- xl = FIRST_CHILD(this);
- }
-
- STACK_WIND (frame, glupy_rmdir_cbk, xl, xl->fops->rmdir,
- loc, xflags, xdata);
-}
-
-void
-unwind_rmdir (call_frame_t *frame, long cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
-{
- frame->local = NULL;
- STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno,
- preparent, postparent, xdata);
-}
-
-void
-set_rmdir_fop (long py_this, fop_rmdir_t fop)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->fops[GLUPY_RMDIR] = (long)fop;
-}
-
-void
-set_rmdir_cbk (long py_this, fop_rmdir_cbk_t cbk)
-{
- glupy_private_t *priv = ((xlator_t *)py_this)->private;
-
- priv->cbks[GLUPY_RMDIR] = (long)cbk;
-}
-
-
-/* NON-FOP-SPECIFIC CODE */
-
-
-long
-get_id (call_frame_t *frame)
-{
- return (long)(frame->local);
-}
-
-uint64_t
-get_rootunique (call_frame_t *frame)
-{
- return frame->root->unique;
-}
-
-int32_t
-mem_acct_init (xlator_t *this)
-{
- int ret = -1;
-
- if (!this)
- return ret;
-
- ret = xlator_mem_acct_init (this, gf_glupy_mt_end);
-
- if (ret != 0) {
- gf_log(this->name, GF_LOG_ERROR, "Memory accounting init"
- " failed");
- return ret;
- }
-
- return ret;
-}
-
-int32_t
-init (xlator_t *this)
-{
- glupy_private_t *priv = NULL;
- char *module_name = NULL;
- PyObject *py_mod_name = NULL;
- PyObject *py_init_func = NULL;
- PyObject *py_args = NULL;
- PyObject *syspath = NULL;
- PyObject *path = NULL;
- PyObject *error_type = NULL;
- PyObject *error_msg = NULL;
- PyObject *error_bt = NULL;
- static gf_boolean_t py_inited = _gf_false;
- void * err_cleanup = &&err_return;
- char libpython[16];
-
- if (dict_get_str(this->options,"module-name",&module_name) != 0) {
- gf_log (this->name, GF_LOG_ERROR, "missing module-name");
- return -1;
- }
-
- priv = GF_CALLOC (1, sizeof (glupy_private_t), gf_glupy_mt_priv);
- if (!priv) {
- goto *err_cleanup;
- }
- this->private = priv;
- err_cleanup = &&err_free_priv;
-
- if (!py_inited) {
- /* FIXME:
- * This hack is necessary because glusterfs (rightly) loads
- * glupy.so with RTLD_LOCAL but glupy needs libpython to be
- * loaded with RTLD_GLOBAL even though glupy is correctly
- * linked with libpython.
- * This is needed because one of the internal modules of
- * python 2.x (lib-dynload/_struct.so) does not explicitly
- * link with libpython.
- */
- snprintf(libpython, sizeof(libpython), "libpython%d.%d.so",
- PY_MAJOR_VERSION, PY_MINOR_VERSION);
- if (!dlopen (libpython, RTLD_NOW|RTLD_GLOBAL)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- LG_MSG_DLOPEN_FAILED, "dlopen(%s) failed: %s",
- libpython, dlerror ());
- }
-
- /*
- * This must be done before Py_Initialize(),
- * because it will duplicate the environment,
- * and fail to see later environment updates.
- */
- setenv("PATH_GLUSTERFS_GLUPY_MODULE",
- PATH_GLUSTERFS_GLUPY_MODULE, 1);
-
- Py_Initialize();
- PyEval_InitThreads();
-
- (void)pthread_key_create(&gil_init_key,NULL);
- (void)pthread_setspecific(gil_init_key,(void *)1);
-
- /* PyEval_InitThreads takes this "for" us. No thanks. */
- PyEval_ReleaseLock();
- py_inited = _gf_true;
- }
-
- /* Adjust python's path */
- syspath = PySys_GetObject("path");
- path = PyString_FromString(GLUSTER_PYTHON_PATH);
- PyList_Append(syspath, path);
- Py_DECREF(path);
-
- py_mod_name = PyString_FromString(module_name);
- if (!py_mod_name) {
- gf_log (this->name, GF_LOG_ERROR, "could not create name");
- if (PyErr_Occurred()) {
- PyErr_Fetch (&error_type, &error_msg, &error_bt);
- gf_log (this->name, GF_LOG_ERROR, "Python error: %s",
- PyString_AsString(error_msg));
- }
- goto *err_cleanup;
- }
-
- gf_log (this->name, GF_LOG_DEBUG, "py_mod_name = %s", module_name);
- priv->py_module = PyImport_Import(py_mod_name);
- Py_DECREF(py_mod_name);
- if (!priv->py_module) {
- gf_log (this->name, GF_LOG_ERROR, "Python import of %s failed",
- module_name);
- if (PyErr_Occurred()) {
- PyErr_Fetch (&error_type, &error_msg, &error_bt);
- gf_log (this->name, GF_LOG_ERROR, "Python error: %s",
- PyString_AsString(error_msg));
- }
- goto *err_cleanup;
- }
- gf_log (this->name, GF_LOG_INFO, "Import of %s succeeded", module_name);
- err_cleanup = &&err_deref_module;
-
- py_init_func = PyObject_GetAttrString(priv->py_module, "xlator");
- if (!py_init_func || !PyCallable_Check(py_init_func)) {
- gf_log (this->name, GF_LOG_ERROR, "missing init func");
- if (PyErr_Occurred()) {
- PyErr_Fetch (&error_type, &error_msg, &error_bt);
- gf_log (this->name, GF_LOG_ERROR, "Python error: %s",
- PyString_AsString(error_msg));
- }
- goto *err_cleanup;
- }
- err_cleanup = &&err_deref_init;
-
- py_args = PyTuple_New(1);
- if (!py_args) {
- gf_log (this->name, GF_LOG_ERROR, "could not create args");
- if (PyErr_Occurred()) {
- PyErr_Fetch (&error_type, &error_msg, &error_bt);
- gf_log (this->name, GF_LOG_ERROR, "Python error: %s",
- PyString_AsString(error_msg));
- }
- goto *err_cleanup;
- }
- PyTuple_SetItem(py_args,0,PyLong_FromLong((long)this));
-
- /* TBD: pass in list of children */
- priv->py_xlator = PyObject_CallObject(py_init_func, py_args);
- Py_DECREF(py_args);
- if (!priv->py_xlator) {
- gf_log (this->name, GF_LOG_ERROR, "Python init failed");
- if (PyErr_Occurred()) {
- PyErr_Fetch (&error_type, &error_msg, &error_bt);
- gf_log (this->name, GF_LOG_ERROR, "Python error: %s",
- PyString_AsString(error_msg));
- }
- goto *err_cleanup;
- }
- gf_log (this->name, GF_LOG_DEBUG, "init returned %p", priv->py_xlator);
-
- return 0;
-
-err_deref_init:
- Py_DECREF(py_init_func);
-err_deref_module:
- Py_DECREF(priv->py_module);
-err_free_priv:
- GF_FREE(priv);
-err_return:
- return -1;
-}
-
-void
-fini (xlator_t *this)
-{
- glupy_private_t *priv = this->private;
-
- if (!priv)
- return;
- Py_DECREF(priv->py_xlator);
- Py_DECREF(priv->py_module);
- this->private = NULL;
- GF_FREE (priv);
-
- return;
-}
-
-struct xlator_fops fops = {
- .lookup = glupy_lookup,
- .create = glupy_create,
- .open = glupy_open,
- .readv = glupy_readv,
- .writev = glupy_writev,
- .opendir = glupy_opendir,
- .readdir = glupy_readdir,
- .stat = glupy_stat,
- .fstat = glupy_fstat,
- .setxattr = glupy_setxattr,
- .getxattr = glupy_getxattr,
- .fsetxattr = glupy_fsetxattr,
- .fgetxattr = glupy_fgetxattr,
- .removexattr = glupy_removexattr,
- .fremovexattr = glupy_fremovexattr,
- .link = glupy_link,
- .unlink = glupy_unlink,
- .readlink = glupy_readlink,
- .symlink = glupy_symlink,
- .mkdir = glupy_mkdir,
- .rmdir = glupy_rmdir,
- .statfs = glupy_statfs,
- .readdirp = glupy_readdirp
-};
-
-struct xlator_cbks cbks = {
-};
-
-struct volume_options options[] = {
- { .key = {NULL} },
-};
diff --git a/xlators/features/glupy/src/glupy.h b/xlators/features/glupy/src/glupy.h
deleted file mode 100644
index 1488c55c331..00000000000
--- a/xlators/features/glupy/src/glupy.h
+++ /dev/null
@@ -1,56 +0,0 @@
-/*
- Copyright (c) 2006-2014 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
-*/
-
-#ifndef __GLUPY_H__
-#define __GLUPY_H__
-
-#include "mem-types.h"
-
-enum {
- GLUPY_LOOKUP = 0,
- GLUPY_CREATE,
- GLUPY_OPEN,
- GLUPY_READV,
- GLUPY_WRITEV,
- GLUPY_OPENDIR,
- GLUPY_READDIR,
- GLUPY_READDIRP,
- GLUPY_STAT,
- GLUPY_FSTAT,
- GLUPY_STATFS,
- GLUPY_SETXATTR,
- GLUPY_GETXATTR,
- GLUPY_FSETXATTR,
- GLUPY_FGETXATTR,
- GLUPY_REMOVEXATTR,
- GLUPY_FREMOVEXATTR,
- GLUPY_LINK,
- GLUPY_UNLINK,
- GLUPY_READLINK,
- GLUPY_SYMLINK,
- GLUPY_MKNOD,
- GLUPY_MKDIR,
- GLUPY_RMDIR,
- GLUPY_N_FUNCS
-};
-
-typedef struct {
- PyObject *py_module;
- PyObject *py_xlator;
- long fops[GLUPY_N_FUNCS];
- long cbks[GLUPY_N_FUNCS];
-} glupy_private_t;
-
-enum gf_glupy_mem_types_ {
- gf_glupy_mt_priv = gf_common_mt_end + 1,
- gf_glupy_mt_end
-};
-
-#endif /* __GLUPY_H__ */
diff --git a/xlators/features/glupy/src/glupy.sym b/xlators/features/glupy/src/glupy.sym
deleted file mode 100644
index 55d9a300108..00000000000
--- a/xlators/features/glupy/src/glupy.sym
+++ /dev/null
@@ -1,101 +0,0 @@
-init
-fini
-fops
-cbks
-options
-notify
-mem_acct_init
-reconfigure
-dumpops
-set_lookup_fop
-set_lookup_cbk
-set_create_fop
-set_create_cbk
-set_open_fop
-set_open_cbk
-set_readv_fop
-set_readv_cbk
-set_writev_fop
-set_writev_cbk
-set_opendir_fop
-set_opendir_cbk
-set_readdir_fop
-set_readdir_cbk
-set_readdirp_fop
-set_readdirp_cbk
-set_stat_fop
-set_stat_cbk
-set_fstat_fop
-set_fstat_cbk
-set_statfs_fop
-set_statfs_cbk
-set_setxattr_fop
-set_setxattr_cbk
-set_getxattr_fop
-set_getxattr_cbk
-set_fsetxattr_fop
-set_fsetxattr_cbk
-set_fgetxattr_fop
-set_fgetxattr_cbk
-set_removexattr_fop
-set_removexattr_cbk
-set_fremovexattr_fop
-set_fremovexattr_cbk
-set_link_fop
-set_link_cbk
-set_symlink_fop
-set_symlink_cbk
-set_readlink_fop
-set_readlink_cbk
-set_unlink_fop
-set_unlink_cbk
-set_mkdir_fop
-set_mkdir_cbk
-set_rmdir_fop
-set_rmdir_cbk
-wind_lookup
-wind_create
-wind_open
-wind_readv
-wind_writev
-wind_opendir
-wind_readdir
-wind_readdirp
-wind_stat
-wind_fstat
-wind_statfs
-wind_setxattr
-wind_getxattr
-wind_fsetxattr
-wind_fgetxattr
-wind_removexattr
-wind_fremovexattr
-wind_link
-wind_symlink
-wind_readlink
-wind_unlink
-wind_mkdir
-wind_rmdir
-unwind_lookup
-unwind_create
-unwind_open
-unwind_readv
-unwind_writev
-unwind_opendir
-unwind_readdir
-unwind_readdirp
-unwind_stat
-unwind_fstat
-unwind_statfs
-unwind_setxattr
-unwind_getxattr
-unwind_fsetxattr
-unwind_fgetxattr
-unwind_removexattr
-unwind_fremovexattr
-unwind_link
-unwind_symlink
-unwind_readlink
-unwind_unlink
-unwind_mkdir
-unwind_rmdir
diff --git a/xlators/features/glupy/src/glupy/Makefile.am b/xlators/features/glupy/src/glupy/Makefile.am
deleted file mode 100644
index 573d2da12e1..00000000000
--- a/xlators/features/glupy/src/glupy/Makefile.am
+++ /dev/null
@@ -1,5 +0,0 @@
-# Install __init__.py into the Python site-packages area
-pyglupydir = @BUILD_PYTHON_SITE_PACKAGES@/gluster/glupy
-pyglupy_PYTHON = __init__.py
-
-CLEANFILES =
diff --git a/xlators/features/glupy/src/glupy/__init__.py b/xlators/features/glupy/src/glupy/__init__.py
deleted file mode 100644
index b9fc3700fa6..00000000000
--- a/xlators/features/glupy/src/glupy/__init__.py
+++ /dev/null
@@ -1,852 +0,0 @@
-##
-## Copyright (c) 2006-2014 Red Hat, Inc. <http://www.redhat.com>
-## This file is part of GlusterFS.
-##
-## This file is licensed to you under your choice of the GNU Lesser
-## General Public License, version 3 or any later version (LGPLv3 or
-## later), or the GNU General Public License, version 2 (GPLv2), in all
-## cases as published by the Free Software Foundation.
-##
-
-import sys
-import os
-from ctypes import *
-
-dl = CDLL(os.getenv("PATH_GLUSTERFS_GLUPY_MODULE", ""),RTLD_GLOBAL)
-
-
-class call_frame_t (Structure):
- pass
-
-class dev_t (Structure):
- pass
-
-
-class dict_t (Structure):
- pass
-
-
-class gf_dirent_t (Structure):
- pass
-
-
-class iobref_t (Structure):
- pass
-
-
-class iovec_t (Structure):
- pass
-
-
-class list_head (Structure):
- pass
-
-list_head._fields_ = [
- ("next", POINTER(list_head)),
- ("prev", POINTER(list_head))
- ]
-
-
-class rwxperm_t (Structure):
- _fields_ = [
- ("read", c_uint8, 1),
- ("write", c_uint8, 1),
- ("execn", c_uint8, 1)
- ]
-
-
-class statvfs_t (Structure):
- pass
-
-
-class xlator_t (Structure):
- pass
-
-
-class ia_prot_t (Structure):
- _fields_ = [
- ("suid", c_uint8, 1),
- ("sgid", c_uint8, 1),
- ("sticky", c_uint8, 1),
- ("owner", rwxperm_t),
- ("group", rwxperm_t),
- ("other", rwxperm_t)
- ]
-
-# For checking file type.
-(IA_INVAL, IA_IFREG, IA_IFDIR, IA_IFLNK, IA_IFBLK, IA_IFCHR, IA_IFIFO,
- IA_IFSOCK) = xrange(8)
-
-
-class iatt_t (Structure):
- _fields_ = [
- ("ia_no", c_uint64),
- ("ia_gfid", c_ubyte * 16),
- ("ia_dev", c_uint64),
- ("ia_type", c_uint),
- ("ia_prot", ia_prot_t),
- ("ia_nlink", c_uint32),
- ("ia_uid", c_uint32),
- ("ia_gid", c_uint32),
- ("ia_rdev", c_uint64),
- ("ia_size", c_uint64),
- ("ia_blksize", c_uint32),
- ("ia_blocks", c_uint64),
- ("ia_atime", c_uint32 ),
- ("ia_atime_nsec", c_uint32),
- ("ia_mtime", c_uint32),
- ("ia_mtime_nsec", c_uint32),
- ("ia_ctime", c_uint32),
- ("ia_ctime_nsec", c_uint32)
- ]
-
-
-class mem_pool (Structure):
- _fields_ = [
- ("list", list_head),
- ("hot_count", c_int),
- ("cold_count", c_int),
- ("lock", c_void_p),
- ("padded_sizeof_type", c_ulong),
- ("pool", c_void_p),
- ("pool_end", c_void_p),
- ("real_sizeof_type", c_int),
- ("alloc_count", c_uint64),
- ("pool_misses", c_uint64),
- ("max_alloc", c_int),
- ("curr_stdalloc", c_int),
- ("max_stdalloc", c_int),
- ("name", c_char_p),
- ("global_list", list_head)
- ]
-
-
-class U_ctx_key_inode (Union):
- _fields_ = [
- ("key", c_uint64),
- ("xl_key", POINTER(xlator_t))
- ]
-
-
-class U_ctx_value1 (Union):
- _fields_ = [
- ("value1", c_uint64),
- ("ptr1", c_void_p)
- ]
-
-
-class U_ctx_value2 (Union):
- _fields_ = [
- ("value2", c_uint64),
- ("ptr2", c_void_p)
- ]
-
-class inode_ctx (Structure):
- _anonymous_ = ("u_key","u_value1","u_value2",)
- _fields_ = [
- ("u_key", U_ctx_key_inode),
- ("u_value1", U_ctx_value1),
- ("u_value2", U_ctx_value2)
- ]
-
-class inode_t (Structure):
- pass
-
-class inode_table_t (Structure):
- _fields_ = [
- ("lock", c_void_p),
- ("hashsize", c_size_t),
- ("name", c_char_p),
- ("root", POINTER(inode_t)),
- ("xl", POINTER(xlator_t)),
- ("lru_limit", c_uint32),
- ("inode_hash", POINTER(list_head)),
- ("name_hash", POINTER(list_head)),
- ("active", list_head),
- ("active_size", c_uint32),
- ("lru", list_head),
- ("lru_size", c_uint32),
- ("purge", list_head),
- ("purge_size", c_uint32),
- ("inode_pool", POINTER(mem_pool)),
- ("dentry_pool", POINTER(mem_pool)),
- ("fd_mem_pool", POINTER(mem_pool))
- ]
-
-inode_t._fields_ = [
- ("table", POINTER(inode_table_t)),
- ("gfid", c_ubyte * 16),
- ("lock", c_void_p),
- ("nlookup", c_uint64),
- ("fd_count", c_uint32),
- ("ref", c_uint32),
- ("ia_type", c_uint),
- ("fd_list", list_head),
- ("dentry_list", list_head),
- ("hashv", list_head),
- ("listv", list_head),
- ("ctx", POINTER(inode_ctx))
- ]
-
-
-
-class U_ctx_key_fd (Union):
- _fields_ = [
- ("key", c_uint64),
- ("xl_key", c_void_p)
- ]
-
-class fd_lk_ctx (Structure):
- _fields_ = [
- ("lk_list", list_head),
- ("ref", c_int),
- ("lock", c_void_p)
- ]
-
-class fd_ctx (Structure):
- _anonymous_ = ("u_key","u_value1")
- _fields_ = [
- ("u_key", U_ctx_key_fd),
- ("u_value1", U_ctx_value1)
- ]
-
-class fd_t (Structure):
- _fields_ = [
- ("pid", c_uint64),
- ("flags", c_int32),
- ("refcount", c_int32),
- ("inode_list", list_head),
- ("inode", POINTER(inode_t)),
- ("lock", c_void_p),
- ("ctx", POINTER(fd_ctx)),
- ("xl_count", c_int),
- ("lk_ctx", POINTER(fd_lk_ctx)),
- ("anonymous", c_uint)
- ]
-
-class loc_t (Structure):
- _fields_ = [
- ("path", c_char_p),
- ("name", c_char_p),
- ("inode", POINTER(inode_t)),
- ("parent", POINTER(inode_t)),
- ("gfid", c_ubyte * 16),
- ("pargfid", c_ubyte * 16),
- ]
-
-
-
-def _init_op (a_class, fop, cbk, wind, unwind):
- # Decorators, used by translators. We could pass the signatures as
- # parameters, but it's actually kind of nice to keep them around for
- # inspection.
- a_class.fop_type = apply(CFUNCTYPE,a_class.fop_sig)
- a_class.cbk_type = apply(CFUNCTYPE,a_class.cbk_sig)
- # Dispatch-function registration.
- fop.restype = None
- fop.argtypes = [ c_long, a_class.fop_type ]
- # Callback-function registration.
- cbk.restype = None
- cbk.argtypes = [ c_long, a_class.cbk_type ]
- # STACK_WIND function.
- wind.restype = None
- wind.argtypes = list(a_class.fop_sig[1:])
- # STACK_UNWIND function.
- unwind.restype = None
- unwind.argtypes = list(a_class.cbk_sig[1:])
-
-class OpLookup:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(loc_t), POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(inode_t), POINTER(iatt_t),
- POINTER(dict_t), POINTER(iatt_t))
-_init_op (OpLookup, dl.set_lookup_fop, dl.set_lookup_cbk,
- dl.wind_lookup, dl.unwind_lookup)
-
-class OpCreate:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(loc_t), c_int, c_uint, c_uint, POINTER(fd_t),
- POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(fd_t), POINTER(inode_t),
- POINTER(iatt_t), POINTER(iatt_t), POINTER(iatt_t),
- POINTER(dict_t))
-_init_op (OpCreate, dl.set_create_fop, dl.set_create_cbk,
- dl.wind_create, dl.unwind_create)
-
-class OpOpen:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(loc_t), c_int, POINTER(fd_t), POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(fd_t), POINTER(dict_t))
-_init_op (OpOpen, dl.set_open_fop, dl.set_open_cbk,
- dl.wind_open, dl.unwind_open)
-
-class OpReadv:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(fd_t), c_size_t, c_long, c_uint32, POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(iovec_t), c_int, POINTER(iatt_t),
- POINTER(iobref_t), POINTER(dict_t))
-_init_op (OpReadv, dl.set_readv_fop, dl.set_readv_cbk,
- dl.wind_readv, dl.unwind_readv)
-class OpWritev:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(fd_t), POINTER(iovec_t), c_int, c_long, c_uint32,
- POINTER(iobref_t), POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(iatt_t), POINTER(iatt_t),
- POINTER(dict_t))
-_init_op (OpWritev, dl.set_writev_fop, dl.set_writev_cbk,
- dl.wind_writev, dl.unwind_writev)
-
-class OpOpendir:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(loc_t), POINTER(fd_t) ,POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(fd_t), POINTER(dict_t))
-_init_op (OpOpendir, dl.set_opendir_fop, dl.set_opendir_cbk,
- dl.wind_opendir, dl.unwind_opendir)
-
-class OpReaddir:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(fd_t), c_size_t, c_long, POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(gf_dirent_t), POINTER(dict_t))
-_init_op (OpReaddir, dl.set_readdir_fop, dl.set_readdir_cbk,
- dl.wind_readdir, dl.unwind_readdir)
-
-class OpReaddirp:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(fd_t), c_size_t, c_long, POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(gf_dirent_t), POINTER(dict_t))
-_init_op (OpReaddirp, dl.set_readdirp_fop, dl.set_readdirp_cbk,
- dl.wind_readdirp, dl.unwind_readdirp)
-
-class OpStat:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(loc_t), POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(iatt_t), POINTER(dict_t))
-_init_op (OpStat, dl.set_stat_fop, dl.set_stat_cbk,
- dl.wind_stat, dl.unwind_stat)
-
-class OpFstat:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(fd_t), POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(iatt_t), POINTER(dict_t))
-_init_op (OpFstat, dl.set_fstat_fop, dl.set_fstat_cbk,
- dl.wind_fstat, dl.unwind_fstat)
-
-class OpStatfs:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(loc_t), POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(statvfs_t), POINTER(dict_t))
-_init_op (OpStatfs, dl.set_statfs_fop, dl.set_statfs_cbk,
- dl.wind_statfs, dl.unwind_statfs)
-
-
-class OpSetxattr:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(loc_t), POINTER(dict_t), c_int32,
- POINTER (dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(dict_t))
-_init_op (OpSetxattr, dl.set_setxattr_fop, dl.set_setxattr_cbk,
- dl.wind_setxattr, dl.unwind_setxattr)
-
-class OpGetxattr:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(loc_t), c_char_p, POINTER (dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(dict_t), POINTER(dict_t))
-_init_op (OpGetxattr, dl.set_getxattr_fop, dl.set_getxattr_cbk,
- dl.wind_getxattr, dl.unwind_getxattr)
-
-class OpFsetxattr:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(fd_t), POINTER(dict_t), c_int32,
- POINTER (dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(dict_t))
-_init_op (OpFsetxattr, dl.set_fsetxattr_fop, dl.set_fsetxattr_cbk,
- dl.wind_fsetxattr, dl.unwind_fsetxattr)
-
-class OpFgetxattr:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(fd_t), c_char_p, POINTER (dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(dict_t), POINTER(dict_t))
-_init_op (OpFgetxattr, dl.set_fgetxattr_fop, dl.set_fgetxattr_cbk,
- dl.wind_fgetxattr, dl.unwind_fgetxattr)
-
-class OpRemovexattr:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(loc_t), c_char_p, POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(dict_t))
-_init_op (OpRemovexattr, dl.set_removexattr_fop, dl.set_removexattr_cbk,
- dl.wind_removexattr, dl.unwind_removexattr)
-
-
-class OpFremovexattr:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(fd_t), c_char_p, POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(dict_t))
-_init_op (OpFremovexattr, dl.set_fremovexattr_fop, dl.set_fremovexattr_cbk,
- dl.wind_fremovexattr, dl.unwind_fremovexattr)
-
-class OpLink:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(loc_t), POINTER(loc_t), POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(inode_t), POINTER(iatt_t),
- POINTER(iatt_t), POINTER(iatt_t), POINTER(dict_t))
-_init_op (OpLink, dl.set_link_fop, dl.set_link_cbk,
- dl.wind_link, dl.unwind_link)
-
-class OpSymlink:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- c_char_p, POINTER(loc_t), c_uint, POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(inode_t), POINTER(iatt_t),
- POINTER(iatt_t), POINTER(iatt_t), POINTER(dict_t))
-_init_op (OpSymlink, dl.set_symlink_fop, dl.set_symlink_cbk,
- dl.wind_symlink, dl.unwind_symlink)
-
-class OpUnlink:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(loc_t), c_int, POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(iatt_t), POINTER(iatt_t),
- POINTER(dict_t))
-_init_op (OpUnlink, dl.set_unlink_fop, dl.set_unlink_cbk,
- dl.wind_unlink, dl.unwind_unlink)
-
-class OpReadlink:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(loc_t), c_size_t, POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, c_char_p, POINTER(iatt_t), POINTER(dict_t))
-_init_op (OpReadlink, dl.set_readlink_fop, dl.set_readlink_cbk,
- dl.wind_readlink, dl.unwind_readlink)
-
-class OpMkdir:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(loc_t), c_uint, c_uint, POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(inode_t), POINTER(iatt_t),
- POINTER(iatt_t), POINTER(iatt_t), POINTER(dict_t))
-_init_op (OpMkdir, dl.set_mkdir_fop, dl.set_mkdir_cbk,
- dl.wind_mkdir, dl.unwind_mkdir)
-
-class OpRmdir:
- fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
- POINTER(loc_t), c_int, POINTER(dict_t))
- cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
- c_int, c_int, POINTER(iatt_t), POINTER(iatt_t),
- POINTER(dict_t))
-_init_op (OpRmdir, dl.set_rmdir_fop, dl.set_rmdir_cbk,
- dl.wind_rmdir, dl.unwind_rmdir)
-
-
-class Translator:
- def __init__ (self, c_this):
- # This is only here to keep references to the stubs we create,
- # because ctypes doesn't and glupy.so can't because it doesn't
- # get a pointer to the actual Python object. It's a dictionary
- # instead of a list in case we ever allow changing fops/cbks
- # after initialization and need to look them up.
- self.stub_refs = {}
- funcs = dir(self.__class__)
- if "lookup_fop" in funcs:
- @OpLookup.fop_type
- def stub (frame, this, loc, xdata, s=self):
- return s.lookup_fop (frame, this, loc, xdata)
- self.stub_refs["lookup_fop"] = stub
- dl.set_lookup_fop(c_this,stub)
- if "lookup_cbk" in funcs:
- @OpLookup.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno, inode,
- buf, xdata, postparent, s=self):
- return s.lookup_cbk(frame, cookie, this, op_ret,
- op_errno, inode, buf, xdata,
- postparent)
- self.stub_refs["lookup_cbk"] = stub
- dl.set_lookup_cbk(c_this,stub)
- if "create_fop" in funcs:
- @OpCreate.fop_type
- def stub (frame, this, loc, flags, mode, umask, fd,
- xdata, s=self):
- return s.create_fop (frame, this, loc, flags,
- mode, umask, fd, xdata)
- self.stub_refs["create_fop"] = stub
- dl.set_create_fop(c_this,stub)
- if "create_cbk" in funcs:
- @OpCreate.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno, fd,
- inode, buf, preparent, postparent, xdata,
- s=self):
- return s.create_cbk (frame, cookie, this,
- op_ret, op_errno, fd,
- inode, buf, preparent,
- postparent, xdata)
- self.stub_refs["create_cbk"] = stub
- dl.set_create_cbk(c_this,stub)
- if "open_fop" in funcs:
- @OpOpen.fop_type
- def stub (frame, this, loc, flags, fd,
- xdata, s=self):
- return s.open_fop (frame, this, loc, flags,
- fd, xdata)
- self.stub_refs["open_fop"] = stub
- dl.set_open_fop(c_this,stub)
- if "open_cbk" in funcs:
- @OpOpen.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno, fd,
- xdata, s=self):
- return s.open_cbk (frame, cookie, this,
- op_ret, op_errno, fd,
- xdata)
- self.stub_refs["open_cbk"] = stub
- dl.set_open_cbk(c_this,stub)
- if "readv_fop" in funcs:
- @OpReadv.fop_type
- def stub (frame, this, fd, size, offset, flags,
- xdata, s=self):
- return s.readv_fop (frame, this, fd, size,
- offset, flags, xdata)
- self.stub_refs["readv_fop"] = stub
- dl.set_readv_fop(c_this,stub)
- if "readv_cbk" in funcs:
- @OpReadv.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno,
- vector, count, stbuf, iobref, xdata,
- s=self):
- return s.readv_cbk (frame, cookie, this,
- op_ret, op_errno, vector,
- count, stbuf, iobref,
- xdata)
- self.stub_refs["readv_cbk"] = stub
- dl.set_readv_cbk(c_this,stub)
- if "writev_fop" in funcs:
- @OpWritev.fop_type
- def stub (frame, this, fd, vector, count,
- offset, flags, iobref, xdata, s=self):
- return s.writev_fop (frame, this, fd, vector,
- count, offset, flags,
- iobref, xdata)
- self.stub_refs["writev_fop"] = stub
- dl.set_writev_fop(c_this,stub)
- if "writev_cbk" in funcs:
- @OpWritev.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno,
- prebuf, postbuf, xdata, s=self):
- return s.writev_cbk (frame, cookie, this,
- op_ret, op_errno, prebuf,
- postbuf, xdata)
- self.stub_refs["writev_cbk"] = stub
- dl.set_writev_cbk(c_this,stub)
- if "opendir_fop" in funcs:
- @OpOpendir.fop_type
- def stub (frame, this, loc, fd, xdata, s=self):
- return s.opendir_fop (frame, this, loc, fd,
- xdata)
- self.stub_refs["opendir_fop"] = stub
- dl.set_opendir_fop(c_this,stub)
- if "opendir_cbk" in funcs:
- @OpOpendir.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno, fd,
- xdata, s=self):
- return s.opendir_cbk(frame, cookie, this,
- op_ret, op_errno, fd,
- xdata)
- self.stub_refs["opendir_cbk"] = stub
- dl.set_opendir_cbk(c_this,stub)
- if "readdir_fop" in funcs:
- @OpReaddir.fop_type
- def stub (frame, this, fd, size, offset, xdata, s=self):
- return s.readdir_fop (frame, this, fd, size,
- offset, xdata)
- self.stub_refs["readdir_fop"] = stub
- dl.set_readdir_fop(c_this,stub)
- if "readdir_cbk" in funcs:
- @OpReaddir.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno,
- entries, xdata, s=self):
- return s.readdir_cbk(frame, cookie, this,
- op_ret, op_errno, entries,
- xdata)
- self.stub_refs["readdir_cbk"] = stub
- dl.set_readdir_cbk(c_this,stub)
- if "readdirp_fop" in funcs:
- @OpReaddirp.fop_type
- def stub (frame, this, fd, size, offset, xdata, s=self):
- return s.readdirp_fop (frame, this, fd, size,
- offset, xdata)
- self.stub_refs["readdirp_fop"] = stub
- dl.set_readdirp_fop(c_this,stub)
- if "readdirp_cbk" in funcs:
- @OpReaddirp.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno,
- entries, xdata, s=self):
- return s.readdirp_cbk (frame, cookie, this,
- op_ret, op_errno,
- entries, xdata)
- self.stub_refs["readdirp_cbk"] = stub
- dl.set_readdirp_cbk(c_this,stub)
- if "stat_fop" in funcs:
- @OpStat.fop_type
- def stub (frame, this, loc, xdata, s=self):
- return s.stat_fop (frame, this, loc, xdata)
- self.stub_refs["stat_fop"] = stub
- dl.set_stat_fop(c_this,stub)
- if "stat_cbk" in funcs:
- @OpStat.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno, buf,
- xdata, s=self):
- return s.stat_cbk(frame, cookie, this, op_ret,
- op_errno, buf, xdata)
- self.stub_refs["stat_cbk"] = stub
- dl.set_stat_cbk(c_this,stub)
- if "fstat_fop" in funcs:
- @OpFstat.fop_type
- def stub (frame, this, fd, xdata, s=self):
- return s.fstat_fop (frame, this, fd, xdata)
- self.stub_refs["fstat_fop"] = stub
- dl.set_fstat_fop(c_this,stub)
- if "fstat_cbk" in funcs:
- @OpFstat.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno, buf,
- xdata, s=self):
- return s.fstat_cbk(frame, cookie, this, op_ret,
- op_errno, buf, xdata)
- self.stub_refs["fstat_cbk"] = stub
- dl.set_fstat_cbk(c_this,stub)
- if "statfs_fop" in funcs:
- @OpStatfs.fop_type
- def stub (frame, this, loc, xdata, s=self):
- return s.statfs_fop (frame, this, loc, xdata)
- self.stub_refs["statfs_fop"] = stub
- dl.set_statfs_fop(c_this,stub)
- if "statfs_cbk" in funcs:
- @OpStatfs.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno, buf,
- xdata, s=self):
- return s.statfs_cbk (frame, cookie, this,
- op_ret, op_errno, buf,
- xdata)
- self.stub_refs["statfs_cbk"] = stub
- dl.set_statfs_cbk(c_this,stub)
- if "setxattr_fop" in funcs:
- @OpSetxattr.fop_type
- def stub (frame, this, loc, dictionary, flags, xdata,
- s=self):
- return s.setxattr_fop (frame, this, loc,
- dictionary, flags,
- xdata)
- self.stub_refs["setxattr_fop"] = stub
- dl.set_setxattr_fop(c_this,stub)
- if "setxattr_cbk" in funcs:
- @OpSetxattr.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno, xdata,
- s=self):
- return s.setxattr_cbk(frame, cookie, this,
- op_ret, op_errno, xdata)
- self.stub_refs["setxattr_cbk"] = stub
- dl.set_setxattr_cbk(c_this,stub)
- if "getxattr_fop" in funcs:
- @OpGetxattr.fop_type
- def stub (frame, this, loc, name, xdata, s=self):
- return s.getxattr_fop (frame, this, loc, name,
- xdata)
- self.stub_refs["getxattr_fop"] = stub
- dl.set_getxattr_fop(c_this,stub)
- if "getxattr_cbk" in funcs:
- @OpGetxattr.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno,
- dictionary, xdata, s=self):
- return s.getxattr_cbk(frame, cookie, this,
- op_ret, op_errno,
- dictionary, xdata)
- self.stub_refs["getxattr_cbk"] = stub
- dl.set_getxattr_cbk(c_this,stub)
- if "fsetxattr_fop" in funcs:
- @OpFsetxattr.fop_type
- def stub (frame, this, fd, dictionary, flags, xdata,
- s=self):
- return s.fsetxattr_fop (frame, this, fd,
- dictionary, flags,
- xdata)
- self.stub_refs["fsetxattr_fop"] = stub
- dl.set_fsetxattr_fop(c_this,stub)
- if "fsetxattr_cbk" in funcs:
- @OpFsetxattr.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno, xdata,
- s=self):
- return s.fsetxattr_cbk(frame, cookie, this,
- op_ret, op_errno, xdata)
- self.stub_refs["fsetxattr_cbk"] = stub
- dl.set_fsetxattr_cbk(c_this,stub)
- if "fgetxattr_fop" in funcs:
- @OpFgetxattr.fop_type
- def stub (frame, this, fd, name, xdata, s=self):
- return s.fgetxattr_fop (frame, this, fd, name,
- xdata)
- self.stub_refs["fgetxattr_fop"] = stub
- dl.set_fgetxattr_fop(c_this,stub)
- if "fgetxattr_cbk" in funcs:
- @OpFgetxattr.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno,
- dictionary, xdata, s=self):
- return s.fgetxattr_cbk(frame, cookie, this,
- op_ret, op_errno,
- dictionary, xdata)
- self.stub_refs["fgetxattr_cbk"] = stub
- dl.set_fgetxattr_cbk(c_this,stub)
- if "removexattr_fop" in funcs:
- @OpRemovexattr.fop_type
- def stub (frame, this, loc, name, xdata, s=self):
- return s.removexattr_fop (frame, this, loc,
- name, xdata)
- self.stub_refs["removexattr_fop"] = stub
- dl.set_removexattr_fop(c_this,stub)
- if "removexattr_cbk" in funcs:
- @OpRemovexattr.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno,
- xdata, s=self):
- return s.removexattr_cbk(frame, cookie, this,
- op_ret, op_errno,
- xdata)
- self.stub_refs["removexattr_cbk"] = stub
- dl.set_removexattr_cbk(c_this,stub)
- if "fremovexattr_fop" in funcs:
- @OpFremovexattr.fop_type
- def stub (frame, this, fd, name, xdata, s=self):
- return s.fremovexattr_fop (frame, this, fd,
- name, xdata)
- self.stub_refs["fremovexattr_fop"] = stub
- dl.set_fremovexattr_fop(c_this,stub)
- if "fremovexattr_cbk" in funcs:
- @OpFremovexattr.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno,
- xdata, s=self):
- return s.fremovexattr_cbk(frame, cookie, this,
- op_ret, op_errno,
- xdata)
- self.stub_refs["fremovexattr_cbk"] = stub
- dl.set_fremovexattr_cbk(c_this,stub)
- if "link_fop" in funcs:
- @OpLink.fop_type
- def stub (frame, this, oldloc, newloc,
- xdata, s=self):
- return s.link_fop (frame, this, oldloc,
- newloc, xdata)
- self.stub_refs["link_fop"] = stub
- dl.set_link_fop(c_this,stub)
- if "link_cbk" in funcs:
- @OpLink.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno,
- inode, buf, preparent, postparent, xdata,
- s=self):
- return s.link_cbk (frame, cookie, this,
- op_ret, op_errno, inode,
- buf, preparent,
- postparent, xdata)
- self.stub_refs["link_cbk"] = stub
- dl.set_link_cbk(c_this,stub)
- if "symlink_fop" in funcs:
- @OpSymlink.fop_type
- def stub (frame, this, linkname, loc,
- umask, xdata, s=self):
- return s.symlink_fop (frame, this, linkname,
- loc, umask, xdata)
- self.stub_refs["symlink_fop"] = stub
- dl.set_symlink_fop(c_this,stub)
- if "symlink_cbk" in funcs:
- @OpSymlink.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno,
- inode, buf, preparent, postparent, xdata,
- s=self):
- return s.symlink_cbk (frame, cookie, this,
- op_ret, op_errno, inode,
- buf, preparent,
- postparent, xdata)
- self.stub_refs["symlink_cbk"] = stub
- dl.set_symlink_cbk(c_this,stub)
- if "unlink_fop" in funcs:
- @OpUnlink.fop_type
- def stub (frame, this, loc, xflags,
- xdata, s=self):
- return s.unlink_fop (frame, this, loc,
- xflags, xdata)
- self.stub_refs["unlink_fop"] = stub
- dl.set_unlink_fop(c_this,stub)
- if "unlink_cbk" in funcs:
- @OpUnlink.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno,
- preparent, postparent, xdata, s=self):
- return s.unlink_cbk (frame, cookie, this,
- op_ret, op_errno,
- preparent, postparent,
- xdata)
- self.stub_refs["unlink_cbk"] = stub
- dl.set_unlink_cbk(c_this,stub)
- if "readlink_fop" in funcs:
- @OpReadlink.fop_type
- def stub (frame, this, loc, size,
- xdata, s=self):
- return s.readlink_fop (frame, this, loc,
- size, xdata)
- self.stub_refs["readlink_fop"] = stub
- dl.set_readlink_fop(c_this,stub)
- if "readlink_cbk" in funcs:
- @OpReadlink.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno,
- path, buf, xdata, s=self):
- return s.readlink_cbk (frame, cookie, this,
- op_ret, op_errno,
- path, buf, xdata)
- self.stub_refs["readlink_cbk"] = stub
- dl.set_readlink_cbk(c_this,stub)
- if "mkdir_fop" in funcs:
- @OpMkdir.fop_type
- def stub (frame, this, loc, mode, umask, xdata,
- s=self):
- return s.mkdir_fop (frame, this, loc, mode,
- umask, xdata)
- self.stub_refs["mkdir_fop"] = stub
- dl.set_mkdir_fop(c_this,stub)
- if "mkdir_cbk" in funcs:
- @OpMkdir.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno, inode,
- buf, preparent, postparent, xdata, s=self):
- return s.mkdir_cbk (frame, cookie, this,
- op_ret, op_errno, inode,
- buf, preparent,
- postparent, xdata)
- self.stub_refs["mkdir_cbk"] = stub
- dl.set_mkdir_cbk(c_this,stub)
- if "rmdir_fop" in funcs:
- @OpRmdir.fop_type
- def stub (frame, this, loc, xflags,
- xdata, s=self):
- return s.rmdir_fop (frame, this, loc,
- xflags, xdata)
- self.stub_refs["rmdir_fop"] = stub
- dl.set_rmdir_fop(c_this,stub)
- if "rmdir_cbk" in funcs:
- @OpRmdir.cbk_type
- def stub (frame, cookie, this, op_ret, op_errno,
- preparent, postparent, xdata, s=self):
- return s.rmdir_cbk (frame, cookie, this,
- op_ret, op_errno,
- preparent, postparent,
- xdata)
- self.stub_refs["rmdir_cbk"] = stub
- dl.set_rmdir_cbk(c_this,stub)
diff --git a/xlators/features/glupy/src/setup.py.in b/xlators/features/glupy/src/setup.py.in
deleted file mode 100644
index 611e9695f76..00000000000
--- a/xlators/features/glupy/src/setup.py.in
+++ /dev/null
@@ -1,24 +0,0 @@
-from distutils.core import setup
-
-DESC = """GlusterFS is a distributed file-system capable of scaling to
-several petabytes. It aggregates various storage bricks over Infiniband
-RDMA or TCP/IP interconnect into one large parallel network file system.
-GlusterFS is one of the most sophisticated file systems in terms of
-features and extensibility. It borrows a powerful concept called
-Translators from GNU Hurd kernel. Much of the code in GlusterFS is in
-user space and easily manageable.
-
-This package contains Glupy, the Python translator interface for GlusterFS."""
-
-setup(
- name='glusterfs-glupy',
- version='@PACKAGE_VERSION@',
- description='Glupy is the Python translator interface for GlusterFS',
- long_description=DESC,
- author='Gluster Community',
- author_email='gluster-devel@gluster.org',
- license='LGPLv3',
- url='http://gluster.org/',
- package_dir={'gluster':''},
- packages=['gluster']
-)
diff --git a/xlators/features/index/src/index-mem-types.h b/xlators/features/index/src/index-mem-types.h
index ca291cfba7e..58833d0ec9b 100644
--- a/xlators/features/index/src/index-mem-types.h
+++ b/xlators/features/index/src/index-mem-types.h
@@ -8,16 +8,16 @@
cases as published by the Free Software Foundation.
*/
-#ifndef __QUIESCE_MEM_TYPES_H__
-#define __QUIESCE_MEM_TYPES_H__
+#ifndef __INDEX_MEM_TYPES_H__
+#define __INDEX_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_index_mem_types_ {
- gf_index_mt_priv_t = gf_common_mt_end + 1,
- gf_index_inode_ctx_t = gf_common_mt_end + 2,
- gf_index_fd_ctx_t = gf_common_mt_end + 3,
- gf_index_mt_local_t = gf_common_mt_end + 4,
- gf_index_mt_end
+ gf_index_mt_priv_t = gf_common_mt_end + 1,
+ gf_index_inode_ctx_t,
+ gf_index_fd_ctx_t,
+ gf_index_mt_local_t,
+ gf_index_mt_end
};
#endif
diff --git a/xlators/features/index/src/index-messages.h b/xlators/features/index/src/index-messages.h
index 7ab2bb0ad4b..364f17cd34e 100644
--- a/xlators/features/index/src/index-messages.h
+++ b/xlators/features/index/src/index-messages.h
@@ -11,7 +11,7 @@
#ifndef _INDEX_MESSAGES_H_
#define _INDEX_MESSAGES_H_
-#include "glfs-message-id.h"
+#include <glusterfs/glfs-message-id.h>
/* To add new message IDs, append new identifiers at the end of the list.
*
@@ -23,17 +23,11 @@
* glfs-message-id.h.
*/
-GLFS_MSGID(INDEX,
- INDEX_MSG_INDEX_DIR_CREATE_FAILED,
- INDEX_MSG_INDEX_READDIR_FAILED,
- INDEX_MSG_INDEX_ADD_FAILED,
- INDEX_MSG_INDEX_DEL_FAILED,
- INDEX_MSG_DICT_SET_FAILED,
- INDEX_MSG_INODE_CTX_GET_SET_FAILED,
- INDEX_MSG_INVALID_ARGS,
- INDEX_MSG_FD_OP_FAILED,
- INDEX_MSG_WORKER_THREAD_CREATE_FAILED,
- INDEX_MSG_INVALID_GRAPH
-);
+GLFS_MSGID(INDEX, INDEX_MSG_INDEX_DIR_CREATE_FAILED,
+ INDEX_MSG_INDEX_READDIR_FAILED, INDEX_MSG_INDEX_ADD_FAILED,
+ INDEX_MSG_INDEX_DEL_FAILED, INDEX_MSG_DICT_SET_FAILED,
+ INDEX_MSG_INODE_CTX_GET_SET_FAILED, INDEX_MSG_INVALID_ARGS,
+ INDEX_MSG_FD_OP_FAILED, INDEX_MSG_WORKER_THREAD_CREATE_FAILED,
+ INDEX_MSG_INVALID_GRAPH);
#endif /* !_INDEX_MESSAGES_H_ */
diff --git a/xlators/features/index/src/index.c b/xlators/features/index/src/index.c
index 152b52dfc33..4abb2c73ce5 100644
--- a/xlators/features/index/src/index.c
+++ b/xlators/features/index/src/index.c
@@ -8,11 +8,11 @@
cases as published by the Free Software Foundation.
*/
#include "index.h"
-#include "options.h"
+#include <glusterfs/options.h>
#include "glusterfs3-xdr.h"
-#include "syscall.h"
-#include "syncop.h"
-#include "common-utils.h"
+#include <glusterfs/syscall.h>
+#include <glusterfs/syncop.h>
+#include <glusterfs/common-utils.h>
#include "index-messages.h"
#include <ftw.h>
#include <libgen.h> /* for dirname() */
@@ -23,2569 +23,2660 @@
#define ENTRY_CHANGES_SUBDIR "entry-changes"
struct index_syncop_args {
- inode_t *parent;
- gf_dirent_t *entries;
- char *path;
+ inode_t *parent;
+ gf_dirent_t *entries;
+ char *path;
};
static char *index_vgfid_xattrs[XATTROP_TYPE_END] = {
- [XATTROP] = GF_XATTROP_INDEX_GFID,
- [DIRTY] = GF_XATTROP_DIRTY_GFID,
- [ENTRY_CHANGES] = GF_XATTROP_ENTRY_CHANGES_GFID
-};
+ [XATTROP] = GF_XATTROP_INDEX_GFID,
+ [DIRTY] = GF_XATTROP_DIRTY_GFID,
+ [ENTRY_CHANGES] = GF_XATTROP_ENTRY_CHANGES_GFID};
static char *index_subdirs[XATTROP_TYPE_END] = {
- [XATTROP] = XATTROP_SUBDIR,
- [DIRTY] = DIRTY_SUBDIR,
- [ENTRY_CHANGES] = ENTRY_CHANGES_SUBDIR
-};
+ [XATTROP] = XATTROP_SUBDIR,
+ [DIRTY] = DIRTY_SUBDIR,
+ [ENTRY_CHANGES] = ENTRY_CHANGES_SUBDIR};
int
-index_get_type_from_vgfid (index_priv_t *priv, uuid_t vgfid)
+index_get_type_from_vgfid(index_priv_t *priv, uuid_t vgfid)
{
- int i = 0;
+ int i = 0;
- for (i = 0; i < XATTROP_TYPE_END; i++) {
- if (gf_uuid_compare (priv->internal_vgfid[i], vgfid) == 0)
- return i;
- }
- return -1;
+ for (i = 0; i < XATTROP_TYPE_END; i++) {
+ if (gf_uuid_compare(priv->internal_vgfid[i], vgfid) == 0)
+ return i;
+ }
+ return -1;
}
gf_boolean_t
-index_is_virtual_gfid (index_priv_t *priv, uuid_t vgfid)
+index_is_virtual_gfid(index_priv_t *priv, uuid_t vgfid)
{
- if (index_get_type_from_vgfid (priv, vgfid) < 0)
- return _gf_false;
- return _gf_true;
+ if (index_get_type_from_vgfid(priv, vgfid) < 0)
+ return _gf_false;
+ return _gf_true;
}
static int
-__index_inode_ctx_get (inode_t *inode, xlator_t *this, index_inode_ctx_t **ctx)
-{
- int ret = 0;
- index_inode_ctx_t *ictx = NULL;
- uint64_t tmpctx = 0;
-
- ret = __inode_ctx_get (inode, this, &tmpctx);
- if (!ret) {
- ictx = (index_inode_ctx_t *) (long) tmpctx;
- goto out;
- }
- ictx = GF_CALLOC (1, sizeof (*ictx), gf_index_inode_ctx_t);
- if (!ictx) {
- ret = -1;
- goto out;
- }
-
- INIT_LIST_HEAD (&ictx->callstubs);
- ret = __inode_ctx_put (inode, this, (uint64_t)ictx);
- if (ret) {
- GF_FREE (ictx);
- ictx = NULL;
- goto out;
- }
+__index_inode_ctx_get(inode_t *inode, xlator_t *this, index_inode_ctx_t **ctx)
+{
+ int ret = 0;
+ index_inode_ctx_t *ictx = NULL;
+ uint64_t tmpctx = 0;
+
+ ret = __inode_ctx_get(inode, this, &tmpctx);
+ if (!ret) {
+ ictx = (index_inode_ctx_t *)(long)tmpctx;
+ goto out;
+ }
+ ictx = GF_CALLOC(1, sizeof(*ictx), gf_index_inode_ctx_t);
+ if (!ictx) {
+ ret = -1;
+ goto out;
+ }
+
+ INIT_LIST_HEAD(&ictx->callstubs);
+ ret = __inode_ctx_put(inode, this, (uint64_t)(uintptr_t)ictx);
+ if (ret) {
+ GF_FREE(ictx);
+ ictx = NULL;
+ goto out;
+ }
out:
- if (ictx)
- *ctx = ictx;
- return ret;
+ if (ictx)
+ *ctx = ictx;
+ return ret;
}
static int
-index_inode_ctx_get (inode_t *inode, xlator_t *this, index_inode_ctx_t **ctx)
+index_inode_ctx_get(inode_t *inode, xlator_t *this, index_inode_ctx_t **ctx)
{
- int ret = 0;
+ int ret = 0;
- LOCK (&inode->lock);
- {
- ret = __index_inode_ctx_get (inode, this, ctx);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __index_inode_ctx_get(inode, this, ctx);
+ }
+ UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
static gf_boolean_t
-index_is_subdir_of_entry_changes (xlator_t *this, inode_t *inode)
+index_is_subdir_of_entry_changes(xlator_t *this, inode_t *inode)
{
- index_inode_ctx_t *ctx = NULL;
- int ret = 0;
+ index_inode_ctx_t *ctx = NULL;
+ int ret = 0;
- if (!inode)
- return _gf_false;
-
- ret = index_inode_ctx_get (inode, this, &ctx);
- if ((ret == 0) && !gf_uuid_is_null (ctx->virtual_pargfid))
- return _gf_true;
+ if (!inode)
return _gf_false;
+
+ ret = index_inode_ctx_get(inode, this, &ctx);
+ if ((ret == 0) && !gf_uuid_is_null(ctx->virtual_pargfid))
+ return _gf_true;
+ return _gf_false;
}
static int
-index_get_type_from_vgfid_xattr (const char *name)
+index_get_type_from_vgfid_xattr(const char *name)
{
- int i = 0;
+ int i = 0;
- for (i = 0; i < XATTROP_TYPE_END; i++) {
- if (strcmp (name, index_vgfid_xattrs[i]) == 0)
- return i;
- }
- return -1;
+ for (i = 0; i < XATTROP_TYPE_END; i++) {
+ if (strcmp(name, index_vgfid_xattrs[i]) == 0)
+ return i;
+ }
+ return -1;
}
gf_boolean_t
-index_is_fop_on_internal_inode (xlator_t *this, inode_t *inode, uuid_t gfid)
+index_is_fop_on_internal_inode(xlator_t *this, inode_t *inode, uuid_t gfid)
{
- index_priv_t *priv = this->private;
- uuid_t vgfid = {0};
+ index_priv_t *priv = this->private;
+ uuid_t vgfid = {0};
- if (!inode)
- return _gf_false;
+ if (!inode)
+ return _gf_false;
- if (gfid && !gf_uuid_is_null (gfid))
- gf_uuid_copy (vgfid, gfid);
- else
- gf_uuid_copy (vgfid, inode->gfid);
+ if (gfid && !gf_uuid_is_null(gfid))
+ gf_uuid_copy(vgfid, gfid);
+ else
+ gf_uuid_copy(vgfid, inode->gfid);
- if (index_is_virtual_gfid (priv, vgfid))
- return _gf_true;
- if (index_is_subdir_of_entry_changes (this, inode))
- return _gf_true;
- return _gf_false;
+ if (index_is_virtual_gfid(priv, vgfid))
+ return _gf_true;
+ if (index_is_subdir_of_entry_changes(this, inode))
+ return _gf_true;
+ return _gf_false;
}
static gf_boolean_t
-index_is_vgfid_xattr (const char *name)
+index_is_vgfid_xattr(const char *name)
{
- if (index_get_type_from_vgfid_xattr (name) < 0)
- return _gf_false;
- return _gf_true;
+ if (index_get_type_from_vgfid_xattr(name) < 0)
+ return _gf_false;
+ return _gf_true;
}
call_stub_t *
-__index_dequeue (struct list_head *callstubs)
+__index_dequeue(struct list_head *callstubs)
{
- call_stub_t *stub = NULL;
+ call_stub_t *stub = NULL;
- if (!list_empty (callstubs)) {
- stub = list_entry (callstubs->next, call_stub_t, list);
- list_del_init (&stub->list);
- }
+ if (!list_empty(callstubs)) {
+ stub = list_entry(callstubs->next, call_stub_t, list);
+ list_del_init(&stub->list);
+ }
- return stub;
+ return stub;
}
static void
-__index_enqueue (struct list_head *callstubs, call_stub_t *stub)
+__index_enqueue(struct list_head *callstubs, call_stub_t *stub)
{
- list_add_tail (&stub->list, callstubs);
+ list_add_tail(&stub->list, callstubs);
}
static void
-worker_enqueue (xlator_t *this, call_stub_t *stub)
+worker_enqueue(xlator_t *this, call_stub_t *stub)
{
- index_priv_t *priv = NULL;
+ index_priv_t *priv = NULL;
- priv = this->private;
- pthread_mutex_lock (&priv->mutex);
- {
- __index_enqueue (&priv->callstubs, stub);
- pthread_cond_signal (&priv->cond);
- }
- pthread_mutex_unlock (&priv->mutex);
+ priv = this->private;
+ pthread_mutex_lock(&priv->mutex);
+ {
+ __index_enqueue(&priv->callstubs, stub);
+ GF_ATOMIC_INC(priv->stub_cnt);
+ pthread_cond_signal(&priv->cond);
+ }
+ pthread_mutex_unlock(&priv->mutex);
}
void *
-index_worker (void *data)
-{
- index_priv_t *priv = NULL;
- xlator_t *this = NULL;
- call_stub_t *stub = NULL;
- gf_boolean_t bye = _gf_false;
-
-
- THIS = data;
- this = data;
- priv = this->private;
-
- for (;;) {
- pthread_mutex_lock (&priv->mutex);
- {
- while (list_empty (&priv->callstubs)) {
- if (priv->down) {
- bye = _gf_true;/*Avoid wait*/
- break;
- }
- (void) pthread_cond_wait (&priv->cond,
- &priv->mutex);
- if (priv->down) {
- bye = _gf_true;
- break;
- }
- }
- if (!bye)
- stub = __index_dequeue (&priv->callstubs);
+index_worker(void *data)
+{
+ index_priv_t *priv = NULL;
+ xlator_t *this = NULL;
+ call_stub_t *stub = NULL;
+ gf_boolean_t bye = _gf_false;
+
+ THIS = data;
+ this = data;
+ priv = this->private;
+
+ for (;;) {
+ pthread_mutex_lock(&priv->mutex);
+ {
+ while (list_empty(&priv->callstubs)) {
+ if (priv->down) {
+ bye = _gf_true; /*Avoid wait*/
+ break;
}
- pthread_mutex_unlock (&priv->mutex);
+ (void)pthread_cond_wait(&priv->cond, &priv->mutex);
+ if (priv->down) {
+ bye = _gf_true;
+ break;
+ }
+ }
+ if (!bye)
+ stub = __index_dequeue(&priv->callstubs);
+ if (bye) {
+ priv->curr_count--;
+ if (priv->curr_count == 0)
+ pthread_cond_broadcast(&priv->cond);
+ }
+ }
+ pthread_mutex_unlock(&priv->mutex);
- if (stub) /* guard against spurious wakeups */
- call_resume (stub);
- stub = NULL;
- if (bye)
- break;
+ if (stub) { /* guard against spurious wakeups */
+ call_resume(stub);
+ GF_ATOMIC_DEC(priv->stub_cnt);
}
+ stub = NULL;
+ if (bye)
+ break;
+ }
- return NULL;
+ return NULL;
}
static void
-make_index_dir_path (char *base, const char *subdir,
- char *index_dir, size_t len)
+make_index_dir_path(char *base, const char *subdir, char *index_dir, size_t len)
{
- snprintf (index_dir, len, "%s/%s", base, subdir);
+ snprintf(index_dir, len, "%s/%s", base, subdir);
}
int
-index_dir_create (xlator_t *this, const char *subdir)
-{
- int ret = 0;
- struct stat st = {0};
- char fullpath[PATH_MAX] = {0};
- char path[PATH_MAX] = {0};
- char *dir = NULL;
- index_priv_t *priv = NULL;
- size_t len = 0;
- size_t pathlen = 0;
-
- priv = this->private;
- make_index_dir_path (priv->index_basepath, subdir, fullpath,
- sizeof (fullpath));
- ret = sys_stat (fullpath, &st);
- if (!ret) {
- if (!S_ISDIR (st.st_mode))
- ret = -2;
- goto out;
- }
-
- pathlen = strlen (fullpath);
- if ((pathlen > 1) && fullpath[pathlen - 1] == '/')
- fullpath[pathlen - 1] = '\0';
- dir = strchr (fullpath, '/');
- while (dir) {
- dir = strchr (dir + 1, '/');
- if (dir)
- len = pathlen - strlen (dir);
- else
- len = pathlen;
- strncpy (path, fullpath, len);
- path[len] = '\0';
- ret = sys_mkdir (path, 0600);
- if (ret && (errno != EEXIST))
- goto out;
- }
- ret = 0;
+index_dir_create(xlator_t *this, const char *subdir)
+{
+ int ret = 0;
+ struct stat st = {0};
+ char fullpath[PATH_MAX] = {0};
+ char path[PATH_MAX] = {0};
+ char *dir = NULL;
+ index_priv_t *priv = NULL;
+ size_t len = 0;
+ size_t pathlen = 0;
+
+ priv = this->private;
+ make_index_dir_path(priv->index_basepath, subdir, fullpath,
+ sizeof(fullpath));
+ ret = sys_stat(fullpath, &st);
+ if (!ret) {
+ if (!S_ISDIR(st.st_mode))
+ ret = -2;
+ goto out;
+ }
+
+ pathlen = strlen(fullpath);
+ if ((pathlen > 1) && fullpath[pathlen - 1] == '/')
+ fullpath[pathlen - 1] = '\0';
+ dir = strchr(fullpath, '/');
+ while (dir) {
+ dir = strchr(dir + 1, '/');
+ if (dir)
+ len = pathlen - strlen(dir);
+ else
+ len = pathlen;
+ strncpy(path, fullpath, len);
+ path[len] = '\0';
+ ret = sys_mkdir(path, 0600);
+ if (ret && (errno != EEXIST))
+ goto out;
+ }
+ ret = 0;
out:
- if (ret == -1) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- INDEX_MSG_INDEX_DIR_CREATE_FAILED, "%s/%s: Failed to "
- "create", priv->index_basepath, subdir);
- } else if (ret == -2) {
- gf_msg (this->name, GF_LOG_ERROR, ENOTDIR,
- INDEX_MSG_INDEX_DIR_CREATE_FAILED, "%s/%s: Failed to "
- "create, path exists, not a directory ",
- priv->index_basepath, subdir);
- }
- return ret;
+ if (ret == -1) {
+ gf_msg(this->name, GF_LOG_ERROR, errno,
+ INDEX_MSG_INDEX_DIR_CREATE_FAILED,
+ "%s/%s: Failed to "
+ "create",
+ priv->index_basepath, subdir);
+ } else if (ret == -2) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOTDIR,
+ INDEX_MSG_INDEX_DIR_CREATE_FAILED,
+ "%s/%s: Failed to "
+ "create, path exists, not a directory ",
+ priv->index_basepath, subdir);
+ }
+ return ret;
}
void
-index_get_index (index_priv_t *priv, uuid_t index)
+index_get_index(index_priv_t *priv, uuid_t index)
{
- LOCK (&priv->lock);
- {
- gf_uuid_copy (index, priv->index);
- }
- UNLOCK (&priv->lock);
+ LOCK(&priv->lock);
+ {
+ gf_uuid_copy(index, priv->index);
+ }
+ UNLOCK(&priv->lock);
}
void
-index_generate_index (index_priv_t *priv, uuid_t index)
+index_generate_index(index_priv_t *priv, uuid_t index)
{
- LOCK (&priv->lock);
- {
- //To prevent duplicate generates.
- //This method fails if number of contending threads is greater
- //than MAX_LINK count of the fs
- if (!gf_uuid_compare (priv->index, index))
- gf_uuid_generate (priv->index);
- gf_uuid_copy (index, priv->index);
- }
- UNLOCK (&priv->lock);
+ LOCK(&priv->lock);
+ {
+ // To prevent duplicate generates.
+ // This method fails if number of contending threads is greater
+ // than MAX_LINK count of the fs
+ if (!gf_uuid_compare(priv->index, index))
+ gf_uuid_generate(priv->index);
+ gf_uuid_copy(index, priv->index);
+ }
+ UNLOCK(&priv->lock);
}
static void
-make_index_path (char *base, const char *subdir, uuid_t index,
- char *index_path, size_t len)
+make_index_path(char *base, const char *subdir, uuid_t index, char *index_path,
+ size_t len)
{
- make_index_dir_path (base, subdir, index_path, len);
- snprintf (index_path + strlen (index_path), len - strlen (index_path),
- "/%s-%s", subdir, uuid_utoa (index));
+ make_index_dir_path(base, subdir, index_path, len);
+ snprintf(index_path + strlen(index_path), len - strlen(index_path),
+ "/%s-%s", subdir, uuid_utoa(index));
}
static void
-make_gfid_path (char *base, const char *subdir, uuid_t gfid,
- char *gfid_path, size_t len)
+make_gfid_path(char *base, const char *subdir, uuid_t gfid, char *gfid_path,
+ size_t len)
{
- make_index_dir_path (base, subdir, gfid_path, len);
- snprintf (gfid_path + strlen (gfid_path), len - strlen (gfid_path),
- "/%s", uuid_utoa (gfid));
+ make_index_dir_path(base, subdir, gfid_path, len);
+ snprintf(gfid_path + strlen(gfid_path), len - strlen(gfid_path), "/%s",
+ uuid_utoa(gfid));
}
static void
-make_file_path (char *base, const char *subdir, const char *filename,
- char *file_path, size_t len)
+make_file_path(char *base, const char *subdir, const char *filename,
+ char *file_path, size_t len)
{
- make_index_dir_path (base, subdir, file_path, len);
- snprintf (file_path + strlen (file_path), len - strlen (file_path),
- "/%s", filename);
+ make_index_dir_path(base, subdir, file_path, len);
+ snprintf(file_path + strlen(file_path), len - strlen(file_path), "/%s",
+ filename);
}
static int
-is_index_file_current (char *filename, uuid_t priv_index, char *subdir)
+is_index_file_current(char *filename, uuid_t priv_index, char *subdir)
{
- char current_index[GF_UUID_BUF_SIZE + 16] = {0, };
+ char current_index[GF_UUID_BUF_SIZE + 16] = {
+ 0,
+ };
- snprintf (current_index, sizeof current_index,
- "%s-%s", subdir, uuid_utoa(priv_index));
- return (!strcmp(filename, current_index));
+ snprintf(current_index, sizeof current_index, "%s-%s", subdir,
+ uuid_utoa(priv_index));
+ return (!strcmp(filename, current_index));
}
static void
-check_delete_stale_index_file (xlator_t *this, char *filename, char *subdir)
+check_delete_stale_index_file(xlator_t *this, char *filename, char *subdir)
{
- int ret = 0;
- struct stat st = {0};
- char filepath[PATH_MAX] = {0};
- index_priv_t *priv = NULL;
+ int ret = 0;
+ struct stat st = {0};
+ char filepath[PATH_MAX] = {0};
+ index_priv_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (is_index_file_current (filename, priv->index, subdir))
- return;
+ if (is_index_file_current(filename, priv->index, subdir))
+ return;
- make_file_path (priv->index_basepath, subdir,
- filename, filepath, sizeof (filepath));
- ret = sys_stat (filepath, &st);
- if (!ret && st.st_nlink == 1)
- sys_unlink (filepath);
+ make_file_path(priv->index_basepath, subdir, filename, filepath,
+ sizeof(filepath));
+ ret = sys_stat(filepath, &st);
+ if (!ret && st.st_nlink == 1)
+ sys_unlink(filepath);
}
static void
-index_set_link_count (index_priv_t *priv, int64_t count,
- index_xattrop_type_t type)
+index_set_link_count(index_priv_t *priv, int64_t count,
+ index_xattrop_type_t type)
{
- switch (type) {
+ switch (type) {
case XATTROP:
- LOCK (&priv->lock);
- {
- priv->pending_count = count;
- }
- UNLOCK (&priv->lock);
- break;
+ LOCK(&priv->lock);
+ {
+ priv->pending_count = count;
+ }
+ UNLOCK(&priv->lock);
+ break;
default:
- break;
- }
+ break;
+ }
}
static void
-index_get_link_count (index_priv_t *priv, int64_t *count,
- index_xattrop_type_t type)
+index_get_link_count(index_priv_t *priv, int64_t *count,
+ index_xattrop_type_t type)
{
- switch (type) {
+ switch (type) {
case XATTROP:
- LOCK (&priv->lock);
- {
- *count = priv->pending_count;
- }
- UNLOCK (&priv->lock);
- break;
+ LOCK(&priv->lock);
+ {
+ *count = priv->pending_count;
+ }
+ UNLOCK(&priv->lock);
+ break;
default:
- break;
- }
+ break;
+ }
}
static void
-index_dec_link_count (index_priv_t *priv, index_xattrop_type_t type)
+index_dec_link_count(index_priv_t *priv, index_xattrop_type_t type)
{
- switch (type) {
+ switch (type) {
case XATTROP:
- LOCK (&priv->lock);
- {
- priv->pending_count--;
- if (priv->pending_count == 0)
- priv->pending_count--;
- }
- UNLOCK (&priv->lock);
- break;
+ LOCK(&priv->lock);
+ {
+ priv->pending_count--;
+ if (priv->pending_count == 0)
+ priv->pending_count--;
+ }
+ UNLOCK(&priv->lock);
+ break;
default:
- break;
- }
+ break;
+ }
}
-char*
-index_get_subdir_from_type (index_xattrop_type_t type)
+char *
+index_get_subdir_from_type(index_xattrop_type_t type)
{
- if (type < XATTROP || type >= XATTROP_TYPE_END)
- return NULL;
- return index_subdirs[type];
+ if (type < XATTROP || type >= XATTROP_TYPE_END)
+ return NULL;
+ return index_subdirs[type];
}
-char*
-index_get_subdir_from_vgfid (index_priv_t *priv, uuid_t vgfid)
+char *
+index_get_subdir_from_vgfid(index_priv_t *priv, uuid_t vgfid)
{
- return index_get_subdir_from_type (index_get_type_from_vgfid (priv,
- vgfid));
+ return index_get_subdir_from_type(index_get_type_from_vgfid(priv, vgfid));
}
static int
-index_fill_readdir (fd_t *fd, index_fd_ctx_t *fctx, DIR *dir, off_t off,
- size_t size, gf_dirent_t *entries)
-{
- off_t in_case = -1;
- off_t last_off = 0;
- size_t filled = 0;
- int count = 0;
- struct dirent *entry = NULL;
- struct dirent scratch[2] = {{0,},};
- int32_t this_size = -1;
- gf_dirent_t *this_entry = NULL;
- xlator_t *this = NULL;
-
- this = THIS;
- if (!off) {
- rewinddir (dir);
- } else {
- seekdir (dir, off);
+index_fill_readdir(fd_t *fd, index_fd_ctx_t *fctx, DIR *dir, off_t off,
+ size_t size, gf_dirent_t *entries)
+{
+ off_t in_case = -1;
+ off_t last_off = 0;
+ size_t filled = 0;
+ int count = 0;
+ struct dirent *entry = NULL;
+ struct dirent scratch[2] = {
+ {
+ 0,
+ },
+ };
+ int32_t this_size = -1;
+ gf_dirent_t *this_entry = NULL;
+ xlator_t *this = NULL;
+
+ this = THIS;
+ if (!off) {
+ rewinddir(dir);
+ } else {
+ seekdir(dir, off);
#ifndef GF_LINUX_HOST_OS
- if ((u_long)telldir(dir) != off && off != fctx->dir_eof) {
- gf_msg (THIS->name, GF_LOG_ERROR, EINVAL,
- INDEX_MSG_INDEX_READDIR_FAILED,
- "seekdir(0x%llx) failed on dir=%p: "
- "Invalid argument (offset reused from "
- "another DIR * structure?)", off, dir);
- errno = EINVAL;
- count = -1;
- goto out;
- }
-#endif /* GF_LINUX_HOST_OS */
+ if ((u_long)telldir(dir) != off && off != fctx->dir_eof) {
+ gf_msg(THIS->name, GF_LOG_ERROR, EINVAL,
+ INDEX_MSG_INDEX_READDIR_FAILED,
+ "seekdir(0x%llx) failed on dir=%p: "
+ "Invalid argument (offset reused from "
+ "another DIR * structure?)",
+ off, dir);
+ errno = EINVAL;
+ count = -1;
+ goto out;
}
+#endif /* GF_LINUX_HOST_OS */
+ }
- while (filled <= size) {
- in_case = (u_long)telldir (dir);
+ while (filled <= size) {
+ in_case = (u_long)telldir(dir);
- if (in_case == -1) {
- gf_msg (THIS->name, GF_LOG_ERROR, errno,
- INDEX_MSG_INDEX_READDIR_FAILED,
- "telldir failed on dir=%p", dir);
- goto out;
- }
+ if (in_case == -1) {
+ gf_msg(THIS->name, GF_LOG_ERROR, errno,
+ INDEX_MSG_INDEX_READDIR_FAILED, "telldir failed on dir=%p",
+ dir);
+ goto out;
+ }
- errno = 0;
- entry = sys_readdir (dir, scratch);
- if (!entry || errno != 0) {
- if (errno == EBADF) {
- gf_msg (THIS->name, GF_LOG_WARNING, errno,
- INDEX_MSG_INDEX_READDIR_FAILED,
- "readdir failed on dir=%p", dir);
- goto out;
- }
- break;
- }
+ errno = 0;
+ entry = sys_readdir(dir, scratch);
+ if (!entry || errno != 0) {
+ if (errno == EBADF) {
+ gf_msg(THIS->name, GF_LOG_WARNING, errno,
+ INDEX_MSG_INDEX_READDIR_FAILED,
+ "readdir failed on dir=%p", dir);
+ goto out;
+ }
+ break;
+ }
- if (!strncmp (entry->d_name, XATTROP_SUBDIR"-",
- strlen (XATTROP_SUBDIR"-"))) {
- check_delete_stale_index_file (this, entry->d_name,
- XATTROP_SUBDIR);
- continue;
- } else if (!strncmp (entry->d_name, DIRTY_SUBDIR"-",
- strlen (DIRTY_SUBDIR"-"))) {
- check_delete_stale_index_file (this, entry->d_name,
- DIRTY_SUBDIR);
- continue;
- }
+ if (!strncmp(entry->d_name, XATTROP_SUBDIR "-",
+ strlen(XATTROP_SUBDIR "-"))) {
+ check_delete_stale_index_file(this, entry->d_name, XATTROP_SUBDIR);
+ continue;
+ } else if (!strncmp(entry->d_name, DIRTY_SUBDIR "-",
+ strlen(DIRTY_SUBDIR "-"))) {
+ check_delete_stale_index_file(this, entry->d_name, DIRTY_SUBDIR);
+ continue;
+ }
- this_size = max (sizeof (gf_dirent_t),
- sizeof (gfs3_dirplist))
- + strlen (entry->d_name) + 1;
+ this_size = max(sizeof(gf_dirent_t), sizeof(gfs3_dirplist)) +
+ strlen(entry->d_name) + 1;
- if (this_size + filled > size) {
- seekdir (dir, in_case);
+ if (this_size + filled > size) {
+ seekdir(dir, in_case);
#ifndef GF_LINUX_HOST_OS
- if ((u_long)telldir(dir) != in_case &&
- in_case != fctx->dir_eof) {
- gf_msg (THIS->name, GF_LOG_ERROR, EINVAL,
- INDEX_MSG_INDEX_READDIR_FAILED,
- "seekdir(0x%llx) failed on dir=%p: "
- "Invalid argument (offset reused from "
- "another DIR * structure?)",
- in_case, dir);
- errno = EINVAL;
- count = -1;
- goto out;
- }
+ if ((u_long)telldir(dir) != in_case && in_case != fctx->dir_eof) {
+ gf_msg(THIS->name, GF_LOG_ERROR, EINVAL,
+ INDEX_MSG_INDEX_READDIR_FAILED,
+ "seekdir(0x%llx) failed on dir=%p: "
+ "Invalid argument (offset reused from "
+ "another DIR * structure?)",
+ in_case, dir);
+ errno = EINVAL;
+ count = -1;
+ goto out;
+ }
#endif /* GF_LINUX_HOST_OS */
- break;
- }
+ break;
+ }
- this_entry = gf_dirent_for_name (entry->d_name);
+ this_entry = gf_dirent_for_name(entry->d_name);
- if (!this_entry) {
- gf_msg (THIS->name, GF_LOG_ERROR, errno,
- INDEX_MSG_INDEX_READDIR_FAILED,
- "could not create gf_dirent for entry %s",
- entry->d_name);
- goto out;
- }
- /*
- * we store the offset of next entry here, which is
- * probably not intended, but code using syncop_readdir()
- * (glfs-heal.c, afr-self-heald.c, pump.c) rely on it
- * for directory read resumption.
- */
- last_off = (u_long)telldir(dir);
- this_entry->d_off = last_off;
- this_entry->d_ino = entry->d_ino;
-
- list_add_tail (&this_entry->list, &entries->list);
-
- filled += this_size;
- count ++;
+ if (!this_entry) {
+ gf_msg(THIS->name, GF_LOG_ERROR, errno,
+ INDEX_MSG_INDEX_READDIR_FAILED,
+ "could not create gf_dirent for entry %s", entry->d_name);
+ goto out;
}
+ /*
+ * we store the offset of next entry here, which is
+ * probably not intended, but code using syncop_readdir()
+ * (glfs-heal.c, afr-self-heald.c, pump.c) rely on it
+ * for directory read resumption.
+ */
+ last_off = (u_long)telldir(dir);
+ this_entry->d_off = last_off;
+ this_entry->d_ino = entry->d_ino;
- errno = 0;
+ list_add_tail(&this_entry->list, &entries->list);
- if ((!sys_readdir (dir, scratch) && (errno == 0))) {
- /* Indicate EOF */
- errno = ENOENT;
- /* Remember EOF offset for later detection */
- fctx->dir_eof = last_off;
- }
+ filled += this_size;
+ count++;
+ }
+
+ errno = 0;
+
+ if ((!sys_readdir(dir, scratch) && (errno == 0))) {
+ /* Indicate EOF */
+ errno = ENOENT;
+ /* Remember EOF offset for later detection */
+ fctx->dir_eof = last_off;
+ }
out:
- return count;
+ return count;
}
int
-index_link_to_base (xlator_t *this, char *fpath, const char *subdir)
-{
- int ret = 0;
- int fd = 0;
- int op_errno = 0;
- uuid_t index = {0};
- index_priv_t *priv = this->private;
- char base[PATH_MAX] = {0};
-
- index_get_index (priv, index);
- make_index_path (priv->index_basepath, subdir,
- index, base, sizeof (base));
-
- ret = sys_link (base, fpath);
- if (!ret || (errno == EEXIST)) {
- ret = 0;
- goto out;
- }
+index_link_to_base(xlator_t *this, char *fpath, const char *subdir)
+{
+ int ret = 0;
+ int fd = 0;
+ int op_errno = 0;
+ uuid_t index = {0};
+ index_priv_t *priv = this->private;
+ char base[PATH_MAX] = {0};
- op_errno = errno;
- if (op_errno == ENOENT) {
- ret = index_dir_create (this, subdir);
- if (ret) {
- op_errno = errno;
- goto out;
- }
- } else if (op_errno == EMLINK) {
- index_generate_index (priv, index);
- make_index_path (priv->index_basepath, subdir,
- index, base, sizeof (base));
- } else {
- goto out;
- }
+ index_get_index(priv, index);
+ make_index_path(priv->index_basepath, subdir, index, base, sizeof(base));
- op_errno = 0;
- fd = sys_creat (base, 0);
- if ((fd < 0) && (errno != EEXIST)) {
- op_errno = errno;
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- INDEX_MSG_INDEX_ADD_FAILED, "%s: Not able to "
- "create index", fpath);
- goto out;
- }
-
- if (fd >= 0)
- sys_close (fd);
+ ret = sys_link(base, fpath);
+ if (!ret || (errno == EEXIST)) {
+ ret = 0;
+ goto out;
+ }
- ret = sys_link (base, fpath);
- if (ret && (errno != EEXIST)) {
- op_errno = errno;
- gf_msg (this->name, GF_LOG_ERROR, errno,
- INDEX_MSG_INDEX_ADD_FAILED, "%s: Not able to "
- "add to index", fpath);
- goto out;
- }
+ op_errno = errno;
+ if (op_errno == ENOENT) {
+ ret = index_dir_create(this, subdir);
+ if (ret) {
+ op_errno = errno;
+ goto out;
+ }
+ } else if (op_errno == EMLINK) {
+ index_generate_index(priv, index);
+ make_index_path(priv->index_basepath, subdir, index, base,
+ sizeof(base));
+ } else {
+ goto out;
+ }
+
+ op_errno = 0;
+ fd = sys_creat(base, 0);
+ if ((fd < 0) && (errno != EEXIST)) {
+ op_errno = errno;
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, INDEX_MSG_INDEX_ADD_FAILED,
+ "%s: Not able to "
+ "create index",
+ fpath);
+ goto out;
+ }
+
+ if (fd >= 0)
+ sys_close(fd);
+
+ ret = sys_link(base, fpath);
+ if (ret && (errno != EEXIST)) {
+ op_errno = errno;
+ gf_msg(this->name, GF_LOG_ERROR, errno, INDEX_MSG_INDEX_ADD_FAILED,
+ "%s: Not able to "
+ "add to index",
+ fpath);
+ goto out;
+ }
out:
- return -op_errno;
+ return -op_errno;
}
int
-index_add (xlator_t *this, uuid_t gfid, const char *subdir,
- index_xattrop_type_t type)
+index_add(xlator_t *this, uuid_t gfid, const char *subdir,
+ index_xattrop_type_t type)
{
- char gfid_path[PATH_MAX] = {0};
- int ret = -1;
- index_priv_t *priv = NULL;
- struct stat st = {0};
+ char gfid_path[PATH_MAX] = {0};
+ int ret = -1;
+ index_priv_t *priv = NULL;
+ struct stat st = {0};
- priv = this->private;
+ priv = this->private;
- if (gf_uuid_is_null (gfid)) {
- GF_ASSERT (0);
- goto out;
- }
+ if (gf_uuid_is_null(gfid)) {
+ GF_ASSERT(0);
+ goto out;
+ }
- make_gfid_path (priv->index_basepath, subdir, gfid,
- gfid_path, sizeof (gfid_path));
+ make_gfid_path(priv->index_basepath, subdir, gfid, gfid_path,
+ sizeof(gfid_path));
- ret = sys_stat (gfid_path, &st);
- if (!ret)
- goto out;
- ret = index_link_to_base (this, gfid_path, subdir);
+ ret = sys_stat(gfid_path, &st);
+ if (!ret)
+ goto out;
+ ret = index_link_to_base(this, gfid_path, subdir);
out:
- return ret;
+ return ret;
}
int
-index_del (xlator_t *this, uuid_t gfid, const char *subdir, int type)
-{
- int32_t op_errno __attribute__((unused)) = 0;
- index_priv_t *priv = NULL;
- int ret = 0;
- char gfid_path[PATH_MAX] = {0};
- char rename_dst[PATH_MAX] = {0,};
- uuid_t uuid;
-
- priv = this->private;
- GF_ASSERT_AND_GOTO_WITH_ERROR (this->name, !gf_uuid_is_null (gfid),
- out, op_errno, EINVAL);
- make_gfid_path (priv->index_basepath, subdir, gfid,
- gfid_path, sizeof (gfid_path));
-
- if ((strcmp (subdir, ENTRY_CHANGES_SUBDIR)) == 0) {
- ret = sys_rmdir (gfid_path);
- /* rmdir above could fail with ENOTEMPTY if the indices under
- * it were created when granular-entry-heal was enabled, whereas
- * the actual heal that happened was non-granular (or full) in
- * nature, resulting in name indices getting left out. To
- * clean up this directory without it affecting the IO path perf,
- * the directory is renamed to a unique name under
- * indices/entry-changes. Self-heal will pick up this entry
- * during crawl and on lookup into the file system figure that
- * the index is stale and subsequently wipe it out using rmdir().
- */
- if ((ret) && (errno == ENOTEMPTY)) {
- gf_uuid_generate (uuid);
- make_gfid_path (priv->index_basepath, subdir, uuid,
- rename_dst, sizeof (rename_dst));
- ret = sys_rename (gfid_path, rename_dst);
- }
- } else {
- ret = sys_unlink (gfid_path);
- }
-
- if (ret && (errno != ENOENT)) {
- gf_msg (this->name, GF_LOG_ERROR, errno,
- INDEX_MSG_INDEX_DEL_FAILED, "%s: failed to delete"
- " from index", gfid_path);
- ret = -errno;
- goto out;
- }
-
- index_dec_link_count (priv, type);
- ret = 0;
+index_del(xlator_t *this, uuid_t gfid, const char *subdir, int type)
+{
+ int32_t op_errno __attribute__((unused)) = 0;
+ index_priv_t *priv = NULL;
+ int ret = 0;
+ char gfid_path[PATH_MAX] = {0};
+ char rename_dst[PATH_MAX] = {
+ 0,
+ };
+ uuid_t uuid;
+
+ priv = this->private;
+ GF_ASSERT_AND_GOTO_WITH_ERROR(this->name, !gf_uuid_is_null(gfid), out,
+ op_errno, EINVAL);
+ make_gfid_path(priv->index_basepath, subdir, gfid, gfid_path,
+ sizeof(gfid_path));
+
+ if ((strcmp(subdir, ENTRY_CHANGES_SUBDIR)) == 0) {
+ ret = sys_rmdir(gfid_path);
+ /* rmdir above could fail with ENOTEMPTY if the indices under
+ * it were created when granular-entry-heal was enabled, whereas
+ * the actual heal that happened was non-granular (or full) in
+ * nature, resulting in name indices getting left out. To
+ * clean up this directory without it affecting the IO path perf,
+ * the directory is renamed to a unique name under
+ * indices/entry-changes. Self-heal will pick up this entry
+ * during crawl and on lookup into the file system figure that
+ * the index is stale and subsequently wipe it out using rmdir().
+ */
+ if ((ret) && (errno == ENOTEMPTY)) {
+ gf_uuid_generate(uuid);
+ make_gfid_path(priv->index_basepath, subdir, uuid, rename_dst,
+ sizeof(rename_dst));
+ ret = sys_rename(gfid_path, rename_dst);
+ }
+ } else {
+ ret = sys_unlink(gfid_path);
+ }
+
+ if (ret && (errno != ENOENT)) {
+ gf_msg(this->name, GF_LOG_ERROR, errno, INDEX_MSG_INDEX_DEL_FAILED,
+ "%s: failed to delete"
+ " from index",
+ gfid_path);
+ ret = -errno;
+ goto out;
+ }
+
+ index_dec_link_count(priv, type);
+ ret = 0;
out:
- return ret;
+ return ret;
}
static gf_boolean_t
-_is_xattr_in_watchlist (dict_t *d, char *k, data_t *v, void *tmp)
+_is_xattr_in_watchlist(dict_t *d, char *k, data_t *v, void *tmp)
{
- if (!strncmp (k, tmp, strlen (k)))
- return _gf_true;
+ if (!strncmp(k, tmp, strlen(k)))
+ return _gf_true;
- return _gf_false;
+ return _gf_false;
}
static gf_boolean_t
-is_xattr_in_watchlist (dict_t *this, char *key, data_t *value, void *matchdata)
+is_xattr_in_watchlist(dict_t *this, char *key, data_t *value, void *matchdata)
{
- int ret = -1;
+ int ret = -1;
- //matchdata is a list of xattrs
- //key is strncmp'ed with each xattr in matchdata.
- //ret will be 0 if key pattern is not present in the matchdata
- //else ret will be count number of xattrs the key pattern-matches with.
- ret = dict_foreach_match (matchdata, _is_xattr_in_watchlist, key,
- dict_null_foreach_fn, NULL);
+ // matchdata is a list of xattrs
+ // key is strncmp'ed with each xattr in matchdata.
+ // ret will be 0 if key pattern is not present in the matchdata
+ // else ret will be count number of xattrs the key pattern-matches with.
+ ret = dict_foreach_match(matchdata, _is_xattr_in_watchlist, key,
+ dict_null_foreach_fn, NULL);
- if (ret > 0)
- return _gf_true;
- return _gf_false;
+ if (ret > 0)
+ return _gf_true;
+ return _gf_false;
}
static int
-index_find_xattr_type (dict_t *d, char *k, data_t *v)
+index_find_xattr_type(dict_t *d, char *k, data_t *v)
{
- int idx = -1;
- index_priv_t *priv = THIS->private;
+ int idx = -1;
+ index_priv_t *priv = THIS->private;
- if (priv->dirty_watchlist && is_xattr_in_watchlist (d, k, v,
- priv->dirty_watchlist))
- idx = DIRTY;
- else if (priv->pending_watchlist && is_xattr_in_watchlist (d, k, v,
- priv->pending_watchlist))
- idx = XATTROP;
+ if (priv->dirty_watchlist &&
+ is_xattr_in_watchlist(d, k, v, priv->dirty_watchlist))
+ idx = DIRTY;
+ else if (priv->pending_watchlist &&
+ is_xattr_in_watchlist(d, k, v, priv->pending_watchlist))
+ idx = XATTROP;
- return idx;
+ return idx;
}
int
-index_fill_zero_array (dict_t *d, char *k, data_t *v, void *adata)
+index_fill_zero_array(dict_t *d, char *k, data_t *v, void *adata)
{
- int idx = -1;
- int *zfilled = adata;
- //zfilled array contains `state` for all types xattrs.
- //state : whether the gfid file of this file exists in
- //corresponding xattr directory or not.
+ int idx = -1;
+ int *zfilled = adata;
+ // zfilled array contains `state` for all types xattrs.
+ // state : whether the gfid file of this file exists in
+ // corresponding xattr directory or not.
- idx = index_find_xattr_type (d, k, v);
- if (idx == -1)
- return 0;
- zfilled[idx] = 0;
+ idx = index_find_xattr_type(d, k, v);
+ if (idx == -1)
return 0;
+ zfilled[idx] = 0;
+ return 0;
}
static int
-_check_key_is_zero_filled (dict_t *d, char *k, data_t *v,
- void *tmp)
+_check_key_is_zero_filled(dict_t *d, char *k, data_t *v, void *tmp)
{
- int *zfilled = tmp;
- int idx = -1;
+ int *zfilled = tmp;
+ int idx = -1;
- idx = index_find_xattr_type (d, k, v);
- if (idx == -1)
- return 0;
-
- /* Along with checking that the value of a key is zero filled
- * the key's corresponding index should be assigned
- * appropriate value.
- * zfilled[idx] will be 0(false) if value not zero.
- * will be 1(true) if value is zero.
- */
- if (mem_0filled ((const char*)v->data, v->len)) {
- zfilled[idx] = 0;
- return 0;
- }
+ idx = index_find_xattr_type(d, k, v);
+ if (idx == -1)
+ return 0;
- /* If zfilled[idx] was previously 0, it means at least
- * one xattr of its "kind" is non-zero. Keep its value
- * the same.
- */
- if (zfilled[idx])
- zfilled[idx] = 1;
+ /* Along with checking that the value of a key is zero filled
+ * the key's corresponding index should be assigned
+ * appropriate value.
+ * zfilled[idx] will be 0(false) if value not zero.
+ * will be 1(true) if value is zero.
+ */
+ if (mem_0filled((const char *)v->data, v->len)) {
+ zfilled[idx] = 0;
return 0;
+ }
+
+ /* If zfilled[idx] was previously 0, it means at least
+ * one xattr of its "kind" is non-zero. Keep its value
+ * the same.
+ */
+ if (zfilled[idx])
+ zfilled[idx] = 1;
+ return 0;
}
int
-index_entry_create (xlator_t *this, inode_t *inode, char *filename)
-{
- int ret = -1;
- int op_errno = 0;
- char pgfid_path[PATH_MAX] = {0};
- char entry_path[PATH_MAX] = {0};
- index_priv_t *priv = NULL;
- index_inode_ctx_t *ctx = NULL;
-
- priv = this->private;
-
- GF_ASSERT_AND_GOTO_WITH_ERROR (this->name,
- !gf_uuid_is_null (inode->gfid), out,
- op_errno, EINVAL);
- GF_ASSERT_AND_GOTO_WITH_ERROR (this->name, filename, out, op_errno,
- EINVAL);
-
- ret = index_inode_ctx_get (inode, this, &ctx);
- if (ret) {
- op_errno = EINVAL;
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- INDEX_MSG_INODE_CTX_GET_SET_FAILED,
- "Not able to get inode ctx for %s",
- uuid_utoa (inode->gfid));
- goto out;
- }
-
- make_gfid_path (priv->index_basepath, ENTRY_CHANGES_SUBDIR,
- inode->gfid, pgfid_path, sizeof (pgfid_path));
-
- if (ctx->state[ENTRY_CHANGES] != IN) {
- ret = sys_mkdir (pgfid_path, 0600);
- if (ret != 0 && errno != EEXIST) {
- op_errno = errno;
- goto out;
- }
- ctx->state[ENTRY_CHANGES] = IN;
- }
-
- op_errno = 0;
-
- snprintf (entry_path, sizeof(entry_path), "%s/%s", pgfid_path,
- filename);
- ret = index_link_to_base (this, entry_path, ENTRY_CHANGES_SUBDIR);
+index_entry_create(xlator_t *this, inode_t *inode, char *filename)
+{
+ int ret = -1;
+ int op_errno = 0;
+ char pgfid_path[PATH_MAX] = {0};
+ char entry_path[PATH_MAX] = {0};
+ index_priv_t *priv = NULL;
+ index_inode_ctx_t *ctx = NULL;
+ int32_t len = 0;
+
+ priv = this->private;
+
+ GF_ASSERT_AND_GOTO_WITH_ERROR(this->name, !gf_uuid_is_null(inode->gfid),
+ out, op_errno, EINVAL);
+ GF_ASSERT_AND_GOTO_WITH_ERROR(this->name, filename, out, op_errno, EINVAL);
+
+ ret = index_inode_ctx_get(inode, this, &ctx);
+ if (ret) {
+ op_errno = EINVAL;
+ gf_msg(this->name, GF_LOG_ERROR, op_errno,
+ INDEX_MSG_INODE_CTX_GET_SET_FAILED,
+ "Not able to get inode ctx for %s", uuid_utoa(inode->gfid));
+ goto out;
+ }
+
+ make_gfid_path(priv->index_basepath, ENTRY_CHANGES_SUBDIR, inode->gfid,
+ pgfid_path, sizeof(pgfid_path));
+
+ if (ctx->state[ENTRY_CHANGES] != IN) {
+ ret = sys_mkdir(pgfid_path, 0600);
+ if (ret != 0 && errno != EEXIST) {
+ op_errno = errno;
+ goto out;
+ }
+ ctx->state[ENTRY_CHANGES] = IN;
+ }
+
+ if (strchr(filename, '/')) {
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL, INDEX_MSG_INDEX_ADD_FAILED,
+ "Got invalid entry (%s) for pargfid path (%s)", filename,
+ pgfid_path);
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ len = snprintf(entry_path, sizeof(entry_path), "%s/%s", pgfid_path,
+ filename);
+ if ((len < 0) || (len >= sizeof(entry_path))) {
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ op_errno = 0;
+
+ ret = index_link_to_base(this, entry_path, ENTRY_CHANGES_SUBDIR);
out:
- if (op_errno)
- ret = -op_errno;
- return ret;
+ if (op_errno)
+ ret = -op_errno;
+ return ret;
}
int
-index_entry_delete (xlator_t *this, uuid_t pgfid, char *filename)
-{
- int ret = 0;
- int op_errno = 0;
- char pgfid_path[PATH_MAX] = {0};
- char entry_path[PATH_MAX] = {0};
- index_priv_t *priv = NULL;
-
- priv = this->private;
-
- GF_ASSERT_AND_GOTO_WITH_ERROR (this->name, !gf_uuid_is_null (pgfid),
- out, op_errno, EINVAL);
- GF_ASSERT_AND_GOTO_WITH_ERROR (this->name, filename, out, op_errno,
- EINVAL);
-
- make_gfid_path (priv->index_basepath, ENTRY_CHANGES_SUBDIR, pgfid,
- pgfid_path, sizeof (pgfid_path));
- snprintf (entry_path, sizeof(entry_path), "%s/%s", pgfid_path,
- filename);
-
- ret = sys_unlink (entry_path);
- if (ret && (errno != ENOENT)) {
- op_errno = errno;
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- INDEX_MSG_INDEX_DEL_FAILED,
- "%s: failed to delete from index/entry-changes",
- entry_path);
- }
+index_entry_delete(xlator_t *this, uuid_t pgfid, char *filename)
+{
+ int ret = 0;
+ int op_errno = 0;
+ char pgfid_path[PATH_MAX] = {0};
+ char entry_path[PATH_MAX] = {0};
+ index_priv_t *priv = NULL;
+ int32_t len = 0;
+
+ priv = this->private;
+
+ GF_ASSERT_AND_GOTO_WITH_ERROR(this->name, !gf_uuid_is_null(pgfid), out,
+ op_errno, EINVAL);
+ GF_ASSERT_AND_GOTO_WITH_ERROR(this->name, filename, out, op_errno, EINVAL);
+
+ make_gfid_path(priv->index_basepath, ENTRY_CHANGES_SUBDIR, pgfid,
+ pgfid_path, sizeof(pgfid_path));
+
+ if (strchr(filename, '/')) {
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL, INDEX_MSG_INDEX_DEL_FAILED,
+ "Got invalid entry (%s) for pargfid path (%s)", filename,
+ pgfid_path);
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ len = snprintf(entry_path, sizeof(entry_path), "%s/%s", pgfid_path,
+ filename);
+ if ((len < 0) || (len >= sizeof(entry_path))) {
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ ret = sys_unlink(entry_path);
+ if (ret && (errno != ENOENT)) {
+ op_errno = errno;
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, INDEX_MSG_INDEX_DEL_FAILED,
+ "%s: failed to delete from index/entry-changes", entry_path);
+ }
out:
- return -op_errno;
+ return -op_errno;
}
int
-index_entry_action (xlator_t *this, inode_t *inode, dict_t *xdata, char *key)
+index_entry_action(xlator_t *this, inode_t *inode, dict_t *xdata, char *key)
{
- int ret = 0;
- char *filename = NULL;
+ int ret = 0;
+ char *filename = NULL;
- ret = dict_get_str (xdata, key, &filename);
- if (ret != 0) {
- ret = 0;
- goto out;
- }
+ ret = dict_get_str(xdata, key, &filename);
+ if (ret != 0) {
+ ret = 0;
+ goto out;
+ }
- if (strcmp (key, GF_XATTROP_ENTRY_IN_KEY) == 0)
- ret = index_entry_create (this, inode, filename);
- else if (strcmp (key, GF_XATTROP_ENTRY_OUT_KEY) == 0)
- ret = index_entry_delete (this, inode->gfid, filename);
+ if (strcmp(key, GF_XATTROP_ENTRY_IN_KEY) == 0)
+ ret = index_entry_create(this, inode, filename);
+ else if (strcmp(key, GF_XATTROP_ENTRY_OUT_KEY) == 0)
+ ret = index_entry_delete(this, inode->gfid, filename);
out:
- return ret;
+ return ret;
}
void
-_index_action (xlator_t *this, inode_t *inode, int *zfilled)
-{
- int ret = 0;
- int i = 0;
- index_inode_ctx_t *ctx = NULL;
- char *subdir = NULL;
-
- ret = index_inode_ctx_get (inode, this, &ctx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, EINVAL,
- INDEX_MSG_INODE_CTX_GET_SET_FAILED, "Not able to get"
- " inode context for %s.", uuid_utoa (inode->gfid));
- goto out;
- }
-
- for (i = 0; i < XATTROP_TYPE_END; i++) {
- subdir = index_get_subdir_from_type (i);
- if (zfilled[i] == 1) {
- if (ctx->state[i] == NOTIN)
- continue;
- ret = index_del (this, inode->gfid, subdir, i);
- if (!ret)
- ctx->state[i] = NOTIN;
- } else if (zfilled[i] == 0){
- if (ctx->state[i] == IN)
- continue;
- ret = index_add (this, inode->gfid, subdir, i);
- if (!ret)
- ctx->state[i] = IN;
- }
- }
+_index_action(xlator_t *this, inode_t *inode, int *zfilled)
+{
+ int ret = 0;
+ int i = 0;
+ index_inode_ctx_t *ctx = NULL;
+ char *subdir = NULL;
+
+ ret = index_inode_ctx_get(inode, this, &ctx);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL,
+ INDEX_MSG_INODE_CTX_GET_SET_FAILED,
+ "Not able to get"
+ " inode context for %s.",
+ uuid_utoa(inode->gfid));
+ goto out;
+ }
+
+ for (i = 0; i < XATTROP_TYPE_END; i++) {
+ subdir = index_get_subdir_from_type(i);
+ if (zfilled[i] == 1) {
+ if (ctx->state[i] == NOTIN)
+ continue;
+ ret = index_del(this, inode->gfid, subdir, i);
+ if (!ret)
+ ctx->state[i] = NOTIN;
+ } else if (zfilled[i] == 0) {
+ if (ctx->state[i] == IN)
+ continue;
+ ret = index_add(this, inode->gfid, subdir, i);
+ if (!ret)
+ ctx->state[i] = IN;
+ }
+ }
out:
- return;
+ return;
}
static void
-index_init_state (xlator_t *this, inode_t *inode, index_inode_ctx_t *ctx,
- char *subdir)
+index_init_state(xlator_t *this, inode_t *inode, index_inode_ctx_t *ctx,
+ char *subdir)
{
- int ret = -1;
- char pgfid_path[PATH_MAX] = {0};
- struct stat st = {0};
- index_priv_t *priv = NULL;
+ int ret = -1;
+ char pgfid_path[PATH_MAX] = {0};
+ struct stat st = {0};
+ index_priv_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- make_gfid_path (priv->index_basepath, subdir, inode->gfid, pgfid_path,
- sizeof (pgfid_path));
+ make_gfid_path(priv->index_basepath, subdir, inode->gfid, pgfid_path,
+ sizeof(pgfid_path));
- ret = sys_stat (pgfid_path, &st);
- if (ret == 0)
- ctx->state[ENTRY_CHANGES] = IN;
- else if (ret != 0 && errno == ENOENT)
- ctx->state[ENTRY_CHANGES] = NOTIN;
+ ret = sys_stat(pgfid_path, &st);
+ if (ret == 0)
+ ctx->state[ENTRY_CHANGES] = IN;
+ else if (ret != 0 && errno == ENOENT)
+ ctx->state[ENTRY_CHANGES] = NOTIN;
- return;
+ return;
}
void
-xattrop_index_action (xlator_t *this, index_local_t *local, dict_t *xattr,
- dict_match_t match, void *match_data)
-{
- int ret = 0;
- int zfilled[XATTROP_TYPE_END] = {0,};
- int8_t value = 0;
- char *subdir = NULL;
- dict_t *req_xdata = NULL;
- inode_t *inode = NULL;
- index_inode_ctx_t *ctx = NULL;
-
- inode = local->inode;
- req_xdata = local->xdata;
-
- memset (zfilled, -1, sizeof (zfilled));
- ret = dict_foreach_match (xattr, match, match_data,
- _check_key_is_zero_filled, zfilled);
- _index_action (this, inode, zfilled);
-
- if (req_xdata) {
- ret = index_entry_action (this, inode, req_xdata,
- GF_XATTROP_ENTRY_OUT_KEY);
-
- ret = dict_get_int8 (req_xdata, GF_XATTROP_PURGE_INDEX, &value);
- if ((ret) || (value == 0))
- goto out;
- }
-
- if (zfilled[XATTROP] != 1)
- goto out;
-
- if (inode->ia_type != IA_IFDIR)
- goto out;
-
- subdir = index_get_subdir_from_type (ENTRY_CHANGES);
- ret = index_inode_ctx_get (inode, this, &ctx);
- if (ctx->state[ENTRY_CHANGES] == UNKNOWN)
- index_init_state (this, inode, ctx, subdir);
- if (ctx->state[ENTRY_CHANGES] == IN) {
- ret = index_del (this, inode->gfid, subdir,
- ENTRY_CHANGES);
- ctx->state[ENTRY_CHANGES] = NOTIN;
- }
+xattrop_index_action(xlator_t *this, index_local_t *local, dict_t *xattr,
+ dict_match_t match, void *match_data)
+{
+ int ret = 0;
+ int zfilled[XATTROP_TYPE_END] = {
+ 0,
+ };
+ int8_t value = 0;
+ char *subdir = NULL;
+ dict_t *req_xdata = NULL;
+ inode_t *inode = NULL;
+ index_inode_ctx_t *ctx = NULL;
+
+ inode = local->inode;
+ req_xdata = local->xdata;
+
+ memset(zfilled, -1, sizeof(zfilled));
+ ret = dict_foreach_match(xattr, match, match_data,
+ _check_key_is_zero_filled, zfilled);
+ _index_action(this, inode, zfilled);
+
+ if (req_xdata) {
+ ret = index_entry_action(this, inode, req_xdata,
+ GF_XATTROP_ENTRY_OUT_KEY);
+
+ ret = dict_get_int8(req_xdata, GF_XATTROP_PURGE_INDEX, &value);
+ if ((ret) || (value == 0))
+ goto out;
+ }
+
+ if (zfilled[XATTROP] != 1)
+ goto out;
+
+ if (inode->ia_type != IA_IFDIR)
+ goto out;
+
+ subdir = index_get_subdir_from_type(ENTRY_CHANGES);
+ ret = index_inode_ctx_get(inode, this, &ctx);
+ if (ctx->state[ENTRY_CHANGES] == UNKNOWN)
+ index_init_state(this, inode, ctx, subdir);
+ if (ctx->state[ENTRY_CHANGES] == IN) {
+ ret = index_del(this, inode->gfid, subdir, ENTRY_CHANGES);
+ ctx->state[ENTRY_CHANGES] = NOTIN;
+ }
out:
- return;
+ return;
}
static gf_boolean_t
-index_xattrop_track (xlator_t *this, gf_xattrop_flags_t flags, dict_t *dict)
+index_xattrop_track(xlator_t *this, gf_xattrop_flags_t flags, dict_t *dict)
{
- index_priv_t *priv = this->private;
+ index_priv_t *priv = this->private;
- if (flags == GF_XATTROP_ADD_ARRAY)
- return _gf_true;
+ if (flags == GF_XATTROP_ADD_ARRAY)
+ return _gf_true;
- if (flags != GF_XATTROP_ADD_ARRAY64)
- return _gf_false;
+ if (flags != GF_XATTROP_ADD_ARRAY64)
+ return _gf_false;
- if (!priv->pending_watchlist)
- return _gf_false;
+ if (!priv->pending_watchlist)
+ return _gf_false;
- if (dict_foreach_match (dict, is_xattr_in_watchlist,
- priv->pending_watchlist, dict_null_foreach_fn,
- NULL) > 0)
- return _gf_true;
+ if (dict_foreach_match(dict, is_xattr_in_watchlist, priv->pending_watchlist,
+ dict_null_foreach_fn, NULL) > 0)
+ return _gf_true;
- return _gf_false;
+ return _gf_false;
}
int
-index_inode_path (xlator_t *this, inode_t *inode, char *dirpath, size_t len)
-{
- char *subdir = NULL;
- int ret = 0;
- index_priv_t *priv = NULL;
- index_inode_ctx_t *ictx = NULL;
-
- priv = this->private;
- if (!index_is_fop_on_internal_inode (this, inode, NULL)) {
- ret = -EINVAL;
- goto out;
- }
-
- subdir = index_get_subdir_from_vgfid (priv, inode->gfid);
- if (subdir) {
- if (len <= strlen (priv->index_basepath) + 1 /*'/'*/ +
- strlen (subdir)) {
- ret = -EINVAL;
- goto out;
- }
- make_index_dir_path (priv->index_basepath, subdir,
- dirpath, len);
- } else {
- ret = index_inode_ctx_get (inode, this, &ictx);
- if (ret)
- goto out;
- if (gf_uuid_is_null (ictx->virtual_pargfid)) {
- ret = -EINVAL;
- goto out;
- }
- make_index_dir_path (priv->index_basepath, ENTRY_CHANGES_SUBDIR,
- dirpath, len);
- if (len <= strlen (dirpath) + 1 /*'/'*/ + strlen (UUID0_STR)) {
- ret = -EINVAL;
- goto out;
- }
- strcat (dirpath, "/");
- strcat (dirpath, uuid_utoa (ictx->virtual_pargfid));
- }
+index_inode_path(xlator_t *this, inode_t *inode, char *dirpath, size_t len)
+{
+ char *subdir = NULL;
+ int ret = 0;
+ index_priv_t *priv = NULL;
+ index_inode_ctx_t *ictx = NULL;
+
+ priv = this->private;
+ if (!index_is_fop_on_internal_inode(this, inode, NULL)) {
+ ret = -EINVAL;
+ goto out;
+ }
+
+ subdir = index_get_subdir_from_vgfid(priv, inode->gfid);
+ if (subdir) {
+ if (len <= strlen(priv->index_basepath) + 1 /*'/'*/ + strlen(subdir)) {
+ ret = -EINVAL;
+ goto out;
+ }
+ make_index_dir_path(priv->index_basepath, subdir, dirpath, len);
+ } else {
+ ret = index_inode_ctx_get(inode, this, &ictx);
+ if (ret)
+ goto out;
+ if (gf_uuid_is_null(ictx->virtual_pargfid)) {
+ ret = -EINVAL;
+ goto out;
+ }
+ make_index_dir_path(priv->index_basepath, ENTRY_CHANGES_SUBDIR, dirpath,
+ len);
+ if (len <= strlen(dirpath) + 1 /*'/'*/ + SLEN(UUID0_STR)) {
+ ret = -EINVAL;
+ goto out;
+ }
+ strcat(dirpath, "/");
+ strcat(dirpath, uuid_utoa(ictx->virtual_pargfid));
+ }
out:
- return ret;
+ return ret;
}
int
-__index_fd_ctx_get (fd_t *fd, xlator_t *this, index_fd_ctx_t **ctx)
+__index_fd_ctx_get(fd_t *fd, xlator_t *this, index_fd_ctx_t **ctx)
{
- int ret = 0;
- index_fd_ctx_t *fctx = NULL;
- uint64_t tmpctx = 0;
- char dirpath[PATH_MAX] = {0};
-
- ret = __fd_ctx_get (fd, this, &tmpctx);
- if (!ret) {
- fctx = (index_fd_ctx_t*) (long) tmpctx;
- *ctx = fctx;
- goto out;
- }
-
- ret = index_inode_path (this, fd->inode, dirpath, sizeof (dirpath));
- if (ret)
- goto out;
-
- fctx = GF_CALLOC (1, sizeof (*fctx), gf_index_fd_ctx_t);
- if (!fctx) {
- ret = -ENOMEM;
- goto out;
- }
-
- fctx->dir = sys_opendir (dirpath);
- if (!fctx->dir) {
- ret = -errno;
- GF_FREE (fctx);
- fctx = NULL;
- goto out;
- }
- fctx->dir_eof = -1;
+ int ret = 0;
+ index_fd_ctx_t *fctx = NULL;
+ uint64_t tmpctx = 0;
+ char dirpath[PATH_MAX] = {0};
- ret = __fd_ctx_set (fd, this, (uint64_t)(long)fctx);
- if (ret) {
- (void) sys_closedir (fctx->dir);
- GF_FREE (fctx);
- fctx = NULL;
- ret = -EINVAL;
- goto out;
- }
+ ret = __fd_ctx_get(fd, this, &tmpctx);
+ if (!ret) {
+ fctx = (index_fd_ctx_t *)(long)tmpctx;
*ctx = fctx;
+ goto out;
+ }
+
+ ret = index_inode_path(this, fd->inode, dirpath, sizeof(dirpath));
+ if (ret)
+ goto out;
+
+ fctx = GF_CALLOC(1, sizeof(*fctx), gf_index_fd_ctx_t);
+ if (!fctx) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ fctx->dir = sys_opendir(dirpath);
+ if (!fctx->dir) {
+ ret = -errno;
+ GF_FREE(fctx);
+ fctx = NULL;
+ goto out;
+ }
+ fctx->dir_eof = -1;
+
+ ret = __fd_ctx_set(fd, this, (uint64_t)(long)fctx);
+ if (ret) {
+ (void)sys_closedir(fctx->dir);
+ GF_FREE(fctx);
+ fctx = NULL;
+ ret = -EINVAL;
+ goto out;
+ }
+ *ctx = fctx;
out:
- return ret;
+ return ret;
}
int
-index_fd_ctx_get (fd_t *fd, xlator_t *this, index_fd_ctx_t **ctx)
+index_fd_ctx_get(fd_t *fd, xlator_t *this, index_fd_ctx_t **ctx)
{
- int ret = 0;
- LOCK (&fd->lock);
- {
- ret = __index_fd_ctx_get (fd, this, ctx);
- }
- UNLOCK (&fd->lock);
- return ret;
+ int ret = 0;
+ LOCK(&fd->lock);
+ {
+ ret = __index_fd_ctx_get(fd, this, ctx);
+ }
+ UNLOCK(&fd->lock);
+ return ret;
}
-//new - Not NULL means start a fop
-//new - NULL means done processing the fop
+// new - Not NULL means start a fop
+// new - NULL means done processing the fop
void
-index_queue_process (xlator_t *this, inode_t *inode, call_stub_t *new)
+index_queue_process(xlator_t *this, inode_t *inode, call_stub_t *new)
{
- call_stub_t *stub = NULL;
- index_inode_ctx_t *ctx = NULL;
- int ret = 0;
- call_frame_t *frame = NULL;
+ call_stub_t *stub = NULL;
+ index_inode_ctx_t *ctx = NULL;
+ int ret = 0;
+ call_frame_t *frame = NULL;
- LOCK (&inode->lock);
- {
- ret = __index_inode_ctx_get (inode, this, &ctx);
- if (ret)
- goto unlock;
-
- if (new) {
- __index_enqueue (&ctx->callstubs, new);
- new = NULL;
- } else {
- ctx->processing = _gf_false;
- }
+ LOCK(&inode->lock);
+ {
+ ret = __index_inode_ctx_get(inode, this, &ctx);
+ if (ret)
+ goto unlock;
- if (!ctx->processing) {
- stub = __index_dequeue (&ctx->callstubs);
- if (stub)
- ctx->processing = _gf_true;
- else
- ctx->processing = _gf_false;
- }
+ if (new) {
+ __index_enqueue(&ctx->callstubs, new);
+ new = NULL;
+ } else {
+ ctx->processing = _gf_false;
+ }
+
+ if (!ctx->processing) {
+ stub = __index_dequeue(&ctx->callstubs);
+ if (stub)
+ ctx->processing = _gf_true;
+ else
+ ctx->processing = _gf_false;
}
+ }
unlock:
- UNLOCK (&inode->lock);
-
- if (ret && new) {
- frame = new->frame;
- if (new->fop == GF_FOP_XATTROP) {
- INDEX_STACK_UNWIND (xattrop, frame, -1, ENOMEM,
- NULL, NULL);
- } else if (new->fop == GF_FOP_FXATTROP) {
- INDEX_STACK_UNWIND (fxattrop, frame, -1, ENOMEM,
- NULL, NULL);
- }
- call_stub_destroy (new);
- } else if (stub) {
- call_resume (stub);
+ UNLOCK(&inode->lock);
+
+ if (ret && new) {
+ frame = new->frame;
+ if (new->fop == GF_FOP_XATTROP) {
+ INDEX_STACK_UNWIND(xattrop, frame, -1, ENOMEM, NULL, NULL);
+ } else if (new->fop == GF_FOP_FXATTROP) {
+ INDEX_STACK_UNWIND(fxattrop, frame, -1, ENOMEM, NULL, NULL);
}
- return;
+ call_stub_destroy(new);
+ } else if (stub) {
+ call_resume(stub);
+ }
+ return;
}
static int
-xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xattr,
- dict_t *xdata, dict_match_t match, dict_t *matchdata)
+xattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xattr, dict_t *xdata, dict_match_t match,
+ dict_t *matchdata)
{
- inode_t *inode = NULL;
- index_local_t *local = NULL;
+ inode_t *inode = NULL;
+ index_local_t *local = NULL;
- local = frame->local;
- inode = inode_ref (local->inode);
+ local = frame->local;
+ inode = inode_ref(local->inode);
- if (op_ret < 0)
- goto out;
+ if (op_ret < 0)
+ goto out;
- xattrop_index_action (this, local, xattr, match, matchdata);
+ xattrop_index_action(this, local, xattr, match, matchdata);
out:
- INDEX_STACK_UNWIND (xattrop, frame, op_ret, op_errno, xattr, xdata);
- index_queue_process (this, inode, NULL);
- inode_unref (inode);
+ INDEX_STACK_UNWIND(xattrop, frame, op_ret, op_errno, xattr, xdata);
+ index_queue_process(this, inode, NULL);
+ inode_unref(inode);
- return 0;
+ return 0;
}
int32_t
-index_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xattr,
- dict_t *xdata)
+index_xattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
{
- index_priv_t *priv = this->private;
+ index_priv_t *priv = this->private;
- xattrop_cbk (frame, cookie, this, op_ret, op_errno,
- xattr, xdata, is_xattr_in_watchlist,
- priv->complete_watchlist);
- return 0;
+ xattrop_cbk(frame, cookie, this, op_ret, op_errno, xattr, xdata,
+ is_xattr_in_watchlist, priv->complete_watchlist);
+ return 0;
}
int32_t
-index_xattrop64_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xattr,
- dict_t *xdata)
+index_xattrop64_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xattr,
+ dict_t *xdata)
{
- index_priv_t *priv = this->private;
+ index_priv_t *priv = this->private;
- return xattrop_cbk (frame, cookie, this, op_ret, op_errno, xattr, xdata,
- is_xattr_in_watchlist, priv->pending_watchlist);
+ return xattrop_cbk(frame, cookie, this, op_ret, op_errno, xattr, xdata,
+ is_xattr_in_watchlist, priv->pending_watchlist);
}
void
-index_xattrop_do (call_frame_t *frame, xlator_t *this, loc_t *loc,
- fd_t *fd, gf_xattrop_flags_t optype, dict_t *xattr,
- dict_t *xdata)
-{
- int ret = -1;
- int zfilled[XATTROP_TYPE_END] = {0,};
- index_local_t *local = NULL;
- fop_xattrop_cbk_t x_cbk = NULL;
-
- local = frame->local;
-
- if (optype == GF_XATTROP_ADD_ARRAY)
- x_cbk = index_xattrop_cbk;
- else
- x_cbk = index_xattrop64_cbk;
-
- //In wind phase bring the gfid into index. This way if the brick crashes
- //just after posix performs xattrop before _cbk reaches index xlator
- //we will still have the gfid in index.
- memset (zfilled, -1, sizeof (zfilled));
-
- /* Foreach xattr, set corresponding index of zfilled to 1
- * zfilled[index] = 1 implies the xattr's value is zero filled
- * and should be added in its corresponding subdir.
- *
- * zfilled should be set to 1 only for those index that
- * exist in xattr variable. This is to distinguish
- * between different types of volumes.
- * For e.g., if the check is not made,
- * zfilled[DIRTY] is set to 1 for EC volumes,
- * index file will be tried to create in indices/dirty dir
- * which doesn't exist for an EC volume.
- */
- ret = dict_foreach (xattr, index_fill_zero_array, zfilled);
-
- _index_action (this, local->inode, zfilled);
- if (xdata)
- ret = index_entry_action (this, local->inode, xdata,
- GF_XATTROP_ENTRY_IN_KEY);
- if (ret < 0) {
- x_cbk (frame, NULL, this, -1, -ret, NULL, NULL);
- return;
- }
+index_xattrop_do(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
+{
+ int ret = -1;
+ int zfilled[XATTROP_TYPE_END] = {
+ 0,
+ };
+ index_local_t *local = NULL;
+ fop_xattrop_cbk_t x_cbk = NULL;
+
+ local = frame->local;
+
+ if (optype == GF_XATTROP_ADD_ARRAY)
+ x_cbk = index_xattrop_cbk;
+ else
+ x_cbk = index_xattrop64_cbk;
+
+ // In wind phase bring the gfid into index. This way if the brick crashes
+ // just after posix performs xattrop before _cbk reaches index xlator
+ // we will still have the gfid in index.
+ memset(zfilled, -1, sizeof(zfilled));
+
+ /* Foreach xattr, set corresponding index of zfilled to 1
+ * zfilled[index] = 1 implies the xattr's value is zero filled
+ * and should be added in its corresponding subdir.
+ *
+ * zfilled should be set to 1 only for those index that
+ * exist in xattr variable. This is to distinguish
+ * between different types of volumes.
+ * For e.g., if the check is not made,
+ * zfilled[DIRTY] is set to 1 for EC volumes,
+ * index file will be tried to create in indices/dirty dir
+ * which doesn't exist for an EC volume.
+ */
+ ret = dict_foreach(xattr, index_fill_zero_array, zfilled);
+
+ _index_action(this, local->inode, zfilled);
+ if (xdata)
+ ret = index_entry_action(this, local->inode, xdata,
+ GF_XATTROP_ENTRY_IN_KEY);
+ if (ret < 0) {
+ x_cbk(frame, NULL, this, -1, -ret, NULL, NULL);
+ return;
+ }
- if (loc)
- STACK_WIND (frame, x_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->xattrop,
- loc, optype, xattr, xdata);
- else
- STACK_WIND (frame, x_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fxattrop, fd,
- optype, xattr, xdata);
+ if (loc)
+ STACK_WIND(frame, x_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->xattrop, loc, optype, xattr, xdata);
+ else
+ STACK_WIND(frame, x_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fxattrop, fd, optype, xattr, xdata);
}
int
-index_xattrop_wrapper (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
+index_xattrop_wrapper(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
- index_xattrop_do (frame, this, loc, NULL, optype, xattr, xdata);
- return 0;
+ index_xattrop_do(frame, this, loc, NULL, optype, xattr, xdata);
+ return 0;
}
int
-index_fxattrop_wrapper (call_frame_t *frame, xlator_t *this, fd_t *fd,
- gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
+index_fxattrop_wrapper(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
- index_xattrop_do (frame, this, NULL, fd, optype, xattr, xdata);
- return 0;
+ index_xattrop_do(frame, this, NULL, fd, optype, xattr, xdata);
+ return 0;
}
int32_t
-index_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+index_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- call_stub_t *stub = NULL;
- index_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ index_local_t *local = NULL;
- if (!index_xattrop_track (this, flags, dict))
- goto out;
+ if (!index_xattrop_track(this, flags, dict))
+ goto out;
- local = mem_get0 (this->local_pool);
- if (!local)
- goto err;
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
- frame->local = local;
- local->inode = inode_ref (loc->inode);
- if (xdata)
- local->xdata = dict_ref (xdata);
- stub = fop_xattrop_stub (frame, index_xattrop_wrapper,
- loc, flags, dict, xdata);
+ frame->local = local;
+ local->inode = inode_ref(loc->inode);
+ if (xdata)
+ local->xdata = dict_ref(xdata);
+ stub = fop_xattrop_stub(frame, index_xattrop_wrapper, loc, flags, dict,
+ xdata);
err:
- if ((!local) || (!stub)) {
- INDEX_STACK_UNWIND (xattrop, frame, -1, ENOMEM, NULL, NULL);
- return 0;
- }
-
- index_queue_process (this, loc->inode, stub);
+ if ((!local) || (!stub)) {
+ INDEX_STACK_UNWIND(xattrop, frame, -1, ENOMEM, NULL, NULL);
return 0;
+ }
+
+ index_queue_process(this, loc->inode, stub);
+ return 0;
out:
- STACK_WIND (frame, default_xattrop_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->xattrop, loc, flags, dict, xdata);
- return 0;
+ STACK_WIND(frame, default_xattrop_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->xattrop, loc, flags, dict, xdata);
+ return 0;
}
int32_t
-index_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
- gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+index_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- call_stub_t *stub = NULL;
- index_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ index_local_t *local = NULL;
- if (!index_xattrop_track (this, flags, dict))
- goto out;
+ if (!index_xattrop_track(this, flags, dict))
+ goto out;
- local = mem_get0 (this->local_pool);
- if (!local)
- goto err;
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
- frame->local = local;
- local->inode = inode_ref (fd->inode);
- if (xdata)
- local->xdata = dict_ref (xdata);
- stub = fop_fxattrop_stub (frame, index_fxattrop_wrapper,
- fd, flags, dict, xdata);
+ frame->local = local;
+ local->inode = inode_ref(fd->inode);
+ if (xdata)
+ local->xdata = dict_ref(xdata);
+ stub = fop_fxattrop_stub(frame, index_fxattrop_wrapper, fd, flags, dict,
+ xdata);
err:
- if ((!local) || (!stub)) {
- INDEX_STACK_UNWIND (fxattrop, frame, -1, ENOMEM, NULL, xdata);
- return 0;
- }
-
- index_queue_process (this, fd->inode, stub);
+ if ((!local) || (!stub)) {
+ INDEX_STACK_UNWIND(fxattrop, frame, -1, ENOMEM, NULL, xdata);
return 0;
+ }
+
+ index_queue_process(this, fd->inode, stub);
+ return 0;
out:
- STACK_WIND (frame, default_fxattrop_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fxattrop, fd, flags, dict, xdata);
- return 0;
+ STACK_WIND(frame, default_fxattrop_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fxattrop, fd, flags, dict, xdata);
+ return 0;
}
uint64_t
-index_entry_count (xlator_t *this, char *subdir)
+index_entry_count(xlator_t *this, char *subdir)
{
- uint64_t count = 0;
- index_priv_t *priv = NULL;
- DIR *dirp = NULL;
- struct dirent *entry = NULL;
- struct dirent scratch[2] = {{0,},};
- char index_dir[PATH_MAX] = {0,};
+ uint64_t count = 0;
+ index_priv_t *priv = NULL;
+ DIR *dirp = NULL;
+ struct dirent *entry = NULL;
+ struct dirent scratch[2] = {
+ {
+ 0,
+ },
+ };
+ char index_dir[PATH_MAX] = {
+ 0,
+ };
- priv = this->private;
+ priv = this->private;
- make_index_dir_path (priv->index_basepath, subdir,
- index_dir, sizeof (index_dir));
+ make_index_dir_path(priv->index_basepath, subdir, index_dir,
+ sizeof(index_dir));
- dirp = sys_opendir (index_dir);
- if (!dirp)
- return 0;
+ dirp = sys_opendir(index_dir);
+ if (!dirp)
+ return 0;
- for (;;) {
- errno = 0;
- entry = sys_readdir (dirp, scratch);
- if (!entry || errno != 0)
- break;
+ for (;;) {
+ errno = 0;
+ entry = sys_readdir(dirp, scratch);
+ if (!entry || errno != 0)
+ break;
- if (strcmp (entry->d_name, ".") == 0 ||
- strcmp (entry->d_name, "..") == 0)
- continue;
+ if (strcmp(entry->d_name, ".") == 0 || strcmp(entry->d_name, "..") == 0)
+ continue;
- if (!strncmp (entry->d_name, subdir, strlen (subdir)))
- continue;
+ if (!strncmp(entry->d_name, subdir, strlen(subdir)))
+ continue;
- count++;
- }
+ count++;
+ }
- (void) sys_closedir (dirp);
+ (void)sys_closedir(dirp);
- return count;
+ return count;
}
int32_t
-index_getxattr_wrapper (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
-{
- index_priv_t *priv = NULL;
- dict_t *xattr = NULL;
- int ret = 0;
- int vgfid_type = 0;
- uint64_t count = 0;
+index_getxattr_wrapper(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
+{
+ index_priv_t *priv = NULL;
+ dict_t *xattr = NULL;
+ int ret = 0;
+ int vgfid_type = 0;
+ uint64_t count = 0;
+
+ priv = this->private;
+
+ xattr = dict_new();
+ if (!xattr) {
+ ret = -ENOMEM;
+ goto done;
+ }
+
+ vgfid_type = index_get_type_from_vgfid_xattr(name);
+ if (vgfid_type >= 0) {
+ ret = dict_set_static_bin(xattr, (char *)name,
+ priv->internal_vgfid[vgfid_type],
+ sizeof(priv->internal_vgfid[vgfid_type]));
+ if (ret) {
+ ret = -EINVAL;
+ gf_msg(this->name, GF_LOG_ERROR, -ret, INDEX_MSG_DICT_SET_FAILED,
+ "xattrop index "
+ "gfid set failed");
+ goto done;
+ }
+ }
- priv = this->private;
+ /* TODO: Need to check what kind of link-counts are needed for
+ * ENTRY-CHANGES before refactor of this block with array*/
+ if (strcmp(name, GF_XATTROP_INDEX_COUNT) == 0) {
+ count = index_entry_count(this, XATTROP_SUBDIR);
- xattr = dict_new ();
- if (!xattr) {
- ret = -ENOMEM;
- goto done;
+ ret = dict_set_uint64(xattr, (char *)name, count);
+ if (ret) {
+ ret = -EINVAL;
+ gf_msg(this->name, GF_LOG_ERROR, -ret, INDEX_MSG_DICT_SET_FAILED,
+ "xattrop index "
+ "count set failed");
+ goto done;
}
+ } else if (strcmp(name, GF_XATTROP_DIRTY_COUNT) == 0) {
+ count = index_entry_count(this, DIRTY_SUBDIR);
- vgfid_type = index_get_type_from_vgfid_xattr (name);
- if (vgfid_type >= 0) {
- ret = dict_set_static_bin (xattr, (char *)name,
- priv->internal_vgfid[vgfid_type],
- sizeof (priv->internal_vgfid[vgfid_type]));
- if (ret) {
- ret = -EINVAL;
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- INDEX_MSG_DICT_SET_FAILED, "xattrop index "
- "gfid set failed");
- goto done;
- }
+ ret = dict_set_uint64(xattr, (char *)name, count);
+ if (ret) {
+ ret = -EINVAL;
+ gf_msg(this->name, GF_LOG_ERROR, -ret, INDEX_MSG_DICT_SET_FAILED,
+ "dirty index "
+ "count set failed");
+ goto done;
}
-
- /* TODO: Need to check what kind of link-counts are needed for
- * ENTRY-CHANGES before refactor of this block with array*/
- if (strcmp (name, GF_XATTROP_INDEX_COUNT) == 0) {
- count = index_entry_count (this, XATTROP_SUBDIR);
-
- ret = dict_set_uint64 (xattr, (char *)name, count);
- if (ret) {
- ret = -EINVAL;
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- INDEX_MSG_DICT_SET_FAILED, "xattrop index "
- "count set failed");
- goto done;
- }
- } else if (strcmp (name, GF_XATTROP_DIRTY_COUNT) == 0) {
- count = index_entry_count (this, DIRTY_SUBDIR);
-
- ret = dict_set_uint64 (xattr, (char *)name, count);
- if (ret) {
- ret = -EINVAL;
- gf_msg (this->name, GF_LOG_ERROR, -ret,
- INDEX_MSG_DICT_SET_FAILED, "dirty index "
- "count set failed");
- goto done;
- }
- }
+ }
done:
- if (ret)
- STACK_UNWIND_STRICT (getxattr, frame, -1, -ret, xattr, NULL);
- else
- STACK_UNWIND_STRICT (getxattr, frame, 0, 0, xattr, NULL);
+ if (ret)
+ STACK_UNWIND_STRICT(getxattr, frame, -1, -ret, xattr, NULL);
+ else
+ STACK_UNWIND_STRICT(getxattr, frame, 0, 0, xattr, NULL);
- if (xattr)
- dict_unref (xattr);
+ if (xattr)
+ dict_unref(xattr);
- return 0;
+ return 0;
}
static int
-index_save_pargfid_for_entry_changes (xlator_t *this, loc_t *loc, char *path)
+index_save_pargfid_for_entry_changes(xlator_t *this, loc_t *loc, char *path)
{
- index_priv_t *priv = NULL;
- index_inode_ctx_t *ctx = NULL;
- int ret = 0;
-
- priv = this->private;
- if (gf_uuid_compare (loc->pargfid,
- priv->internal_vgfid[ENTRY_CHANGES]))
- return 0;
+ index_priv_t *priv = NULL;
+ index_inode_ctx_t *ctx = NULL;
+ int ret = 0;
- ret = index_inode_ctx_get (loc->inode, this, &ctx);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, EINVAL,
- INDEX_MSG_INODE_CTX_GET_SET_FAILED,
- "Unable to get inode context for %s", path);
- return -EINVAL;
- }
- ret = gf_uuid_parse (loc->name, ctx->virtual_pargfid);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, EINVAL,
- INDEX_MSG_INODE_CTX_GET_SET_FAILED, "Unable to store "
- "virtual gfid in inode context for %s", path);
- return -EINVAL;
- }
+ priv = this->private;
+ if (!loc)
+ return -1;
+ if (gf_uuid_compare(loc->pargfid, priv->internal_vgfid[ENTRY_CHANGES]))
return 0;
+
+ ret = index_inode_ctx_get(loc->inode, this, &ctx);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL,
+ INDEX_MSG_INODE_CTX_GET_SET_FAILED,
+ "Unable to get inode context for %s", path);
+ return -EINVAL;
+ }
+ ret = gf_uuid_parse(loc->name, ctx->virtual_pargfid);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL,
+ INDEX_MSG_INODE_CTX_GET_SET_FAILED,
+ "Unable to store "
+ "virtual gfid in inode context for %s",
+ path);
+ return -EINVAL;
+ }
+ return 0;
}
int32_t
-index_lookup_wrapper (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xattr_req)
-{
- index_priv_t *priv = NULL;
- struct stat lstatbuf = {0};
- int ret = 0;
- int32_t op_errno = EINVAL;
- int32_t op_ret = -1;
- uint64_t val = IA_INVAL;
- char path[PATH_MAX] = {0};
- struct iatt stbuf = {0, };
- struct iatt postparent = {0,};
- dict_t *xattr = NULL;
- gf_boolean_t is_dir = _gf_false;
- char *subdir = NULL;
- loc_t iloc = {0};
-
- priv = this->private;
- loc_copy (&iloc, loc);
-
- VALIDATE_OR_GOTO (loc, done);
- if (index_is_fop_on_internal_inode (this, loc->parent, loc->pargfid)) {
- subdir = index_get_subdir_from_vgfid (priv, loc->pargfid);
- ret = index_inode_path (this, loc->parent, path, sizeof (path));
- if (ret < 0) {
- op_errno = -ret;
- goto done;
- }
- strcat (path, "/");
- strcat (path, (char *)loc->name);
- } else if (index_is_virtual_gfid (priv, loc->gfid)) {
- subdir = index_get_subdir_from_vgfid (priv, loc->gfid);
- make_index_dir_path (priv->index_basepath, subdir,
- path, sizeof (path));
- is_dir = _gf_true;
-
- if ((xattr_req) &&
- (dict_get (xattr_req, GF_INDEX_IA_TYPE_GET_REQ))) {
- if (0 == strcmp (subdir,
- index_get_subdir_from_type(ENTRY_CHANGES)))
- val = IA_IFDIR;
- else
- val = IA_IFREG;
- }
- } else {
- if (!inode_is_linked (loc->inode)) {
- inode_unref (iloc.inode);
- iloc.inode = inode_find (loc->inode->table, loc->gfid);
- }
- ret = index_inode_path (this, iloc.inode, path,
- sizeof (path));
- if (ret < 0) {
- op_errno = -ret;
- goto done;
- }
- }
- ret = sys_lstat (path, &lstatbuf);
- if (ret) {
- gf_msg_debug (this->name, errno, "Stat failed on %s dir ",
- path);
- op_errno = errno;
- goto done;
- } else if (!S_ISDIR (lstatbuf.st_mode) && is_dir) {
- op_errno = ENOTDIR;
- gf_msg_debug (this->name, op_errno, "Stat failed on %s dir, "
- "not a directory", path);
- goto done;
- }
- xattr = dict_new ();
- if (!xattr) {
- op_errno = ENOMEM;
- goto done;
+index_lookup_wrapper(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ dict_t *xattr_req)
+{
+ index_priv_t *priv = NULL;
+ struct stat lstatbuf = {0};
+ int ret = 0;
+ int32_t op_errno = EINVAL;
+ int32_t op_ret = -1;
+ uint64_t val = IA_INVAL;
+ char path[PATH_MAX] = {0};
+ struct iatt stbuf = {
+ 0,
+ };
+ struct iatt postparent = {
+ 0,
+ };
+ dict_t *xattr = NULL;
+ gf_boolean_t is_dir = _gf_false;
+ char *subdir = NULL;
+ loc_t iloc = {0};
+
+ priv = this->private;
+ loc_copy(&iloc, loc);
+
+ VALIDATE_OR_GOTO(loc, done);
+ if (index_is_fop_on_internal_inode(this, loc->parent, loc->pargfid)) {
+ subdir = index_get_subdir_from_vgfid(priv, loc->pargfid);
+ ret = index_inode_path(this, loc->parent, path, sizeof(path));
+ if (ret < 0) {
+ op_errno = -ret;
+ goto done;
}
+ ret = snprintf(path + strlen(path), PATH_MAX - strlen(path), "/%s",
+ loc->name);
- if (val != IA_INVAL) {
- ret = dict_set_uint64 (xattr, GF_INDEX_IA_TYPE_GET_RSP, val);
- if (ret) {
- op_ret = -1;
- op_errno = -ret;
- goto done;
- }
+ if ((ret < 0) || (ret > (PATH_MAX - strlen(path)))) {
+ op_errno = EINVAL;
+ op_ret = -1;
+ goto done;
}
- iatt_from_stat (&stbuf, &lstatbuf);
- if (is_dir || inode_is_linked (iloc.inode))
- loc_gfid (&iloc, stbuf.ia_gfid);
- else
- gf_uuid_generate (stbuf.ia_gfid);
-
- ret = index_save_pargfid_for_entry_changes (this, &iloc, path);
- if (ret) {
- op_ret = -1;
- op_errno = -ret;
- goto done;
- }
+ } else if (index_is_virtual_gfid(priv, loc->gfid)) {
+ subdir = index_get_subdir_from_vgfid(priv, loc->gfid);
+ make_index_dir_path(priv->index_basepath, subdir, path, sizeof(path));
+ is_dir = _gf_true;
- stbuf.ia_ino = -1;
- op_ret = 0;
+ if ((xattr_req) && (dict_get(xattr_req, GF_INDEX_IA_TYPE_GET_REQ))) {
+ if (0 == strcmp(subdir, index_get_subdir_from_type(ENTRY_CHANGES)))
+ val = IA_IFDIR;
+ else
+ val = IA_IFREG;
+ }
+ } else {
+ if (!inode_is_linked(loc->inode)) {
+ inode_unref(iloc.inode);
+ iloc.inode = inode_find(loc->inode->table, loc->gfid);
+ }
+ ret = index_inode_path(this, iloc.inode, path, sizeof(path));
+ if (ret < 0) {
+ op_errno = -ret;
+ goto done;
+ }
+ }
+ ret = sys_lstat(path, &lstatbuf);
+ if (ret) {
+ gf_msg_debug(this->name, errno, "Stat failed on %s dir ", path);
+ op_errno = errno;
+ goto done;
+ } else if (!S_ISDIR(lstatbuf.st_mode) && is_dir) {
+ op_errno = ENOTDIR;
+ gf_msg_debug(this->name, op_errno,
+ "Stat failed on %s dir, "
+ "not a directory",
+ path);
+ goto done;
+ }
+ xattr = dict_new();
+ if (!xattr) {
+ op_errno = ENOMEM;
+ goto done;
+ }
+
+ if (val != IA_INVAL) {
+ ret = dict_set_uint64(xattr, GF_INDEX_IA_TYPE_GET_RSP, val);
+ if (ret) {
+ op_ret = -1;
+ op_errno = -ret;
+ goto done;
+ }
+ }
+
+ iatt_from_stat(&stbuf, &lstatbuf);
+ if (is_dir || inode_is_linked(iloc.inode))
+ loc_gfid(&iloc, stbuf.ia_gfid);
+ else
+ gf_uuid_generate(stbuf.ia_gfid);
+
+ ret = index_save_pargfid_for_entry_changes(this, &iloc, path);
+ if (ret) {
+ op_ret = -1;
+ op_errno = -ret;
+ goto done;
+ }
+
+ stbuf.ia_ino = -1;
+ op_ret = 0;
done:
- STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno,
- loc->inode, &stbuf, xattr, &postparent);
- if (xattr)
- dict_unref (xattr);
- loc_wipe (&iloc);
- return 0;
+ STACK_UNWIND_STRICT(lookup, frame, op_ret, op_errno,
+ loc ? loc->inode : NULL, &stbuf, xattr, &postparent);
+ if (xattr)
+ dict_unref(xattr);
+ loc_wipe(&iloc);
+ return 0;
}
int
-index_get_gfid_type (void *opaque)
-{
- gf_dirent_t *entry = NULL;
- xlator_t *this = THIS;
- struct index_syncop_args *args = opaque;
- loc_t loc = {0};
- struct iatt iatt = {0};
- int ret = 0;
-
- list_for_each_entry (entry, &args->entries->list, list) {
- if (strcmp (entry->d_name, ".") == 0 ||
- strcmp (entry->d_name, "..") == 0)
- continue;
-
- loc_wipe (&loc);
-
- entry->d_type = IA_INVAL;
- if (gf_uuid_parse (entry->d_name, loc.gfid))
- continue;
-
- loc.inode = inode_find (args->parent->table, loc.gfid);
- if (loc.inode) {
- entry->d_type = loc.inode->ia_type;
- continue;
- }
- loc.inode = inode_new (args->parent->table);
- if (!loc.inode)
- continue;
- ret = syncop_lookup (FIRST_CHILD (this), &loc, &iatt, 0, 0, 0);
- if (ret == 0)
- entry->d_type = iatt.ia_type;
- }
- loc_wipe (&loc);
+index_get_gfid_type(void *opaque)
+{
+ gf_dirent_t *entry = NULL;
+ xlator_t *this = THIS;
+ struct index_syncop_args *args = opaque;
+ loc_t loc = {0};
+ struct iatt iatt = {0};
+ int ret = 0;
- return 0;
-}
+ list_for_each_entry(entry, &args->entries->list, list)
+ {
+ if (strcmp(entry->d_name, ".") == 0 || strcmp(entry->d_name, "..") == 0)
+ continue;
-int32_t
-index_readdir_wrapper (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t off, dict_t *xdata)
-{
- index_fd_ctx_t *fctx = NULL;
- index_priv_t *priv = NULL;
- DIR *dir = NULL;
- int ret = -1;
- int32_t op_ret = -1;
- int32_t op_errno = 0;
- int count = 0;
- gf_dirent_t entries;
- struct index_syncop_args args = {0};
-
- priv = this->private;
- INIT_LIST_HEAD (&entries.list);
-
- ret = index_fd_ctx_get (fd, this, &fctx);
- if (ret < 0) {
- op_errno = -ret;
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- INDEX_MSG_FD_OP_FAILED, "pfd is NULL, fd=%p", fd);
- goto done;
- }
+ loc_wipe(&loc);
+
+ entry->d_type = gf_d_type_from_ia_type(IA_INVAL);
+ entry->d_stat.ia_type = IA_INVAL;
+ if (gf_uuid_parse(entry->d_name, loc.gfid))
+ continue;
- dir = fctx->dir;
- if (!dir) {
- op_errno = EINVAL;
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- INDEX_MSG_INDEX_READDIR_FAILED,
- "dir is NULL for fd=%p", fd);
- goto done;
+ loc.inode = inode_find(args->parent->table, loc.gfid);
+ if (loc.inode) {
+ entry->d_stat.ia_type = loc.inode->ia_type;
+ entry->d_type = gf_d_type_from_ia_type(loc.inode->ia_type);
+ continue;
}
+ loc.inode = inode_new(args->parent->table);
+ if (!loc.inode)
+ continue;
+ ret = syncop_lookup(FIRST_CHILD(this), &loc, &iatt, 0, 0, 0);
+ if (ret == 0) {
+ entry->d_type = gf_d_type_from_ia_type(iatt.ia_type);
+ entry->d_stat = iatt;
+ }
+ }
+ loc_wipe(&loc);
- count = index_fill_readdir (fd, fctx, dir, off, size, &entries);
+ return 0;
+}
- /* pick ENOENT to indicate EOF */
- op_errno = errno;
- op_ret = count;
- if (index_is_virtual_gfid (priv, fd->inode->gfid) &&
- xdata && dict_get (xdata, "get-gfid-type")) {
- args.parent = fd->inode;
- args.entries = &entries;
- ret = synctask_new (this->ctx->env, index_get_gfid_type,
- NULL, NULL, &args);
- }
+int32_t
+index_readdir_wrapper(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ size_t size, off_t off, dict_t *xdata)
+{
+ index_fd_ctx_t *fctx = NULL;
+ index_priv_t *priv = NULL;
+ DIR *dir = NULL;
+ int ret = -1;
+ int32_t op_ret = -1;
+ int32_t op_errno = 0;
+ int count = 0;
+ gf_dirent_t entries;
+ struct index_syncop_args args = {0};
+
+ priv = this->private;
+ INIT_LIST_HEAD(&entries.list);
+
+ ret = index_fd_ctx_get(fd, this, &fctx);
+ if (ret < 0) {
+ op_errno = -ret;
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, INDEX_MSG_FD_OP_FAILED,
+ "pfd is NULL, fd=%p", fd);
+ goto done;
+ }
+
+ dir = fctx->dir;
+ if (!dir) {
+ op_errno = EINVAL;
+ gf_msg(this->name, GF_LOG_WARNING, op_errno,
+ INDEX_MSG_INDEX_READDIR_FAILED, "dir is NULL for fd=%p", fd);
+ goto done;
+ }
+
+ count = index_fill_readdir(fd, fctx, dir, off, size, &entries);
+
+ /* pick ENOENT to indicate EOF */
+ op_errno = errno;
+ op_ret = count;
+ if (index_is_virtual_gfid(priv, fd->inode->gfid) && xdata &&
+ dict_get(xdata, "get-gfid-type")) {
+ args.parent = fd->inode;
+ args.entries = &entries;
+ ret = synctask_new(this->ctx->env, index_get_gfid_type, NULL, NULL,
+ &args);
+ }
done:
- STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, &entries, NULL);
- gf_dirent_free (&entries);
- return 0;
+ STACK_UNWIND_STRICT(readdir, frame, op_ret, op_errno, &entries, NULL);
+ gf_dirent_free(&entries);
+ return 0;
}
int
-deletion_handler (const char *fpath, const struct stat *sb, int typeflag,
- struct FTW *ftwbuf)
+deletion_handler(const char *fpath, const struct stat *sb, int typeflag,
+ struct FTW *ftwbuf)
{
- ia_type_t type = IA_INVAL;
+ ia_type_t type = IA_INVAL;
- switch (sb->st_mode & S_IFMT) {
+ switch (sb->st_mode & S_IFMT) {
case S_IFREG:
- sys_unlink (fpath);
- break;
+ sys_unlink(fpath);
+ break;
case S_IFDIR:
- sys_rmdir (fpath);
- break;
+ sys_rmdir(fpath);
+ break;
default:
- type = ia_type_from_st_mode (sb->st_mode);
- gf_msg (THIS->name, GF_LOG_WARNING, EINVAL,
- INDEX_MSG_INVALID_ARGS,
- "%s neither a regular file nor a directory - type:%s",
- fpath, gf_inode_type_to_str (type));
- break;
- }
- return 0;
+ type = ia_type_from_st_mode(sb->st_mode);
+ gf_msg(THIS->name, GF_LOG_WARNING, EINVAL, INDEX_MSG_INVALID_ARGS,
+ "%s neither a regular file nor a directory - type:%s", fpath,
+ gf_inode_type_to_str(type));
+ break;
+ }
+ return 0;
}
static int
-index_wipe_index_subdir (void *opaque)
+index_wipe_index_subdir(void *opaque)
{
- struct index_syncop_args *args = opaque;
+ struct index_syncop_args *args = opaque;
- nftw (args->path, deletion_handler, 1, FTW_DEPTH | FTW_PHYS);
- return 0;
+ nftw(args->path, deletion_handler, 1, FTW_DEPTH | FTW_PHYS);
+ return 0;
}
static void
-index_get_parent_iatt (struct iatt *parent, char *path, loc_t *loc,
- int32_t *op_ret, int32_t *op_errno)
-{
- int ret = -1;
- struct stat lstatbuf = {0,};
-
- ret = sys_lstat (path, &lstatbuf);
- if (ret < 0) {
- *op_ret = -1;
- *op_errno = errno;
- return;
- }
+index_get_parent_iatt(struct iatt *parent, char *path, loc_t *loc,
+ int32_t *op_ret, int32_t *op_errno)
+{
+ int ret = -1;
+ struct stat lstatbuf = {
+ 0,
+ };
+
+ ret = sys_lstat(path, &lstatbuf);
+ if (ret < 0) {
+ *op_ret = -1;
+ *op_errno = errno;
+ return;
+ }
- iatt_from_stat (parent, &lstatbuf);
- gf_uuid_copy (parent->ia_gfid, loc->pargfid);
- parent->ia_ino = -1;
+ iatt_from_stat(parent, &lstatbuf);
+ gf_uuid_copy(parent->ia_gfid, loc->pargfid);
+ parent->ia_ino = -1;
- return;
+ return;
}
int
-index_rmdir_wrapper (call_frame_t *frame, xlator_t *this, loc_t *loc, int flag,
- dict_t *xdata)
-{
- int ret = 0;
- int32_t op_ret = 0;
- int32_t op_errno = 0;
- char *subdir = NULL;
- char index_dir[PATH_MAX] = {0};
- char index_subdir[PATH_MAX] = {0};
- uuid_t gfid = {0};
- struct iatt preparent = {0};
- struct iatt postparent = {0};
- index_priv_t *priv = NULL;
- index_xattrop_type_t type = XATTROP_TYPE_UNSET;
- struct index_syncop_args args = {0,};
-
- priv = this->private;
-
- type = index_get_type_from_vgfid (priv, loc->pargfid);
- subdir = index_get_subdir_from_vgfid (priv, loc->pargfid);
- make_index_dir_path (priv->index_basepath, subdir,
- index_dir, sizeof (index_dir));
-
- index_get_parent_iatt (&preparent, index_dir, loc, &op_ret, &op_errno);
- if (op_ret < 0)
- goto done;
-
- gf_uuid_parse (loc->name, gfid);
- make_gfid_path (priv->index_basepath, subdir, gfid, index_subdir,
- sizeof (index_subdir));
-
- if (flag == 0) {
- ret = index_del (this, gfid, subdir, type);
- if (ret < 0) {
- op_ret = -1;
- op_errno = -ret;
- goto done;
- }
- } else {
- args.path = index_subdir;
- ret = synctask_new (this->ctx->env, index_wipe_index_subdir,
- NULL, NULL, &args);
+index_rmdir_wrapper(call_frame_t *frame, xlator_t *this, loc_t *loc, int flag,
+ dict_t *xdata)
+{
+ int ret = 0;
+ int32_t op_ret = 0;
+ int32_t op_errno = 0;
+ char *subdir = NULL;
+ char index_dir[PATH_MAX] = {0};
+ char index_subdir[PATH_MAX] = {0};
+ uuid_t gfid = {0};
+ struct iatt preparent = {0};
+ struct iatt postparent = {0};
+ index_priv_t *priv = NULL;
+ index_xattrop_type_t type = XATTROP_TYPE_UNSET;
+ struct index_syncop_args args = {
+ 0,
+ };
+
+ priv = this->private;
+
+ type = index_get_type_from_vgfid(priv, loc->pargfid);
+ subdir = index_get_subdir_from_vgfid(priv, loc->pargfid);
+ make_index_dir_path(priv->index_basepath, subdir, index_dir,
+ sizeof(index_dir));
+
+ index_get_parent_iatt(&preparent, index_dir, loc, &op_ret, &op_errno);
+ if (op_ret < 0)
+ goto done;
+
+ gf_uuid_parse(loc->name, gfid);
+ make_gfid_path(priv->index_basepath, subdir, gfid, index_subdir,
+ sizeof(index_subdir));
+
+ if (flag == 0) {
+ ret = index_del(this, gfid, subdir, type);
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = -ret;
+ goto done;
}
+ } else {
+ args.path = index_subdir;
+ ret = synctask_new(this->ctx->env, index_wipe_index_subdir, NULL, NULL,
+ &args);
+ }
- index_get_parent_iatt (&postparent, index_dir, loc, &op_ret, &op_errno);
- if (op_ret < 0)
- goto done;
+ index_get_parent_iatt(&postparent, index_dir, loc, &op_ret, &op_errno);
+ if (op_ret < 0)
+ goto done;
done:
- INDEX_STACK_UNWIND (rmdir, frame, op_ret, op_errno, &preparent,
- &postparent, xdata);
- return 0;
+ INDEX_STACK_UNWIND(rmdir, frame, op_ret, op_errno, &preparent, &postparent,
+ xdata);
+ return 0;
}
int
-index_unlink_wrapper (call_frame_t *frame, xlator_t *this, loc_t *loc, int flag,
- dict_t *xdata)
-{
- index_priv_t *priv = NULL;
- index_inode_ctx_t *ictx = NULL;
- int32_t op_ret = 0;
- int32_t op_errno = 0;
- int ret = 0;
- index_xattrop_type_t type = XATTROP_TYPE_UNSET;
- struct iatt preparent = {0};
- struct iatt postparent = {0};
- char index_dir[PATH_MAX] = {0};
- char filepath[PATH_MAX] = {0};
- uuid_t gfid = {0};
- char *subdir = NULL;
-
- priv = this->private;
- type = index_get_type_from_vgfid (priv, loc->pargfid);
- ret = index_inode_path (this, loc->parent, index_dir,
- sizeof (index_dir));
- if (ret < 0) {
- op_ret = -1;
- op_errno = -ret;
- goto done;
- }
-
- index_get_parent_iatt (&preparent, index_dir, loc, &op_ret, &op_errno);
- if (op_ret < 0)
- goto done;
-
- if (type <= XATTROP_TYPE_UNSET) {
- ret = index_inode_ctx_get (loc->parent, this, &ictx);
- if ((ret == 0) && gf_uuid_is_null (ictx->virtual_pargfid)) {
- ret = -EINVAL;
- }
- if (ret == 0) {
- ret = index_entry_delete (this, ictx->virtual_pargfid,
- (char *)loc->name);
- }
- } else if (type == ENTRY_CHANGES) {
- make_file_path (priv->index_basepath, ENTRY_CHANGES_SUBDIR,
- (char *)loc->name, filepath, sizeof (filepath));
- ret = sys_unlink (filepath);
- } else {
- subdir = index_get_subdir_from_type (type);
- gf_uuid_parse (loc->name, gfid);
- ret = index_del (this, gfid, subdir, type);
- }
- if (ret < 0) {
- op_ret = -1;
- op_errno = -ret;
- goto done;
- }
-
- index_get_parent_iatt (&postparent, index_dir, loc, &op_ret, &op_errno);
- if (op_ret < 0)
- goto done;
+index_unlink_wrapper(call_frame_t *frame, xlator_t *this, loc_t *loc, int flag,
+ dict_t *xdata)
+{
+ index_priv_t *priv = NULL;
+ index_inode_ctx_t *ictx = NULL;
+ int32_t op_ret = 0;
+ int32_t op_errno = 0;
+ int ret = 0;
+ index_xattrop_type_t type = XATTROP_TYPE_UNSET;
+ struct iatt preparent = {0};
+ struct iatt postparent = {0};
+ char index_dir[PATH_MAX] = {0};
+ char filepath[PATH_MAX] = {0};
+ uuid_t gfid = {0};
+ char *subdir = NULL;
+
+ priv = this->private;
+ type = index_get_type_from_vgfid(priv, loc->pargfid);
+ ret = index_inode_path(this, loc->parent, index_dir, sizeof(index_dir));
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = -ret;
+ goto done;
+ }
+
+ index_get_parent_iatt(&preparent, index_dir, loc, &op_ret, &op_errno);
+ if (op_ret < 0)
+ goto done;
+
+ if (type <= XATTROP_TYPE_UNSET) {
+ ret = index_inode_ctx_get(loc->parent, this, &ictx);
+ if ((ret == 0) && gf_uuid_is_null(ictx->virtual_pargfid)) {
+ ret = -EINVAL;
+ }
+ if (ret == 0) {
+ ret = index_entry_delete(this, ictx->virtual_pargfid,
+ (char *)loc->name);
+ }
+ } else if (type == ENTRY_CHANGES) {
+ make_file_path(priv->index_basepath, ENTRY_CHANGES_SUBDIR,
+ (char *)loc->name, filepath, sizeof(filepath));
+ ret = sys_unlink(filepath);
+ } else {
+ subdir = index_get_subdir_from_type(type);
+ gf_uuid_parse(loc->name, gfid);
+ ret = index_del(this, gfid, subdir, type);
+ }
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = -ret;
+ goto done;
+ }
+
+ index_get_parent_iatt(&postparent, index_dir, loc, &op_ret, &op_errno);
+ if (op_ret < 0)
+ goto done;
done:
- INDEX_STACK_UNWIND (unlink, frame, op_ret, op_errno, &preparent,
- &postparent, xdata);
- return 0;
+ INDEX_STACK_UNWIND(unlink, frame, op_ret, op_errno, &preparent, &postparent,
+ xdata);
+ return 0;
}
int32_t
-index_getxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
+index_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- call_stub_t *stub = NULL;
+ call_stub_t *stub = NULL;
- if (!name || (!index_is_vgfid_xattr (name) &&
- strcmp (GF_XATTROP_INDEX_COUNT, name) &&
- strcmp (GF_XATTROP_DIRTY_COUNT, name)))
- goto out;
+ if (!name ||
+ (!index_is_vgfid_xattr(name) && strcmp(GF_XATTROP_INDEX_COUNT, name) &&
+ strcmp(GF_XATTROP_DIRTY_COUNT, name)))
+ goto out;
- stub = fop_getxattr_stub (frame, index_getxattr_wrapper, loc, name,
- xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (getxattr, frame, -1, ENOMEM, NULL, NULL);
- return 0;
- }
- worker_enqueue (this, stub);
+ stub = fop_getxattr_stub(frame, index_getxattr_wrapper, loc, name, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(getxattr, frame, -1, ENOMEM, NULL, NULL);
return 0;
+ }
+ worker_enqueue(this, stub);
+ return 0;
out:
- STACK_WIND (frame, default_getxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
- return 0;
+ STACK_WIND(frame, default_getxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+ return 0;
}
int64_t
-index_fetch_link_count (xlator_t *this, index_xattrop_type_t type)
-{
- index_priv_t *priv = this->private;
- char *subdir = NULL;
- struct stat lstatbuf = {0,};
- int ret = -1;
- int64_t count = -1;
- DIR *dirp = NULL;
- struct dirent *entry = NULL;
- struct dirent scratch[2] = {{0,},};
- char index_dir[PATH_MAX] = {0,};
- char index_path[PATH_MAX] = {0,};
-
- subdir = index_get_subdir_from_type (type);
- make_index_dir_path (priv->index_basepath, subdir,
- index_dir, sizeof (index_dir));
-
- dirp = sys_opendir (index_dir);
- if (!dirp)
- goto out;
+index_fetch_link_count(xlator_t *this, index_xattrop_type_t type)
+{
+ index_priv_t *priv = this->private;
+ char *subdir = NULL;
+ struct stat lstatbuf = {
+ 0,
+ };
+ int ret = -1;
+ int64_t count = -1;
+ DIR *dirp = NULL;
+ struct dirent *entry = NULL;
+ struct dirent scratch[2] = {
+ {
+ 0,
+ },
+ };
+ char index_dir[PATH_MAX] = {
+ 0,
+ };
+ char index_path[PATH_MAX] = {
+ 0,
+ };
+
+ subdir = index_get_subdir_from_type(type);
+ make_index_dir_path(priv->index_basepath, subdir, index_dir,
+ sizeof(index_dir));
+
+ dirp = sys_opendir(index_dir);
+ if (!dirp)
+ goto out;
+
+ for (;;) {
+ errno = 0;
+ entry = sys_readdir(dirp, scratch);
+ if (!entry || errno != 0) {
+ if (count == -1)
+ count = 0;
+ goto out;
+ }
- for (;;) {
- errno = 0;
- entry = sys_readdir (dirp, scratch);
- if (!entry || errno != 0) {
- if (count == -1)
- count = 0;
- goto out;
- }
+ if (strcmp(entry->d_name, ".") == 0 || strcmp(entry->d_name, "..") == 0)
+ continue;
- if (strcmp (entry->d_name, ".") == 0 ||
- strcmp (entry->d_name, "..") == 0)
- continue;
-
- make_file_path (priv->index_basepath, subdir,
- entry->d_name, index_path, sizeof(index_path));
-
- ret = sys_lstat (index_path, &lstatbuf);
- if (ret < 0) {
- count = -2;
- continue;
- } else {
- count = lstatbuf.st_nlink - 1;
- if (count == 0)
- continue;
- else
- break;
- }
+ make_file_path(priv->index_basepath, subdir, entry->d_name, index_path,
+ sizeof(index_path));
+
+ ret = sys_lstat(index_path, &lstatbuf);
+ if (ret < 0) {
+ count = -2;
+ continue;
+ } else {
+ count = lstatbuf.st_nlink - 1;
+ if (count == 0)
+ continue;
+ else
+ break;
}
+ }
out:
- if (dirp)
- (void) sys_closedir (dirp);
- return count;
+ if (dirp)
+ (void)sys_closedir(dirp);
+ return count;
}
-dict_t*
-index_fill_link_count (xlator_t *this, dict_t *xdata)
+dict_t *
+index_fill_link_count(xlator_t *this, dict_t *xdata)
{
- int ret = -1;
- index_priv_t *priv = NULL;
- int64_t count = -1;
+ int ret = -1;
+ index_priv_t *priv = NULL;
+ int64_t count = -1;
- priv = this->private;
- xdata = (xdata) ? dict_ref (xdata) : dict_new ();
- if (!xdata)
- goto out;
+ priv = this->private;
+ xdata = (xdata) ? dict_ref(xdata) : dict_new();
+ if (!xdata)
+ goto out;
- index_get_link_count (priv, &count, XATTROP);
- if (count < 0) {
- count = index_fetch_link_count (this, XATTROP);
- index_set_link_count (priv, count, XATTROP);
- }
+ index_get_link_count(priv, &count, XATTROP);
+ if (count < 0) {
+ count = index_fetch_link_count(this, XATTROP);
+ index_set_link_count(priv, count, XATTROP);
+ }
- if (count == 0) {
- ret = dict_set_int8 (xdata, "link-count", 0);
- if (ret < 0)
- gf_msg (this->name, GF_LOG_ERROR, EINVAL,
- INDEX_MSG_DICT_SET_FAILED,
- "Unable to set link-count");
- } else {
- ret = dict_set_int8 (xdata, "link-count", 1);
- if (ret < 0)
- gf_msg (this->name, GF_LOG_ERROR, EINVAL,
- INDEX_MSG_DICT_SET_FAILED,
- "Unable to set link-count");
- }
+ if (count == 0) {
+ ret = dict_set_int8(xdata, "link-count", 0);
+ if (ret < 0)
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL, INDEX_MSG_DICT_SET_FAILED,
+ "Unable to set link-count");
+ } else {
+ ret = dict_set_int8(xdata, "link-count", 1);
+ if (ret < 0)
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL, INDEX_MSG_DICT_SET_FAILED,
+ "Unable to set link-count");
+ }
out:
- return xdata;
+ return xdata;
}
int32_t
-index_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+index_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata, struct iatt *postparent)
{
-
- xdata = index_fill_link_count (this, xdata);
- STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
- xdata, postparent);
- if (xdata)
- dict_unref (xdata);
- return 0;
+ xdata = index_fill_link_count(this, xdata);
+ STACK_UNWIND_STRICT(lookup, frame, op_ret, op_errno, inode, buf, xdata,
+ postparent);
+ if (xdata)
+ dict_unref(xdata);
+ return 0;
}
int32_t
-index_lookup (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xattr_req)
-{
- inode_t *inode = NULL;
- call_stub_t *stub = NULL;
- char *flag = NULL;
- int ret = -1;
-
- if (!index_is_fop_on_internal_inode (this, loc->parent, loc->pargfid) &&
- !index_is_fop_on_internal_inode (this, loc->inode, loc->gfid)) {
- if (!inode_is_linked (loc->inode)) {
- inode = inode_find (loc->inode->table, loc->gfid);
- if (!index_is_fop_on_internal_inode (this, inode,
- loc->gfid)) {
- inode_unref (inode);
- goto normal;
- }
- inode_unref (inode);
- } else {
- goto normal;
- }
+index_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr_req)
+{
+ inode_t *inode = NULL;
+ call_stub_t *stub = NULL;
+ char *flag = NULL;
+ int ret = -1;
+
+ if (!index_is_fop_on_internal_inode(this, loc->parent, loc->pargfid) &&
+ !index_is_fop_on_internal_inode(this, loc->inode, loc->gfid)) {
+ if (!inode_is_linked(loc->inode)) {
+ inode = inode_find(loc->inode->table, loc->gfid);
+ if (!index_is_fop_on_internal_inode(this, inode, loc->gfid)) {
+ inode_unref(inode);
+ goto normal;
+ }
+ inode_unref(inode);
+ } else {
+ goto normal;
}
+ }
- stub = fop_lookup_stub (frame, index_lookup_wrapper, loc, xattr_req);
- if (!stub) {
- STACK_UNWIND_STRICT (lookup, frame, -1, ENOMEM, loc->inode,
- NULL, NULL, NULL);
- return 0;
- }
- worker_enqueue (this, stub);
+ stub = fop_lookup_stub(frame, index_lookup_wrapper, loc, xattr_req);
+ if (!stub) {
+ STACK_UNWIND_STRICT(lookup, frame, -1, ENOMEM, loc->inode, NULL, NULL,
+ NULL);
return 0;
+ }
+ worker_enqueue(this, stub);
+ return 0;
normal:
- ret = dict_get_str (xattr_req, "link-count", &flag);
- if ((ret == 0) && (strcmp (flag, GF_XATTROP_INDEX_COUNT) == 0)) {
- STACK_WIND (frame, index_lookup_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
- } else {
- STACK_WIND (frame, default_lookup_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
- }
+ ret = dict_get_str_sizen(xattr_req, "link-count", &flag);
+ if ((ret == 0) && (strcmp(flag, GF_XATTROP_INDEX_COUNT) == 0)) {
+ STACK_WIND(frame, index_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
+ } else {
+ STACK_WIND(frame, default_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
+ }
- return 0;
+ return 0;
}
int32_t
-index_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- dict_t *xdata)
-{
- xdata = index_fill_link_count (this, xdata);
- STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, buf, xdata);
- if (xdata)
- dict_unref (xdata);
- return 0;
+index_fstat_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
+{
+ xdata = index_fill_link_count(this, xdata);
+ STACK_UNWIND_STRICT(fstat, frame, op_ret, op_errno, buf, xdata);
+ if (xdata)
+ dict_unref(xdata);
+ return 0;
}
int32_t
-index_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+index_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- int ret = -1;
- char *flag = NULL;
+ int ret = -1;
+ char *flag = NULL;
- ret = dict_get_str (xdata, "link-count", &flag);
- if ((ret == 0) && (strcmp (flag, GF_XATTROP_INDEX_COUNT) == 0)) {
- STACK_WIND (frame, index_fstat_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fstat, fd, xdata);
- } else {
- STACK_WIND (frame, default_fstat_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fstat, fd, xdata);
- }
+ ret = dict_get_str(xdata, "link-count", &flag);
+ if ((ret == 0) && (strcmp(flag, GF_XATTROP_INDEX_COUNT) == 0)) {
+ STACK_WIND(frame, index_fstat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fstat, fd, xdata);
+ } else {
+ STACK_WIND(frame, default_fstat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fstat, fd, xdata);
+ }
- return 0;
+ return 0;
}
int32_t
-index_opendir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, fd_t *fd, dict_t *xdata)
+index_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata)
{
- if (!index_is_fop_on_internal_inode (this, fd->inode, NULL))
- goto normal;
+ if (!index_is_fop_on_internal_inode(this, fd->inode, NULL))
+ goto normal;
- frame->local = NULL;
- STACK_UNWIND_STRICT (opendir, frame, 0, 0, fd, NULL);
- return 0;
+ frame->local = NULL;
+ STACK_UNWIND_STRICT(opendir, frame, 0, 0, fd, NULL);
+ return 0;
normal:
- STACK_WIND (frame, default_opendir_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->opendir, loc, fd, xdata);
- return 0;
+ STACK_WIND(frame, default_opendir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->opendir, loc, fd, xdata);
+ return 0;
}
int32_t
-index_readdir (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t off, dict_t *xdata)
+index_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *xdata)
{
- call_stub_t *stub = NULL;
+ call_stub_t *stub = NULL;
- if (!index_is_fop_on_internal_inode (this, fd->inode, NULL))
- goto out;
+ if (!index_is_fop_on_internal_inode(this, fd->inode, NULL))
+ goto out;
- stub = fop_readdir_stub (frame, index_readdir_wrapper, fd, size, off,
- xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (readdir, frame, -1, ENOMEM, NULL, NULL);
- return 0;
- }
- worker_enqueue (this, stub);
+ stub = fop_readdir_stub(frame, index_readdir_wrapper, fd, size, off, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(readdir, frame, -1, ENOMEM, NULL, NULL);
return 0;
+ }
+ worker_enqueue(this, stub);
+ return 0;
out:
- STACK_WIND (frame, default_readdir_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdir, fd, size, off, xdata);
- return 0;
+ STACK_WIND(frame, default_readdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdir, fd, size, off, xdata);
+ return 0;
}
int
-index_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
- dict_t *xdata)
+index_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
{
- call_stub_t *stub = NULL;
+ call_stub_t *stub = NULL;
- if (!index_is_fop_on_internal_inode (this, loc->parent, NULL))
- goto out;
+ if (!index_is_fop_on_internal_inode(this, loc->parent, NULL))
+ goto out;
- stub = fop_unlink_stub (frame, index_unlink_wrapper, loc, xflag, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (unlink, frame, -1, ENOMEM, NULL, NULL,
- NULL);
- return 0;
- }
- worker_enqueue (this, stub);
+ stub = fop_unlink_stub(frame, index_unlink_wrapper, loc, xflag, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(unlink, frame, -1, ENOMEM, NULL, NULL, NULL);
return 0;
+ }
+ worker_enqueue(this, stub);
+ return 0;
out:
- STACK_WIND (frame, default_unlink_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
- return 0;
+ STACK_WIND(frame, default_unlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
+ return 0;
}
int
-index_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- dict_t *xdata)
+index_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ dict_t *xdata)
{
- call_stub_t *stub = NULL;
+ call_stub_t *stub = NULL;
- if (!index_is_fop_on_internal_inode (this, loc->parent, NULL))
- goto out;
+ if (!index_is_fop_on_internal_inode(this, loc->parent, NULL))
+ goto out;
- stub = fop_rmdir_stub (frame, index_rmdir_wrapper, loc, flags, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (rmdir, frame, -1, ENOMEM, NULL, NULL,
- NULL);
- return 0;
- }
- worker_enqueue (this, stub);
+ stub = fop_rmdir_stub(frame, index_rmdir_wrapper, loc, flags, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(rmdir, frame, -1, ENOMEM, NULL, NULL, NULL);
return 0;
+ }
+ worker_enqueue(this, stub);
+ return 0;
out:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rmdir, loc, flags, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->rmdir,
+ loc, flags, xdata);
+ return 0;
}
int
-index_make_xattrop_watchlist (xlator_t *this, index_priv_t *priv,
- char *watchlist, index_xattrop_type_t type)
-{
- char *delim = NULL;
- char *dup_watchlist = NULL;
- char *key = NULL;
- char *saveptr = NULL;
- dict_t *xattrs = NULL;
- data_t *dummy = NULL;
- int ret = 0;
-
- if (!watchlist)
- return 0;
-
- dup_watchlist = gf_strdup (watchlist);
- if (!dup_watchlist)
- return -1;
-
- xattrs = dict_new ();
- if (!xattrs) {
- ret = -1;
- goto out;
- }
+index_make_xattrop_watchlist(xlator_t *this, index_priv_t *priv,
+ char *watchlist, index_xattrop_type_t type)
+{
+ char *delim = NULL;
+ char *dup_watchlist = NULL;
+ char *key = NULL;
+ char *saveptr = NULL;
+ dict_t *xattrs = NULL;
+ data_t *dummy = NULL;
+ int ret = 0;
+
+ if (!watchlist)
+ return 0;
- dummy = int_to_data (1);
- if (!dummy) {
- ret = -1;
- goto out;
- }
+ dup_watchlist = gf_strdup(watchlist);
+ if (!dup_watchlist)
+ return -1;
- data_ref (dummy);
+ xattrs = dict_new();
+ if (!xattrs) {
+ ret = -1;
+ goto out;
+ }
- delim = ",";
- key = strtok_r (dup_watchlist, delim, &saveptr);
- while (key) {
- if (strlen (key) == 0) {
- ret = -1;
- goto out;
- }
+ dummy = int_to_data(1);
+ if (!dummy) {
+ ret = -1;
+ goto out;
+ }
- ret = dict_set (xattrs, key, dummy);
- if (ret)
- goto out;
+ data_ref(dummy);
- key = strtok_r (NULL, delim, &saveptr);
+ delim = ",";
+ key = strtok_r(dup_watchlist, delim, &saveptr);
+ while (key) {
+ if (strlen(key) == 0) {
+ ret = -1;
+ goto out;
}
- switch (type) {
+ ret = dict_set(xattrs, key, dummy);
+ if (ret)
+ goto out;
+
+ key = strtok_r(NULL, delim, &saveptr);
+ }
+
+ switch (type) {
case DIRTY:
- priv->dirty_watchlist = dict_copy_with_ref (xattrs,
- priv->dirty_watchlist);
- if (!priv->dirty_watchlist) {
- ret = -1;
- goto out;
- }
- break;
+ priv->dirty_watchlist = dict_copy_with_ref(xattrs,
+ priv->dirty_watchlist);
+ if (!priv->dirty_watchlist) {
+ ret = -1;
+ goto out;
+ }
+ break;
case XATTROP:
- priv->pending_watchlist = dict_copy_with_ref (xattrs,
- priv->pending_watchlist);
- if (!priv->pending_watchlist) {
- ret = -1;
- goto out;
- }
- break;
+ priv->pending_watchlist = dict_copy_with_ref(
+ xattrs, priv->pending_watchlist);
+ if (!priv->pending_watchlist) {
+ ret = -1;
+ goto out;
+ }
+ break;
default:
- break;
- }
+ break;
+ }
- ret = 0;
+ ret = 0;
out:
- if (xattrs)
- dict_unref (xattrs);
+ if (xattrs)
+ dict_unref(xattrs);
- GF_FREE (dup_watchlist);
+ GF_FREE(dup_watchlist);
- if (dummy)
- data_unref (dummy);
+ if (dummy)
+ data_unref(dummy);
- return ret;
+ return ret;
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- ret = xlator_mem_acct_init (this, gf_index_mt_end + 1);
+ ret = xlator_mem_acct_init(this, gf_index_mt_end + 1);
- return ret;
+ return ret;
}
int
-init (xlator_t *this)
-{
- int i = 0;
- int ret = -1;
- int64_t count = -1;
- index_priv_t *priv = NULL;
- pthread_attr_t w_attr;
- gf_boolean_t mutex_inited = _gf_false;
- gf_boolean_t cond_inited = _gf_false;
- gf_boolean_t attr_inited = _gf_false;
- char *watchlist = NULL;
- char *dirtylist = NULL;
- char *pendinglist = NULL;
- char *index_base_parent = NULL;
- char *tmp = NULL;
-
- if (!this->children || this->children->next) {
- gf_msg (this->name, GF_LOG_ERROR, EINVAL,
- INDEX_MSG_INVALID_GRAPH,
- "'index' not configured with exactly one child");
- goto out;
- }
-
- if (!this->parents) {
- gf_msg (this->name, GF_LOG_WARNING, EINVAL,
- INDEX_MSG_INVALID_GRAPH,
- "dangling volume. check volfile ");
- }
-
- priv = GF_CALLOC (1, sizeof (*priv), gf_index_mt_priv_t);
- if (!priv)
- goto out;
-
- LOCK_INIT (&priv->lock);
- if ((ret = pthread_cond_init(&priv->cond, NULL)) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ret,
- INDEX_MSG_INVALID_ARGS,
- "pthread_cond_init failed");
- goto out;
- }
- cond_inited = _gf_true;
-
- if ((ret = pthread_mutex_init(&priv->mutex, NULL)) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ret,
- INDEX_MSG_INVALID_ARGS,
- "pthread_mutex_init failed");
- goto out;
- }
- mutex_inited = _gf_true;
-
- if ((ret = pthread_attr_init (&w_attr)) != 0) {
- gf_msg (this->name, GF_LOG_ERROR, ret,
- INDEX_MSG_INVALID_ARGS,
- "pthread_attr_init failed");
- goto out;
- }
- attr_inited = _gf_true;
-
- ret = pthread_attr_setstacksize (&w_attr, INDEX_THREAD_STACK_SIZE);
- if (ret == EINVAL) {
- gf_msg (this->name, GF_LOG_WARNING, ret,
- INDEX_MSG_INVALID_ARGS,
- "Using default thread stack size");
- }
-
- GF_OPTION_INIT ("index-base", priv->index_basepath, path, out);
- tmp = gf_strdup(priv->index_basepath);
- index_base_parent = dirname(tmp);
- if (gf_lstat_dir (index_base_parent, NULL) != 0) {
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, errno,
- INDEX_MSG_INDEX_DIR_CREATE_FAILED,
- "Failed to find parent dir (%s) of index basepath %s.",
- index_base_parent, priv->index_basepath);
- goto out;
- }
-
- GF_OPTION_INIT ("xattrop64-watchlist", watchlist, str, out);
- ret = index_make_xattrop_watchlist (this, priv, watchlist,
- XATTROP);
- if (ret)
- goto out;
-
- GF_OPTION_INIT ("xattrop-dirty-watchlist", dirtylist, str, out);
- ret = index_make_xattrop_watchlist (this, priv, dirtylist,
- DIRTY);
- if (ret)
- goto out;
-
- GF_OPTION_INIT ("xattrop-pending-watchlist", pendinglist, str, out);
- ret = index_make_xattrop_watchlist (this, priv, pendinglist,
- XATTROP);
- if (ret)
- goto out;
-
- if (priv->dirty_watchlist)
- priv->complete_watchlist = dict_copy_with_ref (priv->dirty_watchlist,
+init(xlator_t *this)
+{
+ int i = 0;
+ int ret = -1;
+ int64_t count = -1;
+ index_priv_t *priv = NULL;
+ pthread_attr_t w_attr;
+ gf_boolean_t mutex_inited = _gf_false;
+ gf_boolean_t cond_inited = _gf_false;
+ gf_boolean_t attr_inited = _gf_false;
+ char *watchlist = NULL;
+ char *dirtylist = NULL;
+ char *pendinglist = NULL;
+ char *index_base_parent = NULL;
+ char *tmp = NULL;
+
+ if (!this->children || this->children->next) {
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL, INDEX_MSG_INVALID_GRAPH,
+ "'index' not configured with exactly one child");
+ goto out;
+ }
+
+ if (!this->parents) {
+ gf_msg(this->name, GF_LOG_WARNING, EINVAL, INDEX_MSG_INVALID_GRAPH,
+ "dangling volume. check volfile ");
+ }
+
+ priv = GF_CALLOC(1, sizeof(*priv), gf_index_mt_priv_t);
+ if (!priv)
+ goto out;
+
+ LOCK_INIT(&priv->lock);
+ if ((ret = pthread_cond_init(&priv->cond, NULL)) != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, ret, INDEX_MSG_INVALID_ARGS,
+ "pthread_cond_init failed");
+ goto out;
+ }
+ cond_inited = _gf_true;
+
+ if ((ret = pthread_mutex_init(&priv->mutex, NULL)) != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, ret, INDEX_MSG_INVALID_ARGS,
+ "pthread_mutex_init failed");
+ goto out;
+ }
+ mutex_inited = _gf_true;
+
+ if ((ret = pthread_attr_init(&w_attr)) != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, ret, INDEX_MSG_INVALID_ARGS,
+ "pthread_attr_init failed");
+ goto out;
+ }
+ attr_inited = _gf_true;
+
+ ret = pthread_attr_setstacksize(&w_attr, INDEX_THREAD_STACK_SIZE);
+ if (ret == EINVAL) {
+ gf_msg(this->name, GF_LOG_WARNING, ret, INDEX_MSG_INVALID_ARGS,
+ "Using default thread stack size");
+ }
+
+ GF_OPTION_INIT("index-base", priv->index_basepath, path, out);
+ tmp = gf_strdup(priv->index_basepath);
+ index_base_parent = dirname(tmp);
+ if (gf_lstat_dir(index_base_parent, NULL) != 0) {
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, errno,
+ INDEX_MSG_INDEX_DIR_CREATE_FAILED,
+ "Failed to find parent dir (%s) of index basepath %s.",
+ index_base_parent, priv->index_basepath);
+ goto out;
+ }
+
+ GF_OPTION_INIT("xattrop64-watchlist", watchlist, str, out);
+ ret = index_make_xattrop_watchlist(this, priv, watchlist, XATTROP);
+ if (ret)
+ goto out;
+
+ GF_OPTION_INIT("xattrop-dirty-watchlist", dirtylist, str, out);
+ ret = index_make_xattrop_watchlist(this, priv, dirtylist, DIRTY);
+ if (ret)
+ goto out;
+
+ GF_OPTION_INIT("xattrop-pending-watchlist", pendinglist, str, out);
+ ret = index_make_xattrop_watchlist(this, priv, pendinglist, XATTROP);
+ if (ret)
+ goto out;
+
+ if (priv->dirty_watchlist)
+ priv->complete_watchlist = dict_copy_with_ref(priv->dirty_watchlist,
priv->complete_watchlist);
- if (priv->pending_watchlist)
- priv->complete_watchlist = dict_copy_with_ref (priv->pending_watchlist,
+ if (priv->pending_watchlist)
+ priv->complete_watchlist = dict_copy_with_ref(priv->pending_watchlist,
priv->complete_watchlist);
- gf_uuid_generate (priv->index);
- for (i = 0; i < XATTROP_TYPE_END; i++)
- gf_uuid_generate (priv->internal_vgfid[i]);
-
- INIT_LIST_HEAD (&priv->callstubs);
+ gf_uuid_generate(priv->index);
+ for (i = 0; i < XATTROP_TYPE_END; i++)
+ gf_uuid_generate(priv->internal_vgfid[i]);
- this->local_pool = mem_pool_new (index_local_t, 64);
- if (!this->local_pool) {
- ret = -1;
- goto out;
- }
+ INIT_LIST_HEAD(&priv->callstubs);
+ GF_ATOMIC_INIT(priv->stub_cnt, 0);
- this->private = priv;
+ this->local_pool = mem_pool_new(index_local_t, 64);
+ if (!this->local_pool) {
+ ret = -1;
+ goto out;
+ }
- ret = index_dir_create (this, XATTROP_SUBDIR);
- if (ret < 0)
- goto out;
+ this->private = priv;
- if (priv->dirty_watchlist) {
- ret = index_dir_create (this, DIRTY_SUBDIR);
- if (ret < 0)
- goto out;
- }
+ ret = index_dir_create(this, XATTROP_SUBDIR);
+ if (ret < 0)
+ goto out;
- ret = index_dir_create (this, ENTRY_CHANGES_SUBDIR);
+ if (priv->dirty_watchlist) {
+ ret = index_dir_create(this, DIRTY_SUBDIR);
if (ret < 0)
- goto out;
-
- /*init indices files counts*/
- count = index_fetch_link_count (this, XATTROP);
- index_set_link_count (priv, count, XATTROP);
- priv->down = _gf_false;
-
- ret = gf_thread_create (&priv->thread, &w_attr, index_worker, this,
- "idxwrker");
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, ret,
- INDEX_MSG_WORKER_THREAD_CREATE_FAILED,
- "Failed to create worker thread, aborting");
- goto out;
- }
-
- ret = 0;
+ goto out;
+ }
+
+ ret = index_dir_create(this, ENTRY_CHANGES_SUBDIR);
+ if (ret < 0)
+ goto out;
+
+ /*init indices files counts*/
+ count = index_fetch_link_count(this, XATTROP);
+ index_set_link_count(priv, count, XATTROP);
+ priv->down = _gf_false;
+
+ priv->curr_count = 0;
+ ret = gf_thread_create(&priv->thread, &w_attr, index_worker, this,
+ "idxwrker");
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, ret,
+ INDEX_MSG_WORKER_THREAD_CREATE_FAILED,
+ "Failed to create worker thread, aborting");
+ goto out;
+ }
+ priv->curr_count++;
+ ret = 0;
out:
- GF_FREE(tmp);
-
- if (ret) {
- if (cond_inited)
- pthread_cond_destroy (&priv->cond);
- if (mutex_inited)
- pthread_mutex_destroy (&priv->mutex);
- if (priv && priv->dirty_watchlist)
- dict_unref (priv->dirty_watchlist);
- if (priv && priv->pending_watchlist)
- dict_unref (priv->pending_watchlist);
- if (priv && priv->complete_watchlist)
- dict_unref (priv->complete_watchlist);
- if (priv)
- GF_FREE (priv);
- this->private = NULL;
- mem_pool_destroy (this->local_pool);
- }
+ GF_FREE(tmp);
+
+ if (ret) {
+ if (cond_inited)
+ pthread_cond_destroy(&priv->cond);
+ if (mutex_inited)
+ pthread_mutex_destroy(&priv->mutex);
+ if (priv && priv->dirty_watchlist)
+ dict_unref(priv->dirty_watchlist);
+ if (priv && priv->pending_watchlist)
+ dict_unref(priv->pending_watchlist);
+ if (priv && priv->complete_watchlist)
+ dict_unref(priv->complete_watchlist);
+ if (priv)
+ GF_FREE(priv);
+ this->private = NULL;
+ mem_pool_destroy(this->local_pool);
+ this->local_pool = NULL;
+ }
- if (attr_inited)
- pthread_attr_destroy (&w_attr);
- return ret;
+ if (attr_inited)
+ pthread_attr_destroy(&w_attr);
+ return ret;
}
void
-fini (xlator_t *this)
-{
- index_priv_t *priv = NULL;
-
- priv = this->private;
- if (!priv)
- goto out;
- this->private = NULL;
- LOCK_DESTROY (&priv->lock);
- pthread_cond_destroy (&priv->cond);
- pthread_mutex_destroy (&priv->mutex);
- if (priv->dirty_watchlist)
- dict_unref (priv->dirty_watchlist);
- if (priv->pending_watchlist)
- dict_unref (priv->pending_watchlist);
- if (priv->complete_watchlist)
- dict_unref (priv->complete_watchlist);
- GF_FREE (priv);
- mem_pool_destroy (this->local_pool);
+fini(xlator_t *this)
+{
+ index_priv_t *priv = NULL;
+
+ priv = this->private;
+ if (!priv)
+ goto out;
+
+ priv->down = _gf_true;
+ pthread_cond_broadcast(&priv->cond);
+ if (priv->thread) {
+ gf_thread_cleanup_xint(priv->thread);
+ priv->thread = 0;
+ }
+ this->private = NULL;
+ LOCK_DESTROY(&priv->lock);
+ pthread_cond_destroy(&priv->cond);
+ pthread_mutex_destroy(&priv->mutex);
+ if (priv->dirty_watchlist)
+ dict_unref(priv->dirty_watchlist);
+ if (priv->pending_watchlist)
+ dict_unref(priv->pending_watchlist);
+ if (priv->complete_watchlist)
+ dict_unref(priv->complete_watchlist);
+ GF_FREE(priv);
+
+ if (this->local_pool) {
+ mem_pool_destroy(this->local_pool);
this->local_pool = NULL;
+ }
out:
- return;
+ return;
}
int
-index_forget (xlator_t *this, inode_t *inode)
+index_forget(xlator_t *this, inode_t *inode)
{
- uint64_t tmp_cache = 0;
- if (!inode_ctx_del (inode, this, &tmp_cache))
- GF_FREE ((index_inode_ctx_t*) (long)tmp_cache);
+ uint64_t tmp_cache = 0;
+ if (!inode_ctx_del(inode, this, &tmp_cache))
+ GF_FREE((index_inode_ctx_t *)(long)tmp_cache);
- return 0;
+ return 0;
}
int32_t
-index_releasedir (xlator_t *this, fd_t *fd)
+index_releasedir(xlator_t *this, fd_t *fd)
{
- index_fd_ctx_t *fctx = NULL;
- uint64_t ctx = 0;
- int ret = 0;
+ index_fd_ctx_t *fctx = NULL;
+ uint64_t ctx = 0;
+ int ret = 0;
- ret = fd_ctx_del (fd, this, &ctx);
- if (ret < 0)
- goto out;
+ ret = fd_ctx_del(fd, this, &ctx);
+ if (ret < 0)
+ goto out;
- fctx = (index_fd_ctx_t*) (long) ctx;
- if (fctx->dir) {
- ret = sys_closedir (fctx->dir);
- if (ret)
- gf_msg (this->name, GF_LOG_ERROR, errno,
- INDEX_MSG_FD_OP_FAILED,
- "closedir error");
- }
+ fctx = (index_fd_ctx_t *)(long)ctx;
+ if (fctx->dir) {
+ ret = sys_closedir(fctx->dir);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, errno, INDEX_MSG_FD_OP_FAILED,
+ "closedir error");
+ }
- GF_FREE (fctx);
+ GF_FREE(fctx);
out:
- return 0;
+ return 0;
}
int32_t
-index_release (xlator_t *this, fd_t *fd)
+index_release(xlator_t *this, fd_t *fd)
{
- index_fd_ctx_t *fctx = NULL;
- uint64_t ctx = 0;
- int ret = 0;
+ index_fd_ctx_t *fctx = NULL;
+ uint64_t ctx = 0;
+ int ret = 0;
- ret = fd_ctx_del (fd, this, &ctx);
- if (ret < 0)
- goto out;
+ ret = fd_ctx_del(fd, this, &ctx);
+ if (ret < 0)
+ goto out;
- fctx = (index_fd_ctx_t*) (long) ctx;
- GF_FREE (fctx);
+ fctx = (index_fd_ctx_t *)(long)ctx;
+ GF_FREE(fctx);
out:
- return 0;
+ return 0;
}
int
-notify (xlator_t *this, int event, void *data, ...)
+notify(xlator_t *this, int event, void *data, ...)
{
- int ret = 0;
- index_priv_t *priv = NULL;
+ int ret = 0;
+ index_priv_t *priv = NULL;
+ uint64_t stub_cnt = 0;
+ xlator_t *victim = data;
+ struct timespec sleep_till = {
+ 0,
+ };
- if (!this)
- return 0;
+ if (!this)
+ return 0;
+
+ priv = this->private;
+ if (!priv)
+ return 0;
- priv = this->private;
- if (!priv)
- return 0;
+ if ((event == GF_EVENT_PARENT_DOWN) && victim->cleanup_starting) {
+ stub_cnt = GF_ATOMIC_GET(priv->stub_cnt);
+ timespec_now_realtime(&sleep_till);
+ sleep_till.tv_sec += 1;
- switch (event) {
- case GF_EVENT_CLEANUP:
- priv->down = _gf_true;
- pthread_cond_broadcast (&priv->cond);
- break;
+ /* Wait for draining stub from queue before notify PARENT_DOWN */
+ pthread_mutex_lock(&priv->mutex);
+ {
+ while (stub_cnt) {
+ (void)pthread_cond_timedwait(&priv->cond, &priv->mutex,
+ &sleep_till);
+ stub_cnt = GF_ATOMIC_GET(priv->stub_cnt);
+ }
+ }
+ pthread_mutex_unlock(&priv->mutex);
+ gf_log(this->name, GF_LOG_INFO,
+ "Notify GF_EVENT_PARENT_DOWN for brick %s", victim->name);
+ }
+
+ if ((event == GF_EVENT_CHILD_DOWN) && victim->cleanup_starting) {
+ pthread_mutex_lock(&priv->mutex);
+ {
+ priv->down = _gf_true;
+ pthread_cond_broadcast(&priv->cond);
+ while (priv->curr_count)
+ pthread_cond_wait(&priv->cond, &priv->mutex);
}
+ pthread_mutex_unlock(&priv->mutex);
+
+ gf_log(this->name, GF_LOG_INFO,
+ "Notify GF_EVENT_CHILD_DOWN for brick %s", victim->name);
+ }
- ret = default_notify (this, event, data);
- return ret;
+ ret = default_notify(this, event, data);
+ return ret;
}
struct xlator_fops fops = {
- .xattrop = index_xattrop,
- .fxattrop = index_fxattrop,
-
- //interface functions follow
- .getxattr = index_getxattr,
- .lookup = index_lookup,
- .opendir = index_opendir,
- .readdir = index_readdir,
- .unlink = index_unlink,
- .rmdir = index_rmdir,
- .fstat = index_fstat,
+ .xattrop = index_xattrop,
+ .fxattrop = index_fxattrop,
+
+ // interface functions follow
+ .getxattr = index_getxattr,
+ .lookup = index_lookup,
+ .opendir = index_opendir,
+ .readdir = index_readdir,
+ .unlink = index_unlink,
+ .rmdir = index_rmdir,
+ .fstat = index_fstat,
};
struct xlator_dumpops dumpops;
-struct xlator_cbks cbks = {
- .forget = index_forget,
- .release = index_release,
- .releasedir = index_releasedir
-};
+struct xlator_cbks cbks = {.forget = index_forget,
+ .release = index_release,
+ .releasedir = index_releasedir};
struct volume_options options[] = {
- { .key = {"index-base" },
- .type = GF_OPTION_TYPE_PATH,
- .description = "path where the index files need to be stored",
- .default_value = "{{ brick.path }}/.glusterfs/indices"
- },
- { .key = {"xattrop64-watchlist" },
- .type = GF_OPTION_TYPE_STR,
- .description = "Comma separated list of xattrs that are watched",
- .default_value = "trusted.ec.dirty"
- },
- { .key = {"xattrop-dirty-watchlist" },
- .type = GF_OPTION_TYPE_STR,
- .description = "Comma separated list of xattrs that are watched",
- .default_value = "trusted.afr.dirty"
- },
- { .key = {"xattrop-pending-watchlist" },
- .type = GF_OPTION_TYPE_STR,
- .description = "Comma separated list of xattrs that are watched",
- .default_value = "trusted.afr.{{ volume.name }}"
- },
- { .key = {NULL} },
+ {.key = {"index-base"},
+ .type = GF_OPTION_TYPE_PATH,
+ .description = "path where the index files need to be stored",
+ .default_value = "{{ brick.path }}/.glusterfs/indices"},
+ {.key = {"xattrop64-watchlist"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Comma separated list of xattrs that are watched",
+ .default_value = "trusted.ec.dirty"},
+ {.key = {"xattrop-dirty-watchlist"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Comma separated list of xattrs that are watched",
+ .default_value = "trusted.afr.dirty"},
+ {.key = {"xattrop-pending-watchlist"},
+ .type = GF_OPTION_TYPE_STR,
+ .description = "Comma separated list of xattrs that are watched",
+ .default_value = "trusted.afr.{{ volume.name }}"},
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .notify = notify,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .dumpops = &dumpops,
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "index",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/features/index/src/index.h b/xlators/features/index/src/index.h
index ae9091d69c9..a2b6e6e2570 100644
--- a/xlators/features/index/src/index.h
+++ b/xlators/features/index/src/index.h
@@ -11,78 +11,76 @@
#ifndef __INDEX_H__
#define __INDEX_H__
-#include "xlator.h"
-#include "call-stub.h"
-#include "defaults.h"
-#include "byte-order.h"
-#include "common-utils.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/call-stub.h>
+#include <glusterfs/defaults.h>
+#include <glusterfs/byte-order.h>
+#include <glusterfs/common-utils.h>
#include "index-mem-types.h"
-#define INDEX_THREAD_STACK_SIZE ((size_t)(1024*1024))
+#define INDEX_THREAD_STACK_SIZE ((size_t)(1024 * 1024))
-typedef enum {
- UNKNOWN,
- IN,
- NOTIN
-} index_state_t;
+typedef enum { UNKNOWN, IN, NOTIN } index_state_t;
typedef enum {
- XATTROP_TYPE_UNSET = -1,
- XATTROP,
- DIRTY,
- ENTRY_CHANGES,
- XATTROP_TYPE_END
+ XATTROP_TYPE_UNSET = -1,
+ XATTROP,
+ DIRTY,
+ ENTRY_CHANGES,
+ XATTROP_TYPE_END
} index_xattrop_type_t;
typedef struct index_inode_ctx {
- gf_boolean_t processing;
- struct list_head callstubs;
- int state[XATTROP_TYPE_END];
- uuid_t virtual_pargfid; /* virtual gfid of dir under
- .glusterfs/indices/entry-changes. */
+ gf_boolean_t processing;
+ struct list_head callstubs;
+ int state[XATTROP_TYPE_END];
+ uuid_t virtual_pargfid; /* virtual gfid of dir under
+ .glusterfs/indices/entry-changes. */
} index_inode_ctx_t;
typedef struct index_fd_ctx {
- DIR *dir;
- off_t dir_eof;
+ DIR *dir;
+ off_t dir_eof;
} index_fd_ctx_t;
typedef struct index_priv {
- char *index_basepath;
- char *dirty_basepath;
- uuid_t index;
- gf_lock_t lock;
- uuid_t internal_vgfid[XATTROP_TYPE_END];
- struct list_head callstubs;
- pthread_mutex_t mutex;
- pthread_cond_t cond;
- dict_t *dirty_watchlist;
- dict_t *pending_watchlist;
- dict_t *complete_watchlist;
- int64_t pending_count;
- pthread_t thread;
- gf_boolean_t down;
+ char *index_basepath;
+ char *dirty_basepath;
+ uuid_t index;
+ gf_lock_t lock;
+ uuid_t internal_vgfid[XATTROP_TYPE_END];
+ struct list_head callstubs;
+ pthread_mutex_t mutex;
+ pthread_cond_t cond;
+ dict_t *dirty_watchlist;
+ dict_t *pending_watchlist;
+ dict_t *complete_watchlist;
+ int64_t pending_count;
+ pthread_t thread;
+ gf_boolean_t down;
+ gf_atomic_t stub_cnt;
+ int32_t curr_count;
} index_priv_t;
typedef struct index_local {
- inode_t *inode;
- dict_t *xdata;
+ inode_t *inode;
+ dict_t *xdata;
} index_local_t;
-#define INDEX_STACK_UNWIND(fop, frame, params ...) \
-do { \
- index_local_t *__local = NULL; \
- if (frame) { \
- __local = frame->local; \
- frame->local = NULL; \
- } \
- STACK_UNWIND_STRICT (fop, frame, params); \
- if (__local) { \
- inode_unref (__local->inode); \
- if (__local->xdata) \
- dict_unref (__local->xdata); \
- mem_put (__local); \
- } \
-} while (0)
+#define INDEX_STACK_UNWIND(fop, frame, params...) \
+ do { \
+ index_local_t *__local = NULL; \
+ if (frame) { \
+ __local = frame->local; \
+ frame->local = NULL; \
+ } \
+ STACK_UNWIND_STRICT(fop, frame, params); \
+ if (__local) { \
+ inode_unref(__local->inode); \
+ if (__local->xdata) \
+ dict_unref(__local->xdata); \
+ mem_put(__local); \
+ } \
+ } while (0)
#endif
diff --git a/xlators/features/leases/src/leases-internal.c b/xlators/features/leases/src/leases-internal.c
index 475576eac79..56dee244281 100644
--- a/xlators/features/leases/src/leases-internal.c
+++ b/xlators/features/leases/src/leases-internal.c
@@ -15,7 +15,6 @@
#include "leases.h"
-
/* Mutex locks used in this xlator and their order of acquisition:
* Check lease conflict:
* lease_ctx lock
@@ -53,1299 +52,1361 @@
* _gf_false - lease lock option disabled
*/
gf_boolean_t
-is_leases_enabled (xlator_t *this)
+is_leases_enabled(xlator_t *this)
{
- leases_private_t *priv = NULL;
- gf_boolean_t is_enabled = _gf_false;
+ leases_private_t *priv = NULL;
+ gf_boolean_t is_enabled = _gf_false;
- GF_VALIDATE_OR_GOTO ("leases", this, out);
+ GF_VALIDATE_OR_GOTO("leases", this, out);
- if (this->private) {
- priv = (leases_private_t *)this->private;
- is_enabled = priv->leases_enabled;
- }
+ if (this->private) {
+ priv = (leases_private_t *)this->private;
+ is_enabled = priv->leases_enabled;
+ }
out:
- return is_enabled;
+ return is_enabled;
}
-
/*
* Get the recall_leaselk_timeout
* Return Value:
* timeout value(in seconds) set as an option to this xlator.
* -1 error case
*/
-int32_t
-get_recall_lease_timeout (xlator_t *this)
+static int32_t
+get_recall_lease_timeout(xlator_t *this)
{
- leases_private_t *priv = NULL;
- int32_t timeout = -1;
+ leases_private_t *priv = NULL;
+ int32_t timeout = -1;
- GF_VALIDATE_OR_GOTO ("leases", this, out);
+ GF_VALIDATE_OR_GOTO("leases", this, out);
- if (this->private) {
- priv = (leases_private_t *)this->private;
- timeout = priv->recall_lease_timeout;
- }
+ if (this->private) {
+ priv = (leases_private_t *)this->private;
+ timeout = priv->recall_lease_timeout;
+ }
out:
- return timeout;
+ return timeout;
}
-
static void
-__dump_leases_info (xlator_t *this, lease_inode_ctx_t *lease_ctx)
+__dump_leases_info(xlator_t *this, lease_inode_ctx_t *lease_ctx)
{
- lease_id_entry_t *lease_entry = NULL;
- lease_id_entry_t *tmp = NULL;
-
- GF_VALIDATE_OR_GOTO ("leases", this, out);
- GF_VALIDATE_OR_GOTO ("leases", lease_ctx, out);
-
- gf_msg_debug (this->name, 0, "Lease held on this inode, lease_type: %d,"
- " lease_cnt:%"PRIu64", RD lease:%d, RW lease:%d, "
- "openfd cnt:%"PRIu64, lease_ctx->lease_type,
- lease_ctx->lease_cnt,
- lease_ctx->lease_type_cnt[GF_RD_LEASE],
- lease_ctx->lease_type_cnt[GF_RW_LEASE],
- lease_ctx->openfd_cnt);
-
- list_for_each_entry_safe (lease_entry, tmp,
- &lease_ctx->lease_id_list,
- lease_id_list) {
- gf_msg_debug (this->name, 0, "Leases held by client: %s, lease "
- "ID:%s, RD lease:%d, RW lease:%d, lease_type: %d, "
- "lease_cnt:%"PRIu64, lease_entry->client_uid,
- lease_entry->lease_id,
- lease_entry->lease_type_cnt[GF_RD_LEASE],
- lease_entry->lease_type_cnt[GF_RW_LEASE],
- lease_entry->lease_type, lease_entry->lease_cnt);
- }
+ lease_id_entry_t *lease_entry = NULL;
+ lease_id_entry_t *tmp = NULL;
+
+ GF_VALIDATE_OR_GOTO("leases", this, out);
+ GF_VALIDATE_OR_GOTO("leases", lease_ctx, out);
+
+ gf_msg_debug(this->name, 0,
+ "Lease held on this inode, lease_type: %d,"
+ " lease_cnt:%" PRIu64
+ ", RD lease:%d, RW lease:%d, "
+ "openfd cnt:%" PRIu64,
+ lease_ctx->lease_type, lease_ctx->lease_cnt,
+ lease_ctx->lease_type_cnt[GF_RD_LEASE],
+ lease_ctx->lease_type_cnt[GF_RW_LEASE], lease_ctx->openfd_cnt);
+
+ list_for_each_entry_safe(lease_entry, tmp, &lease_ctx->lease_id_list,
+ lease_id_list)
+ {
+ gf_msg_debug(this->name, 0,
+ "Leases held by client: %s, lease "
+ "ID:%s, RD lease:%d, RW lease:%d, lease_type: %d, "
+ "lease_cnt:%" PRIu64,
+ lease_entry->client_uid, lease_entry->lease_id,
+ lease_entry->lease_type_cnt[GF_RD_LEASE],
+ lease_entry->lease_type_cnt[GF_RW_LEASE],
+ lease_entry->lease_type, lease_entry->lease_cnt);
+ }
out:
- return;
+ return;
}
-
static int
-__lease_ctx_set (inode_t *inode, xlator_t *this)
+__lease_ctx_set(inode_t *inode, xlator_t *this)
{
- lease_inode_ctx_t *inode_ctx = NULL;
- int ret = -1;
- uint64_t ctx = 0;
-
- GF_VALIDATE_OR_GOTO ("leases", inode, out);
- GF_VALIDATE_OR_GOTO ("leases", this, out);
-
- ret = __inode_ctx_get (inode, this, &ctx);
- if (!ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, LEASE_MSG_INVAL_INODE_CTX,
- "inode_ctx_get failed");
- goto out;
- }
-
- inode_ctx = GF_CALLOC (1, sizeof (*inode_ctx),
- gf_leases_mt_lease_inode_ctx_t);
- GF_CHECK_ALLOC (inode_ctx, ret, out);
-
- pthread_mutex_init (&inode_ctx->lock, NULL);
- INIT_LIST_HEAD (&inode_ctx->lease_id_list);
- INIT_LIST_HEAD (&inode_ctx->blocked_list);
-
- inode_ctx->lease_cnt = 0;
-
- ret = __inode_ctx_set (inode, this, (uint64_t *) inode_ctx);
- if (ret) {
- GF_FREE (inode_ctx);
- gf_msg (this->name, GF_LOG_INFO, 0, LEASE_MSG_INVAL_INODE_CTX,
- "failed to set inode ctx (%p)", inode);
- }
+ lease_inode_ctx_t *inode_ctx = NULL;
+ int ret = -1;
+ uint64_t ctx = 0;
+
+ GF_VALIDATE_OR_GOTO("leases", inode, out);
+ GF_VALIDATE_OR_GOTO("leases", this, out);
+
+ ret = __inode_ctx_get(inode, this, &ctx);
+ if (!ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, LEASE_MSG_INVAL_INODE_CTX,
+ "inode_ctx_get failed");
+ goto out;
+ }
+
+ inode_ctx = GF_CALLOC(1, sizeof(*inode_ctx),
+ gf_leases_mt_lease_inode_ctx_t);
+ GF_CHECK_ALLOC(inode_ctx, ret, out);
+
+ pthread_mutex_init(&inode_ctx->lock, NULL);
+ INIT_LIST_HEAD(&inode_ctx->lease_id_list);
+ INIT_LIST_HEAD(&inode_ctx->blocked_list);
+
+ inode_ctx->lease_cnt = 0;
+
+ ret = __inode_ctx_set(inode, this, (uint64_t *)inode_ctx);
+ if (ret) {
+ GF_FREE(inode_ctx);
+ gf_msg(this->name, GF_LOG_INFO, 0, LEASE_MSG_INVAL_INODE_CTX,
+ "failed to set inode ctx (%p)", inode);
+ }
out:
- return ret;
+ return ret;
}
-
static lease_inode_ctx_t *
-__lease_ctx_get (inode_t *inode, xlator_t *this)
+__lease_ctx_get(inode_t *inode, xlator_t *this)
{
- lease_inode_ctx_t *inode_ctx = NULL;
- uint64_t ctx = 0;
- int ret = 0;
+ lease_inode_ctx_t *inode_ctx = NULL;
+ uint64_t ctx = 0;
+ int ret = 0;
- GF_VALIDATE_OR_GOTO ("leases", inode, out);
- GF_VALIDATE_OR_GOTO ("leases", this, out);
+ GF_VALIDATE_OR_GOTO("leases", inode, out);
+ GF_VALIDATE_OR_GOTO("leases", this, out);
- ret = __inode_ctx_get (inode, this, &ctx);
+ ret = __inode_ctx_get(inode, this, &ctx);
+ if (ret < 0) {
+ ret = __lease_ctx_set(inode, this);
+ if (ret < 0)
+ goto out;
+
+ ret = __inode_ctx_get(inode, this, &ctx);
if (ret < 0) {
- ret = __lease_ctx_set (inode, this);
- if (ret < 0)
- goto out;
-
- ret = __inode_ctx_get (inode, this, &ctx);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, 0, LEASE_MSG_INVAL_INODE_CTX,
- "failed to get inode ctx (%p)", inode);
- goto out;
- }
+ gf_msg(this->name, GF_LOG_WARNING, 0, LEASE_MSG_INVAL_INODE_CTX,
+ "failed to get inode ctx (%p)", inode);
+ goto out;
}
+ }
- inode_ctx = (lease_inode_ctx_t *)(long) ctx;
+ inode_ctx = (lease_inode_ctx_t *)(long)ctx;
out:
- return inode_ctx;
+ return inode_ctx;
}
-
lease_inode_ctx_t *
-lease_ctx_get (inode_t *inode, xlator_t *this)
+lease_ctx_get(inode_t *inode, xlator_t *this)
{
- lease_inode_ctx_t *inode_ctx = NULL;
+ lease_inode_ctx_t *inode_ctx = NULL;
- GF_VALIDATE_OR_GOTO ("leases", inode, out);
- GF_VALIDATE_OR_GOTO ("leases", this, out);
+ GF_VALIDATE_OR_GOTO("leases", inode, out);
+ GF_VALIDATE_OR_GOTO("leases", this, out);
- LOCK (&inode->lock);
- {
- inode_ctx = __lease_ctx_get (inode, this);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ inode_ctx = __lease_ctx_get(inode, this);
+ }
+ UNLOCK(&inode->lock);
out:
- return inode_ctx;
+ return inode_ctx;
}
-
static lease_id_entry_t *
-new_lease_id_entry (call_frame_t *frame, const char *lease_id)
+new_lease_id_entry(call_frame_t *frame, const char *lease_id)
{
- lease_id_entry_t *lease_entry = NULL;
-
- GF_VALIDATE_OR_GOTO ("leases", frame, out);
- GF_VALIDATE_OR_GOTO ("leases", lease_id, out);
+ lease_id_entry_t *lease_entry = NULL;
- lease_entry = GF_CALLOC (1, sizeof (*lease_entry),
- gf_leases_mt_lease_id_entry_t);
- if (!lease_entry) {
- gf_msg (frame->this->name, GF_LOG_ERROR, ENOMEM, LEASE_MSG_NO_MEM,
- "Memory allocation for lease_entry failed");
- return NULL;
- }
-
- INIT_LIST_HEAD (&lease_entry->lease_id_list);
- lease_entry->lease_type = NONE;
- lease_entry->lease_cnt = 0;
- lease_entry->recall_time =
- get_recall_lease_timeout (frame->this);
- lease_entry->client_uid = gf_strdup (frame->root->client->client_uid);
- if (!lease_entry->client_uid) {
- gf_msg (frame->this->name, GF_LOG_ERROR, ENOMEM, LEASE_MSG_NO_MEM,
- "Memory allocation for client_uid failed");
- GF_FREE (lease_entry);
- lease_entry = NULL;
- goto out;
- }
+ GF_VALIDATE_OR_GOTO("leases", frame, out);
+ GF_VALIDATE_OR_GOTO("leases", lease_id, out);
- memcpy (lease_entry->lease_id, lease_id, LEASE_ID_SIZE);
+ lease_entry = GF_CALLOC(1, sizeof(*lease_entry),
+ gf_leases_mt_lease_id_entry_t);
+ if (!lease_entry) {
+ gf_msg(frame->this->name, GF_LOG_ERROR, ENOMEM, LEASE_MSG_NO_MEM,
+ "Memory allocation for lease_entry failed");
+ return NULL;
+ }
+
+ INIT_LIST_HEAD(&lease_entry->lease_id_list);
+ lease_entry->lease_type = NONE;
+ lease_entry->lease_cnt = 0;
+ lease_entry->recall_time = get_recall_lease_timeout(frame->this);
+ lease_entry->client_uid = gf_strdup(frame->root->client->client_uid);
+ if (!lease_entry->client_uid) {
+ gf_msg(frame->this->name, GF_LOG_ERROR, ENOMEM, LEASE_MSG_NO_MEM,
+ "Memory allocation for client_uid failed");
+ GF_FREE(lease_entry);
+ lease_entry = NULL;
+ goto out;
+ }
+
+ memcpy(lease_entry->lease_id, lease_id, LEASE_ID_SIZE);
out:
- return lease_entry;
+ return lease_entry;
}
-
static void
-__destroy_lease_id_entry (lease_id_entry_t *lease_entry)
+__destroy_lease_id_entry(lease_id_entry_t *lease_entry)
{
- GF_VALIDATE_OR_GOTO ("leases", lease_entry, out);
+ GF_VALIDATE_OR_GOTO("leases", lease_entry, out);
- list_del_init (&lease_entry->lease_id_list);
- GF_FREE (lease_entry->client_uid);
- GF_FREE (lease_entry);
+ list_del_init(&lease_entry->lease_id_list);
+ GF_FREE(lease_entry->client_uid);
+ GF_FREE(lease_entry);
out:
- return;
+ return;
}
-
static inline gf_boolean_t
-__is_same_lease_id (const char *k1, const char *k2)
+__is_same_lease_id(const char *k1, const char *k2)
{
- if (memcmp(k1, k2, strlen(k1)) == 0)
- return _gf_true;
+ if (memcmp(k1, k2, strlen(k1)) == 0)
+ return _gf_true;
- return _gf_false;
+ return _gf_false;
}
-
/* Checks if there are any leases, other than the leases taken
* by the given lease_id
*/
static gf_boolean_t
-__another_lease_found (lease_inode_ctx_t *lease_ctx, const char *lease_id)
+__another_lease_found(lease_inode_ctx_t *lease_ctx, const char *lease_id)
{
- lease_id_entry_t *lease_entry = NULL;
- lease_id_entry_t *tmp = NULL;
- gf_boolean_t found_lease = _gf_false;
-
- GF_VALIDATE_OR_GOTO ("leases", lease_id, out);
- GF_VALIDATE_OR_GOTO ("leases", lease_ctx, out);
-
- list_for_each_entry_safe (lease_entry, tmp,
- &lease_ctx->lease_id_list,
- lease_id_list) {
-
- if (!__is_same_lease_id (lease_id, lease_entry->lease_id)) {
- if (lease_entry->lease_cnt > 0) {
- found_lease = _gf_true;
- break;
- }
- }
+ lease_id_entry_t *lease_entry = NULL;
+ lease_id_entry_t *tmp = NULL;
+ gf_boolean_t found_lease = _gf_false;
+
+ GF_VALIDATE_OR_GOTO("leases", lease_id, out);
+ GF_VALIDATE_OR_GOTO("leases", lease_ctx, out);
+
+ list_for_each_entry_safe(lease_entry, tmp, &lease_ctx->lease_id_list,
+ lease_id_list)
+ {
+ if (!__is_same_lease_id(lease_id, lease_entry->lease_id)) {
+ if (lease_entry->lease_cnt > 0) {
+ found_lease = _gf_true;
+ break;
+ }
}
+ }
out:
- return found_lease;
+ return found_lease;
}
-
/* Returns the lease_id_entry for a given lease_id and a given inode.
* Return values:
* NULL - If no client entry found
* lease_id_entry_t* - a pointer to the client entry if found
*/
static lease_id_entry_t *
-__get_lease_id_entry (lease_inode_ctx_t *lease_ctx, const char *lease_id)
+__get_lease_id_entry(lease_inode_ctx_t *lease_ctx, const char *lease_id)
{
- lease_id_entry_t *lease_entry = NULL;
- lease_id_entry_t *tmp = NULL;
- lease_id_entry_t *found = NULL;
-
- GF_VALIDATE_OR_GOTO ("leases", lease_id, out);
- GF_VALIDATE_OR_GOTO ("leases", lease_ctx, out);
-
- list_for_each_entry_safe (lease_entry, tmp,
- &lease_ctx->lease_id_list,
- lease_id_list) {
-
- if (__is_same_lease_id (lease_id, lease_entry->lease_id)) {
- found = lease_entry;
- gf_msg_debug ("leases", 0, "lease ID entry found "
- "Client UID:%s, lease id:%s",
- lease_entry->client_uid,
- leaseid_utoa (lease_entry->lease_id));
- break;
- }
+ lease_id_entry_t *lease_entry = NULL;
+ lease_id_entry_t *tmp = NULL;
+ lease_id_entry_t *found = NULL;
+
+ GF_VALIDATE_OR_GOTO("leases", lease_id, out);
+ GF_VALIDATE_OR_GOTO("leases", lease_ctx, out);
+
+ list_for_each_entry_safe(lease_entry, tmp, &lease_ctx->lease_id_list,
+ lease_id_list)
+ {
+ if (__is_same_lease_id(lease_id, lease_entry->lease_id)) {
+ found = lease_entry;
+ gf_msg_debug("leases", 0,
+ "lease ID entry found "
+ "Client UID:%s, lease id:%s",
+ lease_entry->client_uid,
+ leaseid_utoa(lease_entry->lease_id));
+ break;
}
+ }
out:
- return found;
+ return found;
}
-
/* Returns the lease_id_entry for a given lease_id and a given inode,
* if none found creates one.
* Return values:
* lease_id_entry_t* - a pointer to the client entry
*/
static lease_id_entry_t *
-__get_or_new_lease_entry (call_frame_t *frame, const char *lease_id,
- lease_inode_ctx_t *lease_ctx)
+__get_or_new_lease_entry(call_frame_t *frame, const char *lease_id,
+ lease_inode_ctx_t *lease_ctx)
{
- lease_id_entry_t *lease_entry = NULL;
-
- GF_VALIDATE_OR_GOTO ("leases", frame, out);
- GF_VALIDATE_OR_GOTO ("leases", lease_id, out);
- GF_VALIDATE_OR_GOTO ("leases", lease_ctx, out);
-
- lease_entry = __get_lease_id_entry (lease_ctx, lease_id);
- if (!lease_entry) { /* create one */
- lease_entry = new_lease_id_entry (frame, lease_id);
- if (!lease_entry)
- goto out;
-
- list_add_tail (&lease_entry->lease_id_list,
- &lease_ctx->lease_id_list);
-
- gf_msg_debug (frame->this->name, 0, "lease ID entry added,"
- " Client UID:%s, lease id:%s",
- lease_entry->client_uid,
- leaseid_utoa (lease_entry->lease_id));
- }
+ lease_id_entry_t *lease_entry = NULL;
+
+ GF_VALIDATE_OR_GOTO("leases", frame, out);
+ GF_VALIDATE_OR_GOTO("leases", lease_id, out);
+ GF_VALIDATE_OR_GOTO("leases", lease_ctx, out);
+
+ lease_entry = __get_lease_id_entry(lease_ctx, lease_id);
+ if (!lease_entry) { /* create one */
+ lease_entry = new_lease_id_entry(frame, lease_id);
+ if (!lease_entry)
+ goto out;
+
+ list_add_tail(&lease_entry->lease_id_list, &lease_ctx->lease_id_list);
+
+ gf_msg_debug(frame->this->name, 0,
+ "lease ID entry added,"
+ " Client UID:%s, lease id:%s",
+ lease_entry->client_uid,
+ leaseid_utoa(lease_entry->lease_id));
+ }
out:
- return lease_entry;
+ return lease_entry;
}
-
static lease_inode_t *
-new_lease_inode (inode_t *inode)
+new_lease_inode(inode_t *inode)
{
- lease_inode_t *l_inode = NULL;
+ lease_inode_t *l_inode = GF_MALLOC(sizeof(*l_inode),
+ gf_leases_mt_lease_inode_t);
+ if (!l_inode)
+ goto out;
- l_inode = GF_CALLOC (1, sizeof (*l_inode), gf_leases_mt_lease_inode_t);
- if (!l_inode)
- goto out;
-
- INIT_LIST_HEAD (&l_inode->list);
- l_inode->inode = inode_ref (inode);
+ INIT_LIST_HEAD(&l_inode->list);
+ l_inode->inode = inode_ref(inode);
out:
- return l_inode;
+ return l_inode;
}
-
static void
-__destroy_lease_inode (lease_inode_t *l_inode)
+__destroy_lease_inode(lease_inode_t *l_inode)
{
- list_del_init (&l_inode->list);
- inode_unref (l_inode->inode);
- GF_FREE (l_inode);
+ list_del_init(&l_inode->list);
+ inode_unref(l_inode->inode);
+ GF_FREE(l_inode);
}
-
static lease_client_t *
-new_lease_client (const char *client_uid)
+new_lease_client(const char *client_uid)
{
- lease_client_t *clnt = NULL;
-
- clnt = GF_CALLOC (1, sizeof (*clnt), gf_leases_mt_lease_client_t);
- if (!clnt)
- goto out;
-
- INIT_LIST_HEAD (&clnt->client_list);
- INIT_LIST_HEAD (&clnt->inode_list);
- clnt->client_uid = gf_strdup (client_uid);
+ lease_client_t *clnt = GF_MALLOC(sizeof(*clnt),
+ gf_leases_mt_lease_client_t);
+ if (!clnt)
+ goto out;
+
+ INIT_LIST_HEAD(&clnt->client_list);
+ INIT_LIST_HEAD(&clnt->inode_list);
+ clnt->client_uid = gf_strdup(client_uid);
out:
- return clnt;
+ return clnt;
}
-
static void
-__destroy_lease_client (lease_client_t *clnt)
+__destroy_lease_client(lease_client_t *clnt)
{
- list_del_init (&clnt->inode_list);
- list_del_init (&clnt->client_list);
- GF_FREE (clnt);
+ list_del_init(&clnt->inode_list);
+ list_del_init(&clnt->client_list);
+ GF_FREE(clnt);
- return;
+ return;
}
-
static lease_client_t *
-__get_lease_client (xlator_t *this, leases_private_t *priv,
- const char *client_uid)
+__get_lease_client(xlator_t *this, leases_private_t *priv,
+ const char *client_uid)
{
- lease_client_t *clnt = NULL;
- lease_client_t *tmp = NULL;
- lease_client_t *found = NULL;
-
- list_for_each_entry_safe (clnt, tmp, &priv->client_list, client_list) {
- if ((strcmp (clnt->client_uid, client_uid) == 0)) {
- found = clnt;
- gf_msg_debug (this->name, 0, "Client:%s already found "
- "in the cleanup list", client_uid);
- break;
- }
+ lease_client_t *clnt = NULL;
+ lease_client_t *tmp = NULL;
+ lease_client_t *found = NULL;
+
+ list_for_each_entry_safe(clnt, tmp, &priv->client_list, client_list)
+ {
+ if ((strcmp(clnt->client_uid, client_uid) == 0)) {
+ found = clnt;
+ gf_msg_debug(this->name, 0,
+ "Client:%s already found "
+ "in the cleanup list",
+ client_uid);
+ break;
}
- return found;
+ }
+ return found;
}
-
static lease_client_t *
-__get_or_new_lease_client (xlator_t *this, leases_private_t *priv,
- const char *client_uid)
+__get_or_new_lease_client(xlator_t *this, leases_private_t *priv,
+ const char *client_uid)
{
- lease_client_t *found = NULL;
-
- found = __get_lease_client (this, priv, client_uid);
- if (!found) {
- found = new_lease_client (client_uid);
- if (!found)
- goto out;
- list_add_tail (&found->client_list, &priv->client_list);
- gf_msg_debug (this->name, 0, "Adding a new client:%s entry "
- "to the cleanup list", client_uid);
- }
+ lease_client_t *found = NULL;
+
+ found = __get_lease_client(this, priv, client_uid);
+ if (!found) {
+ found = new_lease_client(client_uid);
+ if (!found)
+ goto out;
+ list_add_tail(&found->client_list, &priv->client_list);
+ gf_msg_debug(this->name, 0,
+ "Adding a new client:%s entry "
+ "to the cleanup list",
+ client_uid);
+ }
out:
- return found;
+ return found;
}
-
static int
-add_inode_to_client_list (xlator_t *this, inode_t *inode, const char *client_uid)
+add_inode_to_client_list(xlator_t *this, inode_t *inode, const char *client_uid)
{
- int ret = 0;
- leases_private_t *priv = NULL;
- lease_client_t *clnt = NULL;
- lease_inode_t *lease_inode = NULL;
-
- priv = this->private;
- pthread_mutex_lock (&priv->mutex);
- {
- clnt = __get_or_new_lease_client (this, priv, client_uid);
- GF_CHECK_ALLOC (clnt, ret, out);
-
- lease_inode = new_lease_inode (inode);
- GF_CHECK_ALLOC (lease_inode, ret, out);
-
- list_add_tail (&clnt->inode_list, &lease_inode->list);
- gf_msg_debug (this->name, 0,
- "Added a new inode:%p to the client(%s) "
- "cleanup list, gfid(%s)", inode, client_uid,
- uuid_utoa (inode->gfid));
+ leases_private_t *priv = this->private;
+ lease_client_t *clnt = NULL;
+
+ lease_inode_t *lease_inode = new_lease_inode(inode);
+ if (!lease_inode)
+ return -ENOMEM;
+
+ pthread_mutex_lock(&priv->mutex);
+ {
+ clnt = __get_or_new_lease_client(this, priv, client_uid);
+ if (!clnt) {
+ pthread_mutex_unlock(&priv->mutex);
+ __destroy_lease_inode(lease_inode);
+ return -ENOMEM;
}
-out:
- pthread_mutex_unlock (&priv->mutex);
- return ret;
+ list_add_tail(&clnt->inode_list, &lease_inode->list);
+ }
+ pthread_mutex_unlock(&priv->mutex);
+ gf_msg_debug(this->name, 0,
+ "Added a new inode:%p to the client(%s) "
+ "cleanup list, gfid(%s)",
+ inode, client_uid, uuid_utoa(inode->gfid));
+ return 0;
}
-
/* Add lease entry to the corresponding client entry.
* Return values:
* 0 Success
* -1 Failure
*/
static int
-__add_lease (call_frame_t *frame, inode_t *inode, lease_inode_ctx_t *lease_ctx,
- const char *client_uid, struct gf_lease *lease)
+__add_lease(call_frame_t *frame, inode_t *inode, lease_inode_ctx_t *lease_ctx,
+ const char *client_uid, struct gf_lease *lease)
{
- lease_id_entry_t *lease_entry = NULL;
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO ("leases", frame, out);
- GF_VALIDATE_OR_GOTO ("leases", client_uid, out);
- GF_VALIDATE_OR_GOTO ("leases", lease_ctx, out);
- GF_VALIDATE_OR_GOTO ("leases", inode, out);
- GF_VALIDATE_OR_GOTO ("leases", lease, out);
-
- gf_msg_trace (frame->this->name, 0,
- "Granting lease lock to client %s with lease id %s"
- " on gfid(%s)", client_uid, leaseid_utoa (lease->lease_id),
- uuid_utoa (inode->gfid));
-
- lease_entry = __get_or_new_lease_entry (frame, lease->lease_id, lease_ctx);
- if (!lease_entry) {
- errno = ENOMEM;
- goto out;
- }
-
- lease_entry->lease_type_cnt[lease->lease_type]++;
- lease_entry->lease_cnt++;
- lease_entry->lease_type |= lease->lease_type;
- /* If this is the first lease taken by the client on the file, then
- * add this inode/file to the client disconnect cleanup list
- */
- if (lease_entry->lease_cnt == 1) {
- add_inode_to_client_list (frame->this, inode, client_uid);
- }
-
- lease_ctx->lease_cnt++;
- lease_ctx->lease_type_cnt[lease->lease_type]++;
- lease_ctx->lease_type |= lease->lease_type;
-
- /* Take a ref for the first lock taken on this inode. Corresponding
- * unref when all the leases are unlocked or during DISCONNECT
- * Ref is required because the inode on which lease is acquired should
- * not be deleted when lru cleanup kicks in*/
- if (lease_ctx->lease_cnt == 1) {
- lease_ctx->inode = inode_ref (inode);
- }
-
- ret = 0;
+ lease_id_entry_t *lease_entry = NULL;
+ int ret = -1;
+
+ GF_VALIDATE_OR_GOTO("leases", frame, out);
+ GF_VALIDATE_OR_GOTO("leases", client_uid, out);
+ GF_VALIDATE_OR_GOTO("leases", lease_ctx, out);
+ GF_VALIDATE_OR_GOTO("leases", inode, out);
+ GF_VALIDATE_OR_GOTO("leases", lease, out);
+
+ gf_msg_trace(frame->this->name, 0,
+ "Granting lease lock to client %s with lease id %s"
+ " on gfid(%s)",
+ client_uid, leaseid_utoa(lease->lease_id),
+ uuid_utoa(inode->gfid));
+
+ lease_entry = __get_or_new_lease_entry(frame, lease->lease_id, lease_ctx);
+ if (!lease_entry) {
+ errno = ENOMEM;
+ goto out;
+ }
+
+ lease_entry->lease_type_cnt[lease->lease_type]++;
+ lease_entry->lease_cnt++;
+ lease_entry->lease_type |= lease->lease_type;
+ /* If this is the first lease taken by the client on the file, then
+ * add this inode/file to the client disconnect cleanup list
+ */
+ if (lease_entry->lease_cnt == 1) {
+ add_inode_to_client_list(frame->this, inode, client_uid);
+ }
+
+ lease_ctx->lease_cnt++;
+ lease_ctx->lease_type_cnt[lease->lease_type]++;
+ lease_ctx->lease_type |= lease->lease_type;
+
+ /* Take a ref for the first lock taken on this inode. Corresponding
+ * unref when all the leases are unlocked or during DISCONNECT
+ * Ref is required because the inode on which lease is acquired should
+ * not be deleted when lru cleanup kicks in*/
+ if (lease_ctx->lease_cnt == 1) {
+ lease_ctx->inode = inode_ref(inode);
+ }
+
+ ret = 0;
out:
- return ret;
+ return ret;
}
-
static gf_boolean_t
-__is_clnt_lease_none (const char *client_uid, lease_inode_ctx_t *lease_ctx)
+__is_clnt_lease_none(const char *client_uid, lease_inode_ctx_t *lease_ctx)
{
- gf_boolean_t lease_none = _gf_true;
- lease_id_entry_t *lease_entry = NULL;
- lease_id_entry_t *tmp = NULL;
-
- list_for_each_entry_safe (lease_entry, tmp,
- &lease_ctx->lease_id_list,
- lease_id_list) {
- if ((strcmp (client_uid, lease_entry->client_uid) == 0)
- && (lease_entry->lease_cnt != 0)) {
- lease_none = _gf_false;
- break;
- }
+ gf_boolean_t lease_none = _gf_true;
+ lease_id_entry_t *lease_entry = NULL;
+ lease_id_entry_t *tmp = NULL;
+
+ list_for_each_entry_safe(lease_entry, tmp, &lease_ctx->lease_id_list,
+ lease_id_list)
+ {
+ if ((strcmp(client_uid, lease_entry->client_uid) == 0) &&
+ (lease_entry->lease_cnt != 0)) {
+ lease_none = _gf_false;
+ break;
}
+ }
- return lease_none;
+ return lease_none;
}
static int
-__remove_inode_from_clnt_list (xlator_t *this, lease_client_t *clnt,
- inode_t *inode)
+__remove_inode_from_clnt_list(xlator_t *this, lease_client_t *clnt,
+ inode_t *inode)
{
- int ret = -1;
- lease_inode_t *l_inode = NULL;
- lease_inode_t *tmp1 = NULL;
-
- list_for_each_entry_safe (l_inode, tmp1,
- &clnt->inode_list,
- list) {
- if (l_inode->inode == inode) {
- __destroy_lease_inode (l_inode);
- gf_msg_debug (this->name, 0,
- "Removed the inode from the client cleanup list");
- ret = 0;
- }
+ int ret = -1;
+ lease_inode_t *l_inode = NULL;
+ lease_inode_t *tmp1 = NULL;
+
+ list_for_each_entry_safe(l_inode, tmp1, &clnt->inode_list, list)
+ {
+ if (l_inode->inode == inode) {
+ __destroy_lease_inode(l_inode);
+ gf_msg_debug(this->name, 0,
+ "Removed the inode from the client cleanup list");
+ ret = 0;
}
- /* TODO: Remove the client entry from the cleanup list */
+ }
+ /* TODO: Remove the client entry from the cleanup list */
- return ret;
+ return ret;
}
-
static int
-remove_from_clnt_list (xlator_t *this, const char *client_uid, inode_t *inode)
+remove_from_clnt_list(xlator_t *this, const char *client_uid, inode_t *inode)
{
- leases_private_t *priv = NULL;
- int ret = -1;
- lease_client_t *clnt = NULL;
-
- priv = this->private;
- if (!priv)
- goto out;
-
- pthread_mutex_lock (&priv->mutex);
- {
- clnt = __get_lease_client (this, priv, client_uid);
- if (!clnt) {
- gf_msg (this->name, GF_LOG_ERROR, 0, LEASE_MSG_CLNT_NOTFOUND,
- "There is no client entry found in the cleanup list");
- pthread_mutex_unlock (&priv->mutex);
- goto out;
- }
- ret = __remove_inode_from_clnt_list (this, clnt, inode);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, LEASE_MSG_INODE_NOTFOUND,
- "There is no inode entry found in the cleanup list");
- }
+ leases_private_t *priv = NULL;
+ int ret = -1;
+ lease_client_t *clnt = NULL;
+
+ priv = this->private;
+ if (!priv)
+ goto out;
+
+ pthread_mutex_lock(&priv->mutex);
+ {
+ clnt = __get_lease_client(this, priv, client_uid);
+ if (!clnt) {
+ pthread_mutex_unlock(&priv->mutex);
+ gf_msg(this->name, GF_LOG_ERROR, 0, LEASE_MSG_CLNT_NOTFOUND,
+ "There is no client entry found in the cleanup list");
+ goto out;
+ }
+ ret = __remove_inode_from_clnt_list(this, clnt, inode);
+ if (ret) {
+ pthread_mutex_unlock(&priv->mutex);
+ gf_msg(this->name, GF_LOG_ERROR, 0, LEASE_MSG_INODE_NOTFOUND,
+ "There is no inode entry found in the cleanup list");
+ goto out;
}
- pthread_mutex_unlock (&priv->mutex);
+ }
+ pthread_mutex_unlock(&priv->mutex);
out:
- return ret;
+ return ret;
}
-
/* Remove lease entry in the corresponding client entry.
*/
static int
-__remove_lease (xlator_t *this, inode_t *inode, lease_inode_ctx_t *lease_ctx,
- const char *client_uid, struct gf_lease *lease)
+__remove_lease(xlator_t *this, inode_t *inode, lease_inode_ctx_t *lease_ctx,
+ const char *client_uid, struct gf_lease *lease)
{
- lease_id_entry_t *lease_entry = NULL;
- int ret = 0;
- int32_t lease_type = 0;
- leases_private_t *priv = NULL;
-
- GF_VALIDATE_OR_GOTO ("leases", lease_ctx, out);
- GF_VALIDATE_OR_GOTO ("leases", lease, out);
-
- priv = this->private;
-
- gf_msg_trace (this->name, 0, "Removing lease entry for client: %s, "
- "lease type:%d, lease id:%s", client_uid, lease->lease_type,
- leaseid_utoa (lease->lease_id));
-
- lease_entry = __get_lease_id_entry (lease_ctx, lease->lease_id);
- if (!lease_entry || !(lease_entry->lease_type & lease->lease_type)) {
- gf_msg (this->name, GF_LOG_INFO, 0, LEASE_MSG_INVAL_UNLK_LEASE,
- "Got unlock lease request from client:%s, but has no "
- "corresponding lock", client_uid);
- ret = -EINVAL;
- errno = EINVAL;
- goto out;
- }
-
- lease_type = lease->lease_type;
- lease_entry->lease_type_cnt[lease_type]--;
- lease_entry->lease_cnt--;
-
- lease_ctx->lease_type_cnt[lease_type]--;
- lease_ctx->lease_cnt--;
-
- if (lease_entry->lease_type_cnt[lease_type] == 0)
- lease_entry->lease_type = lease_entry->lease_type & (~lease_type);
-
- if (lease_ctx->lease_type_cnt[lease_type] == 0)
- lease_ctx->lease_type = lease_ctx->lease_type & (~lease_type);
-
- if (lease_entry->lease_cnt == 0) {
- if (__is_clnt_lease_none (client_uid, lease_ctx)) {
- gf_msg_trace (this->name, 0, "Client(%s) has no leases"
- " on gfid (%s), hence removing the inode"
- " from the client cleanup list",
- client_uid, uuid_utoa (inode->gfid));
- remove_from_clnt_list (this, client_uid, lease_ctx->inode);
- }
- __destroy_lease_id_entry (lease_entry);
+ lease_id_entry_t *lease_entry = NULL;
+ int ret = 0;
+ int32_t lease_type = 0;
+ leases_private_t *priv = NULL;
+
+ GF_VALIDATE_OR_GOTO("leases", lease_ctx, out);
+ GF_VALIDATE_OR_GOTO("leases", lease, out);
+
+ priv = this->private;
+
+ gf_msg_trace(this->name, 0,
+ "Removing lease entry for client: %s, "
+ "lease type:%d, lease id:%s",
+ client_uid, lease->lease_type, leaseid_utoa(lease->lease_id));
+
+ /* There could be a race where in server recalled the lease and by the time
+ * client sends lease_unlock request, server may have revoked it. To handle
+ * such cases, if lease doesnt exist treat it as noop and return success.
+ */
+ lease_entry = __get_lease_id_entry(lease_ctx, lease->lease_id);
+ if (!lease_entry) {
+ gf_msg(this->name, GF_LOG_INFO, 0, LEASE_MSG_INVAL_UNLK_LEASE,
+ "Got unlock lease request from client:%s, but has no "
+ "corresponding lock",
+ client_uid);
+ ret = 0;
+ goto out;
+ }
+
+ if (!(lease_entry->lease_type & lease->lease_type)) {
+ gf_msg(this->name, GF_LOG_INFO, 0, LEASE_MSG_INVAL_UNLK_LEASE,
+ "Got unlock lease request from client:%s for an invalid "
+ "lease_type",
+ client_uid);
+ ret = -EINVAL;
+ errno = EINVAL;
+ goto out;
+ }
+ lease_type = lease->lease_type;
+ lease_entry->lease_type_cnt[lease_type]--;
+ lease_entry->lease_cnt--;
+
+ lease_ctx->lease_type_cnt[lease_type]--;
+ lease_ctx->lease_cnt--;
+
+ if (lease_entry->lease_type_cnt[lease_type] == 0)
+ lease_entry->lease_type = lease_entry->lease_type & (~lease_type);
+
+ if (lease_ctx->lease_type_cnt[lease_type] == 0)
+ lease_ctx->lease_type = lease_ctx->lease_type & (~lease_type);
+
+ if (lease_entry->lease_cnt == 0) {
+ if (__is_clnt_lease_none(client_uid, lease_ctx)) {
+ gf_msg_trace(this->name, 0,
+ "Client(%s) has no leases"
+ " on gfid (%s), hence removing the inode"
+ " from the client cleanup list",
+ client_uid, uuid_utoa(inode->gfid));
+ remove_from_clnt_list(this, client_uid, lease_ctx->inode);
}
+ __destroy_lease_id_entry(lease_entry);
+ lease_ctx->blocked_fops_resuming = _gf_true;
+ }
- if (lease_ctx->lease_cnt == 0 && lease_ctx->timer) {
- ret = gf_tw_del_timer (priv->timer_wheel, lease_ctx->timer);
- lease_ctx->recall_in_progress = _gf_false;
- }
+ if (lease_ctx->lease_cnt == 0 && lease_ctx->timer) {
+ ret = gf_tw_del_timer(priv->timer_wheel, lease_ctx->timer);
+ lease_ctx->recall_in_progress = _gf_false;
+ lease_ctx->timer = NULL;
+ }
out:
- return ret;
+ return ret;
}
-
static gf_boolean_t
-__is_lease_grantable (xlator_t *this, lease_inode_ctx_t *lease_ctx,
- struct gf_lease *lease, inode_t *inode)
+__is_lease_grantable(xlator_t *this, lease_inode_ctx_t *lease_ctx,
+ struct gf_lease *lease, inode_t *inode)
{
- uint32_t fd_count = 0;
- int32_t flags = 0;
- fd_t *iter_fd = NULL;
- gf_boolean_t grant = _gf_false;
- int ret = 0;
- lease_fd_ctx_t *fd_ctx = NULL;
- uint64_t ctx = 0;
-
- GF_VALIDATE_OR_GOTO ("leases", lease_ctx, out);
- GF_VALIDATE_OR_GOTO ("leases", lease, out);
- GF_VALIDATE_OR_GOTO ("leases", inode, out);
-
- if (lease_ctx->recall_in_progress) {
- gf_msg_debug (this->name, 0, "Recall in progress, hence "
- "failing the lease request");
+ uint32_t fd_count = 0;
+ int32_t flags = 0;
+ fd_t *iter_fd = NULL;
+ gf_boolean_t grant = _gf_false;
+ int ret = 0;
+ lease_fd_ctx_t *fd_ctx = NULL;
+ uint64_t ctx = 0;
+
+ GF_VALIDATE_OR_GOTO("leases", lease_ctx, out);
+ GF_VALIDATE_OR_GOTO("leases", lease, out);
+ GF_VALIDATE_OR_GOTO("leases", inode, out);
+
+ if (lease_ctx->recall_in_progress) {
+ gf_msg_debug(this->name, 0,
+ "Recall in progress, hence "
+ "failing the lease request");
+ grant = _gf_false;
+ goto out;
+ }
+
+ if (lease_ctx->blocked_fops_resuming) {
+ gf_msg_debug(this->name, 0,
+ "Previously blocked fops resuming, hence "
+ "failing the lease request");
+ grant = _gf_false;
+ goto out;
+ }
+
+ LOCK(&inode->lock);
+ {
+ list_for_each_entry(iter_fd, &inode->fd_list, inode_list)
+ {
+ ret = fd_ctx_get(iter_fd, this, &ctx);
+ if (ret < 0) {
grant = _gf_false;
+ UNLOCK(&inode->lock);
+ gf_msg(this->name, GF_LOG_ERROR, 0, LEASE_MSG_INVAL_FD_CTX,
+ "Unable to get fd ctx");
goto out;
+ }
+ fd_ctx = (lease_fd_ctx_t *)(long)ctx;
+
+ /* Check for open fd conflict, note that open fds from
+ * the same lease id is not checked for conflict, as it is
+ * lease id based lease.
+ */
+ if (fd_ctx->client_uid != NULL &&
+ !__is_same_lease_id(fd_ctx->lease_id, lease->lease_id)) {
+ fd_count++;
+ flags |= iter_fd->flags;
+ }
}
+ }
+ UNLOCK(&inode->lock);
- LOCK (&inode->lock);
- {
- list_for_each_entry (iter_fd, &inode->fd_list, inode_list) {
- ret = fd_ctx_get (iter_fd, this, &ctx);
- if (ret < 0) {
- grant = _gf_false;
- UNLOCK (&inode->lock);
- gf_msg (this->name, GF_LOG_ERROR, 0,
- LEASE_MSG_INVAL_FD_CTX,
- "Unable to get fd ctx");
- goto out;
- }
- fd_ctx = (lease_fd_ctx_t *)(long) ctx;
-
- /* Check for open fd conflict, note that open fds from
- * the same lease id is not checked for conflict, as it is
- * lease id based lease.
- */
- if (fd_ctx->client_uid != NULL
- && !__is_same_lease_id (fd_ctx->lease_id, lease->lease_id)) {
- fd_count++;
- flags |= iter_fd->flags;
- }
- }
- }
- UNLOCK (&inode->lock);
-
- gf_msg_debug (this->name, 0, "open fd count:%d flags:%d",
- fd_count, flags);
+ gf_msg_debug(this->name, 0, "open fd count:%d flags:%d", fd_count, flags);
- __dump_leases_info (this, lease_ctx);
+ __dump_leases_info(this, lease_ctx);
- switch (lease->lease_type) {
+ switch (lease->lease_type) {
case GF_RD_LEASE:
- /* check open fd conflict */
- if ((fd_count > 0) && ((flags & O_WRONLY) || (flags & O_RDWR))) {
- grant = _gf_false;
- break;
- }
-
- /* check for conflict with existing leases */
- if (lease_ctx->lease_type == NONE ||
- lease_ctx->lease_type == GF_RD_LEASE ||
- !(__another_lease_found (lease_ctx, lease->lease_id)))
- grant = _gf_true;
- else
- grant = _gf_false;
+ /* check open fd conflict */
+ if ((fd_count > 0) && ((flags & O_WRONLY) || (flags & O_RDWR))) {
+ grant = _gf_false;
break;
+ }
+
+ /* check for conflict with existing leases */
+ if (lease_ctx->lease_type == NONE ||
+ lease_ctx->lease_type == GF_RD_LEASE ||
+ !(__another_lease_found(lease_ctx, lease->lease_id)))
+ grant = _gf_true;
+ else
+ grant = _gf_false;
+ break;
case GF_RW_LEASE:
- /* check open fd conflict; conflict if there are any fds open
- * other than the client on which the lease is requested. */
- if (fd_count > 0) {
- grant = _gf_false;
- break;
- }
-
- /* check existing lease conflict */
- if (lease_ctx->lease_type == NONE ||
- !(__another_lease_found (lease_ctx, lease->lease_id)))
- grant = _gf_true;
- else
- grant = _gf_false;
+ /* check open fd conflict; conflict if there are any fds open
+ * other than the client on which the lease is requested. */
+ if (fd_count > 0) {
+ grant = _gf_false;
break;
+ }
+
+ /* check existing lease conflict */
+ if (lease_ctx->lease_type == NONE ||
+ !(__another_lease_found(lease_ctx, lease->lease_id)))
+ grant = _gf_true;
+ else
+ grant = _gf_false;
+ break;
default:
- gf_msg (this->name, GF_LOG_ERROR, EINVAL, LEASE_MSG_INVAL_LEASE_TYPE,
- "Invalid lease type specified");
- break;
- }
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL, LEASE_MSG_INVAL_LEASE_TYPE,
+ "Invalid lease type specified");
+ break;
+ }
out:
- return grant;
+ return grant;
}
-
static void
-do_blocked_fops (xlator_t *this, lease_inode_ctx_t *lease_ctx)
+do_blocked_fops(xlator_t *this, lease_inode_ctx_t *lease_ctx)
{
- struct list_head wind_list;
- fop_stub_t *blk_fop = NULL;
- fop_stub_t *tmp = NULL;
-
- INIT_LIST_HEAD (&wind_list);
-
- pthread_mutex_lock (&lease_ctx->lock);
- {
- list_for_each_entry_safe (blk_fop, tmp,
- &lease_ctx->blocked_list, list) {
- list_del_init (&blk_fop->list);
- list_add_tail (&blk_fop->list, &wind_list);
- }
- }
- pthread_mutex_unlock (&lease_ctx->lock);
-
- gf_msg_trace (this->name, 0, "Executing the blocked stubs on gfid(%s)",
- uuid_utoa (lease_ctx->inode->gfid));
- list_for_each_entry_safe (blk_fop, tmp, &wind_list, list) {
- list_del_init (&blk_fop->list);
- gf_msg_trace (this->name, 0, "Executing fop:%d", blk_fop->stub->fop);
- call_resume (blk_fop->stub);
- GF_FREE (blk_fop);
+ struct list_head wind_list;
+ fop_stub_t *blk_fop = NULL;
+ fop_stub_t *tmp = NULL;
+
+ INIT_LIST_HEAD(&wind_list);
+
+ pthread_mutex_lock(&lease_ctx->lock);
+ {
+ if (!lease_ctx->blocked_fops_resuming) {
+ /* lease_ctx->blocked_fops_resuming will be set
+ * only when the last lease is released. That
+ * is when we need to resume blocked fops and unref
+ * the inode taken in __add_lease (when lease_cnt == 1).
+ * Return otherwise.
+ */
+ pthread_mutex_unlock(&lease_ctx->lock);
+ return;
}
- pthread_mutex_lock (&lease_ctx->lock);
+ list_for_each_entry_safe(blk_fop, tmp, &lease_ctx->blocked_list, list)
{
- lease_ctx->lease_type = NONE;
- inode_unref (lease_ctx->inode);
- lease_ctx->inode = NULL;
+ list_del_init(&blk_fop->list);
+ list_add_tail(&blk_fop->list, &wind_list);
}
- pthread_mutex_unlock (&lease_ctx->lock);
-
- return;
+ }
+ pthread_mutex_unlock(&lease_ctx->lock);
+
+ gf_msg_trace(this->name, 0, "Executing the blocked stubs on gfid(%s)",
+ uuid_utoa(lease_ctx->inode->gfid));
+ list_for_each_entry_safe(blk_fop, tmp, &wind_list, list)
+ {
+ list_del_init(&blk_fop->list);
+ gf_msg_trace(this->name, 0, "Executing fop:%d", blk_fop->stub->fop);
+ call_resume(blk_fop->stub);
+ GF_FREE(blk_fop);
+ }
+
+ pthread_mutex_lock(&lease_ctx->lock);
+ {
+ lease_ctx->lease_type = NONE;
+ /* unref the inode taken in __add_lease
+ * (when lease_cnt == 1) */
+ lease_ctx->blocked_fops_resuming = _gf_false;
+ inode_unref(lease_ctx->inode);
+ lease_ctx->inode = NULL;
+ }
+ pthread_mutex_unlock(&lease_ctx->lock);
+
+ return;
}
-
void
-recall_lease_timer_handler (struct gf_tw_timer_list *timer,
- void *data, unsigned long calltime)
+recall_lease_timer_handler(struct gf_tw_timer_list *timer, void *data,
+ unsigned long calltime)
{
- inode_t *inode = NULL;
- lease_inode_t *lease_inode = NULL;
- leases_private_t *priv = NULL;
- lease_timer_data_t *timer_data = NULL;
-
- timer_data = data;
-
- priv = timer_data->this->private;
- inode = timer_data->inode;
- pthread_mutex_lock (&priv->mutex);
- {
- lease_inode = new_lease_inode (inode);
- if (!lease_inode) {
- errno = ENOMEM;
- goto out;
- }
- list_add_tail (&lease_inode->list, &priv->recall_list);
- pthread_cond_broadcast (&priv->cond);
- }
+ inode_t *inode = NULL;
+ lease_inode_t *lease_inode = NULL;
+ leases_private_t *priv = NULL;
+ lease_timer_data_t *timer_data = NULL;
+
+ timer_data = data;
+
+ priv = timer_data->this->private;
+ inode = timer_data->inode;
+ lease_inode = new_lease_inode(inode);
+ if (!lease_inode) {
+ errno = ENOMEM;
+ goto out;
+ }
+ pthread_mutex_lock(&priv->mutex);
+ {
+ list_add_tail(&lease_inode->list, &priv->recall_list);
+ pthread_cond_broadcast(&priv->cond);
+ }
+ pthread_mutex_unlock(&priv->mutex);
out:
- pthread_mutex_unlock (&priv->mutex);
+ /* unref the inode_ref taken by timer_data in __recall_lease */
+ inode_unref(timer_data->inode);
- GF_FREE (timer);
+ GF_FREE(timer);
}
-
static void
-__recall_lease (xlator_t *this, lease_inode_ctx_t *lease_ctx)
+__recall_lease(xlator_t *this, lease_inode_ctx_t *lease_ctx)
{
- lease_id_entry_t *lease_entry = NULL;
- lease_id_entry_t *tmp = NULL;
- struct gf_upcall up_req = {0,};
- struct gf_upcall_recall_lease recall_req = {0,};
- int notify_ret = -1;
- struct gf_tw_timer_list *timer = NULL;
- leases_private_t *priv = NULL;
- lease_timer_data_t *timer_data = NULL;
-
- if (lease_ctx->recall_in_progress) {
- gf_msg_debug (this->name, 0, "Lease recall is already in "
- "progress, hence not sending another recall");
- goto out;
- }
-
- priv = this->private;
- list_for_each_entry_safe (lease_entry, tmp,
- &lease_ctx->lease_id_list,
- lease_id_list) {
- gf_uuid_copy (up_req.gfid, lease_ctx->inode->gfid);
- up_req.client_uid = lease_entry->client_uid;
- up_req.event_type = GF_UPCALL_RECALL_LEASE;
- up_req.data = &recall_req;
-
- notify_ret = this->notify (this, GF_EVENT_UPCALL, &up_req);
- if (notify_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0, LEASE_MSG_RECALL_FAIL,
- "Recall notification to client: %s failed",
- lease_entry->client_uid);
- /* Do not return from here, continue registering the timer,
- this is required mostly o keep replicas in sync*/
- } else {
- gf_msg_debug (this->name, 0, "Recall lease (all)"
- "notification sent to client %s",
- lease_entry->client_uid);
- }
-
- lease_ctx->recall_in_progress = _gf_true;
- lease_entry->recall_time = time (NULL);
- }
- timer = GF_CALLOC (1, sizeof (*timer),
- gf_common_mt_tw_timer_list);
- if (!timer) {
- goto out;
- }
- timer_data = GF_CALLOC (1, sizeof (*timer_data),
- gf_leases_mt_timer_data_t);
- if (!timer_data) {
- GF_FREE (timer);
- goto out;
+ lease_id_entry_t *lease_entry = NULL;
+ lease_id_entry_t *tmp = NULL;
+ struct gf_upcall up_req = {
+ 0,
+ };
+ struct gf_upcall_recall_lease recall_req = {
+ 0,
+ };
+ int notify_ret = -1;
+ struct gf_tw_timer_list *timer = NULL;
+ leases_private_t *priv = NULL;
+ lease_timer_data_t *timer_data = NULL;
+ time_t recall_time;
+
+ if (lease_ctx->recall_in_progress) {
+ gf_msg_debug(this->name, 0,
+ "Lease recall is already in "
+ "progress, hence not sending another recall");
+ goto out;
+ }
+
+ priv = this->private;
+ recall_time = gf_time();
+ list_for_each_entry_safe(lease_entry, tmp, &lease_ctx->lease_id_list,
+ lease_id_list)
+ {
+ gf_uuid_copy(up_req.gfid, lease_ctx->inode->gfid);
+ up_req.client_uid = lease_entry->client_uid;
+ up_req.event_type = GF_UPCALL_RECALL_LEASE;
+ up_req.data = &recall_req;
+
+ notify_ret = this->notify(this, GF_EVENT_UPCALL, &up_req);
+ if (notify_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, LEASE_MSG_RECALL_FAIL,
+ "Recall notification to client: %s failed",
+ lease_entry->client_uid);
+ /* Do not return from here, continue registering the timer,
+ this is required mostly o keep replicas in sync*/
+ } else {
+ gf_msg_debug(this->name, 0,
+ "Recall lease (all)"
+ "notification sent to client %s",
+ lease_entry->client_uid);
}
- timer_data->inode = inode_ref (lease_ctx->inode);
- timer_data->this = this;
- timer->data = timer_data;
-
- INIT_LIST_HEAD (&timer->entry);
- timer->expires = get_recall_lease_timeout (this);
- timer->function = recall_lease_timer_handler;
- lease_ctx->timer = timer;
- gf_tw_add_timer (priv->timer_wheel, timer);
- gf_msg_trace (this->name, 0, "Registering timer " "%p, after "
- "sending recall", timer);
+ lease_ctx->recall_in_progress = _gf_true;
+ lease_entry->recall_time = recall_time;
+ }
+ timer = GF_MALLOC(sizeof(*timer), gf_common_mt_tw_timer_list);
+ if (!timer) {
+ goto out;
+ }
+ timer_data = GF_MALLOC(sizeof(lease_timer_data_t),
+ gf_leases_mt_timer_data_t);
+ if (!timer_data) {
+ GF_FREE(timer);
+ goto out;
+ }
+
+ timer_data->inode = inode_ref(lease_ctx->inode);
+ timer_data->this = this;
+ timer->data = timer_data;
+
+ INIT_LIST_HEAD(&timer->entry);
+ timer->expires = get_recall_lease_timeout(this);
+ timer->function = recall_lease_timer_handler;
+ lease_ctx->timer = timer;
+ gf_tw_add_timer(priv->timer_wheel, timer);
+ gf_msg_trace(this->name, 0,
+ "Registering timer "
+ "%p, after "
+ "sending recall",
+ timer);
out:
- return;
+ return;
}
-
/* ret = 0; STACK_UNWIND Success
* ret = -1; STACK_UNWIND failure
*/
int
-process_lease_req (call_frame_t *frame, xlator_t *this,
- inode_t *inode, struct gf_lease *lease)
+process_lease_req(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ struct gf_lease *lease)
{
- int ret = 0;
- char *client_uid = NULL;
- lease_inode_ctx_t *lease_ctx = NULL;
-
- GF_VALIDATE_OR_GOTO ("leases", frame, out);
- GF_VALIDATE_OR_GOTO ("leases", this, out);
- GF_VALIDATE_OR_GOTO ("leases", inode, out);
- GF_VALIDATE_OR_GOTO ("leases", lease, out);
-
- client_uid = frame->root->client->client_uid;
-
- if (!is_valid_lease_id (lease->lease_id)) {
- gf_msg (this->name, GF_LOG_ERROR, EINVAL,
- LEASE_MSG_INVAL_LEASE_ID, "Invalid lease id, from"
- "client:%s", client_uid);
- ret = -EINVAL;
- errno = EINVAL;
- goto out;
- }
-
- lease_ctx = lease_ctx_get (inode, this);
- if (!lease_ctx) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- LEASE_MSG_NO_MEM, "Unable to create/get inode ctx, "
- "inode:%p", inode);
- ret = -ENOMEM;
- errno = ENOMEM;
- goto out;
- }
-
- gf_msg_debug (this->name, 0, "Lease request from client: %s, "
- "lease type:%d, lease cmd:%d, lease ID:%s, gfid:%s",
- client_uid, lease->lease_type, lease->cmd,
- leaseid_utoa (lease->lease_id), uuid_utoa (inode->gfid));
+ int ret = 0;
+ char *client_uid = NULL;
+ lease_inode_ctx_t *lease_ctx = NULL;
+
+ GF_VALIDATE_OR_GOTO("leases", frame, out);
+ GF_VALIDATE_OR_GOTO("leases", this, out);
+ GF_VALIDATE_OR_GOTO("leases", inode, out);
+ GF_VALIDATE_OR_GOTO("leases", lease, out);
+
+ client_uid = frame->root->client->client_uid;
+
+ if (!is_valid_lease_id(lease->lease_id)) {
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL, LEASE_MSG_INVAL_LEASE_ID,
+ "Invalid lease id, from"
+ "client:%s",
+ client_uid);
+ ret = -EINVAL;
+ errno = EINVAL;
+ goto out;
+ }
+
+ lease_ctx = lease_ctx_get(inode, this);
+ if (!lease_ctx) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, LEASE_MSG_NO_MEM,
+ "Unable to create/get inode ctx, "
+ "inode:%p",
+ inode);
+ ret = -ENOMEM;
+ errno = ENOMEM;
+ goto out;
+ }
+
+ gf_msg_debug(this->name, 0,
+ "Lease request from client: %s, "
+ "lease type:%d, lease cmd:%d, lease ID:%s, gfid:%s",
+ client_uid, lease->lease_type, lease->cmd,
+ leaseid_utoa(lease->lease_id), uuid_utoa(inode->gfid));
+
+ pthread_mutex_lock(&lease_ctx->lock);
+ {
+ switch (lease->cmd) {
+ case GF_GET_LEASE:
+ lease->lease_type = lease_ctx->lease_type;
+ gf_msg_debug(this->name, 0,
+ "Get lease, existing lease"
+ "type: %d",
+ lease_ctx->lease_type);
+ /*TODO:Should it consider lease id or client_uid?*/
+ break;
- pthread_mutex_lock (&lease_ctx->lock);
- {
- switch (lease->cmd) {
- case GF_GET_LEASE:
- lease->lease_type = lease_ctx->lease_type;
- gf_msg_debug (this->name, 0, "Get lease, existing lease"
- "type: %d", lease_ctx->lease_type);
- /*TODO:Should it consider lease id or client_uid?*/
- break;
-
- case GF_SET_LEASE:
- if (__is_lease_grantable (this, lease_ctx, lease, inode)) {
- __add_lease (frame, inode, lease_ctx,
- client_uid, lease);
- ret = 0;
- } else {
- gf_msg_debug (this->name, GF_LOG_DEBUG,
- "Not granting the conflicting lease"
- " request from %s on gfid(%s)",
- client_uid, uuid_utoa (inode->gfid));
- __recall_lease (this, lease_ctx);
- ret = -1;
- }
- break;
- case GF_UNLK_LEASE:
- ret = __remove_lease (this, inode, lease_ctx,
- client_uid, lease);
- if ((ret == 0) && (lease_ctx->lease_cnt == 0)) {
- pthread_mutex_unlock (&lease_ctx->lock);
- goto unblock;
- }
- break;
- default:
- ret = -EINVAL;
- break;
+ case GF_SET_LEASE:
+ if (__is_lease_grantable(this, lease_ctx, lease, inode)) {
+ __add_lease(frame, inode, lease_ctx, client_uid, lease);
+ ret = 0;
+ } else {
+ gf_msg_debug(this->name, GF_LOG_DEBUG,
+ "Not granting the conflicting lease"
+ " request from %s on gfid(%s)",
+ client_uid, uuid_utoa(inode->gfid));
+ __recall_lease(this, lease_ctx);
+ ret = -1;
+ }
+ break;
+ case GF_UNLK_LEASE:
+ ret = __remove_lease(this, inode, lease_ctx, client_uid, lease);
+ if ((ret >= 0) && (lease_ctx->lease_cnt == 0)) {
+ pthread_mutex_unlock(&lease_ctx->lock);
+ goto unblock;
}
+ break;
+ default:
+ ret = -EINVAL;
+ break;
}
- pthread_mutex_unlock (&lease_ctx->lock);
+ }
+ pthread_mutex_unlock(&lease_ctx->lock);
- return ret;
+ return ret;
unblock:
- do_blocked_fops (this, lease_ctx);
+ do_blocked_fops(this, lease_ctx);
out:
- return ret;
+ return ret;
}
-
/* ret = 1 conflict
* ret = 0 no conflict
*/
gf_boolean_t
-__check_lease_conflict (call_frame_t *frame, lease_inode_ctx_t *lease_ctx,
- const char *lease_id, gf_boolean_t is_write)
+__check_lease_conflict(call_frame_t *frame, lease_inode_ctx_t *lease_ctx,
+ const char *lease_id, gf_boolean_t is_write)
{
- gf_lease_types_t lease_type = {0,};
- gf_boolean_t conflicts = _gf_false;
- lease_id_entry_t *lease_entry = NULL;
-
- GF_VALIDATE_OR_GOTO ("leases", frame, out);
- GF_VALIDATE_OR_GOTO ("leases", lease_ctx, out);
- GF_VALIDATE_OR_GOTO ("leases", lease_id, out);
-
- lease_type = lease_ctx->lease_type;
-
- /* If the fop is rename or unlink conflict the lease even if its
- * from the same client??
- */
- if ((frame->root->op == GF_FOP_RENAME) ||
- (frame->root->op == GF_FOP_UNLINK)) {
- conflicts = _gf_true;
- goto recall;
- }
-
- /* TODO: If lease_id is not sent, fall back to client uid conflict check?
- * Or set conflicts = true if lease_id is 0 when there is an existing
- * lease */
- switch (lease_type) {
+ gf_lease_types_t lease_type = {
+ 0,
+ };
+ gf_boolean_t conflicts = _gf_false;
+ lease_id_entry_t *lease_entry = NULL;
+
+ GF_VALIDATE_OR_GOTO("leases", frame, out);
+ GF_VALIDATE_OR_GOTO("leases", lease_ctx, out);
+
+ lease_type = lease_ctx->lease_type;
+
+ /* If the fop is rename or unlink conflict the lease even if its
+ * from the same client??
+ */
+ if ((frame->root->op == GF_FOP_RENAME) ||
+ (frame->root->op == GF_FOP_UNLINK)) {
+ conflicts = _gf_true;
+ goto recall;
+ }
+
+ /* As internal fops are used to maintain data integrity but do not
+ * make modififications to the client data, no need to conflict with
+ * them.
+ *
+ * @todo: like for locks, even lease state has to be handled by
+ * rebalance or self-heal daemon process. */
+ if (frame->root->pid < 0) {
+ conflicts = _gf_false;
+ goto recall;
+ }
+
+ /* If lease_id is not sent, set conflicts = true if there is
+ * an existing lease */
+ if (!lease_id && (lease_ctx->lease_cnt > 0)) {
+ conflicts = _gf_true;
+ goto recall;
+ }
+
+ switch (lease_type) {
case (GF_RW_LEASE | GF_RD_LEASE):
case GF_RW_LEASE:
- lease_entry = __get_lease_id_entry (lease_ctx, lease_id);
- if (lease_entry && (lease_entry->lease_type & GF_RW_LEASE))
- conflicts = _gf_false;
- else
- conflicts = _gf_true;
- break;
+ lease_entry = __get_lease_id_entry(lease_ctx, lease_id);
+ if (lease_entry && (lease_entry->lease_type & GF_RW_LEASE))
+ conflicts = _gf_false;
+ else
+ conflicts = _gf_true;
+ break;
case GF_RD_LEASE:
- if (is_write && __another_lease_found(lease_ctx, lease_id))
- conflicts = _gf_true;
- else
- conflicts = _gf_false;
- break;
+ if (is_write && __another_lease_found(lease_ctx, lease_id))
+ conflicts = _gf_true;
+ else
+ conflicts = _gf_false;
+ break;
default:
- break;
- }
+ break;
+ }
recall:
- /* If there is a conflict found and recall is not already sent to all
- * the clients, then send recall to each of the client holding lease.
- */
- if (conflicts)
- __recall_lease (frame->this, lease_ctx);
+ /* If there is a conflict found and recall is not already sent to all
+ * the clients, then send recall to each of the client holding lease.
+ */
+ if (conflicts)
+ __recall_lease(frame->this, lease_ctx);
out:
- return conflicts;
+ return conflicts;
}
-
/* Return values:
* -1 : error, unwind the fop
* WIND_FOP: No conflict, wind the fop
* BLOCK_FOP: Found a conflicting lease, block the fop
*/
int
-check_lease_conflict (call_frame_t *frame, inode_t *inode,
- const char *lease_id, uint32_t fop_flags)
+check_lease_conflict(call_frame_t *frame, inode_t *inode, const char *lease_id,
+ uint32_t fop_flags)
{
- lease_inode_ctx_t *lease_ctx = NULL;
- gf_boolean_t is_blocking_fop = _gf_false;
- gf_boolean_t is_write_fop = _gf_false;
- gf_boolean_t conflicts = _gf_false;
- int ret = WIND_FOP;
-
- lease_ctx = lease_ctx_get (inode, frame->this);
- if (!lease_ctx) {
- gf_msg (frame->this->name, GF_LOG_WARNING, ENOMEM,
- LEASE_MSG_NO_MEM,
- "Unable to create/get inode ctx");
- ret = -1;
- errno = ENOMEM;
- goto out;
+ lease_inode_ctx_t *lease_ctx = NULL;
+ gf_boolean_t is_blocking_fop = _gf_false;
+ gf_boolean_t is_write_fop = _gf_false;
+ gf_boolean_t conflicts = _gf_false;
+ int ret = WIND_FOP;
+
+ lease_ctx = lease_ctx_get(inode, frame->this);
+ if (!lease_ctx) {
+ gf_msg(frame->this->name, GF_LOG_WARNING, ENOMEM, LEASE_MSG_NO_MEM,
+ "Unable to create/get inode ctx");
+ ret = -1;
+ errno = ENOMEM;
+ goto out;
+ }
+
+ is_blocking_fop = ((fop_flags & BLOCKING_FOP) != 0);
+ is_write_fop = ((fop_flags & DATA_MODIFY_FOP) != 0);
+
+ pthread_mutex_lock(&lease_ctx->lock);
+ {
+ if (lease_ctx->lease_type == NONE) {
+ pthread_mutex_unlock(&lease_ctx->lock);
+ gf_msg_debug(frame->this->name, 0,
+ "No leases found continuing with the"
+ " fop:%s",
+ gf_fop_list[frame->root->op]);
+ ret = WIND_FOP;
+ goto out;
}
-
- is_blocking_fop = ((fop_flags & BLOCKING_FOP) != 0);
- is_write_fop = ((fop_flags & DATA_MODIFY_FOP) != 0);
-
- pthread_mutex_lock (&lease_ctx->lock);
- {
- if (lease_ctx->lease_type == NONE) {
- gf_msg_debug (frame->this->name, 0,
- "No leases found continuing with the"
- " fop:%s", gf_fop_list[frame->root->op]);
- ret = WIND_FOP;
- goto unlock;
- }
- conflicts = __check_lease_conflict (frame, lease_ctx,
- lease_id, is_write_fop);
- if (conflicts) {
- if (is_blocking_fop) {
- gf_msg_debug (frame->this->name, 0, "Fop: %s "
- "conflicting existing "
- "lease: %d, blocking the"
- "fop", gf_fop_list[frame->root->op],
- lease_ctx->lease_type);
- ret = BLOCK_FOP;
- } else {
- gf_msg_debug (frame->this->name, 0, "Fop: %s "
- "conflicting existing "
- "lease: %d, sending "
- "EAGAIN",
- gf_fop_list[frame->root->op],
- lease_ctx->lease_type);
- errno = EAGAIN;
- ret = -1;
- }
- }
+ conflicts = __check_lease_conflict(frame, lease_ctx, lease_id,
+ is_write_fop);
+ if (conflicts) {
+ if (is_blocking_fop) {
+ gf_msg_debug(frame->this->name, 0,
+ "Fop: %s "
+ "conflicting existing "
+ "lease: %d, blocking the"
+ "fop",
+ gf_fop_list[frame->root->op],
+ lease_ctx->lease_type);
+ ret = BLOCK_FOP;
+ } else {
+ gf_msg_debug(frame->this->name, 0,
+ "Fop: %s "
+ "conflicting existing "
+ "lease: %d, sending "
+ "EAGAIN",
+ gf_fop_list[frame->root->op],
+ lease_ctx->lease_type);
+ errno = EAGAIN;
+ ret = -1;
+ }
}
-unlock:
- pthread_mutex_unlock (&lease_ctx->lock);
+ }
+ pthread_mutex_unlock(&lease_ctx->lock);
out:
- return ret;
+ return ret;
}
-
static int
-remove_clnt_leases (const char *client_uid, inode_t *inode, xlator_t *this)
+remove_clnt_leases(const char *client_uid, inode_t *inode, xlator_t *this)
{
- lease_inode_ctx_t *lease_ctx = NULL;
- lease_id_entry_t *lease_entry = NULL;
- lease_id_entry_t *tmp = NULL;
- int ret = 0;
- int i = 0;
-
- lease_ctx = lease_ctx_get (inode, this);
- if (!lease_ctx) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- LEASE_MSG_INVAL_INODE_CTX,
- "Unable to create/get inode ctx");
- ret = -1;
- errno = ENOMEM;
- goto out;
- }
-
- pthread_mutex_lock (&lease_ctx->lock);
+ lease_inode_ctx_t *lease_ctx = NULL;
+ lease_id_entry_t *lease_entry = NULL;
+ lease_id_entry_t *tmp = NULL;
+ int ret = 0;
+ int i = 0;
+
+ lease_ctx = lease_ctx_get(inode, this);
+ if (!lease_ctx) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, LEASE_MSG_INVAL_INODE_CTX,
+ "Unable to create/get inode ctx");
+ ret = -1;
+ errno = ENOMEM;
+ goto out;
+ }
+
+ pthread_mutex_lock(&lease_ctx->lock);
+ {
+ list_for_each_entry_safe(lease_entry, tmp, &lease_ctx->lease_id_list,
+ lease_id_list)
{
- list_for_each_entry_safe (lease_entry, tmp,
- &lease_ctx->lease_id_list,
- lease_id_list) {
- if (strcmp (client_uid, lease_entry->client_uid) == 0) {
- for (i = 0; i < GF_LEASE_MAX_TYPE; i++) {
- lease_ctx->lease_type_cnt[i] -= lease_entry->lease_type_cnt[i];
- }
- lease_ctx->lease_cnt -= lease_entry->lease_cnt;
- __destroy_lease_id_entry (lease_entry);
- if (lease_ctx->lease_cnt == 0) {
- pthread_mutex_unlock (&lease_ctx->lock);
- goto unblock;
- }
- }
+ if (strcmp(client_uid, lease_entry->client_uid) == 0) {
+ for (i = 0; i < GF_LEASE_MAX_TYPE; i++) {
+ lease_ctx->lease_type_cnt[i] -= lease_entry
+ ->lease_type_cnt[i];
+ }
+ lease_ctx->lease_cnt -= lease_entry->lease_cnt;
+ __destroy_lease_id_entry(lease_entry);
+ if (lease_ctx->lease_cnt == 0) {
+ lease_ctx->blocked_fops_resuming = _gf_true;
+ pthread_mutex_unlock(&lease_ctx->lock);
+ goto unblock;
}
+ }
}
- pthread_mutex_unlock (&lease_ctx->lock);
+ }
+ pthread_mutex_unlock(&lease_ctx->lock);
out:
- return ret;
+ return ret;
unblock:
- do_blocked_fops (this, lease_ctx);
- return ret;
+ do_blocked_fops(this, lease_ctx);
+ return ret;
}
-
int
-cleanup_client_leases (xlator_t *this, const char *client_uid)
+cleanup_client_leases(xlator_t *this, const char *client_uid)
{
- lease_client_t *clnt = NULL;
- lease_client_t *tmp = NULL;
- struct list_head cleanup_list = {0, };
- lease_inode_t *l_inode = NULL;
- lease_inode_t *tmp1 = NULL;
- leases_private_t *priv = NULL;
- int ret = 0;
-
- priv = this->private;
- if (!priv) {
- ret = -1;
- errno = EINVAL;
- goto out;
- }
-
- INIT_LIST_HEAD (&cleanup_list);
- pthread_mutex_lock (&priv->mutex);
+ lease_client_t *clnt = NULL;
+ lease_client_t *tmp = NULL;
+ struct list_head cleanup_list = {
+ 0,
+ };
+ lease_inode_t *l_inode = NULL;
+ lease_inode_t *tmp1 = NULL;
+ leases_private_t *priv = NULL;
+ int ret = 0;
+
+ priv = this->private;
+ if (!priv) {
+ ret = -1;
+ errno = EINVAL;
+ goto out;
+ }
+
+ INIT_LIST_HEAD(&cleanup_list);
+ pthread_mutex_lock(&priv->mutex);
+ {
+ list_for_each_entry_safe(clnt, tmp, &priv->client_list, client_list)
{
- list_for_each_entry_safe (clnt, tmp, &priv->client_list, client_list) {
- if ((strcmp (clnt->client_uid, client_uid) == 0)) {
- list_for_each_entry_safe (l_inode, tmp1,
- &clnt->inode_list, list) {
- list_del_init (&l_inode->list);
- list_add_tail (&l_inode->list, &cleanup_list);
- }
- break;
- }
- __destroy_lease_client (clnt);
+ if ((strcmp(clnt->client_uid, client_uid) == 0)) {
+ list_for_each_entry_safe(l_inode, tmp1, &clnt->inode_list, list)
+ {
+ list_del_init(&l_inode->list);
+ list_add_tail(&l_inode->list, &cleanup_list);
}
+ __destroy_lease_client(clnt);
+ break;
+ }
}
- pthread_mutex_unlock (&priv->mutex);
-
- l_inode = tmp1 = NULL;
- list_for_each_entry_safe (l_inode, tmp1, &cleanup_list, list) {
- remove_clnt_leases (client_uid, l_inode->inode, this);
- }
+ }
+ pthread_mutex_unlock(&priv->mutex);
+
+ l_inode = tmp1 = NULL;
+ list_for_each_entry_safe(l_inode, tmp1, &cleanup_list, list)
+ {
+ remove_clnt_leases(client_uid, l_inode->inode, this);
+ __destroy_lease_inode(l_inode);
+ }
out:
- return ret;
+ return ret;
}
-
static void
-__remove_all_leases (xlator_t *this, lease_inode_ctx_t *lease_ctx)
+__remove_all_leases(xlator_t *this, lease_inode_ctx_t *lease_ctx)
{
- int i = 0;
- lease_id_entry_t *lease_entry = NULL;
- lease_id_entry_t *tmp = NULL;
-
- __dump_leases_info (this, lease_ctx);
-
- list_for_each_entry_safe (lease_entry, tmp,
- &lease_ctx->lease_id_list,
- lease_id_list) {
- lease_entry->lease_cnt = 0;
- remove_from_clnt_list (this, lease_entry->client_uid, lease_ctx->inode);
- __destroy_lease_id_entry (lease_entry);
- }
- INIT_LIST_HEAD (&lease_ctx->lease_id_list);
- for (i = 0; i <= GF_LEASE_MAX_TYPE; i++)
- lease_ctx->lease_type_cnt[i] = 0;
- lease_ctx->lease_type = 0;
- lease_ctx->lease_cnt = 0;
- lease_ctx->recall_in_progress = _gf_false;
- inode_unref (lease_ctx->inode);
- lease_ctx->timer = NULL;
+ int i = 0;
+ lease_id_entry_t *lease_entry = NULL;
+ lease_id_entry_t *tmp = NULL;
- /* TODO:
- * - Mark the corresponding fd bad. Could be done on client side
- * as a result of recall
- * - Free the lease_ctx
- */
+ if (lease_ctx->lease_cnt == 0) {
+ /* No leases to remove. Return */
return;
-}
+ }
+ __dump_leases_info(this, lease_ctx);
+ list_for_each_entry_safe(lease_entry, tmp, &lease_ctx->lease_id_list,
+ lease_id_list)
+ {
+ lease_entry->lease_cnt = 0;
+ remove_from_clnt_list(this, lease_entry->client_uid, lease_ctx->inode);
+ __destroy_lease_id_entry(lease_entry);
+ }
+ INIT_LIST_HEAD(&lease_ctx->lease_id_list);
+ for (i = 0; i <= GF_LEASE_MAX_TYPE; i++)
+ lease_ctx->lease_type_cnt[i] = 0;
+ lease_ctx->lease_type = 0;
+ lease_ctx->lease_cnt = 0;
+ lease_ctx->recall_in_progress = _gf_false;
+ lease_ctx->timer = NULL;
+ lease_ctx->blocked_fops_resuming = _gf_true;
+
+ /* TODO:
+ * - Mark the corresponding fd bad. Could be done on client side
+ * as a result of recall
+ * - Free the lease_ctx
+ */
+ return;
+}
static int
-remove_all_leases (xlator_t *this, inode_t *inode)
+remove_all_leases(xlator_t *this, inode_t *inode)
{
- lease_inode_ctx_t *lease_ctx = NULL;
- int ret = 0;
-
- GF_VALIDATE_OR_GOTO ("leases", inode, out);
-
- lease_ctx = lease_ctx_get (inode, this);
- if (!lease_ctx) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- LEASE_MSG_INVAL_INODE_CTX,
- "Unable to create/get inode ctx");
- ret = -1;
- errno = ENOMEM;
- goto out;
- }
-
- pthread_mutex_lock (&lease_ctx->lock);
- {
- __remove_all_leases (this, lease_ctx);
- }
- pthread_mutex_unlock (&lease_ctx->lock);
-
- do_blocked_fops (this, lease_ctx);
+ lease_inode_ctx_t *lease_ctx = NULL;
+ int ret = 0;
+
+ GF_VALIDATE_OR_GOTO("leases", inode, out);
+
+ lease_ctx = lease_ctx_get(inode, this);
+ if (!lease_ctx) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, LEASE_MSG_INVAL_INODE_CTX,
+ "Unable to create/get inode ctx");
+ ret = -1;
+ errno = ENOMEM;
+ goto out;
+ }
+
+ pthread_mutex_lock(&lease_ctx->lock);
+ {
+ __remove_all_leases(this, lease_ctx);
+ }
+ pthread_mutex_unlock(&lease_ctx->lock);
+
+ do_blocked_fops(this, lease_ctx);
out:
- return ret;
+ return ret;
}
-
void *
-expired_recall_cleanup (void *data)
+expired_recall_cleanup(void *data)
{
- struct timespec sleep_till = {0, };
- struct list_head recall_cleanup_list;
- lease_inode_t *recall_entry = NULL;
- lease_inode_t *tmp = NULL;
- leases_private_t *priv = NULL;
- xlator_t *this = NULL;
-
- GF_VALIDATE_OR_GOTO ("leases", data, out);
-
- this = data;
- priv = this->private;
-
- gf_msg_debug (this->name, 0, "Started the expired_recall_cleanup thread");
-
- while (1) {
- pthread_mutex_lock (&priv->mutex);
+ struct timespec sleep_till = {
+ 0,
+ };
+ struct list_head recall_cleanup_list;
+ lease_inode_t *recall_entry = NULL;
+ lease_inode_t *tmp = NULL;
+ leases_private_t *priv = NULL;
+ xlator_t *this = NULL;
+ time_t time_now;
+
+ GF_VALIDATE_OR_GOTO("leases", data, out);
+
+ this = data;
+ priv = this->private;
+
+ gf_msg_debug(this->name, 0, "Started the expired_recall_cleanup thread");
+
+ while (1) {
+ time_now = gf_time();
+ pthread_mutex_lock(&priv->mutex);
+ {
+ if (priv->fini) {
+ pthread_mutex_unlock(&priv->mutex);
+ goto out;
+ }
+ INIT_LIST_HEAD(&recall_cleanup_list);
+ if (list_empty(&priv->recall_list)) {
+ sleep_till.tv_sec = time_now + 600;
+ pthread_cond_timedwait(&priv->cond, &priv->mutex, &sleep_till);
+ }
+ if (!list_empty(&priv->recall_list)) {
+ gf_msg_debug(this->name, 0, "Found expired recalls");
+ list_for_each_entry_safe(recall_entry, tmp, &priv->recall_list,
+ list)
{
- if (priv->fini) {
- pthread_mutex_unlock (&priv->mutex);
- goto out;
- }
- INIT_LIST_HEAD (&recall_cleanup_list);
- if (list_empty (&priv->recall_list)) {
- sleep_till.tv_sec = time (NULL) + 600;
- pthread_cond_timedwait (&priv->cond, &priv->mutex,
- &sleep_till);
- }
- if (!list_empty (&priv->recall_list)) {
- gf_msg_debug (this->name, 0, "Found expired recalls");
- list_for_each_entry_safe (recall_entry, tmp,
- &priv->recall_list, list) {
- list_del_init (&recall_entry->list);
- list_add_tail (&recall_entry->list, &recall_cleanup_list);
- }
- }
- }
- pthread_mutex_unlock (&priv->mutex);
-
- recall_entry = tmp = NULL;
- list_for_each_entry_safe (recall_entry, tmp, &recall_cleanup_list, list) {
- gf_msg_debug (this->name, 0, "Recall lease was sent on"
- " inode:%p, recall timer has expired"
- " and clients haven't unlocked the lease"
- " hence cleaning up leases on the inode",
- recall_entry->inode);
- remove_all_leases (this, recall_entry->inode);
- list_del_init (&recall_entry->list);
+ list_del_init(&recall_entry->list);
+ list_add_tail(&recall_entry->list, &recall_cleanup_list);
}
+ }
}
+ pthread_mutex_unlock(&priv->mutex);
+
+ recall_entry = tmp = NULL;
+ list_for_each_entry_safe(recall_entry, tmp, &recall_cleanup_list, list)
+ {
+ gf_msg_debug(this->name, 0,
+ "Recall lease was sent on"
+ " inode:%p, recall timer has expired"
+ " and clients haven't unlocked the lease"
+ " hence cleaning up leases on the inode",
+ recall_entry->inode);
+ remove_all_leases(this, recall_entry->inode);
+ /* no need to take priv->mutex lock as this entry
+ * reference is removed from global recall list. */
+ __destroy_lease_inode(recall_entry);
+ }
+ }
out:
- return NULL;
+ return NULL;
}
diff --git a/xlators/features/leases/src/leases-mem-types.h b/xlators/features/leases/src/leases-mem-types.h
index d1a59c1db2e..25664b44156 100644
--- a/xlators/features/leases/src/leases-mem-types.h
+++ b/xlators/features/leases/src/leases-mem-types.h
@@ -11,18 +11,17 @@
#ifndef __LEASES_MEM_TYPES_H__
#define __LEASES_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_leases_mem_types_ {
- gf_leases_mt_conf_t = gf_common_mt_end + 1,
- gf_leases_mt_private_t,
- gf_leases_mt_lease_client_t,
- gf_leases_mt_lease_inode_t,
- gf_leases_mt_fd_ctx_t,
- gf_leases_mt_lease_inode_ctx_t,
- gf_leases_mt_lease_id_entry_t,
- gf_leases_mt_fop_stub_t,
- gf_leases_mt_timer_data_t,
- gf_leases_mt_end
+ gf_leases_mt_private_t = gf_common_mt_end + 1,
+ gf_leases_mt_lease_client_t,
+ gf_leases_mt_lease_inode_t,
+ gf_leases_mt_fd_ctx_t,
+ gf_leases_mt_lease_inode_ctx_t,
+ gf_leases_mt_lease_id_entry_t,
+ gf_leases_mt_fop_stub_t,
+ gf_leases_mt_timer_data_t,
+ gf_leases_mt_end
};
#endif
diff --git a/xlators/features/leases/src/leases-messages.h b/xlators/features/leases/src/leases-messages.h
index f4dc169b8d6..da696b832de 100644
--- a/xlators/features/leases/src/leases-messages.h
+++ b/xlators/features/leases/src/leases-messages.h
@@ -11,7 +11,7 @@
#ifndef _LEASES_MESSAGES_H_
#define _LEASES_MESSAGES_H_
-#include "glfs-message-id.h"
+#include <glusterfs/glfs-message-id.h>
/* To add new message IDs, append new identifiers at the end of the list.
*
@@ -23,18 +23,11 @@
* glfs-message-id.h.
*/
-GLFS_MSGID(LEASES,
- LEASE_MSG_NO_MEM,
- LEASE_MSG_RECALL_FAIL,
- LEASE_MSG_INVAL_LEASE_ID,
- LEASE_MSG_INVAL_UNLK_LEASE,
- LEASE_MSG_INVAL_INODE_CTX,
- LEASE_MSG_NOT_ENABLED,
- LEASE_MSG_NO_TIMER_WHEEL,
- LEASE_MSG_CLNT_NOTFOUND,
- LEASE_MSG_INODE_NOTFOUND,
- LEASE_MSG_INVAL_FD_CTX,
- LEASE_MSG_INVAL_LEASE_TYPE
-);
+GLFS_MSGID(LEASES, LEASE_MSG_NO_MEM, LEASE_MSG_RECALL_FAIL,
+ LEASE_MSG_INVAL_LEASE_ID, LEASE_MSG_INVAL_UNLK_LEASE,
+ LEASE_MSG_INVAL_INODE_CTX, LEASE_MSG_NOT_ENABLED,
+ LEASE_MSG_NO_TIMER_WHEEL, LEASE_MSG_CLNT_NOTFOUND,
+ LEASE_MSG_INODE_NOTFOUND, LEASE_MSG_INVAL_FD_CTX,
+ LEASE_MSG_INVAL_LEASE_TYPE);
#endif /* !_LEASES_MESSAGES_H_ */
diff --git a/xlators/features/leases/src/leases.c b/xlators/features/leases/src/leases.c
index a0423491d9e..04bee50ba3f 100644
--- a/xlators/features/leases/src/leases.c
+++ b/xlators/features/leases/src/leases.c
@@ -16,1174 +16,1153 @@
#include "leases.h"
int32_t
-leases_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+leases_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
+ STACK_UNWIND_STRICT(open, frame, op_ret, op_errno, fd, xdata);
- return 0;
+ return 0;
}
-
int32_t
-leases_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- fd_t *fd, dict_t *xdata)
+leases_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
{
- uint32_t fop_flags = 0;
- int32_t op_errno = 0;
- int ret = 0;
- lease_fd_ctx_t *fd_ctx = NULL;
- char *lease_id = NULL;
-
- EXIT_IF_LEASES_OFF (this, out);
-
- fd_ctx = GF_CALLOC (1, sizeof (*fd_ctx), gf_leases_mt_fd_ctx_t);
- if (!fd_ctx) {
- op_errno = ENOMEM;
- goto err;
- }
-
- fd_ctx->client_uid = gf_strdup (frame->root->client->client_uid);
- if (!fd_ctx->client_uid) {
- op_errno = ENOMEM;
- goto err;
- }
-
- GET_FLAGS (frame->root->op, flags);
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- if (lease_id != NULL)
- memcpy (fd_ctx->lease_id, lease_id, LEASE_ID_SIZE);
- else
- memset (fd_ctx->lease_id, 0, LEASE_ID_SIZE);
-
- ret = fd_ctx_set (fd, this, (uint64_t)fd_ctx);
- if (ret) {
- op_errno = ENOMEM;
- goto err;
- }
-
- ret = check_lease_conflict (frame, fd->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ uint32_t fop_flags = 0;
+ int32_t op_errno = EINVAL;
+ int ret = 0;
+ lease_fd_ctx_t *fd_ctx = NULL;
+ char *lease_id = NULL;
+
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
+
+ fd_ctx = GF_CALLOC(1, sizeof(*fd_ctx), gf_leases_mt_fd_ctx_t);
+ if (!fd_ctx) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ fd_ctx->client_uid = gf_strdup(frame->root->client->client_uid);
+ if (!fd_ctx->client_uid) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ GET_FLAGS(frame->root->op, flags);
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ if (lease_id != NULL)
+ memcpy(fd_ctx->lease_id, lease_id, LEASE_ID_SIZE);
+ else
+ memset(fd_ctx->lease_id, 0, LEASE_ID_SIZE);
+
+ ret = fd_ctx_set(fd, this, (uint64_t)(uintptr_t)fd_ctx);
+ if (ret) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ ret = check_lease_conflict(frame, fd->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (fd->inode, open, frame, this,
- loc, flags, fd, xdata);
- return 0;
+ LEASE_BLOCK_FOP(fd->inode, open, frame, this, loc, flags, fd, xdata);
+ return 0;
out:
- STACK_WIND (frame, leases_open_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->open,
- loc, flags, fd, xdata);
- return 0;
+ STACK_WIND(frame, leases_open_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
+ return 0;
err:
- if (fd_ctx) {
- GF_FREE (fd_ctx->client_uid);
- GF_FREE (fd_ctx);
- }
+ if (fd_ctx) {
+ GF_FREE(fd_ctx->client_uid);
+ GF_FREE(fd_ctx);
+ }
- op_errno = (op_errno == -1) ? errno : op_errno;
- STACK_UNWIND_STRICT (open, frame, -1, op_errno, NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(open, frame, -1, op_errno, NULL, NULL);
+ return 0;
}
int32_t
-leases_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+leases_writev_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata)
{
- STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno,
- prebuf, postbuf, xdata);
+ STACK_UNWIND_STRICT(writev, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
- return 0;
+ return 0;
}
-
int32_t
-leases_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int count, off_t off, uint32_t flags,
- struct iobref *iobref, dict_t *xdata)
+leases_writev(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int count, off_t off, uint32_t flags,
+ struct iobref *iobref, dict_t *xdata)
{
- uint32_t fop_flags = 0;
- char *lease_id = NULL;
- int ret = 0;
+ uint32_t fop_flags = 0;
+ char *lease_id = NULL;
+ int ret = 0;
- EXIT_IF_LEASES_OFF (this, out);
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- GET_FLAGS (frame->root->op, fd->flags);
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ GET_FLAGS(frame->root->op, fd->flags);
- ret = check_lease_conflict (frame, fd->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ ret = check_lease_conflict(frame, fd->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (fd->inode, writev, frame, this, fd, vector, count,
- off, flags, iobref, xdata);
- return 0;
+ LEASE_BLOCK_FOP(fd->inode, writev, frame, this, fd, vector, count, off,
+ flags, iobref, xdata);
+ return 0;
out:
- STACK_WIND (frame, leases_writev_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->writev,
- fd, vector, count, off, flags, iobref, xdata);
- return 0;
+ STACK_WIND(frame, leases_writev_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector, count, off, flags,
+ iobref, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (writev, frame, -1, errno, NULL, NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(writev, frame, -1, errno, NULL, NULL, NULL);
+ return 0;
}
-
int32_t
-leases_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno,
- struct iovec *vector, int count, struct iatt *stbuf,
- struct iobref *iobref, dict_t *xdata)
+leases_readv_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iovec *vector, int count,
+ struct iatt *stbuf, struct iobref *iobref, dict_t *xdata)
{
- STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector,
- count, stbuf, iobref, xdata);
+ STACK_UNWIND_STRICT(readv, frame, op_ret, op_errno, vector, count, stbuf,
+ iobref, xdata);
- return 0;
+ return 0;
}
int32_t
-leases_readv (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset,
- uint32_t flags, dict_t *xdata)
+leases_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- uint32_t fop_flags = 0;
- char *lease_id = NULL;
- int ret = 0;
+ uint32_t fop_flags = 0;
+ char *lease_id = NULL;
+ int ret = 0;
- EXIT_IF_LEASES_OFF (this, out);
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- GET_FLAGS (frame->root->op, fd->flags);
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ GET_FLAGS(frame->root->op, fd->flags);
- ret = check_lease_conflict (frame, fd->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ ret = check_lease_conflict(frame, fd->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (fd->inode, readv, frame, this,
- fd, size, offset, flags, xdata);
- return 0;
+ LEASE_BLOCK_FOP(fd->inode, readv, frame, this, fd, size, offset, flags,
+ xdata);
+ return 0;
out:
- STACK_WIND (frame, leases_readv_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->readv,
- fd, size, offset, flags, xdata);
- return 0;
+ STACK_WIND(frame, leases_readv_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readv, fd, size, offset, flags, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (readv, frame, -1, errno, NULL, 0,
- NULL, NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(readv, frame, -1, errno, NULL, 0, NULL, NULL, NULL);
+ return 0;
}
int32_t
-leases_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
- dict_t *xdata)
+leases_lk_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
{
- STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, lock, xdata);
+ STACK_UNWIND_STRICT(lk, frame, op_ret, op_errno, lock, xdata);
- return 0;
+ return 0;
}
int32_t
-leases_lk (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
+leases_lk(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
+ struct gf_flock *flock, dict_t *xdata)
{
- uint32_t fop_flags = 0;
- char *lease_id = NULL;
- int ret = 0;
+ uint32_t fop_flags = 0;
+ char *lease_id = NULL;
+ int ret = 0;
- EXIT_IF_LEASES_OFF (this, out);
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- GET_FLAGS_LK (cmd, flock->l_type, fd->flags);
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ GET_FLAGS_LK(cmd, flock->l_type, fd->flags);
- ret = check_lease_conflict (frame, fd->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ ret = check_lease_conflict(frame, fd->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (fd->inode, lk, frame, this,
- fd, cmd, flock, xdata);
- return 0;
+ LEASE_BLOCK_FOP(fd->inode, lk, frame, this, fd, cmd, flock, xdata);
+ return 0;
out:
- STACK_WIND (frame, leases_lk_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lk,
- fd, cmd, flock, xdata);
- return 0;
+ STACK_WIND(frame, leases_lk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lk, fd, cmd, flock, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (lk, frame, -1, errno, NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(lk, frame, -1, errno, NULL, NULL);
+ return 0;
}
int32_t
-leases_lease (call_frame_t *frame, xlator_t *this,
- loc_t *loc, struct gf_lease *lease, dict_t *xdata)
+leases_lease(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct gf_lease *lease, dict_t *xdata)
{
- int32_t op_errno = 0;
- int ret = 0;
- struct gf_lease nullease = {0, };
- int32_t op_ret = 0;
-
- EXIT_IF_LEASES_OFF (this, out);
-
- ret = process_lease_req (frame, this, loc->inode, lease);
- if (ret < 0) {
- op_errno = -ret;
- op_ret = -1;
- }
- goto unwind;
+ int32_t op_errno = 0;
+ int ret = 0;
+ struct gf_lease nullease = {
+ 0,
+ };
+ int32_t op_ret = 0;
+
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
+
+ ret = process_lease_req(frame, this, loc->inode, lease);
+ if (ret < 0) {
+ op_errno = -ret;
+ op_ret = -1;
+ }
+ goto unwind;
out:
- gf_msg (this->name, GF_LOG_ERROR, EINVAL, LEASE_MSG_NOT_ENABLED,
- "\"features/leases\" translator is not enabled. "
- "You need to enable it for proper functioning of your "
- "application");
- op_errno = ENOSYS;
- op_ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, EINVAL, LEASE_MSG_NOT_ENABLED,
+ "\"features/leases\" translator is not enabled. "
+ "You need to enable it for proper functioning of your "
+ "application");
+ op_errno = ENOSYS;
+ op_ret = -1;
unwind:
- STACK_UNWIND_STRICT (lease, frame, op_ret, op_errno,
- (op_errno == ENOSYS) ? &nullease : lease, xdata);
- return 0;
+ STACK_UNWIND_STRICT(lease, frame, op_ret, op_errno,
+ (op_errno == ENOSYS) ? &nullease : lease, xdata);
+ return 0;
}
int32_t
-leases_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+leases_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno,
- prebuf, postbuf, xdata);
+ STACK_UNWIND_STRICT(truncate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
- return 0;
+ return 0;
}
int32_t
-leases_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
- dict_t *xdata)
+leases_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- uint32_t fop_flags = 0;
- char *lease_id = NULL;
- int ret = 0;
+ uint32_t fop_flags = 0;
+ char *lease_id = NULL;
+ int ret = 0;
- EXIT_IF_LEASES_OFF (this, out);
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- GET_FLAGS (frame->root->op, 0);
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ GET_FLAGS(frame->root->op, 0);
- ret = check_lease_conflict (frame, loc->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ ret = check_lease_conflict(frame, loc->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (loc->inode, truncate, frame, this,
- loc, offset, xdata);
- return 0;
+ LEASE_BLOCK_FOP(loc->inode, truncate, frame, this, loc, offset, xdata);
+ return 0;
out:
- STACK_WIND (frame, leases_truncate_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->truncate,
- loc, offset, xdata);
- return 0;
+ STACK_WIND(frame, leases_truncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (truncate, frame, -1, errno, NULL, NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(truncate, frame, -1, errno, NULL, NULL, NULL);
+ return 0;
}
int32_t
-leases_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *statpre,
- struct iatt *statpost, dict_t *xdata)
+leases_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
{
- STACK_UNWIND_STRICT (setattr, frame, op_ret, op_errno,
- statpre, statpost, xdata);
+ STACK_UNWIND_STRICT(setattr, frame, op_ret, op_errno, statpre, statpost,
+ xdata);
- return 0;
+ return 0;
}
int32_t
-leases_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+leases_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- uint32_t fop_flags = 0;
- char *lease_id = NULL;
- int ret = 0;
+ uint32_t fop_flags = 0;
+ char *lease_id = NULL;
+ int ret = 0;
- EXIT_IF_LEASES_OFF (this, out);
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- GET_FLAGS (frame->root->op, 0);
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ GET_FLAGS(frame->root->op, 0);
- ret = check_lease_conflict (frame, loc->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ ret = check_lease_conflict(frame, loc->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (loc->inode, setattr, frame, this,
- loc, stbuf, valid, xdata);
- return 0;
+ LEASE_BLOCK_FOP(loc->inode, setattr, frame, this, loc, stbuf, valid, xdata);
+ return 0;
out:
- STACK_WIND (frame, leases_setattr_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->setattr,
- loc, stbuf, valid, xdata);
- return 0;
+ STACK_WIND(frame, leases_setattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (setattr, frame, -1, errno, NULL, NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(setattr, frame, -1, errno, NULL, NULL, NULL);
+ return 0;
}
int32_t
-leases_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent,
- dict_t *xdata)
+leases_rename_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
+ struct iatt *preoldparent, struct iatt *postoldparent,
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
{
- STACK_UNWIND_STRICT (rename, frame, op_ret, op_errno,
- stbuf, preoldparent, postoldparent,
- prenewparent, postnewparent, xdata);
+ STACK_UNWIND_STRICT(rename, frame, op_ret, op_errno, stbuf, preoldparent,
+ postoldparent, prenewparent, postnewparent, xdata);
- return 0;
+ return 0;
}
int32_t
-leases_rename (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+leases_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- uint32_t fop_flags = 0;
- char *lease_id = NULL;
- int ret = 0;
+ uint32_t fop_flags = 0;
+ char *lease_id = NULL;
+ int ret = 0;
- EXIT_IF_LEASES_OFF (this, out);
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
- /* should the lease be also checked for newloc */
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- GET_FLAGS (frame->root->op, 0);
+ /* should the lease be also checked for newloc */
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ GET_FLAGS(frame->root->op, 0);
- ret = check_lease_conflict (frame, oldloc->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ ret = check_lease_conflict(frame, oldloc->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (oldloc->inode, rename, frame, this,
- oldloc, newloc, xdata);
- return 0;
+ LEASE_BLOCK_FOP(oldloc->inode, rename, frame, this, oldloc, newloc, xdata);
+ return 0;
out:
- STACK_WIND (frame, leases_rename_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->rename,
- oldloc, newloc, xdata);
- return 0;
+ STACK_WIND(frame, leases_rename_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (rename, frame, -1, errno, NULL,
- NULL, NULL, NULL, NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(rename, frame, -1, errno, NULL, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
int32_t
-leases_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+leases_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno,
- preparent, postparent, xdata);
+ STACK_UNWIND_STRICT(unlink, frame, op_ret, op_errno, preparent, postparent,
+ xdata);
- return 0;
+ return 0;
}
int32_t
-leases_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
- dict_t *xdata)
+leases_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
{
- uint32_t fop_flags = 0;
- char *lease_id = NULL;
- int ret = 0;
+ uint32_t fop_flags = 0;
+ char *lease_id = NULL;
+ int ret = 0;
- EXIT_IF_LEASES_OFF (this, out);
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- GET_FLAGS (frame->root->op, 0);
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ GET_FLAGS(frame->root->op, 0);
- ret = check_lease_conflict (frame, loc->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ ret = check_lease_conflict(frame, loc->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (loc->inode, unlink, frame, this,
- loc, xflag, xdata);
- return 0;
+ LEASE_BLOCK_FOP(loc->inode, unlink, frame, this, loc, xflag, xdata);
+ return 0;
out:
- STACK_WIND (frame, leases_unlink_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->unlink,
- loc, xflag, xdata);
- return 0;
+ STACK_WIND(frame, leases_unlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (unlink, frame, -1, errno, NULL, NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(unlink, frame, -1, errno, NULL, NULL, NULL);
+ return 0;
}
int32_t
-leases_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
+leases_link_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, inode_t *inode, struct iatt *stbuf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
{
- STACK_UNWIND_STRICT (link, frame, op_ret, op_errno,
- inode, stbuf, preparent, postparent, xdata);
+ STACK_UNWIND_STRICT(link, frame, op_ret, op_errno, inode, stbuf, preparent,
+ postparent, xdata);
- return 0;
+ return 0;
}
int32_t
-leases_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
- loc_t *newloc, dict_t *xdata)
+leases_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- uint32_t fop_flags = 0;
- char *lease_id = NULL;
- int ret = 0;
+ uint32_t fop_flags = 0;
+ char *lease_id = NULL;
+ int ret = 0;
- EXIT_IF_LEASES_OFF (this, out);
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- GET_FLAGS (frame->root->op, 0);
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ GET_FLAGS(frame->root->op, 0);
- ret = check_lease_conflict (frame, oldloc->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ ret = check_lease_conflict(frame, oldloc->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (oldloc->inode, link, frame, this,
- oldloc, newloc, xdata);
- return 0;
+ LEASE_BLOCK_FOP(oldloc->inode, link, frame, this, oldloc, newloc, xdata);
+ return 0;
out:
- STACK_WIND (frame, leases_link_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->link,
- oldloc, newloc, xdata);
- return 0;
+ STACK_WIND(frame, leases_link_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (link, frame, -1, errno, NULL,
- NULL, NULL, NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(link, frame, -1, errno, NULL, NULL, NULL, NULL, NULL);
+ return 0;
}
int32_t
-leases_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, fd_t *fd, inode_t *inode,
- struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+leases_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, fd_t *fd, inode_t *inode, struct iatt *stbuf,
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd,
- inode, stbuf, preparent, postparent, xdata);
+ STACK_UNWIND_STRICT(create, frame, op_ret, op_errno, fd, inode, stbuf,
+ preparent, postparent, xdata);
- return 0;
+ return 0;
}
int32_t
-leases_create (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, mode_t mode,
- mode_t umask, fd_t *fd, dict_t *xdata)
+leases_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- uint32_t fop_flags = 0;
- char *lease_id = NULL;
- int ret = 0;
+ uint32_t fop_flags = 0;
+ char *lease_id = NULL;
+ int ret = 0;
- EXIT_IF_LEASES_OFF (this, out);
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- GET_FLAGS (frame->root->op, flags);
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ GET_FLAGS(frame->root->op, flags);
- ret = check_lease_conflict (frame, fd->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ ret = check_lease_conflict(frame, fd->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (fd->inode, create, frame, this,
- loc, flags, mode, umask, fd, xdata);
- return 0;
+ LEASE_BLOCK_FOP(fd->inode, create, frame, this, loc, flags, mode, umask, fd,
+ xdata);
+ return 0;
out:
- STACK_WIND (frame, leases_create_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->create,
- loc, flags, mode, umask, fd, xdata);
- return 0;
+ STACK_WIND(frame, leases_create_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, flags, mode, umask, fd,
+ xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (create, frame, -1, errno, NULL, NULL, NULL,
- NULL, NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(create, frame, -1, errno, NULL, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
int32_t
-leases_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf,
- dict_t *xdata)
+leases_fsync_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- STACK_UNWIND_STRICT (fsync, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
- return 0;
+ STACK_UNWIND_STRICT(fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+ return 0;
}
int32_t
-leases_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int32_t flags, dict_t *xdata)
+leases_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+ dict_t *xdata)
{
- uint32_t fop_flags = 0;
- char *lease_id = NULL;
- int ret = 0;
+ uint32_t fop_flags = 0;
+ char *lease_id = NULL;
+ int ret = 0;
- EXIT_IF_LEASES_OFF (this, out);
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- GET_FLAGS (frame->root->op, fd->flags);
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ GET_FLAGS(frame->root->op, fd->flags);
- ret = check_lease_conflict (frame, fd->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ ret = check_lease_conflict(frame, fd->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (fd->inode, fsync, frame, this,
- fd, flags, xdata);
- return 0;
+ LEASE_BLOCK_FOP(fd->inode, fsync, frame, this, fd, flags, xdata);
+ return 0;
out:
- STACK_WIND (frame, leases_fsync_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsync, fd, flags, xdata);
- return 0;
+ STACK_WIND(frame, leases_fsync_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsync, fd, flags, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (fsync, frame, -1, errno, NULL, NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(fsync, frame, -1, errno, NULL, NULL, NULL);
+ return 0;
}
int32_t
-leases_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf,
- dict_t *xdata)
+leases_ftruncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
- return 0;
+ STACK_UNWIND_STRICT(ftruncate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
int32_t
-leases_ftruncate (call_frame_t *frame, xlator_t *this,
- fd_t *fd, off_t offset, dict_t *xdata)
+leases_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- uint32_t fop_flags = 0;
- char *lease_id = NULL;
- int ret = 0;
+ uint32_t fop_flags = 0;
+ char *lease_id = NULL;
+ int ret = 0;
- EXIT_IF_LEASES_OFF (this, out);
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- GET_FLAGS (frame->root->op, 0); /* TODO:fd->flags?*/
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ GET_FLAGS(frame->root->op, 0); /* TODO:fd->flags?*/
- ret = check_lease_conflict (frame, fd->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ ret = check_lease_conflict(frame, fd->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (fd->inode, ftruncate, frame, this,
- fd, offset, xdata);
- return 0;
+ LEASE_BLOCK_FOP(fd->inode, ftruncate, frame, this, fd, offset, xdata);
+ return 0;
out:
- STACK_WIND (frame, leases_ftruncate_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->ftruncate,
- fd, offset, xdata);
- return 0;
+ STACK_WIND(frame, leases_ftruncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (ftruncate, frame, -1, errno, NULL,
- NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(ftruncate, frame, -1, errno, NULL, NULL, NULL);
+ return 0;
}
int32_t
-leases_fsetattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *statpre,
- struct iatt *statpost, dict_t *xdata)
+leases_fsetattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
{
- STACK_UNWIND_STRICT (fsetattr, frame, op_ret, op_errno,
- statpre, statpost, xdata);
- return 0;
+ STACK_UNWIND_STRICT(fsetattr, frame, op_ret, op_errno, statpre, statpost,
+ xdata);
+ return 0;
}
int32_t
-leases_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+leases_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- uint32_t fop_flags = 0;
- char *lease_id = NULL;
- int ret = 0;
+ uint32_t fop_flags = 0;
+ char *lease_id = NULL;
+ int ret = 0;
- EXIT_IF_LEASES_OFF (this, out);
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- GET_FLAGS (frame->root->op, fd->flags);
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ GET_FLAGS(frame->root->op, fd->flags);
- ret = check_lease_conflict (frame, fd->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ ret = check_lease_conflict(frame, fd->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (fd->inode, fsetattr, frame, this,
- fd, stbuf, valid, xdata);
- return 0;
+ LEASE_BLOCK_FOP(fd->inode, fsetattr, frame, this, fd, stbuf, valid, xdata);
+ return 0;
out:
- STACK_WIND (frame, leases_fsetattr_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->fsetattr,
- fd, stbuf, valid, xdata);
- return 0;
+ STACK_WIND(frame, leases_fsetattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetattr, fd, stbuf, valid, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (fsetattr, frame, -1, errno, NULL,
- NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(fsetattr, frame, -1, errno, NULL, NULL, NULL);
+ return 0;
}
int32_t
-leases_fallocate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *pre,
- struct iatt *post, dict_t *xdata)
+leases_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *pre,
+ struct iatt *post, dict_t *xdata)
{
- STACK_UNWIND_STRICT (fallocate, frame, op_ret, op_errno, pre,
- post, xdata);
+ STACK_UNWIND_STRICT(fallocate, frame, op_ret, op_errno, pre, post, xdata);
- return 0;
+ return 0;
}
int32_t
-leases_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int32_t mode, off_t offset, size_t len, dict_t *xdata)
+leases_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+ off_t offset, size_t len, dict_t *xdata)
{
- uint32_t fop_flags = 0;
- char *lease_id = NULL;
- int ret = 0;
+ uint32_t fop_flags = 0;
+ char *lease_id = NULL;
+ int ret = 0;
- EXIT_IF_LEASES_OFF (this, out);
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- GET_FLAGS (frame->root->op, fd->flags);
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ GET_FLAGS(frame->root->op, fd->flags);
- ret = check_lease_conflict (frame, fd->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ ret = check_lease_conflict(frame, fd->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (fd->inode, fallocate, frame, this,
- fd, mode, offset, len, xdata);
- return 0;
+ LEASE_BLOCK_FOP(fd->inode, fallocate, frame, this, fd, mode, offset, len,
+ xdata);
+ return 0;
out:
- STACK_WIND (frame, leases_fallocate_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->fallocate,
- fd, mode, offset, len, xdata);
- return 0;
+ STACK_WIND(frame, leases_fallocate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fallocate, fd, mode, offset, len,
+ xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (fallocate, frame, -1, errno, NULL,
- NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(fallocate, frame, -1, errno, NULL, NULL, NULL);
+ return 0;
}
int32_t
-leases_discard_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *pre,
- struct iatt *post, dict_t *xdata)
+leases_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *pre,
+ struct iatt *post, dict_t *xdata)
{
- STACK_UNWIND_STRICT (discard, frame, op_ret, op_errno, pre,
- post, xdata);
+ STACK_UNWIND_STRICT(discard, frame, op_ret, op_errno, pre, post, xdata);
- return 0;
+ return 0;
}
int32_t
-leases_discard (call_frame_t *frame, xlator_t *this, fd_t *fd,
- off_t offset, size_t len, dict_t *xdata)
+leases_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
{
- uint32_t fop_flags = 0;
- char *lease_id = NULL;
- int ret = 0;
+ uint32_t fop_flags = 0;
+ char *lease_id = NULL;
+ int ret = 0;
- EXIT_IF_LEASES_OFF (this, out);
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- GET_FLAGS (frame->root->op, fd->flags);
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ GET_FLAGS(frame->root->op, fd->flags);
- ret = check_lease_conflict (frame, fd->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ ret = check_lease_conflict(frame, fd->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (fd->inode, discard, frame, this,
- fd, offset, len, xdata);
- return 0;
+ LEASE_BLOCK_FOP(fd->inode, discard, frame, this, fd, offset, len, xdata);
+ return 0;
out:
- STACK_WIND (frame, leases_discard_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->discard,
- fd, offset, len, xdata);
- return 0;
+ STACK_WIND(frame, leases_discard_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->discard, fd, offset, len, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (discard, frame, -1, errno, NULL,
- NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(discard, frame, -1, errno, NULL, NULL, NULL);
+ return 0;
}
int32_t
-leases_zerofill_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *pre,
- struct iatt *post, dict_t *xdata)
+leases_zerofill_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *pre,
+ struct iatt *post, dict_t *xdata)
{
- STACK_UNWIND_STRICT (zerofill, frame, op_ret, op_errno, pre,
- post, xdata);
+ STACK_UNWIND_STRICT(zerofill, frame, op_ret, op_errno, pre, post, xdata);
- return 0;
+ return 0;
}
int
-leases_zerofill (call_frame_t *frame, xlator_t *this, fd_t *fd,
- off_t offset, off_t len, dict_t *xdata)
+leases_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ off_t len, dict_t *xdata)
{
- uint32_t fop_flags = 0;
- char *lease_id = NULL;
- int ret = 0;
+ uint32_t fop_flags = 0;
+ char *lease_id = NULL;
+ int ret = 0;
- EXIT_IF_LEASES_OFF (this, out);
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- GET_FLAGS (frame->root->op, fd->flags);
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ GET_FLAGS(frame->root->op, fd->flags);
- ret = check_lease_conflict (frame, fd->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ ret = check_lease_conflict(frame, fd->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (fd->inode, zerofill, frame, this,
- fd, offset, len, xdata);
- return 0;
+ LEASE_BLOCK_FOP(fd->inode, zerofill, frame, this, fd, offset, len, xdata);
+ return 0;
out:
- STACK_WIND (frame, leases_zerofill_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->zerofill,
- fd, offset, len, xdata);
- return 0;
+ STACK_WIND(frame, leases_zerofill_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->zerofill, fd, offset, len, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (zerofill, frame, -1, errno, NULL,
- NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(zerofill, frame, -1, errno, NULL, NULL, NULL);
+ return 0;
}
int
-leases_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+leases_flush_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno, xdata);
+ STACK_UNWIND_STRICT(flush, frame, op_ret, op_errno, xdata);
- return 0;
+ return 0;
}
int
-leases_flush (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *xdata)
+leases_flush(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- uint32_t fop_flags = 0;
- char *lease_id = NULL;
- int ret = 0;
- lease_fd_ctx_t *fd_ctx = NULL;
- uint64_t ctx = 0;
-
- EXIT_IF_LEASES_OFF (this, out);
-
- GET_LEASE_ID (xdata, lease_id, frame->root->client->client_uid);
- GET_FLAGS (frame->root->op, fd->flags);
-
- ret = check_lease_conflict (frame, fd->inode, lease_id, fop_flags);
- if (ret < 0)
- goto err;
- else if (ret == BLOCK_FOP)
- goto block;
- else if (ret == WIND_FOP)
- goto out;
+ uint32_t fop_flags = 0;
+ char *lease_id = NULL;
+ int ret = 0;
+ lease_fd_ctx_t *fd_ctx = NULL;
+ uint64_t ctx = 0;
+
+ EXIT_IF_LEASES_OFF(this, out);
+ EXIT_IF_INTERNAL_FOP(frame, xdata, out);
+
+ GET_LEASE_ID(xdata, lease_id, frame->root->client->client_uid);
+ GET_FLAGS(frame->root->op, fd->flags);
+
+ ret = check_lease_conflict(frame, fd->inode, lease_id, fop_flags);
+ if (ret < 0)
+ goto err;
+ else if (ret == BLOCK_FOP)
+ goto block;
+ else if (ret == WIND_FOP)
+ goto out;
block:
- LEASE_BLOCK_FOP (fd->inode, flush, frame, this,
- fd, xdata);
- return 0;
+ LEASE_BLOCK_FOP(fd->inode, flush, frame, this, fd, xdata);
+ return 0;
out:
- /* *
- * currently release is not called after the close fop from the
- * application. Hence lease fd ctx is resetted on here.
- * This is actually not the right way, since flush can be called
- * not only from the close op.
- * TODO :
- * - Either identify the flush is called from close call on fd from
- * from the application.
- * OR
- * - Find why release is not called post the last close call
- */
- ret = fd_ctx_get (fd, this, &ctx);
- if (ret == 0) {
- fd_ctx = (lease_fd_ctx_t *)(long)ctx;
- if (fd_ctx->client_uid) {
- GF_FREE (fd_ctx->client_uid);
- fd_ctx->client_uid = NULL;
- }
- memset (fd_ctx->lease_id, 0, LEASE_ID_SIZE);
+ /* *
+ * currently release is not called after the close fop from the
+ * application. Hence lease fd ctx is reset on here.
+ * This is actually not the right way, since flush can be called
+ * not only from the close op.
+ * TODO :
+ * - Either identify the flush is called from close call on fd from
+ * from the application.
+ * OR
+ * - Find why release is not called post the last close call
+ */
+ ret = fd_ctx_get(fd, this, &ctx);
+ if (ret == 0) {
+ fd_ctx = (lease_fd_ctx_t *)(long)ctx;
+ if (fd_ctx->client_uid) {
+ GF_FREE(fd_ctx->client_uid);
+ fd_ctx->client_uid = NULL;
}
- STACK_WIND (frame, leases_flush_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->flush, fd, xdata);
- return 0;
+ memset(fd_ctx->lease_id, 0, LEASE_ID_SIZE);
+ }
+ STACK_WIND(frame, leases_flush_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->flush, fd, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (create, frame, -1, errno, NULL,
- NULL, NULL, NULL, NULL, NULL);
- return 0;
+ STACK_UNWIND_STRICT(create, frame, -1, errno, NULL, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- if (!this)
- return ret;
-
- ret = xlator_mem_acct_init (this, gf_leases_mt_end + 1);
+ if (!this)
+ return ret;
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM, LEASE_MSG_NO_MEM,
- "mem account init failed");
- return ret;
- }
+ ret = xlator_mem_acct_init(this, gf_leases_mt_end + 1);
+ if (ret != 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, LEASE_MSG_NO_MEM,
+ "mem account init failed");
return ret;
+ }
+
+ return ret;
}
static int
-leases_init_priv (xlator_t *this)
+leases_init_priv(xlator_t *this)
{
- int ret = 0;
- leases_private_t *priv = NULL;
+ int ret = 0;
+ leases_private_t *priv = NULL;
- priv = this->private;
- GF_ASSERT (priv);
+ priv = this->private;
+ GF_ASSERT(priv);
+ if (!priv->timer_wheel) {
+ priv->timer_wheel = glusterfs_ctx_tw_get(this->ctx);
if (!priv->timer_wheel) {
- priv->timer_wheel = glusterfs_ctx_tw_get (this->ctx);
- if (!priv->timer_wheel) {
- ret = -1;
- goto out;
- }
+ ret = -1;
+ goto out;
}
+ }
- if (!priv->inited_recall_thr) {
- ret = gf_thread_create (&priv->recall_thr, NULL,
- expired_recall_cleanup, this,
- "leasercl");
- if (!ret)
- priv->inited_recall_thr = _gf_true;
- }
+ if (!priv->inited_recall_thr) {
+ ret = gf_thread_create(&priv->recall_thr, NULL, expired_recall_cleanup,
+ this, "leasercl");
+ if (!ret)
+ priv->inited_recall_thr = _gf_true;
+ }
out:
- return ret;
+ return ret;
}
int
-reconfigure (xlator_t *this, dict_t *options)
+reconfigure(xlator_t *this, dict_t *options)
{
- leases_private_t *priv = NULL;
- int ret = -1;
+ leases_private_t *priv = NULL;
+ int ret = -1;
- priv = this->private;
- GF_ASSERT (priv);
+ priv = this->private;
+ GF_ASSERT(priv);
- /* TODO: In case of reconfigure, if its enabling the leases
- * its not an issue, but if its disabling the leases, there
- * is more to it, like recall all the existing leases, wait
- * for unlock of all the leases etc., hence not supporting the
- * reconfigure for now.
+ /* TODO: In case of reconfigure, if its enabling the leases
+ * its not an issue, but if its disabling the leases, there
+ * is more to it, like recall all the existing leases, wait
+ * for unlock of all the leases etc., hence not supporting the
+ * reconfigure for now.
- GF_OPTION_RECONF ("leases", priv->leases_enabled,
- options, bool, out);
+ GF_OPTION_RECONF ("leases", priv->leases_enabled,
+ options, bool, out);
- if (priv->leases_enabled) {
- ret = leases_init_priv (this);
- if (ret)
- goto out;
- }
- */
+ if (priv->leases_enabled) {
+ ret = leases_init_priv (this);
+ if (ret)
+ goto out;
+ }
+ */
- GF_OPTION_RECONF ("lease-lock-recall-timeout",
- priv->recall_lease_timeout,
- options, int32, out);
+ GF_OPTION_RECONF("lease-lock-recall-timeout", priv->recall_lease_timeout,
+ options, int32, out);
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-init (xlator_t *this)
+init(xlator_t *this)
{
- int ret = -1;
- leases_private_t *priv = NULL;
-
- priv = GF_CALLOC (1, sizeof (*priv),
- gf_leases_mt_private_t);
- if (!priv) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM, LEASE_MSG_NO_MEM,
- "Leases init failed");
- goto out;
- }
+ int ret = -1;
+ leases_private_t *priv = NULL;
- GF_OPTION_INIT ("leases", priv->leases_enabled,
- bool, out);
- GF_OPTION_INIT ("lease-lock-recall-timeout",
- priv->recall_lease_timeout, int32, out);
- pthread_mutex_init (&priv->mutex, NULL);
- INIT_LIST_HEAD (&priv->client_list);
- INIT_LIST_HEAD (&priv->recall_list);
+ priv = GF_CALLOC(1, sizeof(*priv), gf_leases_mt_private_t);
+ if (!priv) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, LEASE_MSG_NO_MEM,
+ "Leases init failed");
+ goto out;
+ }
- this->private = priv;
+ GF_OPTION_INIT("leases", priv->leases_enabled, bool, out);
+ GF_OPTION_INIT("lease-lock-recall-timeout", priv->recall_lease_timeout,
+ int32, out);
+ pthread_mutex_init(&priv->mutex, NULL);
+ INIT_LIST_HEAD(&priv->client_list);
+ INIT_LIST_HEAD(&priv->recall_list);
- if (priv->leases_enabled) {
- ret = leases_init_priv (this);
- if (ret)
- goto out;
- }
+ this->private = priv;
- ret = 0;
+ if (priv->leases_enabled) {
+ ret = leases_init_priv(this);
+ if (ret)
+ goto out;
+ }
+
+ ret = 0;
out:
- if (ret) {
- GF_FREE (priv);
- this->private = NULL;
- }
+ if (ret) {
+ GF_FREE(priv);
+ this->private = NULL;
+ }
- return ret;
+ return ret;
}
-int
-fini (xlator_t *this)
+void
+fini(xlator_t *this)
{
- leases_private_t *priv = NULL;
-
- priv = this->private;
- if (!priv) {
- return 0;
- }
- this->private = NULL;
-
- priv->fini = _gf_true;
- pthread_cond_broadcast (&priv->cond);
- pthread_join (priv->recall_thr, NULL);
-
+ leases_private_t *priv = NULL;
+
+ priv = this->private;
+ if (!priv) {
+ return;
+ }
+ this->private = NULL;
+
+ priv->fini = _gf_true;
+ pthread_cond_broadcast(&priv->cond);
+ if (priv->recall_thr) {
+ gf_thread_cleanup_xint(priv->recall_thr);
+ priv->recall_thr = 0;
priv->inited_recall_thr = _gf_false;
+ }
- GF_FREE (priv);
+ if (priv->timer_wheel) {
+ glusterfs_ctx_tw_put(this->ctx);
+ }
- glusterfs_ctx_tw_put (this->ctx);
-
- return 0;
+ GF_FREE(priv);
+ return;
}
static int
-leases_forget (xlator_t *this, inode_t *inode)
+leases_forget(xlator_t *this, inode_t *inode)
{
- /* TODO:leases_cleanup_inode_ctx (this, inode); */
- return 0;
+ /* TODO:leases_cleanup_inode_ctx (this, inode); */
+ return 0;
}
static int
-leases_release (xlator_t *this, fd_t *fd)
+leases_release(xlator_t *this, fd_t *fd)
{
- int ret = -1;
- uint64_t tmp = 0;
- lease_fd_ctx_t *fd_ctx = NULL;
+ int ret = -1;
+ uint64_t tmp = 0;
+ lease_fd_ctx_t *fd_ctx = NULL;
- if (fd == NULL) {
- goto out;
- }
+ if (fd == NULL) {
+ goto out;
+ }
- gf_log (this->name, GF_LOG_TRACE,
- "Releasing all leases with fd %p", fd);
+ gf_log(this->name, GF_LOG_TRACE, "Releasing all leases with fd %p", fd);
- ret = fd_ctx_del (fd, this, &tmp);
- if (ret) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Could not get fdctx");
- goto out;
- }
+ ret = fd_ctx_del(fd, this, &tmp);
+ if (ret) {
+ gf_log(this->name, GF_LOG_DEBUG, "Could not get fdctx");
+ goto out;
+ }
- fd_ctx = (lease_fd_ctx_t *)(long)tmp;
- if (fd_ctx)
- GF_FREE (fd_ctx);
+ fd_ctx = (lease_fd_ctx_t *)(long)tmp;
+ if (fd_ctx)
+ GF_FREE(fd_ctx);
out:
- return ret;
-
+ return ret;
}
static int
-leases_clnt_disconnect_cbk (xlator_t *this, client_t *client)
+leases_clnt_disconnect_cbk(xlator_t *this, client_t *client)
{
- int ret = 0;
+ int ret = 0;
- EXIT_IF_LEASES_OFF (this, out);
+ EXIT_IF_LEASES_OFF(this, out);
- ret = cleanup_client_leases (this, client->client_uid);
+ ret = cleanup_client_leases(this, client->client_uid);
out:
- return ret;
+ return ret;
}
struct xlator_fops fops = {
- /* Metadata modifying fops */
- .fsetattr = leases_fsetattr,
- .setattr = leases_setattr,
-
- /* File Data reading fops */
- .open = leases_open,
- .readv = leases_readv,
-
- /* File Data modifying fops */
- .truncate = leases_truncate,
- .ftruncate = leases_ftruncate,
- .writev = leases_writev,
- .zerofill = leases_zerofill,
- .fallocate = leases_fallocate,
- .discard = leases_discard,
- .lk = leases_lk,
- .fsync = leases_fsync,
- .flush = leases_flush,
- .lease = leases_lease,
-
- /* Directory Data modifying fops */
- .create = leases_create,
- .rename = leases_rename,
- .unlink = leases_unlink,
- .link = leases_link,
+ /* Metadata modifying fops */
+ .fsetattr = leases_fsetattr,
+ .setattr = leases_setattr,
+
+ /* File Data reading fops */
+ .open = leases_open,
+ .readv = leases_readv,
+
+ /* File Data modifying fops */
+ .truncate = leases_truncate,
+ .ftruncate = leases_ftruncate,
+ .writev = leases_writev,
+ .zerofill = leases_zerofill,
+ .fallocate = leases_fallocate,
+ .discard = leases_discard,
+ .lk = leases_lk,
+ .fsync = leases_fsync,
+ .flush = leases_flush,
+ .lease = leases_lease,
+
+ /* Directory Data modifying fops */
+ .create = leases_create,
+ .rename = leases_rename,
+ .unlink = leases_unlink,
+ .link = leases_link,
#ifdef NOT_SUPPORTED
- /* internal lk fops */
- .inodelk = leases_inodelk,
- .finodelk = leases_finodelk,
- .entrylk = leases_entrylk,
- .fentrylk = leases_fentrylk,
-
- /* Internal special fops*/
- .xattrop = leases_xattrop,
- .fxattrop = leases_fxattrop,
+ /* internal lk fops */
+ .inodelk = leases_inodelk,
+ .finodelk = leases_finodelk,
+ .entrylk = leases_entrylk,
+ .fentrylk = leases_fentrylk,
+
+ /* Internal special fops*/
+ .xattrop = leases_xattrop,
+ .fxattrop = leases_fxattrop,
#endif
};
struct xlator_cbks cbks = {
- .forget = leases_forget,
- .release = leases_release,
- .client_disconnect = leases_clnt_disconnect_cbk,
+ .forget = leases_forget,
+ .release = leases_release,
+ .client_disconnect = leases_clnt_disconnect_cbk,
};
struct volume_options options[] = {
- { .key = {"leases"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .op_version = {GD_OP_VERSION_3_8_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .description = "When \"on\", enables leases support"
- },
- { .key = {"lease-lock-recall-timeout"},
- .type = GF_OPTION_TYPE_INT,
- .default_value = RECALL_LEASE_LK_TIMEOUT,
- .op_version = {GD_OP_VERSION_3_8_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .description = "After 'timeout' seconds since the recall_lease"
- " request has been sent to the client, the lease lock"
- " will be forcefully purged by the server."
- },
- { .key = {NULL} },
+ {.key = {"leases"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .op_version = {GD_OP_VERSION_3_8_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .description = "When \"on\", enables leases support"},
+ {.key = {"lease-lock-recall-timeout"},
+ .type = GF_OPTION_TYPE_INT,
+ .default_value = RECALL_LEASE_LK_TIMEOUT,
+ .op_version = {GD_OP_VERSION_3_8_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .description = "After 'timeout' seconds since the recall_lease"
+ " request has been sent to the client, the lease lock"
+ " will be forcefully purged by the server."},
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "leases",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/features/leases/src/leases.h b/xlators/features/leases/src/leases.h
index 80dbcbb06b9..a6e8a6824cc 100644
--- a/xlators/features/leases/src/leases.h
+++ b/xlators/features/leases/src/leases.h
@@ -16,15 +16,15 @@
#include "config.h"
#endif
-#include "common-utils.h"
-#include "glusterfs.h"
-#include "xlator.h"
-#include "call-stub.h"
-#include "logging.h"
-#include "client_t.h"
-#include "lkowner.h"
-#include "locking.h"
-#include "upcall-utils.h"
+#include <glusterfs/common-utils.h>
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/call-stub.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/client_t.h>
+#include <glusterfs/lkowner.h>
+#include <glusterfs/locking.h>
+#include <glusterfs/upcall-utils.h>
#include "timer-wheel.h"
#include "leases-mem-types.h"
#include "leases-messages.h"
@@ -39,24 +39,35 @@
#define BLOCK_FOP 0x0001
#define WIND_FOP 0x0002
-#define EXIT_IF_LEASES_OFF(this, label) do { \
+#define EXIT_IF_LEASES_OFF(this, label) \
+ do { \
if (!is_leases_enabled(this)) \
- goto label; \
-} while (0)
+ goto label; \
+ } while (0)
-#define GET_LEASE_ID(xdata, lease_id, client_uid) do { \
- int ret_val = -1; \
- ret_val = dict_get_bin (xdata, "lease-id", (void **)&lease_id); \
+#define EXIT_IF_INTERNAL_FOP(frame, xdata, label) \
+ do { \
+ if (frame->root->pid < 0) \
+ goto label; \
+ if (xdata && dict_get(xdata, GLUSTERFS_INTERNAL_FOP_KEY)) \
+ goto label; \
+ } while (0)
+
+#define GET_LEASE_ID(xdata, lease_id, client_uid) \
+ do { \
+ int ret_val = -1; \
+ ret_val = dict_get_bin(xdata, "lease-id", (void **)&lease_id); \
if (ret_val) { \
- ret_val = 0; \
- gf_msg_debug ("leases", 0, "Lease id is not set for client:%s", client_uid); \
+ ret_val = 0; \
+ gf_msg_debug("leases", 0, "Lease id is not set for client:%s", \
+ client_uid); \
} \
-} while (0)
+ } while (0)
#define GET_FLAGS(fop, fd_flags) \
-do { \
+ do { \
if ((fd_flags & (O_WRONLY | O_RDWR)) && fop == GF_FOP_OPEN) \
- fop_flags = DATA_MODIFY_FOP; \
+ fop_flags = DATA_MODIFY_FOP; \
\
if (fop == GF_FOP_UNLINK || fop == GF_FOP_RENAME || \
fop == GF_FOP_TRUNCATE || fop == GF_FOP_FTRUNCATE || \
@@ -65,185 +76,184 @@ do { \
fop == GF_FOP_DISCARD || fop == GF_FOP_ZEROFILL || \
fop == GF_FOP_SETATTR || fop == GF_FOP_FSETATTR || \
fop == GF_FOP_LINK) \
- fop_flags = DATA_MODIFY_FOP; \
+ fop_flags = DATA_MODIFY_FOP; \
\
if (!(fd_flags & (O_NONBLOCK | O_NDELAY))) \
- fop_flags |= BLOCKING_FOP; \
+ fop_flags |= BLOCKING_FOP; \
\
-} while (0) \
-
+ } while (0)
#define GET_FLAGS_LK(cmd, l_type, fd_flags) \
-do { \
+ do { \
/* TODO: handle F_RESLK_LCK and other glusterfs_lk_recovery_cmds_t */ \
- if ((cmd == F_SETLKW || cmd == F_SETLKW64 || \
- cmd == F_SETLK || cmd == F_SETLK64) && \
+ if ((cmd == F_SETLKW || cmd == F_SETLKW64 || cmd == F_SETLK || \
+ cmd == F_SETLK64) && \
l_type == F_WRLCK) \
- fop_flags = DATA_MODIFY_FOP; \
+ fop_flags = DATA_MODIFY_FOP; \
\
if (fd_flags & (O_NONBLOCK | O_NDELAY) && \
(cmd == F_SETLKW || cmd == F_SETLKW64)) \
- fop_flags |= BLOCKING_FOP; \
+ fop_flags |= BLOCKING_FOP; \
\
-} while (0) \
+ } while (0)
-#define LEASE_BLOCK_FOP(inode, fop_name, frame, this, params ...) \
-do { \
- call_stub_t *__stub = NULL; \
- fop_stub_t *blk_fop = NULL; \
- lease_inode_ctx_t *lease_ctx = NULL; \
+#define LEASE_BLOCK_FOP(inode, fop_name, frame, this, params...) \
+ do { \
+ call_stub_t *__stub = NULL; \
+ fop_stub_t *blk_fop = NULL; \
+ lease_inode_ctx_t *lease_ctx = NULL; \
\
- __stub = fop_##fop_name##_stub (frame, default_##fop_name##_resume, \
- params); \
+ __stub = fop_##fop_name##_stub(frame, default_##fop_name##_resume, \
+ params); \
if (!__stub) { \
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM, \
- LEASE_MSG_NO_MEM, \
- "Unable to create stub"); \
- ret = -ENOMEM; \
- goto __out; \
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, LEASE_MSG_NO_MEM, \
+ "Unable to create stub"); \
+ ret = -ENOMEM; \
+ goto __out; \
} \
\
- blk_fop = GF_CALLOC (1, sizeof (*blk_fop), \
- gf_leases_mt_fop_stub_t); \
+ blk_fop = GF_CALLOC(1, sizeof(*blk_fop), gf_leases_mt_fop_stub_t); \
if (!blk_fop) { \
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM, \
- LEASE_MSG_NO_MEM, \
- "Unable to create lease fop stub"); \
- ret = -ENOMEM; \
- goto __out; \
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, LEASE_MSG_NO_MEM, \
+ "Unable to create lease fop stub"); \
+ ret = -ENOMEM; \
+ goto __out; \
} \
\
- lease_ctx = lease_ctx_get (inode, this); \
+ lease_ctx = lease_ctx_get(inode, this); \
if (!lease_ctx) { \
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM, \
- LEASE_MSG_NO_MEM, \
- "Unable to create/get inode ctx"); \
- ret = -ENOMEM; \
- goto __out; \
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, LEASE_MSG_NO_MEM, \
+ "Unable to create/get inode ctx"); \
+ ret = -ENOMEM; \
+ goto __out; \
} \
\
blk_fop->stub = __stub; \
- pthread_mutex_lock (&lease_ctx->lock); \
+ pthread_mutex_lock(&lease_ctx->lock); \
{ \
- /*TODO: If the lease is unlocked btw check lease conflict and \
- * by now, then this fop shouldn't be add to the blocked fop \
- * list, can use generation number for the same?*/ \
- list_add_tail (&blk_fop->list, &lease_ctx->blocked_list); \
+ /*TODO: If the lease is unlocked btw check lease conflict and \
+ * by now, then this fop shouldn't be add to the blocked fop \
+ * list, can use generation number for the same?*/ \
+ list_add_tail(&blk_fop->list, &lease_ctx->blocked_list); \
} \
- pthread_mutex_unlock (&lease_ctx->lock); \
+ pthread_mutex_unlock(&lease_ctx->lock); \
\
-__out: \
+ __out: \
if (ret < 0) { \
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM, LEASE_MSG_NO_MEM, \
- "Unable to create stub for blocking the fop:%s (%s)", \
- gf_fop_list[frame->root->op], strerror(ENOMEM)); \
- if (__stub != NULL) { \
- call_stub_destroy (__stub); \
- } \
- GF_FREE (blk_fop); \
- goto err; \
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, LEASE_MSG_NO_MEM, \
+ "Unable to create stub for blocking the fop:%s (%s)", \
+ gf_fop_list[frame->root->op], strerror(ENOMEM)); \
+ if (__stub != NULL) { \
+ call_stub_destroy(__stub); \
+ } \
+ GF_FREE(blk_fop); \
+ goto err; \
} \
-} while (0) \
+ } while (0)
struct _leases_private {
- gf_boolean_t leases_enabled;
- int32_t recall_lease_timeout;
- struct list_head client_list;
- struct list_head recall_list;
- struct tvec_base *timer_wheel; /* timer wheel where the recall request
- is qued and waits for unlock/expiry */
- gf_boolean_t fini;
- pthread_t recall_thr;
- gf_boolean_t inited_recall_thr;
- pthread_mutex_t mutex;
- pthread_cond_t cond;
+ struct list_head client_list;
+ struct list_head recall_list;
+ struct tvec_base *timer_wheel; /* timer wheel where the recall request
+ is qued and waits for unlock/expiry */
+ pthread_t recall_thr;
+ pthread_mutex_t mutex;
+ pthread_cond_t cond;
+ int32_t recall_lease_timeout;
+ gf_boolean_t inited_recall_thr;
+ gf_boolean_t fini;
+ gf_boolean_t leases_enabled;
+
+ char _pad[1]; /* manual padding */
};
typedef struct _leases_private leases_private_t;
struct _lease_client {
- char *client_uid;
- struct list_head client_list;
- struct list_head inode_list;
+ char *client_uid;
+ struct list_head client_list;
+ struct list_head inode_list;
};
typedef struct _lease_client lease_client_t;
struct _lease_inode {
- inode_t *inode;
- struct list_head list; /* This can be part of both inode_list and recall_list */
+ inode_t *inode;
+ struct list_head
+ list; /* This can be part of both inode_list and recall_list */
};
typedef struct _lease_inode lease_inode_t;
struct _lease_fd_ctx {
- char *client_uid;
- char lease_id[LEASE_ID_SIZE];
+ char *client_uid;
+ char lease_id[LEASE_ID_SIZE];
};
typedef struct _lease_fd_ctx lease_fd_ctx_t;
struct _lease_inode_ctx {
- struct list_head lease_id_list; /* clients that have taken leases */
- int lease_type_cnt[GF_LEASE_MAX_TYPE+1];
- int lease_type; /* Types of leases acquired */
- uint64_t lease_cnt; /* Total number of leases on this inode */
- uint64_t openfd_cnt; /* number of fds open */
- gf_boolean_t recall_in_progress; /* if lease recall is sent on this inode */
- struct list_head blocked_list; /* List of fops blocked until the
- lease recall is complete */
- inode_t *inode; /* this represents the inode on which the
- lock was taken, required mainly during
- disconnect cleanup */
- struct gf_tw_timer_list *timer;
- pthread_mutex_t lock;
+ struct list_head lease_id_list; /* clients that have taken leases */
+ int lease_type_cnt[GF_LEASE_MAX_TYPE + 1];
+ uint64_t lease_cnt; /* Total number of leases on this inode */
+ uint64_t openfd_cnt; /* number of fds open */
+ struct list_head blocked_list; /* List of fops blocked until the
+ lease recall is complete */
+ inode_t *inode; /* this represents the inode on which the
+ lock was taken, required mainly during
+ disconnect cleanup */
+ struct gf_tw_timer_list *timer;
+ pthread_mutex_t lock;
+ int lease_type; /* Types of leases acquired */
+ gf_boolean_t recall_in_progress; /* if lease recall is sent on this inode */
+ gf_boolean_t blocked_fops_resuming; /* if blocked fops are being resumed */
+
+ char _pad[2]; /* manual padding */
};
typedef struct _lease_inode_ctx lease_inode_ctx_t;
struct _lease_id_entry {
- struct list_head lease_id_list;
- char lease_id[LEASE_ID_SIZE];
- char *client_uid; /* uid of the client that has
- taken the lease */
- int lease_type_cnt[GF_LEASE_MAX_TYPE+1]; /* count of each lease type */
- int lease_type; /* Union of all the leases taken
- under the given lease id */
- uint64_t lease_cnt; /* Number of leases taken under the
- given lease id */
- time_t recall_time; /* time @ which recall was sent */
+ struct list_head lease_id_list;
+ char lease_id[LEASE_ID_SIZE];
+ char *client_uid; /* uid of the client that has
+ taken the lease */
+ int lease_type_cnt[GF_LEASE_MAX_TYPE + 1]; /* count of each lease type */
+ uint64_t lease_cnt; /* Number of leases taken under the
+ given lease id */
+ time_t recall_time; /* time @ which recall was sent */
+ int lease_type; /* Union of all the leases taken
+ under the given lease id */
+ char _pad[4]; /* manual padding */
};
typedef struct _lease_id_entry lease_id_entry_t;
/* Required? as stub itself will have list */
struct __fop_stub {
- struct list_head list;
- call_stub_t *stub;
+ struct list_head list;
+ call_stub_t *stub;
};
typedef struct __fop_stub fop_stub_t;
struct __lease_timer_data {
- inode_t *inode;
- xlator_t *this;
+ inode_t *inode;
+ xlator_t *this;
};
typedef struct __lease_timer_data lease_timer_data_t;
gf_boolean_t
-is_leases_enabled (xlator_t *this);
-
-int32_t
-get_recall_lease_timeout (xlator_t *this);
+is_leases_enabled(xlator_t *this);
lease_inode_ctx_t *
-lease_ctx_get (inode_t *inode, xlator_t *this);
+lease_ctx_get(inode_t *inode, xlator_t *this);
int
-process_lease_req (call_frame_t *frame, xlator_t *this,
- inode_t *inode, struct gf_lease *lease);
+process_lease_req(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ struct gf_lease *lease);
int
-check_lease_conflict (call_frame_t *frame, inode_t *inode,
- const char *lease_id, uint32_t fop_flags);
+check_lease_conflict(call_frame_t *frame, inode_t *inode, const char *lease_id,
+ uint32_t fop_flags);
int
-cleanup_client_leases (xlator_t *this, const char *client_uid);
+cleanup_client_leases(xlator_t *this, const char *client_uid);
void *
-expired_recall_cleanup (void *data);
+expired_recall_cleanup(void *data);
#endif /* _LEASES_H */
diff --git a/xlators/features/locks/src/clear.c b/xlators/features/locks/src/clear.c
index 1609fc416d2..ab1eac68a53 100644
--- a/xlators/features/locks/src/clear.c
+++ b/xlators/features/locks/src/clear.c
@@ -12,440 +12,449 @@
#include <limits.h>
#include <pthread.h>
-#include "glusterfs.h"
-#include "compat.h"
-#include "xlator.h"
-#include "logging.h"
-#include "common-utils.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/compat.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/common-utils.h>
#include "locks.h"
#include "common.h"
-#include "statedump.h"
+#include <glusterfs/statedump.h>
#include "clear.h"
+const char *clrlk_type_names[CLRLK_TYPE_MAX] = {
+ [CLRLK_INODE] = "inode",
+ [CLRLK_ENTRY] = "entry",
+ [CLRLK_POSIX] = "posix",
+};
+
int
-clrlk_get_kind (char *kind)
+clrlk_get_kind(char *kind)
{
- char *clrlk_kinds[CLRLK_KIND_MAX] = {"dummy", "blocked", "granted",
- "all"};
- int ret_kind = CLRLK_KIND_MAX;
- int i = 0;
-
- for (i = CLRLK_BLOCKED; i < CLRLK_KIND_MAX; i++) {
- if (!strcmp (clrlk_kinds[i], kind)) {
- ret_kind = i;
- break;
- }
+ char *clrlk_kinds[CLRLK_KIND_MAX] = {"dummy", "blocked", "granted", "all"};
+ int ret_kind = CLRLK_KIND_MAX;
+ int i = 0;
+
+ for (i = CLRLK_BLOCKED; i < CLRLK_KIND_MAX; i++) {
+ if (!strcmp(clrlk_kinds[i], kind)) {
+ ret_kind = i;
+ break;
}
+ }
- return ret_kind;
+ return ret_kind;
}
int
-clrlk_get_type (char *type)
+clrlk_get_type(char *type)
{
- char *clrlk_types[CLRLK_TYPE_MAX] = {"inode", "entry", "posix"};
- int ret_type = CLRLK_TYPE_MAX;
- int i = 0;
-
- for (i = CLRLK_INODE; i < CLRLK_TYPE_MAX; i++) {
- if (!strcmp (clrlk_types[i], type)) {
- ret_type = i;
- break;
- }
+ char *clrlk_types[CLRLK_TYPE_MAX] = {"inode", "entry", "posix"};
+ int ret_type = CLRLK_TYPE_MAX;
+ int i = 0;
+
+ for (i = CLRLK_INODE; i < CLRLK_TYPE_MAX; i++) {
+ if (!strcmp(clrlk_types[i], type)) {
+ ret_type = i;
+ break;
}
+ }
- return ret_type;
+ return ret_type;
}
int
-clrlk_get_lock_range (char *range_str, struct gf_flock *ulock,
- gf_boolean_t *chk_range)
+clrlk_get_lock_range(char *range_str, struct gf_flock *ulock,
+ gf_boolean_t *chk_range)
{
- int ret = -1;
-
- if (!chk_range)
- goto out;
-
- if (!range_str) {
- ret = 0;
- *chk_range = _gf_false;
- goto out;
- }
+ int ret = -1;
- if (sscanf (range_str, "%hd,%"PRId64"-""%"PRId64, &ulock->l_whence,
- &ulock->l_start, &ulock->l_len) != 3) {
- goto out;
- }
+ if (!chk_range)
+ goto out;
+ if (!range_str) {
ret = 0;
- *chk_range = _gf_true;
+ *chk_range = _gf_false;
+ goto out;
+ }
+
+ if (sscanf(range_str,
+ "%hd,%" PRId64 "-"
+ "%" PRId64,
+ &ulock->l_whence, &ulock->l_start, &ulock->l_len) != 3) {
+ goto out;
+ }
+
+ ret = 0;
+ *chk_range = _gf_true;
out:
- return ret;
+ return ret;
}
int
-clrlk_parse_args (const char* cmd, clrlk_args *args)
+clrlk_parse_args(const char *cmd, clrlk_args *args)
{
- char *opts = NULL;
- char *cur = NULL;
- char *tok = NULL;
- char *sptr = NULL;
- char *free_ptr = NULL;
- char kw[KW_MAX] = {[KW_TYPE] = 't',
- [KW_KIND] = 'k',
- };
- int ret = -1;
- int i = 0;
-
- GF_ASSERT (cmd);
- free_ptr = opts = GF_CALLOC (1, strlen (cmd), gf_common_mt_char);
- if (!opts)
- goto out;
-
- if (sscanf (cmd, GF_XATTR_CLRLK_CMD".%s", opts) < 1) {
- ret = -1;
- goto out;
+ char *opts = NULL;
+ char *cur = NULL;
+ char *tok = NULL;
+ char *sptr = NULL;
+ char *free_ptr = NULL;
+ char kw[KW_MAX] = {
+ [KW_TYPE] = 't',
+ [KW_KIND] = 'k',
+ };
+ int ret = -1;
+ int i = 0;
+
+ GF_ASSERT(cmd);
+ free_ptr = opts = GF_CALLOC(1, strlen(cmd), gf_common_mt_char);
+ if (!opts)
+ goto out;
+
+ if (sscanf(cmd, GF_XATTR_CLRLK_CMD ".%s", opts) < 1) {
+ ret = -1;
+ goto out;
+ }
+
+ /*clr_lk_prefix.ttype.kkind.args, args - type specific*/
+ cur = opts;
+ for (i = 0; i < KW_MAX && (tok = strtok_r(cur, ".", &sptr));
+ cur = NULL, i++) {
+ if (tok[0] != kw[i]) {
+ ret = -1;
+ goto out;
}
-
- /*clr_lk_prefix.ttype.kkind.args, args - type specific*/
- cur = opts;
- for (i = 0; i < KW_MAX && (tok = strtok_r (cur, ".", &sptr));
- cur = NULL, i++) {
- if (tok[0] != kw[i]) {
- ret = -1;
- goto out;
- }
- if (i == KW_TYPE)
- args->type = clrlk_get_type (tok+1);
- if (i == KW_KIND)
- args->kind = clrlk_get_kind (tok+1);
- }
-
- if ((args->type == CLRLK_TYPE_MAX) || (args->kind == CLRLK_KIND_MAX))
- goto out;
-
- /*optional args, neither range nor basename can 'legally' contain
- * "/" in them*/
- tok = strtok_r (NULL, "/", &sptr);
- if (tok)
- args->opts = gf_strdup (tok);
-
- ret = 0;
+ if (i == KW_TYPE)
+ args->type = clrlk_get_type(tok + 1);
+ if (i == KW_KIND)
+ args->kind = clrlk_get_kind(tok + 1);
+ }
+
+ if ((args->type == CLRLK_TYPE_MAX) || (args->kind == CLRLK_KIND_MAX))
+ goto out;
+
+ /*optional args, neither range nor basename can 'legally' contain
+ * "/" in them*/
+ tok = strtok_r(NULL, "/", &sptr);
+ if (tok)
+ args->opts = gf_strdup(tok);
+
+ ret = 0;
out:
- GF_FREE (free_ptr);
- return ret;
+ GF_FREE(free_ptr);
+ return ret;
}
int
-clrlk_clear_posixlk (xlator_t *this, pl_inode_t *pl_inode, clrlk_args *args,
- int *blkd, int *granted, int *op_errno)
+clrlk_clear_posixlk(xlator_t *this, pl_inode_t *pl_inode, clrlk_args *args,
+ int *blkd, int *granted, int *op_errno)
{
- posix_lock_t *plock = NULL;
- posix_lock_t *tmp = NULL;
- struct gf_flock ulock = {0, };
- int ret = -1;
- int bcount = 0;
- int gcount = 0;
- gf_boolean_t chk_range = _gf_false;
-
- if (clrlk_get_lock_range (args->opts, &ulock, &chk_range)) {
- *op_errno = EINVAL;
- goto out;
- }
-
- pthread_mutex_lock (&pl_inode->mutex);
+ posix_lock_t *plock = NULL;
+ posix_lock_t *tmp = NULL;
+ struct gf_flock ulock = {
+ 0,
+ };
+ int ret = -1;
+ int bcount = 0;
+ int gcount = 0;
+ gf_boolean_t chk_range = _gf_false;
+
+ if (clrlk_get_lock_range(args->opts, &ulock, &chk_range)) {
+ *op_errno = EINVAL;
+ goto out;
+ }
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ list_for_each_entry_safe(plock, tmp, &pl_inode->ext_list, list)
{
- list_for_each_entry_safe (plock, tmp, &pl_inode->ext_list,
- list) {
- if ((plock->blocked &&
- !(args->kind & CLRLK_BLOCKED)) ||
- (!plock->blocked &&
- !(args->kind & CLRLK_GRANTED)))
- continue;
-
- if (chk_range &&
- (plock->user_flock.l_whence != ulock.l_whence
- || plock->user_flock.l_start != ulock.l_start
- || plock->user_flock.l_len != ulock.l_len))
- continue;
-
- list_del_init (&plock->list);
- if (plock->blocked) {
- bcount++;
- pl_trace_out (this, plock->frame, NULL, NULL,
- F_SETLKW, &plock->user_flock,
- -1, EAGAIN, NULL);
-
- STACK_UNWIND_STRICT (lk, plock->frame, -1, EAGAIN,
- &plock->user_flock, NULL);
-
- } else {
- gcount++;
- }
- __destroy_lock(plock);
- }
+ if ((plock->blocked && !(args->kind & CLRLK_BLOCKED)) ||
+ (!plock->blocked && !(args->kind & CLRLK_GRANTED)))
+ continue;
+
+ if (chk_range && (plock->user_flock.l_whence != ulock.l_whence ||
+ plock->user_flock.l_start != ulock.l_start ||
+ plock->user_flock.l_len != ulock.l_len))
+ continue;
+
+ list_del_init(&plock->list);
+ if (plock->blocked) {
+ bcount++;
+ pl_trace_out(this, plock->frame, NULL, NULL, F_SETLKW,
+ &plock->user_flock, -1, EINTR, NULL);
+
+ STACK_UNWIND_STRICT(lk, plock->frame, -1, EINTR,
+ &plock->user_flock, NULL);
+
+ } else {
+ gcount++;
+ }
+ __destroy_lock(plock);
}
- pthread_mutex_unlock (&pl_inode->mutex);
- grant_blocked_locks (this, pl_inode);
- ret = 0;
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+ grant_blocked_locks(this, pl_inode);
+ ret = 0;
out:
- *blkd = bcount;
- *granted = gcount;
- return ret;
+ *blkd = bcount;
+ *granted = gcount;
+ return ret;
}
/* Returns 0 on success and -1 on failure */
int
-clrlk_clear_inodelk (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom,
- clrlk_args *args, int *blkd, int *granted, int *op_errno)
+clrlk_clear_inodelk(xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom,
+ clrlk_args *args, int *blkd, int *granted, int *op_errno)
{
- posix_locks_private_t *priv;
- pl_inode_lock_t *ilock = NULL;
- pl_inode_lock_t *tmp = NULL;
- struct gf_flock ulock = {0, };
- int ret = -1;
- int bcount = 0;
- int gcount = 0;
- gf_boolean_t chk_range = _gf_false;
- struct list_head *pcontend = NULL;
- struct list_head released;
- struct list_head contend;
- struct timespec now = { };
-
- INIT_LIST_HEAD (&released);
-
- priv = this->private;
- if (priv->notify_contention) {
- pcontend = &contend;
- INIT_LIST_HEAD (pcontend);
- timespec_now(&now);
- }
-
- if (clrlk_get_lock_range (args->opts, &ulock, &chk_range)) {
- *op_errno = EINVAL;
- goto out;
- }
-
- if (args->kind & CLRLK_BLOCKED)
- goto blkd;
-
- if (args->kind & CLRLK_GRANTED)
- goto granted;
+ posix_locks_private_t *priv;
+ pl_inode_lock_t *ilock = NULL;
+ pl_inode_lock_t *tmp = NULL;
+ struct gf_flock ulock = {
+ 0,
+ };
+ int ret = -1;
+ int bcount = 0;
+ int gcount = 0;
+ gf_boolean_t chk_range = _gf_false;
+ struct list_head *pcontend = NULL;
+ struct list_head released;
+ struct list_head contend;
+ struct timespec now = {};
+
+ INIT_LIST_HEAD(&released);
+
+ priv = this->private;
+ if (priv->notify_contention) {
+ pcontend = &contend;
+ INIT_LIST_HEAD(pcontend);
+ timespec_now(&now);
+ }
+
+ if (clrlk_get_lock_range(args->opts, &ulock, &chk_range)) {
+ *op_errno = EINVAL;
+ goto out;
+ }
+
+ if (args->kind & CLRLK_BLOCKED)
+ goto blkd;
+
+ if (args->kind & CLRLK_GRANTED)
+ goto granted;
blkd:
- pthread_mutex_lock (&pl_inode->mutex);
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ list_for_each_entry_safe(ilock, tmp, &dom->blocked_inodelks,
+ blocked_locks)
{
- list_for_each_entry_safe (ilock, tmp, &dom->blocked_inodelks,
- blocked_locks) {
- if (chk_range &&
- (ilock->user_flock.l_whence != ulock.l_whence
- || ilock->user_flock.l_start != ulock.l_start
- || ilock->user_flock.l_len != ulock.l_len))
- continue;
-
- bcount++;
- list_del_init (&ilock->client_list);
- list_del_init (&ilock->blocked_locks);
- list_add (&ilock->blocked_locks, &released);
- }
- }
- pthread_mutex_unlock (&pl_inode->mutex);
-
- list_for_each_entry_safe (ilock, tmp, &released, blocked_locks) {
- list_del_init (&ilock->blocked_locks);
- pl_trace_out (this, ilock->frame, NULL, NULL, F_SETLKW,
- &ilock->user_flock, -1, EAGAIN,
- ilock->volume);
- STACK_UNWIND_STRICT (inodelk, ilock->frame, -1,
- EAGAIN, NULL);
- //No need to take lock as the locks are only in one list
- __pl_inodelk_unref (ilock);
+ if (chk_range && (ilock->user_flock.l_whence != ulock.l_whence ||
+ ilock->user_flock.l_start != ulock.l_start ||
+ ilock->user_flock.l_len != ulock.l_len))
+ continue;
+
+ bcount++;
+ list_del_init(&ilock->client_list);
+ list_del_init(&ilock->blocked_locks);
+ list_add(&ilock->blocked_locks, &released);
}
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
- if (!(args->kind & CLRLK_GRANTED)) {
- ret = 0;
- goto out;
+ if (!list_empty(&released)) {
+ list_for_each_entry_safe(ilock, tmp, &released, blocked_locks)
+ {
+ list_del_init(&ilock->blocked_locks);
+ pl_trace_out(this, ilock->frame, NULL, NULL, F_SETLKW,
+ &ilock->user_flock, -1, EAGAIN, ilock->volume);
+ STACK_UNWIND_STRICT(inodelk, ilock->frame, -1, EAGAIN, NULL);
+ // No need to take lock as the locks are only in one list
+ __pl_inodelk_unref(ilock);
}
+ }
+
+ if (!(args->kind & CLRLK_GRANTED)) {
+ ret = 0;
+ goto out;
+ }
granted:
- pthread_mutex_lock (&pl_inode->mutex);
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ list_for_each_entry_safe(ilock, tmp, &dom->inodelk_list, list)
{
- list_for_each_entry_safe (ilock, tmp, &dom->inodelk_list,
- list) {
- if (chk_range &&
- (ilock->user_flock.l_whence != ulock.l_whence
- || ilock->user_flock.l_start != ulock.l_start
- || ilock->user_flock.l_len != ulock.l_len))
- continue;
-
- gcount++;
- list_del_init (&ilock->client_list);
- list_del_init (&ilock->list);
- list_add (&ilock->list, &released);
- }
+ if (chk_range && (ilock->user_flock.l_whence != ulock.l_whence ||
+ ilock->user_flock.l_start != ulock.l_start ||
+ ilock->user_flock.l_len != ulock.l_len))
+ continue;
+
+ gcount++;
+ list_del_init(&ilock->client_list);
+ list_del_init(&ilock->list);
+ list_add(&ilock->list, &released);
}
- pthread_mutex_unlock (&pl_inode->mutex);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
- list_for_each_entry_safe (ilock, tmp, &released, list) {
- list_del_init (&ilock->list);
- //No need to take lock as the locks are only in one list
- __pl_inodelk_unref (ilock);
- }
+ list_for_each_entry_safe(ilock, tmp, &released, list)
+ {
+ list_del_init(&ilock->list);
+ // No need to take lock as the locks are only in one list
+ __pl_inodelk_unref(ilock);
+ }
- ret = 0;
+ ret = 0;
out:
- grant_blocked_inode_locks (this, pl_inode, dom, &now, pcontend);
- if (pcontend != NULL) {
- inodelk_contention_notify(this, pcontend);
- }
- *blkd = bcount;
- *granted = gcount;
- return ret;
+ grant_blocked_inode_locks(this, pl_inode, dom, &now, pcontend);
+ if (pcontend != NULL) {
+ inodelk_contention_notify(this, pcontend);
+ }
+ *blkd = bcount;
+ *granted = gcount;
+ return ret;
}
/* Returns 0 on success and -1 on failure */
int
-clrlk_clear_entrylk (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom,
- clrlk_args *args, int *blkd, int *granted, int *op_errno)
+clrlk_clear_entrylk(xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom,
+ clrlk_args *args, int *blkd, int *granted, int *op_errno)
{
- posix_locks_private_t *priv;
- pl_entry_lock_t *elock = NULL;
- pl_entry_lock_t *tmp = NULL;
- int bcount = 0;
- int gcount = 0;
- int ret = -1;
- struct list_head *pcontend = NULL;
- struct list_head removed;
- struct list_head released;
- struct list_head contend;
- struct timespec now;
-
- INIT_LIST_HEAD (&released);
-
- priv = this->private;
- if (priv->notify_contention) {
- pcontend = &contend;
- INIT_LIST_HEAD (pcontend);
- timespec_now(&now);
- }
-
- if (args->kind & CLRLK_BLOCKED)
- goto blkd;
-
- if (args->kind & CLRLK_GRANTED)
- goto granted;
+ posix_locks_private_t *priv;
+ pl_entry_lock_t *elock = NULL;
+ pl_entry_lock_t *tmp = NULL;
+ int bcount = 0;
+ int gcount = 0;
+ int ret = -1;
+ struct list_head *pcontend = NULL;
+ struct list_head removed;
+ struct list_head released;
+ struct list_head contend;
+ struct timespec now;
+
+ INIT_LIST_HEAD(&released);
+
+ priv = this->private;
+ if (priv->notify_contention) {
+ pcontend = &contend;
+ INIT_LIST_HEAD(pcontend);
+ timespec_now(&now);
+ }
+
+ if (args->kind & CLRLK_BLOCKED)
+ goto blkd;
+
+ if (args->kind & CLRLK_GRANTED)
+ goto granted;
blkd:
- pthread_mutex_lock (&pl_inode->mutex);
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ list_for_each_entry_safe(elock, tmp, &dom->blocked_entrylks,
+ blocked_locks)
{
- list_for_each_entry_safe (elock, tmp, &dom->blocked_entrylks,
- blocked_locks) {
- if (args->opts) {
- if (!elock->basename ||
- strcmp (elock->basename, args->opts))
- continue;
- }
-
- bcount++;
-
- list_del_init (&elock->client_list);
- list_del_init (&elock->blocked_locks);
- list_add_tail (&elock->blocked_locks, &released);
- }
- }
- pthread_mutex_unlock (&pl_inode->mutex);
+ if (args->opts) {
+ if (!elock->basename || strcmp(elock->basename, args->opts))
+ continue;
+ }
- list_for_each_entry_safe (elock, tmp, &released, blocked_locks) {
- list_del_init (&elock->blocked_locks);
- entrylk_trace_out (this, elock->frame, elock->volume, NULL, NULL,
- elock->basename, ENTRYLK_LOCK, elock->type,
- -1, EAGAIN);
- STACK_UNWIND_STRICT (entrylk, elock->frame, -1, EAGAIN, NULL);
+ bcount++;
- __pl_entrylk_unref (elock);
+ list_del_init(&elock->client_list);
+ list_del_init(&elock->blocked_locks);
+ list_add_tail(&elock->blocked_locks, &released);
}
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+
+ if (!list_empty(&released)) {
+ list_for_each_entry_safe(elock, tmp, &released, blocked_locks)
+ {
+ list_del_init(&elock->blocked_locks);
+ entrylk_trace_out(this, elock->frame, elock->volume, NULL, NULL,
+ elock->basename, ENTRYLK_LOCK, elock->type, -1,
+ EAGAIN);
+ STACK_UNWIND_STRICT(entrylk, elock->frame, -1, EAGAIN, NULL);
- if (!(args->kind & CLRLK_GRANTED)) {
- ret = 0;
- goto out;
+ __pl_entrylk_unref(elock);
}
+ }
+
+ if (!(args->kind & CLRLK_GRANTED)) {
+ ret = 0;
+ goto out;
+ }
granted:
- INIT_LIST_HEAD (&removed);
- pthread_mutex_lock (&pl_inode->mutex);
+ INIT_LIST_HEAD(&removed);
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ list_for_each_entry_safe(elock, tmp, &dom->entrylk_list, domain_list)
{
- list_for_each_entry_safe (elock, tmp, &dom->entrylk_list,
- domain_list) {
- if (args->opts) {
- if (!elock->basename ||
- strcmp (elock->basename, args->opts))
- continue;
- }
-
- gcount++;
- list_del_init (&elock->client_list);
- list_del_init (&elock->domain_list);
- list_add_tail (&elock->domain_list, &removed);
-
- __pl_entrylk_unref (elock);
- }
- }
- pthread_mutex_unlock (&pl_inode->mutex);
+ if (args->opts) {
+ if (!elock->basename || strcmp(elock->basename, args->opts))
+ continue;
+ }
- grant_blocked_entry_locks (this, pl_inode, dom, &now, pcontend);
- if (pcontend != NULL) {
- entrylk_contention_notify(this, pcontend);
+ gcount++;
+ list_del_init(&elock->client_list);
+ list_del_init(&elock->domain_list);
+ list_add_tail(&elock->domain_list, &removed);
+
+ __pl_entrylk_unref(elock);
}
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
- ret = 0;
+ grant_blocked_entry_locks(this, pl_inode, dom, &now, pcontend);
+ if (pcontend != NULL) {
+ entrylk_contention_notify(this, pcontend);
+ }
+
+ ret = 0;
out:
- *blkd = bcount;
- *granted = gcount;
- return ret;
+ *blkd = bcount;
+ *granted = gcount;
+ return ret;
}
int
-clrlk_clear_lks_in_all_domains (xlator_t *this, pl_inode_t *pl_inode,
- clrlk_args *args, int *blkd, int *granted,
- int *op_errno)
+clrlk_clear_lks_in_all_domains(xlator_t *this, pl_inode_t *pl_inode,
+ clrlk_args *args, int *blkd, int *granted,
+ int *op_errno)
{
- pl_dom_list_t *dom = NULL;
- int ret = -1;
- int tmp_bcount = 0;
- int tmp_gcount = 0;
-
- if (list_empty (&pl_inode->dom_list)) {
- ret = 0;
- goto out;
- }
+ pl_dom_list_t *dom = NULL;
+ int ret = -1;
+ int tmp_bcount = 0;
+ int tmp_gcount = 0;
- list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
- tmp_bcount = tmp_gcount = 0;
-
- switch (args->type)
- {
- case CLRLK_INODE:
- ret = clrlk_clear_inodelk (this, pl_inode, dom, args,
- &tmp_bcount, &tmp_gcount,
- op_errno);
- if (ret)
- goto out;
- break;
- case CLRLK_ENTRY:
- ret = clrlk_clear_entrylk (this, pl_inode, dom, args,
- &tmp_bcount, &tmp_gcount,
- op_errno);
- if (ret)
- goto out;
- break;
- }
-
- *blkd += tmp_bcount;
- *granted += tmp_gcount;
+ if (list_empty(&pl_inode->dom_list)) {
+ ret = 0;
+ goto out;
+ }
+
+ list_for_each_entry(dom, &pl_inode->dom_list, inode_list)
+ {
+ tmp_bcount = tmp_gcount = 0;
+
+ switch (args->type) {
+ case CLRLK_INODE:
+ ret = clrlk_clear_inodelk(this, pl_inode, dom, args,
+ &tmp_bcount, &tmp_gcount, op_errno);
+ if (ret)
+ goto out;
+ break;
+ case CLRLK_ENTRY:
+ ret = clrlk_clear_entrylk(this, pl_inode, dom, args,
+ &tmp_bcount, &tmp_gcount, op_errno);
+ if (ret)
+ goto out;
+ break;
}
- ret = 0;
+ *blkd += tmp_bcount;
+ *granted += tmp_gcount;
+ }
+
+ ret = 0;
out:
- return ret;
+ return ret;
}
diff --git a/xlators/features/locks/src/clear.h b/xlators/features/locks/src/clear.h
index 78fc5ae3398..bc118cb1b81 100644
--- a/xlators/features/locks/src/clear.h
+++ b/xlators/features/locks/src/clear.h
@@ -10,62 +10,64 @@
#ifndef __CLEAR_H__
#define __CLEAR_H__
-#include "compat-errno.h"
-#include "stack.h"
-#include "call-stub.h"
+#include <glusterfs/compat-errno.h>
+#include <glusterfs/stack.h>
+#include <glusterfs/call-stub.h>
#include "locks.h"
typedef enum {
- CLRLK_INODE,
- CLRLK_ENTRY,
- CLRLK_POSIX,
- CLRLK_TYPE_MAX
+ CLRLK_INODE,
+ CLRLK_ENTRY,
+ CLRLK_POSIX,
+ CLRLK_TYPE_MAX
} clrlk_type;
+extern const char *clrlk_type_names[];
+
typedef enum {
- CLRLK_BLOCKED = 1,
- CLRLK_GRANTED,
- CLRLK_ALL,
- CLRLK_KIND_MAX
+ CLRLK_BLOCKED = 1,
+ CLRLK_GRANTED,
+ CLRLK_ALL,
+ CLRLK_KIND_MAX
} clrlk_kind;
typedef enum {
- KW_TYPE,
- KW_KIND,
- /*add new keywords here*/
- KW_MAX
+ KW_TYPE,
+ KW_KIND,
+ /*add new keywords here*/
+ KW_MAX
} clrlk_opts;
struct _clrlk_args;
typedef struct _clrlk_args clrlk_args;
struct _clrlk_args {
- int type;
- int kind;
- char *opts;
+ int type;
+ int kind;
+ char *opts;
};
int
-clrlk_get__kind (char *kind);
+clrlk_get__kind(char *kind);
int
-clrlk_get_type (char *type);
+clrlk_get_type(char *type);
int
-clrlk_get_lock_range (char *range_str, struct gf_flock *ulock,
- gf_boolean_t *chk_range);
+clrlk_get_lock_range(char *range_str, struct gf_flock *ulock,
+ gf_boolean_t *chk_range);
int
-clrlk_parse_args (const char* cmd, clrlk_args *args);
+clrlk_parse_args(const char *cmd, clrlk_args *args);
int
-clrlk_clear_posixlk (xlator_t *this, pl_inode_t *pl_inode, clrlk_args *args,
- int *blkd, int *granted, int *op_errno);
+clrlk_clear_posixlk(xlator_t *this, pl_inode_t *pl_inode, clrlk_args *args,
+ int *blkd, int *granted, int *op_errno);
int
-clrlk_clear_inodelk (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom,
- clrlk_args *args, int *blkd, int *granted, int *op_errno);
+clrlk_clear_inodelk(xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom,
+ clrlk_args *args, int *blkd, int *granted, int *op_errno);
int
-clrlk_clear_entrylk (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom,
- clrlk_args *args, int *blkd, int *granted, int *op_errno);
+clrlk_clear_entrylk(xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom,
+ clrlk_args *args, int *blkd, int *granted, int *op_errno);
int
-clrlk_clear_lks_in_all_domains (xlator_t *this, pl_inode_t *pl_inode,
- clrlk_args *args, int *blkd, int *granted,
- int *op_errno);
+clrlk_clear_lks_in_all_domains(xlator_t *this, pl_inode_t *pl_inode,
+ clrlk_args *args, int *blkd, int *granted,
+ int *op_errno);
#endif /* __CLEAR_H__ */
diff --git a/xlators/features/locks/src/common.c b/xlators/features/locks/src/common.c
index c22dff08382..a2c6be93e03 100644
--- a/xlators/features/locks/src/common.c
+++ b/xlators/features/locks/src/common.c
@@ -12,737 +12,775 @@
#include <limits.h>
#include <pthread.h>
-#include "glusterfs.h"
-#include "compat.h"
-#include "xlator.h"
-#include "logging.h"
-#include "common-utils.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/compat.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/syncop.h>
#include "locks.h"
#include "common.h"
-
static int
-__is_lock_grantable (pl_inode_t *pl_inode, posix_lock_t *lock);
+__is_lock_grantable(pl_inode_t *pl_inode, posix_lock_t *lock);
static void
-__insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock);
+__insert_and_merge(pl_inode_t *pl_inode, posix_lock_t *lock);
static int
-pl_send_prelock_unlock (xlator_t *this, pl_inode_t *pl_inode,
- posix_lock_t *old_lock);
+pl_send_prelock_unlock(xlator_t *this, pl_inode_t *pl_inode,
+ posix_lock_t *old_lock);
static pl_dom_list_t *
-__allocate_domain (const char *volume)
+__allocate_domain(const char *volume)
{
- pl_dom_list_t *dom = NULL;
+ pl_dom_list_t *dom = NULL;
- dom = GF_CALLOC (1, sizeof (*dom),
- gf_locks_mt_pl_dom_list_t);
- if (!dom)
- goto out;
+ dom = GF_CALLOC(1, sizeof(*dom), gf_locks_mt_pl_dom_list_t);
+ if (!dom)
+ goto out;
- dom->domain = gf_strdup(volume);
- if (!dom->domain)
- goto out;
+ dom->domain = gf_strdup(volume);
+ if (!dom->domain)
+ goto out;
- gf_log ("posix-locks", GF_LOG_TRACE,
- "New domain allocated: %s", dom->domain);
+ gf_log("posix-locks", GF_LOG_TRACE, "New domain allocated: %s",
+ dom->domain);
- INIT_LIST_HEAD (&dom->inode_list);
- INIT_LIST_HEAD (&dom->entrylk_list);
- INIT_LIST_HEAD (&dom->blocked_entrylks);
- INIT_LIST_HEAD (&dom->inodelk_list);
- INIT_LIST_HEAD (&dom->blocked_inodelks);
+ INIT_LIST_HEAD(&dom->inode_list);
+ INIT_LIST_HEAD(&dom->entrylk_list);
+ INIT_LIST_HEAD(&dom->blocked_entrylks);
+ INIT_LIST_HEAD(&dom->inodelk_list);
+ INIT_LIST_HEAD(&dom->blocked_inodelks);
out:
- if (dom && (NULL == dom->domain)) {
- GF_FREE (dom);
- dom = NULL;
- }
+ if (dom && (NULL == dom->domain)) {
+ GF_FREE(dom);
+ dom = NULL;
+ }
- return dom;
+ return dom;
}
/* Returns domain for the lock. If domain is not present,
* allocates a domain and returns it
*/
pl_dom_list_t *
-get_domain (pl_inode_t *pl_inode, const char *volume)
+get_domain(pl_inode_t *pl_inode, const char *volume)
{
- pl_dom_list_t *dom = NULL;
+ pl_dom_list_t *dom = NULL;
- GF_VALIDATE_OR_GOTO ("posix-locks", pl_inode, out);
- GF_VALIDATE_OR_GOTO ("posix-locks", volume, out);
+ GF_VALIDATE_OR_GOTO("posix-locks", pl_inode, out);
+ GF_VALIDATE_OR_GOTO("posix-locks", volume, out);
- pthread_mutex_lock (&pl_inode->mutex);
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ list_for_each_entry(dom, &pl_inode->dom_list, inode_list)
{
- list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
- if (strcmp (dom->domain, volume) == 0)
- goto unlock;
- }
-
- dom = __allocate_domain (volume);
- if (dom)
- list_add (&dom->inode_list, &pl_inode->dom_list);
+ if (strcmp(dom->domain, volume) == 0)
+ goto unlock;
}
+
+ dom = __allocate_domain(volume);
+ if (dom)
+ list_add(&dom->inode_list, &pl_inode->dom_list);
+ }
unlock:
- pthread_mutex_unlock (&pl_inode->mutex);
- if (dom) {
- gf_log ("posix-locks", GF_LOG_TRACE, "Domain %s found", volume);
- } else {
- gf_log ("posix-locks", GF_LOG_TRACE, "Domain %s not found", volume);
- }
+ pthread_mutex_unlock(&pl_inode->mutex);
+ if (dom) {
+ gf_log("posix-locks", GF_LOG_TRACE, "Domain %s found", volume);
+ } else {
+ gf_log("posix-locks", GF_LOG_TRACE, "Domain %s not found", volume);
+ }
out:
- return dom;
+ return dom;
}
unsigned long
-fd_to_fdnum (fd_t *fd)
+fd_to_fdnum(fd_t *fd)
{
- return ((unsigned long) fd);
+ return ((unsigned long)fd);
}
fd_t *
-fd_from_fdnum (posix_lock_t *lock)
+fd_from_fdnum(posix_lock_t *lock)
{
- return ((fd_t *) lock->fd_num);
+ return ((fd_t *)lock->fd_num);
}
int
-__pl_inode_is_empty (pl_inode_t *pl_inode)
+__pl_inode_is_empty(pl_inode_t *pl_inode)
{
- return (list_empty (&pl_inode->ext_list));
+ return (list_empty(&pl_inode->ext_list));
}
void
-pl_print_locker (char *str, int size, xlator_t *this, call_frame_t *frame)
+pl_print_locker(char *str, int size, xlator_t *this, call_frame_t *frame)
{
- snprintf (str, size, "Pid=%llu, lk-owner=%s, Client=%p, Frame=%llu",
- (unsigned long long) frame->root->pid,
- lkowner_utoa (&frame->root->lk_owner),
- frame->root->client,
- (unsigned long long) frame->root->unique);
+ snprintf(str, size, "Pid=%llu, lk-owner=%s, Client=%p, Frame=%llu",
+ (unsigned long long)frame->root->pid,
+ lkowner_utoa(&frame->root->lk_owner), frame->root->client,
+ (unsigned long long)frame->root->unique);
}
-
void
-pl_print_lockee (char *str, int size, fd_t *fd, loc_t *loc)
+pl_print_lockee(char *str, int size, fd_t *fd, loc_t *loc)
{
- inode_t *inode = NULL;
- char *ipath = NULL;
- int ret = 0;
+ inode_t *inode = NULL;
+ char *ipath = NULL;
+ int ret = 0;
- if (fd)
- inode = fd->inode;
- if (loc)
- inode = loc->inode;
+ if (fd)
+ inode = fd->inode;
+ if (loc)
+ inode = loc->inode;
- if (!inode) {
- snprintf (str, size, "<nul>");
- return;
- }
+ if (!inode) {
+ snprintf(str, size, "<nul>");
+ return;
+ }
- if (loc && loc->path) {
- ipath = gf_strdup (loc->path);
- } else {
- ret = inode_path (inode, NULL, &ipath);
- if (ret <= 0)
- ipath = NULL;
- }
+ if (loc && loc->path) {
+ ipath = gf_strdup(loc->path);
+ } else {
+ ret = inode_path(inode, NULL, &ipath);
+ if (ret <= 0)
+ ipath = NULL;
+ }
- snprintf (str, size, "gfid=%s, fd=%p, path=%s",
- uuid_utoa (inode->gfid), fd,
- ipath ? ipath : "<nul>");
+ snprintf(str, size, "gfid=%s, fd=%p, path=%s", uuid_utoa(inode->gfid), fd,
+ ipath ? ipath : "<nul>");
- GF_FREE (ipath);
+ GF_FREE(ipath);
}
-
void
-pl_print_lock (char *str, int size, int cmd,
- struct gf_flock *flock, gf_lkowner_t *owner)
+pl_print_lock(char *str, int size, int cmd, struct gf_flock *flock,
+ gf_lkowner_t *owner)
{
- char *cmd_str = NULL;
- char *type_str = NULL;
+ char *cmd_str = NULL;
+ char *type_str = NULL;
- switch (cmd) {
+ switch (cmd) {
#if F_GETLK != F_GETLK64
case F_GETLK64:
#endif
case F_GETLK:
- cmd_str = "GETLK";
- break;
+ cmd_str = "GETLK";
+ break;
#if F_SETLK != F_SETLK64
case F_SETLK64:
#endif
case F_SETLK:
- cmd_str = "SETLK";
- break;
+ cmd_str = "SETLK";
+ break;
#if F_SETLKW != F_SETLKW64
case F_SETLKW64:
#endif
case F_SETLKW:
- cmd_str = "SETLKW";
- break;
+ cmd_str = "SETLKW";
+ break;
default:
- cmd_str = "UNKNOWN";
- break;
- }
+ cmd_str = "UNKNOWN";
+ break;
+ }
- switch (flock->l_type) {
+ switch (flock->l_type) {
case F_RDLCK:
- type_str = "READ";
- break;
+ type_str = "READ";
+ break;
case F_WRLCK:
- type_str = "WRITE";
- break;
+ type_str = "WRITE";
+ break;
case F_UNLCK:
- type_str = "UNLOCK";
- break;
+ type_str = "UNLOCK";
+ break;
default:
- type_str = "UNKNOWN";
- break;
- }
-
- snprintf (str, size, "lock=FCNTL, cmd=%s, type=%s, "
- "start=%llu, len=%llu, pid=%llu, lk-owner=%s",
- cmd_str, type_str, (unsigned long long) flock->l_start,
- (unsigned long long) flock->l_len,
- (unsigned long long) flock->l_pid,
- lkowner_utoa (owner));
+ type_str = "UNKNOWN";
+ break;
+ }
+
+ snprintf(str, size,
+ "lock=FCNTL, cmd=%s, type=%s, "
+ "start=%llu, len=%llu, pid=%llu, lk-owner=%s",
+ cmd_str, type_str, (unsigned long long)flock->l_start,
+ (unsigned long long)flock->l_len, (unsigned long long)flock->l_pid,
+ lkowner_utoa(owner));
}
-
void
-pl_trace_in (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
- int cmd, struct gf_flock *flock, const char *domain)
+pl_trace_in(xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc, int cmd,
+ struct gf_flock *flock, const char *domain)
{
- posix_locks_private_t *priv = NULL;
- char pl_locker[256];
- char pl_lockee[256];
- char pl_lock[256];
-
- priv = this->private;
+ posix_locks_private_t *priv = this->private;
+ char pl_locker[256];
+ char pl_lockee[256];
+ char pl_lock[256];
- if (!priv->trace)
- return;
+ if (!priv->trace)
+ return;
- pl_print_locker (pl_locker, 256, this, frame);
- pl_print_lockee (pl_lockee, 256, fd, loc);
- if (domain)
- pl_print_inodelk (pl_lock, 256, cmd, flock, domain);
- else
- pl_print_lock (pl_lock, 256, cmd, flock, &frame->root->lk_owner);
+ pl_print_locker(pl_locker, 256, this, frame);
+ pl_print_lockee(pl_lockee, 256, fd, loc);
+ if (domain)
+ pl_print_inodelk(pl_lock, 256, cmd, flock, domain);
+ else
+ pl_print_lock(pl_lock, 256, cmd, flock, &frame->root->lk_owner);
- gf_log (this->name, GF_LOG_INFO,
- "[REQUEST] Locker = {%s} Lockee = {%s} Lock = {%s}",
- pl_locker, pl_lockee, pl_lock);
+ gf_log(this->name, GF_LOG_INFO,
+ "[REQUEST] Locker = {%s} Lockee = {%s} Lock = {%s}", pl_locker,
+ pl_lockee, pl_lock);
}
-
void
-pl_print_verdict (char *str, int size, int op_ret, int op_errno)
+pl_print_verdict(char *str, int size, int op_ret, int op_errno)
{
- char *verdict = NULL;
-
- if (op_ret == 0) {
- verdict = "GRANTED";
- } else {
- switch (op_errno) {
- case EAGAIN:
- verdict = "TRYAGAIN";
- break;
- default:
- verdict = strerror (op_errno);
- }
+ char *verdict = NULL;
+
+ if (op_ret == 0) {
+ verdict = "GRANTED";
+ } else {
+ switch (op_errno) {
+ case EAGAIN:
+ verdict = "TRYAGAIN";
+ break;
+ default:
+ verdict = strerror(op_errno);
}
+ }
- snprintf (str, size, "%s", verdict);
+ snprintf(str, size, "%s", verdict);
}
-
void
-pl_trace_out (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
- int cmd, struct gf_flock *flock, int op_ret, int op_errno, const char *domain)
+pl_trace_out(xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc, int cmd,
+ struct gf_flock *flock, int op_ret, int op_errno,
+ const char *domain)
{
- posix_locks_private_t *priv = NULL;
- char pl_locker[256];
- char pl_lockee[256];
- char pl_lock[256];
- char verdict[32];
+ posix_locks_private_t *priv = NULL;
+ char pl_locker[256];
+ char pl_lockee[256];
+ char pl_lock[256];
+ char verdict[32];
- priv = this->private;
+ priv = this->private;
- if (!priv->trace)
- return;
+ if (!priv->trace)
+ return;
- pl_print_locker (pl_locker, 256, this, frame);
- pl_print_lockee (pl_lockee, 256, fd, loc);
- if (domain)
- pl_print_inodelk (pl_lock, 256, cmd, flock, domain);
- else
- pl_print_lock (pl_lock, 256, cmd, flock, &frame->root->lk_owner);
+ pl_print_locker(pl_locker, 256, this, frame);
+ pl_print_lockee(pl_lockee, 256, fd, loc);
+ if (domain)
+ pl_print_inodelk(pl_lock, 256, cmd, flock, domain);
+ else
+ pl_print_lock(pl_lock, 256, cmd, flock, &frame->root->lk_owner);
- pl_print_verdict (verdict, 32, op_ret, op_errno);
+ pl_print_verdict(verdict, 32, op_ret, op_errno);
- gf_log (this->name, GF_LOG_INFO,
- "[%s] Locker = {%s} Lockee = {%s} Lock = {%s}",
- verdict, pl_locker, pl_lockee, pl_lock);
+ gf_log(this->name, GF_LOG_INFO,
+ "[%s] Locker = {%s} Lockee = {%s} Lock = {%s}", verdict, pl_locker,
+ pl_lockee, pl_lock);
}
-
void
-pl_trace_block (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
- int cmd, struct gf_flock *flock, const char *domain)
+pl_trace_block(xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
+ int cmd, struct gf_flock *flock, const char *domain)
{
- posix_locks_private_t *priv = NULL;
- char pl_locker[256];
- char pl_lockee[256];
- char pl_lock[256];
+ posix_locks_private_t *priv = this->private;
+ char pl_locker[256];
+ char pl_lockee[256];
+ char pl_lock[256];
- priv = this->private;
-
- if (!priv->trace)
- return;
+ if (!priv->trace)
+ return;
- pl_print_locker (pl_locker, 256, this, frame);
- pl_print_lockee (pl_lockee, 256, fd, loc);
- if (domain)
- pl_print_inodelk (pl_lock, 256, cmd, flock, domain);
- else
- pl_print_lock (pl_lock, 256, cmd, flock, &frame->root->lk_owner);
+ pl_print_locker(pl_locker, 256, this, frame);
+ pl_print_lockee(pl_lockee, 256, fd, loc);
+ if (domain)
+ pl_print_inodelk(pl_lock, 256, cmd, flock, domain);
+ else
+ pl_print_lock(pl_lock, 256, cmd, flock, &frame->root->lk_owner);
- gf_log (this->name, GF_LOG_INFO,
- "[BLOCKED] Locker = {%s} Lockee = {%s} Lock = {%s}",
- pl_locker, pl_lockee, pl_lock);
+ gf_log(this->name, GF_LOG_INFO,
+ "[BLOCKED] Locker = {%s} Lockee = {%s} Lock = {%s}", pl_locker,
+ pl_lockee, pl_lock);
}
-
void
-pl_trace_flush (xlator_t *this, call_frame_t *frame, fd_t *fd)
+pl_trace_flush(xlator_t *this, call_frame_t *frame, fd_t *fd)
{
- posix_locks_private_t *priv = NULL;
- char pl_locker[256];
- char pl_lockee[256];
- pl_inode_t *pl_inode = NULL;
+ posix_locks_private_t *priv = NULL;
+ char pl_locker[256];
+ char pl_lockee[256];
+ pl_inode_t *pl_inode = NULL;
- priv = this->private;
+ priv = this->private;
- if (!priv->trace)
- return;
+ if (!priv->trace)
+ return;
- pl_inode = pl_inode_get (this, fd->inode);
+ pl_inode = pl_inode_get(this, fd->inode, NULL);
- if (pl_inode && __pl_inode_is_empty (pl_inode))
- return;
+ if (pl_inode && __pl_inode_is_empty(pl_inode))
+ return;
- pl_print_locker (pl_locker, 256, this, frame);
- pl_print_lockee (pl_lockee, 256, fd, NULL);
+ pl_print_locker(pl_locker, 256, this, frame);
+ pl_print_lockee(pl_lockee, 256, fd, NULL);
- gf_log (this->name, GF_LOG_INFO,
- "[FLUSH] Locker = {%s} Lockee = {%s}",
- pl_locker, pl_lockee);
+ gf_log(this->name, GF_LOG_INFO, "[FLUSH] Locker = {%s} Lockee = {%s}",
+ pl_locker, pl_lockee);
}
void
-pl_trace_release (xlator_t *this, fd_t *fd)
+pl_trace_release(xlator_t *this, fd_t *fd)
{
- posix_locks_private_t *priv = NULL;
- char pl_lockee[256];
+ posix_locks_private_t *priv = NULL;
+ char pl_lockee[256];
- priv = this->private;
+ priv = this->private;
- if (!priv->trace)
- return;
+ if (!priv->trace)
+ return;
- pl_print_lockee (pl_lockee, 256, fd, NULL);
+ pl_print_lockee(pl_lockee, 256, fd, NULL);
- gf_log (this->name, GF_LOG_INFO,
- "[RELEASE] Lockee = {%s}", pl_lockee);
+ gf_log(this->name, GF_LOG_INFO, "[RELEASE] Lockee = {%s}", pl_lockee);
}
-
void
-pl_update_refkeeper (xlator_t *this, inode_t *inode)
+pl_update_refkeeper(xlator_t *this, inode_t *inode)
{
- pl_inode_t *pl_inode = NULL;
- int is_empty = 0;
- int need_unref = 0;
- int need_ref = 0;
+ pl_inode_t *pl_inode = NULL;
+ int is_empty = 0;
+ int need_unref = 0;
+ int need_ref = 0;
- pl_inode = pl_inode_get (this, inode);
+ pl_inode = pl_inode_get(this, inode, NULL);
+ if (!pl_inode)
+ return;
- pthread_mutex_lock (&pl_inode->mutex);
- {
- is_empty = __pl_inode_is_empty (pl_inode);
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ is_empty = __pl_inode_is_empty(pl_inode);
- if (is_empty && pl_inode->refkeeper) {
- need_unref = 1;
- pl_inode->refkeeper = NULL;
- }
+ if (is_empty && pl_inode->refkeeper) {
+ need_unref = 1;
+ pl_inode->refkeeper = NULL;
+ }
- if (!is_empty && !pl_inode->refkeeper) {
- need_ref = 1;
- pl_inode->refkeeper = inode;
- }
+ if (!is_empty && !pl_inode->refkeeper) {
+ need_ref = 1;
+ pl_inode->refkeeper = inode;
}
- pthread_mutex_unlock (&pl_inode->mutex);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
- if (need_unref)
- inode_unref (inode);
+ if (need_unref)
+ inode_unref(inode);
- if (need_ref)
- inode_ref (inode);
+ if (need_ref)
+ inode_ref(inode);
}
+/* Get lock enforcement info from disk */
+int
+pl_fetch_mlock_info_from_disk(xlator_t *this, pl_inode_t *pl_inode,
+ pl_local_t *local)
+{
+ dict_t *xdata_rsp = NULL;
+ int ret = 0;
+ int op_ret = 0;
+
+ if (!local) {
+ return -1;
+ }
+
+ if (local->fd) {
+ op_ret = syncop_fgetxattr(this, local->fd, &xdata_rsp,
+ GF_ENFORCE_MANDATORY_LOCK, NULL, NULL);
+ } else {
+ op_ret = syncop_getxattr(this, &local->loc[0], &xdata_rsp,
+ GF_ENFORCE_MANDATORY_LOCK, NULL, NULL);
+ }
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ if (op_ret >= 0) {
+ pl_inode->mlock_enforced = _gf_true;
+ pl_inode->check_mlock_info = _gf_false;
+ } else {
+ gf_msg(this->name, GF_LOG_WARNING, -op_ret, 0,
+ "getxattr failed with %d", op_ret);
+ pl_inode->mlock_enforced = _gf_false;
+
+ if (-op_ret == ENODATA) {
+ pl_inode->check_mlock_info = _gf_false;
+ } else {
+ pl_inode->check_mlock_info = _gf_true;
+ }
+ }
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+
+ return ret;
+}
pl_inode_t *
-pl_inode_get (xlator_t *this, inode_t *inode)
+pl_inode_get(xlator_t *this, inode_t *inode, pl_local_t *local)
{
- uint64_t tmp_pl_inode = 0;
- pl_inode_t *pl_inode = NULL;
- int ret = 0;
+ uint64_t tmp_pl_inode = 0;
+ pl_inode_t *pl_inode = NULL;
+ int ret = 0;
+
+ LOCK(&inode->lock);
+ {
+ ret = __inode_ctx_get(inode, this, &tmp_pl_inode);
+ if (ret == 0) {
+ pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
+ goto unlock;
+ }
- LOCK (&inode->lock);
- {
- ret = __inode_ctx_get (inode, this, &tmp_pl_inode);
- if (ret == 0) {
- pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
- goto unlock;
- }
- pl_inode = GF_CALLOC (1, sizeof (*pl_inode),
- gf_locks_mt_pl_inode_t);
- if (!pl_inode) {
- goto unlock;
- }
+ pl_inode = GF_CALLOC(1, sizeof(*pl_inode), gf_locks_mt_pl_inode_t);
+ if (!pl_inode) {
+ goto unlock;
+ }
- gf_log (this->name, GF_LOG_TRACE,
- "Allocating new pl inode");
-
- pthread_mutex_init (&pl_inode->mutex, NULL);
-
- INIT_LIST_HEAD (&pl_inode->dom_list);
- INIT_LIST_HEAD (&pl_inode->ext_list);
- INIT_LIST_HEAD (&pl_inode->rw_list);
- INIT_LIST_HEAD (&pl_inode->reservelk_list);
- INIT_LIST_HEAD (&pl_inode->blocked_reservelks);
- INIT_LIST_HEAD (&pl_inode->blocked_calls);
- INIT_LIST_HEAD (&pl_inode->metalk_list);
- INIT_LIST_HEAD (&pl_inode->queued_locks);
- gf_uuid_copy (pl_inode->gfid, inode->gfid);
-
- ret = __inode_ctx_put (inode, this, (uint64_t)(long)(pl_inode));
- if (ret) {
- pthread_mutex_destroy (&pl_inode->mutex);
- GF_FREE (pl_inode);
- pl_inode = NULL;
- goto unlock;
- }
+ gf_log(this->name, GF_LOG_TRACE, "Allocating new pl inode");
+
+ pthread_mutex_init(&pl_inode->mutex, NULL);
+ pthread_cond_init(&pl_inode->check_fop_wind_count, 0);
+
+ INIT_LIST_HEAD(&pl_inode->dom_list);
+ INIT_LIST_HEAD(&pl_inode->ext_list);
+ INIT_LIST_HEAD(&pl_inode->rw_list);
+ INIT_LIST_HEAD(&pl_inode->reservelk_list);
+ INIT_LIST_HEAD(&pl_inode->blocked_reservelks);
+ INIT_LIST_HEAD(&pl_inode->blocked_calls);
+ INIT_LIST_HEAD(&pl_inode->metalk_list);
+ INIT_LIST_HEAD(&pl_inode->queued_locks);
+ INIT_LIST_HEAD(&pl_inode->waiting);
+ gf_uuid_copy(pl_inode->gfid, inode->gfid);
+
+ pl_inode->check_mlock_info = _gf_true;
+ pl_inode->mlock_enforced = _gf_false;
+
+ /* -2 means never looked up. -1 means something went wrong and link
+ * tracking is disabled. */
+ pl_inode->links = -2;
+
+ ret = __inode_ctx_put(inode, this, (uint64_t)(long)(pl_inode));
+ if (ret) {
+ pthread_mutex_destroy(&pl_inode->mutex);
+ GF_FREE(pl_inode);
+ pl_inode = NULL;
+ goto unlock;
}
+ }
unlock:
- UNLOCK (&inode->lock);
+ UNLOCK(&inode->lock);
- return pl_inode;
-}
+ if ((pl_inode != NULL) && pl_is_mandatory_locking_enabled(pl_inode) &&
+ pl_inode->check_mlock_info && local) {
+ /* Note: The lock enforcement information per file can be stored in the
+ attribute flag of stat(x) in posix. With that there won't be a need
+ for doing getxattr post a reboot
+ */
+ pl_fetch_mlock_info_from_disk(this, pl_inode, local);
+ }
+ return pl_inode;
+}
/* Create a new posix_lock_t */
posix_lock_t *
-new_posix_lock (struct gf_flock *flock, client_t *client, pid_t client_pid,
- gf_lkowner_t *owner, fd_t *fd, uint32_t lk_flags, int blocking)
+new_posix_lock(struct gf_flock *flock, client_t *client, pid_t client_pid,
+ gf_lkowner_t *owner, fd_t *fd, uint32_t lk_flags, int blocking,
+ int32_t *op_errno)
{
- posix_lock_t *lock = NULL;
+ posix_lock_t *lock = NULL;
- GF_VALIDATE_OR_GOTO ("posix-locks", flock, out);
- GF_VALIDATE_OR_GOTO ("posix-locks", client, out);
- GF_VALIDATE_OR_GOTO ("posix-locks", fd, out);
+ GF_VALIDATE_OR_GOTO("posix-locks", flock, out);
+ GF_VALIDATE_OR_GOTO("posix-locks", client, out);
+ GF_VALIDATE_OR_GOTO("posix-locks", fd, out);
- lock = GF_CALLOC (1, sizeof (posix_lock_t),
- gf_locks_mt_posix_lock_t);
- if (!lock) {
- goto out;
- }
+ if (!pl_is_lk_owner_valid(owner, client)) {
+ *op_errno = EINVAL;
+ goto out;
+ }
- lock->fl_start = flock->l_start;
- lock->fl_type = flock->l_type;
+ lock = GF_CALLOC(1, sizeof(posix_lock_t), gf_locks_mt_posix_lock_t);
+ if (!lock) {
+ *op_errno = ENOMEM;
+ goto out;
+ }
- if (flock->l_len == 0)
- lock->fl_end = LLONG_MAX;
- else
- lock->fl_end = flock->l_start + flock->l_len - 1;
+ lock->fl_start = flock->l_start;
+ lock->fl_type = flock->l_type;
- lock->client = client;
+ if (flock->l_len == 0)
+ lock->fl_end = LLONG_MAX;
+ else
+ lock->fl_end = flock->l_start + flock->l_len - 1;
- lock->client_uid = gf_strdup (client->client_uid);
- if (lock->client_uid == NULL) {
- GF_FREE (lock);
- lock = NULL;
- goto out;
- }
+ lock->client = client;
- lock->fd_num = fd_to_fdnum (fd);
- lock->fd = fd;
- lock->client_pid = client_pid;
- lock->owner = *owner;
- lock->lk_flags = lk_flags;
+ lock->client_uid = gf_strdup(client->client_uid);
+ if (lock->client_uid == NULL) {
+ GF_FREE(lock);
+ lock = NULL;
+ *op_errno = ENOMEM;
+ goto out;
+ }
- lock->blocking = blocking;
- memcpy (&lock->user_flock, flock, sizeof (lock->user_flock));
+ lock->fd_num = fd_to_fdnum(fd);
+ lock->fd = fd;
+ lock->client_pid = client_pid;
+ lock->owner = *owner;
+ lock->lk_flags = lk_flags;
- INIT_LIST_HEAD (&lock->list);
+ lock->blocking = blocking;
+ memcpy(&lock->user_flock, flock, sizeof(lock->user_flock));
+
+ INIT_LIST_HEAD(&lock->list);
out:
- return lock;
+ return lock;
}
-
/* Delete a lock from the inode's lock list */
void
-__delete_lock (posix_lock_t *lock)
+__delete_lock(posix_lock_t *lock)
{
- list_del_init (&lock->list);
+ list_del_init(&lock->list);
}
-
/* Destroy a posix_lock */
void
-__destroy_lock (posix_lock_t *lock)
+__destroy_lock(posix_lock_t *lock)
{
- GF_FREE (lock->client_uid);
- GF_FREE (lock);
+ GF_FREE(lock->client_uid);
+ GF_FREE(lock);
}
static posix_lock_t *
__copy_lock(posix_lock_t *src)
{
- posix_lock_t *dst;
-
- dst = GF_CALLOC(1, sizeof(posix_lock_t), gf_locks_mt_posix_lock_t);
- if (dst != NULL) {
- memcpy (dst, src, sizeof(posix_lock_t));
- dst->client_uid = gf_strdup(src->client_uid);
- if (dst->client_uid == NULL) {
- GF_FREE(dst);
- dst = NULL;
- }
- INIT_LIST_HEAD (&dst->list);
+ posix_lock_t *dst;
+
+ dst = GF_MALLOC(sizeof(posix_lock_t), gf_locks_mt_posix_lock_t);
+ if (dst != NULL) {
+ memcpy(dst, src, sizeof(posix_lock_t));
+ dst->client_uid = gf_strdup(src->client_uid);
+ if (dst->client_uid == NULL) {
+ GF_FREE(dst);
+ dst = NULL;
}
- return dst;
+ if (dst != NULL)
+ INIT_LIST_HEAD(&dst->list);
+ }
+
+ return dst;
}
/* Convert a posix_lock to a struct gf_flock */
void
-posix_lock_to_flock (posix_lock_t *lock, struct gf_flock *flock)
+posix_lock_to_flock(posix_lock_t *lock, struct gf_flock *flock)
{
- flock->l_pid = lock->user_flock.l_pid;
- flock->l_type = lock->fl_type;
- flock->l_start = lock->fl_start;
- flock->l_owner = lock->owner;
-
- if (lock->fl_end == LLONG_MAX)
- flock->l_len = 0;
- else
- flock->l_len = lock->fl_end - lock->fl_start + 1;
+ flock->l_pid = lock->user_flock.l_pid;
+ flock->l_type = lock->fl_type;
+ flock->l_start = lock->fl_start;
+ flock->l_owner = lock->owner;
+
+ if (lock->fl_end == LLONG_MAX)
+ flock->l_len = 0;
+ else
+ flock->l_len = lock->fl_end - lock->fl_start + 1;
}
/* Insert the lock into the inode's lock list */
static void
-__insert_lock (pl_inode_t *pl_inode, posix_lock_t *lock)
+__insert_lock(pl_inode_t *pl_inode, posix_lock_t *lock)
{
- if (lock->blocked)
- gettimeofday (&lock->blkd_time, NULL);
- else
- gettimeofday (&lock->granted_time, NULL);
+ if (lock->blocked)
+ lock->blkd_time = gf_time();
+ else
+ lock->granted_time = gf_time();
- list_add_tail (&lock->list, &pl_inode->ext_list);
-
- return;
+ list_add_tail(&lock->list, &pl_inode->ext_list);
}
-
/* Return true if the locks overlap, false otherwise */
int
-locks_overlap (posix_lock_t *l1, posix_lock_t *l2)
+locks_overlap(posix_lock_t *l1, posix_lock_t *l2)
{
- /*
- Note:
- FUSE always gives us absolute offsets, so no need to worry
- about SEEK_CUR or SEEK_END
- */
+ /*
+ Note:
+ FUSE always gives us absolute offsets, so no need to worry
+ about SEEK_CUR or SEEK_END
+ */
- return ((l1->fl_end >= l2->fl_start) &&
- (l2->fl_end >= l1->fl_start));
+ return ((l1->fl_end >= l2->fl_start) && (l2->fl_end >= l1->fl_start));
}
-
/* Return true if the locks have the same owner */
int
-same_owner (posix_lock_t *l1, posix_lock_t *l2)
+same_owner(posix_lock_t *l1, posix_lock_t *l2)
{
-
- return (is_same_lkowner (&l1->owner, &l2->owner) &&
- (l1->client == l2->client));
-
+ return (is_same_lkowner(&l1->owner, &l2->owner) &&
+ (l1->client == l2->client));
}
-
/* Delete all F_UNLCK locks */
void
-__delete_unlck_locks (pl_inode_t *pl_inode)
+__delete_unlck_locks(pl_inode_t *pl_inode)
{
- posix_lock_t *l = NULL;
- posix_lock_t *tmp = NULL;
-
- list_for_each_entry_safe (l, tmp, &pl_inode->ext_list, list) {
- if (l->fl_type == F_UNLCK) {
- __delete_lock (l);
- __destroy_lock (l);
- }
+ posix_lock_t *l = NULL;
+ posix_lock_t *tmp = NULL;
+
+ list_for_each_entry_safe(l, tmp, &pl_inode->ext_list, list)
+ {
+ if (l->fl_type == F_UNLCK) {
+ __delete_lock(l);
+ __destroy_lock(l);
}
+ }
}
-
/* Add two locks */
static posix_lock_t *
-add_locks (posix_lock_t *l1, posix_lock_t *l2, posix_lock_t *dst)
+add_locks(posix_lock_t *l1, posix_lock_t *l2, posix_lock_t *dst)
{
- posix_lock_t *sum = NULL;
+ posix_lock_t *sum = NULL;
- sum = __copy_lock (dst);
- if (!sum)
- return NULL;
+ sum = __copy_lock(dst);
+ if (!sum)
+ return NULL;
- sum->fl_start = min (l1->fl_start, l2->fl_start);
- sum->fl_end = max (l1->fl_end, l2->fl_end);
+ sum->fl_start = min(l1->fl_start, l2->fl_start);
+ sum->fl_end = max(l1->fl_end, l2->fl_end);
- posix_lock_to_flock (sum, &sum->user_flock);
+ posix_lock_to_flock(sum, &sum->user_flock);
- return sum;
+ return sum;
}
/* Subtract two locks */
struct _values {
- posix_lock_t *locks[3];
+ posix_lock_t *locks[3];
};
/* {big} must always be contained inside {small} */
static struct _values
-subtract_locks (posix_lock_t *big, posix_lock_t *small)
+subtract_locks(posix_lock_t *big, posix_lock_t *small)
{
+ struct _values v = {.locks = {0, 0, 0}};
- struct _values v = { .locks = {0, 0, 0} };
-
- if ((big->fl_start == small->fl_start) &&
- (big->fl_end == small->fl_end)) {
- /* both edges coincide with big */
- v.locks[0] = __copy_lock(big);
- if (!v.locks[0]) {
- goto out;
- }
-
- v.locks[0]->fl_type = small->fl_type;
- v.locks[0]->user_flock.l_type = small->fl_type;
- goto done;
+ if ((big->fl_start == small->fl_start) && (big->fl_end == small->fl_end)) {
+ /* both edges coincide with big */
+ v.locks[0] = __copy_lock(big);
+ if (!v.locks[0]) {
+ goto out;
}
- if ((small->fl_start > big->fl_start) &&
- (small->fl_end < big->fl_end)) {
- /* both edges lie inside big */
- v.locks[0] = __copy_lock(big);
- v.locks[1] = __copy_lock(small);
- v.locks[2] = __copy_lock(big);
- if ((v.locks[0] == NULL) || (v.locks[1] == NULL) ||
- (v.locks[2] == NULL)) {
- goto out;
- }
-
- v.locks[0]->fl_end = small->fl_start - 1;
- v.locks[2]->fl_start = small->fl_end + 1;
- posix_lock_to_flock (v.locks[0], &v.locks[0]->user_flock);
- posix_lock_to_flock (v.locks[2], &v.locks[2]->user_flock);
- goto done;
-
+ v.locks[0]->fl_type = small->fl_type;
+ v.locks[0]->user_flock.l_type = small->fl_type;
+ goto done;
+ }
+
+ if ((small->fl_start > big->fl_start) && (small->fl_end < big->fl_end)) {
+ /* both edges lie inside big */
+ v.locks[0] = __copy_lock(big);
+ v.locks[1] = __copy_lock(small);
+ v.locks[2] = __copy_lock(big);
+ if ((v.locks[0] == NULL) || (v.locks[1] == NULL) ||
+ (v.locks[2] == NULL)) {
+ goto out;
}
- /* one edge coincides with big */
- if (small->fl_start == big->fl_start) {
- v.locks[0] = __copy_lock(big);
- v.locks[1] = __copy_lock(small);
- if ((v.locks[0] == NULL) || (v.locks[1] == NULL)) {
- goto out;
- }
-
- v.locks[0]->fl_start = small->fl_end + 1;
- posix_lock_to_flock (v.locks[0], &v.locks[0]->user_flock);
- goto done;
+ v.locks[0]->fl_end = small->fl_start - 1;
+ v.locks[2]->fl_start = small->fl_end + 1;
+ posix_lock_to_flock(v.locks[0], &v.locks[0]->user_flock);
+ posix_lock_to_flock(v.locks[2], &v.locks[2]->user_flock);
+ goto done;
+ }
+
+ /* one edge coincides with big */
+ if (small->fl_start == big->fl_start) {
+ v.locks[0] = __copy_lock(big);
+ v.locks[1] = __copy_lock(small);
+ if ((v.locks[0] == NULL) || (v.locks[1] == NULL)) {
+ goto out;
}
- if (small->fl_end == big->fl_end) {
- v.locks[0] = __copy_lock(big);
- v.locks[1] = __copy_lock(small);
- if ((v.locks[0] == NULL) || (v.locks[1] == NULL)) {
- goto out;
- }
+ v.locks[0]->fl_start = small->fl_end + 1;
+ posix_lock_to_flock(v.locks[0], &v.locks[0]->user_flock);
+ goto done;
+ }
- v.locks[0]->fl_end = small->fl_start - 1;
- posix_lock_to_flock (v.locks[0], &v.locks[0]->user_flock);
- goto done;
+ if (small->fl_end == big->fl_end) {
+ v.locks[0] = __copy_lock(big);
+ v.locks[1] = __copy_lock(small);
+ if ((v.locks[0] == NULL) || (v.locks[1] == NULL)) {
+ goto out;
}
- GF_ASSERT (0);
- gf_log ("posix-locks", GF_LOG_ERROR, "Unexpected case in subtract_locks");
+ v.locks[0]->fl_end = small->fl_start - 1;
+ posix_lock_to_flock(v.locks[0], &v.locks[0]->user_flock);
+ goto done;
+ }
+
+ GF_ASSERT(0);
+ gf_log("posix-locks", GF_LOG_ERROR, "Unexpected case in subtract_locks");
out:
- if (v.locks[0]) {
- __destroy_lock(v.locks[0]);
- v.locks[0] = NULL;
- }
- if (v.locks[1]) {
- __destroy_lock(v.locks[1]);
- v.locks[1] = NULL;
- }
- if (v.locks[2]) {
- __destroy_lock(v.locks[2]);
- v.locks[2] = NULL;
- }
+ if (v.locks[0]) {
+ __destroy_lock(v.locks[0]);
+ v.locks[0] = NULL;
+ }
+ if (v.locks[1]) {
+ __destroy_lock(v.locks[1]);
+ v.locks[1] = NULL;
+ }
+ if (v.locks[2]) {
+ __destroy_lock(v.locks[2]);
+ v.locks[2] = NULL;
+ }
done:
- return v;
+ return v;
}
static posix_lock_t *
-first_conflicting_overlap (pl_inode_t *pl_inode, posix_lock_t *lock)
+first_conflicting_overlap(pl_inode_t *pl_inode, posix_lock_t *lock)
{
- posix_lock_t *l = NULL;
- posix_lock_t *conf = NULL;
+ posix_lock_t *l = NULL;
+ posix_lock_t *conf = NULL;
- pthread_mutex_lock (&pl_inode->mutex);
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ list_for_each_entry(l, &pl_inode->ext_list, list)
{
- list_for_each_entry (l, &pl_inode->ext_list, list) {
- if (l->blocked)
- continue;
-
- if (locks_overlap (l, lock)) {
- if (same_owner (l, lock))
- continue;
-
- if ((l->fl_type == F_WRLCK) ||
- (lock->fl_type == F_WRLCK)) {
- conf = l;
- goto unlock;
- }
- }
+ if (l->blocked)
+ continue;
+
+ if (locks_overlap(l, lock)) {
+ if (same_owner(l, lock))
+ continue;
+
+ if ((l->fl_type == F_WRLCK) || (lock->fl_type == F_WRLCK)) {
+ conf = l;
+ goto unlock;
}
+ }
}
+ }
unlock:
- pthread_mutex_unlock (&pl_inode->mutex);
+ pthread_mutex_unlock(&pl_inode->mutex);
- return conf;
+ return conf;
}
/*
@@ -751,364 +789,803 @@ unlock:
If {begin} is NULL, then start from the beginning of the list
*/
static posix_lock_t *
-first_overlap (pl_inode_t *pl_inode, posix_lock_t *lock)
+first_overlap(pl_inode_t *pl_inode, posix_lock_t *lock)
{
- posix_lock_t *l = NULL;
+ posix_lock_t *l = NULL;
- list_for_each_entry (l, &pl_inode->ext_list, list) {
- if (l->blocked)
- continue;
+ list_for_each_entry(l, &pl_inode->ext_list, list)
+ {
+ if (l->blocked)
+ continue;
- if (locks_overlap (l, lock))
- return l;
- }
+ if (locks_overlap(l, lock))
+ return l;
+ }
- return NULL;
+ return NULL;
}
-
-
/* Return true if lock is grantable */
static int
-__is_lock_grantable (pl_inode_t *pl_inode, posix_lock_t *lock)
+__is_lock_grantable(pl_inode_t *pl_inode, posix_lock_t *lock)
{
- posix_lock_t *l = NULL;
- int ret = 1;
-
- list_for_each_entry (l, &pl_inode->ext_list, list) {
- if (!l->blocked && locks_overlap (lock, l)) {
- if (((l->fl_type == F_WRLCK)
- || (lock->fl_type == F_WRLCK))
- && (lock->fl_type != F_UNLCK)
- && !same_owner (l, lock)) {
- ret = 0;
- break;
- }
- }
+ posix_lock_t *l = NULL;
+ int ret = 1;
+
+ list_for_each_entry(l, &pl_inode->ext_list, list)
+ {
+ if (!l->blocked && locks_overlap(lock, l)) {
+ if (((l->fl_type == F_WRLCK) || (lock->fl_type == F_WRLCK)) &&
+ (lock->fl_type != F_UNLCK) && !same_owner(l, lock)) {
+ ret = 0;
+ break;
+ }
}
- return ret;
+ }
+ return ret;
}
-
-extern void do_blocked_rw (pl_inode_t *);
-
+extern void
+do_blocked_rw(pl_inode_t *);
static void
-__insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock)
+__insert_and_merge(pl_inode_t *pl_inode, posix_lock_t *lock)
{
- posix_lock_t *conf = NULL;
- posix_lock_t *t = NULL;
- posix_lock_t *sum = NULL;
- int i = 0;
- struct _values v = { .locks = {0, 0, 0} };
-
- list_for_each_entry_safe (conf, t, &pl_inode->ext_list, list) {
- if (conf->blocked)
- continue;
- if (!locks_overlap (conf, lock))
- continue;
-
- if (same_owner (conf, lock)) {
- if (conf->fl_type == lock->fl_type &&
- conf->lk_flags == lock->lk_flags) {
- sum = add_locks (lock, conf, lock);
+ posix_lock_t *conf = NULL;
+ posix_lock_t *t = NULL;
+ posix_lock_t *sum = NULL;
+ int i = 0;
+ struct _values v = {.locks = {0, 0, 0}};
+
+ list_for_each_entry_safe(conf, t, &pl_inode->ext_list, list)
+ {
+ if (conf->blocked)
+ continue;
+ if (!locks_overlap(conf, lock))
+ continue;
+
+ if (same_owner(conf, lock)) {
+ if (conf->fl_type == lock->fl_type &&
+ conf->lk_flags == lock->lk_flags) {
+ sum = add_locks(lock, conf, lock);
+
+ __delete_lock(conf);
+ __destroy_lock(conf);
- __delete_lock (conf);
- __destroy_lock (conf);
+ __destroy_lock(lock);
+ INIT_LIST_HEAD(&sum->list);
+ posix_lock_to_flock(sum, &sum->user_flock);
+ __insert_and_merge(pl_inode, sum);
- __destroy_lock (lock);
- INIT_LIST_HEAD (&sum->list);
- posix_lock_to_flock (sum, &sum->user_flock);
- __insert_and_merge (pl_inode, sum);
+ return;
+ } else {
+ sum = add_locks(lock, conf, conf);
- return;
- } else {
- sum = add_locks (lock, conf, conf);
+ v = subtract_locks(sum, lock);
- v = subtract_locks (sum, lock);
+ __delete_lock(conf);
+ __destroy_lock(conf);
- __delete_lock (conf);
- __destroy_lock (conf);
+ __delete_lock(lock);
+ __destroy_lock(lock);
- __delete_lock (lock);
- __destroy_lock (lock);
+ __destroy_lock(sum);
- __destroy_lock (sum);
+ for (i = 0; i < 3; i++) {
+ if (!v.locks[i])
+ continue;
- for (i = 0; i < 3; i++) {
- if (!v.locks[i])
- continue;
+ __insert_and_merge(pl_inode, v.locks[i]);
+ }
- __insert_and_merge (pl_inode,
- v.locks[i]);
- }
+ __delete_unlck_locks(pl_inode);
+ return;
+ }
+ }
- __delete_unlck_locks (pl_inode);
- return;
- }
- }
+ if (lock->fl_type == F_UNLCK) {
+ continue;
+ }
- if (lock->fl_type == F_UNLCK) {
- continue;
- }
+ if ((conf->fl_type == F_RDLCK) && (lock->fl_type == F_RDLCK)) {
+ __insert_lock(pl_inode, lock);
+ return;
+ }
+ }
+
+ /* no conflicts, so just insert */
+ if (lock->fl_type != F_UNLCK) {
+ __insert_lock(pl_inode, lock);
+ } else {
+ __destroy_lock(lock);
+ }
+}
- if ((conf->fl_type == F_RDLCK) && (lock->fl_type == F_RDLCK)) {
- __insert_lock (pl_inode, lock);
- return;
- }
+void
+__grant_blocked_locks(xlator_t *this, pl_inode_t *pl_inode,
+ struct list_head *granted)
+{
+ struct list_head tmp_list;
+ posix_lock_t *l = NULL;
+ posix_lock_t *tmp = NULL;
+ posix_lock_t *conf = NULL;
+
+ INIT_LIST_HEAD(&tmp_list);
+
+ list_for_each_entry_safe(l, tmp, &pl_inode->ext_list, list)
+ {
+ if (l->blocked) {
+ conf = first_overlap(pl_inode, l);
+ if (conf)
+ continue;
+
+ l->blocked = 0;
+ list_move_tail(&l->list, &tmp_list);
}
+ }
+
+ list_for_each_entry_safe(l, tmp, &tmp_list, list)
+ {
+ list_del_init(&l->list);
+
+ if (__is_lock_grantable(pl_inode, l)) {
+ conf = GF_CALLOC(1, sizeof(*conf), gf_locks_mt_posix_lock_t);
+
+ if (!conf) {
+ l->blocked = 1;
+ __insert_lock(pl_inode, l);
+ continue;
+ }
- /* no conflicts, so just insert */
- if (lock->fl_type != F_UNLCK) {
- __insert_lock (pl_inode, lock);
+ conf->frame = l->frame;
+ l->frame = NULL;
+
+ posix_lock_to_flock(l, &conf->user_flock);
+
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s (pid=%d) lk-owner:%s %" PRId64 " - %" PRId64
+ " => Granted",
+ l->fl_type == F_UNLCK ? "Unlock" : "Lock", l->client_pid,
+ lkowner_utoa(&l->owner), l->user_flock.l_start,
+ l->user_flock.l_len);
+
+ __insert_and_merge(pl_inode, l);
+
+ list_add(&conf->list, granted);
} else {
- __destroy_lock (lock);
+ l->blocked = 1;
+ __insert_lock(pl_inode, l);
}
+ }
}
-
void
-__grant_blocked_locks (xlator_t *this, pl_inode_t *pl_inode, struct list_head *granted)
+grant_blocked_locks(xlator_t *this, pl_inode_t *pl_inode)
{
- struct list_head tmp_list;
- posix_lock_t *l = NULL;
- posix_lock_t *tmp = NULL;
- posix_lock_t *conf = NULL;
+ struct list_head granted_list;
+ posix_lock_t *tmp = NULL;
+ posix_lock_t *lock = NULL;
+ pl_local_t *local = NULL;
+ INIT_LIST_HEAD(&granted_list);
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ __grant_blocked_locks(this, pl_inode, &granted_list);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+
+ list_for_each_entry_safe(lock, tmp, &granted_list, list)
+ {
+ list_del_init(&lock->list);
+
+ pl_trace_out(this, lock->frame, NULL, NULL, F_SETLKW, &lock->user_flock,
+ 0, 0, NULL);
+ local = lock->frame->local;
+ PL_STACK_UNWIND_AND_FREE(local, lk, lock->frame, 0, 0,
+ &lock->user_flock, NULL);
+ __destroy_lock(lock);
+ }
+
+ return;
+}
- INIT_LIST_HEAD (&tmp_list);
+static int
+pl_send_prelock_unlock(xlator_t *this, pl_inode_t *pl_inode,
+ posix_lock_t *old_lock)
+{
+ struct gf_flock flock = {
+ 0,
+ };
+ posix_lock_t *unlock_lock = NULL;
+ int32_t op_errno = 0;
- list_for_each_entry_safe (l, tmp, &pl_inode->ext_list, list) {
- if (l->blocked) {
- conf = first_overlap (pl_inode, l);
- if (conf)
- continue;
+ struct list_head granted_list;
+ posix_lock_t *tmp = NULL;
+ posix_lock_t *lock = NULL;
+ pl_local_t *local = NULL;
- l->blocked = 0;
- list_move_tail (&l->list, &tmp_list);
- }
- }
+ int ret = -1;
- list_for_each_entry_safe (l, tmp, &tmp_list, list) {
- list_del_init (&l->list);
+ INIT_LIST_HEAD(&granted_list);
- if (__is_lock_grantable (pl_inode, l)) {
- conf = GF_CALLOC (1, sizeof (*conf),
- gf_locks_mt_posix_lock_t);
+ flock.l_type = F_UNLCK;
+ flock.l_whence = old_lock->user_flock.l_whence;
+ flock.l_start = old_lock->user_flock.l_start;
+ flock.l_len = old_lock->user_flock.l_len;
+ flock.l_pid = old_lock->user_flock.l_pid;
- if (!conf) {
- l->blocked = 1;
- __insert_lock (pl_inode, l);
- continue;
- }
+ unlock_lock = new_posix_lock(&flock, old_lock->client, old_lock->client_pid,
+ &old_lock->owner, old_lock->fd,
+ old_lock->lk_flags, 0, &op_errno);
+ GF_VALIDATE_OR_GOTO(this->name, unlock_lock, out);
+ ret = 0;
- conf->frame = l->frame;
- l->frame = NULL;
+ __insert_and_merge(pl_inode, unlock_lock);
- posix_lock_to_flock (l, &conf->user_flock);
+ __grant_blocked_locks(this, pl_inode, &granted_list);
- gf_log (this->name, GF_LOG_TRACE,
- "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => Granted",
- l->fl_type == F_UNLCK ? "Unlock" : "Lock",
- l->client_pid, lkowner_utoa (&l->owner),
- l->user_flock.l_start,
- l->user_flock.l_len);
+ list_for_each_entry_safe(lock, tmp, &granted_list, list)
+ {
+ list_del_init(&lock->list);
- __insert_and_merge (pl_inode, l);
+ pl_trace_out(this, lock->frame, NULL, NULL, F_SETLKW, &lock->user_flock,
+ 0, 0, NULL);
+ local = lock->frame->local;
+ PL_STACK_UNWIND_AND_FREE(local, lk, lock->frame, 0, 0,
+ &lock->user_flock, NULL);
+ __destroy_lock(lock);
+ }
- list_add (&conf->list, granted);
- } else {
- l->blocked = 1;
- __insert_lock (pl_inode, l);
- }
- }
+out:
+ return ret;
}
-
-void
-grant_blocked_locks (xlator_t *this, pl_inode_t *pl_inode)
+int
+pl_setlk(xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
+ int can_block)
{
- struct list_head granted_list;
- posix_lock_t *tmp = NULL;
- posix_lock_t *lock = NULL;
+ int ret = 0;
- INIT_LIST_HEAD (&granted_list);
+ errno = 0;
- pthread_mutex_lock (&pl_inode->mutex);
- {
- __grant_blocked_locks (this, pl_inode, &granted_list);
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ /* Send unlock before the actual lock to
+ prevent lock upgrade / downgrade
+ problems only if:
+ - it is a blocking call
+ - it has other conflicting locks
+ */
+
+ if (can_block && !(__is_lock_grantable(pl_inode, lock))) {
+ ret = pl_send_prelock_unlock(this, pl_inode, lock);
+ if (ret)
+ gf_log(this->name, GF_LOG_DEBUG,
+ "Could not send pre-lock "
+ "unlock");
}
- pthread_mutex_unlock (&pl_inode->mutex);
- list_for_each_entry_safe (lock, tmp, &granted_list, list) {
- list_del_init (&lock->list);
+ if (__is_lock_grantable(pl_inode, lock)) {
+ if (pl_metalock_is_active(pl_inode)) {
+ __pl_queue_lock(pl_inode, lock);
+ pthread_mutex_unlock(&pl_inode->mutex);
+ ret = -2;
+ goto out;
+ }
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s (pid=%d) lk-owner:%s %" PRId64 " - %" PRId64 " => OK",
+ lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+ lock->client_pid, lkowner_utoa(&lock->owner),
+ lock->user_flock.l_start, lock->user_flock.l_len);
+ __insert_and_merge(pl_inode, lock);
+ } else if (can_block) {
+ if (pl_metalock_is_active(pl_inode)) {
+ __pl_queue_lock(pl_inode, lock);
+ pthread_mutex_unlock(&pl_inode->mutex);
+ ret = -2;
+ goto out;
+ }
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s (pid=%d) lk-owner:%s %" PRId64 " - %" PRId64
+ " => Blocked",
+ lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+ lock->client_pid, lkowner_utoa(&lock->owner),
+ lock->user_flock.l_start, lock->user_flock.l_len);
+
+ pl_trace_block(this, lock->frame, NULL, NULL, F_SETLKW,
+ &lock->user_flock, NULL);
+
+ lock->blocked = 1;
+ __insert_lock(pl_inode, lock);
+ ret = -1;
+ } else {
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s (pid=%d) lk-owner:%s %" PRId64 " - %" PRId64 " => NOK",
+ lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+ lock->client_pid, lkowner_utoa(&lock->owner),
+ lock->user_flock.l_start, lock->user_flock.l_len);
+ errno = EAGAIN;
+ ret = -1;
+ }
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
- pl_trace_out (this, lock->frame, NULL, NULL, F_SETLKW,
- &lock->user_flock, 0, 0, NULL);
+ grant_blocked_locks(this, pl_inode);
- STACK_UNWIND_STRICT (lk, lock->frame, 0, 0,
- &lock->user_flock, NULL);
+ do_blocked_rw(pl_inode);
+out:
+ return ret;
+}
+
+posix_lock_t *
+pl_getlk(pl_inode_t *pl_inode, posix_lock_t *lock)
+{
+ posix_lock_t *conf = first_conflicting_overlap(pl_inode, lock);
+ if (conf == NULL) {
+ lock->fl_type = F_UNLCK;
+ return lock;
+ }
+
+ return conf;
+}
+
+gf_boolean_t
+pl_does_monkey_want_stuck_lock()
+{
+ long int monkey_unlock_rand = 0;
+ long int monkey_unlock_rand_rem = 0;
+
+ /* coverity[DC.WEAK_CRYPTO] */
+ monkey_unlock_rand = random();
+ monkey_unlock_rand_rem = monkey_unlock_rand % 100;
+ if (monkey_unlock_rand_rem == 0)
+ return _gf_true;
+ return _gf_false;
+}
+
+int
+pl_lock_preempt(pl_inode_t *pl_inode, posix_lock_t *reqlock)
+{
+ posix_lock_t *lock = NULL;
+ posix_lock_t *i = NULL;
+ pl_rw_req_t *rw = NULL;
+ pl_rw_req_t *itr = NULL;
+ struct list_head unwind_blist = {
+ 0,
+ };
+ struct list_head unwind_rw_list = {
+ 0,
+ };
+ int ret = 0;
+
+ INIT_LIST_HEAD(&unwind_blist);
+ INIT_LIST_HEAD(&unwind_rw_list);
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ /*
+ - go through the lock list
+ - remove all locks from different owners
+ - same owner locks will be added or substracted based on
+ the new request
+ - add the new lock
+ */
+ list_for_each_entry_safe(lock, i, &pl_inode->ext_list, list)
+ {
+ if (lock->blocked) {
+ list_del_init(&lock->list);
+ list_add(&lock->list, &unwind_blist);
+ continue;
+ }
+
+ if (locks_overlap(lock, reqlock)) {
+ if (same_owner(lock, reqlock))
+ continue;
+
+ /* remove conflicting locks */
+ list_del_init(&lock->list);
+ __delete_lock(lock);
__destroy_lock(lock);
+ }
+ }
+
+ __insert_and_merge(pl_inode, reqlock);
+
+ list_for_each_entry_safe(rw, itr, &pl_inode->rw_list, list)
+ {
+ list_del_init(&rw->list);
+ list_add(&rw->list, &unwind_rw_list);
}
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+
+ /* unwind blocked locks */
+ list_for_each_entry_safe(lock, i, &unwind_blist, list)
+ {
+ PL_STACK_UNWIND_AND_FREE(((pl_local_t *)lock->frame->local), lk,
+ lock->frame, -1, EBUSY, &lock->user_flock,
+ NULL);
+ __destroy_lock(lock);
+ }
+
+ /* unwind blocked IOs */
+ list_for_each_entry_safe(rw, itr, &unwind_rw_list, list)
+ {
+ pl_clean_local(rw->stub->frame->local);
+ call_unwind_error(rw->stub, -1, EBUSY);
+ }
+
+ return ret;
+}
+
+/* Return true in case we need to ensure mandatory-locking
+ * semantics under different modes.
+ */
+gf_boolean_t
+pl_is_mandatory_locking_enabled(pl_inode_t *pl_inode)
+{
+ posix_locks_private_t *priv = THIS->private;
+
+ if (priv->mandatory_mode == MLK_FILE_BASED && pl_inode->mandatory)
+ return _gf_true;
+ else if (priv->mandatory_mode == MLK_FORCED ||
+ priv->mandatory_mode == MLK_OPTIMAL)
+ return _gf_true;
+ return _gf_false;
+}
+
+void
+pl_clean_local(pl_local_t *local)
+{
+ if (!local)
return;
+
+ if (local->inodelk_dom_count_req)
+ data_unref(local->inodelk_dom_count_req);
+ loc_wipe(&local->loc[0]);
+ loc_wipe(&local->loc[1]);
+ if (local->fd)
+ fd_unref(local->fd);
+ if (local->inode)
+ inode_unref(local->inode);
+ mem_put(local);
}
-static int
-pl_send_prelock_unlock (xlator_t *this, pl_inode_t *pl_inode,
- posix_lock_t *old_lock)
+/*
+TODO: detach local initialization from PL_LOCAL_GET_REQUESTS and add it here
+*/
+int
+pl_local_init(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
{
- struct gf_flock flock = {0,};
- posix_lock_t *unlock_lock = NULL;
+ pl_local_t *local = NULL;
+
+ if (!loc && !fd) {
+ return -1;
+ }
+
+ if (!frame->local) {
+ local = mem_get0(this->local_pool);
+ if (!local) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, 0,
+ "mem allocation failed");
+ return -1;
+ }
- struct list_head granted_list;
- posix_lock_t *tmp = NULL;
- posix_lock_t *lock = NULL;
+ local->inode = (loc ? inode_ref(loc->inode) : inode_ref(fd->inode));
- int ret = -1;
+ frame->local = local;
+ }
- INIT_LIST_HEAD (&granted_list);
+ return 0;
+}
- flock.l_type = F_UNLCK;
- flock.l_whence = old_lock->user_flock.l_whence;
- flock.l_start = old_lock->user_flock.l_start;
- flock.l_len = old_lock->user_flock.l_len;
- flock.l_pid = old_lock->user_flock.l_pid;
+gf_boolean_t
+pl_is_lk_owner_valid(gf_lkowner_t *owner, client_t *client)
+{
+ if (client && (client->opversion < GD_OP_VERSION_7_0)) {
+ return _gf_true;
+ }
+ if (is_lk_owner_null(owner)) {
+ return _gf_false;
+ }
+ return _gf_true;
+}
- unlock_lock = new_posix_lock (&flock, old_lock->client,
- old_lock->client_pid, &old_lock->owner,
- old_lock->fd, old_lock->lk_flags, 0);
- GF_VALIDATE_OR_GOTO (this->name, unlock_lock, out);
- ret = 0;
+static int32_t
+pl_inode_from_loc(loc_t *loc, inode_t **pinode)
+{
+ inode_t *inode = NULL;
+ int32_t error = 0;
+
+ if (loc->inode != NULL) {
+ inode = inode_ref(loc->inode);
+ goto done;
+ }
+
+ if (loc->parent == NULL) {
+ error = EINVAL;
+ goto done;
+ }
+
+ if (!gf_uuid_is_null(loc->gfid)) {
+ inode = inode_find(loc->parent->table, loc->gfid);
+ if (inode != NULL) {
+ goto done;
+ }
+ }
- __insert_and_merge (pl_inode, unlock_lock);
+ if (loc->name == NULL) {
+ error = EINVAL;
+ goto done;
+ }
- __grant_blocked_locks (this, pl_inode, &granted_list);
+ inode = inode_grep(loc->parent->table, loc->parent, loc->name);
+ if (inode == NULL) {
+ /* We haven't found any inode. This means that the file doesn't exist
+ * or that even if it exists, we don't have any knowledge about it, so
+ * we don't have locks on it either, which is fine for our purposes. */
+ goto done;
+ }
- list_for_each_entry_safe (lock, tmp, &granted_list, list) {
- list_del_init (&lock->list);
+done:
+ *pinode = inode;
- pl_trace_out (this, lock->frame, NULL, NULL, F_SETLKW,
- &lock->user_flock, 0, 0, NULL);
+ return error;
+}
- STACK_UNWIND_STRICT (lk, lock->frame, 0, 0,
- &lock->user_flock, NULL);
+static gf_boolean_t
+pl_inode_has_owners(xlator_t *xl, client_t *client, pl_inode_t *pl_inode,
+ struct timespec *now, struct list_head *contend)
+{
+ pl_dom_list_t *dom;
+ pl_inode_lock_t *lock;
+ gf_boolean_t has_owners = _gf_false;
- __destroy_lock(lock);
+ list_for_each_entry(dom, &pl_inode->dom_list, inode_list)
+ {
+ list_for_each_entry(lock, &dom->inodelk_list, list)
+ {
+ /* If the lock belongs to the same client, we assume it's related
+ * to the same operation, so we allow the removal to continue. */
+ if (lock->client == client) {
+ continue;
+ }
+ /* If the lock belongs to an internal process, we don't block the
+ * removal. */
+ if (lock->client_pid < 0) {
+ continue;
+ }
+ if (contend == NULL) {
+ return _gf_true;
+ }
+ has_owners = _gf_true;
+ inodelk_contention_notify_check(xl, lock, now, contend);
}
+ }
-out:
- return ret;
+ return has_owners;
}
-int
-pl_setlk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
- int can_block)
+int32_t
+pl_inode_remove_prepare(xlator_t *xl, call_frame_t *frame, loc_t *loc,
+ pl_inode_t **ppl_inode, struct list_head *contend)
{
- int ret = 0;
+ struct timespec now;
+ inode_t *inode;
+ pl_inode_t *pl_inode;
+ int32_t error;
+
+ pl_inode = NULL;
+
+ error = pl_inode_from_loc(loc, &inode);
+ if ((error != 0) || (inode == NULL)) {
+ goto done;
+ }
+
+ pl_inode = pl_inode_get(xl, inode, NULL);
+ if (pl_inode == NULL) {
+ inode_unref(inode);
+ error = ENOMEM;
+ goto done;
+ }
+
+ /* pl_inode_from_loc() already increments ref count for inode, so
+ * we only assign here our reference. */
+ pl_inode->inode = inode;
+
+ timespec_now(&now);
+
+ pthread_mutex_lock(&pl_inode->mutex);
+
+ if (pl_inode->removed) {
+ error = ESTALE;
+ goto unlock;
+ }
+
+ if (pl_inode_has_owners(xl, frame->root->client, pl_inode, &now, contend)) {
+ error = -1;
+ /* We skip the unlock here because the caller must create a stub when
+ * we return -1 and do a call to pl_inode_remove_complete(), which
+ * assumes the lock is still acquired and will release it once
+ * everything else is prepared. */
+ goto done;
+ }
+
+ pl_inode->is_locked = _gf_true;
+ pl_inode->remove_running++;
- errno = 0;
+unlock:
+ pthread_mutex_unlock(&pl_inode->mutex);
- pthread_mutex_lock (&pl_inode->mutex);
- {
- /* Send unlock before the actual lock to
- prevent lock upgrade / downgrade
- problems only if:
- - it is a blocking call
- - it has other conflicting locks
- */
-
- if (can_block &&
- !(__is_lock_grantable (pl_inode, lock))) {
- ret = pl_send_prelock_unlock (this, pl_inode,
- lock);
- if (ret)
- gf_log (this->name, GF_LOG_DEBUG,
- "Could not send pre-lock "
- "unlock");
- }
+done:
+ *ppl_inode = pl_inode;
- if (__is_lock_grantable (pl_inode, lock)) {
- if (pl_metalock_is_active (pl_inode)) {
- __pl_queue_lock (pl_inode, lock, can_block);
- pthread_mutex_unlock (&pl_inode->mutex);
- ret = -2;
- goto out;
- }
- gf_log (this->name, GF_LOG_TRACE,
- "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => OK",
- lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
- lock->client_pid,
- lkowner_utoa (&lock->owner),
- lock->user_flock.l_start,
- lock->user_flock.l_len);
- __insert_and_merge (pl_inode, lock);
- } else if (can_block) {
- if (pl_metalock_is_active (pl_inode)) {
- __pl_queue_lock (pl_inode, lock, can_block);
- pthread_mutex_unlock (&pl_inode->mutex);
- ret = -2;
- goto out;
- }
- gf_log (this->name, GF_LOG_TRACE,
- "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => Blocked",
- lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
- lock->client_pid,
- lkowner_utoa (&lock->owner),
- lock->user_flock.l_start,
- lock->user_flock.l_len);
- lock->blocked = 1;
- __insert_lock (pl_inode, lock);
- ret = -1;
- } else {
- gf_log (this->name, GF_LOG_TRACE,
- "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => NOK",
- lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
- lock->client_pid,
- lkowner_utoa (&lock->owner),
- lock->user_flock.l_start,
- lock->user_flock.l_len);
- errno = EAGAIN;
- ret = -1;
- }
+ return error;
+}
+
+int32_t
+pl_inode_remove_complete(xlator_t *xl, pl_inode_t *pl_inode, call_stub_t *stub,
+ struct list_head *contend)
+{
+ pl_inode_lock_t *lock;
+ int32_t error = -1;
+
+ if (stub != NULL) {
+ list_add_tail(&stub->list, &pl_inode->waiting);
+ pl_inode->is_locked = _gf_true;
+ } else {
+ error = ENOMEM;
+
+ while (!list_empty(contend)) {
+ lock = list_first_entry(contend, pl_inode_lock_t, list);
+ list_del_init(&lock->list);
+ __pl_inodelk_unref(lock);
}
- pthread_mutex_unlock (&pl_inode->mutex);
+ }
- grant_blocked_locks (this, pl_inode);
+ pthread_mutex_unlock(&pl_inode->mutex);
- do_blocked_rw (pl_inode);
+ if (error < 0) {
+ inodelk_contention_notify(xl, contend);
+ }
-out:
- return ret;
+ inode_unref(pl_inode->inode);
+
+ return error;
}
+void
+pl_inode_remove_wake(struct list_head *list)
+{
+ call_stub_t *stub;
-posix_lock_t *
-pl_getlk (pl_inode_t *pl_inode, posix_lock_t *lock)
+ while (!list_empty(list)) {
+ stub = list_first_entry(list, call_stub_t, list);
+ list_del_init(&stub->list);
+
+ call_resume(stub);
+ }
+}
+
+void
+pl_inode_remove_cbk(xlator_t *xl, pl_inode_t *pl_inode, int32_t error)
{
- posix_lock_t *conf = NULL;
+ struct list_head contend, granted;
+ struct timespec now;
+ pl_dom_list_t *dom;
+
+ if (pl_inode == NULL) {
+ return;
+ }
+
+ INIT_LIST_HEAD(&contend);
+ INIT_LIST_HEAD(&granted);
+ timespec_now(&now);
- conf = first_conflicting_overlap (pl_inode, lock);
+ pthread_mutex_lock(&pl_inode->mutex);
- if (conf == NULL) {
- lock->fl_type = F_UNLCK;
- return lock;
+ if (error == 0) {
+ if (pl_inode->links >= 0) {
+ pl_inode->links--;
}
+ if (pl_inode->links == 0) {
+ pl_inode->removed = _gf_true;
+ }
+ }
+
+ pl_inode->remove_running--;
+
+ if ((pl_inode->remove_running == 0) && list_empty(&pl_inode->waiting)) {
+ pl_inode->is_locked = _gf_false;
+
+ list_for_each_entry(dom, &pl_inode->dom_list, inode_list)
+ {
+ __grant_blocked_inode_locks(xl, pl_inode, &granted, dom, &now,
+ &contend);
+ }
+ }
+
+ pthread_mutex_unlock(&pl_inode->mutex);
+
+ unwind_granted_inodes(xl, pl_inode, &granted);
- return conf;
+ inodelk_contention_notify(xl, &contend);
+
+ inode_unref(pl_inode->inode);
}
-gf_boolean_t
-pl_does_monkey_want_stuck_lock()
+void
+pl_inode_remove_unlocked(xlator_t *xl, pl_inode_t *pl_inode,
+ struct list_head *list)
{
- long int monkey_unlock_rand = 0;
- long int monkey_unlock_rand_rem = 0;
+ call_stub_t *stub, *tmp;
- /* coverty[DC.WEAK_CRYPTO] */
- monkey_unlock_rand = random ();
- monkey_unlock_rand_rem = monkey_unlock_rand % 100;
- if (monkey_unlock_rand_rem == 0)
- return _gf_true;
- return _gf_false;
+ if (!pl_inode->is_locked) {
+ return;
+ }
+
+ list_for_each_entry_safe(stub, tmp, &pl_inode->waiting, list)
+ {
+ if (!pl_inode_has_owners(xl, stub->frame->root->client, pl_inode, NULL,
+ NULL)) {
+ list_move_tail(&stub->list, list);
+ }
+ }
+}
+
+/* This function determines if an inodelk attempt can be done now or it needs
+ * to wait.
+ *
+ * Possible return values:
+ * < 0: An error occurred. Currently only -ESTALE can be returned if the
+ * inode has been deleted previously by unlink/rmdir/rename
+ * = 0: The lock can be attempted.
+ * > 0: The lock needs to wait because a conflicting remove operation is
+ * ongoing.
+ */
+int32_t
+pl_inode_remove_inodelk(pl_inode_t *pl_inode, pl_inode_lock_t *lock)
+{
+ pl_dom_list_t *dom;
+ pl_inode_lock_t *ilock;
+
+ /* If the inode has been deleted, we won't allow any lock. */
+ if (pl_inode->removed) {
+ return -ESTALE;
+ }
+
+ /* We only synchronize with locks made for regular operations coming from
+ * the user. Locks done for internal purposes are hard to control and could
+ * lead to long delays or deadlocks quite easily. */
+ if (lock->client_pid < 0) {
+ return 0;
+ }
+ if (!pl_inode->is_locked) {
+ return 0;
+ }
+ if (pl_inode->remove_running > 0) {
+ return 1;
+ }
+
+ list_for_each_entry(dom, &pl_inode->dom_list, inode_list)
+ {
+ list_for_each_entry(ilock, &dom->inodelk_list, list)
+ {
+ /* If a lock from the same client is already granted, we allow this
+ * one to continue. This is necessary to prevent deadlocks when
+ * multiple locks are taken for the same operation.
+ *
+ * On the other side it's unlikely that the same client sends
+ * completely unrelated locks for the same inode.
+ */
+ if (ilock->client == lock->client) {
+ return 0;
+ }
+ }
+ }
+
+ return 1;
}
diff --git a/xlators/features/locks/src/common.h b/xlators/features/locks/src/common.h
index 50c156feb38..281223bf3b8 100644
--- a/xlators/features/locks/src/common.h
+++ b/xlators/features/locks/src/common.h
@@ -10,166 +10,253 @@
#ifndef __COMMON_H__
#define __COMMON_H__
-#include "lkowner.h"
/*dump locks format strings */
-#define RANGE_FMT "type=%s, whence=%hd, start=%llu, len=%llu"
-#define ENTRY_FMT "type=%s on basename=%s"
-#define DUMP_GEN_FMT "pid = %llu, owner=%s, client=%p"
-#define GRNTD_AT "granted at %s"
-#define BLKD_AT "blocked at %s"
-#define CONN_ID "connection-id=%s"
-#define DUMP_BLKD_FMT DUMP_GEN_FMT", "CONN_ID", "BLKD_AT
-#define DUMP_GRNTD_FMT DUMP_GEN_FMT", "CONN_ID", "GRNTD_AT
-#define DUMP_BLKD_GRNTD_FMT DUMP_GEN_FMT", "CONN_ID", "BLKD_AT", "GRNTD_AT
-
-#define ENTRY_BLKD_FMT ENTRY_FMT", "DUMP_BLKD_FMT
-#define ENTRY_GRNTD_FMT ENTRY_FMT", "DUMP_GRNTD_FMT
-#define ENTRY_BLKD_GRNTD_FMT ENTRY_FMT", "DUMP_BLKD_GRNTD_FMT
-
-#define RANGE_BLKD_FMT RANGE_FMT", "DUMP_BLKD_FMT
-#define RANGE_GRNTD_FMT RANGE_FMT", "DUMP_GRNTD_FMT
-#define RANGE_BLKD_GRNTD_FMT RANGE_FMT", "DUMP_BLKD_GRNTD_FMT
+#define RANGE_FMT "type=%s, whence=%hd, start=%llu, len=%llu"
+#define ENTRY_FMT "type=%s on basename=%s"
+#define DUMP_GEN_FMT "pid = %llu, owner=%s, client=%p"
+#define GRNTD_AT "granted at %s"
+#define BLKD_AT "blocked at %s"
+#define CONN_ID "connection-id=%s"
+#define DUMP_BLKD_FMT DUMP_GEN_FMT ", " CONN_ID ", " BLKD_AT
+#define DUMP_GRNTD_FMT DUMP_GEN_FMT ", " CONN_ID ", " GRNTD_AT
+#define DUMP_BLKD_GRNTD_FMT DUMP_GEN_FMT ", " CONN_ID ", " BLKD_AT ", " GRNTD_AT
+
+#define ENTRY_BLKD_FMT ENTRY_FMT ", " DUMP_BLKD_FMT
+#define ENTRY_GRNTD_FMT ENTRY_FMT ", " DUMP_GRNTD_FMT
+#define ENTRY_BLKD_GRNTD_FMT ENTRY_FMT ", " DUMP_BLKD_GRNTD_FMT
+
+#define RANGE_BLKD_FMT RANGE_FMT ", " DUMP_BLKD_FMT
+#define RANGE_GRNTD_FMT RANGE_FMT ", " DUMP_GRNTD_FMT
+#define RANGE_BLKD_GRNTD_FMT RANGE_FMT ", " DUMP_BLKD_GRNTD_FMT
#define SET_FLOCK_PID(flock, lock) ((flock)->l_pid = lock->client_pid)
+#define PL_STACK_UNWIND_AND_FREE(__local, fop, frame, op_ret, params...) \
+ do { \
+ frame->local = NULL; \
+ STACK_UNWIND_STRICT(fop, frame, op_ret, params); \
+ if (__local) { \
+ if (__local->inodelk_dom_count_req) \
+ data_unref(__local->inodelk_dom_count_req); \
+ loc_wipe(&__local->loc[0]); \
+ loc_wipe(&__local->loc[1]); \
+ if (__local->fd) \
+ fd_unref(__local->fd); \
+ if (__local->inode) \
+ inode_unref(__local->inode); \
+ if (__local->xdata) { \
+ dict_unref(__local->xdata); \
+ __local->xdata = NULL; \
+ } \
+ mem_put(__local); \
+ } \
+ } while (0)
posix_lock_t *
-new_posix_lock (struct gf_flock *flock, client_t *client, pid_t client_pid,
- gf_lkowner_t *owner, fd_t *fd, uint32_t lk_flags,
- int can_block);
+new_posix_lock(struct gf_flock *flock, client_t *client, pid_t client_pid,
+ gf_lkowner_t *owner, fd_t *fd, uint32_t lk_flags, int blocking,
+ int32_t *op_errno);
pl_inode_t *
-pl_inode_get (xlator_t *this, inode_t *inode);
+pl_inode_get(xlator_t *this, inode_t *inode, pl_local_t *local);
posix_lock_t *
-pl_getlk (pl_inode_t *inode, posix_lock_t *lock);
+pl_getlk(pl_inode_t *inode, posix_lock_t *lock);
int
-pl_setlk (xlator_t *this, pl_inode_t *inode, posix_lock_t *lock,
- int can_block);
+pl_setlk(xlator_t *this, pl_inode_t *inode, posix_lock_t *lock, int can_block);
+
+int
+pl_lock_preempt(pl_inode_t *pl_inode, posix_lock_t *reqlock);
void
-grant_blocked_locks (xlator_t *this, pl_inode_t *inode);
+grant_blocked_locks(xlator_t *this, pl_inode_t *inode);
void
-posix_lock_to_flock (posix_lock_t *lock, struct gf_flock *flock);
+posix_lock_to_flock(posix_lock_t *lock, struct gf_flock *flock);
int
-locks_overlap (posix_lock_t *l1, posix_lock_t *l2);
+locks_overlap(posix_lock_t *l1, posix_lock_t *l2);
int
-same_owner (posix_lock_t *l1, posix_lock_t *l2);
+same_owner(posix_lock_t *l1, posix_lock_t *l2);
-void __delete_lock (posix_lock_t *);
+void
+__delete_lock(posix_lock_t *);
-void __destroy_lock (posix_lock_t *);
+void
+__destroy_lock(posix_lock_t *);
pl_dom_list_t *
-get_domain (pl_inode_t *pl_inode, const char *volume);
+get_domain(pl_inode_t *pl_inode, const char *volume);
+
+void
+grant_blocked_inode_locks(xlator_t *this, pl_inode_t *pl_inode,
+ pl_dom_list_t *dom, struct timespec *now,
+ struct list_head *contend);
void
-grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode,
- pl_dom_list_t *dom, struct timespec *now,
- struct list_head *contend);
+inodelk_contention_notify(xlator_t *this, struct list_head *contend);
void
-inodelk_contention_notify (xlator_t *this, struct list_head *contend);
+__delete_inode_lock(pl_inode_lock_t *lock);
void
-__delete_inode_lock (pl_inode_lock_t *lock);
+__pl_inodelk_unref(pl_inode_lock_t *lock);
void
-__pl_inodelk_unref (pl_inode_lock_t *lock);
+__grant_blocked_inode_locks(xlator_t *this, pl_inode_t *pl_inode,
+ struct list_head *granted, pl_dom_list_t *dom,
+ struct timespec *now, struct list_head *contend);
void
-grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode,
- pl_dom_list_t *dom, struct timespec *now,
- struct list_head *contend);
+unwind_granted_inodes(xlator_t *this, pl_inode_t *pl_inode,
+ struct list_head *granted);
void
-entrylk_contention_notify (xlator_t *this, struct list_head *contend);
+grant_blocked_entry_locks(xlator_t *this, pl_inode_t *pl_inode,
+ pl_dom_list_t *dom, struct timespec *now,
+ struct list_head *contend);
-void pl_update_refkeeper (xlator_t *this, inode_t *inode);
+void
+entrylk_contention_notify(xlator_t *this, struct list_head *contend);
+
+void
+pl_update_refkeeper(xlator_t *this, inode_t *inode);
int32_t
-__get_inodelk_count (xlator_t *this, pl_inode_t *pl_inode, char *domname);
+__get_inodelk_count(xlator_t *this, pl_inode_t *pl_inode, char *domname);
int32_t
-get_inodelk_count (xlator_t *this, inode_t *inode, char *domname);
+get_inodelk_count(xlator_t *this, inode_t *inode, char *domname);
int32_t
-__get_entrylk_count (xlator_t *this, pl_inode_t *pl_inode);
+__get_entrylk_count(xlator_t *this, pl_inode_t *pl_inode);
int32_t
-get_entrylk_count (xlator_t *this, inode_t *inode);
+get_entrylk_count(xlator_t *this, inode_t *inode);
-void pl_trace_in (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
- int cmd, struct gf_flock *flock, const char *domain);
+void
+pl_trace_in(xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc, int cmd,
+ struct gf_flock *flock, const char *domain);
-void pl_trace_out (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
- int cmd, struct gf_flock *flock, int op_ret, int op_errno, const char *domain);
+void
+pl_trace_out(xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc, int cmd,
+ struct gf_flock *flock, int op_ret, int op_errno,
+ const char *domain);
-void pl_trace_block (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
- int cmd, struct gf_flock *flock, const char *domain);
+void
+pl_trace_block(xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
+ int cmd, struct gf_flock *flock, const char *domain);
-void pl_trace_flush (xlator_t *this, call_frame_t *frame, fd_t *fd);
+void
+pl_trace_flush(xlator_t *this, call_frame_t *frame, fd_t *fd);
-void entrylk_trace_in (xlator_t *this, call_frame_t *frame, const char *volume,
- fd_t *fd, loc_t *loc, const char *basename,
- entrylk_cmd cmd, entrylk_type type);
+void
+entrylk_trace_in(xlator_t *this, call_frame_t *frame, const char *volume,
+ fd_t *fd, loc_t *loc, const char *basename, entrylk_cmd cmd,
+ entrylk_type type);
-void entrylk_trace_out (xlator_t *this, call_frame_t *frame, const char *volume,
- fd_t *fd, loc_t *loc, const char *basename,
- entrylk_cmd cmd, entrylk_type type,
- int op_ret, int op_errno);
+void
+entrylk_trace_out(xlator_t *this, call_frame_t *frame, const char *volume,
+ fd_t *fd, loc_t *loc, const char *basename, entrylk_cmd cmd,
+ entrylk_type type, int op_ret, int op_errno);
-void entrylk_trace_block (xlator_t *this, call_frame_t *frame, const char *volume,
- fd_t *fd, loc_t *loc, const char *basename,
- entrylk_cmd cmd, entrylk_type type);
+void
+entrylk_trace_block(xlator_t *this, call_frame_t *frame, const char *volume,
+ fd_t *fd, loc_t *loc, const char *basename, entrylk_cmd cmd,
+ entrylk_type type);
void
-pl_print_verdict (char *str, int size, int op_ret, int op_errno);
+pl_print_verdict(char *str, int size, int op_ret, int op_errno);
void
-pl_print_lockee (char *str, int size, fd_t *fd, loc_t *loc);
+pl_print_lockee(char *str, int size, fd_t *fd, loc_t *loc);
void
-pl_print_locker (char *str, int size, xlator_t *this, call_frame_t *frame);
+pl_print_locker(char *str, int size, xlator_t *this, call_frame_t *frame);
void
-pl_print_inodelk (char *str, int size, int cmd, struct gf_flock *flock, const char *domain);
+pl_print_inodelk(char *str, int size, int cmd, struct gf_flock *flock,
+ const char *domain);
void
-pl_trace_release (xlator_t *this, fd_t *fd);
+pl_trace_release(xlator_t *this, fd_t *fd);
unsigned long
-fd_to_fdnum (fd_t *fd);
+fd_to_fdnum(fd_t *fd);
fd_t *
-fd_from_fdnum (posix_lock_t *lock);
+fd_from_fdnum(posix_lock_t *lock);
int
-pl_reserve_setlk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
- int can_block);
+pl_reserve_setlk(xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
+ int can_block);
int
-reservelks_equal (posix_lock_t *l1, posix_lock_t *l2);
+reservelks_equal(posix_lock_t *l1, posix_lock_t *l2);
int
-pl_verify_reservelk (xlator_t *this, pl_inode_t *pl_inode,
- posix_lock_t *lock, int can_block);
+pl_verify_reservelk(xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
+ int can_block);
int
-pl_reserve_unlock (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *reqlock);
+pl_reserve_unlock(xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *reqlock);
int32_t
-check_entrylk_on_basename (xlator_t *this, inode_t *parent, char *basename);
+check_entrylk_on_basename(xlator_t *this, inode_t *parent, char *basename);
-void __pl_inodelk_unref (pl_inode_lock_t *lock);
-void __pl_entrylk_unref (pl_entry_lock_t *lock);
+void
+__pl_inodelk_unref(pl_inode_lock_t *lock);
+void
+__pl_entrylk_unref(pl_entry_lock_t *lock);
int
-pl_metalock_is_active (pl_inode_t *pl_inode);
+pl_metalock_is_active(pl_inode_t *pl_inode);
-int
-__pl_queue_lock (pl_inode_t *pl_inode, posix_lock_t *reqlock, int can_block);
+void
+__pl_queue_lock(pl_inode_t *pl_inode, posix_lock_t *reqlock);
+
+void
+inodelk_contention_notify_check(xlator_t *xl, pl_inode_lock_t *lock,
+ struct timespec *now,
+ struct list_head *contend);
+
+void
+entrylk_contention_notify_check(xlator_t *xl, pl_entry_lock_t *lock,
+ struct timespec *now,
+ struct list_head *contend);
gf_boolean_t
pl_does_monkey_want_stuck_lock();
+
+gf_boolean_t
+pl_is_mandatory_locking_enabled(pl_inode_t *pl_inode);
+
+void
+pl_clean_local(pl_local_t *local);
+
+int
+pl_local_init(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd);
+
+gf_boolean_t
+pl_is_lk_owner_valid(gf_lkowner_t *owner, client_t *client);
+
+int32_t
+pl_inode_remove_prepare(xlator_t *xl, call_frame_t *frame, loc_t *loc,
+ pl_inode_t **ppl_inode, struct list_head *contend);
+
+int32_t
+pl_inode_remove_complete(xlator_t *xl, pl_inode_t *pl_inode, call_stub_t *stub,
+ struct list_head *contend);
+
+void
+pl_inode_remove_wake(struct list_head *list);
+
+void
+pl_inode_remove_cbk(xlator_t *xl, pl_inode_t *pl_inode, int32_t error);
+
+void
+pl_inode_remove_unlocked(xlator_t *xl, pl_inode_t *pl_inode,
+ struct list_head *list);
+
+int32_t
+pl_inode_remove_inodelk(pl_inode_t *pl_inode, pl_inode_lock_t *lock);
+
#endif /* __COMMON_H__ */
diff --git a/xlators/features/locks/src/entrylk.c b/xlators/features/locks/src/entrylk.c
index 008d05a34c4..fd772c850dd 100644
--- a/xlators/features/locks/src/entrylk.c
+++ b/xlators/features/locks/src/entrylk.c
@@ -7,13 +7,13 @@
later), or the GNU General Public License, version 2 (GPLv2), in all
cases as published by the Free Software Foundation.
*/
-#include "glusterfs.h"
-#include "compat.h"
-#include "xlator.h"
-#include "logging.h"
-#include "common-utils.h"
-#include "list.h"
-#include "upcall-utils.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/compat.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/common-utils.h>
+#include <glusterfs/list.h>
+#include <glusterfs/upcall-utils.h>
#include "locks.h"
#include "clear.h"
@@ -21,59 +21,63 @@
#include "pl-messages.h"
void
-__pl_entrylk_unref (pl_entry_lock_t *lock)
+__pl_entrylk_unref(pl_entry_lock_t *lock)
{
- lock->ref--;
- if (!lock->ref) {
- GF_FREE ((char *)lock->basename);
- GF_FREE (lock->connection_id);
- GF_FREE (lock);
- }
+ lock->ref--;
+ if (!lock->ref) {
+ GF_FREE((char *)lock->basename);
+ GF_FREE(lock->connection_id);
+ GF_FREE(lock);
+ }
}
-
static void
-__pl_entrylk_ref (pl_entry_lock_t *lock)
+__pl_entrylk_ref(pl_entry_lock_t *lock)
{
- lock->ref++;
+ lock->ref++;
}
-
static pl_entry_lock_t *
-new_entrylk_lock (pl_inode_t *pinode, const char *basename, entrylk_type type,
- const char *domain, call_frame_t *frame, char *conn_id)
+new_entrylk_lock(pl_inode_t *pinode, const char *basename, entrylk_type type,
+ const char *domain, call_frame_t *frame, char *conn_id,
+ int32_t *op_errno)
{
- pl_entry_lock_t *newlock = NULL;
-
- newlock = GF_CALLOC (1, sizeof (pl_entry_lock_t),
- gf_locks_mt_pl_entry_lock_t);
- if (!newlock) {
- goto out;
- }
-
- newlock->basename = basename ? gf_strdup (basename) : NULL;
- newlock->type = type;
- newlock->client = frame->root->client;
- newlock->client_pid = frame->root->pid;
- newlock->volume = domain;
- newlock->owner = frame->root->lk_owner;
- newlock->frame = frame;
- newlock->this = frame->this;
-
- if (conn_id) {
- newlock->connection_id = gf_strdup (conn_id);
- }
-
- INIT_LIST_HEAD (&newlock->domain_list);
- INIT_LIST_HEAD (&newlock->blocked_locks);
- INIT_LIST_HEAD (&newlock->client_list);
-
- __pl_entrylk_ref (newlock);
+ pl_entry_lock_t *newlock = NULL;
+
+ if (!pl_is_lk_owner_valid(&frame->root->lk_owner, frame->root->client)) {
+ *op_errno = EINVAL;
+ goto out;
+ }
+
+ newlock = GF_CALLOC(1, sizeof(pl_entry_lock_t),
+ gf_locks_mt_pl_entry_lock_t);
+ if (!newlock) {
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ newlock->basename = basename ? gf_strdup(basename) : NULL;
+ newlock->type = type;
+ newlock->client = frame->root->client;
+ newlock->client_pid = frame->root->pid;
+ newlock->volume = domain;
+ newlock->owner = frame->root->lk_owner;
+ newlock->frame = frame;
+ newlock->this = frame->this;
+
+ if (conn_id) {
+ newlock->connection_id = gf_strdup(conn_id);
+ }
+
+ INIT_LIST_HEAD(&newlock->domain_list);
+ INIT_LIST_HEAD(&newlock->blocked_locks);
+ INIT_LIST_HEAD(&newlock->client_list);
+
+ __pl_entrylk_ref(newlock);
out:
- return newlock;
+ return newlock;
}
-
/**
* all_names - does a basename represent all names?
* @basename: name to check
@@ -88,228 +92,222 @@ out:
*/
static int
-names_conflict (const char *n1, const char *n2)
+names_conflict(const char *n1, const char *n2)
{
- return all_names (n1) || all_names (n2) || !strcmp (n1, n2);
+ return all_names(n1) || all_names(n2) || !strcmp(n1, n2);
}
-
static int
-__same_entrylk_owner (pl_entry_lock_t *l1, pl_entry_lock_t *l2)
+__same_entrylk_owner(pl_entry_lock_t *l1, pl_entry_lock_t *l2)
{
- return (is_same_lkowner (&l1->owner, &l2->owner) &&
- (l1->client == l2->client));
+ return (is_same_lkowner(&l1->owner, &l2->owner) &&
+ (l1->client == l2->client));
}
/* Just as in inodelk, allow conflicting name locks from same (lk_owner, conn)*/
static int
-__conflicting_entrylks (pl_entry_lock_t *l1, pl_entry_lock_t *l2)
+__conflicting_entrylks(pl_entry_lock_t *l1, pl_entry_lock_t *l2)
{
- if (names_conflict (l1->basename, l2->basename)
- && !__same_entrylk_owner (l1, l2))
- return 1;
+ if (names_conflict(l1->basename, l2->basename) &&
+ !__same_entrylk_owner(l1, l2))
+ return 1;
- return 0;
+ return 0;
}
/* See comments in inodelk.c for details */
static inline gf_boolean_t
-__stale_entrylk (xlator_t *this, pl_entry_lock_t *candidate_lock,
+__stale_entrylk(xlator_t *this, pl_entry_lock_t *candidate_lock,
pl_entry_lock_t *requested_lock, time_t *lock_age_sec)
{
- posix_locks_private_t *priv = NULL;
- struct timeval curr;
- gettimeofday (&curr, NULL);
-
- priv = this->private;
-
- /* Question: Should we just prune them all given the
- * chance? Or just the locks we are attempting to acquire?
- */
- if (names_conflict (candidate_lock->basename,
- requested_lock->basename)) {
- *lock_age_sec = curr.tv_sec -
- candidate_lock->granted_time.tv_sec;
- if (*lock_age_sec > priv->revocation_secs)
- return _gf_true;
- }
- return _gf_false;
+ posix_locks_private_t *priv = NULL;
+
+ priv = this->private;
+
+ /* Question: Should we just prune them all given the
+ * chance? Or just the locks we are attempting to acquire?
+ */
+ if (names_conflict(candidate_lock->basename, requested_lock->basename)) {
+ *lock_age_sec = gf_time() - candidate_lock->granted_time;
+ if (*lock_age_sec > priv->revocation_secs)
+ return _gf_true;
+ }
+ return _gf_false;
}
/* See comments in inodelk.c for details */
static gf_boolean_t
-__entrylk_prune_stale (xlator_t *this, pl_inode_t *pinode, pl_dom_list_t *dom,
- pl_entry_lock_t *lock)
+__entrylk_prune_stale(xlator_t *this, pl_inode_t *pinode, pl_dom_list_t *dom,
+ pl_entry_lock_t *lock)
{
- posix_locks_private_t *priv = NULL;
- pl_entry_lock_t *tmp = NULL;
- pl_entry_lock_t *lk = NULL;
- gf_boolean_t revoke_lock = _gf_false;
- int bcount = 0;
- int gcount = 0;
- int op_errno = 0;
- clrlk_args args;
- args.opts = NULL;
- time_t lk_age_sec = 0;
- uint32_t max_blocked = 0;
- char *reason_str = NULL;
-
- priv = this->private;
- args.type = CLRLK_ENTRY;
- if (priv->revocation_clear_all == _gf_true)
- args.kind = CLRLK_ALL;
- else
- args.kind = CLRLK_GRANTED;
-
-
- if (list_empty (&dom->entrylk_list))
- goto out;
-
- pthread_mutex_lock (&pinode->mutex);
- lock->pinode = pinode;
- list_for_each_entry_safe (lk, tmp, &dom->entrylk_list, domain_list) {
- if (__stale_entrylk (this, lk, lock, &lk_age_sec) == _gf_true) {
- revoke_lock = _gf_true;
- reason_str = "age";
- break;
- }
+ posix_locks_private_t *priv = NULL;
+ pl_entry_lock_t *tmp = NULL;
+ pl_entry_lock_t *lk = NULL;
+ gf_boolean_t revoke_lock = _gf_false;
+ int bcount = 0;
+ int gcount = 0;
+ int op_errno = 0;
+ clrlk_args args;
+ args.opts = NULL;
+ time_t lk_age_sec = 0;
+ uint32_t max_blocked = 0;
+ char *reason_str = NULL;
+
+ priv = this->private;
+ args.type = CLRLK_ENTRY;
+ if (priv->revocation_clear_all == _gf_true)
+ args.kind = CLRLK_ALL;
+ else
+ args.kind = CLRLK_GRANTED;
+
+ if (list_empty(&dom->entrylk_list))
+ goto out;
+
+ pthread_mutex_lock(&pinode->mutex);
+ lock->pinode = pinode;
+ list_for_each_entry_safe(lk, tmp, &dom->entrylk_list, domain_list)
+ {
+ if (__stale_entrylk(this, lk, lock, &lk_age_sec) == _gf_true) {
+ revoke_lock = _gf_true;
+ reason_str = "age";
+ break;
}
- max_blocked = priv->revocation_max_blocked;
- if (max_blocked != 0 && revoke_lock == _gf_false) {
- list_for_each_entry_safe (lk, tmp, &dom->blocked_entrylks,
- blocked_locks) {
- max_blocked--;
- if (max_blocked == 0) {
- revoke_lock = _gf_true;
- reason_str = "max blocked";
- break;
- }
- }
+ }
+ max_blocked = priv->revocation_max_blocked;
+ if (max_blocked != 0 && revoke_lock == _gf_false) {
+ list_for_each_entry_safe(lk, tmp, &dom->blocked_entrylks, blocked_locks)
+ {
+ max_blocked--;
+ if (max_blocked == 0) {
+ revoke_lock = _gf_true;
+ reason_str = "max blocked";
+ break;
+ }
}
- pthread_mutex_unlock (&pinode->mutex);
+ }
+ pthread_mutex_unlock(&pinode->mutex);
out:
- if (revoke_lock == _gf_true) {
- clrlk_clear_entrylk (this, pinode, dom, &args, &bcount, &gcount,
- &op_errno);
- gf_log (this->name, GF_LOG_WARNING,
- "Lock revocation [reason: %s; gfid: %s; domain: %s; "
- "age: %ld sec] - Entry lock revoked: %d granted & %d "
- "blocked locks cleared", reason_str,
- uuid_utoa (pinode->gfid), dom->domain, lk_age_sec,
- gcount, bcount);
- }
-
- return revoke_lock;
+ if (revoke_lock == _gf_true) {
+ clrlk_clear_entrylk(this, pinode, dom, &args, &bcount, &gcount,
+ &op_errno);
+ gf_log(this->name, GF_LOG_WARNING,
+ "Lock revocation [reason: %s; gfid: %s; domain: %s; "
+ "age: %ld sec] - Entry lock revoked: %d granted & %d "
+ "blocked locks cleared",
+ reason_str, uuid_utoa(pinode->gfid), dom->domain, lk_age_sec,
+ gcount, bcount);
+ }
+
+ return revoke_lock;
}
-static gf_boolean_t
-__entrylk_needs_contention_notify(xlator_t *this, pl_entry_lock_t *lock,
- struct timespec *now)
+void
+entrylk_contention_notify_check(xlator_t *this, pl_entry_lock_t *lock,
+ struct timespec *now, struct list_head *contend)
{
- posix_locks_private_t *priv;
- int64_t elapsed;
+ posix_locks_private_t *priv;
+ int64_t elapsed;
- priv = this->private;
+ priv = this->private;
- /* If this lock is in a list, it means that we are about to send a
- * notification for it, so no need to do anything else. */
- if (!list_empty(&lock->contend)) {
- return _gf_false;
- }
-
- elapsed = now->tv_sec;
- elapsed -= lock->contention_time.tv_sec;
- if (now->tv_nsec < lock->contention_time.tv_nsec) {
- elapsed--;
- }
- if (elapsed < priv->notify_contention_delay) {
- return _gf_false;
- }
+ /* If this lock is in a list, it means that we are about to send a
+ * notification for it, so no need to do anything else. */
+ if (!list_empty(&lock->contend)) {
+ return;
+ }
+
+ elapsed = now->tv_sec;
+ elapsed -= lock->contention_time.tv_sec;
+ if (now->tv_nsec < lock->contention_time.tv_nsec) {
+ elapsed--;
+ }
+ if (elapsed < priv->notify_contention_delay) {
+ return;
+ }
- /* All contention notifications will be sent outside of the locked
- * region. This means that currently granted locks might have already
- * been unlocked by that time. To avoid the lock or the inode to be
- * destroyed before we process them, we take an additional reference
- * on both. */
- inode_ref(lock->pinode->inode);
- __pl_entrylk_ref(lock);
+ /* All contention notifications will be sent outside of the locked
+ * region. This means that currently granted locks might have already
+ * been unlocked by that time. To avoid the lock or the inode to be
+ * destroyed before we process them, we take an additional reference
+ * on both. */
+ inode_ref(lock->pinode->inode);
+ __pl_entrylk_ref(lock);
- lock->contention_time = *now;
+ lock->contention_time = *now;
- return _gf_true;
+ list_add_tail(&lock->contend, contend);
}
void
entrylk_contention_notify(xlator_t *this, struct list_head *contend)
{
- struct gf_upcall up;
- struct gf_upcall_entrylk_contention lc;
- pl_entry_lock_t *lock;
- pl_inode_t *pl_inode;
- client_t *client;
- gf_boolean_t notify;
-
- while (!list_empty(contend)) {
- lock = list_first_entry(contend, pl_entry_lock_t, contend);
-
- pl_inode = lock->pinode;
-
- pthread_mutex_lock(&pl_inode->mutex);
-
- /* If the lock has already been released, no notification is
- * sent. We clear the notification time in this case. */
- notify = !list_empty(&lock->domain_list);
- if (!notify) {
- lock->contention_time.tv_sec = 0;
- lock->contention_time.tv_nsec = 0;
- } else {
- lc.type = lock->type;
- lc.name = lock->basename;
- lc.pid = lock->client_pid;
- lc.domain = lock->volume;
- lc.xdata = NULL;
-
- gf_uuid_copy(up.gfid, lock->pinode->gfid);
- client = (client_t *)lock->client;
- if (client == NULL) {
- /* A NULL client can be found if the entrylk
- * was issued by a server side xlator. */
- up.client_uid = NULL;
- } else {
- up.client_uid = client->client_uid;
- }
- }
+ struct gf_upcall up;
+ struct gf_upcall_entrylk_contention lc;
+ pl_entry_lock_t *lock;
+ pl_inode_t *pl_inode;
+ client_t *client;
+ gf_boolean_t notify;
+
+ while (!list_empty(contend)) {
+ lock = list_first_entry(contend, pl_entry_lock_t, contend);
+
+ pl_inode = lock->pinode;
+
+ pthread_mutex_lock(&pl_inode->mutex);
+
+ /* If the lock has already been released, no notification is
+ * sent. We clear the notification time in this case. */
+ notify = !list_empty(&lock->domain_list);
+ if (!notify) {
+ lock->contention_time.tv_sec = 0;
+ lock->contention_time.tv_nsec = 0;
+ } else {
+ lc.type = lock->type;
+ lc.name = lock->basename;
+ lc.pid = lock->client_pid;
+ lc.domain = lock->volume;
+ lc.xdata = NULL;
+
+ gf_uuid_copy(up.gfid, lock->pinode->gfid);
+ client = (client_t *)lock->client;
+ if (client == NULL) {
+ /* A NULL client can be found if the entrylk
+ * was issued by a server side xlator. */
+ up.client_uid = NULL;
+ } else {
+ up.client_uid = client->client_uid;
+ }
+ }
- pthread_mutex_unlock(&pl_inode->mutex);
-
- if (notify) {
- up.event_type = GF_UPCALL_ENTRYLK_CONTENTION;
- up.data = &lc;
-
- if (this->notify(this, GF_EVENT_UPCALL, &up) < 0) {
- gf_msg_debug(this->name, 0,
- "Entrylk contention notification "
- "failed");
- } else {
- gf_msg_debug(this->name, 0,
- "Entrylk contention notification "
- "sent");
- }
- }
+ pthread_mutex_unlock(&pl_inode->mutex);
+
+ if (notify) {
+ up.event_type = GF_UPCALL_ENTRYLK_CONTENTION;
+ up.data = &lc;
+
+ if (this->notify(this, GF_EVENT_UPCALL, &up) < 0) {
+ gf_msg_debug(this->name, 0,
+ "Entrylk contention notification "
+ "failed");
+ } else {
+ gf_msg_debug(this->name, 0,
+ "Entrylk contention notification "
+ "sent");
+ }
+ }
- pthread_mutex_lock(&pl_inode->mutex);
+ pthread_mutex_lock(&pl_inode->mutex);
- list_del_init(&lock->contend);
- __pl_entrylk_unref(lock);
+ list_del_init(&lock->contend);
+ __pl_entrylk_unref(lock);
- pthread_mutex_unlock(&pl_inode->mutex);
+ pthread_mutex_unlock(&pl_inode->mutex);
- inode_unref(pl_inode->inode);
- }
+ inode_unref(pl_inode->inode);
+ }
}
-
/**
* entrylk_grantable - is this lock grantable?
* @inode: inode in which to look
@@ -317,189 +315,188 @@ entrylk_contention_notify(xlator_t *this, struct list_head *contend)
* @type: type of lock
*/
static pl_entry_lock_t *
-__entrylk_grantable (xlator_t *this, pl_dom_list_t *dom, pl_entry_lock_t *lock,
- struct timespec *now, struct list_head *contend)
+__entrylk_grantable(xlator_t *this, pl_dom_list_t *dom, pl_entry_lock_t *lock,
+ struct timespec *now, struct list_head *contend)
{
- pl_entry_lock_t *tmp = NULL;
- pl_entry_lock_t *ret = NULL;
-
- list_for_each_entry (tmp, &dom->entrylk_list, domain_list) {
- if (__conflicting_entrylks (tmp, lock)) {
- if (ret == NULL) {
- ret = tmp;
- if (contend == NULL) {
- break;
- }
- }
- if (__entrylk_needs_contention_notify(this, tmp, now)) {
- list_add_tail(&tmp->contend, contend);
- }
+ pl_entry_lock_t *tmp = NULL;
+ pl_entry_lock_t *ret = NULL;
+
+ list_for_each_entry(tmp, &dom->entrylk_list, domain_list)
+ {
+ if (__conflicting_entrylks(tmp, lock)) {
+ if (ret == NULL) {
+ ret = tmp;
+ if (contend == NULL) {
+ break;
}
+ }
+ entrylk_contention_notify_check(this, tmp, now, contend);
}
+ }
- return ret;
+ return ret;
}
static pl_entry_lock_t *
-__blocked_entrylk_conflict (pl_dom_list_t *dom, pl_entry_lock_t *lock)
+__blocked_entrylk_conflict(pl_dom_list_t *dom, pl_entry_lock_t *lock)
{
- pl_entry_lock_t *tmp = NULL;
+ pl_entry_lock_t *tmp = NULL;
- list_for_each_entry (tmp, &dom->blocked_entrylks, blocked_locks) {
- if (names_conflict (tmp->basename, lock->basename))
- return lock;
- }
+ list_for_each_entry(tmp, &dom->blocked_entrylks, blocked_locks)
+ {
+ if (names_conflict(tmp->basename, lock->basename))
+ return lock;
+ }
- return NULL;
+ return NULL;
}
static int
-__owner_has_lock (pl_dom_list_t *dom, pl_entry_lock_t *newlock)
+__owner_has_lock(pl_dom_list_t *dom, pl_entry_lock_t *newlock)
{
- pl_entry_lock_t *lock = NULL;
+ pl_entry_lock_t *lock = NULL;
- list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
- if (__same_entrylk_owner (lock, newlock))
- return 1;
- }
+ list_for_each_entry(lock, &dom->entrylk_list, domain_list)
+ {
+ if (__same_entrylk_owner(lock, newlock))
+ return 1;
+ }
- list_for_each_entry (lock, &dom->blocked_entrylks, blocked_locks) {
- if (__same_entrylk_owner (lock, newlock))
- return 1;
- }
+ list_for_each_entry(lock, &dom->blocked_entrylks, blocked_locks)
+ {
+ if (__same_entrylk_owner(lock, newlock))
+ return 1;
+ }
- return 0;
+ return 0;
}
static int
-names_equal (const char *n1, const char *n2)
+names_equal(const char *n1, const char *n2)
{
- return (n1 == NULL && n2 == NULL) || (n1 && n2 && !strcmp (n1, n2));
+ return (n1 == NULL && n2 == NULL) || (n1 && n2 && !strcmp(n1, n2));
}
void
-pl_print_entrylk (char *str, int size, entrylk_cmd cmd, entrylk_type type,
- const char *basename, const char *domain)
+pl_print_entrylk(char *str, int size, entrylk_cmd cmd, entrylk_type type,
+ const char *basename, const char *domain)
{
- char *cmd_str = NULL;
- char *type_str = NULL;
+ char *cmd_str = NULL;
+ char *type_str = NULL;
- switch (cmd) {
+ switch (cmd) {
case ENTRYLK_LOCK:
- cmd_str = "LOCK";
- break;
+ cmd_str = "LOCK";
+ break;
case ENTRYLK_LOCK_NB:
- cmd_str = "LOCK_NB";
- break;
+ cmd_str = "LOCK_NB";
+ break;
case ENTRYLK_UNLOCK:
- cmd_str = "UNLOCK";
- break;
+ cmd_str = "UNLOCK";
+ break;
default:
- cmd_str = "UNKNOWN";
- break;
- }
+ cmd_str = "UNKNOWN";
+ break;
+ }
- switch (type) {
+ switch (type) {
case ENTRYLK_RDLCK:
- type_str = "READ";
- break;
+ type_str = "READ";
+ break;
case ENTRYLK_WRLCK:
- type_str = "WRITE";
- break;
+ type_str = "WRITE";
+ break;
default:
- type_str = "UNKNOWN";
- break;
- }
+ type_str = "UNKNOWN";
+ break;
+ }
- snprintf (str, size, "lock=ENTRYLK, cmd=%s, type=%s, basename=%s, domain: %s",
- cmd_str, type_str, basename, domain);
+ snprintf(str, size,
+ "lock=ENTRYLK, cmd=%s, type=%s, basename=%s, domain: %s", cmd_str,
+ type_str, basename, domain);
}
-
void
-entrylk_trace_in (xlator_t *this, call_frame_t *frame, const char *domain,
- fd_t *fd, loc_t *loc, const char *basename,
- entrylk_cmd cmd, entrylk_type type)
+entrylk_trace_in(xlator_t *this, call_frame_t *frame, const char *domain,
+ fd_t *fd, loc_t *loc, const char *basename, entrylk_cmd cmd,
+ entrylk_type type)
{
- posix_locks_private_t *priv = NULL;
- char pl_locker[256];
- char pl_lockee[256];
- char pl_entrylk[256];
+ posix_locks_private_t *priv = NULL;
+ char pl_locker[256];
+ char pl_lockee[256];
+ char pl_entrylk[256];
- priv = this->private;
+ priv = this->private;
- if (!priv->trace)
- return;
+ if (!priv->trace)
+ return;
- pl_print_locker (pl_locker, 256, this, frame);
- pl_print_lockee (pl_lockee, 256, fd, loc);
- pl_print_entrylk (pl_entrylk, 256, cmd, type, basename, domain);
+ pl_print_locker(pl_locker, 256, this, frame);
+ pl_print_lockee(pl_lockee, 256, fd, loc);
+ pl_print_entrylk(pl_entrylk, 256, cmd, type, basename, domain);
- gf_log (this->name, GF_LOG_INFO,
- "[REQUEST] Locker = {%s} Lockee = {%s} Lock = {%s}",
- pl_locker, pl_lockee, pl_entrylk);
+ gf_log(this->name, GF_LOG_INFO,
+ "[REQUEST] Locker = {%s} Lockee = {%s} Lock = {%s}", pl_locker,
+ pl_lockee, pl_entrylk);
}
-
void
-entrylk_trace_out (xlator_t *this, call_frame_t *frame, const char *domain,
- fd_t *fd, loc_t *loc, const char *basename,
- entrylk_cmd cmd, entrylk_type type, int op_ret, int op_errno)
+entrylk_trace_out(xlator_t *this, call_frame_t *frame, const char *domain,
+ fd_t *fd, loc_t *loc, const char *basename, entrylk_cmd cmd,
+ entrylk_type type, int op_ret, int op_errno)
{
- posix_locks_private_t *priv = NULL;
- char pl_locker[256];
- char pl_lockee[256];
- char pl_entrylk[256];
- char verdict[32];
+ posix_locks_private_t *priv = NULL;
+ char pl_locker[256];
+ char pl_lockee[256];
+ char pl_entrylk[256];
+ char verdict[32];
- priv = this->private;
+ priv = this->private;
- if (!priv->trace)
- return;
+ if (!priv->trace)
+ return;
- pl_print_locker (pl_locker, 256, this, frame);
- pl_print_lockee (pl_lockee, 256, fd, loc);
- pl_print_entrylk (pl_entrylk, 256, cmd, type, basename, domain);
- pl_print_verdict (verdict, 32, op_ret, op_errno);
+ pl_print_locker(pl_locker, 256, this, frame);
+ pl_print_lockee(pl_lockee, 256, fd, loc);
+ pl_print_entrylk(pl_entrylk, 256, cmd, type, basename, domain);
+ pl_print_verdict(verdict, 32, op_ret, op_errno);
- gf_log (this->name, GF_LOG_INFO,
- "[%s] Locker = {%s} Lockee = {%s} Lock = {%s}",
- verdict, pl_locker, pl_lockee, pl_entrylk);
+ gf_log(this->name, GF_LOG_INFO,
+ "[%s] Locker = {%s} Lockee = {%s} Lock = {%s}", verdict, pl_locker,
+ pl_lockee, pl_entrylk);
}
-
void
-entrylk_trace_block (xlator_t *this, call_frame_t *frame, const char *volume,
- fd_t *fd, loc_t *loc, const char *basename,
- entrylk_cmd cmd, entrylk_type type)
+entrylk_trace_block(xlator_t *this, call_frame_t *frame, const char *volume,
+ fd_t *fd, loc_t *loc, const char *basename, entrylk_cmd cmd,
+ entrylk_type type)
{
- posix_locks_private_t *priv = NULL;
- char pl_locker[256];
- char pl_lockee[256];
- char pl_entrylk[256];
+ posix_locks_private_t *priv = NULL;
+ char pl_locker[256];
+ char pl_lockee[256];
+ char pl_entrylk[256];
- priv = this->private;
+ priv = this->private;
- if (!priv->trace)
- return;
+ if (!priv->trace)
+ return;
- pl_print_locker (pl_locker, 256, this, frame);
- pl_print_lockee (pl_lockee, 256, fd, loc);
- pl_print_entrylk (pl_entrylk, 256, cmd, type, basename, volume);
+ pl_print_locker(pl_locker, 256, this, frame);
+ pl_print_lockee(pl_lockee, 256, fd, loc);
+ pl_print_entrylk(pl_entrylk, 256, cmd, type, basename, volume);
- gf_log (this->name, GF_LOG_INFO,
- "[BLOCKED] Locker = {%s} Lockee = {%s} Lock = {%s}",
- pl_locker, pl_lockee, pl_entrylk);
+ gf_log(this->name, GF_LOG_INFO,
+ "[BLOCKED] Locker = {%s} Lockee = {%s} Lock = {%s}", pl_locker,
+ pl_lockee, pl_entrylk);
}
/**
- * __find_most_matching_lock - find the lock struct which most matches in order of:
- * lock on the exact basename ||
- * an all_names lock
+ * __find_most_matching_lock - find the lock struct which most matches in order
+ * of: lock on the exact basename || an all_names lock
*
*
* @inode: inode in which to look
@@ -507,58 +504,57 @@ entrylk_trace_block (xlator_t *this, call_frame_t *frame, const char *volume,
*/
static pl_entry_lock_t *
-__find_most_matching_lock (pl_dom_list_t *dom, const char *basename)
+__find_most_matching_lock(pl_dom_list_t *dom, const char *basename)
{
- pl_entry_lock_t *lock;
- pl_entry_lock_t *all = NULL;
- pl_entry_lock_t *exact = NULL;
-
- if (list_empty (&dom->entrylk_list))
- return NULL;
-
- list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
- if (all_names (lock->basename))
- all = lock;
- else if (names_equal (lock->basename, basename))
- exact = lock;
- }
+ pl_entry_lock_t *lock;
+ pl_entry_lock_t *all = NULL;
+ pl_entry_lock_t *exact = NULL;
- return (exact ? exact : all);
+ if (list_empty(&dom->entrylk_list))
+ return NULL;
+
+ list_for_each_entry(lock, &dom->entrylk_list, domain_list)
+ {
+ if (all_names(lock->basename))
+ all = lock;
+ else if (names_equal(lock->basename, basename))
+ exact = lock;
+ }
+
+ return (exact ? exact : all);
}
-static pl_entry_lock_t*
-__find_matching_lock (pl_dom_list_t *dom, pl_entry_lock_t *lock)
+static pl_entry_lock_t *
+__find_matching_lock(pl_dom_list_t *dom, pl_entry_lock_t *lock)
{
- pl_entry_lock_t *tmp = NULL;
-
- list_for_each_entry (tmp, &dom->entrylk_list, domain_list) {
- if (names_equal (lock->basename, tmp->basename)
- && __same_entrylk_owner (lock, tmp)
- && (lock->type == tmp->type))
- return tmp;
- }
- return NULL;
+ pl_entry_lock_t *tmp = NULL;
+
+ list_for_each_entry(tmp, &dom->entrylk_list, domain_list)
+ {
+ if (names_equal(lock->basename, tmp->basename) &&
+ __same_entrylk_owner(lock, tmp) && (lock->type == tmp->type))
+ return tmp;
+ }
+ return NULL;
}
static int
__lock_blocked_add(xlator_t *this, pl_inode_t *pinode, pl_dom_list_t *dom,
pl_entry_lock_t *lock, int nonblock)
{
- struct timeval now;
-
- gettimeofday(&now, NULL);
+ if (nonblock)
+ goto out;
- if (nonblock)
- goto out;
-
- lock->blkd_time = now;
- list_add_tail (&lock->blocked_locks, &dom->blocked_entrylks);
+ lock->blkd_time = gf_time();
+ list_add_tail(&lock->blocked_locks, &dom->blocked_entrylks);
- gf_msg_trace (this->name, 0, "Blocking lock: {pinode=%p, basename=%s}",
- pinode, lock->basename);
+ gf_msg_trace(this->name, 0, "Blocking lock: {pinode=%p, basename=%s}",
+ pinode, lock->basename);
+ entrylk_trace_block(this, lock->frame, NULL, NULL, NULL, lock->basename,
+ ENTRYLK_LOCK, lock->type);
out:
- return -EAGAIN;
+ return -EAGAIN;
}
/**
@@ -573,49 +569,49 @@ out:
*/
int
-__lock_entrylk (xlator_t *this, pl_inode_t *pinode, pl_entry_lock_t *lock,
- int nonblock, pl_dom_list_t *dom, struct timespec *now,
- struct list_head *contend)
+__lock_entrylk(xlator_t *this, pl_inode_t *pinode, pl_entry_lock_t *lock,
+ int nonblock, pl_dom_list_t *dom, struct timespec *now,
+ struct list_head *contend)
{
- pl_entry_lock_t *conf = NULL;
- int ret = -EAGAIN;
-
- conf = __entrylk_grantable (this, dom, lock, now, contend);
- if (conf) {
- ret = __lock_blocked_add(this, pinode, dom, lock, nonblock);
- goto out;
+ pl_entry_lock_t *conf = NULL;
+ int ret = -EAGAIN;
+
+ conf = __entrylk_grantable(this, dom, lock, now, contend);
+ if (conf) {
+ ret = __lock_blocked_add(this, pinode, dom, lock, nonblock);
+ goto out;
+ }
+
+ /* To prevent blocked locks starvation, check if there are any blocked
+ * locks thay may conflict with this lock. If there is then don't grant
+ * the lock. BUT grant the lock if the owner already has lock to allow
+ * nested locks.
+ * Example: SHD from Machine1 takes (gfid, basename=257-length-name)
+ * and is granted.
+ * SHD from machine2 takes (gfid, basename=NULL) and is blocked.
+ * When SHD from Machine1 takes (gfid, basename=NULL) it needs to be
+ * granted, without which self-heal can't progress.
+ * TODO: Find why 'owner_has_lock' is checked even for blocked locks.
+ */
+ if (__blocked_entrylk_conflict(dom, lock) &&
+ !(__owner_has_lock(dom, lock))) {
+ if (nonblock == 0) {
+ gf_log(this->name, GF_LOG_DEBUG,
+ "Lock is grantable, but blocking to prevent "
+ "starvation");
}
- /* To prevent blocked locks starvation, check if there are any blocked
- * locks thay may conflict with this lock. If there is then don't grant
- * the lock. BUT grant the lock if the owner already has lock to allow
- * nested locks.
- * Example: SHD from Machine1 takes (gfid, basename=257-length-name)
- * and is granted.
- * SHD from machine2 takes (gfid, basename=NULL) and is blocked.
- * When SHD from Machine1 takes (gfid, basename=NULL) it needs to be
- * granted, without which self-heal can't progress.
- * TODO: Find why 'owner_has_lock' is checked even for blocked locks.
- */
- if (__blocked_entrylk_conflict (dom, lock) &&
- !(__owner_has_lock (dom, lock))) {
- if (nonblock == 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Lock is grantable, but blocking to prevent "
- "starvation");
- }
+ ret = __lock_blocked_add(this, pinode, dom, lock, nonblock);
+ goto out;
+ }
- ret = __lock_blocked_add(this, pinode, dom, lock, nonblock);
- goto out;
- }
+ __pl_entrylk_ref(lock);
+ lock->granted_time = gf_time();
+ list_add(&lock->domain_list, &dom->entrylk_list);
- __pl_entrylk_ref (lock);
- gettimeofday (&lock->granted_time, NULL);
- list_add (&lock->domain_list, &dom->entrylk_list);
-
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
/**
@@ -626,327 +622,322 @@ out:
*/
pl_entry_lock_t *
-__unlock_entrylk (pl_dom_list_t *dom, pl_entry_lock_t *lock)
+__unlock_entrylk(pl_dom_list_t *dom, pl_entry_lock_t *lock)
{
- pl_entry_lock_t *ret_lock = NULL;
+ pl_entry_lock_t *ret_lock = NULL;
- ret_lock = __find_matching_lock (dom, lock);
+ ret_lock = __find_matching_lock(dom, lock);
- if (ret_lock) {
- list_del_init (&ret_lock->domain_list);
- } else {
- gf_log ("locks", GF_LOG_ERROR, "unlock on %s "
- "(type=ENTRYLK_WRLCK) attempted but no matching lock "
- "found", lock->basename);
- }
+ if (ret_lock) {
+ list_del_init(&ret_lock->domain_list);
+ } else {
+ gf_log("locks", GF_LOG_ERROR,
+ "unlock on %s "
+ "(type=ENTRYLK_WRLCK) attempted but no matching lock "
+ "found",
+ lock->basename);
+ }
- return ret_lock;
+ return ret_lock;
}
int32_t
-check_entrylk_on_basename (xlator_t *this, inode_t *parent, char *basename)
+check_entrylk_on_basename(xlator_t *this, inode_t *parent, char *basename)
{
- int32_t entrylk = 0;
- pl_inode_t *pinode = 0;
- pl_dom_list_t *dom = NULL;
- pl_entry_lock_t *conf = NULL;
-
- pinode = pl_inode_get (this, parent);
- if (!pinode)
- goto out;
- pthread_mutex_lock (&pinode->mutex);
+ int32_t entrylk = 0;
+ pl_dom_list_t *dom = NULL;
+ pl_entry_lock_t *conf = NULL;
+
+ pl_inode_t *pinode = pl_inode_get(this, parent, NULL);
+ if (!pinode)
+ goto out;
+ pthread_mutex_lock(&pinode->mutex);
+ {
+ list_for_each_entry(dom, &pinode->dom_list, inode_list)
{
- list_for_each_entry (dom, &pinode->dom_list, inode_list) {
- conf = __find_most_matching_lock (dom, basename);
- if (conf && conf->basename) {
- entrylk = 1;
- break;
- }
- }
+ conf = __find_most_matching_lock(dom, basename);
+ if (conf && conf->basename) {
+ entrylk = 1;
+ break;
+ }
}
- pthread_mutex_unlock (&pinode->mutex);
+ }
+ pthread_mutex_unlock(&pinode->mutex);
out:
- return entrylk;
+ return entrylk;
}
void
-__grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode,
- pl_dom_list_t *dom, struct list_head *granted,
- struct timespec *now, struct list_head *contend)
+__grant_blocked_entry_locks(xlator_t *this, pl_inode_t *pl_inode,
+ pl_dom_list_t *dom, struct list_head *granted,
+ struct timespec *now, struct list_head *contend)
{
- int bl_ret = 0;
- pl_entry_lock_t *bl = NULL;
- pl_entry_lock_t *tmp = NULL;
-
- struct list_head blocked_list;
+ int bl_ret = 0;
+ pl_entry_lock_t *bl = NULL;
+ pl_entry_lock_t *tmp = NULL;
- INIT_LIST_HEAD (&blocked_list);
- list_splice_init (&dom->blocked_entrylks, &blocked_list);
+ struct list_head blocked_list;
- list_for_each_entry_safe (bl, tmp, &blocked_list, blocked_locks) {
+ INIT_LIST_HEAD(&blocked_list);
+ list_splice_init(&dom->blocked_entrylks, &blocked_list);
- list_del_init (&bl->blocked_locks);
+ list_for_each_entry_safe(bl, tmp, &blocked_list, blocked_locks)
+ {
+ list_del_init(&bl->blocked_locks);
- bl_ret = __lock_entrylk (bl->this, pl_inode, bl, 0, dom, now,
- contend);
+ bl_ret = __lock_entrylk(bl->this, pl_inode, bl, 0, dom, now, contend);
- if (bl_ret == 0) {
- list_add (&bl->blocked_locks, granted);
- }
+ if (bl_ret == 0) {
+ list_add_tail(&bl->blocked_locks, granted);
}
- return;
+ }
}
/* Grants locks if possible which are blocked on a lock */
void
-grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode,
- pl_dom_list_t *dom, struct timespec *now,
- struct list_head *contend)
+grant_blocked_entry_locks(xlator_t *this, pl_inode_t *pl_inode,
+ pl_dom_list_t *dom, struct timespec *now,
+ struct list_head *contend)
{
- struct list_head granted_list;
- pl_entry_lock_t *tmp = NULL;
- pl_entry_lock_t *lock = NULL;
-
- INIT_LIST_HEAD (&granted_list);
-
- pthread_mutex_lock (&pl_inode->mutex);
+ struct list_head granted_list;
+ pl_entry_lock_t *tmp = NULL;
+ pl_entry_lock_t *lock = NULL;
+
+ INIT_LIST_HEAD(&granted_list);
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ __grant_blocked_entry_locks(this, pl_inode, dom, &granted_list, now,
+ contend);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+
+ list_for_each_entry_safe(lock, tmp, &granted_list, blocked_locks)
+ {
+ entrylk_trace_out(this, lock->frame, NULL, NULL, NULL, lock->basename,
+ ENTRYLK_LOCK, lock->type, 0, 0);
+
+ STACK_UNWIND_STRICT(entrylk, lock->frame, 0, 0, NULL);
+ lock->frame = NULL;
+ }
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ list_for_each_entry_safe(lock, tmp, &granted_list, blocked_locks)
{
- __grant_blocked_entry_locks (this, pl_inode, dom,
- &granted_list, now, contend);
+ list_del_init(&lock->blocked_locks);
+ __pl_entrylk_unref(lock);
}
- pthread_mutex_unlock (&pl_inode->mutex);
-
- list_for_each_entry_safe (lock, tmp, &granted_list, blocked_locks) {
- entrylk_trace_out (this, lock->frame, NULL, NULL, NULL,
- lock->basename, ENTRYLK_LOCK, lock->type,
- 0, 0);
-
- STACK_UNWIND_STRICT (entrylk, lock->frame, 0, 0, NULL);
- lock->frame = NULL;
- }
-
- pthread_mutex_lock (&pl_inode->mutex);
- {
- list_for_each_entry_safe (lock, tmp, &granted_list, blocked_locks) {
- list_del_init (&lock->blocked_locks);
- __pl_entrylk_unref (lock);
- }
- }
- pthread_mutex_unlock (&pl_inode->mutex);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
}
-
/* Common entrylk code called by pl_entrylk and pl_fentrylk */
int
-pl_common_entrylk (call_frame_t *frame, xlator_t *this,
- const char *volume, inode_t *inode, const char *basename,
- entrylk_cmd cmd, entrylk_type type, loc_t *loc, fd_t *fd,
- dict_t *xdata)
+pl_common_entrylk(call_frame_t *frame, xlator_t *this, const char *volume,
+ inode_t *inode, const char *basename, entrylk_cmd cmd,
+ entrylk_type type, loc_t *loc, fd_t *fd, dict_t *xdata)
{
- int32_t op_ret = -1;
- int32_t op_errno = 0;
- int ret = -1;
- char unwind = 1;
- GF_UNUSED int dict_ret = -1;
- pl_inode_t *pinode = NULL;
- pl_entry_lock_t *reqlock = NULL;
- pl_entry_lock_t *unlocked = NULL;
- pl_dom_list_t *dom = NULL;
- char *conn_id = NULL;
- pl_ctx_t *ctx = NULL;
- int nonblock = 0;
- gf_boolean_t need_inode_unref = _gf_false;
- posix_locks_private_t *priv = NULL;
- struct list_head *pcontend = NULL;
- struct list_head contend;
- struct timespec now = { };
-
- priv = this->private;
-
- if (priv->notify_contention) {
- pcontend = &contend;
- INIT_LIST_HEAD(pcontend);
- timespec_now(&now);
+ int32_t op_ret = -1;
+ int32_t op_errno = 0;
+ int ret = -1;
+ char unwind = 1;
+ GF_UNUSED int dict_ret = -1;
+ pl_inode_t *pinode = NULL;
+ pl_entry_lock_t *reqlock = NULL;
+ pl_entry_lock_t *unlocked = NULL;
+ pl_dom_list_t *dom = NULL;
+ char *conn_id = NULL;
+ pl_ctx_t *ctx = NULL;
+ int nonblock = 0;
+ gf_boolean_t need_inode_unref = _gf_false;
+ posix_locks_private_t *priv = NULL;
+ struct list_head *pcontend = NULL;
+ struct list_head contend;
+ struct timespec now = {};
+
+ priv = this->private;
+
+ if (priv->notify_contention) {
+ pcontend = &contend;
+ INIT_LIST_HEAD(pcontend);
+ timespec_now(&now);
+ }
+
+ if (xdata)
+ dict_ret = dict_get_str(xdata, "connection-id", &conn_id);
+
+ pinode = pl_inode_get(this, inode, NULL);
+ if (!pinode) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ if (frame->root->client) {
+ ctx = pl_ctx_get(frame->root->client, this);
+ if (!ctx) {
+ op_errno = ENOMEM;
+ gf_log(this->name, GF_LOG_INFO, "pl_ctx_get() failed");
+ goto unwind;
}
-
- if (xdata)
- dict_ret = dict_get_str (xdata, "connection-id", &conn_id);
-
- pinode = pl_inode_get (this, inode);
- if (!pinode) {
- op_errno = ENOMEM;
- goto out;
- }
-
- if (frame->root->client) {
- ctx = pl_ctx_get (frame->root->client, this);
- if (!ctx) {
- op_errno = ENOMEM;
- gf_log (this->name, GF_LOG_INFO, "pl_ctx_get() failed");
- goto unwind;
- }
- }
-
- dom = get_domain (pinode, volume);
- if (!dom){
- op_errno = ENOMEM;
+ }
+
+ dom = get_domain(pinode, volume);
+ if (!dom) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ entrylk_trace_in(this, frame, volume, fd, loc, basename, cmd, type);
+
+ reqlock = new_entrylk_lock(pinode, basename, type, dom->domain, frame,
+ conn_id, &op_errno);
+ if (!reqlock) {
+ op_ret = -1;
+ goto unwind;
+ }
+
+ /* Ideally, AFTER a successful lock (both blocking and non-blocking) or
+ * an unsuccessful blocking lock operation, the inode needs to be ref'd.
+ *
+ * But doing so might give room to a race where the lock-requesting
+ * client could send a DISCONNECT just before this thread refs the inode
+ * after the locking is done, and the epoll thread could unref the inode
+ * in cleanup which means the inode's refcount would come down to 0, and
+ * the call to pl_forget() at this point destroys @pinode. Now when
+ * the io-thread executing this function tries to access pinode,
+ * it could crash on account of illegal memory access.
+ *
+ * To get around this problem, the inode is ref'd once even before
+ * adding the lock into client_list as a precautionary measure.
+ * This way even if there are DISCONNECTs, there will always be 1 extra
+ * ref on the inode, so @pinode is still alive until after the
+ * current stack unwinds.
+ */
+ pinode->inode = inode_ref(inode);
+ if (priv->revocation_secs != 0) {
+ if (cmd != ENTRYLK_UNLOCK) {
+ __entrylk_prune_stale(this, pinode, dom, reqlock);
+ } else if (priv->monkey_unlocking == _gf_true) {
+ if (pl_does_monkey_want_stuck_lock()) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "MONKEY LOCKING (forcing stuck lock)!");
+ op_ret = 0;
+ need_inode_unref = _gf_true;
+ pthread_mutex_lock(&pinode->mutex);
+ {
+ __pl_entrylk_unref(reqlock);
+ }
+ pthread_mutex_unlock(&pinode->mutex);
goto out;
+ }
}
+ }
- entrylk_trace_in (this, frame, volume, fd, loc, basename, cmd, type);
+ switch (cmd) {
+ case ENTRYLK_LOCK_NB:
+ nonblock = 1;
+ /* fall through */
+ case ENTRYLK_LOCK:
+ if (ctx)
+ pthread_mutex_lock(&ctx->lock);
+ pthread_mutex_lock(&pinode->mutex);
+ {
+ reqlock->pinode = pinode;
+
+ ret = __lock_entrylk(this, pinode, reqlock, nonblock, dom, &now,
+ pcontend);
+ if (ret == 0) {
+ reqlock->frame = NULL;
+ op_ret = 0;
+ } else {
+ op_errno = -ret;
+ }
- reqlock = new_entrylk_lock (pinode, basename, type, dom->domain, frame,
- conn_id);
- if (!reqlock) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
+ if (ctx && (!ret || !nonblock))
+ list_add(&reqlock->client_list, &ctx->entrylk_lockers);
- /* Ideally, AFTER a successful lock (both blocking and non-blocking) or
- * an unsuccessful blocking lock operation, the inode needs to be ref'd.
- *
- * But doing so might give room to a race where the lock-requesting
- * client could send a DISCONNECT just before this thread refs the inode
- * after the locking is done, and the epoll thread could unref the inode
- * in cleanup which means the inode's refcount would come down to 0, and
- * the call to pl_forget() at this point destroys @pinode. Now when
- * the io-thread executing this function tries to access pinode,
- * it could crash on account of illegal memory access.
- *
- * To get around this problem, the inode is ref'd once even before
- * adding the lock into client_list as a precautionary measure.
- * This way even if there are DISCONNECTs, there will always be 1 extra
- * ref on the inode, so @pinode is still alive until after the
- * current stack unwinds.
- */
- pinode->inode = inode_ref (inode);
- if (priv->revocation_secs != 0) {
- if (cmd != ENTRYLK_UNLOCK) {
- __entrylk_prune_stale (this, pinode, dom, reqlock);
- } else if (priv->monkey_unlocking == _gf_true) {
- if (pl_does_monkey_want_stuck_lock ()) {
- gf_log (this->name, GF_LOG_WARNING,
- "MONKEY LOCKING (forcing stuck lock)!");
- op_ret = 0;
- need_inode_unref = _gf_true;
- pthread_mutex_lock (&pinode->mutex);
- {
- __pl_entrylk_unref (reqlock);
- }
- pthread_mutex_unlock (&pinode->mutex);
- goto out;
- }
+ if (ret == -EAGAIN && !nonblock) {
+ /* blocked */
+ unwind = 0;
+ } else {
+ __pl_entrylk_unref(reqlock);
}
- }
- switch (cmd) {
- case ENTRYLK_LOCK_NB:
- nonblock = 1;
- /* fall through */
- case ENTRYLK_LOCK:
- if (ctx)
- pthread_mutex_lock (&ctx->lock);
- pthread_mutex_lock (&pinode->mutex);
- {
- reqlock->pinode = pinode;
-
- ret = __lock_entrylk (this, pinode, reqlock, nonblock,
- dom, &now, pcontend);
- if (ret == 0) {
- reqlock->frame = NULL;
- op_ret = 0;
- } else {
- op_errno = -ret;
- }
-
- if (ctx && (!ret || !nonblock))
- list_add (&reqlock->client_list,
- &ctx->entrylk_lockers);
-
- if (ret == -EAGAIN && !nonblock) {
- /* blocked */
- unwind = 0;
- } else {
- __pl_entrylk_unref (reqlock);
- }
-
- /* For all but the case where a non-blocking lock
- * attempt fails, the extra ref taken before the switch
- * block must be negated.
- */
- if ((ret == -EAGAIN) && (nonblock))
- need_inode_unref = _gf_true;
- }
- pthread_mutex_unlock (&pinode->mutex);
- if (ctx)
- pthread_mutex_unlock (&ctx->lock);
- break;
+ /* For all but the case where a non-blocking lock
+ * attempt fails, the extra ref taken before the switch
+ * block must be negated.
+ */
+ if ((ret == -EAGAIN) && (nonblock))
+ need_inode_unref = _gf_true;
+ }
+ pthread_mutex_unlock(&pinode->mutex);
+ if (ctx)
+ pthread_mutex_unlock(&ctx->lock);
+ break;
case ENTRYLK_UNLOCK:
- if (ctx)
- pthread_mutex_lock (&ctx->lock);
- pthread_mutex_lock (&pinode->mutex);
- {
- /* Irrespective of whether unlock succeeds or not,
- * the extra inode ref that was done before the switch
- * block must be negated. Towards this,
- * @need_inode_unref flag is set unconditionally here.
- */
- need_inode_unref = _gf_true;
- unlocked = __unlock_entrylk (dom, reqlock);
- if (unlocked) {
- list_del_init (&unlocked->client_list);
- __pl_entrylk_unref (unlocked);
- op_ret = 0;
- } else {
- op_errno = EINVAL;
- }
- __pl_entrylk_unref (reqlock);
+ if (ctx)
+ pthread_mutex_lock(&ctx->lock);
+ pthread_mutex_lock(&pinode->mutex);
+ {
+ /* Irrespective of whether unlock succeeds or not,
+ * the extra inode ref that was done before the switch
+ * block must be negated. Towards this,
+ * @need_inode_unref flag is set unconditionally here.
+ */
+ need_inode_unref = _gf_true;
+ unlocked = __unlock_entrylk(dom, reqlock);
+ if (unlocked) {
+ list_del_init(&unlocked->client_list);
+ __pl_entrylk_unref(unlocked);
+ op_ret = 0;
+ } else {
+ op_errno = EINVAL;
}
- pthread_mutex_unlock (&pinode->mutex);
- if (ctx)
- pthread_mutex_unlock (&ctx->lock);
+ __pl_entrylk_unref(reqlock);
+ }
+ pthread_mutex_unlock(&pinode->mutex);
+ if (ctx)
+ pthread_mutex_unlock(&ctx->lock);
- grant_blocked_entry_locks (this, pinode, dom, &now, pcontend);
+ grant_blocked_entry_locks(this, pinode, dom, &now, pcontend);
- break;
+ break;
default:
- need_inode_unref = _gf_true;
- gf_log (this->name, GF_LOG_ERROR,
- "Unexpected case in entrylk (cmd=%d). Please file"
- "a bug report at http://bugs.gluster.com", cmd);
- goto out;
- }
- /* The following (extra) unref corresponds to the ref that
- * was done at the time the lock was granted.
- */
- if ((cmd == ENTRYLK_UNLOCK) && (op_ret == 0))
- inode_unref (pinode->inode);
+ need_inode_unref = _gf_true;
+ gf_log(this->name, GF_LOG_ERROR,
+ "Unexpected case in entrylk (cmd=%d). Please file"
+ "a bug report at http://bugs.gluster.com",
+ cmd);
+ goto out;
+ }
+ /* The following (extra) unref corresponds to the ref that
+ * was done at the time the lock was granted.
+ */
+ if ((cmd == ENTRYLK_UNLOCK) && (op_ret == 0))
+ inode_unref(pinode->inode);
out:
- if (need_inode_unref)
- inode_unref (pinode->inode);
+ if (need_inode_unref)
+ inode_unref(pinode->inode);
- if (unwind) {
- entrylk_trace_out (this, frame, volume, fd, loc, basename,
- cmd, type, op_ret, op_errno);
-unwind:
- STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno, NULL);
- } else {
- entrylk_trace_block (this, frame, volume, fd, loc, basename,
- cmd, type);
- }
+ if (unwind) {
+ entrylk_trace_out(this, frame, volume, fd, loc, basename, cmd, type,
+ op_ret, op_errno);
+ unwind:
+ STACK_UNWIND_STRICT(entrylk, frame, op_ret, op_errno, NULL);
+ }
- if (pcontend != NULL) {
- entrylk_contention_notify(this, pcontend);
- }
+ if (pcontend != NULL) {
+ entrylk_contention_notify(this, pcontend);
+ }
- return 0;
+ return 0;
}
/**
@@ -956,17 +947,16 @@ unwind:
*/
int
-pl_entrylk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, const char *basename,
- entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
+pl_entrylk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata)
{
- pl_common_entrylk (frame, this, volume, loc->inode, basename, cmd,
- type, loc, NULL, xdata);
+ pl_common_entrylk(frame, this, volume, loc->inode, basename, cmd, type, loc,
+ NULL, xdata);
- return 0;
+ return 0;
}
-
/**
* pl_fentrylk:
*
@@ -974,191 +964,190 @@ pl_entrylk (call_frame_t *frame, xlator_t *this,
*/
int
-pl_fentrylk (call_frame_t *frame, xlator_t *this,
- const char *volume, fd_t *fd, const char *basename,
- entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
+pl_fentrylk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata)
{
- pl_common_entrylk (frame, this, volume, fd->inode, basename, cmd,
- type, NULL, fd, xdata);
+ pl_common_entrylk(frame, this, volume, fd->inode, basename, cmd, type, NULL,
+ fd, xdata);
- return 0;
+ return 0;
}
-
static void
-pl_entrylk_log_cleanup (pl_entry_lock_t *lock)
+pl_entrylk_log_cleanup(pl_entry_lock_t *lock)
{
- pl_inode_t *pinode = NULL;
+ pl_inode_t *pinode = NULL;
- pinode = lock->pinode;
+ pinode = lock->pinode;
- gf_log (THIS->name, GF_LOG_WARNING,
- "releasing lock on %s held by "
- "{client=%p, pid=%"PRId64" lk-owner=%s}",
- uuid_utoa (pinode->gfid), lock->client,
- (uint64_t) lock->client_pid, lkowner_utoa (&lock->owner));
+ gf_log(THIS->name, GF_LOG_WARNING,
+ "releasing lock on %s held by "
+ "{client=%p, pid=%" PRId64 " lk-owner=%s}",
+ uuid_utoa(pinode->gfid), lock->client, (uint64_t)lock->client_pid,
+ lkowner_utoa(&lock->owner));
}
-
/* Release all entrylks from this client */
int
-pl_entrylk_client_cleanup (xlator_t *this, pl_ctx_t *ctx)
+pl_entrylk_client_cleanup(xlator_t *this, pl_ctx_t *ctx)
{
- posix_locks_private_t *priv;
- pl_entry_lock_t *tmp = NULL;
- pl_entry_lock_t *l = NULL;
- pl_dom_list_t *dom = NULL;
- pl_inode_t *pinode = NULL;
- struct list_head *pcontend = NULL;
- struct list_head released;
- struct list_head unwind;
- struct list_head contend;
- struct timespec now = { };
-
- INIT_LIST_HEAD (&released);
- INIT_LIST_HEAD (&unwind);
-
- priv = this->private;
- if (priv->notify_contention) {
- pcontend = &contend;
- INIT_LIST_HEAD (pcontend);
- timespec_now(&now);
- }
-
- pthread_mutex_lock (&ctx->lock);
+ posix_locks_private_t *priv;
+ pl_entry_lock_t *tmp = NULL;
+ pl_entry_lock_t *l = NULL;
+ pl_dom_list_t *dom = NULL;
+ pl_inode_t *pinode = NULL;
+ struct list_head *pcontend = NULL;
+ struct list_head released;
+ struct list_head unwind;
+ struct list_head contend;
+ struct timespec now = {};
+
+ INIT_LIST_HEAD(&released);
+ INIT_LIST_HEAD(&unwind);
+
+ priv = this->private;
+ if (priv->notify_contention) {
+ pcontend = &contend;
+ INIT_LIST_HEAD(pcontend);
+ timespec_now(&now);
+ }
+
+ pthread_mutex_lock(&ctx->lock);
+ {
+ list_for_each_entry_safe(l, tmp, &ctx->entrylk_lockers, client_list)
{
- list_for_each_entry_safe (l, tmp, &ctx->entrylk_lockers,
- client_list) {
- pl_entrylk_log_cleanup (l);
-
- pinode = l->pinode;
-
- pthread_mutex_lock (&pinode->mutex);
- {
- /* If the entrylk object is part of granted list but not
- * blocked list, then perform the following actions:
- * i. delete the object from granted list;
- * ii. grant other locks (from other clients) that may
- * have been blocked on this entrylk; and
- * iii. unref the object.
- *
- * If the entrylk object (L1) is part of both granted
- * and blocked lists, then this means that a parallel
- * unlock on another entrylk (L2 say) may have 'granted'
- * L1 and added it to 'granted' list in
- * __grant_blocked_entry_locks() (although using the
- * 'blocked_locks' member). In that case, the cleanup
- * codepath must try and grant other overlapping
- * blocked entrylks from other clients, now that L1 is
- * out of their way and then unref L1 in the end, and
- * leave it to the other thread (the one executing
- * unlock codepath) to unwind L1's frame, delete it from
- * blocked_locks list, and perform the last unref on L1.
- *
- * If the entrylk object (L1) is part of blocked list
- * only, the cleanup code path must:
- * i. delete it from the blocked_locks list inside
- * this critical section,
- * ii. unwind its frame with EAGAIN,
- * iii. try and grant blocked entry locks from other
- * clients that were otherwise grantable, but were
- * blocked to avoid leaving L1 to starve forever.
- * iv. unref the object.
- */
- list_del_init (&l->client_list);
-
- if (!list_empty (&l->domain_list)) {
- list_del_init (&l->domain_list);
- list_add_tail (&l->client_list,
- &released);
- } else {
- list_del_init (&l->blocked_locks);
- list_add_tail (&l->client_list,
- &unwind);
- }
- }
- pthread_mutex_unlock (&pinode->mutex);
+ pl_entrylk_log_cleanup(l);
+
+ pinode = l->pinode;
+
+ pthread_mutex_lock(&pinode->mutex);
+ {
+ /* If the entrylk object is part of granted list but not
+ * blocked list, then perform the following actions:
+ * i. delete the object from granted list;
+ * ii. grant other locks (from other clients) that may
+ * have been blocked on this entrylk; and
+ * iii. unref the object.
+ *
+ * If the entrylk object (L1) is part of both granted
+ * and blocked lists, then this means that a parallel
+ * unlock on another entrylk (L2 say) may have 'granted'
+ * L1 and added it to 'granted' list in
+ * __grant_blocked_entry_locks() (although using the
+ * 'blocked_locks' member). In that case, the cleanup
+ * codepath must try and grant other overlapping
+ * blocked entrylks from other clients, now that L1 is
+ * out of their way and then unref L1 in the end, and
+ * leave it to the other thread (the one executing
+ * unlock codepath) to unwind L1's frame, delete it from
+ * blocked_locks list, and perform the last unref on L1.
+ *
+ * If the entrylk object (L1) is part of blocked list
+ * only, the cleanup code path must:
+ * i. delete it from the blocked_locks list inside
+ * this critical section,
+ * ii. unwind its frame with EAGAIN,
+ * iii. try and grant blocked entry locks from other
+ * clients that were otherwise grantable, but were
+ * blocked to avoid leaving L1 to starve forever.
+ * iv. unref the object.
+ */
+ list_del_init(&l->client_list);
+
+ if (!list_empty(&l->domain_list)) {
+ list_del_init(&l->domain_list);
+ list_add_tail(&l->client_list, &released);
+ } else {
+ list_del_init(&l->blocked_locks);
+ list_add_tail(&l->client_list, &unwind);
}
+ }
+ pthread_mutex_unlock(&pinode->mutex);
}
- pthread_mutex_unlock (&ctx->lock);
+ }
+ pthread_mutex_unlock(&ctx->lock);
- list_for_each_entry_safe (l, tmp, &unwind, client_list) {
- list_del_init (&l->client_list);
+ if (!list_empty(&unwind)) {
+ list_for_each_entry_safe(l, tmp, &unwind, client_list)
+ {
+ list_del_init(&l->client_list);
- if (l->frame)
- STACK_UNWIND_STRICT (entrylk, l->frame, -1, EAGAIN,
- NULL);
- list_add_tail (&l->client_list, &released);
+ if (l->frame)
+ STACK_UNWIND_STRICT(entrylk, l->frame, -1, EAGAIN, NULL);
+ list_add_tail(&l->client_list, &released);
}
+ }
- list_for_each_entry_safe (l, tmp, &released, client_list) {
- list_del_init (&l->client_list);
+ if (!list_empty(&released)) {
+ list_for_each_entry_safe(l, tmp, &released, client_list)
+ {
+ list_del_init(&l->client_list);
- pinode = l->pinode;
+ pinode = l->pinode;
- dom = get_domain (pinode, l->volume);
+ dom = get_domain(pinode, l->volume);
- grant_blocked_entry_locks (this, pinode, dom, &now, pcontend);
+ grant_blocked_entry_locks(this, pinode, dom, &now, pcontend);
- pthread_mutex_lock (&pinode->mutex);
- {
- __pl_entrylk_unref (l);
- }
- pthread_mutex_unlock (&pinode->mutex);
+ pthread_mutex_lock(&pinode->mutex);
+ {
+ __pl_entrylk_unref(l);
+ }
+ pthread_mutex_unlock(&pinode->mutex);
- inode_unref (pinode->inode);
+ inode_unref(pinode->inode);
}
+ }
- if (pcontend != NULL) {
- entrylk_contention_notify(this, pcontend);
- }
+ if (pcontend != NULL) {
+ entrylk_contention_notify(this, pcontend);
+ }
- return 0;
+ return 0;
}
-
int32_t
-__get_entrylk_count (xlator_t *this, pl_inode_t *pl_inode)
+__get_entrylk_count(xlator_t *this, pl_inode_t *pl_inode)
{
- int32_t count = 0;
- pl_entry_lock_t *lock = NULL;
- pl_dom_list_t *dom = NULL;
+ int32_t count = 0;
+ pl_entry_lock_t *lock = NULL;
+ pl_dom_list_t *dom = NULL;
- list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
- list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
- count++;
- }
-
- list_for_each_entry (lock, &dom->blocked_entrylks, blocked_locks) {
- count++;
- }
+ list_for_each_entry(dom, &pl_inode->dom_list, inode_list)
+ {
+ list_for_each_entry(lock, &dom->entrylk_list, domain_list) { count++; }
+ list_for_each_entry(lock, &dom->blocked_entrylks, blocked_locks)
+ {
+ count++;
}
+ }
- return count;
+ return count;
}
int32_t
-get_entrylk_count (xlator_t *this, inode_t *inode)
+get_entrylk_count(xlator_t *this, inode_t *inode)
{
- pl_inode_t *pl_inode = NULL;
- uint64_t tmp_pl_inode = 0;
- int ret = 0;
- int32_t count = 0;
+ pl_inode_t *pl_inode = NULL;
+ uint64_t tmp_pl_inode = 0;
+ int ret = 0;
+ int32_t count = 0;
- ret = inode_ctx_get (inode, this, &tmp_pl_inode);
- if (ret != 0) {
- goto out;
- }
+ ret = inode_ctx_get(inode, this, &tmp_pl_inode);
+ if (ret != 0) {
+ goto out;
+ }
- pl_inode = (pl_inode_t *)(long) tmp_pl_inode;
+ pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
- pthread_mutex_lock (&pl_inode->mutex);
- {
- count = __get_entrylk_count (this, pl_inode);
- }
- pthread_mutex_unlock (&pl_inode->mutex);
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ count = __get_entrylk_count(this, pl_inode);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
out:
- return count;
+ return count;
}
diff --git a/xlators/features/locks/src/inodelk.c b/xlators/features/locks/src/inodelk.c
index 890ac8b6d00..d4e51d6e0a1 100644
--- a/xlators/features/locks/src/inodelk.c
+++ b/xlators/features/locks/src/inodelk.c
@@ -7,129 +7,127 @@
later), or the GNU General Public License, version 2 (GPLv2), in all
cases as published by the Free Software Foundation.
*/
-#include "glusterfs.h"
-#include "compat.h"
-#include "xlator.h"
-#include "logging.h"
-#include "common-utils.h"
-#include "list.h"
-#include "upcall-utils.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/compat.h>
+#include <glusterfs/dict.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/list.h>
+#include <glusterfs/upcall-utils.h>
#include "locks.h"
#include "clear.h"
#include "common.h"
-#include "pl-messages.h"
void
-__delete_inode_lock (pl_inode_lock_t *lock)
+__delete_inode_lock(pl_inode_lock_t *lock)
{
- list_del_init (&lock->list);
+ list_del_init(&lock->list);
}
static void
-__pl_inodelk_ref (pl_inode_lock_t *lock)
+__pl_inodelk_ref(pl_inode_lock_t *lock)
{
- lock->ref++;
+ lock->ref++;
}
void
-__pl_inodelk_unref (pl_inode_lock_t *lock)
+__pl_inodelk_unref(pl_inode_lock_t *lock)
{
- lock->ref--;
- if (!lock->ref) {
- GF_FREE (lock->connection_id);
- GF_FREE (lock);
- }
+ lock->ref--;
+ if (!lock->ref) {
+ GF_FREE(lock->connection_id);
+ GF_FREE(lock);
+ }
}
-/* Check if 2 inodelks are conflicting on type. Only 2 shared locks don't conflict */
+/* Check if 2 inodelks are conflicting on type. Only 2 shared locks don't
+ * conflict */
static int
-inodelk_type_conflict (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
+inodelk_type_conflict(pl_inode_lock_t *l1, pl_inode_lock_t *l2)
{
- if (l2->fl_type == F_WRLCK || l1->fl_type == F_WRLCK)
- return 1;
+ if (l2->fl_type == F_WRLCK || l1->fl_type == F_WRLCK)
+ return 1;
- return 0;
+ return 0;
}
void
-pl_print_inodelk (char *str, int size, int cmd, struct gf_flock *flock, const char *domain)
+pl_print_inodelk(char *str, int size, int cmd, struct gf_flock *flock,
+ const char *domain)
{
- char *cmd_str = NULL;
- char *type_str = NULL;
+ char *cmd_str = NULL;
+ char *type_str = NULL;
- switch (cmd) {
+ switch (cmd) {
#if F_GETLK != F_GETLK64
case F_GETLK64:
#endif
case F_GETLK:
- cmd_str = "GETLK";
- break;
+ cmd_str = "GETLK";
+ break;
#if F_SETLK != F_SETLK64
case F_SETLK64:
#endif
case F_SETLK:
- cmd_str = "SETLK";
- break;
+ cmd_str = "SETLK";
+ break;
#if F_SETLKW != F_SETLKW64
case F_SETLKW64:
#endif
case F_SETLKW:
- cmd_str = "SETLKW";
- break;
+ cmd_str = "SETLKW";
+ break;
default:
- cmd_str = "UNKNOWN";
- break;
- }
+ cmd_str = "UNKNOWN";
+ break;
+ }
- switch (flock->l_type) {
+ switch (flock->l_type) {
case F_RDLCK:
- type_str = "READ";
- break;
+ type_str = "READ";
+ break;
case F_WRLCK:
- type_str = "WRITE";
- break;
+ type_str = "WRITE";
+ break;
case F_UNLCK:
- type_str = "UNLOCK";
- break;
+ type_str = "UNLOCK";
+ break;
default:
- type_str = "UNKNOWN";
- break;
- }
-
- snprintf (str, size, "lock=INODELK, cmd=%s, type=%s, "
- "domain: %s, start=%llu, len=%llu, pid=%llu",
- cmd_str, type_str, domain,
- (unsigned long long) flock->l_start,
- (unsigned long long) flock->l_len,
- (unsigned long long) flock->l_pid);
+ type_str = "UNKNOWN";
+ break;
+ }
+
+ snprintf(str, size,
+ "lock=INODELK, cmd=%s, type=%s, "
+ "domain: %s, start=%llu, len=%llu, pid=%llu",
+ cmd_str, type_str, domain, (unsigned long long)flock->l_start,
+ (unsigned long long)flock->l_len,
+ (unsigned long long)flock->l_pid);
}
/* Determine if the two inodelks overlap reach other's lock regions */
static int
-inodelk_overlap (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
+inodelk_overlap(pl_inode_lock_t *l1, pl_inode_lock_t *l2)
{
- return ((l1->fl_end >= l2->fl_start) &&
- (l2->fl_end >= l1->fl_start));
+ return ((l1->fl_end >= l2->fl_start) && (l2->fl_end >= l1->fl_start));
}
/* Returns true if the 2 inodelks have the same owner */
static int
-same_inodelk_owner (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
+same_inodelk_owner(pl_inode_lock_t *l1, pl_inode_lock_t *l2)
{
- return (is_same_lkowner (&l1->owner, &l2->owner) &&
- (l1->client == l2->client));
+ return (is_same_lkowner(&l1->owner, &l2->owner) &&
+ (l1->client == l2->client));
}
/* Returns true if the 2 inodelks conflict with each other */
static int
-inodelk_conflict (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
+inodelk_conflict(pl_inode_lock_t *l1, pl_inode_lock_t *l2)
{
- return (inodelk_overlap (l1, l2) &&
- inodelk_type_conflict (l1, l2));
+ return (inodelk_overlap(l1, l2) && inodelk_type_conflict(l1, l2));
}
/*
@@ -138,24 +136,21 @@ inodelk_conflict (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
* true if it exceeds the configured threshold, false otherwise.
*/
static inline gf_boolean_t
-__stale_inodelk (xlator_t *this, pl_inode_lock_t *candidate_lock,
+__stale_inodelk(xlator_t *this, pl_inode_lock_t *candidate_lock,
pl_inode_lock_t *requested_lock, time_t *lock_age_sec)
{
- posix_locks_private_t *priv = NULL;
- struct timeval curr;
-
- priv = this->private;
- gettimeofday (&curr, NULL);
- /* Question: Should we just prune them all given the
- * chance? Or just the locks we are attempting to acquire?
- */
- if (inodelk_conflict (candidate_lock, requested_lock)) {
- *lock_age_sec = curr.tv_sec -
- candidate_lock->granted_time.tv_sec;
- if (*lock_age_sec > priv->revocation_secs)
- return _gf_true;
- }
- return _gf_false;
+ posix_locks_private_t *priv = NULL;
+
+ priv = this->private;
+ /* Question: Should we just prune them all given the
+ * chance? Or just the locks we are attempting to acquire?
+ */
+ if (inodelk_conflict(candidate_lock, requested_lock)) {
+ *lock_age_sec = gf_time() - candidate_lock->granted_time;
+ if (*lock_age_sec > priv->revocation_secs)
+ return _gf_true;
+ }
+ return _gf_false;
}
/* Examine any locks held on this inode and potentially revoke the lock
@@ -167,992 +162,1013 @@ __stale_inodelk (xlator_t *this, pl_inode_lock_t *candidate_lock,
* are beyond a threshold.
*/
static gf_boolean_t
-__inodelk_prune_stale (xlator_t *this, pl_inode_t *pinode, pl_dom_list_t *dom,
- pl_inode_lock_t *lock)
+__inodelk_prune_stale(xlator_t *this, pl_inode_t *pinode, pl_dom_list_t *dom,
+ pl_inode_lock_t *lock)
{
- posix_locks_private_t *priv = NULL;
- pl_inode_lock_t *tmp = NULL;
- pl_inode_lock_t *lk = NULL;
- gf_boolean_t revoke_lock = _gf_false;
- int bcount = 0;
- int gcount = 0;
- int op_errno = 0;
- clrlk_args args;
- args.opts = NULL;
- time_t lk_age_sec = 0;
- uint32_t max_blocked = 0;
- char *reason_str = NULL;
-
- priv = this->private;
-
- args.type = CLRLK_INODE;
- if (priv->revocation_clear_all == _gf_true)
- args.kind = CLRLK_ALL;
- else
- args.kind = CLRLK_GRANTED;
-
- if (list_empty (&dom->inodelk_list))
- goto out;
-
- pthread_mutex_lock (&pinode->mutex);
- list_for_each_entry_safe (lk, tmp, &dom->inodelk_list, list) {
- if (__stale_inodelk (this, lk, lock, &lk_age_sec) == _gf_true) {
- revoke_lock = _gf_true;
- reason_str = "age";
- break;
- }
+ posix_locks_private_t *priv = NULL;
+ pl_inode_lock_t *tmp = NULL;
+ pl_inode_lock_t *lk = NULL;
+ gf_boolean_t revoke_lock = _gf_false;
+ int bcount = 0;
+ int gcount = 0;
+ int op_errno = 0;
+ clrlk_args args;
+ args.opts = NULL;
+ time_t lk_age_sec = 0;
+ uint32_t max_blocked = 0;
+ char *reason_str = NULL;
+
+ priv = this->private;
+
+ args.type = CLRLK_INODE;
+ if (priv->revocation_clear_all == _gf_true)
+ args.kind = CLRLK_ALL;
+ else
+ args.kind = CLRLK_GRANTED;
+
+ if (list_empty(&dom->inodelk_list))
+ goto out;
+
+ pthread_mutex_lock(&pinode->mutex);
+ list_for_each_entry_safe(lk, tmp, &dom->inodelk_list, list)
+ {
+ if (__stale_inodelk(this, lk, lock, &lk_age_sec) == _gf_true) {
+ revoke_lock = _gf_true;
+ reason_str = "age";
+ break;
}
+ }
- max_blocked = priv->revocation_max_blocked;
- if (max_blocked != 0 && revoke_lock == _gf_false) {
- list_for_each_entry_safe (lk, tmp, &dom->blocked_inodelks,
- blocked_locks) {
- max_blocked--;
- if (max_blocked == 0) {
- revoke_lock = _gf_true;
- reason_str = "max blocked";
- break;
- }
- }
+ max_blocked = priv->revocation_max_blocked;
+ if (max_blocked != 0 && revoke_lock == _gf_false) {
+ list_for_each_entry_safe(lk, tmp, &dom->blocked_inodelks, blocked_locks)
+ {
+ max_blocked--;
+ if (max_blocked == 0) {
+ revoke_lock = _gf_true;
+ reason_str = "max blocked";
+ break;
+ }
}
- pthread_mutex_unlock (&pinode->mutex);
+ }
+ pthread_mutex_unlock(&pinode->mutex);
out:
- if (revoke_lock == _gf_true) {
- clrlk_clear_inodelk (this, pinode, dom, &args, &bcount, &gcount,
- &op_errno);
- gf_log (this->name, GF_LOG_WARNING,
- "Lock revocation [reason: %s; gfid: %s; domain: %s; "
- "age: %ld sec] - Inode lock revoked: %d granted & %d "
- "blocked locks cleared",
- reason_str, uuid_utoa (pinode->gfid), dom->domain,
- lk_age_sec, gcount, bcount);
- }
- return revoke_lock;
+ if (revoke_lock == _gf_true) {
+ clrlk_clear_inodelk(this, pinode, dom, &args, &bcount, &gcount,
+ &op_errno);
+ gf_log(this->name, GF_LOG_WARNING,
+ "Lock revocation [reason: %s; gfid: %s; domain: %s; "
+ "age: %ld sec] - Inode lock revoked: %d granted & %d "
+ "blocked locks cleared",
+ reason_str, uuid_utoa(pinode->gfid), dom->domain, lk_age_sec,
+ gcount, bcount);
+ }
+ return revoke_lock;
}
-static gf_boolean_t
-__inodelk_needs_contention_notify(xlator_t *this, pl_inode_lock_t *lock,
- struct timespec *now)
+void
+inodelk_contention_notify_check(xlator_t *this, pl_inode_lock_t *lock,
+ struct timespec *now, struct list_head *contend)
{
- posix_locks_private_t *priv;
- int64_t elapsed;
+ posix_locks_private_t *priv;
+ int64_t elapsed;
- priv = this->private;
+ priv = this->private;
- /* If this lock is in a list, it means that we are about to send a
- * notification for it, so no need to do anything else. */
- if (!list_empty(&lock->contend)) {
- return _gf_false;
- }
-
- elapsed = now->tv_sec;
- elapsed -= lock->contention_time.tv_sec;
- if (now->tv_nsec < lock->contention_time.tv_nsec) {
- elapsed--;
- }
- if (elapsed < priv->notify_contention_delay) {
- return _gf_false;
- }
+ /* If this lock is in a list, it means that we are about to send a
+ * notification for it, so no need to do anything else. */
+ if (!list_empty(&lock->contend)) {
+ return;
+ }
+
+ elapsed = now->tv_sec;
+ elapsed -= lock->contention_time.tv_sec;
+ if (now->tv_nsec < lock->contention_time.tv_nsec) {
+ elapsed--;
+ }
+ if (elapsed < priv->notify_contention_delay) {
+ return;
+ }
- /* All contention notifications will be sent outside of the locked
- * region. This means that currently granted locks might have already
- * been unlocked by that time. To avoid the lock or the inode to be
- * destroyed before we process them, we take an additional reference
- * on both. */
- inode_ref(lock->pl_inode->inode);
- __pl_inodelk_ref(lock);
+ /* All contention notifications will be sent outside of the locked
+ * region. This means that currently granted locks might have already
+ * been unlocked by that time. To avoid the lock or the inode to be
+ * destroyed before we process them, we take an additional reference
+ * on both. */
+ inode_ref(lock->pl_inode->inode);
+ __pl_inodelk_ref(lock);
- lock->contention_time = *now;
+ lock->contention_time = *now;
- return _gf_true;
+ list_add_tail(&lock->contend, contend);
}
void
inodelk_contention_notify(xlator_t *this, struct list_head *contend)
{
- struct gf_upcall up;
- struct gf_upcall_inodelk_contention lc;
- pl_inode_lock_t *lock;
- pl_inode_t *pl_inode;
- client_t *client;
- gf_boolean_t notify;
-
- while (!list_empty(contend)) {
- lock = list_first_entry(contend, pl_inode_lock_t, contend);
-
- pl_inode = lock->pl_inode;
+ struct gf_upcall up;
+ struct gf_upcall_inodelk_contention lc;
+ pl_inode_lock_t *lock;
+ pl_inode_t *pl_inode;
+ client_t *client;
+ gf_boolean_t notify;
- pthread_mutex_lock(&pl_inode->mutex);
+ while (!list_empty(contend)) {
+ lock = list_first_entry(contend, pl_inode_lock_t, contend);
- /* If the lock has already been released, no notification is
- * sent. We clear the notification time in this case. */
- notify = !list_empty(&lock->list);
- if (!notify) {
- lock->contention_time.tv_sec = 0;
- lock->contention_time.tv_nsec = 0;
- } else {
- memcpy(&lc.flock, &lock->user_flock, sizeof(lc.flock));
- lc.pid = lock->client_pid;
- lc.domain = lock->volume;
- lc.xdata = NULL;
-
- gf_uuid_copy(up.gfid, lock->pl_inode->gfid);
- client = (client_t *)lock->client;
- if (client == NULL) {
- /* A NULL client can be found if the inodelk
- * was issued by a server side xlator. */
- up.client_uid = NULL;
- } else {
- up.client_uid = client->client_uid;
- }
- }
+ pl_inode = lock->pl_inode;
- pthread_mutex_unlock(&pl_inode->mutex);
+ pthread_mutex_lock(&pl_inode->mutex);
+
+ /* If the lock has already been released, no notification is
+ * sent. We clear the notification time in this case. */
+ notify = !list_empty(&lock->list);
+ if (!notify) {
+ lock->contention_time.tv_sec = 0;
+ lock->contention_time.tv_nsec = 0;
+ } else {
+ memcpy(&lc.flock, &lock->user_flock, sizeof(lc.flock));
+ lc.pid = lock->client_pid;
+ lc.domain = lock->volume;
+ lc.xdata = NULL;
+
+ gf_uuid_copy(up.gfid, lock->pl_inode->gfid);
+ client = (client_t *)lock->client;
+ if (client == NULL) {
+ /* A NULL client can be found if the inodelk
+ * was issued by a server side xlator. */
+ up.client_uid = NULL;
+ } else {
+ up.client_uid = client->client_uid;
+ }
+ }
- if (notify) {
- up.event_type = GF_UPCALL_INODELK_CONTENTION;
- up.data = &lc;
-
- if (this->notify(this, GF_EVENT_UPCALL, &up) < 0) {
- gf_msg_debug(this->name, 0,
- "Inodelk contention notification "
- "failed");
- } else {
- gf_msg_debug(this->name, 0,
- "Inodelk contention notification "
- "sent");
- }
- }
+ pthread_mutex_unlock(&pl_inode->mutex);
+
+ if (notify) {
+ up.event_type = GF_UPCALL_INODELK_CONTENTION;
+ up.data = &lc;
+
+ if (this->notify(this, GF_EVENT_UPCALL, &up) < 0) {
+ gf_msg_debug(this->name, 0,
+ "Inodelk contention notification "
+ "failed");
+ } else {
+ gf_msg_debug(this->name, 0,
+ "Inodelk contention notification "
+ "sent");
+ }
+ }
- pthread_mutex_lock(&pl_inode->mutex);
+ pthread_mutex_lock(&pl_inode->mutex);
- list_del_init(&lock->contend);
- __pl_inodelk_unref(lock);
+ list_del_init(&lock->contend);
+ __pl_inodelk_unref(lock);
- pthread_mutex_unlock(&pl_inode->mutex);
+ pthread_mutex_unlock(&pl_inode->mutex);
- inode_unref(pl_inode->inode);
- }
+ inode_unref(pl_inode->inode);
+ }
}
/* Determine if lock is grantable or not */
static pl_inode_lock_t *
-__inodelk_grantable (xlator_t *this, pl_dom_list_t *dom, pl_inode_lock_t *lock,
- struct timespec *now, struct list_head *contend)
+__inodelk_grantable(xlator_t *this, pl_dom_list_t *dom, pl_inode_lock_t *lock,
+ struct timespec *now, struct list_head *contend)
{
- pl_inode_lock_t *l = NULL;
- pl_inode_lock_t *ret = NULL;
-
- list_for_each_entry (l, &dom->inodelk_list, list){
- if (inodelk_conflict (lock, l) &&
- !same_inodelk_owner (lock, l)) {
- if (ret == NULL) {
- ret = l;
- if (contend == NULL) {
- break;
- }
- }
- if (__inodelk_needs_contention_notify(this, l, now)) {
- list_add_tail(&l->contend, contend);
- }
+ pl_inode_lock_t *l = NULL;
+ pl_inode_lock_t *ret = NULL;
+
+ list_for_each_entry(l, &dom->inodelk_list, list)
+ {
+ if (inodelk_conflict(lock, l) && !same_inodelk_owner(lock, l)) {
+ if (ret == NULL) {
+ ret = l;
+ if (contend == NULL) {
+ break;
}
+ }
+ inodelk_contention_notify_check(this, l, now, contend);
}
+ }
- return ret;
+ return ret;
}
static pl_inode_lock_t *
-__blocked_lock_conflict (pl_dom_list_t *dom, pl_inode_lock_t *lock)
+__blocked_lock_conflict(pl_dom_list_t *dom, pl_inode_lock_t *lock)
{
- pl_inode_lock_t *l = NULL;
+ pl_inode_lock_t *l = NULL;
- list_for_each_entry (l, &dom->blocked_inodelks, blocked_locks) {
- if (inodelk_conflict (lock, l)) {
- return l;
- }
+ list_for_each_entry(l, &dom->blocked_inodelks, blocked_locks)
+ {
+ if (inodelk_conflict(lock, l)) {
+ return l;
}
+ }
- return NULL;
+ return NULL;
}
static int
-__owner_has_lock (pl_dom_list_t *dom, pl_inode_lock_t *newlock)
+__owner_has_lock(pl_dom_list_t *dom, pl_inode_lock_t *newlock)
{
- pl_inode_lock_t *lock = NULL;
+ pl_inode_lock_t *lock = NULL;
- list_for_each_entry (lock, &dom->inodelk_list, list) {
- if (same_inodelk_owner (lock, newlock))
- return 1;
- }
+ list_for_each_entry(lock, &dom->inodelk_list, list)
+ {
+ if (same_inodelk_owner(lock, newlock))
+ return 1;
+ }
- list_for_each_entry (lock, &dom->blocked_inodelks, blocked_locks) {
- if (same_inodelk_owner (lock, newlock))
- return 1;
- }
+ list_for_each_entry(lock, &dom->blocked_inodelks, blocked_locks)
+ {
+ if (same_inodelk_owner(lock, newlock))
+ return 1;
+ }
- return 0;
+ return 0;
}
static int
__lock_blocked_add(xlator_t *this, pl_dom_list_t *dom, pl_inode_lock_t *lock,
int can_block)
{
- struct timeval now;
-
- gettimeofday(&now, NULL);
-
- if (can_block == 0) {
- goto out;
- }
-
- lock->blkd_time = now;
- list_add_tail (&lock->blocked_locks, &dom->blocked_inodelks);
-
- gf_msg_trace (this->name, 0, "%s (pid=%d) (lk-owner=%s) %"PRId64" - "
- "%"PRId64" => Blocked",
- lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
- lock->client_pid, lkowner_utoa (&lock->owner),
- lock->user_flock.l_start, lock->user_flock.l_len);
-
+ if (can_block == 0) {
+ goto out;
+ }
+
+ lock->blkd_time = gf_time();
+ list_add_tail(&lock->blocked_locks, &dom->blocked_inodelks);
+
+ gf_msg_trace(this->name, 0,
+ "%s (pid=%d) (lk-owner=%s) %" PRId64
+ " - "
+ "%" PRId64 " => Blocked",
+ lock->fl_type == F_UNLCK ? "Unlock" : "Lock", lock->client_pid,
+ lkowner_utoa(&lock->owner), lock->user_flock.l_start,
+ lock->user_flock.l_len);
+
+ pl_trace_block(this, lock->frame, NULL, NULL, F_SETLKW, &lock->user_flock,
+ lock->volume);
out:
- return -EAGAIN;
+ return -EAGAIN;
}
/* Determines if lock can be granted and adds the lock. If the lock
* is blocking, adds it to the blocked_inodelks list of the domain.
*/
static int
-__lock_inodelk (xlator_t *this, pl_inode_t *pl_inode, pl_inode_lock_t *lock,
- int can_block, pl_dom_list_t *dom, struct timespec *now,
- struct list_head *contend)
+__lock_inodelk(xlator_t *this, pl_inode_t *pl_inode, pl_inode_lock_t *lock,
+ int can_block, pl_dom_list_t *dom, struct timespec *now,
+ struct list_head *contend)
{
- pl_inode_lock_t *conf = NULL;
- int ret = -EINVAL;
-
- conf = __inodelk_grantable (this, dom, lock, now, contend);
- if (conf) {
- ret = __lock_blocked_add(this, dom, lock, can_block);
- goto out;
- }
-
- /* To prevent blocked locks starvation, check if there are any blocked
- * locks thay may conflict with this lock. If there is then don't grant
- * the lock. BUT grant the lock if the owner already has lock to allow
- * nested locks.
- * Example:
- * SHD from Machine1 takes (gfid, 0-infinity) and is granted.
- * SHD from machine2 takes (gfid, 0-infinity) and is blocked.
- * When SHD from Machine1 takes (gfid, 0-128KB) it
- * needs to be granted, without which the earlier lock on 0-infinity
- * will not be unlocked by SHD from Machine1.
- * TODO: Find why 'owner_has_lock' is checked even for blocked locks.
- */
- if (__blocked_lock_conflict (dom, lock) &&
- !(__owner_has_lock (dom, lock))) {
- if (can_block != 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Lock is grantable, but blocking to prevent "
- "starvation");
- }
+ pl_inode_lock_t *conf = NULL;
+ int ret;
- ret = __lock_blocked_add(this, dom, lock, can_block);
- goto out;
+ ret = pl_inode_remove_inodelk(pl_inode, lock);
+ if (ret < 0) {
+ return ret;
+ }
+ if (ret == 0) {
+ conf = __inodelk_grantable(this, dom, lock, now, contend);
+ }
+ if ((ret > 0) || (conf != NULL)) {
+ return __lock_blocked_add(this, dom, lock, can_block);
+ }
+
+ /* To prevent blocked locks starvation, check if there are any blocked
+ * locks thay may conflict with this lock. If there is then don't grant
+ * the lock. BUT grant the lock if the owner already has lock to allow
+ * nested locks.
+ * Example:
+ * SHD from Machine1 takes (gfid, 0-infinity) and is granted.
+ * SHD from machine2 takes (gfid, 0-infinity) and is blocked.
+ * When SHD from Machine1 takes (gfid, 0-128KB) it
+ * needs to be granted, without which the earlier lock on 0-infinity
+ * will not be unlocked by SHD from Machine1.
+ * TODO: Find why 'owner_has_lock' is checked even for blocked locks.
+ */
+ if (__blocked_lock_conflict(dom, lock) && !(__owner_has_lock(dom, lock))) {
+ if (can_block != 0) {
+ gf_log(this->name, GF_LOG_DEBUG,
+ "Lock is grantable, but blocking to prevent "
+ "starvation");
}
- __pl_inodelk_ref (lock);
- gettimeofday (&lock->granted_time, NULL);
- list_add (&lock->list, &dom->inodelk_list);
- ret = 0;
+ return __lock_blocked_add(this, dom, lock, can_block);
+ }
+ __pl_inodelk_ref(lock);
+ lock->granted_time = gf_time();
+ list_add(&lock->list, &dom->inodelk_list);
-out:
- return ret;
+ return 0;
}
/* Return true if the two inodelks have exactly same lock boundaries */
static int
-inodelks_equal (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
+inodelks_equal(pl_inode_lock_t *l1, pl_inode_lock_t *l2)
{
- if ((l1->fl_start == l2->fl_start) &&
- (l1->fl_end == l2->fl_end))
- return 1;
+ if ((l1->fl_start == l2->fl_start) && (l1->fl_end == l2->fl_end))
+ return 1;
- return 0;
+ return 0;
}
-
static pl_inode_lock_t *
-find_matching_inodelk (pl_inode_lock_t *lock, pl_dom_list_t *dom)
+find_matching_inodelk(pl_inode_lock_t *lock, pl_dom_list_t *dom)
{
- pl_inode_lock_t *l = NULL;
- list_for_each_entry (l, &dom->inodelk_list, list) {
- if (inodelks_equal (l, lock) &&
- same_inodelk_owner (l, lock))
- return l;
- }
- return NULL;
+ pl_inode_lock_t *l = NULL;
+ list_for_each_entry(l, &dom->inodelk_list, list)
+ {
+ if (inodelks_equal(l, lock) && same_inodelk_owner(l, lock))
+ return l;
+ }
+ return NULL;
}
/* Set F_UNLCK removes a lock which has the exact same lock boundaries
* as the UNLCK lock specifies. If such a lock is not found, returns invalid
*/
static pl_inode_lock_t *
-__inode_unlock_lock (xlator_t *this, pl_inode_lock_t *lock, pl_dom_list_t *dom)
+__inode_unlock_lock(xlator_t *this, pl_inode_lock_t *lock, pl_dom_list_t *dom)
{
-
- pl_inode_lock_t *conf = NULL;
-
- conf = find_matching_inodelk (lock, dom);
- if (!conf) {
- gf_log (this->name, GF_LOG_ERROR,
- " Matching lock not found for unlock %llu-%llu, by %s "
- "on %p", (unsigned long long)lock->fl_start,
- (unsigned long long)lock->fl_end,
- lkowner_utoa (&lock->owner), lock->client);
- goto out;
- }
- __delete_inode_lock (conf);
- gf_log (this->name, GF_LOG_DEBUG,
- " Matching lock found for unlock %llu-%llu, by %s on %p",
- (unsigned long long)lock->fl_start,
- (unsigned long long)lock->fl_end, lkowner_utoa (&lock->owner),
- lock->client);
+ pl_inode_lock_t *conf = NULL;
+ inode_t *inode = NULL;
+
+ inode = lock->pl_inode->inode;
+
+ conf = find_matching_inodelk(lock, dom);
+ if (!conf) {
+ gf_log(this->name, GF_LOG_ERROR,
+ " Matching lock not found for unlock %llu-%llu, by %s "
+ "on %p for gfid:%s",
+ (unsigned long long)lock->fl_start,
+ (unsigned long long)lock->fl_end, lkowner_utoa(&lock->owner),
+ lock->client, inode ? uuid_utoa(inode->gfid) : "UNKNOWN");
+ goto out;
+ }
+ __delete_inode_lock(conf);
+ gf_log(this->name, GF_LOG_DEBUG,
+ " Matching lock found for unlock %llu-%llu, by %s on %p for gfid:%s",
+ (unsigned long long)lock->fl_start, (unsigned long long)lock->fl_end,
+ lkowner_utoa(&lock->owner), lock->client,
+ inode ? uuid_utoa(inode->gfid) : "UNKNOWN");
out:
- return conf;
+ return conf;
}
-
-static void
-__grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode,
- struct list_head *granted, pl_dom_list_t *dom,
- struct timespec *now, struct list_head *contend)
+void
+__grant_blocked_inode_locks(xlator_t *this, pl_inode_t *pl_inode,
+ struct list_head *granted, pl_dom_list_t *dom,
+ struct timespec *now, struct list_head *contend)
{
- int bl_ret = 0;
- pl_inode_lock_t *bl = NULL;
- pl_inode_lock_t *tmp = NULL;
-
- struct list_head blocked_list;
+ pl_inode_lock_t *bl = NULL;
+ pl_inode_lock_t *tmp = NULL;
- INIT_LIST_HEAD (&blocked_list);
- list_splice_init (&dom->blocked_inodelks, &blocked_list);
+ struct list_head blocked_list;
- list_for_each_entry_safe (bl, tmp, &blocked_list, blocked_locks) {
+ INIT_LIST_HEAD(&blocked_list);
+ list_splice_init(&dom->blocked_inodelks, &blocked_list);
- list_del_init (&bl->blocked_locks);
+ list_for_each_entry_safe(bl, tmp, &blocked_list, blocked_locks)
+ {
+ list_del_init(&bl->blocked_locks);
- bl_ret = __lock_inodelk (this, pl_inode, bl, 1, dom, now,
- contend);
+ bl->status = __lock_inodelk(this, pl_inode, bl, 1, dom, now, contend);
- if (bl_ret == 0) {
- list_add (&bl->blocked_locks, granted);
- }
+ if (bl->status != -EAGAIN) {
+ list_add_tail(&bl->blocked_locks, granted);
}
- return;
+ }
}
-/* Grant all inodelks blocked on a lock */
void
-grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode,
- pl_dom_list_t *dom, struct timespec *now,
- struct list_head *contend)
+unwind_granted_inodes(xlator_t *this, pl_inode_t *pl_inode,
+ struct list_head *granted)
{
- struct list_head granted;
- pl_inode_lock_t *lock;
- pl_inode_lock_t *tmp;
+ pl_inode_lock_t *lock;
+ pl_inode_lock_t *tmp;
+ int32_t op_ret;
+ int32_t op_errno;
+
+ list_for_each_entry_safe(lock, tmp, granted, blocked_locks)
+ {
+ if (lock->status == 0) {
+ op_ret = 0;
+ op_errno = 0;
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s (pid=%d) (lk-owner=%s) %" PRId64 " - %" PRId64
+ " => Granted",
+ lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+ lock->client_pid, lkowner_utoa(&lock->owner),
+ lock->user_flock.l_start, lock->user_flock.l_len);
+ } else {
+ op_ret = -1;
+ op_errno = -lock->status;
+ }
+ pl_trace_out(this, lock->frame, NULL, NULL, F_SETLKW, &lock->user_flock,
+ op_ret, op_errno, lock->volume);
- INIT_LIST_HEAD (&granted);
+ STACK_UNWIND_STRICT(inodelk, lock->frame, op_ret, op_errno, NULL);
+ lock->frame = NULL;
+ }
- pthread_mutex_lock (&pl_inode->mutex);
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ list_for_each_entry_safe(lock, tmp, granted, blocked_locks)
{
- __grant_blocked_inode_locks (this, pl_inode, &granted, dom,
- now, contend);
+ list_del_init(&lock->blocked_locks);
+ __pl_inodelk_unref(lock);
}
- pthread_mutex_unlock (&pl_inode->mutex);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+}
- list_for_each_entry_safe (lock, tmp, &granted, blocked_locks) {
- gf_log (this->name, GF_LOG_TRACE,
- "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => Granted",
- lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
- lock->client_pid,
- lkowner_utoa (&lock->owner),
- lock->user_flock.l_start,
- lock->user_flock.l_len);
+/* Grant all inodelks blocked on a lock */
+void
+grant_blocked_inode_locks(xlator_t *this, pl_inode_t *pl_inode,
+ pl_dom_list_t *dom, struct timespec *now,
+ struct list_head *contend)
+{
+ struct list_head granted;
- pl_trace_out (this, lock->frame, NULL, NULL, F_SETLKW,
- &lock->user_flock, 0, 0, lock->volume);
+ INIT_LIST_HEAD(&granted);
- STACK_UNWIND_STRICT (inodelk, lock->frame, 0, 0, NULL);
- lock->frame = NULL;
- }
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ __grant_blocked_inode_locks(this, pl_inode, &granted, dom, now,
+ contend);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
- pthread_mutex_lock (&pl_inode->mutex);
- {
- list_for_each_entry_safe (lock, tmp, &granted, blocked_locks) {
- list_del_init (&lock->blocked_locks);
- __pl_inodelk_unref (lock);
- }
- }
- pthread_mutex_unlock (&pl_inode->mutex);
+ unwind_granted_inodes(this, pl_inode, &granted);
}
-
static void
-pl_inodelk_log_cleanup (pl_inode_lock_t *lock)
+pl_inodelk_log_cleanup(pl_inode_lock_t *lock)
{
- pl_inode_t *pl_inode = NULL;
+ pl_inode_t *pl_inode = NULL;
- pl_inode = lock->pl_inode;
+ pl_inode = lock->pl_inode;
- gf_log (THIS->name, GF_LOG_WARNING, "releasing lock on %s held by "
- "{client=%p, pid=%"PRId64" lk-owner=%s}",
- uuid_utoa (pl_inode->gfid), lock->client,
- (uint64_t) lock->client_pid, lkowner_utoa (&lock->owner));
+ gf_log(THIS->name, GF_LOG_WARNING,
+ "releasing lock on %s held by "
+ "{client=%p, pid=%" PRId64 " lk-owner=%s}",
+ uuid_utoa(pl_inode->gfid), lock->client, (uint64_t)lock->client_pid,
+ lkowner_utoa(&lock->owner));
}
-
/* Release all inodelks from this client */
int
-pl_inodelk_client_cleanup (xlator_t *this, pl_ctx_t *ctx)
+pl_inodelk_client_cleanup(xlator_t *this, pl_ctx_t *ctx)
{
- posix_locks_private_t *priv;
- pl_inode_lock_t *tmp = NULL;
- pl_inode_lock_t *l = NULL;
- pl_dom_list_t *dom = NULL;
- pl_inode_t *pl_inode = NULL;
- struct list_head *pcontend = NULL;
- struct list_head released;
- struct list_head unwind;
- struct list_head contend;
- struct timespec now = { };
-
- priv = this->private;
-
- INIT_LIST_HEAD (&released);
- INIT_LIST_HEAD (&unwind);
-
- if (priv->notify_contention) {
- pcontend = &contend;
- INIT_LIST_HEAD (pcontend);
- timespec_now(&now);
- }
-
- pthread_mutex_lock (&ctx->lock);
+ posix_locks_private_t *priv;
+ pl_inode_lock_t *tmp = NULL;
+ pl_inode_lock_t *l = NULL;
+ pl_dom_list_t *dom = NULL;
+ pl_inode_t *pl_inode = NULL;
+ struct list_head *pcontend = NULL;
+ struct list_head released;
+ struct list_head unwind;
+ struct list_head contend;
+ struct timespec now = {};
+
+ priv = this->private;
+
+ INIT_LIST_HEAD(&released);
+ INIT_LIST_HEAD(&unwind);
+
+ if (priv->notify_contention) {
+ pcontend = &contend;
+ INIT_LIST_HEAD(pcontend);
+ timespec_now(&now);
+ }
+
+ pthread_mutex_lock(&ctx->lock);
+ {
+ list_for_each_entry_safe(l, tmp, &ctx->inodelk_lockers, client_list)
{
- list_for_each_entry_safe (l, tmp, &ctx->inodelk_lockers,
- client_list) {
- pl_inodelk_log_cleanup (l);
-
- pl_inode = l->pl_inode;
-
- pthread_mutex_lock (&pl_inode->mutex);
- {
- /* If the inodelk object is part of granted list but not
- * blocked list, then perform the following actions:
- * i. delete the object from granted list;
- * ii. grant other locks (from other clients) that may
- * have been blocked on this inodelk; and
- * iii. unref the object.
- *
- * If the inodelk object (L1) is part of both granted
- * and blocked lists, then this means that a parallel
- * unlock on another inodelk (L2 say) may have 'granted'
- * L1 and added it to 'granted' list in
- * __grant_blocked_node_locks() (although using the
- * 'blocked_locks' member). In that case, the cleanup
- * codepath must try and grant other overlapping
- * blocked inodelks from other clients, now that L1 is
- * out of their way and then unref L1 in the end, and
- * leave it to the other thread (the one executing
- * unlock codepath) to unwind L1's frame, delete it from
- * blocked_locks list, and perform the last unref on L1.
- *
- * If the inodelk object (L1) is part of blocked list
- * only, the cleanup code path must:
- * i. delete it from the blocked_locks list inside
- * this critical section,
- * ii. unwind its frame with EAGAIN,
- * iii. try and grant blocked inode locks from other
- * clients that were otherwise grantable, but just
- * got blocked to avoid leaving L1 to starve
- * forever.
- * iv. unref the object.
- */
- list_del_init (&l->client_list);
-
- if (!list_empty (&l->list)) {
- __delete_inode_lock (l);
- list_add_tail (&l->client_list,
- &released);
- } else {
- list_del_init(&l->blocked_locks);
- list_add_tail (&l->client_list,
- &unwind);
- }
- }
- pthread_mutex_unlock (&pl_inode->mutex);
+ pl_inodelk_log_cleanup(l);
+
+ pl_inode = l->pl_inode;
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ /* If the inodelk object is part of granted list but not
+ * blocked list, then perform the following actions:
+ * i. delete the object from granted list;
+ * ii. grant other locks (from other clients) that may
+ * have been blocked on this inodelk; and
+ * iii. unref the object.
+ *
+ * If the inodelk object (L1) is part of both granted
+ * and blocked lists, then this means that a parallel
+ * unlock on another inodelk (L2 say) may have 'granted'
+ * L1 and added it to 'granted' list in
+ * __grant_blocked_inode_locks() (although using the
+ * 'blocked_locks' member). In that case, the cleanup
+ * codepath must try and grant other overlapping
+ * blocked inodelks from other clients, now that L1 is
+ * out of their way and then unref L1 in the end, and
+ * leave it to the other thread (the one executing
+ * unlock codepath) to unwind L1's frame, delete it from
+ * blocked_locks list, and perform the last unref on L1.
+ *
+ * If the inodelk object (L1) is part of blocked list
+ * only, the cleanup code path must:
+ * i. delete it from the blocked_locks list inside
+ * this critical section,
+ * ii. unwind its frame with EAGAIN,
+ * iii. try and grant blocked inode locks from other
+ * clients that were otherwise grantable, but just
+ * got blocked to avoid leaving L1 to starve
+ * forever.
+ * iv. unref the object.
+ */
+ list_del_init(&l->client_list);
+
+ if (!list_empty(&l->list)) {
+ __delete_inode_lock(l);
+ list_add_tail(&l->client_list, &released);
+ } else {
+ list_del_init(&l->blocked_locks);
+ list_add_tail(&l->client_list, &unwind);
}
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
}
- pthread_mutex_unlock (&ctx->lock);
+ }
+ pthread_mutex_unlock(&ctx->lock);
- list_for_each_entry_safe (l, tmp, &unwind, client_list) {
- list_del_init (&l->client_list);
+ if (!list_empty(&unwind)) {
+ list_for_each_entry_safe(l, tmp, &unwind, client_list)
+ {
+ list_del_init(&l->client_list);
- if (l->frame)
- STACK_UNWIND_STRICT (inodelk, l->frame, -1, EAGAIN,
- NULL);
- list_add_tail (&l->client_list, &released);
+ if (l->frame)
+ STACK_UNWIND_STRICT(inodelk, l->frame, -1, EAGAIN, NULL);
+ list_add_tail(&l->client_list, &released);
}
+ }
- list_for_each_entry_safe (l, tmp, &released, client_list) {
- list_del_init (&l->client_list);
+ if (!list_empty(&released)) {
+ list_for_each_entry_safe(l, tmp, &released, client_list)
+ {
+ list_del_init(&l->client_list);
- pl_inode = l->pl_inode;
+ pl_inode = l->pl_inode;
- dom = get_domain (pl_inode, l->volume);
+ dom = get_domain(pl_inode, l->volume);
- grant_blocked_inode_locks (this, pl_inode, dom, &now,
- pcontend);
+ grant_blocked_inode_locks(this, pl_inode, dom, &now, pcontend);
- pthread_mutex_lock (&pl_inode->mutex);
- {
- __pl_inodelk_unref (l);
- }
- pthread_mutex_unlock (&pl_inode->mutex);
- inode_unref (pl_inode->inode);
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ __pl_inodelk_unref(l);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+ inode_unref(pl_inode->inode);
}
+ }
- if (pcontend != NULL) {
- inodelk_contention_notify(this, pcontend);
- }
+ if (pcontend != NULL) {
+ inodelk_contention_notify(this, pcontend);
+ }
- return 0;
+ return 0;
}
-
static int
-pl_inode_setlk (xlator_t *this, pl_ctx_t *ctx, pl_inode_t *pl_inode,
- pl_inode_lock_t *lock, int can_block, pl_dom_list_t *dom,
- inode_t *inode)
+pl_inode_setlk(xlator_t *this, pl_ctx_t *ctx, pl_inode_t *pl_inode,
+ pl_inode_lock_t *lock, int can_block, pl_dom_list_t *dom,
+ inode_t *inode)
{
- posix_locks_private_t *priv = NULL;
- int ret = -EINVAL;
- pl_inode_lock_t *retlock = NULL;
- gf_boolean_t unref = _gf_true;
- gf_boolean_t need_inode_unref = _gf_false;
- struct list_head *pcontend = NULL;
- struct list_head contend;
- struct timespec now = { };
- short fl_type;
-
- lock->pl_inode = pl_inode;
- fl_type = lock->fl_type;
-
- priv = this->private;
-
- /* Ideally, AFTER a successful lock (both blocking and non-blocking) or
- * an unsuccessful blocking lock operation, the inode needs to be ref'd.
- *
- * But doing so might give room to a race where the lock-requesting
- * client could send a DISCONNECT just before this thread refs the inode
- * after the locking is done, and the epoll thread could unref the inode
- * in cleanup which means the inode's refcount would come down to 0, and
- * the call to pl_forget() at this point destroys @pl_inode. Now when
- * the io-thread executing this function tries to access pl_inode,
- * it could crash on account of illegal memory access.
- *
- * To get around this problem, the inode is ref'd once even before
- * adding the lock into client_list as a precautionary measure.
- * This way even if there are DISCONNECTs, there will always be 1 extra
- * ref on the inode, so @pl_inode is still alive until after the
- * current stack unwinds.
- */
- pl_inode->inode = inode_ref (inode);
-
- if (priv->revocation_secs != 0) {
- if (lock->fl_type != F_UNLCK) {
- __inodelk_prune_stale (this, pl_inode, dom, lock);
- } else if (priv->monkey_unlocking == _gf_true) {
- if (pl_does_monkey_want_stuck_lock ()) {
- pthread_mutex_lock (&pl_inode->mutex);
- {
- __pl_inodelk_unref (lock);
- }
- pthread_mutex_unlock (&pl_inode->mutex);
- inode_unref (pl_inode->inode);
- gf_log (this->name, GF_LOG_WARNING,
- "MONKEY LOCKING (forcing stuck lock)!");
- return 0;
- }
+ posix_locks_private_t *priv = NULL;
+ int ret = -EINVAL;
+ pl_inode_lock_t *retlock = NULL;
+ gf_boolean_t unref = _gf_true;
+ gf_boolean_t need_inode_unref = _gf_false;
+ struct list_head *pcontend = NULL;
+ struct list_head contend;
+ struct list_head wake;
+ struct timespec now = {};
+ short fl_type;
+
+ lock->pl_inode = pl_inode;
+ fl_type = lock->fl_type;
+
+ priv = this->private;
+
+ /* Ideally, AFTER a successful lock (both blocking and non-blocking) or
+ * an unsuccessful blocking lock operation, the inode needs to be ref'd.
+ *
+ * But doing so might give room to a race where the lock-requesting
+ * client could send a DISCONNECT just before this thread refs the inode
+ * after the locking is done, and the epoll thread could unref the inode
+ * in cleanup which means the inode's refcount would come down to 0, and
+ * the call to pl_forget() at this point destroys @pl_inode. Now when
+ * the io-thread executing this function tries to access pl_inode,
+ * it could crash on account of illegal memory access.
+ *
+ * To get around this problem, the inode is ref'd once even before
+ * adding the lock into client_list as a precautionary measure.
+ * This way even if there are DISCONNECTs, there will always be 1 extra
+ * ref on the inode, so @pl_inode is still alive until after the
+ * current stack unwinds.
+ */
+ pl_inode->inode = inode_ref(inode);
+
+ if (priv->revocation_secs != 0) {
+ if (lock->fl_type != F_UNLCK) {
+ __inodelk_prune_stale(this, pl_inode, dom, lock);
+ } else if (priv->monkey_unlocking == _gf_true) {
+ if (pl_does_monkey_want_stuck_lock()) {
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ __pl_inodelk_unref(lock);
}
+ pthread_mutex_unlock(&pl_inode->mutex);
+ inode_unref(pl_inode->inode);
+ gf_log(this->name, GF_LOG_WARNING,
+ "MONKEY LOCKING (forcing stuck lock)!");
+ return 0;
+ }
}
-
- if (priv->notify_contention) {
- pcontend = &contend;
- INIT_LIST_HEAD(pcontend);
- timespec_now(&now);
- }
-
- if (ctx)
- pthread_mutex_lock (&ctx->lock);
- pthread_mutex_lock (&pl_inode->mutex);
- {
- if (lock->fl_type != F_UNLCK) {
- ret = __lock_inodelk (this, pl_inode, lock, can_block,
- dom, &now, pcontend);
- if (ret == 0) {
- lock->frame = NULL;
- gf_log (this->name, GF_LOG_TRACE,
- "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => OK",
- lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
- lock->client_pid,
- lkowner_utoa (&lock->owner),
- lock->fl_start,
- lock->fl_end);
- } else if (ret == -EAGAIN) {
- gf_log (this->name, GF_LOG_TRACE,
- "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => NOK",
- lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
- lock->client_pid,
- lkowner_utoa (&lock->owner),
- lock->user_flock.l_start,
- lock->user_flock.l_len);
- if (can_block)
- unref = _gf_false;
- /* For all but the case where a non-blocking
- * lock attempt fails, the extra ref taken at
- * the start of this function must be negated.
- */
- else
- need_inode_unref = _gf_true;
- }
-
- if (ctx && (!ret || can_block))
- list_add_tail (&lock->client_list,
- &ctx->inodelk_lockers);
- } else {
- /* Irrespective of whether unlock succeeds or not,
- * the extra inode ref that was done at the start of
- * this function must be negated. Towards this,
- * @need_inode_unref flag is set unconditionally here.
- */
- need_inode_unref = _gf_true;
- retlock = __inode_unlock_lock (this, lock, dom);
- if (!retlock) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Bad Unlock issued on Inode lock");
- ret = -EINVAL;
- goto out;
- }
- list_del_init (&retlock->client_list);
- __pl_inodelk_unref (retlock);
-
- ret = 0;
+ }
+
+ if (priv->notify_contention) {
+ pcontend = &contend;
+ INIT_LIST_HEAD(pcontend);
+ timespec_now(&now);
+ }
+
+ INIT_LIST_HEAD(&wake);
+
+ if (ctx)
+ pthread_mutex_lock(&ctx->lock);
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ if (lock->fl_type != F_UNLCK) {
+ ret = __lock_inodelk(this, pl_inode, lock, can_block, dom, &now,
+ pcontend);
+ if (ret == 0) {
+ lock->frame = NULL;
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s (pid=%d) (lk-owner=%s) %" PRId64 " - %" PRId64
+ " => OK",
+ lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+ lock->client_pid, lkowner_utoa(&lock->owner),
+ lock->fl_start, lock->fl_end);
+ } else if (ret == -EAGAIN) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s (pid=%d) (lk-owner=%s) %" PRId64 " - %" PRId64
+ " => NOK",
+ lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+ lock->client_pid, lkowner_utoa(&lock->owner),
+ lock->user_flock.l_start, lock->user_flock.l_len);
+ if (can_block) {
+ unref = _gf_false;
}
-out:
- if (unref)
- __pl_inodelk_unref (lock);
- }
- pthread_mutex_unlock (&pl_inode->mutex);
- if (ctx)
- pthread_mutex_unlock (&ctx->lock);
-
- /* The following (extra) unref corresponds to the ref that
- * was done at the time the lock was granted.
- */
- if ((fl_type == F_UNLCK) && (ret == 0)) {
- inode_unref (pl_inode->inode);
- grant_blocked_inode_locks (this, pl_inode, dom, &now,
- pcontend);
- }
+ }
+ /* For all but the case where a non-blocking lock attempt fails
+ * with -EAGAIN, the extra ref taken at the start of this function
+ * must be negated. */
+ need_inode_unref = (ret != 0) && ((ret != -EAGAIN) || !can_block);
+ if (ctx && !need_inode_unref) {
+ list_add_tail(&lock->client_list, &ctx->inodelk_lockers);
+ }
+ } else {
+ /* Irrespective of whether unlock succeeds or not,
+ * the extra inode ref that was done at the start of
+ * this function must be negated. Towards this,
+ * @need_inode_unref flag is set unconditionally here.
+ */
+ need_inode_unref = _gf_true;
+ retlock = __inode_unlock_lock(this, lock, dom);
+ if (!retlock) {
+ gf_log(this->name, GF_LOG_DEBUG,
+ "Bad Unlock issued on Inode lock");
+ ret = -EINVAL;
+ goto out;
+ }
+ list_del_init(&retlock->client_list);
+ __pl_inodelk_unref(retlock);
- if (need_inode_unref) {
- inode_unref (pl_inode->inode);
- }
+ pl_inode_remove_unlocked(this, pl_inode, &wake);
- if (pcontend != NULL) {
- inodelk_contention_notify(this, pcontend);
+ ret = 0;
}
-
- return ret;
+ out:
+ if (unref)
+ __pl_inodelk_unref(lock);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+ if (ctx)
+ pthread_mutex_unlock(&ctx->lock);
+
+ pl_inode_remove_wake(&wake);
+
+ /* The following (extra) unref corresponds to the ref that
+ * was done at the time the lock was granted.
+ */
+ if ((fl_type == F_UNLCK) && (ret == 0)) {
+ inode_unref(pl_inode->inode);
+ grant_blocked_inode_locks(this, pl_inode, dom, &now, pcontend);
+ }
+
+ if (need_inode_unref) {
+ inode_unref(pl_inode->inode);
+ }
+
+ if (pcontend != NULL) {
+ inodelk_contention_notify(this, pcontend);
+ }
+
+ return ret;
}
/* Create a new inode_lock_t */
-pl_inode_lock_t *
-new_inode_lock (struct gf_flock *flock, client_t *client, pid_t client_pid,
- call_frame_t *frame, xlator_t *this, const char *volume,
- char *conn_id)
+static pl_inode_lock_t *
+new_inode_lock(struct gf_flock *flock, client_t *client, pid_t client_pid,
+ call_frame_t *frame, xlator_t *this, const char *volume,
+ char *conn_id, int32_t *op_errno)
{
- pl_inode_lock_t *lock = NULL;
-
- lock = GF_CALLOC (1, sizeof (*lock),
- gf_locks_mt_pl_inode_lock_t);
- if (!lock) {
- return NULL;
- }
-
- lock->fl_start = flock->l_start;
- lock->fl_type = flock->l_type;
+ pl_inode_lock_t *lock = NULL;
+
+ if (!pl_is_lk_owner_valid(&frame->root->lk_owner, frame->root->client)) {
+ *op_errno = EINVAL;
+ goto out;
+ }
+
+ lock = GF_CALLOC(1, sizeof(*lock), gf_locks_mt_pl_inode_lock_t);
+ if (!lock) {
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ lock->fl_start = flock->l_start;
+ lock->fl_type = flock->l_type;
+
+ if (flock->l_len == 0)
+ lock->fl_end = LLONG_MAX;
+ else
+ lock->fl_end = flock->l_start + flock->l_len - 1;
+
+ lock->client = client;
+ lock->client_pid = client_pid;
+ lock->volume = volume;
+ lock->owner = frame->root->lk_owner;
+ lock->frame = frame;
+ lock->this = this;
+
+ if (conn_id) {
+ lock->connection_id = gf_strdup(conn_id);
+ }
+
+ INIT_LIST_HEAD(&lock->list);
+ INIT_LIST_HEAD(&lock->blocked_locks);
+ INIT_LIST_HEAD(&lock->client_list);
+ INIT_LIST_HEAD(&lock->contend);
+ __pl_inodelk_ref(lock);
- if (flock->l_len == 0)
- lock->fl_end = LLONG_MAX;
- else
- lock->fl_end = flock->l_start + flock->l_len - 1;
-
- lock->client = client;
- lock->client_pid = client_pid;
- lock->volume = volume;
- lock->owner = frame->root->lk_owner;
- lock->frame = frame;
- lock->this = this;
-
- if (conn_id) {
- lock->connection_id = gf_strdup (conn_id);
- }
-
- INIT_LIST_HEAD (&lock->list);
- INIT_LIST_HEAD (&lock->blocked_locks);
- INIT_LIST_HEAD (&lock->client_list);
- INIT_LIST_HEAD (&lock->contend);
- __pl_inodelk_ref (lock);
-
- return lock;
+out:
+ return lock;
}
int32_t
-_pl_convert_volume (const char *volume, char **res)
+_pl_convert_volume(const char *volume, char **res)
{
- char *mdata_vol = NULL;
- int ret = 0;
+ char *mdata_vol = NULL;
+ int ret = 0;
- mdata_vol = strrchr (volume, ':');
- //if the volume already ends with :metadata don't bother
- if (mdata_vol && (strcmp (mdata_vol, ":metadata") == 0))
- return 0;
-
- ret = gf_asprintf (res, "%s:metadata", volume);
- if (ret <= 0)
- return ENOMEM;
+ mdata_vol = strrchr(volume, ':');
+ // if the volume already ends with :metadata don't bother
+ if (mdata_vol && (strcmp(mdata_vol, ":metadata") == 0))
return 0;
+
+ ret = gf_asprintf(res, "%s:metadata", volume);
+ if (ret <= 0)
+ return ENOMEM;
+ return 0;
}
int32_t
-_pl_convert_volume_for_special_range (struct gf_flock *flock,
- const char *volume, char **res)
+_pl_convert_volume_for_special_range(struct gf_flock *flock, const char *volume,
+ char **res)
{
- int32_t ret = 0;
+ int32_t ret = 0;
- if ((flock->l_start == LLONG_MAX -1) &&
- (flock->l_len == 0)) {
- ret = _pl_convert_volume (volume, res);
- }
+ if ((flock->l_start == LLONG_MAX - 1) && (flock->l_len == 0)) {
+ ret = _pl_convert_volume(volume, res);
+ }
- return ret;
+ return ret;
}
/* Common inodelk code called from pl_inodelk and pl_finodelk */
int
-pl_common_inodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, inode_t *inode, int32_t cmd,
- struct gf_flock *flock, loc_t *loc, fd_t *fd, dict_t *xdata)
+pl_common_inodelk(call_frame_t *frame, xlator_t *this, const char *volume,
+ inode_t *inode, int32_t cmd, struct gf_flock *flock,
+ loc_t *loc, fd_t *fd, dict_t *xdata)
{
- int32_t op_ret = -1;
- int32_t op_errno = 0;
- int ret = -1;
- GF_UNUSED int dict_ret = -1;
- int can_block = 0;
- pl_inode_t * pinode = NULL;
- pl_inode_lock_t * reqlock = NULL;
- pl_dom_list_t * dom = NULL;
- char *res = NULL;
- char *res1 = NULL;
- char *conn_id = NULL;
- pl_ctx_t *ctx = NULL;
-
- if (xdata)
- dict_ret = dict_get_str (xdata, "connection-id", &conn_id);
-
- VALIDATE_OR_GOTO (frame, out);
- VALIDATE_OR_GOTO (inode, unwind);
- VALIDATE_OR_GOTO (flock, unwind);
-
- if ((flock->l_start < 0) || (flock->l_len < 0)) {
- op_errno = EINVAL;
- goto unwind;
- }
-
- op_errno = _pl_convert_volume_for_special_range (flock, volume, &res);
- if (op_errno)
- goto unwind;
- if (res)
- volume = res;
-
- pl_trace_in (this, frame, fd, loc, cmd, flock, volume);
-
- if (frame->root->client) {
- ctx = pl_ctx_get (frame->root->client, this);
- if (!ctx) {
- op_errno = ENOMEM;
- gf_log (this->name, GF_LOG_INFO, "pl_ctx_get() failed");
- goto unwind;
- }
- }
-
- pinode = pl_inode_get (this, inode);
- if (!pinode) {
- op_errno = ENOMEM;
- goto unwind;
+ int32_t op_ret = -1;
+ int32_t op_errno = 0;
+ int ret = -1;
+ GF_UNUSED int dict_ret = -1;
+ int can_block = 0;
+ short lock_type = 0;
+ pl_inode_t *pinode = NULL;
+ pl_inode_lock_t *reqlock = NULL;
+ pl_dom_list_t *dom = NULL;
+ char *res = NULL;
+ char *res1 = NULL;
+ char *conn_id = NULL;
+ pl_ctx_t *ctx = NULL;
+
+ if (xdata)
+ dict_ret = dict_get_str(xdata, "connection-id", &conn_id);
+
+ VALIDATE_OR_GOTO(frame, out);
+ VALIDATE_OR_GOTO(inode, unwind);
+ VALIDATE_OR_GOTO(flock, unwind);
+
+ if ((flock->l_start < 0) || (flock->l_len < 0)) {
+ op_errno = EINVAL;
+ goto unwind;
+ }
+
+ op_errno = _pl_convert_volume_for_special_range(flock, volume, &res);
+ if (op_errno)
+ goto unwind;
+ if (res)
+ volume = res;
+
+ pl_trace_in(this, frame, fd, loc, cmd, flock, volume);
+
+ if (frame->root->client) {
+ ctx = pl_ctx_get(frame->root->client, this);
+ if (!ctx) {
+ op_errno = ENOMEM;
+ gf_log(this->name, GF_LOG_INFO, "pl_ctx_get() failed");
+ goto unwind;
}
+ }
- dom = get_domain (pinode, volume);
- if (!dom) {
- op_errno = ENOMEM;
- goto unwind;
- }
+ pinode = pl_inode_get(this, inode, NULL);
+ if (!pinode) {
+ op_errno = ENOMEM;
+ goto unwind;
+ }
- reqlock = new_inode_lock (flock, frame->root->client, frame->root->pid,
- frame, this, dom->domain, conn_id);
+ dom = get_domain(pinode, volume);
+ if (!dom) {
+ op_errno = ENOMEM;
+ goto unwind;
+ }
- if (!reqlock) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
+ reqlock = new_inode_lock(flock, frame->root->client, frame->root->pid,
+ frame, this, dom->domain, conn_id, &op_errno);
+ if (!reqlock) {
+ op_ret = -1;
+ goto unwind;
+ }
- switch (cmd) {
+ switch (cmd) {
case F_SETLKW:
- can_block = 1;
+ can_block = 1;
- /* fall through */
+ /* fall through */
case F_SETLK:
- memcpy (&reqlock->user_flock, flock, sizeof (struct gf_flock));
- ret = pl_inode_setlk (this, ctx, pinode, reqlock, can_block,
- dom, inode);
-
- if (ret < 0) {
- if ((can_block) && (F_UNLCK != flock->l_type)) {
- pl_trace_block (this, frame, fd, loc,
- cmd, flock, volume);
- goto out;
- }
- gf_log (this->name, GF_LOG_TRACE, "returning EAGAIN");
- op_errno = -ret;
- goto unwind;
+ lock_type = flock->l_type;
+ memcpy(&reqlock->user_flock, flock, sizeof(struct gf_flock));
+ ret = pl_inode_setlk(this, ctx, pinode, reqlock, can_block, dom,
+ inode);
+
+ if (ret < 0) {
+ if (ret == -EAGAIN) {
+ if (can_block && (F_UNLCK != lock_type)) {
+ goto out;
+ }
+ gf_log(this->name, GF_LOG_TRACE, "returning EAGAIN");
+ } else {
+ gf_log(this->name, GF_LOG_TRACE, "returning %d", ret);
}
- break;
+ op_errno = -ret;
+ goto unwind;
+ }
+ break;
default:
- op_errno = ENOTSUP;
- gf_log (this->name, GF_LOG_DEBUG,
- "Lock command F_GETLK not supported for [f]inodelk "
- "(cmd=%d)",
- cmd);
- goto unwind;
- }
+ op_errno = ENOTSUP;
+ gf_log(this->name, GF_LOG_DEBUG,
+ "Lock command F_GETLK not supported for [f]inodelk "
+ "(cmd=%d)",
+ cmd);
+ goto unwind;
+ }
- op_ret = 0;
+ op_ret = 0;
unwind:
- if (flock != NULL)
- pl_trace_out (this, frame, fd, loc, cmd, flock, op_ret,
- op_errno, volume);
+ if (flock != NULL)
+ pl_trace_out(this, frame, fd, loc, cmd, flock, op_ret, op_errno,
+ volume);
- STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno, NULL);
+ STACK_UNWIND_STRICT(inodelk, frame, op_ret, op_errno, NULL);
out:
- GF_FREE (res);
- GF_FREE (res1);
- return 0;
+ GF_FREE(res);
+ GF_FREE(res1);
+ return 0;
}
int
-pl_inodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *flock,
- dict_t *xdata)
+pl_inodelk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ int32_t cmd, struct gf_flock *flock, dict_t *xdata)
{
- pl_common_inodelk (frame, this, volume, loc->inode, cmd, flock,
- loc, NULL, xdata);
+ pl_common_inodelk(frame, this, volume, loc->inode, cmd, flock, loc, NULL,
+ xdata);
- return 0;
+ return 0;
}
int
-pl_finodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *flock,
- dict_t *xdata)
+pl_finodelk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ int32_t cmd, struct gf_flock *flock, dict_t *xdata)
{
- pl_common_inodelk (frame, this, volume, fd->inode, cmd, flock,
- NULL, fd, xdata);
-
- return 0;
+ pl_common_inodelk(frame, this, volume, fd->inode, cmd, flock, NULL, fd,
+ xdata);
+ return 0;
}
static int32_t
-__get_inodelk_dom_count (pl_dom_list_t *dom)
+__get_inodelk_dom_count(pl_dom_list_t *dom)
{
- pl_inode_lock_t *lock = NULL;
- int32_t count = 0;
-
- list_for_each_entry (lock, &dom->inodelk_list, list) {
- count++;
- }
- list_for_each_entry (lock, &dom->blocked_inodelks, blocked_locks) {
- count++;
- }
- return count;
+ pl_inode_lock_t *lock = NULL;
+ int32_t count = 0;
+
+ list_for_each_entry(lock, &dom->inodelk_list, list) { count++; }
+ list_for_each_entry(lock, &dom->blocked_inodelks, blocked_locks)
+ {
+ count++;
+ }
+ return count;
}
/* Returns the no. of locks (blocked/granted) held on a given domain name
* If @domname is NULL, returns the no. of locks in all the domains present.
* If @domname is non-NULL and non-existent, returns 0 */
int32_t
-__get_inodelk_count (xlator_t *this, pl_inode_t *pl_inode, char *domname)
+__get_inodelk_count(xlator_t *this, pl_inode_t *pl_inode, char *domname)
{
- int32_t count = 0;
- pl_dom_list_t *dom = NULL;
-
- list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
- if (domname) {
- if (strcmp (domname, dom->domain) == 0) {
- count = __get_inodelk_dom_count (dom);
- goto out;
- }
-
- } else {
- /* Counting locks from all domains */
- count += __get_inodelk_dom_count (dom);
+ int32_t count = 0;
+ pl_dom_list_t *dom = NULL;
+
+ list_for_each_entry(dom, &pl_inode->dom_list, inode_list)
+ {
+ if (domname) {
+ if (strcmp(domname, dom->domain) == 0) {
+ count = __get_inodelk_dom_count(dom);
+ goto out;
+ }
- }
+ } else {
+ /* Counting locks from all domains */
+ count += __get_inodelk_dom_count(dom);
}
+ }
out:
- return count;
+ return count;
}
int32_t
-get_inodelk_count (xlator_t *this, inode_t *inode, char *domname)
+get_inodelk_count(xlator_t *this, inode_t *inode, char *domname)
{
- pl_inode_t *pl_inode = NULL;
- uint64_t tmp_pl_inode = 0;
- int ret = 0;
- int32_t count = 0;
+ pl_inode_t *pl_inode = NULL;
+ uint64_t tmp_pl_inode = 0;
+ int ret = 0;
+ int32_t count = 0;
- ret = inode_ctx_get (inode, this, &tmp_pl_inode);
- if (ret != 0) {
- goto out;
- }
+ ret = inode_ctx_get(inode, this, &tmp_pl_inode);
+ if (ret != 0) {
+ goto out;
+ }
- pl_inode = (pl_inode_t *)(long) tmp_pl_inode;
+ pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
- pthread_mutex_lock (&pl_inode->mutex);
- {
- count = __get_inodelk_count (this, pl_inode, domname);
- }
- pthread_mutex_unlock (&pl_inode->mutex);
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ count = __get_inodelk_count(this, pl_inode, domname);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
out:
- return count;
+ return count;
}
diff --git a/xlators/features/locks/src/locks-mem-types.h b/xlators/features/locks/src/locks-mem-types.h
index a48b35c2044..a76605027b3 100644
--- a/xlators/features/locks/src/locks-mem-types.h
+++ b/xlators/features/locks/src/locks-mem-types.h
@@ -11,20 +11,18 @@
#ifndef __LOCKS_MEM_TYPES_H__
#define __LOCKS_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_locks_mem_types_ {
- gf_locks_mt_pl_dom_list_t = gf_common_mt_end + 1,
- gf_locks_mt_pl_inode_t,
- gf_locks_mt_posix_lock_t,
- gf_locks_mt_pl_entry_lock_t,
- gf_locks_mt_pl_inode_lock_t,
- gf_locks_mt_truncate_ops,
- gf_locks_mt_pl_rw_req_t,
- gf_locks_mt_posix_locks_private_t,
- gf_locks_mt_pl_fdctx_t,
- gf_locks_mt_pl_meta_lock_t,
- gf_locks_mt_end
+ gf_locks_mt_pl_dom_list_t = gf_common_mt_end + 1,
+ gf_locks_mt_pl_inode_t,
+ gf_locks_mt_posix_lock_t,
+ gf_locks_mt_pl_entry_lock_t,
+ gf_locks_mt_pl_inode_lock_t,
+ gf_locks_mt_pl_rw_req_t,
+ gf_locks_mt_posix_locks_private_t,
+ gf_locks_mt_pl_fdctx_t,
+ gf_locks_mt_pl_meta_lock_t,
+ gf_locks_mt_end
};
#endif
-
diff --git a/xlators/features/locks/src/locks.h b/xlators/features/locks/src/locks.h
index c2edfff8f00..c868eb494a2 100644
--- a/xlators/features/locks/src/locks.h
+++ b/xlators/features/locks/src/locks.h
@@ -10,243 +10,283 @@
#ifndef __POSIX_LOCKS_H__
#define __POSIX_LOCKS_H__
-#include "compat-errno.h"
-#include "stack.h"
-#include "call-stub.h"
+#include <glusterfs/compat-errno.h>
+#include <glusterfs/stack.h>
+#include <glusterfs/call-stub.h>
#include "locks-mem-types.h"
-#include "client_t.h"
+#include <glusterfs/client_t.h>
-#include "lkowner.h"
+#include <glusterfs/lkowner.h>
typedef enum {
- MLK_NONE,
- MLK_FILE_BASED,
- MLK_FORCED,
- MLK_OPTIMAL
+ MLK_NONE,
+ MLK_FILE_BASED,
+ MLK_FORCED,
+ MLK_OPTIMAL
} mlk_mode_t; /* defines different mandatory locking modes*/
struct __pl_fd;
struct __posix_lock {
- struct list_head list;
+ struct list_head list;
- short fl_type;
- off_t fl_start;
- off_t fl_end;
- uint32_t lk_flags;
+ off_t fl_start;
+ off_t fl_end;
+ uint32_t lk_flags;
- short blocked; /* waiting to acquire */
- struct gf_flock user_flock; /* the flock supplied by the user */
- xlator_t *this; /* required for blocked locks */
- unsigned long fd_num;
+ short fl_type;
+ short blocked; /* waiting to acquire */
+ struct gf_flock user_flock; /* the flock supplied by the user */
+ xlator_t *this; /* required for blocked locks */
+ unsigned long fd_num;
- fd_t *fd;
- call_frame_t *frame;
+ fd_t *fd;
+ call_frame_t *frame;
- struct timeval blkd_time; /*time at which lock was queued into blkd list*/
- struct timeval granted_time; /*time at which lock was queued into active list*/
+ time_t blkd_time; /* time at which lock was queued into blkd list */
+ time_t granted_time; /* time at which lock was queued into active list */
- /* These two together serve to uniquely identify each process
- across nodes */
+ /* These two together serve to uniquely identify each process
+ across nodes */
- void *client; /* to identify client node */
+ void *client; /* to identify client node */
- /* This field uniquely identifies the client the lock belongs to. As
- * lock migration is handled by rebalance, the client_t object will be
- * overwritten by rebalance and can't be deemed as the owner of the
- * lock on destination. Hence, the below field is migrated from
- * source to destination by lock_migration_info_t and updated on the
- * destination. So that on client-server disconnection, server can
- * cleanup the locks proper;y. */
+ /* This field uniquely identifies the client the lock belongs to. As
+ * lock migration is handled by rebalance, the client_t object will be
+ * overwritten by rebalance and can't be deemed as the owner of the
+ * lock on destination. Hence, the below field is migrated from
+ * source to destination by lock_migration_info_t and updated on the
+ * destination. So that on client-server disconnection, server can
+ * cleanup the locks proper;y. */
- char *client_uid;
- gf_lkowner_t owner;
- pid_t client_pid; /* pid of client process */
+ char *client_uid;
+ gf_lkowner_t owner;
+ pid_t client_pid; /* pid of client process */
- int blocking;
+ int blocking;
};
typedef struct __posix_lock posix_lock_t;
struct __pl_inode_lock {
- struct list_head list;
- struct list_head blocked_locks; /* list_head pointing to blocked_inodelks */
- struct list_head contend; /* list of contending locks */
- int ref;
+ struct list_head list;
+ struct list_head blocked_locks; /* list_head pointing to blocked_inodelks */
+ struct list_head contend; /* list of contending locks */
+ int ref;
- short fl_type;
- off_t fl_start;
- off_t fl_end;
+ off_t fl_start;
+ off_t fl_end;
- const char *volume;
+ const char *volume;
- struct gf_flock user_flock; /* the flock supplied by the user */
- xlator_t *this; /* required for blocked locks */
- struct __pl_inode *pl_inode;
+ struct gf_flock user_flock; /* the flock supplied by the user */
+ xlator_t *this; /* required for blocked locks */
+ struct __pl_inode *pl_inode;
- call_frame_t *frame;
+ call_frame_t *frame;
- struct timeval blkd_time; /*time at which lock was queued into blkd list*/
- struct timeval granted_time; /*time at which lock was queued into active list*/
- /*last time at wich lock contention was detected and notified*/
- struct timespec contention_time;
+ time_t blkd_time; /* time at which lock was queued into blkd list */
+ time_t granted_time; /* time at which lock was queued into active list */
- /* These two together serve to uniquely identify each process
- across nodes */
+ /*last time at which lock contention was detected and notified*/
+ struct timespec contention_time;
- void *client; /* to identify client node */
- gf_lkowner_t owner;
- pid_t client_pid; /* pid of client process */
+ /* These two together serve to uniquely identify each process
+ across nodes */
- char *connection_id; /* stores the client connection id */
+ void *client; /* to identify client node */
+ gf_lkowner_t owner;
+ pid_t client_pid; /* pid of client process */
- struct list_head client_list; /* list of all locks from a client */
+ char *connection_id; /* stores the client connection id */
+
+ struct list_head client_list; /* list of all locks from a client */
+ short fl_type;
+
+ int32_t status; /* Error code when we try to grant a lock in blocked
+ state */
};
typedef struct __pl_inode_lock pl_inode_lock_t;
struct _pl_rw_req {
- struct list_head list;
- call_stub_t *stub;
- posix_lock_t region;
+ struct list_head list;
+ call_stub_t *stub;
+ posix_lock_t region;
};
typedef struct _pl_rw_req pl_rw_req_t;
struct _pl_dom_list {
- struct list_head inode_list; /* list_head back to pl_inode_t */
- const char *domain;
- struct list_head entrylk_list; /* List of entry locks */
- struct list_head blocked_entrylks; /* List of all blocked entrylks */
- struct list_head inodelk_list; /* List of inode locks */
- struct list_head blocked_inodelks; /* List of all blocked inodelks */
+ struct list_head inode_list; /* list_head back to pl_inode_t */
+ const char *domain;
+ struct list_head entrylk_list; /* List of entry locks */
+ struct list_head blocked_entrylks; /* List of all blocked entrylks */
+ struct list_head inodelk_list; /* List of inode locks */
+ struct list_head blocked_inodelks; /* List of all blocked inodelks */
};
typedef struct _pl_dom_list pl_dom_list_t;
struct __entry_lock {
- struct list_head domain_list; /* list_head back to pl_dom_list_t */
- struct list_head blocked_locks; /* list_head back to blocked_entrylks */
- struct list_head contend; /* list of contending locks */
- int ref;
+ struct list_head domain_list; /* list_head back to pl_dom_list_t */
+ struct list_head blocked_locks; /* list_head back to blocked_entrylks */
+ struct list_head contend; /* list of contending locks */
+ int ref;
- call_frame_t *frame;
- xlator_t *this;
- struct __pl_inode *pinode;
+ call_frame_t *frame;
+ xlator_t *this;
+ struct __pl_inode *pinode;
- const char *volume;
+ const char *volume;
- const char *basename;
- entrylk_type type;
+ const char *basename;
- struct timeval blkd_time; /*time at which lock was queued into blkd list*/
- struct timeval granted_time; /*time at which lock was queued into active list*/
- /*last time at wich lock contention was detected and notified*/
- struct timespec contention_time;
+ time_t blkd_time; /* time at which lock was queued into blkd list */
+ time_t granted_time; /* time at which lock was queued into active list */
- void *client;
- gf_lkowner_t owner;
- pid_t client_pid; /* pid of client process */
+ /*last time at which lock contention was detected and notified*/
+ struct timespec contention_time;
- char *connection_id; /* stores the client connection id */
+ void *client;
+ gf_lkowner_t owner;
+ pid_t client_pid; /* pid of client process */
- struct list_head client_list; /* list of all locks from a client */
+ char *connection_id; /* stores the client connection id */
+
+ struct list_head client_list; /* list of all locks from a client */
+ entrylk_type type;
};
typedef struct __entry_lock pl_entry_lock_t;
-
/* The "simulated" inode. This contains a list of all the locks associated
with this file */
struct __pl_inode {
- pthread_mutex_t mutex;
-
- struct list_head dom_list; /* list of domains */
- struct list_head ext_list; /* list of fcntl locks */
- struct list_head rw_list; /* list of waiting r/w requests */
- struct list_head reservelk_list; /* list of reservelks */
- struct list_head blocked_reservelks; /* list of blocked reservelks */
- struct list_head blocked_calls; /* List of blocked lock calls while a reserve is held*/
- struct list_head metalk_list; /* Meta lock list */
- /* This is to store the incoming lock
- requests while meta lock is enabled */
- struct list_head queued_locks;
- int mandatory; /* if mandatory locking is enabled */
-
- inode_t *refkeeper; /* hold refs on an inode while locks are
- held to prevent pruning */
- uuid_t gfid; /* placeholder for gfid of the inode */
- inode_t *inode; /* pointer to be used for ref and unref
- of inode_t as long as there are
- locks on it */
- gf_boolean_t migrated;
+ pthread_mutex_t mutex;
+
+ struct list_head dom_list; /* list of domains */
+ struct list_head ext_list; /* list of fcntl locks */
+ struct list_head rw_list; /* list of waiting r/w requests */
+ struct list_head reservelk_list; /* list of reservelks */
+ struct list_head blocked_reservelks; /* list of blocked reservelks */
+ struct list_head blocked_calls; /* List of blocked lock calls while a
+ reserve is held*/
+ struct list_head metalk_list; /* Meta lock list */
+ struct list_head queued_locks; /* This is to store the incoming lock
+ requests while meta lock is enabled */
+ struct list_head waiting; /* List of pending fops waiting to unlink/rmdir
+ the inode. */
+ int mandatory; /* if mandatory locking is enabled */
+
+ inode_t *refkeeper; /* hold refs on an inode while locks are
+ held to prevent pruning */
+ uuid_t gfid; /* placeholder for gfid of the inode */
+ inode_t *inode; /* pointer to be used for ref and unref
+ of inode_t as long as there are
+ locks on it */
+ gf_boolean_t migrated;
+
+ /* Flag to indicate whether to read mlock-enforce xattr from disk */
+ gf_boolean_t check_mlock_info;
+
+ /* Mandatory_lock enforce: IO will be allowed if and only if the lkowner has
+ held the lock.
+
+ Note: An xattr is set on the file to recover this information post
+ reboot. If client does not want mandatory lock to be enforced, then it
+ should remove this xattr explicitly
+ */
+ gf_boolean_t mlock_enforced;
+ /* There are scenarios where mandatory lock is granted but there are IOs
+ pending at posix level. To avoid this before preempting the previous lock
+ owner, we wait for all the fops to be unwound.
+ */
+ int fop_wind_count;
+ pthread_cond_t check_fop_wind_count;
+
+ gf_boolean_t track_fop_wind_count;
+
+ int32_t links; /* Number of hard links the inode has. */
+ uint32_t remove_running; /* Number of remove operations running. */
+ gf_boolean_t is_locked; /* Regular locks will be blocked. */
+ gf_boolean_t removed; /* The inode has been deleted. */
};
typedef struct __pl_inode pl_inode_t;
struct __pl_metalk {
- pthread_mutex_t mutex;
- /* For pl_inode meta lock list */
- struct list_head list;
- /* For pl_ctx_t list */
- struct list_head client_list;
- char *client_uid;
-
- pl_inode_t *pl_inode;
- int ref;
+ pthread_mutex_t mutex;
+ /* For pl_inode meta lock list */
+ struct list_head list;
+ /* For pl_ctx_t list */
+ struct list_head client_list;
+ char *client_uid;
+
+ pl_inode_t *pl_inode;
+ int ref;
};
typedef struct __pl_metalk pl_meta_lock_t;
typedef struct {
- mlk_mode_t mandatory_mode; /* holds current mandatory locking mode */
- gf_boolean_t trace; /* trace lock requests in and out */
- char *brickname;
- gf_boolean_t monkey_unlocking;
- uint32_t revocation_secs;
- gf_boolean_t revocation_clear_all;
- uint32_t revocation_max_blocked;
- gf_boolean_t notify_contention;
- uint32_t notify_contention_delay;
+ char *brickname;
+ uint32_t revocation_secs;
+ uint32_t revocation_max_blocked;
+ uint32_t notify_contention_delay;
+ mlk_mode_t mandatory_mode; /* holds current mandatory locking mode */
+ gf_boolean_t trace; /* trace lock requests in and out */
+ gf_boolean_t monkey_unlocking;
+ gf_boolean_t revocation_clear_all;
+ gf_boolean_t notify_contention;
+ gf_boolean_t mlock_enforced;
} posix_locks_private_t;
-
typedef struct {
- gf_boolean_t entrylk_count_req;
- gf_boolean_t inodelk_count_req;
- gf_boolean_t posixlk_count_req;
- gf_boolean_t parent_entrylk_req;
- data_t *inodelk_dom_count_req;
-
- dict_t *xdata;
- loc_t loc[2];
- fd_t *fd;
- off_t offset;
- glusterfs_fop_t op;
+ data_t *inodelk_dom_count_req;
+
+ dict_t *xdata;
+ loc_t loc[2];
+ fd_t *fd;
+ inode_t *inode;
+ off_t offset;
+ glusterfs_fop_t op;
+ gf_boolean_t entrylk_count_req;
+ gf_boolean_t inodelk_count_req;
+ gf_boolean_t posixlk_count_req;
+ gf_boolean_t parent_entrylk_req;
+ gf_boolean_t multiple_dom_lk_requests;
+ int update_mlock_enforced_flag;
} pl_local_t;
-
typedef struct {
- struct list_head locks_list;
+ struct list_head locks_list;
} pl_fdctx_t;
-
struct _locker {
- struct list_head lockers;
- char *volume;
- inode_t *inode;
- gf_lkowner_t owner;
+ struct list_head lockers;
+ char *volume;
+ inode_t *inode;
+ gf_lkowner_t owner;
};
typedef struct _locks_ctx {
- pthread_mutex_t lock;
- struct list_head inodelk_lockers;
- struct list_head entrylk_lockers;
- struct list_head metalk_list;
+ pthread_mutex_t lock;
+ struct list_head inodelk_lockers;
+ struct list_head entrylk_lockers;
+ struct list_head metalk_list;
} pl_ctx_t;
+typedef struct _multi_dom_lk_data {
+ xlator_t *this;
+ inode_t *inode;
+ dict_t *xdata_rsp;
+ gf_boolean_t keep_max;
+} multi_dom_lk_data;
+
+typedef enum { DECREMENT, INCREMENT } pl_count_op_t;
pl_ctx_t *
-pl_ctx_get (client_t *client, xlator_t *xlator);
+pl_ctx_get(client_t *client, xlator_t *xlator);
int
-pl_inodelk_client_cleanup (xlator_t *this, pl_ctx_t *ctx);
+pl_inodelk_client_cleanup(xlator_t *this, pl_ctx_t *ctx);
int
-pl_entrylk_client_cleanup (xlator_t *this, pl_ctx_t *ctx);
+pl_entrylk_client_cleanup(xlator_t *this, pl_ctx_t *ctx);
#endif /* __POSIX_LOCKS_H__ */
diff --git a/xlators/features/locks/src/pl-messages.h b/xlators/features/locks/src/pl-messages.h
index e5a276f35b5..e2d3d7ca974 100644
--- a/xlators/features/locks/src/pl-messages.h
+++ b/xlators/features/locks/src/pl-messages.h
@@ -11,7 +11,7 @@
#ifndef _PL_MESSAGES_H_
#define _PL_MESSAGES_H_
-#include "glfs-message-id.h"
+#include <glusterfs/glfs-message-id.h>
/* To add new message IDs, append new identifiers at the end of the list.
*
@@ -23,10 +23,7 @@
* glfs-message-id.h.
*/
-GLFS_MSGID(PL,
- PL_MSG_LOCK_NUMBER,
- PL_MSG_INODELK_CONTENTION_FAILED,
- PL_MSG_ENTRYLK_CONTENTION_FAILED
-);
+GLFS_MSGID(PL, PL_MSG_LOCK_NUMBER, PL_MSG_INODELK_CONTENTION_FAILED,
+ PL_MSG_ENTRYLK_CONTENTION_FAILED);
#endif /* !_PL_MESSAGES_H_ */
diff --git a/xlators/features/locks/src/posix.c b/xlators/features/locks/src/posix.c
index 57753dac588..cf0ae4c57dd 100644
--- a/xlators/features/locks/src/posix.c
+++ b/xlators/features/locks/src/posix.c
@@ -12,361 +12,554 @@
#include <limits.h>
#include <pthread.h>
-#include "glusterfs.h"
-#include "compat.h"
-#include "xlator.h"
-#include "logging.h"
-#include "common-utils.h"
+#include <glusterfs/compat.h>
+#include <glusterfs/logging.h>
#include "locks.h"
#include "common.h"
-#include "statedump.h"
+#include <glusterfs/statedump.h>
#include "clear.h"
-#include "defaults.h"
-#include "syncop.h"
-#include "pl-messages.h"
+#include <glusterfs/defaults.h>
+#include <glusterfs/syncop.h>
#ifndef LLONG_MAX
#define LLONG_MAX LONG_LONG_MAX /* compat with old gcc */
-#endif /* LLONG_MAX */
+#endif /* LLONG_MAX */
/* Forward declarations */
-void do_blocked_rw (pl_inode_t *);
-static int __rw_allowable (pl_inode_t *, posix_lock_t *, glusterfs_fop_t);
-static int format_brickname(char *);
-int pl_lockinfo_get_brickname (xlator_t *, inode_t *, int32_t *);
-static int fetch_pathinfo(xlator_t *, inode_t *, int32_t *, char **);
-
-#define PL_STACK_UNWIND(fop, xdata, frame, op_ret, params ...) \
- do { \
- pl_local_t *__local = NULL; \
- inode_t *__parent = NULL; \
- inode_t *__inode = NULL; \
- char *__name = NULL; \
- dict_t *__unref = NULL; \
- int __i = 0 ; \
- __local = frame->local; \
- if (op_ret >= 0 && pl_needs_xdata_response (frame->local)) {\
- if (xdata) \
- dict_ref (xdata); \
- else \
- xdata = dict_new(); \
- if (xdata) { \
- __unref = xdata; \
- while (__local->fd || __local->loc[__i].inode) { \
- pl_get_xdata_rsp_args (__local, \
- #fop, &__parent, &__inode, \
- &__name, __i); \
- pl_set_xdata_response (frame->this, \
- __local, __parent, __inode, __name, \
- xdata, __i > 0); \
- if (__local->fd || __i == 1) \
- break; \
- __i++; \
- } \
- } \
- } \
- frame->local = NULL; \
- STACK_UNWIND_STRICT (fop, frame, op_ret, params); \
- if (__local) { \
- if (__local->inodelk_dom_count_req) \
- data_unref (__local->inodelk_dom_count_req);\
- loc_wipe (&__local->loc[0]); \
- loc_wipe (&__local->loc[1]); \
- if (__local->fd) \
- fd_unref (__local->fd); \
- mem_put (__local); \
- } \
- if (__unref) \
- dict_unref (__unref); \
- } while (0)
-
-#define PL_LOCAL_GET_REQUESTS(frame, this, xdata, __fd, __loc, __newloc)\
- do { \
- if (pl_has_xdata_requests (xdata)) { \
- frame->local = mem_get0 (this->local_pool); \
- pl_local_t *__local = frame->local; \
- if (__local) { \
- if (__fd) { \
- __local->fd = fd_ref (__fd); \
- } else { \
- if (__loc) \
- loc_copy (&__local->loc[0],\
- __loc); \
- if (__newloc) \
- loc_copy (&__local->loc[1],\
- __newloc); \
- } \
- pl_get_xdata_requests (__local, xdata); \
- } \
- } \
- } while (0)
+void
+do_blocked_rw(pl_inode_t *);
+static int
+__rw_allowable(pl_inode_t *, posix_lock_t *, glusterfs_fop_t);
+static int
+format_brickname(char *);
+int
+pl_lockinfo_get_brickname(xlator_t *, inode_t *, int32_t *);
+static int
+fetch_pathinfo(xlator_t *, inode_t *, int32_t *, char **);
+
+/*
+ * The client is always requesting data, but older
+ * servers were not returning it. Newer ones are, so
+ * the client is receiving a mix of NULL and non-NULL
+ * xdata in the answers when bricks are of different
+ * versions. This triggers a bug in older clients.
+ * To prevent that, we avoid returning extra xdata to
+ * older clients (making the newer brick to behave as
+ * an old brick).
+ */
+#define PL_STACK_UNWIND_FOR_CLIENT(fop, xdata, frame, op_ret, params...) \
+ do { \
+ pl_local_t *__local = NULL; \
+ if (frame->root->client && \
+ (frame->root->client->opversion < GD_OP_VERSION_3_10_0)) { \
+ __local = frame->local; \
+ PL_STACK_UNWIND_AND_FREE(__local, fop, frame, op_ret, params); \
+ } else { \
+ PL_STACK_UNWIND(fop, xdata, frame, op_ret, params); \
+ } \
+ } while (0)
+
+#define PL_STACK_UNWIND(fop, xdata, frame, op_ret, params...) \
+ do { \
+ pl_local_t *__local = NULL; \
+ inode_t *__parent = NULL; \
+ inode_t *__inode = NULL; \
+ char *__name = NULL; \
+ dict_t *__unref = NULL; \
+ int __i = 0; \
+ __local = frame->local; \
+ if (op_ret >= 0 && pl_needs_xdata_response(frame->local)) { \
+ if (xdata) \
+ dict_ref(xdata); \
+ else \
+ xdata = dict_new(); \
+ if (xdata) { \
+ __unref = xdata; \
+ while (__local->fd || __local->loc[__i].inode) { \
+ pl_get_xdata_rsp_args(__local, #fop, &__parent, &__inode, \
+ &__name, __i); \
+ pl_set_xdata_response(frame->this, __local, __parent, \
+ __inode, __name, xdata, __i > 0); \
+ if (__local->fd || __i == 1) \
+ break; \
+ __i++; \
+ } \
+ } \
+ } \
+ PL_STACK_UNWIND_AND_FREE(__local, fop, frame, op_ret, params); \
+ if (__unref) \
+ dict_unref(__unref); \
+ } while (0)
+
+#define PL_LOCAL_GET_REQUESTS(frame, this, xdata, __fd, __loc, __newloc) \
+ do { \
+ if (pl_has_xdata_requests(xdata)) { \
+ if (!frame->local) \
+ frame->local = mem_get0(this->local_pool); \
+ pl_local_t *__local = frame->local; \
+ if (__local) { \
+ if (__fd) { \
+ __local->fd = fd_ref(__fd); \
+ __local->inode = inode_ref(__fd->inode); \
+ } else { \
+ if (__loc) \
+ loc_copy(&__local->loc[0], __loc); \
+ if (__newloc) \
+ loc_copy(&__local->loc[1], __newloc); \
+ __local->inode = inode_ref(__local->loc[0].inode); \
+ } \
+ pl_get_xdata_requests(__local, xdata); \
+ } \
+ } \
+ } while (0)
+
+#define PL_CHECK_LOCK_ENFORCE_KEY(frame, dict, name, this, loc, fd, priv) \
+ do { \
+ if ((dict && (dict_get(dict, GF_ENFORCE_MANDATORY_LOCK))) || \
+ (name && (strcmp(name, GF_ENFORCE_MANDATORY_LOCK) == 0))) { \
+ inode_t *__inode = (loc ? loc->inode : fd->inode); \
+ pl_inode_t *__pl_inode = pl_inode_get(this, __inode, NULL); \
+ if (__pl_inode == NULL) { \
+ op_ret = -1; \
+ op_errno = ENOMEM; \
+ goto unwind; \
+ } \
+ if (!pl_is_mandatory_locking_enabled(__pl_inode) || \
+ !priv->mlock_enforced) { \
+ op_ret = -1; \
+ gf_msg(this->name, GF_LOG_DEBUG, EINVAL, 0, \
+ "option %s would need mandatory lock to be enabled " \
+ "and feature.enforce-mandatory-lock option to be set " \
+ "to on", \
+ GF_ENFORCE_MANDATORY_LOCK); \
+ op_errno = EINVAL; \
+ goto unwind; \
+ } \
+ \
+ op_ret = pl_local_init(frame, this, loc, fd); \
+ if (op_ret) { \
+ op_errno = ENOMEM; \
+ goto unwind; \
+ } \
+ \
+ ((pl_local_t *)(frame->local))->update_mlock_enforced_flag = 1; \
+ } \
+ } while (0)
+
+#define PL_INODE_REMOVE(_fop, _frame, _xl, _loc1, _loc2, _cont, _cbk, \
+ _args...) \
+ ({ \
+ struct list_head contend; \
+ pl_inode_t *__pl_inode; \
+ call_stub_t *__stub; \
+ int32_t __error; \
+ INIT_LIST_HEAD(&contend); \
+ __error = pl_inode_remove_prepare(_xl, _frame, _loc2 ? _loc2 : _loc1, \
+ &__pl_inode, &contend); \
+ if (__error < 0) { \
+ __stub = fop_##_fop##_stub(_frame, _cont, ##_args); \
+ __error = pl_inode_remove_complete(_xl, __pl_inode, __stub, \
+ &contend); \
+ } else if (__error == 0) { \
+ PL_LOCAL_GET_REQUESTS(_frame, _xl, xdata, ((fd_t *)NULL), _loc1, \
+ _loc2); \
+ STACK_WIND_COOKIE(_frame, _cbk, __pl_inode, FIRST_CHILD(_xl), \
+ FIRST_CHILD(_xl)->fops->_fop, ##_args); \
+ } \
+ __error; \
+ })
gf_boolean_t
-pl_has_xdata_requests (dict_t *xdata)
-{
- char *reqs[] = {GLUSTERFS_ENTRYLK_COUNT, GLUSTERFS_INODELK_COUNT,
- GLUSTERFS_INODELK_DOM_COUNT, GLUSTERFS_POSIXLK_COUNT,
- GLUSTERFS_PARENT_ENTRYLK, NULL};
- int i = 0;
+pl_has_xdata_requests(dict_t *xdata)
+{
+ static char *reqs[] = {GLUSTERFS_ENTRYLK_COUNT,
+ GLUSTERFS_INODELK_COUNT,
+ GLUSTERFS_INODELK_DOM_COUNT,
+ GLUSTERFS_POSIXLK_COUNT,
+ GLUSTERFS_PARENT_ENTRYLK,
+ GLUSTERFS_MULTIPLE_DOM_LK_CNT_REQUESTS,
+ NULL};
+ static int reqs_size[] = {SLEN(GLUSTERFS_ENTRYLK_COUNT),
+ SLEN(GLUSTERFS_INODELK_COUNT),
+ SLEN(GLUSTERFS_INODELK_DOM_COUNT),
+ SLEN(GLUSTERFS_POSIXLK_COUNT),
+ SLEN(GLUSTERFS_PARENT_ENTRYLK),
+ SLEN(GLUSTERFS_MULTIPLE_DOM_LK_CNT_REQUESTS),
+ 0};
+ int i = 0;
+
+ if (!xdata)
+ return _gf_false;
- if (!xdata)
- return _gf_false;
+ for (i = 0; reqs[i]; i++)
+ if (dict_getn(xdata, reqs[i], reqs_size[i]))
+ return _gf_true;
- for (i = 0; reqs[i]; i++)
- if (dict_get (xdata, reqs[i]))
- return _gf_true;
+ return _gf_false;
+}
- return _gf_false;
+static int
+dict_delete_domain_key(dict_t *dict, char *key, data_t *value, void *data)
+{
+ dict_del(dict, key);
+ return 0;
}
void
-pl_get_xdata_requests (pl_local_t *local, dict_t *xdata)
+pl_get_xdata_requests(pl_local_t *local, dict_t *xdata)
{
- if (!local || !xdata)
- return;
-
- if (dict_get (xdata, GLUSTERFS_ENTRYLK_COUNT)) {
- local->entrylk_count_req = 1;
- dict_del (xdata, GLUSTERFS_ENTRYLK_COUNT);
- }
- if (dict_get (xdata, GLUSTERFS_INODELK_COUNT)) {
- local->inodelk_count_req = 1;
- dict_del (xdata, GLUSTERFS_INODELK_COUNT);
- }
-
- local->inodelk_dom_count_req = dict_get (xdata, GLUSTERFS_INODELK_DOM_COUNT);
- if (local->inodelk_dom_count_req) {
- data_ref (local->inodelk_dom_count_req);
- dict_del (xdata, GLUSTERFS_INODELK_DOM_COUNT);
- }
-
- if (dict_get (xdata, GLUSTERFS_POSIXLK_COUNT)) {
- local->posixlk_count_req = 1;
- dict_del (xdata, GLUSTERFS_POSIXLK_COUNT);
- }
+ if (!local || !xdata)
+ return;
- if (dict_get (xdata, GLUSTERFS_PARENT_ENTRYLK)) {
- local->parent_entrylk_req = 1;
- dict_del (xdata, GLUSTERFS_PARENT_ENTRYLK);
- }
+ GF_ASSERT(local->xdata == NULL);
+ local->xdata = dict_copy_with_ref(xdata, NULL);
+
+ if (dict_get_sizen(xdata, GLUSTERFS_ENTRYLK_COUNT)) {
+ local->entrylk_count_req = 1;
+ dict_del_sizen(xdata, GLUSTERFS_ENTRYLK_COUNT);
+ }
+ if (dict_get_sizen(xdata, GLUSTERFS_INODELK_COUNT)) {
+ local->inodelk_count_req = 1;
+ dict_del_sizen(xdata, GLUSTERFS_INODELK_COUNT);
+ }
+ if (dict_get_sizen(xdata, GLUSTERFS_MULTIPLE_DOM_LK_CNT_REQUESTS)) {
+ local->multiple_dom_lk_requests = 1;
+ dict_del_sizen(xdata, GLUSTERFS_MULTIPLE_DOM_LK_CNT_REQUESTS);
+ dict_foreach_fnmatch(xdata, GLUSTERFS_INODELK_DOM_PREFIX "*",
+ dict_delete_domain_key, NULL);
+ }
+
+ local->inodelk_dom_count_req = dict_get_sizen(xdata,
+ GLUSTERFS_INODELK_DOM_COUNT);
+ if (local->inodelk_dom_count_req) {
+ data_ref(local->inodelk_dom_count_req);
+ dict_del_sizen(xdata, GLUSTERFS_INODELK_DOM_COUNT);
+ }
+
+ if (dict_get_sizen(xdata, GLUSTERFS_POSIXLK_COUNT)) {
+ local->posixlk_count_req = 1;
+ dict_del_sizen(xdata, GLUSTERFS_POSIXLK_COUNT);
+ }
+
+ if (dict_get_sizen(xdata, GLUSTERFS_PARENT_ENTRYLK)) {
+ local->parent_entrylk_req = 1;
+ dict_del_sizen(xdata, GLUSTERFS_PARENT_ENTRYLK);
+ }
}
gf_boolean_t
-pl_needs_xdata_response (pl_local_t *local)
+pl_needs_xdata_response(pl_local_t *local)
{
- if (!local)
- return _gf_false;
-
- if (local->parent_entrylk_req)
- return _gf_true;
-
- if (local->entrylk_count_req)
- return _gf_true;
-
- if (local->inodelk_dom_count_req)
- return _gf_true;
+ if (!local)
+ return _gf_false;
- if (local->inodelk_count_req)
- return _gf_true;
+ if (local->parent_entrylk_req || local->entrylk_count_req ||
+ local->inodelk_dom_count_req || local->inodelk_count_req ||
+ local->posixlk_count_req || local->multiple_dom_lk_requests)
+ return _gf_true;
- if (local->posixlk_count_req)
- return _gf_true;
- return _gf_false;
+ return _gf_false;
}
void
-pl_get_xdata_rsp_args (pl_local_t *local, char *fop, inode_t **parent,
- inode_t **inode, char **name, int i)
-{
- if (strcmp (fop, "lookup") == 0) {
- *parent = local->loc[0].parent;
- *inode = local->loc[0].inode;
- *name = (char *)local->loc[0].name;
+pl_get_xdata_rsp_args(pl_local_t *local, char *fop, inode_t **parent,
+ inode_t **inode, char **name, int i)
+{
+ if (strcmp(fop, "lookup") == 0) {
+ *parent = local->loc[0].parent;
+ *inode = local->loc[0].inode;
+ *name = (char *)local->loc[0].name;
+ } else {
+ if (local->fd) {
+ *inode = local->fd->inode;
} else {
- if (local->fd) {
- *inode = local->fd->inode;
- } else {
- *inode = local->loc[i].parent;
- }
+ *inode = local->loc[i].parent;
}
+ }
}
-int32_t
-__get_posixlk_count (xlator_t *this, pl_inode_t *pl_inode)
+static inline int
+pl_track_io_fop_count(pl_local_t *local, xlator_t *this, pl_count_op_t op)
{
- posix_lock_t *lock = NULL;
- int32_t count = 0;
+ pl_inode_t *pl_inode = NULL;
+
+ if (!local)
+ return -1;
- list_for_each_entry (lock, &pl_inode->ext_list, list) {
+ pl_inode = pl_inode_get(this, local->inode, NULL);
+ if (!pl_inode)
+ return -1;
- count++;
+ if (pl_inode->mlock_enforced && pl_inode->track_fop_wind_count) {
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ if (op == DECREMENT) {
+ pl_inode->fop_wind_count--;
+ /* fop_wind_count can go negative when lock enforcement is
+ * enabled on unwind path of an IO. Hence the "<" comparision.
+ */
+ if (pl_inode->fop_wind_count <= 0) {
+ pthread_cond_broadcast(&pl_inode->check_fop_wind_count);
+ pl_inode->track_fop_wind_count = _gf_false;
+ pl_inode->fop_wind_count = 0;
+ }
+ } else {
+ pl_inode->fop_wind_count++;
+ }
}
+ pthread_mutex_unlock(&pl_inode->mutex);
+ }
- return count;
+ return 0;
+}
+
+static int32_t
+__get_posixlk_count(pl_inode_t *pl_inode)
+{
+ posix_lock_t *lock = NULL;
+ int32_t count = 0;
+
+ list_for_each_entry(lock, &pl_inode->ext_list, list) { count++; }
+
+ return count;
}
int32_t
-get_posixlk_count (xlator_t *this, inode_t *inode)
+get_posixlk_count(xlator_t *this, inode_t *inode)
{
- pl_inode_t *pl_inode = NULL;
- uint64_t tmp_pl_inode = 0;
- int ret = 0;
- int32_t count = 0;
+ pl_inode_t *pl_inode = NULL;
+ uint64_t tmp_pl_inode = 0;
+ int32_t count = 0;
- ret = inode_ctx_get (inode, this, &tmp_pl_inode);
- if (ret != 0) {
- goto out;
- }
+ int ret = inode_ctx_get(inode, this, &tmp_pl_inode);
+ if (ret != 0) {
+ goto out;
+ }
- pl_inode = (pl_inode_t *)(long) tmp_pl_inode;
+ pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
- pthread_mutex_lock (&pl_inode->mutex);
- {
- count = __get_posixlk_count (this, pl_inode);
- }
- pthread_mutex_unlock (&pl_inode->mutex);
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ count = __get_posixlk_count(pl_inode);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
out:
- return count;
+ return count;
}
void
-pl_parent_entrylk_xattr_fill (xlator_t *this, inode_t *parent,
- char *basename, dict_t *dict, gf_boolean_t keep_max)
-{
- int32_t entrylk = 0;
- int32_t maxcount = -1;
- int ret = -1;
-
- if (!parent || !basename || !strlen (basename))
- goto out;
- if (keep_max) {
- ret = dict_get_int32 (dict, GLUSTERFS_PARENT_ENTRYLK, &maxcount);
- }
- entrylk = check_entrylk_on_basename (this, parent, basename);
- if (maxcount >= entrylk)
- return;
+pl_parent_entrylk_xattr_fill(xlator_t *this, inode_t *parent, char *basename,
+ dict_t *dict, gf_boolean_t keep_max)
+{
+ int32_t entrylk = 0;
+ int32_t maxcount = -1;
+ int ret = -1;
+
+ if (!parent || !basename)
+ goto out;
+ if (keep_max) {
+ ret = dict_get_int32_sizen(dict, GLUSTERFS_PARENT_ENTRYLK, &maxcount);
+ if (ret < 0)
+ gf_msg_debug(this->name, 0, " Failed to fetch the value for key %s",
+ GLUSTERFS_PARENT_ENTRYLK);
+ }
+ entrylk = check_entrylk_on_basename(this, parent, basename);
+ if (maxcount >= entrylk)
+ return;
out:
- ret = dict_set_int32 (dict, GLUSTERFS_PARENT_ENTRYLK, entrylk);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- " dict_set failed on key %s", GLUSTERFS_PARENT_ENTRYLK);
- }
+ ret = dict_set_int32_sizen(dict, GLUSTERFS_PARENT_ENTRYLK, entrylk);
+ if (ret < 0) {
+ gf_msg_debug(this->name, 0, " dict_set failed on key %s",
+ GLUSTERFS_PARENT_ENTRYLK);
+ }
}
void
-pl_entrylk_xattr_fill (xlator_t *this, inode_t *inode,
- dict_t *dict, gf_boolean_t keep_max)
-{
- int32_t count = 0;
- int32_t maxcount = -1;
- int ret = -1;
-
- if (keep_max) {
- ret = dict_get_int32 (dict, GLUSTERFS_ENTRYLK_COUNT, &maxcount);
- }
- count = get_entrylk_count (this, inode);
- if (maxcount >= count)
- return;
-
- ret = dict_set_int32 (dict, GLUSTERFS_ENTRYLK_COUNT, count);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- " dict_set failed on key %s", GLUSTERFS_ENTRYLK_COUNT);
- }
+pl_entrylk_xattr_fill(xlator_t *this, inode_t *inode, dict_t *dict,
+ gf_boolean_t keep_max)
+{
+ int32_t count = 0;
+ int32_t maxcount = -1;
+ int ret = -1;
+
+ if (keep_max) {
+ ret = dict_get_int32_sizen(dict, GLUSTERFS_ENTRYLK_COUNT, &maxcount);
+ if (ret < 0)
+ gf_msg_debug(this->name, 0, " Failed to fetch the value for key %s",
+ GLUSTERFS_ENTRYLK_COUNT);
+ }
+ count = get_entrylk_count(this, inode);
+ if (maxcount >= count)
+ return;
+ ret = dict_set_int32_sizen(dict, GLUSTERFS_ENTRYLK_COUNT, count);
+ if (ret < 0) {
+ gf_msg_debug(this->name, 0, " dict_set failed on key %s",
+ GLUSTERFS_ENTRYLK_COUNT);
+ }
}
void
-pl_inodelk_xattr_fill (xlator_t *this, inode_t *inode, dict_t *dict,
- char *domname, gf_boolean_t keep_max)
-{
- int32_t count = 0;
- int32_t maxcount = -1;
- int ret = -1;
+pl_inodelk_xattr_fill(xlator_t *this, inode_t *inode, dict_t *dict,
+ char *domname, gf_boolean_t keep_max)
+{
+ int32_t count = 0;
+ int32_t maxcount = -1;
+ int ret = -1;
+
+ if (keep_max) {
+ ret = dict_get_int32_sizen(dict, GLUSTERFS_INODELK_COUNT, &maxcount);
+ if (ret < 0)
+ gf_msg_debug(this->name, 0, " Failed to fetch the value for key %s",
+ GLUSTERFS_INODELK_COUNT);
+ }
+ count = get_inodelk_count(this, inode, domname);
+ if (maxcount >= count)
+ return;
- if (keep_max) {
- ret = dict_get_int32 (dict, GLUSTERFS_INODELK_COUNT, &maxcount);
- }
- count = get_inodelk_count (this, inode, domname);
- if (maxcount >= count)
- return;
+ ret = dict_set_int32_sizen(dict, GLUSTERFS_INODELK_COUNT, count);
+ if (ret < 0) {
+ gf_msg_debug(this->name, 0,
+ "Failed to set count for "
+ "key %s",
+ GLUSTERFS_INODELK_COUNT);
+ }
- ret = dict_set_int32 (dict, GLUSTERFS_INODELK_COUNT, count);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG, "Failed to set count for "
- "key %s", GLUSTERFS_INODELK_COUNT);
- }
+ return;
+}
+void
+pl_posixlk_xattr_fill(xlator_t *this, inode_t *inode, dict_t *dict,
+ gf_boolean_t keep_max)
+{
+ int32_t count = 0;
+ int32_t maxcount = -1;
+ int ret = -1;
+
+ if (keep_max) {
+ ret = dict_get_int32_sizen(dict, GLUSTERFS_POSIXLK_COUNT, &maxcount);
+ if (ret < 0)
+ gf_msg_debug(this->name, 0, " Failed to fetch the value for key %s",
+ GLUSTERFS_POSIXLK_COUNT);
+ }
+ count = get_posixlk_count(this, inode);
+ if (maxcount >= count)
return;
+
+ ret = dict_set_int32_sizen(dict, GLUSTERFS_POSIXLK_COUNT, count);
+ if (ret < 0) {
+ gf_msg_debug(this->name, 0, " dict_set failed on key %s",
+ GLUSTERFS_POSIXLK_COUNT);
+ }
}
void
-pl_posixlk_xattr_fill (xlator_t *this, inode_t *inode,
- dict_t *dict, gf_boolean_t keep_max)
+pl_inodelk_xattr_fill_each(xlator_t *this, inode_t *inode, dict_t *dict,
+ char *domname, gf_boolean_t keep_max, char *key)
+{
+ int32_t count = 0;
+ int32_t maxcount = -1;
+ int ret = -1;
+
+ if (keep_max) {
+ ret = dict_get_int32(dict, key, &maxcount);
+ if (ret < 0)
+ gf_msg_debug(this->name, 0, " Failed to fetch the value for key %s",
+ GLUSTERFS_INODELK_COUNT);
+ }
+ count = get_inodelk_count(this, inode, domname);
+ if (maxcount >= count)
+ return;
+
+ ret = dict_set_int32(dict, key, count);
+ if (ret < 0) {
+ gf_msg_debug(this->name, 0,
+ "Failed to set count for "
+ "key %s",
+ key);
+ }
+
+ return;
+}
+
+static int
+pl_inodelk_xattr_fill_multiple(dict_t *this, char *key, data_t *value,
+ void *data)
{
- int32_t count = 0;
- int32_t maxcount = -1;
- int ret = -1;
+ multi_dom_lk_data *d = data;
+ char *tmp_key = NULL;
+ char *save_ptr = NULL;
- if (keep_max) {
- ret = dict_get_int32 (dict, GLUSTERFS_POSIXLK_COUNT, &maxcount);
- }
- count = get_posixlk_count (this, inode);
- if (maxcount >= count)
- return;
+ tmp_key = gf_strdup(key);
+ if (!tmp_key)
+ return -1;
- ret = dict_set_int32 (dict, GLUSTERFS_POSIXLK_COUNT, count);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- " dict_set failed on key %s", GLUSTERFS_POSIXLK_COUNT);
- }
+ strtok_r(tmp_key, ":", &save_ptr);
+ if (!*save_ptr) {
+ if (tmp_key)
+ GF_FREE(tmp_key);
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, EINVAL,
+ "Could not tokenize domain string from key %s", key);
+ return -1;
+ }
+ pl_inodelk_xattr_fill_each(d->this, d->inode, d->xdata_rsp, save_ptr,
+ d->keep_max, key);
+ if (tmp_key)
+ GF_FREE(tmp_key);
+
+ return 0;
}
void
-pl_set_xdata_response (xlator_t *this, pl_local_t *local, inode_t *parent,
- inode_t *inode, char *name, dict_t *xdata, gf_boolean_t max_lock)
+pl_fill_multiple_dom_lk_requests(xlator_t *this, pl_local_t *local,
+ inode_t *inode, dict_t *dict,
+ gf_boolean_t keep_max)
{
- if (!xdata || !local)
- return;
+ multi_dom_lk_data data;
- if (local->parent_entrylk_req && parent && name && strlen (name))
- pl_parent_entrylk_xattr_fill (this, parent, name, xdata, max_lock);
+ data.this = this;
+ data.inode = inode;
+ data.xdata_rsp = dict;
+ data.keep_max = keep_max;
- if (local->entrylk_count_req && inode)
- pl_entrylk_xattr_fill (this, inode, xdata, max_lock);
+ dict_foreach_fnmatch(local->xdata, GLUSTERFS_INODELK_DOM_PREFIX "*",
+ pl_inodelk_xattr_fill_multiple, &data);
+}
- if (local->inodelk_dom_count_req && inode)
- pl_inodelk_xattr_fill (this, inode, xdata,
- data_to_str (local->inodelk_dom_count_req), max_lock);
+void
+pl_set_xdata_response(xlator_t *this, pl_local_t *local, inode_t *parent,
+ inode_t *inode, char *name, dict_t *xdata,
+ gf_boolean_t max_lock)
+{
+ if (!xdata || !local)
+ return;
- if (local->inodelk_count_req && inode)
- pl_inodelk_xattr_fill (this, inode, xdata, NULL, max_lock);
+ if (local->parent_entrylk_req && parent && name && name[0] != '\0')
+ pl_parent_entrylk_xattr_fill(this, parent, name, xdata, max_lock);
- if (local->posixlk_count_req && inode)
- pl_posixlk_xattr_fill (this, inode, xdata, max_lock);
-}
+ if (!inode)
+ return;
-/* Return true in case we need to ensure mandatory-locking
- * semnatics under different modes.
- */
-gf_boolean_t
-pl_is_mandatory_locking_enabled (pl_inode_t *pl_inode)
-{
- posix_locks_private_t *priv = NULL;
+ if (local->entrylk_count_req)
+ pl_entrylk_xattr_fill(this, inode, xdata, max_lock);
- priv = THIS->private;
+ if (local->inodelk_dom_count_req)
+ pl_inodelk_xattr_fill(this, inode, xdata,
+ data_to_str(local->inodelk_dom_count_req),
+ max_lock);
- if (priv->mandatory_mode == MLK_FILE_BASED && pl_inode->mandatory)
- return _gf_true;
- else if (priv->mandatory_mode == MLK_FORCED ||
- priv->mandatory_mode == MLK_OPTIMAL)
- return _gf_true;
+ if (local->inodelk_count_req)
+ pl_inodelk_xattr_fill(this, inode, xdata, NULL, max_lock);
- return _gf_false;
+ if (local->posixlk_count_req)
+ pl_posixlk_xattr_fill(this, inode, xdata, max_lock);
+
+ if (local->multiple_dom_lk_requests)
+ pl_fill_multiple_dom_lk_requests(this, local, inode, xdata, max_lock);
}
/* Checks whether the region where fop is acting upon conflicts
@@ -375,4155 +568,4528 @@ pl_is_mandatory_locking_enabled (pl_inode_t *pl_inode)
* indicate block/fail the fop.
*/
int
-pl_is_fop_allowed (pl_inode_t *pl_inode, posix_lock_t *region, fd_t *fd,
- glusterfs_fop_t op, gf_boolean_t *can_block)
-{
- int ret = 0;
-
- if (!__rw_allowable (pl_inode, region, op)) {
- if ((!fd) || (fd && (fd->flags & O_NONBLOCK))) {
- gf_log ("locks", GF_LOG_TRACE, "returning EAGAIN"
- " because fd is O_NONBLOCK");
- *can_block = _gf_false;
- } else
- *can_block = _gf_true;
- } else
- ret = 1;
+pl_is_fop_allowed(pl_inode_t *pl_inode, posix_lock_t *region, fd_t *fd,
+ glusterfs_fop_t op, gf_boolean_t *can_block)
+{
+ int ret = 0;
+
+ if (!__rw_allowable(pl_inode, region, op)) {
+ if (pl_inode->mlock_enforced) {
+ *can_block = _gf_false;
+ } else if ((!fd) || (fd && (fd->flags & O_NONBLOCK))) {
+ gf_log("locks", GF_LOG_TRACE,
+ "returning EAGAIN"
+ " because fd is O_NONBLOCK");
+ *can_block = _gf_false;
+ } else {
+ *can_block = _gf_true;
+ }
+ } else {
+ ret = 1;
+ }
- return ret;
+ return ret;
}
static pl_fdctx_t *
-pl_new_fdctx ()
+pl_new_fdctx()
{
- pl_fdctx_t *fdctx = NULL;
-
- fdctx = GF_CALLOC (1, sizeof (*fdctx),
- gf_locks_mt_pl_fdctx_t);
- GF_VALIDATE_OR_GOTO ("posix-locks", fdctx, out);
+ pl_fdctx_t *fdctx = GF_MALLOC(sizeof(*fdctx), gf_locks_mt_pl_fdctx_t);
+ GF_VALIDATE_OR_GOTO("posix-locks", fdctx, out);
- INIT_LIST_HEAD (&fdctx->locks_list);
+ INIT_LIST_HEAD(&fdctx->locks_list);
out:
- return fdctx;
+ return fdctx;
}
static pl_fdctx_t *
-pl_check_n_create_fdctx (xlator_t *this, fd_t *fd)
+pl_check_n_create_fdctx(xlator_t *this, fd_t *fd)
{
- int ret = 0;
- uint64_t tmp = 0;
- pl_fdctx_t *fdctx = NULL;
+ int ret = 0;
+ uint64_t tmp = 0;
+ pl_fdctx_t *fdctx = NULL;
- GF_VALIDATE_OR_GOTO ("posix-locks", this, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
+ GF_VALIDATE_OR_GOTO("posix-locks", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
- LOCK (&fd->lock);
- {
- ret = __fd_ctx_get (fd, this, &tmp);
- if ((ret != 0) || (tmp == 0)) {
- fdctx = pl_new_fdctx ();
- if (fdctx == NULL) {
- goto unlock;
- }
- }
+ LOCK(&fd->lock);
+ {
+ ret = __fd_ctx_get(fd, this, &tmp);
+ if ((ret != 0) || (tmp == 0)) {
+ fdctx = pl_new_fdctx();
+ if (fdctx == NULL) {
+ goto unlock;
+ }
+ }
- ret = __fd_ctx_set (fd, this, (uint64_t)(long)fdctx);
- if (ret != 0) {
- GF_FREE (fdctx);
- fdctx = NULL;
- gf_log (this->name, GF_LOG_DEBUG,
- "failed to set fd ctx");
- }
+ ret = __fd_ctx_set(fd, this, (uint64_t)(long)fdctx);
+ if (ret != 0) {
+ GF_FREE(fdctx);
+ fdctx = NULL;
+ UNLOCK(&fd->lock);
+ gf_log(this->name, GF_LOG_DEBUG, "failed to set fd ctx");
+ goto out;
}
+ }
unlock:
- UNLOCK (&fd->lock);
+ UNLOCK(&fd->lock);
out:
- return fdctx;
+ return fdctx;
}
int32_t
-pl_discard_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+pl_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- STACK_UNWIND_STRICT (discard, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
- return 0;
+ pl_track_io_fop_count(frame->local, this, DECREMENT);
+
+ PL_STACK_UNWIND(discard, xdata, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
int
-pl_discard_cont (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- size_t len, dict_t *xdata)
+pl_discard_cont(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
{
- STACK_WIND (frame, pl_discard_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->discard, fd, offset, len, xdata);
- return 0;
+ pl_track_io_fop_count(frame->local, this, INCREMENT);
+
+ STACK_WIND(frame, pl_discard_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->discard, fd, offset, len, xdata);
+ return 0;
}
int32_t
-pl_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- size_t len, dict_t *xdata)
-{
- pl_inode_t *pl_inode = NULL;
- pl_rw_req_t *rw = NULL;
- posix_lock_t region = {.list = {0, }, };
- gf_boolean_t enabled = _gf_false;
- gf_boolean_t can_block = _gf_true;
- int op_ret = 0;
- int op_errno = 0;
- int allowed = 1;
-
- GF_VALIDATE_OR_GOTO ("locks", this, unwind);
+pl_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
+{
+ pl_local_t *local = NULL;
+ pl_inode_t *pl_inode = NULL;
+ pl_rw_req_t *rw = NULL;
+ posix_lock_t region = {
+ .list =
+ {
+ 0,
+ },
+ };
+ gf_boolean_t enabled = _gf_false;
+ gf_boolean_t can_block = _gf_true;
+ int op_ret = 0;
+ int op_errno = 0;
+ int allowed = 1;
+
+ GF_VALIDATE_OR_GOTO("locks", this, unwind);
+
+ local = mem_get0(this->local_pool);
+ if (!local) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ frame->local = local;
+ local->inode = inode_ref(fd->inode);
+ local->fd = fd_ref(fd);
+
+ pl_inode = pl_inode_get(this, fd->inode, local);
+ if (!pl_inode) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ if (frame->root->pid < 0)
+ enabled = _gf_false;
+ else
+ enabled = pl_is_mandatory_locking_enabled(pl_inode);
+
+ if (enabled) {
+ region.fl_start = offset;
+ region.fl_end = offset + len - 1;
+ region.client = frame->root->client;
+ region.fd_num = fd_to_fdnum(fd);
+ region.client_pid = frame->root->pid;
+ region.owner = frame->root->lk_owner;
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ allowed = pl_is_fop_allowed(pl_inode, &region, fd, GF_FOP_DISCARD,
+ &can_block);
+ if (allowed == 1) {
+ if (pl_inode->mlock_enforced &&
+ pl_inode->track_fop_wind_count) {
+ pl_inode->fop_wind_count++;
+ }
+ goto unlock;
+ } else if (!can_block) {
+ op_errno = EAGAIN;
+ op_ret = -1;
+ goto unlock;
+ }
- pl_inode = pl_inode_get (this, fd->inode);
- if (!pl_inode) {
- op_ret = -1;
+ rw = GF_MALLOC(sizeof(*rw), gf_locks_mt_pl_rw_req_t);
+ if (!rw) {
op_errno = ENOMEM;
- goto unwind;
- }
-
- enabled = pl_is_mandatory_locking_enabled (pl_inode);
-
- if (frame->root->pid < 0)
- enabled = _gf_false;
-
- if (enabled) {
- region.fl_start = offset;
- region.fl_end = offset + len - 1;
- region.client = frame->root->client;
- region.fd_num = fd_to_fdnum(fd);
- region.client_pid = frame->root->pid;
- region.owner = frame->root->lk_owner;
-
- pthread_mutex_lock (&pl_inode->mutex);
- {
- allowed = pl_is_fop_allowed (pl_inode, &region, fd,
- GF_FOP_DISCARD,
- &can_block);
- if (allowed == 1)
- goto unlock;
- else if (!can_block) {
- op_errno = EAGAIN;
- op_ret = -1;
- goto unlock;
- }
-
- rw = GF_CALLOC (1, sizeof (*rw),
- gf_locks_mt_pl_rw_req_t);
- if (!rw) {
- op_errno = ENOMEM;
- op_ret = -1;
- goto unlock;
- }
+ op_ret = -1;
+ goto unlock;
+ }
- rw->stub = fop_discard_stub (frame, pl_discard_cont,
- fd, offset, len, xdata);
- if (!rw->stub) {
- op_errno = ENOMEM;
- op_ret = -1;
- GF_FREE (rw);
- goto unlock;
- }
+ rw->stub = fop_discard_stub(frame, pl_discard_cont, fd, offset, len,
+ xdata);
+ if (!rw->stub) {
+ op_errno = ENOMEM;
+ op_ret = -1;
+ GF_FREE(rw);
+ goto unlock;
+ }
- rw->region = region;
+ rw->region = region;
- list_add_tail (&rw->list, &pl_inode->rw_list);
- }
- unlock:
- pthread_mutex_unlock (&pl_inode->mutex);
+ list_add_tail(&rw->list, &pl_inode->rw_list);
}
+ unlock:
+ pthread_mutex_unlock(&pl_inode->mutex);
+ }
- if (allowed == 1)
- STACK_WIND (frame, pl_discard_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->discard, fd, offset,
- len, xdata);
+ if (allowed == 1)
+ STACK_WIND(frame, pl_discard_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->discard, fd, offset, len, xdata);
unwind:
- if (op_ret == -1)
- STACK_UNWIND_STRICT (discard, frame, op_ret, op_errno,
- NULL, NULL, NULL);
+ if (op_ret == -1)
+ PL_STACK_UNWIND(discard, xdata, frame, op_ret, op_errno, NULL, NULL,
+ NULL);
- return 0;
+ return 0;
}
int32_t
-pl_zerofill_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+pl_zerofill_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- STACK_UNWIND_STRICT (zerofill, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
- return 0;
+ pl_track_io_fop_count(frame->local, this, DECREMENT);
+
+ PL_STACK_UNWIND(zerofill, xdata, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
int
-pl_zerofill_cont (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- off_t len, dict_t *xdata)
+pl_zerofill_cont(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ off_t len, dict_t *xdata)
{
- STACK_WIND (frame, pl_zerofill_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->zerofill, fd, offset, len, xdata);
- return 0;
+ pl_track_io_fop_count(frame->local, this, INCREMENT);
+
+ STACK_WIND(frame, pl_zerofill_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->zerofill, fd, offset, len, xdata);
+ return 0;
}
int32_t
-pl_zerofill (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- off_t len, dict_t *xdata)
-{
- pl_inode_t *pl_inode = NULL;
- pl_rw_req_t *rw = NULL;
- posix_lock_t region = {.list = {0, }, };
- gf_boolean_t enabled = _gf_false;
- gf_boolean_t can_block = _gf_true;
- int op_ret = 0;
- int op_errno = 0;
- int allowed = 1;
-
- GF_VALIDATE_OR_GOTO ("locks", this, unwind);
+pl_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ off_t len, dict_t *xdata)
+{
+ pl_local_t *local = NULL;
+ pl_inode_t *pl_inode = NULL;
+ pl_rw_req_t *rw = NULL;
+ posix_lock_t region = {
+ .list =
+ {
+ 0,
+ },
+ };
+ gf_boolean_t enabled = _gf_false;
+ gf_boolean_t can_block = _gf_true;
+ int op_ret = 0;
+ int op_errno = 0;
+ int allowed = 1;
+
+ GF_VALIDATE_OR_GOTO("locks", this, unwind);
+
+ local = mem_get0(this->local_pool);
+ if (!local) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ frame->local = local;
+ local->inode = inode_ref(fd->inode);
+ local->fd = fd_ref(fd);
+
+ pl_inode = pl_inode_get(this, fd->inode, local);
+ if (!pl_inode) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ if (frame->root->pid < 0)
+ enabled = _gf_false;
+ else
+ enabled = pl_is_mandatory_locking_enabled(pl_inode);
+
+ if (enabled) {
+ region.fl_start = offset;
+ region.fl_end = offset + len - 1;
+ region.client = frame->root->client;
+ region.fd_num = fd_to_fdnum(fd);
+ region.client_pid = frame->root->pid;
+ region.owner = frame->root->lk_owner;
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ allowed = pl_is_fop_allowed(pl_inode, &region, fd, GF_FOP_ZEROFILL,
+ &can_block);
+ if (allowed == 1) {
+ if (pl_inode->mlock_enforced &&
+ pl_inode->track_fop_wind_count) {
+ pl_inode->fop_wind_count++;
+ }
+ goto unlock;
+ } else if (!can_block) {
+ op_errno = EAGAIN;
+ op_ret = -1;
+ goto unlock;
+ }
- pl_inode = pl_inode_get (this, fd->inode);
- if (!pl_inode) {
- op_ret = -1;
+ rw = GF_MALLOC(sizeof(*rw), gf_locks_mt_pl_rw_req_t);
+ if (!rw) {
op_errno = ENOMEM;
- goto unwind;
- }
-
- enabled = pl_is_mandatory_locking_enabled (pl_inode);
-
- if (frame->root->pid < 0)
- enabled = _gf_false;
-
- if (enabled) {
- region.fl_start = offset;
- region.fl_end = offset + len - 1;
- region.client = frame->root->client;
- region.fd_num = fd_to_fdnum(fd);
- region.client_pid = frame->root->pid;
- region.owner = frame->root->lk_owner;
-
- pthread_mutex_lock (&pl_inode->mutex);
- {
- allowed = pl_is_fop_allowed (pl_inode, &region, fd,
- GF_FOP_ZEROFILL,
- &can_block);
- if (allowed == 1)
- goto unlock;
- else if (!can_block) {
- op_errno = EAGAIN;
- op_ret = -1;
- goto unlock;
- }
-
- rw = GF_CALLOC (1, sizeof (*rw),
- gf_locks_mt_pl_rw_req_t);
- if (!rw) {
- op_errno = ENOMEM;
- op_ret = -1;
- goto unlock;
- }
+ op_ret = -1;
+ goto unlock;
+ }
- rw->stub = fop_zerofill_stub (frame, pl_zerofill_cont,
- fd, offset, len, xdata);
- if (!rw->stub) {
- op_errno = ENOMEM;
- op_ret = -1;
- GF_FREE (rw);
- goto unlock;
- }
+ rw->stub = fop_zerofill_stub(frame, pl_zerofill_cont, fd, offset,
+ len, xdata);
+ if (!rw->stub) {
+ op_errno = ENOMEM;
+ op_ret = -1;
+ GF_FREE(rw);
+ goto unlock;
+ }
- rw->region = region;
+ rw->region = region;
- list_add_tail (&rw->list, &pl_inode->rw_list);
- }
- unlock:
- pthread_mutex_unlock (&pl_inode->mutex);
+ list_add_tail(&rw->list, &pl_inode->rw_list);
}
+ unlock:
+ pthread_mutex_unlock(&pl_inode->mutex);
+ }
- if (allowed == 1)
- STACK_WIND (frame, pl_zerofill_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->zerofill, fd, offset,
- len, xdata);
+ if (allowed == 1)
+ STACK_WIND(frame, pl_zerofill_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->zerofill, fd, offset, len, xdata);
unwind:
- if (op_ret == -1)
- STACK_UNWIND_STRICT (zerofill, frame, op_ret, op_errno,
- NULL, NULL, NULL);
+ if (op_ret == -1)
+ PL_STACK_UNWIND(zerofill, xdata, frame, op_ret, op_errno, NULL, NULL,
+ NULL);
- return 0;
+ return 0;
}
int
-pl_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+pl_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- pl_local_t *local = NULL;
+ pl_local_t *local = frame->local;
- local = frame->local;
+ pl_track_io_fop_count(local, this, DECREMENT);
- if (local->op == GF_FOP_TRUNCATE)
- loc_wipe (&local->loc[0]);
-
- if (local->xdata)
- dict_unref (local->xdata);
- if (local->fd)
- fd_unref (local->fd);
-
- if (local->op == GF_FOP_TRUNCATE)
- STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno,
- prebuf, postbuf, xdata);
- else
- STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno,
- prebuf, postbuf, xdata);
- return 0;
+ if (local->op == GF_FOP_TRUNCATE)
+ PL_STACK_UNWIND(truncate, xdata, frame, op_ret, op_errno, prebuf,
+ postbuf, xdata);
+ else
+ PL_STACK_UNWIND(ftruncate, xdata, frame, op_ret, op_errno, prebuf,
+ postbuf, xdata);
+ return 0;
}
int
-pl_ftruncate_cont (call_frame_t *frame, xlator_t *this, fd_t *fd,
- off_t offset, dict_t *xdata)
+pl_ftruncate_cont(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- STACK_WIND (frame, pl_truncate_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
- return 0;
+ pl_track_io_fop_count(frame->local, this, INCREMENT);
+
+ STACK_WIND(frame, pl_truncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+ return 0;
}
int
-pl_truncate_cont (call_frame_t *frame, xlator_t *this, loc_t *loc,
- off_t offset, dict_t *xdata)
+pl_truncate_cont(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- STACK_WIND (frame, pl_truncate_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
- return 0;
+ pl_track_io_fop_count(frame->local, this, INCREMENT);
+
+ STACK_WIND(frame, pl_truncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
+ return 0;
}
static int
-truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- dict_t *xdata)
-{
- pl_local_t *local = NULL;
- inode_t *inode = NULL;
- pl_inode_t *pl_inode = NULL;
- pl_rw_req_t *rw = NULL;
- posix_lock_t region = {.list = {0, }, };
- gf_boolean_t enabled = _gf_false;
- gf_boolean_t can_block = _gf_true;
- int allowed = 1;
-
- GF_VALIDATE_OR_GOTO ("locks", this, unwind);
- local = frame->local;
-
- if (op_ret != 0) {
- gf_log (this->name, GF_LOG_ERROR,
- "got error (errno=%d, stderror=%s) from child",
- op_errno, strerror (op_errno));
- goto unwind;
- }
+truncate_stat_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
+{
+ pl_local_t *local = frame->local;
+ inode_t *inode = NULL;
+ pl_inode_t *pl_inode = NULL;
+ pl_rw_req_t *rw = NULL;
+ posix_lock_t region = {
+ .list =
+ {
+ 0,
+ },
+ };
+ gf_boolean_t enabled = _gf_false;
+ gf_boolean_t can_block = _gf_true;
+ int allowed = 1;
+
+ GF_VALIDATE_OR_GOTO("locks", this, unwind);
+
+ if (op_ret != 0) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "got error (errno=%d, stderror=%s) from child", op_errno,
+ strerror(op_errno));
+ goto unwind;
+ }
+
+ if (local->op == GF_FOP_TRUNCATE)
+ inode = local->loc[0].inode;
+ else
+ inode = local->fd->inode;
+
+ local->inode = inode_ref(inode);
+
+ pl_inode = pl_inode_get(this, inode, local);
+ if (!pl_inode) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ if (frame->root->pid < 0)
+ enabled = _gf_false;
+ else
+ enabled = pl_is_mandatory_locking_enabled(pl_inode);
+
+ if (enabled) {
+ region.fl_start = local->offset;
+ region.fl_end = LLONG_MAX;
+ region.client = frame->root->client;
+ region.fd_num = fd_to_fdnum(local->fd);
+ region.client_pid = frame->root->pid;
+ region.owner = frame->root->lk_owner;
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ allowed = pl_is_fop_allowed(pl_inode, &region, local->fd, local->op,
+ &can_block);
- if (local->op == GF_FOP_TRUNCATE)
- inode = local->loc[0].inode;
- else
- inode = local->fd->inode;
+ if (allowed == 1) {
+ if (pl_inode->mlock_enforced &&
+ pl_inode->track_fop_wind_count) {
+ pl_inode->fop_wind_count++;
+ }
+ goto unlock;
+ } else if (!can_block) {
+ op_errno = EAGAIN;
+ op_ret = -1;
+ goto unlock;
+ }
- pl_inode = pl_inode_get (this, inode);
- if (!pl_inode) {
- op_ret = -1;
+ rw = GF_MALLOC(sizeof(*rw), gf_locks_mt_pl_rw_req_t);
+ if (!rw) {
op_errno = ENOMEM;
- goto unwind;
- }
-
- enabled = pl_is_mandatory_locking_enabled (pl_inode);
-
- if (frame->root->pid < 0)
- enabled = _gf_false;
-
- if (enabled) {
- region.fl_start = local->offset;
- region.fl_end = LLONG_MAX;
- region.client = frame->root->client;
- region.fd_num = fd_to_fdnum(local->fd);
- region.client_pid = frame->root->pid;
- region.owner = frame->root->lk_owner;
- pthread_mutex_lock (&pl_inode->mutex);
- {
- allowed = pl_is_fop_allowed (pl_inode, &region,
- local->fd, local->op,
- &can_block);
-
- if (allowed == 1)
- goto unlock;
- else if (!can_block) {
- op_errno = EAGAIN;
- op_ret = -1;
- goto unlock;
- }
-
- rw = GF_CALLOC (1, sizeof (*rw),
- gf_locks_mt_pl_rw_req_t);
- if (!rw) {
- op_errno = ENOMEM;
- op_ret = -1;
- goto unlock;
- }
-
- if (local->op == GF_FOP_TRUNCATE)
- rw->stub = fop_truncate_stub (frame,
- pl_truncate_cont, &local->loc[0],
- local->offset, local->xdata);
- else
- rw->stub = fop_ftruncate_stub (frame,
- pl_ftruncate_cont, local->fd,
- local->offset, local->xdata);
- if (!rw->stub) {
- op_errno = ENOMEM;
- op_ret = -1;
- GF_FREE (rw);
- goto unlock;
- }
+ op_ret = -1;
+ goto unlock;
+ }
+
+ if (local->op == GF_FOP_TRUNCATE)
+ rw->stub = fop_truncate_stub(frame, pl_truncate_cont,
+ &local->loc[0], local->offset,
+ local->xdata);
+ else
+ rw->stub = fop_ftruncate_stub(frame, pl_ftruncate_cont,
+ local->fd, local->offset,
+ local->xdata);
+ if (!rw->stub) {
+ op_errno = ENOMEM;
+ op_ret = -1;
+ GF_FREE(rw);
+ goto unlock;
+ }
- rw->region = region;
+ rw->region = region;
- list_add_tail (&rw->list, &pl_inode->rw_list);
- }
- unlock:
- pthread_mutex_unlock (&pl_inode->mutex);
+ list_add_tail(&rw->list, &pl_inode->rw_list);
}
+ unlock:
+ pthread_mutex_unlock(&pl_inode->mutex);
+ }
- if (allowed == 1) {
- switch (local->op) {
- case GF_FOP_TRUNCATE:
- STACK_WIND (frame, pl_truncate_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->truncate,
- &local->loc[0], local->offset, local->xdata);
- break;
- case GF_FOP_FTRUNCATE:
- STACK_WIND (frame, pl_truncate_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->ftruncate,
- local->fd, local->offset, local->xdata);
- break;
- default:
- break;
- }
+ if (allowed == 1) {
+ switch (local->op) {
+ case GF_FOP_TRUNCATE:
+ STACK_WIND(frame, pl_truncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, &local->loc[0],
+ local->offset, local->xdata);
+ break;
+ case GF_FOP_FTRUNCATE:
+ STACK_WIND(frame, pl_truncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, local->fd,
+ local->offset, local->xdata);
+ break;
+ default:
+ break;
}
+ }
unwind:
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_ERROR, "truncate failed with "
- "ret: %d, error: %s", op_ret, strerror (op_errno));
- if (local->op == GF_FOP_TRUNCATE)
- loc_wipe (&local->loc[0]);
- if (local->xdata)
- dict_unref (local->xdata);
- if (local->fd)
- fd_unref (local->fd);
-
- switch (local->op) {
- case GF_FOP_TRUNCATE:
- STACK_UNWIND_STRICT (truncate, frame, op_ret,
- op_errno, buf, NULL, xdata);
- break;
- case GF_FOP_FTRUNCATE:
- STACK_UNWIND_STRICT (ftruncate, frame, op_ret,
- op_errno, buf, NULL, xdata);
- break;
- default:
- break;
- }
+ if (op_ret == -1) {
+ gf_log(this ? this->name : "locks", GF_LOG_ERROR,
+ "truncate failed with "
+ "ret: %d, error: %s",
+ op_ret, strerror(op_errno));
+
+ switch (local->op) {
+ case GF_FOP_TRUNCATE:
+ PL_STACK_UNWIND(truncate, xdata, frame, op_ret, op_errno, buf,
+ NULL, xdata);
+ break;
+ case GF_FOP_FTRUNCATE:
+ PL_STACK_UNWIND(ftruncate, xdata, frame, op_ret, op_errno, buf,
+ NULL, xdata);
+ break;
+ default:
+ break;
}
- return 0;
+ }
+ return 0;
}
int
-pl_truncate (call_frame_t *frame, xlator_t *this,
- loc_t *loc, off_t offset, dict_t *xdata)
+pl_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- pl_local_t *local = NULL;
- int ret = -1;
+ pl_local_t *local = NULL;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO ("locks", this, unwind);
+ GF_VALIDATE_OR_GOTO("locks", this, unwind);
- local = mem_get0 (this->local_pool);
- GF_VALIDATE_OR_GOTO (this->name, local, unwind);
+ local = mem_get0(this->local_pool);
+ GF_VALIDATE_OR_GOTO(this->name, local, unwind);
- local->op = GF_FOP_TRUNCATE;
- local->offset = offset;
- loc_copy (&local->loc[0], loc);
- if (xdata)
- local->xdata = dict_ref (xdata);
+ local->op = GF_FOP_TRUNCATE;
+ local->offset = offset;
+ loc_copy(&local->loc[0], loc);
+ if (xdata)
+ local->xdata = dict_ref(xdata);
- frame->local = local;
+ frame->local = local;
+
+ STACK_WIND(frame, truncate_stat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->stat, loc, NULL);
+ ret = 0;
- STACK_WIND (frame, truncate_stat_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->stat, loc, NULL);
- ret = 0;
unwind:
- if (ret == -1) {
- gf_log (this->name, GF_LOG_ERROR, "truncate on %s failed with"
- " ret: %d, error: %s", loc->path, -1,
- strerror (ENOMEM));
- STACK_UNWIND_STRICT (truncate, frame, -1, ENOMEM,
- NULL, NULL, NULL);
- }
- return 0;
+ if (ret == -1) {
+ gf_log(this ? this->name : "locks", GF_LOG_ERROR,
+ "truncate on %s failed with"
+ " ret: %d, error: %s",
+ loc->path, -1, strerror(ENOMEM));
+ STACK_UNWIND_STRICT(truncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+ }
+ return 0;
}
int
-pl_ftruncate (call_frame_t *frame, xlator_t *this,
- fd_t *fd, off_t offset, dict_t *xdata)
+pl_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- pl_local_t *local = NULL;
- int ret = -1;
+ pl_local_t *local = NULL;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO ("locks", this, unwind);
- local = mem_get0 (this->local_pool);
- GF_VALIDATE_OR_GOTO (this->name, local, unwind);
+ GF_VALIDATE_OR_GOTO("locks", this, unwind);
+ local = mem_get0(this->local_pool);
+ GF_VALIDATE_OR_GOTO(this->name, local, unwind);
- local->op = GF_FOP_FTRUNCATE;
- local->offset = offset;
- local->fd = fd_ref (fd);
- if (xdata)
- local->xdata = dict_ref (xdata);
+ local->op = GF_FOP_FTRUNCATE;
+ local->offset = offset;
+ local->fd = fd_ref(fd);
+ if (xdata)
+ local->xdata = dict_ref(xdata);
- frame->local = local;
+ frame->local = local;
- STACK_WIND (frame, truncate_stat_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fstat, fd, xdata);
- ret = 0;
+ STACK_WIND(frame, truncate_stat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fstat, fd, xdata);
+ ret = 0;
unwind:
- if (ret == -1) {
- gf_log (this->name, GF_LOG_ERROR, "ftruncate failed with"
- " ret: %d, error: %s", -1, strerror (ENOMEM));
- STACK_UNWIND_STRICT (ftruncate, frame, -1, ENOMEM,
- NULL, NULL, NULL);
- }
- return 0;
+ if (ret == -1) {
+ gf_log(this ? this->name : "locks", GF_LOG_ERROR,
+ "ftruncate failed with"
+ " ret: %d, error: %s",
+ -1, strerror(ENOMEM));
+ STACK_UNWIND_STRICT(ftruncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+ }
+ return 0;
}
int
-pl_locks_by_fd (pl_inode_t *pl_inode, fd_t *fd)
+pl_locks_by_fd(pl_inode_t *pl_inode, fd_t *fd)
{
- posix_lock_t *l = NULL;
- int found = 0;
-
- pthread_mutex_lock (&pl_inode->mutex);
- {
+ posix_lock_t *l = NULL;
+ int found = 0;
- list_for_each_entry (l, &pl_inode->ext_list, list) {
- if (l->fd_num == fd_to_fdnum(fd)) {
- found = 1;
- break;
- }
- }
-
- }
- pthread_mutex_unlock (&pl_inode->mutex);
- return found;
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ list_for_each_entry(l, &pl_inode->ext_list, list)
+ {
+ if (l->fd_num == fd_to_fdnum(fd)) {
+ found = 1;
+ break;
+ }
+ }
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+ return found;
}
static void
-delete_locks_of_fd (xlator_t *this, pl_inode_t *pl_inode, fd_t *fd)
+delete_locks_of_fd(xlator_t *this, pl_inode_t *pl_inode, fd_t *fd)
{
- posix_lock_t *tmp = NULL;
- posix_lock_t *l = NULL;
-
- struct list_head blocked_list;
-
- INIT_LIST_HEAD (&blocked_list);
+ posix_lock_t *tmp = NULL;
+ posix_lock_t *l = NULL;
- pthread_mutex_lock (&pl_inode->mutex);
- {
+ struct list_head blocked_list;
- list_for_each_entry_safe (l, tmp, &pl_inode->ext_list, list) {
- if (l->fd_num == fd_to_fdnum(fd)) {
- if (l->blocked) {
- list_move_tail (&l->list, &blocked_list);
- continue;
- }
- __delete_lock (l);
- __destroy_lock (l);
- }
- }
+ INIT_LIST_HEAD(&blocked_list);
- }
- pthread_mutex_unlock (&pl_inode->mutex);
-
- list_for_each_entry_safe (l, tmp, &blocked_list, list) {
- list_del_init(&l->list);
- STACK_UNWIND_STRICT (lk, l->frame, -1, EAGAIN, &l->user_flock,
- NULL);
- __destroy_lock (l);
- }
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ list_for_each_entry_safe(l, tmp, &pl_inode->ext_list, list)
+ {
+ if (l->fd_num == fd_to_fdnum(fd)) {
+ if (l->blocked) {
+ list_move_tail(&l->list, &blocked_list);
+ continue;
+ }
+ __delete_lock(l);
+ __destroy_lock(l);
+ }
+ }
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
- grant_blocked_locks (this, pl_inode);
+ list_for_each_entry_safe(l, tmp, &blocked_list, list)
+ {
+ list_del_init(&l->list);
+ STACK_UNWIND_STRICT(lk, l->frame, -1, EAGAIN, &l->user_flock, NULL);
+ __destroy_lock(l);
+ }
- do_blocked_rw (pl_inode);
+ grant_blocked_locks(this, pl_inode);
+ do_blocked_rw(pl_inode);
}
static void
-__delete_locks_of_owner (pl_inode_t *pl_inode,
- client_t *client, gf_lkowner_t *owner)
-{
- posix_lock_t *tmp = NULL;
- posix_lock_t *l = NULL;
-
- /* TODO: what if it is a blocked lock with pending l->frame */
-
- list_for_each_entry_safe (l, tmp, &pl_inode->ext_list, list) {
- if (l->blocked)
- continue;
- if ((l->client == client) &&
- is_same_lkowner (&l->owner, owner)) {
- gf_log ("posix-locks", GF_LOG_TRACE,
- " Flushing lock"
- "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" state: %s",
- l->fl_type == F_UNLCK ? "Unlock" : "Lock",
- l->client_pid,
- lkowner_utoa (&l->owner),
- l->user_flock.l_start,
- l->user_flock.l_len,
- l->blocked == 1 ? "Blocked" : "Active");
-
- __delete_lock (l);
- __destroy_lock (l);
- }
- }
+__delete_locks_of_owner(pl_inode_t *pl_inode, client_t *client,
+ gf_lkowner_t *owner)
+{
+ posix_lock_t *tmp = NULL;
+ posix_lock_t *l = NULL;
- return;
-}
+ /* TODO: what if it is a blocked lock with pending l->frame */
+ list_for_each_entry_safe(l, tmp, &pl_inode->ext_list, list)
+ {
+ if (l->blocked)
+ continue;
+ if ((l->client == client) && is_same_lkowner(&l->owner, owner)) {
+ gf_log("posix-locks", GF_LOG_TRACE,
+ " Flushing lock"
+ "%s (pid=%d) (lk-owner=%s) %" PRId64 " - %" PRId64
+ " state: %s",
+ l->fl_type == F_UNLCK ? "Unlock" : "Lock", l->client_pid,
+ lkowner_utoa(&l->owner), l->user_flock.l_start,
+ l->user_flock.l_len, l->blocked == 1 ? "Blocked" : "Active");
-int32_t
-pl_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
-{
- STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ __delete_lock(l);
+ __destroy_lock(l);
+ }
+ }
+ return;
}
int32_t
-pl_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
-{
- int32_t op_errno = EINVAL;
- int op_ret = -1;
- int32_t bcount = 0;
- int32_t gcount = 0;
- char key[PATH_MAX] = {0, };
- char *lk_summary = NULL;
- pl_inode_t *pl_inode = NULL;
- dict_t *dict = NULL;
- clrlk_args args = {0,};
- char *brickname = NULL;
-
- if (!name)
- goto usual;
-
- if (strncmp (name, GF_XATTR_CLRLK_CMD, strlen (GF_XATTR_CLRLK_CMD)))
- goto usual;
-
- if (clrlk_parse_args (name, &args)) {
- op_errno = EINVAL;
- goto out;
+pl_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+ STACK_UNWIND_STRICT(getxattr, frame, op_ret, op_errno, dict, xdata);
+ return 0;
+}
+
+static int32_t
+pl_getxattr_clrlk(xlator_t *this, const char *name, inode_t *inode,
+ dict_t **dict, int32_t *op_errno)
+{
+ int32_t bcount = 0;
+ int32_t gcount = 0;
+ char *key = NULL;
+ char *lk_summary = NULL;
+ pl_inode_t *pl_inode = NULL;
+ clrlk_args args = {
+ 0,
+ };
+ char *brickname = NULL;
+ int32_t op_ret = -1;
+
+ *op_errno = EINVAL;
+
+ if (clrlk_parse_args(name, &args)) {
+ *op_errno = EINVAL;
+ goto out;
+ }
+
+ *dict = dict_new();
+ if (!*dict) {
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ pl_inode = pl_inode_get(this, inode, NULL);
+ if (!pl_inode) {
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ switch (args.type) {
+ case CLRLK_INODE:
+ case CLRLK_ENTRY:
+ op_ret = clrlk_clear_lks_in_all_domains(this, pl_inode, &args,
+ &bcount, &gcount, op_errno);
+ break;
+ case CLRLK_POSIX:
+ op_ret = clrlk_clear_posixlk(this, pl_inode, &args, &bcount,
+ &gcount, op_errno);
+ break;
+ default:
+ op_ret = -1;
+ *op_errno = EINVAL;
+ }
+ if (op_ret) {
+ if (args.type >= CLRLK_TYPE_MAX) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "clear locks: invalid lock type %d", args.type);
+ } else {
+ gf_log(this->name, GF_LOG_ERROR,
+ "clear locks of type %s failed: %s",
+ clrlk_type_names[args.type], strerror(*op_errno));
}
- dict = dict_new ();
- if (!dict) {
- op_errno = ENOMEM;
- goto out;
- }
+ goto out;
+ }
- pl_inode = pl_inode_get (this, loc->inode);
- if (!pl_inode) {
- op_errno = ENOMEM;
- goto out;
- }
+ op_ret = fetch_pathinfo(this, inode, op_errno, &brickname);
+ if (op_ret) {
+ gf_log(this->name, GF_LOG_WARNING, "Couldn't get brickname");
+ } else {
+ op_ret = format_brickname(brickname);
+ if (op_ret) {
+ gf_log(this->name, GF_LOG_WARNING, "Couldn't format brickname");
+ GF_FREE(brickname);
+ brickname = NULL;
+ }
+ }
+
+ if (!gcount && !bcount) {
+ if (gf_asprintf(&lk_summary, "No locks cleared.") == -1) {
+ op_ret = -1;
+ *op_errno = ENOMEM;
+ goto out;
+ }
+ } else if (gf_asprintf(&lk_summary,
+ "%s: %s blocked locks=%d "
+ "granted locks=%d",
+ (brickname == NULL) ? this->name : brickname,
+ clrlk_type_names[args.type], bcount, gcount) == -1) {
+ op_ret = -1;
+ *op_errno = ENOMEM;
+ goto out;
+ }
+ gf_log(this->name, GF_LOG_DEBUG, "%s", lk_summary);
+
+ key = gf_strdup(name);
+ if (!key) {
+ op_ret = -1;
+ goto out;
+ }
+ if (dict_set_dynstr(*dict, key, lk_summary)) {
+ op_ret = -1;
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ op_ret = 0;
- switch (args.type) {
- case CLRLK_INODE:
- case CLRLK_ENTRY:
- op_ret = clrlk_clear_lks_in_all_domains (this, pl_inode,
- &args, &bcount,
- &gcount,
- &op_errno);
- if (op_ret)
- goto out;
- break;
- case CLRLK_POSIX:
- op_ret = clrlk_clear_posixlk (this, pl_inode, &args,
- &bcount, &gcount,
- &op_errno);
- if (op_ret)
- goto out;
- break;
- case CLRLK_TYPE_MAX:
- op_errno = EINVAL;
- goto out;
- }
+out:
+ GF_FREE(brickname);
+ GF_FREE(args.opts);
+ GF_FREE(key);
+ if (op_ret) {
+ GF_FREE(lk_summary);
+ }
- op_ret = fetch_pathinfo (this, loc->inode, &op_errno, &brickname);
- if (op_ret) {
- gf_log (this->name, GF_LOG_WARNING,
- "Couldn't get brickname");
- } else {
- op_ret = format_brickname(brickname);
- if (op_ret) {
- gf_log (this->name, GF_LOG_WARNING,
- "Couldn't format brickname");
- GF_FREE(brickname);
- brickname = NULL;
- }
- }
+ return op_ret;
+}
- if (!gcount && !bcount) {
- if (gf_asprintf (&lk_summary, "No locks cleared.") == -1) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto out;
- }
- } else if (gf_asprintf (&lk_summary, "%s: %s blocked locks=%d "
- "granted locks=%d",
- (brickname == NULL)? this->name : brickname,
- (args.type == CLRLK_INODE)? "inode":
- (args.type == CLRLK_ENTRY)? "entry":
- (args.type == CLRLK_POSIX)? "posix": " ",
- bcount, gcount) == -1) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto out;
- }
+int32_t
+pl_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, const char *name,
+ dict_t *xdata)
+{
+ int32_t op_errno = EINVAL;
+ int32_t op_ret = -1;
+ dict_t *dict = NULL;
- strncpy (key, name, strlen (name));
- if (dict_set_dynstr (dict, key, lk_summary)) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto out;
- }
+ if (!name)
+ goto usual;
- op_ret = 0;
-out:
- GF_FREE(brickname);
- STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
+ if (strncmp(name, GF_XATTR_CLRLK_CMD, SLEN(GF_XATTR_CLRLK_CMD)))
+ goto usual;
- GF_FREE (args.opts);
- if (op_ret && lk_summary)
- GF_FREE (lk_summary);
- if (dict)
- dict_unref (dict);
- return 0;
+ op_ret = pl_getxattr_clrlk(this, name, loc->inode, &dict, &op_errno);
+
+ STACK_UNWIND_STRICT(getxattr, frame, op_ret, op_errno, dict, xdata);
+
+ if (dict)
+ dict_unref(dict);
+ return 0;
usual:
- STACK_WIND (frame, pl_getxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
- return 0;
+ STACK_WIND(frame, pl_getxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+ return 0;
}
static int
format_brickname(char *brickname)
{
- int ret = -1;
- char *hostname = NULL;
- char *volume = NULL;
- char *saveptr = NULL;
+ int ret = -1;
+ char *hostname = NULL;
+ char *volume = NULL;
+ char *saveptr = NULL;
- if (!brickname)
- goto out;
+ if (!brickname)
+ goto out;
- strtok_r(brickname, ":", &saveptr);
- hostname = gf_strdup(strtok_r(NULL, ":", &saveptr));
- if (hostname == NULL)
- goto out;
- volume = gf_strdup(strtok_r(NULL, ".", &saveptr));
- if (volume == NULL)
- goto out;
+ strtok_r(brickname, ":", &saveptr);
+ hostname = gf_strdup(strtok_r(NULL, ":", &saveptr));
+ if (hostname == NULL)
+ goto out;
+ volume = gf_strdup(strtok_r(NULL, ".", &saveptr));
+ if (volume == NULL)
+ goto out;
- sprintf(brickname, "%s:%s", hostname, volume);
+ sprintf(brickname, "%s:%s", hostname, volume);
- ret = 0;
+ ret = 0;
out:
- GF_FREE(hostname);
- GF_FREE(volume);
- return ret;
+ GF_FREE(hostname);
+ GF_FREE(volume);
+ return ret;
}
static int
-fetch_pathinfo (xlator_t *this, inode_t *inode, int32_t *op_errno,
- char **brickname)
-{
- int ret = -1;
- loc_t loc = {0, };
- dict_t *dict = NULL;
-
- if (!brickname)
- goto out;
-
- if (!op_errno)
- goto out;
-
- gf_uuid_copy (loc.gfid, inode->gfid);
- loc.inode = inode_ref (inode);
-
- ret = syncop_getxattr (FIRST_CHILD(this), &loc, &dict,
- GF_XATTR_PATHINFO_KEY, NULL, NULL);
- if (ret < 0) {
- *op_errno = -ret;
- ret = -1;
- goto out;
- }
-
- ret = dict_get_str (dict, GF_XATTR_PATHINFO_KEY, brickname);
- if (ret)
- goto out;
-
- *brickname = gf_strdup(*brickname);
- if (*brickname == NULL) {
- ret = -1;
- goto out;
- }
-
- ret = 0;
+fetch_pathinfo(xlator_t *this, inode_t *inode, int32_t *op_errno,
+ char **brickname)
+{
+ int ret = -1;
+ loc_t loc = {
+ 0,
+ };
+ dict_t *dict = NULL;
+
+ if (!brickname)
+ goto out;
+
+ if (!op_errno)
+ goto out;
+
+ gf_uuid_copy(loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+
+ ret = syncop_getxattr(FIRST_CHILD(this), &loc, &dict, GF_XATTR_PATHINFO_KEY,
+ NULL, NULL);
+ if (ret < 0) {
+ *op_errno = -ret;
+ ret = -1;
+ goto out;
+ }
+
+ ret = dict_get_str_sizen(dict, GF_XATTR_PATHINFO_KEY, brickname);
+ if (ret)
+ goto out;
+
+ *brickname = gf_strdup(*brickname);
+ if (*brickname == NULL) {
+ ret = -1;
+ goto out;
+ }
+
+ ret = 0;
out:
- if (dict != NULL) {
- dict_unref (dict);
- }
- loc_wipe(&loc);
+ if (dict != NULL) {
+ dict_unref(dict);
+ }
+ loc_wipe(&loc);
- return ret;
+ return ret;
}
-
int
-pl_lockinfo_get_brickname (xlator_t *this, inode_t *inode, int32_t *op_errno)
+pl_lockinfo_get_brickname(xlator_t *this, inode_t *inode, int32_t *op_errno)
{
- int ret = -1;
- posix_locks_private_t *priv = NULL;
- char *brickname = NULL;
- char *end = NULL;
- char *tmp = NULL;
-
- priv = this->private;
+ posix_locks_private_t *priv = this->private;
+ char *brickname = NULL;
+ char *end = NULL;
+ char *tmp = NULL;
- ret = fetch_pathinfo (this, inode, op_errno, &brickname);
- if (ret)
- goto out;
-
- end = strrchr (brickname, ':');
- if (!end) {
- GF_FREE(brickname);
- ret = -1;
- goto out;
- }
-
- tmp = brickname;
- brickname = gf_strndup (brickname, (end - brickname));
- if (brickname == NULL) {
- ret = -1;
- goto out;
- }
+ int ret = fetch_pathinfo(this, inode, op_errno, &brickname);
+ if (ret)
+ goto out;
- priv->brickname = brickname;
- ret = 0;
+ end = strrchr(brickname, ':');
+ if (!end) {
+ GF_FREE(brickname);
+ ret = -1;
+ goto out;
+ }
+
+ tmp = brickname;
+ brickname = gf_strndup(brickname, (end - brickname));
+ if (brickname == NULL) {
+ ret = -1;
+ goto out;
+ }
+
+ priv->brickname = brickname;
+ ret = 0;
out:
- GF_FREE(tmp);
- return ret;
+ GF_FREE(tmp);
+ return ret;
}
char *
-pl_lockinfo_key (xlator_t *this, inode_t *inode, int32_t *op_errno)
+pl_lockinfo_key(xlator_t *this, inode_t *inode, int32_t *op_errno)
{
- posix_locks_private_t *priv = NULL;
- char *key = NULL;
- int ret = 0;
-
- priv = this->private;
+ posix_locks_private_t *priv = this->private;
+ char *key = NULL;
+ int ret = 0;
- if (priv->brickname == NULL) {
- ret = pl_lockinfo_get_brickname (this, inode, op_errno);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "cannot get brickname");
- goto out;
- }
+ if (priv->brickname == NULL) {
+ ret = pl_lockinfo_get_brickname(this, inode, op_errno);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_WARNING, "cannot get brickname");
+ goto out;
}
+ }
- key = priv->brickname;
+ key = priv->brickname;
out:
- return key;
+ return key;
}
int32_t
-pl_fgetxattr_handle_lockinfo (xlator_t *this, fd_t *fd,
- dict_t *dict, int32_t *op_errno)
+pl_fgetxattr_handle_lockinfo(xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t *op_errno)
{
- pl_inode_t *pl_inode = NULL;
- char *key = NULL, *buf = NULL;
- int32_t op_ret = 0;
- unsigned long fdnum = 0;
- int32_t len = 0;
- dict_t *tmp = NULL;
-
- pl_inode = pl_inode_get (this, fd->inode);
-
- if (!pl_inode) {
- gf_log (this->name, GF_LOG_DEBUG, "Could not get inode.");
- *op_errno = EBADFD;
- op_ret = -1;
- goto out;
- }
+ char *key = NULL, *buf = NULL;
+ int32_t op_ret = 0;
+ unsigned long fdnum = 0;
+ int32_t len = 0;
+ dict_t *tmp = NULL;
- if (!pl_locks_by_fd (pl_inode, fd)) {
- op_ret = 0;
- goto out;
- }
+ pl_inode_t *pl_inode = pl_inode_get(this, fd->inode, NULL);
- fdnum = fd_to_fdnum (fd);
+ if (!pl_inode) {
+ gf_log(this->name, GF_LOG_DEBUG, "Could not get inode.");
+ *op_errno = EBADFD;
+ op_ret = -1;
+ goto out;
+ }
- key = pl_lockinfo_key (this, fd->inode, op_errno);
- if (key == NULL) {
- op_ret = -1;
- goto out;
- }
-
- tmp = dict_new ();
- if (tmp == NULL) {
- op_ret = -1;
- *op_errno = ENOMEM;
- goto out;
- }
-
- op_ret = dict_set_uint64 (tmp, key, fdnum);
- if (op_ret < 0) {
- *op_errno = -op_ret;
- op_ret = -1;
- gf_log (this->name, GF_LOG_WARNING, "setting lockinfo value "
- "(%lu) for fd (ptr:%p inode-gfid:%s) failed (%s)",
- fdnum, fd, uuid_utoa (fd->inode->gfid),
- strerror (*op_errno));
- goto out;
- }
-
- len = dict_serialized_length (tmp);
- if (len < 0) {
- *op_errno = -op_ret;
- op_ret = -1;
- gf_log (this->name, GF_LOG_WARNING,
- "dict_serialized_length failed (%s) while handling "
- "lockinfo for fd (ptr:%p inode-gfid:%s)",
- strerror (*op_errno), fd, uuid_utoa (fd->inode->gfid));
- goto out;
- }
-
- buf = GF_CALLOC (1, len, gf_common_mt_char);
- if (buf == NULL) {
- op_ret = -1;
- *op_errno = ENOMEM;
- goto out;
- }
-
- op_ret = dict_serialize (tmp, buf);
- if (op_ret < 0) {
- *op_errno = -op_ret;
- op_ret = -1;
- gf_log (this->name, GF_LOG_WARNING,
- "dict_serialize failed (%s) while handling lockinfo "
- "for fd (ptr: %p inode-gfid:%s)", strerror (*op_errno),
- fd, uuid_utoa (fd->inode->gfid));
- goto out;
- }
-
- op_ret = dict_set_dynptr (dict, GF_XATTR_LOCKINFO_KEY, buf, len);
- if (op_ret < 0) {
- *op_errno = -op_ret;
- op_ret = -1;
- gf_log (this->name, GF_LOG_WARNING, "setting lockinfo value "
- "(%lu) for fd (ptr:%p inode-gfid:%s) failed (%s)",
- fdnum, fd, uuid_utoa (fd->inode->gfid),
- strerror (*op_errno));
- goto out;
- }
-
- buf = NULL;
+ if (!pl_locks_by_fd(pl_inode, fd)) {
+ op_ret = 0;
+ goto out;
+ }
+
+ fdnum = fd_to_fdnum(fd);
+
+ key = pl_lockinfo_key(this, fd->inode, op_errno);
+ if (key == NULL) {
+ op_ret = -1;
+ goto out;
+ }
+
+ tmp = dict_new();
+ if (tmp == NULL) {
+ op_ret = -1;
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ op_ret = dict_set_uint64(tmp, key, fdnum);
+ if (op_ret < 0) {
+ *op_errno = -op_ret;
+ op_ret = -1;
+ gf_log(this->name, GF_LOG_WARNING,
+ "setting lockinfo value "
+ "(%lu) for fd (ptr:%p inode-gfid:%s) failed (%s)",
+ fdnum, fd, uuid_utoa(fd->inode->gfid), strerror(*op_errno));
+ goto out;
+ }
+
+ op_ret = dict_allocate_and_serialize(tmp, (char **)&buf,
+ (unsigned int *)&len);
+ if (op_ret != 0) {
+ *op_errno = -op_ret;
+ op_ret = -1;
+ gf_log(this->name, GF_LOG_WARNING,
+ "dict_serialized_length failed (%s) while handling "
+ "lockinfo for fd (ptr:%p inode-gfid:%s)",
+ strerror(*op_errno), fd, uuid_utoa(fd->inode->gfid));
+ goto out;
+ }
+
+ op_ret = dict_set_dynptr(dict, GF_XATTR_LOCKINFO_KEY, buf, len);
+ if (op_ret < 0) {
+ *op_errno = -op_ret;
+ op_ret = -1;
+ gf_log(this->name, GF_LOG_WARNING,
+ "setting lockinfo value "
+ "(%lu) for fd (ptr:%p inode-gfid:%s) failed (%s)",
+ fdnum, fd, uuid_utoa(fd->inode->gfid), strerror(*op_errno));
+ goto out;
+ }
+
+ buf = NULL;
out:
- if (tmp != NULL) {
- dict_unref (tmp);
- }
+ if (tmp != NULL) {
+ dict_unref(tmp);
+ }
- if (buf != NULL) {
- GF_FREE (buf);
- }
+ if (buf != NULL) {
+ GF_FREE(buf);
+ }
- return op_ret;
+ return op_ret;
}
-
int32_t
-pl_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
+pl_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
+ dict_t *xdata)
{
- int32_t op_ret = 0, op_errno = 0;
- dict_t *dict = NULL;
+ int32_t op_ret = 0, op_errno = 0;
+ dict_t *dict = NULL;
- if (!name) {
- goto usual;
+ if (!name) {
+ goto usual;
+ }
+
+ if (strcmp(name, GF_XATTR_LOCKINFO_KEY) == 0) {
+ dict = dict_new();
+ if (dict == NULL) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
}
- if (strcmp (name, GF_XATTR_LOCKINFO_KEY) == 0) {
- dict = dict_new ();
- if (dict == NULL) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
+ op_ret = pl_fgetxattr_handle_lockinfo(this, fd, dict, &op_errno);
+ if (op_ret < 0) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "getting lockinfo on fd (ptr:%p inode-gfid:%s) "
+ "failed (%s)",
+ fd, uuid_utoa(fd->inode->gfid), strerror(op_errno));
+ }
- op_ret = pl_fgetxattr_handle_lockinfo (this, fd, dict,
- &op_errno);
- if (op_ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "getting lockinfo on fd (ptr:%p inode-gfid:%s) "
- "failed (%s)", fd, uuid_utoa (fd->inode->gfid),
- strerror (op_errno));
- }
+ goto unwind;
+ } else if (strncmp(name, GF_XATTR_CLRLK_CMD, SLEN(GF_XATTR_CLRLK_CMD)) ==
+ 0) {
+ op_ret = pl_getxattr_clrlk(this, name, fd->inode, &dict, &op_errno);
- goto unwind;
- } else {
- goto usual;
- }
+ goto unwind;
+ } else {
+ goto usual;
+ }
unwind:
- STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict, NULL);
- if (dict != NULL) {
- dict_unref (dict);
- }
+ STACK_UNWIND_STRICT(fgetxattr, frame, op_ret, op_errno, dict, NULL);
+ if (dict != NULL) {
+ dict_unref(dict);
+ }
- return 0;
+ return 0;
usual:
- STACK_WIND (frame, default_fgetxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
- return 0;
+ STACK_WIND(frame, default_fgetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
+ return 0;
}
int32_t
-pl_migrate_locks (call_frame_t *frame, fd_t *newfd, uint64_t oldfd_num,
- int32_t *op_errno)
-{
- pl_inode_t *pl_inode = NULL;
- uint64_t newfd_num = 0;
- posix_lock_t *l = NULL;
- int32_t op_ret = 0;
-
- newfd_num = fd_to_fdnum (newfd);
-
- pl_inode = pl_inode_get (frame->this, newfd->inode);
- if (pl_inode == NULL) {
- op_ret = -1;
- *op_errno = EBADFD;
- goto out;
- }
-
- pthread_mutex_lock (&pl_inode->mutex);
+pl_migrate_locks(call_frame_t *frame, fd_t *newfd, uint64_t oldfd_num,
+ int32_t *op_errno)
+{
+ posix_lock_t *l = NULL;
+ int32_t op_ret = 0;
+ uint64_t newfd_num = fd_to_fdnum(newfd);
+
+ pl_inode_t *pl_inode = pl_inode_get(frame->this, newfd->inode, NULL);
+ if (pl_inode == NULL) {
+ op_ret = -1;
+ *op_errno = EBADFD;
+ goto out;
+ }
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ list_for_each_entry(l, &pl_inode->ext_list, list)
{
- list_for_each_entry (l, &pl_inode->ext_list, list) {
- if (l->fd_num == oldfd_num) {
- l->fd_num = newfd_num;
- l->client = frame->root->client;
- }
- }
+ if (l->fd_num == oldfd_num) {
+ l->fd_num = newfd_num;
+ l->client = frame->root->client;
+ }
}
- pthread_mutex_unlock (&pl_inode->mutex);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
- op_ret = 0;
+ op_ret = 0;
out:
- return op_ret;
+ return op_ret;
}
int32_t
-pl_fsetxattr_handle_lockinfo (call_frame_t *frame, fd_t *fd, char *lockinfo_buf,
- int len, int32_t *op_errno)
-{
- int32_t op_ret = -1;
- dict_t *lockinfo = NULL;
- uint64_t oldfd_num = 0;
- char *key = NULL;
-
- lockinfo = dict_new ();
- if (lockinfo == NULL) {
- op_ret = -1;
- *op_errno = ENOMEM;
- goto out;
- }
+pl_fsetxattr_handle_lockinfo(call_frame_t *frame, fd_t *fd, char *lockinfo_buf,
+ int len, int32_t *op_errno)
+{
+ int32_t op_ret = -1;
+ uint64_t oldfd_num = 0;
+ char *key = NULL;
+
+ dict_t *lockinfo = dict_new();
+ if (lockinfo == NULL) {
+ op_ret = -1;
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ op_ret = dict_unserialize(lockinfo_buf, len, &lockinfo);
+ if (op_ret < 0) {
+ *op_errno = -op_ret;
+ op_ret = -1;
+ goto out;
+ }
+
+ key = pl_lockinfo_key(frame->this, fd->inode, op_errno);
+ if (key == NULL) {
+ op_ret = -1;
+ goto out;
+ }
+
+ op_ret = dict_get_uint64(lockinfo, key, &oldfd_num);
+
+ if (oldfd_num == 0) {
+ op_ret = 0;
+ goto out;
+ }
+
+ op_ret = pl_migrate_locks(frame, fd, oldfd_num, op_errno);
+ if (op_ret < 0) {
+ gf_log(frame->this->name, GF_LOG_WARNING,
+ "migration of locks from oldfd (ptr:%p) to newfd "
+ "(ptr:%p) (inode-gfid:%s)",
+ (void *)(uintptr_t)oldfd_num, fd, uuid_utoa(fd->inode->gfid));
+ goto out;
+ }
- op_ret = dict_unserialize (lockinfo_buf, len, &lockinfo);
- if (op_ret < 0) {
- *op_errno = -op_ret;
- op_ret = -1;
- goto out;
- }
+out:
+ dict_unref(lockinfo);
- key = pl_lockinfo_key (frame->this, fd->inode, op_errno);
- if (key == NULL) {
- op_ret = -1;
- goto out;
- }
+ return op_ret;
+}
- op_ret = dict_get_uint64 (lockinfo, key, &oldfd_num);
+int32_t
+pl_fsetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+ pl_local_t *local = NULL;
+ pl_inode_t *pl_inode = NULL;
- if (oldfd_num == 0) {
- op_ret = 0;
- goto out;
+ local = frame->local;
+ if (local && local->update_mlock_enforced_flag && op_ret != -1) {
+ pl_inode = pl_inode_get(this, local->inode, NULL);
+ if (!pl_inode) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
}
- op_ret = pl_migrate_locks (frame, fd, oldfd_num, op_errno);
- if (op_ret < 0) {
- gf_log (frame->this->name, GF_LOG_WARNING,
- "migration of locks from oldfd (ptr:%p) to newfd "
- "(ptr:%p) (inode-gfid:%s)", (void *)oldfd_num, fd,
- uuid_utoa (fd->inode->gfid));
- goto out;
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ pl_inode->mlock_enforced = _gf_true;
+ pl_inode->check_mlock_info = _gf_false;
}
+ pthread_mutex_unlock(&pl_inode->mutex);
+ }
-out:
- dict_unref (lockinfo);
-
- return op_ret;
-}
-
-int32_t
-pl_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- PL_STACK_UNWIND (fsetxattr, xdata, frame, op_ret, op_errno, xdata);
- return 0;
+unwind:
+ PL_STACK_UNWIND_FOR_CLIENT(fsetxattr, xdata, frame, op_ret, op_errno,
+ xdata);
+ return 0;
}
int32_t
-pl_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
- int32_t flags, dict_t *xdata)
-{
- int32_t op_ret = 0, op_errno = 0;
- void *lockinfo_buf = NULL;
- int len = 0;
+pl_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata)
+{
+ int32_t op_errno = 0;
+ void *lockinfo_buf = NULL;
+ int len = 0;
+ char *name = NULL;
+ posix_locks_private_t *priv = this->private;
+
+ int32_t op_ret = dict_get_ptr_and_len(dict, GF_XATTR_LOCKINFO_KEY,
+ &lockinfo_buf, &len);
+ if (lockinfo_buf == NULL) {
+ goto usual;
+ }
+
+ op_ret = pl_fsetxattr_handle_lockinfo(frame, fd, lockinfo_buf, len,
+ &op_errno);
+ if (op_ret < 0) {
+ goto unwind;
+ }
- op_ret = dict_get_ptr_and_len (dict, GF_XATTR_LOCKINFO_KEY,
- &lockinfo_buf, &len);
- if (lockinfo_buf == NULL) {
- goto usual;
- }
+usual:
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
- op_ret = pl_fsetxattr_handle_lockinfo (frame, fd, lockinfo_buf, len,
- &op_errno);
- if (op_ret < 0) {
- goto unwind;
- }
+ PL_CHECK_LOCK_ENFORCE_KEY(frame, dict, name, this, ((loc_t *)NULL), fd,
+ priv);
-usual:
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, fd, NULL, NULL);
- STACK_WIND (frame, pl_fsetxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
- return 0;
+ STACK_WIND(frame, pl_fsetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+ return 0;
unwind:
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, NULL);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(fsetxattr, xdata, frame, op_ret, op_errno, NULL);
+
+ return 0;
}
int32_t
-pl_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+pl_opendir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- pl_fdctx_t *fdctx = NULL;
+ pl_fdctx_t *fdctx = NULL;
- if (op_ret < 0)
- goto unwind;
+ if (op_ret < 0)
+ goto unwind;
- fdctx = pl_check_n_create_fdctx (this, fd);
- if (!fdctx) {
- op_errno = ENOMEM;
- op_ret = -1;
- goto unwind;
- }
+ fdctx = pl_check_n_create_fdctx(this, fd);
+ if (!fdctx) {
+ op_errno = ENOMEM;
+ op_ret = -1;
+ goto unwind;
+ }
unwind:
- PL_STACK_UNWIND (opendir, xdata, frame, op_ret, op_errno, fd, xdata);
+ PL_STACK_UNWIND(opendir, xdata, frame, op_ret, op_errno, fd, xdata);
- return 0;
+ return 0;
}
int32_t
-pl_opendir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, fd_t *fd, dict_t *xdata)
+pl_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, fd, NULL, NULL);
- STACK_WIND (frame, pl_opendir_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->opendir, loc, fd, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
+ STACK_WIND(frame, pl_opendir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->opendir, loc, fd, xdata);
+ return 0;
}
int
-pl_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+pl_flush_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
+ PL_STACK_UNWIND_FOR_CLIENT(flush, xdata, frame, op_ret, op_errno, xdata);
- if (frame->root->client &&
- (frame->root->client->opversion < GD_OP_VERSION_3_10_0)) {
- STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno, xdata);
- } else {
- PL_STACK_UNWIND (flush, xdata, frame, op_ret, op_errno, xdata);
- }
-
- return 0;
+ return 0;
}
-
int
-pl_flush (call_frame_t *frame, xlator_t *this,
- fd_t *fd, dict_t *xdata)
+pl_flush(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- pl_inode_t *pl_inode = NULL;
-
- pl_inode = pl_inode_get (this, fd->inode);
+ pl_inode_t *pl_inode = pl_inode_get(this, fd->inode, NULL);
+ if (!pl_inode) {
+ gf_log(this->name, GF_LOG_DEBUG, "Could not get inode.");
+ STACK_UNWIND_STRICT(flush, frame, -1, EBADFD, NULL);
+ return 0;
+ }
- if (!pl_inode) {
- gf_log (this->name, GF_LOG_DEBUG, "Could not get inode.");
- STACK_UNWIND_STRICT (flush, frame, -1, EBADFD, NULL);
- return 0;
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ if (pl_inode->migrated) {
+ pthread_mutex_unlock(&pl_inode->mutex);
+ STACK_UNWIND_STRICT(flush, frame, -1, EREMOTE, NULL);
+ return 0;
}
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
- pthread_mutex_lock (&pl_inode->mutex);
- {
- if (pl_inode->migrated) {
- pthread_mutex_unlock (&pl_inode->mutex);
- STACK_UNWIND_STRICT (flush, frame, -1, EREMOTE,
- NULL);
- return 0;
- }
- }
- pthread_mutex_unlock (&pl_inode->mutex);
-
- pl_trace_flush (this, frame, fd);
-
- if (frame->root->lk_owner.len == 0) {
- /* Handle special case when protocol/server sets lk-owner to zero.
- * This usually happens due to a client disconnection. Hence, free
- * all locks opened with this fd.
- */
- gf_log (this->name, GF_LOG_TRACE,
- "Releasing all locks with fd %p", fd);
- delete_locks_of_fd (this, pl_inode, fd);
- goto wind;
+ pl_trace_flush(this, frame, fd);
- }
- pthread_mutex_lock (&pl_inode->mutex);
- {
- __delete_locks_of_owner (pl_inode, frame->root->client,
- &frame->root->lk_owner);
- }
- pthread_mutex_unlock (&pl_inode->mutex);
+ if (frame->root->lk_owner.len == 0) {
+ /* Handle special case when protocol/server sets lk-owner to zero.
+ * This usually happens due to a client disconnection. Hence, free
+ * all locks opened with this fd.
+ */
+ gf_log(this->name, GF_LOG_TRACE, "Releasing all locks with fd %p", fd);
+ delete_locks_of_fd(this, pl_inode, fd);
+ goto wind;
+ }
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ __delete_locks_of_owner(pl_inode, frame->root->client,
+ &frame->root->lk_owner);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
- grant_blocked_locks (this, pl_inode);
+ grant_blocked_locks(this, pl_inode);
- do_blocked_rw (pl_inode);
+ do_blocked_rw(pl_inode);
wind:
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, fd, NULL, NULL);
- STACK_WIND (frame, pl_flush_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->flush, fd, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
+ STACK_WIND(frame, pl_flush_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->flush, fd, xdata);
+ return 0;
}
-
int
-pl_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+pl_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- pl_fdctx_t *fdctx = NULL;
+ pl_fdctx_t *fdctx = NULL;
- if (op_ret < 0)
- goto unwind;
+ if (op_ret < 0)
+ goto unwind;
- fdctx = pl_check_n_create_fdctx (this, fd);
- if (!fdctx) {
- op_errno = ENOMEM;
- op_ret = -1;
- goto unwind;
- }
+ fdctx = pl_check_n_create_fdctx(this, fd);
+ if (!fdctx) {
+ op_errno = ENOMEM;
+ op_ret = -1;
+ goto unwind;
+ }
unwind:
- STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
+ STACK_UNWIND_STRICT(open, frame, op_ret, op_errno, fd, xdata);
- return 0;
+ return 0;
}
int
-pl_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- fd_t *fd, dict_t *xdata)
-{
- int op_ret = -1;
- int op_errno = EINVAL;
- pl_inode_t *pl_inode = NULL;
- posix_lock_t *l = NULL;
- posix_locks_private_t *priv = NULL;
-
- priv = this->private;
-
- GF_VALIDATE_OR_GOTO ("locks", this, unwind);
-
- op_ret = 0, op_errno = 0;
- pl_inode = pl_inode_get (this, fd->inode);
-
- /* As per design, under forced and file-based mandatory locking modes
- * it doesn't matter whether inodes's lock list contain advisory or
- * mandatory type locks. So we just check whether inode's lock list is
- * empty or not to make sure that no locks are being held for the file.
- * Whereas under optimal mandatory locking mode, we strictly fail open
- * if and only if lock list contain mandatory locks.
- */
- if (((priv->mandatory_mode == MLK_FILE_BASED) && pl_inode->mandatory) ||
- priv->mandatory_mode == MLK_FORCED) {
- if (fd->flags & O_TRUNC) {
- pthread_mutex_lock (&pl_inode->mutex);
- {
- if (!list_empty (&pl_inode->ext_list)) {
- op_ret = -1;
- op_errno = EAGAIN;
- }
- }
- pthread_mutex_unlock (&pl_inode->mutex);
+pl_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
+{
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ pl_inode_t *pl_inode = NULL;
+ posix_lock_t *l = NULL;
+ posix_locks_private_t *priv = this->private;
+
+ GF_VALIDATE_OR_GOTO("locks", this, unwind);
+
+ op_ret = 0, op_errno = 0;
+ pl_inode = pl_inode_get(this, fd->inode, NULL);
+ if (!pl_inode) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, ENOMEM, "Could not get inode");
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ /* As per design, under forced and file-based mandatory locking modes
+ * it doesn't matter whether inodes's lock list contain advisory or
+ * mandatory type locks. So we just check whether inode's lock list is
+ * empty or not to make sure that no locks are being held for the file.
+ * Whereas under optimal mandatory locking mode, we strictly fail open
+ * if and only if lock list contain mandatory locks.
+ */
+ if (((priv->mandatory_mode == MLK_FILE_BASED) && pl_inode->mandatory) ||
+ priv->mandatory_mode == MLK_FORCED) {
+ if (fd->flags & O_TRUNC) {
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ if (!list_empty(&pl_inode->ext_list)) {
+ op_ret = -1;
+ op_errno = EAGAIN;
}
- } else if (priv->mandatory_mode == MLK_OPTIMAL) {
- if (fd->flags & O_TRUNC) {
- pthread_mutex_lock (&pl_inode->mutex);
- {
- list_for_each_entry (l, &pl_inode->ext_list, list) {
- if ((l->lk_flags & GF_LK_MANDATORY)) {
- op_ret = -1;
- op_errno = EAGAIN;
- break;
- }
- }
- }
- pthread_mutex_unlock (&pl_inode->mutex);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+ }
+ } else if (priv->mandatory_mode == MLK_OPTIMAL) {
+ if (fd->flags & O_TRUNC) {
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ list_for_each_entry(l, &pl_inode->ext_list, list)
+ {
+ if ((l->lk_flags & GF_LK_MANDATORY)) {
+ op_ret = -1;
+ op_errno = EAGAIN;
+ break;
+ }
}
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
}
+ }
unwind:
- if (op_ret == -1)
- STACK_UNWIND_STRICT (open, frame, op_ret, op_errno,
- NULL, NULL);
- else
- STACK_WIND (frame, pl_open_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->open,
- loc, flags, fd, xdata);
- return 0;
+ if (op_ret == -1)
+ STACK_UNWIND_STRICT(open, frame, op_ret, op_errno, NULL, NULL);
+ else
+ STACK_WIND(frame, pl_open_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
+ return 0;
}
-
int
-pl_create_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- fd_t *fd, inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
+pl_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, fd_t *fd, inode_t *inode, struct iatt *buf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
{
- pl_fdctx_t *fdctx = NULL;
+ pl_fdctx_t *fdctx = NULL;
- if (op_ret < 0)
- goto unwind;
+ if (op_ret < 0)
+ goto unwind;
- fdctx = pl_check_n_create_fdctx (this, fd);
- if (!fdctx) {
- op_errno = ENOMEM;
- op_ret = -1;
- goto unwind;
- }
+ fdctx = pl_check_n_create_fdctx(this, fd);
+ if (!fdctx) {
+ op_errno = ENOMEM;
+ op_ret = -1;
+ goto unwind;
+ }
unwind:
- PL_STACK_UNWIND (create, xdata, frame, op_ret, op_errno, fd, inode, buf,
- preparent, postparent, xdata);
+ PL_STACK_UNWIND(create, xdata, frame, op_ret, op_errno, fd, inode, buf,
+ preparent, postparent, xdata);
- return 0;
+ return 0;
}
-
int
-pl_create (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, mode_t mode, mode_t umask, fd_t *fd,
- dict_t *xdata)
+pl_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, loc, NULL);
- STACK_WIND (frame, pl_create_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->create,
- loc, flags, mode, umask, fd, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
+
+ STACK_WIND(frame, pl_create_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, flags, mode, umask, fd,
+ xdata);
+ return 0;
}
int
-pl_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iovec *vector, int32_t count, struct iatt *stbuf,
- struct iobref *iobref, dict_t *xdata)
+pl_readv_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iovec *vector, int32_t count,
+ struct iatt *stbuf, struct iobref *iobref, dict_t *xdata)
{
- PL_STACK_UNWIND (readv, xdata, frame, op_ret, op_errno,
- vector, count, stbuf, iobref, xdata);
+ pl_track_io_fop_count(frame->local, this, DECREMENT);
- return 0;
+ PL_STACK_UNWIND(readv, xdata, frame, op_ret, op_errno, vector, count, stbuf,
+ iobref, xdata);
+
+ return 0;
}
int
-pl_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+pl_writev_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata)
{
- PL_STACK_UNWIND (writev, xdata, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
+ pl_track_io_fop_count(frame->local, this, DECREMENT);
- return 0;
+ PL_STACK_UNWIND(writev, xdata, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+
+ return 0;
}
void
-do_blocked_rw (pl_inode_t *pl_inode)
+do_blocked_rw(pl_inode_t *pl_inode)
{
- struct list_head wind_list;
- pl_rw_req_t *rw = NULL;
- pl_rw_req_t *tmp = NULL;
+ struct list_head wind_list;
+ pl_rw_req_t *rw = NULL;
+ pl_rw_req_t *tmp = NULL;
- INIT_LIST_HEAD (&wind_list);
+ INIT_LIST_HEAD(&wind_list);
- pthread_mutex_lock (&pl_inode->mutex);
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ list_for_each_entry_safe(rw, tmp, &pl_inode->rw_list, list)
{
- list_for_each_entry_safe (rw, tmp, &pl_inode->rw_list, list) {
- if (__rw_allowable (pl_inode, &rw->region,
- rw->stub->fop)) {
- list_del_init (&rw->list);
- list_add_tail (&rw->list, &wind_list);
- }
+ if (__rw_allowable(pl_inode, &rw->region, rw->stub->fop)) {
+ list_del_init(&rw->list);
+ list_add_tail(&rw->list, &wind_list);
+ if (pl_inode->mlock_enforced &&
+ pl_inode->track_fop_wind_count) {
+ pl_inode->fop_wind_count++;
}
+ }
}
- pthread_mutex_unlock (&pl_inode->mutex);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
- list_for_each_entry_safe (rw, tmp, &wind_list, list) {
- list_del_init (&rw->list);
- call_resume (rw->stub);
- GF_FREE (rw);
- }
+ list_for_each_entry_safe(rw, tmp, &wind_list, list)
+ {
+ list_del_init(&rw->list);
+ call_resume(rw->stub);
+ GF_FREE(rw);
+ }
- return;
+ return;
}
-static int
-__rw_allowable (pl_inode_t *pl_inode, posix_lock_t *region,
- glusterfs_fop_t op)
-{
- posix_lock_t *l = NULL;
- posix_locks_private_t *priv = NULL;
- int ret = 1;
-
- priv = THIS->private;
-
- list_for_each_entry (l, &pl_inode->ext_list, list) {
- if (!l->blocked && locks_overlap (l, region)
- && !same_owner (l, region)) {
- if ((op == GF_FOP_READ) && (l->fl_type != F_WRLCK))
- continue;
- /* Check for mandatory lock under optimal
- * mandatory-locking mode */
- if (priv->mandatory_mode == MLK_OPTIMAL
- && !(l->lk_flags & GF_LK_MANDATORY))
- continue;
- ret = 0;
- break;
- }
- }
-
- return ret;
-}
+/* when mandatory lock is enforced:
+ If an IO request comes on a region which is out of the boundary of the
+ granted mandatory lock, it will be rejected.
-int
-pl_readv_cont (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, uint32_t flags, dict_t *xdata)
+ Note: There is no IO blocking with mandatory lock enforced as it may be
+ a stale data from an old client.
+ */
+gf_boolean_t static within_range(posix_lock_t *existing, posix_lock_t *new)
{
- STACK_WIND (frame, pl_readv_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->readv,
- fd, size, offset, flags, xdata);
+ if (existing->fl_start <= new->fl_start && existing->fl_end >= new->fl_end)
+ return _gf_true;
- return 0;
+ return _gf_false;
}
-int
-pl_readv (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata)
+static int
+__rw_allowable(pl_inode_t *pl_inode, posix_lock_t *region, glusterfs_fop_t op)
{
- pl_inode_t *pl_inode = NULL;
- pl_rw_req_t *rw = NULL;
- posix_lock_t region = {.list = {0, }, };
- gf_boolean_t enabled = _gf_false;
- gf_boolean_t can_block = _gf_true;
- int op_ret = 0;
- int op_errno = 0;
- int allowed = 1;
+ posix_lock_t *l = NULL;
+ posix_locks_private_t *priv = THIS->private;
+ int ret = 1;
- GF_VALIDATE_OR_GOTO ("locks", this, unwind);
+ if (pl_inode->mlock_enforced) {
+ list_for_each_entry(l, &pl_inode->ext_list, list)
+ {
+ /*
+ with lock enforced (fencing) there should not be any blocking
+ lock coexisting.
+ */
+ if (same_owner(l, region)) {
+ /* Should range check be strict for same owner with fencing? */
+ if (locks_overlap(l, region)) {
+ if (within_range(l, region)) {
+ return 1;
+ } else {
+ /*
+ Should we allow read fop if it does not fit it in the
+ range?
+ if (op == GF_FOP_READ && l->fl_type != F_WRLCK) {
+ return 1;
+ }
+ */
+ return 0;
+ }
+ }
+ } else {
+ if (locks_overlap(l, region)) {
+ /*
+ with fencing should a read from a different owner be
+ allowed if the mandatory lock taken is F_RDLCK?
+ if (op == GF_FOP_READ && l->fl_type != F_WRLCK) {
+ return 1;
+ }
+ */
+ return 0;
+ }
+ }
+ }
- pl_inode = pl_inode_get (this, fd->inode);
- if (!pl_inode) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
+ /* No lock has been taken by this owner */
+ return 0;
+ }
+
+ list_for_each_entry(l, &pl_inode->ext_list, list)
+ {
+ if (!l->blocked && locks_overlap(l, region) && !same_owner(l, region)) {
+ if ((op == GF_FOP_READ) && (l->fl_type != F_WRLCK))
+ continue;
+ /* Check for mandatory lock under optimal
+ * mandatory-locking mode */
+ if (priv->mandatory_mode == MLK_OPTIMAL &&
+ !(l->lk_flags & GF_LK_MANDATORY))
+ continue;
+ ret = 0;
+ break;
}
+ }
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, fd, NULL, NULL);
- enabled = pl_is_mandatory_locking_enabled (pl_inode);
+ return ret;
+}
- if (frame->root->pid < 0)
- enabled = _gf_false;
+int
+pl_readv_cont(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
+{
+ pl_track_io_fop_count(frame->local, this, INCREMENT);
- if (enabled) {
- region.fl_start = offset;
- region.fl_end = offset + size - 1;
- region.client = frame->root->client;
- region.fd_num = fd_to_fdnum(fd);
- region.client_pid = frame->root->pid;
- region.owner = frame->root->lk_owner;
+ STACK_WIND(frame, pl_readv_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readv, fd, size, offset, flags, xdata);
- pthread_mutex_lock (&pl_inode->mutex);
- {
- allowed = pl_is_fop_allowed (pl_inode, &region, fd,
- GF_FOP_READ, &can_block);
- if (allowed == 1)
- goto unlock;
- else if (!can_block) {
- op_errno = EAGAIN;
- op_ret = -1;
- goto unlock;
- }
+ return 0;
+}
- rw = GF_CALLOC (1, sizeof (*rw),
- gf_locks_mt_pl_rw_req_t);
- if (!rw) {
- op_errno = ENOMEM;
- op_ret = -1;
- goto unlock;
- }
+int
+pl_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
+{
+ pl_local_t *local = NULL;
+ pl_inode_t *pl_inode = NULL;
+ pl_rw_req_t *rw = NULL;
+ posix_lock_t region = {
+ .list =
+ {
+ 0,
+ },
+ };
+ gf_boolean_t enabled = _gf_false;
+ gf_boolean_t can_block = _gf_true;
+ int op_ret = 0;
+ int op_errno = 0;
+ int allowed = 1;
+
+ GF_VALIDATE_OR_GOTO("locks", this, unwind);
+
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
+
+ if (!frame->local) {
+ frame->local = mem_get0(this->local_pool);
+ local = frame->local;
+ local->inode = inode_ref(fd->inode);
+ local->fd = fd_ref(fd);
+ }
+
+ pl_inode = pl_inode_get(this, fd->inode, local);
+ if (!pl_inode) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ if (frame->root->pid < 0)
+ enabled = _gf_false;
+ else
+ enabled = pl_is_mandatory_locking_enabled(pl_inode);
+
+ if (enabled) {
+ region.fl_start = offset;
+ region.fl_end = offset + size - 1;
+ region.client = frame->root->client;
+ region.fd_num = fd_to_fdnum(fd);
+ region.client_pid = frame->root->pid;
+ region.owner = frame->root->lk_owner;
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ allowed = pl_is_fop_allowed(pl_inode, &region, fd, GF_FOP_READ,
+ &can_block);
+ if (allowed == 1) {
+ if (pl_inode->mlock_enforced &&
+ pl_inode->track_fop_wind_count) {
+ pl_inode->fop_wind_count++;
+ }
+ goto unlock;
+ } else if (!can_block) {
+ op_errno = EAGAIN;
+ op_ret = -1;
+ goto unlock;
+ }
- rw->stub = fop_readv_stub (frame, pl_readv_cont,
- fd, size, offset, flags,
- xdata);
- if (!rw->stub) {
- op_errno = ENOMEM;
- op_ret = -1;
- GF_FREE (rw);
- goto unlock;
- }
+ rw = GF_MALLOC(sizeof(*rw), gf_locks_mt_pl_rw_req_t);
+ if (!rw) {
+ op_errno = ENOMEM;
+ op_ret = -1;
+ goto unlock;
+ }
- rw->region = region;
+ rw->stub = fop_readv_stub(frame, pl_readv_cont, fd, size, offset,
+ flags, xdata);
+ if (!rw->stub) {
+ op_errno = ENOMEM;
+ op_ret = -1;
+ GF_FREE(rw);
+ goto unlock;
+ }
- list_add_tail (&rw->list, &pl_inode->rw_list);
- }
- unlock:
- pthread_mutex_unlock (&pl_inode->mutex);
- }
+ rw->region = region;
- if (allowed == 1) {
- STACK_WIND (frame, pl_readv_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->readv,
- fd, size, offset, flags, xdata);
+ list_add_tail(&rw->list, &pl_inode->rw_list);
}
+ unlock:
+ pthread_mutex_unlock(&pl_inode->mutex);
+ }
+
+ if (allowed == 1) {
+ STACK_WIND(frame, pl_readv_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readv, fd, size, offset, flags,
+ xdata);
+ }
unwind:
- if (op_ret == -1)
- STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno,
- NULL, 0, NULL, NULL, NULL);
+ if (op_ret == -1)
+ PL_STACK_UNWIND(readv, xdata, frame, op_ret, op_errno, NULL, 0, NULL,
+ NULL, NULL);
- return 0;
+ return 0;
}
int
-pl_writev_cont (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int count, off_t offset,
- uint32_t flags, struct iobref *iobref, dict_t *xdata)
+pl_writev_cont(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int count, off_t offset, uint32_t flags,
+ struct iobref *iobref, dict_t *xdata)
{
- STACK_WIND (frame, pl_writev_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->writev,
- fd, vector, count, offset, flags, iobref, xdata);
+ pl_track_io_fop_count(frame->local, this, INCREMENT);
- return 0;
+ STACK_WIND(frame, pl_writev_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector, count, offset,
+ flags, iobref, xdata);
+
+ return 0;
}
int
-pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count, off_t offset,
- uint32_t flags, struct iobref *iobref, dict_t *xdata)
-{
- pl_inode_t *pl_inode = NULL;
- pl_rw_req_t *rw = NULL;
- posix_lock_t region = {.list = {0, }, };
- gf_boolean_t enabled = _gf_false;
- gf_boolean_t can_block = _gf_true;
- int op_ret = 0;
- int op_errno = 0;
- int allowed = 1;
-
- GF_VALIDATE_OR_GOTO ("locks", this, unwind);
-
- pl_inode = pl_inode_get (this, fd->inode);
- if (!pl_inode) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
-
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, fd, NULL, NULL);
- enabled = pl_is_mandatory_locking_enabled (pl_inode);
-
- if (frame->root->pid < 0)
- enabled = _gf_false;
-
- if (enabled) {
- region.fl_start = offset;
- region.fl_end = offset + iov_length (vector, count) - 1;
- region.client = frame->root->client;
- region.fd_num = fd_to_fdnum(fd);
- region.client_pid = frame->root->pid;
- region.owner = frame->root->lk_owner;
+pl_writev(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
+ int32_t count, off_t offset, uint32_t flags, struct iobref *iobref,
+ dict_t *xdata)
+{
+ pl_local_t *local = NULL;
+ pl_inode_t *pl_inode = NULL;
+ pl_rw_req_t *rw = NULL;
+ posix_lock_t region = {
+ .list =
+ {
+ 0,
+ },
+ };
+ gf_boolean_t enabled = _gf_false;
+ gf_boolean_t can_block = _gf_true;
+ int op_ret = 0;
+ int op_errno = 0;
+ int allowed = 1;
+
+ GF_VALIDATE_OR_GOTO("locks", this, unwind);
+
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
+
+ if (!frame->local) {
+ frame->local = mem_get0(this->local_pool);
+ local = frame->local;
+ local->inode = inode_ref(fd->inode);
+ local->fd = fd_ref(fd);
+ }
+
+ pl_inode = pl_inode_get(this, fd->inode, local);
+ if (!pl_inode) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ if (frame->root->pid < 0)
+ enabled = _gf_false;
+ else
+ enabled = pl_is_mandatory_locking_enabled(pl_inode);
+
+ if (enabled) {
+ region.fl_start = offset;
+ region.fl_end = offset + iov_length(vector, count) - 1;
+ region.client = frame->root->client;
+ region.fd_num = fd_to_fdnum(fd);
+ region.client_pid = frame->root->pid;
+ region.owner = frame->root->lk_owner;
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ allowed = pl_is_fop_allowed(pl_inode, &region, fd, GF_FOP_WRITE,
+ &can_block);
+ if (allowed == 1) {
+ if (pl_inode->mlock_enforced &&
+ pl_inode->track_fop_wind_count) {
+ pl_inode->fop_wind_count++;
+ }
+ goto unlock;
+ } else if (!can_block) {
+ if (pl_inode->mlock_enforced) {
+ op_errno = EBUSY;
+ } else {
+ op_errno = EAGAIN;
+ }
- pthread_mutex_lock (&pl_inode->mutex);
- {
- allowed = pl_is_fop_allowed (pl_inode, &region, fd,
- GF_FOP_WRITE, &can_block);
- if (allowed == 1)
- goto unlock;
- else if (!can_block) {
- op_errno = EAGAIN;
- op_ret = -1;
- goto unlock;
- }
+ op_ret = -1;
+ goto unlock;
+ }
- rw = GF_CALLOC (1, sizeof (*rw),
- gf_locks_mt_pl_rw_req_t);
- if (!rw) {
- op_errno = ENOMEM;
- op_ret = -1;
- goto unlock;
- }
+ rw = GF_MALLOC(sizeof(*rw), gf_locks_mt_pl_rw_req_t);
+ if (!rw) {
+ op_errno = ENOMEM;
+ op_ret = -1;
+ goto unlock;
+ }
- rw->stub = fop_writev_stub (frame, pl_writev_cont,
- fd, vector, count, offset,
- flags, iobref, xdata);
- if (!rw->stub) {
- op_errno = ENOMEM;
- op_ret = -1;
- GF_FREE (rw);
- goto unlock;
- }
+ rw->stub = fop_writev_stub(frame, pl_writev_cont, fd, vector, count,
+ offset, flags, iobref, xdata);
+ if (!rw->stub) {
+ op_errno = ENOMEM;
+ op_ret = -1;
+ GF_FREE(rw);
+ goto unlock;
+ }
- rw->region = region;
+ rw->region = region;
- list_add_tail (&rw->list, &pl_inode->rw_list);
- }
- unlock:
- pthread_mutex_unlock (&pl_inode->mutex);
+ list_add_tail(&rw->list, &pl_inode->rw_list);
}
+ unlock:
+ pthread_mutex_unlock(&pl_inode->mutex);
+ }
- if (allowed == 1) {
- STACK_WIND (frame, pl_writev_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->writev,
- fd, vector, count, offset, flags, iobref, xdata);
- }
+ if (allowed == 1) {
+ STACK_WIND(frame, pl_writev_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector, count, offset,
+ flags, iobref, xdata);
+ }
unwind:
- if (op_ret == -1)
- STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno,
- NULL, NULL, NULL);
+ if (op_ret == -1)
+ PL_STACK_UNWIND(writev, xdata, frame, op_ret, op_errno, NULL, NULL,
+ NULL);
- return 0;
+ return 0;
}
static int
-__fd_has_locks (pl_inode_t *pl_inode, fd_t *fd)
+__fd_has_locks(pl_inode_t *pl_inode, fd_t *fd)
{
- int found = 0;
- posix_lock_t *l = NULL;
+ posix_lock_t *l = NULL;
- list_for_each_entry (l, &pl_inode->ext_list, list) {
- if (l->fd_num == fd_to_fdnum(fd)) {
- found = 1;
- break;
- }
+ list_for_each_entry(l, &pl_inode->ext_list, list)
+ {
+ if (l->fd_num == fd_to_fdnum(fd)) {
+ return 1;
}
+ }
- return found;
+ return 0;
}
static posix_lock_t *
-lock_dup (posix_lock_t *lock)
+lock_dup(posix_lock_t *lock)
{
- posix_lock_t *new_lock = NULL;
-
- new_lock = new_posix_lock (&lock->user_flock, lock->client,
- lock->client_pid, &lock->owner,
- (fd_t *)lock->fd_num, lock->lk_flags,
- lock->blocking);
- return new_lock;
+ int32_t op_errno = 0;
+ return new_posix_lock(&lock->user_flock, lock->client, lock->client_pid,
+ &lock->owner, (fd_t *)lock->fd_num, lock->lk_flags,
+ lock->blocking, &op_errno);
}
static int
-__dup_locks_to_fdctx (pl_inode_t *pl_inode, fd_t *fd,
- pl_fdctx_t *fdctx)
-{
- posix_lock_t *l = NULL;
- posix_lock_t *duplock = NULL;
- int ret = 0;
-
- list_for_each_entry (l, &pl_inode->ext_list, list) {
- if (l->fd_num == fd_to_fdnum(fd)) {
- duplock = lock_dup (l);
- if (!duplock) {
- ret = -1;
- break;
- }
+__dup_locks_to_fdctx(pl_inode_t *pl_inode, fd_t *fd, pl_fdctx_t *fdctx)
+{
+ posix_lock_t *l = NULL;
+ posix_lock_t *duplock = NULL;
+ int ret = 0;
- list_add_tail (&duplock->list, &fdctx->locks_list);
- }
+ list_for_each_entry(l, &pl_inode->ext_list, list)
+ {
+ if (l->fd_num == fd_to_fdnum(fd)) {
+ duplock = lock_dup(l);
+ if (!duplock) {
+ ret = -1;
+ break;
+ }
+
+ list_add_tail(&duplock->list, &fdctx->locks_list);
}
+ }
- return ret;
+ return ret;
}
static int
-__copy_locks_to_fdctx (pl_inode_t *pl_inode, fd_t *fd,
- pl_fdctx_t *fdctx)
+__copy_locks_to_fdctx(pl_inode_t *pl_inode, fd_t *fd, pl_fdctx_t *fdctx)
{
- int ret = 0;
-
- ret = __dup_locks_to_fdctx (pl_inode, fd, fdctx);
- if (ret)
- goto out;
-
-out:
- return ret;
-
+ return __dup_locks_to_fdctx(pl_inode, fd, fdctx);
}
static void
-pl_mark_eol_lock (posix_lock_t *lock)
+pl_mark_eol_lock(posix_lock_t *lock)
{
- lock->user_flock.l_type = GF_LK_EOL;
- return;
+ lock->user_flock.l_type = GF_LK_EOL;
+ return;
}
static posix_lock_t *
-__get_next_fdctx_lock (pl_fdctx_t *fdctx)
+__get_next_fdctx_lock(pl_fdctx_t *fdctx)
{
- posix_lock_t *lock = NULL;
+ posix_lock_t *lock = NULL;
- GF_ASSERT (fdctx);
+ GF_ASSERT(fdctx);
- if (list_empty (&fdctx->locks_list)) {
- gf_log (THIS->name, GF_LOG_DEBUG,
- "fdctx lock list empty");
- goto out;
- }
+ if (list_empty(&fdctx->locks_list)) {
+ gf_log(THIS->name, GF_LOG_DEBUG, "fdctx lock list empty");
+ goto out;
+ }
- lock = list_entry (fdctx->locks_list.next, typeof (*lock),
- list);
+ lock = list_entry(fdctx->locks_list.next, typeof(*lock), list);
- GF_ASSERT (lock);
+ GF_ASSERT(lock);
- list_del_init (&lock->list);
+ list_del_init(&lock->list);
out:
- return lock;
+ return lock;
}
static int
-__set_next_lock_fd (pl_fdctx_t *fdctx, posix_lock_t *reqlock)
+__set_next_lock_fd(pl_fdctx_t *fdctx, posix_lock_t *reqlock)
{
- posix_lock_t *lock = NULL;
- int ret = 0;
+ posix_lock_t *lock = NULL;
+ int ret = 0;
- GF_ASSERT (fdctx);
+ GF_ASSERT(fdctx);
- lock = __get_next_fdctx_lock (fdctx);
- if (!lock) {
- gf_log (THIS->name, GF_LOG_DEBUG,
- "marking EOL in reqlock");
- pl_mark_eol_lock (reqlock);
- goto out;
- }
+ lock = __get_next_fdctx_lock(fdctx);
+ if (!lock) {
+ gf_log(THIS->name, GF_LOG_DEBUG, "marking EOL in reqlock");
+ pl_mark_eol_lock(reqlock);
+ goto out;
+ }
- reqlock->user_flock = lock->user_flock;
- reqlock->fl_start = lock->fl_start;
- reqlock->fl_type = lock->fl_type;
- reqlock->fl_end = lock->fl_end;
- reqlock->owner = lock->owner;
+ reqlock->user_flock = lock->user_flock;
+ reqlock->fl_start = lock->fl_start;
+ reqlock->fl_type = lock->fl_type;
+ reqlock->fl_end = lock->fl_end;
+ reqlock->owner = lock->owner;
out:
- if (lock)
- __destroy_lock (lock);
+ if (lock)
+ __destroy_lock(lock);
- return ret;
+ return ret;
}
static int
-pl_getlk_fd (xlator_t *this, pl_inode_t *pl_inode,
- fd_t *fd, posix_lock_t *reqlock)
+pl_getlk_fd(xlator_t *this, pl_inode_t *pl_inode, fd_t *fd,
+ posix_lock_t *reqlock)
{
- uint64_t tmp = 0;
- pl_fdctx_t *fdctx = NULL;
- int ret = 0;
+ uint64_t tmp = 0;
+ pl_fdctx_t *fdctx = NULL;
+ int ret = 0;
- pthread_mutex_lock (&pl_inode->mutex);
- {
- if (!__fd_has_locks (pl_inode, fd)) {
- gf_log (this->name, GF_LOG_DEBUG,
- "fd=%p has no active locks", fd);
- ret = 0;
- goto unlock;
- }
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ if (!__fd_has_locks(pl_inode, fd)) {
+ pthread_mutex_unlock(&pl_inode->mutex);
+ gf_log(this->name, GF_LOG_DEBUG, "fd=%p has no active locks", fd);
+ ret = 0;
+ goto out;
+ }
- gf_log (this->name, GF_LOG_DEBUG,
- "There are active locks on fd");
+ gf_log(this->name, GF_LOG_DEBUG, "There are active locks on fd");
- ret = fd_ctx_get (fd, this, &tmp);
- fdctx = (pl_fdctx_t *)(long) tmp;
+ ret = fd_ctx_get(fd, this, &tmp);
+ fdctx = (pl_fdctx_t *)(long)tmp;
- if (list_empty (&fdctx->locks_list)) {
- gf_log (this->name, GF_LOG_TRACE,
- "no fdctx -> copying all locks on fd");
+ if (list_empty(&fdctx->locks_list)) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "no fdctx -> copying all locks on fd");
- ret = __copy_locks_to_fdctx (pl_inode, fd, fdctx);
- if (ret) {
- goto unlock;
- }
+ ret = __copy_locks_to_fdctx(pl_inode, fd, fdctx);
+ if (ret) {
+ goto unlock;
+ }
- ret = __set_next_lock_fd (fdctx, reqlock);
+ ret = __set_next_lock_fd(fdctx, reqlock);
- } else {
- gf_log (this->name, GF_LOG_TRACE,
- "fdctx present -> returning the next lock");
- ret = __set_next_lock_fd (fdctx, reqlock);
- if (ret) {
- gf_log (this->name, GF_LOG_DEBUG,
- "could not get next lock of fd");
- goto unlock;
- }
- }
+ } else {
+ gf_log(this->name, GF_LOG_TRACE,
+ "fdctx present -> returning the next lock");
+ ret = __set_next_lock_fd(fdctx, reqlock);
+ if (ret) {
+ pthread_mutex_unlock(&pl_inode->mutex);
+ gf_log(this->name, GF_LOG_DEBUG,
+ "could not get next lock of fd");
+ goto out;
+ }
}
+ }
unlock:
- pthread_mutex_unlock (&pl_inode->mutex);
- return ret;
-
+ pthread_mutex_unlock(&pl_inode->mutex);
+out:
+ return ret;
}
int
-pl_metalock_is_active (pl_inode_t *pl_inode)
+pl_metalock_is_active(pl_inode_t *pl_inode)
{
- if (list_empty (&pl_inode->metalk_list))
- return 0;
- else
- return 1;
+ if (list_empty(&pl_inode->metalk_list))
+ return 0;
+ else
+ return 1;
}
-int
-__pl_queue_lock (pl_inode_t *pl_inode, posix_lock_t *reqlock, int can_block)
+void
+__pl_queue_lock(pl_inode_t *pl_inode, posix_lock_t *reqlock)
{
- list_add_tail (&reqlock->list, &pl_inode->queued_locks);
-
- return 0;
+ list_add_tail(&reqlock->list, &pl_inode->queued_locks);
}
int
-pl_lk (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
-{
- pl_inode_t *pl_inode = NULL;
- int op_ret = 0;
- int op_errno = 0;
- int can_block = 0;
- posix_lock_t *reqlock = NULL;
- posix_lock_t *conf = NULL;
- int ret = 0;
- uint32_t lk_flags = 0;
- posix_locks_private_t *priv = NULL;
-
- priv = this->private;
-
- ret = dict_get_uint32 (xdata, GF_LOCK_MODE, &lk_flags);
- if (ret == 0) {
- if (priv->mandatory_mode == MLK_NONE)
- gf_log (this->name, GF_LOG_DEBUG, "Lock flags received "
- "in a non-mandatory locking environment, "
- "continuing");
- else
- gf_log (this->name, GF_LOG_DEBUG, "Lock flags received, "
- "continuing");
- }
-
- if ((flock->l_start < 0) ||
- ((flock->l_start + flock->l_len) < 0)) {
- op_ret = -1;
- op_errno = EINVAL;
- goto unwind;
- }
-
- /* As per 'man 3 fcntl', the value of l_len may be
- * negative. In such cases, lock request should be
- * considered for the range starting at 'l_start+l_len'
- * and ending at 'l_start-1'. Update the fields accordingly.
- */
- if (flock->l_len < 0) {
- flock->l_start += flock->l_len;
- flock->l_len = labs (flock->l_len);
- }
-
- pl_inode = pl_inode_get (this, fd->inode);
- if (!pl_inode) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
+pl_lk(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
+ struct gf_flock *flock, dict_t *xdata)
+{
+ pl_inode_t *pl_inode = NULL;
+ int op_ret = 0;
+ int op_errno = 0;
+ int can_block = 0;
+ posix_lock_t *reqlock = NULL;
+ posix_lock_t *conf = NULL;
+ uint32_t lk_flags = 0;
+ posix_locks_private_t *priv = this->private;
+ pl_local_t *local = NULL;
+ short lock_type = 0;
+
+ int ret = dict_get_uint32(xdata, GF_LOCK_MODE, &lk_flags);
+ if (ret == 0) {
+ if (priv->mandatory_mode == MLK_NONE)
+ gf_log(this->name, GF_LOG_DEBUG,
+ "Lock flags received "
+ "in a non-mandatory locking environment, "
+ "continuing");
+ else
+ gf_log(this->name, GF_LOG_DEBUG,
+ "Lock flags received, "
+ "continuing");
+ }
+
+ if ((flock->l_start < 0) || ((flock->l_start + flock->l_len) < 0)) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto unwind;
+ }
+
+ /* As per 'man 3 fcntl', the value of l_len may be
+ * negative. In such cases, lock request should be
+ * considered for the range starting at 'l_start+l_len'
+ * and ending at 'l_start-1'. Update the fields accordingly.
+ */
+ if (flock->l_len < 0) {
+ flock->l_start += flock->l_len;
+ flock->l_len = labs(flock->l_len);
+ }
+
+ local = mem_get0(this->local_pool);
+ if (!local) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ } else {
+ frame->local = local;
+ local->fd = fd_ref(fd);
+ }
- reqlock = new_posix_lock (flock, frame->root->client, frame->root->pid,
- &frame->root->lk_owner, fd, lk_flags,
- can_block);
+ pl_inode = pl_inode_get(this, fd->inode, local);
+ if (!pl_inode) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
- if (!reqlock) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
+ reqlock = new_posix_lock(flock, frame->root->client, frame->root->pid,
+ &frame->root->lk_owner, fd, lk_flags, can_block,
+ &op_errno);
- pl_trace_in (this, frame, fd, NULL, cmd, flock, NULL);
+ if (!reqlock) {
+ op_ret = -1;
+ goto unwind;
+ }
- switch (cmd) {
+ pl_trace_in(this, frame, fd, NULL, cmd, flock, NULL);
+ switch (cmd) {
case F_RESLK_LCKW:
- can_block = 1;
+ can_block = 1;
- /* fall through */
+ /* fall through */
case F_RESLK_LCK:
- reqlock->frame = frame;
- reqlock->this = this;
+ reqlock->frame = frame;
+ reqlock->this = this;
- ret = pl_reserve_setlk (this, pl_inode, reqlock,
- can_block);
- if (ret < 0) {
- if (can_block)
- goto out;
+ ret = pl_reserve_setlk(this, pl_inode, reqlock, can_block);
+ if (ret < 0) {
+ if (can_block)
+ goto out;
- op_ret = -1;
- op_errno = -ret;
- __destroy_lock (reqlock);
- goto unwind;
- }
- /* Finally a getlk and return the call */
- conf = pl_getlk (pl_inode, reqlock);
- if (conf)
- posix_lock_to_flock (conf, flock);
- break;
+ op_ret = -1;
+ op_errno = -ret;
+ __destroy_lock(reqlock);
+ goto unwind;
+ }
+ /* Finally a getlk and return the call */
+ conf = pl_getlk(pl_inode, reqlock);
+ if (conf)
+ posix_lock_to_flock(conf, flock);
+ break;
case F_RESLK_UNLCK:
- reqlock->frame = frame;
- reqlock->this = this;
- ret = pl_reserve_unlock (this, pl_inode, reqlock);
- if (ret < 0) {
- op_ret = -1;
- op_errno = -ret;
- }
- __destroy_lock (reqlock);
- goto unwind;
+ reqlock->frame = frame;
+ reqlock->this = this;
+ ret = pl_reserve_unlock(this, pl_inode, reqlock);
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = -ret;
+ }
+ __destroy_lock(reqlock);
+ goto unwind;
- break;
+ break;
case F_GETLK_FD:
- reqlock->frame = frame;
- reqlock->this = this;
- ret = pl_verify_reservelk (this, pl_inode, reqlock, can_block);
- GF_ASSERT (ret >= 0);
-
- ret = pl_getlk_fd (this, pl_inode, fd, reqlock);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "getting locks on fd failed");
- op_ret = -1;
- op_errno = ENOLCK;
- goto unwind;
- }
+ reqlock->frame = frame;
+ reqlock->this = this;
+ ret = pl_verify_reservelk(this, pl_inode, reqlock, can_block);
+ GF_ASSERT(ret >= 0);
+
+ ret = pl_getlk_fd(this, pl_inode, fd, reqlock);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_DEBUG, "getting locks on fd failed");
+ op_ret = -1;
+ op_errno = ENOLCK;
+ goto unwind;
+ }
- gf_log (this->name, GF_LOG_TRACE,
- "Replying with a lock on fd for healing");
+ gf_log(this->name, GF_LOG_TRACE,
+ "Replying with a lock on fd for healing");
- posix_lock_to_flock (reqlock, flock);
- __destroy_lock (reqlock);
+ posix_lock_to_flock(reqlock, flock);
+ __destroy_lock(reqlock);
- break;
+ break;
#if F_GETLK != F_GETLK64
case F_GETLK64:
#endif
case F_GETLK:
- conf = pl_getlk (pl_inode, reqlock);
- posix_lock_to_flock (conf, flock);
- __destroy_lock (reqlock);
+ conf = pl_getlk(pl_inode, reqlock);
+ posix_lock_to_flock(conf, flock);
+ __destroy_lock(reqlock);
- break;
+ break;
#if F_SETLKW != F_SETLKW64
case F_SETLKW64:
#endif
case F_SETLKW:
- can_block = 1;
- reqlock->frame = frame;
- reqlock->this = this;
- reqlock->blocking = can_block;
- /* fall through */
+ can_block = 1;
+ reqlock->frame = frame;
+ reqlock->this = this;
+ reqlock->blocking = can_block;
+ /* fall through */
#if F_SETLK != F_SETLK64
case F_SETLK64:
#endif
case F_SETLK:
- reqlock->frame = frame;
- reqlock->this = this;
+ reqlock->frame = frame;
+ reqlock->this = this;
+ lock_type = flock->l_type;
- pthread_mutex_lock (&pl_inode->mutex);
- {
- if (pl_inode->migrated) {
- op_errno = EREMOTE;
- pthread_mutex_unlock (&pl_inode->mutex);
- STACK_UNWIND_STRICT (lk, frame, -1,
- op_errno, flock, xdata);
-
- __destroy_lock (reqlock);
- goto out;
- }
- }
- pthread_mutex_unlock (&pl_inode->mutex);
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ if (pl_inode->migrated) {
+ op_errno = EREMOTE;
+ pthread_mutex_unlock(&pl_inode->mutex);
+ STACK_UNWIND_STRICT(lk, frame, -1, op_errno, flock, xdata);
- ret = pl_verify_reservelk (this, pl_inode, reqlock, can_block);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_TRACE,
- "Lock blocked due to conflicting reserve lock");
- goto out;
+ __destroy_lock(reqlock);
+ goto out;
}
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
- ret = pl_setlk (this, pl_inode, reqlock, can_block);
+ ret = pl_verify_reservelk(this, pl_inode, reqlock, can_block);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "Lock blocked due to conflicting reserve lock");
+ goto out;
+ }
+
+ if (reqlock->fl_type != F_UNLCK && pl_inode->mlock_enforced) {
+ ret = pl_lock_preempt(pl_inode, reqlock);
if (ret == -1) {
- if ((can_block) && (F_UNLCK != flock->l_type)) {
- pl_trace_block (this, frame, fd, NULL, cmd, flock, NULL);
- goto out;
- }
- gf_log (this->name, GF_LOG_DEBUG, "returning EAGAIN");
- op_ret = -1;
- op_errno = EAGAIN;
- __destroy_lock (reqlock);
- } else if (ret == -2) {
- goto out;
- } else if ((0 == ret) && (F_UNLCK == flock->l_type)) {
- /* For NLM's last "unlock on fd" detection */
- if (pl_locks_by_fd (pl_inode, fd))
- flock->l_type = F_RDLCK;
- else
- flock->l_type = F_UNLCK;
+ gf_log(this->name, GF_LOG_ERROR, "lock preempt failed");
+ op_ret = -1;
+ op_errno = EAGAIN;
+ __destroy_lock(reqlock);
+ goto out;
}
- }
-unwind:
- pl_trace_out (this, frame, fd, NULL, cmd, flock, op_ret, op_errno, NULL);
- pl_update_refkeeper (this, fd->inode);
+ pl_trace_block(this, frame, fd, NULL, cmd, flock, NULL);
+ goto unwind;
+ }
+
+ ret = pl_setlk(this, pl_inode, reqlock, can_block);
+ if (ret == -1) {
+ if ((can_block) && (F_UNLCK != lock_type)) {
+ goto out;
+ }
+ gf_log(this->name, GF_LOG_DEBUG, "returning EAGAIN");
+ op_ret = -1;
+ op_errno = EAGAIN;
+ __destroy_lock(reqlock);
+ } else if (ret == -2) {
+ goto out;
+ } else if ((0 == ret) && (F_UNLCK == flock->l_type)) {
+ /* For NLM's last "unlock on fd" detection */
+ if (pl_locks_by_fd(pl_inode, fd))
+ flock->l_type = F_RDLCK;
+ else
+ flock->l_type = F_UNLCK;
+ }
+ }
+unwind:
+ pl_trace_out(this, frame, fd, NULL, cmd, flock, op_ret, op_errno, NULL);
+ pl_update_refkeeper(this, fd->inode);
- STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, flock, xdata);
+ PL_STACK_UNWIND(lk, xdata, frame, op_ret, op_errno, flock, xdata);
out:
- return 0;
+ return 0;
}
-
/* TODO: this function just logs, no action required?? */
int
-pl_forget (xlator_t *this,
- inode_t *inode)
+pl_forget(xlator_t *this, inode_t *inode)
{
- pl_inode_t *pl_inode = NULL;
-
- posix_lock_t *ext_tmp = NULL;
- posix_lock_t *ext_l = NULL;
- struct list_head posixlks_released;
+ pl_inode_t *pl_inode = NULL;
- pl_inode_lock_t *ino_tmp = NULL;
- pl_inode_lock_t *ino_l = NULL;
- struct list_head inodelks_released;
+ posix_lock_t *ext_tmp = NULL;
+ posix_lock_t *ext_l = NULL;
+ struct list_head posixlks_released;
- pl_rw_req_t *rw_tmp = NULL;
- pl_rw_req_t *rw_req = NULL;
+ pl_inode_lock_t *ino_tmp = NULL;
+ pl_inode_lock_t *ino_l = NULL;
+ struct list_head inodelks_released;
- pl_entry_lock_t *entry_tmp = NULL;
- pl_entry_lock_t *entry_l = NULL;
- struct list_head entrylks_released;
+ pl_rw_req_t *rw_tmp = NULL;
+ pl_rw_req_t *rw_req = NULL;
- pl_dom_list_t *dom = NULL;
- pl_dom_list_t *dom_tmp = NULL;
+ pl_entry_lock_t *entry_tmp = NULL;
+ pl_entry_lock_t *entry_l = NULL;
+ struct list_head entrylks_released;
- INIT_LIST_HEAD (&posixlks_released);
- INIT_LIST_HEAD (&inodelks_released);
- INIT_LIST_HEAD (&entrylks_released);
+ pl_dom_list_t *dom = NULL;
+ pl_dom_list_t *dom_tmp = NULL;
- pl_inode = pl_inode_get (this, inode);
+ INIT_LIST_HEAD(&posixlks_released);
+ INIT_LIST_HEAD(&inodelks_released);
+ INIT_LIST_HEAD(&entrylks_released);
- pthread_mutex_lock (&pl_inode->mutex);
- {
+ pl_inode = pl_inode_get(this, inode, NULL);
+ if (!pl_inode)
+ return 0;
- if (!list_empty (&pl_inode->rw_list)) {
- gf_log (this->name, GF_LOG_WARNING,
- "Pending R/W requests found, releasing.");
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ if (!list_empty(&pl_inode->rw_list)) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "Pending R/W requests found, releasing.");
+
+ list_for_each_entry_safe(rw_req, rw_tmp, &pl_inode->rw_list, list)
+ {
+ list_del(&rw_req->list);
+ call_stub_destroy(rw_req->stub);
+ GF_FREE(rw_req);
+ }
+ }
+
+ if (!list_empty(&pl_inode->ext_list)) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "Pending fcntl locks found, releasing.");
+
+ list_for_each_entry_safe(ext_l, ext_tmp, &pl_inode->ext_list, list)
+ {
+ __delete_lock(ext_l);
+ if (ext_l->blocked) {
+ list_add_tail(&ext_l->list, &posixlks_released);
+ continue;
+ }
+ __destroy_lock(ext_l);
+ }
+ }
- list_for_each_entry_safe (rw_req, rw_tmp, &pl_inode->rw_list,
- list) {
+ list_for_each_entry_safe(dom, dom_tmp, &pl_inode->dom_list, inode_list)
+ {
+ if (!list_empty(&dom->inodelk_list)) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "Pending inode locks found, releasing.");
- list_del (&rw_req->list);
- call_stub_destroy(rw_req->stub);
- GF_FREE (rw_req);
- }
+ list_for_each_entry_safe(ino_l, ino_tmp, &dom->inodelk_list,
+ list)
+ {
+ __delete_inode_lock(ino_l);
+ __pl_inodelk_unref(ino_l);
}
- if (!list_empty (&pl_inode->ext_list)) {
- gf_log (this->name, GF_LOG_WARNING,
- "Pending fcntl locks found, releasing.");
+ list_splice_init(&dom->blocked_inodelks, &inodelks_released);
+ }
+ if (!list_empty(&dom->entrylk_list)) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "Pending entry locks found, releasing.");
- list_for_each_entry_safe (ext_l, ext_tmp, &pl_inode->ext_list,
- list) {
+ list_for_each_entry_safe(entry_l, entry_tmp, &dom->entrylk_list,
+ domain_list)
+ {
+ list_del_init(&entry_l->domain_list);
- __delete_lock (ext_l);
- if (ext_l->blocked) {
- list_add_tail (&ext_l->list, &posixlks_released);
- continue;
- }
- __destroy_lock (ext_l);
- }
+ GF_FREE((char *)entry_l->basename);
+ GF_FREE(entry_l->connection_id);
+ GF_FREE(entry_l);
}
+ list_splice_init(&dom->blocked_entrylks, &entrylks_released);
+ }
- list_for_each_entry_safe (dom, dom_tmp, &pl_inode->dom_list, inode_list) {
+ list_del(&dom->inode_list);
+ gf_log("posix-locks", GF_LOG_TRACE, " Cleaning up domain: %s",
+ dom->domain);
+ GF_FREE((char *)(dom->domain));
+ GF_FREE(dom);
+ }
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
- if (!list_empty (&dom->inodelk_list)) {
- gf_log (this->name, GF_LOG_WARNING,
- "Pending inode locks found, releasing.");
+ if (!list_empty(&posixlks_released)) {
+ list_for_each_entry_safe(ext_l, ext_tmp, &posixlks_released, list)
+ {
+ STACK_UNWIND_STRICT(lk, ext_l->frame, -1, 0, &ext_l->user_flock,
+ NULL);
+ __destroy_lock(ext_l);
+ }
+ }
- list_for_each_entry_safe (ino_l, ino_tmp, &dom->inodelk_list, list) {
- __delete_inode_lock (ino_l);
- __pl_inodelk_unref (ino_l);
- }
+ if (!list_empty(&inodelks_released)) {
+ list_for_each_entry_safe(ino_l, ino_tmp, &inodelks_released,
+ blocked_locks)
+ {
+ STACK_UNWIND_STRICT(inodelk, ino_l->frame, -1, 0, NULL);
+ __pl_inodelk_unref(ino_l);
+ }
+ }
- list_splice_init (&dom->blocked_inodelks, &inodelks_released);
+ if (!list_empty(&entrylks_released)) {
+ list_for_each_entry_safe(entry_l, entry_tmp, &entrylks_released,
+ blocked_locks)
+ {
+ STACK_UNWIND_STRICT(entrylk, entry_l->frame, -1, 0, NULL);
+ GF_FREE((char *)entry_l->basename);
+ GF_FREE(entry_l->connection_id);
+ GF_FREE(entry_l);
+ }
+ }
+ pthread_mutex_destroy(&pl_inode->mutex);
- }
- if (!list_empty (&dom->entrylk_list)) {
- gf_log (this->name, GF_LOG_WARNING,
- "Pending entry locks found, releasing.");
+ GF_FREE(pl_inode);
- list_for_each_entry_safe (entry_l, entry_tmp, &dom->entrylk_list, domain_list) {
- list_del_init (&entry_l->domain_list);
+ return 0;
+}
- GF_FREE ((char *)entry_l->basename);
- GF_FREE (entry_l->connection_id);
- GF_FREE (entry_l);
- }
+int
+pl_release(xlator_t *this, fd_t *fd)
+{
+ pl_inode_t *pl_inode = NULL;
+ uint64_t tmp_pl_inode = 0;
+ int ret = -1;
+ uint64_t tmp = 0;
+ pl_fdctx_t *fdctx = NULL;
- list_splice_init (&dom->blocked_entrylks, &entrylks_released);
- }
+ if (fd == NULL) {
+ goto out;
+ }
- list_del (&dom->inode_list);
- gf_log ("posix-locks", GF_LOG_TRACE,
- " Cleaning up domain: %s", dom->domain);
- GF_FREE ((char *)(dom->domain));
- GF_FREE (dom);
- }
+ ret = inode_ctx_get(fd->inode, this, &tmp_pl_inode);
+ if (ret != 0)
+ goto clean;
- }
- pthread_mutex_unlock (&pl_inode->mutex);
+ pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
- list_for_each_entry_safe (ext_l, ext_tmp, &posixlks_released, list) {
+ pl_trace_release(this, fd);
- STACK_UNWIND_STRICT (lk, ext_l->frame, -1, 0,
- &ext_l->user_flock, NULL);
- __destroy_lock (ext_l);
- }
+ gf_log(this->name, GF_LOG_TRACE, "Releasing all locks with fd %p", fd);
- list_for_each_entry_safe (ino_l, ino_tmp, &inodelks_released, blocked_locks) {
+ delete_locks_of_fd(this, pl_inode, fd);
+ pl_update_refkeeper(this, fd->inode);
- STACK_UNWIND_STRICT (inodelk, ino_l->frame, -1, 0, NULL);
- __pl_inodelk_unref (ino_l);
- }
+clean:
+ ret = fd_ctx_del(fd, this, &tmp);
+ if (ret) {
+ gf_log(this->name, GF_LOG_DEBUG, "Could not get fdctx");
+ goto out;
+ }
- list_for_each_entry_safe (entry_l, entry_tmp, &entrylks_released, blocked_locks) {
+ fdctx = (pl_fdctx_t *)(long)tmp;
- STACK_UNWIND_STRICT (entrylk, entry_l->frame, -1, 0, NULL);
- GF_FREE ((char *)entry_l->basename);
- GF_FREE (entry_l->connection_id);
- GF_FREE (entry_l);
+ GF_FREE(fdctx);
+out:
+ return ret;
+}
- }
+int
+pl_releasedir(xlator_t *this, fd_t *fd)
+{
+ int ret = -1;
+ uint64_t tmp = 0;
+ pl_fdctx_t *fdctx = NULL;
- pthread_mutex_destroy(&pl_inode->mutex);
+ if (fd == NULL) {
+ goto out;
+ }
- GF_FREE (pl_inode);
+ ret = fd_ctx_del(fd, this, &tmp);
+ if (ret) {
+ gf_log(this->name, GF_LOG_DEBUG, "Could not get fdctx");
+ goto out;
+ }
- return 0;
+ fdctx = (pl_fdctx_t *)(long)tmp;
+
+ GF_FREE(fdctx);
+out:
+ return ret;
}
-int
-pl_release (xlator_t *this, fd_t *fd)
+static int32_t
+pl_request_link_count(dict_t **pxdata)
{
- pl_inode_t *pl_inode = NULL;
- uint64_t tmp_pl_inode = 0;
- int ret = -1;
- uint64_t tmp = 0;
- pl_fdctx_t *fdctx = NULL;
+ dict_t *xdata;
- if (fd == NULL) {
- goto out;
+ xdata = *pxdata;
+ if (xdata == NULL) {
+ xdata = dict_new();
+ if (xdata == NULL) {
+ return ENOMEM;
}
+ } else {
+ dict_ref(xdata);
+ }
- ret = inode_ctx_get (fd->inode, this, &tmp_pl_inode);
- if (ret != 0)
- goto clean;
+ if (dict_set_uint32(xdata, GET_LINK_COUNT, 0) != 0) {
+ dict_unref(xdata);
+ return ENOMEM;
+ }
- pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
+ *pxdata = xdata;
- pl_trace_release (this, fd);
+ return 0;
+}
- gf_log (this->name, GF_LOG_TRACE,
- "Releasing all locks with fd %p", fd);
+static int32_t
+pl_check_link_count(dict_t *xdata)
+{
+ int32_t count;
- delete_locks_of_fd (this, pl_inode, fd);
- pl_update_refkeeper (this, fd->inode);
+ /* In case we are unable to read the link count from xdata, we take a
+ * conservative approach and return -2, which will prevent the inode from
+ * being considered deleted. In fact it will cause link tracking for this
+ * inode to be disabled completely to avoid races. */
-clean:
- ret = fd_ctx_del (fd, this, &tmp);
- if (ret) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Could not get fdctx");
- goto out;
- }
+ if (xdata == NULL) {
+ return -2;
+ }
- fdctx = (pl_fdctx_t *)(long)tmp;
+ if (dict_get_int32(xdata, GET_LINK_COUNT, &count) != 0) {
+ return -2;
+ }
- GF_FREE (fdctx);
-out:
- return ret;
+ return count;
}
-int
-pl_releasedir (xlator_t *this, fd_t *fd)
+int32_t
+pl_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode, struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
{
- int ret = -1;
- uint64_t tmp = 0;
- pl_fdctx_t *fdctx = NULL;
+ pl_inode_t *pl_inode;
- if (fd == NULL) {
- goto out;
+ if (op_ret >= 0) {
+ pl_inode = pl_inode_get(this, inode, NULL);
+ if (pl_inode == NULL) {
+ PL_STACK_UNWIND(lookup, xdata, frame, -1, ENOMEM, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
- ret = fd_ctx_del (fd, this, &tmp);
- if (ret) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Could not get fdctx");
- goto out;
+ pthread_mutex_lock(&pl_inode->mutex);
+
+ /* We only update the link count if we previously didn't know it.
+ * Doing it always can lead to races since lookup is not executed
+ * atomically most of the times. */
+ if (pl_inode->links == -2) {
+ pl_inode->links = pl_check_link_count(xdata);
+ if (buf->ia_type == IA_IFDIR) {
+ /* Directories have at least 2 links. To avoid special handling
+ * for directories, we simply decrement the value here to make
+ * them equivalent to regular files. */
+ pl_inode->links--;
+ }
}
- fdctx = (pl_fdctx_t *)(long)tmp;
+ pthread_mutex_unlock(&pl_inode->mutex);
+ }
- GF_FREE (fdctx);
-out:
- return ret;
+ PL_STACK_UNWIND(lookup, xdata, frame, op_ret, op_errno, inode, buf, xdata,
+ postparent);
+ return 0;
}
int32_t
-pl_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+pl_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- PL_STACK_UNWIND (lookup, xdata, frame, op_ret, op_errno, inode, buf,
- xdata, postparent);
- return 0;
-}
+ int32_t error;
-int32_t
-pl_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
-{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, loc, NULL);
- STACK_WIND (frame, pl_lookup_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, loc, xdata);
- return 0;
+ error = pl_request_link_count(&xdata);
+ if (error == 0) {
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, ((fd_t *)NULL), loc, NULL);
+ STACK_WIND(frame, pl_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, xdata);
+ dict_unref(xdata);
+ } else {
+ STACK_UNWIND_STRICT(lookup, frame, -1, error, NULL, NULL, NULL, NULL);
+ }
+ return 0;
}
int32_t
-pl_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
- int32_t op_errno, struct iatt *buf, dict_t *xdata)
+pl_fstat_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *buf, dict_t *xdata)
{
- PL_STACK_UNWIND (fstat, xdata, frame, op_ret, op_errno, buf, xdata);
- return 0;
+ PL_STACK_UNWIND(fstat, xdata, frame, op_ret, op_errno, buf, xdata);
+ return 0;
}
int32_t
-pl_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+pl_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, fd, NULL, NULL);
- STACK_WIND (frame, pl_fstat_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fstat, fd, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
+ STACK_WIND(frame, pl_fstat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fstat, fd, xdata);
+ return 0;
}
int
-pl_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, gf_dirent_t *entries, dict_t *xdata)
+pl_readdirp_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+ int op_errno, gf_dirent_t *entries, dict_t *xdata)
{
- pl_local_t *local = NULL;
- gf_dirent_t *entry = NULL;
+ pl_local_t *local = NULL;
+ gf_dirent_t *entry = NULL;
- if (op_ret <= 0)
- goto unwind;
+ if (op_ret <= 0)
+ goto unwind;
- local = frame->local;
- if (!local)
- goto unwind;
+ local = frame->local;
+ if (!local)
+ goto unwind;
- list_for_each_entry (entry, &entries->list, list) {
- pl_set_xdata_response (this, local, local->fd->inode,
- entry->inode, entry->d_name,
- entry->dict, 0);
- }
+ list_for_each_entry(entry, &entries->list, list)
+ {
+ pl_set_xdata_response(this, local, local->fd->inode, entry->inode,
+ entry->d_name, entry->dict, 0);
+ }
unwind:
- PL_STACK_UNWIND (readdirp, xdata, frame, op_ret, op_errno, entries,
- xdata);
+ PL_STACK_UNWIND(readdirp, xdata, frame, op_ret, op_errno, entries, xdata);
- return 0;
+ return 0;
}
int
-pl_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, dict_t *xdata)
+pl_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, fd, NULL, NULL);
- STACK_WIND (frame, pl_readdirp_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->readdirp,
- fd, size, offset, xdata);
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
+ STACK_WIND(frame, pl_readdirp_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdirp, fd, size, offset, xdata);
- return 0;
+ return 0;
}
lock_migration_info_t *
-gf_mig_info_for_lock (posix_lock_t *lock)
+gf_mig_info_for_lock(posix_lock_t *lock)
{
- lock_migration_info_t *new = NULL;
+ lock_migration_info_t *new = GF_MALLOC(sizeof(lock_migration_info_t),
+ gf_common_mt_lock_mig);
+ if (new == NULL) {
+ goto out;
+ }
- new = GF_CALLOC (1, sizeof (lock_migration_info_t),
- gf_common_mt_lock_mig);
- if (new == NULL) {
- goto out;
- }
-
- INIT_LIST_HEAD (&new->list);
+ INIT_LIST_HEAD(&new->list);
- posix_lock_to_flock (lock, &new->flock);
+ posix_lock_to_flock(lock, &new->flock);
- new->lk_flags = lock->lk_flags;
+ new->lk_flags = lock->lk_flags;
- new->client_uid = gf_strdup (lock->client_uid);
+ new->client_uid = gf_strdup(lock->client_uid);
out:
- return new;
+ return new;
}
int
-pl_fill_active_locks (pl_inode_t *pl_inode, lock_migration_info_t *lmi)
+pl_fill_active_locks(pl_inode_t *pl_inode, lock_migration_info_t *lmi)
{
- posix_lock_t *temp = NULL;
- lock_migration_info_t *newlock = NULL;
- int count = 0;
+ posix_lock_t *temp = NULL;
+ lock_migration_info_t *newlock = NULL;
+ int count = 0;
- pthread_mutex_lock (&pl_inode->mutex);
- {
- if (list_empty (&pl_inode->ext_list)) {
- count = 0;
- goto out;
- }
-
- list_for_each_entry (temp, &pl_inode->ext_list, list) {
-
- if (temp->blocked)
- continue;
-
- newlock = gf_mig_info_for_lock (temp);
- if (!newlock) {
- gf_msg (THIS->name, GF_LOG_ERROR, 0, 0,
- "lock_dup failed");
- count = -1;
- goto out;
- }
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ if (list_empty(&pl_inode->ext_list)) {
+ count = 0;
+ goto unlock;
+ }
- list_add_tail (&newlock->list, &lmi->list);
- count++;
- }
+ list_for_each_entry(temp, &pl_inode->ext_list, list)
+ {
+ if (temp->blocked)
+ continue;
+
+ newlock = gf_mig_info_for_lock(temp);
+ if (!newlock) {
+ pthread_mutex_unlock(&pl_inode->mutex);
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, 0, "lock_dup failed");
+ count = -1;
+ goto out;
+ }
+ list_add_tail(&newlock->list, &lmi->list);
+ count++;
}
+ }
+unlock:
+ pthread_mutex_unlock(&pl_inode->mutex);
out:
- pthread_mutex_unlock (&pl_inode->mutex);
- return count;
+ return count;
}
/* This function reads only active locks */
static int
-pl_getactivelk (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+pl_getactivelk(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- pl_inode_t *pl_inode = NULL;
- lock_migration_info_t locks;
- int op_ret = 0;
- int op_errno = 0;
- int count = 0;
+ pl_inode_t *pl_inode = NULL;
+ lock_migration_info_t locks;
+ int op_ret = 0;
+ int op_errno = 0;
+ int count = 0;
- INIT_LIST_HEAD (&locks.list);
+ INIT_LIST_HEAD(&locks.list);
- pl_inode = pl_inode_get (this, loc->inode);
- if (!pl_inode) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0,
- "pl_inode_get failed");
+ pl_inode = pl_inode_get(this, loc->inode, NULL);
+ if (!pl_inode) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "pl_inode_get failed");
- op_ret = -1;
- op_errno = ENOMEM;
- goto out;
- }
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto out;
+ }
- count = pl_fill_active_locks (pl_inode, &locks);
+ count = pl_fill_active_locks(pl_inode, &locks);
- op_ret = count;
+ op_ret = count;
out:
- STACK_UNWIND_STRICT (getactivelk, frame, op_ret, op_errno, &locks,
- NULL);
+ STACK_UNWIND_STRICT(getactivelk, frame, op_ret, op_errno, &locks, NULL);
- gf_free_mig_locks (&locks);
+ gf_free_mig_locks(&locks);
- return 0;
+ return 0;
}
void
-pl_metalk_unref (pl_meta_lock_t *lock)
+pl_metalk_unref(pl_meta_lock_t *lock)
{
- lock->ref--;
- if (!lock->ref) {
- GF_FREE (lock->client_uid);
- GF_FREE (lock);
- }
+ lock->ref--;
+ if (!lock->ref) {
+ GF_FREE(lock->client_uid);
+ GF_FREE(lock);
+ }
}
-
void
-__pl_metalk_ref (pl_meta_lock_t *lock)
+__pl_metalk_ref(pl_meta_lock_t *lock)
{
- lock->ref++;
+ lock->ref++;
}
pl_meta_lock_t *
-new_meta_lock (call_frame_t *frame, xlator_t *this)
-{
- pl_meta_lock_t *lock = NULL;
-
- lock = GF_CALLOC (1, sizeof (*lock),
- gf_locks_mt_pl_meta_lock_t);
-
- if (!lock) {
- gf_msg (this->name, GF_LOG_ERROR, 0, ENOMEM, "mem allocation"
- " failed for meta lock");
- goto out;
- }
-
- INIT_LIST_HEAD (&lock->list);
- INIT_LIST_HEAD (&lock->client_list);
-
- lock->client_uid = gf_strdup (frame->root->client->client_uid);
- if (!lock->client_uid) {
- gf_msg (this->name, GF_LOG_ERROR, 0, ENOMEM, "mem allocation"
- " failed for client_uid");
- GF_FREE (lock);
- lock = NULL;
- goto out;
- }
-
- __pl_metalk_ref (lock);
+new_meta_lock(call_frame_t *frame, xlator_t *this)
+{
+ pl_meta_lock_t *lock = GF_CALLOC(1, sizeof(*lock),
+ gf_locks_mt_pl_meta_lock_t);
+
+ if (!lock) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, ENOMEM,
+ "mem allocation"
+ " failed for meta lock");
+ goto out;
+ }
+
+ INIT_LIST_HEAD(&lock->list);
+ INIT_LIST_HEAD(&lock->client_list);
+
+ lock->client_uid = gf_strdup(frame->root->client->client_uid);
+ if (!lock->client_uid) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, ENOMEM,
+ "mem allocation"
+ " failed for client_uid");
+ GF_FREE(lock);
+ lock = NULL;
+ goto out;
+ }
+
+ __pl_metalk_ref(lock);
out:
- return lock;
+ return lock;
}
int
-pl_insert_metalk (pl_inode_t *pl_inode, pl_ctx_t *ctx, pl_meta_lock_t *lock)
+pl_insert_metalk(pl_inode_t *pl_inode, pl_ctx_t *ctx, pl_meta_lock_t *lock)
{
- int ret = 0;
+ int ret = 0;
- if (!pl_inode || !ctx || !lock) {
- gf_msg (THIS->name, GF_LOG_INFO, 0, 0, "NULL parameter");
- ret = -1;
- goto out;
- }
+ if (!pl_inode || !ctx || !lock) {
+ gf_msg(THIS->name, GF_LOG_INFO, 0, 0, "NULL parameter");
+ ret = -1;
+ goto out;
+ }
- lock->pl_inode = pl_inode;
+ lock->pl_inode = pl_inode;
- /* refer function pl_inode_setlk for more info for this ref.
- * This should be unrefed on meta-unlock triggered by rebalance or
- * in cleanup with client disconnect*/
- /*TODO: unref this in cleanup code for disconnect and meta-unlock*/
- pl_inode->inode = inode_ref (pl_inode->inode);
+ /* refer function pl_inode_setlk for more info for this ref.
+ * This should be unrefed on meta-unlock triggered by rebalance or
+ * in cleanup with client disconnect*/
+ /*TODO: unref this in cleanup code for disconnect and meta-unlock*/
+ pl_inode->inode = inode_ref(pl_inode->inode);
- /* NOTE:In case of a client-server disconnect we need to cleanup metalk.
- * Hence, adding the metalk to pl_ctx_t as well. The mutex lock order
- * should always be on ctx and then on pl_inode*/
+ /* NOTE:In case of a client-server disconnect we need to cleanup metalk.
+ * Hence, adding the metalk to pl_ctx_t as well. The mutex lock order
+ * should always be on ctx and then on pl_inode*/
- pthread_mutex_lock (&ctx->lock);
+ pthread_mutex_lock(&ctx->lock);
+ {
+ pthread_mutex_lock(&pl_inode->mutex);
{
- pthread_mutex_lock (&pl_inode->mutex);
- {
- list_add_tail (&lock->list, &pl_inode->metalk_list);
- }
- pthread_mutex_unlock (&pl_inode->mutex);
-
- list_add_tail (&lock->client_list, &ctx->metalk_list);
+ list_add_tail(&lock->list, &pl_inode->metalk_list);
}
- pthread_mutex_unlock (&ctx->lock);
+ pthread_mutex_unlock(&pl_inode->mutex);
+
+ list_add_tail(&lock->client_list, &ctx->metalk_list);
+ }
+ pthread_mutex_unlock(&ctx->lock);
out:
- return ret;
+ return ret;
}
int32_t
-pl_metalk (call_frame_t *frame, xlator_t *this, inode_t *inode)
-{
- pl_inode_t *pl_inode = NULL;
- int ret = 0;
- pl_meta_lock_t *reqlk = NULL;
- pl_ctx_t *ctx = NULL;
-
- pl_inode = pl_inode_get (this, inode);
- if (!pl_inode) {
- gf_msg (this->name, GF_LOG_ERROR, 0, ENOMEM,
- "pl_inode mem allocation failedd");
-
- ret = -1;
- goto out;
- }
-
- if (frame->root->client) {
- ctx = pl_ctx_get (frame->root->client, this);
- if (!ctx) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0,
- "pl_ctx_get failed");
-
- ret = -1;
- goto out;
-
- }
- } else {
- gf_msg (this->name, GF_LOG_INFO, 0, 0, "frame-root-client "
- "is NULL");
-
- ret = -1;
- goto out;
- }
-
- reqlk = new_meta_lock (frame, this);
- if (!reqlk) {
- ret = -1;
- goto out;
- }
-
- ret = pl_insert_metalk (pl_inode, ctx, reqlk);
- if (ret < 0) {
- pl_metalk_unref (reqlk);
- }
+pl_metalk(call_frame_t *frame, xlator_t *this, inode_t *inode)
+{
+ pl_inode_t *pl_inode = NULL;
+ int ret = 0;
+ pl_meta_lock_t *reqlk = NULL;
+ pl_ctx_t *ctx = NULL;
+
+ pl_inode = pl_inode_get(this, inode, NULL);
+ if (!pl_inode) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, ENOMEM,
+ "pl_inode mem allocation failedd");
+
+ ret = -1;
+ goto out;
+ }
+
+ /* Non rebalance process trying to do metalock */
+ if (frame->root->pid != GF_CLIENT_PID_DEFRAG) {
+ ret = -1;
+ goto out;
+ }
+
+ /* Note: In the current scheme of glusterfs where lock migration is
+ * experimental, (ideally) the rebalance process which is migrating
+ * the file should request for a metalock. Hence, the metalock count
+ * should not be more than one for an inode. In future, if there is a
+ * need for meta-lock from other clients, the following block can be
+ * removed.
+ *
+ * Since pl_metalk is called as part of setxattr operation, any client
+ * process(non-rebalance) residing outside trusted network can exhaust
+ * memory of the server node by issuing setxattr repetitively on the
+ * metalock key. The following code makes sure that more than
+ * one metalock cannot be granted on an inode*/
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ if (pl_metalock_is_active(pl_inode)) {
+ ret = -1;
+ }
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+
+ if (ret == -1) {
+ gf_msg(this->name, GF_LOG_WARNING, EINVAL, 0,
+ "More than one meta-lock cannot be granted on"
+ " the inode");
+ goto out;
+ }
+
+ if (frame->root->client) {
+ ctx = pl_ctx_get(frame->root->client, this);
+ if (!ctx) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "pl_ctx_get failed");
+
+ ret = -1;
+ goto out;
+ }
+ } else {
+ gf_msg(this->name, GF_LOG_INFO, 0, 0,
+ "frame-root-client "
+ "is NULL");
+
+ ret = -1;
+ goto out;
+ }
+
+ reqlk = new_meta_lock(frame, this);
+ if (!reqlk) {
+ ret = -1;
+ goto out;
+ }
+
+ ret = pl_insert_metalk(pl_inode, ctx, reqlk);
+ if (ret < 0) {
+ pl_metalk_unref(reqlk);
+ }
out:
- return ret;
+ return ret;
}
-void
-__unwind_queued_locks (xlator_t *this, pl_inode_t *pl_inode,
- struct list_head *tmp_list)
+static void
+__unwind_queued_locks(pl_inode_t *pl_inode, struct list_head *tmp_list)
{
- if (list_empty (&pl_inode->queued_locks))
- return;
+ if (list_empty(&pl_inode->queued_locks))
+ return;
- list_splice_init (&pl_inode->queued_locks, tmp_list);
+ list_splice_init(&pl_inode->queued_locks, tmp_list);
}
-void
-__unwind_blocked_locks (xlator_t *this, pl_inode_t *pl_inode,
- struct list_head *tmp_list)
+static void
+__unwind_blocked_locks(pl_inode_t *pl_inode, struct list_head *tmp_list)
{
- posix_lock_t *lock = NULL;
- posix_lock_t *tmp = NULL;
+ posix_lock_t *lock = NULL;
+ posix_lock_t *tmp = NULL;
- if (list_empty (&pl_inode->ext_list))
- return;
-
- list_for_each_entry_safe (lock, tmp, &pl_inode->ext_list, list) {
+ if (list_empty(&pl_inode->ext_list))
+ return;
- if (!lock->blocking)
- continue;
+ list_for_each_entry_safe(lock, tmp, &pl_inode->ext_list, list)
+ {
+ if (!lock->blocking)
+ continue;
- list_del_init (&lock->list);
- list_add_tail (&lock->list, tmp_list);
- }
+ list_del_init(&lock->list);
+ list_add_tail(&lock->list, tmp_list);
+ }
}
int
-pl_metaunlock (call_frame_t *frame, xlator_t *this, inode_t *inode,
- dict_t *dict)
-{
- pl_inode_t *pl_inode = NULL;
- int ret = 0;
- pl_meta_lock_t *meta_lock = NULL;
- pl_meta_lock_t *tmp_metalk = NULL;
- pl_ctx_t *ctx = NULL;
- posix_lock_t *posix_lock = NULL;
- posix_lock_t *tmp_posixlk = NULL;
- struct list_head tmp_posixlk_list;
-
- INIT_LIST_HEAD (&tmp_posixlk_list);
-
- if (frame->root->client) {
- ctx = pl_ctx_get (frame->root->client, this);
- if (!ctx) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0,
- "pl_ctx_get failed");
-
- ret = -1;
- goto out;
- }
- } else {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0, "frame-root-client is "
- "NULL");
- ret = -1;
- goto out;
- }
-
- pl_inode = pl_inode_get (this, inode);
- if (!pl_inode) {
- ret = -1;
- goto out;
- }
-
- pthread_mutex_lock (&ctx->lock);
+pl_metaunlock(call_frame_t *frame, xlator_t *this, inode_t *inode, dict_t *dict)
+{
+ pl_inode_t *pl_inode = NULL;
+ int ret = 0;
+ pl_meta_lock_t *meta_lock = NULL;
+ pl_meta_lock_t *tmp_metalk = NULL;
+ pl_ctx_t *ctx = NULL;
+ posix_lock_t *posix_lock = NULL;
+ posix_lock_t *tmp_posixlk = NULL;
+ struct list_head tmp_posixlk_list;
+
+ INIT_LIST_HEAD(&tmp_posixlk_list);
+
+ if (frame->root->client) {
+ ctx = pl_ctx_get(frame->root->client, this);
+ if (!ctx) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "pl_ctx_get failed");
+
+ ret = -1;
+ goto out;
+ }
+ } else {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0,
+ "frame-root-client is "
+ "NULL");
+ ret = -1;
+ goto out;
+ }
+
+ pl_inode = pl_inode_get(this, inode, NULL);
+ if (!pl_inode) {
+ ret = -1;
+ goto out;
+ }
+
+ pthread_mutex_lock(&ctx->lock);
+ {
+ pthread_mutex_lock(&pl_inode->mutex);
{
- pthread_mutex_lock (&pl_inode->mutex);
- {
- /* Unwind queued locks regardless of migration status */
- __unwind_queued_locks (this, pl_inode,
- &tmp_posixlk_list);
-
- /* Unwind blocked locks only for successful migration */
- if (dict_get (dict, "status")) {
-
- /* unwind all blocked locks */
- __unwind_blocked_locks (this, pl_inode,
- &tmp_posixlk_list);
- }
-
- /* unlock metalk */
- /* if this list is empty then pl_inode->metalk_list
- * should be empty too. meta lock should in all cases
- * be added/removed from both pl_ctx_t and pl_inode */
+ /* Unwind queued locks regardless of migration status */
+ __unwind_queued_locks(pl_inode, &tmp_posixlk_list);
- if (list_empty (&ctx->metalk_list))
- goto unlock;
+ /* Unwind blocked locks only for successful migration */
+ if (dict_get_sizen(dict, "status")) {
+ /* unwind all blocked locks */
+ __unwind_blocked_locks(pl_inode, &tmp_posixlk_list);
+ }
- list_for_each_entry_safe (meta_lock, tmp_metalk,
- &ctx->metalk_list,
- client_list) {
- list_del_init (&meta_lock->client_list);
+ /* unlock metalk */
+ /* if this list is empty then pl_inode->metalk_list
+ * should be empty too. meta lock should in all cases
+ * be added/removed from both pl_ctx_t and pl_inode */
- pl_inode = meta_lock->pl_inode;
+ if (list_empty(&ctx->metalk_list))
+ goto unlock;
- list_del_init (&meta_lock->list);
+ list_for_each_entry_safe(meta_lock, tmp_metalk, &ctx->metalk_list,
+ client_list)
+ {
+ list_del_init(&meta_lock->client_list);
- pl_metalk_unref (meta_lock);
+ pl_inode = meta_lock->pl_inode;
- /* The corresponding ref is taken in
- * pl_insert_metalk*/
- inode_unref (pl_inode->inode);
- }
+ list_del_init(&meta_lock->list);
- if (dict_get (dict, "status"))
- pl_inode->migrated = _gf_true;
- else
- pl_inode->migrated = _gf_false;
- }
-unlock:
+ pl_metalk_unref(meta_lock);
- pthread_mutex_unlock (&pl_inode->mutex);
+ /* The corresponding ref is taken in
+ * pl_insert_metalk*/
+ inode_unref(pl_inode->inode);
+ }
+ if (dict_get_sizen(dict, "status"))
+ pl_inode->migrated = _gf_true;
+ else
+ pl_inode->migrated = _gf_false;
}
- pthread_mutex_unlock (&ctx->lock);
+ unlock:
+
+ pthread_mutex_unlock(&pl_inode->mutex);
+ }
+ pthread_mutex_unlock(&ctx->lock);
out:
- list_for_each_entry_safe (posix_lock, tmp_posixlk, &tmp_posixlk_list,
- list) {
- list_del_init (&posix_lock->list);
+ list_for_each_entry_safe(posix_lock, tmp_posixlk, &tmp_posixlk_list, list)
+ {
+ list_del_init(&posix_lock->list);
- STACK_UNWIND_STRICT (lk, posix_lock->frame, -1, EREMOTE,
- &posix_lock->user_flock, NULL);
+ STACK_UNWIND_STRICT(lk, posix_lock->frame, -1, EREMOTE,
+ &posix_lock->user_flock, NULL);
- __destroy_lock(posix_lock);
- }
+ __destroy_lock(posix_lock);
+ }
- return ret;
+ return ret;
}
int32_t
-pl_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- PL_STACK_UNWIND (setxattr, xdata, frame, op_ret, op_errno, xdata);
- return 0;
+pl_setxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+ pl_local_t *local = NULL;
+ pl_inode_t *pl_inode = NULL;
+ local = frame->local;
+ if (local && local->update_mlock_enforced_flag && op_ret != -1) {
+ pl_inode = pl_inode_get(this, local->inode, NULL);
+ if (!pl_inode) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ while (pl_inode->fop_wind_count > 0) {
+ gf_msg(this->name, GF_LOG_INFO, 0, 0,
+ "waiting for existing fops (count %d) to drain for "
+ "gfid %s",
+ pl_inode->fop_wind_count, uuid_utoa(pl_inode->gfid));
+ pthread_cond_wait(&pl_inode->check_fop_wind_count,
+ &pl_inode->mutex);
+ }
+ pl_inode->mlock_enforced = _gf_true;
+ pl_inode->check_mlock_info = _gf_false;
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+ }
+
+unwind:
+ PL_STACK_UNWIND_FOR_CLIENT(setxattr, xdata, frame, op_ret, op_errno, xdata);
+ return 0;
}
int32_t
-pl_setxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *dict, int flags, dict_t *xdata)
+pl_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int flags, dict_t *xdata)
{
- int op_ret = 0;
- int op_errno = 0;
- dict_t *xdata_rsp = NULL;
+ int op_ret = 0;
+ int op_errno = EINVAL;
+ dict_t *xdata_rsp = NULL;
+ char *name = NULL;
+ posix_locks_private_t *priv = this->private;
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, loc, NULL);
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, ((fd_t *)NULL), loc, NULL);
- if (dict_get (dict, GF_META_LOCK_KEY)) {
+ if (dict_get_sizen(dict, GF_META_LOCK_KEY)) {
+ op_ret = pl_metalk(frame, this, loc->inode);
- op_ret = pl_metalk (frame, this, loc->inode);
+ } else if (dict_get_sizen(dict, GF_META_UNLOCK_KEY)) {
+ op_ret = pl_metaunlock(frame, this, loc->inode, dict);
+ } else {
+ goto usual;
+ }
- } else if (dict_get (dict, GF_META_UNLOCK_KEY)) {
+ PL_STACK_UNWIND_FOR_CLIENT(setxattr, xdata_rsp, frame, op_ret, op_errno,
+ xdata_rsp);
+ return 0;
- op_ret = pl_metaunlock (frame, this, loc->inode, dict);
+usual:
+ PL_CHECK_LOCK_ENFORCE_KEY(frame, dict, name, this, loc, ((fd_t *)NULL),
+ priv);
- } else {
- goto usual;
- }
+ STACK_WIND(frame, pl_setxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, loc, dict, flags, xdata);
+ return 0;
- PL_STACK_UNWIND (setxattr, xdata_rsp, frame, op_ret, op_errno,
- xdata_rsp);
- return 0;
+unwind:
+ PL_STACK_UNWIND_FOR_CLIENT(setxattr, xdata, frame, op_ret, op_errno, xdata);
-usual:
- STACK_WIND (frame, pl_setxattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setxattr,
- loc, dict, flags, xdata);
- return 0;
+ return 0;
}
void
-pl_dump_lock (char *str, int size, struct gf_flock *flock,
- gf_lkowner_t *owner, void *trans, char *conn_id,
- time_t *granted_time, time_t *blkd_time, gf_boolean_t active)
-{
- char *type_str = NULL;
- char granted[256] = {0,};
- char blocked[256] = {0,};
-
- if (granted_time)
- gf_time_fmt (granted, sizeof (granted), *granted_time,
- gf_timefmt_FT);
- if (blkd_time)
- gf_time_fmt (blocked, sizeof (blocked), *blkd_time,
- gf_timefmt_FT);
- switch (flock->l_type) {
+pl_dump_lock(char *str, int size, struct gf_flock *flock, gf_lkowner_t *owner,
+ void *trans, char *conn_id, time_t *granted_time,
+ time_t *blkd_time, gf_boolean_t active)
+{
+ char *type_str = NULL;
+ char granted[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ char blocked[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+
+ if (granted_time)
+ gf_time_fmt(granted, sizeof(granted), *granted_time, gf_timefmt_FT);
+ if (blkd_time)
+ gf_time_fmt(blocked, sizeof(blocked), *blkd_time, gf_timefmt_FT);
+ switch (flock->l_type) {
case F_RDLCK:
- type_str = "READ";
- break;
+ type_str = "READ";
+ break;
case F_WRLCK:
- type_str = "WRITE";
- break;
+ type_str = "WRITE";
+ break;
case F_UNLCK:
- type_str = "UNLOCK";
- break;
+ type_str = "UNLOCK";
+ break;
default:
- type_str = "UNKNOWN";
- break;
- }
-
- if (active) {
- if (blkd_time && *blkd_time == 0) {
- snprintf (str, size, RANGE_GRNTD_FMT,
- type_str, flock->l_whence,
- (unsigned long long) flock->l_start,
- (unsigned long long) flock->l_len,
- (unsigned long long) flock->l_pid,
- lkowner_utoa (owner), trans, conn_id,
- granted);
- } else {
- snprintf (str, size, RANGE_BLKD_GRNTD_FMT,
- type_str, flock->l_whence,
- (unsigned long long) flock->l_start,
- (unsigned long long) flock->l_len,
- (unsigned long long) flock->l_pid,
- lkowner_utoa (owner), trans, conn_id,
- blocked, granted);
- }
+ type_str = "UNKNOWN";
+ break;
+ }
+
+ if (active) {
+ if (blkd_time && *blkd_time == 0) {
+ snprintf(str, size, RANGE_GRNTD_FMT, type_str, flock->l_whence,
+ (unsigned long long)flock->l_start,
+ (unsigned long long)flock->l_len,
+ (unsigned long long)flock->l_pid, lkowner_utoa(owner),
+ trans, conn_id, granted);
} else {
- snprintf (str, size, RANGE_BLKD_FMT,
- type_str, flock->l_whence,
- (unsigned long long) flock->l_start,
- (unsigned long long) flock->l_len,
- (unsigned long long) flock->l_pid,
- lkowner_utoa (owner), trans, conn_id, blocked);
+ snprintf(str, size, RANGE_BLKD_GRNTD_FMT, type_str, flock->l_whence,
+ (unsigned long long)flock->l_start,
+ (unsigned long long)flock->l_len,
+ (unsigned long long)flock->l_pid, lkowner_utoa(owner),
+ trans, conn_id, blocked, granted);
}
-
+ } else {
+ snprintf(str, size, RANGE_BLKD_FMT, type_str, flock->l_whence,
+ (unsigned long long)flock->l_start,
+ (unsigned long long)flock->l_len,
+ (unsigned long long)flock->l_pid, lkowner_utoa(owner), trans,
+ conn_id, blocked);
+ }
}
void
-__dump_entrylks (pl_inode_t *pl_inode)
-{
- pl_dom_list_t *dom = NULL;
- pl_entry_lock_t *lock = NULL;
- char blocked[256] = {0,};
- char granted[256] = {0,};
- int count = 0;
- char key[GF_DUMP_MAX_BUF_LEN] = {0,};
- char *k = "xlator.feature.locks.lock-dump.domain.entrylk";
-
- char tmp[4098];
-
- list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
-
- count = 0;
-
- gf_proc_dump_build_key(key,
- "lock-dump.domain",
- "domain");
- gf_proc_dump_write(key, "%s", dom->domain);
-
- list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
-
- gf_time_fmt (granted, sizeof (granted),
- lock->granted_time.tv_sec, gf_timefmt_FT);
- gf_proc_dump_build_key(key, k,
- "entrylk[%d](ACTIVE)", count );
- if (lock->blkd_time.tv_sec == 0) {
- snprintf (tmp, sizeof (tmp), ENTRY_GRNTD_FMT,
- lock->type == ENTRYLK_RDLCK ?
- "ENTRYLK_RDLCK" : "ENTRYLK_WRLCK",
- lock->basename,
- (unsigned long long) lock->client_pid,
- lkowner_utoa (&lock->owner),
- lock->client,
- lock->connection_id, granted);
- } else {
- gf_time_fmt (blocked, sizeof (blocked),
- lock->blkd_time.tv_sec,
- gf_timefmt_FT);
- snprintf (tmp, sizeof (tmp),
- ENTRY_BLKD_GRNTD_FMT,
- lock->type == ENTRYLK_RDLCK ?
- "ENTRYLK_RDLCK" : "ENTRYLK_WRLCK",
- lock->basename,
- (unsigned long long) lock->client_pid,
- lkowner_utoa (&lock->owner),
- lock->client,
- lock->connection_id,
- blocked, granted);
- }
-
- gf_proc_dump_write(key, tmp);
-
- count++;
- }
-
- list_for_each_entry (lock, &dom->blocked_entrylks,
- blocked_locks) {
-
- gf_time_fmt (blocked, sizeof (blocked),
- lock->blkd_time.tv_sec, gf_timefmt_FT);
-
- gf_proc_dump_build_key(key, k,
- "entrylk[%d](BLOCKED)", count );
- snprintf (tmp, sizeof (tmp), ENTRY_BLKD_FMT,
- lock->type == ENTRYLK_RDLCK ?
- "ENTRYLK_RDLCK" : "ENTRYLK_WRLCK",
- lock->basename,
- (unsigned long long) lock->client_pid,
- lkowner_utoa (&lock->owner), lock->client,
- lock->connection_id, blocked);
+__dump_entrylks(pl_inode_t *pl_inode)
+{
+ pl_dom_list_t *dom = NULL;
+ pl_entry_lock_t *lock = NULL;
+ char blocked[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ char granted[GF_TIMESTR_SIZE] = {
+ 0,
+ };
+ int count = 0;
+ char key[GF_DUMP_MAX_BUF_LEN] = {
+ 0,
+ };
+ char *k = "xlator.feature.locks.lock-dump.domain.entrylk";
+
+ char tmp[4098];
+
+ list_for_each_entry(dom, &pl_inode->dom_list, inode_list)
+ {
+ count = 0;
+
+ gf_proc_dump_build_key(key, "lock-dump.domain", "domain");
+ gf_proc_dump_write(key, "%s", dom->domain);
+
+ list_for_each_entry(lock, &dom->entrylk_list, domain_list)
+ {
+ gf_time_fmt(granted, sizeof(granted), lock->granted_time,
+ gf_timefmt_FT);
+ gf_proc_dump_build_key(key, k, "entrylk[%d](ACTIVE)", count);
+ if (lock->blkd_time == 0) {
+ snprintf(tmp, sizeof(tmp), ENTRY_GRNTD_FMT,
+ lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK"
+ : "ENTRYLK_WRLCK",
+ lock->basename, (unsigned long long)lock->client_pid,
+ lkowner_utoa(&lock->owner), lock->client,
+ lock->connection_id, granted);
+ } else {
+ gf_time_fmt(blocked, sizeof(blocked), lock->blkd_time,
+ gf_timefmt_FT);
+ snprintf(tmp, sizeof(tmp), ENTRY_BLKD_GRNTD_FMT,
+ lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK"
+ : "ENTRYLK_WRLCK",
+ lock->basename, (unsigned long long)lock->client_pid,
+ lkowner_utoa(&lock->owner), lock->client,
+ lock->connection_id, blocked, granted);
+ }
+
+ gf_proc_dump_write(key, "%s", tmp);
+
+ count++;
+ }
+
+ list_for_each_entry(lock, &dom->blocked_entrylks, blocked_locks)
+ {
+ gf_time_fmt(blocked, sizeof(blocked), lock->blkd_time,
+ gf_timefmt_FT);
- gf_proc_dump_write(key, tmp);
+ gf_proc_dump_build_key(key, k, "entrylk[%d](BLOCKED)", count);
+ snprintf(
+ tmp, sizeof(tmp), ENTRY_BLKD_FMT,
+ lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" : "ENTRYLK_WRLCK",
+ lock->basename, (unsigned long long)lock->client_pid,
+ lkowner_utoa(&lock->owner), lock->client, lock->connection_id,
+ blocked);
- count++;
- }
+ gf_proc_dump_write(key, "%s", tmp);
+ count++;
}
+ }
}
void
-dump_entrylks (pl_inode_t *pl_inode)
+dump_entrylks(pl_inode_t *pl_inode)
{
- pthread_mutex_lock (&pl_inode->mutex);
- {
- __dump_entrylks (pl_inode);
- }
- pthread_mutex_unlock (&pl_inode->mutex);
-
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ __dump_entrylks(pl_inode);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
}
void
-__dump_inodelks (pl_inode_t *pl_inode)
+__dump_inodelks(pl_inode_t *pl_inode)
{
- pl_dom_list_t *dom = NULL;
- pl_inode_lock_t *lock = NULL;
- int count = 0;
- char key[GF_DUMP_MAX_BUF_LEN];
-
- char tmp[4098];
-
- list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
+ pl_dom_list_t *dom = NULL;
+ pl_inode_lock_t *lock = NULL;
+ int count = 0;
+ char key[GF_DUMP_MAX_BUF_LEN];
- count = 0;
+ char tmp[4098];
- gf_proc_dump_build_key(key,
- "lock-dump.domain",
- "domain");
- gf_proc_dump_write(key, "%s", dom->domain);
+ list_for_each_entry(dom, &pl_inode->dom_list, inode_list)
+ {
+ count = 0;
- list_for_each_entry (lock, &dom->inodelk_list, list) {
+ gf_proc_dump_build_key(key, "lock-dump.domain", "domain");
+ gf_proc_dump_write(key, "%s", dom->domain);
- gf_proc_dump_build_key(key,
- "inodelk",
- "inodelk[%d](ACTIVE)",count );
-
- SET_FLOCK_PID (&lock->user_flock, lock);
- pl_dump_lock (tmp, sizeof (tmp), &lock->user_flock,
- &lock->owner,
- lock->client, lock->connection_id,
- &lock->granted_time.tv_sec,
- &lock->blkd_time.tv_sec,
- _gf_true);
- gf_proc_dump_write(key, tmp);
-
- count++;
- }
+ list_for_each_entry(lock, &dom->inodelk_list, list)
+ {
+ gf_proc_dump_build_key(key, "inodelk", "inodelk[%d](ACTIVE)",
+ count);
- list_for_each_entry (lock, &dom->blocked_inodelks, blocked_locks) {
+ SET_FLOCK_PID(&lock->user_flock, lock);
+ pl_dump_lock(tmp, sizeof(tmp), &lock->user_flock, &lock->owner,
+ lock->client, lock->connection_id, &lock->granted_time,
+ &lock->blkd_time, _gf_true);
+ gf_proc_dump_write(key, "%s", tmp);
- gf_proc_dump_build_key(key,
- "inodelk",
- "inodelk[%d](BLOCKED)",count );
- SET_FLOCK_PID (&lock->user_flock, lock);
- pl_dump_lock (tmp, sizeof (tmp), &lock->user_flock,
- &lock->owner,
- lock->client, lock->connection_id,
- 0, &lock->blkd_time.tv_sec,
- _gf_false);
- gf_proc_dump_write(key, tmp);
+ count++;
+ }
- count++;
- }
+ list_for_each_entry(lock, &dom->blocked_inodelks, blocked_locks)
+ {
+ gf_proc_dump_build_key(key, "inodelk", "inodelk[%d](BLOCKED)",
+ count);
+ SET_FLOCK_PID(&lock->user_flock, lock);
+ pl_dump_lock(tmp, sizeof(tmp), &lock->user_flock, &lock->owner,
+ lock->client, lock->connection_id, 0, &lock->blkd_time,
+ _gf_false);
+ gf_proc_dump_write(key, "%s", tmp);
+ count++;
}
-
+ }
}
void
-dump_inodelks (pl_inode_t *pl_inode)
+dump_inodelks(pl_inode_t *pl_inode)
{
- pthread_mutex_lock (&pl_inode->mutex);
- {
- __dump_inodelks (pl_inode);
- }
- pthread_mutex_unlock (&pl_inode->mutex);
-
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ __dump_inodelks(pl_inode);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
}
void
-__dump_posixlks (pl_inode_t *pl_inode)
+__dump_posixlks(pl_inode_t *pl_inode)
{
- posix_lock_t *lock = NULL;
- int count = 0;
- char key[GF_DUMP_MAX_BUF_LEN];
+ posix_lock_t *lock = NULL;
+ int count = 0;
+ char key[GF_DUMP_MAX_BUF_LEN];
- char tmp[4098];
+ char tmp[4098];
- list_for_each_entry (lock, &pl_inode->ext_list, list) {
+ list_for_each_entry(lock, &pl_inode->ext_list, list)
+ {
+ SET_FLOCK_PID(&lock->user_flock, lock);
+ gf_proc_dump_build_key(key, "posixlk", "posixlk[%d](%s)", count,
+ lock->blocked ? "BLOCKED" : "ACTIVE");
+ pl_dump_lock(tmp, sizeof(tmp), &lock->user_flock, &lock->owner,
+ lock->client, lock->client_uid, &lock->granted_time,
+ &lock->blkd_time, (lock->blocked) ? _gf_false : _gf_true);
+ gf_proc_dump_write(key, "%s", tmp);
- SET_FLOCK_PID (&lock->user_flock, lock);
- gf_proc_dump_build_key(key,
- "posixlk",
- "posixlk[%d](%s)",
- count,
- lock->blocked ? "BLOCKED" : "ACTIVE");
- pl_dump_lock (tmp, sizeof (tmp), &lock->user_flock,
- &lock->owner, lock->client, NULL,
- &lock->granted_time.tv_sec, &lock->blkd_time.tv_sec,
- (lock->blocked)? _gf_false: _gf_true);
- gf_proc_dump_write(key, tmp);
-
- count++;
- }
+ count++;
+ }
}
void
-dump_posixlks (pl_inode_t *pl_inode)
+dump_posixlks(pl_inode_t *pl_inode)
{
- pthread_mutex_lock (&pl_inode->mutex);
- {
- __dump_posixlks (pl_inode);
- }
- pthread_mutex_unlock (&pl_inode->mutex);
-
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ __dump_posixlks(pl_inode);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
}
int32_t
-pl_dump_inode_priv (xlator_t *this, inode_t *inode)
-{
-
- int ret = -1;
- uint64_t tmp_pl_inode = 0;
- pl_inode_t *pl_inode = NULL;
- char *pathname = NULL;
- gf_boolean_t section_added = _gf_false;
-
- int count = 0;
-
- if (!inode) {
- errno = EINVAL;
- goto out;
- }
-
- ret = TRY_LOCK (&inode->lock);
+pl_dump_inode_priv(xlator_t *this, inode_t *inode)
+{
+ int ret = -1;
+ uint64_t tmp_pl_inode = 0;
+ pl_inode_t *pl_inode = NULL;
+ char *pathname = NULL;
+ gf_boolean_t section_added = _gf_false;
+
+ int count = 0;
+
+ if (!inode) {
+ errno = EINVAL;
+ goto out;
+ }
+
+ ret = TRY_LOCK(&inode->lock);
+ if (ret)
+ goto out;
+ {
+ ret = __inode_ctx_get(inode, this, &tmp_pl_inode);
if (ret)
- goto out;
- {
- ret = __inode_ctx_get (inode, this, &tmp_pl_inode);
- if (ret)
- goto unlock;
- }
+ goto unlock;
+ }
unlock:
- UNLOCK (&inode->lock);
- if (ret)
- goto out;
+ UNLOCK(&inode->lock);
+ if (ret)
+ goto out;
- pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
- if (!pl_inode) {
- ret = -1;
- goto out;
- }
+ pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
+ if (!pl_inode) {
+ ret = -1;
+ goto out;
+ }
- gf_proc_dump_add_section("xlator.features.locks.%s.inode", this->name);
- section_added = _gf_true;
+ gf_proc_dump_add_section("xlator.features.locks.%s.inode", this->name);
+ section_added = _gf_true;
- /*We are safe to call __inode_path since we have the
- * inode->table->lock */
- __inode_path (inode, NULL, &pathname);
- if (pathname)
- gf_proc_dump_write ("path", "%s", pathname);
+ /*We are safe to call __inode_path since we have the
+ * inode->table->lock */
+ __inode_path(inode, NULL, &pathname);
+ if (pathname)
+ gf_proc_dump_write("path", "%s", pathname);
- gf_proc_dump_write("mandatory", "%d", pl_inode->mandatory);
+ gf_proc_dump_write("mandatory", "%d", pl_inode->mandatory);
- ret = pthread_mutex_trylock (&pl_inode->mutex);
- if (ret)
- goto out;
- {
- count = __get_entrylk_count (this, pl_inode);
- if (count) {
- gf_proc_dump_write("entrylk-count", "%d", count);
- __dump_entrylks (pl_inode);
- }
+ ret = pthread_mutex_trylock(&pl_inode->mutex);
+ if (ret)
+ goto out;
+ {
+ count = __get_entrylk_count(this, pl_inode);
+ if (count) {
+ gf_proc_dump_write("entrylk-count", "%d", count);
+ __dump_entrylks(pl_inode);
+ }
- count = __get_inodelk_count (this, pl_inode, NULL);
- if (count) {
- gf_proc_dump_write("inodelk-count", "%d", count);
- __dump_inodelks (pl_inode);
- }
+ count = __get_inodelk_count(this, pl_inode, NULL);
+ if (count) {
+ gf_proc_dump_write("inodelk-count", "%d", count);
+ __dump_inodelks(pl_inode);
+ }
- count = __get_posixlk_count (this, pl_inode);
- if (count) {
- gf_proc_dump_write("posixlk-count", "%d", count);
- __dump_posixlks (pl_inode);
- }
+ count = __get_posixlk_count(pl_inode);
+ if (count) {
+ gf_proc_dump_write("posixlk-count", "%d", count);
+ __dump_posixlks(pl_inode);
}
- pthread_mutex_unlock (&pl_inode->mutex);
+
+ gf_proc_dump_write("links", "%d", pl_inode->links);
+ gf_proc_dump_write("removes_pending", "%u", pl_inode->remove_running);
+ gf_proc_dump_write("removed", "%u", pl_inode->removed);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
out:
- GF_FREE (pathname);
-
- if (ret && inode) {
- if (!section_added)
- gf_proc_dump_add_section ("xlator.features.locks.%s."
- "inode", this->name);
- gf_proc_dump_write ("Unable to print lock state", "(Lock "
- "acquisition failure) %s",
- uuid_utoa (inode->gfid));
- }
- return ret;
+ GF_FREE(pathname);
+
+ if (ret && inode) {
+ if (!section_added)
+ gf_proc_dump_add_section(
+ "xlator.features.locks.%s."
+ "inode",
+ this->name);
+ gf_proc_dump_write("Unable to print lock state",
+ "(Lock "
+ "acquisition failure) %s",
+ uuid_utoa(inode->gfid));
+ }
+ return ret;
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
-
- if (!this)
- return ret;
+ int ret = -1;
- ret = xlator_mem_acct_init (this, gf_locks_mt_end + 1);
+ if (!this)
+ return ret;
- if (ret != 0) {
- gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
- "failed");
- return ret;
- }
+ ret = xlator_mem_acct_init(this, gf_locks_mt_end + 1);
+ if (ret != 0) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Memory accounting init"
+ "failed");
return ret;
-}
+ }
+ return ret;
+}
-pl_ctx_t*
-pl_ctx_get (client_t *client, xlator_t *xlator)
+pl_ctx_t *
+pl_ctx_get(client_t *client, xlator_t *xlator)
{
- void *tmp = NULL;
- pl_ctx_t *ctx = NULL;
- pl_ctx_t *setted_ctx = NULL;
+ void *tmp = NULL;
+ pl_ctx_t *ctx = NULL;
+ pl_ctx_t *setted_ctx = NULL;
- client_ctx_get (client, xlator, &tmp);
+ client_ctx_get(client, xlator, &tmp);
- ctx = tmp;
+ ctx = tmp;
- if (ctx != NULL)
- goto out;
+ if (ctx != NULL)
+ goto out;
- ctx = GF_CALLOC (1, sizeof (pl_ctx_t), gf_locks_mt_posix_lock_t);
+ ctx = GF_CALLOC(1, sizeof(pl_ctx_t), gf_locks_mt_posix_lock_t);
- if (ctx == NULL)
- goto out;
+ if (ctx == NULL)
+ goto out;
- pthread_mutex_init (&ctx->lock, NULL);
- INIT_LIST_HEAD (&ctx->inodelk_lockers);
- INIT_LIST_HEAD (&ctx->entrylk_lockers);
- INIT_LIST_HEAD (&ctx->metalk_list);
+ pthread_mutex_init(&ctx->lock, NULL);
+ INIT_LIST_HEAD(&ctx->inodelk_lockers);
+ INIT_LIST_HEAD(&ctx->entrylk_lockers);
+ INIT_LIST_HEAD(&ctx->metalk_list);
- setted_ctx = client_ctx_set (client, xlator, ctx);
- if (ctx != setted_ctx) {
- pthread_mutex_destroy (&ctx->lock);
- GF_FREE (ctx);
- ctx = setted_ctx;
- }
+ setted_ctx = client_ctx_set(client, xlator, ctx);
+ if (ctx != setted_ctx) {
+ pthread_mutex_destroy(&ctx->lock);
+ GF_FREE(ctx);
+ ctx = setted_ctx;
+ }
out:
- return ctx;
+ return ctx;
}
int
-pl_metalk_client_cleanup (xlator_t *this, pl_ctx_t *ctx)
-{
- pl_meta_lock_t *meta_lock = NULL;
- pl_meta_lock_t *tmp_metalk = NULL;
- pl_inode_t *pl_inode = NULL;
- posix_lock_t *posix_lock = NULL;
- posix_lock_t *tmp_posixlk = NULL;
- struct list_head tmp_posixlk_list;
-
- INIT_LIST_HEAD (&tmp_posixlk_list);
-
- pthread_mutex_lock (&ctx->lock);
+pl_metalk_client_cleanup(xlator_t *this, pl_ctx_t *ctx)
+{
+ pl_meta_lock_t *meta_lock = NULL;
+ pl_meta_lock_t *tmp_metalk = NULL;
+ pl_inode_t *pl_inode = NULL;
+ posix_lock_t *posix_lock = NULL;
+ posix_lock_t *tmp_posixlk = NULL;
+ struct list_head tmp_posixlk_list;
+
+ INIT_LIST_HEAD(&tmp_posixlk_list);
+
+ pthread_mutex_lock(&ctx->lock);
+ {
+ /* if this list is empty then pl_inode->metalk_list should be
+ * empty too. meta lock should in all cases be added/removed
+ * from both pl_ctx_t and pl_inode */
+ if (list_empty(&ctx->metalk_list))
+ goto unlock;
+
+ list_for_each_entry_safe(meta_lock, tmp_metalk, &ctx->metalk_list,
+ client_list)
{
+ list_del_init(&meta_lock->client_list);
- /* if this list is empty then pl_inode->metalk_list should be
- * empty too. meta lock should in all cases be added/removed
- * from both pl_ctx_t and pl_inode */
- if (list_empty (&ctx->metalk_list))
- goto unlock;
-
- list_for_each_entry_safe (meta_lock, tmp_metalk,
- &ctx->metalk_list, client_list) {
- list_del_init (&meta_lock->client_list);
-
- pl_inode = meta_lock->pl_inode;
+ pl_inode = meta_lock->pl_inode;
- pthread_mutex_lock (&pl_inode->mutex);
+ pthread_mutex_lock(&pl_inode->mutex);
- {
+ {
+ /* Since the migration status is unknown here
+ * unwind all queued and blocked locks to check
+ * migration status and find the correct
+ * destination */
+ __unwind_queued_locks(pl_inode, &tmp_posixlk_list);
- /* Since the migration status is unknown here
- * unwind all queued and blocked locks to check
- * migration status and find the correct
- * destination */
- __unwind_queued_locks (this, pl_inode,
- &tmp_posixlk_list);
+ __unwind_blocked_locks(pl_inode, &tmp_posixlk_list);
- __unwind_blocked_locks (this, pl_inode,
- &tmp_posixlk_list);
+ list_del_init(&meta_lock->list);
- list_del_init (&meta_lock->list);
+ pl_metalk_unref(meta_lock);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
- pl_metalk_unref (meta_lock);
-
- }
- pthread_mutex_unlock (&pl_inode->mutex);
-
- /* The corresponding ref is taken in
- * pl_insert_metalk*/
- inode_unref (pl_inode->inode);
- }
+ /* The corresponding ref is taken in
+ * pl_insert_metalk*/
+ inode_unref(pl_inode->inode);
}
+ }
unlock:
- pthread_mutex_unlock (&ctx->lock);
+ pthread_mutex_unlock(&ctx->lock);
- list_for_each_entry_safe (posix_lock, tmp_posixlk, &tmp_posixlk_list,
- list) {
- list_del_init (&posix_lock->list);
+ list_for_each_entry_safe(posix_lock, tmp_posixlk, &tmp_posixlk_list, list)
+ {
+ list_del_init(&posix_lock->list);
- STACK_UNWIND_STRICT (lk, posix_lock->frame, -1, EREMOTE,
- &posix_lock->user_flock, NULL);
+ STACK_UNWIND_STRICT(lk, posix_lock->frame, -1, EREMOTE,
+ &posix_lock->user_flock, NULL);
- __destroy_lock(posix_lock);
- }
- return 0;
+ __destroy_lock(posix_lock);
+ }
+ return 0;
}
static int
-pl_client_disconnect_cbk (xlator_t *this, client_t *client)
+pl_client_disconnect_cbk(xlator_t *this, client_t *client)
{
- pl_ctx_t *pl_ctx = NULL;
-
- pl_ctx = pl_ctx_get (client, this);
-
- if (pl_ctx) {
- pl_inodelk_client_cleanup (this, pl_ctx);
- pl_entrylk_client_cleanup (this, pl_ctx);
- pl_metalk_client_cleanup (this, pl_ctx);
- }
+ pl_ctx_t *pl_ctx = pl_ctx_get(client, this);
+ if (pl_ctx) {
+ pl_inodelk_client_cleanup(this, pl_ctx);
+ pl_entrylk_client_cleanup(this, pl_ctx);
+ pl_metalk_client_cleanup(this, pl_ctx);
+ }
- return 0;
+ return 0;
}
-
static int
-pl_client_destroy_cbk (xlator_t *this, client_t *client)
+pl_client_destroy_cbk(xlator_t *this, client_t *client)
{
- void *tmp = NULL;
- pl_ctx_t *pl_ctx = NULL;
+ void *tmp = NULL;
+ pl_ctx_t *pl_ctx = NULL;
- pl_client_disconnect_cbk (this, client);
+ pl_client_disconnect_cbk(this, client);
- client_ctx_del (client, this, &tmp);
+ client_ctx_del(client, this, &tmp);
- if (tmp == NULL)
- return 0;
+ if (tmp == NULL)
+ return 0;
- pl_ctx = tmp;
+ pl_ctx = tmp;
- GF_ASSERT (list_empty(&pl_ctx->inodelk_lockers));
- GF_ASSERT (list_empty(&pl_ctx->entrylk_lockers));
+ GF_ASSERT(list_empty(&pl_ctx->inodelk_lockers));
+ GF_ASSERT(list_empty(&pl_ctx->entrylk_lockers));
- pthread_mutex_destroy (&pl_ctx->lock);
- GF_FREE (pl_ctx);
+ pthread_mutex_destroy(&pl_ctx->lock);
+ GF_FREE(pl_ctx);
- return 0;
+ return 0;
}
int
-reconfigure (xlator_t *this, dict_t *options)
+reconfigure(xlator_t *this, dict_t *options)
{
- posix_locks_private_t *priv = NULL;
- int ret = -1;
+ posix_locks_private_t *priv = this->private;
+ int ret = -1;
+ char *tmp_str = NULL;
- priv = this->private;
+ GF_OPTION_RECONF("trace", priv->trace, options, bool, out);
- GF_OPTION_RECONF ("trace", priv->trace, options, bool, out);
+ GF_OPTION_RECONF("monkey-unlocking", priv->monkey_unlocking, options, bool,
+ out);
- GF_OPTION_RECONF ("monkey-unlocking", priv->monkey_unlocking, options,
- bool, out);
+ GF_OPTION_RECONF("revocation-secs", priv->revocation_secs, options, uint32,
+ out);
- GF_OPTION_RECONF ("revocation-secs",
- priv->revocation_secs, options,
- uint32, out);
+ GF_OPTION_RECONF("revocation-clear-all", priv->revocation_clear_all,
+ options, bool, out);
- GF_OPTION_RECONF ("revocation-clear-all", priv->revocation_clear_all,
- options, bool, out);
+ GF_OPTION_RECONF("revocation-max-blocked", priv->revocation_max_blocked,
+ options, uint32, out);
- GF_OPTION_RECONF ("revocation-max-blocked",
- priv->revocation_max_blocked, options,
- uint32, out);
+ GF_OPTION_RECONF("notify-contention", priv->notify_contention, options,
+ bool, out);
- GF_OPTION_RECONF ("notify-contention", priv->notify_contention,
- options, bool, out);
+ GF_OPTION_RECONF("notify-contention-delay", priv->notify_contention_delay,
+ options, uint32, out);
- GF_OPTION_RECONF ("notify-contention-delay",
- priv->notify_contention_delay, options, uint32, out);
+ GF_OPTION_RECONF("mandatory-locking", tmp_str, options, str, out);
- ret = 0;
+ GF_OPTION_RECONF("enforce-mandatory-lock", priv->mlock_enforced, options,
+ bool, out);
+
+ if (!strcmp(tmp_str, "forced"))
+ priv->mandatory_mode = MLK_FORCED;
+ else if (!strcmp(tmp_str, "file"))
+ priv->mandatory_mode = MLK_FILE_BASED;
+ else if (!strcmp(tmp_str, "optimal"))
+ priv->mandatory_mode = MLK_OPTIMAL;
+ else
+ priv->mandatory_mode = MLK_NONE;
+
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-init (xlator_t *this)
+init(xlator_t *this)
{
- posix_locks_private_t *priv = NULL;
- xlator_list_t *trav = NULL;
- char *tmp_str = NULL;
- int ret = -1;
+ posix_locks_private_t *priv = NULL;
+ xlator_list_t *trav = NULL;
+ char *tmp_str = NULL;
+ int ret = -1;
- if (!this->children || this->children->next) {
- gf_log (this->name, GF_LOG_CRITICAL,
- "FATAL: posix-locks should have exactly one child");
- goto out;
- }
+ if (!this->children || this->children->next) {
+ gf_log(this->name, GF_LOG_CRITICAL,
+ "FATAL: posix-locks should have exactly one child");
+ goto out;
+ }
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "Volume is dangling. Please check the volume file.");
- }
+ if (!this->parents) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "Volume is dangling. Please check the volume file.");
+ }
- trav = this->children;
- while (trav->xlator->children)
- trav = trav->xlator->children;
+ trav = this->children;
+ while (trav->xlator->children)
+ trav = trav->xlator->children;
- if (strncmp ("storage/", trav->xlator->type, 8)) {
- gf_log (this->name, GF_LOG_CRITICAL,
- "'locks' translator is not loaded over a storage "
- "translator");
- goto out;
- }
+ if (strncmp("storage/", trav->xlator->type, 8)) {
+ gf_log(this->name, GF_LOG_CRITICAL,
+ "'locks' translator is not loaded over a storage "
+ "translator");
+ goto out;
+ }
- priv = GF_CALLOC (1, sizeof (*priv),
- gf_locks_mt_posix_locks_private_t);
+ priv = GF_CALLOC(1, sizeof(*priv), gf_locks_mt_posix_locks_private_t);
- GF_OPTION_INIT ("mandatory-locking", tmp_str, str, out);
- if (!strcmp (tmp_str, "forced"))
- priv->mandatory_mode = MLK_FORCED;
- else if (!strcmp (tmp_str, "file"))
- priv->mandatory_mode = MLK_FILE_BASED;
- else if (!strcmp (tmp_str, "optimal"))
- priv->mandatory_mode = MLK_OPTIMAL;
- else
- priv->mandatory_mode = MLK_NONE;
- tmp_str = NULL;
+ GF_OPTION_INIT("mandatory-locking", tmp_str, str, out);
+ if (!strcmp(tmp_str, "forced"))
+ priv->mandatory_mode = MLK_FORCED;
+ else if (!strcmp(tmp_str, "file"))
+ priv->mandatory_mode = MLK_FILE_BASED;
+ else if (!strcmp(tmp_str, "optimal"))
+ priv->mandatory_mode = MLK_OPTIMAL;
+ else
+ priv->mandatory_mode = MLK_NONE;
- GF_OPTION_INIT ("trace", priv->trace, bool, out);
+ tmp_str = NULL;
- GF_OPTION_INIT ("monkey-unlocking", priv->monkey_unlocking,
- bool, out);
+ GF_OPTION_INIT("trace", priv->trace, bool, out);
- GF_OPTION_INIT ("revocation-secs", priv->revocation_secs,
- uint32, out);
+ GF_OPTION_INIT("monkey-unlocking", priv->monkey_unlocking, bool, out);
- GF_OPTION_INIT ("revocation-clear-all", priv->revocation_clear_all,
- bool, out);
+ GF_OPTION_INIT("revocation-secs", priv->revocation_secs, uint32, out);
- GF_OPTION_INIT ("revocation-max-blocked", priv->revocation_max_blocked,
- uint32, out);
+ GF_OPTION_INIT("revocation-clear-all", priv->revocation_clear_all, bool,
+ out);
- GF_OPTION_INIT ("notify-contention", priv->notify_contention, bool,
- out);
+ GF_OPTION_INIT("revocation-max-blocked", priv->revocation_max_blocked,
+ uint32, out);
- GF_OPTION_INIT ("notify-contention-delay",
- priv->notify_contention_delay, uint32, out);
+ GF_OPTION_INIT("notify-contention", priv->notify_contention, bool, out);
- this->local_pool = mem_pool_new (pl_local_t, 32);
- if (!this->local_pool) {
- ret = -1;
- gf_log (this->name, GF_LOG_ERROR,
- "failed to create local_t's memory pool");
- goto out;
- }
+ GF_OPTION_INIT("notify-contention-delay", priv->notify_contention_delay,
+ uint32, out);
+
+ GF_OPTION_INIT("enforce-mandatory-lock", priv->mlock_enforced, bool, out);
- this->private = priv;
- ret = 0;
+ this->local_pool = mem_pool_new(pl_local_t, 32);
+ if (!this->local_pool) {
+ ret = -1;
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to create local_t's memory pool");
+ goto out;
+ }
+
+ this->private = priv;
+ ret = 0;
out:
- if (ret) {
- GF_FREE (priv);
- }
- return ret;
+ if (ret) {
+ GF_FREE(priv);
+ }
+ return ret;
}
-int
-fini (xlator_t *this)
+void
+fini(xlator_t *this)
{
- posix_locks_private_t *priv = NULL;
-
- priv = this->private;
- if (!priv)
- return 0;
- this->private = NULL;
- GF_FREE (priv->brickname);
- GF_FREE (priv);
+ posix_locks_private_t *priv = this->private;
+ if (!priv)
+ return;
+ this->private = NULL;
+ if (this->local_pool) {
+ mem_pool_destroy(this->local_pool);
+ this->local_pool = NULL;
+ }
+ GF_FREE(priv->brickname);
+ GF_FREE(priv);
- return 0;
+ return;
}
int
-pl_inodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *flock,
- dict_t *xdata);
+pl_inodelk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ int32_t cmd, struct gf_flock *flock, dict_t *xdata);
int
-pl_finodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *flock,
- dict_t *xdata);
+pl_finodelk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ int32_t cmd, struct gf_flock *flock, dict_t *xdata);
int
-pl_entrylk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, const char *basename,
- entrylk_cmd cmd, entrylk_type type, dict_t *xdata);
+pl_entrylk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata);
int
-pl_fentrylk (call_frame_t *frame, xlator_t *this,
- const char *volume, fd_t *fd, const char *basename,
- entrylk_cmd cmd, entrylk_type type, dict_t *xdata);
+pl_fentrylk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata);
int32_t
-pl_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent,
- dict_t *xdata)
+pl_rename_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *buf, struct iatt *preoldparent,
+ struct iatt *postoldparent, struct iatt *prenewparent,
+ struct iatt *postnewparent, dict_t *xdata)
{
- PL_STACK_UNWIND (rename, xdata, frame, op_ret, op_errno,
- buf, preoldparent, postoldparent, prenewparent,
- postnewparent, xdata);
- return 0;
+ pl_inode_remove_cbk(this, cookie, op_ret < 0 ? op_errno : 0);
+
+ PL_STACK_UNWIND(rename, xdata, frame, op_ret, op_errno, buf, preoldparent,
+ postoldparent, prenewparent, postnewparent, xdata);
+
+ return 0;
}
int32_t
-pl_rename (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+pl_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, oldloc, newloc);
+ int32_t error;
- STACK_WIND (frame, pl_rename_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->rename, oldloc,
- newloc, xdata);
- return 0;
+ error = PL_INODE_REMOVE(rename, frame, this, oldloc, newloc, pl_rename,
+ pl_rename_cbk, oldloc, newloc, xdata);
+ if (error > 0) {
+ STACK_UNWIND_STRICT(rename, frame, -1, error, NULL, NULL, NULL, NULL,
+ NULL, NULL);
+ }
+
+ return 0;
}
posix_lock_t *
-gf_lkmig_info_to_posix_lock (call_frame_t *frame,
- lock_migration_info_t *lmi)
+gf_lkmig_info_to_posix_lock(call_frame_t *frame, lock_migration_info_t *lmi)
{
- posix_lock_t *lock = NULL;
+ posix_lock_t *lock = GF_CALLOC(1, sizeof(posix_lock_t),
+ gf_locks_mt_posix_lock_t);
+ if (!lock)
+ goto out;
- lock = GF_CALLOC (1, sizeof (posix_lock_t), gf_locks_mt_posix_lock_t);
- if (!lock)
- goto out;
+ lock->fl_start = lmi->flock.l_start;
+ lock->fl_type = lmi->flock.l_type;
- lock->fl_start = lmi->flock.l_start;
- lock->fl_type = lmi->flock.l_type;
+ if (lmi->flock.l_len == 0)
+ lock->fl_end = LLONG_MAX;
+ else
+ lock->fl_end = lmi->flock.l_start + lmi->flock.l_len - 1;
- if (lmi->flock.l_len == 0)
- lock->fl_end = LLONG_MAX;
- else
- lock->fl_end = lmi->flock.l_start + lmi->flock.l_len - 1;
-
- lock->client = frame->root->client;
+ lock->client = frame->root->client;
- lock->lk_flags = lmi->lk_flags;
+ lock->lk_flags = lmi->lk_flags;
- lock->client_uid = gf_strdup (lmi->client_uid);
- if (lock->client_uid == NULL) {
- GF_FREE (lock);
- lock = NULL;
- goto out;
- }
+ lock->client_uid = gf_strdup(lmi->client_uid);
+ if (lock->client_uid == NULL) {
+ GF_FREE(lock);
+ lock = NULL;
+ goto out;
+ }
- lock->client_pid = lmi->flock.l_pid;
- lock->owner = lmi->flock.l_owner;
+ lock->client_pid = lmi->flock.l_pid;
+ lock->owner = lmi->flock.l_owner;
- INIT_LIST_HEAD (&lock->list);
+ INIT_LIST_HEAD(&lock->list);
out:
- return lock;
+ return lock;
}
/* This function is supposed to write the active locks from the source brick(in
* rebalance context) and write here. Hence, will add the locks directly to the
* pl_inode->ext_list*/
int
-pl_write_active_locks (call_frame_t *frame, pl_inode_t *pl_inode,
- lock_migration_info_t *locklist)
+pl_write_active_locks(call_frame_t *frame, pl_inode_t *pl_inode,
+ lock_migration_info_t *locklist)
{
- posix_lock_t *newlock = NULL;
- lock_migration_info_t *temp = NULL;
- int ret = 0;
+ posix_lock_t *newlock = NULL;
+ lock_migration_info_t *temp = NULL;
+ int ret = 0;
- pthread_mutex_lock (&pl_inode->mutex);
- {
- /* Just making sure the activelk list is empty. Should not
- * happen though*/
- if (!list_empty (&pl_inode->ext_list)) {
-
- gf_msg (THIS->name, GF_LOG_ERROR, 0, 0,
- "invalid locks found");
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ /* Just making sure the activelk list is empty. Should not
+ * happen though*/
+ if (!list_empty(&pl_inode->ext_list)) {
+ pthread_mutex_unlock(&pl_inode->mutex);
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, 0, "invalid locks found");
- ret = -1;
- goto out;
- }
+ ret = -1;
+ goto out;
+ }
- /* This list also should not be empty */
- if (list_empty (&locklist->list)) {
- gf_msg (THIS->name, GF_LOG_ERROR, 0, 0,
- "empty lock list");
+ /* This list also should not be empty */
+ if (list_empty(&locklist->list)) {
+ pthread_mutex_unlock(&pl_inode->mutex);
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, 0, "empty lock list");
- ret = -1;
- goto out;
- }
-
- list_for_each_entry (temp, &locklist->list, list) {
+ ret = -1;
+ goto out;
+ }
- newlock = gf_lkmig_info_to_posix_lock (frame, temp);
- if (!newlock) {
- gf_msg (THIS->name, GF_LOG_ERROR, 0, 0,
- "mem allocation failed for newlock");
+ list_for_each_entry(temp, &locklist->list, list)
+ {
+ newlock = gf_lkmig_info_to_posix_lock(frame, temp);
+ if (!newlock) {
+ pthread_mutex_unlock(&pl_inode->mutex);
+ gf_msg(THIS->name, GF_LOG_ERROR, 0, 0,
+ "mem allocation failed for newlock");
- ret = -1;
- goto out;
- }
- list_add_tail (&newlock->list, &pl_inode->ext_list);
- }
+ ret = -1;
+ goto out;
+ }
+ list_add_tail(&newlock->list, &pl_inode->ext_list);
}
-
+ }
+ /*TODO: What if few lock add failed with ENOMEM. Should the already
+ * added locks be clearted */
+ pthread_mutex_unlock(&pl_inode->mutex);
out:
- /*TODO: What if few lock add failed with ENOMEM. Should the already
- * added locks be clearted */
- pthread_mutex_unlock (&pl_inode->mutex);
-
- return ret;
+ return ret;
}
static int
-pl_setactivelk (call_frame_t *frame, xlator_t *this, loc_t *loc,
- lock_migration_info_t *locklist, dict_t *xdata)
+pl_setactivelk(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ lock_migration_info_t *locklist, dict_t *xdata)
{
- pl_inode_t *pl_inode = NULL;
- int op_ret = 0;
- int op_errno = 0;
- int ret = 0;
+ int op_ret = 0;
+ int op_errno = 0;
+ int ret = 0;
- pl_inode = pl_inode_get (this, loc->inode);
- if (!pl_inode) {
- gf_msg (this->name, GF_LOG_ERROR, 0, 0,
- "pl_inode_get failed");
+ pl_inode_t *pl_inode = pl_inode_get(this, loc->inode, NULL);
+ if (!pl_inode) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, 0, "pl_inode_get failed");
- op_ret = -1;
- op_errno = ENOMEM;
- goto out;
- }
- ret = pl_write_active_locks (frame, pl_inode, locklist);
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto out;
+ }
+ ret = pl_write_active_locks(frame, pl_inode, locklist);
- op_ret = ret;
+ op_ret = ret;
out:
- STACK_UNWIND_STRICT (setactivelk, frame, op_ret, op_errno, NULL);
+ STACK_UNWIND_STRICT(setactivelk, frame, op_ret, op_errno, NULL);
- return 0;
+ return 0;
}
int32_t
-pl_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+pl_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- PL_STACK_UNWIND (unlink, xdata, frame, op_ret, op_errno, preparent,
- postparent, xdata);
- return 0;
+ pl_inode_remove_cbk(this, cookie, op_ret < 0 ? op_errno : 0);
+
+ PL_STACK_UNWIND(unlink, xdata, frame, op_ret, op_errno, preparent,
+ postparent, xdata);
+
+ return 0;
}
int32_t
-pl_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
- dict_t *xdata)
+pl_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, loc, NULL);
- STACK_WIND (frame, pl_unlink_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
- return 0;
+ int32_t error;
+
+ error = PL_INODE_REMOVE(unlink, frame, this, loc, NULL, pl_unlink,
+ pl_unlink_cbk, loc, xflag, xdata);
+ if (error > 0) {
+ STACK_UNWIND_STRICT(unlink, frame, -1, error, NULL, NULL, NULL);
+ }
+
+ return 0;
}
int32_t
-pl_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+pl_mkdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode, struct iatt *buf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
{
- PL_STACK_UNWIND (mkdir, xdata, frame, op_ret, op_errno,
- inode, buf, preparent, postparent, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(mkdir, xdata, frame, op_ret, op_errno, inode,
+ buf, preparent, postparent, xdata);
+ return 0;
}
int
-pl_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc,
- mode_t mode, mode_t umask, dict_t *xdata)
+pl_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, loc, NULL);
- STACK_WIND (frame, pl_mkdir_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->mkdir, loc, mode,
- umask, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, ((fd_t *)NULL), loc, NULL);
+ STACK_WIND(frame, pl_mkdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
+ return 0;
}
int32_t
-pl_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- dict_t *xdata)
+pl_stat_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *buf, dict_t *xdata)
{
- PL_STACK_UNWIND (stat, xdata, frame, op_ret, op_errno, buf, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(stat, xdata, frame, op_ret, op_errno, buf,
+ xdata);
+ return 0;
}
int
-pl_stat (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xdata)
+pl_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, loc, NULL);
- STACK_WIND (frame, pl_stat_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->stat, loc, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, ((fd_t *)NULL), loc, NULL);
+ STACK_WIND(frame, pl_stat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->stat, loc, xdata);
+ return 0;
}
int32_t
-pl_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+pl_mknod_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode, struct iatt *buf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
{
- PL_STACK_UNWIND (mknod, xdata, frame, op_ret, op_errno,
- inode, buf, preparent, postparent, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(mknod, xdata, frame, op_ret, op_errno, inode,
+ buf, preparent, postparent, xdata);
+ return 0;
}
int
-pl_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t rdev,
- mode_t umask, dict_t *xdata)
-{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, loc, NULL);
- STACK_WIND (frame, pl_mknod_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->mknod, loc, mode,
- rdev, umask, xdata);
- return 0;
+pl_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *xdata)
+{
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, ((fd_t *)NULL), loc, NULL);
+ STACK_WIND(frame, pl_mknod_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, loc, mode, rdev, umask, xdata);
+ return 0;
}
int32_t
-pl_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+pl_rmdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- PL_STACK_UNWIND (rmdir, xdata, frame, op_ret, op_errno,
- preparent, postparent, xdata);
- return 0;
+ pl_inode_remove_cbk(this, cookie, op_ret < 0 ? op_errno : 0);
+
+ PL_STACK_UNWIND_FOR_CLIENT(rmdir, xdata, frame, op_ret, op_errno, preparent,
+ postparent, xdata);
+
+ return 0;
}
int
-pl_rmdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int xflags, dict_t *xdata)
+pl_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflags,
+ dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, loc, NULL);
- STACK_WIND (frame, pl_rmdir_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->rmdir, loc,
- xflags, xdata);
- return 0;
+ int32_t error;
+
+ error = PL_INODE_REMOVE(rmdir, frame, this, loc, NULL, pl_rmdir,
+ pl_rmdir_cbk, loc, xflags, xdata);
+ if (error > 0) {
+ STACK_UNWIND_STRICT(rmdir, frame, -1, error, NULL, NULL, NULL);
+ }
+
+ return 0;
}
int32_t
-pl_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+pl_symlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- PL_STACK_UNWIND (symlink, xdata, frame, op_ret, op_errno,
- inode, buf, preparent, postparent, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(symlink, xdata, frame, op_ret, op_errno, inode,
+ buf, preparent, postparent, xdata);
+ return 0;
}
int
-pl_symlink (call_frame_t *frame, xlator_t *this,
- const char *linkname, loc_t *loc,
- mode_t umask, dict_t *xdata)
-{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, loc, NULL);
- STACK_WIND (frame, pl_symlink_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->symlink, linkname,
- loc, umask, xdata);
- return 0;
+pl_symlink(call_frame_t *frame, xlator_t *this, const char *linkname,
+ loc_t *loc, mode_t umask, dict_t *xdata)
+{
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, ((fd_t *)NULL), loc, NULL);
+ STACK_WIND(frame, pl_symlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->symlink, linkname, loc, umask, xdata);
+ return 0;
}
int32_t
-pl_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+pl_link_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode, struct iatt *buf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
{
- PL_STACK_UNWIND (link, xdata, frame, op_ret, op_errno,
- inode, buf, preparent, postparent, xdata);
- return 0;
+ pl_inode_t *pl_inode = (pl_inode_t *)cookie;
+
+ if (op_ret >= 0) {
+ pthread_mutex_lock(&pl_inode->mutex);
+
+ /* TODO: can happen pl_inode->links == 0 ? */
+ if (pl_inode->links >= 0) {
+ pl_inode->links++;
+ }
+
+ pthread_mutex_unlock(&pl_inode->mutex);
+ }
+
+ PL_STACK_UNWIND_FOR_CLIENT(link, xdata, frame, op_ret, op_errno, inode, buf,
+ preparent, postparent, xdata);
+ return 0;
}
int
-pl_link (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+pl_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, oldloc, newloc);
- STACK_WIND (frame, pl_link_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
+ pl_inode_t *pl_inode;
+
+ pl_inode = pl_inode_get(this, oldloc->inode, NULL);
+ if (pl_inode == NULL) {
+ STACK_UNWIND_STRICT(link, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
+ NULL);
return 0;
+ }
+
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, ((fd_t *)NULL), oldloc, newloc);
+ STACK_WIND_COOKIE(frame, pl_link_cbk, pl_inode, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
+ return 0;
}
int32_t
-pl_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf,
- dict_t *xdata)
+pl_fsync_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata)
{
- PL_STACK_UNWIND (fsync, xdata, frame, op_ret, op_errno,
- prebuf, postbuf, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(fsync, xdata, frame, op_ret, op_errno, prebuf,
+ postbuf, xdata);
+ return 0;
}
int
-pl_fsync (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int32_t datasync, dict_t *xdata)
+pl_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+ dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, fd, NULL, NULL);
- STACK_WIND (frame, pl_fsync_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->fsync, fd, datasync, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
+ STACK_WIND(frame, pl_fsync_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsync, fd, datasync, xdata);
+ return 0;
}
int32_t
-pl_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
- dict_t *xdata)
+pl_readdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+ dict_t *xdata)
{
- PL_STACK_UNWIND (readdir, xdata, frame, op_ret, op_errno,
- entries, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(readdir, xdata, frame, op_ret, op_errno, entries,
+ xdata);
+ return 0;
}
-
int
-pl_readdir (call_frame_t *frame, xlator_t *this,
- fd_t *fd, size_t size, off_t offset,
- dict_t *xdata)
+pl_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, fd, NULL, NULL);
- STACK_WIND (frame, pl_readdir_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->readdir, fd,
- size, offset, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
+ STACK_WIND(frame, pl_readdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdir, fd, size, offset, xdata);
+ return 0;
}
int32_t
-pl_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+pl_fsyncdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- PL_STACK_UNWIND (fsyncdir, xdata, frame, op_ret, op_errno, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(fsyncdir, xdata, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-pl_fsyncdir (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int32_t datasync, dict_t *xdata)
+pl_fsyncdir(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+ dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, fd, NULL, NULL);
- STACK_WIND (frame, pl_fsyncdir_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->fsyncdir, fd,
- datasync, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
+ STACK_WIND(frame, pl_fsyncdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsyncdir, fd, datasync, xdata);
+ return 0;
}
int32_t
-pl_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct statvfs *buf,
- dict_t *xdata)
+pl_statfs_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct statvfs *buf, dict_t *xdata)
{
- PL_STACK_UNWIND (statfs, xdata, frame, op_ret, op_errno, buf, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(statfs, xdata, frame, op_ret, op_errno, buf,
+ xdata);
+ return 0;
}
int
-pl_statfs (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xdata)
+pl_statfs(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, loc, NULL);
- STACK_WIND (frame, pl_statfs_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->statfs, loc, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, ((fd_t *)NULL), loc, NULL);
+ STACK_WIND(frame, pl_statfs_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->statfs, loc, xdata);
+ return 0;
}
int32_t
-pl_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+pl_removexattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- PL_STACK_UNWIND (removexattr, xdata, frame, op_ret, op_errno, xdata);
- return 0;
+ pl_local_t *local = NULL;
+ pl_inode_t *pl_inode = NULL;
+
+ local = frame->local;
+ if (local && local->update_mlock_enforced_flag && op_ret != -1) {
+ pl_inode = pl_inode_get(this, local->inode, NULL);
+ if (!pl_inode) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ pl_inode->mlock_enforced = _gf_false;
+ pl_inode->check_mlock_info = _gf_false;
+ pl_inode->track_fop_wind_count = _gf_true;
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+ }
+
+unwind:
+ PL_STACK_UNWIND_FOR_CLIENT(removexattr, xdata, frame, op_ret, op_errno,
+ xdata);
+ return 0;
}
int
-pl_removexattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
+pl_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, loc, NULL);
- STACK_WIND (frame, pl_removexattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
- return 0;
+ int op_ret = 0;
+ int op_errno = EINVAL;
+ posix_locks_private_t *priv = this->private;
+
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, ((fd_t *)NULL), loc, NULL);
+
+ PL_CHECK_LOCK_ENFORCE_KEY(frame, ((dict_t *)NULL), name, this, loc,
+ ((fd_t *)NULL), priv);
+
+ STACK_WIND(frame, pl_removexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
+ return 0;
+
+unwind:
+ PL_STACK_UNWIND_FOR_CLIENT(removexattr, xdata, frame, op_ret, op_errno,
+ NULL);
+
+ return 0;
}
int32_t
-pl_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+pl_fremovexattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- PL_STACK_UNWIND (fremovexattr, xdata, frame, op_ret, op_errno, xdata);
- return 0;
+ pl_local_t *local = NULL;
+ pl_inode_t *pl_inode = NULL;
+
+ local = frame->local;
+ if (local && local->update_mlock_enforced_flag && op_ret != -1) {
+ pl_inode = pl_inode_get(this, local->inode, NULL);
+ if (!pl_inode) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ pl_inode->mlock_enforced = _gf_false;
+ pl_inode->check_mlock_info = _gf_false;
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+ }
+
+unwind:
+ PL_STACK_UNWIND_FOR_CLIENT(fremovexattr, xdata, frame, op_ret, op_errno,
+ xdata);
+ return 0;
}
int
-pl_fremovexattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, const char *name, dict_t *xdata)
+pl_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
+ dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, fd, NULL, NULL);
- STACK_WIND (frame, pl_fremovexattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->fremovexattr, fd, name, xdata);
- return 0;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ posix_locks_private_t *priv = this->private;
+
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
+
+ PL_CHECK_LOCK_ENFORCE_KEY(frame, ((dict_t *)NULL), name, this,
+ ((loc_t *)NULL), fd, priv);
+
+ STACK_WIND(frame, pl_fremovexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fremovexattr, fd, name, xdata);
+ return 0;
+
+unwind:
+ PL_STACK_UNWIND_FOR_CLIENT(fremovexattr, xdata, frame, op_ret, op_errno,
+ NULL);
+ return 0;
}
int32_t
-pl_rchecksum_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, uint32_t weak_cksum,
- uint8_t *strong_cksum, dict_t *xdata)
+pl_rchecksum_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, uint32_t weak_cksum,
+ uint8_t *strong_cksum, dict_t *xdata)
{
- PL_STACK_UNWIND (rchecksum, xdata, frame, op_ret, op_errno,
- weak_cksum, strong_cksum, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(rchecksum, xdata, frame, op_ret, op_errno,
+ weak_cksum, strong_cksum, xdata);
+ return 0;
}
int
-pl_rchecksum (call_frame_t *frame, xlator_t *this,
- fd_t *fd, off_t offset, int32_t len,
- dict_t *xdata)
+pl_rchecksum(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ int32_t len, dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, fd, NULL, NULL);
- STACK_WIND (frame, pl_rchecksum_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->rchecksum, fd, offset, len, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
+ STACK_WIND(frame, pl_rchecksum_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rchecksum, fd, offset, len, xdata);
+ return 0;
}
int32_t
-pl_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
+pl_xattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- PL_STACK_UNWIND (xattrop, xdata, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(xattrop, xdata, frame, op_ret, op_errno, dict,
+ xdata);
+ return 0;
}
int
-pl_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t optype, dict_t *xattr,
- dict_t *xdata)
+pl_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, loc, NULL);
- STACK_WIND (frame, pl_xattrop_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->xattrop, loc,
- optype, xattr, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, ((fd_t *)NULL), loc, NULL);
+ STACK_WIND(frame, pl_xattrop_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->xattrop, loc, optype, xattr, xdata);
+ return 0;
}
int32_t
-pl_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
+pl_fxattrop_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
{
- PL_STACK_UNWIND (fxattrop, xdata, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(fxattrop, xdata, frame, op_ret, op_errno, dict,
+ xdata);
+ return 0;
}
int
-pl_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
- gf_xattrop_flags_t optype, dict_t *xattr,
- dict_t *xdata)
+pl_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, fd, NULL, NULL);
- STACK_WIND (frame, pl_fxattrop_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->fxattrop, fd,
- optype, xattr, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
+ STACK_WIND(frame, pl_fxattrop_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fxattrop, fd, optype, xattr, xdata);
+ return 0;
}
int32_t
-pl_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *statpre,
- struct iatt *statpost,
- dict_t *xdata)
+pl_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
{
- PL_STACK_UNWIND (setattr, xdata, frame, op_ret, op_errno,
- statpre, statpost, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(setattr, xdata, frame, op_ret, op_errno, statpre,
+ statpost, xdata);
+ return 0;
}
int
-pl_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+pl_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *stbuf,
+ int32_t valid, dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, loc, NULL);
- STACK_WIND (frame, pl_setattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, ((fd_t *)NULL), loc, NULL);
+ STACK_WIND(frame, pl_setattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid, xdata);
+ return 0;
}
int32_t
-pl_fsetattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *statpre,
- struct iatt *statpost, dict_t *xdata)
+pl_fsetattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
{
- PL_STACK_UNWIND (fsetattr, xdata, frame, op_ret, op_errno,
- statpre, statpost, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(fsetattr, xdata, frame, op_ret, op_errno,
+ statpre, statpost, xdata);
+ return 0;
}
int
-pl_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+pl_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
+ int32_t valid, dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, fd, NULL, NULL);
- STACK_WIND (frame, pl_fsetattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->fsetattr, fd, stbuf, valid, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
+ STACK_WIND(frame, pl_fsetattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetattr, fd, stbuf, valid, xdata);
+ return 0;
}
int32_t
-pl_fallocate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *pre,
- struct iatt *post, dict_t *xdata)
+pl_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *pre,
+ struct iatt *post, dict_t *xdata)
{
- PL_STACK_UNWIND (fallocate, xdata, frame, op_ret, op_errno,
- pre, post, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(fallocate, xdata, frame, op_ret, op_errno, pre,
+ post, xdata);
+ return 0;
}
int
-pl_fallocate (call_frame_t *frame, xlator_t *this,
- fd_t *fd, int32_t keep_size, off_t offset,
- size_t len, dict_t *xdata)
-{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, fd, NULL, NULL);
- STACK_WIND (frame, pl_fallocate_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->fallocate, fd,
- keep_size, offset, len, xdata);
- return 0;
+pl_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t keep_size,
+ off_t offset, size_t len, dict_t *xdata)
+{
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
+ STACK_WIND(frame, pl_fallocate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fallocate, fd, keep_size, offset, len,
+ xdata);
+ return 0;
}
int32_t
-pl_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, const char *path,
- struct iatt *buf, dict_t *xdata)
+pl_readlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, const char *path,
+ struct iatt *buf, dict_t *xdata)
{
- PL_STACK_UNWIND (readlink, xdata, frame, op_ret, op_errno,
- path, buf, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(readlink, xdata, frame, op_ret, op_errno, path,
+ buf, xdata);
+ return 0;
}
int
-pl_readlink (call_frame_t *frame, xlator_t *this,
- loc_t *loc, size_t size, dict_t *xdata)
+pl_readlink(call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+ dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, loc, NULL);
- STACK_WIND (frame, pl_readlink_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readlink, loc, size, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, ((fd_t *)NULL), loc, NULL);
+ STACK_WIND(frame, pl_readlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readlink, loc, size, xdata);
+ return 0;
}
int32_t
-pl_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+pl_access_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
{
- PL_STACK_UNWIND (access, xdata, frame, op_ret, op_errno, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(access, xdata, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-pl_access (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int32_t mask, dict_t *xdata)
+pl_access(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+ dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, NULL, loc, NULL);
- STACK_WIND (frame, pl_access_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->access, loc, mask, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, ((fd_t *)NULL), loc, NULL);
+ STACK_WIND(frame, pl_access_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->access, loc, mask, xdata);
+ return 0;
}
int32_t
-pl_seek_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, off_t offset,
- dict_t *xdata)
+pl_seek_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, off_t offset, dict_t *xdata)
{
- PL_STACK_UNWIND (seek, xdata, frame, op_ret, op_errno, offset, xdata);
- return 0;
+ PL_STACK_UNWIND_FOR_CLIENT(seek, xdata, frame, op_ret, op_errno, offset,
+ xdata);
+ return 0;
}
int32_t
-pl_seek (call_frame_t *frame, xlator_t *this, fd_t *fd,
- off_t offset, gf_seek_what_t what,
- dict_t *xdata)
+pl_seek(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ gf_seek_what_t what, dict_t *xdata)
{
- PL_LOCAL_GET_REQUESTS (frame, this, xdata, fd, NULL, NULL);
- STACK_WIND (frame, pl_seek_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->seek, fd, offset, what, xdata);
- return 0;
+ PL_LOCAL_GET_REQUESTS(frame, this, xdata, fd, NULL, NULL);
+ STACK_WIND(frame, pl_seek_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->seek, fd, offset, what, xdata);
+ return 0;
}
struct xlator_fops fops = {
- .lookup = pl_lookup,
- .create = pl_create,
- .fstat = pl_fstat,
- .truncate = pl_truncate,
- .ftruncate = pl_ftruncate,
- .discard = pl_discard,
- .zerofill = pl_zerofill,
- .open = pl_open,
- .readv = pl_readv,
- .writev = pl_writev,
- .lk = pl_lk,
- .inodelk = pl_inodelk,
- .finodelk = pl_finodelk,
- .entrylk = pl_entrylk,
- .fentrylk = pl_fentrylk,
- .flush = pl_flush,
- .opendir = pl_opendir,
- .readdirp = pl_readdirp,
- .setxattr = pl_setxattr,
- .fsetxattr = pl_fsetxattr,
- .getxattr = pl_getxattr,
- .fgetxattr = pl_fgetxattr,
- .removexattr = pl_removexattr,
- .fremovexattr = pl_fremovexattr,
- .rename = pl_rename,
- .getactivelk = pl_getactivelk,
- .setactivelk = pl_setactivelk,
- .unlink = pl_unlink,
- .access = pl_access,
- .readlink = pl_readlink,
- .fallocate = pl_fallocate,
- .fsetattr = pl_fsetattr,
- .setattr = pl_setattr,
- .fxattrop = pl_fxattrop,
- .xattrop = pl_xattrop,
- .rchecksum = pl_rchecksum,
- .statfs = pl_statfs,
- .fsyncdir = pl_fsyncdir,
- .readdir = pl_readdir,
- .symlink = pl_symlink,
- .link = pl_link,
- .rmdir = pl_rmdir,
- .mknod = pl_mknod,
- .stat = pl_stat,
- .seek = pl_seek,
+ .lookup = pl_lookup,
+ .create = pl_create,
+ .fstat = pl_fstat,
+ .truncate = pl_truncate,
+ .ftruncate = pl_ftruncate,
+ .discard = pl_discard,
+ .zerofill = pl_zerofill,
+ .open = pl_open,
+ .readv = pl_readv,
+ .writev = pl_writev,
+ .lk = pl_lk,
+ .inodelk = pl_inodelk,
+ .finodelk = pl_finodelk,
+ .entrylk = pl_entrylk,
+ .fentrylk = pl_fentrylk,
+ .flush = pl_flush,
+ .opendir = pl_opendir,
+ .readdirp = pl_readdirp,
+ .setxattr = pl_setxattr,
+ .fsetxattr = pl_fsetxattr,
+ .getxattr = pl_getxattr,
+ .fgetxattr = pl_fgetxattr,
+ .removexattr = pl_removexattr,
+ .fremovexattr = pl_fremovexattr,
+ .rename = pl_rename,
+ .getactivelk = pl_getactivelk,
+ .setactivelk = pl_setactivelk,
+ .unlink = pl_unlink,
+ .access = pl_access,
+ .readlink = pl_readlink,
+ .fallocate = pl_fallocate,
+ .fsetattr = pl_fsetattr,
+ .setattr = pl_setattr,
+ .fxattrop = pl_fxattrop,
+ .xattrop = pl_xattrop,
+ .rchecksum = pl_rchecksum,
+ .statfs = pl_statfs,
+ .fsyncdir = pl_fsyncdir,
+ .readdir = pl_readdir,
+ .symlink = pl_symlink,
+ .link = pl_link,
+ .rmdir = pl_rmdir,
+ .mknod = pl_mknod,
+ .stat = pl_stat,
+ .seek = pl_seek,
};
struct xlator_dumpops dumpops = {
- .inodectx = pl_dump_inode_priv,
+ .inodectx = pl_dump_inode_priv,
};
struct xlator_cbks cbks = {
- .forget = pl_forget,
- .release = pl_release,
- .releasedir = pl_releasedir,
- .client_destroy = pl_client_destroy_cbk,
- .client_disconnect = pl_client_disconnect_cbk,
+ .forget = pl_forget,
+ .release = pl_release,
+ .releasedir = pl_releasedir,
+ .client_destroy = pl_client_destroy_cbk,
+ .client_disconnect = pl_client_disconnect_cbk,
};
struct volume_options options[] = {
- { .key = { "mandatory-locking" },
- .type = GF_OPTION_TYPE_STR,
- .default_value = "off",
- .op_version = {GD_OP_VERSION_3_8_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"locks"},
- .description = "Specifies the mandatory-locking mode. Valid options "
- "are 'file' to use linux style mandatory locks, "
- "'forced' to use volume striclty under mandatory lock "
- "semantics only and 'optimal' to treat advisory and "
- "mandatory locks separately on their own."
- },
- { .key = { "trace" },
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .op_version = {GD_OP_VERSION_3_7_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"locks"},
- .description = "Trace the different lock requests "
- "to logs."
- },
- { .key = { "monkey-unlocking" },
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "false",
- .op_version = {GD_OP_VERSION_3_9_0},
- .flags = OPT_FLAG_SETTABLE,
- .tags = {"locks"},
- .description = "Ignore a random number of unlock requests. Useful "
- "for testing/creating robust lock recovery mechanisms."
- },
- { .key = {"revocation-secs"},
- .type = GF_OPTION_TYPE_INT,
- .min = 0,
- .max = INT_MAX,
- .default_value = "0",
- .op_version = {GD_OP_VERSION_3_9_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"locks"},
- .description = "Maximum time a lock can be taken out, before"
- "being revoked.",
- },
- { .key = {"revocation-clear-all"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "false",
- .op_version = {GD_OP_VERSION_3_9_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"locks"},
- .description = "If set to true, will revoke BOTH granted and blocked "
- "(pending) lock requests if a revocation threshold is "
- "hit.",
- },
- { .key = {"revocation-max-blocked"},
- .type = GF_OPTION_TYPE_INT,
- .min = 0,
- .max = INT_MAX,
- .default_value = "0",
- .op_version = {GD_OP_VERSION_3_9_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {"locks"},
- .description = "A number of blocked lock requests after which a lock "
- "will be revoked to allow the others to proceed. Can "
- "be used in conjunction w/ revocation-clear-all."
- },
- { .key = {"notify-contention"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "no",
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .op_version = { GD_OP_VERSION_4_0_0 },
- .tags = { "locks", "contention" },
- .description = "When this option is enabled and a lock request "
- "conflicts with a currently granted lock, an upcall "
- "notification will be sent to the current owner of "
- "the lock to request it to be released as soon as "
- "possible."
- },
- { .key = {"notify-contention-delay"},
- .type = GF_OPTION_TYPE_INT,
- .min = 0, /* An upcall notification is sent every time a conflict is
- * detected. */
- .max = 60,
- .default_value = "5",
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .op_version = { GD_OP_VERSION_4_0_0 },
- .tags = { "locks", "contention", "timeout" },
- .description = "This value determines the minimum amount of time "
- "(in seconds) between upcall contention notifications "
- "on the same inode. If multiple lock requests are "
- "received during this period, only one upcall will "
- "be sent."
- },
- { .key = {NULL} },
+ {.key = {"mandatory-locking"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = "off",
+ .op_version = {GD_OP_VERSION_3_8_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"locks"},
+ .description = "Specifies the mandatory-locking mode. Valid options "
+ "are 'file' to use linux style mandatory locks, "
+ "'forced' to use volume strictly under mandatory lock "
+ "semantics only and 'optimal' to treat advisory and "
+ "mandatory locks separately on their own."},
+ {.key = {"trace"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .op_version = {GD_OP_VERSION_3_7_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"locks"},
+ .description = "Trace the different lock requests "
+ "to logs."},
+ {.key = {"monkey-unlocking"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "false",
+ .op_version = {GD_OP_VERSION_3_9_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .tags = {"locks"},
+ .description = "Ignore a random number of unlock requests. Useful "
+ "for testing/creating robust lock recovery mechanisms."},
+ {
+ .key = {"revocation-secs"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max = INT_MAX,
+ .default_value = "0",
+ .op_version = {GD_OP_VERSION_3_9_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"locks"},
+ .description = "Maximum time a lock can be taken out, before"
+ "being revoked.",
+ },
+ {
+ .key = {"revocation-clear-all"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "false",
+ .op_version = {GD_OP_VERSION_3_9_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"locks"},
+ .description = "If set to true, will revoke BOTH granted and blocked "
+ "(pending) lock requests if a revocation threshold is "
+ "hit.",
+ },
+ {.key = {"revocation-max-blocked"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0,
+ .max = INT_MAX,
+ .default_value = "0",
+ .op_version = {GD_OP_VERSION_3_9_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {"locks"},
+ .description = "A number of blocked lock requests after which a lock "
+ "will be revoked to allow the others to proceed. Can "
+ "be used in conjunction w/ revocation-clear-all."},
+ {.key = {"notify-contention"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "yes",
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .op_version = {GD_OP_VERSION_4_0_0},
+ .tags = {"locks", "contention"},
+ .description = "When this option is enabled and a lock request "
+ "conflicts with a currently granted lock, an upcall "
+ "notification will be sent to the current owner of "
+ "the lock to request it to be released as soon as "
+ "possible."},
+ {.key = {"notify-contention-delay"},
+ .type = GF_OPTION_TYPE_INT,
+ .min = 0, /* An upcall notification is sent every time a conflict is
+ * detected. */
+ .max = 60,
+ .default_value = "5",
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .op_version = {GD_OP_VERSION_4_0_0},
+ .tags = {"locks", "contention", "timeout"},
+ .description = "This value determines the minimum amount of time "
+ "(in seconds) between upcall contention notifications "
+ "on the same inode. If multiple lock requests are "
+ "received during this period, only one upcall will "
+ "be sent."},
+ {.key = {"enforce-mandatory-lock"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .flags = OPT_FLAG_SETTABLE,
+ .op_version = {GD_OP_VERSION_6_0},
+ .description = "option to enable lock enforcement"},
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .dumpops = &dumpops,
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "locks",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/features/locks/src/reservelk.c b/xlators/features/locks/src/reservelk.c
index 5c7a5b47f3e..604691fd887 100644
--- a/xlators/features/locks/src/reservelk.c
+++ b/xlators/features/locks/src/reservelk.c
@@ -7,419 +7,376 @@
later), or the GNU General Public License, version 2 (GPLv2), in all
cases as published by the Free Software Foundation.
*/
-#include "glusterfs.h"
-#include "compat.h"
-#include "xlator.h"
-#include "logging.h"
-#include "common-utils.h"
-#include "list.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/compat.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/common-utils.h>
+#include <glusterfs/list.h>
#include "locks.h"
#include "common.h"
/* Return true if the two reservelks have exactly same lock boundaries */
int
-reservelks_equal (posix_lock_t *l1, posix_lock_t *l2)
+reservelks_equal(posix_lock_t *l1, posix_lock_t *l2)
{
- if ((l1->fl_start == l2->fl_start) &&
- (l1->fl_end == l2->fl_end))
- return 1;
+ if ((l1->fl_start == l2->fl_start) && (l1->fl_end == l2->fl_end))
+ return 1;
- return 0;
+ return 0;
}
/* Determine if lock is grantable or not */
static posix_lock_t *
-__reservelk_grantable (pl_inode_t *pl_inode, posix_lock_t *lock)
+__reservelk_grantable(pl_inode_t *pl_inode, posix_lock_t *lock)
{
- xlator_t *this = NULL;
- posix_lock_t *l = NULL;
- posix_lock_t *ret_lock = NULL;
-
- this = THIS;
-
- if (list_empty (&pl_inode->reservelk_list)) {
- gf_log (this->name, GF_LOG_TRACE,
- "No reservelks in list");
- goto out;
- }
- list_for_each_entry (l, &pl_inode->reservelk_list, list){
- if (reservelks_equal (lock, l)) {
- ret_lock = l;
- break;
- }
+ xlator_t *this = THIS;
+ posix_lock_t *l = NULL;
+ posix_lock_t *ret_lock = NULL;
+
+ if (list_empty(&pl_inode->reservelk_list)) {
+ gf_log(this->name, GF_LOG_TRACE, "No reservelks in list");
+ goto out;
+ }
+ list_for_each_entry(l, &pl_inode->reservelk_list, list)
+ {
+ if (reservelks_equal(lock, l)) {
+ ret_lock = l;
+ break;
}
+ }
out:
- return ret_lock;
+ return ret_lock;
}
static int
-__same_owner_reservelk (posix_lock_t *l1, posix_lock_t *l2)
+__same_owner_reservelk(posix_lock_t *l1, posix_lock_t *l2)
{
- return (is_same_lkowner (&l1->owner, &l2->owner));
-
+ return (is_same_lkowner(&l1->owner, &l2->owner));
}
static posix_lock_t *
-__matching_reservelk (pl_inode_t *pl_inode, posix_lock_t *lock)
+__matching_reservelk(pl_inode_t *pl_inode, posix_lock_t *lock)
{
- posix_lock_t *l = NULL;
+ posix_lock_t *l = NULL;
- if (list_empty (&pl_inode->reservelk_list)) {
- gf_log ("posix-locks", GF_LOG_TRACE,
- "reservelk list empty");
- return NULL;
- }
+ if (list_empty(&pl_inode->reservelk_list)) {
+ gf_log("posix-locks", GF_LOG_TRACE, "reservelk list empty");
+ return NULL;
+ }
- list_for_each_entry (l, &pl_inode->reservelk_list, list) {
- if (reservelks_equal (l, lock)) {
- gf_log ("posix-locks", GF_LOG_TRACE,
- "equal reservelk found");
- break;
- }
+ list_for_each_entry(l, &pl_inode->reservelk_list, list)
+ {
+ if (reservelks_equal(l, lock)) {
+ gf_log("posix-locks", GF_LOG_TRACE, "equal reservelk found");
+ break;
}
+ }
- return l;
+ return l;
}
static int
-__reservelk_conflict (xlator_t *this, pl_inode_t *pl_inode,
- posix_lock_t *lock)
+__reservelk_conflict(xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock)
{
- posix_lock_t *conf = NULL;
- int ret = 0;
-
- conf = __matching_reservelk (pl_inode, lock);
- if (conf) {
- gf_log (this->name, GF_LOG_TRACE,
- "Matching reservelk found");
- if (__same_owner_reservelk (lock, conf)) {
- list_del_init (&conf->list);
- gf_log (this->name, GF_LOG_TRACE,
- "Removing the matching reservelk for setlk to progress");
- __destroy_lock(conf);
- ret = 0;
- } else {
- gf_log (this->name, GF_LOG_TRACE,
- "Conflicting reservelk found");
- ret = 1;
- }
-
+ int ret = 0;
+
+ posix_lock_t *conf = __matching_reservelk(pl_inode, lock);
+ if (conf) {
+ gf_log(this->name, GF_LOG_TRACE, "Matching reservelk found");
+ if (__same_owner_reservelk(lock, conf)) {
+ list_del_init(&conf->list);
+ gf_log(this->name, GF_LOG_TRACE,
+ "Removing the matching reservelk for setlk to progress");
+ __destroy_lock(conf);
+ ret = 0;
+ } else {
+ gf_log(this->name, GF_LOG_TRACE, "Conflicting reservelk found");
+ ret = 1;
}
- return ret;
-
+ }
+ return ret;
}
int
-pl_verify_reservelk (xlator_t *this, pl_inode_t *pl_inode,
- posix_lock_t *lock, int can_block)
+pl_verify_reservelk(xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
+ const int can_block)
{
- int ret = 0;
-
- pthread_mutex_lock (&pl_inode->mutex);
- {
- if (__reservelk_conflict (this, pl_inode, lock)) {
- gf_log (this->name, GF_LOG_TRACE,
- "Found conflicting reservelk. Blocking until reservelk is unlocked.");
- lock->blocked = can_block;
- list_add_tail (&lock->list, &pl_inode->blocked_calls);
- ret = -1;
- goto unlock;
- }
-
- gf_log (this->name, GF_LOG_TRACE,
- "no conflicting reservelk found. Call continuing");
- ret = 0;
-
+ int ret = 0;
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ if (__reservelk_conflict(this, pl_inode, lock)) {
+ lock->blocked = can_block;
+ list_add_tail(&lock->list, &pl_inode->blocked_calls);
+ pthread_mutex_unlock(&pl_inode->mutex);
+ gf_log(this->name, GF_LOG_TRACE,
+ "Found conflicting reservelk. Blocking until reservelk is "
+ "unlocked.");
+ ret = -1;
+ goto out;
}
-unlock:
- pthread_mutex_unlock (&pl_inode->mutex);
-
- return ret;
-
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+ gf_log(this->name, GF_LOG_TRACE,
+ "no conflicting reservelk found. Call continuing");
+ ret = 0;
+out:
+ return ret;
}
-
/* Determines if lock can be granted and adds the lock. If the lock
* is blocking, adds it to the blocked_reservelks.
*/
static int
-__lock_reservelk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
- int can_block)
+__lock_reservelk(xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
+ const int can_block)
{
- posix_lock_t *conf = NULL;
- int ret = -EINVAL;
-
- conf = __reservelk_grantable (pl_inode, lock);
- if (conf){
- ret = -EAGAIN;
- if (can_block == 0)
- goto out;
+ int ret = -EINVAL;
- list_add_tail (&lock->list, &pl_inode->blocked_reservelks);
+ posix_lock_t *conf = __reservelk_grantable(pl_inode, lock);
+ if (conf) {
+ ret = -EAGAIN;
+ if (can_block == 0)
+ goto out;
- gf_log (this->name, GF_LOG_TRACE,
- "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => Blocked",
- lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
- lock->client_pid,
- lkowner_utoa (&lock->owner),
- lock->user_flock.l_start,
- lock->user_flock.l_len);
+ list_add_tail(&lock->list, &pl_inode->blocked_reservelks);
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s (pid=%d) lk-owner:%s %" PRId64 " - %" PRId64 " => Blocked",
+ lock->fl_type == F_UNLCK ? "Unlock" : "Lock", lock->client_pid,
+ lkowner_utoa(&lock->owner), lock->user_flock.l_start,
+ lock->user_flock.l_len);
- goto out;
- }
+ goto out;
+ }
- list_add (&lock->list, &pl_inode->reservelk_list);
+ list_add(&lock->list, &pl_inode->reservelk_list);
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
static posix_lock_t *
-find_matching_reservelk (posix_lock_t *lock, pl_inode_t *pl_inode)
+find_matching_reservelk(posix_lock_t *lock, pl_inode_t *pl_inode)
{
- posix_lock_t *l = NULL;
- list_for_each_entry (l, &pl_inode->reservelk_list, list) {
- if (reservelks_equal (l, lock))
- return l;
- }
- return NULL;
+ posix_lock_t *l = NULL;
+ list_for_each_entry(l, &pl_inode->reservelk_list, list)
+ {
+ if (reservelks_equal(l, lock))
+ return l;
+ }
+ return NULL;
}
/* Set F_UNLCK removes a lock which has the exact same lock boundaries
* as the UNLCK lock specifies. If such a lock is not found, returns invalid
*/
static posix_lock_t *
-__reserve_unlock_lock (xlator_t *this, posix_lock_t *lock, pl_inode_t *pl_inode)
+__reserve_unlock_lock(xlator_t *this, posix_lock_t *lock, pl_inode_t *pl_inode)
{
-
- posix_lock_t *conf = NULL;
-
- conf = find_matching_reservelk (lock, pl_inode);
- if (!conf) {
- gf_log (this->name, GF_LOG_DEBUG,
- " Matching lock not found for unlock");
- goto out;
- }
- __delete_lock(conf);
- gf_log (this->name, GF_LOG_DEBUG,
- " Matching lock found for unlock");
+ posix_lock_t *conf = find_matching_reservelk(lock, pl_inode);
+ if (!conf) {
+ gf_log(this->name, GF_LOG_DEBUG, " Matching lock not found for unlock");
+ goto out;
+ }
+ __delete_lock(conf);
+ gf_log(this->name, GF_LOG_DEBUG, " Matching lock found for unlock");
out:
- return conf;
-
-
+ return conf;
}
static void
-__grant_blocked_reserve_locks (xlator_t *this, pl_inode_t *pl_inode,
- struct list_head *granted)
+__grant_blocked_reserve_locks(xlator_t *this, pl_inode_t *pl_inode,
+ struct list_head *granted)
{
- int bl_ret = 0;
- posix_lock_t *bl = NULL;
- posix_lock_t *tmp = NULL;
-
- struct list_head blocked_list;
+ int bl_ret = 0;
+ posix_lock_t *bl = NULL;
+ posix_lock_t *tmp = NULL;
- INIT_LIST_HEAD (&blocked_list);
- list_splice_init (&pl_inode->blocked_reservelks, &blocked_list);
+ struct list_head blocked_list;
- list_for_each_entry_safe (bl, tmp, &blocked_list, list) {
+ INIT_LIST_HEAD(&blocked_list);
+ list_splice_init(&pl_inode->blocked_reservelks, &blocked_list);
- list_del_init (&bl->list);
+ list_for_each_entry_safe(bl, tmp, &blocked_list, list)
+ {
+ list_del_init(&bl->list);
- bl_ret = __lock_reservelk (this, pl_inode, bl, 1);
+ bl_ret = __lock_reservelk(this, pl_inode, bl, 1);
- if (bl_ret == 0) {
- list_add (&bl->list, granted);
- }
+ if (bl_ret == 0) {
+ list_add(&bl->list, granted);
}
- return;
+ }
+ return;
}
/* Grant all reservelks blocked on lock(s) */
void
-grant_blocked_reserve_locks (xlator_t *this, pl_inode_t *pl_inode)
+grant_blocked_reserve_locks(xlator_t *this, pl_inode_t *pl_inode)
{
- struct list_head granted;
- posix_lock_t *lock = NULL;
- posix_lock_t *tmp = NULL;
+ struct list_head granted;
+ posix_lock_t *lock = NULL;
+ posix_lock_t *tmp = NULL;
- INIT_LIST_HEAD (&granted);
-
- if (list_empty (&pl_inode->blocked_reservelks)) {
- gf_log (this->name, GF_LOG_TRACE,
- "No blocked locks to be granted");
- return;
- }
-
- pthread_mutex_lock (&pl_inode->mutex);
- {
- __grant_blocked_reserve_locks (this, pl_inode, &granted);
- }
- pthread_mutex_unlock (&pl_inode->mutex);
-
- list_for_each_entry_safe (lock, tmp, &granted, list) {
- gf_log (this->name, GF_LOG_TRACE,
- "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => Granted",
- lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
- lock->client_pid,
- lkowner_utoa (&lock->owner),
- lock->user_flock.l_start,
- lock->user_flock.l_len);
-
- STACK_UNWIND_STRICT (lk, lock->frame, 0, 0, &lock->user_flock,
- NULL);
- }
+ INIT_LIST_HEAD(&granted);
+ if (list_empty(&pl_inode->blocked_reservelks)) {
+ gf_log(this->name, GF_LOG_TRACE, "No blocked locks to be granted");
+ return;
+ }
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ __grant_blocked_reserve_locks(this, pl_inode, &granted);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+
+ list_for_each_entry_safe(lock, tmp, &granted, list)
+ {
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s (pid=%d) (lk-owner=%s) %" PRId64 " - %" PRId64 " => Granted",
+ lock->fl_type == F_UNLCK ? "Unlock" : "Lock", lock->client_pid,
+ lkowner_utoa(&lock->owner), lock->user_flock.l_start,
+ lock->user_flock.l_len);
+
+ STACK_UNWIND_STRICT(lk, lock->frame, 0, 0, &lock->user_flock, NULL);
+ }
}
static void
-__grant_blocked_lock_calls (xlator_t *this, pl_inode_t *pl_inode,
- struct list_head *granted)
+__grant_blocked_lock_calls(xlator_t *this, pl_inode_t *pl_inode,
+ struct list_head *granted)
{
- int bl_ret = 0;
- posix_lock_t *bl = NULL;
- posix_lock_t *tmp = NULL;
+ int bl_ret = 0;
+ posix_lock_t *bl = NULL;
+ posix_lock_t *tmp = NULL;
- struct list_head blocked_list;
+ struct list_head blocked_list;
- INIT_LIST_HEAD (&blocked_list);
- list_splice_init (&pl_inode->blocked_reservelks, &blocked_list);
+ INIT_LIST_HEAD(&blocked_list);
+ list_splice_init(&pl_inode->blocked_reservelks, &blocked_list);
- list_for_each_entry_safe (bl, tmp, &blocked_list, list) {
+ list_for_each_entry_safe(bl, tmp, &blocked_list, list)
+ {
+ list_del_init(&bl->list);
- list_del_init (&bl->list);
+ bl_ret = pl_verify_reservelk(this, pl_inode, bl, bl->blocked);
- bl_ret = pl_verify_reservelk (this, pl_inode, bl, bl->blocked);
-
- if (bl_ret == 0) {
- list_add_tail (&bl->list, granted);
- }
+ if (bl_ret == 0) {
+ list_add_tail(&bl->list, granted);
}
- return;
+ }
+ return;
}
void
-grant_blocked_lock_calls (xlator_t *this, pl_inode_t *pl_inode)
+grant_blocked_lock_calls(xlator_t *this, pl_inode_t *pl_inode)
{
- struct list_head granted;
- posix_lock_t *lock = NULL;
- posix_lock_t *tmp = NULL;
- fd_t *fd = NULL;
-
- int can_block = 0;
- int32_t cmd = 0;
- int ret = 0;
-
- if (list_empty (&pl_inode->blocked_calls)) {
- gf_log (this->name, GF_LOG_TRACE,
- "No blocked lock calls to be granted");
- return;
- }
+ struct list_head granted;
+ posix_lock_t *lock = NULL;
+ posix_lock_t *tmp = NULL;
+ fd_t *fd = NULL;
- pthread_mutex_lock (&pl_inode->mutex);
- {
- __grant_blocked_lock_calls (this, pl_inode, &granted);
- }
- pthread_mutex_unlock (&pl_inode->mutex);
-
- list_for_each_entry_safe (lock, tmp, &granted, list) {
- fd = fd_from_fdnum (lock);
-
- if (lock->blocked) {
- can_block = 1;
- cmd = F_SETLKW;
- }
- else
- cmd = F_SETLK;
-
- lock->blocked = 0;
- ret = pl_setlk (this, pl_inode, lock, can_block);
- if (ret == -1) {
- if (can_block) {
- pl_trace_block (this, lock->frame, fd, NULL,
- cmd, &lock->user_flock, NULL);
- continue;
- } else {
- gf_log (this->name, GF_LOG_DEBUG, "returning EAGAIN");
- pl_trace_out (this, lock->frame, fd, NULL, cmd,
- &lock->user_flock, -1, EAGAIN, NULL);
- pl_update_refkeeper (this, fd->inode);
- STACK_UNWIND_STRICT (lk, lock->frame, -1,
- EAGAIN, &lock->user_flock,
- NULL);
- __destroy_lock (lock);
- }
- }
+ int can_block = 0;
+ int32_t cmd = 0;
+ int ret = 0;
+ if (list_empty(&pl_inode->blocked_calls)) {
+ gf_log(this->name, GF_LOG_TRACE, "No blocked lock calls to be granted");
+ return;
+ }
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ __grant_blocked_lock_calls(this, pl_inode, &granted);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+
+ list_for_each_entry_safe(lock, tmp, &granted, list)
+ {
+ fd = fd_from_fdnum(lock);
+
+ if (lock->blocked) {
+ can_block = 1;
+ cmd = F_SETLKW;
+ } else
+ cmd = F_SETLK;
+
+ lock->blocked = 0;
+ ret = pl_setlk(this, pl_inode, lock, can_block);
+ if (ret == -1) {
+ if (can_block) {
+ continue;
+ } else {
+ gf_log(this->name, GF_LOG_DEBUG, "returning EAGAIN");
+ pl_trace_out(this, lock->frame, fd, NULL, cmd,
+ &lock->user_flock, -1, EAGAIN, NULL);
+ pl_update_refkeeper(this, fd->inode);
+ STACK_UNWIND_STRICT(lk, lock->frame, -1, EAGAIN,
+ &lock->user_flock, NULL);
+ __destroy_lock(lock);
+ }
}
-
+ }
}
-
int
-pl_reserve_unlock (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock)
+pl_reserve_unlock(xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock)
{
- posix_lock_t *retlock = NULL;
- int ret = -1;
-
- pthread_mutex_lock (&pl_inode->mutex);
- {
- retlock = __reserve_unlock_lock (this, lock, pl_inode);
- if (!retlock) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Bad Unlock issued on Inode lock");
- ret = -EINVAL;
- goto out;
- }
-
- gf_log (this->name, GF_LOG_TRACE,
- "Reservelk Unlock successful");
- __destroy_lock(retlock);
- ret = 0;
+ posix_lock_t *retlock = NULL;
+ int ret = -1;
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ retlock = __reserve_unlock_lock(this, lock, pl_inode);
+ if (!retlock) {
+ pthread_mutex_unlock(&pl_inode->mutex);
+ gf_log(this->name, GF_LOG_DEBUG, "Bad Unlock issued on Inode lock");
+ ret = -EINVAL;
+ goto out;
}
-out:
- pthread_mutex_unlock (&pl_inode->mutex);
- grant_blocked_reserve_locks (this, pl_inode);
- grant_blocked_lock_calls (this, pl_inode);
-
- return ret;
+ gf_log(this->name, GF_LOG_TRACE, "Reservelk Unlock successful");
+ __destroy_lock(retlock);
+ ret = 0;
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+out:
+ grant_blocked_reserve_locks(this, pl_inode);
+ grant_blocked_lock_calls(this, pl_inode);
+ return ret;
}
int
-pl_reserve_setlk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
- int can_block)
+pl_reserve_setlk(xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
+ int can_block)
{
- int ret = -EINVAL;
-
- pthread_mutex_lock (&pl_inode->mutex);
- {
-
- ret = __lock_reservelk (this, pl_inode, lock, can_block);
- if (ret < 0)
- gf_log (this->name, GF_LOG_TRACE,
- "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => NOK",
- lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
- lock->client_pid,
- lkowner_utoa (&lock->owner),
- lock->user_flock.l_start,
- lock->user_flock.l_len);
- else
- gf_log (this->name, GF_LOG_TRACE,
- "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => OK",
- lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
- lock->client_pid,
- lkowner_utoa (&lock->owner),
- lock->fl_start,
- lock->fl_end);
-
- }
- pthread_mutex_unlock (&pl_inode->mutex);
- return ret;
+ int ret = -EINVAL;
+
+ pthread_mutex_lock(&pl_inode->mutex);
+ {
+ ret = __lock_reservelk(this, pl_inode, lock, can_block);
+ }
+ pthread_mutex_unlock(&pl_inode->mutex);
+
+ if (ret < 0)
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s (pid=%d) (lk-owner=%s) %" PRId64 " - %" PRId64 " => NOK",
+ lock->fl_type == F_UNLCK ? "Unlock" : "Lock", lock->client_pid,
+ lkowner_utoa(&lock->owner), lock->user_flock.l_start,
+ lock->user_flock.l_len);
+ else
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s (pid=%d) (lk-owner=%s) %" PRId64 " - %" PRId64 " => OK",
+ lock->fl_type == F_UNLCK ? "Unlock" : "Lock", lock->client_pid,
+ lkowner_utoa(&lock->owner), lock->fl_start, lock->fl_end);
+
+ return ret;
}
diff --git a/xlators/features/locks/tests/unit-test.c b/xlators/features/locks/tests/unit-test.c
index 9468f9e1865..d285b12b5aa 100644
--- a/xlators/features/locks/tests/unit-test.c
+++ b/xlators/features/locks/tests/unit-test.c
@@ -7,53 +7,71 @@
later), or the GNU General Public License, version 2 (GPLv2), in all
cases as published by the Free Software Foundation.
*/
-#include "glusterfs.h"
-#include "compat.h"
-#include "xlator.h"
-#include "logging.h"
-#include "common-utils.h"
-#include "list.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/compat.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/common-utils.h>
+#include <glusterfs/list.h>
#include "locks.h"
#include "common.h"
-#define expect(cond) if (!(cond)) { goto out; }
+#define expect(cond) \
+ if (!(cond)) { \
+ goto out; \
+ }
-extern int lock_name (pl_inode_t *, const char *, entrylk_type);
-extern int unlock_name (pl_inode_t *, const char *, entrylk_type);
+extern int
+lock_name(pl_inode_t *, const char *, entrylk_type);
+extern int
+unlock_name(pl_inode_t *, const char *, entrylk_type);
-int main (int argc, char **argv)
+int
+main(int argc, char **argv)
{
- int ret = 1;
- int r = -1;
-
- pl_inode_t *pinode = CALLOC (sizeof (pl_inode_t), 1);
- pthread_mutex_init (&pinode->dir_lock_mutex, NULL);
- INIT_LIST_HEAD (&pinode->gf_dir_locks);
-
- r = lock_name (pinode, NULL, ENTRYLK_WRLCK); expect (r == 0);
- {
- r = lock_name (pinode, "foo", ENTRYLK_WRLCK); expect (r == -EAGAIN);
- }
- r = unlock_name (pinode, NULL, ENTRYLK_WRLCK); expect (r == 0);
-
- r = lock_name (pinode, "foo", ENTRYLK_RDLCK); expect (r == 0);
- {
- r = lock_name (pinode, "foo", ENTRYLK_RDLCK); expect (r == 0);
- {
- r = lock_name (pinode, "foo", ENTRYLK_WRLCK); expect (r == -EAGAIN);
- }
- r = unlock_name (pinode, "foo", ENTRYLK_RDLCK); expect (r == 0);
- }
- r = unlock_name (pinode, "foo", ENTRYLK_RDLCK); expect (r == 0);
-
- r = lock_name (pinode, "foo", ENTRYLK_WRLCK); expect (r == 0);
- r = unlock_name (pinode, "foo", ENTRYLK_WRLCK); expect (r == 0);
-
- r = lock_name (pinode, "baz", ENTRYLK_WRLCK); expect (r == 0);
- r = lock_name (pinode, "baz", ENTRYLK_RDLCK); expect (r == -EAGAIN);
-
- ret = 0;
+ int ret = 1;
+ int r = -1;
+
+ pl_inode_t *pinode = CALLOC(sizeof(pl_inode_t), 1);
+ pthread_mutex_init(&pinode->dir_lock_mutex, NULL);
+ INIT_LIST_HEAD(&pinode->gf_dir_locks);
+
+ r = lock_name(pinode, NULL, ENTRYLK_WRLCK);
+ expect(r == 0);
+ {
+ r = lock_name(pinode, "foo", ENTRYLK_WRLCK);
+ expect(r == -EAGAIN);
+ }
+ r = unlock_name(pinode, NULL, ENTRYLK_WRLCK);
+ expect(r == 0);
+
+ r = lock_name(pinode, "foo", ENTRYLK_RDLCK);
+ expect(r == 0);
+ {
+ r = lock_name(pinode, "foo", ENTRYLK_RDLCK);
+ expect(r == 0);
+ {
+ r = lock_name(pinode, "foo", ENTRYLK_WRLCK);
+ expect(r == -EAGAIN);
+ }
+ r = unlock_name(pinode, "foo", ENTRYLK_RDLCK);
+ expect(r == 0);
+ }
+ r = unlock_name(pinode, "foo", ENTRYLK_RDLCK);
+ expect(r == 0);
+
+ r = lock_name(pinode, "foo", ENTRYLK_WRLCK);
+ expect(r == 0);
+ r = unlock_name(pinode, "foo", ENTRYLK_WRLCK);
+ expect(r == 0);
+
+ r = lock_name(pinode, "baz", ENTRYLK_WRLCK);
+ expect(r == 0);
+ r = lock_name(pinode, "baz", ENTRYLK_RDLCK);
+ expect(r == -EAGAIN);
+
+ ret = 0;
out:
- return ret;
+ return ret;
}
diff --git a/xlators/features/marker/src/marker-common.c b/xlators/features/marker/src/marker-common.c
index 6ec5e3cc8bc..9c9047005d6 100644
--- a/xlators/features/marker/src/marker-common.c
+++ b/xlators/features/marker/src/marker-common.c
@@ -11,55 +11,47 @@
#include "marker-common.h"
marker_inode_ctx_t *
-marker_inode_ctx_new ()
+marker_inode_ctx_new()
{
- marker_inode_ctx_t *ctx = NULL;
+ marker_inode_ctx_t *ctx = NULL;
- ctx = GF_CALLOC (1, sizeof (marker_inode_ctx_t),
- gf_marker_mt_marker_inode_ctx_t);
- if (ctx == NULL)
- goto out;
+ ctx = GF_CALLOC(1, sizeof(marker_inode_ctx_t),
+ gf_marker_mt_marker_inode_ctx_t);
+ if (ctx == NULL)
+ goto out;
- ctx->quota_ctx = NULL;
+ ctx->quota_ctx = NULL;
out:
- return ctx;
+ return ctx;
}
int32_t
-marker_force_inode_ctx_get (inode_t *inode, xlator_t *this,
- marker_inode_ctx_t **ctx)
+marker_force_inode_ctx_get(inode_t *inode, xlator_t *this,
+ marker_inode_ctx_t **ctx)
{
- int32_t ret = -1;
- uint64_t ctx_int = 0;
-
- LOCK (&inode->lock);
- {
- ret = __inode_ctx_get (inode, this, &ctx_int);
- if (ret == 0)
- *ctx = (marker_inode_ctx_t *) (unsigned long)ctx_int;
- else {
- *ctx = marker_inode_ctx_new ();
- if (*ctx == NULL)
- goto unlock;
-
- ret = __inode_ctx_put (inode, this,
- (uint64_t )(unsigned long) *ctx);
- if (ret == -1) {
- GF_FREE (*ctx);
- goto unlock;
- }
- ret = 0;
- }
+ int32_t ret = -1;
+ uint64_t ctx_int = 0;
+
+ LOCK(&inode->lock);
+ {
+ ret = __inode_ctx_get(inode, this, &ctx_int);
+ if (ret == 0)
+ *ctx = (marker_inode_ctx_t *)(unsigned long)ctx_int;
+ else {
+ *ctx = marker_inode_ctx_new();
+ if (*ctx == NULL)
+ goto unlock;
+
+ ret = __inode_ctx_put(inode, this, (uint64_t)(unsigned long)*ctx);
+ if (ret == -1) {
+ GF_FREE(*ctx);
+ goto unlock;
+ }
+ ret = 0;
}
-unlock: UNLOCK (&inode->lock);
+ }
+unlock:
+ UNLOCK(&inode->lock);
- return ret;
-}
-
-int
-marker_filter_quota_xattr (dict_t *dict, char *key,
- data_t *value, void *data)
-{
- dict_del (dict, key);
- return 0;
+ return ret;
}
diff --git a/xlators/features/marker/src/marker-common.h b/xlators/features/marker/src/marker-common.h
index fbe9a2ce4b0..7f8cffe7d35 100644
--- a/xlators/features/marker/src/marker-common.h
+++ b/xlators/features/marker/src/marker-common.h
@@ -10,12 +10,10 @@
#ifndef _MARKER_COMMON_H
#define _MARKER_COMMON_H
-#include "xlator.h"
+#include <glusterfs/xlator.h>
#include "marker.h"
int32_t
-marker_force_inode_ctx_get (inode_t *, xlator_t *, marker_inode_ctx_t **);
+marker_force_inode_ctx_get(inode_t *, xlator_t *, marker_inode_ctx_t **);
-int
-marker_filter_quota_xattr (dict_t *, char *, data_t *, void *);
#endif
diff --git a/xlators/features/marker/src/marker-mem-types.h b/xlators/features/marker/src/marker-mem-types.h
index dc5ad16ed76..aedfdb4a1b7 100644
--- a/xlators/features/marker/src/marker-mem-types.h
+++ b/xlators/features/marker/src/marker-mem-types.h
@@ -10,18 +10,19 @@
#ifndef __MARKER_MEM_TYPES_H__
#define __MARKER_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_marker_mem_types_ {
- gf_marker_mt_marker_conf_t = gf_common_mt_end + 1,
- gf_marker_mt_loc_t,
- gf_marker_mt_volume_mark,
- gf_marker_mt_int64_t,
- gf_marker_mt_quota_inode_ctx_t,
- gf_marker_mt_marker_inode_ctx_t,
- gf_marker_mt_inode_contribution_t,
- gf_marker_mt_quota_meta_t,
- gf_marker_mt_quota_synctask_t,
- gf_marker_mt_end
+ /* Those are used by ALLOCATE_OR_GOTO macro */
+ gf_marker_mt_marker_conf_t = gf_common_mt_end + 1,
+ gf_marker_mt_loc_t,
+ gf_marker_mt_volume_mark,
+ gf_marker_mt_int64_t,
+ gf_marker_mt_quota_inode_ctx_t,
+ gf_marker_mt_marker_inode_ctx_t,
+ gf_marker_mt_inode_contribution_t,
+ gf_marker_mt_quota_meta_t,
+ gf_marker_mt_quota_synctask_t,
+ gf_marker_mt_end
};
#endif
diff --git a/xlators/features/marker/src/marker-quota-helper.c b/xlators/features/marker/src/marker-quota-helper.c
index e3e218f8af5..ecd85d67b2b 100644
--- a/xlators/features/marker/src/marker-quota-helper.c
+++ b/xlators/features/marker/src/marker-quota-helper.c
@@ -7,473 +7,374 @@
later), or the GNU General Public License, version 2 (GPLv2), in all
cases as published by the Free Software Foundation.
*/
-#include "locking.h"
+#include <glusterfs/locking.h>
#include "marker-quota.h"
#include "marker-common.h"
#include "marker-quota-helper.h"
#include "marker-mem-types.h"
int
-mq_loc_fill (loc_t *loc, inode_t *inode, inode_t *parent, char *path)
+mq_loc_fill(loc_t *loc, inode_t *inode, inode_t *parent, char *path)
{
- int ret = -1;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
- GF_VALIDATE_OR_GOTO ("marker", inode, out);
- GF_VALIDATE_OR_GOTO ("marker", path, out);
- /* Not checking for parent because while filling
- * loc of root, parent will be NULL
- */
+ GF_VALIDATE_OR_GOTO("marker", loc, out);
+ GF_VALIDATE_OR_GOTO("marker", inode, out);
+ GF_VALIDATE_OR_GOTO("marker", path, out);
+ /* Not checking for parent because while filling
+ * loc of root, parent will be NULL
+ */
- if (inode) {
- loc->inode = inode_ref (inode);
- }
+ if (inode) {
+ loc->inode = inode_ref(inode);
+ }
- if (parent)
- loc->parent = inode_ref (parent);
+ if (parent)
+ loc->parent = inode_ref(parent);
- if (!gf_uuid_is_null (inode->gfid))
- gf_uuid_copy (loc->gfid, inode->gfid);
+ if (!gf_uuid_is_null(inode->gfid))
+ gf_uuid_copy(loc->gfid, inode->gfid);
- loc->path = gf_strdup (path);
- if (!loc->path) {
- gf_log ("loc fill", GF_LOG_ERROR, "strdup failed");
- goto out;
- }
+ loc->path = gf_strdup(path);
+ if (!loc->path) {
+ gf_log("loc fill", GF_LOG_ERROR, "strdup failed");
+ goto out;
+ }
- loc->name = strrchr (loc->path, '/');
- if (loc->name)
- loc->name++;
- else
- goto out;
+ loc->name = strrchr(loc->path, '/');
+ if (loc->name)
+ loc->name++;
+ else
+ goto out;
- ret = 0;
+ ret = 0;
out:
- if (ret < 0)
- loc_wipe (loc);
+ if (ret < 0)
+ loc_wipe(loc);
- return ret;
+ return ret;
}
-
int32_t
-mq_inode_loc_fill (const char *parent_gfid, inode_t *inode, loc_t *loc)
+mq_inode_loc_fill(const char *parent_gfid, inode_t *inode, loc_t *loc)
{
- char *resolvedpath = NULL;
- inode_t *parent = NULL;
- quota_inode_ctx_t *ctx = NULL;
- xlator_t *this = NULL;
- int ret = -1;
-
- this = THIS;
-
- if (inode == NULL) {
- gf_log_callingfn ("marker", GF_LOG_ERROR, "loc fill failed, "
- "inode is NULL");
- return ret;
- }
+ char *resolvedpath = NULL;
+ inode_t *parent = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ xlator_t *this = NULL;
+ int ret = -1;
+
+ this = THIS;
+
+ if (inode == NULL) {
+ gf_log_callingfn("marker", GF_LOG_ERROR,
+ "loc fill failed, "
+ "inode is NULL");
+ return ret;
+ }
- if (loc == NULL)
- return ret;
+ if (loc == NULL)
+ return ret;
- if ((inode) && __is_root_gfid (inode->gfid)) {
- loc->parent = NULL;
- goto ignore_parent;
- }
+ if ((inode) && __is_root_gfid(inode->gfid)) {
+ loc->parent = NULL;
+ goto ignore_parent;
+ }
- if (parent_gfid == NULL)
- parent = inode_parent (inode, 0, NULL);
- else
- parent = inode_find (inode->table,
- (unsigned char *) parent_gfid);
+ if (parent_gfid == NULL)
+ parent = inode_parent(inode, 0, NULL);
+ else
+ parent = inode_find(inode->table, (unsigned char *)parent_gfid);
- if (parent == NULL) {
- gf_log ("marker", GF_LOG_ERROR, "parent is NULL for %s",
- uuid_utoa(inode->gfid));
- goto err;
- }
+ if (parent == NULL) {
+ gf_log("marker", GF_LOG_ERROR, "parent is NULL for %s",
+ uuid_utoa(inode->gfid));
+ goto err;
+ }
ignore_parent:
- ret = inode_path (inode, NULL, &resolvedpath);
- if (ret < 0) {
- gf_log ("marker", GF_LOG_ERROR, "failed to resolve path for %s",
- uuid_utoa(inode->gfid));
- goto err;
- }
-
- ret = mq_loc_fill (loc, inode, parent, resolvedpath);
- if (ret < 0)
- goto err;
-
- ret = mq_inode_ctx_get (inode, this, &ctx);
- if (ret < 0 || ctx == NULL)
- ctx = mq_inode_ctx_new (inode, this);
- if (ctx == NULL) {
- gf_log (this->name, GF_LOG_WARNING, "mq_inode_ctx_new "
- "failed for %s", uuid_utoa (inode->gfid));
- ret = -1;
- goto err;
- }
- ret = 0;
+ ret = inode_path(inode, NULL, &resolvedpath);
+ if (ret < 0) {
+ gf_log("marker", GF_LOG_ERROR, "failed to resolve path for %s",
+ uuid_utoa(inode->gfid));
+ goto err;
+ }
+
+ ret = mq_loc_fill(loc, inode, parent, resolvedpath);
+ if (ret < 0)
+ goto err;
+
+ ret = mq_inode_ctx_get(inode, this, &ctx);
+ if (ret < 0 || ctx == NULL)
+ ctx = mq_inode_ctx_new(inode, this);
+ if (ctx == NULL) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "mq_inode_ctx_new "
+ "failed for %s",
+ uuid_utoa(inode->gfid));
+ ret = -1;
+ goto err;
+ }
+ ret = 0;
err:
- if (parent)
- inode_unref (parent);
+ if (parent)
+ inode_unref(parent);
- GF_FREE (resolvedpath);
+ GF_FREE(resolvedpath);
- return ret;
+ return ret;
}
-
quota_inode_ctx_t *
-mq_alloc_inode_ctx ()
+mq_alloc_inode_ctx()
{
- int32_t ret = -1;
- quota_inode_ctx_t *ctx = NULL;
-
- QUOTA_ALLOC (ctx, quota_inode_ctx_t, ret);
- if (ret == -1)
- goto out;
-
- ctx->size = 0;
- ctx->dirty = 0;
- ctx->updation_status = _gf_false;
- LOCK_INIT (&ctx->lock);
- INIT_LIST_HEAD (&ctx->contribution_head);
+ int32_t ret = -1;
+ quota_inode_ctx_t *ctx = NULL;
+
+ QUOTA_ALLOC(ctx, quota_inode_ctx_t, ret);
+ if (ret == -1)
+ goto out;
+
+ ctx->size = 0;
+ ctx->dirty = 0;
+ ctx->updation_status = _gf_false;
+ LOCK_INIT(&ctx->lock);
+ INIT_LIST_HEAD(&ctx->contribution_head);
out:
- return ctx;
+ return ctx;
}
static void
-mq_contri_fini (inode_contribution_t *contri)
+mq_contri_fini(inode_contribution_t *contri)
{
- LOCK_DESTROY (&contri->lock);
- GF_FREE (contri);
+ LOCK_DESTROY(&contri->lock);
+ GF_FREE(contri);
}
-inode_contribution_t*
-mq_contri_init (inode_t *inode)
+inode_contribution_t *
+mq_contri_init(inode_t *inode)
{
- inode_contribution_t *contri = NULL;
- int32_t ret = 0;
+ inode_contribution_t *contri = NULL;
+ int32_t ret = 0;
- QUOTA_ALLOC (contri, inode_contribution_t, ret);
- if (ret == -1)
- goto out;
+ QUOTA_ALLOC(contri, inode_contribution_t, ret);
+ if (ret == -1)
+ goto out;
- GF_REF_INIT (contri, mq_contri_fini);
+ GF_REF_INIT(contri, mq_contri_fini);
- contri->contribution = 0;
- contri->file_count = 0;
- contri->dir_count = 0;
- gf_uuid_copy (contri->gfid, inode->gfid);
+ contri->contribution = 0;
+ contri->file_count = 0;
+ contri->dir_count = 0;
+ gf_uuid_copy(contri->gfid, inode->gfid);
- LOCK_INIT (&contri->lock);
- INIT_LIST_HEAD (&contri->contri_list);
+ LOCK_INIT(&contri->lock);
+ INIT_LIST_HEAD(&contri->contri_list);
out:
- return contri;
+ return contri;
}
inode_contribution_t *
-mq_get_contribution_node (inode_t *inode, quota_inode_ctx_t *ctx)
+mq_get_contribution_node(inode_t *inode, quota_inode_ctx_t *ctx)
{
- inode_contribution_t *contri = NULL;
- inode_contribution_t *temp = NULL;
+ inode_contribution_t *contri = NULL;
+ inode_contribution_t *temp = NULL;
+
+ if (!inode || !ctx)
+ goto out;
- if (!inode || !ctx)
- goto out;
+ LOCK(&ctx->lock);
+ {
+ if (list_empty(&ctx->contribution_head))
+ goto unlock;
- LOCK (&ctx->lock);
+ list_for_each_entry(temp, &ctx->contribution_head, contri_list)
{
- if (list_empty (&ctx->contribution_head))
- goto unlock;
-
- list_for_each_entry (temp, &ctx->contribution_head,
- contri_list) {
- if (gf_uuid_compare (temp->gfid, inode->gfid) == 0) {
- contri = temp;
- GF_REF_GET (contri);
- break;
- }
- }
+ if (gf_uuid_compare(temp->gfid, inode->gfid) == 0) {
+ contri = temp;
+ GF_REF_GET(contri);
+ break;
+ }
}
+ }
unlock:
- UNLOCK (&ctx->lock);
+ UNLOCK(&ctx->lock);
out:
- return contri;
+ return contri;
}
inode_contribution_t *
-__mq_add_new_contribution_node (xlator_t *this, quota_inode_ctx_t *ctx,
- loc_t *loc)
+__mq_add_new_contribution_node(xlator_t *this, quota_inode_ctx_t *ctx,
+ loc_t *loc)
{
- inode_contribution_t *contribution = NULL;
-
- if (!loc->parent) {
- if (!gf_uuid_is_null (loc->pargfid))
- loc->parent = inode_find (loc->inode->table,
- loc->pargfid);
-
- if (!loc->parent)
- loc->parent = inode_parent (loc->inode, loc->pargfid,
- loc->name);
- if (!loc->parent)
- goto out;
- }
-
- list_for_each_entry (contribution, &ctx->contribution_head,
- contri_list) {
- if (loc->parent &&
- gf_uuid_compare (contribution->gfid, loc->parent->gfid) == 0) {
- goto out;
- }
+ inode_contribution_t *contribution = NULL;
+
+ if (!loc->parent) {
+ if (!gf_uuid_is_null(loc->pargfid))
+ loc->parent = inode_find(loc->inode->table, loc->pargfid);
+
+ if (!loc->parent)
+ loc->parent = inode_parent(loc->inode, loc->pargfid, loc->name);
+ if (!loc->parent)
+ goto out;
+ }
+
+ list_for_each_entry(contribution, &ctx->contribution_head, contri_list)
+ {
+ if (loc->parent &&
+ gf_uuid_compare(contribution->gfid, loc->parent->gfid) == 0) {
+ goto out;
}
+ }
- contribution = mq_contri_init (loc->parent);
- if (contribution == NULL)
- goto out;
+ contribution = mq_contri_init(loc->parent);
+ if (contribution == NULL)
+ goto out;
- list_add_tail (&contribution->contri_list, &ctx->contribution_head);
+ list_add_tail(&contribution->contri_list, &ctx->contribution_head);
out:
- return contribution;
+ return contribution;
}
-
inode_contribution_t *
-mq_add_new_contribution_node (xlator_t *this, quota_inode_ctx_t *ctx,
- loc_t *loc)
+mq_add_new_contribution_node(xlator_t *this, quota_inode_ctx_t *ctx, loc_t *loc)
{
- inode_contribution_t *contribution = NULL;
+ inode_contribution_t *contribution = NULL;
- if ((ctx == NULL) || (loc == NULL))
- return NULL;
+ if ((ctx == NULL) || (loc == NULL))
+ return NULL;
- if (((loc->path) && (strcmp (loc->path, "/") == 0))
- || (!loc->path && gf_uuid_is_null (loc->pargfid)))
- return NULL;
+ if (((loc->path) && (strcmp(loc->path, "/") == 0)) ||
+ (!loc->path && gf_uuid_is_null(loc->pargfid)))
+ return NULL;
- LOCK (&ctx->lock);
- {
- contribution = __mq_add_new_contribution_node (this, ctx, loc);
- if (contribution)
- GF_REF_GET (contribution);
- }
- UNLOCK (&ctx->lock);
+ LOCK(&ctx->lock);
+ {
+ contribution = __mq_add_new_contribution_node(this, ctx, loc);
+ if (contribution)
+ GF_REF_GET(contribution);
+ }
+ UNLOCK(&ctx->lock);
- return contribution;
+ return contribution;
}
-
int32_t
-mq_dict_set_contribution (xlator_t *this, dict_t *dict, loc_t *loc,
- uuid_t gfid, char *contri_key)
+mq_dict_set_contribution(xlator_t *this, dict_t *dict, loc_t *loc, uuid_t gfid,
+ char *contri_key)
{
- int32_t ret = -1;
- char key[QUOTA_KEY_MAX] = {0, };
-
- GF_VALIDATE_OR_GOTO ("marker", this, out);
- GF_VALIDATE_OR_GOTO ("marker", dict, out);
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
-
- if (gfid && !gf_uuid_is_null(gfid)) {
- GET_CONTRI_KEY (this, key, gfid, ret);
- } else if (loc->parent) {
- GET_CONTRI_KEY (this, key, loc->parent->gfid, ret);
- } else {
- /* nameless lookup, fetch contributions to all parents */
- GET_CONTRI_KEY (this, key, NULL, ret);
+ int32_t ret = -1;
+ char key[QUOTA_KEY_MAX] = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("marker", this, out);
+ GF_VALIDATE_OR_GOTO("marker", dict, out);
+ GF_VALIDATE_OR_GOTO("marker", loc, out);
+
+ if (gfid && !gf_uuid_is_null(gfid)) {
+ GET_CONTRI_KEY(this, key, gfid, ret);
+ } else if (loc->parent) {
+ GET_CONTRI_KEY(this, key, loc->parent->gfid, ret);
+ } else {
+ /* nameless lookup, fetch contributions to all parents */
+ GET_CONTRI_KEY(this, key, NULL, ret);
+ }
+
+ if (ret < 0)
+ goto out;
+
+ ret = dict_set_int64(dict, key, 0);
+ if (ret < 0)
+ goto out;
+
+ if (contri_key)
+ if (snprintf(contri_key, QUOTA_KEY_MAX, "%s", key) >= QUOTA_KEY_MAX) {
+ ret = -1;
+ goto out;
}
- if (ret < 0)
- goto out;
-
- ret = dict_set_int64 (dict, key, 0);
- if (ret < 0)
- goto out;
-
- if (contri_key)
- strncpy (contri_key, key, QUOTA_KEY_MAX);
-
out:
- if (ret < 0)
- gf_log_callingfn (this->name, GF_LOG_ERROR, "dict set failed");
+ if (ret < 0)
+ gf_log_callingfn(this ? this->name : "Marker", GF_LOG_ERROR,
+ "dict set failed");
- return ret;
+ return ret;
}
-
int32_t
-mq_inode_ctx_get (inode_t *inode, xlator_t *this,
- quota_inode_ctx_t **ctx)
+mq_inode_ctx_get(inode_t *inode, xlator_t *this, quota_inode_ctx_t **ctx)
{
- int32_t ret = -1;
- uint64_t ctx_int = 0;
- marker_inode_ctx_t *mark_ctx = NULL;
+ int32_t ret = -1;
+ uint64_t ctx_int = 0;
+ marker_inode_ctx_t *mark_ctx = NULL;
- GF_VALIDATE_OR_GOTO ("marker", inode, out);
- GF_VALIDATE_OR_GOTO ("marker", this, out);
- GF_VALIDATE_OR_GOTO ("marker", ctx, out);
+ GF_VALIDATE_OR_GOTO("marker", inode, out);
+ GF_VALIDATE_OR_GOTO("marker", this, out);
+ GF_VALIDATE_OR_GOTO("marker", ctx, out);
- ret = inode_ctx_get (inode, this, &ctx_int);
- if (ret < 0) {
- ret = -1;
- *ctx = NULL;
- goto out;
- }
+ ret = inode_ctx_get(inode, this, &ctx_int);
+ if (ret < 0) {
+ ret = -1;
+ *ctx = NULL;
+ goto out;
+ }
- mark_ctx = (marker_inode_ctx_t *) (unsigned long)ctx_int;
- if (mark_ctx->quota_ctx == NULL) {
- ret = -1;
- goto out;
- }
+ mark_ctx = (marker_inode_ctx_t *)(unsigned long)ctx_int;
+ if (mark_ctx->quota_ctx == NULL) {
+ ret = -1;
+ goto out;
+ }
- *ctx = mark_ctx->quota_ctx;
+ *ctx = mark_ctx->quota_ctx;
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
-
quota_inode_ctx_t *
-__mq_inode_ctx_new (inode_t *inode, xlator_t *this)
+__mq_inode_ctx_new(inode_t *inode, xlator_t *this)
{
- int32_t ret = -1;
- quota_inode_ctx_t *quota_ctx = NULL;
- marker_inode_ctx_t *mark_ctx = NULL;
-
- ret = marker_force_inode_ctx_get (inode, this, &mark_ctx);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR,
- "marker_force_inode_ctx_get() failed");
- goto out;
+ int32_t ret = -1;
+ quota_inode_ctx_t *quota_ctx = NULL;
+ marker_inode_ctx_t *mark_ctx = NULL;
+
+ ret = marker_force_inode_ctx_get(inode, this, &mark_ctx);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_ERROR, "marker_force_inode_ctx_get() failed");
+ goto out;
+ }
+
+ LOCK(&inode->lock);
+ {
+ if (mark_ctx->quota_ctx == NULL) {
+ quota_ctx = mq_alloc_inode_ctx();
+ if (quota_ctx == NULL) {
+ ret = -1;
+ goto unlock;
+ }
+ mark_ctx->quota_ctx = quota_ctx;
+ } else {
+ quota_ctx = mark_ctx->quota_ctx;
}
- LOCK (&inode->lock);
- {
- if (mark_ctx->quota_ctx == NULL) {
- quota_ctx = mq_alloc_inode_ctx ();
- if (quota_ctx == NULL) {
- ret = -1;
- goto unlock;
- }
- mark_ctx->quota_ctx = quota_ctx;
- } else {
- quota_ctx = mark_ctx->quota_ctx;
- }
-
- ret = 0;
- }
+ ret = 0;
+ }
unlock:
- UNLOCK (&inode->lock);
+ UNLOCK(&inode->lock);
out:
- return quota_ctx;
+ return quota_ctx;
}
-
quota_inode_ctx_t *
-mq_inode_ctx_new (inode_t * inode, xlator_t *this)
-{
- return __mq_inode_ctx_new (inode, this);
-}
-
-quota_local_t *
-mq_local_new ()
-{
- quota_local_t *local = NULL;
-
- local = mem_get0 (THIS->local_pool);
- if (!local)
- goto out;
-
- local->ref = 1;
- LOCK_INIT (&local->lock);
-
- local->ctx = NULL;
- local->contri = NULL;
-
-out:
- return local;
-}
-
-quota_local_t *
-mq_local_ref (quota_local_t *local)
-{
- LOCK (&local->lock);
- {
- local->ref ++;
- }
- UNLOCK (&local->lock);
-
- return local;
-}
-
-
-int32_t
-mq_local_unref (xlator_t *this, quota_local_t *local)
-{
- int32_t ref = 0;
- if (local == NULL)
- goto out;
-
- QUOTA_SAFE_DECREMENT (&local->lock, local->ref, ref);
-
- if (ref != 0)
- goto out;
-
- if (local->fd != NULL)
- fd_unref (local->fd);
-
- if (local->contri)
- GF_REF_PUT (local->contri);
-
- if (local->xdata)
- dict_unref (local->xdata);
-
- loc_wipe (&local->loc);
-
- loc_wipe (&local->parent_loc);
-
- LOCK_DESTROY (&local->lock);
-
- mem_put (local);
-out:
- return 0;
-}
-
-
-inode_contribution_t *
-mq_get_contribution_from_loc (xlator_t *this, loc_t *loc)
+mq_inode_ctx_new(inode_t *inode, xlator_t *this)
{
- int32_t ret = 0;
- quota_inode_ctx_t *ctx = NULL;
- inode_contribution_t *contribution = NULL;
-
- ret = mq_inode_ctx_get (loc->inode, this, &ctx);
- if (ret < 0) {
- gf_log_callingfn (this->name, GF_LOG_WARNING,
- "cannot get marker-quota context from inode "
- "(gfid:%s, path:%s)",
- uuid_utoa (loc->inode->gfid), loc->path);
- goto err;
- }
-
- contribution = mq_get_contribution_node (loc->parent, ctx);
- if (contribution == NULL) {
- gf_log_callingfn (this->name, GF_LOG_WARNING,
- "inode (gfid:%s, path:%s) has "
- "no contribution towards parent (gfid:%s)",
- uuid_utoa (loc->inode->gfid),
- loc->path, uuid_utoa (loc->parent->gfid));
- goto err;
- }
-
-err:
- return contribution;
+ return __mq_inode_ctx_new(inode, this);
}
diff --git a/xlators/features/marker/src/marker-quota-helper.h b/xlators/features/marker/src/marker-quota-helper.h
index 444ba7ca71e..d4091dd2180 100644
--- a/xlators/features/marker/src/marker-quota-helper.h
+++ b/xlators/features/marker/src/marker-quota-helper.h
@@ -13,66 +13,54 @@
#include "marker.h"
-#define QUOTA_FREE_CONTRIBUTION_NODE(ctx, _contribution) \
- do { \
- LOCK (&ctx->lock); \
- { \
- list_del_init (&_contribution->contri_list); \
- GF_REF_PUT (_contribution); \
- } \
- UNLOCK (&ctx->lock); \
- } while (0)
-
-#define QUOTA_SAFE_INCREMENT(lock, var) \
- do { \
- LOCK (lock); \
- var ++; \
- UNLOCK (lock); \
- } while (0)
-
-#define QUOTA_SAFE_DECREMENT(lock, var, value) \
- do { \
- LOCK (lock); \
- { \
- value = --var; \
- } \
- UNLOCK (lock); \
- } while (0)
+#define QUOTA_FREE_CONTRIBUTION_NODE(ctx, _contribution) \
+ do { \
+ LOCK(&ctx->lock); \
+ { \
+ list_del_init(&_contribution->contri_list); \
+ GF_REF_PUT(_contribution); \
+ } \
+ UNLOCK(&ctx->lock); \
+ } while (0)
+
+#define QUOTA_SAFE_INCREMENT(lock, var) \
+ do { \
+ LOCK(lock); \
+ var++; \
+ UNLOCK(lock); \
+ } while (0)
+
+#define QUOTA_SAFE_DECREMENT(lock, var, value) \
+ do { \
+ LOCK(lock); \
+ { \
+ value = --var; \
+ } \
+ UNLOCK(lock); \
+ } while (0)
inode_contribution_t *
-mq_add_new_contribution_node (xlator_t *, quota_inode_ctx_t *, loc_t *);
+mq_add_new_contribution_node(xlator_t *, quota_inode_ctx_t *, loc_t *);
int32_t
-mq_dict_set_contribution (xlator_t *, dict_t *, loc_t *, uuid_t, char *);
+mq_dict_set_contribution(xlator_t *, dict_t *, loc_t *, uuid_t, char *);
quota_inode_ctx_t *
-mq_inode_ctx_new (inode_t *, xlator_t *);
+mq_inode_ctx_new(inode_t *, xlator_t *);
int32_t
-mq_inode_ctx_get (inode_t *, xlator_t *, quota_inode_ctx_t **);
+mq_inode_ctx_get(inode_t *, xlator_t *, quota_inode_ctx_t **);
int32_t
-mq_delete_contribution_node (dict_t *, char *, inode_contribution_t *);
+mq_delete_contribution_node(dict_t *, char *, inode_contribution_t *);
int32_t
-mq_inode_loc_fill (const char *, inode_t *, loc_t *);
-
-quota_local_t *
-mq_local_new ();
-
-quota_local_t *
-mq_local_ref (quota_local_t *);
-
-int32_t
-mq_local_unref (xlator_t *, quota_local_t *);
-
-inode_contribution_t*
-mq_contri_init (inode_t *inode);
+mq_inode_loc_fill(const char *, inode_t *, loc_t *);
inode_contribution_t *
-mq_get_contribution_node (inode_t *, quota_inode_ctx_t *);
+mq_contri_init(inode_t *inode);
inode_contribution_t *
-mq_get_contribution_from_loc (xlator_t *this, loc_t *loc);
+mq_get_contribution_node(inode_t *, quota_inode_ctx_t *);
#endif
diff --git a/xlators/features/marker/src/marker-quota.c b/xlators/features/marker/src/marker-quota.c
index b6bf2fb36a4..3de2ea1c92c 100644
--- a/xlators/features/marker/src/marker-quota.c
+++ b/xlators/features/marker/src/marker-quota.c
@@ -7,282 +7,263 @@
later), or the GNU General Public License, version 2 (GPLv2), in all
cases as published by the Free Software Foundation.
*/
-#include "dict.h"
-#include "xlator.h"
-#include "defaults.h"
+#include <glusterfs/dict.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
#include "libxlator.h"
-#include "common-utils.h"
-#include "byte-order.h"
+#include <glusterfs/common-utils.h>
+#include <glusterfs/byte-order.h>
#include "marker-quota.h"
#include "marker-quota-helper.h"
-#include "syncop.h"
-#include "quota-common-utils.h"
+#include <glusterfs/syncop.h>
+#include <glusterfs/quota-common-utils.h>
int
-mq_loc_copy (loc_t *dst, loc_t *src)
+mq_loc_copy(loc_t *dst, loc_t *src)
{
- int ret = -1;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO ("marker", dst, out);
- GF_VALIDATE_OR_GOTO ("marker", src, out);
+ GF_VALIDATE_OR_GOTO("marker", dst, out);
+ GF_VALIDATE_OR_GOTO("marker", src, out);
- if (src->inode == NULL ||
- ((src->parent == NULL) && (gf_uuid_is_null (src->pargfid))
- && !__is_root_gfid (src->inode->gfid))) {
- gf_log ("marker", GF_LOG_WARNING,
- "src loc is not valid");
- goto out;
- }
+ if (src->inode == NULL ||
+ ((src->parent == NULL) && (gf_uuid_is_null(src->pargfid)) &&
+ !__is_root_gfid(src->inode->gfid))) {
+ gf_log("marker", GF_LOG_WARNING, "src loc is not valid");
+ goto out;
+ }
- ret = loc_copy (dst, src);
+ ret = loc_copy(dst, src);
out:
- return ret;
+ return ret;
}
static void
-mq_set_ctx_status (quota_inode_ctx_t *ctx, gf_boolean_t *flag,
- gf_boolean_t status)
+mq_set_ctx_status(quota_inode_ctx_t *ctx, gf_boolean_t *flag,
+ gf_boolean_t status)
{
- LOCK (&ctx->lock);
- {
- *flag = status;
- }
- UNLOCK (&ctx->lock);
+ LOCK(&ctx->lock);
+ {
+ *flag = status;
+ }
+ UNLOCK(&ctx->lock);
}
static void
-mq_test_and_set_ctx_status (quota_inode_ctx_t *ctx, gf_boolean_t *flag,
- gf_boolean_t *status)
+mq_test_and_set_ctx_status(quota_inode_ctx_t *ctx, gf_boolean_t *flag,
+ gf_boolean_t *status)
{
- gf_boolean_t temp = _gf_false;
-
- LOCK (&ctx->lock);
- {
- temp = *status;
- *status = *flag;
- *flag = temp;
- }
- UNLOCK (&ctx->lock);
+ gf_boolean_t temp = _gf_false;
+
+ LOCK(&ctx->lock);
+ {
+ temp = *status;
+ *status = *flag;
+ *flag = temp;
+ }
+ UNLOCK(&ctx->lock);
}
static void
-mq_get_ctx_status (quota_inode_ctx_t *ctx, gf_boolean_t *flag,
- gf_boolean_t *status)
+mq_get_ctx_status(quota_inode_ctx_t *ctx, gf_boolean_t *flag,
+ gf_boolean_t *status)
{
- LOCK (&ctx->lock);
- {
- *status = *flag;
- }
- UNLOCK (&ctx->lock);
-}
-
-int32_t
-mq_get_ctx_updation_status (quota_inode_ctx_t *ctx,
- gf_boolean_t *status)
-{
- GF_VALIDATE_OR_GOTO ("marker", ctx, out);
- GF_VALIDATE_OR_GOTO ("marker", status, out);
-
- mq_get_ctx_status (ctx, &ctx->updation_status, status);
- return 0;
-out:
- return -1;
+ LOCK(&ctx->lock);
+ {
+ *status = *flag;
+ }
+ UNLOCK(&ctx->lock);
}
int32_t
-mq_set_ctx_updation_status (quota_inode_ctx_t *ctx,
- gf_boolean_t status)
+mq_get_ctx_updation_status(quota_inode_ctx_t *ctx, gf_boolean_t *status)
{
- GF_VALIDATE_OR_GOTO ("marker", ctx, out);
+ GF_VALIDATE_OR_GOTO("marker", ctx, out);
+ GF_VALIDATE_OR_GOTO("marker", status, out);
- mq_set_ctx_status (ctx, &ctx->updation_status, status);
- return 0;
+ mq_get_ctx_status(ctx, &ctx->updation_status, status);
+ return 0;
out:
- return -1;
+ return -1;
}
int32_t
-mq_test_and_set_ctx_updation_status (quota_inode_ctx_t *ctx,
- gf_boolean_t *status)
+mq_set_ctx_updation_status(quota_inode_ctx_t *ctx, gf_boolean_t status)
{
- GF_VALIDATE_OR_GOTO ("marker", ctx, out);
- GF_VALIDATE_OR_GOTO ("marker", status, out);
+ GF_VALIDATE_OR_GOTO("marker", ctx, out);
- mq_test_and_set_ctx_status (ctx, &ctx->updation_status, status);
- return 0;
+ mq_set_ctx_status(ctx, &ctx->updation_status, status);
+ return 0;
out:
- return -1;
+ return -1;
}
int32_t
-mq_set_ctx_create_status (quota_inode_ctx_t *ctx,
- gf_boolean_t status)
+mq_test_and_set_ctx_updation_status(quota_inode_ctx_t *ctx,
+ gf_boolean_t *status)
{
- GF_VALIDATE_OR_GOTO ("marker", ctx, out);
+ GF_VALIDATE_OR_GOTO("marker", ctx, out);
+ GF_VALIDATE_OR_GOTO("marker", status, out);
- mq_set_ctx_status (ctx, &ctx->create_status, status);
- return 0;
+ mq_test_and_set_ctx_status(ctx, &ctx->updation_status, status);
+ return 0;
out:
- return -1;
+ return -1;
}
int32_t
-mq_test_and_set_ctx_create_status (quota_inode_ctx_t *ctx,
- gf_boolean_t *status)
+mq_set_ctx_create_status(quota_inode_ctx_t *ctx, gf_boolean_t status)
{
- GF_VALIDATE_OR_GOTO ("marker", ctx, out);
- GF_VALIDATE_OR_GOTO ("marker", status, out);
+ GF_VALIDATE_OR_GOTO("marker", ctx, out);
- mq_test_and_set_ctx_status (ctx, &ctx->create_status, status);
- return 0;
+ mq_set_ctx_status(ctx, &ctx->create_status, status);
+ return 0;
out:
- return -1;
+ return -1;
}
int32_t
-mq_set_ctx_dirty_status (quota_inode_ctx_t *ctx,
- gf_boolean_t status)
+mq_test_and_set_ctx_create_status(quota_inode_ctx_t *ctx, gf_boolean_t *status)
{
- GF_VALIDATE_OR_GOTO ("marker", ctx, out);
+ GF_VALIDATE_OR_GOTO("marker", ctx, out);
+ GF_VALIDATE_OR_GOTO("marker", status, out);
- mq_set_ctx_status (ctx, &ctx->dirty_status, status);
- return 0;
+ mq_test_and_set_ctx_status(ctx, &ctx->create_status, status);
+ return 0;
out:
- return -1;
+ return -1;
}
-int32_t
-mq_test_and_set_ctx_dirty_status (quota_inode_ctx_t *ctx,
- gf_boolean_t *status)
+static void
+mq_set_ctx_dirty_status(quota_inode_ctx_t *ctx, gf_boolean_t status)
{
- GF_VALIDATE_OR_GOTO ("marker", ctx, out);
- GF_VALIDATE_OR_GOTO ("marker", status, out);
+ GF_VALIDATE_OR_GOTO("marker", ctx, out);
- mq_test_and_set_ctx_status (ctx, &ctx->dirty_status, status);
- return 0;
+ mq_set_ctx_status(ctx, &ctx->dirty_status, status);
out:
- return -1;
+ return;
}
int
-mq_build_ancestry (xlator_t *this, loc_t *loc)
+mq_build_ancestry(xlator_t *this, loc_t *loc)
{
- int32_t ret = -1;
- fd_t *fd = NULL;
- gf_dirent_t entries;
- gf_dirent_t *entry = NULL;
- dict_t *xdata = NULL;
- inode_t *tmp_parent = NULL;
- inode_t *tmp_inode = NULL;
- inode_t *linked_inode = NULL;
- quota_inode_ctx_t *ctx = NULL;
-
- INIT_LIST_HEAD (&entries.list);
-
- xdata = dict_new ();
- if (xdata == NULL) {
- gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
- ret = -ENOMEM;
- goto out;
- }
-
- ret = dict_set_int8 (xdata, GET_ANCESTRY_DENTRY_KEY, 1);
- if (ret < 0)
- goto out;
-
- fd = fd_anonymous (loc->inode);
- if (fd == NULL) {
- gf_log (this->name, GF_LOG_ERROR, "fd creation failed");
- ret = -ENOMEM;
- goto out;
- }
-
- fd_bind (fd);
+ int32_t ret = -1;
+ fd_t *fd = NULL;
+ gf_dirent_t entries;
+ gf_dirent_t *entry = NULL;
+ dict_t *xdata = NULL;
+ inode_t *tmp_parent = NULL;
+ inode_t *tmp_inode = NULL;
+ inode_t *linked_inode = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+
+ INIT_LIST_HEAD(&entries.list);
+
+ xdata = dict_new();
+ if (xdata == NULL) {
+ gf_log(this->name, GF_LOG_ERROR, "dict_new failed");
+ ret = -ENOMEM;
+ goto out;
+ }
- ret = syncop_readdirp (this, fd, 131072, 0, &entries, xdata, NULL);
- if (ret < 0) {
- gf_log (this->name, (-ret == ENOENT || -ret == ESTALE)
- ? GF_LOG_DEBUG:GF_LOG_ERROR, "readdirp failed "
- "for %s: %s", loc->path, strerror (-ret));
- goto out;
- }
+ ret = dict_set_int8(xdata, GET_ANCESTRY_DENTRY_KEY, 1);
+ if (ret < 0)
+ goto out;
- if (list_empty (&entries.list)) {
- ret = -1;
- goto out;
- }
+ fd = fd_anonymous(loc->inode);
+ if (fd == NULL) {
+ gf_log(this->name, GF_LOG_ERROR, "fd creation failed");
+ ret = -ENOMEM;
+ goto out;
+ }
- list_for_each_entry (entry, &entries.list, list) {
- if (__is_root_gfid (entry->inode->gfid)) {
- /* The list contains a sub-list for each possible path
- * to the target inode. Each sub-list starts with the
- * root entry of the tree and is followed by the child
- * entries for a particular path to the target entry.
- * The root entry is an implied sub-list delimiter,
- * as it denotes we have started processing a new path.
- * Reset the parent pointer and continue
- */
-
- tmp_parent = NULL;
- } else {
- linked_inode = inode_link (entry->inode, tmp_parent,
- entry->d_name,
- &entry->d_stat);
- if (linked_inode) {
- tmp_inode = entry->inode;
- entry->inode = linked_inode;
- inode_unref (tmp_inode);
- } else {
- gf_log (this->name, GF_LOG_ERROR,
- "inode link failed");
- ret = -EINVAL;
- goto out;
- }
- }
-
- ctx = mq_inode_ctx_new (entry->inode, this);
- if (ctx == NULL) {
- gf_log (this->name, GF_LOG_WARNING, "mq_inode_ctx_new "
- "failed for %s",
- uuid_utoa (entry->inode->gfid));
- ret = -ENOMEM;
- goto out;
- }
-
- /* For non-directory, posix_get_ancestry_non_directory returns
- * all hard-links that are represented by nodes adjacent to
- * each other in the dentry-list.
- * (Unlike the directory case where adjacent nodes either have
- * a parent/child relationship or belong to different paths).
- */
- if (entry->inode->ia_type == IA_IFDIR)
- tmp_parent = entry->inode;
- }
+ fd_bind(fd);
- if (loc->parent)
- inode_unref (loc->parent);
+ ret = syncop_readdirp(this, fd, 131072, 0, &entries, xdata, NULL);
+ if (ret < 0) {
+ gf_log(this->name,
+ (-ret == ENOENT || -ret == ESTALE) ? GF_LOG_DEBUG : GF_LOG_ERROR,
+ "readdirp failed "
+ "for %s: %s",
+ loc->path, strerror(-ret));
+ goto out;
+ }
- loc->parent = inode_parent (loc->inode, 0, NULL);
- if (loc->parent == NULL) {
- ret = -1;
- goto out;
- }
+ if (list_empty(&entries.list)) {
+ ret = -1;
+ goto out;
+ }
+
+ list_for_each_entry(entry, &entries.list, list)
+ {
+ if (__is_root_gfid(entry->inode->gfid)) {
+ /* The list contains a sub-list for each possible path
+ * to the target inode. Each sub-list starts with the
+ * root entry of the tree and is followed by the child
+ * entries for a particular path to the target entry.
+ * The root entry is an implied sub-list delimiter,
+ * as it denotes we have started processing a new path.
+ * Reset the parent pointer and continue
+ */
+
+ tmp_parent = NULL;
+ } else {
+ linked_inode = inode_link(entry->inode, tmp_parent, entry->d_name,
+ &entry->d_stat);
+ if (linked_inode) {
+ tmp_inode = entry->inode;
+ entry->inode = linked_inode;
+ inode_unref(tmp_inode);
+ } else {
+ gf_log(this->name, GF_LOG_ERROR, "inode link failed");
+ ret = -EINVAL;
+ goto out;
+ }
+ }
+
+ ctx = mq_inode_ctx_new(entry->inode, this);
+ if (ctx == NULL) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "mq_inode_ctx_new "
+ "failed for %s",
+ uuid_utoa(entry->inode->gfid));
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ /* For non-directory, posix_get_ancestry_non_directory returns
+ * all hard-links that are represented by nodes adjacent to
+ * each other in the dentry-list.
+ * (Unlike the directory case where adjacent nodes either have
+ * a parent/child relationship or belong to different paths).
+ */
+ if (entry->inode->ia_type == IA_IFDIR)
+ tmp_parent = entry->inode;
+ }
+
+ if (loc->parent)
+ inode_unref(loc->parent);
+
+ loc->parent = inode_parent(loc->inode, 0, NULL);
+ if (loc->parent == NULL) {
+ ret = -1;
+ goto out;
+ }
- ret = 0;
+ ret = 0;
out:
- gf_dirent_free (&entries);
+ gf_dirent_free(&entries);
- if (fd)
- fd_unref (fd);
+ if (fd)
+ fd_unref(fd);
- if (xdata)
- dict_unref (xdata);
+ if (xdata)
+ dict_unref(xdata);
- return ret;
+ return ret;
}
-
/* This function should be used only in inspect_directory and inspect_file
* function to heal quota xattrs.
* Inode quota feature is introduced in 3.7.
@@ -292,1914 +273,2025 @@ out:
* This function returns success even is inode-quota xattrs are missing and
* hence no healing performed.
*/
-int32_t
-_quota_dict_get_meta (xlator_t *this, dict_t *dict, char *key,
- quota_meta_t *meta, ia_type_t ia_type,
- gf_boolean_t add_delta)
+static int32_t
+_quota_dict_get_meta(xlator_t *this, dict_t *dict, char *key, const int keylen,
+ quota_meta_t *meta, ia_type_t ia_type,
+ gf_boolean_t add_delta)
{
- int32_t ret = 0;
- marker_conf_t *priv = NULL;
-
- priv = this->private;
-
- ret = quota_dict_get_inode_meta (dict, key, meta);
- if (ret == -2 && (priv->feature_enabled & GF_INODE_QUOTA) == 0) {
- /* quota_dict_get_inode_meta returns -2 if
- * inode quota xattrs are not present.
- * if inode quota self heal is turned off,
- * then we should skip healing inode quotas
- */
-
- gf_log (this->name, GF_LOG_DEBUG, "inode quota disabled. "
- "inode quota self heal will not be performed");
- ret = 0;
- if (add_delta) {
- if (ia_type == IA_IFDIR)
- meta->dir_count = 1;
- else
- meta->file_count = 1;
- }
+ int32_t ret = 0;
+ marker_conf_t *priv = NULL;
+
+ priv = this->private;
+
+ ret = quota_dict_get_inode_meta(dict, key, keylen, meta);
+ if (ret == -2 && (priv->feature_enabled & GF_INODE_QUOTA) == 0) {
+ /* quota_dict_get_inode_meta returns -2 if
+ * inode quota xattrs are not present.
+ * if inode quota self heal is turned off,
+ * then we should skip healing inode quotas
+ */
+
+ gf_log(this->name, GF_LOG_DEBUG,
+ "inode quota disabled. "
+ "inode quota self heal will not be performed");
+ ret = 0;
+ if (add_delta) {
+ if (ia_type == IA_IFDIR)
+ meta->dir_count = 1;
+ else
+ meta->file_count = 1;
}
+ }
- return ret;
+ return ret;
}
int32_t
-quota_dict_set_size_meta (xlator_t *this, dict_t *dict,
- const quota_meta_t *meta)
+quota_dict_set_size_meta(xlator_t *this, dict_t *dict, const quota_meta_t *meta)
{
- int32_t ret = -ENOMEM;
- quota_meta_t *value = NULL;
- char size_key[QUOTA_KEY_MAX] = {0, };
-
- value = GF_CALLOC (2, sizeof (quota_meta_t), gf_common_quota_meta_t);
- if (value == NULL) {
- goto out;
- }
- value[0].size = hton64 (meta->size);
- value[0].file_count = hton64 (meta->file_count);
- value[0].dir_count = hton64 (meta->dir_count);
+ int32_t ret = -ENOMEM;
+ quota_meta_t *value = NULL;
+ char size_key[QUOTA_KEY_MAX] = {
+ 0,
+ };
+
+ value = GF_MALLOC(2 * sizeof(quota_meta_t), gf_common_quota_meta_t);
+ if (value == NULL) {
+ goto out;
+ }
+ value[0].size = hton64(meta->size);
+ value[0].file_count = hton64(meta->file_count);
+ value[0].dir_count = hton64(meta->dir_count);
- value[1].size = 0;
- value[1].file_count = 0;
- value[1].dir_count = hton64 (1);
+ value[1].size = 0;
+ value[1].file_count = 0;
+ value[1].dir_count = hton64(1);
- GET_SIZE_KEY (this, size_key, ret);
- if (ret < 0)
- goto out;
- ret = dict_set_bin (dict, size_key, value,
- (sizeof (quota_meta_t) * 2));
- if (ret < 0) {
- gf_log_callingfn ("quota", GF_LOG_ERROR, "dict set failed");
- GF_FREE (value);
- }
+ GET_SIZE_KEY(this, size_key, ret);
+ if (ret < 0)
+ goto out;
+ ret = dict_set_bin(dict, size_key, value, (sizeof(quota_meta_t) * 2));
+ if (ret < 0) {
+ gf_log_callingfn("quota", GF_LOG_ERROR, "dict set failed");
+ GF_FREE(value);
+ }
out:
- return ret;
+ return ret;
}
void
-mq_compute_delta (quota_meta_t *delta, const quota_meta_t *op1,
- const quota_meta_t *op2)
+mq_compute_delta(quota_meta_t *delta, const quota_meta_t *op1,
+ const quota_meta_t *op2)
{
- delta->size = op1->size - op2->size;
- delta->file_count = op1->file_count - op2->file_count;
- delta->dir_count = op1->dir_count - op2->dir_count;
+ delta->size = op1->size - op2->size;
+ delta->file_count = op1->file_count - op2->file_count;
+ delta->dir_count = op1->dir_count - op2->dir_count;
}
void
-mq_add_meta (quota_meta_t *dst, const quota_meta_t *src)
+mq_add_meta(quota_meta_t *dst, const quota_meta_t *src)
{
- dst->size += src->size;
- dst->file_count += src->file_count;
- dst->dir_count += src->dir_count;
+ dst->size += src->size;
+ dst->file_count += src->file_count;
+ dst->dir_count += src->dir_count;
}
void
-mq_sub_meta (quota_meta_t *dst, const quota_meta_t *src)
+mq_sub_meta(quota_meta_t *dst, const quota_meta_t *src)
{
- if (src == NULL) {
- dst->size = -dst->size;
- dst->file_count = -dst->file_count;
- dst->dir_count = -dst->dir_count;
- } else {
- dst->size = src->size - dst->size;
- dst->file_count = src->file_count - dst->file_count;
- dst->dir_count = src->dir_count - dst->dir_count;
- }
+ if (src == NULL) {
+ dst->size = -dst->size;
+ dst->file_count = -dst->file_count;
+ dst->dir_count = -dst->dir_count;
+ } else {
+ dst->size = src->size - dst->size;
+ dst->file_count = src->file_count - dst->file_count;
+ dst->dir_count = src->dir_count - dst->dir_count;
+ }
}
int32_t
-mq_are_xattrs_set (xlator_t *this, loc_t *loc, gf_boolean_t *contri_set,
- gf_boolean_t *size_set)
+mq_are_xattrs_set(xlator_t *this, loc_t *loc, gf_boolean_t *contri_set,
+ gf_boolean_t *size_set)
{
- int32_t ret = -1;
- char contri_key[QUOTA_KEY_MAX] = {0, };
- char size_key[QUOTA_KEY_MAX] = {0, };
- quota_meta_t meta = {0, };
- struct iatt stbuf = {0,};
- dict_t *dict = NULL;
- dict_t *rsp_dict = NULL;
-
- dict = dict_new ();
- if (dict == NULL) {
- gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
- goto out;
- }
-
- ret = mq_req_xattr (this, loc, dict, contri_key, size_key);
- if (ret < 0)
- goto out;
+ int32_t ret = -1;
+ char contri_key[QUOTA_KEY_MAX] = {
+ 0,
+ };
+ char size_key[QUOTA_KEY_MAX] = {
+ 0,
+ };
+ quota_meta_t meta = {
+ 0,
+ };
+ struct iatt stbuf = {
+ 0,
+ };
+ dict_t *dict = NULL;
+ dict_t *rsp_dict = NULL;
+
+ dict = dict_new();
+ if (dict == NULL) {
+ gf_log(this->name, GF_LOG_ERROR, "dict_new failed");
+ goto out;
+ }
- ret = syncop_lookup (FIRST_CHILD(this), loc, &stbuf, NULL,
- dict, &rsp_dict);
- if (ret < 0) {
- gf_log_callingfn (this->name, (-ret == ENOENT || -ret == ESTALE)
- ? GF_LOG_DEBUG:GF_LOG_ERROR, "lookup failed "
- "for %s: %s", loc->path, strerror (-ret));
- goto out;
- }
+ ret = mq_req_xattr(this, loc, dict, contri_key, size_key);
+ if (ret < 0)
+ goto out;
- if (rsp_dict == NULL)
- goto out;
+ ret = syncop_lookup(FIRST_CHILD(this), loc, &stbuf, NULL, dict, &rsp_dict);
+ if (ret < 0) {
+ gf_log_callingfn(
+ this->name,
+ (-ret == ENOENT || -ret == ESTALE) ? GF_LOG_DEBUG : GF_LOG_ERROR,
+ "lookup failed "
+ "for %s: %s",
+ loc->path, strerror(-ret));
+ goto out;
+ }
- *contri_set = _gf_true;
- *size_set = _gf_true;
- if (loc->inode->ia_type == IA_IFDIR) {
- ret = quota_dict_get_inode_meta (rsp_dict, size_key, &meta);
- if (ret < 0 || meta.dir_count == 0)
- *size_set = _gf_false;
- }
+ if (rsp_dict == NULL)
+ goto out;
- if (!loc_is_root(loc)) {
- ret = quota_dict_get_inode_meta (rsp_dict, contri_key, &meta);
- if (ret < 0)
- *contri_set = _gf_false;
- }
+ *contri_set = _gf_true;
+ *size_set = _gf_true;
+ if (loc->inode->ia_type == IA_IFDIR) {
+ ret = quota_dict_get_inode_meta(rsp_dict, size_key, strlen(size_key),
+ &meta);
+ if (ret < 0 || meta.dir_count == 0)
+ *size_set = _gf_false;
+ }
+
+ if (!loc_is_root(loc)) {
+ ret = quota_dict_get_inode_meta(rsp_dict, contri_key,
+ strlen(contri_key), &meta);
+ if (ret < 0)
+ *contri_set = _gf_false;
+ }
- ret = 0;
+ ret = 0;
out:
- if (dict)
- dict_unref (dict);
+ if (dict)
+ dict_unref(dict);
- if (rsp_dict)
- dict_unref (rsp_dict);
+ if (rsp_dict)
+ dict_unref(rsp_dict);
- return ret;
+ return ret;
}
int32_t
-mq_create_size_xattrs (xlator_t *this, quota_inode_ctx_t *ctx, loc_t *loc)
+mq_create_size_xattrs(xlator_t *this, quota_inode_ctx_t *ctx, loc_t *loc)
{
- int32_t ret = -1;
- quota_meta_t size = {0, };
- dict_t *dict = NULL;
+ int32_t ret = -1;
+ quota_meta_t size = {
+ 0,
+ };
+ dict_t *dict = NULL;
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
- GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
+ GF_VALIDATE_OR_GOTO("marker", loc, out);
+ GF_VALIDATE_OR_GOTO("marker", loc->inode, out);
- if (loc->inode->ia_type != IA_IFDIR) {
- ret = 0;
- goto out;
- }
-
- dict = dict_new ();
- if (!dict) {
- gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
- ret = -1;
- goto out;
- }
+ if (loc->inode->ia_type != IA_IFDIR) {
+ ret = 0;
+ goto out;
+ }
- ret = quota_dict_set_size_meta (this, dict, &size);
- if (ret < 0)
- goto out;
+ dict = dict_new();
+ if (!dict) {
+ gf_log(this->name, GF_LOG_ERROR, "dict_new failed");
+ ret = -1;
+ goto out;
+ }
- ret = syncop_xattrop (FIRST_CHILD(this), loc,
- GF_XATTROP_ADD_ARRAY64_WITH_DEFAULT, dict, NULL,
- NULL, NULL);
+ ret = quota_dict_set_size_meta(this, dict, &size);
+ if (ret < 0)
+ goto out;
- if (ret < 0) {
- gf_log_callingfn (this->name, (-ret == ENOENT || -ret == ESTALE)
- ? GF_LOG_DEBUG:GF_LOG_ERROR, "xattrop failed "
- "for %s: %s", loc->path, strerror (-ret));
- goto out;
- }
+ ret = syncop_xattrop(FIRST_CHILD(this), loc,
+ GF_XATTROP_ADD_ARRAY64_WITH_DEFAULT, dict, NULL, NULL,
+ NULL);
+
+ if (ret < 0) {
+ gf_log_callingfn(
+ this->name,
+ (-ret == ENOENT || -ret == ESTALE) ? GF_LOG_DEBUG : GF_LOG_ERROR,
+ "xattrop failed "
+ "for %s: %s",
+ loc->path, strerror(-ret));
+ goto out;
+ }
out:
- if (dict)
- dict_unref (dict);
+ if (dict)
+ dict_unref(dict);
- return ret;
+ return ret;
}
int32_t
-mq_lock (xlator_t *this, loc_t *loc, short l_type)
+mq_lock(xlator_t *this, loc_t *loc, short l_type)
{
- struct gf_flock lock = {0, };
- int32_t ret = -1;
-
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
- GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
-
- gf_log (this->name, GF_LOG_DEBUG, "set lock type %d on %s",
- l_type, loc->path);
-
- lock.l_len = 0;
- lock.l_start = 0;
- lock.l_type = l_type;
- lock.l_whence = SEEK_SET;
-
- ret = syncop_inodelk (FIRST_CHILD(this), this->name, loc, F_SETLKW,
- &lock, NULL, NULL);
- if (ret < 0)
- gf_log_callingfn (this->name, (-ret == ENOENT || -ret == ESTALE)
- ? GF_LOG_DEBUG:GF_LOG_ERROR, "inodelk failed "
- "for %s: %s", loc->path, strerror (-ret));
+ struct gf_flock lock = {
+ 0,
+ };
+ int32_t ret = -1;
+
+ GF_VALIDATE_OR_GOTO("marker", loc, out);
+ GF_VALIDATE_OR_GOTO("marker", loc->inode, out);
+
+ gf_log(this->name, GF_LOG_DEBUG, "set lock type %d on %s", l_type,
+ loc->path);
+
+ lock.l_len = 0;
+ lock.l_start = 0;
+ lock.l_type = l_type;
+ lock.l_whence = SEEK_SET;
+
+ ret = syncop_inodelk(FIRST_CHILD(this), this->name, loc, F_SETLKW, &lock,
+ NULL, NULL);
+ if (ret < 0)
+ gf_log_callingfn(
+ this->name,
+ (-ret == ENOENT || -ret == ESTALE) ? GF_LOG_DEBUG : GF_LOG_ERROR,
+ "inodelk failed "
+ "for %s: %s",
+ loc->path, strerror(-ret));
out:
- return ret;
+ return ret;
}
int32_t
-mq_get_dirty (xlator_t *this, loc_t *loc, int32_t *dirty)
+mq_get_dirty(xlator_t *this, loc_t *loc, int32_t *dirty)
{
- int32_t ret = -1;
- int8_t value = 0;
- dict_t *dict = NULL;
- dict_t *rsp_dict = NULL;
- struct iatt stbuf = {0,};
-
- dict = dict_new ();
- if (dict == NULL) {
- gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
- goto out;
- }
-
- ret = dict_set_int64 (dict, QUOTA_DIRTY_KEY, 0);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING, "dict set failed");
- goto out;
- }
+ int32_t ret = -1;
+ int8_t value = 0;
+ dict_t *dict = NULL;
+ dict_t *rsp_dict = NULL;
+ struct iatt stbuf = {
+ 0,
+ };
+
+ dict = dict_new();
+ if (dict == NULL) {
+ gf_log(this->name, GF_LOG_ERROR, "dict_new failed");
+ goto out;
+ }
- ret = syncop_lookup (FIRST_CHILD(this), loc, &stbuf, NULL,
- dict, &rsp_dict);
- if (ret < 0) {
- gf_log_callingfn (this->name, (-ret == ENOENT || -ret == ESTALE)
- ? GF_LOG_DEBUG:GF_LOG_ERROR, "lookup failed "
- "for %s: %s", loc->path, strerror (-ret));
- goto out;
- }
+ ret = dict_set_int64(dict, QUOTA_DIRTY_KEY, 0);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_WARNING, "dict set failed");
+ goto out;
+ }
+
+ ret = syncop_lookup(FIRST_CHILD(this), loc, &stbuf, NULL, dict, &rsp_dict);
+ if (ret < 0) {
+ gf_log_callingfn(
+ this->name,
+ (-ret == ENOENT || -ret == ESTALE) ? GF_LOG_DEBUG : GF_LOG_ERROR,
+ "lookup failed "
+ "for %s: %s",
+ loc->path, strerror(-ret));
+ goto out;
+ }
- ret = dict_get_int8 (rsp_dict, QUOTA_DIRTY_KEY, &value);
- if (ret < 0)
- goto out;
+ ret = dict_get_int8(rsp_dict, QUOTA_DIRTY_KEY, &value);
+ if (ret < 0)
+ goto out;
- *dirty = value;
+ *dirty = value;
out:
- if (dict)
- dict_unref (dict);
+ if (dict)
+ dict_unref(dict);
- if (rsp_dict)
- dict_unref (rsp_dict);
+ if (rsp_dict)
+ dict_unref(rsp_dict);
- return ret;
+ return ret;
}
int32_t
-mq_get_set_dirty (xlator_t *this, loc_t *loc, int32_t dirty,
- int32_t *prev_dirty)
+mq_get_set_dirty(xlator_t *this, loc_t *loc, int32_t dirty, int32_t *prev_dirty)
{
- int32_t ret = -1;
- int8_t value = 0;
- quota_inode_ctx_t *ctx = NULL;
- dict_t *dict = NULL;
- dict_t *rsp_dict = NULL;
-
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
- GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
- GF_VALIDATE_OR_GOTO ("marker", prev_dirty, out);
-
- ret = mq_inode_ctx_get (loc->inode, this, &ctx);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "failed to get inode ctx for "
- "%s", loc->path);
- goto out;
- }
-
- dict = dict_new ();
- if (!dict) {
- gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
- ret = -1;
- goto out;
- }
-
- ret = dict_set_int8 (dict, QUOTA_DIRTY_KEY, dirty);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "dict_set failed");
- goto out;
- }
-
- ret = syncop_xattrop (FIRST_CHILD(this), loc, GF_XATTROP_GET_AND_SET,
- dict, NULL, NULL, &rsp_dict);
- if (ret < 0) {
- gf_log_callingfn (this->name, (-ret == ENOENT || -ret == ESTALE)
- ? GF_LOG_DEBUG:GF_LOG_ERROR, "xattrop failed "
- "for %s: %s", loc->path, strerror (-ret));
- goto out;
- }
+ int32_t ret = -1;
+ int8_t value = 0;
+ quota_inode_ctx_t *ctx = NULL;
+ dict_t *dict = NULL;
+ dict_t *rsp_dict = NULL;
+
+ GF_VALIDATE_OR_GOTO("marker", loc, out);
+ GF_VALIDATE_OR_GOTO("marker", loc->inode, out);
+ GF_VALIDATE_OR_GOTO("marker", prev_dirty, out);
+
+ ret = mq_inode_ctx_get(loc->inode, this, &ctx);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to get inode ctx for "
+ "%s",
+ loc->path);
+ goto out;
+ }
- *prev_dirty = 0;
- if (rsp_dict) {
- ret = dict_get_int8 (rsp_dict, QUOTA_DIRTY_KEY, &value);
- if (ret == 0)
- *prev_dirty = value;
- }
+ dict = dict_new();
+ if (!dict) {
+ gf_log(this->name, GF_LOG_ERROR, "dict_new failed");
+ ret = -1;
+ goto out;
+ }
- LOCK (&ctx->lock);
- {
- ctx->dirty = dirty;
- }
- UNLOCK (&ctx->lock);
- ret = 0;
+ ret = dict_set_int8(dict, QUOTA_DIRTY_KEY, dirty);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_ERROR, "dict_set failed");
+ goto out;
+ }
+
+ ret = syncop_xattrop(FIRST_CHILD(this), loc, GF_XATTROP_GET_AND_SET, dict,
+ NULL, NULL, &rsp_dict);
+ if (ret < 0) {
+ gf_log_callingfn(
+ this->name,
+ (-ret == ENOENT || -ret == ESTALE) ? GF_LOG_DEBUG : GF_LOG_ERROR,
+ "xattrop failed "
+ "for %s: %s",
+ loc->path, strerror(-ret));
+ goto out;
+ }
+
+ *prev_dirty = 0;
+ if (rsp_dict) {
+ ret = dict_get_int8(rsp_dict, QUOTA_DIRTY_KEY, &value);
+ if (ret == 0)
+ *prev_dirty = value;
+ }
+
+ LOCK(&ctx->lock);
+ {
+ ctx->dirty = dirty;
+ }
+ UNLOCK(&ctx->lock);
+ ret = 0;
out:
- if (dict)
- dict_unref (dict);
+ if (dict)
+ dict_unref(dict);
- if (rsp_dict)
- dict_unref (rsp_dict);
+ if (rsp_dict)
+ dict_unref(rsp_dict);
- return ret;
+ return ret;
}
int32_t
-mq_mark_dirty (xlator_t *this, loc_t *loc, int32_t dirty)
+mq_mark_dirty(xlator_t *this, loc_t *loc, int32_t dirty)
{
- int32_t ret = -1;
- dict_t *dict = NULL;
- quota_inode_ctx_t *ctx = NULL;
-
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
- GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
-
- ret = mq_inode_ctx_get (loc->inode, this, &ctx);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "failed to get inode ctx for "
- "%s", loc->path);
- ret = 0;
- goto out;
- }
-
- dict = dict_new ();
- if (!dict) {
- ret = -1;
- gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
- goto out;
- }
+ int32_t ret = -1;
+ dict_t *dict = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+
+ GF_VALIDATE_OR_GOTO("marker", loc, out);
+ GF_VALIDATE_OR_GOTO("marker", loc->inode, out);
+
+ ret = mq_inode_ctx_get(loc->inode, this, &ctx);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to get inode ctx for "
+ "%s",
+ loc->path);
+ ret = 0;
+ goto out;
+ }
- ret = dict_set_int8 (dict, QUOTA_DIRTY_KEY, dirty);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "dict_set failed");
- goto out;
- }
+ dict = dict_new();
+ if (!dict) {
+ ret = -1;
+ gf_log(this->name, GF_LOG_ERROR, "dict_new failed");
+ goto out;
+ }
- ret = syncop_setxattr (FIRST_CHILD(this), loc, dict, 0, NULL, NULL);
- if (ret < 0) {
- gf_log_callingfn (this->name, (-ret == ENOENT || -ret == ESTALE)
- ? GF_LOG_DEBUG:GF_LOG_ERROR, "setxattr dirty = %d "
- "failed for %s: %s", dirty, loc->path, strerror (-ret));
- goto out;
- }
+ ret = dict_set_int8(dict, QUOTA_DIRTY_KEY, dirty);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_ERROR, "dict_set failed");
+ goto out;
+ }
+
+ ret = syncop_setxattr(FIRST_CHILD(this), loc, dict, 0, NULL, NULL);
+ if (ret < 0) {
+ gf_log_callingfn(
+ this->name,
+ (-ret == ENOENT || -ret == ESTALE) ? GF_LOG_DEBUG : GF_LOG_ERROR,
+ "setxattr dirty = %d "
+ "failed for %s: %s",
+ dirty, loc->path, strerror(-ret));
+ goto out;
+ }
- LOCK (&ctx->lock);
- {
- ctx->dirty = dirty;
- }
- UNLOCK (&ctx->lock);
+ LOCK(&ctx->lock);
+ {
+ ctx->dirty = dirty;
+ }
+ UNLOCK(&ctx->lock);
out:
- if (dict)
- dict_unref (dict);
+ if (dict)
+ dict_unref(dict);
- return ret;
+ return ret;
}
int32_t
-_mq_get_metadata (xlator_t *this, loc_t *loc, quota_meta_t *contri,
- quota_meta_t *size, uuid_t contri_gfid)
+_mq_get_metadata(xlator_t *this, loc_t *loc, quota_meta_t *contri,
+ quota_meta_t *size, uuid_t contri_gfid)
{
- int32_t ret = -1;
- quota_meta_t meta = {0, };
- char contri_key[QUOTA_KEY_MAX] = {0, };
- char size_key[QUOTA_KEY_MAX] = {0, };
- dict_t *dict = NULL;
- dict_t *rsp_dict = NULL;
- struct iatt stbuf = {0,};
-
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
- GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
-
- if (size == NULL && contri == NULL)
- goto out;
+ int32_t ret = -1;
+ quota_meta_t meta = {
+ 0,
+ };
+ char contri_key[QUOTA_KEY_MAX] = {
+ 0,
+ };
+ char size_key[QUOTA_KEY_MAX] = {
+ 0,
+ };
+ int keylen = 0;
+ dict_t *dict = NULL;
+ dict_t *rsp_dict = NULL;
+ struct iatt stbuf = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("marker", loc, out);
+ GF_VALIDATE_OR_GOTO("marker", loc->inode, out);
+
+ if (size == NULL && contri == NULL)
+ goto out;
- dict = dict_new ();
- if (dict == NULL) {
- gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
- goto out;
- }
+ dict = dict_new();
+ if (dict == NULL) {
+ gf_log(this->name, GF_LOG_ERROR, "dict_new failed");
+ goto out;
+ }
- if (size && loc->inode->ia_type == IA_IFDIR) {
- GET_SIZE_KEY (this, size_key, ret);
- if (ret < 0)
- goto out;
- ret = dict_set_int64 (dict, size_key, 0);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "dict_set failed.");
- goto out;
- }
+ if (size && loc->inode->ia_type == IA_IFDIR) {
+ GET_SIZE_KEY(this, size_key, keylen);
+ if (keylen < 0)
+ goto out;
+ ret = dict_set_int64(dict, size_key, 0);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_ERROR, "dict_set failed.");
+ goto out;
}
+ }
- if (contri && !loc_is_root(loc)) {
- ret = mq_dict_set_contribution (this, dict, loc, contri_gfid,
- contri_key);
- if (ret < 0)
- goto out;
- }
+ if (contri && !loc_is_root(loc)) {
+ ret = mq_dict_set_contribution(this, dict, loc, contri_gfid,
+ contri_key);
+ if (ret < 0)
+ goto out;
+ }
+
+ ret = syncop_lookup(FIRST_CHILD(this), loc, &stbuf, NULL, dict, &rsp_dict);
+ if (ret < 0) {
+ gf_log_callingfn(
+ this->name,
+ (-ret == ENOENT || -ret == ESTALE) ? GF_LOG_DEBUG : GF_LOG_ERROR,
+ "lookup failed "
+ "for %s: %s",
+ loc->path, strerror(-ret));
+ goto out;
+ }
- ret = syncop_lookup (FIRST_CHILD(this), loc, &stbuf, NULL,
- dict, &rsp_dict);
- if (ret < 0) {
- gf_log_callingfn (this->name, (-ret == ENOENT || -ret == ESTALE)
- ? GF_LOG_DEBUG:GF_LOG_ERROR, "lookup failed "
- "for %s: %s", loc->path, strerror (-ret));
+ if (size) {
+ if (loc->inode->ia_type == IA_IFDIR) {
+ ret = quota_dict_get_meta(rsp_dict, size_key, keylen, &meta);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_ERROR, "dict_get failed.");
goto out;
- }
+ }
- if (size) {
- if (loc->inode->ia_type == IA_IFDIR) {
- ret = quota_dict_get_meta (rsp_dict, size_key,
- &meta);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR,
- "dict_get failed.");
- goto out;
- }
-
- size->size = meta.size;
- size->file_count = meta.file_count;
- size->dir_count = meta.dir_count;
- } else {
- size->size = stbuf.ia_blocks * 512;
- size->file_count = 1;
- size->dir_count = 0;
- }
+ size->size = meta.size;
+ size->file_count = meta.file_count;
+ size->dir_count = meta.dir_count;
+ } else {
+ size->size = stbuf.ia_blocks * 512;
+ size->file_count = 1;
+ size->dir_count = 0;
}
+ }
- if (contri && !loc_is_root(loc)) {
- ret = quota_dict_get_meta (rsp_dict, contri_key, &meta);
- if (ret < 0) {
- contri->size = 0;
- contri->file_count = 0;
- contri->dir_count = 0;
- } else {
- contri->size = meta.size;
- contri->file_count = meta.file_count;
- contri->dir_count = meta.dir_count;
- }
+ if (contri && !loc_is_root(loc)) {
+ ret = quota_dict_get_meta(rsp_dict, contri_key, strlen(contri_key),
+ &meta);
+ if (ret < 0) {
+ contri->size = 0;
+ contri->file_count = 0;
+ contri->dir_count = 0;
+ } else {
+ contri->size = meta.size;
+ contri->file_count = meta.file_count;
+ contri->dir_count = meta.dir_count;
}
+ }
- ret = 0;
+ ret = 0;
out:
- if (dict)
- dict_unref (dict);
+ if (dict)
+ dict_unref(dict);
- if (rsp_dict)
- dict_unref (rsp_dict);
+ if (rsp_dict)
+ dict_unref(rsp_dict);
- return ret;
+ return ret;
}
int32_t
-mq_get_metadata (xlator_t *this, loc_t *loc, quota_meta_t *contri,
- quota_meta_t *size, quota_inode_ctx_t *ctx,
- inode_contribution_t *contribution)
+mq_get_metadata(xlator_t *this, loc_t *loc, quota_meta_t *contri,
+ quota_meta_t *size, quota_inode_ctx_t *ctx,
+ inode_contribution_t *contribution)
{
- int32_t ret = -1;
+ int32_t ret = -1;
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
- GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
- GF_VALIDATE_OR_GOTO ("marker", ctx, out);
- GF_VALIDATE_OR_GOTO ("marker", contribution, out);
+ GF_VALIDATE_OR_GOTO("marker", loc, out);
+ GF_VALIDATE_OR_GOTO("marker", loc->inode, out);
+ GF_VALIDATE_OR_GOTO("marker", ctx, out);
+ GF_VALIDATE_OR_GOTO("marker", contribution, out);
- if (size == NULL && contri == NULL) {
- ret = 0;
- goto out;
- }
+ if (size == NULL && contri == NULL) {
+ ret = 0;
+ goto out;
+ }
- ret = _mq_get_metadata (this, loc, contri, size, contribution->gfid);
- if (ret < 0)
- goto out;
+ ret = _mq_get_metadata(this, loc, contri, size, contribution->gfid);
+ if (ret < 0)
+ goto out;
- if (size) {
- LOCK (&ctx->lock);
- {
- ctx->size = size->size;
- ctx->file_count = size->file_count;
- ctx->dir_count = size->dir_count;
- }
- UNLOCK (&ctx->lock);
+ if (size) {
+ LOCK(&ctx->lock);
+ {
+ ctx->size = size->size;
+ ctx->file_count = size->file_count;
+ ctx->dir_count = size->dir_count;
}
+ UNLOCK(&ctx->lock);
+ }
- if (contri) {
- LOCK (&contribution->lock);
- {
- contribution->contribution = contri->size;
- contribution->file_count = contri->file_count;
- contribution->dir_count = contri->dir_count;
- }
- UNLOCK (&contribution->lock);
+ if (contri) {
+ LOCK(&contribution->lock);
+ {
+ contribution->contribution = contri->size;
+ contribution->file_count = contri->file_count;
+ contribution->dir_count = contri->dir_count;
}
+ UNLOCK(&contribution->lock);
+ }
out:
- return ret;
-}
-
-int32_t
-mq_get_size (xlator_t *this, loc_t *loc, quota_meta_t *size)
-{
- return _mq_get_metadata (this, loc, NULL, size, 0);
+ return ret;
}
int32_t
-mq_get_contri (xlator_t *this, loc_t *loc, quota_meta_t *contri,
- uuid_t contri_gfid)
+mq_get_delta(xlator_t *this, loc_t *loc, quota_meta_t *delta,
+ quota_inode_ctx_t *ctx, inode_contribution_t *contribution)
{
- return _mq_get_metadata (this, loc, contri, NULL, contri_gfid);
-}
-
-int32_t
-mq_get_delta (xlator_t *this, loc_t *loc, quota_meta_t *delta,
- quota_inode_ctx_t *ctx, inode_contribution_t *contribution)
-{
- int32_t ret = -1;
- quota_meta_t size = {0, };
- quota_meta_t contri = {0, };
-
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
- GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
- GF_VALIDATE_OR_GOTO ("marker", ctx, out);
- GF_VALIDATE_OR_GOTO ("marker", contribution, out);
-
- ret = mq_get_metadata (this, loc, &contri, &size, ctx, contribution);
- if (ret < 0)
- goto out;
+ int32_t ret = -1;
+ quota_meta_t size = {
+ 0,
+ };
+ quota_meta_t contri = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("marker", loc, out);
+ GF_VALIDATE_OR_GOTO("marker", loc->inode, out);
+ GF_VALIDATE_OR_GOTO("marker", ctx, out);
+ GF_VALIDATE_OR_GOTO("marker", contribution, out);
+
+ ret = mq_get_metadata(this, loc, &contri, &size, ctx, contribution);
+ if (ret < 0)
+ goto out;
- mq_compute_delta (delta, &size, &contri);
+ mq_compute_delta(delta, &size, &contri);
out:
- return ret;
+ return ret;
}
int32_t
-mq_remove_contri (xlator_t *this, loc_t *loc, quota_inode_ctx_t *ctx,
- inode_contribution_t *contri, quota_meta_t *delta,
- uint32_t nlink)
+mq_remove_contri(xlator_t *this, loc_t *loc, quota_inode_ctx_t *ctx,
+ inode_contribution_t *contri, quota_meta_t *delta,
+ uint32_t nlink)
{
- int32_t ret = -1;
- char contri_key[QUOTA_KEY_MAX] = {0, };
+ int32_t ret = -1;
+ char contri_key[QUOTA_KEY_MAX] = {
+ 0,
+ };
- if (nlink == 1) {
- /*File was a last link and has been deleted */
- ret = 0;
- goto done;
- }
-
- GET_CONTRI_KEY (this, contri_key, contri->gfid, ret);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "get contri_key "
- "failed for %s", uuid_utoa(contri->gfid));
- goto out;
- }
-
- ret = syncop_removexattr (FIRST_CHILD(this), loc, contri_key, 0, NULL);
- if (ret < 0) {
- if (-ret == ENOENT || -ret == ESTALE || -ret == ENODATA ||
- -ret == ENOATTR) {
- /* Remove contri in done when unlink operation is
- * performed, so return success on ENOENT/ESTSLE
- * rename operation removes xattr earlier,
- * so return success on ENODATA
- */
- ret = 0;
- } else {
- gf_log_callingfn (this->name, GF_LOG_ERROR,
- "removexattr %s failed for %s: %s",
- contri_key, loc->path,
- strerror (-ret));
- goto out;
- }
+ if (nlink == 1) {
+ /*File was a last link and has been deleted */
+ ret = 0;
+ goto done;
+ }
+
+ GET_CONTRI_KEY(this, contri_key, contri->gfid, ret);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "get contri_key "
+ "failed for %s",
+ uuid_utoa(contri->gfid));
+ goto out;
+ }
+
+ ret = syncop_removexattr(FIRST_CHILD(this), loc, contri_key, 0, NULL);
+ if (ret < 0) {
+ if (-ret == ENOENT || -ret == ESTALE || -ret == ENODATA ||
+ -ret == ENOATTR) {
+ /* Remove contri in done when unlink operation is
+ * performed, so return success on ENOENT/ESTSLE
+ * rename operation removes xattr earlier,
+ * so return success on ENODATA
+ */
+ ret = 0;
+ } else {
+ gf_log_callingfn(this->name, GF_LOG_ERROR,
+ "removexattr %s failed for %s: %s", contri_key,
+ loc->path, strerror(-ret));
+ goto out;
}
+ }
done:
- LOCK (&contri->lock);
- {
- contri->contribution += delta->size;
- contri->file_count += delta->file_count;
- contri->dir_count += delta->dir_count;
- }
- UNLOCK (&contri->lock);
+ LOCK(&contri->lock);
+ {
+ contri->contribution += delta->size;
+ contri->file_count += delta->file_count;
+ contri->dir_count += delta->dir_count;
+ }
+ UNLOCK(&contri->lock);
- ret = 0;
+ ret = 0;
out:
- QUOTA_FREE_CONTRIBUTION_NODE (ctx, contri);
+ QUOTA_FREE_CONTRIBUTION_NODE(ctx, contri);
- return ret;
+ return ret;
}
int32_t
-mq_update_contri (xlator_t *this, loc_t *loc, inode_contribution_t *contri,
- quota_meta_t *delta)
+mq_update_contri(xlator_t *this, loc_t *loc, inode_contribution_t *contri,
+ quota_meta_t *delta)
{
- int32_t ret = -1;
- char contri_key[QUOTA_KEY_MAX] = {0, };
- dict_t *dict = NULL;
-
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
- GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
- GF_VALIDATE_OR_GOTO ("marker", delta, out);
- GF_VALIDATE_OR_GOTO ("marker", contri, out);
-
- if (quota_meta_is_null (delta)) {
- ret = 0;
- goto out;
- }
-
- dict = dict_new ();
- if (!dict) {
- gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
- ret = -1;
- goto out;
- }
+ int32_t ret = -1;
+ char contri_key[QUOTA_KEY_MAX] = {
+ 0,
+ };
+ dict_t *dict = NULL;
+
+ GF_VALIDATE_OR_GOTO("marker", loc, out);
+ GF_VALIDATE_OR_GOTO("marker", loc->inode, out);
+ GF_VALIDATE_OR_GOTO("marker", delta, out);
+ GF_VALIDATE_OR_GOTO("marker", contri, out);
+
+ if (quota_meta_is_null(delta)) {
+ ret = 0;
+ goto out;
+ }
- GET_CONTRI_KEY (this, contri_key, contri->gfid, ret);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "get contri_key "
- "failed for %s", uuid_utoa(contri->gfid));
- goto out;
- }
+ dict = dict_new();
+ if (!dict) {
+ gf_log(this->name, GF_LOG_ERROR, "dict_new failed");
+ ret = -1;
+ goto out;
+ }
+
+ GET_CONTRI_KEY(this, contri_key, contri->gfid, ret);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "get contri_key "
+ "failed for %s",
+ uuid_utoa(contri->gfid));
+ goto out;
+ }
- ret = quota_dict_set_meta (dict, contri_key, delta,
- loc->inode->ia_type);
- if (ret < 0)
- goto out;
+ ret = quota_dict_set_meta(dict, contri_key, delta, loc->inode->ia_type);
+ if (ret < 0)
+ goto out;
- ret = syncop_xattrop(FIRST_CHILD(this), loc, GF_XATTROP_ADD_ARRAY64,
- dict, NULL, NULL, NULL);
- if (ret < 0) {
- gf_log_callingfn (this->name, (-ret == ENOENT || -ret == ESTALE)
- ? GF_LOG_DEBUG:GF_LOG_ERROR, "xattrop failed "
- "for %s: %s", loc->path, strerror (-ret));
- goto out;
- }
+ ret = syncop_xattrop(FIRST_CHILD(this), loc, GF_XATTROP_ADD_ARRAY64, dict,
+ NULL, NULL, NULL);
+ if (ret < 0) {
+ gf_log_callingfn(
+ this->name,
+ (-ret == ENOENT || -ret == ESTALE) ? GF_LOG_DEBUG : GF_LOG_ERROR,
+ "xattrop failed "
+ "for %s: %s",
+ loc->path, strerror(-ret));
+ goto out;
+ }
- LOCK (&contri->lock);
- {
- contri->contribution += delta->size;
- contri->file_count += delta->file_count;
- contri->dir_count += delta->dir_count;
- }
- UNLOCK (&contri->lock);
+ LOCK(&contri->lock);
+ {
+ contri->contribution += delta->size;
+ contri->file_count += delta->file_count;
+ contri->dir_count += delta->dir_count;
+ }
+ UNLOCK(&contri->lock);
out:
- if (dict)
- dict_unref (dict);
+ if (dict)
+ dict_unref(dict);
- return ret;
+ return ret;
}
int32_t
-mq_update_size (xlator_t *this, loc_t *loc, quota_meta_t *delta)
+mq_update_size(xlator_t *this, loc_t *loc, quota_meta_t *delta)
{
- int32_t ret = -1;
- quota_inode_ctx_t *ctx = NULL;
- dict_t *dict = NULL;
+ int32_t ret = -1;
+ quota_inode_ctx_t *ctx = NULL;
+ dict_t *dict = NULL;
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
- GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
- GF_VALIDATE_OR_GOTO ("marker", delta, out);
+ GF_VALIDATE_OR_GOTO("marker", loc, out);
+ GF_VALIDATE_OR_GOTO("marker", loc->inode, out);
+ GF_VALIDATE_OR_GOTO("marker", delta, out);
- if (quota_meta_is_null (delta)) {
- ret = 0;
- goto out;
- }
-
- ret = mq_inode_ctx_get (loc->inode, this, &ctx);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "failed to get inode ctx for "
- "%s", loc->path);
- goto out;
- }
-
- dict = dict_new ();
- if (!dict) {
- gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
- ret = -1;
- goto out;
- }
+ if (quota_meta_is_null(delta)) {
+ ret = 0;
+ goto out;
+ }
+
+ ret = mq_inode_ctx_get(loc->inode, this, &ctx);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to get inode ctx for "
+ "%s",
+ loc->path);
+ goto out;
+ }
- ret = quota_dict_set_size_meta (this, dict, delta);
- if (ret < 0)
- goto out;
+ dict = dict_new();
+ if (!dict) {
+ gf_log(this->name, GF_LOG_ERROR, "dict_new failed");
+ ret = -1;
+ goto out;
+ }
- ret = syncop_xattrop(FIRST_CHILD(this), loc,
- GF_XATTROP_ADD_ARRAY64_WITH_DEFAULT, dict, NULL,
- NULL, NULL);
- if (ret < 0) {
- gf_log_callingfn (this->name, (-ret == ENOENT || -ret == ESTALE)
- ? GF_LOG_DEBUG:GF_LOG_ERROR, "xattrop failed "
- "for %s: %s", loc->path, strerror (-ret));
- goto out;
- }
+ ret = quota_dict_set_size_meta(this, dict, delta);
+ if (ret < 0)
+ goto out;
- LOCK (&ctx->lock);
- {
- ctx->size += delta->size;
- ctx->file_count += delta->file_count;
- if (ctx->dir_count == 0)
- ctx->dir_count += delta->dir_count + 1;
- else
- ctx->dir_count += delta->dir_count;
- }
- UNLOCK (&ctx->lock);
+ ret = syncop_xattrop(FIRST_CHILD(this), loc,
+ GF_XATTROP_ADD_ARRAY64_WITH_DEFAULT, dict, NULL, NULL,
+ NULL);
+ if (ret < 0) {
+ gf_log_callingfn(
+ this->name,
+ (-ret == ENOENT || -ret == ESTALE) ? GF_LOG_DEBUG : GF_LOG_ERROR,
+ "xattrop failed "
+ "for %s: %s",
+ loc->path, strerror(-ret));
+ goto out;
+ }
+
+ LOCK(&ctx->lock);
+ {
+ ctx->size += delta->size;
+ ctx->file_count += delta->file_count;
+ if (ctx->dir_count == 0)
+ ctx->dir_count += delta->dir_count + 1;
+ else
+ ctx->dir_count += delta->dir_count;
+ }
+ UNLOCK(&ctx->lock);
out:
- if (dict)
- dict_unref (dict);
+ if (dict)
+ dict_unref(dict);
- return ret;
+ return ret;
}
int
-mq_synctask_cleanup (int ret, call_frame_t *frame, void *opaque)
+mq_synctask_cleanup(int ret, call_frame_t *frame, void *opaque)
{
- quota_synctask_t *args = NULL;
+ quota_synctask_t *args = NULL;
- GF_ASSERT (opaque);
+ GF_ASSERT(opaque);
- args = (quota_synctask_t *) opaque;
- loc_wipe (&args->loc);
+ args = (quota_synctask_t *)opaque;
+ loc_wipe(&args->loc);
- if (args->stub)
- call_resume (args->stub);
+ if (args->stub)
+ call_resume(args->stub);
- if (!args->is_static)
- GF_FREE (args);
+ if (!args->is_static)
+ GF_FREE(args);
- return 0;
+ return 0;
}
int
-mq_synctask1 (xlator_t *this, synctask_fn_t task, gf_boolean_t spawn,
- loc_t *loc, quota_meta_t *contri, uint32_t nlink,
- call_stub_t *stub)
+mq_synctask1(xlator_t *this, synctask_fn_t task, gf_boolean_t spawn, loc_t *loc,
+ quota_meta_t *contri, uint32_t nlink, call_stub_t *stub)
{
- int32_t ret = -1;
- quota_synctask_t *args = NULL;
- quota_synctask_t static_args = {0, };
-
- if (spawn) {
- QUOTA_ALLOC_OR_GOTO (args, quota_synctask_t, ret, out);
- args->is_static = _gf_false;
- } else {
- args = &static_args;
- args->is_static = _gf_true;
- }
-
- args->this = this;
- args->stub = stub;
- loc_copy (&args->loc, loc);
- args->ia_nlink = nlink;
-
- if (contri) {
- args->contri = *contri;
- } else {
- args->contri.size = -1;
- args->contri.file_count = -1;
- args->contri.dir_count = -1;
- }
-
- if (spawn) {
- ret = synctask_new1 (this->ctx->env, 1024 * 16, task,
- mq_synctask_cleanup, NULL, args);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "Failed to spawn "
- "new synctask");
- mq_synctask_cleanup (ret, NULL, args);
- }
- } else {
- ret = task (args);
- mq_synctask_cleanup (ret, NULL, args);
- }
+ int32_t ret = -1;
+ quota_synctask_t *args = NULL;
+ quota_synctask_t static_args = {
+ 0,
+ };
+
+ if (spawn) {
+ QUOTA_ALLOC_OR_GOTO(args, quota_synctask_t, ret, out);
+ args->is_static = _gf_false;
+ } else {
+ args = &static_args;
+ args->is_static = _gf_true;
+ }
+
+ args->this = this;
+ args->stub = stub;
+ loc_copy(&args->loc, loc);
+ args->ia_nlink = nlink;
+
+ if (contri) {
+ args->contri = *contri;
+ } else {
+ args->contri.size = -1;
+ args->contri.file_count = -1;
+ args->contri.dir_count = -1;
+ }
+
+ if (spawn) {
+ ret = synctask_new1(this->ctx->env, 1024 * 16, task,
+ mq_synctask_cleanup, NULL, args);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Failed to spawn "
+ "new synctask");
+ mq_synctask_cleanup(ret, NULL, args);
+ }
+ } else {
+ ret = task(args);
+ mq_synctask_cleanup(ret, NULL, args);
+ }
out:
- return ret;
+ return ret;
}
int
-mq_synctask (xlator_t *this, synctask_fn_t task, gf_boolean_t spawn, loc_t *loc)
+mq_synctask(xlator_t *this, synctask_fn_t task, gf_boolean_t spawn, loc_t *loc)
{
- return mq_synctask1 (this, task, spawn, loc, NULL, -1, NULL);
+ return mq_synctask1(this, task, spawn, loc, NULL, -1, NULL);
}
int32_t
-mq_prevalidate_txn (xlator_t *this, loc_t *origin_loc, loc_t *loc,
- quota_inode_ctx_t **ctx, struct iatt *buf)
+mq_prevalidate_txn(xlator_t *this, loc_t *origin_loc, loc_t *loc,
+ quota_inode_ctx_t **ctx, struct iatt *buf)
{
- int32_t ret = -1;
- quota_inode_ctx_t *ctxtmp = NULL;
+ int32_t ret = -1;
+ quota_inode_ctx_t *ctxtmp = NULL;
- if (buf) {
- if (buf->ia_type == IA_IFREG && IS_DHT_LINKFILE_MODE(buf))
- goto out;
+ if (buf) {
+ if (buf->ia_type == IA_IFREG && IS_DHT_LINKFILE_MODE(buf))
+ goto out;
- if (buf->ia_type != IA_IFREG && buf->ia_type != IA_IFLNK &&
- buf->ia_type != IA_IFDIR)
- goto out;
- }
+ if (buf->ia_type != IA_IFREG && buf->ia_type != IA_IFLNK &&
+ buf->ia_type != IA_IFDIR)
+ goto out;
+ }
- if (origin_loc == NULL || origin_loc->inode == NULL ||
- gf_uuid_is_null(origin_loc->inode->gfid))
- goto out;
+ if (origin_loc == NULL || origin_loc->inode == NULL ||
+ gf_uuid_is_null(origin_loc->inode->gfid))
+ goto out;
- loc_copy (loc, origin_loc);
+ loc_copy(loc, origin_loc);
- if (gf_uuid_is_null (loc->gfid))
- gf_uuid_copy (loc->gfid, loc->inode->gfid);
+ if (gf_uuid_is_null(loc->gfid))
+ gf_uuid_copy(loc->gfid, loc->inode->gfid);
- if (!loc_is_root(loc) && loc->parent == NULL)
- loc->parent = inode_parent (loc->inode, 0, NULL);
+ if (!loc_is_root(loc) && loc->parent == NULL)
+ loc->parent = inode_parent(loc->inode, 0, NULL);
- ret = mq_inode_ctx_get (loc->inode, this, &ctxtmp);
- if (ret < 0) {
- gf_log_callingfn (this->name, GF_LOG_WARNING, "inode ctx for "
- "is NULL for %s", loc->path);
- goto out;
- }
- if (ctx)
- *ctx = ctxtmp;
+ ret = mq_inode_ctx_get(loc->inode, this, &ctxtmp);
+ if (ret < 0) {
+ gf_log_callingfn(this->name, GF_LOG_WARNING,
+ "inode ctx for "
+ "is NULL for %s",
+ loc->path);
+ goto out;
+ }
+ if (ctx)
+ *ctx = ctxtmp;
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-mq_create_xattrs_task (void *opaque)
+mq_create_xattrs_task(void *opaque)
{
- int32_t ret = -1;
- gf_boolean_t locked = _gf_false;
- gf_boolean_t contri_set = _gf_false;
- gf_boolean_t size_set = _gf_false;
- gf_boolean_t need_txn = _gf_false;
- quota_synctask_t *args = NULL;
- quota_inode_ctx_t *ctx = NULL;
- xlator_t *this = NULL;
- loc_t *loc = NULL;
- gf_boolean_t status = _gf_false;
-
- GF_ASSERT (opaque);
-
- args = (quota_synctask_t *) opaque;
- loc = &args->loc;
- this = args->this;
- THIS = this;
-
- ret = mq_inode_ctx_get (loc->inode, this, &ctx);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING, "Failed to"
- "get inode ctx, aborting quota create txn");
- goto out;
- }
+ int32_t ret = -1;
+ gf_boolean_t locked = _gf_false;
+ gf_boolean_t contri_set = _gf_false;
+ gf_boolean_t size_set = _gf_false;
+ gf_boolean_t need_txn = _gf_false;
+ quota_synctask_t *args = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ xlator_t *this = NULL;
+ loc_t *loc = NULL;
+ gf_boolean_t status = _gf_false;
+
+ GF_ASSERT(opaque);
+
+ args = (quota_synctask_t *)opaque;
+ loc = &args->loc;
+ this = args->this;
+ THIS = this;
+
+ ret = mq_inode_ctx_get(loc->inode, this, &ctx);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "Failed to"
+ "get inode ctx, aborting quota create txn");
+ goto out;
+ }
- if (loc->inode->ia_type == IA_IFDIR) {
- /* lock not required for files */
- ret = mq_lock (this, loc, F_WRLCK);
- if (ret < 0)
- goto out;
- locked = _gf_true;
- }
+ if (loc->inode->ia_type == IA_IFDIR) {
+ /* lock not required for files */
+ ret = mq_lock(this, loc, F_WRLCK);
+ if (ret < 0)
+ goto out;
+ locked = _gf_true;
+ }
- ret = mq_are_xattrs_set (this, loc, &contri_set, &size_set);
- if (ret < 0 || (contri_set && size_set))
- goto out;
+ ret = mq_are_xattrs_set(this, loc, &contri_set, &size_set);
+ if (ret < 0 || (contri_set && size_set))
+ goto out;
- mq_set_ctx_create_status (ctx, _gf_false);
- status = _gf_true;
+ mq_set_ctx_create_status(ctx, _gf_false);
+ status = _gf_true;
- if (loc->inode->ia_type == IA_IFDIR && size_set == _gf_false) {
- ret = mq_create_size_xattrs (this, ctx, loc);
- if (ret < 0)
- goto out;
- }
+ if (loc->inode->ia_type == IA_IFDIR && size_set == _gf_false) {
+ ret = mq_create_size_xattrs(this, ctx, loc);
+ if (ret < 0)
+ goto out;
+ }
- need_txn = _gf_true;
+ need_txn = _gf_true;
out:
- if (locked)
- ret = mq_lock (this, loc, F_UNLCK);
+ if (locked)
+ ret = mq_lock(this, loc, F_UNLCK);
- if (status == _gf_false)
- mq_set_ctx_create_status (ctx, _gf_false);
+ if (status == _gf_false)
+ mq_set_ctx_create_status(ctx, _gf_false);
- if (need_txn)
- ret = mq_initiate_quota_blocking_txn (this, loc, NULL);
+ if (need_txn)
+ ret = mq_initiate_quota_blocking_txn(this, loc, NULL);
- return ret;
+ return ret;
}
static int
-_mq_create_xattrs_txn (xlator_t *this, loc_t *origin_loc, struct iatt *buf,
- gf_boolean_t spawn)
+_mq_create_xattrs_txn(xlator_t *this, loc_t *origin_loc, struct iatt *buf,
+ gf_boolean_t spawn)
{
- int32_t ret = -1;
- quota_inode_ctx_t *ctx = NULL;
- gf_boolean_t status = _gf_true;
- loc_t loc = {0, };
- inode_contribution_t *contribution = NULL;
-
- ret = mq_prevalidate_txn (this, origin_loc, &loc, &ctx, buf);
- if (ret < 0)
- goto out;
+ int32_t ret = -1;
+ quota_inode_ctx_t *ctx = NULL;
+ gf_boolean_t status = _gf_true;
+ loc_t loc = {
+ 0,
+ };
+ inode_contribution_t *contribution = NULL;
+
+ ret = mq_prevalidate_txn(this, origin_loc, &loc, &ctx, buf);
+ if (ret < 0)
+ goto out;
- ret = mq_test_and_set_ctx_create_status (ctx, &status);
- if (ret < 0 || status == _gf_true)
- goto out;
+ ret = mq_test_and_set_ctx_create_status(ctx, &status);
+ if (ret < 0 || status == _gf_true)
+ goto out;
- if (!loc_is_root(&loc) && loc.parent) {
- contribution = mq_add_new_contribution_node (this, ctx, &loc);
- if (contribution == NULL) {
- gf_log (this->name, GF_LOG_WARNING,
- "cannot add a new contribution node "
- "(%s)", uuid_utoa (loc.gfid));
- ret = -1;
- goto out;
- } else {
- GF_REF_PUT (contribution);
- }
+ if (!loc_is_root(&loc) && loc.parent) {
+ contribution = mq_add_new_contribution_node(this, ctx, &loc);
+ if (contribution == NULL) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "cannot add a new contribution node "
+ "(%s)",
+ uuid_utoa(loc.gfid));
+ ret = -1;
+ goto out;
+ } else {
+ GF_REF_PUT(contribution);
}
+ }
- ret = mq_synctask (this, mq_create_xattrs_task, spawn, &loc);
+ ret = mq_synctask(this, mq_create_xattrs_task, spawn, &loc);
out:
- if (ret < 0 && status == _gf_false)
- mq_set_ctx_create_status (ctx, _gf_false);
-
- loc_wipe (&loc);
- return ret;
-}
-
-int
-mq_create_xattrs_txn (xlator_t *this, loc_t *loc, struct iatt *buf)
-{
- int32_t ret = -1;
+ if (ret < 0 && status == _gf_false)
+ mq_set_ctx_create_status(ctx, _gf_false);
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
- GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
-
- ret = _mq_create_xattrs_txn (this, loc, buf, _gf_true);
-out:
- return ret;
+ loc_wipe(&loc);
+ return ret;
}
int
-mq_create_xattrs_blocking_txn (xlator_t *this, loc_t *loc, struct iatt *buf)
+mq_create_xattrs_txn(xlator_t *this, loc_t *loc, struct iatt *buf)
{
- int32_t ret = -1;
+ int32_t ret = -1;
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
- GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
+ GF_VALIDATE_OR_GOTO("marker", loc, out);
+ GF_VALIDATE_OR_GOTO("marker", loc->inode, out);
- ret = _mq_create_xattrs_txn (this, loc, buf, _gf_false);
+ ret = _mq_create_xattrs_txn(this, loc, buf, _gf_true);
out:
- return ret;
+ return ret;
}
int32_t
-mq_reduce_parent_size_task (void *opaque)
+mq_reduce_parent_size_task(void *opaque)
{
- int32_t ret = -1;
- int32_t prev_dirty = 0;
- quota_inode_ctx_t *ctx = NULL;
- quota_inode_ctx_t *parent_ctx = NULL;
- inode_contribution_t *contribution = NULL;
- quota_meta_t delta = {0, };
- quota_meta_t contri = {0, };
- loc_t parent_loc = {0,};
- gf_boolean_t locked = _gf_false;
- gf_boolean_t dirty = _gf_false;
- quota_synctask_t *args = NULL;
- xlator_t *this = NULL;
- loc_t *loc = NULL;
- gf_boolean_t remove_xattr = _gf_true;
- uint32_t nlink = 0;
-
- GF_ASSERT (opaque);
-
- args = (quota_synctask_t *) opaque;
- loc = &args->loc;
- contri = args->contri;
- nlink = args->ia_nlink;
- this = args->this;
- THIS = this;
-
- ret = mq_inode_loc_fill (NULL, loc->parent, &parent_loc);
+ int32_t ret = -1;
+ int32_t prev_dirty = 0;
+ quota_inode_ctx_t *ctx = NULL;
+ quota_inode_ctx_t *parent_ctx = NULL;
+ inode_contribution_t *contribution = NULL;
+ quota_meta_t delta = {
+ 0,
+ };
+ quota_meta_t contri = {
+ 0,
+ };
+ loc_t parent_loc = {
+ 0,
+ };
+ gf_boolean_t locked = _gf_false;
+ gf_boolean_t dirty = _gf_false;
+ quota_synctask_t *args = NULL;
+ xlator_t *this = NULL;
+ loc_t *loc = NULL;
+ gf_boolean_t remove_xattr = _gf_true;
+ uint32_t nlink = 0;
+
+ GF_ASSERT(opaque);
+
+ args = (quota_synctask_t *)opaque;
+ loc = &args->loc;
+ contri = args->contri;
+ nlink = args->ia_nlink;
+ this = args->this;
+ THIS = this;
+
+ ret = mq_inode_loc_fill(NULL, loc->parent, &parent_loc);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "parent_loc fill failed for "
+ "child inode %s: ",
+ uuid_utoa(loc->inode->gfid));
+ goto out;
+ }
+
+ ret = mq_lock(this, &parent_loc, F_WRLCK);
+ if (ret < 0)
+ goto out;
+ locked = _gf_true;
+
+ if (contri.size >= 0) {
+ /* contri parameter is supplied only for rename operation.
+ * remove xattr is alreday performed, we need to skip
+ * removexattr for rename operation
+ */
+ remove_xattr = _gf_false;
+ delta.size = contri.size;
+ delta.file_count = contri.file_count;
+ delta.dir_count = contri.dir_count;
+ } else {
+ remove_xattr = _gf_true;
+
+ ret = mq_inode_ctx_get(loc->inode, this, &ctx);
if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "parent_loc fill failed for "
- "child inode %s: ", uuid_utoa (loc->inode->gfid));
- goto out;
+ gf_log_callingfn(this->name, GF_LOG_WARNING,
+ "ctx for"
+ " the node %s is NULL",
+ loc->path);
+ goto out;
}
- ret = mq_lock (this, &parent_loc, F_WRLCK);
- if (ret < 0)
- goto out;
- locked = _gf_true;
+ contribution = mq_get_contribution_node(loc->parent, ctx);
+ if (contribution == NULL) {
+ ret = -1;
+ gf_log(this->name, GF_LOG_DEBUG,
+ "contribution for the node %s is NULL", loc->path);
+ goto out;
+ }
- if (contri.size >= 0) {
- /* contri paramater is supplied only for rename operation.
- * remove xattr is alreday performed, we need to skip
- * removexattr for rename operation
- */
- remove_xattr = _gf_false;
- delta.size = contri.size;
- delta.file_count = contri.file_count;
- delta.dir_count = contri.dir_count;
- } else {
- remove_xattr = _gf_true;
-
- ret = mq_inode_ctx_get (loc->inode, this, &ctx);
- if (ret < 0) {
- gf_log_callingfn (this->name, GF_LOG_WARNING, "ctx for"
- " the node %s is NULL", loc->path);
- goto out;
- }
-
- contribution = mq_get_contribution_node (loc->parent, ctx);
- if (contribution == NULL) {
- ret = -1;
- gf_log (this->name, GF_LOG_DEBUG,
- "contribution for the node %s is NULL",
- loc->path);
- goto out;
- }
-
- LOCK (&contribution->lock);
- {
- delta.size = contribution->contribution;
- delta.file_count = contribution->file_count;
- delta.dir_count = contribution->dir_count;
- }
- UNLOCK (&contribution->lock);
+ LOCK(&contribution->lock);
+ {
+ delta.size = contribution->contribution;
+ delta.file_count = contribution->file_count;
+ delta.dir_count = contribution->dir_count;
}
+ UNLOCK(&contribution->lock);
+ }
- ret = mq_get_set_dirty (this, &parent_loc, 1, &prev_dirty);
- if (ret < 0)
- goto out;
- dirty = _gf_true;
+ ret = mq_get_set_dirty(this, &parent_loc, 1, &prev_dirty);
+ if (ret < 0)
+ goto out;
+ dirty = _gf_true;
- mq_sub_meta (&delta, NULL);
+ mq_sub_meta(&delta, NULL);
- if (remove_xattr) {
- ret = mq_remove_contri (this, loc, ctx, contribution, &delta,
- nlink);
- if (ret < 0)
- goto out;
- }
+ if (remove_xattr) {
+ ret = mq_remove_contri(this, loc, ctx, contribution, &delta, nlink);
+ if (ret < 0)
+ goto out;
+ }
- if (quota_meta_is_null (&delta))
- goto out;
+ if (quota_meta_is_null(&delta))
+ goto out;
- ret = mq_update_size (this, &parent_loc, &delta);
- if (ret < 0)
- goto out;
+ ret = mq_update_size(this, &parent_loc, &delta);
+ if (ret < 0)
+ goto out;
out:
- if (dirty) {
- if (ret < 0 || prev_dirty) {
- /* On failure clear dirty status flag.
- * In the next lookup inspect_directory_xattr
- * can set the status flag and fix the
- * dirty directory.
- * Do the same if dir was dirty before
- * the txn
- */
- ret = mq_inode_ctx_get (parent_loc.inode, this,
- &parent_ctx);
- if (ret == 0)
- mq_set_ctx_dirty_status (parent_ctx, _gf_false);
- } else {
- ret = mq_mark_dirty (this, &parent_loc, 0);
- }
+ if (dirty) {
+ if (ret < 0 || prev_dirty) {
+ /* On failure clear dirty status flag.
+ * In the next lookup inspect_directory_xattr
+ * can set the status flag and fix the
+ * dirty directory.
+ * Do the same if dir was dirty before
+ * the txn
+ */
+ ret = mq_inode_ctx_get(parent_loc.inode, this, &parent_ctx);
+ if (ret == 0)
+ mq_set_ctx_dirty_status(parent_ctx, _gf_false);
+ } else {
+ ret = mq_mark_dirty(this, &parent_loc, 0);
}
+ }
- if (locked)
- ret = mq_lock (this, &parent_loc, F_UNLCK);
+ if (locked)
+ ret = mq_lock(this, &parent_loc, F_UNLCK);
- if (ret >= 0)
- ret = mq_initiate_quota_blocking_txn (this, &parent_loc, NULL);
+ if (ret >= 0)
+ ret = mq_initiate_quota_blocking_txn(this, &parent_loc, NULL);
- loc_wipe (&parent_loc);
+ loc_wipe(&parent_loc);
- if (contribution)
- GF_REF_PUT (contribution);
+ if (contribution)
+ GF_REF_PUT(contribution);
- return ret;
+ return ret;
}
int32_t
-mq_reduce_parent_size_txn (xlator_t *this, loc_t *origin_loc,
- quota_meta_t *contri, uint32_t nlink,
- call_stub_t *stub)
+mq_reduce_parent_size_txn(xlator_t *this, loc_t *origin_loc,
+ quota_meta_t *contri, uint32_t nlink,
+ call_stub_t *stub)
{
- int32_t ret = -1;
- loc_t loc = {0, };
- gf_boolean_t resume_stub = _gf_true;
+ int32_t ret = -1;
+ loc_t loc = {
+ 0,
+ };
+ gf_boolean_t resume_stub = _gf_true;
- GF_VALIDATE_OR_GOTO ("marker", this, out);
- GF_VALIDATE_OR_GOTO ("marker", origin_loc, out);
+ GF_VALIDATE_OR_GOTO("marker", this, out);
+ GF_VALIDATE_OR_GOTO("marker", origin_loc, out);
- ret = mq_prevalidate_txn (this, origin_loc, &loc, NULL, NULL);
- if (ret < 0)
- goto out;
+ ret = mq_prevalidate_txn(this, origin_loc, &loc, NULL, NULL);
+ if (ret < 0)
+ goto out;
- if (loc_is_root(&loc)) {
- ret = 0;
- goto out;
- }
+ if (loc_is_root(&loc)) {
+ ret = 0;
+ goto out;
+ }
- resume_stub = _gf_false;
- ret = mq_synctask1 (this, mq_reduce_parent_size_task, _gf_true, &loc,
- contri, nlink, stub);
+ resume_stub = _gf_false;
+ ret = mq_synctask1(this, mq_reduce_parent_size_task, _gf_true, &loc, contri,
+ nlink, stub);
out:
- loc_wipe (&loc);
+ loc_wipe(&loc);
- if (resume_stub && stub)
- call_resume (stub);
+ if (resume_stub && stub)
+ call_resume(stub);
- if (ret)
- gf_log_callingfn (this->name, GF_LOG_ERROR,
- "mq_reduce_parent_size_txn failed");
+ if (ret)
+ gf_log_callingfn(this ? this->name : "Marker", GF_LOG_ERROR,
+ "mq_reduce_parent_size_txn failed");
- return ret;
+ return ret;
}
int
-mq_initiate_quota_task (void *opaque)
+mq_initiate_quota_task(void *opaque)
{
- int32_t ret = -1;
- int32_t prev_dirty = 0;
- loc_t child_loc = {0,};
- loc_t parent_loc = {0,};
- gf_boolean_t locked = _gf_false;
- gf_boolean_t dirty = _gf_false;
- gf_boolean_t status = _gf_false;
- quota_meta_t delta = {0, };
- quota_synctask_t *args = NULL;
- xlator_t *this = NULL;
- loc_t *loc = NULL;
- inode_contribution_t *contri = NULL;
- quota_inode_ctx_t *ctx = NULL;
- quota_inode_ctx_t *parent_ctx = NULL;
- inode_t *tmp_parent = NULL;
-
- GF_VALIDATE_OR_GOTO ("marker", opaque, out);
-
- args = (quota_synctask_t *) opaque;
- loc = &args->loc;
- this = args->this;
-
- GF_VALIDATE_OR_GOTO ("marker", this, out);
- THIS = this;
-
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
-
- ret = mq_loc_copy (&child_loc, loc);
+ int32_t ret = -1;
+ int32_t prev_dirty = 0;
+ loc_t child_loc = {
+ 0,
+ };
+ loc_t parent_loc = {
+ 0,
+ };
+ gf_boolean_t locked = _gf_false;
+ gf_boolean_t dirty = _gf_false;
+ gf_boolean_t status = _gf_false;
+ quota_meta_t delta = {
+ 0,
+ };
+ quota_synctask_t *args = NULL;
+ xlator_t *this = NULL;
+ loc_t *loc = NULL;
+ inode_contribution_t *contri = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ quota_inode_ctx_t *parent_ctx = NULL;
+ inode_t *tmp_parent = NULL;
+
+ GF_VALIDATE_OR_GOTO("marker", opaque, out);
+
+ args = (quota_synctask_t *)opaque;
+ loc = &args->loc;
+ this = args->this;
+
+ GF_VALIDATE_OR_GOTO("marker", this, out);
+ THIS = this;
+
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
+
+ ret = mq_loc_copy(&child_loc, loc);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_ERROR, "loc copy failed");
+ goto out;
+ }
+
+ while (!__is_root_gfid(child_loc.gfid)) {
+ ret = mq_inode_ctx_get(child_loc.inode, this, &ctx);
if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "loc copy failed");
+ gf_log(this->name, GF_LOG_WARNING,
+ "inode ctx get failed for %s, "
+ "aborting update txn",
+ child_loc.path);
+ goto out;
+ }
+
+ /* To improve performance, abort current transaction
+ * if one is already in progress for same inode
+ */
+ if (status == _gf_true) {
+ /* status will already set before txn start,
+ * so it should not be set in first
+ * loop iteration
+ */
+ ret = mq_test_and_set_ctx_updation_status(ctx, &status);
+ if (ret < 0 || status == _gf_true)
+ goto out;
+ }
+
+ if (child_loc.parent == NULL) {
+ ret = mq_build_ancestry(this, &child_loc);
+ if (ret < 0 || child_loc.parent == NULL) {
+ /* If application performs parallel remove
+ * operations on same set of files/directories
+ * then we may get ENOENT/ESTALE
+ */
+ gf_log(this->name,
+ (-ret == ENOENT || -ret == ESTALE) ? GF_LOG_DEBUG
+ : GF_LOG_ERROR,
+ "build ancestry failed for inode %s",
+ uuid_utoa(child_loc.inode->gfid));
+ ret = -1;
goto out;
+ }
+ }
+
+ ret = mq_inode_loc_fill(NULL, child_loc.parent, &parent_loc);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "parent_loc fill "
+ "failed for child inode %s: ",
+ uuid_utoa(child_loc.inode->gfid));
+ goto out;
}
- while (!__is_root_gfid (child_loc.gfid)) {
+ ret = mq_lock(this, &parent_loc, F_WRLCK);
+ if (ret < 0)
+ goto out;
+ locked = _gf_true;
- ret = mq_inode_ctx_get (child_loc.inode, this, &ctx);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "inode ctx get failed for %s, "
- "aborting update txn", child_loc.path);
- goto out;
- }
+ mq_set_ctx_updation_status(ctx, _gf_false);
+ status = _gf_true;
- /* To improve performance, abort current transaction
- * if one is already in progress for same inode
+ /* Contribution node can be NULL in below scenarios and
+ create if needed:
+
+ Scenario 1)
+ In this case create a new contribution node
+ Suppose hard link for a file f1 present in a directory d1 is
+ created in the directory d2 (as f2). Now, since d2's
+ contribution is not there in f1's inode ctx, d2's
+ contribution xattr won't be created and will create problems
+ for quota operations.
+
+ Don't create contribution if parent has been changed after
+ taking a lock, this can happen when rename is performed
+ and writes is still in-progress for the same file
+
+ Scenario 2)
+ When a rename operation is performed, contribution node
+ for olp path will be removed.
+
+ Create contribution node only if oldparent is same as
+ newparent.
+ Consider below example
+ 1) rename FOP invoked on file 'x'
+ 2) write is still in progress for file 'x'
+ 3) rename takes a lock on old-parent
+ 4) write-update txn blocked on old-parent to acquire lock
+ 5) in rename_cbk, contri xattrs are removed and contribution
+ is deleted and lock is released
+ 6) now write-update txn gets the lock and updates the
+ wrong parent as it was holding lock on old parent
+ so validate parent once the lock is acquired
+
+ For more information on this problem, please see
+ doc for marker_rename in file marker.c
+ */
+ contri = mq_get_contribution_node(child_loc.parent, ctx);
+ if (contri == NULL) {
+ tmp_parent = inode_parent(child_loc.inode, 0, NULL);
+ if (tmp_parent == NULL) {
+ /* This can happen if application performs
+ * parallel remove operations on same set
+ * of files/directories
*/
- if (status == _gf_true) {
- /* status will already set before txn start,
- * so it should not be set in first
- * loop iteration
- */
- ret = mq_test_and_set_ctx_updation_status (ctx,
- &status);
- if (ret < 0 || status == _gf_true)
- goto out;
- }
-
- if (child_loc.parent == NULL) {
- ret = mq_build_ancestry (this, &child_loc);
- if (ret < 0 || child_loc.parent == NULL) {
- /* If application performs parallel remove
- * operations on same set of files/directories
- * then we may get ENOENT/ESTALE
- */
- gf_log (this->name,
- (-ret == ENOENT || -ret == ESTALE)
- ? GF_LOG_DEBUG:GF_LOG_ERROR,
- "build ancestry failed for inode %s",
- uuid_utoa (child_loc.inode->gfid));
- ret = -1;
- goto out;
- }
- }
-
- ret = mq_inode_loc_fill (NULL, child_loc.parent, &parent_loc);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "parent_loc fill "
- "failed for child inode %s: ",
- uuid_utoa (child_loc.inode->gfid));
- goto out;
- }
-
- ret = mq_lock (this, &parent_loc, F_WRLCK);
- if (ret < 0)
- goto out;
- locked = _gf_true;
-
- mq_set_ctx_updation_status (ctx, _gf_false);
- status = _gf_true;
-
- /* Contribution node can be NULL in below scenarios and
- create if needed:
-
- Scenario 1)
- In this case create a new contribution node
- Suppose hard link for a file f1 present in a directory d1 is
- created in the directory d2 (as f2). Now, since d2's
- contribution is not there in f1's inode ctx, d2's
- contribution xattr wont be created and will create problems
- for quota operations.
-
- Don't create contribution if parent has been changed after
- taking a lock, this can happen when rename is performed
- and writes is still in-progress for the same file
-
- Scenario 2)
- When a rename operation is performed, contribution node
- for olp path will be removed.
-
- Create contribution node only if oldparent is same as
- newparent.
- Consider below example
- 1) rename FOP invoked on file 'x'
- 2) write is still in progress for file 'x'
- 3) rename takes a lock on old-parent
- 4) write-update txn blocked on old-parent to acquire lock
- 5) in rename_cbk, contri xattrs are removed and contribution
- is deleted and lock is released
- 6) now write-update txn gets the lock and updates the
- wrong parent as it was holding lock on old parent
- so validate parent once the lock is acquired
-
- For more information on this problem, please see
- doc for marker_rename in file marker.c
- */
- contri = mq_get_contribution_node (child_loc.parent, ctx);
- if (contri == NULL) {
- tmp_parent = inode_parent (child_loc.inode, 0, NULL);
- if (tmp_parent == NULL) {
- /* This can happen if application performs
- * parallel remove operations on same set
- * of files/directories
- */
- gf_log (this->name, GF_LOG_WARNING, "parent is "
- "NULL for inode %s",
- uuid_utoa (child_loc.inode->gfid));
- ret = -1;
- goto out;
- }
- if (gf_uuid_compare(tmp_parent->gfid,
- parent_loc.gfid)) {
- /* abort txn if parent has changed */
- ret = 0;
- goto out;
- }
-
- inode_unref (tmp_parent);
- tmp_parent = NULL;
-
- contri = mq_add_new_contribution_node (this, ctx,
- &child_loc);
- if (contri == NULL) {
- gf_log (this->name, GF_LOG_ERROR, "Failed to "
- "create contribution node for %s, "
- "abort update txn", child_loc.path);
- ret = -1;
- goto out;
- }
- }
-
- ret = mq_get_delta (this, &child_loc, &delta, ctx, contri);
- if (ret < 0)
- goto out;
-
- if (quota_meta_is_null (&delta))
- goto out;
-
- ret = mq_get_set_dirty (this, &parent_loc, 1, &prev_dirty);
- if (ret < 0)
- goto out;
- dirty = _gf_true;
-
- ret = mq_update_contri (this, &child_loc, contri, &delta);
- if (ret < 0)
- goto out;
-
- ret = mq_update_size (this, &parent_loc, &delta);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_DEBUG, "rollback "
- "contri updation");
- mq_sub_meta (&delta, NULL);
- mq_update_contri (this, &child_loc, contri, &delta);
- goto out;
- }
-
- if (prev_dirty == 0) {
- ret = mq_mark_dirty (this, &parent_loc, 0);
- } else {
- ret = mq_inode_ctx_get (parent_loc.inode, this,
- &parent_ctx);
- if (ret == 0)
- mq_set_ctx_dirty_status (parent_ctx, _gf_false);
- }
- dirty = _gf_false;
- prev_dirty = 0;
-
- ret = mq_lock (this, &parent_loc, F_UNLCK);
- locked = _gf_false;
-
- if (__is_root_gfid (parent_loc.gfid))
- break;
-
- /* Repeate above steps upwards till the root */
- loc_wipe (&child_loc);
- ret = mq_loc_copy (&child_loc, &parent_loc);
- if (ret < 0)
- goto out;
-
- loc_wipe (&parent_loc);
- GF_REF_PUT (contri);
- contri = NULL;
+ gf_log(this->name, GF_LOG_WARNING,
+ "parent is "
+ "NULL for inode %s",
+ uuid_utoa(child_loc.inode->gfid));
+ ret = -1;
+ goto out;
+ }
+ if (gf_uuid_compare(tmp_parent->gfid, parent_loc.gfid)) {
+ /* abort txn if parent has changed */
+ ret = 0;
+ goto out;
+ }
+
+ inode_unref(tmp_parent);
+ tmp_parent = NULL;
+
+ contri = mq_add_new_contribution_node(this, ctx, &child_loc);
+ if (contri == NULL) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Failed to "
+ "create contribution node for %s, "
+ "abort update txn",
+ child_loc.path);
+ ret = -1;
+ goto out;
+ }
}
-out:
- if (dirty) {
- if (ret < 0 || prev_dirty) {
- /* On failure clear dirty status flag.
- * In the next lookup inspect_directory_xattr
- * can set the status flag and fix the
- * dirty directory.
- * Do the same if the dir was dirty before
- * txn
- */
- ret = mq_inode_ctx_get (parent_loc.inode, this,
- &parent_ctx);
- if (ret == 0)
- mq_set_ctx_dirty_status (parent_ctx, _gf_false);
- } else {
- ret = mq_mark_dirty (this, &parent_loc, 0);
- }
+ ret = mq_get_delta(this, &child_loc, &delta, ctx, contri);
+ if (ret < 0)
+ goto out;
+
+ if (quota_meta_is_null(&delta))
+ goto out;
+
+ ret = mq_get_set_dirty(this, &parent_loc, 1, &prev_dirty);
+ if (ret < 0)
+ goto out;
+ dirty = _gf_true;
+
+ ret = mq_update_contri(this, &child_loc, contri, &delta);
+ if (ret < 0)
+ goto out;
+
+ ret = mq_update_size(this, &parent_loc, &delta);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_DEBUG,
+ "rollback "
+ "contri updation");
+ mq_sub_meta(&delta, NULL);
+ mq_update_contri(this, &child_loc, contri, &delta);
+ goto out;
}
- if (locked)
- ret = mq_lock (this, &parent_loc, F_UNLCK);
+ if (prev_dirty == 0) {
+ ret = mq_mark_dirty(this, &parent_loc, 0);
+ } else {
+ ret = mq_inode_ctx_get(parent_loc.inode, this, &parent_ctx);
+ if (ret == 0)
+ mq_set_ctx_dirty_status(parent_ctx, _gf_false);
+ }
+ dirty = _gf_false;
+ prev_dirty = 0;
- if (ctx && status == _gf_false)
- mq_set_ctx_updation_status (ctx, _gf_false);
+ ret = mq_lock(this, &parent_loc, F_UNLCK);
+ locked = _gf_false;
- loc_wipe (&child_loc);
- loc_wipe (&parent_loc);
+ if (__is_root_gfid(parent_loc.gfid))
+ break;
- if (tmp_parent)
- inode_unref (tmp_parent);
+ /* Repeate above steps upwards till the root */
+ loc_wipe(&child_loc);
+ ret = mq_loc_copy(&child_loc, &parent_loc);
+ if (ret < 0)
+ goto out;
- if (contri)
- GF_REF_PUT (contri);
+ loc_wipe(&parent_loc);
+ GF_REF_PUT(contri);
+ contri = NULL;
+ }
- return 0;
+out:
+ if ((dirty) && (ret < 0)) {
+ /* On failure clear dirty status flag.
+ * In the next lookup inspect_directory_xattr
+ * can set the status flag and fix the
+ * dirty directory.
+ * Do the same if the dir was dirty before
+ * txn
+ */
+ ret = mq_inode_ctx_get(parent_loc.inode, this, &parent_ctx);
+ if (ret == 0)
+ mq_set_ctx_dirty_status(parent_ctx, _gf_false);
+ }
+
+ if (locked)
+ ret = mq_lock(this, &parent_loc, F_UNLCK);
+
+ if (ctx && status == _gf_false)
+ mq_set_ctx_updation_status(ctx, _gf_false);
+
+ loc_wipe(&child_loc);
+ loc_wipe(&parent_loc);
+
+ if (tmp_parent)
+ inode_unref(tmp_parent);
+
+ if (contri)
+ GF_REF_PUT(contri);
+
+ return 0;
}
int
-_mq_initiate_quota_txn (xlator_t *this, loc_t *origin_loc, struct iatt *buf,
- gf_boolean_t spawn)
+_mq_initiate_quota_txn(xlator_t *this, loc_t *origin_loc, struct iatt *buf,
+ gf_boolean_t spawn)
{
- int32_t ret = -1;
- quota_inode_ctx_t *ctx = NULL;
- gf_boolean_t status = _gf_true;
- loc_t loc = {0,};
-
- ret = mq_prevalidate_txn (this, origin_loc, &loc, &ctx, buf);
- if (ret < 0)
- goto out;
+ int32_t ret = -1;
+ quota_inode_ctx_t *ctx = NULL;
+ gf_boolean_t status = _gf_true;
+ loc_t loc = {
+ 0,
+ };
+
+ ret = mq_prevalidate_txn(this, origin_loc, &loc, &ctx, buf);
+ if (ret < 0)
+ goto out;
- if (loc_is_root(&loc)) {
- ret = 0;
- goto out;
- }
+ if (loc_is_root(&loc)) {
+ ret = 0;
+ goto out;
+ }
- ret = mq_test_and_set_ctx_updation_status (ctx, &status);
- if (ret < 0 || status == _gf_true)
- goto out;
+ ret = mq_test_and_set_ctx_updation_status(ctx, &status);
+ if (ret < 0 || status == _gf_true)
+ goto out;
- ret = mq_synctask (this, mq_initiate_quota_task, spawn, &loc);
+ ret = mq_synctask(this, mq_initiate_quota_task, spawn, &loc);
out:
- if (ret < 0 && status == _gf_false)
- mq_set_ctx_updation_status (ctx, _gf_false);
+ if (ret < 0 && status == _gf_false)
+ mq_set_ctx_updation_status(ctx, _gf_false);
- loc_wipe (&loc);
- return ret;
+ loc_wipe(&loc);
+ return ret;
}
int
-mq_initiate_quota_txn (xlator_t *this, loc_t *loc, struct iatt *buf)
+mq_initiate_quota_txn(xlator_t *this, loc_t *loc, struct iatt *buf)
{
- int32_t ret = -1;
+ int32_t ret = -1;
- GF_VALIDATE_OR_GOTO ("marker", this, out);
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
- GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
+ GF_VALIDATE_OR_GOTO("marker", this, out);
+ GF_VALIDATE_OR_GOTO("marker", loc, out);
+ GF_VALIDATE_OR_GOTO("marker", loc->inode, out);
- ret = _mq_initiate_quota_txn (this, loc, buf, _gf_true);
+ ret = _mq_initiate_quota_txn(this, loc, buf, _gf_true);
out:
- return ret;
+ return ret;
}
int
-mq_initiate_quota_blocking_txn (xlator_t *this, loc_t *loc, struct iatt *buf)
+mq_initiate_quota_blocking_txn(xlator_t *this, loc_t *loc, struct iatt *buf)
{
- int32_t ret = -1;
+ int32_t ret = -1;
- GF_VALIDATE_OR_GOTO ("marker", this, out);
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
- GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
+ GF_VALIDATE_OR_GOTO("marker", this, out);
+ GF_VALIDATE_OR_GOTO("marker", loc, out);
+ GF_VALIDATE_OR_GOTO("marker", loc->inode, out);
- ret = _mq_initiate_quota_txn (this, loc, buf, _gf_false);
+ ret = _mq_initiate_quota_txn(this, loc, buf, _gf_false);
out:
- return ret;
+ return ret;
}
int
-mq_update_dirty_inode_task (void *opaque)
+mq_update_dirty_inode_task(void *opaque)
{
- int32_t ret = -1;
- fd_t *fd = NULL;
- off_t offset = 0;
- gf_dirent_t entries;
- gf_dirent_t *entry = NULL;
- gf_boolean_t locked = _gf_false;
- gf_boolean_t updated = _gf_false;
- int32_t dirty = 0;
- quota_meta_t contri = {0, };
- quota_meta_t size = {0, };
- quota_meta_t contri_sum = {0, };
- quota_meta_t delta = {0, };
- quota_synctask_t *args = NULL;
- xlator_t *this = NULL;
- loc_t *loc = NULL;
- quota_inode_ctx_t *ctx = NULL;
- dict_t *xdata = NULL;
- char contri_key[QUOTA_KEY_MAX] = {0, };
-
- GF_ASSERT (opaque);
-
- args = (quota_synctask_t *) opaque;
- loc = &args->loc;
- this = args->this;
- THIS = this;
- INIT_LIST_HEAD (&entries.list);
-
- ret = mq_inode_ctx_get (loc->inode, this, &ctx);
- if (ret < 0)
- goto out;
+ int32_t ret = -1;
+ fd_t *fd = NULL;
+ off_t offset = 0;
+ gf_dirent_t entries;
+ gf_dirent_t *entry = NULL;
+ gf_boolean_t locked = _gf_false;
+ gf_boolean_t updated = _gf_false;
+ int32_t dirty = 0;
+ quota_meta_t contri = {
+ 0,
+ };
+ quota_meta_t size = {
+ 0,
+ };
+ quota_meta_t contri_sum = {
+ 0,
+ };
+ quota_meta_t delta = {
+ 0,
+ };
+ quota_synctask_t *args = NULL;
+ xlator_t *this = NULL;
+ loc_t *loc = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ dict_t *xdata = NULL;
+ char contri_key[QUOTA_KEY_MAX] = {
+ 0,
+ };
+ int keylen = 0;
+
+ GF_ASSERT(opaque);
+
+ args = (quota_synctask_t *)opaque;
+ loc = &args->loc;
+ this = args->this;
+ THIS = this;
+ INIT_LIST_HEAD(&entries.list);
+
+ ret = mq_inode_ctx_get(loc->inode, this, &ctx);
+ if (ret < 0)
+ goto out;
- GET_CONTRI_KEY (this, contri_key, loc->gfid, ret);
- if (ret < 0)
- goto out;
+ GET_CONTRI_KEY(this, contri_key, loc->gfid, keylen);
+ if (keylen < 0) {
+ ret = keylen;
+ goto out;
+ }
- xdata = dict_new ();
- if (xdata == NULL) {
- gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
- ret = -1;
- goto out;
- }
+ xdata = dict_new();
+ if (xdata == NULL) {
+ gf_log(this->name, GF_LOG_ERROR, "dict_new failed");
+ ret = -1;
+ goto out;
+ }
- ret = dict_set_int64 (xdata, contri_key, 0);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "dict_set failed");
- goto out;
- }
+ ret = dict_set_int64(xdata, contri_key, 0);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_ERROR, "dict_set failed");
+ goto out;
+ }
- ret = mq_lock (this, loc, F_WRLCK);
- if (ret < 0)
- goto out;
- locked = _gf_true;
+ ret = mq_lock(this, loc, F_WRLCK);
+ if (ret < 0)
+ goto out;
+ locked = _gf_true;
- ret = mq_get_dirty (this, loc, &dirty);
- if (ret < 0 || dirty == 0) {
- ret = 0;
- goto out;
- }
+ ret = mq_get_dirty(this, loc, &dirty);
+ if (ret < 0 || dirty == 0) {
+ ret = 0;
+ goto out;
+ }
- fd = fd_create (loc->inode, 0);
- if (!fd) {
- gf_log (this->name, GF_LOG_ERROR, "Failed to create fd");
- ret = -1;
- goto out;
- }
+ fd = fd_create(loc->inode, 0);
+ if (!fd) {
+ gf_log(this->name, GF_LOG_ERROR, "Failed to create fd");
+ ret = -1;
+ goto out;
+ }
+
+ ret = syncop_opendir(this, loc, fd, NULL, NULL);
+ if (ret < 0) {
+ gf_log(this->name,
+ (-ret == ENOENT || -ret == ESTALE) ? GF_LOG_DEBUG : GF_LOG_ERROR,
+ "opendir failed "
+ "for %s: %s",
+ loc->path, strerror(-ret));
+ goto out;
+ }
- ret = syncop_opendir (this, loc, fd, NULL, NULL);
+ fd_bind(fd);
+ while ((ret = syncop_readdirp(this, fd, 131072, offset, &entries, xdata,
+ NULL)) != 0) {
if (ret < 0) {
- gf_log (this->name, (-ret == ENOENT || -ret == ESTALE)
- ? GF_LOG_DEBUG:GF_LOG_ERROR, "opendir failed "
- "for %s: %s", loc->path, strerror (-ret));
- goto out;
+ gf_log(this->name,
+ (-ret == ENOENT || -ret == ESTALE) ? GF_LOG_DEBUG
+ : GF_LOG_ERROR,
+ "readdirp failed "
+ "for %s: %s",
+ loc->path, strerror(-ret));
+ goto out;
}
- fd_bind (fd);
- while ((ret = syncop_readdirp (this, fd, 131072, offset, &entries,
- xdata, NULL)) != 0) {
- if (ret < 0) {
- gf_log (this->name, (-ret == ENOENT || -ret == ESTALE)
- ? GF_LOG_DEBUG:GF_LOG_ERROR, "readdirp failed "
- "for %s: %s", loc->path, strerror (-ret));
- goto out;
- }
-
- if (list_empty (&entries.list))
- break;
+ if (list_empty(&entries.list))
+ break;
- list_for_each_entry (entry, &entries.list, list) {
- offset = entry->d_off;
-
- if (!strcmp (entry->d_name, ".") ||
- !strcmp (entry->d_name, ".."))
- continue;
+ list_for_each_entry(entry, &entries.list, list)
+ {
+ offset = entry->d_off;
- memset (&contri, 0, sizeof (contri));
- quota_dict_get_meta (entry->dict, contri_key, &contri);
- if (quota_meta_is_null (&contri))
- continue;
+ if (!strcmp(entry->d_name, ".") || !strcmp(entry->d_name, ".."))
+ continue;
- mq_add_meta (&contri_sum, &contri);
- }
+ memset(&contri, 0, sizeof(contri));
+ quota_dict_get_meta(entry->dict, contri_key, keylen, &contri);
+ if (quota_meta_is_null(&contri))
+ continue;
- gf_dirent_free (&entries);
+ mq_add_meta(&contri_sum, &contri);
}
- /* Inculde for self */
- contri_sum.dir_count++;
- ret = mq_get_size (this, loc, &size);
- if (ret < 0)
- goto out;
+ gf_dirent_free(&entries);
+ }
+ /* Inculde for self */
+ contri_sum.dir_count++;
- mq_compute_delta (&delta, &contri_sum, &size);
+ ret = _mq_get_metadata(this, loc, NULL, &size, 0);
+ if (ret < 0)
+ goto out;
- if (quota_meta_is_null (&delta))
- goto out;
+ mq_compute_delta(&delta, &contri_sum, &size);
- gf_log (this->name, GF_LOG_INFO, "calculated size = %"PRId64
- ", original size = %"PRIu64 ", diff = %"PRIu64
- ", path = %s ", contri_sum.size, size.size, delta.size,
- loc->path);
+ if (quota_meta_is_null(&delta))
+ goto out;
- gf_log (this->name, GF_LOG_INFO, "calculated f_count = %"PRId64
- ", original f_count = %"PRIu64 ", diff = %"PRIu64
- ", path = %s ", contri_sum.file_count, size.file_count,
- delta.file_count, loc->path);
+ gf_log(this->name, GF_LOG_INFO,
+ "calculated size = %" PRId64 ", original size = %" PRIu64
+ ", diff = %" PRIu64 ", path = %s ",
+ contri_sum.size, size.size, delta.size, loc->path);
- gf_log (this->name, GF_LOG_INFO, "calculated d_count = %"PRId64
- ", original d_count = %"PRIu64 ", diff = %"PRIu64
- ", path = %s ", contri_sum.dir_count, size.dir_count,
- delta.dir_count, loc->path);
+ gf_log(this->name, GF_LOG_INFO,
+ "calculated f_count = %" PRId64 ", original f_count = %" PRIu64
+ ", diff = %" PRIu64 ", path = %s ",
+ contri_sum.file_count, size.file_count, delta.file_count, loc->path);
+ gf_log(this->name, GF_LOG_INFO,
+ "calculated d_count = %" PRId64 ", original d_count = %" PRIu64
+ ", diff = %" PRIu64 ", path = %s ",
+ contri_sum.dir_count, size.dir_count, delta.dir_count, loc->path);
- ret = mq_update_size (this, loc, &delta);
- if (ret < 0)
- goto out;
+ ret = mq_update_size(this, loc, &delta);
+ if (ret < 0)
+ goto out;
- updated = _gf_true;
+ updated = _gf_true;
out:
- gf_dirent_free (&entries);
+ gf_dirent_free(&entries);
- if (fd)
- fd_unref (fd);
+ if (fd)
+ fd_unref(fd);
- if (xdata)
- dict_unref (xdata);
+ if (xdata)
+ dict_unref(xdata);
- if (ret < 0) {
- /* On failure clear dirty status flag.
- * In the next lookup inspect_directory_xattr
- * can set the status flag and fix the
- * dirty directory
- */
- if (ctx)
- mq_set_ctx_dirty_status (ctx, _gf_false);
- } else if (dirty) {
- mq_mark_dirty (this, loc, 0);
- }
+ if (ret < 0) {
+ /* On failure clear dirty status flag.
+ * In the next lookup inspect_directory_xattr
+ * can set the status flag and fix the
+ * dirty directory
+ */
+ if (ctx)
+ mq_set_ctx_dirty_status(ctx, _gf_false);
+ } else if (dirty) {
+ mq_mark_dirty(this, loc, 0);
+ }
- if (locked)
- mq_lock (this, loc, F_UNLCK);
+ if (locked)
+ mq_lock(this, loc, F_UNLCK);
- if (updated)
- mq_initiate_quota_blocking_txn (this, loc, NULL);
+ if (updated)
+ mq_initiate_quota_blocking_txn(this, loc, NULL);
- return ret;
+ return ret;
}
int32_t
-mq_update_dirty_inode_txn (xlator_t *this, loc_t *loc, quota_inode_ctx_t *ctx)
+mq_update_dirty_inode_txn(xlator_t *this, loc_t *loc, quota_inode_ctx_t *ctx)
{
- int32_t ret = -1;
- gf_boolean_t status = _gf_true;
+ int32_t ret = -1;
+ gf_boolean_t status = _gf_true;
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
- GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
+ GF_VALIDATE_OR_GOTO("marker", loc, out);
+ GF_VALIDATE_OR_GOTO("marker", loc->inode, out);
- ret = mq_test_and_set_ctx_dirty_status (ctx, &status);
- if (ret < 0 || status == _gf_true)
- goto out;
+ mq_test_and_set_ctx_status(ctx, &ctx->dirty_status, &status);
+ if (status == _gf_true)
+ goto out;
- ret = mq_synctask (this, mq_update_dirty_inode_task, _gf_true, loc);
+ ret = mq_synctask(this, mq_update_dirty_inode_task, _gf_true, loc);
out:
- if (ret < 0 && status == _gf_false)
- mq_set_ctx_dirty_status (ctx, _gf_false);
+ if (ret < 0 && status == _gf_false)
+ mq_set_ctx_dirty_status(ctx, _gf_false);
- return ret;
+ return ret;
}
int32_t
-mq_inspect_directory_xattr (xlator_t *this, quota_inode_ctx_t *ctx,
- inode_contribution_t *contribution, loc_t *loc,
- dict_t *dict, struct iatt buf)
+mq_inspect_directory_xattr(xlator_t *this, quota_inode_ctx_t *ctx,
+ inode_contribution_t *contribution, loc_t *loc,
+ dict_t *dict)
{
- int32_t ret = -1;
- int8_t dirty = -1;
- quota_meta_t size = {0, };
- quota_meta_t contri = {0, };
- quota_meta_t delta = {0, };
- char contri_key[QUOTA_KEY_MAX] = {0, };
- char size_key[QUOTA_KEY_MAX] = {0, };
- gf_boolean_t status = _gf_false;
-
- ret = dict_get_int8 (dict, QUOTA_DIRTY_KEY, &dirty);
- if (ret < 0) {
- /* dirty is set only on the first file write operation
- * so ignore this error
- */
- ret = 0;
- dirty = 0;
- }
+ int32_t ret = -1;
+ int8_t dirty = -1;
+ quota_meta_t size = {
+ 0,
+ };
+ quota_meta_t contri = {
+ 0,
+ };
+ quota_meta_t delta = {
+ 0,
+ };
+ char contri_key[QUOTA_KEY_MAX] = {
+ 0,
+ };
+ char size_key[QUOTA_KEY_MAX] = {
+ 0,
+ };
+ int keylen = 0;
+ gf_boolean_t status = _gf_false;
+
+ ret = dict_get_int8(dict, QUOTA_DIRTY_KEY, &dirty);
+ if (ret < 0) {
+ /* dirty is set only on the first file write operation
+ * so ignore this error
+ */
+ ret = 0;
+ dirty = 0;
+ }
- GET_SIZE_KEY (this, size_key, ret);
- if (ret < 0)
- goto out;
- ret = _quota_dict_get_meta (this, dict, size_key, &size,
- IA_IFDIR, _gf_false);
+ GET_SIZE_KEY(this, size_key, keylen);
+ if (keylen < 0) {
+ ret = -1;
+ goto out;
+ }
+ ret = _quota_dict_get_meta(this, dict, size_key, keylen, &size, IA_IFDIR,
+ _gf_false);
+ if (ret < 0)
+ goto create_xattr;
+
+ if (!contribution)
+ goto create_xattr;
+
+ if (!loc_is_root(loc)) {
+ GET_CONTRI_KEY(this, contri_key, contribution->gfid, keylen);
+ if (keylen < 0) {
+ ret = -1;
+ goto out;
+ }
+ ret = _quota_dict_get_meta(this, dict, contri_key, keylen, &contri,
+ IA_IFDIR, _gf_false);
if (ret < 0)
- goto create_xattr;
-
- if (!loc_is_root(loc)) {
- GET_CONTRI_KEY (this, contri_key, contribution->gfid, ret);
- if (ret < 0)
- goto out;
-
- ret = _quota_dict_get_meta (this, dict, contri_key, &contri,
- IA_IFDIR, _gf_false);
- if (ret < 0)
- goto create_xattr;
-
- LOCK (&contribution->lock);
- {
- contribution->contribution = contri.size;
- contribution->file_count = contri.file_count;
- contribution->dir_count = contri.dir_count;
- }
- UNLOCK (&contribution->lock);
- }
+ goto create_xattr;
- LOCK (&ctx->lock);
+ LOCK(&contribution->lock);
{
- ctx->size = size.size;
- ctx->file_count = size.file_count;
- ctx->dir_count = size.dir_count;
- ctx->dirty = dirty;
- }
- UNLOCK (&ctx->lock);
-
- ret = mq_get_ctx_updation_status (ctx, &status);
- if (ret < 0 || status == _gf_true) {
- /* If the update txn is in progress abort inspection */
- ret = 0;
- goto out;
- }
+ contribution->contribution = contri.size;
+ contribution->file_count = contri.file_count;
+ contribution->dir_count = contri.dir_count;
+ }
+ UNLOCK(&contribution->lock);
+ }
+
+ LOCK(&ctx->lock);
+ {
+ ctx->size = size.size;
+ ctx->file_count = size.file_count;
+ ctx->dir_count = size.dir_count;
+ ctx->dirty = dirty;
+ }
+ UNLOCK(&ctx->lock);
+
+ ret = mq_get_ctx_updation_status(ctx, &status);
+ if (ret < 0 || status == _gf_true) {
+ /* If the update txn is in progress abort inspection */
+ ret = 0;
+ goto out;
+ }
- mq_compute_delta (&delta, &size, &contri);
+ mq_compute_delta(&delta, &size, &contri);
- if (dirty) {
- ret = mq_update_dirty_inode_txn (this, loc, ctx);
- goto out;
- }
+ if (dirty) {
+ ret = mq_update_dirty_inode_txn(this, loc, ctx);
+ goto out;
+ }
- if (!loc_is_root(loc) &&
- !quota_meta_is_null (&delta))
- mq_initiate_quota_txn (this, loc, NULL);
+ if (!loc_is_root(loc) && !quota_meta_is_null(&delta))
+ mq_initiate_quota_txn(this, loc, NULL);
- ret = 0;
- goto out;
+ ret = 0;
+ goto out;
create_xattr:
- if (ret < 0)
- ret = mq_create_xattrs_txn (this, loc, NULL);
+ if (ret < 0)
+ ret = mq_create_xattrs_txn(this, loc, NULL);
out:
- return ret;
+ return ret;
}
int32_t
-mq_inspect_file_xattr (xlator_t *this, quota_inode_ctx_t *ctx,
- inode_contribution_t *contribution, loc_t *loc,
- dict_t *dict, struct iatt buf)
+mq_inspect_file_xattr(xlator_t *this, quota_inode_ctx_t *ctx,
+ inode_contribution_t *contribution, loc_t *loc,
+ dict_t *dict, struct iatt *buf)
{
- int32_t ret = -1;
- quota_meta_t size = {0, };
- quota_meta_t contri = {0, };
- quota_meta_t delta = {0, };
- char contri_key[QUOTA_KEY_MAX] = {0, };
- gf_boolean_t status = _gf_false;
-
- LOCK (&ctx->lock);
- {
- ctx->size = 512 * buf.ia_blocks;
- ctx->file_count = 1;
- ctx->dir_count = 0;
+ int32_t ret = -1;
+ quota_meta_t size = {
+ 0,
+ };
+ quota_meta_t contri = {
+ 0,
+ };
+ quota_meta_t delta = {
+ 0,
+ };
+ char contri_key[QUOTA_KEY_MAX] = {
+ 0,
+ };
+ int keylen = 0;
+ gf_boolean_t status = _gf_false;
+
+ if (!buf || !contribution || !ctx)
+ goto out;
- size.size = ctx->size;
- size.file_count = ctx->file_count;
- size.dir_count = ctx->dir_count;
+ LOCK(&ctx->lock);
+ {
+ ctx->size = 512 * buf->ia_blocks;
+ ctx->file_count = 1;
+ ctx->dir_count = 0;
+
+ size.size = ctx->size;
+ size.file_count = ctx->file_count;
+ size.dir_count = ctx->dir_count;
+ }
+ UNLOCK(&ctx->lock);
+
+ GET_CONTRI_KEY(this, contri_key, contribution->gfid, keylen);
+ if (keylen < 0) {
+ ret = -1;
+ goto out;
+ }
+
+ ret = _quota_dict_get_meta(this, dict, contri_key, keylen, &contri,
+ IA_IFREG, _gf_true);
+ if (ret < 0) {
+ ret = mq_create_xattrs_txn(this, loc, NULL);
+ } else {
+ LOCK(&contribution->lock);
+ {
+ contribution->contribution = contri.size;
+ contribution->file_count = contri.file_count;
+ contribution->dir_count = contri.dir_count;
}
- UNLOCK (&ctx->lock);
-
- GET_CONTRI_KEY (this, contri_key, contribution->gfid, ret);
- if (ret < 0)
- goto out;
+ UNLOCK(&contribution->lock);
- ret = _quota_dict_get_meta (this, dict, contri_key, &contri,
- IA_IFREG, _gf_true);
- if (ret < 0) {
- ret = mq_create_xattrs_txn (this, loc, NULL);
- } else {
- LOCK (&contribution->lock);
- {
- contribution->contribution = contri.size;
- contribution->file_count = contri.file_count;
- contribution->dir_count = contri.dir_count;
- }
- UNLOCK (&contribution->lock);
-
- ret = mq_get_ctx_updation_status (ctx, &status);
- if (ret < 0 || status == _gf_true) {
- /* If the update txn is in progress abort inspection */
- ret = 0;
- goto out;
- }
-
- mq_compute_delta (&delta, &size, &contri);
- if (!quota_meta_is_null (&delta))
- mq_initiate_quota_txn (this, loc, NULL);
+ ret = mq_get_ctx_updation_status(ctx, &status);
+ if (ret < 0 || status == _gf_true) {
+ /* If the update txn is in progress abort inspection */
+ ret = 0;
+ goto out;
}
- /* TODO: revist this code when fixing hardlinks */
+
+ mq_compute_delta(&delta, &size, &contri);
+ if (!quota_meta_is_null(&delta))
+ mq_initiate_quota_txn(this, loc, NULL);
+ }
+ /* TODO: revist this code when fixing hardlinks */
out:
- return ret;
+ return ret;
}
int32_t
-mq_xattr_state (xlator_t *this, loc_t *origin_loc, dict_t *dict,
- struct iatt buf)
+mq_xattr_state(xlator_t *this, loc_t *origin_loc, dict_t *dict,
+ struct iatt *buf)
{
- int32_t ret = -1;
- quota_inode_ctx_t *ctx = NULL;
- loc_t loc = {0, };
- inode_contribution_t *contribution = NULL;
-
- ret = mq_prevalidate_txn (this, origin_loc, &loc, &ctx, &buf);
- if (ret < 0 || loc.parent == NULL)
- goto out;
-
- if (!loc_is_root(&loc)) {
- contribution = mq_add_new_contribution_node (this, ctx, &loc);
- if (contribution == NULL) {
- if (!gf_uuid_is_null (loc.inode->gfid))
- gf_log (this->name, GF_LOG_WARNING,
- "cannot add a new contribution node "
- "(%s)", uuid_utoa (loc.gfid));
- ret = -1;
- goto out;
- }
- }
+ int32_t ret = -1;
+ quota_inode_ctx_t *ctx = NULL;
+ loc_t loc = {
+ 0,
+ };
+ inode_contribution_t *contribution = NULL;
+
+ ret = mq_prevalidate_txn(this, origin_loc, &loc, &ctx, buf);
+ if (ret < 0 || loc.parent == NULL)
+ goto out;
- if (buf.ia_type == IA_IFDIR)
- mq_inspect_directory_xattr (this, ctx, contribution, &loc, dict,
- buf);
+ if (!loc_is_root(&loc)) {
+ contribution = mq_add_new_contribution_node(this, ctx, &loc);
+ if (contribution == NULL) {
+ if (!gf_uuid_is_null(loc.inode->gfid))
+ gf_log(this->name, GF_LOG_WARNING,
+ "cannot add a new contribution node "
+ "(%s)",
+ uuid_utoa(loc.gfid));
+ ret = -1;
+ goto out;
+ }
+ if (buf->ia_type == IA_IFDIR)
+ mq_inspect_directory_xattr(this, ctx, contribution, &loc, dict);
else
- mq_inspect_file_xattr (this, ctx, contribution, &loc, dict,
- buf);
+ mq_inspect_file_xattr(this, ctx, contribution, &loc, dict, buf);
+ } else {
+ mq_inspect_directory_xattr(this, ctx, 0, &loc, dict);
+ }
out:
- loc_wipe (&loc);
+ loc_wipe(&loc);
- if (contribution)
- GF_REF_PUT (contribution);
+ if (contribution)
+ GF_REF_PUT(contribution);
- return ret;
+ return ret;
}
int32_t
-mq_req_xattr (xlator_t *this, loc_t *loc, dict_t *dict,
- char *contri_key, char *size_key)
+mq_req_xattr(xlator_t *this, loc_t *loc, dict_t *dict, char *contri_key,
+ char *size_key)
{
- int32_t ret = -1;
- char key[QUOTA_KEY_MAX] = {0, };
-
- GF_VALIDATE_OR_GOTO ("marker", this, out);
- GF_VALIDATE_OR_GOTO ("marker", loc, out);
- GF_VALIDATE_OR_GOTO ("marker", dict, out);
-
- if (!loc_is_root(loc)) {
- ret = mq_dict_set_contribution (this, dict, loc, NULL,
- contri_key);
- if (ret < 0)
- goto out;
- }
+ int32_t ret = -1;
+ char key[QUOTA_KEY_MAX] = {
+ 0,
+ };
- GET_SIZE_KEY (this, key, ret);
- if (ret < 0)
- goto out;
- if (size_key)
- strncpy (size_key, key, QUOTA_KEY_MAX);
+ GF_VALIDATE_OR_GOTO("marker", this, out);
+ GF_VALIDATE_OR_GOTO("marker", loc, out);
+ GF_VALIDATE_OR_GOTO("marker", dict, out);
- ret = dict_set_uint64 (dict, key, 0);
+ if (!loc_is_root(loc)) {
+ ret = mq_dict_set_contribution(this, dict, loc, NULL, contri_key);
if (ret < 0)
- goto out;
+ goto out;
+ }
- ret = dict_set_int8 (dict, QUOTA_DIRTY_KEY, 0);
+ GET_SIZE_KEY(this, key, ret);
+ if (ret < 0)
+ goto out;
+ if (size_key)
+ if (snprintf(size_key, QUOTA_KEY_MAX, "%s", key) >= QUOTA_KEY_MAX) {
+ ret = -1;
+ goto out;
+ }
-out:
- if (ret < 0)
- gf_log_callingfn (this->name, GF_LOG_ERROR, "dict set failed");
+ ret = dict_set_uint64(dict, key, 0);
+ if (ret < 0)
+ goto out;
- return ret;
-}
+ ret = dict_set_int8(dict, QUOTA_DIRTY_KEY, 0);
+out:
+ if (ret < 0)
+ gf_log_callingfn(this ? this->name : "Marker", GF_LOG_ERROR,
+ "dict set failed");
+ return ret;
+}
int32_t
-mq_forget (xlator_t *this, quota_inode_ctx_t *ctx)
+mq_forget(xlator_t *this, quota_inode_ctx_t *ctx)
{
- inode_contribution_t *contri = NULL;
- inode_contribution_t *next = NULL;
+ inode_contribution_t *contri = NULL;
+ inode_contribution_t *next = NULL;
- GF_VALIDATE_OR_GOTO ("marker", this, out);
- GF_VALIDATE_OR_GOTO ("marker", ctx, out);
+ GF_VALIDATE_OR_GOTO("marker", this, out);
+ GF_VALIDATE_OR_GOTO("marker", ctx, out);
- list_for_each_entry_safe (contri, next, &ctx->contribution_head,
- contri_list) {
- list_del_init (&contri->contri_list);
- GF_REF_PUT (contri);
- }
+ list_for_each_entry_safe(contri, next, &ctx->contribution_head, contri_list)
+ {
+ list_del_init(&contri->contri_list);
+ GF_REF_PUT(contri);
+ }
- LOCK_DESTROY (&ctx->lock);
- GF_FREE (ctx);
+ LOCK_DESTROY(&ctx->lock);
+ GF_FREE(ctx);
out:
- return 0;
+ return 0;
}
diff --git a/xlators/features/marker/src/marker-quota.h b/xlators/features/marker/src/marker-quota.h
index 51e062537b8..4bbf6878b22 100644
--- a/xlators/features/marker/src/marker-quota.h
+++ b/xlators/features/marker/src/marker-quota.h
@@ -10,147 +10,131 @@
#ifndef _MARKER_QUOTA_H
#define _MARKER_QUOTA_H
-#include "xlator.h"
+#include <glusterfs/xlator.h>
#include "marker-mem-types.h"
-#include "refcount.h"
-#include "quota-common-utils.h"
-#include "call-stub.h"
+#include <glusterfs/refcount.h>
+#include <glusterfs/quota-common-utils.h>
+#include <glusterfs/call-stub.h>
#define QUOTA_XATTR_PREFIX "trusted.glusterfs"
#define QUOTA_DIRTY_KEY "trusted.glusterfs.quota.dirty"
-#define CONTRIBUTION "contri"
+#define CONTRIBUTION "contri"
#define QUOTA_KEY_MAX 512
#define READDIR_BUF 4096
-
-#define QUOTA_STACK_DESTROY(_frame, _this) \
- do { \
- quota_local_t *_local = NULL; \
- _local = _frame->local; \
- _frame->local = NULL; \
- STACK_DESTROY (_frame->root); \
- mq_local_unref (_this, _local); \
- } while (0)
-
-
-#define QUOTA_ALLOC(var, type, ret) \
- do { \
- ret = 0; \
- var = GF_CALLOC (sizeof (type), 1, \
- gf_marker_mt_##type); \
- if (!var) { \
- ret = -1; \
- } \
- } while (0);
-
-#define QUOTA_ALLOC_OR_GOTO(var, type, ret, label) \
- do { \
- var = GF_CALLOC (sizeof (type), 1, \
- gf_marker_mt_##type); \
- if (!var) { \
- gf_log ("", GF_LOG_ERROR, \
- "out of memory"); \
- ret = -1; \
- goto label; \
- } \
- ret = 0; \
- } while (0);
-
-#define GET_QUOTA_KEY(_this, var, key, _ret) \
- do { \
- marker_conf_t *_priv = _this->private; \
- if (_priv->version > 0) \
- _ret = snprintf (var, QUOTA_KEY_MAX, "%s.%d", \
- key, _priv->version); \
- else \
- _ret = snprintf (var, QUOTA_KEY_MAX, "%s", key); \
- } while (0)
-
-#define GET_CONTRI_KEY(_this, var, _gfid, _ret) \
- do { \
- char _tmp_var[QUOTA_KEY_MAX] = {0, }; \
- if (_gfid != NULL) { \
- char _gfid_unparsed[40]; \
- gf_uuid_unparse (_gfid, _gfid_unparsed); \
- _ret = snprintf (_tmp_var, QUOTA_KEY_MAX, \
- QUOTA_XATTR_PREFIX \
- ".%s.%s." CONTRIBUTION, \
- "quota", _gfid_unparsed); \
- } else { \
- _ret = snprintf (_tmp_var, QUOTA_KEY_MAX, \
- QUOTA_XATTR_PREFIX \
- ".%s.." CONTRIBUTION, \
- "quota"); \
- } \
- GET_QUOTA_KEY (_this, var, _tmp_var, _ret); \
- } while (0)
-
-#define GET_SIZE_KEY(_this, var, _ret) \
- { \
- GET_QUOTA_KEY (_this, var, QUOTA_SIZE_KEY, _ret); \
- }
-
-#define QUOTA_SAFE_INCREMENT(lock, var) \
- do { \
- LOCK (lock); \
- var ++; \
- UNLOCK (lock); \
- } while (0)
+#define QUOTA_ALLOC(var, type, ret) \
+ do { \
+ ret = 0; \
+ var = GF_CALLOC(sizeof(type), 1, gf_marker_mt_##type); \
+ if (!var) { \
+ ret = -1; \
+ } \
+ } while (0);
+
+#define QUOTA_ALLOC_OR_GOTO(var, type, ret, label) \
+ do { \
+ var = GF_CALLOC(sizeof(type), 1, gf_marker_mt_##type); \
+ if (!var) { \
+ gf_log("", GF_LOG_ERROR, "out of memory"); \
+ ret = -1; \
+ goto label; \
+ } \
+ ret = 0; \
+ } while (0);
+
+#define GET_QUOTA_KEY(_this, var, key, _ret) \
+ do { \
+ marker_conf_t *_priv = _this->private; \
+ if (_priv->version > 0) \
+ _ret = snprintf(var, QUOTA_KEY_MAX, "%s.%d", key, _priv->version); \
+ else \
+ _ret = snprintf(var, QUOTA_KEY_MAX, "%s", key); \
+ } while (0)
+
+#define GET_CONTRI_KEY(_this, var, _gfid, _ret) \
+ do { \
+ char _tmp_var[QUOTA_KEY_MAX] = { \
+ 0, \
+ }; \
+ if (_gfid != NULL) { \
+ char _gfid_unparsed[40]; \
+ gf_uuid_unparse(_gfid, _gfid_unparsed); \
+ _ret = snprintf(_tmp_var, QUOTA_KEY_MAX, \
+ QUOTA_XATTR_PREFIX ".%s.%s." CONTRIBUTION, \
+ "quota", _gfid_unparsed); \
+ } else { \
+ _ret = snprintf(_tmp_var, QUOTA_KEY_MAX, \
+ QUOTA_XATTR_PREFIX ".%s.." CONTRIBUTION, "quota"); \
+ } \
+ GET_QUOTA_KEY(_this, var, _tmp_var, _ret); \
+ } while (0)
+
+#define GET_SIZE_KEY(_this, var, _ret) \
+ { \
+ GET_QUOTA_KEY(_this, var, QUOTA_SIZE_KEY, _ret); \
+ }
+
+#define QUOTA_SAFE_INCREMENT(lock, var) \
+ do { \
+ LOCK(lock); \
+ var++; \
+ UNLOCK(lock); \
+ } while (0)
struct quota_inode_ctx {
- int64_t size;
- int64_t file_count;
- int64_t dir_count;
- int8_t dirty;
- gf_boolean_t create_status;
- gf_boolean_t updation_status;
- gf_boolean_t dirty_status;
- gf_lock_t lock;
- struct list_head contribution_head;
+ int64_t size;
+ int64_t file_count;
+ int64_t dir_count;
+ int8_t dirty;
+ gf_boolean_t create_status;
+ gf_boolean_t updation_status;
+ gf_boolean_t dirty_status;
+ gf_lock_t lock;
+ struct list_head contribution_head;
};
typedef struct quota_inode_ctx quota_inode_ctx_t;
struct quota_synctask {
- xlator_t *this;
- loc_t loc;
- quota_meta_t contri;
- gf_boolean_t is_static;
- uint32_t ia_nlink;
- call_stub_t *stub;
+ xlator_t *this;
+ loc_t loc;
+ quota_meta_t contri;
+ gf_boolean_t is_static;
+ uint32_t ia_nlink;
+ call_stub_t *stub;
};
typedef struct quota_synctask quota_synctask_t;
struct inode_contribution {
- struct list_head contri_list;
- int64_t contribution;
- int64_t file_count;
- int64_t dir_count;
- uuid_t gfid;
- gf_lock_t lock;
- GF_REF_DECL;
+ struct list_head contri_list;
+ int64_t contribution;
+ int64_t file_count;
+ int64_t dir_count;
+ uuid_t gfid;
+ gf_lock_t lock;
+ GF_REF_DECL;
};
typedef struct inode_contribution inode_contribution_t;
int32_t
-mq_req_xattr (xlator_t *, loc_t *, dict_t *, char *, char *);
+mq_req_xattr(xlator_t *, loc_t *, dict_t *, char *, char *);
int32_t
-mq_xattr_state (xlator_t *, loc_t *, dict_t *, struct iatt);
+mq_xattr_state(xlator_t *, loc_t *, dict_t *, struct iatt *);
int
-mq_initiate_quota_txn (xlator_t *, loc_t *, struct iatt *);
+mq_initiate_quota_txn(xlator_t *, loc_t *, struct iatt *);
int
-mq_initiate_quota_blocking_txn (xlator_t *, loc_t *, struct iatt *);
+mq_initiate_quota_blocking_txn(xlator_t *, loc_t *, struct iatt *);
int
-mq_create_xattrs_txn (xlator_t *this, loc_t *loc, struct iatt *buf);
+mq_create_xattrs_txn(xlator_t *this, loc_t *loc, struct iatt *buf);
int32_t
-mq_reduce_parent_size_txn (xlator_t *, loc_t *, quota_meta_t *,
- uint32_t nlink, call_stub_t *stub);
+mq_reduce_parent_size_txn(xlator_t *, loc_t *, quota_meta_t *, uint32_t nlink,
+ call_stub_t *stub);
int32_t
-mq_forget (xlator_t *, quota_inode_ctx_t *);
+mq_forget(xlator_t *, quota_inode_ctx_t *);
#endif
diff --git a/xlators/features/marker/src/marker.c b/xlators/features/marker/src/marker.c
index 95297922ec5..1375ccc498c 100644
--- a/xlators/features/marker/src/marker.c
+++ b/xlators/features/marker/src/marker.c
@@ -7,34 +7,34 @@
later), or the GNU General Public License, version 2 (GPLv2), in all
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "defaults.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
#include "libxlator.h"
#include "marker.h"
#include "marker-mem-types.h"
#include "marker-quota.h"
#include "marker-quota-helper.h"
#include "marker-common.h"
-#include "byte-order.h"
-#include "syncop.h"
-#include "syscall.h"
+#include <glusterfs/byte-order.h>
+#include <glusterfs/syncop.h>
+#include <glusterfs/syscall.h>
#include <fnmatch.h>
#define _GF_UID_GID_CHANGED 1
static char *mq_ext_xattrs[] = {
- QUOTA_SIZE_KEY,
- QUOTA_LIMIT_KEY,
- QUOTA_LIMIT_OBJECTS_KEY,
- NULL,
+ QUOTA_SIZE_KEY,
+ QUOTA_LIMIT_KEY,
+ QUOTA_LIMIT_OBJECTS_KEY,
+ NULL,
};
void
-fini (xlator_t *this);
+fini(xlator_t *this);
int32_t
-marker_start_setxattr (call_frame_t *, xlator_t *);
+marker_start_setxattr(call_frame_t *, xlator_t *);
/* When client/quotad request for quota xattrs,
* replace the key-name by adding the version number
@@ -46,1578 +46,1566 @@ marker_start_setxattr (call_frame_t *, xlator_t *);
* version for the key-name
*/
int
-marker_key_replace_with_ver (xlator_t *this, dict_t *dict)
+marker_key_replace_with_ver(xlator_t *this, dict_t *dict)
{
- int ret = -1;
- int i = 0;
- marker_conf_t *priv = NULL;
- char key[QUOTA_KEY_MAX] = {0, };
+ int ret = -1;
+ int i = 0;
+ marker_conf_t *priv = NULL;
+ char key[QUOTA_KEY_MAX] = {
+ 0,
+ };
- priv = this->private;
+ priv = this->private;
- if (dict == NULL || priv->version <= 0) {
- ret = 0;
+ if (dict == NULL || priv->version <= 0) {
+ ret = 0;
+ goto out;
+ }
+
+ for (i = 0; mq_ext_xattrs[i]; i++) {
+ if (dict_get(dict, mq_ext_xattrs[i])) {
+ GET_QUOTA_KEY(this, key, mq_ext_xattrs[i], ret);
+ if (ret < 0)
goto out;
- }
- for (i = 0; mq_ext_xattrs[i]; i++) {
- if (dict_get (dict, mq_ext_xattrs[i])) {
- GET_QUOTA_KEY (this, key, mq_ext_xattrs[i], ret);
- if (ret < 0)
- goto out;
-
- ret = dict_set (dict, key,
- dict_get (dict, mq_ext_xattrs[i]));
- if (ret < 0)
- goto out;
-
- dict_del (dict, mq_ext_xattrs[i]);
- }
+ ret = dict_set(dict, key, dict_get(dict, mq_ext_xattrs[i]));
+ if (ret < 0)
+ goto out;
+
+ dict_del(dict, mq_ext_xattrs[i]);
}
+ }
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-marker_key_set_ver (xlator_t *this, dict_t *dict)
+marker_key_set_ver(xlator_t *this, dict_t *dict)
{
- int ret = -1;
- int i = -1;
- marker_conf_t *priv = NULL;
- char key[QUOTA_KEY_MAX] = {0, };
+ int ret = -1;
+ int i = -1;
+ marker_conf_t *priv = NULL;
+ char key[QUOTA_KEY_MAX] = {
+ 0,
+ };
- priv = this->private;
+ priv = this->private;
- if (dict == NULL || priv->version <= 0) {
- ret = 0;
- goto out;
- }
+ if (dict == NULL || priv->version <= 0) {
+ ret = 0;
+ goto out;
+ }
- for (i = 0; mq_ext_xattrs[i]; i++) {
- GET_QUOTA_KEY (this, key, mq_ext_xattrs[i], ret);
- if (ret < 0)
- goto out;
+ for (i = 0; mq_ext_xattrs[i]; i++) {
+ GET_QUOTA_KEY(this, key, mq_ext_xattrs[i], ret);
+ if (ret < 0)
+ goto out;
- if (dict_get (dict, key))
- dict_set (dict, mq_ext_xattrs[i], dict_get (dict, key));
- }
+ if (dict_get(dict, key))
+ dict_set(dict, mq_ext_xattrs[i], dict_get(dict, key));
+ }
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
marker_local_t *
-marker_local_ref (marker_local_t *local)
+marker_local_ref(marker_local_t *local)
{
- GF_VALIDATE_OR_GOTO ("marker", local, err);
+ GF_VALIDATE_OR_GOTO("marker", local, err);
- LOCK (&local->lock);
- {
- local->ref++;
- }
- UNLOCK (&local->lock);
+ LOCK(&local->lock);
+ {
+ local->ref++;
+ }
+ UNLOCK(&local->lock);
- return local;
+ return local;
err:
- return NULL;
+ return NULL;
}
int
-marker_loc_fill (loc_t *loc, inode_t *inode, inode_t *parent, char *path)
+marker_loc_fill(loc_t *loc, inode_t *inode, inode_t *parent, char *path)
{
- int ret = -1;
+ int ret = -1;
- if (!loc)
- return ret;
+ if (!loc)
+ return ret;
- if (inode) {
- loc->inode = inode_ref (inode);
- if (gf_uuid_is_null (loc->gfid)) {
- gf_uuid_copy (loc->gfid, loc->inode->gfid);
- }
+ if (inode) {
+ loc->inode = inode_ref(inode);
+ if (gf_uuid_is_null(loc->gfid)) {
+ gf_uuid_copy(loc->gfid, loc->inode->gfid);
}
+ }
- if (parent)
- loc->parent = inode_ref (parent);
-
- if (path) {
- loc->path = gf_strdup (path);
- if (!loc->path) {
- gf_log ("loc fill", GF_LOG_ERROR, "strdup failed");
- goto loc_wipe;
- }
-
- loc->name = strrchr (loc->path, '/');
- if (loc->name)
- loc->name++;
+ if (parent)
+ loc->parent = inode_ref(parent);
+
+ if (path) {
+ loc->path = gf_strdup(path);
+ if (!loc->path) {
+ gf_log("loc fill", GF_LOG_ERROR, "strdup failed");
+ goto loc_wipe;
}
- ret = 0;
+ loc->name = strrchr(loc->path, '/');
+ if (loc->name)
+ loc->name++;
+ }
+
+ ret = 0;
loc_wipe:
- if (ret < 0)
- loc_wipe (loc);
+ if (ret < 0)
+ loc_wipe(loc);
- return ret;
+ return ret;
}
int
-_marker_inode_loc_fill (inode_t *inode, inode_t *parent, char *name, loc_t *loc)
+_marker_inode_loc_fill(inode_t *inode, inode_t *parent, char *name, loc_t *loc)
{
- char *resolvedpath = NULL;
- int ret = -1;
- gf_boolean_t free_parent = _gf_false;
+ char *resolvedpath = NULL;
+ int ret = -1;
+ gf_boolean_t free_parent = _gf_false;
- if ((!inode) || (!loc))
- return ret;
+ if ((!inode) || (!loc))
+ return ret;
- if (parent && name)
- ret = inode_path (parent, name, &resolvedpath);
- else
- ret = inode_path (inode, NULL, &resolvedpath);
- if (ret < 0)
- goto err;
+ if (parent && name)
+ ret = inode_path(parent, name, &resolvedpath);
+ else
+ ret = inode_path(inode, NULL, &resolvedpath);
+ if (ret < 0)
+ goto err;
- if (parent == NULL) {
- parent = inode_parent (inode, NULL, NULL);
- free_parent = _gf_true;
- }
+ if (parent == NULL) {
+ parent = inode_parent(inode, NULL, NULL);
+ free_parent = _gf_true;
+ }
- ret = marker_loc_fill (loc, inode, parent, resolvedpath);
- if (ret < 0)
- goto err;
+ ret = marker_loc_fill(loc, inode, parent, resolvedpath);
+ if (ret < 0)
+ goto err;
err:
- if (free_parent)
- inode_unref (parent);
+ if (free_parent)
+ inode_unref(parent);
- GF_FREE (resolvedpath);
+ GF_FREE(resolvedpath);
- return ret;
+ return ret;
}
int
-marker_inode_loc_fill (inode_t *inode, loc_t *loc)
+marker_inode_loc_fill(inode_t *inode, loc_t *loc)
{
- return _marker_inode_loc_fill (inode, NULL, NULL, loc);
+ return _marker_inode_loc_fill(inode, NULL, NULL, loc);
}
int32_t
-marker_trav_parent (marker_local_t *local)
+marker_trav_parent(marker_local_t *local)
{
- int32_t ret = 0;
- loc_t loc = {0, };
- inode_t *parent = NULL;
- int8_t need_unref = 0;
+ int32_t ret = 0;
+ loc_t loc = {
+ 0,
+ };
+ inode_t *parent = NULL;
+ int8_t need_unref = 0;
- if (!local->loc.parent) {
- parent = inode_parent (local->loc.inode, NULL, NULL);
- if (parent)
- need_unref = 1;
- } else
- parent = local->loc.parent;
+ if (!local->loc.parent) {
+ parent = inode_parent(local->loc.inode, NULL, NULL);
+ if (parent)
+ need_unref = 1;
+ } else
+ parent = local->loc.parent;
- ret = marker_inode_loc_fill (parent, &loc);
+ ret = marker_inode_loc_fill(parent, &loc);
- if (ret < 0) {
- ret = -1;
- goto out;
- }
+ if (ret < 0) {
+ ret = -1;
+ goto out;
+ }
- loc_wipe (&local->loc);
+ loc_wipe(&local->loc);
- local->loc = loc;
+ local->loc = loc;
out:
- if (need_unref)
- inode_unref (parent);
+ if (need_unref)
+ inode_unref(parent);
- return ret;
+ return ret;
}
-int32_t
-marker_error_handler (xlator_t *this, marker_local_t *local, int32_t op_errno)
+void
+marker_error_handler(xlator_t *this, marker_local_t *local, int32_t op_errno)
{
- marker_conf_t *priv = NULL;
- const char *path = NULL;
-
- priv = (marker_conf_t *) this->private;
- path = local
- ? (local->loc.path
- ? local->loc.path : uuid_utoa(local->loc.gfid))
- : "<nul>";
-
- gf_log (this->name, GF_LOG_CRITICAL,
- "Indexing gone corrupt at %s (reason: %s)."
- " Geo-replication slave content needs to be revalidated",
- path, strerror (op_errno));
- sys_unlink (priv->timestamp_file);
+ marker_conf_t *priv = (marker_conf_t *)this->private;
+ const char *path = local ? ((local->loc.path) ? local->loc.path
+ : uuid_utoa(local->loc.gfid))
+ : "<nul>";
- return 0;
+ gf_log(this->name, GF_LOG_CRITICAL,
+ "Indexing gone corrupt at %s (reason: %s)."
+ " Geo-replication slave content needs to be revalidated",
+ path, strerror(op_errno));
+ sys_unlink(priv->timestamp_file);
}
int32_t
-marker_local_unref (marker_local_t *local)
+marker_local_unref(marker_local_t *local)
{
- int32_t var = 0;
-
- if (local == NULL)
- return -1;
+ int32_t var = 0;
- LOCK (&local->lock);
- {
- var = --local->ref;
- }
- UNLOCK (&local->lock);
-
- if (var != 0)
- goto out;
+ if (local == NULL)
+ return -1;
- loc_wipe (&local->loc);
- loc_wipe (&local->parent_loc);
- if (local->xdata)
- dict_unref (local->xdata);
+ LOCK(&local->lock);
+ {
+ var = --local->ref;
+ }
+ UNLOCK(&local->lock);
- if (local->lk_frame) {
- STACK_DESTROY (local->lk_frame->root);
- local->lk_frame = NULL;
- }
+ if (var != 0)
+ goto out;
- if (local->oplocal) {
- marker_local_unref (local->oplocal);
- local->oplocal = NULL;
- }
- mem_put (local);
+ loc_wipe(&local->loc);
+ loc_wipe(&local->parent_loc);
+ if (local->xdata)
+ dict_unref(local->xdata);
+
+ if (local->lk_frame) {
+ STACK_DESTROY(local->lk_frame->root);
+ local->lk_frame = NULL;
+ }
+
+ if (local->oplocal) {
+ marker_local_unref(local->oplocal);
+ local->oplocal = NULL;
+ }
+ mem_put(local);
out:
- return 0;
+ return 0;
}
int32_t
-stat_stampfile (xlator_t *this, marker_conf_t *priv,
- struct volume_mark **status)
+stat_stampfile(xlator_t *this, marker_conf_t *priv, struct volume_mark **status)
{
- struct stat buf = {0, };
- struct volume_mark *vol_mark = NULL;
+ struct stat buf = {
+ 0,
+ };
+ struct volume_mark *vol_mark = NULL;
- vol_mark = GF_CALLOC (sizeof (struct volume_mark), 1,
- gf_marker_mt_volume_mark);
+ vol_mark = GF_CALLOC(sizeof(struct volume_mark), 1,
+ gf_marker_mt_volume_mark);
- vol_mark->major = 1;
- vol_mark->minor = 0;
+ vol_mark->major = 1;
+ vol_mark->minor = 0;
- GF_ASSERT (sizeof (priv->volume_uuid_bin) == 16);
- memcpy (vol_mark->uuid, priv->volume_uuid_bin, 16);
+ GF_ASSERT(sizeof(priv->volume_uuid_bin) == 16);
+ memcpy(vol_mark->uuid, priv->volume_uuid_bin, 16);
- if (sys_stat (priv->timestamp_file, &buf) != -1) {
- vol_mark->retval = 0;
- vol_mark->sec = htonl (buf.st_mtime);
- vol_mark->usec = htonl (ST_MTIM_NSEC (&buf)/1000);
- } else
- vol_mark->retval = 1;
+ if (sys_stat(priv->timestamp_file, &buf) != -1) {
+ vol_mark->retval = 0;
+ vol_mark->sec = htonl(buf.st_mtime);
+ vol_mark->usec = htonl(ST_MTIM_NSEC(&buf) / 1000);
+ } else
+ vol_mark->retval = 1;
- *status = vol_mark;
+ *status = vol_mark;
- return 0;
+ return 0;
}
int32_t
-marker_getxattr_stampfile_cbk (call_frame_t *frame, xlator_t *this,
- const char *name, struct volume_mark *vol_mark,
- dict_t *xdata)
+marker_getxattr_stampfile_cbk(call_frame_t *frame, xlator_t *this,
+ const char *name, struct volume_mark *vol_mark,
+ dict_t *xdata)
{
- int32_t ret = -1;
- dict_t *dict = NULL;
+ int32_t ret = -1;
+ dict_t *dict = NULL;
- if (vol_mark == NULL){
- STACK_UNWIND_STRICT (getxattr, frame, -1, ENOMEM, NULL, NULL);
+ if (vol_mark == NULL) {
+ STACK_UNWIND_STRICT(getxattr, frame, -1, ENOMEM, NULL, NULL);
- goto out;
- }
+ goto out;
+ }
- dict = dict_new ();
+ dict = dict_new();
- ret = dict_set_bin (dict, (char *)name, vol_mark,
- sizeof (struct volume_mark));
- if (ret) {
- GF_FREE (vol_mark);
- gf_log (this->name, GF_LOG_WARNING, "failed to set key %s",
- name);
- }
+ ret = dict_set_bin(dict, (char *)name, vol_mark,
+ sizeof(struct volume_mark));
+ if (ret) {
+ GF_FREE(vol_mark);
+ gf_log(this->name, GF_LOG_WARNING, "failed to set key %s", name);
+ }
- STACK_UNWIND_STRICT (getxattr, frame, 0, 0, dict, xdata);
+ STACK_UNWIND_STRICT(getxattr, frame, 0, 0, dict, xdata);
- if (dict)
- dict_unref (dict);
+ if (dict)
+ dict_unref(dict);
out:
- return 0;
+ return 0;
}
gf_boolean_t
-call_from_special_client (call_frame_t *frame, xlator_t *this, const char *name)
+call_from_special_client(call_frame_t *frame, xlator_t *this, const char *name)
{
- struct volume_mark *vol_mark = NULL;
- marker_conf_t *priv = NULL;
- gf_boolean_t is_true = _gf_true;
+ struct volume_mark *vol_mark = NULL;
+ marker_conf_t *priv = NULL;
+ gf_boolean_t is_true = _gf_true;
- priv = (marker_conf_t *)this->private;
+ priv = (marker_conf_t *)this->private;
- if (frame->root->pid != GF_CLIENT_PID_GSYNCD || name == NULL ||
- strcmp (name, MARKER_XATTR_PREFIX "." VOLUME_MARK) != 0) {
- is_true = _gf_false;
- goto out;
- }
+ if (frame->root->pid != GF_CLIENT_PID_GSYNCD || name == NULL ||
+ strcmp(name, MARKER_XATTR_PREFIX "." VOLUME_MARK) != 0) {
+ is_true = _gf_false;
+ goto out;
+ }
- stat_stampfile (this, priv, &vol_mark);
+ stat_stampfile(this, priv, &vol_mark);
- marker_getxattr_stampfile_cbk (frame, this, name, vol_mark, NULL);
+ marker_getxattr_stampfile_cbk(frame, this, name, vol_mark, NULL);
out:
- return is_true;
+ return is_true;
}
static gf_boolean_t
-_is_quota_internal_xattr (dict_t *d, char *k, data_t *v, void *data)
+_is_quota_internal_xattr(dict_t *d, char *k, data_t *v, void *data)
{
- int i = 0;
- char **external_xattrs = data;
+ int i = 0;
+ char **external_xattrs = data;
- for (i = 0; external_xattrs && external_xattrs[i]; i++) {
- if (strcmp (k, external_xattrs[i]) == 0)
- return _gf_false;
- }
+ for (i = 0; external_xattrs && external_xattrs[i]; i++) {
+ if (strcmp(k, external_xattrs[i]) == 0)
+ return _gf_false;
+ }
- if (fnmatch ("trusted.glusterfs.quota*", k, 0) == 0)
- return _gf_true;
+ if (fnmatch("trusted.glusterfs.quota*", k, 0) == 0)
+ return _gf_true;
- /* It would be nice if posix filters pgfid xattrs. But since marker
- * also takes up responsibility to clean these up, adding the filtering
- * here (Check 'quota_xattr_cleaner')
- */
- if (fnmatch (PGFID_XATTR_KEY_PREFIX"*", k, 0) == 0)
- return _gf_true;
+ /* It would be nice if posix filters pgfid xattrs. But since marker
+ * also takes up responsibility to clean these up, adding the filtering
+ * here (Check 'quota_xattr_cleaner')
+ */
+ if (fnmatch(PGFID_XATTR_KEY_PREFIX "*", k, 0) == 0)
+ return _gf_true;
- return _gf_false;
+ return _gf_false;
}
static void
-marker_filter_internal_xattrs (xlator_t *this, dict_t *xattrs)
+marker_filter_internal_xattrs(xlator_t *this, dict_t *xattrs)
{
- marker_conf_t *priv = NULL;
- char **ext = NULL;
+ marker_conf_t *priv = NULL;
+ char **ext = NULL;
- priv = this->private;
- if (priv->feature_enabled & GF_QUOTA)
- ext = mq_ext_xattrs;
+ priv = this->private;
+ if (priv->feature_enabled & GF_QUOTA)
+ ext = mq_ext_xattrs;
- dict_foreach_match (xattrs, _is_quota_internal_xattr, ext,
- dict_remove_foreach_fn, NULL);
+ dict_foreach_match(xattrs, _is_quota_internal_xattr, ext,
+ dict_remove_foreach_fn, NULL);
}
static void
-marker_filter_gsyncd_xattrs (call_frame_t *frame,
- xlator_t *this, dict_t *xattrs)
+marker_filter_gsyncd_xattrs(call_frame_t *frame, xlator_t *this, dict_t *xattrs)
{
- marker_conf_t *priv = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
- GF_ASSERT (priv);
- GF_ASSERT (frame);
+ priv = this->private;
+ GF_ASSERT(priv);
+ GF_ASSERT(frame);
- if (xattrs && frame->root->pid != GF_CLIENT_PID_GSYNCD) {
- GF_REMOVE_INTERNAL_XATTR (GF_XATTR_XTIME_PATTERN, xattrs);
- }
- return;
+ if (xattrs && frame->root->pid != GF_CLIENT_PID_GSYNCD) {
+ GF_REMOVE_INTERNAL_XATTR(GF_XATTR_XTIME_PATTERN, xattrs);
+ }
+ return;
}
int32_t
-marker_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
-{
- int32_t ret = -1;
- if (op_ret < 0)
- goto unwind;
-
- ret = marker_key_set_ver (this, dict);
- if (ret < 0) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
-
- if (cookie) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Filtering the quota extended attributes");
-
- /* If the getxattr is from a non special client, then do not
- copy the quota related xattrs (except the quota limit key
- i.e trusted.glusterfs.quota.limit-set which has been set by
- glusterd on the directory on which quota limit is set.) for
- directories. Let the healing of xattrs happen upon lookup.
- NOTE: setting of trusted.glusterfs.quota.limit-set as of now
- happens from glusterd. It should be moved to quotad. Also
- trusted.glusterfs.quota.limit-set is set on directory which
- is permanent till quota is removed on that directory or limit
- is changed. So let that xattr be healed by other xlators
- properly whenever directory healing is done.
- */
- /*
- * Except limit-set xattr, rest of the xattrs are maintained
- * by quota xlator. Don't expose them to other xlators.
- * This filter makes sure quota xattrs are not healed as part of
- * metadata self-heal
- */
- marker_filter_internal_xattrs (frame->this, dict);
- }
+marker_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
+{
+ int32_t ret = -1;
+ if (op_ret < 0)
+ goto unwind;
+
+ ret = marker_key_set_ver(this, dict);
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ if (cookie) {
+ gf_log(this->name, GF_LOG_DEBUG,
+ "Filtering the quota extended attributes");
+
+ /* If the getxattr is from a non special client, then do not
+ copy the quota related xattrs (except the quota limit key
+ i.e trusted.glusterfs.quota.limit-set which has been set by
+ glusterd on the directory on which quota limit is set.) for
+ directories. Let the healing of xattrs happen upon lookup.
+ NOTE: setting of trusted.glusterfs.quota.limit-set as of now
+ happens from glusterd. It should be moved to quotad. Also
+ trusted.glusterfs.quota.limit-set is set on directory which
+ is permanent till quota is removed on that directory or limit
+ is changed. So let that xattr be healed by other xlators
+ properly whenever directory healing is done.
+ */
+ /*
+ * Except limit-set xattr, rest of the xattrs are maintained
+ * by quota xlator. Don't expose them to other xlators.
+ * This filter makes sure quota xattrs are not healed as part of
+ * metadata self-heal
+ */
+ marker_filter_internal_xattrs(frame->this, dict);
+ }
- /* Filter gsyncd xtime xattr for non gsyncd clients */
- marker_filter_gsyncd_xattrs (frame, frame->this, dict);
+ /* Filter gsyncd xtime xattr for non gsyncd clients */
+ marker_filter_gsyncd_xattrs(frame, frame->this, dict);
unwind:
- MARKER_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ MARKER_STACK_UNWIND(getxattr, frame, op_ret, op_errno, dict, xdata);
+ return 0;
}
int32_t
-marker_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
-{
- gf_boolean_t is_true = _gf_false;
- marker_conf_t *priv = NULL;
- unsigned long cookie = 0;
- marker_local_t *local = NULL;
- char key[QUOTA_KEY_MAX] = {0, };
- int32_t ret = -1;
- int32_t i = 0;
-
- priv = this->private;
-
- if (name) {
- for (i = 0; mq_ext_xattrs[i]; i++) {
- if (strcmp (name, mq_ext_xattrs[i]))
- continue;
-
- GET_QUOTA_KEY (this, key, mq_ext_xattrs[i], ret);
- if (ret < 0)
- goto out;
- name = key;
- break;
- }
- }
+marker_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
+{
+ gf_boolean_t is_true = _gf_false;
+ marker_conf_t *priv = NULL;
+ unsigned long cookie = 0;
+ marker_local_t *local = NULL;
+ char key[QUOTA_KEY_MAX] = {
+ 0,
+ };
+ int32_t ret = -1;
+ int32_t i = 0;
+
+ priv = this->private;
+
+ if (name) {
+ for (i = 0; mq_ext_xattrs[i]; i++) {
+ if (strcmp(name, mq_ext_xattrs[i]))
+ continue;
- frame->local = mem_get0 (this->local_pool);
- local = frame->local;
- if (local == NULL)
+ GET_QUOTA_KEY(this, key, mq_ext_xattrs[i], ret);
+ if (ret < 0)
goto out;
+ name = key;
+ break;
+ }
+ }
- MARKER_INIT_LOCAL (frame, local);
-
- if ((loc_copy (&local->loc, loc)) < 0)
- goto out;
-
- gf_log (this->name, GF_LOG_DEBUG, "USER:PID = %d", frame->root->pid);
-
- if (priv && priv->feature_enabled & GF_XTIME)
- is_true = call_from_special_client (frame, this, name);
-
- if (is_true == _gf_false) {
- if (name == NULL) {
- /* Signifies that marker translator
- * has to filter the quota's xattr's,
- * this is to prevent afr from performing
- * self healing on marker-quota xattrs'
- */
- cookie = 1;
- }
- STACK_WIND_COOKIE (frame, marker_getxattr_cbk,
- (void *)cookie,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr,
- loc, name, xdata);
+ frame->local = mem_get0(this->local_pool);
+ local = frame->local;
+ if (local == NULL)
+ goto out;
+
+ MARKER_INIT_LOCAL(frame, local);
+
+ if ((loc_copy(&local->loc, loc)) < 0)
+ goto out;
+
+ gf_log(this->name, GF_LOG_DEBUG, "USER:PID = %d", frame->root->pid);
+
+ if (priv && priv->feature_enabled & GF_XTIME)
+ is_true = call_from_special_client(frame, this, name);
+
+ if (is_true == _gf_false) {
+ if (name == NULL) {
+ /* Signifies that marker translator
+ * has to filter the quota's xattr's,
+ * this is to prevent afr from performing
+ * self healing on marker-quota xattrs'
+ */
+ cookie = 1;
}
+ STACK_WIND_COOKIE(frame, marker_getxattr_cbk, (void *)cookie,
+ FIRST_CHILD(this), FIRST_CHILD(this)->fops->getxattr,
+ loc, name, xdata);
+ }
- return 0;
+ return 0;
out:
- MARKER_STACK_UNWIND (getxattr, frame, -1, ENOMEM, NULL, NULL);
- return 0;
+ MARKER_STACK_UNWIND(getxattr, frame, -1, ENOMEM, NULL, NULL);
+ return 0;
}
int32_t
-marker_setxattr_done (call_frame_t *frame)
+marker_setxattr_done(call_frame_t *frame)
{
- marker_local_t *local = NULL;
+ marker_local_t *local = NULL;
- local = (marker_local_t *) frame->local;
+ local = (marker_local_t *)frame->local;
- frame->local = NULL;
+ frame->local = NULL;
- STACK_DESTROY (frame->root);
+ STACK_DESTROY(frame->root);
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int
-marker_specific_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+marker_specific_setxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- int32_t ret = 0;
- int32_t done = 0;
- marker_local_t *local = NULL;
+ int32_t ret = 0;
+ int32_t done = 1;
+ marker_local_t *local = NULL;
- local = (marker_local_t*) frame->local;
+ local = (marker_local_t *)frame->local;
- if (op_ret == -1 && op_errno == ENOSPC) {
- marker_error_handler (this, local, op_errno);
- done = 1;
- goto out;
- }
+ if (op_ret == -1 && op_errno == ENOSPC) {
+ marker_error_handler(this, local, op_errno);
+ goto out;
+ }
- if (local) {
- if (local->loc.path && strcmp (local->loc.path, "/") == 0) {
- done = 1;
- goto out;
- }
- if (__is_root_gfid (local->loc.gfid)) {
- done = 1;
- goto out;
- }
+ if (local) {
+ if (local->loc.path && strcmp(local->loc.path, "/") == 0) {
+ goto out;
}
-
- ret = (local) ? marker_trav_parent (local) : -1;
-
- if (ret == -1) {
- gf_log (this->name, GF_LOG_DEBUG, "Error occurred "
- "while traversing to the parent, stopping marker");
-
- done = 1;
-
- goto out;
+ if (__is_root_gfid(local->loc.gfid)) {
+ goto out;
}
+ }
- marker_start_setxattr (frame, this);
+ ret = (local) ? marker_trav_parent(local) : -1;
+ if (ret == -1) {
+ gf_log(this->name, GF_LOG_DEBUG,
+ "Error occurred "
+ "while traversing to the parent, stopping marker");
+ goto out;
+ }
+
+ marker_start_setxattr(frame, this);
+ done = 0;
out:
- if (done) {
- marker_setxattr_done (frame);
- }
+ if (done) {
+ marker_setxattr_done(frame);
+ }
- return 0;
+ return 0;
}
int32_t
-marker_start_setxattr (call_frame_t *frame, xlator_t *this)
+marker_start_setxattr(call_frame_t *frame, xlator_t *this)
{
- int32_t ret = -1;
- dict_t *dict = NULL;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = -1;
+ dict_t *dict = NULL;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- local = (marker_local_t*) frame->local;
+ local = (marker_local_t *)frame->local;
- if (!local)
- goto out;
+ if (!local)
+ goto out;
- dict = dict_new ();
+ dict = dict_new();
- if (!dict)
- goto out;
+ if (!dict)
+ goto out;
- if (local->loc.inode && gf_uuid_is_null (local->loc.gfid))
- gf_uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+ if (local->loc.inode && gf_uuid_is_null(local->loc.gfid))
+ gf_uuid_copy(local->loc.gfid, local->loc.inode->gfid);
- GF_UUID_ASSERT (local->loc.gfid);
+ GF_UUID_ASSERT(local->loc.gfid);
- ret = dict_set_static_bin (dict, priv->marker_xattr,
- (void *)local->timebuf, 8);
- if (ret) {
- gf_log (this->name, GF_LOG_WARNING,
- "failed to set marker xattr (%s)", local->loc.path);
- goto out;
- }
+ ret = dict_set_static_bin(dict, priv->marker_xattr, (void *)local->timebuf,
+ 8);
+ if (ret) {
+ gf_log(this->name, GF_LOG_WARNING, "failed to set marker xattr (%s)",
+ local->loc.path);
+ goto out;
+ }
- STACK_WIND (frame, marker_specific_setxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setxattr, &local->loc, dict, 0,
- NULL);
+ STACK_WIND(frame, marker_specific_setxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, &local->loc, dict, 0, NULL);
- ret = 0;
+ ret = 0;
out:
- if (dict)
- dict_unref (dict);
+ if (dict)
+ dict_unref(dict);
- return ret;
+ return ret;
}
void
-marker_gettimeofday (marker_local_t *local)
+marker_gettimeofday(marker_local_t *local)
{
- struct timeval tv = {0, };
+ struct timeval tv = {
+ 0,
+ };
- gettimeofday (&tv, NULL);
+ gettimeofday(&tv, NULL);
- local->timebuf [0] = htonl (tv.tv_sec);
- local->timebuf [1] = htonl (tv.tv_usec);
+ local->timebuf[0] = htonl(tv.tv_sec);
+ local->timebuf[1] = htonl(tv.tv_usec);
- return;
+ return;
}
int32_t
-marker_create_frame (xlator_t *this, marker_local_t *local)
+marker_create_frame(xlator_t *this, marker_local_t *local)
{
- call_frame_t *frame = NULL;
+ call_frame_t *frame = NULL;
- frame = create_frame (this, this->ctx->pool);
+ frame = create_frame(this, this->ctx->pool);
- if (!frame)
- return -1;
+ if (!frame)
+ return -1;
- frame->local = (void *) local;
+ frame->local = (void *)local;
- marker_start_setxattr (frame, this);
+ marker_start_setxattr(frame, this);
- return 0;
+ return 0;
}
int32_t
-marker_xtime_update_marks (xlator_t *this, marker_local_t *local)
+marker_xtime_update_marks(xlator_t *this, marker_local_t *local)
{
- marker_conf_t *priv = NULL;
+ marker_conf_t *priv = NULL;
- GF_VALIDATE_OR_GOTO ("marker", this, out);
- GF_VALIDATE_OR_GOTO (this->name, local, out);
+ GF_VALIDATE_OR_GOTO("marker", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, local, out);
- priv = this->private;
+ priv = this->private;
- if ((local->pid == GF_CLIENT_PID_GSYNCD
- && !(priv->feature_enabled & GF_XTIME_GSYNC_FORCE))
- || (local->pid == GF_CLIENT_PID_DEFRAG))
- goto out;
+ if ((local->pid == GF_CLIENT_PID_GSYNCD &&
+ !(priv->feature_enabled & GF_XTIME_GSYNC_FORCE)) ||
+ (local->pid == GF_CLIENT_PID_DEFRAG))
+ goto out;
- marker_gettimeofday (local);
+ marker_gettimeofday(local);
- marker_local_ref (local);
+ marker_local_ref(local);
- marker_create_frame (this, local);
+ marker_create_frame(this, local);
out:
- return 0;
+ return 0;
}
-
int32_t
-marker_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+marker_mkdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- marker_conf_t *priv = NULL;
- marker_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
+ marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE, "error occurred "
- "while creating directory %s", strerror (op_errno));
- }
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "error occurred "
+ "while creating directory %s",
+ strerror(op_errno));
+ }
- local = (marker_local_t *) frame->local;
+ local = (marker_local_t *)frame->local;
- frame->local = NULL;
- priv = this->private;
-
- if (op_ret >= 0 && inode && (priv->feature_enabled & GF_QUOTA)) {
- ctx = mq_inode_ctx_new (inode, this);
- if (ctx == NULL) {
- gf_log (this->name, GF_LOG_WARNING, "mq_inode_ctx_new "
- "failed for %s", uuid_utoa (inode->gfid));
- op_ret = -1;
- op_errno = ENOMEM;
- }
+ frame->local = NULL;
+ priv = this->private;
+
+ if (op_ret >= 0 && inode && (priv->feature_enabled & GF_QUOTA)) {
+ ctx = mq_inode_ctx_new(inode, this);
+ if (ctx == NULL) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "mq_inode_ctx_new "
+ "failed for %s",
+ uuid_utoa(inode->gfid));
+ op_ret = -1;
+ op_errno = ENOMEM;
}
+ }
- STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno, inode,
- buf, preparent, postparent, xdata);
+ STACK_UNWIND_STRICT(mkdir, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
- if (op_ret == -1 || local == NULL)
- goto out;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- if (gf_uuid_is_null (local->loc.gfid))
- gf_uuid_copy (local->loc.gfid, buf->ia_gfid);
+ if (gf_uuid_is_null(local->loc.gfid))
+ gf_uuid_copy(local->loc.gfid, buf->ia_gfid);
- if (priv->feature_enabled & GF_QUOTA)
- mq_create_xattrs_txn (this, &local->loc, NULL);
+ if (priv->feature_enabled & GF_QUOTA)
+ mq_create_xattrs_txn(this, &local->loc, NULL);
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
out:
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int
-marker_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- mode_t umask, dict_t *xdata)
+marker_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = loc_copy (&local->loc, loc);
+ ret = loc_copy(&local->loc, loc);
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
wind:
- STACK_WIND (frame, marker_mkdir_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
+ STACK_WIND(frame, marker_mkdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
- return 0;
+ return 0;
err:
- MARKER_STACK_UNWIND (mkdir, frame, -1, ENOMEM, NULL,
- NULL, NULL, NULL, NULL);
+ MARKER_STACK_UNWIND(mkdir, frame, -1, ENOMEM, NULL, NULL, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
int32_t
-marker_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+marker_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
- quota_inode_ctx_t *ctx = NULL;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
+ quota_inode_ctx_t *ctx = NULL;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE, "error occurred "
- "while creating file %s", strerror (op_errno));
- }
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "error occurred "
+ "while creating file %s",
+ strerror(op_errno));
+ }
- local = (marker_local_t *) frame->local;
+ local = (marker_local_t *)frame->local;
- frame->local = NULL;
- priv = this->private;
-
- if (op_ret >= 0 && inode && (priv->feature_enabled & GF_QUOTA)) {
- ctx = mq_inode_ctx_new (inode, this);
- if (ctx == NULL) {
- gf_log (this->name, GF_LOG_WARNING, "mq_inode_ctx_new "
- "failed for %s", uuid_utoa (inode->gfid));
- op_ret = -1;
- op_errno = ENOMEM;
- }
+ frame->local = NULL;
+ priv = this->private;
+
+ if (op_ret >= 0 && inode && (priv->feature_enabled & GF_QUOTA)) {
+ ctx = mq_inode_ctx_new(inode, this);
+ if (ctx == NULL) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "mq_inode_ctx_new "
+ "failed for %s",
+ uuid_utoa(inode->gfid));
+ op_ret = -1;
+ op_errno = ENOMEM;
}
+ }
- STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
- preparent, postparent, xdata);
+ STACK_UNWIND_STRICT(create, frame, op_ret, op_errno, fd, inode, buf,
+ preparent, postparent, xdata);
- if (op_ret == -1 || local == NULL)
- goto out;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- if (gf_uuid_is_null (local->loc.gfid))
- gf_uuid_copy (local->loc.gfid, buf->ia_gfid);
+ if (gf_uuid_is_null(local->loc.gfid))
+ gf_uuid_copy(local->loc.gfid, buf->ia_gfid);
- if (priv->feature_enabled & GF_QUOTA)
- mq_create_xattrs_txn (this, &local->loc, buf);
+ if (priv->feature_enabled & GF_QUOTA)
+ mq_create_xattrs_txn(this, &local->loc, buf);
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
out:
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int32_t
-marker_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+marker_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = loc_copy (&local->loc, loc);
+ ret = loc_copy(&local->loc, loc);
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
wind:
- STACK_WIND (frame, marker_create_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->create, loc, flags, mode, umask,
- fd, xdata);
- return 0;
+ STACK_WIND(frame, marker_create_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, flags, mode, umask, fd,
+ xdata);
+ return 0;
err:
- MARKER_STACK_UNWIND (create, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
- NULL, NULL);
+ MARKER_STACK_UNWIND(create, frame, -1, ENOMEM, NULL, NULL, NULL, NULL, NULL,
+ NULL);
- return 0;
+ return 0;
}
-
int32_t
-marker_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+marker_writev_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- marker_conf_t *priv = NULL;
- marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE, "error occurred "
- "while write, %s", strerror (op_errno));
- }
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "error occurred "
+ "while write, %s",
+ strerror(op_errno));
+ }
- local = (marker_local_t *) frame->local;
+ local = (marker_local_t *)frame->local;
- frame->local = NULL;
+ frame->local = NULL;
- STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
+ STACK_UNWIND_STRICT(writev, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
- if (op_ret == -1 || local == NULL)
- goto out;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled & GF_QUOTA)
- mq_initiate_quota_txn (this, &local->loc, postbuf);
+ if (priv->feature_enabled & GF_QUOTA)
+ mq_initiate_quota_txn(this, &local->loc, postbuf);
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
out:
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int32_t
-marker_writev (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- struct iovec *vector,
- int32_t count,
- off_t offset, uint32_t flags,
- struct iobref *iobref, dict_t *xdata)
+marker_writev(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int32_t count, off_t offset, uint32_t flags,
+ struct iobref *iobref, dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = marker_inode_loc_fill (fd->inode, &local->loc);
+ ret = marker_inode_loc_fill(fd->inode, &local->loc);
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
wind:
- STACK_WIND (frame, marker_writev_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->writev, fd, vector, count, offset,
- flags, iobref, xdata);
- return 0;
+ STACK_WIND(frame, marker_writev_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector, count, offset,
+ flags, iobref, xdata);
+ return 0;
err:
- MARKER_STACK_UNWIND (writev, frame, -1, ENOMEM, NULL, NULL, NULL);
+ MARKER_STACK_UNWIND(writev, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
int32_t
-marker_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+marker_rmdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- marker_conf_t *priv = NULL;
- marker_local_t *local = NULL;
- call_stub_t *stub = NULL;
+ marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
+ call_stub_t *stub = NULL;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE, "error occurred "
- "rmdir %s", strerror (op_errno));
- }
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "error occurred "
+ "rmdir %s",
+ strerror(op_errno));
+ }
- local = (marker_local_t *) frame->local;
+ local = (marker_local_t *)frame->local;
- frame->local = NULL;
- priv = this->private;
+ frame->local = NULL;
+ priv = this->private;
- if (op_ret == -1 || local == NULL)
- goto out;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
- if (priv->feature_enabled & GF_QUOTA) {
- /* If a 'rm -rf' is performed by a client, rmdir can be faster
- than marker background mq_reduce_parent_size_txn.
- In this case, as part of rmdir parent child association
- will be removed in the server protocol.
- This can lead to mq_reduce_parent_size_txn failures.
-
- So perform mq_reduce_parent_size_txn in foreground
- and unwind to server once txn is complete
- */
-
- stub = fop_rmdir_cbk_stub (frame, default_rmdir_cbk, op_ret,
- op_errno, preparent, postparent,
- xdata);
- mq_reduce_parent_size_txn (this, &local->loc, NULL, 1, stub);
-
- if (stub) {
- marker_local_unref (local);
- return 0;
- }
+ if (priv->feature_enabled & GF_QUOTA) {
+ /* If a 'rm -rf' is performed by a client, rmdir can be faster
+ than marker background mq_reduce_parent_size_txn.
+ In this case, as part of rmdir parent child association
+ will be removed in the server protocol.
+ This can lead to mq_reduce_parent_size_txn failures.
+
+ So perform mq_reduce_parent_size_txn in foreground
+ and unwind to server once txn is complete
+ */
+
+ stub = fop_rmdir_cbk_stub(frame, default_rmdir_cbk, op_ret, op_errno,
+ preparent, postparent, xdata);
+ mq_reduce_parent_size_txn(this, &local->loc, NULL, 1, stub);
+
+ if (stub) {
+ marker_local_unref(local);
+ return 0;
}
+ }
out:
- STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno, preparent,
- postparent, xdata);
+ STACK_UNWIND_STRICT(rmdir, frame, op_ret, op_errno, preparent, postparent,
+ xdata);
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int32_t
-marker_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
- dict_t *xdata)
+marker_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = loc_copy (&local->loc, loc);
+ ret = loc_copy(&local->loc, loc);
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
wind:
- STACK_WIND (frame, marker_rmdir_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rmdir, loc, flags, xdata);
- return 0;
+ STACK_WIND(frame, marker_rmdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rmdir, loc, flags, xdata);
+ return 0;
err:
- MARKER_STACK_UNWIND (rmdir, frame, -1, ENOMEM, NULL, NULL, NULL);
+ MARKER_STACK_UNWIND(rmdir, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
int32_t
-marker_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+marker_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- marker_conf_t *priv = NULL;
- marker_local_t *local = NULL;
- uint32_t nlink = -1;
- GF_UNUSED int32_t ret = 0;
- call_stub_t *stub = NULL;
+ marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
+ uint32_t nlink = -1;
+ GF_UNUSED int32_t ret = 0;
+ call_stub_t *stub = NULL;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE,
- "%s occurred in unlink", strerror (op_errno));
- }
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE, "%s occurred in unlink",
+ strerror(op_errno));
+ }
- local = (marker_local_t *) frame->local;
+ local = (marker_local_t *)frame->local;
- frame->local = NULL;
- priv = this->private;
+ frame->local = NULL;
+ priv = this->private;
- if (op_ret == -1 || local == NULL)
- goto out;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
- if (priv->feature_enabled & GF_QUOTA) {
- if (local->skip_txn)
- goto out;
-
- if (xdata) {
- ret = dict_get_uint32 (xdata,
- GF_RESPONSE_LINK_COUNT_XDATA, &nlink);
- if (ret) {
- gf_log (this->name, GF_LOG_TRACE,
- "dict get failed %s ",
- strerror (-ret));
- }
- }
-
- /* If a 'rm -rf' is performed by a client, unlink can be faster
- than marker background mq_reduce_parent_size_txn.
- In this case, as part of unlink parent child association
- will be removed in the server protocol.
- This can lead to mq_reduce_parent_size_txn failures.
-
- So perform mq_reduce_parent_size_txn in foreground
- and unwind to server once txn is complete
- */
-
- stub = fop_unlink_cbk_stub (frame, default_unlink_cbk, op_ret,
- op_errno, preparent, postparent,
- xdata);
- mq_reduce_parent_size_txn (this, &local->loc, NULL, nlink,
- stub);
-
- if (stub) {
- marker_local_unref (local);
- return 0;
- }
+ if (priv->feature_enabled & GF_QUOTA) {
+ if (local->skip_txn)
+ goto out;
+
+ if (xdata) {
+ ret = dict_get_uint32(xdata, GF_RESPONSE_LINK_COUNT_XDATA, &nlink);
+ if (ret) {
+ gf_log(this->name, GF_LOG_TRACE, "dict get failed %s ",
+ strerror(-ret));
+ }
}
+ /* If a 'rm -rf' is performed by a client, unlink can be faster
+ than marker background mq_reduce_parent_size_txn.
+ In this case, as part of unlink parent child association
+ will be removed in the server protocol.
+ This can lead to mq_reduce_parent_size_txn failures.
+
+ So perform mq_reduce_parent_size_txn in foreground
+ and unwind to server once txn is complete
+ */
+
+ stub = fop_unlink_cbk_stub(frame, default_unlink_cbk, op_ret, op_errno,
+ preparent, postparent, xdata);
+ mq_reduce_parent_size_txn(this, &local->loc, NULL, nlink, stub);
+
+ if (stub) {
+ marker_local_unref(local);
+ return 0;
+ }
+ }
+
out:
- STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno, preparent,
- postparent, xdata);
+ STACK_UNWIND_STRICT(unlink, frame, op_ret, op_errno, preparent, postparent,
+ xdata);
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
-
int32_t
-marker_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
- dict_t *xdata)
+marker_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
- gf_boolean_t dict_free = _gf_false;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
+ gf_boolean_t dict_free = _gf_false;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto unlink_wind;
+ if (priv->feature_enabled == 0)
+ goto unlink_wind;
- local = mem_get0 (this->local_pool);
- local->xflag = xflag;
- if (xdata)
- local->xdata = dict_ref (xdata);
- MARKER_INIT_LOCAL (frame, local);
+ local = mem_get0(this->local_pool);
+ local->xflag = xflag;
+ if (xdata)
+ local->xdata = dict_ref(xdata);
+ MARKER_INIT_LOCAL(frame, local);
- ret = loc_copy (&local->loc, loc);
+ ret = loc_copy(&local->loc, loc);
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
- if (xdata && dict_get (xdata, GLUSTERFS_MARKER_DONT_ACCOUNT_KEY)) {
- local->skip_txn = 1;
- goto unlink_wind;
- }
+ if (xdata && dict_get(xdata, GLUSTERFS_MARKER_DONT_ACCOUNT_KEY)) {
+ local->skip_txn = 1;
+ goto unlink_wind;
+ }
- if (xdata == NULL) {
- xdata = dict_new ();
- dict_free = _gf_true;
- }
+ if (xdata == NULL) {
+ xdata = dict_new();
+ dict_free = _gf_true;
+ }
- ret = dict_set_int32 (xdata, GF_REQUEST_LINK_COUNT_XDATA, 1);
- if (ret < 0)
- goto err;
+ ret = dict_set_int32(xdata, GF_REQUEST_LINK_COUNT_XDATA, 1);
+ if (ret < 0)
+ goto err;
unlink_wind:
- STACK_WIND (frame, marker_unlink_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
- goto out;
+ STACK_WIND(frame, marker_unlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
+ goto out;
err:
- MARKER_STACK_UNWIND (unlink, frame, -1, ENOMEM, NULL, NULL, NULL);
+ MARKER_STACK_UNWIND(unlink, frame, -1, ENOMEM, NULL, NULL, NULL);
out:
- if (dict_free)
- dict_unref (xdata);
- return 0;
+ if (dict_free)
+ dict_unref(xdata);
+ return 0;
}
-
int32_t
-marker_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+marker_link_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
- "linking a file ", strerror (op_errno));
- }
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- local = (marker_local_t *) frame->local;
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s occurred while "
+ "linking a file ",
+ strerror(op_errno));
+ }
- frame->local = NULL;
+ local = (marker_local_t *)frame->local;
- STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
+ frame->local = NULL;
- if (op_ret == -1 || local == NULL)
- goto out;
+ STACK_UNWIND_STRICT(link, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
- priv = this->private;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- if (priv->feature_enabled & GF_QUOTA) {
- if (!local->skip_txn)
- mq_create_xattrs_txn (this, &local->loc, buf);
- }
+ priv = this->private;
+ if (priv->feature_enabled & GF_QUOTA) {
+ if (!local->skip_txn)
+ mq_create_xattrs_txn(this, &local->loc, buf);
+ }
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
out:
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int32_t
-marker_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
+marker_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = loc_copy (&local->loc, newloc);
+ ret = loc_copy(&local->loc, newloc);
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
- if (xdata && dict_get (xdata, GLUSTERFS_MARKER_DONT_ACCOUNT_KEY))
- local->skip_txn = 1;
+ if (xdata && dict_get(xdata, GLUSTERFS_MARKER_DONT_ACCOUNT_KEY))
+ local->skip_txn = 1;
wind:
- STACK_WIND (frame, marker_link_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
- return 0;
+ STACK_WIND(frame, marker_link_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
+ return 0;
err:
- MARKER_STACK_UNWIND (link, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
- NULL);
+ MARKER_STACK_UNWIND(link, frame, -1, ENOMEM, NULL, NULL, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
int32_t
-marker_rename_done (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+marker_rename_done(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- marker_local_t *local = NULL, *oplocal = NULL;
- loc_t newloc = {0, };
- marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL, *oplocal = NULL;
+ loc_t newloc = {
+ 0,
+ };
+ marker_conf_t *priv = NULL;
- local = frame->local;
- oplocal = local->oplocal;
+ local = frame->local;
+ oplocal = local->oplocal;
- priv = this->private;
+ priv = this->private;
- frame->local = NULL;
+ frame->local = NULL;
- if (op_ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "inodelk (UNLOCK) failed on path:%s (gfid:%s) (%s)",
- oplocal->parent_loc.path,
- uuid_utoa (oplocal->parent_loc.inode->gfid),
- strerror (op_errno));
- }
+ if (op_ret < 0) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "inodelk (UNLOCK) failed on path:%s (gfid:%s) (%s)",
+ oplocal->parent_loc.path,
+ uuid_utoa(oplocal->parent_loc.inode->gfid), strerror(op_errno));
+ }
- if (local->err != 0)
- goto err;
-
- mq_reduce_parent_size_txn (this, &oplocal->loc, &oplocal->contribution,
- -1, NULL);
+ if (local->err != 0)
+ goto err;
- if (local->loc.inode != NULL) {
- /* If destination file exits before rename, it would have
- * been unlinked while renaming a file
- */
- mq_reduce_parent_size_txn (this, &local->loc, NULL,
- local->ia_nlink, NULL);
- }
-
- newloc.inode = inode_ref (oplocal->loc.inode);
- newloc.path = gf_strdup (local->loc.path);
- newloc.name = strrchr (newloc.path, '/');
- if (newloc.name)
- newloc.name++;
- newloc.parent = inode_ref (local->loc.parent);
+ mq_reduce_parent_size_txn(this, &oplocal->loc, &oplocal->contribution, -1,
+ NULL);
- mq_create_xattrs_txn (this, &newloc, &local->buf);
-
- loc_wipe (&newloc);
-
- if (priv->feature_enabled & GF_XTIME) {
- if (!local->loc.inode)
- local->loc.inode = inode_ref (oplocal->loc.inode);
- //update marks on oldpath
- gf_uuid_copy (local->loc.gfid, oplocal->loc.inode->gfid);
- marker_xtime_update_marks (this, oplocal);
- marker_xtime_update_marks (this, local);
- }
+ if (local->loc.inode != NULL) {
+ /* If destination file exits before rename, it would have
+ * been unlinked while renaming a file
+ */
+ mq_reduce_parent_size_txn(this, &local->loc, NULL, local->ia_nlink,
+ NULL);
+ }
+
+ newloc.inode = inode_ref(oplocal->loc.inode);
+ newloc.path = gf_strdup(local->loc.path);
+ newloc.name = strrchr(newloc.path, '/');
+ if (newloc.name)
+ newloc.name++;
+ newloc.parent = inode_ref(local->loc.parent);
+
+ mq_create_xattrs_txn(this, &newloc, &local->buf);
+
+ loc_wipe(&newloc);
+
+ if (priv->feature_enabled & GF_XTIME) {
+ if (!local->loc.inode)
+ local->loc.inode = inode_ref(oplocal->loc.inode);
+ // update marks on oldpath
+ gf_uuid_copy(local->loc.gfid, oplocal->loc.inode->gfid);
+ marker_xtime_update_marks(this, oplocal);
+ marker_xtime_update_marks(this, local);
+ }
err:
- marker_local_unref (local);
- marker_local_unref (oplocal);
+ marker_local_unref(local);
+ marker_local_unref(oplocal);
- return 0;
+ return 0;
}
-
void
-marker_rename_release_oldp_lock (marker_local_t *local, xlator_t *this)
+marker_rename_release_oldp_lock(marker_local_t *local, xlator_t *this)
{
- marker_local_t *oplocal = NULL;
- call_frame_t *lk_frame = NULL;
- struct gf_flock lock = {0, };
+ marker_local_t *oplocal = NULL;
+ call_frame_t *lk_frame = NULL;
+ struct gf_flock lock = {
+ 0,
+ };
- oplocal = local->oplocal;
- lk_frame = local->lk_frame;
+ oplocal = local->oplocal;
+ lk_frame = local->lk_frame;
- if (lk_frame == NULL)
- goto err;
+ if (lk_frame == NULL)
+ goto err;
- lock.l_type = F_UNLCK;
- lock.l_whence = SEEK_SET;
- lock.l_start = 0;
- lock.l_len = 0;
- lock.l_pid = 0;
+ lock.l_type = F_UNLCK;
+ lock.l_whence = SEEK_SET;
+ lock.l_start = 0;
+ lock.l_len = 0;
+ lock.l_pid = 0;
- STACK_WIND (lk_frame,
- marker_rename_done,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->inodelk,
- this->name, &oplocal->parent_loc, F_SETLKW, &lock, NULL);
+ STACK_WIND(lk_frame, marker_rename_done, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->inodelk, this->name,
+ &oplocal->parent_loc, F_SETLKW, &lock, NULL);
- return;
+ return;
err:
- marker_local_unref (local);
- marker_local_unref (oplocal);
+ marker_local_unref(local);
+ marker_local_unref(oplocal);
}
-
int32_t
-marker_rename_unwind (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+marker_rename_unwind(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- marker_local_t *local = NULL;
- marker_local_t *oplocal = NULL;
- quota_inode_ctx_t *ctx = NULL;
- inode_contribution_t *contri = NULL;
+ marker_local_t *local = NULL;
+ marker_local_t *oplocal = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ inode_contribution_t *contri = NULL;
- local = frame->local;
- oplocal = local->oplocal;
- frame->local = NULL;
+ local = frame->local;
+ oplocal = local->oplocal;
+ frame->local = NULL;
- //Reset frame uid and gid if set.
- if (cookie == (void *) _GF_UID_GID_CHANGED)
- MARKER_RESET_UID_GID (frame, frame->root, local);
-
- if (op_ret < 0)
- local->err = op_errno ? op_errno : EINVAL;
-
- if (local->stub != NULL) {
- /* Remove contribution node from in-memory even if
- * remove-xattr has failed as the rename is already performed
- * if local->stub is set, which means rename was sucessful
- */
- (void) mq_inode_ctx_get (oplocal->loc.inode, this, &ctx);
- if (ctx) {
- contri = mq_get_contribution_node (oplocal->loc.parent,
- ctx);
- if (contri) {
- QUOTA_FREE_CONTRIBUTION_NODE (ctx, contri);
- GF_REF_PUT (contri);
- }
- }
-
- call_resume (local->stub);
- local->stub = NULL;
- local->err = 0;
- } else if (local->err != 0) {
- STACK_UNWIND_STRICT (rename, frame, -1, local->err, NULL, NULL,
- NULL, NULL, NULL, NULL);
- } else {
- gf_log (this->name, GF_LOG_CRITICAL,
- "continuation stub to unwind the call is absent, hence "
- "call will be hung (call-stack id = %"PRIu64")",
- frame->root->unique);
- }
+ // Reset frame uid and gid if set.
+ if (cookie == (void *)_GF_UID_GID_CHANGED)
+ MARKER_RESET_UID_GID(frame, frame->root, local);
- /* If there are in-progress writes on old-path when during rename
- * operation, update txn will update the wrong path if lock
- * is released before rename unwind.
- * So release lock only after rename unwind
- */
- marker_rename_release_oldp_lock (local, this);
+ if (op_ret < 0)
+ local->err = op_errno ? op_errno : EINVAL;
- return 0;
+ if (local->stub != NULL) {
+ /* Remove contribution node from in-memory even if
+ * remove-xattr has failed as the rename is already performed
+ * if local->stub is set, which means rename was successful
+ */
+ (void)mq_inode_ctx_get(oplocal->loc.inode, this, &ctx);
+ if (ctx) {
+ contri = mq_get_contribution_node(oplocal->loc.parent, ctx);
+ if (contri) {
+ QUOTA_FREE_CONTRIBUTION_NODE(ctx, contri);
+ GF_REF_PUT(contri);
+ }
+ }
+
+ call_resume(local->stub);
+ local->stub = NULL;
+ local->err = 0;
+ } else if (local->err != 0) {
+ STACK_UNWIND_STRICT(rename, frame, -1, local->err, NULL, NULL, NULL,
+ NULL, NULL, NULL);
+ } else {
+ gf_log(this->name, GF_LOG_CRITICAL,
+ "continuation stub to unwind the call is absent, hence "
+ "call will be hung (call-stack id = %" PRIu64 ")",
+ frame->root->unique);
+ }
+
+ /* If there are in-progress writes on old-path when during rename
+ * operation, update txn will update the wrong path if lock
+ * is released before rename unwind.
+ * So release lock only after rename unwind
+ */
+ marker_rename_release_oldp_lock(local, this);
+
+ return 0;
}
-
int32_t
-marker_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent,
- dict_t *xdata)
-{
- marker_conf_t *priv = NULL;
- marker_local_t *local = NULL;
- marker_local_t *oplocal = NULL;
- call_stub_t *stub = NULL;
- int32_t ret = 0;
- char contri_key[QUOTA_KEY_MAX] = {0, };
- loc_t newloc = {0, };
-
- local = (marker_local_t *) frame->local;
+marker_rename_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ struct iatt *preoldparent, struct iatt *postoldparent,
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
+{
+ marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
+ marker_local_t *oplocal = NULL;
+ call_stub_t *stub = NULL;
+ int32_t ret = 0;
+ char contri_key[QUOTA_KEY_MAX] = {
+ 0,
+ };
+ loc_t newloc = {
+ 0,
+ };
+
+ local = (marker_local_t *)frame->local;
+
+ if (local != NULL) {
+ oplocal = local->oplocal;
+ }
+ priv = this->private;
+
+ if (op_ret < 0) {
if (local != NULL) {
- oplocal = local->oplocal;
+ local->err = op_errno;
}
- priv = this->private;
-
- if (op_ret < 0) {
- if (local != NULL) {
- local->err = op_errno;
- }
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s occurred while "
+ "renaming a file ",
+ strerror(op_errno));
+ }
- gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
- "renaming a file ", strerror (op_errno));
+ if (priv->feature_enabled & GF_QUOTA) {
+ if ((op_ret < 0) || (local == NULL)) {
+ goto quota_err;
}
- if (priv->feature_enabled & GF_QUOTA) {
- if ((op_ret < 0) || (local == NULL)) {
- goto quota_err;
- }
-
- local->ia_nlink = 0;
- if (xdata)
- ret = dict_get_uint32 (xdata,
- GF_RESPONSE_LINK_COUNT_XDATA,
- &local->ia_nlink);
-
- local->buf = *buf;
- stub = fop_rename_cbk_stub (frame, default_rename_cbk, op_ret,
- op_errno, buf, preoldparent,
- postoldparent, prenewparent,
- postnewparent, xdata);
- if (stub == NULL) {
- local->err = ENOMEM;
- goto quota_err;
- }
-
- local->stub = stub;
-
- GET_CONTRI_KEY (this, contri_key, oplocal->loc.parent->gfid,
- ret);
- if (ret < 0) {
- local->err = ENOMEM;
- goto quota_err;
- }
-
- /* Removexattr requires uid and gid to be 0,
- * reset them in the callback.
- */
- MARKER_SET_UID_GID (frame, local, frame->root);
-
- newloc.inode = inode_ref (oplocal->loc.inode);
- newloc.path = gf_strdup (local->loc.path);
- newloc.name = strrchr (newloc.path, '/');
- if (newloc.name)
- newloc.name++;
- newloc.parent = inode_ref (local->loc.parent);
- gf_uuid_copy (newloc.gfid, oplocal->loc.inode->gfid);
-
- STACK_WIND_COOKIE (frame, marker_rename_unwind,
- frame->cookie, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->removexattr,
- &newloc, contri_key, NULL);
-
- loc_wipe (&newloc);
- } else {
- frame->local = NULL;
-
- STACK_UNWIND_STRICT (rename, frame, op_ret, op_errno, buf,
- preoldparent, postoldparent, prenewparent,
- postnewparent, xdata);
-
- if ((op_ret < 0) || (local == NULL)) {
- goto out;
- }
-
- if (priv->feature_enabled & GF_XTIME) {
- //update marks on oldpath
- if (!local->loc.inode)
- local->loc.inode = inode_ref (oplocal->loc.inode);
- gf_uuid_copy (local->loc.gfid, oplocal->loc.inode->gfid);
- marker_xtime_update_marks (this, oplocal);
- marker_xtime_update_marks (this, local);
- }
+ local->ia_nlink = 0;
+ if (xdata)
+ ret = dict_get_uint32(xdata, GF_RESPONSE_LINK_COUNT_XDATA,
+ &local->ia_nlink);
+
+ local->buf = *buf;
+ stub = fop_rename_cbk_stub(frame, default_rename_cbk, op_ret, op_errno,
+ buf, preoldparent, postoldparent,
+ prenewparent, postnewparent, xdata);
+ if (stub == NULL) {
+ local->err = ENOMEM;
+ goto quota_err;
}
-out:
- if (!(priv->feature_enabled & GF_QUOTA)) {
- marker_local_unref (local);
- marker_local_unref (oplocal);
+ local->stub = stub;
+
+ GET_CONTRI_KEY(this, contri_key, oplocal->loc.parent->gfid, ret);
+ if (ret < 0) {
+ local->err = ENOMEM;
+ goto quota_err;
}
- return 0;
+ /* Removexattr requires uid and gid to be 0,
+ * reset them in the callback.
+ */
+ MARKER_SET_UID_GID(frame, local, frame->root);
-quota_err:
- marker_rename_unwind (frame, NULL, this, 0, 0, NULL);
- return 0;
-}
+ newloc.inode = inode_ref(oplocal->loc.inode);
+ newloc.path = gf_strdup(local->loc.path);
+ newloc.name = strrchr(newloc.path, '/');
+ if (newloc.name)
+ newloc.name++;
+ newloc.parent = inode_ref(local->loc.parent);
+ gf_uuid_copy(newloc.gfid, oplocal->loc.inode->gfid);
+ STACK_WIND_COOKIE(
+ frame, marker_rename_unwind, frame->cookie, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, &newloc, contri_key, NULL);
-int32_t
-marker_do_rename (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
-{
- marker_local_t *local = NULL;
- marker_local_t *oplocal = NULL;
- char contri_key[QUOTA_KEY_MAX] = {0, };
- int32_t ret = 0;
- quota_meta_t contribution = {0, };
+ loc_wipe(&newloc);
+ } else {
+ frame->local = NULL;
- local = frame->local;
- oplocal = local->oplocal;
+ STACK_UNWIND_STRICT(rename, frame, op_ret, op_errno, buf, preoldparent,
+ postoldparent, prenewparent, postnewparent, xdata);
- //Reset frame uid and gid if set.
- if (cookie == (void *) _GF_UID_GID_CHANGED)
- MARKER_RESET_UID_GID (frame, frame->root, local);
-
- if ((op_ret < 0) && (op_errno != ENOATTR) && (op_errno != ENODATA)) {
- local->err = op_errno ? op_errno : EINVAL;
- gf_log (this->name, GF_LOG_WARNING,
- "fetching contribution values from %s (gfid:%s) "
- "failed (%s)", oplocal->loc.path,
- uuid_utoa (oplocal->loc.inode->gfid),
- strerror (op_errno));
- goto err;
+ if ((op_ret < 0) || (local == NULL)) {
+ goto out;
}
- GET_CONTRI_KEY (this, contri_key, oplocal->loc.parent->gfid, ret);
- if (ret < 0) {
- local->err = errno ? errno : ENOMEM;
- goto err;
+ if (priv->feature_enabled & GF_XTIME) {
+ // update marks on oldpath
+ if (!local->loc.inode)
+ local->loc.inode = inode_ref(oplocal->loc.inode);
+ gf_uuid_copy(local->loc.gfid, oplocal->loc.inode->gfid);
+ marker_xtime_update_marks(this, oplocal);
+ marker_xtime_update_marks(this, local);
}
- quota_dict_get_meta (dict, contri_key, &contribution);
- oplocal->contribution = contribution;
+ }
- STACK_WIND (frame, marker_rename_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rename, &oplocal->loc,
- &local->loc, local->xdata);
+out:
+ if (!(priv->feature_enabled & GF_QUOTA)) {
+ marker_local_unref(local);
+ marker_local_unref(oplocal);
+ }
- return 0;
+ return 0;
-err:
- marker_rename_unwind (frame, NULL, this, 0, 0, NULL);
- return 0;
+quota_err:
+ marker_rename_unwind(frame, NULL, this, 0, 0, NULL);
+ return 0;
}
int32_t
-marker_get_oldpath_contribution (call_frame_t *lk_frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, dict_t *xdata)
-{
- call_frame_t *frame = NULL;
- marker_local_t *local = NULL;
- marker_local_t *oplocal = NULL;
- char contri_key[QUOTA_KEY_MAX] = {0, };
- int32_t ret = 0;
+marker_do_rename(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+ marker_local_t *local = NULL;
+ marker_local_t *oplocal = NULL;
+ char contri_key[QUOTA_KEY_MAX] = {
+ 0,
+ };
+ int keylen = 0;
+ quota_meta_t contribution = {
+ 0,
+ };
+
+ local = frame->local;
+ oplocal = local->oplocal;
+
+ // Reset frame uid and gid if set.
+ if (cookie == (void *)_GF_UID_GID_CHANGED)
+ MARKER_RESET_UID_GID(frame, frame->root, local);
+
+ if ((op_ret < 0) && (op_errno != ENOATTR) && (op_errno != ENODATA)) {
+ local->err = op_errno ? op_errno : EINVAL;
+ gf_log(this->name, GF_LOG_WARNING,
+ "fetching contribution values from %s (gfid:%s) "
+ "failed (%s)",
+ oplocal->loc.path, uuid_utoa(oplocal->loc.inode->gfid),
+ strerror(op_errno));
+ goto err;
+ }
+
+ GET_CONTRI_KEY(this, contri_key, oplocal->loc.parent->gfid, keylen);
+ if (keylen < 0) {
+ local->err = errno ? errno : ENOMEM;
+ goto err;
+ }
+ quota_dict_get_meta(dict, contri_key, keylen, &contribution);
+ oplocal->contribution = contribution;
+
+ STACK_WIND(frame, marker_rename_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, &oplocal->loc, &local->loc,
+ local->xdata);
+
+ return 0;
- local = lk_frame->local;
- oplocal = local->oplocal;
- frame = local->frame;
-
- if (op_ret < 0) {
- local->err = op_errno ? op_errno : EINVAL;
- gf_log (this->name, GF_LOG_WARNING,
- "cannot hold inodelk on %s (gfid:%s) (%s)",
- oplocal->loc.path, uuid_utoa (oplocal->loc.inode->gfid),
- strerror (op_errno));
- goto err;
+err:
+ marker_rename_unwind(frame, NULL, this, 0, 0, NULL);
+ return 0;
+}
- STACK_DESTROY (local->lk_frame->root);
- local->lk_frame = NULL;
+int32_t
+marker_get_oldpath_contribution(call_frame_t *lk_frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *xdata)
+{
+ call_frame_t *frame = NULL;
+ marker_local_t *local = NULL;
+ marker_local_t *oplocal = NULL;
+ char contri_key[QUOTA_KEY_MAX] = {
+ 0,
+ };
+ int32_t ret = 0;
+
+ local = lk_frame->local;
+ oplocal = local->oplocal;
+ frame = local->frame;
+
+ if (op_ret < 0) {
+ local->err = op_errno ? op_errno : EINVAL;
+ gf_log(this->name, GF_LOG_WARNING,
+ "cannot hold inodelk on %s (gfid:%s) (%s)", oplocal->loc.path,
+ uuid_utoa(oplocal->loc.inode->gfid), strerror(op_errno));
+ if (local->lk_frame) {
+ STACK_DESTROY(local->lk_frame->root);
+ local->lk_frame = NULL;
}
+ goto err;
+ }
- GET_CONTRI_KEY (this, contri_key, oplocal->loc.parent->gfid, ret);
- if (ret < 0) {
- local->err = errno ? errno : ENOMEM;
- goto err;
- }
+ GET_CONTRI_KEY(this, contri_key, oplocal->loc.parent->gfid, ret);
+ if (ret < 0) {
+ local->err = errno ? errno : ENOMEM;
+ goto err;
+ }
- /* getxattr requires uid and gid to be 0,
- * reset them in the callback.
- */
- MARKER_SET_UID_GID (frame, local, frame->root);
+ /* getxattr requires uid and gid to be 0,
+ * reset them in the callback.
+ */
+ MARKER_SET_UID_GID(frame, local, frame->root);
- if (gf_uuid_is_null (oplocal->loc.gfid))
- gf_uuid_copy (oplocal->loc.gfid,
- oplocal->loc.inode->gfid);
+ if (gf_uuid_is_null(oplocal->loc.gfid))
+ gf_uuid_copy(oplocal->loc.gfid, oplocal->loc.inode->gfid);
- GF_UUID_ASSERT (oplocal->loc.gfid);
+ GF_UUID_ASSERT(oplocal->loc.gfid);
- STACK_WIND_COOKIE (frame, marker_do_rename,
- frame->cookie, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr,
- &oplocal->loc, contri_key, NULL);
+ STACK_WIND_COOKIE(frame, marker_do_rename, frame->cookie, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->getxattr, &oplocal->loc,
+ contri_key, NULL);
- return 0;
+ return 0;
err:
- marker_rename_unwind (frame, NULL, this, 0, 0, NULL);
- return 0;
+ marker_rename_unwind(frame, NULL, this, 0, 0, NULL);
+ return 0;
}
-
/* For a marker_rename FOP, following is the algorithm used for Quota
* accounting. The use-case considered is:
* 1. rename (src, dst)
@@ -1641,7 +1629,7 @@ err:
* b) we should subtract from src-parent exactly what we contributed to
* src-parent
* So, We hold a lock on src-parent to block any parallel transcations on
- * src-inode (since thats the one which survives rename).
+ * src-inode (since that's the one which survives rename).
*
* If there are any parallel transactions on dst-inode they keep succeeding
* till the association of dst-inode with dst-parent is broken because of an
@@ -1703,628 +1691,631 @@ err:
* 10) create contribution xattr to dst-parent on src-inode.
*/
int32_t
-marker_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
- loc_t *newloc, dict_t *xdata)
+marker_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_local_t *oplocal = NULL;
- marker_conf_t *priv = NULL;
- struct gf_flock lock = {0, };
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_local_t *oplocal = NULL;
+ marker_conf_t *priv = NULL;
+ struct gf_flock lock = {
+ 0,
+ };
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto rename_wind;
+ if (priv->feature_enabled == 0)
+ goto rename_wind;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- oplocal = mem_get0 (this->local_pool);
+ oplocal = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, oplocal);
+ MARKER_INIT_LOCAL(frame, oplocal);
- frame->local = local;
+ frame->local = local;
- local->oplocal = marker_local_ref (oplocal);
+ local->oplocal = marker_local_ref(oplocal);
- ret = loc_copy (&local->loc, newloc);
- if (ret < 0)
- goto err;
+ ret = loc_copy(&local->loc, newloc);
+ if (ret < 0)
+ goto err;
- ret = loc_copy (&oplocal->loc, oldloc);
- if (ret < 0)
- goto err;
+ ret = loc_copy(&oplocal->loc, oldloc);
+ if (ret < 0)
+ goto err;
- if (!(priv->feature_enabled & GF_QUOTA)) {
- goto rename_wind;
- }
+ if (!(priv->feature_enabled & GF_QUOTA)) {
+ goto rename_wind;
+ }
- ret = mq_inode_loc_fill (NULL, newloc->parent, &local->parent_loc);
- if (ret < 0)
- goto err;
+ ret = mq_inode_loc_fill(NULL, newloc->parent, &local->parent_loc);
+ if (ret < 0)
+ goto err;
- ret = mq_inode_loc_fill (NULL, oldloc->parent, &oplocal->parent_loc);
- if (ret < 0)
- goto err;
+ ret = mq_inode_loc_fill(NULL, oldloc->parent, &oplocal->parent_loc);
+ if (ret < 0)
+ goto err;
- lock.l_len = 0;
- lock.l_start = 0;
- lock.l_type = F_WRLCK;
- lock.l_whence = SEEK_SET;
+ lock.l_len = 0;
+ lock.l_start = 0;
+ lock.l_type = F_WRLCK;
+ lock.l_whence = SEEK_SET;
- local->xdata = xdata ? dict_ref (xdata) : dict_new ();
- ret = dict_set_int32 (local->xdata, GF_REQUEST_LINK_COUNT_XDATA, 1);
- if (ret < 0)
- goto err;
+ local->xdata = xdata ? dict_ref(xdata) : dict_new();
+ ret = dict_set_int32(local->xdata, GF_REQUEST_LINK_COUNT_XDATA, 1);
+ if (ret < 0)
+ goto err;
- local->frame = frame;
- local->lk_frame = create_frame (this, this->ctx->pool);
- if (local->lk_frame == NULL)
- goto err;
+ local->frame = frame;
+ local->lk_frame = create_frame(this, this->ctx->pool);
+ if (local->lk_frame == NULL)
+ goto err;
- local->lk_frame->root->uid = 0;
- local->lk_frame->root->gid = 0;
- local->lk_frame->local = local;
- set_lk_owner_from_ptr (&local->lk_frame->root->lk_owner,
- local->lk_frame->root);
+ local->lk_frame->root->uid = 0;
+ local->lk_frame->root->gid = 0;
+ local->lk_frame->local = local;
+ set_lk_owner_from_ptr(&local->lk_frame->root->lk_owner,
+ local->lk_frame->root);
- STACK_WIND (local->lk_frame,
- marker_get_oldpath_contribution,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->inodelk,
- this->name, &oplocal->parent_loc,
- F_SETLKW, &lock, NULL);
+ STACK_WIND(local->lk_frame, marker_get_oldpath_contribution,
+ FIRST_CHILD(this), FIRST_CHILD(this)->fops->inodelk, this->name,
+ &oplocal->parent_loc, F_SETLKW, &lock, NULL);
- return 0;
+ return 0;
rename_wind:
- STACK_WIND (frame, marker_rename_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
+ STACK_WIND(frame, marker_rename_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
- return 0;
+ return 0;
err:
- MARKER_STACK_UNWIND (rename, frame, -1, ENOMEM, NULL,
- NULL, NULL, NULL, NULL, NULL);
- marker_local_unref (oplocal);
+ MARKER_STACK_UNWIND(rename, frame, -1, ENOMEM, NULL, NULL, NULL, NULL, NULL,
+ NULL);
+ marker_local_unref(oplocal);
- return 0;
+ return 0;
}
-
int32_t
-marker_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+marker_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
- "truncating a file ", strerror (op_errno));
- }
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- local = (marker_local_t *) frame->local;
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s occurred while "
+ "truncating a file ",
+ strerror(op_errno));
+ }
- frame->local = NULL;
+ local = (marker_local_t *)frame->local;
- STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
+ frame->local = NULL;
- if (op_ret == -1 || local == NULL)
- goto out;
+ STACK_UNWIND_STRICT(truncate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
- priv = this->private;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- if (priv->feature_enabled & GF_QUOTA) {
- /* DHT Rebalance process, at the end of migration will
- * first make the src file as a linkto file and then
- * truncate the file. By doing a truncate after making the
- * src file as linkto file, the contri which is already
- * accounted is left over.
- * So, we need to account for the linkto file when a truncate
- * happens, thereby updating the contri properly.
- * By passing NULL for postbuf, mq_prevalidate does not check
- * for linkto file.
- * Same happens with ftruncate as well.
- */
- if (postbuf && IS_DHT_LINKFILE_MODE (postbuf))
- mq_initiate_quota_txn (this, &local->loc, NULL);
- else
- mq_initiate_quota_txn (this, &local->loc, postbuf);
- }
+ priv = this->private;
+
+ if (priv->feature_enabled & GF_QUOTA) {
+ /* DHT Rebalance process, at the end of migration will
+ * first make the src file as a linkto file and then
+ * truncate the file. By doing a truncate after making the
+ * src file as linkto file, the contri which is already
+ * accounted is left over.
+ * So, we need to account for the linkto file when a truncate
+ * happens, thereby updating the contri properly.
+ * By passing NULL for postbuf, mq_prevalidate does not check
+ * for linkto file.
+ * Same happens with ftruncate as well.
+ */
+ if (postbuf && IS_DHT_LINKFILE_MODE(postbuf))
+ mq_initiate_quota_txn(this, &local->loc, NULL);
+ else
+ mq_initiate_quota_txn(this, &local->loc, postbuf);
+ }
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
out:
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int32_t
-marker_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
- dict_t *xdata)
+marker_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = loc_copy (&local->loc, loc);
+ ret = loc_copy(&local->loc, loc);
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
wind:
- STACK_WIND (frame, marker_truncate_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
- return 0;
+ STACK_WIND(frame, marker_truncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
+ return 0;
err:
- MARKER_STACK_UNWIND (truncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+ MARKER_STACK_UNWIND(truncate, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
int32_t
-marker_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+marker_ftruncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
- "truncating a file ", strerror (op_errno));
- }
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s occurred while "
+ "truncating a file ",
+ strerror(op_errno));
+ }
- local = (marker_local_t *) frame->local;
+ local = (marker_local_t *)frame->local;
- frame->local = NULL;
+ frame->local = NULL;
- STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
+ STACK_UNWIND_STRICT(ftruncate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
- if (op_ret == -1 || local == NULL)
- goto out;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled & GF_QUOTA) {
- if (postbuf && IS_DHT_LINKFILE_MODE (postbuf))
- mq_initiate_quota_txn (this, &local->loc, NULL);
- else
- mq_initiate_quota_txn (this, &local->loc, postbuf);
- }
+ if (priv->feature_enabled & GF_QUOTA) {
+ if (postbuf && IS_DHT_LINKFILE_MODE(postbuf))
+ mq_initiate_quota_txn(this, &local->loc, NULL);
+ else
+ mq_initiate_quota_txn(this, &local->loc, postbuf);
+ }
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
out:
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int32_t
-marker_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- dict_t *xdata)
+marker_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = marker_inode_loc_fill (fd->inode, &local->loc);
+ ret = marker_inode_loc_fill(fd->inode, &local->loc);
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
wind:
- STACK_WIND (frame, marker_ftruncate_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
- return 0;
+ STACK_WIND(frame, marker_ftruncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+ return 0;
err:
- MARKER_STACK_UNWIND (ftruncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+ MARKER_STACK_UNWIND(ftruncate, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
int32_t
-marker_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- marker_conf_t *priv = NULL;
- marker_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
- "creating symlinks ", strerror (op_errno));
- }
+marker_symlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
- local = (marker_local_t *) frame->local;
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s occurred while "
+ "creating symlinks ",
+ strerror(op_errno));
+ }
- frame->local = NULL;
- priv = this->private;
-
- if (op_ret >= 0 && inode && (priv->feature_enabled & GF_QUOTA)) {
- ctx = mq_inode_ctx_new (inode, this);
- if (ctx == NULL) {
- gf_log (this->name, GF_LOG_WARNING, "mq_inode_ctx_new "
- "failed for %s", uuid_utoa (inode->gfid));
- op_ret = -1;
- op_errno = ENOMEM;
- }
+ local = (marker_local_t *)frame->local;
+
+ frame->local = NULL;
+ priv = this->private;
+
+ if (op_ret >= 0 && inode && (priv->feature_enabled & GF_QUOTA)) {
+ ctx = mq_inode_ctx_new(inode, this);
+ if (ctx == NULL) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "mq_inode_ctx_new "
+ "failed for %s",
+ uuid_utoa(inode->gfid));
+ op_ret = -1;
+ op_errno = ENOMEM;
}
+ }
- STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
+ STACK_UNWIND_STRICT(symlink, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
- if (op_ret == -1 || local == NULL)
- goto out;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- if (gf_uuid_is_null (local->loc.gfid))
- gf_uuid_copy (local->loc.gfid, buf->ia_gfid);
+ if (gf_uuid_is_null(local->loc.gfid))
+ gf_uuid_copy(local->loc.gfid, buf->ia_gfid);
- if (priv->feature_enabled & GF_QUOTA) {
- mq_create_xattrs_txn (this, &local->loc, buf);
- }
+ if (priv->feature_enabled & GF_QUOTA) {
+ mq_create_xattrs_txn(this, &local->loc, buf);
+ }
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
out:
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int
-marker_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
- loc_t *loc, mode_t umask, dict_t *xdata)
+marker_symlink(call_frame_t *frame, xlator_t *this, const char *linkpath,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = loc_copy (&local->loc, loc);
+ ret = loc_copy(&local->loc, loc);
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
wind:
- STACK_WIND (frame, marker_symlink_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->symlink, linkpath, loc, umask,
- xdata);
- return 0;
+ STACK_WIND(frame, marker_symlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->symlink, linkpath, loc, umask, xdata);
+ return 0;
err:
- MARKER_STACK_UNWIND (symlink, frame, -1, ENOMEM, NULL,
- NULL, NULL, NULL, NULL);
+ MARKER_STACK_UNWIND(symlink, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
+ NULL);
- return 0;
+ return 0;
}
-
int32_t
-marker_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+marker_mknod_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
- quota_inode_ctx_t *ctx = NULL;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
+ quota_inode_ctx_t *ctx = NULL;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE, "%s occurred with "
- "mknod ", strerror (op_errno));
- }
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s occurred with "
+ "mknod ",
+ strerror(op_errno));
+ }
- local = (marker_local_t *) frame->local;
+ local = (marker_local_t *)frame->local;
- frame->local = NULL;
- priv = this->private;
-
- if (op_ret >= 0 && inode && (priv->feature_enabled & GF_QUOTA)) {
- ctx = mq_inode_ctx_new (inode, this);
- if (ctx == NULL) {
- gf_log (this->name, GF_LOG_WARNING, "mq_inode_ctx_new "
- "failed for %s", uuid_utoa (inode->gfid));
- op_ret = -1;
- op_errno = ENOMEM;
- }
+ frame->local = NULL;
+ priv = this->private;
+
+ if (op_ret >= 0 && inode && (priv->feature_enabled & GF_QUOTA)) {
+ ctx = mq_inode_ctx_new(inode, this);
+ if (ctx == NULL) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "mq_inode_ctx_new "
+ "failed for %s",
+ uuid_utoa(inode->gfid));
+ op_ret = -1;
+ op_errno = ENOMEM;
}
+ }
- STACK_UNWIND_STRICT (mknod, frame, op_ret, op_errno, inode,
- buf, preparent, postparent, xdata);
+ STACK_UNWIND_STRICT(mknod, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
- if (op_ret == -1 || local == NULL)
- goto out;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- if (gf_uuid_is_null (local->loc.gfid))
- gf_uuid_copy (local->loc.gfid, buf->ia_gfid);
+ if (gf_uuid_is_null(local->loc.gfid))
+ gf_uuid_copy(local->loc.gfid, buf->ia_gfid);
- if ((priv->feature_enabled & GF_QUOTA) && (S_ISREG (local->mode))) {
- mq_create_xattrs_txn (this, &local->loc, buf);
- }
+ if ((priv->feature_enabled & GF_QUOTA) && (S_ISREG(local->mode))) {
+ mq_create_xattrs_txn(this, &local->loc, buf);
+ }
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
out:
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int
-marker_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- dev_t rdev, mode_t umask, dict_t *xdata)
+marker_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = loc_copy (&local->loc, loc);
+ ret = loc_copy(&local->loc, loc);
- local->mode = mode;
+ local->mode = mode;
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
wind:
- STACK_WIND (frame, marker_mknod_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mknod, loc, mode, rdev, umask,
- xdata);
- return 0;
+ STACK_WIND(frame, marker_mknod_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, loc, mode, rdev, umask, xdata);
+ return 0;
err:
- MARKER_STACK_UNWIND (mknod, frame, -1, ENOMEM, NULL,
- NULL, NULL, NULL, NULL);
+ MARKER_STACK_UNWIND(mknod, frame, -1, ENOMEM, NULL, NULL, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
int32_t
-marker_fallocate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+marker_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
- "fallocating a file ", strerror (op_errno));
- }
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s occurred while "
+ "fallocating a file ",
+ strerror(op_errno));
+ }
- local = (marker_local_t *) frame->local;
+ local = (marker_local_t *)frame->local;
- frame->local = NULL;
+ frame->local = NULL;
- STACK_UNWIND_STRICT (fallocate, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
+ STACK_UNWIND_STRICT(fallocate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
- if (op_ret == -1 || local == NULL)
- goto out;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled & GF_QUOTA)
- mq_initiate_quota_txn (this, &local->loc, postbuf);
+ if (priv->feature_enabled & GF_QUOTA)
+ mq_initiate_quota_txn(this, &local->loc, postbuf);
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
out:
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int32_t
marker_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
- off_t offset, size_t len, dict_t *xdata)
+ off_t offset, size_t len, dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = marker_inode_loc_fill (fd->inode, &local->loc);
+ ret = marker_inode_loc_fill(fd->inode, &local->loc);
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
wind:
- STACK_WIND (frame, marker_fallocate_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fallocate, fd, mode, offset, len,
- xdata);
- return 0;
+ STACK_WIND(frame, marker_fallocate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fallocate, fd, mode, offset, len,
+ xdata);
+ return 0;
err:
- MARKER_STACK_UNWIND (fallocate, frame, -1, ENOMEM, NULL, NULL, NULL);
+ MARKER_STACK_UNWIND(fallocate, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
int32_t
marker_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
struct iatt *postbuf, dict_t *xdata)
{
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE, "%s occurred during discard",
- strerror (op_errno));
- }
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE, "%s occurred during discard",
+ strerror(op_errno));
+ }
- local = (marker_local_t *) frame->local;
+ local = (marker_local_t *)frame->local;
- frame->local = NULL;
+ frame->local = NULL;
- STACK_UNWIND_STRICT (discard, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
+ STACK_UNWIND_STRICT(discard, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
- if (op_ret == -1 || local == NULL)
- goto out;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled & GF_QUOTA)
- mq_initiate_quota_txn (this, &local->loc, postbuf);
+ if (priv->feature_enabled & GF_QUOTA)
+ mq_initiate_quota_txn(this, &local->loc, postbuf);
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
out:
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int32_t
marker_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- size_t len, dict_t *xdata)
+ size_t len, dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = marker_inode_loc_fill (fd->inode, &local->loc);
+ ret = marker_inode_loc_fill(fd->inode, &local->loc);
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
wind:
- STACK_WIND (frame, marker_discard_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->discard, fd, offset, len, xdata);
- return 0;
+ STACK_WIND(frame, marker_discard_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->discard, fd, offset, len, xdata);
+ return 0;
err:
- MARKER_STACK_UNWIND (discard, frame, -1, ENOMEM, NULL, NULL, NULL);
+ MARKER_STACK_UNWIND(discard, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ return 0;
}
int32_t
marker_zerofill_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE, "%s occurred during zerofill",
- strerror (op_errno));
- }
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE, "%s occurred during zerofill",
+ strerror(op_errno));
+ }
- local = (marker_local_t *) frame->local;
+ local = (marker_local_t *)frame->local;
- frame->local = NULL;
+ frame->local = NULL;
- STACK_UNWIND_STRICT (zerofill, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
+ STACK_UNWIND_STRICT(zerofill, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
- if (op_ret == -1 || local == NULL)
- goto out;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled & GF_QUOTA)
- mq_initiate_quota_txn (this, &local->loc, postbuf);
+ if (priv->feature_enabled & GF_QUOTA)
+ mq_initiate_quota_txn(this, &local->loc, postbuf);
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
out:
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int32_t
marker_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- off_t len, dict_t *xdata)
+ off_t len, dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = marker_inode_loc_fill (fd->inode, &local->loc);
+ ret = marker_inode_loc_fill(fd->inode, &local->loc);
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
wind:
- STACK_WIND (frame, marker_zerofill_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->zerofill, fd, offset, len, xdata);
- return 0;
+ STACK_WIND(frame, marker_zerofill_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->zerofill, fd, offset, len, xdata);
+ return 0;
err:
- MARKER_STACK_UNWIND (zerofill, frame, -1, ENOMEM, NULL, NULL, NULL);
+ MARKER_STACK_UNWIND(zerofill, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
/* when a call from the special client is received on
* key trusted.glusterfs.volume-mark with value "RESET"
* or if the value is 0length, update the change the
@@ -2332,1212 +2323,1246 @@ err:
* timestamp file.
*/
int32_t
-call_from_sp_client_to_reset_tmfile (call_frame_t *frame,
- xlator_t *this,
- dict_t *dict)
+call_from_sp_client_to_reset_tmfile(call_frame_t *frame, xlator_t *this,
+ dict_t *dict)
{
- int32_t fd = 0;
- int32_t op_ret = 0;
- int32_t op_errno = 0;
- data_t *data = NULL;
- marker_conf_t *priv = NULL;
+ int32_t fd = 0;
+ int32_t op_ret = 0;
+ int32_t op_errno = 0;
+ data_t *data = NULL;
+ marker_conf_t *priv = NULL;
+
+ if (frame == NULL || this == NULL || dict == NULL)
+ return -1;
- if (frame == NULL || this == NULL || dict == NULL)
- return -1;
+ priv = this->private;
- priv = this->private;
+ data = dict_get(dict, "trusted.glusterfs.volume-mark");
+ if (data == NULL)
+ return -1;
- data = dict_get (dict, "trusted.glusterfs.volume-mark");
- if (data == NULL)
- return -1;
+ if (frame->root->pid != GF_CLIENT_PID_GSYNCD) {
+ op_ret = -1;
+ op_errno = EPERM;
- if (frame->root->pid != GF_CLIENT_PID_GSYNCD) {
- op_ret = -1;
- op_errno = EPERM;
+ goto out;
+ }
- goto out;
+ if (data->len == 0 ||
+ (data->len == 5 && memcmp(data->data, "RESET", 5) == 0)) {
+ fd = open(priv->timestamp_file, O_WRONLY | O_TRUNC);
+ if (fd != -1) {
+ /* TODO check whether the O_TRUNC would update the
+ * timestamps on a zero length file on all machies.
+ */
+ sys_close(fd);
}
- if (data->len == 0 || (data->len == 5 &&
- memcmp (data->data, "RESET", 5) == 0)) {
- fd = open (priv->timestamp_file, O_WRONLY|O_TRUNC);
- if (fd != -1) {
- /* TODO check whether the O_TRUNC would update the
- * timestamps on a zero length file on all machies.
- */
- sys_close (fd);
- }
-
- if (fd != -1 || errno == ENOENT) {
- op_ret = 0;
- op_errno = 0;
- } else {
- op_ret = -1;
- op_errno = errno;
- }
+ if (fd != -1 || errno == ENOENT) {
+ op_ret = 0;
+ op_errno = 0;
} else {
- op_ret = -1;
- op_errno = EINVAL;
+ op_ret = -1;
+ op_errno = errno;
}
+ } else {
+ op_ret = -1;
+ op_errno = EINVAL;
+ }
out:
- STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, NULL);
+ STACK_UNWIND_STRICT(setxattr, frame, op_ret, op_errno, NULL);
- return 0;
+ return 0;
}
-
int32_t
-marker_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+marker_setxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE, "%s occurred in "
- "setxattr ", strerror (op_errno));
- }
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s occurred in "
+ "setxattr ",
+ strerror(op_errno));
+ }
- local = (marker_local_t *) frame->local;
+ local = (marker_local_t *)frame->local;
- frame->local = NULL;
+ frame->local = NULL;
- STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
+ STACK_UNWIND_STRICT(setxattr, frame, op_ret, op_errno, xdata);
- if (op_ret == -1 || local == NULL)
- goto out;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
out:
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int
-remove_quota_keys (dict_t *dict, char *k, data_t *v, void *data)
-{
- call_frame_t *frame = data;
- marker_local_t *local = frame->local;
- xlator_t *this = frame->this;
- marker_conf_t *priv = NULL;
- char ver_str[NAME_MAX] = {0,};
- char *dot = NULL;
- int ret = -1;
-
- priv = this->private;
-
- /* If quota is enabled immediately after disable.
- * quota healing starts creating new xattrs
- * before completing the cleanup operation.
- * So we should check if the xattr is the new.
- * Do not remove xattr if its xattr
- * version is same as current version
- */
- if ((priv->feature_enabled & GF_QUOTA) && priv->version > 0) {
- snprintf (ver_str, sizeof (ver_str), ".%d", priv->version);
- dot = strrchr (k, '.');
- if (dot && !strcmp(dot, ver_str))
- return 0;
- }
-
- ret = syncop_removexattr (FIRST_CHILD (this), &local->loc, k, 0, NULL);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "%s: Failed to remove "
- "extended attribute: %s", local->loc.path, k);
- return -1;
- }
- return 0;
+remove_quota_keys(dict_t *dict, char *k, data_t *v, void *data)
+{
+ call_frame_t *frame = data;
+ marker_local_t *local = frame->local;
+ xlator_t *this = frame->this;
+ marker_conf_t *priv = NULL;
+ char ver_str[NAME_MAX] = {
+ 0,
+ };
+ char *dot = NULL;
+ int ret = -1;
+
+ priv = this->private;
+
+ /* If quota is enabled immediately after disable.
+ * quota healing starts creating new xattrs
+ * before completing the cleanup operation.
+ * So we should check if the xattr is the new.
+ * Do not remove xattr if its xattr
+ * version is same as current version
+ */
+ if ((priv->feature_enabled & GF_QUOTA) && priv->version > 0) {
+ snprintf(ver_str, sizeof(ver_str), ".%d", priv->version);
+ dot = strrchr(k, '.');
+ if (dot && !strcmp(dot, ver_str))
+ return 0;
+ }
+
+ ret = syncop_removexattr(FIRST_CHILD(this), &local->loc, k, 0, NULL);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "%s: Failed to remove "
+ "extended attribute: %s",
+ local->loc.path, k);
+ return -1;
+ }
+ return 0;
}
int
-quota_xattr_cleaner_cbk (int ret, call_frame_t *frame, void *args)
+quota_xattr_cleaner_cbk(int ret, call_frame_t *frame, void *args)
{
- dict_t *xdata = args;
- int op_ret = -1;
- int op_errno = 0;
+ dict_t *xdata = args;
+ int op_ret = -1;
+ int op_errno = 0;
- op_ret = (ret < 0)? -1: 0;
- op_errno = -ret;
+ op_ret = (ret < 0) ? -1 : 0;
+ op_errno = -ret;
- MARKER_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
- return ret;
+ MARKER_STACK_UNWIND(setxattr, frame, op_ret, op_errno, xdata);
+ return ret;
}
int
-quota_xattr_cleaner (void *args)
+quota_xattr_cleaner(void *args)
{
- struct synctask *task = NULL;
- call_frame_t *frame = NULL;
- xlator_t *this = NULL;
- marker_local_t *local = NULL;
- dict_t *xdata = NULL;
- int ret = -1;
+ struct synctask *task = NULL;
+ call_frame_t *frame = NULL;
+ xlator_t *this = NULL;
+ marker_local_t *local = NULL;
+ dict_t *xdata = NULL;
+ int ret = -1;
- task = synctask_get ();
- if (!task)
- goto out;
+ task = synctask_get();
+ if (!task)
+ goto out;
- frame = task->frame;
- this = frame->this;
- local = frame->local;
+ frame = task->frame;
+ this = frame->this;
+ local = frame->local;
- ret = syncop_listxattr (FIRST_CHILD(this), &local->loc, &xdata, NULL,
- NULL);
- if (ret == -1) {
- ret = -errno;
- goto out;
- }
+ ret = syncop_listxattr(FIRST_CHILD(this), &local->loc, &xdata, NULL, NULL);
+ if (ret == -1) {
+ ret = -errno;
+ goto out;
+ }
- ret = dict_foreach_fnmatch (xdata, "trusted.glusterfs.quota.*",
- remove_quota_keys, frame);
- if (ret == -1) {
- ret = -errno;
- goto out;
- }
- ret = dict_foreach_fnmatch (xdata, PGFID_XATTR_KEY_PREFIX"*",
- remove_quota_keys, frame);
- if (ret == -1) {
- ret = -errno;
- goto out;
- }
+ ret = dict_foreach_fnmatch(xdata, "trusted.glusterfs.quota.*",
+ remove_quota_keys, frame);
+ if (ret == -1) {
+ ret = -errno;
+ goto out;
+ }
+ ret = dict_foreach_fnmatch(xdata, PGFID_XATTR_KEY_PREFIX "*",
+ remove_quota_keys, frame);
+ if (ret == -1) {
+ ret = -errno;
+ goto out;
+ }
- ret = 0;
+ ret = 0;
out:
- if (xdata)
- dict_unref (xdata);
+ if (xdata)
+ dict_unref(xdata);
- return ret;
+ return ret;
}
int
-marker_do_xattr_cleanup (call_frame_t *frame, xlator_t *this, dict_t *xdata,
+marker_do_xattr_cleanup(call_frame_t *frame, xlator_t *this, dict_t *xdata,
loc_t *loc)
{
- int ret = -1;
- marker_local_t *local = NULL;
+ int ret = -1;
+ marker_local_t *local = NULL;
- local = mem_get0 (this->local_pool);
- if (!local)
- goto out;
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto out;
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- loc_copy (&local->loc, loc);
- ret = synctask_new (this->ctx->env, quota_xattr_cleaner,
- quota_xattr_cleaner_cbk, frame, xdata);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "Failed to create synctask "
- "for cleaning up quota extended attributes");
- goto out;
- }
+ loc_copy(&local->loc, loc);
+ ret = synctask_new(this->ctx->env, quota_xattr_cleaner,
+ quota_xattr_cleaner_cbk, frame, xdata);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Failed to create synctask "
+ "for cleaning up quota extended attributes");
+ goto out;
+ }
- ret = 0;
+ ret = 0;
out:
- if (ret)
- MARKER_STACK_UNWIND (setxattr, frame, -1, ENOMEM, xdata);
+ if (ret)
+ MARKER_STACK_UNWIND(setxattr, frame, -1, ENOMEM, xdata);
- return ret;
+ return ret;
}
static gf_boolean_t
-marker_xattr_cleanup_cmd (dict_t *dict)
+marker_xattr_cleanup_cmd(dict_t *dict)
{
- return (dict_get (dict, VIRTUAL_QUOTA_XATTR_CLEANUP_KEY) != NULL);
+ return (dict_get(dict, VIRTUAL_QUOTA_XATTR_CLEANUP_KEY) != NULL);
}
int32_t
-marker_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
- int32_t flags, dict_t *xdata)
+marker_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
- int op_errno = ENOMEM;
-
- priv = this->private;
-
- if (marker_xattr_cleanup_cmd (dict)) {
- if (frame->root->uid != 0 || frame->root->gid != 0) {
- op_errno = EPERM;
- ret = -1;
- goto err;
- }
-
- /* The following function does the cleanup and then unwinds the
- * corresponding call*/
- loc_path (loc, NULL);
- marker_do_xattr_cleanup (frame, this, xdata, loc);
- return 0;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
+ int op_errno = ENOMEM;
+
+ priv = this->private;
+
+ if (marker_xattr_cleanup_cmd(dict)) {
+ if (frame->root->uid != 0 || frame->root->gid != 0) {
+ op_errno = EPERM;
+ ret = -1;
+ goto err;
}
- ret = marker_key_replace_with_ver (this, dict);
- if (ret < 0)
- goto err;
+ /* The following function does the cleanup and then unwinds the
+ * corresponding call*/
+ loc_path(loc, NULL);
+ marker_do_xattr_cleanup(frame, this, xdata, loc);
+ return 0;
+ }
- if (priv->feature_enabled == 0)
- goto wind;
+ ret = marker_key_replace_with_ver(this, dict);
+ if (ret < 0)
+ goto err;
- ret = call_from_sp_client_to_reset_tmfile (frame, this, dict);
- if (ret == 0)
- return 0;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
+ ret = call_from_sp_client_to_reset_tmfile(frame, this, dict);
+ if (ret == 0)
+ return 0;
- MARKER_INIT_LOCAL (frame, local);
+ local = mem_get0(this->local_pool);
- ret = loc_copy (&local->loc, loc);
+ MARKER_INIT_LOCAL(frame, local);
- if (ret == -1)
- goto err;
+ ret = loc_copy(&local->loc, loc);
+
+ if (ret == -1)
+ goto err;
wind:
- STACK_WIND (frame, marker_setxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setxattr, loc, dict, flags, xdata);
- return 0;
+ STACK_WIND(frame, marker_setxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, loc, dict, flags, xdata);
+ return 0;
err:
- MARKER_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
+ MARKER_STACK_UNWIND(setxattr, frame, -1, op_errno, NULL);
- return 0;
+ return 0;
}
-
int32_t
-marker_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+marker_fsetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE, "%s occurred in "
- "fsetxattr", strerror (op_errno));
- }
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s occurred in "
+ "fsetxattr",
+ strerror(op_errno));
+ }
- local = (marker_local_t *) frame->local;
+ local = (marker_local_t *)frame->local;
- frame->local = NULL;
+ frame->local = NULL;
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
+ STACK_UNWIND_STRICT(fsetxattr, frame, op_ret, op_errno, xdata);
- if (op_ret == -1 || local == NULL)
- goto out;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
out:
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int32_t
-marker_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
- int32_t flags, dict_t *xdata)
+marker_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- ret = call_from_sp_client_to_reset_tmfile (frame, this, dict);
- if (ret == 0)
- return 0;
+ ret = call_from_sp_client_to_reset_tmfile(frame, this, dict);
+ if (ret == 0)
+ return 0;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = marker_inode_loc_fill (fd->inode, &local->loc);
+ ret = marker_inode_loc_fill(fd->inode, &local->loc);
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
wind:
- STACK_WIND (frame, marker_fsetxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
- return 0;
+ STACK_WIND(frame, marker_fsetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+ return 0;
err:
- MARKER_STACK_UNWIND (fsetxattr, frame, -1, ENOMEM, NULL);
+ MARKER_STACK_UNWIND(fsetxattr, frame, -1, ENOMEM, NULL);
- return 0;
+ return 0;
}
-
int32_t
-marker_fsetattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *statpre,
- struct iatt *statpost, dict_t *xdata)
+marker_fsetattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
{
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE, "%s occurred in "
- "fsetattr ", strerror (op_errno));
- }
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s occurred in "
+ "fsetattr ",
+ strerror(op_errno));
+ }
- local = (marker_local_t *) frame->local;
+ local = (marker_local_t *)frame->local;
- frame->local = NULL;
+ frame->local = NULL;
- STACK_UNWIND_STRICT (fsetattr, frame, op_ret, op_errno, statpre,
- statpost, xdata);
+ STACK_UNWIND_STRICT(fsetattr, frame, op_ret, op_errno, statpre, statpost,
+ xdata);
- if (op_ret == -1 || local == NULL)
- goto out;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
out:
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
-
int32_t
-marker_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+marker_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = marker_inode_loc_fill (fd->inode, &local->loc);
+ ret = marker_inode_loc_fill(fd->inode, &local->loc);
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
wind:
- STACK_WIND (frame, marker_fsetattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fsetattr, fd, stbuf, valid, xdata);
- return 0;
+ STACK_WIND(frame, marker_fsetattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetattr, fd, stbuf, valid, xdata);
+ return 0;
err:
- MARKER_STACK_UNWIND (fsetattr, frame, -1, ENOMEM, NULL, NULL, NULL);
+ MARKER_STACK_UNWIND(fsetattr, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
int32_t
-marker_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *statpre,
- struct iatt *statpost, dict_t *xdata)
+marker_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
{
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- local = (marker_local_t *) frame->local;
+ local = (marker_local_t *)frame->local;
- frame->local = NULL;
+ frame->local = NULL;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE,
- "%s occurred during setattr of %s",
- strerror (op_errno),
- (local ? local->loc.path : "<nul>"));
- }
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE, "%s occurred during setattr of %s",
+ strerror(op_errno), (local ? local->loc.path : "<nul>"));
+ }
- STACK_UNWIND_STRICT (setattr, frame, op_ret, op_errno, statpre,
- statpost, xdata);
+ STACK_UNWIND_STRICT(setattr, frame, op_ret, op_errno, statpre, statpost,
+ xdata);
- if (op_ret == -1 || local == NULL)
- goto out;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
out:
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int32_t
-marker_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+marker_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = loc_copy (&local->loc, loc);
+ ret = loc_copy(&local->loc, loc);
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
wind:
- STACK_WIND (frame, marker_setattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setattr, loc, stbuf, valid, xdata);
- return 0;
+ STACK_WIND(frame, marker_setattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid, xdata);
+ return 0;
err:
- MARKER_STACK_UNWIND (setattr, frame, -1, ENOMEM, NULL, NULL, NULL);
+ MARKER_STACK_UNWIND(setattr, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ return 0;
}
-
int32_t
-marker_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+marker_removexattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE,
- "%s occurred while "
- "removing extended attribute",
- strerror (op_errno));
- }
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE,
+ "%s occurred while "
+ "removing extended attribute",
+ strerror(op_errno));
+ }
- local = (marker_local_t *) frame->local;
+ local = (marker_local_t *)frame->local;
- frame->local = NULL;
+ frame->local = NULL;
- STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
+ STACK_UNWIND_STRICT(removexattr, frame, op_ret, op_errno, xdata);
- if (op_ret == -1 || local == NULL)
- goto out;
+ if (op_ret == -1 || local == NULL)
+ goto out;
- priv = this->private;
+ priv = this->private;
- if (priv->feature_enabled & GF_XTIME)
- marker_xtime_update_marks (this, local);
+ if (priv->feature_enabled & GF_XTIME)
+ marker_xtime_update_marks(this, local);
out:
- marker_local_unref (local);
+ marker_local_unref(local);
- return 0;
+ return 0;
}
int32_t
-marker_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
-{
- int32_t ret = -1;
- int32_t i = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
- char key[QUOTA_KEY_MAX] = {0, };
-
- priv = this->private;
-
- if (name) {
- for (i = 0; mq_ext_xattrs[i]; i++) {
- if (strcmp (name, mq_ext_xattrs[i]))
- continue;
-
- GET_QUOTA_KEY (this, key, mq_ext_xattrs[i], ret);
- if (ret < 0)
- goto err;
- name = key;
- break;
- }
+marker_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
+{
+ int32_t ret = -1;
+ int32_t i = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
+ char key[QUOTA_KEY_MAX] = {
+ 0,
+ };
+
+ priv = this->private;
+
+ if (name) {
+ for (i = 0; mq_ext_xattrs[i]; i++) {
+ if (strcmp(name, mq_ext_xattrs[i]))
+ continue;
+
+ GET_QUOTA_KEY(this, key, mq_ext_xattrs[i], ret);
+ if (ret < 0)
+ goto err;
+ name = key;
+ break;
}
+ }
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
+ local = mem_get0(this->local_pool);
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = loc_copy (&local->loc, loc);
+ ret = loc_copy(&local->loc, loc);
- if (ret == -1)
- goto err;
+ if (ret == -1)
+ goto err;
wind:
- STACK_WIND (frame, marker_removexattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
- return 0;
+ STACK_WIND(frame, marker_removexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
+ return 0;
err:
- MARKER_STACK_UNWIND (removexattr, frame, -1, ENOMEM, NULL);
+ MARKER_STACK_UNWIND(removexattr, frame, -1, ENOMEM, NULL);
- return 0;
+ return 0;
}
static gf_boolean_t
-__has_quota_xattrs (dict_t *xattrs)
+__has_quota_xattrs(dict_t *xattrs)
{
- if (dict_foreach_match (xattrs, _is_quota_internal_xattr, NULL,
- dict_null_foreach_fn, NULL) > 0)
- return _gf_true;
+ if (dict_foreach_match(xattrs, _is_quota_internal_xattr, NULL,
+ dict_null_foreach_fn, NULL) > 0)
+ return _gf_true;
- return _gf_false;
+ return _gf_false;
}
int32_t
-marker_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *dict, struct iatt *postparent)
-{
- marker_conf_t *priv = NULL;
- marker_local_t *local = NULL;
- dict_t *xattrs = NULL;
- quota_inode_ctx_t *ctx = NULL;
- int32_t ret = -1;
-
- priv = this->private;
- local = (marker_local_t *) frame->local;
- frame->local = NULL;
-
- if (op_ret == -1) {
- gf_log (this->name, GF_LOG_TRACE, "lookup failed with %s",
- strerror (op_errno));
- goto unwind;
- }
-
- ret = marker_key_set_ver (this, dict);
- if (ret < 0) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
-
- if (dict && __has_quota_xattrs (dict)) {
- xattrs = dict_copy_with_ref (dict, NULL);
- if (!xattrs) {
- op_ret = -1;
- op_errno = ENOMEM;
- } else {
- marker_filter_internal_xattrs (this, xattrs);
- }
- } else if (dict) {
- xattrs = dict_ref (dict);
+marker_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *dict, struct iatt *postparent)
+{
+ marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
+ dict_t *xattrs = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ int32_t ret = -1;
+
+ priv = this->private;
+ local = (marker_local_t *)frame->local;
+ frame->local = NULL;
+
+ if (op_ret == -1) {
+ gf_log(this->name, GF_LOG_TRACE, "lookup failed with %s",
+ strerror(op_errno));
+ goto unwind;
+ }
+
+ ret = marker_key_set_ver(this, dict);
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ if (dict && __has_quota_xattrs(dict)) {
+ xattrs = dict_copy_with_ref(dict, NULL);
+ if (!xattrs) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ } else {
+ marker_filter_internal_xattrs(this, xattrs);
}
+ } else if (dict) {
+ xattrs = dict_ref(dict);
+ }
- if (op_ret >= 0 && inode && (priv->feature_enabled & GF_QUOTA)) {
- ctx = mq_inode_ctx_new (inode, this);
- if (ctx == NULL) {
- gf_log (this->name, GF_LOG_WARNING, "mq_inode_ctx_new "
- "failed for %s", uuid_utoa (inode->gfid));
- op_ret = -1;
- op_errno = ENOMEM;
- }
+ if (op_ret >= 0 && inode && (priv->feature_enabled & GF_QUOTA)) {
+ ctx = mq_inode_ctx_new(inode, this);
+ if (ctx == NULL) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "mq_inode_ctx_new "
+ "failed for %s",
+ uuid_utoa(inode->gfid));
+ op_ret = -1;
+ op_errno = ENOMEM;
}
+ }
unwind:
- STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
- xattrs, postparent);
-
- if (op_ret == -1 || local == NULL)
- goto out;
+ STACK_UNWIND_STRICT(lookup, frame, op_ret, op_errno, inode, buf, xattrs,
+ postparent);
- /* copy the gfid from the stat structure instead of inode,
- * since if the lookup is fresh lookup, then the inode
- * would have not yet linked to the inode table which happens
- * in protocol/server.
- */
- if (gf_uuid_is_null (local->loc.gfid))
- gf_uuid_copy (local->loc.gfid, buf->ia_gfid);
+ if (op_ret == -1 || local == NULL)
+ goto out;
+ /* copy the gfid from the stat structure instead of inode,
+ * since if the lookup is fresh lookup, then the inode
+ * would have not yet linked to the inode table which happens
+ * in protocol/server.
+ */
+ if (gf_uuid_is_null(local->loc.gfid))
+ gf_uuid_copy(local->loc.gfid, buf->ia_gfid);
- if (priv->feature_enabled & GF_QUOTA) {
- mq_xattr_state (this, &local->loc, dict, *buf);
- }
+ if (priv->feature_enabled & GF_QUOTA) {
+ mq_xattr_state(this, &local->loc, dict, buf);
+ }
out:
- marker_local_unref (local);
- if (xattrs)
- dict_unref (xattrs);
+ marker_local_unref(local);
+ if (xattrs)
+ dict_unref(xattrs);
- return 0;
+ return 0;
}
int32_t
-marker_lookup (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *xattr_req)
+marker_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ dict_t *xattr_req)
{
- int32_t ret = 0;
- marker_local_t *local = NULL;
- marker_conf_t *priv = NULL;
+ int32_t ret = 0;
+ marker_local_t *local = NULL;
+ marker_conf_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- xattr_req = xattr_req ? dict_ref (xattr_req) : dict_new ();
- if (!xattr_req)
- goto err;
+ xattr_req = xattr_req ? dict_ref(xattr_req) : dict_new();
+ if (!xattr_req)
+ goto err;
- ret = marker_key_replace_with_ver (this, xattr_req);
- if (ret < 0)
- goto err;
+ ret = marker_key_replace_with_ver(this, xattr_req);
+ if (ret < 0)
+ goto err;
- if (priv->feature_enabled == 0)
- goto wind;
+ if (priv->feature_enabled == 0)
+ goto wind;
- local = mem_get0 (this->local_pool);
- if (local == NULL)
- goto err;
+ local = mem_get0(this->local_pool);
+ if (local == NULL)
+ goto err;
- MARKER_INIT_LOCAL (frame, local);
+ MARKER_INIT_LOCAL(frame, local);
- ret = loc_copy (&local->loc, loc);
- if (ret == -1)
- goto err;
+ ret = loc_copy(&local->loc, loc);
+ if (ret == -1)
+ goto err;
- if ((priv->feature_enabled & GF_QUOTA))
- mq_req_xattr (this, loc, xattr_req, NULL, NULL);
+ if ((priv->feature_enabled & GF_QUOTA))
+ mq_req_xattr(this, loc, xattr_req, NULL, NULL);
wind:
- STACK_WIND (frame, marker_lookup_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
+ STACK_WIND(frame, marker_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
- dict_unref (xattr_req);
+ dict_unref(xattr_req);
- return 0;
+ return 0;
err:
- MARKER_STACK_UNWIND (lookup, frame, -1, ENOMEM, NULL, NULL, NULL, NULL);
+ MARKER_STACK_UNWIND(lookup, frame, -1, ENOMEM, NULL, NULL, NULL, NULL);
- if (xattr_req)
- dict_unref (xattr_req);
+ if (xattr_req)
+ dict_unref(xattr_req);
- return 0;
+ return 0;
}
-
int
-marker_build_ancestry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, gf_dirent_t *entries,
- dict_t *xdata)
+marker_build_ancestry_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, gf_dirent_t *entries,
+ dict_t *xdata)
{
- gf_dirent_t *entry = NULL;
- quota_inode_ctx_t *ctx = NULL;
- int ret = -1;
+ gf_dirent_t *entry = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ int ret = -1;
- if ((op_ret <= 0) || (entries == NULL)) {
- goto out;
- }
+ if ((op_ret <= 0) || (entries == NULL)) {
+ goto out;
+ }
- list_for_each_entry (entry, &entries->list, list) {
- if (entry->inode == NULL)
- continue;
-
- ret = marker_key_set_ver (this, entry->dict);
- if (ret < 0) {
- op_ret = -1;
- op_errno = ENOMEM;
- break;
- }
-
- ctx = mq_inode_ctx_new (entry->inode, this);
- if (ctx == NULL)
- gf_log (this->name, GF_LOG_WARNING, "mq_inode_ctx_new "
- "failed for %s",
- uuid_utoa (entry->inode->gfid));
+ list_for_each_entry(entry, &entries->list, list)
+ {
+ if (entry->inode == NULL)
+ continue;
+
+ ret = marker_key_set_ver(this, entry->dict);
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ break;
}
+ ctx = mq_inode_ctx_new(entry->inode, this);
+ if (ctx == NULL)
+ gf_log(this->name, GF_LOG_WARNING,
+ "mq_inode_ctx_new "
+ "failed for %s",
+ uuid_utoa(entry->inode->gfid));
+ }
+
out:
- STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries, xdata);
- return 0;
+ STACK_UNWIND_STRICT(readdirp, frame, op_ret, op_errno, entries, xdata);
+ return 0;
}
int
-marker_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, gf_dirent_t *entries,
- dict_t *xdata)
-{
- gf_dirent_t *entry = NULL;
- marker_conf_t *priv = NULL;
- marker_local_t *local = NULL;
- loc_t loc = {0, };
- int ret = -1;
- char *resolvedpath = NULL;
- quota_inode_ctx_t *ctx = NULL;
-
- if (op_ret <= 0)
- goto unwind;
-
- priv = this->private;
- local = frame->local;
-
- if (!(priv->feature_enabled & GF_QUOTA) || (local == NULL)) {
- goto unwind;
+marker_readdirp_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, gf_dirent_t *entries,
+ dict_t *xdata)
+{
+ gf_dirent_t *entry = NULL;
+ marker_conf_t *priv = NULL;
+ marker_local_t *local = NULL;
+ loc_t loc = {
+ 0,
+ };
+ int ret = -1;
+ char *resolvedpath = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+
+ if (op_ret <= 0)
+ goto unwind;
+
+ priv = this->private;
+ local = frame->local;
+
+ if (!(priv->feature_enabled & GF_QUOTA) || (local == NULL)) {
+ goto unwind;
+ }
+
+ list_for_each_entry(entry, &entries->list, list)
+ {
+ if ((strcmp(entry->d_name, ".") == 0) ||
+ (strcmp(entry->d_name, "..") == 0) || entry->inode == NULL)
+ continue;
+
+ loc.parent = inode_ref(local->loc.inode);
+ loc.inode = inode_ref(entry->inode);
+ ret = inode_path(loc.parent, entry->d_name, &resolvedpath);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to get the "
+ "path for the entry %s",
+ entry->d_name);
+ loc_wipe(&loc);
+ continue;
}
- list_for_each_entry (entry, &entries->list, list) {
- if ((strcmp (entry->d_name, ".") == 0) ||
- (strcmp (entry->d_name, "..") == 0) ||
- entry->inode == NULL)
- continue;
-
- loc.parent = inode_ref (local->loc.inode);
- loc.inode = inode_ref (entry->inode);
- ret = inode_path (loc.parent, entry->d_name, &resolvedpath);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "failed to get the "
- "path for the entry %s", entry->d_name);
- loc_wipe (&loc);
- continue;
- }
-
- loc.path = resolvedpath;
- resolvedpath = NULL;
-
- ctx = mq_inode_ctx_new (loc.inode, this);
- if (ctx == NULL)
- gf_log (this->name, GF_LOG_WARNING, "mq_inode_ctx_new "
- "failed for %s", uuid_utoa (loc.inode->gfid));
-
- mq_xattr_state (this, &loc, entry->dict, entry->d_stat);
- loc_wipe (&loc);
-
- ret = marker_key_set_ver (this, entry->dict);
- if (ret < 0) {
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
+ loc.path = resolvedpath;
+ resolvedpath = NULL;
+
+ ctx = mq_inode_ctx_new(loc.inode, this);
+ if (ctx == NULL)
+ gf_log(this->name, GF_LOG_WARNING,
+ "mq_inode_ctx_new "
+ "failed for %s",
+ uuid_utoa(loc.inode->gfid));
+
+ mq_xattr_state(this, &loc, entry->dict, &entry->d_stat);
+ loc_wipe(&loc);
+
+ ret = marker_key_set_ver(this, entry->dict);
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
}
+ }
unwind:
- MARKER_STACK_UNWIND (readdirp, frame, op_ret, op_errno, entries, xdata);
+ MARKER_STACK_UNWIND(readdirp, frame, op_ret, op_errno, entries, xdata);
- return 0;
+ return 0;
}
int
-marker_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, dict_t *dict)
-{
- marker_conf_t *priv = NULL;
- loc_t loc = {0, };
- marker_local_t *local = NULL;
- int ret = -1;
+marker_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *dict)
+{
+ marker_conf_t *priv = NULL;
+ loc_t loc = {
+ 0,
+ };
+ marker_local_t *local = NULL;
+ int ret = -1;
+
+ priv = this->private;
+
+ dict = dict ? dict_ref(dict) : dict_new();
+ if (!dict)
+ goto unwind;
+
+ ret = marker_key_replace_with_ver(this, dict);
+ if (ret < 0)
+ goto unwind;
+
+ if (dict_get(dict, GET_ANCESTRY_DENTRY_KEY)) {
+ STACK_WIND(frame, marker_build_ancestry_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdirp, fd, size, offset, dict);
+ } else {
+ if (priv->feature_enabled & GF_QUOTA) {
+ local = mem_get0(this->local_pool);
- priv = this->private;
+ MARKER_INIT_LOCAL(frame, local);
- dict = dict ? dict_ref(dict) : dict_new();
- if (!dict)
- goto unwind;
+ loc.parent = local->loc.inode = inode_ref(fd->inode);
- ret = marker_key_replace_with_ver (this, dict);
- if (ret < 0)
- goto unwind;
+ mq_req_xattr(this, &loc, dict, NULL, NULL);
+ }
- if (dict_get (dict, GET_ANCESTRY_DENTRY_KEY)) {
- STACK_WIND (frame, marker_build_ancestry_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdirp,
- fd, size, offset, dict);
- } else {
- if (priv->feature_enabled & GF_QUOTA) {
- local = mem_get0 (this->local_pool);
+ STACK_WIND(frame, marker_readdirp_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdirp, fd, size, offset, dict);
+ }
- MARKER_INIT_LOCAL (frame, local);
+ dict_unref(dict);
+ return 0;
+unwind:
+ MARKER_STACK_UNWIND(readdirp, frame, -1, ENOMEM, NULL, NULL);
+ return 0;
+}
- loc.parent = local->loc.inode = inode_ref (fd->inode);
+int32_t
+mem_acct_init(xlator_t *this)
+{
+ int ret = -1;
- mq_req_xattr (this, &loc, dict, NULL, NULL);
- }
+ if (!this)
+ return ret;
- STACK_WIND (frame, marker_readdirp_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdirp,
- fd, size, offset, dict);
- }
+ ret = xlator_mem_acct_init(this, gf_marker_mt_end + 1);
- dict_unref (dict);
- return 0;
-unwind:
- MARKER_STACK_UNWIND (readdirp, frame, -1, ENOMEM, NULL, NULL);
- return 0;
+ if (ret != 0) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Memory accounting init"
+ " failed");
+ return ret;
+ }
+
+ return ret;
}
int32_t
-mem_acct_init (xlator_t *this)
+init_xtime_priv(xlator_t *this, dict_t *options)
{
- int ret = -1;
+ int32_t ret = -1;
+ marker_conf_t *priv = NULL;
+ char *tmp_opt = NULL;
- if (!this)
- return ret;
+ GF_VALIDATE_OR_GOTO("marker", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, options, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
- ret = xlator_mem_acct_init (this, gf_marker_mt_end + 1);
+ priv = this->private;
- if (ret != 0) {
- gf_log(this->name, GF_LOG_ERROR, "Memory accounting init"
- " failed");
- return ret;
- }
+ ret = dict_get_str(options, "volume-uuid", &tmp_opt);
- return ret;
-}
+ if (ret) {
+ priv->volume_uuid = NULL;
+ tmp_opt = "";
+ gf_log(this->name, GF_LOG_ERROR,
+ "please specify the volume-uuid"
+ "in the translator options");
-int32_t
-init_xtime_priv (xlator_t *this, dict_t *options)
-{
- data_t *data = NULL;
- int32_t ret = -1;
- marker_conf_t *priv = NULL;
-
- GF_VALIDATE_OR_GOTO ("marker", this, out);
- GF_VALIDATE_OR_GOTO (this->name, options, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
-
- priv = this->private;
- data = dict_get (options, VOLUME_UUID);
- if (data != NULL) {
- priv->volume_uuid = data->data;
-
- ret = gf_uuid_parse (priv->volume_uuid, priv->volume_uuid_bin);
- if (ret == -1) {
- gf_log (this->name, GF_LOG_ERROR,
- "invalid volume uuid %s", priv->volume_uuid);
- goto out;
- }
-
- ret = gf_asprintf (&(priv->marker_xattr), "%s.%s.%s",
- MARKER_XATTR_PREFIX, priv->volume_uuid,
- XTIME);
-
- if (ret == -1) {
- priv->marker_xattr = NULL;
- goto out;
- }
-
- gf_log (this->name, GF_LOG_DEBUG,
- "volume-uuid = %s", priv->volume_uuid);
- } else {
- priv->volume_uuid = NULL;
+ return -1;
+ }
+ gf_asprintf(&priv->volume_uuid, "%s", tmp_opt);
- gf_log (this->name, GF_LOG_ERROR,
- "please specify the volume-uuid"
- "in the translator options");
+ ret = gf_uuid_parse(priv->volume_uuid, priv->volume_uuid_bin);
- return -1;
- }
- data = dict_get (options, TIMESTAMP_FILE);
- if (data != NULL) {
- priv->timestamp_file = data->data;
+ if (ret == -1) {
+ gf_log(this->name, GF_LOG_ERROR, "invalid volume uuid %s",
+ priv->volume_uuid);
+ goto out;
+ }
- gf_log (this->name, GF_LOG_DEBUG,
- "the timestamp-file is = %s",
- priv->timestamp_file);
+ ret = gf_asprintf(&(priv->marker_xattr), "%s.%s.%s", MARKER_XATTR_PREFIX,
+ priv->volume_uuid, XTIME);
- } else {
- priv->timestamp_file = NULL;
+ if (ret == -1) {
+ priv->marker_xattr = NULL;
+ goto out;
+ }
- gf_log (this->name, GF_LOG_ERROR,
- "please specify the timestamp-file"
- "in the translator options");
+ gf_log(this->name, GF_LOG_DEBUG, "volume-uuid = %s", priv->volume_uuid);
- goto out;
- }
+ ret = dict_get_str(options, "timestamp-file", &tmp_opt);
+ if (ret) {
+ priv->timestamp_file = NULL;
+ tmp_opt = "";
- ret = 0;
+ gf_log(this->name, GF_LOG_ERROR,
+ "please specify the timestamp-file"
+ "in the translator options");
+
+ goto out;
+ }
+
+ ret = gf_asprintf(&priv->timestamp_file, "%s", tmp_opt);
+ if (ret == -1) {
+ priv->timestamp_file = NULL;
+ goto out;
+ }
+
+ gf_log(this->name, GF_LOG_DEBUG, "the timestamp-file is = %s",
+ priv->timestamp_file);
+
+ ret = 0;
out:
- return ret;
+ return ret;
}
void
-marker_xtime_priv_cleanup (xlator_t *this)
+marker_xtime_priv_cleanup(xlator_t *this)
{
- marker_conf_t *priv = NULL;
+ marker_conf_t *priv = NULL;
- GF_VALIDATE_OR_GOTO ("marker", this, out);
+ GF_VALIDATE_OR_GOTO("marker", this, out);
- priv = (marker_conf_t *) this->private;
+ priv = (marker_conf_t *)this->private;
- GF_VALIDATE_OR_GOTO (this->name, priv, out);
+ GF_VALIDATE_OR_GOTO(this->name, priv, out);
- GF_FREE (priv->volume_uuid);
+ GF_FREE(priv->volume_uuid);
- GF_FREE (priv->timestamp_file);
+ GF_FREE(priv->timestamp_file);
- GF_FREE (priv->marker_xattr);
+ GF_FREE(priv->marker_xattr);
out:
- return;
+ return;
}
void
-marker_priv_cleanup (xlator_t *this)
+marker_priv_cleanup(xlator_t *this)
{
- marker_conf_t *priv = NULL;
+ marker_conf_t *priv = NULL;
+
+ GF_VALIDATE_OR_GOTO("marker", this, out);
- GF_VALIDATE_OR_GOTO ("marker", this, out);
+ priv = (marker_conf_t *)this->private;
- priv = (marker_conf_t *) this->private;
+ GF_VALIDATE_OR_GOTO(this->name, priv, out);
- GF_VALIDATE_OR_GOTO (this->name, priv, out);
+ marker_xtime_priv_cleanup(this);
- marker_xtime_priv_cleanup (this);
+ LOCK_DESTROY(&priv->lock);
- LOCK_DESTROY (&priv->lock);
+ GF_FREE(priv);
+
+ if (this->local_pool) {
+ mem_pool_destroy(this->local_pool);
+ this->local_pool = NULL;
+ }
- GF_FREE (priv);
out:
- return;
+ return;
}
int32_t
-reconfigure (xlator_t *this, dict_t *options)
+reconfigure(xlator_t *this, dict_t *options)
{
- int32_t ret = 0;
- data_t *data = NULL;
- gf_boolean_t flag = _gf_false;
- marker_conf_t *priv = NULL;
- int32_t version = 0;
+ int32_t ret = 0;
+ data_t *data = NULL;
+ gf_boolean_t flag = _gf_false;
+ marker_conf_t *priv = NULL;
+ int32_t version = 0;
- GF_ASSERT (this);
- GF_ASSERT (this->private);
+ GF_ASSERT(this);
+ GF_ASSERT(this->private);
- priv = this->private;
+ priv = this->private;
- priv->feature_enabled = 0;
+ priv->feature_enabled = 0;
- GF_VALIDATE_OR_GOTO (this->name, options, out);
+ GF_VALIDATE_OR_GOTO(this->name, options, out);
- data = dict_get (options, "quota");
- if (data) {
- ret = gf_string2boolean (data->data, &flag);
- if (ret == 0 && flag == _gf_true)
- priv->feature_enabled |= GF_QUOTA;
- }
+ data = dict_get(options, "quota");
+ if (data) {
+ ret = gf_string2boolean(data->data, &flag);
+ if (ret == 0 && flag == _gf_true)
+ priv->feature_enabled |= GF_QUOTA;
+ }
- data = dict_get (options, "inode-quota");
- if (data) {
- ret = gf_string2boolean (data->data, &flag);
- if (ret == 0 && flag == _gf_true)
- priv->feature_enabled |= GF_INODE_QUOTA;
- }
-
- data = dict_get (options, "quota-version");
- if (data)
- ret = gf_string2int32 (data->data, &version);
+ data = dict_get(options, "inode-quota");
+ if (data) {
+ ret = gf_string2boolean(data->data, &flag);
+ if (ret == 0 && flag == _gf_true)
+ priv->feature_enabled |= GF_INODE_QUOTA;
+ }
- if (priv->feature_enabled) {
- if (version >= 0)
- priv->version = version;
- else
- gf_log (this->name, GF_LOG_ERROR, "Invalid quota "
- "version %d", priv->version);
- }
+ data = dict_get(options, "quota-version");
+ if (data)
+ ret = gf_string2int32(data->data, &version);
- data = dict_get (options, "xtime");
- if (data) {
- ret = gf_string2boolean (data->data, &flag);
- if (ret == 0 && flag == _gf_true) {
- marker_xtime_priv_cleanup (this);
-
- ret = init_xtime_priv (this, options);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_WARNING,
- "failed to initialize xtime private, "
- "xtime updation will fail");
- } else {
- priv->feature_enabled |= GF_XTIME;
- data = dict_get (options, "gsync-force-xtime");
- if (!data)
- goto out;
- ret = gf_string2boolean (data->data, &flag);
- if (ret == 0 && flag)
- priv->feature_enabled |= GF_XTIME_GSYNC_FORCE;
- }
- }
- }
+ if (priv->feature_enabled) {
+ if (version >= 0)
+ priv->version = version;
+ else
+ gf_log(this->name, GF_LOG_ERROR,
+ "Invalid quota "
+ "version %d",
+ priv->version);
+ }
+
+ data = dict_get(options, "xtime");
+ if (data) {
+ ret = gf_string2boolean(data->data, &flag);
+ if (ret == 0 && flag == _gf_true) {
+ marker_xtime_priv_cleanup(this);
+
+ ret = init_xtime_priv(this, options);
+ if (ret < 0) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "failed to initialize xtime private, "
+ "xtime updation will fail");
+ } else {
+ priv->feature_enabled |= GF_XTIME;
+ data = dict_get(options, "gsync-force-xtime");
+ if (!data)
+ goto out;
+ ret = gf_string2boolean(data->data, &flag);
+ if (ret == 0 && flag)
+ priv->feature_enabled |= GF_XTIME_GSYNC_FORCE;
+ }
+ }
+ }
out:
- return ret;
+ return ret;
}
-
int32_t
-init (xlator_t *this)
-{
- dict_t *options = NULL;
- data_t *data = NULL;
- int32_t ret = 0;
- gf_boolean_t flag = _gf_false;
- marker_conf_t *priv = NULL;
-
- if (!this->children) {
- gf_log (this->name, GF_LOG_ERROR,
- "marker translator needs subvolume defined.");
- return -1;
- }
+init(xlator_t *this)
+{
+ dict_t *options = NULL;
+ data_t *data = NULL;
+ int32_t ret = 0;
+ gf_boolean_t flag = _gf_false;
+ marker_conf_t *priv = NULL;
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "Volume is dangling.");
- return -1;
- }
+ if (!this->children) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "marker translator needs subvolume defined.");
+ return -1;
+ }
- options = this->options;
+ if (!this->parents) {
+ gf_log(this->name, GF_LOG_WARNING, "Volume is dangling.");
+ return -1;
+ }
- ALLOCATE_OR_GOTO (this->private, marker_conf_t, err);
+ options = this->options;
- priv = this->private;
+ ALLOCATE_OR_GOTO(this->private, marker_conf_t, err);
- priv->feature_enabled = 0;
- priv->version = 0;
+ priv = this->private;
- LOCK_INIT (&priv->lock);
+ priv->feature_enabled = 0;
+ priv->version = 0;
- data = dict_get (options, "quota");
- if (data) {
- ret = gf_string2boolean (data->data, &flag);
- if (ret == 0 && flag == _gf_true)
- priv->feature_enabled |= GF_QUOTA;
- }
+ LOCK_INIT(&priv->lock);
- data = dict_get (options, "inode-quota");
- if (data) {
- ret = gf_string2boolean (data->data, &flag);
- if (ret == 0 && flag == _gf_true)
- priv->feature_enabled |= GF_INODE_QUOTA;
- }
+ data = dict_get(options, "quota");
+ if (data) {
+ ret = gf_string2boolean(data->data, &flag);
+ if (ret == 0 && flag == _gf_true)
+ priv->feature_enabled |= GF_QUOTA;
+ }
- data = dict_get (options, "quota-version");
- if (data)
- ret = gf_string2int32 (data->data, &priv->version);
+ data = dict_get(options, "inode-quota");
+ if (data) {
+ ret = gf_string2boolean(data->data, &flag);
+ if (ret == 0 && flag == _gf_true)
+ priv->feature_enabled |= GF_INODE_QUOTA;
+ }
- if (priv->feature_enabled && priv->version < 0) {
- gf_log (this->name, GF_LOG_ERROR, "Invalid quota version %d",
- priv->version);
- goto err;
- }
+ data = dict_get(options, "quota-version");
+ if (data)
+ ret = gf_string2int32(data->data, &priv->version);
- data = dict_get (options, "xtime");
- if (data) {
- ret = gf_string2boolean (data->data, &flag);
- if (ret == 0 && flag == _gf_true) {
- ret = init_xtime_priv (this, options);
- if (ret < 0)
- goto err;
-
- priv->feature_enabled |= GF_XTIME;
- data = dict_get (options, "gsync-force-xtime");
- if (!data)
- goto cont;
- ret = gf_string2boolean (data->data, &flag);
- if (ret == 0 && flag)
- priv->feature_enabled |= GF_XTIME_GSYNC_FORCE;
- }
- }
+ if ((ret == 0) && priv->feature_enabled && priv->version < 0) {
+ gf_log(this->name, GF_LOG_ERROR, "Invalid quota version %d",
+ priv->version);
+ goto err;
+ }
- cont:
- this->local_pool = mem_pool_new (marker_local_t, 128);
- if (!this->local_pool) {
- gf_log (this->name, GF_LOG_ERROR,
- "failed to create local_t's memory pool");
+ data = dict_get(options, "xtime");
+ if (data) {
+ ret = gf_string2boolean(data->data, &flag);
+ if (ret == 0 && flag == _gf_true) {
+ ret = init_xtime_priv(this, options);
+ if (ret < 0)
goto err;
- }
- return 0;
+ priv->feature_enabled |= GF_XTIME;
+ data = dict_get(options, "gsync-force-xtime");
+ if (!data)
+ goto cont;
+ ret = gf_string2boolean(data->data, &flag);
+ if (ret == 0 && flag)
+ priv->feature_enabled |= GF_XTIME_GSYNC_FORCE;
+ }
+ }
+
+cont:
+ this->local_pool = mem_pool_new(marker_local_t, 128);
+ if (!this->local_pool) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to create local_t's memory pool");
+ goto err;
+ }
+
+ return 0;
err:
- marker_priv_cleanup (this);
+ marker_priv_cleanup(this);
- return -1;
+ return -1;
}
int32_t
-marker_forget (xlator_t *this, inode_t *inode)
+marker_forget(xlator_t *this, inode_t *inode)
{
- marker_inode_ctx_t *ctx = NULL;
- uint64_t value = 0;
+ marker_inode_ctx_t *ctx = NULL;
+ uint64_t value = 0;
- if (inode_ctx_del (inode, this, &value) != 0)
- goto out;
+ if (inode_ctx_del(inode, this, &value) != 0)
+ goto out;
- ctx = (marker_inode_ctx_t *)(unsigned long)value;
- if (ctx == NULL) {
- goto out;
- }
+ ctx = (marker_inode_ctx_t *)(unsigned long)value;
+ if (ctx == NULL) {
+ goto out;
+ }
- mq_forget (this, ctx->quota_ctx);
+ mq_forget(this, ctx->quota_ctx);
- GF_FREE (ctx);
+ GF_FREE(ctx);
out:
- return 0;
+ return 0;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- marker_priv_cleanup (this);
+ marker_priv_cleanup(this);
}
struct xlator_fops fops = {
- .lookup = marker_lookup,
- .create = marker_create,
- .mkdir = marker_mkdir,
- .writev = marker_writev,
- .truncate = marker_truncate,
- .ftruncate = marker_ftruncate,
- .symlink = marker_symlink,
- .link = marker_link,
- .unlink = marker_unlink,
- .rmdir = marker_rmdir,
- .rename = marker_rename,
- .mknod = marker_mknod,
- .setxattr = marker_setxattr,
- .fsetxattr = marker_fsetxattr,
- .setattr = marker_setattr,
- .fsetattr = marker_fsetattr,
- .removexattr = marker_removexattr,
- .getxattr = marker_getxattr,
- .readdirp = marker_readdirp,
- .fallocate = marker_fallocate,
- .discard = marker_discard,
- .zerofill = marker_zerofill,
+ .lookup = marker_lookup,
+ .create = marker_create,
+ .mkdir = marker_mkdir,
+ .writev = marker_writev,
+ .truncate = marker_truncate,
+ .ftruncate = marker_ftruncate,
+ .symlink = marker_symlink,
+ .link = marker_link,
+ .unlink = marker_unlink,
+ .rmdir = marker_rmdir,
+ .rename = marker_rename,
+ .mknod = marker_mknod,
+ .setxattr = marker_setxattr,
+ .fsetxattr = marker_fsetxattr,
+ .setattr = marker_setattr,
+ .fsetattr = marker_fsetattr,
+ .removexattr = marker_removexattr,
+ .getxattr = marker_getxattr,
+ .readdirp = marker_readdirp,
+ .fallocate = marker_fallocate,
+ .discard = marker_discard,
+ .zerofill = marker_zerofill,
};
-struct xlator_cbks cbks = {
- .forget = marker_forget
-};
+struct xlator_cbks cbks = {.forget = marker_forget};
struct volume_options options[] = {
- {.key = {"volume-uuid"},
- .default_value = "{{ volume.id }}"},
- {.key = {"timestamp-file"}},
- {.key = {"quota"},
- .op_version = {1},
- .flags = OPT_FLAG_NONE,
- .tags = {},
- },
- {.key = {"inode-quota"},
- .op_version = {1},
- .flags = OPT_FLAG_NONE,
- .tags = {},
- },
- {.key = {"xtime"},
- .op_version = {1},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_FORCE,
- .tags = {},
- },
- {.key = {"gsync-force-xtime"},
- .op_version = {2},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_FORCE,
- .tags = {},
- },
- {.key = {"quota-version"},
- .flags = OPT_FLAG_NONE,
- },
- {.key = {NULL}}
+ {.key = {"volume-uuid"}, .default_value = "{{ volume.id }}"},
+ {.key = {"timestamp-file"}},
+ {
+ .key = {"quota"},
+ .op_version = {1},
+ .flags = OPT_FLAG_NONE,
+ .tags = {},
+ },
+ {
+ .key = {"inode-quota"},
+ .op_version = {1},
+ .flags = OPT_FLAG_NONE,
+ .tags = {},
+ },
+ {
+ .key = {"xtime"},
+ .op_version = {1},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_FORCE,
+ .tags = {},
+ },
+ {
+ .key = {"gsync-force-xtime"},
+ .op_version = {2},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_FORCE,
+ .tags = {},
+ },
+ {
+ .key = {"quota-version"},
+ .flags = OPT_FLAG_NONE,
+ },
+ {.key = {NULL}}};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "marker",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/features/marker/src/marker.h b/xlators/features/marker/src/marker.h
index 4726880b82f..4821094c14b 100644
--- a/xlators/features/marker/src/marker.h
+++ b/xlators/features/marker/src/marker.h
@@ -11,138 +11,137 @@
#define _MARKER_H
#include "marker-quota.h"
-#include "xlator.h"
-#include "defaults.h"
-#include "compat-uuid.h"
-#include "call-stub.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
+#include <glusterfs/compat-uuid.h>
+#include <glusterfs/call-stub.h>
#define MARKER_XATTR_PREFIX "trusted.glusterfs"
-#define XTIME "xtime"
-#define VOLUME_MARK "volume-mark"
-#define VOLUME_UUID "volume-uuid"
-#define TIMESTAMP_FILE "timestamp-file"
+#define XTIME "xtime"
+#define VOLUME_MARK "volume-mark"
+#define VOLUME_UUID "volume-uuid"
+#define TIMESTAMP_FILE "timestamp-file"
enum {
- GF_QUOTA = 1,
- GF_XTIME = 2,
- GF_XTIME_GSYNC_FORCE = 4,
- GF_INODE_QUOTA = 8,
+ GF_QUOTA = 1,
+ GF_XTIME = 2,
+ GF_XTIME_GSYNC_FORCE = 4,
+ GF_INODE_QUOTA = 8,
};
/*initialize the local variable*/
-#define MARKER_INIT_LOCAL(_frame,_local) do { \
- _frame->local = _local; \
- _local->pid = _frame->root->pid; \
- memset (&_local->loc, 0, sizeof (loc_t)); \
- _local->ref = 1; \
- _local->uid = -1; \
- _local->gid = -1; \
- LOCK_INIT (&_local->lock); \
- _local->oplocal = NULL; \
- } while (0)
+#define MARKER_INIT_LOCAL(_frame, _local) \
+ do { \
+ _frame->local = _local; \
+ _local->pid = _frame->root->pid; \
+ memset(&_local->loc, 0, sizeof(loc_t)); \
+ _local->ref = 1; \
+ _local->uid = -1; \
+ _local->gid = -1; \
+ LOCK_INIT(&_local->lock); \
+ _local->oplocal = NULL; \
+ } while (0)
/* try alloc and if it fails, goto label */
-#define ALLOCATE_OR_GOTO(var, type, label) do { \
- var = GF_CALLOC (sizeof (type), 1, \
- gf_marker_mt_##type); \
- if (!var) { \
- gf_log (this->name, GF_LOG_ERROR, \
- "out of memory :("); \
- goto label; \
- } \
- } while (0)
-
-#define _MARKER_SET_UID_GID(dest, src) \
- do { \
- if (src->uid != -1 && \
- src->gid != -1) { \
- dest->uid = src->uid; \
- dest->gid = src->gid; \
- } \
- } while (0)
-
-#define MARKER_SET_UID_GID(frame, dest, src) \
- do { \
- _MARKER_SET_UID_GID (dest, src); \
- frame->root->uid = 0; \
- frame->root->gid = 0; \
- frame->cookie = (void *) _GF_UID_GID_CHANGED; \
- } while (0)
-
-#define MARKER_RESET_UID_GID(frame, dest, src) \
- do { \
- _MARKER_SET_UID_GID (dest, src); \
- frame->cookie = NULL; \
- } while (0)
-
-#define MARKER_STACK_UNWIND(fop, frame, params...) \
- do { \
- quota_local_t *_local = NULL; \
- if (frame) { \
- _local = frame->local; \
- frame->local = NULL; \
- } \
- STACK_UNWIND_STRICT (fop, frame, params); \
- if (_local) \
- marker_local_unref (_local); \
- } while (0)
-
-struct marker_local{
- uint32_t timebuf[2];
- pid_t pid;
- loc_t loc;
- loc_t parent_loc;
- uid_t uid;
- gid_t gid;
- int32_t ref;
- uint32_t ia_nlink;
- struct iatt buf;
- gf_lock_t lock;
- mode_t mode;
- int32_t err;
- call_stub_t *stub;
- call_frame_t *lk_frame;
- quota_meta_t contribution;
- struct marker_local *oplocal;
-
- /* marker quota specific */
- int64_t delta;
- int64_t d_off;
- int64_t sum;
- int64_t size;
- int32_t hl_count;
- int32_t dentry_child_count;
-
- fd_t *fd;
- call_frame_t *frame;
-
- quota_inode_ctx_t *ctx;
- inode_contribution_t *contri;
-
- int xflag;
- dict_t *xdata;
- gf_boolean_t skip_txn;
+#define ALLOCATE_OR_GOTO(var, type, label) \
+ do { \
+ var = GF_CALLOC(sizeof(type), 1, gf_marker_mt_##type); \
+ if (!var) { \
+ gf_log(this->name, GF_LOG_ERROR, "out of memory :("); \
+ goto label; \
+ } \
+ } while (0)
+
+#define _MARKER_SET_UID_GID(dest, src) \
+ do { \
+ if (src->uid != -1 && src->gid != -1) { \
+ dest->uid = src->uid; \
+ dest->gid = src->gid; \
+ } \
+ } while (0)
+
+#define MARKER_SET_UID_GID(frame, dest, src) \
+ do { \
+ _MARKER_SET_UID_GID(dest, src); \
+ frame->root->uid = 0; \
+ frame->root->gid = 0; \
+ frame->cookie = (void *)_GF_UID_GID_CHANGED; \
+ } while (0)
+
+#define MARKER_RESET_UID_GID(frame, dest, src) \
+ do { \
+ _MARKER_SET_UID_GID(dest, src); \
+ frame->cookie = NULL; \
+ } while (0)
+
+#define MARKER_STACK_UNWIND(fop, frame, params...) \
+ do { \
+ quota_local_t *_local = NULL; \
+ if (frame) { \
+ _local = frame->local; \
+ frame->local = NULL; \
+ } \
+ STACK_UNWIND_STRICT(fop, frame, params); \
+ if (_local) \
+ marker_local_unref(_local); \
+ } while (0)
+
+struct marker_local {
+ uint32_t timebuf[2];
+ pid_t pid;
+ loc_t loc;
+ loc_t parent_loc;
+ uid_t uid;
+ gid_t gid;
+ int32_t ref;
+ uint32_t ia_nlink;
+ struct iatt buf;
+ gf_lock_t lock;
+ mode_t mode;
+ int32_t err;
+ call_stub_t *stub;
+ call_frame_t *lk_frame;
+ quota_meta_t contribution;
+ struct marker_local *oplocal;
+
+ /* marker quota specific */
+ int64_t delta;
+ int64_t d_off;
+ int64_t sum;
+ int64_t size;
+ int32_t hl_count;
+ int32_t dentry_child_count;
+
+ fd_t *fd;
+ call_frame_t *frame;
+
+ quota_inode_ctx_t *ctx;
+ inode_contribution_t *contri;
+
+ int xflag;
+ dict_t *xdata;
+ gf_boolean_t skip_txn;
};
typedef struct marker_local marker_local_t;
#define quota_local_t marker_local_t
struct marker_inode_ctx {
- struct quota_inode_ctx *quota_ctx;
+ struct quota_inode_ctx *quota_ctx;
};
typedef struct marker_inode_ctx marker_inode_ctx_t;
-struct marker_conf{
- char feature_enabled;
- char *size_key;
- char *dirty_key;
- char *volume_uuid;
- uuid_t volume_uuid_bin;
- char *timestamp_file;
- char *marker_xattr;
- uint64_t quota_lk_owner;
- gf_lock_t lock;
- int32_t version;
+struct marker_conf {
+ char feature_enabled;
+ char *size_key;
+ char *dirty_key;
+ char *volume_uuid;
+ uuid_t volume_uuid_bin;
+ char *timestamp_file;
+ char *marker_xattr;
+ uint64_t quota_lk_owner;
+ gf_lock_t lock;
+ int32_t version;
};
typedef struct marker_conf marker_conf_t;
diff --git a/xlators/encryption/crypt/Makefile.am b/xlators/features/metadisp/Makefile.am
index d471a3f9243..a985f42a877 100644
--- a/xlators/encryption/crypt/Makefile.am
+++ b/xlators/features/metadisp/Makefile.am
@@ -1,3 +1,3 @@
SUBDIRS = src
-CLEANFILES =
+CLEANFILES =
diff --git a/xlators/features/metadisp/src/Makefile.am b/xlators/features/metadisp/src/Makefile.am
new file mode 100644
index 00000000000..1520ad8c424
--- /dev/null
+++ b/xlators/features/metadisp/src/Makefile.am
@@ -0,0 +1,38 @@
+noinst_PYTHON = gen-fops.py
+
+EXTRA_DIST = fops-tmpl.c
+
+xlator_LTLIBRARIES = metadisp.la
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+nodist_metadisp_la_SOURCES = fops.c
+
+BUILT_SOURCES = fops.c
+
+metadisp_la_LDFLAGS = -module $(GF_XLATOR_DEFAULT_LDFLAGS)
+
+metadisp_la_SOURCES = metadisp.c \
+ metadisp-unlink.c \
+ metadisp-stat.c \
+ metadisp-lookup.c \
+ metadisp-readdir.c \
+ metadisp-create.c \
+ metadisp-open.c \
+ metadisp-fsync.c \
+ metadisp-setattr.c \
+ backend.c
+
+metadisp_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+noinst_HEADERS = metadisp.h metadisp-fops.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
+ -I$(top_srcdir)/rpc/xdr/src -I$(top_builddir)/rpc/xdr/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+fops.c: fops-tmpl.c $(top_srcdir)/libglusterfs/src/generator.py gen-fops.py
+ PYTHONPATH=$(top_srcdir)/libglusterfs/src \
+ $(PYTHON) $(srcdir)/gen-fops.py $(srcdir)/fops-tmpl.c > $@
+
+CLEANFILES = $(nodist_metadisp_la_SOURCES)
diff --git a/xlators/features/metadisp/src/backend.c b/xlators/features/metadisp/src/backend.c
new file mode 100644
index 00000000000..ee2c25bfaa7
--- /dev/null
+++ b/xlators/features/metadisp/src/backend.c
@@ -0,0 +1,45 @@
+#define GFID_STR_LEN 37
+
+#include "metadisp.h"
+
+/*
+ * backend.c
+ *
+ * functions responsible for converting user-facing paths to backend-style
+ * "/$GFID" paths.
+ */
+
+int32_t
+build_backend_loc(uuid_t gfid, loc_t *src_loc, loc_t *dst_loc)
+{
+ static uuid_t root = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1};
+ char gfid_buf[GFID_STR_LEN + 1] = {
+ 0,
+ };
+ char *path = NULL;
+
+ GF_VALIDATE_OR_GOTO("metadisp", src_loc, out);
+ GF_VALIDATE_OR_GOTO("metadisp", dst_loc, out);
+
+ loc_copy(dst_loc, src_loc);
+ memcpy(dst_loc->pargfid, root, sizeof(root));
+ GF_FREE((char *)dst_loc->path); // we are overwriting path so nuke
+ // whatever loc_copy gave us
+
+ uuid_utoa_r(gfid, gfid_buf);
+
+ path = GF_CALLOC(GFID_STR_LEN + 1, sizeof(char),
+ gf_common_mt_char); // freed via loc_wipe
+
+ path[0] = '/';
+ strncpy(path + 1, gfid_buf, GFID_STR_LEN);
+ path[GFID_STR_LEN] = 0;
+ dst_loc->path = path;
+ if (src_loc->name)
+ dst_loc->name = strrchr(dst_loc->path, '/');
+ if (dst_loc->name)
+ dst_loc->name++;
+ return 0;
+out:
+ return -1;
+}
diff --git a/xlators/features/metadisp/src/fops-tmpl.c b/xlators/features/metadisp/src/fops-tmpl.c
new file mode 100644
index 00000000000..4385b7dd5b7
--- /dev/null
+++ b/xlators/features/metadisp/src/fops-tmpl.c
@@ -0,0 +1,10 @@
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include <glusterfs/xlator.h>
+#include "metadisp.h"
+#include "metadisp-fops.h"
+
+#pragma generate
diff --git a/xlators/features/metadisp/src/gen-fops.py b/xlators/features/metadisp/src/gen-fops.py
new file mode 100644
index 00000000000..8b5e120fdec
--- /dev/null
+++ b/xlators/features/metadisp/src/gen-fops.py
@@ -0,0 +1,160 @@
+#!/usr/bin/python
+
+import sys
+from generator import fop_subs, generate
+
+FN_METADATA_CHILD_GENERIC = """
+int32_t
+metadisp_@NAME@ (call_frame_t *frame, xlator_t *this,
+ @LONG_ARGS@)
+{
+ METADISP_TRACE("@NAME@ metadata");
+ STACK_WIND (frame, default_@NAME@_cbk,
+ METADATA_CHILD(this), METADATA_CHILD(this)->fops->@NAME@,
+ @SHORT_ARGS@);
+ return 0;
+}
+"""
+
+FN_GENERIC_TEMPLATE = """
+int32_t
+metadisp_@NAME@ (call_frame_t *frame, xlator_t *this,
+ @LONG_ARGS@)
+{
+ METADISP_TRACE("@NAME@ generic");
+ STACK_WIND (frame, default_@NAME@_cbk,
+ DATA_CHILD(this), DATA_CHILD(this)->fops->@NAME@,
+ @SHORT_ARGS@);
+ return 0;
+}
+"""
+
+FN_DATAFD_TEMPLATE = """
+int32_t
+metadisp_@NAME@ (call_frame_t *frame, xlator_t *this,
+ @LONG_ARGS@)
+{
+ METADISP_TRACE("@NAME@ datafd");
+ xlator_t *child = NULL;
+ child = DATA_CHILD(this);
+ STACK_WIND (frame, default_@NAME@_cbk,
+ child, child->fops->@NAME@,
+ @SHORT_ARGS@);
+ return 0;
+}
+"""
+
+FN_DATALOC_TEMPLATE = """
+int32_t
+metadisp_@NAME@ (call_frame_t *frame, xlator_t *this,
+ @LONG_ARGS@)
+{
+ METADISP_TRACE("@NAME@ dataloc");
+ loc_t backend_loc = {
+ 0,
+ };
+ if (build_backend_loc(loc->gfid, loc, &backend_loc)) {
+ goto unwind;
+ }
+ xlator_t *child = NULL;
+ child = DATA_CHILD(this);
+ STACK_WIND (frame, default_@NAME@_cbk,
+ child, child->fops->@NAME@,
+ @SHORT_ARGS@);
+ return 0;
+
+unwind:
+ STACK_UNWIND_STRICT(lookup, frame, -1, EINVAL, NULL, NULL, NULL, NULL);
+ return 0;
+}
+"""
+
+FOPS_LINE_TEMPLATE = "\t.@NAME@ = metadisp_@NAME@,"
+
+skipped = [
+ "readdir",
+ "readdirp",
+ "lookup",
+ "fsync",
+ "stat",
+ "open",
+ "create",
+ "unlink",
+ "setattr",
+ # TODO: implement "inodelk",
+]
+
+
+def gen_fops():
+ done = skipped
+
+ #
+ # these are fops that wind to the DATA_CHILD
+ #
+ # NOTE: re-written in order from google doc:
+ # https://docs.google.com/document/d/1KEwVtSNvDhs4qb63gWx2ulCp5GJjge77NGJk4p_Ms4Q
+ for name in [
+ "writev",
+ "readv",
+ "ftruncate",
+ "zerofill",
+ "discard",
+ "seek",
+ "fstat",
+ ]:
+ done = done + [name]
+ print(generate(FN_DATAFD_TEMPLATE, name, fop_subs))
+
+ for name in ["truncate"]:
+ done = done + [name]
+ print(generate(FN_DATALOC_TEMPLATE, name, fop_subs))
+
+ # these are fops that operate solely on dentries, folders,
+ # or extended attributes. Therefore, they must always
+ # wind to METADATA_CHILD and should never perform
+ # any path rewriting
+ #
+ # NOTE: re-written in order from google doc:
+ # https://docs.google.com/document/d/1KEwVtSNvDhs4qb63gWx2ulCp5GJjge77NGJk4p_Ms4Q
+ for name in [
+ "mkdir",
+ "symlink",
+ "link",
+ "rename",
+ "mknod",
+ "opendir",
+ # "readdir, # special-cased
+ # "readdirp, # special-cased
+ "fsyncdir",
+ # "setattr", # special-cased
+ "readlink",
+ "fentrylk",
+ "access",
+ # TODO: these wind to both,
+ # data for backend-attributes and metadata for the rest
+ "xattrop",
+ "setxattr",
+ "getxattr",
+ "removexattr",
+ "fgetxattr",
+ "fsetxattr",
+ "fremovexattr",
+ ]:
+
+ done = done + [name]
+ print(generate(FN_METADATA_CHILD_GENERIC, name, fop_subs))
+
+ print("struct xlator_fops fops = {")
+ for name in done:
+ print(generate(FOPS_LINE_TEMPLATE, name, fop_subs))
+
+ print("};")
+
+
+for l in open(sys.argv[1], "r").readlines():
+ if l.find("#pragma generate") != -1:
+ print("/* BEGIN GENERATED CODE - DO NOT MODIFY */")
+ gen_fops()
+ print("/* END GENERATED CODE */")
+ else:
+ print(l[:-1])
diff --git a/xlators/features/metadisp/src/metadisp-create.c b/xlators/features/metadisp/src/metadisp-create.c
new file mode 100644
index 00000000000..f8c9798dd59
--- /dev/null
+++ b/xlators/features/metadisp/src/metadisp-create.c
@@ -0,0 +1,101 @@
+#include "metadisp.h"
+#include <glusterfs/call-stub.h>
+
+/**
+ * Create, like stat, is a two-step process. We send a create
+ * to the METADATA_CHILD, then send another create to the DATA_CHILD.
+ *
+ * We do the metadata child first to ensure that the ACLs are enforced.
+ */
+
+int32_t
+metadisp_create_dentry_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd,
+ inode_t *inode, struct iatt *buf,
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
+{
+ STACK_UNWIND_STRICT(create, frame, op_ret, op_errno, fd, inode, buf,
+ preparent, postparent, xdata);
+ return 0;
+}
+
+int32_t
+metadisp_create_resume(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ int32_t flags, mode_t mode, mode_t umask, fd_t *fd,
+ dict_t *xdata)
+{
+ // create the backend data inode
+ STACK_WIND(frame, metadisp_create_dentry_cbk, DATA_CHILD(this),
+ DATA_CHILD(this)->fops->create, loc, flags, mode, umask, fd,
+ xdata);
+ return 0;
+}
+
+int32_t
+metadisp_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ METADISP_TRACE("%d %d", op_ret, op_errno);
+ call_stub_t *stub = cookie;
+ if (op_ret != 0) {
+ STACK_UNWIND_STRICT(create, frame, op_ret, op_errno, fd, inode, buf,
+ preparent, postparent, xdata);
+ return 0;
+ }
+
+ if (stub == NULL) {
+ goto unwind;
+ }
+
+ if (stub->poison) {
+ call_stub_destroy(stub);
+ return 0;
+ }
+
+ call_resume(stub);
+ return 0;
+
+unwind:
+ STACK_UNWIND_STRICT(create, frame, -1, EINVAL, NULL, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
+}
+
+int32_t
+metadisp_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+{
+ METADISP_TRACE(".");
+
+ loc_t backend_loc = {
+ 0,
+ };
+ call_stub_t *stub = NULL;
+ uuid_t *gfid_req = NULL;
+
+ RESOLVE_GFID_REQ(xdata, gfid_req, out);
+
+ if (build_backend_loc(*gfid_req, loc, &backend_loc)) {
+ goto unwind;
+ }
+
+ frame->local = loc;
+
+ stub = fop_create_stub(frame, metadisp_create_resume, &backend_loc, flags,
+ mode, umask, fd, xdata);
+
+ STACK_WIND_COOKIE(frame, metadisp_create_cbk, stub, METADATA_CHILD(this),
+ METADATA_CHILD(this)->fops->create, loc, flags, mode,
+ umask, fd, xdata);
+ return 0;
+
+unwind:
+ STACK_UNWIND_STRICT(create, frame, -1, EINVAL, NULL, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
+out:
+ return -1;
+}
diff --git a/xlators/features/metadisp/src/metadisp-fops.h b/xlators/features/metadisp/src/metadisp-fops.h
new file mode 100644
index 00000000000..56dd427cf34
--- /dev/null
+++ b/xlators/features/metadisp/src/metadisp-fops.h
@@ -0,0 +1,51 @@
+#ifndef GF_METADISP_FOPS_H_
+#define GF_METADISP_FOPS_H_
+
+#include <glusterfs/xlator.h>
+#include <glusterfs/dict.h>
+#include <glusterfs/glusterfs.h>
+
+#include <sys/types.h>
+
+/* fops in here are defined in their own file. Every other fop is just defined
+ * inline of fops.c */
+
+int
+metadisp_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *xdata);
+
+int
+metadisp_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *dict);
+
+int
+metadisp_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata);
+
+int
+metadisp_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata);
+
+int
+metadisp_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata);
+
+int
+metadisp_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata);
+
+int
+metadisp_inodelk(call_frame_t *frame, xlator_t *this, const char *volume,
+ loc_t *loc, int32_t cmd, struct gf_flock *lock, dict_t *xdata);
+
+int
+metadisp_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+ dict_t *xdata);
+
+int
+metadisp_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata);
+
+int
+metadisp_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata);
+
+#endif
diff --git a/xlators/features/metadisp/src/metadisp-fsync.c b/xlators/features/metadisp/src/metadisp-fsync.c
new file mode 100644
index 00000000000..2e46fa84eac
--- /dev/null
+++ b/xlators/features/metadisp/src/metadisp-fsync.c
@@ -0,0 +1,54 @@
+
+#include "metadisp.h"
+#include <glusterfs/call-stub.h>
+
+int32_t
+metadisp_fsync_resume(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ int32_t flags, dict_t *xdata)
+{
+ STACK_WIND(frame, default_fsync_cbk, DATA_CHILD(this),
+ DATA_CHILD(this)->fops->fsync, fd, flags, xdata);
+ return 0;
+}
+
+int32_t
+metadisp_fsync_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ call_stub_t *stub = NULL;
+ if (cookie) {
+ stub = cookie;
+ }
+
+ if (op_ret != 0) {
+ goto unwind;
+ }
+
+ if (stub->poison) {
+ call_stub_destroy(stub);
+ stub = NULL;
+ return 0;
+ }
+
+ call_resume(stub);
+ return 0;
+
+unwind:
+ if (stub) {
+ call_stub_destroy(stub);
+ }
+ STACK_UNWIND_STRICT(fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+ return 0;
+}
+
+int32_t
+metadisp_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+ dict_t *xdata)
+{
+ call_stub_t *stub = NULL;
+ stub = fop_fsync_stub(frame, metadisp_fsync_resume, fd, flags, xdata);
+ STACK_WIND_COOKIE(frame, metadisp_fsync_cbk, stub, METADATA_CHILD(this),
+ METADATA_CHILD(this)->fops->fsync, fd, flags, xdata);
+ return 0;
+}
diff --git a/xlators/features/metadisp/src/metadisp-lookup.c b/xlators/features/metadisp/src/metadisp-lookup.c
new file mode 100644
index 00000000000..27d90c9f746
--- /dev/null
+++ b/xlators/features/metadisp/src/metadisp-lookup.c
@@ -0,0 +1,90 @@
+#include "metadisp.h"
+#include <glusterfs/call-stub.h>
+
+/**
+ * Lookup, like stat, is a two-step process for grabbing the metadata details
+ * as well as the data details.
+ */
+
+int32_t
+metadisp_backend_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
+{
+ METADISP_TRACE("backend_lookup_cbk");
+ if (op_errno == ENOENT) {
+ op_errno = ENODATA;
+ op_ret = -1;
+ }
+ STACK_UNWIND_STRICT(lookup, frame, op_ret, op_errno, inode, buf, xdata,
+ postparent);
+ return 0;
+}
+
+int32_t
+metadisp_backend_lookup_resume(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ dict_t *xdata)
+{
+ METADISP_TRACE("backend_lookup_resume");
+ loc_t backend_loc = {
+ 0,
+ };
+ if (build_backend_loc(loc->gfid, loc, &backend_loc)) {
+ goto unwind;
+ }
+
+ STACK_WIND(frame, metadisp_backend_lookup_cbk, DATA_CHILD(this),
+ DATA_CHILD(this)->fops->lookup, &backend_loc, xdata);
+ return 0;
+
+unwind:
+ STACK_UNWIND_STRICT(lookup, frame, -1, EINVAL, NULL, NULL, NULL, NULL);
+ return 0;
+}
+
+int32_t
+metadisp_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+{
+ METADISP_TRACE("%d %d", op_ret, op_errno);
+ call_stub_t *stub = NULL;
+ stub = cookie;
+
+ if (op_ret != 0) {
+ goto unwind;
+ }
+
+ if (!IA_ISREG(buf->ia_type)) {
+ goto unwind;
+ } else if (!stub) {
+ op_errno = EINVAL;
+ goto unwind;
+ }
+
+ METADISP_TRACE("resuming stub");
+
+ // memcpy(stub->args.loc.gfid, buf->ia_gfid, sizeof(uuid_t));
+ call_resume(stub);
+ return 0;
+unwind:
+ METADISP_TRACE("unwinding %d %d", op_ret, op_errno);
+ STACK_UNWIND_STRICT(lookup, frame, op_ret, op_errno, inode, buf, xdata,
+ postparent);
+ if (stub) {
+ call_stub_destroy(stub);
+ }
+ return 0;
+}
+
+int32_t
+metadisp_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+ METADISP_TRACE("lookup");
+ call_stub_t *stub = NULL;
+ stub = fop_lookup_stub(frame, metadisp_backend_lookup_resume, loc, xdata);
+ STACK_WIND_COOKIE(frame, metadisp_lookup_cbk, stub, METADATA_CHILD(this),
+ METADATA_CHILD(this)->fops->lookup, loc, xdata);
+ return 0;
+}
diff --git a/xlators/features/metadisp/src/metadisp-open.c b/xlators/features/metadisp/src/metadisp-open.c
new file mode 100644
index 00000000000..64814afe636
--- /dev/null
+++ b/xlators/features/metadisp/src/metadisp-open.c
@@ -0,0 +1,70 @@
+#include <glusterfs/call-stub.h>
+#include "metadisp.h"
+
+int32_t
+metadisp_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+ METADISP_TRACE("got open results %d %d", op_ret, op_errno);
+
+ call_stub_t *stub = NULL;
+ if (cookie) {
+ stub = cookie;
+ }
+
+ if (op_ret != 0) {
+ goto unwind;
+ }
+
+ if (!stub) {
+ goto unwind;
+ }
+
+ if (stub->poison) {
+ call_stub_destroy(stub);
+ stub = NULL;
+ return 0;
+ }
+
+ call_resume(stub);
+ return 0;
+
+unwind:
+ if (stub) {
+ call_stub_destroy(stub);
+ }
+ STACK_UNWIND_STRICT(open, frame, op_ret, op_errno, fd, xdata);
+ return 0;
+}
+
+int32_t
+metadisp_open_resume(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ int32_t flags, fd_t *fd, dict_t *xdata)
+{
+ STACK_WIND_COOKIE(frame, metadisp_open_cbk, NULL, DATA_CHILD(this),
+ DATA_CHILD(this)->fops->open, loc, flags, fd, xdata);
+ return 0;
+}
+
+int32_t
+metadisp_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
+{
+ call_stub_t *stub = NULL;
+ loc_t backend_loc = {
+ 0,
+ };
+
+ if (build_backend_loc(loc->gfid, loc, &backend_loc)) {
+ goto unwind;
+ }
+
+ stub = fop_open_stub(frame, metadisp_open_resume, &backend_loc, flags, fd,
+ xdata);
+ STACK_WIND_COOKIE(frame, metadisp_open_cbk, stub, METADATA_CHILD(this),
+ METADATA_CHILD(this)->fops->open, loc, flags, fd, xdata);
+ return 0;
+unwind:
+ STACK_UNWIND_STRICT(open, frame, -1, EINVAL, NULL, NULL);
+ return 0;
+}
diff --git a/xlators/features/metadisp/src/metadisp-readdir.c b/xlators/features/metadisp/src/metadisp-readdir.c
new file mode 100644
index 00000000000..5f840b1e88f
--- /dev/null
+++ b/xlators/features/metadisp/src/metadisp-readdir.c
@@ -0,0 +1,65 @@
+#include "metadisp.h"
+
+/**
+ * With a change to the posix xlator, readdir and readdirp are shockingly
+ * simple.
+ *
+ * The issue with separating the backend data of the files
+ * with the metadata is that readdirs must now read from multiple sources
+ * to coalesce the directory entries.
+ *
+ * The way we do this is to tell the METADATA_CHILD that when it's
+ * running readdirp, each file entry should have a stat wound to
+ * 'stat-source-of-truth'.
+ *
+ * see metadisp_stat for how it handles winds _from_posix.
+ */
+
+int32_t
+metadisp_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *xdata)
+{
+ METADISP_TRACE(".");
+ /*
+ * Always use readdirp, even if the original was readdir. Why? Because NFS.
+ * There are multiple translations between Gluster, UNIX, and NFS stat
+ * structures in that path. One of them uses the type etc. from the stat
+ * structure, which is only filled in by readdirp. If we use readdir, the
+ * entries do actually go all the way back to the client and are visible in
+ * getdents, but then the readdir throws them away because of the
+ * uninitialized type.
+ */
+ GF_UNUSED int32_t ret;
+ if (!xdata) {
+ xdata = dict_new();
+ }
+
+ // ret = dict_set_int32 (xdata, "list-xattr", 1);
+
+ // I'm my own source of truth!
+ ret = dict_set_static_ptr(xdata, "stat-source-of-truth", (void *)this);
+
+ STACK_WIND(frame, default_readdirp_cbk, METADATA_CHILD(this),
+ METADATA_CHILD(this)->fops->readdirp, fd, size, off, xdata);
+
+ return 0;
+}
+
+int32_t
+metadisp_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *xdata)
+{
+ METADISP_TRACE(".");
+ if (!xdata) {
+ xdata = dict_new();
+ }
+ GF_UNUSED int32_t ret;
+ // ret = dict_set_int32 (xdata, "list-xattr", 1);
+
+ // I'm my own source of truth!
+ ret = dict_set_static_ptr(xdata, "stat-source-of-truth", (void *)this);
+
+ STACK_WIND(frame, default_readdirp_cbk, METADATA_CHILD(this),
+ METADATA_CHILD(this)->fops->readdirp, fd, size, off, xdata);
+ return 0;
+}
diff --git a/xlators/features/metadisp/src/metadisp-setattr.c b/xlators/features/metadisp/src/metadisp-setattr.c
new file mode 100644
index 00000000000..6991cf644f3
--- /dev/null
+++ b/xlators/features/metadisp/src/metadisp-setattr.c
@@ -0,0 +1,90 @@
+#include "metadisp.h"
+#include <glusterfs/call-stub.h>
+
+int32_t
+metadisp_backend_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *statpre, struct iatt *statpost,
+ dict_t *xdata)
+
+{
+ METADISP_TRACE("backend_setattr_cbk");
+ if (op_errno == ENOENT) {
+ op_errno = ENODATA;
+ op_ret = -1;
+ }
+ STACK_UNWIND_STRICT(setattr, frame, op_ret, op_errno, statpre, statpost,
+ xdata);
+ return 0;
+}
+
+int32_t
+metadisp_backend_setattr_resume(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct iatt *stbuf, int32_t valid,
+ dict_t *xdata)
+
+{
+ METADISP_TRACE("backend_setattr_resume");
+ loc_t backend_loc = {
+ 0,
+ };
+ if (build_backend_loc(loc->gfid, loc, &backend_loc)) {
+ goto unwind;
+ }
+
+ STACK_WIND(frame, metadisp_backend_setattr_cbk, DATA_CHILD(this),
+ DATA_CHILD(this)->fops->setattr, &backend_loc, stbuf, valid,
+ xdata);
+ return 0;
+
+unwind:
+ STACK_UNWIND_STRICT(setattr, frame, -1, EINVAL, NULL, NULL, NULL);
+ return 0;
+}
+
+int32_t
+metadisp_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
+{
+ METADISP_TRACE("%d %d", op_ret, op_errno);
+ call_stub_t *stub = NULL;
+ stub = cookie;
+
+ if (op_ret != 0) {
+ goto unwind;
+ }
+
+ if (!IA_ISREG(statpost->ia_type)) {
+ goto unwind;
+ } else if (!stub) {
+ op_errno = EINVAL;
+ goto unwind;
+ }
+
+ METADISP_TRACE("resuming stub");
+ call_resume(stub);
+ return 0;
+unwind:
+ METADISP_TRACE("unwinding %d %d", op_ret, op_errno);
+ STACK_UNWIND_STRICT(setattr, frame, op_ret, op_errno, statpre, statpost,
+ xdata);
+ if (stub) {
+ call_stub_destroy(stub);
+ }
+ return 0;
+}
+
+int32_t
+metadisp_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
+{
+ METADISP_TRACE("setattr");
+ call_stub_t *stub = NULL;
+ stub = fop_setattr_stub(frame, metadisp_backend_setattr_resume, loc, stbuf,
+ valid, xdata);
+ STACK_WIND_COOKIE(frame, metadisp_setattr_cbk, stub, METADATA_CHILD(this),
+ METADATA_CHILD(this)->fops->setattr, loc, stbuf, valid,
+ xdata);
+ return 0;
+}
diff --git a/xlators/features/metadisp/src/metadisp-stat.c b/xlators/features/metadisp/src/metadisp-stat.c
new file mode 100644
index 00000000000..b06d0dbcddd
--- /dev/null
+++ b/xlators/features/metadisp/src/metadisp-stat.c
@@ -0,0 +1,124 @@
+#include "metadisp.h"
+#include <glusterfs/call-stub.h>
+
+/**
+ * The stat flow in METADISP is complicated because we must
+ * do ensure a few things:
+ * 1. stat, on the path within the metadata layer,
+ * MUST get the backend FD of the data layer.
+ * --- we wind to the metadata layer, then the data layer.
+ *
+ * 2. the metadata layer MUST be able to ask the data
+ * layer for stat information.
+ * --- this is 'syncop-internal-from-posix'
+ *
+ * 3. when the metadata exists BUT the data is missing,
+ * we MUST mark the backend file as bad and heal it.
+ */
+
+int32_t
+metadisp_stat_backend_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
+{
+ METADISP_TRACE("got backend stat results %d %d", op_ret, op_errno);
+ if (op_errno == ENOENT) {
+ STACK_UNWIND_STRICT(open, frame, -1, ENODATA, NULL, NULL);
+ return 0;
+ }
+ STACK_UNWIND_STRICT(stat, frame, op_ret, op_errno, buf, xdata);
+ return 0;
+}
+
+int32_t
+metadisp_stat_resume(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ dict_t *xdata)
+{
+ METADISP_TRACE("winding stat to path %s", loc->path);
+ if (gf_uuid_is_null(loc->gfid)) {
+ METADISP_TRACE("bad object, sending EUCLEAN");
+ STACK_UNWIND_STRICT(open, frame, -1, EUCLEAN, NULL, NULL);
+ return 0;
+ }
+
+ STACK_WIND(frame, metadisp_stat_backend_cbk, SECOND_CHILD(this),
+ SECOND_CHILD(this)->fops->stat, loc, xdata);
+ return 0;
+}
+
+int32_t
+metadisp_stat_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
+{
+ call_stub_t *stub = NULL;
+
+ METADISP_TRACE("got stat results %d %d", op_ret, op_errno);
+
+ if (cookie) {
+ stub = cookie;
+ }
+
+ if (op_ret != 0) {
+ goto unwind;
+ }
+
+ // only use the stub for the files
+ if (!IA_ISREG(buf->ia_type)) {
+ goto unwind;
+ }
+
+ if (stub->poison) {
+ call_stub_destroy(stub);
+ stub = NULL;
+ return 0;
+ }
+
+ call_resume(stub);
+ return 0;
+
+unwind:
+ if (stub) {
+ call_stub_destroy(stub);
+ }
+ STACK_UNWIND_STRICT(stat, frame, op_ret, op_errno, buf, xdata);
+ return 0;
+}
+
+int32_t
+metadisp_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+ call_stub_t *stub = NULL;
+ int32_t ret = 0;
+ loc_t backend_loc = {
+ 0,
+ };
+ METADISP_FILTER_ROOT(stat, loc, xdata);
+
+ if (build_backend_loc(loc->gfid, loc, &backend_loc)) {
+ goto unwind;
+ }
+
+ if (dict_get_int32(xdata, "syncop-internal-from-posix", &ret) == 0) {
+ // if we've just been sent a stat from posix, then we know
+ // that we must send down a stat for a file to the second child.
+ //
+ // that means we can skip the stat for the first child and just
+ // send to the data disk.
+ METADISP_TRACE("got syncop-internal-from-posix");
+ STACK_WIND(frame, default_stat_cbk, DATA_CHILD(this),
+ DATA_CHILD(this)->fops->stat, &backend_loc, xdata);
+ return 0;
+ }
+
+ // we do not know if the request is for a file, folder, etc. wind
+ // to first child to find out.
+ stub = fop_stat_stub(frame, metadisp_stat_resume, &backend_loc, xdata);
+ METADISP_TRACE("winding stat to first child %s", loc->path);
+ STACK_WIND_COOKIE(frame, metadisp_stat_cbk, stub, METADATA_CHILD(this),
+ METADATA_CHILD(this)->fops->stat, loc, xdata);
+ return 0;
+unwind:
+ STACK_UNWIND_STRICT(stat, frame, -1, EINVAL, NULL, NULL);
+ return 0;
+}
diff --git a/xlators/features/metadisp/src/metadisp-unlink.c b/xlators/features/metadisp/src/metadisp-unlink.c
new file mode 100644
index 00000000000..1f6a8eb35ce
--- /dev/null
+++ b/xlators/features/metadisp/src/metadisp-unlink.c
@@ -0,0 +1,160 @@
+
+#include "metadisp.h"
+#include <glusterfs/call-stub.h>
+
+/**
+ * The unlink flow in metadisp is complicated because we must
+ * do ensure that UNLINK causes both the metadata objects
+ * to get removed and the data objects to get removed.
+ */
+
+int32_t
+metadisp_unlink_resume(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ int xflag, dict_t *xdata)
+{
+ METADISP_TRACE("winding backend unlink to path %s", loc->path);
+ STACK_WIND(frame, default_unlink_cbk, DATA_CHILD(this),
+ DATA_CHILD(this)->fops->unlink, loc, xflag, xdata);
+ return 0;
+}
+
+int32_t
+metadisp_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ METADISP_TRACE(". %d %d", op_ret, op_errno);
+
+ int ret = 0;
+ call_stub_t *stub = NULL;
+ int nlink = 0;
+
+ if (cookie) {
+ stub = cookie;
+ }
+
+ if (op_ret != 0) {
+ goto unwind;
+ }
+
+ if (stub->poison) {
+ call_stub_destroy(stub);
+ stub = NULL;
+ return 0;
+ }
+
+ ret = dict_get_uint32(xdata, GF_RESPONSE_LINK_COUNT_XDATA, &nlink);
+ if (ret != 0) {
+ op_errno = EINVAL;
+ op_ret = -1;
+ goto unwind;
+ }
+ METADISP_TRACE("frontend hardlink count %d %d", ret, nlink);
+ if (nlink > 1) {
+ goto unwind;
+ }
+
+ call_resume(stub);
+ return 0;
+
+unwind:
+ if (stub) {
+ call_stub_destroy(stub);
+ }
+ STACK_UNWIND_STRICT(unlink, frame, op_ret, op_errno, preparent, postparent,
+ xdata);
+ return 0;
+}
+
+int32_t
+metadisp_unlink_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
+{
+ call_stub_t *stub = NULL;
+
+ if (cookie) {
+ stub = cookie;
+ }
+
+ if (op_ret != 0) {
+ goto unwind;
+ }
+
+ // fail fast on empty gfid so we don't loop forever
+ if (gf_uuid_is_null(buf->ia_gfid)) {
+ op_ret = -1;
+ op_errno = ENODATA;
+ goto unwind;
+ }
+
+ // fill gfid since the stub is incomplete
+ memcpy(stub->args.loc.gfid, buf->ia_gfid, sizeof(uuid_t));
+ memcpy(stub->args.loc.pargfid, postparent->ia_gfid, sizeof(uuid_t));
+
+ if (stub->poison) {
+ call_stub_destroy(stub);
+ stub = NULL;
+ return 0;
+ }
+
+ call_resume(stub);
+ return 0;
+
+unwind:
+ if (stub) {
+ call_stub_destroy(stub);
+ }
+ STACK_UNWIND_STRICT(unlink, frame, op_ret, op_errno, NULL, NULL, NULL);
+ return 0;
+}
+
+int32_t
+metadisp_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
+{
+ call_stub_t *stub = NULL;
+ loc_t backend_loc = {
+ 0,
+ };
+
+ if (gf_uuid_is_null(loc->gfid)) {
+ METADISP_TRACE("winding lookup for unlink to path %s", loc->path);
+
+ // loop back to ourselves after a lookup
+ stub = fop_unlink_stub(frame, metadisp_unlink, loc, xflag, xdata);
+ STACK_WIND_COOKIE(frame, metadisp_unlink_lookup_cbk, stub,
+ METADATA_CHILD(this),
+ METADATA_CHILD(this)->fops->lookup, loc, xdata);
+ return 0;
+ }
+
+ if (build_backend_loc(loc->gfid, loc, &backend_loc)) {
+ goto unwind;
+ }
+
+ //
+ // ensure we get the link count on the unlink response, so we can
+ // account for hardlinks before winding to the backend.
+ // NOTE:
+ // multiple xlators use GF_REQUEST_LINK_COUNT_XDATA. confirmation
+ // is needed to ensure that multiple requests will work in the same
+ // xlator stack.
+ //
+ if (!xdata) {
+ xdata = dict_new();
+ }
+ dict_set_int32(xdata, GF_REQUEST_LINK_COUNT_XDATA, 1);
+
+ METADISP_TRACE("winding frontend unlink to path %s", loc->path);
+ stub = fop_unlink_stub(frame, metadisp_unlink_resume, &backend_loc, xflag,
+ xdata);
+
+ STACK_WIND_COOKIE(frame, metadisp_unlink_cbk, stub, METADATA_CHILD(this),
+ METADATA_CHILD(this)->fops->unlink, loc, xflag, xdata);
+ return 0;
+unwind:
+ STACK_UNWIND_STRICT(unlink, frame, -1, EINVAL, NULL, NULL, NULL);
+ return 0;
+}
diff --git a/xlators/features/metadisp/src/metadisp.c b/xlators/features/metadisp/src/metadisp.c
new file mode 100644
index 00000000000..3c8f150cebc
--- /dev/null
+++ b/xlators/features/metadisp/src/metadisp.c
@@ -0,0 +1,46 @@
+#include <glusterfs/call-stub.h>
+
+#include "metadisp.h"
+#include "metadisp-fops.h"
+
+int32_t
+init(xlator_t *this)
+{
+ if (!this->children) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "not configured with children. exiting");
+ return -1;
+ }
+
+ if (!this->parents) {
+ gf_log(this->name, GF_LOG_WARNING, "dangling volume. check volfile ");
+ }
+
+ return 0;
+}
+
+void
+fini(xlator_t *this)
+{
+ return;
+}
+
+/* defined in fops.c */
+struct xlator_fops fops;
+
+struct xlator_cbks cbks = {};
+
+struct volume_options options[] = {
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .op_version = {1},
+ .identifier = "metadisp",
+ .category = GF_EXPERIMENTAL,
+};
diff --git a/xlators/features/metadisp/src/metadisp.h b/xlators/features/metadisp/src/metadisp.h
new file mode 100644
index 00000000000..c8fd7a13c04
--- /dev/null
+++ b/xlators/features/metadisp/src/metadisp.h
@@ -0,0 +1,45 @@
+/*
+ Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+*/
+#ifndef GF_METADISP_H_
+#define GF_METADISP_H_
+
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/dict.h>
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
+
+#define METADATA_CHILD(_this) FIRST_CHILD(_this)
+#define DATA_CHILD(_this) SECOND_CHILD(_this)
+
+int32_t
+build_backend_loc(uuid_t gfid, loc_t *src_loc, loc_t *dst_loc);
+
+#define METADISP_TRACE(_args...) gf_log("metadisp", GF_LOG_INFO, _args)
+
+#define METADISP_FILTER_ROOT(_op, _args...) \
+ if (strcmp(loc->path, "/") == 0) { \
+ STACK_WIND(frame, default_##_op##_cbk, METADATA_CHILD(this), \
+ METADATA_CHILD(this)->fops->_op, _args); \
+ return 0; \
+ }
+
+#define METADISP_FILTER_ROOT_BY_GFID(_op, _gfid, _args...) \
+ if (__is_root_gfid(_gfid)) { \
+ STACK_WIND(frame, default_##_op##_cbk, METADATA_CHILD(this), \
+ METADATA_CHILD(this)->fops->_op, _args); \
+ return 0; \
+ }
+
+#define RESOLVE_GFID_REQ(_dict, _dest, _lbl) \
+ VALIDATE_OR_GOTO(dict_get_ptr(_dict, "gfid-req", (void **)&_dest) == 0, \
+ _lbl)
+
+#endif /* __TEMPLATE_H__ */
diff --git a/xlators/features/namespace/src/namespace.c b/xlators/features/namespace/src/namespace.c
index cabd247da0c..86c5ebee900 100644
--- a/xlators/features/namespace/src/namespace.c
+++ b/xlators/features/namespace/src/namespace.c
@@ -15,17 +15,16 @@
#include <sys/types.h>
-#include "defaults.h"
-#include "glusterfs.h"
-#include "hashfn.h"
-#include "logging.h"
+#include <glusterfs/defaults.h>
+#include <glusterfs/hashfn.h>
+#include <glusterfs/logging.h>
#include "namespace.h"
/* Return codes for common path parsing functions. */
enum _path_parse_result {
- PATH_PARSE_RESULT_NO_PATH = 0,
- PATH_PARSE_RESULT_FOUND = 1,
- PATH_PARSE_RESULT_IS_GFID = 2,
+ PATH_PARSE_RESULT_NO_PATH = 0,
+ PATH_PARSE_RESULT_FOUND = 1,
+ PATH_PARSE_RESULT_IS_GFID = 2,
};
typedef enum _path_parse_result path_parse_result_t;
@@ -33,48 +32,50 @@ typedef enum _path_parse_result path_parse_result_t;
/* Clean up an ns_local struct. Wipe a loc (its inode is ref'd, so we're good.)
*/
static inline void
-ns_local_cleanup (ns_local_t *local)
+ns_local_cleanup(ns_local_t *local)
{
- if (!local) {
- return;
- }
+ if (!local) {
+ return;
+ }
- loc_wipe (&local->loc);
- GF_FREE (local);
+ loc_wipe(&local->loc);
+ GF_FREE(local);
}
/* Create a new ns_local. We ref the inode, fake a new loc struct, and stash
* the stub given to us. */
static inline ns_local_t *
-ns_local_new (call_stub_t *stub, inode_t *inode)
+ns_local_new(call_stub_t *stub, inode_t *inode)
{
- ns_local_t *local = NULL;
- loc_t loc = {0, };
+ ns_local_t *local = NULL;
+ loc_t loc = {
+ 0,
+ };
- if (!stub || !inode) {
- goto out;
- }
+ if (!stub || !inode) {
+ goto out;
+ }
- local = GF_CALLOC (1, sizeof (ns_local_t), 0);
- if (local == NULL) {
- goto out;
- }
+ local = GF_CALLOC(1, sizeof(ns_local_t), 0);
+ if (local == NULL) {
+ goto out;
+ }
- /* Set up a fake loc_t struct to give to the getxattr call. */
- gf_uuid_copy (loc.gfid, inode->gfid);
- loc.inode = inode_ref (inode);
+ /* Set up a fake loc_t struct to give to the getxattr call. */
+ gf_uuid_copy(loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
- /* If for some reason inode_ref() fails, then just give up. */
- if (!loc.inode) {
- GF_FREE (local);
- goto out;
- }
+ /* If for some reason inode_ref() fails, then just give up. */
+ if (!loc.inode) {
+ GF_FREE(local);
+ goto out;
+ }
- local->stub = stub;
- local->loc = loc;
+ local->stub = stub;
+ local->loc = loc;
out:
- return local;
+ return local;
}
/* Try parsing a path string. If the path string is a GFID, then return
@@ -82,110 +83,109 @@ out:
* return PATH_PARSE_RESULT_NO_PATH and set the hash to 1. Otherwise, hash the
* namespace and store it in the info struct. */
static path_parse_result_t
-parse_path (ns_info_t *info, const char *path)
+parse_path(ns_info_t *info, const char *path)
{
- int len = 0;
- const char *ns_begin = path;
- const char *ns_end = NULL;
-
- if (!path || strlen (path) == 0) {
- return PATH_PARSE_RESULT_NO_PATH;
- }
-
- if (path[0] == '<') {
- return PATH_PARSE_RESULT_IS_GFID;
- }
-
- /* Right now we only want the top-level directory, so
- * skip the initial '/' and read until the next '/'. */
- while (*ns_begin == '/') {
- ns_begin++;
- }
-
- /* ns_end will point to the next '/' or NULL if there is no delimiting
- * '/' (i.e. "/directory" or the top level "/") */
- ns_end = strchr (ns_begin, '/');
- len = ns_end ? (ns_end - ns_begin) : strlen (ns_begin);
-
- if (len != 0) {
- info->hash = SuperFastHash (ns_begin, len);
- } else {
- /* If our substring is empty, then we can hash '/' instead.
- * '/' is used in the namespace config for the top-level
- * namespace. */
- info->hash = SuperFastHash ("/", 1);
- }
-
- info->found = _gf_true;
- return PATH_PARSE_RESULT_FOUND;
+ int len = 0;
+ const char *ns_begin = path;
+ const char *ns_end = NULL;
+
+ if (!path || strlen(path) == 0) {
+ return PATH_PARSE_RESULT_NO_PATH;
+ }
+
+ if (path[0] == '<') {
+ return PATH_PARSE_RESULT_IS_GFID;
+ }
+
+ /* Right now we only want the top-level directory, so
+ * skip the initial '/' and read until the next '/'. */
+ while (*ns_begin == '/') {
+ ns_begin++;
+ }
+
+ /* ns_end will point to the next '/' or NULL if there is no delimiting
+ * '/' (i.e. "/directory" or the top level "/") */
+ ns_end = strchr(ns_begin, '/');
+ len = ns_end ? (ns_end - ns_begin) : strlen(ns_begin);
+
+ if (len != 0) {
+ info->hash = SuperFastHash(ns_begin, len);
+ } else {
+ /* If our substring is empty, then we can hash '/' instead.
+ * '/' is used in the namespace config for the top-level
+ * namespace. */
+ info->hash = SuperFastHash("/", 1);
+ }
+
+ info->found = _gf_true;
+ return PATH_PARSE_RESULT_FOUND;
}
/* Cache namespace info stored in the stack (info) into the inode. */
static int
-ns_inode_ctx_put (inode_t *inode, xlator_t *this, ns_info_t *info)
+ns_inode_ctx_put(inode_t *inode, xlator_t *this, ns_info_t *info)
{
- ns_info_t *cached_ns_info = NULL;
- uint64_t ns_as_64 = 0;
- int ret = -1;
-
- if (!inode || !this) {
- gf_log (this->name, GF_LOG_WARNING,
- "Need a valid inode and xlator to cache ns_info.");
- ret = -1;
- goto out;
- }
+ ns_info_t *cached_ns_info = NULL;
+ uint64_t ns_as_64 = 0;
+ int ret = -1;
- cached_ns_info = GF_CALLOC (1, sizeof (ns_info_t), 0);
+ if (!inode || !this) {
+ gf_log(this ? this->name : "namespace", GF_LOG_WARNING,
+ "Need a valid inode and xlator to cache ns_info.");
+ ret = -1;
+ goto out;
+ }
- /* If we've run out of memory, then return ENOMEM. */
- if (cached_ns_info == NULL) {
- gf_log (this->name, GF_LOG_WARNING,
- "No memory to cache ns_info.");
- ret = -(ENOMEM);
- goto out;
- }
+ cached_ns_info = GF_CALLOC(1, sizeof(ns_info_t), 0);
- *cached_ns_info = *info;
- ns_as_64 = (uint64_t)cached_ns_info;
+ /* If we've run out of memory, then return ENOMEM. */
+ if (cached_ns_info == NULL) {
+ gf_log(this->name, GF_LOG_WARNING, "No memory to cache ns_info.");
+ ret = -(ENOMEM);
+ goto out;
+ }
- ret = inode_ctx_put (inode, this, ns_as_64);
+ *cached_ns_info = *info;
+ ns_as_64 = (uint64_t)(uintptr_t)cached_ns_info;
- if (ret) {
- goto out;
- }
+ ret = inode_ctx_put(inode, this, ns_as_64);
+
+ if (ret) {
+ goto out;
+ }
- ret = 0;
+ ret = 0;
out:
- if (ret && cached_ns_info) {
- GF_FREE (cached_ns_info);
- }
+ if (ret && cached_ns_info) {
+ GF_FREE(cached_ns_info);
+ }
- return ret;
+ return ret;
}
/* Retrieve namespace info cached in the inode into the stack for use in later
* translators. */
static int
-ns_inode_ctx_get (inode_t *inode, xlator_t *this, ns_info_t *info)
+ns_inode_ctx_get(inode_t *inode, xlator_t *this, ns_info_t *info)
{
- ns_info_t *cached_ns_info = NULL;
- uint64_t ns_as_64 = 0;
- int ret = -1;
+ ns_info_t *cached_ns_info = NULL;
+ uint64_t ns_as_64 = 0;
+ int ret = -1;
- if (!inode) {
- ret = -ENOENT;
- goto out;
- }
+ if (!inode) {
+ ret = -ENOENT;
+ goto out;
+ }
- ret = inode_ctx_get (inode, this, &ns_as_64);
+ ret = inode_ctx_get(inode, this, &ns_as_64);
- if (!ret) {
- cached_ns_info = (ns_info_t *)ns_as_64;
- *info = *cached_ns_info;
- }
+ if (!ret) {
+ cached_ns_info = (ns_info_t *)(uintptr_t)ns_as_64;
+ *info = *cached_ns_info;
+ }
out:
- return ret;
+ return ret;
}
/* This callback is the top of the unwind path of our attempt to get the path
@@ -194,76 +194,75 @@ out:
* default stub that we provide in the local, since there's nothing else to do
* once we've gotten the namespace hash. */
int32_t
-get_path_resume_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
+get_path_resume_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- path_parse_result_t ret = PATH_PARSE_RESULT_NO_PATH;
- call_frame_t *resume_frame = NULL;
- ns_local_t *local = NULL;
- call_stub_t *stub = NULL;
- ns_info_t *info = NULL;
- char *path = NULL;
-
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- local = frame->local;
-
- GF_VALIDATE_OR_GOTO (this->name, local, out);
- stub = local->stub;
-
- GF_VALIDATE_OR_GOTO (this->name, stub, out);
- /* Get the ns_info from the frame that we will eventually resume,
- * not the frame that we're going to destroy (frame). */
- resume_frame = stub->frame;
-
- GF_VALIDATE_OR_GOTO (this->name, resume_frame, out);
- GF_VALIDATE_OR_GOTO (this->name, resume_frame->root, out);
- info = &resume_frame->root->ns_info;
-
- GF_VALIDATE_OR_GOTO (this->name, dict, out);
-
- /* If we get a value back for the GET_ANCESTRY_PATH_KEY, then we
- * try to access it and parse it like a path. */
- if (!op_ret && !dict_get_str (dict, GET_ANCESTRY_PATH_KEY, &path)) {
- gf_log (this->name, GF_LOG_DEBUG, "G>P %s retrieved path %s",
- uuid_utoa (local->loc.gfid), path);
- /* Now let's parse a path, finally. */
- ret = parse_path (info, path);
- }
-
- if (ret == PATH_PARSE_RESULT_FOUND) {
- /* If we finally found namespace, then stash it. */
- ns_inode_ctx_put (local->loc.inode, this, info);
-
- gf_log (this->name, GF_LOG_DEBUG,
- "G>P %s %10u namespace found %s",
- uuid_utoa (local->loc.inode->gfid), info->hash, path);
- } else if (ret == PATH_PARSE_RESULT_NO_PATH) {
- gf_log (this->name, GF_LOG_WARNING, "G>P %s has no path",
- uuid_utoa (local->loc.inode->gfid));
- } else if (ret == PATH_PARSE_RESULT_IS_GFID) {
- gf_log (this->name, GF_LOG_WARNING,
- "G>P %s winding failed, still have gfid",
- uuid_utoa (local->loc.inode->gfid));
- }
+ path_parse_result_t ret = PATH_PARSE_RESULT_NO_PATH;
+ call_frame_t *resume_frame = NULL;
+ ns_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ ns_info_t *info = NULL;
+ char *path = NULL;
+
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ local = frame->local;
+
+ GF_VALIDATE_OR_GOTO(this->name, local, out);
+ stub = local->stub;
+
+ GF_VALIDATE_OR_GOTO(this->name, stub, out);
+ /* Get the ns_info from the frame that we will eventually resume,
+ * not the frame that we're going to destroy (frame). */
+ resume_frame = stub->frame;
+
+ GF_VALIDATE_OR_GOTO(this->name, resume_frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, resume_frame->root, out);
+ info = &resume_frame->root->ns_info;
+
+ GF_VALIDATE_OR_GOTO(this->name, dict, out);
+
+ /* If we get a value back for the GET_ANCESTRY_PATH_KEY, then we
+ * try to access it and parse it like a path. */
+ if (!op_ret && !dict_get_str(dict, GET_ANCESTRY_PATH_KEY, &path)) {
+ gf_log(this->name, GF_LOG_DEBUG, "G>P %s retrieved path %s",
+ uuid_utoa(local->loc.gfid), path);
+ /* Now let's parse a path, finally. */
+ ret = parse_path(info, path);
+ }
+
+ if (ret == PATH_PARSE_RESULT_FOUND) {
+ /* If we finally found namespace, then stash it. */
+ ns_inode_ctx_put(local->loc.inode, this, info);
+
+ gf_log(this->name, GF_LOG_DEBUG, "G>P %s %10u namespace found %s",
+ uuid_utoa(local->loc.inode->gfid), info->hash, path);
+ } else if (ret == PATH_PARSE_RESULT_NO_PATH) {
+ gf_log(this->name, GF_LOG_WARNING, "G>P %s has no path",
+ uuid_utoa(local->loc.inode->gfid));
+ } else if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "G>P %s winding failed, still have gfid",
+ uuid_utoa(local->loc.inode->gfid));
+ }
out:
- /* Make sure to clean up local finally. */
+ /* Make sure to clean up local finally. */
- if (frame) {
- frame->local = NULL;
- STACK_DESTROY (frame->root);
- }
+ if (frame) {
+ frame->local = NULL;
+ STACK_DESTROY(frame->root);
+ }
- if (local) {
- ns_local_cleanup (local);
- }
+ if (local) {
+ ns_local_cleanup(local);
+ }
- if (stub) {
- call_resume (stub);
- }
+ if (stub) {
+ call_resume(stub);
+ }
- return 0;
+ return 0;
}
/* This function tries first to set a namespace based on the information that
@@ -271,133 +270,129 @@ out:
* namespace in the inode, then trying to parse the path string in the `loc_t`
* struct. If this fails, then it will try to call inode_path. */
static path_parse_result_t
-set_ns_from_loc (const char *fn, call_frame_t *frame, xlator_t *this,
- loc_t *loc)
+set_ns_from_loc(const char *fn, call_frame_t *frame, xlator_t *this, loc_t *loc)
{
- path_parse_result_t ret = PATH_PARSE_RESULT_NO_PATH;
- ns_private_t *priv = (ns_private_t *)this->private;
- ns_info_t *info = &frame->root->ns_info;
- char *path = NULL;
+ path_parse_result_t ret = PATH_PARSE_RESULT_NO_PATH;
+ ns_private_t *priv = (ns_private_t *)this->private;
+ ns_info_t *info = &frame->root->ns_info;
+ char *path = NULL;
+
+ info->hash = 0;
+ info->found = _gf_false;
- info->hash = 0;
- info->found = _gf_false;
+ if (!priv->tag_namespaces) {
+ return ret;
+ }
+
+ /* This is our first pass at trying to get a path. Try getting
+ * from the inode context, then from the loc's path itself. */
+ if (!loc || !loc->path || !loc->inode) {
+ ret = PATH_PARSE_RESULT_NO_PATH;
+ } else if (!ns_inode_ctx_get(loc->inode, this, info)) {
+ ret = PATH_PARSE_RESULT_FOUND;
+ } else {
+ ret = parse_path(info, loc->path);
+ gf_log(this->name, GF_LOG_DEBUG, "%s: LOC retrieved path %s", fn,
+ loc->path);
- if (!priv->tag_namespaces) {
- return ret;
+ if (ret == PATH_PARSE_RESULT_FOUND) {
+ ns_inode_ctx_put(loc->inode, this, info);
}
+ }
- /* This is our first pass at trying to get a path. Try getting
- * from the inode context, then from the loc's path itself. */
- if (!loc || !loc->path || !loc->inode) {
- ret = PATH_PARSE_RESULT_NO_PATH;
- } else if (!ns_inode_ctx_get (loc->inode, this, info)) {
- ret = PATH_PARSE_RESULT_FOUND;
- } else {
- ret = parse_path (info, loc->path);
- gf_log (this->name, GF_LOG_DEBUG, "%s: LOC retrieved path %s",
- fn, loc->path);
-
- if (ret == PATH_PARSE_RESULT_FOUND) {
- ns_inode_ctx_put (loc->inode, this, info);
- }
+ /* Keep trying by calling inode_path next, making sure to copy
+ the loc's gfid into its inode if necessary. */
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ if (gf_uuid_is_null(loc->inode->gfid)) {
+ gf_uuid_copy(loc->inode->gfid, loc->gfid);
}
- /* Keep trying by calling inode_path next, making sure to copy
- the loc's gfid into its inode if necessary. */
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- if (gf_uuid_is_null (loc->inode->gfid)) {
- gf_uuid_copy (loc->inode->gfid, loc->gfid);
- }
-
- if (inode_path (loc->inode, NULL, &path) >= 0 && path) {
- ret = parse_path (info, loc->path);
- gf_log (this->name, GF_LOG_DEBUG,
- "%s: LOC retrieved path %s", fn, path);
-
- if (ret == PATH_PARSE_RESULT_FOUND) {
- ns_inode_ctx_put (loc->inode, this, info);
- }
- }
-
- if (path) {
- GF_FREE (path);
- }
+ if (inode_path(loc->inode, NULL, &path) >= 0 && path) {
+ ret = parse_path(info, loc->path);
+ gf_log(this->name, GF_LOG_DEBUG, "%s: LOC retrieved path %s", fn,
+ path);
+
+ if (ret == PATH_PARSE_RESULT_FOUND) {
+ ns_inode_ctx_put(loc->inode, this, info);
+ }
}
- /* Report our status, and if we have a GFID, we'll eventually try a
- * GET_ANCESTRY_PATH_KEY wind when we return from this function. */
- if (ret == PATH_PARSE_RESULT_FOUND) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s: LOC %s %10u namespace found for %s", fn,
- uuid_utoa (loc->inode->gfid), info->hash, loc->path);
- } else if (ret == PATH_PARSE_RESULT_NO_PATH) {
- gf_log (this->name, GF_LOG_WARNING, "%s: LOC has no path", fn);
- } else if (ret == PATH_PARSE_RESULT_IS_GFID) {
- /* Make sure to copy the inode's gfid for the eventual wind. */
- if (gf_uuid_is_null (loc->inode->gfid)) {
- gf_uuid_copy (loc->inode->gfid, loc->gfid);
- }
-
- gf_log (this->name, GF_LOG_DEBUG,
- "%s: LOC %s winding, looking for path", fn,
- uuid_utoa (loc->inode->gfid));
+ if (path) {
+ GF_FREE(path);
}
+ }
- return ret;
+ /* Report our status, and if we have a GFID, we'll eventually try a
+ * GET_ANCESTRY_PATH_KEY wind when we return from this function. */
+ if (ret == PATH_PARSE_RESULT_FOUND) {
+ gf_log(this->name, GF_LOG_DEBUG,
+ "%s: LOC %s %10u namespace found for %s", fn,
+ uuid_utoa(loc->inode->gfid), info->hash, loc->path);
+ } else if (ret == PATH_PARSE_RESULT_NO_PATH) {
+ gf_log(this->name, GF_LOG_WARNING, "%s: LOC has no path", fn);
+ } else if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ /* Make sure to copy the inode's gfid for the eventual wind. */
+ if (gf_uuid_is_null(loc->inode->gfid)) {
+ gf_uuid_copy(loc->inode->gfid, loc->gfid);
+ }
+
+ gf_log(this->name, GF_LOG_DEBUG, "%s: LOC %s winding, looking for path",
+ fn, uuid_utoa(loc->inode->gfid));
+ }
+
+ return ret;
}
/* This function tries first to set a namespace based on the information that
* it can retrieve from an `fd_t`. This includes first looking for a cached
* namespace in the inode, then trying to call inode_path manually. */
static path_parse_result_t
-set_ns_from_fd (const char *fn, call_frame_t *frame, xlator_t *this, fd_t *fd)
+set_ns_from_fd(const char *fn, call_frame_t *frame, xlator_t *this, fd_t *fd)
{
- path_parse_result_t ret = PATH_PARSE_RESULT_NO_PATH;
- ns_private_t *priv = (ns_private_t *)this->private;
- ns_info_t *info = &frame->root->ns_info;
- char *path = NULL;
-
- info->hash = 0;
- info->found = _gf_false;
+ path_parse_result_t ret = PATH_PARSE_RESULT_NO_PATH;
+ ns_private_t *priv = (ns_private_t *)this->private;
+ ns_info_t *info = &frame->root->ns_info;
+ char *path = NULL;
- if (!priv->tag_namespaces) {
- return ret;
- }
-
- /* This is our first pass at trying to get a path. Try getting
- * from the inode context, then inode_path. */
- if (!fd || !fd->inode) {
- ret = PATH_PARSE_RESULT_NO_PATH;
- } else if (!ns_inode_ctx_get (fd->inode, this, info)) {
- ret = PATH_PARSE_RESULT_FOUND;
- } else if (inode_path (fd->inode, NULL, &path) >= 0 && path) {
- ret = parse_path (info, path);
- gf_log (this->name, GF_LOG_DEBUG, "%s: FD retrieved path %s",
- fn, path);
-
- if (ret == PATH_PARSE_RESULT_FOUND) {
- ns_inode_ctx_put (fd->inode, this, info);
- }
- }
+ info->hash = 0;
+ info->found = _gf_false;
- if (path) {
- GF_FREE (path);
- }
+ if (!priv->tag_namespaces) {
+ return ret;
+ }
+
+ /* This is our first pass at trying to get a path. Try getting
+ * from the inode context, then inode_path. */
+ if (!fd || !fd->inode) {
+ ret = PATH_PARSE_RESULT_NO_PATH;
+ } else if (!ns_inode_ctx_get(fd->inode, this, info)) {
+ ret = PATH_PARSE_RESULT_FOUND;
+ } else if (inode_path(fd->inode, NULL, &path) >= 0 && path) {
+ ret = parse_path(info, path);
+ gf_log(this->name, GF_LOG_DEBUG, "%s: FD retrieved path %s", fn, path);
- /* Report our status, and if we have a GFID, we'll eventually try a
- * GET_ANCESTRY_PATH_KEY wind when we return from this function. */
if (ret == PATH_PARSE_RESULT_FOUND) {
- gf_log (this->name, GF_LOG_DEBUG, "%s: FD %s %10u namespace found",
- fn, uuid_utoa (fd->inode->gfid), info->hash);
- } else if (ret == PATH_PARSE_RESULT_NO_PATH) {
- gf_log (this->name, GF_LOG_WARNING, "%s: FD has no path", fn);
- } else if (ret == PATH_PARSE_RESULT_IS_GFID) {
- gf_log (this->name, GF_LOG_DEBUG,
- "%s: FD %s winding, looking for path",
- fn, uuid_utoa (fd->inode->gfid));
- }
-
- return ret;
+ ns_inode_ctx_put(fd->inode, this, info);
+ }
+ }
+
+ if (path) {
+ GF_FREE(path);
+ }
+
+ /* Report our status, and if we have a GFID, we'll eventually try a
+ * GET_ANCESTRY_PATH_KEY wind when we return from this function. */
+ if (ret == PATH_PARSE_RESULT_FOUND) {
+ gf_log(this->name, GF_LOG_DEBUG, "%s: FD %s %10u namespace found", fn,
+ uuid_utoa(fd->inode->gfid), info->hash);
+ } else if (ret == PATH_PARSE_RESULT_NO_PATH) {
+ gf_log(this->name, GF_LOG_WARNING, "%s: FD has no path", fn);
+ } else if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ gf_log(this->name, GF_LOG_DEBUG, "%s: FD %s winding, looking for path",
+ fn, uuid_utoa(fd->inode->gfid));
+ }
+
+ return ret;
}
/* This macro does the work of winding down a call of `getxattr` in the case
@@ -405,963 +400,945 @@ set_ns_from_fd (const char *fn, call_frame_t *frame, xlator_t *this, fd_t *fd)
* called `wind` and the existence of several basic variables (frame, this),
* but otherwise is general enough for any fop (fd- or loc-based.) */
#define GET_ANCESTRY_PATH_WIND(fop, inode, args...) \
- do { \
- ns_info_t *info = &frame->root->ns_info; \
- call_frame_t *new_frame = NULL; \
- ns_local_t *local = NULL; \
- call_stub_t *stub = NULL; \
+ do { \
+ ns_info_t *info = &frame->root->ns_info; \
+ call_frame_t *new_frame = NULL; \
+ ns_local_t *local = NULL; \
+ call_stub_t *stub = NULL; \
\
- gf_log (this->name, GF_LOG_DEBUG, \
- " %s winding, looking for path", \
- uuid_utoa (inode->gfid)); \
+ gf_log(this->name, GF_LOG_DEBUG, " %s winding, looking for path", \
+ uuid_utoa(inode->gfid)); \
\
- new_frame = create_frame (this, this->ctx->pool); \
- if (!new_frame) { \
- gf_log (this->name, GF_LOG_ERROR, \
- "Cannot allocate new call frame."); \
- goto wind; \
- } \
+ new_frame = create_frame(this, this->ctx->pool); \
+ if (!new_frame) { \
+ gf_log(this->name, GF_LOG_ERROR, \
+ "Cannot allocate new call frame."); \
+ goto wind; \
+ } \
\
- stub = fop_##fop##_stub (frame, default_##fop, args); \
- if (!stub) { \
- gf_log (this->name, GF_LOG_ERROR, \
- "Cannot allocate function stub."); \
- goto wind; \
- } \
+ stub = fop_##fop##_stub(frame, default_##fop, args); \
+ if (!stub) { \
+ gf_log(this->name, GF_LOG_ERROR, \
+ "Cannot allocate function stub."); \
+ goto wind; \
+ } \
\
- new_frame->root->uid = 0; \
- new_frame->root->gid = 0; \
- /* Put a phony "not found" NS info into this call. */ \
- new_frame->root->ns_info = *info; \
+ new_frame->root->uid = 0; \
+ new_frame->root->gid = 0; \
+ /* Put a phony "not found" NS info into this call. */ \
+ new_frame->root->ns_info = *info; \
\
- local = ns_local_new (stub, inode); \
- if (!local) { \
- gf_log (this->name, GF_LOG_ERROR, \
- "Cannot allocate function local."); \
- goto wind; \
- } \
+ local = ns_local_new(stub, inode); \
+ if (!local) { \
+ gf_log(this->name, GF_LOG_ERROR, \
+ "Cannot allocate function local."); \
+ goto wind; \
+ } \
\
- new_frame->local = local; \
- /* After allocating a new frame, a call stub (to \
- * resume our current fop), and a local variables \
- * struct (for our loc to getxattr and our resume \
- * stub), call getxattr and unwind to get_path_resume_cbk. \
- */ \
- STACK_WIND (new_frame, get_path_resume_cbk, \
- FIRST_CHILD (this), \
- FIRST_CHILD (this)->fops->getxattr, &local->loc, \
- GET_ANCESTRY_PATH_KEY, NULL); \
- } while (0)
+ new_frame->local = local; \
+ /* After allocating a new frame, a call stub (to \
+ * resume our current fop), and a local variables \
+ * struct (for our loc to getxattr and our resume \
+ * stub), call getxattr and unwind to get_path_resume_cbk. \
+ */ \
+ STACK_WIND(new_frame, get_path_resume_cbk, FIRST_CHILD(this), \
+ FIRST_CHILD(this)->fops->getxattr, &local->loc, \
+ GET_ANCESTRY_PATH_KEY, NULL); \
+ } while (0)
int32_t
-ns_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflags,
- dict_t *xdata)
+ns_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflags,
+ dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (rmdir, loc->inode, loc, xflags, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_rmdir_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->rmdir, loc, xflags, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(rmdir, loc->inode, loc, xflags, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_rmdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rmdir, loc, xflags, xdata);
+ return 0;
}
int32_t
-ns_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflags,
- dict_t *xdata)
+ns_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflags,
+ dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (unlink, loc->inode, loc, xflags, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_unlink_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->unlink, loc, xflags, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(unlink, loc->inode, loc, xflags, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_unlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, xflags, xdata);
+ return 0;
}
int32_t
-ns_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
+ns_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, newloc);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this,
+ newloc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (rename, newloc->inode, oldloc, newloc,
- xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_rename_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->rename, oldloc, newloc, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(rename, newloc->inode, oldloc, newloc, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_rename_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
+ return 0;
}
int32_t
-ns_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
+ns_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, newloc);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this,
+ newloc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (link, newloc->inode, oldloc, newloc,
- xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_link_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->link, oldloc, newloc, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(link, newloc->inode, oldloc, newloc, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_link_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
+ return 0;
}
int32_t
-ns_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- mode_t umask, dict_t *xdata)
+ns_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (mkdir, loc->inode, loc, mode, umask,
- xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_mkdir_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->mkdir, loc, mode, umask, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(mkdir, loc->inode, loc, mode, umask, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_mkdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
+ return 0;
}
int32_t
-ns_symlink (call_frame_t *frame, xlator_t *this, const char *linkname,
- loc_t *loc, mode_t umask, dict_t *xdata)
+ns_symlink(call_frame_t *frame, xlator_t *this, const char *linkname,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (symlink, loc->inode, linkname, loc,
- umask, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_symlink_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->symlink, linkname, loc, umask,
- xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(symlink, loc->inode, linkname, loc, umask,
+ xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_symlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->symlink, linkname, loc, umask, xdata);
+ return 0;
}
int32_t
-ns_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- dev_t dev, mode_t umask, dict_t *xdata)
+ns_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t dev, mode_t umask, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (mknod, loc->inode, loc, mode, dev,
- umask, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_mknod_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->mknod, loc, mode, dev, umask,
- xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(mknod, loc->inode, loc, mode, dev, umask, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_mknod_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, loc, mode, dev, umask, xdata);
+ return 0;
}
int32_t
-ns_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+ns_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (create, loc->inode, loc, flags, mode,
- umask, fd, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_create_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->create, loc, flags, mode, umask,
- fd, xdata);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(create, loc->inode, loc, flags, mode, umask, fd,
+ xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_create_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, flags, mode, umask, fd,
+ xdata);
+ return 0;
}
int32_t
-ns_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
- int32_t valid, dict_t *xdata)
+ns_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
+ int32_t valid, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (fsetattr, fd->inode, fd, stbuf, valid,
- xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_fsetattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fsetattr, fd, stbuf, valid,
- xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(fsetattr, fd->inode, fd, stbuf, valid, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_fsetattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetattr, fd, stbuf, valid, xdata);
+ return 0;
}
int32_t
-ns_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *stbuf,
- int32_t valid, dict_t *xdata)
+ns_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *stbuf,
+ int32_t valid, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (setattr, loc->inode, loc, stbuf, valid,
- xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_setattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setattr, loc, stbuf, valid,
- xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(setattr, loc->inode, loc, stbuf, valid, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_setattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid, xdata);
+ return 0;
}
int32_t
-ns_fremovexattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
+ns_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
+ dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (fremovexattr, fd->inode, fd, name,
- xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_fremovexattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fremovexattr, fd, name, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(fremovexattr, fd->inode, fd, name, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_fremovexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fremovexattr, fd, name, xdata);
+ return 0;
}
int32_t
-ns_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+ns_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (removexattr, loc->inode, loc, name,
- xdata);
- return 0;
- }
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(removexattr, loc->inode, loc, name, xdata);
+ return 0;
+ }
wind:
- STACK_WIND (frame, default_removexattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->removexattr, loc, name, xdata);
+ STACK_WIND(frame, default_removexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
+ return 0;
+}
+
+int32_t
+ns_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata)
+{
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
+
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(setxattr, loc->inode, loc, dict, flags, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_setxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, loc, dict, flags, xdata);
+ return 0;
}
int32_t
-ns_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ns_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
int32_t flags, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (setxattr, loc->inode, loc, dict, flags,
- xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_setxattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setxattr, loc, dict, flags,
- xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(fsetxattr, fd->inode, fd, dict, flags, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_fsetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+ return 0;
}
int32_t
-ns_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
- int32_t flags, dict_t *xdata)
+ns_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (fsetxattr, fd->inode, fd, dict, flags,
- xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_fsetxattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fsetxattr, fd, dict, flags,
- xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(truncate, loc->inode, loc, offset, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_truncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
+ return 0;
}
int32_t
-ns_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ns_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (truncate, loc->inode, loc, offset,
- xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_truncate_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->truncate, loc, offset, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(ftruncate, fd->inode, fd, offset, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_ftruncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+ return 0;
}
int32_t
-ns_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- dict_t *xdata)
+ns_writev(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
+ int32_t count, off_t offset, uint32_t flags, struct iobref *iobref,
+ dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (ftruncate, fd->inode, fd, offset,
- xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_ftruncate_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->ftruncate, fd, offset, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(writev, fd->inode, fd, vector, count, offset,
+ flags, iobref, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_writev_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector, count, offset,
+ flags, iobref, xdata);
+ return 0;
}
int32_t
-ns_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
- int32_t count, off_t offset, uint32_t flags, struct iobref *iobref,
- dict_t *xdata)
+ns_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (writev, fd->inode, fd, vector, count,
- offset, flags, iobref, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_writev_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->writev, fd, vector, count, offset,
- flags, iobref, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(lookup, loc->inode, loc, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, xdata);
+ return 0;
}
int32_t
-ns_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+ns_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (lookup, loc->inode, loc, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_lookup_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->lookup, loc, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(stat, loc->inode, loc, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_stat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->stat, loc, xdata);
+ return 0;
}
int32_t
-ns_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+ns_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (stat, loc->inode, loc, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_stat_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->stat, loc, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(fstat, fd->inode, fd, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_fstat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fstat, fd, xdata);
+ return 0;
}
int32_t
-ns_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+ns_readlink(call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+ dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (fstat, fd->inode, fd, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_fstat_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fstat, fd, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(readlink, loc->inode, loc, size, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_readlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readlink, loc, size, xdata);
+ return 0;
}
int32_t
-ns_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
- dict_t *xdata)
+ns_access(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+ dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (readlink, loc->inode, loc, size, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_readlink_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->readlink, loc, size, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(access, loc->inode, loc, mask, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_access_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->access, loc, mask, xdata);
+ return 0;
}
int32_t
-ns_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
- dict_t *xdata)
+ns_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (access, loc->inode, loc, mask, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_access_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->access, loc, mask, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(open, fd->inode, loc, flags, fd, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_open_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
+ return 0;
}
int32_t
-ns_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- fd_t *fd, dict_t *xdata)
+ns_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (open, fd->inode, loc, flags, fd, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_open_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->open, loc, flags, fd, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(readv, fd->inode, fd, size, offset, flags,
+ xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_readv_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readv, fd, size, offset, flags, xdata);
+ return 0;
}
int32_t
-ns_readv (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, uint32_t flags, dict_t *xdata)
+ns_flush(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (readv, fd->inode, fd, size, offset,
- flags, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_readv_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->readv, fd, size, offset, flags,
- xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(flush, fd->inode, fd, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_flush_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->flush, fd, xdata);
+ return 0;
}
int32_t
-ns_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+ns_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+ dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (flush, fd->inode, fd, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_flush_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->flush, fd, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(fsync, fd->inode, fd, datasync, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_fsync_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsync, fd, datasync, xdata);
+ return 0;
}
int32_t
-ns_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
- dict_t *xdata)
+ns_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (fsync, fd->inode, fd, datasync, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_fsync_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fsync, fd, datasync, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(opendir, loc->inode, loc, fd, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_opendir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->opendir, loc, fd, xdata);
+ return 0;
}
int32_t
-ns_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ns_fsyncdir(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
dict_t *xdata)
+
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (opendir, loc->inode, loc, fd, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_opendir_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->opendir, loc, fd, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(fsyncdir, fd->inode, fd, datasync, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_fsyncdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsyncdir, fd, datasync, xdata);
+ return 0;
}
int32_t
-ns_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
- dict_t *xdata)
-
+ns_rchecksum(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ int32_t len, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (fsyncdir, fd->inode, fd, datasync,
- xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_fsyncdir_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fsyncdir, fd, datasync, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(rchecksum, fd->inode, fd, offset, len, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_rchecksum_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rchecksum, fd, offset, len, xdata);
+ return 0;
}
int32_t
-ns_rchecksum (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- int32_t len, dict_t *xdata)
+ns_statfs(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (rchecksum, fd->inode, fd, offset, len,
- xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_rchecksum_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->rchecksum, fd, offset, len,
- xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(statfs, loc->inode, loc, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_statfs_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->statfs, loc, xdata);
+ return 0;
}
int32_t
-ns_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+ns_inodelk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ int32_t cmd, struct gf_flock *flock, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (statfs, loc->inode, loc, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_statfs_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->statfs, loc, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(inodelk, loc->inode, volume, loc, cmd, flock,
+ xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_inodelk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->inodelk, volume, loc, cmd, flock,
+ xdata);
+ return 0;
}
int32_t
-ns_inodelk (call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ns_finodelk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
int32_t cmd, struct gf_flock *flock, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (inodelk, loc->inode, volume, loc, cmd,
- flock, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_inodelk_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->inodelk, volume, loc, cmd, flock,
- xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(finodelk, fd->inode, volume, fd, cmd, flock,
+ xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_finodelk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->finodelk, volume, fd, cmd, flock,
+ xdata);
+ return 0;
}
int32_t
-ns_finodelk (call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
- int32_t cmd, struct gf_flock *flock, dict_t *xdata)
+ns_entrylk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (finodelk, fd->inode, volume, fd, cmd,
- flock, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_finodelk_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->finodelk, volume, fd, cmd, flock,
- xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(entrylk, loc->inode, volume, loc, basename, cmd,
+ type, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, volume, loc, basename, cmd,
+ type, xdata);
+ return 0;
}
int32_t
-ns_entrylk (call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ns_fentrylk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
const char *basename, entrylk_cmd cmd, entrylk_type type,
dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (entrylk, loc->inode, volume, loc,
- basename, cmd, type, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_entrylk_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->entrylk, volume, loc, basename,
- cmd, type, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(fentrylk, fd->inode, volume, fd, basename, cmd,
+ type, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_fentrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fentrylk, volume, fd, basename, cmd,
+ type, xdata);
+ return 0;
}
int32_t
-ns_fentrylk (call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
- const char *basename, entrylk_cmd cmd, entrylk_type type,
+ns_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (fentrylk, fd->inode, volume, fd,
- basename, cmd, type, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_fentrylk_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fentrylk, volume, fd, basename,
- cmd, type, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(fgetxattr, fd->inode, fd, name, xdata);
return 0;
-}
-
-int32_t
-ns_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
- dict_t *xdata)
-{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
-
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (fgetxattr, fd->inode, fd, name, xdata);
- return 0;
- }
+ }
wind:
- STACK_WIND (frame, default_fgetxattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fgetxattr, fd, name, xdata);
- return 0;
+ STACK_WIND(frame, default_fgetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
+ return 0;
}
int32_t
-ns_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, const char *name,
- dict_t *xdata)
+ns_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, const char *name,
+ dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (getxattr, loc->inode, loc, name, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_getxattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->getxattr, loc, name, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(getxattr, loc->inode, loc, name, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_getxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+ return 0;
}
int32_t
-ns_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
- struct gf_flock *flock, dict_t *xdata)
+ns_lk(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
+ struct gf_flock *flock, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (lk, fd->inode, fd, cmd, flock, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_lk_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->lk, fd, cmd, flock, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(lk, fd->inode, fd, cmd, flock, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_lk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lk, fd, cmd, flock, xdata);
+ return 0;
}
int32_t
-ns_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, dict_t *xdata)
+ns_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (readdir, fd->inode, fd, size, offset,
- xdata);
- return 0;
- }
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(readdir, fd->inode, fd, size, offset, xdata);
+ return 0;
+ }
wind:
- STACK_WIND (frame, default_readdir_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->readdir, fd, size, offset, xdata);
+ STACK_WIND(frame, default_readdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdir, fd, size, offset, xdata);
- return 0;
+ return 0;
}
int32_t
-ns_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, dict_t *dict)
+ns_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *dict)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (readdirp, fd->inode, fd, size, offset,
- dict);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_readdirp_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->readdirp, fd, size, offset, dict);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(readdirp, fd->inode, fd, size, offset, dict);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_readdirp_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdirp, fd, size, offset, dict);
+ return 0;
}
int32_t
-ns_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+ns_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_loc (__FUNCTION__, frame, this, loc);
+ path_parse_result_t ret = set_ns_from_loc(__FUNCTION__, frame, this, loc);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (xattrop, loc->inode, loc, flags, dict,
- xdata);
- return 0;
- }
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(xattrop, loc->inode, loc, flags, dict, xdata);
+ return 0;
+ }
wind:
- STACK_WIND (frame, default_xattrop_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->xattrop, loc, flags, dict, xdata);
+ STACK_WIND(frame, default_xattrop_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->xattrop, loc, flags, dict, xdata);
- return 0;
+ return 0;
}
int32_t
-ns_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
- gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+ns_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (fxattrop, fd->inode, fd, flags, dict,
- xdata);
- return 0;
- }
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(fxattrop, fd->inode, fd, flags, dict, xdata);
+ return 0;
+ }
wind:
- STACK_WIND (frame, default_fxattrop_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fxattrop, fd, flags, dict, xdata);
+ STACK_WIND(frame, default_fxattrop_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fxattrop, fd, flags, dict, xdata);
- return 0;
+ return 0;
}
-
int32_t
-ns_getspec (call_frame_t *frame, xlator_t *this, const char *key, int32_t flag)
+ns_getspec(call_frame_t *frame, xlator_t *this, const char *key, int32_t flag)
{
- STACK_WIND (frame, default_getspec_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->getspec, key, flag);
- return 0;
+ STACK_WIND(frame, default_getspec_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->getspec, key, flag);
+ return 0;
}
int32_t
-ns_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t keep_size,
- off_t offset, size_t len, dict_t *xdata)
+ns_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t keep_size,
+ off_t offset, size_t len, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (fallocate, fd->inode, fd, keep_size,
- offset, len, xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_fallocate_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fallocate, fd, keep_size, offset,
- len, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(fallocate, fd->inode, fd, keep_size, offset, len,
+ xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_fallocate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fallocate, fd, keep_size, offset, len,
+ xdata);
+ return 0;
}
int32_t
-ns_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- size_t len, dict_t *xdata)
+ns_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (discard, fd->inode, fd, offset, len,
- xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_discard_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->discard, fd, offset, len, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(discard, fd->inode, fd, offset, len, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_discard_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->discard, fd, offset, len, xdata);
+ return 0;
}
int32_t
-ns_zerofill (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- off_t len, dict_t *xdata)
+ns_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ off_t len, dict_t *xdata)
{
- path_parse_result_t ret = set_ns_from_fd (__FUNCTION__, frame, this, fd);
+ path_parse_result_t ret = set_ns_from_fd(__FUNCTION__, frame, this, fd);
- if (ret == PATH_PARSE_RESULT_IS_GFID) {
- GET_ANCESTRY_PATH_WIND (zerofill, fd->inode, fd, offset, len,
- xdata);
- return 0;
- }
-wind:
- STACK_WIND (frame, default_zerofill_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->zerofill, fd, offset, len, xdata);
+ if (ret == PATH_PARSE_RESULT_IS_GFID) {
+ GET_ANCESTRY_PATH_WIND(zerofill, fd->inode, fd, offset, len, xdata);
return 0;
+ }
+wind:
+ STACK_WIND(frame, default_zerofill_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->zerofill, fd, offset, len, xdata);
+ return 0;
}
int
-ns_forget (xlator_t *this, inode_t *inode)
+ns_forget(xlator_t *this, inode_t *inode)
{
- uint64_t ns_as_64 = 0;
- ns_info_t *info = NULL;
+ uint64_t ns_as_64 = 0;
+ ns_info_t *info = NULL;
- inode_ctx_del (inode, this, &ns_as_64);
+ inode_ctx_del(inode, this, &ns_as_64);
- if (!ns_as_64) {
- return 0;
- }
+ if (!ns_as_64) {
+ return 0;
+ }
- info = (ns_info_t *)ns_as_64;
- GF_FREE (info);
+ info = (ns_info_t *)(uintptr_t)ns_as_64;
+ GF_FREE(info);
- return 0;
+ return 0;
}
int32_t
-init (xlator_t *this)
+init(xlator_t *this)
{
- int32_t ret = -1;
- ns_private_t *priv = NULL;
+ int32_t ret = -1;
+ ns_private_t *priv = NULL;
- GF_VALIDATE_OR_GOTO (GF_NAMESPACE, this, out);
+ GF_VALIDATE_OR_GOTO(GF_NAMESPACE, this, out);
- if (!this->children || this->children->next) {
- gf_log (this->name, GF_LOG_ERROR,
- "translator needs a single subvolume.");
- goto out;
- }
+ if (!this->children || this->children->next) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "translator needs a single subvolume.");
+ goto out;
+ }
- if (!this->parents) {
- gf_log (this->name, GF_LOG_ERROR,
- "dangling volume. please check volfile.");
- goto out;
- }
+ if (!this->parents) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "dangling volume. please check volfile.");
+ goto out;
+ }
- priv = GF_CALLOC (1, sizeof (ns_private_t), 0);
+ priv = GF_CALLOC(1, sizeof(ns_private_t), 0);
- if (!priv) {
- gf_log (this->name, GF_LOG_ERROR,
- "Can't allocate ns_priv structure.");
- goto out;
- }
+ if (!priv) {
+ gf_log(this->name, GF_LOG_ERROR, "Can't allocate ns_priv structure.");
+ goto out;
+ }
- GF_OPTION_INIT ("tag-namespaces", priv->tag_namespaces, bool, out);
+ GF_OPTION_INIT("tag-namespaces", priv->tag_namespaces, bool, out);
- gf_log (this->name, GF_LOG_INFO, "Namespace xlator loaded");
- this->private = priv;
- ret = 0;
+ gf_log(this->name, GF_LOG_INFO, "Namespace xlator loaded");
+ this->private = priv;
+ ret = 0;
out:
- if (ret) {
- GF_FREE (priv);
- }
+ if (ret) {
+ GF_FREE(priv);
+ }
- return ret;
+ return ret;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- GF_FREE (this->private);
+ GF_FREE(this->private);
}
int
-reconfigure (xlator_t *this, dict_t *options)
+reconfigure(xlator_t *this, dict_t *options)
{
- int ret = -1;
- ns_private_t *priv = NULL;
+ int ret = -1;
+ ns_private_t *priv = NULL;
- GF_VALIDATE_OR_GOTO (this->name, this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
- GF_VALIDATE_OR_GOTO (this->name, options, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
+ GF_VALIDATE_OR_GOTO(this->name, options, out);
- priv = (ns_private_t *)this->private;
+ priv = (ns_private_t *)this->private;
- GF_OPTION_RECONF ("tag-namespaces", priv->tag_namespaces, options, bool,
- out);
+ GF_OPTION_RECONF("tag-namespaces", priv->tag_namespaces, options, bool,
+ out);
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
struct xlator_fops fops = {
- .lookup = ns_lookup,
- .stat = ns_stat,
- .fstat = ns_fstat,
- .truncate = ns_truncate,
- .ftruncate = ns_ftruncate,
- .access = ns_access,
- .readlink = ns_readlink,
- .mknod = ns_mknod,
- .mkdir = ns_mkdir,
- .unlink = ns_unlink,
- .rmdir = ns_rmdir,
- .symlink = ns_symlink,
- .rename = ns_rename,
- .link = ns_link,
- .create = ns_create,
- .open = ns_open,
- .readv = ns_readv,
- .writev = ns_writev,
- .flush = ns_flush,
- .fsync = ns_fsync,
- .opendir = ns_opendir,
- .readdir = ns_readdir,
- .readdirp = ns_readdirp,
- .fsyncdir = ns_fsyncdir,
- .statfs = ns_statfs,
- .setxattr = ns_setxattr,
- .getxattr = ns_getxattr,
- .fsetxattr = ns_fsetxattr,
- .fgetxattr = ns_fgetxattr,
- .removexattr = ns_removexattr,
- .fremovexattr = ns_fremovexattr,
- .lk = ns_lk,
- .inodelk = ns_inodelk,
- .finodelk = ns_finodelk,
- .entrylk = ns_entrylk,
- .fentrylk = ns_fentrylk,
- .rchecksum = ns_rchecksum,
- .xattrop = ns_xattrop,
- .fxattrop = ns_fxattrop,
- .setattr = ns_setattr,
- .fsetattr = ns_fsetattr,
- .getspec = ns_getspec,
- .fallocate = ns_fallocate,
- .discard = ns_discard,
- .zerofill = ns_zerofill,
+ .lookup = ns_lookup,
+ .stat = ns_stat,
+ .fstat = ns_fstat,
+ .truncate = ns_truncate,
+ .ftruncate = ns_ftruncate,
+ .access = ns_access,
+ .readlink = ns_readlink,
+ .mknod = ns_mknod,
+ .mkdir = ns_mkdir,
+ .unlink = ns_unlink,
+ .rmdir = ns_rmdir,
+ .symlink = ns_symlink,
+ .rename = ns_rename,
+ .link = ns_link,
+ .create = ns_create,
+ .open = ns_open,
+ .readv = ns_readv,
+ .writev = ns_writev,
+ .flush = ns_flush,
+ .fsync = ns_fsync,
+ .opendir = ns_opendir,
+ .readdir = ns_readdir,
+ .readdirp = ns_readdirp,
+ .fsyncdir = ns_fsyncdir,
+ .statfs = ns_statfs,
+ .setxattr = ns_setxattr,
+ .getxattr = ns_getxattr,
+ .fsetxattr = ns_fsetxattr,
+ .fgetxattr = ns_fgetxattr,
+ .removexattr = ns_removexattr,
+ .fremovexattr = ns_fremovexattr,
+ .lk = ns_lk,
+ .inodelk = ns_inodelk,
+ .finodelk = ns_finodelk,
+ .entrylk = ns_entrylk,
+ .fentrylk = ns_fentrylk,
+ .rchecksum = ns_rchecksum,
+ .xattrop = ns_xattrop,
+ .fxattrop = ns_fxattrop,
+ .setattr = ns_setattr,
+ .fsetattr = ns_fsetattr,
+ .getspec = ns_getspec,
+ .fallocate = ns_fallocate,
+ .discard = ns_discard,
+ .zerofill = ns_zerofill,
};
struct xlator_cbks cbks = {
- .forget = ns_forget,
+ .forget = ns_forget,
};
struct xlator_dumpops dumpops;
struct volume_options options[] = {
- {
- .key = { "tag-namespaces" },
- .type = GF_OPTION_TYPE_BOOL,
+ {
+ .key = {"tag-namespaces"},
+ .type = GF_OPTION_TYPE_BOOL,
.default_value = "off",
- .description = "This option enables this translator's functionality "
+ .description = "This option enables this translator's functionality "
"that tags every fop with a namespace hash for later "
"throttling, stats collection, logging, etc.",
- .op_version = {GD_OP_VERSION_4_1_0},
- .tags = {"namespace"},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
- },
- {.key = { NULL } },
+ .op_version = {GD_OP_VERSION_4_1_0},
+ .tags = {"namespace"},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
+ },
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .reconfigure = reconfigure,
+ .op_version = {GD_OP_VERSION_3_12_0},
+ .dumpops = &dumpops,
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "namespace",
+ .category = GF_TECH_PREVIEW,
};
diff --git a/xlators/features/namespace/src/namespace.h b/xlators/features/namespace/src/namespace.h
index 1dbec50e024..3a9b84d6426 100644
--- a/xlators/features/namespace/src/namespace.h
+++ b/xlators/features/namespace/src/namespace.h
@@ -6,18 +6,18 @@
#include "config.h"
#endif
-#include "xlator.h"
-#include "call-stub.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/call-stub.h>
#define GF_NAMESPACE "namespace"
typedef struct {
- gf_boolean_t tag_namespaces;
+ gf_boolean_t tag_namespaces;
} ns_private_t;
typedef struct {
- loc_t loc; /* We store a "fake" loc_t for the getxattr wind. */
- call_stub_t *stub; /* A stub back to the function we're resuming. */
+ loc_t loc; /* We store a "fake" loc_t for the getxattr wind. */
+ call_stub_t *stub; /* A stub back to the function we're resuming. */
} ns_local_t;
#endif /* __NAMESPACE_H__ */
diff --git a/xlators/features/quiesce/src/quiesce-mem-types.h b/xlators/features/quiesce/src/quiesce-mem-types.h
index 31346c1a794..416456b13af 100644
--- a/xlators/features/quiesce/src/quiesce-mem-types.h
+++ b/xlators/features/quiesce/src/quiesce-mem-types.h
@@ -11,11 +11,11 @@
#ifndef __QUIESCE_MEM_TYPES_H__
#define __QUIESCE_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_quiesce_mem_types_ {
- gf_quiesce_mt_priv_t = gf_common_mt_end + 1,
- gf_quiesce_mt_failover_hosts,
- gf_quiesce_mt_end
+ gf_quiesce_mt_priv_t = gf_common_mt_end + 1,
+ gf_quiesce_mt_failover_hosts,
+ gf_quiesce_mt_end
};
#endif
diff --git a/xlators/features/quiesce/src/quiesce-messages.h b/xlators/features/quiesce/src/quiesce-messages.h
index 8af3b10b06e..32ffd409807 100644
--- a/xlators/features/quiesce/src/quiesce-messages.h
+++ b/xlators/features/quiesce/src/quiesce-messages.h
@@ -11,7 +11,7 @@
#ifndef __QUIESCE_MESSAGES_H__
#define __QUIESCE_MESSAGES_H__
-#include "glfs-message-id.h"
+#include <glusterfs/glfs-message-id.h>
/* To add new message IDs, append new identifiers at the end of the list.
*
@@ -23,9 +23,6 @@
* glfs-message-id.h.
*/
-GLFS_MSGID(QUIESCE,
- QUIESCE_MSG_INVAL_HOST,
- QUIESCE_MSG_FAILOVER_FAILED
-);
+GLFS_MSGID(QUIESCE, QUIESCE_MSG_INVAL_HOST, QUIESCE_MSG_FAILOVER_FAILED);
#endif /* __NL_CACHE_MESSAGES_H__ */
diff --git a/xlators/features/quiesce/src/quiesce.c b/xlators/features/quiesce/src/quiesce.c
index 95c0c8b0829..0e5eb60a16f 100644
--- a/xlators/features/quiesce/src/quiesce.c
+++ b/xlators/features/quiesce/src/quiesce.c
@@ -8,833 +8,814 @@
cases as published by the Free Software Foundation.
*/
#include "quiesce.h"
-#include "defaults.h"
-#include "call-stub.h"
+#include <glusterfs/defaults.h>
+#include <glusterfs/call-stub.h>
/* TODO: */
/* Think about 'writev/_*_lk/setattr/xattrop/' fops to do re-transmittion */
void
-gf_quiesce_timeout (void *data);
-
+gf_quiesce_timeout(void *data);
/* Quiesce Specific Functions */
void
-gf_quiesce_local_wipe (xlator_t *this, quiesce_local_t *local)
+gf_quiesce_local_wipe(xlator_t *this, quiesce_local_t *local)
{
- if (!local || !this || !this->private)
- return;
+ if (!local || !this || !this->private)
+ return;
- if (local->loc.inode)
- loc_wipe (&local->loc);
- if (local->fd)
- fd_unref (local->fd);
- GF_FREE (local->name);
- GF_FREE (local->volname);
- if (local->dict)
- dict_unref (local->dict);
- if (local->iobref)
- iobref_unref (local->iobref);
- GF_FREE (local->vector);
+ if (local->loc.inode)
+ loc_wipe(&local->loc);
+ if (local->fd)
+ fd_unref(local->fd);
+ GF_FREE(local->name);
+ GF_FREE(local->volname);
+ if (local->dict)
+ dict_unref(local->dict);
+ if (local->iobref)
+ iobref_unref(local->iobref);
+ GF_FREE(local->vector);
- mem_put (local);
+ mem_put(local);
}
void
-__gf_quiesce_start_timer (xlator_t *this, quiesce_priv_t *priv)
+__gf_quiesce_start_timer(xlator_t *this, quiesce_priv_t *priv)
{
- struct timespec timeout = {0,};
+ struct timespec timeout = {
+ 0,
+ };
- if (!priv->timer) {
- timeout.tv_sec = priv->timeout;
- timeout.tv_nsec = 0;
+ if (!priv->timer) {
+ timeout.tv_sec = priv->timeout;
+ timeout.tv_nsec = 0;
- priv->timer = gf_timer_call_after (this->ctx,
- timeout,
- gf_quiesce_timeout,
- (void *) this);
- if (priv->timer == NULL) {
- gf_log (this->name, GF_LOG_ERROR,
- "Cannot create timer");
- }
+ priv->timer = gf_timer_call_after(this->ctx, timeout,
+ gf_quiesce_timeout, (void *)this);
+ if (priv->timer == NULL) {
+ gf_log(this->name, GF_LOG_ERROR, "Cannot create timer");
}
+ }
}
static void
-__gf_quiesce_cleanup_failover_hosts (xlator_t *this, quiesce_priv_t *priv)
+__gf_quiesce_cleanup_failover_hosts(xlator_t *this, quiesce_priv_t *priv)
{
- quiesce_failover_hosts_t *tmp = NULL;
- quiesce_failover_hosts_t *failover_host = NULL;
+ quiesce_failover_hosts_t *tmp = NULL;
+ quiesce_failover_hosts_t *failover_host = NULL;
- list_for_each_entry_safe (failover_host, tmp,
- &priv->failover_list, list) {
- GF_FREE (failover_host->addr);
- list_del (&failover_host->list);
- GF_FREE (failover_host);
- }
- return;
+ list_for_each_entry_safe(failover_host, tmp, &priv->failover_list, list)
+ {
+ GF_FREE(failover_host->addr);
+ list_del(&failover_host->list);
+ GF_FREE(failover_host);
+ }
+ return;
}
void
-gf_quiesce_populate_failover_hosts (xlator_t *this, quiesce_priv_t *priv,
- const char *value)
-{
- char *dup_val = NULL;
- char *addr_tok = NULL;
- char *save_ptr = NULL;
- quiesce_failover_hosts_t *failover_host = NULL;
-
- if (!value)
- goto out;
-
- dup_val = gf_strdup (value);
- if (!dup_val)
- goto out;
-
- LOCK (&priv->lock);
- {
- if (!list_empty (&priv->failover_list))
- __gf_quiesce_cleanup_failover_hosts (this, priv);
- addr_tok = strtok_r (dup_val, ",", &save_ptr);
- while (addr_tok) {
- if (!valid_internet_address (addr_tok, _gf_true)) {
- gf_msg (this->name, GF_LOG_INFO, 0,
- QUIESCE_MSG_INVAL_HOST, "Specified "
- "invalid internet address:%s",
- addr_tok);
- continue;
- }
- failover_host = GF_CALLOC (1, sizeof(*failover_host),
- gf_quiesce_mt_failover_hosts);
- failover_host->addr = gf_strdup (addr_tok);
- INIT_LIST_HEAD (&failover_host->list);
- list_add (&failover_host->list, &priv->failover_list);
- addr_tok = strtok_r (NULL, ",", &save_ptr);
- }
- }
- UNLOCK (&priv->lock);
- GF_FREE (dup_val);
+gf_quiesce_populate_failover_hosts(xlator_t *this, quiesce_priv_t *priv,
+ const char *value)
+{
+ char *dup_val = NULL;
+ char *addr_tok = NULL;
+ char *save_ptr = NULL;
+ quiesce_failover_hosts_t *failover_host = NULL;
+
+ if (!value)
+ goto out;
+
+ dup_val = gf_strdup(value);
+ if (!dup_val)
+ goto out;
+
+ addr_tok = strtok_r(dup_val, ",", &save_ptr);
+ LOCK(&priv->lock);
+ {
+ if (!list_empty(&priv->failover_list))
+ __gf_quiesce_cleanup_failover_hosts(this, priv);
+
+ while (addr_tok) {
+ if (!valid_internet_address(addr_tok, _gf_true, _gf_false)) {
+ gf_msg(this->name, GF_LOG_INFO, 0, QUIESCE_MSG_INVAL_HOST,
+ "Specified "
+ "invalid internet address:%s",
+ addr_tok);
+ continue;
+ }
+ failover_host = GF_CALLOC(1, sizeof(*failover_host),
+ gf_quiesce_mt_failover_hosts);
+ failover_host->addr = gf_strdup(addr_tok);
+ INIT_LIST_HEAD(&failover_host->list);
+ list_add(&failover_host->list, &priv->failover_list);
+ addr_tok = strtok_r(NULL, ",", &save_ptr);
+ }
+ }
+ UNLOCK(&priv->lock);
+ GF_FREE(dup_val);
out:
- return;
+ return;
}
int32_t
-gf_quiesce_failover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+gf_quiesce_failover_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
+ quiesce_priv_t *priv = NULL;
- if (op_ret < 0) {
- /* Failure here doesn't mean the failover to another host didn't
- * succeed, we will know if failover succeeds or not by the
- * CHILD_UP/CHILD_DOWN event. A failure here indicates something
- * went wrong with the submission of failover command, hence
- * just abort the failover attempts without retrying with other
- * hosts.
- */
- gf_msg (this->name, GF_LOG_INFO, op_errno,
- QUIESCE_MSG_FAILOVER_FAILED,
- "Initiating failover to host:%s failed:", (char *)cookie);
- }
+ if (op_ret < 0) {
+ /* Failure here doesn't mean the failover to another host didn't
+ * succeed, we will know if failover succeeds or not by the
+ * CHILD_UP/CHILD_DOWN event. A failure here indicates something
+ * went wrong with the submission of failover command, hence
+ * just abort the failover attempts without retrying with other
+ * hosts.
+ */
+ gf_msg(this->name, GF_LOG_INFO, op_errno, QUIESCE_MSG_FAILOVER_FAILED,
+ "Initiating failover to host:%s failed:", (char *)cookie);
+ }
- GF_FREE (cookie);
- STACK_DESTROY (frame->root);
+ GF_FREE(cookie);
+ STACK_DESTROY(frame->root);
- priv = this->private;
- __gf_quiesce_start_timer (this, priv);
+ priv = this->private;
+ __gf_quiesce_start_timer(this, priv);
- return 0;
+ return 0;
}
int
-__gf_quiesce_perform_failover (xlator_t *this)
-{
- int ret = 0;
- call_frame_t *frame = NULL;
- dict_t *dict = NULL;
- quiesce_priv_t *priv = NULL;
- quiesce_failover_hosts_t *failover_host = NULL;
- quiesce_failover_hosts_t *host = NULL;
-
- priv = this->private;
-
- if (priv->pass_through) {
- gf_msg_trace (this->name, 0, "child is up, hence not "
- "performing any failover");
- goto out;
- }
-
- list_for_each_entry (failover_host, &priv->failover_list, list) {
- if (failover_host->tried == 0) {
- host = failover_host;
- failover_host->tried = 1;
- break;
- }
- }
- if (!host) {
- /*TODO: Keep trying until any of the gfproxy comes back up.
- Currently it tries failing over once for each host,
- if it doesn't succeed then returns error to mount point
- list_for_each_entry (failover_host,
- &priv->failover_list, list) {
- failover_host->tried = 0;
- }*/
- gf_msg_debug (this->name, 0, "all the failover hosts have "
- "been tried and looks like didn't succeed");
- ret = -1;
- goto out;
- }
-
- frame = create_frame (this, this->ctx->pool);
-
- dict = dict_new ();
-
- ret = dict_set_dynstr (dict, CLIENT_CMD_CONNECT,
- gf_strdup (host->addr));
-
- gf_msg_trace (this->name, 0, "Initiating failover to:%s",
- host->addr);
-
- STACK_WIND_COOKIE (frame, gf_quiesce_failover_cbk, NULL,
- FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setxattr,
- NULL, dict, 0, NULL);
+__gf_quiesce_perform_failover(xlator_t *this)
+{
+ int ret = 0;
+ call_frame_t *frame = NULL;
+ dict_t *dict = NULL;
+ quiesce_priv_t *priv = NULL;
+ quiesce_failover_hosts_t *failover_host = NULL;
+ quiesce_failover_hosts_t *host = NULL;
+
+ priv = this->private;
+
+ if (priv->pass_through) {
+ gf_msg_trace(this->name, 0,
+ "child is up, hence not "
+ "performing any failover");
+ goto out;
+ }
+
+ list_for_each_entry(failover_host, &priv->failover_list, list)
+ {
+ if (failover_host->tried == 0) {
+ host = failover_host;
+ failover_host->tried = 1;
+ break;
+ }
+ }
+ if (!host) {
+ /*TODO: Keep trying until any of the gfproxy comes back up.
+ Currently it tries failing over once for each host,
+ if it doesn't succeed then returns error to mount point
+ list_for_each_entry (failover_host,
+ &priv->failover_list, list) {
+ failover_host->tried = 0;
+ }*/
+ gf_msg_debug(this->name, 0,
+ "all the failover hosts have "
+ "been tried and looks like didn't succeed");
+ ret = -1;
+ goto out;
+ }
+
+ frame = create_frame(this, this->ctx->pool);
+ if (!frame) {
+ gf_msg_debug(this->name, 0, "failed to create the frame");
+ ret = -1;
+ goto out;
+ }
+
+ dict = dict_new();
+
+ ret = dict_set_dynstr(dict, CLIENT_CMD_CONNECT, gf_strdup(host->addr));
+
+ gf_msg_trace(this->name, 0, "Initiating failover to:%s", host->addr);
+
+ STACK_WIND_COOKIE(frame, gf_quiesce_failover_cbk, NULL, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, NULL, dict, 0, NULL);
out:
- if (dict)
- dict_unref (dict);
+ if (dict)
+ dict_unref(dict);
- return ret;
+ return ret;
}
call_stub_t *
-gf_quiesce_dequeue (xlator_t *this)
+gf_quiesce_dequeue(xlator_t *this)
{
- call_stub_t *stub = NULL;
- quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- if (!priv || list_empty (&priv->req))
- return NULL;
+ if (!priv || list_empty(&priv->req))
+ return NULL;
- LOCK (&priv->lock);
- {
- stub = list_entry (priv->req.next, call_stub_t, list);
- list_del_init (&stub->list);
- priv->queue_size--;
- }
- UNLOCK (&priv->lock);
+ LOCK(&priv->lock);
+ {
+ stub = list_entry(priv->req.next, call_stub_t, list);
+ list_del_init(&stub->list);
+ priv->queue_size--;
+ }
+ UNLOCK(&priv->lock);
- return stub;
+ return stub;
}
void *
-gf_quiesce_dequeue_start (void *data)
+gf_quiesce_dequeue_start(void *data)
{
- xlator_t *this = NULL;
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ xlator_t *this = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- this = data;
- priv = this->private;
- THIS = this;
+ this = data;
+ priv = this->private;
+ THIS = this;
- while (!list_empty (&priv->req)) {
- stub = gf_quiesce_dequeue (this);
- if (stub) {
- call_resume (stub);
- }
+ while (!list_empty(&priv->req)) {
+ stub = gf_quiesce_dequeue(this);
+ if (stub) {
+ call_resume(stub);
}
+ }
- return 0;
+ return 0;
}
-
void
-gf_quiesce_timeout (void *data)
+gf_quiesce_timeout(void *data)
{
- xlator_t *this = NULL;
- quiesce_priv_t *priv = NULL;
- int ret = -1;
+ xlator_t *this = NULL;
+ quiesce_priv_t *priv = NULL;
+ int ret = -1;
- this = data;
- priv = this->private;
- THIS = this;
+ this = data;
+ priv = this->private;
+ THIS = this;
- LOCK (&priv->lock);
- {
- priv->timer = NULL;
- if (priv->pass_through) {
- UNLOCK (&priv->lock);
- goto out;
- }
- ret = __gf_quiesce_perform_failover (THIS);
+ LOCK(&priv->lock);
+ {
+ priv->timer = NULL;
+ if (priv->pass_through) {
+ UNLOCK(&priv->lock);
+ goto out;
}
- UNLOCK (&priv->lock);
+ ret = __gf_quiesce_perform_failover(THIS);
+ }
+ UNLOCK(&priv->lock);
- if (ret < 0) {
- priv->pass_through = _gf_true;
- gf_quiesce_dequeue_start (this);
- }
+ if (ret < 0) {
+ priv->pass_through = _gf_true;
+ gf_quiesce_dequeue_start(this);
+ }
out:
- return;
+ return;
}
void
-gf_quiesce_enqueue (xlator_t *this, call_stub_t *stub)
+gf_quiesce_enqueue(xlator_t *this, call_stub_t *stub)
{
- quiesce_priv_t *priv = NULL;
-
- priv = this->private;
- if (!priv) {
- gf_log_callingfn (this->name, GF_LOG_ERROR,
- "this->private == NULL");
- return;
- }
-
- LOCK (&priv->lock);
- {
- list_add_tail (&stub->list, &priv->req);
- priv->queue_size++;
- __gf_quiesce_start_timer (this, priv);
- }
- UNLOCK (&priv->lock);
+ quiesce_priv_t *priv = NULL;
+ priv = this->private;
+ if (!priv) {
+ gf_log_callingfn(this->name, GF_LOG_ERROR, "this->private == NULL");
return;
-}
+ }
+ LOCK(&priv->lock);
+ {
+ list_add_tail(&stub->list, &priv->req);
+ priv->queue_size++;
+ __gf_quiesce_start_timer(this, priv);
+ }
+ UNLOCK(&priv->lock);
+ return;
+}
/* _CBK function section */
int32_t
-quiesce_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *dict, struct iatt *postparent)
+quiesce_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *dict, struct iatt *postparent)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
-
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_lookup_stub (frame, default_lookup_resume,
- &local->loc, local->dict);
- if (!stub) {
- STACK_UNWIND_STRICT (lookup, frame, -1, ENOMEM,
- NULL, NULL, NULL, NULL);
- goto out;
- }
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_lookup_stub(frame, default_lookup_resume, &local->loc,
+ local->dict);
+ if (!stub) {
+ STACK_UNWIND_STRICT(lookup, frame, -1, ENOMEM, NULL, NULL, NULL,
+ NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
- dict, postparent);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(lookup, frame, op_ret, op_errno, inode, buf, dict,
+ postparent);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
int32_t
-quiesce_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- dict_t *xdata)
+quiesce_stat_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_stat_stub (frame, default_stat_resume,
- &local->loc, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (stat, frame, -1, ENOMEM,
- NULL, NULL);
- goto out;
- }
-
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_stat_stub(frame, default_stat_resume, &local->loc, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(stat, frame, -1, ENOMEM, NULL, NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (stat, frame, op_ret, op_errno, buf, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(stat, frame, op_ret, op_errno, buf, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
int32_t
-quiesce_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+quiesce_access_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_access_stub (frame, default_access_resume,
- &local->loc, local->flag, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (access, frame, -1, ENOMEM, NULL);
- goto out;
- }
-
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_access_stub(frame, default_access_resume, &local->loc,
+ local->flag, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(access, frame, -1, ENOMEM, NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (access, frame, op_ret, op_errno, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(access, frame, op_ret, op_errno, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
int32_t
-quiesce_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, const char *path,
- struct iatt *buf, dict_t *xdata)
+quiesce_readlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, const char *path,
+ struct iatt *buf, dict_t *xdata)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
-
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_readlink_stub (frame, default_readlink_resume,
- &local->loc, local->size, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (readlink, frame, -1, ENOMEM,
- NULL, NULL, NULL);
- goto out;
- }
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_readlink_stub(frame, default_readlink_resume, &local->loc,
+ local->size, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(readlink, frame, -1, ENOMEM, NULL, NULL, NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, path, buf, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(readlink, frame, op_ret, op_errno, path, buf, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
int32_t
-quiesce_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+quiesce_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_open_stub (frame, default_open_resume,
- &local->loc, local->flag, local->fd,
- xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (open, frame, -1, ENOMEM,
- NULL, NULL);
- goto out;
- }
-
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_open_stub(frame, default_open_resume, &local->loc,
+ local->flag, local->fd, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(open, frame, -1, ENOMEM, NULL, NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(open, frame, op_ret, op_errno, fd, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
int32_t
-quiesce_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iovec *vector,
- int32_t count, struct iatt *stbuf, struct iobref *iobref, dict_t *xdata)
+quiesce_readv_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iovec *vector,
+ int32_t count, struct iatt *stbuf, struct iobref *iobref,
+ dict_t *xdata)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
-
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_readv_stub (frame, default_readv_resume,
- local->fd, local->size, local->offset,
- local->io_flag, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (readv, frame, -1, ENOMEM,
- NULL, 0, NULL, NULL, NULL);
- goto out;
- }
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_readv_stub(frame, default_readv_resume, local->fd,
+ local->size, local->offset, local->io_flag,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(readv, frame, -1, ENOMEM, NULL, 0, NULL, NULL,
+ NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector, count,
- stbuf, iobref, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(readv, frame, op_ret, op_errno, vector, count, stbuf,
+ iobref, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
int32_t
-quiesce_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+quiesce_flush_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
-
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_flush_stub (frame, default_flush_resume,
- local->fd, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (flush, frame, -1, ENOMEM, NULL);
- goto out;
- }
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_flush_stub(frame, default_flush_resume, local->fd, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(flush, frame, -1, ENOMEM, NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(flush, frame, op_ret, op_errno, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
-
-
int32_t
-quiesce_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+quiesce_fsync_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
-
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_fsync_stub (frame, default_fsync_resume,
- local->fd, local->flag, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (fsync, frame, -1, ENOMEM,
- NULL, NULL, NULL);
- goto out;
- }
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_fsync_stub(frame, default_fsync_resume, local->fd,
+ local->flag, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(fsync, frame, -1, ENOMEM, NULL, NULL, NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
int32_t
-quiesce_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
+quiesce_fstat_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_fstat_stub (frame, default_fstat_resume,
- local->fd, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (fstat, frame, -1, ENOMEM,
- NULL, NULL);
- goto out;
- }
-
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_fstat_stub(frame, default_fstat_resume, local->fd, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(fstat, frame, -1, ENOMEM, NULL, NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, buf, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(fstat, frame, op_ret, op_errno, buf, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
int32_t
-quiesce_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+quiesce_opendir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_opendir_stub (frame, default_opendir_resume,
- &local->loc, local->fd, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (opendir, frame, -1, ENOMEM,
- NULL, NULL);
- goto out;
- }
-
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_opendir_stub(frame, default_opendir_resume, &local->loc,
+ local->fd, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(opendir, frame, -1, ENOMEM, NULL, NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(opendir, frame, op_ret, op_errno, fd, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
int32_t
-quiesce_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+quiesce_fsyncdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_fsyncdir_stub (frame, default_fsyncdir_resume,
- local->fd, local->flag, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (fsyncdir, frame, -1, ENOMEM, NULL);
- goto out;
- }
-
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_fsyncdir_stub(frame, default_fsyncdir_resume, local->fd,
+ local->flag, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(fsyncdir, frame, -1, ENOMEM, NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (fsyncdir, frame, op_ret, op_errno, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(fsyncdir, frame, op_ret, op_errno, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
int32_t
-quiesce_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct statvfs *buf, dict_t *xdata)
+quiesce_statfs_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct statvfs *buf,
+ dict_t *xdata)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
-
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_statfs_stub (frame, default_statfs_resume,
- &local->loc, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (statfs, frame, -1, ENOMEM,
- NULL, NULL);
- goto out;
- }
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_statfs_stub(frame, default_statfs_resume, &local->loc,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(statfs, frame, -1, ENOMEM, NULL, NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, buf, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(statfs, frame, op_ret, op_errno, buf, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
int32_t
-quiesce_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+quiesce_fgetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_fgetxattr_stub (frame, default_fgetxattr_resume,
- local->fd, local->name, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (fgetxattr, frame, -1, ENOMEM,
- NULL, NULL);
- goto out;
- }
-
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_fgetxattr_stub(frame, default_fgetxattr_resume, local->fd,
+ local->name, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(fgetxattr, frame, -1, ENOMEM, NULL, NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(fgetxattr, frame, op_ret, op_errno, dict, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
-
int32_t
-quiesce_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+quiesce_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_getxattr_stub (frame, default_getxattr_resume,
- &local->loc, local->name, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (getxattr, frame, -1, ENOMEM,
- NULL, NULL);
- goto out;
- }
-
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_getxattr_stub(frame, default_getxattr_resume, &local->loc,
+ local->name, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(getxattr, frame, -1, ENOMEM, NULL, NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(getxattr, frame, op_ret, op_errno, dict, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
-
int32_t
-quiesce_rchecksum_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, uint32_t weak_checksum,
- uint8_t *strong_checksum, dict_t *xdata)
+quiesce_rchecksum_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, uint32_t weak_checksum,
+ uint8_t *strong_checksum, dict_t *xdata)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_rchecksum_stub (frame, default_rchecksum_resume,
- local->fd, local->offset, local->flag, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (rchecksum, frame, -1, ENOMEM,
- 0, NULL, NULL);
- goto out;
- }
-
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_rchecksum_stub(frame, default_rchecksum_resume, local->fd,
+ local->offset, local->flag, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(rchecksum, frame, -1, ENOMEM, 0, NULL, NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (rchecksum, frame, op_ret, op_errno, weak_checksum,
- strong_checksum, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(rchecksum, frame, op_ret, op_errno, weak_checksum,
+ strong_checksum, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
-
int32_t
-quiesce_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *entries, dict_t *xdata)
+quiesce_readdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+ dict_t *xdata)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_readdir_stub (frame, default_readdir_resume,
- local->fd, local->size, local->offset, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (readdir, frame, -1, ENOMEM,
- NULL, NULL);
- goto out;
- }
-
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_readdir_stub(frame, default_readdir_resume, local->fd,
+ local->size, local->offset, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(readdir, frame, -1, ENOMEM, NULL, NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, entries, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(readdir, frame, op_ret, op_errno, entries, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
-
int32_t
-quiesce_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *entries, dict_t *xdata)
+quiesce_readdirp_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+ dict_t *xdata)
{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_readdirp_stub (frame, default_readdirp_resume,
- local->fd, local->size, local->offset,
- local->dict);
- if (!stub) {
- STACK_UNWIND_STRICT (readdirp, frame, -1, ENOMEM,
- NULL, NULL);
- goto out;
- }
-
- gf_quiesce_enqueue (this, stub);
- goto out;
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_readdirp_stub(frame, default_readdirp_resume, local->fd,
+ local->size, local->offset, local->dict);
+ if (!stub) {
+ STACK_UNWIND_STRICT(readdirp, frame, -1, ENOMEM, NULL, NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(readdirp, frame, op_ret, op_errno, entries, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
-
#if 0
int32_t
@@ -1180,1740 +1161,1544 @@ out:
#endif /* if 0 */
-
/* FOP */
/* No retransmittion */
int32_t
-quiesce_removexattr (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- const char *name, dict_t *xdata)
+quiesce_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->pass_through) {
- STACK_WIND (frame,
- default_removexattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->removexattr,
- loc,
- name, xdata);
- return 0;
- }
+ if (priv->pass_through) {
+ STACK_WIND(frame, default_removexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
+ return 0;
+ }
- stub = fop_removexattr_stub (frame, default_removexattr_resume,
- loc, name, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (removexattr, frame, -1, ENOMEM, NULL);
- return 0;
- }
+ stub = fop_removexattr_stub(frame, default_removexattr_resume, loc, name,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(removexattr, frame, -1, ENOMEM, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int32_t
-quiesce_truncate (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- off_t offset, dict_t *xdata)
+quiesce_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->pass_through) {
- STACK_WIND (frame,
- default_truncate_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->truncate,
- loc,
- offset, xdata);
- return 0;
- }
+ if (priv->pass_through) {
+ STACK_WIND(frame, default_fremovexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fremovexattr, fd, name, xdata);
+ return 0;
+ }
- stub = fop_truncate_stub (frame, default_truncate_resume, loc, offset, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (truncate, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
- }
+ stub = fop_fremovexattr_stub(frame, default_fremovexattr_resume, fd, name,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(fremovexattr, frame, -1, ENOMEM, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int32_t
-quiesce_fsetxattr (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- dict_t *dict,
- int32_t flags, dict_t *xdata)
+quiesce_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->pass_through) {
- STACK_WIND (frame,
- default_fsetxattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetxattr,
- fd,
- dict,
- flags, xdata);
- return 0;
- }
+ if (priv->pass_through) {
+ STACK_WIND(frame, default_truncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
+ return 0;
+ }
- stub = fop_fsetxattr_stub (frame, default_fsetxattr_resume,
- fd, dict, flags, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (fsetxattr, frame, -1, ENOMEM, NULL);
- return 0;
- }
+ stub = fop_truncate_stub(frame, default_truncate_resume, loc, offset,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(truncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int32_t
-quiesce_setxattr (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- dict_t *dict,
- int32_t flags, dict_t *xdata)
+quiesce_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->pass_through) {
- STACK_WIND (frame,
- default_setxattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setxattr,
- loc,
- dict,
- flags, xdata);
- return 0;
- }
+ if (priv->pass_through) {
+ STACK_WIND(frame, default_fsetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+ return 0;
+ }
- stub = fop_setxattr_stub (frame, default_setxattr_resume,
- loc, dict, flags, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (setxattr, frame, -1, ENOMEM, NULL);
- return 0;
- }
+ stub = fop_fsetxattr_stub(frame, default_fsetxattr_resume, fd, dict, flags,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(fsetxattr, frame, -1, ENOMEM, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int32_t
-quiesce_create (call_frame_t *frame, xlator_t *this,
- loc_t *loc, int32_t flags, mode_t mode,
- mode_t umask, fd_t *fd, dict_t *xdata)
+quiesce_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->pass_through) {
- /* Don't send O_APPEND below, as write() re-transmittions can
- fail with O_APPEND */
- STACK_WIND (frame, default_create_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->create,
- loc, (flags & ~O_APPEND), mode, umask, fd, xdata);
- return 0;
- }
+ if (priv->pass_through) {
+ STACK_WIND(frame, default_setxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, loc, dict, flags, xdata);
+ return 0;
+ }
- stub = fop_create_stub (frame, default_create_resume,
- loc, (flags & ~O_APPEND), mode, umask, fd, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (create, frame, -1, ENOMEM,
- NULL, NULL, NULL, NULL, NULL, NULL);
- return 0;
- }
+ stub = fop_setxattr_stub(frame, default_setxattr_resume, loc, dict, flags,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(setxattr, frame, -1, ENOMEM, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int32_t
-quiesce_link (call_frame_t *frame,
- xlator_t *this,
- loc_t *oldloc,
- loc_t *newloc, dict_t *xdata)
+quiesce_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->pass_through) {
- STACK_WIND (frame,
- default_link_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->link,
- oldloc, newloc, xdata);
- return 0;
- }
+ if (priv->pass_through) {
+ /* Don't send O_APPEND below, as write() re-transmittions can
+ fail with O_APPEND */
+ STACK_WIND(frame, default_create_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, (flags & ~O_APPEND),
+ mode, umask, fd, xdata);
+ return 0;
+ }
- stub = fop_link_stub (frame, default_link_resume, oldloc, newloc, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (link, frame, -1, ENOMEM,
- NULL, NULL, NULL, NULL, NULL);
- return 0;
- }
+ stub = fop_create_stub(frame, default_create_resume, loc,
+ (flags & ~O_APPEND), mode, umask, fd, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(create, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
+ NULL, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int32_t
-quiesce_rename (call_frame_t *frame,
- xlator_t *this,
- loc_t *oldloc,
- loc_t *newloc, dict_t *xdata)
+quiesce_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->pass_through) {
- STACK_WIND (frame,
- default_rename_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rename,
- oldloc, newloc, xdata);
- return 0;
- }
+ if (priv->pass_through) {
+ STACK_WIND(frame, default_link_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
+ return 0;
+ }
- stub = fop_rename_stub (frame, default_rename_resume, oldloc, newloc, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (rename, frame, -1, ENOMEM,
- NULL, NULL, NULL, NULL, NULL, NULL);
- return 0;
- }
+ stub = fop_link_stub(frame, default_link_resume, oldloc, newloc, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(link, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
+int32_t
+quiesce_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata)
+{
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+
+ priv = this->private;
+
+ if (priv->pass_through) {
+ STACK_WIND(frame, default_rename_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
+ return 0;
+ }
+
+ stub = fop_rename_stub(frame, default_rename_resume, oldloc, newloc, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(rename, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
+ NULL, NULL);
+ return 0;
+ }
+
+ gf_quiesce_enqueue(this, stub);
+
+ return 0;
+}
int
-quiesce_symlink (call_frame_t *frame, xlator_t *this,
- const char *linkpath, loc_t *loc, mode_t umask, dict_t *xdata)
+quiesce_symlink(call_frame_t *frame, xlator_t *this, const char *linkpath,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->pass_through) {
- STACK_WIND (frame, default_symlink_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->symlink,
- linkpath, loc, umask, xdata);
- return 0;
- }
+ if (priv->pass_through) {
+ STACK_WIND(frame, default_symlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->symlink, linkpath, loc, umask,
+ xdata);
+ return 0;
+ }
- stub = fop_symlink_stub (frame, default_symlink_resume,
- linkpath, loc, umask, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (symlink, frame, -1, ENOMEM,
- NULL, NULL, NULL, NULL, NULL);
- return 0;
- }
+ stub = fop_symlink_stub(frame, default_symlink_resume, linkpath, loc, umask,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(symlink, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
-
int
-quiesce_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags, dict_t *xdata)
+quiesce_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->pass_through) {
- STACK_WIND (frame, default_rmdir_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rmdir,
- loc, flags, xdata);
- return 0;
- }
+ if (priv->pass_through) {
+ STACK_WIND(frame, default_rmdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rmdir, loc, flags, xdata);
+ return 0;
+ }
- stub = fop_rmdir_stub (frame, default_rmdir_resume, loc, flags, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (rmdir, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
- }
+ stub = fop_rmdir_stub(frame, default_rmdir_resume, loc, flags, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(rmdir, frame, -1, ENOMEM, NULL, NULL, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int32_t
-quiesce_unlink (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc, int xflag, dict_t *xdata)
+quiesce_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->pass_through) {
- STACK_WIND (frame,
- default_unlink_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->unlink,
- loc, xflag, xdata);
- return 0;
- }
+ if (priv->pass_through) {
+ STACK_WIND(frame, default_unlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
+ return 0;
+ }
- stub = fop_unlink_stub (frame, default_unlink_resume, loc, xflag, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (unlink, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
- }
+ stub = fop_unlink_stub(frame, default_unlink_resume, loc, xflag, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(unlink, frame, -1, ENOMEM, NULL, NULL, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int
-quiesce_mkdir (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata)
+quiesce_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->pass_through) {
- STACK_WIND (frame, default_mkdir_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mkdir,
- loc, mode, umask, xdata);
- return 0;
- }
+ if (priv->pass_through) {
+ STACK_WIND(frame, default_mkdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
+ return 0;
+ }
- stub = fop_mkdir_stub (frame, default_mkdir_resume,
- loc, mode, umask, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (mkdir, frame, -1, ENOMEM,
- NULL, NULL, NULL, NULL, NULL);
- return 0;
- }
+ stub = fop_mkdir_stub(frame, default_mkdir_resume, loc, mode, umask, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(mkdir, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
-
int
-quiesce_mknod (call_frame_t *frame, xlator_t *this,
- loc_t *loc, mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata)
+quiesce_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->pass_through) {
- STACK_WIND (frame, default_mknod_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mknod,
- loc, mode, rdev, umask, xdata);
- return 0;
- }
+ if (priv->pass_through) {
+ STACK_WIND(frame, default_mknod_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, loc, mode, rdev, umask,
+ xdata);
+ return 0;
+ }
- stub = fop_mknod_stub (frame, default_mknod_resume,
- loc, mode, rdev, umask, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (mknod, frame, -1, ENOMEM,
- NULL, NULL, NULL, NULL, NULL);
- return 0;
- }
+ stub = fop_mknod_stub(frame, default_mknod_resume, loc, mode, rdev, umask,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(mknod, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int32_t
-quiesce_ftruncate (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- off_t offset, dict_t *xdata)
+quiesce_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv->pass_through) {
- STACK_WIND (frame,
- default_ftruncate_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->ftruncate,
- fd,
- offset, xdata);
- return 0;
- }
+ if (priv->pass_through) {
+ STACK_WIND(frame, default_ftruncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+ return 0;
+ }
- stub = fop_ftruncate_stub (frame, default_ftruncate_resume, fd, offset, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (ftruncate, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
- }
+ stub = fop_ftruncate_stub(frame, default_ftruncate_resume, fd, offset,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(ftruncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
/* Re-transmittion */
int32_t
-quiesce_readlink (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- size_t size, dict_t *xdata)
+quiesce_readlink(call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+ dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
-
- priv = this->private;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- loc_dup (loc, &local->loc);
- local->size = size;
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_readlink_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readlink,
- loc,
- size, xdata);
- return 0;
- }
+ priv = this->private;
- stub = fop_readlink_stub (frame, default_readlink_resume, loc, size, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (readlink, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ loc_dup(loc, &local->loc);
+ local->size = size;
+ frame->local = local;
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_readlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readlink, loc, size, xdata);
+ return 0;
+ }
+ stub = fop_readlink_stub(frame, default_readlink_resume, loc, size, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(readlink, frame, -1, ENOMEM, NULL, NULL, NULL);
return 0;
-}
+ }
+
+ gf_quiesce_enqueue(this, stub);
+ return 0;
+}
int32_t
-quiesce_access (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- int32_t mask, dict_t *xdata)
+quiesce_access(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+ dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- priv = this->private;
-
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- loc_dup (loc, &local->loc);
- local->flag = mask;
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_access_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->access,
- loc,
- mask, xdata);
- return 0;
- }
+ priv = this->private;
- stub = fop_access_stub (frame, default_access_resume, loc, mask, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (access, frame, -1, ENOMEM, NULL);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ loc_dup(loc, &local->loc);
+ local->flag = mask;
+ frame->local = local;
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_access_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->access, loc, mask, xdata);
+ return 0;
+ }
+ stub = fop_access_stub(frame, default_access_resume, loc, mask, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(access, frame, -1, ENOMEM, NULL);
return 0;
+ }
+
+ gf_quiesce_enqueue(this, stub);
+
+ return 0;
}
int32_t
-quiesce_fgetxattr (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- const char *name, dict_t *xdata)
+quiesce_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- local->fd = fd_ref (fd);
- if (name)
- local->name = gf_strdup (name);
-
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_fgetxattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fgetxattr,
- fd,
- name, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ local->fd = fd_ref(fd);
+ if (name)
+ local->name = gf_strdup(name);
- stub = fop_fgetxattr_stub (frame, default_fgetxattr_resume, fd, name, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (fgetxattr, frame, -1, ENOMEM, NULL, NULL);
- return 0;
- }
+ frame->local = local;
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_fgetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
+ return 0;
+ }
+ stub = fop_fgetxattr_stub(frame, default_fgetxattr_resume, fd, name, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(fgetxattr, frame, -1, ENOMEM, NULL, NULL);
return 0;
+ }
+
+ gf_quiesce_enqueue(this, stub);
+
+ return 0;
}
int32_t
-quiesce_statfs (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc, dict_t *xdata)
+quiesce_statfs(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- loc_dup (loc, &local->loc);
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_statfs_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->statfs,
- loc, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ loc_dup(loc, &local->loc);
+ frame->local = local;
- stub = fop_statfs_stub (frame, default_statfs_resume, loc, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (statfs, frame, -1, ENOMEM, NULL, NULL);
- return 0;
- }
-
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_statfs_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->statfs, loc, xdata);
+ return 0;
+ }
+ stub = fop_statfs_stub(frame, default_statfs_resume, loc, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(statfs, frame, -1, ENOMEM, NULL, NULL);
return 0;
+ }
+
+ gf_quiesce_enqueue(this, stub);
+
+ return 0;
}
int32_t
-quiesce_fsyncdir (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- int32_t flags, dict_t *xdata)
+quiesce_fsyncdir(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+ dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- local->fd = fd_ref (fd);
- local->flag = flags;
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_fsyncdir_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsyncdir,
- fd,
- flags, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ local->fd = fd_ref(fd);
+ local->flag = flags;
+ frame->local = local;
- stub = fop_fsyncdir_stub (frame, default_fsyncdir_resume, fd, flags, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (fsyncdir, frame, -1, ENOMEM, NULL);
- return 0;
- }
-
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_fsyncdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsyncdir, fd, flags, xdata);
+ return 0;
+ }
+ stub = fop_fsyncdir_stub(frame, default_fsyncdir_resume, fd, flags, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(fsyncdir, frame, -1, ENOMEM, NULL);
return 0;
+ }
+
+ gf_quiesce_enqueue(this, stub);
+
+ return 0;
}
int32_t
-quiesce_opendir (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc, fd_t *fd, dict_t *xdata)
+quiesce_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
-
- priv = this->private;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- loc_dup (loc, &local->loc);
- local->fd = fd_ref (fd);
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_opendir_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->opendir,
- loc, fd, xdata);
- return 0;
- }
+ priv = this->private;
- stub = fop_opendir_stub (frame, default_opendir_resume, loc, fd, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (opendir, frame, -1, ENOMEM, NULL, NULL);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ loc_dup(loc, &local->loc);
+ local->fd = fd_ref(fd);
+ frame->local = local;
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_opendir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->opendir, loc, fd, xdata);
+ return 0;
+ }
+ stub = fop_opendir_stub(frame, default_opendir_resume, loc, fd, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(opendir, frame, -1, ENOMEM, NULL, NULL);
return 0;
+ }
+
+ gf_quiesce_enqueue(this, stub);
+
+ return 0;
}
int32_t
-quiesce_fstat (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd, dict_t *xdata)
+quiesce_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
-
- priv = this->private;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- local->fd = fd_ref (fd);
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_fstat_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fstat,
- fd, xdata);
- return 0;
- }
+ priv = this->private;
- stub = fop_fstat_stub (frame, default_fstat_resume, fd, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (fstat, frame, -1, ENOMEM, NULL, NULL);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ local->fd = fd_ref(fd);
+ frame->local = local;
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_fstat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fstat, fd, xdata);
+ return 0;
+ }
+ stub = fop_fstat_stub(frame, default_fstat_resume, fd, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(fstat, frame, -1, ENOMEM, NULL, NULL);
return 0;
+ }
+
+ gf_quiesce_enqueue(this, stub);
+
+ return 0;
}
int32_t
-quiesce_fsync (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- int32_t flags, dict_t *xdata)
+quiesce_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+ dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- local->fd = fd_ref (fd);
- local->flag = flags;
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_fsync_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsync,
- fd,
- flags, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ local->fd = fd_ref(fd);
+ local->flag = flags;
+ frame->local = local;
- stub = fop_fsync_stub (frame, default_fsync_resume, fd, flags, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (fsync, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
- }
-
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_fsync_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsync, fd, flags, xdata);
+ return 0;
+ }
+ stub = fop_fsync_stub(frame, default_fsync_resume, fd, flags, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(fsync, frame, -1, ENOMEM, NULL, NULL, NULL);
return 0;
+ }
+
+ gf_quiesce_enqueue(this, stub);
+
+ return 0;
}
int32_t
-quiesce_flush (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd, dict_t *xdata)
+quiesce_flush(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- local->fd = fd_ref (fd);
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_flush_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->flush,
- fd, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ local->fd = fd_ref(fd);
+ frame->local = local;
- stub = fop_flush_stub (frame, default_flush_resume, fd, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (flush, frame, -1, ENOMEM, NULL);
- return 0;
- }
-
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_flush_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->flush, fd, xdata);
+ return 0;
+ }
+ stub = fop_flush_stub(frame, default_flush_resume, fd, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(flush, frame, -1, ENOMEM, NULL);
return 0;
+ }
+
+ gf_quiesce_enqueue(this, stub);
+
+ return 0;
}
int32_t
-quiesce_writev (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- struct iovec *vector,
- int32_t count,
- off_t off, uint32_t flags,
- struct iobref *iobref, dict_t *xdata)
+quiesce_writev(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int32_t count, off_t off, uint32_t flags,
+ struct iobref *iobref, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- STACK_WIND (frame,
- default_writev_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->writev,
- fd,
- vector,
- count,
- off, flags,
- iobref, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ STACK_WIND(frame, default_writev_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector, count, off,
+ flags, iobref, xdata);
+ return 0;
+ }
- stub = fop_writev_stub (frame, default_writev_resume,
- fd, vector, count, off, flags, iobref, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (writev, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
- }
+ stub = fop_writev_stub(frame, default_writev_resume, fd, vector, count, off,
+ flags, iobref, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(writev, frame, -1, ENOMEM, NULL, NULL, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int32_t
-quiesce_readv (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- size_t size,
- off_t offset, uint32_t flags, dict_t *xdata)
+quiesce_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- local->fd = fd_ref (fd);
- local->size = size;
- local->offset = offset;
- local->io_flag = flags;
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_readv_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readv,
- fd,
- size,
- offset, flags, xdata);
- return 0;
- }
-
- stub = fop_readv_stub (frame, default_readv_resume, fd, size, offset,
- flags, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (readv, frame, -1, ENOMEM,
- NULL, 0, NULL, NULL, NULL);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ local->fd = fd_ref(fd);
+ local->size = size;
+ local->offset = offset;
+ local->io_flag = flags;
+ frame->local = local;
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_readv_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readv, fd, size, offset, flags,
+ xdata);
+ return 0;
+ }
+ stub = fop_readv_stub(frame, default_readv_resume, fd, size, offset, flags,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(readv, frame, -1, ENOMEM, NULL, 0, NULL, NULL,
+ NULL);
return 0;
-}
+ }
+ gf_quiesce_enqueue(this, stub);
+
+ return 0;
+}
int32_t
-quiesce_open (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- int32_t flags, fd_t *fd,
- dict_t *xdata)
+quiesce_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- loc_dup (loc, &local->loc);
- local->fd = fd_ref (fd);
-
- /* Don't send O_APPEND below, as write() re-transmittions can
- fail with O_APPEND */
- local->flag = (flags & ~O_APPEND);
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_open_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->open,
- loc, (flags & ~O_APPEND), fd, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ loc_dup(loc, &local->loc);
+ local->fd = fd_ref(fd);
- stub = fop_open_stub (frame, default_open_resume, loc,
- (flags & ~O_APPEND), fd, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (open, frame, -1, ENOMEM, NULL, NULL);
- return 0;
- }
+ /* Don't send O_APPEND below, as write() re-transmittions can
+ fail with O_APPEND */
+ local->flag = (flags & ~O_APPEND);
+ frame->local = local;
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_open_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->open, loc, (flags & ~O_APPEND), fd,
+ xdata);
+ return 0;
+ }
+ stub = fop_open_stub(frame, default_open_resume, loc, (flags & ~O_APPEND),
+ fd, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(open, frame, -1, ENOMEM, NULL, NULL);
return 0;
+ }
+
+ gf_quiesce_enqueue(this, stub);
+
+ return 0;
}
int32_t
-quiesce_getxattr (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- const char *name, dict_t *xdata)
+quiesce_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- loc_dup (loc, &local->loc);
- if (name)
- local->name = gf_strdup (name);
-
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_getxattr_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr,
- loc,
- name, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ loc_dup(loc, &local->loc);
+ if (name)
+ local->name = gf_strdup(name);
- stub = fop_getxattr_stub (frame, default_getxattr_resume, loc, name, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (getxattr, frame, -1, ENOMEM, NULL, NULL);
- return 0;
- }
+ frame->local = local;
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_getxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+ return 0;
+ }
+ stub = fop_getxattr_stub(frame, default_getxattr_resume, loc, name, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(getxattr, frame, -1, ENOMEM, NULL, NULL);
return 0;
-}
+ }
+
+ gf_quiesce_enqueue(this, stub);
+ return 0;
+}
int32_t
-quiesce_xattrop (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- gf_xattrop_flags_t flags,
- dict_t *dict, dict_t *xdata)
+quiesce_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- STACK_WIND (frame,
- default_xattrop_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->xattrop,
- loc,
- flags,
- dict, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ STACK_WIND(frame, default_xattrop_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->xattrop, loc, flags, dict, xdata);
+ return 0;
+ }
- stub = fop_xattrop_stub (frame, default_xattrop_resume,
- loc, flags, dict, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (xattrop, frame, -1, ENOMEM, NULL, NULL);
- return 0;
- }
+ stub = fop_xattrop_stub(frame, default_xattrop_resume, loc, flags, dict,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(xattrop, frame, -1, ENOMEM, NULL, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int32_t
-quiesce_fxattrop (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- gf_xattrop_flags_t flags,
- dict_t *dict, dict_t *xdata)
+quiesce_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- STACK_WIND (frame,
- default_fxattrop_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fxattrop,
- fd,
- flags,
- dict, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ STACK_WIND(frame, default_fxattrop_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fxattrop, fd, flags, dict, xdata);
+ return 0;
+ }
- stub = fop_fxattrop_stub (frame, default_fxattrop_resume,
- fd, flags, dict, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (fxattrop, frame, -1, ENOMEM, NULL, NULL);
- return 0;
- }
+ stub = fop_fxattrop_stub(frame, default_fxattrop_resume, fd, flags, dict,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(fxattrop, frame, -1, ENOMEM, NULL, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int32_t
-quiesce_lk (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- int32_t cmd,
- struct gf_flock *lock, dict_t *xdata)
+quiesce_lk(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
+ struct gf_flock *lock, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- STACK_WIND (frame,
- default_lk_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lk,
- fd,
- cmd,
- lock, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ STACK_WIND(frame, default_lk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lk, fd, cmd, lock, xdata);
+ return 0;
+ }
- stub = fop_lk_stub (frame, default_lk_resume, fd, cmd, lock, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (lk, frame, -1, ENOMEM, NULL, NULL);
- return 0;
- }
+ stub = fop_lk_stub(frame, default_lk_resume, fd, cmd, lock, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(lk, frame, -1, ENOMEM, NULL, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
-
int32_t
-quiesce_inodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, int32_t cmd,
- struct gf_flock *lock, dict_t *xdata)
+quiesce_inodelk(call_frame_t *frame, xlator_t *this, const char *volume,
+ loc_t *loc, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- STACK_WIND (frame,
- default_inodelk_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->inodelk,
- volume, loc, cmd, lock, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ STACK_WIND(frame, default_inodelk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->inodelk, volume, loc, cmd, lock,
+ xdata);
+ return 0;
+ }
- stub = fop_inodelk_stub (frame, default_inodelk_resume,
- volume, loc, cmd, lock, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (inodelk, frame, -1, ENOMEM, NULL);
- return 0;
- }
+ stub = fop_inodelk_stub(frame, default_inodelk_resume, volume, loc, cmd,
+ lock, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(inodelk, frame, -1, ENOMEM, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int32_t
-quiesce_finodelk (call_frame_t *frame, xlator_t *this,
- const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
+quiesce_finodelk(call_frame_t *frame, xlator_t *this, const char *volume,
+ fd_t *fd, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- STACK_WIND (frame,
- default_finodelk_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->finodelk,
- volume, fd, cmd, lock, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ STACK_WIND(frame, default_finodelk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->finodelk, volume, fd, cmd, lock,
+ xdata);
+ return 0;
+ }
- stub = fop_finodelk_stub (frame, default_finodelk_resume,
- volume, fd, cmd, lock, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (finodelk, frame, -1, ENOMEM, NULL);
- return 0;
- }
+ stub = fop_finodelk_stub(frame, default_finodelk_resume, volume, fd, cmd,
+ lock, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(finodelk, frame, -1, ENOMEM, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int32_t
-quiesce_entrylk (call_frame_t *frame, xlator_t *this,
- const char *volume, loc_t *loc, const char *basename,
- entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
+quiesce_entrylk(call_frame_t *frame, xlator_t *this, const char *volume,
+ loc_t *loc, const char *basename, entrylk_cmd cmd,
+ entrylk_type type, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- STACK_WIND (frame, default_entrylk_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->entrylk,
- volume, loc, basename, cmd, type, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ STACK_WIND(frame, default_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, volume, loc, basename, cmd,
+ type, xdata);
+ return 0;
+ }
- stub = fop_entrylk_stub (frame, default_entrylk_resume,
- volume, loc, basename, cmd, type, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (entrylk, frame, -1, ENOMEM, NULL);
- return 0;
- }
+ stub = fop_entrylk_stub(frame, default_entrylk_resume, volume, loc,
+ basename, cmd, type, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(entrylk, frame, -1, ENOMEM, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int32_t
-quiesce_fentrylk (call_frame_t *frame, xlator_t *this,
- const char *volume, fd_t *fd, const char *basename,
- entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
+quiesce_fentrylk(call_frame_t *frame, xlator_t *this, const char *volume,
+ fd_t *fd, const char *basename, entrylk_cmd cmd,
+ entrylk_type type, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- STACK_WIND (frame, default_fentrylk_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fentrylk,
- volume, fd, basename, cmd, type, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ STACK_WIND(frame, default_fentrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fentrylk, volume, fd, basename, cmd,
+ type, xdata);
+ return 0;
+ }
- stub = fop_fentrylk_stub (frame, default_fentrylk_resume,
- volume, fd, basename, cmd, type, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (fentrylk, frame, -1, ENOMEM, NULL);
- return 0;
- }
+ stub = fop_fentrylk_stub(frame, default_fentrylk_resume, volume, fd,
+ basename, cmd, type, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(fentrylk, frame, -1, ENOMEM, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int32_t
-quiesce_rchecksum (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd, off_t offset,
- int32_t len, dict_t *xdata)
+quiesce_rchecksum(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ int32_t len, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- priv = this->private;
-
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- local->fd = fd_ref (fd);
- local->offset = offset;
- local->flag = len;
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_rchecksum_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rchecksum,
- fd, offset, len, xdata);
- return 0;
- }
+ priv = this->private;
- stub = fop_rchecksum_stub (frame, default_rchecksum_resume,
- fd, offset, len, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (rchecksum, frame, -1, ENOMEM, 0, NULL, NULL);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ local->fd = fd_ref(fd);
+ local->offset = offset;
+ local->flag = len;
+ frame->local = local;
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_rchecksum_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rchecksum, fd, offset, len, xdata);
+ return 0;
+ }
+ stub = fop_rchecksum_stub(frame, default_rchecksum_resume, fd, offset, len,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(rchecksum, frame, -1, ENOMEM, 0, NULL, NULL);
return 0;
-}
+ }
+
+ gf_quiesce_enqueue(this, stub);
+ return 0;
+}
int32_t
-quiesce_readdir (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- size_t size,
- off_t off, dict_t *xdata)
+quiesce_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- local->fd = fd_ref (fd);
- local->size = size;
- local->offset = off;
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_readdir_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdir,
- fd, size, off, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ local->fd = fd_ref(fd);
+ local->size = size;
+ local->offset = off;
+ frame->local = local;
- stub = fop_readdir_stub (frame, default_readdir_resume, fd, size, off, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (readdir, frame, -1, ENOMEM, NULL, NULL);
- return 0;
- }
-
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_readdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdir, fd, size, off, xdata);
+ return 0;
+ }
+ stub = fop_readdir_stub(frame, default_readdir_resume, fd, size, off,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(readdir, frame, -1, ENOMEM, NULL, NULL);
return 0;
-}
+ }
+ gf_quiesce_enqueue(this, stub);
+
+ return 0;
+}
int32_t
-quiesce_readdirp (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- size_t size,
- off_t off, dict_t *dict)
+quiesce_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *dict)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- priv = this->private;
-
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- local->fd = fd_ref (fd);
- local->size = size;
- local->offset = off;
- local->dict = dict_ref (dict);
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_readdirp_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdirp,
- fd, size, off, dict);
- return 0;
- }
+ priv = this->private;
- stub = fop_readdirp_stub (frame, default_readdirp_resume, fd, size,
- off, dict);
- if (!stub) {
- STACK_UNWIND_STRICT (readdirp, frame, -1, ENOMEM, NULL, NULL);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ local->fd = fd_ref(fd);
+ local->size = size;
+ local->offset = off;
+ local->dict = dict_ref(dict);
+ frame->local = local;
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_readdirp_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdirp, fd, size, off, dict);
+ return 0;
+ }
+ stub = fop_readdirp_stub(frame, default_readdirp_resume, fd, size, off,
+ dict);
+ if (!stub) {
+ STACK_UNWIND_STRICT(readdirp, frame, -1, ENOMEM, NULL, NULL);
return 0;
+ }
+
+ gf_quiesce_enqueue(this, stub);
+
+ return 0;
}
int32_t
-quiesce_setattr (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- struct iatt *stbuf,
- int32_t valid, dict_t *xdata)
+quiesce_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- STACK_WIND (frame,
- default_setattr_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setattr,
- loc, stbuf, valid, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ STACK_WIND(frame, default_setattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid, xdata);
+ return 0;
+ }
- stub = fop_setattr_stub (frame, default_setattr_resume,
- loc, stbuf, valid, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (setattr, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
- }
+ stub = fop_setattr_stub(frame, default_setattr_resume, loc, stbuf, valid,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(setattr, frame, -1, ENOMEM, NULL, NULL, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
-
int32_t
-quiesce_stat (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc, dict_t *xdata)
+quiesce_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- priv = this->private;
-
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- loc_dup (loc, &local->loc);
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_stat_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->stat,
- loc, xdata);
- return 0;
- }
+ priv = this->private;
- stub = fop_stat_stub (frame, default_stat_resume, loc, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (stat, frame, -1, ENOMEM, NULL, NULL);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ loc_dup(loc, &local->loc);
+ frame->local = local;
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_stat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->stat, loc, xdata);
+ return 0;
+ }
+ stub = fop_stat_stub(frame, default_stat_resume, loc, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(stat, frame, -1, ENOMEM, NULL, NULL);
return 0;
+ }
+
+ gf_quiesce_enqueue(this, stub);
+
+ return 0;
}
int32_t
-quiesce_lookup (call_frame_t *frame,
- xlator_t *this,
- loc_t *loc,
- dict_t *xattr_req)
+quiesce_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ dict_t *xattr_req)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- priv = this->private;
-
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- loc_dup (loc, &local->loc);
- local->dict = dict_ref (xattr_req);
- frame->local = local;
-
- STACK_WIND (frame,
- quiesce_lookup_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup,
- loc, xattr_req);
- return 0;
- }
+ priv = this->private;
- stub = fop_lookup_stub (frame, default_lookup_resume, loc, xattr_req);
- if (!stub) {
- STACK_UNWIND_STRICT (lookup, frame, -1, ENOMEM,
- NULL, NULL, NULL, NULL);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ loc_dup(loc, &local->loc);
+ local->dict = dict_ref(xattr_req);
+ frame->local = local;
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
+ return 0;
+ }
+ stub = fop_lookup_stub(frame, default_lookup_resume, loc, xattr_req);
+ if (!stub) {
+ STACK_UNWIND_STRICT(lookup, frame, -1, ENOMEM, NULL, NULL, NULL, NULL);
return 0;
+ }
+
+ gf_quiesce_enqueue(this, stub);
+
+ return 0;
}
int32_t
-quiesce_fsetattr (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd,
- struct iatt *stbuf,
- int32_t valid, dict_t *xdata)
+quiesce_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- STACK_WIND (frame,
- default_fsetattr_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fsetattr,
- fd, stbuf, valid, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ STACK_WIND(frame, default_fsetattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetattr, fd, stbuf, valid, xdata);
+ return 0;
+ }
- stub = fop_fsetattr_stub (frame, default_fsetattr_resume,
- fd, stbuf, valid, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (fsetattr, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
- }
+ stub = fop_fsetattr_stub(frame, default_fsetattr_resume, fd, stbuf, valid,
+ xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(fsetattr, frame, -1, ENOMEM, NULL, NULL, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
int32_t
-quiesce_fallocate (call_frame_t *frame,
- xlator_t *this,
- fd_t *fd, int32_t mode,
- off_t offset, size_t len, dict_t *xdata)
+quiesce_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+ off_t offset, size_t len, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- local->fd = fd_ref (fd);
- local->offset = offset;
- local->len = len;
- local->flag = mode;
-
- frame->local = local;
-
- STACK_WIND (frame,
- default_fallocate_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fallocate,
- fd, mode, offset, len, xdata);
- return 0;
- }
+ if (priv && priv->pass_through) {
+ STACK_WIND(frame, default_fallocate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fallocate, fd, mode, offset, len,
+ xdata);
+ return 0;
+ }
- stub = fop_fallocate_stub (frame, default_fallocate_resume, fd,
- mode, offset, len, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (fallocate, frame, -1, ENOMEM,
- NULL, NULL, NULL);
- return 0;
- }
+ stub = fop_fallocate_stub(frame, default_fallocate_resume, fd, mode, offset,
+ len, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(fallocate, frame, -1, ENOMEM, NULL, NULL, NULL);
+ return 0;
+ }
- gf_quiesce_enqueue (this, stub);
+ gf_quiesce_enqueue(this, stub);
- return 0;
+ return 0;
}
-
int
-quiesce_seek_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, off_t offset,
- dict_t *xdata)
-{
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
-
- local = frame->local;
- frame->local = NULL;
- if ((op_ret == -1) && (op_errno == ENOTCONN)) {
- /* Re-transmit (by putting in the queue) */
- stub = fop_seek_stub (frame, default_seek_resume,
- local->fd, local->offset,
- local->what, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (seek, frame, -1, ENOMEM, 0, NULL);
- goto out;
- }
-
- gf_quiesce_enqueue (this, stub);
- goto out;
+quiesce_seek_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, off_t offset, dict_t *xdata)
+{
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
+
+ local = frame->local;
+ frame->local = NULL;
+ if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+ /* Re-transmit (by putting in the queue) */
+ stub = fop_seek_stub(frame, default_seek_resume, local->fd,
+ local->offset, local->what, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(seek, frame, -1, ENOMEM, 0, NULL);
+ goto out;
}
- STACK_UNWIND_STRICT (seek, frame, op_ret, op_errno, offset, xdata);
+ gf_quiesce_enqueue(this, stub);
+ goto out;
+ }
+
+ STACK_UNWIND_STRICT(seek, frame, op_ret, op_errno, offset, xdata);
out:
- gf_quiesce_local_wipe (this, local);
+ gf_quiesce_local_wipe(this, local);
- return 0;
+ return 0;
}
int
-quiesce_seek (call_frame_t *frame, xlator_t *this, fd_t *fd,
- off_t offset, gf_seek_what_t what, dict_t *xdata)
+quiesce_seek(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ gf_seek_what_t what, dict_t *xdata)
{
- quiesce_priv_t *priv = NULL;
- call_stub_t *stub = NULL;
- quiesce_local_t *local = NULL;
+ quiesce_priv_t *priv = NULL;
+ call_stub_t *stub = NULL;
+ quiesce_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- if (priv && priv->pass_through) {
- local = mem_get0 (priv->local_pool);
- local->fd = fd_ref (fd);
- local->offset = offset;
- local->what = what;
+ if (priv && priv->pass_through) {
+ local = mem_get0(priv->local_pool);
+ local->fd = fd_ref(fd);
+ local->offset = offset;
+ local->what = what;
- frame->local = local;
+ frame->local = local;
- STACK_WIND (frame,
- quiesce_seek_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->seek,
- fd, offset, what, xdata);
- return 0;
- }
-
- stub = fop_seek_stub (frame, default_seek_resume, fd,
- offset, what, xdata);
- if (!stub) {
- STACK_UNWIND_STRICT (seek, frame, -1, ENOMEM, 0, NULL);
- return 0;
- }
-
- gf_quiesce_enqueue (this, stub);
+ STACK_WIND(frame, quiesce_seek_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->seek, fd, offset, what, xdata);
+ return 0;
+ }
+ stub = fop_seek_stub(frame, default_seek_resume, fd, offset, what, xdata);
+ if (!stub) {
+ STACK_UNWIND_STRICT(seek, frame, -1, ENOMEM, 0, NULL);
return 0;
-}
+ }
+ gf_quiesce_enqueue(this, stub);
+ return 0;
+}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- ret = xlator_mem_acct_init (this, gf_quiesce_mt_end + 1);
+ ret = xlator_mem_acct_init(this, gf_quiesce_mt_end + 1);
- return ret;
+ return ret;
}
int
-reconfigure (xlator_t *this, dict_t *options)
+reconfigure(xlator_t *this, dict_t *options)
{
- int32_t ret = -1;
- quiesce_priv_t *priv = NULL;
+ int32_t ret = -1;
+ quiesce_priv_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- GF_OPTION_RECONF("timeout", priv->timeout, options, time, out);
- GF_OPTION_RECONF ("failover-hosts", priv->failover_hosts, options,
- str, out);
- gf_quiesce_populate_failover_hosts (this, priv, priv->failover_hosts);
+ GF_OPTION_RECONF("timeout", priv->timeout, options, time, out);
+ GF_OPTION_RECONF("failover-hosts", priv->failover_hosts, options, str, out);
+ gf_quiesce_populate_failover_hosts(this, priv, priv->failover_hosts);
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-init (xlator_t *this)
+init(xlator_t *this)
{
- int ret = -1;
- quiesce_priv_t *priv = NULL;
+ int ret = -1;
+ quiesce_priv_t *priv = NULL;
- if (!this->children || this->children->next) {
- gf_log (this->name, GF_LOG_ERROR,
- "'quiesce' not configured with exactly one child");
- goto out;
- }
+ if (!this->children || this->children->next) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "'quiesce' not configured with exactly one child");
+ goto out;
+ }
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "dangling volume. check volfile ");
- }
+ if (!this->parents) {
+ gf_log(this->name, GF_LOG_WARNING, "dangling volume. check volfile ");
+ }
- priv = GF_CALLOC (1, sizeof (*priv), gf_quiesce_mt_priv_t);
- if (!priv)
- goto out;
+ priv = GF_CALLOC(1, sizeof(*priv), gf_quiesce_mt_priv_t);
+ if (!priv)
+ goto out;
- INIT_LIST_HEAD (&priv->failover_list);
+ INIT_LIST_HEAD(&priv->failover_list);
- GF_OPTION_INIT ("timeout", priv->timeout, time, out);
- GF_OPTION_INIT ("failover-hosts", priv->failover_hosts, str, out);
- gf_quiesce_populate_failover_hosts (this, priv, priv->failover_hosts);
+ GF_OPTION_INIT("timeout", priv->timeout, time, out);
+ GF_OPTION_INIT("failover-hosts", priv->failover_hosts, str, out);
+ gf_quiesce_populate_failover_hosts(this, priv, priv->failover_hosts);
- priv->local_pool = mem_pool_new (quiesce_local_t,
- GF_FOPS_EXPECTED_IN_PARALLEL);
+ priv->local_pool = mem_pool_new(quiesce_local_t,
+ GF_FOPS_EXPECTED_IN_PARALLEL);
- LOCK_INIT (&priv->lock);
- priv->pass_through = _gf_false;
+ LOCK_INIT(&priv->lock);
+ priv->pass_through = _gf_false;
- INIT_LIST_HEAD (&priv->req);
+ INIT_LIST_HEAD(&priv->req);
- this->private = priv;
- ret = 0;
+ this->private = priv;
+ ret = 0;
out:
- return ret;
+ return ret;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- quiesce_priv_t *priv = NULL;
+ quiesce_priv_t *priv = NULL;
- priv = this->private;
- if (!priv)
- goto out;
- this->private = NULL;
+ priv = this->private;
+ if (!priv)
+ goto out;
+ this->private = NULL;
- mem_pool_destroy (priv->local_pool);
- LOCK_DESTROY (&priv->lock);
- GF_FREE (priv);
+ mem_pool_destroy(priv->local_pool);
+ priv->local_pool = NULL;
+ LOCK_DESTROY(&priv->lock);
+ GF_FREE(priv);
out:
- return;
+ return;
}
int
-notify (xlator_t *this, int event, void *data, ...)
-{
- int ret = 0;
- quiesce_priv_t *priv = NULL;
-
- priv = this->private;
- if (!priv)
- goto out;
-
- switch (event) {
- case GF_EVENT_CHILD_UP:
- {
- ret = gf_thread_create (&priv->thr, NULL,
- gf_quiesce_dequeue_start,
- this, "quiesce");
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "failed to create the quiesce-dequeue thread");
- }
-
- LOCK (&priv->lock);
- {
- priv->pass_through = _gf_true;
- }
- UNLOCK (&priv->lock);
- break;
+notify(xlator_t *this, int event, void *data, ...)
+{
+ int ret = 0;
+ quiesce_priv_t *priv = NULL;
+
+ priv = this->private;
+ if (!priv)
+ goto out;
+
+ switch (event) {
+ case GF_EVENT_CHILD_UP: {
+ ret = gf_thread_create(&priv->thr, NULL, gf_quiesce_dequeue_start,
+ this, "quiesce");
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to create the quiesce-dequeue thread");
+ }
+
+ LOCK(&priv->lock);
+ {
+ priv->pass_through = _gf_true;
+ }
+ UNLOCK(&priv->lock);
+ break;
}
case GF_EVENT_CHILD_DOWN:
- LOCK (&priv->lock);
- {
- priv->pass_through = _gf_false;
- __gf_quiesce_start_timer (this, priv);
-
- }
- UNLOCK (&priv->lock);
- break;
+ LOCK(&priv->lock);
+ {
+ priv->pass_through = _gf_false;
+ __gf_quiesce_start_timer(this, priv);
+ }
+ UNLOCK(&priv->lock);
+ break;
default:
- break;
- }
+ break;
+ }
- ret = default_notify (this, event, data);
+ ret = default_notify(this, event, data);
out:
- return ret;
+ return ret;
}
-
struct xlator_fops fops = {
- /* write/modifying fops */
- .mknod = quiesce_mknod,
- .create = quiesce_create,
- .truncate = quiesce_truncate,
- .ftruncate = quiesce_ftruncate,
- .setxattr = quiesce_setxattr,
- .removexattr = quiesce_removexattr,
- .symlink = quiesce_symlink,
- .unlink = quiesce_unlink,
- .link = quiesce_link,
- .mkdir = quiesce_mkdir,
- .rmdir = quiesce_rmdir,
- .rename = quiesce_rename,
- .fallocate = quiesce_fallocate,
-
- /* The below calls are known to change state, hence
- re-transmittion is not advised */
- .lk = quiesce_lk,
- .inodelk = quiesce_inodelk,
- .finodelk = quiesce_finodelk,
- .entrylk = quiesce_entrylk,
- .fentrylk = quiesce_fentrylk,
- .xattrop = quiesce_xattrop,
- .fxattrop = quiesce_fxattrop,
- .setattr = quiesce_setattr,
- .fsetattr = quiesce_fsetattr,
-
- /* Special case, re-transmittion is not harmful *
- * as offset is properly sent from above layers */
- /* TODO: not re-transmitted as of now */
- .writev = quiesce_writev,
-
- /* re-transmittable fops */
- .lookup = quiesce_lookup,
- .stat = quiesce_stat,
- .fstat = quiesce_fstat,
- .access = quiesce_access,
- .readlink = quiesce_readlink,
- .getxattr = quiesce_getxattr,
- .open = quiesce_open,
- .readv = quiesce_readv,
- .flush = quiesce_flush,
- .fsync = quiesce_fsync,
- .statfs = quiesce_statfs,
- .opendir = quiesce_opendir,
- .readdir = quiesce_readdir,
- .readdirp = quiesce_readdirp,
- .fsyncdir = quiesce_fsyncdir,
- .seek = quiesce_seek,
+ /* write/modifying fops */
+ .mknod = quiesce_mknod,
+ .create = quiesce_create,
+ .truncate = quiesce_truncate,
+ .ftruncate = quiesce_ftruncate,
+ .setxattr = quiesce_setxattr,
+ .fsetxattr = quiesce_fsetxattr,
+ .removexattr = quiesce_removexattr,
+ .fremovexattr = quiesce_fremovexattr,
+ .symlink = quiesce_symlink,
+ .unlink = quiesce_unlink,
+ .link = quiesce_link,
+ .mkdir = quiesce_mkdir,
+ .rmdir = quiesce_rmdir,
+ .rename = quiesce_rename,
+ .fallocate = quiesce_fallocate,
+
+ /* The below calls are known to change state, hence
+ re-transmittion is not advised */
+ .lk = quiesce_lk,
+ .inodelk = quiesce_inodelk,
+ .finodelk = quiesce_finodelk,
+ .entrylk = quiesce_entrylk,
+ .fentrylk = quiesce_fentrylk,
+ .xattrop = quiesce_xattrop,
+ .fxattrop = quiesce_fxattrop,
+ .setattr = quiesce_setattr,
+ .fsetattr = quiesce_fsetattr,
+
+ /* Special case, re-transmittion is not harmful *
+ * as offset is properly sent from above layers */
+ /* TODO: not re-transmitted as of now */
+ .writev = quiesce_writev,
+
+ /* re-transmittable fops */
+ .lookup = quiesce_lookup,
+ .stat = quiesce_stat,
+ .fstat = quiesce_fstat,
+ .access = quiesce_access,
+ .readlink = quiesce_readlink,
+ .getxattr = quiesce_getxattr,
+ .fgetxattr = quiesce_fgetxattr,
+ .open = quiesce_open,
+ .readv = quiesce_readv,
+ .flush = quiesce_flush,
+ .fsync = quiesce_fsync,
+ .statfs = quiesce_statfs,
+ .opendir = quiesce_opendir,
+ .readdir = quiesce_readdir,
+ .readdirp = quiesce_readdirp,
+ .fsyncdir = quiesce_fsyncdir,
+ .seek = quiesce_seek,
};
struct xlator_dumpops dumpops;
-
struct xlator_cbks cbks;
-
struct volume_options options[] = {
- { .key = {"timeout"},
- .type = GF_OPTION_TYPE_TIME,
- .default_value = "45",
- .description = "After 'timeout' seconds since the time 'quiesce' "
- "option was set to \"!pass-through\", acknowledgements to file "
- "operations are no longer quiesced and previously "
- "quiesced acknowledgements are sent to the application",
- .op_version = { GD_OP_VERSION_4_0_0 },
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- },
- { .key = {"failover-hosts"},
- .type = GF_OPTION_TYPE_INTERNET_ADDRESS_LIST,
- .op_version = { GD_OP_VERSION_4_0_0 },
- .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .description = "It is a comma separated list of hostname/IP "
- "addresses. It Specifies the list of hosts where "
- "the gfproxy daemons are running, to which the "
- "the thin clients can failover to."
- },
- { .key = {NULL} },
+ {
+ .key = {"timeout"},
+ .type = GF_OPTION_TYPE_TIME,
+ .default_value = "45",
+ .description =
+ "After 'timeout' seconds since the time 'quiesce' "
+ "option was set to \"!pass-through\", acknowledgements to file "
+ "operations are no longer quiesced and previously "
+ "quiesced acknowledgements are sent to the application",
+ .op_version = {GD_OP_VERSION_4_0_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ },
+ {.key = {"failover-hosts"},
+ .type = GF_OPTION_TYPE_INTERNET_ADDRESS_LIST,
+ .op_version = {GD_OP_VERSION_4_0_0},
+ .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .description = "It is a comma separated list of hostname/IP "
+ "addresses. It Specifies the list of hosts where "
+ "the gfproxy daemons are running, to which the "
+ "the thin clients can failover to."},
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .notify = notify,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {GD_OP_VERSION_3_12_0},
+ .dumpops = &dumpops,
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "quiesce",
+ .category = GF_TECH_PREVIEW,
};
diff --git a/xlators/features/quiesce/src/quiesce.h b/xlators/features/quiesce/src/quiesce.h
index c084801c6c6..6ab2af40a56 100644
--- a/xlators/features/quiesce/src/quiesce.h
+++ b/xlators/features/quiesce/src/quiesce.h
@@ -13,53 +13,53 @@
#include "quiesce-mem-types.h"
#include "quiesce-messages.h"
-#include "xlator.h"
-#include "timer.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/timer.h>
#define GF_FOPS_EXPECTED_IN_PARALLEL 512
typedef struct {
- struct list_head list;
- char *addr;
- gf_boolean_t tried; /* indicates attempted connecting */
+ struct list_head list;
+ char *addr;
+ gf_boolean_t tried; /* indicates attempted connecting */
} quiesce_failover_hosts_t;
typedef struct {
- gf_timer_t *timer;
- gf_boolean_t pass_through;
- gf_lock_t lock;
- struct list_head req;
- int queue_size;
- pthread_t thr;
- struct mem_pool *local_pool;
- uint32_t timeout;
- char *failover_hosts;
- struct list_head failover_list;
+ gf_timer_t *timer;
+ gf_boolean_t pass_through;
+ gf_lock_t lock;
+ struct list_head req;
+ int queue_size;
+ pthread_t thr;
+ struct mem_pool *local_pool;
+ uint32_t timeout;
+ char *failover_hosts;
+ struct list_head failover_list;
} quiesce_priv_t;
typedef struct {
- fd_t *fd;
- char *name;
- char *volname;
- loc_t loc;
- off_t size;
- off_t offset;
- mode_t mode;
- int32_t flag;
- struct iatt stbuf;
- struct iovec *vector;
- struct iobref *iobref;
- dict_t *dict;
- struct gf_flock flock;
- entrylk_cmd cmd;
- entrylk_type type;
- gf_xattrop_flags_t xattrop_flags;
- int32_t wbflags;
- uint32_t io_flag;
- /* for fallocate */
- size_t len;
- /* for lseek */
- gf_seek_what_t what;
+ fd_t *fd;
+ char *name;
+ char *volname;
+ loc_t loc;
+ off_t size;
+ off_t offset;
+ mode_t mode;
+ int32_t flag;
+ struct iatt stbuf;
+ struct iovec *vector;
+ struct iobref *iobref;
+ dict_t *dict;
+ struct gf_flock flock;
+ entrylk_cmd cmd;
+ entrylk_type type;
+ gf_xattrop_flags_t xattrop_flags;
+ int32_t wbflags;
+ uint32_t io_flag;
+ /* for fallocate */
+ size_t len;
+ /* for lseek */
+ gf_seek_what_t what;
} quiesce_local_t;
#endif
diff --git a/xlators/features/quota/src/Makefile.am b/xlators/features/quota/src/Makefile.am
index ec20e3ea49d..1c2dcef0ca3 100644
--- a/xlators/features/quota/src/Makefile.am
+++ b/xlators/features/quota/src/Makefile.am
@@ -4,7 +4,7 @@ endif
xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
quota_la_LDFLAGS = -module $(GF_XLATOR_DEFAULT_LDFLAGS)
-quotad_la_LDFLAGS = -module -avoid-version -export-symbols $(top_srcdir)/xlators/features/quota/src/quotad.sym
+quotad_la_LDFLAGS = -module $(GF_XLATOR_DEFAULT_LDFLAGS)
quota_la_SOURCES = quota.c quota-enforcer-client.c
quota_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la \
@@ -27,6 +27,3 @@ AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
AM_CFLAGS = -Wall $(GF_CFLAGS)
CLEANFILES =
-
-EXTRA_DIST = quotad.sym
-
diff --git a/xlators/features/quota/src/quota-enforcer-client.c b/xlators/features/quota/src/quota-enforcer-client.c
index 8011bacdf8c..480d64ade27 100644
--- a/xlators/features/quota/src/quota-enforcer-client.c
+++ b/xlators/features/quota/src/quota-enforcer-client.c
@@ -32,460 +32,472 @@
#include <malloc.h>
#endif
-#ifdef HAVE_MALLOC_STATS
-#ifdef DEBUG
-#include <mcheck.h>
-#endif
-#endif
-
#include "quota.h"
#include "quota-messages.h"
extern struct rpc_clnt_program quota_enforcer_clnt;
int32_t
-quota_validate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata, struct iatt *postparent);
+quota_validate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata, struct iatt *postparent);
int
-quota_enforcer_submit_request (void *req, call_frame_t *frame,
- rpc_clnt_prog_t *prog,
- int procnum, struct iobref *iobref,
- xlator_t *this, fop_cbk_fn_t cbkfn,
- xdrproc_t xdrproc)
+quota_enforcer_submit_request(void *req, call_frame_t *frame,
+ rpc_clnt_prog_t *prog, int procnum,
+ struct iobref *iobref, xlator_t *this,
+ fop_cbk_fn_t cbkfn, xdrproc_t xdrproc)
{
- int ret = -1;
- int count = 0;
- struct iovec iov = {0, };
- struct iobuf *iobuf = NULL;
- char new_iobref = 0;
- ssize_t xdr_size = 0;
- quota_priv_t *priv = NULL;
-
- GF_ASSERT (this);
-
- priv = this->private;
-
- if (req) {
- xdr_size = xdr_sizeof (xdrproc, req);
- iobuf = iobuf_get2 (this->ctx->iobuf_pool, xdr_size);
- if (!iobuf) {
- goto out;
- }
-
- if (!iobref) {
- iobref = iobref_new ();
- if (!iobref) {
- goto out;
- }
-
- new_iobref = 1;
- }
-
- iobref_add (iobref, iobuf);
-
- iov.iov_base = iobuf->ptr;
- iov.iov_len = iobuf_size (iobuf);
-
- /* Create the xdr payload */
- ret = xdr_serialize_generic (iov, req, xdrproc);
- if (ret == -1) {
- goto out;
- }
- iov.iov_len = ret;
- count = 1;
+ int ret = -1;
+ int count = 0;
+ struct iovec iov = {
+ 0,
+ };
+ struct iobuf *iobuf = NULL;
+ char new_iobref = 0;
+ ssize_t xdr_size = 0;
+ quota_priv_t *priv = NULL;
+
+ GF_ASSERT(this);
+
+ priv = this->private;
+
+ if (req) {
+ xdr_size = xdr_sizeof(xdrproc, req);
+ iobuf = iobuf_get2(this->ctx->iobuf_pool, xdr_size);
+ if (!iobuf) {
+ goto out;
+ }
+
+ if (!iobref) {
+ iobref = iobref_new();
+ if (!iobref) {
+ goto out;
+ }
+
+ new_iobref = 1;
}
- /* Send the msg */
- ret = rpc_clnt_submit (priv->rpc_clnt, prog, procnum, cbkfn,
- &iov, count,
- NULL, 0, iobref, frame, NULL, 0, NULL, 0, NULL);
- ret = 0;
+ iobref_add(iobref, iobuf);
+
+ iov.iov_base = iobuf->ptr;
+ iov.iov_len = iobuf_size(iobuf);
+
+ /* Create the xdr payload */
+ ret = xdr_serialize_generic(iov, req, xdrproc);
+ if (ret == -1) {
+ goto out;
+ }
+ iov.iov_len = ret;
+ count = 1;
+ }
+
+ /* Send the msg */
+ ret = rpc_clnt_submit(priv->rpc_clnt, prog, procnum, cbkfn, &iov, count,
+ NULL, 0, iobref, frame, NULL, 0, NULL, 0, NULL);
+ ret = 0;
out:
- if (new_iobref)
- iobref_unref (iobref);
- if (iobuf)
- iobuf_unref (iobuf);
+ if (new_iobref)
+ iobref_unref(iobref);
+ if (iobuf)
+ iobuf_unref(iobuf);
- return ret;
+ return ret;
}
int
-quota_enforcer_lookup_cbk (struct rpc_req *req, struct iovec *iov,
- int count, void *myframe)
+quota_enforcer_lookup_cbk(struct rpc_req *req, struct iovec *iov, int count,
+ void *myframe)
{
- quota_local_t *local = NULL;
- call_frame_t *frame = NULL;
- int ret = 0;
- gfs3_lookup_rsp rsp = {0,};
- struct iatt stbuf = {0,};
- struct iatt postparent = {0,};
- int op_errno = EINVAL;
- dict_t *xdata = NULL;
- inode_t *inode = NULL;
- xlator_t *this = NULL;
- quota_priv_t *priv = NULL;
- struct timespec retry_delay = {0,};
- gf_timer_t *timer = NULL;
-
- this = THIS;
-
- frame = myframe;
- local = frame->local;
- inode = local->validate_loc.inode;
- priv = this->private;
-
- if (-1 == req->rpc_status) {
- rsp.op_ret = -1;
- op_errno = ENOTCONN;
- goto out;
- }
+ quota_local_t *local = NULL;
+ call_frame_t *frame = NULL;
+ int ret = 0;
+ gfs3_lookup_rsp rsp = {
+ 0,
+ };
+ struct iatt stbuf = {
+ 0,
+ };
+ struct iatt postparent = {
+ 0,
+ };
+ int op_errno = EINVAL;
+ dict_t *xdata = NULL;
+ inode_t *inode = NULL;
+ xlator_t *this = NULL;
+ quota_priv_t *priv = NULL;
+ struct timespec retry_delay = {
+ 0,
+ };
+ gf_timer_t *timer = NULL;
+
+ this = THIS;
+
+ frame = myframe;
+ local = frame->local;
+ inode = local->validate_loc.inode;
+ priv = this->private;
+
+ if (-1 == req->rpc_status) {
+ rsp.op_ret = -1;
+ op_errno = ENOTCONN;
+ goto out;
+ }
+
+ ret = xdr_to_generic(*iov, &rsp, (xdrproc_t)xdr_gfs3_lookup_rsp);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, Q_MSG_XDR_DECODING_FAILED,
+ "XDR decoding failed");
+ rsp.op_ret = -1;
+ op_errno = EINVAL;
+ goto out;
+ }
- ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gfs3_lookup_rsp);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- Q_MSG_XDR_DECODING_FAILED,
- "XDR decoding failed");
- rsp.op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
+ op_errno = gf_error_to_errno(rsp.op_errno);
+ gf_stat_to_iatt(&rsp.postparent, &postparent);
- op_errno = gf_error_to_errno (rsp.op_errno);
- gf_stat_to_iatt (&rsp.postparent, &postparent);
+ if (rsp.op_ret == -1)
+ goto out;
- if (rsp.op_ret == -1)
- goto out;
+ rsp.op_ret = -1;
+ gf_stat_to_iatt(&rsp.stat, &stbuf);
+
+ GF_PROTOCOL_DICT_UNSERIALIZE(frame->this, xdata, (rsp.xdata.xdata_val),
+ (rsp.xdata.xdata_len), rsp.op_ret, op_errno,
+ out);
+ if ((!gf_uuid_is_null(inode->gfid)) &&
+ (gf_uuid_compare(stbuf.ia_gfid, inode->gfid) != 0)) {
+ gf_msg_debug(frame->this->name, ESTALE, "gfid changed for %s",
+ local->validate_loc.path);
rsp.op_ret = -1;
- gf_stat_to_iatt (&rsp.stat, &stbuf);
-
- GF_PROTOCOL_DICT_UNSERIALIZE (frame->this, xdata, (rsp.xdata.xdata_val),
- (rsp.xdata.xdata_len), rsp.op_ret,
- op_errno, out);
-
- if ((!gf_uuid_is_null (inode->gfid))
- && (gf_uuid_compare (stbuf.ia_gfid, inode->gfid) != 0)) {
- gf_msg_debug (frame->this->name, ESTALE,
- "gfid changed for %s", local->validate_loc.path);
- rsp.op_ret = -1;
- op_errno = ESTALE;
- goto out;
- }
+ op_errno = ESTALE;
+ goto out;
+ }
- rsp.op_ret = 0;
+ rsp.op_ret = 0;
out:
- rsp.op_errno = op_errno;
-
- /* We need to retry connecting to quotad on ENOTCONN error.
- * Suppose if there are two volumes vol1 and vol2,
- * and quota is enabled and limit is set on vol1.
- * Now if IO is happening on vol1 and quota is enabled/disabled
- * on vol2, quotad gets restarted and client will receive
- * ENOTCONN in the IO path of vol1
- */
- if (rsp.op_ret == -1 && rsp.op_errno == ENOTCONN) {
- if (local->quotad_conn_retry >= 12) {
- priv->quotad_conn_status = 1;
- gf_log (this->name, GF_LOG_WARNING, "failed to connect "
- "to quotad after retry count %d)",
- local->quotad_conn_retry);
- } else {
- local->quotad_conn_retry++;
- }
-
- if (priv->quotad_conn_status == 0) {
- /* retry connecting after 5secs for 12 retries
- * (up to 60sec).
- */
- gf_log (this->name, GF_LOG_DEBUG, "retry connecting to "
- "quotad (retry count %d)",
- local->quotad_conn_retry);
-
- retry_delay.tv_sec = 5;
- retry_delay.tv_nsec = 0;
- timer = gf_timer_call_after (this->ctx, retry_delay,
- _quota_enforcer_lookup,
- (void *) frame);
- if (timer == NULL) {
- gf_log (this->name, GF_LOG_WARNING, "failed to "
- "set quota_enforcer_lookup with timer");
- } else {
- goto clean;
- }
- }
+ rsp.op_errno = op_errno;
+
+ /* We need to retry connecting to quotad on ENOTCONN error.
+ * Suppose if there are two volumes vol1 and vol2,
+ * and quota is enabled and limit is set on vol1.
+ * Now if IO is happening on vol1 and quota is enabled/disabled
+ * on vol2, quotad gets restarted and client will receive
+ * ENOTCONN in the IO path of vol1
+ */
+ if (rsp.op_ret == -1 && rsp.op_errno == ENOTCONN) {
+ if (local->quotad_conn_retry >= 12) {
+ priv->quotad_conn_status = 1;
+ gf_log(this->name, GF_LOG_WARNING,
+ "failed to connect "
+ "to quotad after retry count %d)",
+ local->quotad_conn_retry);
} else {
- priv->quotad_conn_status = 0;
+ local->quotad_conn_retry++;
}
- if (rsp.op_ret == -1) {
- /* any error other than ENOENT */
- if (rsp.op_errno != ENOENT)
- gf_msg (this->name, GF_LOG_WARNING, rsp.op_errno,
- Q_MSG_LOOKUP_FAILED,
- "Getting cluster-wide size of directory failed "
- "(path: %s gfid:%s)", local->validate_loc.path,
- loc_gfid_utoa (&local->validate_loc));
- else
- gf_msg_trace (this->name, ENOENT,
- "not found on remote node");
-
- } else if (local->quotad_conn_retry) {
- gf_log (this->name, GF_LOG_DEBUG, "connected to quotad after "
- "retry count %d", local->quotad_conn_retry);
+ if (priv->quotad_conn_status == 0) {
+ /* retry connecting after 5secs for 12 retries
+ * (up to 60sec).
+ */
+ gf_log(this->name, GF_LOG_DEBUG,
+ "retry connecting to "
+ "quotad (retry count %d)",
+ local->quotad_conn_retry);
+
+ retry_delay.tv_sec = 5;
+ retry_delay.tv_nsec = 0;
+ timer = gf_timer_call_after(this->ctx, retry_delay,
+ _quota_enforcer_lookup, (void *)frame);
+ if (timer == NULL) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "failed to "
+ "set quota_enforcer_lookup with timer");
+ } else {
+ goto clean;
+ }
}
+ } else {
+ priv->quotad_conn_status = 0;
+ }
+
+ if (rsp.op_ret == -1) {
+ /* any error other than ENOENT */
+ if (rsp.op_errno != ENOENT)
+ gf_msg(
+ this->name, GF_LOG_WARNING, rsp.op_errno, Q_MSG_LOOKUP_FAILED,
+ "Getting cluster-wide size of directory failed "
+ "(path: %s gfid:%s)",
+ local->validate_loc.path, loc_gfid_utoa(&local->validate_loc));
+ else
+ gf_msg_trace(this->name, ENOENT, "not found on remote node");
- local->validate_cbk (frame, NULL, this, rsp.op_ret, rsp.op_errno, inode,
- &stbuf, xdata, &postparent);
+ } else if (local->quotad_conn_retry) {
+ gf_log(this->name, GF_LOG_DEBUG,
+ "connected to quotad after "
+ "retry count %d",
+ local->quotad_conn_retry);
+ }
+
+ local->validate_cbk(frame, NULL, this, rsp.op_ret, rsp.op_errno, inode,
+ &stbuf, xdata, &postparent);
clean:
- if (xdata)
- dict_unref (xdata);
+ if (xdata)
+ dict_unref(xdata);
- free (rsp.xdata.xdata_val);
+ free(rsp.xdata.xdata_val);
- return 0;
+ return 0;
}
void
-_quota_enforcer_lookup (void *data)
+_quota_enforcer_lookup(void *data)
{
- quota_local_t *local = NULL;
- gfs3_lookup_req req = {{0,},};
- int ret = 0;
- int op_errno = ESTALE;
- quota_priv_t *priv = NULL;
- call_frame_t *frame = NULL;
- loc_t *loc = NULL;
- xlator_t *this = NULL;
- char *dir_path = NULL;
-
- frame = data;
- local = frame->local;
- this = local->this;
- loc = &local->validate_loc;
-
- priv = this->private;
-
- if (!(loc && loc->inode))
- goto unwind;
-
- if (!gf_uuid_is_null (loc->inode->gfid))
- memcpy (req.gfid, loc->inode->gfid, 16);
- else
- memcpy (req.gfid, loc->gfid, 16);
-
- if (local->validate_xdata) {
- GF_PROTOCOL_DICT_SERIALIZE (this, local->validate_xdata,
- (&req.xdata.xdata_val),
- req.xdata.xdata_len,
- op_errno, unwind);
- }
-
- if (loc->name)
- req.bname = (char *)loc->name;
- else
- req.bname = "";
-
- if (loc->path)
- dir_path = (char *)loc->path;
- else
- dir_path = "";
-
- ret = quota_enforcer_submit_request (&req, frame,
- priv->quota_enforcer,
- GF_AGGREGATOR_LOOKUP,
- NULL, this,
- quota_enforcer_lookup_cbk,
- (xdrproc_t)xdr_gfs3_lookup_req);
-
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- Q_MSG_RPC_SUBMIT_FAILED, "Couldn't send the request to "
- "fetch cluster wide size of directory (path:%s gfid:%s)"
- , dir_path, req.gfid);
- }
-
- GF_FREE (req.xdata.xdata_val);
-
- return;
+ quota_local_t *local = NULL;
+ gfs3_lookup_req req = {
+ {
+ 0,
+ },
+ };
+ int ret = 0;
+ int op_errno = ESTALE;
+ quota_priv_t *priv = NULL;
+ call_frame_t *frame = NULL;
+ loc_t *loc = NULL;
+ xlator_t *this = NULL;
+ char *dir_path = NULL;
+
+ frame = data;
+ local = frame->local;
+ this = local->this;
+ loc = &local->validate_loc;
+
+ priv = this->private;
+
+ if (!(loc && loc->inode))
+ goto unwind;
+
+ if (!gf_uuid_is_null(loc->inode->gfid))
+ memcpy(req.gfid, loc->inode->gfid, 16);
+ else
+ memcpy(req.gfid, loc->gfid, 16);
+
+ if (local->validate_xdata) {
+ GF_PROTOCOL_DICT_SERIALIZE(this, local->validate_xdata,
+ (&req.xdata.xdata_val), req.xdata.xdata_len,
+ op_errno, unwind);
+ }
+
+ if (loc->name)
+ req.bname = (char *)loc->name;
+ else
+ req.bname = "";
+
+ if (loc->path)
+ dir_path = (char *)loc->path;
+ else
+ dir_path = "";
+
+ ret = quota_enforcer_submit_request(
+ &req, frame, priv->quota_enforcer, GF_AGGREGATOR_LOOKUP, NULL, this,
+ quota_enforcer_lookup_cbk, (xdrproc_t)xdr_gfs3_lookup_req);
+
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, Q_MSG_RPC_SUBMIT_FAILED,
+ "Couldn't send the request to "
+ "fetch cluster wide size of directory (path:%s gfid:%s)",
+ dir_path, req.gfid);
+ }
+
+ GF_FREE(req.xdata.xdata_val);
+
+ return;
unwind:
- local->validate_cbk (frame, NULL, this, -1, op_errno, NULL, NULL, NULL,
- NULL);
+ local->validate_cbk(frame, NULL, this, -1, op_errno, NULL, NULL, NULL,
+ NULL);
- GF_FREE (req.xdata.xdata_val);
+ GF_FREE(req.xdata.xdata_val);
- return;
+ return;
}
int
-quota_enforcer_lookup (call_frame_t *frame, xlator_t *this, dict_t *xdata,
- fop_lookup_cbk_t validate_cbk)
+quota_enforcer_lookup(call_frame_t *frame, xlator_t *this, dict_t *xdata,
+ fop_lookup_cbk_t validate_cbk)
{
- quota_local_t *local = NULL;
+ quota_local_t *local = NULL;
- if (!frame || !this)
- goto unwind;
+ if (!frame || !this)
+ goto unwind;
- local = frame->local;
- local->this = this;
- local->validate_cbk = validate_cbk;
- local->validate_xdata = dict_ref (xdata);
+ local = frame->local;
+ local->this = this;
+ local->validate_cbk = validate_cbk;
+ local->validate_xdata = dict_ref(xdata);
- _quota_enforcer_lookup (frame);
+ _quota_enforcer_lookup(frame);
- return 0;
+ return 0;
unwind:
- validate_cbk (frame, NULL, this, -1, ESTALE, NULL, NULL, NULL, NULL);
+ validate_cbk(frame, NULL, this, -1, ESTALE, NULL, NULL, NULL, NULL);
- return 0;
+ return 0;
}
int
-quota_enforcer_notify (struct rpc_clnt *rpc, void *mydata,
- rpc_clnt_event_t event, void *data)
+quota_enforcer_notify(struct rpc_clnt *rpc, void *mydata,
+ rpc_clnt_event_t event, void *data)
{
- xlator_t *this = NULL;
- int ret = 0;
-
- this = mydata;
-
- switch (event) {
- case RPC_CLNT_CONNECT:
- {
- gf_msg_trace (this->name, 0, "got RPC_CLNT_CONNECT");
- break;
+ xlator_t *this = NULL;
+ int ret = 0;
+ quota_priv_t *priv = NULL;
+
+ this = mydata;
+ priv = this->private;
+ switch (event) {
+ case RPC_CLNT_CONNECT: {
+ pthread_mutex_lock(&priv->conn_mutex);
+ {
+ priv->conn_status = _gf_true;
+ }
+ pthread_mutex_unlock(&priv->conn_mutex);
+ gf_msg_trace(this->name, 0, "got RPC_CLNT_CONNECT");
+ break;
}
- case RPC_CLNT_DISCONNECT:
- {
- gf_msg_trace (this->name, 0, "got RPC_CLNT_DISCONNECT");
- break;
+ case RPC_CLNT_DISCONNECT: {
+ pthread_mutex_lock(&priv->conn_mutex);
+ {
+ priv->conn_status = _gf_false;
+ pthread_cond_signal(&priv->conn_cond);
+ }
+ pthread_mutex_unlock(&priv->conn_mutex);
+ gf_msg_trace(this->name, 0, "got RPC_CLNT_DISCONNECT");
+ break;
}
default:
- gf_msg_trace (this->name, 0,
- "got some other RPC event %d", event);
- ret = 0;
- break;
- }
+ gf_msg_trace(this->name, 0, "got some other RPC event %d", event);
+ ret = 0;
+ break;
+ }
- return ret;
+ return ret;
}
int
-quota_enforcer_blocking_connect (rpc_clnt_t *rpc)
+quota_enforcer_blocking_connect(rpc_clnt_t *rpc)
{
- dict_t *options = NULL;
- int ret = -1;
+ dict_t *options = NULL;
+ int ret = -1;
- options = dict_new ();
- if (options == NULL)
- goto out;
+ options = dict_new();
+ if (options == NULL)
+ goto out;
- ret = dict_set_str (options, "non-blocking-io", "no");
- if (ret)
- goto out;
+ ret = dict_set_sizen_str_sizen(options, "non-blocking-io", "no");
+ if (ret)
+ goto out;
- rpc->conn.trans->reconfigure (rpc->conn.trans, options);
+ rpc->conn.trans->reconfigure(rpc->conn.trans, options);
- rpc_clnt_start (rpc);
+ rpc_clnt_start(rpc);
- ret = dict_set_str (options, "non-blocking-io", "yes");
- if (ret)
- goto out;
+ ret = dict_set_sizen_str_sizen(options, "non-blocking-io", "yes");
+ if (ret)
+ goto out;
- rpc->conn.trans->reconfigure (rpc->conn.trans, options);
+ rpc->conn.trans->reconfigure(rpc->conn.trans, options);
- ret = 0;
+ ret = 0;
out:
- if (options)
- dict_unref (options);
+ if (options)
+ dict_unref(options);
- return ret;
+ return ret;
}
-//Returns a started rpc_clnt. Creates a new rpc_clnt if quota_priv doesn't have
-//one already
+// Returns a started rpc_clnt. Creates a new rpc_clnt if quota_priv doesn't have
+// one already
struct rpc_clnt *
-quota_enforcer_init (xlator_t *this, dict_t *options)
+quota_enforcer_init(xlator_t *this, dict_t *options)
{
- struct rpc_clnt *rpc = NULL;
- quota_priv_t *priv = NULL;
- int ret = -1;
+ struct rpc_clnt *rpc = NULL;
+ quota_priv_t *priv = NULL;
+ int ret = -1;
- priv = this->private;
+ priv = this->private;
- LOCK (&priv->lock);
- {
- if (priv->rpc_clnt) {
- ret = 0;
- rpc = priv->rpc_clnt;
- }
+ LOCK(&priv->lock);
+ {
+ if (priv->rpc_clnt) {
+ ret = 0;
+ rpc = priv->rpc_clnt;
}
- UNLOCK (&priv->lock);
+ }
+ UNLOCK(&priv->lock);
- if (rpc)
- goto out;
+ if (rpc)
+ goto out;
- priv->quota_enforcer = &quota_enforcer_clnt;
+ priv->quota_enforcer = &quota_enforcer_clnt;
- ret = dict_set_str (options, "transport.address-family", "unix");
- if (ret)
- goto out;
+ ret = dict_set_sizen_str_sizen(options, "transport.address-family", "unix");
+ if (ret)
+ goto out;
- ret = dict_set_str (options, "transport-type", "socket");
- if (ret)
- goto out;
+ ret = dict_set_sizen_str_sizen(options, "transport-type", "socket");
+ if (ret)
+ goto out;
- ret = dict_set_str (options, "transport.socket.connect-path",
- "/var/run/gluster/quotad.socket");
- if (ret)
- goto out;
+ ret = dict_set_sizen_str_sizen(options, "transport.socket.connect-path",
+ "/var/run/gluster/quotad.socket");
+ if (ret)
+ goto out;
- rpc = rpc_clnt_new (options, this, this->name, 16);
- if (!rpc) {
- ret = -1;
- goto out;
- }
+ rpc = rpc_clnt_new(options, this, this->name, 16);
+ if (!rpc) {
+ ret = -1;
+ goto out;
+ }
- ret = rpc_clnt_register_notify (rpc, quota_enforcer_notify, this);
- if (ret) {
- gf_msg ("quota", GF_LOG_ERROR, 0,
- Q_MSG_RPCCLNT_REGISTER_NOTIFY_FAILED,
- "failed to register notify");
- goto out;
- }
+ ret = rpc_clnt_register_notify(rpc, quota_enforcer_notify, this);
+ if (ret) {
+ gf_msg("quota", GF_LOG_ERROR, 0, Q_MSG_RPCCLNT_REGISTER_NOTIFY_FAILED,
+ "failed to register notify");
+ goto out;
+ }
- ret = quota_enforcer_blocking_connect (rpc);
- if (ret)
- goto out;
+ ret = quota_enforcer_blocking_connect(rpc);
+ if (ret)
+ goto out;
- ret = 0;
+ ret = 0;
out:
- if (ret) {
- if (rpc)
- rpc_clnt_unref (rpc);
- rpc = NULL;
- }
+ if (ret) {
+ if (rpc)
+ rpc_clnt_unref(rpc);
+ rpc = NULL;
+ }
- return rpc;
- }
+ return rpc;
+}
struct rpc_clnt_procedure quota_enforcer_actors[GF_AGGREGATOR_MAXVALUE] = {
- [GF_AGGREGATOR_NULL] = {"NULL", NULL},
- [GF_AGGREGATOR_LOOKUP] = {"LOOKUP", NULL},
+ [GF_AGGREGATOR_NULL] = {"NULL", NULL},
+ [GF_AGGREGATOR_LOOKUP] = {"LOOKUP", NULL},
};
struct rpc_clnt_program quota_enforcer_clnt = {
- .progname = "Quota enforcer",
- .prognum = GLUSTER_AGGREGATOR_PROGRAM,
- .progver = GLUSTER_AGGREGATOR_VERSION,
- .numproc = GF_AGGREGATOR_MAXVALUE,
- .proctable = quota_enforcer_actors,
+ .progname = "Quota enforcer",
+ .prognum = GLUSTER_AGGREGATOR_PROGRAM,
+ .progver = GLUSTER_AGGREGATOR_VERSION,
+ .numproc = GF_AGGREGATOR_MAXVALUE,
+ .proctable = quota_enforcer_actors,
};
diff --git a/xlators/features/quota/src/quota-mem-types.h b/xlators/features/quota/src/quota-mem-types.h
index 97d9165681f..782a7de96bb 100644
--- a/xlators/features/quota/src/quota-mem-types.h
+++ b/xlators/features/quota/src/quota-mem-types.h
@@ -10,21 +10,21 @@
#ifndef __QUOTA_MEM_TYPES_H__
#define __QUOTA_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_quota_mem_types_ {
- gf_quota_mt_quota_priv_t = gf_common_mt_end + 1,
- gf_quota_mt_quota_inode_ctx_t,
- gf_quota_mt_loc_t,
- gf_quota_mt_char,
- gf_quota_mt_int64_t,
- gf_quota_mt_int32_t,
- gf_quota_mt_limits_t,
- gf_quota_mt_quota_dentry_t,
- gf_quota_mt_quota_limits_level_t,
- gf_quota_mt_qd_vols_conf_t,
- gf_quota_mt_aggregator_state_t,
- gf_quota_mt_end
+ /* Those are used by QUOTA_ALLOC_OR_GOTO macro */
+ gf_quota_mt_quota_priv_t = gf_common_mt_end + 1,
+ gf_quota_mt_quota_inode_ctx_t,
+ gf_quota_mt_loc_t,
+ gf_quota_mt_char,
+ gf_quota_mt_int64_t,
+ gf_quota_mt_int32_t,
+ gf_quota_mt_limits_t,
+ gf_quota_mt_quota_dentry_t,
+ gf_quota_mt_quota_limits_level_t,
+ gf_quota_mt_qd_vols_conf_t,
+ gf_quota_mt_aggregator_state_t,
+ gf_quota_mt_end
};
#endif
-
diff --git a/xlators/features/quota/src/quota-messages.h b/xlators/features/quota/src/quota-messages.h
index 85f5abf7b29..d434ed75e76 100644
--- a/xlators/features/quota/src/quota-messages.h
+++ b/xlators/features/quota/src/quota-messages.h
@@ -11,7 +11,7 @@
#ifndef _QUOTA_MESSAGES_H_
#define _QUOTA_MESSAGES_H_
-#include "glfs-message-id.h"
+#include <glusterfs/glfs-message-id.h>
/* To add new message IDs, append new identifiers at the end of the list.
*
@@ -23,33 +23,17 @@
* glfs-message-id.h.
*/
-GLFS_MSGID(QUOTA,
- Q_MSG_ENFORCEMENT_FAILED,
- Q_MSG_ENOMEM,
- Q_MSG_PARENT_NULL,
- Q_MSG_CROSSED_SOFT_LIMIT,
- Q_MSG_QUOTA_ENFORCER_RPC_INIT_FAILED,
- Q_MSG_REMOTE_OPERATION_FAILED,
- Q_MSG_FAILED_TO_SEND_FOP,
- Q_MSG_INVALID_VOLFILE,
- Q_MSG_INODE_PARENT_NOT_FOUND,
- Q_MSG_XDR_DECODE_ERROR,
- Q_MSG_DICT_UNSERIALIZE_FAIL,
- Q_MSG_DICT_SERIALIZE_FAIL,
- Q_MSG_RPCSVC_INIT_FAILED,
- Q_MSG_RPCSVC_LISTENER_CREATION_FAILED,
- Q_MSG_RPCSVC_REGISTER_FAILED,
- Q_MSG_XDR_DECODING_FAILED,
- Q_MSG_RPCCLNT_REGISTER_NOTIFY_FAILED,
- Q_MSG_ANCESTRY_BUILD_FAILED,
- Q_MSG_SIZE_KEY_MISSING,
- Q_MSG_INODE_CTX_GET_FAILED,
- Q_MSG_INODE_CTX_SET_FAILED,
- Q_MSG_LOOKUP_FAILED,
- Q_MSG_RPC_SUBMIT_FAILED,
- Q_MSG_ENFORCEMENT_SKIPPED,
- Q_MSG_INTERNAL_FOP_KEY_MISSING
-);
+GLFS_MSGID(QUOTA, Q_MSG_ENFORCEMENT_FAILED, Q_MSG_ENOMEM, Q_MSG_PARENT_NULL,
+ Q_MSG_CROSSED_SOFT_LIMIT, Q_MSG_QUOTA_ENFORCER_RPC_INIT_FAILED,
+ Q_MSG_REMOTE_OPERATION_FAILED, Q_MSG_FAILED_TO_SEND_FOP,
+ Q_MSG_INVALID_VOLFILE, Q_MSG_INODE_PARENT_NOT_FOUND,
+ Q_MSG_XDR_DECODE_ERROR, Q_MSG_DICT_UNSERIALIZE_FAIL,
+ Q_MSG_DICT_SERIALIZE_FAIL, Q_MSG_RPCSVC_INIT_FAILED,
+ Q_MSG_RPCSVC_LISTENER_CREATION_FAILED, Q_MSG_RPCSVC_REGISTER_FAILED,
+ Q_MSG_XDR_DECODING_FAILED, Q_MSG_RPCCLNT_REGISTER_NOTIFY_FAILED,
+ Q_MSG_ANCESTRY_BUILD_FAILED, Q_MSG_SIZE_KEY_MISSING,
+ Q_MSG_INODE_CTX_GET_FAILED, Q_MSG_INODE_CTX_SET_FAILED,
+ Q_MSG_LOOKUP_FAILED, Q_MSG_RPC_SUBMIT_FAILED,
+ Q_MSG_ENFORCEMENT_SKIPPED, Q_MSG_INTERNAL_FOP_KEY_MISSING);
#endif /* !_QUOTA_MESSAGES_H_ */
-
diff --git a/xlators/features/quota/src/quota.c b/xlators/features/quota/src/quota.c
index 2b440bb5827..18df9ae6d19 100644
--- a/xlators/features/quota/src/quota.c
+++ b/xlators/features/quota/src/quota.c
@@ -7,4766 +7,4636 @@
later), or the GNU General Public License, version 2 (GPLv2), in all
cases as published by the Free Software Foundation.
*/
-#include <fnmatch.h>
#include "quota.h"
-#include "common-utils.h"
-#include "defaults.h"
-#include "statedump.h"
-#include "quota-common-utils.h"
+#include <glusterfs/statedump.h>
#include "quota-messages.h"
-#include "events.h"
+#include <glusterfs/events.h>
struct volume_options options[];
static int32_t
-__quota_init_inode_ctx (inode_t *inode, xlator_t *this,
- quota_inode_ctx_t **context)
+__quota_init_inode_ctx(inode_t *inode, xlator_t *this,
+ quota_inode_ctx_t **context)
{
- int32_t ret = -1;
- quota_inode_ctx_t *ctx = NULL;
+ int32_t ret = -1;
+ quota_inode_ctx_t *ctx = NULL;
- if (inode == NULL) {
- goto out;
- }
+ if (inode == NULL) {
+ goto out;
+ }
- QUOTA_ALLOC_OR_GOTO (ctx, quota_inode_ctx_t, out);
+ QUOTA_ALLOC_OR_GOTO(ctx, quota_inode_ctx_t, out);
- LOCK_INIT(&ctx->lock);
+ LOCK_INIT(&ctx->lock);
- if (context != NULL) {
- *context = ctx;
- }
+ if (context != NULL) {
+ *context = ctx;
+ }
- INIT_LIST_HEAD (&ctx->parents);
+ INIT_LIST_HEAD(&ctx->parents);
- ret = __inode_ctx_put (inode, this, (uint64_t )(long)ctx);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- Q_MSG_INODE_CTX_SET_FAILED, "cannot set quota context "
- "in inode (gfid:%s)", uuid_utoa (inode->gfid));
- GF_FREE (ctx);
- }
+ ret = __inode_ctx_put(inode, this, (uint64_t)(long)ctx);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, Q_MSG_INODE_CTX_SET_FAILED,
+ "cannot set quota context "
+ "in inode (gfid:%s)",
+ uuid_utoa(inode->gfid));
+ GF_FREE(ctx);
+ }
out:
- return ret;
+ return ret;
}
-
static int32_t
-quota_inode_ctx_get (inode_t *inode, xlator_t *this,
- quota_inode_ctx_t **ctx, char create_if_absent)
+quota_inode_ctx_get(inode_t *inode, xlator_t *this, quota_inode_ctx_t **ctx,
+ char create_if_absent)
{
- int32_t ret = 0;
- uint64_t ctx_int;
+ int32_t ret = 0;
+ uint64_t ctx_int;
- LOCK (&inode->lock);
- {
- ret = __inode_ctx_get (inode, this, &ctx_int);
+ LOCK(&inode->lock);
+ {
+ ret = __inode_ctx_get(inode, this, &ctx_int);
- if ((ret == 0) && (ctx != NULL)) {
- *ctx = (quota_inode_ctx_t *) (unsigned long)ctx_int;
- } else if (create_if_absent) {
- ret = __quota_init_inode_ctx (inode, this, ctx);
- }
+ if ((ret == 0) && (ctx != NULL)) {
+ *ctx = (quota_inode_ctx_t *)(unsigned long)ctx_int;
+ } else if (create_if_absent) {
+ ret = __quota_init_inode_ctx(inode, this, ctx);
}
- UNLOCK (&inode->lock);
+ }
+ UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
int
-quota_loc_fill (loc_t *loc, inode_t *inode, inode_t *parent, char *path)
+quota_loc_fill(loc_t *loc, inode_t *inode, inode_t *parent, char *path)
{
- int ret = -1;
+ int ret = -1;
- if (!loc || (inode == NULL))
- return ret;
+ if (!loc || (inode == NULL))
+ return ret;
- if (inode) {
- loc->inode = inode_ref (inode);
- gf_uuid_copy (loc->gfid, inode->gfid);
- }
+ if (inode) {
+ loc->inode = inode_ref(inode);
+ gf_uuid_copy(loc->gfid, inode->gfid);
+ }
- if (parent) {
- loc->parent = inode_ref (parent);
- }
+ if (parent) {
+ loc->parent = inode_ref(parent);
+ }
- if (path != NULL) {
- loc->path = gf_strdup (path);
+ if (path != NULL) {
+ loc->path = gf_strdup(path);
- loc->name = strrchr (loc->path, '/');
- if (loc->name) {
- loc->name++;
- }
+ loc->name = strrchr(loc->path, '/');
+ if (loc->name) {
+ loc->name++;
}
+ }
- ret = 0;
+ ret = 0;
- return ret;
+ return ret;
}
-
int
-quota_inode_loc_fill (inode_t *inode, loc_t *loc)
+quota_inode_loc_fill(inode_t *inode, loc_t *loc)
{
- char *resolvedpath = NULL;
- inode_t *parent = NULL;
- int ret = -1;
- xlator_t *this = NULL;
+ char *resolvedpath = NULL;
+ inode_t *parent = NULL;
+ int ret = -1;
+ xlator_t *this = NULL;
- if ((!inode) || (!loc)) {
- return ret;
- }
+ if ((!inode) || (!loc)) {
+ return ret;
+ }
- this = THIS;
+ this = THIS;
- if ((inode) && __is_root_gfid (inode->gfid)) {
- loc->parent = NULL;
- goto ignore_parent;
- }
+ if ((inode) && __is_root_gfid(inode->gfid)) {
+ loc->parent = NULL;
+ goto ignore_parent;
+ }
- parent = inode_parent (inode, 0, NULL);
- if (!parent) {
- gf_msg_debug (this->name, 0, "cannot find parent for "
- "inode (gfid:%s)", uuid_utoa (inode->gfid));
- }
+ parent = inode_parent(inode, 0, NULL);
+ if (!parent) {
+ gf_msg_debug(this->name, 0,
+ "cannot find parent for "
+ "inode (gfid:%s)",
+ uuid_utoa(inode->gfid));
+ }
ignore_parent:
- ret = inode_path (inode, NULL, &resolvedpath);
- if (ret < 0) {
- gf_msg_debug (this->name, 0, "cannot construct path for "
- "inode (gfid:%s)", uuid_utoa (inode->gfid));
- }
-
- ret = quota_loc_fill (loc, inode, parent, resolvedpath);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
- "cannot fill loc");
- goto err;
- }
+ ret = inode_path(inode, NULL, &resolvedpath);
+ if (ret < 0) {
+ gf_msg_debug(this->name, 0,
+ "cannot construct path for "
+ "inode (gfid:%s)",
+ uuid_utoa(inode->gfid));
+ }
+
+ ret = quota_loc_fill(loc, inode, parent, resolvedpath);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "cannot fill loc");
+ goto err;
+ }
err:
- if (parent) {
- inode_unref (parent);
- }
+ if (parent) {
+ inode_unref(parent);
+ }
- GF_FREE (resolvedpath);
+ GF_FREE(resolvedpath);
- return ret;
+ return ret;
}
-
int32_t
-quota_local_cleanup (quota_local_t *local)
+quota_local_cleanup(quota_local_t *local)
{
- if (local == NULL) {
- goto out;
- }
+ if (local == NULL) {
+ goto out;
+ }
- loc_wipe (&local->loc);
- loc_wipe (&local->newloc);
- loc_wipe (&local->oldloc);
- loc_wipe (&local->validate_loc);
+ loc_wipe(&local->loc);
+ loc_wipe(&local->newloc);
+ loc_wipe(&local->oldloc);
+ loc_wipe(&local->validate_loc);
- inode_unref (local->inode);
+ inode_unref(local->inode);
- if (local->xdata)
- dict_unref (local->xdata);
+ if (local->xdata)
+ dict_unref(local->xdata);
- if (local->validate_xdata)
- dict_unref (local->validate_xdata);
+ if (local->validate_xdata)
+ dict_unref(local->validate_xdata);
- if (local->stub)
- call_stub_destroy (local->stub);
+ if (local->stub)
+ call_stub_destroy(local->stub);
- LOCK_DESTROY (&local->lock);
+ LOCK_DESTROY(&local->lock);
- mem_put (local);
+ mem_put(local);
out:
- return 0;
+ return 0;
}
-
static quota_local_t *
-quota_local_new ()
+quota_local_new()
{
- quota_local_t *local = NULL;
- local = mem_get0 (THIS->local_pool);
- if (local == NULL)
- goto out;
+ quota_local_t *local = NULL;
+ local = mem_get0(THIS->local_pool);
+ if (local == NULL)
+ goto out;
- LOCK_INIT (&local->lock);
- local->space_available = -1;
+ LOCK_INIT(&local->lock);
+ local->space_available = -1;
out:
- return local;
+ return local;
}
-
quota_dentry_t *
-__quota_dentry_new (quota_inode_ctx_t *ctx, char *name, uuid_t par)
+__quota_dentry_new(quota_inode_ctx_t *ctx, char *name, uuid_t par)
{
- quota_dentry_t *dentry = NULL;
- GF_UNUSED int32_t ret = 0;
+ quota_dentry_t *dentry = NULL;
+ GF_UNUSED int32_t ret = 0;
- QUOTA_ALLOC_OR_GOTO (dentry, quota_dentry_t, err);
+ QUOTA_ALLOC_OR_GOTO(dentry, quota_dentry_t, err);
- INIT_LIST_HEAD (&dentry->next);
+ INIT_LIST_HEAD(&dentry->next);
- dentry->name = gf_strdup (name);
- if (dentry->name == NULL) {
- GF_FREE (dentry);
- dentry = NULL;
- goto err;
- }
+ dentry->name = gf_strdup(name);
+ if (dentry->name == NULL) {
+ GF_FREE(dentry);
+ dentry = NULL;
+ goto err;
+ }
- gf_uuid_copy (dentry->par, par);
+ gf_uuid_copy(dentry->par, par);
- if (ctx != NULL)
- list_add_tail (&dentry->next, &ctx->parents);
+ if (ctx != NULL)
+ list_add_tail(&dentry->next, &ctx->parents);
err:
- return dentry;
+ return dentry;
}
-
void
-__quota_dentry_free (quota_dentry_t *dentry)
+__quota_dentry_free(quota_dentry_t *dentry)
{
- if (dentry == NULL) {
- goto out;
- }
+ if (dentry == NULL) {
+ goto out;
+ }
- list_del_init (&dentry->next);
+ list_del_init(&dentry->next);
- GF_FREE (dentry->name);
- GF_FREE (dentry);
+ GF_FREE(dentry->name);
+ GF_FREE(dentry);
out:
- return;
+ return;
}
void
-__quota_dentry_del (quota_inode_ctx_t *ctx, const char *name, uuid_t par)
+__quota_dentry_del(quota_inode_ctx_t *ctx, const char *name, uuid_t par)
{
- quota_dentry_t *dentry = NULL;
- quota_dentry_t *tmp = NULL;
-
- list_for_each_entry_safe (dentry, tmp, &ctx->parents, next) {
- if ((strcmp (dentry->name, name) == 0) &&
- (gf_uuid_compare (dentry->par, par) == 0)) {
- __quota_dentry_free (dentry);
- break;
- }
+ quota_dentry_t *dentry = NULL;
+ quota_dentry_t *tmp = NULL;
+
+ list_for_each_entry_safe(dentry, tmp, &ctx->parents, next)
+ {
+ if ((strcmp(dentry->name, name) == 0) &&
+ (gf_uuid_compare(dentry->par, par) == 0)) {
+ __quota_dentry_free(dentry);
+ break;
}
+ }
}
void
-quota_dentry_del (quota_inode_ctx_t *ctx, const char *name, uuid_t par)
+quota_dentry_del(quota_inode_ctx_t *ctx, const char *name, uuid_t par)
{
- LOCK (&ctx->lock);
- {
- __quota_dentry_del (ctx, name, par);
- }
- UNLOCK (&ctx->lock);
+ LOCK(&ctx->lock);
+ {
+ __quota_dentry_del(ctx, name, par);
+ }
+ UNLOCK(&ctx->lock);
}
-static inode_t*
-__quota_inode_parent (inode_t *inode, uuid_t pargfid, const char *name)
+static inode_t *
+__quota_inode_parent(inode_t *inode, uuid_t pargfid, const char *name)
{
- inode_t *parent = NULL;
+ inode_t *parent = NULL;
- parent = inode_parent (inode, pargfid, name);
- inode_unref (inode);
- return parent;
+ parent = inode_parent(inode, pargfid, name);
+ inode_unref(inode);
+ return parent;
}
-static inode_t*
-quota_inode_parent (inode_t *inode, uuid_t pargfid, const char *name)
+static inode_t *
+quota_inode_parent(inode_t *inode, uuid_t pargfid, const char *name)
{
- inode_t *parent = NULL;
+ inode_t *parent = NULL;
- parent = __quota_inode_parent (inode, pargfid, name);
- if (!parent)
- gf_msg_callingfn (THIS->name, GF_LOG_ERROR, 0,
- Q_MSG_PARENT_NULL,
- "Failed to find "
- "ancestor for inode (%s)",
- uuid_utoa(inode->gfid));
+ parent = __quota_inode_parent(inode, pargfid, name);
+ if (!parent)
+ gf_msg_callingfn(THIS->name, GF_LOG_ERROR, 0, Q_MSG_PARENT_NULL,
+ "Failed to find "
+ "ancestor for inode (%s)",
+ uuid_utoa(inode->gfid));
- return parent;
+ return parent;
}
int32_t
-quota_inode_depth (inode_t *inode)
+quota_inode_depth(inode_t *inode)
{
- int depth = 0;
- inode_t *cur_inode = NULL;
-
- cur_inode = inode_ref (inode);
- while (cur_inode && !__is_root_gfid (cur_inode->gfid)) {
- depth++;
- cur_inode = quota_inode_parent (cur_inode, 0 , NULL);
- if (!cur_inode)
- depth = -1;
- }
+ int depth = 0;
+ inode_t *cur_inode = NULL;
+
+ cur_inode = inode_ref(inode);
+ while (cur_inode && !__is_root_gfid(cur_inode->gfid)) {
+ depth++;
+ cur_inode = quota_inode_parent(cur_inode, 0, NULL);
+ if (!cur_inode)
+ depth = -1;
+ }
- if (cur_inode)
- inode_unref (cur_inode);
+ if (cur_inode)
+ inode_unref(cur_inode);
- return depth;
+ return depth;
}
-int32_t quota_find_common_ancestor (inode_t *inode1, inode_t *inode2,
- uuid_t *common_ancestor)
+int32_t
+quota_find_common_ancestor(inode_t *inode1, inode_t *inode2,
+ uuid_t *common_ancestor)
{
- int32_t depth1 = 0;
- int32_t depth2 = 0;
- int32_t ret = -1;
- inode_t *cur_inode1 = NULL;
- inode_t *cur_inode2 = NULL;
-
- depth1 = quota_inode_depth (inode1);
- if (depth1 < 0)
- goto out;
-
- depth2 = quota_inode_depth (inode2);
- if (depth2 < 0)
- goto out;
-
- cur_inode1 = inode_ref (inode1);
- cur_inode2 = inode_ref (inode2);
-
- while (cur_inode1 && depth1 > depth2) {
- cur_inode1 = quota_inode_parent (cur_inode1, 0 , NULL);
- depth1--;
- }
-
- while (cur_inode2 && depth2 > depth1) {
- cur_inode2 = quota_inode_parent (cur_inode2, 0 , NULL);
- depth2--;
- }
-
- while (depth1 && cur_inode1 && cur_inode2 && cur_inode1 != cur_inode2) {
- cur_inode1 = quota_inode_parent (cur_inode1, 0 , NULL);
- cur_inode2 = quota_inode_parent (cur_inode2, 0 , NULL);
- depth1--;
- }
-
- if (cur_inode1 && cur_inode2) {
- gf_uuid_copy (*common_ancestor, cur_inode1->gfid);
- ret = 0;
- }
+ int32_t depth1 = 0;
+ int32_t depth2 = 0;
+ int32_t ret = -1;
+ inode_t *cur_inode1 = NULL;
+ inode_t *cur_inode2 = NULL;
+
+ depth1 = quota_inode_depth(inode1);
+ if (depth1 < 0)
+ goto out;
+
+ depth2 = quota_inode_depth(inode2);
+ if (depth2 < 0)
+ goto out;
+
+ cur_inode1 = inode_ref(inode1);
+ cur_inode2 = inode_ref(inode2);
+
+ while (cur_inode1 && depth1 > depth2) {
+ cur_inode1 = quota_inode_parent(cur_inode1, 0, NULL);
+ depth1--;
+ }
+
+ while (cur_inode2 && depth2 > depth1) {
+ cur_inode2 = quota_inode_parent(cur_inode2, 0, NULL);
+ depth2--;
+ }
+
+ while (depth1 && cur_inode1 && cur_inode2 && cur_inode1 != cur_inode2) {
+ cur_inode1 = quota_inode_parent(cur_inode1, 0, NULL);
+ cur_inode2 = quota_inode_parent(cur_inode2, 0, NULL);
+ depth1--;
+ }
+
+ if (cur_inode1 && cur_inode2) {
+ gf_uuid_copy(*common_ancestor, cur_inode1->gfid);
+ ret = 0;
+ }
out:
- if (cur_inode1)
- inode_unref (cur_inode1);
+ if (cur_inode1)
+ inode_unref(cur_inode1);
- if (cur_inode2)
- inode_unref (cur_inode2);
+ if (cur_inode2)
+ inode_unref(cur_inode2);
- return ret;
- }
+ return ret;
+}
void
-check_ancestory_continue (struct list_head *parents, inode_t *inode,
- int32_t op_ret, int32_t op_errno, void *data)
+check_ancestory_continue(struct list_head *parents, inode_t *inode,
+ int32_t op_ret, int32_t op_errno, void *data)
{
- call_frame_t *frame = NULL;
- quota_local_t *local = NULL;
- uint32_t link_count = 0;
-
- frame = data;
- local = frame->local;
-
- if (parents && list_empty (parents)) {
- gf_msg (THIS->name, GF_LOG_WARNING, EIO,
- Q_MSG_ANCESTRY_BUILD_FAILED,
- "Couldn't build ancestry for inode (gfid:%s). "
- "Without knowing ancestors till root, quota "
- "cannot be enforced. "
- "Hence, failing fop with EIO",
- uuid_utoa (inode->gfid));
- op_errno = EIO;
- op_ret = -1;
- }
-
- LOCK (&local->lock);
- {
- link_count = --local->link_count;
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- }
+ call_frame_t *frame = NULL;
+ quota_local_t *local = NULL;
+ uint32_t link_count = 0;
+
+ frame = data;
+ local = frame->local;
+
+ if (parents && list_empty(parents)) {
+ gf_msg(THIS->name, GF_LOG_WARNING, EIO, Q_MSG_ANCESTRY_BUILD_FAILED,
+ "Couldn't build ancestry for inode (gfid:%s). "
+ "Without knowing ancestors till root, quota "
+ "cannot be enforced. "
+ "Hence, failing fop with EIO",
+ uuid_utoa(inode->gfid));
+ op_errno = EIO;
+ op_ret = -1;
+ }
+
+ LOCK(&local->lock);
+ {
+ link_count = --local->link_count;
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
}
- UNLOCK (&local->lock);
+ }
+ UNLOCK(&local->lock);
- if (link_count == 0)
- local->fop_continue_cbk (frame);
+ if (link_count == 0)
+ local->fop_continue_cbk(frame);
}
void
-check_ancestory (call_frame_t *frame, inode_t *inode)
+check_ancestory(call_frame_t *frame, inode_t *inode)
{
- inode_t *cur_inode = NULL;
- inode_t *parent = NULL;
-
- cur_inode = inode_ref (inode);
- while (cur_inode && !__is_root_gfid (cur_inode->gfid)) {
- parent = inode_parent (cur_inode, 0, NULL);
- if (!parent) {
- quota_build_ancestry (cur_inode,
- check_ancestory_continue, frame);
- inode_unref (cur_inode);
- return;
- }
- inode_unref (cur_inode);
- cur_inode = parent;
- }
+ inode_t *cur_inode = NULL;
+ inode_t *parent = NULL;
- if (cur_inode) {
- inode_unref (cur_inode);
- check_ancestory_continue (NULL, NULL, 0, 0, frame);
- } else {
- check_ancestory_continue (NULL, NULL, -1, ESTALE, frame);
- }
+ cur_inode = inode_ref(inode);
+ while (cur_inode && !__is_root_gfid(cur_inode->gfid)) {
+ parent = inode_parent(cur_inode, 0, NULL);
+ if (!parent) {
+ quota_build_ancestry(cur_inode, check_ancestory_continue, frame);
+ inode_unref(cur_inode);
+ return;
+ }
+ inode_unref(cur_inode);
+ cur_inode = parent;
+ }
+
+ if (cur_inode) {
+ inode_unref(cur_inode);
+ check_ancestory_continue(NULL, NULL, 0, 0, frame);
+ } else {
+ check_ancestory_continue(NULL, NULL, -1, ESTALE, frame);
+ }
}
void
-check_ancestory_2_cbk (struct list_head *parents, inode_t *inode,
- int32_t op_ret, int32_t op_errno, void *data)
+check_ancestory_2_cbk(struct list_head *parents, inode_t *inode, int32_t op_ret,
+ int32_t op_errno, void *data)
{
- inode_t *this_inode = NULL;
- quota_inode_ctx_t *ctx = NULL;
+ inode_t *this_inode = NULL;
+ quota_inode_ctx_t *ctx = NULL;
- this_inode = data;
+ this_inode = data;
- if (op_ret < 0)
- goto out;
+ if (op_ret < 0)
+ goto out;
- if (parents == NULL || list_empty (parents)) {
- gf_msg (THIS->name, GF_LOG_WARNING, 0,
- Q_MSG_ENFORCEMENT_FAILED,
- "Couldn't build ancestry for inode (gfid:%s). "
- "Without knowing ancestors till root, quota "
- "cannot be enforced.",
- uuid_utoa (this_inode->gfid));
- goto out;
- }
+ if (parents == NULL || list_empty(parents)) {
+ gf_msg(THIS->name, GF_LOG_WARNING, 0, Q_MSG_ENFORCEMENT_FAILED,
+ "Couldn't build ancestry for inode (gfid:%s). "
+ "Without knowing ancestors till root, quota "
+ "cannot be enforced.",
+ uuid_utoa(this_inode->gfid));
+ goto out;
+ }
- quota_inode_ctx_get (this_inode, THIS, &ctx, 0);
- if (ctx)
- ctx->ancestry_built = _gf_true;
+ quota_inode_ctx_get(this_inode, THIS, &ctx, 0);
+ if (ctx)
+ ctx->ancestry_built = _gf_true;
out:
- inode_unref (this_inode);
+ inode_unref(this_inode);
}
void
-check_ancestory_2 (xlator_t *this, quota_local_t *local, inode_t *inode)
+check_ancestory_2(xlator_t *this, quota_local_t *local, inode_t *inode)
{
- inode_t *cur_inode = NULL;
- inode_t *parent = NULL;
- quota_inode_ctx_t *ctx = NULL;
- char *name = NULL;
- uuid_t pgfid = {0};
-
- name = (char *) local->loc.name;
- if (local->loc.parent) {
- gf_uuid_copy (pgfid, local->loc.parent->gfid);
- parent = local->loc.parent;
- }
-
- cur_inode = inode_ref (inode);
- while (cur_inode && !__is_root_gfid (cur_inode->gfid)) {
- quota_inode_ctx_get (cur_inode, this, &ctx, 0);
- /* build ancestry is required only on the first lookup,
- * so stop crawling when the inode_ctx is set for an inode
- */
- if (ctx && ctx->ancestry_built)
- goto setctx;
-
- parent = inode_parent (cur_inode, pgfid, name);
- if (!parent) {
- quota_build_ancestry (cur_inode, check_ancestory_2_cbk,
- inode_ref (inode));
- goto out;
- }
+ inode_t *cur_inode = NULL;
+ inode_t *parent = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ char *name = NULL;
+ uuid_t pgfid = {0};
+
+ name = (char *)local->loc.name;
+ if (local->loc.parent) {
+ gf_uuid_copy(pgfid, local->loc.parent->gfid);
+ }
+
+ cur_inode = inode_ref(inode);
+ while (cur_inode && !__is_root_gfid(cur_inode->gfid)) {
+ quota_inode_ctx_get(cur_inode, this, &ctx, 0);
+ /* build ancestry is required only on the first lookup,
+ * so stop crawling when the inode_ctx is set for an inode
+ */
+ if (ctx && ctx->ancestry_built)
+ goto setctx;
- if (name != NULL) {
- name = NULL;
- gf_uuid_clear (pgfid);
- }
+ parent = inode_parent(cur_inode, pgfid, name);
+ if (!parent) {
+ quota_build_ancestry(cur_inode, check_ancestory_2_cbk,
+ inode_ref(inode));
+ goto out;
+ }
- inode_unref (cur_inode);
- cur_inode = parent;
+ if (name != NULL) {
+ name = NULL;
+ gf_uuid_clear(pgfid);
}
+ inode_unref(cur_inode);
+ cur_inode = parent;
+ }
+
setctx:
- if (cur_inode && cur_inode != inode) {
- quota_inode_ctx_get (inode, this, &ctx, 0);
- if (ctx)
- ctx->ancestry_built = _gf_true;
- }
+ if (cur_inode && cur_inode != inode) {
+ quota_inode_ctx_get(inode, this, &ctx, 0);
+ if (ctx)
+ ctx->ancestry_built = _gf_true;
+ }
out:
- if (cur_inode)
- inode_unref (cur_inode);
+ if (cur_inode)
+ inode_unref(cur_inode);
}
static void
-quota_link_count_decrement (call_frame_t *frame)
+quota_link_count_decrement(call_frame_t *frame)
{
- call_frame_t *tmpframe = NULL;
- quota_local_t *local = NULL;
- call_stub_t *stub = NULL;
- int link_count = -1;
-
- local = frame->local;
- if (local && local->par_frame) {
- local = local->par_frame->local;
- tmpframe = frame;
- }
-
- if (local == NULL)
- goto out;
-
- LOCK (&local->lock);
- {
- link_count = --local->link_count;
- if (link_count == 0) {
- stub = local->stub;
- local->stub = NULL;
- }
- }
- UNLOCK (&local->lock);
-
- if (stub != NULL) {
- call_resume (stub);
- }
+ call_frame_t *tmpframe = NULL;
+ quota_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+ int link_count = -1;
+
+ local = frame->local;
+ if (local && local->par_frame) {
+ local = local->par_frame->local;
+ tmpframe = frame;
+ }
+
+ if (local == NULL)
+ goto out;
+
+ LOCK(&local->lock);
+ {
+ link_count = --local->link_count;
+ if (link_count == 0) {
+ stub = local->stub;
+ local->stub = NULL;
+ }
+ }
+ UNLOCK(&local->lock);
+
+ if (stub != NULL) {
+ call_resume(stub);
+ }
out:
- if (tmpframe) {
- local = tmpframe->local;
- tmpframe->local = NULL;
+ if (tmpframe) {
+ local = tmpframe->local;
+ tmpframe->local = NULL;
- STACK_DESTROY (frame->root);
- if (local)
- quota_local_cleanup (local);
- }
+ STACK_DESTROY(frame->root);
+ if (local)
+ quota_local_cleanup(local);
+ }
- return;
+ return;
}
static void
-quota_handle_validate_error (call_frame_t *frame, int32_t op_ret,
- int32_t op_errno)
+quota_handle_validate_error(call_frame_t *frame, int32_t op_ret,
+ int32_t op_errno)
{
- quota_local_t *local;
+ quota_local_t *local;
- local = frame->local;
- if (local && local->par_frame)
- local = local->par_frame->local;
+ local = frame->local;
+ if (local && local->par_frame)
+ local = local->par_frame->local;
- if (local == NULL)
- goto out;
+ if (local == NULL)
+ goto out;
- LOCK (&local->lock);
+ if (op_ret < 0) {
+ LOCK(&local->lock);
{
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- }
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
}
- UNLOCK (&local->lock);
-
- /* we abort checking limits on this path to root */
- quota_link_count_decrement (frame);
+ UNLOCK(&local->lock);
+ }
+ /* we abort checking limits on this path to root */
+ quota_link_count_decrement(frame);
out:
- return;
+ return;
}
int32_t
-quota_validate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+quota_validate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata, struct iatt *postparent)
{
- quota_local_t *local = NULL;
- int32_t ret = 0;
- quota_inode_ctx_t *ctx = NULL;
- uint64_t value = 0;
- quota_meta_t size = {0,};
-
- local = frame->local;
-
- if (op_ret < 0) {
- goto unwind;
- }
-
- GF_ASSERT (local);
- GF_ASSERT (frame);
- GF_VALIDATE_OR_GOTO_WITH_ERROR ("quota", this, unwind, op_errno,
- EINVAL);
- GF_VALIDATE_OR_GOTO_WITH_ERROR (this->name, xdata, unwind, op_errno,
- EINVAL);
-
- ret = inode_ctx_get (local->validate_loc.inode, this, &value);
-
- ctx = (quota_inode_ctx_t *)(unsigned long)value;
- if ((ret == -1) || (ctx == NULL)) {
- gf_msg (this->name, GF_LOG_WARNING, EINVAL,
- Q_MSG_INODE_CTX_GET_FAILED, "quota context is"
- " not present in inode (gfid:%s)",
- uuid_utoa (local->validate_loc.inode->gfid));
- op_errno = EINVAL;
- goto unwind;
- }
-
- ret = quota_dict_get_meta (xdata, QUOTA_SIZE_KEY, &size);
- if (ret == -1) {
- gf_msg (this->name, GF_LOG_WARNING, EINVAL,
- Q_MSG_SIZE_KEY_MISSING, "quota size key not present "
- "in dict");
- op_errno = EINVAL;
- }
-
- local->just_validated = 1; /* so that we don't go into infinite
- * loop of validation and checking
- * limit when timeout is zero.
- */
- LOCK (&ctx->lock);
- {
- ctx->size = size.size;
- ctx->file_count = size.file_count;
- ctx->dir_count = size.dir_count;
- gettimeofday (&ctx->tv, NULL);
- }
- UNLOCK (&ctx->lock);
-
- quota_check_limit (frame, local->validate_loc.inode, this);
- return 0;
+ quota_local_t *local = NULL;
+ int32_t ret = 0;
+ quota_inode_ctx_t *ctx = NULL;
+ uint64_t value = 0;
+ quota_meta_t size = {
+ 0,
+ };
+
+ local = frame->local;
+
+ if (op_ret < 0) {
+ goto unwind;
+ }
+
+ GF_ASSERT(local);
+ GF_ASSERT(frame);
+ GF_VALIDATE_OR_GOTO_WITH_ERROR("quota", this, unwind, op_errno, EINVAL);
+ GF_VALIDATE_OR_GOTO_WITH_ERROR(this->name, xdata, unwind, op_errno, EINVAL);
+
+ ret = inode_ctx_get(local->validate_loc.inode, this, &value);
+
+ ctx = (quota_inode_ctx_t *)(unsigned long)value;
+ if ((ret == -1) || (ctx == NULL)) {
+ gf_msg(this->name, GF_LOG_WARNING, EINVAL, Q_MSG_INODE_CTX_GET_FAILED,
+ "quota context is"
+ " not present in inode (gfid:%s)",
+ uuid_utoa(local->validate_loc.inode->gfid));
+ op_errno = EINVAL;
+ goto unwind;
+ }
+
+ ret = quota_dict_get_meta(xdata, QUOTA_SIZE_KEY, SLEN(QUOTA_SIZE_KEY),
+ &size);
+ if (ret == -1) {
+ gf_msg(this->name, GF_LOG_WARNING, EINVAL, Q_MSG_SIZE_KEY_MISSING,
+ "quota size key not present "
+ "in dict");
+ op_errno = EINVAL;
+ }
+
+ local->just_validated = 1; /* so that we don't go into infinite
+ * loop of validation and checking
+ * limit when timeout is zero.
+ */
+ LOCK(&ctx->lock);
+ {
+ ctx->size = size.size;
+ ctx->validate_time = gf_time();
+ ctx->file_count = size.file_count;
+ ctx->dir_count = size.dir_count;
+ }
+ UNLOCK(&ctx->lock);
+
+ quota_check_limit(frame, local->validate_loc.inode, this);
+ return 0;
unwind:
- quota_handle_validate_error (frame, op_ret, op_errno);
- return 0;
+ quota_handle_validate_error(frame, op_ret, op_errno);
+ return 0;
}
-
-static uint64_t
-quota_time_elapsed (struct timeval *now, struct timeval *then)
+static inline gf_boolean_t
+quota_timeout(time_t t, uint32_t timeout)
{
- return (now->tv_sec - then->tv_sec);
-}
-
-
-int32_t
-quota_timeout (struct timeval *tv, int32_t timeout)
-{
- struct timeval now = {0,};
- int32_t timed_out = 0;
-
- gettimeofday (&now, NULL);
-
- if (quota_time_elapsed (&now, tv) >= timeout) {
- timed_out = 1;
- }
-
- return timed_out;
+ return (gf_time() - t) >= timeout;
}
/* Return: 1 if new entry added
* 0 no entry added
+ * -1 on errors
*/
static int32_t
-quota_add_parent (struct list_head *list, char *name, uuid_t pgfid)
+quota_add_parent(struct list_head *list, char *name, uuid_t pgfid)
{
- quota_dentry_t *entry = NULL;
- gf_boolean_t found = _gf_false;
+ quota_dentry_t *entry = NULL;
+ gf_boolean_t found = _gf_false;
+ int ret = 0;
- if (list == NULL) {
+ if (!list_empty(list)) {
+ list_for_each_entry(entry, list, next)
+ {
+ if (gf_uuid_compare(pgfid, entry->par) == 0) {
+ found = _gf_true;
goto out;
+ }
}
+ }
- list_for_each_entry (entry, list, next) {
- if (gf_uuid_compare (pgfid, entry->par) == 0) {
- found = _gf_true;
- goto out;
- }
- }
-
- entry = __quota_dentry_new (NULL, name, pgfid);
- if (entry)
- list_add_tail (&entry->next, list);
+ entry = __quota_dentry_new(NULL, name, pgfid);
+ if (entry)
+ list_add_tail(&entry->next, list);
+ else
+ ret = -1;
out:
- if (found)
- return 0;
- else
- return 1;
-
+ if (found)
+ return 0;
+ else if (ret == 0)
+ return 1;
+ else
+ return -1;
}
/* This function iterates the parent list in inode
* context and add unique parent to the list
- * Returns number of dentry added to the list
+ * Returns number of dentry added to the list, or -1 on errors
*/
static int32_t
-quota_add_parents_from_ctx (quota_inode_ctx_t *ctx, struct list_head *list)
+quota_add_parents_from_ctx(quota_inode_ctx_t *ctx, struct list_head *list)
{
- int ret = 0;
- quota_dentry_t *dentry = NULL;
- int32_t count = 0;
+ int ret = 0;
+ quota_dentry_t *dentry = NULL;
+ int32_t count = 0;
- if (ctx == NULL || list == NULL)
- goto out;
+ if (ctx == NULL || list == NULL)
+ goto out;
- LOCK (&ctx->lock);
+ LOCK(&ctx->lock);
+ {
+ list_for_each_entry(dentry, &ctx->parents, next)
{
- list_for_each_entry (dentry, &ctx->parents, next) {
- ret = quota_add_parent (list, dentry->name,
- dentry->par);
-
- if (ret == 1)
- count++;
- }
+ ret = quota_add_parent(list, dentry->name, dentry->par);
+ if (ret == 1)
+ count++;
+ else if (ret == -1)
+ break;
}
- UNLOCK (&ctx->lock);
+ }
+ UNLOCK(&ctx->lock);
out:
- return count;
+ return (ret == -1) ? -1 : count;
}
int32_t
-quota_build_ancestry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- gf_dirent_t *entries, dict_t *xdata)
+quota_build_ancestry_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+ dict_t *xdata)
{
- inode_t *parent = NULL;
- inode_t *tmp_parent = NULL;
- inode_t *linked_inode = NULL;
- inode_t *tmp_inode = NULL;
- gf_dirent_t *entry = NULL;
- loc_t loc = {0, };
- quota_dentry_t *dentry = NULL;
- quota_dentry_t *tmp = NULL;
- quota_inode_ctx_t *ctx = NULL;
- struct list_head parents = {0, };
- quota_local_t *local = NULL;
-
- INIT_LIST_HEAD (&parents);
-
- local = frame->local;
- frame->local = NULL;
-
- if (op_ret < 0)
- goto err;
+ inode_t *parent = NULL;
+ inode_t *tmp_parent = NULL;
+ inode_t *linked_inode = NULL;
+ inode_t *tmp_inode = NULL;
+ gf_dirent_t *entry = NULL;
+ loc_t loc = {
+ 0,
+ };
+ quota_dentry_t *dentry = NULL;
+ quota_dentry_t *tmp = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ struct list_head parents;
+ quota_local_t *local = NULL;
+ int ret;
+
+ INIT_LIST_HEAD(&parents);
+
+ local = frame->local;
+ frame->local = NULL;
+
+ if (op_ret < 0)
+ goto err;
+
+ if ((op_ret > 0) && (entries != NULL)) {
+ list_for_each_entry(entry, &entries->list, list)
+ {
+ if (__is_root_gfid(entry->inode->gfid)) {
+ /* The list contains a sub-list for each
+ * possible path to the target inode. Each
+ * sub-list starts with the root entry of the
+ * tree and is followed by the child entries
+ * for a particular path to the target entry.
+ * The root entry is an implied sub-list
+ * delimiter, as it denotes we have started
+ * processing a new path. Reset the parent
+ * pointer and continue
+ */
- if ((op_ret > 0) && (entries != NULL)) {
- list_for_each_entry (entry, &entries->list, list) {
- if (__is_root_gfid (entry->inode->gfid)) {
- /* The list contains a sub-list for each
- * possible path to the target inode. Each
- * sub-list starts with the root entry of the
- * tree and is followed by the child entries
- * for a particular path to the target entry.
- * The root entry is an implied sub-list
- * delimiter, as it denotes we have started
- * processing a new path. Reset the parent
- * pointer and continue
- */
-
- tmp_parent = NULL;
- } else {
- /* For a non-root entry, link this inode */
- linked_inode = inode_link (entry->inode,
- tmp_parent,
- entry->d_name,
- &entry->d_stat);
- if (linked_inode) {
- tmp_inode = entry->inode;
- entry->inode = linked_inode;
- inode_unref (tmp_inode);
- } else {
- gf_msg (this->name, GF_LOG_WARNING,
- EINVAL, Q_MSG_PARENT_NULL,
- "inode link failed");
- op_errno = EINVAL;
- goto err;
- }
- }
-
- gf_uuid_copy (loc.gfid, entry->d_stat.ia_gfid);
-
- loc.inode = inode_ref (entry->inode);
- loc.parent = inode_ref (tmp_parent);
- loc.name = entry->d_name;
-
- quota_fill_inodectx (this, entry->inode, entry->dict,
- &loc, &entry->d_stat, &op_errno);
-
- /* For non-directory, posix_get_ancestry_non_directory
- * returns all hard-links that are represented by nodes
- * adjacent to each other in the dentry-list.
- * (Unlike the directory case where adjacent nodes
- * either have a parent/child relationship or belong to
- * different paths).
- */
- if (entry->inode->ia_type == IA_IFDIR)
- tmp_parent = entry->inode;
-
- loc_wipe (&loc);
+ tmp_parent = NULL;
+ } else {
+ /* For a non-root entry, link this inode */
+ linked_inode = inode_link(entry->inode, tmp_parent,
+ entry->d_name, &entry->d_stat);
+ if (linked_inode) {
+ tmp_inode = entry->inode;
+ entry->inode = linked_inode;
+ inode_unref(tmp_inode);
+ } else {
+ gf_msg(this->name, GF_LOG_WARNING, EINVAL,
+ Q_MSG_PARENT_NULL, "inode link failed");
+ op_errno = EINVAL;
+ goto err;
}
+ }
+
+ gf_uuid_copy(loc.gfid, entry->d_stat.ia_gfid);
+
+ loc.inode = inode_ref(entry->inode);
+ loc.parent = inode_ref(tmp_parent);
+ loc.name = entry->d_name;
+
+ quota_fill_inodectx(this, entry->inode, entry->dict, &loc,
+ &entry->d_stat, &op_errno);
+
+ /* For non-directory, posix_get_ancestry_non_directory
+ * returns all hard-links that are represented by nodes
+ * adjacent to each other in the dentry-list.
+ * (Unlike the directory case where adjacent nodes
+ * either have a parent/child relationship or belong to
+ * different paths).
+ */
+ if (entry->inode->ia_type == IA_IFDIR)
+ tmp_parent = entry->inode;
+
+ loc_wipe(&loc);
+ }
+ }
+
+ parent = inode_parent(local->loc.inode, 0, NULL);
+ if (parent == NULL) {
+ gf_msg(this->name, GF_LOG_WARNING, EINVAL, Q_MSG_PARENT_NULL,
+ "parent is NULL");
+ op_errno = EINVAL;
+ goto err;
+ }
+
+ quota_inode_ctx_get(local->loc.inode, this, &ctx, 0);
+
+ ret = quota_add_parents_from_ctx(ctx, &parents);
+ if (ret == -1) {
+ op_errno = errno;
+ goto err;
+ }
+
+ if (list_empty(&parents)) {
+ /* we built ancestry for a directory */
+ list_for_each_entry(entry, &entries->list, list)
+ {
+ if (entry->inode == local->loc.inode)
+ break;
}
- parent = inode_parent (local->loc.inode, 0, NULL);
- if (parent == NULL) {
- gf_msg (this->name, GF_LOG_WARNING, EINVAL,
- Q_MSG_PARENT_NULL, "parent is NULL");
- op_errno = EINVAL;
- goto err;
- }
-
- quota_inode_ctx_get (local->loc.inode, this, &ctx, 0);
-
- quota_add_parents_from_ctx (ctx, &parents);
+ /* Getting assertion here, need to investigate
+ comment for now
+ GF_ASSERT (&entry->list != &entries->list);
+ */
- if (list_empty (&parents)) {
- /* we built ancestry for a directory */
- list_for_each_entry (entry, &entries->list, list) {
- if (entry->inode == local->loc.inode)
- break;
- }
-
- /* Getting assertion here, need to investigate
- comment for now
- GF_ASSERT (&entry->list != &entries->list);
- */
-
- quota_add_parent (&parents, entry->d_name, parent->gfid);
+ ret = quota_add_parent(&parents, entry->d_name, parent->gfid);
+ if (ret == -1) {
+ op_errno = errno;
+ goto err;
}
+ }
- local->ancestry_cbk (&parents, local->loc.inode, 0, 0,
- local->ancestry_data);
- goto cleanup;
+ local->ancestry_cbk(&parents, local->loc.inode, 0, 0, local->ancestry_data);
+ goto cleanup;
err:
- local->ancestry_cbk (NULL, NULL, -1, op_errno, local->ancestry_data);
+ local->ancestry_cbk(NULL, NULL, -1, op_errno, local->ancestry_data);
cleanup:
- STACK_DESTROY (frame->root);
- quota_local_cleanup (local);
+ STACK_DESTROY(frame->root);
+ quota_local_cleanup(local);
- if (parent != NULL) {
- inode_unref (parent);
- parent = NULL;
- }
+ if (parent != NULL) {
+ inode_unref(parent);
+ parent = NULL;
+ }
- list_for_each_entry_safe (dentry, tmp, &parents, next) {
- __quota_dentry_free (dentry);
+ if (!list_empty(&parents)) {
+ list_for_each_entry_safe(dentry, tmp, &parents, next)
+ {
+ __quota_dentry_free(dentry);
}
+ }
- return 0;
+ return 0;
}
int
-quota_build_ancestry (inode_t *inode, quota_ancestry_built_t ancestry_cbk,
- void *data)
+quota_build_ancestry(inode_t *inode, quota_ancestry_built_t ancestry_cbk,
+ void *data)
{
- fd_t *fd = NULL;
- quota_local_t *local = NULL;
- call_frame_t *new_frame = NULL;
- int op_errno = ENOMEM;
- int op_ret = -1;
- xlator_t *this = NULL;
- dict_t *xdata_req = NULL;
-
- this = THIS;
-
- xdata_req = dict_new ();
- if (xdata_req == NULL)
- goto err;
+ fd_t *fd = NULL;
+ quota_local_t *local = NULL;
+ call_frame_t *new_frame = NULL;
+ int op_errno = ENOMEM;
+ int op_ret = -1;
+ xlator_t *this = NULL;
+ dict_t *xdata_req = NULL;
+
+ this = THIS;
+
+ xdata_req = dict_new();
+ if (xdata_req == NULL)
+ goto err;
+
+ fd = fd_anonymous(inode);
+ if (fd == NULL)
+ goto err;
+
+ new_frame = create_frame(this, this->ctx->pool);
+ if (new_frame == NULL)
+ goto err;
+
+ local = quota_local_new();
+ if (local == NULL)
+ goto err;
+
+ new_frame->root->uid = new_frame->root->gid = 0;
+ new_frame->local = local;
+ local->ancestry_cbk = ancestry_cbk;
+ local->ancestry_data = data;
+ local->loc.inode = inode_ref(inode);
+
+ op_ret = dict_set_int8(xdata_req, QUOTA_LIMIT_KEY, 1);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto err;
+ }
+
+ op_ret = dict_set_int8(xdata_req, QUOTA_LIMIT_OBJECTS_KEY, 1);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto err;
+ }
+
+ op_ret = dict_set_int8(xdata_req, GET_ANCESTRY_DENTRY_KEY, 1);
+ if (op_ret < 0) {
+ op_errno = -op_ret;
+ goto err;
+ }
+
+ /* This would ask posix layer to construct dentry chain till root
+ * We don't need to do a opendir, we can use the anonymous fd
+ * here for the readidrp.
+ * avoiding opendir also reduces the window size where another FOP
+ * can be executed before completion of build ancestry
+ */
+ STACK_WIND(new_frame, quota_build_ancestry_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdirp, fd, 0, 0, xdata_req);
+
+ op_ret = 0;
- fd = fd_anonymous (inode);
- if (fd == NULL)
- goto err;
-
- new_frame = create_frame (this, this->ctx->pool);
- if (new_frame == NULL)
- goto err;
-
- local = quota_local_new ();
- if (local == NULL)
- goto err;
+err:
+ if (fd)
+ fd_unref(fd);
- new_frame->root->uid = new_frame->root->gid = 0;
- new_frame->local = local;
- local->ancestry_cbk = ancestry_cbk;
- local->ancestry_data = data;
- local->loc.inode = inode_ref (inode);
+ if (xdata_req)
+ dict_unref(xdata_req);
- op_ret = dict_set_int8 (xdata_req, QUOTA_LIMIT_KEY, 1);
- if (op_ret < 0) {
- op_errno = -op_ret;
- goto err;
- }
+ if (op_ret < 0) {
+ ancestry_cbk(NULL, NULL, -1, op_errno, data);
- op_ret = dict_set_int8 (xdata_req, QUOTA_LIMIT_OBJECTS_KEY, 1);
- if (op_ret < 0) {
- op_errno = -op_ret;
- goto err;
+ if (new_frame) {
+ local = new_frame->local;
+ new_frame->local = NULL;
+ STACK_DESTROY(new_frame->root);
}
- op_ret = dict_set_int8 (xdata_req, GET_ANCESTRY_DENTRY_KEY, 1);
- if (op_ret < 0) {
- op_errno = -op_ret;
- goto err;
- }
+ if (local)
+ quota_local_cleanup(local);
+ }
- /* This would ask posix layer to construct dentry chain till root
- * We don't need to do a opendir, we can use the anonymous fd
- * here for the readidrp.
- * avoiding opendir also reduces the window size where another FOP
- * can be executed before completion of build ancestry
- */
- STACK_WIND (new_frame, quota_build_ancestry_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdirp, fd, 0, 0, xdata_req);
+ return 0;
+}
- op_ret = 0;
+int
+quota_validate(call_frame_t *frame, inode_t *inode, xlator_t *this,
+ fop_lookup_cbk_t cbk_fn)
+{
+ quota_local_t *local = NULL;
+ int ret = 0;
+ dict_t *xdata = NULL;
+ quota_priv_t *priv = NULL;
-err:
- if (fd)
- fd_unref (fd);
+ local = frame->local;
+ priv = this->private;
- if (xdata_req)
- dict_unref (xdata_req);
+ LOCK(&local->lock);
+ {
+ loc_wipe(&local->validate_loc);
- if (op_ret < 0) {
- ancestry_cbk (NULL, NULL, -1, op_errno, data);
+ ret = quota_inode_loc_fill(inode, &local->validate_loc);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENFORCEMENT_FAILED,
+ "cannot fill loc for inode (gfid:%s), hence "
+ "aborting quota-checks and continuing with fop",
+ uuid_utoa(inode->gfid));
+ }
+ }
+ UNLOCK(&local->lock);
+
+ if (ret < 0) {
+ ret = -ENOMEM;
+ goto err;
+ }
+
+ xdata = dict_new();
+ if (xdata == NULL) {
+ ret = -ENOMEM;
+ goto err;
+ }
+
+ ret = dict_set_int8(xdata, QUOTA_SIZE_KEY, 1);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "dict set failed");
+ ret = -ENOMEM;
+ goto err;
+ }
+
+ ret = dict_set_str(xdata, "volume-uuid", priv->volume_uuid);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "dict set failed");
+ ret = -ENOMEM;
+ goto err;
+ }
+
+ ret = quota_enforcer_lookup(frame, this, xdata, cbk_fn);
+ if (ret < 0) {
+ ret = -ENOTCONN;
+ goto err;
+ }
+
+ ret = 0;
+err:
+ if (xdata)
+ dict_unref(xdata);
- if (new_frame) {
- local = new_frame->local;
- new_frame->local = NULL;
- STACK_DESTROY (new_frame->root);
- }
+ return ret;
+}
- if (local)
- quota_local_cleanup (local);
+void
+quota_check_limit_continuation(struct list_head *parents, inode_t *inode,
+ int32_t op_ret, int32_t op_errno, void *data)
+{
+ call_frame_t *frame = NULL;
+ xlator_t *this = NULL;
+ quota_local_t *local = NULL;
+ quota_local_t *par_local = NULL;
+ quota_dentry_t *entry = NULL;
+ inode_t *parent = NULL;
+ int parent_count = 0;
+
+ frame = data;
+ local = frame->local;
+ this = THIS;
+
+ if (local->par_frame)
+ par_local = local->par_frame->local;
+ else
+ par_local = local;
+
+ if ((op_ret < 0) || list_empty(parents)) {
+ if (op_ret >= 0) {
+ gf_msg(this->name, GF_LOG_WARNING, EIO, Q_MSG_ANCESTRY_BUILD_FAILED,
+ "Couldn't build ancestry for inode (gfid:%s). "
+ "Without knowing ancestors till root, quota"
+ "cannot be enforced. "
+ "Hence, failing fop with EIO",
+ uuid_utoa(inode->gfid));
+ op_errno = EIO;
+ }
+
+ quota_handle_validate_error(frame, -1, op_errno);
+ goto out;
+ }
+
+ list_for_each_entry(entry, parents, next) { parent_count++; }
+
+ LOCK(&par_local->lock);
+ {
+ par_local->link_count += (parent_count - 1);
+ }
+ UNLOCK(&par_local->lock);
+
+ if (local->par_frame) {
+ list_for_each_entry(entry, parents, next)
+ {
+ parent = inode_find(inode->table, entry->par);
+ quota_check_limit(frame, parent, this);
+ inode_unref(parent);
+ }
+ } else {
+ list_for_each_entry(entry, parents, next)
+ {
+ parent = do_quota_check_limit(frame, inode, this, entry, _gf_true);
+ if (parent)
+ inode_unref(parent);
+ else
+ quota_link_count_decrement(frame);
}
+ }
- return 0;
+out:
+ return;
}
-int
-quota_validate (call_frame_t *frame, inode_t *inode, xlator_t *this,
- fop_lookup_cbk_t cbk_fn)
+int32_t
+quota_check_object_limit(call_frame_t *frame, quota_inode_ctx_t *ctx,
+ quota_priv_t *priv, inode_t *_inode, xlator_t *this,
+ int32_t *op_errno, int just_validated,
+ quota_local_t *local, gf_boolean_t *skip_check)
{
- quota_local_t *local = NULL;
- int ret = 0;
- dict_t *xdata = NULL;
- quota_priv_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- LOCK (&local->lock);
+ int32_t ret = -1;
+ uint32_t timeout = 0;
+ char need_validate = 0;
+ gf_boolean_t hard_limit_exceeded = 0;
+ int64_t object_aggr_count = 0;
+
+ GF_ASSERT(frame);
+ GF_ASSERT(priv);
+ GF_ASSERT(_inode);
+ GF_ASSERT(this);
+ GF_ASSERT(local);
+
+ if (ctx != NULL && (ctx->object_hard_lim > 0 || ctx->object_soft_lim)) {
+ LOCK(&ctx->lock);
{
- loc_wipe (&local->validate_loc);
-
- ret = quota_inode_loc_fill (inode, &local->validate_loc);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENFORCEMENT_FAILED,
- "cannot fill loc for inode (gfid:%s), hence "
- "aborting quota-checks and continuing with fop",
- uuid_utoa (inode->gfid));
- }
- }
- UNLOCK (&local->lock);
+ timeout = priv->soft_timeout;
- if (ret < 0) {
- ret = -ENOMEM;
- goto err;
- }
+ object_aggr_count = ctx->file_count + ctx->dir_count + 1;
+ if (((ctx->object_soft_lim >= 0) &&
+ (object_aggr_count) > ctx->object_soft_lim)) {
+ timeout = priv->hard_timeout;
+ }
- xdata = dict_new ();
- if (xdata == NULL) {
- ret = -ENOMEM;
- goto err;
+ if (!just_validated && quota_timeout(ctx->validate_time, timeout)) {
+ need_validate = 1;
+ } else if ((object_aggr_count) > ctx->object_hard_lim) {
+ hard_limit_exceeded = 1;
+ }
}
+ UNLOCK(&ctx->lock);
- ret = dict_set_int8 (xdata, QUOTA_SIZE_KEY, 1);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "dict set failed");
- ret = -ENOMEM;
- goto err;
+ if (need_validate && *skip_check != _gf_true) {
+ *skip_check = _gf_true;
+ ret = quota_validate(frame, _inode, this, quota_validate_cbk);
+ if (ret < 0) {
+ *op_errno = -ret;
+ *skip_check = _gf_false;
+ }
+ goto out;
}
- ret = dict_set_str (xdata, "volume-uuid", priv->volume_uuid);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "dict set failed");
- ret = -ENOMEM;
- goto err;
+ if (hard_limit_exceeded) {
+ local->op_ret = -1;
+ local->op_errno = EDQUOT;
+ *op_errno = EDQUOT;
+ goto out;
}
- ret = quota_enforcer_lookup (frame, this, xdata, cbk_fn);
- if (ret < 0) {
- ret = -ENOTCONN;
- goto err;
- }
+ /*We log usage only if quota limit is configured on
+ that inode
+ */
+ quota_log_usage(this, ctx, _inode, 0);
+ }
- ret = 0;
-err:
- if (xdata)
- dict_unref (xdata);
+ ret = 0;
- return ret;
+out:
+ return ret;
}
-void
-quota_check_limit_continuation (struct list_head *parents, inode_t *inode,
- int32_t op_ret, int32_t op_errno, void *data)
+int32_t
+quota_check_size_limit(call_frame_t *frame, quota_inode_ctx_t *ctx,
+ quota_priv_t *priv, inode_t *_inode, xlator_t *this,
+ int32_t *op_errno, int just_validated, int64_t delta,
+ quota_local_t *local, gf_boolean_t *skip_check)
{
- call_frame_t *frame = NULL;
- xlator_t *this = NULL;
- quota_local_t *local = NULL;
- quota_local_t *par_local = NULL;
- quota_dentry_t *entry = NULL;
- inode_t *parent = NULL;
- int parent_count = 0;
-
- frame = data;
- local = frame->local;
- this = THIS;
-
- if (local->par_frame)
- par_local = local->par_frame->local;
- else
- par_local = local;
-
-
- if ((op_ret < 0) || list_empty (parents)) {
- if (op_ret >= 0) {
- gf_msg (this->name, GF_LOG_WARNING, EIO,
- Q_MSG_ANCESTRY_BUILD_FAILED,
- "Couldn't build ancestry for inode (gfid:%s). "
- "Without knowing ancestors till root, quota"
- "cannot be enforced. "
- "Hence, failing fop with EIO",
- uuid_utoa (inode->gfid));
- op_errno = EIO;
- }
+ int32_t ret = -1;
+ uint32_t timeout = 0;
+ char need_validate = 0;
+ gf_boolean_t hard_limit_exceeded = 0;
+ int64_t space_available = 0;
+ int64_t wouldbe_size = 0;
+
+ GF_ASSERT(frame);
+ GF_ASSERT(priv);
+ GF_ASSERT(_inode);
+ GF_ASSERT(this);
+ GF_ASSERT(local);
+
+ if (ctx != NULL && (ctx->hard_lim > 0 || ctx->soft_lim > 0)) {
+ wouldbe_size = ctx->size + delta;
+
+ LOCK(&ctx->lock);
+ {
+ timeout = priv->soft_timeout;
- quota_handle_validate_error (frame, -1, op_errno);
- goto out;
- }
+ if ((ctx->soft_lim >= 0) && (wouldbe_size > ctx->soft_lim)) {
+ timeout = priv->hard_timeout;
+ }
- list_for_each_entry (entry, parents, next) {
- parent_count++;
+ if (!just_validated && quota_timeout(ctx->validate_time, timeout)) {
+ need_validate = 1;
+ } else if (wouldbe_size >= ctx->hard_lim) {
+ hard_limit_exceeded = 1;
+ }
}
+ UNLOCK(&ctx->lock);
- LOCK (&par_local->lock);
- {
- par_local->link_count += (parent_count - 1);
+ if (need_validate && *skip_check != _gf_true) {
+ *skip_check = _gf_true;
+ ret = quota_validate(frame, _inode, this, quota_validate_cbk);
+ if (ret < 0) {
+ *op_errno = -ret;
+ *skip_check = _gf_false;
+ }
+ goto out;
}
- UNLOCK (&par_local->lock);
- if (local->par_frame) {
- list_for_each_entry (entry, parents, next) {
- parent = inode_find (inode->table, entry->par);
- quota_check_limit (frame, parent, this);
- inode_unref (parent);
- }
- } else {
- list_for_each_entry (entry, parents, next) {
- parent = do_quota_check_limit (frame, inode, this,
- entry, _gf_true);
- if (parent)
- inode_unref (parent);
- else
- quota_link_count_decrement (frame);
- }
- }
+ if (hard_limit_exceeded) {
+ local->op_ret = -1;
+ local->op_errno = EDQUOT;
-out:
- return;
-}
+ space_available = ctx->hard_lim - ctx->size;
-int32_t
-quota_check_object_limit (call_frame_t *frame, quota_inode_ctx_t *ctx,
- quota_priv_t *priv, inode_t *_inode, xlator_t *this,
- int32_t *op_errno, int just_validated,
- quota_local_t *local, gf_boolean_t *skip_check)
-{
- int32_t ret = -1;
- uint32_t timeout = 0;
- char need_validate = 0;
- gf_boolean_t hard_limit_exceeded = 0;
- int64_t object_aggr_count = 0;
-
- GF_ASSERT (frame);
- GF_ASSERT (priv);
- GF_ASSERT (_inode);
- GF_ASSERT (this);
- GF_ASSERT (local);
-
- if (ctx != NULL && (ctx->object_hard_lim > 0 ||
- ctx->object_soft_lim)) {
- LOCK (&ctx->lock);
- {
- timeout = priv->soft_timeout;
-
- object_aggr_count = ctx->file_count +
- ctx->dir_count + 1;
- if (((ctx->object_soft_lim >= 0)
- && (object_aggr_count) >
- ctx->object_soft_lim)) {
- timeout = priv->hard_timeout;
- }
-
- if (!just_validated
- && quota_timeout (&ctx->tv, timeout)) {
- need_validate = 1;
- } else if ((object_aggr_count) >
- ctx->object_hard_lim) {
- hard_limit_exceeded = 1;
- }
- }
- UNLOCK (&ctx->lock);
-
- if (need_validate && *skip_check != _gf_true) {
- *skip_check = _gf_true;
- ret = quota_validate (frame, _inode, this,
- quota_validate_cbk);
- if (ret < 0) {
- *op_errno = -ret;
- *skip_check = _gf_false;
- }
- goto out;
- }
+ if (space_available < 0)
+ space_available = 0;
- if (hard_limit_exceeded) {
- local->op_ret = -1;
- local->op_errno = EDQUOT;
- *op_errno = EDQUOT;
- goto out;
- }
+ if ((local->space_available < 0) ||
+ (local->space_available > space_available)) {
+ local->space_available = space_available;
+ }
- /*We log usage only if quota limit is configured on
- that inode
- */
- quota_log_usage (this, ctx, _inode, 0);
+ if (space_available == 0) {
+ *op_errno = EDQUOT;
+ goto out;
+ }
}
- ret = 0;
+ /* We log usage only if quota limit is configured on
+ that inode. */
+ quota_log_usage(this, ctx, _inode, delta);
+ }
+ ret = 0;
out:
- return ret;
+ return ret;
}
-
int32_t
-quota_check_size_limit (call_frame_t *frame, quota_inode_ctx_t *ctx,
- quota_priv_t *priv, inode_t *_inode, xlator_t *this,
- int32_t *op_errno, int just_validated, int64_t delta,
- quota_local_t *local, gf_boolean_t *skip_check)
+quota_check_limit(call_frame_t *frame, inode_t *inode, xlator_t *this)
{
- int32_t ret = -1;
- uint32_t timeout = 0;
- char need_validate = 0;
- gf_boolean_t hard_limit_exceeded = 0;
- int64_t space_available = 0;
- int64_t wouldbe_size = 0;
-
- GF_ASSERT (frame);
- GF_ASSERT (priv);
- GF_ASSERT (_inode);
- GF_ASSERT (this);
- GF_ASSERT (local);
-
- if (ctx != NULL && (ctx->hard_lim > 0 || ctx->soft_lim > 0)) {
- wouldbe_size = ctx->size + delta;
-
- LOCK (&ctx->lock);
- {
- timeout = priv->soft_timeout;
-
- if ((ctx->soft_lim >= 0)
- && (wouldbe_size > ctx->soft_lim)) {
- timeout = priv->hard_timeout;
- }
-
- if (!just_validated
- && quota_timeout (&ctx->tv, timeout)) {
- need_validate = 1;
- } else if (wouldbe_size >= ctx->hard_lim) {
- hard_limit_exceeded = 1;
- }
- }
- UNLOCK (&ctx->lock);
-
- if (need_validate && *skip_check != _gf_true) {
- *skip_check = _gf_true;
- ret = quota_validate (frame, _inode, this,
- quota_validate_cbk);
- if (ret < 0) {
- *op_errno = -ret;
- *skip_check = _gf_false;
- }
- goto out;
- }
-
- if (hard_limit_exceeded) {
- local->op_ret = -1;
- local->op_errno = EDQUOT;
-
- space_available = ctx->hard_lim - ctx->size;
+ int32_t ret = -1, op_errno = EINVAL;
+ inode_t *_inode = NULL, *parent = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ quota_priv_t *priv = NULL;
+ quota_local_t *local = NULL;
+ quota_local_t *par_local = NULL;
+ char just_validated = 0;
+ int64_t delta = 0;
+ int8_t object_delta = 0;
+ uint64_t value = 0;
+ gf_boolean_t skip_check = _gf_false;
+
+ GF_VALIDATE_OR_GOTO("quota", this, err);
+ GF_VALIDATE_OR_GOTO(this->name, frame, err);
+ GF_VALIDATE_OR_GOTO(this->name, inode, err);
+
+ local = frame->local;
+ GF_VALIDATE_OR_GOTO(this->name, local, err);
+
+ if (local->par_frame) {
+ par_local = local->par_frame->local;
+ GF_VALIDATE_OR_GOTO(this->name, par_local, err);
+ } else {
+ par_local = local;
+ }
+
+ delta = par_local->delta;
+ object_delta = par_local->object_delta;
+
+ GF_VALIDATE_OR_GOTO(this->name, par_local->stub, err);
+ /* Allow all the trusted clients
+ * Don't block the gluster internal processes like rebalance, gsyncd,
+ * self heal etc from the disk quotas.
+ *
+ * Method: Allow all the clients with PID negative. This is by the
+ * assumption that any kernel assigned pid doesn't have the negative
+ * number.
+ */
+ if (0 > frame->root->pid) {
+ ret = 0;
+ quota_link_count_decrement(frame);
+ goto done;
+ }
- if (space_available < 0)
- space_available = 0;
+ priv = this->private;
- if ((local->space_available < 0)
- || (local->space_available
- > space_available)){
- local->space_available
- = space_available;
+ inode_ctx_get(inode, this, &value);
+ ctx = (quota_inode_ctx_t *)(unsigned long)value;
- }
+ _inode = inode_ref(inode);
- if (space_available == 0) {
- *op_errno = EDQUOT;
- goto out;
- }
- }
+ LOCK(&local->lock);
+ {
+ just_validated = local->just_validated;
+ local->just_validated = 0;
+ }
+ UNLOCK(&local->lock);
- /* We log usage only if quota limit is configured on
- that inode. */
- quota_log_usage (this, ctx, _inode, delta);
+ do {
+ /* In a rename operation, enforce should be stopped at common
+ ancestor */
+ if (!gf_uuid_is_null(par_local->common_ancestor) &&
+ !gf_uuid_compare(_inode->gfid, par_local->common_ancestor)) {
+ quota_link_count_decrement(frame);
+ break;
}
- ret = 0;
-out:
- return ret;
-}
+ if (object_delta <= 0)
+ goto skip_check_object_limit;
+ ret = quota_check_object_limit(frame, ctx, priv, _inode, this,
+ &op_errno, just_validated, par_local,
+ &skip_check);
+ if (skip_check == _gf_true)
+ goto done;
-int32_t
-quota_check_limit (call_frame_t *frame, inode_t *inode, xlator_t *this)
-{
- int32_t ret = -1, op_errno = EINVAL;
- inode_t *_inode = NULL, *parent = NULL;
- quota_inode_ctx_t *ctx = NULL;
- quota_priv_t *priv = NULL;
- quota_local_t *local = NULL;
- quota_local_t *par_local = NULL;
- char just_validated = 0;
- int64_t delta = 0;
- int8_t object_delta = 0;
- uint64_t value = 0;
- gf_boolean_t skip_check = _gf_false;
-
- GF_VALIDATE_OR_GOTO ("quota", this, err);
- GF_VALIDATE_OR_GOTO (this->name, frame, err);
- GF_VALIDATE_OR_GOTO (this->name, inode, err);
-
- local = frame->local;
- GF_VALIDATE_OR_GOTO (this->name, local, err);
-
- if (local->par_frame) {
- par_local = local->par_frame->local;
- GF_VALIDATE_OR_GOTO (this->name, par_local, err);
- } else {
- par_local = local;
- }
-
- delta = par_local->delta;
- object_delta = par_local->object_delta;
-
- GF_VALIDATE_OR_GOTO (this->name, par_local->stub, err);
- /* Allow all the trusted clients
- * Don't block the gluster internal processes like rebalance, gsyncd,
- * self heal etc from the disk quotas.
- *
- * Method: Allow all the clients with PID negative. This is by the
- * assumption that any kernel assigned pid doesn't have the negative
- * number.
- */
- if (0 > frame->root->pid) {
- ret = 0;
- quota_link_count_decrement (frame);
- goto done;
+ if (ret) {
+ if (op_errno != EDQUOT)
+ gf_msg(this->name, GF_LOG_ERROR, 0, Q_MSG_ENFORCEMENT_FAILED,
+ "Failed to "
+ "check quota object limit");
+ goto err;
}
- priv = this->private;
-
- inode_ctx_get (inode, this, &value);
- ctx = (quota_inode_ctx_t *)(unsigned long)value;
+ skip_check_object_limit:
+ ret = quota_check_size_limit(frame, ctx, priv, _inode, this, &op_errno,
+ just_validated, delta, par_local,
+ &skip_check);
+ if (skip_check == _gf_true)
+ goto done;
- _inode = inode_ref (inode);
-
- LOCK (&local->lock);
- {
- just_validated = local->just_validated;
- local->just_validated = 0;
+ if (ret) {
+ if (op_errno != EDQUOT)
+ gf_msg(this->name, GF_LOG_ERROR, 0, Q_MSG_ENFORCEMENT_FAILED,
+ "Failed to "
+ "check quota size limit");
+ goto err;
}
- UNLOCK (&local->lock);
-
- do {
- /* In a rename operation, enforce should be stopped at common
- ancestor */
- if (!gf_uuid_is_null (par_local->common_ancestor) &&
- !gf_uuid_compare (_inode->gfid, par_local->common_ancestor)
- ) {
- quota_link_count_decrement (frame);
- break;
- }
-
- if (object_delta <= 0)
- goto skip_check_object_limit;
-
- ret = quota_check_object_limit (frame, ctx, priv, _inode, this,
- &op_errno, just_validated,
- par_local, &skip_check);
- if (skip_check == _gf_true)
- goto done;
-
- if (ret) {
- if (op_errno != EDQUOT)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- Q_MSG_ENFORCEMENT_FAILED, "Failed to "
- "check quota object limit");
- goto err;
- }
-skip_check_object_limit:
- ret = quota_check_size_limit (frame, ctx, priv, _inode, this,
- &op_errno, just_validated, delta,
- par_local, &skip_check);
- if (skip_check == _gf_true)
- goto done;
-
- if (ret) {
- if (op_errno != EDQUOT)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- Q_MSG_ENFORCEMENT_FAILED, "Failed to "
- "check quota size limit");
- goto err;
- }
+ if (__is_root_gfid(_inode->gfid)) {
+ quota_link_count_decrement(frame);
+ break;
+ }
- if (__is_root_gfid (_inode->gfid)) {
- quota_link_count_decrement (frame);
- break;
- }
+ parent = inode_parent(_inode, 0, NULL);
+ if (parent == NULL) {
+ ret = quota_build_ancestry(_inode, quota_check_limit_continuation,
+ frame);
+ if (ret < 0) {
+ op_errno = -ret;
+ goto err;
+ }
- parent = inode_parent (_inode, 0, NULL);
- if (parent == NULL) {
- ret = quota_build_ancestry (_inode,
- quota_check_limit_continuation,
- frame);
- if (ret < 0) {
- op_errno = -ret;
- goto err;
- }
-
- break;
- }
+ break;
+ }
- inode_unref (_inode);
- _inode = parent;
- just_validated = 0;
+ inode_unref(_inode);
+ _inode = parent;
+ just_validated = 0;
- value = 0;
- inode_ctx_get (_inode, this, &value);
- ctx = (quota_inode_ctx_t *)(unsigned long)value;
- } while (1);
+ value = 0;
+ inode_ctx_get(_inode, this, &value);
+ ctx = (quota_inode_ctx_t *)(unsigned long)value;
+ } while (1);
done:
- if (_inode != NULL) {
- inode_unref (_inode);
- _inode = NULL;
- }
- return 0;
+ if (_inode != NULL) {
+ inode_unref(_inode);
+ _inode = NULL;
+ }
+ return 0;
err:
- quota_handle_validate_error (frame, -1, op_errno);
+ quota_handle_validate_error(frame, -1, op_errno);
- inode_unref (_inode);
- return 0;
+ inode_unref(_inode);
+ return 0;
}
inode_t *
-do_quota_check_limit (call_frame_t *frame, inode_t *inode, xlator_t *this,
- quota_dentry_t *dentry, gf_boolean_t force)
+do_quota_check_limit(call_frame_t *frame, inode_t *inode, xlator_t *this,
+ quota_dentry_t *dentry, gf_boolean_t force)
{
- int32_t ret = -1;
- inode_t *parent = NULL;
- call_frame_t *new_frame = NULL;
- quota_local_t *new_local = NULL;
-
- parent = inode_parent (inode, dentry->par, dentry->name);
- if (parent == NULL) {
- if (force)
- parent = inode_find (inode->table, dentry->par);
- else
- goto out;
- }
- if (parent == NULL)
- goto out;
+ int32_t ret = -1;
+ inode_t *parent = NULL;
+ call_frame_t *new_frame = NULL;
+ quota_local_t *new_local = NULL;
+
+ parent = inode_parent(inode, dentry->par, dentry->name);
+ if (parent == NULL) {
+ if (force)
+ parent = inode_find(inode->table, dentry->par);
+ else
+ goto out;
+ }
+ if (parent == NULL)
+ goto out;
- new_frame = copy_frame (frame);
- if (new_frame == NULL)
- goto out;
+ new_frame = copy_frame(frame);
+ if (new_frame == NULL)
+ goto out;
- new_local = quota_local_new ();
- if (new_local == NULL)
- goto out;
+ new_local = quota_local_new();
+ if (new_local == NULL)
+ goto out;
- new_frame->local = new_local;
- new_local->par_frame = frame;
+ new_frame->local = new_local;
+ new_local->par_frame = frame;
- quota_check_limit (new_frame, parent, this);
+ quota_check_limit(new_frame, parent, this);
- ret = 0;
+ ret = 0;
out:
- if (ret < 0) {
- if (parent) {
- /* Caller should decrement link_count, in case parent is
- * NULL
- */
- quota_handle_validate_error (frame, -1, ENOMEM);
- }
-
- if (new_frame) {
- new_frame->local = NULL;
- STACK_DESTROY (new_frame->root);
- }
+ if (ret < 0) {
+ if (parent) {
+ /* Caller should decrement link_count, in case parent is
+ * NULL
+ */
+ quota_handle_validate_error(frame, -1, ENOMEM);
+ }
- if (new_local)
- quota_local_cleanup (new_local);
+ if (new_frame) {
+ new_frame->local = NULL;
+ STACK_DESTROY(new_frame->root);
}
+ }
- return parent;
+ return parent;
}
static int
-quota_get_limits (xlator_t *this, dict_t *dict, int64_t *hard_lim,
- int64_t *soft_lim, int64_t *object_hard_limit,
- int64_t *object_soft_limit)
+quota_get_limits(xlator_t *this, dict_t *dict, int64_t *hard_lim,
+ int64_t *soft_lim, int64_t *object_hard_limit,
+ int64_t *object_soft_limit)
{
- quota_limits_t *limit = NULL;
- quota_limits_t *object_limit = NULL;
- quota_priv_t *priv = NULL;
- int64_t soft_lim_percent = 0;
- int64_t *ptr = NULL;
- int ret = 0;
-
- if ((this == NULL) || (dict == NULL) || (hard_lim == NULL)
- || (soft_lim == NULL))
- goto out;
+ quota_limits_t *limit = NULL;
+ quota_limits_t *object_limit = NULL;
+ quota_priv_t *priv = NULL;
+ int64_t soft_lim_percent = 0;
+ int64_t *ptr = NULL;
+ int ret = 0;
- priv = this->private;
+ if ((this == NULL) || (dict == NULL) || (hard_lim == NULL) ||
+ (soft_lim == NULL))
+ goto out;
- ret = dict_get_bin (dict, QUOTA_LIMIT_KEY, (void **) &ptr);
- limit = (quota_limits_t *)ptr;
+ priv = this->private;
- if (limit) {
- *hard_lim = ntoh64 (limit->hl);
- soft_lim_percent = ntoh64 (limit->sl);
- }
+ ret = dict_get_bin(dict, QUOTA_LIMIT_KEY, (void **)&ptr);
+ limit = (quota_limits_t *)ptr;
- if (soft_lim_percent < 0) {
- soft_lim_percent = priv->default_soft_lim;
- }
+ if (limit) {
+ *hard_lim = ntoh64(limit->hl);
+ soft_lim_percent = ntoh64(limit->sl);
+ }
- if ((*hard_lim > 0) && (soft_lim_percent > 0)) {
- *soft_lim = (soft_lim_percent * (*hard_lim))/100;
- }
+ if (soft_lim_percent < 0) {
+ soft_lim_percent = priv->default_soft_lim;
+ }
- ret = dict_get_bin (dict, QUOTA_LIMIT_OBJECTS_KEY, (void **) &ptr);
- if (ret)
- return 0;
- object_limit = (quota_limits_t *)ptr;
+ if ((*hard_lim > 0) && (soft_lim_percent > 0)) {
+ *soft_lim = (soft_lim_percent * (*hard_lim)) / 100;
+ }
- if (object_limit) {
- *object_hard_limit = ntoh64 (object_limit->hl);
- soft_lim_percent = ntoh64 (object_limit->sl);
- }
+ ret = dict_get_bin(dict, QUOTA_LIMIT_OBJECTS_KEY, (void **)&ptr);
+ if (ret)
+ return 0;
+ object_limit = (quota_limits_t *)ptr;
- if (soft_lim_percent < 0) {
- soft_lim_percent = priv->default_soft_lim;
- }
+ if (object_limit) {
+ *object_hard_limit = ntoh64(object_limit->hl);
+ soft_lim_percent = ntoh64(object_limit->sl);
+ }
- if ((*object_hard_limit > 0) && (soft_lim_percent > 0)) {
- *object_soft_limit = (soft_lim_percent *
- (*object_hard_limit))/100;
- }
+ if (soft_lim_percent < 0) {
+ soft_lim_percent = priv->default_soft_lim;
+ }
+
+ if ((*object_hard_limit > 0) && (soft_lim_percent > 0)) {
+ *object_soft_limit = (soft_lim_percent * (*object_hard_limit)) / 100;
+ }
out:
- return 0;
+ return 0;
}
int
-quota_fill_inodectx (xlator_t *this, inode_t *inode, dict_t *dict,
- loc_t *loc, struct iatt *buf, int32_t *op_errno)
+quota_fill_inodectx(xlator_t *this, inode_t *inode, dict_t *dict, loc_t *loc,
+ struct iatt *buf, int32_t *op_errno)
{
- int32_t ret = -1;
- char found = 0;
- quota_inode_ctx_t *ctx = NULL;
- quota_dentry_t *dentry = NULL;
- uint64_t value = 0;
- int64_t hard_lim = 0;
- int64_t soft_lim = 0;
- int64_t object_hard_limit = 0;
- int64_t object_soft_limit = 0;
-
- quota_get_limits (this, dict, &hard_lim, &soft_lim, &object_hard_limit,
- &object_soft_limit);
-
- inode_ctx_get (inode, this, &value);
- ctx = (quota_inode_ctx_t *)(unsigned long)value;
-
- if ((((ctx == NULL) || (ctx->hard_lim == hard_lim))
- && (hard_lim < 0) && !QUOTA_REG_OR_LNK_FILE (buf->ia_type))) {
- ret = 0;
- goto out;
- }
-
- ret = quota_inode_ctx_get (inode, this, &ctx, 1);
- if ((ret == -1) || (ctx == NULL)) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_INODE_CTX_GET_FAILED, "cannot create quota "
- "context in inode(gfid:%s)", uuid_utoa (inode->gfid));
- ret = -1;
- *op_errno = ENOMEM;
- goto out;
- }
-
- LOCK (&ctx->lock);
+ int32_t ret = -1;
+ char found = 0;
+ quota_inode_ctx_t *ctx = NULL;
+ quota_dentry_t *dentry = NULL;
+ uint64_t value = 0;
+ int64_t hard_lim = 0;
+ int64_t soft_lim = 0;
+ int64_t object_hard_limit = 0;
+ int64_t object_soft_limit = 0;
+
+ quota_get_limits(this, dict, &hard_lim, &soft_lim, &object_hard_limit,
+ &object_soft_limit);
+
+ inode_ctx_get(inode, this, &value);
+ ctx = (quota_inode_ctx_t *)(unsigned long)value;
+
+ if ((((ctx == NULL) || (ctx->hard_lim == hard_lim)) && (hard_lim < 0) &&
+ !QUOTA_REG_OR_LNK_FILE(buf->ia_type))) {
+ ret = 0;
+ goto out;
+ }
+
+ ret = quota_inode_ctx_get(inode, this, &ctx, 1);
+ if ((ret == -1) || (ctx == NULL)) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_INODE_CTX_GET_FAILED,
+ "cannot create quota "
+ "context in inode(gfid:%s)",
+ uuid_utoa(inode->gfid));
+ ret = -1;
+ *op_errno = ENOMEM;
+ goto out;
+ }
+
+ LOCK(&ctx->lock);
+ {
+ ctx->hard_lim = hard_lim;
+ ctx->soft_lim = soft_lim;
+ ctx->object_hard_lim = object_hard_limit;
+ ctx->object_soft_lim = object_soft_limit;
+
+ ctx->buf = *buf;
+
+ if (!QUOTA_REG_OR_LNK_FILE(buf->ia_type)) {
+ goto unlock;
+ }
+
+ /* do nothing if it is a nameless lookup */
+ if (loc->name == NULL || !loc->parent)
+ goto unlock;
+
+ list_for_each_entry(dentry, &ctx->parents, next)
{
- ctx->hard_lim = hard_lim;
- ctx->soft_lim = soft_lim;
- ctx->object_hard_lim = object_hard_limit;
- ctx->object_soft_lim = object_soft_limit;
-
- ctx->buf = *buf;
-
- if (!QUOTA_REG_OR_LNK_FILE (buf->ia_type)) {
- goto unlock;
- }
-
- /* do nothing if it is a nameless lookup */
- if (loc->name == NULL || !loc->parent)
- goto unlock;
-
- list_for_each_entry (dentry, &ctx->parents, next) {
- if ((strcmp (dentry->name, loc->name) == 0) &&
- (gf_uuid_compare (loc->parent->gfid,
- dentry->par) == 0)) {
- found = 1;
- break;
- }
- }
+ if ((strcmp(dentry->name, loc->name) == 0) &&
+ (gf_uuid_compare(loc->parent->gfid, dentry->par) == 0)) {
+ found = 1;
+ break;
+ }
+ }
- if (!found) {
- dentry = __quota_dentry_new (ctx,
- (char *)loc->name,
- loc->parent->gfid);
- if (dentry == NULL) {
- /*
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM,
- "cannot create a new dentry (par:%"
+ if (!found) {
+ dentry = __quota_dentry_new(ctx, (char *)loc->name,
+ loc->parent->gfid);
+ if (dentry == NULL) {
+ /*
+ gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
+ Q_MSG_ENOMEM,
+ "cannot create a new dentry (par:%"
- PRId64", name:%s) for inode(ino:%"
- PRId64", gfid:%s)",
- uuid_utoa (local->loc.inode->gfid));
- */
- ret = -1;
- *op_errno = ENOMEM;
- goto unlock;
- }
- }
+ */
+ ret = -1;
+ *op_errno = ENOMEM;
+ goto unlock;
+ }
}
+ }
unlock:
- UNLOCK (&ctx->lock);
+ UNLOCK(&ctx->lock);
out:
- return ret;
+ return ret;
}
/*
* return _gf_true if enforcement is needed and _gf_false otherwise
*/
gf_boolean_t
-should_quota_enforce (xlator_t *this, dict_t *dict, glusterfs_fop_t fop)
+should_quota_enforce(xlator_t *this, dict_t *dict, glusterfs_fop_t fop)
{
- int ret = 0;
+ int ret = 0;
- ret = dict_check_flag(dict, GF_INTERNAL_CTX_KEY, GF_DHT_HEAL_DIR);
+ ret = dict_check_flag(dict, GF_INTERNAL_CTX_KEY, GF_DHT_HEAL_DIR);
- if (fop == GF_FOP_MKDIR && ret == DICT_FLAG_SET) {
- return _gf_false;
- } else if (ret == -ENOENT) {
- gf_msg (this->name, GF_LOG_DEBUG, EINVAL,
- Q_MSG_INTERNAL_FOP_KEY_MISSING,
- "No internal fop context present");
- goto out;
- }
+ if (fop == GF_FOP_MKDIR && ret == DICT_FLAG_SET) {
+ return _gf_false;
+ } else if (ret == -ENOENT) {
+ gf_msg(this->name, GF_LOG_DEBUG, EINVAL, Q_MSG_INTERNAL_FOP_KEY_MISSING,
+ "No internal fop context present");
+ goto out;
+ }
out:
- return _gf_true;
+ return _gf_true;
}
int32_t
-quota_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *dict, struct iatt *postparent)
+quota_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *dict, struct iatt *postparent)
{
- quota_local_t *local = NULL;
- inode_t *this_inode = NULL;
+ quota_local_t *local = NULL;
+ inode_t *this_inode = NULL;
- local = frame->local;
- frame->local = NULL;
+ local = frame->local;
+ frame->local = NULL;
- if (op_ret >= 0 && inode) {
- this_inode = inode_ref (inode);
+ if (op_ret >= 0 && inode) {
+ this_inode = inode_ref(inode);
- op_ret = quota_fill_inodectx (this, inode, dict, &local->loc,
- buf, &op_errno);
- if (op_ret < 0)
- op_errno = ENOMEM;
- }
+ op_ret = quota_fill_inodectx(this, inode, dict, &local->loc, buf,
+ &op_errno);
+ if (op_ret < 0)
+ op_errno = ENOMEM;
+ }
- QUOTA_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, buf,
- dict, postparent);
+ QUOTA_STACK_UNWIND(lookup, frame, op_ret, op_errno, inode, buf, dict,
+ postparent);
- if (op_ret < 0 || this_inode == NULL || gf_uuid_is_null(this_inode->gfid))
- goto out;
+ if (op_ret < 0 || this_inode == NULL || gf_uuid_is_null(this_inode->gfid))
+ goto out;
- check_ancestory_2 (this, local, this_inode);
+ check_ancestory_2(this, local, this_inode);
out:
- if (this_inode)
- inode_unref (this_inode);
+ if (this_inode)
+ inode_unref(this_inode);
- quota_local_cleanup (local);
+ quota_local_cleanup(local);
- return 0;
+ return 0;
}
int32_t
-quota_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *xattr_req)
+quota_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr_req)
{
- quota_priv_t *priv = NULL;
- int32_t ret = -1;
- quota_local_t *local = NULL;
+ quota_priv_t *priv = NULL;
+ int32_t ret = -1;
+ quota_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
- xattr_req = xattr_req ? dict_ref(xattr_req) : dict_new();
- if (!xattr_req)
- goto err;
+ xattr_req = xattr_req ? dict_ref(xattr_req) : dict_new();
+ if (!xattr_req)
+ goto err;
- local = quota_local_new ();
- if (local == NULL) {
- goto err;
- }
+ local = quota_local_new();
+ if (local == NULL) {
+ goto err;
+ }
- frame->local = local;
- loc_copy (&local->loc, loc);
+ frame->local = local;
+ loc_copy(&local->loc, loc);
- ret = dict_set_int8 (xattr_req, QUOTA_LIMIT_KEY, 1);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "dict set of key for "
- "hard-limit failed");
- goto err;
- }
+ ret = dict_set_int8(xattr_req, QUOTA_LIMIT_KEY, 1);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "dict set of key for "
+ "hard-limit failed");
+ goto err;
+ }
- ret = dict_set_int8 (xattr_req, QUOTA_LIMIT_OBJECTS_KEY, 1);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
- "dict set of key for quota object limit failed");
- goto err;
- }
+ ret = dict_set_int8(xattr_req, QUOTA_LIMIT_OBJECTS_KEY, 1);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "dict set of key for quota object limit failed");
+ goto err;
+ }
- STACK_WIND (frame, quota_lookup_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
+ STACK_WIND(frame, quota_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
- ret = 0;
+ ret = 0;
err:
- if (xattr_req)
- dict_unref (xattr_req);
+ if (xattr_req)
+ dict_unref(xattr_req);
- if (ret < 0) {
- QUOTA_STACK_UNWIND (lookup, frame, -1, ENOMEM,
- NULL, NULL, NULL, NULL);
- }
+ if (ret < 0) {
+ QUOTA_STACK_UNWIND(lookup, frame, -1, ENOMEM, NULL, NULL, NULL, NULL);
+ }
- return 0;
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->lookup,
+ loc, xattr_req);
+ return 0;
}
int32_t
-quota_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+quota_writev_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- int32_t ret = 0;
- uint64_t ctx_int = 0;
- quota_inode_ctx_t *ctx = NULL;
- quota_local_t *local = NULL;
-
- local = frame->local;
-
- if ((op_ret < 0) || (local == NULL) || (postbuf == NULL)) {
- goto out;
- }
-
- ret = inode_ctx_get (local->loc.inode, this, &ctx_int);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- Q_MSG_INODE_CTX_GET_FAILED, "%s: failed to get the "
- "context", local->loc.path);
- goto out;
- }
-
- ctx = (quota_inode_ctx_t *)(unsigned long) ctx_int;
-
- if (ctx == NULL) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- Q_MSG_INODE_CTX_GET_FAILED,
- "quota context not set in %s (gfid:%s)",
- local->loc.path, uuid_utoa (local->loc.inode->gfid));
- goto out;
- }
-
- LOCK (&ctx->lock);
- {
- ctx->buf = *postbuf;
- }
- UNLOCK (&ctx->lock);
+ int32_t ret = 0;
+ uint64_t ctx_int = 0;
+ quota_inode_ctx_t *ctx = NULL;
+ quota_local_t *local = NULL;
+
+ local = frame->local;
+
+ if ((op_ret < 0) || (local == NULL) || (postbuf == NULL)) {
+ goto out;
+ }
+
+ ret = inode_ctx_get(local->loc.inode, this, &ctx_int);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, Q_MSG_INODE_CTX_GET_FAILED,
+ "%s: failed to get the "
+ "context",
+ local->loc.path);
+ goto out;
+ }
+
+ ctx = (quota_inode_ctx_t *)(unsigned long)ctx_int;
+
+ if (ctx == NULL) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, Q_MSG_INODE_CTX_GET_FAILED,
+ "quota context not set in %s (gfid:%s)", local->loc.path,
+ uuid_utoa(local->loc.inode->gfid));
+ goto out;
+ }
+
+ LOCK(&ctx->lock);
+ {
+ ctx->buf = *postbuf;
+ }
+ UNLOCK(&ctx->lock);
out:
- QUOTA_STACK_UNWIND (writev, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
+ QUOTA_STACK_UNWIND(writev, frame, op_ret, op_errno, prebuf, postbuf, xdata);
- return 0;
+ return 0;
}
static int gf_quota_enforcer_log;
int32_t
-quota_writev_helper (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count, off_t off,
- uint32_t flags, struct iobref *iobref, dict_t *xdata)
+quota_writev_helper(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int32_t count, off_t off,
+ uint32_t flags, struct iobref *iobref, dict_t *xdata)
{
- quota_local_t *local = NULL;
- int32_t op_errno = EINVAL;
- struct iovec *new_vector = NULL;
- int32_t new_count = 0;
-
- local = frame->local;
-
- GF_VALIDATE_OR_GOTO ("quota", local, unwind);
-
- if (local->op_ret == -1) {
- op_errno = local->op_errno;
-
- if ((op_errno == EDQUOT) && (local->space_available > 0)) {
- new_count = iov_subset (vector, count, 0,
- local->space_available, NULL);
-
- new_vector = GF_CALLOC (new_count,
- sizeof (struct iovec),
- gf_common_mt_iovec);
- if (new_vector == NULL) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unwind;
- }
-
- new_count = iov_subset (vector, count, 0,
- local->space_available,
- new_vector);
-
- vector = new_vector;
- count = new_count;
- } else if (op_errno == ENOENT || op_errno == ESTALE) {
- /* We may get ENOENT/ESTALE in case of below scenario
- * fd = open file.txt
- * unlink file.txt
- * write on fd
- * Here build_ancestry can fail as the file is removed.
- * For now ignore ENOENT/ESTALE with writes on active fd
- * We need to re-visit this code once we understand
- * how other file-system behave in this scenario
- */
- gf_msg_debug (this->name, 0, "quota enforcer failed "
- "with ENOENT/ESTALE on %s, cannot check "
- "quota limits and allowing writes",
- uuid_utoa (fd->inode->gfid));
- } else if ((op_errno == EINVAL) &&
- !inode_parent (local->loc.inode, 0, NULL)) {
- /* We may get INVAL with parent == NULL,
- * in case of below scenario
- * 1. enable quota
- * 2. glusterfsd stop/start
- * 3. nameless lookup
- * 4. write on fd
- * Here build_ancestry can fail as the file's pgfid
- * is't exist.
- * For now ignore EINVAL with writes on active fd
- * untils the pgfid is created at name lookup
- */
- GF_LOG_OCCASIONALLY (gf_quota_enforcer_log, this->name,
- GF_LOG_CRITICAL, "Quota cannot be enforced as "
+ quota_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
+ struct iovec *new_vector = NULL;
+ int32_t new_count = 0;
+
+ local = frame->local;
+
+ GF_VALIDATE_OR_GOTO("quota", local, unwind);
+
+ if (local->op_ret == -1) {
+ op_errno = local->op_errno;
+
+ if ((op_errno == EDQUOT) && (local->space_available > 0)) {
+ new_count = iov_subset(vector, count, 0, local->space_available,
+ &new_vector, 0);
+ if (new_count < 0) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ vector = new_vector;
+ count = new_count;
+ } else if (op_errno == ENOENT || op_errno == ESTALE) {
+ /* We may get ENOENT/ESTALE in case of below scenario
+ * fd = open file.txt
+ * unlink file.txt
+ * write on fd
+ * Here build_ancestry can fail as the file is removed.
+ * For now ignore ENOENT/ESTALE with writes on active fd
+ * We need to re-visit this code once we understand
+ * how other file-system behave in this scenario
+ */
+ gf_msg_debug(this->name, 0,
+ "quota enforcer failed "
+ "with ENOENT/ESTALE on %s, cannot check "
+ "quota limits and allowing writes",
+ uuid_utoa(fd->inode->gfid));
+ } else if ((op_errno == EINVAL) &&
+ !inode_parent(local->loc.inode, 0, NULL)) {
+ /* We may get INVAL with parent == NULL,
+ * in case of below scenario
+ * 1. enable quota
+ * 2. glusterfsd stop/start
+ * 3. nameless lookup
+ * 4. write on fd
+ * Here build_ancestry can fail as the file's pgfid
+ * is't exist.
+ * For now ignore EINVAL with writes on active fd
+ * untils the pgfid is created at name lookup
+ */
+ GF_LOG_OCCASIONALLY(gf_quota_enforcer_log, this->name,
+ GF_LOG_CRITICAL,
+ "Quota cannot be enforced as "
"parent is not available and writes are being "
"allowed without checking whether they are "
"within quota limits. This can happen if Quota "
"crawl is not complete. If crawl has been "
"completed, please file a bug.");
- } else {
- goto unwind;
- }
+ } else {
+ goto unwind;
}
+ }
- STACK_WIND (frame, quota_writev_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->writev, fd,
- vector, count, off, flags, iobref, xdata);
+ STACK_WIND(frame, quota_writev_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector, count, off, flags,
+ iobref, xdata);
- if (new_vector != NULL)
- GF_FREE (new_vector);
+ if (new_vector != NULL)
+ GF_FREE(new_vector);
- return 0;
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(writev, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
-
int32_t
-quota_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count, off_t off,
- uint32_t flags, struct iobref *iobref, dict_t *xdata)
+quota_writev(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int32_t count, off_t off, uint32_t flags,
+ struct iobref *iobref, dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- int32_t op_errno = EINVAL;
- int32_t parents = 0;
- int32_t fail_count = 0;
- uint64_t size = 0;
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
- quota_dentry_t *dentry = NULL, *tmp = NULL;
- call_stub_t *stub = NULL;
- struct list_head head = {0, };
- inode_t *par_inode = NULL;
-
- priv = this->private;
-
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
-
- INIT_LIST_HEAD (&head);
-
- GF_ASSERT (frame);
- GF_VALIDATE_OR_GOTO ("quota", this, unwind);
- GF_VALIDATE_OR_GOTO (this->name, fd, unwind);
-
- local = quota_local_new ();
- if (local == NULL) {
- goto unwind;
- }
-
- frame->local = local;
- local->loc.inode = inode_ref (fd->inode);
-
- (void) quota_inode_ctx_get (fd->inode, this, &ctx, 0);
- if (ctx == NULL) {
- gf_msg_debug (this->name, 0, "quota context is NULL on inode"
- " (%s). If quota is not enabled recently and "
- "crawler has finished crawling, its an error",
- uuid_utoa (fd->inode->gfid));
- }
-
- stub = fop_writev_stub (frame, quota_writev_helper, fd, vector, count,
- off, flags, iobref, xdata);
- if (stub == NULL) {
- op_errno = ENOMEM;
- goto unwind;
- }
-
- priv = this->private;
- GF_VALIDATE_OR_GOTO (this->name, priv, unwind);
-
- size = iov_length (vector, count);
-
- parents = quota_add_parents_from_ctx (ctx, &head);
+ quota_priv_t *priv = NULL;
+ int32_t op_errno = EINVAL;
+ int32_t parents = 0;
+ int32_t fail_count = 0;
+ uint64_t size = 0;
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ quota_dentry_t *dentry = NULL, *tmp = NULL;
+ call_stub_t *stub = NULL;
+ struct list_head head;
+ inode_t *par_inode = NULL;
+
+ priv = this->private;
+
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
+
+ INIT_LIST_HEAD(&head);
+
+ GF_ASSERT(frame);
+ GF_VALIDATE_OR_GOTO("quota", this, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, fd, unwind);
+
+ local = quota_local_new();
+ if (local == NULL) {
+ goto unwind;
+ }
+
+ frame->local = local;
+ local->loc.inode = inode_ref(fd->inode);
+
+ (void)quota_inode_ctx_get(fd->inode, this, &ctx, 0);
+ if (ctx == NULL) {
+ gf_msg_debug(this->name, 0,
+ "quota context is NULL on inode"
+ " (%s). If quota is not enabled recently and "
+ "crawler has finished crawling, its an error",
+ uuid_utoa(fd->inode->gfid));
+ }
+
+ stub = fop_writev_stub(frame, quota_writev_helper, fd, vector, count, off,
+ flags, iobref, xdata);
+ if (stub == NULL) {
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ priv = this->private;
+ GF_VALIDATE_OR_GOTO(this->name, priv, unwind);
+
+ parents = quota_add_parents_from_ctx(ctx, &head);
+ if (parents == -1) {
+ op_errno = errno;
+ goto unwind;
+ }
+
+ size = iov_length(vector, count);
+
+ LOCK(&local->lock);
+ {
+ local->delta = size;
+ local->object_delta = 0;
+ local->link_count = (parents != 0) ? parents : 1;
+ local->stub = stub;
+ }
+ UNLOCK(&local->lock);
- LOCK (&local->lock);
+ if (parents == 0) {
+ /* nameless lookup on this inode, allow quota to reconstruct
+ * ancestry as part of check_limit.
+ */
+ quota_check_limit(frame, fd->inode, this);
+ } else {
+ list_for_each_entry_safe(dentry, tmp, &head, next)
{
- local->delta = size;
- local->object_delta = 0;
- local->link_count = (parents != 0) ? parents : 1;
- local->stub = stub;
+ par_inode = do_quota_check_limit(frame, fd->inode, this, dentry,
+ _gf_false);
+ if (par_inode == NULL) {
+ if (ctx) {
+ /* remove stale entry from inode ctx */
+ quota_dentry_del(ctx, dentry->name, dentry->par);
+ parents--;
+ fail_count++;
+ }
+ } else {
+ inode_unref(par_inode);
+ }
+ __quota_dentry_free(dentry);
}
- UNLOCK (&local->lock);
if (parents == 0) {
- /* nameless lookup on this inode, allow quota to reconstruct
- * ancestry as part of check_limit.
- */
- quota_check_limit (frame, fd->inode, this);
- } else {
- list_for_each_entry_safe (dentry, tmp, &head, next) {
- par_inode = do_quota_check_limit (frame, fd->inode,
- this, dentry,
- _gf_false);
- if (par_inode == NULL) {
- /* remove stale entry from inode ctx */
- quota_dentry_del (ctx, dentry->name,
- dentry->par);
- parents--;
- fail_count++;
- } else {
- inode_unref (par_inode);
- }
- __quota_dentry_free (dentry);
- }
-
- if (parents == 0) {
- LOCK (&local->lock);
- {
- local->link_count++;
- }
- UNLOCK (&local->lock);
- quota_check_limit (frame, fd->inode, this);
- }
+ LOCK(&local->lock);
+ {
+ local->link_count++;
+ }
+ UNLOCK(&local->lock);
+ quota_check_limit(frame, fd->inode, this);
+ }
- while (fail_count != 0) {
- quota_link_count_decrement (frame);
- fail_count--;
- }
+ while (fail_count != 0) {
+ quota_link_count_decrement(frame);
+ fail_count--;
}
+ }
- return 0;
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(writev, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->writev, fd,
- vector, count, off, flags, iobref, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->writev,
+ fd, vector, count, off, flags, iobref, xdata);
+ return 0;
}
-
int32_t
-quota_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+quota_mkdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- QUOTA_STACK_UNWIND (mkdir, frame, op_ret, op_errno, inode,
- buf, preparent, postparent, xdata);
- return 0;
+ QUOTA_STACK_UNWIND(mkdir, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
+ return 0;
}
-
int32_t
-quota_mkdir_helper (call_frame_t *frame, xlator_t *this, loc_t *loc,
- mode_t mode, mode_t umask, dict_t *xdata)
+quota_mkdir_helper(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- quota_local_t *local = NULL;
- int32_t op_errno = EINVAL;
+ quota_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
- local = frame->local;
+ local = frame->local;
- GF_VALIDATE_OR_GOTO ("quota", local, unwind);
+ GF_VALIDATE_OR_GOTO("quota", local, unwind);
- op_errno = local->op_errno;
+ op_errno = local->op_errno;
- if (local->op_ret == -1) {
- goto unwind;
- }
+ if (local->op_ret == -1) {
+ goto unwind;
+ }
- STACK_WIND (frame, quota_mkdir_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->mkdir, loc,
- mode, umask, xdata);
+ STACK_WIND(frame, quota_mkdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
- return 0;
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
int32_t
-quota_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- mode_t umask, dict_t *xdata)
+quota_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- int32_t ret = 0, op_errno = 0;
- quota_local_t *local = NULL;
- call_stub_t *stub = NULL;
-
- priv = this->private;
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
-
- if (!should_quota_enforce(this, xdata, GF_FOP_MKDIR)) {
- gf_msg (this->name, GF_LOG_DEBUG, 0,
- Q_MSG_ENFORCEMENT_SKIPPED,
- "Enforcement has been skipped(internal fop).");
- goto off;
- }
-
- local = quota_local_new ();
- if (local == NULL) {
- op_errno = ENOMEM;
- goto err;
- }
-
- frame->local = local;
-
- ret = loc_copy (&local->loc, loc);
- if (ret) {
- op_errno = ENOMEM;
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "loc_copy failed");
- goto err;
- }
-
- stub = fop_mkdir_stub (frame, quota_mkdir_helper, loc, mode, umask,
- xdata);
- if (stub == NULL) {
- op_errno = ENOMEM;
- goto err;
- }
-
- LOCK (&local->lock);
- {
- local->stub = stub;
- local->delta = 0;
- local->object_delta = 1;
- local->link_count = 1;
- }
- UNLOCK (&local->lock);
+ quota_priv_t *priv = NULL;
+ int32_t ret = 0, op_errno = 0;
+ quota_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+
+ priv = this->private;
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
+
+ if (!should_quota_enforce(this, xdata, GF_FOP_MKDIR)) {
+ gf_msg(this->name, GF_LOG_DEBUG, 0, Q_MSG_ENFORCEMENT_SKIPPED,
+ "Enforcement has been skipped(internal fop).");
+ goto off;
+ }
+
+ local = quota_local_new();
+ if (local == NULL) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ frame->local = local;
+
+ ret = loc_copy(&local->loc, loc);
+ if (ret) {
+ op_errno = ENOMEM;
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "loc_copy failed");
+ goto err;
+ }
+
+ stub = fop_mkdir_stub(frame, quota_mkdir_helper, loc, mode, umask, xdata);
+ if (stub == NULL) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ LOCK(&local->lock);
+ {
+ local->stub = stub;
+ local->delta = 0;
+ local->object_delta = 1;
+ local->link_count = 1;
+ }
+ UNLOCK(&local->lock);
- quota_check_limit (frame, loc->parent, this);
- return 0;
+ quota_check_limit(frame, loc->parent, this);
+ return 0;
err:
- QUOTA_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL,
- NULL, NULL);
+ QUOTA_STACK_UNWIND(mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
- return 0;
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mkdir,
- loc, mode, umask, xdata);
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->mkdir,
+ loc, mode, umask, xdata);
- return 0;
+ return 0;
}
-
int32_t
-quota_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+quota_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- int32_t ret = -1;
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
- quota_dentry_t *dentry = NULL;
-
- local = frame->local;
- if (op_ret < 0) {
- goto unwind;
- }
-
- ret = quota_inode_ctx_get (inode, this, &ctx, 1);
- if ((ret == -1) || (ctx == NULL)) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_INODE_CTX_GET_FAILED, "cannot create quota "
- "context in inode(gfid:%s)", uuid_utoa (inode->gfid));
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
-
- LOCK (&ctx->lock);
- {
- ctx->buf = *buf;
-
- dentry = __quota_dentry_new (ctx, (char *)local->loc.name,
- local->loc.parent->gfid);
- if (dentry == NULL) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "cannot create a new dentry "
- "(name:%s) for inode(gfid:%s)", local->loc.name,
- uuid_utoa (local->loc.inode->gfid));
- op_ret = -1;
- op_errno = ENOMEM;
- goto unlock;
- }
- }
+ int32_t ret = -1;
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ quota_dentry_t *dentry = NULL;
+
+ local = frame->local;
+ if (op_ret < 0) {
+ goto unwind;
+ }
+
+ ret = quota_inode_ctx_get(inode, this, &ctx, 1);
+ if ((ret == -1) || (ctx == NULL)) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_INODE_CTX_GET_FAILED,
+ "cannot create quota "
+ "context in inode(gfid:%s)",
+ uuid_utoa(inode->gfid));
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ LOCK(&ctx->lock);
+ {
+ ctx->buf = *buf;
+
+ dentry = __quota_dentry_new(ctx, (char *)local->loc.name,
+ local->loc.parent->gfid);
+ if (dentry == NULL) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "cannot create a new dentry "
+ "(name:%s) for inode(gfid:%s)",
+ local->loc.name, uuid_utoa(local->loc.inode->gfid));
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unlock;
+ }
+ }
unlock:
- UNLOCK (&ctx->lock);
+ UNLOCK(&ctx->lock);
unwind:
- QUOTA_STACK_UNWIND (create, frame, op_ret, op_errno, fd, inode, buf,
- preparent, postparent, xdata);
- return 0;
+ QUOTA_STACK_UNWIND(create, frame, op_ret, op_errno, fd, inode, buf,
+ preparent, postparent, xdata);
+ return 0;
}
-
int32_t
-quota_create_helper (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int32_t flags, mode_t mode, mode_t umask, fd_t *fd,
- dict_t *xdata)
+quota_create_helper(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ int32_t flags, mode_t mode, mode_t umask, fd_t *fd,
+ dict_t *xdata)
{
- quota_local_t *local = NULL;
- int32_t op_errno = EINVAL;
+ quota_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
- local = frame->local;
+ local = frame->local;
- GF_VALIDATE_OR_GOTO ("quota", local, unwind);
-
- if (local->op_ret == -1) {
- op_errno = local->op_errno;
- goto unwind;
- }
+ GF_VALIDATE_OR_GOTO("quota", local, unwind);
+ if (local->op_ret == -1) {
+ op_errno = local->op_errno;
+ goto unwind;
+ }
- STACK_WIND (frame, quota_create_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->create, loc,
- flags, mode, umask, fd, xdata);
- return 0;
+ STACK_WIND(frame, quota_create_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, flags, mode, umask, fd,
+ xdata);
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(create, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL, NULL);
+ return 0;
}
-
int32_t
-quota_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+quota_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- int32_t ret = -1;
- quota_local_t *local = NULL;
- int32_t op_errno = 0;
- call_stub_t *stub = NULL;
-
- priv = this->private;
-
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
- QUOTA_WIND_FOR_INTERNAL_FOP (xdata, off);
-
- local = quota_local_new ();
- if (local == NULL) {
- op_errno = ENOMEM;
- goto err;
- }
-
- frame->local = local;
-
- ret = loc_copy (&local->loc, loc);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "loc_copy failed");
- op_errno = ENOMEM;
- goto err;
- }
-
- stub = fop_create_stub (frame, quota_create_helper, loc, flags, mode,
- umask, fd, xdata);
- if (stub == NULL) {
- goto err;
- }
-
- LOCK (&local->lock);
- {
- local->link_count = 1;
- local->stub = stub;
- local->delta = 0;
- local->object_delta = 1;
- }
- UNLOCK (&local->lock);
+ quota_priv_t *priv = NULL;
+ int32_t ret = -1;
+ quota_local_t *local = NULL;
+ int32_t op_errno = 0;
+ call_stub_t *stub = NULL;
+
+ priv = this->private;
+
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
+ QUOTA_WIND_FOR_INTERNAL_FOP(xdata, off);
+
+ local = quota_local_new();
+ if (local == NULL) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ frame->local = local;
+
+ ret = loc_copy(&local->loc, loc);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "loc_copy failed");
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ stub = fop_create_stub(frame, quota_create_helper, loc, flags, mode, umask,
+ fd, xdata);
+ if (stub == NULL) {
+ goto err;
+ }
+
+ LOCK(&local->lock);
+ {
+ local->link_count = 1;
+ local->stub = stub;
+ local->delta = 0;
+ local->object_delta = 1;
+ }
+ UNLOCK(&local->lock);
- quota_check_limit (frame, loc->parent, this);
- return 0;
+ quota_check_limit(frame, loc->parent, this);
+ return 0;
err:
- QUOTA_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL,
- NULL, NULL, NULL);
+ QUOTA_STACK_UNWIND(create, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL, NULL);
- return 0;
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->create, loc,
- flags, mode, umask, fd, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->create,
+ loc, flags, mode, umask, fd, xdata);
+ return 0;
}
-
int32_t
-quota_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+quota_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
- uint64_t value = 0;
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ uint64_t value = 0;
- if (op_ret < 0) {
- goto out;
- }
+ if (op_ret < 0) {
+ goto out;
+ }
- local = (quota_local_t *) frame->local;
+ local = (quota_local_t *)frame->local;
- inode_ctx_get (local->loc.inode, this, &value);
- ctx = (quota_inode_ctx_t *)(unsigned long)value;
+ inode_ctx_get(local->loc.inode, this, &value);
+ ctx = (quota_inode_ctx_t *)(unsigned long)value;
- if (ctx == NULL) {
- gf_msg (this->name, GF_LOG_INFO, EINVAL,
- Q_MSG_INODE_CTX_GET_FAILED,
- "quota context not set inode (gfid:%s)",
- uuid_utoa (local->loc.gfid));
- goto out;
- }
+ if (ctx == NULL) {
+ gf_msg(this->name, GF_LOG_INFO, EINVAL, Q_MSG_INODE_CTX_GET_FAILED,
+ "quota context not set inode (gfid:%s)",
+ uuid_utoa(local->loc.gfid));
+ goto out;
+ }
- quota_dentry_del (ctx, local->loc.name, local->loc.parent->gfid);
+ quota_dentry_del(ctx, local->loc.name, local->loc.parent->gfid);
out:
- QUOTA_STACK_UNWIND (unlink, frame, op_ret, op_errno, preparent,
- postparent, xdata);
- return 0;
+ QUOTA_STACK_UNWIND(unlink, frame, op_ret, op_errno, preparent, postparent,
+ xdata);
+ return 0;
}
-
int32_t
-quota_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
- dict_t *xdata)
+quota_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- int32_t ret = -1;
- quota_local_t *local = NULL;
+ quota_priv_t *priv = NULL;
+ int32_t ret = -1;
+ quota_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
- local = quota_local_new ();
- if (local == NULL) {
- goto err;
- }
+ local = quota_local_new();
+ if (local == NULL) {
+ goto err;
+ }
- frame->local = local;
+ frame->local = local;
- ret = loc_copy (&local->loc, loc);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "loc_copy failed");
- goto err;
- }
+ ret = loc_copy(&local->loc, loc);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "loc_copy failed");
+ goto err;
+ }
- STACK_WIND (frame, quota_unlink_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
+ STACK_WIND(frame, quota_unlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
- ret = 0;
+ ret = 0;
err:
- if (ret == -1) {
- QUOTA_STACK_UNWIND (unlink, frame, -1, 0, NULL, NULL, NULL);
- }
+ if (ret == -1) {
+ QUOTA_STACK_UNWIND(unlink, frame, -1, 0, NULL, NULL, NULL);
+ }
- return 0;
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->unlink,
+ loc, xflag, xdata);
+ return 0;
}
-
int32_t
-quota_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+quota_link_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- int32_t ret = -1;
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
- quota_dentry_t *dentry = NULL;
- char found = 0;
-
- if (op_ret < 0) {
- goto out;
- }
-
- local = (quota_local_t *) frame->local;
-
- ret = quota_inode_ctx_get (inode, this, &ctx, 0);
- if ((ret == -1) || (ctx == NULL)) {
- gf_msg_debug (this->name, 0, "quota context is NULL on inode"
- " (%s). If quota is not enabled recently and "
- "crawler has finished crawling, its an error",
- uuid_utoa (inode->gfid));
- goto out;
- }
-
- LOCK (&ctx->lock);
+ int32_t ret = -1;
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ quota_dentry_t *dentry = NULL;
+ char found = 0;
+
+ if (op_ret < 0) {
+ goto out;
+ }
+
+ local = (quota_local_t *)frame->local;
+
+ ret = quota_inode_ctx_get(inode, this, &ctx, 0);
+ if ((ret == -1) || (ctx == NULL)) {
+ gf_msg_debug(this->name, 0,
+ "quota context is NULL on inode"
+ " (%s). If quota is not enabled recently and "
+ "crawler has finished crawling, its an error",
+ uuid_utoa(inode->gfid));
+ goto out;
+ }
+
+ LOCK(&ctx->lock);
+ {
+ list_for_each_entry(dentry, &ctx->parents, next)
{
- list_for_each_entry (dentry, &ctx->parents, next) {
- if ((strcmp (dentry->name, local->loc.name) == 0) &&
- (gf_uuid_compare (local->loc.parent->gfid,
- dentry->par) == 0)) {
- found = 1;
-
- gf_msg_debug (this->name, 0, "new entry being"
- " linked (name:%s) for inode "
- "(gfid:%s) is already present "
- "in inode-dentry-list",
- dentry->name,
- uuid_utoa (local->loc.inode->gfid));
- break;
- }
- }
-
- if (!found) {
- dentry = __quota_dentry_new (ctx,
- (char *)local->loc.name,
- local->loc.parent->gfid);
- if (dentry == NULL) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM,
- "cannot create a new dentry (name:%s)"
- "for inode(gfid:%s)", local->loc.name,
- uuid_utoa (local->loc.inode->gfid));
- op_ret = -1;
- op_errno = ENOMEM;
- goto unlock;
- }
- }
-
- ctx->buf = *buf;
+ if ((strcmp(dentry->name, local->loc.name) == 0) &&
+ (gf_uuid_compare(local->loc.parent->gfid, dentry->par) == 0)) {
+ found = 1;
+
+ gf_msg_debug(this->name, 0,
+ "new entry being"
+ " linked (name:%s) for inode "
+ "(gfid:%s) is already present "
+ "in inode-dentry-list",
+ dentry->name, uuid_utoa(local->loc.inode->gfid));
+ break;
+ }
+ }
+
+ if (!found) {
+ dentry = __quota_dentry_new(ctx, (char *)local->loc.name,
+ local->loc.parent->gfid);
+ if (dentry == NULL) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "cannot create a new dentry (name:%s)"
+ "for inode(gfid:%s)",
+ local->loc.name, uuid_utoa(local->loc.inode->gfid));
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unlock;
+ }
}
+
+ ctx->buf = *buf;
+ }
unlock:
- UNLOCK (&ctx->lock);
+ UNLOCK(&ctx->lock);
out:
- QUOTA_STACK_UNWIND (link, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
+ QUOTA_STACK_UNWIND(link, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
- return 0;
+ return 0;
}
-
int32_t
-quota_link_helper (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
- loc_t *newloc, dict_t *xdata)
+quota_link_helper(call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata)
{
- quota_local_t *local = NULL;
- int32_t op_errno = EINVAL;
+ quota_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
- local = frame->local;
+ local = frame->local;
- GF_VALIDATE_OR_GOTO ("quota", local, unwind);
+ GF_VALIDATE_OR_GOTO("quota", local, unwind);
- op_errno = local->op_errno;
+ op_errno = local->op_errno;
- if (local->op_ret == -1) {
- goto unwind;
- }
+ if (local->op_ret == -1) {
+ goto unwind;
+ }
- STACK_WIND (frame, quota_link_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->link, oldloc,
- newloc, xdata);
- return 0;
+ STACK_WIND(frame, quota_link_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+ return 0;
}
void
-quota_link_continue (call_frame_t *frame)
+quota_link_continue(call_frame_t *frame)
{
- int32_t ret = -1;
- int32_t op_errno = EIO;
- quota_local_t *local = NULL;
- uuid_t common_ancestor = {0};
- xlator_t *this = NULL;
- quota_inode_ctx_t *ctx = NULL;
- inode_t *src_parent = NULL;
- inode_t *dst_parent = NULL;
-
- local = frame->local;
- this = THIS;
-
- if (local->op_ret < 0) {
- op_errno = local->op_errno;
- goto err;
- }
-
- if (local->xdata &&
- dict_get (local->xdata, GLUSTERFS_INTERNAL_FOP_KEY)) {
- /* Treat link as rename, crawl upwards only till common ancestor
- */
- ret = quota_find_common_ancestor (local->oldloc.inode,
- local->newloc.parent,
- &common_ancestor);
- if (ret < 0 || gf_uuid_is_null(common_ancestor)) {
- gf_msg (this->name, GF_LOG_ERROR, ESTALE,
- Q_MSG_ANCESTRY_BUILD_FAILED, "failed to get "
- "common_ancestor for %s and %s",
- local->oldloc.path, local->newloc.path);
- op_errno = ESTALE;
- goto err;
- }
- } else {
- /* Treat link as a new file.
- * TODO: Currently marker accounts twice for the links created
- * across directories.
- * This needs re-visit if marker accounts only once
- * for the links created across directories
- */
- if (local->oldloc.parent)
- src_parent = inode_ref (local->oldloc.parent);
- else
- src_parent = inode_parent (local->oldloc.inode, 0,
- NULL);
- dst_parent = local->newloc.parent;
-
- /* No need to check quota limit if src and dst parents are same
- */
- if (src_parent == dst_parent ||
- gf_uuid_compare (src_parent->gfid, dst_parent->gfid) == 0) {
- inode_unref (src_parent);
- goto wind;
- }
-
- inode_unref (src_parent);
- }
+ int32_t ret = -1;
+ int32_t op_errno = EIO;
+ quota_local_t *local = NULL;
+ uuid_t common_ancestor = {0};
+ xlator_t *this = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ inode_t *src_parent = NULL;
+ inode_t *dst_parent = NULL;
+
+ local = frame->local;
+ this = THIS;
+
+ if (local->op_ret < 0) {
+ op_errno = local->op_errno;
+ goto err;
+ }
- quota_inode_ctx_get (local->oldloc.inode, this, &ctx, 0);
- if (ctx == NULL) {
- gf_msg_debug (this->name, 0, "quota context is NULL on inode"
- " (%s). If quota is not enabled recently and "
- "crawler has finished crawling, its an error",
- uuid_utoa (local->oldloc.inode->gfid));
- }
+ if (local->xdata && dict_get(local->xdata, GLUSTERFS_INTERNAL_FOP_KEY)) {
+ /* Treat link as rename, crawl upwards only till common ancestor
+ */
+ ret = quota_find_common_ancestor(
+ local->oldloc.inode, local->newloc.parent, &common_ancestor);
+ if (ret < 0 || gf_uuid_is_null(common_ancestor)) {
+ gf_msg(this->name, GF_LOG_ERROR, ESTALE,
+ Q_MSG_ANCESTRY_BUILD_FAILED,
+ "failed to get "
+ "common_ancestor for %s and %s",
+ local->oldloc.path, local->newloc.path);
+ op_errno = ESTALE;
+ goto err;
+ }
+ } else {
+ /* Treat link as a new file.
+ * TODO: Currently marker accounts twice for the links created
+ * across directories.
+ * This needs re-visit if marker accounts only once
+ * for the links created across directories
+ */
+ if (local->oldloc.parent)
+ src_parent = inode_ref(local->oldloc.parent);
+ else
+ src_parent = inode_parent(local->oldloc.inode, 0, NULL);
+ dst_parent = local->newloc.parent;
- LOCK (&local->lock);
- {
- local->link_count = 1;
- local->delta = (ctx != NULL) ? ctx->buf.ia_blocks * 512 : 0;
- local->object_delta = 1;
- gf_uuid_copy (local->common_ancestor, common_ancestor);
- }
- UNLOCK (&local->lock);
+ /* No need to check quota limit if src and dst parents are same
+ */
+ if (src_parent == dst_parent ||
+ gf_uuid_compare(src_parent->gfid, dst_parent->gfid) == 0) {
+ inode_unref(src_parent);
+ goto wind;
+ }
+
+ inode_unref(src_parent);
+ }
+
+ quota_inode_ctx_get(local->oldloc.inode, this, &ctx, 0);
+ if (ctx == NULL) {
+ gf_msg_debug(this->name, 0,
+ "quota context is NULL on inode"
+ " (%s). If quota is not enabled recently and "
+ "crawler has finished crawling, its an error",
+ uuid_utoa(local->oldloc.inode->gfid));
+ }
+
+ LOCK(&local->lock);
+ {
+ local->link_count = 1;
+ local->delta = (ctx != NULL) ? ctx->buf.ia_blocks * 512 : 0;
+ local->object_delta = 1;
+ gf_uuid_copy(local->common_ancestor, common_ancestor);
+ }
+ UNLOCK(&local->lock);
- quota_check_limit (frame, local->newloc.parent, this);
- return;
+ quota_check_limit(frame, local->newloc.parent, this);
+ return;
err:
- QUOTA_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL);
- return;
+ QUOTA_STACK_UNWIND(link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+ return;
wind:
- STACK_WIND (frame, quota_link_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->link, &(local->oldloc),
- &(local->newloc), local->xdata);
- return;
+ STACK_WIND(frame, quota_link_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->link, &(local->oldloc),
+ &(local->newloc), local->xdata);
+ return;
}
int32_t
-quota_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
+quota_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- int32_t ret = -1;
- int32_t op_errno = ENOMEM;
- quota_local_t *local = NULL;
- call_stub_t *stub = NULL;
-
- priv = this->private;
-
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
-
- local = quota_local_new ();
- if (local == NULL) {
- goto err;
- }
-
- frame->local = (void *) local;
-
- if (xdata)
- local->xdata = dict_ref (xdata);
-
- ret = loc_copy (&local->loc, newloc);
- if (ret == -1) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "loc_copy failed");
- goto err;
- }
-
- ret = loc_copy (&local->oldloc, oldloc);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
- "loc_copy failed");
- goto err;
- }
-
- ret = loc_copy (&local->newloc, newloc);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
- "loc_copy failed");
- goto err;
- }
-
- /* No need to check quota limit if src and dst parents are same */
- if (oldloc->parent && newloc->parent &&
- !gf_uuid_compare(oldloc->parent->gfid, newloc->parent->gfid)) {
- gf_msg_debug (this->name, GF_LOG_DEBUG, "link %s -> %s are "
- "in the same directory, so skip check limit",
- oldloc->path, newloc->path);
- goto wind;
- }
-
- stub = fop_link_stub (frame, quota_link_helper, oldloc, newloc, xdata);
- if (stub == NULL) {
- goto err;
- }
-
- LOCK (&local->lock);
- {
- local->link_count = 2;
- local->fop_continue_cbk = quota_link_continue;
- local->stub = stub;
- }
- UNLOCK (&local->lock);
+ quota_priv_t *priv = NULL;
+ int32_t ret = -1;
+ int32_t op_errno = ENOMEM;
+ quota_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+
+ priv = this->private;
+
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
+
+ local = quota_local_new();
+ if (local == NULL) {
+ goto err;
+ }
+
+ frame->local = (void *)local;
+
+ if (xdata)
+ local->xdata = dict_ref(xdata);
+
+ ret = loc_copy(&local->loc, newloc);
+ if (ret == -1) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "loc_copy failed");
+ goto err;
+ }
+
+ ret = loc_copy(&local->oldloc, oldloc);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "loc_copy failed");
+ goto err;
+ }
+
+ ret = loc_copy(&local->newloc, newloc);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "loc_copy failed");
+ goto err;
+ }
+
+ /* No need to check quota limit if src and dst parents are same */
+ if (oldloc->parent && newloc->parent &&
+ !gf_uuid_compare(oldloc->parent->gfid, newloc->parent->gfid)) {
+ gf_msg_debug(this->name, GF_LOG_DEBUG,
+ "link %s -> %s are "
+ "in the same directory, so skip check limit",
+ oldloc->path, newloc->path);
+ goto wind;
+ }
+
+ stub = fop_link_stub(frame, quota_link_helper, oldloc, newloc, xdata);
+ if (stub == NULL) {
+ goto err;
+ }
+
+ LOCK(&local->lock);
+ {
+ local->link_count = 2;
+ local->fop_continue_cbk = quota_link_continue;
+ local->stub = stub;
+ }
+ UNLOCK(&local->lock);
- check_ancestory (frame, newloc->parent);
+ check_ancestory(frame, newloc->parent);
- /* source parent can be NULL, so do check_ancestry on a file */
- if (oldloc->parent)
- check_ancestory (frame, oldloc->parent);
- else
- check_ancestory (frame, oldloc->inode);
+ /* source parent can be NULL, so do check_ancestry on a file */
+ if (oldloc->parent)
+ check_ancestory(frame, oldloc->parent);
+ else
+ check_ancestory(frame, oldloc->inode);
- return 0;
+ return 0;
err:
- QUOTA_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->link, oldloc,
- newloc, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->link,
+ oldloc, newloc, xdata);
+ return 0;
wind:
- STACK_WIND (frame, quota_link_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->link, oldloc,
- newloc, xdata);
- return 0;
+ STACK_WIND(frame, quota_link_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
+ return 0;
}
-
int32_t
-quota_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent,
- dict_t *xdata)
+quota_rename_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ struct iatt *preoldparent, struct iatt *postoldparent,
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
{
- int32_t ret = -1;
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
- quota_dentry_t *old_dentry = NULL, *dentry = NULL;
- char new_dentry_found = 0;
+ int32_t ret = -1;
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ quota_dentry_t *old_dentry = NULL, *dentry = NULL;
+ char new_dentry_found = 0;
- if (op_ret < 0) {
- goto out;
- }
+ if (op_ret < 0) {
+ goto out;
+ }
- local = frame->local;
+ local = frame->local;
- GF_VALIDATE_OR_GOTO ("quota", local, out);
+ GF_VALIDATE_OR_GOTO("quota", local, out);
- if (!QUOTA_REG_OR_LNK_FILE (local->oldloc.inode->ia_type))
- goto out;
+ if (!QUOTA_REG_OR_LNK_FILE(local->oldloc.inode->ia_type))
+ goto out;
- ret = quota_inode_ctx_get (local->oldloc.inode, this, &ctx, 0);
- if ((ret == -1) || (ctx == NULL)) {
- gf_msg_debug (this->name, 0, "quota context is NULL on inode"
- " (%s). If quota is not enabled recently and "
- "crawler has finished crawling, its an error",
- uuid_utoa (local->oldloc.inode->gfid));
+ ret = quota_inode_ctx_get(local->oldloc.inode, this, &ctx, 0);
+ if ((ret == -1) || (ctx == NULL)) {
+ gf_msg_debug(this->name, 0,
+ "quota context is NULL on inode"
+ " (%s). If quota is not enabled recently and "
+ "crawler has finished crawling, its an error",
+ uuid_utoa(local->oldloc.inode->gfid));
- goto out;
- }
+ goto out;
+ }
- LOCK (&ctx->lock);
+ LOCK(&ctx->lock);
+ {
+ list_for_each_entry(dentry, &ctx->parents, next)
{
- list_for_each_entry (dentry, &ctx->parents, next) {
- if ((strcmp (dentry->name, local->oldloc.name) == 0) &&
- (gf_uuid_compare (local->oldloc.parent->gfid,
- dentry->par) == 0)) {
- old_dentry = dentry;
- } else if ((strcmp (dentry->name,
- local->newloc.name) == 0) &&
- (gf_uuid_compare (local->newloc.parent->gfid,
- dentry->par) == 0)) {
- new_dentry_found = 1;
- gf_msg_debug (this->name, 0, "new entry being "
- "linked (name:%s) for inode (gfid:%s) "
- "is in inode-dentry-list", dentry->name,
- uuid_utoa (local->oldloc.inode->gfid));
- }
-
- if (old_dentry && new_dentry_found)
- break;
- }
-
- if (old_dentry != NULL) {
- __quota_dentry_free (old_dentry);
- } else {
- gf_msg_debug (this->name, 0, "dentry corresponding"
- "the path just renamed (name:%s) is not"
- " present", local->oldloc.name);
- }
-
- if (!new_dentry_found) {
- dentry = __quota_dentry_new (ctx,
- (char *)local->newloc.name,
- local->newloc.parent->gfid);
- if (dentry == NULL) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM,
- "cannot create a new dentry (name:%s) "
- "for inode(gfid:%s)",
- local->newloc.name,
- uuid_utoa (local->newloc.inode->gfid));
- op_ret = -1;
- op_errno = ENOMEM;
- goto unlock;
- }
- }
-
- ctx->buf = *buf;
+ if ((strcmp(dentry->name, local->oldloc.name) == 0) &&
+ (gf_uuid_compare(local->oldloc.parent->gfid, dentry->par) ==
+ 0)) {
+ old_dentry = dentry;
+ } else if ((strcmp(dentry->name, local->newloc.name) == 0) &&
+ (gf_uuid_compare(local->newloc.parent->gfid,
+ dentry->par) == 0)) {
+ new_dentry_found = 1;
+ gf_msg_debug(this->name, 0,
+ "new entry being "
+ "linked (name:%s) for inode (gfid:%s) "
+ "is in inode-dentry-list",
+ dentry->name,
+ uuid_utoa(local->oldloc.inode->gfid));
+ }
+
+ if (old_dentry && new_dentry_found)
+ break;
+ }
+
+ if (old_dentry != NULL) {
+ __quota_dentry_free(old_dentry);
+ } else {
+ gf_msg_debug(this->name, 0,
+ "dentry corresponding"
+ "the path just renamed (name:%s) is not"
+ " present",
+ local->oldloc.name);
+ }
+
+ if (!new_dentry_found) {
+ dentry = __quota_dentry_new(ctx, (char *)local->newloc.name,
+ local->newloc.parent->gfid);
+ if (dentry == NULL) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "cannot create a new dentry (name:%s) "
+ "for inode(gfid:%s)",
+ local->newloc.name,
+ uuid_utoa(local->newloc.inode->gfid));
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unlock;
+ }
}
+
+ ctx->buf = *buf;
+ }
unlock:
- UNLOCK (&ctx->lock);
+ UNLOCK(&ctx->lock);
out:
- QUOTA_STACK_UNWIND (rename, frame, op_ret, op_errno, buf, preoldparent,
- postoldparent, prenewparent, postnewparent, xdata);
+ QUOTA_STACK_UNWIND(rename, frame, op_ret, op_errno, buf, preoldparent,
+ postoldparent, prenewparent, postnewparent, xdata);
- return 0;
+ return 0;
}
-
int32_t
-quota_rename_helper (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
- loc_t *newloc, dict_t *xdata)
+quota_rename_helper(call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata)
{
- quota_local_t *local = NULL;
- int32_t op_errno = EINVAL;
+ quota_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
- local = frame->local;
+ local = frame->local;
- GF_VALIDATE_OR_GOTO ("quota", local, unwind);
+ GF_VALIDATE_OR_GOTO("quota", local, unwind);
- op_errno = local->op_errno;
+ op_errno = local->op_errno;
- if (local->op_ret == -1) {
- goto unwind;
- }
+ if (local->op_ret == -1) {
+ goto unwind;
+ }
- STACK_WIND (frame, quota_rename_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->rename, oldloc,
- newloc, xdata);
+ STACK_WIND(frame, quota_rename_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
- return 0;
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL, NULL);
+ return 0;
}
-
static int32_t
-quota_rename_get_size_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata,
- struct iatt *postparent)
+quota_rename_get_size_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
{
- quota_local_t *local = NULL;
- int32_t ret = 0;
- int64_t *size = 0;
-
- GF_ASSERT (frame);
- GF_VALIDATE_OR_GOTO_WITH_ERROR ("quota", this, out, op_errno,
- EINVAL);
- GF_VALIDATE_OR_GOTO_WITH_ERROR (this->name, xdata, out, op_errno,
- EINVAL);
- local = frame->local;
- GF_ASSERT (local);
- local->link_count = 1;
-
- if (op_ret < 0)
- goto out;
-
-
- ret = dict_get_bin (xdata, QUOTA_SIZE_KEY, (void **) &size);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, EINVAL,
- Q_MSG_SIZE_KEY_MISSING, "size key not present in dict");
- op_errno = EINVAL;
- goto out;
- }
- local->delta = ntoh64 (*size);
- local->object_delta = 1;
- quota_check_limit (frame, local->newloc.parent, this);
- return 0;
+ quota_local_t *local = NULL;
+ int32_t ret = 0;
+ int64_t *size = 0;
+
+ GF_ASSERT(frame);
+ GF_VALIDATE_OR_GOTO_WITH_ERROR("quota", this, out, op_errno, EINVAL);
+ GF_VALIDATE_OR_GOTO_WITH_ERROR(this->name, xdata, out, op_errno, EINVAL);
+ local = frame->local;
+ GF_ASSERT(local);
+ local->link_count = 1;
+
+ if (op_ret < 0)
+ goto out;
+
+ ret = dict_get_bin(xdata, QUOTA_SIZE_KEY, (void **)&size);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, EINVAL, Q_MSG_SIZE_KEY_MISSING,
+ "size key not present in dict");
+ op_errno = EINVAL;
+ goto out;
+ }
+ local->delta = ntoh64(*size);
+ local->object_delta = 1;
+ quota_check_limit(frame, local->newloc.parent, this);
+ return 0;
out:
- quota_handle_validate_error (frame, -1, op_errno);
- return 0;
+ quota_handle_validate_error(frame, -1, op_errno);
+ return 0;
}
void
-quota_rename_continue (call_frame_t *frame)
+quota_rename_continue(call_frame_t *frame)
{
- int32_t ret = -1;
- int32_t op_errno = EIO;
- quota_local_t *local = NULL;
- uuid_t common_ancestor = {0};
- xlator_t *this = NULL;
- quota_inode_ctx_t *ctx = NULL;
-
- local = frame->local;
- this = THIS;
-
- if (local->op_ret < 0) {
- op_errno = local->op_errno;
- goto err;
- }
-
- ret = quota_find_common_ancestor (local->oldloc.parent,
- local->newloc.parent,
- &common_ancestor);
- if (ret < 0 || gf_uuid_is_null(common_ancestor)) {
- gf_msg (this->name, GF_LOG_ERROR, ESTALE,
- Q_MSG_ANCESTRY_BUILD_FAILED, "failed to get "
- "common_ancestor for %s and %s",
- local->oldloc.path, local->newloc.path);
- op_errno = ESTALE;
- goto err;
- }
-
- LOCK (&local->lock);
- {
- local->link_count = 1;
- gf_uuid_copy (local->common_ancestor, common_ancestor);
- }
- UNLOCK (&local->lock);
+ int32_t ret = -1;
+ int32_t op_errno = EIO;
+ quota_local_t *local = NULL;
+ uuid_t common_ancestor = {0};
+ xlator_t *this = NULL;
+ quota_inode_ctx_t *ctx = NULL;
- if (QUOTA_REG_OR_LNK_FILE (local->oldloc.inode->ia_type)) {
- ret = quota_inode_ctx_get (local->oldloc.inode, this, &ctx, 0);
- if (ctx == NULL) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- Q_MSG_INODE_CTX_GET_FAILED,
- "quota context not set in inode (gfid:%s), "
- "considering file size as zero while enforcing "
- "quota on new ancestry",
- uuid_utoa (local->oldloc.inode->gfid));
+ local = frame->local;
+ this = THIS;
+ if (local->op_ret < 0) {
+ op_errno = local->op_errno;
+ goto err;
+ }
+
+ ret = quota_find_common_ancestor(local->oldloc.parent, local->newloc.parent,
+ &common_ancestor);
+ if (ret < 0 || gf_uuid_is_null(common_ancestor)) {
+ gf_msg(this->name, GF_LOG_ERROR, ESTALE, Q_MSG_ANCESTRY_BUILD_FAILED,
+ "failed to get "
+ "common_ancestor for %s and %s",
+ local->oldloc.path, local->newloc.path);
+ op_errno = ESTALE;
+ goto err;
+ }
+
+ LOCK(&local->lock);
+ {
+ local->link_count = 1;
+ gf_uuid_copy(local->common_ancestor, common_ancestor);
+ }
+ UNLOCK(&local->lock);
- local->delta = 0;
- local->object_delta = 1;
- } else {
-
- /* FIXME: We need to account for the size occupied by
- * this inode on the target directory. To avoid double
- * accounting, we need to modify enforcer to perform
- * quota_check_limit only up till the least common
- * ancestor directory inode*/
-
- /* FIXME: The following code assumes that regular files
- * and link files are present, in their entirety, in a
- * single brick. This *assumption is invalid in the
- * case of stripe.*/
+ if (QUOTA_REG_OR_LNK_FILE(local->oldloc.inode->ia_type)) {
+ ret = quota_inode_ctx_get(local->oldloc.inode, this, &ctx, 0);
+ if (ctx == NULL) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, Q_MSG_INODE_CTX_GET_FAILED,
+ "quota context not set in inode (gfid:%s), "
+ "considering file size as zero while enforcing "
+ "quota on new ancestry",
+ uuid_utoa(local->oldloc.inode->gfid));
+
+ local->delta = 0;
+ local->object_delta = 1;
+ } else {
+ /* FIXME: We need to account for the size occupied by
+ * this inode on the target directory. To avoid double
+ * accounting, we need to modify enforcer to perform
+ * quota_check_limit only up till the least common
+ * ancestor directory inode*/
- local->delta = ctx->buf.ia_blocks * 512;
- local->object_delta = 1;
- }
+ /* FIXME: The following code assumes that regular files
+ * and link files are present, in their entirety, in a
+ * single brick. This *assumption is invalid in the
+ * case of stripe.*/
- } else if (IA_ISDIR (local->oldloc.inode->ia_type)) {
- ret = quota_validate (frame, local->oldloc.inode, this,
- quota_rename_get_size_cbk);
- if (ret){
- op_errno = -ret;
- goto err;
- }
+ local->delta = ctx->buf.ia_blocks * 512;
+ local->object_delta = 1;
+ }
- return;
+ } else if (IA_ISDIR(local->oldloc.inode->ia_type)) {
+ ret = quota_validate(frame, local->oldloc.inode, this,
+ quota_rename_get_size_cbk);
+ if (ret) {
+ op_errno = -ret;
+ goto err;
}
- quota_check_limit (frame, local->newloc.parent, this);
return;
+ }
-err:
- QUOTA_STACK_UNWIND (rename, frame, -1, op_errno, NULL,
- NULL, NULL, NULL, NULL, NULL);
- return;
+ quota_check_limit(frame, local->newloc.parent, this);
+ return;
+err:
+ QUOTA_STACK_UNWIND(rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL, NULL);
+ return;
}
int32_t
-quota_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
- loc_t *newloc, dict_t *xdata)
+quota_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- int32_t ret = -1;
- int32_t op_errno = ENOMEM;
- quota_local_t *local = NULL;
- call_stub_t *stub = NULL;
-
- priv = this->private;
-
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
-
- local = quota_local_new ();
- if (local == NULL) {
- goto err;
- }
-
- frame->local = local;
-
- ret = loc_copy (&local->oldloc, oldloc);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
- "loc_copy failed");
- goto err;
- }
-
- ret = loc_copy (&local->newloc, newloc);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
- "loc_copy failed");
- goto err;
- }
-
- /* No need to check quota limit if src and dst parents are same */
- if (oldloc->parent && newloc->parent &&
- !gf_uuid_compare(oldloc->parent->gfid, newloc->parent->gfid)) {
- gf_msg_debug (this->name, 0, "rename %s -> %s are "
- "in the same directory, so skip check limit",
- oldloc->path, newloc->path);
- goto wind;
- }
-
- stub = fop_rename_stub (frame, quota_rename_helper, oldloc, newloc,
- xdata);
- if (stub == NULL) {
- goto err;
- }
-
- LOCK (&local->lock);
- {
- /* link_count here tell how many check_ancestry should be done
- * before continuing the FOP
- */
- local->link_count = 2;
- local->stub = stub;
- local->fop_continue_cbk = quota_rename_continue;
- }
- UNLOCK (&local->lock);
+ quota_priv_t *priv = NULL;
+ int32_t ret = -1;
+ int32_t op_errno = ENOMEM;
+ quota_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+
+ priv = this->private;
+
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
+
+ local = quota_local_new();
+ if (local == NULL) {
+ goto err;
+ }
+
+ frame->local = local;
+
+ ret = loc_copy(&local->oldloc, oldloc);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "loc_copy failed");
+ goto err;
+ }
+
+ ret = loc_copy(&local->newloc, newloc);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "loc_copy failed");
+ goto err;
+ }
+
+ /* No need to check quota limit if src and dst parents are same */
+ if (oldloc->parent && newloc->parent &&
+ !gf_uuid_compare(oldloc->parent->gfid, newloc->parent->gfid)) {
+ gf_msg_debug(this->name, 0,
+ "rename %s -> %s are "
+ "in the same directory, so skip check limit",
+ oldloc->path, newloc->path);
+ goto wind;
+ }
+
+ stub = fop_rename_stub(frame, quota_rename_helper, oldloc, newloc, xdata);
+ if (stub == NULL) {
+ goto err;
+ }
+
+ LOCK(&local->lock);
+ {
+ /* link_count here tell how many check_ancestry should be done
+ * before continuing the FOP
+ */
+ local->link_count = 2;
+ local->stub = stub;
+ local->fop_continue_cbk = quota_rename_continue;
+ }
+ UNLOCK(&local->lock);
- check_ancestory (frame, newloc->parent);
- check_ancestory (frame, oldloc->parent);
- return 0;
+ check_ancestory(frame, newloc->parent);
+ check_ancestory(frame, oldloc->parent);
+ return 0;
err:
- QUOTA_STACK_UNWIND (rename, frame, -1, op_errno, NULL,
- NULL, NULL, NULL, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL, NULL);
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rename, oldloc,
- newloc, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->rename,
+ oldloc, newloc, xdata);
+ return 0;
wind:
- STACK_WIND (frame, quota_rename_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->rename, oldloc,
- newloc, xdata);
- return 0;
+ STACK_WIND(frame, quota_rename_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
+ return 0;
}
-
int32_t
-quota_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+quota_symlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
- quota_dentry_t *dentry = NULL;
-
- if (op_ret < 0) {
- goto out;
- }
-
- local = frame->local;
-
- quota_inode_ctx_get (local->loc.inode, this, &ctx, 1);
- if (ctx == NULL) {
- gf_msg_debug (this->name, 0, "quota context is NULL on inode"
- " (%s). If quota is not enabled recently and "
- "crawler has finished crawling, its an error",
- uuid_utoa (local->loc.inode->gfid));
-
- goto out;
- }
-
- LOCK (&ctx->lock);
- {
- ctx->buf = *buf;
-
- dentry = __quota_dentry_new (ctx, (char *)local->loc.name,
- local->loc.parent->gfid);
- if (dentry == NULL) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "cannot create "
- "a new dentry (name:%s) for inode(gfid:%s)",
- local->loc.name,
- uuid_utoa (local->loc.inode->gfid));
- op_ret = -1;
- op_errno = ENOMEM;
- }
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ quota_dentry_t *dentry = NULL;
+ int32_t ret = -1;
+
+ if (op_ret < 0) {
+ goto out;
+ }
+
+ local = frame->local;
+
+ ret = quota_inode_ctx_get(local->loc.inode, this, &ctx, 1);
+ if ((ret == -1) || (ctx == NULL)) {
+ gf_msg_debug(this->name, 0,
+ "quota context is NULL on inode"
+ " (%s). If quota is not enabled recently and "
+ "crawler has finished crawling, its an error",
+ uuid_utoa(local->loc.inode->gfid));
+
+ goto out;
+ }
+
+ LOCK(&ctx->lock);
+ {
+ ctx->buf = *buf;
+
+ dentry = __quota_dentry_new(ctx, (char *)local->loc.name,
+ local->loc.parent->gfid);
+ if (dentry == NULL) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "cannot create "
+ "a new dentry (name:%s) for inode(gfid:%s)",
+ local->loc.name, uuid_utoa(local->loc.inode->gfid));
+ op_ret = -1;
+ op_errno = ENOMEM;
}
- UNLOCK (&ctx->lock);
+ }
+ UNLOCK(&ctx->lock);
out:
- QUOTA_STACK_UNWIND (symlink, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
+ QUOTA_STACK_UNWIND(symlink, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
- return 0;
+ return 0;
}
-
int
-quota_symlink_helper (call_frame_t *frame, xlator_t *this, const char *linkpath,
- loc_t *loc, mode_t umask, dict_t *xdata)
+quota_symlink_helper(call_frame_t *frame, xlator_t *this, const char *linkpath,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
- quota_local_t *local = NULL;
- int32_t op_errno = EINVAL;
+ quota_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
- local = frame->local;
+ local = frame->local;
- GF_VALIDATE_OR_GOTO ("quota", local, unwind);
+ GF_VALIDATE_OR_GOTO("quota", local, unwind);
- if (local->op_ret == -1) {
- op_errno = local->op_errno;
- goto unwind;
- }
+ if (local->op_ret == -1) {
+ op_errno = local->op_errno;
+ goto unwind;
+ }
- STACK_WIND (frame, quota_symlink_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->symlink,
- linkpath, loc, umask, xdata);
- return 0;
+ STACK_WIND(frame, quota_symlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->symlink, linkpath, loc, umask, xdata);
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (symlink, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(symlink, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
-
int
-quota_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
- loc_t *loc, mode_t umask, dict_t *xdata)
+quota_symlink(call_frame_t *frame, xlator_t *this, const char *linkpath,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- int32_t ret = -1;
- int32_t op_errno = ENOMEM;
- quota_local_t *local = NULL;
- call_stub_t *stub = NULL;
+ quota_priv_t *priv = NULL;
+ int32_t ret = -1;
+ int32_t op_errno = ENOMEM;
+ quota_local_t *local = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
+ priv = this->private;
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
- local = quota_local_new ();
- if (local == NULL) {
- goto err;
- }
+ local = quota_local_new();
+ if (local == NULL) {
+ goto err;
+ }
- frame->local = local;
+ frame->local = local;
- ret = loc_copy (&local->loc, loc);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "loc_copy failed");
- goto err;
- }
+ ret = loc_copy(&local->loc, loc);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "loc_copy failed");
+ goto err;
+ }
- stub = fop_symlink_stub (frame, quota_symlink_helper, linkpath, loc,
- umask, xdata);
- if (stub == NULL) {
- goto err;
- }
+ stub = fop_symlink_stub(frame, quota_symlink_helper, linkpath, loc, umask,
+ xdata);
+ if (stub == NULL) {
+ goto err;
+ }
- LOCK (&local->lock);
- {
- local->stub = stub;
- local->delta = strlen (linkpath);
- local->object_delta = 1;
- local->link_count = 1;
- }
- UNLOCK (&local->lock);
+ LOCK(&local->lock);
+ {
+ local->stub = stub;
+ local->delta = strlen(linkpath);
+ local->object_delta = 1;
+ local->link_count = 1;
+ }
+ UNLOCK(&local->lock);
- quota_check_limit (frame, loc->parent, this);
- return 0;
+ quota_check_limit(frame, loc->parent, this);
+ return 0;
err:
- QUOTA_STACK_UNWIND (symlink, frame, -1, op_errno, NULL, NULL, NULL,
- NULL, NULL);
+ QUOTA_STACK_UNWIND(symlink, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
- return 0;
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->symlink,
- linkpath, loc, umask, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->symlink,
+ linkpath, loc, umask, xdata);
+ return 0;
}
-
int32_t
-quota_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+quota_truncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
- if (op_ret < 0) {
- goto out;
- }
+ if (op_ret < 0) {
+ goto out;
+ }
- local = frame->local;
+ local = frame->local;
- GF_VALIDATE_OR_GOTO ("quota", local, out);
+ GF_VALIDATE_OR_GOTO("quota", local, out);
- quota_inode_ctx_get (local->loc.inode, this, &ctx, 0);
- if (ctx == NULL) {
- gf_msg_debug (this->name, 0, "quota context is NULL on inode"
- " (%s). If quota is not enabled recently and "
- "crawler has finished crawling, its an error",
- uuid_utoa (local->loc.inode->gfid));
- goto out;
- }
+ quota_inode_ctx_get(local->loc.inode, this, &ctx, 0);
+ if (ctx == NULL) {
+ gf_msg_debug(this->name, 0,
+ "quota context is NULL on inode"
+ " (%s). If quota is not enabled recently and "
+ "crawler has finished crawling, its an error",
+ uuid_utoa(local->loc.inode->gfid));
+ goto out;
+ }
- LOCK (&ctx->lock);
- {
- ctx->buf = *postbuf;
- }
- UNLOCK (&ctx->lock);
+ LOCK(&ctx->lock);
+ {
+ ctx->buf = *postbuf;
+ }
+ UNLOCK(&ctx->lock);
out:
- QUOTA_STACK_UNWIND (truncate, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
- return 0;
+ QUOTA_STACK_UNWIND(truncate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
-
int32_t
-quota_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
- dict_t *xdata)
+quota_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- int32_t ret = -1;
- quota_local_t *local = NULL;
+ quota_priv_t *priv = NULL;
+ int32_t ret = -1;
+ quota_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
- local = quota_local_new ();
- if (local == NULL) {
- goto err;
- }
+ local = quota_local_new();
+ if (local == NULL) {
+ goto err;
+ }
- frame->local = local;
+ frame->local = local;
- ret = loc_copy (&local->loc, loc);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "loc_copy failed");
- goto err;
- }
+ ret = loc_copy(&local->loc, loc);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "loc_copy failed");
+ goto err;
+ }
- STACK_WIND (frame, quota_truncate_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
+ STACK_WIND(frame, quota_truncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
- return 0;
+ return 0;
err:
- QUOTA_STACK_UNWIND (truncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+ QUOTA_STACK_UNWIND(truncate, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->truncate,
+ loc, offset, xdata);
+ return 0;
}
-
int32_t
-quota_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+quota_ftruncate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
- if (op_ret < 0) {
- goto out;
- }
+ if (op_ret < 0) {
+ goto out;
+ }
- local = frame->local;
+ local = frame->local;
- GF_VALIDATE_OR_GOTO ("quota", local, out);
+ GF_VALIDATE_OR_GOTO("quota", local, out);
- quota_inode_ctx_get (local->loc.inode, this, &ctx, 0);
- if (ctx == NULL) {
- gf_msg_debug (this->name, 0, "quota context is NULL on inode"
- " (%s). If quota is not enabled recently and "
- "crawler has finished crawling, its an error",
- uuid_utoa (local->loc.inode->gfid));
- goto out;
- }
+ quota_inode_ctx_get(local->loc.inode, this, &ctx, 0);
+ if (ctx == NULL) {
+ gf_msg_debug(this->name, 0,
+ "quota context is NULL on inode"
+ " (%s). If quota is not enabled recently and "
+ "crawler has finished crawling, its an error",
+ uuid_utoa(local->loc.inode->gfid));
+ goto out;
+ }
- LOCK (&ctx->lock);
- {
- ctx->buf = *postbuf;
- }
- UNLOCK (&ctx->lock);
+ LOCK(&ctx->lock);
+ {
+ ctx->buf = *postbuf;
+ }
+ UNLOCK(&ctx->lock);
out:
- QUOTA_STACK_UNWIND (ftruncate, frame, op_ret, op_errno, prebuf,
- postbuf, xdata);
- return 0;
+ QUOTA_STACK_UNWIND(ftruncate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
}
-
int32_t
-quota_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- dict_t *xdata)
+quota_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- quota_local_t *local = NULL;
+ quota_priv_t *priv = NULL;
+ quota_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
- local = quota_local_new ();
- if (local == NULL)
- goto err;
+ local = quota_local_new();
+ if (local == NULL)
+ goto err;
- frame->local = local;
+ frame->local = local;
- local->loc.inode = inode_ref (fd->inode);
+ local->loc.inode = inode_ref(fd->inode);
- STACK_WIND (frame, quota_ftruncate_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->ftruncate, fd,
- offset, xdata);
+ STACK_WIND(frame, quota_ftruncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
- return 0;
+ return 0;
err:
- QUOTA_STACK_UNWIND (ftruncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+ QUOTA_STACK_UNWIND(ftruncate, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->ftruncate, fd,
- offset, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+ return 0;
}
-
-int32_t
-quota_send_dir_limit_to_cli (call_frame_t *frame, xlator_t *this,
- inode_t *inode, const char *name)
+static int32_t
+quota_send_dir_limit_to_cli(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ const char *name, const int namelen)
{
- int32_t ret = 0;
- char dir_limit [1024] = {0, };
- dict_t *dict = NULL;
- quota_inode_ctx_t *ctx = NULL;
- uint64_t value = 0;
- quota_priv_t *priv = NULL;
-
- priv = this->private;
- if (!priv->is_quota_on) {
- snprintf (dir_limit, 1024, "Quota is disabled please turn on");
- goto dict_set;
- }
-
- ret = inode_ctx_get (inode, this, &value);
- if (ret < 0)
- goto out;
-
- ctx = (quota_inode_ctx_t *)(unsigned long)value;
- snprintf (dir_limit, 1024, "%"PRId64",%"PRId64, ctx->size,
- ctx->hard_lim);
+ int32_t ret = 0;
+ int dir_limit_len = 0;
+ char dir_limit[64] = {
+ 0,
+ };
+ dict_t *dict = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ uint64_t value = 0;
+ quota_priv_t *priv = NULL;
+
+ priv = this->private;
+ if (!priv->is_quota_on) {
+ dir_limit_len = snprintf(dir_limit, sizeof(dir_limit),
+ "Quota is disabled please turn on");
+ goto dict_set;
+ }
+
+ ret = inode_ctx_get(inode, this, &value);
+ if (ret < 0)
+ goto out;
+
+ ctx = (quota_inode_ctx_t *)(unsigned long)value;
+ dir_limit_len = snprintf(dir_limit, sizeof(dir_limit),
+ "%" PRId64 ",%" PRId64, ctx->size, ctx->hard_lim);
dict_set:
- dict = dict_new ();
- if (dict == NULL) {
- ret = -1;
- goto out;
- }
+ dict = dict_new();
+ if (dict == NULL) {
+ ret = -1;
+ goto out;
+ }
- ret = dict_set_str (dict, (char *) name, dir_limit);
- if (ret < 0)
- goto out;
+ ret = dict_set_nstrn(dict, (char *)name, namelen, dir_limit, dir_limit_len);
+ if (ret < 0)
+ goto out;
- gf_msg_debug (this->name, 0, "str = %s", dir_limit);
+ gf_msg_debug(this->name, 0, "str = %s", dir_limit);
- QUOTA_STACK_UNWIND (getxattr, frame, 0, 0, dict, NULL);
+ QUOTA_STACK_UNWIND(getxattr, frame, 0, 0, dict, NULL);
- ret = 0;
+ ret = 0;
out:
- if (dict)
- dict_unref (dict);
- return ret;
+ if (dict)
+ dict_unref(dict);
+ return ret;
}
-
int32_t
-quota_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
+quota_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
+ dict_t *xdata)
{
- int32_t ret = 0;
+ int32_t ret = 0;
- if (name && strcasecmp (name, "trusted.limit.list") == 0) {
- ret = quota_send_dir_limit_to_cli (frame, this, fd->inode,
- name);
- if (ret == 0) {
- return 0;
- }
+ if (name && strcasecmp(name, "trusted.limit.list") == 0) {
+ ret = quota_send_dir_limit_to_cli(frame, this, fd->inode,
+ "trusted.limit.list",
+ SLEN("trusted.limit.list"));
+ if (ret == 0) {
+ return 0;
}
+ }
- STACK_WIND (frame, default_fgetxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
- return 0;
+ STACK_WIND(frame, default_fgetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
+ return 0;
}
-
int32_t
-quota_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+quota_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- int32_t ret = 0;
-
- if ((name != NULL) && strcasecmp (name, "trusted.limit.list") == 0) {
- ret = quota_send_dir_limit_to_cli (frame, this, loc->inode,
- name);
- if (ret == 0)
- return 0;
- }
-
- STACK_WIND (frame, default_getxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
- return 0;
+ int32_t ret = 0;
+
+ if ((name != NULL) && strcasecmp(name, "trusted.limit.list") == 0) {
+ ret = quota_send_dir_limit_to_cli(frame, this, loc->inode,
+ "trusted.limit.list",
+ SLEN("trusted.limit.list"));
+ if (ret == 0)
+ return 0;
+ }
+
+ STACK_WIND(frame, default_getxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+ return 0;
}
-
int32_t
-quota_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- dict_t *xdata)
+quota_stat_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
{
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
- if (op_ret < 0) {
- goto out;
- }
+ if (op_ret < 0) {
+ goto out;
+ }
- local = frame->local;
+ local = frame->local;
- GF_VALIDATE_OR_GOTO ("quota", local, out);
+ GF_VALIDATE_OR_GOTO("quota", local, out);
- quota_inode_ctx_get (local->loc.inode, this, &ctx, 0);
- if (ctx == NULL) {
- if (!IA_ISDIR (buf->ia_type)) {
- gf_msg_debug (this->name, 0, "quota context is NULL on inode"
- " (%s). If quota is not enabled recently and "
- "crawler has finished crawling, its an error",
- uuid_utoa (local->loc.inode->gfid));
- }
-
- goto out;
+ quota_inode_ctx_get(local->loc.inode, this, &ctx, 0);
+ if (ctx == NULL) {
+ if (!IA_ISDIR(buf->ia_type)) {
+ gf_msg_debug(this->name, 0,
+ "quota context is NULL on inode"
+ " (%s). If quota is not enabled recently and "
+ "crawler has finished crawling, its an error",
+ uuid_utoa(local->loc.inode->gfid));
}
- LOCK (&ctx->lock);
- {
- if (buf)
- ctx->buf = *buf;
- }
- UNLOCK (&ctx->lock);
+ goto out;
+ }
+
+ if (buf) {
+ LOCK(&ctx->lock);
+ ctx->buf = *buf;
+ UNLOCK(&ctx->lock);
+ }
out:
- QUOTA_STACK_UNWIND (stat, frame, op_ret, op_errno, buf, xdata);
- return 0;
+ QUOTA_STACK_UNWIND(stat, frame, op_ret, op_errno, buf, xdata);
+ return 0;
}
-
int32_t
-quota_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+quota_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- quota_local_t *local = NULL;
- int32_t ret = -1;
+ quota_priv_t *priv = NULL;
+ quota_local_t *local = NULL;
+ int32_t ret = -1;
- priv = this->private;
+ priv = this->private;
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
- local = quota_local_new ();
- if (local == NULL) {
- goto unwind;
- }
+ local = quota_local_new();
+ if (local == NULL) {
+ goto unwind;
+ }
- frame->local = local;
- ret = loc_copy (&local->loc, loc);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "loc_copy failed");
- goto unwind;
- }
+ frame->local = local;
+ ret = loc_copy(&local->loc, loc);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "loc_copy failed");
+ goto unwind;
+ }
- STACK_WIND (frame, quota_stat_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->stat, loc,
- xdata);
- return 0;
+ STACK_WIND(frame, quota_stat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->stat, loc, xdata);
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (stat, frame, -1, ENOMEM, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(stat, frame, -1, ENOMEM, NULL, NULL);
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->stat, loc,
- xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->stat,
+ loc, xdata);
+ return 0;
}
-
int32_t
-quota_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- dict_t *xdata)
+quota_fstat_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
{
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
- if (op_ret < 0) {
- goto out;
- }
+ if (op_ret < 0) {
+ goto out;
+ }
- local = frame->local;
+ local = frame->local;
- GF_VALIDATE_OR_GOTO ("quota", local, out);
-
- quota_inode_ctx_get (local->loc.inode, this, &ctx, 0);
- if (ctx == NULL) {
- if (!IA_ISDIR (buf->ia_type)) {
- gf_msg_debug (this->name, 0, "quota context is NULL on inode"
- " (%s). If quota is not enabled recently and "
- "crawler has finished crawling, its an error",
- uuid_utoa (local->loc.inode->gfid));
- }
+ GF_VALIDATE_OR_GOTO("quota", local, out);
- goto out;
+ quota_inode_ctx_get(local->loc.inode, this, &ctx, 0);
+ if (ctx == NULL) {
+ if (!IA_ISDIR(buf->ia_type)) {
+ gf_msg_debug(this->name, 0,
+ "quota context is NULL on inode"
+ " (%s). If quota is not enabled recently and "
+ "crawler has finished crawling, its an error",
+ uuid_utoa(local->loc.inode->gfid));
}
- LOCK (&ctx->lock);
- {
- if (buf)
- ctx->buf = *buf;
- }
- UNLOCK (&ctx->lock);
+ goto out;
+ }
+
+ if (buf) {
+ LOCK(&ctx->lock);
+ ctx->buf = *buf;
+ UNLOCK(&ctx->lock);
+ }
out:
- QUOTA_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf, xdata);
- return 0;
+ QUOTA_STACK_UNWIND(fstat, frame, op_ret, op_errno, buf, xdata);
+ return 0;
}
-
int32_t
-quota_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+quota_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- quota_local_t *local = NULL;
+ quota_priv_t *priv = NULL;
+ quota_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
- local = quota_local_new ();
- if (local == NULL) {
- goto unwind;
- }
+ local = quota_local_new();
+ if (local == NULL) {
+ goto unwind;
+ }
- frame->local = local;
+ frame->local = local;
- local->loc.inode = inode_ref (fd->inode);
+ local->loc.inode = inode_ref(fd->inode);
- STACK_WIND (frame, quota_fstat_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->fstat, fd,
- xdata);
- return 0;
+ STACK_WIND(frame, quota_fstat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fstat, fd, xdata);
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (fstat, frame, -1, ENOMEM, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(fstat, frame, -1, ENOMEM, NULL, NULL);
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fstat, fd,
- xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->fstat,
+ fd, xdata);
+ return 0;
}
-
int32_t
-quota_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, const char *path,
- struct iatt *buf, dict_t *xdata)
+quota_readlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, const char *path,
+ struct iatt *buf, dict_t *xdata)
{
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
- if (op_ret < 0) {
- goto out;
- }
+ if (op_ret < 0) {
+ goto out;
+ }
- local = frame->local;
+ local = frame->local;
- GF_VALIDATE_OR_GOTO ("quota", local, out);
+ GF_VALIDATE_OR_GOTO("quota", local, out);
- quota_inode_ctx_get (local->loc.inode, this, &ctx, 0);
- if (ctx == NULL) {
- gf_msg_debug (this->name, 0, "quota context is NULL on inode"
- " (%s). If quota is not enabled recently and "
- "crawler has finished crawling, its an error",
- uuid_utoa (local->loc.inode->gfid));
- goto out;
- }
+ quota_inode_ctx_get(local->loc.inode, this, &ctx, 0);
+ if (ctx == NULL) {
+ gf_msg_debug(this->name, 0,
+ "quota context is NULL on inode"
+ " (%s). If quota is not enabled recently and "
+ "crawler has finished crawling, its an error",
+ uuid_utoa(local->loc.inode->gfid));
+ goto out;
+ }
- LOCK (&ctx->lock);
- {
- ctx->buf = *buf;
- }
- UNLOCK (&ctx->lock);
+ LOCK(&ctx->lock);
+ {
+ ctx->buf = *buf;
+ }
+ UNLOCK(&ctx->lock);
out:
- QUOTA_STACK_UNWIND (readlink, frame, op_ret, op_errno, path, buf,
- xdata);
- return 0;
+ QUOTA_STACK_UNWIND(readlink, frame, op_ret, op_errno, path, buf, xdata);
+ return 0;
}
-
int32_t
-quota_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
- dict_t *xdata)
+quota_readlink(call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+ dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- quota_local_t *local = NULL;
- int32_t ret = -1;
+ quota_priv_t *priv = NULL;
+ quota_local_t *local = NULL;
+ int32_t ret = -1;
- priv = this->private;
+ priv = this->private;
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
- local = quota_local_new ();
- if (local == NULL) {
- goto unwind;
- }
+ local = quota_local_new();
+ if (local == NULL) {
+ goto unwind;
+ }
- frame->local = local;
+ frame->local = local;
- ret = loc_copy (&local->loc, loc);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "loc_copy failed");
- goto unwind;
- }
+ ret = loc_copy(&local->loc, loc);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "loc_copy failed");
+ goto unwind;
+ }
- STACK_WIND (frame, quota_readlink_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->readlink, loc,
- size, xdata);
- return 0;
+ STACK_WIND(frame, quota_readlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readlink, loc, size, xdata);
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (readlink, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(readlink, frame, -1, ENOMEM, NULL, NULL, NULL);
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readlink, loc,
- size, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->readlink,
+ loc, size, xdata);
+ return 0;
}
-
int32_t
-quota_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iovec *vector,
- int32_t count, struct iatt *buf, struct iobref *iobref,
- dict_t *xdata)
+quota_readv_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iovec *vector,
+ int32_t count, struct iatt *buf, struct iobref *iobref,
+ dict_t *xdata)
{
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
- if (op_ret < 0) {
- goto out;
- }
+ if (op_ret < 0) {
+ goto out;
+ }
- local = frame->local;
+ local = frame->local;
- GF_VALIDATE_OR_GOTO ("quota", local, out);
+ GF_VALIDATE_OR_GOTO("quota", local, out);
- quota_inode_ctx_get (local->loc.inode, this, &ctx, 0);
- if (ctx == NULL) {
- gf_msg_debug (this->name, 0, "quota context is NULL on inode"
- " (%s). If quota is not enabled recently and "
- "crawler has finished crawling, its an error",
- uuid_utoa (local->loc.inode->gfid));
- goto out;
- }
+ quota_inode_ctx_get(local->loc.inode, this, &ctx, 0);
+ if (ctx == NULL) {
+ gf_msg_debug(this->name, 0,
+ "quota context is NULL on inode"
+ " (%s). If quota is not enabled recently and "
+ "crawler has finished crawling, its an error",
+ uuid_utoa(local->loc.inode->gfid));
+ goto out;
+ }
- LOCK (&ctx->lock);
- {
- ctx->buf = *buf;
- }
- UNLOCK (&ctx->lock);
+ LOCK(&ctx->lock);
+ {
+ ctx->buf = *buf;
+ }
+ UNLOCK(&ctx->lock);
out:
- QUOTA_STACK_UNWIND (readv, frame, op_ret, op_errno, vector, count,
- buf, iobref, xdata);
- return 0;
+ QUOTA_STACK_UNWIND(readv, frame, op_ret, op_errno, vector, count, buf,
+ iobref, xdata);
+ return 0;
}
-
int32_t
-quota_readv (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, uint32_t flags, dict_t *xdata)
+quota_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- quota_local_t *local = NULL;
+ quota_priv_t *priv = NULL;
+ quota_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
- local = quota_local_new ();
- if (local == NULL) {
- goto unwind;
- }
+ local = quota_local_new();
+ if (local == NULL) {
+ goto unwind;
+ }
- frame->local = local;
+ frame->local = local;
- local->loc.inode = inode_ref (fd->inode);
+ local->loc.inode = inode_ref(fd->inode);
- STACK_WIND (frame, quota_readv_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->readv, fd,
- size, offset, flags, xdata);
- return 0;
+ STACK_WIND(frame, quota_readv_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readv, fd, size, offset, flags, xdata);
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (readv, frame, -1, ENOMEM, NULL, -1, NULL, NULL,
- NULL);
- return 0;
+ QUOTA_STACK_UNWIND(readv, frame, -1, ENOMEM, NULL, -1, NULL, NULL, NULL);
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readv, fd,
- size, offset, flags, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->readv,
+ fd, size, offset, flags, xdata);
+ return 0;
}
-
int32_t
-quota_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+quota_fsync_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
{
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
- if (op_ret < 0) {
- goto out;
- }
+ if (op_ret < 0) {
+ goto out;
+ }
- local = frame->local;
+ local = frame->local;
- GF_VALIDATE_OR_GOTO ("quota", local, out);
+ GF_VALIDATE_OR_GOTO("quota", local, out);
- quota_inode_ctx_get (local->loc.inode, this, &ctx, 0);
- if (ctx == NULL) {
- gf_msg_debug (this->name, 0, "quota context is NULL on inode"
- " (%s). If quota is not enabled recently and "
- "crawler has finished crawling, its an error",
- uuid_utoa (local->loc.inode->gfid));
- goto out;
- }
+ quota_inode_ctx_get(local->loc.inode, this, &ctx, 0);
+ if (ctx == NULL) {
+ gf_msg_debug(this->name, 0,
+ "quota context is NULL on inode"
+ " (%s). If quota is not enabled recently and "
+ "crawler has finished crawling, its an error",
+ uuid_utoa(local->loc.inode->gfid));
+ goto out;
+ }
- LOCK (&ctx->lock);
- {
- ctx->buf = *postbuf;
- }
- UNLOCK (&ctx->lock);
+ LOCK(&ctx->lock);
+ {
+ ctx->buf = *postbuf;
+ }
+ UNLOCK(&ctx->lock);
out:
- QUOTA_STACK_UNWIND (fsync, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
- return 0;
+ QUOTA_STACK_UNWIND(fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+ return 0;
}
-
int32_t
-quota_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
- dict_t *xdata)
+quota_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+ dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- quota_local_t *local = NULL;
+ quota_priv_t *priv = NULL;
+ quota_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
- local = quota_local_new ();
- if (local == NULL) {
- goto unwind;
- }
+ local = quota_local_new();
+ if (local == NULL) {
+ goto unwind;
+ }
- local->loc.inode = inode_ref (fd->inode);
+ local->loc.inode = inode_ref(fd->inode);
- frame->local = local;
+ frame->local = local;
- STACK_WIND (frame, quota_fsync_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsync, fd,
- flags, xdata);
- return 0;
+ STACK_WIND(frame, quota_fsync_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsync, fd, flags, xdata);
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (fsync, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(fsync, frame, -1, ENOMEM, NULL, NULL, NULL);
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsync, fd,
- flags, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->fsync,
+ fd, flags, xdata);
+ return 0;
}
-
int32_t
-quota_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *statpre,
- struct iatt *statpost, dict_t *xdata)
+quota_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
{
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
- if (op_ret < 0) {
- goto out;
- }
+ if (op_ret < 0) {
+ goto out;
+ }
- local = frame->local;
+ local = frame->local;
- GF_VALIDATE_OR_GOTO ("quota", local, out);
-
- quota_inode_ctx_get (local->loc.inode, this, &ctx, 0);
- if (ctx == NULL) {
- if (!IA_ISDIR (statpost->ia_type)) {
- gf_msg_debug (this->name, 0, "quota context is NULL on inode"
- " (%s). If quota is not enabled recently and "
- "crawler has finished crawling, its an error",
- uuid_utoa (local->loc.inode->gfid));
- }
+ GF_VALIDATE_OR_GOTO("quota", local, out);
- goto out;
+ quota_inode_ctx_get(local->loc.inode, this, &ctx, 0);
+ if (ctx == NULL) {
+ if (!IA_ISDIR(statpost->ia_type)) {
+ gf_msg_debug(this->name, 0,
+ "quota context is NULL on inode"
+ " (%s). If quota is not enabled recently and "
+ "crawler has finished crawling, its an error",
+ uuid_utoa(local->loc.inode->gfid));
}
- LOCK (&ctx->lock);
- {
- if (statpost)
- ctx->buf = *statpost;
- }
- UNLOCK (&ctx->lock);
+ goto out;
+ }
+
+ if (statpost) {
+ LOCK(&ctx->lock);
+ ctx->buf = *statpost;
+ UNLOCK(&ctx->lock);
+ }
out:
- QUOTA_STACK_UNWIND (setattr, frame, op_ret, op_errno, statpre,
- statpost, xdata);
- return 0;
+ QUOTA_STACK_UNWIND(setattr, frame, op_ret, op_errno, statpre, statpost,
+ xdata);
+ return 0;
}
-
int32_t
-quota_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+quota_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- quota_local_t *local = NULL;
- int32_t ret = -1;
+ quota_priv_t *priv = NULL;
+ quota_local_t *local = NULL;
+ int32_t ret = -1;
- priv = this->private;
+ priv = this->private;
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
- local = quota_local_new ();
- if (local == NULL) {
- goto unwind;
- }
+ local = quota_local_new();
+ if (local == NULL) {
+ goto unwind;
+ }
- frame->local = local;
+ frame->local = local;
- ret = loc_copy (&local->loc, loc);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "loc_copy failed");
- goto unwind;
- }
+ ret = loc_copy(&local->loc, loc);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "loc_copy failed");
+ goto unwind;
+ }
- STACK_WIND (frame, quota_setattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setattr, loc,
- stbuf, valid, xdata);
- return 0;
+ STACK_WIND(frame, quota_setattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid, xdata);
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (setattr, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(setattr, frame, -1, ENOMEM, NULL, NULL, NULL);
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setattr, loc,
- stbuf, valid, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->setattr,
+ loc, stbuf, valid, xdata);
+ return 0;
}
-
int32_t
-quota_fsetattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *statpre,
- struct iatt *statpost, dict_t *xdata)
+quota_fsetattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+ struct iatt *statpost, dict_t *xdata)
{
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
- if (op_ret < 0) {
- goto out;
- }
+ if (op_ret < 0) {
+ goto out;
+ }
- local = frame->local;
+ local = frame->local;
- GF_VALIDATE_OR_GOTO ("quota", local, out);
-
- quota_inode_ctx_get (local->loc.inode, this, &ctx, 0);
- if (ctx == NULL) {
- if (!IA_ISDIR (statpost->ia_type)) {
- gf_msg_debug (this->name, 0, "quota context is NULL on inode"
- " (%s). If quota is not enabled recently and "
- "crawler has finished crawling, its an error",
- uuid_utoa (local->loc.inode->gfid));
- }
+ GF_VALIDATE_OR_GOTO("quota", local, out);
- goto out;
+ quota_inode_ctx_get(local->loc.inode, this, &ctx, 0);
+ if (ctx == NULL) {
+ if (!IA_ISDIR(statpost->ia_type)) {
+ gf_msg_debug(this->name, 0,
+ "quota context is NULL on inode"
+ " (%s). If quota is not enabled recently and "
+ "crawler has finished crawling, its an error",
+ uuid_utoa(local->loc.inode->gfid));
}
- LOCK (&ctx->lock);
- {
- ctx->buf = *statpost;
- }
- UNLOCK (&ctx->lock);
+ goto out;
+ }
+
+ LOCK(&ctx->lock);
+ {
+ ctx->buf = *statpost;
+ }
+ UNLOCK(&ctx->lock);
out:
- QUOTA_STACK_UNWIND (fsetattr, frame, op_ret, op_errno, statpre,
- statpost, xdata);
- return 0;
+ QUOTA_STACK_UNWIND(fsetattr, frame, op_ret, op_errno, statpre, statpost,
+ xdata);
+ return 0;
}
-
int32_t
-quota_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+quota_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- quota_local_t *local = NULL;
+ quota_priv_t *priv = NULL;
+ quota_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
- local = quota_local_new ();
- if (local == NULL) {
- goto unwind;
- }
+ local = quota_local_new();
+ if (local == NULL) {
+ goto unwind;
+ }
- frame->local = local;
+ frame->local = local;
- local->loc.inode = inode_ref (fd->inode);
+ local->loc.inode = inode_ref(fd->inode);
- STACK_WIND (frame, quota_fsetattr_cbk,
- FIRST_CHILD (this), FIRST_CHILD (this)->fops->fsetattr, fd,
- stbuf, valid, xdata);
- return 0;
+ STACK_WIND(frame, quota_fsetattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetattr, fd, stbuf, valid, xdata);
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (fsetattr, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(fsetattr, frame, -1, ENOMEM, NULL, NULL, NULL);
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fsetattr, fd,
- stbuf, valid, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->fsetattr,
+ fd, stbuf, valid, xdata);
+ return 0;
}
-
int32_t
-quota_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+quota_mknod_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- int32_t ret = -1;
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
- quota_dentry_t *dentry = NULL;
-
- local = frame->local;
- if (op_ret < 0) {
- goto unwind;
- }
-
- ret = quota_inode_ctx_get (inode, this, &ctx, 1);
- if ((ret == -1) || (ctx == NULL)) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- Q_MSG_INODE_CTX_GET_FAILED,
- "cannot create quota context in "
- "inode(gfid:%s)", uuid_utoa (inode->gfid));
- op_ret = -1;
- op_errno = ENOMEM;
- goto unwind;
- }
-
- LOCK (&ctx->lock);
- {
- ctx->buf = *buf;
-
- dentry = __quota_dentry_new (ctx, (char *)local->loc.name,
- local->loc.parent->gfid);
- if (dentry == NULL) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "cannot create a new dentry "
- "(name:%s) for inode(gfid:%s)", local->loc.name,
- uuid_utoa (local->loc.inode->gfid));
- op_ret = -1;
- op_errno = ENOMEM;
- goto unlock;
- }
- }
+ int32_t ret = -1;
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ quota_dentry_t *dentry = NULL;
+
+ local = frame->local;
+ if (op_ret < 0) {
+ goto unwind;
+ }
+
+ ret = quota_inode_ctx_get(inode, this, &ctx, 1);
+ if ((ret == -1) || (ctx == NULL)) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, Q_MSG_INODE_CTX_GET_FAILED,
+ "cannot create quota context in "
+ "inode(gfid:%s)",
+ uuid_utoa(inode->gfid));
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ LOCK(&ctx->lock);
+ {
+ ctx->buf = *buf;
+
+ dentry = __quota_dentry_new(ctx, (char *)local->loc.name,
+ local->loc.parent->gfid);
+ if (dentry == NULL) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "cannot create a new dentry "
+ "(name:%s) for inode(gfid:%s)",
+ local->loc.name, uuid_utoa(local->loc.inode->gfid));
+ op_ret = -1;
+ op_errno = ENOMEM;
+ goto unlock;
+ }
+ }
unlock:
- UNLOCK (&ctx->lock);
+ UNLOCK(&ctx->lock);
unwind:
- QUOTA_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode,
- buf, preparent, postparent, xdata);
- return 0;
+ QUOTA_STACK_UNWIND(mknod, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
+ return 0;
}
-
int
-quota_mknod_helper (call_frame_t *frame, xlator_t *this, loc_t *loc,
- mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata)
+quota_mknod_helper(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *xdata)
{
- quota_local_t *local = NULL;
- int32_t op_errno = EINVAL;
+ quota_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
- local = frame->local;
+ local = frame->local;
- GF_VALIDATE_OR_GOTO ("quota", local, unwind);
+ GF_VALIDATE_OR_GOTO("quota", local, unwind);
- if (local->op_ret == -1) {
- op_errno = local->op_errno;
- goto unwind;
- }
+ if (local->op_ret == -1) {
+ op_errno = local->op_errno;
+ goto unwind;
+ }
- STACK_WIND (frame, quota_mknod_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->mknod, loc,
- mode, rdev, umask, xdata);
+ STACK_WIND(frame, quota_mknod_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, loc, mode, rdev, umask, xdata);
- return 0;
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (mknod, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
-
int
-quota_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- dev_t rdev, mode_t umask, dict_t *xdata)
+quota_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- int32_t ret = -1;
- quota_local_t *local = NULL;
- call_stub_t *stub = NULL;
-
- priv = this->private;
-
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
- QUOTA_WIND_FOR_INTERNAL_FOP (xdata, off);
-
- local = quota_local_new ();
- if (local == NULL) {
- goto err;
- }
-
- frame->local = local;
-
- ret = loc_copy (&local->loc, loc);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "loc_copy failed");
- goto err;
- }
-
- stub = fop_mknod_stub (frame, quota_mknod_helper, loc, mode, rdev,
- umask, xdata);
- if (stub == NULL) {
- goto err;
- }
+ quota_priv_t *priv = NULL;
+ int32_t ret = -1;
+ quota_local_t *local = NULL;
+ call_stub_t *stub = NULL;
+
+ priv = this->private;
+
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
+ QUOTA_WIND_FOR_INTERNAL_FOP(xdata, off);
+
+ local = quota_local_new();
+ if (local == NULL) {
+ goto err;
+ }
+
+ frame->local = local;
+
+ ret = loc_copy(&local->loc, loc);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "loc_copy failed");
+ goto err;
+ }
+
+ stub = fop_mknod_stub(frame, quota_mknod_helper, loc, mode, rdev, umask,
+ xdata);
+ if (stub == NULL) {
+ goto err;
+ }
+
+ LOCK(&local->lock);
+ {
+ local->link_count = 1;
+ local->stub = stub;
+ local->delta = 0;
+ local->object_delta = 1;
+ }
+ UNLOCK(&local->lock);
- LOCK (&local->lock);
- {
- local->link_count = 1;
- local->stub = stub;
- local->delta = 0;
- local->object_delta = 1;
- }
- UNLOCK (&local->lock);
-
- quota_check_limit (frame, loc->parent, this);
- return 0;
+ quota_check_limit(frame, loc->parent, this);
+ return 0;
err:
- QUOTA_STACK_UNWIND (mknod, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
- NULL);
- return 0;
+ QUOTA_STACK_UNWIND(mknod, frame, -1, ENOMEM, NULL, NULL, NULL, NULL, NULL);
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mknod, loc,
- mode, rdev, umask, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->mknod,
+ loc, mode, rdev, umask, xdata);
+ return 0;
}
int
-quota_setxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno, dict_t *xdata)
+quota_setxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
{
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
- int ret = 0;
-
- if (op_ret < 0) {
- goto out;
- }
-
- local = frame->local;
- if (!local)
- goto out;
-
- ret = quota_inode_ctx_get (local->loc.inode, this, &ctx, 1);
- if ((ret < 0) || (ctx == NULL)) {
- op_errno = -1;
- goto out;
- }
-
- LOCK (&ctx->lock);
- {
- ctx->hard_lim = local->limit.hl;
- ctx->soft_lim = local->limit.sl;
- ctx->object_hard_lim = local->object_limit.hl;
- ctx->object_soft_lim = local->object_limit.sl;
- }
- UNLOCK (&ctx->lock);
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ int ret = 0;
+
+ if (op_ret < 0) {
+ goto out;
+ }
+
+ local = frame->local;
+ if (!local)
+ goto out;
+
+ ret = quota_inode_ctx_get(local->loc.inode, this, &ctx, 1);
+ if ((ret < 0) || (ctx == NULL)) {
+ op_errno = -1;
+ goto out;
+ }
+
+ LOCK(&ctx->lock);
+ {
+ ctx->hard_lim = local->limit.hl;
+ ctx->soft_lim = local->limit.sl;
+ ctx->object_hard_lim = local->object_limit.hl;
+ ctx->object_soft_lim = local->object_limit.sl;
+ }
+ UNLOCK(&ctx->lock);
out:
- QUOTA_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
- return 0;
+ QUOTA_STACK_UNWIND(setxattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-quota_setxattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, dict_t *dict, int flags, dict_t *xdata)
+quota_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int flags, dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- int op_errno = EINVAL;
- int op_ret = -1;
- int64_t hard_lim = -1;
- int64_t soft_lim = -1;
- int64_t object_hard_limit = -1;
- int64_t object_soft_limit = -1;
- quota_local_t *local = NULL;
- gf_boolean_t internal_fop = _gf_false;
-
- priv = this->private;
-
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (loc, err);
-
- if (xdata && dict_get (xdata, GLUSTERFS_INTERNAL_FOP_KEY))
- internal_fop = _gf_true;
-
- if (frame->root->pid >= 0 && internal_fop == _gf_false) {
- GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.quota*", dict,
- op_errno, err);
- GF_IF_INTERNAL_XATTR_GOTO ("trusted.pgfid*", dict, op_errno,
- err);
- }
-
- quota_get_limits (this, dict, &hard_lim, &soft_lim, &object_hard_limit,
- &object_soft_limit);
-
- if (hard_lim > 0 || object_hard_limit > 0) {
- local = quota_local_new ();
- if (local == NULL) {
- op_errno = ENOMEM;
- goto err;
- }
- frame->local = local;
- loc_copy (&local->loc, loc);
- }
-
- if (hard_lim > 0) {
- local->limit.hl = hard_lim;
- local->limit.sl = soft_lim;
- }
-
- if (object_hard_limit > 0) {
- local->object_limit.hl = object_hard_limit;
- local->object_limit.sl = object_soft_limit;
+ quota_priv_t *priv = NULL;
+ int op_errno = EINVAL;
+ int op_ret = -1;
+ int64_t hard_lim = -1;
+ int64_t soft_lim = -1;
+ int64_t object_hard_limit = -1;
+ int64_t object_soft_limit = -1;
+ quota_local_t *local = NULL;
+ gf_boolean_t internal_fop = _gf_false;
+
+ priv = this->private;
+
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(loc, err);
+
+ if (xdata && dict_get_sizen(xdata, GLUSTERFS_INTERNAL_FOP_KEY))
+ internal_fop = _gf_true;
+
+ if (frame->root->pid >= 0 && internal_fop == _gf_false) {
+ GF_IF_INTERNAL_XATTR_GOTO("trusted.glusterfs.quota*", dict, op_errno,
+ err);
+ GF_IF_INTERNAL_XATTR_GOTO("trusted.pgfid*", dict, op_errno, err);
+ }
+
+ quota_get_limits(this, dict, &hard_lim, &soft_lim, &object_hard_limit,
+ &object_soft_limit);
+
+ if (hard_lim > 0 || object_hard_limit > 0) {
+ local = quota_local_new();
+ if (local == NULL) {
+ op_errno = ENOMEM;
+ goto err;
}
-
- STACK_WIND (frame, quota_setxattr_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->setxattr, loc,
- dict, flags, xdata);
- return 0;
+ frame->local = local;
+ loc_copy(&local->loc, loc);
+ }
+
+ if (hard_lim > 0) {
+ local->limit.hl = hard_lim;
+ local->limit.sl = soft_lim;
+ }
+
+ if (object_hard_limit > 0) {
+ local->object_limit.hl = object_hard_limit;
+ local->object_limit.sl = object_soft_limit;
+ }
+
+ STACK_WIND(frame, quota_setxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, loc, dict, flags, xdata);
+ return 0;
err:
- QUOTA_STACK_UNWIND (setxattr, frame, op_ret, op_errno, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(setxattr, frame, op_ret, op_errno, NULL);
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setxattr, loc,
- dict, flags, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->setxattr,
+ loc, dict, flags, xdata);
+ return 0;
}
int
-quota_fsetxattr_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int op_ret, int op_errno, dict_t *xdata)
+quota_fsetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
{
- quota_inode_ctx_t *ctx = NULL;
- quota_local_t *local = NULL;
-
- if (op_ret < 0)
- goto out;
-
- local = frame->local;
- if (!local)
- goto out;
-
- op_ret = quota_inode_ctx_get (local->loc.inode, this, &ctx, 1);
- if ((op_ret < 0) || (ctx == NULL)) {
- op_errno = ENOMEM;
- goto out;
- }
-
- LOCK (&ctx->lock);
- {
- ctx->hard_lim = local->limit.hl;
- ctx->soft_lim = local->limit.sl;
- ctx->object_hard_lim = local->object_limit.hl;
- ctx->object_soft_lim = local->object_limit.sl;
- }
- UNLOCK (&ctx->lock);
+ quota_inode_ctx_t *ctx = NULL;
+ quota_local_t *local = NULL;
+
+ if (op_ret < 0)
+ goto out;
+
+ local = frame->local;
+ if (!local)
+ goto out;
+
+ op_ret = quota_inode_ctx_get(local->loc.inode, this, &ctx, 1);
+ if ((op_ret < 0) || (ctx == NULL)) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ LOCK(&ctx->lock);
+ {
+ ctx->hard_lim = local->limit.hl;
+ ctx->soft_lim = local->limit.sl;
+ ctx->object_hard_lim = local->object_limit.hl;
+ ctx->object_soft_lim = local->object_limit.sl;
+ }
+ UNLOCK(&ctx->lock);
out:
- QUOTA_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, xdata);
- return 0;
+ QUOTA_STACK_UNWIND(fsetxattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-quota_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- dict_t *dict, int flags, dict_t *xdata)
+quota_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int flags, dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- quota_local_t *local = NULL;
- int64_t hard_lim = -1;
- int64_t soft_lim = -1;
- int64_t object_hard_limit = -1;
- int64_t object_soft_limit = -1;
-
- priv = this->private;
-
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- if (0 <= frame->root->pid) {
- GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.quota*",
- dict, op_errno, err);
- GF_IF_INTERNAL_XATTR_GOTO ("trusted.pgfid*", dict,
- op_errno, err);
- }
-
- quota_get_limits (this, dict, &hard_lim, &soft_lim, &object_hard_limit,
- &object_soft_limit);
-
- if (hard_lim > 0 || object_hard_limit > 0) {
- local = quota_local_new ();
- if (local == NULL) {
- op_errno = ENOMEM;
- goto err;
- }
- frame->local = local;
- local->loc.inode = inode_ref (fd->inode);
- }
-
- if (hard_lim > 0) {
- local->limit.hl = hard_lim;
- local->limit.sl = soft_lim;
- }
-
- if (object_hard_limit > 0) {
- local->object_limit.hl = object_hard_limit;
- local->object_limit.sl = object_soft_limit;
+ quota_priv_t *priv = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ quota_local_t *local = NULL;
+ int64_t hard_lim = -1;
+ int64_t soft_lim = -1;
+ int64_t object_hard_limit = -1;
+ int64_t object_soft_limit = -1;
+
+ priv = this->private;
+
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+
+ if (0 <= frame->root->pid) {
+ GF_IF_INTERNAL_XATTR_GOTO("trusted.glusterfs.quota*", dict, op_errno,
+ err);
+ GF_IF_INTERNAL_XATTR_GOTO("trusted.pgfid*", dict, op_errno, err);
+ }
+
+ quota_get_limits(this, dict, &hard_lim, &soft_lim, &object_hard_limit,
+ &object_soft_limit);
+
+ if (hard_lim > 0 || object_hard_limit > 0) {
+ local = quota_local_new();
+ if (local == NULL) {
+ op_errno = ENOMEM;
+ goto err;
}
-
- STACK_WIND (frame, quota_fsetxattr_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->fsetxattr, fd,
- dict, flags, xdata);
- return 0;
+ frame->local = local;
+ local->loc.inode = inode_ref(fd->inode);
+ }
+
+ if (hard_lim > 0) {
+ local->limit.hl = hard_lim;
+ local->limit.sl = soft_lim;
+ }
+
+ if (object_hard_limit > 0) {
+ local->object_limit.hl = object_hard_limit;
+ local->object_limit.sl = object_soft_limit;
+ }
+
+ STACK_WIND(frame, quota_fsetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+ return 0;
err:
- QUOTA_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(fsetxattr, frame, op_ret, op_errno, NULL);
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetxattr, fd,
- dict, flags, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+ return 0;
}
-
int
-quota_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+quota_removexattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- QUOTA_STACK_UNWIND (removexattr, frame, op_ret, op_errno, xdata);
- return 0;
+ QUOTA_STACK_UNWIND(removexattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-quota_removexattr (call_frame_t *frame, xlator_t *this,
- loc_t *loc, const char *name, dict_t *xdata)
+quota_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- int32_t op_errno = EINVAL;
+ quota_priv_t *priv = NULL;
+ int32_t op_errno = EINVAL;
- priv = this->private;
+ priv = this->private;
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
- VALIDATE_OR_GOTO (this, err);
+ VALIDATE_OR_GOTO(this, err);
- /* all quota xattrs can be cleaned up by doing setxattr on special key.
- * Hence its ok that we don't allow removexattr on quota keys here.
- */
- if (frame->root->pid >= 0) {
- GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.quota*",
- name, op_errno, err);
- GF_IF_NATIVE_XATTR_GOTO ("trusted.pgfid*", name,
- op_errno, err);
- }
+ /* all quota xattrs can be cleaned up by doing setxattr on special key.
+ * Hence its ok that we don't allow removexattr on quota keys here.
+ */
+ if (frame->root->pid >= 0) {
+ GF_IF_NATIVE_XATTR_GOTO("trusted.glusterfs.quota*", name, op_errno,
+ err);
+ GF_IF_NATIVE_XATTR_GOTO("trusted.pgfid*", name, op_errno, err);
+ }
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (loc, err);
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(loc, err);
- STACK_WIND (frame, quota_removexattr_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->removexattr,
- loc, name, xdata);
- return 0;
+ STACK_WIND(frame, quota_removexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
+ return 0;
err:
- QUOTA_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(removexattr, frame, -1, op_errno, NULL);
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->removexattr,
- loc, name, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
+ return 0;
}
-
int
-quota_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+quota_fremovexattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- QUOTA_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, xdata);
- return 0;
+ QUOTA_STACK_UNWIND(fremovexattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-quota_fremovexattr (call_frame_t *frame, xlator_t *this,
- fd_t *fd, const char *name, dict_t *xdata)
+quota_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata)
{
- quota_priv_t *priv = NULL;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
-
- priv = this->private;
-
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
-
- VALIDATE_OR_GOTO (frame, err);
- VALIDATE_OR_GOTO (this, err);
- VALIDATE_OR_GOTO (fd, err);
-
- if (frame->root->pid >= 0) {
- GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.quota*",
- name, op_errno, err);
- GF_IF_NATIVE_XATTR_GOTO ("trusted.pgfid*", name,
- op_errno, err);
- }
- STACK_WIND (frame, quota_fremovexattr_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->fremovexattr,
- fd, name, xdata);
- return 0;
+ quota_priv_t *priv = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+
+ priv = this->private;
+
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
+
+ VALIDATE_OR_GOTO(frame, err);
+ VALIDATE_OR_GOTO(this, err);
+ VALIDATE_OR_GOTO(fd, err);
+
+ if (frame->root->pid >= 0) {
+ GF_IF_NATIVE_XATTR_GOTO("trusted.glusterfs.quota*", name, op_errno,
+ err);
+ GF_IF_NATIVE_XATTR_GOTO("trusted.pgfid*", name, op_errno, err);
+ }
+ STACK_WIND(frame, quota_fremovexattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fremovexattr, fd, name, xdata);
+ return 0;
err:
- QUOTA_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(fremovexattr, frame, op_ret, op_errno, NULL);
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fremovexattr,
- fd, name, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fremovexattr, fd, name, xdata);
+ return 0;
}
-
int32_t
-quota_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct statvfs *buf,
- dict_t *xdata)
+quota_statfs_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct statvfs *buf,
+ dict_t *xdata)
{
- inode_t *inode = NULL;
- uint64_t value = 0;
- int64_t usage = -1;
- int64_t avail = -1;
- int64_t blocks = 0;
- quota_inode_ctx_t *ctx = NULL;
- int ret = 0;
-
- inode = cookie;
-
- /* This fop will fail mostly in case of client disconnect,
- * which is already logged. Hence, not logging here */
- if (op_ret == -1)
- goto unwind;
- /*
- * We should never get here unless quota_statfs (below) sent us a
- * cookie, and it would only do so if the value was non-NULL. This
- * check is therefore just routine defensive coding.
- */
-
- GF_VALIDATE_OR_GOTO ("quota", inode, unwind);
-
- inode_ctx_get (inode, this, &value);
- ctx = (quota_inode_ctx_t *)(unsigned long)value;
- if (!ctx || ctx->hard_lim <= 0)
- goto unwind;
-
- { /* statfs is adjusted in this code block */
- usage = (ctx->size) / buf->f_bsize;
-
- blocks = ctx->hard_lim / buf->f_bsize;
- buf->f_blocks = blocks;
-
- avail = buf->f_blocks - usage;
- avail = max (avail, 0);
-
- buf->f_bfree = avail;
- /*
- * We have to assume that the total assigned quota
- * won't cause us to dip into the reserved space,
- * because dealing with the overcommitted cases is
- * just too hairy (especially when different bricks
- * might be using different reserved percentages and
- * such).
- */
- buf->f_bavail = buf->f_bfree;
- }
+ inode_t *inode = NULL;
+ uint64_t value = 0;
+ int64_t usage = -1;
+ int64_t avail = -1;
+ int64_t blocks = 0;
+ quota_inode_ctx_t *ctx = NULL;
+ int ret = 0;
+
+ inode = cookie;
+
+ /* This fop will fail mostly in case of client disconnect,
+ * which is already logged. Hence, not logging here */
+ if (op_ret == -1)
+ goto unwind;
+ /*
+ * We should never get here unless quota_statfs (below) sent us a
+ * cookie, and it would only do so if the value was non-NULL. This
+ * check is therefore just routine defensive coding.
+ */
+
+ GF_VALIDATE_OR_GOTO("quota", inode, unwind);
+
+ inode_ctx_get(inode, this, &value);
+ ctx = (quota_inode_ctx_t *)(unsigned long)value;
+ if (!ctx || ctx->hard_lim <= 0)
+ goto unwind;
+
+ { /* statfs is adjusted in this code block */
+ usage = (ctx->size) / buf->f_bsize;
+
+ blocks = ctx->hard_lim / buf->f_bsize;
+ buf->f_blocks = blocks;
+
+ avail = buf->f_blocks - usage;
+ avail = max(avail, 0);
+
+ buf->f_bfree = avail;
+ /*
+ * We have to assume that the total assigned quota
+ * won't cause us to dip into the reserved space,
+ * because dealing with the overcommitted cases is
+ * just too hairy (especially when different bricks
+ * might be using different reserved percentages and
+ * such).
+ */
+ buf->f_bavail = buf->f_bfree;
+ }
- xdata = xdata ? dict_ref(xdata) : dict_new();
- if (!xdata)
- goto unwind;
+ xdata = xdata ? dict_ref(xdata) : dict_new();
+ if (!xdata)
+ goto unwind;
- ret = dict_set_int8 (xdata, "quota-deem-statfs", 1);
- if (-1 == ret)
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- Q_MSG_ENOMEM, "Dict set failed, deem-statfs option may "
- "have no effect");
+ ret = dict_set_int8(xdata, "quota-deem-statfs", 1);
+ if (-1 == ret)
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, Q_MSG_ENOMEM,
+ "Dict set failed, deem-statfs option may "
+ "have no effect");
unwind:
- QUOTA_STACK_UNWIND (statfs, frame, op_ret, op_errno, buf, xdata);
+ QUOTA_STACK_UNWIND(statfs, frame, op_ret, op_errno, buf, xdata);
- if (xdata)
- dict_unref (xdata);
+ if (xdata)
+ dict_unref(xdata);
- return 0;
+ return 0;
}
-
int32_t
-quota_statfs_helper (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *xdata)
+quota_statfs_helper(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ dict_t *xdata)
{
- quota_local_t *local = frame->local;
- int op_errno = EINVAL;
+ quota_local_t *local = frame->local;
+ int op_errno = EINVAL;
- GF_VALIDATE_OR_GOTO ("quota", local, err);
+ GF_VALIDATE_OR_GOTO("quota", local, err);
- if (-1 == local->op_ret) {
- op_errno = local->op_errno;
- goto err;
- }
+ if (-1 == local->op_ret) {
+ op_errno = local->op_errno;
+ goto err;
+ }
- STACK_WIND_COOKIE (frame, quota_statfs_cbk, local->inode,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->statfs, loc, xdata);
- return 0;
+ STACK_WIND_COOKIE(frame, quota_statfs_cbk, local->inode, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->statfs, loc, xdata);
+ return 0;
err:
- QUOTA_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
+ QUOTA_STACK_UNWIND(statfs, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
int32_t
-quota_statfs_validate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata,
- struct iatt *postparent)
+quota_statfs_validate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
{
- quota_local_t *local = NULL;
- int32_t ret = 0;
- quota_inode_ctx_t *ctx = NULL;
- uint64_t value = 0;
- quota_meta_t size = {0,};
-
- local = frame->local;
-
- if (op_ret < 0)
- goto resume;
-
- GF_ASSERT (local);
- GF_ASSERT (frame);
- GF_VALIDATE_OR_GOTO_WITH_ERROR ("quota", this, resume, op_errno,
- EINVAL);
- GF_VALIDATE_OR_GOTO_WITH_ERROR (this->name, xdata, resume, op_errno,
- EINVAL);
-
- ret = inode_ctx_get (local->validate_loc.inode, this, &value);
-
- ctx = (quota_inode_ctx_t *)(unsigned long)value;
- if ((ret == -1) || (ctx == NULL)) {
- gf_msg (this->name, GF_LOG_WARNING, EINVAL,
- Q_MSG_INODE_CTX_GET_FAILED,
- "quota context is not present in inode (gfid:%s)",
- uuid_utoa (local->validate_loc.inode->gfid));
- op_errno = EINVAL;
- goto resume;
- }
-
- ret = quota_dict_get_meta (xdata, QUOTA_SIZE_KEY, &size);
- if (ret == -1) {
- gf_msg (this->name, GF_LOG_WARNING, EINVAL,
- Q_MSG_SIZE_KEY_MISSING, "size key not present in "
- "dict");
- op_errno = EINVAL;
- }
-
- LOCK (&ctx->lock);
- {
- ctx->size = size.size;
- ctx->file_count = size.file_count;
- ctx->dir_count = size.dir_count;
- gettimeofday (&ctx->tv, NULL);
- }
- UNLOCK (&ctx->lock);
+ quota_local_t *local = NULL;
+ int32_t ret = 0;
+ quota_inode_ctx_t *ctx = NULL;
+ uint64_t value = 0;
+ quota_meta_t size = {
+ 0,
+ };
+
+ local = frame->local;
+
+ if (op_ret < 0)
+ goto resume;
+
+ GF_ASSERT(local);
+ GF_ASSERT(frame);
+ GF_VALIDATE_OR_GOTO_WITH_ERROR("quota", this, resume, op_errno, EINVAL);
+ GF_VALIDATE_OR_GOTO_WITH_ERROR(this->name, xdata, resume, op_errno, EINVAL);
+
+ ret = inode_ctx_get(local->validate_loc.inode, this, &value);
+
+ ctx = (quota_inode_ctx_t *)(unsigned long)value;
+ if ((ret == -1) || (ctx == NULL)) {
+ gf_msg(this->name, GF_LOG_WARNING, EINVAL, Q_MSG_INODE_CTX_GET_FAILED,
+ "quota context is not present in inode (gfid:%s)",
+ uuid_utoa(local->validate_loc.inode->gfid));
+ op_errno = EINVAL;
+ goto resume;
+ }
+
+ ret = quota_dict_get_meta(xdata, QUOTA_SIZE_KEY, SLEN(QUOTA_SIZE_KEY),
+ &size);
+ if (ret == -1) {
+ gf_msg(this->name, GF_LOG_WARNING, EINVAL, Q_MSG_SIZE_KEY_MISSING,
+ "size key not present in "
+ "dict");
+ op_errno = EINVAL;
+ }
+
+ LOCK(&ctx->lock);
+ {
+ ctx->size = size.size;
+ ctx->validate_time = gf_time();
+ ctx->file_count = size.file_count;
+ ctx->dir_count = size.dir_count;
+ }
+ UNLOCK(&ctx->lock);
resume:
- quota_link_count_decrement (frame);
- return 0;
+ local->op_errno = op_errno;
+ quota_link_count_decrement(frame);
+ return 0;
}
void
-quota_get_limit_dir_continuation (struct list_head *parents, inode_t *inode,
- int32_t op_ret, int32_t op_errno, void *data)
+quota_get_limit_dir_continuation(struct list_head *parents, inode_t *inode,
+ int32_t op_ret, int32_t op_errno, void *data)
{
- call_frame_t *frame = NULL;
- xlator_t *this = NULL;
- quota_dentry_t *entry = NULL;
- inode_t *parent = NULL;
-
- frame = data;
- this = THIS;
-
- if ((op_ret < 0) || list_empty (parents)) {
- if (op_ret >= 0) {
- gf_msg (this->name, GF_LOG_WARNING, EIO,
- Q_MSG_ANCESTRY_BUILD_FAILED,
- "Couldn't build ancestry for inode (gfid:%s). "
- "Without knowing ancestors till root, quota "
- "cannot be enforced. "
- "Hence, failing fop with EIO",
- uuid_utoa (inode->gfid));
- op_errno = EIO;
- }
+ call_frame_t *frame = NULL;
+ xlator_t *this = NULL;
+ quota_dentry_t *entry = NULL;
+ inode_t *parent = NULL;
- quota_handle_validate_error (frame, -1, op_errno);
- goto out;
+ frame = data;
+ this = THIS;
+
+ if ((op_ret < 0) || list_empty(parents)) {
+ if (op_ret >= 0) {
+ gf_msg(this->name, GF_LOG_WARNING, EIO, Q_MSG_ANCESTRY_BUILD_FAILED,
+ "Couldn't build ancestry for inode (gfid:%s). "
+ "Without knowing ancestors till root, quota "
+ "cannot be enforced. "
+ "Hence, failing fop with EIO",
+ uuid_utoa(inode->gfid));
+ op_errno = EIO;
}
- entry = list_entry (parents, quota_dentry_t, next);
- parent = inode_find (inode->table, entry->par);
+ quota_handle_validate_error(frame, -1, op_errno);
+ goto out;
+ }
- quota_get_limit_dir (frame, parent, this);
+ entry = list_entry(parents, quota_dentry_t, next);
+ parent = inode_find(inode->table, entry->par);
- inode_unref (parent);
+ quota_get_limit_dir(frame, parent, this);
+
+ inode_unref(parent);
out:
- return;
+ return;
}
void
-quota_statfs_continue (call_frame_t *frame, xlator_t *this, inode_t *inode)
+quota_statfs_continue(call_frame_t *frame, xlator_t *this, inode_t *inode)
{
- quota_local_t *local = frame->local;
- int ret = -1;
-
- LOCK (&local->lock);
- {
- local->inode = inode_ref (inode);
- }
- UNLOCK (&local->lock);
-
- ret = quota_validate (frame, local->inode, this,
- quota_statfs_validate_cbk);
- if (0 > ret)
- quota_handle_validate_error (frame, -1, -ret);
+ quota_local_t *local = frame->local;
+ int ret = -1;
+
+ LOCK(&local->lock);
+ {
+ local->inode = inode_ref(inode);
+ }
+ UNLOCK(&local->lock);
+
+ ret = quota_validate(frame, local->inode, this, quota_statfs_validate_cbk);
+ if (0 > ret)
+ quota_handle_validate_error(frame, -1, -ret);
}
void
-quota_get_limit_dir (call_frame_t *frame, inode_t *cur_inode, xlator_t *this)
+quota_get_limit_dir(call_frame_t *frame, inode_t *cur_inode, xlator_t *this)
{
- inode_t *inode = NULL;
- inode_t *parent = NULL;
- uint64_t value = 0;
- quota_inode_ctx_t *ctx = NULL;
- quota_local_t *local = frame->local;
+ inode_t *inode = NULL;
+ inode_t *parent = NULL;
+ uint64_t value = 0;
+ quota_inode_ctx_t *ctx = NULL;
+ quota_local_t *local = frame->local;
- if (!cur_inode)
- goto out;
+ if (!cur_inode)
+ goto out;
- inode = inode_ref (cur_inode);
- while (inode) {
- value = 0;
- inode_ctx_get (inode, this, &value);
+ inode = inode_ref(cur_inode);
+ while (inode) {
+ value = 0;
+ inode_ctx_get(inode, this, &value);
- if (value) {
- ctx = (quota_inode_ctx_t *)(unsigned long)value;
- if (ctx->hard_lim > 0)
- break;
- }
+ if (value) {
+ ctx = (quota_inode_ctx_t *)(unsigned long)value;
+ if (ctx->hard_lim > 0)
+ break;
+ }
- if (__is_root_gfid (inode->gfid))
- goto off;
+ if (__is_root_gfid(inode->gfid))
+ goto off;
- parent = inode_parent (inode, 0, NULL);
- if (!parent) {
- (void) quota_build_ancestry (inode,
- quota_get_limit_dir_continuation,
- frame);
- goto out;
- }
-
- inode_unref (inode);
- inode = parent;
+ parent = inode_parent(inode, 0, NULL);
+ if (!parent) {
+ (void)quota_build_ancestry(inode, quota_get_limit_dir_continuation,
+ frame);
+ goto out;
}
- quota_statfs_continue (frame, this, inode);
- inode_unref (inode);
- return;
+ inode_unref(inode);
+ inode = parent;
+ }
+
+ quota_statfs_continue(frame, this, inode);
+ inode_unref(inode);
+ return;
off:
- gf_msg_debug (this->name, 0,
- "No limit set on the inode or it's parents.");
+ gf_msg_debug(this->name, 0, "No limit set on the inode or it's parents.");
- QUOTA_STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->statfs,
- &local->loc, local->xdata);
+ QUOTA_STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->statfs, &local->loc,
+ local->xdata);
out:
- inode_unref (inode);
+ inode_unref(inode);
- return;
+ return;
}
int32_t
-quota_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+quota_statfs(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- int op_errno = 0;
- int ret = -1;
- int8_t ignore_deem_statfs = 0;
- quota_priv_t *priv = NULL;
- quota_local_t *local = NULL;
- call_stub_t *stub = NULL;
+ int op_errno = 0;
+ int ret = -1;
+ int8_t ignore_deem_statfs = 0;
+ quota_priv_t *priv = NULL;
+ quota_local_t *local = NULL;
+ call_stub_t *stub = NULL;
- priv = this->private;
- GF_ASSERT (loc);
+ priv = this->private;
+ GF_ASSERT(loc);
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
- ret = dict_get_int8 (xdata, GF_INTERNAL_IGNORE_DEEM_STATFS,
- &ignore_deem_statfs);
- ret = 0;
+ ret = dict_get_int8(xdata, GF_INTERNAL_IGNORE_DEEM_STATFS,
+ &ignore_deem_statfs);
+ ret = 0;
- if (ignore_deem_statfs)
- goto off;
+ if (ignore_deem_statfs)
+ goto off;
- if (priv->consider_statfs && loc->inode) {
- local = quota_local_new ();
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
- frame->local = local;
+ if (priv->consider_statfs && loc->inode) {
+ local = quota_local_new();
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+ frame->local = local;
- ret = loc_copy (&local->loc, loc);
- if (-1 == ret) {
- op_errno = ENOMEM;
- goto err;
- }
+ ret = loc_copy(&local->loc, loc);
+ if (-1 == ret) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- if (xdata)
- local->xdata = dict_ref (xdata);
+ if (xdata)
+ local->xdata = dict_ref(xdata);
- stub = fop_statfs_stub (frame, quota_statfs_helper,
- &local->loc, local->xdata);
- if (!stub) {
- op_errno = ENOMEM;
- goto err;
- }
+ stub = fop_statfs_stub(frame, quota_statfs_helper, &local->loc,
+ local->xdata);
+ if (!stub) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- LOCK (&local->lock);
- {
- local->link_count = 1;
- local->stub = stub;
- }
- UNLOCK (&local->lock);
+ LOCK(&local->lock);
+ {
+ local->link_count = 1;
+ local->stub = stub;
+ }
+ UNLOCK(&local->lock);
- quota_get_limit_dir (frame, loc->inode, this);
+ quota_get_limit_dir(frame, loc->inode, this);
- return 0;
- }
+ return 0;
+ }
- /*
- * We have to make sure that we never get to quota_statfs_cbk
- * with a cookie that points to something other than an inode,
- * which is exactly what would happen with STACK_UNWIND using
- * that as a callback. Therefore, use default_statfs_cbk in
- * this case instead.
- *
- * Also if the option deem-statfs is not set to "on" don't
- * bother calculating quota limit on / in statfs_cbk.
- */
- if (priv->consider_statfs)
- gf_log (this->name, GF_LOG_ERROR,
- "Missing inode, can't adjust for quota");
+ /*
+ * We have to make sure that we never get to quota_statfs_cbk
+ * with a cookie that points to something other than an inode,
+ * which is exactly what would happen with STACK_UNWIND using
+ * that as a callback. Therefore, use default_statfs_cbk in
+ * this case instead.
+ *
+ * Also if the option deem-statfs is not set to "on" don't
+ * bother calculating quota limit on / in statfs_cbk.
+ */
+ if (priv->consider_statfs)
+ gf_log(this->name, GF_LOG_ERROR,
+ "Missing inode, can't adjust for quota");
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->statfs, loc, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->statfs,
+ loc, xdata);
+ return 0;
err:
- QUOTA_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
+ QUOTA_STACK_UNWIND(statfs, frame, -1, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
int
-quota_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, gf_dirent_t *entries,
- dict_t *xdata)
+quota_readdirp_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, gf_dirent_t *entries,
+ dict_t *xdata)
{
- gf_dirent_t *entry = NULL;
- quota_local_t *local = NULL;
- loc_t loc = {0, };
+ gf_dirent_t *entry = NULL;
+ quota_local_t *local = NULL;
+ loc_t loc = {
+ 0,
+ };
- if (op_ret <= 0)
- goto unwind;
+ if (op_ret <= 0)
+ goto unwind;
- local = frame->local;
+ local = frame->local;
- list_for_each_entry (entry, &entries->list, list) {
- if ((strcmp (entry->d_name, ".") == 0) ||
- (strcmp (entry->d_name, "..") == 0) ||
- entry->inode == NULL)
- continue;
+ list_for_each_entry(entry, &entries->list, list)
+ {
+ if ((strcmp(entry->d_name, ".") == 0) ||
+ (strcmp(entry->d_name, "..") == 0) || entry->inode == NULL)
+ continue;
- gf_uuid_copy (loc.gfid, entry->d_stat.ia_gfid);
- loc.inode = inode_ref (entry->inode);
- loc.parent = inode_ref (local->loc.inode);
- gf_uuid_copy (loc.pargfid, loc.parent->gfid);
- loc.name = entry->d_name;
+ gf_uuid_copy(loc.gfid, entry->d_stat.ia_gfid);
+ loc.inode = inode_ref(entry->inode);
+ loc.parent = inode_ref(local->loc.inode);
+ gf_uuid_copy(loc.pargfid, loc.parent->gfid);
+ loc.name = entry->d_name;
- quota_fill_inodectx (this, entry->inode, entry->dict,
- &loc, &entry->d_stat, &op_errno);
+ quota_fill_inodectx(this, entry->inode, entry->dict, &loc,
+ &entry->d_stat, &op_errno);
- loc_wipe (&loc);
- }
+ loc_wipe(&loc);
+ }
unwind:
- QUOTA_STACK_UNWIND (readdirp, frame, op_ret, op_errno, entries, xdata);
+ QUOTA_STACK_UNWIND(readdirp, frame, op_ret, op_errno, entries, xdata);
- return 0;
+ return 0;
}
int
-quota_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, dict_t *dict)
+quota_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *dict)
{
- quota_priv_t *priv = NULL;
- int ret = 0;
- gf_boolean_t new_dict = _gf_false;
- quota_local_t *local = NULL;
+ quota_priv_t *priv = NULL;
+ int ret = 0;
+ gf_boolean_t new_dict = _gf_false;
+ quota_local_t *local = NULL;
- priv = this->private;
+ priv = this->private;
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
- local = quota_local_new ();
+ local = quota_local_new();
- if (local == NULL) {
- goto err;
- }
+ if (local == NULL) {
+ goto err;
+ }
- frame->local = local;
+ frame->local = local;
- local->loc.inode = inode_ref (fd->inode);
+ local->loc.inode = inode_ref(fd->inode);
- if (dict == NULL) {
- dict = dict_new ();
- new_dict = _gf_true;
- }
+ if (dict == NULL) {
+ dict = dict_new();
+ new_dict = _gf_true;
+ }
- if (dict) {
- ret = dict_set_int8 (dict, QUOTA_LIMIT_KEY, 1);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM,
- "dict set of key for hard-limit");
- goto err;
- }
+ if (dict) {
+ ret = dict_set_int8(dict, QUOTA_LIMIT_KEY, 1);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "dict set of key for hard-limit");
+ goto err;
}
+ }
- if (dict) {
- ret = dict_set_int8 (dict, QUOTA_LIMIT_OBJECTS_KEY, 1);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "dict set of key for hard-limit "
- "failed");
- goto err;
- }
+ if (dict) {
+ ret = dict_set_int8(dict, QUOTA_LIMIT_OBJECTS_KEY, 1);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "dict set of key for hard-limit "
+ "failed");
+ goto err;
}
+ }
- STACK_WIND (frame, quota_readdirp_cbk,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->readdirp, fd,
- size, offset, dict);
+ STACK_WIND(frame, quota_readdirp_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdirp, fd, size, offset, dict);
- if (new_dict) {
- dict_unref (dict);
- }
+ if (new_dict) {
+ dict_unref(dict);
+ }
- return 0;
+ return 0;
err:
- STACK_UNWIND_STRICT (readdirp, frame, -1, EINVAL, NULL, NULL);
+ STACK_UNWIND_STRICT(readdirp, frame, -1, EINVAL, NULL, NULL);
- if (new_dict) {
- dict_unref (dict);
- }
+ if (new_dict) {
+ dict_unref(dict);
+ }
- return 0;
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdirp, fd,
- size, offset, dict);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->readdirp,
+ fd, size, offset, dict);
+ return 0;
}
int32_t
@@ -4774,606 +4644,693 @@ quota_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
struct iatt *postbuf, dict_t *xdata)
{
- int32_t ret = 0;
- uint64_t ctx_int = 0;
- quota_inode_ctx_t *ctx = NULL;
- quota_local_t *local = NULL;
-
- local = frame->local;
-
- if ((op_ret < 0) || (local == NULL)) {
- goto out;
- }
-
- ret = inode_ctx_get (local->loc.inode, this, &ctx_int);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- Q_MSG_INODE_CTX_GET_FAILED,
- "%s: failed to get the context", local->loc.path);
- goto out;
- }
-
- ctx = (quota_inode_ctx_t *)(unsigned long) ctx_int;
-
- if (ctx == NULL) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- Q_MSG_INODE_CTX_GET_FAILED,
- "quota context not set in %s (gfid:%s)",
- local->loc.path, uuid_utoa (local->loc.inode->gfid));
- goto out;
- }
-
- LOCK (&ctx->lock);
- {
- ctx->buf = *postbuf;
- }
- UNLOCK (&ctx->lock);
+ int32_t ret = 0;
+ uint64_t ctx_int = 0;
+ quota_inode_ctx_t *ctx = NULL;
+ quota_local_t *local = NULL;
+
+ local = frame->local;
+
+ if ((op_ret < 0) || (local == NULL)) {
+ goto out;
+ }
+
+ ret = inode_ctx_get(local->loc.inode, this, &ctx_int);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, Q_MSG_INODE_CTX_GET_FAILED,
+ "%s: failed to get the context", local->loc.path);
+ goto out;
+ }
+
+ ctx = (quota_inode_ctx_t *)(unsigned long)ctx_int;
+
+ if (ctx == NULL) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, Q_MSG_INODE_CTX_GET_FAILED,
+ "quota context not set in %s (gfid:%s)", local->loc.path,
+ uuid_utoa(local->loc.inode->gfid));
+ goto out;
+ }
+
+ LOCK(&ctx->lock);
+ {
+ ctx->buf = *postbuf;
+ }
+ UNLOCK(&ctx->lock);
out:
- QUOTA_STACK_UNWIND (fallocate, frame, op_ret, op_errno, prebuf, postbuf,
- xdata);
+ QUOTA_STACK_UNWIND(fallocate, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
- return 0;
+ return 0;
}
-
int32_t
-quota_fallocate_helper (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int32_t mode, off_t offset, size_t len, dict_t *xdata)
+quota_fallocate_helper(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ int32_t mode, off_t offset, size_t len, dict_t *xdata)
{
- quota_local_t *local = NULL;
- int32_t op_errno = EINVAL;
-
- local = frame->local;
-
- GF_VALIDATE_OR_GOTO ("quota", local, unwind);
-
- if (local->op_ret == -1) {
- op_errno = local->op_errno;
- if (op_errno == ENOENT || op_errno == ESTALE) {
- /* We may get ENOENT/ESTALE in case of below scenario
- * fd = open file.txt
- * unlink file.txt
- * fallocate on fd
- * Here build_ancestry can fail as the file is removed.
- * For now ignore ENOENT/ESTALE on active fd
- * We need to re-visit this code once we understand
- * how other file-system behave in this scenario
- */
- gf_msg_debug (this->name, 0, "quota enforcer failed "
- "with ENOENT/ESTALE on %s, cannot check "
- "quota limits and allowing fallocate",
- uuid_utoa (fd->inode->gfid));
- } else {
- goto unwind;
- }
+ quota_local_t *local = NULL;
+ int32_t op_errno = EINVAL;
+
+ local = frame->local;
+
+ GF_VALIDATE_OR_GOTO("quota", local, unwind);
+
+ if (local->op_ret == -1) {
+ op_errno = local->op_errno;
+ if (op_errno == ENOENT || op_errno == ESTALE) {
+ /* We may get ENOENT/ESTALE in case of below scenario
+ * fd = open file.txt
+ * unlink file.txt
+ * fallocate on fd
+ * Here build_ancestry can fail as the file is removed.
+ * For now ignore ENOENT/ESTALE on active fd
+ * We need to re-visit this code once we understand
+ * how other file-system behave in this scenario
+ */
+ gf_msg_debug(this->name, 0,
+ "quota enforcer failed "
+ "with ENOENT/ESTALE on %s, cannot check "
+ "quota limits and allowing fallocate",
+ uuid_utoa(fd->inode->gfid));
+ } else {
+ goto unwind;
}
+ }
- STACK_WIND (frame, quota_fallocate_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fallocate, fd, mode, offset, len,
- xdata);
- return 0;
+ STACK_WIND(frame, quota_fallocate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fallocate, fd, mode, offset, len,
+ xdata);
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(fallocate, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
}
-
int32_t
quota_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
- off_t offset, size_t len, dict_t *xdata)
+ off_t offset, size_t len, dict_t *xdata)
{
- int32_t op_errno = EINVAL;
- int32_t parents = 0;
- int32_t fail_count = 0;
- quota_local_t *local = NULL;
- quota_inode_ctx_t *ctx = NULL;
- quota_priv_t *priv = NULL;
- quota_dentry_t *dentry = NULL;
- quota_dentry_t *tmp = NULL;
- call_stub_t *stub = NULL;
- struct list_head head = {0, };
- inode_t *par_inode = NULL;
-
- priv = this->private;
- GF_VALIDATE_OR_GOTO (this->name, priv, unwind);
-
- WIND_IF_QUOTAOFF (priv->is_quota_on, off);
-
- INIT_LIST_HEAD (&head);
-
- GF_ASSERT (frame);
- GF_VALIDATE_OR_GOTO ("quota", this, unwind);
- GF_VALIDATE_OR_GOTO (this->name, fd, unwind);
-
- local = quota_local_new ();
- if (local == NULL) {
- goto unwind;
- }
-
- frame->local = local;
- local->loc.inode = inode_ref (fd->inode);
-
- (void) quota_inode_ctx_get (fd->inode, this, &ctx, 0);
- if (ctx == NULL) {
- gf_msg_debug (this->name, 0, "quota context is NULL on inode"
- " (%s). If quota is not enabled recently and "
- "crawler has finished crawling, its an error",
- uuid_utoa (local->loc.inode->gfid));
- }
-
- stub = fop_fallocate_stub(frame, quota_fallocate_helper, fd, mode,
- offset, len, xdata);
- if (stub == NULL) {
- op_errno = ENOMEM;
- goto unwind;
+ int32_t op_errno = EINVAL;
+ int32_t parents = 0;
+ int32_t fail_count = 0;
+ quota_local_t *local = NULL;
+ quota_inode_ctx_t *ctx = NULL;
+ quota_priv_t *priv = NULL;
+ quota_dentry_t *dentry = NULL;
+ quota_dentry_t *tmp = NULL;
+ call_stub_t *stub = NULL;
+ struct list_head head = {
+ 0,
+ };
+ inode_t *par_inode = NULL;
+
+ priv = this->private;
+ GF_VALIDATE_OR_GOTO(this->name, priv, unwind);
+
+ WIND_IF_QUOTAOFF(priv->is_quota_on, off);
+
+ INIT_LIST_HEAD(&head);
+
+ GF_ASSERT(frame);
+ GF_VALIDATE_OR_GOTO("quota", this, unwind);
+ GF_VALIDATE_OR_GOTO(this->name, fd, unwind);
+
+ local = quota_local_new();
+ if (local == NULL) {
+ goto unwind;
+ }
+
+ frame->local = local;
+ local->loc.inode = inode_ref(fd->inode);
+
+ (void)quota_inode_ctx_get(fd->inode, this, &ctx, 0);
+ if (ctx == NULL) {
+ gf_msg_debug(this->name, 0,
+ "quota context is NULL on inode"
+ " (%s). If quota is not enabled recently and "
+ "crawler has finished crawling, its an error",
+ uuid_utoa(local->loc.inode->gfid));
+ }
+
+ stub = fop_fallocate_stub(frame, quota_fallocate_helper, fd, mode, offset,
+ len, xdata);
+ if (stub == NULL) {
+ op_errno = ENOMEM;
+ goto unwind;
+ }
+
+ priv = this->private;
+ GF_VALIDATE_OR_GOTO(this->name, priv, unwind);
+
+ parents = quota_add_parents_from_ctx(ctx, &head);
+ if (parents == -1) {
+ op_errno = errno;
+ goto unwind;
+ }
+
+ /*
+ * Note that by using len as the delta we're assuming the range from
+ * offset to offset+len has not already been allocated. This can result
+ * in ENOSPC errors attempting to allocate an already allocated range.
+ */
+ local->delta = len;
+ local->object_delta = 0;
+ local->stub = stub;
+ local->link_count = parents;
+
+ if (parents == 0) {
+ local->link_count = 1;
+ quota_check_limit(frame, fd->inode, this);
+ } else {
+ list_for_each_entry_safe(dentry, tmp, &head, next)
+ {
+ par_inode = do_quota_check_limit(frame, fd->inode, this, dentry,
+ _gf_false);
+ if (par_inode == NULL) {
+ /* remove stale entry from inode_ctx */
+ quota_dentry_del(ctx, dentry->name, dentry->par);
+ parents--;
+ fail_count++;
+ } else {
+ inode_unref(par_inode);
+ }
+ __quota_dentry_free(dentry);
}
- priv = this->private;
- GF_VALIDATE_OR_GOTO (this->name, priv, unwind);
-
- parents = quota_add_parents_from_ctx (ctx, &head);
-
- /*
- * Note that by using len as the delta we're assuming the range from
- * offset to offset+len has not already been allocated. This can result
- * in ENOSPC errors attempting to allocate an already allocated range.
- */
- local->delta = len;
- local->object_delta = 0;
- local->stub = stub;
- local->link_count = parents;
-
if (parents == 0) {
- local->link_count = 1;
- quota_check_limit (frame, fd->inode, this);
- } else {
- list_for_each_entry_safe (dentry, tmp, &head, next) {
- par_inode = do_quota_check_limit (frame, fd->inode,
- this, dentry,
- _gf_false);
- if (par_inode == NULL) {
- /* remove stale entry from inode_ctx */
- quota_dentry_del (ctx, dentry->name,
- dentry->par);
- parents--;
- fail_count++;
- } else {
- inode_unref (par_inode);
- }
- __quota_dentry_free (dentry);
- }
-
- if (parents == 0) {
- LOCK (&local->lock);
- {
- local->link_count++;
- }
- UNLOCK (&local->lock);
- quota_check_limit (frame, fd->inode, this);
- }
+ LOCK(&local->lock);
+ {
+ local->link_count++;
+ }
+ UNLOCK(&local->lock);
+ quota_check_limit(frame, fd->inode, this);
+ }
- while (fail_count != 0) {
- quota_link_count_decrement (frame);
- fail_count--;
- }
+ while (fail_count != 0) {
+ quota_link_count_decrement(frame);
+ fail_count--;
}
+ }
- return 0;
+ return 0;
unwind:
- QUOTA_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
- return 0;
+ QUOTA_STACK_UNWIND(fallocate, frame, -1, op_errno, NULL, NULL, NULL);
+ return 0;
off:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fallocate, fd, mode, offset,
- len, xdata);
- return 0;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fallocate, fd, mode, offset, len,
+ xdata);
+ return 0;
}
void
-quota_log_helper (char **usage_str, int64_t cur_size, inode_t *inode,
- char **path, struct timeval *cur_time)
+quota_log_helper(char **usage_str, int64_t cur_size, inode_t *inode,
+ char **path, time_t *cur_time)
{
- xlator_t *this = THIS;
+ xlator_t *this = THIS;
- if (!usage_str || !inode || !path || !cur_time) {
- gf_log (this->name, GF_LOG_ERROR, "Received null argument");
- return;
- }
+ if (!usage_str || !inode || !path || !cur_time) {
+ gf_log(this->name, GF_LOG_ERROR, "Received null argument");
+ return;
+ }
- *usage_str = gf_uint64_2human_readable (cur_size);
- if (!(*usage_str))
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM, Q_MSG_ENOMEM,
- "integer to string conversion failed Reason"
- ":\"Cannot allocate memory\"");
+ *usage_str = gf_uint64_2human_readable(cur_size);
+ if (!(*usage_str))
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, Q_MSG_ENOMEM,
+ "integer to string conversion failed Reason"
+ ":\"Cannot allocate memory\"");
- inode_path (inode, NULL, path);
- if (!(*path))
- *path = uuid_utoa (inode->gfid);
+ inode_path(inode, NULL, path);
+ if (!(*path))
+ *path = uuid_utoa(inode->gfid);
- gettimeofday (cur_time, NULL);
+ *cur_time = gf_time();
}
/* Logs if
-* i. Usage crossed soft limit
-* ii. Usage above soft limit and alert-time elapsed
-*/
+ * i. Usage crossed soft limit
+ * ii. Usage above soft limit and alert-time elapsed
+ */
void
-quota_log_usage (xlator_t *this, quota_inode_ctx_t *ctx, inode_t *inode,
- int64_t delta)
+quota_log_usage(xlator_t *this, quota_inode_ctx_t *ctx, inode_t *inode,
+ int64_t delta)
{
- struct timeval cur_time = {0,};
- char *usage_str = NULL;
- char *path = NULL;
- int64_t cur_size = 0;
- quota_priv_t *priv = NULL;
+ time_t cur_time = 0;
+ char *usage_str = NULL;
+ char *path = NULL;
+ int64_t cur_size = 0;
+ quota_priv_t *priv = NULL;
- priv = this->private;
- cur_size = ctx->size + delta;
+ priv = this->private;
+ cur_size = ctx->size + delta;
- if ((ctx->soft_lim <= 0) || cur_size < ctx->soft_lim)
- return;
+ if ((ctx->soft_lim <= 0) || cur_size < ctx->soft_lim)
+ return;
- /* Usage crossed/reached soft limit */
- if (DID_REACH_LIMIT (ctx->soft_lim, ctx->size, cur_size)) {
+ /* Usage crossed/reached soft limit */
+ if (DID_REACH_LIMIT(ctx->soft_lim, ctx->size, cur_size)) {
+ quota_log_helper(&usage_str, cur_size, inode, &path, &cur_time);
- quota_log_helper (&usage_str, cur_size, inode,
- &path, &cur_time);
+ gf_msg(this->name, GF_LOG_ALERT, 0, Q_MSG_CROSSED_SOFT_LIMIT,
+ "Usage crossed soft limit: "
+ "%s used by %s",
+ usage_str, path);
- gf_msg (this->name, GF_LOG_ALERT, 0,
- Q_MSG_CROSSED_SOFT_LIMIT, "Usage crossed soft limit: "
- "%s used by %s", usage_str, path);
+ gf_event(EVENT_QUOTA_CROSSED_SOFT_LIMIT,
+ "Usage=%s;volume=%s;"
+ "path=%s",
+ usage_str, priv->volume_uuid, path);
- gf_event (EVENT_QUOTA_CROSSED_SOFT_LIMIT, "Usage=%s;volume=%s;"
- "path=%s", usage_str, priv->volume_uuid, path);
+ ctx->prev_log_time = cur_time;
- ctx->prev_log = cur_time;
+ }
+ /* Usage is above soft limit */
+ else if (cur_size > ctx->soft_lim &&
+ quota_timeout(ctx->prev_log_time, priv->log_timeout)) {
+ quota_log_helper(&usage_str, cur_size, inode, &path, &cur_time);
- }
- /* Usage is above soft limit */
- else if (cur_size > ctx->soft_lim &&
- quota_timeout (&ctx->prev_log, priv->log_timeout)) {
+ gf_msg(this->name, GF_LOG_ALERT, 0, Q_MSG_CROSSED_SOFT_LIMIT,
+ "Usage is above soft limit: %s used by %s", usage_str, path);
- quota_log_helper (&usage_str, cur_size, inode,
- &path, &cur_time);
+ gf_event(EVENT_QUOTA_CROSSED_SOFT_LIMIT,
+ "Usage=%s;volume=%s;"
+ "path=%s",
+ usage_str, priv->volume_uuid, path);
- gf_msg (this->name, GF_LOG_ALERT, 0, Q_MSG_CROSSED_SOFT_LIMIT,
- "Usage is above soft limit: %s used by %s",
- usage_str, path);
+ ctx->prev_log_time = cur_time;
+ }
- gf_event (EVENT_QUOTA_CROSSED_SOFT_LIMIT, "Usage=%s;volume=%s;"
- "path=%s", usage_str, priv->volume_uuid, path);
+ if (path)
+ GF_FREE(path);
- ctx->prev_log = cur_time;
- }
-
- if (usage_str)
- GF_FREE (usage_str);
+ if (usage_str)
+ GF_FREE(usage_str);
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
-
- if (!this)
- return ret;
+ int ret = -1;
- ret = xlator_mem_acct_init (this, gf_quota_mt_end + 1);
+ if (!this)
+ return ret;
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
- "Memory accounting init failed");
- return ret;
- }
+ ret = xlator_mem_acct_init(this, gf_quota_mt_end + 1);
+ if (ret != 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "Memory accounting init failed");
return ret;
-}
+ }
+ return ret;
+}
int32_t
-quota_forget (xlator_t *this, inode_t *inode)
+quota_forget(xlator_t *this, inode_t *inode)
{
- int32_t ret = 0;
- uint64_t ctx_int = 0;
- quota_inode_ctx_t *ctx = NULL;
- quota_dentry_t *dentry = NULL, *tmp;
+ int32_t ret = 0;
+ uint64_t ctx_int = 0;
+ quota_inode_ctx_t *ctx = NULL;
+ quota_dentry_t *dentry = NULL, *tmp;
- ret = inode_ctx_del (inode, this, &ctx_int);
+ ret = inode_ctx_del(inode, this, &ctx_int);
- if (ret < 0) {
- return 0;
- }
+ if (ret < 0) {
+ return 0;
+ }
- ctx = (quota_inode_ctx_t *) (long)ctx_int;
+ ctx = (quota_inode_ctx_t *)(long)ctx_int;
- LOCK (&ctx->lock);
+ LOCK(&ctx->lock);
+ {
+ list_for_each_entry_safe(dentry, tmp, &ctx->parents, next)
{
- list_for_each_entry_safe (dentry, tmp, &ctx->parents, next) {
- __quota_dentry_free (dentry);
- }
+ __quota_dentry_free(dentry);
}
- UNLOCK (&ctx->lock);
+ }
+ UNLOCK(&ctx->lock);
- LOCK_DESTROY (&ctx->lock);
+ LOCK_DESTROY(&ctx->lock);
- GF_FREE (ctx);
+ GF_FREE(ctx);
- return 0;
+ return 0;
}
-int32_t
-init (xlator_t *this)
+int
+notify(xlator_t *this, int event, void *data, ...)
{
- int32_t ret = -1;
- quota_priv_t *priv = NULL;
- rpc_clnt_t *rpc = NULL;
-
- if ((this->children == NULL)
- || this->children->next) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- Q_MSG_INVALID_VOLFILE,
- "FATAL: quota (%s) not configured with "
- "exactly one child", this->name);
- return -1;
- }
-
- if (this->parents == NULL) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- Q_MSG_INVALID_VOLFILE,
- "dangling volume. check volfile");
- }
-
- QUOTA_ALLOC_OR_GOTO (priv, quota_priv_t, err);
-
- LOCK_INIT (&priv->lock);
-
- this->private = priv;
-
- GF_OPTION_INIT ("deem-statfs", priv->consider_statfs, bool, err);
- GF_OPTION_INIT ("server-quota", priv->is_quota_on, bool, err);
- GF_OPTION_INIT ("default-soft-limit", priv->default_soft_lim, percent,
- err);
- GF_OPTION_INIT ("soft-timeout", priv->soft_timeout, time, err);
- GF_OPTION_INIT ("hard-timeout", priv->hard_timeout, time, err);
- GF_OPTION_INIT ("alert-time", priv->log_timeout, time, err);
- GF_OPTION_INIT ("volume-uuid", priv->volume_uuid, str, err);
-
- this->local_pool = mem_pool_new (quota_local_t, 64);
- if (!this->local_pool) {
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM,
- Q_MSG_ENOMEM, "failed to create local_t's memory pool");
- goto err;
+ quota_priv_t *priv = NULL;
+ int ret = 0;
+ rpc_clnt_t *rpc = NULL;
+ gf_boolean_t conn_status = _gf_true;
+ xlator_t *victim = data;
+
+ priv = this->private;
+ if (!priv || !priv->is_quota_on)
+ goto out;
+
+ if (event == GF_EVENT_PARENT_DOWN) {
+ rpc = priv->rpc_clnt;
+ if (rpc) {
+ rpc_clnt_disable(rpc);
+ pthread_mutex_lock(&priv->conn_mutex);
+ {
+ conn_status = priv->conn_status;
+ while (conn_status) {
+ (void)pthread_cond_wait(&priv->conn_cond,
+ &priv->conn_mutex);
+ conn_status = priv->conn_status;
+ }
+ }
+ pthread_mutex_unlock(&priv->conn_mutex);
+ gf_log(this->name, GF_LOG_INFO,
+ "Notify GF_EVENT_PARENT_DOWN for brick %s", victim->name);
}
+ }
- if (priv->is_quota_on) {
- rpc = quota_enforcer_init (this, this->options);
- if (rpc == NULL) {
- ret = -1;
- gf_msg (this->name, GF_LOG_WARNING, 0,
- Q_MSG_QUOTA_ENFORCER_RPC_INIT_FAILED,
- "quota enforcer rpc init failed");
- goto err;
- }
+out:
+ ret = default_notify(this, event, data);
+ return ret;
+}
- LOCK (&priv->lock);
- {
- priv->rpc_clnt = rpc;
- }
- UNLOCK (&priv->lock);
+int32_t
+init(xlator_t *this)
+{
+ int32_t ret = -1;
+ quota_priv_t *priv = NULL;
+ rpc_clnt_t *rpc = NULL;
+
+ if ((this->children == NULL) || this->children->next) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, Q_MSG_INVALID_VOLFILE,
+ "FATAL: quota (%s) not configured with "
+ "exactly one child",
+ this->name);
+ return -1;
+ }
+
+ if (this->parents == NULL) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, Q_MSG_INVALID_VOLFILE,
+ "dangling volume. check volfile");
+ }
+
+ QUOTA_ALLOC_OR_GOTO(priv, quota_priv_t, err);
+
+ LOCK_INIT(&priv->lock);
+
+ this->private = priv;
+
+ GF_OPTION_INIT("deem-statfs", priv->consider_statfs, bool, err);
+ GF_OPTION_INIT("server-quota", priv->is_quota_on, bool, err);
+ GF_OPTION_INIT("default-soft-limit", priv->default_soft_lim, percent, err);
+ GF_OPTION_INIT("soft-timeout", priv->soft_timeout, time, err);
+ GF_OPTION_INIT("hard-timeout", priv->hard_timeout, time, err);
+ GF_OPTION_INIT("alert-time", priv->log_timeout, time, err);
+ GF_OPTION_INIT("volume-uuid", priv->volume_uuid, str, err);
+
+ this->local_pool = mem_pool_new(quota_local_t, 64);
+ if (!this->local_pool) {
+ ret = -1;
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, Q_MSG_ENOMEM,
+ "failed to create local_t's memory pool");
+ goto err;
+ }
+
+ pthread_mutex_init(&priv->conn_mutex, NULL);
+ pthread_cond_init(&priv->conn_cond, NULL);
+ priv->conn_status = _gf_false;
+
+ if (priv->is_quota_on) {
+ rpc = quota_enforcer_init(this, this->options);
+ if (rpc == NULL) {
+ ret = -1;
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ Q_MSG_QUOTA_ENFORCER_RPC_INIT_FAILED,
+ "quota enforcer rpc init failed");
+ goto err;
+ }
+
+ LOCK(&priv->lock);
+ {
+ priv->rpc_clnt = rpc;
}
+ UNLOCK(&priv->lock);
+ }
- ret = 0;
+ ret = 0;
err:
- return ret;
+ return ret;
}
int
-reconfigure (xlator_t *this, dict_t *options)
+reconfigure(xlator_t *this, dict_t *options)
{
- int32_t ret = -1;
- quota_priv_t *priv = NULL;
- gf_boolean_t quota_on = _gf_false;
- rpc_clnt_t *rpc = NULL;
-
- priv = this->private;
-
- GF_OPTION_RECONF ("deem-statfs", priv->consider_statfs, options, bool,
- out);
- GF_OPTION_RECONF ("server-quota", quota_on, options, bool,
- out);
- GF_OPTION_RECONF ("default-soft-limit", priv->default_soft_lim,
- options, percent, out);
- GF_OPTION_RECONF ("alert-time", priv->log_timeout, options,
- time, out);
- GF_OPTION_RECONF ("soft-timeout", priv->soft_timeout, options,
- time, out);
- GF_OPTION_RECONF ("hard-timeout", priv->hard_timeout, options,
- time, out);
-
- if (quota_on) {
- priv->rpc_clnt = quota_enforcer_init (this,
- this->options);
- if (priv->rpc_clnt == NULL) {
- ret = -1;
- gf_msg (this->name, GF_LOG_WARNING, 0,
- Q_MSG_QUOTA_ENFORCER_RPC_INIT_FAILED,
- "quota enforcer rpc init failed");
- goto out;
- }
+ int32_t ret = -1;
+ quota_priv_t *priv = NULL;
+ gf_boolean_t quota_on = _gf_false;
+ rpc_clnt_t *rpc = NULL;
+
+ priv = this->private;
+
+ GF_OPTION_RECONF("deem-statfs", priv->consider_statfs, options, bool, out);
+ GF_OPTION_RECONF("server-quota", quota_on, options, bool, out);
+ GF_OPTION_RECONF("default-soft-limit", priv->default_soft_lim, options,
+ percent, out);
+ GF_OPTION_RECONF("alert-time", priv->log_timeout, options, time, out);
+ GF_OPTION_RECONF("soft-timeout", priv->soft_timeout, options, time, out);
+ GF_OPTION_RECONF("hard-timeout", priv->hard_timeout, options, time, out);
+
+ if (quota_on) {
+ priv->rpc_clnt = quota_enforcer_init(this, this->options);
+ if (priv->rpc_clnt == NULL) {
+ ret = -1;
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ Q_MSG_QUOTA_ENFORCER_RPC_INIT_FAILED,
+ "quota enforcer rpc init failed");
+ goto out;
+ }
+
+ } else {
+ LOCK(&priv->lock);
+ {
+ rpc = priv->rpc_clnt;
+ priv->rpc_clnt = NULL;
+ }
+ UNLOCK(&priv->lock);
- } else {
- LOCK (&priv->lock);
- {
- rpc = priv->rpc_clnt;
- priv->rpc_clnt = NULL;
- }
- UNLOCK (&priv->lock);
-
- if (rpc != NULL) {
- // Quotad is shutdown when there is no started volume
- // which has quota enabled. So, we should disable the
- // enforcer client when quota is disabled on a volume,
- // to avoid spurious reconnect attempts to a service
- // (quotad), that is known to be down.
- rpc_clnt_unref (rpc);
- }
+ if (rpc != NULL) {
+ // Quotad is shutdown when there is no started volume
+ // which has quota enabled. So, we should disable the
+ // enforcer client when quota is disabled on a volume,
+ // to avoid spurious reconnect attempts to a service
+ // (quotad), that is known to be down.
+ rpc_clnt_unref(rpc);
}
+ }
- priv->is_quota_on = quota_on;
+ priv->is_quota_on = quota_on;
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int32_t
-quota_priv_dump (xlator_t *this)
+quota_priv_dump(xlator_t *this)
{
- quota_priv_t *priv = NULL;
- int32_t ret = -1;
-
-
- GF_ASSERT (this);
-
- priv = this->private;
-
- gf_proc_dump_add_section ("xlators.features.quota.priv", this->name);
-
- ret = TRY_LOCK (&priv->lock);
- if (ret)
- goto out;
- else {
- gf_proc_dump_write("soft-timeout", "%d", priv->soft_timeout);
- gf_proc_dump_write("hard-timeout", "%d", priv->hard_timeout);
- gf_proc_dump_write("alert-time", "%d", priv->log_timeout);
- gf_proc_dump_write("quota-on", "%d", priv->is_quota_on);
- gf_proc_dump_write("statfs", "%d", priv->consider_statfs);
- gf_proc_dump_write("volume-uuid", "%s", priv->volume_uuid);
- gf_proc_dump_write("validation-count", "%ld",
- priv->validation_count);
- }
- UNLOCK (&priv->lock);
+ quota_priv_t *priv = NULL;
+ int32_t ret = -1;
+
+ GF_ASSERT(this);
+
+ priv = this->private;
+ if (!priv)
+ goto out;
+
+ gf_proc_dump_add_section("xlators.features.quota.priv");
+
+ ret = TRY_LOCK(&priv->lock);
+ if (ret)
+ goto out;
+ else {
+ gf_proc_dump_write("soft-timeout", "%u", priv->soft_timeout);
+ gf_proc_dump_write("hard-timeout", "%u", priv->hard_timeout);
+ gf_proc_dump_write("alert-time", "%u", priv->log_timeout);
+ gf_proc_dump_write("quota-on", "%d", priv->is_quota_on);
+ gf_proc_dump_write("statfs", "%d", priv->consider_statfs);
+ gf_proc_dump_write("volume-uuid", "%s", priv->volume_uuid);
+ gf_proc_dump_write("validation-count", "%" PRIu64,
+ priv->validation_count);
+ }
+ UNLOCK(&priv->lock);
out:
- return 0;
+ return 0;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
+ quota_priv_t *priv = NULL;
+ rpc_clnt_t *rpc = NULL;
+
+ priv = this->private;
+ if (!priv)
return;
+ rpc = priv->rpc_clnt;
+ priv->rpc_clnt = NULL;
+ if (rpc) {
+ rpc_clnt_connection_cleanup(&rpc->conn);
+ rpc_clnt_unref(rpc);
+ }
+
+ this->private = NULL;
+ LOCK_DESTROY(&priv->lock);
+ pthread_mutex_destroy(&priv->conn_mutex);
+ pthread_cond_destroy(&priv->conn_cond);
+
+ GF_FREE(priv);
+ if (this->local_pool) {
+ mem_pool_destroy(this->local_pool);
+ this->local_pool = NULL;
+ }
+ return;
}
-
struct xlator_fops fops = {
- .statfs = quota_statfs,
- .lookup = quota_lookup,
- .writev = quota_writev,
- .create = quota_create,
- .mkdir = quota_mkdir,
- .truncate = quota_truncate,
- .ftruncate = quota_ftruncate,
- .unlink = quota_unlink,
- .symlink = quota_symlink,
- .link = quota_link,
- .rename = quota_rename,
- .getxattr = quota_getxattr,
- .fgetxattr = quota_fgetxattr,
- .stat = quota_stat,
- .fstat = quota_fstat,
- .readlink = quota_readlink,
- .readv = quota_readv,
- .fsync = quota_fsync,
- .setattr = quota_setattr,
- .fsetattr = quota_fsetattr,
- .mknod = quota_mknod,
- .setxattr = quota_setxattr,
- .fsetxattr = quota_fsetxattr,
- .removexattr = quota_removexattr,
- .fremovexattr = quota_fremovexattr,
- .readdirp = quota_readdirp,
- .fallocate = quota_fallocate,
+ .statfs = quota_statfs,
+ .lookup = quota_lookup,
+ .writev = quota_writev,
+ .create = quota_create,
+ .mkdir = quota_mkdir,
+ .truncate = quota_truncate,
+ .ftruncate = quota_ftruncate,
+ .unlink = quota_unlink,
+ .symlink = quota_symlink,
+ .link = quota_link,
+ .rename = quota_rename,
+ .getxattr = quota_getxattr,
+ .fgetxattr = quota_fgetxattr,
+ .stat = quota_stat,
+ .fstat = quota_fstat,
+ .readlink = quota_readlink,
+ .readv = quota_readv,
+ .fsync = quota_fsync,
+ .setattr = quota_setattr,
+ .fsetattr = quota_fsetattr,
+ .mknod = quota_mknod,
+ .setxattr = quota_setxattr,
+ .fsetxattr = quota_fsetxattr,
+ .removexattr = quota_removexattr,
+ .fremovexattr = quota_fremovexattr,
+ .readdirp = quota_readdirp,
+ .fallocate = quota_fallocate,
};
-struct xlator_cbks cbks = {
- .forget = quota_forget
-};
+struct xlator_cbks cbks = {.forget = quota_forget};
struct xlator_dumpops dumpops = {
- .priv = quota_priv_dump,
+ .priv = quota_priv_dump,
};
struct volume_options options[] = {
- {.key = {"deem-statfs"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .description = "If set to on, it takes quota limits into"
- " consideration while estimating fs size. (df command)"
- " (Default is on).",
- .op_version = {2},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {},
- },
- {.key = {"server-quota"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- .description = "Skip the quota enforcement if the feature is"
- " not turned on. This is not a user exposed option.",
- .flags = OPT_FLAG_NONE,
- },
- {.key = {"default-soft-limit"},
- .type = GF_OPTION_TYPE_PERCENT,
- .default_value = "80%",
- .op_version = {3},
- .description = "Soft limit is expressed as a proportion of hard limit."
- " Default-soft-limit is the proportion used when the "
- " user does not supply any soft limit value.",
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {},
- },
- {.key = {"soft-timeout"},
- .type = GF_OPTION_TYPE_TIME,
- .min = 0,
- .max = 1800,
- .default_value = "60",
- .description = "quota caches the directory sizes on client. "
- "soft-timeout indicates the timeout for the validity of"
- " cache before soft-limit has been crossed.",
- .op_version = {3},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {},
- },
- {.key = {"hard-timeout"},
- .type = GF_OPTION_TYPE_TIME,
- .min = 0,
- .max = 60,
- .default_value = "5",
- .description = "quota caches the directory sizes on client. "
- "hard-timeout indicates the timeout for the validity of"
- " cache after soft-limit has been crossed.",
- .op_version = {3},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .tags = {},
- },
- { .key = {"volume-uuid"},
- .type = GF_OPTION_TYPE_STR,
- .default_value = "{{ volume.id }}",
- .description = "uuid of the volume this brick is part of."
- },
- { .key = {"alert-time"},
- .type = GF_OPTION_TYPE_TIME,
- .min = 0,
- .max = 7*86400,
- .default_value = "86400",
- .op_version = {3},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
- .description = "Frequency of limit breach messages in log.",
- .tags = {},
- },
- {.key = {NULL}}
+ {
+ .key = {"enable"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "enable is the volume option that can be used "
+ "to turn on quota.",
+ .op_version = {1},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .level = OPT_STATUS_BASIC,
+ .tags = {},
+ },
+ {
+ .key = {"deem-statfs"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "If set to on, it takes quota limits into"
+ " consideration while estimating fs size. (df command)"
+ " (Default is on).",
+ .op_version = {2},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {},
+ },
+ {
+ .key = {"server-quota"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "Skip the quota enforcement if the feature is"
+ " not turned on. This is not a user exposed option.",
+ .flags = OPT_FLAG_NONE,
+ },
+ {
+ .key = {"default-soft-limit"},
+ .type = GF_OPTION_TYPE_PERCENT,
+ .default_value = "80%",
+ .op_version = {3},
+ .description = "Soft limit is expressed as a proportion of hard limit."
+ " Default-soft-limit is the proportion used when the "
+ " user does not supply any soft limit value.",
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {},
+ },
+ {
+ .key = {"soft-timeout"},
+ .type = GF_OPTION_TYPE_TIME,
+ .min = 0,
+ .max = 1800,
+ .default_value = "60",
+ .description = "quota caches the directory sizes on client. "
+ "soft-timeout indicates the timeout for the validity of"
+ " cache before soft-limit has been crossed.",
+ .op_version = {3},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {},
+ },
+ {
+ .key = {"hard-timeout"},
+ .type = GF_OPTION_TYPE_TIME,
+ .min = 0,
+ .max = 60,
+ .default_value = "5",
+ .description = "quota caches the directory sizes on client. "
+ "hard-timeout indicates the timeout for the validity of"
+ " cache after soft-limit has been crossed.",
+ .op_version = {3},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .tags = {},
+ },
+ {.key = {"volume-uuid"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = "{{ volume.id }}",
+ .description = "uuid of the volume this brick is part of."},
+ {
+ .key = {"alert-time"},
+ .type = GF_OPTION_TYPE_TIME,
+ .min = 0,
+ .max = 7 * 86400,
+ .default_value = "86400",
+ .op_version = {3},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC,
+ .description = "Frequency of limit breach messages in log.",
+ .tags = {},
+ },
+ {.key = {NULL}}};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .notify = notify,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "quota",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/features/quota/src/quota.h b/xlators/features/quota/src/quota.h
index b99ca820e7f..0395d78c9ef 100644
--- a/xlators/features/quota/src/quota.h
+++ b/xlators/features/quota/src/quota.h
@@ -10,267 +10,257 @@
#ifndef _QUOTA_H
#define _QUOTA_H
-#include "xlator.h"
-#include "call-stub.h"
-#include "defaults.h"
-#include "common-utils.h"
+#include <glusterfs/call-stub.h>
#include "quota-mem-types.h"
-#include "glusterfs.h"
-#include "compat.h"
-#include "logging.h"
-#include "dict.h"
-#include "event.h"
+#include <glusterfs/glusterfs.h>
+#include <glusterfs/compat.h>
+#include <glusterfs/logging.h>
+#include <glusterfs/dict.h>
+#include <glusterfs/gf-event.h>
#include "rpcsvc.h"
#include "rpc-clnt.h"
-#include "byte-order.h"
+#include <glusterfs/byte-order.h>
#include "glusterfs3-xdr.h"
#include "glusterfs3.h"
#include "xdr-generic.h"
-#include "compat-errno.h"
+#include <glusterfs/compat-errno.h>
#include "protocol-common.h"
-#include "quota-common-utils.h"
+#include <glusterfs/quota-common-utils.h>
#include "quota-messages.h"
-#define DIRTY "dirty"
-#define SIZE "size"
-#define CONTRIBUTION "contri"
-#define VAL_LENGTH 8
-#define READDIR_BUF 4096
+#define DIRTY "dirty"
+#define SIZE "size"
+#define CONTRIBUTION "contri"
+#define VAL_LENGTH 8
+#define READDIR_BUF 4096
#ifndef UUID_CANONICAL_FORM_LEN
#define UUID_CANONICAL_FORM_LEN 36
#endif
-#define WIND_IF_QUOTAOFF(is_quota_on, label) \
- if (!is_quota_on) \
- goto label;
-
-#define QUOTA_WIND_FOR_INTERNAL_FOP(xdata, label) \
- do { \
- if (xdata && dict_get (xdata, GLUSTERFS_INTERNAL_FOP_KEY)) \
- goto label; \
- } while (0)
-
-#define DID_REACH_LIMIT(lim, prev_size, cur_size) \
- ((cur_size) >= (lim) && (prev_size) < (lim))
-
-#define QUOTA_SAFE_INCREMENT(lock, var) \
- do { \
- LOCK (lock); \
- var ++; \
- UNLOCK (lock); \
- } while (0)
-
-#define QUOTA_SAFE_DECREMENT(lock, var) \
- do { \
- LOCK (lock); \
- var --; \
- UNLOCK (lock); \
- } while (0)
-
-#define QUOTA_ALLOC_OR_GOTO(var, type, label) \
- do { \
- var = GF_CALLOC (sizeof (type), 1, \
- gf_quota_mt_##type); \
- if (!var) { \
- gf_msg ("", GF_LOG_ERROR, \
- ENOMEM, Q_MSG_ENOMEM, \
- "out of memory"); \
- ret = -1; \
- goto label; \
- } \
- } while (0);
-
-#define QUOTA_STACK_WIND_TAIL(frame, params...) \
- do { \
- quota_local_t *_local = NULL; \
- \
- if (frame) { \
- _local = frame->local; \
- frame->local = NULL; \
- } \
- \
- STACK_WIND_TAIL (frame, params); \
- \
- if (_local) \
- quota_local_cleanup (_local); \
- } while (0)
-
-#define QUOTA_STACK_UNWIND(fop, frame, params...) \
- do { \
- quota_local_t *_local = NULL; \
- if (frame) { \
- _local = frame->local; \
- frame->local = NULL; \
- } \
- STACK_UNWIND_STRICT (fop, frame, params); \
- quota_local_cleanup (_local); \
- } while (0)
-
-#define QUOTA_FREE_CONTRIBUTION_NODE(_contribution) \
- do { \
- list_del (&_contribution->contri_list); \
- GF_FREE (_contribution); \
- } while (0)
-
-#define GET_CONTRI_KEY(var, _vol_name, _gfid, _ret) \
- do { \
- char _gfid_unparsed[40]; \
- if (_gfid != NULL) { \
- gf_uuid_unparse (_gfid, _gfid_unparsed);\
- _ret = gf_asprintf (var, QUOTA_XATTR_PREFIX \
- "%s.%s." CONTRIBUTION, \
- _vol_name, _gfid_unparsed); \
- } else { \
- _ret = gf_asprintf (var, QUOTA_XATTR_PREFIX \
- "%s.." CONTRIBUTION, \
- _vol_name); \
- } \
- } while (0)
-
-
-#define GET_CONTRI_KEY_OR_GOTO(var, _vol_name, _gfid, label) \
- do { \
- GET_CONTRI_KEY(var, _vol_name, _gfid, ret); \
- if (ret == -1) \
- goto label; \
- } while (0)
-
-#define GET_DIRTY_KEY_OR_GOTO(var, _vol_name, label) \
- do { \
- ret = gf_asprintf (var, QUOTA_XATTR_PREFIX \
- "%s." DIRTY, _vol_name); \
- if (ret == -1) \
- goto label; \
- } while (0)
-
-#define QUOTA_REG_OR_LNK_FILE(ia_type) \
- (IA_ISREG (ia_type) || IA_ISLNK (ia_type))
-
-
+#define WIND_IF_QUOTAOFF(is_quota_on, label) \
+ if (!is_quota_on) \
+ goto label;
+
+#define QUOTA_WIND_FOR_INTERNAL_FOP(xdata, label) \
+ do { \
+ if (xdata && dict_get_sizen(xdata, GLUSTERFS_INTERNAL_FOP_KEY)) \
+ goto label; \
+ } while (0)
+
+#define DID_REACH_LIMIT(lim, prev_size, cur_size) \
+ ((cur_size) >= (lim) && (prev_size) < (lim))
+
+#define QUOTA_SAFE_INCREMENT(lock, var) \
+ do { \
+ LOCK(lock); \
+ var++; \
+ UNLOCK(lock); \
+ } while (0)
+
+#define QUOTA_SAFE_DECREMENT(lock, var) \
+ do { \
+ LOCK(lock); \
+ var--; \
+ UNLOCK(lock); \
+ } while (0)
+
+#define QUOTA_ALLOC_OR_GOTO(var, type, label) \
+ do { \
+ var = GF_CALLOC(sizeof(type), 1, gf_quota_mt_##type); \
+ if (!var) { \
+ gf_msg("", GF_LOG_ERROR, ENOMEM, Q_MSG_ENOMEM, "out of memory"); \
+ ret = -1; \
+ goto label; \
+ } \
+ } while (0);
+
+#define QUOTA_STACK_WIND_TAIL(frame, params...) \
+ do { \
+ quota_local_t *_local = NULL; \
+ \
+ if (frame) { \
+ _local = frame->local; \
+ frame->local = NULL; \
+ } \
+ \
+ STACK_WIND_TAIL(frame, params); \
+ \
+ if (_local) \
+ quota_local_cleanup(_local); \
+ } while (0)
+
+#define QUOTA_STACK_UNWIND(fop, frame, params...) \
+ do { \
+ quota_local_t *_local = NULL; \
+ if (frame) { \
+ _local = frame->local; \
+ frame->local = NULL; \
+ } \
+ STACK_UNWIND_STRICT(fop, frame, params); \
+ quota_local_cleanup(_local); \
+ } while (0)
+
+#define QUOTA_FREE_CONTRIBUTION_NODE(_contribution) \
+ do { \
+ list_del(&_contribution->contri_list); \
+ GF_FREE(_contribution); \
+ } while (0)
+
+#define GET_CONTRI_KEY(var, _vol_name, _gfid, _ret) \
+ do { \
+ char _gfid_unparsed[40]; \
+ if (_gfid != NULL) { \
+ gf_uuid_unparse(_gfid, _gfid_unparsed); \
+ _ret = gf_asprintf(var, QUOTA_XATTR_PREFIX "%s.%s." CONTRIBUTION, \
+ _vol_name, _gfid_unparsed); \
+ } else { \
+ _ret = gf_asprintf(var, QUOTA_XATTR_PREFIX "%s.." CONTRIBUTION, \
+ _vol_name); \
+ } \
+ } while (0)
+
+#define GET_CONTRI_KEY_OR_GOTO(var, _vol_name, _gfid, label) \
+ do { \
+ GET_CONTRI_KEY(var, _vol_name, _gfid, ret); \
+ if (ret == -1) \
+ goto label; \
+ } while (0)
+
+#define GET_DIRTY_KEY_OR_GOTO(var, _vol_name, label) \
+ do { \
+ ret = gf_asprintf(var, QUOTA_XATTR_PREFIX "%s." DIRTY, _vol_name); \
+ if (ret == -1) \
+ goto label; \
+ } while (0)
+
+#define QUOTA_REG_OR_LNK_FILE(ia_type) (IA_ISREG(ia_type) || IA_ISLNK(ia_type))
struct quota_dentry {
- char *name;
- uuid_t par;
- struct list_head next;
+ char *name;
+ uuid_t par;
+ struct list_head next;
};
typedef struct quota_dentry quota_dentry_t;
struct quota_inode_ctx {
- int64_t size;
- int64_t hard_lim;
- int64_t soft_lim;
- int64_t file_count;
- int64_t dir_count;
- int64_t object_hard_lim;
- int64_t object_soft_lim;
- struct iatt buf;
- struct list_head parents;
- struct timeval tv;
- struct timeval prev_log;
- gf_boolean_t ancestry_built;
- gf_lock_t lock;
+ int64_t size;
+ int64_t hard_lim;
+ int64_t soft_lim;
+ int64_t file_count;
+ int64_t dir_count;
+ int64_t object_hard_lim;
+ int64_t object_soft_lim;
+ struct iatt buf;
+ struct list_head parents;
+ time_t validate_time;
+ time_t prev_log_time;
+ gf_boolean_t ancestry_built;
+ gf_lock_t lock;
};
typedef struct quota_inode_ctx quota_inode_ctx_t;
-typedef void
-(*quota_ancestry_built_t) (struct list_head *parents, inode_t *inode,
- int32_t op_ret, int32_t op_errno, void *data);
+typedef void (*quota_ancestry_built_t)(struct list_head *parents,
+ inode_t *inode, int32_t op_ret,
+ int32_t op_errno, void *data);
-typedef void
-(*quota_fop_continue_t) (call_frame_t *frame);
+typedef void (*quota_fop_continue_t)(call_frame_t *frame);
struct quota_local {
- gf_lock_t lock;
- uint32_t link_count;
- loc_t loc;
- loc_t oldloc;
- loc_t newloc;
- loc_t validate_loc;
- int64_t delta;
- int8_t object_delta;
- int32_t op_ret;
- int32_t op_errno;
- int64_t size;
- char just_validated;
- fop_lookup_cbk_t validate_cbk;
- quota_fop_continue_t fop_continue_cbk;
- inode_t *inode;
- uuid_t common_ancestor; /* Used by quota_rename */
- call_stub_t *stub;
- struct iobref *iobref;
- quota_limits_t limit;
- quota_limits_t object_limit;
- int64_t space_available;
- quota_ancestry_built_t ancestry_cbk;
- void *ancestry_data;
- dict_t *xdata;
- dict_t *validate_xdata;
- int32_t quotad_conn_retry;
- xlator_t *this;
- call_frame_t *par_frame;
+ gf_lock_t lock;
+ uint32_t link_count;
+ loc_t loc;
+ loc_t oldloc;
+ loc_t newloc;
+ loc_t validate_loc;
+ int64_t delta;
+ int8_t object_delta;
+ int32_t op_ret;
+ int32_t op_errno;
+ int64_t size;
+ char just_validated;
+ fop_lookup_cbk_t validate_cbk;
+ quota_fop_continue_t fop_continue_cbk;
+ inode_t *inode;
+ uuid_t common_ancestor; /* Used by quota_rename */
+ call_stub_t *stub;
+ struct iobref *iobref;
+ quota_limits_t limit;
+ quota_limits_t object_limit;
+ int64_t space_available;
+ quota_ancestry_built_t ancestry_cbk;
+ void *ancestry_data;
+ dict_t *xdata;
+ dict_t *validate_xdata;
+ int32_t quotad_conn_retry;
+ xlator_t *this;
+ call_frame_t *par_frame;
};
-typedef struct quota_local quota_local_t;
+typedef struct quota_local quota_local_t;
struct quota_priv {
- uint32_t soft_timeout;
- uint32_t hard_timeout;
- uint32_t log_timeout;
- double default_soft_lim;
- gf_boolean_t is_quota_on;
- gf_boolean_t consider_statfs;
- gf_lock_t lock;
- rpc_clnt_prog_t *quota_enforcer;
- struct rpcsvc_program *quotad_aggregator;
- struct rpc_clnt *rpc_clnt;
- rpcsvc_t *rpcsvc;
- inode_table_t *itable;
- char *volume_uuid;
- uint64_t validation_count;
- int32_t quotad_conn_status;
+ /* FIXME: consider time_t for timeouts. */
+ uint32_t soft_timeout;
+ uint32_t hard_timeout;
+ uint32_t log_timeout;
+ double default_soft_lim;
+ gf_boolean_t is_quota_on;
+ gf_boolean_t consider_statfs;
+ gf_lock_t lock;
+ rpc_clnt_prog_t *quota_enforcer;
+ struct rpcsvc_program *quotad_aggregator;
+ struct rpc_clnt *rpc_clnt;
+ rpcsvc_t *rpcsvc;
+ inode_table_t *itable;
+ char *volume_uuid;
+ uint64_t validation_count;
+ int32_t quotad_conn_status;
+ pthread_mutex_t conn_mutex;
+ pthread_cond_t conn_cond;
+ gf_boolean_t conn_status;
};
-typedef struct quota_priv quota_priv_t;
+typedef struct quota_priv quota_priv_t;
int
-quota_enforcer_lookup (call_frame_t *frame, xlator_t *this, dict_t *xdata,
- fop_lookup_cbk_t cbk);
+quota_enforcer_lookup(call_frame_t *frame, xlator_t *this, dict_t *xdata,
+ fop_lookup_cbk_t cbk);
void
-_quota_enforcer_lookup (void *data);
+_quota_enforcer_lookup(void *data);
struct rpc_clnt *
-quota_enforcer_init (xlator_t *this, dict_t *options);
+quota_enforcer_init(xlator_t *this, dict_t *options);
void
-quota_log_usage (xlator_t *this, quota_inode_ctx_t *ctx, inode_t *inode,
- int64_t delta);
+quota_log_usage(xlator_t *this, quota_inode_ctx_t *ctx, inode_t *inode,
+ int64_t delta);
int
-quota_build_ancestry (inode_t *inode, quota_ancestry_built_t ancestry_cbk,
- void *data);
+quota_build_ancestry(inode_t *inode, quota_ancestry_built_t ancestry_cbk,
+ void *data);
void
-quota_get_limit_dir (call_frame_t *frame, inode_t *cur_inode, xlator_t *this);
+quota_get_limit_dir(call_frame_t *frame, inode_t *cur_inode, xlator_t *this);
int32_t
-quota_check_limit (call_frame_t *frame, inode_t *inode, xlator_t *this);
+quota_check_limit(call_frame_t *frame, inode_t *inode, xlator_t *this);
inode_t *
-do_quota_check_limit (call_frame_t *frame, inode_t *inode, xlator_t *this,
- quota_dentry_t *dentry, gf_boolean_t force);
+do_quota_check_limit(call_frame_t *frame, inode_t *inode, xlator_t *this,
+ quota_dentry_t *dentry, gf_boolean_t force);
int
-quota_fill_inodectx (xlator_t *this, inode_t *inode, dict_t *dict,
- loc_t *loc, struct iatt *buf, int32_t *op_errno);
+quota_fill_inodectx(xlator_t *this, inode_t *inode, dict_t *dict, loc_t *loc,
+ struct iatt *buf, int32_t *op_errno);
int32_t
-quota_check_size_limit (call_frame_t *frame, quota_inode_ctx_t *ctx,
- quota_priv_t *priv, inode_t *_inode, xlator_t *this,
- int32_t *op_errno, int just_validated, int64_t delta,
- quota_local_t *local, gf_boolean_t *skip_check);
+quota_check_size_limit(call_frame_t *frame, quota_inode_ctx_t *ctx,
+ quota_priv_t *priv, inode_t *_inode, xlator_t *this,
+ int32_t *op_errno, int just_validated, int64_t delta,
+ quota_local_t *local, gf_boolean_t *skip_check);
int32_t
-quota_check_object_limit (call_frame_t *frame, quota_inode_ctx_t *ctx,
- quota_priv_t *priv, inode_t *_inode, xlator_t *this,
- int32_t *op_errno, int just_validated,
- quota_local_t *local, gf_boolean_t *skip_check);
+quota_check_object_limit(call_frame_t *frame, quota_inode_ctx_t *ctx,
+ quota_priv_t *priv, inode_t *_inode, xlator_t *this,
+ int32_t *op_errno, int just_validated,
+ quota_local_t *local, gf_boolean_t *skip_check);
#endif
diff --git a/xlators/features/quota/src/quotad-aggregator.c b/xlators/features/quota/src/quotad-aggregator.c
index c27b67136a3..75d47867b5b 100644
--- a/xlators/features/quota/src/quotad-aggregator.c
+++ b/xlators/features/quota/src/quotad-aggregator.c
@@ -13,445 +13,482 @@
#include "quotad-helpers.h"
#include "quotad-aggregator.h"
-struct rpcsvc_program quotad_aggregator_prog;
+static char *qd_ext_xattrs[] = {
+ QUOTA_SIZE_KEY,
+ QUOTA_LIMIT_KEY,
+ QUOTA_LIMIT_OBJECTS_KEY,
+ NULL,
+};
+
+static struct rpcsvc_program quotad_aggregator_prog;
struct iobuf *
-quotad_serialize_reply (rpcsvc_request_t *req, void *arg, struct iovec *outmsg,
- xdrproc_t xdrproc)
+quotad_serialize_reply(rpcsvc_request_t *req, void *arg, struct iovec *outmsg,
+ xdrproc_t xdrproc)
{
- struct iobuf *iob = NULL;
- ssize_t retlen = 0;
- ssize_t xdr_size = 0;
-
- GF_VALIDATE_OR_GOTO ("server", req, ret);
-
- /* First, get the io buffer into which the reply in arg will
- * be serialized.
+ struct iobuf *iob = NULL;
+ ssize_t retlen = 0;
+ ssize_t xdr_size = 0;
+
+ GF_VALIDATE_OR_GOTO("server", req, ret);
+
+ /* First, get the io buffer into which the reply in arg will
+ * be serialized.
+ */
+ if (arg && xdrproc) {
+ xdr_size = xdr_sizeof(xdrproc, arg);
+ iob = iobuf_get2(req->svc->ctx->iobuf_pool, xdr_size);
+ if (!iob) {
+ gf_log_callingfn(THIS->name, GF_LOG_ERROR, "Failed to get iobuf");
+ goto ret;
+ };
+
+ iobuf_to_iovec(iob, outmsg);
+ /* Use the given serializer to translate the given C structure
+ * in arg to XDR format which will be written into the buffer
+ * in outmsg.
*/
- if (arg && xdrproc) {
- xdr_size = xdr_sizeof (xdrproc, arg);
- iob = iobuf_get2 (req->svc->ctx->iobuf_pool, xdr_size);
- if (!iob) {
- gf_log_callingfn (THIS->name, GF_LOG_ERROR,
- "Failed to get iobuf");
- goto ret;
- };
-
- iobuf_to_iovec (iob, outmsg);
- /* Use the given serializer to translate the given C structure
- * in arg to XDR format which will be written into the buffer
- * in outmsg.
- */
- /* retlen is used to received the error since size_t is unsigned and we
- * need -1 for error notification during encoding.
- */
-
- retlen = xdr_serialize_generic (*outmsg, arg, xdrproc);
- if (retlen == -1) {
- /* Failed to Encode 'GlusterFS' msg in RPC is not exactly
- failure of RPC return values.. Client should get
- notified about this, so there are no missing frames */
- gf_log_callingfn ("", GF_LOG_ERROR, "Failed to encode message");
- req->rpc_err = GARBAGE_ARGS;
- retlen = 0;
- }
+ /* retlen is used to received the error since size_t is unsigned and we
+ * need -1 for error notification during encoding.
+ */
+
+ retlen = xdr_serialize_generic(*outmsg, arg, xdrproc);
+ if (retlen == -1) {
+ /* Failed to Encode 'GlusterFS' msg in RPC is not exactly
+ failure of RPC return values.. Client should get
+ notified about this, so there are no missing frames */
+ gf_log_callingfn("", GF_LOG_ERROR, "Failed to encode message");
+ req->rpc_err = GARBAGE_ARGS;
+ retlen = 0;
}
- outmsg->iov_len = retlen;
+ }
+ outmsg->iov_len = retlen;
ret:
- return iob;
+ return iob;
}
int
-quotad_aggregator_submit_reply (call_frame_t *frame, rpcsvc_request_t *req,
- void *arg, struct iovec *payload,
- int payloadcount, struct iobref *iobref,
- xdrproc_t xdrproc)
+quotad_aggregator_submit_reply(call_frame_t *frame, rpcsvc_request_t *req,
+ void *arg, struct iovec *payload,
+ int payloadcount, struct iobref *iobref,
+ xdrproc_t xdrproc)
{
- struct iobuf *iob = NULL;
- int ret = -1;
- struct iovec rsp = {0,};
- quotad_aggregator_state_t *state = NULL;
- char new_iobref = 0;
+ struct iobuf *iob = NULL;
+ int ret = -1;
+ struct iovec rsp = {
+ 0,
+ };
+ quotad_aggregator_state_t *state = NULL;
+ char new_iobref = 0;
- GF_VALIDATE_OR_GOTO ("server", req, ret);
+ GF_VALIDATE_OR_GOTO("server", req, ret);
- if (frame) {
- state = frame->root->state;
- frame->local = NULL;
- }
+ if (frame) {
+ state = frame->root->state;
+ frame->local = NULL;
+ }
+ if (!iobref) {
+ iobref = iobref_new();
if (!iobref) {
- iobref = iobref_new ();
- if (!iobref) {
- goto ret;
- }
-
- new_iobref = 1;
+ goto ret;
}
- iob = quotad_serialize_reply (req, arg, &rsp, xdrproc);
- if (!iob) {
- gf_msg ("", GF_LOG_ERROR, 0, Q_MSG_DICT_SERIALIZE_FAIL,
- "Failed to serialize reply");
- goto ret;
- }
+ new_iobref = 1;
+ }
+
+ iob = quotad_serialize_reply(req, arg, &rsp, xdrproc);
+ if (!iob) {
+ gf_msg("", GF_LOG_ERROR, 0, Q_MSG_DICT_SERIALIZE_FAIL,
+ "Failed to serialize reply");
+ goto ret;
+ }
- iobref_add (iobref, iob);
+ iobref_add(iobref, iob);
- ret = rpcsvc_submit_generic (req, &rsp, 1, payload, payloadcount,
- iobref);
+ ret = rpcsvc_submit_generic(req, &rsp, 1, payload, payloadcount, iobref);
- iobuf_unref (iob);
+ iobuf_unref(iob);
- ret = 0;
+ ret = 0;
ret:
- if (state) {
- quotad_aggregator_free_state (state);
- }
+ if (state) {
+ quotad_aggregator_free_state(state);
+ }
- if (frame)
- STACK_DESTROY (frame->root);
+ if (frame)
+ STACK_DESTROY(frame->root);
- if (new_iobref) {
- iobref_unref (iobref);
- }
+ if (new_iobref) {
+ iobref_unref(iobref);
+ }
- return ret;
+ return ret;
}
int
-quotad_aggregator_getlimit_cbk (xlator_t *this, call_frame_t *frame,
- void *lookup_rsp)
+quotad_aggregator_getlimit_cbk(xlator_t *this, call_frame_t *frame,
+ void *lookup_rsp)
{
- gfs3_lookup_rsp *rsp = lookup_rsp;
- gf_cli_rsp cli_rsp = {0,};
- dict_t *xdata = NULL;
- quotad_aggregator_state_t *state = NULL;
- int ret = -1;
- int type = 0;
-
- GF_PROTOCOL_DICT_UNSERIALIZE (frame->this, xdata,
- (rsp->xdata.xdata_val),
- (rsp->xdata.xdata_len), rsp->op_ret,
- rsp->op_errno, out);
-
- if (xdata) {
- state = frame->root->state;
- ret = dict_get_int32 (state->xdata, "type", &type);
- if (ret < 0)
- goto out;
-
- ret = dict_set_int32 (xdata, "type", type);
- if (ret < 0)
- goto out;
- }
+ gfs3_lookup_rsp *rsp = lookup_rsp;
+ gf_cli_rsp cli_rsp = {
+ 0,
+ };
+ dict_t *xdata = NULL;
+ quotad_aggregator_state_t *state = NULL;
+ int ret = -1;
+ int type = 0;
+
+ if (!rsp || (rsp->op_ret == -1))
+ goto reply;
+
+ GF_PROTOCOL_DICT_UNSERIALIZE(frame->this, xdata, (rsp->xdata.xdata_val),
+ (rsp->xdata.xdata_len), rsp->op_ret,
+ rsp->op_errno, out);
+
+ if (xdata) {
+ state = frame->root->state;
+ ret = dict_get_int32n(state->req_xdata, "type", SLEN("type"), &type);
+ if (ret < 0)
+ goto out;
+
+ ret = dict_set_int32_sizen(xdata, "type", type);
+ if (ret < 0)
+ goto out;
+ }
- ret = 0;
+ ret = 0;
out:
- rsp->op_ret = ret;
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- Q_MSG_DICT_UNSERIALIZE_FAIL,
- "failed to unserialize "
- "nameless lookup rsp");
- goto reply;
- }
- cli_rsp.op_ret = rsp->op_ret;
- cli_rsp.op_errno = rsp->op_errno;
- cli_rsp.op_errstr = "";
- if (xdata) {
- GF_PROTOCOL_DICT_SERIALIZE (frame->this, xdata,
- (&cli_rsp.dict.dict_val),
- (cli_rsp.dict.dict_len),
- cli_rsp.op_errno, reply);
- }
+ rsp->op_ret = ret;
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, Q_MSG_DICT_UNSERIALIZE_FAIL,
+ "failed to unserialize "
+ "nameless lookup rsp");
+ goto reply;
+ }
+ cli_rsp.op_ret = rsp->op_ret;
+ cli_rsp.op_errno = rsp->op_errno;
+ cli_rsp.op_errstr = "";
+ if (xdata) {
+ GF_PROTOCOL_DICT_SERIALIZE(frame->this, xdata, (&cli_rsp.dict.dict_val),
+ (cli_rsp.dict.dict_len), cli_rsp.op_errno,
+ reply);
+ }
reply:
- quotad_aggregator_submit_reply (frame, frame->local, (void*)&cli_rsp, NULL, 0,
- NULL, (xdrproc_t)xdr_gf_cli_rsp);
+ quotad_aggregator_submit_reply(frame, (frame) ? frame->local : NULL,
+ (void *)&cli_rsp, NULL, 0, NULL,
+ (xdrproc_t)xdr_gf_cli_rsp);
- dict_unref (xdata);
- GF_FREE (cli_rsp.dict.dict_val);
- return 0;
+ dict_unref(xdata);
+ GF_FREE(cli_rsp.dict.dict_val);
+ return 0;
}
int
-quotad_aggregator_getlimit (rpcsvc_request_t *req)
+quotad_aggregator_getlimit(rpcsvc_request_t *req)
{
- call_frame_t *frame = NULL;
- gf_cli_req cli_req = {{0}, };
- gf_cli_rsp cli_rsp = {0};
- gfs3_lookup_req args = {{0,},};
- quotad_aggregator_state_t *state = NULL;
- xlator_t *this = NULL;
- dict_t *dict = NULL;
- int ret = -1, op_errno = 0;
- char *gfid_str = NULL;
- uuid_t gfid = {0};
-
- GF_VALIDATE_OR_GOTO ("quotad-aggregator", req, err);
-
- this = THIS;
-
- ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
- if (ret < 0) {
- //failed to decode msg;
- gf_msg ("this->name", GF_LOG_ERROR, 0, Q_MSG_XDR_DECODE_ERROR,
- "xdr decoding error");
- req->rpc_err = GARBAGE_ARGS;
- goto err;
- }
-
- if (cli_req.dict.dict_len) {
- dict = dict_new ();
- ret = dict_unserialize (cli_req.dict.dict_val,
- cli_req.dict.dict_len, &dict);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- Q_MSG_DICT_UNSERIALIZE_FAIL,
- "Failed to unserialize req-buffer to "
- "dictionary");
- goto err;
- }
- }
-
- ret = dict_get_str (dict, "gfid", &gfid_str);
- if (ret) {
- goto err;
- }
-
- gf_uuid_parse ((const char*)gfid_str, gfid);
-
- frame = quotad_aggregator_get_frame_from_req (req);
- if (frame == NULL) {
- cli_rsp.op_errno = ENOMEM;
- goto errx;
- }
- state = frame->root->state;
- state->xdata = dict;
-
- ret = dict_set_int32 (state->xdata, QUOTA_LIMIT_KEY, 42);
- if (ret)
- goto err;
-
- ret = dict_set_int32 (state->xdata, QUOTA_LIMIT_OBJECTS_KEY, 42);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM, Q_MSG_ENOMEM,
- "Failed to set QUOTA_LIMIT_OBJECTS_KEY");
- goto err;
- }
-
- ret = dict_set_int32 (state->xdata, QUOTA_SIZE_KEY, 42);
- if (ret)
- goto err;
-
- ret = dict_set_int32 (state->xdata, GET_ANCESTRY_PATH_KEY, 42);
- if (ret)
- goto err;
-
- memcpy (&args.gfid, &gfid, 16);
-
- args.bname = alloca (req->msg[0].iov_len);
- args.xdata.xdata_val = alloca (req->msg[0].iov_len);
-
- ret = qd_nameless_lookup (this, frame, &args, state->xdata,
- quotad_aggregator_getlimit_cbk);
- if (ret) {
- cli_rsp.op_errno = ret;
- goto errx;
+ call_frame_t *frame = NULL;
+ gf_cli_req cli_req = {
+ {0},
+ };
+ gf_cli_rsp cli_rsp = {0};
+ quotad_aggregator_state_t *state = NULL;
+ xlator_t *this = NULL;
+ dict_t *dict = NULL;
+ int ret = -1, op_errno = 0;
+ char *gfid_str = NULL;
+ uuid_t gfid = {0};
+ char *volume_uuid = NULL;
+
+ GF_VALIDATE_OR_GOTO("quotad-aggregator", req, err);
+
+ this = THIS;
+
+ cli_req.dict.dict_val = alloca(req->msg[0].iov_len);
+
+ ret = xdr_to_generic(req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+ if (ret < 0) {
+ // failed to decode msg;
+ gf_msg("this->name", GF_LOG_ERROR, 0, Q_MSG_XDR_DECODE_ERROR,
+ "xdr decoding error");
+ req->rpc_err = GARBAGE_ARGS;
+ goto err;
+ }
+
+ if (cli_req.dict.dict_len) {
+ dict = dict_new();
+ ret = dict_unserialize(cli_req.dict.dict_val, cli_req.dict.dict_len,
+ &dict);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, Q_MSG_DICT_UNSERIALIZE_FAIL,
+ "Failed to unserialize req-buffer to "
+ "dictionary");
+ goto err;
}
-
- return ret;
+ }
+
+ ret = dict_get_strn(dict, "gfid", SLEN("gfid"), &gfid_str);
+ if (ret) {
+ goto err;
+ }
+
+ ret = dict_get_strn(dict, "volume-uuid", SLEN("volume-uuid"), &volume_uuid);
+ if (ret) {
+ goto err;
+ }
+
+ gf_uuid_parse((const char *)gfid_str, gfid);
+
+ frame = quotad_aggregator_get_frame_from_req(req);
+ if (frame == NULL) {
+ cli_rsp.op_errno = ENOMEM;
+ goto errx;
+ }
+ state = frame->root->state;
+ state->req_xdata = dict;
+ state->xdata = dict_new();
+ dict = NULL;
+
+ ret = dict_set_int32_sizen(state->xdata, QUOTA_LIMIT_KEY, 42);
+ if (ret)
+ goto err;
+
+ ret = dict_set_int32_sizen(state->xdata, QUOTA_LIMIT_OBJECTS_KEY, 42);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, Q_MSG_ENOMEM,
+ "Failed to set QUOTA_LIMIT_OBJECTS_KEY");
+ goto err;
+ }
+
+ ret = dict_set_int32_sizen(state->xdata, QUOTA_SIZE_KEY, 42);
+ if (ret)
+ goto err;
+
+ ret = dict_set_int32_sizen(state->xdata, GET_ANCESTRY_PATH_KEY, 42);
+ if (ret)
+ goto err;
+
+ ret = qd_nameless_lookup(this, frame, (char *)gfid, state->xdata,
+ volume_uuid, quotad_aggregator_getlimit_cbk);
+ if (ret) {
+ cli_rsp.op_errno = ret;
+ goto errx;
+ }
+
+ return ret;
err:
- cli_rsp.op_errno = op_errno;
+ cli_rsp.op_errno = op_errno;
errx:
- cli_rsp.op_ret = -1;
- cli_rsp.op_errstr = "";
-
- quotad_aggregator_getlimit_cbk (this, frame, &cli_rsp);
- if (dict)
- dict_unref (dict);
+ cli_rsp.op_ret = -1;
+ cli_rsp.op_errstr = "";
- return ret;
+ quotad_aggregator_getlimit_cbk(this, frame, &cli_rsp);
+ if (dict)
+ dict_unref(dict);
+ return ret;
}
int
-quotad_aggregator_lookup_cbk (xlator_t *this, call_frame_t *frame,
- void *rsp)
+quotad_aggregator_lookup_cbk(xlator_t *this, call_frame_t *frame, void *rsp)
{
- quotad_aggregator_submit_reply (frame, frame->local, rsp, NULL, 0, NULL,
- (xdrproc_t)xdr_gfs3_lookup_rsp);
+ quotad_aggregator_submit_reply(frame, frame ? frame->local : NULL, rsp,
+ NULL, 0, NULL,
+ (xdrproc_t)xdr_gfs3_lookup_rsp);
- return 0;
+ return 0;
}
-
int
-quotad_aggregator_lookup (rpcsvc_request_t *req)
+quotad_aggregator_lookup(rpcsvc_request_t *req)
{
- call_frame_t *frame = NULL;
- gfs3_lookup_req args = {{0,},};
- int ret = -1, op_errno = 0;
- gfs3_lookup_rsp rsp = {0,};
- quotad_aggregator_state_t *state = NULL;
- xlator_t *this = NULL;
-
- GF_VALIDATE_OR_GOTO ("quotad-aggregator", req, err);
-
- this = THIS;
-
- args.bname = alloca (req->msg[0].iov_len);
- args.xdata.xdata_val = alloca (req->msg[0].iov_len);
-
- ret = xdr_to_generic (req->msg[0], &args,
- (xdrproc_t)xdr_gfs3_lookup_req);
- if (ret < 0) {
- rsp.op_errno = EINVAL;
+ call_frame_t *frame = NULL;
+ gfs3_lookup_req args = {
+ {
+ 0,
+ },
+ };
+ int i = 0, ret = -1, op_errno = 0;
+ gfs3_lookup_rsp rsp = {
+ 0,
+ };
+ quotad_aggregator_state_t *state = NULL;
+ xlator_t *this = NULL;
+ dict_t *dict = NULL;
+ char *volume_uuid = NULL;
+
+ GF_VALIDATE_OR_GOTO("quotad-aggregator", req, err);
+
+ this = THIS;
+
+ args.bname = alloca(req->msg[0].iov_len);
+ args.xdata.xdata_val = alloca(req->msg[0].iov_len);
+
+ ret = xdr_to_generic(req->msg[0], &args, (xdrproc_t)xdr_gfs3_lookup_req);
+ if (ret < 0) {
+ rsp.op_errno = EINVAL;
+ goto err;
+ }
+
+ frame = quotad_aggregator_get_frame_from_req(req);
+ if (frame == NULL) {
+ rsp.op_errno = ENOMEM;
+ goto err;
+ }
+
+ state = frame->root->state;
+
+ GF_PROTOCOL_DICT_UNSERIALIZE(this, dict, (args.xdata.xdata_val),
+ (args.xdata.xdata_len), ret, op_errno, err);
+
+ ret = dict_get_str(dict, "volume-uuid", &volume_uuid);
+ if (ret) {
+ goto err;
+ }
+
+ state->xdata = dict_new();
+
+ for (i = 0; qd_ext_xattrs[i]; i++) {
+ if (dict_get(dict, qd_ext_xattrs[i])) {
+ ret = dict_set_uint32(state->xdata, qd_ext_xattrs[i], 1);
+ if (ret < 0)
goto err;
}
+ }
- frame = quotad_aggregator_get_frame_from_req (req);
- if (frame == NULL) {
- rsp.op_errno = ENOMEM;
- goto err;
- }
+ ret = qd_nameless_lookup(this, frame, args.gfid, state->xdata, volume_uuid,
+ quotad_aggregator_lookup_cbk);
+ if (ret) {
+ rsp.op_errno = ret;
+ goto err;
+ }
- state = frame->root->state;
+ if (dict)
+ dict_unref(dict);
- GF_PROTOCOL_DICT_UNSERIALIZE (this, state->xdata,
- (args.xdata.xdata_val),
- (args.xdata.xdata_len), ret,
- op_errno, err);
-
-
- ret = qd_nameless_lookup (this, frame, &args, state->xdata,
- quotad_aggregator_lookup_cbk);
- if (ret) {
- rsp.op_errno = ret;
- goto err;
- }
-
- return ret;
+ return ret;
err:
- rsp.op_ret = -1;
- rsp.op_errno = op_errno;
+ rsp.op_ret = -1;
+ rsp.op_errno = op_errno;
- quotad_aggregator_lookup_cbk (this, frame, &rsp);
- return ret;
+ quotad_aggregator_lookup_cbk(this, frame, &rsp);
+ if (dict)
+ dict_unref(dict);
+
+ return ret;
}
int
-quotad_aggregator_rpc_notify (rpcsvc_t *rpc, void *xl, rpcsvc_event_t event,
- void *data)
+quotad_aggregator_rpc_notify(rpcsvc_t *rpc, void *xl, rpcsvc_event_t event,
+ void *data)
{
- if (!xl || !data) {
- gf_log_callingfn ("server", GF_LOG_WARNING,
- "Calling rpc_notify without initializing");
- goto out;
- }
+ if (!xl || !data) {
+ gf_log_callingfn("server", GF_LOG_WARNING,
+ "Calling rpc_notify without initializing");
+ goto out;
+ }
- switch (event) {
+ switch (event) {
case RPCSVC_EVENT_ACCEPT:
- break;
+ break;
case RPCSVC_EVENT_DISCONNECT:
- break;
+ break;
default:
- break;
- }
+ break;
+ }
out:
- return 0;
+ return 0;
}
int
-quotad_aggregator_init (xlator_t *this)
+quotad_aggregator_init(xlator_t *this)
{
- quota_priv_t *priv = NULL;
- int ret = -1;
+ quota_priv_t *priv = NULL;
+ int ret = -1;
- priv = this->private;
+ priv = this->private;
- if (priv->rpcsvc) {
- /* Listener already created */
- return 0;
- }
-
- ret = dict_set_str (this->options, "transport.address-family", "unix");
- if (ret)
- goto out;
-
- ret = dict_set_str (this->options, "transport-type", "socket");
- if (ret)
- goto out;
-
- ret = dict_set_str (this->options, "transport.socket.listen-path",
- "/var/run/gluster/quotad.socket");
- if (ret)
- goto out;
-
- /* RPC related */
- priv->rpcsvc = rpcsvc_init (this, this->ctx, this->options, 0);
- if (priv->rpcsvc == NULL) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- Q_MSG_RPCSVC_INIT_FAILED,
- "creation of rpcsvc failed");
- ret = -1;
- goto out;
- }
-
- ret = rpcsvc_create_listeners (priv->rpcsvc, this->options,
- this->name);
- if (ret < 1) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- Q_MSG_RPCSVC_LISTENER_CREATION_FAILED,
- "creation of listener failed");
- ret = -1;
- goto out;
- }
-
- priv->quotad_aggregator = &quotad_aggregator_prog;
- quotad_aggregator_prog.options = this->options;
-
- ret = rpcsvc_program_register (priv->rpcsvc, &quotad_aggregator_prog,
- _gf_false);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- Q_MSG_RPCSVC_REGISTER_FAILED,
- "registration of program (name:%s, prognum:%d, "
- "progver:%d) failed", quotad_aggregator_prog.progname,
- quotad_aggregator_prog.prognum,
- quotad_aggregator_prog.progver);
- goto out;
- }
-
- ret = 0;
+ if (priv->rpcsvc) {
+ /* Listener already created */
+ return 0;
+ }
+
+ ret = dict_set_nstrn(this->options, "transport.address-family",
+ SLEN("transport.address-family"), "unix",
+ SLEN("unix"));
+ if (ret)
+ goto out;
+
+ ret = dict_set_nstrn(this->options, "transport-type",
+ SLEN("transport-type"), "socket", SLEN("socket"));
+ if (ret)
+ goto out;
+
+ ret = dict_set_nstrn(this->options, "transport.socket.listen-path",
+ SLEN("transport.socket.listen-path"),
+ "/var/run/gluster/quotad.socket",
+ SLEN("/var/run/gluster/quotad.socket"));
+ if (ret)
+ goto out;
+
+ /* RPC related */
+ priv->rpcsvc = rpcsvc_init(this, this->ctx, this->options, 0);
+ if (priv->rpcsvc == NULL) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, Q_MSG_RPCSVC_INIT_FAILED,
+ "creation of rpcsvc failed");
+ ret = -1;
+ goto out;
+ }
+
+ ret = rpcsvc_create_listeners(priv->rpcsvc, this->options, this->name);
+ if (ret < 1) {
+ gf_msg(this->name, GF_LOG_WARNING, 0,
+ Q_MSG_RPCSVC_LISTENER_CREATION_FAILED,
+ "creation of listener failed");
+ ret = -1;
+ goto out;
+ }
+
+ priv->quotad_aggregator = &quotad_aggregator_prog;
+ quotad_aggregator_prog.options = this->options;
+
+ ret = rpcsvc_program_register(priv->rpcsvc, &quotad_aggregator_prog,
+ _gf_false);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, Q_MSG_RPCSVC_REGISTER_FAILED,
+ "registration of program (name:%s, prognum:%d, "
+ "progver:%d) failed",
+ quotad_aggregator_prog.progname, quotad_aggregator_prog.prognum,
+ quotad_aggregator_prog.progver);
+ goto out;
+ }
+
+ ret = 0;
out:
- if (ret && priv->rpcsvc) {
- GF_FREE (priv->rpcsvc);
- priv->rpcsvc = NULL;
- }
+ if (ret && priv->rpcsvc) {
+ GF_FREE(priv->rpcsvc);
+ priv->rpcsvc = NULL;
+ }
- return ret;
+ return ret;
}
-rpcsvc_actor_t quotad_aggregator_actors[GF_AGGREGATOR_MAXVALUE] = {
- [GF_AGGREGATOR_NULL] = {"NULL", GF_AGGREGATOR_NULL, NULL, NULL, 0,
- DRC_NA},
- [GF_AGGREGATOR_LOOKUP] = {"LOOKUP", GF_AGGREGATOR_NULL,
- quotad_aggregator_lookup, NULL, 0, DRC_NA},
- [GF_AGGREGATOR_GETLIMIT] = {"GETLIMIT", GF_AGGREGATOR_GETLIMIT,
- quotad_aggregator_getlimit, NULL, 0, DRC_NA},
+static rpcsvc_actor_t quotad_aggregator_actors[GF_AGGREGATOR_MAXVALUE] = {
+ [GF_AGGREGATOR_NULL] = {"NULL", NULL, NULL, GF_AGGREGATOR_NULL, DRC_NA, 0},
+ [GF_AGGREGATOR_LOOKUP] = {"LOOKUP", quotad_aggregator_lookup, NULL,
+ GF_AGGREGATOR_NULL, DRC_NA, 0},
+ [GF_AGGREGATOR_GETLIMIT] = {"GETLIMIT", quotad_aggregator_getlimit, NULL,
+ GF_AGGREGATOR_GETLIMIT, DRC_NA, 0},
};
-
-struct rpcsvc_program quotad_aggregator_prog = {
- .progname = "GlusterFS 3.3",
- .prognum = GLUSTER_AGGREGATOR_PROGRAM,
- .progver = GLUSTER_AGGREGATOR_VERSION,
- .numactors = GF_AGGREGATOR_MAXVALUE,
- .actors = quotad_aggregator_actors
-};
+static struct rpcsvc_program quotad_aggregator_prog = {
+ .progname = "GlusterFS 3.3",
+ .prognum = GLUSTER_AGGREGATOR_PROGRAM,
+ .progver = GLUSTER_AGGREGATOR_VERSION,
+ .numactors = GF_AGGREGATOR_MAXVALUE,
+ .actors = quotad_aggregator_actors};
diff --git a/xlators/features/quota/src/quotad-aggregator.h b/xlators/features/quota/src/quotad-aggregator.h
index 5ddea5b3c46..706592c7d50 100644
--- a/xlators/features/quota/src/quotad-aggregator.h
+++ b/xlators/features/quota/src/quotad-aggregator.h
@@ -12,26 +12,27 @@
#define _QUOTAD_AGGREGATOR_H
#include "quota.h"
-#include "stack.h"
+#include <glusterfs/stack.h>
#include "glusterfs3-xdr.h"
-#include "inode.h"
+#include <glusterfs/inode.h>
typedef struct {
- void *pool;
- xlator_t *this;
- xlator_t *active_subvol;
- inode_table_t *itable;
- loc_t loc;
- dict_t *xdata;
+ void *pool;
+ xlator_t *this;
+ xlator_t *active_subvol;
+ inode_table_t *itable;
+ loc_t loc;
+ dict_t *xdata;
+ dict_t *req_xdata;
} quotad_aggregator_state_t;
-typedef int (*quotad_aggregator_lookup_cbk_t) (xlator_t *this,
- call_frame_t *frame,
- void *rsp);
+typedef int (*quotad_aggregator_lookup_cbk_t)(xlator_t *this,
+ call_frame_t *frame, void *rsp);
int
-qd_nameless_lookup (xlator_t *this, call_frame_t *frame, gfs3_lookup_req *req,
- dict_t *xdata, quotad_aggregator_lookup_cbk_t lookup_cbk);
+qd_nameless_lookup(xlator_t *this, call_frame_t *frame, char *gfid,
+ dict_t *xdata, char *volume_uuid,
+ quotad_aggregator_lookup_cbk_t lookup_cbk);
int
-quotad_aggregator_init (xlator_t *this);
+quotad_aggregator_init(xlator_t *this);
#endif
diff --git a/xlators/features/quota/src/quotad-helpers.c b/xlators/features/quota/src/quotad-helpers.c
index 70298fc87f5..51ff1d7e98d 100644
--- a/xlators/features/quota/src/quotad-helpers.c
+++ b/xlators/features/quota/src/quotad-helpers.c
@@ -11,97 +11,97 @@
#include "quotad-helpers.h"
quotad_aggregator_state_t *
-get_quotad_aggregator_state (xlator_t *this, rpcsvc_request_t *req)
+get_quotad_aggregator_state(xlator_t *this, rpcsvc_request_t *req)
{
- quotad_aggregator_state_t *state = NULL;
- xlator_t *active_subvol = NULL;
- quota_priv_t *priv = NULL;
+ quotad_aggregator_state_t *state = NULL;
+ xlator_t *active_subvol = NULL;
+ quota_priv_t *priv = NULL;
- state = (void *)GF_CALLOC (1, sizeof (*state),
- gf_quota_mt_aggregator_state_t);
- if (!state)
- return NULL;
+ state = (void *)GF_CALLOC(1, sizeof(*state),
+ gf_quota_mt_aggregator_state_t);
+ if (!state)
+ return NULL;
- state->this = THIS;
- priv = this->private;
+ state->this = THIS;
+ priv = this->private;
- LOCK (&priv->lock);
- {
- active_subvol = state->active_subvol = FIRST_CHILD (this);
- }
- UNLOCK (&priv->lock);
+ LOCK(&priv->lock);
+ {
+ active_subvol = state->active_subvol = FIRST_CHILD(this);
+ }
+ UNLOCK(&priv->lock);
- if (active_subvol->itable == NULL)
- active_subvol->itable = inode_table_new (4096, active_subvol);
+ if (active_subvol->itable == NULL)
+ active_subvol->itable = inode_table_new(4096, active_subvol);
- state->itable = active_subvol->itable;
+ state->itable = active_subvol->itable;
- state->pool = this->ctx->pool;
+ state->pool = this->ctx->pool;
- return state;
+ return state;
}
void
-quotad_aggregator_free_state (quotad_aggregator_state_t *state)
+quotad_aggregator_free_state(quotad_aggregator_state_t *state)
{
- if (state->xdata)
- dict_unref (state->xdata);
+ if (state->xdata)
+ dict_unref(state->xdata);
- GF_FREE (state);
+ if (state->req_xdata)
+ dict_unref(state->req_xdata);
+
+ GF_FREE(state);
}
call_frame_t *
-quotad_aggregator_alloc_frame (rpcsvc_request_t *req)
+quotad_aggregator_alloc_frame(rpcsvc_request_t *req)
{
- call_frame_t *frame = NULL;
- quotad_aggregator_state_t *state = NULL;
- xlator_t *this = NULL;
+ call_frame_t *frame = NULL;
+ quotad_aggregator_state_t *state = NULL;
+ xlator_t *this = NULL;
- GF_VALIDATE_OR_GOTO ("server", req, out);
- GF_VALIDATE_OR_GOTO ("server", req->trans, out);
- GF_VALIDATE_OR_GOTO ("server", req->svc, out);
- GF_VALIDATE_OR_GOTO ("server", req->svc->ctx, out);
+ GF_VALIDATE_OR_GOTO("server", req, out);
+ GF_VALIDATE_OR_GOTO("server", req->trans, out);
+ GF_VALIDATE_OR_GOTO("server", req->svc, out);
+ GF_VALIDATE_OR_GOTO("server", req->svc->ctx, out);
- this = req->svc->xl;
+ this = req->svc->xl;
- frame = create_frame (this, req->svc->ctx->pool);
- if (!frame)
- goto out;
+ frame = create_frame(this, req->svc->ctx->pool);
+ if (!frame)
+ goto out;
- state = get_quotad_aggregator_state (this, req);
- if (!state)
- goto out;
+ state = get_quotad_aggregator_state(this, req);
+ if (!state)
+ goto out;
- frame->root->state = state;
- frame->root->unique = 0;
+ frame->root->state = state;
- frame->this = this;
+ frame->this = this;
out:
- return frame;
+ return frame;
}
call_frame_t *
-quotad_aggregator_get_frame_from_req (rpcsvc_request_t *req)
+quotad_aggregator_get_frame_from_req(rpcsvc_request_t *req)
{
- call_frame_t *frame = NULL;
-
- GF_VALIDATE_OR_GOTO ("server", req, out);
+ call_frame_t *frame = NULL;
- frame = quotad_aggregator_alloc_frame (req);
- if (!frame)
- goto out;
+ GF_VALIDATE_OR_GOTO("server", req, out);
- frame->root->op = req->procnum;
+ frame = quotad_aggregator_alloc_frame(req);
+ if (!frame)
+ goto out;
- frame->root->unique = req->xid;
+ frame->root->op = req->procnum;
- frame->root->uid = req->uid;
- frame->root->gid = req->gid;
- frame->root->pid = req->pid;
+ frame->root->uid = req->uid;
+ frame->root->gid = req->gid;
+ frame->root->pid = req->pid;
- frame->root->lk_owner = req->lk_owner;
+ frame->root->lk_owner = req->lk_owner;
- frame->local = req;
+ frame->local = req;
out:
- return frame;
+ return frame;
}
diff --git a/xlators/features/quota/src/quotad-helpers.h b/xlators/features/quota/src/quotad-helpers.h
index a10fb7fa82a..bcb39fe845e 100644
--- a/xlators/features/quota/src/quotad-helpers.h
+++ b/xlators/features/quota/src/quotad-helpers.h
@@ -16,9 +16,9 @@
#include "quotad-aggregator.h"
void
-quotad_aggregator_free_state (quotad_aggregator_state_t *state);
+quotad_aggregator_free_state(quotad_aggregator_state_t *state);
call_frame_t *
-quotad_aggregator_get_frame_from_req (rpcsvc_request_t *req);
+quotad_aggregator_get_frame_from_req(rpcsvc_request_t *req);
#endif
diff --git a/xlators/features/quota/src/quotad.c b/xlators/features/quota/src/quotad.c
index 2c9949fa127..643f25c9c2a 100644
--- a/xlators/features/quota/src/quotad.c
+++ b/xlators/features/quota/src/quotad.c
@@ -9,232 +9,237 @@
*/
#include "quota.h"
#include "quotad-aggregator.h"
-#include "common-utils.h"
int
-qd_notify (xlator_t *this, int32_t event, void *data, ...)
+qd_notify(xlator_t *this, int32_t event, void *data, ...)
{
- switch (event) {
+ switch (event) {
case GF_EVENT_PARENT_UP:
- quotad_aggregator_init (this);
- }
+ quotad_aggregator_init(this);
+ }
- default_notify (this, event, data);
- return 0;
+ default_notify(this, event, data);
+ return 0;
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
-
- if (!this)
- return ret;
+ int ret = -1;
- ret = xlator_mem_acct_init (this, gf_quota_mt_end + 1);
+ if (!this)
+ return ret;
- if (0 != ret) {
- gf_log (this->name, GF_LOG_WARNING, "Memory accounting "
- "init failed");
- return ret;
- }
+ ret = xlator_mem_acct_init(this, gf_quota_mt_end + 1);
+ if (0 != ret) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "Memory accounting "
+ "init failed");
return ret;
+ }
+
+ return ret;
}
int32_t
-qd_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+qd_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode, struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
{
- quotad_aggregator_lookup_cbk_t lookup_cbk = NULL;
- gfs3_lookup_rsp rsp = {0, };
+ quotad_aggregator_lookup_cbk_t lookup_cbk = NULL;
+ gfs3_lookup_rsp rsp = {
+ 0,
+ };
- lookup_cbk = cookie;
+ lookup_cbk = cookie;
- rsp.op_ret = op_ret;
- rsp.op_errno = op_errno;
+ rsp.op_ret = op_ret;
+ rsp.op_errno = op_errno;
- gf_stat_from_iatt (&rsp.postparent, postparent);
+ gf_stat_from_iatt(&rsp.postparent, postparent);
- GF_PROTOCOL_DICT_SERIALIZE (this, xdata, (&rsp.xdata.xdata_val),
- rsp.xdata.xdata_len, rsp.op_errno, out);
+ GF_PROTOCOL_DICT_SERIALIZE(this, xdata, (&rsp.xdata.xdata_val),
+ rsp.xdata.xdata_len, rsp.op_errno, out);
- gf_stat_from_iatt (&rsp.stat, buf);
+ gf_stat_from_iatt(&rsp.stat, buf);
out:
- lookup_cbk (this, frame, &rsp);
+ lookup_cbk(this, frame, &rsp);
- GF_FREE (rsp.xdata.xdata_val);
+ GF_FREE(rsp.xdata.xdata_val);
- inode_unref (inode);
+ inode_unref(inode);
- return 0;
+ return 0;
}
xlator_t *
-qd_find_subvol (xlator_t *this, char *volume_uuid)
+qd_find_subvol(xlator_t *this, char *volume_uuid)
{
- xlator_list_t *child = NULL;
- xlator_t *subvol = NULL;
- char key[1024];
- char *optstr = NULL;
-
- if (!this || !volume_uuid)
- goto out;
-
- for (child = this->children; child; child = child->next) {
- snprintf(key, 1024, "%s.volume-id", child->xlator->name);
- if (dict_get_str(this->options, key, &optstr) < 0)
- continue;
-
- if (strcmp (optstr, volume_uuid) == 0) {
- subvol = child->xlator;
- break;
- }
+ xlator_list_t *child = NULL;
+ xlator_t *subvol = NULL;
+ char key[1024];
+ int keylen = 0;
+ char *optstr = NULL;
+
+ if (!this || !volume_uuid)
+ goto out;
+
+ for (child = this->children; child; child = child->next) {
+ keylen = snprintf(key, sizeof(key), "%s.volume-id",
+ child->xlator->name);
+ if (dict_get_strn(this->options, key, keylen, &optstr) < 0)
+ continue;
+
+ if (strcmp(optstr, volume_uuid) == 0) {
+ subvol = child->xlator;
+ break;
}
+ }
out:
- return subvol;
+ return subvol;
}
int
-qd_nameless_lookup (xlator_t *this, call_frame_t *frame, gfs3_lookup_req *req,
- dict_t *xdata, quotad_aggregator_lookup_cbk_t lookup_cbk)
+qd_nameless_lookup(xlator_t *this, call_frame_t *frame, char *gfid,
+ dict_t *xdata, char *volume_uuid,
+ quotad_aggregator_lookup_cbk_t lookup_cbk)
{
- gfs3_lookup_rsp rsp = {0, };
- int op_errno = 0, ret = -1;
- loc_t loc = {0, };
- quotad_aggregator_state_t *state = NULL;
- xlator_t *subvol = NULL;
- char *volume_uuid = NULL;
-
- state = frame->root->state;
-
- frame->root->op = GF_FOP_LOOKUP;
-
- loc.inode = inode_new (state->itable);
- if (loc.inode == NULL) {
- op_errno = ENOMEM;
- goto out;
- }
-
- memcpy (loc.gfid, req->gfid, 16);
-
- ret = dict_get_str (xdata, "volume-uuid", &volume_uuid);
- if (ret < 0) {
- op_errno = EINVAL;
- goto out;
- }
-
- ret = dict_set_int8 (xdata, QUOTA_READ_ONLY_KEY, 1);
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- Q_MSG_ENOMEM, "dict set failed");
- ret = -ENOMEM;
- goto out;
- }
-
- subvol = qd_find_subvol (this, volume_uuid);
- if (subvol == NULL) {
- op_errno = EINVAL;
- goto out;
- }
-
- STACK_WIND_COOKIE (frame, qd_lookup_cbk, lookup_cbk, subvol,
- subvol->fops->lookup, &loc, xdata);
- return 0;
+ gfs3_lookup_rsp rsp = {
+ 0,
+ };
+ int op_errno = 0, ret = -1;
+ loc_t loc = {
+ 0,
+ };
+ quotad_aggregator_state_t *state = NULL;
+ xlator_t *subvol = NULL;
+
+ state = frame->root->state;
+
+ frame->root->op = GF_FOP_LOOKUP;
+
+ loc.inode = inode_new(state->itable);
+ if (loc.inode == NULL) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ memcpy(loc.gfid, gfid, 16);
+
+ ret = dict_set_int8(xdata, QUOTA_READ_ONLY_KEY, 1);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, Q_MSG_ENOMEM,
+ "dict set failed");
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ subvol = qd_find_subvol(this, volume_uuid);
+ if (subvol == NULL) {
+ op_errno = EINVAL;
+ goto out;
+ }
+
+ STACK_WIND_COOKIE(frame, qd_lookup_cbk, lookup_cbk, subvol,
+ subvol->fops->lookup, &loc, xdata);
+ return 0;
out:
- rsp.op_ret = -1;
- rsp.op_errno = op_errno;
+ rsp.op_ret = -1;
+ rsp.op_errno = op_errno;
- lookup_cbk (this, frame, &rsp);
+ lookup_cbk(this, frame, &rsp);
- inode_unref (loc.inode);
- return 0;
+ inode_unref(loc.inode);
+ return 0;
}
int
-qd_reconfigure (xlator_t *this, dict_t *options)
+qd_reconfigure(xlator_t *this, dict_t *options)
{
- /* As of now quotad is restarted upon alteration of volfile */
- return 0;
+ /* As of now quotad is restarted upon alteration of volfile */
+ return 0;
}
void
-qd_fini (xlator_t *this)
+qd_fini(xlator_t *this)
{
- quota_priv_t *priv = NULL;
+ quota_priv_t *priv = NULL;
- if (this == NULL || this->private == NULL)
- goto out;
+ if (this == NULL || this->private == NULL)
+ goto out;
- priv = this->private;
+ priv = this->private;
- if (priv->rpcsvc) {
- GF_FREE (priv->rpcsvc);
- priv->rpcsvc = NULL;
- }
+ if (priv->rpcsvc) {
+ GF_FREE(priv->rpcsvc);
+ priv->rpcsvc = NULL;
+ }
- GF_FREE (priv);
+ GF_FREE(priv);
out:
- return;
+ return;
}
int32_t
-qd_init (xlator_t *this)
+qd_init(xlator_t *this)
{
- int32_t ret = -1;
- quota_priv_t *priv = NULL;
-
- if (NULL == this->children) {
- gf_log (this->name, GF_LOG_ERROR,
- "FATAL: quota (%s) not configured for min of 1 child",
- this->name);
- ret = -1;
- goto err;
- }
+ int32_t ret = -1;
+ quota_priv_t *priv = NULL;
+
+ if (NULL == this->children) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "FATAL: quota (%s) not configured for min of 1 child",
+ this->name);
+ ret = -1;
+ goto err;
+ }
- QUOTA_ALLOC_OR_GOTO (priv, quota_priv_t, err);
- LOCK_INIT (&priv->lock);
+ QUOTA_ALLOC_OR_GOTO(priv, quota_priv_t, err);
+ LOCK_INIT(&priv->lock);
- this->private = priv;
+ this->private = priv;
- ret = 0;
+ ret = 0;
err:
- if (ret) {
- GF_FREE (priv);
- }
- return ret;
+ if (ret) {
+ GF_FREE(priv);
+ }
+ return ret;
}
-class_methods_t class_methods = {
- .init = qd_init,
- .fini = qd_fini,
- .reconfigure = qd_reconfigure,
- .notify = qd_notify
-};
+struct xlator_fops fops = {};
-struct xlator_fops fops = {
-};
+struct xlator_cbks cbks = {};
-struct xlator_cbks cbks = {
+struct volume_options options[] = {
+ {.key = {"transport-type"},
+ .value = {"rpc", "rpc-over-rdma", "tcp", "socket", "ib-verbs", "unix",
+ "ib-sdp", "tcp/server", "ib-verbs/server", "rdma",
+ "rdma*([ \t]),*([ \t])socket", "rdma*([ \t]),*([ \t])tcp",
+ "tcp*([ \t]),*([ \t])rdma", "socket*([ \t]),*([ \t])rdma"},
+ .type = GF_OPTION_TYPE_STR},
+ {
+ .key = {"transport.*"},
+ .type = GF_OPTION_TYPE_ANY,
+ },
+ {.key = {NULL}},
};
-struct volume_options options[] = {
- { .key = {"transport-type"},
- .value = {"rpc", "rpc-over-rdma", "tcp", "socket", "ib-verbs",
- "unix", "ib-sdp", "tcp/server", "ib-verbs/server", "rdma",
- "rdma*([ \t]),*([ \t])socket",
- "rdma*([ \t]),*([ \t])tcp",
- "tcp*([ \t]),*([ \t])rdma",
- "socket*([ \t]),*([ \t])rdma"},
- .type = GF_OPTION_TYPE_STR
- },
- { .key = {"transport.*"},
- .type = GF_OPTION_TYPE_ANY,
- },
- {.key = {NULL}}
+xlator_api_t xlator_api = {
+ .init = qd_init,
+ .fini = qd_fini,
+ .reconfigure = qd_reconfigure,
+ .notify = qd_notify,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1},
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "quotad",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/features/quota/src/quotad.sym b/xlators/features/quota/src/quotad.sym
deleted file mode 100644
index 0829ffe1584..00000000000
--- a/xlators/features/quota/src/quotad.sym
+++ /dev/null
@@ -1,7 +0,0 @@
-fops
-cbks
-class_methods
-options
-mem_acct_init
-reconfigure
-dumpops
diff --git a/xlators/features/read-only/src/read-only-common.c b/xlators/features/read-only/src/read-only-common.c
index f4851a4be40..9640e7e3eee 100644
--- a/xlators/features/read-only/src/read-only-common.c
+++ b/xlators/features/read-only/src/read-only-common.c
@@ -9,427 +9,398 @@
*/
#include "read-only.h"
#include "read-only-mem-types.h"
-#include "defaults.h"
+#include <glusterfs/defaults.h>
gf_boolean_t
-is_readonly_or_worm_enabled (call_frame_t *frame, xlator_t *this)
+is_readonly_or_worm_enabled(call_frame_t *frame, xlator_t *this)
{
- read_only_priv_t *priv = NULL;
- gf_boolean_t readonly_or_worm_enabled = _gf_false;
+ read_only_priv_t *priv = NULL;
+ gf_boolean_t readonly_or_worm_enabled = _gf_false;
- priv = this->private;
- GF_ASSERT (priv);
+ priv = this->private;
+ GF_ASSERT(priv);
- readonly_or_worm_enabled = priv->readonly_or_worm_enabled;
+ readonly_or_worm_enabled = priv->readonly_or_worm_enabled;
- if (frame->root->pid < GF_CLIENT_PID_MAX)
- readonly_or_worm_enabled = _gf_false;
+ if (frame->root->pid < GF_CLIENT_PID_MAX)
+ readonly_or_worm_enabled = _gf_false;
- return readonly_or_worm_enabled;
+ return readonly_or_worm_enabled;
}
static int
-_check_key_is_zero_filled (dict_t *d, char *k, data_t *v,
- void *tmp)
+_check_key_is_zero_filled(dict_t *d, char *k, data_t *v, void *tmp)
{
- if (mem_0filled ((const char *)v->data, v->len)) {
- /* -1 means, no more iterations, treat as 'break' */
- return -1;
- }
- return 0;
+ if (mem_0filled((const char *)v->data, v->len)) {
+ /* -1 means, no more iterations, treat as 'break' */
+ return -1;
+ }
+ return 0;
}
int32_t
-ro_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+ro_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- gf_boolean_t allzero = _gf_false;
- int ret = 0;
-
- ret = dict_foreach (dict, _check_key_is_zero_filled, NULL);
- if (ret == 0)
- allzero = _gf_true;
-
- if (is_readonly_or_worm_enabled (frame, this) && !allzero)
- STACK_UNWIND_STRICT (xattrop, frame, -1, EROFS, NULL, xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->xattrop,
- loc, flags, dict, xdata);
- return 0;
+ gf_boolean_t allzero = _gf_false;
+ int ret = 0;
+
+ ret = dict_foreach(dict, _check_key_is_zero_filled, NULL);
+ if (ret == 0)
+ allzero = _gf_true;
+
+ if (is_readonly_or_worm_enabled(frame, this) && !allzero)
+ STACK_UNWIND_STRICT(xattrop, frame, -1, EROFS, NULL, xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->xattrop, loc, flags, dict,
+ xdata);
+ return 0;
}
int32_t
-ro_fxattrop (call_frame_t *frame, xlator_t *this,
- fd_t *fd, gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+ro_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
{
- gf_boolean_t allzero = _gf_false;
- int ret = 0;
+ gf_boolean_t allzero = _gf_false;
+ int ret = 0;
- ret = dict_foreach (dict, _check_key_is_zero_filled, NULL);
- if (ret == 0)
- allzero = _gf_true;
+ ret = dict_foreach(dict, _check_key_is_zero_filled, NULL);
+ if (ret == 0)
+ allzero = _gf_true;
- if (is_readonly_or_worm_enabled (frame, this) && !allzero)
- STACK_UNWIND_STRICT (fxattrop, frame, -1, EROFS, NULL, xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->fxattrop,
- fd, flags, dict, xdata);
+ if (is_readonly_or_worm_enabled(frame, this) && !allzero)
+ STACK_UNWIND_STRICT(fxattrop, frame, -1, EROFS, NULL, xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fxattrop, fd, flags, dict,
+ xdata);
- return 0;
+ return 0;
}
int32_t
-ro_entrylk (call_frame_t *frame, xlator_t *this, const char *volume,
- loc_t *loc, const char *basename, entrylk_cmd cmd,
- entrylk_type type, dict_t *xdata)
+ro_entrylk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata)
{
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- volume, loc, basename, cmd, type, xdata);
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->entrylk,
+ volume, loc, basename, cmd, type, xdata);
- return 0;
+ return 0;
}
int32_t
-ro_fentrylk (call_frame_t *frame, xlator_t *this, const char *volume,
- fd_t *fd, const char *basename, entrylk_cmd cmd, entrylk_type type,
- dict_t *xdata)
+ro_fentrylk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata)
{
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->fentrylk,
- volume, fd, basename, cmd, type, xdata);
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->fentrylk,
+ volume, fd, basename, cmd, type, xdata);
- return 0;
+ return 0;
}
int32_t
-ro_inodelk (call_frame_t *frame, xlator_t *this, const char *volume,
- loc_t *loc, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
+ro_inodelk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ int32_t cmd, struct gf_flock *lock, dict_t *xdata)
{
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->inodelk,
- volume, loc, cmd, lock, xdata);
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->inodelk,
+ volume, loc, cmd, lock, xdata);
- return 0;
+ return 0;
}
int32_t
-ro_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
- fd_t *fd, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
+ro_finodelk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ int32_t cmd, struct gf_flock *lock, dict_t *xdata)
{
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->finodelk,
- volume, fd, cmd, lock, xdata);
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->finodelk,
+ volume, fd, cmd, lock, xdata);
- return 0;
+ return 0;
}
int32_t
-ro_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int cmd,
- struct gf_flock *flock, dict_t *xdata)
+ro_lk(call_frame_t *frame, xlator_t *this, fd_t *fd, int cmd,
+ struct gf_flock *flock, dict_t *xdata)
{
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->lk, fd, cmd, flock,
- xdata);
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->lk, fd,
+ cmd, flock, xdata);
- return 0;
+ return 0;
}
int32_t
-ro_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+ro_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *stbuf,
+ int32_t valid, dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (setattr, frame, -1, EROFS, NULL, NULL,
- xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->setattr, loc, stbuf,
- valid, xdata);
-
- return 0;
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(setattr, frame, -1, EROFS, NULL, NULL, xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid,
+ xdata);
+
+ return 0;
}
int32_t
-ro_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+ro_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
+ int32_t valid, dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (fsetattr, frame, -1, EROFS, NULL, NULL,
- xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->fsetattr, fd, stbuf,
- valid, xdata);
-
- return 0;
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(fsetattr, frame, -1, EROFS, NULL, NULL, xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetattr, fd, stbuf, valid,
+ xdata);
+
+ return 0;
}
-
int32_t
-ro_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset, dict_t *xdata)
+ro_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (truncate, frame, -1, EROFS, NULL, NULL,
- xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->truncate, loc, offset,
- xdata);
-
- return 0;
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(truncate, frame, -1, EROFS, NULL, NULL, xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
+
+ return 0;
}
int32_t
-ro_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset, dict_t *xdata)
+ro_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (ftruncate, frame, -1, EROFS, NULL, NULL,
- xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->ftruncate, fd, offset,
- xdata);
-
- return 0;
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(ftruncate, frame, -1, EROFS, NULL, NULL, xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+
+ return 0;
}
int32_t
-ro_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
- off_t offset, size_t len, dict_t *xdata)
+ro_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+ off_t offset, size_t len, dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (fallocate, frame, -1, EROFS, NULL, NULL,
- xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->fallocate, fd, mode,
- offset, len, xdata);
- return 0;
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(fallocate, frame, -1, EROFS, NULL, NULL, xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fallocate, fd, mode, offset,
+ len, xdata);
+ return 0;
}
-
int
-ro_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- dev_t rdev, mode_t umask, dict_t *xdata)
+ro_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (mknod, frame, -1, EROFS, NULL, NULL, NULL,
- NULL, xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->mknod, loc, mode,
- rdev, umask, xdata);
-
- return 0;
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(mknod, frame, -1, EROFS, NULL, NULL, NULL, NULL,
+ xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, loc, mode, rdev, umask,
+ xdata);
+
+ return 0;
}
-
int
-ro_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- mode_t umask, dict_t *xdata)
+ro_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (mkdir, frame, -1, EROFS, NULL, NULL, NULL,
- NULL, xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->mkdir, loc, mode,
- umask, xdata);
-
- return 0;
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(mkdir, frame, -1, EROFS, NULL, NULL, NULL, NULL,
+ xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mkdir, loc, mode, umask,
+ xdata);
+
+ return 0;
}
int32_t
-ro_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
- dict_t *xdata)
+ro_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (unlink, frame, -1, EROFS, NULL, NULL,
- xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->unlink, loc, xflag,
- xdata);
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(unlink, frame, -1, EROFS, NULL, NULL, xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
- return 0;
+ return 0;
}
-
int
-ro_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
- dict_t *xdata)
+ro_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (rmdir, frame, -1, EROFS, NULL, NULL,
- xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->rmdir, loc, flags,
- xdata);
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(rmdir, frame, -1, EROFS, NULL, NULL, xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rmdir, loc, flags, xdata);
- return 0;
+ return 0;
}
-
int
-ro_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
- loc_t *loc, mode_t umask, dict_t *xdata)
+ro_symlink(call_frame_t *frame, xlator_t *this, const char *linkpath,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (symlink, frame, -1, EROFS, NULL, NULL,
- NULL, NULL, xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->symlink, linkpath,
- loc, umask, xdata);
-
- return 0;
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(symlink, frame, -1, EROFS, NULL, NULL, NULL, NULL,
+ xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->symlink, linkpath, loc, umask,
+ xdata);
+
+ return 0;
}
-
-
int32_t
-ro_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
+ro_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (rename, frame, -1, EROFS, NULL, NULL, NULL,
- NULL, NULL, xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->rename, oldloc,
- newloc, xdata);
-
- return 0;
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(rename, frame, -1, EROFS, NULL, NULL, NULL, NULL,
+ NULL, xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
+
+ return 0;
}
-
int32_t
-ro_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+ro_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (link, frame, -1, EROFS, NULL, NULL, NULL,
- NULL, xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->link, oldloc, newloc,
- xdata);
-
- return 0;
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(link, frame, -1, EROFS, NULL, NULL, NULL, NULL,
+ xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->link,
+ oldloc, newloc, xdata);
+
+ return 0;
}
int32_t
-ro_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+ro_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (create, frame, -1, EROFS, NULL, NULL, NULL,
- NULL, NULL, xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->create, loc, flags,
- mode, umask, fd, xdata);
-
- return 0;
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(create, frame, -1, EROFS, NULL, NULL, NULL, NULL,
+ NULL, xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, flags, mode,
+ umask, fd, xdata);
+
+ return 0;
}
-
static int32_t
-ro_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
- int32_t op_errno, fd_t *fd, dict_t *xdata)
+ro_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
- return 0;
+ STACK_UNWIND_STRICT(open, frame, op_ret, op_errno, fd, xdata);
+ return 0;
}
int32_t
-ro_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- fd_t *fd, dict_t *xdata)
+ro_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this) &&
- (((flags & O_ACCMODE) == O_WRONLY) ||
- ((flags & O_ACCMODE) == O_RDWR))) {
- STACK_UNWIND_STRICT (open, frame, -1, EROFS, NULL, xdata);
- return 0;
- }
-
- STACK_WIND (frame, ro_open_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
- return 0;
+ if (is_readonly_or_worm_enabled(frame, this) &&
+ (((flags & O_ACCMODE) == O_WRONLY) ||
+ ((flags & O_ACCMODE) == O_RDWR))) {
+ STACK_UNWIND_STRICT(open, frame, -1, EROFS, NULL, xdata);
+ return 0;
+ }
+
+ STACK_WIND(frame, ro_open_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
+ return 0;
}
int32_t
-ro_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
- int32_t flags, dict_t *xdata)
+ro_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (fsetxattr, frame, -1, EROFS, xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->fsetxattr, fd, dict,
- flags, xdata);
-
- return 0;
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(fsetxattr, frame, -1, EROFS, xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags,
+ xdata);
+
+ return 0;
}
int32_t
-ro_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
- dict_t *xdata)
+ro_fsyncdir(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+ dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (fsyncdir, frame, -1, EROFS, xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->fsyncdir, fd, flags,
- xdata);
-
- return 0;
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(fsyncdir, frame, -1, EROFS, xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsyncdir, fd, flags, xdata);
+
+ return 0;
}
int32_t
-ro_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
- int32_t count, off_t off, uint32_t flags, struct iobref *iobref,
- dict_t *xdata)
+ro_writev(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
+ int32_t count, off_t off, uint32_t flags, struct iobref *iobref,
+ dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (writev, frame, -1, EROFS, NULL, NULL,
- xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->writev, fd, vector,
- count, off, flags, iobref, xdata);
-
- return 0;
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(writev, frame, -1, EROFS, NULL, NULL, xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector, count, off,
+ flags, iobref, xdata);
+
+ return 0;
}
-
int32_t
-ro_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
- int32_t flags, dict_t *xdata)
+ro_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (setxattr, frame, -1, EROFS, xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->setxattr, loc, dict,
- flags, xdata);
-
- return 0;
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(setxattr, frame, -1, EROFS, xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, loc, dict, flags,
+ xdata);
+
+ return 0;
}
int32_t
-ro_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+ro_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this))
- STACK_UNWIND_STRICT (removexattr, frame, -1, EROFS, xdata);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->removexattr, loc,
- name, xdata);
+ if (is_readonly_or_worm_enabled(frame, this))
+ STACK_UNWIND_STRICT(removexattr, frame, -1, EROFS, xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
- return 0;
+ return 0;
}
diff --git a/xlators/features/read-only/src/read-only-common.h b/xlators/features/read-only/src/read-only-common.h
index ec629bd5660..5561961ffa2 100644
--- a/xlators/features/read-only/src/read-only-common.h
+++ b/xlators/features/read-only/src/read-only-common.h
@@ -7,111 +7,115 @@
later), or the GNU General Public License, version 2 (GPLv2), in all
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "defaults.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
gf_boolean_t
-is_readonly_or_worm_enabled (call_frame_t *frame, xlator_t *this);
+is_readonly_or_worm_enabled(call_frame_t *frame, xlator_t *this);
int32_t
-ro_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
- gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata);
+ro_xattrop(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata);
int32_t
-ro_fxattrop (call_frame_t *frame, xlator_t *this,
- fd_t *fd, gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata);
+ro_fxattrop(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata);
int32_t
-ro_entrylk (call_frame_t *frame, xlator_t *this, const char *volume,
- loc_t *loc, const char *basename, entrylk_cmd cmd,
- entrylk_type type, dict_t *xdata);
+ro_entrylk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata);
int32_t
-ro_fentrylk (call_frame_t *frame, xlator_t *this, const char *volume,
- fd_t *fd, const char *basename, entrylk_cmd cmd, entrylk_type
- type, dict_t *xdata);
+ro_fentrylk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ const char *basename, entrylk_cmd cmd, entrylk_type type,
+ dict_t *xdata);
int32_t
-ro_inodelk (call_frame_t *frame, xlator_t *this, const char *volume,
- loc_t *loc, int32_t cmd, struct gf_flock *lock, dict_t *xdata);
+ro_inodelk(call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc,
+ int32_t cmd, struct gf_flock *lock, dict_t *xdata);
int32_t
-ro_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
- fd_t *fd, int32_t cmd, struct gf_flock *lock, dict_t *xdata);
+ro_finodelk(call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+ int32_t cmd, struct gf_flock *lock, dict_t *xdata);
int32_t
-ro_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int cmd,
- struct gf_flock *flock, dict_t *xdata);
+ro_lk(call_frame_t *frame, xlator_t *this, fd_t *fd, int cmd,
+ struct gf_flock *flock, dict_t *xdata);
int32_t
-ro_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata);
+ro_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *stbuf,
+ int32_t valid, dict_t *xdata);
int32_t
-ro_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iatt *stbuf, int32_t valid, dict_t *xdata);
-
+ro_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
+ int32_t valid, dict_t *xdata);
int32_t
-ro_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset, dict_t *xdata);
+ro_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata);
int32_t
-ro_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset, dict_t *xdata);
+ro_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata);
int
-ro_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- dev_t rdev, mode_t umask, dict_t *xdata);
+ro_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *xdata);
int
-ro_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- mode_t umask, dict_t *xdata);
+ro_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata);
int32_t
-ro_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
- dict_t *xdata);
-
-int
-ro_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ro_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
dict_t *xdata);
+int
+ro_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata);
int
-ro_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
- loc_t *loc, mode_t umask, dict_t *xdata);
+ro_symlink(call_frame_t *frame, xlator_t *this, const char *linkpath,
+ loc_t *loc, mode_t umask, dict_t *xdata);
int32_t
-ro_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc, dict_t *xdata);
+ro_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata);
int32_t
-ro_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc, dict_t *xdata);
+ro_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata);
int32_t
-ro_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata);
+ro_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata);
int32_t
-ro_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- fd_t *fd, dict_t *xdata);
+ro_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata);
int32_t
-ro_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
- int32_t flags, dict_t *xdata);
+ro_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata);
int32_t
-ro_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags, dict_t *xdata);
+ro_fsyncdir(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+ dict_t *xdata);
int32_t
-ro_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
- int32_t count, off_t off, uint32_t flags, struct iobref *iobref, dict_t *xdata);
+ro_writev(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
+ int32_t count, off_t off, uint32_t flags, struct iobref *iobref,
+ dict_t *xdata);
int32_t
-ro_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
- int32_t flags, dict_t *xdata);
+ro_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata);
int32_t
-ro_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata);
+ro_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata);
int32_t
-ro_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
- off_t offset, size_t len, dict_t *xdata);
+ro_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+ off_t offset, size_t len, dict_t *xdata);
diff --git a/xlators/features/read-only/src/read-only-mem-types.h b/xlators/features/read-only/src/read-only-mem-types.h
index 940700a017d..c67d6c02cd0 100644
--- a/xlators/features/read-only/src/read-only-mem-types.h
+++ b/xlators/features/read-only/src/read-only-mem-types.h
@@ -11,10 +11,10 @@
#ifndef __READONLY_MEM_TYPES_H__
#define __READONLY_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_read_only_mem_types_ {
- gf_read_only_mt_priv_t = gf_common_mt_end + 1,
- gf_read_only_mt_end
+ gf_read_only_mt_priv_t = gf_common_mt_end + 1,
+ gf_read_only_mt_end
};
#endif
diff --git a/xlators/features/read-only/src/read-only.c b/xlators/features/read-only/src/read-only.c
index 207b2b1a1d6..48654998e63 100644
--- a/xlators/features/read-only/src/read-only.c
+++ b/xlators/features/read-only/src/read-only.c
@@ -7,131 +7,138 @@
later), or the GNU General Public License, version 2 (GPLv2), in all
cases as published by the Free Software Foundation.
*/
-#include "defaults.h"
#include "read-only-common.h"
#include "read-only-mem-types.h"
#include "read-only.h"
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- ret = xlator_mem_acct_init (this, gf_read_only_mt_end + 1);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR, "Memory accounting "
- "initialization failed.");
+ ret = xlator_mem_acct_init(this, gf_read_only_mt_end + 1);
+ if (ret)
+ gf_log(this->name, GF_LOG_ERROR,
+ "Memory accounting "
+ "initialization failed.");
- return ret;
+ return ret;
}
int32_t
-init (xlator_t *this)
+init(xlator_t *this)
{
- int ret = -1;
- read_only_priv_t *priv = NULL;
+ int ret = -1;
+ read_only_priv_t *priv = NULL;
- if (!this->children || this->children->next) {
- gf_log (this->name, GF_LOG_ERROR,
- "translator not configured with exactly one child");
- return -1;
- }
+ if (!this->children || this->children->next) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "translator not configured with exactly one child");
+ return -1;
+ }
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "dangling volume. check volfile ");
- }
+ if (!this->parents) {
+ gf_log(this->name, GF_LOG_WARNING, "dangling volume. check volfile ");
+ }
- priv = GF_CALLOC (1, sizeof (*priv), gf_read_only_mt_priv_t);
- if (!priv)
- goto out;
+ priv = GF_CALLOC(1, sizeof(*priv), gf_read_only_mt_priv_t);
+ if (!priv)
+ goto out;
- this->private = priv;
+ this->private = priv;
- GF_OPTION_INIT ("read-only", priv->readonly_or_worm_enabled, bool, out);
+ GF_OPTION_INIT("read-only", priv->readonly_or_worm_enabled, bool, out);
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-reconfigure (xlator_t *this, dict_t *options)
+reconfigure(xlator_t *this, dict_t *options)
{
- read_only_priv_t *priv = NULL;
- int ret = -1;
- gf_boolean_t readonly_or_worm_enabled = _gf_false;
+ read_only_priv_t *priv = NULL;
+ int ret = -1;
+ gf_boolean_t readonly_or_worm_enabled = _gf_false;
- priv = this->private;
- GF_ASSERT (priv);
+ priv = this->private;
+ GF_ASSERT(priv);
- GF_OPTION_RECONF ("read-only", readonly_or_worm_enabled, options, bool,
- out);
- priv->readonly_or_worm_enabled = readonly_or_worm_enabled;
- ret = 0;
+ GF_OPTION_RECONF("read-only", readonly_or_worm_enabled, options, bool, out);
+ priv->readonly_or_worm_enabled = readonly_or_worm_enabled;
+ ret = 0;
out:
- gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret);
- return ret;
+ gf_log(this->name, GF_LOG_DEBUG, "returning %d", ret);
+ return ret;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- read_only_priv_t *priv = NULL;
+ read_only_priv_t *priv = NULL;
- priv = this->private;
- if (!priv)
- return;
+ priv = this->private;
+ if (!priv)
+ return;
- this->private = NULL;
- GF_FREE (priv);
+ this->private = NULL;
+ GF_FREE(priv);
- return;
+ return;
}
-
struct xlator_fops fops = {
- .mknod = ro_mknod,
- .mkdir = ro_mkdir,
- .unlink = ro_unlink,
- .rmdir = ro_rmdir,
- .symlink = ro_symlink,
- .rename = ro_rename,
- .link = ro_link,
- .truncate = ro_truncate,
- .open = ro_open,
- .writev = ro_writev,
- .setxattr = ro_setxattr,
- .fsetxattr = ro_fsetxattr,
- .removexattr = ro_removexattr,
- .fsyncdir = ro_fsyncdir,
- .ftruncate = ro_ftruncate,
- .create = ro_create,
- .setattr = ro_setattr,
- .fsetattr = ro_fsetattr,
- .xattrop = ro_xattrop,
- .fxattrop = ro_fxattrop,
- .inodelk = ro_inodelk,
- .finodelk = ro_finodelk,
- .entrylk = ro_entrylk,
- .fentrylk = ro_fentrylk,
- .lk = ro_lk,
- .fallocate = ro_fallocate,
+ .mknod = ro_mknod,
+ .mkdir = ro_mkdir,
+ .unlink = ro_unlink,
+ .rmdir = ro_rmdir,
+ .symlink = ro_symlink,
+ .rename = ro_rename,
+ .link = ro_link,
+ .truncate = ro_truncate,
+ .open = ro_open,
+ .writev = ro_writev,
+ .setxattr = ro_setxattr,
+ .fsetxattr = ro_fsetxattr,
+ .removexattr = ro_removexattr,
+ .fsyncdir = ro_fsyncdir,
+ .ftruncate = ro_ftruncate,
+ .create = ro_create,
+ .setattr = ro_setattr,
+ .fsetattr = ro_fsetattr,
+ .xattrop = ro_xattrop,
+ .fxattrop = ro_fxattrop,
+ .inodelk = ro_inodelk,
+ .finodelk = ro_finodelk,
+ .entrylk = ro_entrylk,
+ .fentrylk = ro_fentrylk,
+ .lk = ro_lk,
+ .fallocate = ro_fallocate,
};
-struct xlator_cbks cbks = {
-};
+struct xlator_cbks cbks = {};
struct volume_options options[] = {
- { .key = {"read-only"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- /*.validate_fn = validate_boolean,*/
- .op_version = {1},
- .flags = OPT_FLAG_SETTABLE,
- .description = "When \"on\", makes a volume read-only. It is turned "
- "\"off\" by default."
- },
- { .key = {NULL}
- },
+ {.key = {"read-only"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ /*.validate_fn = validate_boolean,*/
+ .op_version = {1},
+ .flags = OPT_FLAG_SETTABLE,
+ .description = "When \"on\", makes a volume read-only. It is turned "
+ "\"off\" by default."},
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "read-only",
+ .category = GF_TECH_PREVIEW,
};
diff --git a/xlators/features/read-only/src/read-only.h b/xlators/features/read-only/src/read-only.h
index aae625f5eaf..aced5d3c577 100644
--- a/xlators/features/read-only/src/read-only.h
+++ b/xlators/features/read-only/src/read-only.h
@@ -11,28 +11,27 @@
#ifndef __READONLY_H__
#define __READONLY_H__
-#include "read-only-mem-types.h"
-#include "xlator.h"
-
+#include <stdint.h> // for uint64_t, uint8_t
+#include <sys/time.h> // for time_t
+#include "glusterfs/glusterfs.h" // for gf_boolean_t
typedef struct {
- uint8_t worm : 1;
- uint8_t retain : 1;
- uint8_t legal_hold :1;
- uint8_t ret_mode : 1;
- uint64_t ret_period;
- uint64_t auto_commit_period;
+ uint8_t worm : 1;
+ uint8_t retain : 1;
+ uint8_t legal_hold : 1;
+ uint8_t ret_mode : 1;
+ int64_t ret_period;
+ int64_t auto_commit_period;
} worm_reten_state_t;
-
typedef struct {
- gf_boolean_t readonly_or_worm_enabled;
- gf_boolean_t worm_file;
- gf_boolean_t worm_files_deletable;
- uint64_t reten_period;
- uint64_t com_period;
- char *reten_mode;
- time_t start_time;
+ gf_boolean_t readonly_or_worm_enabled;
+ gf_boolean_t worm_file;
+ gf_boolean_t worm_files_deletable;
+ int64_t reten_period;
+ int64_t com_period;
+ int reten_mode;
+ time_t start_time;
} read_only_priv_t;
#endif
diff --git a/xlators/features/read-only/src/worm-helper.c b/xlators/features/read-only/src/worm-helper.c
index 0770f9d77c4..df45f2a940b 100644
--- a/xlators/features/read-only/src/worm-helper.c
+++ b/xlators/features/read-only/src/worm-helper.c
@@ -9,8 +9,8 @@
*/
#include "read-only-mem-types.h"
#include "read-only.h"
-#include "xlator.h"
-#include "syncop.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/syncop.h>
#include "worm-helper.h"
/*Function to check whether file is read-only.
@@ -18,277 +18,259 @@
* the write protection bits for all the users of the file.
* Return true if all the write bits are disabled,false otherwise*/
gf_boolean_t
-gf_worm_write_disabled (struct iatt *stbuf)
+gf_worm_write_disabled(struct iatt *stbuf)
{
- gf_boolean_t ret = _gf_false;
+ gf_boolean_t ret = _gf_false;
- GF_VALIDATE_OR_GOTO ("worm", stbuf, out);
+ GF_VALIDATE_OR_GOTO("worm", stbuf, out);
- if (stbuf->ia_prot.owner.write == 0 &&
- stbuf->ia_prot.group.write == 0 &&
- stbuf->ia_prot.other.write == 0)
- ret = _gf_true;
+ if (stbuf->ia_prot.owner.write == 0 && stbuf->ia_prot.group.write == 0 &&
+ stbuf->ia_prot.other.write == 0)
+ ret = _gf_true;
out:
- return ret;
+ return ret;
}
-
int32_t
-worm_init_state (xlator_t *this, gf_boolean_t fop_with_fd, void *file_ptr)
+worm_init_state(xlator_t *this, gf_boolean_t fop_with_fd, void *file_ptr)
{
- int ret = -1;
- uint64_t start_time = 0;
- dict_t *dict = NULL;
-
- GF_VALIDATE_OR_GOTO ("worm", this, out);
- GF_VALIDATE_OR_GOTO (this->name, file_ptr, out);
-
- start_time = time (NULL);
- dict = dict_new ();
- if (!dict) {
- gf_log (this->name, GF_LOG_ERROR, "Error creating the dict");
- goto out;
- }
- ret = dict_set_uint64 (dict, "trusted.start_time", start_time);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "Error in setting the dict");
- goto out;
- }
- if (fop_with_fd)
- ret = syncop_fsetxattr (this, (fd_t *)file_ptr, dict, 0,
- NULL, NULL);
- else
- ret = syncop_setxattr (this, (loc_t *)file_ptr, dict, 0, NULL,
- NULL);
+ int ret = -1;
+ uint64_t start_time = 0;
+ dict_t *dict = NULL;
+
+ GF_VALIDATE_OR_GOTO("worm", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, file_ptr, out);
+
+ start_time = gf_time();
+ dict = dict_new();
+ if (!dict) {
+ gf_log(this->name, GF_LOG_ERROR, "Error creating the dict");
+ goto out;
+ }
+ ret = dict_set_uint64(dict, "trusted.start_time", start_time);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR, "Error in setting the dict");
+ goto out;
+ }
+ if (fop_with_fd)
+ ret = syncop_fsetxattr(this, (fd_t *)file_ptr, dict, 0, NULL, NULL);
+ else
+ ret = syncop_setxattr(this, (loc_t *)file_ptr, dict, 0, NULL, NULL);
out:
- if (dict)
- dict_unref (dict);
- return ret;
+ if (dict)
+ dict_unref(dict);
+ return ret;
}
-
/*Function to set the retention state for a file.
* It loads the WORM/Retention state into the retention_state pointer.*/
int32_t
-worm_set_state (xlator_t *this, gf_boolean_t fop_with_fd, void *file_ptr,
- worm_reten_state_t *retention_state, struct iatt *stbuf)
+worm_set_state(xlator_t *this, gf_boolean_t fop_with_fd, void *file_ptr,
+ worm_reten_state_t *retention_state, struct iatt *stbuf)
{
- read_only_priv_t *priv = NULL;
- struct iatt stpre = {0,};
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO ("worm", this, out);
- GF_VALIDATE_OR_GOTO (this->name, file_ptr, out);
- GF_VALIDATE_OR_GOTO (this->name, retention_state, out);
- GF_VALIDATE_OR_GOTO (this->name, stbuf, out);
-
- priv = this->private;
- GF_ASSERT (priv);
- retention_state->worm = 1;
- retention_state->retain = 1;
- retention_state->legal_hold = 0;
- if (strcmp (priv->reten_mode, "relax") == 0)
- retention_state->ret_mode = 0;
- else
- retention_state->ret_mode = 1;
- retention_state->ret_period = priv->reten_period;
- retention_state->auto_commit_period = priv->com_period;
- if (fop_with_fd)
- ret = syncop_fstat (this, (fd_t *)file_ptr, &stpre, NULL, NULL);
- else
- ret = syncop_stat (this, (loc_t *)file_ptr, &stpre, NULL, NULL);
- if (ret)
- goto out;
- stbuf->ia_mtime = stpre.ia_mtime;
- stbuf->ia_atime = time (NULL) + retention_state->ret_period;
-
- if (fop_with_fd)
- ret = syncop_fsetattr (this, (fd_t *)file_ptr, stbuf,
- GF_SET_ATTR_ATIME, NULL, NULL,
- NULL, NULL);
- else
- ret = syncop_setattr (this, (loc_t *)file_ptr, stbuf,
- GF_SET_ATTR_ATIME, NULL, NULL,
- NULL, NULL);
- if (ret)
- goto out;
-
- ret = gf_worm_set_xattr (this, retention_state, fop_with_fd, file_ptr);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "Error setting xattr");
- goto out;
- }
- ret = 0;
+ read_only_priv_t *priv = NULL;
+ struct iatt stpre = {
+ 0,
+ };
+ int ret = -1;
+
+ GF_VALIDATE_OR_GOTO("worm", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, file_ptr, out);
+ GF_VALIDATE_OR_GOTO(this->name, retention_state, out);
+ GF_VALIDATE_OR_GOTO(this->name, stbuf, out);
+
+ priv = this->private;
+ GF_ASSERT(priv);
+ retention_state->worm = 1;
+ retention_state->retain = 1;
+ retention_state->legal_hold = 0;
+ retention_state->ret_mode = priv->reten_mode;
+ retention_state->ret_period = priv->reten_period;
+ retention_state->auto_commit_period = priv->com_period;
+ if (fop_with_fd)
+ ret = syncop_fstat(this, (fd_t *)file_ptr, &stpre, NULL, NULL);
+ else
+ ret = syncop_stat(this, (loc_t *)file_ptr, &stpre, NULL, NULL);
+ if (ret)
+ goto out;
+ stbuf->ia_mtime = stpre.ia_mtime;
+ stbuf->ia_atime = gf_time() + retention_state->ret_period;
+
+ if (fop_with_fd)
+ ret = syncop_fsetattr(this, (fd_t *)file_ptr, stbuf, GF_SET_ATTR_ATIME,
+ NULL, NULL, NULL, NULL);
+ else
+ ret = syncop_setattr(this, (loc_t *)file_ptr, stbuf, GF_SET_ATTR_ATIME,
+ NULL, NULL, NULL, NULL);
+ if (ret)
+ goto out;
+
+ ret = gf_worm_set_xattr(this, retention_state, fop_with_fd, file_ptr);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR, "Error setting xattr");
+ goto out;
+ }
+ ret = 0;
out:
- return ret;
+ return ret;
}
-
/*This function gets the state of the WORM/Retention xattr and loads it in the
* dict pointer.*/
int32_t
-worm_get_state (xlator_t *this, gf_boolean_t fop_with_fd, void *file_ptr,
- worm_reten_state_t *reten_state)
+worm_get_state(xlator_t *this, gf_boolean_t fop_with_fd, void *file_ptr,
+ worm_reten_state_t *reten_state)
{
- dict_t *dict = NULL;
- char *val = NULL;
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO ("worm", this, out);
- GF_VALIDATE_OR_GOTO (this->name, file_ptr, out);
- GF_VALIDATE_OR_GOTO (this->name, reten_state, out);
-
- if (fop_with_fd)
- ret = syncop_fgetxattr (this, (fd_t *)file_ptr, &dict,
- "trusted.reten_state", NULL, NULL);
- else
- ret = syncop_getxattr (this, (loc_t *)file_ptr, &dict,
- "trusted.reten_state", NULL, NULL);
- if (ret < 0 || !dict) {
- ret = -1;
- goto out;
- }
- ret = dict_get_str (dict, "trusted.reten_state", &val);
- if (ret) {
- ret = -2;
- gf_log (this->name, GF_LOG_ERROR, "Empty val");
- }
- gf_worm_deserialize_state (val, reten_state);
+ dict_t *dict = NULL;
+ char *val = NULL;
+ int ret = -1;
+
+ GF_VALIDATE_OR_GOTO("worm", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, file_ptr, out);
+ GF_VALIDATE_OR_GOTO(this->name, reten_state, out);
+
+ if (fop_with_fd)
+ ret = syncop_fgetxattr(this, (fd_t *)file_ptr, &dict,
+ "trusted.reten_state", NULL, NULL);
+ else
+ ret = syncop_getxattr(this, (loc_t *)file_ptr, &dict,
+ "trusted.reten_state", NULL, NULL);
+ if (ret < 0 || !dict) {
+ ret = -1;
+ goto out;
+ }
+ ret = dict_get_str(dict, "trusted.reten_state", &val);
+ if (ret) {
+ ret = -2;
+ gf_log(this->name, GF_LOG_ERROR, "Empty val");
+ }
+ gf_worm_deserialize_state(val, reten_state);
out:
- if (dict)
- dict_unref (dict);
- return ret;
+ if (dict)
+ dict_unref(dict);
+ return ret;
}
-
/*Function to lookup the current state of the WORM/Retention profile.
* Based on the retain value and the access time of the file, the transition
* from WORM/Retention to WORM is made.*/
void
-gf_worm_state_lookup (xlator_t *this, gf_boolean_t fop_with_fd, void *file_ptr,
- worm_reten_state_t *reten_state, struct iatt *stbuf)
+gf_worm_state_lookup(xlator_t *this, gf_boolean_t fop_with_fd, void *file_ptr,
+ worm_reten_state_t *reten_state, struct iatt *stbuf)
{
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO ("worm", this, out);
- GF_VALIDATE_OR_GOTO (this->name, file_ptr, out);
- GF_VALIDATE_OR_GOTO (this->name, reten_state, out);
- GF_VALIDATE_OR_GOTO (this->name, stbuf, out);
-
- stbuf->ia_atime -= reten_state->ret_period;
- reten_state->retain = 0;
- reten_state->ret_period = 0;
- reten_state->auto_commit_period = 0;
- ret = gf_worm_set_xattr (this, reten_state, fop_with_fd, file_ptr);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "Error setting xattr");
- goto out;
- }
-
- if (fop_with_fd)
- ret = syncop_fsetattr (this, (fd_t *)file_ptr, stbuf,
- GF_SET_ATTR_ATIME, NULL, NULL,
- NULL, NULL);
- else
- ret = syncop_setattr (this, (loc_t *)file_ptr, stbuf,
- GF_SET_ATTR_ATIME, NULL, NULL,
- NULL, NULL);
- if (ret)
- goto out;
- gf_log (this->name, GF_LOG_INFO, "Retention state reset");
+ int ret = -1;
+
+ GF_VALIDATE_OR_GOTO("worm", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, file_ptr, out);
+ GF_VALIDATE_OR_GOTO(this->name, reten_state, out);
+ GF_VALIDATE_OR_GOTO(this->name, stbuf, out);
+
+ stbuf->ia_atime -= reten_state->ret_period;
+ reten_state->retain = 0;
+ reten_state->ret_period = 0;
+ reten_state->auto_commit_period = 0;
+ ret = gf_worm_set_xattr(this, reten_state, fop_with_fd, file_ptr);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR, "Error setting xattr");
+ goto out;
+ }
+
+ if (fop_with_fd)
+ ret = syncop_fsetattr(this, (fd_t *)file_ptr, stbuf, GF_SET_ATTR_ATIME,
+ NULL, NULL, NULL, NULL);
+ else
+ ret = syncop_setattr(this, (loc_t *)file_ptr, stbuf, GF_SET_ATTR_ATIME,
+ NULL, NULL, NULL, NULL);
+ if (ret)
+ goto out;
+ gf_log(this->name, GF_LOG_INFO, "Retention state reset");
out:
- return;
+ return;
}
-
/*This function serializes and stores the WORM/Retention state of a file in an
* uint64_t variable by setting the bits using the bitwise operations.*/
void
-gf_worm_serialize_state (worm_reten_state_t *reten_state, char *val)
+gf_worm_serialize_state(worm_reten_state_t *reten_state, char *val)
{
- uint32_t state = 0;
+ uint32_t state = 0;
- GF_VALIDATE_OR_GOTO ("worm", reten_state, out);
- GF_VALIDATE_OR_GOTO ("worm", val, out);
+ GF_VALIDATE_OR_GOTO("worm", reten_state, out);
+ GF_VALIDATE_OR_GOTO("worm", val, out);
- state |= reten_state->worm << 0;
- state |= reten_state->retain << 1;
- state |= reten_state->legal_hold << 2;
- state |= reten_state->ret_mode << 3;
- sprintf (val, "%d/%"PRIu64"/%"PRIu64, state, reten_state->ret_period,
- reten_state->auto_commit_period);
+ state |= reten_state->worm << 0;
+ state |= reten_state->retain << 1;
+ state |= reten_state->legal_hold << 2;
+ state |= reten_state->ret_mode << 3;
+ sprintf(val, "%d/%" PRIu64 "/%" PRIu64, state, reten_state->ret_period,
+ reten_state->auto_commit_period);
out:
- return;
+ return;
}
-
/*This function deserializes the data stored in the xattr of the file and loads
* the value to the reten_state structure.*/
void
-gf_worm_deserialize_state (char *val, worm_reten_state_t *reten_state)
+gf_worm_deserialize_state(char *val, worm_reten_state_t *reten_state)
{
- char *token = NULL;
- uint32_t state = 0;
-
- GF_VALIDATE_OR_GOTO ("worm", val, out);
- GF_VALIDATE_OR_GOTO ("worm", reten_state, out);
-
- token = strtok (val, "/");
- state = atoi (token);
- reten_state->worm = (state >> 0) & 1;
- reten_state->retain = (state >> 1) & 1;
- reten_state->legal_hold = (state >> 2) & 1;
- reten_state->ret_mode = (state >> 3) & 1;
- token = strtok (NULL, "/");
- reten_state->ret_period = atoi (token);
- token = strtok (NULL, "/");
- reten_state->auto_commit_period = atoi (token);
+ char *token = NULL;
+ uint32_t state = 0;
+
+ GF_VALIDATE_OR_GOTO("worm", val, out);
+ GF_VALIDATE_OR_GOTO("worm", reten_state, out);
+
+ token = strtok(val, "/");
+ state = atoi(token);
+ reten_state->worm = (state >> 0) & 1;
+ reten_state->retain = (state >> 1) & 1;
+ reten_state->legal_hold = (state >> 2) & 1;
+ reten_state->ret_mode = (state >> 3) & 1;
+ token = strtok(NULL, "/");
+ reten_state->ret_period = atoi(token);
+ token = strtok(NULL, "/");
+ reten_state->auto_commit_period = atoi(token);
out:
- return;
+ return;
}
-
/*Function to set the xattr for a file.
* If the xattr is already present then it will replace that.*/
int32_t
-gf_worm_set_xattr (xlator_t *this, worm_reten_state_t *reten_state,
- gf_boolean_t fop_with_fd, void *file_ptr)
+gf_worm_set_xattr(xlator_t *this, worm_reten_state_t *reten_state,
+ gf_boolean_t fop_with_fd, void *file_ptr)
{
- char val[100] = "";
- int ret = -1;
- dict_t *dict = NULL;
-
- GF_VALIDATE_OR_GOTO ("worm", this, out);
- GF_VALIDATE_OR_GOTO (this->name, reten_state, out);
- GF_VALIDATE_OR_GOTO (this->name, file_ptr, out);
-
- gf_worm_serialize_state (reten_state, val);
- dict = dict_new ();
- if (!dict) {
- gf_log (this->name, GF_LOG_ERROR, "Error creating the dict");
- goto out;
- }
- ret = dict_set_str (dict, "trusted.reten_state", val);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "Error in setting the dict");
- goto out;
- }
- if (fop_with_fd)
- ret = syncop_fsetxattr (this, (fd_t *)file_ptr, dict, 0,
- NULL, NULL);
- else
- ret = syncop_setxattr (this, (loc_t *)file_ptr, dict, 0, NULL,
- NULL);
+ char val[100] = "";
+ int ret = -1;
+ dict_t *dict = NULL;
+
+ GF_VALIDATE_OR_GOTO("worm", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, reten_state, out);
+ GF_VALIDATE_OR_GOTO(this->name, file_ptr, out);
+
+ gf_worm_serialize_state(reten_state, val);
+ dict = dict_new();
+ if (!dict) {
+ gf_log(this->name, GF_LOG_ERROR, "Error creating the dict");
+ goto out;
+ }
+ ret = dict_set_str(dict, "trusted.reten_state", val);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR, "Error in setting the dict");
+ goto out;
+ }
+ if (fop_with_fd)
+ ret = syncop_fsetxattr(this, (fd_t *)file_ptr, dict, 0, NULL, NULL);
+ else
+ ret = syncop_setxattr(this, (loc_t *)file_ptr, dict, 0, NULL, NULL);
out:
- if (dict)
- dict_unref (dict);
- return ret;
+ if (dict)
+ dict_unref(dict);
+ return ret;
}
-
-/*This function checks whether a file's timeout is happend for the state
+/*This function checks whether a file's timeout is happened for the state
* transition and if yes, then it will do the transition from the current state
* to the appropriate state. It also decides whether to continue or to block
* the FOP.
@@ -299,115 +281,115 @@ out:
* 2: Blocks the FOP if any operation fails while doing the state transition or
* fails to get the state of the file.*/
int
-gf_worm_state_transition (xlator_t *this, gf_boolean_t fop_with_fd,
- void *file_ptr, glusterfs_fop_t op)
+gf_worm_state_transition(xlator_t *this, gf_boolean_t fop_with_fd,
+ void *file_ptr, glusterfs_fop_t op)
{
- int op_errno = EROFS;
- int ret = -1;
- uint64_t com_period = 0;
- uint64_t start_time = 0;
- dict_t *dict = NULL;
- worm_reten_state_t reten_state = {0,};
- read_only_priv_t *priv = NULL;
- struct iatt stbuf = {0,};
-
- priv = this->private;
- GF_ASSERT (priv);
-
- if (fop_with_fd)
- ret = syncop_fgetxattr (this, (fd_t *)file_ptr, &dict,
- "trusted.start_time", NULL, NULL);
- else
- ret = syncop_getxattr (this, (loc_t *)file_ptr, &dict,
- "trusted.start_time", NULL, NULL);
- if (ret < 0 || !dict) {
- op_errno = ret;
- gf_msg (this->name, GF_LOG_ERROR, -ret, 0,
- "Error getting xattr");
- goto out;
- }
- ret = dict_get_uint64 (dict, "trusted.start_time", &start_time);
- if (ret) {
- op_errno = ret;
- gf_msg (this->name, GF_LOG_ERROR, -ret, 0,
- "Error getting start time");
- goto out;
- }
-
- com_period = priv->com_period;
- if (fop_with_fd)
- ret = syncop_fstat (this, (fd_t *)file_ptr, &stbuf, NULL, NULL);
- else
- ret = syncop_stat (this, (loc_t *)file_ptr, &stbuf, NULL, NULL);
- if (ret) {
+ int op_errno = EROFS;
+ int ret = -1;
+ time_t now = 0;
+ uint64_t com_period = 0;
+ uint64_t start_time = 0;
+ dict_t *dict = NULL;
+ worm_reten_state_t reten_state = {
+ 0,
+ };
+ read_only_priv_t *priv = NULL;
+ struct iatt stbuf = {
+ 0,
+ };
+
+ priv = this->private;
+ GF_ASSERT(priv);
+
+ if (fop_with_fd)
+ ret = syncop_fgetxattr(this, (fd_t *)file_ptr, &dict,
+ "trusted.start_time", NULL, NULL);
+ else
+ ret = syncop_getxattr(this, (loc_t *)file_ptr, &dict,
+ "trusted.start_time", NULL, NULL);
+ if (ret < 0 || !dict) {
+ op_errno = ret;
+ gf_msg(this->name, GF_LOG_ERROR, -ret, 0, "Error getting xattr");
+ goto out;
+ }
+ ret = dict_get_uint64(dict, "trusted.start_time", &start_time);
+ if (ret) {
+ op_errno = ret;
+ gf_msg(this->name, GF_LOG_ERROR, -ret, 0, "Error getting start time");
+ goto out;
+ }
+
+ com_period = priv->com_period;
+ if (fop_with_fd)
+ ret = syncop_fstat(this, (fd_t *)file_ptr, &stbuf, NULL, NULL);
+ else
+ ret = syncop_stat(this, (loc_t *)file_ptr, &stbuf, NULL, NULL);
+ if (ret) {
+ op_errno = ret;
+ gf_msg(this->name, GF_LOG_ERROR, -ret, 0, "Error getting file stat");
+ goto out;
+ }
+
+ ret = worm_get_state(this, fop_with_fd, file_ptr, &reten_state);
+ if (ret == -2) {
+ op_errno = ret;
+ gf_msg(this->name, GF_LOG_ERROR, -ret, 0,
+ "Error getting worm/retention state");
+ goto out;
+ }
+
+ now = gf_time();
+
+ if (ret == -1 && (now - start_time) >= com_period) {
+ if ((now - stbuf.ia_mtime) >= com_period) {
+ ret = worm_set_state(this, fop_with_fd, file_ptr, &reten_state,
+ &stbuf);
+ if (ret) {
op_errno = ret;
- gf_msg (this->name, GF_LOG_ERROR, -ret, 0,
- "Error getting file stat");
- goto out;
- }
-
- ret = worm_get_state (this, fop_with_fd, file_ptr, &reten_state);
- if (ret == -2) {
- op_errno = ret;
- gf_msg (this->name, GF_LOG_ERROR, -ret, 0,
- "Error getting worm/retention state");
- goto out;
- }
-
- if (ret == -1 && (time (NULL) - start_time) >= com_period) {
- if ((time (NULL) - stbuf.ia_mtime) >= com_period) {
- ret = worm_set_state(this, fop_with_fd, file_ptr,
- &reten_state, &stbuf);
- if (ret) {
- op_errno = ret;
- gf_msg (this->name, GF_LOG_ERROR, -ret, 0,
- "Error setting worm/retention state");
- goto out;
- }
- goto out;
- } else {
- op_errno = 0;
- goto out;
- }
- } else if (ret == -1 && (time (NULL) - start_time)
- < com_period) {
- op_errno = 0;
- goto out;
- } else if (reten_state.retain &&
- ((time (NULL) >= stbuf.ia_atime))) {
- gf_worm_state_lookup (this, fop_with_fd, file_ptr,
- &reten_state, &stbuf);
- }
- if (reten_state.worm && !reten_state.retain &&
- priv->worm_files_deletable && op == GF_FOP_UNLINK) {
- op_errno = 0;
+ gf_msg(this->name, GF_LOG_ERROR, -ret, 0,
+ "Error setting worm/retention state");
goto out;
+ }
+ goto out;
+ } else {
+ op_errno = 0;
+ goto out;
}
+ } else if (ret == -1 && (now - start_time) < com_period) {
+ op_errno = 0;
+ goto out;
+ } else if (reten_state.retain && ((now >= stbuf.ia_atime))) {
+ gf_worm_state_lookup(this, fop_with_fd, file_ptr, &reten_state, &stbuf);
+ }
+ if (reten_state.worm && !reten_state.retain && priv->worm_files_deletable &&
+ op == GF_FOP_UNLINK) {
+ op_errno = 0;
+ goto out;
+ }
out:
- if (dict)
- dict_unref (dict);
- return op_errno;
+ if (dict)
+ dict_unref(dict);
+ return op_errno;
}
-
/*Function to check whether a file is independently WORMed (i.e., file level
* WORM is set on the file). */
int32_t
-is_wormfile (xlator_t *this, gf_boolean_t fop_with_fd, void *file_ptr)
+is_wormfile(xlator_t *this, gf_boolean_t fop_with_fd, void *file_ptr)
{
- int ret = -1;
- dict_t *dict = NULL;
-
- if (fop_with_fd)
- ret = syncop_fgetxattr (this, (fd_t *)file_ptr, &dict,
- "trusted.worm_file", NULL, NULL);
- else
- ret = syncop_getxattr (this, (loc_t *)file_ptr, &dict,
- "trusted.worm_file", NULL, NULL);
- if (dict) {
- ret = 0;
- dict_unref (dict);
- }
- return ret;
+ int ret = -1;
+ dict_t *dict = NULL;
+
+ if (fop_with_fd)
+ ret = syncop_fgetxattr(this, (fd_t *)file_ptr, &dict,
+ "trusted.worm_file", NULL, NULL);
+ else
+ ret = syncop_getxattr(this, (loc_t *)file_ptr, &dict,
+ "trusted.worm_file", NULL, NULL);
+ if (dict) {
+ ret = 0;
+ dict_unref(dict);
+ }
+ return ret;
}
diff --git a/xlators/features/read-only/src/worm-helper.h b/xlators/features/read-only/src/worm-helper.h
index 745df8294c3..b42f8d2b40c 100644
--- a/xlators/features/read-only/src/worm-helper.h
+++ b/xlators/features/read-only/src/worm-helper.h
@@ -8,30 +8,37 @@
cases as published by the Free Software Foundation.
*/
-gf_boolean_t gf_worm_write_disabled (struct iatt *stbuf);
+gf_boolean_t
+gf_worm_write_disabled(struct iatt *stbuf);
-int32_t worm_init_state (xlator_t *this, gf_boolean_t fop_with_fd,
- void *file_ptr);
+int32_t
+worm_init_state(xlator_t *this, gf_boolean_t fop_with_fd, void *file_ptr);
-int32_t worm_set_state (xlator_t *this, gf_boolean_t fop_with_fd,
- void *file_ptr, worm_reten_state_t *retention_state,
- struct iatt *stbuf);
+int32_t
+worm_set_state(xlator_t *this, gf_boolean_t fop_with_fd, void *file_ptr,
+ worm_reten_state_t *retention_state, struct iatt *stbuf);
-int32_t worm_get_state (xlator_t *this, gf_boolean_t fop_with_fd,
- void *file_ptr, worm_reten_state_t *reten_state);
+int32_t
+worm_get_state(xlator_t *this, gf_boolean_t fop_with_fd, void *file_ptr,
+ worm_reten_state_t *reten_state);
-void gf_worm_state_lookup (xlator_t *this, gf_boolean_t fop_with_fd,
- void *file_ptr, worm_reten_state_t *reten_state,
- struct iatt *stbuf);
+void
+gf_worm_state_lookup(xlator_t *this, gf_boolean_t fop_with_fd, void *file_ptr,
+ worm_reten_state_t *reten_state, struct iatt *stbuf);
-void gf_worm_serialize_state (worm_reten_state_t *reten_state, char *val);
+void
+gf_worm_serialize_state(worm_reten_state_t *reten_state, char *val);
-void gf_worm_deserialize_state (char *val, worm_reten_state_t *reten_state);
+void
+gf_worm_deserialize_state(char *val, worm_reten_state_t *reten_state);
-int32_t gf_worm_set_xattr (xlator_t *this, worm_reten_state_t *reten_state,
- gf_boolean_t fop_with_fd, void *file_ptr);
+int32_t
+gf_worm_set_xattr(xlator_t *this, worm_reten_state_t *reten_state,
+ gf_boolean_t fop_with_fd, void *file_ptr);
-int gf_worm_state_transition (xlator_t *this, gf_boolean_t fop_with_fd,
- void *file_ptr, glusterfs_fop_t op);
+int
+gf_worm_state_transition(xlator_t *this, gf_boolean_t fop_with_fd,
+ void *file_ptr, glusterfs_fop_t op);
-int32_t is_wormfile (xlator_t *this, gf_boolean_t fop_with_fd, void *file_ptr);
+int32_t
+is_wormfile(xlator_t *this, gf_boolean_t fop_with_fd, void *file_ptr);
diff --git a/xlators/features/read-only/src/worm.c b/xlators/features/read-only/src/worm.c
index 088dd5a1b23..1cc5526d5cd 100644
--- a/xlators/features/read-only/src/worm.c
+++ b/xlators/features/read-only/src/worm.c
@@ -7,679 +7,716 @@
later), or the GNU General Public License, version 2 (GPLv2), in all
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "defaults.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/defaults.h>
#include "read-only-common.h"
#include "read-only-mem-types.h"
#include "read-only.h"
-#include "syncop.h"
+#include <glusterfs/syncop.h>
#include "worm-helper.h"
-
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- ret = xlator_mem_acct_init (this, gf_read_only_mt_end + 1);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR, "Memory accounting "
- "initialization failed.");
+ ret = xlator_mem_acct_init(this, gf_read_only_mt_end + 1);
+ if (ret)
+ gf_log(this->name, GF_LOG_ERROR,
+ "Memory accounting "
+ "initialization failed.");
- return ret;
+ return ret;
}
-
static int32_t
-worm_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- fd_t *fd, dict_t *xdata)
+worm_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
{
- if (is_readonly_or_worm_enabled (frame, this) &&
- (flags & (O_WRONLY | O_RDWR | O_APPEND | O_TRUNC))) {
- STACK_UNWIND_STRICT (open, frame, -1, EROFS, NULL, NULL);
- return 0;
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
+ if (is_readonly_or_worm_enabled(frame, this) &&
+ (flags & (O_WRONLY | O_RDWR | O_APPEND | O_TRUNC))) {
+ STACK_UNWIND_STRICT(open, frame, -1, EROFS, NULL, NULL);
return 0;
-}
+ }
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->open,
+ loc, flags, fd, xdata);
+ return 0;
+}
static int32_t
-worm_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
+worm_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- int op_errno = EROFS;
- read_only_priv_t *priv = NULL;
-
- priv = this->private;
- GF_ASSERT (priv);
- if (is_readonly_or_worm_enabled (frame, this))
- goto out;
- if (!priv->worm_file || (frame->root->pid < 0)) {
- op_errno = 0;
- goto out;
- }
+ int op_errno = EROFS;
+ read_only_priv_t *priv = NULL;
+
+ priv = this->private;
+ GF_ASSERT(priv);
+ if (is_readonly_or_worm_enabled(frame, this))
+ goto out;
+ if (!priv->worm_file || (frame->root->pid < 0)) {
+ op_errno = 0;
+ goto out;
+ }
- gf_uuid_copy (oldloc->gfid, oldloc->inode->gfid);
- if (is_wormfile (this, _gf_false, oldloc)) {
- op_errno = 0;
- goto out;
- }
- op_errno = gf_worm_state_transition (this, _gf_false, oldloc,
- GF_FOP_LINK);
+ gf_uuid_copy(oldloc->gfid, oldloc->inode->gfid);
+ if (is_wormfile(this, _gf_false, oldloc)) {
+ op_errno = 0;
+ goto out;
+ }
+ op_errno = gf_worm_state_transition(this, _gf_false, oldloc, GF_FOP_LINK);
out:
- if (op_errno) {
- if (op_errno < 0)
- op_errno = EROFS;
- STACK_UNWIND_STRICT (link, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL);
- }
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->link,
- oldloc, newloc, xdata);
- return 0;
+ if (op_errno) {
+ if (op_errno < 0)
+ op_errno = EROFS;
+ STACK_UNWIND_STRICT(link, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
+ } else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->link,
+ oldloc, newloc, xdata);
+ return 0;
}
-
static int32_t
-worm_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- dict_t *xdata)
+worm_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ dict_t *xdata)
{
- int op_errno = EROFS;
- read_only_priv_t *priv = NULL;
-
- priv = this->private;
- GF_ASSERT (priv);
- if (is_readonly_or_worm_enabled (frame, this)) {
- goto out;
- }
- if (!priv->worm_file || (frame->root->pid < 0)) {
- op_errno = 0;
- goto out;
- }
+ int op_errno = EROFS;
+ read_only_priv_t *priv = NULL;
+
+ priv = this->private;
+ GF_ASSERT(priv);
+ if (is_readonly_or_worm_enabled(frame, this)) {
+ goto out;
+ }
+ if (!priv->worm_file || (frame->root->pid < 0)) {
+ op_errno = 0;
+ goto out;
+ }
- gf_uuid_copy (loc->gfid, loc->inode->gfid);
- if (is_wormfile (this, _gf_false, loc)) {
- op_errno = 0;
- goto out;
- }
- op_errno = gf_worm_state_transition (this, _gf_false, loc,
- GF_FOP_UNLINK);
+ gf_uuid_copy(loc->gfid, loc->inode->gfid);
+ if (is_wormfile(this, _gf_false, loc)) {
+ op_errno = 0;
+ goto out;
+ }
+ op_errno = gf_worm_state_transition(this, _gf_false, loc, GF_FOP_UNLINK);
out:
- if (op_errno) {
- if (op_errno < 0)
- op_errno = EROFS;
- STACK_UNWIND_STRICT (unlink, frame, -1, op_errno, NULL, NULL,
- NULL);
- }
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->unlink,
- loc, flags, xdata);
- return 0;
+ if (op_errno) {
+ if (op_errno < 0)
+ op_errno = EROFS;
+ STACK_UNWIND_STRICT(unlink, frame, -1, op_errno, NULL, NULL, NULL);
+ } else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, flags, xdata);
+ return 0;
}
-
static int32_t
-worm_rename (call_frame_t *frame, xlator_t *this,
- loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+worm_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- int op_errno = EROFS;
- read_only_priv_t *priv = NULL;
+ int op_errno = EROFS;
+ read_only_priv_t *priv = NULL;
+
+ priv = this->private;
+ GF_ASSERT(priv);
+ if (is_readonly_or_worm_enabled(frame, this))
+ goto out;
+ if (!priv->worm_file || (frame->root->pid < 0)) {
+ op_errno = 0;
+ goto out;
+ }
- priv = this->private;
- GF_ASSERT (priv);
- if (is_readonly_or_worm_enabled (frame, this))
- goto out;
- if (!priv->worm_file || (frame->root->pid < 0)) {
+ gf_uuid_copy(oldloc->gfid, oldloc->inode->gfid);
+ if (is_wormfile(this, _gf_false, oldloc)) {
+ op_errno = 0;
+ goto check_newloc;
+ }
+ op_errno = gf_worm_state_transition(this, _gf_false, oldloc, GF_FOP_RENAME);
+
+ if (op_errno == 0) {
+ check_newloc:
+ if (newloc->inode != NULL) {
+ gf_uuid_copy(newloc->gfid, newloc->inode->gfid);
+ if (is_wormfile(this, _gf_false, newloc)) {
op_errno = 0;
goto out;
+ }
+ op_errno = gf_worm_state_transition(this, _gf_false, newloc,
+ GF_FOP_RENAME);
}
-
- gf_uuid_copy (oldloc->gfid, oldloc->inode->gfid);
- if (is_wormfile (this, _gf_false, oldloc)) {
- op_errno = 0;
- goto check_newloc;
- }
- op_errno = gf_worm_state_transition (this, _gf_false, oldloc,
- GF_FOP_RENAME);
-
- if (op_errno == 0) {
-check_newloc:
- if (newloc->inode != NULL) {
- gf_uuid_copy (newloc->gfid, newloc->inode->gfid);
- if (is_wormfile (this, _gf_false, newloc)) {
- op_errno = 0;
- goto out;
- }
- op_errno = gf_worm_state_transition (this, _gf_false,
- newloc,
- GF_FOP_RENAME);
- }
- }
+ }
out:
- if (op_errno) {
- if (op_errno < 0)
- op_errno = EROFS;
- STACK_UNWIND_STRICT (rename, frame, -1, op_errno, NULL,
- NULL, NULL, NULL, NULL, NULL);
- }
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->rename,
- oldloc, newloc, xdata);
- return 0;
+ if (op_errno) {
+ if (op_errno < 0)
+ op_errno = EROFS;
+ STACK_UNWIND_STRICT(rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL, NULL);
+ } else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
+ return 0;
}
-
static int32_t
-worm_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
- dict_t *xdata)
+worm_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- int op_errno = EROFS;
- read_only_priv_t *priv = NULL;
-
- priv = this->private;
- GF_ASSERT (priv);
- if (is_readonly_or_worm_enabled (frame, this))
- goto out;
- if (!priv->worm_file || (frame->root->pid < 0)) {
- op_errno = 0;
- goto out;
- }
+ int op_errno = EROFS;
+ read_only_priv_t *priv = NULL;
+
+ priv = this->private;
+ GF_ASSERT(priv);
+ if (is_readonly_or_worm_enabled(frame, this))
+ goto out;
+ if (!priv->worm_file || (frame->root->pid < 0)) {
+ op_errno = 0;
+ goto out;
+ }
- if (is_wormfile (this, _gf_false, loc)) {
- op_errno = 0;
- goto out;
- }
- op_errno = gf_worm_state_transition (this, _gf_false, loc,
- GF_FOP_TRUNCATE);
+ if (is_wormfile(this, _gf_false, loc)) {
+ op_errno = 0;
+ goto out;
+ }
+ op_errno = gf_worm_state_transition(this, _gf_false, loc, GF_FOP_TRUNCATE);
out:
- if (op_errno) {
- if (op_errno < 0)
- op_errno = EROFS;
- STACK_UNWIND_STRICT (truncate, frame, -1, op_errno, NULL, NULL,
- NULL);
- }
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->truncate,
- loc, offset, xdata);
- return 0;
+ if (op_errno) {
+ if (op_errno < 0)
+ op_errno = EROFS;
+ STACK_UNWIND_STRICT(truncate, frame, -1, op_errno, NULL, NULL, NULL);
+ } else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
+ return 0;
}
-
static int32_t
-worm_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- dict_t *xdata)
+worm_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- int op_errno = EROFS;
- read_only_priv_t *priv = NULL;
-
- priv = this->private;
- GF_ASSERT (priv);
- if (is_readonly_or_worm_enabled (frame, this))
- goto out;
- if (!priv->worm_file || (frame->root->pid < 0)) {
- op_errno = 0;
- goto out;
- }
+ int op_errno = EROFS;
+ read_only_priv_t *priv = NULL;
+
+ priv = this->private;
+ GF_ASSERT(priv);
+ if (is_readonly_or_worm_enabled(frame, this))
+ goto out;
+ if (!priv->worm_file || (frame->root->pid < 0)) {
+ op_errno = 0;
+ goto out;
+ }
- if (is_wormfile (this, _gf_true, fd)) {
- op_errno = 0;
- goto out;
- }
- op_errno = gf_worm_state_transition (this, _gf_true, fd,
- GF_FOP_FTRUNCATE);
+ if (is_wormfile(this, _gf_true, fd)) {
+ op_errno = 0;
+ goto out;
+ }
+ op_errno = gf_worm_state_transition(this, _gf_true, fd, GF_FOP_FTRUNCATE);
out:
- if (op_errno) {
- if (op_errno < 0)
- op_errno = EROFS;
- STACK_UNWIND_STRICT (ftruncate, frame, -1, op_errno, NULL, NULL,
- NULL);
- }
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->ftruncate,
- fd, offset, xdata);
- return 0;
+ if (op_errno) {
+ if (op_errno < 0)
+ op_errno = EROFS;
+ STACK_UNWIND_STRICT(ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
+ } else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+ return 0;
}
-
static int32_t
-worm_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+worm_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- gf_boolean_t rd_only = _gf_false;
- worm_reten_state_t reten_state = {0,};
- struct iatt stpre = {0,};
- read_only_priv_t *priv = NULL;
- int op_errno = EROFS;
- int ret = -1;
-
- priv = this->private;
- GF_ASSERT (priv);
- if (!priv->worm_file) {
- op_errno = 0;
- goto out;
- }
+ gf_boolean_t rd_only = _gf_false;
+ worm_reten_state_t reten_state = {
+ 0,
+ };
+ struct iatt stpre = {
+ 0,
+ };
+ read_only_priv_t *priv = NULL;
+ int op_errno = EROFS;
+ int ret = -1;
+
+ priv = this->private;
+ GF_ASSERT(priv);
+ if (!priv->worm_file) {
+ op_errno = 0;
+ goto out;
+ }
- if (is_wormfile (this, _gf_false, loc)) {
- op_errno = 0;
+ if (is_wormfile(this, _gf_false, loc)) {
+ op_errno = 0;
+ goto out;
+ }
+ if (valid & GF_SET_ATTR_MODE) {
+ rd_only = gf_worm_write_disabled(stbuf);
+ if (!rd_only) {
+ op_errno = 0;
+ goto out;
+ }
+
+ ret = worm_set_state(this, _gf_false, loc, &reten_state, stbuf);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR, "Error setting worm state");
+ goto out;
+ }
+ } else if (valid & GF_SET_ATTR_ATIME) {
+ ret = worm_get_state(this, _gf_false, loc, &reten_state);
+ if (ret) {
+ op_errno = 0;
+ goto out;
+ }
+ if (reten_state.retain) {
+ ret = syncop_stat(this, loc, &stpre, NULL, NULL);
+ if (ret)
goto out;
- }
- if (valid & GF_SET_ATTR_MODE) {
- rd_only = gf_worm_write_disabled (stbuf);
- if (!rd_only) {
- op_errno = 0;
- goto out;
- }
-
- ret = worm_set_state (this, _gf_false, loc,
- &reten_state, stbuf);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "Error setting worm state");
- goto out;
+ if (reten_state.ret_mode == 0) {
+ if (stbuf->ia_atime < stpre.ia_mtime) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Cannot set atime less than "
+ "the mtime for a WORM-Retained "
+ "file");
+ goto out;
}
- } else if (valid & GF_SET_ATTR_ATIME) {
- ret = worm_get_state (this, _gf_false, loc, &reten_state);
- if (ret) {
- op_errno = 0;
- goto out;
- }
- if (reten_state.retain) {
- ret = syncop_stat (this, loc, &stpre, NULL, NULL);
- if (ret)
- goto out;
- if (reten_state.ret_mode == 0) {
- if (stbuf->ia_atime < stpre.ia_mtime) {
- gf_log (this->name, GF_LOG_ERROR,
- "Cannot set atime less than "
- "the mtime for a WORM-Retained "
- "file");
- goto out;
- }
- } else {
- if (stbuf->ia_atime < stpre.ia_atime) {
- gf_log (this->name, GF_LOG_ERROR,
- "Cannot decrease the atime of a"
- " WORM-Retained file in "
- "Enterprise mode");
- goto out;
- }
- }
- stbuf->ia_mtime = stpre.ia_mtime;
+ } else {
+ if (stbuf->ia_atime < stpre.ia_atime) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Cannot decrease the atime of a"
+ " WORM-Retained file in "
+ "Enterprise mode");
+ goto out;
}
+ }
+ reten_state.ret_period = reten_state.ret_period + stbuf->ia_atime -
+ stpre.ia_atime;
+ ret = gf_worm_set_xattr(this, &reten_state, _gf_false, loc);
+ if (ret) {
+ goto out;
+ }
+ stbuf->ia_mtime = stpre.ia_mtime;
}
- op_errno = 0;
+ }
+ op_errno = 0;
out:
- if (op_errno)
- STACK_UNWIND_STRICT (setattr, frame, -1, EROFS, NULL, NULL,
- NULL);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setattr,
- loc, stbuf, valid, xdata);
- return 0;
+ if (op_errno)
+ STACK_UNWIND_STRICT(setattr, frame, -1, EROFS, NULL, NULL, NULL);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid,
+ xdata);
+ return 0;
}
-
static int32_t
-worm_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+worm_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
+ int32_t valid, dict_t *xdata)
{
- gf_boolean_t rd_only = _gf_false;
- worm_reten_state_t reten_state = {0,};
- struct iatt stpre = {0,};
- read_only_priv_t *priv = NULL;
- int op_errno = EROFS;
- int ret = -1;
-
- priv = this->private;
- GF_ASSERT (priv);
- if (!priv->worm_file) {
- op_errno = 0;
- goto out;
- }
+ gf_boolean_t rd_only = _gf_false;
+ worm_reten_state_t reten_state = {
+ 0,
+ };
+ struct iatt stpre = {
+ 0,
+ };
+ read_only_priv_t *priv = NULL;
+ int op_errno = EROFS;
+ int ret = -1;
+
+ priv = this->private;
+ GF_ASSERT(priv);
+ if (!priv->worm_file) {
+ op_errno = 0;
+ goto out;
+ }
- if (is_wormfile (this, _gf_true, fd)) {
- op_errno = 0;
+ if (is_wormfile(this, _gf_true, fd)) {
+ op_errno = 0;
+ goto out;
+ }
+ if (valid & GF_SET_ATTR_MODE) {
+ rd_only = gf_worm_write_disabled(stbuf);
+ if (!rd_only) {
+ op_errno = 0;
+ goto out;
+ }
+
+ ret = worm_set_state(this, _gf_true, fd, &reten_state, stbuf);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR, "Error setting worm state");
+ goto out;
+ }
+ } else if (valid & GF_SET_ATTR_ATIME) {
+ ret = worm_get_state(this, _gf_true, fd, &reten_state);
+ if (ret) {
+ op_errno = 0;
+ goto out;
+ }
+ if (reten_state.retain) {
+ ret = syncop_fstat(this, fd, &stpre, NULL, NULL);
+ if (ret)
goto out;
- }
- if (valid & GF_SET_ATTR_MODE) {
- rd_only = gf_worm_write_disabled (stbuf);
- if (!rd_only) {
- op_errno = 0;
- goto out;
+ if (reten_state.ret_mode == 0) {
+ if (stbuf->ia_atime < stpre.ia_mtime) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Cannot set atime less than "
+ "the mtime for a WORM-Retained "
+ "file");
+ goto out;
}
-
- ret = worm_set_state (this, _gf_true, fd,
- &reten_state, stbuf);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "Error setting worm state");
- goto out;
- }
- } else if (valid & GF_SET_ATTR_ATIME) {
- ret = worm_get_state (this, _gf_true, fd, &reten_state);
- if (ret) {
- op_errno = 0;
- goto out;
- }
- if (reten_state.retain) {
- ret = syncop_fstat (this, fd, &stpre, NULL, NULL);
- if (ret)
- goto out;
- if (reten_state.ret_mode == 0) {
- if (stbuf->ia_atime < stpre.ia_mtime) {
- gf_log (this->name, GF_LOG_ERROR,
- "Cannot set atime less than "
- "the mtime for a WORM-Retained "
- "file");
- goto out;
- }
- } else {
- if (stbuf->ia_atime < stpre.ia_atime) {
- gf_log (this->name, GF_LOG_ERROR,
- "Cannot decrease the atime of a"
- " WORM-Retained file in "
- "Enterprise mode");
- goto out;
- }
- }
- stbuf->ia_mtime = stpre.ia_mtime;
+ } else {
+ if (stbuf->ia_atime < stpre.ia_atime) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Cannot decrease the atime of a"
+ " WORM-Retained file in "
+ "Enterprise mode");
+ goto out;
}
+ }
+ reten_state.ret_period = reten_state.ret_period + stbuf->ia_atime -
+ stpre.ia_atime;
+ ret = gf_worm_set_xattr(this, &reten_state, _gf_true, fd);
+ if (ret) {
+ goto out;
+ }
+
+ stbuf->ia_mtime = stpre.ia_mtime;
}
- op_errno = 0;
+ }
+ op_errno = 0;
out:
- if (op_errno)
- STACK_UNWIND_STRICT (fsetattr, frame, -1, op_errno, NULL, NULL,
- NULL);
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fsetattr,
- fd, stbuf, valid, xdata);
- return 0;
+ if (op_errno)
+ STACK_UNWIND_STRICT(fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetattr, fd, stbuf, valid,
+ xdata);
+ return 0;
}
-
static int32_t
-worm_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count, off_t offset, uint32_t flags,
- struct iobref *iobref, dict_t *xdata)
+worm_writev(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
+ int32_t count, off_t offset, uint32_t flags, struct iobref *iobref,
+ dict_t *xdata)
{
- read_only_priv_t *priv = NULL;
- int op_errno = EROFS;
+ read_only_priv_t *priv = NULL;
+ int op_errno = EROFS;
- priv = this->private;
- GF_ASSERT (priv);
- if (!priv->worm_file || (frame->root->pid < 0)) {
- op_errno = 0;
- goto out;
- }
- if (is_wormfile (this, _gf_true, fd)) {
- op_errno = 0;
- goto out;
- }
- op_errno = gf_worm_state_transition (this, _gf_true, fd, GF_FOP_WRITE);
+ priv = this->private;
+ GF_ASSERT(priv);
+ if (!priv->worm_file || (frame->root->pid < 0)) {
+ op_errno = 0;
+ goto out;
+ }
+ if (is_wormfile(this, _gf_true, fd)) {
+ op_errno = 0;
+ goto out;
+ }
+ op_errno = gf_worm_state_transition(this, _gf_true, fd, GF_FOP_WRITE);
out:
- if (op_errno) {
- if (op_errno < 0)
- op_errno = EROFS;
- STACK_UNWIND_STRICT (writev, frame, -1, op_errno, NULL, NULL,
- NULL);
- }
- else
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->writev,
- fd, vector, count, offset, flags, iobref,
- xdata);
- return 0;
+ if (op_errno) {
+ if (op_errno < 0)
+ op_errno = EROFS;
+ STACK_UNWIND_STRICT(writev, frame, -1, op_errno, NULL, NULL, NULL);
+ } else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector, count,
+ offset, flags, iobref, xdata);
+ return 0;
}
static int32_t
-worm_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd,
- inode_t *inode, struct iatt *buf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+worm_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- int ret = 0;
- read_only_priv_t *priv = NULL;
- dict_t *dict = NULL;
-
- priv = this->private;
- GF_ASSERT (priv);
- if (priv->worm_file) {
- dict = dict_new ();
- if (!dict) {
- gf_log (this->name, GF_LOG_ERROR, "Error creating the "
- "dict");
- goto out;
- }
- ret = dict_set_int8 (dict, "trusted.worm_file", 1);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "Error in setting "
- "the dict");
- goto out;
- }
- ret = syncop_fsetxattr (this, fd, dict, 0, NULL, NULL);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "Error setting xattr");
- goto out;
- }
- ret = worm_init_state (this, _gf_true, fd);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "Error initializing state");
- }
- }
+ int ret = 0;
+ read_only_priv_t *priv = NULL;
+ // In case of an error exit because fd can be NULL and this would
+ // cause an segfault when performing fsetxattr . We explicitly
+ // unwind to avoid future problems
+ if (op_ret < 0) {
+ goto out;
+ }
+
+ priv = this->private;
+ GF_ASSERT(priv);
+ if (priv->worm_file) {
+ ret = fd_ctx_set(fd, this, 1);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Failed to set the fd ctx "
+ "for gfid:%s . Worm feature may not work for the gfid",
+ uuid_utoa(inode->gfid));
+ }
+ ret = worm_init_state(this, _gf_true, fd);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR, "Error initializing state");
+ }
+ }
out:
- STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
- preparent, postparent, xdata);
- if (dict)
- dict_unref (dict);
- return ret;
+ STACK_UNWIND_STRICT(create, frame, op_ret, op_errno, fd, inode, buf,
+ preparent, postparent, xdata);
+ return ret;
}
-
static int32_t
-worm_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+worm_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- STACK_WIND (frame, worm_create_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->create, loc, flags,
- mode, umask, fd, xdata);
- return 0;
+ STACK_WIND(frame, worm_create_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, flags, mode, umask, fd,
+ xdata);
+ return 0;
}
+static void
+set_reten_mode(read_only_priv_t *priv, char *reten_mode)
+{
+ if (strcmp(reten_mode, "relax") == 0)
+ priv->reten_mode = 0;
+ else
+ priv->reten_mode = 1;
+}
int32_t
-init (xlator_t *this)
+init(xlator_t *this)
{
- int ret = -1;
- read_only_priv_t *priv = NULL;
-
- if (!this->children || this->children->next) {
- gf_log (this->name, GF_LOG_ERROR,
- "translator not configured with exactly one child");
- return -1;
- }
-
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "dangling volume. check volfile ");
- }
-
- this->local_pool = mem_pool_new (read_only_priv_t, 64);
- if (!this->local_pool) {
- ret = -1;
- gf_log (this->name, GF_LOG_ERROR,
- "failed to create read_only_priv_t's memory pool");
- goto out;
- }
-
- priv = mem_get0 (this->local_pool);
- if (!priv) {
- gf_log (this->name, GF_LOG_ERROR, "Error allocating priv");
- goto out;
- }
-
- this->private = priv;
-
- GF_OPTION_INIT ("worm", priv->readonly_or_worm_enabled,
- bool, out);
- GF_OPTION_INIT ("worm-file-level", priv->worm_file, bool, out);
- GF_OPTION_INIT ("default-retention-period", priv->reten_period,
- uint64, out);
- GF_OPTION_INIT ("auto-commit-period", priv->com_period, uint64, out);
- GF_OPTION_INIT ("retention-mode", priv->reten_mode, str, out);
- GF_OPTION_INIT ("worm-files-deletable", priv->worm_files_deletable,
- bool, out);
-
- ret = 0;
+ int ret = -1;
+ read_only_priv_t *priv = NULL;
+ char *reten_mode = NULL;
+
+ if (!this->children || this->children->next) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "translator not configured with exactly one child");
+ return -1;
+ }
+
+ if (!this->parents) {
+ gf_log(this->name, GF_LOG_WARNING, "dangling volume. check volfile ");
+ }
+
+ this->local_pool = mem_pool_new(read_only_priv_t, 64);
+ if (!this->local_pool) {
+ ret = -1;
+ gf_log(this->name, GF_LOG_ERROR,
+ "failed to create read_only_priv_t's memory pool");
+ goto out;
+ }
+
+ priv = mem_get0(this->local_pool);
+ if (!priv) {
+ gf_log(this->name, GF_LOG_ERROR, "Error allocating priv");
+ goto out;
+ }
+
+ this->private = priv;
+
+ GF_OPTION_INIT("worm", priv->readonly_or_worm_enabled, bool, out);
+ GF_OPTION_INIT("worm-file-level", priv->worm_file, bool, out);
+ GF_OPTION_INIT("default-retention-period", priv->reten_period, int64, out);
+ GF_OPTION_INIT("auto-commit-period", priv->com_period, int64, out);
+ GF_OPTION_INIT("retention-mode", reten_mode, str, out);
+ set_reten_mode(priv, reten_mode);
+ GF_OPTION_INIT("worm-files-deletable", priv->worm_files_deletable, bool,
+ out);
+
+ ret = 0;
out:
- return ret;
+ return ret;
}
-
int
-reconfigure (xlator_t *this, dict_t *options)
+reconfigure(xlator_t *this, dict_t *options)
{
- read_only_priv_t *priv = NULL;
- int ret = -1;
-
- priv = this->private;
- GF_ASSERT (priv);
-
- GF_OPTION_RECONF ("worm", priv->readonly_or_worm_enabled,
- options, bool, out);
- GF_OPTION_RECONF ("worm-file-level", priv->worm_file, options, bool,
- out);
- GF_OPTION_RECONF ("default-retention-period", priv->reten_period,
- options, uint64, out);
- GF_OPTION_RECONF ("retention-mode", priv->reten_mode, options, str,
- out);
- GF_OPTION_RECONF ("auto-commit-period", priv->com_period, options,
- uint64, out);
- GF_OPTION_RECONF ("worm-files-deletable", priv->worm_files_deletable,
- options, bool, out);
- ret = 0;
+ read_only_priv_t *priv = NULL;
+ char *reten_mode = NULL;
+ int ret = -1;
+
+ priv = this->private;
+ GF_ASSERT(priv);
+
+ GF_OPTION_RECONF("worm", priv->readonly_or_worm_enabled, options, bool,
+ out);
+ GF_OPTION_RECONF("worm-file-level", priv->worm_file, options, bool, out);
+ GF_OPTION_RECONF("default-retention-period", priv->reten_period, options,
+ int64, out);
+ GF_OPTION_RECONF("retention-mode", reten_mode, options, str, out);
+ set_reten_mode(priv, reten_mode);
+ GF_OPTION_RECONF("auto-commit-period", priv->com_period, options, int64,
+ out);
+ GF_OPTION_RECONF("worm-files-deletable", priv->worm_files_deletable,
+ options, bool, out);
+ ret = 0;
out:
- gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret);
- return ret;
+ gf_log(this->name, GF_LOG_DEBUG, "returning %d", ret);
+ return ret;
}
-
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- read_only_priv_t *priv = NULL;
-
- priv = this->private;
- if (!priv)
- goto out;
- mem_put (priv);
- this->private = NULL;
- mem_pool_destroy (this->local_pool);
+ read_only_priv_t *priv = NULL;
+
+ priv = this->private;
+ if (!priv)
+ goto out;
+ mem_put(priv);
+ this->private = NULL;
+ mem_pool_destroy(this->local_pool);
+ this->local_pool = NULL;
out:
- return;
+ return;
}
-
struct xlator_fops fops = {
- .open = worm_open,
- .writev = worm_writev,
- .setattr = worm_setattr,
- .fsetattr = worm_fsetattr,
- .rename = worm_rename,
- .link = worm_link,
- .unlink = worm_unlink,
- .truncate = worm_truncate,
- .ftruncate = worm_ftruncate,
- .create = worm_create,
-
- .rmdir = ro_rmdir,
- .removexattr = ro_removexattr,
- .fsyncdir = ro_fsyncdir,
- .xattrop = ro_xattrop,
- .inodelk = ro_inodelk,
- .finodelk = ro_finodelk,
- .entrylk = ro_entrylk,
- .fentrylk = ro_fentrylk,
- .lk = ro_lk,
+ .open = worm_open,
+ .writev = worm_writev,
+ .setattr = worm_setattr,
+ .fsetattr = worm_fsetattr,
+ .rename = worm_rename,
+ .link = worm_link,
+ .unlink = worm_unlink,
+ .truncate = worm_truncate,
+ .ftruncate = worm_ftruncate,
+ .create = worm_create,
+
+ .rmdir = ro_rmdir,
+ .removexattr = ro_removexattr,
+ .fsyncdir = ro_fsyncdir,
+ .xattrop = ro_xattrop,
+ .inodelk = ro_inodelk,
+ .finodelk = ro_finodelk,
+ .entrylk = ro_entrylk,
+ .fentrylk = ro_fentrylk,
+ .lk = ro_lk,
};
+int32_t
+worm_release(xlator_t *this, fd_t *fd)
+{
+ dict_t *dict = NULL;
+ int ret = -1;
+ dict = dict_new();
+ uint64_t value = 0;
+ loc_t loc = {
+ 0,
+ };
+ read_only_priv_t *priv = NULL;
+ priv = this->private;
+
+ if (priv->worm_file) {
+ if (!dict) {
+ gf_log(this->name, GF_LOG_ERROR, "Error creating the dict");
+ goto out;
+ }
-struct xlator_cbks cbks;
+ ret = fd_ctx_get(fd, this, &value);
+ if (ret) {
+ gf_log(this->name, GF_LOG_DEBUG, "Failed to get the fd ctx");
+ }
+ if (!value) {
+ goto out;
+ }
+ ret = dict_set_int8(dict, "trusted.worm_file", 1);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "Error in setting "
+ "the dict");
+ goto out;
+ }
+
+ loc.inode = inode_ref(fd->inode);
+ gf_uuid_copy(loc.gfid, fd->inode->gfid);
+ ret = syncop_setxattr(this, &loc, dict, 0, NULL, NULL);
+ if (ret) {
+ gf_log(this->name, GF_LOG_ERROR, "Error setting xattr");
+ goto out;
+ }
+
+ gf_worm_state_transition(this, _gf_false, &loc, GF_FOP_WRITE);
+ }
+
+out:
+ loc_wipe(&loc);
+ if (dict)
+ dict_unref(dict);
+ return 0;
+}
+
+struct xlator_cbks cbks = {
+ .release = worm_release,
+};
struct volume_options options[] = {
- { .key = {"worm"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- /*.validate_fn = validate_boolean,*/
- .op_version = {2},
- .flags = OPT_FLAG_SETTABLE,
- .description = "When \"on\", makes a volume get write once read many "
- " feature. It is turned \"off\" by default."
- },
- { .key = {"worm-file-level"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "off",
- /*.validate_fn = validate_boolean,*/
- .op_version = {GD_OP_VERSION_3_8_0},
- .flags = OPT_FLAG_SETTABLE,
- .description = "When \"on\", activates the file level worm. "
- "It is turned \"off\" by default."
- },
- { .key = {"worm-files-deletable"},
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- /*.validate_fn = validate_boolean,*/
- .op_version = {GD_OP_VERSION_3_13_0},
- .flags = OPT_FLAG_SETTABLE,
- .description = "When \"off\", doesn't allow the Worm files"
- "to be deleted. It is turned \"on\" by default."
- },
- { .key = {"default-retention-period"},
- .type = GF_OPTION_TYPE_TIME,
- .default_value = "120",
- /*.validate_fn = validate_worm_period,*/
- .op_version = {GD_OP_VERSION_3_8_0},
- .flags = OPT_FLAG_SETTABLE,
- .description = "The default retention period for the files."
- },
- { .key = {"retention-mode"},
- .type = GF_OPTION_TYPE_STR,
- .default_value = "relax",
- /*.validate_fn = validate_reten_mode,*/
- .op_version = {GD_OP_VERSION_3_8_0},
- .flags = OPT_FLAG_SETTABLE,
- .description = "The mode of retention (relax/enterprise). "
- "It is relax by default."
- },
- { .key = {"auto-commit-period"},
- .type = GF_OPTION_TYPE_TIME,
- .default_value = "180",
- /*.validate_fn = validate_worm_period,*/
- .op_version = {GD_OP_VERSION_3_8_0},
- .flags = OPT_FLAG_SETTABLE,
- .description = "Auto commit period for the files."
- },
- { .key = {NULL}
- },
+ {.key = {"worm"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ /*.validate_fn = validate_boolean,*/
+ .op_version = {2},
+ .flags = OPT_FLAG_SETTABLE,
+ .description = "When \"on\", makes a volume get write once read many "
+ " feature. It is turned \"off\" by default."},
+ {.key = {"worm-file-level"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ /*.validate_fn = validate_boolean,*/
+ .op_version = {GD_OP_VERSION_3_8_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .description = "When \"on\", activates the file level worm. "
+ "It is turned \"off\" by default."},
+ {.key = {"worm-files-deletable"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ /*.validate_fn = validate_boolean,*/
+ .op_version = {GD_OP_VERSION_3_13_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .description = "When \"off\", doesn't allow the Worm files"
+ "to be deleted. It is turned \"on\" by default."},
+ {.key = {"default-retention-period"},
+ .type = GF_OPTION_TYPE_TIME,
+ .default_value = "120",
+ /*.validate_fn = validate_worm_period,*/
+ .op_version = {GD_OP_VERSION_3_8_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .description = "The default retention period for the files."},
+ {.key = {"retention-mode"},
+ .type = GF_OPTION_TYPE_STR,
+ .default_value = "relax",
+ /*.validate_fn = validate_reten_mode,*/
+ .op_version = {GD_OP_VERSION_3_8_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .description = "The mode of retention (relax/enterprise). "
+ "It is relax by default."},
+ {.key = {"auto-commit-period"},
+ .type = GF_OPTION_TYPE_TIME,
+ .default_value = "180",
+ /*.validate_fn = validate_worm_period,*/
+ .op_version = {GD_OP_VERSION_3_8_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .description = "Auto commit period for the files."},
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "worm",
+ .category = GF_TECH_PREVIEW,
};
diff --git a/xlators/features/sdfs/src/sdfs-messages.h b/xlators/features/sdfs/src/sdfs-messages.h
index 6c7a9d90667..3053efa8935 100644
--- a/xlators/features/sdfs/src/sdfs-messages.h
+++ b/xlators/features/sdfs/src/sdfs-messages.h
@@ -11,7 +11,7 @@
#ifndef _DFS_MESSAGES_H_
#define _DFS_MESSAGES_H_
-#include "glfs-message-id.h"
+#include <glusterfs/glfs-message-id.h>
/* file bit-rot-bitd-messages.h
* brief SDFS log-message IDs and their descriptions
@@ -39,16 +39,14 @@
* holes.
*/
-#define GLFS_SDFS_BASE GLFS_MSGID_COMP_SDFS
-#define GLFS_SDFS_NUM_MESSAGES 2
-#define GLFS_MSGID_END (GLFS_SDFS_BASE + \
- GLFS_SDFS_NUM_MESSAGES + 1)
+#define GLFS_SDFS_BASE GLFS_MSGID_COMP_SDFS
+#define GLFS_SDFS_NUM_MESSAGES 2
+#define GLFS_MSGID_END (GLFS_SDFS_BASE + GLFS_SDFS_NUM_MESSAGES + 1)
/* Messaged with message IDs */
-#define glfs_msg_start_x GLFS_DFS_BASE, "Invalid: Start of messages"
+#define glfs_msg_start_x GLFS_DFS_BASE, "Invalid: Start of messages"
/*------------*/
-
-#define SDFS_MSG_ENTRYLK_ERROR (GLFS_SDFS_BASE + 1)
+#define SDFS_MSG_ENTRYLK_ERROR (GLFS_SDFS_BASE + 1)
/*!
* @messageid
* @diagnosis
@@ -56,7 +54,7 @@
*
*/
-#define SDFS_MSG_MKDIR_ERROR (GLFS_SDFS_BASE + 2)
+#define SDFS_MSG_MKDIR_ERROR (GLFS_SDFS_BASE + 2)
/*!
* @messageid
* @diagnosis
diff --git a/xlators/features/sdfs/src/sdfs.c b/xlators/features/sdfs/src/sdfs.c
index 0c45cad6238..aaf13f0852e 100644
--- a/xlators/features/sdfs/src/sdfs.c
+++ b/xlators/features/sdfs/src/sdfs.c
@@ -11,1347 +11,1469 @@
#include "sdfs.h"
static int
-sdfs_frame_return (call_frame_t *frame)
+sdfs_frame_return(call_frame_t *frame)
{
- sdfs_local_t *local = NULL;
+ sdfs_local_t *local = NULL;
- if (!frame)
- return -1;
+ if (!frame)
+ return -1;
- local = frame->local;
+ local = frame->local;
- return GF_ATOMIC_DEC (local->call_cnt);
+ return GF_ATOMIC_DEC(local->call_cnt);
}
static void
-sdfs_lock_free (sdfs_entry_lock_t *entrylk)
+sdfs_lock_free(sdfs_entry_lock_t *entrylk)
{
- if (entrylk == NULL)
- goto out;
+ if (entrylk == NULL)
+ goto out;
- loc_wipe (&entrylk->parent_loc);
- GF_FREE (entrylk->basename);
+ loc_wipe(&entrylk->parent_loc);
+ GF_FREE(entrylk->basename);
out:
- return;
+ return;
}
static void
-sdfs_lock_array_free (sdfs_lock_t *lock)
+sdfs_lock_array_free(sdfs_lock_t *lock)
{
- sdfs_entry_lock_t *entrylk = NULL;
- int i = 0;
+ sdfs_entry_lock_t *entrylk = NULL;
+ int i = 0;
- if (lock == NULL)
- goto out;
+ if (lock == NULL)
+ goto out;
- for (i = 0; i < lock->lock_count; i++) {
- entrylk = &lock->entrylk[i];
- sdfs_lock_free (entrylk);
- }
+ for (i = 0; i < lock->lock_count; i++) {
+ entrylk = &lock->entrylk[i];
+ sdfs_lock_free(entrylk);
+ }
out:
- return;
+ return;
}
static void
-sdfs_local_cleanup (sdfs_local_t *local)
+sdfs_local_cleanup(sdfs_local_t *local)
{
- if (!local)
- return;
+ if (!local)
+ return;
- loc_wipe (&local->loc);
- loc_wipe (&local->parent_loc);
+ loc_wipe(&local->loc);
+ loc_wipe(&local->parent_loc);
- if (local->stub) {
- call_stub_destroy (local->stub);
- local->stub = NULL;
- }
+ if (local->stub) {
+ call_stub_destroy(local->stub);
+ local->stub = NULL;
+ }
- sdfs_lock_array_free (local->lock);
- GF_FREE (local->lock);
+ sdfs_lock_array_free(local->lock);
+ GF_FREE(local->lock);
- mem_put (local);
+ mem_put(local);
}
static int
-sdfs_build_parent_loc (loc_t *parent, loc_t *child)
+sdfs_build_parent_loc(loc_t *parent, loc_t *child)
{
- int ret = -1;
- char *path = NULL;
+ int ret = -1;
+ char *path = NULL;
+
+ if (!child->parent) {
+ goto out;
+ }
+ parent->inode = inode_ref(child->parent);
+ path = gf_strdup(child->path);
+ if (!path) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ parent->path = dirname(path);
+ if (!parent->path) {
+ goto out;
+ }
+
+ gf_uuid_copy(parent->gfid, child->pargfid);
+ return 0;
- if (!child->parent) {
- goto out;
- }
- parent->inode = inode_ref (child->parent);
- path = gf_strdup (child->path);
- if (!path) {
- ret = -ENOMEM;
- goto out;
- }
+out:
+ GF_FREE(path);
+ return ret;
+}
- parent->path = dirname(path);
- if (!parent->path) {
- goto out;
- }
+static sdfs_local_t *
+sdfs_local_init(call_frame_t *frame, xlator_t *this)
+{
+ sdfs_local_t *local = NULL;
- gf_uuid_copy (parent->gfid, child->pargfid);
- return 0;
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto out;
+ frame->local = local;
out:
- GF_FREE (path);
- return ret;
+ return local;
}
-static sdfs_local_t *
-sdfs_local_init (call_frame_t *frame, xlator_t *this)
+static int
+sdfs_get_new_frame_common(call_frame_t *frame, call_frame_t **new_frame)
{
- sdfs_local_t *local = NULL;
+ int ret = -1;
+ sdfs_local_t *local = NULL;
+ client_t *client = NULL;
- local = mem_get0 (this->local_pool);
- if (!local)
- goto out;
+ *new_frame = copy_frame(frame);
+ if (!*new_frame) {
+ goto err;
+ }
- frame->local = local;
-out:
- return local;
+ client = frame->root->client;
+ gf_client_ref(client);
+ (*new_frame)->root->client = client;
+
+ local = sdfs_local_init(*new_frame, THIS);
+ if (!local) {
+ goto err;
+ }
+
+ local->main_frame = frame;
+ /*Set unique lk-owner for the fop*/
+ set_lk_owner_from_ptr(&(*new_frame)->root->lk_owner, (*new_frame)->root);
+
+ ret = 0;
+err:
+ if ((ret == -1) && (*new_frame)) {
+ SDFS_STACK_DESTROY((*new_frame));
+ *new_frame = NULL;
+ }
+
+ return ret;
}
static int
-sdfs_get_new_frame (call_frame_t *frame, loc_t *loc, call_frame_t **new_frame)
+sdfs_get_new_frame(call_frame_t *frame, loc_t *loc, call_frame_t **new_frame)
{
- int ret = -1;
- sdfs_local_t *local = NULL;
- client_t *client = NULL;
+ int ret = -1;
+ sdfs_local_t *local = NULL;
- *new_frame = copy_frame (frame);
- if (!*new_frame) {
- goto err;
- }
+ ret = sdfs_get_new_frame_common(frame, new_frame);
+ if (ret < 0) {
+ goto err;
+ }
- client = frame->root->client;
- gf_client_ref (client);
- (*new_frame)->root->client = client;
- local = sdfs_local_init (*new_frame, THIS);
- if (!local) {
- goto err;
- }
+ local = (*new_frame)->local;
- local->main_frame = frame;
+ ret = sdfs_build_parent_loc(&local->parent_loc, loc);
+ if (ret) {
+ goto err;
+ }
- ret = sdfs_build_parent_loc (&local->parent_loc, loc);
- if (ret) {
- goto err;
- }
+ ret = loc_copy(&local->loc, loc);
+ if (ret == -1) {
+ goto err;
+ }
- ret = loc_copy (&local->loc, loc);
- if (ret == -1) {
- goto err;
- }
+ ret = 0;
+err:
+ if (ret && (*new_frame)) {
+ SDFS_STACK_DESTROY((*new_frame));
+ *new_frame = NULL;
+ ret = -1;
+ }
+
+ return ret;
+}
- ret = 0;
+static int
+sdfs_get_new_frame_readdirp(call_frame_t *frame, fd_t *fd,
+ call_frame_t **new_frame)
+{
+ int ret = -1;
+ sdfs_local_t *local = NULL;
+
+ ret = sdfs_get_new_frame_common(frame, new_frame);
+ if (ret < 0) {
+ goto err;
+ }
+
+ local = (*new_frame)->local;
+ local->parent_loc.inode = inode_ref(fd->inode);
+ gf_uuid_copy(local->parent_loc.gfid, fd->inode->gfid);
+
+ ret = 0;
err:
- if (ret == -1) {
- SDFS_STACK_DESTROY (frame);
- }
- return ret;
+ return ret;
}
int
-sdfs_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+sdfs_entrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- call_stub_t *stub = NULL;
+ sdfs_local_t *local = NULL;
+ call_stub_t *stub = NULL;
- local = frame->local;
+ local = frame->local;
- local->op_ret = op_ret;
- local->op_errno = op_errno;
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
- if (local->stub) {
- stub = local->stub;
- local->stub = NULL;
- call_resume (stub);
- } else {
- if (op_ret < 0)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SDFS_MSG_ENTRYLK_ERROR,
- "Unlocking entry lock failed for %s",
- local->loc.name);
-
- SDFS_STACK_DESTROY (frame);
- }
+ if (local->stub) {
+ stub = local->stub;
+ local->stub = NULL;
+ call_resume(stub);
+ } else {
+ if (op_ret < 0)
+ gf_msg(this->name, GF_LOG_ERROR, 0, SDFS_MSG_ENTRYLK_ERROR,
+ "Unlocking entry lock failed for %s", local->loc.name);
- return 0;
+ SDFS_STACK_DESTROY(frame);
+ }
+
+ return 0;
}
int
-sdfs_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+sdfs_mkdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
+ sdfs_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- STACK_UNWIND_STRICT (mkdir, local->main_frame, op_ret, op_errno, inode,
- stbuf, preparent, postparent, xdata);
+ STACK_UNWIND_STRICT(mkdir, local->main_frame, op_ret, op_errno, inode,
+ stbuf, preparent, postparent, xdata);
- local->main_frame = NULL;
- STACK_WIND (frame, sdfs_entrylk_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &local->parent_loc, local->loc.name,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
- return 0;
+ local->main_frame = NULL;
+ STACK_WIND(frame, sdfs_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &local->parent_loc,
+ local->loc.name, ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
+ return 0;
}
int
-sdfs_mkdir_helper (call_frame_t *frame, xlator_t *this, loc_t *loc,
- mode_t mode, mode_t umask, dict_t *xdata)
+sdfs_mkdir_helper(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
- int op_errno = -1;
+ sdfs_local_t *local = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
+ int op_errno = -1;
- local = frame->local;
+ local = frame->local;
- gf_uuid_unparse(loc->pargfid, gfid);
+ gf_uuid_unparse(loc->pargfid, gfid);
- if (local->op_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SDFS_MSG_ENTRYLK_ERROR,
- "Acquiring entry lock failed for directory %s "
- "with parent gfid %s", local->loc.name, gfid);
- op_errno = local->op_errno;
- goto err;
- }
+ if (local->op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SDFS_MSG_ENTRYLK_ERROR,
+ "Acquiring entry lock failed for directory %s "
+ "with parent gfid %s",
+ local->loc.name, gfid);
+ op_errno = local->op_errno;
+ goto err;
+ }
- STACK_WIND (frame, sdfs_mkdir_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mkdir, loc,
- mode, umask, xdata);
+ STACK_WIND(frame, sdfs_mkdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
- return 0;
+ return 0;
err:
- STACK_UNWIND_STRICT (mkdir, local->main_frame, -1, op_errno,
- NULL, NULL, NULL, NULL, NULL);
+ STACK_UNWIND_STRICT(mkdir, local->main_frame, -1, op_errno, NULL, NULL,
+ NULL, NULL, NULL);
- local->main_frame = NULL;
- SDFS_STACK_DESTROY (frame);
- return 0;
+ local->main_frame = NULL;
+ SDFS_STACK_DESTROY(frame);
+ return 0;
}
int
-sdfs_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- mode_t umask, dict_t *xdata)
+sdfs_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- call_frame_t *new_frame = NULL;
- call_stub_t *stub = NULL;
- int op_errno = 0;
-
- if (-1 == sdfs_get_new_frame (frame, loc, &new_frame)) {
- op_errno = ENOMEM;
- goto err;
- }
-
- stub = fop_mkdir_stub (new_frame, sdfs_mkdir_helper, loc, mode,
- umask, xdata);
- if (!stub) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local = new_frame->local;
- local->stub = stub;
-
- STACK_WIND (new_frame, sdfs_entrylk_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &local->parent_loc, local->loc.name,
- ENTRYLK_LOCK, ENTRYLK_WRLCK, xdata);
-
- return 0;
+ sdfs_local_t *local = NULL;
+ call_frame_t *new_frame = NULL;
+ call_stub_t *stub = NULL;
+ int op_errno = 0;
+
+ if (-1 == sdfs_get_new_frame(frame, loc, &new_frame)) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ stub = fop_mkdir_stub(new_frame, sdfs_mkdir_helper, loc, mode, umask,
+ xdata);
+ if (!stub) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local = new_frame->local;
+ local->stub = stub;
+
+ STACK_WIND(new_frame, sdfs_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &local->parent_loc,
+ local->loc.name, ENTRYLK_LOCK, ENTRYLK_WRLCK, xdata);
+
+ return 0;
err:
- STACK_UNWIND_STRICT (mkdir, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL);
+ STACK_UNWIND_STRICT(mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
- if (new_frame)
- SDFS_STACK_DESTROY (new_frame);
+ if (new_frame)
+ SDFS_STACK_DESTROY(new_frame);
- return 0;
+ return 0;
}
int
-sdfs_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+sdfs_rmdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
+ sdfs_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- STACK_UNWIND_STRICT (rmdir, local->main_frame, op_ret, op_errno,
- preparent, postparent, xdata);
+ STACK_UNWIND_STRICT(rmdir, local->main_frame, op_ret, op_errno, preparent,
+ postparent, xdata);
- local->main_frame = NULL;
- STACK_WIND (frame, sdfs_entrylk_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &local->parent_loc, local->loc.name,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
- return 0;
+ local->main_frame = NULL;
+ STACK_WIND(frame, sdfs_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &local->parent_loc,
+ local->loc.name, ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
+ return 0;
}
int
-sdfs_rmdir_helper (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int flags, dict_t *xdata)
+sdfs_rmdir_helper(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
+ sdfs_local_t *local = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- local = frame->local;
+ local = frame->local;
- gf_uuid_unparse(loc->pargfid, gfid);
+ gf_uuid_unparse(loc->pargfid, gfid);
- if (local->op_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SDFS_MSG_ENTRYLK_ERROR,
- "Acquiring entry lock failed for directory %s "
- "with parent gfid %s", local->loc.name, gfid);
- goto err;
- }
+ if (local->op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SDFS_MSG_ENTRYLK_ERROR,
+ "Acquiring entry lock failed for directory %s "
+ "with parent gfid %s",
+ local->loc.name, gfid);
+ goto err;
+ }
- STACK_WIND (frame, sdfs_rmdir_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rmdir, loc,
- flags, xdata);
+ STACK_WIND(frame, sdfs_rmdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rmdir, loc, flags, xdata);
- return 0;
+ return 0;
err:
- STACK_UNWIND_STRICT (rmdir, local->main_frame, -1, local->op_errno,
- NULL, NULL, NULL);
+ STACK_UNWIND_STRICT(rmdir, local->main_frame, -1, local->op_errno, NULL,
+ NULL, NULL);
- local->main_frame = NULL;
- SDFS_STACK_DESTROY (frame);
- return 0;
+ local->main_frame = NULL;
+ SDFS_STACK_DESTROY(frame);
+ return 0;
}
int
-sdfs_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
- dict_t *xdata)
+sdfs_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- call_frame_t *new_frame = NULL;
- call_stub_t *stub = NULL;
- int op_errno = 0;
-
- if (-1 == sdfs_get_new_frame (frame, loc, &new_frame)) {
- op_errno = ENOMEM;
- goto err;
- }
-
- stub = fop_rmdir_stub (new_frame, sdfs_rmdir_helper, loc, flags, xdata);
- if (!stub) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local = new_frame->local;
- local->stub = stub;
-
- STACK_WIND (new_frame, sdfs_entrylk_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &local->parent_loc, local->loc.name,
- ENTRYLK_LOCK, ENTRYLK_WRLCK, xdata);
-
- return 0;
+ sdfs_local_t *local = NULL;
+ call_frame_t *new_frame = NULL;
+ call_stub_t *stub = NULL;
+ int op_errno = 0;
+
+ if (-1 == sdfs_get_new_frame(frame, loc, &new_frame)) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ stub = fop_rmdir_stub(new_frame, sdfs_rmdir_helper, loc, flags, xdata);
+ if (!stub) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local = new_frame->local;
+ local->stub = stub;
+
+ STACK_WIND(new_frame, sdfs_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &local->parent_loc,
+ local->loc.name, ENTRYLK_LOCK, ENTRYLK_WRLCK, xdata);
+
+ return 0;
err:
- STACK_UNWIND_STRICT (rmdir, frame, -1, op_errno, NULL, NULL,
- NULL);
+ STACK_UNWIND_STRICT(rmdir, frame, -1, op_errno, NULL, NULL, NULL);
- if (new_frame)
- SDFS_STACK_DESTROY (new_frame);
+ if (new_frame)
+ SDFS_STACK_DESTROY(new_frame);
- return 0;
+ return 0;
}
int
-sdfs_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- fd_t *fd, inode_t *inode, struct iatt *stbuf,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+sdfs_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
+ sdfs_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- STACK_UNWIND_STRICT (create, local->main_frame, op_ret, op_errno, fd,
- inode, stbuf, preparent, postparent, xdata);
+ STACK_UNWIND_STRICT(create, local->main_frame, op_ret, op_errno, fd, inode,
+ stbuf, preparent, postparent, xdata);
- local->main_frame = NULL;
- STACK_WIND (frame, sdfs_entrylk_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &local->parent_loc, local->loc.name,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
- return 0;
+ local->main_frame = NULL;
+ STACK_WIND(frame, sdfs_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &local->parent_loc,
+ local->loc.name, ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
+ return 0;
}
int
-sdfs_create_helper (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int32_t flags, mode_t mode, mode_t umask, fd_t *fd,
- dict_t *xdata)
+sdfs_create_helper(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ int32_t flags, mode_t mode, mode_t umask, fd_t *fd,
+ dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
+ sdfs_local_t *local = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- local = frame->local;
+ local = frame->local;
- gf_uuid_unparse(loc->pargfid, gfid);
+ gf_uuid_unparse(loc->pargfid, gfid);
- if (local->op_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SDFS_MSG_ENTRYLK_ERROR,
- "Acquiring entry lock failed for directory %s "
- "with parent gfid %s", local->loc.name, gfid);
- goto err;
- }
+ if (local->op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SDFS_MSG_ENTRYLK_ERROR,
+ "Acquiring entry lock failed for directory %s "
+ "with parent gfid %s",
+ local->loc.name, gfid);
+ goto err;
+ }
- STACK_WIND (frame, sdfs_create_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->create, loc, flags,
- mode, umask, fd, xdata);
+ STACK_WIND(frame, sdfs_create_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, flags, mode, umask, fd,
+ xdata);
- return 0;
+ return 0;
err:
- STACK_UNWIND_STRICT (create, local->main_frame, -1, local->op_errno,
- NULL, NULL, NULL, NULL, NULL, NULL);
+ STACK_UNWIND_STRICT(create, local->main_frame, -1, local->op_errno, NULL,
+ NULL, NULL, NULL, NULL, NULL);
- local->main_frame = NULL;
- SDFS_STACK_DESTROY (frame);
- return 0;
+ local->main_frame = NULL;
+ SDFS_STACK_DESTROY(frame);
+ return 0;
}
int
-sdfs_create (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int32_t flags, mode_t mode, mode_t umask,
- fd_t *fd, dict_t *xdata)
+sdfs_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- call_frame_t *new_frame = NULL;
- call_stub_t *stub = NULL;
- int op_errno = 0;
-
- if (-1 == sdfs_get_new_frame (frame, loc, &new_frame)) {
- op_errno = ENOMEM;
- goto err;
- }
-
- stub = fop_create_stub (new_frame, sdfs_create_helper, loc,
- flags, mode, umask, fd, xdata);
- if (!stub) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local = new_frame->local;
- local->stub = stub;
-
- STACK_WIND (new_frame, sdfs_entrylk_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &local->parent_loc, local->loc.name,
- ENTRYLK_LOCK, ENTRYLK_WRLCK, xdata);
-
- return 0;
+ sdfs_local_t *local = NULL;
+ call_frame_t *new_frame = NULL;
+ call_stub_t *stub = NULL;
+ int op_errno = 0;
+
+ if (-1 == sdfs_get_new_frame(frame, loc, &new_frame)) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ stub = fop_create_stub(new_frame, sdfs_create_helper, loc, flags, mode,
+ umask, fd, xdata);
+ if (!stub) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local = new_frame->local;
+ local->stub = stub;
+
+ STACK_WIND(new_frame, sdfs_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &local->parent_loc,
+ local->loc.name, ENTRYLK_LOCK, ENTRYLK_WRLCK, xdata);
+
+ return 0;
err:
- STACK_UNWIND_STRICT (create, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL, NULL);
+ STACK_UNWIND_STRICT(create, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL, NULL);
- if (new_frame)
- SDFS_STACK_DESTROY (new_frame);
+ if (new_frame)
+ SDFS_STACK_DESTROY(new_frame);
- return 0;
+ return 0;
}
int
-sdfs_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+sdfs_unlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
+ sdfs_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- STACK_UNWIND_STRICT (unlink, local->main_frame, op_ret, op_errno,
- preparent, postparent, xdata);
+ STACK_UNWIND_STRICT(unlink, local->main_frame, op_ret, op_errno, preparent,
+ postparent, xdata);
- local->main_frame = NULL;
- STACK_WIND (frame, sdfs_entrylk_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &local->parent_loc, local->loc.name,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
- return 0;
+ local->main_frame = NULL;
+ STACK_WIND(frame, sdfs_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &local->parent_loc,
+ local->loc.name, ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
+ return 0;
}
int
-sdfs_unlink_helper (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int flags, dict_t *xdata)
+sdfs_unlink_helper(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
+ sdfs_local_t *local = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- local = frame->local;
+ local = frame->local;
- gf_uuid_unparse(loc->pargfid, gfid);
+ gf_uuid_unparse(loc->pargfid, gfid);
- if (local->op_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SDFS_MSG_ENTRYLK_ERROR,
- "Acquiring entry lock failed for directory %s "
- "with parent gfid %s", local->loc.name, gfid);
- goto err;
- }
+ if (local->op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SDFS_MSG_ENTRYLK_ERROR,
+ "Acquiring entry lock failed for directory %s "
+ "with parent gfid %s",
+ local->loc.name, gfid);
+ goto err;
+ }
- STACK_WIND (frame, sdfs_unlink_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->unlink, loc, flags, xdata);
+ STACK_WIND(frame, sdfs_unlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, flags, xdata);
- return 0;
+ return 0;
err:
- STACK_UNWIND_STRICT (unlink, local->main_frame, -1, local->op_errno,
- NULL, NULL, NULL);
+ STACK_UNWIND_STRICT(unlink, local->main_frame, -1, local->op_errno, NULL,
+ NULL, NULL);
- local->main_frame = NULL;
- SDFS_STACK_DESTROY (frame);
- return 0;
+ local->main_frame = NULL;
+ SDFS_STACK_DESTROY(frame);
+ return 0;
}
int
-sdfs_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc,
- int flags, dict_t *xdata)
+sdfs_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- call_frame_t *new_frame = NULL;
- call_stub_t *stub = NULL;
- int op_errno = 0;
-
- if (-1 == sdfs_get_new_frame (frame, loc, &new_frame)) {
- op_errno = ENOMEM;
- goto err;
- }
-
- stub = fop_unlink_stub (new_frame, sdfs_unlink_helper, loc,
- flags, xdata);
- if (!stub) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local = new_frame->local;
- local->stub = stub;
-
- STACK_WIND (new_frame, sdfs_entrylk_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &local->parent_loc, local->loc.name,
- ENTRYLK_LOCK, ENTRYLK_WRLCK, xdata);
-
- return 0;
+ sdfs_local_t *local = NULL;
+ call_frame_t *new_frame = NULL;
+ call_stub_t *stub = NULL;
+ int op_errno = 0;
+
+ if (-1 == sdfs_get_new_frame(frame, loc, &new_frame)) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ stub = fop_unlink_stub(new_frame, sdfs_unlink_helper, loc, flags, xdata);
+ if (!stub) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local = new_frame->local;
+ local->stub = stub;
+
+ STACK_WIND(new_frame, sdfs_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &local->parent_loc,
+ local->loc.name, ENTRYLK_LOCK, ENTRYLK_WRLCK, xdata);
+
+ return 0;
err:
- STACK_UNWIND_STRICT (unlink, frame, -1, op_errno, NULL, NULL,
- NULL);
+ STACK_UNWIND_STRICT(unlink, frame, -1, op_errno, NULL, NULL, NULL);
- if (new_frame)
- SDFS_STACK_DESTROY (new_frame);
+ if (new_frame)
+ SDFS_STACK_DESTROY(new_frame);
- return 0;
+ return 0;
}
int
-sdfs_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+sdfs_symlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
+ sdfs_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- STACK_UNWIND_STRICT (link, local->main_frame, op_ret, op_errno, inode,
- stbuf, preparent, postparent, xdata);
+ STACK_UNWIND_STRICT(link, local->main_frame, op_ret, op_errno, inode, stbuf,
+ preparent, postparent, xdata);
- local->main_frame = NULL;
- STACK_WIND (frame, sdfs_entrylk_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &local->parent_loc, local->loc.name,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
- return 0;
+ local->main_frame = NULL;
+ STACK_WIND(frame, sdfs_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &local->parent_loc,
+ local->loc.name, ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
+ return 0;
}
int
-sdfs_symlink_helper (call_frame_t *frame, xlator_t *this,
- const char *linkname, loc_t *loc, mode_t umask,
- dict_t *xdata)
+sdfs_symlink_helper(call_frame_t *frame, xlator_t *this, const char *linkname,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
+ sdfs_local_t *local = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- local = frame->local;
+ local = frame->local;
- gf_uuid_unparse(loc->pargfid, gfid);
+ gf_uuid_unparse(loc->pargfid, gfid);
- if (local->op_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SDFS_MSG_ENTRYLK_ERROR,
- "Acquiring entry lock failed for directory %s "
- "with parent gfid %s", local->loc.name, gfid);
- goto err;
- }
+ if (local->op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SDFS_MSG_ENTRYLK_ERROR,
+ "Acquiring entry lock failed for directory %s "
+ "with parent gfid %s",
+ local->loc.name, gfid);
+ goto err;
+ }
- STACK_WIND (frame, sdfs_symlink_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->symlink, linkname, loc,
- umask, xdata);
+ STACK_WIND(frame, sdfs_symlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->symlink, linkname, loc, umask, xdata);
- return 0;
+ return 0;
err:
- STACK_UNWIND_STRICT (link, local->main_frame, -1, local->op_errno,
- NULL, NULL, NULL, NULL, NULL);
+ STACK_UNWIND_STRICT(link, local->main_frame, -1, local->op_errno, NULL,
+ NULL, NULL, NULL, NULL);
- local->main_frame = NULL;
- SDFS_STACK_DESTROY (frame);
- return 0;
+ local->main_frame = NULL;
+ SDFS_STACK_DESTROY(frame);
+ return 0;
}
int
-sdfs_symlink (call_frame_t *frame, xlator_t *this, const char *linkname,
- loc_t *loc, mode_t umask, dict_t *xdata)
+sdfs_symlink(call_frame_t *frame, xlator_t *this, const char *linkname,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- call_frame_t *new_frame = NULL;
- call_stub_t *stub = NULL;
- int op_errno = 0;
-
- if (-1 == sdfs_get_new_frame (frame, loc, &new_frame)) {
- op_errno = ENOMEM;
- goto err;
- }
-
- stub = fop_symlink_stub (new_frame, sdfs_symlink_helper, linkname, loc,
- umask, xdata);
- if (!stub) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local = new_frame->local;
- local->stub = stub;
-
- STACK_WIND (new_frame, sdfs_entrylk_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &local->parent_loc, local->loc.name,
- ENTRYLK_LOCK, ENTRYLK_WRLCK, xdata);
-
- return 0;
+ sdfs_local_t *local = NULL;
+ call_frame_t *new_frame = NULL;
+ call_stub_t *stub = NULL;
+ int op_errno = 0;
+
+ if (-1 == sdfs_get_new_frame(frame, loc, &new_frame)) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ stub = fop_symlink_stub(new_frame, sdfs_symlink_helper, linkname, loc,
+ umask, xdata);
+ if (!stub) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local = new_frame->local;
+ local->stub = stub;
+
+ STACK_WIND(new_frame, sdfs_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &local->parent_loc,
+ local->loc.name, ENTRYLK_LOCK, ENTRYLK_WRLCK, xdata);
+
+ return 0;
err:
- STACK_UNWIND_STRICT (link, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL);
+ STACK_UNWIND_STRICT(link, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
- if (new_frame)
- SDFS_STACK_DESTROY (new_frame);
+ if (new_frame)
+ SDFS_STACK_DESTROY(new_frame);
- return 0;
+ return 0;
}
int
-sdfs_common_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+sdfs_common_entrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- int this_call_cnt = 0;
- int lk_index = 0;
- sdfs_lock_t *locks = NULL;
- call_stub_t *stub = NULL;
-
- local = frame->local;
- locks = local->lock;
- lk_index = (long) cookie;
-
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- } else {
- locks->entrylk->locked[lk_index] = _gf_true;
- }
-
- this_call_cnt = sdfs_frame_return (frame);
- if (this_call_cnt > 0) {
- gf_log (this->name, GF_LOG_DEBUG,
- "As there are more callcnt (%d) returning without WIND",
- this_call_cnt);
- return 0;
- }
+ sdfs_local_t *local = NULL;
+ int this_call_cnt = 0;
+ int lk_index = 0;
+ sdfs_lock_t *locks = NULL;
+ call_stub_t *stub = NULL;
- if (local->stub) {
- stub = local->stub;
- local->stub = NULL;
- call_resume (stub);
- } else {
- if (local->op_ret < 0)
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SDFS_MSG_ENTRYLK_ERROR,
- "unlocking entry lock failed ");
- SDFS_STACK_DESTROY (frame);
- }
+ local = frame->local;
+ locks = local->lock;
+ lk_index = (long)cookie;
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ } else {
+ locks->entrylk->locked[lk_index] = _gf_true;
+ }
+
+ this_call_cnt = sdfs_frame_return(frame);
+ if (this_call_cnt > 0) {
+ gf_log(this->name, GF_LOG_DEBUG,
+ "As there are more callcnt (%d) returning without WIND",
+ this_call_cnt);
return 0;
+ }
+
+ if (local->stub) {
+ stub = local->stub;
+ local->stub = NULL;
+ call_resume(stub);
+ } else {
+ if (local->op_ret < 0)
+ gf_msg(this->name, GF_LOG_ERROR, 0, SDFS_MSG_ENTRYLK_ERROR,
+ "unlocking entry lock failed ");
+ SDFS_STACK_DESTROY(frame);
+ }
+
+ return 0;
}
int
-sdfs_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+sdfs_link_cbk(call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode, struct iatt *stbuf,
+ struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- sdfs_lock_t *lock = NULL;
- int i = 0;
- int lock_count = 0;
-
- local = frame->local;
- lock = local->lock;
-
- STACK_UNWIND_STRICT (link, local->main_frame, op_ret, op_errno, inode,
- stbuf, preparent, postparent, xdata);
-
- local->main_frame = NULL;
- lock_count = lock->lock_count;
- for (i = 0; i < lock_count; i++) {
- STACK_WIND_COOKIE (frame, sdfs_common_entrylk_cbk,
- (void *)(long) i,
- FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &lock->entrylk[i].parent_loc,
- lock->entrylk[i].basename,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
- }
-
- return 0;
+ sdfs_local_t *local = NULL;
+ sdfs_lock_t *lock = NULL;
+ int i = 0;
+ int lock_count = 0;
+
+ local = frame->local;
+ lock = local->lock;
+
+ STACK_UNWIND_STRICT(link, local->main_frame, op_ret, op_errno, inode, stbuf,
+ preparent, postparent, xdata);
+
+ local->main_frame = NULL;
+ lock_count = lock->lock_count;
+ for (i = 0; i < lock_count; i++) {
+ STACK_WIND_COOKIE(frame, sdfs_common_entrylk_cbk, (void *)(long)i,
+ FIRST_CHILD(this), FIRST_CHILD(this)->fops->entrylk,
+ this->name, &lock->entrylk[i].parent_loc,
+ lock->entrylk[i].basename, ENTRYLK_UNLOCK,
+ ENTRYLK_WRLCK, xdata);
+ }
+
+ return 0;
}
int
-sdfs_link_helper (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
- loc_t *newloc, dict_t *xdata)
+sdfs_link_helper(call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- sdfs_lock_t *locks = NULL;
- gf_boolean_t stack_destroy = _gf_true;
- int lock_count = 0;
- int i = 0;
-
- local = frame->local;
- locks = local->lock;
-
- if (local->op_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SDFS_MSG_ENTRYLK_ERROR,
- "Acquiring entry lock failed");
- goto err;
- }
+ sdfs_local_t *local = NULL;
+ sdfs_lock_t *locks = NULL;
+ gf_boolean_t stack_destroy = _gf_true;
+ int lock_count = 0;
+ int i = 0;
- STACK_WIND (frame, sdfs_link_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->link, oldloc, newloc,
- xdata);
+ local = frame->local;
+ locks = local->lock;
- return 0;
-err:
- STACK_UNWIND_STRICT (link, local->main_frame, -1, local->op_errno,
- NULL, NULL, NULL, NULL, NULL);
+ if (local->op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SDFS_MSG_ENTRYLK_ERROR,
+ "Acquiring entry lock failed");
+ goto err;
+ }
- local->main_frame = NULL;
- for (i = 0; i < locks->lock_count && locks->entrylk->locked[i]; i++) {
- lock_count++;
- }
- GF_ATOMIC_INIT (local->call_cnt, lock_count);
-
- for (i = 0; i < lock_count; i++) {
- if (!locks->entrylk->locked[i]) {
- lock_count++;
- continue;
- }
-
- stack_destroy = _gf_false;
- STACK_WIND (frame, sdfs_common_entrylk_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &locks->entrylk[i].parent_loc,
- locks->entrylk[i].basename,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
+ STACK_WIND(frame, sdfs_link_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
+
+ return 0;
+err:
+ STACK_UNWIND_STRICT(link, local->main_frame, -1, local->op_errno, NULL,
+ NULL, NULL, NULL, NULL);
+
+ local->main_frame = NULL;
+ for (i = 0; i < locks->lock_count && locks->entrylk->locked[i]; i++) {
+ lock_count++;
+ }
+ GF_ATOMIC_INIT(local->call_cnt, lock_count);
+
+ for (i = 0; i < lock_count; i++) {
+ if (!locks->entrylk->locked[i]) {
+ lock_count++;
+ continue;
}
- if (stack_destroy)
- SDFS_STACK_DESTROY (frame);
+ stack_destroy = _gf_false;
+ STACK_WIND(frame, sdfs_common_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name,
+ &locks->entrylk[i].parent_loc, locks->entrylk[i].basename,
+ ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
+ }
- return 0;
+ if (stack_destroy)
+ SDFS_STACK_DESTROY(frame);
+
+ return 0;
}
static int
-sdfs_init_entry_lock (sdfs_entry_lock_t *lock, loc_t *loc)
+sdfs_init_entry_lock(sdfs_entry_lock_t *lock, loc_t *loc)
{
- int ret = 0;
+ int ret = 0;
- ret = sdfs_build_parent_loc (&lock->parent_loc, loc);
- if (ret)
- return -1;
+ ret = sdfs_build_parent_loc(&lock->parent_loc, loc);
+ if (ret)
+ return -1;
- lock->basename = gf_strdup (loc->name);
- if (!lock->basename)
- return -1;
+ lock->basename = gf_strdup(loc->name);
+ if (!lock->basename)
+ return -1;
- return 0;
+ return 0;
}
int
-sdfs_entry_lock_cmp (const void *l1, const void *l2)
+sdfs_entry_lock_cmp(const void *l1, const void *l2)
{
- const sdfs_entry_lock_t *r1 = l1;
- const sdfs_entry_lock_t *r2 = l2;
- int ret = 0;
- uuid_t gfid1 = {0};
- uuid_t gfid2 = {0};
-
- loc_gfid ((loc_t *)&r1->parent_loc, gfid1);
- loc_gfid ((loc_t *)&r2->parent_loc, gfid2);
- ret = gf_uuid_compare (gfid1, gfid2);
- /*Entrylks with NULL basename are the 'smallest'*/
- if (ret == 0) {
- if (!r1->basename)
- return -1;
- if (!r2->basename)
- return 1;
- ret = strcmp (r1->basename, r2->basename);
- }
-
- if (ret <= 0)
- return -1;
- else
- return 1;
+ const sdfs_entry_lock_t *r1 = l1;
+ const sdfs_entry_lock_t *r2 = l2;
+ int ret = 0;
+ uuid_t gfid1 = {0};
+ uuid_t gfid2 = {0};
+
+ loc_gfid((loc_t *)&r1->parent_loc, gfid1);
+ loc_gfid((loc_t *)&r2->parent_loc, gfid2);
+ ret = gf_uuid_compare(gfid1, gfid2);
+ /*Entrylks with NULL basename are the 'smallest'*/
+ if (ret == 0) {
+ if (!r1->basename)
+ return -1;
+ if (!r2->basename)
+ return 1;
+ ret = strcmp(r1->basename, r2->basename);
+ }
+
+ if (ret <= 0)
+ return -1;
+ else
+ return 1;
}
int
-sdfs_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
- loc_t *newloc, dict_t *xdata)
+sdfs_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- call_frame_t *new_frame = NULL;
- call_stub_t *stub = NULL;
- sdfs_lock_t *lock = NULL;
- client_t *client = NULL;
- int ret = 0;
- int op_errno = 0;
-
- new_frame = copy_frame (frame);
- if (!new_frame) {
- op_errno = ENOMEM;
- goto err;
- }
-
- gf_client_ref (client);
- new_frame->root->client = client;
- local = sdfs_local_init (new_frame, this);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
+ sdfs_local_t *local = NULL;
+ call_frame_t *new_frame = NULL;
+ call_stub_t *stub = NULL;
+ sdfs_lock_t *lock = NULL;
+ client_t *client = NULL;
+ int ret = 0;
+ int op_errno = ENOMEM;
+
+ new_frame = copy_frame(frame);
+ if (!new_frame) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+ /*Set unique lk-owner for the fop*/
+ set_lk_owner_from_ptr(&new_frame->root->lk_owner, new_frame->root);
+
+ gf_client_ref(client);
+ new_frame->root->client = client;
+ local = sdfs_local_init(new_frame, this);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->main_frame = frame;
+
+ lock = GF_CALLOC(1, sizeof(*lock), gf_common_mt_char);
+ if (!lock)
+ goto err;
+
+ local->lock = lock;
+
+ ret = sdfs_init_entry_lock(&lock->entrylk[0], newloc);
+ if (ret)
+ goto err;
+
+ ++lock->lock_count;
+
+ local->lock = lock;
+ GF_ATOMIC_INIT(local->call_cnt, lock->lock_count);
+
+ ret = loc_copy(&local->loc, newloc);
+ if (ret == -1) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ stub = fop_link_stub(new_frame, sdfs_link_helper, oldloc, newloc, xdata);
+ if (!stub) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->stub = stub;
+
+ STACK_WIND_COOKIE(new_frame, sdfs_common_entrylk_cbk, 0, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name,
+ &lock->entrylk[0].parent_loc, lock->entrylk[0].basename,
+ ENTRYLK_LOCK, ENTRYLK_WRLCK, xdata);
+
+ return 0;
+err:
- local->main_frame = frame;
+ STACK_UNWIND_STRICT(link, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
- lock = GF_CALLOC (1, sizeof (*lock), gf_common_mt_char);
- if (!lock)
- goto err;
+ if (new_frame)
+ SDFS_STACK_DESTROY(new_frame);
- local->lock = lock;
+ return 0;
+}
- ret = sdfs_init_entry_lock (&lock->entrylk[0], newloc);
- if (ret)
- goto err;
+int
+sdfs_mknod_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ sdfs_local_t *local = NULL;
- ++lock->lock_count;
+ local = frame->local;
- local->lock = lock;
- GF_ATOMIC_INIT (local->call_cnt, lock->lock_count);
+ STACK_UNWIND_STRICT(mknod, local->main_frame, op_ret, op_errno, inode,
+ stbuf, preparent, postparent, xdata);
- loc_copy (&local->loc, newloc);
- if (ret == -1) {
- op_errno = ENOMEM;
- goto err;
- }
+ local->main_frame = NULL;
+ STACK_WIND(frame, sdfs_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &local->parent_loc,
+ local->loc.name, ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
+ return 0;
+}
- stub = fop_link_stub (new_frame, sdfs_link_helper, oldloc,
- newloc, xdata);
- if (!stub) {
- op_errno = ENOMEM;
- goto err;
- }
+int
+sdfs_mknod_helper(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *xdata)
+{
+ sdfs_local_t *local = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- local->stub = stub;
+ local = frame->local;
- STACK_WIND_COOKIE (new_frame, sdfs_common_entrylk_cbk,
- 0, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &lock->entrylk[0].parent_loc,
- lock->entrylk[0].basename, ENTRYLK_LOCK,
- ENTRYLK_WRLCK, xdata);
+ gf_uuid_unparse(loc->pargfid, gfid);
- return 0;
-err:
+ if (local->op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SDFS_MSG_ENTRYLK_ERROR,
+ "Acquiring entry lock failed for directory %s "
+ "with parent gfid %s",
+ local->loc.name, gfid);
+ goto err;
+ }
- STACK_UNWIND_STRICT (link, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL);
+ STACK_WIND(frame, sdfs_mknod_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, loc, mode, rdev, umask, xdata);
- if (new_frame)
- SDFS_STACK_DESTROY (new_frame);
+ return 0;
+err:
+ STACK_UNWIND_STRICT(mknod, local->main_frame, -1, local->op_errno, NULL,
+ NULL, NULL, NULL, NULL);
- return 0;
+ local->main_frame = NULL;
+ SDFS_STACK_DESTROY(frame);
+ return 0;
}
int
-sdfs_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+sdfs_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
+ sdfs_local_t *local = NULL;
+ call_frame_t *new_frame = NULL;
+ call_stub_t *stub = NULL;
+ int op_errno = 0;
+
+ if (-1 == sdfs_get_new_frame(frame, loc, &new_frame)) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ stub = fop_mknod_stub(new_frame, sdfs_mknod_helper, loc, mode, rdev, umask,
+ xdata);
+ if (!stub) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local = new_frame->local;
+ local->stub = stub;
+
+ STACK_WIND(new_frame, sdfs_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &local->parent_loc,
+ local->loc.name, ENTRYLK_LOCK, ENTRYLK_WRLCK, xdata);
+
+ return 0;
+err:
+ STACK_UNWIND_STRICT(mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
- local = frame->local;
+ if (new_frame)
+ SDFS_STACK_DESTROY(new_frame);
- STACK_UNWIND_STRICT (mknod, local->main_frame, op_ret, op_errno, inode,
- stbuf, preparent, postparent, xdata);
+ return 0;
+}
- local->main_frame = NULL;
- STACK_WIND (frame, sdfs_entrylk_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &local->parent_loc, local->loc.name,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
- return 0;
+int
+sdfs_rename_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
+ struct iatt *preoldparent, struct iatt *postoldparent,
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
+{
+ sdfs_local_t *local = NULL;
+ sdfs_lock_t *lock = NULL;
+ int i = 0;
+ int call_cnt = 0;
+
+ local = frame->local;
+ lock = local->lock;
+ GF_ATOMIC_INIT(local->call_cnt, lock->lock_count);
+
+ STACK_UNWIND_STRICT(rename, local->main_frame, op_ret, op_errno, stbuf,
+ preoldparent, postoldparent, prenewparent,
+ postnewparent, xdata);
+
+ local->main_frame = NULL;
+ call_cnt = GF_ATOMIC_GET(local->call_cnt);
+
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND_COOKIE(frame, sdfs_common_entrylk_cbk, (void *)(long)i,
+ FIRST_CHILD(this), FIRST_CHILD(this)->fops->entrylk,
+ this->name, &lock->entrylk[i].parent_loc,
+ lock->entrylk[i].basename, ENTRYLK_UNLOCK,
+ ENTRYLK_WRLCK, xdata);
+ }
+
+ return 0;
}
int
-sdfs_mknod_helper (call_frame_t *frame, xlator_t *this, loc_t *loc,
- mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata)
+sdfs_rename_helper(call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+ loc_t *newloc, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
+ sdfs_local_t *local = NULL;
+ sdfs_lock_t *lock = NULL;
+ gf_boolean_t stack_destroy = _gf_true;
+ int lock_count = 0;
+ int i = 0;
- local = frame->local;
+ local = frame->local;
+ lock = local->lock;
- gf_uuid_unparse(loc->pargfid, gfid);
+ if (local->op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SDFS_MSG_ENTRYLK_ERROR,
+ "Acquiring entry lock failed ");
+ goto err;
+ }
- if (local->op_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SDFS_MSG_ENTRYLK_ERROR,
- "Acquiring entry lock failed for directory %s "
- "with parent gfid %s", local->loc.name, gfid);
- goto err;
- }
+ STACK_WIND(frame, sdfs_rename_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
- STACK_WIND (frame, sdfs_mknod_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mknod, loc, mode, rdev,
- umask, xdata);
+ return 0;
- return 0;
err:
- STACK_UNWIND_STRICT (mknod, local->main_frame, -1, local->op_errno,
- NULL, NULL, NULL, NULL, NULL);
+ STACK_UNWIND_STRICT(rename, local->main_frame, -1, local->op_errno, NULL,
+ NULL, NULL, NULL, NULL, NULL);
+
+ local->main_frame = NULL;
+ for (i = 0; i < lock->lock_count && lock->entrylk->locked[i]; i++) {
+ lock_count++;
+ }
+ GF_ATOMIC_INIT(local->call_cnt, lock_count);
+
+ for (i = 0; i < lock_count; i++) {
+ if (!lock->entrylk->locked[i]) {
+ lock_count++;
+ continue;
+ }
+ stack_destroy = _gf_false;
+ STACK_WIND(frame, sdfs_common_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name,
+ &lock->entrylk[i].parent_loc, lock->entrylk[i].basename,
+ ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
+ }
- local->main_frame = NULL;
- SDFS_STACK_DESTROY (frame);
- return 0;
+ if (stack_destroy)
+ SDFS_STACK_DESTROY(frame);
+
+ return 0;
}
int
-sdfs_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- dev_t rdev, mode_t umask, dict_t *xdata)
+sdfs_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- call_frame_t *new_frame = NULL;
- call_stub_t *stub = NULL;
- int op_errno = 0;
-
- if (-1 == sdfs_get_new_frame (frame, loc, &new_frame)) {
- op_errno = ENOMEM;
- goto err;
- }
-
- stub = fop_mknod_stub (new_frame, sdfs_mknod_helper, loc, mode,
- rdev, umask, xdata);
- if (!stub) {
- op_errno = ENOMEM;
- goto err;
- }
-
- local = new_frame->local;
- local->stub = stub;
-
- STACK_WIND (new_frame, sdfs_entrylk_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &local->parent_loc, local->loc.name,
- ENTRYLK_LOCK, ENTRYLK_WRLCK, xdata);
-
- return 0;
+ sdfs_local_t *local = NULL;
+ sdfs_lock_t *lock = NULL;
+ call_frame_t *new_frame = NULL;
+ call_stub_t *stub = NULL;
+ client_t *client = NULL;
+ int ret = 0;
+ int op_errno = ENOMEM;
+ int i = 0;
+ int call_cnt = 0;
+
+ new_frame = copy_frame(frame);
+ if (!new_frame) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+ /*Set unique lk-owner for the fop*/
+ set_lk_owner_from_ptr(&new_frame->root->lk_owner, new_frame->root);
+
+ gf_client_ref(client);
+ new_frame->root->client = client;
+ local = sdfs_local_init(new_frame, this);
+ if (!local) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->main_frame = frame;
+
+ lock = GF_CALLOC(1, sizeof(*lock), gf_common_mt_char);
+ if (!lock)
+ goto err;
+
+ local->lock = lock;
+
+ ret = sdfs_init_entry_lock(&lock->entrylk[0], oldloc);
+ if (ret)
+ goto err;
+ lock->entrylk->locked[0] = _gf_false;
+
+ ++lock->lock_count;
+
+ ret = sdfs_init_entry_lock(&lock->entrylk[1], newloc);
+ if (ret)
+ goto err;
+ lock->entrylk->locked[1] = _gf_false;
+
+ ++lock->lock_count;
+
+ qsort(lock->entrylk, lock->lock_count, sizeof(*lock->entrylk),
+ sdfs_entry_lock_cmp);
+
+ local->lock = lock;
+ GF_ATOMIC_INIT(local->call_cnt, lock->lock_count);
+
+ stub = fop_rename_stub(new_frame, sdfs_rename_helper, oldloc, newloc,
+ xdata);
+ if (!stub) {
+ op_errno = ENOMEM;
+ goto err;
+ }
+
+ local->stub = stub;
+ call_cnt = GF_ATOMIC_GET(local->call_cnt);
+ for (i = 0; i < call_cnt; i++) {
+ STACK_WIND_COOKIE(new_frame, sdfs_common_entrylk_cbk, (void *)(long)i,
+ FIRST_CHILD(this), FIRST_CHILD(this)->fops->entrylk,
+ this->name, &lock->entrylk[i].parent_loc,
+ lock->entrylk[i].basename, ENTRYLK_LOCK,
+ ENTRYLK_WRLCK, xdata);
+ }
+
+ return 0;
err:
- STACK_UNWIND_STRICT (mknod, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL);
- if (new_frame)
- SDFS_STACK_DESTROY (new_frame);
+ STACK_UNWIND_STRICT(rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+ NULL, NULL);
- return 0;
+ if (new_frame)
+ SDFS_STACK_DESTROY(new_frame);
+
+ return 0;
}
int
-sdfs_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent,
- dict_t *xdata)
+sdfs_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *stbuf, dict_t *xdata, struct iatt *postparent)
{
- sdfs_local_t *local = NULL;
- sdfs_lock_t *lock = NULL;
- int i = 0;
- int call_cnt = 0;
-
- local = frame->local;
- lock = local->lock;
- GF_ATOMIC_INIT (local->call_cnt, lock->lock_count);
-
- STACK_UNWIND_STRICT (rename, local->main_frame, op_ret, op_errno, stbuf,
- preoldparent, postoldparent, prenewparent,
- postnewparent, xdata);
-
- local->main_frame = NULL;
- call_cnt = GF_ATOMIC_GET (local->call_cnt);
-
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND_COOKIE (frame, sdfs_common_entrylk_cbk,
- (void *)(long) i,
- FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &lock->entrylk[i].parent_loc,
- lock->entrylk[i].basename,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
- }
+ sdfs_local_t *local = NULL;
+ local = frame->local;
+
+ if (!local->loc.parent) {
+ sdfs_local_cleanup(local);
+ frame->local = NULL;
+ STACK_UNWIND_STRICT(lookup, frame, op_ret, op_errno, inode, stbuf,
+ xdata, postparent);
return 0;
+ }
+
+ STACK_UNWIND_STRICT(lookup, local->main_frame, op_ret, op_errno, inode,
+ stbuf, xdata, postparent);
+
+ local->main_frame = NULL;
+ STACK_WIND(frame, sdfs_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &local->parent_loc,
+ local->loc.name, ENTRYLK_UNLOCK, ENTRYLK_RDLCK, xdata);
+ return 0;
}
int
-sdfs_rename_helper (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
- loc_t *newloc, dict_t *xdata)
+sdfs_lookup_helper(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- sdfs_lock_t *lock = NULL;
- gf_boolean_t stack_destroy = _gf_true;
- int lock_count = 0;
- int i = 0;
-
- local = frame->local;
- lock = local->lock;
-
- if (local->op_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SDFS_MSG_ENTRYLK_ERROR,
- "Acquiring entry lock failed ");
- goto err;
- }
+ sdfs_local_t *local = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- STACK_WIND (frame, sdfs_rename_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rename, oldloc, newloc,
- xdata);
+ local = frame->local;
- return 0;
+ gf_uuid_unparse(loc->pargfid, gfid);
-err:
- STACK_UNWIND_STRICT (rename, local->main_frame, -1, local->op_errno,
- NULL, NULL, NULL, NULL, NULL, NULL);
+ if (local->op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SDFS_MSG_ENTRYLK_ERROR,
+ "Acquiring entry lock failed for directory %s "
+ "with parent gfid %s",
+ local->loc.name, gfid);
+ goto err;
+ }
- local->main_frame = NULL;
- for (i = 0; i < lock->lock_count && lock->entrylk->locked[i]; i++) {
- lock_count++;
- }
- GF_ATOMIC_INIT (local->call_cnt, lock_count);
-
- for (i = 0; i < lock_count; i++) {
- if (!lock->entrylk->locked[i]) {
- lock_count++;
- continue;
- }
- stack_destroy = _gf_false;
- STACK_WIND (frame, sdfs_common_entrylk_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &lock->entrylk[i].parent_loc,
- lock->entrylk[i].basename,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK, xdata);
- }
+ STACK_WIND(frame, sdfs_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, xdata);
- if (stack_destroy)
- SDFS_STACK_DESTROY (frame);
+ return 0;
+err:
+ STACK_UNWIND_STRICT(lookup, local->main_frame, -1, local->op_errno, NULL,
+ NULL, NULL, NULL);
+ local->main_frame = NULL;
- return 0;
+ SDFS_STACK_DESTROY(frame);
+ return 0;
}
int
-sdfs_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
- loc_t *newloc, dict_t *xdata)
+sdfs_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- sdfs_lock_t *lock = NULL;
- call_frame_t *new_frame = NULL;
- call_stub_t *stub = NULL;
- client_t *client = NULL;
- int ret = 0;
- int op_errno = -1;
- int i = 0;
- int call_cnt = 0;
-
- new_frame = copy_frame (frame);
- if (!new_frame) {
- op_errno = ENOMEM;
- goto err;
- }
+ sdfs_local_t *local = NULL;
+ call_frame_t *new_frame = NULL;
+ call_stub_t *stub = NULL;
+ int op_errno = 0;
- gf_client_ref (client);
- new_frame->root->client = client;
- local = sdfs_local_init (new_frame, this);
+ if (!loc->parent) {
+ local = sdfs_local_init(frame, this);
if (!local) {
- op_errno = ENOMEM;
- goto err;
+ op_errno = ENOMEM;
+ goto err;
}
- local->main_frame = frame;
-
- lock = GF_CALLOC (1, sizeof (*lock), gf_common_mt_char);
- if (!lock)
- goto err;
-
- local->lock = lock;
-
- ret = sdfs_init_entry_lock (&lock->entrylk[0], oldloc);
- if (ret)
- goto err;
- lock->entrylk->locked[0] = _gf_false;
-
- ++lock->lock_count;
-
- ret = sdfs_init_entry_lock (&lock->entrylk[1], newloc);
- if (ret)
- goto err;
- lock->entrylk->locked[1] = _gf_false;
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, xdata);
+ return 0;
+ }
- ++lock->lock_count;
+ if (-1 == sdfs_get_new_frame(frame, loc, &new_frame)) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- qsort (lock->entrylk, lock->lock_count, sizeof (*lock->entrylk),
- sdfs_entry_lock_cmp);
+ stub = fop_lookup_stub(new_frame, sdfs_lookup_helper, loc, xdata);
+ if (!stub) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- local->lock = lock;
- GF_ATOMIC_INIT (local->call_cnt, lock->lock_count);
+ local = new_frame->local;
+ local->stub = stub;
- stub = fop_rename_stub (new_frame, sdfs_rename_helper, oldloc,
- newloc, xdata);
- if (!stub) {
- op_errno = ENOMEM;
- goto err;
- }
+ STACK_WIND(new_frame, sdfs_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &local->parent_loc,
+ local->loc.name, ENTRYLK_LOCK, ENTRYLK_RDLCK, xdata);
- local->stub = stub;
- call_cnt = GF_ATOMIC_GET (local->call_cnt);
- for (i = 0; i < call_cnt; i++) {
- STACK_WIND_COOKIE (new_frame, sdfs_common_entrylk_cbk,
- (void *)(long) i,
- FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &lock->entrylk[i].parent_loc,
- lock->entrylk[i].basename,
- ENTRYLK_LOCK, ENTRYLK_WRLCK, xdata);
- }
+ return 0;
- return 0;
err:
+ STACK_UNWIND_STRICT(lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
- STACK_UNWIND_STRICT (rename, frame, -1, op_errno, NULL, NULL,
- NULL, NULL, NULL, NULL);
+ if (new_frame)
+ SDFS_STACK_DESTROY(new_frame);
- if (new_frame)
- SDFS_STACK_DESTROY (new_frame);
-
- return 0;
+ return 0;
}
-int
-sdfs_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *stbuf, dict_t *xdata,
- struct iatt *postparent)
+int32_t
+sdfs_readdirp_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+ dict_t *xdata)
{
- sdfs_local_t *local = NULL;
-
- local = frame->local;
-
- if (!local->loc.parent) {
- sdfs_local_cleanup (local);
- frame->local = NULL;
- STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno,
- inode, stbuf, xdata, postparent);
- return 0;
- }
+ sdfs_local_t *local = NULL;
- STACK_UNWIND_STRICT (lookup, local->main_frame, op_ret, op_errno, inode,
- stbuf, xdata, postparent);
+ local = frame->local;
+ STACK_UNWIND_STRICT(readdirp, local->main_frame, op_ret, op_errno, entries,
+ xdata);
- local->main_frame = NULL;
- STACK_WIND (frame, sdfs_entrylk_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &local->parent_loc, local->loc.name,
- ENTRYLK_UNLOCK, ENTRYLK_RDLCK, xdata);
- return 0;
+ local->main_frame = NULL;
+ STACK_WIND(frame, sdfs_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &local->parent_loc,
+ NULL, ENTRYLK_UNLOCK, ENTRYLK_RDLCK, xdata);
+ return 0;
}
-int
-sdfs_lookup_helper (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *xdata)
+int32_t
+sdfs_readdirp_helper(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- char gfid[GF_UUID_BUF_SIZE] = {0};
+ sdfs_local_t *local = NULL;
+ char gfid[GF_UUID_BUF_SIZE] = {0};
- local = frame->local;
+ local = frame->local;
- gf_uuid_unparse(loc->pargfid, gfid);
+ gf_uuid_unparse(fd->inode->gfid, gfid);
- if (local->op_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SDFS_MSG_ENTRYLK_ERROR,
- "Acquiring entry lock failed for directory %s "
- "with parent gfid %s", local->loc.name, gfid);
- goto err;
- }
+ if (local->op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SDFS_MSG_ENTRYLK_ERROR,
+ "Acquiring entry lock failed for directory %s "
+ "with parent gfid %s",
+ local->loc.name, gfid);
+ goto err;
+ }
- STACK_WIND (frame, sdfs_lookup_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, loc, xdata);
+ STACK_WIND(frame, sdfs_readdirp_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdirp, fd, size, off, xdata);
- return 0;
+ return 0;
err:
- STACK_UNWIND_STRICT (lookup, local->main_frame, -1, local->op_errno,
- NULL, NULL, NULL, NULL);
- local->main_frame = NULL;
+ STACK_UNWIND_STRICT(readdirp, local->main_frame, -1, local->op_errno, NULL,
+ NULL);
- SDFS_STACK_DESTROY (frame);
- return 0;
+ local->main_frame = NULL;
+
+ SDFS_STACK_DESTROY(frame);
+ return 0;
}
-int
-sdfs_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *xdata)
+int32_t
+sdfs_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *xdata)
{
- sdfs_local_t *local = NULL;
- call_frame_t *new_frame = NULL;
- call_stub_t *stub = NULL;
- int op_errno = 0;
-
- if (!loc->parent) {
- local = sdfs_local_init (frame, this);
- if (!local) {
- op_errno = ENOMEM;
- goto err;
- }
-
- STACK_WIND_TAIL(frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup,
- loc, xdata);
- return 0;
- }
+ sdfs_local_t *local = NULL;
+ call_frame_t *new_frame = NULL;
+ call_stub_t *stub = NULL;
+ int op_errno = 0;
- if (-1 == sdfs_get_new_frame (frame, loc, &new_frame)) {
- op_errno = ENOMEM;
- goto err;
- }
+ if (-1 == sdfs_get_new_frame_readdirp(frame, fd, &new_frame)) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- stub = fop_lookup_stub (new_frame, sdfs_lookup_helper, loc,
- xdata);
- if (!stub) {
- op_errno = ENOMEM;
- goto err;
- }
+ stub = fop_readdirp_stub(new_frame, sdfs_readdirp_helper, fd, size, off,
+ xdata);
+ if (!stub) {
+ op_errno = ENOMEM;
+ goto err;
+ }
- local = new_frame->local;
- local->stub = stub;
+ local = new_frame->local;
+ local->stub = stub;
- STACK_WIND (new_frame, sdfs_entrylk_cbk,
- FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->entrylk,
- this->name, &local->parent_loc, local->loc.name,
- ENTRYLK_LOCK, ENTRYLK_RDLCK, xdata);
+ STACK_WIND(new_frame, sdfs_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &local->parent_loc,
+ NULL, ENTRYLK_LOCK, ENTRYLK_RDLCK, xdata);
- return 0;
+ return 0;
err:
- STACK_UNWIND_STRICT (lookup, frame, -1, op_errno, NULL, NULL,
- NULL, NULL);
+ STACK_UNWIND_STRICT(readdirp, frame, -1, op_errno, NULL, NULL);
- if (new_frame)
- SDFS_STACK_DESTROY (new_frame);
+ if (new_frame)
+ SDFS_STACK_DESTROY(new_frame);
- return 0;
+ return 0;
}
int
-init (xlator_t *this)
+init(xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- if (!this->children || this->children->next) {
- gf_log (this->name, GF_LOG_ERROR,
- "'dentry-fop-serializer' not configured with exactly one child");
- goto out;
- }
+ if (!this->children || this->children->next) {
+ gf_log(this->name, GF_LOG_ERROR,
+ "'dentry-fop-serializer' not configured with exactly one child");
+ goto out;
+ }
- if (!this->parents) {
- gf_log (this->name, GF_LOG_WARNING,
- "dangling volume. check volfile ");
- }
+ if (!this->parents) {
+ gf_log(this->name, GF_LOG_WARNING, "dangling volume. check volfile ");
+ }
- this->local_pool = mem_pool_new (sdfs_local_t, 512);
- if (!this->local_pool) {
- goto out;
- }
+ this->local_pool = mem_pool_new(sdfs_local_t, 512);
+ if (!this->local_pool) {
+ goto out;
+ }
+
+ GF_OPTION_INIT("pass-through", this->pass_through, bool, out);
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-fini (xlator_t *this)
+reconfigure(xlator_t *this, dict_t *options)
{
- mem_pool_destroy (this->local_pool);
+ int ret = -1;
- return 0;
+ GF_OPTION_RECONF("pass-through", this->pass_through, options, bool, out);
+
+ ret = 0;
+out:
+ return ret;
}
+void
+fini(xlator_t *this)
+{
+ mem_pool_destroy(this->local_pool);
+ this->local_pool = NULL;
+ return;
+}
struct xlator_fops fops = {
- .mkdir = sdfs_mkdir,
- .rmdir = sdfs_rmdir,
- .create = sdfs_create,
- .unlink = sdfs_unlink,
- .symlink = sdfs_symlink,
- .link = sdfs_link,
- .mknod = sdfs_mknod,
- .rename = sdfs_rename,
- .lookup = sdfs_lookup,
+ .mkdir = sdfs_mkdir,
+ .rmdir = sdfs_rmdir,
+ .create = sdfs_create,
+ .unlink = sdfs_unlink,
+ .symlink = sdfs_symlink,
+ .link = sdfs_link,
+ .mknod = sdfs_mknod,
+ .rename = sdfs_rename,
+ .lookup = sdfs_lookup,
+ .readdirp = sdfs_readdirp,
};
struct xlator_cbks cbks;
+struct volume_options options[] = {
+ {.key = {"pass-through"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "true",
+ .op_version = {GD_OP_VERSION_4_1_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_DOC | OPT_FLAG_CLIENT_OPT,
+ .tags = {"sdfs"},
+ .description = "Enable/Disable dentry serialize functionality"},
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .reconfigure = reconfigure,
+ .op_version = {GD_OP_VERSION_4_0_0},
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "sdfs",
+ .category = GF_TECH_PREVIEW,
+};
diff --git a/xlators/features/sdfs/src/sdfs.h b/xlators/features/sdfs/src/sdfs.h
index d28257eda5e..dded5a2d7fc 100644
--- a/xlators/features/sdfs/src/sdfs.h
+++ b/xlators/features/sdfs/src/sdfs.h
@@ -8,42 +8,42 @@
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
-#include "call-stub.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/call-stub.h>
#include "sdfs-messages.h"
-#include "atomic.h"
+#include <glusterfs/atomic.h>
-#define SDFS_LOCK_COUNT_MAX 2
+#define SDFS_LOCK_COUNT_MAX 2
-typedef struct{
- loc_t parent_loc;
- char *basename;
- int locked[SDFS_LOCK_COUNT_MAX];
+typedef struct {
+ loc_t parent_loc;
+ char *basename;
+ int locked[SDFS_LOCK_COUNT_MAX];
} sdfs_entry_lock_t;
typedef struct {
- sdfs_entry_lock_t entrylk[SDFS_LOCK_COUNT_MAX];
- int lock_count;
+ sdfs_entry_lock_t entrylk[SDFS_LOCK_COUNT_MAX];
+ int lock_count;
} sdfs_lock_t;
struct sdfs_local {
- call_frame_t *main_frame;
- loc_t loc;
- loc_t parent_loc;
- call_stub_t *stub;
- sdfs_lock_t *lock;
- int op_ret;
- int op_errno;
- gf_atomic_t call_cnt;
+ call_frame_t *main_frame;
+ loc_t loc;
+ loc_t parent_loc;
+ call_stub_t *stub;
+ sdfs_lock_t *lock;
+ int op_ret;
+ int op_errno;
+ gf_atomic_t call_cnt;
};
typedef struct sdfs_local sdfs_local_t;
-#define SDFS_STACK_DESTROY(frame) do { \
- sdfs_local_t *__local = NULL; \
- __local = frame->local; \
- frame->local = NULL; \
- gf_client_unref (frame->root->client); \
- STACK_DESTROY (frame->root); \
- sdfs_local_cleanup (__local); \
- } while (0)
-
+#define SDFS_STACK_DESTROY(frame) \
+ do { \
+ sdfs_local_t *__local = NULL; \
+ __local = frame->local; \
+ frame->local = NULL; \
+ gf_client_unref(frame->root->client); \
+ STACK_DESTROY(frame->root); \
+ sdfs_local_cleanup(__local); \
+ } while (0)
diff --git a/xlators/features/selinux/src/selinux-mem-types.h b/xlators/features/selinux/src/selinux-mem-types.h
index f9f356dd39a..553e59e5a9d 100644
--- a/xlators/features/selinux/src/selinux-mem-types.h
+++ b/xlators/features/selinux/src/selinux-mem-types.h
@@ -10,11 +10,10 @@
#ifndef __SELINUX_MEM_TYPES_H__
#define __SELINUX_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_selinux_mem_types_ {
- gf_selinux_mt_selinux_priv_t = gf_common_mt_end + 1,
- gf_selinux_mt_end
+ gf_selinux_mt_selinux_priv_t = gf_common_mt_end + 1,
+ gf_selinux_mt_end
};
#endif
-
diff --git a/xlators/features/selinux/src/selinux-messages.h b/xlators/features/selinux/src/selinux-messages.h
index 3a1663433ab..f49a54f956c 100644
--- a/xlators/features/selinux/src/selinux-messages.h
+++ b/xlators/features/selinux/src/selinux-messages.h
@@ -11,7 +11,7 @@
#ifndef _SELINUX_MESSAGES_H__
#define _SELINUX_MESSAGES_H__
-#include "glfs-message-id.h"
+#include <glusterfs/glfs-message-id.h>
/* To add new message IDs, append new identifiers at the end of the list.
*
@@ -23,13 +23,8 @@
* glfs-message-id.h.
*/
-GLFS_MSGID(SL,
- SL_MSG_INVALID_VOLFILE,
- SL_MSG_ENOMEM,
- SL_MSG_MEM_ACCT_INIT_FAILED,
- SL_MSG_SELINUX_GLUSTER_XATTR_MISSING,
- SL_MSG_SELINUX_XATTR_MISSING
-);
+GLFS_MSGID(SL, SL_MSG_INVALID_VOLFILE, SL_MSG_ENOMEM,
+ SL_MSG_MEM_ACCT_INIT_FAILED, SL_MSG_SELINUX_GLUSTER_XATTR_MISSING,
+ SL_MSG_SELINUX_XATTR_MISSING);
#endif /*_SELINUX_MESSAGES_H */
-
diff --git a/xlators/features/selinux/src/selinux.c b/xlators/features/selinux/src/selinux.c
index 2dd70599a86..9b1b4b55e1a 100644
--- a/xlators/features/selinux/src/selinux.c
+++ b/xlators/features/selinux/src/selinux.c
@@ -8,320 +8,316 @@
cases as published by the Free Software Foundation.
*/
-#include "xlator.h"
+#include <glusterfs/xlator.h>
#include "selinux.h"
#include "selinux-messages.h"
#include "selinux-mem-types.h"
-#include "compat-errno.h"
+#include <glusterfs/compat-errno.h>
static int
-selinux_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
+selinux_fgetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
{
- int ret = 0;
- char *name = cookie;
-
- if (op_errno == 0 && dict && name && (!strcmp(name, SELINUX_GLUSTER_XATTR))) {
- ret = dict_rename_key (dict, SELINUX_GLUSTER_XATTR,
- SELINUX_XATTR);
- if (ret < 0)
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- SL_MSG_SELINUX_GLUSTER_XATTR_MISSING,
- "getxattr failed for %s", SELINUX_XATTR);
-
- }
-
- STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno,
- dict, xdata);
- return ret;
+ int ret = 0;
+ char *name = cookie;
+
+ if (op_errno == 0 && dict && name &&
+ (!strcmp(name, SELINUX_GLUSTER_XATTR))) {
+ ret = dict_rename_key(dict, SELINUX_GLUSTER_XATTR, SELINUX_XATTR);
+ if (ret < 0)
+ gf_msg(this->name, GF_LOG_ERROR, op_errno,
+ SL_MSG_SELINUX_GLUSTER_XATTR_MISSING,
+ "getxattr failed for %s", SELINUX_XATTR);
+ }
+
+ STACK_UNWIND_STRICT(fgetxattr, frame, op_ret, op_errno, dict, xdata);
+ return ret;
}
-
static int
-selinux_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
+selinux_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata)
{
- selinux_priv_t *priv = NULL;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- char *xattr_name = (char *) name;
+ selinux_priv_t *priv = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ char *xattr_name = (char *)name;
- priv = this->private;
+ priv = this->private;
- GF_VALIDATE_OR_GOTO ("selinux", priv, err);
+ GF_VALIDATE_OR_GOTO("selinux", priv, err);
- /* name can be NULL for listxattr calls */
- if (!priv->selinux_enabled || !name)
- goto off;
+ /* name can be NULL for listxattr calls */
+ if (!priv->selinux_enabled || !name)
+ goto off;
- if (strcmp (name, SELINUX_XATTR) == 0)
- xattr_name = SELINUX_GLUSTER_XATTR;
+ if (strcmp(name, SELINUX_XATTR) == 0)
+ xattr_name = SELINUX_GLUSTER_XATTR;
off:
- STACK_WIND_COOKIE (frame, selinux_fgetxattr_cbk, xattr_name,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fgetxattr, fd, xattr_name,
- xdata);
- return 0;
+ STACK_WIND_COOKIE(frame, selinux_fgetxattr_cbk, xattr_name,
+ FIRST_CHILD(this), FIRST_CHILD(this)->fops->fgetxattr, fd,
+ xattr_name, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, NULL, xdata);
+ STACK_UNWIND_STRICT(fgetxattr, frame, op_ret, op_errno, NULL, xdata);
- return 0;
+ return 0;
}
static int
-selinux_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
+selinux_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
{
- int ret = 0;
- char *name = cookie;
-
- if (op_errno == 0 && dict && name && (!strcmp(name, SELINUX_GLUSTER_XATTR))) {
- ret = dict_rename_key (dict, SELINUX_GLUSTER_XATTR,
- SELINUX_XATTR);
- if (ret < 0)
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- SL_MSG_SELINUX_GLUSTER_XATTR_MISSING,
- "getxattr failed for %s", SELINUX_XATTR);
+ int ret = 0;
+ char *name = cookie;
- }
+ if (op_errno == 0 && dict && name &&
+ (!strcmp(name, SELINUX_GLUSTER_XATTR))) {
+ ret = dict_rename_key(dict, SELINUX_GLUSTER_XATTR, SELINUX_XATTR);
+ if (ret < 0)
+ gf_msg(this->name, GF_LOG_ERROR, op_errno,
+ SL_MSG_SELINUX_GLUSTER_XATTR_MISSING,
+ "getxattr failed for %s", SELINUX_XATTR);
+ }
- STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
+ STACK_UNWIND_STRICT(getxattr, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ return 0;
}
-
static int
-selinux_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+selinux_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- selinux_priv_t *priv = NULL;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- char *xattr_name = (char *) name;
+ selinux_priv_t *priv = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ char *xattr_name = (char *)name;
- priv = this->private;
+ priv = this->private;
- GF_VALIDATE_OR_GOTO ("selinux", priv, err);
+ GF_VALIDATE_OR_GOTO("selinux", priv, err);
- /* name can be NULL for listxattr calls */
- if (!priv->selinux_enabled || !name)
- goto off;
+ /* name can be NULL for listxattr calls */
+ if (!priv->selinux_enabled || !name)
+ goto off;
- if (strcmp (name, SELINUX_XATTR) == 0)
- xattr_name = SELINUX_GLUSTER_XATTR;
+ if (strcmp(name, SELINUX_XATTR) == 0)
+ xattr_name = SELINUX_GLUSTER_XATTR;
off:
- STACK_WIND_COOKIE (frame, selinux_getxattr_cbk, xattr_name,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr, loc, xattr_name,
- xdata);
- return 0;
+ STACK_WIND_COOKIE(frame, selinux_getxattr_cbk, xattr_name,
+ FIRST_CHILD(this), FIRST_CHILD(this)->fops->getxattr, loc,
+ xattr_name, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, NULL, xdata);
- return 0;
+ STACK_UNWIND_STRICT(getxattr, frame, op_ret, op_errno, NULL, xdata);
+ return 0;
}
static int
-selinux_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+selinux_fsetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
{
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
- return 0;
-
+ STACK_UNWIND_STRICT(fsetxattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
-
static int
-selinux_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
- int flags, dict_t *xdata)
+selinux_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int flags, dict_t *xdata)
{
- selinux_priv_t *priv = NULL;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- int32_t ret = -1;
+ selinux_priv_t *priv = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ int32_t ret = -1;
- priv = this->private;
+ priv = this->private;
- GF_VALIDATE_OR_GOTO ("selinux", priv, err);
+ GF_VALIDATE_OR_GOTO("selinux", priv, err);
- if (!priv->selinux_enabled && !dict)
- goto off;
+ if (!priv->selinux_enabled && !dict)
+ goto off;
- ret = dict_rename_key (dict, SELINUX_XATTR, SELINUX_GLUSTER_XATTR);
- if (ret < 0 && ret != -ENODATA)
- goto err;
+ ret = dict_rename_key(dict, SELINUX_XATTR, SELINUX_GLUSTER_XATTR);
+ if (ret < 0 && ret != -ENODATA)
+ goto err;
off:
- STACK_WIND (frame, selinux_fsetxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags,
- xdata);
-
+ STACK_WIND(frame, selinux_fsetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
- return 0;
+ return 0;
err:
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
- return 0;
-
+ STACK_UNWIND_STRICT(fsetxattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
static int
-selinux_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int op_ret, int op_errno, dict_t *xdata)
+selinux_setxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int op_ret, int op_errno, dict_t *xdata)
{
- STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
- return 0;
+ STACK_UNWIND_STRICT(setxattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
-
static int
-selinux_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *dict, int flags, dict_t *xdata)
+selinux_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int flags, dict_t *xdata)
{
- selinux_priv_t *priv = NULL;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- int32_t ret = -1;
+ selinux_priv_t *priv = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ int32_t ret = -1;
- priv = this->private;
+ priv = this->private;
- GF_VALIDATE_OR_GOTO ("selinux", priv, err);
+ GF_VALIDATE_OR_GOTO("selinux", priv, err);
- if (!priv->selinux_enabled && !dict)
- goto off;
+ if (!priv->selinux_enabled && !dict)
+ goto off;
- ret = dict_rename_key (dict, SELINUX_XATTR, SELINUX_GLUSTER_XATTR);
- if (ret < 0 && ret != -ENODATA)
- goto err;
+ ret = dict_rename_key(dict, SELINUX_XATTR, SELINUX_GLUSTER_XATTR);
+ if (ret < 0 && ret != -ENODATA)
+ goto err;
off:
- STACK_WIND (frame, selinux_setxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setxattr, loc, dict, flags,
- xdata);
- return 0;
+ STACK_WIND(frame, selinux_setxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, loc, dict, flags, xdata);
+ return 0;
err:
- STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
- return 0;
+ STACK_UNWIND_STRICT(setxattr, frame, op_ret, op_errno, xdata);
+ return 0;
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO("selinux", this, out);
+ GF_VALIDATE_OR_GOTO("selinux", this, out);
- ret = xlator_mem_acct_init (this, gf_selinux_mt_end + 1);
+ ret = xlator_mem_acct_init(this, gf_selinux_mt_end + 1);
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SL_MSG_MEM_ACCT_INIT_FAILED,
- "Memory accounting init failed");
- return ret;
- }
-out:
+ if (ret != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SL_MSG_MEM_ACCT_INIT_FAILED,
+ "Memory accounting init failed");
return ret;
+ }
+out:
+ return ret;
}
int32_t
-init (xlator_t *this)
+init(xlator_t *this)
{
- int32_t ret = -1;
- selinux_priv_t *priv = NULL;
-
- GF_VALIDATE_OR_GOTO ("selinux", this, out);
-
- if (!this->children || this->children->next) {
- gf_msg (this->name, GF_LOG_WARNING, 0, SL_MSG_INVALID_VOLFILE,
- "Error: SELinux (%s) not configured with exactly one "
- "child", this->name);
- return -1;
- }
-
- if (this->parents == NULL) {
- gf_msg (this->name, GF_LOG_WARNING, 0, SL_MSG_INVALID_VOLFILE,
- "Dangling volume. Please check the volfile");
- }
-
- priv = GF_CALLOC (1, sizeof (*priv), gf_selinux_mt_selinux_priv_t);
- if (!priv) {
- gf_log (this->name, GF_LOG_ERROR, "out of memory");
- ret = ENOMEM;
- goto out;
- }
-
- GF_OPTION_INIT ("selinux", priv->selinux_enabled, bool, out);
-
- this->local_pool = mem_pool_new (selinux_priv_t, 64);
- if (!this->local_pool) {
- ret = -1;
- gf_msg (this->name, GF_LOG_ERROR, ENOMEM, SL_MSG_ENOMEM,
- "Failed to create local_t's memory pool");
- goto out;
- }
-
- this->private = (void *)priv;
- ret = 0;
+ int32_t ret = -1;
+ selinux_priv_t *priv = NULL;
+
+ GF_VALIDATE_OR_GOTO("selinux", this, out);
+
+ if (!this->children || this->children->next) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, SL_MSG_INVALID_VOLFILE,
+ "Error: SELinux (%s) not configured with exactly one "
+ "child",
+ this->name);
+ return -1;
+ }
+
+ if (this->parents == NULL) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, SL_MSG_INVALID_VOLFILE,
+ "Dangling volume. Please check the volfile");
+ }
+
+ priv = GF_CALLOC(1, sizeof(*priv), gf_selinux_mt_selinux_priv_t);
+ if (!priv) {
+ gf_log(this->name, GF_LOG_ERROR, "out of memory");
+ goto out;
+ }
+
+ GF_OPTION_INIT("selinux", priv->selinux_enabled, bool, out);
+
+ this->local_pool = mem_pool_new(selinux_priv_t, 64);
+ if (!this->local_pool) {
+ gf_msg(this->name, GF_LOG_ERROR, ENOMEM, SL_MSG_ENOMEM,
+ "Failed to create local_t's memory pool");
+ goto out;
+ }
+
+ this->private = (void *)priv;
+ ret = 0;
out:
- if (ret) {
- if (priv) {
- GF_FREE (priv);
- }
- mem_pool_destroy (this->local_pool);
- }
- return ret;
+ if (ret) {
+ GF_FREE(priv);
+ mem_pool_destroy(this->local_pool);
+ this->local_pool = NULL;
+ }
+ return ret;
}
int
-reconfigure (xlator_t *this, dict_t *options)
+reconfigure(xlator_t *this, dict_t *options)
{
- int32_t ret = -1;
- selinux_priv_t *priv = NULL;
+ int32_t ret = -1;
+ selinux_priv_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- GF_OPTION_RECONF ("selinux", priv->selinux_enabled, options,
- bool, out);
+ GF_OPTION_RECONF("selinux", priv->selinux_enabled, options, bool, out);
- ret = 0;
+ ret = 0;
out:
- return ret;
-
+ return ret;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- selinux_priv_t *priv = NULL;
+ selinux_priv_t *priv = NULL;
- priv = this->private;
- GF_FREE (priv);
+ priv = this->private;
+ GF_FREE(priv);
- mem_pool_destroy (this->local_pool);
+ mem_pool_destroy(this->local_pool);
+ this->local_pool = NULL;
- return;
+ return;
}
struct xlator_fops fops = {
- .getxattr = selinux_getxattr,
- .fgetxattr = selinux_fgetxattr,
- .setxattr = selinux_setxattr,
- .fsetxattr = selinux_fsetxattr,
+ .getxattr = selinux_getxattr,
+ .fgetxattr = selinux_fgetxattr,
+ .setxattr = selinux_setxattr,
+ .fsetxattr = selinux_fsetxattr,
};
-struct xlator_cbks cbks = {
-};
+struct xlator_cbks cbks = {};
struct volume_options options[] = {
- { .key = { "selinux" },
- .type = GF_OPTION_TYPE_BOOL,
- .default_value = "on",
- .description = "Enable/disable selinux translator",
- .op_version = {GD_OP_VERSION_3_11_0},
- .flags = OPT_FLAG_SETTABLE,
- .tags = {"security" , "linux"},
- },
- { .key = { NULL }, }
+ {
+ .key = {"selinux"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "on",
+ .description = "Enable/disable selinux translator",
+ .op_version = {GD_OP_VERSION_3_11_0},
+ .flags = OPT_FLAG_SETTABLE,
+ .tags = {"security", "linux"},
+ },
+ {
+ .key = {NULL},
+ }};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "selinux",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/features/selinux/src/selinux.h b/xlators/features/selinux/src/selinux.h
index 2ff57b5080a..1bbdad3bb36 100644
--- a/xlators/features/selinux/src/selinux.h
+++ b/xlators/features/selinux/src/selinux.h
@@ -10,13 +10,13 @@
#ifndef __SELINUX_H__
#define __SELINUX_H__
-#include "common-utils.h"
+#include <glusterfs/common-utils.h>
#define SELINUX_XATTR "security.selinux"
#define SELINUX_GLUSTER_XATTR "trusted.glusterfs.selinux"
struct selinux_priv {
- gf_boolean_t selinux_enabled;
+ gf_boolean_t selinux_enabled;
};
typedef struct selinux_priv selinux_priv_t;
diff --git a/xlators/features/shard/src/shard-mem-types.h b/xlators/features/shard/src/shard-mem-types.h
index 77f0cee7f58..1fe7e2e2798 100644
--- a/xlators/features/shard/src/shard-mem-types.h
+++ b/xlators/features/shard/src/shard-mem-types.h
@@ -10,14 +10,15 @@
#ifndef __SHARD_MEM_TYPES_H__
#define __SHARD_MEM_TYPES_H__
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum gf_shard_mem_types_ {
- gf_shard_mt_priv_t = gf_common_mt_end + 1,
- gf_shard_mt_inode_list,
- gf_shard_mt_inode_ctx_t,
- gf_shard_mt_iovec,
- gf_shard_mt_int64_t,
- gf_shard_mt_end
+ gf_shard_mt_priv_t = gf_common_mt_end + 1,
+ gf_shard_mt_inode_list,
+ gf_shard_mt_inode_ctx_t,
+ gf_shard_mt_iovec,
+ gf_shard_mt_int64_t,
+ gf_shard_mt_uint64_t,
+ gf_shard_mt_end
};
#endif
diff --git a/xlators/features/shard/src/shard-messages.h b/xlators/features/shard/src/shard-messages.h
index d7296fc214f..2d0867eb136 100644
--- a/xlators/features/shard/src/shard-messages.h
+++ b/xlators/features/shard/src/shard-messages.h
@@ -11,7 +11,7 @@
#ifndef _SHARD_MESSAGES_H_
#define _SHARD_MESSAGES_H_
-#include "glfs-message-id.h"
+#include <glusterfs/glfs-message-id.h>
/* To add new message IDs, append new identifiers at the end of the list.
*
@@ -23,26 +23,17 @@
* glfs-message-id.h.
*/
-GLFS_MSGID(SHARD,
- SHARD_MSG_BASE_FILE_LOOKUP_FAILED,
- SHARD_MSG_DICT_SET_FAILED,
- SHARD_MSG_DOT_SHARD_NODIR,
- SHARD_MSG_FD_CTX_SET_FAILED,
- SHARD_MSG_INODE_CTX_GET_FAILED,
- SHARD_MSG_INODE_CTX_SET_FAILED,
- SHARD_MSG_INODE_PATH_FAILED,
- SHARD_MSG_INTERNAL_XATTR_MISSING,
- SHARD_MSG_INVALID_VOLFILE,
- SHARD_MSG_LOOKUP_SHARD_FAILED,
- SHARD_MSG_MEM_ACCT_INIT_FAILED,
- SHARD_MSG_NULL_THIS,
- SHARD_MSG_SIZE_SET_FAILED,
- SHARD_MSG_STAT_FAILED,
- SHARD_MSG_TRUNCATE_LAST_SHARD_FAILED,
- SHARD_MSG_UPDATE_FILE_SIZE_FAILED,
- SHARD_MSG_FOP_NOT_SUPPORTED,
- SHARD_MSG_INVALID_FOP,
- SHARD_MSG_MEMALLOC_FAILED
-);
+GLFS_MSGID(SHARD, SHARD_MSG_BASE_FILE_LOOKUP_FAILED, SHARD_MSG_DICT_OP_FAILED,
+ SHARD_MSG_DOT_SHARD_NODIR, SHARD_MSG_FD_CTX_SET_FAILED,
+ SHARD_MSG_INODE_CTX_GET_FAILED, SHARD_MSG_INODE_CTX_SET_FAILED,
+ SHARD_MSG_INODE_PATH_FAILED, SHARD_MSG_INTERNAL_XATTR_MISSING,
+ SHARD_MSG_INVALID_VOLFILE, SHARD_MSG_LOOKUP_SHARD_FAILED,
+ SHARD_MSG_MEM_ACCT_INIT_FAILED, SHARD_MSG_NULL_THIS,
+ SHARD_MSG_SIZE_SET_FAILED, SHARD_MSG_STAT_FAILED,
+ SHARD_MSG_TRUNCATE_LAST_SHARD_FAILED,
+ SHARD_MSG_UPDATE_FILE_SIZE_FAILED, SHARD_MSG_FOP_NOT_SUPPORTED,
+ SHARD_MSG_INVALID_FOP, SHARD_MSG_MEMALLOC_FAILED,
+ SHARD_MSG_FOP_FAILED, SHARD_MSG_SHARDS_DELETION_FAILED,
+ SHARD_MSG_SHARD_DELETION_COMPLETED);
#endif /* !_SHARD_MESSAGES_H_ */
diff --git a/xlators/features/shard/src/shard.c b/xlators/features/shard/src/shard.c
index 5d493e5e675..e5f93063943 100644
--- a/xlators/features/shard/src/shard.c
+++ b/xlators/features/shard/src/shard.c
@@ -12,2396 +12,2741 @@
#include "shard.h"
#include "shard-mem-types.h"
-#include "byte-order.h"
-#include "defaults.h"
-#include "statedump.h"
+#include <glusterfs/byte-order.h>
+#include <glusterfs/defaults.h>
+#include <glusterfs/statedump.h>
static gf_boolean_t
-__is_shard_dir (uuid_t gfid)
+__is_shard_dir(uuid_t gfid)
{
- shard_priv_t *priv = THIS->private;
+ shard_priv_t *priv = THIS->private;
- if (gf_uuid_compare (gfid, priv->dot_shard_gfid) == 0)
- return _gf_true;
+ if (gf_uuid_compare(gfid, priv->dot_shard_gfid) == 0)
+ return _gf_true;
- return _gf_false;
+ return _gf_false;
}
static gf_boolean_t
-__is_gsyncd_on_shard_dir (call_frame_t *frame, loc_t *loc)
+__is_gsyncd_on_shard_dir(call_frame_t *frame, loc_t *loc)
{
- if (frame->root->pid == GF_CLIENT_PID_GSYNCD &&
- (__is_shard_dir (loc->pargfid) ||
- (loc->parent && __is_shard_dir(loc->parent->gfid))))
- return _gf_true;
+ if (frame->root->pid == GF_CLIENT_PID_GSYNCD &&
+ (__is_shard_dir(loc->pargfid) ||
+ (loc->parent && __is_shard_dir(loc->parent->gfid))))
+ return _gf_true;
- return _gf_false;
+ return _gf_false;
}
void
-shard_make_block_bname (int block_num, uuid_t gfid, char *buf, size_t len)
+shard_make_block_bname(int block_num, uuid_t gfid, char *buf, size_t len)
{
- char gfid_str[GF_UUID_BUF_SIZE] = {0,};
+ char gfid_str[GF_UUID_BUF_SIZE] = {
+ 0,
+ };
- gf_uuid_unparse (gfid, gfid_str);
- snprintf (buf, len, "%s.%d", gfid_str, block_num);
+ gf_uuid_unparse(gfid, gfid_str);
+ snprintf(buf, len, "%s.%d", gfid_str, block_num);
}
void
-shard_make_block_abspath (int block_num, uuid_t gfid, char *filepath,
- size_t len)
+shard_make_block_abspath(int block_num, uuid_t gfid, char *filepath, size_t len)
{
- char gfid_str[GF_UUID_BUF_SIZE] = {0,};
+ char gfid_str[GF_UUID_BUF_SIZE] = {
+ 0,
+ };
- gf_uuid_unparse (gfid, gfid_str);
- snprintf (filepath, len, "/%s/%s.%d", GF_SHARD_DIR, gfid_str,
- block_num);
+ gf_uuid_unparse(gfid, gfid_str);
+ snprintf(filepath, len, "/%s/%s.%d", GF_SHARD_DIR, gfid_str, block_num);
}
int
-__shard_inode_ctx_get (inode_t *inode, xlator_t *this, shard_inode_ctx_t **ctx)
+__shard_inode_ctx_get(inode_t *inode, xlator_t *this, shard_inode_ctx_t **ctx)
{
- int ret = -1;
- uint64_t ctx_uint = 0;
- shard_inode_ctx_t *ctx_p = NULL;
+ int ret = -1;
+ uint64_t ctx_uint = 0;
+ shard_inode_ctx_t *ctx_p = NULL;
- ret = __inode_ctx_get (inode, this, &ctx_uint);
- if (ret == 0) {
- *ctx = (shard_inode_ctx_t *) ctx_uint;
- return ret;
- }
+ ret = __inode_ctx_get(inode, this, &ctx_uint);
+ if (ret == 0) {
+ *ctx = (shard_inode_ctx_t *)(uintptr_t)ctx_uint;
+ return ret;
+ }
- ctx_p = GF_CALLOC (1, sizeof (*ctx_p), gf_shard_mt_inode_ctx_t);
- if (!ctx_p)
- return ret;
+ ctx_p = GF_CALLOC(1, sizeof(*ctx_p), gf_shard_mt_inode_ctx_t);
+ if (!ctx_p)
+ return ret;
- INIT_LIST_HEAD (&ctx_p->ilist);
- INIT_LIST_HEAD (&ctx_p->to_fsync_list);
+ INIT_LIST_HEAD(&ctx_p->ilist);
+ INIT_LIST_HEAD(&ctx_p->to_fsync_list);
- ret = __inode_ctx_set (inode, this, (uint64_t *)&ctx_p);
- if (ret < 0) {
- GF_FREE (ctx_p);
- return ret;
- }
+ ctx_uint = (uint64_t)(uintptr_t)ctx_p;
+ ret = __inode_ctx_set(inode, this, &ctx_uint);
+ if (ret < 0) {
+ GF_FREE(ctx_p);
+ return ret;
+ }
- *ctx = ctx_p;
+ *ctx = ctx_p;
- return ret;
+ return ret;
}
int
-shard_inode_ctx_get (inode_t *inode, xlator_t *this, shard_inode_ctx_t **ctx)
+shard_inode_ctx_get(inode_t *inode, xlator_t *this, shard_inode_ctx_t **ctx)
{
- int ret = 0;
+ int ret = 0;
- LOCK(&inode->lock);
- {
- ret = __shard_inode_ctx_get (inode, this, ctx);
- }
- UNLOCK(&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __shard_inode_ctx_get(inode, this, ctx);
+ }
+ UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
int
-__shard_inode_ctx_set (inode_t *inode, xlator_t *this, struct iatt *stbuf,
- uint64_t block_size, int32_t valid)
+__shard_inode_ctx_set(inode_t *inode, xlator_t *this, struct iatt *stbuf,
+ uint64_t block_size, int32_t valid)
{
- int ret = -1;
- shard_inode_ctx_t *ctx = NULL;
-
- ret = __shard_inode_ctx_get (inode, this, &ctx);
- if (ret)
- return ret;
+ int ret = -1;
+ shard_inode_ctx_t *ctx = NULL;
- if (valid & SHARD_MASK_BLOCK_SIZE)
- ctx->block_size = block_size;
+ ret = __shard_inode_ctx_get(inode, this, &ctx);
+ if (ret)
+ return ret;
- if (!stbuf)
- return 0;
+ if (valid & SHARD_MASK_BLOCK_SIZE)
+ ctx->block_size = block_size;
- if (valid & SHARD_MASK_PROT)
- ctx->stat.ia_prot = stbuf->ia_prot;
+ if (valid & SHARD_MASK_PROT)
+ ctx->stat.ia_prot = stbuf->ia_prot;
- if (valid & SHARD_MASK_NLINK)
- ctx->stat.ia_nlink = stbuf->ia_nlink;
+ if (valid & SHARD_MASK_NLINK)
+ ctx->stat.ia_nlink = stbuf->ia_nlink;
- if (valid & SHARD_MASK_UID)
- ctx->stat.ia_uid = stbuf->ia_uid;
+ if (valid & SHARD_MASK_UID)
+ ctx->stat.ia_uid = stbuf->ia_uid;
- if (valid & SHARD_MASK_GID)
- ctx->stat.ia_gid = stbuf->ia_gid;
+ if (valid & SHARD_MASK_GID)
+ ctx->stat.ia_gid = stbuf->ia_gid;
- if (valid & SHARD_MASK_SIZE)
- ctx->stat.ia_size = stbuf->ia_size;
+ if (valid & SHARD_MASK_SIZE)
+ ctx->stat.ia_size = stbuf->ia_size;
- if (valid & SHARD_MASK_BLOCKS)
- ctx->stat.ia_blocks = stbuf->ia_blocks;
+ if (valid & SHARD_MASK_BLOCKS)
+ ctx->stat.ia_blocks = stbuf->ia_blocks;
- if (valid & SHARD_MASK_TIMES) {
- SHARD_TIME_UPDATE (ctx->stat.ia_mtime, ctx->stat.ia_mtime_nsec,
- stbuf->ia_mtime, stbuf->ia_mtime_nsec);
- SHARD_TIME_UPDATE (ctx->stat.ia_ctime, ctx->stat.ia_ctime_nsec,
- stbuf->ia_ctime, stbuf->ia_ctime_nsec);
- SHARD_TIME_UPDATE (ctx->stat.ia_atime, ctx->stat.ia_atime_nsec,
- stbuf->ia_atime, stbuf->ia_atime_nsec);
- }
+ if (valid & SHARD_MASK_TIMES) {
+ SHARD_TIME_UPDATE(ctx->stat.ia_mtime, ctx->stat.ia_mtime_nsec,
+ stbuf->ia_mtime, stbuf->ia_mtime_nsec);
+ SHARD_TIME_UPDATE(ctx->stat.ia_ctime, ctx->stat.ia_ctime_nsec,
+ stbuf->ia_ctime, stbuf->ia_ctime_nsec);
+ SHARD_TIME_UPDATE(ctx->stat.ia_atime, ctx->stat.ia_atime_nsec,
+ stbuf->ia_atime, stbuf->ia_atime_nsec);
+ }
- if (valid & SHARD_MASK_OTHERS) {
- ctx->stat.ia_ino = stbuf->ia_ino;
- gf_uuid_copy (ctx->stat.ia_gfid, stbuf->ia_gfid);
- ctx->stat.ia_dev = stbuf->ia_dev;
- ctx->stat.ia_type = stbuf->ia_type;
- ctx->stat.ia_rdev = stbuf->ia_rdev;
- ctx->stat.ia_blksize = stbuf->ia_blksize;
- }
+ if (valid & SHARD_MASK_OTHERS) {
+ ctx->stat.ia_ino = stbuf->ia_ino;
+ gf_uuid_copy(ctx->stat.ia_gfid, stbuf->ia_gfid);
+ ctx->stat.ia_dev = stbuf->ia_dev;
+ ctx->stat.ia_type = stbuf->ia_type;
+ ctx->stat.ia_rdev = stbuf->ia_rdev;
+ ctx->stat.ia_blksize = stbuf->ia_blksize;
+ }
- if (valid & SHARD_MASK_REFRESH_RESET)
- ctx->refresh = _gf_false;
+ if (valid & SHARD_MASK_REFRESH_RESET)
+ ctx->refresh = _gf_false;
- return 0;
+ return 0;
}
int
-shard_inode_ctx_set (inode_t *inode, xlator_t *this, struct iatt *stbuf,
- uint64_t block_size, int32_t valid)
+shard_inode_ctx_set(inode_t *inode, xlator_t *this, struct iatt *stbuf,
+ uint64_t block_size, int32_t valid)
{
- int ret = -1;
+ int ret = -1;
- LOCK (&inode->lock);
- {
- ret = __shard_inode_ctx_set (inode, this, stbuf, block_size,
- valid);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __shard_inode_ctx_set(inode, this, stbuf, block_size, valid);
+ }
+ UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
int
-__shard_inode_ctx_set_refreshed_flag (inode_t *inode, xlator_t *this)
+__shard_inode_ctx_set_refresh_flag(inode_t *inode, xlator_t *this)
{
- int ret = -1;
- shard_inode_ctx_t *ctx = NULL;
+ int ret = -1;
+ shard_inode_ctx_t *ctx = NULL;
- ret = __shard_inode_ctx_get (inode, this, &ctx);
- if (ret)
- return ret;
+ ret = __shard_inode_ctx_get(inode, this, &ctx);
+ if (ret)
+ return ret;
- ctx->refreshed = _gf_true;
- return 0;
-}
+ ctx->refresh = _gf_true;
+ return 0;
+}
int
-shard_inode_ctx_set_refreshed_flag (inode_t *inode, xlator_t *this)
+shard_inode_ctx_set_refresh_flag(inode_t *inode, xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- LOCK (&inode->lock);
- {
- ret = __shard_inode_ctx_set_refreshed_flag (inode, this);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __shard_inode_ctx_set_refresh_flag(inode, this);
+ }
+ UNLOCK(&inode->lock);
+
+ return ret;
+}
+
+int
+__shard_inode_ctx_mark_dir_refreshed(inode_t *inode, xlator_t *this)
+{
+ int ret = -1;
+ shard_inode_ctx_t *ctx = NULL;
+ ret = __shard_inode_ctx_get(inode, this, &ctx);
+ if (ret)
return ret;
+
+ ctx->refreshed = _gf_true;
+ return 0;
}
int
-__shard_inode_ctx_add_to_fsync_list (inode_t *base_inode, xlator_t *this,
- inode_t *shard_inode)
+shard_inode_ctx_mark_dir_refreshed(inode_t *inode, xlator_t *this)
{
- int ret = -1;
- shard_inode_ctx_t *base_ictx = NULL;
- shard_inode_ctx_t *shard_ictx = NULL;
+ int ret = -1;
- ret = __shard_inode_ctx_get (base_inode, this, &base_ictx);
- if (ret)
- return ret;
+ LOCK(&inode->lock);
+ {
+ ret = __shard_inode_ctx_mark_dir_refreshed(inode, this);
+ }
+ UNLOCK(&inode->lock);
- ret = __shard_inode_ctx_get (shard_inode, this, &shard_ictx);
- if (ret)
- return ret;
+ return ret;
+}
- if (shard_ictx->fsync_needed) {
- shard_ictx->fsync_needed++;
- return 1;
- }
+int
+__shard_inode_ctx_add_to_fsync_list(inode_t *base_inode, xlator_t *this,
+ inode_t *shard_inode)
+{
+ int ret = -1;
+ shard_inode_ctx_t *base_ictx = NULL;
+ shard_inode_ctx_t *shard_ictx = NULL;
- list_add_tail (&shard_ictx->to_fsync_list, &base_ictx->to_fsync_list);
- shard_ictx->inode = shard_inode;
+ ret = __shard_inode_ctx_get(base_inode, this, &base_ictx);
+ if (ret)
+ return ret;
+
+ ret = __shard_inode_ctx_get(shard_inode, this, &shard_ictx);
+ if (ret)
+ return ret;
+
+ if (shard_ictx->fsync_needed) {
shard_ictx->fsync_needed++;
- base_ictx->fsync_count++;
- shard_ictx->base_inode = base_inode;
+ return 1;
+ }
- return 0;
+ list_add_tail(&shard_ictx->to_fsync_list, &base_ictx->to_fsync_list);
+ shard_ictx->inode = shard_inode;
+ shard_ictx->fsync_needed++;
+ base_ictx->fsync_count++;
+ shard_ictx->base_inode = base_inode;
+
+ return 0;
}
int
-shard_inode_ctx_add_to_fsync_list (inode_t *base_inode, xlator_t *this,
- inode_t *shard_inode)
+shard_inode_ctx_add_to_fsync_list(inode_t *base_inode, xlator_t *this,
+ inode_t *shard_inode)
{
- int ret = -1;
+ int ret = -1;
- /* This ref acts as a refkeepr on the base inode. We
- * need to keep this inode alive as it holds the head
- * of the to_fsync_list.
- */
- inode_ref (base_inode);
+ /* This ref acts as a refkeepr on the base inode. We
+ * need to keep this inode alive as it holds the head
+ * of the to_fsync_list.
+ */
+ inode_ref(base_inode);
+ inode_ref(shard_inode);
- LOCK (&base_inode->lock);
- LOCK (&shard_inode->lock);
- {
- ret = __shard_inode_ctx_add_to_fsync_list (base_inode, this,
- shard_inode);
- }
- UNLOCK (&shard_inode->lock);
- UNLOCK (&base_inode->lock);
+ LOCK(&base_inode->lock);
+ LOCK(&shard_inode->lock);
+ {
+ ret = __shard_inode_ctx_add_to_fsync_list(base_inode, this,
+ shard_inode);
+ }
+ UNLOCK(&shard_inode->lock);
+ UNLOCK(&base_inode->lock);
- /* Unref the base inode corresponding to the ref above, if the shard is
- * found to be already part of the fsync list.
- */
- if (ret != 0)
- inode_unref (base_inode);
- return ret;
+ /* Unref the base inode corresponding to the ref above, if the shard is
+ * found to be already part of the fsync list.
+ */
+ if (ret != 0) {
+ inode_unref(base_inode);
+ inode_unref(shard_inode);
+ }
+ return ret;
}
gf_boolean_t
-__shard_inode_ctx_needs_lookup (inode_t *inode, xlator_t *this)
+__shard_inode_ctx_needs_lookup(inode_t *inode, xlator_t *this)
{
- int ret = -1;
- shard_inode_ctx_t *ctx = NULL;
+ int ret = -1;
+ shard_inode_ctx_t *ctx = NULL;
- ret = __shard_inode_ctx_get (inode, this, &ctx);
- /* If inode ctx get fails, better to err on the side of caution and
- * try again? Unless the failure is due to mem-allocation.
- */
- if (ret)
- return _gf_true;
+ ret = __shard_inode_ctx_get(inode, this, &ctx);
+ /* If inode ctx get fails, better to err on the side of caution and
+ * try again? Unless the failure is due to mem-allocation.
+ */
+ if (ret)
+ return _gf_true;
- return !ctx->refreshed;
+ return !ctx->refreshed;
}
gf_boolean_t
-shard_inode_ctx_needs_lookup (inode_t *inode, xlator_t *this)
+shard_inode_ctx_needs_lookup(inode_t *inode, xlator_t *this)
{
- gf_boolean_t flag = _gf_false;
+ gf_boolean_t flag = _gf_false;
- LOCK (&inode->lock);
- {
- flag = __shard_inode_ctx_needs_lookup (inode, this);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ flag = __shard_inode_ctx_needs_lookup(inode, this);
+ }
+ UNLOCK(&inode->lock);
- return flag;
+ return flag;
}
int
-__shard_inode_ctx_invalidate (inode_t *inode, xlator_t *this, struct iatt *stbuf)
+__shard_inode_ctx_invalidate(inode_t *inode, xlator_t *this, struct iatt *stbuf)
{
- int ret = -1;
- shard_inode_ctx_t *ctx = NULL;
+ int ret = -1;
+ shard_inode_ctx_t *ctx = NULL;
- ret = __shard_inode_ctx_get (inode, this, &ctx);
- if (ret)
- return ret;
+ ret = __shard_inode_ctx_get(inode, this, &ctx);
+ if (ret)
+ return ret;
- if ((stbuf->ia_size != ctx->stat.ia_size) ||
- (stbuf->ia_blocks != ctx->stat.ia_blocks))
- ctx->refresh = _gf_true;
+ if ((stbuf->ia_size != ctx->stat.ia_size) ||
+ (stbuf->ia_blocks != ctx->stat.ia_blocks))
+ ctx->refresh = _gf_true;
- return 0;
+ return 0;
}
int
-shard_inode_ctx_invalidate (inode_t *inode, xlator_t *this, struct iatt *stbuf)
+shard_inode_ctx_invalidate(inode_t *inode, xlator_t *this, struct iatt *stbuf)
{
- int ret = -1;
+ int ret = -1;
- LOCK (&inode->lock);
- {
- ret = __shard_inode_ctx_invalidate (inode, this, stbuf);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __shard_inode_ctx_invalidate(inode, this, stbuf);
+ }
+ UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
int
-__shard_inode_ctx_get_block_size (inode_t *inode, xlator_t *this,
- uint64_t *block_size)
+__shard_inode_ctx_get_block_size(inode_t *inode, xlator_t *this,
+ uint64_t *block_size)
{
- int ret = -1;
- uint64_t ctx_uint = 0;
- shard_inode_ctx_t *ctx = NULL;
+ int ret = -1;
+ uint64_t ctx_uint = 0;
+ shard_inode_ctx_t *ctx = NULL;
- ret = __inode_ctx_get (inode, this, &ctx_uint);
- if (ret < 0)
- return ret;
+ ret = __inode_ctx_get(inode, this, &ctx_uint);
+ if (ret < 0)
+ return ret;
- ctx = (shard_inode_ctx_t *) ctx_uint;
+ ctx = (shard_inode_ctx_t *)(uintptr_t)ctx_uint;
- *block_size = ctx->block_size;
+ *block_size = ctx->block_size;
- return 0;
+ return 0;
}
int
-shard_inode_ctx_get_block_size (inode_t *inode, xlator_t *this,
- uint64_t *block_size)
+shard_inode_ctx_get_block_size(inode_t *inode, xlator_t *this,
+ uint64_t *block_size)
{
- int ret = -1;
+ int ret = -1;
- LOCK (&inode->lock);
- {
- ret = __shard_inode_ctx_get_block_size (inode, this,
- block_size);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __shard_inode_ctx_get_block_size(inode, this, block_size);
+ }
+ UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
int
-__shard_inode_ctx_get_fsync_count (inode_t *inode, xlator_t *this,
- int *fsync_count)
+__shard_inode_ctx_get_fsync_count(inode_t *inode, xlator_t *this,
+ int *fsync_count)
{
- int ret = -1;
- uint64_t ctx_uint = 0;
- shard_inode_ctx_t *ctx = NULL;
+ int ret = -1;
+ uint64_t ctx_uint = 0;
+ shard_inode_ctx_t *ctx = NULL;
- ret = __inode_ctx_get (inode, this, &ctx_uint);
- if (ret < 0)
- return ret;
+ ret = __inode_ctx_get(inode, this, &ctx_uint);
+ if (ret < 0)
+ return ret;
- ctx = (shard_inode_ctx_t *) ctx_uint;
+ ctx = (shard_inode_ctx_t *)(uintptr_t)ctx_uint;
- *fsync_count = ctx->fsync_needed;
+ *fsync_count = ctx->fsync_needed;
- return 0;
+ return 0;
}
int
-shard_inode_ctx_get_fsync_count (inode_t *inode, xlator_t *this,
+shard_inode_ctx_get_fsync_count(inode_t *inode, xlator_t *this,
int *fsync_count)
{
- int ret = -1;
+ int ret = -1;
- LOCK (&inode->lock);
- {
- ret = __shard_inode_ctx_get_fsync_count (inode, this,
- fsync_count);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __shard_inode_ctx_get_fsync_count(inode, this, fsync_count);
+ }
+ UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
int
-__shard_inode_ctx_get_all (inode_t *inode, xlator_t *this,
- shard_inode_ctx_t *ctx_out)
+__shard_inode_ctx_get_all(inode_t *inode, xlator_t *this,
+ shard_inode_ctx_t *ctx_out)
{
- int ret = -1;
- uint64_t ctx_uint = 0;
- shard_inode_ctx_t *ctx = NULL;
+ int ret = -1;
+ uint64_t ctx_uint = 0;
+ shard_inode_ctx_t *ctx = NULL;
- ret = __inode_ctx_get (inode, this, &ctx_uint);
- if (ret < 0)
- return ret;
+ ret = __inode_ctx_get(inode, this, &ctx_uint);
+ if (ret < 0)
+ return ret;
- ctx = (shard_inode_ctx_t *) ctx_uint;
+ ctx = (shard_inode_ctx_t *)(uintptr_t)ctx_uint;
- memcpy (ctx_out, ctx, sizeof (shard_inode_ctx_t));
- return 0;
+ memcpy(ctx_out, ctx, sizeof(shard_inode_ctx_t));
+ return 0;
}
int
-shard_inode_ctx_get_all (inode_t *inode, xlator_t *this,
- shard_inode_ctx_t *ctx_out)
+shard_inode_ctx_get_all(inode_t *inode, xlator_t *this,
+ shard_inode_ctx_t *ctx_out)
{
- int ret = -1;
+ int ret = -1;
- LOCK (&inode->lock);
- {
- ret = __shard_inode_ctx_get_all (inode, this, ctx_out);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __shard_inode_ctx_get_all(inode, this, ctx_out);
+ }
+ UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
int
-__shard_inode_ctx_fill_iatt_from_cache (inode_t *inode, xlator_t *this,
- struct iatt *buf,
- gf_boolean_t *need_refresh)
+__shard_inode_ctx_fill_iatt_from_cache(inode_t *inode, xlator_t *this,
+ struct iatt *buf,
+ gf_boolean_t *need_refresh)
{
- int ret = -1;
- uint64_t ctx_uint = 0;
- shard_inode_ctx_t *ctx = NULL;
+ int ret = -1;
+ uint64_t ctx_uint = 0;
+ shard_inode_ctx_t *ctx = NULL;
- ret = __inode_ctx_get (inode, this, &ctx_uint);
- if (ret < 0)
- return ret;
+ ret = __inode_ctx_get(inode, this, &ctx_uint);
+ if (ret < 0)
+ return ret;
- ctx = (shard_inode_ctx_t *) ctx_uint;
+ ctx = (shard_inode_ctx_t *)(uintptr_t)ctx_uint;
- if (ctx->refresh == _gf_false)
- *buf = ctx->stat;
- else
- *need_refresh = _gf_true;
+ if (ctx->refresh == _gf_false)
+ *buf = ctx->stat;
+ else
+ *need_refresh = _gf_true;
- return 0;
+ return 0;
}
int
-shard_inode_ctx_fill_iatt_from_cache (inode_t *inode, xlator_t *this,
- struct iatt *buf,
- gf_boolean_t *need_refresh)
+shard_inode_ctx_fill_iatt_from_cache(inode_t *inode, xlator_t *this,
+ struct iatt *buf,
+ gf_boolean_t *need_refresh)
{
- int ret = -1;
+ int ret = -1;
- LOCK (&inode->lock);
- {
- ret = __shard_inode_ctx_fill_iatt_from_cache (inode, this, buf,
- need_refresh);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __shard_inode_ctx_fill_iatt_from_cache(inode, this, buf,
+ need_refresh);
+ }
+ UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
void
-shard_local_wipe (shard_local_t *local)
+shard_local_wipe(shard_local_t *local)
{
- int i = 0;
- int count = 0;
+ int i = 0;
+ int count = 0;
- count = local->num_blocks;
+ count = local->num_blocks;
- syncbarrier_destroy (&local->barrier);
- loc_wipe (&local->loc);
- loc_wipe (&local->dot_shard_loc);
- loc_wipe (&local->loc2);
- loc_wipe (&local->tmp_loc);
+ syncbarrier_destroy(&local->barrier);
+ loc_wipe(&local->loc);
+ loc_wipe(&local->dot_shard_loc);
+ loc_wipe(&local->dot_shard_rm_loc);
+ loc_wipe(&local->loc2);
+ loc_wipe(&local->tmp_loc);
+ loc_wipe(&local->int_inodelk.loc);
+ loc_wipe(&local->int_entrylk.loc);
+ loc_wipe(&local->newloc);
- if (local->fd)
- fd_unref (local->fd);
+ if (local->name)
+ GF_FREE(local->name);
- if (local->xattr_req)
- dict_unref (local->xattr_req);
- if (local->xattr_rsp)
- dict_unref (local->xattr_rsp);
+ if (local->int_entrylk.basename)
+ GF_FREE(local->int_entrylk.basename);
+ if (local->fd)
+ fd_unref(local->fd);
- for (i = 0; i < count; i++) {
- if (!local->inode_list)
- break;
+ if (local->xattr_req)
+ dict_unref(local->xattr_req);
+ if (local->xattr_rsp)
+ dict_unref(local->xattr_rsp);
- if (local->inode_list[i])
- inode_unref (local->inode_list[i]);
- }
+ for (i = 0; i < count; i++) {
+ if (!local->inode_list)
+ break;
+
+ if (local->inode_list[i])
+ inode_unref(local->inode_list[i]);
+ }
- GF_FREE (local->inode_list);
+ GF_FREE(local->inode_list);
- GF_FREE (local->vector);
- if (local->iobref)
- iobref_unref (local->iobref);
- if (local->list_inited)
- gf_dirent_free (&local->entries_head);
+ GF_FREE(local->vector);
+ if (local->iobref)
+ iobref_unref(local->iobref);
+ if (local->list_inited)
+ gf_dirent_free(&local->entries_head);
+ if (local->inodelk_frame)
+ SHARD_STACK_DESTROY(local->inodelk_frame);
+ if (local->entrylk_frame)
+ SHARD_STACK_DESTROY(local->entrylk_frame);
}
int
-shard_modify_size_and_block_count (struct iatt *stbuf, dict_t *dict)
+shard_modify_size_and_block_count(struct iatt *stbuf, dict_t *dict)
{
- int ret = -1;
- void *size_attr = NULL;
- uint64_t size_array[4];
+ int ret = -1;
+ void *size_attr = NULL;
+ uint64_t size_array[4];
- ret = dict_get_ptr (dict, GF_XATTR_SHARD_FILE_SIZE, &size_attr);
- if (ret) {
- gf_msg_callingfn (THIS->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INTERNAL_XATTR_MISSING, "Failed to "
- "get "GF_XATTR_SHARD_FILE_SIZE" for %s",
- uuid_utoa (stbuf->ia_gfid));
- return ret;
- }
+ ret = dict_get_ptr(dict, GF_XATTR_SHARD_FILE_SIZE, &size_attr);
+ if (ret) {
+ gf_msg_callingfn(THIS->name, GF_LOG_ERROR, 0,
+ SHARD_MSG_INTERNAL_XATTR_MISSING,
+ "Failed to "
+ "get " GF_XATTR_SHARD_FILE_SIZE " for %s",
+ uuid_utoa(stbuf->ia_gfid));
+ return ret;
+ }
- memcpy (size_array, size_attr, sizeof (size_array));
+ memcpy(size_array, size_attr, sizeof(size_array));
- stbuf->ia_size = ntoh64 (size_array[0]);
- stbuf->ia_blocks = ntoh64 (size_array[2]);
+ stbuf->ia_size = ntoh64(size_array[0]);
+ stbuf->ia_blocks = ntoh64(size_array[2]);
- return 0;
+ return 0;
}
int
-shard_call_count_return (call_frame_t *frame)
+shard_call_count_return(call_frame_t *frame)
{
- int call_count = 0;
- shard_local_t *local = NULL;
+ int call_count = 0;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- LOCK (&frame->lock);
- {
- call_count = --local->call_count;
- }
- UNLOCK (&frame->lock);
+ LOCK(&frame->lock);
+ {
+ call_count = --local->call_count;
+ }
+ UNLOCK(&frame->lock);
- return call_count;
+ return call_count;
}
static char *
-shard_internal_dir_string (shard_internal_dir_type_t type)
+shard_internal_dir_string(shard_internal_dir_type_t type)
{
- char *str = NULL;
+ char *str = NULL;
- switch (type) {
+ switch (type) {
case SHARD_INTERNAL_DIR_DOT_SHARD:
- str = ".shard";
- break;
+ str = GF_SHARD_DIR;
+ break;
+ case SHARD_INTERNAL_DIR_DOT_SHARD_REMOVE_ME:
+ str = GF_SHARD_REMOVE_ME_DIR;
+ break;
default:
- break;
- }
- return str;
+ break;
+ }
+ return str;
}
static int
-shard_init_internal_dir_loc (xlator_t *this, shard_local_t *local,
- shard_internal_dir_type_t type)
+shard_init_internal_dir_loc(xlator_t *this, shard_local_t *local,
+ shard_internal_dir_type_t type)
{
- int ret = -1;
- char *bname = NULL;
- loc_t *internal_dir_loc = NULL;
+ int ret = -1;
+ char *bname = NULL;
+ inode_t *parent = NULL;
+ loc_t *internal_dir_loc = NULL;
+ shard_priv_t *priv = NULL;
- if (!local)
- return -1;
+ priv = this->private;
+ if (!local)
+ return -1;
- switch (type) {
+ switch (type) {
case SHARD_INTERNAL_DIR_DOT_SHARD:
- internal_dir_loc = &local->dot_shard_loc;
- bname = GF_SHARD_DIR;
- break;
+ internal_dir_loc = &local->dot_shard_loc;
+ bname = GF_SHARD_DIR;
+ parent = inode_ref(this->itable->root);
+ break;
+ case SHARD_INTERNAL_DIR_DOT_SHARD_REMOVE_ME:
+ internal_dir_loc = &local->dot_shard_rm_loc;
+ bname = GF_SHARD_REMOVE_ME_DIR;
+ parent = inode_ref(priv->dot_shard_inode);
+ break;
default:
- break;
- }
-
- internal_dir_loc->inode = inode_new (this->itable);
- internal_dir_loc->parent = inode_ref (this->itable->root);
- ret = inode_path (internal_dir_loc->parent, bname,
- (char **)&internal_dir_loc->path);
- if (ret < 0 || !(internal_dir_loc->inode)) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_PATH_FAILED,
- "Inode path failed on %s", bname);
- goto out;
- }
-
- internal_dir_loc->name = strrchr (internal_dir_loc->path, '/');
- if (internal_dir_loc->name)
- internal_dir_loc->name++;
-
- ret = 0;
+ break;
+ }
+
+ internal_dir_loc->inode = inode_new(this->itable);
+ internal_dir_loc->parent = parent;
+ ret = inode_path(internal_dir_loc->parent, bname,
+ (char **)&internal_dir_loc->path);
+ if (ret < 0 || !(internal_dir_loc->inode)) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_PATH_FAILED,
+ "Inode path failed on %s", bname);
+ goto out;
+ }
+
+ internal_dir_loc->name = strrchr(internal_dir_loc->path, '/');
+ if (internal_dir_loc->name)
+ internal_dir_loc->name++;
+
+ ret = 0;
out:
- return ret;
+ return ret;
}
inode_t *
-__shard_update_shards_inode_list (inode_t *linked_inode, xlator_t *this,
- inode_t *base_inode, int block_num)
-{
- char block_bname[256] = {0,};
- inode_t *lru_inode = NULL;
- shard_priv_t *priv = NULL;
- shard_inode_ctx_t *ctx = NULL;
- shard_inode_ctx_t *lru_inode_ctx = NULL;
- shard_inode_ctx_t *lru_base_inode_ctx = NULL;
- inode_t *fsync_inode = NULL;
- inode_t *lru_base_inode = NULL;
- gf_boolean_t do_fsync = _gf_false;
-
- priv = this->private;
-
- shard_inode_ctx_get (linked_inode, this, &ctx);
-
- if (list_empty (&ctx->ilist)) {
- if (priv->inode_count + 1 <= SHARD_MAX_INODES) {
- /* If this inode was linked here for the first time (indicated
- * by empty list), and if there is still space in the priv list,
- * add this ctx to the tail of the list.
+__shard_update_shards_inode_list(inode_t *linked_inode, xlator_t *this,
+ inode_t *base_inode, int block_num,
+ uuid_t gfid)
+{
+ char block_bname[256] = {
+ 0,
+ };
+ inode_t *lru_inode = NULL;
+ shard_priv_t *priv = NULL;
+ shard_inode_ctx_t *ctx = NULL;
+ shard_inode_ctx_t *lru_inode_ctx = NULL;
+ shard_inode_ctx_t *lru_base_inode_ctx = NULL;
+ inode_t *fsync_inode = NULL;
+ inode_t *lru_base_inode = NULL;
+ gf_boolean_t do_fsync = _gf_false;
+
+ priv = this->private;
+
+ shard_inode_ctx_get(linked_inode, this, &ctx);
+
+ if (list_empty(&ctx->ilist)) {
+ if (priv->inode_count + 1 <= priv->lru_limit) {
+ /* If this inode was linked here for the first time (indicated
+ * by empty list), and if there is still space in the priv list,
+ * add this ctx to the tail of the list.
+ */
+ /* For as long as an inode is in lru list, we try to
+ * keep it alive by holding a ref on it.
+ */
+ inode_ref(linked_inode);
+ if (base_inode)
+ gf_uuid_copy(ctx->base_gfid, base_inode->gfid);
+ else
+ gf_uuid_copy(ctx->base_gfid, gfid);
+ ctx->block_num = block_num;
+ list_add_tail(&ctx->ilist, &priv->ilist_head);
+ priv->inode_count++;
+ ctx->base_inode = inode_ref(base_inode);
+ } else {
+ /*If on the other hand there is no available slot for this inode
+ * in the list, delete the lru inode from the head of the list,
+ * unlink it. And in its place add this new inode into the list.
+ */
+ lru_inode_ctx = list_first_entry(&priv->ilist_head,
+ shard_inode_ctx_t, ilist);
+ GF_ASSERT(lru_inode_ctx->block_num > 0);
+ lru_base_inode = lru_inode_ctx->base_inode;
+ list_del_init(&lru_inode_ctx->ilist);
+ lru_inode = inode_find(linked_inode->table,
+ lru_inode_ctx->stat.ia_gfid);
+ /* If the lru inode was part of the pending-fsync list,
+ * the base inode needs to be unref'd, the lru inode
+ * deleted from fsync list and fsync'd in a new frame,
+ * and then unlinked in memory and forgotten.
+ */
+ if (!lru_base_inode)
+ goto after_fsync_check;
+ LOCK(&lru_base_inode->lock);
+ LOCK(&lru_inode->lock);
+ {
+ if (!list_empty(&lru_inode_ctx->to_fsync_list)) {
+ list_del_init(&lru_inode_ctx->to_fsync_list);
+ lru_inode_ctx->fsync_needed = 0;
+ do_fsync = _gf_true;
+ __shard_inode_ctx_get(lru_base_inode, this,
+ &lru_base_inode_ctx);
+ lru_base_inode_ctx->fsync_count--;
+ }
+ }
+ UNLOCK(&lru_inode->lock);
+ UNLOCK(&lru_base_inode->lock);
+
+ after_fsync_check:
+ if (!do_fsync) {
+ shard_make_block_bname(lru_inode_ctx->block_num,
+ lru_inode_ctx->base_gfid, block_bname,
+ sizeof(block_bname));
+ /* The following unref corresponds to the ref held at
+ * the time the shard was added to the lru list.
*/
- /* For as long as an inode is in lru list, we try to
- * keep it alive by holding a ref on it.
- */
- inode_ref (linked_inode);
- gf_uuid_copy (ctx->base_gfid, base_inode->gfid);
- ctx->block_num = block_num;
- list_add_tail (&ctx->ilist, &priv->ilist_head);
- priv->inode_count++;
- ctx->base_inode = base_inode;
- } else {
- /*If on the other hand there is no available slot for this inode
- * in the list, delete the lru inode from the head of the list,
- * unlink it. And in its place add this new inode into the list.
+ inode_unref(lru_inode);
+ inode_unlink(lru_inode, priv->dot_shard_inode, block_bname);
+ inode_forget(lru_inode, 0);
+ } else {
+ /* The following unref corresponds to the ref
+ * held when the shard was added to fsync list.
*/
- lru_inode_ctx = list_first_entry (&priv->ilist_head,
- shard_inode_ctx_t,
- ilist);
- GF_ASSERT (lru_inode_ctx->block_num > 0);
- lru_base_inode = lru_inode_ctx->base_inode;
- list_del_init (&lru_inode_ctx->ilist);
- lru_inode = inode_find (linked_inode->table,
- lru_inode_ctx->stat.ia_gfid);
- /* If the lru inode was part of the pending-fsync list,
- * the base inode needs to be unref'd, the lru inode
- * deleted from fsync list and fsync'd in a new frame,
- * and then unlinked in memory and forgotten.
- */
- LOCK (&lru_base_inode->lock);
- LOCK (&lru_inode->lock);
- {
- if (!list_empty(&lru_inode_ctx->to_fsync_list)) {
- list_del_init (&lru_inode_ctx->to_fsync_list);
- lru_inode_ctx->fsync_needed = 0;
- do_fsync = _gf_true;
- __shard_inode_ctx_get (lru_base_inode, this, &lru_base_inode_ctx);
- lru_base_inode_ctx->fsync_count--;
- }
- }
- UNLOCK (&lru_inode->lock);
- UNLOCK (&lru_base_inode->lock);
-
- if (!do_fsync) {
- shard_make_block_bname (lru_inode_ctx->block_num,
- lru_inode_ctx->base_gfid,
- block_bname,
- sizeof (block_bname));
- /* The following unref corresponds to the ref held at
- * the time the shard was added to the lru list.
- */
- inode_unref (lru_inode);
- inode_unlink (lru_inode, priv->dot_shard_inode,
- block_bname);
- inode_forget (lru_inode, 0);
- } else {
- fsync_inode = lru_inode;
- inode_unref (lru_base_inode);
- }
- /* The following unref corresponds to the ref
- * held by inode_find() above.
- */
- inode_unref (lru_inode);
- /* For as long as an inode is in lru list, we try to
- * keep it alive by holding a ref on it.
- */
- inode_ref (linked_inode);
- gf_uuid_copy (ctx->base_gfid, base_inode->gfid);
- ctx->block_num = block_num;
- ctx->base_inode = base_inode;
- list_add_tail (&ctx->ilist, &priv->ilist_head);
- }
- } else {
- /* If this is not the first time this inode is being operated on, move
+ inode_unref(lru_inode);
+ fsync_inode = lru_inode;
+ if (lru_base_inode)
+ inode_unref(lru_base_inode);
+ }
+ /* The following unref corresponds to the ref
+ * held by inode_find() above.
+ */
+ inode_unref(lru_inode);
+
+ /* The following unref corresponds to the ref held on the base shard
+ * at the time of adding shard inode to lru list
+ */
+ if (lru_base_inode)
+ inode_unref(lru_base_inode);
+
+ /* For as long as an inode is in lru list, we try to
+ * keep it alive by holding a ref on it.
+ */
+ inode_ref(linked_inode);
+ if (base_inode)
+ gf_uuid_copy(ctx->base_gfid, base_inode->gfid);
+ else
+ gf_uuid_copy(ctx->base_gfid, gfid);
+ ctx->block_num = block_num;
+ ctx->base_inode = inode_ref(base_inode);
+ list_add_tail(&ctx->ilist, &priv->ilist_head);
+ }
+ } else {
+ /* If this is not the first time this inode is being operated on, move
* it to the most recently used end of the list.
*/
- list_move_tail (&ctx->ilist, &priv->ilist_head);
- }
- return fsync_inode;
-}
-
-int
-shard_common_inode_write_failure_unwind (glusterfs_fop_t fop,
- call_frame_t *frame, int32_t op_ret,
- int32_t op_errno)
-{
- switch (fop) {
+ list_move_tail(&ctx->ilist, &priv->ilist_head);
+ }
+ return fsync_inode;
+}
+
+int
+shard_common_failure_unwind(glusterfs_fop_t fop, call_frame_t *frame,
+ int32_t op_ret, int32_t op_errno)
+{
+ switch (fop) {
+ case GF_FOP_LOOKUP:
+ SHARD_STACK_UNWIND(lookup, frame, op_ret, op_errno, NULL, NULL,
+ NULL, NULL);
+ break;
+ case GF_FOP_STAT:
+ SHARD_STACK_UNWIND(stat, frame, op_ret, op_errno, NULL, NULL);
+ break;
+ case GF_FOP_FSTAT:
+ SHARD_STACK_UNWIND(fstat, frame, op_ret, op_errno, NULL, NULL);
+ break;
+ case GF_FOP_TRUNCATE:
+ SHARD_STACK_UNWIND(truncate, frame, op_ret, op_errno, NULL, NULL,
+ NULL);
+ break;
+ case GF_FOP_FTRUNCATE:
+ SHARD_STACK_UNWIND(ftruncate, frame, op_ret, op_errno, NULL, NULL,
+ NULL);
+ break;
+ case GF_FOP_MKNOD:
+ SHARD_STACK_UNWIND(mknod, frame, op_ret, op_errno, NULL, NULL, NULL,
+ NULL, NULL);
+ break;
+ case GF_FOP_LINK:
+ SHARD_STACK_UNWIND(link, frame, op_ret, op_errno, NULL, NULL, NULL,
+ NULL, NULL);
+ break;
+ case GF_FOP_CREATE:
+ SHARD_STACK_UNWIND(create, frame, op_ret, op_errno, NULL, NULL,
+ NULL, NULL, NULL, NULL);
+ break;
+ case GF_FOP_UNLINK:
+ SHARD_STACK_UNWIND(unlink, frame, op_ret, op_errno, NULL, NULL,
+ NULL);
+ break;
+ case GF_FOP_RENAME:
+ SHARD_STACK_UNWIND(rename, frame, op_ret, op_errno, NULL, NULL,
+ NULL, NULL, NULL, NULL);
+ break;
case GF_FOP_WRITE:
- SHARD_STACK_UNWIND (writev, frame, op_ret, op_errno,
- NULL, NULL, NULL);
- break;
+ SHARD_STACK_UNWIND(writev, frame, op_ret, op_errno, NULL, NULL,
+ NULL);
+ break;
case GF_FOP_FALLOCATE:
- SHARD_STACK_UNWIND (fallocate, frame, op_ret, op_errno,
- NULL, NULL, NULL);
- break;
+ SHARD_STACK_UNWIND(fallocate, frame, op_ret, op_errno, NULL, NULL,
+ NULL);
+ break;
case GF_FOP_ZEROFILL:
- SHARD_STACK_UNWIND (zerofill, frame, op_ret, op_errno,
- NULL, NULL, NULL);
- break;
+ SHARD_STACK_UNWIND(zerofill, frame, op_ret, op_errno, NULL, NULL,
+ NULL);
+ break;
case GF_FOP_DISCARD:
- SHARD_STACK_UNWIND (discard, frame, op_ret, op_errno,
- NULL, NULL, NULL);
- break;
+ SHARD_STACK_UNWIND(discard, frame, op_ret, op_errno, NULL, NULL,
+ NULL);
+ break;
+ case GF_FOP_READ:
+ SHARD_STACK_UNWIND(readv, frame, op_ret, op_errno, NULL, -1, NULL,
+ NULL, NULL);
+ break;
+ case GF_FOP_FSYNC:
+ SHARD_STACK_UNWIND(fsync, frame, op_ret, op_errno, NULL, NULL,
+ NULL);
+ break;
+ case GF_FOP_REMOVEXATTR:
+ SHARD_STACK_UNWIND(removexattr, frame, op_ret, op_errno, NULL);
+ break;
+ case GF_FOP_FREMOVEXATTR:
+ SHARD_STACK_UNWIND(fremovexattr, frame, op_ret, op_errno, NULL);
+ break;
+ case GF_FOP_FGETXATTR:
+ SHARD_STACK_UNWIND(fgetxattr, frame, op_ret, op_errno, NULL, NULL);
+ break;
+ case GF_FOP_GETXATTR:
+ SHARD_STACK_UNWIND(getxattr, frame, op_ret, op_errno, NULL, NULL);
+ break;
+ case GF_FOP_FSETXATTR:
+ SHARD_STACK_UNWIND(fsetxattr, frame, op_ret, op_errno, NULL);
+ break;
+ case GF_FOP_SETXATTR:
+ SHARD_STACK_UNWIND(setxattr, frame, op_ret, op_errno, NULL);
+ break;
+ case GF_FOP_SETATTR:
+ SHARD_STACK_UNWIND(setattr, frame, op_ret, op_errno, NULL, NULL,
+ NULL);
+ break;
+ case GF_FOP_FSETATTR:
+ SHARD_STACK_UNWIND(fsetattr, frame, op_ret, op_errno, NULL, NULL,
+ NULL);
+ break;
+ case GF_FOP_SEEK:
+ SHARD_STACK_UNWIND(seek, frame, op_ret, op_errno, 0, NULL);
+ break;
default:
- gf_msg (THIS->name, GF_LOG_WARNING, 0, SHARD_MSG_INVALID_FOP,
- "Invalid fop id = %d", fop);
- break;
- }
- return 0;
+ gf_msg(THIS->name, GF_LOG_WARNING, 0, SHARD_MSG_INVALID_FOP,
+ "Invalid fop id = %d", fop);
+ break;
+ }
+ return 0;
}
int
-shard_common_inode_write_success_unwind (glusterfs_fop_t fop,
- call_frame_t *frame, int32_t op_ret)
+shard_common_inode_write_success_unwind(glusterfs_fop_t fop,
+ call_frame_t *frame, int32_t op_ret)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = frame->local;
- local = frame->local;
+ /* the below 3 variables are required because, in SHARD_STACK_UNWIND()
+ macro, there is a check for local being null. So many static analyzers
+ backtrace the code with assumption of possible (local == NULL) case,
+ and complains for below lines. By handling it like below, we overcome
+ the warnings */
- switch (fop) {
+ struct iatt *prebuf = ((local) ? &local->prebuf : NULL);
+ struct iatt *postbuf = ((local) ? &local->postbuf : NULL);
+ dict_t *xattr_rsp = ((local) ? local->xattr_rsp : NULL);
+
+ switch (fop) {
case GF_FOP_WRITE:
- SHARD_STACK_UNWIND (writev, frame, op_ret, 0, &local->prebuf,
- &local->postbuf, local->xattr_rsp);
- break;
+ SHARD_STACK_UNWIND(writev, frame, op_ret, 0, prebuf, postbuf,
+ xattr_rsp);
+ break;
case GF_FOP_FALLOCATE:
- SHARD_STACK_UNWIND (fallocate, frame, op_ret, 0, &local->prebuf,
- &local->postbuf, local->xattr_rsp);
- break;
+ SHARD_STACK_UNWIND(fallocate, frame, op_ret, 0, prebuf, postbuf,
+ xattr_rsp);
+ break;
case GF_FOP_ZEROFILL:
- SHARD_STACK_UNWIND (zerofill, frame, op_ret, 0, &local->prebuf,
- &local->postbuf, local->xattr_rsp);
- break;
+ SHARD_STACK_UNWIND(zerofill, frame, op_ret, 0, prebuf, postbuf,
+ xattr_rsp);
+ break;
case GF_FOP_DISCARD:
- SHARD_STACK_UNWIND (discard, frame, op_ret, 0, &local->prebuf,
- &local->postbuf, local->xattr_rsp);
- break;
+ SHARD_STACK_UNWIND(discard, frame, op_ret, 0, prebuf, postbuf,
+ xattr_rsp);
+ break;
default:
- gf_msg (THIS->name, GF_LOG_WARNING, 0, SHARD_MSG_INVALID_FOP,
- "Invalid fop id = %d", fop);
- break;
- }
- return 0;
-}
-
-int
-shard_evicted_inode_fsync_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct iatt *prebuf, struct iatt *postbuf,
- dict_t *xdata)
-{
- char block_bname[256] = {0,};
- fd_t *anon_fd = cookie;
- inode_t *shard_inode = NULL;
- shard_inode_ctx_t *ctx = NULL;
- shard_priv_t *priv = NULL;
-
- priv = this->private;
- shard_inode = anon_fd->inode;
-
- if (op_ret < 0) {
- gf_msg (this->name, GF_LOG_WARNING, op_errno,
- SHARD_MSG_MEMALLOC_FAILED, "fsync failed on shard");
- goto out;
- }
-
- LOCK (&priv->lock);
- LOCK(&shard_inode->lock);
- {
- __shard_inode_ctx_get (shard_inode, this, &ctx);
- if ((list_empty(&ctx->to_fsync_list)) &&
- (list_empty(&ctx->ilist))) {
- shard_make_block_bname (ctx->block_num,
- shard_inode->gfid, block_bname,
- sizeof (block_bname));
- inode_unlink (shard_inode, priv->dot_shard_inode,
- block_bname);
- /* The following unref corresponds to the ref held by
- * inode_link() at the time the shard was created or
- * looked up
- */
- inode_unref (shard_inode);
- inode_forget (shard_inode, 0);
- }
- }
- UNLOCK(&shard_inode->lock);
- UNLOCK(&priv->lock);
+ gf_msg(THIS->name, GF_LOG_WARNING, 0, SHARD_MSG_INVALID_FOP,
+ "Invalid fop id = %d", fop);
+ break;
+ }
+ return 0;
+}
+
+int
+shard_evicted_inode_fsync_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata)
+{
+ char block_bname[256] = {
+ 0,
+ };
+ fd_t *anon_fd = cookie;
+ inode_t *shard_inode = NULL;
+ shard_inode_ctx_t *ctx = NULL;
+ shard_priv_t *priv = NULL;
+
+ priv = this->private;
+
+ if (anon_fd == NULL || op_ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, op_errno, SHARD_MSG_MEMALLOC_FAILED,
+ "fsync failed on shard");
+ goto out;
+ }
+ shard_inode = anon_fd->inode;
+
+ LOCK(&priv->lock);
+ LOCK(&shard_inode->lock);
+ {
+ __shard_inode_ctx_get(shard_inode, this, &ctx);
+ if ((list_empty(&ctx->to_fsync_list)) && (list_empty(&ctx->ilist))) {
+ shard_make_block_bname(ctx->block_num, shard_inode->gfid,
+ block_bname, sizeof(block_bname));
+ inode_unlink(shard_inode, priv->dot_shard_inode, block_bname);
+ /* The following unref corresponds to the ref held by
+ * inode_link() at the time the shard was created or
+ * looked up
+ */
+ inode_unref(shard_inode);
+ inode_forget(shard_inode, 0);
+ }
+ }
+ UNLOCK(&shard_inode->lock);
+ UNLOCK(&priv->lock);
out:
- if (anon_fd)
- fd_unref (anon_fd);
- STACK_DESTROY (frame->root);
- return 0;
-}
-
-int
-shard_initiate_evicted_inode_fsync (xlator_t *this, inode_t *inode)
-{
- fd_t *anon_fd = NULL;
- call_frame_t *fsync_frame = NULL;
-
- fsync_frame = create_frame (this, this->ctx->pool);
- if (!fsync_frame) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- SHARD_MSG_MEMALLOC_FAILED, "Failed to create new frame "
- "to fsync shard");
- return -1;
- }
-
- anon_fd = fd_anonymous (inode);
- if (!anon_fd) {
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- SHARD_MSG_MEMALLOC_FAILED, "Failed to create anon fd to"
- " fsync shard");
- STACK_DESTROY (fsync_frame->root);
- return -1;
- }
-
- STACK_WIND_COOKIE (fsync_frame, shard_evicted_inode_fsync_cbk, anon_fd,
- FIRST_CHILD(this), FIRST_CHILD(this)->fops->fsync,
- anon_fd, 1, NULL);
- return 0;
-}
-
-int
-shard_common_resolve_shards (call_frame_t *frame, xlator_t *this,
- shard_post_resolve_fop_handler_t post_res_handler)
-{
- int i = -1;
- uint32_t shard_idx_iter = 0;
- char path[PATH_MAX] = {0,};
- inode_t *inode = NULL;
- inode_t *res_inode = NULL;
- inode_t *fsync_inode = NULL;
- shard_priv_t *priv = NULL;
- shard_local_t *local = NULL;
-
- priv = this->private;
- local = frame->local;
- local->call_count = 0;
- shard_idx_iter = local->first_block;
- res_inode = local->resolver_base_inode;
-
- if (local->op_ret < 0)
- goto out;
-
- while (shard_idx_iter <= local->last_block) {
- i++;
- if (shard_idx_iter == 0) {
- local->inode_list[i] = inode_ref (res_inode);
- shard_idx_iter++;
- continue;
- }
-
- shard_make_block_abspath (shard_idx_iter, res_inode->gfid, path,
- sizeof(path));
-
- inode = NULL;
- inode = inode_resolve (this->itable, path);
- if (inode) {
- gf_msg_debug (this->name, 0, "Shard %d already "
- "present. gfid=%s. Saving inode for future.",
- shard_idx_iter, uuid_utoa(inode->gfid));
- local->inode_list[i] = inode;
- /* Let the ref on the inodes that are already present
- * in inode table still be held so that they don't get
- * forgotten by the time the fop reaches the actual
- * write stage.
- */
- LOCK(&priv->lock);
- {
- fsync_inode = __shard_update_shards_inode_list (inode,
- this,
- res_inode,
- shard_idx_iter);
- }
- UNLOCK(&priv->lock);
- shard_idx_iter++;
- if (fsync_inode)
- shard_initiate_evicted_inode_fsync (this,
- fsync_inode);
- continue;
- } else {
- local->call_count++;
- shard_idx_iter++;
- }
+ if (anon_fd)
+ fd_unref(anon_fd);
+ STACK_DESTROY(frame->root);
+ return 0;
+}
+
+int
+shard_initiate_evicted_inode_fsync(xlator_t *this, inode_t *inode)
+{
+ fd_t *anon_fd = NULL;
+ call_frame_t *fsync_frame = NULL;
+
+ fsync_frame = create_frame(this, this->ctx->pool);
+ if (!fsync_frame) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, SHARD_MSG_MEMALLOC_FAILED,
+ "Failed to create new frame "
+ "to fsync shard");
+ return -1;
+ }
+
+ anon_fd = fd_anonymous(inode);
+ if (!anon_fd) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, SHARD_MSG_MEMALLOC_FAILED,
+ "Failed to create anon fd to"
+ " fsync shard");
+ STACK_DESTROY(fsync_frame->root);
+ return -1;
+ }
+
+ STACK_WIND_COOKIE(fsync_frame, shard_evicted_inode_fsync_cbk, anon_fd,
+ FIRST_CHILD(this), FIRST_CHILD(this)->fops->fsync,
+ anon_fd, 1, NULL);
+ return 0;
+}
+
+int
+shard_common_inode_write_post_lookup_shards_handler(call_frame_t *frame,
+ xlator_t *this);
+
+int
+shard_common_resolve_shards(call_frame_t *frame, xlator_t *this,
+ shard_post_resolve_fop_handler_t post_res_handler)
+{
+ int i = -1;
+ uint32_t shard_idx_iter = 0;
+ char path[PATH_MAX] = {
+ 0,
+ };
+ uuid_t gfid = {
+ 0,
+ };
+ inode_t *inode = NULL;
+ inode_t *res_inode = NULL;
+ inode_t *fsync_inode = NULL;
+ shard_priv_t *priv = NULL;
+ shard_local_t *local = NULL;
+ uint64_t resolve_count = 0;
+
+ priv = this->private;
+ local = frame->local;
+ local->call_count = 0;
+ shard_idx_iter = local->first_block;
+ res_inode = local->resolver_base_inode;
+
+ if ((local->op_ret < 0) || (local->resolve_not))
+ goto out;
+
+ /* If this prealloc FOP is for fresh file creation, then the size of the
+ * file will be 0. Then there will be no shards associated with this file.
+ * So we can skip the lookup process for the shards which do not exists
+ * and directly issue mknod to crete shards.
+ *
+ * In case the prealloc fop is to extend the preallocated file to bigger
+ * size then just lookup and populate inodes of existing shards and
+ * update the create count
+ */
+ if (local->fop == GF_FOP_FALLOCATE) {
+ if (!local->prebuf.ia_size) {
+ local->inode_list[0] = inode_ref(res_inode);
+ local->create_count = local->last_block;
+ shard_common_inode_write_post_lookup_shards_handler(frame, this);
+ return 0;
+ }
+ if (local->prebuf.ia_size < local->total_size)
+ local->create_count = local->last_block -
+ ((local->prebuf.ia_size - 1) /
+ local->block_size);
+ }
+
+ resolve_count = local->last_block - local->create_count;
+
+ if (res_inode)
+ gf_uuid_copy(gfid, res_inode->gfid);
+ else
+ gf_uuid_copy(gfid, local->base_gfid);
+
+ while (shard_idx_iter <= resolve_count) {
+ i++;
+ if (shard_idx_iter == 0) {
+ local->inode_list[i] = inode_ref(res_inode);
+ shard_idx_iter++;
+ continue;
+ }
+
+ shard_make_block_abspath(shard_idx_iter, gfid, path, sizeof(path));
+
+ inode = NULL;
+ inode = inode_resolve(this->itable, path);
+ if (inode) {
+ gf_msg_debug(this->name, 0,
+ "Shard %d already "
+ "present. gfid=%s. Saving inode for future.",
+ shard_idx_iter, uuid_utoa(inode->gfid));
+ local->inode_list[i] = inode;
+ /* Let the ref on the inodes that are already present
+ * in inode table still be held so that they don't get
+ * forgotten by the time the fop reaches the actual
+ * write stage.
+ */
+ LOCK(&priv->lock);
+ {
+ fsync_inode = __shard_update_shards_inode_list(
+ inode, this, res_inode, shard_idx_iter, gfid);
+ }
+ UNLOCK(&priv->lock);
+ shard_idx_iter++;
+ if (fsync_inode)
+ shard_initiate_evicted_inode_fsync(this, fsync_inode);
+ continue;
+ } else {
+ local->call_count++;
+ shard_idx_iter++;
}
+ }
out:
- post_res_handler (frame, this);
- return 0;
+ post_res_handler(frame, this);
+ return 0;
}
int
-shard_update_file_size_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
+shard_update_file_size_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- inode_t *inode = NULL;
- shard_local_t *local = NULL;
+ inode_t *inode = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if ((local->fd) && (local->fd->inode))
- inode = local->fd->inode;
- else if (local->loc.inode)
- inode = local->loc.inode;
+ if ((local->fd) && (local->fd->inode))
+ inode = local->fd->inode;
+ else if (local->loc.inode)
+ inode = local->loc.inode;
- if (op_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- SHARD_MSG_UPDATE_FILE_SIZE_FAILED, "Update to file size"
- " xattr failed on %s", uuid_utoa (inode->gfid));
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto err;
- }
+ if (op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno,
+ SHARD_MSG_UPDATE_FILE_SIZE_FAILED,
+ "Update to file size"
+ " xattr failed on %s",
+ uuid_utoa(inode->gfid));
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto err;
+ }
- if (shard_modify_size_and_block_count (&local->postbuf, dict)) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto err;
- }
+ if (shard_modify_size_and_block_count(&local->postbuf, dict)) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto err;
+ }
err:
- local->post_update_size_handler (frame, this);
- return 0;
+ local->post_update_size_handler(frame, this);
+ return 0;
}
int
-shard_set_size_attrs (int64_t size, int64_t block_count, int64_t **size_attr_p)
+shard_set_size_attrs(int64_t size, int64_t block_count, int64_t **size_attr_p)
{
- int ret = -1;
- int64_t *size_attr = NULL;
+ int ret = -1;
+ int64_t *size_attr = NULL;
- if (!size_attr_p)
- goto out;
+ if (!size_attr_p)
+ goto out;
- size_attr = GF_CALLOC (4, sizeof (int64_t), gf_shard_mt_int64_t);
- if (!size_attr)
- goto out;
+ size_attr = GF_CALLOC(4, sizeof(int64_t), gf_shard_mt_int64_t);
+ if (!size_attr)
+ goto out;
- size_attr[0] = hton64 (size);
- /* As sharding evolves, it _may_ be necessary to embed more pieces of
- * information within the same xattr. So allocating slots for them in
- * advance. For now, only bytes 0-63 and 128-191 which would make up the
- * current size and block count respectively of the file are valid.
- */
- size_attr[2] = hton64 (block_count);
+ size_attr[0] = hton64(size);
+ /* As sharding evolves, it _may_ be necessary to embed more pieces of
+ * information within the same xattr. So allocating slots for them in
+ * advance. For now, only bytes 0-63 and 128-191 which would make up the
+ * current size and block count respectively of the file are valid.
+ */
+ size_attr[2] = hton64(block_count);
- *size_attr_p = size_attr;
+ *size_attr_p = size_attr;
- ret = 0;
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-shard_update_file_size (call_frame_t *frame, xlator_t *this, fd_t *fd,
- loc_t *loc,
- shard_post_update_size_fop_handler_t handler)
+shard_update_file_size(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ loc_t *loc, shard_post_update_size_fop_handler_t handler)
{
- int ret = -1;
- int64_t *size_attr = NULL;
- inode_t *inode = NULL;
- shard_local_t *local = NULL;
- dict_t *xattr_req = NULL;
+ int ret = -1;
+ int64_t *size_attr = NULL;
+ int64_t delta_blocks = 0;
+ inode_t *inode = NULL;
+ shard_local_t *local = NULL;
+ dict_t *xattr_req = NULL;
- local = frame->local;
- local->post_update_size_handler = handler;
+ local = frame->local;
+ local->post_update_size_handler = handler;
- xattr_req = dict_new ();
- if (!xattr_req) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto out;
- }
-
- if (fd)
- inode = fd->inode;
- else
- inode = loc->inode;
-
- /* If both size and block count have not changed, then skip the xattrop.
- */
- if ((local->delta_size + local->hole_size == 0) &&
- (local->delta_blocks == 0)) {
- goto out;
- }
-
- ret = shard_set_size_attrs (local->delta_size + local->hole_size,
- local->delta_blocks, &size_attr);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, SHARD_MSG_SIZE_SET_FAILED,
- "Failed to set size attrs for %s",
- uuid_utoa (inode->gfid));
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto out;
- }
-
- ret = dict_set_bin (xattr_req, GF_XATTR_SHARD_FILE_SIZE, size_attr,
- 8 * 4);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, SHARD_MSG_DICT_SET_FAILED,
- "Failed to set key %s into dict. gfid=%s",
- GF_XATTR_SHARD_FILE_SIZE, uuid_utoa (inode->gfid));
- GF_FREE (size_attr);
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto out;
- }
+ xattr_req = dict_new();
+ if (!xattr_req) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto out;
+ }
+
+ if (fd)
+ inode = fd->inode;
+ else
+ inode = loc->inode;
+
+ /* If both size and block count have not changed, then skip the xattrop.
+ */
+ delta_blocks = GF_ATOMIC_GET(local->delta_blocks);
+ if ((local->delta_size + local->hole_size == 0) && (delta_blocks == 0)) {
+ goto out;
+ }
+
+ ret = shard_set_size_attrs(local->delta_size + local->hole_size,
+ delta_blocks, &size_attr);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_SIZE_SET_FAILED,
+ "Failed to set size attrs for %s", uuid_utoa(inode->gfid));
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto out;
+ }
+
+ ret = dict_set_bin(xattr_req, GF_XATTR_SHARD_FILE_SIZE, size_attr, 8 * 4);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_DICT_OP_FAILED,
+ "Failed to set key %s into dict. gfid=%s",
+ GF_XATTR_SHARD_FILE_SIZE, uuid_utoa(inode->gfid));
+ GF_FREE(size_attr);
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto out;
+ }
- if (fd)
- STACK_WIND (frame, shard_update_file_size_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fxattrop, fd,
- GF_XATTROP_ADD_ARRAY64, xattr_req, NULL);
- else
- STACK_WIND (frame, shard_update_file_size_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->xattrop, loc,
- GF_XATTROP_ADD_ARRAY64, xattr_req, NULL);
+ if (fd)
+ STACK_WIND(frame, shard_update_file_size_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fxattrop, fd,
+ GF_XATTROP_ADD_ARRAY64, xattr_req, NULL);
+ else
+ STACK_WIND(frame, shard_update_file_size_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->xattrop, loc,
+ GF_XATTROP_ADD_ARRAY64, xattr_req, NULL);
- dict_unref (xattr_req);
- return 0;
+ dict_unref(xattr_req);
+ return 0;
out:
- if (xattr_req)
- dict_unref (xattr_req);
- handler (frame, this);
- return 0;
-
+ if (xattr_req)
+ dict_unref(xattr_req);
+ handler(frame, this);
+ return 0;
}
static inode_t *
-shard_link_internal_dir_inode (shard_local_t *local, inode_t *inode,
- struct iatt *buf, shard_internal_dir_type_t type)
+shard_link_internal_dir_inode(shard_local_t *local, inode_t *inode,
+ struct iatt *buf, shard_internal_dir_type_t type)
{
- inode_t *linked_inode = NULL;
- shard_priv_t *priv = NULL;
- char *bname = NULL;
- inode_t **priv_inode = NULL;
+ inode_t *linked_inode = NULL;
+ shard_priv_t *priv = NULL;
+ char *bname = NULL;
+ inode_t **priv_inode = NULL;
+ inode_t *parent = NULL;
- priv = THIS->private;
+ priv = THIS->private;
- switch (type) {
+ switch (type) {
case SHARD_INTERNAL_DIR_DOT_SHARD:
- bname = ".shard";
- priv_inode = &priv->dot_shard_inode;
- break;
+ bname = GF_SHARD_DIR;
+ priv_inode = &priv->dot_shard_inode;
+ parent = inode->table->root;
+ break;
+ case SHARD_INTERNAL_DIR_DOT_SHARD_REMOVE_ME:
+ bname = GF_SHARD_REMOVE_ME_DIR;
+ priv_inode = &priv->dot_shard_rm_inode;
+ parent = priv->dot_shard_inode;
+ break;
default:
- break;
- }
+ break;
+ }
- linked_inode = inode_link (inode, inode->table->root, bname, buf);
- inode_lookup (linked_inode);
- *priv_inode = linked_inode;
- return linked_inode;
+ linked_inode = inode_link(inode, parent, bname, buf);
+ inode_lookup(linked_inode);
+ *priv_inode = linked_inode;
+ return linked_inode;
}
-
int
-shard_refresh_internal_dir_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata,
- struct iatt *postparent)
+shard_refresh_internal_dir_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ inode_t *inode, struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
{
- shard_local_t *local = NULL;
- inode_t *linked_inode = NULL;
- shard_internal_dir_type_t type = (shard_internal_dir_type_t) cookie;
-
- local = frame->local;
+ shard_local_t *local = NULL;
+ inode_t *linked_inode = NULL;
+ shard_internal_dir_type_t type = (shard_internal_dir_type_t)cookie;
- if (op_ret) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto out;
- }
+ local = frame->local;
- /* To-Do: Fix refcount increment per call to
- * shard_link_internal_dir_inode().
- */
- linked_inode = shard_link_internal_dir_inode (local, inode, buf, type);
- shard_inode_ctx_set_refreshed_flag (linked_inode, this);
+ if (op_ret) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto out;
+ }
+
+ /* To-Do: Fix refcount increment per call to
+ * shard_link_internal_dir_inode().
+ */
+ linked_inode = shard_link_internal_dir_inode(local, inode, buf, type);
+ shard_inode_ctx_mark_dir_refreshed(linked_inode, this);
out:
- shard_common_resolve_shards (frame, this, local->post_res_handler);
- return 0;
+ shard_common_resolve_shards(frame, this, local->post_res_handler);
+ return 0;
}
int
-shard_refresh_internal_dir (call_frame_t *frame, xlator_t *this,
- shard_internal_dir_type_t type)
+shard_refresh_internal_dir(call_frame_t *frame, xlator_t *this,
+ shard_internal_dir_type_t type)
{
- loc_t loc = {0,};
- inode_t *inode = NULL;
- shard_priv_t *priv = NULL;
- shard_local_t *local = NULL;
- uuid_t gfid = {0,};
+ loc_t loc = {
+ 0,
+ };
+ inode_t *inode = NULL;
+ shard_priv_t *priv = NULL;
+ shard_local_t *local = NULL;
+ uuid_t gfid = {
+ 0,
+ };
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- switch (type) {
+ switch (type) {
case SHARD_INTERNAL_DIR_DOT_SHARD:
- gf_uuid_copy (gfid, priv->dot_shard_gfid);
- break;
+ gf_uuid_copy(gfid, priv->dot_shard_gfid);
+ break;
+ case SHARD_INTERNAL_DIR_DOT_SHARD_REMOVE_ME:
+ gf_uuid_copy(gfid, priv->dot_shard_rm_gfid);
+ break;
default:
- break;
- }
+ break;
+ }
- inode = inode_find (this->itable, gfid);
+ inode = inode_find(this->itable, gfid);
- if (!shard_inode_ctx_needs_lookup (inode, this)) {
- local->op_ret = 0;
- goto out;
- }
+ if (!shard_inode_ctx_needs_lookup(inode, this)) {
+ local->op_ret = 0;
+ goto out;
+ }
- /* Plain assignment because the ref is already taken above through
- * call to inode_find()
- */
- loc.inode = inode;
- gf_uuid_copy (loc.gfid, gfid);
+ /* Plain assignment because the ref is already taken above through
+ * call to inode_find()
+ */
+ loc.inode = inode;
+ gf_uuid_copy(loc.gfid, gfid);
- STACK_WIND_COOKIE (frame, shard_refresh_internal_dir_cbk,
- (void *)(long) type, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, &loc, NULL);
- loc_wipe (&loc);
+ STACK_WIND_COOKIE(frame, shard_refresh_internal_dir_cbk, (void *)(long)type,
+ FIRST_CHILD(this), FIRST_CHILD(this)->fops->lookup, &loc,
+ NULL);
+ loc_wipe(&loc);
- return 0;
+ return 0;
out:
- shard_common_resolve_shards (frame, this, local->post_res_handler);
- return 0;
+ shard_common_resolve_shards(frame, this, local->post_res_handler);
+ return 0;
}
int
-shard_lookup_internal_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata,
- struct iatt *postparent)
+shard_lookup_internal_dir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
{
- inode_t *link_inode = NULL;
- shard_local_t *local = NULL;
- shard_internal_dir_type_t type = (shard_internal_dir_type_t) cookie;
+ inode_t *link_inode = NULL;
+ shard_local_t *local = NULL;
+ shard_internal_dir_type_t type = (shard_internal_dir_type_t)cookie;
- local = frame->local;
+ local = frame->local;
- if (op_ret) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto unwind;
- }
-
- if (!IA_ISDIR (buf->ia_type)) {
- gf_msg (this->name, GF_LOG_CRITICAL, 0,
- SHARD_MSG_DOT_SHARD_NODIR, "%s already exists and "
- "is not a directory. Please remove it from all bricks "
- "and try again", shard_internal_dir_string (type));
- local->op_ret = -1;
- local->op_errno = EIO;
- goto unwind;
- }
-
- link_inode = shard_link_internal_dir_inode (local, inode, buf, type);
- if (link_inode != inode) {
- shard_refresh_internal_dir (frame, this, type);
- } else {
- shard_inode_ctx_set_refreshed_flag (link_inode, this);
- shard_common_resolve_shards (frame, this,
- local->post_res_handler);
- }
- return 0;
+ if (op_ret) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto unwind;
+ }
+
+ if (!IA_ISDIR(buf->ia_type)) {
+ gf_msg(this->name, GF_LOG_CRITICAL, 0, SHARD_MSG_DOT_SHARD_NODIR,
+ "%s already exists and "
+ "is not a directory. Please remove it from all bricks "
+ "and try again",
+ shard_internal_dir_string(type));
+ local->op_ret = -1;
+ local->op_errno = EIO;
+ goto unwind;
+ }
+
+ link_inode = shard_link_internal_dir_inode(local, inode, buf, type);
+ if (link_inode != inode) {
+ shard_refresh_internal_dir(frame, this, type);
+ } else {
+ shard_inode_ctx_mark_dir_refreshed(link_inode, this);
+ shard_common_resolve_shards(frame, this, local->post_res_handler);
+ }
+ return 0;
unwind:
- local->post_res_handler (frame, this);
- return 0;
+ local->post_res_handler(frame, this);
+ return 0;
}
int
-shard_lookup_internal_dir (call_frame_t *frame, xlator_t *this,
- shard_post_resolve_fop_handler_t post_res_handler,
- shard_internal_dir_type_t type)
+shard_lookup_internal_dir(call_frame_t *frame, xlator_t *this,
+ shard_post_resolve_fop_handler_t post_res_handler,
+ shard_internal_dir_type_t type)
{
- int ret = -1;
- dict_t *xattr_req = NULL;
- shard_priv_t *priv = NULL;
- shard_local_t *local = NULL;
- uuid_t *gfid = NULL;
- loc_t *loc = NULL;
- gf_boolean_t free_gfid = _gf_true;
+ int ret = -1;
+ dict_t *xattr_req = NULL;
+ shard_priv_t *priv = NULL;
+ shard_local_t *local = NULL;
+ uuid_t *gfid = NULL;
+ loc_t *loc = NULL;
+ gf_boolean_t free_gfid = _gf_true;
- local = frame->local;
- priv = this->private;
- local->post_res_handler = post_res_handler;
+ local = frame->local;
+ priv = this->private;
+ local->post_res_handler = post_res_handler;
- gfid = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t);
- if (!gfid)
- goto err;
+ gfid = GF_MALLOC(sizeof(uuid_t), gf_common_mt_uuid_t);
+ if (!gfid)
+ goto err;
- xattr_req = dict_new ();
- if (!xattr_req) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto err;
- }
+ xattr_req = dict_new();
+ if (!xattr_req) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto err;
+ }
- switch (type) {
+ switch (type) {
case SHARD_INTERNAL_DIR_DOT_SHARD:
- gf_uuid_copy (*gfid, priv->dot_shard_gfid);
- loc = &local->dot_shard_loc;
- break;
+ gf_uuid_copy(*gfid, priv->dot_shard_gfid);
+ loc = &local->dot_shard_loc;
+ break;
+ case SHARD_INTERNAL_DIR_DOT_SHARD_REMOVE_ME:
+ gf_uuid_copy(*gfid, priv->dot_shard_rm_gfid);
+ loc = &local->dot_shard_rm_loc;
+ break;
default:
- break;
- }
-
- ret = dict_set_gfuuid (xattr_req, "gfid-req", *gfid, false);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, SHARD_MSG_DICT_SET_FAILED,
- "Failed to set gfid of %s into dict",
- shard_internal_dir_string (type));
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto err;
- } else {
- free_gfid = _gf_false;
- }
+ bzero(*gfid, sizeof(uuid_t));
+ break;
+ }
+
+ ret = dict_set_gfuuid(xattr_req, "gfid-req", *gfid, false);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_DICT_OP_FAILED,
+ "Failed to set gfid of %s into dict",
+ shard_internal_dir_string(type));
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto err;
+ } else {
+ free_gfid = _gf_false;
+ }
- STACK_WIND_COOKIE (frame, shard_lookup_internal_dir_cbk,
- (void *) (long) type, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
+ STACK_WIND_COOKIE(frame, shard_lookup_internal_dir_cbk, (void *)(long)type,
+ FIRST_CHILD(this), FIRST_CHILD(this)->fops->lookup, loc,
+ xattr_req);
- dict_unref (xattr_req);
- return 0;
+ dict_unref(xattr_req);
+ return 0;
err:
- if (xattr_req)
- dict_unref (xattr_req);
- if (free_gfid)
- GF_FREE (gfid);
- post_res_handler (frame, this);
- return 0;
+ if (xattr_req)
+ dict_unref(xattr_req);
+ if (free_gfid)
+ GF_FREE(gfid);
+ post_res_handler(frame, this);
+ return 0;
}
static void
-shard_inode_ctx_update (inode_t *inode, xlator_t *this, dict_t *xdata,
- struct iatt *buf)
-{
- int ret = 0;
- uint64_t size = 0;
- void *bsize = NULL;
-
- if (shard_inode_ctx_get_block_size (inode, this, &size)) {
- /* Fresh lookup */
- ret = dict_get_ptr (xdata, GF_XATTR_SHARD_BLOCK_SIZE, &bsize);
- if (!ret)
- size = ntoh64 (*((uint64_t *)bsize));
- /* If the file is sharded, set its block size, otherwise just
- * set 0.
- */
-
- shard_inode_ctx_set (inode, this, buf, size,
- SHARD_MASK_BLOCK_SIZE);
- }
- /* If the file is sharded, also set the remaining attributes,
- * except for ia_size and ia_blocks.
+shard_inode_ctx_update(inode_t *inode, xlator_t *this, dict_t *xdata,
+ struct iatt *buf)
+{
+ int ret = 0;
+ uint64_t size = 0;
+ void *bsize = NULL;
+
+ if (shard_inode_ctx_get_block_size(inode, this, &size)) {
+ /* Fresh lookup */
+ ret = dict_get_ptr(xdata, GF_XATTR_SHARD_BLOCK_SIZE, &bsize);
+ if (!ret)
+ size = ntoh64(*((uint64_t *)bsize));
+ /* If the file is sharded, set its block size, otherwise just
+ * set 0.
*/
- if (size) {
- shard_inode_ctx_set (inode, this, buf, 0, SHARD_LOOKUP_MASK);
- (void) shard_inode_ctx_invalidate (inode, this, buf);
- }
+
+ shard_inode_ctx_set(inode, this, buf, size, SHARD_MASK_BLOCK_SIZE);
+ }
+ /* If the file is sharded, also set the remaining attributes,
+ * except for ia_size and ia_blocks.
+ */
+ if (size) {
+ shard_inode_ctx_set(inode, this, buf, 0, SHARD_LOOKUP_MASK);
+ (void)shard_inode_ctx_invalidate(inode, this, buf);
+ }
}
int
-shard_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata, struct iatt *postparent)
-{
- if (op_ret < 0)
- goto unwind;
+shard_delete_shards(void *opaque);
- if (IA_ISDIR (buf->ia_type))
- goto unwind;
+int
+shard_delete_shards_cbk(int ret, call_frame_t *frame, void *data);
- /* Also, if the file is sharded, get the file size and block cnt xattr,
- * and store them in the stbuf appropriately.
- */
+int
+shard_start_background_deletion(xlator_t *this)
+{
+ int ret = 0;
+ gf_boolean_t i_cleanup = _gf_true;
+ shard_priv_t *priv = NULL;
+ call_frame_t *cleanup_frame = NULL;
- if (dict_get (xdata, GF_XATTR_SHARD_FILE_SIZE) &&
- frame->root->pid != GF_CLIENT_PID_GSYNCD)
- shard_modify_size_and_block_count (buf, xdata);
-
- /* If this was a fresh lookup, there are two possibilities:
- * 1) If the file is sharded (indicated by the presence of block size
- * xattr), store this block size, along with rdev and mode in its
- * inode ctx.
- * 2) If the file is not sharded, store size along with rdev and mode
- * (which are anyway don't cares) in inode ctx. Since @ctx_tmp is
- * already initialised to all zeroes, nothing more needs to be done.
- */
+ priv = this->private;
- (void) shard_inode_ctx_update (inode, this, xdata, buf);
+ LOCK(&priv->lock);
+ {
+ switch (priv->bg_del_state) {
+ case SHARD_BG_DELETION_NONE:
+ i_cleanup = _gf_true;
+ priv->bg_del_state = SHARD_BG_DELETION_LAUNCHING;
+ break;
+ case SHARD_BG_DELETION_LAUNCHING:
+ i_cleanup = _gf_false;
+ break;
+ case SHARD_BG_DELETION_IN_PROGRESS:
+ priv->bg_del_state = SHARD_BG_DELETION_LAUNCHING;
+ i_cleanup = _gf_false;
+ break;
+ default:
+ break;
+ }
+ }
+ UNLOCK(&priv->lock);
+ if (!i_cleanup)
+ return 0;
+
+ cleanup_frame = create_frame(this, this->ctx->pool);
+ if (!cleanup_frame) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, SHARD_MSG_MEMALLOC_FAILED,
+ "Failed to create "
+ "new frame to delete shards");
+ ret = -ENOMEM;
+ goto err;
+ }
+
+ set_lk_owner_from_ptr(&cleanup_frame->root->lk_owner, cleanup_frame->root);
+
+ ret = synctask_new(this->ctx->env, shard_delete_shards,
+ shard_delete_shards_cbk, cleanup_frame, cleanup_frame);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_WARNING, errno,
+ SHARD_MSG_SHARDS_DELETION_FAILED,
+ "failed to create task to do background "
+ "cleanup of shards");
+ STACK_DESTROY(cleanup_frame->root);
+ goto err;
+ }
+ return 0;
-unwind:
- SHARD_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, buf,
- xdata, postparent);
- return 0;
+err:
+ LOCK(&priv->lock);
+ {
+ priv->bg_del_state = SHARD_BG_DELETION_NONE;
+ }
+ UNLOCK(&priv->lock);
+ return ret;
}
int
-shard_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *xattr_req)
+shard_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata, struct iatt *postparent)
{
- int ret = -1;
- int32_t op_errno = ENOMEM;
- uint64_t block_size = 0;
- shard_local_t *local = NULL;
+ int ret = -1;
+ shard_priv_t *priv = NULL;
+ gf_boolean_t i_start_cleanup = _gf_false;
- if (frame->root->pid != GF_CLIENT_PID_GSYNCD) {
- SHARD_ENTRY_FOP_CHECK (loc, op_errno, err);
- }
-
- local = mem_get0 (this->local_pool);
- if (!local)
- goto err;
-
- frame->local = local;
+ priv = this->private;
- loc_copy (&local->loc, loc);
+ if (op_ret < 0)
+ goto unwind;
- local->xattr_req = xattr_req ? dict_ref (xattr_req) : dict_new ();
- if (!local->xattr_req)
- goto err;
+ if (IA_ISDIR(buf->ia_type))
+ goto unwind;
- if (shard_inode_ctx_get_block_size (loc->inode, this, &block_size)) {
- ret = dict_set_uint64 (local->xattr_req,
- GF_XATTR_SHARD_BLOCK_SIZE, 0);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- SHARD_MSG_DICT_SET_FAILED, "Failed to set dict"
- " value: key:%s for path %s",
- GF_XATTR_SHARD_BLOCK_SIZE, loc->path);
- goto err;
- }
- }
+ /* Also, if the file is sharded, get the file size and block cnt xattr,
+ * and store them in the stbuf appropriately.
+ */
- if (frame->root->pid != GF_CLIENT_PID_GSYNCD) {
- ret = dict_set_uint64 (local->xattr_req,
- GF_XATTR_SHARD_FILE_SIZE, 8 * 4);
- if (ret) {
- gf_msg (this->name, GF_LOG_WARNING, 0,
- SHARD_MSG_DICT_SET_FAILED,
- "Failed to set dict value: key:%s for path %s.",
- GF_XATTR_SHARD_FILE_SIZE, loc->path);
- goto err;
- }
- }
+ if (dict_get(xdata, GF_XATTR_SHARD_FILE_SIZE) &&
+ frame->root->pid != GF_CLIENT_PID_GSYNCD)
+ shard_modify_size_and_block_count(buf, xdata);
- if ((xattr_req) && (dict_get (xattr_req, GF_CONTENT_KEY)))
- dict_del (xattr_req, GF_CONTENT_KEY);
+ /* If this was a fresh lookup, there are two possibilities:
+ * 1) If the file is sharded (indicated by the presence of block size
+ * xattr), store this block size, along with rdev and mode in its
+ * inode ctx.
+ * 2) If the file is not sharded, store size along with rdev and mode
+ * (which are anyway don't cares) in inode ctx. Since @ctx_tmp is
+ * already initialised to all zeroes, nothing more needs to be done.
+ */
- STACK_WIND (frame, shard_lookup_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->lookup, loc, local->xattr_req);
+ (void)shard_inode_ctx_update(inode, this, xdata, buf);
- return 0;
+ LOCK(&priv->lock);
+ {
+ if (priv->first_lookup_done == _gf_false) {
+ priv->first_lookup_done = _gf_true;
+ i_start_cleanup = _gf_true;
+ }
+ }
+ UNLOCK(&priv->lock);
+ if (!i_start_cleanup)
+ goto unwind;
-err:
- SHARD_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL,
- NULL, NULL);
- return 0;
+ ret = shard_start_background_deletion(this);
+ if (ret < 0) {
+ LOCK(&priv->lock);
+ {
+ priv->first_lookup_done = _gf_false;
+ }
+ UNLOCK(&priv->lock);
+ }
+unwind:
+ SHARD_STACK_UNWIND(lookup, frame, op_ret, op_errno, inode, buf, xdata,
+ postparent);
+ return 0;
}
int
-shard_lookup_base_file_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata,
- struct iatt *postparent)
+shard_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr_req)
{
- int ret = -1;
- int32_t mask = SHARD_INODE_WRITE_MASK;
- shard_local_t *local = NULL;
- shard_inode_ctx_t ctx = {0,};
+ int ret = -1;
+ int32_t op_errno = ENOMEM;
+ uint64_t block_size = 0;
+ shard_local_t *local = NULL;
- local = frame->local;
+ this->itable = loc->inode->table;
+ if ((frame->root->pid != GF_CLIENT_PID_GSYNCD) &&
+ (frame->root->pid != GF_CLIENT_PID_GLFS_HEAL)) {
+ SHARD_ENTRY_FOP_CHECK(loc, op_errno, err);
+ }
- if (op_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- SHARD_MSG_BASE_FILE_LOOKUP_FAILED, "Lookup on base file"
- " failed : %s", loc_gfid_utoa (&(local->loc)));
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto unwind;
- }
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
- local->prebuf = *buf;
- if (shard_modify_size_and_block_count (&local->prebuf, xdata)) {
- local->op_ret = -1;
- local->op_errno = EINVAL;
- goto unwind;
- }
+ frame->local = local;
+
+ loc_copy(&local->loc, loc);
- if (shard_inode_ctx_get_all (inode, this, &ctx))
- mask = SHARD_ALL_MASK;
+ local->xattr_req = xattr_req ? dict_ref(xattr_req) : dict_new();
+ if (!local->xattr_req)
+ goto err;
- ret = shard_inode_ctx_set (inode, this, &local->prebuf, 0,
- (mask | SHARD_MASK_REFRESH_RESET));
+ if (shard_inode_ctx_get_block_size(loc->inode, this, &block_size)) {
+ ret = dict_set_uint64(local->xattr_req, GF_XATTR_SHARD_BLOCK_SIZE, 0);
if (ret) {
- gf_msg (this->name, GF_LOG_ERROR,
- SHARD_MSG_INODE_CTX_SET_FAILED, 0, "Failed to set inode"
- " write params into inode ctx for %s",
- uuid_utoa (buf->ia_gfid));
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto unwind;
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_DICT_OP_FAILED,
+ "Failed to set dict"
+ " value: key:%s for path %s",
+ GF_XATTR_SHARD_BLOCK_SIZE, loc->path);
+ goto err;
+ }
+ }
+
+ if (frame->root->pid != GF_CLIENT_PID_GSYNCD) {
+ ret = dict_set_uint64(local->xattr_req, GF_XATTR_SHARD_FILE_SIZE,
+ 8 * 4);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_DICT_OP_FAILED,
+ "Failed to set dict value: key:%s for path %s.",
+ GF_XATTR_SHARD_FILE_SIZE, loc->path);
+ goto err;
}
+ }
+
+ if ((xattr_req) && (dict_get(xattr_req, GF_CONTENT_KEY)))
+ dict_del(xattr_req, GF_CONTENT_KEY);
+
+ STACK_WIND(frame, shard_lookup_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, local->xattr_req);
+ return 0;
+err:
+ shard_common_failure_unwind(GF_FOP_LOOKUP, frame, -1, op_errno);
+ return 0;
+}
+
+int
+shard_set_iattr_invoke_post_handler(call_frame_t *frame, xlator_t *this,
+ inode_t *inode, int32_t op_ret,
+ int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
+{
+ int ret = -1;
+ int32_t mask = SHARD_INODE_WRITE_MASK;
+ shard_local_t *local = frame->local;
+ shard_inode_ctx_t ctx = {
+ 0,
+ };
+
+ if (op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno,
+ SHARD_MSG_BASE_FILE_LOOKUP_FAILED,
+ "Lookup on base file"
+ " failed : %s",
+ uuid_utoa(inode->gfid));
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto unwind;
+ }
+
+ local->prebuf = *buf;
+ if (shard_modify_size_and_block_count(&local->prebuf, xdata)) {
+ local->op_ret = -1;
+ local->op_errno = EINVAL;
+ goto unwind;
+ }
+
+ if (shard_inode_ctx_get_all(inode, this, &ctx))
+ mask = SHARD_ALL_MASK;
+
+ ret = shard_inode_ctx_set(inode, this, &local->prebuf, 0,
+ (mask | SHARD_MASK_REFRESH_RESET));
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, SHARD_MSG_INODE_CTX_SET_FAILED, 0,
+ "Failed to set inode"
+ " write params into inode ctx for %s",
+ uuid_utoa(buf->ia_gfid));
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto unwind;
+ }
unwind:
- local->handler (frame, this);
- return 0;
+ local->handler(frame, this);
+ return 0;
}
int
-shard_lookup_base_file (call_frame_t *frame, xlator_t *this, loc_t *loc,
- shard_post_fop_handler_t handler)
+shard_fstat_base_file_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
{
- int ret = -1;
- shard_local_t *local = NULL;
- dict_t *xattr_req = NULL;
- gf_boolean_t need_refresh = _gf_false;
+ shard_local_t *local = frame->local;
- local = frame->local;
- local->handler = handler;
+ shard_set_iattr_invoke_post_handler(frame, this, local->fd->inode, op_ret,
+ op_errno, buf, xdata);
+ return 0;
+}
- ret = shard_inode_ctx_fill_iatt_from_cache (loc->inode, this,
- &local->prebuf,
- &need_refresh);
- /* By this time, inode ctx should have been created either in create,
- * mknod, readdirp or lookup. If not it is a bug!
- */
- if ((ret == 0) && (need_refresh == _gf_false)) {
- gf_msg_debug (this->name, 0, "Skipping lookup on base file: %s"
- "Serving prebuf off the inode ctx cache",
- uuid_utoa (loc->gfid));
- goto out;
- }
+int
+shard_lookup_base_file_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
+{
+ /* In case of op_ret < 0, inode passed to this function will be NULL
+ ex: in case of op_errno = ENOENT. So refer prefilled inode data
+ which is part of local.
+ Note: Reassigning/overriding the inode passed to this cbk with inode
+ which is part of *struct shard_local_t* won't cause any issue as
+ both inodes have same reference/address as of the inode passed */
+ inode = ((shard_local_t *)frame->local)->loc.inode;
- xattr_req = dict_new ();
- if (!xattr_req) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto out;
- }
+ shard_set_iattr_invoke_post_handler(frame, this, inode, op_ret, op_errno,
+ buf, xdata);
+ return 0;
+}
- SHARD_MD_READ_FOP_INIT_REQ_DICT (this, xattr_req, loc->gfid,
- local, out);
+/* This function decides whether to make file based lookup or
+ * fd based lookup (fstat) depending on the 3rd and 4th arg.
+ * If fd != NULL and loc == NULL then call is for fstat
+ * If fd == NULL and loc != NULL then call is for file based
+ * lookup. Please pass args based on the requirement.
+ */
+int
+shard_refresh_base_file(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ fd_t *fd, shard_post_fop_handler_t handler)
+{
+ int ret = -1;
+ inode_t *inode = NULL;
+ shard_local_t *local = NULL;
+ dict_t *xattr_req = NULL;
+ gf_boolean_t need_refresh = _gf_false;
+
+ local = frame->local;
+ local->handler = handler;
+ inode = fd ? fd->inode : loc->inode;
+
+ ret = shard_inode_ctx_fill_iatt_from_cache(inode, this, &local->prebuf,
+ &need_refresh);
+ /* By this time, inode ctx should have been created either in create,
+ * mknod, readdirp or lookup. If not it is a bug!
+ */
+ if ((ret == 0) && (need_refresh == _gf_false)) {
+ gf_msg_debug(this->name, 0,
+ "Skipping lookup on base file: %s"
+ "Serving prebuf off the inode ctx cache",
+ uuid_utoa(inode->gfid));
+ goto out;
+ }
+
+ xattr_req = dict_new();
+ if (!xattr_req) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto out;
+ }
- STACK_WIND (frame, shard_lookup_base_file_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
+ SHARD_MD_READ_FOP_INIT_REQ_DICT(this, xattr_req, inode->gfid, local, out);
- dict_unref (xattr_req);
- return 0;
+ if (fd)
+ STACK_WIND(frame, shard_fstat_base_file_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fstat, fd, xattr_req);
+ else
+ STACK_WIND(frame, shard_lookup_base_file_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
-out:
- if (xattr_req)
- dict_unref (xattr_req);
- handler (frame, this);
- return 0;
+ dict_unref(xattr_req);
+ return 0;
+out:
+ if (xattr_req)
+ dict_unref(xattr_req);
+ handler(frame, this);
+ return 0;
}
int
-shard_post_fstat_handler (call_frame_t *frame, xlator_t *this)
+shard_post_fstat_handler(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->op_ret >= 0)
- shard_inode_ctx_set (local->fd->inode, this, &local->prebuf, 0,
- SHARD_LOOKUP_MASK);
+ if (local->op_ret >= 0)
+ shard_inode_ctx_set(local->fd->inode, this, &local->prebuf, 0,
+ SHARD_LOOKUP_MASK);
- SHARD_STACK_UNWIND (fstat, frame, local->op_ret, local->op_errno,
- &local->prebuf, local->xattr_rsp);
- return 0;
+ SHARD_STACK_UNWIND(fstat, frame, local->op_ret, local->op_errno,
+ &local->prebuf, local->xattr_rsp);
+ return 0;
}
int
-shard_post_stat_handler (call_frame_t *frame, xlator_t *this)
+shard_post_stat_handler(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->op_ret >= 0)
- shard_inode_ctx_set (local->loc.inode, this, &local->prebuf, 0,
- SHARD_LOOKUP_MASK);
+ if (local->op_ret >= 0)
+ shard_inode_ctx_set(local->loc.inode, this, &local->prebuf, 0,
+ SHARD_LOOKUP_MASK);
- SHARD_STACK_UNWIND (stat, frame, local->op_ret, local->op_errno,
- &local->prebuf, local->xattr_rsp);
- return 0;
+ SHARD_STACK_UNWIND(stat, frame, local->op_ret, local->op_errno,
+ &local->prebuf, local->xattr_rsp);
+ return 0;
}
int
-shard_common_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- dict_t *xdata)
+shard_common_stat_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
{
- inode_t *inode = NULL;
- shard_local_t *local = NULL;
+ inode_t *inode = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- SHARD_MSG_STAT_FAILED, "stat failed: %s",
- local->fd ? uuid_utoa (local->fd->inode->gfid)
- : uuid_utoa ((local->loc.inode)->gfid));
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto unwind;
- }
+ if (op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, SHARD_MSG_STAT_FAILED,
+ "stat failed: %s",
+ local->fd ? uuid_utoa(local->fd->inode->gfid)
+ : uuid_utoa((local->loc.inode)->gfid));
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto unwind;
+ }
- local->prebuf = *buf;
- if (shard_modify_size_and_block_count (&local->prebuf, xdata)) {
- local->op_ret = -1;
- local->op_errno = EINVAL;
- goto unwind;
- }
- local->xattr_rsp = dict_ref (xdata);
+ local->prebuf = *buf;
+ if (shard_modify_size_and_block_count(&local->prebuf, xdata)) {
+ local->op_ret = -1;
+ local->op_errno = EINVAL;
+ goto unwind;
+ }
+ local->xattr_rsp = dict_ref(xdata);
- if (local->loc.inode)
- inode = local->loc.inode;
- else
- inode = local->fd->inode;
+ if (local->loc.inode)
+ inode = local->loc.inode;
+ else
+ inode = local->fd->inode;
- shard_inode_ctx_invalidate (inode, this, &local->prebuf);
+ shard_inode_ctx_invalidate(inode, this, &local->prebuf);
unwind:
- local->handler (frame, this);
- return 0;
+ local->handler(frame, this);
+ return 0;
}
int
-shard_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+shard_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- int ret = -1;
- uint64_t block_size = 0;
- shard_local_t *local = NULL;
-
- if ((IA_ISDIR (loc->inode->ia_type)) ||
- (IA_ISLNK (loc->inode->ia_type))) {
- STACK_WIND (frame, default_stat_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->stat, loc, xdata);
- return 0;
- }
-
- ret = shard_inode_ctx_get_block_size (loc->inode, this, &block_size);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_CTX_GET_FAILED, "Failed to get block "
- "size from inode ctx of %s",
- uuid_utoa (loc->inode->gfid));
- goto err;
- }
+ int ret = -1;
+ uint64_t block_size = 0;
+ shard_local_t *local = NULL;
- if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
- STACK_WIND (frame, default_stat_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->stat, loc, xdata);
- return 0;
- }
+ if ((IA_ISDIR(loc->inode->ia_type)) || (IA_ISLNK(loc->inode->ia_type))) {
+ STACK_WIND(frame, default_stat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->stat, loc, xdata);
+ return 0;
+ }
- local = mem_get0 (this->local_pool);
- if (!local)
- goto err;
+ ret = shard_inode_ctx_get_block_size(loc->inode, this, &block_size);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_CTX_GET_FAILED,
+ "Failed to get block "
+ "size from inode ctx of %s",
+ uuid_utoa(loc->inode->gfid));
+ goto err;
+ }
- frame->local = local;
+ if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
+ STACK_WIND(frame, default_stat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->stat, loc, xdata);
+ return 0;
+ }
- local->handler = shard_post_stat_handler;
- loc_copy (&local->loc, loc);
- local->xattr_req = (xdata) ? dict_ref (xdata) : dict_new ();
- if (!local->xattr_req)
- goto err;
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
- SHARD_MD_READ_FOP_INIT_REQ_DICT (this, local->xattr_req,
- local->loc.gfid, local, err);
+ frame->local = local;
- STACK_WIND (frame, shard_common_stat_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->stat, loc, local->xattr_req);
+ local->handler = shard_post_stat_handler;
+ loc_copy(&local->loc, loc);
+ local->xattr_req = (xdata) ? dict_ref(xdata) : dict_new();
+ if (!local->xattr_req)
+ goto err;
- return 0;
+ SHARD_MD_READ_FOP_INIT_REQ_DICT(this, local->xattr_req, local->loc.gfid,
+ local, err);
+ STACK_WIND(frame, shard_common_stat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->stat, loc, local->xattr_req);
+ return 0;
err:
- SHARD_STACK_UNWIND (stat, frame, -1, ENOMEM, NULL, NULL);
- return 0;
+ shard_common_failure_unwind(GF_FOP_STAT, frame, -1, ENOMEM);
+ return 0;
}
int
-shard_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+shard_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- int ret = -1;
- uint64_t block_size = 0;
- shard_local_t *local = NULL;
+ int ret = -1;
+ uint64_t block_size = 0;
+ shard_local_t *local = NULL;
- if ((IA_ISDIR (fd->inode->ia_type)) ||
- (IA_ISLNK (fd->inode->ia_type))) {
- STACK_WIND (frame, default_fstat_cbk, FIRST_CHILD(this),
- FIRST_CHILD (this)->fops->fstat, fd, xdata);
- return 0;
- }
-
- ret = shard_inode_ctx_get_block_size (fd->inode, this, &block_size);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_CTX_GET_FAILED, "Failed to get block "
- "size from inode ctx of %s",
- uuid_utoa (fd->inode->gfid));
- goto err;
- }
+ if ((IA_ISDIR(fd->inode->ia_type)) || (IA_ISLNK(fd->inode->ia_type))) {
+ STACK_WIND(frame, default_fstat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fstat, fd, xdata);
+ return 0;
+ }
- if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
- STACK_WIND (frame, default_fstat_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fstat, fd, xdata);
- return 0;
- }
+ ret = shard_inode_ctx_get_block_size(fd->inode, this, &block_size);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_CTX_GET_FAILED,
+ "Failed to get block "
+ "size from inode ctx of %s",
+ uuid_utoa(fd->inode->gfid));
+ goto err;
+ }
- if (!this->itable)
- this->itable = fd->inode->table;
+ if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
+ STACK_WIND(frame, default_fstat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fstat, fd, xdata);
+ return 0;
+ }
- local = mem_get0 (this->local_pool);
- if (!local)
- goto err;
+ if (!this->itable)
+ this->itable = fd->inode->table;
- frame->local = local;
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
- local->handler = shard_post_fstat_handler;
- local->fd = fd_ref (fd);
- local->xattr_req = (xdata) ? dict_ref (xdata) : dict_new ();
- if (!local->xattr_req)
- goto err;
+ frame->local = local;
- SHARD_MD_READ_FOP_INIT_REQ_DICT (this, local->xattr_req,
- fd->inode->gfid, local, err);
+ local->handler = shard_post_fstat_handler;
+ local->fd = fd_ref(fd);
+ local->xattr_req = (xdata) ? dict_ref(xdata) : dict_new();
+ if (!local->xattr_req)
+ goto err;
- STACK_WIND (frame, shard_common_stat_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fstat, fd, local->xattr_req);
- return 0;
+ SHARD_MD_READ_FOP_INIT_REQ_DICT(this, local->xattr_req, fd->inode->gfid,
+ local, err);
+ STACK_WIND(frame, shard_common_stat_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fstat, fd, local->xattr_req);
+ return 0;
err:
- SHARD_STACK_UNWIND (fstat, frame, -1, ENOMEM, NULL, NULL);
- return 0;
+ shard_common_failure_unwind(GF_FOP_FSTAT, frame, -1, ENOMEM);
+ return 0;
}
int
-shard_post_update_size_truncate_handler (call_frame_t *frame, xlator_t *this)
+shard_post_update_size_truncate_handler(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->fop == GF_FOP_TRUNCATE)
- SHARD_STACK_UNWIND (truncate, frame, local->op_ret,
- local->op_errno, &local->prebuf,
- &local->postbuf, NULL);
- else
- SHARD_STACK_UNWIND (ftruncate, frame, local->op_ret,
- local->op_errno, &local->prebuf,
- &local->postbuf, NULL);
- return 0;
+ if (local->fop == GF_FOP_TRUNCATE)
+ SHARD_STACK_UNWIND(truncate, frame, local->op_ret, local->op_errno,
+ &local->prebuf, &local->postbuf, NULL);
+ else
+ SHARD_STACK_UNWIND(ftruncate, frame, local->op_ret, local->op_errno,
+ &local->prebuf, &local->postbuf, NULL);
+ return 0;
}
int
-shard_truncate_last_shard_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret, int32_t op_errno,
- struct iatt *prebuf, struct iatt *postbuf,
- dict_t *xdata)
+shard_truncate_last_shard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *prebuf, struct iatt *postbuf,
+ dict_t *xdata)
{
- inode_t *inode = NULL;
- shard_local_t *local = NULL;
+ inode_t *inode = NULL;
+ int64_t delta_blocks = 0;
+ shard_local_t *local = NULL;
- local = frame->local;
-
- SHARD_UNSET_ROOT_FS_ID (frame, local);
-
- inode = (local->fop == GF_FOP_TRUNCATE) ? local->loc.inode
- : local->fd->inode;
- if (op_ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- SHARD_MSG_TRUNCATE_LAST_SHARD_FAILED, "truncate on last"
- " shard failed : %s", uuid_utoa (inode->gfid));
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto err;
- }
-
- local->postbuf.ia_size = local->offset;
- local->postbuf.ia_blocks -= (prebuf->ia_blocks - postbuf->ia_blocks);
- /* Let the delta be negative. We want xattrop to do subtraction */
- local->delta_size = local->postbuf.ia_size - local->prebuf.ia_size;
- local->delta_blocks = postbuf->ia_blocks - prebuf->ia_blocks;
- local->hole_size = 0;
+ local = frame->local;
- shard_inode_ctx_set (inode, this, postbuf, 0, SHARD_MASK_TIMES);
-
- shard_update_file_size (frame, this, NULL, &local->loc,
- shard_post_update_size_truncate_handler);
- return 0;
+ SHARD_UNSET_ROOT_FS_ID(frame, local);
+ inode = (local->fop == GF_FOP_TRUNCATE) ? local->loc.inode
+ : local->fd->inode;
+ if (op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno,
+ SHARD_MSG_TRUNCATE_LAST_SHARD_FAILED,
+ "truncate on last"
+ " shard failed : %s",
+ uuid_utoa(inode->gfid));
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto err;
+ }
+
+ local->postbuf.ia_size = local->offset;
+ /* Let the delta be negative. We want xattrop to do subtraction */
+ local->delta_size = local->postbuf.ia_size - local->prebuf.ia_size;
+ delta_blocks = GF_ATOMIC_ADD(local->delta_blocks,
+ postbuf->ia_blocks - prebuf->ia_blocks);
+ GF_ASSERT(delta_blocks <= 0);
+ local->postbuf.ia_blocks += delta_blocks;
+ local->hole_size = 0;
+
+ shard_inode_ctx_set(inode, this, &local->postbuf, 0, SHARD_MASK_TIMES);
+ shard_update_file_size(frame, this, NULL, &local->loc,
+ shard_post_update_size_truncate_handler);
+ return 0;
err:
- if (local->fop == GF_FOP_TRUNCATE)
- SHARD_STACK_UNWIND (truncate, frame, local->op_ret,
- local->op_errno, NULL, NULL, NULL);
- else
- SHARD_STACK_UNWIND (ftruncate, frame, local->op_ret,
- local->op_errno, NULL, NULL, NULL);
- return 0;
+ shard_common_failure_unwind(local->fop, frame, local->op_ret,
+ local->op_errno);
+ return 0;
}
int
-shard_truncate_last_shard (call_frame_t *frame, xlator_t *this, inode_t *inode)
+shard_truncate_last_shard(call_frame_t *frame, xlator_t *this, inode_t *inode)
{
- size_t last_shard_size_after = 0;
- loc_t loc = {0,};
- shard_local_t *local = NULL;
+ size_t last_shard_size_after = 0;
+ loc_t loc = {
+ 0,
+ };
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- /* A NULL inode could be due to the fact that the last shard which
- * needs to be truncated does not exist due to it lying in a hole
- * region. So the only thing left to do in that case would be an
- * update to file size xattr.
- */
- if (!inode) {
- gf_msg_debug (this->name, 0, "Last shard to be truncated absent"
- " in backend: %s. Directly proceeding to update "
- "file size", uuid_utoa (inode->gfid));
- shard_update_file_size (frame, this, NULL, &local->loc,
- shard_post_update_size_truncate_handler);
- return 0;
- }
+ /* A NULL inode could be due to the fact that the last shard which
+ * needs to be truncated does not exist due to it lying in a hole
+ * region. So the only thing left to do in that case would be an
+ * update to file size xattr.
+ */
+ if (!inode) {
+ gf_msg_debug(this->name, 0,
+ "Last shard to be truncated absent in backend: %" PRIu64
+ " of gfid %s. Directly proceeding to update file size",
+ local->first_block, uuid_utoa(local->loc.inode->gfid));
+ shard_update_file_size(frame, this, NULL, &local->loc,
+ shard_post_update_size_truncate_handler);
+ return 0;
+ }
- SHARD_SET_ROOT_FS_ID (frame, local);
+ SHARD_SET_ROOT_FS_ID(frame, local);
- loc.inode = inode_ref (inode);
- gf_uuid_copy (loc.gfid, inode->gfid);
+ loc.inode = inode_ref(inode);
+ gf_uuid_copy(loc.gfid, inode->gfid);
- last_shard_size_after = (local->offset % local->block_size);
+ last_shard_size_after = (local->offset % local->block_size);
- STACK_WIND (frame, shard_truncate_last_shard_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->truncate, &loc,
- last_shard_size_after, NULL);
- loc_wipe (&loc);
- return 0;
+ STACK_WIND(frame, shard_truncate_last_shard_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, &loc, last_shard_size_after,
+ NULL);
+ loc_wipe(&loc);
+ return 0;
}
-int
-shard_unlink_shards_do_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata);
-
void
-shard_unlink_block_inode (shard_local_t *local, int shard_block_num);
+shard_unlink_block_inode(shard_local_t *local, int shard_block_num);
int
-shard_truncate_htol_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+shard_truncate_htol_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- int call_count = 0;
- int shard_block_num = (long) cookie;
- shard_local_t *local = NULL;
-
- local = frame->local;
+ int ret = 0;
+ int call_count = 0;
+ int shard_block_num = (long)cookie;
+ uint64_t block_count = 0;
+ shard_local_t *local = NULL;
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto done;
- }
+ local = frame->local;
- shard_unlink_block_inode (local, shard_block_num);
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto done;
+ }
+ ret = dict_get_uint64(xdata, GF_GET_FILE_BLOCK_COUNT, &block_count);
+ if (!ret) {
+ GF_ATOMIC_SUB(local->delta_blocks, block_count);
+ } else {
+ /* dict_get failed possibly due to a heterogeneous cluster? */
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_DICT_OP_FAILED,
+ "Failed to get key %s from dict during truncate of gfid %s",
+ GF_GET_FILE_BLOCK_COUNT,
+ uuid_utoa(local->resolver_base_inode->gfid));
+ }
+
+ shard_unlink_block_inode(local, shard_block_num);
done:
- call_count = shard_call_count_return (frame);
- if (call_count == 0) {
- SHARD_UNSET_ROOT_FS_ID (frame, local);
- shard_truncate_last_shard (frame, this, local->inode_list[0]);
- }
- return 0;
-}
-
-int
-shard_truncate_htol (call_frame_t *frame, xlator_t *this, inode_t *inode)
-{
- int i = 1;
- int ret = -1;
- int call_count = 0;
- uint32_t cur_block = 0;
- uint32_t last_block = 0;
- char path[PATH_MAX] = {0,};
- char *bname = NULL;
- loc_t loc = {0,};
- gf_boolean_t wind_failed = _gf_false;
- shard_local_t *local = NULL;
- shard_priv_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
-
- cur_block = local->first_block + 1;
- last_block = local->last_block;
-
- /* Determine call count */
- for (i = 1; i < local->num_blocks; i++) {
- if (!local->inode_list[i])
- continue;
- call_count++;
- }
-
- if (!call_count) {
- /* Call count = 0 implies that all of the shards that need to be
- * unlinked do not exist. So shard xlator would now proceed to
- * do the final truncate + size updates.
- */
- gf_msg_debug (this->name, 0, "Shards to be unlinked as part of "
- "truncate absent in backend: %s. Directly "
- "proceeding to update file size",
- uuid_utoa (inode->gfid));
- local->postbuf.ia_size = local->offset;
- local->postbuf.ia_blocks = local->prebuf.ia_blocks;
- local->delta_size = local->postbuf.ia_size -
- local->prebuf.ia_size;
- local->delta_blocks = 0;
- local->hole_size = 0;
- shard_update_file_size (frame, this, local->fd, &local->loc,
- shard_post_update_size_truncate_handler);
- return 0;
- }
-
- local->call_count = call_count;
- i = 1;
-
- SHARD_SET_ROOT_FS_ID (frame, local);
- while (cur_block <= last_block) {
- if (!local->inode_list[i]) {
- cur_block++;
- i++;
- continue;
- }
- if (wind_failed) {
- shard_truncate_htol_cbk (frame, (void *)(long) cur_block,
- this, -1, ENOMEM, NULL, NULL,
- NULL);
- goto next;
- }
-
- shard_make_block_abspath (cur_block, inode->gfid, path,
- sizeof (path));
- bname = strrchr (path, '/') + 1;
- loc.parent = inode_ref (priv->dot_shard_inode);
- ret = inode_path (loc.parent, bname, (char **)&(loc.path));
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_PATH_FAILED, "Inode path failed"
- " on %s. Base file gfid = %s", bname,
- uuid_utoa (inode->gfid));
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- loc_wipe (&loc);
- wind_failed = _gf_true;
- shard_truncate_htol_cbk (frame, (void *)(long) cur_block,
- this, -1, ENOMEM, NULL, NULL,
- NULL);
- goto next;
- }
- loc.name = strrchr (loc.path, '/');
- if (loc.name)
- loc.name++;
- loc.inode = inode_ref (local->inode_list[i]);
-
- STACK_WIND_COOKIE (frame, shard_truncate_htol_cbk,
- (void *) (long) cur_block, FIRST_CHILD(this),
- FIRST_CHILD (this)->fops->unlink, &loc,
- 0, NULL);
- loc_wipe (&loc);
-next:
- i++;
- cur_block++;
- if (!--call_count)
- break;
- }
+ call_count = shard_call_count_return(frame);
+ if (call_count == 0) {
+ SHARD_UNSET_ROOT_FS_ID(frame, local);
+ shard_truncate_last_shard(frame, this, local->inode_list[0]);
+ }
+ return 0;
+}
+
+int
+shard_truncate_htol(call_frame_t *frame, xlator_t *this, inode_t *inode)
+{
+ int i = 1;
+ int ret = -1;
+ int call_count = 0;
+ uint32_t cur_block = 0;
+ uint32_t last_block = 0;
+ char path[PATH_MAX] = {
+ 0,
+ };
+ char *bname = NULL;
+ loc_t loc = {
+ 0,
+ };
+ gf_boolean_t wind_failed = _gf_false;
+ shard_local_t *local = NULL;
+ shard_priv_t *priv = NULL;
+ dict_t *xdata_req = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ cur_block = local->first_block + 1;
+ last_block = local->last_block;
+
+ /* Determine call count */
+ for (i = 1; i < local->num_blocks; i++) {
+ if (!local->inode_list[i])
+ continue;
+ call_count++;
+ }
+
+ if (!call_count) {
+ /* Call count = 0 implies that all of the shards that need to be
+ * unlinked do not exist. So shard xlator would now proceed to
+ * do the final truncate + size updates.
+ */
+ gf_msg_debug(this->name, 0,
+ "Shards to be unlinked as part of "
+ "truncate absent in backend: %s. Directly "
+ "proceeding to update file size",
+ uuid_utoa(inode->gfid));
+ local->postbuf.ia_size = local->offset;
+ local->postbuf.ia_blocks = local->prebuf.ia_blocks;
+ local->delta_size = local->postbuf.ia_size - local->prebuf.ia_size;
+ GF_ATOMIC_INIT(local->delta_blocks, 0);
+ local->hole_size = 0;
+ shard_update_file_size(frame, this, local->fd, &local->loc,
+ shard_post_update_size_truncate_handler);
+ return 0;
+ }
+
+ local->call_count = call_count;
+ i = 1;
+ xdata_req = dict_new();
+ if (!xdata_req) {
+ shard_common_failure_unwind(local->fop, frame, -1, ENOMEM);
+ return 0;
+ }
+ ret = dict_set_uint64(xdata_req, GF_GET_FILE_BLOCK_COUNT, 8 * 8);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_DICT_OP_FAILED,
+ "Failed to set key %s into dict during truncate of %s",
+ GF_GET_FILE_BLOCK_COUNT,
+ uuid_utoa(local->resolver_base_inode->gfid));
+ dict_unref(xdata_req);
+ shard_common_failure_unwind(local->fop, frame, -1, ENOMEM);
+ return 0;
+ }
+
+ SHARD_SET_ROOT_FS_ID(frame, local);
+ while (cur_block <= last_block) {
+ if (!local->inode_list[i]) {
+ cur_block++;
+ i++;
+ continue;
+ }
+ if (wind_failed) {
+ shard_truncate_htol_cbk(frame, (void *)(long)cur_block, this, -1,
+ ENOMEM, NULL, NULL, NULL);
+ goto next;
+ }
+
+ shard_make_block_abspath(cur_block, inode->gfid, path, sizeof(path));
+ bname = strrchr(path, '/') + 1;
+ loc.parent = inode_ref(priv->dot_shard_inode);
+ ret = inode_path(loc.parent, bname, (char **)&(loc.path));
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_PATH_FAILED,
+ "Inode path failed"
+ " on %s. Base file gfid = %s",
+ bname, uuid_utoa(inode->gfid));
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ loc_wipe(&loc);
+ wind_failed = _gf_true;
+ shard_truncate_htol_cbk(frame, (void *)(long)cur_block, this, -1,
+ ENOMEM, NULL, NULL, NULL);
+ goto next;
+ }
+ loc.name = strrchr(loc.path, '/');
+ if (loc.name)
+ loc.name++;
+ loc.inode = inode_ref(local->inode_list[i]);
+
+ STACK_WIND_COOKIE(frame, shard_truncate_htol_cbk,
+ (void *)(long)cur_block, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, &loc, 0, xdata_req);
+ loc_wipe(&loc);
+ next:
+ i++;
+ cur_block++;
+ if (!--call_count)
+ break;
+ }
+ dict_unref(xdata_req);
+ return 0;
+}
+
+int
+shard_truncate_do(call_frame_t *frame, xlator_t *this)
+{
+ shard_local_t *local = NULL;
+
+ local = frame->local;
+
+ if (local->num_blocks == 1) {
+ /* This means that there are no shards to be unlinked.
+ * The fop boils down to truncating the last shard, updating
+ * the size and unwinding.
+ */
+ shard_truncate_last_shard(frame, this, local->inode_list[0]);
return 0;
-
+ } else {
+ shard_truncate_htol(frame, this, local->loc.inode);
+ }
+ return 0;
}
int
-shard_truncate_do (call_frame_t *frame, xlator_t *this)
+shard_post_lookup_shards_truncate_handler(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->num_blocks == 1) {
- /* This means that there are no shards to be unlinked.
- * The fop boils down to truncating the last shard, updating
- * the size and unwinding.
- */
- shard_truncate_last_shard (frame, this,
- local->inode_list[0]);
- return 0;
- } else {
- shard_truncate_htol (frame, this, local->loc.inode);
- }
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(local->fop, frame, local->op_ret,
+ local->op_errno);
return 0;
-}
-
-int
-shard_post_lookup_shards_truncate_handler (call_frame_t *frame, xlator_t *this)
-{
- shard_local_t *local = NULL;
-
- local = frame->local;
+ }
- if (local->op_ret < 0) {
- if (local->fop == GF_FOP_TRUNCATE)
- SHARD_STACK_UNWIND (truncate, frame, local->op_ret,
- local->op_errno, NULL, NULL, NULL);
- else
- SHARD_STACK_UNWIND (ftruncate, frame, local->op_ret,
- local->op_errno, NULL, NULL, NULL);
- return 0;
- }
-
- shard_truncate_do (frame, this);
- return 0;
+ shard_truncate_do(frame, this);
+ return 0;
}
void
-shard_link_block_inode (shard_local_t *local, int block_num, inode_t *inode,
- struct iatt *buf)
-{
- int list_index = 0;
- char block_bname[256] = {0,};
- inode_t *linked_inode = NULL;
- xlator_t *this = NULL;
- inode_t *fsync_inode = NULL;
- shard_priv_t *priv = NULL;
-
- this = THIS;
- priv = this->private;
+shard_link_block_inode(shard_local_t *local, int block_num, inode_t *inode,
+ struct iatt *buf)
+{
+ int list_index = 0;
+ char block_bname[256] = {
+ 0,
+ };
+ uuid_t gfid = {
+ 0,
+ };
+ inode_t *linked_inode = NULL;
+ xlator_t *this = NULL;
+ inode_t *fsync_inode = NULL;
+ shard_priv_t *priv = NULL;
+ inode_t *base_inode = NULL;
+
+ this = THIS;
+ priv = this->private;
+ if (local->loc.inode) {
+ gf_uuid_copy(gfid, local->loc.inode->gfid);
+ base_inode = local->loc.inode;
+ } else if (local->resolver_base_inode) {
+ gf_uuid_copy(gfid, local->resolver_base_inode->gfid);
+ base_inode = local->resolver_base_inode;
+ } else {
+ gf_uuid_copy(gfid, local->base_gfid);
+ }
- shard_make_block_bname (block_num, (local->loc.inode)->gfid,
- block_bname, sizeof (block_bname));
+ shard_make_block_bname(block_num, gfid, block_bname, sizeof(block_bname));
- shard_inode_ctx_set (inode, this, buf, 0, SHARD_LOOKUP_MASK);
- linked_inode = inode_link (inode, priv->dot_shard_inode, block_bname,
- buf);
- inode_lookup (linked_inode);
- list_index = block_num - local->first_block;
- local->inode_list[list_index] = linked_inode;
+ shard_inode_ctx_set(inode, this, buf, 0, SHARD_LOOKUP_MASK);
+ linked_inode = inode_link(inode, priv->dot_shard_inode, block_bname, buf);
+ inode_lookup(linked_inode);
+ list_index = block_num - local->first_block;
+ local->inode_list[list_index] = linked_inode;
- LOCK(&priv->lock);
- {
- fsync_inode = __shard_update_shards_inode_list (linked_inode,
- this,
- local->loc.inode,
- block_num);
- }
- UNLOCK(&priv->lock);
- if (fsync_inode)
- shard_initiate_evicted_inode_fsync (this, fsync_inode);
+ LOCK(&priv->lock);
+ {
+ fsync_inode = __shard_update_shards_inode_list(
+ linked_inode, this, base_inode, block_num, gfid);
+ }
+ UNLOCK(&priv->lock);
+ if (fsync_inode)
+ shard_initiate_evicted_inode_fsync(this, fsync_inode);
}
int
-shard_common_lookup_shards_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata,
- struct iatt *postparent)
+shard_common_lookup_shards_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret, int32_t op_errno,
+ inode_t *inode, struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
{
- int call_count = 0;
- int shard_block_num = (long) cookie;
- shard_local_t *local = NULL;
-
- local = frame->local;
-
- if (op_ret < 0) {
- /* Ignore absence of shards in the backend in truncate fop. */
- switch (local->fop) {
- case GF_FOP_TRUNCATE:
- case GF_FOP_FTRUNCATE:
- case GF_FOP_RENAME:
- case GF_FOP_UNLINK:
- if (op_errno == ENOENT)
- goto done;
- break;
- case GF_FOP_WRITE:
- case GF_FOP_READ:
- case GF_FOP_ZEROFILL:
- case GF_FOP_DISCARD:
- case GF_FOP_FALLOCATE:
- if ((!local->first_lookup_done) &&
- (op_errno == ENOENT)) {
- LOCK (&frame->lock);
- {
- local->create_count++;
- }
- UNLOCK (&frame->lock);
- goto done;
- }
- break;
- default:
- break;
+ int call_count = 0;
+ int shard_block_num = (long)cookie;
+ uuid_t gfid = {
+ 0,
+ };
+ shard_local_t *local = NULL;
+
+ local = frame->local;
+ if (local->resolver_base_inode)
+ gf_uuid_copy(gfid, local->resolver_base_inode->gfid);
+ else
+ gf_uuid_copy(gfid, local->base_gfid);
+
+ if (op_ret < 0) {
+ /* Ignore absence of shards in the backend in truncate fop. */
+ switch (local->fop) {
+ case GF_FOP_TRUNCATE:
+ case GF_FOP_FTRUNCATE:
+ case GF_FOP_RENAME:
+ case GF_FOP_UNLINK:
+ if (op_errno == ENOENT)
+ goto done;
+ break;
+ case GF_FOP_WRITE:
+ case GF_FOP_READ:
+ case GF_FOP_ZEROFILL:
+ case GF_FOP_DISCARD:
+ case GF_FOP_FALLOCATE:
+ if ((!local->first_lookup_done) && (op_errno == ENOENT)) {
+ LOCK(&frame->lock);
+ {
+ local->create_count++;
+ }
+ UNLOCK(&frame->lock);
+ goto done;
}
-
- /* else */
- gf_msg (this->name, GF_LOG_ERROR, op_errno,
- SHARD_MSG_LOOKUP_SHARD_FAILED, "Lookup on shard %d "
- "failed. Base file gfid = %s", shard_block_num,
- (local->fop == GF_FOP_RENAME) ?
- uuid_utoa (local->loc2.inode->gfid)
- : uuid_utoa (local->loc.inode->gfid));
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto done;
+ break;
+ default:
+ break;
}
- shard_link_block_inode (local, shard_block_num, inode, buf);
+ /* else */
+ gf_msg(this->name, GF_LOG_ERROR, op_errno,
+ SHARD_MSG_LOOKUP_SHARD_FAILED,
+ "Lookup on shard %d "
+ "failed. Base file gfid = %s",
+ shard_block_num, uuid_utoa(gfid));
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto done;
+ }
-done:
- call_count = shard_call_count_return (frame);
- if (local->lookup_shards_barriered) {
- syncbarrier_wake (&local->barrier);
- return 0;
- } else {
- if (call_count == 0) {
- if (!local->first_lookup_done)
- local->first_lookup_done = _gf_true;
- if (local->op_ret < 0)
- goto unwind;
- else
- local->pls_fop_handler (frame, this);
- }
- }
- return 0;
+ shard_link_block_inode(local, shard_block_num, inode, buf);
-unwind:
- local->pls_fop_handler (frame, this);
+done:
+ if (local->lookup_shards_barriered) {
+ syncbarrier_wake(&local->barrier);
return 0;
+ } else {
+ call_count = shard_call_count_return(frame);
+ if (call_count == 0) {
+ if (!local->first_lookup_done)
+ local->first_lookup_done = _gf_true;
+ local->pls_fop_handler(frame, this);
+ }
+ }
+ return 0;
}
-dict_t*
-shard_create_gfid_dict (dict_t *dict)
+dict_t *
+shard_create_gfid_dict(dict_t *dict)
{
- int ret = 0;
- dict_t *new = NULL;
- unsigned char *gfid = NULL;
+ int ret = 0;
+ dict_t *new = NULL;
+ unsigned char *gfid = NULL;
- new = dict_copy_with_ref (dict, NULL);
- if (!new)
- return NULL;
+ new = dict_copy_with_ref(dict, NULL);
+ if (!new)
+ return NULL;
- gfid = GF_CALLOC (1, sizeof (uuid_t), gf_common_mt_char);
- if (!gfid) {
- ret = -1;
- goto out;
- }
+ gfid = GF_MALLOC(sizeof(uuid_t), gf_common_mt_char);
+ if (!gfid) {
+ ret = -1;
+ goto out;
+ }
- gf_uuid_generate (gfid);
+ gf_uuid_generate(gfid);
- ret = dict_set_gfuuid (new, "gfid-req", gfid, false);
+ ret = dict_set_gfuuid(new, "gfid-req", gfid, false);
out:
- if (ret) {
- dict_unref (new);
- new = NULL;
- GF_FREE (gfid);
- }
-
- return new;
-}
-
-int
-shard_common_lookup_shards (call_frame_t *frame, xlator_t *this, inode_t *inode,
- shard_post_lookup_shards_fop_handler_t handler)
-{
- int i = 0;
- int ret = 0;
- int count = 0;
- int call_count = 0;
- int32_t shard_idx_iter = 0;
- int last_block = 0;
- char path[PATH_MAX] = {0,};
- char *bname = NULL;
- loc_t loc = {0,};
- shard_local_t *local = NULL;
- shard_priv_t *priv = NULL;
- gf_boolean_t wind_failed = _gf_false;
- dict_t *xattr_req = NULL;
-
- priv = this->private;
- local = frame->local;
- count = call_count = local->call_count;
- shard_idx_iter = local->first_block;
- last_block = local->last_block;
- local->pls_fop_handler = handler;
- if (local->lookup_shards_barriered)
- local->barrier.waitfor = local->call_count;
-
- while (shard_idx_iter <= last_block) {
- if (local->inode_list[i]) {
- i++;
- shard_idx_iter++;
- continue;
- }
-
- if (wind_failed) {
- shard_common_lookup_shards_cbk (frame,
- (void *) (long) shard_idx_iter,
- this, -1, ENOMEM, NULL, NULL,
- NULL, NULL);
- goto next;
- }
+ if (ret) {
+ dict_unref(new);
+ new = NULL;
+ GF_FREE(gfid);
+ }
+
+ return new;
+}
+
+int
+shard_common_lookup_shards(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ shard_post_lookup_shards_fop_handler_t handler)
+{
+ int i = 0;
+ int ret = 0;
+ int count = 0;
+ int call_count = 0;
+ int32_t shard_idx_iter = 0;
+ int lookup_count = 0;
+ char path[PATH_MAX] = {
+ 0,
+ };
+ char *bname = NULL;
+ uuid_t gfid = {
+ 0,
+ };
+ loc_t loc = {
+ 0,
+ };
+ shard_local_t *local = NULL;
+ shard_priv_t *priv = NULL;
+ gf_boolean_t wind_failed = _gf_false;
+ dict_t *xattr_req = NULL;
+
+ priv = this->private;
+ local = frame->local;
+ count = call_count = local->call_count;
+ shard_idx_iter = local->first_block;
+ lookup_count = local->last_block - local->create_count;
+ local->pls_fop_handler = handler;
+ if (local->lookup_shards_barriered)
+ local->barrier.waitfor = local->call_count;
+
+ if (inode)
+ gf_uuid_copy(gfid, inode->gfid);
+ else
+ gf_uuid_copy(gfid, local->base_gfid);
+
+ while (shard_idx_iter <= lookup_count) {
+ if (local->inode_list[i]) {
+ i++;
+ shard_idx_iter++;
+ continue;
+ }
+
+ if (wind_failed) {
+ shard_common_lookup_shards_cbk(frame, (void *)(long)shard_idx_iter,
+ this, -1, ENOMEM, NULL, NULL, NULL,
+ NULL);
+ goto next;
+ }
+
+ shard_make_block_abspath(shard_idx_iter, gfid, path, sizeof(path));
+
+ bname = strrchr(path, '/') + 1;
+ loc.inode = inode_new(this->itable);
+ loc.parent = inode_ref(priv->dot_shard_inode);
+ gf_uuid_copy(loc.pargfid, priv->dot_shard_gfid);
+ ret = inode_path(loc.parent, bname, (char **)&(loc.path));
+ if (ret < 0 || !(loc.inode)) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_PATH_FAILED,
+ "Inode path failed"
+ " on %s, base file gfid = %s",
+ bname, uuid_utoa(gfid));
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ loc_wipe(&loc);
+ wind_failed = _gf_true;
+ shard_common_lookup_shards_cbk(frame, (void *)(long)shard_idx_iter,
+ this, -1, ENOMEM, NULL, NULL, NULL,
+ NULL);
+ goto next;
+ }
+
+ loc.name = strrchr(loc.path, '/');
+ if (loc.name)
+ loc.name++;
+
+ xattr_req = shard_create_gfid_dict(local->xattr_req);
+ if (!xattr_req) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ wind_failed = _gf_true;
+ loc_wipe(&loc);
+ shard_common_lookup_shards_cbk(frame, (void *)(long)shard_idx_iter,
+ this, -1, ENOMEM, NULL, NULL, NULL,
+ NULL);
+ goto next;
+ }
+
+ STACK_WIND_COOKIE(frame, shard_common_lookup_shards_cbk,
+ (void *)(long)shard_idx_iter, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, &loc, xattr_req);
+ loc_wipe(&loc);
+ dict_unref(xattr_req);
+ next:
+ shard_idx_iter++;
+ i++;
+
+ if (!--call_count)
+ break;
+ }
+ if (local->lookup_shards_barriered) {
+ syncbarrier_wait(&local->barrier, count);
+ local->pls_fop_handler(frame, this);
+ }
+ return 0;
+}
+
+int
+shard_post_resolve_truncate_handler(call_frame_t *frame, xlator_t *this)
+{
+ shard_local_t *local = NULL;
+
+ local = frame->local;
+
+ if (local->op_ret < 0) {
+ if (local->op_errno == ENOENT) {
+ /* If lookup on /.shard fails with ENOENT, it means that
+ * the file was 0-byte in size but truncated sometime in
+ * the past to a higher size which is reflected in the
+ * size xattr, and now being truncated to a lower size.
+ * In this case, the only thing that needs to be done is
+ * to update the size xattr of the file and unwind.
+ */
+ local->first_block = local->last_block = 0;
+ local->num_blocks = 1;
+ local->call_count = 0;
+ local->op_ret = 0;
+ local->postbuf.ia_size = local->offset;
+ shard_update_file_size(frame, this, local->fd, &local->loc,
+ shard_post_update_size_truncate_handler);
+ return 0;
+ } else {
+ shard_common_failure_unwind(local->fop, frame, local->op_ret,
+ local->op_errno);
+ return 0;
+ }
+ }
+
+ if (!local->call_count)
+ shard_truncate_do(frame, this);
+ else
+ shard_common_lookup_shards(frame, this, local->loc.inode,
+ shard_post_lookup_shards_truncate_handler);
+
+ return 0;
+}
+
+int
+shard_truncate_begin(call_frame_t *frame, xlator_t *this)
+{
+ int ret = 0;
+ shard_local_t *local = NULL;
+ shard_priv_t *priv = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ /* First participant block here is the lowest numbered block that would
+ * hold the last byte of the file post successful truncation.
+ * Last participant block is the block that contains the last byte in
+ * the current state of the file.
+ * If (first block == last_block):
+ * then that means that the file only needs truncation of the
+ * first (or last since both are same) block.
+ * Else
+ * if (new_size % block_size == 0)
+ * then that means there is no truncate to be done with
+ * only shards from first_block + 1 through the last
+ * block needing to be unlinked.
+ * else
+ * both truncate of the first block and unlink of the
+ * remaining shards until end of file is required.
+ */
+ local->first_block = (local->offset == 0)
+ ? 0
+ : get_lowest_block(local->offset - 1,
+ local->block_size);
+ local->last_block = get_highest_block(0, local->prebuf.ia_size,
+ local->block_size);
+
+ local->num_blocks = local->last_block - local->first_block + 1;
+ GF_ASSERT(local->num_blocks > 0);
+ local->resolver_base_inode = (local->fop == GF_FOP_TRUNCATE)
+ ? local->loc.inode
+ : local->fd->inode;
+
+ if ((local->first_block == 0) && (local->num_blocks == 1)) {
+ if (local->fop == GF_FOP_TRUNCATE)
+ STACK_WIND(frame, shard_truncate_last_shard_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, &local->loc,
+ local->offset, local->xattr_req);
+ else
+ STACK_WIND(frame, shard_truncate_last_shard_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, local->fd,
+ local->offset, local->xattr_req);
+ return 0;
+ }
- shard_make_block_abspath (shard_idx_iter, inode->gfid, path,
- sizeof(path));
-
- bname = strrchr (path, '/') + 1;
- loc.inode = inode_new (this->itable);
- loc.parent = inode_ref (priv->dot_shard_inode);
- gf_uuid_copy (loc.pargfid, priv->dot_shard_gfid);
- ret = inode_path (loc.parent, bname, (char **) &(loc.path));
- if (ret < 0 || !(loc.inode)) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_PATH_FAILED, "Inode path failed"
- " on %s, base file gfid = %s", bname,
- uuid_utoa (inode->gfid));
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- loc_wipe (&loc);
- wind_failed = _gf_true;
- shard_common_lookup_shards_cbk (frame,
- (void *) (long) shard_idx_iter,
- this, -1, ENOMEM, NULL, NULL,
- NULL, NULL);
- goto next;
- }
+ local->inode_list = GF_CALLOC(local->num_blocks, sizeof(inode_t *),
+ gf_shard_mt_inode_list);
+ if (!local->inode_list)
+ goto err;
- loc.name = strrchr (loc.path, '/');
- if (loc.name)
- loc.name++;
-
- xattr_req = shard_create_gfid_dict (local->xattr_req);
- if (!xattr_req) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- wind_failed = _gf_true;
- loc_wipe (&loc);
- shard_common_lookup_shards_cbk (frame,
- (void *) (long) shard_idx_iter,
- this, -1, ENOMEM, NULL, NULL,
- NULL, NULL);
- goto next;
- }
+ local->dot_shard_loc.inode = inode_find(this->itable, priv->dot_shard_gfid);
+ if (!local->dot_shard_loc.inode) {
+ ret = shard_init_internal_dir_loc(this, local,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
+ if (ret)
+ goto err;
+ shard_lookup_internal_dir(frame, this,
+ shard_post_resolve_truncate_handler,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
+ } else {
+ local->post_res_handler = shard_post_resolve_truncate_handler;
+ shard_refresh_internal_dir(frame, this, SHARD_INTERNAL_DIR_DOT_SHARD);
+ }
+ return 0;
- STACK_WIND_COOKIE (frame, shard_common_lookup_shards_cbk,
- (void *) (long) shard_idx_iter,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->lookup, &loc,
- xattr_req);
- loc_wipe (&loc);
- dict_unref (xattr_req);
-next:
- shard_idx_iter++;
- i++;
-
- if (!--call_count)
- break;
- }
- if (local->lookup_shards_barriered)
- syncbarrier_wait (&local->barrier, count);
- return 0;
+err:
+ shard_common_failure_unwind(local->fop, frame, -1, ENOMEM);
+ return 0;
}
int
-shard_post_resolve_truncate_handler (call_frame_t *frame, xlator_t *this)
+shard_post_lookup_truncate_handler(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
-
- local = frame->local;
-
- if (local->op_ret < 0) {
- if (local->op_errno == ENOENT) {
- /* If lookup on /.shard fails with ENOENT, it means that
- * the file was 0-byte in size but truncated sometime in
- * the past to a higher size which is reflected in the
- * size xattr, and now being truncated to a lower size.
- * In this case, the only thing that needs to be done is
- * to update the size xattr of the file and unwind.
- */
- local->first_block = local->last_block = 0;
- local->num_blocks = 1;
- local->call_count = 0;
- local->op_ret = 0;
- local->postbuf.ia_size = local->offset;
- shard_update_file_size (frame, this, local->fd,
- &local->loc,
- shard_post_update_size_truncate_handler);
- return 0;
- } else {
- if (local->fop == GF_FOP_TRUNCATE)
- SHARD_STACK_UNWIND (truncate, frame,
- local->op_ret,
- local->op_errno, NULL, NULL,
- NULL);
- else
- SHARD_STACK_UNWIND (ftruncate, frame,
- local->op_ret,
- local->op_errno, NULL, NULL,
- NULL);
- return 0;
- }
- }
+ shard_local_t *local = NULL;
+ struct iatt tmp_stbuf = {
+ 0,
+ };
- if (!local->call_count)
- shard_truncate_do (frame, this);
- else
- shard_common_lookup_shards (frame, this, local->loc.inode,
- shard_post_lookup_shards_truncate_handler);
+ local = frame->local;
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(local->fop, frame, local->op_ret,
+ local->op_errno);
return 0;
-}
+ }
-int
-shard_truncate_begin (call_frame_t *frame, xlator_t *this)
-{
- int ret = 0;
- shard_local_t *local = NULL;
- shard_priv_t *priv = NULL;
-
- priv = this->private;
- local = frame->local;
+ local->postbuf = tmp_stbuf = local->prebuf;
- /* First participant block here is the lowest numbered block that would
- * hold the last byte of the file post successful truncation.
- * Last participant block is the block that contains the last byte in
- * the current state of the file.
- * If (first block == last_block):
- * then that means that the file only needs truncation of the
- * first (or last since both are same) block.
- * Else
- * if (new_size % block_size == 0)
- * then that means there is no truncate to be done with
- * only shards from first_block + 1 through the last
- * block needing to be unlinked.
- * else
- * both truncate of the first block and unlink of the
- * remaining shards until end of file is required.
+ if (local->prebuf.ia_size == local->offset) {
+ /* If the file size is same as requested size, unwind the call
+ * immediately.
*/
- local->first_block = (local->offset == 0) ? 0
- : get_lowest_block (local->offset - 1,
- local->block_size);
- local->last_block = get_highest_block (0, local->prebuf.ia_size,
- local->block_size);
-
- local->num_blocks = local->last_block - local->first_block + 1;
- local->resolver_base_inode = (local->fop == GF_FOP_TRUNCATE) ?
- local->loc.inode : local->fd->inode;
-
- if ((local->first_block == 0) && (local->num_blocks == 1)) {
- if (local->fop == GF_FOP_TRUNCATE)
- STACK_WIND (frame, shard_truncate_last_shard_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->truncate,
- &local->loc, local->offset,
- local->xattr_req);
- else
- STACK_WIND (frame, shard_truncate_last_shard_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->ftruncate,
- local->fd, local->offset, local->xattr_req);
- return 0;
- }
-
- local->inode_list = GF_CALLOC (local->num_blocks, sizeof (inode_t *),
- gf_shard_mt_inode_list);
- if (!local->inode_list)
- goto err;
-
- local->dot_shard_loc.inode = inode_find (this->itable,
- priv->dot_shard_gfid);
- if (!local->dot_shard_loc.inode) {
- ret = shard_init_internal_dir_loc (this, local,
- SHARD_INTERNAL_DIR_DOT_SHARD);
- if (ret)
- goto err;
- shard_lookup_internal_dir (frame, this,
- shard_post_resolve_truncate_handler,
- SHARD_INTERNAL_DIR_DOT_SHARD);
- } else {
- local->post_res_handler = shard_post_resolve_truncate_handler;
- shard_refresh_internal_dir (frame, this,
- SHARD_INTERNAL_DIR_DOT_SHARD);
- }
- return 0;
-
-err:
if (local->fop == GF_FOP_TRUNCATE)
- SHARD_STACK_UNWIND (truncate, frame, -1, ENOMEM, NULL, NULL,
- NULL);
+ SHARD_STACK_UNWIND(truncate, frame, 0, 0, &local->prebuf,
+ &local->postbuf, NULL);
else
- SHARD_STACK_UNWIND (ftruncate, frame, -1, ENOMEM, NULL, NULL,
- NULL);
-
- return 0;
-}
-
-int
-shard_post_lookup_truncate_handler (call_frame_t *frame, xlator_t *this)
-{
- shard_local_t *local = NULL;
- struct iatt tmp_stbuf = {0,};
-
- local = frame->local;
-
- if (local->op_ret < 0) {
- if (local->fop == GF_FOP_TRUNCATE)
- SHARD_STACK_UNWIND (truncate, frame, local->op_ret,
- local->op_errno, NULL, NULL, NULL);
- else
- SHARD_STACK_UNWIND (ftruncate, frame, local->op_ret,
- local->op_errno, NULL, NULL, NULL);
-
- return 0;
- }
-
- local->postbuf = tmp_stbuf = local->prebuf;
-
- if (local->prebuf.ia_size == local->offset) {
- /* If the file size is same as requested size, unwind the call
- * immediately.
- */
- if (local->fop == GF_FOP_TRUNCATE)
- SHARD_STACK_UNWIND (truncate, frame, 0, 0,
- &local->prebuf, &local->postbuf,
- NULL);
- else
- SHARD_STACK_UNWIND (ftruncate, frame, 0, 0,
- &local->prebuf, &local->postbuf,
- NULL);
- } else if (local->offset > local->prebuf.ia_size) {
- /* If the truncate is from a lower to a higher size, set the
- * new size xattr and unwind.
- */
- local->hole_size = local->offset - local->prebuf.ia_size;
- local->delta_size = 0;
- local->delta_blocks = 0;
- local->postbuf.ia_size = local->offset;
- tmp_stbuf.ia_size = local->offset;
- shard_inode_ctx_set (local->loc.inode, this, &tmp_stbuf, 0,
- SHARD_INODE_WRITE_MASK);
- shard_update_file_size (frame, this, NULL, &local->loc,
- shard_post_update_size_truncate_handler);
- } else {
- /* ... else
- * i. unlink all shards that need to be unlinked.
- * ii. truncate the last of the shards.
- * iii. update the new size using setxattr.
- * and unwind the fop.
- */
- local->hole_size = 0;
- local->delta_size = (local->offset - local->prebuf.ia_size);
- local->delta_blocks = 0;
- tmp_stbuf.ia_size = local->offset;
- shard_inode_ctx_set (local->loc.inode, this, &tmp_stbuf, 0,
- SHARD_INODE_WRITE_MASK);
- shard_truncate_begin (frame, this);
- }
- return 0;
+ SHARD_STACK_UNWIND(ftruncate, frame, 0, 0, &local->prebuf,
+ &local->postbuf, NULL);
+ } else if (local->offset > local->prebuf.ia_size) {
+ /* If the truncate is from a lower to a higher size, set the
+ * new size xattr and unwind.
+ */
+ local->hole_size = local->offset - local->prebuf.ia_size;
+ local->delta_size = 0;
+ GF_ATOMIC_INIT(local->delta_blocks, 0);
+ local->postbuf.ia_size = local->offset;
+ tmp_stbuf.ia_size = local->offset;
+ shard_inode_ctx_set(local->loc.inode, this, &tmp_stbuf, 0,
+ SHARD_INODE_WRITE_MASK);
+ shard_update_file_size(frame, this, NULL, &local->loc,
+ shard_post_update_size_truncate_handler);
+ } else {
+ /* ... else
+ * i. unlink all shards that need to be unlinked.
+ * ii. truncate the last of the shards.
+ * iii. update the new size using setxattr.
+ * and unwind the fop.
+ */
+ local->hole_size = 0;
+ local->delta_size = (local->offset - local->prebuf.ia_size);
+ GF_ATOMIC_INIT(local->delta_blocks, 0);
+ tmp_stbuf.ia_size = local->offset;
+ shard_inode_ctx_set(local->loc.inode, this, &tmp_stbuf, 0,
+ SHARD_INODE_WRITE_MASK);
+ shard_truncate_begin(frame, this);
+ }
+ return 0;
}
/* TO-DO:
@@ -2409,3390 +2754,4629 @@ shard_post_lookup_truncate_handler (call_frame_t *frame, xlator_t *this)
*/
int
-shard_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
- dict_t *xdata)
+shard_truncate(call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+ dict_t *xdata)
{
- int ret = -1;
- uint64_t block_size = 0;
- shard_local_t *local = NULL;
+ int ret = -1;
+ uint64_t block_size = 0;
+ shard_local_t *local = NULL;
- ret = shard_inode_ctx_get_block_size (loc->inode, this, &block_size);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_CTX_GET_FAILED, "Failed to get block "
- "size from inode ctx of %s",
- uuid_utoa (loc->inode->gfid));
- goto err;
- }
+ ret = shard_inode_ctx_get_block_size(loc->inode, this, &block_size);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_CTX_GET_FAILED,
+ "Failed to get block "
+ "size from inode ctx of %s",
+ uuid_utoa(loc->inode->gfid));
+ goto err;
+ }
- if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
- STACK_WIND (frame, default_truncate_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->truncate, loc, offset,
- xdata);
- return 0;
- }
+ if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
+ STACK_WIND(frame, default_truncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
+ return 0;
+ }
- if (!this->itable)
- this->itable = loc->inode->table;
+ if (!this->itable)
+ this->itable = loc->inode->table;
- local = mem_get0 (this->local_pool);
- if (!local)
- goto err;
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
- frame->local = local;
+ frame->local = local;
- ret = syncbarrier_init (&local->barrier);
- if (ret)
- goto err;
- loc_copy (&local->loc, loc);
- local->offset = offset;
- local->block_size = block_size;
- local->fop = GF_FOP_TRUNCATE;
- local->xattr_req = (xdata) ? dict_ref (xdata) : dict_new ();
- if (!local->xattr_req)
- goto err;
- local->resolver_base_inode = loc->inode;
-
- shard_lookup_base_file (frame, this, &local->loc,
- shard_post_lookup_truncate_handler);
- return 0;
+ ret = syncbarrier_init(&local->barrier);
+ if (ret)
+ goto err;
+ loc_copy(&local->loc, loc);
+ local->offset = offset;
+ local->block_size = block_size;
+ local->fop = GF_FOP_TRUNCATE;
+ local->xattr_req = (xdata) ? dict_ref(xdata) : dict_new();
+ if (!local->xattr_req)
+ goto err;
+ local->resolver_base_inode = loc->inode;
+ GF_ATOMIC_INIT(local->delta_blocks, 0);
+
+ shard_refresh_base_file(frame, this, &local->loc, NULL,
+ shard_post_lookup_truncate_handler);
+ return 0;
err:
- SHARD_STACK_UNWIND (truncate, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ shard_common_failure_unwind(GF_FOP_TRUNCATE, frame, -1, ENOMEM);
+ return 0;
}
int
-shard_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- dict_t *xdata)
+shard_ftruncate(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ dict_t *xdata)
{
- int ret = -1;
- uint64_t block_size = 0;
- shard_local_t *local = NULL;
-
- ret = shard_inode_ctx_get_block_size (fd->inode, this, &block_size);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_CTX_GET_FAILED, "Failed to get block "
- "size from inode ctx of %s",
- uuid_utoa (fd->inode->gfid));
- goto err;
- }
-
- if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
- STACK_WIND (frame, default_ftruncate_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->ftruncate, fd, offset,
- xdata);
- return 0;
- }
-
- if (!this->itable)
- this->itable = fd->inode->table;
-
- local = mem_get0 (this->local_pool);
- if (!local)
- goto err;
-
- frame->local = local;
- ret = syncbarrier_init (&local->barrier);
- if (ret)
- goto err;
- local->fd = fd_ref (fd);
- local->offset = offset;
- local->block_size = block_size;
- local->xattr_req = (xdata) ? dict_ref (xdata) : dict_new ();
- if (!local->xattr_req)
- goto err;
- local->fop = GF_FOP_FTRUNCATE;
-
- local->loc.inode = inode_ref (fd->inode);
- gf_uuid_copy (local->loc.gfid, fd->inode->gfid);
- local->resolver_base_inode = fd->inode;
-
- shard_lookup_base_file (frame, this, &local->loc,
- shard_post_lookup_truncate_handler);
- return 0;
+ int ret = -1;
+ uint64_t block_size = 0;
+ shard_local_t *local = NULL;
+
+ ret = shard_inode_ctx_get_block_size(fd->inode, this, &block_size);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_CTX_GET_FAILED,
+ "Failed to get block "
+ "size from inode ctx of %s",
+ uuid_utoa(fd->inode->gfid));
+ goto err;
+ }
+
+ if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
+ STACK_WIND(frame, default_ftruncate_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+ return 0;
+ }
+
+ if (!this->itable)
+ this->itable = fd->inode->table;
+
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
+
+ frame->local = local;
+ ret = syncbarrier_init(&local->barrier);
+ if (ret)
+ goto err;
+ local->fd = fd_ref(fd);
+ local->offset = offset;
+ local->block_size = block_size;
+ local->xattr_req = (xdata) ? dict_ref(xdata) : dict_new();
+ if (!local->xattr_req)
+ goto err;
+ local->fop = GF_FOP_FTRUNCATE;
+
+ local->loc.inode = inode_ref(fd->inode);
+ gf_uuid_copy(local->loc.gfid, fd->inode->gfid);
+ local->resolver_base_inode = fd->inode;
+ GF_ATOMIC_INIT(local->delta_blocks, 0);
+
+ shard_refresh_base_file(frame, this, NULL, fd,
+ shard_post_lookup_truncate_handler);
+ return 0;
err:
-
- SHARD_STACK_UNWIND (ftruncate, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ shard_common_failure_unwind(GF_FOP_FTRUNCATE, frame, -1, ENOMEM);
+ return 0;
}
int
-shard_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+shard_mknod_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- int ret = -1;
- shard_local_t *local = NULL;
+ int ret = -1;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret == -1)
- goto unwind;
+ if (op_ret == -1)
+ goto unwind;
- ret = shard_inode_ctx_set (inode, this, buf, ntoh64 (local->block_size),
- SHARD_ALL_MASK);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- SHARD_MSG_INODE_CTX_SET_FAILED, "Failed to set inode "
- "ctx for %s", uuid_utoa (inode->gfid));
+ ret = shard_inode_ctx_set(inode, this, buf, local->block_size,
+ SHARD_ALL_MASK);
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_INODE_CTX_SET_FAILED,
+ "Failed to set inode "
+ "ctx for %s",
+ uuid_utoa(inode->gfid));
unwind:
- SHARD_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
+ SHARD_STACK_UNWIND(mknod, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
- return 0;
+ return 0;
}
int
-shard_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- dev_t rdev, mode_t umask, dict_t *xdata)
+shard_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *xdata)
{
- shard_local_t *local = NULL;
-
- local = mem_get0 (this->local_pool);
- if (!local)
- goto err;
+ shard_priv_t *priv = NULL;
+ shard_local_t *local = NULL;
- frame->local = local;
- if (!__is_gsyncd_on_shard_dir (frame, loc)) {
- SHARD_INODE_CREATE_INIT (this, local, xdata, loc, err);
- }
+ priv = this->private;
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
- STACK_WIND (frame, shard_mknod_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->mknod, loc, mode, rdev, umask,
- xdata);
- return 0;
+ frame->local = local;
+ local->block_size = priv->block_size;
+ if (!__is_gsyncd_on_shard_dir(frame, loc)) {
+ SHARD_INODE_CREATE_INIT(this, local->block_size, xdata, loc, 0, 0, err);
+ }
+ STACK_WIND(frame, shard_mknod_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, loc, mode, rdev, umask, xdata);
+ return 0;
err:
- SHARD_STACK_UNWIND (mknod, frame, -1, ENOMEM, NULL, NULL, NULL,
- NULL, NULL);
- return 0;
-
+ shard_common_failure_unwind(GF_FOP_MKNOD, frame, -1, ENOMEM);
+ return 0;
}
int32_t
-shard_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent,
- dict_t *xdata)
+shard_link_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
- if (op_ret < 0)
- goto err;
+ local = frame->local;
+ if (op_ret < 0)
+ goto err;
- shard_inode_ctx_set (inode, this, buf, 0,
- SHARD_MASK_NLINK | SHARD_MASK_TIMES);
- buf->ia_size = local->prebuf.ia_size;
- buf->ia_blocks = local->prebuf.ia_blocks;
+ shard_inode_ctx_set(inode, this, buf, 0,
+ SHARD_MASK_NLINK | SHARD_MASK_TIMES);
+ buf->ia_size = local->prebuf.ia_size;
+ buf->ia_blocks = local->prebuf.ia_blocks;
- SHARD_STACK_UNWIND (link, frame, op_ret, op_errno, inode, buf,
- preparent, postparent, xdata);
- return 0;
+ SHARD_STACK_UNWIND(link, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
+ return 0;
err:
- SHARD_STACK_UNWIND (link, frame, op_ret, op_errno, inode, NULL, NULL,
- NULL, NULL);
- return 0;
+ shard_common_failure_unwind(GF_FOP_LINK, frame, op_ret, op_errno);
+ return 0;
}
int
-shard_post_lookup_link_handler (call_frame_t *frame, xlator_t *this)
+shard_post_lookup_link_handler(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->op_ret < 0) {
- SHARD_STACK_UNWIND (link, frame, local->op_ret, local->op_errno,
- NULL, NULL, NULL, NULL, NULL);
- return 0;
- }
-
- STACK_WIND (frame, shard_link_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->link, &local->loc, &local->loc2,
- local->xattr_req);
+ if (local->op_ret < 0) {
+ SHARD_STACK_UNWIND(link, frame, local->op_ret, local->op_errno, NULL,
+ NULL, NULL, NULL, NULL);
return 0;
+ }
+
+ STACK_WIND(frame, shard_link_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->link, &local->loc, &local->loc2,
+ local->xattr_req);
+ return 0;
}
int32_t
-shard_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
-{
- int ret = -1;
- uint64_t block_size = 0;
- shard_local_t *local = NULL;
-
- ret = shard_inode_ctx_get_block_size (oldloc->inode, this, &block_size);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_CTX_GET_FAILED, "Failed to get block "
- "size from inode ctx of %s",
- uuid_utoa (oldloc->inode->gfid));
- goto err;
- }
-
- if (!block_size) {
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->link, oldloc, newloc,
- xdata);
- return 0;
- }
+shard_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
+{
+ int ret = -1;
+ uint64_t block_size = 0;
+ shard_local_t *local = NULL;
+
+ ret = shard_inode_ctx_get_block_size(oldloc->inode, this, &block_size);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_CTX_GET_FAILED,
+ "Failed to get block "
+ "size from inode ctx of %s",
+ uuid_utoa(oldloc->inode->gfid));
+ goto err;
+ }
+
+ if (!block_size) {
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this), FIRST_CHILD(this)->fops->link,
+ oldloc, newloc, xdata);
+ return 0;
+ }
+
+ if (!this->itable)
+ this->itable = oldloc->inode->table;
+
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
+
+ frame->local = local;
+
+ loc_copy(&local->loc, oldloc);
+ loc_copy(&local->loc2, newloc);
+ local->xattr_req = (xdata) ? dict_ref(xdata) : dict_new();
+ if (!local->xattr_req)
+ goto err;
+
+ shard_refresh_base_file(frame, this, &local->loc, NULL,
+ shard_post_lookup_link_handler);
+ return 0;
+err:
+ shard_common_failure_unwind(GF_FOP_LINK, frame, -1, ENOMEM);
+ return 0;
+}
- if (!this->itable)
- this->itable = oldloc->inode->table;
+int
+shard_unlink_shards_do(call_frame_t *frame, xlator_t *this, inode_t *inode);
- local = mem_get0 (this->local_pool);
- if (!local)
- goto err;
+int
+shard_post_lookup_shards_unlink_handler(call_frame_t *frame, xlator_t *this)
+{
+ shard_local_t *local = NULL;
+ uuid_t gfid = {
+ 0,
+ };
- frame->local = local;
+ local = frame->local;
- loc_copy (&local->loc, oldloc);
- loc_copy (&local->loc2, newloc);
- local->xattr_req = (xdata) ? dict_ref (xdata) : dict_new ();
- if (!local->xattr_req)
- goto err;
+ if (local->resolver_base_inode)
+ gf_uuid_copy(gfid, local->resolver_base_inode->gfid);
+ else
+ gf_uuid_copy(gfid, local->base_gfid);
- shard_lookup_base_file (frame, this, &local->loc,
- shard_post_lookup_link_handler);
+ if ((local->op_ret < 0) && (local->op_errno != ENOENT)) {
+ gf_msg(this->name, GF_LOG_ERROR, local->op_errno, SHARD_MSG_FOP_FAILED,
+ "failed to delete shards of %s", uuid_utoa(gfid));
return 0;
+ }
+ local->op_ret = 0;
+ local->op_errno = 0;
-err:
- SHARD_STACK_UNWIND (link, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
- NULL);
- return 0;
+ shard_unlink_shards_do(frame, this, local->resolver_base_inode);
+ return 0;
}
int
-shard_unlink_shards_do (call_frame_t *frame, xlator_t *this, inode_t *inode);
-
-int
-shard_post_lookup_shards_unlink_handler (call_frame_t *frame, xlator_t *this)
+shard_post_resolve_unlink_handler(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
+ local->lookup_shards_barriered = _gf_true;
- if ((local->op_ret < 0) && (local->op_errno != ENOENT)) {
- if (local->fop == GF_FOP_UNLINK)
- SHARD_STACK_UNWIND (unlink, frame, local->op_ret,
- local->op_errno, NULL, NULL, NULL);
- else
- SHARD_STACK_UNWIND (rename, frame, local->op_ret,
- local->op_errno, NULL, NULL, NULL,
- NULL, NULL, NULL);
- return 0;
- }
- local->op_ret = 0;
- local->op_errno = 0;
-
- shard_unlink_shards_do (frame, this, local->resolver_base_inode);
- return 0;
+ if (!local->call_count)
+ shard_unlink_shards_do(frame, this, local->resolver_base_inode);
+ else
+ shard_common_lookup_shards(frame, this, local->resolver_base_inode,
+ shard_post_lookup_shards_unlink_handler);
+ return 0;
}
-int
-shard_rename_cbk (call_frame_t *frame, xlator_t *this);
+void
+shard_unlink_block_inode(shard_local_t *local, int shard_block_num)
+{
+ char block_bname[256] = {
+ 0,
+ };
+ uuid_t gfid = {
+ 0,
+ };
+ inode_t *inode = NULL;
+ inode_t *base_inode = NULL;
+ xlator_t *this = NULL;
+ shard_priv_t *priv = NULL;
+ shard_inode_ctx_t *ctx = NULL;
+ shard_inode_ctx_t *base_ictx = NULL;
+ int unref_base_inode = 0;
+ int unref_shard_inode = 0;
+
+ this = THIS;
+ priv = this->private;
+
+ inode = local->inode_list[shard_block_num - local->first_block];
+ shard_inode_ctx_get(inode, this, &ctx);
+ base_inode = ctx->base_inode;
+ if (base_inode)
+ gf_uuid_copy(gfid, base_inode->gfid);
+ else
+ gf_uuid_copy(gfid, ctx->base_gfid);
+ shard_make_block_bname(shard_block_num, gfid, block_bname,
+ sizeof(block_bname));
+
+ LOCK(&priv->lock);
+ if (base_inode)
+ LOCK(&base_inode->lock);
+ LOCK(&inode->lock);
+ {
+ __shard_inode_ctx_get(inode, this, &ctx);
+ if (!list_empty(&ctx->ilist)) {
+ list_del_init(&ctx->ilist);
+ priv->inode_count--;
+ unref_base_inode++;
+ unref_shard_inode++;
+ GF_ASSERT(priv->inode_count >= 0);
+ }
+ if (ctx->fsync_needed) {
+ unref_base_inode++;
+ unref_shard_inode++;
+ list_del_init(&ctx->to_fsync_list);
+ if (base_inode) {
+ __shard_inode_ctx_get(base_inode, this, &base_ictx);
+ base_ictx->fsync_count--;
+ }
+ }
+ }
+ UNLOCK(&inode->lock);
+ if (base_inode)
+ UNLOCK(&base_inode->lock);
-int32_t
-shard_unlink_cbk (call_frame_t *frame, xlator_t *this);
+ inode_unlink(inode, priv->dot_shard_inode, block_bname);
+ inode_ref_reduce_by_n(inode, unref_shard_inode);
+ inode_forget(inode, 0);
+
+ if (base_inode && unref_base_inode)
+ inode_ref_reduce_by_n(base_inode, unref_base_inode);
+ UNLOCK(&priv->lock);
+}
int
-shard_post_resolve_unlink_handler (call_frame_t *frame, xlator_t *this)
+shard_rename_cbk(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->op_ret < 0) {
- if (local->op_errno == ENOENT) {
- /* If lookup on /.shard fails with ENOENT, it probably
- * means that the file is being unlinked before it
- * could grow beyond its first block. In this case,
- * unlink boils down to unlinking the base file and
- * unwinding the call.
- */
- local->op_ret = 0;
- local->first_block = local->last_block = 0;
- local->num_blocks = 1;
- if (local->fop == GF_FOP_UNLINK)
- shard_unlink_cbk (frame, this);
- else
- shard_rename_cbk (frame, this);
- return 0;
- } else {
- if (local->fop == GF_FOP_UNLINK)
- SHARD_STACK_UNWIND (unlink, frame,
- local->op_ret,
- local->op_errno, NULL, NULL,
- NULL);
- else
- shard_rename_cbk (frame, this);
- return 0;
- }
- }
+ SHARD_STACK_UNWIND(rename, frame, local->op_ret, local->op_errno,
+ &local->prebuf, &local->preoldparent,
+ &local->postoldparent, &local->prenewparent,
+ &local->postnewparent, local->xattr_rsp);
+ return 0;
+}
- if (!local->call_count)
- shard_unlink_shards_do (frame, this,
- local->resolver_base_inode);
- else
- shard_common_lookup_shards (frame, this,
- local->resolver_base_inode,
- shard_post_lookup_shards_unlink_handler);
- return 0;
+int32_t
+shard_unlink_cbk(call_frame_t *frame, xlator_t *this)
+{
+ shard_local_t *local = frame->local;
+
+ SHARD_STACK_UNWIND(unlink, frame, local->op_ret, local->op_errno,
+ &local->preoldparent, &local->postoldparent,
+ local->xattr_rsp);
+ return 0;
}
int
-shard_unlink_base_file_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+shard_unlink_shards_do_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- int ret = 0;
- uint32_t link_count = 0;
- shard_local_t *local = NULL;
- shard_priv_t *priv = NULL;
+ int shard_block_num = (long)cookie;
+ shard_local_t *local = NULL;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
- if (op_ret < 0) {
- SHARD_STACK_UNWIND (unlink, frame, op_ret, op_errno, NULL, NULL,
- NULL);
- return 0;
- }
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto done;
+ }
- /* Because link() does not create links for all but the
- * base shard, unlink() must delete these shards only when the
- * link count is 1. We can return safely now.
- */
- if ((xdata) && (!dict_get_uint32 (xdata, GET_LINK_COUNT, &link_count))
- && (link_count > 1))
- goto unwind;
-
- local->first_block = get_lowest_block (0, local->block_size);
- local->last_block = get_highest_block (0, local->prebuf.ia_size,
- local->block_size);
- local->num_blocks = local->last_block - local->first_block + 1;
- local->resolver_base_inode = local->loc.inode;
-
- /* num_blocks = 1 implies that the file has not crossed its
- * shard block size. So unlink boils down to unlinking just the
- * base file. We can safely return now.
+ shard_unlink_block_inode(local, shard_block_num);
+done:
+ syncbarrier_wake(&local->barrier);
+ return 0;
+}
+
+int
+shard_unlink_shards_do(call_frame_t *frame, xlator_t *this, inode_t *inode)
+{
+ int i = 0;
+ int ret = -1;
+ int count = 0;
+ uint32_t cur_block = 0;
+ uint32_t cur_block_idx = 0; /*this is idx into inode_list[] array */
+ char *bname = NULL;
+ char path[PATH_MAX] = {
+ 0,
+ };
+ uuid_t gfid = {
+ 0,
+ };
+ loc_t loc = {
+ 0,
+ };
+ gf_boolean_t wind_failed = _gf_false;
+ shard_local_t *local = NULL;
+ shard_priv_t *priv = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ if (inode)
+ gf_uuid_copy(gfid, inode->gfid);
+ else
+ gf_uuid_copy(gfid, local->base_gfid);
+
+ for (i = 0; i < local->num_blocks; i++) {
+ if (!local->inode_list[i])
+ continue;
+ count++;
+ }
+
+ if (!count) {
+ /* callcount = 0 implies that all of the shards that need to be
+ * unlinked are non-existent (in other words the file is full of
+ * holes).
*/
- if (local->num_blocks == 1)
- goto unwind;
+ gf_msg_debug(this->name, 0,
+ "All shards that need to be "
+ "unlinked are non-existent: %s",
+ uuid_utoa(gfid));
+ return 0;
+ }
- local->inode_list = GF_CALLOC (local->num_blocks, sizeof (inode_t *),
- gf_shard_mt_inode_list);
- if (!local->inode_list)
- goto unwind;
+ SHARD_SET_ROOT_FS_ID(frame, local);
+ local->barrier.waitfor = count;
+ cur_block = cur_block_idx + local->first_block;
- /* Save the xdata and preparent and postparent iatts now. This will be
- * used at the time of unwinding the call to the parent xl.
- */
- local->preoldparent = *preparent;
- local->postoldparent = *postparent;
- if (xdata)
- local->xattr_rsp = dict_ref (xdata);
+ while (cur_block_idx < local->num_blocks) {
+ if (!local->inode_list[cur_block_idx])
+ goto next;
- local->dot_shard_loc.inode = inode_find (this->itable,
- priv->dot_shard_gfid);
- if (!local->dot_shard_loc.inode) {
- ret = shard_init_internal_dir_loc (this, local,
- SHARD_INTERNAL_DIR_DOT_SHARD);
- if (ret)
- goto unwind;
- shard_lookup_internal_dir (frame, this,
- shard_post_resolve_unlink_handler,
- SHARD_INTERNAL_DIR_DOT_SHARD);
- } else {
- local->post_res_handler = shard_post_resolve_unlink_handler;
- shard_refresh_internal_dir (frame, this,
- SHARD_INTERNAL_DIR_DOT_SHARD);
+ if (wind_failed) {
+ shard_unlink_shards_do_cbk(frame, (void *)(long)cur_block, this, -1,
+ ENOMEM, NULL, NULL, NULL);
+ goto next;
}
- return 0;
+ shard_make_block_abspath(cur_block, gfid, path, sizeof(path));
+ bname = strrchr(path, '/') + 1;
+ loc.parent = inode_ref(priv->dot_shard_inode);
+ ret = inode_path(loc.parent, bname, (char **)&(loc.path));
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_PATH_FAILED,
+ "Inode path failed"
+ " on %s, base file gfid = %s",
+ bname, uuid_utoa(gfid));
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ loc_wipe(&loc);
+ wind_failed = _gf_true;
+ shard_unlink_shards_do_cbk(frame, (void *)(long)cur_block, this, -1,
+ ENOMEM, NULL, NULL, NULL);
+ goto next;
+ }
+
+ loc.name = strrchr(loc.path, '/');
+ if (loc.name)
+ loc.name++;
+ loc.inode = inode_ref(local->inode_list[cur_block_idx]);
+
+ STACK_WIND_COOKIE(frame, shard_unlink_shards_do_cbk,
+ (void *)(long)cur_block, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, &loc, local->xflag,
+ local->xattr_req);
+ loc_wipe(&loc);
+ next:
+ cur_block++;
+ cur_block_idx++;
+ }
+ syncbarrier_wait(&local->barrier, count);
+ SHARD_UNSET_ROOT_FS_ID(frame, local);
+ return 0;
+}
+
+int
+shard_regulated_shards_deletion(call_frame_t *cleanup_frame, xlator_t *this,
+ int now, int first_block, gf_dirent_t *entry)
+{
+ int i = 0;
+ int ret = 0;
+ shard_local_t *local = NULL;
+ uuid_t gfid = {
+ 0,
+ };
+
+ local = cleanup_frame->local;
+
+ local->inode_list = GF_CALLOC(now, sizeof(inode_t *),
+ gf_shard_mt_inode_list);
+ if (!local->inode_list)
+ return -ENOMEM;
+
+ local->first_block = first_block;
+ local->last_block = first_block + now - 1;
+ local->num_blocks = now;
+ gf_uuid_parse(entry->d_name, gfid);
+ gf_uuid_copy(local->base_gfid, gfid);
+ local->resolver_base_inode = inode_find(this->itable, gfid);
+ local->call_count = 0;
+ ret = syncbarrier_init(&local->barrier);
+ if (ret) {
+ GF_FREE(local->inode_list);
+ local->inode_list = NULL;
+ inode_unref(local->resolver_base_inode);
+ local->resolver_base_inode = NULL;
+ return -errno;
+ }
+ shard_common_resolve_shards(cleanup_frame, this,
+ shard_post_resolve_unlink_handler);
+
+ for (i = 0; i < local->num_blocks; i++) {
+ if (local->inode_list[i])
+ inode_unref(local->inode_list[i]);
+ }
+ GF_FREE(local->inode_list);
+ local->inode_list = NULL;
+ if (local->op_ret)
+ ret = -local->op_errno;
+ syncbarrier_destroy(&local->barrier);
+ inode_unref(local->resolver_base_inode);
+ local->resolver_base_inode = NULL;
+ STACK_RESET(cleanup_frame->root);
+ return ret;
+}
+
+int
+__shard_delete_shards_of_entry(call_frame_t *cleanup_frame, xlator_t *this,
+ gf_dirent_t *entry, inode_t *inode)
+{
+ int ret = 0;
+ int shard_count = 0;
+ int first_block = 0;
+ int now = 0;
+ uint64_t size = 0;
+ uint64_t block_size = 0;
+ uint64_t size_array[4] = {
+ 0,
+ };
+ void *bsize = NULL;
+ void *size_attr = NULL;
+ dict_t *xattr_rsp = NULL;
+ loc_t loc = {
+ 0,
+ };
+ shard_local_t *local = NULL;
+ shard_priv_t *priv = NULL;
+
+ priv = this->private;
+ local = cleanup_frame->local;
+ ret = dict_reset(local->xattr_req);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_DICT_OP_FAILED,
+ "Failed to reset dict");
+ ret = -ENOMEM;
+ goto err;
+ }
+
+ ret = dict_set_uint64(local->xattr_req, GF_XATTR_SHARD_BLOCK_SIZE, 0);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_DICT_OP_FAILED,
+ "Failed to set dict value: key:%s", GF_XATTR_SHARD_BLOCK_SIZE);
+ ret = -ENOMEM;
+ goto err;
+ }
+
+ ret = dict_set_uint64(local->xattr_req, GF_XATTR_SHARD_FILE_SIZE, 8 * 4);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_DICT_OP_FAILED,
+ "Failed to set dict value: key:%s", GF_XATTR_SHARD_FILE_SIZE);
+ ret = -ENOMEM;
+ goto err;
+ }
+
+ loc.inode = inode_ref(inode);
+ loc.parent = inode_ref(priv->dot_shard_rm_inode);
+ ret = inode_path(loc.parent, entry->d_name, (char **)&(loc.path));
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_PATH_FAILED,
+ "Inode path failed on %s", entry->d_name);
+ ret = -ENOMEM;
+ goto err;
+ }
+
+ loc.name = strrchr(loc.path, '/');
+ if (loc.name)
+ loc.name++;
+ ret = syncop_lookup(FIRST_CHILD(this), &loc, NULL, NULL, local->xattr_req,
+ &xattr_rsp);
+ if (ret)
+ goto err;
+
+ ret = dict_get_ptr(xattr_rsp, GF_XATTR_SHARD_BLOCK_SIZE, &bsize);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_DICT_OP_FAILED,
+ "Failed to get dict value: key:%s", GF_XATTR_SHARD_BLOCK_SIZE);
+ goto err;
+ }
+ block_size = ntoh64(*((uint64_t *)bsize));
+
+ ret = dict_get_ptr(xattr_rsp, GF_XATTR_SHARD_FILE_SIZE, &size_attr);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_DICT_OP_FAILED,
+ "Failed to get dict value: key:%s", GF_XATTR_SHARD_FILE_SIZE);
+ goto err;
+ }
+
+ memcpy(size_array, size_attr, sizeof(size_array));
+ size = ntoh64(size_array[0]);
+
+ shard_count = (size / block_size) - 1;
+ if (shard_count < 0) {
+ gf_msg_debug(this->name, 0,
+ "Size of %s hasn't grown beyond "
+ "its shard-block-size. Nothing to delete. "
+ "Returning",
+ entry->d_name);
+ /* File size < shard-block-size, so nothing to delete */
+ ret = 0;
+ goto delete_marker;
+ }
+ if ((size % block_size) > 0)
+ shard_count++;
+
+ if (shard_count == 0) {
+ gf_msg_debug(this->name, 0,
+ "Size of %s is exactly equal to "
+ "its shard-block-size. Nothing to delete. "
+ "Returning",
+ entry->d_name);
+ ret = 0;
+ goto delete_marker;
+ }
+ gf_msg_debug(this->name, 0,
+ "base file = %s, "
+ "shard-block-size=%" PRIu64 ", file-size=%" PRIu64
+ ", "
+ "shard_count=%d",
+ entry->d_name, block_size, size, shard_count);
+
+ /* Perform a gfid-based lookup to see if gfid corresponding to marker
+ * file's base name exists.
+ */
+ loc_wipe(&loc);
+ loc.inode = inode_new(this->itable);
+ if (!loc.inode) {
+ ret = -ENOMEM;
+ goto err;
+ }
+ gf_uuid_parse(entry->d_name, loc.gfid);
+ ret = syncop_lookup(FIRST_CHILD(this), &loc, NULL, NULL, NULL, NULL);
+ if (!ret) {
+ gf_msg_debug(this->name, 0,
+ "Base shard corresponding to gfid "
+ "%s is present. Skipping shard deletion. "
+ "Returning",
+ entry->d_name);
+ ret = 0;
+ goto delete_marker;
+ }
-unwind:
- SHARD_STACK_UNWIND (unlink, frame, op_ret, op_errno, preparent,
- postparent, xdata);
- return 0;
+ first_block = 1;
+
+ while (shard_count) {
+ if (shard_count < local->deletion_rate) {
+ now = shard_count;
+ shard_count = 0;
+ } else {
+ now = local->deletion_rate;
+ shard_count -= local->deletion_rate;
+ }
+
+ gf_msg_debug(this->name, 0,
+ "deleting %d shards starting from "
+ "block %d of gfid %s",
+ now, first_block, entry->d_name);
+ ret = shard_regulated_shards_deletion(cleanup_frame, this, now,
+ first_block, entry);
+ if (ret)
+ goto err;
+ first_block += now;
+ }
+
+delete_marker:
+ loc_wipe(&loc);
+ loc.inode = inode_ref(inode);
+ loc.parent = inode_ref(priv->dot_shard_rm_inode);
+ ret = inode_path(loc.parent, entry->d_name, (char **)&(loc.path));
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_PATH_FAILED,
+ "Inode path failed on %s", entry->d_name);
+ ret = -ENOMEM;
+ goto err;
+ }
+ loc.name = strrchr(loc.path, '/');
+ if (loc.name)
+ loc.name++;
+ ret = syncop_unlink(FIRST_CHILD(this), &loc, NULL, NULL);
+ if (ret)
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_SHARDS_DELETION_FAILED,
+ "Failed to delete %s "
+ "from /%s",
+ entry->d_name, GF_SHARD_REMOVE_ME_DIR);
+err:
+ if (xattr_rsp)
+ dict_unref(xattr_rsp);
+ loc_wipe(&loc);
+ return ret;
+}
+
+int
+shard_delete_shards_of_entry(call_frame_t *cleanup_frame, xlator_t *this,
+ gf_dirent_t *entry, inode_t *inode)
+{
+ int ret = -1;
+ loc_t loc = {
+ 0,
+ };
+ shard_priv_t *priv = NULL;
+
+ priv = this->private;
+ loc.inode = inode_ref(priv->dot_shard_rm_inode);
+
+ ret = syncop_entrylk(FIRST_CHILD(this), this->name, &loc, entry->d_name,
+ ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL, NULL);
+ if (ret < 0) {
+ if (ret == -EAGAIN) {
+ ret = 0;
+ }
+ goto out;
+ }
+ {
+ ret = __shard_delete_shards_of_entry(cleanup_frame, this, entry, inode);
+ }
+ syncop_entrylk(FIRST_CHILD(this), this->name, &loc, entry->d_name,
+ ENTRYLK_UNLOCK, ENTRYLK_WRLCK, NULL, NULL);
+out:
+ loc_wipe(&loc);
+ return ret;
}
int
-shard_unlink_base_file (call_frame_t *frame, xlator_t *this)
+shard_delete_shards_cbk(int ret, call_frame_t *frame, void *data)
{
- shard_local_t *local = NULL;
-
- local = frame->local;
-
- if (dict_set_uint32 (local->xattr_req, GET_LINK_COUNT, 0))
- gf_msg (this->name, GF_LOG_WARNING, 0,
- SHARD_MSG_DICT_SET_FAILED, "Failed to set "
- GET_LINK_COUNT" in dict");
-
- /* To-Do: Request open-fd count on base file */
- STACK_WIND (frame, shard_unlink_base_file_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->unlink, &local->loc, local->xflag,
- local->xattr_req);
- return 0;
+ SHARD_STACK_DESTROY(frame);
+ return 0;
}
-void
-shard_unlink_block_inode (shard_local_t *local, int shard_block_num)
+int
+shard_resolve_internal_dir(xlator_t *this, shard_local_t *local,
+ shard_internal_dir_type_t type)
{
- char block_bname[256] = {0,};
- inode_t *inode = NULL;
- inode_t *base_inode = NULL;
- xlator_t *this = NULL;
- shard_priv_t *priv = NULL;
- shard_inode_ctx_t *ctx = NULL;
- shard_inode_ctx_t *base_ictx = NULL;
- gf_boolean_t unlink_unref_forget = _gf_false;
+ int ret = 0;
+ char *bname = NULL;
+ loc_t *loc = NULL;
+ shard_priv_t *priv = NULL;
+ uuid_t gfid = {
+ 0,
+ };
+ struct iatt stbuf = {
+ 0,
+ };
- this = THIS;
- priv = this->private;
+ priv = this->private;
- inode = local->inode_list[shard_block_num - local->first_block];
- base_inode = local->resolver_base_inode;
-
- shard_make_block_bname (shard_block_num, (local->loc.inode)->gfid,
- block_bname, sizeof (block_bname));
+ switch (type) {
+ case SHARD_INTERNAL_DIR_DOT_SHARD:
+ loc = &local->dot_shard_loc;
+ gf_uuid_copy(gfid, priv->dot_shard_gfid);
+ bname = GF_SHARD_DIR;
+ break;
+ case SHARD_INTERNAL_DIR_DOT_SHARD_REMOVE_ME:
+ loc = &local->dot_shard_rm_loc;
+ gf_uuid_copy(gfid, priv->dot_shard_rm_gfid);
+ bname = GF_SHARD_REMOVE_ME_DIR;
+ break;
+ default:
+ break;
+ }
+ loc->inode = inode_find(this->itable, gfid);
+ if (!loc->inode) {
+ ret = shard_init_internal_dir_loc(this, local, type);
+ if (ret)
+ goto err;
+ ret = dict_reset(local->xattr_req);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_DICT_OP_FAILED,
+ "Failed to reset "
+ "dict");
+ ret = -ENOMEM;
+ goto err;
+ }
+ ret = dict_set_gfuuid(local->xattr_req, "gfid-req", gfid, true);
+ ret = syncop_lookup(FIRST_CHILD(this), loc, &stbuf, NULL,
+ local->xattr_req, NULL);
+ if (ret < 0) {
+ if (ret != -ENOENT)
+ gf_msg(this->name, GF_LOG_ERROR, -ret,
+ SHARD_MSG_SHARDS_DELETION_FAILED,
+ "Lookup on %s failed, exiting", bname);
+ goto err;
+ } else {
+ shard_link_internal_dir_inode(local, loc->inode, &stbuf, type);
+ }
+ }
+ ret = 0;
+err:
+ return ret;
+}
+
+int
+shard_lookup_marker_entry(xlator_t *this, shard_local_t *local,
+ gf_dirent_t *entry)
+{
+ int ret = 0;
+ loc_t loc = {
+ 0,
+ };
+
+ loc.inode = inode_new(this->itable);
+ if (!loc.inode) {
+ ret = -ENOMEM;
+ goto err;
+ }
+ loc.parent = inode_ref(local->fd->inode);
+
+ ret = inode_path(loc.parent, entry->d_name, (char **)&(loc.path));
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_PATH_FAILED,
+ "Inode path failed on %s", entry->d_name);
+ ret = -ENOMEM;
+ goto err;
+ }
+
+ loc.name = strrchr(loc.path, '/');
+ if (loc.name)
+ loc.name++;
+
+ ret = syncop_lookup(FIRST_CHILD(this), &loc, NULL, NULL, NULL, NULL);
+ if (ret < 0) {
+ goto err;
+ }
+ entry->inode = inode_ref(loc.inode);
+ ret = 0;
+err:
+ loc_wipe(&loc);
+ return ret;
+}
+
+int
+shard_delete_shards(void *opaque)
+{
+ int ret = 0;
+ off_t offset = 0;
+ loc_t loc = {
+ 0,
+ };
+ inode_t *link_inode = NULL;
+ xlator_t *this = NULL;
+ shard_priv_t *priv = NULL;
+ shard_local_t *local = NULL;
+ gf_dirent_t entries;
+ gf_dirent_t *entry = NULL;
+ call_frame_t *cleanup_frame = NULL;
+ gf_boolean_t done = _gf_false;
+
+ this = THIS;
+ priv = this->private;
+ INIT_LIST_HEAD(&entries.list);
+
+ cleanup_frame = opaque;
+
+ local = mem_get0(this->local_pool);
+ if (!local) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, SHARD_MSG_MEMALLOC_FAILED,
+ "Failed to create local to "
+ "delete shards");
+ ret = -ENOMEM;
+ goto err;
+ }
+ cleanup_frame->local = local;
+ local->fop = GF_FOP_UNLINK;
+
+ local->xattr_req = dict_new();
+ if (!local->xattr_req) {
+ ret = -ENOMEM;
+ goto err;
+ }
+ local->deletion_rate = priv->deletion_rate;
+
+ ret = shard_resolve_internal_dir(this, local, SHARD_INTERNAL_DIR_DOT_SHARD);
+ if (ret == -ENOENT) {
+ gf_msg_debug(this->name, 0,
+ ".shard absent. Nothing to"
+ " delete. Exiting");
+ ret = 0;
+ goto err;
+ } else if (ret < 0) {
+ goto err;
+ }
+
+ ret = shard_resolve_internal_dir(this, local,
+ SHARD_INTERNAL_DIR_DOT_SHARD_REMOVE_ME);
+ if (ret == -ENOENT) {
+ gf_msg_debug(this->name, 0,
+ ".remove_me absent. "
+ "Nothing to delete. Exiting");
+ ret = 0;
+ goto err;
+ } else if (ret < 0) {
+ goto err;
+ }
+
+ local->fd = fd_anonymous(local->dot_shard_rm_loc.inode);
+ if (!local->fd) {
+ ret = -ENOMEM;
+ goto err;
+ }
+
+ for (;;) {
+ offset = 0;
LOCK(&priv->lock);
- LOCK(&base_inode->lock);
- LOCK(&inode->lock);
{
- __shard_inode_ctx_get (inode, this, &ctx);
- if (!list_empty (&ctx->ilist)) {
- list_del_init (&ctx->ilist);
- priv->inode_count--;
- GF_ASSERT (priv->inode_count >= 0);
- unlink_unref_forget = _gf_true;
+ if (priv->bg_del_state == SHARD_BG_DELETION_LAUNCHING) {
+ priv->bg_del_state = SHARD_BG_DELETION_IN_PROGRESS;
+ } else if (priv->bg_del_state == SHARD_BG_DELETION_IN_PROGRESS) {
+ priv->bg_del_state = SHARD_BG_DELETION_NONE;
+ done = _gf_true;
+ }
+ }
+ UNLOCK(&priv->lock);
+ if (done)
+ break;
+ while (
+ (ret = syncop_readdirp(FIRST_CHILD(this), local->fd, 131072, offset,
+ &entries, local->xattr_req, NULL))) {
+ if (ret > 0)
+ ret = 0;
+ list_for_each_entry(entry, &entries.list, list)
+ {
+ offset = entry->d_off;
+
+ if (!strcmp(entry->d_name, ".") || !strcmp(entry->d_name, ".."))
+ continue;
+
+ if (!entry->inode) {
+ ret = shard_lookup_marker_entry(this, local, entry);
+ if (ret < 0)
+ continue;
}
- if (ctx->fsync_needed) {
- inode_unref (base_inode);
- list_del_init (&ctx->to_fsync_list);
- __shard_inode_ctx_get (base_inode, this, &base_ictx);
- base_ictx->fsync_count--;
+ link_inode = inode_link(entry->inode, local->fd->inode,
+ entry->d_name, &entry->d_stat);
+
+ gf_msg_debug(this->name, 0,
+ "Initiating deletion of "
+ "shards of gfid %s",
+ entry->d_name);
+ ret = shard_delete_shards_of_entry(cleanup_frame, this, entry,
+ link_inode);
+ inode_unlink(link_inode, local->fd->inode, entry->d_name);
+ inode_unref(link_inode);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, -ret,
+ SHARD_MSG_SHARDS_DELETION_FAILED,
+ "Failed to clean up shards of gfid %s",
+ entry->d_name);
+ continue;
}
+ gf_msg(this->name, GF_LOG_INFO, 0,
+ SHARD_MSG_SHARD_DELETION_COMPLETED,
+ "Deleted "
+ "shards of gfid=%s from backend",
+ entry->d_name);
+ }
+ gf_dirent_free(&entries);
+ if (ret)
+ break;
}
- UNLOCK(&inode->lock);
- UNLOCK(&base_inode->lock);
- if (unlink_unref_forget) {
- inode_unlink (inode, priv->dot_shard_inode, block_bname);
- inode_unref (inode);
- inode_forget (inode, 0);
- }
- UNLOCK(&priv->lock);
+ }
+ ret = 0;
+ loc_wipe(&loc);
+ return ret;
+
+err:
+ LOCK(&priv->lock);
+ {
+ priv->bg_del_state = SHARD_BG_DELETION_NONE;
+ }
+ UNLOCK(&priv->lock);
+ loc_wipe(&loc);
+ return ret;
}
int
-shard_rename_cbk (call_frame_t *frame, xlator_t *this);
-
-int32_t
-shard_unlink_cbk (call_frame_t *frame, xlator_t *this)
+shard_unlock_inodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- shard_local_t *local = frame->local;
-
- SHARD_STACK_UNWIND (unlink, frame, local->op_ret, local->op_errno,
- &local->preoldparent, &local->postoldparent,
- local->xattr_rsp);
- return 0;
+ if (op_ret)
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, SHARD_MSG_FOP_FAILED,
+ "Unlock failed. Please check brick logs for "
+ "more details");
+ SHARD_STACK_DESTROY(frame);
+ return 0;
}
int
-shard_unlink_shards_do_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- struct iatt *preparent, struct iatt *postparent,
- dict_t *xdata)
+shard_unlock_inodelk(call_frame_t *frame, xlator_t *this)
{
- int call_count = 0;
- int shard_block_num = (long) cookie;
- shard_local_t *local = NULL;
-
- local = frame->local;
-
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto done;
- }
-
- shard_unlink_block_inode (local, shard_block_num);
+ loc_t *loc = NULL;
+ call_frame_t *lk_frame = NULL;
+ shard_local_t *local = NULL;
+ shard_local_t *lk_local = NULL;
+ shard_inodelk_t *lock = NULL;
-done:
- call_count = shard_call_count_return (frame);
- if (local->unlink_shards_barriered) {
- syncbarrier_wake (&local->barrier);
- } else {
-
- if (call_count == 0) {
- SHARD_UNSET_ROOT_FS_ID (frame, local);
+ local = frame->local;
+ lk_frame = local->inodelk_frame;
+ lk_local = lk_frame->local;
+ local->inodelk_frame = NULL;
+ loc = &local->int_inodelk.loc;
+ lock = &lk_local->int_inodelk;
+ lock->flock.l_type = F_UNLCK;
- if (local->fop == GF_FOP_UNLINK)
- shard_unlink_cbk (frame, this);
- else if (local->fop == GF_FOP_RENAME)
- shard_rename_cbk (frame, this);
- }
- }
- return 0;
+ STACK_WIND(lk_frame, shard_unlock_inodelk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->inodelk, lock->domain, loc, F_SETLK,
+ &lock->flock, NULL);
+ local->int_inodelk.acquired_lock = _gf_false;
+ return 0;
}
int
-shard_unlink_shards_do (call_frame_t *frame, xlator_t *this, inode_t *inode)
+shard_rename_src_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ struct iatt *preoldparent, struct iatt *postoldparent,
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata);
+int
+shard_rename_src_base_file(call_frame_t *frame, xlator_t *this)
{
- int i = 0;
- int ret = -1;
- int count = 0;
- int call_count = 0;
- uint32_t last_block = 0;
- uint32_t cur_block = 0;
- char *bname = NULL;
- char path[PATH_MAX] = {0,};
- loc_t loc = {0,};
- gf_boolean_t wind_failed = _gf_false;
- shard_local_t *local = NULL;
- shard_priv_t *priv = NULL;
+ int ret = 0;
+ loc_t *dst_loc = NULL;
+ loc_t tmp_loc = {
+ 0,
+ };
+ shard_local_t *local = frame->local;
- priv = this->private;
- local = frame->local;
+ if (local->dst_block_size) {
+ tmp_loc.parent = inode_ref(local->loc2.parent);
+ ret = inode_path(tmp_loc.parent, local->loc2.name,
+ (char **)&tmp_loc.path);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_PATH_FAILED,
+ "Inode path failed"
+ " on pargfid=%s bname=%s",
+ uuid_utoa(tmp_loc.parent->gfid), local->loc2.name);
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto err;
+ }
+
+ tmp_loc.name = strrchr(tmp_loc.path, '/');
+ if (tmp_loc.name)
+ tmp_loc.name++;
+ dst_loc = &tmp_loc;
+ } else {
+ dst_loc = &local->loc2;
+ }
+
+ /* To-Do: Request open-fd count on dst base file */
+ STACK_WIND(frame, shard_rename_src_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, &local->loc, dst_loc,
+ local->xattr_req);
+ loc_wipe(&tmp_loc);
+ return 0;
+err:
+ loc_wipe(&tmp_loc);
+ shard_common_failure_unwind(local->fop, frame, local->op_ret,
+ local->op_errno);
+ return 0;
+}
- /* local->num_blocks includes the base file block. This function only
- * deletes the shards under /.shard. So subtract num_blocks by 1.
- */
- local->call_count = call_count = local->num_blocks - 1;
- last_block = local->last_block;
+int
+shard_unlink_base_file(call_frame_t *frame, xlator_t *this);
- /* Ignore the inode associated with the base file and start counting
- * from 1.
- */
- for (i = 1; i < local->num_blocks; i++) {
- if (!local->inode_list[i])
- continue;
- count++;
- }
+int
+shard_set_size_attrs_on_marker_file_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
+{
+ shard_priv_t *priv = NULL;
+ shard_local_t *local = NULL;
- if (!count) {
- /* callcount = 0 implies that all of the shards that need to be
- * unlinked are non-existent (in other words the file is full of
- * holes). So shard xlator can simply return the fop to its
- * parent now.
- */
- gf_msg_debug (this->name, 0, "All shards that need to be "
- "unlinked are non-existent: %s",
- uuid_utoa (inode->gfid));
- local->num_blocks = 1;
- if (local->fop == GF_FOP_UNLINK) {
- shard_unlink_cbk (frame, this);
- } else if (local->fop == GF_FOP_RENAME) {
- gf_msg_debug (this->name, 0, "Resuming rename()");
- shard_rename_cbk (frame, this);
- }
- return 0;
- }
+ priv = this->private;
+ local = frame->local;
+ if (op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, SHARD_MSG_FOP_FAILED,
+ "Xattrop on marker file failed "
+ "while performing %s; entry gfid=%s",
+ gf_fop_string(local->fop), local->newloc.name);
+ goto err;
+ }
- local->call_count = call_count = count;
- cur_block = 1;
- SHARD_SET_ROOT_FS_ID (frame, local);
- if (local->unlink_shards_barriered)
- local->barrier.waitfor = count;
+ inode_unlink(local->newloc.inode, priv->dot_shard_rm_inode,
+ local->newloc.name);
- /* Ignore the base file and start iterating from the first block shard.
- */
- while (cur_block <= last_block) {
- if (!local->inode_list[cur_block]) {
- cur_block++;
- continue;
- }
+ if (local->fop == GF_FOP_UNLINK)
+ shard_unlink_base_file(frame, this);
+ else if (local->fop == GF_FOP_RENAME)
+ shard_rename_src_base_file(frame, this);
+ return 0;
+err:
+ shard_common_failure_unwind(local->fop, frame, op_ret, op_errno);
+ return 0;
+}
+
+int
+shard_set_size_attrs_on_marker_file(call_frame_t *frame, xlator_t *this)
+{
+ int op_errno = ENOMEM;
+ uint64_t bs = 0;
+ dict_t *xdata = NULL;
+ shard_local_t *local = NULL;
+
+ local = frame->local;
+ xdata = dict_new();
+ if (!xdata)
+ goto err;
+
+ if (local->fop == GF_FOP_UNLINK)
+ bs = local->block_size;
+ else if (local->fop == GF_FOP_RENAME)
+ bs = local->dst_block_size;
+ SHARD_INODE_CREATE_INIT(this, bs, xdata, &local->newloc,
+ local->prebuf.ia_size, 0, err);
+ STACK_WIND(frame, shard_set_size_attrs_on_marker_file_cbk,
+ FIRST_CHILD(this), FIRST_CHILD(this)->fops->xattrop,
+ &local->newloc, GF_XATTROP_GET_AND_SET, xdata, NULL);
+ dict_unref(xdata);
+ return 0;
+err:
+ if (xdata)
+ dict_unref(xdata);
+ shard_common_failure_unwind(local->fop, frame, -1, op_errno);
+ return 0;
+}
+
+int
+shard_lookup_marker_file_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
+{
+ inode_t *linked_inode = NULL;
+ shard_priv_t *priv = NULL;
+ shard_local_t *local = NULL;
+
+ local = frame->local;
+ priv = this->private;
+
+ if (op_ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, SHARD_MSG_FOP_FAILED,
+ "Lookup on marker file failed "
+ "while performing %s; entry gfid=%s",
+ gf_fop_string(local->fop), local->newloc.name);
+ goto err;
+ }
+
+ linked_inode = inode_link(inode, priv->dot_shard_rm_inode,
+ local->newloc.name, buf);
+ inode_unref(local->newloc.inode);
+ local->newloc.inode = linked_inode;
+ shard_set_size_attrs_on_marker_file(frame, this);
+ return 0;
+err:
+ shard_common_failure_unwind(local->fop, frame, op_ret, op_errno);
+ return 0;
+}
- if (wind_failed) {
- shard_unlink_shards_do_cbk (frame,
- (void *) (long) cur_block,
- this, -1, ENOMEM, NULL,
- NULL, NULL);
- goto next;
- }
+int
+shard_lookup_marker_file(call_frame_t *frame, xlator_t *this)
+{
+ int op_errno = ENOMEM;
+ dict_t *xattr_req = NULL;
+ shard_local_t *local = NULL;
- shard_make_block_abspath (cur_block, inode->gfid, path,
- sizeof (path));
- bname = strrchr (path, '/') + 1;
- loc.parent = inode_ref (priv->dot_shard_inode);
- ret = inode_path (loc.parent, bname, (char **) &(loc.path));
- if (ret < 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_PATH_FAILED, "Inode path failed"
- " on %s, base file gfid = %s", bname,
- uuid_utoa (inode->gfid));
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- loc_wipe (&loc);
- wind_failed = _gf_true;
- shard_unlink_shards_do_cbk (frame,
- (void *) (long) cur_block,
- this, -1, ENOMEM, NULL,
- NULL, NULL);
- goto next;
- }
+ local = frame->local;
- loc.name = strrchr (loc.path, '/');
- if (loc.name)
- loc.name++;
- loc.inode = inode_ref (local->inode_list[cur_block]);
-
- STACK_WIND_COOKIE (frame, shard_unlink_shards_do_cbk,
- (void *) (long) cur_block, FIRST_CHILD(this),
- FIRST_CHILD (this)->fops->unlink, &loc,
- local->xflag, local->xattr_req);
- loc_wipe (&loc);
-
-next:
- cur_block++;
- if (!--call_count)
- break;
- }
- if (local->unlink_shards_barriered)
- syncbarrier_wait (&local->barrier, count);
+ xattr_req = shard_create_gfid_dict(local->xattr_req);
+ if (!xattr_req)
+ goto err;
- return 0;
+ STACK_WIND(frame, shard_lookup_marker_file_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->lookup, &local->newloc, xattr_req);
+ dict_unref(xattr_req);
+ return 0;
+err:
+ shard_common_failure_unwind(local->fop, frame, -1, op_errno);
+ return 0;
}
int
-shard_post_lookup_unlink_handler (call_frame_t *frame, xlator_t *this)
+shard_create_marker_file_under_remove_me_cbk(
+ call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+ int32_t op_errno, inode_t *inode, struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- shard_local_t *local = NULL;
+ inode_t *linked_inode = NULL;
+ shard_priv_t *priv = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
+ priv = this->private;
- if (local->op_ret < 0) {
- SHARD_STACK_UNWIND (unlink, frame, local->op_ret,
- local->op_errno, NULL, NULL, NULL);
- return 0;
+ SHARD_UNSET_ROOT_FS_ID(frame, local);
+ if (op_ret < 0) {
+ if ((op_errno != EEXIST) && (op_errno != ENODATA)) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, SHARD_MSG_FOP_FAILED,
+ "Marker file creation "
+ "failed while performing %s; entry gfid=%s",
+ gf_fop_string(local->fop), local->newloc.name);
+ goto err;
+ } else {
+ shard_lookup_marker_file(frame, this);
+ return 0;
}
+ }
- shard_unlink_base_file (frame, this);
- return 0;
+ linked_inode = inode_link(inode, priv->dot_shard_rm_inode,
+ local->newloc.name, buf);
+ inode_unref(local->newloc.inode);
+ local->newloc.inode = linked_inode;
+
+ if (local->fop == GF_FOP_UNLINK)
+ shard_unlink_base_file(frame, this);
+ else if (local->fop == GF_FOP_RENAME)
+ shard_rename_src_base_file(frame, this);
+ return 0;
+err:
+ shard_common_failure_unwind(local->fop, frame, -1, local->op_errno);
+ return 0;
+}
+
+int
+shard_create_marker_file_under_remove_me(call_frame_t *frame, xlator_t *this,
+ loc_t *loc)
+{
+ int ret = 0;
+ int op_errno = ENOMEM;
+ uint64_t bs = 0;
+ char g1[64] = {
+ 0,
+ };
+ char g2[64] = {
+ 0,
+ };
+ dict_t *xattr_req = NULL;
+ shard_priv_t *priv = NULL;
+ shard_local_t *local = NULL;
+
+ priv = this->private;
+ local = frame->local;
+
+ SHARD_SET_ROOT_FS_ID(frame, local);
+
+ xattr_req = shard_create_gfid_dict(local->xattr_req);
+ if (!xattr_req)
+ goto err;
+
+ local->newloc.inode = inode_new(this->itable);
+ local->newloc.parent = inode_ref(priv->dot_shard_rm_inode);
+ ret = inode_path(local->newloc.parent, uuid_utoa(loc->inode->gfid),
+ (char **)&local->newloc.path);
+ if (ret < 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_PATH_FAILED,
+ "Inode path failed on "
+ "pargfid=%s bname=%s",
+ uuid_utoa_r(priv->dot_shard_rm_gfid, g1),
+ uuid_utoa_r(loc->inode->gfid, g2));
+ goto err;
+ }
+ local->newloc.name = strrchr(local->newloc.path, '/');
+ if (local->newloc.name)
+ local->newloc.name++;
+
+ if (local->fop == GF_FOP_UNLINK)
+ bs = local->block_size;
+ else if (local->fop == GF_FOP_RENAME)
+ bs = local->dst_block_size;
+
+ SHARD_INODE_CREATE_INIT(this, bs, xattr_req, &local->newloc,
+ local->prebuf.ia_size, 0, err);
+
+ STACK_WIND(frame, shard_create_marker_file_under_remove_me_cbk,
+ FIRST_CHILD(this), FIRST_CHILD(this)->fops->mknod,
+ &local->newloc, 0, 0, 0644, xattr_req);
+ dict_unref(xattr_req);
+ return 0;
+
+err:
+ if (xattr_req)
+ dict_unref(xattr_req);
+ shard_create_marker_file_under_remove_me_cbk(frame, 0, this, -1, op_errno,
+ NULL, NULL, NULL, NULL, NULL);
+ return 0;
}
int
-shard_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
- dict_t *xdata)
+shard_unlock_entrylk(call_frame_t *frame, xlator_t *this);
+
+int
+shard_unlink_base_file_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *preparent, struct iatt *postparent,
+ dict_t *xdata)
{
- int ret = -1;
- uint64_t block_size = 0;
- shard_local_t *local = NULL;
+ int ret = 0;
+ shard_local_t *local = NULL;
- ret = shard_inode_ctx_get_block_size (loc->inode, this, &block_size);
- if ((ret) && (!IA_ISLNK(loc->inode->ia_type))) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_CTX_GET_FAILED, "Failed to get block "
- "size from inode ctx of %s",
- uuid_utoa (loc->inode->gfid));
- goto err;
- }
+ local = frame->local;
+
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ } else {
+ shard_inode_ctx_set_refresh_flag(local->int_inodelk.loc.inode, this);
+ local->preoldparent = *preparent;
+ local->postoldparent = *postparent;
+ if (xdata)
+ local->xattr_rsp = dict_ref(xdata);
+ if (local->cleanup_required)
+ shard_start_background_deletion(this);
+ }
- if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
- STACK_WIND (frame, default_unlink_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
- return 0;
+ if (local->entrylk_frame) {
+ ret = shard_unlock_entrylk(frame, this);
+ if (ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -ret;
}
+ }
- local = mem_get0 (this->local_pool);
- if (!local)
- goto err;
+ ret = shard_unlock_inodelk(frame, this);
+ if (ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -ret;
+ }
- frame->local = local;
+ shard_unlink_cbk(frame, this);
+ return 0;
+}
- loc_copy (&local->loc, loc);
- local->xflag = xflag;
- local->xattr_req = (xdata) ? dict_ref (xdata) : dict_new ();
- local->block_size = block_size;
- local->resolver_base_inode = loc->inode;
- local->fop = GF_FOP_UNLINK;
- if (!this->itable)
- this->itable = (local->loc.inode)->table;
+int
+shard_unlink_base_file(call_frame_t *frame, xlator_t *this)
+{
+ shard_local_t *local = frame->local;
- shard_lookup_base_file (frame, this, &local->loc,
- shard_post_lookup_unlink_handler);
- return 0;
-err:
- SHARD_STACK_UNWIND (unlink, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ /* To-Do: Request open-fd count on base file */
+ STACK_WIND(frame, shard_unlink_base_file_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, &local->loc, local->xflag,
+ local->xattr_req);
+ return 0;
+}
+int
+shard_unlock_entrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+ if (op_ret)
+ gf_msg(this->name, GF_LOG_ERROR, op_errno, SHARD_MSG_FOP_FAILED,
+ "Unlock failed. Please check brick logs for "
+ "more details");
+ SHARD_STACK_DESTROY(frame);
+ return 0;
}
int
-shard_rename_cbk (call_frame_t *frame, xlator_t *this)
+shard_unlock_entrylk(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
+ loc_t *loc = NULL;
+ call_frame_t *lk_frame = NULL;
+ shard_local_t *local = NULL;
+ shard_local_t *lk_local = NULL;
+ shard_entrylk_t *lock = NULL;
- local = frame->local;
+ local = frame->local;
+ lk_frame = local->entrylk_frame;
+ lk_local = lk_frame->local;
+ local->entrylk_frame = NULL;
+ lock = &lk_local->int_entrylk;
+ loc = &lock->loc;
- SHARD_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
- &local->prebuf, &local->preoldparent,
- &local->postoldparent, &local->prenewparent,
- &local->postnewparent, local->xattr_rsp);
- return 0;
+ STACK_WIND(lk_frame, shard_unlock_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, loc,
+ lk_local->int_entrylk.basename, ENTRYLK_UNLOCK, ENTRYLK_WRLCK,
+ NULL);
+ local->int_entrylk.acquired_lock = _gf_false;
+ return 0;
}
int
-shard_rename_unlink_dst_shards_do (call_frame_t *frame, xlator_t *this)
+shard_post_entrylk_fop_handler(call_frame_t *frame, xlator_t *this)
{
- int ret = -1;
- uint32_t link_count = 0;
- shard_local_t *local = NULL;
- shard_priv_t *priv = NULL;
-
- local = frame->local;
- priv = this->private;
+ shard_local_t *local = NULL;
- local->first_block = get_lowest_block (0, local->dst_block_size);
- local->last_block = get_highest_block (0, local->postbuf.ia_size,
- local->dst_block_size);
- local->num_blocks = local->last_block - local->first_block + 1;
- local->resolver_base_inode = local->loc2.inode;
-
- if ((local->xattr_rsp) &&
- (!dict_get_uint32 (local->xattr_rsp, GET_LINK_COUNT, &link_count))
- && (link_count > 1)) {
- shard_rename_cbk (frame, this);
- return 0;
- }
+ local = frame->local;
- if (local->num_blocks == 1) {
- shard_rename_cbk (frame, this);
- return 0;
- }
+ switch (local->fop) {
+ case GF_FOP_UNLINK:
+ case GF_FOP_RENAME:
+ shard_create_marker_file_under_remove_me(frame, this,
+ &local->int_inodelk.loc);
+ break;
+ default:
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_INVALID_FOP,
+ "post-entrylk handler not defined. This case should not"
+ " be hit");
+ break;
+ }
+ return 0;
+}
+
+int
+shard_acquire_entrylk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+ call_frame_t *main_frame = NULL;
+ shard_local_t *local = NULL;
+ shard_local_t *main_local = NULL;
+
+ local = frame->local;
+ main_frame = local->main_frame;
+ main_local = main_frame->local;
+
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(main_local->fop, main_frame, op_ret,
+ op_errno);
+ return 0;
+ }
+ main_local->int_entrylk.acquired_lock = _gf_true;
+ shard_post_entrylk_fop_handler(main_frame, this);
+ return 0;
+}
+
+int
+shard_acquire_entrylk(call_frame_t *frame, xlator_t *this, inode_t *inode,
+ uuid_t gfid)
+{
+ char gfid_str[GF_UUID_BUF_SIZE] = {
+ 0,
+ };
+ shard_local_t *local = NULL;
+ shard_local_t *entrylk_local = NULL;
+ shard_entrylk_t *int_entrylk = NULL;
+ call_frame_t *entrylk_frame = NULL;
+
+ local = frame->local;
+ entrylk_frame = create_frame(this, this->ctx->pool);
+ if (!entrylk_frame) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, SHARD_MSG_MEMALLOC_FAILED,
+ "Failed to create new frame "
+ "to lock marker file");
+ goto err;
+ }
+
+ entrylk_local = mem_get0(this->local_pool);
+ if (!entrylk_local) {
+ STACK_DESTROY(entrylk_frame->root);
+ goto err;
+ }
+
+ entrylk_frame->local = entrylk_local;
+ entrylk_local->main_frame = frame;
+ int_entrylk = &entrylk_local->int_entrylk;
+
+ int_entrylk->loc.inode = inode_ref(inode);
+ set_lk_owner_from_ptr(&entrylk_frame->root->lk_owner, entrylk_frame->root);
+ local->entrylk_frame = entrylk_frame;
+ gf_uuid_unparse(gfid, gfid_str);
+ int_entrylk->basename = gf_strdup(gfid_str);
+
+ STACK_WIND(entrylk_frame, shard_acquire_entrylk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->entrylk, this->name, &int_entrylk->loc,
+ int_entrylk->basename, ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
+ return 0;
+err:
+ shard_common_failure_unwind(local->fop, frame, -1, ENOMEM);
+ return 0;
+}
- local->inode_list = GF_CALLOC (local->num_blocks, sizeof (inode_t *),
- gf_shard_mt_inode_list);
- if (!local->inode_list)
- goto out;
+int
+shard_post_lookup_base_shard_rm_handler(call_frame_t *frame, xlator_t *this)
+{
+ shard_local_t *local = NULL;
+ shard_priv_t *priv = NULL;
- local->dot_shard_loc.inode = inode_find (this->itable,
- priv->dot_shard_gfid);
- if (!local->dot_shard_loc.inode) {
- ret = shard_init_internal_dir_loc (this, local,
- SHARD_INTERNAL_DIR_DOT_SHARD);
- if (ret)
- goto out;
- shard_lookup_internal_dir (frame, this,
- shard_post_resolve_unlink_handler,
- SHARD_INTERNAL_DIR_DOT_SHARD);
- } else {
- local->post_res_handler = shard_post_resolve_unlink_handler;
- shard_refresh_internal_dir (frame, this,
- SHARD_INTERNAL_DIR_DOT_SHARD);
- }
+ priv = this->private;
+ local = frame->local;
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(local->fop, frame, -1, local->op_errno);
return 0;
+ }
-out:
- SHARD_STACK_UNWIND (rename, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
- NULL, NULL);
- return 0;
+ if (local->prebuf.ia_nlink > 1) {
+ gf_msg_debug(this->name, 0,
+ "link count on %s > 1:%d, "
+ "performing rename()/unlink()",
+ local->int_inodelk.loc.path, local->prebuf.ia_nlink);
+ if (local->fop == GF_FOP_RENAME)
+ shard_rename_src_base_file(frame, this);
+ else if (local->fop == GF_FOP_UNLINK)
+ shard_unlink_base_file(frame, this);
+ } else {
+ gf_msg_debug(this->name, 0,
+ "link count on %s = 1, creating "
+ "file under .remove_me",
+ local->int_inodelk.loc.path);
+ local->cleanup_required = _gf_true;
+ shard_acquire_entrylk(frame, this, priv->dot_shard_rm_inode,
+ local->prebuf.ia_gfid);
+ }
+ return 0;
}
int
-shard_post_rename_lookup_handler (call_frame_t *frame, xlator_t *this)
+shard_post_inodelk_fop_handler(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->op_ret < 0) {
- SHARD_STACK_UNWIND (rename, frame, local->op_ret,
- local->op_errno, NULL, NULL, NULL, NULL,
- NULL, NULL);
- return 0;
- }
+ switch (local->fop) {
+ case GF_FOP_UNLINK:
+ case GF_FOP_RENAME:
+ shard_refresh_base_file(frame, this, &local->int_inodelk.loc, NULL,
+ shard_post_lookup_base_shard_rm_handler);
+ break;
+ default:
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_INVALID_FOP,
+ "post-inodelk handler not defined. This case should not"
+ " be hit");
+ break;
+ }
+ return 0;
+}
+
+int
+shard_acquire_inodelk_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+ call_frame_t *main_frame = NULL;
+ shard_local_t *local = NULL;
+ shard_local_t *main_local = NULL;
+
+ local = frame->local;
+ main_frame = local->main_frame;
+ main_local = main_frame->local;
+
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(main_local->fop, main_frame, op_ret,
+ op_errno);
+ return 0;
+ }
+ main_local->int_inodelk.acquired_lock = _gf_true;
+ shard_post_inodelk_fop_handler(main_frame, this);
+ return 0;
+}
+
+int
+shard_acquire_inodelk(call_frame_t *frame, xlator_t *this, loc_t *loc)
+{
+ call_frame_t *lk_frame = NULL;
+ shard_local_t *local = NULL;
+ shard_local_t *lk_local = NULL;
+ shard_inodelk_t *int_inodelk = NULL;
+
+ local = frame->local;
+ lk_frame = create_frame(this, this->ctx->pool);
+ if (!lk_frame) {
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM, SHARD_MSG_MEMALLOC_FAILED,
+ "Failed to create new frame "
+ "to lock base shard");
+ goto err;
+ }
+ lk_local = mem_get0(this->local_pool);
+ if (!lk_local) {
+ STACK_DESTROY(lk_frame->root);
+ goto err;
+ }
+
+ lk_frame->local = lk_local;
+ lk_local->main_frame = frame;
+ int_inodelk = &lk_local->int_inodelk;
+
+ int_inodelk->flock.l_len = 0;
+ int_inodelk->flock.l_start = 0;
+ int_inodelk->domain = this->name;
+ int_inodelk->flock.l_type = F_WRLCK;
+ loc_copy(&local->int_inodelk.loc, loc);
+ set_lk_owner_from_ptr(&lk_frame->root->lk_owner, lk_frame->root);
+ local->inodelk_frame = lk_frame;
+
+ STACK_WIND(lk_frame, shard_acquire_inodelk_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->inodelk, int_inodelk->domain,
+ &local->int_inodelk.loc, F_SETLKW, &int_inodelk->flock, NULL);
+ return 0;
+err:
+ shard_common_failure_unwind(local->fop, frame, -1, ENOMEM);
+ return 0;
+}
- if (local->dst_block_size)
- shard_rename_unlink_dst_shards_do (frame, this);
- else
- shard_rename_cbk (frame, this);
+int
+shard_post_mkdir_rm_handler(call_frame_t *frame, xlator_t *this)
+{
+ loc_t *loc = NULL;
+ shard_local_t *local = NULL;
+
+ local = frame->local;
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(local->fop, frame, -1, local->op_errno);
return 0;
+ }
+ if (local->fop == GF_FOP_UNLINK)
+ loc = &local->loc;
+ else if (local->fop == GF_FOP_RENAME)
+ loc = &local->loc2;
+ shard_acquire_inodelk(frame, this, loc);
+ return 0;
}
int
-shard_rename_src_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- struct iatt *preoldparent, struct iatt *postoldparent,
- struct iatt *prenewparent, struct iatt *postnewparent,
- dict_t *xdata)
+shard_mkdir_internal_dir(call_frame_t *frame, xlator_t *this,
+ shard_post_resolve_fop_handler_t handler,
+ shard_internal_dir_type_t type);
+int
+shard_pre_mkdir_rm_handler(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto err;
- }
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(local->fop, frame, -1, local->op_errno);
+ return 0;
+ }
+ shard_mkdir_internal_dir(frame, this, shard_post_mkdir_rm_handler,
+ SHARD_INTERNAL_DIR_DOT_SHARD_REMOVE_ME);
+ return 0;
+}
- local->prebuf = *buf;
- local->preoldparent = *preoldparent;
- local->postoldparent = *postoldparent;
- local->prenewparent = *prenewparent;
- local->postnewparent = *postnewparent;
- if (xdata)
- local->xattr_rsp = dict_ref (xdata);
+void
+shard_begin_rm_resolution(call_frame_t *frame, xlator_t *this)
+{
+ shard_priv_t *priv = NULL;
+ shard_local_t *local = NULL;
- /* Now the base file is looked up to gather the ia_size and ia_blocks.*/
+ priv = this->private;
+ local = frame->local;
- if (local->block_size) {
- local->tmp_loc.inode = inode_new (this->itable);
- gf_uuid_copy (local->tmp_loc.gfid, (local->loc.inode)->gfid);
- shard_lookup_base_file (frame, this, &local->tmp_loc,
- shard_post_rename_lookup_handler);
+ local->dot_shard_rm_loc.inode = inode_find(this->itable,
+ priv->dot_shard_rm_gfid);
+ if (!local->dot_shard_rm_loc.inode) {
+ local->dot_shard_loc.inode = inode_find(this->itable,
+ priv->dot_shard_gfid);
+ if (!local->dot_shard_loc.inode) {
+ shard_mkdir_internal_dir(frame, this, shard_pre_mkdir_rm_handler,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
} else {
- shard_rename_unlink_dst_shards_do (frame, this);
+ local->post_res_handler = shard_pre_mkdir_rm_handler;
+ shard_refresh_internal_dir(frame, this,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
}
+ } else {
+ local->post_res_handler = shard_post_mkdir_rm_handler;
+ shard_refresh_internal_dir(frame, this,
+ SHARD_INTERNAL_DIR_DOT_SHARD_REMOVE_ME);
+ }
+}
- return 0;
+int
+shard_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+ dict_t *xdata)
+{
+ int ret = -1;
+ uint64_t block_size = 0;
+ shard_local_t *local = NULL;
+
+ ret = shard_inode_ctx_get_block_size(loc->inode, this, &block_size);
+ if ((ret) && (!IA_ISLNK(loc->inode->ia_type))) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_CTX_GET_FAILED,
+ "Failed to get block "
+ "size from inode ctx of %s",
+ uuid_utoa(loc->inode->gfid));
+ goto err;
+ }
+
+ if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
+ STACK_WIND(frame, default_unlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
+ return 0;
+ }
+
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
+
+ frame->local = local;
+
+ loc_copy(&local->loc, loc);
+ local->xflag = xflag;
+ local->xattr_req = (xdata) ? dict_ref(xdata) : dict_new();
+ local->block_size = block_size;
+ local->resolver_base_inode = loc->inode;
+ local->fop = GF_FOP_UNLINK;
+ if (!this->itable)
+ this->itable = (local->loc.inode)->table;
+
+ local->resolve_not = _gf_true;
+ shard_begin_rm_resolution(frame, this);
+ return 0;
err:
- SHARD_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno, NULL,
- NULL, NULL, NULL, NULL, NULL);
- return 0;
+ shard_common_failure_unwind(GF_FOP_UNLINK, frame, -1, ENOMEM);
+ return 0;
}
int
-shard_rename_src_base_file (call_frame_t *frame, xlator_t *this)
+shard_post_rename_lookup_handler(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
-
- local = frame->local;
-
- if (dict_set_uint32 (local->xattr_req, GET_LINK_COUNT, 0))
- gf_msg (this->name, GF_LOG_WARNING, 0,
- SHARD_MSG_DICT_SET_FAILED, "Failed to set "
- GET_LINK_COUNT" in dict");
-
- /* To-Do: Request open-fd count on dst base file */
- STACK_WIND (frame, shard_rename_src_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rename, &local->loc, &local->loc2,
- local->xattr_req);
- return 0;
+ shard_rename_cbk(frame, this);
+ return 0;
}
int
-shard_post_lookup_dst_base_file_handler (call_frame_t *frame, xlator_t *this)
+shard_rename_src_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ struct iatt *preoldparent, struct iatt *postoldparent,
+ struct iatt *prenewparent, struct iatt *postnewparent,
+ dict_t *xdata)
{
- shard_local_t *local = NULL;
+ int ret = 0;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->op_ret < 0) {
- SHARD_STACK_UNWIND (rename, frame, local->op_ret,
- local->op_errno, NULL, NULL, NULL, NULL,
- NULL, NULL);
- return 0;
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto err;
+ }
+ /* Set ctx->refresh to TRUE to force a lookup on disk when
+ * shard_lookup_base_file() is called next to refresh the hard link
+ * count in ctx. Note that this is applicable only to the case where
+ * the rename dst is already existent and sharded.
+ */
+ if ((local->dst_block_size) && (!local->cleanup_required))
+ shard_inode_ctx_set_refresh_flag(local->int_inodelk.loc.inode, this);
+
+ local->prebuf = *buf;
+ local->preoldparent = *preoldparent;
+ local->postoldparent = *postoldparent;
+ local->prenewparent = *prenewparent;
+ local->postnewparent = *postnewparent;
+ if (xdata)
+ local->xattr_rsp = dict_ref(xdata);
+
+ if (local->dst_block_size) {
+ if (local->entrylk_frame) {
+ ret = shard_unlock_entrylk(frame, this);
+ if (ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -ret;
+ }
}
- /* Save dst base file attributes into postbuf so the information is not
- * lost when it is overwritten after lookup on base file of src in
- * shard_lookup_base_file_cbk().
- */
- local->postbuf = local->prebuf;
- shard_rename_src_base_file (frame, this);
- return 0;
+ ret = shard_unlock_inodelk(frame, this);
+ if (ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = -ret;
+ goto err;
+ }
+ if (local->cleanup_required)
+ shard_start_background_deletion(this);
+ }
+
+ /* Now the base file of src, if sharded, is looked up to gather ia_size
+ * and ia_blocks.*/
+ if (local->block_size) {
+ local->tmp_loc.inode = inode_new(this->itable);
+ gf_uuid_copy(local->tmp_loc.gfid, (local->loc.inode)->gfid);
+ shard_refresh_base_file(frame, this, &local->tmp_loc, NULL,
+ shard_post_rename_lookup_handler);
+ } else {
+ shard_rename_cbk(frame, this);
+ }
+ return 0;
+err:
+ shard_common_failure_unwind(local->fop, frame, local->op_ret,
+ local->op_errno);
+ return 0;
}
int
-shard_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
- dict_t *xdata)
+shard_post_lookup_dst_base_file_handler(call_frame_t *frame, xlator_t *this)
{
- int ret = -1;
- uint64_t block_size = 0;
- uint64_t dst_block_size = 0;
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- if (IA_ISDIR (oldloc->inode->ia_type)) {
- STACK_WIND (frame, default_rename_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rename, oldloc, newloc,
- xdata);
- return 0;
- }
+ local = frame->local;
- ret = shard_inode_ctx_get_block_size (oldloc->inode, this, &block_size);
- if ((ret) && (!IA_ISLNK (oldloc->inode->ia_type))) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_CTX_GET_FAILED, "Failed to get block "
- "size from inode ctx of %s",
- uuid_utoa (oldloc->inode->gfid));
- goto err;
- }
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(local->fop, frame, local->op_ret,
+ local->op_errno);
+ return 0;
+ }
- if (newloc->inode)
- ret = shard_inode_ctx_get_block_size (newloc->inode, this,
- &dst_block_size);
- /* The following stack_wind covers the case where:
- * a. the src file is not sharded and dst doesn't exist, OR
- * b. the src and dst both exist but are not sharded.
- */
- if (((!block_size) && (!dst_block_size)) ||
- frame->root->pid == GF_CLIENT_PID_GSYNCD) {
- STACK_WIND (frame, default_rename_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->rename, oldloc, newloc,
- xdata);
- return 0;
- }
+ /* Save dst base file attributes into postbuf so the information is not
+ * lost when it is overwritten after lookup on base file of src in
+ * shard_lookup_base_file_cbk().
+ */
+ local->postbuf = local->prebuf;
+ shard_rename_src_base_file(frame, this);
+ return 0;
+}
- local = mem_get0 (this->local_pool);
- if (!local)
- goto err;
-
- frame->local = local;
- loc_copy (&local->loc, oldloc);
- loc_copy (&local->loc2, newloc);
- local->resolver_base_inode = newloc->inode;
- local->fop = GF_FOP_RENAME;
- local->xattr_req = (xdata) ? dict_ref (xdata) : dict_new();
- if (!local->xattr_req)
- goto err;
-
- local->block_size = block_size;
- local->dst_block_size = dst_block_size;
- if (!this->itable)
- this->itable = (local->loc.inode)->table;
-
- if (local->dst_block_size)
- /* The if block covers the case where the dst file exists and is
- * sharded. So it is important to look up this inode, record its
- * size, before renaming src to dst, so as to NOT lose this
- * information.
- */
- shard_lookup_base_file (frame, this, &local->loc2,
- shard_post_lookup_dst_base_file_handler);
- else
- /* The following block covers the case where the dst either
- * doesn't exist or is NOT sharded. In this case, shard xlator
- * would go ahead and rename src to dst.
- */
- shard_rename_src_base_file (frame, this);
- return 0;
+int
+shard_rename(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+ dict_t *xdata)
+{
+ int ret = -1;
+ uint64_t block_size = 0;
+ uint64_t dst_block_size = 0;
+ shard_local_t *local = NULL;
+
+ if (IA_ISDIR(oldloc->inode->ia_type)) {
+ STACK_WIND(frame, default_rename_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
+ return 0;
+ }
+
+ ret = shard_inode_ctx_get_block_size(oldloc->inode, this, &block_size);
+ if ((ret) && (!IA_ISLNK(oldloc->inode->ia_type))) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_CTX_GET_FAILED,
+ "Failed to get block "
+ "size from inode ctx of %s",
+ uuid_utoa(oldloc->inode->gfid));
+ goto err;
+ }
+
+ if (newloc->inode)
+ ret = shard_inode_ctx_get_block_size(newloc->inode, this,
+ &dst_block_size);
+
+ /* The following stack_wind covers the case where:
+ * a. the src file is not sharded and dst doesn't exist, OR
+ * b. the src and dst both exist but are not sharded.
+ */
+ if (((!block_size) && (!dst_block_size)) ||
+ frame->root->pid == GF_CLIENT_PID_GSYNCD) {
+ STACK_WIND(frame, default_rename_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
+ return 0;
+ }
+
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
+
+ frame->local = local;
+ loc_copy(&local->loc, oldloc);
+ loc_copy(&local->loc2, newloc);
+ local->resolver_base_inode = newloc->inode;
+ local->fop = GF_FOP_RENAME;
+ local->xattr_req = (xdata) ? dict_ref(xdata) : dict_new();
+ if (!local->xattr_req)
+ goto err;
+
+ local->block_size = block_size;
+ local->dst_block_size = dst_block_size;
+ if (!this->itable)
+ this->itable = (local->loc.inode)->table;
+ local->resolve_not = _gf_true;
+
+ /* The following if-block covers the case where the dst file exists
+ * and is sharded.
+ */
+ if (local->dst_block_size) {
+ shard_begin_rm_resolution(frame, this);
+ } else {
+ /* The following block covers the case where the dst either doesn't
+ * exist or is NOT sharded but the src is sharded. In this case, shard
+ * xlator would go ahead and rename src to dst. Once done, it would also
+ * lookup the base shard of src to get the ia_size and ia_blocks xattr
+ * values.
+ */
+ shard_rename_src_base_file(frame, this);
+ }
+ return 0;
err:
- SHARD_STACK_UNWIND (rename, frame, -1, ENOMEM, NULL, NULL, NULL,
- NULL, NULL, NULL);
- return 0;
-
+ shard_common_failure_unwind(GF_FOP_RENAME, frame, -1, ENOMEM);
+ return 0;
}
-
int
-shard_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
- struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+shard_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- int ret = -1;
- shard_local_t *local = NULL;
+ int ret = -1;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret == -1)
- goto unwind;
+ if (op_ret == -1)
+ goto unwind;
- ret = shard_inode_ctx_set (inode, this, stbuf,
- ntoh64 (local->block_size), SHARD_ALL_MASK);
- if (ret)
- gf_msg (this->name, GF_LOG_WARNING, 0,
- SHARD_MSG_INODE_CTX_SET_FAILED, "Failed to set inode "
- "ctx for %s", uuid_utoa (inode->gfid));
+ ret = shard_inode_ctx_set(inode, this, stbuf, local->block_size,
+ SHARD_ALL_MASK);
+ if (ret)
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_INODE_CTX_SET_FAILED,
+ "Failed to set inode "
+ "ctx for %s",
+ uuid_utoa(inode->gfid));
unwind:
- SHARD_STACK_UNWIND (create, frame, op_ret, op_errno, fd, inode, stbuf,
- preparent, postparent, xdata);
- return 0;
+ SHARD_STACK_UNWIND(create, frame, op_ret, op_errno, fd, inode, stbuf,
+ preparent, postparent, xdata);
+ return 0;
}
int
-shard_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+shard_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- shard_local_t *local = NULL;
-
- local = mem_get0 (this->local_pool);
- if (!local)
- goto err;
+ shard_priv_t *priv = NULL;
+ shard_local_t *local = NULL;
- frame->local = local;
+ priv = this->private;
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
- if (!__is_gsyncd_on_shard_dir (frame, loc)) {
- SHARD_INODE_CREATE_INIT (this, local, xdata, loc, err);
- }
+ frame->local = local;
+ local->block_size = priv->block_size;
- STACK_WIND (frame, shard_create_cbk, FIRST_CHILD (this),
- FIRST_CHILD(this)->fops->create, loc, flags, mode, umask,
- fd, xdata);
- return 0;
+ if (!__is_gsyncd_on_shard_dir(frame, loc)) {
+ SHARD_INODE_CREATE_INIT(this, local->block_size, xdata, loc, 0, 0, err);
+ }
+ STACK_WIND(frame, shard_create_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, flags, mode, umask, fd,
+ xdata);
+ return 0;
err:
- SHARD_STACK_UNWIND (create, frame, -1, ENOMEM, NULL, NULL, NULL,
- NULL, NULL, NULL);
- return 0;
-
+ shard_common_failure_unwind(GF_FOP_CREATE, frame, -1, ENOMEM);
+ return 0;
}
int
-shard_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+shard_open_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- /* To-Do: Handle open with O_TRUNC under locks */
- SHARD_STACK_UNWIND (open, frame, op_ret, op_errno, fd, xdata);
- return 0;
+ /* To-Do: Handle open with O_TRUNC under locks */
+ SHARD_STACK_UNWIND(open, frame, op_ret, op_errno, fd, xdata);
+ return 0;
}
int
-shard_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- fd_t *fd, dict_t *xdata)
+shard_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
{
- STACK_WIND (frame, shard_open_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
- return 0;
+ STACK_WIND(frame, shard_open_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
+ return 0;
}
int
-shard_readv_do_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iovec *vector,
- int32_t count, struct iatt *stbuf, struct iobref *iobref,
- dict_t *xdata)
+shard_readv_do_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iovec *vector,
+ int32_t count, struct iatt *stbuf, struct iobref *iobref,
+ dict_t *xdata)
{
- int i = 0;
- int call_count = 0;
- void *address = NULL;
- uint64_t block_num = 0;
- off_t off = 0;
- struct iovec vec = {0,};
- shard_local_t *local = NULL;
- fd_t *anon_fd = cookie;
- shard_inode_ctx_t *ctx = NULL;
+ int i = 0;
+ int call_count = 0;
+ void *address = NULL;
+ uint64_t block_num = 0;
+ off_t off = 0;
+ struct iovec vec = {
+ 0,
+ };
+ shard_local_t *local = NULL;
+ fd_t *anon_fd = cookie;
+ shard_inode_ctx_t *ctx = NULL;
+
+ local = frame->local;
+
+ /* If shard has already seen a failure here before, there is no point
+ * in aggregating subsequent reads, so just go to out.
+ */
+ if (local->op_ret < 0)
+ goto out;
+
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto out;
+ }
- local = frame->local;
+ if (local->op_ret >= 0)
+ local->op_ret += op_ret;
- /* If shard has already seen a failure here before, there is no point
- * in aggregating subsequent reads, so just go to out.
+ shard_inode_ctx_get(anon_fd->inode, this, &ctx);
+ block_num = ctx->block_num;
+
+ if (block_num == local->first_block) {
+ address = local->iobuf->ptr;
+ } else {
+ /* else
+ * address to start writing to = beginning of buffer +
+ * number of bytes until end of first block +
+ * + block_size times number of blocks
+ * between the current block and the first
*/
- if (local->op_ret < 0)
- goto out;
+ address = (char *)local->iobuf->ptr +
+ (local->block_size - (local->offset % local->block_size)) +
+ ((block_num - local->first_block - 1) * local->block_size);
+ }
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto out;
- }
-
- if (local->op_ret >= 0)
- local->op_ret += op_ret;
-
- shard_inode_ctx_get (anon_fd->inode, this, &ctx);
- block_num = ctx->block_num;
-
- if (block_num == local->first_block) {
- address = local->iobuf->ptr;
- } else {
- /* else
- * address to start writing to = beginning of buffer +
- * number of bytes until end of first block +
- * + block_size times number of blocks
- * between the current block and the first
- */
- address = (char *) local->iobuf->ptr + (local->block_size -
- (local->offset % local->block_size)) +
- ((block_num - local->first_block - 1) *
- local->block_size);
- }
-
- for (i = 0; i < count; i++) {
- address = (char *) address + off;
- memcpy (address, vector[i].iov_base, vector[i].iov_len);
- off += vector[i].iov_len;
- }
+ for (i = 0; i < count; i++) {
+ address = (char *)address + off;
+ memcpy(address, vector[i].iov_base, vector[i].iov_len);
+ off += vector[i].iov_len;
+ }
out:
- if (anon_fd)
- fd_unref (anon_fd);
- call_count = shard_call_count_return (frame);
- if (call_count == 0) {
- SHARD_UNSET_ROOT_FS_ID (frame, local);
- if (local->op_ret < 0) {
- SHARD_STACK_UNWIND (readv, frame, local->op_ret,
- local->op_errno, NULL, 0, NULL,
- NULL, NULL);
- } else {
- if (xdata)
- local->xattr_rsp = dict_ref (xdata);
- vec.iov_base = local->iobuf->ptr;
- vec.iov_len = local->total_size;
- SHARD_STACK_UNWIND (readv, frame, local->total_size,
- local->op_errno, &vec, 1,
- &local->prebuf, local->iobref,
- local->xattr_rsp);
- return 0;
- }
+ if (anon_fd)
+ fd_unref(anon_fd);
+ call_count = shard_call_count_return(frame);
+ if (call_count == 0) {
+ SHARD_UNSET_ROOT_FS_ID(frame, local);
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(GF_FOP_READ, frame, local->op_ret,
+ local->op_errno);
+ } else {
+ if (xdata)
+ local->xattr_rsp = dict_ref(xdata);
+ vec.iov_base = local->iobuf->ptr;
+ if (local->offset + local->req_size > local->prebuf.ia_size)
+ local->total_size = local->prebuf.ia_size - local->offset;
+ vec.iov_len = local->total_size;
+ local->op_ret = local->total_size;
+ SHARD_STACK_UNWIND(readv, frame, local->op_ret, local->op_errno,
+ &vec, 1, &local->prebuf, local->iobref,
+ local->xattr_rsp);
+ return 0;
}
+ }
- return 0;
+ return 0;
}
int
-shard_readv_do (call_frame_t *frame, xlator_t *this)
+shard_readv_do(call_frame_t *frame, xlator_t *this)
{
- int i = 0;
- int call_count = 0;
- int last_block = 0;
- int cur_block = 0;
- off_t orig_offset = 0;
- off_t shard_offset = 0;
- size_t read_size = 0;
- size_t remaining_size = 0;
- fd_t *fd = NULL;
- fd_t *anon_fd = NULL;
- shard_local_t *local = NULL;
- gf_boolean_t wind_failed = _gf_false;
+ int i = 0;
+ int call_count = 0;
+ int last_block = 0;
+ int cur_block = 0;
+ off_t orig_offset = 0;
+ off_t shard_offset = 0;
+ size_t read_size = 0;
+ size_t remaining_size = 0;
+ fd_t *fd = NULL;
+ fd_t *anon_fd = NULL;
+ shard_local_t *local = NULL;
+ gf_boolean_t wind_failed = _gf_false;
- local = frame->local;
- fd = local->fd;
+ local = frame->local;
+ fd = local->fd;
- orig_offset = local->offset;
- cur_block = local->first_block;
- last_block = local->last_block;
- remaining_size = local->total_size;
- local->call_count = call_count = local->num_blocks;
+ orig_offset = local->offset;
+ cur_block = local->first_block;
+ last_block = local->last_block;
+ remaining_size = local->total_size;
+ local->call_count = call_count = local->num_blocks;
- SHARD_SET_ROOT_FS_ID (frame, local);
+ SHARD_SET_ROOT_FS_ID(frame, local);
- if (fd->flags & O_DIRECT)
- local->flags = O_DIRECT;
+ if (fd->flags & O_DIRECT)
+ local->flags = O_DIRECT;
- while (cur_block <= last_block) {
- if (wind_failed) {
- shard_readv_do_cbk (frame, (void *) (long) 0, this, -1,
- ENOMEM, NULL, 0, NULL, NULL, NULL);
- goto next;
- }
-
- shard_offset = orig_offset % local->block_size;
- read_size = local->block_size - shard_offset;
- if (read_size > remaining_size)
- read_size = remaining_size;
-
- remaining_size -= read_size;
-
- if (cur_block == 0) {
- anon_fd = fd_ref (fd);
- } else {
- anon_fd = fd_anonymous (local->inode_list[i]);
- if (!anon_fd) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- wind_failed = _gf_true;
- shard_readv_do_cbk (frame,
- (void *) (long) anon_fd,
- this, -1, ENOMEM, NULL, 0,
- NULL, NULL, NULL);
- goto next;
- }
- }
-
- STACK_WIND_COOKIE (frame, shard_readv_do_cbk, anon_fd,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readv, anon_fd,
- read_size, shard_offset, local->flags,
- local->xattr_req);
-
- orig_offset += read_size;
-next:
- cur_block++;
- i++;
- call_count--;
+ while (cur_block <= last_block) {
+ if (wind_failed) {
+ shard_readv_do_cbk(frame, (void *)(long)0, this, -1, ENOMEM, NULL,
+ 0, NULL, NULL, NULL);
+ goto next;
}
- return 0;
-}
-int
-shard_common_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- int shard_block_num = (long) cookie;
- int call_count = 0;
- shard_local_t *local = NULL;
+ shard_offset = orig_offset % local->block_size;
+ read_size = local->block_size - shard_offset;
+ if (read_size > remaining_size)
+ read_size = remaining_size;
- local = frame->local;
+ remaining_size -= read_size;
- if (op_ret < 0) {
- if (op_errno == EEXIST) {
- LOCK (&frame->lock);
- {
- local->eexist_count++;
- }
- UNLOCK (&frame->lock);
- } else {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- }
- gf_msg_debug (this->name, 0, "mknod of shard %d "
- "failed: %s", shard_block_num, strerror (op_errno));
- goto done;
+ if (cur_block == 0) {
+ anon_fd = fd_ref(fd);
+ } else {
+ anon_fd = fd_anonymous(local->inode_list[i]);
+ if (!anon_fd) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ wind_failed = _gf_true;
+ shard_readv_do_cbk(frame, (void *)(long)anon_fd, this, -1,
+ ENOMEM, NULL, 0, NULL, NULL, NULL);
+ goto next;
+ }
}
- shard_link_block_inode (local, shard_block_num, inode, buf);
-
-done:
- call_count = shard_call_count_return (frame);
- if (call_count == 0) {
- SHARD_UNSET_ROOT_FS_ID (frame, local);
- local->create_count = 0;
- local->post_mknod_handler (frame, this);
- }
+ STACK_WIND_COOKIE(frame, shard_readv_do_cbk, anon_fd, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readv, anon_fd, read_size,
+ shard_offset, local->flags, local->xattr_req);
- return 0;
+ orig_offset += read_size;
+ next:
+ cur_block++;
+ i++;
+ call_count--;
+ }
+ return 0;
}
int
-shard_common_resume_mknod (call_frame_t *frame, xlator_t *this,
- shard_post_mknod_fop_handler_t post_mknod_handler)
-{
- int i = 0;
- int shard_idx_iter = 0;
- int last_block = 0;
- int ret = 0;
- int call_count = 0;
- char path[PATH_MAX] = {0,};
- mode_t mode = 0;
- char *bname = NULL;
- shard_priv_t *priv = NULL;
- shard_inode_ctx_t ctx_tmp = {0,};
- shard_local_t *local = NULL;
- gf_boolean_t wind_failed = _gf_false;
- fd_t *fd = NULL;
- loc_t loc = {0,};
- dict_t *xattr_req = NULL;
-
- local = frame->local;
- priv = this->private;
- fd = local->fd;
- shard_idx_iter = local->first_block;
- last_block = local->last_block;
- call_count = local->call_count = local->create_count;
- local->post_mknod_handler = post_mknod_handler;
-
- SHARD_SET_ROOT_FS_ID (frame, local);
-
- ret = shard_inode_ctx_get_all (fd->inode, this, &ctx_tmp);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_CTX_GET_FAILED, "Failed to get inode "
- "ctx for %s", uuid_utoa (fd->inode->gfid));
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- goto err;
- }
- mode = st_mode_from_ia (ctx_tmp.stat.ia_prot, ctx_tmp.stat.ia_type);
+shard_common_mknod_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ int shard_block_num = (long)cookie;
+ int call_count = 0;
+ shard_local_t *local = NULL;
- while (shard_idx_iter <= last_block) {
- if (local->inode_list[i]) {
- shard_idx_iter++;
- i++;
- continue;
- }
+ local = frame->local;
- if (wind_failed) {
- shard_common_mknod_cbk (frame,
- (void *) (long) shard_idx_iter,
- this, -1, ENOMEM, NULL, NULL,
- NULL, NULL, NULL);
- goto next;
- }
+ if (op_ret < 0) {
+ if (op_errno == EEXIST) {
+ LOCK(&frame->lock);
+ {
+ local->eexist_count++;
+ }
+ UNLOCK(&frame->lock);
+ } else {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ }
+ gf_msg_debug(this->name, 0,
+ "mknod of shard %d "
+ "failed: %s",
+ shard_block_num, strerror(op_errno));
+ goto done;
+ }
- shard_make_block_abspath (shard_idx_iter, fd->inode->gfid,
- path, sizeof(path));
+ shard_link_block_inode(local, shard_block_num, inode, buf);
- xattr_req = shard_create_gfid_dict (local->xattr_req);
- if (!xattr_req) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- wind_failed = _gf_true;
- shard_common_mknod_cbk (frame,
- (void *) (long) shard_idx_iter,
- this, -1, ENOMEM, NULL, NULL,
- NULL, NULL, NULL);
- goto next;
- }
+done:
+ call_count = shard_call_count_return(frame);
+ if (call_count == 0) {
+ SHARD_UNSET_ROOT_FS_ID(frame, local);
+ local->create_count = 0;
+ local->post_mknod_handler(frame, this);
+ }
+
+ return 0;
+}
+
+int
+shard_common_resume_mknod(call_frame_t *frame, xlator_t *this,
+ shard_post_mknod_fop_handler_t post_mknod_handler)
+{
+ int i = 0;
+ int shard_idx_iter = 0;
+ int last_block = 0;
+ int ret = 0;
+ int call_count = 0;
+ char path[PATH_MAX] = {
+ 0,
+ };
+ mode_t mode = 0;
+ char *bname = NULL;
+ shard_priv_t *priv = NULL;
+ shard_inode_ctx_t ctx_tmp = {
+ 0,
+ };
+ shard_local_t *local = NULL;
+ gf_boolean_t wind_failed = _gf_false;
+ fd_t *fd = NULL;
+ loc_t loc = {
+ 0,
+ };
+ dict_t *xattr_req = NULL;
+
+ local = frame->local;
+ priv = this->private;
+ fd = local->fd;
+ shard_idx_iter = local->first_block;
+ last_block = local->last_block;
+ call_count = local->call_count = local->create_count;
+ local->post_mknod_handler = post_mknod_handler;
+
+ SHARD_SET_ROOT_FS_ID(frame, local);
+
+ ret = shard_inode_ctx_get_all(fd->inode, this, &ctx_tmp);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_CTX_GET_FAILED,
+ "Failed to get inode "
+ "ctx for %s",
+ uuid_utoa(fd->inode->gfid));
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto err;
+ }
+ mode = st_mode_from_ia(ctx_tmp.stat.ia_prot, ctx_tmp.stat.ia_type);
- bname = strrchr (path, '/') + 1;
- loc.inode = inode_new (this->itable);
- loc.parent = inode_ref (priv->dot_shard_inode);
- ret = inode_path (loc.parent, bname,
- (char **) &(loc.path));
- if (ret < 0 || !(loc.inode)) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_PATH_FAILED, "Inode path failed"
- "on %s, base file gfid = %s", bname,
- uuid_utoa (fd->inode->gfid));
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- wind_failed = _gf_true;
- loc_wipe (&loc);
- dict_unref (xattr_req);
- shard_common_mknod_cbk (frame,
- (void *) (long) shard_idx_iter,
- this, -1, ENOMEM, NULL, NULL,
- NULL, NULL, NULL);
- goto next;
- }
+ while (shard_idx_iter <= last_block) {
+ if (local->inode_list[i]) {
+ shard_idx_iter++;
+ i++;
+ continue;
+ }
- loc.name = strrchr (loc.path, '/');
- if (loc.name)
- loc.name++;
-
- STACK_WIND_COOKIE (frame, shard_common_mknod_cbk,
- (void *) (long) shard_idx_iter,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mknod, &loc,
- mode, ctx_tmp.stat.ia_rdev, 0, xattr_req);
- loc_wipe (&loc);
- dict_unref (xattr_req);
-
-next:
- shard_idx_iter++;
- i++;
- if (!--call_count)
- break;
+ if (wind_failed) {
+ shard_common_mknod_cbk(frame, (void *)(long)shard_idx_iter, this,
+ -1, ENOMEM, NULL, NULL, NULL, NULL, NULL);
+ goto next;
}
- return 0;
+ shard_make_block_abspath(shard_idx_iter, fd->inode->gfid, path,
+ sizeof(path));
+
+ xattr_req = shard_create_gfid_dict(local->xattr_req);
+ if (!xattr_req) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ wind_failed = _gf_true;
+ shard_common_mknod_cbk(frame, (void *)(long)shard_idx_iter, this,
+ -1, ENOMEM, NULL, NULL, NULL, NULL, NULL);
+ goto next;
+ }
+
+ bname = strrchr(path, '/') + 1;
+ loc.inode = inode_new(this->itable);
+ loc.parent = inode_ref(priv->dot_shard_inode);
+ ret = inode_path(loc.parent, bname, (char **)&(loc.path));
+ if (ret < 0 || !(loc.inode)) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_PATH_FAILED,
+ "Inode path failed"
+ "on %s, base file gfid = %s",
+ bname, uuid_utoa(fd->inode->gfid));
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ wind_failed = _gf_true;
+ loc_wipe(&loc);
+ dict_unref(xattr_req);
+ shard_common_mknod_cbk(frame, (void *)(long)shard_idx_iter, this,
+ -1, ENOMEM, NULL, NULL, NULL, NULL, NULL);
+ goto next;
+ }
+
+ loc.name = strrchr(loc.path, '/');
+ if (loc.name)
+ loc.name++;
+
+ STACK_WIND_COOKIE(frame, shard_common_mknod_cbk,
+ (void *)(long)shard_idx_iter, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, &loc, mode,
+ ctx_tmp.stat.ia_rdev, 0, xattr_req);
+ loc_wipe(&loc);
+ dict_unref(xattr_req);
+
+ next:
+ shard_idx_iter++;
+ i++;
+ if (!--call_count)
+ break;
+ }
+
+ return 0;
err:
- /*
- * This block is for handling failure in shard_inode_ctx_get_all().
- * Failures in the while-loop are handled within the loop.
- */
- SHARD_UNSET_ROOT_FS_ID (frame, local);
- post_mknod_handler (frame, this);
- return 0;
+ /*
+ * This block is for handling failure in shard_inode_ctx_get_all().
+ * Failures in the while-loop are handled within the loop.
+ */
+ SHARD_UNSET_ROOT_FS_ID(frame, local);
+ post_mknod_handler(frame, this);
+ return 0;
}
int
-shard_post_mknod_readv_handler (call_frame_t *frame, xlator_t *this);
+shard_post_mknod_readv_handler(call_frame_t *frame, xlator_t *this);
int
-shard_post_lookup_shards_readv_handler (call_frame_t *frame, xlator_t *this)
+shard_post_lookup_shards_readv_handler(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->op_ret < 0) {
- SHARD_STACK_UNWIND (readv, frame, local->op_ret,
- local->op_errno, NULL, 0, NULL, NULL, NULL);
- return 0;
- }
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(GF_FOP_READ, frame, local->op_ret,
+ local->op_errno);
+ return 0;
+ }
- if (local->create_count) {
- shard_common_resume_mknod (frame, this,
- shard_post_mknod_readv_handler);
- } else {
- shard_readv_do (frame, this);
- }
+ if (local->create_count) {
+ shard_common_resume_mknod(frame, this, shard_post_mknod_readv_handler);
+ } else {
+ shard_readv_do(frame, this);
+ }
- return 0;
+ return 0;
}
int
-shard_post_mknod_readv_handler (call_frame_t *frame, xlator_t *this)
+shard_post_mknod_readv_handler(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->op_ret < 0) {
- SHARD_STACK_UNWIND (readv, frame, local->op_ret,
- local->op_errno, NULL, 0, NULL, NULL, NULL);
- return 0;
- }
-
- if (!local->eexist_count) {
- shard_readv_do (frame, this);
- } else {
- local->call_count = local->eexist_count;
- shard_common_lookup_shards (frame, this, local->loc.inode,
- shard_post_lookup_shards_readv_handler);
- }
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(GF_FOP_READ, frame, local->op_ret,
+ local->op_errno);
return 0;
+ }
+
+ if (!local->eexist_count) {
+ shard_readv_do(frame, this);
+ } else {
+ local->call_count = local->eexist_count;
+ shard_common_lookup_shards(frame, this, local->loc.inode,
+ shard_post_lookup_shards_readv_handler);
+ }
+ return 0;
}
int
-shard_post_resolve_readv_handler (call_frame_t *frame, xlator_t *this)
+shard_post_resolve_readv_handler(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
-
- local = frame->local;
+ shard_local_t *local = NULL;
- if (local->op_ret < 0) {
- if (local->op_errno != ENOENT) {
- SHARD_STACK_UNWIND (readv, frame, local->op_ret,
- local->op_errno, NULL, 0, NULL,
- NULL, NULL);
- return 0;
- } else {
- struct iovec vec = {0,};
-
- vec.iov_base = local->iobuf->ptr;
- vec.iov_len = local->total_size;
- SHARD_STACK_UNWIND (readv, frame, local->total_size,
- 0, &vec, 1, &local->prebuf,
- local->iobref, NULL);
- return 0;
- }
- }
+ local = frame->local;
- if (local->call_count) {
- shard_common_lookup_shards (frame, this,
- local->resolver_base_inode,
- shard_post_lookup_shards_readv_handler);
+ if (local->op_ret < 0) {
+ if (local->op_errno != ENOENT) {
+ shard_common_failure_unwind(GF_FOP_READ, frame, local->op_ret,
+ local->op_errno);
+ return 0;
} else {
- shard_readv_do (frame, this);
+ struct iovec vec = {
+ 0,
+ };
+
+ vec.iov_base = local->iobuf->ptr;
+ vec.iov_len = local->total_size;
+ local->op_ret = local->total_size;
+ SHARD_STACK_UNWIND(readv, frame, local->op_ret, 0, &vec, 1,
+ &local->prebuf, local->iobref, NULL);
+ return 0;
}
+ }
- return 0;
+ if (local->call_count) {
+ shard_common_lookup_shards(frame, this, local->resolver_base_inode,
+ shard_post_lookup_shards_readv_handler);
+ } else {
+ shard_readv_do(frame, this);
+ }
+
+ return 0;
}
int
-shard_post_lookup_readv_handler (call_frame_t *frame, xlator_t *this)
+shard_post_lookup_readv_handler(call_frame_t *frame, xlator_t *this)
{
- int ret = 0;
- struct iobuf *iobuf = NULL;
- shard_local_t *local = NULL;
- shard_priv_t *priv = NULL;
+ int ret = 0;
+ struct iobuf *iobuf = NULL;
+ shard_local_t *local = NULL;
+ shard_priv_t *priv = NULL;
- priv = this->private;
- local = frame->local;
-
- if (local->op_ret < 0) {
- SHARD_STACK_UNWIND (readv, frame, local->op_ret,
- local->op_errno, NULL, 0, NULL, NULL, NULL);
- return 0;
- }
+ priv = this->private;
+ local = frame->local;
- if (local->offset >= local->prebuf.ia_size) {
- /* If the read is being performed past the end of the file,
- * unwind the FOP with 0 bytes read as status.
- */
- struct iovec vec = {0,};
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(GF_FOP_READ, frame, local->op_ret,
+ local->op_errno);
+ return 0;
+ }
- iobuf = iobuf_get2 (this->ctx->iobuf_pool, local->req_size);
- if (!iobuf)
- goto err;
+ if (local->offset >= local->prebuf.ia_size) {
+ /* If the read is being performed past the end of the file,
+ * unwind the FOP with 0 bytes read as status.
+ */
+ struct iovec vec = {
+ 0,
+ };
- vec.iov_base = iobuf->ptr;
- vec.iov_len = 0;
- local->iobref = iobref_new ();
- iobref_add (local->iobref, iobuf);
- iobuf_unref (iobuf);
+ iobuf = iobuf_get2(this->ctx->iobuf_pool, local->req_size);
+ if (!iobuf)
+ goto err;
- SHARD_STACK_UNWIND (readv, frame, 0, 0, &vec, 1, &local->prebuf,
- local->iobref, NULL);
- return 0;
- }
+ vec.iov_base = iobuf->ptr;
+ vec.iov_len = 0;
+ local->iobref = iobref_new();
+ iobref_add(local->iobref, iobuf);
+ iobuf_unref(iobuf);
- local->first_block = get_lowest_block (local->offset,
- local->block_size);
+ SHARD_STACK_UNWIND(readv, frame, 0, 0, &vec, 1, &local->prebuf,
+ local->iobref, NULL);
+ return 0;
+ }
- local->total_size = local->req_size;
+ local->first_block = get_lowest_block(local->offset, local->block_size);
- local->last_block = get_highest_block (local->offset, local->total_size,
- local->block_size);
+ local->total_size = local->req_size;
- local->num_blocks = local->last_block - local->first_block + 1;
- local->resolver_base_inode = local->loc.inode;
+ local->last_block = get_highest_block(local->offset, local->total_size,
+ local->block_size);
- local->inode_list = GF_CALLOC (local->num_blocks, sizeof (inode_t *),
- gf_shard_mt_inode_list);
- if (!local->inode_list)
- goto err;
+ local->num_blocks = local->last_block - local->first_block + 1;
+ GF_ASSERT(local->num_blocks > 0);
+ local->resolver_base_inode = local->loc.inode;
- iobuf = iobuf_get2 (this->ctx->iobuf_pool, local->total_size);
- if (!iobuf)
- goto err;
+ local->inode_list = GF_CALLOC(local->num_blocks, sizeof(inode_t *),
+ gf_shard_mt_inode_list);
+ if (!local->inode_list)
+ goto err;
- local->iobref = iobref_new ();
- if (!local->iobref) {
- iobuf_unref (iobuf);
- goto err;
- }
+ iobuf = iobuf_get2(this->ctx->iobuf_pool, local->total_size);
+ if (!iobuf)
+ goto err;
- if (iobref_add (local->iobref, iobuf) != 0) {
- iobuf_unref (iobuf);
- goto err;
- }
+ local->iobref = iobref_new();
+ if (!local->iobref) {
+ iobuf_unref(iobuf);
+ goto err;
+ }
- iobuf_unref (iobuf);
- local->iobuf = iobuf;
- memset (iobuf->ptr, 0, local->total_size);
+ if (iobref_add(local->iobref, iobuf) != 0) {
+ iobuf_unref(iobuf);
+ goto err;
+ }
- local->dot_shard_loc.inode = inode_find (this->itable,
- priv->dot_shard_gfid);
- if (!local->dot_shard_loc.inode) {
- ret = shard_init_internal_dir_loc (this, local,
- SHARD_INTERNAL_DIR_DOT_SHARD);
- if (ret)
- goto err;
- shard_lookup_internal_dir (frame, this,
- shard_post_resolve_readv_handler,
- SHARD_INTERNAL_DIR_DOT_SHARD);
- } else {
- local->post_res_handler = shard_post_resolve_readv_handler;
- shard_refresh_internal_dir (frame, this,
- SHARD_INTERNAL_DIR_DOT_SHARD);
- }
- return 0;
+ memset(iobuf->ptr, 0, local->total_size);
+ iobuf_unref(iobuf);
+ local->iobuf = iobuf;
+ local->dot_shard_loc.inode = inode_find(this->itable, priv->dot_shard_gfid);
+ if (!local->dot_shard_loc.inode) {
+ ret = shard_init_internal_dir_loc(this, local,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
+ if (ret)
+ goto err;
+ shard_lookup_internal_dir(frame, this, shard_post_resolve_readv_handler,
+ SHARD_INTERNAL_DIR_DOT_SHARD);
+ } else {
+ local->post_res_handler = shard_post_resolve_readv_handler;
+ shard_refresh_internal_dir(frame, this, SHARD_INTERNAL_DIR_DOT_SHARD);
+ }
+ return 0;
err:
- SHARD_STACK_UNWIND (readv, frame, -1, ENOMEM, NULL, 0, NULL, NULL,
- NULL);
- return 0;
+ shard_common_failure_unwind(GF_FOP_READ, frame, -1, ENOMEM);
+ return 0;
}
int
-shard_readv (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, uint32_t flags, dict_t *xdata)
+shard_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- int ret = 0;
- uint64_t block_size = 0;
- shard_local_t *local = NULL;
+ int ret = 0;
+ uint64_t block_size = 0;
+ shard_local_t *local = NULL;
- ret = shard_inode_ctx_get_block_size (fd->inode, this, &block_size);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_CTX_GET_FAILED, "Failed to get block "
- "size for %s from its inode ctx",
- uuid_utoa (fd->inode->gfid));
- goto err;
- }
+ ret = shard_inode_ctx_get_block_size(fd->inode, this, &block_size);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_CTX_GET_FAILED,
+ "Failed to get block "
+ "size for %s from its inode ctx",
+ uuid_utoa(fd->inode->gfid));
+ goto err;
+ }
- if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
- /* block_size = 0 means that the file was created before
- * sharding was enabled on the volume.
- */
- STACK_WIND (frame, default_readv_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readv, fd, size, offset,
- flags, xdata);
- return 0;
- }
-
- if (!this->itable)
- this->itable = fd->inode->table;
-
- local = mem_get0 (this->local_pool);
- if (!local)
- goto err;
-
- frame->local = local;
-
- ret = syncbarrier_init (&local->barrier);
- if (ret)
- goto err;
- local->fd = fd_ref (fd);
- local->block_size = block_size;
- local->offset = offset;
- local->req_size = size;
- local->flags = flags;
- local->fop = GF_FOP_READ;
- local->xattr_req = (xdata) ? dict_ref (xdata) : dict_new ();
- if (!local->xattr_req)
- goto err;
-
- local->loc.inode = inode_ref (fd->inode);
- gf_uuid_copy (local->loc.gfid, fd->inode->gfid);
-
- shard_lookup_base_file (frame, this, &local->loc,
- shard_post_lookup_readv_handler);
- return 0;
+ if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
+ /* block_size = 0 means that the file was created before
+ * sharding was enabled on the volume.
+ */
+ STACK_WIND(frame, default_readv_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readv, fd, size, offset, flags,
+ xdata);
+ return 0;
+ }
+
+ if (!this->itable)
+ this->itable = fd->inode->table;
+
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
+
+ frame->local = local;
+
+ ret = syncbarrier_init(&local->barrier);
+ if (ret)
+ goto err;
+ local->fd = fd_ref(fd);
+ local->block_size = block_size;
+ local->offset = offset;
+ local->req_size = size;
+ local->flags = flags;
+ local->fop = GF_FOP_READ;
+ local->xattr_req = (xdata) ? dict_ref(xdata) : dict_new();
+ if (!local->xattr_req)
+ goto err;
+
+ local->loc.inode = inode_ref(fd->inode);
+ gf_uuid_copy(local->loc.gfid, fd->inode->gfid);
+
+ shard_refresh_base_file(frame, this, NULL, fd,
+ shard_post_lookup_readv_handler);
+ return 0;
err:
- SHARD_STACK_UNWIND (readv, frame, -1, ENOMEM, NULL, 0, NULL, NULL,
- NULL);
- return 0;
+ shard_common_failure_unwind(GF_FOP_READ, frame, -1, ENOMEM);
+ return 0;
}
int
-shard_common_inode_write_post_update_size_handler (call_frame_t *frame,
- xlator_t *this)
+shard_common_inode_write_post_update_size_handler(call_frame_t *frame,
+ xlator_t *this)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->op_ret < 0) {
- shard_common_inode_write_failure_unwind (local->fop, frame,
- local->op_ret,
- local->op_errno);
- } else {
- shard_common_inode_write_success_unwind (local->fop, frame,
- local->written_size);
- }
- return 0;
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(local->fop, frame, local->op_ret,
+ local->op_errno);
+ } else {
+ shard_common_inode_write_success_unwind(local->fop, frame,
+ local->written_size);
+ }
+ return 0;
}
static gf_boolean_t
-shard_is_appending_write (shard_local_t *local)
-{
- if (local->fop != GF_FOP_WRITE)
- return _gf_false;
- if (local->flags & O_APPEND)
- return _gf_true;
- if (local->fd->flags & O_APPEND)
- return _gf_true;
+shard_is_appending_write(shard_local_t *local)
+{
+ if (local->fop != GF_FOP_WRITE)
return _gf_false;
+ if (local->flags & O_APPEND)
+ return _gf_true;
+ if (local->fd->flags & O_APPEND)
+ return _gf_true;
+ return _gf_false;
}
int
-__shard_get_delta_size_from_inode_ctx (shard_local_t *local, inode_t *inode,
- xlator_t *this)
+__shard_get_delta_size_from_inode_ctx(shard_local_t *local, inode_t *inode,
+ xlator_t *this)
{
- int ret = -1;
- uint64_t ctx_uint = 0;
- shard_inode_ctx_t *ctx = NULL;
+ int ret = -1;
+ uint64_t ctx_uint = 0;
+ shard_inode_ctx_t *ctx = NULL;
- ret = __inode_ctx_get (inode, this, &ctx_uint);
- if (ret < 0)
- return ret;
+ ret = __inode_ctx_get(inode, this, &ctx_uint);
+ if (ret < 0)
+ return ret;
- ctx = (shard_inode_ctx_t *) ctx_uint;
+ ctx = (shard_inode_ctx_t *)(uintptr_t)ctx_uint;
- if (shard_is_appending_write (local)) {
- local->delta_size = local->total_size;
- } else if (local->offset + local->total_size > ctx->stat.ia_size) {
- local->delta_size = (local->offset + local->total_size) -
- ctx->stat.ia_size;
- } else {
- local->delta_size = 0;
- }
- ctx->stat.ia_size += (local->delta_size);
- local->postbuf = ctx->stat;
+ if (shard_is_appending_write(local)) {
+ local->delta_size = local->total_size;
+ } else if (local->offset + local->total_size > ctx->stat.ia_size) {
+ local->delta_size = (local->offset + local->total_size) -
+ ctx->stat.ia_size;
+ } else {
+ local->delta_size = 0;
+ }
+ ctx->stat.ia_size += (local->delta_size);
+ local->postbuf = ctx->stat;
- return 0;
+ return 0;
}
int
-shard_get_delta_size_from_inode_ctx (shard_local_t *local, inode_t *inode,
- xlator_t *this)
+shard_get_delta_size_from_inode_ctx(shard_local_t *local, inode_t *inode,
+ xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- LOCK (&inode->lock);
- {
- ret = __shard_get_delta_size_from_inode_ctx (local, inode,
- this);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __shard_get_delta_size_from_inode_ctx(local, inode, this);
+ }
+ UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
int
-shard_common_inode_write_do_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, struct iatt *pre,
- struct iatt *post, dict_t *xdata)
+shard_common_inode_write_do_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, struct iatt *pre,
+ struct iatt *post, dict_t *xdata)
{
- int call_count = 0;
- fd_t *anon_fd = cookie;
- shard_local_t *local = NULL;
- glusterfs_fop_t fop = 0;
-
- local = frame->local;
- fop = local->fop;
+ int call_count = 0;
+ fd_t *anon_fd = cookie;
+ shard_local_t *local = NULL;
+ glusterfs_fop_t fop = 0;
- LOCK (&frame->lock);
- {
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- } else {
- local->written_size += op_ret;
- local->delta_blocks += (post->ia_blocks -
- pre->ia_blocks);
- local->delta_size += (post->ia_size - pre->ia_size);
- shard_inode_ctx_set (local->fd->inode, this, post, 0,
- SHARD_MASK_TIMES);
- if (local->fd->inode != anon_fd->inode)
- shard_inode_ctx_add_to_fsync_list (local->fd->inode,
- this,
- anon_fd->inode);
- }
- }
- UNLOCK (&frame->lock);
+ local = frame->local;
+ fop = local->fop;
- if (anon_fd)
- fd_unref (anon_fd);
-
- call_count = shard_call_count_return (frame);
- if (call_count == 0) {
- SHARD_UNSET_ROOT_FS_ID (frame, local);
- if (local->op_ret < 0) {
- shard_common_inode_write_failure_unwind (fop, frame,
- local->op_ret,
- local->op_errno);
- } else {
- shard_get_delta_size_from_inode_ctx (local,
- local->fd->inode,
- this);
- local->hole_size = 0;
- if (xdata)
- local->xattr_rsp = dict_ref (xdata);
- shard_update_file_size (frame, this, local->fd, NULL,
- shard_common_inode_write_post_update_size_handler);
- }
+ LOCK(&frame->lock);
+ {
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ } else {
+ local->written_size += op_ret;
+ GF_ATOMIC_ADD(local->delta_blocks,
+ post->ia_blocks - pre->ia_blocks);
+ local->delta_size += (post->ia_size - pre->ia_size);
+ shard_inode_ctx_set(local->fd->inode, this, post, 0,
+ SHARD_MASK_TIMES);
+ if (local->fd->inode != anon_fd->inode)
+ shard_inode_ctx_add_to_fsync_list(local->fd->inode, this,
+ anon_fd->inode);
+ }
+ }
+ UNLOCK(&frame->lock);
+
+ if (anon_fd)
+ fd_unref(anon_fd);
+
+ call_count = shard_call_count_return(frame);
+ if (call_count == 0) {
+ SHARD_UNSET_ROOT_FS_ID(frame, local);
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(fop, frame, local->op_ret,
+ local->op_errno);
+ } else {
+ shard_get_delta_size_from_inode_ctx(local, local->fd->inode, this);
+ local->hole_size = 0;
+ if (xdata)
+ local->xattr_rsp = dict_ref(xdata);
+ shard_update_file_size(
+ frame, this, local->fd, NULL,
+ shard_common_inode_write_post_update_size_handler);
}
+ }
- return 0;
+ return 0;
}
int
-shard_common_inode_write_wind (call_frame_t *frame, xlator_t *this,
- fd_t *fd, struct iovec *vec, int count,
- off_t shard_offset, size_t size)
+shard_common_inode_write_wind(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vec, int count, off_t shard_offset,
+ size_t size)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- switch (local->fop) {
+ switch (local->fop) {
case GF_FOP_WRITE:
- STACK_WIND_COOKIE (frame, shard_common_inode_write_do_cbk, fd,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->writev, fd, vec,
- count, shard_offset, local->flags,
- local->iobref, local->xattr_req);
- break;
+ STACK_WIND_COOKIE(
+ frame, shard_common_inode_write_do_cbk, fd, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vec, count, shard_offset,
+ local->flags, local->iobref, local->xattr_req);
+ break;
case GF_FOP_FALLOCATE:
- STACK_WIND_COOKIE (frame, shard_common_inode_write_do_cbk, fd,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fallocate, fd,
- local->flags, shard_offset, size,
- local->xattr_req);
- break;
+ STACK_WIND_COOKIE(
+ frame, shard_common_inode_write_do_cbk, fd, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fallocate, fd, local->flags,
+ shard_offset, size, local->xattr_req);
+ break;
case GF_FOP_ZEROFILL:
- STACK_WIND_COOKIE (frame, shard_common_inode_write_do_cbk, fd,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->zerofill, fd,
- shard_offset, size, local->xattr_req);
- break;
+ STACK_WIND_COOKIE(frame, shard_common_inode_write_do_cbk, fd,
+ FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->zerofill, fd,
+ shard_offset, size, local->xattr_req);
+ break;
case GF_FOP_DISCARD:
- STACK_WIND_COOKIE (frame, shard_common_inode_write_do_cbk, fd,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->discard, fd,
- shard_offset, size, local->xattr_req);
- break;
+ STACK_WIND_COOKIE(frame, shard_common_inode_write_do_cbk, fd,
+ FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->discard, fd,
+ shard_offset, size, local->xattr_req);
+ break;
default:
- gf_msg (this->name, GF_LOG_WARNING, 0, SHARD_MSG_INVALID_FOP,
- "Invalid fop id = %d", local->fop);
- break;
- }
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_INVALID_FOP,
+ "Invalid fop id = %d", local->fop);
+ break;
+ }
+ return 0;
+}
+
+int
+shard_common_inode_write_do(call_frame_t *frame, xlator_t *this)
+{
+ int i = 0;
+ int count = 0;
+ int call_count = 0;
+ int last_block = 0;
+ uint32_t cur_block = 0;
+ fd_t *fd = NULL;
+ fd_t *anon_fd = NULL;
+ shard_local_t *local = NULL;
+ struct iovec *vec = NULL;
+ gf_boolean_t wind_failed = _gf_false;
+ gf_boolean_t odirect = _gf_false;
+ off_t orig_offset = 0;
+ off_t shard_offset = 0;
+ off_t vec_offset = 0;
+ size_t remaining_size = 0;
+ size_t shard_write_size = 0;
+
+ local = frame->local;
+ fd = local->fd;
+
+ orig_offset = local->offset;
+ remaining_size = local->total_size;
+ cur_block = local->first_block;
+ local->call_count = call_count = local->num_blocks;
+ last_block = local->last_block;
+
+ SHARD_SET_ROOT_FS_ID(frame, local);
+
+ if (dict_set_uint32(local->xattr_req, GLUSTERFS_WRITE_UPDATE_ATOMIC, 4)) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_DICT_OP_FAILED,
+ "Failed to set " GLUSTERFS_WRITE_UPDATE_ATOMIC
+ " into "
+ "dict: %s",
+ uuid_utoa(fd->inode->gfid));
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ local->call_count = 1;
+ shard_common_inode_write_do_cbk(frame, (void *)(long)0, this, -1,
+ ENOMEM, NULL, NULL, NULL);
return 0;
-}
+ }
-int
-shard_common_inode_write_do (call_frame_t *frame, xlator_t *this)
-{
- int i = 0;
- int count = 0;
- int call_count = 0;
- int last_block = 0;
- uint32_t cur_block = 0;
- fd_t *fd = NULL;
- fd_t *anon_fd = NULL;
- shard_local_t *local = NULL;
- struct iovec *vec = NULL;
- gf_boolean_t wind_failed = _gf_false;
- gf_boolean_t odirect = _gf_false;
- off_t orig_offset = 0;
- off_t shard_offset = 0;
- off_t vec_offset = 0;
- size_t remaining_size = 0;
- size_t shard_write_size = 0;
-
- local = frame->local;
- fd = local->fd;
-
- orig_offset = local->offset;
- remaining_size = local->total_size;
- cur_block = local->first_block;
- local->call_count = call_count = local->num_blocks;
- last_block = local->last_block;
-
- SHARD_SET_ROOT_FS_ID (frame, local);
-
- if (dict_set_uint32 (local->xattr_req,
- GLUSTERFS_WRITE_UPDATE_ATOMIC, 4)) {
- gf_msg (this->name, GF_LOG_ERROR, 0, SHARD_MSG_DICT_SET_FAILED,
- "Failed to set "GLUSTERFS_WRITE_UPDATE_ATOMIC" into "
- "dict: %s", uuid_utoa (fd->inode->gfid));
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- local->call_count = 1;
- shard_common_inode_write_do_cbk (frame, (void *)(long)0, this,
- -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ if ((fd->flags & O_DIRECT) && (local->fop == GF_FOP_WRITE))
+ odirect = _gf_true;
+
+ while (cur_block <= last_block) {
+ if (wind_failed) {
+ shard_common_inode_write_do_cbk(frame, (void *)(long)0, this, -1,
+ ENOMEM, NULL, NULL, NULL);
+ goto next;
}
- if ((fd->flags & O_DIRECT) && (local->fop == GF_FOP_WRITE))
- odirect = _gf_true;
+ shard_offset = orig_offset % local->block_size;
+ shard_write_size = local->block_size - shard_offset;
+ if (shard_write_size > remaining_size)
+ shard_write_size = remaining_size;
- while (cur_block <= last_block) {
- if (wind_failed) {
- shard_common_inode_write_do_cbk (frame,
- (void *) (long) 0,
- this, -1, ENOMEM, NULL,
- NULL, NULL);
- goto next;
- }
+ remaining_size -= shard_write_size;
- shard_offset = orig_offset % local->block_size;
- shard_write_size = local->block_size - shard_offset;
- if (shard_write_size > remaining_size)
- shard_write_size = remaining_size;
-
- remaining_size -= shard_write_size;
-
- if (local->fop == GF_FOP_WRITE) {
- count = iov_subset (local->vector, local->count,
- vec_offset,
- vec_offset + shard_write_size,
- NULL);
-
- vec = GF_CALLOC (count, sizeof (struct iovec),
- gf_shard_mt_iovec);
- if (!vec) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- wind_failed = _gf_true;
- GF_FREE (vec);
- shard_common_inode_write_do_cbk (frame,
- (void *) (long) 0,
- this, -1,
- ENOMEM, NULL,
- NULL, NULL);
- goto next;
- }
- count = iov_subset (local->vector, local->count,
- vec_offset,
- vec_offset + shard_write_size, vec);
- }
+ if (local->fop == GF_FOP_WRITE) {
+ vec = NULL;
+ count = iov_subset(local->vector, local->count, vec_offset,
+ shard_write_size, &vec, 0);
+ if (count < 0) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ wind_failed = _gf_true;
+ shard_common_inode_write_do_cbk(frame, (void *)(long)0, this,
+ -1, ENOMEM, NULL, NULL, NULL);
+ goto next;
+ }
+ }
- if (cur_block == 0) {
- anon_fd = fd_ref (fd);
- } else {
- anon_fd = fd_anonymous (local->inode_list[i]);
- if (!anon_fd) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- wind_failed = _gf_true;
- GF_FREE (vec);
- shard_common_inode_write_do_cbk (frame,
- (void *) (long) anon_fd,
- this, -1,
- ENOMEM, NULL,
- NULL, NULL);
- goto next;
- }
-
- if (local->fop == GF_FOP_WRITE) {
- if (odirect)
- local->flags = O_DIRECT;
- else
- local->flags = GF_ANON_FD_FLAGS;
- }
- }
+ if (cur_block == 0) {
+ anon_fd = fd_ref(fd);
+ } else {
+ anon_fd = fd_anonymous(local->inode_list[i]);
+ if (!anon_fd) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ wind_failed = _gf_true;
+ GF_FREE(vec);
+ shard_common_inode_write_do_cbk(frame, (void *)(long)anon_fd,
+ this, -1, ENOMEM, NULL, NULL,
+ NULL);
+ goto next;
+ }
- shard_common_inode_write_wind (frame, this, anon_fd,
- vec, count, shard_offset,
- shard_write_size);
- if (vec)
- vec_offset += shard_write_size;
- orig_offset += shard_write_size;
- GF_FREE (vec);
- vec = NULL;
-next:
- cur_block++;
- i++;
- call_count--;
- }
- return 0;
+ if (local->fop == GF_FOP_WRITE) {
+ if (odirect)
+ local->flags = O_DIRECT;
+ else
+ local->flags = GF_ANON_FD_FLAGS;
+ }
+ }
+
+ shard_common_inode_write_wind(frame, this, anon_fd, vec, count,
+ shard_offset, shard_write_size);
+ if (vec)
+ vec_offset += shard_write_size;
+ orig_offset += shard_write_size;
+ GF_FREE(vec);
+ vec = NULL;
+ next:
+ cur_block++;
+ i++;
+ call_count--;
+ }
+ return 0;
}
int
-shard_common_inode_write_post_mknod_handler (call_frame_t *frame,
- xlator_t *this);
+shard_common_inode_write_post_mknod_handler(call_frame_t *frame,
+ xlator_t *this);
int
-shard_common_inode_write_post_lookup_shards_handler (call_frame_t *frame,
- xlator_t *this)
+shard_common_inode_write_post_lookup_shards_handler(call_frame_t *frame,
+ xlator_t *this)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->op_ret < 0) {
- shard_common_inode_write_failure_unwind (local->fop, frame,
- local->op_ret,
- local->op_errno);
- return 0;
- }
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(local->fop, frame, local->op_ret,
+ local->op_errno);
+ return 0;
+ }
- if (local->create_count) {
- shard_common_resume_mknod (frame, this,
- shard_common_inode_write_post_mknod_handler);
- } else {
- shard_common_inode_write_do (frame, this);
- }
+ if (local->create_count) {
+ shard_common_resume_mknod(frame, this,
+ shard_common_inode_write_post_mknod_handler);
+ } else {
+ shard_common_inode_write_do(frame, this);
+ }
- return 0;
+ return 0;
}
int
-shard_common_inode_write_post_mknod_handler (call_frame_t *frame,
- xlator_t *this)
+shard_common_inode_write_post_mknod_handler(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->op_ret < 0) {
- shard_common_inode_write_failure_unwind (local->fop, frame,
- local->op_ret,
- local->op_errno);
- return 0;
- }
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(local->fop, frame, local->op_ret,
+ local->op_errno);
+ return 0;
+ }
- if (!local->eexist_count) {
- shard_common_inode_write_do (frame, this);
- } else {
- local->call_count = local->eexist_count;
- shard_common_lookup_shards (frame, this, local->loc.inode,
- shard_common_inode_write_post_lookup_shards_handler);
- }
+ if (!local->eexist_count) {
+ shard_common_inode_write_do(frame, this);
+ } else {
+ local->call_count = local->eexist_count;
+ shard_common_lookup_shards(
+ frame, this, local->loc.inode,
+ shard_common_inode_write_post_lookup_shards_handler);
+ }
- return 0;
+ return 0;
}
int
-shard_mkdir_internal_dir (call_frame_t *frame, xlator_t *this,
- shard_post_resolve_fop_handler_t handler,
- shard_internal_dir_type_t type);
-int
-shard_common_inode_write_post_resolve_handler (call_frame_t *frame,
- xlator_t *this)
+shard_common_inode_write_post_resolve_handler(call_frame_t *frame,
+ xlator_t *this)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->op_ret < 0) {
- shard_common_inode_write_failure_unwind (local->fop, frame,
- local->op_ret,
- local->op_errno);
- return 0;
- }
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(local->fop, frame, local->op_ret,
+ local->op_errno);
+ return 0;
+ }
- if (local->call_count) {
- shard_common_lookup_shards (frame, this,
- local->resolver_base_inode,
- shard_common_inode_write_post_lookup_shards_handler);
- } else {
- shard_common_inode_write_do (frame, this);
- }
+ if (local->call_count) {
+ shard_common_lookup_shards(
+ frame, this, local->resolver_base_inode,
+ shard_common_inode_write_post_lookup_shards_handler);
+ } else if (local->create_count) {
+ shard_common_inode_write_post_lookup_shards_handler(frame, this);
+ } else {
+ shard_common_inode_write_do(frame, this);
+ }
- return 0;
+ return 0;
}
int
-shard_common_inode_write_post_lookup_handler (call_frame_t *frame,
- xlator_t *this)
+shard_common_inode_write_post_lookup_handler(call_frame_t *frame,
+ xlator_t *this)
{
- shard_local_t *local = frame->local;
- shard_priv_t *priv = this->private;
+ shard_local_t *local = frame->local;
+ shard_priv_t *priv = this->private;
- if (local->op_ret < 0) {
- shard_common_inode_write_failure_unwind (local->fop, frame,
- local->op_ret,
- local->op_errno);
- return 0;
- }
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(local->fop, frame, local->op_ret,
+ local->op_errno);
+ return 0;
+ }
- local->postbuf = local->prebuf;
-
- /*Adjust offset to EOF so that correct shard is chosen for append*/
- if (shard_is_appending_write (local))
- local->offset = local->prebuf.ia_size;
-
- local->first_block = get_lowest_block (local->offset,
- local->block_size);
- local->last_block = get_highest_block (local->offset, local->total_size,
- local->block_size);
- local->num_blocks = local->last_block - local->first_block + 1;
- local->inode_list = GF_CALLOC (local->num_blocks, sizeof (inode_t *),
- gf_shard_mt_inode_list);
- if (!local->inode_list) {
- shard_common_inode_write_failure_unwind (local->fop, frame,
- -1, ENOMEM);
- return 0;
- }
+ local->postbuf = local->prebuf;
- gf_msg_trace (this->name, 0, "%s: gfid=%s first_block=%"PRIu32" "
- "last_block=%"PRIu32" num_blocks=%"PRIu32" offset=%"PRId64
- " total_size=%zu flags=%"PRId32"",
- gf_fop_list[local->fop],
- uuid_utoa (local->resolver_base_inode->gfid),
- local->first_block, local->last_block, local->num_blocks,
- local->offset, local->total_size, local->flags);
+ /*Adjust offset to EOF so that correct shard is chosen for append*/
+ if (shard_is_appending_write(local))
+ local->offset = local->prebuf.ia_size;
- local->dot_shard_loc.inode = inode_find (this->itable,
- priv->dot_shard_gfid);
+ local->first_block = get_lowest_block(local->offset, local->block_size);
+ local->last_block = get_highest_block(local->offset, local->total_size,
+ local->block_size);
+ local->num_blocks = local->last_block - local->first_block + 1;
+ GF_ASSERT(local->num_blocks > 0);
+ local->inode_list = GF_CALLOC(local->num_blocks, sizeof(inode_t *),
+ gf_shard_mt_inode_list);
+ if (!local->inode_list) {
+ shard_common_failure_unwind(local->fop, frame, -1, ENOMEM);
+ return 0;
+ }
- if (!local->dot_shard_loc.inode) {
- /*change handler*/
- shard_mkdir_internal_dir (frame, this,
+ gf_msg_trace(this->name, 0,
+ "%s: gfid=%s first_block=%" PRIu64
+ " "
+ "last_block=%" PRIu64 " num_blocks=%" PRIu64 " offset=%" PRId64
+ " total_size=%zu flags=%" PRId32 "",
+ gf_fop_list[local->fop],
+ uuid_utoa(local->resolver_base_inode->gfid),
+ local->first_block, local->last_block, local->num_blocks,
+ local->offset, local->total_size, local->flags);
+
+ local->dot_shard_loc.inode = inode_find(this->itable, priv->dot_shard_gfid);
+
+ if (!local->dot_shard_loc.inode) {
+ /*change handler*/
+ shard_mkdir_internal_dir(frame, this,
shard_common_inode_write_post_resolve_handler,
- SHARD_INTERNAL_DIR_DOT_SHARD);
- } else {
- /*change handler*/
- local->post_res_handler =
- shard_common_inode_write_post_resolve_handler;
- shard_refresh_internal_dir (frame, this,
- SHARD_INTERNAL_DIR_DOT_SHARD);
- }
- return 0;
+ SHARD_INTERNAL_DIR_DOT_SHARD);
+ } else {
+ /*change handler*/
+ local->post_res_handler = shard_common_inode_write_post_resolve_handler;
+ shard_refresh_internal_dir(frame, this, SHARD_INTERNAL_DIR_DOT_SHARD);
+ }
+ return 0;
}
int
-shard_mkdir_internal_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
-{
- inode_t *link_inode = NULL;
- shard_local_t *local = NULL;
- shard_internal_dir_type_t type = (shard_internal_dir_type_t) cookie;
-
- local = frame->local;
-
- SHARD_UNSET_ROOT_FS_ID (frame, local);
-
- if (op_ret == -1) {
- if (op_errno != EEXIST) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto unwind;
- } else {
- gf_msg_debug (this->name, 0, "mkdir on %s failed "
- "with EEXIST. Attempting lookup now",
- shard_internal_dir_string (type));
- shard_lookup_internal_dir (frame, this,
- local->post_res_handler,
- type);
- return 0;
- }
- }
+shard_mkdir_internal_dir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
+{
+ inode_t *link_inode = NULL;
+ shard_local_t *local = NULL;
+ shard_internal_dir_type_t type = (shard_internal_dir_type_t)cookie;
+
+ local = frame->local;
- link_inode = shard_link_internal_dir_inode (local, inode, buf, type);
- if (link_inode != inode) {
- shard_refresh_internal_dir (frame, this, type);
+ SHARD_UNSET_ROOT_FS_ID(frame, local);
+
+ if (op_ret == -1) {
+ if (op_errno != EEXIST) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto unwind;
} else {
- shard_inode_ctx_set_refreshed_flag (link_inode, this);
- shard_common_resolve_shards (frame, this,
- local->post_res_handler);
- }
- return 0;
+ gf_msg_debug(this->name, 0,
+ "mkdir on %s failed "
+ "with EEXIST. Attempting lookup now",
+ shard_internal_dir_string(type));
+ shard_lookup_internal_dir(frame, this, local->post_res_handler,
+ type);
+ return 0;
+ }
+ }
+
+ link_inode = shard_link_internal_dir_inode(local, inode, buf, type);
+ if (link_inode != inode) {
+ shard_refresh_internal_dir(frame, this, type);
+ } else {
+ shard_inode_ctx_mark_dir_refreshed(link_inode, this);
+ shard_common_resolve_shards(frame, this, local->post_res_handler);
+ }
+ return 0;
unwind:
- shard_common_resolve_shards (frame, this, local->post_res_handler);
- return 0;
+ shard_common_resolve_shards(frame, this, local->post_res_handler);
+ return 0;
}
int
-shard_mkdir_internal_dir (call_frame_t *frame, xlator_t *this,
- shard_post_resolve_fop_handler_t handler,
- shard_internal_dir_type_t type)
+shard_mkdir_internal_dir(call_frame_t *frame, xlator_t *this,
+ shard_post_resolve_fop_handler_t handler,
+ shard_internal_dir_type_t type)
{
- int ret = -1;
- shard_local_t *local = NULL;
- shard_priv_t *priv = NULL;
- dict_t *xattr_req = NULL;
- uuid_t *gfid = NULL;
- loc_t *loc = NULL;
- gf_boolean_t free_gfid = _gf_true;
+ int ret = -1;
+ shard_local_t *local = NULL;
+ shard_priv_t *priv = NULL;
+ dict_t *xattr_req = NULL;
+ uuid_t *gfid = NULL;
+ loc_t *loc = NULL;
+ gf_boolean_t free_gfid = _gf_true;
- local = frame->local;
- priv = this->private;
+ local = frame->local;
+ priv = this->private;
- local->post_res_handler = handler;
- gfid = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t);
- if (!gfid)
- goto err;
+ local->post_res_handler = handler;
+ gfid = GF_MALLOC(sizeof(uuid_t), gf_common_mt_uuid_t);
+ if (!gfid)
+ goto err;
- switch (type) {
+ switch (type) {
case SHARD_INTERNAL_DIR_DOT_SHARD:
- gf_uuid_copy (*gfid, priv->dot_shard_gfid);
- loc = &local->dot_shard_loc;
- break;
+ gf_uuid_copy(*gfid, priv->dot_shard_gfid);
+ loc = &local->dot_shard_loc;
+ break;
+ case SHARD_INTERNAL_DIR_DOT_SHARD_REMOVE_ME:
+ gf_uuid_copy(*gfid, priv->dot_shard_rm_gfid);
+ loc = &local->dot_shard_rm_loc;
+ break;
default:
- break;
- }
-
- xattr_req = dict_new ();
- if (!xattr_req)
- goto err;
-
- ret = shard_init_internal_dir_loc (this, local, type);
- if (ret)
- goto err;
-
- ret = dict_set_gfuuid (xattr_req, "gfid-req", *gfid, false);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0, SHARD_MSG_DICT_SET_FAILED,
- "Failed to set gfid-req for %s",
- shard_internal_dir_string (type));
- goto err;
- } else {
- free_gfid = _gf_false;
- }
-
- SHARD_SET_ROOT_FS_ID (frame, local);
-
- STACK_WIND_COOKIE (frame, shard_mkdir_internal_dir_cbk,
- (void *)(long) type, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->mkdir, loc, 0755, 0,
- xattr_req);
- dict_unref (xattr_req);
- return 0;
+ bzero(*gfid, sizeof(uuid_t));
+ break;
+ }
+
+ xattr_req = dict_new();
+ if (!xattr_req)
+ goto err;
+
+ ret = shard_init_internal_dir_loc(this, local, type);
+ if (ret)
+ goto err;
+
+ ret = dict_set_gfuuid(xattr_req, "gfid-req", *gfid, false);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_DICT_OP_FAILED,
+ "Failed to set gfid-req for %s",
+ shard_internal_dir_string(type));
+ goto err;
+ } else {
+ free_gfid = _gf_false;
+ }
+
+ SHARD_SET_ROOT_FS_ID(frame, local);
+
+ STACK_WIND_COOKIE(frame, shard_mkdir_internal_dir_cbk, (void *)(long)type,
+ FIRST_CHILD(this), FIRST_CHILD(this)->fops->mkdir, loc,
+ 0755, 0, xattr_req);
+ dict_unref(xattr_req);
+ return 0;
err:
- if (xattr_req)
- dict_unref (xattr_req);
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- if (free_gfid)
- GF_FREE (gfid);
- handler (frame, this);
- return 0;
+ if (xattr_req)
+ dict_unref(xattr_req);
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ if (free_gfid)
+ GF_FREE(gfid);
+ handler(frame, this);
+ return 0;
}
int
-shard_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
+shard_flush_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- /* To-Do: Wind flush on all shards of the file */
- SHARD_STACK_UNWIND (flush, frame, op_ret, op_errno, xdata);
- return 0;
+ /* To-Do: Wind flush on all shards of the file */
+ SHARD_STACK_UNWIND(flush, frame, op_ret, op_errno, xdata);
+ return 0;
}
int
-shard_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+shard_flush(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- STACK_WIND (frame, shard_flush_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->flush, fd, xdata);
- return 0;
+ STACK_WIND(frame, shard_flush_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->flush, fd, xdata);
+ return 0;
}
int
-__shard_get_timestamps_from_inode_ctx (shard_local_t *local, inode_t *inode,
- xlator_t *this)
+__shard_get_timestamps_from_inode_ctx(shard_local_t *local, inode_t *inode,
+ xlator_t *this)
{
- int ret = -1;
- uint64_t ctx_uint = 0;
- shard_inode_ctx_t *ctx = NULL;
+ int ret = -1;
+ uint64_t ctx_uint = 0;
+ shard_inode_ctx_t *ctx = NULL;
- ret = __inode_ctx_get (inode, this, &ctx_uint);
- if (ret < 0)
- return ret;
+ ret = __inode_ctx_get(inode, this, &ctx_uint);
+ if (ret < 0)
+ return ret;
- ctx = (shard_inode_ctx_t *) ctx_uint;
+ ctx = (shard_inode_ctx_t *)(uintptr_t)ctx_uint;
- local->postbuf.ia_ctime = ctx->stat.ia_ctime;
- local->postbuf.ia_ctime_nsec = ctx->stat.ia_ctime_nsec;
- local->postbuf.ia_atime = ctx->stat.ia_atime;
- local->postbuf.ia_atime_nsec = ctx->stat.ia_atime_nsec;
- local->postbuf.ia_mtime = ctx->stat.ia_mtime;
- local->postbuf.ia_mtime_nsec = ctx->stat.ia_mtime_nsec;
+ local->postbuf.ia_ctime = ctx->stat.ia_ctime;
+ local->postbuf.ia_ctime_nsec = ctx->stat.ia_ctime_nsec;
+ local->postbuf.ia_atime = ctx->stat.ia_atime;
+ local->postbuf.ia_atime_nsec = ctx->stat.ia_atime_nsec;
+ local->postbuf.ia_mtime = ctx->stat.ia_mtime;
+ local->postbuf.ia_mtime_nsec = ctx->stat.ia_mtime_nsec;
- return 0;
+ return 0;
}
int
-shard_get_timestamps_from_inode_ctx (shard_local_t *local, inode_t *inode,
- xlator_t *this)
+shard_get_timestamps_from_inode_ctx(shard_local_t *local, inode_t *inode,
+ xlator_t *this)
{
- int ret = 0;
+ int ret = 0;
- LOCK (&inode->lock);
- {
- ret = __shard_get_timestamps_from_inode_ctx (local, inode,
- this);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __shard_get_timestamps_from_inode_ctx(local, inode, this);
+ }
+ UNLOCK(&inode->lock);
- return ret;
+ return ret;
}
int
-shard_fsync_shards_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+shard_fsync_shards_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
struct iatt *postbuf, dict_t *xdata)
{
- int call_count = 0;
- uint64_t fsync_count = 0;
- fd_t *anon_fd = cookie;
- shard_local_t *local = NULL;
- shard_inode_ctx_t *ctx = NULL;
- shard_inode_ctx_t *base_ictx = NULL;
- inode_t *base_inode = NULL;
+ int call_count = 0;
+ uint64_t fsync_count = 0;
+ fd_t *anon_fd = cookie;
+ shard_local_t *local = NULL;
+ shard_inode_ctx_t *ctx = NULL;
+ shard_inode_ctx_t *base_ictx = NULL;
+ inode_t *base_inode = NULL;
+ gf_boolean_t unref_shard_inode = _gf_false;
- local = frame->local;
- base_inode = local->fd->inode;
+ local = frame->local;
+ base_inode = local->fd->inode;
- if (local->op_ret < 0)
- goto out;
+ if (local->op_ret < 0)
+ goto out;
- LOCK (&frame->lock);
- {
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- UNLOCK (&frame->lock);
- goto out;
- }
- shard_inode_ctx_set (local->fd->inode, this, postbuf, 0,
- SHARD_MASK_TIMES);
- }
- UNLOCK (&frame->lock);
- fd_ctx_get (anon_fd, this, &fsync_count);
+ LOCK(&frame->lock);
+ {
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ UNLOCK(&frame->lock);
+ goto out;
+ }
+ shard_inode_ctx_set(local->fd->inode, this, postbuf, 0,
+ SHARD_MASK_TIMES);
+ }
+ UNLOCK(&frame->lock);
+ fd_ctx_get(anon_fd, this, &fsync_count);
out:
- if (base_inode != anon_fd->inode) {
- LOCK (&base_inode->lock);
- LOCK (&anon_fd->inode->lock);
- {
- __shard_inode_ctx_get (anon_fd->inode, this, &ctx);
- __shard_inode_ctx_get (base_inode, this, &base_ictx);
- if (op_ret == 0)
- ctx->fsync_needed -= fsync_count;
- GF_ASSERT (ctx->fsync_needed >= 0);
- if (ctx->fsync_needed != 0) {
- list_add_tail (&ctx->to_fsync_list,
- &base_ictx->to_fsync_list);
- base_ictx->fsync_count++;
- }
- }
- UNLOCK (&anon_fd->inode->lock);
- UNLOCK (&base_inode->lock);
- }
- if (anon_fd)
- fd_unref (anon_fd);
-
- call_count = shard_call_count_return (frame);
- if (call_count != 0)
- return 0;
-
- if (local->op_ret < 0) {
- SHARD_STACK_UNWIND (fsync, frame, local->op_ret,
- local->op_errno, NULL, NULL, NULL);
- } else {
- shard_get_timestamps_from_inode_ctx (local, base_inode, this);
- SHARD_STACK_UNWIND (fsync, frame, local->op_ret,
- local->op_errno, &local->prebuf,
- &local->postbuf, local->xattr_rsp);
- }
- return 0;
-}
-
-int
-shard_post_lookup_fsync_handler (call_frame_t *frame, xlator_t *this)
-{
- int ret = 0;
- int call_count = 0;
- int fsync_count = 0;
- fd_t *anon_fd = NULL;
- inode_t *base_inode = NULL;
- shard_local_t *local = NULL;
- shard_inode_ctx_t *ctx = NULL;
- shard_inode_ctx_t *iter = NULL;
- struct list_head copy = {0,};
- shard_inode_ctx_t *tmp = NULL;
-
- local = frame->local;
- base_inode = local->fd->inode;
- local->postbuf = local->prebuf;
- INIT_LIST_HEAD (&copy);
-
- if (local->op_ret < 0) {
- SHARD_STACK_UNWIND (fsync, frame, local->op_ret,
- local->op_errno, NULL, NULL, NULL);
- return 0;
- }
-
- LOCK (&base_inode->lock);
+ if (anon_fd && (base_inode != anon_fd->inode)) {
+ LOCK(&base_inode->lock);
+ LOCK(&anon_fd->inode->lock);
{
- __shard_inode_ctx_get (base_inode, this, &ctx);
- list_splice_init (&ctx->to_fsync_list, &copy);
- call_count = ctx->fsync_count;
- ctx->fsync_count = 0;
- }
- UNLOCK (&base_inode->lock);
-
- local->call_count = ++call_count;
-
- /* Send fsync() on the base shard first */
- anon_fd = fd_ref (local->fd);
- STACK_WIND_COOKIE (frame, shard_fsync_shards_cbk, anon_fd,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsync, anon_fd,
- local->datasync, local->xattr_req);
+ __shard_inode_ctx_get(anon_fd->inode, this, &ctx);
+ __shard_inode_ctx_get(base_inode, this, &base_ictx);
+ if (op_ret == 0)
+ ctx->fsync_needed -= fsync_count;
+ GF_ASSERT(ctx->fsync_needed >= 0);
+ if (ctx->fsync_needed != 0) {
+ list_add_tail(&ctx->to_fsync_list, &base_ictx->to_fsync_list);
+ base_ictx->fsync_count++;
+ } else {
+ unref_shard_inode = _gf_true;
+ }
+ }
+ UNLOCK(&anon_fd->inode->lock);
+ UNLOCK(&base_inode->lock);
+ }
+
+ if (unref_shard_inode)
+ inode_unref(anon_fd->inode);
+ if (anon_fd)
+ fd_unref(anon_fd);
+
+ call_count = shard_call_count_return(frame);
+ if (call_count != 0)
+ return 0;
+
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(GF_FOP_FSYNC, frame, local->op_ret,
+ local->op_errno);
+ } else {
+ shard_get_timestamps_from_inode_ctx(local, base_inode, this);
+ SHARD_STACK_UNWIND(fsync, frame, local->op_ret, local->op_errno,
+ &local->prebuf, &local->postbuf, local->xattr_rsp);
+ }
+ return 0;
+}
+
+int
+shard_post_lookup_fsync_handler(call_frame_t *frame, xlator_t *this)
+{
+ int ret = 0;
+ int call_count = 0;
+ int fsync_count = 0;
+ fd_t *anon_fd = NULL;
+ inode_t *base_inode = NULL;
+ shard_local_t *local = NULL;
+ shard_inode_ctx_t *ctx = NULL;
+ shard_inode_ctx_t *iter = NULL;
+ struct list_head copy = {
+ 0,
+ };
+ shard_inode_ctx_t *tmp = NULL;
+
+ local = frame->local;
+ base_inode = local->fd->inode;
+ local->postbuf = local->prebuf;
+ INIT_LIST_HEAD(&copy);
+
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(GF_FOP_FSYNC, frame, local->op_ret,
+ local->op_errno);
+ return 0;
+ }
+
+ LOCK(&base_inode->lock);
+ {
+ __shard_inode_ctx_get(base_inode, this, &ctx);
+ list_splice_init(&ctx->to_fsync_list, &copy);
+ call_count = ctx->fsync_count;
+ ctx->fsync_count = 0;
+ }
+ UNLOCK(&base_inode->lock);
+
+ local->call_count = ++call_count;
+
+ /* Send fsync() on the base shard first */
+ anon_fd = fd_ref(local->fd);
+ STACK_WIND_COOKIE(frame, shard_fsync_shards_cbk, anon_fd, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsync, anon_fd, local->datasync,
+ local->xattr_req);
+ call_count--;
+ anon_fd = NULL;
+
+ list_for_each_entry_safe(iter, tmp, &copy, to_fsync_list)
+ {
+ list_del_init(&iter->to_fsync_list);
+ fsync_count = 0;
+ shard_inode_ctx_get_fsync_count(iter->inode, this, &fsync_count);
+ GF_ASSERT(fsync_count > 0);
+ anon_fd = fd_anonymous(iter->inode);
+ if (!anon_fd) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ gf_msg(this->name, GF_LOG_WARNING, ENOMEM,
+ SHARD_MSG_MEMALLOC_FAILED,
+ "Failed to create "
+ "anon fd to fsync shard");
+ shard_fsync_shards_cbk(frame, (void *)(long)anon_fd, this, -1,
+ ENOMEM, NULL, NULL, NULL);
+ continue;
+ }
+
+ ret = fd_ctx_set(anon_fd, this, fsync_count);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_FD_CTX_SET_FAILED,
+ "Failed to set fd "
+ "ctx for shard inode gfid=%s",
+ uuid_utoa(iter->inode->gfid));
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ shard_fsync_shards_cbk(frame, (void *)(long)anon_fd, this, -1,
+ ENOMEM, NULL, NULL, NULL);
+ continue;
+ }
+ STACK_WIND_COOKIE(frame, shard_fsync_shards_cbk, anon_fd,
+ FIRST_CHILD(this), FIRST_CHILD(this)->fops->fsync,
+ anon_fd, local->datasync, local->xattr_req);
call_count--;
- anon_fd = NULL;
-
- list_for_each_entry_safe (iter, tmp, &copy, to_fsync_list) {
- list_del_init (&iter->to_fsync_list);
- fsync_count = 0;
- shard_inode_ctx_get_fsync_count (iter->inode, this,
- &fsync_count);
- GF_ASSERT (fsync_count > 0);
- anon_fd = fd_anonymous (iter->inode);
- if (!anon_fd) {
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- gf_msg (this->name, GF_LOG_WARNING, ENOMEM,
- SHARD_MSG_MEMALLOC_FAILED, "Failed to create "
- "anon fd to fsync shard");
- shard_fsync_shards_cbk (frame, (void *) (long) anon_fd,
- this, -1, ENOMEM, NULL, NULL,
- NULL);
- continue;
- }
+ }
- ret = fd_ctx_set (anon_fd, this, fsync_count);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_FD_CTX_SET_FAILED, "Failed to set fd "
- "ctx for shard inode gfid=%s",
- uuid_utoa (iter->inode->gfid));
- local->op_ret = -1;
- local->op_errno = ENOMEM;
- shard_fsync_shards_cbk (frame, (void *) (long) anon_fd,
- this, -1, ENOMEM, NULL, NULL,
- NULL);
- continue;
- }
- STACK_WIND_COOKIE (frame, shard_fsync_shards_cbk, anon_fd,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsync, anon_fd,
- local->datasync, local->xattr_req);
- call_count--;
- }
-
- return 0;
+ return 0;
}
int
-shard_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
- dict_t *xdata)
+shard_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+ dict_t *xdata)
{
- int ret = 0;
- uint64_t block_size = 0;
- shard_local_t *local = NULL;
+ int ret = 0;
+ uint64_t block_size = 0;
+ shard_local_t *local = NULL;
- ret = shard_inode_ctx_get_block_size (fd->inode, this, &block_size);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_CTX_GET_FAILED, "Failed to get block "
- "size for %s from its inode ctx",
- uuid_utoa (fd->inode->gfid));
- goto err;
- }
+ ret = shard_inode_ctx_get_block_size(fd->inode, this, &block_size);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_CTX_GET_FAILED,
+ "Failed to get block "
+ "size for %s from its inode ctx",
+ uuid_utoa(fd->inode->gfid));
+ goto err;
+ }
- if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
- STACK_WIND (frame, default_fsync_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsync, fd, datasync,
- xdata);
- return 0;
- }
+ if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
+ STACK_WIND(frame, default_fsync_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsync, fd, datasync, xdata);
+ return 0;
+ }
- if (!this->itable)
- this->itable = fd->inode->table;
+ if (!this->itable)
+ this->itable = fd->inode->table;
- local = mem_get0 (this->local_pool);
- if (!local)
- goto err;
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
- frame->local = local;
+ frame->local = local;
- local->fd = fd_ref (fd);
- local->fop = GF_FOP_FSYNC;
- local->datasync = datasync;
- local->xattr_req = (xdata) ? dict_ref (xdata) : dict_new ();
- if (!local->xattr_req)
- goto err;
+ local->fd = fd_ref(fd);
+ local->fop = GF_FOP_FSYNC;
+ local->datasync = datasync;
+ local->xattr_req = (xdata) ? dict_ref(xdata) : dict_new();
+ if (!local->xattr_req)
+ goto err;
- local->loc.inode = inode_ref (fd->inode);
- gf_uuid_copy (local->loc.gfid, fd->inode->gfid);
+ local->loc.inode = inode_ref(fd->inode);
+ gf_uuid_copy(local->loc.gfid, fd->inode->gfid);
- shard_lookup_base_file (frame, this, &local->loc,
- shard_post_lookup_fsync_handler);
- return 0;
+ shard_refresh_base_file(frame, this, NULL, fd,
+ shard_post_lookup_fsync_handler);
+ return 0;
err:
- SHARD_STACK_UNWIND (fsync, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ shard_common_failure_unwind(GF_FOP_FSYNC, frame, -1, ENOMEM);
+ return 0;
}
int
-shard_readdir_past_dot_shard_cbk (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, gf_dirent_t *orig_entries,
- dict_t *xdata)
+shard_readdir_past_dot_shard_cbk(call_frame_t *frame, void *cookie,
+ xlator_t *this, int32_t op_ret,
+ int32_t op_errno, gf_dirent_t *orig_entries,
+ dict_t *xdata)
{
- gf_dirent_t *entry = NULL;
- gf_dirent_t *tmp = NULL;
- shard_local_t *local = NULL;
+ gf_dirent_t *entry = NULL;
+ gf_dirent_t *tmp = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0)
- goto unwind;
+ if (op_ret < 0)
+ goto unwind;
- list_for_each_entry_safe (entry, tmp, (&orig_entries->list), list) {
+ list_for_each_entry_safe(entry, tmp, (&orig_entries->list), list)
+ {
+ list_del_init(&entry->list);
+ list_add_tail(&entry->list, &local->entries_head.list);
- list_del_init (&entry->list);
- list_add_tail (&entry->list, &local->entries_head.list);
+ if (!entry->dict)
+ continue;
- if (!entry->dict)
- continue;
+ if (IA_ISDIR(entry->d_stat.ia_type))
+ continue;
- if (IA_ISDIR (entry->d_stat.ia_type))
- continue;
-
- if (dict_get (entry->dict, GF_XATTR_SHARD_FILE_SIZE))
- shard_modify_size_and_block_count (&entry->d_stat,
- entry->dict);
- if (!entry->inode)
- continue;
+ if (dict_get(entry->dict, GF_XATTR_SHARD_FILE_SIZE))
+ shard_modify_size_and_block_count(&entry->d_stat, entry->dict);
+ if (!entry->inode)
+ continue;
- shard_inode_ctx_update (entry->inode, this, entry->dict,
- &entry->d_stat);
- }
- local->op_ret += op_ret;
+ shard_inode_ctx_update(entry->inode, this, entry->dict, &entry->d_stat);
+ }
+ local->op_ret += op_ret;
unwind:
- if (local->fop == GF_FOP_READDIR)
- SHARD_STACK_UNWIND (readdir, frame, local->op_ret,
- local->op_errno,
- &local->entries_head, xdata);
- else
- SHARD_STACK_UNWIND (readdirp, frame, op_ret, op_errno,
- &local->entries_head, xdata);
- return 0;
+ if (local->fop == GF_FOP_READDIR)
+ SHARD_STACK_UNWIND(readdir, frame, local->op_ret, local->op_errno,
+ &local->entries_head, xdata);
+ else
+ SHARD_STACK_UNWIND(readdirp, frame, op_ret, op_errno,
+ &local->entries_head, xdata);
+ return 0;
}
int32_t
-shard_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *orig_entries,
- dict_t *xdata)
+shard_readdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *orig_entries,
+ dict_t *xdata)
{
- fd_t *fd = NULL;
- gf_dirent_t *entry = NULL;
- gf_dirent_t *tmp = NULL;
- shard_local_t *local = NULL;
- gf_boolean_t last_entry = _gf_false;
+ fd_t *fd = NULL;
+ gf_dirent_t *entry = NULL;
+ gf_dirent_t *tmp = NULL;
+ shard_local_t *local = NULL;
+ gf_boolean_t last_entry = _gf_false;
- local = frame->local;
- fd = local->fd;
+ local = frame->local;
+ fd = local->fd;
- if (op_ret < 0)
- goto unwind;
+ if (op_ret < 0)
+ goto unwind;
- list_for_each_entry_safe (entry, tmp, (&orig_entries->list), list) {
- if (last_entry)
- last_entry = _gf_false;
+ list_for_each_entry_safe(entry, tmp, (&orig_entries->list), list)
+ {
+ if (last_entry)
+ last_entry = _gf_false;
- if (__is_root_gfid (fd->inode->gfid) &&
- !(strcmp (entry->d_name, GF_SHARD_DIR))) {
- local->offset = entry->d_off;
- op_ret--;
- last_entry = _gf_true;
- continue;
- }
+ if (__is_root_gfid(fd->inode->gfid) &&
+ !(strcmp(entry->d_name, GF_SHARD_DIR))) {
+ local->offset = entry->d_off;
+ op_ret--;
+ last_entry = _gf_true;
+ continue;
+ }
- list_del_init (&entry->list);
- list_add_tail (&entry->list, &local->entries_head.list);
+ list_del_init(&entry->list);
+ list_add_tail(&entry->list, &local->entries_head.list);
- if (!entry->dict)
- continue;
+ if (!entry->dict)
+ continue;
- if (IA_ISDIR (entry->d_stat.ia_type))
- continue;
-
- if (dict_get (entry->dict, GF_XATTR_SHARD_FILE_SIZE) &&
- frame->root->pid != GF_CLIENT_PID_GSYNCD)
- shard_modify_size_and_block_count (&entry->d_stat,
- entry->dict);
+ if (IA_ISDIR(entry->d_stat.ia_type))
+ continue;
- if (!entry->inode)
- continue;
+ if (dict_get(entry->dict, GF_XATTR_SHARD_FILE_SIZE) &&
+ frame->root->pid != GF_CLIENT_PID_GSYNCD)
+ shard_modify_size_and_block_count(&entry->d_stat, entry->dict);
- shard_inode_ctx_update (entry->inode, this, entry->dict,
- &entry->d_stat);
- }
+ if (!entry->inode)
+ continue;
- local->op_ret = op_ret;
+ shard_inode_ctx_update(entry->inode, this, entry->dict, &entry->d_stat);
+ }
- if (last_entry) {
- if (local->fop == GF_FOP_READDIR)
- STACK_WIND (frame, shard_readdir_past_dot_shard_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdir, local->fd,
- local->readdir_size, local->offset,
- local->xattr_req);
- else
- STACK_WIND (frame, shard_readdir_past_dot_shard_cbk,
- FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdirp,
- local->fd, local->readdir_size,
- local->offset, local->xattr_req);
- return 0;
- }
+ local->op_ret = op_ret;
-unwind:
+ if (last_entry) {
if (local->fop == GF_FOP_READDIR)
- SHARD_STACK_UNWIND (readdir, frame, op_ret, op_errno,
- &local->entries_head, xdata);
+ STACK_WIND(frame, shard_readdir_past_dot_shard_cbk,
+ FIRST_CHILD(this), FIRST_CHILD(this)->fops->readdir,
+ local->fd, local->readdir_size, local->offset,
+ local->xattr_req);
else
- SHARD_STACK_UNWIND (readdirp, frame, op_ret, op_errno,
- &local->entries_head, xdata);
+ STACK_WIND(frame, shard_readdir_past_dot_shard_cbk,
+ FIRST_CHILD(this), FIRST_CHILD(this)->fops->readdirp,
+ local->fd, local->readdir_size, local->offset,
+ local->xattr_req);
return 0;
-}
+ }
-
-int
-shard_readdir_do (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, int whichop, dict_t *xdata)
-{
- int ret = 0;
- shard_local_t *local = NULL;
-
- local = mem_get0 (this->local_pool);
- if (!local) {
- goto err;
+unwind:
+ if (local->fop == GF_FOP_READDIR)
+ SHARD_STACK_UNWIND(readdir, frame, op_ret, op_errno,
+ &local->entries_head, xdata);
+ else
+ SHARD_STACK_UNWIND(readdirp, frame, op_ret, op_errno,
+ &local->entries_head, xdata);
+ return 0;
+}
+
+int
+shard_readdir_do(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, int whichop, dict_t *xdata)
+{
+ int ret = 0;
+ shard_local_t *local = NULL;
+
+ local = mem_get0(this->local_pool);
+ if (!local) {
+ goto err;
+ }
+
+ frame->local = local;
+
+ local->fd = fd_ref(fd);
+ local->fop = whichop;
+ local->readdir_size = size;
+ INIT_LIST_HEAD(&local->entries_head.list);
+ local->list_inited = _gf_true;
+
+ if (whichop == GF_FOP_READDIR) {
+ STACK_WIND(frame, shard_readdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdir, fd, size, offset, xdata);
+ } else {
+ local->xattr_req = (xdata) ? dict_ref(xdata) : dict_new();
+ SHARD_MD_READ_FOP_INIT_REQ_DICT(this, local->xattr_req, fd->inode->gfid,
+ local, err);
+ ret = dict_set_uint64(local->xattr_req, GF_XATTR_SHARD_BLOCK_SIZE, 0);
+ if (ret) {
+ gf_log(this->name, GF_LOG_WARNING,
+ "Failed to set "
+ "dict value: key:%s, directory gfid=%s",
+ GF_XATTR_SHARD_BLOCK_SIZE, uuid_utoa(fd->inode->gfid));
+ goto err;
}
- frame->local = local;
+ STACK_WIND(frame, shard_readdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->readdirp, fd, size, offset,
+ local->xattr_req);
+ }
- local->fd = fd_ref (fd);
- local->fop = whichop;
- local->readdir_size = size;
- INIT_LIST_HEAD (&local->entries_head.list);
- local->list_inited = _gf_true;
-
- if (whichop == GF_FOP_READDIR) {
- STACK_WIND (frame, shard_readdir_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdir, fd, size, offset,
- xdata);
- } else {
- local->xattr_req = (xdata) ? dict_ref (xdata) : dict_new ();
- SHARD_MD_READ_FOP_INIT_REQ_DICT (this, local->xattr_req,
- fd->inode->gfid, local, err);
- ret = dict_set_uint64 (local->xattr_req,
- GF_XATTR_SHARD_BLOCK_SIZE, 0);
- if (ret) {
- gf_log (this->name, GF_LOG_WARNING, "Failed to set "
- "dict value: key:%s, directory gfid=%s",
- GF_XATTR_SHARD_BLOCK_SIZE,
- uuid_utoa (fd->inode->gfid));
- goto err;
- }
-
- STACK_WIND (frame, shard_readdir_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->readdirp, fd, size, offset,
- local->xattr_req);
- }
-
- return 0;
+ return 0;
err:
- STACK_UNWIND_STRICT (readdir, frame, -1, ENOMEM,
- NULL, NULL);
- return 0;
-
+ STACK_UNWIND_STRICT(readdir, frame, -1, ENOMEM, NULL, NULL);
+ return 0;
}
-
int32_t
-shard_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, dict_t *xdata)
+shard_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *xdata)
{
- shard_readdir_do (frame, this, fd, size, offset, GF_FOP_READDIR, xdata);
- return 0;
+ shard_readdir_do(frame, this, fd, size, offset, GF_FOP_READDIR, xdata);
+ return 0;
}
-
int32_t
-shard_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, dict_t *xdata)
+shard_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, dict_t *xdata)
{
- shard_readdir_do (frame, this, fd, size, offset, GF_FOP_READDIRP,
- xdata);
- return 0;
+ shard_readdir_do(frame, this, fd, size, offset, GF_FOP_READDIRP, xdata);
+ return 0;
}
int32_t
-shard_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+shard_modify_and_set_iatt_in_dict(dict_t *xdata, shard_local_t *local,
+ char *key)
{
- int op_errno = EINVAL;
-
- if (frame->root->pid != GF_CLIENT_PID_GSYNCD) {
- GF_IF_NATIVE_XATTR_GOTO (SHARD_XATTR_PREFIX"*",
- name, op_errno, out);
- }
-
- if (xdata && (frame->root->pid != GF_CLIENT_PID_GSYNCD)) {
- dict_del (xdata, GF_XATTR_SHARD_BLOCK_SIZE);
- dict_del (xdata, GF_XATTR_SHARD_FILE_SIZE);
- }
+ int ret = 0;
+ struct iatt *tmpbuf = NULL;
+ struct iatt *stbuf = NULL;
+ data_t *data = NULL;
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->removexattr, loc, name,
- xdata);
+ if (!xdata)
return 0;
-out:
- SHARD_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
+ data = dict_get(xdata, key);
+ if (!data)
return 0;
+
+ tmpbuf = data_to_iatt(data, key);
+ stbuf = GF_MALLOC(sizeof(struct iatt), gf_common_mt_char);
+ if (stbuf == NULL) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto err;
+ }
+ *stbuf = *tmpbuf;
+ stbuf->ia_size = local->prebuf.ia_size;
+ stbuf->ia_blocks = local->prebuf.ia_blocks;
+ ret = dict_set_iatt(xdata, key, stbuf, false);
+ if (ret < 0) {
+ local->op_ret = -1;
+ local->op_errno = ENOMEM;
+ goto err;
+ }
+ return 0;
+
+err:
+ GF_FREE(stbuf);
+ return -1;
}
int32_t
-shard_fremovexattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
+shard_common_remove_xattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- int op_errno = EINVAL;
-
- if (frame->root->pid != GF_CLIENT_PID_GSYNCD) {
- GF_IF_NATIVE_XATTR_GOTO (SHARD_XATTR_PREFIX"*",
- name, op_errno, out);
- }
+ int ret = -1;
+ shard_local_t *local = NULL;
- if (xdata && (frame->root->pid != GF_CLIENT_PID_GSYNCD)) {
- dict_del (xdata, GF_XATTR_SHARD_BLOCK_SIZE);
- dict_del (xdata, GF_XATTR_SHARD_FILE_SIZE);
- }
+ local = frame->local;
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fremovexattr, fd, name,
- xdata);
- return 0;
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto err;
+ }
+
+ ret = shard_modify_and_set_iatt_in_dict(xdata, local, GF_PRESTAT);
+ if (ret < 0)
+ goto err;
+
+ ret = shard_modify_and_set_iatt_in_dict(xdata, local, GF_POSTSTAT);
+ if (ret < 0)
+ goto err;
+
+ if (local->fd)
+ SHARD_STACK_UNWIND(fremovexattr, frame, local->op_ret, local->op_errno,
+ xdata);
+ else
+ SHARD_STACK_UNWIND(removexattr, frame, local->op_ret, local->op_errno,
+ xdata);
+ return 0;
-out:
- SHARD_STACK_UNWIND (fremovexattr, frame, -1, op_errno, NULL);
- return 0;
+err:
+ shard_common_failure_unwind(local->fop, frame, local->op_ret,
+ local->op_errno);
+ return 0;
}
int32_t
-shard_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *dict,
- dict_t *xdata)
+shard_post_lookup_remove_xattr_handler(call_frame_t *frame, xlator_t *this)
{
- if (op_ret < 0)
- goto unwind;
+ shard_local_t *local = NULL;
- if (dict && (frame->root->pid != GF_CLIENT_PID_GSYNCD)) {
- dict_del (dict, GF_XATTR_SHARD_BLOCK_SIZE);
- dict_del (dict, GF_XATTR_SHARD_FILE_SIZE);
- }
+ local = frame->local;
-unwind:
- SHARD_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, dict, xdata);
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(local->fop, frame, local->op_ret,
+ local->op_errno);
return 0;
+ }
+
+ if (local->fd)
+ STACK_WIND(frame, shard_common_remove_xattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fremovexattr, local->fd,
+ local->name, local->xattr_req);
+ else
+ STACK_WIND(frame, shard_common_remove_xattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, &local->loc,
+ local->name, local->xattr_req);
+ return 0;
}
int32_t
-shard_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- const char *name, dict_t *xdata)
-{
- int op_errno = EINVAL;
-
- if ((frame->root->pid != GF_CLIENT_PID_GSYNCD) &&
- (name) && (!strncmp (name, SHARD_XATTR_PREFIX,
- strlen (SHARD_XATTR_PREFIX)))) {
- op_errno = ENODATA;
- goto out;
- }
-
- STACK_WIND (frame, shard_fgetxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
+shard_common_remove_xattr(call_frame_t *frame, xlator_t *this,
+ glusterfs_fop_t fop, loc_t *loc, fd_t *fd,
+ const char *name, dict_t *xdata)
+{
+ int ret = -1;
+ int op_errno = ENOMEM;
+ uint64_t block_size = 0;
+ shard_local_t *local = NULL;
+ inode_t *inode = loc ? loc->inode : fd->inode;
+
+ if ((IA_ISDIR(inode->ia_type)) || (IA_ISLNK(inode->ia_type))) {
+ if (loc)
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, loc, name,
+ xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fremovexattr, fd, name,
+ xdata);
return 0;
-
-out:
- SHARD_STACK_UNWIND (fgetxattr, frame, -1, op_errno, NULL, NULL);
+ }
+
+ /* If shard's special xattrs are attempted to be removed,
+ * fail the fop with EPERM (except if the client is gsyncd).
+ */
+ if (frame->root->pid != GF_CLIENT_PID_GSYNCD) {
+ GF_IF_NATIVE_XATTR_GOTO(SHARD_XATTR_PREFIX "*", name, op_errno, err);
+ }
+
+ /* Repeat the same check for bulk-removexattr */
+ if (xdata && (frame->root->pid != GF_CLIENT_PID_GSYNCD)) {
+ dict_del(xdata, GF_XATTR_SHARD_BLOCK_SIZE);
+ dict_del(xdata, GF_XATTR_SHARD_FILE_SIZE);
+ }
+
+ ret = shard_inode_ctx_get_block_size(inode, this, &block_size);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_CTX_GET_FAILED,
+ "Failed to get block size from inode ctx of %s",
+ uuid_utoa(inode->gfid));
+ goto err;
+ }
+
+ if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
+ if (loc)
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->removexattr, loc, name,
+ xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fremovexattr, fd, name,
+ xdata);
return 0;
+ }
+
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
+
+ frame->local = local;
+ local->fop = fop;
+ if (loc) {
+ if (loc_copy(&local->loc, loc) != 0)
+ goto err;
+ }
+
+ if (fd) {
+ local->fd = fd_ref(fd);
+ local->loc.inode = inode_ref(fd->inode);
+ gf_uuid_copy(local->loc.gfid, fd->inode->gfid);
+ }
+
+ if (name) {
+ local->name = gf_strdup(name);
+ if (!local->name)
+ goto err;
+ }
+
+ if (xdata)
+ local->xattr_req = dict_ref(xdata);
+
+ shard_refresh_base_file(frame, this, loc, fd,
+ shard_post_lookup_remove_xattr_handler);
+ return 0;
+err:
+ shard_common_failure_unwind(fop, frame, -1, op_errno);
+ return 0;
}
+int32_t
+shard_removexattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
+{
+ shard_common_remove_xattr(frame, this, GF_FOP_REMOVEXATTR, loc, NULL, name,
+ xdata);
+ return 0;
+}
+
+int32_t
+shard_fremovexattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ const char *name, dict_t *xdata)
+{
+ shard_common_remove_xattr(frame, this, GF_FOP_FREMOVEXATTR, NULL, fd, name,
+ xdata);
+ return 0;
+}
int32_t
-shard_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+shard_fgetxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, dict_t *dict,
dict_t *xdata)
{
- if (op_ret < 0)
- goto unwind;
+ if (op_ret < 0)
+ goto unwind;
- if (dict && (frame->root->pid != GF_CLIENT_PID_GSYNCD)) {
- dict_del (dict, GF_XATTR_SHARD_BLOCK_SIZE);
- dict_del (dict, GF_XATTR_SHARD_FILE_SIZE);
- }
+ if (dict && (frame->root->pid != GF_CLIENT_PID_GSYNCD)) {
+ dict_del(dict, GF_XATTR_SHARD_BLOCK_SIZE);
+ dict_del(dict, GF_XATTR_SHARD_FILE_SIZE);
+ }
unwind:
- SHARD_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
- return 0;
+ SHARD_STACK_UNWIND(fgetxattr, frame, op_ret, op_errno, dict, xdata);
+ return 0;
}
int32_t
-shard_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+shard_fgetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, const char *name,
+ dict_t *xdata)
{
- int op_errno = EINVAL;
+ int op_errno = EINVAL;
- if ((frame->root->pid != GF_CLIENT_PID_GSYNCD) &&
- (name) && (!strncmp (name, SHARD_XATTR_PREFIX,
- strlen (SHARD_XATTR_PREFIX)))) {
- op_errno = ENODATA;
- goto out;
- }
-
- STACK_WIND (frame, shard_getxattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
- return 0;
+ if ((frame->root->pid != GF_CLIENT_PID_GSYNCD) && (name) &&
+ (!strncmp(name, SHARD_XATTR_PREFIX, SLEN(SHARD_XATTR_PREFIX)))) {
+ op_errno = ENODATA;
+ goto out;
+ }
+ STACK_WIND(frame, shard_fgetxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
+ return 0;
out:
- SHARD_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
- return 0;
+ shard_common_failure_unwind(GF_FOP_FGETXATTR, frame, -1, op_errno);
+ return 0;
}
int32_t
-shard_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
- int32_t flags, dict_t *xdata)
+shard_getxattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *dict,
+ dict_t *xdata)
{
- int op_errno = EINVAL;
-
- if (frame->root->pid != GF_CLIENT_PID_GSYNCD) {
- GF_IF_INTERNAL_XATTR_GOTO (SHARD_XATTR_PREFIX"*", dict,
- op_errno, out);
- }
+ if (op_ret < 0)
+ goto unwind;
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags,
- xdata);
- return 0;
+ if (dict && (frame->root->pid != GF_CLIENT_PID_GSYNCD)) {
+ dict_del(dict, GF_XATTR_SHARD_BLOCK_SIZE);
+ dict_del(dict, GF_XATTR_SHARD_FILE_SIZE);
+ }
-out:
- SHARD_STACK_UNWIND (fsetxattr, frame, -1, op_errno, NULL);
- return 0;
+unwind:
+ SHARD_STACK_UNWIND(getxattr, frame, op_ret, op_errno, dict, xdata);
+ return 0;
}
int32_t
-shard_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
- int32_t flags, dict_t *xdata)
+shard_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- int op_errno = EINVAL;
+ int op_errno = EINVAL;
- if (frame->root->pid != GF_CLIENT_PID_GSYNCD) {
- GF_IF_INTERNAL_XATTR_GOTO (SHARD_XATTR_PREFIX"*", dict,
- op_errno, out);
- }
-
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setxattr, loc, dict, flags,
- xdata);
- return 0;
+ if ((frame->root->pid != GF_CLIENT_PID_GSYNCD) && (name) &&
+ (!strncmp(name, SHARD_XATTR_PREFIX, sizeof(SHARD_XATTR_PREFIX) - 1))) {
+ op_errno = ENODATA;
+ goto out;
+ }
+ STACK_WIND(frame, shard_getxattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+ return 0;
out:
- SHARD_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
- return 0;
+ shard_common_failure_unwind(GF_FOP_GETXATTR, frame, -1, op_errno);
+ return 0;
}
-int
-shard_post_setattr_handler (call_frame_t *frame, xlator_t *this)
+int32_t
+shard_common_set_xattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- shard_local_t *local = NULL;
+ int ret = -1;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (local->fop == GF_FOP_SETATTR) {
- if (local->op_ret >= 0)
- shard_inode_ctx_set (local->loc.inode, this,
- &local->postbuf, 0,
- SHARD_LOOKUP_MASK);
- SHARD_STACK_UNWIND (setattr, frame, local->op_ret,
- local->op_errno, &local->prebuf,
- &local->postbuf, local->xattr_rsp);
- } else if (local->fop == GF_FOP_FSETATTR) {
- if (local->op_ret >= 0)
- shard_inode_ctx_set (local->fd->inode, this,
- &local->postbuf, 0,
- SHARD_LOOKUP_MASK);
- SHARD_STACK_UNWIND (fsetattr, frame, local->op_ret,
- local->op_errno, &local->prebuf,
- &local->postbuf, local->xattr_rsp);
- }
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto err;
+ }
+
+ ret = shard_modify_and_set_iatt_in_dict(xdata, local, GF_PRESTAT);
+ if (ret < 0)
+ goto err;
+
+ ret = shard_modify_and_set_iatt_in_dict(xdata, local, GF_POSTSTAT);
+ if (ret < 0)
+ goto err;
+
+ if (local->fd)
+ SHARD_STACK_UNWIND(fsetxattr, frame, local->op_ret, local->op_errno,
+ xdata);
+ else
+ SHARD_STACK_UNWIND(setxattr, frame, local->op_ret, local->op_errno,
+ xdata);
+ return 0;
- return 0;
+err:
+ shard_common_failure_unwind(local->fop, frame, local->op_ret,
+ local->op_errno);
+ return 0;
}
-int
-shard_common_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
- struct iatt *postbuf, dict_t *xdata)
+int32_t
+shard_post_lookup_set_xattr_handler(call_frame_t *frame, xlator_t *this)
{
- shard_local_t *local = NULL;
+ shard_local_t *local = NULL;
- local = frame->local;
+ local = frame->local;
- if (op_ret < 0) {
- local->op_ret = op_ret;
- local->op_errno = op_errno;
- goto unwind;
- }
+ if (local->op_ret < 0) {
+ shard_common_failure_unwind(local->fop, frame, local->op_ret,
+ local->op_errno);
+ return 0;
+ }
- local->prebuf = *prebuf;
- if (shard_modify_size_and_block_count (&local->prebuf, xdata)) {
- local->op_ret = -1;
- local->op_errno = EINVAL;
- goto unwind;
- }
- if (xdata)
- local->xattr_rsp = dict_ref (xdata);
- local->postbuf = *postbuf;
- local->postbuf.ia_size = local->prebuf.ia_size;
- local->postbuf.ia_blocks = local->prebuf.ia_blocks;
+ if (local->fd)
+ STACK_WIND(frame, shard_common_set_xattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, local->fd,
+ local->xattr_req, local->flags, local->xattr_rsp);
+ else
+ STACK_WIND(frame, shard_common_set_xattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, &local->loc,
+ local->xattr_req, local->flags, local->xattr_rsp);
+ return 0;
+}
-unwind:
- local->handler (frame, this);
+int32_t
+shard_common_set_xattr(call_frame_t *frame, xlator_t *this, glusterfs_fop_t fop,
+ loc_t *loc, fd_t *fd, dict_t *dict, int32_t flags,
+ dict_t *xdata)
+{
+ int ret = -1;
+ int op_errno = ENOMEM;
+ uint64_t block_size = 0;
+ shard_local_t *local = NULL;
+ inode_t *inode = loc ? loc->inode : fd->inode;
+
+ if ((IA_ISDIR(inode->ia_type)) || (IA_ISLNK(inode->ia_type))) {
+ if (loc)
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, loc, dict, flags,
+ xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags,
+ xdata);
+ return 0;
+ }
+
+ /* Sharded or not, if shard's special xattrs are attempted to be set,
+ * fail the fop with EPERM (except if the client is gsyncd.
+ */
+ if (frame->root->pid != GF_CLIENT_PID_GSYNCD) {
+ GF_IF_INTERNAL_XATTR_GOTO(SHARD_XATTR_PREFIX "*", dict, op_errno, err);
+ }
+
+ ret = shard_inode_ctx_get_block_size(inode, this, &block_size);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_CTX_GET_FAILED,
+ "Failed to get block size from inode ctx of %s",
+ uuid_utoa(inode->gfid));
+ goto err;
+ }
+
+ if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
+ if (loc)
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, loc, dict, flags,
+ xdata);
+ else
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags,
+ xdata);
return 0;
+ }
+
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
+
+ frame->local = local;
+ local->fop = fop;
+ if (loc) {
+ if (loc_copy(&local->loc, loc) != 0)
+ goto err;
+ }
+
+ if (fd) {
+ local->fd = fd_ref(fd);
+ local->loc.inode = inode_ref(fd->inode);
+ gf_uuid_copy(local->loc.gfid, fd->inode->gfid);
+ }
+ local->flags = flags;
+ /* Reusing local->xattr_req and local->xattr_rsp to store the setxattr dict
+ * and the xdata dict
+ */
+ if (dict)
+ local->xattr_req = dict_ref(dict);
+ if (xdata)
+ local->xattr_rsp = dict_ref(xdata);
+
+ shard_refresh_base_file(frame, this, loc, fd,
+ shard_post_lookup_set_xattr_handler);
+ return 0;
+err:
+ shard_common_failure_unwind(fop, frame, -1, op_errno);
+ return 0;
}
-int
-shard_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+int32_t
+shard_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- int ret = -1;
- uint64_t block_size = 0;
- shard_local_t *local = NULL;
+ shard_common_set_xattr(frame, this, GF_FOP_FSETXATTR, NULL, fd, dict, flags,
+ xdata);
+ return 0;
+}
- if ((IA_ISDIR (loc->inode->ia_type)) ||
- (IA_ISLNK (loc->inode->ia_type))) {
- STACK_WIND (frame, default_setattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setattr, loc, stbuf,
- valid, xdata);
- return 0;
- }
+int32_t
+shard_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata)
+{
+ shard_common_set_xattr(frame, this, GF_FOP_SETXATTR, loc, NULL, dict, flags,
+ xdata);
+ return 0;
+}
- ret = shard_inode_ctx_get_block_size (loc->inode, this, &block_size);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_CTX_GET_FAILED,
- "Failed to get block size from inode ctx of %s",
- uuid_utoa (loc->inode->gfid));
- goto err;
- }
+int
+shard_post_setattr_handler(call_frame_t *frame, xlator_t *this)
+{
+ shard_local_t *local = NULL;
- if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
- STACK_WIND (frame, default_setattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setattr, loc, stbuf,
- valid, xdata);
- return 0;
- }
+ local = frame->local;
- local = mem_get0 (this->local_pool);
- if (!local)
- goto err;
+ if (local->fop == GF_FOP_SETATTR) {
+ if (local->op_ret >= 0)
+ shard_inode_ctx_set(local->loc.inode, this, &local->postbuf, 0,
+ SHARD_LOOKUP_MASK);
+ SHARD_STACK_UNWIND(setattr, frame, local->op_ret, local->op_errno,
+ &local->prebuf, &local->postbuf, local->xattr_rsp);
+ } else if (local->fop == GF_FOP_FSETATTR) {
+ if (local->op_ret >= 0)
+ shard_inode_ctx_set(local->fd->inode, this, &local->postbuf, 0,
+ SHARD_LOOKUP_MASK);
+ SHARD_STACK_UNWIND(fsetattr, frame, local->op_ret, local->op_errno,
+ &local->prebuf, &local->postbuf, local->xattr_rsp);
+ }
- frame->local = local;
+ return 0;
+}
- local->handler = shard_post_setattr_handler;
- local->xattr_req = (xdata) ? dict_ref (xdata) : dict_new ();
- if (!local->xattr_req)
- goto err;
- local->fop = GF_FOP_SETATTR;
- loc_copy (&local->loc, loc);
+int
+shard_common_setattr_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+ struct iatt *postbuf, dict_t *xdata)
+{
+ shard_local_t *local = NULL;
- SHARD_MD_READ_FOP_INIT_REQ_DICT (this, local->xattr_req,
- local->loc.gfid, local, err);
+ local = frame->local;
- STACK_WIND (frame, shard_common_setattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid,
- local->xattr_req);
+ if (op_ret < 0) {
+ local->op_ret = op_ret;
+ local->op_errno = op_errno;
+ goto unwind;
+ }
- return 0;
+ local->prebuf = *prebuf;
+ if (shard_modify_size_and_block_count(&local->prebuf, xdata)) {
+ local->op_ret = -1;
+ local->op_errno = EINVAL;
+ goto unwind;
+ }
+ if (xdata)
+ local->xattr_rsp = dict_ref(xdata);
+ local->postbuf = *postbuf;
+ local->postbuf.ia_size = local->prebuf.ia_size;
+ local->postbuf.ia_blocks = local->prebuf.ia_blocks;
-err:
- SHARD_STACK_UNWIND (setattr, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+unwind:
+ local->handler(frame, this);
+ return 0;
}
int
-shard_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+shard_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- int ret = -1;
- uint64_t block_size = 0;
- shard_local_t *local = NULL;
-
- if ((IA_ISDIR (fd->inode->ia_type)) ||
- (IA_ISLNK (fd->inode->ia_type))) {
- STACK_WIND (frame, default_fsetattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD (this)->fops->fsetattr, fd, stbuf,
- valid, xdata);
- return 0;
- }
+ int ret = -1;
+ uint64_t block_size = 0;
+ shard_local_t *local = NULL;
- ret = shard_inode_ctx_get_block_size (fd->inode, this, &block_size);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_CTX_GET_FAILED,
- "Failed to get block size from inode ctx of %s",
- uuid_utoa (fd->inode->gfid));
- goto err;
- }
-
- if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
- STACK_WIND (frame, default_fsetattr_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fsetattr, fd, stbuf,
- valid, xdata);
- return 0;
- }
+ if ((IA_ISDIR(loc->inode->ia_type)) || (IA_ISLNK(loc->inode->ia_type))) {
+ STACK_WIND(frame, default_setattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid, xdata);
+ return 0;
+ }
- if (!this->itable)
- this->itable = fd->inode->table;
+ ret = shard_inode_ctx_get_block_size(loc->inode, this, &block_size);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_CTX_GET_FAILED,
+ "Failed to get block size from inode ctx of %s",
+ uuid_utoa(loc->inode->gfid));
+ goto err;
+ }
- local = mem_get0 (this->local_pool);
- if (!local)
- goto err;
+ if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
+ STACK_WIND(frame, default_setattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid, xdata);
+ return 0;
+ }
- frame->local = local;
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
- local->handler = shard_post_setattr_handler;
- local->xattr_req = (xdata) ? dict_ref (xdata) : dict_new ();
- if (!local->xattr_req)
- goto err;
- local->fop = GF_FOP_FSETATTR;
- local->fd = fd_ref (fd);
+ frame->local = local;
- SHARD_MD_READ_FOP_INIT_REQ_DICT (this, local->xattr_req,
- fd->inode->gfid, local, err);
+ local->handler = shard_post_setattr_handler;
+ local->xattr_req = (xdata) ? dict_ref(xdata) : dict_new();
+ if (!local->xattr_req)
+ goto err;
+ local->fop = GF_FOP_SETATTR;
+ loc_copy(&local->loc, loc);
- STACK_WIND (frame, shard_common_setattr_cbk, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fsetattr, fd, stbuf, valid,
- local->xattr_req);
- return 0;
+ SHARD_MD_READ_FOP_INIT_REQ_DICT(this, local->xattr_req, local->loc.gfid,
+ local, err);
+ STACK_WIND(frame, shard_common_setattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid,
+ local->xattr_req);
+ return 0;
err:
- SHARD_STACK_UNWIND (fsetattr, frame, -1, ENOMEM, NULL, NULL, NULL);
- return 0;
+ shard_common_failure_unwind(GF_FOP_SETATTR, frame, -1, ENOMEM);
+ return 0;
}
int
-shard_common_inode_write_begin (call_frame_t *frame, xlator_t *this,
- glusterfs_fop_t fop, fd_t *fd,
- struct iovec *vector, int32_t count,
- off_t offset, uint32_t flags, size_t len,
- struct iobref *iobref, dict_t *xdata)
+shard_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- int ret = 0;
- int i = 0;
- uint64_t block_size = 0;
- shard_local_t *local = NULL;
+ int ret = -1;
+ uint64_t block_size = 0;
+ shard_local_t *local = NULL;
- ret = shard_inode_ctx_get_block_size (fd->inode, this, &block_size);
- if (ret) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_INODE_CTX_GET_FAILED, "Failed to get block "
- "size for %s from its inode ctx",
- uuid_utoa (fd->inode->gfid));
- goto out;
- }
+ if ((IA_ISDIR(fd->inode->ia_type)) || (IA_ISLNK(fd->inode->ia_type))) {
+ STACK_WIND(frame, default_fsetattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetattr, fd, stbuf, valid, xdata);
+ return 0;
+ }
- if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
- /* block_size = 0 means that the file was created before
- * sharding was enabled on the volume.
- */
- switch (fop) {
- case GF_FOP_WRITE:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->writev, fd,
- vector, count, offset, flags, iobref,
- xdata);
- break;
- case GF_FOP_FALLOCATE:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->fallocate, fd,
- flags, offset, len, xdata);
- break;
- case GF_FOP_ZEROFILL:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->zerofill,
- fd, offset, len, xdata);
- break;
- case GF_FOP_DISCARD:
- STACK_WIND_TAIL (frame, FIRST_CHILD(this),
- FIRST_CHILD(this)->fops->discard,
- fd, offset, len, xdata);
- break;
- default:
- gf_msg (this->name, GF_LOG_WARNING, 0, SHARD_MSG_INVALID_FOP,
- "Invalid fop id = %d", fop);
- break;
- }
- return 0;
- }
+ ret = shard_inode_ctx_get_block_size(fd->inode, this, &block_size);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_CTX_GET_FAILED,
+ "Failed to get block size from inode ctx of %s",
+ uuid_utoa(fd->inode->gfid));
+ goto err;
+ }
- if (!this->itable)
- this->itable = fd->inode->table;
+ if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
+ STACK_WIND(frame, default_fsetattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetattr, fd, stbuf, valid, xdata);
+ return 0;
+ }
- local = mem_get0 (this->local_pool);
- if (!local)
- goto out;
+ if (!this->itable)
+ this->itable = fd->inode->table;
- frame->local = local;
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto err;
- ret = syncbarrier_init (&local->barrier);
- if (ret)
- goto out;
- local->xattr_req = (xdata) ? dict_ref (xdata) : dict_new ();
- if (!local->xattr_req)
- goto out;
-
- if (vector) {
- local->vector = iov_dup (vector, count);
- if (!local->vector)
- goto out;
- for (i = 0; i < count; i++)
- local->total_size += vector[i].iov_len;
- local->count = count;
- } else {
- local->total_size = len;
- }
+ frame->local = local;
- local->fop = fop;
- local->offset = offset;
- local->flags = flags;
- if (iobref)
- local->iobref = iobref_ref (iobref);
- local->fd = fd_ref (fd);
- local->block_size = block_size;
- local->resolver_base_inode = local->fd->inode;
+ local->handler = shard_post_setattr_handler;
+ local->xattr_req = (xdata) ? dict_ref(xdata) : dict_new();
+ if (!local->xattr_req)
+ goto err;
+ local->fop = GF_FOP_FSETATTR;
+ local->fd = fd_ref(fd);
- local->loc.inode = inode_ref (fd->inode);
- gf_uuid_copy (local->loc.gfid, fd->inode->gfid);
+ SHARD_MD_READ_FOP_INIT_REQ_DICT(this, local->xattr_req, fd->inode->gfid,
+ local, err);
- shard_lookup_base_file (frame, this, &local->loc,
- shard_common_inode_write_post_lookup_handler);
+ STACK_WIND(frame, shard_common_setattr_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetattr, fd, stbuf, valid,
+ local->xattr_req);
+ return 0;
+err:
+ shard_common_failure_unwind(GF_FOP_FSETATTR, frame, -1, ENOMEM);
+ return 0;
+}
+
+int
+shard_common_inode_write_begin(call_frame_t *frame, xlator_t *this,
+ glusterfs_fop_t fop, fd_t *fd,
+ struct iovec *vector, int32_t count,
+ off_t offset, uint32_t flags, size_t len,
+ struct iobref *iobref, dict_t *xdata)
+{
+ int ret = 0;
+ int i = 0;
+ uint64_t block_size = 0;
+ shard_local_t *local = NULL;
+
+ ret = shard_inode_ctx_get_block_size(fd->inode, this, &block_size);
+ if (ret) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INODE_CTX_GET_FAILED,
+ "Failed to get block "
+ "size for %s from its inode ctx",
+ uuid_utoa(fd->inode->gfid));
+ goto out;
+ }
+
+ if (!block_size || frame->root->pid == GF_CLIENT_PID_GSYNCD) {
+ /* block_size = 0 means that the file was created before
+ * sharding was enabled on the volume.
+ */
+ switch (fop) {
+ case GF_FOP_WRITE:
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->writev, fd, vector,
+ count, offset, flags, iobref, xdata);
+ break;
+ case GF_FOP_FALLOCATE:
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fallocate, fd, flags,
+ offset, len, xdata);
+ break;
+ case GF_FOP_ZEROFILL:
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->zerofill, fd, offset,
+ len, xdata);
+ break;
+ case GF_FOP_DISCARD:
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->discard, fd, offset,
+ len, xdata);
+ break;
+ default:
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_INVALID_FOP,
+ "Invalid fop id = %d", fop);
+ break;
+ }
return 0;
+ }
+
+ if (!this->itable)
+ this->itable = fd->inode->table;
+
+ local = mem_get0(this->local_pool);
+ if (!local)
+ goto out;
+
+ frame->local = local;
+
+ ret = syncbarrier_init(&local->barrier);
+ if (ret)
+ goto out;
+ local->xattr_req = (xdata) ? dict_ref(xdata) : dict_new();
+ if (!local->xattr_req)
+ goto out;
+
+ if (vector) {
+ local->vector = iov_dup(vector, count);
+ if (!local->vector)
+ goto out;
+ for (i = 0; i < count; i++)
+ local->total_size += vector[i].iov_len;
+ local->count = count;
+ } else {
+ local->total_size = len;
+ }
+
+ local->fop = fop;
+ local->offset = offset;
+ local->flags = flags;
+ if (iobref)
+ local->iobref = iobref_ref(iobref);
+ local->fd = fd_ref(fd);
+ local->block_size = block_size;
+ local->resolver_base_inode = local->fd->inode;
+ GF_ATOMIC_INIT(local->delta_blocks, 0);
+
+ local->loc.inode = inode_ref(fd->inode);
+ gf_uuid_copy(local->loc.gfid, fd->inode->gfid);
+
+ shard_refresh_base_file(frame, this, NULL, fd,
+ shard_common_inode_write_post_lookup_handler);
+ return 0;
out:
- shard_common_inode_write_failure_unwind (fop, frame, -1, ENOMEM);
- return 0;
+ shard_common_failure_unwind(fop, frame, -1, ENOMEM);
+ return 0;
}
int
-shard_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
- struct iovec *vector, int32_t count, off_t offset, uint32_t flags,
- struct iobref *iobref, dict_t *xdata)
+shard_writev(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ struct iovec *vector, int32_t count, off_t offset, uint32_t flags,
+ struct iobref *iobref, dict_t *xdata)
{
- shard_common_inode_write_begin (frame, this, GF_FOP_WRITE, fd, vector,
- count, offset, flags, 0, iobref, xdata);
- return 0;
+ shard_common_inode_write_begin(frame, this, GF_FOP_WRITE, fd, vector, count,
+ offset, flags, 0, iobref, xdata);
+ return 0;
}
int
-shard_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd,
- int32_t keep_size, off_t offset, size_t len, dict_t *xdata)
+shard_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd,
+ int32_t keep_size, off_t offset, size_t len, dict_t *xdata)
{
- if ((keep_size != 0) && (keep_size != FALLOC_FL_ZERO_RANGE) &&
- (keep_size != (FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE)))
- goto out;
-
- shard_common_inode_write_begin (frame, this, GF_FOP_FALLOCATE, fd, NULL,
- 0, offset, keep_size, len, NULL, xdata);
- return 0;
+ if ((keep_size != 0) && (keep_size != FALLOC_FL_ZERO_RANGE) &&
+ (keep_size != (FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE)))
+ goto out;
+ shard_common_inode_write_begin(frame, this, GF_FOP_FALLOCATE, fd, NULL, 0,
+ offset, keep_size, len, NULL, xdata);
+ return 0;
out:
- SHARD_STACK_UNWIND (fallocate, frame, -1, ENOTSUP, NULL, NULL, NULL);
- return 0;
+ shard_common_failure_unwind(GF_FOP_FALLOCATE, frame, -1, ENOTSUP);
+ return 0;
}
int
-shard_zerofill (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- off_t len, dict_t *xdata)
+shard_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ off_t len, dict_t *xdata)
{
- shard_common_inode_write_begin (frame, this, GF_FOP_ZEROFILL, fd, NULL,
- 0, offset, 0, len, NULL, xdata);
- return 0;
+ shard_common_inode_write_begin(frame, this, GF_FOP_ZEROFILL, fd, NULL, 0,
+ offset, 0, len, NULL, xdata);
+ return 0;
}
int
-shard_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- size_t len, dict_t *xdata)
+shard_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ size_t len, dict_t *xdata)
{
- shard_common_inode_write_begin (frame, this, GF_FOP_DISCARD, fd, NULL,
- 0, offset, 0, len, NULL, xdata);
- return 0;
+ shard_common_inode_write_begin(frame, this, GF_FOP_DISCARD, fd, NULL, 0,
+ offset, 0, len, NULL, xdata);
+ return 0;
}
int32_t
-shard_seek (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
- gf_seek_what_t what, dict_t *xdata)
+shard_seek(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+ gf_seek_what_t what, dict_t *xdata)
{
- /* TBD */
- gf_msg (this->name, GF_LOG_INFO, ENOTSUP, SHARD_MSG_FOP_NOT_SUPPORTED,
- "seek called on %s.", uuid_utoa (fd->inode->gfid));
- SHARD_STACK_UNWIND (seek, frame, -1, ENOTSUP, 0, NULL);
- return 0;
+ /* TBD */
+ gf_msg(this->name, GF_LOG_INFO, ENOTSUP, SHARD_MSG_FOP_NOT_SUPPORTED,
+ "seek called on %s.", uuid_utoa(fd->inode->gfid));
+ shard_common_failure_unwind(GF_FOP_SEEK, frame, -1, ENOTSUP);
+ return 0;
}
int32_t
-mem_acct_init (xlator_t *this)
+mem_acct_init(xlator_t *this)
{
- int ret = -1;
+ int ret = -1;
- if (!this)
- return ret;
-
- ret = xlator_mem_acct_init (this, gf_shard_mt_end + 1);
+ if (!this)
+ return ret;
- if (ret != 0) {
- gf_msg (this->name, GF_LOG_ERROR, 0,
- SHARD_MSG_MEM_ACCT_INIT_FAILED, "Memory accounting init"
- "failed");
- return ret;
- }
+ ret = xlator_mem_acct_init(this, gf_shard_mt_end + 1);
+ if (ret != 0) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_MEM_ACCT_INIT_FAILED,
+ "Memory accounting init"
+ "failed");
return ret;
+ }
+
+ return ret;
}
int
-init (xlator_t *this)
+init(xlator_t *this)
{
- int ret = -1;
- shard_priv_t *priv = NULL;
+ int ret = -1;
+ shard_priv_t *priv = NULL;
- if (!this) {
- gf_msg ("shard", GF_LOG_ERROR, 0, SHARD_MSG_NULL_THIS,
- "this is NULL. init() failed");
- return -1;
- }
+ if (!this) {
+ gf_msg("shard", GF_LOG_ERROR, 0, SHARD_MSG_NULL_THIS,
+ "this is NULL. init() failed");
+ return -1;
+ }
- if (!this->parents) {
- gf_msg (this->name, GF_LOG_ERROR, 0, SHARD_MSG_INVALID_VOLFILE,
- "Dangling volume. Check volfile");
- goto out;
- }
+ if (!this->parents) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INVALID_VOLFILE,
+ "Dangling volume. Check volfile");
+ goto out;
+ }
- if (!this->children || this->children->next) {
- gf_msg (this->name, GF_LOG_ERROR, 0, SHARD_MSG_INVALID_VOLFILE,
- "shard not configured with exactly one sub-volume. "
- "Check volfile");
- goto out;
- }
+ if (!this->children || this->children->next) {
+ gf_msg(this->name, GF_LOG_ERROR, 0, SHARD_MSG_INVALID_VOLFILE,
+ "shard not configured with exactly one sub-volume. "
+ "Check volfile");
+ goto out;
+ }
- priv = GF_CALLOC (1, sizeof (shard_priv_t), gf_shard_mt_priv_t);
- if (!priv)
- goto out;
+ priv = GF_CALLOC(1, sizeof(shard_priv_t), gf_shard_mt_priv_t);
+ if (!priv)
+ goto out;
- GF_OPTION_INIT ("shard-block-size", priv->block_size, size_uint64, out);
+ GF_OPTION_INIT("shard-block-size", priv->block_size, size_uint64, out);
- this->local_pool = mem_pool_new (shard_local_t, 128);
- if (!this->local_pool) {
- ret = -1;
- goto out;
- }
- gf_uuid_parse (SHARD_ROOT_GFID, priv->dot_shard_gfid);
+ GF_OPTION_INIT("shard-deletion-rate", priv->deletion_rate, uint32, out);
- this->private = priv;
- LOCK_INIT (&priv->lock);
- INIT_LIST_HEAD (&priv->ilist_head);
- ret = 0;
-out:
- if (ret) {
- GF_FREE (priv);
- mem_pool_destroy (this->local_pool);
- }
+ GF_OPTION_INIT("shard-lru-limit", priv->lru_limit, uint64, out);
- return ret;
+ this->local_pool = mem_pool_new(shard_local_t, 128);
+ if (!this->local_pool) {
+ ret = -1;
+ goto out;
+ }
+ gf_uuid_parse(SHARD_ROOT_GFID, priv->dot_shard_gfid);
+ gf_uuid_parse(DOT_SHARD_REMOVE_ME_GFID, priv->dot_shard_rm_gfid);
+
+ this->private = priv;
+ LOCK_INIT(&priv->lock);
+ INIT_LIST_HEAD(&priv->ilist_head);
+ ret = 0;
+out:
+ if (ret) {
+ GF_FREE(priv);
+ mem_pool_destroy(this->local_pool);
+ }
+ return ret;
}
void
-fini (xlator_t *this)
+fini(xlator_t *this)
{
- shard_priv_t *priv = NULL;
+ shard_priv_t *priv = NULL;
+
+ GF_VALIDATE_OR_GOTO("shard", this, out);
- GF_VALIDATE_OR_GOTO ("shard", this, out);
+ /*Itable was not created by shard, hence setting to NULL.*/
+ this->itable = NULL;
- mem_pool_destroy (this->local_pool);
- this->local_pool = NULL;
+ mem_pool_destroy(this->local_pool);
+ this->local_pool = NULL;
- priv = this->private;
- if (!priv)
- goto out;
+ priv = this->private;
+ if (!priv)
+ goto out;
- this->private = NULL;
- LOCK_DESTROY (&priv->lock);
- GF_FREE (priv);
+ this->private = NULL;
+ LOCK_DESTROY(&priv->lock);
+ GF_FREE(priv);
out:
- return;
+ return;
}
int
-reconfigure (xlator_t *this, dict_t *options)
+reconfigure(xlator_t *this, dict_t *options)
{
- int ret = -1;
- shard_priv_t *priv = NULL;
+ int ret = -1;
+ shard_priv_t *priv = NULL;
- priv = this->private;
+ priv = this->private;
- GF_OPTION_RECONF ("shard-block-size", priv->block_size, options, size,
- out);
+ GF_OPTION_RECONF("shard-block-size", priv->block_size, options, size, out);
- ret = 0;
+ GF_OPTION_RECONF("shard-deletion-rate", priv->deletion_rate, options,
+ uint32, out);
+ ret = 0;
out:
- return ret;
+ return ret;
}
int
-shard_forget (xlator_t *this, inode_t *inode)
+shard_forget(xlator_t *this, inode_t *inode)
{
- uint64_t ctx_uint = 0;
- shard_inode_ctx_t *ctx = NULL;
- shard_priv_t *priv = NULL;
+ uint64_t ctx_uint = 0;
+ shard_inode_ctx_t *ctx = NULL;
+ shard_priv_t *priv = NULL;
- priv = this->private;
- inode_ctx_del (inode, this, &ctx_uint);
- if (!ctx_uint)
- return 0;
+ priv = this->private;
+ if (!priv)
+ return 0;
- ctx = (shard_inode_ctx_t *)ctx_uint;
+ inode_ctx_del(inode, this, &ctx_uint);
+ if (!ctx_uint)
+ return 0;
- /* When LRU limit reaches inode will be forcefully removed from the
- * table, inode needs to be removed from LRU of shard as well.
- */
- if (!list_empty (&ctx->ilist)) {
- LOCK(&priv->lock);
- {
- list_del_init (&ctx->ilist);
- priv->inode_count--;
- }
- UNLOCK(&priv->lock);
+ ctx = (shard_inode_ctx_t *)(uintptr_t)ctx_uint;
+
+ /* When LRU limit reaches inode will be forcefully removed from the
+ * table, inode needs to be removed from LRU of shard as well.
+ */
+ if (!list_empty(&ctx->ilist)) {
+ LOCK(&priv->lock);
+ {
+ list_del_init(&ctx->ilist);
+ priv->inode_count--;
}
- GF_FREE (ctx);
+ UNLOCK(&priv->lock);
+ }
+ GF_FREE(ctx);
- return 0;
+ return 0;
}
int
-shard_release (xlator_t *this, fd_t *fd)
+shard_release(xlator_t *this, fd_t *fd)
{
- /* TBD */
- return 0;
+ /* TBD */
+ return 0;
}
int
-shard_priv_dump (xlator_t *this)
+shard_priv_dump(xlator_t *this)
{
- shard_priv_t *priv = NULL;
- char key_prefix[GF_DUMP_MAX_BUF_LEN] = {0,};
- char *str = NULL;
+ shard_priv_t *priv = NULL;
+ char key_prefix[GF_DUMP_MAX_BUF_LEN] = {
+ 0,
+ };
+ char *str = NULL;
- priv = this->private;
+ priv = this->private;
- snprintf (key_prefix, GF_DUMP_MAX_BUF_LEN, "%s.%s", this->type,
- this->name);
- gf_proc_dump_add_section (key_prefix);
- str = gf_uint64_2human_readable (priv->block_size);
- gf_proc_dump_write ("shard-block-size", "%s", str);
- gf_proc_dump_write ("inode-count", "%d", priv->inode_count);
- gf_proc_dump_write ("ilist_head", "%p", &priv->ilist_head);
- gf_proc_dump_write ("lru-max-limit", "%d", SHARD_MAX_INODES);
+ snprintf(key_prefix, GF_DUMP_MAX_BUF_LEN, "%s.%s", this->type, this->name);
+ gf_proc_dump_add_section("%s", key_prefix);
+ str = gf_uint64_2human_readable(priv->block_size);
+ gf_proc_dump_write("shard-block-size", "%s", str);
+ gf_proc_dump_write("inode-count", "%d", priv->inode_count);
+ gf_proc_dump_write("ilist_head", "%p", &priv->ilist_head);
+ gf_proc_dump_write("lru-max-limit", "%" PRIu64, priv->lru_limit);
- GF_FREE (str);
+ GF_FREE(str);
- return 0;
+ return 0;
}
int
-shard_releasedir (xlator_t *this, fd_t *fd)
+shard_releasedir(xlator_t *this, fd_t *fd)
{
- return 0;
+ return 0;
}
struct xlator_fops fops = {
- .lookup = shard_lookup,
- .open = shard_open,
- .flush = shard_flush,
- .fsync = shard_fsync,
- .stat = shard_stat,
- .fstat = shard_fstat,
- .getxattr = shard_getxattr,
- .fgetxattr = shard_fgetxattr,
- .readv = shard_readv,
- .writev = shard_writev,
- .truncate = shard_truncate,
- .ftruncate = shard_ftruncate,
- .setxattr = shard_setxattr,
- .fsetxattr = shard_fsetxattr,
- .setattr = shard_setattr,
- .fsetattr = shard_fsetattr,
- .removexattr = shard_removexattr,
- .fremovexattr = shard_fremovexattr,
- .fallocate = shard_fallocate,
- .discard = shard_discard,
- .zerofill = shard_zerofill,
- .readdir = shard_readdir,
- .readdirp = shard_readdirp,
- .create = shard_create,
- .mknod = shard_mknod,
- .link = shard_link,
- .unlink = shard_unlink,
- .rename = shard_rename,
- .seek = shard_seek,
+ .lookup = shard_lookup,
+ .open = shard_open,
+ .flush = shard_flush,
+ .fsync = shard_fsync,
+ .stat = shard_stat,
+ .fstat = shard_fstat,
+ .getxattr = shard_getxattr,
+ .fgetxattr = shard_fgetxattr,
+ .readv = shard_readv,
+ .writev = shard_writev,
+ .truncate = shard_truncate,
+ .ftruncate = shard_ftruncate,
+ .setxattr = shard_setxattr,
+ .fsetxattr = shard_fsetxattr,
+ .setattr = shard_setattr,
+ .fsetattr = shard_fsetattr,
+ .removexattr = shard_removexattr,
+ .fremovexattr = shard_fremovexattr,
+ .fallocate = shard_fallocate,
+ .discard = shard_discard,
+ .zerofill = shard_zerofill,
+ .readdir = shard_readdir,
+ .readdirp = shard_readdirp,
+ .create = shard_create,
+ .mknod = shard_mknod,
+ .link = shard_link,
+ .unlink = shard_unlink,
+ .rename = shard_rename,
+ .seek = shard_seek,
};
struct xlator_cbks cbks = {
- .forget = shard_forget,
- .release = shard_release,
- .releasedir = shard_releasedir,
+ .forget = shard_forget,
+ .release = shard_release,
+ .releasedir = shard_releasedir,
};
struct xlator_dumpops dumpops = {
- .priv = shard_priv_dump,
+ .priv = shard_priv_dump,
};
struct volume_options options[] = {
- { .key = {"shard-block-size"},
- .type = GF_OPTION_TYPE_SIZET,
- .op_version = {GD_OP_VERSION_3_7_0},
- .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
- .tags = {"shard"},
- .default_value = "64MB",
- .min = SHARD_MIN_BLOCK_SIZE,
- .max = SHARD_MAX_BLOCK_SIZE,
- .description = "The size unit used to break a file into multiple "
- "chunks",
- },
- { .key = {NULL} },
+ {
+ .key = {"shard"},
+ .type = GF_OPTION_TYPE_BOOL,
+ .default_value = "off",
+ .description = "enable/disable shard",
+ .op_version = {GD_OP_VERSION_6_0},
+ .flags = OPT_FLAG_SETTABLE,
+ },
+ {
+ .key = {"shard-block-size"},
+ .type = GF_OPTION_TYPE_SIZET,
+ .op_version = {GD_OP_VERSION_3_7_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
+ .tags = {"shard"},
+ .default_value = "64MB",
+ .min = SHARD_MIN_BLOCK_SIZE,
+ .max = SHARD_MAX_BLOCK_SIZE,
+ .description = "The size unit used to break a file into multiple "
+ "chunks",
+ },
+ {
+ .key = {"shard-deletion-rate"},
+ .type = GF_OPTION_TYPE_INT,
+ .op_version = {GD_OP_VERSION_5_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT | OPT_FLAG_DOC,
+ .tags = {"shard"},
+ .default_value = "100",
+ .min = 100,
+ .max = INT_MAX,
+ .description = "The number of shards to send deletes on at a time",
+ },
+ {
+ .key = {"shard-lru-limit"},
+ .type = GF_OPTION_TYPE_INT,
+ .op_version = {GD_OP_VERSION_5_0},
+ .flags = OPT_FLAG_SETTABLE | OPT_FLAG_CLIENT_OPT,
+ .tags = {"shard"},
+ .default_value = "16384",
+ .min = 20,
+ .max = INT_MAX,
+ .description = "The number of resolved shard inodes to keep in "
+ "memory. A higher number means shards that are "
+ "resolved will remain in memory longer, avoiding "
+ "frequent lookups on them when they participate in "
+ "file operations. The option also has a bearing on "
+ "amount of memory consumed by these inodes and their "
+ "internal metadata",
+ },
+ {.key = {NULL}},
+};
+
+xlator_api_t xlator_api = {
+ .init = init,
+ .fini = fini,
+ .reconfigure = reconfigure,
+ .mem_acct_init = mem_acct_init,
+ .op_version = {1}, /* Present from the initial version */
+ .dumpops = &dumpops,
+ .fops = &fops,
+ .cbks = &cbks,
+ .options = options,
+ .identifier = "shard",
+ .category = GF_MAINTAINED,
};
diff --git a/xlators/features/shard/src/shard.h b/xlators/features/shard/src/shard.h
index 225caa0c0c2..4fe181b64d5 100644
--- a/xlators/features/shard/src/shard.h
+++ b/xlators/features/shard/src/shard.h
@@ -8,282 +8,341 @@
cases as published by the Free Software Foundation.
*/
-
#ifndef __SHARD_H__
#define __SHARD_H__
-#include "xlator.h"
-#include "compat-errno.h"
+#include <glusterfs/xlator.h>
+#include <glusterfs/compat-errno.h>
#include "shard-messages.h"
-#include "syncop.h"
+#include <glusterfs/syncop.h>
#define GF_SHARD_DIR ".shard"
-#define SHARD_MIN_BLOCK_SIZE (4 * GF_UNIT_MB)
-#define SHARD_MAX_BLOCK_SIZE (4 * GF_UNIT_TB)
+#define GF_SHARD_REMOVE_ME_DIR ".remove_me"
+#define SHARD_MIN_BLOCK_SIZE (4 * GF_UNIT_MB)
+#define SHARD_MAX_BLOCK_SIZE (4 * GF_UNIT_TB)
#define SHARD_XATTR_PREFIX "trusted.glusterfs.shard."
#define GF_XATTR_SHARD_BLOCK_SIZE "trusted.glusterfs.shard.block-size"
-#define SHARD_INODE_LRU_LIMIT 4096
-#define SHARD_MAX_INODES 16384
/**
* Bit masks for the valid flag, which is used while updating ctx
-**/
-#define SHARD_MASK_BLOCK_SIZE (1 << 0)
-#define SHARD_MASK_PROT (1 << 1)
-#define SHARD_MASK_NLINK (1 << 2)
-#define SHARD_MASK_UID (1 << 3)
-#define SHARD_MASK_GID (1 << 4)
-#define SHARD_MASK_SIZE (1 << 6)
-#define SHARD_MASK_BLOCKS (1 << 7)
-#define SHARD_MASK_TIMES (1 << 8)
-#define SHARD_MASK_OTHERS (1 << 9)
-#define SHARD_MASK_REFRESH_RESET (1 << 10)
-
-#define SHARD_INODE_WRITE_MASK (SHARD_MASK_SIZE | SHARD_MASK_BLOCKS \
- | SHARD_MASK_TIMES)
+ **/
+#define SHARD_MASK_BLOCK_SIZE (1 << 0)
+#define SHARD_MASK_PROT (1 << 1)
+#define SHARD_MASK_NLINK (1 << 2)
+#define SHARD_MASK_UID (1 << 3)
+#define SHARD_MASK_GID (1 << 4)
+#define SHARD_MASK_SIZE (1 << 6)
+#define SHARD_MASK_BLOCKS (1 << 7)
+#define SHARD_MASK_TIMES (1 << 8)
+#define SHARD_MASK_OTHERS (1 << 9)
+#define SHARD_MASK_REFRESH_RESET (1 << 10)
-#define SHARD_LOOKUP_MASK (SHARD_MASK_PROT | SHARD_MASK_NLINK | SHARD_MASK_UID \
- | SHARD_MASK_GID | SHARD_MASK_TIMES \
- | SHARD_MASK_OTHERS)
+#define SHARD_INODE_WRITE_MASK \
+ (SHARD_MASK_SIZE | SHARD_MASK_BLOCKS | SHARD_MASK_TIMES)
-#define SHARD_ALL_MASK (SHARD_MASK_BLOCK_SIZE | SHARD_MASK_PROT \
- | SHARD_MASK_NLINK | SHARD_MASK_UID | SHARD_MASK_GID \
- | SHARD_MASK_SIZE | SHARD_MASK_BLOCKS \
- | SHARD_MASK_TIMES | SHARD_MASK_OTHERS)
+#define SHARD_LOOKUP_MASK \
+ (SHARD_MASK_PROT | SHARD_MASK_NLINK | SHARD_MASK_UID | SHARD_MASK_GID | \
+ SHARD_MASK_TIMES | SHARD_MASK_OTHERS)
+#define SHARD_ALL_MASK \
+ (SHARD_MASK_BLOCK_SIZE | SHARD_MASK_PROT | SHARD_MASK_NLINK | \
+ SHARD_MASK_UID | SHARD_MASK_GID | SHARD_MASK_SIZE | SHARD_MASK_BLOCKS | \
+ SHARD_MASK_TIMES | SHARD_MASK_OTHERS)
#define get_lowest_block(off, shard_size) ((off) / (shard_size))
-#define get_highest_block(off, len, shard_size) \
- (((((off)+(len)) == 0)?0:((off)+(len)-1)) / (shard_size))
+#define get_highest_block(off, len, shard_size) \
+ (((((off) + (len)) == 0) ? 0 : ((off) + (len)-1)) / (shard_size))
+
+int
+shard_unlock_inodelk(call_frame_t *frame, xlator_t *this);
+
+int
+shard_unlock_entrylk(call_frame_t *frame, xlator_t *this);
-#define SHARD_ENTRY_FOP_CHECK(loc, op_errno, label) do { \
- if ((loc->name && !strcmp (GF_SHARD_DIR, loc->name)) && \
- (((loc->parent) && \
- __is_root_gfid (loc->parent->gfid)) || \
- __is_root_gfid (loc->pargfid))) { \
- op_errno = EPERM; \
- goto label; \
- } \
- \
- if ((loc->parent && \
- __is_shard_dir (loc->parent->gfid)) || \
- __is_shard_dir (loc->pargfid)) { \
- op_errno = EPERM; \
- goto label; \
- } \
-} while (0)
+#define SHARD_ENTRY_FOP_CHECK(loc, op_errno, label) \
+ do { \
+ if ((loc->name && !strcmp(GF_SHARD_DIR, loc->name)) && \
+ (((loc->parent) && __is_root_gfid(loc->parent->gfid)) || \
+ __is_root_gfid(loc->pargfid))) { \
+ op_errno = EPERM; \
+ goto label; \
+ } \
+ \
+ if ((loc->parent && __is_shard_dir(loc->parent->gfid)) || \
+ __is_shard_dir(loc->pargfid)) { \
+ op_errno = EPERM; \
+ goto label; \
+ } \
+ } while (0)
-#define SHARD_INODE_OP_CHECK(gfid, err, label) do { \
- if (__is_shard_dir(gfid)) { \
- err = EPERM; \
- goto label; \
- } \
-} while (0)
+#define SHARD_INODE_OP_CHECK(gfid, err, label) \
+ do { \
+ if (__is_shard_dir(gfid)) { \
+ err = EPERM; \
+ goto label; \
+ } \
+ } while (0)
-#define SHARD_STACK_UNWIND(fop, frame, params ...) do { \
- shard_local_t *__local = NULL; \
- if (frame) { \
- __local = frame->local; \
- frame->local = NULL; \
- } \
- STACK_UNWIND_STRICT (fop, frame, params); \
- if (__local) { \
- shard_local_wipe (__local); \
- mem_put (__local); \
- } \
-} while (0)
+#define SHARD_STACK_UNWIND(fop, frame, params...) \
+ do { \
+ shard_local_t *__local = NULL; \
+ if (frame) { \
+ __local = frame->local; \
+ if (__local && __local->int_inodelk.acquired_lock) \
+ shard_unlock_inodelk(frame, frame->this); \
+ if (__local && __local->int_entrylk.acquired_lock) \
+ shard_unlock_entrylk(frame, frame->this); \
+ frame->local = NULL; \
+ } \
+ STACK_UNWIND_STRICT(fop, frame, params); \
+ if (__local) { \
+ shard_local_wipe(__local); \
+ mem_put(__local); \
+ } \
+ } while (0)
+#define SHARD_STACK_DESTROY(frame) \
+ do { \
+ shard_local_t *__local = NULL; \
+ __local = frame->local; \
+ frame->local = NULL; \
+ STACK_DESTROY(frame->root); \
+ if (__local) { \
+ shard_local_wipe(__local); \
+ mem_put(__local); \
+ } \
+ } while (0);
-#define SHARD_INODE_CREATE_INIT(this, local, xattr_req, loc, label) do { \
- int __ret = -1; \
- int64_t *__size_attr = NULL; \
- shard_priv_t *__priv = NULL; \
- \
- __priv = this->private; \
- \
- local->block_size = hton64 (__priv->block_size); \
- __ret = dict_set_static_bin (xattr_req, GF_XATTR_SHARD_BLOCK_SIZE, \
- &local->block_size, \
- sizeof (local->block_size)); \
- if (__ret) { \
- gf_msg (this->name, GF_LOG_WARNING, 0, \
- SHARD_MSG_DICT_SET_FAILED, "Failed to set key: %s " \
- "on path %s", GF_XATTR_SHARD_BLOCK_SIZE, loc->path); \
- goto label; \
- } \
- \
- __ret = shard_set_size_attrs (0, 0, &__size_attr); \
- if (__ret) \
- goto label; \
- \
- __ret = dict_set_bin (xattr_req, GF_XATTR_SHARD_FILE_SIZE, \
- __size_attr, 8 * 4); \
- if (__ret) { \
- gf_msg (this->name, GF_LOG_WARNING, 0, \
- SHARD_MSG_DICT_SET_FAILED, "Failed to set key: %s " \
- "on path %s", GF_XATTR_SHARD_FILE_SIZE, loc->path); \
- GF_FREE (__size_attr); \
- goto label; \
- } \
-} while (0)
+#define SHARD_INODE_CREATE_INIT(this, block_size, xattr_req, loc, size, \
+ block_count, label) \
+ do { \
+ int __ret = -1; \
+ int64_t *__size_attr = NULL; \
+ uint64_t *__bs = 0; \
+ \
+ __bs = GF_MALLOC(sizeof(uint64_t), gf_shard_mt_uint64_t); \
+ if (!__bs) \
+ goto label; \
+ *__bs = hton64(block_size); \
+ __ret = dict_set_bin(xattr_req, GF_XATTR_SHARD_BLOCK_SIZE, __bs, \
+ sizeof(*__bs)); \
+ if (__ret) { \
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_DICT_OP_FAILED, \
+ "Failed to set key: %s " \
+ "on path %s", \
+ GF_XATTR_SHARD_BLOCK_SIZE, (loc)->path); \
+ GF_FREE(__bs); \
+ goto label; \
+ } \
+ \
+ __ret = shard_set_size_attrs(size, block_count, &__size_attr); \
+ if (__ret) \
+ goto label; \
+ \
+ __ret = dict_set_bin(xattr_req, GF_XATTR_SHARD_FILE_SIZE, __size_attr, \
+ 8 * 4); \
+ if (__ret) { \
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_DICT_OP_FAILED, \
+ "Failed to set key: %s " \
+ "on path %s", \
+ GF_XATTR_SHARD_FILE_SIZE, (loc)->path); \
+ GF_FREE(__size_attr); \
+ goto label; \
+ } \
+ } while (0)
+#define SHARD_MD_READ_FOP_INIT_REQ_DICT(this, dict, gfid, local, label) \
+ do { \
+ int __ret = -1; \
+ \
+ __ret = dict_set_uint64(dict, GF_XATTR_SHARD_FILE_SIZE, 8 * 4); \
+ if (__ret) { \
+ local->op_ret = -1; \
+ local->op_errno = ENOMEM; \
+ gf_msg(this->name, GF_LOG_WARNING, 0, SHARD_MSG_DICT_OP_FAILED, \
+ "Failed to set dict value:" \
+ " key:%s for %s.", \
+ GF_XATTR_SHARD_FILE_SIZE, uuid_utoa(gfid)); \
+ goto label; \
+ } \
+ } while (0)
-#define SHARD_MD_READ_FOP_INIT_REQ_DICT(this, dict, gfid, local, label) do { \
- int __ret = -1; \
- \
- __ret = dict_set_uint64 (dict, GF_XATTR_SHARD_FILE_SIZE, 8 * 4); \
- if (__ret) { \
- local->op_ret = -1; \
- local->op_errno = ENOMEM; \
- gf_msg (this->name, GF_LOG_WARNING, 0, \
- SHARD_MSG_DICT_SET_FAILED, "Failed to set dict value:"\
- " key:%s for %s.", GF_XATTR_SHARD_FILE_SIZE, \
- uuid_utoa (gfid)); \
- goto label; \
- } \
-} while (0)
+#define SHARD_SET_ROOT_FS_ID(frame, local) \
+ do { \
+ if (!local->is_set_fsid) { \
+ local->uid = frame->root->uid; \
+ local->gid = frame->root->gid; \
+ frame->root->uid = 0; \
+ frame->root->gid = 0; \
+ local->is_set_fsid = _gf_true; \
+ } \
+ } while (0)
-#define SHARD_SET_ROOT_FS_ID(frame, local) do { \
- if (!local->is_set_fsid) { \
- local->uid = frame->root->uid; \
- local->gid = frame->root->gid; \
- frame->root->uid = 0; \
- frame->root->gid = 0; \
- local->is_set_fsid = _gf_true; \
- } \
-} while (0)
+#define SHARD_UNSET_ROOT_FS_ID(frame, local) \
+ do { \
+ if (local->is_set_fsid) { \
+ frame->root->uid = local->uid; \
+ frame->root->gid = local->gid; \
+ local->is_set_fsid = _gf_false; \
+ } \
+ } while (0)
-#define SHARD_UNSET_ROOT_FS_ID(frame, local) do { \
- if (local->is_set_fsid) { \
- frame->root->uid = local->uid; \
- frame->root->gid = local->gid; \
- local->is_set_fsid = _gf_false; \
- } \
-} while (0)
+#define SHARD_TIME_UPDATE(ctx_sec, ctx_nsec, new_sec, new_nsec) \
+ do { \
+ if (ctx_sec == new_sec) \
+ ctx_nsec = new_nsec = max(new_nsec, ctx_nsec); \
+ else if (ctx_sec > new_sec) { \
+ new_sec = ctx_sec; \
+ new_nsec = ctx_nsec; \
+ } else { \
+ ctx_sec = new_sec; \
+ ctx_nsec = new_nsec; \
+ } \
+ } while (0)
-#define SHARD_TIME_UPDATE(ctx_sec, ctx_nsec, new_sec, new_nsec) do { \
- if (ctx_sec == new_sec) \
- ctx_nsec = new_nsec = max (new_nsec, ctx_nsec); \
- else if (ctx_sec > new_sec) { \
- new_sec = ctx_sec; \
- new_nsec = ctx_nsec; \
- } else { \
- ctx_sec = new_sec; \
- ctx_nsec = new_nsec; \
- } \
- } while (0)
+typedef enum {
+ SHARD_BG_DELETION_NONE = 0,
+ SHARD_BG_DELETION_LAUNCHING,
+ SHARD_BG_DELETION_IN_PROGRESS,
+} shard_bg_deletion_state_t;
+/* rm = "remove me" */
typedef struct shard_priv {
- uint64_t block_size;
- uuid_t dot_shard_gfid;
- inode_t *dot_shard_inode;
- gf_lock_t lock;
- int inode_count;
- struct list_head ilist_head;
+ uint64_t block_size;
+ uuid_t dot_shard_gfid;
+ uuid_t dot_shard_rm_gfid;
+ inode_t *dot_shard_inode;
+ inode_t *dot_shard_rm_inode;
+ gf_lock_t lock;
+ int inode_count;
+ struct list_head ilist_head;
+ uint32_t deletion_rate;
+ shard_bg_deletion_state_t bg_del_state;
+ gf_boolean_t first_lookup_done;
+ uint64_t lru_limit;
} shard_priv_t;
typedef struct {
- loc_t *loc;
- short type;
- char *domain;
-} shard_lock_t;
+ loc_t loc;
+ char *domain;
+ struct gf_flock flock;
+ gf_boolean_t acquired_lock;
+} shard_inodelk_t;
+
+typedef struct {
+ loc_t loc;
+ char *domain;
+ char *basename;
+ entrylk_cmd cmd;
+ entrylk_type type;
+ gf_boolean_t acquired_lock;
+} shard_entrylk_t;
+
+typedef int32_t (*shard_post_fop_handler_t)(call_frame_t *frame,
+ xlator_t *this);
+typedef int32_t (*shard_post_resolve_fop_handler_t)(call_frame_t *frame,
+ xlator_t *this);
+typedef int32_t (*shard_post_lookup_shards_fop_handler_t)(call_frame_t *frame,
+ xlator_t *this);
-typedef int32_t (*shard_post_fop_handler_t) (call_frame_t *frame,
- xlator_t *this);
-typedef int32_t (*shard_post_resolve_fop_handler_t) (call_frame_t *frame,
- xlator_t *this);
-typedef int32_t (*shard_post_lookup_shards_fop_handler_t) (call_frame_t *frame,
- xlator_t *this);
+typedef int32_t (*shard_post_mknod_fop_handler_t)(call_frame_t *frame,
+ xlator_t *this);
-typedef int32_t (*shard_post_mknod_fop_handler_t) (call_frame_t *frame,
- xlator_t *this);
+typedef int32_t (*shard_post_update_size_fop_handler_t)(call_frame_t *frame,
+ xlator_t *this);
-typedef int32_t (*shard_post_update_size_fop_handler_t) (call_frame_t *frame,
- xlator_t *this);
typedef struct shard_local {
- int op_ret;
- int op_errno;
- int first_block;
- int last_block;
- int num_blocks;
- int call_count;
- int eexist_count;
- int create_count;
- int xflag;
- int count;
- uint32_t flags;
- uint32_t uid;
- uint32_t gid;
- uint64_t block_size;
- uint64_t dst_block_size;
- int32_t datasync;
- off_t offset;
- size_t total_size;
- size_t written_size;
- size_t hole_size;
- size_t req_size;
- size_t readdir_size;
- int64_t delta_size;
- int delta_blocks;
- loc_t loc;
- loc_t dot_shard_loc;
- loc_t loc2;
- loc_t tmp_loc;
- fd_t *fd;
- dict_t *xattr_req;
- dict_t *xattr_rsp;
- inode_t **inode_list;
- glusterfs_fop_t fop;
- struct iatt prebuf;
- struct iatt postbuf;
- struct iatt preoldparent;
- struct iatt postoldparent;
- struct iatt prenewparent;
- struct iatt postnewparent;
- struct iovec *vector;
- struct iobref *iobref;
- struct iobuf *iobuf;
- gf_dirent_t entries_head;
- gf_boolean_t is_set_fsid;
- gf_boolean_t list_inited;
- shard_post_fop_handler_t handler;
- shard_post_lookup_shards_fop_handler_t pls_fop_handler;
- shard_post_resolve_fop_handler_t post_res_handler;
- shard_post_mknod_fop_handler_t post_mknod_handler;
- shard_post_update_size_fop_handler_t post_update_size_handler;
- struct {
- int lock_count;
- fop_inodelk_cbk_t inodelk_cbk;
- shard_lock_t *shard_lock;
- } lock;
- inode_t *resolver_base_inode;
- gf_boolean_t first_lookup_done;
- syncbarrier_t barrier;
- gf_boolean_t lookup_shards_barriered;
- gf_boolean_t unlink_shards_barriered;
+ int op_ret;
+ int op_errno;
+ uint64_t first_block;
+ uint64_t last_block;
+ uint64_t num_blocks;
+ int call_count;
+ int eexist_count;
+ int create_count;
+ int xflag;
+ int count;
+ uint32_t flags;
+ uint32_t uid;
+ uint32_t gid;
+ uint64_t block_size;
+ uint64_t dst_block_size;
+ int32_t datasync;
+ off_t offset;
+ size_t total_size;
+ size_t written_size;
+ size_t hole_size;
+ size_t req_size;
+ size_t readdir_size;
+ int64_t delta_size;
+ gf_atomic_t delta_blocks;
+ loc_t loc;
+ loc_t dot_shard_loc;
+ loc_t dot_shard_rm_loc;
+ loc_t loc2;
+ loc_t tmp_loc;
+ fd_t *fd;
+ dict_t *xattr_req;
+ dict_t *xattr_rsp;
+ inode_t **inode_list;
+ glusterfs_fop_t fop;
+ struct iatt prebuf;
+ struct iatt postbuf;
+ struct iatt preoldparent;
+ struct iatt postoldparent;
+ struct iatt prenewparent;
+ struct iatt postnewparent;
+ struct iovec *vector;
+ struct iobref *iobref;
+ struct iobuf *iobuf;
+ gf_dirent_t entries_head;
+ gf_boolean_t is_set_fsid;
+ gf_boolean_t list_inited;
+ shard_post_fop_handler_t handler;
+ shard_post_lookup_shards_fop_handler_t pls_fop_handler;
+ shard_post_resolve_fop_handler_t post_res_handler;
+ shard_post_mknod_fop_handler_t post_mknod_handler;
+ shard_post_update_size_fop_handler_t post_update_size_handler;
+ shard_inodelk_t int_inodelk;
+ shard_entrylk_t int_entrylk;
+ inode_t *resolver_base_inode;
+ gf_boolean_t first_lookup_done;
+ syncbarrier_t barrier;
+ gf_boolean_t lookup_shards_barriered;
+ gf_boolean_t unlink_shards_barriered;
+ gf_boolean_t resolve_not;
+ loc_t newloc;
+ call_frame_t *main_frame;
+ call_frame_t *inodelk_frame;
+ call_frame_t *entrylk_frame;
+ uint32_t deletion_rate;
+ gf_boolean_t cleanup_required;
+ uuid_t base_gfid;
+ char *name;
} shard_local_t;
typedef struct shard_inode_ctx {
- uint64_t block_size; /* The block size with which this inode is
- sharded */
- struct iatt stat;
- gf_boolean_t refresh;
- /* The following members of inode ctx will be applicable only to the
- * individual shards' ctx and never the base file ctx.
- */
- struct list_head ilist;
- uuid_t base_gfid;
- int block_num;
- gf_boolean_t refreshed;
- struct list_head to_fsync_list;
- int fsync_needed;
- inode_t *inode;
- int fsync_count;
- inode_t *base_inode;
+ uint64_t block_size; /* The block size with which this inode is
+ sharded */
+ struct iatt stat;
+ gf_boolean_t refresh;
+ /* The following members of inode ctx will be applicable only to the
+ * individual shards' ctx and never the base file ctx.
+ */
+ struct list_head ilist;
+ uuid_t base_gfid;
+ int block_num;
+ gf_boolean_t refreshed;
+ struct list_head to_fsync_list;
+ int fsync_needed;
+ inode_t *inode;
+ int fsync_count;
+ inode_t *base_inode;
} shard_inode_ctx_t;
typedef enum {
- SHARD_INTERNAL_DIR_DOT_SHARD = 1,
+ SHARD_INTERNAL_DIR_DOT_SHARD = 1,
+ SHARD_INTERNAL_DIR_DOT_SHARD_REMOVE_ME,
} shard_internal_dir_type_t;
#endif /* __SHARD_H__ */
diff --git a/xlators/features/snapview-client/src/Makefile.am b/xlators/features/snapview-client/src/Makefile.am
index a40d2b291ad..fa08656c537 100644
--- a/xlators/features/snapview-client/src/Makefile.am
+++ b/xlators/features/snapview-client/src/Makefile.am
@@ -6,7 +6,7 @@ snapview_client_la_LDFLAGS = -module $(GF_XLATOR_DEFAULT_LDFLAGS)
snapview_client_la_SOURCES = snapview-client.c
snapview_client_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-noinst_HEADERS = snapview-client.h snapview-client-mem-types.h
+noinst_HEADERS = snapview-client.h snapview-client-mem-types.h snapview-client-messages.h
AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
-I$(top_srcdir)/rpc/xdr/src -I$(top_builddir)/rpc/xdr/src
diff --git a/xlators/features/snapview-client/src/snapview-client-mem-types.h b/xlators/features/snapview-client/src/snapview-client-mem-types.h
index 1a0158d950e..3c3ab555a55 100644
--- a/xlators/features/snapview-client/src/snapview-client-mem-types.h
+++ b/xlators/features/snapview-client/src/snapview-client-mem-types.h
@@ -11,14 +11,14 @@
#ifndef _SVC_MEM_TYPES_H
#define _SVC_MEM_TYPES_H
-#include "mem-types.h"
+#include <glusterfs/mem-types.h>
enum svc_mem_types {
- gf_svc_mt_svc_private_t = gf_common_mt_end + 1,
- gf_svc_mt_svc_local_t,
- gf_svc_mt_svc_inode_t,
- gf_svc_mt_svc_fd_t,
- gf_svc_mt_end
+ gf_svc_mt_svc_private_t = gf_common_mt_end + 1,
+ gf_svc_mt_svc_local_t,
+ gf_svc_mt_svc_inode_t,
+ gf_svc_mt_svc_fd_t,
+ gf_svc_mt_end
};
#endif
diff --git a/xlators/features/snapview-client/src/snapview-client-messages.h b/xlators/features/snapview-client/src/snapview-client-messages.h
new file mode 100644
index 00000000000..c02fb154930
--- /dev/null
+++ b/xlators/features/snapview-client/src/snapview-client-messages.h
@@ -0,0 +1,71 @@
+/*
+ Copyright (c) 2018 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
+ */
+
+#ifndef _SNAPVIEW_CLIENT_MESSAGES_H_
+#define _SNAPVIEW_CLIENT_MESSAGES_H_
+
+#include <glusterfs/glfs-message-id.h>
+
+/* To add new message IDs, append new identifiers at the end of the list.
+ *
+ * Never remove a message ID. If it's not used anymore, you can rename it or
+ * leave it as it is, but not delete it. This is to prevent reutilization of
+ * IDs by other messages.
+ *
+ * The component name must match one of the entries defined in
+ * glfs-message-id.h.
+ */
+
+GLFS_MSGID(SNAPVIEW_CLIENT, SVC_MSG_NO_MEMORY, SVC_MSG_MEM_ACNT_FAILED,
+ SVC_MSG_SET_INODE_CONTEXT_FAILED, SVC_MSG_GET_INODE_CONTEXT_FAILED,
+ SVC_MSG_DELETE_INODE_CONTEXT_FAILED, SVC_MSG_SET_FD_CONTEXT_FAILED,
+ SVC_MSG_GET_FD_CONTEXT_FAILED, SVC_MSG_DICT_SET_FAILED,
+ SVC_MSG_SUBVOLUME_NULL, SVC_MSG_NO_CHILD_FOR_XLATOR,
+ SVC_MSG_XLATOR_CHILDREN_WRONG, SVC_MSG_NORMAL_GRAPH_LOOKUP_FAIL,
+ SVC_MSG_SNAPVIEW_GRAPH_LOOKUP_FAIL, SVC_MSG_OPENDIR_SPECIAL_DIR,
+ SVC_MSG_RENAME_SNAPSHOT_ENTRY, SVC_MSG_LINK_SNAPSHOT_ENTRY,
+ SVC_MSG_COPY_ENTRY_POINT_FAILED, SVC_MSG_ENTRY_POINT_SPECIAL_DIR,
+ SVC_MSG_STR_LEN, SVC_MSG_INVALID_ENTRY_POINT, SVC_MSG_NULL_PRIV,
+ SVC_MSG_PRIV_DESTROY_FAILED, SVC_MSG_ALLOC_FD_FAILED,
+ SVC_MSG_ALLOC_INODE_FAILED, SVC_MSG_NULL_SPECIAL_DIR,
+ SVC_MSG_MEM_POOL_GET_FAILED);
+
+#define SVC_MSG_ALLOC_FD_FAILED_STR "failed to allocate new fd context"
+#define SVC_MSG_SET_FD_CONTEXT_FAILED_STR "failed to set fd context"
+#define SVC_MSG_STR_LEN_STR \
+ "destination buffer size is less than the length of entry point name"
+#define SVC_MSG_NORMAL_GRAPH_LOOKUP_FAIL_STR "lookup failed on normal graph"
+#define SVC_MSG_SNAPVIEW_GRAPH_LOOKUP_FAIL_STR "lookup failed on snapview graph"
+#define SVC_MSG_SET_INODE_CONTEXT_FAILED_STR "failed to set inode context"
+#define SVC_MSG_NO_MEMORY_STR "failed to allocate memory"
+#define SVC_MSG_COPY_ENTRY_POINT_FAILED_STR \
+ "failed to copy the entry point string"
+#define SVC_MSG_GET_FD_CONTEXT_FAILED_STR "fd context not found"
+#define SVC_MSG_GET_INODE_CONTEXT_FAILED_STR "failed to get inode context"
+#define SVC_MSG_ALLOC_INODE_FAILED_STR "failed to allocate new inode"
+#define SVC_MSG_DICT_SET_FAILED_STR "failed to set dict"
+#define SVC_MSG_RENAME_SNAPSHOT_ENTRY_STR \
+ "rename happening on a entry residing in snapshot"
+#define SVC_MSG_DELETE_INODE_CONTEXT_FAILED_STR "failed to delete inode context"
+#define SVC_MSG_NULL_PRIV_STR "priv NULL"
+#define SVC_MSG_INVALID_ENTRY_POINT_STR "not a valid entry point"
+#define SVC_MSG_MEM_ACNT_FAILED_STR "Memory accouting init failed"
+#define SVC_MSG_NO_CHILD_FOR_XLATOR_STR "configured without any child"
+#define SVC_MSG_XLATOR_CHILDREN_WRONG_STR \
+ "snap-view-client has got wrong subvolumes. It can have only 2"
+#define SVC_MSG_ENTRY_POINT_SPECIAL_DIR_STR \
+ "entry point directory cannot be part of special directory"
+#define SVC_MSG_NULL_SPECIAL_DIR_STR "null special directory"
+#define SVC_MSG_MEM_POOL_GET_FAILED_STR \
+ "could not get mem pool for frame->local"
+#define SVC_MSG_PRIV_DESTROY_FAILED_STR "failed to destroy private"
+#define SVC_MSG_LINK_SNAPSHOT_ENTRY_STR \
+ "link happening on a entry residin gin snapshot"
+#endif /* !_SNAPVIEW_CLIENT_MESSAGES_H_ */
diff --git a/xlators/features/snapview-client/src/snapview-client.c b/xlators/features/snapview-client/src/snapview-client.c
index 63ed160bd71..486c5179d5b 100644
--- a/xlators/features/snapview-client/src/snapview-client.c
+++ b/xlators/features/snapview-client/src/snapview-client.c
@@ -1,539 +1,593 @@
- /*
- Copyright (c) 2014 Red Hat, Inc. <http://www.redhat.com>
- This file is part of GlusterFS.
-
- This file is licensed to you under your choice of the GNU Lesser
- General Public License, version 3 or any later version (LGPLv3 or
- later), or the GNU General Public License, version 2 (GPLv2), in all
- cases as published by the Free Software Foundation.
+/*
+ Copyright (c) 2014 Red Hat, Inc. <http://www.redhat.com>
+ This file is part of GlusterFS.
+
+ This file is licensed to you under your choice of the GNU Lesser
+ General Public License, version 3 or any later version (LGPLv3 or
+ later), or the GNU General Public License, version 2 (GPLv2), in all
+ cases as published by the Free Software Foundation.
*/
#include "snapview-client.h"
-#include "inode.h"
-#include "byte-order.h"
-
+#include <glusterfs/inode.h>
+#include <glusterfs/byte-order.h>
static void
-svc_local_free (svc_local_t *local)
+svc_local_free(svc_local_t *local)
{
- if (local) {
- loc_wipe (&local->loc);
- if (local->fd)
- fd_unref (local->fd);
- if (local->xdata)
- dict_unref (local->xdata);
- mem_put (local);
- }
+ if (local) {
+ loc_wipe(&local->loc);
+ if (local->fd)
+ fd_unref(local->fd);
+ if (local->xdata)
+ dict_unref(local->xdata);
+ mem_put(local);
+ }
}
static xlator_t *
-svc_get_subvolume (xlator_t *this, int inode_type)
+svc_get_subvolume(xlator_t *this, int inode_type)
{
- xlator_t *subvolume = NULL;
+ xlator_t *subvolume = NULL;
- GF_VALIDATE_OR_GOTO ("snapview-client", this, out);
+ GF_VALIDATE_OR_GOTO("snapview-client", this, out);
- if (inode_type == VIRTUAL_INODE)
- subvolume = SECOND_CHILD (this);
- else
- subvolume = FIRST_CHILD (this);
+ if (inode_type == VIRTUAL_INODE)
+ subvolume = SECOND_CHILD(this);
+ else
+ subvolume = FIRST_CHILD(this);
out:
- return subvolume;
+ return subvolume;
}
static int32_t
-__svc_inode_ctx_set (xlator_t *this, inode_t *inode, int inode_type)
+__svc_inode_ctx_set(xlator_t *this, inode_t *inode, int inode_type)
{
- uint64_t value = 0;
- int32_t ret = -1;
+ uint64_t value = 0;
+ int32_t ret = -1;
- GF_VALIDATE_OR_GOTO ("snapview-client", this, out);
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO("snapview-client", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- value = inode_type;
+ value = inode_type;
- ret = __inode_ctx_set (inode, this, &value);
+ ret = __inode_ctx_set(inode, this, &value);
out:
- return ret;
+ return ret;
}
static int
-__svc_inode_ctx_get (xlator_t *this, inode_t *inode, int *inode_type)
+__svc_inode_ctx_get(xlator_t *this, inode_t *inode, int *inode_type)
{
- uint64_t value = 0;
- int ret = -1;
+ uint64_t value = 0;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO ("snapview-client", this, out);
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO("snapview-client", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- ret = __inode_ctx_get (inode, this, &value);
- if (ret < 0)
- goto out;
+ ret = __inode_ctx_get(inode, this, &value);
+ if (ret < 0)
+ goto out;
- *inode_type = (int)(value);
+ *inode_type = (int)(value);
out:
- return ret;
+ return ret;
}
static int
-svc_inode_ctx_get (xlator_t *this, inode_t *inode, int *inode_type)
+svc_inode_ctx_get(xlator_t *this, inode_t *inode, int *inode_type)
{
- int ret = -1;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO ("snapview-client", this, out);
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO("snapview-client", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- LOCK (&inode->lock);
- {
- ret = __svc_inode_ctx_get (this, inode, inode_type);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __svc_inode_ctx_get(this, inode, inode_type);
+ }
+ UNLOCK(&inode->lock);
out:
- return ret;
+ return ret;
}
static int32_t
-svc_inode_ctx_set (xlator_t *this, inode_t *inode, int inode_type)
+svc_inode_ctx_set(xlator_t *this, inode_t *inode, int inode_type)
{
- int32_t ret = -1;
+ int32_t ret = -1;
- GF_VALIDATE_OR_GOTO ("snapview-client", this, out);
- GF_VALIDATE_OR_GOTO (this->name, inode, out);
+ GF_VALIDATE_OR_GOTO("snapview-client", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, inode, out);
- LOCK (&inode->lock);
- {
- ret = __svc_inode_ctx_set (this, inode, inode_type);
- }
- UNLOCK (&inode->lock);
+ LOCK(&inode->lock);
+ {
+ ret = __svc_inode_ctx_set(this, inode, inode_type);
+ }
+ UNLOCK(&inode->lock);
out:
- return ret;
+ return ret;
}
static svc_fd_t *
-svc_fd_new (void)
+svc_fd_new(void)
{
- svc_fd_t *svc_fd = NULL;
+ svc_fd_t *svc_fd = NULL;
- svc_fd = GF_CALLOC (1, sizeof (*svc_fd), gf_svc_mt_svc_fd_t);
+ svc_fd = GF_CALLOC(1, sizeof(*svc_fd), gf_svc_mt_svc_fd_t);
- return svc_fd;
+ return svc_fd;
}
static svc_fd_t *
-__svc_fd_ctx_get (xlator_t *this, fd_t *fd)
+__svc_fd_ctx_get(xlator_t *this, fd_t *fd)
{
- svc_fd_t *svc_fd = NULL;
- uint64_t value = 0;
- int ret = -1;
+ svc_fd_t *svc_fd = NULL;
+ uint64_t value = 0;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO ("snapview-client", this, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
+ GF_VALIDATE_OR_GOTO("snapview-client", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
- ret = __fd_ctx_get (fd, this, &value);
- if (ret)
- return NULL;
+ ret = __fd_ctx_get(fd, this, &value);
+ if (ret)
+ return NULL;
- svc_fd = (svc_fd_t *) ((long) value);
+ svc_fd = (svc_fd_t *)((long)value);
out:
- return svc_fd;
+ return svc_fd;
}
static svc_fd_t *
-svc_fd_ctx_get (xlator_t *this, fd_t *fd)
+svc_fd_ctx_get(xlator_t *this, fd_t *fd)
{
- svc_fd_t *svc_fd = NULL;
+ svc_fd_t *svc_fd = NULL;
- GF_VALIDATE_OR_GOTO ("snapview-client", this, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
+ GF_VALIDATE_OR_GOTO("snapview-client", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
- LOCK (&fd->lock);
- {
- svc_fd = __svc_fd_ctx_get (this, fd);
- }
- UNLOCK (&fd->lock);
+ LOCK(&fd->lock);
+ {
+ svc_fd = __svc_fd_ctx_get(this, fd);
+ }
+ UNLOCK(&fd->lock);
out:
- return svc_fd;
+ return svc_fd;
}
static int
-__svc_fd_ctx_set (xlator_t *this, fd_t *fd, svc_fd_t *svc_fd)
+__svc_fd_ctx_set(xlator_t *this, fd_t *fd, svc_fd_t *svc_fd)
{
- uint64_t value = 0;
- int ret = -1;
+ uint64_t value = 0;
+ int ret = -1;
- GF_VALIDATE_OR_GOTO ("snapview-client", this, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
- GF_VALIDATE_OR_GOTO (this->name, svc_fd, out);
+ GF_VALIDATE_OR_GOTO("snapview-client", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
+ GF_VALIDATE_OR_GOTO(this->name, svc_fd, out);
- value = (uint64_t)(long) svc_fd;
+ value = (uint64_t)(long)svc_fd;
- ret = __fd_ctx_set (fd, this, value);
+ ret = __fd_ctx_set(fd, this, value);
out:
- return ret;
+ return ret;
}
static svc_fd_t *
-__svc_fd_ctx_get_or_new (xlator_t *this, fd_t *fd)
+__svc_fd_ctx_get_or_new(xlator_t *this, fd_t *fd)
{
- svc_fd_t *svc_fd = NULL;
- int ret = -1;
- inode_t *inode = NULL;
-
- GF_VALIDATE_OR_GOTO ("snapview-client", this, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
+ svc_fd_t *svc_fd = NULL;
+ int ret = -1;
+ inode_t *inode = NULL;
- inode = fd->inode;
- svc_fd = __svc_fd_ctx_get (this, fd);
- if (svc_fd) {
- ret = 0;
- goto out;
- }
+ GF_VALIDATE_OR_GOTO("snapview-client", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
- svc_fd = svc_fd_new ();
- if (!svc_fd) {
- gf_log (this->name, GF_LOG_ERROR, "failed to allocate new fd "
- "context for gfid %s", uuid_utoa (inode->gfid));
- goto out;
- }
-
- ret = __svc_fd_ctx_set (this, fd, svc_fd);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "failed to set fd context "
- "for gfid %s", uuid_utoa (inode->gfid));
- ret = -1;
- }
+ inode = fd->inode;
+ svc_fd = __svc_fd_ctx_get(this, fd);
+ if (svc_fd) {
+ ret = 0;
+ goto out;
+ }
+
+ svc_fd = svc_fd_new();
+ if (!svc_fd) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, SVC_MSG_ALLOC_FD_FAILED,
+ "gfid=%s", uuid_utoa(inode->gfid), NULL);
+ goto out;
+ }
+
+ ret = __svc_fd_ctx_set(this, fd, svc_fd);
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, SVC_MSG_SET_FD_CONTEXT_FAILED,
+ "gfid=%s", uuid_utoa(inode->gfid), NULL);
+ ret = -1;
+ }
out:
- if (ret) {
- GF_FREE (svc_fd);
- svc_fd = NULL;
- }
+ if (ret) {
+ GF_FREE(svc_fd);
+ svc_fd = NULL;
+ }
- return svc_fd;
+ return svc_fd;
}
static svc_fd_t *
-svc_fd_ctx_get_or_new (xlator_t *this, fd_t *fd)
+svc_fd_ctx_get_or_new(xlator_t *this, fd_t *fd)
{
- svc_fd_t *svc_fd = NULL;
+ svc_fd_t *svc_fd = NULL;
- GF_VALIDATE_OR_GOTO ("snapview-client", this, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
+ GF_VALIDATE_OR_GOTO("snapview-client", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
- LOCK (&fd->lock);
- {
- svc_fd = __svc_fd_ctx_get_or_new (this, fd);
- }
- UNLOCK (&fd->lock);
+ LOCK(&fd->lock);
+ {
+ svc_fd = __svc_fd_ctx_get_or_new(this, fd);
+ }
+ UNLOCK(&fd->lock);
out:
- return svc_fd;
+ return svc_fd;
}
-
-static int32_t
-gf_svc_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+/**
+ * @this: xlator
+ * @entry_point: pointer to the buffer provided by consumer
+ *
+ * This function is mainly for copying the entry point name
+ * (stored as string in priv->path) to a buffer point to by
+ * @entry_point within the lock. It is for the consumer to
+ * allocate the memory for the buffer.
+ *
+ * This function is called by all the functions (or fops)
+ * who need to use priv->path for avoiding the race.
+ * For example, either in lookup or in any other fop,
+ * while priv->path is being accessed, a reconfigure can
+ * happen to change priv->path. This ensures that, a lock
+ * is taken before accessing priv->path.
+ **/
+int
+gf_svc_get_entry_point(xlator_t *this, char *entry_point, size_t dest_size)
{
- svc_local_t *local = NULL;
- xlator_t *subvolume = NULL;
- gf_boolean_t do_unwind = _gf_true;
- int inode_type = -1;
- int ret = -1;
-
- local = frame->local;
- subvolume = local->subvolume;
- if (!subvolume) {
- gf_log_callingfn (this->name, GF_LOG_ERROR, "path: %s, "
- "gfid: %s ", local->loc.path,
- inode?uuid_utoa (inode->gfid):"");
- GF_ASSERT (0);
- }
+ int ret = -1;
+ svc_private_t *priv = NULL;
- /* There is a possibility that, the client process just came online
- and does not have the inode on which the lookup came. In that case,
- the fresh inode created from fuse for the lookup fop, wont have
- the inode context set without which svc cannot decide where to
- STACK_WIND to. So by default it decides to send the fop to the
- regular subvolume (i.e first child of the xlator). If lookup fails
- on the regular volume, then there is a possibility that the lookup
- is happening on a virtual inode (i.e history data residing in snaps).
- So if lookup fails with ENOENT and the inode context is not there,
- then send the lookup to the 2nd child of svc.
-
- If there are any changes in volfile/client-restarted then inode-ctx
- is lost. In this case if nameless lookup fails with ESTALE,
- then send the lookup to the 2nd child of svc.
- */
- if (op_ret) {
- if (subvolume == FIRST_CHILD (this)) {
- gf_log (this->name,
- (op_errno == ENOENT || op_errno == ESTALE)
- ? GF_LOG_DEBUG:GF_LOG_ERROR,
- "Lookup failed on normal graph with error %s",
- strerror (op_errno));
- } else {
- gf_log (this->name,
- (op_errno == ENOENT || op_errno == ESTALE)
- ? GF_LOG_DEBUG:GF_LOG_ERROR,
- "Lookup failed on snapview graph with error %s",
- strerror (op_errno));
- goto out;
- }
-
- if ((op_errno == ENOENT || op_errno == ESTALE) &&
- !gf_uuid_is_null (local->loc.gfid)) {
- if (inode != NULL)
- ret = svc_inode_ctx_get (this, inode,
- &inode_type);
-
- if (ret < 0 || inode == NULL) {
- gf_log (this->name, GF_LOG_DEBUG,
- "Lookup on normal graph failed. "
- "Sending lookup to snapview-server");
-
- subvolume = SECOND_CHILD (this);
- local->subvolume = subvolume;
- STACK_WIND (frame, gf_svc_lookup_cbk,
- subvolume, subvolume->fops->lookup,
- &local->loc, xdata);
- do_unwind = _gf_false;
- }
- }
+ GF_VALIDATE_OR_GOTO("snapview-client", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, entry_point, out);
- goto out;
- }
+ priv = this->private;
- if (subvolume == FIRST_CHILD (this))
- inode_type = NORMAL_INODE;
- else
- inode_type = VIRTUAL_INODE;
-
- ret = svc_inode_ctx_set (this, inode, inode_type);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR, "failed to set inode type"
- "into the context");
-
-out:
- if (do_unwind) {
- SVC_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, buf,
- xdata, postparent);
+ LOCK(&priv->lock);
+ {
+ if (dest_size <= strlen(priv->path)) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, SVC_MSG_STR_LEN,
+ "dest-size=%zu", dest_size, "priv-path-len=%zu",
+ strlen(priv->path), "path=%s", priv->path, NULL);
+ } else {
+ snprintf(entry_point, dest_size, "%s", priv->path);
+ ret = 0;
}
+ }
+ UNLOCK(&priv->lock);
- return 0;
+out:
+ return ret;
}
static int32_t
-gf_svc_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+gf_svc_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata, struct iatt *postparent)
{
- int32_t ret = -1;
- svc_local_t *local = NULL;
- xlator_t *subvolume = NULL;
- int op_ret = -1;
- int op_errno = EINVAL;
- inode_t *parent = NULL;
- svc_private_t *priv = NULL;
- dict_t *new_xdata = NULL;
- int inode_type = -1;
- int parent_type = -1;
- gf_boolean_t wind = _gf_false;
-
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
-
- priv = this->private;
-
- ret = svc_inode_ctx_get (this, loc->inode, &inode_type);
- if (!__is_root_gfid (loc->gfid)) {
- if (loc->parent) {
- parent = inode_ref (loc->parent);
- ret = svc_inode_ctx_get (this, loc->parent,
- &parent_type);
- } else {
- parent = inode_parent (loc->inode, loc->pargfid, NULL);
- if (parent)
- ret = svc_inode_ctx_get (this, parent,
- &parent_type);
- }
- }
-
- local = mem_get0 (this->local_pool);
- if (!local) {
- gf_log (this->name, GF_LOG_ERROR, "failed to allocate local");
- op_ret = -1;
- op_errno = ENOMEM;
- goto out;
- }
-
- frame->local = local;
- loc_copy (&local->loc, loc);
-
- if (__is_root_gfid (loc->inode->gfid)) {
- subvolume = FIRST_CHILD (this);
- GF_ASSERT (subvolume);
+ svc_local_t *local = NULL;
+ xlator_t *subvolume = NULL;
+ gf_boolean_t do_unwind = _gf_true;
+ int inode_type = -1;
+ int ret = -1;
+
+ local = frame->local;
+ subvolume = local->subvolume;
+ if (!subvolume) {
+ gf_msg_callingfn(this->name, GF_LOG_ERROR, 0, SVC_MSG_SUBVOLUME_NULL,
+ "path: %s gfid: %s ", local->loc.path,
+ inode ? uuid_utoa(inode->gfid) : "");
+ GF_ASSERT(0);
+ }
+
+ /* There is a possibility that, the client process just came online
+ and does not have the inode on which the lookup came. In that case,
+ the fresh inode created from fuse for the lookup fop, won't have
+ the inode context set without which svc cannot decide where to
+ STACK_WIND to. So by default it decides to send the fop to the
+ regular subvolume (i.e first child of the xlator). If lookup fails
+ on the regular volume, then there is a possibility that the lookup
+ is happening on a virtual inode (i.e history data residing in snaps).
+ So if lookup fails with ENOENT and the inode context is not there,
+ then send the lookup to the 2nd child of svc.
+
+ If there are any changes in volfile/client-restarted then inode-ctx
+ is lost. In this case if nameless lookup fails with ESTALE,
+ then send the lookup to the 2nd child of svc.
+ */
+ if (op_ret) {
+ if (subvolume == FIRST_CHILD(this)) {
+ gf_smsg(this->name,
+ (op_errno == ENOENT || op_errno == ESTALE) ? GF_LOG_DEBUG
+ : GF_LOG_ERROR,
+ op_errno, SVC_MSG_NORMAL_GRAPH_LOOKUP_FAIL, "error=%s",
+ strerror(op_errno), NULL);
+ } else {
+ gf_smsg(this->name,
+ (op_errno == ENOENT || op_errno == ESTALE) ? GF_LOG_DEBUG
+ : GF_LOG_ERROR,
+ op_errno, SVC_MSG_SNAPVIEW_GRAPH_LOOKUP_FAIL, "error=%s",
+ strerror(op_errno), NULL);
+ goto out;
+ }
+
+ if ((op_errno == ENOENT || op_errno == ESTALE) &&
+ !gf_uuid_is_null(local->loc.gfid)) {
+ if (inode != NULL)
+ ret = svc_inode_ctx_get(this, inode, &inode_type);
+
+ if (ret < 0 || inode == NULL) {
+ gf_msg_debug(this->name, 0,
+ "Lookup on normal graph failed. "
+ " Sending lookup to snapview-server");
+ subvolume = SECOND_CHILD(this);
local->subvolume = subvolume;
- wind = _gf_true;
- goto out;
+ STACK_WIND(frame, gf_svc_lookup_cbk, subvolume,
+ subvolume->fops->lookup, &local->loc, xdata);
+ do_unwind = _gf_false;
+ }
}
- /* nfs sends nameless lookups directly using the gfid. In that case
- loc->name will be NULL. So check if loc->name is NULL. If so, then
- try to get the subvolume using inode context. But if the inode has
- not been looked up yet, then send the lookup call to the first
- subvolume.
- */
-
- if (!loc->name) {
- if (gf_uuid_is_null (loc->inode->gfid)) {
- subvolume = FIRST_CHILD (this);
- local->subvolume = subvolume;
- wind = _gf_true;
- goto out;
- } else {
- if (inode_type >= 0)
- subvolume = svc_get_subvolume (this,
- inode_type);
- else
- subvolume = FIRST_CHILD (this);
- local->subvolume = subvolume;
- wind = _gf_true;
- goto out;
- }
- }
+ goto out;
+ }
- if (strcmp (loc->name, priv->path)) {
- if (parent_type == VIRTUAL_INODE) {
- subvolume = SECOND_CHILD (this);
- } else {
- /*
- * Either parent type is normal graph, or the parent
- * type is uncertain.
- */
- subvolume = FIRST_CHILD (this);
- }
- local->subvolume = subvolume;
- } else {
- subvolume = SECOND_CHILD (this);
- local->subvolume = subvolume;
- if (parent_type == NORMAL_INODE) {
- /* Indication of whether the lookup is happening on the
- entry point or not, to the snapview-server.
- */
- SVC_ENTRY_POINT_SET (this, xdata, op_ret, op_errno,
- new_xdata, priv, ret, out);
- }
- }
+ if (subvolume == FIRST_CHILD(this))
+ inode_type = NORMAL_INODE;
+ else
+ inode_type = VIRTUAL_INODE;
- wind = _gf_true;
+ ret = svc_inode_ctx_set(this, inode, inode_type);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, 0, SVC_MSG_SET_INODE_CONTEXT_FAILED,
+ "gfid=%s", uuid_utoa(inode->gfid), NULL);
out:
- if (wind)
- STACK_WIND (frame, gf_svc_lookup_cbk, subvolume,
- subvolume->fops->lookup, loc, xdata);
- else
- SVC_STACK_UNWIND (lookup, frame, op_ret, op_errno, NULL,
- NULL, NULL, NULL);
- if (new_xdata)
- dict_unref (new_xdata);
+ if (do_unwind) {
+ SVC_STACK_UNWIND(lookup, frame, op_ret, op_errno, inode, buf, xdata,
+ postparent);
+ }
- if (parent)
- inode_unref (parent);
-
- return 0;
+ return 0;
}
static int32_t
-gf_svc_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+gf_svc_lookup(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- xlator_t *subvolume = NULL;
- int32_t ret = -1;
- int inode_type = -1;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- gf_boolean_t wind = _gf_false;
- svc_private_t *priv = NULL;
- const char *path = NULL;
- int path_len = -1;
- int snap_len = -1;
- loc_t root_loc = {0,};
- loc_t *temp_loc = NULL;
-
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
-
- priv = this->private;
- SVC_GET_SUBVOL_FROM_CTX (this, op_ret, op_errno, inode_type, ret,
- loc->inode, subvolume, out);
- path_len = strlen (loc->path);
- snap_len = strlen (priv->path);
- temp_loc = loc;
-
- if (path_len >= snap_len && inode_type == VIRTUAL_INODE) {
- path = &loc->path[path_len - snap_len];
- if (!strcmp (path, priv->path)) {
- /*
- * statfs call for virtual snap directory.
- * Sent the fops to parent volume by removing
- * virtual directory from path
- */
- subvolume = FIRST_CHILD (this);
- root_loc.path = gf_strdup("/");
- gf_uuid_clear(root_loc.gfid);
- root_loc.gfid[15] = 1;
- root_loc.inode = inode_ref (loc->inode->table->root);
- temp_loc = &root_loc;
- }
+ int32_t ret = -1;
+ svc_local_t *local = NULL;
+ xlator_t *subvolume = NULL;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ inode_t *parent = NULL;
+ dict_t *new_xdata = NULL;
+ int inode_type = -1;
+ int parent_type = -1;
+ gf_boolean_t wind = _gf_false;
+ char entry_point[NAME_MAX + 1] = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
+
+ ret = svc_inode_ctx_get(this, loc->inode, &inode_type);
+ if (!__is_root_gfid(loc->gfid)) {
+ if (loc->parent) {
+ parent = inode_ref(loc->parent);
+ ret = svc_inode_ctx_get(this, loc->parent, &parent_type);
+ } else {
+ parent = inode_parent(loc->inode, loc->pargfid, NULL);
+ if (parent)
+ ret = svc_inode_ctx_get(this, parent, &parent_type);
+ }
+ }
+
+ local = mem_get0(this->local_pool);
+ if (!local) {
+ op_ret = -1;
+ op_errno = ENOMEM;
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno, SVC_MSG_NO_MEMORY, NULL);
+ goto out;
+ }
+
+ frame->local = local;
+ loc_copy(&local->loc, loc);
+
+ if (__is_root_gfid(loc->inode->gfid)) {
+ subvolume = FIRST_CHILD(this);
+ GF_ASSERT(subvolume);
+ local->subvolume = subvolume;
+ wind = _gf_true;
+ goto out;
+ }
+
+ /* nfs sends nameless lookups directly using the gfid. In that case
+ loc->name will be NULL. So check if loc->name is NULL. If so, then
+ try to get the subvolume using inode context. But if the inode has
+ not been looked up yet, then send the lookup call to the first
+ subvolume.
+ */
+
+ if (!loc->name) {
+ if (gf_uuid_is_null(loc->inode->gfid)) {
+ subvolume = FIRST_CHILD(this);
+ local->subvolume = subvolume;
+ wind = _gf_true;
+ goto out;
+ } else {
+ if (inode_type >= 0)
+ subvolume = svc_get_subvolume(this, inode_type);
+ else
+ subvolume = FIRST_CHILD(this);
+ local->subvolume = subvolume;
+ wind = _gf_true;
+ goto out;
+ }
+ }
+
+ if (gf_svc_get_entry_point(this, entry_point, sizeof(entry_point))) {
+ gf_smsg(this->name, GF_LOG_WARNING, op_errno,
+ SVC_MSG_COPY_ENTRY_POINT_FAILED, NULL);
+ goto out;
+ }
+
+ if (strcmp(loc->name, entry_point)) {
+ if (parent_type == VIRTUAL_INODE) {
+ subvolume = SECOND_CHILD(this);
+ } else {
+ /*
+ * Either parent type is normal graph, or the parent
+ * type is uncertain.
+ */
+ subvolume = FIRST_CHILD(this);
+ }
+ local->subvolume = subvolume;
+ } else {
+ subvolume = SECOND_CHILD(this);
+ local->subvolume = subvolume;
+ if (parent_type == NORMAL_INODE) {
+ /* Indication of whether the lookup is happening on the
+ entry point or not, to the snapview-server.
+ */
+ SVC_ENTRY_POINT_SET(this, xdata, op_ret, op_errno, new_xdata, ret,
+ out);
}
+ }
- STACK_WIND_TAIL (frame, subvolume, subvolume->fops->statfs,
- temp_loc, xdata);
- if (temp_loc == &root_loc)
- loc_wipe (temp_loc);
+ wind = _gf_true;
- wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (statfs, frame, op_ret, op_errno,
- NULL, NULL);
- return 0;
+ if (wind)
+ STACK_WIND(frame, gf_svc_lookup_cbk, subvolume, subvolume->fops->lookup,
+ loc, xdata);
+ else
+ SVC_STACK_UNWIND(lookup, frame, op_ret, op_errno, NULL, NULL, NULL,
+ NULL);
+ if (new_xdata)
+ dict_unref(new_xdata);
+
+ if (parent)
+ inode_unref(parent);
+
+ return 0;
}
static int32_t
-gf_svc_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, struct iatt *buf,
- dict_t *xdata)
+gf_svc_statfs(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- /* Consider a testcase:
- * #mount -t nfs host1:/vol1 /mnt
- * #ls /mnt
- * #ls /mnt/.snaps (As expected this fails)
- * #gluster volume set vol1 features.uss enable
- * Now `ls /mnt/.snaps` should work,
- * but fails with No such file or directory.
- * This is because NFS client caches the list of files in
- * a directory. This cache is updated if there are any changes
- * in the directory attributes. To solve this problem change
- * a attribute 'ctime' when USS is enabled
- */
- if (op_ret == 0 && IA_ISDIR(buf->ia_type))
- buf->ia_ctime_nsec++;
+ xlator_t *subvolume = NULL;
+ int32_t ret = -1;
+ int inode_type = -1;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ gf_boolean_t wind = _gf_false;
+ svc_private_t *priv = NULL;
+ const char *path = NULL;
+ int path_len = -1;
+ int snap_len = -1;
+ loc_t root_loc = {
+ 0,
+ };
+ loc_t *temp_loc = NULL;
+ char entry_point[NAME_MAX + 1] = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
+
+ priv = this->private;
+ SVC_GET_SUBVOL_FROM_CTX(this, op_ret, op_errno, inode_type, ret, loc->inode,
+ subvolume, out);
+ path_len = strlen(loc->path);
+ snap_len = strlen(priv->path);
+ temp_loc = loc;
+
+ if (path_len >= snap_len && inode_type == VIRTUAL_INODE) {
+ path = &loc->path[path_len - snap_len];
+ if (gf_svc_get_entry_point(this, entry_point, sizeof(entry_point))) {
+ gf_smsg(this->name, GF_LOG_WARNING, op_errno,
+ SVC_MSG_COPY_ENTRY_POINT_FAILED, NULL);
+ goto out;
+ }
+
+ if (!strcmp(path, entry_point)) {
+ /*
+ * statfs call for virtual snap directory.
+ * Sent the fops to parent volume by removing
+ * virtual directory from path
+ */
+ subvolume = FIRST_CHILD(this);
+ root_loc.path = gf_strdup("/");
+ gf_uuid_clear(root_loc.gfid);
+ root_loc.gfid[15] = 1;
+ root_loc.inode = inode_ref(loc->inode->table->root);
+ temp_loc = &root_loc;
+ }
+ }
+
+ STACK_WIND_TAIL(frame, subvolume, subvolume->fops->statfs, temp_loc, xdata);
+ if (temp_loc == &root_loc)
+ loc_wipe(temp_loc);
+
+ wind = _gf_true;
+out:
+ if (!wind)
+ SVC_STACK_UNWIND(statfs, frame, op_ret, op_errno, NULL, NULL);
+ return 0;
+}
- SVC_STACK_UNWIND (stat, frame, op_ret, op_errno, buf, xdata);
- return 0;
+static int32_t
+gf_svc_stat_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, struct iatt *buf,
+ dict_t *xdata)
+{
+ /* TODO: FIX ME
+ * Consider a testcase:
+ * #mount -t nfs host1:/vol1 /mnt
+ * #ls /mnt
+ * #ls /mnt/.snaps (As expected this fails)
+ * #gluster volume set vol1 features.uss enable
+ * Now `ls /mnt/.snaps` should work, but fails with No such file or
+ * directory. This is because NFS client (gNFS) caches the list of files
+ * in a directory. This cache is updated if there are any changes in the
+ * directory attributes. So, one way to solve this problem is to change
+ * 'ctime' attribute when USS is enabled as below.
+ *
+ * if (op_ret == 0 && IA_ISDIR(buf->ia_type))
+ * buf->ia_ctime_nsec++;
+ *
+ * But this is not the ideal solution as applications see the unexpected
+ * ctime change causing failures.
+ */
+
+ SVC_STACK_UNWIND(stat, frame, op_ret, op_errno, buf, xdata);
+ return 0;
}
/* should all the fops be handled like lookup is supposed to be
@@ -541,122 +595,119 @@ gf_svc_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
be sent and in the call back update the contexts.
*/
static int32_t
-gf_svc_stat (call_frame_t *frame, xlator_t *this, loc_t *loc,
- dict_t *xdata)
+gf_svc_stat(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
{
- int32_t ret = -1;
- int inode_type = -1;
- xlator_t *subvolume = NULL;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- gf_boolean_t wind = _gf_false;
+ int32_t ret = -1;
+ int inode_type = -1;
+ xlator_t *subvolume = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ gf_boolean_t wind = _gf_false;
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
- SVC_GET_SUBVOL_FROM_CTX (this, op_ret, op_errno, inode_type, ret,
- loc->inode, subvolume, out);
+ SVC_GET_SUBVOL_FROM_CTX(this, op_ret, op_errno, inode_type, ret, loc->inode,
+ subvolume, out);
- STACK_WIND (frame, gf_svc_stat_cbk, subvolume,
- subvolume->fops->stat, loc, xdata);
+ STACK_WIND(frame, gf_svc_stat_cbk, subvolume, subvolume->fops->stat, loc,
+ xdata);
- wind = _gf_true;
+ wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (stat, frame, op_ret, op_errno,
- NULL, NULL);
- return 0;
+ if (!wind)
+ SVC_STACK_UNWIND(stat, frame, op_ret, op_errno, NULL, NULL);
+ return 0;
}
static int32_t
-gf_svc_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+gf_svc_fstat(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
{
- int32_t ret = -1;
- int inode_type = -1;
- xlator_t *subvolume = NULL;
- int32_t op_ret = -1;
- int32_t op_errno = EINVAL;
- gf_boolean_t wind = _gf_false;
+ int32_t ret = -1;
+ int inode_type = -1;
+ xlator_t *subvolume = NULL;
+ int32_t op_ret = -1;
+ int32_t op_errno = EINVAL;
+ gf_boolean_t wind = _gf_false;
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
- GF_VALIDATE_OR_GOTO (this->name, fd->inode, out);
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd->inode, out);
- SVC_GET_SUBVOL_FROM_CTX (this, op_ret, op_errno, inode_type, ret,
- fd->inode, subvolume, out);
+ SVC_GET_SUBVOL_FROM_CTX(this, op_ret, op_errno, inode_type, ret, fd->inode,
+ subvolume, out);
- STACK_WIND_TAIL (frame, subvolume, subvolume->fops->fstat, fd, xdata);
+ STACK_WIND_TAIL(frame, subvolume, subvolume->fops->fstat, fd, xdata);
- wind = _gf_true;
+ wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (fstat, frame, op_ret, op_errno, NULL, NULL);
+ if (!wind)
+ SVC_STACK_UNWIND(fstat, frame, op_ret, op_errno, NULL, NULL);
- return ret;
+ return ret;
}
static int32_t
-gf_svc_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+gf_svc_opendir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
{
- svc_fd_t *svc_fd = NULL;
- svc_local_t *local = NULL;
- svc_private_t *priv = NULL;
- gf_boolean_t special_dir = _gf_false;
- char path[PATH_MAX] = {0, };
-
- GF_VALIDATE_OR_GOTO ("snapview-client", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
-
- if (op_ret)
- goto out;
+ svc_fd_t *svc_fd = NULL;
+ svc_local_t *local = NULL;
+ svc_private_t *priv = NULL;
+ gf_boolean_t special_dir = _gf_false;
+ char path[PATH_MAX] = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("snapview-client", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
+
+ if (op_ret)
+ goto out;
+
+ priv = this->private;
+ local = frame->local;
+
+ if (local->subvolume == FIRST_CHILD(this) && priv->special_dir &&
+ strcmp(priv->special_dir, "")) {
+ if (!__is_root_gfid(fd->inode->gfid))
+ snprintf(path, sizeof(path), "%s/.", priv->special_dir);
+ else
+ snprintf(path, sizeof(path), "/.");
- priv = this->private;
- local = frame->local;
-
- if (local->subvolume == FIRST_CHILD (this) && priv->special_dir
- && strcmp (priv->special_dir, "")) {
- if (!__is_root_gfid (fd->inode->gfid))
- snprintf (path, sizeof (path), "%s/.",
- priv->special_dir);
- else
- snprintf (path, sizeof (path), "/.");
-
- if (!strcmp (local->loc.path, priv->special_dir) ||
- !strcmp (local->loc.path, path)) {
- gf_log_callingfn (this->name, GF_LOG_DEBUG,
- "got opendir on special "
- "directory %s (%s)", path,
- uuid_utoa (fd->inode->gfid));
- special_dir = _gf_true;
- }
+ if (!strcmp(local->loc.path, priv->special_dir) ||
+ !strcmp(local->loc.path, path)) {
+ gf_msg_debug(this->name, 0,
+ "got opendir on special directory"
+ " %s (gfid: %s)",
+ path, uuid_utoa(fd->inode->gfid));
+ special_dir = _gf_true;
}
+ }
- if (special_dir) {
- svc_fd = svc_fd_ctx_get_or_new (this, fd);
- if (!svc_fd) {
- gf_log (this->name, GF_LOG_ERROR,
- "fd context not found for %s",
- uuid_utoa (fd->inode->gfid));
- goto out;
- }
-
- svc_fd->last_offset = -1;
- svc_fd->special_dir = special_dir;
+ if (special_dir) {
+ svc_fd = svc_fd_ctx_get_or_new(this, fd);
+ if (!svc_fd) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, SVC_MSG_GET_FD_CONTEXT_FAILED,
+ "gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto out;
}
+ svc_fd->last_offset = -1;
+ svc_fd->special_dir = special_dir;
+ }
+
out:
- STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd, xdata);
+ STACK_UNWIND_STRICT(opendir, frame, op_ret, op_errno, fd, xdata);
- return 0;
+ return 0;
}
-
/* If the inode represents a directory which is actually
present in a snapshot, then opendir on that directory
should be sent to the snap-view-server which opens
@@ -668,93 +719,91 @@ out:
svc has to do things that open-behind is doing.
*/
static int32_t
-gf_svc_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
- dict_t *xdata)
+gf_svc_opendir(call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+ dict_t *xdata)
{
- int32_t ret = -1;
- int inode_type = -1;
- xlator_t *subvolume = NULL;
- int op_ret = -1;
- int op_errno = EINVAL;
- gf_boolean_t wind = _gf_false;
- svc_local_t *local = NULL;
-
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
-
- local = mem_get0 (this->local_pool);
- if (!local) {
- gf_log (this->name, GF_LOG_ERROR, "failed to allocate memory "
- "for local (path: %s, gfid: %s)", loc->path,
- uuid_utoa (fd->inode->gfid));
- op_errno = ENOMEM;
- goto out;
- }
-
- SVC_GET_SUBVOL_FROM_CTX (this, op_ret, op_errno, inode_type, ret,
- loc->inode, subvolume, out);
-
- loc_copy (&local->loc, loc);
- local->subvolume = subvolume;
- frame->local = local;
-
- STACK_WIND (frame, gf_svc_opendir_cbk, subvolume,
- subvolume->fops->opendir, loc, fd, xdata);
-
- wind = _gf_true;
+ int32_t ret = -1;
+ int inode_type = -1;
+ xlator_t *subvolume = NULL;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ gf_boolean_t wind = _gf_false;
+ svc_local_t *local = NULL;
+
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
+
+ local = mem_get0(this->local_pool);
+ if (!local) {
+ op_errno = ENOMEM;
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno, SVC_MSG_NO_MEMORY,
+ "path=%s", loc->path, "gfid=%s", uuid_utoa(fd->inode->gfid),
+ NULL);
+ goto out;
+ }
+ loc_copy(&local->loc, loc);
+ frame->local = local;
+
+ SVC_GET_SUBVOL_FROM_CTX(this, op_ret, op_errno, inode_type, ret, loc->inode,
+ subvolume, out);
+ local->subvolume = subvolume;
+
+ STACK_WIND(frame, gf_svc_opendir_cbk, subvolume, subvolume->fops->opendir,
+ loc, fd, xdata);
+
+ wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (opendir, frame, op_ret, op_errno, NULL, NULL);
+ if (!wind)
+ SVC_STACK_UNWIND(opendir, frame, op_ret, op_errno, NULL, NULL);
- return 0;
+ return 0;
}
static int32_t
-gf_svc_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- struct iatt *stbuf, int32_t valid, dict_t *xdata)
+gf_svc_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ struct iatt *stbuf, int32_t valid, dict_t *xdata)
{
- int32_t ret = -1;
- int inode_type = -1;
- int op_ret = -1;
- int op_errno = EINVAL;
- gf_boolean_t wind = _gf_false;
-
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
-
- ret = svc_inode_ctx_get (this, loc->inode, &inode_type);
- if (ret < 0) {
- op_ret = -1;
- op_errno = EINVAL;
- gf_log (this->name, GF_LOG_ERROR, "failed to get the inode "
- "context for %s (gfid: %s)", loc->path,
- uuid_utoa (loc->inode->gfid));
- goto out;
- }
-
- if (inode_type == NORMAL_INODE) {
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setattr, loc, stbuf,
- valid, xdata);
- } else {
- op_ret = -1;
- op_errno = EROFS;
- goto out;
- }
-
- wind = _gf_true;
+ int32_t ret = -1;
+ int inode_type = -1;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ gf_boolean_t wind = _gf_false;
+
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
+
+ ret = svc_inode_ctx_get(this, loc->inode, &inode_type);
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno,
+ SVC_MSG_GET_INODE_CONTEXT_FAILED, "path=%s", loc->path,
+ "gfid= %s", uuid_utoa(loc->inode->gfid), NULL);
+ goto out;
+ }
+
+ if (inode_type == NORMAL_INODE) {
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setattr, loc, stbuf, valid,
+ xdata);
+ } else {
+ op_ret = -1;
+ op_errno = EROFS;
+ goto out;
+ }
+
+ wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (setattr, frame, op_ret, op_errno,
- NULL, NULL, NULL);
- return 0;
+ if (!wind)
+ SVC_STACK_UNWIND(setattr, frame, op_ret, op_errno, NULL, NULL, NULL);
+ return 0;
}
/* XXX: This function is currently not used. Remove "#if 0" when required */
@@ -776,10 +825,12 @@ gf_svc_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
ret = svc_inode_ctx_get (this, fd->inode, &inode_type);
if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "failed to get the inode "
- "context for %s", uuid_utoa (fd->inode->gfid));
op_ret = -1;
op_errno = EINVAL;
+ gf_msg (this->name, GF_LOG_ERROR, op_errno,
+ SVC_MSG_GET_INODE_CONTEXT_FAILED, "failed to "
+ "get the inode context for %s",
+ uuid_utoa (fd->inode->gfid));
goto out;
}
@@ -804,82 +855,88 @@ out:
#endif /* gf_svc_fsetattr() is not used */
static int32_t
-gf_svc_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
- const char *name, dict_t *xdata)
+gf_svc_getxattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
+ const char *name, dict_t *xdata)
{
- int32_t ret = -1;
- int inode_type = -1;
- xlator_t *subvolume = NULL;
- int op_ret = -1;
- int op_errno = EINVAL;
- gf_boolean_t wind = _gf_false;
- svc_private_t *priv = NULL;
- char attrname[PATH_MAX] = "";
- char attrval[64] = "";
- dict_t *dict = NULL;
+ int32_t ret = -1;
+ int inode_type = -1;
+ xlator_t *subvolume = NULL;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ gf_boolean_t wind = _gf_false;
+ svc_private_t *priv = NULL;
+ char attrname[PATH_MAX] = "";
+ char attrval[64] = "";
+ dict_t *dict = NULL;
+ char entry_point[NAME_MAX + 1] = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
+ priv = this->private;
+ GF_VALIDATE_OR_GOTO(this->name, priv, out);
+
+ /*
+ * Samba sends this special key for case insensitive
+ * filename check. This request comes with a parent
+ * path and with a special key GF_XATTR_GET_REAL_FILENAME_KEY.
+ * e.g. "glusterfs.get_real_filename:.snaps".
+ * If the name variable matches this key then we have
+ * to send back .snaps as the real filename.
+ */
+ if (!name)
+ goto stack_wind;
+
+ sscanf(name, "%[^:]:%[^@]", attrname, attrval);
+ strcat(attrname, ":");
+
+ if (!strcmp(attrname, GF_XATTR_GET_REAL_FILENAME_KEY)) {
+ if (gf_svc_get_entry_point(this, entry_point, sizeof(entry_point))) {
+ gf_smsg(this->name, GF_LOG_WARNING, op_errno,
+ SVC_MSG_COPY_ENTRY_POINT_FAILED, NULL);
+ goto out;
+ }
+
+ if (!strcasecmp(attrval, entry_point)) {
+ dict = dict_new();
+ if (NULL == dict) {
+ op_errno = ENOMEM;
+ goto out;
+ }
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
- priv = this->private;
- GF_VALIDATE_OR_GOTO (this->name, priv, out);
-
- /*
- * Samba sends this special key for case insensitive
- * filename check. This request comes with a parent
- * path and with a special key GF_XATTR_GET_REAL_FILENAME_KEY.
- * e.g. "glusterfs.get_real_filename:.snaps".
- * If the name variable matches this key then we have
- * to send back .snaps as the real filename.
- */
- if (!name)
- goto stack_wind;
-
- sscanf (name, "%[^:]:%[^@]", attrname, attrval);
- strcat (attrname, ":");
-
- if (!strcmp (attrname, GF_XATTR_GET_REAL_FILENAME_KEY)) {
- if (!strcasecmp (attrval, priv->path)) {
- dict = dict_new ();
- if (NULL == dict) {
- op_errno = ENOMEM;
- goto out;
- }
-
- ret = dict_set_dynstr_with_alloc (dict,
- (char *)name,
- priv->path);
-
- if (ret) {
- op_errno = ENOMEM;
- goto out;
- }
-
- op_errno = 0;
- op_ret = strlen (priv->path) + 1;
- /* We should return from here */
- goto out;
- }
+ ret = dict_set_dynstr_with_alloc(dict, (char *)name, entry_point);
+
+ if (ret) {
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ op_errno = 0;
+ op_ret = strlen(entry_point) + 1;
+ /* We should return from here */
+ goto out;
}
+ }
stack_wind:
- SVC_GET_SUBVOL_FROM_CTX (this, op_ret, op_errno, inode_type, ret,
- loc->inode, subvolume, out);
+ SVC_GET_SUBVOL_FROM_CTX(this, op_ret, op_errno, inode_type, ret, loc->inode,
+ subvolume, out);
- STACK_WIND_TAIL (frame, subvolume, subvolume->fops->getxattr, loc, name,
- xdata);
+ STACK_WIND_TAIL(frame, subvolume, subvolume->fops->getxattr, loc, name,
+ xdata);
- wind = _gf_true;
+ wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (getxattr, frame, op_ret, op_errno,
- dict, NULL);
+ if (!wind)
+ SVC_STACK_UNWIND(getxattr, frame, op_ret, op_errno, dict, NULL);
- if (dict)
- dict_unref (dict);
+ if (dict)
+ dict_unref(dict);
- return 0;
+ return 0;
}
/* XXX: This function is currently not used. Mark it '#if 0' when required */
@@ -917,279 +974,286 @@ out:
#endif /* gf_svc_fgetxattr() is not used */
static int32_t
-gf_svc_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
- int32_t flags, dict_t *xdata)
+gf_svc_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- int32_t ret = -1;
- int inode_type = -1;
- int op_ret = -1;
- int op_errno = EINVAL;
- gf_boolean_t wind = _gf_false;
-
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
-
- ret = svc_inode_ctx_get (this, loc->inode, &inode_type);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "failed to get inode context "
- "for %s (gfid: %s)", loc->name,
- uuid_utoa (loc->inode->gfid));
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
-
- if (inode_type == NORMAL_INODE) {
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->setxattr, loc, dict,
- flags, xdata);
- } else {
- op_ret = -1;
- op_errno = EROFS;
- goto out;
- }
-
- wind = _gf_true;
+ int32_t ret = -1;
+ int inode_type = -1;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ gf_boolean_t wind = _gf_false;
+
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
+
+ ret = svc_inode_ctx_get(this, loc->inode, &inode_type);
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno,
+ SVC_MSG_GET_INODE_CONTEXT_FAILED, "name=%s", loc->name,
+ "gfid=%s", uuid_utoa(loc->inode->gfid), NULL);
+ goto out;
+ }
+
+ if (inode_type == NORMAL_INODE) {
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->setxattr, loc, dict, flags,
+ xdata);
+ } else {
+ op_ret = -1;
+ op_errno = EROFS;
+ goto out;
+ }
+
+ wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (setxattr, frame, op_ret, op_errno,
- NULL);
+ if (!wind)
+ SVC_STACK_UNWIND(setxattr, frame, op_ret, op_errno, NULL);
- return 0;
+ return 0;
}
static int32_t
-gf_svc_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
- int32_t flags, dict_t *xdata)
+gf_svc_fsetxattr(call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+ int32_t flags, dict_t *xdata)
{
- int32_t ret = -1;
- int inode_type = -1;
- int op_ret = -1;
- int op_errno = EINVAL;
- gf_boolean_t wind = _gf_false;
-
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
- GF_VALIDATE_OR_GOTO (this->name, fd->inode, out);
-
- ret = svc_inode_ctx_get (this, fd->inode, &inode_type);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "failed to get inode context "
- "for %s", uuid_utoa (fd->inode->gfid));
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
-
- if (inode_type == NORMAL_INODE) {
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->fsetxattr, fd, dict,
- flags, xdata);
- } else {
- op_ret = -1;
- op_errno = EROFS;
- goto out;
- }
-
- wind = _gf_true;
+ int32_t ret = -1;
+ int inode_type = -1;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ gf_boolean_t wind = _gf_false;
+
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd->inode, out);
+
+ ret = svc_inode_ctx_get(this, fd->inode, &inode_type);
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno,
+ SVC_MSG_GET_INODE_CONTEXT_FAILED, "gfid=%s",
+ uuid_utoa(fd->inode->gfid), NULL);
+ goto out;
+ }
+
+ if (inode_type == NORMAL_INODE) {
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags,
+ xdata);
+ } else {
+ op_ret = -1;
+ op_errno = EROFS;
+ goto out;
+ }
+
+ wind = _gf_true;
out:
- if (!wind)
- STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno,
- NULL);
+ if (!wind)
+ STACK_UNWIND_STRICT(fsetxattr, frame, op_ret, op_errno, NULL);
- return 0;
+ return 0;
}
static int32_t
-gf_svc_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
- dict_t *xdata)
+gf_svc_rmdir(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
{
- int inode_type = -1;
- int ret = -1;
- int op_ret = -1;
- int op_errno = EINVAL;
- gf_boolean_t wind = _gf_false;
-
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
-
- ret = svc_inode_ctx_get (this, loc->inode, &inode_type);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "failed to get the inode "
- "context for %s (gfid: %s)", loc->name,
- uuid_utoa (loc->inode->gfid));
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
-
- if (inode_type == NORMAL_INODE) {
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->rmdir, loc, flags,
- xdata);
- } else {
- op_ret = -1;
- op_errno = EROFS;
- goto out;
- }
-
- wind = _gf_true;
+ int inode_type = -1;
+ int ret = -1;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ gf_boolean_t wind = _gf_false;
+
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
+
+ ret = svc_inode_ctx_get(this, loc->inode, &inode_type);
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno,
+ SVC_MSG_GET_INODE_CONTEXT_FAILED, "name=%s", loc->name,
+ "gfid=%s", uuid_utoa(loc->inode->gfid), NULL);
+ goto out;
+ }
+
+ if (inode_type == NORMAL_INODE) {
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->rmdir, loc, flags, xdata);
+ } else {
+ op_ret = -1;
+ op_errno = EROFS;
+ goto out;
+ }
+
+ wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (rmdir, frame, op_ret, op_errno,
- NULL, NULL, NULL);
- return 0;
+ if (!wind)
+ SVC_STACK_UNWIND(rmdir, frame, op_ret, op_errno, NULL, NULL, NULL);
+ return 0;
}
static int32_t
-gf_svc_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+gf_svc_mkdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- int inode_type = -1;
- int ret = -1;
+ int inode_type = -1;
+ int ret = -1;
- if (op_ret < 0)
- goto out;
-
- inode_type = NORMAL_INODE;
- ret = svc_inode_ctx_set (this, inode, inode_type);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR, "failed to set inode "
- "context");
+ if (op_ret < 0)
+ goto out;
+ inode_type = NORMAL_INODE;
+ ret = svc_inode_ctx_set(this, inode, inode_type);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, 0, SVC_MSG_SET_INODE_CONTEXT_FAILED,
+ NULL);
out:
- SVC_STACK_UNWIND (mkdir, frame, op_ret, op_errno, inode,
- buf, preparent, postparent, xdata);
- return 0;
+ SVC_STACK_UNWIND(mkdir, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
+ return 0;
}
static int32_t
-gf_svc_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- mode_t umask, dict_t *xdata)
+gf_svc_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ mode_t umask, dict_t *xdata)
{
- int parent_type = -1;
- int ret = -1;
- int op_ret = -1;
- int op_errno = EINVAL;
- svc_private_t *priv = NULL;
- gf_boolean_t wind = _gf_false;
-
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
-
- priv = this->private;
-
- ret = svc_inode_ctx_get (this, loc->parent, &parent_type);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "failed to get the inode "
- "context for %s", uuid_utoa (loc->parent->gfid));
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
-
- if (strcmp (loc->name, priv->path) && parent_type == NORMAL_INODE) {
- STACK_WIND (frame, gf_svc_mkdir_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->mkdir, loc, mode,
- umask, xdata);
- } else {
- op_ret = -1;
- op_errno = EROFS;
- goto out;
- }
-
- wind = _gf_true;
+ int parent_type = -1;
+ int ret = -1;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ gf_boolean_t wind = _gf_false;
+ char entry_point[NAME_MAX + 1] = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
+
+ ret = svc_inode_ctx_get(this, loc->parent, &parent_type);
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno,
+ SVC_MSG_GET_INODE_CONTEXT_FAILED, "gfid=%s",
+ uuid_utoa(loc->parent->gfid), NULL);
+ goto out;
+ }
+
+ if (gf_svc_get_entry_point(this, entry_point, sizeof(entry_point))) {
+ gf_smsg(this->name, GF_LOG_WARNING, op_errno,
+ SVC_MSG_COPY_ENTRY_POINT_FAILED, NULL);
+ goto out;
+ }
+
+ if (strcmp(loc->name, entry_point) && parent_type == NORMAL_INODE) {
+ STACK_WIND(frame, gf_svc_mkdir_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
+ } else {
+ op_ret = -1;
+ op_errno = EROFS;
+ goto out;
+ }
+
+ wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (mkdir, frame, op_ret, op_errno, NULL, NULL,
- NULL, NULL, NULL);
- return 0;
+ if (!wind)
+ SVC_STACK_UNWIND(mkdir, frame, op_ret, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
static int32_t
-gf_svc_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+gf_svc_mknod_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- int inode_type = -1;
- int ret = -1;
+ int inode_type = -1;
+ int ret = -1;
- if (op_ret < 0)
- goto out;
+ if (op_ret < 0)
+ goto out;
- inode_type = NORMAL_INODE;
- ret = svc_inode_ctx_set (this, inode, inode_type);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR, "failed to set inode "
- "context");
+ inode_type = NORMAL_INODE;
+ ret = svc_inode_ctx_set(this, inode, inode_type);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, 0, SVC_MSG_SET_INODE_CONTEXT_FAILED,
+ NULL);
out:
- SVC_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode,
- buf, preparent, postparent, xdata);
- return 0;
+ SVC_STACK_UNWIND(mknod, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
+ return 0;
}
static int32_t
-gf_svc_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
- dev_t rdev, mode_t umask, dict_t *xdata)
+gf_svc_mknod(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+ dev_t rdev, mode_t umask, dict_t *xdata)
{
- int parent_type = -1;
- int ret = -1;
- int op_ret = -1;
- int op_errno = EINVAL;
- svc_private_t *priv = NULL;
- gf_boolean_t wind = _gf_false;
-
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
-
- priv = this->private;
-
- ret = svc_inode_ctx_get (this, loc->parent, &parent_type);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "failed to get the inode "
- "context for %s", uuid_utoa (loc->parent->gfid));
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
-
- if (strcmp (loc->name, priv->path) && parent_type == NORMAL_INODE) {
- STACK_WIND (frame, gf_svc_mknod_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->mknod, loc, mode,
- rdev, umask, xdata);
- } else {
- op_ret = -1;
- op_errno = EROFS;
- goto out;
- }
-
- wind = _gf_true;
+ int parent_type = -1;
+ int ret = -1;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ gf_boolean_t wind = _gf_false;
+ char entry_point[NAME_MAX + 1] = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
+
+ ret = svc_inode_ctx_get(this, loc->parent, &parent_type);
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno,
+ SVC_MSG_GET_INODE_CONTEXT_FAILED, "gfid=%s",
+ uuid_utoa(loc->parent->gfid), NULL);
+ goto out;
+ }
+
+ if (gf_svc_get_entry_point(this, entry_point, sizeof(entry_point))) {
+ gf_smsg(this->name, GF_LOG_WARNING, op_errno,
+ SVC_MSG_COPY_ENTRY_POINT_FAILED, NULL);
+ goto out;
+ }
+
+ if (strcmp(loc->name, entry_point) && parent_type == NORMAL_INODE) {
+ STACK_WIND(frame, gf_svc_mknod_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->mknod, loc, mode, rdev, umask,
+ xdata);
+ } else {
+ op_ret = -1;
+ op_errno = EROFS;
+ goto out;
+ }
+
+ wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (mknod, frame, op_ret, op_errno, NULL, NULL,
- NULL, NULL, NULL);
- return 0;
+ if (!wind)
+ SVC_STACK_UNWIND(mknod, frame, op_ret, op_errno, NULL, NULL, NULL, NULL,
+ NULL);
+ return 0;
}
/* If the flags of the open call contain O_WRONLY or O_RDWR and the inode is
@@ -1197,425 +1261,447 @@ out:
STACK_WIND the call to the first child of svc xlator.
*/
static int32_t
-gf_svc_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- fd_t *fd, dict_t *xdata)
+gf_svc_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ fd_t *fd, dict_t *xdata)
{
- xlator_t *subvolume = NULL;
- int inode_type = -1;
- int op_ret = -1;
- int op_errno = EINVAL;
- int ret = -1;
- gf_boolean_t wind = _gf_false;
-
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
-
- /* Another way is to STACK_WIND to normal subvolume, if inode
- type is not there in the context. If the file actually resides
- in snapshots, then ENOENT would be returned. Needs more analysis.
- */
- SVC_GET_SUBVOL_FROM_CTX (this, op_ret, op_errno, inode_type, ret,
- loc->inode, subvolume, out);
-
- if (((flags & O_ACCMODE) == O_WRONLY) ||
- ((flags & O_ACCMODE) == O_RDWR)) {
- if (subvolume != FIRST_CHILD (this)) {
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
- }
-
- STACK_WIND_TAIL (frame, subvolume, subvolume->fops->open, loc,
- flags, fd, xdata);
-
- wind = _gf_true;
+ xlator_t *subvolume = NULL;
+ int inode_type = -1;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ int ret = -1;
+ gf_boolean_t wind = _gf_false;
+
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
+
+ /* Another way is to STACK_WIND to normal subvolume, if inode
+ type is not there in the context. If the file actually resides
+ in snapshots, then ENOENT would be returned. Needs more analysis.
+ */
+ SVC_GET_SUBVOL_FROM_CTX(this, op_ret, op_errno, inode_type, ret, loc->inode,
+ subvolume, out);
+
+ if (((flags & O_ACCMODE) == O_WRONLY) || ((flags & O_ACCMODE) == O_RDWR)) {
+ if (subvolume != FIRST_CHILD(this)) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ goto out;
+ }
+ }
+
+ STACK_WIND_TAIL(frame, subvolume, subvolume->fops->open, loc, flags, fd,
+ xdata);
+
+ wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (open, frame, op_ret, op_errno, NULL,
- NULL);
- return 0;
+ if (!wind)
+ SVC_STACK_UNWIND(open, frame, op_ret, op_errno, NULL, NULL);
+ return 0;
}
static int32_t
-gf_svc_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
- struct iatt *stbuf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+gf_svc_create_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+ struct iatt *stbuf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- int inode_type = -1;
- int ret = -1;
+ int inode_type = -1;
+ int ret = -1;
- if (op_ret < 0)
- goto out;
+ if (op_ret < 0)
+ goto out;
- inode_type = NORMAL_INODE;
- ret = svc_inode_ctx_set (this, inode, inode_type);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR, "failed to set inode "
- "context");
+ inode_type = NORMAL_INODE;
+ ret = svc_inode_ctx_set(this, inode, inode_type);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, 0, SVC_MSG_SET_INODE_CONTEXT_FAILED,
+ NULL);
out:
- SVC_STACK_UNWIND (create, frame, op_ret, op_errno, fd,
- inode, stbuf, preparent, postparent, xdata);
+ SVC_STACK_UNWIND(create, frame, op_ret, op_errno, fd, inode, stbuf,
+ preparent, postparent, xdata);
- return 0;
+ return 0;
}
static int32_t
-gf_svc_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
- mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+gf_svc_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+ mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
{
- int parent_type = -1;
- int ret = -1;
- int op_ret = -1;
- int op_errno = EINVAL;
- svc_private_t *priv = NULL;
- gf_boolean_t wind = _gf_false;
-
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
-
- priv = this->private;
-
- ret = svc_inode_ctx_get (this, loc->parent, &parent_type);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "failed to get the inode "
- "context for %s", uuid_utoa (loc->parent->gfid));
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
-
- if (strcmp (loc->name, priv->path) && parent_type == NORMAL_INODE) {
- STACK_WIND (frame, gf_svc_create_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->create, loc, flags,
- mode, umask, fd, xdata);
- } else {
- op_ret = -1;
- op_errno = EROFS;
- goto out;
- }
-
- wind = _gf_true;
+ int parent_type = -1;
+ int ret = -1;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ gf_boolean_t wind = _gf_false;
+ char entry_point[NAME_MAX + 1] = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
+
+ ret = svc_inode_ctx_get(this, loc->parent, &parent_type);
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno,
+ SVC_MSG_GET_INODE_CONTEXT_FAILED, "gfid=%s",
+ uuid_utoa(loc->parent->gfid), NULL);
+ goto out;
+ }
+
+ if (gf_svc_get_entry_point(this, entry_point, sizeof(entry_point))) {
+ gf_smsg(this->name, GF_LOG_WARNING, op_errno,
+ SVC_MSG_COPY_ENTRY_POINT_FAILED, NULL);
+ goto out;
+ }
+
+ if (strcmp(loc->name, entry_point) && parent_type == NORMAL_INODE) {
+ STACK_WIND(frame, gf_svc_create_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->create, loc, flags, mode, umask, fd,
+ xdata);
+ } else {
+ op_ret = -1;
+ op_errno = EROFS;
+ goto out;
+ }
+
+ wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (create, frame, op_ret, op_errno,
- NULL, NULL, NULL, NULL, NULL, NULL);
- return 0;
+ if (!wind)
+ SVC_STACK_UNWIND(create, frame, op_ret, op_errno, NULL, NULL, NULL,
+ NULL, NULL, NULL);
+ return 0;
}
static int32_t
-gf_svc_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, struct iatt *preparent,
- struct iatt *postparent, dict_t *xdata)
+gf_svc_symlink_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, struct iatt *preparent,
+ struct iatt *postparent, dict_t *xdata)
{
- int inode_type = -1;
- int ret = -1;
+ int inode_type = -1;
+ int ret = -1;
- if (op_ret < 0)
- goto out;
+ if (op_ret < 0)
+ goto out;
- inode_type = NORMAL_INODE;
- ret = svc_inode_ctx_set (this, inode, inode_type);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR, "failed to set inode "
- "context");
+ inode_type = NORMAL_INODE;
+ ret = svc_inode_ctx_set(this, inode, inode_type);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, 0, SVC_MSG_SET_INODE_CONTEXT_FAILED,
+ NULL);
out:
- SVC_STACK_UNWIND (symlink, frame, op_ret, op_errno, inode,
- buf, preparent, postparent, xdata);
+ SVC_STACK_UNWIND(symlink, frame, op_ret, op_errno, inode, buf, preparent,
+ postparent, xdata);
- return 0;
+ return 0;
}
static int32_t
-gf_svc_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
- loc_t *loc, mode_t umask, dict_t *xdata)
+gf_svc_symlink(call_frame_t *frame, xlator_t *this, const char *linkpath,
+ loc_t *loc, mode_t umask, dict_t *xdata)
{
- int parent_type = -1;
- int op_ret = -1;
- int op_errno = EINVAL;
- int ret = -1;
- svc_private_t *priv = NULL;
- gf_boolean_t wind = _gf_false;
-
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
-
- priv = this->private;
-
- ret = svc_inode_ctx_get (this, loc->parent, &parent_type);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "failed to get the inode "
- "context for %s", uuid_utoa (loc->parent->gfid));
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
-
- if (strcmp (loc->name, priv->path) && parent_type == NORMAL_INODE) {
- STACK_WIND (frame, gf_svc_symlink_cbk, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->symlink, linkpath, loc,
- umask, xdata);
- } else {
- op_ret = -1;
- op_errno = EROFS;
- goto out;
- }
-
- wind = _gf_true;
+ int parent_type = -1;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ int ret = -1;
+ gf_boolean_t wind = _gf_false;
+ char entry_point[NAME_MAX + 1] = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
+
+ ret = svc_inode_ctx_get(this, loc->parent, &parent_type);
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno,
+ SVC_MSG_GET_INODE_CONTEXT_FAILED, "gfid=%s",
+ uuid_utoa(loc->parent->gfid), NULL);
+ goto out;
+ }
+
+ if (gf_svc_get_entry_point(this, entry_point, sizeof(entry_point))) {
+ gf_smsg(this->name, GF_LOG_WARNING, op_errno,
+ SVC_MSG_COPY_ENTRY_POINT_FAILED, NULL);
+ goto out;
+ }
+
+ if (strcmp(loc->name, entry_point) && parent_type == NORMAL_INODE) {
+ STACK_WIND(frame, gf_svc_symlink_cbk, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->symlink, linkpath, loc, umask,
+ xdata);
+ } else {
+ op_ret = -1;
+ op_errno = EROFS;
+ goto out;
+ }
+
+ wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (symlink, frame, op_ret, op_errno,
- NULL, NULL, NULL, NULL, NULL);
- return 0;
+ if (!wind)
+ SVC_STACK_UNWIND(symlink, frame, op_ret, op_errno, NULL, NULL, NULL,
+ NULL, NULL);
+ return 0;
}
static int32_t
-gf_svc_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
- dict_t *xdata)
+gf_svc_unlink(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+ dict_t *xdata)
{
- int inode_type = -1;
- int op_ret = -1;
- int op_errno = EINVAL;
- int ret = -1;
- gf_boolean_t wind = _gf_false;
-
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
-
- ret = svc_inode_ctx_get (this, loc->inode, &inode_type);
- if (ret < 0) {
- gf_log (this->name, GF_LOG_ERROR, "failed to get the inode "
- "context for %s", uuid_utoa (loc->parent->gfid));
- op_ret = -1;
- op_errno = EINVAL;
- goto out;
- }
-
- if (inode_type == NORMAL_INODE) {
- STACK_WIND_TAIL (frame, FIRST_CHILD (this),
- FIRST_CHILD (this)->fops->unlink, loc, flags,
- xdata);
- } else {
- op_ret = -1;
- op_errno = EROFS;
- goto out;
- }
-
- wind = _gf_true;
+ int inode_type = -1;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ int ret = -1;
+ gf_boolean_t wind = _gf_false;
+
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
+
+ ret = svc_inode_ctx_get(this, loc->inode, &inode_type);
+ if (ret < 0) {
+ op_ret = -1;
+ op_errno = EINVAL;
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno,
+ SVC_MSG_GET_INODE_CONTEXT_FAILED, "gfid=%s",
+ uuid_utoa(loc->parent->gfid), NULL);
+ goto out;
+ }
+
+ if (inode_type == NORMAL_INODE) {
+ STACK_WIND_TAIL(frame, FIRST_CHILD(this),
+ FIRST_CHILD(this)->fops->unlink, loc, flags, xdata);
+ } else {
+ op_ret = -1;
+ op_errno = EROFS;
+ goto out;
+ }
+
+ wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (unlink, frame, op_ret, op_errno, NULL, NULL,
- NULL);
- return 0;
+ if (!wind)
+ SVC_STACK_UNWIND(unlink, frame, op_ret, op_errno, NULL, NULL, NULL);
+ return 0;
}
static int32_t
-gf_svc_readv (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t offset, uint32_t flags, dict_t *xdata)
+gf_svc_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t offset, uint32_t flags, dict_t *xdata)
{
- int inode_type = -1;
- xlator_t *subvolume = NULL;
- int ret = -1;
- int op_ret = -1;
- int op_errno = EINVAL;
- gf_boolean_t wind = _gf_false;
+ int inode_type = -1;
+ xlator_t *subvolume = NULL;
+ int ret = -1;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ gf_boolean_t wind = _gf_false;
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
- GF_VALIDATE_OR_GOTO (this->name, fd->inode, out);
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd->inode, out);
- SVC_GET_SUBVOL_FROM_CTX (this, op_ret, op_errno, inode_type, ret,
- fd->inode, subvolume, out);
+ SVC_GET_SUBVOL_FROM_CTX(this, op_ret, op_errno, inode_type, ret, fd->inode,
+ subvolume, out);
- STACK_WIND_TAIL (frame, subvolume, subvolume->fops->readv,
- fd, size, offset, flags, xdata);
+ STACK_WIND_TAIL(frame, subvolume, subvolume->fops->readv, fd, size, offset,
+ flags, xdata);
- wind = _gf_true;
+ wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (readv, frame, op_ret, op_errno, NULL, 0, NULL,
- NULL, NULL);
- return 0;
+ if (!wind)
+ SVC_STACK_UNWIND(readv, frame, op_ret, op_errno, NULL, 0, NULL, NULL,
+ NULL);
+ return 0;
}
static int32_t
-gf_svc_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
- dict_t *xdata)
+gf_svc_readlink(call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+ dict_t *xdata)
{
- int inode_type = -1;
- xlator_t *subvolume = NULL;
- int ret = -1;
- int op_ret = -1;
- int op_errno = EINVAL;
- gf_boolean_t wind = _gf_false;
+ int inode_type = -1;
+ xlator_t *subvolume = NULL;
+ int ret = -1;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ gf_boolean_t wind = _gf_false;
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
- SVC_GET_SUBVOL_FROM_CTX (this, op_ret, op_errno, inode_type, ret,
- loc->inode, subvolume, out);
+ SVC_GET_SUBVOL_FROM_CTX(this, op_ret, op_errno, inode_type, ret, loc->inode,
+ subvolume, out);
- STACK_WIND_TAIL (frame, subvolume, subvolume->fops->readlink, loc, size,
- xdata);
+ STACK_WIND_TAIL(frame, subvolume, subvolume->fops->readlink, loc, size,
+ xdata);
- wind = _gf_true;
+ wind = _gf_true;
out:
- if (!wind)
- STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, NULL, NULL,
- NULL);
- return 0;
+ if (!wind)
+ STACK_UNWIND_STRICT(readlink, frame, op_ret, op_errno, NULL, NULL,
+ NULL);
+ return 0;
}
static int32_t
-gf_svc_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
- dict_t *xdata)
+gf_svc_access(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+ dict_t *xdata)
{
- int ret = -1;
- int inode_type = -1;
- xlator_t *subvolume = NULL;
- int op_ret = -1;
- int op_errno = EINVAL;
- gf_boolean_t wind = _gf_false;
+ int ret = -1;
+ int inode_type = -1;
+ xlator_t *subvolume = NULL;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ gf_boolean_t wind = _gf_false;
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, loc, out);
- GF_VALIDATE_OR_GOTO (this->name, loc->inode, out);
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc, out);
+ GF_VALIDATE_OR_GOTO(this->name, loc->inode, out);
- SVC_GET_SUBVOL_FROM_CTX (this, op_ret, op_errno, inode_type, ret,
- loc->inode, subvolume, out);
+ SVC_GET_SUBVOL_FROM_CTX(this, op_ret, op_errno, inode_type, ret, loc->inode,
+ subvolume, out);
- STACK_WIND_TAIL (frame, subvolume, subvolume->fops->access, loc, mask,
- xdata);
+ STACK_WIND_TAIL(frame, subvolume, subvolume->fops->access, loc, mask,
+ xdata);
- wind = _gf_true;
+ wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (access, frame, op_ret, op_errno, NULL);
+ if (!wind)
+ SVC_STACK_UNWIND(access, frame, op_ret, op_errno, NULL);
- return 0;
+ return 0;
}
int32_t
-gf_svc_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
- dict_t *xdata)
+gf_svc_readdir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+ dict_t *xdata)
{
- gf_dirent_t *entry = NULL;
- gf_dirent_t *tmpentry = NULL;
- svc_local_t *local = NULL;
- svc_private_t *priv = NULL;
-
- if (op_ret < 0)
- goto out;
-
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
-
- priv = this->private;
- local = frame->local;
-
- /* If .snaps pre-exists, then it should not be listed
- * in the NORMAL INODE directory when USS is enabled,
- * so filter the .snaps entry if exists.
- * However it is OK to list .snaps in VIRTUAL world
- */
- if (local->subvolume != FIRST_CHILD (this))
- goto out;
-
- list_for_each_entry_safe (entry, tmpentry, &entries->list, list) {
- if (strcmp(priv->path, entry->d_name) == 0)
- gf_dirent_entry_free (entry);
- }
+ gf_dirent_t *entry = NULL;
+ gf_dirent_t *tmpentry = NULL;
+ svc_local_t *local = NULL;
+ char entry_point[NAME_MAX + 1] = {
+ 0,
+ };
+
+ if (op_ret < 0)
+ goto out;
+
+ local = frame->local;
+
+ /* If .snaps pre-exists, then it should not be listed
+ * in the NORMAL INODE directory when USS is enabled,
+ * so filter the .snaps entry if exists.
+ * However it is OK to list .snaps in VIRTUAL world
+ */
+ if (local->subvolume != FIRST_CHILD(this))
+ goto out;
+
+ /*
+ * Better to goto out if getting the entry point
+ * fails. We might end up sending the directory
+ * entry for the snapview entry point in the readdir
+ * response. But, the intention is to avoid the race
+ * condition where priv->path is being changed in
+ * reconfigure while this is accessing it.
+ */
+ if (gf_svc_get_entry_point(this, entry_point, sizeof(entry_point))) {
+ gf_smsg(this->name, GF_LOG_WARNING, op_errno,
+ SVC_MSG_COPY_ENTRY_POINT_FAILED, NULL);
+ goto out;
+ }
+
+ list_for_each_entry_safe(entry, tmpentry, &entries->list, list)
+ {
+ if (strcmp(entry_point, entry->d_name) == 0)
+ gf_dirent_entry_free(entry);
+ }
out:
- SVC_STACK_UNWIND (readdir, frame, op_ret, op_errno, entries, xdata);
- return 0;
+ SVC_STACK_UNWIND(readdir, frame, op_ret, op_errno, entries, xdata);
+ return 0;
}
static int32_t
-gf_svc_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
- off_t off, dict_t *xdata)
+gf_svc_readdir(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+ off_t off, dict_t *xdata)
{
- int inode_type = -1;
- xlator_t *subvolume = NULL;
- svc_local_t *local = NULL;
- int ret = -1;
- int op_ret = -1;
- int op_errno = EINVAL;
- gf_boolean_t wind = _gf_false;
- svc_fd_t *svc_fd = NULL;
- gf_dirent_t entries;
-
- INIT_LIST_HEAD (&entries);
-
- GF_VALIDATE_OR_GOTO ("svc", this, out);
- GF_VALIDATE_OR_GOTO (this->name, frame, out);
- GF_VALIDATE_OR_GOTO (this->name, fd, out);
- GF_VALIDATE_OR_GOTO (this->name, fd->inode, out);
-
- svc_fd = svc_fd_ctx_get_or_new (this, fd);
- if (!svc_fd)
- gf_log (this->name, GF_LOG_ERROR, "failed to get the fd "
- "context for the inode %s",
- uuid_utoa (fd->inode->gfid));
- else {
- if (svc_fd->entry_point_handled && off == svc_fd->last_offset) {
- op_ret = 0;
- op_errno = ENOENT;
- goto out;
- }
- }
-
- SVC_GET_SUBVOL_FROM_CTX (this, op_ret, op_errno, inode_type, ret,
- fd->inode, subvolume, out);
-
- local = mem_get0 (this->local_pool);
- if (!local) {
- gf_log (this->name, GF_LOG_ERROR, "failed to allocate local");
- goto out;
- }
- local->subvolume = subvolume;
- frame->local = local;
-
- STACK_WIND (frame, gf_svc_readdir_cbk, subvolume,
- subvolume->fops->readdir, fd, size, off, xdata);
-
- wind = _gf_true;
+ int inode_type = -1;
+ xlator_t *subvolume = NULL;
+ svc_local_t *local = NULL;
+ int ret = -1;
+ int op_ret = -1;
+ int op_errno = EINVAL;
+ gf_boolean_t wind = _gf_false;
+ svc_fd_t *svc_fd = NULL;
+ gf_dirent_t entries;
+
+ INIT_LIST_HEAD(&entries);
+
+ GF_VALIDATE_OR_GOTO("svc", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, frame, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd, out);
+ GF_VALIDATE_OR_GOTO(this->name, fd->inode, out);
+
+ svc_fd = svc_fd_ctx_get_or_new(this, fd);
+ if (!svc_fd)
+ gf_smsg(this->name, GF_LOG_ERROR, 0, SVC_MSG_GET_FD_CONTEXT_FAILED,
+ "gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ else {
+ if (svc_fd->entry_point_handled && off == svc_fd->last_offset) {
+ op_ret = 0;
+ op_errno = ENOENT;
+ goto out;
+ }
+ }
+
+ SVC_GET_SUBVOL_FROM_CTX(this, op_ret, op_errno, inode_type, ret, fd->inode,
+ subvolume, out);
+
+ local = mem_get0(this->local_pool);
+ if (!local) {
+ gf_smsg(this->name, GF_LOG_ERROR, op_errno, SVC_MSG_NO_MEMORY,
+ "inode-gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto out;
+ }
+ local->subvolume = subvolume;
+ frame->local = local;
+
+ STACK_WIND(frame, gf_svc_readdir_cbk, subvolume, subvolume->fops->readdir,
+ fd, size, off, xdata);
+
+ wind = _gf_true;
out:
- if (!wind)
- SVC_STACK_UNWIND (readdir, frame, op_ret, op_errno, &entries,
- NULL);
+ if (!wind)
+ SVC_STACK_UNWIND(readdir, frame, op_ret, op_errno, &entries, NULL);
- gf_dirent_free (&entries);
+ gf_dirent_free(&entries);
- return 0;
+ return 0;
}
/*
@@ -1643,902 +1729,1063 @@ out:
*/
static int32_t
-gf_svc_readdirp_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, inode_t *inode,
- struct iatt *buf, dict_t *xdata,
- struct iatt *postparent)
+gf_svc_readdirp_lookup_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, inode_t *inode,
+ struct iatt *buf, dict_t *xdata,
+ struct iatt *postparent)
{
- gf_dirent_t entries;
- gf_dirent_t *entry = NULL;
- svc_private_t *private = NULL;
- svc_fd_t *svc_fd = NULL;
- svc_local_t *local = NULL;
- int inode_type = -1;
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO ("snapview-client", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
-
- private = this->private;
- INIT_LIST_HEAD (&entries.list);
-
- local = frame->local;
-
- if (op_ret) {
- if (op_errno == ESTALE && !local->revalidate) {
- local->revalidate = 1;
- ret = gf_svc_special_dir_revalidate_lookup (frame,
- this,
- xdata);
-
- if (!ret)
- return 0;
- }
- op_ret = 0;
- op_errno = ENOENT;
- goto out;
- }
-
- svc_fd = svc_fd_ctx_get (this, local->fd);
- if (!svc_fd) {
- gf_log (this->name, GF_LOG_ERROR, "failed to get the fd "
- "context for the inode %s",
- uuid_utoa (local->fd->inode->gfid));
- op_ret = 0;
- op_errno = ENOENT;
- goto out;
- }
-
- entry = gf_dirent_for_name (private->path);
- if (!entry) {
- gf_log (this->name, GF_LOG_ERROR, "failed to allocate memory "
- "for the entry %s", private->path);
- op_ret = 0;
- op_errno = ENOMEM;
- goto out;
- }
-
- entry->inode = inode_ref (inode);
- entry->d_off = svc_fd->last_offset + 22;
- entry->d_ino = buf->ia_ino;
- entry->d_type = DT_DIR;
- entry->d_stat = *buf;
- inode_type = VIRTUAL_INODE;
- ret = svc_inode_ctx_set (this, entry->inode, inode_type);
- if (ret)
- gf_log (this->name, GF_LOG_ERROR, "failed to set the inode "
- "context");
-
- list_add_tail (&entry->list, &entries.list);
- op_ret = 1;
- svc_fd->last_offset = entry->d_off;
- svc_fd->entry_point_handled = _gf_true;
+ gf_dirent_t entries;
+ gf_dirent_t *entry = NULL;
+ svc_fd_t *svc_fd = NULL;
+ svc_local_t *local = NULL;
+ int inode_type = -1;
+ int ret = -1;
+ char entry_point[NAME_MAX + 1] = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("snapview-client", this, out);
+
+ INIT_LIST_HEAD(&entries.list);
+
+ local = frame->local;
+
+ if (op_ret) {
+ if (op_errno == ESTALE && !local->revalidate) {
+ local->revalidate = 1;
+ ret = gf_svc_special_dir_revalidate_lookup(frame, this, xdata);
+
+ if (!ret)
+ return 0;
+ }
+ op_ret = 0;
+ op_errno = ENOENT;
+ goto out;
+ }
+
+ svc_fd = svc_fd_ctx_get(this, local->fd);
+ if (!svc_fd) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, SVC_MSG_GET_FD_CONTEXT_FAILED,
+ "gfid=%s", uuid_utoa(local->fd->inode->gfid), NULL);
+ op_ret = 0;
+ op_errno = ENOENT;
+ goto out;
+ }
+
+ if (gf_svc_get_entry_point(this, entry_point, sizeof(entry_point))) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, SVC_MSG_COPY_ENTRY_POINT_FAILED,
+ NULL);
+ op_ret = 0;
+ op_errno = ENOENT;
+ goto out;
+ }
+
+ entry = gf_dirent_for_name(entry_point);
+ if (!entry) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, SVC_MSG_NO_MEMORY,
+ "entry-point=%s", entry_point, NULL);
+ op_ret = 0;
+ op_errno = ENOMEM;
+ goto out;
+ }
+
+ entry->inode = inode_ref(inode);
+ entry->d_off = svc_fd->last_offset + 22;
+ entry->d_ino = buf->ia_ino;
+ entry->d_type = DT_DIR;
+ entry->d_stat = *buf;
+ inode_type = VIRTUAL_INODE;
+ ret = svc_inode_ctx_set(this, entry->inode, inode_type);
+ if (ret)
+ gf_smsg(this->name, GF_LOG_ERROR, 0, SVC_MSG_SET_INODE_CONTEXT_FAILED,
+ "entry-name=%s", entry->d_name, NULL);
+
+ list_add_tail(&entry->list, &entries.list);
+ op_ret = 1;
+ svc_fd->last_offset = entry->d_off;
+ svc_fd->entry_point_handled = _gf_true;
out:
- SVC_STACK_UNWIND (readdirp, frame, op_ret, op_errno, &entries,
- local ? local->xdata : NULL);
+ SVC_STACK_UNWIND(readdirp, frame, op_ret, op_errno, &entries,
+ local ? local->xdata : NULL);
- gf_dirent_free (&entries);
+ gf_dirent_free(&entries);
- return 0;
+ return 0;
}
int
-gf_svc_special_dir_revalidate_lookup (call_frame_t *frame, xlator_t *this,
- dict_t *xdata)
+gf_svc_special_dir_revalidate_lookup(call_frame_t *frame, xlator_t *this,
+ dict_t *xdata)
{
- svc_private_t *private = NULL;
- svc_local_t *local = NULL;
- loc_t *loc = NULL;
- dict_t *tmp_xdata = NULL;
- char *path = NULL;
- int ret = -1;
-
- GF_VALIDATE_OR_GOTO ("snapview-client", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
-
- private = this->private;
-
- local = frame->local;
- loc = &local->loc;
-
- if (local->xdata) {
- dict_unref (local->xdata);
- local->xdata = NULL;
- }
-
- if (xdata)
- local->xdata = dict_ref (xdata);
-
-
- inode_unref (loc->inode);
- loc->inode = inode_new (loc->parent->table);
- if (!loc->inode) {
- gf_log (this->name, GF_LOG_ERROR, "failed to "
- "allocate new inode");
- goto out;
- }
-
- gf_uuid_copy (local->loc.gfid, loc->inode->gfid);
- ret = inode_path (loc->parent, private->path, &path);
- if (ret < 0)
- goto out;
-
- if (loc->path)
- GF_FREE ((char *)loc->path);
-
- loc->path = gf_strdup (path);
- if (loc->path) {
- if (!loc->name ||
- (loc->name && !strcmp (loc->name, ""))) {
- loc->name = strrchr (loc->path, '/');
- if (loc->name)
- loc->name++;
- }
- } else
- loc->path = NULL;
-
- tmp_xdata = dict_new ();
- if (!tmp_xdata) {
- ret = -1;
- goto out;
- }
-
- ret = dict_set_str (tmp_xdata, "entry-point", "true");
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "failed to set dict");
- goto out;
- }
-
- STACK_WIND (frame, gf_svc_readdirp_lookup_cbk,
- SECOND_CHILD (this),
- SECOND_CHILD (this)->fops->lookup, loc,
- tmp_xdata);
+ svc_local_t *local = NULL;
+ loc_t *loc = NULL;
+ dict_t *tmp_xdata = NULL;
+ char *path = NULL;
+ int ret = -1;
+ char entry_point[NAME_MAX + 1] = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("snapview-client", this, out);
+
+ local = frame->local;
+ loc = &local->loc;
+
+ if (local->xdata) {
+ dict_unref(local->xdata);
+ local->xdata = NULL;
+ }
+
+ if (xdata)
+ local->xdata = dict_ref(xdata);
+
+ inode_unref(loc->inode);
+ loc->inode = inode_new(loc->parent->table);
+ if (!loc->inode) {
+ gf_smsg(this->name, GF_LOG_ERROR, ENOMEM, SVC_MSG_ALLOC_INODE_FAILED,
+ NULL);
+ goto out;
+ }
+
+ if (gf_svc_get_entry_point(this, entry_point, sizeof(entry_point))) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0, SVC_MSG_COPY_ENTRY_POINT_FAILED,
+ NULL);
+ goto out;
+ }
+
+ gf_uuid_copy(local->loc.gfid, loc->inode->gfid);
+ ret = inode_path(loc->parent, entry_point, &path);
+ if (ret < 0)
+ goto out;
+
+ if (loc->path)
+ GF_FREE((char *)loc->path);
+
+ loc->path = gf_strdup(path);
+ if (loc->path) {
+ if (!loc->name || (loc->name && !strcmp(loc->name, ""))) {
+ loc->name = strrchr(loc->path, '/');
+ if (loc->name)
+ loc->name++;
+ }
+ } else
+ loc->path = NULL;
+
+ tmp_xdata = dict_new();
+ if (!tmp_xdata) {
+ ret = -1;
+ goto out;
+ }
+
+ ret = dict_set_str(tmp_xdata, "entry-point", "true");
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, SVC_MSG_DICT_SET_FAILED, NULL);
+ goto out;
+ }
+
+ STACK_WIND(frame, gf_svc_readdirp_lookup_cbk, SECOND_CHILD(this),
+ SECOND_CHILD(this)->fops->lookup, loc, tmp_xdata);
out:
- if (tmp_xdata)
- dict_unref (tmp_xdata);
+ if (tmp_xdata)
+ dict_unref(tmp_xdata);
- GF_FREE (path);
- return ret;
+ GF_FREE(path);
+ return ret;
}
static gf_boolean_t
-gf_svc_readdir_on_special_dir (call_frame_t *frame, void *cookie,
- xlator_t *this, int32_t op_ret,
- int32_t op_errno, gf_dirent_t *entries,
- dict_t *xdata)
+gf_svc_readdir_on_special_dir(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ gf_dirent_t *entries, dict_t *xdata)
{
- svc_local_t *local = NULL;
- svc_private_t *private = NULL;
- inode_t *inode = NULL;
- fd_t *fd = NULL;
- char *path = NULL;
- loc_t *loc = NULL;
- dict_t *tmp_xdata = NULL;
- int ret = -1;
- gf_boolean_t unwind = _gf_true;
- svc_fd_t *svc_fd = NULL;
-
- GF_VALIDATE_OR_GOTO ("snapview-client", this, out);
- GF_VALIDATE_OR_GOTO (this->name, this->private, out);
-
- private = this->private;
- local = frame->local;
-
- loc = &local->loc;
- fd = local->fd;
- svc_fd = svc_fd_ctx_get (this, fd);
- if (!svc_fd) {
- gf_log (this->name, GF_LOG_ERROR, "failed to get the fd "
- "context for the inode %s",
- uuid_utoa (fd->inode->gfid));
+ svc_local_t *local = NULL;
+ svc_private_t *private = NULL;
+ inode_t *inode = NULL;
+ fd_t *fd = NULL;
+ char *path = NULL;
+ loc_t *loc = NULL;
+ dict_t *tmp_xdata = NULL;
+ int ret = -1;
+ gf_boolean_t unwind = _gf_true;
+ svc_fd_t *svc_fd = NULL;
+ char entry_point[NAME_MAX + 1] = {
+ 0,
+ };
+
+ GF_VALIDATE_OR_GOTO("snapview-client", this, out);
+ GF_VALIDATE_OR_GOTO(this->name, this->private, out);
+
+ private
+ = this->private;
+ local = frame->local;
+
+ loc = &local->loc;
+ fd = local->fd;
+ svc_fd = svc_fd_ctx_get(this, fd);
+ if (!svc_fd) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, SVC_MSG_GET_FD_CONTEXT_FAILED,
+ "gfid=%s", uuid_utoa(fd->inode->gfid), NULL);
+ goto out;
+ }
+
+ /*
+ * check if its end of readdir operation from posix, if special_dir
+ * option is set, if readdir is done on special directory and if
+ * readdirp is from normal regular graph.
+ */
+
+ if (!private->show_entry_point)
+ goto out;
+
+ if (op_ret == 0 && op_errno == ENOENT && private->special_dir &&
+ strcmp(private->special_dir, "") && svc_fd->special_dir &&
+ local->subvolume == FIRST_CHILD(this)) {
+ if (gf_svc_get_entry_point(this, entry_point, sizeof(entry_point))) {
+ gf_smsg(this->name, GF_LOG_WARNING, 0,
+ SVC_MSG_GET_FD_CONTEXT_FAILED, NULL);
+ goto out;
+ }
+
+ inode = inode_grep(fd->inode->table, fd->inode, entry_point);
+ if (!inode) {
+ inode = inode_new(fd->inode->table);
+ if (!inode) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, SVC_MSG_ALLOC_INODE_FAILED,
+ NULL);
goto out;
+ }
}
- /*
- * check if its end of readdir operation from posix, if special_dir
- * option is set, if readdir is done on special directory and if
- * readdirp is from normal regular graph.
- */
+ gf_uuid_copy(local->loc.pargfid, fd->inode->gfid);
+ gf_uuid_copy(local->loc.gfid, inode->gfid);
+ if (gf_uuid_is_null(inode->gfid))
+ ret = inode_path(fd->inode, entry_point, &path);
+ else
+ ret = inode_path(inode, NULL, &path);
- if (!private->show_entry_point)
- goto out;
+ if (ret < 0)
+ goto out;
+ loc->path = gf_strdup(path);
+ if (loc->path) {
+ if (!loc->name || (loc->name && !strcmp(loc->name, ""))) {
+ loc->name = strrchr(loc->path, '/');
+ if (loc->name)
+ loc->name++;
+ }
+ }
+
+ loc->inode = inode;
+ loc->parent = inode_ref(fd->inode);
+ tmp_xdata = dict_new();
+ if (!tmp_xdata)
+ goto out;
+ ret = dict_set_str(tmp_xdata, "entry-point", "true");
+ if (ret) {
+ gf_smsg(this->name, GF_LOG_ERROR, 0, SVC_MSG_DICT_SET_FAILED, NULL);
+ goto out;
+ }
- if (op_ret == 0 && op_errno == ENOENT && private->special_dir &&
- strcmp (private->special_dir, "") && svc_fd->special_dir &&
- local->subvolume == FIRST_CHILD (this)) {
- inode = inode_grep (fd->inode->table, fd->inode,
- private->path);
- if (!inode) {
- inode = inode_new (fd->inode->table);
- if (!inode) {
- gf_log (this->name, GF_LOG_ERROR, "failed to "
- "allocate new inode");
- goto out;
- }
- }
-
- gf_uuid_copy (local->loc.pargfid, fd->inode->gfid);
- gf_uuid_copy (local->loc.gfid, inode->gfid);
- if (gf_uuid_is_null (inode->gfid))
- ret = inode_path (fd->inode, private->path, &path);
- else
- ret = inode_path (inode, NULL, &path);
-
- if (ret < 0)
- goto out;
- loc->path = gf_strdup (path);
- if (loc->path) {
- if (!loc->name ||
- (loc->name && !strcmp (loc->name, ""))) {
- loc->name = strrchr (loc->path, '/');
- if (loc->name)
- loc->name++;
- }
- }
-
- loc->inode = inode;
- loc->parent = inode_ref (fd->inode);
- tmp_xdata = dict_new ();
- if (!tmp_xdata)
- goto out;
- ret = dict_set_str (tmp_xdata, "entry-point", "true");
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "failed to set dict");
- goto out;
- }
-
- local->cookie = cookie;
- if (local->xdata) {
- dict_unref (local->xdata);
- local->xdata = NULL;
- }
- if (xdata)
- local->xdata = dict_ref (xdata);
-
- STACK_WIND (frame, gf_svc_readdirp_lookup_cbk,
- SECOND_CHILD (this),
- SECOND_CHILD (this)->fops->lookup, loc, tmp_xdata);
- unwind = _gf_false;
+ local->cookie = cookie;
+ if (local->xdata) {
+ dict_unref(local->xdata);
+ local->xdata = NULL;
}
+ if (xdata)
+ local->xdata = dict_ref(xdata);
+
+ STACK_WIND(frame, gf_svc_readdirp_lookup_cbk, SECOND_CHILD(this),
+ SECOND_CHILD(this)->fops->lookup, loc, tmp_xdata);
+ unwind = _gf_false;
+ }
out:
- if (tmp_xdata)
- dict_unref (tmp_xdata);
+ if (tmp_xdata)
+ dict_unref(tmp_xdata);
- GF_FREE (path);
- return unwind;
+ GF_FREE(path);
+ return unwind;
}
static int32_t
-gf_svc_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno,
- gf_dirent_t *entries, dict_t *xdata)
+gf_svc_readdirp_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,